sctp_timer.c revision 234297
1/*-
2 * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2011, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2011, by Michael Tuexen. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * a) Redistributions of source code must retain the above copyright notice,
10 *    this list of conditions and the following disclaimer.
11 *
12 * b) Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *    the documentation and/or other materials provided with the distribution.
15 *
16 * c) Neither the name of Cisco Systems, Inc. nor the names of its
17 *    contributors may be used to endorse or promote products derived
18 *    from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33/* $KAME: sctp_timer.c,v 1.29 2005/03/06 16:04:18 itojun Exp $	 */
34
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: head/sys/netinet/sctp_timer.c 234297 2012-04-14 21:01:44Z tuexen $");
37
38#define _IP_VHL
39#include <netinet/sctp_os.h>
40#include <netinet/sctp_pcb.h>
41#ifdef INET6
42#endif
43#include <netinet/sctp_var.h>
44#include <netinet/sctp_sysctl.h>
45#include <netinet/sctp_timer.h>
46#include <netinet/sctputil.h>
47#include <netinet/sctp_output.h>
48#include <netinet/sctp_header.h>
49#include <netinet/sctp_indata.h>
50#include <netinet/sctp_asconf.h>
51#include <netinet/sctp_input.h>
52#include <netinet/sctp.h>
53#include <netinet/sctp_uio.h>
54#include <netinet/udp.h>
55
56
57void
58sctp_audit_retranmission_queue(struct sctp_association *asoc)
59{
60	struct sctp_tmit_chunk *chk;
61
62	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
63	    asoc->sent_queue_retran_cnt,
64	    asoc->sent_queue_cnt);
65	asoc->sent_queue_retran_cnt = 0;
66	asoc->sent_queue_cnt = 0;
67	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
68		if (chk->sent == SCTP_DATAGRAM_RESEND) {
69			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
70		}
71		asoc->sent_queue_cnt++;
72	}
73	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
74		if (chk->sent == SCTP_DATAGRAM_RESEND) {
75			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
76		}
77	}
78	TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) {
79		if (chk->sent == SCTP_DATAGRAM_RESEND) {
80			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
81		}
82	}
83	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
84	    asoc->sent_queue_retran_cnt,
85	    asoc->sent_queue_cnt);
86}
87
88int
89sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
90    struct sctp_nets *net, uint16_t threshold)
91{
92	if (net) {
93		net->error_count++;
94		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
95		    net, net->error_count,
96		    net->failure_threshold);
97		if (net->error_count > net->failure_threshold) {
98			/* We had a threshold failure */
99			if (net->dest_state & SCTP_ADDR_REACHABLE) {
100				net->dest_state &= ~SCTP_ADDR_REACHABLE;
101				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
102				net->dest_state &= ~SCTP_ADDR_PF;
103				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
104				    stcb,
105				    SCTP_FAILED_THRESHOLD,
106				    (void *)net, SCTP_SO_NOT_LOCKED);
107			}
108		} else if ((net->pf_threshold < net->failure_threshold) &&
109		    (net->error_count > net->pf_threshold)) {
110			if (!(net->dest_state & SCTP_ADDR_PF)) {
111				net->dest_state |= SCTP_ADDR_PF;
112				net->last_active = sctp_get_tick_count();
113				sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
114				sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
115				sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net);
116			}
117		}
118	}
119	if (stcb == NULL)
120		return (0);
121
122	if (net) {
123		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
124			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
125				sctp_misc_ints(SCTP_THRESHOLD_INCR,
126				    stcb->asoc.overall_error_count,
127				    (stcb->asoc.overall_error_count + 1),
128				    SCTP_FROM_SCTP_TIMER,
129				    __LINE__);
130			}
131			stcb->asoc.overall_error_count++;
132		}
133	} else {
134		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
135			sctp_misc_ints(SCTP_THRESHOLD_INCR,
136			    stcb->asoc.overall_error_count,
137			    (stcb->asoc.overall_error_count + 1),
138			    SCTP_FROM_SCTP_TIMER,
139			    __LINE__);
140		}
141		stcb->asoc.overall_error_count++;
142	}
143	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
144	    &stcb->asoc, stcb->asoc.overall_error_count,
145	    (uint32_t) threshold,
146	    ((net == NULL) ? (uint32_t) 0 : (uint32_t) net->dest_state));
147	/*
148	 * We specifically do not do >= to give the assoc one more change
149	 * before we fail it.
150	 */
151	if (stcb->asoc.overall_error_count > threshold) {
152		/* Abort notification sends a ULP notify */
153		struct mbuf *oper;
154
155		oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
156		    0, M_DONTWAIT, 1, MT_DATA);
157		if (oper) {
158			struct sctp_paramhdr *ph;
159			uint32_t *ippp;
160
161			SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
162			    sizeof(uint32_t);
163			ph = mtod(oper, struct sctp_paramhdr *);
164			ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
165			ph->param_length = htons(SCTP_BUF_LEN(oper));
166			ippp = (uint32_t *) (ph + 1);
167			*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_1);
168		}
169		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_1;
170		sctp_abort_an_association(inp, stcb, SCTP_FAILED_THRESHOLD, oper, SCTP_SO_NOT_LOCKED);
171		return (1);
172	}
173	return (0);
174}
175
176/*
177 * sctp_find_alternate_net() returns a non-NULL pointer as long
178 * the argument net is non-NULL.
179 */
180struct sctp_nets *
181sctp_find_alternate_net(struct sctp_tcb *stcb,
182    struct sctp_nets *net,
183    int mode)
184{
185	/* Find and return an alternate network if possible */
186	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
187	int once;
188
189	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
190	int min_errors = -1;
191	uint32_t max_cwnd = 0;
192
193	if (stcb->asoc.numnets == 1) {
194		/* No others but net */
195		return (TAILQ_FIRST(&stcb->asoc.nets));
196	}
197	/*
198	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
199	 * net algorithm. This algorithm chooses the active destination (not
200	 * in PF state) with the largest cwnd value. If all destinations are
201	 * in PF state, unreachable, or unconfirmed, choose the desination
202	 * that is in PF state with the lowest error count. In case of a
203	 * tie, choose the destination that was most recently active.
204	 */
205	if (mode == 2) {
206		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
207			/*
208			 * JRS 5/14/07 - If the destination is unreachable
209			 * or unconfirmed, skip it.
