1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
5 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
6 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
10 *
11 * a) Redistributions of source code must retain the above copyright notice,
12 *    this list of conditions and the following disclaimer.
13 *
14 * b) Redistributions in binary form must reproduce the above copyright
15 *    notice, this list of conditions and the following disclaimer in
16 *    the documentation and/or other materials provided with the distribution.
17 *
18 * c) Neither the name of Cisco Systems, Inc. nor the names of its
19 *    contributors may be used to endorse or promote products derived
20 *    from this software without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
24 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
26 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
29 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
30 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
31 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
32 * THE POSSIBILITY OF SUCH DAMAGE.
33 */
34
35#define _IP_VHL
36#include <netinet/sctp_os.h>
37#include <netinet/sctp_pcb.h>
38
39#include <netinet/sctp_var.h>
40#include <netinet/sctp_sysctl.h>
41#include <netinet/sctp_timer.h>
42#include <netinet/sctputil.h>
43#include <netinet/sctp_output.h>
44#include <netinet/sctp_header.h>
45#include <netinet/sctp_indata.h>
46#include <netinet/sctp_asconf.h>
47#include <netinet/sctp_input.h>
48#include <netinet/sctp.h>
49#include <netinet/sctp_uio.h>
50#if defined(INET) || defined(INET6)
51#include <netinet/udp.h>
52#endif
53
54void
55sctp_audit_retranmission_queue(struct sctp_association *asoc)
56{
57	struct sctp_tmit_chunk *chk;
58
59	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit invoked on send queue cnt:%d onqueue:%d\n",
60	    asoc->sent_queue_retran_cnt,
61	    asoc->sent_queue_cnt);
62	asoc->sent_queue_retran_cnt = 0;
63	asoc->sent_queue_cnt = 0;
64	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
65		if (chk->sent == SCTP_DATAGRAM_RESEND) {
66			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
67		}
68		asoc->sent_queue_cnt++;
69	}
70	TAILQ_FOREACH(chk, &asoc->control_send_queue, sctp_next) {
71		if (chk->sent == SCTP_DATAGRAM_RESEND) {
72			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
73		}
74	}
75	TAILQ_FOREACH(chk, &asoc->asconf_send_queue, sctp_next) {
76		if (chk->sent == SCTP_DATAGRAM_RESEND) {
77			sctp_ucount_incr(asoc->sent_queue_retran_cnt);
78		}
79	}
80	SCTPDBG(SCTP_DEBUG_TIMER4, "Audit completes retran:%d onqueue:%d\n",
81	    asoc->sent_queue_retran_cnt,
82	    asoc->sent_queue_cnt);
83}
84
85static int
86sctp_threshold_management(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
87    struct sctp_nets *net, uint16_t threshold)
88{
89	KASSERT(stcb != NULL, ("stcb is NULL"));
90	SCTP_TCB_LOCK_ASSERT(stcb);
91
92	if (net != NULL) {
93		net->error_count++;
94		SCTPDBG(SCTP_DEBUG_TIMER4, "Error count for %p now %d thresh:%d\n",
95		    (void *)net, net->error_count,
96		    net->failure_threshold);
97		if (net->error_count > net->failure_threshold) {
98			/* We had a threshold failure */
99			if (net->dest_state & SCTP_ADDR_REACHABLE) {
100				net->dest_state &= ~SCTP_ADDR_REACHABLE;
101				net->dest_state &= ~SCTP_ADDR_REQ_PRIMARY;
102				net->dest_state &= ~SCTP_ADDR_PF;
103				sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
104				    stcb, 0,
105				    (void *)net, SCTP_SO_NOT_LOCKED);
106			}
107		} else if ((net->pf_threshold < net->failure_threshold) &&
108		    (net->error_count > net->pf_threshold)) {
109			if ((net->dest_state & SCTP_ADDR_PF) == 0) {
110				net->dest_state |= SCTP_ADDR_PF;
111				net->last_active = sctp_get_tick_count();
112				sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
113				sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT,
114				    inp, stcb, net,
115				    SCTP_FROM_SCTP_TIMER + SCTP_LOC_1);
116				sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net);
117			}
118		}
119		if ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) {
120			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
121				sctp_misc_ints(SCTP_THRESHOLD_INCR,
122				    stcb->asoc.overall_error_count,
123				    (stcb->asoc.overall_error_count + 1),
124				    SCTP_FROM_SCTP_TIMER,
125				    __LINE__);
126			}
127			stcb->asoc.overall_error_count++;
128		}
129	} else {
130		if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_THRESHOLD_LOGGING) {
131			sctp_misc_ints(SCTP_THRESHOLD_INCR,
132			    stcb->asoc.overall_error_count,
133			    (stcb->asoc.overall_error_count + 1),
134			    SCTP_FROM_SCTP_TIMER,
135			    __LINE__);
136		}
137		stcb->asoc.overall_error_count++;
138	}
139	SCTPDBG(SCTP_DEBUG_TIMER4, "Overall error count for %p now %d thresh:%u state:%x\n",
140	    (void *)&stcb->asoc, stcb->asoc.overall_error_count,
141	    (uint32_t)threshold,
142	    ((net == NULL) ? (uint32_t)0 : (uint32_t)net->dest_state));
143	/*
144	 * We specifically do not do >= to give the assoc one more change
145	 * before we fail it.
146	 */
147	if (stcb->asoc.overall_error_count > threshold) {
148		/* Abort notification sends a ULP notify */
149		struct mbuf *op_err;
150
151		op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
152		    "Association error counter exceeded");
153		inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_2;
154		sctp_abort_an_association(inp, stcb, op_err, true, SCTP_SO_NOT_LOCKED);
155		return (1);
156	}
157	return (0);
158}
159
160/*
161 * sctp_find_alternate_net() returns a non-NULL pointer as long as there
162 * exists nets, which are not being deleted.
163 */
164struct sctp_nets *
165sctp_find_alternate_net(struct sctp_tcb *stcb,
166    struct sctp_nets *net,
167    int mode)
168{
169	/* Find and return an alternate network if possible */
170	struct sctp_nets *alt, *mnet, *min_errors_net = NULL, *max_cwnd_net = NULL;
171	bool looped;
172
173	/* JRS 5/14/07 - Initialize min_errors to an impossible value. */
174	int min_errors = -1;
175	uint32_t max_cwnd = 0;
176
177	if (stcb->asoc.numnets == 1) {
178		/* No selection can be made. */
179		return (TAILQ_FIRST(&stcb->asoc.nets));
180	}
181	/*
182	 * JRS 5/14/07 - If mode is set to 2, use the CMT PF find alternate
183	 * net algorithm. This algorithm chooses the active destination (not
184	 * in PF state) with the largest cwnd value. If all destinations are
185	 * in PF state, unreachable, or unconfirmed, choose the destination
186	 * that is in PF state with the lowest error count. In case of a
187	 * tie, choose the destination that was most recently active.
188	 */
189	if (mode == 2) {
190		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
191			/*
192			 * JRS 5/14/07 - If the destination is unreachable
193			 * or unconfirmed, skip it.
194			 */
195			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
196			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
197				continue;
198			}
199			/*
200			 * JRS 5/14/07 -  If the destination is reachable
201			 * but in PF state, compare the error count of the
202			 * destination to the minimum error count seen thus
203			 * far. Store the destination with the lower error
204			 * count.  If the error counts are equal, store the
205			 * destination that was most recently active.
