xref: /freebsd/sys/netinet/sctp_timer.c (revision 4a5216a6dc0c3ce4cf5f2d3ee8af0c3ff3402c4f)
1 /*-
2  * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
3  *
4  * Redistribution and use in source and binary forms, with or without
5  * modification, are permitted provided that the following conditions are met:
6  *
7  * a) Redistributions of source code must retain the above copyright notice,
8  *   this list of conditions and the following disclaimer.
9  *
10  * b) Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in
12  *   the documentation and/or other materials provided with the distribution.
13  *
14  * c) Neither the name of Cisco Systems, Inc. nor the names of its
15  *    contributors may be used to endorse or promote products derived
16  *    from this software without specific prior written permission.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
20  * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
22  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
28  * THE POSSIBILITY OF SUCH DAMAGE.
29  */
30 
31 /* $KAME: sctp_timer.c,v 1.29 2005/03/06 16:04:18 itojun Exp $	 */
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 
53 
54 
55 void
56 sctp_early_fr_timer(struct sctp_inpcb *inp,
57     struct sctp_tcb *stcb,
58     struct sctp_nets *net)
59 {
60 	struct sctp_tmit_chunk *chk, *tp2;
61 	struct timeval now, min_wait, tv;
62 	unsigned int cur_rtt, cnt = 0, cnt_resend = 0;
63 
64 	/* an early FR is occuring. */
65 	(void)SCTP_GETTIME_TIMEVAL(&now);
66 	/* get cur rto in micro-seconds */
67 	if (net->lastsa == 0) {
68 		/* Hmm no rtt estimate yet? */
69 		cur_rtt = stcb->asoc.initial_rto >> 2;
70 	} else {
71 
72 		cur_rtt = ((net->lastsa >> 2) + net->lastsv) >> 1;
73 	}
74 	if (cur_rtt < SCTP_BASE_SYSCTL(sctp_early_fr_msec)) {
75 		cur_rtt = SCTP_BASE_SYSCTL(sctp_early_fr_msec);
76 	}
77 	cur_rtt *= 1000;
78 	tv.tv_sec = cur_rtt / 1000000;
79 	tv.tv_usec = cur_rtt % 1000000;
80 	min_wait = now;
81 	timevalsub(&min_wait, &tv);
82 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
83 		/*
84 		 * if we hit here, we don't have enough seconds on the clock
85 		 * to account for the RTO. We just let the lower seconds be
86 		 * the bounds and don't worry about it. This may mean we
87 		 * will mark a lot more than we should.
88 		 */
89 		min_wait.tv_sec = min_wait.tv_usec = 0;
90 	}
91 	chk = TAILQ_LAST(&stcb->asoc.sent_queue, sctpchunk_listhead);
92 	for (; chk != NULL; chk = tp2) {
93 		tp2 = TAILQ_PREV(chk, sctpchunk_listhead, sctp_next);
94 		if (chk->whoTo != net) {
95 			continue;
96 		}
97 		if (chk->sent == SCTP_DATAGRAM_RESEND)
98 			cnt_resend++;
99 		else if ((chk->sent > SCTP_DATAGRAM_UNSENT) &&
100 		    (chk->sent < SCTP_DATAGRAM_RESEND)) {
101 			/* pending, may need retran */
102 			if (chk->sent_rcv_time.tv_sec > min_wait.tv_sec) {
103 				/*
104 				 * we have reached a chunk that was sent
105 				 * some seconds past our min.. forget it we
106 				 * will find no more to send.
107 				 */
108 				continue;
109 			} else if (chk->sent_rcv_time.tv_sec == min_wait.tv_sec) {
110 				/*
111 				 * we must look at the micro seconds to
112 				 * know.
113 				 */
114 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
115 					/*
116 					 * ok it was sent after our boundary
117 					 * time.
118 					 */
119 					continue;
120 				}
121 			}
122 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_EARLYFR_LOGGING_ENABLE) {
123 				sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
124 				    4, SCTP_FR_MARKED_EARLY);
125 			}
126 			SCTP_STAT_INCR(sctps_earlyfrmrkretrans);
127 			chk->sent = SCTP_DATAGRAM_RESEND;
128 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
129 			/* double book size since we are doing an early FR */
130 			chk->book_size_scale++;
131 			cnt += chk->send_size;
132 			if ((cnt + net->flight_size) > net->cwnd) {
133 				/* Mark all we could possibly resend */
134 				break;
135 			}
136 		}
137 	}
138 	if (cnt) {
139 		/*
140 		 * JRS - Use the congestion control given in the congestion
141 		 * control module
142 		 */
143 		stcb->asoc.cc_functions.sctp_cwnd_update_after_fr_timer(inp, stcb, net);
144 	} else if (cnt_resend) {
145 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_EARLY_FR_TMR, SCTP_SO_NOT_LOCKED);
146 	}
147 	/* Restart it? */
148 	if (net->flight_size < net->cwnd) {
149 		SCTP_STAT_INCR(sctps_earlyfrstrtmr);
150 		sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net);
151 	}
152 }
153 
154 void
155 sctp_audit_retranmission_queue(struct sctp_association *asoc)
156 {
157 	struct sctp_tmit_chunk *chk;
158 
159 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
160 	    asoc->sent_queue_retran_cnt,
161 	    asoc->sent_queue_cnt);
162 	asoc->sent_queue_retran_cnt = 0;
163 	asoc->sent_queue_cnt = 0;
164 	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
165 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
166 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
167 		}
168 		asoc->sent_queue_cnt++;
169 	}
170 	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
171 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
172 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
173 		}
174 	}
175 	TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) {
176 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
177 			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
178 		}
179 	}
180 	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
181 	    asoc->sent_queue_retran_cnt,
182 	    asoc->sent_queue_cnt);
183 }
184 
185 int
186 sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
187     struct sctp_nets *net, uint16_t threshold)
188 {
189 	if (net) {
190 		net->error_count++;
191 		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
192 		    net, net->error_count,
193 		    net->failure_threshold);
194 		if (net->error_count > net->failure_threshold) {
195 			/* We had a threshold failure */
196 			if (net->dest_state & SCTP_ADDR_REACHABLE) {
197 				net->dest_state &= ~SCTP_ADDR_REACHABLE;
198 				net->dest_state |= SCTP_ADDR_NOT_REACHABLE;
199 				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
200 				if (net == stcb->asoc.primary_destination) {
201 					net->dest_state |= SCTP_ADDR_WAS_PRIMARY;
202 				}
203 				/*
204 				 * JRS 5/14/07 - If a destination is
205 				 * unreachable, the PF bit is turned off.
206 				 * This allows an unambiguous use of the PF
207 				 * bit for destinations that are reachable
208 				 * but potentially failed. If the
209 				 * destination is set to the unreachable
210 				 * state, also set the destination to the PF
211 				 * state.
212 				 */
213 				/*
214 				 * Add debug message here if destination is
215 				 * not in PF state.
216 				 */
217 				/* Stop any running T3 timers here? */
218 				if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf)) {
219 					net->dest_state &= ~SCTP_ADDR_PF;
220 					SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from PF to unreachable.\n",
221 					    net);
222 				}
223 				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
224 				    stcb,
225 				    SCTP_FAILED_THRESHOLD,
226 				    (void *)net, SCTP_SO_NOT_LOCKED);
227 			}
228 		}
229 		/*********HOLD THIS COMMENT FOR PATCH OF ALTERNATE
230 		 *********ROUTING CODE
231 		 */
232 		/*********HOLD THIS COMMENT FOR END OF PATCH OF ALTERNATE
233 		 *********ROUTING CODE
234 		 */
235 	}
236 	if (stcb == NULL)
237 		return (0);
238 
239 	if (net) {
240 		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
241 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
242 				sctp_misc_ints(SCTP_THRESHOLD_INCR,
243 				    stcb->asoc.overall_error_count,
244 				    (stcb->asoc.overall_error_count + 1),
245 				    SCTP_FROM_SCTP_TIMER,
246 				    __LINE__);
247 			}
248 			stcb->asoc.overall_error_count++;
249 		}
250 	} else {
251 		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
252 			sctp_misc_ints(SCTP_THRESHOLD_INCR,
253 			    stcb->asoc.overall_error_count,
254 			    (stcb->asoc.overall_error_count + 1),
255 			    SCTP_FROM_SCTP_TIMER,
256 			    __LINE__);
257 		}
258 		stcb->asoc.overall_error_count++;
259 	}
260 	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
261 	    &stcb->asoc, stcb->asoc.overall_error_count,
262 	    (uint32_t) threshold,
263 	    ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state));
264 	/*
265 	 * We specifically do not do >= to give the assoc one more change
266 	 * before we fail it.
