xref: /freebsd/sys/netinet/sctp_timer.c (revision e39e854e27f53a784c3982cbeb68f4ad1cfd9162)
1 /*-
2  * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
3  * Copyright (c) 2008-2011, by Randall Stewart. All rights reserved.
4  * Copyright (c) 2008-2011, 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 /* $KAME: sctp_timer.c,v 1.29 2005/03/06 16:04:18 itojun Exp $	 */
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 #include <netinet/udp.h>
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 		    net, net->error_count,
96 		    net->failure_threshold);
97 		if (net->error_count > net->failure_threshold) {
98 			/* We had a threshold failure */
99 			if (net->dest_state & SCTP_ADDR_REACHABLE) {
100 				net->dest_state &= ~SCTP_ADDR_REACHABLE;
101 				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
102 				net->dest_state &= ~SCTP_ADDR_PF;
103 				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
104 				    stcb, 0,
105 				    (void *)net, SCTP_SO_NOT_LOCKED);
106 			}
107 		} else if ((net->pf_threshold < net->failure_threshold) &&
108 		    (net->error_count > net->pf_threshold)) {
109 			if (!(net->dest_state & SCTP_ADDR_PF)) {
110 				net->dest_state |= SCTP_ADDR_PF;
111 				net->last_active = sctp_get_tick_count();
112 				sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
113 				sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
114 				sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net);
115 			}
116 		}
117 	}
118 	if (stcb == NULL)
119 		return (0);
120 
121 	if (net) {
122 		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
123 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
124 				sctp_misc_ints(SCTP_THRESHOLD_INCR,
125 				    stcb->asoc.overall_error_count,
126 				    (stcb->asoc.overall_error_count + 1),
127 				    SCTP_FROM_SCTP_TIMER,
128 				    __LINE__);
129 			}
130 			stcb->asoc.overall_error_count++;
131 		}
132 	} else {
133 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
134 			sctp_misc_ints(SCTP_THRESHOLD_INCR,
135 			    stcb->asoc.overall_error_count,
136 			    (stcb->asoc.overall_error_count + 1),
137 			    SCTP_FROM_SCTP_TIMER,
138 			    __LINE__);
139 		}
140 		stcb->asoc.overall_error_count++;
141 	}
142 	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
143 	    &stcb->asoc, stcb->asoc.overall_error_count,
144 	    (uint32_t) threshold,
145 	    ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state));
146 	/*
147 	 * We specifically do not do >= to give the assoc one more change
148 	 * before we fail it.
149 	 */
150 	if (stcb->asoc.overall_error_count > threshold) {
151 		/* Abort notification sends a ULP notify */
152 		struct mbuf *oper;
153 
154 		oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
155 		    0, M_DONTWAIT, 1, MT_DATA);
156 		if (oper) {
157 			struct sctp_paramhdr *ph;
158 			uint32_t *ippp;
159 
160 			SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
161 			    sizeof(uint32_t);
162 			ph = mtod(oper, struct sctp_paramhdr *);
163 			ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
164 			ph->param_length = htons(SCTP_BUF_LEN(oper));
165 			ippp = (uint32_t *) (ph + 1);
166 			*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_1);
167 		}
168 		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_1;
169 		sctp_abort_an_association(inp, stcb, oper, SCTP_SO_NOT_LOCKED);
170 		return (1);
171 	}
172 	return (0);
173 }
174 
175 /*
176  * sctp_find_alternate_net() returns a non-NULL pointer as long
177  * the argument net is non-NULL.
178  */
179 struct sctp_nets *
180 sctp_find_alternate_net(struct sctp_tcb *stcb,
181     struct sctp_nets *net,
182     int mode)
183 {
184 	/* Find and return an alternate network if possible */
185 	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
186 	int once;
187 
188 	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
189 	int min_errors = -1;
190 	uint32_t max_cwnd = 0;
191 
192 	if (stcb->asoc.numnets == 1) {
193 		/* No others but net */
194 		return (TAILQ_FIRST(&stcb->asoc.nets));
195 	}
196 	/*
197 	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
198 	 * net algorithm. This algorithm chooses the active destination (not
199 	 * in PF state) with the largest cwnd value. If all destinations are
200 	 * in PF state, unreachable, or unconfirmed, choose the desination
201 	 * that is in PF state with the lowest error count. In case of a
202 	 * tie, choose the destination that was most recently active.
203 	 */
204 	if (mode == 2) {
205 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
206 			/*
207 			 * JRS 5/14/07 - If the destination is unreachable
208 			 * or unconfirmed, skip it.
209 			 */
210 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
211 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
212 				continue;
213 			}
214 			/*
215 			 * JRS 5/14/07 -  If the destination is reachable
216 			 * but in PF state, compare the error count of the
217 			 * destination to the minimum error count seen thus
218 			 * far. Store the destination with the lower error
219 			 * count.  If the error counts are equal, store the
220 			 * destination that was most recently active.
221 			 */
222 			if (mnet->dest_state & SCTP_ADDR_PF) {
223 				/*
224 				 * JRS 5/14/07 - If the destination under
225 				 * consideration is the current destination,
226 				 * work as if the error count is one higher.
227 				 * The actual error count will not be
228 				 * incremented until later in the t3
229 				 * handler.
230 				 */
231 				if (mnet == net) {
232 					if (min_errors == -1) {
233 						min_errors = mnet->error_count + 1;
234 						min_errors_net = mnet;
235 					} else if (mnet->error_count + 1 < min_errors) {
236 						min_errors = mnet->error_count + 1;
237 						min_errors_net = mnet;
238 					} else if (mnet->error_count + 1 == min_errors
239 					    && mnet->last_active > min_errors_net->last_active) {
240 						min_errors_net = mnet;
241 						min_errors = mnet->error_count + 1;
242 					}
243 					continue;
244 				} else {
245 					if (min_errors == -1) {
246 						min_errors = mnet->error_count;
247 						min_errors_net = mnet;
248 					} else if (mnet->error_count < min_errors) {
249 						min_errors = mnet->error_count;
250 						min_errors_net = mnet;
251 					} else if (mnet->error_count == min_errors
252 					    && mnet->last_active > min_errors_net->last_active) {
253 						min_errors_net = mnet;
254 						min_errors = mnet->error_count;
255 					}
256 					continue;
257 				}
258 			}
259 			/*
260 			 * JRS 5/14/07 - If the destination is reachable and
261 			 * not in PF state, compare the cwnd of the
262 			 * destination to the highest cwnd seen thus far.