206			 */
207			if (mnet->dest_state & SCTP_ADDR_PF) {
208				/*
209				 * JRS 5/14/07 - If the destination under
210				 * consideration is the current destination,
211				 * work as if the error count is one higher.
212				 * The actual error count will not be
213				 * incremented until later in the t3
214				 * handler.
215				 */
216				if (mnet == net) {
217					if (min_errors == -1) {
218						min_errors = mnet->error_count + 1;
219						min_errors_net = mnet;
220					} else if (mnet->error_count + 1 < min_errors) {
221						min_errors = mnet->error_count + 1;
222						min_errors_net = mnet;
223					} else if (mnet->error_count + 1 == min_errors
224					    && mnet->last_active > min_errors_net->last_active) {
225						min_errors_net = mnet;
226						min_errors = mnet->error_count + 1;
227					}
228					continue;
229				} else {
230					if (min_errors == -1) {
231						min_errors = mnet->error_count;
232						min_errors_net = mnet;
233					} else if (mnet->error_count < min_errors) {
234						min_errors = mnet->error_count;
235						min_errors_net = mnet;
236					} else if (mnet->error_count == min_errors
237					    && mnet->last_active > min_errors_net->last_active) {
238						min_errors_net = mnet;
239						min_errors = mnet->error_count;
240					}
241					continue;
242				}
243			}
244			/*
245			 * JRS 5/14/07 - If the destination is reachable and
246			 * not in PF state, compare the cwnd of the
247			 * destination to the highest cwnd seen thus far.
248			 * Store the destination with the higher cwnd value.
249			 * If the cwnd values are equal, randomly choose one
250			 * of the two destinations.
251			 */
252			if (max_cwnd < mnet->cwnd) {
253				max_cwnd_net = mnet;
254				max_cwnd = mnet->cwnd;
255			} else if (max_cwnd == mnet->cwnd) {
256				uint32_t rndval;
257				uint8_t this_random;
258
259				if (stcb->asoc.hb_random_idx > 3) {
260					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
261					memcpy(stcb->asoc.hb_random_values, &rndval, sizeof(stcb->asoc.hb_random_values));
262					this_random = stcb->asoc.hb_random_values[0];
263					stcb->asoc.hb_random_idx++;
264					stcb->asoc.hb_ect_randombit = 0;
265				} else {
266					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
267					stcb->asoc.hb_random_idx++;
268					stcb->asoc.hb_ect_randombit = 0;
269				}
270				if (this_random % 2 == 1) {
271					max_cwnd_net = mnet;
272					max_cwnd = mnet->cwnd;	/* Useless? */
273				}
274			}
275		}
276		if (max_cwnd_net == NULL) {
277			if (min_errors_net == NULL) {
278				return (net);
279			}
280			return (min_errors_net);
281		} else {
282			return (max_cwnd_net);
283		}
284	}			/* JRS 5/14/07 - If mode is set to 1, use the
285				 * CMT policy for choosing an alternate net. */
286	else if (mode == 1) {
287		TAILQ_FOREACH(mnet, &stcb->asoc.nets, sctp_next) {
288			if (((mnet->dest_state & SCTP_ADDR_REACHABLE) != SCTP_ADDR_REACHABLE) ||
289			    (mnet->dest_state & SCTP_ADDR_UNCONFIRMED)) {
290				/*
291				 * will skip ones that are not-reachable or
292				 * unconfirmed
293				 */
294				continue;
295			}
296			if (max_cwnd < mnet->cwnd) {
297				max_cwnd_net = mnet;
298				max_cwnd = mnet->cwnd;
299			} else if (max_cwnd == mnet->cwnd) {
300				uint32_t rndval;
301				uint8_t this_random;
302
303				if (stcb->asoc.hb_random_idx > 3) {
304					rndval = sctp_select_initial_TSN(&stcb->sctp_ep->sctp_ep);
305					memcpy(stcb->asoc.hb_random_values, &rndval,
306					    sizeof(stcb->asoc.hb_random_values));
307					this_random = stcb->asoc.hb_random_values[0];
308					stcb->asoc.hb_random_idx = 0;
309					stcb->asoc.hb_ect_randombit = 0;
310				} else {
311					this_random = stcb->asoc.hb_random_values[stcb->asoc.hb_random_idx];
312					stcb->asoc.hb_random_idx++;
313					stcb->asoc.hb_ect_randombit = 0;
314				}
315				if (this_random % 2) {
316					max_cwnd_net = mnet;
317					max_cwnd = mnet->cwnd;
318				}
319			}
320		}
321		if (max_cwnd_net) {
322			return (max_cwnd_net);
323		}
324	}
325	/* Look for an alternate net, which is active. */
326	if ((net != NULL) && ((net->dest_state & SCTP_ADDR_BEING_DELETED) == 0)) {
327		alt = TAILQ_NEXT(net, sctp_next);
328	} else {
329		alt = TAILQ_FIRST(&stcb->asoc.nets);
330	}
331	looped = false;
332	for (;;) {
333		if (alt == NULL) {
334			if (!looped) {
335				alt = TAILQ_FIRST(&stcb->asoc.nets);
336				looped = true;
337			}
338			/* Definitely out of candidates. */
339			if (alt == NULL) {
340				break;
341			}
342		}
343		if (alt->ro.ro_nh == NULL) {
344			if (alt->ro._s_addr) {
345				sctp_free_ifa(alt->ro._s_addr);
346				alt->ro._s_addr = NULL;
347			}
348			alt->src_addr_selected = 0;
349		}
350		if (((alt->dest_state & SCTP_ADDR_REACHABLE) == SCTP_ADDR_REACHABLE) &&
351		    (alt->ro.ro_nh != NULL) &&
352		    ((alt->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) &&
353		    (alt != net)) {
354			/* Found an alternate net, which is reachable. */
355			break;
356		}
357		alt = TAILQ_NEXT(alt, sctp_next);
358	}
359
360	if (alt == NULL) {
361		/*
362		 * In case no active alternate net has been found, look for
363		 * an alternate net, which is confirmed.
364		 */
365		if ((net != NULL) && ((net->dest_state & SCTP_ADDR_BEING_DELETED) == 0)) {
366			alt = TAILQ_NEXT(net, sctp_next);
367		} else {
368			alt = TAILQ_FIRST(&stcb->asoc.nets);
369		}
370		looped = false;
371		for (;;) {
372			if (alt == NULL) {
373				if (!looped) {
374					alt = TAILQ_FIRST(&stcb->asoc.nets);
375					looped = true;
376				}
377				/* Definitely out of candidates. */
378				if (alt == NULL) {
379					break;
380				}
381			}
382			if (((alt->dest_state & SCTP_ADDR_UNCONFIRMED) == 0) &&
383			    (alt != net)) {
384				/*
385				 * Found an alternate net, which is
386				 * confirmed.
387				 */
388				break;
389			}
390			alt = TAILQ_NEXT(alt, sctp_next);
391		}
392	}
393	if (alt == NULL) {
394		/*
395		 * In case no confirmed alternate net has been found, just
396		 * return net, if it is not being deleted. In the other case
397		 * just return the first net.