267 	 */
268 	if (stcb->asoc.overall_error_count > threshold) {
269 		/* Abort notification sends a ULP notify */
270 		struct mbuf *oper;
271 
272 		oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
273 		    0, M_DONTWAIT, 1, MT_DATA);
274 		if (oper) {
275 			struct sctp_paramhdr *ph;
276 			uint32_t *ippp;
277 
278 			SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
279 			    sizeof(uint32_t);
280 			ph = mtod(oper, struct sctp_paramhdr *);
281 			ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
282 			ph->param_length = htons(SCTP_BUF_LEN(oper));
283 			ippp = (uint32_t *) (ph + 1);
284 			*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_1);
285 		}
286 		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_1;
287 		sctp_abort_an_association(inp, stcb, SCTP_FAILED_THRESHOLD, oper, SCTP_SO_NOT_LOCKED);
288 		return (1);
289 	}
290 	return (0);
291 }
292 
293 struct sctp_nets *
294 sctp_find_alternate_net(struct sctp_tcb *stcb,
295     struct sctp_nets *net,
296     int mode)
297 {
298 	/* Find and return an alternate network if possible */
299 	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
300 	int once;
301 
302 	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
303 	int min_errors = -1;
304 	uint32_t max_cwnd = 0;
305 
306 	if (stcb->asoc.numnets == 1) {
307 		/* No others but net */
308 		return (TAILQ_FIRST(&stcb->asoc.nets));
309 	}
310 	/*
311 	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
312 	 * net algorithm. This algorithm chooses the active destination (not
313 	 * in PF state) with the largest cwnd value. If all destinations are
314 	 * in PF state, unreachable, or unconfirmed, choose the desination
315 	 * that is in PF state with the lowest error count. In case of a
316 	 * tie, choose the destination that was most recently active.
317 	 */
318 	if (mode == 2) {
319 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
320 			/*
321 			 * JRS 5/14/07 - If the destination is unreachable
322 			 * or unconfirmed, skip it.
323 			 */
324 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
325 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
326 				continue;
327 			}
328 			/*
329 			 * JRS 5/14/07 -  If the destination is reachable
330 			 * but in PF state, compare the error count of the
331 			 * destination to the minimum error count seen thus
332 			 * far. Store the destination with the lower error
333 			 * count.  If the error counts are equal, store the
334 			 * destination that was most recently active.
335 			 */
336 			if (mnet->dest_state & SCTP_ADDR_PF) {
337 				/*
338 				 * JRS 5/14/07 - If the destination under
339 				 * consideration is the current destination,
340 				 * work as if the error count is one higher.
341 				 * The actual error count will not be
342 				 * incremented until later in the t3
343 				 * handler.
344 				 */
345 				if (mnet == net) {
346 					if (min_errors == -1) {
347 						min_errors = mnet->error_count + 1;
348 						min_errors_net = mnet;
349 					} else if (mnet->error_count + 1 < min_errors) {
350 						min_errors = mnet->error_count + 1;
351 						min_errors_net = mnet;
352 					} else if (mnet->error_count + 1 == min_errors
353 					    && mnet->last_active > min_errors_net->last_active) {
354 						min_errors_net = mnet;
355 						min_errors = mnet->error_count + 1;
356 					}
357 					continue;
358 				} else {
359 					if (min_errors == -1) {
360 						min_errors = mnet->error_count;
361 						min_errors_net = mnet;
362 					} else if (mnet->error_count < min_errors) {
363 						min_errors = mnet->error_count;
364 						min_errors_net = mnet;
365 					} else if (mnet->error_count == min_errors
366 					    && mnet->last_active > min_errors_net->last_active) {
367 						min_errors_net = mnet;
368 						min_errors = mnet->error_count;
369 					}
370 					continue;
371 				}
372 			}
373 			/*
374 			 * JRS 5/14/07 - If the destination is reachable and
375 			 * not in PF state, compare the cwnd of the
376 			 * destination to the highest cwnd seen thus far.
377 			 * Store the destination with the higher cwnd value.
378 			 * If the cwnd values are equal, randomly choose one
379 			 * of the two destinations.
380 			 */
381 			if (max_cwnd < mnet->cwnd) {
382 				max_cwnd_net = mnet;
383 				max_cwnd = mnet->cwnd;
384 			} else if (max_cwnd == mnet->cwnd) {
385 				uint32_t rndval;
386 				uint8_t this_random;
387 
388 				if (stcb->asoc.hb_random_idx > 3) {
389 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
390 					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
391 					this_random = stcb->asoc.hb_random_values[0];
392 					stcb->asoc.hb_random_idx++;
393 					stcb->asoc.hb_ect_randombit = 0;
394 				} else {
395 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
396 					stcb->asoc.hb_random_idx++;
397 					stcb->asoc.hb_ect_randombit = 0;
398 				}
399 				if (this_random % 2 == 1) {
400 					max_cwnd_net = mnet;
401 					max_cwnd = mnet->cwnd;	/* Useless? */
402 				}
403 			}
404 		}
405 		/*
406 		 * JRS 5/14/07 - After all destination have been considered
407 		 * as alternates, check to see if there was some active
408 		 * destination (not in PF state).  If not, check to see if
409 		 * there was some PF destination with the minimum number of
410 		 * errors.  If not, return the original destination.  If
411 		 * there is a min_errors_net, remove the PF flag from that
412 		 * destination, set the cwnd to one or two MTUs, and return
413 		 * the destination as an alt. If there was some active
414 		 * destination with a highest cwnd, return the destination
415 		 * as an alt.
416 		 */
417 		if (max_cwnd_net == NULL) {
418 			if (min_errors_net == NULL) {
419 				return (net);
420 			}
421 			min_errors_net->dest_state &= ~SCTP_ADDR_PF;
422 			min_errors_net->cwnd = min_errors_net->mtu * SCTP_BASE_SYSCTL(sctp_cmt_pf);
423 			if (SCTP_OS_TIMER_PENDING(&min_errors_net->rxt_timer.timer)) {
424 				sctp_timer_stop(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep,
425 				    stcb, min_errors_net,
426 				    SCTP_FROM_SCTP_TIMER + SCTP_LOC_2);
427 			}
428 			SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from PF to active with %d errors.\n",
429 			    min_errors_net, min_errors_net->error_count);
430 			return (min_errors_net);
431 		} else {
432 			return (max_cwnd_net);
433 		}
434 	}
435 	/*
436 	 * JRS 5/14/07 - If mode is set to 1, use the CMT policy for
437 	 * choosing an alternate net.
