xref: /freebsd/sys/netinet/sctp_timer.c (revision 0572ccaa4543b0abef8ef81e384c1d04de9f3da1)
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 #include <netinet/udp.h>
53 
54 
55 void
56 sctp_audit_retranmission_queue(struct sctp_association *asoc)
57 {
58 	struct sctp_tmit_chunk *chk;
59 
60 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
61 	    asoc->sent_queue_retran_cnt,
62 	    asoc->sent_queue_cnt);
63 	asoc->sent_queue_retran_cnt = 0;
64 	asoc->sent_queue_cnt = 0;
65 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
66 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
67 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
68 		}
69 		asoc->sent_queue_cnt++;
70 	}
71 	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
72 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
73 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
74 		}
75 	}
76 	TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) {
77 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
78 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
79 		}
80 	}
81 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
82 	    asoc->sent_queue_retran_cnt,
83 	    asoc->sent_queue_cnt);
84 }
85 
86 int
87 sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
88     struct sctp_nets *net, uint16_t threshold)
89 {
90 	if (net) {
91 		net->error_count++;
92 		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
93 		    (void *)net, net->error_count,
94 		    net->failure_threshold);
95 		if (net->error_count > net->failure_threshold) {
96 			/* We had a threshold failure */
97 			if (net->dest_state & SCTP_ADDR_REACHABLE) {
98 				net->dest_state &= ~SCTP_ADDR_REACHABLE;
99 				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
100 				net->dest_state &= ~SCTP_ADDR_PF;
101 				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
102 				    stcb, 0,
103 				    (void *)net, SCTP_SO_NOT_LOCKED);
104 			}
105 		} else if ((net->pf_threshold < net->failure_threshold) &&
106 		    (net->error_count > net->pf_threshold)) {
107 			if (!(net->dest_state & SCTP_ADDR_PF)) {
108 				net->dest_state |= SCTP_ADDR_PF;
109 				net->last_active = sctp_get_tick_count();
110 				sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
111 				sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
112 				sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net);
113 			}
114 		}
115 	}
116 	if (stcb == NULL)
117 		return (0);
118 
119 	if (net) {
120 		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
121 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
122 				sctp_misc_ints(SCTP_THRESHOLD_INCR,
123 				    stcb->asoc.overall_error_count,
124 				    (stcb->asoc.overall_error_count + 1),
125 				    SCTP_FROM_SCTP_TIMER,
126 				    __LINE__);
127 			}
128 			stcb->asoc.overall_error_count++;
129 		}
130 	} else {
131 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
132 			sctp_misc_ints(SCTP_THRESHOLD_INCR,
133 			    stcb->asoc.overall_error_count,
134 			    (stcb->asoc.overall_error_count + 1),
135 			    SCTP_FROM_SCTP_TIMER,
136 			    __LINE__);
137 		}
138 		stcb->asoc.overall_error_count++;
139 	}
140 	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
141 	    (void *)&stcb->asoc, stcb->asoc.overall_error_count,
142 	    (uint32_t) threshold,
143 	    ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state));
144 	/*
145 	 * We specifically do not do >= to give the assoc one more change
146 	 * before we fail it.
147 	 */
148 	if (stcb->asoc.overall_error_count > threshold) {
149 		/* Abort notification sends a ULP notify */
150 		struct mbuf *op_err;
151 
152 		op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION,
153 		    "Association error couter exceeded");
154 		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_1;
155 		sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
156 		return (1);
157 	}
158 	return (0);
159 }
160 
161 /*
162  * sctp_find_alternate_net() returns a non-NULL pointer as long
163  * the argument net is non-NULL.
164  */
165 struct sctp_nets *
166 sctp_find_alternate_net(struct sctp_tcb *stcb,
167     struct sctp_nets *net,
168     int mode)
169 {
170 	/* Find and return an alternate network if possible */
171 	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
172 	int once;
173 
174 	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
175 	int min_errors = -1;
176 	uint32_t max_cwnd = 0;
177 
178 	if (stcb->asoc.numnets == 1) {
179 		/* No others but net */
180 		return (TAILQ_FIRST(&stcb->asoc.nets));
181 	}
182 	/*
183 	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
184 	 * net algorithm. This algorithm chooses the active destination (not
185 	 * in PF state) with the largest cwnd value. If all destinations are
186 	 * in PF state, unreachable, or unconfirmed, choose the desination
187 	 * that is in PF state with the lowest error count. In case of a
188 	 * tie, choose the destination that was most recently active.
189 	 */
190 	if (mode == 2) {
191 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
192 			/*
193 			 * JRS 5/14/07 - If the destination is unreachable
194 			 * or unconfirmed, skip it.
195 			 */
196 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
197 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
198 				continue;
199 			}
200 			/*
201 			 * JRS 5/14/07 -  If the destination is reachable
202 			 * but in PF state, compare the error count of the
203 			 * destination to the minimum error count seen thus
204 			 * far. Store the destination with the lower error
205 			 * count.  If the error counts are equal, store the
206 			 * destination that was most recently active.
207 			 */
208 			if (mnet->dest_state & SCTP_ADDR_PF) {
209 				/*
210 				 * JRS 5/14/07 - If the destination under
211 				 * consideration is the current destination,
212 				 * work as if the error count is one higher.
213 				 * The actual error count will not be
214 				 * incremented until later in the t3
215 				 * handler.
216 				 */
217 				if (mnet == net) {
218 					if (min_errors == -1) {
219 						min_errors = mnet->error_count + 1;
220 						min_errors_net = mnet;
221 					} else if (mnet->error_count + 1 < min_errors) {
222 						min_errors = mnet->error_count + 1;
223 						min_errors_net = mnet;
224 					} else if (mnet->error_count + 1 == min_errors
225 					    && mnet->last_active > min_errors_net->last_active) {
226 						min_errors_net = mnet;
227 						min_errors = mnet->error_count + 1;
228 					}
229 					continue;
230 				} else {
231 					if (min_errors == -1) {
232 						min_errors = mnet->error_count;
233 						min_errors_net = mnet;
234 					} else if (mnet->error_count < min_errors) {
235 						min_errors = mnet->error_count;
236 						min_errors_net = mnet;
237 					} else if (mnet->error_count == min_errors
238 					    && mnet->last_active > min_errors_net->last_active) {
239 						min_errors_net = mnet;
240 						min_errors = mnet->error_count;
241 					}
242 					continue;
243 				}
244 			}
245 			/*
246 			 * JRS 5/14/07 - If the destination is reachable and
247 			 * not in PF state, compare the cwnd of the
248 			 * destination to the highest cwnd seen thus far.