210			 */
211			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
212			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
213				continue;
214			}
215			/*
216			 * JRS 5/14/07 -  If the destination is reachable
217			 * but in PF state, compare the error count of the
218			 * destination to the minimum error count seen thus
219			 * far. Store the destination with the lower error
220			 * count.  If the error counts are equal, store the
221			 * destination that was most recently active.
222			 */
223			if (mnet->dest_state & SCTP_ADDR_PF) {
224				/*
225				 * JRS 5/14/07 - If the destination under
226				 * consideration is the current destination,
227				 * work as if the error count is one higher.
228				 * The actual error count will not be
229				 * incremented until later in the t3
230				 * handler.
231				 */
232				if (mnet == net) {
233					if (min_errors == -1) {
234						min_errors = mnet->error_count + 1;
235						min_errors_net = mnet;
236					} else if (mnet->error_count + 1 < min_errors) {
237						min_errors = mnet->error_count + 1;
238						min_errors_net = mnet;
239					} else if (mnet->error_count + 1 == min_errors
240					    && mnet->last_active > min_errors_net->last_active) {
241						min_errors_net = mnet;
242						min_errors = mnet->error_count + 1;
243					}
244					continue;
245				} else {
246					if (min_errors == -1) {
247						min_errors = mnet->error_count;
248						min_errors_net = mnet;
249					} else if (mnet->error_count < min_errors) {
250						min_errors = mnet->error_count;
251						min_errors_net = mnet;
252					} else if (mnet->error_count == min_errors
253					    && mnet->last_active > min_errors_net->last_active) {
254						min_errors_net = mnet;
255						min_errors = mnet->error_count;
256					}
257					continue;
258				}
259			}
260			/*
261			 * JRS 5/14/07 - If the destination is reachable and
262			 * not in PF state, compare the cwnd of the
263			 * destination to the highest cwnd seen thus far.
264			 * Store the destination with the higher cwnd value.
265			 * If the cwnd values are equal, randomly choose one
266			 * of the two destinations.
267			 */
268			if (max_cwnd < mnet->cwnd) {
269				max_cwnd_net = mnet;
270				max_cwnd = mnet->cwnd;
271			} else if (max_cwnd == mnet->cwnd) {
272				uint32_t rndval;
273				uint8_t this_random;
274
275				if (stcb->asoc.hb_random_idx > 3) {
276					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
277					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
278					this_random = stcb->asoc.hb_random_values[0];
279					stcb->asoc.hb_random_idx++;
280					stcb->asoc.hb_ect_randombit = 0;
281				} else {
282					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
283					stcb->asoc.hb_random_idx++;
284					stcb->asoc.hb_ect_randombit = 0;
285				}
286				if (this_random % 2 == 1) {
287					max_cwnd_net = mnet;
288					max_cwnd = mnet->cwnd;	/* Useless? */
289				}
290			}
291		}
292		if (max_cwnd_net == NULL) {
293			if (min_errors_net == NULL) {
294				return (net);
295			}
296			return (min_errors_net);
297		} else {
298			return (max_cwnd_net);
299		}
300	}
301	/*
302	 * JRS 5/14/07 - If mode is set to 1, use the CMT policy for
303	 * choosing an alternate net.
304	 */
305	else if (mode == 1) {
306		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
307			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
308			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
309				/*
310				 * will skip ones that are not-reachable or
311				 * unconfirmed
312				 */
313				continue;
314			}
315			if (max_cwnd < mnet->cwnd) {
316				max_cwnd_net = mnet;
317				max_cwnd = mnet->cwnd;
318			} else if (max_cwnd == mnet->cwnd) {
319				uint32_t rndval;
320				uint8_t this_random;
321
322				if (stcb->asoc.hb_random_idx > 3) {
323					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
324					memcpy(stcb->asoc.hb_random_values, &rndval,
325					    sizeof(stcb->asoc.hb_random_values));
326					this_random = stcb->asoc.hb_random_values[0];
327					stcb->asoc.hb_random_idx = 0;
328					stcb->asoc.hb_ect_randombit = 0;
329				} else {
330					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
331					stcb->asoc.hb_random_idx++;
332					stcb->asoc.hb_ect_randombit = 0;
333				}
334				if (this_random % 2) {
335					max_cwnd_net = mnet;
336					max_cwnd = mnet->cwnd;
337				}
338			}
339		}
340		if (max_cwnd_net) {
341			return (max_cwnd_net);
342		}
343	}
344	mnet = net;
345	once = 0;
346
347	if (mnet == NULL) {
348		mnet = TAILQ_FIRST(&stcb->asoc.nets);
349		if (mnet == NULL) {
350			return (NULL);
351		}
352	}
353	do {
354		alt = TAILQ_NEXT(mnet, sctp_next);
355		if (alt == NULL) {
356			once++;
357			if (once > 1) {
358				break;
359			}
360			alt = TAILQ_FIRST(&stcb->asoc.nets);
361			if (alt == NULL) {
362				return (NULL);
363			}
364		}
365		if (alt->ro.ro_rt == NULL) {
366			if (alt->ro._s_addr) {
367				sctp_free_ifa(alt->ro._s_addr);
368				alt->ro._s_addr = NULL;
369			}
370			alt->src_addr_selected = 0;
371		}
372		/* sa_ignore NO_NULL_CHK */
373		if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
374		    (alt->ro.ro_rt != NULL) &&
375		    (!(alt->dest_state & SCTP_ADDR_UNCONFIRMED))) {
376			/* Found a reachable address */
377			break;
378		}
379		mnet = alt;
380	} while (alt != NULL);
381
382	if (alt == NULL) {
383		/* Case where NO insv network exists (dormant state) */
384		/* we rotate destinations */
385		once = 0;
386		mnet = net;
387		do {
388			if (mnet == NULL) {
389				return (TAILQ_FIRST(&stcb->asoc.nets));
390			}
391			alt = TAILQ_NEXT(mnet, sctp_next);
392			if (alt == NULL) {
393				once++;
394				if (once > 1) {
395					break;
396				}
397				alt = TAILQ_FIRST(&stcb->asoc.nets);
398			}
399			/* sa_ignore NO_NULL_CHK */
400			if ((!(alt->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
401			    (alt != net)) {
402				/* Found an alternate address */
403				break;
404			}
405			mnet = alt;
406		} while (alt != NULL);
407	}
408	if (alt == NULL) {
409		return (net);
410	}
411	return (alt);
412}
413
414static void
415sctp_backoff_on_timeout(struct sctp_tcb *stcb,
416    struct sctp_nets *net,
417    int win_probe,
