sctp_pcb.c revision 170786
1/*-
2 * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
3 *
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions are met:
6 *
7 * a) Redistributions of source code must retain the above copyright notice,
8 *   this list of conditions and the following disclaimer.
9 *
10 * b) Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in
12 *   the documentation and/or other materials provided with the distribution.
13 *
14 * c) Neither the name of Cisco Systems, Inc. nor the names of its
15 *    contributors may be used to endorse or promote products derived
16 *    from this software without specific prior written permission.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
20 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
22 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
28 * THE POSSIBILITY OF SUCH DAMAGE.
29 */
30
31/* $KAME: sctp_pcb.c,v 1.38 2005/03/06 16:04:18 itojun Exp $	 */
32
33#include <sys/cdefs.h>
34__FBSDID("$FreeBSD: head/sys/netinet/sctp_pcb.c 170786 2007-06-15 19:28:58Z rrs $");
35
36#include <netinet/sctp_os.h>
37#include <sys/proc.h>
38#include <netinet/sctp_var.h>
39#include <netinet/sctp_sysctl.h>
40#include <netinet/sctp_pcb.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp.h>
43#include <netinet/sctp_header.h>
44#include <netinet/sctp_asconf.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_timer.h>
47#include <netinet/sctp_bsd_addr.h>
48
49
50struct sctp_epinfo sctppcbinfo;
51
52/* FIX: we don't handle multiple link local scopes */
53/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
54int
55SCTP6_ARE_ADDR_EQUAL(struct in6_addr *a, struct in6_addr *b)
56{
57	struct in6_addr tmp_a, tmp_b;
58
59	/* use a copy of a and b */
60	tmp_a = *a;
61	tmp_b = *b;
62	in6_clearscope(&tmp_a);
63	in6_clearscope(&tmp_b);
64	return (IN6_ARE_ADDR_EQUAL(&tmp_a, &tmp_b));
65}
66
67void
68sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
69{
70	/*
71	 * We really don't need to lock this, but I will just because it
72	 * does not hurt.
73	 */
74	SCTP_INP_INFO_RLOCK();
75	spcb->ep_count = sctppcbinfo.ipi_count_ep;
76	spcb->asoc_count = sctppcbinfo.ipi_count_asoc;
77	spcb->laddr_count = sctppcbinfo.ipi_count_laddr;
78	spcb->raddr_count = sctppcbinfo.ipi_count_raddr;
79	spcb->chk_count = sctppcbinfo.ipi_count_chunk;
80	spcb->readq_count = sctppcbinfo.ipi_count_readq;
81	spcb->stream_oque = sctppcbinfo.ipi_count_strmoq;
82	spcb->free_chunks = sctppcbinfo.ipi_free_chunks;
83
84	SCTP_INP_INFO_RUNLOCK();
85}
86
87/*
88 * Addresses are added to VRF's (Virtual Router's). For BSD we
89 * have only the default VRF 0. We maintain a hash list of
90 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
91 * these has a list of addresses. When we add a new address
92 * to a VRF we lookup the ifn/ifn_index, if the ifn does
93 * not exist we create it and add it to the list of IFN's
94 * within the VRF. Once we have the sctp_ifn, we add the
95 * address to the list. So we look something like:
96 *
97 * hash-vrf-table
98 *   vrf-> ifn-> ifn -> ifn
99 *   vrf    |
100 *    ...   +--ifa-> ifa -> ifa
101 *   vrf
102 *
103 * We keep these seperate lists since the SCTP subsystem will
104 * point to these from its source address selection nets structure.
105 * When an address is deleted it does not happen right away on
106 * the SCTP side, it gets scheduled. What we do when a
107 * delete happens is immediately remove the address from
108 * the master list and decrement the refcount. As our
109 * addip iterator works through and frees the src address
110 * selection pointing to the sctp_ifa, eventually the refcount
111 * will reach 0 and we will delete it. Note that it is assumed
112 * that any locking on system level ifn/ifa is done at the
113 * caller of these functions and these routines will only
114 * lock the SCTP structures as they add or delete things.
115 *
116 * Other notes on VRF concepts.
117 *  - An endpoint can be in multiple VRF's
118 *  - An association lives within a VRF and only one VRF.
119 *  - Any incoming packet we can deduce the VRF for by
120 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
121 *  - Any downward send call or connect call must supply the
122 *    VRF via ancillary data or via some sort of set default
123 *    VRF socket option call (again for BSD no brainer since
124 *    the VRF is always 0).
125 *  - An endpoint may add multiple VRF's to it.
126 *  - Listening sockets can accept associations in any
127 *    of the VRF's they are in but the assoc will end up
128 *    in only one VRF (gotten from the packet or connect/send).
129 *
130 */
131
132struct sctp_vrf *
133sctp_allocate_vrf(int vrf_id)
134{
135	struct sctp_vrf *vrf = NULL;
136	struct sctp_vrflist *bucket;
137
138	/* First allocate the VRF structure */
139	vrf = sctp_find_vrf(vrf_id);
140	if (vrf) {
141		/* Already allocated */
142		return (vrf);
143	}
144	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
145	    SCTP_M_VRF);
146	if (vrf == NULL) {
147		/* No memory */
148#ifdef INVARIANTS
149		panic("No memory for VRF:%d", vrf_id);
150#endif
151		return (NULL);
152	}
153	/* setup the VRF */
154	memset(vrf, 0, sizeof(struct sctp_vrf));
155	vrf->vrf_id = vrf_id;
156	LIST_INIT(&vrf->ifnlist);
157	vrf->total_ifa_count = 0;
158	vrf->refcount = 0;
159	/* now also setup table ids */
160	SCTP_INIT_VRF_TABLEID(vrf);
161	/* Init the HASH of addresses */
162	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
163	    &vrf->vrf_addr_hashmark);
164	if (vrf->vrf_addr_hash == NULL) {
165		/* No memory */
166#ifdef INVARIANTS
167		panic("No memory for VRF:%d", vrf_id);
168#endif
169		SCTP_FREE(vrf, SCTP_M_VRF);
170		return (NULL);
171	}
172	/* Add it to the hash table */
173	bucket = &sctppcbinfo.sctp_vrfhash[(vrf_id & sctppcbinfo.hashvrfmark)];
174	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
175	atomic_add_int(&sctppcbinfo.ipi_count_vrfs, 1);
176	return (vrf);
177}
178
179
180struct sctp_ifn *
181sctp_find_ifn(void *ifn, uint32_t ifn_index)
182{
183	struct sctp_ifn *sctp_ifnp;
184	struct sctp_ifnlist *hash_ifn_head;
185
186	/*
187	 * We assume the lock is held for the addresses if thats wrong
188	 * problems could occur :-)
189	 */
190	hash_ifn_head = &sctppcbinfo.vrf_ifn_hash[(ifn_index & sctppcbinfo.vrf_ifn_hashmark)];
191	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
192		if (sctp_ifnp->ifn_index == ifn_index) {
193			return (sctp_ifnp);
194		}
195		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
196			return (sctp_ifnp);
197		}
198	}
199	return (NULL);
200}
201
202
203
204struct sctp_vrf *
205sctp_find_vrf(uint32_t vrf_id)
206{
207	struct sctp_vrflist *bucket;
208	struct sctp_vrf *liste;
209
210	bucket = &sctppcbinfo.sctp_vrfhash[(vrf_id & sctppcbinfo.hashvrfmark)];
211	LIST_FOREACH(liste, bucket, next_vrf) {
212		if (vrf_id == liste->vrf_id) {
213			return (liste);
214		}
215	}
216	return (NULL);
217}
218
219void
220sctp_free_vrf(struct sctp_vrf *vrf)
221{
222	int ret;
223
224	ret = atomic_fetchadd_int(&vrf->refcount, -1);
225	if (ret == 1) {
226		/* We zero'd the count */
227		LIST_REMOVE(vrf, next_vrf);
228		SCTP_FREE(vrf, SCTP_M_VRF);
229		atomic_subtract_int(&sctppcbinfo.ipi_count_vrfs, 1);
230	}
231}
232
233void
234sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
235{
236	int ret;
237
238	ret = atomic_fetchadd_int(&sctp_ifnp->refcount, -1);
239	if (ret == 1) {
240		/* We zero'd the count */
241		if (sctp_ifnp->vrf) {
242			sctp_free_vrf(sctp_ifnp->vrf);
243		}
244		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
245		atomic_subtract_int(&sctppcbinfo.ipi_count_ifns, 1);
246	}
247}
248
249void
250sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
251{
252	struct sctp_ifn *sctp_ifnp;
253
254	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
255	if (sctp_ifnp != NULL) {
256		sctp_ifnp->ifn_mtu = mtu;
257	}
258}
259
260
261void
262sctp_free_ifa(struct sctp_ifa *sctp_ifap)
263{
264	int ret;
265
266	ret = atomic_fetchadd_int(&sctp_ifap->refcount, -1);
267	if (ret == 1) {
268		/* We zero'd the count */
269		if (sctp_ifap->ifn_p) {
270			sctp_free_ifn(sctp_ifap->ifn_p);
271		}
272		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
273		atomic_subtract_int(&sctppcbinfo.ipi_count_ifas, 1);
274	}
275}
276
277static void
278sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
279{
280	struct sctp_ifn *found;
281
282	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
283	if (found == NULL) {
284		/* Not in the list.. sorry */
285		return;
286	}
287	if (hold_addr_lock == 0)
288		SCTP_IPI_ADDR_LOCK();
289	LIST_REMOVE(sctp_ifnp, next_bucket);
290	LIST_REMOVE(sctp_ifnp, next_ifn);
291	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
292	    sctp_ifnp->registered_af);
293	if (hold_addr_lock == 0)
294		SCTP_IPI_ADDR_UNLOCK();
295	/* Take away the reference, and possibly free it */
296	sctp_free_ifn(sctp_ifnp);
297}
298
299
300struct sctp_ifa *
301sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
302    uint32_t ifn_type, const char *if_name,
303    void *ifa, struct sockaddr *addr, uint32_t ifa_flags,
304    int dynamic_add)
305{
306	struct sctp_vrf *vrf;
307	struct sctp_ifn *sctp_ifnp = NULL;
308	struct sctp_ifa *sctp_ifap = NULL;
309	struct sctp_ifalist *hash_addr_head;
310	struct sctp_ifnlist *hash_ifn_head;
311	uint32_t hash_of_addr;
312	int new_ifn_af = 0;
313
314	/* How granular do we need the locks to be here? */
315	SCTP_IPI_ADDR_LOCK();
316	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
317	if (sctp_ifnp) {
318		vrf = sctp_ifnp->vrf;
319	} else {
320		vrf = sctp_find_vrf(vrf_id);
321		if (vrf == NULL) {
322			vrf = sctp_allocate_vrf(vrf_id);
323			if (vrf == NULL) {
324				SCTP_IPI_ADDR_UNLOCK();
325				return (NULL);
326			}
327		}
328	}
329	if (sctp_ifnp == NULL) {
330		/*
331		 * build one and add it, can't hold lock until after malloc
332		 * done though.
333		 */
334		SCTP_IPI_ADDR_UNLOCK();
335		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *, sizeof(struct sctp_ifn), SCTP_M_IFN);
336		if (sctp_ifnp == NULL) {
337#ifdef INVARIANTS
338			panic("No memory for IFN:%u", sctp_ifnp->ifn_index);
339#endif
340			return (NULL);
341		}
342		sctp_ifnp->ifn_index = ifn_index;
343		sctp_ifnp->ifn_p = ifn;
344		sctp_ifnp->ifn_type = ifn_type;
345		sctp_ifnp->ifa_count = 0;
346		sctp_ifnp->refcount = 1;
347		sctp_ifnp->vrf = vrf;
348		atomic_add_int(&vrf->refcount, 1);
349		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
350		if (if_name != NULL) {
351			memcpy(sctp_ifnp->ifn_name, if_name, SCTP_IFNAMSIZ);
352		} else {
353			memcpy(sctp_ifnp->ifn_name, "unknown", min(7, SCTP_IFNAMSIZ));
354		}
355		hash_ifn_head = &sctppcbinfo.vrf_ifn_hash[(ifn_index & sctppcbinfo.vrf_ifn_hashmark)];
356		LIST_INIT(&sctp_ifnp->ifalist);
357		SCTP_IPI_ADDR_LOCK();
358		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
359		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
360		atomic_add_int(&sctppcbinfo.ipi_count_ifns, 1);
361		new_ifn_af = 1;
362	}
363	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, 1);
364	if (sctp_ifap) {
365		/* Hmm, it already exists? */
366		if ((sctp_ifap->ifn_p) &&
367		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
368			if (new_ifn_af) {
369				/* Remove the created one that we don't want */
370				sctp_delete_ifn(sctp_ifap->ifn_p, 1);
371			}
372			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
373				/* easy to solve, just switch back to active */
374				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
375				sctp_ifap->ifn_p = sctp_ifnp;
376				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
377		exit_stage_left:
378				SCTP_IPI_ADDR_UNLOCK();
379				return (sctp_ifap);
380			} else {
381				goto exit_stage_left;
382			}
383		} else {
384			if (sctp_ifap->ifn_p) {
385				/*
386				 * The first IFN gets the address,
387				 * duplicates are ignored.