438 	 */
439 	else if (mode == 1) {
440 		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
441 			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
442 			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)
443 			    ) {
444 				/*
445 				 * will skip ones that are not-reachable or
446 				 * unconfirmed
447 				 */
448 				continue;
449 			}
450 			if (max_cwnd < mnet->cwnd) {
451 				max_cwnd_net = mnet;
452 				max_cwnd = mnet->cwnd;
453 			} else if (max_cwnd == mnet->cwnd) {
454 				uint32_t rndval;
455 				uint8_t this_random;
456 
457 				if (stcb->asoc.hb_random_idx > 3) {
458 					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
459 					memcpy(stcb->asoc.hb_random_values, &rndval,
460 					    sizeof(stcb->asoc.hb_random_values));
461 					this_random = stcb->asoc.hb_random_values[0];
462 					stcb->asoc.hb_random_idx = 0;
463 					stcb->asoc.hb_ect_randombit = 0;
464 				} else {
465 					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
466 					stcb->asoc.hb_random_idx++;
467 					stcb->asoc.hb_ect_randombit = 0;
468 				}
469 				if (this_random % 2) {
470 					max_cwnd_net = mnet;
471 					max_cwnd = mnet->cwnd;
472 				}
473 			}
474 		}
475 		if (max_cwnd_net) {
476 			return (max_cwnd_net);
477 		}
478 	}
479 	mnet = net;
480 	once = 0;
481 
482 	if (mnet == NULL) {
483 		mnet = TAILQ_FIRST(&stcb->asoc.nets);
484 	}
485 	do {
486 		alt = TAILQ_NEXT(mnet, sctp_next);
487 		if (alt == NULL) {
488 			once++;
489 			if (once > 1) {
490 				break;
491 			}
492 			alt = TAILQ_FIRST(&stcb->asoc.nets);
493 		}
494 		if (alt->ro.ro_rt == NULL) {
495 			if (alt->ro._s_addr) {
496 				sctp_free_ifa(alt->ro._s_addr);
497 				alt->ro._s_addr = NULL;
498 			}
499 			alt->src_addr_selected = 0;
500 		}
501 		if (
502 		    ((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
503 		    (alt->ro.ro_rt != NULL) &&
504 		/* sa_ignore NO_NULL_CHK */
505 		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))
506 		    ) {
507 			/* Found a reachable address */
508 			break;
509 		}
510 		mnet = alt;
511 	} while (alt != NULL);
512 
513 	if (alt == NULL) {
514 		/* Case where NO insv network exists (dormant state) */
515 		/* we rotate destinations */
516 		once = 0;
517 		mnet = net;
518 		do {
519 			alt = TAILQ_NEXT(mnet, sctp_next);
520 			if (alt == NULL) {
521 				once++;
522 				if (once > 1) {
523 					break;
524 				}
525 				alt = TAILQ_FIRST(&stcb->asoc.nets);
526 			}
527 			/* sa_ignore NO_NULL_CHK */
528 			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
529 			    (alt != net)) {
530 				/* Found an alternate address */
531 				break;
532 			}
533 			mnet = alt;
534 		} while (alt != NULL);
535 	}
536 	if (alt == NULL) {
537 		return (net);
538 	}
539 	return (alt);
540 }
541 
542 
543 
544 static void
545 sctp_backoff_on_timeout(struct sctp_tcb *stcb,
546     struct sctp_nets *net,
547     int win_probe,
548     int num_marked)
549 {
550 	if (net->RTO == 0) {
551 		net->RTO = stcb->asoc.minrto;
552 	}
553 	net->RTO <<= 1;
554 	if (net->RTO > stcb->asoc.maxrto) {
555 		net->RTO = stcb->asoc.maxrto;
556 	}
557 	if ((win_probe == 0) && num_marked) {
558 		/* We don't apply penalty to window probe scenarios */
559 		/* JRS - Use the congestion control given in the CC module */
560 		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
561 	}
562 }
563 
564 void
565 sctp_recover_sent_list(struct sctp_tcb *stcb)
566 {
567 	struct sctp_tmit_chunk *chk, *tp2;
568 	struct sctp_association *asoc;
569 
570 	asoc = &stcb->asoc;
571 	chk = TAILQ_FIRST(&stcb->asoc.sent_queue);
572 	for (; chk != NULL; chk = tp2) {
573 		tp2 = TAILQ_NEXT(chk, sctp_next);
574 		if ((compare_with_wrap(stcb->asoc.last_acked_seq,
575 		    chk->rec.data.TSN_seq,
576 		    MAX_TSN)) ||
577 		    (stcb->asoc.last_acked_seq == chk->rec.data.TSN_seq)) {
578 
579 			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
580 			    chk, chk->rec.data.TSN_seq, stcb->asoc.last_acked_seq);
581 			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
582 			if (chk->pr_sctp_on) {
583 				if (asoc->pr_sctp_cnt != 0)
584 					asoc->pr_sctp_cnt--;
585 			}
586 			if (chk->data) {
587 				/* sa_ignore NO_NULL_CHK */
588 				sctp_free_bufspace(stcb, asoc, chk, 1);
589 				sctp_m_freem(chk->data);
590 				if (PR_SCTP_BUF_ENABLED(chk->flags)) {
591 					asoc->sent_queue_cnt_removeable--;
592 				}
593 			}
594 			chk->data = NULL;
595 			asoc->sent_queue_cnt--;
596 			sctp_free_a_chunk(stcb, chk);
597 		}
598 	}
599 	SCTP_PRINTF("after recover order is as follows\n");
600 	chk = TAILQ_FIRST(&stcb->asoc.sent_queue);
601 	for (; chk != NULL; chk = tp2) {
602 		tp2 = TAILQ_NEXT(chk, sctp_next);
603 		SCTP_PRINTF("chk:%p TSN:%x\n", chk, chk->rec.data.TSN_seq);
604 	}
605 }
606 
607 static int
608 sctp_mark_all_for_resend(struct sctp_tcb *stcb,
609     struct sctp_nets *net,
610     struct sctp_nets *alt,
611     int window_probe,
612     int *num_marked)
613 {
614 
615 	/*
616 	 * Mark all chunks (well not all) that were sent to *net for
617 	 * retransmission. Move them to alt for there destination as well...
618 	 * We only mark chunks that have been outstanding long enough to
619 	 * have received feed-back.
620 	 */
621 	struct sctp_tmit_chunk *chk, *tp2, *could_be_sent = NULL;
622 	struct sctp_nets *lnets;
623 	struct timeval now, min_wait, tv;
624 	int cur_rtt;
625 	int audit_tf, num_mk, fir;
626 	unsigned int cnt_mk;
627 	uint32_t orig_flight, orig_tf;
628 	uint32_t tsnlast, tsnfirst;
629 	int recovery_cnt = 0;
630 
631 
632 	/* none in flight now */
633 	audit_tf = 0;
634 	fir = 0;
635 	/*
636 	 * figure out how long a data chunk must be pending before we can
637 	 * mark it ..
638 	 */
639 	(void)SCTP_GETTIME_TIMEVAL(&now);
640 	/* get cur rto in micro-seconds */
641 	cur_rtt = (((net->lastsa >> 2) + net->lastsv) >> 1);
642 	cur_rtt *= 1000;
643 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
644 		sctp_log_fr(cur_rtt,
645 		    stcb->asoc.peers_rwnd,
646 		    window_probe,
647 		    SCTP_FR_T3_MARK_TIME);
648 		sctp_log_fr(net->flight_size,
649 		    SCTP_OS_TIMER_PENDING(&net->fr_timer.timer),
650 		    SCTP_OS_TIMER_ACTIVE(&net->fr_timer.timer),
651 		    SCTP_FR_CWND_REPORT);
652 		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
653 	}
654 	tv.tv_sec = cur_rtt / 1000000;
655 	tv.tv_usec = cur_rtt % 1000000;
656 	min_wait = now;
657 	timevalsub(&min_wait, &tv);
658 	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
659 		/*
660 		 * if we hit here, we don't have enough seconds on the clock
661 		 * to account for the RTO. We just let the lower seconds be
662 		 * the bounds and don't worry about it. This may mean we
663 		 * will mark a lot more than we should.
664 		 */
665 		min_wait.tv_sec = min_wait.tv_usec = 0;
666 	}
667 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
668 		sctp_log_fr(cur_rtt, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
669 		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
670 	}
671 	/*
672 	 * Our rwnd will be incorrect here since we are not adding back the
673 	 * cnt * mbuf but we will fix that down below.
674 	 */
675 	orig_flight = net->flight_size;
676 	orig_tf = stcb->asoc.total_flight;
677 
678 	net->fast_retran_ip = 0;
679 	/* Now on to each chunk */
680 	num_mk = cnt_mk = 0;
681 	tsnfirst = tsnlast = 0;
682 start_again:
683 	chk = TAILQ_FIRST(&stcb->asoc.sent_queue);
684 	for (; chk != NULL; chk = tp2) {
685 		tp2 = TAILQ_NEXT(chk, sctp_next);
686 		if ((compare_with_wrap(stcb->asoc.last_acked_seq,
687 		    chk->rec.data.TSN_seq,
688 		    MAX_TSN)) ||
689 		    (stcb->asoc.last_acked_seq == chk->rec.data.TSN_seq)) {
690 			/* Strange case our list got out of order? */
691 			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x",
692 			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq);
693 			recovery_cnt++;
694 #ifdef INVARIANTS
695 			panic("last acked >= chk on sent-Q");
696 			/* to keep compiler happy */
697 			goto start_again;
698 #else
699 			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
700 			sctp_recover_sent_list(stcb);
701 			if (recovery_cnt < 10) {
702 				goto start_again;
703 			} else {
704 				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
705 			}
706 #endif
707 		}
708 		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
709 			/*
710 			 * found one to mark: If it is less than
711 			 * DATAGRAM_ACKED it MUST not be a skipped or marked
712 			 * TSN but instead one that is either already set
713 			 * for retransmission OR one that needs
714 			 * retransmission.
715 			 */
716 
717 			/* validate its been outstanding long enough */
718 			if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
719 				sctp_log_fr(chk->rec.data.TSN_seq,
720 				    chk->sent_rcv_time.tv_sec,
721 				    chk->sent_rcv_time.tv_usec,
722 				    SCTP_FR_T3_MARK_TIME);
723 			}
724 			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
725 				/*
726 				 * we have reached a chunk that was sent
727 				 * some seconds past our min.. forget it we
728 				 * will find no more to send.
729 				 */
730 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
731 					sctp_log_fr(0,
732 					    chk->sent_rcv_time.tv_sec,
733 					    chk->sent_rcv_time.tv_usec,
734 					    SCTP_FR_T3_STOPPED);
735 				}
736 				continue;
737 			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
738 			    (window_probe == 0)) {
739 				/*
740 				 * we must look at the micro seconds to
741 				 * know.
742 				 */
743 				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
744 					/*
745 					 * ok it was sent after our boundary
746 					 * time.
