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#ifdef __FreeBSD__
34#include <sys/cdefs.h>
35__FBSDID("$FreeBSD: head/sys/netinet/sctp_timer.c 269448 2014-08-02 21:36:40Z tuexen $");
36#endif
37
38#define _IP_VHL
39#include <netinet/sctp_os.h>
40#include <netinet/sctp_pcb.h>
41#ifdef INET6
42#if defined(__Userspace_os_FreeBSD)
43#include <netinet6/sctp6_var.h>
44#endif
45#endif
46#include <netinet/sctp_var.h>
47#include <netinet/sctp_sysctl.h>
48#include <netinet/sctp_timer.h>
49#include <netinet/sctputil.h>
50#include <netinet/sctp_output.h>
51#include <netinet/sctp_header.h>
52#include <netinet/sctp_indata.h>
53#include <netinet/sctp_asconf.h>
54#include <netinet/sctp_input.h>
55#include <netinet/sctp.h>
56#include <netinet/sctp_uio.h>
57#if defined(INET) || defined(INET6)
58#if !defined(__Userspace_os_Windows)
59#include <netinet/udp.h>
60#endif
61#endif
62
63#if defined(__APPLE__)
64#define APPLE_FILE_NO 6
65#endif
66
67void
68sctp_audit_retranmission_queue(struct sctp_association *asoc)
69{
70	struct sctp_tmit_chunk *chk;
71
72	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
73			asoc->sent_queue_retran_cnt,
74			asoc->sent_queue_cnt);
75	asoc->sent_queue_retran_cnt = 0;
76	asoc->sent_queue_cnt = 0;
77	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
78		if (chk->sent == SCTP_DATAGRAM_RESEND) {
79			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
80		}
81		asoc->sent_queue_cnt++;
82	}
83	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
84		if (chk->sent == SCTP_DATAGRAM_RESEND) {
85			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
86		}
87	}
88	TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) {
89		if (chk->sent == SCTP_DATAGRAM_RESEND) {
90			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
91		}
92	}
93	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
94		asoc->sent_queue_retran_cnt,
95		asoc->sent_queue_cnt);
96}
97
98int
99sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
100    struct sctp_nets *net, uint16_t threshold)
101{
102	if (net) {
103		net->error_count++;
104		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
105			(void *)net, net->error_count,
106			net->failure_threshold);
107		if (net->error_count > net->failure_threshold) {
108			/* We had a threshold failure */
109			if (net->dest_state & SCTP_ADDR_REACHABLE) {
110				net->dest_state &= ~SCTP_ADDR_REACHABLE;
111				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
112				net->dest_state &= ~SCTP_ADDR_PF;
113				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
114				    stcb, 0,
115				    (void *)net, SCTP_SO_NOT_LOCKED);
116			}
117		} else if ((net->pf_threshold < net->failure_threshold) &&
118		           (net->error_count > net->pf_threshold)) {
119			if (!(net->dest_state & SCTP_ADDR_PF)) {
120				net->dest_state |= SCTP_ADDR_PF;
121				net->last_active = sctp_get_tick_count();
122				sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
123				sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
124				sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net);
125			}
126		}
127	}
128	if (stcb == NULL)
129		return (0);
130
131	if (net) {
132		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
133			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
134				sctp_misc_ints(SCTP_THRESHOLD_INCR,
135					       stcb->asoc.overall_error_count,
136					       (stcb->asoc.overall_error_count+1),
137					       SCTP_FROM_SCTP_TIMER,
138					       __LINE__);
139			}
140			stcb->asoc.overall_error_count++;
141		}
142	} else {
143		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
144			sctp_misc_ints(SCTP_THRESHOLD_INCR,
145				       stcb->asoc.overall_error_count,
146				       (stcb->asoc.overall_error_count+1),
147				       SCTP_FROM_SCTP_TIMER,
148				       __LINE__);
149		}
150		stcb->asoc.overall_error_count++;
151	}
152	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
153		(void *)&stcb->asoc, stcb->asoc.overall_error_count,
154		(uint32_t)threshold,
155		((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state));
156	/*
157	 * We specifically do not do >= to give the assoc one more change
158	 * before we fail it.
159	 */
160	if (stcb->asoc.overall_error_count > threshold) {
161		/* Abort notification sends a ULP notify */
162		struct mbuf *op_err;
163
164		op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION,
165		                             "Association error couter exceeded");
166		inp->last_abort_code = SCTP_FROM_SCTP_TIMER+SCTP_LOC_1;
167		sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
168		return (1);
169	}
170	return (0);
171}
172
173/*
174 * sctp_find_alternate_net() returns a non-NULL pointer as long
175 * the argument net is non-NULL.
176 */
177struct sctp_nets *
178sctp_find_alternate_net(struct sctp_tcb *stcb,
179    struct sctp_nets *net,
180    int mode)
181{
182	/* Find and return an alternate network if possible */
183	struct sctp_nets *alt, *mnet, *min_errors_net = NULL , *max_cwnd_net = NULL;
184	int once;
185	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
186	int min_errors = -1;
187	uint32_t max_cwnd = 0;
188
189	if (stcb->asoc.numnets == 1) {
190		/* No others but net */
191		return (TAILQ_FIRST(&stcb->asoc.nets));
192	}
193	/*
194	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate net algorithm.
195	 * This algorithm chooses the active destination (not in PF state) with the largest
196	 * cwnd value. If all destinations are in PF state, unreachable, or unconfirmed, choose
197	 * the desination that is in PF state with the lowest error count. In case of a tie,
198	 * choose the destination that was most recently active.
199	 */
200	if (mode == 2) {
201		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
202			/* JRS 5/14/07 - If the destination is unreachable or unconfirmed, skip it. */
203			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
204			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
205				continue;
206			}
207			/*
208			 * JRS 5/14/07 -  If the destination is reachable but in PF state, compare
209			 *  the error count of the destination to the minimum error count seen thus far.
210			 *  Store the destination with the lower error count.  If the error counts are
211			 *  equal, store the destination that was most recently active.