388				 */
389				if (new_ifn_af) {
390					/*
391					 * Remove the created one that we
392					 * don't want
393					 */
394					sctp_delete_ifn(sctp_ifap->ifn_p, 1);
395				}
396				goto exit_stage_left;
397			} else {
398				/* repair ifnp which was NULL ? */
399				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
400				sctp_ifap->ifn_p = sctp_ifnp;
401				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
402			}
403			goto exit_stage_left;
404		}
405	}
406	SCTP_IPI_ADDR_UNLOCK();
407	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
408	if (sctp_ifap == NULL) {
409#ifdef INVARIANTS
410		panic("No memory for IFA");
411#endif
412		return (NULL);
413	}
414	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
415	sctp_ifap->ifn_p = sctp_ifnp;
416	atomic_add_int(&sctp_ifnp->refcount, 1);
417	sctp_ifap->vrf_id = vrf_id;
418	sctp_ifap->ifa = ifa;
419	memcpy(&sctp_ifap->address, addr, addr->sa_len);
420	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
421	sctp_ifap->flags = ifa_flags;
422	/* Set scope */
423	if (sctp_ifap->address.sa.sa_family == AF_INET) {
424		struct sockaddr_in *sin;
425
426		sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
427		if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
428		    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
429			sctp_ifap->src_is_loop = 1;
430		}
431		if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
432			sctp_ifap->src_is_priv = 1;
433		}
434		sctp_ifnp->num_v4++;
435		if (new_ifn_af)
436			new_ifn_af = AF_INET;
437	} else if (sctp_ifap->address.sa.sa_family == AF_INET6) {
438		/* ok to use deprecated addresses? */
439		struct sockaddr_in6 *sin6;
440
441		sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
442		if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
443		    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
444			sctp_ifap->src_is_loop = 1;
445		}
446		if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
447			sctp_ifap->src_is_priv = 1;
448		}
449		sctp_ifnp->num_v6++;
450		if (new_ifn_af)
451			new_ifn_af = AF_INET6;
452	} else {
453		new_ifn_af = 0;
454	}
455	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
456
457	if ((sctp_ifap->src_is_priv == 0) &&
458	    (sctp_ifap->src_is_loop == 0)) {
459		sctp_ifap->src_is_glob = 1;
460	}
461	SCTP_IPI_ADDR_LOCK();
462	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
463	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
464	sctp_ifap->refcount = 1;
465	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
466	sctp_ifnp->ifa_count++;
467	vrf->total_ifa_count++;
468	atomic_add_int(&sctppcbinfo.ipi_count_ifas, 1);
469	if (new_ifn_af) {
470		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
471		sctp_ifnp->registered_af = new_ifn_af;
472	}
473	SCTP_IPI_ADDR_UNLOCK();
474	if (dynamic_add) {
475		/*
476		 * Bump up the refcount so that when the timer completes it
477		 * will drop back down.
478		 */
479		struct sctp_laddr *wi;
480
481		atomic_add_int(&sctp_ifap->refcount, 1);
482		wi = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
483		if (wi == NULL) {
484			/*
485			 * Gak, what can we do? We have lost an address
486			 * change can you say HOSED?
487			 */
488			SCTPDBG(SCTP_DEBUG_PCB1, "Lost and address change ???\n");
489			/* Opps, must decrement the count */
490			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index);
491			return (NULL);
492		}
493		SCTP_INCR_LADDR_COUNT();
494		bzero(wi, sizeof(*wi));
495		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
496		wi->ifa = sctp_ifap;
497		wi->action = SCTP_ADD_IP_ADDRESS;
498		SCTP_IPI_ITERATOR_WQ_LOCK();
499		/*
500		 * Should this really be a tailq? As it is we will process
501		 * the newest first :-0
502		 */
503		LIST_INSERT_HEAD(&sctppcbinfo.addr_wq, wi, sctp_nxt_addr);
504		SCTP_IPI_ITERATOR_WQ_UNLOCK();
505		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
506		    (struct sctp_inpcb *)NULL,
507		    (struct sctp_tcb *)NULL,
508		    (struct sctp_nets *)NULL);
509	} else {
510		/* it's ready for use */
511		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
512	}
513	return (sctp_ifap);
514}
515
516void
517sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
518    uint32_t ifn_index)
519{
520	struct sctp_vrf *vrf;
521	struct sctp_ifa *sctp_ifap = NULL;
522
523	SCTP_IPI_ADDR_LOCK();
524
525	vrf = sctp_find_vrf(vrf_id);
526	if (vrf == NULL) {
527		SCTP_PRINTF("Can't find vrf_id:%d\n", vrf_id);
528		goto out_now;
529	}
530	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, 1);
531	if (sctp_ifap) {
532		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
533		sctp_ifap->localifa_flags |= SCTP_BEING_DELETED;
534		vrf->total_ifa_count--;
535		LIST_REMOVE(sctp_ifap, next_bucket);
536		LIST_REMOVE(sctp_ifap, next_ifa);
537		if (sctp_ifap->ifn_p) {
538			sctp_ifap->ifn_p->ifa_count--;
539			if (sctp_ifap->address.sa.sa_family == AF_INET6)
540				sctp_ifap->ifn_p->num_v6--;
541			else if (sctp_ifap->address.sa.sa_family == AF_INET)
542				sctp_ifap->ifn_p->num_v4--;
543			if (SCTP_LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
544				sctp_delete_ifn(sctp_ifap->ifn_p, 1);
545			} else {
546				if ((sctp_ifap->ifn_p->num_v6 == 0) &&
547				    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
548					SCTP_DEREGISTER_INTERFACE(ifn_index,
549					    AF_INET6);
550					SCTP_REGISTER_INTERFACE(ifn_index,
551					    AF_INET);
552					sctp_ifap->ifn_p->registered_af = AF_INET;
553				} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
554				    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
555					SCTP_DEREGISTER_INTERFACE(ifn_index,
556					    AF_INET);
557					SCTP_REGISTER_INTERFACE(ifn_index,
558					    AF_INET6);
559					sctp_ifap->ifn_p->registered_af = AF_INET6;
560				}
561			}
562			sctp_free_ifn(sctp_ifap->ifn_p);
563			sctp_ifap->ifn_p = NULL;
564		}
565	}
566#ifdef SCTP_DEBUG
567	else {
568		SCTPDBG(SCTP_DEBUG_PCB1, "Del Addr-ifn:%d Could not find address:",
569		    ifn_index);
570		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
571	}
572#endif
573
574out_now:
575	SCTP_IPI_ADDR_UNLOCK();
576	if (sctp_ifap) {
577		struct sctp_laddr *wi;
578
579		wi = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
580		if (wi == NULL) {
581			/*
582			 * Gak, what can we do? We have lost an address
583			 * change can you say HOSED?
584			 */
585			SCTPDBG(SCTP_DEBUG_PCB1, "Lost and address change ???\n");
586
587			/* Opps, must decrement the count */
588			sctp_free_ifa(sctp_ifap);
589			return;
590		}
591		SCTP_INCR_LADDR_COUNT();
592		bzero(wi, sizeof(*wi));
593		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
594		wi->ifa = sctp_ifap;
595		wi->action = SCTP_DEL_IP_ADDRESS;
596		SCTP_IPI_ITERATOR_WQ_LOCK();
597		/*
598		 * Should this really be a tailq? As it is we will process
599		 * the newest first :-0
600		 */
601		LIST_INSERT_HEAD(&sctppcbinfo.addr_wq, wi, sctp_nxt_addr);
602		SCTP_IPI_ITERATOR_WQ_UNLOCK();
603
604		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
605		    (struct sctp_inpcb *)NULL,
606		    (struct sctp_tcb *)NULL,
607		    (struct sctp_nets *)NULL);
608	}
609	return;
610}
611
612
613
614static struct sctp_tcb *
615sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
616    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
617{
618	/**** ASSUMSES THE CALLER holds the INP_INFO_RLOCK */
619	/*
620	 * If we support the TCP model, then we must now dig through to see
621	 * if we can find our endpoint in the list of tcp ep's.
622	 */
623	uint16_t lport, rport;
624	struct sctppcbhead *ephead;
625	struct sctp_inpcb *inp;
626	struct sctp_laddr *laddr;
627	struct sctp_tcb *stcb;
628	struct sctp_nets *net;
629
630	if ((to == NULL) || (from == NULL)) {
631		return (NULL);
632	}
633	if (to->sa_family == AF_INET && from->sa_family == AF_INET) {
634		lport = ((struct sockaddr_in *)to)->sin_port;
635		rport = ((struct sockaddr_in *)from)->sin_port;
636	} else if (to->sa_family == AF_INET6 && from->sa_family == AF_INET6) {
637		lport = ((struct sockaddr_in6 *)to)->sin6_port;
638		rport = ((struct sockaddr_in6 *)from)->sin6_port;
639	} else {
640		return NULL;
641	}
642	ephead = &sctppcbinfo.sctp_tcpephash[SCTP_PCBHASH_ALLADDR(
643	    (lport + rport), sctppcbinfo.hashtcpmark)];
644	/*
645	 * Ok now for each of the guys in this bucket we must look and see:
646	 * - Does the remote port match. - Does there single association's
647	 * addresses match this address (to). If so we update p_ep to point
648	 * to this ep and return the tcb from it.
649	 */
650	LIST_FOREACH(inp, ephead, sctp_hash) {
651		SCTP_INP_RLOCK(inp);
652		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
653			SCTP_INP_RUNLOCK(inp);
654			continue;
655		}
656		if (lport != inp->sctp_lport) {
657			SCTP_INP_RUNLOCK(inp);
658			continue;
659		}
660		if (inp->def_vrf_id != vrf_id) {
661			SCTP_INP_RUNLOCK(inp);
662			continue;
663		}
664		/* check to see if the ep has one of the addresses */
665		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
666			/* We are NOT bound all, so look further */
667			int match = 0;
668
669			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
670
671				if (laddr->ifa == NULL) {
672					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
673					continue;
674				}
675				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
676					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
677					continue;
678				}
679				if (laddr->ifa->address.sa.sa_family ==
680				    to->sa_family) {
681					/* see if it matches */
682					struct sockaddr_in *intf_addr, *sin;
683
684					intf_addr = &laddr->ifa->address.sin;
685					sin = (struct sockaddr_in *)to;
686					if (from->sa_family == AF_INET) {
687						if (sin->sin_addr.s_addr ==
688						    intf_addr->sin_addr.s_addr) {
689							match = 1;
690							break;
691						}
692					} else {
693						struct sockaddr_in6 *intf_addr6;
694						struct sockaddr_in6 *sin6;
695
696						sin6 = (struct sockaddr_in6 *)
697						    to;
698						intf_addr6 = &laddr->ifa->address.sin6;
699
700						if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
701						    &intf_addr6->sin6_addr)) {
702							match = 1;
703							break;
704						}
705					}
706				}
707			}
708			if (match == 0) {
709				/* This endpoint does not have this address */
710				SCTP_INP_RUNLOCK(inp);
711				continue;
712			}
713		}
714		/*
715		 * Ok if we hit here the ep has the address, does it hold
716		 * the tcb?
717		 */
718
719		stcb = LIST_FIRST(&inp->sctp_asoc_list);
720		if (stcb == NULL) {
721			SCTP_INP_RUNLOCK(inp);
722			continue;
723		}
724		SCTP_TCB_LOCK(stcb);
725		if (stcb->rport != rport) {
726			/* remote port does not match. */
727			SCTP_TCB_UNLOCK(stcb);
728			SCTP_INP_RUNLOCK(inp);
729			continue;
730		}
731		/* Does this TCB have a matching address? */
732		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
733
734			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
735				/* not the same family, can't be a match */
736				continue;
737			}
738			if (from->sa_family == AF_INET) {
739				struct sockaddr_in *sin, *rsin;
740
741				sin = (struct sockaddr_in *)&net->ro._l_addr;
742				rsin = (struct sockaddr_in *)from;
743				if (sin->sin_addr.s_addr ==
744				    rsin->sin_addr.s_addr) {
745					/* found it */
746					if (netp != NULL) {
747						*netp = net;
748					}
749					/* Update the endpoint pointer */
750					*inp_p = inp;
751					SCTP_INP_RUNLOCK(inp);
752					return (stcb);
753				}
754			} else {
755				struct sockaddr_in6 *sin6, *rsin6;
756
757				sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
758				rsin6 = (struct sockaddr_in6 *)from;
759				if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
760				    &rsin6->sin6_addr)) {
761					/* found it */
762					if (netp != NULL) {
763						*netp = net;
764					}
765					/* Update the endpoint pointer */
766					*inp_p = inp;
767					SCTP_INP_RUNLOCK(inp);
768					return (stcb);
769				}
770			}
771		}
772		SCTP_TCB_UNLOCK(stcb);
773		SCTP_INP_RUNLOCK(inp);
774	}
775	return (NULL);
776}
777
778/*
779 * rules for use
780 *
781 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
782 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
783 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
784 * NULL.