263 			 * Store the destination with the higher cwnd value.
264 			 * If the cwnd values are equal, randomly choose one
265 			 * of the two destinations.
266 			 */
267 			if (max_cwnd < mnet->cwnd) {
268 				max_cwnd_net = mnet;
269 				max_cwnd = mnet->cwnd;
270 			} else if (max_cwnd == mnet->cwnd) {
271 				uint32_t rndval;
272 				uint8_t this_random;
273 
274 				if (stcb->asoc.hb_random_idx > 3) {
275 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
276 					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
277 					this_random = stcb->asoc.hb_random_values[0];
278 					stcb->asoc.hb_random_idx++;
279 					stcb->asoc.hb_ect_randombit = 0;
280 				} else {
281 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
282 					stcb->asoc.hb_random_idx++;
283 					stcb->asoc.hb_ect_randombit = 0;
284 				}
285 				if (this_random % 2 == 1) {
286 					max_cwnd_net = mnet;
287 					max_cwnd = mnet->cwnd;	/* Useless? */
288 				}
289 			}
290 		}
291 		if (max_cwnd_net == NULL) {
292 			if (min_errors_net == NULL) {
293 				return (net);
294 			}
295 			return (min_errors_net);
296 		} else {
297 			return (max_cwnd_net);
298 		}
299 	}
300 	/*
301 	 * JRS 5/14/07 - If mode is set to 1, use the CMT policy for
302 	 * choosing an alternate net.
303 	 */
304 	else if (mode == 1) {
305 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
306 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
307 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
308 				/*
309 				 * will skip ones that are not-reachable or
310 				 * unconfirmed
311 				 */
312 				continue;
313 			}
314 			if (max_cwnd < mnet->cwnd) {
315 				max_cwnd_net = mnet;
316 				max_cwnd = mnet->cwnd;
317 			} else if (max_cwnd == mnet->cwnd) {
318 				uint32_t rndval;
319 				uint8_t this_random;
320 
321 				if (stcb->asoc.hb_random_idx > 3) {
322 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
323 					memcpy(stcb->asoc.hb_random_values, &rndval,
324 					    sizeof(stcb->asoc.hb_random_values));
325 					this_random = stcb->asoc.hb_random_values[0];
326 					stcb->asoc.hb_random_idx = 0;
327 					stcb->asoc.hb_ect_randombit = 0;
328 				} else {
329 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
330 					stcb->asoc.hb_random_idx++;
331 					stcb->asoc.hb_ect_randombit = 0;
332 				}
333 				if (this_random % 2) {
334 					max_cwnd_net = mnet;
335 					max_cwnd = mnet->cwnd;
336 				}
337 			}
338 		}
339 		if (max_cwnd_net) {
340 			return (max_cwnd_net);
341 		}
342 	}
343 	mnet = net;
344 	once = 0;
345 
346 	if (mnet == NULL) {
347 		mnet = TAILQ_FIRST(&stcb->asoc.nets);
348 		if (mnet == NULL) {
349 			return (NULL);
350 		}
351 	}
352 	do {
353 		alt = TAILQ_NEXT(mnet, sctp_next);
354 		if (alt == NULL) {
355 			once++;
356 			if (once > 1) {
357 				break;
358 			}
359 			alt = TAILQ_FIRST(&stcb->asoc.nets);
360 			if (alt == NULL) {
361 				return (NULL);
362 			}
363 		}
364 		if (alt->ro.ro_rt == NULL) {
365 			if (alt->ro._s_addr) {
366 				sctp_free_ifa(alt->ro._s_addr);
367 				alt->ro._s_addr = NULL;
368 			}
369 			alt->src_addr_selected = 0;
370 		}
371 		/* sa_ignore NO_NULL_CHK */
372 		if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
373 		    (alt->ro.ro_rt != NULL) &&
374 		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))) {
375 			/* Found a reachable address */
376 			break;
377 		}
378 		mnet = alt;
379 	} while (alt != NULL);
380 
381 	if (alt == NULL) {
382 		/* Case where NO insv network exists (dormant state) */
383 		/* we rotate destinations */
384 		once = 0;
385 		mnet = net;
386 		do {
387 			if (mnet == NULL) {
388 				return (TAILQ_FIRST(&stcb->asoc.nets));
389 			}
390 			alt = TAILQ_NEXT(mnet, sctp_next);
391 			if (alt == NULL) {
392 				once++;
393 				if (once > 1) {
394 					break;
395 				}
396 				alt = TAILQ_FIRST(&stcb->asoc.nets);
397 			}
398 			/* sa_ignore NO_NULL_CHK */
399 			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
400 			    (alt != net)) {
401 				/* Found an alternate address */
402 				break;
403 			}
404 			mnet = alt;
405 		} while (alt != NULL);
406 	}
407 	if (alt == NULL) {
408 		return (net);
409 	}
410 	return (alt);
411 }
412 
413 static void
414 sctp_backoff_on_timeout(struct sctp_tcb *stcb,
415     struct sctp_nets *net,
416     int win_probe,
417     int num_marked, int num_abandoned)
418 {
419 	if (net->RTO == 0) {
420 		net->RTO = stcb->asoc.minrto;
421 	}
422 	net->RTO <<= 1;
423 	if (net->RTO > stcb->asoc.maxrto) {
424 		net->RTO = stcb->asoc.maxrto;
425 	}
426 	if ((win_probe == 0) && (num_marked || num_abandoned)) {
427 		/* We don't apply penalty to window probe scenarios */
428 		/* JRS - Use the congestion control given in the CC module */
429 		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
430 	}
431 }
432 
433 #ifndef INVARIANTS
434 static void
435 sctp_recover_sent_list(struct sctp_tcb *stcb)
436 {
437 	struct sctp_tmit_chunk *chk, *nchk;
438 	struct sctp_association *asoc;
439 
440 	asoc = &stcb->asoc;
441 	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
442 		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.TSN_seq)) {
443 			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
444 			    chk, chk->rec.data.TSN_seq, asoc->last_acked_seq);
445 			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
446 			if (chk->pr_sctp_on) {
447 				if (asoc->pr_sctp_cnt != 0)
448 					asoc->pr_sctp_cnt--;
449 			}
450 			if (chk->data) {
451 				/* sa_ignore NO_NULL_CHK */
452 				sctp_free_bufspace(stcb, asoc, chk, 1);
453 				sctp_m_freem(chk->data);
454 				chk->data = NULL;
455 				if (asoc->peer_supports_prsctp && PR_SCTP_BUF_ENABLED(chk->flags)) {
456 					asoc->sent_queue_cnt_removeable--;
457 				}
458 			}
459 			asoc->sent_queue_cnt--;
460 			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
461 		}
462 	}
463 	SCTP_PRINTF("after recover order is as follows\n");
464 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
465 		SCTP_PRINTF("chk:%p TSN:%x\n", chk, chk->rec.data.TSN_seq);
466 	}
467 }
468 
469 #endif
470 
471 static int
472 sctp_mark_all_for_resend(struct sctp_tcb *stcb,
473     struct sctp_nets *net,
474     struct sctp_nets *alt,
475     int window_probe,
476     int *num_marked,
477     int *num_abandoned)
478 {
479 
480 	/*
481 	 * Mark all chunks (well not all) that were sent to *net for
482 	 * retransmission. Move them to alt for there destination as well...