249 			 * Store the destination with the higher cwnd value.
250 			 * If the cwnd values are equal, randomly choose one
251 			 * of the two destinations.
252 			 */
253 			if (max_cwnd < mnet->cwnd) {
254 				max_cwnd_net = mnet;
255 				max_cwnd = mnet->cwnd;
256 			} else if (max_cwnd == mnet->cwnd) {
257 				uint32_t rndval;
258 				uint8_t this_random;
259 
260 				if (stcb->asoc.hb_random_idx > 3) {
261 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
262 					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
263 					this_random = stcb->asoc.hb_random_values[0];
264 					stcb->asoc.hb_random_idx++;
265 					stcb->asoc.hb_ect_randombit = 0;
266 				} else {
267 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
268 					stcb->asoc.hb_random_idx++;
269 					stcb->asoc.hb_ect_randombit = 0;
270 				}
271 				if (this_random % 2 == 1) {
272 					max_cwnd_net = mnet;
273 					max_cwnd = mnet->cwnd;	/* Useless? */
274 				}
275 			}
276 		}
277 		if (max_cwnd_net == NULL) {
278 			if (min_errors_net == NULL) {
279 				return (net);
280 			}
281 			return (min_errors_net);
282 		} else {
283 			return (max_cwnd_net);
284 		}
285 	}
286 	/*
287 	 * JRS 5/14/07 - If mode is set to 1, use the CMT policy for
288 	 * choosing an alternate net.
289 	 */
290 	else if (mode == 1) {
291 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
292 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
293 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
294 				/*
295 				 * will skip ones that are not-reachable or
296 				 * unconfirmed
297 				 */
298 				continue;
299 			}
300 			if (max_cwnd < mnet->cwnd) {
301 				max_cwnd_net = mnet;
302 				max_cwnd = mnet->cwnd;
303 			} else if (max_cwnd == mnet->cwnd) {
304 				uint32_t rndval;
305 				uint8_t this_random;
306 
307 				if (stcb->asoc.hb_random_idx > 3) {
308 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
309 					memcpy(stcb->asoc.hb_random_values, &rndval,
310 					    sizeof(stcb->asoc.hb_random_values));
311 					this_random = stcb->asoc.hb_random_values[0];
312 					stcb->asoc.hb_random_idx = 0;
313 					stcb->asoc.hb_ect_randombit = 0;
314 				} else {
315 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
316 					stcb->asoc.hb_random_idx++;
317 					stcb->asoc.hb_ect_randombit = 0;
318 				}
319 				if (this_random % 2) {
320 					max_cwnd_net = mnet;
321 					max_cwnd = mnet->cwnd;
322 				}
323 			}
324 		}
325 		if (max_cwnd_net) {
326 			return (max_cwnd_net);
327 		}
328 	}
329 	mnet = net;
330 	once = 0;
331 
332 	if (mnet == NULL) {
333 		mnet = TAILQ_FIRST(&stcb->asoc.nets);
334 		if (mnet == NULL) {
335 			return (NULL);
336 		}
337 	}
338 	do {
339 		alt = TAILQ_NEXT(mnet, sctp_next);
340 		if (alt == NULL) {
341 			once++;
342 			if (once > 1) {
343 				break;
344 			}
345 			alt = TAILQ_FIRST(&stcb->asoc.nets);
346 			if (alt == NULL) {
347 				return (NULL);
348 			}
349 		}
350 		if (alt->ro.ro_rt == NULL) {
351 			if (alt->ro._s_addr) {
352 				sctp_free_ifa(alt->ro._s_addr);
353 				alt->ro._s_addr = NULL;
354 			}
355 			alt->src_addr_selected = 0;
356 		}
357 		/* sa_ignore NO_NULL_CHK */
358 		if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
359 		    (alt->ro.ro_rt != NULL) &&
360 		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))) {
361 			/* Found a reachable address */
362 			break;
363 		}
364 		mnet = alt;
365 	} while (alt != NULL);
366 
367 	if (alt == NULL) {
368 		/* Case where NO insv network exists (dormant state) */
369 		/* we rotate destinations */
370 		once = 0;
371 		mnet = net;
372 		do {
373 			if (mnet == NULL) {
374 				return (TAILQ_FIRST(&stcb->asoc.nets));
375 			}
376 			alt = TAILQ_NEXT(mnet, sctp_next);
377 			if (alt == NULL) {
378 				once++;
379 				if (once > 1) {
380 					break;
381 				}
382 				alt = TAILQ_FIRST(&stcb->asoc.nets);
383 			}
384 			/* sa_ignore NO_NULL_CHK */
385 			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
386 			    (alt != net)) {
387 				/* Found an alternate address */
388 				break;
389 			}
390 			mnet = alt;
391 		} while (alt != NULL);
392 	}
393 	if (alt == NULL) {
394 		return (net);
395 	}
396 	return (alt);
397 }
398 
399 static void
400 sctp_backoff_on_timeout(struct sctp_tcb *stcb,
401     struct sctp_nets *net,
402     int win_probe,
403     int num_marked, int num_abandoned)
404 {
405 	if (net->RTO == 0) {
406 		net->RTO = stcb->asoc.minrto;
407 	}
408 	net->RTO <<= 1;
409 	if (net->RTO > stcb->asoc.maxrto) {
410 		net->RTO = stcb->asoc.maxrto;
411 	}
412 	if ((win_probe == 0) && (num_marked || num_abandoned)) {
413 		/* We don't apply penalty to window probe scenarios */
414 		/* JRS - Use the congestion control given in the CC module */
415 		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
416 	}
417 }
418 
419 #ifndef INVARIANTS
420 static void
421 sctp_recover_sent_list(struct sctp_tcb *stcb)
422 {
423 	struct sctp_tmit_chunk *chk, *nchk;
424 	struct sctp_association *asoc;
425 
426 	asoc = &stcb->asoc;
427 	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
428 		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.TSN_seq)) {
429 			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
430 			    (void *)chk, chk->rec.data.TSN_seq, asoc->last_acked_seq);
431 			if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
432 				if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
433 					asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
434 				}
435 			}
436 			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
437 			if (PR_SCTP_ENABLED(chk->flags)) {
438 				if (asoc->pr_sctp_cnt != 0)
439 					asoc->pr_sctp_cnt--;
440 			}
441 			if (chk->data) {
442 				/* sa_ignore NO_NULL_CHK */
443 				sctp_free_bufspace(stcb, asoc, chk, 1);
444 				sctp_m_freem(chk->data);
445 				chk->data = NULL;
446 				if (asoc->peer_supports_prsctp && PR_SCTP_BUF_ENABLED(chk->flags)) {
447 					asoc->sent_queue_cnt_removeable--;
448 				}
449 			}
450 			asoc->sent_queue_cnt--;
451 			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
452 		}
453 	}
454 	SCTP_PRINTF("after recover order is as follows\n");
455 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
456 		SCTP_PRINTF("chk:%p TSN:%x\n", (void *)chk, chk->rec.data.TSN_seq);
457 	}
458 }
459 
460 #endif
461 
462 static int
463 sctp_mark_all_for_resend(struct sctp_tcb *stcb,
464     struct sctp_nets *net,
465     struct sctp_nets *alt,
466     int window_probe,
467     int *num_marked,
468     int *num_abandoned)
469 {
470 
471 	/*
472 	 * Mark all chunks (well not all) that were sent to *net for
473 	 * retransmission. Move them to alt for there destination as well...