785 */
786
787struct sctp_tcb *
788sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
789    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
790{
791	struct sctpasochead *head;
792	struct sctp_inpcb *inp;
793	struct sctp_tcb *stcb = NULL;
794	struct sctp_nets *net;
795	uint16_t rport;
796
797	inp = *inp_p;
798	if (remote->sa_family == AF_INET) {
799		rport = (((struct sockaddr_in *)remote)->sin_port);
800	} else if (remote->sa_family == AF_INET6) {
801		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
802	} else {
803		return (NULL);
804	}
805	if (locked_tcb) {
806		/*
807		 * UN-lock so we can do proper locking here this occurs when
808		 * called from load_addresses_from_init.
809		 */
810		SCTP_TCB_UNLOCK(locked_tcb);
811	}
812	SCTP_INP_INFO_RLOCK();
813	if (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
814		/*-
815		 * Now either this guy is our listener or it's the
816		 * connector. If it is the one that issued the connect, then
817		 * it's only chance is to be the first TCB in the list. If
818		 * it is the acceptor, then do the special_lookup to hash
819		 * and find the real inp.
820		 */
821		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
822			/* to is peer addr, from is my addr */
823			stcb = sctp_tcb_special_locate(inp_p, remote, local,
824			    netp, inp->def_vrf_id);
825			if ((stcb != NULL) && (locked_tcb == NULL)) {
826				/* we have a locked tcb, lower refcount */
827				SCTP_INP_WLOCK(inp);
828				SCTP_INP_DECR_REF(inp);
829				SCTP_INP_WUNLOCK(inp);
830			}
831			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
832				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
833				SCTP_TCB_LOCK(locked_tcb);
834				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
835			}
836			SCTP_INP_INFO_RUNLOCK();
837			return (stcb);
838		} else {
839			SCTP_INP_WLOCK(inp);
840			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
841				goto null_return;
842			}
843			stcb = LIST_FIRST(&inp->sctp_asoc_list);
844			if (stcb == NULL) {
845				goto null_return;
846			}
847			SCTP_TCB_LOCK(stcb);
848			if (stcb->rport != rport) {
849				/* remote port does not match. */
850				SCTP_TCB_UNLOCK(stcb);
851				goto null_return;
852			}
853			/* now look at the list of remote addresses */
854			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
855#ifdef INVARIANTS
856				if (net == (TAILQ_NEXT(net, sctp_next))) {
857					panic("Corrupt net list");
858				}
859#endif
860				if (net->ro._l_addr.sa.sa_family !=
861				    remote->sa_family) {
862					/* not the same family */
863					continue;
864				}
865				if (remote->sa_family == AF_INET) {
866					struct sockaddr_in *sin, *rsin;
867
868					sin = (struct sockaddr_in *)
869					    &net->ro._l_addr;
870					rsin = (struct sockaddr_in *)remote;
871					if (sin->sin_addr.s_addr ==
872					    rsin->sin_addr.s_addr) {
873						/* found it */
874						if (netp != NULL) {
875							*netp = net;
876						}
877						if (locked_tcb == NULL) {
878							SCTP_INP_DECR_REF(inp);
879						} else if (locked_tcb != stcb) {
880							SCTP_TCB_LOCK(locked_tcb);
881						}
882						SCTP_INP_WUNLOCK(inp);
883						SCTP_INP_INFO_RUNLOCK();
884						return (stcb);
885					}
886				} else if (remote->sa_family == AF_INET6) {
887					struct sockaddr_in6 *sin6, *rsin6;
888
889					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
890					rsin6 = (struct sockaddr_in6 *)remote;
891					if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
892					    &rsin6->sin6_addr)) {
893						/* found it */
894						if (netp != NULL) {
895							*netp = net;
896						}
897						if (locked_tcb == NULL) {
898							SCTP_INP_DECR_REF(inp);
899						} else if (locked_tcb != stcb) {
900							SCTP_TCB_LOCK(locked_tcb);
901						}
902						SCTP_INP_WUNLOCK(inp);
903						SCTP_INP_INFO_RUNLOCK();
904						return (stcb);
905					}
906				}
907			}
908			SCTP_TCB_UNLOCK(stcb);
909		}
910	} else {
911		SCTP_INP_WLOCK(inp);
912		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
913			goto null_return;
914		}
915		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
916		    inp->sctp_hashmark)];
917		if (head == NULL) {
918			goto null_return;
919		}
920		LIST_FOREACH(stcb, head, sctp_tcbhash) {
921			if (stcb->rport != rport) {
922				/* remote port does not match */
923				continue;
924			}
925			/* now look at the list of remote addresses */
926			SCTP_TCB_LOCK(stcb);
927			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
928#ifdef INVARIANTS
929				if (net == (TAILQ_NEXT(net, sctp_next))) {
930					panic("Corrupt net list");
931				}
932#endif
933				if (net->ro._l_addr.sa.sa_family !=
934				    remote->sa_family) {
935					/* not the same family */
936					continue;
937				}
938				if (remote->sa_family == AF_INET) {
939					struct sockaddr_in *sin, *rsin;
940
941					sin = (struct sockaddr_in *)
942					    &net->ro._l_addr;
943					rsin = (struct sockaddr_in *)remote;
944					if (sin->sin_addr.s_addr ==
945					    rsin->sin_addr.s_addr) {
946						/* found it */
947						if (netp != NULL) {
948							*netp = net;
949						}
950						if (locked_tcb == NULL) {
951							SCTP_INP_DECR_REF(inp);
952						} else if (locked_tcb != stcb) {
953							SCTP_TCB_LOCK(locked_tcb);
954						}
955						SCTP_INP_WUNLOCK(inp);
956						SCTP_INP_INFO_RUNLOCK();
957						return (stcb);
958					}
959				} else if (remote->sa_family == AF_INET6) {
960					struct sockaddr_in6 *sin6, *rsin6;
961
962					sin6 = (struct sockaddr_in6 *)
963					    &net->ro._l_addr;
964					rsin6 = (struct sockaddr_in6 *)remote;
965					if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
966					    &rsin6->sin6_addr)) {
967						/* found it */
968						if (netp != NULL) {
969							*netp = net;
970						}
971						if (locked_tcb == NULL) {
972							SCTP_INP_DECR_REF(inp);
973						} else if (locked_tcb != stcb) {
974							SCTP_TCB_LOCK(locked_tcb);
975						}
976						SCTP_INP_WUNLOCK(inp);
977						SCTP_INP_INFO_RUNLOCK();
978						return (stcb);
979					}
980				}
981			}
982			SCTP_TCB_UNLOCK(stcb);
983		}
984	}
985null_return:
986	/* clean up for returning null */
987	if (locked_tcb) {
988		SCTP_TCB_LOCK(locked_tcb);
989	}
990	SCTP_INP_WUNLOCK(inp);
991	SCTP_INP_INFO_RUNLOCK();
992	/* not found */
993	return (NULL);
994}
995
996/*
997 * Find an association for a specific endpoint using the association id given
998 * out in the COMM_UP notification
999 */
1000
1001struct sctp_tcb *
1002sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1003{
1004	/*
1005	 * Use my the assoc_id to find a endpoint
1006	 */
1007	struct sctpasochead *head;
1008	struct sctp_tcb *stcb;
1009	uint32_t id;
1010
1011	if (asoc_id == 0 || inp == NULL) {
1012		return (NULL);
1013	}
1014	SCTP_INP_INFO_RLOCK();
1015	id = (uint32_t) asoc_id;
1016	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(id,
1017	    sctppcbinfo.hashasocmark)];
1018	if (head == NULL) {
1019		/* invalid id TSNH */
1020		SCTP_INP_INFO_RUNLOCK();
1021		return (NULL);
1022	}
1023	LIST_FOREACH(stcb, head, sctp_asocs) {
1024		SCTP_INP_RLOCK(stcb->sctp_ep);
1025		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1026			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1027			SCTP_INP_INFO_RUNLOCK();
1028			return (NULL);
1029		}
1030		if (stcb->asoc.assoc_id == id) {
1031			/* candidate */
1032			if (inp != stcb->sctp_ep) {
1033				/*
1034				 * some other guy has the same id active (id
1035				 * collision ??).
1036				 */
1037				SCTP_INP_RUNLOCK(stcb->sctp_ep);
1038				continue;
1039			}
1040			if (want_lock) {
1041				SCTP_TCB_LOCK(stcb);
1042			}
1043			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1044			SCTP_INP_INFO_RUNLOCK();
1045			return (stcb);
1046		}
1047		SCTP_INP_RUNLOCK(stcb->sctp_ep);
1048	}
1049	/* Ok if we missed here, lets try the restart hash */
1050	head = &sctppcbinfo.sctp_restarthash[SCTP_PCBHASH_ASOC(id, sctppcbinfo.hashrestartmark)];
1051	if (head == NULL) {
1052		/* invalid id TSNH */
1053		SCTP_INP_INFO_RUNLOCK();
1054		return (NULL);
1055	}
1056	LIST_FOREACH(stcb, head, sctp_tcbrestarhash) {
1057		SCTP_INP_RLOCK(stcb->sctp_ep);
1058		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1059			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1060			continue;
1061		}
1062		if (want_lock) {
1063			SCTP_TCB_LOCK(stcb);
1064		}
1065		if (stcb->asoc.assoc_id == id) {
1066			/* candidate */
1067			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1068			if (inp != stcb->sctp_ep) {
1069				/*
1070				 * some other guy has the same id active (id
1071				 * collision ??).
1072				 */
1073				if (want_lock) {
1074					SCTP_TCB_UNLOCK(stcb);
1075				}
1076				continue;
1077			}
1078			SCTP_INP_INFO_RUNLOCK();
1079			return (stcb);
1080		} else {
1081			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1082		}
1083		if (want_lock) {
1084			SCTP_TCB_UNLOCK(stcb);
1085		}
1086	}
1087	SCTP_INP_INFO_RUNLOCK();
1088	return (NULL);
1089}
1090
1091
1092static struct sctp_inpcb *
1093sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1094    uint16_t lport, uint32_t vrf_id)
1095{
1096	struct sctp_inpcb *inp;
1097	struct sockaddr_in *sin;
1098	struct sockaddr_in6 *sin6;
1099	struct sctp_laddr *laddr;
1100	int fnd;
1101
1102	/*
1103	 * Endpoing probe expects that the INP_INFO is locked.
1104	 */
1105	if (nam->sa_family == AF_INET) {
1106		sin = (struct sockaddr_in *)nam;
1107		sin6 = NULL;
1108	} else if (nam->sa_family == AF_INET6) {
1109		sin6 = (struct sockaddr_in6 *)nam;
1110		sin = NULL;
1111	} else {
1112		/* unsupported family */
1113		return (NULL);
1114	}
1115	if (head == NULL)
1116		return (NULL);
1117	LIST_FOREACH(inp, head, sctp_hash) {
1118		SCTP_INP_RLOCK(inp);
1119		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1120			SCTP_INP_RUNLOCK(inp);
1121			continue;
1122		}
1123		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1124		    (inp->sctp_lport == lport)) {
1125			/* got it */
1126			if ((nam->sa_family == AF_INET) &&
1127			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1128			    SCTP_IPV6_V6ONLY(inp)) {
1129				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1130				SCTP_INP_RUNLOCK(inp);
1131				continue;
1132			}
1133			/* A V6 address and the endpoint is NOT bound V6 */
1134			if (nam->sa_family == AF_INET6 &&
1135			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1136				SCTP_INP_RUNLOCK(inp);
1137				continue;
1138			}
1139			/* does a VRF id match? */
1140			fnd = 0;
1141			if (inp->def_vrf_id == vrf_id)
1142				fnd = 1;
1143
1144			SCTP_INP_RUNLOCK(inp);
1145			if (!fnd)
1146				continue;
1147			return (inp);
1148		}
1149		SCTP_INP_RUNLOCK(inp);
1150	}
1151
1152	if ((nam->sa_family == AF_INET) &&
1153	    (sin->sin_addr.s_addr == INADDR_ANY)) {
1154		/* Can't hunt for one that has no address specified */
1155		return (NULL);
1156	} else if ((nam->sa_family == AF_INET6) &&
1157	    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
1158		/* Can't hunt for one that has no address specified */
1159		return (NULL);
1160	}
1161	/*
1162	 * ok, not bound to all so see if we can find a EP bound to this
1163	 * address.