212			 */
213			if (mnet->dest_state & SCTP_ADDR_PF) {
214				/*
215				 * JRS 5/14/07 - If the destination under consideration is the current
216				 *  destination, work as if the error count is one higher.  The
217				 *  actual error count will not be incremented until later in the
218				 *  t3 handler.
219				 */
220				if (mnet == net) {
221					if (min_errors == -1) {
222						min_errors = mnet->error_count + 1;
223						min_errors_net = mnet;
224					} else if (mnet->error_count + 1 < min_errors) {
225						min_errors = mnet->error_count + 1;
226						min_errors_net = mnet;
227					} else if (mnet->error_count + 1 == min_errors
228								&& mnet->last_active > min_errors_net->last_active) {
229						min_errors_net = mnet;
230						min_errors = mnet->error_count + 1;
231					}
232					continue;
233				} else {
234					if (min_errors == -1) {
235						min_errors = mnet->error_count;
236						min_errors_net = mnet;
237					} else if (mnet->error_count < min_errors) {
238						min_errors = mnet->error_count;
239						min_errors_net = mnet;
240					} else if (mnet->error_count == min_errors
241								&& mnet->last_active > min_errors_net->last_active) {
242						min_errors_net = mnet;
243						min_errors = mnet->error_count;
244					}
245					continue;
246				}
247			}
248			/*
249			 * JRS 5/14/07 - If the destination is reachable and not in PF state, compare the
250			 *  cwnd of the destination to the highest cwnd seen thus far.  Store the
251			 *  destination with the higher cwnd value.  If the cwnd values are equal,
252			 *  randomly choose one of the two destinations.
253			 */
254			if (max_cwnd < mnet->cwnd) {
255				max_cwnd_net = mnet;
256				max_cwnd = mnet->cwnd;
257			} else if (max_cwnd == mnet->cwnd) {
258				uint32_t rndval;
259				uint8_t this_random;
260
261				if (stcb->asoc.hb_random_idx > 3) {
262					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
263					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
264					this_random = stcb->asoc.hb_random_values[0];
265					stcb->asoc.hb_random_idx++;
266					stcb->asoc.hb_ect_randombit = 0;
267				} else {
268					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
269					stcb->asoc.hb_random_idx++;
270					stcb->asoc.hb_ect_randombit = 0;
271				}
272				if (this_random % 2 == 1) {
273					max_cwnd_net = mnet;
274					max_cwnd = mnet->cwnd; /* Useless? */
275				}
276			}
277		}
278		if (max_cwnd_net == NULL) {
279			if (min_errors_net == NULL) {
280				return (net);
281			}
282			return (min_errors_net);
283		} else {
284			return (max_cwnd_net);
285		}
286	} /* JRS 5/14/07 - If mode is set to 1, use the CMT policy for choosing an alternate net. */
287	else if (mode == 1) {
288		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
289			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
290			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
291				/*
292				 * will skip ones that are not-reachable or
293				 * unconfirmed
294				 */
295				continue;
296			}
297			if (max_cwnd < mnet->cwnd) {
298				max_cwnd_net = mnet;
299				max_cwnd = mnet->cwnd;
300			} else if (max_cwnd == mnet->cwnd) {
301				uint32_t rndval;
302				uint8_t this_random;
303
304				if (stcb->asoc.hb_random_idx > 3) {
305					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
306					memcpy(stcb->asoc.hb_random_values, &rndval,
307					    sizeof(stcb->asoc.hb_random_values));
308					this_random = stcb->asoc.hb_random_values[0];
309					stcb->asoc.hb_random_idx = 0;
310					stcb->asoc.hb_ect_randombit = 0;
311				} else {
312					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
313					stcb->asoc.hb_random_idx++;
314					stcb->asoc.hb_ect_randombit = 0;
315				}
316				if (this_random % 2) {
317					max_cwnd_net = mnet;
318					max_cwnd = mnet->cwnd;
319				}
320			}
321		}
322		if (max_cwnd_net) {
323			return (max_cwnd_net);
324		}
325	}
326	mnet = net;
327	once = 0;
328
329	if (mnet == NULL) {
330		mnet = TAILQ_FIRST(&stcb->asoc.nets);
331		if (mnet == NULL) {
332			return (NULL);
333		}
334	}
335	do {
336		alt = TAILQ_NEXT(mnet, sctp_next);
337		if (alt == NULL)
338		{
339			once++;
340			if (once > 1) {
341				break;
342			}
343			alt = TAILQ_FIRST(&stcb->asoc.nets);
344			if (alt == NULL) {
345				return (NULL);
346			}
347		}
348		if (alt->ro.ro_rt == NULL) {
349			if (alt->ro._s_addr) {
350				sctp_free_ifa(alt->ro._s_addr);
351				alt->ro._s_addr = NULL;
352			}
353			alt->src_addr_selected = 0;
354		}
355		/*sa_ignore NO_NULL_CHK*/
356		if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
357		    (alt->ro.ro_rt != NULL) &&
358		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))) {
359			/* Found a reachable address */
360			break;
361		}
362		mnet = alt;
363	} while (alt != NULL);
364
365	if (alt == NULL) {
366		/* Case where NO insv network exists (dormant state) */
367		/* we rotate destinations */
368		once = 0;
369		mnet = net;
370		do {
371			if (mnet == NULL) {
372				return (TAILQ_FIRST(&stcb->asoc.nets));
373			}
374			alt = TAILQ_NEXT(mnet, sctp_next);
375			if (alt == NULL) {
376				once++;
377				if (once > 1) {
378					break;
379				}
380				alt = TAILQ_FIRST(&stcb->asoc.nets);
381			}
382			/*sa_ignore NO_NULL_CHK*/
383			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
384			    (alt != net)) {
385				/* Found an alternate address */
386				break;
387			}
388			mnet = alt;
389		} while (alt != NULL);
390	}
391	if (alt == NULL) {
392		return (net);
393	}
394	return (alt);
395}
396
397static void
398sctp_backoff_on_timeout(struct sctp_tcb *stcb,
399    struct sctp_nets *net,
400    int win_probe,
401    int num_marked, int num_abandoned)
402{
403	if (net->RTO == 0) {
404		net->RTO = stcb->asoc.minrto;
405	}
406	net->RTO <<= 1;
407	if (net->RTO > stcb->asoc.maxrto) {
408		net->RTO = stcb->asoc.maxrto;
409	}
410	if ((win_probe == 0) && (num_marked || num_abandoned)) {
411		/* We don't apply penalty to window probe scenarios */
412		/* JRS - Use the congestion control given in the CC module */
413		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