418    int num_marked, int num_abandoned)
419{
420	if (net->RTO == 0) {
421		net->RTO = stcb->asoc.minrto;
422	}
423	net->RTO <<= 1;
424	if (net->RTO > stcb->asoc.maxrto) {
425		net->RTO = stcb->asoc.maxrto;
426	}
427	if ((win_probe == 0) && (num_marked || num_abandoned)) {
428		/* We don't apply penalty to window probe scenarios */
429		/* JRS - Use the congestion control given in the CC module */
430		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
431	}
432}
433
434#ifndef INVARIANTS
435static void
436sctp_recover_sent_list(struct sctp_tcb *stcb)
437{
438	struct sctp_tmit_chunk *chk, *nchk;
439	struct sctp_association *asoc;
440
441	asoc = &stcb->asoc;
442	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
443		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.TSN_seq)) {
444			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
445			    chk, chk->rec.data.TSN_seq, asoc->last_acked_seq);
446			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
447			if (chk->pr_sctp_on) {
448				if (asoc->pr_sctp_cnt != 0)
449					asoc->pr_sctp_cnt--;
450			}
451			if (chk->data) {
452				/* sa_ignore NO_NULL_CHK */
453				sctp_free_bufspace(stcb, asoc, chk, 1);
454				sctp_m_freem(chk->data);
455				chk->data = NULL;
456				if (asoc->peer_supports_prsctp && PR_SCTP_BUF_ENABLED(chk->flags)) {
457					asoc->sent_queue_cnt_removeable--;
458				}
459			}
460			asoc->sent_queue_cnt--;
461			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
462		}
463	}
464	SCTP_PRINTF("after recover order is as follows\n");
465	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
466		SCTP_PRINTF("chk:%p TSN:%x\n", chk, chk->rec.data.TSN_seq);
467	}
468}
469
470#endif
471
472static int
473sctp_mark_all_for_resend(struct sctp_tcb *stcb,
474    struct sctp_nets *net,
475    struct sctp_nets *alt,
476    int window_probe,
477    int *num_marked,
478    int *num_abandoned)
479{
480
481	/*
482	 * Mark all chunks (well not all) that were sent to *net for
483	 * retransmission. Move them to alt for there destination as well...
484	 * We only mark chunks that have been outstanding long enough to
485	 * have received feed-back.
486	 */
487	struct sctp_tmit_chunk *chk, *nchk;
488	struct sctp_nets *lnets;
489	struct timeval now, min_wait, tv;
490	int cur_rto;
491	int cnt_abandoned;
492	int audit_tf, num_mk, fir;
493	unsigned int cnt_mk;
494	uint32_t orig_flight, orig_tf;
495	uint32_t tsnlast, tsnfirst;
496	int recovery_cnt = 0;
497
498
499	/* none in flight now */
500	audit_tf = 0;
501	fir = 0;
502	/*
503	 * figure out how long a data chunk must be pending before we can
504	 * mark it ..
505	 */
506	(void)SCTP_GETTIME_TIMEVAL(&now);
507	/* get cur rto in micro-seconds */
508	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
509	cur_rto *= 1000;
510	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
511		sctp_log_fr(cur_rto,
512		    stcb->asoc.peers_rwnd,
513		    window_probe,
514		    SCTP_FR_T3_MARK_TIME);
515		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
516		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
517	}
518	tv.tv_sec = cur_rto / 1000000;
519	tv.tv_usec = cur_rto % 1000000;
520	min_wait = now;
521	timevalsub(&min_wait, &tv);
522	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
523		/*
524		 * if we hit here, we don't have enough seconds on the clock
525		 * to account for the RTO. We just let the lower seconds be
526		 * the bounds and don't worry about it. This may mean we
527		 * will mark a lot more than we should.
528		 */
529		min_wait.tv_sec = min_wait.tv_usec = 0;
530	}
531	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
532		sctp_log_fr(cur_rto, now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
533		sctp_log_fr(0, min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
534	}
535	/*
536	 * Our rwnd will be incorrect here since we are not adding back the
537	 * cnt * mbuf but we will fix that down below.
538	 */
539	orig_flight = net->flight_size;
540	orig_tf = stcb->asoc.total_flight;
541
542	net->fast_retran_ip = 0;
543	/* Now on to each chunk */
544	cnt_abandoned = 0;
545	num_mk = cnt_mk = 0;
546	tsnfirst = tsnlast = 0;
547#ifndef INVARIANTS
548start_again:
549#endif
550	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
551		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.TSN_seq)) {
552			/* Strange case our list got out of order? */
553			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x",
554			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.TSN_seq);
555			recovery_cnt++;
556#ifdef INVARIANTS
557			panic("last acked >= chk on sent-Q");
558#else
559			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
560			sctp_recover_sent_list(stcb);
561			if (recovery_cnt < 10) {
562				goto start_again;
563			} else {
564				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
565			}
566#endif
567		}
568		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
569			/*
570			 * found one to mark: If it is less than
571			 * DATAGRAM_ACKED it MUST not be a skipped or marked
572			 * TSN but instead one that is either already set
573			 * for retransmission OR one that needs
574			 * retransmission.
575			 */
576
577			/* validate its been outstanding long enough */
578			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
579				sctp_log_fr(chk->rec.data.TSN_seq,
580				    chk->sent_rcv_time.tv_sec,
581				    chk->sent_rcv_time.tv_usec,
582				    SCTP_FR_T3_MARK_TIME);
583			}
584			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
585				/*
586				 * we have reached a chunk that was sent
587				 * some seconds past our min.. forget it we
588				 * will find no more to send.
589				 */
590				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
591					sctp_log_fr(0,
592					    chk->sent_rcv_time.tv_sec,
593					    chk->sent_rcv_time.tv_usec,
594					    SCTP_FR_T3_STOPPED);
595				}
596				continue;
597			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
598			    (window_probe == 0)) {
599				/*
600				 * we must look at the micro seconds to
601				 * know.
602				 */
603				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
604					/*
605					 * ok it was sent after our boundary
606					 * time.