1164	 */
1165	LIST_FOREACH(inp, head, sctp_hash) {
1166		SCTP_INP_RLOCK(inp);
1167		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1168			SCTP_INP_RUNLOCK(inp);
1169			continue;
1170		}
1171		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1172			SCTP_INP_RUNLOCK(inp);
1173			continue;
1174		}
1175		/*
1176		 * Ok this could be a likely candidate, look at all of its
1177		 * addresses
1178		 */
1179		if (inp->sctp_lport != lport) {
1180			SCTP_INP_RUNLOCK(inp);
1181			continue;
1182		}
1183		/* does a VRF id match? */
1184		fnd = 0;
1185		if (inp->def_vrf_id == vrf_id)
1186			fnd = 1;
1187
1188		if (!fnd) {
1189			SCTP_INP_RUNLOCK(inp);
1190			continue;
1191		}
1192		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1193			if (laddr->ifa == NULL) {
1194				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1195				    __FUNCTION__);
1196				continue;
1197			}
1198			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1199			    laddr->ifa);
1200			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1201				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1202				continue;
1203			}
1204			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1205				/* possible, see if it matches */
1206				struct sockaddr_in *intf_addr;
1207
1208				intf_addr = &laddr->ifa->address.sin;
1209				if (nam->sa_family == AF_INET) {
1210					if (sin->sin_addr.s_addr ==
1211					    intf_addr->sin_addr.s_addr) {
1212						SCTP_INP_RUNLOCK(inp);
1213						return (inp);
1214					}
1215				} else if (nam->sa_family == AF_INET6) {
1216					struct sockaddr_in6 *intf_addr6;
1217
1218					intf_addr6 = &laddr->ifa->address.sin6;
1219					if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
1220					    &intf_addr6->sin6_addr)) {
1221						SCTP_INP_RUNLOCK(inp);
1222						return (inp);
1223					}
1224				}
1225			}
1226		}
1227		SCTP_INP_RUNLOCK(inp);
1228	}
1229	return (NULL);
1230}
1231
1232
1233struct sctp_inpcb *
1234sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1235    uint32_t vrf_id)
1236{
1237	/*
1238	 * First we check the hash table to see if someone has this port
1239	 * bound with just the port.
1240	 */
1241	struct sctp_inpcb *inp;
1242	struct sctppcbhead *head;
1243	struct sockaddr_in *sin;
1244	struct sockaddr_in6 *sin6;
1245	int lport;
1246
1247	if (nam->sa_family == AF_INET) {
1248		sin = (struct sockaddr_in *)nam;
1249		lport = ((struct sockaddr_in *)nam)->sin_port;
1250	} else if (nam->sa_family == AF_INET6) {
1251		sin6 = (struct sockaddr_in6 *)nam;
1252		lport = ((struct sockaddr_in6 *)nam)->sin6_port;
1253	} else {
1254		/* unsupported family */
1255		return (NULL);
1256	}
1257	/*
1258	 * I could cheat here and just cast to one of the types but we will
1259	 * do it right. It also provides the check against an Unsupported
1260	 * type too.
1261	 */
1262	/* Find the head of the ALLADDR chain */
1263	if (have_lock == 0) {
1264		SCTP_INP_INFO_RLOCK();
1265
1266	}
1267	head = &sctppcbinfo.sctp_ephash[SCTP_PCBHASH_ALLADDR(lport,
1268	    sctppcbinfo.hashmark)];
1269	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1270
1271	/*
1272	 * If the TCP model exists it could be that the main listening
1273	 * endpoint is gone but there exists a connected socket for this guy
1274	 * yet. If so we can return the first one that we find. This may NOT
1275	 * be the correct one but the sctp_findassociation_ep_addr has
1276	 * further code to look at all TCP models.
1277	 */
1278	if (inp == NULL && find_tcp_pool) {
1279		unsigned int i;
1280
1281		for (i = 0; i < sctppcbinfo.hashtblsize; i++) {
1282			/*
1283			 * This is real gross, but we do NOT have a remote
1284			 * port at this point depending on who is calling.
1285			 * We must therefore look for ANY one that matches
1286			 * our local port :/
1287			 */
1288			head = &sctppcbinfo.sctp_tcpephash[i];
1289			if (LIST_FIRST(head)) {
1290				inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1291				if (inp) {
1292					/* Found one */
1293					break;
1294				}
1295			}
1296		}
1297	}
1298	if (inp) {
1299		SCTP_INP_INCR_REF(inp);
1300	}
1301	if (have_lock == 0) {
1302		SCTP_INP_INFO_RUNLOCK();
1303	}
1304	return (inp);
1305}
1306
1307/*
1308 * Find an association for an endpoint with the pointer to whom you want to
1309 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1310 * need to change the *to to some other struct like a mbuf...
1311 */
1312struct sctp_tcb *
1313sctp_findassociation_addr_sa(struct sockaddr *to, struct sockaddr *from,
1314    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1315    uint32_t vrf_id)
1316{
1317	struct sctp_inpcb *inp = NULL;
1318	struct sctp_tcb *retval;
1319
1320	SCTP_INP_INFO_RLOCK();
1321	if (find_tcp_pool) {
1322		if (inp_p != NULL) {
1323			retval = sctp_tcb_special_locate(inp_p, from, to, netp,
1324			    vrf_id);
1325		} else {
1326			retval = sctp_tcb_special_locate(&inp, from, to, netp,
1327			    vrf_id);
1328		}
1329		if (retval != NULL) {
1330			SCTP_INP_INFO_RUNLOCK();
1331			return (retval);
1332		}
1333	}
1334	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1335	if (inp_p != NULL) {
1336		*inp_p = inp;
1337	}
1338	SCTP_INP_INFO_RUNLOCK();
1339
1340	if (inp == NULL) {
1341		return (NULL);
1342	}
1343	/*
1344	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1345	 * we now place the source address or from in the to of the find
1346	 * endpoint call. Since in reality this chain is used from the
1347	 * inbound packet side.
1348	 */
1349	if (inp_p != NULL) {
1350		retval = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1351		    NULL);
1352	} else {
1353		retval = sctp_findassociation_ep_addr(&inp, from, netp, to,
1354		    NULL);
1355	}
1356	return retval;
1357}
1358
1359
1360/*
1361 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1362 * find all addresses that the sender has specified in any address list. Each
1363 * address will be used to lookup the TCB and see if one exits.
1364 */
1365static struct sctp_tcb *
1366sctp_findassociation_special_addr(struct mbuf *m, int iphlen, int offset,
1367    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1368    struct sockaddr *dest)
1369{
1370	struct sockaddr_in sin4;
1371	struct sockaddr_in6 sin6;
1372	struct sctp_paramhdr *phdr, parm_buf;
1373	struct sctp_tcb *retval;
1374	uint32_t ptype, plen;
1375
1376	memset(&sin4, 0, sizeof(sin4));
1377	memset(&sin6, 0, sizeof(sin6));
1378	sin4.sin_len = sizeof(sin4);
1379	sin4.sin_family = AF_INET;
1380	sin4.sin_port = sh->src_port;
1381	sin6.sin6_len = sizeof(sin6);
1382	sin6.sin6_family = AF_INET6;
1383	sin6.sin6_port = sh->src_port;
1384
1385	retval = NULL;
1386	offset += sizeof(struct sctp_init_chunk);
1387
1388	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
1389	while (phdr != NULL) {
1390		/* now we must see if we want the parameter */
1391		ptype = ntohs(phdr->param_type);
1392		plen = ntohs(phdr->param_length);
1393		if (plen == 0) {
1394			break;
1395		}
1396		if (ptype == SCTP_IPV4_ADDRESS &&
1397		    plen == sizeof(struct sctp_ipv4addr_param)) {
1398			/* Get the rest of the address */
1399			struct sctp_ipv4addr_param ip4_parm, *p4;
1400
1401			phdr = sctp_get_next_param(m, offset,
1402			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
1403			if (phdr == NULL) {
1404				return (NULL);
1405			}
1406			p4 = (struct sctp_ipv4addr_param *)phdr;
1407			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
1408			/* look it up */
1409			retval = sctp_findassociation_ep_addr(inp_p,
1410			    (struct sockaddr *)&sin4, netp, dest, NULL);
1411			if (retval != NULL) {
1412				return (retval);
1413			}
1414		} else if (ptype == SCTP_IPV6_ADDRESS &&
1415		    plen == sizeof(struct sctp_ipv6addr_param)) {
1416			/* Get the rest of the address */
1417			struct sctp_ipv6addr_param ip6_parm, *p6;
1418
1419			phdr = sctp_get_next_param(m, offset,
1420			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
1421			if (phdr == NULL) {
1422				return (NULL);
1423			}
1424			p6 = (struct sctp_ipv6addr_param *)phdr;
1425			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
1426			/* look it up */
1427			retval = sctp_findassociation_ep_addr(inp_p,
1428			    (struct sockaddr *)&sin6, netp, dest, NULL);
1429			if (retval != NULL) {
1430				return (retval);
1431			}
1432		}
1433		offset += SCTP_SIZE32(plen);
1434		phdr = sctp_get_next_param(m, offset, &parm_buf,
1435		    sizeof(parm_buf));
1436	}
1437	return (NULL);
1438}
1439
1440
1441static struct sctp_tcb *
1442sctp_findassoc_by_vtag(struct sockaddr *from, uint32_t vtag,
1443    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
1444    uint16_t lport, int skip_src_check)
1445{
1446	/*
1447	 * Use my vtag to hash. If we find it we then verify the source addr
1448	 * is in the assoc. If all goes well we save a bit on rec of a
1449	 * packet.
1450	 */
1451	struct sctpasochead *head;
1452	struct sctp_nets *net;
1453	struct sctp_tcb *stcb;
1454
1455	*netp = NULL;
1456	*inp_p = NULL;
1457	SCTP_INP_INFO_RLOCK();
1458	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(vtag,
1459	    sctppcbinfo.hashasocmark)];
1460	if (head == NULL) {
1461		/* invalid vtag */
1462		SCTP_INP_INFO_RUNLOCK();
1463		return (NULL);
1464	}
1465	LIST_FOREACH(stcb, head, sctp_asocs) {
1466		SCTP_INP_RLOCK(stcb->sctp_ep);
1467		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1468			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1469			continue;
1470		}
1471		SCTP_TCB_LOCK(stcb);
1472		SCTP_INP_RUNLOCK(stcb->sctp_ep);
1473		if (stcb->asoc.my_vtag == vtag) {
1474			/* candidate */
1475			if (stcb->rport != rport) {
1476				/*
1477				 * we could remove this if vtags are unique
1478				 * across the system.
1479				 */
1480				SCTP_TCB_UNLOCK(stcb);
1481				continue;
1482			}
1483			if (stcb->sctp_ep->sctp_lport != lport) {
1484				/*
1485				 * we could remove this if vtags are unique
1486				 * across the system.
1487				 */
1488				SCTP_TCB_UNLOCK(stcb);
1489				continue;
1490			}
1491			if (skip_src_check) {
1492				*netp = NULL;	/* unknown */
1493				if (inp_p)
1494					*inp_p = stcb->sctp_ep;
1495				SCTP_INP_INFO_RUNLOCK();
1496				return (stcb);
1497			}
1498			net = sctp_findnet(stcb, from);
1499			if (net) {
1500				/* yep its him. */
1501				*netp = net;
1502				SCTP_STAT_INCR(sctps_vtagexpress);
1503				*inp_p = stcb->sctp_ep;
1504				SCTP_INP_INFO_RUNLOCK();
1505				return (stcb);
1506			} else {
1507				/*
1508				 * not him, this should only happen in rare
1509				 * cases so I peg it.
1510				 */
1511				SCTP_STAT_INCR(sctps_vtagbogus);
1512			}
1513		}
1514		SCTP_TCB_UNLOCK(stcb);
1515	}
1516	SCTP_INP_INFO_RUNLOCK();
1517	return (NULL);
1518}
1519
1520/*
1521 * Find an association with the pointer to the inbound IP packet. This can be
1522 * a IPv4 or IPv6 packet.