483 	 * We only mark chunks that have been outstanding long enough to
484 	 * have received feed-back.
485 	 */
486 	struct sctp_tmit_chunk *chk, *nchk;
487 	struct sctp_nets *lnets;
488 	struct timeval now, min_wait, tv;
489 	int cur_rto;
490 	int cnt_abandoned;
491 	int audit_tf, num_mk, fir;
492 	unsigned int cnt_mk;
493 	uint32_t orig_flight, orig_tf;
494 	uint32_t tsnlast, tsnfirst;
495 	int recovery_cnt = 0;
496 
497 
498 	/* none in flight now */
499 	audit_tf = 0;
500 	fir = 0;
501 	/*
502 	 * figure out how long a data chunk must be pending before we can
503 	 * mark it ..
504 	 */
505 	(void)SCTP_GETTIME_TIMEVAL(&now);
506 	/* get cur rto in micro-seconds */
507 	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
508 	cur_rto *= 1000;
509 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
510 		sctp_log_fr(cur_rto,
511 		    stcb->asoc.peers_rwnd,
512 		    window_probe,
513 		    SCTP_FR_T3_MARK_TIME);
514 		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
515 		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
516 	}
517 	tv.tv_sec = cur_rto / 1000000;
518 	tv.tv_usec = cur_rto % 1000000;
519 	min_wait = now;
520 	timevalsub(&min_wait, &tv);
521 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
522 		/*
523 		 * if we hit here, we don't have enough seconds on the clock
524 		 * to account for the RTO. We just let the lower seconds be
525 		 * the bounds and don't worry about it. This may mean we
526 		 * will mark a lot more than we should.
527 		 */
528 		min_wait.tv_sec = min_wait.tv_usec = 0;
529 	}
530 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
531 		sctp_log_fr(cur_rto, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
532 		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
533 	}
534 	/*
535 	 * Our rwnd will be incorrect here since we are not adding back the
536 	 * cnt * mbuf but we will fix that down below.
537 	 */
538 	orig_flight = net->flight_size;
539 	orig_tf = stcb->asoc.total_flight;
540 
541 	net->fast_retran_ip = 0;
542 	/* Now on to each chunk */
543 	cnt_abandoned = 0;
544 	num_mk = cnt_mk = 0;
545 	tsnfirst = tsnlast = 0;
546 #ifndef INVARIANTS
547 start_again:
548 #endif
549 	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
550 		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.TSN_seq)) {
551 			/* Strange case our list got out of order? */
552 			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x",
553 			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq);
554 			recovery_cnt++;
555 #ifdef INVARIANTS
556 			panic("last acked >= chk on sent-Q");
557 #else
558 			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
559 			sctp_recover_sent_list(stcb);
560 			if (recovery_cnt < 10) {
561 				goto start_again;
562 			} else {
563 				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
564 			}
565 #endif
566 		}
567 		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
568 			/*
569 			 * found one to mark: If it is less than
570 			 * DATAGRAM_ACKED it MUST not be a skipped or marked
571 			 * TSN but instead one that is either already set
572 			 * for retransmission OR one that needs
573 			 * retransmission.
574 			 */
575 
576 			/* validate its been outstanding long enough */
577 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
578 				sctp_log_fr(chk->rec.data.TSN_seq,
579 				    chk->sent_rcv_time.tv_sec,
580 				    chk->sent_rcv_time.tv_usec,
581 				    SCTP_FR_T3_MARK_TIME);
582 			}
583 			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
584 				/*
585 				 * we have reached a chunk that was sent
586 				 * some seconds past our min.. forget it we
587 				 * will find no more to send.
588 				 */
589 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
590 					sctp_log_fr(0,
591 					    chk->sent_rcv_time.tv_sec,
592 					    chk->sent_rcv_time.tv_usec,
593 					    SCTP_FR_T3_STOPPED);
594 				}
595 				continue;
596 			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
597 			    (window_probe == 0)) {
598 				/*
599 				 * we must look at the micro seconds to
600 				 * know.
601 				 */
602 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
603 					/*
604 					 * ok it was sent after our boundary
605 					 * time.