607					 */
608					continue;
609				}
610			}
611			if (stcb->asoc.peer_supports_prsctp && PR_SCTP_TTL_ENABLED(chk->flags)) {
612				/* Is it expired? */
613				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
614					/* Yes so drop it */
615					if (chk->data) {
616						(void)sctp_release_pr_sctp_chunk(stcb,
617						    chk,
618						    (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT),
619						    SCTP_SO_NOT_LOCKED);
620						cnt_abandoned++;
621					}
622					continue;
623				}
624			}
625			if (stcb->asoc.peer_supports_prsctp && PR_SCTP_RTX_ENABLED(chk->flags)) {
626				/* Has it been retransmitted tv_sec times? */
627				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
628					if (chk->data) {
629						(void)sctp_release_pr_sctp_chunk(stcb,
630						    chk,
631						    (SCTP_RESPONSE_TO_USER_REQ | SCTP_NOTIFY_DATAGRAM_SENT),
632						    SCTP_SO_NOT_LOCKED);
633						cnt_abandoned++;
634					}
635					continue;
636				}
637			}
638			if (chk->sent < SCTP_DATAGRAM_RESEND) {
639				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
640				num_mk++;
641				if (fir == 0) {
642					fir = 1;
643					tsnfirst = chk->rec.data.TSN_seq;
644				}
645				tsnlast = chk->rec.data.TSN_seq;
646				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
647					sctp_log_fr(chk->rec.data.TSN_seq, chk->snd_count,
648					    0, SCTP_FR_T3_MARKED);
649				}
650				if (chk->rec.data.chunk_was_revoked) {
651					/* deflate the cwnd */
652					chk->whoTo->cwnd -= chk->book_size;
653					chk->rec.data.chunk_was_revoked = 0;
654				}
655				net->marked_retrans++;
656				stcb->asoc.marked_retrans++;
657				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
658					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
659					    chk->whoTo->flight_size,
660					    chk->book_size,
661					    (uintptr_t) chk->whoTo,
662					    chk->rec.data.TSN_seq);
663				}
664				sctp_flight_size_decrease(chk);
665				sctp_total_flight_decrease(stcb, chk);
666				stcb->asoc.peers_rwnd += chk->send_size;
667				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
668			}
669			chk->sent = SCTP_DATAGRAM_RESEND;
670			SCTP_STAT_INCR(sctps_markedretrans);
671
672			/* reset the TSN for striking and other FR stuff */
673			chk->rec.data.doing_fast_retransmit = 0;
674			/* Clear any time so NO RTT is being done */
675
676			if (chk->do_rtt) {
677				if (chk->whoTo->rto_needed == 0) {
678					chk->whoTo->rto_needed = 1;
679				}
680			}
681			chk->do_rtt = 0;
682			if (alt != net) {
683				sctp_free_remote_addr(chk->whoTo);
684				chk->no_fr_allowed = 1;
685				chk->whoTo = alt;
686				atomic_add_int(&alt->ref_count, 1);
687			} else {
688				chk->no_fr_allowed = 0;
689				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
690					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
691				} else {
692					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.TSN_seq;
693				}
694			}
695			/*
696			 * CMT: Do not allow FRs on retransmitted TSNs.
697			 */
698			if (stcb->asoc.sctp_cmt_on_off > 0) {
699				chk->no_fr_allowed = 1;
700			}
701#ifdef THIS_SHOULD_NOT_BE_DONE
702		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
703			/* remember highest acked one */
704			could_be_sent = chk;
705#endif
706		}
707		if (chk->sent == SCTP_DATAGRAM_RESEND) {
708			cnt_mk++;
709		}
710	}
711	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
712		/* we did not subtract the same things? */
713		audit_tf = 1;
714	}
715	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
716		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
717	}
718#ifdef SCTP_DEBUG
719	if (num_mk) {
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:%ld\n",
723		    num_mk, (u_long)stcb->asoc.peers_rwnd);
724		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
725		    tsnlast);
726		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%d\n",
727		    num_mk,
728		    (int)stcb->asoc.peers_rwnd);
729	}
730#endif
731	*num_marked = num_mk;
732	*num_abandoned = cnt_abandoned;
733	/*
734	 * Now check for a ECN Echo that may be stranded And include the
735	 * cnt_mk'd to have all resends in the control queue.
736	 */
737	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
738		if (chk->sent == SCTP_DATAGRAM_RESEND) {
739			cnt_mk++;
740		}
741		if ((chk->whoTo == net) &&
742		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
743			sctp_free_remote_addr(chk->whoTo);
744			chk->whoTo = alt;
745			if (chk->sent != SCTP_DATAGRAM_RESEND) {
746				chk->sent = SCTP_DATAGRAM_RESEND;
747				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
748				cnt_mk++;
749			}
750			atomic_add_int(&alt->ref_count, 1);
751		}
752	}
753#ifdef THIS_SHOULD_NOT_BE_DONE
754	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
755		/* fix it so we retransmit the highest acked anyway */
756		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
757		cnt_mk++;
758		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
759	}
760#endif
761	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
762#ifdef INVARIANTS
763		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
764		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
765#endif
766#ifndef SCTP_AUDITING_ENABLED
767		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
768#endif
769	}
770	if (audit_tf) {
771		SCTPDBG(SCTP_DEBUG_TIMER4,
772		    "Audit total flight due to negative value net:%p\n",
773		    net);
774		stcb->asoc.total_flight = 0;
775		stcb->asoc.total_flight_count = 0;
776		/* Clear all networks flight size */
777		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
778			lnets->flight_size = 0;
779			SCTPDBG(SCTP_DEBUG_TIMER4,
780			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
781			    lnets, lnets->cwnd, lnets->ssthresh);
782		}
783		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
784			if (chk->sent < SCTP_DATAGRAM_RESEND) {
785				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
786					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
787					    chk->whoTo->flight_size,
788					    chk->book_size,
789					    (uintptr_t) chk->whoTo,
790					    chk->rec.data.TSN_seq);
791				}
792				sctp_flight_size_increase(chk);
793				sctp_total_flight_increase(stcb, chk);
794			}
795		}
796	}
797	/* We return 1 if we only have a window probe outstanding */