1523 */
1524struct sctp_tcb *
1525sctp_findassociation_addr(struct mbuf *m, int iphlen, int offset,
1526    struct sctphdr *sh, struct sctp_chunkhdr *ch,
1527    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
1528{
1529	int find_tcp_pool;
1530	struct ip *iph;
1531	struct sctp_tcb *retval;
1532	struct sockaddr_storage to_store, from_store;
1533	struct sockaddr *to = (struct sockaddr *)&to_store;
1534	struct sockaddr *from = (struct sockaddr *)&from_store;
1535	struct sctp_inpcb *inp;
1536
1537	iph = mtod(m, struct ip *);
1538	if (iph->ip_v == IPVERSION) {
1539		/* its IPv4 */
1540		struct sockaddr_in *from4;
1541
1542		from4 = (struct sockaddr_in *)&from_store;
1543		bzero(from4, sizeof(*from4));
1544		from4->sin_family = AF_INET;
1545		from4->sin_len = sizeof(struct sockaddr_in);
1546		from4->sin_addr.s_addr = iph->ip_src.s_addr;
1547		from4->sin_port = sh->src_port;
1548	} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
1549		/* its IPv6 */
1550		struct ip6_hdr *ip6;
1551		struct sockaddr_in6 *from6;
1552
1553		ip6 = mtod(m, struct ip6_hdr *);
1554		from6 = (struct sockaddr_in6 *)&from_store;
1555		bzero(from6, sizeof(*from6));
1556		from6->sin6_family = AF_INET6;
1557		from6->sin6_len = sizeof(struct sockaddr_in6);
1558		from6->sin6_addr = ip6->ip6_src;
1559		from6->sin6_port = sh->src_port;
1560		/* Get the scopes in properly to the sin6 addr's */
1561		/* we probably don't need these operations */
1562		(void)sa6_recoverscope(from6);
1563		sa6_embedscope(from6, ip6_use_defzone);
1564	} else {
1565		/* Currently not supported. */
1566		return (NULL);
1567	}
1568	if (sh->v_tag) {
1569		/* we only go down this path if vtag is non-zero */
1570		retval = sctp_findassoc_by_vtag(from, ntohl(sh->v_tag),
1571		    inp_p, netp, sh->src_port, sh->dest_port, 0);
1572		if (retval) {
1573			return (retval);
1574		}
1575	}
1576	if (iph->ip_v == IPVERSION) {
1577		/* its IPv4 */
1578		struct sockaddr_in *to4;
1579
1580		to4 = (struct sockaddr_in *)&to_store;
1581		bzero(to4, sizeof(*to4));
1582		to4->sin_family = AF_INET;
1583		to4->sin_len = sizeof(struct sockaddr_in);
1584		to4->sin_addr.s_addr = iph->ip_dst.s_addr;
1585		to4->sin_port = sh->dest_port;
1586	} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
1587		/* its IPv6 */
1588		struct ip6_hdr *ip6;
1589		struct sockaddr_in6 *to6;
1590
1591		ip6 = mtod(m, struct ip6_hdr *);
1592		to6 = (struct sockaddr_in6 *)&to_store;
1593		bzero(to6, sizeof(*to6));
1594		to6->sin6_family = AF_INET6;
1595		to6->sin6_len = sizeof(struct sockaddr_in6);
1596		to6->sin6_addr = ip6->ip6_dst;
1597		to6->sin6_port = sh->dest_port;
1598		/* Get the scopes in properly to the sin6 addr's */
1599		/* we probably don't need these operations */
1600		(void)sa6_recoverscope(to6);
1601		sa6_embedscope(to6, ip6_use_defzone);
1602	}
1603	find_tcp_pool = 0;
1604	if ((ch->chunk_type != SCTP_INITIATION) &&
1605	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
1606	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
1607	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
1608		/* Other chunk types go to the tcp pool. */
1609		find_tcp_pool = 1;
1610	}
1611	if (inp_p) {
1612		retval = sctp_findassociation_addr_sa(to, from, inp_p, netp,
1613		    find_tcp_pool, vrf_id);
1614		inp = *inp_p;
1615	} else {
1616		retval = sctp_findassociation_addr_sa(to, from, &inp, netp,
1617		    find_tcp_pool, vrf_id);
1618	}
1619	SCTPDBG(SCTP_DEBUG_PCB1, "retval:%p inp:%p\n", retval, inp);
1620	if (retval == NULL && inp) {
1621		/* Found a EP but not this address */
1622		if ((ch->chunk_type == SCTP_INITIATION) ||
1623		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
1624			/*-
1625			 * special hook, we do NOT return linp or an
1626			 * association that is linked to an existing
1627			 * association that is under the TCP pool (i.e. no
1628			 * listener exists). The endpoint finding routine
1629			 * will always find a listner before examining the
1630			 * TCP pool.
1631			 */
1632			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
1633				if (inp_p) {
1634					*inp_p = NULL;
1635				}
1636				return (NULL);
1637			}
1638			retval = sctp_findassociation_special_addr(m, iphlen,
1639			    offset, sh, &inp, netp, to);
1640			if (inp_p != NULL) {
1641				*inp_p = inp;
1642			}
1643		}
1644	}
1645	SCTPDBG(SCTP_DEBUG_PCB1, "retval is %p\n", retval);
1646	return (retval);
1647}
1648
1649/*
1650 * lookup an association by an ASCONF lookup address.
1651 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
1652 */
1653struct sctp_tcb *
1654sctp_findassociation_ep_asconf(struct mbuf *m, int iphlen, int offset,
1655    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp)
1656{
1657	struct sctp_tcb *stcb;
1658	struct sockaddr_in *sin;
1659	struct sockaddr_in6 *sin6;
1660	struct sockaddr_storage local_store, remote_store;
1661	struct ip *iph;
1662	struct sctp_paramhdr parm_buf, *phdr;
1663	int ptype;
1664	int zero_address = 0;
1665
1666
1667	memset(&local_store, 0, sizeof(local_store));
1668	memset(&remote_store, 0, sizeof(remote_store));
1669
1670	/* First get the destination address setup too. */
1671	iph = mtod(m, struct ip *);
1672	if (iph->ip_v == IPVERSION) {
1673		/* its IPv4 */
1674		sin = (struct sockaddr_in *)&local_store;
1675		sin->sin_family = AF_INET;
1676		sin->sin_len = sizeof(*sin);
1677		sin->sin_port = sh->dest_port;
1678		sin->sin_addr.s_addr = iph->ip_dst.s_addr;
1679	} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
1680		/* its IPv6 */
1681		struct ip6_hdr *ip6;
1682
1683		ip6 = mtod(m, struct ip6_hdr *);
1684		sin6 = (struct sockaddr_in6 *)&local_store;
1685		sin6->sin6_family = AF_INET6;
1686		sin6->sin6_len = sizeof(*sin6);
1687		sin6->sin6_port = sh->dest_port;
1688		sin6->sin6_addr = ip6->ip6_dst;
1689	} else {
1690		return NULL;
1691	}
1692
1693	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
1694	    &parm_buf, sizeof(struct sctp_paramhdr));
1695	if (phdr == NULL) {
1696		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
1697		    __FUNCTION__);
1698		return NULL;
1699	}
1700	ptype = (int)((uint32_t) ntohs(phdr->param_type));
1701	/* get the correlation address */
1702	if (ptype == SCTP_IPV6_ADDRESS) {
1703		/* ipv6 address param */
1704		struct sctp_ipv6addr_param *p6, p6_buf;
1705
1706		if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
1707			return NULL;
1708		}
1709		p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
1710		    offset + sizeof(struct sctp_asconf_chunk),
1711		    &p6_buf.ph, sizeof(*p6));
1712		if (p6 == NULL) {
1713			SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
1714			    __FUNCTION__);
1715			return (NULL);
1716		}
1717		sin6 = (struct sockaddr_in6 *)&remote_store;
1718		sin6->sin6_family = AF_INET6;
1719		sin6->sin6_len = sizeof(*sin6);
1720		sin6->sin6_port = sh->src_port;
1721		memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
1722		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
1723			zero_address = 1;
1724	} else if (ptype == SCTP_IPV4_ADDRESS) {
1725		/* ipv4 address param */
1726		struct sctp_ipv4addr_param *p4, p4_buf;
1727
1728		if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
1729			return NULL;
1730		}
1731		p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
1732		    offset + sizeof(struct sctp_asconf_chunk),
1733		    &p4_buf.ph, sizeof(*p4));
1734		if (p4 == NULL) {
1735			SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
1736			    __FUNCTION__);
1737			return (NULL);
1738		}
1739		sin = (struct sockaddr_in *)&remote_store;
1740		sin->sin_family = AF_INET;
1741		sin->sin_len = sizeof(*sin);
1742		sin->sin_port = sh->src_port;
1743		memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
1744		if (sin->sin_addr.s_addr == INADDR_ANY)
1745			zero_address = 1;
1746	} else {
1747		/* invalid address param type */
1748		return NULL;
1749	}
1750
1751	if (zero_address) {
1752		stcb = sctp_findassoc_by_vtag(NULL, ntohl(sh->v_tag), inp_p,
1753		    netp, sh->src_port, sh->dest_port, 1);
1754		/*
1755		 * printf("findassociation_ep_asconf: zero lookup address
1756		 * finds stcb 0x%x\n", (uint32_t)stcb);
1757		 */
1758	} else {
1759		stcb = sctp_findassociation_ep_addr(inp_p,
1760		    (struct sockaddr *)&remote_store, netp,
1761		    (struct sockaddr *)&local_store, NULL);
1762	}
1763	return (stcb);
1764}
1765
1766
1767/*
1768 * allocate a sctp_inpcb and setup a temporary binding to a port/all
1769 * addresses. This way if we don't get a bind we by default pick a ephemeral
1770 * port with all addresses bound.
1771 */
1772int
1773sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
1774{
1775	/*
1776	 * we get called when a new endpoint starts up. We need to allocate
1777	 * the sctp_inpcb structure from the zone and init it. Mark it as
1778	 * unbound and find a port that we can use as an ephemeral with
1779	 * INADDR_ANY. If the user binds later no problem we can then add in
1780	 * the specific addresses. And setup the default parameters for the
1781	 * EP.