747 					 */
748 					if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
749 						sctp_log_fr(0,
750 						    chk->sent_rcv_time.tv_sec,
751 						    chk->sent_rcv_time.tv_usec,
752 						    SCTP_FR_T3_STOPPED);
753 					}
754 					continue;
755 				}
756 			}
757 			if (PR_SCTP_TTL_ENABLED(chk->flags)) {
758 				/* Is it expired? */
759 				if ((now.tv_sec > chk->rec.data.timetodrop.tv_sec) ||
760 				    ((chk->rec.data.timetodrop.tv_sec == now.tv_sec) &&
761 				    (now.tv_usec > chk->rec.data.timetodrop.tv_usec))) {
762 					/* Yes so drop it */
763 					if (chk->data) {
764 						(void)sctp_release_pr_sctp_chunk(stcb,
765 						    chk,
766 						    (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT),
767 						    &stcb->asoc.sent_queue, SCTP_SO_NOT_LOCKED);
768 					}
769 				}
770 				continue;
771 			}
772 			if (PR_SCTP_RTX_ENABLED(chk->flags)) {
773 				/* Has it been retransmitted tv_sec times? */
774 				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
775 					if (chk->data) {
776 						(void)sctp_release_pr_sctp_chunk(stcb,
777 						    chk,
778 						    (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT),
779 						    &stcb->asoc.sent_queue, SCTP_SO_NOT_LOCKED);
780 					}
781 				}
782 				continue;
783 			}
784 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
785 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
786 				num_mk++;
787 				if (fir == 0) {
788 					fir = 1;
789 					tsnfirst = chk->rec.data.TSN_seq;
790 				}
791 				tsnlast = chk->rec.data.TSN_seq;
792 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
793 					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
794 					    0, SCTP_FR_T3_MARKED);
795 				}
796 				if (chk->rec.data.chunk_was_revoked) {
797 					/* deflate the cwnd */
798 					chk->whoTo->cwnd -= chk->book_size;
799 					chk->rec.data.chunk_was_revoked = 0;
800 				}
801 				net->marked_retrans++;
802 				stcb->asoc.marked_retrans++;
803 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
804 					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
805 					    chk->whoTo->flight_size,
806 					    chk->book_size,
807 					    (uintptr_t) chk->whoTo,
808 					    chk->rec.data.TSN_seq);
809 				}
810 				sctp_flight_size_decrease(chk);
811 				sctp_total_flight_decrease(stcb, chk);
812 				stcb->asoc.peers_rwnd += chk->send_size;
813 				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
814 			}
815 			chk->sent = SCTP_DATAGRAM_RESEND;
816 			SCTP_STAT_INCR(sctps_markedretrans);
817 
818 			/* reset the TSN for striking and other FR stuff */
819 			chk->rec.data.doing_fast_retransmit = 0;
820 			/* Clear any time so NO RTT is being done */
821 			chk->do_rtt = 0;
822 			if (alt != net) {
823 				sctp_free_remote_addr(chk->whoTo);
824 				chk->no_fr_allowed = 1;
825 				chk->whoTo = alt;
826 				atomic_add_int(&alt->ref_count, 1);
827 			} else {
828 				chk->no_fr_allowed = 0;
829 				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
830 					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
831 				} else {
832 					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
833 				}
834 			}
835 			/*
836 			 * CMT: Do not allow FRs on retransmitted TSNs.
837 			 */
838 			if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 1) {
839 				chk->no_fr_allowed = 1;
840 			}
841 		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
842 			/* remember highest acked one */
843 			could_be_sent = chk;
844 		}
845 		if (chk->sent == SCTP_DATAGRAM_RESEND) {
846 			cnt_mk++;
847 		}
848 	}
849 	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
850 		/* we did not subtract the same things? */
851 		audit_tf = 1;
852 	}
853 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_EARLYFR_LOGGING_ENABLE | SCTP_FR_LOGGING_ENABLE)) {
854 		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
855 	}
856 #ifdef SCTP_DEBUG
857 	if (num_mk) {
858 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
859 		    tsnlast);
860 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%ld\n",
861 		    num_mk, (u_long)stcb->asoc.peers_rwnd);
862 		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
863 		    tsnlast);
864 		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n",
865 		    num_mk,
866 		    (int)stcb->asoc.peers_rwnd);
867 	}
868 #endif
869 	*num_marked = num_mk;
870 	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
871 		/* fix it so we retransmit the highest acked anyway */
872 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
873 		cnt_mk++;
874 		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
875 	}
876 	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
877 #ifdef INVARIANTS
878 		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
879 		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
880 #endif
881 #ifndef SCTP_AUDITING_ENABLED
882 		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
883 #endif
884 	}
885 	/* Now check for a ECN Echo that may be stranded */
886 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
887 		if ((chk->whoTo == net) &&
888 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
889 			sctp_free_remote_addr(chk->whoTo);
890 			chk->whoTo = alt;
891 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
892 				chk->sent = SCTP_DATAGRAM_RESEND;
893 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
894 			}
895 			atomic_add_int(&alt->ref_count, 1);
896 		}
897 	}
898 	if (audit_tf) {
899 		SCTPDBG(SCTP_DEBUG_TIMER4,
900 		    "Audit total flight due to negative value net:%p\n",
901 		    net);
902 		stcb->asoc.total_flight = 0;
903 		stcb->asoc.total_flight_count = 0;
904 		/* Clear all networks flight size */
905 		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
906 			lnets->flight_size = 0;
907 			SCTPDBG(SCTP_DEBUG_TIMER4,
908 			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
909 			    lnets, lnets->cwnd, lnets->ssthresh);
910 		}
911 		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
912 			if (chk->sent < SCTP_DATAGRAM_RESEND) {
913 				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
914 					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
915 					    chk->whoTo->flight_size,
916 					    chk->book_size,
917 					    (uintptr_t) chk->whoTo,
918 					    chk->rec.data.TSN_seq);
919 				}
920 				sctp_flight_size_increase(chk);
921 				sctp_total_flight_increase(stcb, chk);
922 			}
923 		}
924 	}
925 	/*
926 	 * Setup the ecn nonce re-sync point. We do this since
927 	 * retranmissions are NOT setup for ECN. This means that do to
928 	 * Karn's rule, we don't know the total of the peers ecn bits.
929 	 */
930 	chk = TAILQ_FIRST(&stcb->asoc.send_queue);
931 	if (chk == NULL) {
932 		stcb->asoc.nonce_resync_tsn = stcb->asoc.sending_seq;
933 	} else {
934 		stcb->asoc.nonce_resync_tsn = chk->rec.data.TSN_seq;
935 	}
936 	stcb->asoc.nonce_wait_for_ecne = 0;
937 	stcb->asoc.nonce_sum_check = 0;
938 	/* We return 1 if we only have a window probe outstanding */
939 	return (0);
940 }
941 
942 static void
943 sctp_move_all_chunks_to_alt(struct sctp_tcb *stcb,
944     struct sctp_nets *net,
945     struct sctp_nets *alt)
946 {
947 	struct sctp_association *asoc;
948 	struct sctp_stream_out *outs;
949 	struct sctp_tmit_chunk *chk;
950 	struct sctp_stream_queue_pending *sp;
951 
952 	if (net == alt)
953 		/* nothing to do */
954 		return;
955 
956 	asoc = &stcb->asoc;
957 
958 	/*
959 	 * now through all the streams checking for chunks sent to our bad
960 	 * network.
961 	 */
962 	TAILQ_FOREACH(outs, &asoc->out_wheel, next_spoke) {
963 		/* now clean up any chunks here */
964 		TAILQ_FOREACH(sp, &outs->outqueue, next) {
965 			if (sp->net == net) {
966 				sctp_free_remote_addr(sp->net);
967 				sp->net = alt;
968 				atomic_add_int(&alt->ref_count, 1);
969 			}
970 		}
971 	}
972 	/* Now check the pending queue */
973 	TAILQ_FOREACH(chk, &asoc->send_queue, sctp_next) {
974 		if (chk->whoTo == net) {
975 			sctp_free_remote_addr(chk->whoTo);
976 			chk->whoTo = alt;
977 			atomic_add_int(&alt->ref_count, 1);
978 		}
979 	}
980 
981 }
982 
983 int
984 sctp_t3rxt_timer(struct sctp_inpcb *inp,
985     struct sctp_tcb *stcb,
986     struct sctp_nets *net)
987 {
988 	struct sctp_nets *alt;
989 	int win_probe, num_mk;
990 
991 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
992 		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
993 	}
994 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
995 		struct sctp_nets *lnet;
996 
997 		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
998 			if (net == lnet) {
999 				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
1000 			} else {
1001 				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
1002 			}
1003 		}
1004 	}
1005 	/* Find an alternate and mark those for retransmission */
1006 	if ((stcb->asoc.peers_rwnd == 0) &&
1007 	    (stcb->asoc.total_flight < net->mtu)) {
1008 		SCTP_STAT_INCR(sctps_timowindowprobe);
1009 		win_probe = 1;
1010 	} else {
1011 		win_probe = 0;
1012 	}
1013 
1014 	/*
1015 	 * JRS 5/14/07 - If CMT PF is on and the destination if not already
1016 	 * in PF state, set the destination to PF state and store the
1017 	 * current time as the time that the destination was last active. In
1018 	 * addition, find an alternate destination with PF-based
1019 	 * find_alt_net().