414	}
415}
416
417#ifndef INVARIANTS
418static void
419sctp_recover_sent_list(struct sctp_tcb *stcb)
420{
421	struct sctp_tmit_chunk *chk, *nchk;
422	struct sctp_association *asoc;
423
424	asoc = &stcb->asoc;
425	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
426		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.TSN_seq)) {
427			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
428			            (void *)chk, chk->rec.data.TSN_seq, asoc->last_acked_seq);
429			if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
430				if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
431					asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
432				}
433			}
434			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
435			if (PR_SCTP_ENABLED(chk->flags)) {
436				if (asoc->pr_sctp_cnt != 0)
437					asoc->pr_sctp_cnt--;
438			}
439			if (chk->data) {
440				/*sa_ignore NO_NULL_CHK*/
441				sctp_free_bufspace(stcb, asoc, chk, 1);
442				sctp_m_freem(chk->data);
443				chk->data = NULL;
444				if (asoc->prsctp_supported && PR_SCTP_BUF_ENABLED(chk->flags)) {
445					asoc->sent_queue_cnt_removeable--;
446				}
447			}
448			asoc->sent_queue_cnt--;
449			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
450		}
451	}
452	SCTP_PRINTF("after recover order is as follows\n");
453	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
454		SCTP_PRINTF("chk:%p TSN:%x\n", (void *)chk, chk->rec.data.TSN_seq);
455	}
456}
457#endif
458
459static int
460sctp_mark_all_for_resend(struct sctp_tcb *stcb,
461    struct sctp_nets *net,
462    struct sctp_nets *alt,
463    int window_probe,
464    int *num_marked,
465    int *num_abandoned)
466{
467
468	/*
469	 * Mark all chunks (well not all) that were sent to *net for
470	 * retransmission. Move them to alt for there destination as well...
471	 * We only mark chunks that have been outstanding long enough to
472	 * have received feed-back.
473	 */
474	struct sctp_tmit_chunk *chk, *nchk;
475	struct sctp_nets *lnets;
476	struct timeval now, min_wait, tv;
477	int cur_rto;
478	int cnt_abandoned;
479	int audit_tf, num_mk, fir;
480	unsigned int cnt_mk;
481	uint32_t orig_flight, orig_tf;
482	uint32_t tsnlast, tsnfirst;
483	int recovery_cnt = 0;
484
485
486	/* none in flight now */
487	audit_tf = 0;
488	fir = 0;
489	/*
490	 * figure out how long a data chunk must be pending before we can
491	 * mark it ..
492	 */
493	(void)SCTP_GETTIME_TIMEVAL(&now);
494	/* get cur rto in micro-seconds */
495	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
496	cur_rto *= 1000;
497	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
498		sctp_log_fr(cur_rto,
499			    stcb->asoc.peers_rwnd,
500			    window_probe,
501			    SCTP_FR_T3_MARK_TIME);
502		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
503		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
504	}
505	tv.tv_sec = cur_rto / 1000000;
506	tv.tv_usec = cur_rto % 1000000;
507#ifndef __FreeBSD__
508	timersub(&now, &tv, &min_wait);
509#else
510	min_wait = now;
511	timevalsub(&min_wait, &tv);
512#endif
513	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
514		/*
515		 * if we hit here, we don't have enough seconds on the clock
516		 * to account for the RTO. We just let the lower seconds be
517		 * the bounds and don't worry about it. This may mean we
518		 * will mark a lot more than we should.
519		 */
520		min_wait.tv_sec = min_wait.tv_usec = 0;
521	}
522	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
523		sctp_log_fr(cur_rto, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
524		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
525	}
526	/*
527	 * Our rwnd will be incorrect here since we are not adding back the
528	 * cnt * mbuf but we will fix that down below.
529	 */
530	orig_flight = net->flight_size;
531	orig_tf = stcb->asoc.total_flight;
532
533	net->fast_retran_ip = 0;
534	/* Now on to each chunk */
535	cnt_abandoned = 0;
536	num_mk = cnt_mk = 0;
537	tsnfirst = tsnlast = 0;
538#ifndef INVARIANTS
539 start_again:
540#endif
541	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
542		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.TSN_seq)) {
543			/* Strange case our list got out of order? */
544			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x\n",
545			            (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq);
546			recovery_cnt++;
547#ifdef INVARIANTS
548			panic("last acked >= chk on sent-Q");
549#else
550			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
551			sctp_recover_sent_list(stcb);
552			if (recovery_cnt < 10) {
553				goto start_again;
554			} else {
555				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
556			}
557#endif
558		}
559		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
560			/*
561			 * found one to mark: If it is less than
562			 * DATAGRAM_ACKED it MUST not be a skipped or marked
563			 * TSN but instead one that is either already set
564			 * for retransmission OR one that needs
565			 * retransmission.
566			 */
567
568			/* validate its been outstanding long enough */
569			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
570				sctp_log_fr(chk->rec.data.TSN_seq,
571					    chk->sent_rcv_time.tv_sec,
572					    chk->sent_rcv_time.tv_usec,
573					    SCTP_FR_T3_MARK_TIME);
574			}
575			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
576				/*
577				 * we have reached a chunk that was sent
578				 * some seconds past our min.. forget it we
579				 * will find no more to send.
580				 */
581				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
582					sctp_log_fr(0,
583						    chk->sent_rcv_time.tv_sec,
584						    chk->sent_rcv_time.tv_usec,
585						    SCTP_FR_T3_STOPPED);
586				}
587				continue;
588			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
589				   (window_probe == 0)) {
590				/*
591				 * we must look at the micro seconds to
592				 * know.
593				 */
594				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
595					/*
596					 * ok it was sent after our boundary
597					 * time.