1782	 */
1783	int i, error;
1784	struct sctp_inpcb *inp;
1785	struct sctp_pcb *m;
1786	struct timeval time;
1787	sctp_sharedkey_t *null_key;
1788
1789	error = 0;
1790
1791	SCTP_INP_INFO_WLOCK();
1792	inp = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_ep, struct sctp_inpcb);
1793	if (inp == NULL) {
1794		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
1795		SCTP_INP_INFO_WUNLOCK();
1796		return (ENOBUFS);
1797	}
1798	/* zap it */
1799	bzero(inp, sizeof(*inp));
1800
1801	/* bump generations */
1802	/* setup socket pointers */
1803	inp->sctp_socket = so;
1804	inp->ip_inp.inp.inp_socket = so;
1805
1806	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
1807	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
1808
1809#ifdef IPSEC
1810	{
1811		struct inpcbpolicy *pcb_sp = NULL;
1812
1813		error = ipsec_init_pcbpolicy(so, &pcb_sp);
1814		/* Arrange to share the policy */
1815		inp->ip_inp.inp.inp_sp = pcb_sp;
1816		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
1817	}
1818	if (error != 0) {
1819		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
1820		SCTP_INP_INFO_WUNLOCK();
1821		return error;
1822	}
1823#endif				/* IPSEC */
1824	SCTP_INCR_EP_COUNT();
1825	inp->ip_inp.inp.inp_ip_ttl = ip_defttl;
1826	SCTP_INP_INFO_WUNLOCK();
1827
1828	so->so_pcb = (caddr_t)inp;
1829
1830	if ((SCTP_SO_TYPE(so) == SOCK_DGRAM) ||
1831	    (SCTP_SO_TYPE(so) == SOCK_SEQPACKET)) {
1832		/* UDP style socket */
1833		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
1834		    SCTP_PCB_FLAGS_UNBOUND);
1835		/* Be sure it is NON-BLOCKING IO for UDP */
1836		/* SCTP_SET_SO_NBIO(so); */
1837	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
1838		/* TCP style socket */
1839		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
1840		    SCTP_PCB_FLAGS_UNBOUND);
1841		/* Be sure we have blocking IO by default */
1842		SCTP_CLEAR_SO_NBIO(so);
1843	} else {
1844		/*
1845		 * unsupported socket type (RAW, etc)- in case we missed it
1846		 * in protosw
1847		 */
1848		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
1849		return (EOPNOTSUPP);
1850	}
1851	sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
1852
1853	inp->sctp_tcbhash = SCTP_HASH_INIT(sctp_pcbtblsize,
1854	    &inp->sctp_hashmark);
1855	if (inp->sctp_tcbhash == NULL) {
1856		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
1857		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
1858		return (ENOBUFS);
1859	}
1860	inp->def_vrf_id = vrf_id;
1861
1862	SCTP_INP_INFO_WLOCK();
1863	SCTP_INP_LOCK_INIT(inp);
1864	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
1865	SCTP_INP_READ_INIT(inp);
1866	SCTP_ASOC_CREATE_LOCK_INIT(inp);
1867	/* lock the new ep */
1868	SCTP_INP_WLOCK(inp);
1869
1870	/* add it to the info area */
1871	LIST_INSERT_HEAD(&sctppcbinfo.listhead, inp, sctp_list);
1872	SCTP_INP_INFO_WUNLOCK();
1873
1874	TAILQ_INIT(&inp->read_queue);
1875	LIST_INIT(&inp->sctp_addr_list);
1876
1877	LIST_INIT(&inp->sctp_asoc_list);
1878
1879#ifdef SCTP_TRACK_FREED_ASOCS
1880	/* TEMP CODE */
1881	LIST_INIT(&inp->sctp_asoc_free_list);
1882#endif
1883	/* Init the timer structure for signature change */
1884	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
1885	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
1886
1887	/* now init the actual endpoint default data */
1888	m = &inp->sctp_ep;
1889
1890	/* setup the base timeout information */
1891	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
1892	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
1893	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(sctp_delayed_sack_time_default);
1894	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(sctp_heartbeat_interval_default);
1895	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(sctp_pmtu_raise_time_default);
1896	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(sctp_shutdown_guard_time_default);
1897	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(sctp_secret_lifetime_default);
1898	/* all max/min max are in ms */
1899	m->sctp_maxrto = sctp_rto_max_default;
1900	m->sctp_minrto = sctp_rto_min_default;
1901	m->initial_rto = sctp_rto_initial_default;
1902	m->initial_init_rto_max = sctp_init_rto_max_default;
1903	m->sctp_sack_freq = sctp_sack_freq_default;
1904
1905	m->max_open_streams_intome = MAX_SCTP_STREAMS;
1906
1907	m->max_init_times = sctp_init_rtx_max_default;
1908	m->max_send_times = sctp_assoc_rtx_max_default;
1909	m->def_net_failure = sctp_path_rtx_max_default;
1910	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
1911	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
1912	m->max_burst = sctp_max_burst_default;
1913	/* number of streams to pre-open on a association */
1914	m->pre_open_stream_count = sctp_nr_outgoing_streams_default;
1915
1916	/* Add adaptation cookie */
1917	m->adaptation_layer_indicator = 0x504C5253;
1918
1919	/* seed random number generator */
1920	m->random_counter = 1;
1921	m->store_at = SCTP_SIGNATURE_SIZE;
1922	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
1923	sctp_fill_random_store(m);
1924
1925	/* Minimum cookie size */
1926	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
1927	    sizeof(struct sctp_state_cookie);
1928	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
1929
1930	/* Setup the initial secret */
1931	(void)SCTP_GETTIME_TIMEVAL(&time);
1932	m->time_of_secret_change = time.tv_sec;
1933
1934	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
1935		m->secret_key[0][i] = sctp_select_initial_TSN(m);
1936	}
1937	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
1938
1939	/* How long is a cookie good for ? */
1940	m->def_cookie_life = MSEC_TO_TICKS(sctp_valid_cookie_life_default);
1941	/*
1942	 * Initialize authentication parameters
1943	 */
1944	m->local_hmacs = sctp_default_supported_hmaclist();
1945	m->local_auth_chunks = sctp_alloc_chunklist();
1946	sctp_auth_set_default_chunks(m->local_auth_chunks);
1947	LIST_INIT(&m->shared_keys);
1948	/* add default NULL key as key id 0 */
1949	null_key = sctp_alloc_sharedkey();
1950	sctp_insert_sharedkey(&m->shared_keys, null_key);
1951	SCTP_INP_WUNLOCK(inp);
1952#ifdef SCTP_LOG_CLOSING
1953	sctp_log_closing(inp, NULL, 12);
1954#endif
1955	return (error);
1956}
1957
1958
1959void
1960sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
1961    struct sctp_tcb *stcb)
1962{
1963	struct sctp_nets *net;
1964	uint16_t lport, rport;
1965	struct sctppcbhead *head;
1966	struct sctp_laddr *laddr, *oladdr;
1967
1968	SCTP_TCB_UNLOCK(stcb);
1969	SCTP_INP_INFO_WLOCK();
1970	SCTP_INP_WLOCK(old_inp);
1971	SCTP_INP_WLOCK(new_inp);
1972	SCTP_TCB_LOCK(stcb);
1973
1974	new_inp->sctp_ep.time_of_secret_change =
1975	    old_inp->sctp_ep.time_of_secret_change;
1976	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
1977	    sizeof(old_inp->sctp_ep.secret_key));
1978	new_inp->sctp_ep.current_secret_number =
1979	    old_inp->sctp_ep.current_secret_number;
1980	new_inp->sctp_ep.last_secret_number =
1981	    old_inp->sctp_ep.last_secret_number;
1982	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
1983
1984	/* make it so new data pours into the new socket */
1985	stcb->sctp_socket = new_inp->sctp_socket;
1986	stcb->sctp_ep = new_inp;
1987
1988	/* Copy the port across */
1989	lport = new_inp->sctp_lport = old_inp->sctp_lport;
1990	rport = stcb->rport;
1991	/* Pull the tcb from the old association */
1992	LIST_REMOVE(stcb, sctp_tcbhash);
1993	LIST_REMOVE(stcb, sctp_tcblist);
1994
1995	/* Now insert the new_inp into the TCP connected hash */
1996	head = &sctppcbinfo.sctp_tcpephash[SCTP_PCBHASH_ALLADDR((lport + rport),
1997	    sctppcbinfo.hashtcpmark)];
1998
1999	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2000	/* Its safe to access */
2001	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2002
2003	/* Now move the tcb into the endpoint list */
2004	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2005	/*
2006	 * Question, do we even need to worry about the ep-hash since we
2007	 * only have one connection? Probably not :> so lets get rid of it
2008	 * and not suck up any kernel memory in that.
2009	 */
2010
2011	/* Ok. Let's restart timer. */
2012	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2013		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2014		    stcb, net);
2015	}
2016
2017	SCTP_INP_INFO_WUNLOCK();
2018	if (new_inp->sctp_tcbhash != NULL) {
2019		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2020		new_inp->sctp_tcbhash = NULL;
2021	}
2022	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2023		/* Subset bound, so copy in the laddr list from the old_inp */
2024		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2025			laddr = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
2026			if (laddr == NULL) {
2027				/*
2028				 * Gak, what can we do? This assoc is really
2029				 * HOSED. We probably should send an abort
2030				 * here.
2031				 */
2032				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2033				continue;
2034			}
2035			SCTP_INCR_LADDR_COUNT();
2036			bzero(laddr, sizeof(*laddr));
2037			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2038			laddr->ifa = oladdr->ifa;
2039			atomic_add_int(&laddr->ifa->refcount, 1);
2040			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2041			    sctp_nxt_addr);
2042			new_inp->laddr_count++;
2043		}
2044	}
2045	/*
2046	 * Now any running timers need to be adjusted since we really don't
2047	 * care if they are running or not just blast in the new_inp into
2048	 * all of them.
2049	 */
2050
2051	stcb->asoc.hb_timer.ep = (void *)new_inp;
2052	stcb->asoc.dack_timer.ep = (void *)new_inp;
2053	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2054	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2055	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2056	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2057	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2058	/* now what about the nets? */
2059	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2060		net->pmtu_timer.ep = (void *)new_inp;
2061		net->rxt_timer.ep = (void *)new_inp;
2062		net->fr_timer.ep = (void *)new_inp;
2063	}
2064	SCTP_INP_WUNLOCK(new_inp);
2065	SCTP_INP_WUNLOCK(old_inp);
2066}
2067
2068static int
2069sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
2070{
2071	struct sctppcbhead *head;
2072	struct sctp_inpcb *t_inp;
2073	int fnd;
2074
2075	head = &sctppcbinfo.sctp_ephash[SCTP_PCBHASH_ALLADDR(lport,
2076	    sctppcbinfo.hashmark)];
2077	LIST_FOREACH(t_inp, head, sctp_hash) {
2078		if (t_inp->sctp_lport != lport) {
2079			continue;
2080		}
2081		/* is it in the VRF in question */
2082		fnd = 0;
2083		if (t_inp->def_vrf_id == vrf_id)
2084			fnd = 1;
2085		if (!fnd)
2086			continue;
2087
2088		/* This one is in use. */
2089		/* check the v6/v4 binding issue */
2090		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
2091		    SCTP_IPV6_V6ONLY(t_inp)) {
2092			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
2093				/* collision in V6 space */
2094				return (1);
2095			} else {
2096				/* inp is BOUND_V4 no conflict */
2097				continue;
2098			}
2099		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
2100			/* t_inp is bound v4 and v6, conflict always */
2101			return (1);
2102		} else {
2103			/* t_inp is bound only V4 */
2104			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
2105			    SCTP_IPV6_V6ONLY(inp)) {
2106				/* no conflict */
2107				continue;
2108			}
2109			/* else fall through to conflict */
2110		}
2111		return (1);
2112	}
2113	return (0);
2114}
2115
2116
2117
2118int
2119sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2120    struct sctp_ifa *sctp_ifap, struct thread *p)
2121{
2122	/* bind a ep to a socket address */
2123	struct sctppcbhead *head;
2124	struct sctp_inpcb *inp, *inp_tmp;
2125	struct inpcb *ip_inp;
2126	int bindall;
2127	uint16_t lport;
2128	int error;
2129	uint32_t vrf_id;
2130
2131	lport = 0;
2132	error = 0;
2133	bindall = 1;
2134	inp = (struct sctp_inpcb *)so->so_pcb;
2135	ip_inp = (struct inpcb *)so->so_pcb;
2136#ifdef SCTP_DEBUG
2137	if (addr) {
2138		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n",
2139		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2140		SCTPDBG(SCTP_DEBUG_PCB1, "Addr :");
2141		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2142	}
2143#endif
2144	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2145		/* already did a bind, subsequent binds NOT allowed ! */
2146		return (EINVAL);
2147	}
2148	if (addr != NULL) {
2149		if (addr->sa_family == AF_INET) {
2150			struct sockaddr_in *sin;
2151
2152			/* IPV6_V6ONLY socket? */
2153			if (SCTP_IPV6_V6ONLY(ip_inp)) {
2154				return (EINVAL);
2155			}
2156			if (addr->sa_len != sizeof(*sin))
2157				return (EINVAL);
2158
2159			sin = (struct sockaddr_in *)addr;
2160			lport = sin->sin_port;
2161
2162			if (sin->sin_addr.s_addr != INADDR_ANY) {
2163				bindall = 0;
2164			}
2165		} else if (addr->sa_family == AF_INET6) {
2166			/* Only for pure IPv6 Address. (No IPv4 Mapped!) */
2167			struct sockaddr_in6 *sin6;
2168
2169			sin6 = (struct sockaddr_in6 *)addr;
2170
2171			if (addr->sa_len != sizeof(*sin6))
2172				return (EINVAL);
2173
2174			lport = sin6->sin6_port;
2175			if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2176				bindall = 0;
2177				/* KAME hack: embed scopeid */
2178				if (sa6_embedscope(sin6, ip6_use_defzone) != 0)
2179					return (EINVAL);
2180			}
2181			/* this must be cleared for ifa_ifwithaddr() */
2182			sin6->sin6_scope_id = 0;
2183		} else {
2184			return (EAFNOSUPPORT);
2185		}
2186	}
2187	/* Setup a vrf_id to be the default for the non-bind-all case. */
2188	vrf_id = inp->def_vrf_id;
2189
2190	SCTP_INP_INFO_WLOCK();
2191	SCTP_INP_WLOCK(inp);
2192	/* increase our count due to the unlock we do */
2193	SCTP_INP_INCR_REF(inp);
2194	if (lport) {
2195		/*
2196		 * Did the caller specify a port? if so we must see if a ep
2197		 * already has this one bound.
2198		 */
2199		/* got to be root to get at low ports */
2200		if (ntohs(lport) < IPPORT_RESERVED) {
2201			if (p && (error =
2202			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2203			    )) {
2204				SCTP_INP_DECR_REF(inp);
2205				SCTP_INP_WUNLOCK(inp);
2206				SCTP_INP_INFO_WUNLOCK();
2207				return (error);
2208			}
2209		}
2210		if (p == NULL) {
2211			SCTP_INP_DECR_REF(inp);
2212			SCTP_INP_WUNLOCK(inp);
2213			SCTP_INP_INFO_WUNLOCK();
2214			return (error);
2215		}
2216		SCTP_INP_WUNLOCK(inp);
2217		if (bindall) {
2218			vrf_id = inp->def_vrf_id;
2219			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2220			if (inp_tmp != NULL) {
2221				/*
2222				 * lock guy returned and lower count note
2223				 * that we are not bound so inp_tmp should
2224				 * NEVER be inp. And it is this inp
2225				 * (inp_tmp) that gets the reference bump,
2226				 * so we must lower it.
2227				 */
2228				SCTP_INP_DECR_REF(inp_tmp);
2229				SCTP_INP_DECR_REF(inp);
2230				/* unlock info */
2231				SCTP_INP_INFO_WUNLOCK();
2232				return (EADDRINUSE);
2233			}
2234		} else {
2235			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2236			if (inp_tmp != NULL) {
2237				/*
2238				 * lock guy returned and lower count note
2239				 * that we are not bound so inp_tmp should
2240				 * NEVER be inp. And it is this inp
2241				 * (inp_tmp) that gets the reference bump,
2242				 * so we must lower it.