606 					 */
607 					continue;
608 				}
609 			}
610 			if (stcb->asoc.peer_supports_prsctp && PR_SCTP_TTL_ENABLED(chk->flags)) {
611 				/* Is it expired? */
612 				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
613 					/* Yes so drop it */
614 					if (chk->data) {
615 						(void)sctp_release_pr_sctp_chunk(stcb,
616 						    chk,
617 						    1,
618 						    SCTP_SO_NOT_LOCKED);
619 						cnt_abandoned++;
620 					}
621 					continue;
622 				}
623 			}
624 			if (stcb->asoc.peer_supports_prsctp && PR_SCTP_RTX_ENABLED(chk->flags)) {
625 				/* Has it been retransmitted tv_sec times? */
626 				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
627 					if (chk->data) {
628 						(void)sctp_release_pr_sctp_chunk(stcb,
629 						    chk,
630 						    1,
631 						    SCTP_SO_NOT_LOCKED);
632 						cnt_abandoned++;
633 					}
634 					continue;
635 				}
636 			}
637 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
638 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
639 				num_mk++;
640 				if (fir == 0) {
641 					fir = 1;
642 					tsnfirst = chk->rec.data.TSN_seq;
643 				}
644 				tsnlast = chk->rec.data.TSN_seq;
645 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
646 					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
647 					    0, SCTP_FR_T3_MARKED);
648 				}
649 				if (chk->rec.data.chunk_was_revoked) {
650 					/* deflate the cwnd */
651 					chk->whoTo->cwnd -= chk->book_size;
652 					chk->rec.data.chunk_was_revoked = 0;
653 				}
654 				net->marked_retrans++;
655 				stcb->asoc.marked_retrans++;
656 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
657 					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
658 					    chk->whoTo->flight_size,
659 					    chk->book_size,
660 					    (uintptr_t) chk->whoTo,
661 					    chk->rec.data.TSN_seq);
662 				}
663 				sctp_flight_size_decrease(chk);
664 				sctp_total_flight_decrease(stcb, chk);
665 				stcb->asoc.peers_rwnd += chk->send_size;
666 				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
667 			}
668 			chk->sent = SCTP_DATAGRAM_RESEND;
669 			SCTP_STAT_INCR(sctps_markedretrans);
670 
671 			/* reset the TSN for striking and other FR stuff */
672 			chk->rec.data.doing_fast_retransmit = 0;
673 			/* Clear any time so NO RTT is being done */
674 
675 			if (chk->do_rtt) {
676 				if (chk->whoTo->rto_needed == 0) {
677 					chk->whoTo->rto_needed = 1;
678 				}
679 			}
680 			chk->do_rtt = 0;
681 			if (alt != net) {
682 				sctp_free_remote_addr(chk->whoTo);
683 				chk->no_fr_allowed = 1;
684 				chk->whoTo = alt;
685 				atomic_add_int(&alt->ref_count, 1);
686 			} else {
687 				chk->no_fr_allowed = 0;
688 				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
689 					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
690 				} else {
691 					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
692 				}
693 			}
694 			/*
695 			 * CMT: Do not allow FRs on retransmitted TSNs.
696 			 */
697 			if (stcb->asoc.sctp_cmt_on_off > 0) {
698 				chk->no_fr_allowed = 1;
699 			}
700 #ifdef THIS_SHOULD_NOT_BE_DONE
701 		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
702 			/* remember highest acked one */
703 			could_be_sent = chk;
704 #endif
705 		}
706 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
707 			cnt_mk++;
708 		}
709 	}
710 	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
711 		/* we did not subtract the same things? */
712 		audit_tf = 1;
713 	}
714 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
715 		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
716 	}
717 #ifdef SCTP_DEBUG
718 	if (num_mk) {
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:%ld\n",
722 		    num_mk, (u_long)stcb->asoc.peers_rwnd);
723 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
724 		    tsnlast);
725 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n",
726 		    num_mk,
727 		    (int)stcb->asoc.peers_rwnd);
728 	}
729 #endif
730 	*num_marked = num_mk;
731 	*num_abandoned = cnt_abandoned;
732 	/*
733 	 * Now check for a ECN Echo that may be stranded And include the
734 	 * cnt_mk'd to have all resends in the control queue.
735 	 */
736 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
737 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
738 			cnt_mk++;
739 		}
740 		if ((chk->whoTo == net) &&
741 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
742 			sctp_free_remote_addr(chk->whoTo);
743 			chk->whoTo = alt;
744 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
745 				chk->sent = SCTP_DATAGRAM_RESEND;
746 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
747 				cnt_mk++;
748 			}
749 			atomic_add_int(&alt->ref_count, 1);
750 		}
751 	}
752 #ifdef THIS_SHOULD_NOT_BE_DONE
753 	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
754 		/* fix it so we retransmit the highest acked anyway */
755 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
756 		cnt_mk++;
757 		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
758 	}
759 #endif
760 	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
761 #ifdef INVARIANTS
762 		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
763 		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
764 #endif
765 #ifndef SCTP_AUDITING_ENABLED
766 		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
767 #endif
768 	}
769 	if (audit_tf) {
770 		SCTPDBG(SCTP_DEBUG_TIMER4,
771 		    "Audit total flight due to negative value net:%p\n",
772 		    net);
773 		stcb->asoc.total_flight = 0;
774 		stcb->asoc.total_flight_count = 0;
775 		/* Clear all networks flight size */
776 		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
777 			lnets->flight_size = 0;
778 			SCTPDBG(SCTP_DEBUG_TIMER4,
779 			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
780 			    lnets, lnets->cwnd, lnets->ssthresh);
781 		}
782 		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
783 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
784 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
785 					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
786 					    chk->whoTo->flight_size,
787 					    chk->book_size,
788 					    (uintptr_t) chk->whoTo,
789 					    chk->rec.data.TSN_seq);
790 				}
791 				sctp_flight_size_increase(chk);
792 				sctp_total_flight_increase(stcb, chk);
793 			}
794 		}
795 	}
796 	/* We return 1 if we only have a window probe outstanding */
797 	return (0);
798 }
799 
800 
801 int