474 	 * We only mark chunks that have been outstanding long enough to
475 	 * have received feed-back.
476 	 */
477 	struct sctp_tmit_chunk *chk, *nchk;
478 	struct sctp_nets *lnets;
479 	struct timeval now, min_wait, tv;
480 	int cur_rto;
481 	int cnt_abandoned;
482 	int audit_tf, num_mk, fir;
483 	unsigned int cnt_mk;
484 	uint32_t orig_flight, orig_tf;
485 	uint32_t tsnlast, tsnfirst;
486 	int recovery_cnt = 0;
487 
488 
489 	/* none in flight now */
490 	audit_tf = 0;
491 	fir = 0;
492 	/*
493 	 * figure out how long a data chunk must be pending before we can
494 	 * mark it ..
495 	 */
496 	(void)SCTP_GETTIME_TIMEVAL(&now);
497 	/* get cur rto in micro-seconds */
498 	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
499 	cur_rto *= 1000;
500 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
501 		sctp_log_fr(cur_rto,
502 		    stcb->asoc.peers_rwnd,
503 		    window_probe,
504 		    SCTP_FR_T3_MARK_TIME);
505 		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
506 		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
507 	}
508 	tv.tv_sec = cur_rto / 1000000;
509 	tv.tv_usec = cur_rto % 1000000;
510 	min_wait = now;
511 	timevalsub(&min_wait, &tv);
512 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
513 		/*
514 		 * if we hit here, we don't have enough seconds on the clock
515 		 * to account for the RTO. We just let the lower seconds be
516 		 * the bounds and don't worry about it. This may mean we
517 		 * will mark a lot more than we should.
518 		 */
519 		min_wait.tv_sec = min_wait.tv_usec = 0;
520 	}
521 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
522 		sctp_log_fr(cur_rto, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
523 		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
524 	}
525 	/*
526 	 * Our rwnd will be incorrect here since we are not adding back the
527 	 * cnt * mbuf but we will fix that down below.
528 	 */
529 	orig_flight = net->flight_size;
530 	orig_tf = stcb->asoc.total_flight;
531 
532 	net->fast_retran_ip = 0;
533 	/* Now on to each chunk */
534 	cnt_abandoned = 0;
535 	num_mk = cnt_mk = 0;
536 	tsnfirst = tsnlast = 0;
537 #ifndef INVARIANTS
538 start_again:
539 #endif
540 	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
541 		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.TSN_seq)) {
542 			/* Strange case our list got out of order? */
543 			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x\n",
544 			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq);
545 			recovery_cnt++;
546 #ifdef INVARIANTS
547 			panic("last acked >= chk on sent-Q");
548 #else
549 			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
550 			sctp_recover_sent_list(stcb);
551 			if (recovery_cnt < 10) {
552 				goto start_again;
553 			} else {
554 				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
555 			}
556 #endif
557 		}
558 		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
559 			/*
560 			 * found one to mark: If it is less than
561 			 * DATAGRAM_ACKED it MUST not be a skipped or marked
562 			 * TSN but instead one that is either already set
563 			 * for retransmission OR one that needs
564 			 * retransmission.
565 			 */
566 
567 			/* validate its been outstanding long enough */
568 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
569 				sctp_log_fr(chk->rec.data.TSN_seq,
570 				    chk->sent_rcv_time.tv_sec,
571 				    chk->sent_rcv_time.tv_usec,
572 				    SCTP_FR_T3_MARK_TIME);
573 			}
574 			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
575 				/*
576 				 * we have reached a chunk that was sent
577 				 * some seconds past our min.. forget it we
578 				 * will find no more to send.
579 				 */
580 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
581 					sctp_log_fr(0,
582 					    chk->sent_rcv_time.tv_sec,
583 					    chk->sent_rcv_time.tv_usec,
584 					    SCTP_FR_T3_STOPPED);
585 				}
586 				continue;
587 			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
588 			    (window_probe == 0)) {
589 				/*
590 				 * we must look at the micro seconds to
591 				 * know.
592 				 */
593 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
594 					/*
595 					 * ok it was sent after our boundary
596 					 * time.