598					 */
599					continue;
600				}
601			}
602			if (stcb->asoc.prsctp_supported && PR_SCTP_TTL_ENABLED(chk->flags)) {
603				/* Is it expired? */
604#ifndef __FreeBSD__
605				if (timercmp(&now, &chk->rec.data.timetodrop, >)) {
606#else
607				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
608#endif
609					/* Yes so drop it */
610					if (chk->data) {
611						(void)sctp_release_pr_sctp_chunk(stcb,
612										 chk,
613										 1,
614										 SCTP_SO_NOT_LOCKED);
615						cnt_abandoned++;
616					}
617					continue;
618				}
619			}
620			if (stcb->asoc.prsctp_supported && PR_SCTP_RTX_ENABLED(chk->flags)) {
621				/* Has it been retransmitted tv_sec times? */
622				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
623					if (chk->data) {
624						(void)sctp_release_pr_sctp_chunk(stcb,
625										 chk,
626										 1,
627										 SCTP_SO_NOT_LOCKED);
628						cnt_abandoned++;
629					}
630					continue;
631				}
632			}
633			if (chk->sent < SCTP_DATAGRAM_RESEND) {
634				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
635				num_mk++;
636				if (fir == 0) {
637					fir = 1;
638					tsnfirst = chk->rec.data.TSN_seq;
639				}
640				tsnlast = chk->rec.data.TSN_seq;
641				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
642					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
643						    0, SCTP_FR_T3_MARKED);
644				}
645
646				if (chk->rec.data.chunk_was_revoked) {
647					/* deflate the cwnd */
648					chk->whoTo->cwnd -= chk->book_size;
649					chk->rec.data.chunk_was_revoked = 0;
650				}
651				net->marked_retrans++;
652				stcb->asoc.marked_retrans++;
653				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
654					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
655						       chk->whoTo->flight_size,
656						       chk->book_size,
657						       (uintptr_t)chk->whoTo,
658						       chk->rec.data.TSN_seq);
659				}
660				sctp_flight_size_decrease(chk);
661				sctp_total_flight_decrease(stcb, chk);
662				stcb->asoc.peers_rwnd += chk->send_size;
663				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
664			}
665			chk->sent = SCTP_DATAGRAM_RESEND;
666			SCTP_STAT_INCR(sctps_markedretrans);
667
668			/* reset the TSN for striking and other FR stuff */
669			chk->rec.data.doing_fast_retransmit = 0;
670			/* Clear any time so NO RTT is being done */
671
672			if (chk->do_rtt) {
673				if (chk->whoTo->rto_needed == 0) {
674					chk->whoTo->rto_needed = 1;
675				}
676			}
677			chk->do_rtt = 0;
678			if (alt != net) {
679				sctp_free_remote_addr(chk->whoTo);
680				chk->no_fr_allowed = 1;
681				chk->whoTo = alt;
682				atomic_add_int(&alt->ref_count, 1);
683			} else {
684				chk->no_fr_allowed = 0;
685				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
686					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
687				} else {
688					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
689				}
690			}
691			/* CMT: Do not allow FRs on retransmitted TSNs.
692			 */
693			if (stcb->asoc.sctp_cmt_on_off > 0) {
694				chk->no_fr_allowed = 1;
695			}
696#ifdef THIS_SHOULD_NOT_BE_DONE
697		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
698			/* remember highest acked one */
699			could_be_sent = chk;
700#endif
701		}
702		if (chk->sent == SCTP_DATAGRAM_RESEND) {
703			cnt_mk++;
704		}
705	}
706	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
707		/* we did not subtract the same things? */
708		audit_tf = 1;
709	}
710
711	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
712		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
713	}
714#ifdef SCTP_DEBUG
715	if (num_mk) {
716		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
717			tsnlast);
718		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%ld\n",
719			num_mk, (u_long)stcb->asoc.peers_rwnd);
720		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
721			tsnlast);
722		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n",
723			num_mk,
724			(int)stcb->asoc.peers_rwnd);
725	}
726#endif
727	*num_marked = num_mk;
728	*num_abandoned = cnt_abandoned;
729	/* Now check for a ECN Echo that may be stranded And
730	 * include the cnt_mk'd to have all resends in the
731	 * control queue.
732	 */
733	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
734		if (chk->sent == SCTP_DATAGRAM_RESEND) {
735			cnt_mk++;
736		}
737		if ((chk->whoTo == net) &&
738		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
739			sctp_free_remote_addr(chk->whoTo);
740			chk->whoTo = alt;
741			if (chk->sent != SCTP_DATAGRAM_RESEND) {
742				chk->sent = SCTP_DATAGRAM_RESEND;
743				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
744				cnt_mk++;
745			}
746			atomic_add_int(&alt->ref_count, 1);
747		}
748	}
749#ifdef THIS_SHOULD_NOT_BE_DONE
750	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
751		/* fix it so we retransmit the highest acked anyway */
752		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
753		cnt_mk++;
754		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
755	}
756#endif
757	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
758#ifdef INVARIANTS
759		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
760			    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
761#endif
762#ifndef SCTP_AUDITING_ENABLED
763		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
764#endif
765	}
766	if (audit_tf) {
767		SCTPDBG(SCTP_DEBUG_TIMER4,
768			"Audit total flight due to negative value net:%p\n",
769			(void *)net);
770		stcb->asoc.total_flight = 0;
771		stcb->asoc.total_flight_count = 0;
772		/* Clear all networks flight size */
773		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
774			lnets->flight_size = 0;
775			SCTPDBG(SCTP_DEBUG_TIMER4,
776				"Net:%p c-f cwnd:%d ssthresh:%d\n",
777				(void *)lnets, lnets->cwnd, lnets->ssthresh);
778		}
779		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
780			if (chk->sent < SCTP_DATAGRAM_RESEND) {
781				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
782					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
783						       chk->whoTo->flight_size,
784						       chk->book_size,
785						       (uintptr_t)chk->whoTo,
786						       chk->rec.data.TSN_seq);
787				}
788
789				sctp_flight_size_increase(chk);
790				sctp_total_flight_increase(stcb, chk);
791			}
792		}
793	}