398		 */
399		if ((net != NULL) && ((net->dest_state & SCTP_ADDR_BEING_DELETED) == 0)) {
400			alt = net;
401		}
402		if (alt == NULL) {
403			alt = TAILQ_FIRST(&stcb->asoc.nets);
404		}
405	}
406	return (alt);
407}
408
409static void
410sctp_backoff_on_timeout(struct sctp_tcb *stcb,
411    struct sctp_nets *net,
412    int win_probe,
413    int num_marked, int num_abandoned)
414{
415	if (net->RTO == 0) {
416		if (net->RTO_measured) {
417			net->RTO = stcb->asoc.minrto;
418		} else {
419			net->RTO = stcb->asoc.initial_rto;
420		}
421	}
422	net->RTO <<= 1;
423	if (net->RTO > stcb->asoc.maxrto) {
424		net->RTO = stcb->asoc.maxrto;
425	}
426	if ((win_probe == 0) && (num_marked || num_abandoned)) {
427		/* We don't apply penalty to window probe scenarios */
428		/* JRS - Use the congestion control given in the CC module */
429		stcb->asoc.cc_functions.sctp_cwnd_update_after_timeout(stcb, net);
430	}
431}
432
433#ifndef INVARIANTS
434static void
435sctp_recover_sent_list(struct sctp_tcb *stcb)
436{
437	struct sctp_tmit_chunk *chk, *nchk;
438	struct sctp_association *asoc;
439
440	asoc = &stcb->asoc;
441	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
442		if (SCTP_TSN_GE(asoc->last_acked_seq, chk->rec.data.tsn)) {
443			SCTP_PRINTF("Found chk:%p tsn:%x <= last_acked_seq:%x\n",
444			    (void *)chk, chk->rec.data.tsn, asoc->last_acked_seq);
445			if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
446				if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) {
447					asoc->strmout[chk->rec.data.sid].chunks_on_queues--;
448				}
449			}
450			if ((asoc->strmout[chk->rec.data.sid].chunks_on_queues == 0) &&
451			    (asoc->strmout[chk->rec.data.sid].state == SCTP_STREAM_RESET_PENDING) &&
452			    TAILQ_EMPTY(&asoc->strmout[chk->rec.data.sid].outqueue)) {
453				asoc->trigger_reset = 1;
454			}
455			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
456			if (PR_SCTP_ENABLED(chk->flags)) {
457				if (asoc->pr_sctp_cnt != 0)
458					asoc->pr_sctp_cnt--;
459			}
460			if (chk->data) {
461				/* sa_ignore NO_NULL_CHK */
462				sctp_free_bufspace(stcb, asoc, chk, 1);
463				sctp_m_freem(chk->data);
464				chk->data = NULL;
465				if (asoc->prsctp_supported && PR_SCTP_BUF_ENABLED(chk->flags)) {
466					asoc->sent_queue_cnt_removeable--;
467				}
468			}
469			asoc->sent_queue_cnt--;
470			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
471		}
472	}
473	SCTP_PRINTF("after recover order is as follows\n");
474	TAILQ_FOREACH(chk, &asoc->sent_queue, sctp_next) {
475		SCTP_PRINTF("chk:%p TSN:%x\n", (void *)chk, chk->rec.data.tsn);
476	}
477}
478#endif
479
480static int
481sctp_mark_all_for_resend(struct sctp_tcb *stcb,
482    struct sctp_nets *net,
483    struct sctp_nets *alt,
484    int window_probe,
485    int *num_marked,
486    int *num_abandoned)
487{
488
489	/*
490	 * Mark all chunks (well not all) that were sent to *net for
491	 * retransmission. Move them to alt for there destination as well...
492	 * We only mark chunks that have been outstanding long enough to
493	 * have received feed-back.
494	 */
495	struct sctp_tmit_chunk *chk, *nchk;
496	struct sctp_nets *lnets;
497	struct timeval now, min_wait, tv;
498	int cur_rto;
499	int cnt_abandoned;
500	int audit_tf, num_mk, fir;
501	unsigned int cnt_mk;
502	uint32_t orig_flight, orig_tf;
503	uint32_t tsnlast, tsnfirst;
504#ifndef INVARIANTS
505	int recovery_cnt = 0;
506#endif
507
508	/* none in flight now */
509	audit_tf = 0;
510	fir = 0;
511	/*
512	 * figure out how long a data chunk must be pending before we can
513	 * mark it ..
514	 */
515	(void)SCTP_GETTIME_TIMEVAL(&now);
516	/* get cur rto in micro-seconds */
517	cur_rto = (net->lastsa >> SCTP_RTT_SHIFT) + net->lastsv;
518	cur_rto *= 1000;
519	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
520		sctp_log_fr(cur_rto,
521		    stcb->asoc.peers_rwnd,
522		    window_probe,
523		    SCTP_FR_T3_MARK_TIME);
524		sctp_log_fr(net->flight_size, 0, 0, SCTP_FR_CWND_REPORT);
525		sctp_log_fr(net->flight_size, net->cwnd, stcb->asoc.total_flight, SCTP_FR_CWND_REPORT);
526	}
527	tv.tv_sec = cur_rto / 1000000;
528	tv.tv_usec = cur_rto % 1000000;
529	min_wait = now;
530	timevalsub(&min_wait, &tv);
531	if (min_wait.tv_sec < 0 || min_wait.tv_usec < 0) {
532		/*
533		 * if we hit here, we don't have enough seconds on the clock
534		 * to account for the RTO. We just let the lower seconds be
535		 * the bounds and don't worry about it. This may mean we
536		 * will mark a lot more than we should.
537		 */
538		min_wait.tv_sec = min_wait.tv_usec = 0;
539	}
540	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
541		sctp_log_fr(cur_rto, (uint32_t)now.tv_sec, now.tv_usec, SCTP_FR_T3_MARK_TIME);
542		sctp_log_fr(0, (uint32_t)min_wait.tv_sec, min_wait.tv_usec, SCTP_FR_T3_MARK_TIME);
543	}
544	/*
545	 * Our rwnd will be incorrect here since we are not adding back the
546	 * cnt * mbuf but we will fix that down below.
547	 */
548	orig_flight = net->flight_size;
549	orig_tf = stcb->asoc.total_flight;
550
551	net->fast_retran_ip = 0;
552	/* Now on to each chunk */
553	cnt_abandoned = 0;
554	num_mk = cnt_mk = 0;
555	tsnfirst = tsnlast = 0;
556#ifndef INVARIANTS
557start_again:
558#endif
559	TAILQ_FOREACH_SAFE(chk, &stcb->asoc.sent_queue, sctp_next, nchk) {
560		if (SCTP_TSN_GE(stcb->asoc.last_acked_seq, chk->rec.data.tsn)) {
561			/* Strange case our list got out of order? */
562			SCTP_PRINTF("Our list is out of order? last_acked:%x chk:%x\n",
563			    (unsigned int)stcb->asoc.last_acked_seq, (unsigned int)chk->rec.data.tsn);
564#ifdef INVARIANTS
565			panic("last acked >= chk on sent-Q");
566#else
567			recovery_cnt++;
568			SCTP_PRINTF("Recover attempts a restart cnt:%d\n", recovery_cnt);
569			sctp_recover_sent_list(stcb);
570			if (recovery_cnt < 10) {
571				goto start_again;
572			} else {
573				SCTP_PRINTF("Recovery fails %d times??\n", recovery_cnt);
574			}
575#endif
576		}
577		if ((chk->whoTo == net) && (chk->sent < SCTP_DATAGRAM_ACKED)) {
578			/*
579			 * found one to mark: If it is less than
580			 * DATAGRAM_ACKED it MUST not be a skipped or marked
581			 * TSN but instead one that is either already set
582			 * for retransmission OR one that needs
583			 * retransmission.