597 					 */
598 					continue;
599 				}
600 			}
601 			if (stcb->asoc.peer_supports_prsctp && PR_SCTP_TTL_ENABLED(chk->flags)) {
602 				/* Is it expired? */
603 				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
604 					/* Yes so drop it */
605 					if (chk->data) {
606 						(void)sctp_release_pr_sctp_chunk(stcb,
607 						    chk,
608 						    1,
609 						    SCTP_SO_NOT_LOCKED);
610 						cnt_abandoned++;
611 					}
612 					continue;
613 				}
614 			}
615 			if (stcb->asoc.peer_supports_prsctp && PR_SCTP_RTX_ENABLED(chk->flags)) {
616 				/* Has it been retransmitted tv_sec times? */
617 				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
618 					if (chk->data) {
619 						(void)sctp_release_pr_sctp_chunk(stcb,
620 						    chk,
621 						    1,
622 						    SCTP_SO_NOT_LOCKED);
623 						cnt_abandoned++;
624 					}
625 					continue;
626 				}
627 			}
628 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
629 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
630 				num_mk++;
631 				if (fir == 0) {
632 					fir = 1;
633 					tsnfirst = chk->rec.data.TSN_seq;
634 				}
635 				tsnlast = chk->rec.data.TSN_seq;
636 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
637 					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
638 					    0, SCTP_FR_T3_MARKED);
639 				}
640 				if (chk->rec.data.chunk_was_revoked) {
641 					/* deflate the cwnd */
642 					chk->whoTo->cwnd -= chk->book_size;
643 					chk->rec.data.chunk_was_revoked = 0;
644 				}
645 				net->marked_retrans++;
646 				stcb->asoc.marked_retrans++;
647 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
648 					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
649 					    chk->whoTo->flight_size,
650 					    chk->book_size,
651 					    (uintptr_t) chk->whoTo,
652 					    chk->rec.data.TSN_seq);
653 				}
654 				sctp_flight_size_decrease(chk);
655 				sctp_total_flight_decrease(stcb, chk);
656 				stcb->asoc.peers_rwnd += chk->send_size;
657 				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
658 			}
659 			chk->sent = SCTP_DATAGRAM_RESEND;
660 			SCTP_STAT_INCR(sctps_markedretrans);
661 
662 			/* reset the TSN for striking and other FR stuff */
663 			chk->rec.data.doing_fast_retransmit = 0;
664 			/* Clear any time so NO RTT is being done */
665 
666 			if (chk->do_rtt) {
667 				if (chk->whoTo->rto_needed == 0) {
668 					chk->whoTo->rto_needed = 1;
669 				}
670 			}
671 			chk->do_rtt = 0;
672 			if (alt != net) {
673 				sctp_free_remote_addr(chk->whoTo);
674 				chk->no_fr_allowed = 1;
675 				chk->whoTo = alt;
676 				atomic_add_int(&alt->ref_count, 1);
677 			} else {
678 				chk->no_fr_allowed = 0;
679 				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
680 					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
681 				} else {
682 					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
683 				}
684 			}
685 			/*
686 			 * CMT: Do not allow FRs on retransmitted TSNs.
687 			 */
688 			if (stcb->asoc.sctp_cmt_on_off > 0) {
689 				chk->no_fr_allowed = 1;
690 			}
691 #ifdef THIS_SHOULD_NOT_BE_DONE
692 		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
693 			/* remember highest acked one */
694 			could_be_sent = chk;
695 #endif
696 		}
697 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
698 			cnt_mk++;
699 		}
700 	}
701 	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
702 		/* we did not subtract the same things? */
703 		audit_tf = 1;
704 	}
705 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
706 		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
707 	}
708 #ifdef SCTP_DEBUG
709 	if (num_mk) {
710 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
711 		    tsnlast);
712 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%ld\n",
713 		    num_mk, (u_long)stcb->asoc.peers_rwnd);
714 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
715 		    tsnlast);
716 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n",
717 		    num_mk,
718 		    (int)stcb->asoc.peers_rwnd);
719 	}
720 #endif
721 	*num_marked = num_mk;
722 	*num_abandoned = cnt_abandoned;
723 	/*
724 	 * Now check for a ECN Echo that may be stranded And include the
725 	 * cnt_mk'd to have all resends in the control queue.
726 	 */
727 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
728 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
729 			cnt_mk++;
730 		}
731 		if ((chk->whoTo == net) &&
732 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
733 			sctp_free_remote_addr(chk->whoTo);
734 			chk->whoTo = alt;
735 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
736 				chk->sent = SCTP_DATAGRAM_RESEND;
737 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
738 				cnt_mk++;
739 			}
740 			atomic_add_int(&alt->ref_count, 1);
741 		}
742 	}
743 #ifdef THIS_SHOULD_NOT_BE_DONE
744 	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
745 		/* fix it so we retransmit the highest acked anyway */
746 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
747 		cnt_mk++;
748 		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
749 	}
750 #endif
751 	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
752 #ifdef INVARIANTS
753 		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
754 		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
755 #endif
756 #ifndef SCTP_AUDITING_ENABLED
757 		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
758 #endif
759 	}
760 	if (audit_tf) {
761 		SCTPDBG(SCTP_DEBUG_TIMER4,
762 		    "Audit total flight due to negative value net:%p\n",
763 		    (void *)net);
764 		stcb->asoc.total_flight = 0;
765 		stcb->asoc.total_flight_count = 0;
766 		/* Clear all networks flight size */
767 		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
768 			lnets->flight_size = 0;
769 			SCTPDBG(SCTP_DEBUG_TIMER4,
770 			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
771 			    (void *)lnets, lnets->cwnd, lnets->ssthresh);
772 		}
773 		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
774 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
775 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
776 					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
777 					    chk->whoTo->flight_size,
778 					    chk->book_size,
779 					    (uintptr_t) chk->whoTo,
780 					    chk->rec.data.TSN_seq);
781 				}
782 				sctp_flight_size_increase(chk);
783 				sctp_total_flight_increase(stcb, chk);
784 			}
785 		}
786 	}
787 	/* We return 1 if we only have a window probe outstanding */