798	return (0);
799}
800
801
802int
803sctp_t3rxt_timer(struct sctp_inpcb *inp,
804    struct sctp_tcb *stcb,
805    struct sctp_nets *net)
806{
807	struct sctp_nets *alt;
808	int win_probe, num_mk, num_abandoned;
809
810	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
811		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
812	}
813	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
814		struct sctp_nets *lnet;
815
816		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
817			if (net == lnet) {
818				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
819			} else {
820				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
821			}
822		}
823	}
824	/* Find an alternate and mark those for retransmission */
825	if ((stcb->asoc.peers_rwnd == 0) &&
826	    (stcb->asoc.total_flight < net->mtu)) {
827		SCTP_STAT_INCR(sctps_timowindowprobe);
828		win_probe = 1;
829	} else {
830		win_probe = 0;
831	}
832
833	if (win_probe == 0) {
834		/* We don't do normal threshold management on window probes */
835		if (sctp_threshold_management(inp, stcb, net,
836		    stcb->asoc.max_send_times)) {
837			/* Association was destroyed */
838			return (1);
839		} else {
840			if (net != stcb->asoc.primary_destination) {
841				/* send a immediate HB if our RTO is stale */
842				struct timeval now;
843				unsigned int ms_goneby;
844
845				(void)SCTP_GETTIME_TIMEVAL(&now);
846				if (net->last_sent_time.tv_sec) {
847					ms_goneby = (now.tv_sec - net->last_sent_time.tv_sec) * 1000;
848				} else {
849					ms_goneby = 0;
850				}
851				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
852					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
853						/*
854						 * no recent feed back in an
855						 * RTO or more, request a
856						 * RTT update
857						 */
858						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
859					}
860				}
861			}
862		}
863	} else {
864		/*
865		 * For a window probe we don't penalize the net's but only
866		 * the association. This may fail it if SACKs are not coming
867		 * back. If sack's are coming with rwnd locked at 0, we will
868		 * continue to hold things waiting for rwnd to raise
869		 */
870		if (sctp_threshold_management(inp, stcb, NULL,
871		    stcb->asoc.max_send_times)) {
872			/* Association was destroyed */
873			return (1);
874		}
875	}
876	if (stcb->asoc.sctp_cmt_on_off > 0) {
877		if (net->pf_threshold < net->failure_threshold) {
878			alt = sctp_find_alternate_net(stcb, net, 2);
879		} else {
880			/*
881			 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is
882			 * being used, then pick dest with largest ssthresh
883			 * for any retransmission.
884			 */
885			alt = sctp_find_alternate_net(stcb, net, 1);
886			/*
887			 * CUCv2: If a different dest is picked for the
888			 * retransmission, then new (rtx-)pseudo_cumack
889			 * needs to be tracked for orig dest. Let CUCv2
890			 * track new (rtx-) pseudo-cumack always.
891			 */
892			net->find_pseudo_cumack = 1;
893			net->find_rtx_pseudo_cumack = 1;
894		}
895	} else {
896		alt = sctp_find_alternate_net(stcb, net, 0);
897	}
898
899	num_mk = 0;
900	num_abandoned = 0;
901	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
902	    &num_mk, &num_abandoned);
903	/* FR Loss recovery just ended with the T3. */
904	stcb->asoc.fast_retran_loss_recovery = 0;
905
906	/* CMT FR loss recovery ended with the T3 */
907	net->fast_retran_loss_recovery = 0;
908	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
909	    (net->flight_size == 0)) {
910		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) (stcb, net);
911	}
912	/*
913	 * setup the sat loss recovery that prevents satellite cwnd advance.
914	 */
915	stcb->asoc.sat_t3_loss_recovery = 1;
916	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
917
918	/* Backoff the timer and cwnd */
919	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
920	if ((!(net->dest_state & SCTP_ADDR_REACHABLE)) ||
921	    (net->dest_state & SCTP_ADDR_PF)) {
922		/* Move all pending over too */
923		sctp_move_chunks_from_net(stcb, net);
924
925		/*
926		 * Get the address that failed, to force a new src address
927		 * selecton and a route allocation.
928		 */
929		if (net->ro._s_addr) {
930			sctp_free_ifa(net->ro._s_addr);
931			net->ro._s_addr = NULL;
932		}
933		net->src_addr_selected = 0;
934
935		/* Force a route allocation too */
936		if (net->ro.ro_rt) {
937			RTFREE(net->ro.ro_rt);
938			net->ro.ro_rt = NULL;
939		}
940		/* Was it our primary? */
941		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
942			/*
943			 * Yes, note it as such and find an alternate note:
944			 * this means HB code must use this to resent the
945			 * primary if it goes active AND if someone does a
946			 * change-primary then this flag must be cleared
947			 * from any net structures.
948			 */
949			if (stcb->asoc.alternate) {
950				sctp_free_remote_addr(stcb->asoc.alternate);
951			}
952			stcb->asoc.alternate = alt;
953			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
954		}
955	}
956	/*
957	 * Special case for cookie-echo'ed case, we don't do output but must
958	 * await the COOKIE-ACK before retransmission
959	 */
960	if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
961		/*
962		 * Here we just reset the timer and start again since we
963		 * have not established the asoc
964		 */
965		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
966		return (0);
967	}
968	if (stcb->asoc.peer_supports_prsctp) {
969		struct sctp_tmit_chunk *lchk;
970
971		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
972		/* C3. See if we need to send a Fwd-TSN */
973		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
974			send_forward_tsn(stcb, &stcb->asoc);
975			if (lchk) {
976				/* Assure a timer is up */
977				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
978			}
979		}
980	}
981	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
982		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
983	}
984	return (0);
985}
986
987int
988sctp_t1init_timer(struct sctp_inpcb *inp,
989    struct sctp_tcb *stcb,
990    struct sctp_nets *net)
991{
992	/* bump the thresholds */
993	if (stcb->asoc.delayed_connection) {
994		/*
995		 * special hook for delayed connection. The library did NOT
996		 * complete the rest of its sends.