794	/* We return 1 if we only have a window probe outstanding */
795	return (0);
796}
797
798
799int
800sctp_t3rxt_timer(struct sctp_inpcb *inp,
801    struct sctp_tcb *stcb,
802    struct sctp_nets *net)
803{
804	struct sctp_nets *alt;
805	int win_probe, num_mk, num_abandoned;
806
807	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
808		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
809	}
810	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
811		struct sctp_nets *lnet;
812
813		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
814			if (net == lnet) {
815				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
816			} else {
817				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
818			}
819		}
820	}
821	/* Find an alternate and mark those for retransmission */
822	if ((stcb->asoc.peers_rwnd == 0) &&
823	    (stcb->asoc.total_flight < net->mtu)) {
824		SCTP_STAT_INCR(sctps_timowindowprobe);
825		win_probe = 1;
826	} else {
827		win_probe = 0;
828	}
829
830	if (win_probe == 0) {
831		/* We don't do normal threshold management on window probes */
832		if (sctp_threshold_management(inp, stcb, net,
833		    stcb->asoc.max_send_times)) {
834			/* Association was destroyed */
835			return (1);
836		} else {
837			if (net != stcb->asoc.primary_destination) {
838				/* send a immediate HB if our RTO is stale */
839				struct timeval now;
840				unsigned int ms_goneby;
841
842				(void)SCTP_GETTIME_TIMEVAL(&now);
843				if (net->last_sent_time.tv_sec) {
844					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
845				} else {
846					ms_goneby = 0;
847				}
848				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
849					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
850						/*
851						 * no recent feed back in an RTO or
852						 * more, request a RTT update
853						 */
854						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
855					}
856				}
857			}
858		}
859	} else {
860		/*
861		 * For a window probe we don't penalize the net's but only
862		 * the association. This may fail it if SACKs are not coming
863		 * back. If sack's are coming with rwnd locked at 0, we will
864		 * continue to hold things waiting for rwnd to raise
865		 */
866		if (sctp_threshold_management(inp, stcb, NULL,
867		    stcb->asoc.max_send_times)) {
868			/* Association was destroyed */
869			return (1);
870		}
871	}
872	if (stcb->asoc.sctp_cmt_on_off > 0) {
873		if (net->pf_threshold < net->failure_threshold) {
874			alt = sctp_find_alternate_net(stcb, net, 2);
875		} else {
876		        /*
877			 * CMT: Using RTX_SSTHRESH policy for CMT.
878			 * If CMT is being used, then pick dest with
879			 * largest ssthresh for any retransmission.
880			 */
881			alt = sctp_find_alternate_net(stcb, net, 1);
882			/*
883			 * CUCv2: If a different dest is picked for
884			 * the retransmission, then new
885			 * (rtx-)pseudo_cumack needs to be tracked
886			 * for orig dest. Let CUCv2 track new (rtx-)
887			 * pseudo-cumack always.
888			 */
889			net->find_pseudo_cumack = 1;
890			net->find_rtx_pseudo_cumack = 1;
891		}
892	} else {
893		alt = sctp_find_alternate_net(stcb, net, 0);
894	}
895
896	num_mk = 0;
897	num_abandoned = 0;
898	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
899				      &num_mk, &num_abandoned);
900	/* FR Loss recovery just ended with the T3. */
901	stcb->asoc.fast_retran_loss_recovery = 0;
902
903	/* CMT FR loss recovery ended with the T3 */
904	net->fast_retran_loss_recovery = 0;
905	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
906	    (net->flight_size == 0)) {
907		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins)(stcb, net);
908	}
909
910	/*
911	 * setup the sat loss recovery that prevents satellite cwnd advance.
912	 */
913	stcb->asoc.sat_t3_loss_recovery = 1;
914	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
915
916	/* Backoff the timer and cwnd */
917	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
918	if ((!(net->dest_state & SCTP_ADDR_REACHABLE)) ||
919	    (net->dest_state & SCTP_ADDR_PF)) {
920		/* Move all pending over too */
921		sctp_move_chunks_from_net(stcb, net);
922
923		/* Get the address that failed, to
924		 * force a new src address selecton and
925		 * a route allocation.
926		 */
927		if (net->ro._s_addr) {
928			sctp_free_ifa(net->ro._s_addr);
929			net->ro._s_addr = NULL;
930		}
931		net->src_addr_selected = 0;
932
933		/* Force a route allocation too */
934		if (net->ro.ro_rt) {
935			RTFREE(net->ro.ro_rt);
936			net->ro.ro_rt = NULL;
937		}
938
939		/* Was it our primary? */
940		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
941			/*
942			 * Yes, note it as such and find an alternate note:
943			 * this means HB code must use this to resent the
944			 * primary if it goes active AND if someone does a
945			 * change-primary then this flag must be cleared
946			 * from any net structures.
947			 */
948			if (stcb->asoc.alternate) {
949				sctp_free_remote_addr(stcb->asoc.alternate);
950			}
951			stcb->asoc.alternate = alt;
952			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
953		}
954	}
955	/*
956	 * Special case for cookie-echo'ed case, we don't do output but must
957	 * await the COOKIE-ACK before retransmission
958	 */
959	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
960		/*
961		 * Here we just reset the timer and start again since we
962		 * have not established the asoc
963		 */
964		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
965		return (0);
966	}
967	if (stcb->asoc.prsctp_supported) {
968		struct sctp_tmit_chunk *lchk;
969
970		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
971		/* C3. See if we need to send a Fwd-TSN */
972		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
973			send_forward_tsn(stcb, &stcb->asoc);
974			if (lchk) {
975				/* Assure a timer is up */
976				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
977			}
978		}
979	}
980	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
981		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
982	}
983	return (0);
984}
985
986int
987sctp_t1init_timer(struct sctp_inpcb *inp,
988    struct sctp_tcb *stcb,
989    struct sctp_nets *net)
990{
991	/* bump the thresholds */
992	if (stcb->asoc.delayed_connection) {
993		/*
994		 * special hook for delayed connection. The library did NOT
995		 * complete the rest of its sends.