788 	return (0);
789 }
790 
791 
792 int
793 sctp_t3rxt_timer(struct sctp_inpcb *inp,
794     struct sctp_tcb *stcb,
795     struct sctp_nets *net)
796 {
797 	struct sctp_nets *alt;
798 	int win_probe, num_mk, num_abandoned;
799 
800 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
801 		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
802 	}
803 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
804 		struct sctp_nets *lnet;
805 
806 		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
807 			if (net == lnet) {
808 				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
809 			} else {
810 				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
811 			}
812 		}
813 	}
814 	/* Find an alternate and mark those for retransmission */
815 	if ((stcb->asoc.peers_rwnd == 0) &&
816 	    (stcb->asoc.total_flight < net->mtu)) {
817 		SCTP_STAT_INCR(sctps_timowindowprobe);
818 		win_probe = 1;
819 	} else {
820 		win_probe = 0;
821 	}
822 
823 	if (win_probe == 0) {
824 		/* We don't do normal threshold management on window probes */
825 		if (sctp_threshold_management(inp, stcb, net,
826 		    stcb->asoc.max_send_times)) {
827 			/* Association was destroyed */
828 			return (1);
829 		} else {
830 			if (net != stcb->asoc.primary_destination) {
831 				/* send a immediate HB if our RTO is stale */
832 				struct timeval now;
833 				unsigned int ms_goneby;
834 
835 				(void)SCTP_GETTIME_TIMEVAL(&now);
836 				if (net->last_sent_time.tv_sec) {
837 					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
838 				} else {
839 					ms_goneby = 0;
840 				}
841 				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
842 					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
843 						/*
844 						 * no recent feed back in an
845 						 * RTO or more, request a
846 						 * RTT update
847 						 */
848 						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
849 					}
850 				}
851 			}
852 		}
853 	} else {
854 		/*
855 		 * For a window probe we don't penalize the net's but only
856 		 * the association. This may fail it if SACKs are not coming
857 		 * back. If sack's are coming with rwnd locked at 0, we will
858 		 * continue to hold things waiting for rwnd to raise
859 		 */
860 		if (sctp_threshold_management(inp, stcb, NULL,
861 		    stcb->asoc.max_send_times)) {
862 			/* Association was destroyed */
863 			return (1);
864 		}
865 	}
866 	if (stcb->asoc.sctp_cmt_on_off > 0) {
867 		if (net->pf_threshold < net->failure_threshold) {
868 			alt = sctp_find_alternate_net(stcb, net, 2);
869 		} else {
870 			/*
871 			 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is
872 			 * being used, then pick dest with largest ssthresh
873 			 * for any retransmission.
874 			 */
875 			alt = sctp_find_alternate_net(stcb, net, 1);
876 			/*
877 			 * CUCv2: If a different dest is picked for the
878 			 * retransmission, then new (rtx-)pseudo_cumack
879 			 * needs to be tracked for orig dest. Let CUCv2
880 			 * track new (rtx-) pseudo-cumack always.
881 			 */
882 			net->find_pseudo_cumack = 1;
883 			net->find_rtx_pseudo_cumack = 1;
884 		}
885 	} else {
886 		alt = sctp_find_alternate_net(stcb, net, 0);
887 	}
888 
889 	num_mk = 0;
890 	num_abandoned = 0;
891 	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
892 	    &num_mk, &num_abandoned);
893 	/* FR Loss recovery just ended with the T3. */
894 	stcb->asoc.fast_retran_loss_recovery = 0;
895 
896 	/* CMT FR loss recovery ended with the T3 */
897 	net->fast_retran_loss_recovery = 0;
898 	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
899 	    (net->flight_size == 0)) {
900 		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) (stcb, net);
901 	}
902 	/*
903 	 * setup the sat loss recovery that prevents satellite cwnd advance.
904 	 */
905 	stcb->asoc.sat_t3_loss_recovery = 1;
906 	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
907 
908 	/* Backoff the timer and cwnd */
909 	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
910 	if ((!(net->dest_state & SCTP_ADDR_REACHABLE)) ||
911 	    (net->dest_state & SCTP_ADDR_PF)) {
912 		/* Move all pending over too */
913 		sctp_move_chunks_from_net(stcb, net);
914 
915 		/*
916 		 * Get the address that failed, to force a new src address
917 		 * selecton and a route allocation.
918 		 */
919 		if (net->ro._s_addr) {
920 			sctp_free_ifa(net->ro._s_addr);
921 			net->ro._s_addr = NULL;
922 		}
923 		net->src_addr_selected = 0;
924 
925 		/* Force a route allocation too */
926 		if (net->ro.ro_rt) {
927 			RTFREE(net->ro.ro_rt);
928 			net->ro.ro_rt = NULL;
929 		}
930 		/* Was it our primary? */
931 		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
932 			/*
933 			 * Yes, note it as such and find an alternate note:
934 			 * this means HB code must use this to resent the
935 			 * primary if it goes active AND if someone does a
936 			 * change-primary then this flag must be cleared
937 			 * from any net structures.
938 			 */
939 			if (stcb->asoc.alternate) {
940 				sctp_free_remote_addr(stcb->asoc.alternate);
941 			}
942 			stcb->asoc.alternate = alt;
943 			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
944 		}
945 	}
946 	/*
947 	 * Special case for cookie-echo'ed case, we don't do output but must
948 	 * await the COOKIE-ACK before retransmission
949 	 */
950 	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
951 		/*
952 		 * Here we just reset the timer and start again since we
953 		 * have not established the asoc
954 		 */
955 		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
956 		return (0);
957 	}
958 	if (stcb->asoc.peer_supports_prsctp) {
959 		struct sctp_tmit_chunk *lchk;
960 
961 		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
962 		/* C3. See if we need to send a Fwd-TSN */
963 		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
964 			send_forward_tsn(stcb, &stcb->asoc);
965 			if (lchk) {
966 				/* Assure a timer is up */
967 				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
968 			}
969 		}
970 	}
971 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
972 		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
973 	}
974 	return (0);
975 }
976 
977 int
978 sctp_t1init_timer(struct sctp_inpcb *inp,
979     struct sctp_tcb *stcb,
980     struct sctp_nets *net)
981 {
982 	/* bump the thresholds */
983 	if (stcb->asoc.delayed_connection) {
984 		/*
985 		 * special hook for delayed connection. The library did NOT
986 		 * complete the rest of its sends.