1020 	 */
1021 	if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf)) {
1022 		if ((net->dest_state & SCTP_ADDR_PF) != SCTP_ADDR_PF) {
1023 			net->dest_state |= SCTP_ADDR_PF;
1024 			net->last_active = sctp_get_tick_count();
1025 			SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from active to PF.\n",
1026 			    net);
1027 		}
1028 		alt = sctp_find_alternate_net(stcb, net, 2);
1029 	} else if (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) {
1030 		/*
1031 		 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is being
1032 		 * used, then pick dest with largest ssthresh for any
1033 		 * retransmission.
1034 		 */
1035 		alt = net;
1036 		alt = sctp_find_alternate_net(stcb, alt, 1);
1037 		/*
1038 		 * CUCv2: If a different dest is picked for the
1039 		 * retransmission, then new (rtx-)pseudo_cumack needs to be
1040 		 * tracked for orig dest. Let CUCv2 track new (rtx-)
1041 		 * pseudo-cumack always.
1042 		 */
1043 		net->find_pseudo_cumack = 1;
1044 		net->find_rtx_pseudo_cumack = 1;
1045 	} else {		/* CMT is OFF */
1046 		alt = sctp_find_alternate_net(stcb, net, 0);
1047 	}
1048 
1049 	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe, &num_mk);
1050 	/* FR Loss recovery just ended with the T3. */
1051 	stcb->asoc.fast_retran_loss_recovery = 0;
1052 
1053 	/* CMT FR loss recovery ended with the T3 */
1054 	net->fast_retran_loss_recovery = 0;
1055 
1056 	/*
1057 	 * setup the sat loss recovery that prevents satellite cwnd advance.
1058 	 */
1059 	stcb->asoc.sat_t3_loss_recovery = 1;
1060 	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
1061 
1062 	/* Backoff the timer and cwnd */
1063 	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk);
1064 	if (win_probe == 0) {
1065 		/* We don't do normal threshold management on window probes */
1066 		if (sctp_threshold_management(inp, stcb, net,
1067 		    stcb->asoc.max_send_times)) {
1068 			/* Association was destroyed */
1069 			return (1);
1070 		} else {
1071 			if (net != stcb->asoc.primary_destination) {
1072 				/* send a immediate HB if our RTO is stale */
1073 				struct timeval now;
1074 				unsigned int ms_goneby;
1075 
1076 				(void)SCTP_GETTIME_TIMEVAL(&now);
1077 				if (net->last_sent_time.tv_sec) {
1078 					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
1079 				} else {
1080 					ms_goneby = 0;
1081 				}
1082 				if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
1083 					/*
1084 					 * no recent feed back in an RTO or
1085 					 * more, request a RTT update
1086 					 */
1087 					if (sctp_send_hb(stcb, 1, net) < 0)
1088 						return 1;
1089 				}
1090 			}
1091 		}
1092 	} else {
1093 		/*
1094 		 * For a window probe we don't penalize the net's but only
1095 		 * the association. This may fail it if SACKs are not coming
1096 		 * back. If sack's are coming with rwnd locked at 0, we will
1097 		 * continue to hold things waiting for rwnd to raise
1098 		 */
1099 		if (sctp_threshold_management(inp, stcb, NULL,
1100 		    stcb->asoc.max_send_times)) {
1101 			/* Association was destroyed */
1102 			return (1);
1103 		}
1104 	}
1105 	if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) {
1106 		/* Move all pending over too */
1107 		sctp_move_all_chunks_to_alt(stcb, net, alt);
1108 
1109 		/*
1110 		 * Get the address that failed, to force a new src address
1111 		 * selecton and a route allocation.
1112 		 */
1113 		if (net->ro._s_addr) {
1114 			sctp_free_ifa(net->ro._s_addr);
1115 			net->ro._s_addr = NULL;
1116 		}
1117 		net->src_addr_selected = 0;
1118 
1119 		/* Force a route allocation too */
1120 		if (net->ro.ro_rt) {
1121 			RTFREE(net->ro.ro_rt);
1122 			net->ro.ro_rt = NULL;
1123 		}
1124 		/* Was it our primary? */
1125 		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
1126 			/*
1127 			 * Yes, note it as such and find an alternate note:
1128 			 * this means HB code must use this to resent the
1129 			 * primary if it goes active AND if someone does a
1130 			 * change-primary then this flag must be cleared
1131 			 * from any net structures.
1132 			 */
1133 			if (sctp_set_primary_addr(stcb,
1134 			    (struct sockaddr *)NULL,
1135 			    alt) == 0) {
1136 				net->dest_state |= SCTP_ADDR_WAS_PRIMARY;
1137 			}
1138 		}
1139 	} else if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && SCTP_BASE_SYSCTL(sctp_cmt_pf) && (net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF) {
1140 		/*
1141 		 * JRS 5/14/07 - If the destination hasn't failed completely
1142 		 * but is in PF state, a PF-heartbeat needs to be sent
1143 		 * manually.
1144 		 */
1145 		if (sctp_send_hb(stcb, 1, net) < 0)
1146 			return 1;
1147 	}
1148 	/*
1149 	 * Special case for cookie-echo'ed case, we don't do output but must
1150 	 * await the COOKIE-ACK before retransmission
1151 	 */
1152 	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1153 		/*
1154 		 * Here we just reset the timer and start again since we
1155 		 * have not established the asoc
1156 		 */
1157 		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
1158 		return (0);
1159 	}
1160 	if (stcb->asoc.peer_supports_prsctp) {
1161 		struct sctp_tmit_chunk *lchk;
1162 
1163 		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
1164 		/* C3. See if we need to send a Fwd-TSN */
1165 		if (compare_with_wrap(stcb->asoc.advanced_peer_ack_point,
1166 		    stcb->asoc.last_acked_seq, MAX_TSN)) {
1167 			/*
1168 			 * ISSUE with ECN, see FWD-TSN processing for notes
1169 			 * on issues that will occur when the ECN NONCE
1170 			 * stuff is put into SCTP for cross checking.
1171 			 */
1172 			send_forward_tsn(stcb, &stcb->asoc);
1173 			if (lchk) {
1174 				/* Assure a timer is up */
1175 				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
1176 			}
1177 		}
1178 	}
1179 	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
1180 		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
1181 	}
1182 	return (0);
1183 }
1184 
1185 int
1186 sctp_t1init_timer(struct sctp_inpcb *inp,
1187     struct sctp_tcb *stcb,
1188     struct sctp_nets *net)
1189 {
1190 	/* bump the thresholds */
1191 	if (stcb->asoc.delayed_connection) {
1192 		/*
1193 		 * special hook for delayed connection. The library did NOT
1194 		 * complete the rest of its sends.
1195 		 */
1196 		stcb->asoc.delayed_connection = 0;
1197 		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1198 		return (0);
1199 	}
1200 	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
1201 		return (0);
1202 	}
1203 	if (sctp_threshold_management(inp, stcb, net,
1204 	    stcb->asoc.max_init_times)) {
1205 		/* Association was destroyed */
1206 		return (1);
1207 	}
1208 	stcb->asoc.dropped_special_cnt = 0;
1209 	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0);
1210 	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1211 		net->RTO = stcb->asoc.initial_init_rto_max;
1212 	}
1213 	if (stcb->asoc.numnets > 1) {
1214 		/* If we have more than one addr use it */
1215 		struct sctp_nets *alt;
1216 
1217 		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1218 		if ((alt != NULL) && (alt != stcb->asoc.primary_destination)) {
1219 			sctp_move_all_chunks_to_alt(stcb, stcb->asoc.primary_destination, alt);
1220 			stcb->asoc.primary_destination = alt;
1221 		}
1222 	}
1223 	/* Send out a new init */
1224 	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1225 	return (0);
1226 }
1227 
1228 /*
1229  * For cookie and asconf we actually need to find and mark for resend, then
1230  * increment the resend counter (after all the threshold management stuff of
1231  * course).