996		 */
997		stcb->asoc.delayed_connection = 0;
998		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
999		return (0);
1000	}
1001	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
1002		return (0);
1003	}
1004	if (sctp_threshold_management(inp, stcb, net,
1005	    stcb->asoc.max_init_times)) {
1006		/* Association was destroyed */
1007		return (1);
1008	}
1009	stcb->asoc.dropped_special_cnt = 0;
1010	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1011	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1012		net->RTO = stcb->asoc.initial_init_rto_max;
1013	}
1014	if (stcb->asoc.numnets > 1) {
1015		/* If we have more than one addr use it */
1016		struct sctp_nets *alt;
1017
1018		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1019		if (alt != stcb->asoc.primary_destination) {
1020			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1021			stcb->asoc.primary_destination = alt;
1022		}
1023	}
1024	/* Send out a new init */
1025	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1026	return (0);
1027}
1028
1029/*
1030 * For cookie and asconf we actually need to find and mark for resend, then
1031 * increment the resend counter (after all the threshold management stuff of
1032 * course).
1033 */
1034int
1035sctp_cookie_timer(struct sctp_inpcb *inp,
1036    struct sctp_tcb *stcb,
1037    struct sctp_nets *net SCTP_UNUSED)
1038{
1039	struct sctp_nets *alt;
1040	struct sctp_tmit_chunk *cookie;
1041
1042	/* first before all else we must find the cookie */
1043	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1044		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1045			break;
1046		}
1047	}
1048	if (cookie == NULL) {
1049		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1050			/* FOOBAR! */
1051			struct mbuf *op_err;
1052
1053			op_err = sctp_generate_cause(SCTP_CAUSE_PROTOCOL_VIOLATION,
1054			                             "Cookie timer expired, but no cookie");
1055			inp->last_abort_code = SCTP_FROM_SCTP_TIMER+SCTP_LOC_4;
1056			sctp_abort_an_association(inp, stcb, op_err, SCTP_SO_NOT_LOCKED);
1057		} else {
1058#ifdef INVARIANTS
1059			panic("Cookie timer expires in wrong state?");
1060#else
1061			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1062			return (0);
1063#endif
1064		}
1065		return (0);
1066	}
1067	/* Ok we found the cookie, threshold management next */
1068	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1069	    stcb->asoc.max_init_times)) {
1070		/* Assoc is over */
1071		return (1);
1072	}
1073	/*
1074	 * cleared theshold management now lets backoff the address & select
1075	 * an alternate
1076	 */
1077	stcb->asoc.dropped_special_cnt = 0;
1078	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0);
1079	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1080	if (alt != cookie->whoTo) {
1081		sctp_free_remote_addr(cookie->whoTo);
1082		cookie->whoTo = alt;
1083		atomic_add_int(&alt->ref_count, 1);
1084	}
1085	/* Now mark the retran info */
1086	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1087		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1088	}
1089	cookie->sent = SCTP_DATAGRAM_RESEND;
1090	/*
1091	 * Now call the output routine to kick out the cookie again, Note we
1092	 * don't mark any chunks for retran so that FR will need to kick in
1093	 * to move these (or a send timer).
1094	 */
1095	return (0);
1096}
1097
1098int
1099sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1100    struct sctp_nets *net)
1101{
1102	struct sctp_nets *alt;
1103	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1104
1105	if (stcb->asoc.stream_reset_outstanding == 0) {
1106		return (0);
1107	}
1108	/* find the existing STRRESET, we use the seq number we sent out on */
1109	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1110	if (strrst == NULL) {
1111		return (0);
1112	}
1113	/* do threshold management */
1114	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1115	    stcb->asoc.max_send_times)) {
1116		/* Assoc is over */
1117		return (1);
1118	}
1119	/*
1120	 * cleared theshold management now lets backoff the address & select
1121	 * an alternate
1122	 */
1123	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0, 0);
1124	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1125	sctp_free_remote_addr(strrst->whoTo);
1126	strrst->whoTo = alt;
1127	atomic_add_int(&alt->ref_count, 1);
1128
1129	/* See if a ECN Echo is also stranded */
1130	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1131		if ((chk->whoTo == net) &&
1132		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1133			sctp_free_remote_addr(chk->whoTo);
1134			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1135				chk->sent = SCTP_DATAGRAM_RESEND;
1136				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1137			}
1138			chk->whoTo = alt;
1139			atomic_add_int(&alt->ref_count, 1);
1140		}
1141	}
1142	if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1143		/*
1144		 * If the address went un-reachable, we need to move to
1145		 * alternates for ALL chk's in queue
1146		 */
1147		sctp_move_chunks_from_net(stcb, net);
1148	}
1149	/* mark the retran info */
1150	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1151		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1152	strrst->sent = SCTP_DATAGRAM_RESEND;
1153
1154	/* restart the timer */
1155	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1156	return (0);
1157}
1158
1159int
1160sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1161		  struct sctp_nets *net)
1162{
1163	struct sctp_nets *alt;
1164	struct sctp_tmit_chunk *asconf, *chk;
1165
1166	/* is this a first send, or a retransmission? */
1167	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1168		/* compose a new ASCONF chunk and send it */
1169		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1170	} else {
1171		/*
1172		 * Retransmission of the existing ASCONF is needed
1173		 */
1174
1175		/* find the existing ASCONF */
1176		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1177		if (asconf == NULL) {
1178			return (0);
1179		}
1180		/* do threshold management */
1181		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1182		    stcb->asoc.max_send_times)) {
1183			/* Assoc is over */
1184			return (1);
1185		}
1186		if (asconf->snd_count > stcb->asoc.max_send_times) {
1187			/*
1188			 * Something is rotten: our peer is not responding to
1189			 * ASCONFs but apparently is to other chunks.  i.e. it
1190			 * is not properly handling the chunk type upper bits.
1191			 * Mark this peer as ASCONF incapable and cleanup.