987 		 */
988 		stcb->asoc.delayed_connection = 0;
989 		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
990 		return (0);
991 	}
992 	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
993 		return (0);
994 	}
995 	if (sctp_threshold_management(inp, stcb, net,
996 	    stcb->asoc.max_init_times)) {
997 		/* Association was destroyed */
998 		return (1);
999 	}
1000 	stcb->asoc.dropped_special_cnt = 0;
1001 	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1002 	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1003 		net->RTO = stcb->asoc.initial_init_rto_max;
1004 	}
1005 	if (stcb->asoc.numnets > 1) {
1006 		/* If we have more than one addr use it */
1007 		struct sctp_nets *alt;
1008 
1009 		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1010 		if (alt != stcb->asoc.primary_destination) {
1011 			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1012 			stcb->asoc.primary_destination = alt;
1013 		}
1014 	}
1015 	/* Send out a new init */
1016 	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1017 	return (0);
1018 }
1019 
1020 /*
1021  * For cookie and asconf we actually need to find and mark for resend, then
1022  * increment the resend counter (after all the threshold management stuff of
1023  * course).
1024  */
1025 int
1026 sctp_cookie_timer(struct sctp_inpcb *inp,
1027     struct sctp_tcb *stcb,
1028     struct sctp_nets *net SCTP_UNUSED)
1029 {
1030 	struct sctp_nets *alt;
1031 	struct sctp_tmit_chunk *cookie;
1032 
1033 	/* first before all else we must find the cookie */
1034 	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1035 		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1036 			break;
1037 		}
1038 	}
1039 	if (cookie == NULL) {
1040 		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1041 			/* FOOBAR! */
1042 			struct mbuf *op_err;
1043 
1044 			op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION,
1045 			    "Cookie timer expired, but no cookie");
1046 			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_4;
1047 			sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
1048 		} else {
1049 #ifdef INVARIANTS
1050 			panic("Cookie timer expires in wrong state?");
1051 #else
1052 			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1053 			return (0);
1054 #endif
1055 		}
1056 		return (0);
1057 	}
1058 	/* Ok we found the cookie, threshold management next */
1059 	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1060 	    stcb->asoc.max_init_times)) {
1061 		/* Assoc is over */
1062 		return (1);
1063 	}
1064 	/*
1065 	 * cleared theshold management now lets backoff the address & select
1066 	 * an alternate
1067 	 */
1068 	stcb->asoc.dropped_special_cnt = 0;
1069 	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0);
1070 	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1071 	if (alt != cookie->whoTo) {
1072 		sctp_free_remote_addr(cookie->whoTo);
1073 		cookie->whoTo = alt;
1074 		atomic_add_int(&alt->ref_count, 1);
1075 	}
1076 	/* Now mark the retran info */
1077 	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1078 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1079 	}
1080 	cookie->sent = SCTP_DATAGRAM_RESEND;
1081 	/*
1082 	 * Now call the output routine to kick out the cookie again, Note we
1083 	 * don't mark any chunks for retran so that FR will need to kick in
1084 	 * to move these (or a send timer).
1085 	 */
1086 	return (0);
1087 }
1088 
1089 int
1090 sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1091     struct sctp_nets *net)
1092 {
1093 	struct sctp_nets *alt;
1094 	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1095 
1096 	if (stcb->asoc.stream_reset_outstanding == 0) {
1097 		return (0);
1098 	}
1099 	/* find the existing STRRESET, we use the seq number we sent out on */
1100 	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1101 	if (strrst == NULL) {
1102 		return (0);
1103 	}
1104 	/* do threshold management */
1105 	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1106 	    stcb->asoc.max_send_times)) {
1107 		/* Assoc is over */
1108 		return (1);
1109 	}
1110 	/*
1111 	 * cleared theshold management now lets backoff the address & select
1112 	 * an alternate
1113 	 */
1114 	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0, 0);
1115 	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1116 	sctp_free_remote_addr(strrst->whoTo);
1117 	strrst->whoTo = alt;
1118 	atomic_add_int(&alt->ref_count, 1);
1119 
1120 	/* See if a ECN Echo is also stranded */
1121 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1122 		if ((chk->whoTo == net) &&
1123 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1124 			sctp_free_remote_addr(chk->whoTo);
1125 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1126 				chk->sent = SCTP_DATAGRAM_RESEND;
1127 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1128 			}
1129 			chk->whoTo = alt;
1130 			atomic_add_int(&alt->ref_count, 1);
1131 		}
1132 	}
1133 	if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1134 		/*
1135 		 * If the address went un-reachable, we need to move to
1136 		 * alternates for ALL chk's in queue
1137 		 */
1138 		sctp_move_chunks_from_net(stcb, net);
1139 	}
1140 	/* mark the retran info */
1141 	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1142 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1143 	strrst->sent = SCTP_DATAGRAM_RESEND;
1144 
1145 	/* restart the timer */
1146 	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1147 	return (0);
1148 }
1149 
1150 int
1151 sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1152     struct sctp_nets *net)
1153 {
1154 	struct sctp_nets *alt;
1155 	struct sctp_tmit_chunk *asconf, *chk;
1156 
1157 	/* is this a first send, or a retransmission? */
1158 	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1159 		/* compose a new ASCONF chunk and send it */
1160 		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1161 	} else {
1162 		/*
1163 		 * Retransmission of the existing ASCONF is needed
1164 		 */
1165 
1166 		/* find the existing ASCONF */
1167 		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1168 		if (asconf == NULL) {
1169 			return (0);
1170 		}
1171 		/* do threshold management */
1172 		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1173 		    stcb->asoc.max_send_times)) {
1174 			/* Assoc is over */
1175 			return (1);
1176 		}
1177 		if (asconf->snd_count > stcb->asoc.max_send_times) {
1178 			/*
1179 			 * Something is rotten: our peer is not responding
1180 			 * to ASCONFs but apparently is to other chunks.
1181 			 * i.e. it is not properly handling the chunk type
1182 			 * upper bits. Mark this peer as ASCONF incapable
1183 			 * and cleanup.