802 sctp_t3rxt_timer(struct sctp_inpcb *inp,
803     struct sctp_tcb *stcb,
804     struct sctp_nets *net)
805 {
806 	struct sctp_nets *alt;
807 	int win_probe, num_mk, num_abandoned;
808 
809 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
810 		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
811 	}
812 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
813 		struct sctp_nets *lnet;
814 
815 		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
816 			if (net == lnet) {
817 				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
818 			} else {
819 				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
820 			}
821 		}
822 	}
823 	/* Find an alternate and mark those for retransmission */
824 	if ((stcb->asoc.peers_rwnd == 0) &&
825 	    (stcb->asoc.total_flight < net->mtu)) {
826 		SCTP_STAT_INCR(sctps_timowindowprobe);
827 		win_probe = 1;
828 	} else {
829 		win_probe = 0;
830 	}
831 
832 	if (win_probe == 0) {
833 		/* We don't do normal threshold management on window probes */
834 		if (sctp_threshold_management(inp, stcb, net,
835 		    stcb->asoc.max_send_times)) {
836 			/* Association was destroyed */
837 			return (1);
838 		} else {
839 			if (net != stcb->asoc.primary_destination) {
840 				/* send a immediate HB if our RTO is stale */
841 				struct timeval now;
842 				unsigned int ms_goneby;
843 
844 				(void)SCTP_GETTIME_TIMEVAL(&now);
845 				if (net->last_sent_time.tv_sec) {
846 					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
847 				} else {
848 					ms_goneby = 0;
849 				}
850 				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
851 					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
852 						/*
853 						 * no recent feed back in an
854 						 * RTO or more, request a
855 						 * RTT update
856 						 */
857 						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
858 					}
859 				}
860 			}
861 		}
862 	} else {
863 		/*
864 		 * For a window probe we don't penalize the net's but only
865 		 * the association. This may fail it if SACKs are not coming
866 		 * back. If sack's are coming with rwnd locked at 0, we will
867 		 * continue to hold things waiting for rwnd to raise
868 		 */
869 		if (sctp_threshold_management(inp, stcb, NULL,
870 		    stcb->asoc.max_send_times)) {
871 			/* Association was destroyed */
872 			return (1);
873 		}
874 	}
875 	if (stcb->asoc.sctp_cmt_on_off > 0) {
876 		if (net->pf_threshold < net->failure_threshold) {
877 			alt = sctp_find_alternate_net(stcb, net, 2);
878 		} else {
879 			/*
880 			 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is
881 			 * being used, then pick dest with largest ssthresh
882 			 * for any retransmission.
883 			 */
884 			alt = sctp_find_alternate_net(stcb, net, 1);
885 			/*
886 			 * CUCv2: If a different dest is picked for the
887 			 * retransmission, then new (rtx-)pseudo_cumack
888 			 * needs to be tracked for orig dest. Let CUCv2
889 			 * track new (rtx-) pseudo-cumack always.
890 			 */
891 			net->find_pseudo_cumack = 1;
892 			net->find_rtx_pseudo_cumack = 1;
893 		}
894 	} else {
895 		alt = sctp_find_alternate_net(stcb, net, 0);
896 	}
897 
898 	num_mk = 0;
899 	num_abandoned = 0;
900 	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
901 	    &num_mk, &num_abandoned);
902 	/* FR Loss recovery just ended with the T3. */
903 	stcb->asoc.fast_retran_loss_recovery = 0;
904 
905 	/* CMT FR loss recovery ended with the T3 */
906 	net->fast_retran_loss_recovery = 0;
907 	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
908 	    (net->flight_size == 0)) {
909 		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) (stcb, net);
910 	}
911 	/*
912 	 * setup the sat loss recovery that prevents satellite cwnd advance.
913 	 */
914 	stcb->asoc.sat_t3_loss_recovery = 1;
915 	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
916 
917 	/* Backoff the timer and cwnd */
918 	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
919 	if ((!(net->dest_state & SCTP_ADDR_REACHABLE)) ||
920 	    (net->dest_state & SCTP_ADDR_PF)) {
921 		/* Move all pending over too */
922 		sctp_move_chunks_from_net(stcb, net);
923 
924 		/*
925 		 * Get the address that failed, to force a new src address
926 		 * selecton and a route allocation.
927 		 */
928 		if (net->ro._s_addr) {
929 			sctp_free_ifa(net->ro._s_addr);
930 			net->ro._s_addr = NULL;
931 		}
932 		net->src_addr_selected = 0;
933 
934 		/* Force a route allocation too */
935 		if (net->ro.ro_rt) {
936 			RTFREE(net->ro.ro_rt);
937 			net->ro.ro_rt = NULL;
938 		}
939 		/* Was it our primary? */
940 		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
941 			/*
942 			 * Yes, note it as such and find an alternate note:
943 			 * this means HB code must use this to resent the
944 			 * primary if it goes active AND if someone does a
945 			 * change-primary then this flag must be cleared
946 			 * from any net structures.
947 			 */
948 			if (stcb->asoc.alternate) {
949 				sctp_free_remote_addr(stcb->asoc.alternate);
950 			}
951 			stcb->asoc.alternate = alt;
952 			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
953 		}
954 	}
955 	/*
956 	 * Special case for cookie-echo'ed case, we don't do output but must
957 	 * await the COOKIE-ACK before retransmission
958 	 */
959 	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
960 		/*
961 		 * Here we just reset the timer and start again since we
962 		 * have not established the asoc
963 		 */
964 		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
965 		return (0);
966 	}
967 	if (stcb->asoc.peer_supports_prsctp) {
968 		struct sctp_tmit_chunk *lchk;
969 
970 		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
971 		/* C3. See if we need to send a Fwd-TSN */
972 		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
973 			send_forward_tsn(stcb, &stcb->asoc);
974 			if (lchk) {
975 				/* Assure a timer is up */
976 				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
977 			}
978 		}
979 	}
980 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
981 		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
982 	}
983 	return (0);
984 }
985 
986 int
987 sctp_t1init_timer(struct sctp_inpcb *inp,
988     struct sctp_tcb *stcb,
989     struct sctp_nets *net)
990 {
991 	/* bump the thresholds */
992 	if (stcb->asoc.delayed_connection) {
993 		/*
994 		 * special hook for delayed connection. The library did NOT
995 		 * complete the rest of its sends.