1192			 */
1193			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1194			sctp_asconf_cleanup(stcb, net);
1195			return (0);
1196		}
1197		/*
1198		 * cleared threshold management, so now backoff the net and
1199		 * select an alternate
1200		 */
1201		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0, 0);
1202		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1203		if (asconf->whoTo != alt) {
1204			sctp_free_remote_addr(asconf->whoTo);
1205			asconf->whoTo = alt;
1206			atomic_add_int(&alt->ref_count, 1);
1207		}
1208
1209		/* See if an ECN Echo is also stranded */
1210		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1211			if ((chk->whoTo == net) &&
1212			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1213				sctp_free_remote_addr(chk->whoTo);
1214				chk->whoTo = alt;
1215				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1216					chk->sent = SCTP_DATAGRAM_RESEND;
1217					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1218				}
1219				atomic_add_int(&alt->ref_count, 1);
1220			}
1221		}
1222		TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) {
1223			if (chk->whoTo != alt) {
1224				sctp_free_remote_addr(chk->whoTo);
1225				chk->whoTo = alt;
1226				atomic_add_int(&alt->ref_count, 1);
1227			}
1228			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1229				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1230			chk->sent = SCTP_DATAGRAM_RESEND;
1231		}
1232		if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1233			/*
1234			 * If the address went un-reachable, we need to move
1235			 * to the alternate for ALL chunks in queue
1236			 */
1237			sctp_move_chunks_from_net(stcb, net);
1238		}
1239		/* mark the retran info */
1240		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1241			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1242		asconf->sent = SCTP_DATAGRAM_RESEND;
1243
1244		/* send another ASCONF if any and we can do */
1245		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1246	}
1247	return (0);
1248}
1249
1250/* Mobility adaptation */
1251void
1252sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1253                       struct sctp_nets *net SCTP_UNUSED)
1254{
1255	if (stcb->asoc.deleted_primary == NULL) {
1256		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1257		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1258		return;
1259	}
1260	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1261	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1262	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1263	stcb->asoc.deleted_primary = NULL;
1264	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1265	return;
1266}
1267
1268/*
1269 * For the shutdown and shutdown-ack, we do not keep one around on the
1270 * control queue. This means we must generate a new one and call the general
1271 * chunk output routine, AFTER having done threshold management.
1272 * It is assumed that net is non-NULL.
1273 */
1274int
1275sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1276    struct sctp_nets *net)
1277{
1278	struct sctp_nets *alt;
1279
1280	/* first threshold managment */
1281	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1282		/* Assoc is over */
1283		return (1);
1284	}
1285	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1286	/* second select an alternative */
1287	alt = sctp_find_alternate_net(stcb, net, 0);
1288
1289	/* third generate a shutdown into the queue for out net */
1290	sctp_send_shutdown(stcb, alt);
1291
1292	/* fourth restart timer */
1293	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1294	return (0);
1295}
1296
1297int
1298sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1299    struct sctp_nets *net)
1300{
1301	struct sctp_nets *alt;
1302
1303	/* first threshold managment */
1304	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1305		/* Assoc is over */
1306		return (1);
1307	}
1308	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1309	/* second select an alternative */
1310	alt = sctp_find_alternate_net(stcb, net, 0);
1311
1312	/* third generate a shutdown into the queue for out net */
1313	sctp_send_shutdown_ack(stcb, alt);
1314
1315	/* fourth restart timer */
1316	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1317	return (0);
1318}
1319
1320static void
1321sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1322    struct sctp_tcb *stcb)
1323{
1324	struct sctp_stream_queue_pending *sp;
1325	unsigned int i, chks_in_queue = 0;
1326	int being_filled = 0;
1327	/*
1328	 * This function is ONLY called when the send/sent queues are empty.
1329	 */
1330	if ((stcb == NULL) || (inp == NULL))
1331		return;
1332
1333	if (stcb->asoc.sent_queue_retran_cnt) {
1334		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1335			    stcb->asoc.sent_queue_retran_cnt);
1336		stcb->asoc.sent_queue_retran_cnt = 0;
1337	}
1338	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1339		/* No stream scheduler information, initialize scheduler */
1340		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 0);
1341		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1342			/* yep, we lost a stream or two */
1343			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1344		} else {
1345			/* no streams lost */
1346			stcb->asoc.total_output_queue_size = 0;
1347		}
1348	}
1349	/* Check to see if some data queued, if so report it */
1350	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1351		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1352			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1353				if (sp->msg_is_complete)
1354					being_filled++;
1355				chks_in_queue++;
1356			}
1357		}
1358	}
1359	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1360		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1361			    stcb->asoc.stream_queue_cnt, chks_in_queue);
1362	}
1363	if (chks_in_queue) {
1364		/* call the output queue function */
1365		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1366		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1367		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1368			/*
1369			 * Probably should go in and make it go back through
1370			 * and add fragments allowed
1371			 */
1372			if (being_filled == 0) {
1373				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1374					    chks_in_queue);
1375			}
1376		}
1377	} else {
1378		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1379			    (u_long)stcb->asoc.total_output_queue_size);
1380		stcb->asoc.total_output_queue_size = 0;
1381	}
1382}
1383
1384int
1385sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1386    struct sctp_nets *net)
1387{
1388	uint8_t net_was_pf;
1389
1390	if (net->dest_state & SCTP_ADDR_PF) {
1391		net_was_pf = 1;
1392	} else {
1393		net_was_pf = 0;
1394	}
1395	if (net->hb_responded == 0) {
1396		if (net->ro._s_addr) {
1397			/* Invalidate the src address if we did not get
1398			 * a response last time.