584			 */
585
586			/* validate its been outstanding long enough */
587			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
588				sctp_log_fr(chk->rec.data.tsn,
589				    (uint32_t)chk->sent_rcv_time.tv_sec,
590				    chk->sent_rcv_time.tv_usec,
591				    SCTP_FR_T3_MARK_TIME);
592			}
593			if ((chk->sent_rcv_time.tv_sec > min_wait.tv_sec) && (window_probe == 0)) {
594				/*
595				 * we have reached a chunk that was sent
596				 * some seconds past our min.. forget it we
597				 * will find no more to send.
598				 */
599				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
600					sctp_log_fr(0,
601					    (uint32_t)chk->sent_rcv_time.tv_sec,
602					    chk->sent_rcv_time.tv_usec,
603					    SCTP_FR_T3_STOPPED);
604				}
605				continue;
606			} else if ((chk->sent_rcv_time.tv_sec == min_wait.tv_sec) &&
607			    (window_probe == 0)) {
608				/*
609				 * we must look at the micro seconds to
610				 * know.
611				 */
612				if (chk->sent_rcv_time.tv_usec >= min_wait.tv_usec) {
613					/*
614					 * ok it was sent after our boundary
615					 * time.
616					 */
617					continue;
618				}
619			}
620			if (stcb->asoc.prsctp_supported && PR_SCTP_TTL_ENABLED(chk->flags)) {
621				/* Is it expired? */
622				if (timevalcmp(&now, &chk->rec.data.timetodrop, >)) {
623					/* Yes so drop it */
624					if (chk->data) {
625						(void)sctp_release_pr_sctp_chunk(stcb,
626						    chk,
627						    1,
628						    SCTP_SO_NOT_LOCKED);
629						cnt_abandoned++;
630					}
631					continue;
632				}
633			}
634			if (stcb->asoc.prsctp_supported && PR_SCTP_RTX_ENABLED(chk->flags)) {
635				/* Has it been retransmitted tv_sec times? */
636				if (chk->snd_count > chk->rec.data.timetodrop.tv_sec) {
637					if (chk->data) {
638						(void)sctp_release_pr_sctp_chunk(stcb,
639						    chk,
640						    1,
641						    SCTP_SO_NOT_LOCKED);
642						cnt_abandoned++;
643					}
644					continue;
645				}
646			}
647			if (chk->sent < SCTP_DATAGRAM_RESEND) {
648				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
649				num_mk++;
650				if (fir == 0) {
651					fir = 1;
652					tsnfirst = chk->rec.data.tsn;
653				}
654				tsnlast = chk->rec.data.tsn;
655				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
656					sctp_log_fr(chk->rec.data.tsn, chk->snd_count,
657					    0, SCTP_FR_T3_MARKED);
658				}
659
660				if (chk->rec.data.chunk_was_revoked) {
661					/* deflate the cwnd */
662					chk->whoTo->cwnd -= chk->book_size;
663					chk->rec.data.chunk_was_revoked = 0;
664				}
665				net->marked_retrans++;
666				stcb->asoc.marked_retrans++;
667				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
668					sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_RSND_TO,
669					    chk->whoTo->flight_size,
670					    chk->book_size,
671					    (uint32_t)(uintptr_t)chk->whoTo,
672					    chk->rec.data.tsn);
673				}
674				sctp_flight_size_decrease(chk);
675				sctp_total_flight_decrease(stcb, chk);
676				stcb->asoc.peers_rwnd += chk->send_size;
677				stcb->asoc.peers_rwnd += SCTP_BASE_SYSCTL(sctp_peer_chunk_oh);
678			}
679			chk->sent = SCTP_DATAGRAM_RESEND;
680			chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
681			SCTP_STAT_INCR(sctps_markedretrans);
682
683			/* reset the TSN for striking and other FR stuff */
684			chk->rec.data.doing_fast_retransmit = 0;
685			/* Clear any time so NO RTT is being done */
686
687			if (chk->do_rtt) {
688				if (chk->whoTo->rto_needed == 0) {
689					chk->whoTo->rto_needed = 1;
690				}
691			}
692			chk->do_rtt = 0;
693			if (alt != net) {
694				sctp_free_remote_addr(chk->whoTo);
695				chk->no_fr_allowed = 1;
696				chk->whoTo = alt;
697				atomic_add_int(&alt->ref_count, 1);
698			} else {
699				chk->no_fr_allowed = 0;
700				if (TAILQ_EMPTY(&stcb->asoc.send_queue)) {
701					chk->rec.data.fast_retran_tsn = stcb->asoc.sending_seq;
702				} else {
703					chk->rec.data.fast_retran_tsn = (TAILQ_FIRST(&stcb->asoc.send_queue))->rec.data.tsn;
704				}
705			}
706			/*
707			 * CMT: Do not allow FRs on retransmitted TSNs.
708			 */
709			if (stcb->asoc.sctp_cmt_on_off > 0) {
710				chk->no_fr_allowed = 1;
711			}
712#ifdef THIS_SHOULD_NOT_BE_DONE
713		} else if (chk->sent == SCTP_DATAGRAM_ACKED) {
714			/* remember highest acked one */
715			could_be_sent = chk;
716#endif
717		}
718		if (chk->sent == SCTP_DATAGRAM_RESEND) {
719			cnt_mk++;
720		}
721	}
722	if ((orig_flight - net->flight_size) != (orig_tf - stcb->asoc.total_flight)) {
723		/* we did not subtract the same things? */
724		audit_tf = 1;
725	}
726
727	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
728		sctp_log_fr(tsnfirst, tsnlast, num_mk, SCTP_FR_T3_TIMEOUT);
729	}
730#ifdef SCTP_DEBUG
731	if (num_mk) {
732		SCTPDBG(SCTP_DEBUG_TIMER1, "LAST TSN marked was %x\n",
733		    tsnlast);
734		SCTPDBG(SCTP_DEBUG_TIMER1, "Num marked for retransmission was %d peer-rwd:%u\n",
735		    num_mk,
736		    stcb->asoc.peers_rwnd);
737	}
738#endif
739	*num_marked = num_mk;
740	*num_abandoned = cnt_abandoned;
741	/*
742	 * Now check for a ECN Echo that may be stranded And include the
743	 * cnt_mk'd to have all resends in the control queue.