1232  */
1233 int
1234 sctp_cookie_timer(struct sctp_inpcb *inp,
1235     struct sctp_tcb *stcb,
1236     struct sctp_nets *net)
1237 {
1238 	struct sctp_nets *alt;
1239 	struct sctp_tmit_chunk *cookie;
1240 
1241 	/* first before all else we must find the cookie */
1242 	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1243 		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1244 			break;
1245 		}
1246 	}
1247 	if (cookie == NULL) {
1248 		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1249 			/* FOOBAR! */
1250 			struct mbuf *oper;
1251 
1252 			oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
1253 			    0, M_DONTWAIT, 1, MT_DATA);
1254 			if (oper) {
1255 				struct sctp_paramhdr *ph;
1256 				uint32_t *ippp;
1257 
1258 				SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
1259 				    sizeof(uint32_t);
1260 				ph = mtod(oper, struct sctp_paramhdr *);
1261 				ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1262 				ph->param_length = htons(SCTP_BUF_LEN(oper));
1263 				ippp = (uint32_t *) (ph + 1);
1264 				*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
1265 			}
1266 			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_4;
1267 			sctp_abort_an_association(inp, stcb, SCTP_INTERNAL_ERROR,
1268 			    oper, SCTP_SO_NOT_LOCKED);
1269 		} else {
1270 #ifdef INVARIANTS
1271 			panic("Cookie timer expires in wrong state?");
1272 #else
1273 			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1274 			return (0);
1275 #endif
1276 		}
1277 		return (0);
1278 	}
1279 	/* Ok we found the cookie, threshold management next */
1280 	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1281 	    stcb->asoc.max_init_times)) {
1282 		/* Assoc is over */
1283 		return (1);
1284 	}
1285 	/*
1286 	 * cleared theshold management now lets backoff the address & select
1287 	 * an alternate
1288 	 */
1289 	stcb->asoc.dropped_special_cnt = 0;
1290 	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0);
1291 	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1292 	if (alt != cookie->whoTo) {
1293 		sctp_free_remote_addr(cookie->whoTo);
1294 		cookie->whoTo = alt;
1295 		atomic_add_int(&alt->ref_count, 1);
1296 	}
1297 	/* Now mark the retran info */
1298 	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1299 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1300 	}
1301 	cookie->sent = SCTP_DATAGRAM_RESEND;
1302 	/*
1303 	 * Now call the output routine to kick out the cookie again, Note we
1304 	 * don't mark any chunks for retran so that FR will need to kick in
1305 	 * to move these (or a send timer).
1306 	 */
1307 	return (0);
1308 }
1309 
1310 int
1311 sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1312     struct sctp_nets *net)
1313 {
1314 	struct sctp_nets *alt;
1315 	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1316 
1317 	if (stcb->asoc.stream_reset_outstanding == 0) {
1318 		return (0);
1319 	}
1320 	/* find the existing STRRESET, we use the seq number we sent out on */
1321 	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1322 	if (strrst == NULL) {
1323 		return (0);
1324 	}
1325 	/* do threshold management */
1326 	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1327 	    stcb->asoc.max_send_times)) {
1328 		/* Assoc is over */
1329 		return (1);
1330 	}
1331 	/*
1332 	 * cleared theshold management now lets backoff the address & select
1333 	 * an alternate
1334 	 */
1335 	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0);
1336 	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1337 	sctp_free_remote_addr(strrst->whoTo);
1338 	strrst->whoTo = alt;
1339 	atomic_add_int(&alt->ref_count, 1);
1340 
1341 	/* See if a ECN Echo is also stranded */
1342 	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1343 		if ((chk->whoTo == net) &&
1344 		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1345 			sctp_free_remote_addr(chk->whoTo);
1346 			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1347 				chk->sent = SCTP_DATAGRAM_RESEND;
1348 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1349 			}
1350 			chk->whoTo = alt;
1351 			atomic_add_int(&alt->ref_count, 1);
1352 		}
1353 	}
1354 	if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) {
1355 		/*
1356 		 * If the address went un-reachable, we need to move to
1357 		 * alternates for ALL chk's in queue
1358 		 */
1359 		sctp_move_all_chunks_to_alt(stcb, net, alt);
1360 	}
1361 	/* mark the retran info */
1362 	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1363 		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1364 	strrst->sent = SCTP_DATAGRAM_RESEND;
1365 
1366 	/* restart the timer */
1367 	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1368 	return (0);
1369 }
1370 
1371 int
1372 sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1373     struct sctp_nets *net)
1374 {
1375 	struct sctp_nets *alt;
1376 	struct sctp_tmit_chunk *asconf, *chk, *nchk;
1377 
1378 	/* is this a first send, or a retransmission? */
1379 	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1380 		/* compose a new ASCONF chunk and send it */
1381 		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1382 	} else {
1383 		/*
1384 		 * Retransmission of the existing ASCONF is needed
1385 		 */
1386 
1387 		/* find the existing ASCONF */
1388 		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1389 		if (asconf == NULL) {
1390 			return (0);
1391 		}
1392 		/* do threshold management */
1393 		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1394 		    stcb->asoc.max_send_times)) {
1395 			/* Assoc is over */
1396 			return (1);
1397 		}
1398 		if (asconf->snd_count > stcb->asoc.max_send_times) {
1399 			/*
1400 			 * Something is rotten: our peer is not responding
1401 			 * to ASCONFs but apparently is to other chunks.
1402 			 * i.e. it is not properly handling the chunk type
1403 			 * upper bits. Mark this peer as ASCONF incapable
1404 			 * and cleanup.
1405 			 */
1406 			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1407 			sctp_asconf_cleanup(stcb, net);
1408 			return (0);
1409 		}
1410 		/*
1411 		 * cleared threshold management, so now backoff the net and
1412 		 * select an alternate
1413 		 */
1414 		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0);
1415 		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1416 		if (asconf->whoTo != alt) {
1417 			sctp_free_remote_addr(asconf->whoTo);
1418 			asconf->whoTo = alt;
1419 			atomic_add_int(&alt->ref_count, 1);
1420 		}
1421 		/* See if an ECN Echo is also stranded */
1422 		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1423 			if ((chk->whoTo == net) &&
1424 			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1425 				sctp_free_remote_addr(chk->whoTo);
1426 				chk->whoTo = alt;
1427 				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1428 					chk->sent = SCTP_DATAGRAM_RESEND;
1429 					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1430 				}
1431 				atomic_add_int(&alt->ref_count, 1);
1432 			}
1433 		}
1434 		for (chk = asconf; chk; chk = nchk) {
1435 			nchk = TAILQ_NEXT(chk, sctp_next);
1436 			if (chk->whoTo != alt) {
1437 				sctp_free_remote_addr(chk->whoTo);
1438 				chk->whoTo = alt;
1439 				atomic_add_int(&alt->ref_count, 1);
1440 			}
1441 			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1442 				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1443 			chk->sent = SCTP_DATAGRAM_RESEND;
1444 		}
1445 		if (net->dest_state & SCTP_ADDR_NOT_REACHABLE) {
1446 			/*
1447 			 * If the address went un-reachable, we need to move
1448 			 * to the alternate for ALL chunks in queue
1449 			 */
1450 			sctp_move_all_chunks_to_alt(stcb, net, alt);
1451 			net = alt;
1452 		}
1453 		/* mark the retran info */
1454 		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1455 			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1456 		asconf->sent = SCTP_DATAGRAM_RESEND;
1457 
1458 		/* send another ASCONF if any and we can do */
1459 		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1460 	}
1461 	return (0);
1462 }
1463 
1464 /* Mobility adaptation */
1465 void
1466 sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1467     struct sctp_nets *net)
1468 {
1469 	if (stcb->asoc.deleted_primary == NULL) {
1470 		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1471 		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1472 		return;
1473 	}
1474 	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1475 	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1476 	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1477 	stcb->asoc.deleted_primary = NULL;
1478 	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1479 	return;
1480 }
1481 
1482 /*
1483  * For the shutdown and shutdown-ack, we do not keep one around on the
1484  * control queue. This means we must generate a new one and call the general
1485  * chunk output routine, AFTER having done threshold management.
1486  */
1487 int
1488 sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1489     struct sctp_nets *net)
1490 {
1491 	struct sctp_nets *alt;
1492 
1493 	/* first threshold managment */
1494 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1495 		/* Assoc is over */
1496 		return (1);
1497 	}
1498 	/* second select an alternative */
1499 	alt = sctp_find_alternate_net(stcb, net, 0);
1500 
1501 	/* third generate a shutdown into the queue for out net */
1502 	if (alt) {
1503 		sctp_send_shutdown(stcb, alt);
1504 	} else {
1505 		/*
1506 		 * if alt is NULL, there is no dest to send to??
1507 		 */
1508 		return (0);
1509 	}
1510 	/* fourth restart timer */
1511 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1512 	return (0);
1513 }
1514 
1515 int
1516 sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1517     struct sctp_nets *net)
1518 {
1519 	struct sctp_nets *alt;
1520 
1521 	/* first threshold managment */
1522 	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1523 		/* Assoc is over */
1524 		return (1);
1525 	}
1526 	/* second select an alternative */
1527 	alt = sctp_find_alternate_net(stcb, net, 0);
1528 
1529 	/* third generate a shutdown into the queue for out net */
1530 	sctp_send_shutdown_ack(stcb, alt);
1531 
1532 	/* fourth restart timer */
1533 	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1534 	return (0);
1535 }
1536 
1537 static void
1538 sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1539     struct sctp_tcb *stcb)
1540 {
1541 	struct sctp_stream_out *outs;
1542 	struct sctp_stream_queue_pending *sp;
1543 	unsigned int chks_in_queue = 0;
1544 	int being_filled = 0;
1545 
1546 	/*
1547 	 * This function is ONLY called when the send/sent queues are empty.