996 		 */
997 		stcb->asoc.delayed_connection = 0;
998 		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
999 		return (0);
1000 	}
1001 	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
1002 		return (0);
1003 	}
1004 	if (sctp_threshold_management(inp, stcb, net,
1005 	    stcb->asoc.max_init_times)) {
1006 		/* Association was destroyed */
1007 		return (1);
1008 	}
1009 	stcb->asoc.dropped_special_cnt = 0;
1010 	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1011 	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1012 		net->RTO = stcb->asoc.initial_init_rto_max;
1013 	}
1014 	if (stcb->asoc.numnets > 1) {
1015 		/* If we have more than one addr use it */
1016 		struct sctp_nets *alt;
1017 
1018 		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1019 		if (alt != stcb->asoc.primary_destination) {
1020 			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1021 			stcb->asoc.primary_destination = alt;
1022 		}
1023 	}
1024 	/* Send out a new init */
1025 	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1026 	return (0);
1027 }
1028 
1029 /*
1030  * For cookie and asconf we actually need to find and mark for resend, then
1031  * increment the resend counter (after all the threshold management stuff of
1032  * course).
1033  */
1034 int
1035 sctp_cookie_timer(struct sctp_inpcb *inp,
1036     struct sctp_tcb *stcb,
1037     struct sctp_nets *net SCTP_UNUSED)
1038 {
1039 	struct sctp_nets *alt;
1040 	struct sctp_tmit_chunk *cookie;
1041 
1042 	/* first before all else we must find the cookie */
1043 	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1044 		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1045 			break;
1046 		}
1047 	}
1048 	if (cookie == NULL) {
1049 		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1050 			/* FOOBAR! */
1051 			struct mbuf *oper;
1052 
1053 			oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
1054 			    0, M_DONTWAIT, 1, MT_DATA);
1055 			if (oper) {
1056 				struct sctp_paramhdr *ph;
1057 				uint32_t *ippp;
1058 
1059 				SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
1060 				    sizeof(uint32_t);
1061 				ph = mtod(oper, struct sctp_paramhdr *);
1062 				ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1063 				ph->param_length = htons(SCTP_BUF_LEN(oper));
1064 				ippp = (uint32_t *) (ph + 1);
1065 				*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
1066 			}
1067 			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_4;
1068 			sctp_abort_an_association(inp, stcb, oper, SCTP_SO_NOT_LOCKED);
1069 		} else {
1070 #ifdef INVARIANTS
1071 			panic("Cookie timer expires in wrong state?");
1072 #else
1073 			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1074 			return (0);
1075 #endif
1076 		}
1077 		return (0);
1078 	}
1079 	/* Ok we found the cookie, threshold management next */
1080 	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1081 	    stcb->asoc.max_init_times)) {
1082 		/* Assoc is over */
1083 		return (1);
1084 	}
1085 	/*
1086 	 * cleared theshold management now lets backoff the address & select
1087 	 * an alternate
1088 	 */
1089 	stcb->asoc.dropped_special_cnt = 0;
1090 	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0);
1091 	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1092 	if (alt != cookie->whoTo) {
1093 		sctp_free_remote_addr(cookie->whoTo);
1094 		cookie->whoTo = alt;
1095 		atomic_add_int(&alt->ref_count, 1);
1096 	}
1097 	/* Now mark the retran info */
1098 	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1099 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1100 	}
1101 	cookie->sent = SCTP_DATAGRAM_RESEND;
1102 	/*
1103 	 * Now call the output routine to kick out the cookie again, Note we
1104 	 * don't mark any chunks for retran so that FR will need to kick in
1105 	 * to move these (or a send timer).
1106 	 */
1107 	return (0);
1108 }
1109 
1110 int
1111 sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1112     struct sctp_nets *net)
1113 {
1114 	struct sctp_nets *alt;
1115 	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1116 
1117 	if (stcb->asoc.stream_reset_outstanding == 0) {
1118 		return (0);
1119 	}
1120 	/* find the existing STRRESET, we use the seq number we sent out on */
1121 	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1122 	if (strrst == NULL) {
1123 		return (0);
1124 	}
1125 	/* do threshold management */
1126 	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1127 	    stcb->asoc.max_send_times)) {
1128 		/* Assoc is over */
1129 		return (1);
1130 	}
1131 	/*
1132 	 * cleared theshold management now lets backoff the address & select
1133 	 * an alternate
1134 	 */
1135 	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0, 0);
1136 	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1137 	sctp_free_remote_addr(strrst->whoTo);
1138 	strrst->whoTo = alt;
1139 	atomic_add_int(&alt->ref_count, 1);
1140 
1141 	/* See if a ECN Echo is also stranded */
1142 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1143 		if ((chk->whoTo == net) &&
1144 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1145 			sctp_free_remote_addr(chk->whoTo);
1146 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1147 				chk->sent = SCTP_DATAGRAM_RESEND;
1148 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1149 			}
1150 			chk->whoTo = alt;
1151 			atomic_add_int(&alt->ref_count, 1);
1152 		}
1153 	}
1154 	if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1155 		/*
1156 		 * If the address went un-reachable, we need to move to
1157 		 * alternates for ALL chk's in queue
1158 		 */
1159 		sctp_move_chunks_from_net(stcb, net);
1160 	}
1161 	/* mark the retran info */
1162 	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1163 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1164 	strrst->sent = SCTP_DATAGRAM_RESEND;
1165 
1166 	/* restart the timer */
1167 	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1168 	return (0);
1169 }
1170 
1171 int
1172 sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1173     struct sctp_nets *net)
1174 {
1175 	struct sctp_nets *alt;
1176 	struct sctp_tmit_chunk *asconf, *chk;
1177 
1178 	/* is this a first send, or a retransmission? */
1179 	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1180 		/* compose a new ASCONF chunk and send it */
1181 		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1182 	} else {
1183 		/*
1184 		 * Retransmission of the existing ASCONF is needed
1185 		 */
1186 
1187 		/* find the existing ASCONF */
1188 		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1189 		if (asconf == NULL) {
1190 			return (0);
1191 		}
1192 		/* do threshold management */
1193 		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1194 		    stcb->asoc.max_send_times)) {
1195 			/* Assoc is over */
1196 			return (1);
1197 		}
1198 		if (asconf->snd_count > stcb->asoc.max_send_times) {
1199 			/*
1200 			 * Something is rotten: our peer is not responding
1201 			 * to ASCONFs but apparently is to other chunks.