744	 */
745	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
746		if (chk->sent == SCTP_DATAGRAM_RESEND) {
747			cnt_mk++;
748		}
749		if ((chk->whoTo == net) &&
750		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
751			sctp_free_remote_addr(chk->whoTo);
752			chk->whoTo = alt;
753			if (chk->sent != SCTP_DATAGRAM_RESEND) {
754				chk->sent = SCTP_DATAGRAM_RESEND;
755				chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
756				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
757				cnt_mk++;
758			}
759			atomic_add_int(&alt->ref_count, 1);
760		}
761	}
762#ifdef THIS_SHOULD_NOT_BE_DONE
763	if ((stcb->asoc.sent_queue_retran_cnt == 0) && (could_be_sent)) {
764		/* fix it so we retransmit the highest acked anyway */
765		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
766		cnt_mk++;
767		could_be_sent->sent = SCTP_DATAGRAM_RESEND;
768	}
769#endif
770	if (stcb->asoc.sent_queue_retran_cnt != cnt_mk) {
771#ifdef INVARIANTS
772		SCTP_PRINTF("Local Audit says there are %d for retran asoc cnt:%d we marked:%d this time\n",
773		    cnt_mk, stcb->asoc.sent_queue_retran_cnt, num_mk);
774#endif
775#ifndef SCTP_AUDITING_ENABLED
776		stcb->asoc.sent_queue_retran_cnt = cnt_mk;
777#endif
778	}
779	if (audit_tf) {
780		SCTPDBG(SCTP_DEBUG_TIMER4,
781		    "Audit total flight due to negative value net:%p\n",
782		    (void *)net);
783		stcb->asoc.total_flight = 0;
784		stcb->asoc.total_flight_count = 0;
785		/* Clear all networks flight size */
786		TAILQ_FOREACH(lnets, &stcb->asoc.nets, sctp_next) {
787			lnets->flight_size = 0;
788			SCTPDBG(SCTP_DEBUG_TIMER4,
789			    "Net:%p c-f cwnd:%d ssthresh:%d\n",
790			    (void *)lnets, lnets->cwnd, lnets->ssthresh);
791		}
792		TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
793			if (chk->sent < SCTP_DATAGRAM_RESEND) {
794				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
795					sctp_misc_ints(SCTP_FLIGHT_LOG_UP,
796					    chk->whoTo->flight_size,
797					    chk->book_size,
798					    (uint32_t)(uintptr_t)chk->whoTo,
799					    chk->rec.data.tsn);
800				}
801
802				sctp_flight_size_increase(chk);
803				sctp_total_flight_increase(stcb, chk);
804			}
805		}
806	}
807	/* We return 1 if we only have a window probe outstanding */
808	return (0);
809}
810
811int
812sctp_t3rxt_timer(struct sctp_inpcb *inp,
813    struct sctp_tcb *stcb,
814    struct sctp_nets *net)
815{
816	struct sctp_nets *alt;
817	int win_probe, num_mk, num_abandoned;
818
819	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FR_LOGGING_ENABLE) {
820		sctp_log_fr(0, 0, 0, SCTP_FR_T3_TIMEOUT);
821	}
822	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
823		struct sctp_nets *lnet;
824
825		TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
826			if (net == lnet) {
827				sctp_log_cwnd(stcb, lnet, 1, SCTP_CWND_LOG_FROM_T3);
828			} else {
829				sctp_log_cwnd(stcb, lnet, 0, SCTP_CWND_LOG_FROM_T3);
830			}
831		}
832	}
833	/* Find an alternate and mark those for retransmission */
834	if ((stcb->asoc.peers_rwnd == 0) &&
835	    (stcb->asoc.total_flight < net->mtu)) {
836		SCTP_STAT_INCR(sctps_timowindowprobe);
837		win_probe = 1;
838	} else {
839		win_probe = 0;
840	}
841
842	if (win_probe == 0) {
843		/* We don't do normal threshold management on window probes */
844		if (sctp_threshold_management(inp, stcb, net,
845		    stcb->asoc.max_send_times)) {
846			/* Association was destroyed */
847			return (1);
848		} else {
849			if (net != stcb->asoc.primary_destination) {
850				/* send a immediate HB if our RTO is stale */
851				struct timeval now;
852				uint32_t ms_goneby;
853
854				(void)SCTP_GETTIME_TIMEVAL(&now);
855				if (net->last_sent_time.tv_sec) {
856					ms_goneby = (uint32_t)(now.tv_sec - net->last_sent_time.tv_sec) * 1000;
857				} else {
858					ms_goneby = 0;
859				}
860				if ((net->dest_state & SCTP_ADDR_PF) == 0) {
861					if ((ms_goneby > net->RTO) || (net->RTO == 0)) {
862						/*
863						 * no recent feed back in an
864						 * RTO or more, request a
865						 * RTT update
866						 */
867						sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
868					}
869				}
870			}
871		}
872	} else {
873		/*
874		 * For a window probe we don't penalize the net's but only
875		 * the association. This may fail it if SACKs are not coming
876		 * back. If sack's are coming with rwnd locked at 0, we will
877		 * continue to hold things waiting for rwnd to raise
878		 */
879		if (sctp_threshold_management(inp, stcb, NULL,
880		    stcb->asoc.max_send_times)) {
881			/* Association was destroyed */
882			return (1);
883		}
884	}
885	if (stcb->asoc.sctp_cmt_on_off > 0) {
886		if (net->pf_threshold < net->failure_threshold) {
887			alt = sctp_find_alternate_net(stcb, net, 2);
888		} else {
889			/*
890			 * CMT: Using RTX_SSTHRESH policy for CMT. If CMT is
891			 * being used, then pick dest with largest ssthresh
892			 * for any retransmission.
893			 */
894			alt = sctp_find_alternate_net(stcb, net, 1);
895			/*
896			 * CUCv2: If a different dest is picked for the
897			 * retransmission, then new (rtx-)pseudo_cumack
898			 * needs to be tracked for orig dest. Let CUCv2
899			 * track new (rtx-) pseudo-cumack always.
900			 */
901			net->find_pseudo_cumack = 1;
902			net->find_rtx_pseudo_cumack = 1;
903		}
904	} else {
905		alt = sctp_find_alternate_net(stcb, net, 0);
906	}
907
908	num_mk = 0;
909	num_abandoned = 0;
910	(void)sctp_mark_all_for_resend(stcb, net, alt, win_probe,
911	    &num_mk, &num_abandoned);
912	/* FR Loss recovery just ended with the T3. */
913	stcb->asoc.fast_retran_loss_recovery = 0;
914
915	/* CMT FR loss recovery ended with the T3 */
916	net->fast_retran_loss_recovery = 0;
917	if ((stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) &&
918	    (net->flight_size == 0)) {
919		(*stcb->asoc.cc_functions.sctp_cwnd_new_transmission_begins) (stcb, net);
920	}
921
922	/*
923	 * setup the sat loss recovery that prevents satellite cwnd advance.
924	 */
925	stcb->asoc.sat_t3_loss_recovery = 1;
926	stcb->asoc.sat_t3_recovery_tsn = stcb->asoc.sending_seq;
927
928	/* Backoff the timer and cwnd */
929	sctp_backoff_on_timeout(stcb, net, win_probe, num_mk, num_abandoned);
930	if (((net->dest_state & SCTP_ADDR_REACHABLE) == 0) ||
931	    (net->dest_state & SCTP_ADDR_PF)) {
932		/* Move all pending over too */
933		sctp_move_chunks_from_net(stcb, net);
934
935		/*
936		 * Get the address that failed, to force a new src address
937		 * selection and a route allocation.
938		 */
939		if (net->ro._s_addr != NULL) {
940			sctp_free_ifa(net->ro._s_addr);
941			net->ro._s_addr = NULL;
942		}
943		net->src_addr_selected = 0;
944
945		/* Force a route allocation too */
946		RO_NHFREE(&net->ro);
947
948		/* Was it our primary? */
949		if ((stcb->asoc.primary_destination == net) && (alt != net)) {
950			/*
951			 * Yes, note it as such and find an alternate note:
952			 * this means HB code must use this to resent the
953			 * primary if it goes active AND if someone does a
954			 * change-primary then this flag must be cleared
955			 * from any net structures.