2243				 */
2244				SCTP_INP_DECR_REF(inp_tmp);
2245				SCTP_INP_DECR_REF(inp);
2246				/* unlock info */
2247				SCTP_INP_INFO_WUNLOCK();
2248				return (EADDRINUSE);
2249			}
2250		}
2251		SCTP_INP_WLOCK(inp);
2252		if (bindall) {
2253			/* verify that no lport is not used by a singleton */
2254			if (sctp_isport_inuse(inp, lport, vrf_id)) {
2255				/* Sorry someone already has this one bound */
2256				SCTP_INP_DECR_REF(inp);
2257				SCTP_INP_WUNLOCK(inp);
2258				SCTP_INP_INFO_WUNLOCK();
2259				return (EADDRINUSE);
2260			}
2261		}
2262	} else {
2263		uint16_t first, last, candidate;
2264		uint16_t count;
2265		int done;
2266
2267		if (ip_inp->inp_flags & INP_HIGHPORT) {
2268			first = ipport_hifirstauto;
2269			last = ipport_hilastauto;
2270		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2271			if (p && (error =
2272			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2273			    )) {
2274				SCTP_INP_DECR_REF(inp);
2275				SCTP_INP_WUNLOCK(inp);
2276				SCTP_INP_INFO_WUNLOCK();
2277				return (error);
2278			}
2279			first = ipport_lowfirstauto;
2280			last = ipport_lowlastauto;
2281		} else {
2282			first = ipport_firstauto;
2283			last = ipport_lastauto;
2284		}
2285		if (first > last) {
2286			uint16_t temp;
2287
2288			temp = first;
2289			first = last;
2290			last = temp;
2291		}
2292		count = last - first + 1;	/* number of candidates */
2293		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2294
2295		done = 0;
2296		while (!done) {
2297			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == 0) {
2298				done = 1;
2299			}
2300			if (!done) {
2301				if (--count == 0) {
2302					SCTP_INP_DECR_REF(inp);
2303					SCTP_INP_WUNLOCK(inp);
2304					SCTP_INP_INFO_WUNLOCK();
2305					return (EADDRINUSE);
2306				}
2307				if (candidate == last)
2308					candidate = first;
2309				else
2310					candidate = candidate + 1;
2311			}
2312		}
2313		lport = htons(candidate);
2314	}
2315	SCTP_INP_DECR_REF(inp);
2316	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2317	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2318		/*
2319		 * this really should not happen. The guy did a non-blocking
2320		 * bind and then did a close at the same time.
2321		 */
2322		SCTP_INP_WUNLOCK(inp);
2323		SCTP_INP_INFO_WUNLOCK();
2324		return (EINVAL);
2325	}
2326	/* ok we look clear to give out this port, so lets setup the binding */
2327	if (bindall) {
2328		/* binding to all addresses, so just set in the proper flags */
2329		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
2330		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2331		/* set the automatic addr changes from kernel flag */
2332		if (sctp_auto_asconf == 0) {
2333			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2334		} else {
2335			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2336		}
2337	} else {
2338		/*
2339		 * bind specific, make sure flags is off and add a new
2340		 * address structure to the sctp_addr_list inside the ep
2341		 * structure.
2342		 *
2343		 * We will need to allocate one and insert it at the head. The
2344		 * socketopt call can just insert new addresses in there as
2345		 * well. It will also have to do the embed scope kame hack
2346		 * too (before adding).
2347		 */
2348		struct sctp_ifa *ifa;
2349		struct sockaddr_storage store_sa;
2350
2351		memset(&store_sa, 0, sizeof(store_sa));
2352		if (addr->sa_family == AF_INET) {
2353			struct sockaddr_in *sin;
2354
2355			sin = (struct sockaddr_in *)&store_sa;
2356			memcpy(sin, addr, sizeof(struct sockaddr_in));
2357			sin->sin_port = 0;
2358		} else if (addr->sa_family == AF_INET6) {
2359			struct sockaddr_in6 *sin6;
2360
2361			sin6 = (struct sockaddr_in6 *)&store_sa;
2362			memcpy(sin6, addr, sizeof(struct sockaddr_in6));
2363			sin6->sin6_port = 0;
2364		}
2365		/*
2366		 * first find the interface with the bound address need to
2367		 * zero out the port to find the address! yuck! can't do
2368		 * this earlier since need port for sctp_pcb_findep()
2369		 */
2370		if (sctp_ifap)
2371			ifa = sctp_ifap;
2372		else
2373			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
2374			    vrf_id, 0);
2375		if (ifa == NULL) {
2376			/* Can't find an interface with that address */
2377			SCTP_INP_WUNLOCK(inp);
2378			SCTP_INP_INFO_WUNLOCK();
2379			return (EADDRNOTAVAIL);
2380		}
2381		if (addr->sa_family == AF_INET6) {
2382			/* GAK, more FIXME IFA lock? */
2383			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
2384				/* Can't bind a non-existent addr. */
2385				SCTP_INP_WUNLOCK(inp);
2386				SCTP_INP_INFO_WUNLOCK();
2387				return (EINVAL);
2388			}
2389		}
2390		/* we're not bound all */
2391		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
2392		/* allow bindx() to send ASCONF's for binding changes */
2393		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2394		/* clear automatic addr changes from kernel flag */
2395		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2396
2397		/* add this address to the endpoint list */
2398		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
2399		if (error != 0) {
2400			SCTP_INP_WUNLOCK(inp);
2401			SCTP_INP_INFO_WUNLOCK();
2402			return (error);
2403		}
2404		inp->laddr_count++;
2405	}
2406	/* find the bucket */
2407	head = &sctppcbinfo.sctp_ephash[SCTP_PCBHASH_ALLADDR(lport,
2408	    sctppcbinfo.hashmark)];
2409	/* put it in the bucket */
2410	LIST_INSERT_HEAD(head, inp, sctp_hash);
2411	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d\n",
2412	    head, ntohs(lport));
2413	/* set in the port */
2414	inp->sctp_lport = lport;
2415
2416	/* turn off just the unbound flag */
2417	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2418	SCTP_INP_WUNLOCK(inp);
2419	SCTP_INP_INFO_WUNLOCK();
2420	return (0);
2421}
2422
2423
2424static void
2425sctp_iterator_inp_being_freed(struct sctp_inpcb *inp, struct sctp_inpcb *inp_next)
2426{
2427	struct sctp_iterator *it;
2428
2429	/*
2430	 * We enter with the only the ITERATOR_LOCK in place and a write
2431	 * lock on the inp_info stuff.
2432	 */
2433
2434	/*
2435	 * Go through all iterators, we must do this since it is possible
2436	 * that some iterator does NOT have the lock, but is waiting for it.
2437	 * And the one that had the lock has either moved in the last
2438	 * iteration or we just cleared it above. We need to find all of
2439	 * those guys. The list of iterators should never be very big
2440	 * though.
2441	 */
2442	TAILQ_FOREACH(it, &sctppcbinfo.iteratorhead, sctp_nxt_itr) {
2443		if (it == inp->inp_starting_point_for_iterator)
2444			/* skip this guy, he's special */
2445			continue;
2446		if (it->inp == inp) {
2447			/*
2448			 * This is tricky and we DON'T lock the iterator.
2449			 * Reason is he's running but waiting for me since
2450			 * inp->inp_starting_point_for_iterator has the lock
2451			 * on me (the guy above we skipped). This tells us
2452			 * its is not running but waiting for
2453			 * inp->inp_starting_point_for_iterator to be
2454			 * released by the guy that does have our INP in a
2455			 * lock.
2456			 */
2457			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
2458				it->inp = NULL;
2459				it->stcb = NULL;
2460			} else {
2461				/* set him up to do the next guy not me */
2462				it->inp = inp_next;
2463				it->stcb = NULL;
2464			}
2465		}
2466	}
2467	it = inp->inp_starting_point_for_iterator;
2468	if (it) {
2469		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
2470			it->inp = NULL;
2471		} else {
2472			it->inp = inp_next;
2473		}
2474		it->stcb = NULL;
2475	}
2476}
2477
2478/* release sctp_inpcb unbind the port */
2479void
2480sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
2481{
2482	/*
2483	 * Here we free a endpoint. We must find it (if it is in the Hash
2484	 * table) and remove it from there. Then we must also find it in the
2485	 * overall list and remove it from there. After all removals are
2486	 * complete then any timer has to be stopped. Then start the actual
2487	 * freeing. a) Any local lists. b) Any associations. c) The hash of
2488	 * all associations. d) finally the ep itself.
2489	 */
2490	struct sctp_pcb *m;
2491	struct sctp_inpcb *inp_save;
2492	struct sctp_tcb *asoc, *nasoc;
2493	struct sctp_laddr *laddr, *nladdr;
2494	struct inpcb *ip_pcb;
2495	struct socket *so;
2496
2497	struct sctp_queued_to_read *sq;
2498
2499
2500	int cnt;
2501	sctp_sharedkey_t *shared_key;
2502
2503
2504#ifdef SCTP_LOG_CLOSING
2505	sctp_log_closing(inp, NULL, 0);
2506#endif
2507
2508	SCTP_ITERATOR_LOCK();
2509	so = inp->sctp_socket;
2510	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2511		/* been here before.. eeks.. get out of here */
2512		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
2513		SCTP_ITERATOR_UNLOCK();
2514#ifdef SCTP_LOG_CLOSING
2515		sctp_log_closing(inp, NULL, 1);
2516#endif
2517		return;
2518	}
2519	SCTP_ASOC_CREATE_LOCK(inp);
2520	SCTP_INP_INFO_WLOCK();
2521
2522	SCTP_INP_WLOCK(inp);
2523	/* First time through we have the socket lock, after that no more. */
2524	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
2525		/*
2526		 * Once we are in we can remove the flag from = 1 is only
2527		 * passed from the actual closing routines that are called
2528		 * via the sockets layer.
2529		 */
2530		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
2531	}
2532	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
2533	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
2534
2535	if (inp->control) {
2536		sctp_m_freem(inp->control);
2537		inp->control = NULL;
2538	}
2539	if (inp->pkt) {
2540		sctp_m_freem(inp->pkt);
2541		inp->pkt = NULL;
2542	}
2543	m = &inp->sctp_ep;
2544	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
2545					 * here but I will be nice :> (i.e.
2546					 * ip_pcb = ep;) */
2547	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
2548		int cnt_in_sd;
2549
2550		cnt_in_sd = 0;
2551		for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
2552		    asoc = nasoc) {
2553			nasoc = LIST_NEXT(asoc, sctp_tcblist);
2554			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2555				/* Skip guys being freed */
2556				asoc->sctp_socket = NULL;
2557				cnt_in_sd++;
2558				continue;
2559			}
2560			if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
2561			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) {
2562				/* Just abandon things in the front states */
2563				if (asoc->asoc.total_output_queue_size == 0) {
2564					sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_2);
2565					continue;
2566				}
2567			}
2568			SCTP_TCB_LOCK(asoc);
2569			/* Disconnect the socket please */
2570			asoc->sctp_socket = NULL;
2571			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
2572			if ((asoc->asoc.size_on_reasm_queue > 0) ||
2573			    (asoc->asoc.control_pdapi) ||
2574			    (asoc->asoc.size_on_all_streams > 0) ||
2575			    (so && (so->so_rcv.sb_cc > 0))
2576			    ) {
2577				/* Left with Data unread */
2578				struct mbuf *op_err;
2579
2580				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2581				    0, M_DONTWAIT, 1, MT_DATA);
2582				if (op_err) {
2583					/* Fill in the user initiated abort */
2584					struct sctp_paramhdr *ph;
2585					uint32_t *ippp;
2586
2587					SCTP_BUF_LEN(op_err) =
2588					    sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
2589					ph = mtod(op_err,
2590					    struct sctp_paramhdr *);
2591					ph->param_type = htons(
2592					    SCTP_CAUSE_USER_INITIATED_ABT);
2593					ph->param_length = htons(SCTP_BUF_LEN(op_err));
2594					ippp = (uint32_t *) (ph + 1);
2595					*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3);
2596				}
2597				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
2598				sctp_send_abort_tcb(asoc, op_err);
2599				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2600				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2601				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2602					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2603				}
2604				sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4);
2605				continue;
2606			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
2607				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
2608				    (asoc->asoc.stream_queue_cnt == 0)
2609			    ) {
2610				if (asoc->asoc.locked_on_sending) {
2611					goto abort_anyway;
2612				}
2613				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
2614				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
2615					/*
2616					 * there is nothing queued to send,
2617					 * so I send shutdown
2618					 */
2619					sctp_send_shutdown(asoc, asoc->asoc.primary_destination);
2620					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2621					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2622						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2623					}
2624					asoc->asoc.state = SCTP_STATE_SHUTDOWN_SENT;
2625					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
2626					    asoc->asoc.primary_destination);
2627					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
2628					    asoc->asoc.primary_destination);
2629					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR);
2630				}
2631			} else {
2632				/* mark into shutdown pending */
2633				struct sctp_stream_queue_pending *sp;
2634
2635				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
2636				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
2637				    asoc->asoc.primary_destination);
2638				if (asoc->asoc.locked_on_sending) {
2639					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
2640					    sctp_streamhead);
2641					if (sp == NULL) {
2642						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
2643						    asoc->asoc.locked_on_sending,
2644						    asoc->asoc.locked_on_sending->stream_no);
2645					} else {
2646						if ((sp->length == 0) && (sp->msg_is_complete == 0))
2647							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
2648					}
2649				}
2650				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
2651				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
2652				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
2653					struct mbuf *op_err;
2654
2655			abort_anyway:
2656					op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2657					    0, M_DONTWAIT, 1, MT_DATA);
2658					if (op_err) {
2659						/*
2660						 * Fill in the user
2661						 * initiated abort
2662						 */
2663						struct sctp_paramhdr *ph;
2664						uint32_t *ippp;
2665
2666						SCTP_BUF_LEN(op_err) =
2667						    (sizeof(struct sctp_paramhdr) +
2668						    sizeof(uint32_t));
2669						ph = mtod(op_err,
2670						    struct sctp_paramhdr *);
2671						ph->param_type = htons(
2672						    SCTP_CAUSE_USER_INITIATED_ABT);
2673						ph->param_length = htons(SCTP_BUF_LEN(op_err));
2674						ippp = (uint32_t *) (ph + 1);
2675						*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5);
2676					}
2677					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
2678					sctp_send_abort_tcb(asoc, op_err);
2679					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2680					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2681					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2682						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2683					}
2684					sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_6);
2685					continue;
2686				}
2687			}
2688			cnt_in_sd++;
2689			SCTP_TCB_UNLOCK(asoc);
2690		}
2691		/* now is there some left in our SHUTDOWN state? */
2692		if (cnt_in_sd) {
2693			SCTP_INP_WUNLOCK(inp);
2694			SCTP_ASOC_CREATE_UNLOCK(inp);
2695			SCTP_INP_INFO_WUNLOCK();
2696			SCTP_ITERATOR_UNLOCK();
2697#ifdef SCTP_LOG_CLOSING
2698			sctp_log_closing(inp, NULL, 2);
2699#endif
2700			return;
2701		}
2702	}
2703	inp->sctp_socket = NULL;
2704	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
2705	    SCTP_PCB_FLAGS_UNBOUND) {
2706		/*
2707		 * ok, this guy has been bound. It's port is somewhere in
2708		 * the sctppcbinfo hash table. Remove it!