1548 	 */
1549 	if ((stcb == NULL) || (inp == NULL))
1550 		return;
1551 
1552 	if (stcb->asoc.sent_queue_retran_cnt) {
1553 		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1554 		    stcb->asoc.sent_queue_retran_cnt);
1555 		stcb->asoc.sent_queue_retran_cnt = 0;
1556 	}
1557 	SCTP_TCB_SEND_LOCK(stcb);
1558 	if (TAILQ_EMPTY(&stcb->asoc.out_wheel)) {
1559 		int i, cnt = 0;
1560 
1561 		/* Check to see if a spoke fell off the wheel */
1562 		for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1563 			if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1564 				sctp_insert_on_wheel(stcb, &stcb->asoc, &stcb->asoc.strmout[i], 1);
1565 				cnt++;
1566 			}
1567 		}
1568 		if (cnt) {
1569 			/* yep, we lost a spoke or two */
1570 			SCTP_PRINTF("Found an additional %d streams NOT on outwheel, corrected\n", cnt);
1571 		} else {
1572 			/* no spokes lost, */
1573 			stcb->asoc.total_output_queue_size = 0;
1574 		}
1575 		SCTP_TCB_SEND_UNLOCK(stcb);
1576 		return;
1577 	}
1578 	SCTP_TCB_SEND_UNLOCK(stcb);
1579 	/* Check to see if some data queued, if so report it */
1580 	TAILQ_FOREACH(outs, &stcb->asoc.out_wheel, next_spoke) {
1581 		if (!TAILQ_EMPTY(&outs->outqueue)) {
1582 			TAILQ_FOREACH(sp, &outs->outqueue, next) {
1583 				if (sp->msg_is_complete)
1584 					being_filled++;
1585 				chks_in_queue++;
1586 			}
1587 		}
1588 	}
1589 	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1590 		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1591 		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1592 	}
1593 	if (chks_in_queue) {
1594 		/* call the output queue function */
1595 		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1596 		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1597 		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1598 			/*
1599 			 * Probably should go in and make it go back through
1600 			 * and add fragments allowed
1601 			 */
1602 			if (being_filled == 0) {
1603 				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1604 				    chks_in_queue);
1605 			}
1606 		}
1607 	} else {
1608 		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1609 		    (u_long)stcb->asoc.total_output_queue_size);
1610 		stcb->asoc.total_output_queue_size = 0;
1611 	}
1612 }
1613 
1614 int
1615 sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1616     struct sctp_nets *net, int cnt_of_unconf)
1617 {
1618 	int ret;
1619 
1620 	if (net) {
1621 		if (net->hb_responded == 0) {
1622 			if (net->ro._s_addr) {
1623 				/*
1624 				 * Invalidate the src address if we did not
1625 				 * get a response last time.
1626 				 */
1627 				sctp_free_ifa(net->ro._s_addr);
1628 				net->ro._s_addr = NULL;
1629 				net->src_addr_selected = 0;
1630 			}
1631 			sctp_backoff_on_timeout(stcb, net, 1, 0);
1632 		}
1633 		/* Zero PBA, if it needs it */
1634 		if (net->partial_bytes_acked) {
1635 			net->partial_bytes_acked = 0;
1636 		}
1637 	}
1638 	if ((stcb->asoc.total_output_queue_size > 0) &&
1639 	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1640 	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1641 		sctp_audit_stream_queues_for_size(inp, stcb);
1642 	}
1643 	/* Send a new HB, this will do threshold managment, pick a new dest */
1644 	if (cnt_of_unconf == 0) {
1645 		if (sctp_send_hb(stcb, 0, NULL) < 0) {
1646 			return (1);
1647 		}
1648 	} else {
1649 		/*
1650 		 * this will send out extra hb's up to maxburst if there are
1651 		 * any unconfirmed addresses.
1652 		 */
1653 		uint32_t cnt_sent = 0;
1654 
1655 		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1656 			if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) &&
1657 			    (net->dest_state & SCTP_ADDR_REACHABLE)) {
1658 				cnt_sent++;
1659 				if (net->hb_responded == 0) {
1660 					/* Did we respond last time? */
1661 					if (net->ro._s_addr) {
1662 						sctp_free_ifa(net->ro._s_addr);
1663 						net->ro._s_addr = NULL;
1664 						net->src_addr_selected = 0;
1665 					}
1666 				}
1667 				ret = sctp_send_hb(stcb, 1, net);
1668 				if (ret < 0)
1669 					return 1;
1670 				else if (ret == 0) {
1671 					break;
1672 				}
1673 				if (cnt_sent >= SCTP_BASE_SYSCTL(sctp_hb_maxburst))
1674 					break;
1675 			}
1676 		}
1677 	}
1678 	return (0);
1679 }
1680 
1681 int
1682 sctp_is_hb_timer_running(struct sctp_tcb *stcb)
1683 {
1684 	if (SCTP_OS_TIMER_PENDING(&stcb->asoc.hb_timer.timer)) {
1685 		/* its running */
1686 		return (1);
1687 	} else {
1688 		/* nope */
1689 		return (0);
1690 	}
1691 }
1692 
1693 int
1694 sctp_is_sack_timer_running(struct sctp_tcb *stcb)
1695 {
1696 	if (SCTP_OS_TIMER_PENDING(&stcb->asoc.dack_timer.timer)) {
1697 		/* its running */
1698 		return (1);
1699 	} else {
1700 		/* nope */
1701 		return (0);
1702 	}
1703 }
1704 
1705 #define SCTP_NUMBER_OF_MTU_SIZES 18
1706 static uint32_t mtu_sizes[] = {
1707 	68,
1708 	296,
1709 	508,
1710 	512,
1711 	544,
1712 	576,
1713 	1006,
1714 	1492,
1715 	1500,
1716 	1536,
1717 	2002,
1718 	2048,
1719 	4352,
1720 	4464,
1721 	8166,
1722 	17914,
1723 	32000,
1724 	65535
1725 };
1726 
1727 
1728 static uint32_t
1729 sctp_getnext_mtu(struct sctp_inpcb *inp, uint32_t cur_mtu)
1730 {
1731 	/* select another MTU that is just bigger than this one */
1732 	int i;
1733 
1734 	for (i = 0; i < SCTP_NUMBER_OF_MTU_SIZES; i++) {
1735 		if (cur_mtu < mtu_sizes[i]) {
1736 			/* no max_mtu is bigger than this one */
1737 			return (mtu_sizes[i]);
1738 		}
1739 	}
1740 	/* here return the highest allowable */
1741 	return (cur_mtu);
1742 }
1743 
1744 
1745 void
1746 sctp_pathmtu_timer(struct sctp_inpcb *inp,
1747     struct sctp_tcb *stcb,
1748     struct sctp_nets *net)
1749 {
1750 	uint32_t next_mtu, mtu;
1751 
1752 	next_mtu = sctp_getnext_mtu(inp, net->mtu);
1753 
1754 	if ((next_mtu > net->mtu) && (net->port == 0)) {
1755 		if ((net->src_addr_selected == 0) ||
1756 		    (net->ro._s_addr == NULL) ||
1757 		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1758 			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1759 				sctp_free_ifa(net->ro._s_addr);
1760 				net->ro._s_addr = NULL;
1761 				net->src_addr_selected = 0;
1762 			} else if (net->ro._s_addr == NULL) {
1763 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1764 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1765 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1766 
1767 					/* KAME hack: embed scopeid */
1768 					(void)sa6_embedscope(sin6, MODULE_GLOBAL(MOD_INET6, ip6_use_defzone));
1769 				}
1770 #endif
1771 
1772 				net->ro._s_addr = sctp_source_address_selection(inp,
1773 				    stcb,
1774 				    (sctp_route_t *) & net->ro,
1775 				    net, 0, stcb->asoc.vrf_id);
1776 #if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1777 				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1778 					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1779 
1780 					(void)sa6_recoverscope(sin6);
1781 				}
1782 #endif				/* INET6 */
1783 			}
1784 			if (net->ro._s_addr)
1785 				net->src_addr_selected = 1;
1786 		}
1787 		if (net->ro._s_addr) {
1788 			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1789 			if (mtu > next_mtu) {
1790 				net->mtu = next_mtu;
1791 			}
1792 		}
1793 	}
1794 	/* restart the timer */
1795 	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1796 }
1797 
1798 void
1799 sctp_autoclose_timer(struct sctp_inpcb *inp,
1800     struct sctp_tcb *stcb,
1801     struct sctp_nets *net)
1802 {
1803 	struct timeval tn, *tim_touse;
1804 	struct sctp_association *asoc;
1805 	int ticks_gone_by;
1806 
1807 	(void)SCTP_GETTIME_TIMEVAL(&tn);
1808 	if (stcb->asoc.sctp_autoclose_ticks &&
1809 	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1810 		/* Auto close is on */
1811 		asoc = &stcb->asoc;
1812 		/* pick the time to use */
1813 		if (asoc->time_last_rcvd.tv_sec >
1814 		    asoc->time_last_sent.tv_sec) {
1815 			tim_touse = &asoc->time_last_rcvd;
1816 		} else {
1817 			tim_touse = &asoc->time_last_sent;
1818 		}
1819 		/* Now has long enough transpired to autoclose? */
1820 		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1821 		if ((ticks_gone_by > 0) &&
1822 		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1823 			/*
1824 			 * autoclose time has hit, call the output routine,
1825 			 * which should do nothing just to be SURE we don't
1826 			 * have hanging data. We can then safely check the
1827 			 * queues and know that we are clear to send
1828 			 * shutdown
1829 			 */
1830 			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1831 			/* Are we clean? */
1832 			if (TAILQ_EMPTY(&asoc->send_queue) &&
1833 			    TAILQ_EMPTY(&asoc->sent_queue)) {
1834 				/*
1835 				 * there is nothing queued to send, so I'm
1836 				 * done...