956			 */
957			if (stcb->asoc.alternate != NULL) {
958				sctp_free_remote_addr(stcb->asoc.alternate);
959			}
960			stcb->asoc.alternate = alt;
961			atomic_add_int(&stcb->asoc.alternate->ref_count, 1);
962		}
963	}
964	/*
965	 * Special case for cookie-echo'ed case, we don't do output but must
966	 * await the COOKIE-ACK before retransmission
967	 */
968	if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) {
969		/*
970		 * Here we just reset the timer and start again since we
971		 * have not established the asoc
972		 */
973		sctp_timer_start(SCTP_TIMER_TYPE_SEND, inp, stcb, net);
974		return (0);
975	}
976	if (stcb->asoc.prsctp_supported) {
977		struct sctp_tmit_chunk *lchk;
978
979		lchk = sctp_try_advance_peer_ack_point(stcb, &stcb->asoc);
980		/* C3. See if we need to send a Fwd-TSN */
981		if (SCTP_TSN_GT(stcb->asoc.advanced_peer_ack_point, stcb->asoc.last_acked_seq)) {
982			send_forward_tsn(stcb, &stcb->asoc);
983			for (; lchk != NULL; lchk = TAILQ_NEXT(lchk, sctp_next)) {
984				if (lchk->whoTo != NULL) {
985					break;
986				}
987			}
988			if (lchk != NULL) {
989				/* Assure a timer is up */
990				sctp_timer_start(SCTP_TIMER_TYPE_SEND, stcb->sctp_ep, stcb, lchk->whoTo);
991			}
992		}
993	}
994	if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
995		sctp_log_cwnd(stcb, net, net->cwnd, SCTP_CWND_LOG_FROM_RTX);
996	}
997	return (0);
998}
999
1000int
1001sctp_t1init_timer(struct sctp_inpcb *inp,
1002    struct sctp_tcb *stcb,
1003    struct sctp_nets *net)
1004{
1005	/* bump the thresholds */
1006	if (stcb->asoc.delayed_connection) {
1007		/*
1008		 * special hook for delayed connection. The library did NOT
1009		 * complete the rest of its sends.
1010		 */
1011		stcb->asoc.delayed_connection = 0;
1012		sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1013		return (0);
1014	}
1015	if (SCTP_GET_STATE(stcb) != SCTP_STATE_COOKIE_WAIT) {
1016		return (0);
1017	}
1018	if (sctp_threshold_management(inp, stcb, net,
1019	    stcb->asoc.max_init_times)) {
1020		/* Association was destroyed */
1021		return (1);
1022	}
1023	stcb->asoc.dropped_special_cnt = 0;
1024	sctp_backoff_on_timeout(stcb, stcb->asoc.primary_destination, 1, 0, 0);
1025	if (stcb->asoc.initial_init_rto_max < net->RTO) {
1026		net->RTO = stcb->asoc.initial_init_rto_max;
1027	}
1028	if (stcb->asoc.numnets > 1) {
1029		/* If we have more than one addr use it */
1030		struct sctp_nets *alt;
1031
1032		alt = sctp_find_alternate_net(stcb, stcb->asoc.primary_destination, 0);
1033		if (alt != stcb->asoc.primary_destination) {
1034			sctp_move_chunks_from_net(stcb, stcb->asoc.primary_destination);
1035			stcb->asoc.primary_destination = alt;
1036		}
1037	}
1038	/* Send out a new init */
1039	sctp_send_initiate(inp, stcb, SCTP_SO_NOT_LOCKED);
1040	return (0);
1041}
1042
1043/*
1044 * For cookie and asconf we actually need to find and mark for resend, then
1045 * increment the resend counter (after all the threshold management stuff of
1046 * course).
1047 */
1048int
1049sctp_cookie_timer(struct sctp_inpcb *inp,
1050    struct sctp_tcb *stcb,
1051    struct sctp_nets *net SCTP_UNUSED)
1052{
1053	struct sctp_nets *alt;
1054	struct sctp_tmit_chunk *cookie;
1055
1056	/* first before all else we must find the cookie */
1057	TAILQ_FOREACH(cookie, &stcb->asoc.control_send_queue, sctp_next) {
1058		if (cookie->rec.chunk_id.id == SCTP_COOKIE_ECHO) {
1059			break;
1060		}
1061	}
1062	if (cookie == NULL) {
1063		if (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED) {
1064			/* FOOBAR! */
1065			struct mbuf *op_err;
1066
1067			op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
1068			    "Cookie timer expired, but no cookie");
1069			inp->last_abort_code = SCTP_FROM_SCTP_TIMER + SCTP_LOC_3;
1070			sctp_abort_an_association(inp, stcb, op_err, false, 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));
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 threshold management, now lets backoff the address and
1089	 * select 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	cookie->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1105	/*
1106	 * Now call the output routine to kick out the cookie again, Note we
1107	 * don't mark any chunks for retran so that FR will need to kick in
1108	 * to move these (or a send timer).
1109	 */
1110	return (0);
1111}
1112
1113int
1114sctp_strreset_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
1115{
1116	struct sctp_nets *alt, *net;
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	net = strrst->whoTo;
1128	/* do threshold management */
1129	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1130		/* Assoc is over */
1131		return (1);
1132	}
1133	/*
1134	 * Cleared threshold management, now lets backoff the address and
1135	 * select an alternate
1136	 */
1137	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1138	alt = sctp_find_alternate_net(stcb, net, 0);
1139	strrst->whoTo = alt;
1140	atomic_add_int(&alt->ref_count, 1);
1141
1142	/* See if a ECN Echo is also stranded */
1143	TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1144		if ((chk->whoTo == net) &&
1145		    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1146			sctp_free_remote_addr(chk->whoTo);
1147			if (chk->sent != SCTP_DATAGRAM_RESEND) {
1148				chk->sent = SCTP_DATAGRAM_RESEND;
1149				chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
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) == 0) {
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	sctp_free_remote_addr(net);
1164
1165	/* mark the retran info */
1166	if (strrst->sent != SCTP_DATAGRAM_RESEND)
1167		sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1168	strrst->sent = SCTP_DATAGRAM_RESEND;
1169	strrst->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1170
1171	/* restart the timer */
1172	sctp_timer_start(SCTP_TIMER_TYPE_STRRESET, inp, stcb, alt);
1173	return (0);
1174}
1175
1176int
1177sctp_asconf_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1178    struct sctp_nets *net)
1179{
1180	struct sctp_nets *alt;
1181	struct sctp_tmit_chunk *asconf, *chk;
1182
1183	/* is this a first send, or a retransmission? */
1184	if (TAILQ_EMPTY(&stcb->asoc.asconf_send_queue)) {
1185		/* compose a new ASCONF chunk and send it */
1186		sctp_send_asconf(stcb, net, SCTP_ADDR_NOT_LOCKED);
1187	} else {
1188		/*
1189		 * Retransmission of the existing ASCONF is needed
1190		 */
1191
1192		/* find the existing ASCONF */
1193		asconf = TAILQ_FIRST(&stcb->asoc.asconf_send_queue);
1194		if (asconf == NULL) {
1195			return (0);
1196		}
1197		net = asconf->whoTo;
1198		/* do threshold management */
1199		if (sctp_threshold_management(inp, stcb, net,
1200		    stcb->asoc.max_send_times)) {
1201			/* Assoc is over */
1202			return (1);
1203		}
1204		if (asconf->snd_count > stcb->asoc.max_send_times) {
1205			/*
1206			 * Something is rotten: our peer is not responding
1207			 * to ASCONFs but apparently is to other chunks.
1208			 * i.e. it is not properly handling the chunk type
1209			 * upper bits. Mark this peer as ASCONF incapable
1210			 * and cleanup.