1399			 */
1400			sctp_free_ifa(net->ro._s_addr);
1401			net->ro._s_addr = NULL;
1402			net->src_addr_selected = 0;
1403		}
1404		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1405		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1406			/* Assoc is over */
1407			return (1);
1408		}
1409	}
1410	/* Zero PBA, if it needs it */
1411	if (net->partial_bytes_acked) {
1412		net->partial_bytes_acked = 0;
1413	}
1414	if ((stcb->asoc.total_output_queue_size > 0) &&
1415	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1416	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1417		sctp_audit_stream_queues_for_size(inp, stcb);
1418	}
1419	if (!(net->dest_state & SCTP_ADDR_NOHB) &&
1420	    !((net_was_pf == 0) && (net->dest_state & SCTP_ADDR_PF))) {
1421		/* when move to PF during threshold mangement, a HB has been
1422		   queued in that routine */
1423		uint32_t ms_gone_by;
1424
1425		if ((net->last_sent_time.tv_sec > 0) ||
1426		    (net->last_sent_time.tv_usec > 0)) {
1427#ifdef __FreeBSD__
1428			struct timeval diff;
1429
1430			SCTP_GETTIME_TIMEVAL(&diff);
1431			timevalsub(&diff, &net->last_sent_time);
1432#else
1433			struct timeval diff, now;
1434
1435			SCTP_GETTIME_TIMEVAL(&now);
1436			timersub(&now, &net->last_sent_time, &diff);
1437#endif
1438			ms_gone_by = (uint32_t)(diff.tv_sec * 1000) +
1439			             (uint32_t)(diff.tv_usec / 1000);
1440		} else {
1441			ms_gone_by = 0xffffffff;
1442		}
1443		if ((ms_gone_by >= net->heart_beat_delay) ||
1444		    (net->dest_state & SCTP_ADDR_PF)) {
1445			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1446		}
1447	}
1448	return (0);
1449}
1450
1451void
1452sctp_pathmtu_timer(struct sctp_inpcb *inp,
1453    struct sctp_tcb *stcb,
1454    struct sctp_nets *net)
1455{
1456	uint32_t next_mtu, mtu;
1457
1458	next_mtu = sctp_get_next_mtu(net->mtu);
1459
1460	if ((next_mtu > net->mtu) && (net->port == 0)) {
1461		if ((net->src_addr_selected == 0) ||
1462		    (net->ro._s_addr == NULL) ||
1463		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1464			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1465				sctp_free_ifa(net->ro._s_addr);
1466				net->ro._s_addr = NULL;
1467				net->src_addr_selected = 0;
1468			} else  if (net->ro._s_addr == NULL) {
1469#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1470				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1471					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1472					/* KAME hack: embed scopeid */
1473#if defined(__APPLE__)
1474#if defined(APPLE_LEOPARD) || defined(APPLE_SNOWLEOPARD)
1475					(void)in6_embedscope(&sin6->sin6_addr, sin6, NULL, NULL);
1476#else
1477					(void)in6_embedscope(&sin6->sin6_addr, sin6, NULL, NULL, NULL);
1478#endif
1479#elif defined(SCTP_KAME)
1480					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1481#else
1482					(void)in6_embedscope(&sin6->sin6_addr, sin6);
1483#endif
1484				}
1485#endif
1486
1487				net->ro._s_addr = sctp_source_address_selection(inp,
1488										stcb,
1489										(sctp_route_t *)&net->ro,
1490										net, 0, stcb->asoc.vrf_id);
1491#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1492				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1493					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1494#ifdef SCTP_KAME
1495					(void)sa6_recoverscope(sin6);
1496#else
1497					(void)in6_recoverscope(sin6, &sin6->sin6_addr, NULL);
1498#endif	/* SCTP_KAME */
1499				}
1500#endif	/* INET6 */
1501			}
1502			if (net->ro._s_addr)
1503				net->src_addr_selected = 1;
1504		}
1505		if (net->ro._s_addr) {
1506			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1507#if defined(INET) || defined(INET6)
1508			if (net->port) {
1509				mtu -= sizeof(struct udphdr);
1510			}
1511#endif
1512			if (mtu > next_mtu) {
1513				net->mtu = next_mtu;
1514			}
1515		}
1516	}
1517	/* restart the timer */
1518	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1519}
1520
1521void
1522sctp_autoclose_timer(struct sctp_inpcb *inp,
1523    struct sctp_tcb *stcb,
1524    struct sctp_nets *net)
1525{
1526	struct timeval tn, *tim_touse;
1527	struct sctp_association *asoc;
1528	int ticks_gone_by;
1529
1530	(void)SCTP_GETTIME_TIMEVAL(&tn);
1531	if (stcb->asoc.sctp_autoclose_ticks &&
1532	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1533		/* Auto close is on */
1534		asoc = &stcb->asoc;
1535		/* pick the time to use */
1536		if (asoc->time_last_rcvd.tv_sec >
1537		    asoc->time_last_sent.tv_sec) {
1538			tim_touse = &asoc->time_last_rcvd;
1539		} else {
1540			tim_touse = &asoc->time_last_sent;
1541		}
1542		/* Now has long enough transpired to autoclose? */
1543		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1544		if ((ticks_gone_by > 0) &&
1545		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1546			/*
1547			 * autoclose time has hit, call the output routine,
1548			 * which should do nothing just to be SURE we don't
1549			 * have hanging data. We can then safely check the
1550			 * queues and know that we are clear to send
1551			 * shutdown
1552			 */
1553			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1554			/* Are we clean? */
1555			if (TAILQ_EMPTY(&asoc->send_queue) &&
1556			    TAILQ_EMPTY(&asoc->sent_queue)) {
1557				/*
1558				 * there is nothing queued to send, so I'm
1559				 * done...
1560				 */
1561				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1562					/* only send SHUTDOWN 1st time thru */
1563					struct sctp_nets *netp;
1564
1565					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1566					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1567						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1568					}
1569					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1570					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1571					sctp_stop_timers_for_shutdown(stcb);
1572					if (stcb->asoc.alternate) {
1573						netp = stcb->asoc.alternate;
1574					} else {
1575						netp = stcb->asoc.primary_destination;
1576					}
1577					sctp_send_shutdown(stcb, netp);
1578					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1579							 stcb->sctp_ep, stcb,
1580							 netp);
1581					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1582							 stcb->sctp_ep, stcb,
1583							 netp);
1584				}
1585			}
1586		} else {
1587			/*
1588			 * No auto close at this time, reset t-o to check
1589			 * later
1590			 */
1591			int tmp;
1592
1593			/* fool the timer startup to use the time left */
1594			tmp = asoc->sctp_autoclose_ticks;
1595			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1596			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1597			    net);
1598			/* restore the real tick value */
1599			asoc->sctp_autoclose_ticks = tmp;
1600		}
1601	}
1602}
1603
1604