997		 */
998		stcb->asoc.delayed_connection = 0;
999		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1000		return (0);
1001	}
1002	if (SCTP_GET_STATE((&stcb->asoc)) != SCTP_STATE_COOKIE_WAIT) {
1003		return (0);
1004	}
1005	if (sctp_threshold_management(inp, stcb, net,
1006	    stcb->asoc.max_init_times)) {
1007		/* Association was destroyed */
1008		return (1);
1009	}
1010	stcb->asoc.dropped_special_cnt = 0;
1011	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1012	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1013		net->RTO = stcb->asoc.initial_init_rto_max;
1014	}
1015	if (stcb->asoc.numnets > 1) {
1016		/* If we have more than one addr use it */
1017		struct sctp_nets *alt;
1018
1019		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1020		if (alt != stcb->asoc.primary_destination) {
1021			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1022			stcb->asoc.primary_destination = alt;
1023		}
1024	}
1025	/* Send out a new init */
1026	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1027	return (0);
1028}
1029
1030/*
1031 * For cookie and asconf we actually need to find and mark for resend, then
1032 * increment the resend counter (after all the threshold management stuff of
1033 * course).
1034 */
1035int
1036sctp_cookie_timer(struct sctp_inpcb *inp,
1037    struct sctp_tcb *stcb,
1038    struct sctp_nets *net SCTP_UNUSED)
1039{
1040	struct sctp_nets *alt;
1041	struct sctp_tmit_chunk *cookie;
1042
1043	/* first before all else we must find the cookie */
1044	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1045		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1046			break;
1047		}
1048	}
1049	if (cookie == NULL) {
1050		if (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_COOKIE_ECHOED) {
1051			/* FOOBAR! */
1052			struct mbuf *oper;
1053
1054			oper = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
1055			    0, M_DONTWAIT, 1, MT_DATA);
1056			if (oper) {
1057				struct sctp_paramhdr *ph;
1058				uint32_t *ippp;
1059
1060				SCTP_BUF_LEN(oper) = sizeof(struct sctp_paramhdr) +
1061				    sizeof(uint32_t);
1062				ph = mtod(oper, struct sctp_paramhdr *);
1063				ph->param_type = htons(SCTP_CAUSE_PROTOCOL_VIOLATION);
1064				ph->param_length = htons(SCTP_BUF_LEN(oper));
1065				ippp = (uint32_t *) (ph + 1);
1066				*ippp = htonl(SCTP_FROM_SCTP_TIMER + SCTP_LOC_3);
1067			}
1068			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_4;
1069			sctp_abort_an_association(inp, stcb, SCTP_INTERNAL_ERROR,
1070			    oper, SCTP_SO_NOT_LOCKED);
1071		} else {
1072#ifdef INVARIANTS
1073			panic("Cookie timer expires in wrong state?");
1074#else
1075			SCTP_PRINTF("Strange in state %d not cookie-echoed yet c-e timer expires?\n", SCTP_GET_STATE(&stcb->asoc));
1076			return (0);
1077#endif
1078		}
1079		return (0);
1080	}
1081	/* Ok we found the cookie, threshold management next */
1082	if (sctp_threshold_management(inp, stcb, cookie->whoTo,
1083	    stcb->asoc.max_init_times)) {
1084		/* Assoc is over */
1085		return (1);
1086	}
1087	/*
1088	 * cleared theshold management now lets backoff the address & select
1089	 * an alternate
1090	 */
1091	stcb->asoc.dropped_special_cnt = 0;
1092	sctp_backoff_on_timeout(stcb, cookie->whoTo, 1, 0, 0);
1093	alt = sctp_find_alternate_net(stcb, cookie->whoTo, 0);
1094	if (alt != cookie->whoTo) {
1095		sctp_free_remote_addr(cookie->whoTo);
1096		cookie->whoTo = alt;
1097		atomic_add_int(&alt->ref_count, 1);
1098	}
1099	/* Now mark the retran info */
1100	if (cookie->sent != SCTP_DATAGRAM_RESEND) {
1101		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1102	}
1103	cookie->sent = SCTP_DATAGRAM_RESEND;
1104	/*
1105	 * Now call the output routine to kick out the cookie again, Note we
1106	 * don't mark any chunks for retran so that FR will need to kick in
1107	 * to move these (or a send timer).
1108	 */
1109	return (0);
1110}
1111
1112int
1113sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1114    struct sctp_nets *net)
1115{
1116	struct sctp_nets *alt;
1117	struct sctp_tmit_chunk *strrst = NULL, *chk = NULL;
1118
1119	if (stcb->asoc.stream_reset_outstanding == 0) {
1120		return (0);
1121	}
1122	/* find the existing STRRESET, we use the seq number we sent out on */
1123	(void)sctp_find_stream_reset(stcb, stcb->asoc.str_reset_seq_out, &strrst);
1124	if (strrst == NULL) {
1125		return (0);
1126	}
1127	/* do threshold management */
1128	if (sctp_threshold_management(inp, stcb, strrst->whoTo,
1129	    stcb->asoc.max_send_times)) {
1130		/* Assoc is over */
1131		return (1);
1132	}
1133	/*
1134	 * cleared theshold management now lets backoff the address & select
1135	 * an alternate
1136	 */
1137	sctp_backoff_on_timeout(stcb, strrst->whoTo, 1, 0, 0);
1138	alt = sctp_find_alternate_net(stcb, strrst->whoTo, 0);
1139	sctp_free_remote_addr(strrst->whoTo);
1140	strrst->whoTo = alt;
1141	atomic_add_int(&alt->ref_count, 1);
1142
1143	/* See if a ECN Echo is also stranded */
1144	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1145		if ((chk->whoTo == net) &&
1146		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1147			sctp_free_remote_addr(chk->whoTo);
1148			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1149				chk->sent = SCTP_DATAGRAM_RESEND;
1150				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1151			}
1152			chk->whoTo = alt;
1153			atomic_add_int(&alt->ref_count, 1);
1154		}
1155	}
1156	if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1157		/*
1158		 * If the address went un-reachable, we need to move to
1159		 * alternates for ALL chk's in queue
1160		 */
1161		sctp_move_chunks_from_net(stcb, net);
1162	}
1163	/* mark the retran info */
1164	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1165		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1166	strrst->sent = SCTP_DATAGRAM_RESEND;
1167
1168	/* restart the timer */
1169	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, strrst->whoTo);
1170	return (0);
1171}
1172
1173int
1174sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1175    struct sctp_nets *net)
1176{
1177	struct sctp_nets *alt;
1178	struct sctp_tmit_chunk *asconf, *chk;
1179
1180	/* is this a first send, or a retransmission? */
1181	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1182		/* compose a new ASCONF chunk and send it */
1183		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1184	} else {
1185		/*
1186		 * Retransmission of the existing ASCONF is needed
1187		 */
1188
1189		/* find the existing ASCONF */
1190		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1191		if (asconf == NULL) {
1192			return (0);
1193		}
1194		/* do threshold management */
1195		if (sctp_threshold_management(inp, stcb, asconf->whoTo,
1196		    stcb->asoc.max_send_times)) {
1197			/* Assoc is over */
1198			return (1);
1199		}
1200		if (asconf->snd_count > stcb->asoc.max_send_times) {
1201			/*
1202			 * Something is rotten: our peer is not responding
1203			 * to ASCONFs but apparently is to other chunks.