1211			 */
1212			SCTPDBG(SCTP_DEBUG_TIMER1, "asconf_timer: Peer has not responded to our repeated ASCONFs\n");
1213			sctp_asconf_cleanup(stcb);
1214			return (0);
1215		}
1216		/*
1217		 * cleared threshold management, so now backoff the net and
1218		 * select an alternate
1219		 */
1220		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1221		alt = sctp_find_alternate_net(stcb, net, 0);
1222		if (asconf->whoTo != alt) {
1223			asconf->whoTo = alt;
1224			atomic_add_int(&alt->ref_count, 1);
1225		}
1226
1227		/* See if an ECN Echo is also stranded */
1228		TAILQ_FOREACH(chk, &stcb->asoc.control_send_queue, sctp_next) {
1229			if ((chk->whoTo == net) &&
1230			    (chk->rec.chunk_id.id == SCTP_ECN_ECHO)) {
1231				sctp_free_remote_addr(chk->whoTo);
1232				chk->whoTo = alt;
1233				if (chk->sent != SCTP_DATAGRAM_RESEND) {
1234					chk->sent = SCTP_DATAGRAM_RESEND;
1235					chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1236					sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1237				}
1238				atomic_add_int(&alt->ref_count, 1);
1239			}
1240		}
1241		TAILQ_FOREACH(chk, &stcb->asoc.asconf_send_queue, sctp_next) {
1242			if (chk->whoTo != alt) {
1243				sctp_free_remote_addr(chk->whoTo);
1244				chk->whoTo = alt;
1245				atomic_add_int(&alt->ref_count, 1);
1246			}
1247			if (asconf->sent != SCTP_DATAGRAM_RESEND && chk->sent != SCTP_DATAGRAM_UNSENT)
1248				sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1249			chk->sent = SCTP_DATAGRAM_RESEND;
1250			chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1251		}
1252		if ((net->dest_state & SCTP_ADDR_REACHABLE) == 0) {
1253			/*
1254			 * If the address went un-reachable, we need to move
1255			 * to the alternate for ALL chunks in queue
1256			 */
1257			sctp_move_chunks_from_net(stcb, net);
1258		}
1259		sctp_free_remote_addr(net);
1260
1261		/* mark the retran info */
1262		if (asconf->sent != SCTP_DATAGRAM_RESEND)
1263			sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
1264		asconf->sent = SCTP_DATAGRAM_RESEND;
1265		asconf->flags |= CHUNK_FLAGS_FRAGMENT_OK;
1266
1267		/* send another ASCONF if any and we can do */
1268		sctp_send_asconf(stcb, alt, SCTP_ADDR_NOT_LOCKED);
1269	}
1270	return (0);
1271}
1272
1273/* Mobility adaptation */
1274void
1275sctp_delete_prim_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
1276{
1277	if (stcb->asoc.deleted_primary == NULL) {
1278		SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: deleted_primary is not stored...\n");
1279		sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1280		return;
1281	}
1282	SCTPDBG(SCTP_DEBUG_ASCONF1, "delete_prim_timer: finished to keep deleted primary ");
1283	SCTPDBG_ADDR(SCTP_DEBUG_ASCONF1, &stcb->asoc.deleted_primary->ro._l_addr.sa);
1284	sctp_free_remote_addr(stcb->asoc.deleted_primary);
1285	stcb->asoc.deleted_primary = NULL;
1286	sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
1287	return;
1288}
1289
1290/*
1291 * For the shutdown and shutdown-ack, we do not keep one around on the
1292 * control queue. This means we must generate a new one and call the general
1293 * chunk output routine, AFTER having done threshold management.
1294 * It is assumed that net is non-NULL.
1295 */
1296int
1297sctp_shutdown_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1298    struct sctp_nets *net)
1299{
1300	struct sctp_nets *alt;
1301
1302	/* first threshold management */
1303	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1304		/* Assoc is over */
1305		return (1);
1306	}
1307	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1308	/* second select an alternative */
1309	alt = sctp_find_alternate_net(stcb, net, 0);
1310
1311	/* third generate a shutdown into the queue for out net */
1312	sctp_send_shutdown(stcb, alt);
1313
1314	/* fourth restart timer */
1315	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, inp, stcb, alt);
1316	return (0);
1317}
1318
1319int
1320sctp_shutdownack_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1321    struct sctp_nets *net)
1322{
1323	struct sctp_nets *alt;
1324
1325	/* first threshold management */
1326	if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1327		/* Assoc is over */
1328		return (1);
1329	}
1330	sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1331	/* second select an alternative */
1332	alt = sctp_find_alternate_net(stcb, net, 0);
1333
1334	/* third generate a shutdown into the queue for out net */
1335	sctp_send_shutdown_ack(stcb, alt);
1336
1337	/* fourth restart timer */
1338	sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNACK, inp, stcb, alt);
1339	return (0);
1340}
1341
1342static void
1343sctp_audit_stream_queues_for_size(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
1344{
1345	struct sctp_stream_queue_pending *sp;
1346	unsigned int i, chks_in_queue = 0;
1347	int being_filled = 0;
1348
1349	KASSERT(inp != NULL, ("inp is NULL"));
1350	KASSERT(stcb != NULL, ("stcb is NULL"));
1351	SCTP_TCB_LOCK_ASSERT(stcb);
1352	KASSERT(TAILQ_EMPTY(&stcb->asoc.send_queue), ("send_queue not empty"));
1353	KASSERT(TAILQ_EMPTY(&stcb->asoc.sent_queue), ("sent_queue not empty"));
1354
1355	if (stcb->asoc.sent_queue_retran_cnt) {
1356		SCTP_PRINTF("Hmm, sent_queue_retran_cnt is non-zero %d\n",
1357		    stcb->asoc.sent_queue_retran_cnt);
1358		stcb->asoc.sent_queue_retran_cnt = 0;
1359	}
1360	if (stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1361		/* No stream scheduler information, initialize scheduler */
1362		stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc);
1363		if (!stcb->asoc.ss_functions.sctp_ss_is_empty(stcb, &stcb->asoc)) {
1364			/* yep, we lost a stream or two */
1365			SCTP_PRINTF("Found additional streams NOT managed by scheduler, corrected\n");
1366		} else {
1367			/* no streams lost */
1368			stcb->asoc.total_output_queue_size = 0;
1369		}
1370	}
1371	/* Check to see if some data queued, if so report it */
1372	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
1373		if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) {
1374			TAILQ_FOREACH(sp, &stcb->asoc.strmout[i].outqueue, next) {
1375				if (sp->msg_is_complete)
1376					being_filled++;
1377				chks_in_queue++;
1378			}
1379		}
1380	}
1381	if (chks_in_queue != stcb->asoc.stream_queue_cnt) {
1382		SCTP_PRINTF("Hmm, stream queue cnt at %d I counted %d in stream out wheel\n",
1383		    stcb->asoc.stream_queue_cnt, chks_in_queue);
1384	}
1385	if (chks_in_queue) {
1386		/* call the output queue function */
1387		sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_NOT_LOCKED);
1388		if ((TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1389		    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1390			/*
1391			 * Probably should go in and make it go back through
1392			 * and add fragments allowed
1393			 */
1394			if (being_filled == 0) {
1395				SCTP_PRINTF("Still nothing moved %d chunks are stuck\n",
1396				    chks_in_queue);
1397			}
1398		}
1399	} else {
1400		SCTP_PRINTF("Found no chunks on any queue tot:%lu\n",
1401		    (u_long)stcb->asoc.total_output_queue_size);
1402		stcb->asoc.total_output_queue_size = 0;
1403	}
1404}
1405
1406int
1407sctp_heartbeat_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb,
1408    struct sctp_nets *net)
1409{
1410	bool net_was_pf;
1411
1412	net_was_pf = (net->dest_state & SCTP_ADDR_PF) != 0;
1413	if (net->hb_responded == 0) {
1414		if (net->ro._s_addr != NULL) {
1415			/*
1416			 * Invalidate the src address if we did not get a
1417			 * response last time.