1202 			 * i.e. it is not properly handling the chunk type
1203 			 * upper bits. Mark this peer as ASCONF incapable
1204 			 * and cleanup.
1205 			 */
1206 			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1207 			sctp_asconf_cleanup(stcb, net);
1208 			return (0);
1209 		}
1210 		/*
1211 		 * cleared threshold management, so now backoff the net and
1212 		 * select an alternate
1213 		 */
1214 		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0, 0);
1215 		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1216 		if (asconf->whoTo != alt) {
1217 			sctp_free_remote_addr(asconf->whoTo);
1218 			asconf->whoTo = alt;
1219 			atomic_add_int(&alt->ref_count, 1);
1220 		}
1221 		/* See if an ECN Echo is also stranded */
1222 		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1223 			if ((chk->whoTo == net) &&
1224 			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1225 				sctp_free_remote_addr(chk->whoTo);
1226 				chk->whoTo = alt;
1227 				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1228 					chk->sent = SCTP_DATAGRAM_RESEND;
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 		}
1244 		if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1245 			/*
1246 			 * If the address went un-reachable, we need to move
1247 			 * to the alternate for ALL chunks in queue
1248 			 */
1249 			sctp_move_chunks_from_net(stcb, net);
1250 		}
1251 		/* mark the retran info */
1252 		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1253 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1254 		asconf->sent = SCTP_DATAGRAM_RESEND;
1255 
1256 		/* send another ASCONF if any and we can do */
1257 		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1258 	}
1259 	return (0);
1260 }
1261 
1262 /* Mobility adaptation */
1263 void
1264 sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1265     struct sctp_nets *net SCTP_UNUSED)
1266 {
1267 	if (stcb->asoc.deleted_primary == NULL) {
1268 		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1269 		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1270 		return;
1271 	}
1272 	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1273 	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1274 	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1275 	stcb->asoc.deleted_primary = NULL;
1276 	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1277 	return;
1278 }
1279 
1280 /*
1281  * For the shutdown and shutdown-ack, we do not keep one around on the
1282  * control queue. This means we must generate a new one and call the general
1283  * chunk output routine, AFTER having done threshold management.
1284  * It is assumed that net is non-NULL.
1285  */
1286 int
1287 sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1288     struct sctp_nets *net)
1289 {
1290 	struct sctp_nets *alt;
1291 
1292 	/* first threshold managment */
1293 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1294 		/* Assoc is over */
1295 		return (1);
1296 	}
1297 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1298 	/* second select an alternative */
1299 	alt = sctp_find_alternate_net(stcb, net, 0);
1300 
1301 	/* third generate a shutdown into the queue for out net */
1302 	sctp_send_shutdown(stcb, alt);
1303 
1304 	/* fourth restart timer */
1305 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1306 	return (0);
1307 }
1308 
1309 int
1310 sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1311     struct sctp_nets *net)
1312 {
1313 	struct sctp_nets *alt;
1314 
1315 	/* first threshold managment */
1316 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1317 		/* Assoc is over */
1318 		return (1);
1319 	}
1320 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1321 	/* second select an alternative */
1322 	alt = sctp_find_alternate_net(stcb, net, 0);
1323 
1324 	/* third generate a shutdown into the queue for out net */
1325 	sctp_send_shutdown_ack(stcb, alt);
1326 
1327 	/* fourth restart timer */
1328 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1329 	return (0);
1330 }
1331 
1332 static void
1333 sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1334     struct sctp_tcb *stcb)
1335 {
1336 	struct sctp_stream_queue_pending *sp;
1337 	unsigned int i, chks_in_queue = 0;
1338 	int being_filled = 0;
1339 
1340 	/*
1341 	 * This function is ONLY called when the send/sent queues are empty.
1342 	 */
1343 	if ((stcb == NULL) || (inp == NULL))
1344 		return;
1345 
1346 	if (stcb->asoc.sent_queue_retran_cnt) {
1347 		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1348 		    stcb->asoc.sent_queue_retran_cnt);
1349 		stcb->asoc.sent_queue_retran_cnt = 0;
1350 	}
1351 	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1352 		/* No stream scheduler information, initialize scheduler */
1353 		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 0);
1354 		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1355 			/* yep, we lost a stream or two */
1356 			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1357 		} else {
1358 			/* no streams lost */
1359 			stcb->asoc.total_output_queue_size = 0;
1360 		}
1361 	}
1362 	/* Check to see if some data queued, if so report it */
1363 	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1364 		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1365 			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1366 				if (sp->msg_is_complete)
1367 					being_filled++;
1368 				chks_in_queue++;
1369 			}
1370 		}
1371 	}
1372 	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1373 		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1374 		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1375 	}
1376 	if (chks_in_queue) {
1377 		/* call the output queue function */
1378 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1379 		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1380 		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1381 			/*
1382 			 * Probably should go in and make it go back through
1383 			 * and add fragments allowed
1384 			 */
1385 			if (being_filled == 0) {
1386 				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1387 				    chks_in_queue);
1388 			}
1389 		}
1390 	} else {
1391 		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1392 		    (u_long)stcb->asoc.total_output_queue_size);
1393 		stcb->asoc.total_output_queue_size = 0;
1394 	}
1395 }
1396 
1397 int
1398 sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1399     struct sctp_nets *net)
1400 {
1401 	uint8_t net_was_pf;
1402 
1403 	if (net->dest_state & SCTP_ADDR_PF) {
1404 		net_was_pf = 1;
1405 	} else {
1406 		net_was_pf = 0;
1407 	}
1408 	if (net->hb_responded == 0) {
1409 		if (net->ro._s_addr) {
1410 			/*
1411 			 * Invalidate the src address if we did not get a
1412 			 * response last time.