1204			 * i.e. it is not properly handling the chunk type
1205			 * upper bits. Mark this peer as ASCONF incapable
1206			 * and cleanup.
1207			 */
1208			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1209			sctp_asconf_cleanup(stcb, net);
1210			return (0);
1211		}
1212		/*
1213		 * cleared threshold management, so now backoff the net and
1214		 * select an alternate
1215		 */
1216		sctp_backoff_on_timeout(stcb, asconf->whoTo, 1, 0, 0);
1217		alt = sctp_find_alternate_net(stcb, asconf->whoTo, 0);
1218		if (asconf->whoTo != alt) {
1219			sctp_free_remote_addr(asconf->whoTo);
1220			asconf->whoTo = alt;
1221			atomic_add_int(&alt->ref_count, 1);
1222		}
1223		/* See if an ECN Echo is also stranded */
1224		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1225			if ((chk->whoTo == net) &&
1226			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1227				sctp_free_remote_addr(chk->whoTo);
1228				chk->whoTo = alt;
1229				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1230					chk->sent = SCTP_DATAGRAM_RESEND;
1231					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1232				}
1233				atomic_add_int(&alt->ref_count, 1);
1234			}
1235		}
1236		TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) {
1237			if (chk->whoTo != alt) {
1238				sctp_free_remote_addr(chk->whoTo);
1239				chk->whoTo = alt;
1240				atomic_add_int(&alt->ref_count, 1);
1241			}
1242			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1243				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1244			chk->sent = SCTP_DATAGRAM_RESEND;
1245		}
1246		if (!(net->dest_state & SCTP_ADDR_REACHABLE)) {
1247			/*
1248			 * If the address went un-reachable, we need to move
1249			 * to the alternate for ALL chunks in queue
1250			 */
1251			sctp_move_chunks_from_net(stcb, net);
1252		}
1253		/* mark the retran info */
1254		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1255			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1256		asconf->sent = SCTP_DATAGRAM_RESEND;
1257
1258		/* send another ASCONF if any and we can do */
1259		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1260	}
1261	return (0);
1262}
1263
1264/* Mobility adaptation */
1265void
1266sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1267    struct sctp_nets *net SCTP_UNUSED)
1268{
1269	if (stcb->asoc.deleted_primary == NULL) {
1270		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1271		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1272		return;
1273	}
1274	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1275	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1276	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1277	stcb->asoc.deleted_primary = NULL;
1278	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1279	return;
1280}
1281
1282/*
1283 * For the shutdown and shutdown-ack, we do not keep one around on the
1284 * control queue. This means we must generate a new one and call the general
1285 * chunk output routine, AFTER having done threshold management.
1286 * It is assumed that net is non-NULL.
1287 */
1288int
1289sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1290    struct sctp_nets *net)
1291{
1292	struct sctp_nets *alt;
1293
1294	/* first threshold managment */
1295	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1296		/* Assoc is over */
1297		return (1);
1298	}
1299	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1300	/* second select an alternative */
1301	alt = sctp_find_alternate_net(stcb, net, 0);
1302
1303	/* third generate a shutdown into the queue for out net */
1304	sctp_send_shutdown(stcb, alt);
1305
1306	/* fourth restart timer */
1307	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1308	return (0);
1309}
1310
1311int
1312sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1313    struct sctp_nets *net)
1314{
1315	struct sctp_nets *alt;
1316
1317	/* first threshold managment */
1318	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1319		/* Assoc is over */
1320		return (1);
1321	}
1322	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1323	/* second select an alternative */
1324	alt = sctp_find_alternate_net(stcb, net, 0);
1325
1326	/* third generate a shutdown into the queue for out net */
1327	sctp_send_shutdown_ack(stcb, alt);
1328
1329	/* fourth restart timer */
1330	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1331	return (0);
1332}
1333
1334static void
1335sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp,
1336    struct sctp_tcb *stcb)
1337{
1338	struct sctp_stream_queue_pending *sp;
1339	unsigned int i, chks_in_queue = 0;
1340	int being_filled = 0;
1341
1342	/*
1343	 * This function is ONLY called when the send/sent queues are empty.
1344	 */
1345	if ((stcb == NULL) || (inp == NULL))
1346		return;
1347
1348	if (stcb->asoc.sent_queue_retran_cnt) {
1349		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1350		    stcb->asoc.sent_queue_retran_cnt);
1351		stcb->asoc.sent_queue_retran_cnt = 0;
1352	}
1353	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1354		/* No stream scheduler information, initialize scheduler */
1355		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 0);
1356		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1357			/* yep, we lost a stream or two */
1358			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1359		} else {
1360			/* no streams lost */
1361			stcb->asoc.total_output_queue_size = 0;
1362		}
1363	}
1364	/* Check to see if some data queued, if so report it */
1365	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1366		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1367			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1368				if (sp->msg_is_complete)
1369					being_filled++;
1370				chks_in_queue++;
1371			}
1372		}
1373	}
1374	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1375		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1376		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1377	}
1378	if (chks_in_queue) {
1379		/* call the output queue function */
1380		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1381		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1382		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1383			/*
1384			 * Probably should go in and make it go back through
1385			 * and add fragments allowed
1386			 */
1387			if (being_filled == 0) {
1388				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1389				    chks_in_queue);
1390			}
1391		}
1392	} else {
1393		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1394		    (u_long)stcb->asoc.total_output_queue_size);
1395		stcb->asoc.total_output_queue_size = 0;
1396	}
1397}
1398
1399int
1400sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1401    struct sctp_nets *net)
1402{
1403	uint8_t net_was_pf;
1404
1405	if (net->dest_state & SCTP_ADDR_PF) {
1406		net_was_pf = 1;
1407	} else {
1408		net_was_pf = 0;
1409	}
1410	if (net->hb_responded == 0) {
1411		if (net->ro._s_addr) {
1412			/*
1413			 * Invalidate the src address if we did not get a
1414			 * response last time.