1418			 */
1419			sctp_free_ifa(net->ro._s_addr);
1420			net->ro._s_addr = NULL;
1421			net->src_addr_selected = 0;
1422		}
1423		sctp_backoff_on_timeout(stcb, net, 1, 0, 0);
1424		if (sctp_threshold_management(inp, stcb, net, stcb->asoc.max_send_times)) {
1425			/* Assoc is over */
1426			return (1);
1427		}
1428	}
1429	/* Zero PBA, if it needs it */
1430	if (net->partial_bytes_acked > 0) {
1431		net->partial_bytes_acked = 0;
1432	}
1433	if ((stcb->asoc.total_output_queue_size > 0) &&
1434	    (TAILQ_EMPTY(&stcb->asoc.send_queue)) &&
1435	    (TAILQ_EMPTY(&stcb->asoc.sent_queue))) {
1436		sctp_audit_stream_queues_for_size(inp, stcb);
1437	}
1438	if ((((net->dest_state & SCTP_ADDR_NOHB) == 0) ||
1439	    (net->dest_state & SCTP_ADDR_UNCONFIRMED)) &&
1440	    (net_was_pf || ((net->dest_state & SCTP_ADDR_PF) == 0))) {
1441		/*
1442		 * When moving to PF during threshold management, a HB has
1443		 * been queued in that routine.
1444		 */
1445		uint32_t ms_gone_by;
1446
1447		if ((net->last_sent_time.tv_sec > 0) ||
1448		    (net->last_sent_time.tv_usec > 0)) {
1449			struct timeval diff;
1450
1451			SCTP_GETTIME_TIMEVAL(&diff);
1452			timevalsub(&diff, &net->last_sent_time);
1453			ms_gone_by = (uint32_t)(diff.tv_sec * 1000) +
1454			    (uint32_t)(diff.tv_usec / 1000);
1455		} else {
1456			ms_gone_by = 0xffffffff;
1457		}
1458		if ((ms_gone_by >= net->heart_beat_delay) ||
1459		    (net->dest_state & SCTP_ADDR_UNCONFIRMED) ||
1460		    (net->dest_state & SCTP_ADDR_PF)) {
1461			sctp_send_hb(stcb, net, SCTP_SO_NOT_LOCKED);
1462		}
1463	}
1464	return (0);
1465}
1466
1467void
1468sctp_pathmtu_timer(struct sctp_inpcb *inp,
1469    struct sctp_tcb *stcb,
1470    struct sctp_nets *net)
1471{
1472	uint32_t next_mtu, mtu;
1473
1474	next_mtu = sctp_get_next_mtu(net->mtu);
1475
1476	if ((next_mtu > net->mtu) && (net->port == 0)) {
1477		if ((net->src_addr_selected == 0) ||
1478		    (net->ro._s_addr == NULL) ||
1479		    (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1480			if ((net->ro._s_addr != NULL) && (net->ro._s_addr->localifa_flags & SCTP_BEING_DELETED)) {
1481				sctp_free_ifa(net->ro._s_addr);
1482				net->ro._s_addr = NULL;
1483				net->src_addr_selected = 0;
1484			} else if (net->ro._s_addr == NULL) {
1485#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1486				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1487					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1488
1489					/* KAME hack: embed scopeid */
1490					(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
1491				}
1492#endif
1493
1494				net->ro._s_addr = sctp_source_address_selection(inp,
1495				    stcb,
1496				    (sctp_route_t *)&net->ro,
1497				    net, 0, stcb->asoc.vrf_id);
1498#if defined(INET6) && defined(SCTP_EMBEDDED_V6_SCOPE)
1499				if (net->ro._l_addr.sa.sa_family == AF_INET6) {
1500					struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1501
1502					(void)sa6_recoverscope(sin6);
1503				}
1504#endif				/* INET6 */
1505			}
1506			if (net->ro._s_addr)
1507				net->src_addr_selected = 1;
1508		}
1509		if (net->ro._s_addr) {
1510			mtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._s_addr.sa, net->ro.ro_nh);
1511#if defined(INET) || defined(INET6)
1512			if (net->port) {
1513				mtu -= sizeof(struct udphdr);
1514			}
1515#endif
1516			if (mtu > next_mtu) {
1517				net->mtu = next_mtu;
1518			} else {
1519				net->mtu = mtu;
1520			}
1521		}
1522	}
1523	/* restart the timer */
1524	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
1525}
1526
1527void
1528sctp_autoclose_timer(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
1529{
1530	struct timeval tn, *tim_touse;
1531	struct sctp_association *asoc;
1532	uint32_t ticks_gone_by;
1533
1534	(void)SCTP_GETTIME_TIMEVAL(&tn);
1535	if (stcb->asoc.sctp_autoclose_ticks > 0 &&
1536	    sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) {
1537		/* Auto close is on */
1538		asoc = &stcb->asoc;
1539		/* pick the time to use */
1540		if (asoc->time_last_rcvd.tv_sec >
1541		    asoc->time_last_sent.tv_sec) {
1542			tim_touse = &asoc->time_last_rcvd;
1543		} else {
1544			tim_touse = &asoc->time_last_sent;
1545		}
1546		/* Now has long enough transpired to autoclose? */
1547		ticks_gone_by = sctp_secs_to_ticks((uint32_t)(tn.tv_sec - tim_touse->tv_sec));
1548		if (ticks_gone_by >= asoc->sctp_autoclose_ticks) {
1549			/*
1550			 * autoclose time has hit, call the output routine,
1551			 * which should do nothing just to be SURE we don't
1552			 * have hanging data. We can then safely check the
1553			 * queues and know that we are clear to send
1554			 * shutdown
1555			 */
1556			sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_AUTOCLOSE_TMR, SCTP_SO_NOT_LOCKED);
1557			/* Are we clean? */
1558			if (TAILQ_EMPTY(&asoc->send_queue) &&
1559			    TAILQ_EMPTY(&asoc->sent_queue)) {
1560				/*
1561				 * there is nothing queued to send, so I'm
1562				 * done...
1563				 */
1564				if (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) {
1565					/* only send SHUTDOWN 1st time thru */
1566					struct sctp_nets *net;
1567
1568					if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) ||
1569					    (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
1570						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
1571					}
1572					SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT);
1573					sctp_stop_timers_for_shutdown(stcb);
1574					if (stcb->asoc.alternate) {
1575						net = stcb->asoc.alternate;
1576					} else {
1577						net = stcb->asoc.primary_destination;
1578					}
1579					sctp_send_shutdown(stcb, net);
1580					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
1581					    stcb->sctp_ep, stcb, net);
1582					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
1583					    stcb->sctp_ep, stcb, NULL);
1584				}
1585			}
1586		} else {
1587			/*
1588			 * No auto close at this time, reset t-o to check
1589			 * later
1590			 */
1591			uint32_t 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, NULL);
1597			/* restore the real tick value */
1598			asoc->sctp_autoclose_ticks = tmp;
1599		}
1600	}
1601}
1602