1413 			 */
1414 			sctp_free_ifa(net->ro._s_addr);
1415 			net->ro._s_addr = NULL;
1416 			net->src_addr_selected = 0;
1417 		}
1418 		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1419 		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1420 			/* Assoc is over */
1421 			return (1);
1422 		}
1423 	}
1424 	/* Zero PBA, if it needs it */
1425 	if (net->partial_bytes_acked) {
1426 		net->partial_bytes_acked = 0;
1427 	}
1428 	if ((stcb->asoc.total_output_queue_size > 0) &&
1429 	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1430 	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1431 		sctp_audit_stream_queues_for_size(inp, stcb);
1432 	}
1433 	if (!(net->dest_state & SCTP_ADDR_NOHB) &&
1434 	    !((net_was_pf == 0) && (net->dest_state & SCTP_ADDR_PF))) {
1435 		/*
1436 		 * when move to PF during threshold mangement, a HB has been
1437 		 * queued in that routine
1438 		 */
1439 		uint32_t ms_gone_by;
1440 
1441 		if ((net->last_sent_time.tv_sec > 0) ||
1442 		    (net->last_sent_time.tv_usec > 0)) {
1443 			struct timeval diff;
1444 
1445 			SCTP_GETTIME_TIMEVAL(&diff);
1446 			timevalsub(&diff, &net->last_sent_time);
1447 			ms_gone_by = (uint32_t) (diff.tv_sec * 1000) +
1448 			    (uint32_t) (diff.tv_usec / 1000);
1449 		} else {
1450 			ms_gone_by = 0xffffffff;
1451 		}
1452 		if ((ms_gone_by >= net->heart_beat_delay) ||
1453 		    (net->dest_state & SCTP_ADDR_PF)) {
1454 			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1455 		}
1456 	}
1457 	return (0);
1458 }
1459 
1460 void
1461 sctp_pathmtu_timer(struct sctp_inpcb *inp,
1462     struct sctp_tcb *stcb,
1463     struct sctp_nets *net)
1464 {
1465 	uint32_t next_mtu, mtu;
1466 
1467 	next_mtu = sctp_get_next_mtu(net->mtu);
1468 
1469 	if ((next_mtu > net->mtu) && (net->port == 0)) {
1470 		if ((net->src_addr_selected == 0) ||
1471 		    (net->ro._s_addr == NULL) ||
1472 		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1473 			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1474 				sctp_free_ifa(net->ro._s_addr);
1475 				net->ro._s_addr = NULL;
1476 				net->src_addr_selected = 0;
1477 			} else if (net->ro._s_addr == NULL) {
1478 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1479 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1480 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1481 
1482 					/* KAME hack: embed scopeid */
1483 					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1484 				}
1485 #endif
1486 
1487 				net->ro._s_addr = sctp_source_address_selection(inp,
1488 				    stcb,
1489 				    (sctp_route_t *) & net->ro,
1490 				    net, 0, stcb->asoc.vrf_id);
1491 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1492 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1493 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1494 
1495 					(void)sa6_recoverscope(sin6);
1496 				}
1497 #endif				/* INET6 */
1498 			}
1499 			if (net->ro._s_addr)
1500 				net->src_addr_selected = 1;
1501 		}
1502 		if (net->ro._s_addr) {
1503 			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1504 			if (net->port) {
1505 				mtu -= sizeof(struct udphdr);
1506 			}
1507 			if (mtu > next_mtu) {
1508 				net->mtu = next_mtu;
1509 			}
1510 		}
1511 	}
1512 	/* restart the timer */
1513 	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1514 }
1515 
1516 void
1517 sctp_autoclose_timer(struct sctp_inpcb *inp,
1518     struct sctp_tcb *stcb,
1519     struct sctp_nets *net)
1520 {
1521 	struct timeval tn, *tim_touse;
1522 	struct sctp_association *asoc;
1523 	int ticks_gone_by;
1524 
1525 	(void)SCTP_GETTIME_TIMEVAL(&tn);
1526 	if (stcb->asoc.sctp_autoclose_ticks &&
1527 	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1528 		/* Auto close is on */
1529 		asoc = &stcb->asoc;
1530 		/* pick the time to use */
1531 		if (asoc->time_last_rcvd.tv_sec >
1532 		    asoc->time_last_sent.tv_sec) {
1533 			tim_touse = &asoc->time_last_rcvd;
1534 		} else {
1535 			tim_touse = &asoc->time_last_sent;
1536 		}
1537 		/* Now has long enough transpired to autoclose? */
1538 		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1539 		if ((ticks_gone_by > 0) &&
1540 		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1541 			/*
1542 			 * autoclose time has hit, call the output routine,
1543 			 * which should do nothing just to be SURE we don't
1544 			 * have hanging data. We can then safely check the
1545 			 * queues and know that we are clear to send
1546 			 * shutdown
1547 			 */
1548 			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1549 			/* Are we clean? */
1550 			if (TAILQ_EMPTY(&asoc->send_queue) &&
1551 			    TAILQ_EMPTY(&asoc->sent_queue)) {
1552 				/*
1553 				 * there is nothing queued to send, so I'm
1554 				 * done...
1555 				 */
1556 				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1557 					/* only send SHUTDOWN 1st time thru */
1558 					struct sctp_nets *netp;
1559 
1560 					if (stcb->asoc.alternate) {
1561 						netp = stcb->asoc.alternate;
1562 					} else {
1563 						netp = stcb->asoc.primary_destination;
1564 					}
1565 					sctp_send_shutdown(stcb, netp);
1566 					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1567 					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1568 						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1569 					}
1570 					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1571 					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1572 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1573 					    stcb->sctp_ep, stcb,
1574 					    netp);
1575 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1576 					    stcb->sctp_ep, stcb,
1577 					    netp);
1578 				}
1579 			}
1580 		} else {
1581 			/*
1582 			 * No auto close at this time, reset t-o to check
1583 			 * later
1584 			 */
1585 			int tmp;
1586 
1587 			/* fool the timer startup to use the time left */
1588 			tmp = asoc->sctp_autoclose_ticks;
1589 			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1590 			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1591 			    net);
1592 			/* restore the real tick value */
1593 			asoc->sctp_autoclose_ticks = tmp;
1594 		}
1595 	}
1596 }
1597