1184 			 */
1185 			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1186 			sctp_asconf_cleanup(stcb, net);
1187 			return (0);
1188 		}
1189 		/*
1190 		 * cleared threshold management, so now backoff the net and
1191 		 * select an alternate
1192 		 */
1193 		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0, 0);
1194 		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1195 		if (asconf->whoTo != alt) {
1196 			sctp_free_remote_addr(asconf->whoTo);
1197 			asconf->whoTo = alt;
1198 			atomic_add_int(&alt->ref_count, 1);
1199 		}
1200 		/* See if an ECN Echo is also stranded */
1201 		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1202 			if ((chk->whoTo == net) &&
1203 			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1204 				sctp_free_remote_addr(chk->whoTo);
1205 				chk->whoTo = alt;
1206 				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1207 					chk->sent = SCTP_DATAGRAM_RESEND;
1208 					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1209 				}
1210 				atomic_add_int(&alt->ref_count, 1);
1211 			}
1212 		}
1213 		TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) {
1214 			if (chk->whoTo != alt) {
1215 				sctp_free_remote_addr(chk->whoTo);
1216 				chk->whoTo = alt;
1217 				atomic_add_int(&alt->ref_count, 1);
1218 			}
1219 			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1220 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1221 			chk->sent = SCTP_DATAGRAM_RESEND;
1222 		}
1223 		if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1224 			/*
1225 			 * If the address went un-reachable, we need to move
1226 			 * to the alternate for ALL chunks in queue
1227 			 */
1228 			sctp_move_chunks_from_net(stcb, net);
1229 		}
1230 		/* mark the retran info */
1231 		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1232 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1233 		asconf->sent = SCTP_DATAGRAM_RESEND;
1234 
1235 		/* send another ASCONF if any and we can do */
1236 		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1237 	}
1238 	return (0);
1239 }
1240 
1241 /* Mobility adaptation */
1242 void
1243 sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1244     struct sctp_nets *net SCTP_UNUSED)
1245 {
1246 	if (stcb->asoc.deleted_primary == NULL) {
1247 		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1248 		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1249 		return;
1250 	}
1251 	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1252 	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1253 	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1254 	stcb->asoc.deleted_primary = NULL;
1255 	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1256 	return;
1257 }
1258 
1259 /*
1260  * For the shutdown and shutdown-ack, we do not keep one around on the
1261  * control queue. This means we must generate a new one and call the general
1262  * chunk output routine, AFTER having done threshold management.
1263  * It is assumed that net is non-NULL.
1264  */
1265 int
1266 sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1267     struct sctp_nets *net)
1268 {
1269 	struct sctp_nets *alt;
1270 
1271 	/* first threshold managment */
1272 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1273 		/* Assoc is over */
1274 		return (1);
1275 	}
1276 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1277 	/* second select an alternative */
1278 	alt = sctp_find_alternate_net(stcb, net, 0);
1279 
1280 	/* third generate a shutdown into the queue for out net */
1281 	sctp_send_shutdown(stcb, alt);
1282 
1283 	/* fourth restart timer */
1284 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1285 	return (0);
1286 }
1287 
1288 int
1289 sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1290     struct sctp_nets *net)
1291 {
1292 	struct sctp_nets *alt;
1293 
1294 	/* first threshold managment */
1295 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1296 		/* Assoc is over */
1297 		return (1);
1298 	}
1299 	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1300 	/* second select an alternative */
1301 	alt = sctp_find_alternate_net(stcb, net, 0);
1302 
1303 	/* third generate a shutdown into the queue for out net */
1304 	sctp_send_shutdown_ack(stcb, alt);
1305 
1306 	/* fourth restart timer */
1307 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1308 	return (0);
1309 }
1310 
1311 static void
1312 sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1313     struct sctp_tcb *stcb)
1314 {
1315 	struct sctp_stream_queue_pending *sp;
1316 	unsigned int i, chks_in_queue = 0;
1317 	int being_filled = 0;
1318 
1319 	/*
1320 	 * This function is ONLY called when the send/sent queues are empty.
1321 	 */
1322 	if ((stcb == NULL) || (inp == NULL))
1323 		return;
1324 
1325 	if (stcb->asoc.sent_queue_retran_cnt) {
1326 		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1327 		    stcb->asoc.sent_queue_retran_cnt);
1328 		stcb->asoc.sent_queue_retran_cnt = 0;
1329 	}
1330 	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1331 		/* No stream scheduler information, initialize scheduler */
1332 		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 0);
1333 		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1334 			/* yep, we lost a stream or two */
1335 			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1336 		} else {
1337 			/* no streams lost */
1338 			stcb->asoc.total_output_queue_size = 0;
1339 		}
1340 	}
1341 	/* Check to see if some data queued, if so report it */
1342 	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1343 		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1344 			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1345 				if (sp->msg_is_complete)
1346 					being_filled++;
1347 				chks_in_queue++;
1348 			}
1349 		}
1350 	}
1351 	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1352 		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1353 		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1354 	}
1355 	if (chks_in_queue) {
1356 		/* call the output queue function */
1357 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1358 		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1359 		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1360 			/*
1361 			 * Probably should go in and make it go back through
1362 			 * and add fragments allowed
1363 			 */
1364 			if (being_filled == 0) {
1365 				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1366 				    chks_in_queue);
1367 			}
1368 		}
1369 	} else {
1370 		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1371 		    (u_long)stcb->asoc.total_output_queue_size);
1372 		stcb->asoc.total_output_queue_size = 0;
1373 	}
1374 }
1375 
1376 int
1377 sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1378     struct sctp_nets *net)
1379 {
1380 	uint8_t net_was_pf;
1381 
1382 	if (net->dest_state & SCTP_ADDR_PF) {
1383 		net_was_pf = 1;
1384 	} else {
1385 		net_was_pf = 0;
1386 	}
1387 	if (net->hb_responded == 0) {
1388 		if (net->ro._s_addr) {
1389 			/*
1390 			 * Invalidate the src address if we did not get a
1391 			 * response last time.