1415			 */
1416			sctp_free_ifa(net->ro._s_addr);
1417			net->ro._s_addr = NULL;
1418			net->src_addr_selected = 0;
1419		}
1420		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1421		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1422			/* Assoc is over */
1423			return (1);
1424		}
1425	}
1426	/* Zero PBA, if it needs it */
1427	if (net->partial_bytes_acked) {
1428		net->partial_bytes_acked = 0;
1429	}
1430	if ((stcb->asoc.total_output_queue_size > 0) &&
1431	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1432	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1433		sctp_audit_stream_queues_for_size(inp, stcb);
1434	}
1435	if (!(net->dest_state & SCTP_ADDR_NOHB) &&
1436	    !((net_was_pf == 0) && (net->dest_state & SCTP_ADDR_PF))) {
1437		/*
1438		 * when move to PF during threshold mangement, a HB has been
1439		 * queued in that routine
1440		 */
1441		uint32_t ms_gone_by;
1442
1443		if ((net->last_sent_time.tv_sec > 0) ||
1444		    (net->last_sent_time.tv_usec > 0)) {
1445			struct timeval diff;
1446
1447			SCTP_GETTIME_TIMEVAL(&diff);
1448			timevalsub(&diff, &net->last_sent_time);
1449			ms_gone_by = (uint32_t) (diff.tv_sec * 1000) +
1450			    (uint32_t) (diff.tv_usec / 1000);
1451		} else {
1452			ms_gone_by = 0xffffffff;
1453		}
1454		if ((ms_gone_by >= net->heart_beat_delay) ||
1455		    (net->dest_state & SCTP_ADDR_PF)) {
1456			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1457		}
1458	}
1459	return (0);
1460}
1461
1462void
1463sctp_pathmtu_timer(struct sctp_inpcb *inp,
1464    struct sctp_tcb *stcb,
1465    struct sctp_nets *net)
1466{
1467	uint32_t next_mtu, mtu;
1468
1469	next_mtu = sctp_get_next_mtu(net->mtu);
1470
1471	if ((next_mtu > net->mtu) && (net->port == 0)) {
1472		if ((net->src_addr_selected == 0) ||
1473		    (net->ro._s_addr == NULL) ||
1474		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1475			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1476				sctp_free_ifa(net->ro._s_addr);
1477				net->ro._s_addr = NULL;
1478				net->src_addr_selected = 0;
1479			} else if (net->ro._s_addr == NULL) {
1480#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1481				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1482					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1483
1484					/* KAME hack: embed scopeid */
1485					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1486				}
1487#endif
1488
1489				net->ro._s_addr = sctp_source_address_selection(inp,
1490				    stcb,
1491				    (sctp_route_t *) & net->ro,
1492				    net, 0, stcb->asoc.vrf_id);
1493#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1494				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1495					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1496
1497					(void)sa6_recoverscope(sin6);
1498				}
1499#endif				/* INET6 */
1500			}
1501			if (net->ro._s_addr)
1502				net->src_addr_selected = 1;
1503		}
1504		if (net->ro._s_addr) {
1505			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_rt);
1506			if (net->port) {
1507				mtu -= sizeof(struct udphdr);
1508			}
1509			if (mtu > next_mtu) {
1510				net->mtu = next_mtu;
1511			}
1512		}
1513	}
1514	/* restart the timer */
1515	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1516}
1517
1518void
1519sctp_autoclose_timer(struct sctp_inpcb *inp,
1520    struct sctp_tcb *stcb,
1521    struct sctp_nets *net)
1522{
1523	struct timeval tn, *tim_touse;
1524	struct sctp_association *asoc;
1525	int ticks_gone_by;
1526
1527	(void)SCTP_GETTIME_TIMEVAL(&tn);
1528	if (stcb->asoc.sctp_autoclose_ticks &&
1529	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1530		/* Auto close is on */
1531		asoc = &stcb->asoc;
1532		/* pick the time to use */
1533		if (asoc->time_last_rcvd.tv_sec >
1534		    asoc->time_last_sent.tv_sec) {
1535			tim_touse = &asoc->time_last_rcvd;
1536		} else {
1537			tim_touse = &asoc->time_last_sent;
1538		}
1539		/* Now has long enough transpired to autoclose? */
1540		ticks_gone_by = SEC_TO_TICKS(tn.tv_sec - tim_touse->tv_sec);
1541		if ((ticks_gone_by > 0) &&
1542		    (ticks_gone_by >= (int)asoc->sctp_autoclose_ticks)) {
1543			/*
1544			 * autoclose time has hit, call the output routine,
1545			 * which should do nothing just to be SURE we don't
1546			 * have hanging data. We can then safely check the
1547			 * queues and know that we are clear to send
1548			 * shutdown
1549			 */
1550			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1551			/* Are we clean? */
1552			if (TAILQ_EMPTY(&asoc->send_queue) &&
1553			    TAILQ_EMPTY(&asoc->sent_queue)) {
1554				/*
1555				 * there is nothing queued to send, so I'm
1556				 * done...
1557				 */
1558				if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
1559					/* only send SHUTDOWN 1st time thru */
1560					struct sctp_nets *netp;
1561
1562					if (stcb->asoc.alternate) {
1563						netp = stcb->asoc.alternate;
1564					} else {
1565						netp = stcb->asoc.primary_destination;
1566					}
1567					sctp_send_shutdown(stcb, netp);
1568					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
1569					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1570						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1571					}
1572					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
1573					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
1574					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1575					    stcb->sctp_ep, stcb,
1576					    netp);
1577					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1578					    stcb->sctp_ep, stcb,
1579					    netp);
1580				}
1581			}
1582		} else {
1583			/*
1584			 * No auto close at this time, reset t-o to check
1585			 * later
1586			 */
1587			int tmp;
1588
1589			/* fool the timer startup to use the time left */
1590			tmp = asoc->sctp_autoclose_ticks;
1591			asoc->sctp_autoclose_ticks -= ticks_gone_by;
1592			sctp_timer_start(SCTP_TIMER_TYPE_AUTOCLOSE, inp, stcb,
1593			    net);
1594			/* restore the real tick value */
1595			asoc->sctp_autoclose_ticks = tmp;
1596		}
1597	}
1598}
1599