1837 				 */
1838 				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1839 					/* only send SHUTDOWN 1st time thru */
1840 					sctp_send_shutdown(stcb, stcb->asoc.primary_destination);
1841 					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1842 					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1843 						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1844 					}
1845 					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1846 					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1847 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1848 					    stcb->sctp_ep, stcb,
1849 					    asoc->primary_destination);
1850 					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1851 					    stcb->sctp_ep, stcb,
1852 					    asoc->primary_destination);
1853 				}
1854 			}
1855 		} else {
1856 			/*
1857 			 * No auto close at this time, reset t-o to check
1858 			 * later
1859 			 */
1860 			int tmp;
1861 
1862 			/* fool the timer startup to use the time left */
1863 			tmp = asoc->sctp_autoclose_ticks;
1864 			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1865 			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1866 			    net);
1867 			/* restore the real tick value */
1868 			asoc->sctp_autoclose_ticks = tmp;
1869 		}
1870 	}
1871 }
1872 
1873 void
1874 sctp_iterator_timer(struct sctp_iterator *it)
1875 {
1876 	int iteration_count = 0;
1877 	int inp_skip = 0;
1878 
1879 	/*
1880 	 * only one iterator can run at a time. This is the only way we can
1881 	 * cleanly pull ep's from underneath all the running interators when
1882 	 * a ep is freed.
1883 	 */
1884 	SCTP_ITERATOR_LOCK();
1885 	if (it->inp == NULL) {
1886 		/* iterator is complete */
1887 done_with_iterator:
1888 		SCTP_ITERATOR_UNLOCK();
1889 		SCTP_INP_INFO_WLOCK();
1890 		TAILQ_REMOVE(&SCTP_BASE_INFO(iteratorhead), it, sctp_nxt_itr);
1891 		/* stopping the callout is not needed, in theory */
1892 		SCTP_INP_INFO_WUNLOCK();
1893 		(void)SCTP_OS_TIMER_STOP(&it->tmr.timer);
1894 		if (it->function_atend != NULL) {
1895 			(*it->function_atend) (it->pointer, it->val);
1896 		}
1897 		SCTP_FREE(it, SCTP_M_ITER);
1898 		return;
1899 	}
1900 select_a_new_ep:
1901 	SCTP_INP_WLOCK(it->inp);
1902 	while (((it->pcb_flags) &&
1903 	    ((it->inp->sctp_flags & it->pcb_flags) != it->pcb_flags)) ||
1904 	    ((it->pcb_features) &&
1905 	    ((it->inp->sctp_features & it->pcb_features) != it->pcb_features))) {
1906 		/* endpoint flags or features don't match, so keep looking */
1907 		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
1908 			SCTP_INP_WUNLOCK(it->inp);
1909 			goto done_with_iterator;
1910 		}
1911 		SCTP_INP_WUNLOCK(it->inp);
1912 		it->inp = LIST_NEXT(it->inp, sctp_list);
1913 		if (it->inp == NULL) {
1914 			goto done_with_iterator;
1915 		}
1916 		SCTP_INP_WLOCK(it->inp);
1917 	}
1918 	if ((it->inp->inp_starting_point_for_iterator != NULL) &&
1919 	    (it->inp->inp_starting_point_for_iterator != it)) {
1920 		SCTP_PRINTF("Iterator collision, waiting for one at %p\n",
1921 		    it->inp);
1922 		SCTP_INP_WUNLOCK(it->inp);
1923 		goto start_timer_return;
1924 	}
1925 	/* mark the current iterator on the endpoint */
1926 	it->inp->inp_starting_point_for_iterator = it;
1927 	SCTP_INP_WUNLOCK(it->inp);
1928 	SCTP_INP_RLOCK(it->inp);
1929 	/* now go through each assoc which is in the desired state */
1930 	if (it->done_current_ep == 0) {
1931 		if (it->function_inp != NULL)
1932 			inp_skip = (*it->function_inp) (it->inp, it->pointer, it->val);
1933 		it->done_current_ep = 1;
1934 	}
1935 	if (it->stcb == NULL) {
1936 		/* run the per instance function */
1937 		it->stcb = LIST_FIRST(&it->inp->sctp_asoc_list);
1938 	}
1939 	SCTP_INP_RUNLOCK(it->inp);
1940 	if ((inp_skip) || it->stcb == NULL) {
1941 		if (it->function_inp_end != NULL) {
1942 			inp_skip = (*it->function_inp_end) (it->inp,
1943 			    it->pointer,
1944 			    it->val);
1945 		}
1946 		goto no_stcb;
1947 	}
1948 	if ((it->stcb) &&
1949 	    (it->stcb->asoc.stcb_starting_point_for_iterator == it)) {
1950 		it->stcb->asoc.stcb_starting_point_for_iterator = NULL;
1951 	}
1952 	while (it->stcb) {
1953 		SCTP_TCB_LOCK(it->stcb);
1954 		if (it->asoc_state && ((it->stcb->asoc.state & it->asoc_state) != it->asoc_state)) {
1955 			/* not in the right state... keep looking */
1956 			SCTP_TCB_UNLOCK(it->stcb);
1957 			goto next_assoc;
1958 		}
1959 		/* mark the current iterator on the assoc */
1960 		it->stcb->asoc.stcb_starting_point_for_iterator = it;
1961 		/* see if we have limited out the iterator loop */
1962 		iteration_count++;
1963 		if (iteration_count > SCTP_ITERATOR_MAX_AT_ONCE) {
1964 	start_timer_return:
1965 			/* set a timer to continue this later */
1966 			if (it->stcb)
1967 				SCTP_TCB_UNLOCK(it->stcb);
1968 			sctp_timer_start(SCTP_TIMER_TYPE_ITERATOR,
1969 			    (struct sctp_inpcb *)it, NULL, NULL);
1970 			SCTP_ITERATOR_UNLOCK();
1971 			return;
1972 		}
1973 		/* run function on this one */
1974 		(*it->function_assoc) (it->inp, it->stcb, it->pointer, it->val);
1975 
1976 		/*
1977 		 * we lie here, it really needs to have its own type but
1978 		 * first I must verify that this won't effect things :-0
1979 		 */
1980 		if (it->no_chunk_output == 0)
1981 			sctp_chunk_output(it->inp, it->stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1982 
1983 		SCTP_TCB_UNLOCK(it->stcb);
1984 next_assoc:
1985 		it->stcb = LIST_NEXT(it->stcb, sctp_tcblist);
1986 		if (it->stcb == NULL) {
1987 			if (it->function_inp_end != NULL) {
1988 				inp_skip = (*it->function_inp_end) (it->inp,
1989 				    it->pointer,
1990 				    it->val);
1991 			}
1992 		}
1993 	}
1994 no_stcb:
1995 	/* done with all assocs on this endpoint, move on to next endpoint */
1996 	it->done_current_ep = 0;
1997 	SCTP_INP_WLOCK(it->inp);
1998 	it->inp->inp_starting_point_for_iterator = NULL;
1999 	SCTP_INP_WUNLOCK(it->inp);
2000 	if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
2001 		it->inp = NULL;
2002 	} else {
2003 		SCTP_INP_INFO_RLOCK();
2004 		it->inp = LIST_NEXT(it->inp, sctp_list);
2005 		SCTP_INP_INFO_RUNLOCK();
2006 	}
2007 	if (it->inp == NULL) {
2008 		goto done_with_iterator;
2009 	}
2010 	goto select_a_new_ep;
2011 }
2012