1392 			 */
1393 			sctp_free_ifa(net->ro._s_addr);
1394 			net->ro._s_addr = NULL;
1395 			net->src_addr_selected = 0;
1396 		}
1397 		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1398 		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1399 			/* Assoc is over */
1400 			return (1);
1401 		}
1402 	}
1403 	/* Zero PBA, if it needs it */
1404 	if (net->partial_bytes_acked) {
1405 		net->partial_bytes_acked = 0;
1406 	}
1407 	if ((stcb->asoc.total_output_queue_size > 0) &&
1408 	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1409 	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1410 		sctp_audit_stream_queues_for_size(inp, stcb);
1411 	}
1412 	if (!(net->dest_state & SCTP_ADDR_NOHB) &&
1413 	    !((net_was_pf == 0) && (net->dest_state & SCTP_ADDR_PF))) {
1414 		/*
1415 		 * when move to PF during threshold mangement, a HB has been
1416 		 * queued in that routine
1417 		 */
1418 		uint32_t ms_gone_by;
1419 
1420 		if ((net->last_sent_time.tv_sec > 0) ||
1421 		    (net->last_sent_time.tv_usec > 0)) {
1422 			struct timeval diff;
1423 
1424 			SCTP_GETTIME_TIMEVAL(&diff);
1425 			timevalsub(&diff, &net->last_sent_time);
1426 			ms_gone_by = (uint32_t) (diff.tv_sec * 1000) +
1427 			    (uint32_t) (diff.tv_usec / 1000);
1428 		} else {
1429 			ms_gone_by = 0xffffffff;
1430 		}
1431 		if ((ms_gone_by >= net->heart_beat_delay) ||
1432 		    (net->dest_state & SCTP_ADDR_PF)) {
1433 			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1434 		}
1435 	}
1436 	return (0);
1437 }
1438 
1439 void
1440 sctp_pathmtu_timer(struct sctp_inpcb *inp,
1441     struct sctp_tcb *stcb,
1442     struct sctp_nets *net)
1443 {
1444 	uint32_t next_mtu, mtu;
1445 
1446 	next_mtu = sctp_get_next_mtu(net->mtu);
1447 
1448 	if ((next_mtu > net->mtu) && (net->port == 0)) {
1449 		if ((net->src_addr_selected == 0) ||
1450 		    (net->ro._s_addr == NULL) ||
1451 		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1452 			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1453 				sctp_free_ifa(net->ro._s_addr);
1454 				net->ro._s_addr = NULL;
1455 				net->src_addr_selected = 0;
1456 			} else if (net->ro._s_addr == NULL) {
1457 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1458 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1459 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1460 
1461 					/* KAME hack: embed scopeid */
1462 					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1463 				}
1464 #endif
1465 
1466 				net->ro._s_addr = sctp_source_address_selection(inp,
1467 				    stcb,
1468 				    (sctp_route_t *) & net->ro,
1469 				    net, 0, stcb->asoc.vrf_id);
1470 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1471 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1472 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1473 
1474 					(void)sa6_recoverscope(sin6);
1475 				}
1476 #endif				/* INET6 */
1477 			}
1478 			if (net->ro._s_addr)
1479 				net->src_addr_selected = 1;
1480 		}
1481 		if (net->ro._s_addr) {
1482 			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1483 			if (net->port) {
1484 				mtu -= sizeof(struct udphdr);
1485 			}
1486 			if (mtu > next_mtu) {
1487 				net->mtu = next_mtu;
1488 			}
1489 		}
1490 	}
1491 	/* restart the timer */
1492 	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1493 }
1494 
1495 void
1496 sctp_autoclose_timer(struct sctp_inpcb *inp,
1497     struct sctp_tcb *stcb,
1498     struct sctp_nets *net)
1499 {
1500 	struct timeval tn, *tim_touse;
1501 	struct sctp_association *asoc;
1502 	int ticks_gone_by;
1503 
1504 	(void)SCTP_GETTIME_TIMEVAL(&tn);
1505 	if (stcb->asoc.sctp_autoclose_ticks &&
1506 	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1507 		/* Auto close is on */
1508 		asoc = &stcb->asoc;
1509 		/* pick the time to use */
1510 		if (asoc->time_last_rcvd.tv_sec >
1511 		    asoc->time_last_sent.tv_sec) {
1512 			tim_touse = &asoc->time_last_rcvd;
1513 		} else {
1514 			tim_touse = &asoc->time_last_sent;
1515 		}
1516 		/* Now has long enough transpired to autoclose? */
1517 		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1518 		if ((ticks_gone_by > 0) &&
1519 		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1520 			/*
1521 			 * autoclose time has hit, call the output routine,
1522 			 * which should do nothing just to be SURE we don't
1523 			 * have hanging data. We can then safely check the
1524 			 * queues and know that we are clear to send
1525 			 * shutdown
1526 			 */
1527 			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1528 			/* Are we clean? */
1529 			if (TAILQ_EMPTY(&asoc->send_queue) &&
1530 			    TAILQ_EMPTY(&asoc->sent_queue)) {
1531 				/*
1532 				 * there is nothing queued to send, so I'm
1533 				 * done...
1534 				 */
1535 				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1536 					/* only send SHUTDOWN 1st time thru */
1537 					struct sctp_nets *netp;
1538 
1539 					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1540 					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1541 						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1542 					}
1543 					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1544 					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1545 					sctp_stop_timers_for_shutdown(stcb);
1546 					if (stcb->asoc.alternate) {
1547 						netp = stcb->asoc.alternate;
1548 					} else {
1549 						netp = stcb->asoc.primary_destination;
1550 					}
1551 					sctp_send_shutdown(stcb, netp);
1552 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1553 					    stcb->sctp_ep, stcb,
1554 					    netp);
1555 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1556 					    stcb->sctp_ep, stcb,
1557 					    netp);
1558 				}
1559 			}
1560 		} else {
1561 			/*
1562 			 * No auto close at this time, reset t-o to check
1563 			 * later
1564 			 */
1565 			int tmp;
1566 
1567 			/* fool the timer startup to use the time left */
1568 			tmp = asoc->sctp_autoclose_ticks;
1569 			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1570 			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1571 			    net);
1572 			/* restore the real tick value */
1573 			asoc->sctp_autoclose_ticks = tmp;
1574 		}
1575 	}
1576 }
1577