2709		 */
2710		LIST_REMOVE(inp, sctp_hash);
2711		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
2712	}
2713	/*
2714	 * If there is a timer running to kill us, forget it, since it may
2715	 * have a contest on the INP lock.. which would cause us to die ...
2716	 */
2717	cnt = 0;
2718	for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
2719	    asoc = nasoc) {
2720		nasoc = LIST_NEXT(asoc, sctp_tcblist);
2721		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2722			cnt++;
2723			continue;
2724		}
2725		/* Free associations that are NOT killing us */
2726		SCTP_TCB_LOCK(asoc);
2727		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
2728		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
2729			struct mbuf *op_err;
2730			uint32_t *ippp;
2731
2732			op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2733			    0, M_DONTWAIT, 1, MT_DATA);
2734			if (op_err) {
2735				/* Fill in the user initiated abort */
2736				struct sctp_paramhdr *ph;
2737
2738				SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) +
2739				    sizeof(uint32_t));
2740				ph = mtod(op_err, struct sctp_paramhdr *);
2741				ph->param_type = htons(
2742				    SCTP_CAUSE_USER_INITIATED_ABT);
2743				ph->param_length = htons(SCTP_BUF_LEN(op_err));
2744				ippp = (uint32_t *) (ph + 1);
2745				*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7);
2746
2747			}
2748			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
2749			sctp_send_abort_tcb(asoc, op_err);
2750			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2751		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2752			cnt++;
2753			SCTP_TCB_UNLOCK(asoc);
2754			continue;
2755		}
2756		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2757		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2758			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2759		}
2760		sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8);
2761	}
2762	if (cnt) {
2763		/* Ok we have someone out there that will kill us */
2764		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2765		SCTP_INP_WUNLOCK(inp);
2766		SCTP_ASOC_CREATE_UNLOCK(inp);
2767		SCTP_INP_INFO_WUNLOCK();
2768		SCTP_ITERATOR_UNLOCK();
2769#ifdef SCTP_LOG_CLOSING
2770		sctp_log_closing(inp, NULL, 3);
2771#endif
2772		return;
2773	}
2774	if ((inp->refcount) || (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
2775		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2776		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
2777		SCTP_INP_WUNLOCK(inp);
2778		SCTP_ASOC_CREATE_UNLOCK(inp);
2779		SCTP_INP_INFO_WUNLOCK();
2780		SCTP_ITERATOR_UNLOCK();
2781#ifdef SCTP_LOG_CLOSING
2782		sctp_log_closing(inp, NULL, 4);
2783#endif
2784		return;
2785	}
2786	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2787	inp->sctp_ep.signature_change.type = 0;
2788	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
2789
2790#ifdef SCTP_LOG_CLOSING
2791	sctp_log_closing(inp, NULL, 5);
2792#endif
2793
2794	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2795	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NONE;
2796	/* Clear the read queue */
2797	/* sa_ignore FREED_MEMORY */
2798	while ((sq = TAILQ_FIRST(&inp->read_queue)) != NULL) {
2799		/* Its only abandoned if it had data left */
2800		if (sq->length)
2801			SCTP_STAT_INCR(sctps_left_abandon);
2802
2803		TAILQ_REMOVE(&inp->read_queue, sq, next);
2804		sctp_free_remote_addr(sq->whoFrom);
2805		if (so)
2806			so->so_rcv.sb_cc -= sq->length;
2807		if (sq->data) {
2808			sctp_m_freem(sq->data);
2809			sq->data = NULL;
2810		}
2811		/*
2812		 * no need to free the net count, since at this point all
2813		 * assoc's are gone.
2814		 */
2815		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, sq);
2816		SCTP_DECR_READQ_COUNT();
2817	}
2818	/* Now the sctp_pcb things */
2819	/*
2820	 * free each asoc if it is not already closed/free. we can't use the
2821	 * macro here since le_next will get freed as part of the
2822	 * sctp_free_assoc() call.
2823	 */
2824	cnt = 0;
2825	if (so) {
2826#ifdef IPSEC
2827		ipsec4_delete_pcbpolicy(ip_pcb);
2828#endif				/* IPSEC */
2829
2830		/* Unlocks not needed since the socket is gone now */
2831	}
2832	if (ip_pcb->inp_options) {
2833		(void)sctp_m_free(ip_pcb->inp_options);
2834		ip_pcb->inp_options = 0;
2835	}
2836	if (ip_pcb->inp_moptions) {
2837		inp_freemoptions(ip_pcb->inp_moptions);
2838		ip_pcb->inp_moptions = 0;
2839	}
2840#ifdef INET6
2841	if (ip_pcb->inp_vflag & INP_IPV6) {
2842		struct in6pcb *in6p;
2843
2844		in6p = (struct in6pcb *)inp;
2845		ip6_freepcbopts(in6p->in6p_outputopts);
2846	}
2847#endif				/* INET6 */
2848	ip_pcb->inp_vflag = 0;
2849	/* free up authentication fields */
2850	if (inp->sctp_ep.local_auth_chunks != NULL)
2851		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
2852	if (inp->sctp_ep.local_hmacs != NULL)
2853		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
2854
2855	shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
2856	while (shared_key) {
2857		LIST_REMOVE(shared_key, next);
2858		sctp_free_sharedkey(shared_key);
2859		/* sa_ignore FREED_MEMORY */
2860		shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
2861	}
2862
2863	inp_save = LIST_NEXT(inp, sctp_list);
2864	LIST_REMOVE(inp, sctp_list);
2865
2866	/* fix any iterators only after out of the list */
2867	sctp_iterator_inp_being_freed(inp, inp_save);
2868	/*
2869	 * if we have an address list the following will free the list of
2870	 * ifaddr's that are set into this ep. Again macro limitations here,
2871	 * since the LIST_FOREACH could be a bad idea.
2872	 */
2873	for ((laddr = LIST_FIRST(&inp->sctp_addr_list)); laddr != NULL;
2874	    laddr = nladdr) {
2875		nladdr = LIST_NEXT(laddr, sctp_nxt_addr);
2876		sctp_remove_laddr(laddr);
2877	}
2878
2879#ifdef SCTP_TRACK_FREED_ASOCS
2880	/* TEMP CODE */
2881	for ((asoc = LIST_FIRST(&inp->sctp_asoc_free_list)); asoc != NULL;
2882	    asoc = nasoc) {
2883		nasoc = LIST_NEXT(asoc, sctp_tcblist);
2884		LIST_REMOVE(asoc, sctp_tcblist);
2885		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, asoc);
2886		SCTP_DECR_ASOC_COUNT();
2887	}
2888	/* *** END TEMP CODE *** */
2889#endif
2890	/* Now lets see about freeing the EP hash table. */
2891	if (inp->sctp_tcbhash != NULL) {
2892		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
2893		inp->sctp_tcbhash = NULL;
2894	}
2895	/* Now we must put the ep memory back into the zone pool */
2896	INP_LOCK_DESTROY(&inp->ip_inp.inp);
2897	SCTP_INP_LOCK_DESTROY(inp);
2898	SCTP_INP_READ_DESTROY(inp);
2899	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
2900	SCTP_INP_INFO_WUNLOCK();
2901
2902	SCTP_ITERATOR_UNLOCK();
2903
2904	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
2905	SCTP_DECR_EP_COUNT();
2906
2907}
2908
2909
2910struct sctp_nets *
2911sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
2912{
2913	struct sctp_nets *net;
2914
2915	/* locate the address */
2916	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2917		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
2918			return (net);
2919	}
2920	return (NULL);
2921}
2922
2923
2924int
2925sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
2926{
2927	struct sctp_ifa *sctp_ifa;
2928
2929	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, 0);
2930	if (sctp_ifa) {
2931		return (1);
2932	} else {
2933		return (0);
2934	}
2935}
2936
2937void
2938sctp_set_initial_cc_param(struct sctp_tcb *stcb, struct sctp_nets *net)
2939{
2940	net->cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND));
2941	/* we always get at LEAST 2 MTU's */
2942	if (net->cwnd < (2 * net->mtu)) {
2943		net->cwnd = 2 * net->mtu;
2944	}
2945	net->ssthresh = stcb->asoc.peers_rwnd;
2946}
2947
2948
2949
2950/*
2951 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
2952 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
2953 * stats of stuff.
2954 */
2955int
2956sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
2957    int set_scope, int from)
2958{
2959	/*
2960	 * The following is redundant to the same lines in the
2961	 * sctp_aloc_assoc() but is needed since other's call the add
2962	 * address function
2963	 */
2964	struct sctp_nets *net, *netfirst;
2965	int addr_inscope;
2966
2967	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
2968	    from);
2969	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
2970
2971	netfirst = sctp_findnet(stcb, newaddr);
2972	if (netfirst) {
2973		/*
2974		 * Lie and return ok, we don't want to make the association
2975		 * go away for this behavior. It will happen in the TCP
2976		 * model in a connected socket. It does not reach the hash
2977		 * table until after the association is built so it can't be
2978		 * found. Mark as reachable, since the initial creation will
2979		 * have been cleared and the NOT_IN_ASSOC flag will have
2980		 * been added... and we don't want to end up removing it
2981		 * back out.
2982		 */
2983		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
2984			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
2985			    SCTP_ADDR_UNCONFIRMED);
2986		} else {
2987			netfirst->dest_state = SCTP_ADDR_REACHABLE;
2988		}
2989
2990		return (0);
2991	}
2992	addr_inscope = 1;
2993	if (newaddr->sa_family == AF_INET) {
2994		struct sockaddr_in *sin;
2995
2996		sin = (struct sockaddr_in *)newaddr;
2997		if (sin->sin_addr.s_addr == 0) {
2998			/* Invalid address */
2999			return (-1);
3000		}
3001		/* zero out the bzero area */
3002		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3003
3004		/* assure len is set */
3005		sin->sin_len = sizeof(struct sockaddr_in);
3006		if (set_scope) {
3007#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3008			stcb->ipv4_local_scope = 1;
3009#else
3010			if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3011				stcb->asoc.ipv4_local_scope = 1;
3012			}
3013#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3014		} else {
3015			/* Validate the address is in scope */
3016			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3017			    (stcb->asoc.ipv4_local_scope == 0)) {
3018				addr_inscope = 0;
3019			}
3020		}
3021	} else if (newaddr->sa_family == AF_INET6) {
3022		struct sockaddr_in6 *sin6;
3023
3024		sin6 = (struct sockaddr_in6 *)newaddr;
3025		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3026			/* Invalid address */
3027			return (-1);
3028		}
3029		/* assure len is set */
3030		sin6->sin6_len = sizeof(struct sockaddr_in6);
3031		if (set_scope) {
3032			if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3033				stcb->asoc.loopback_scope = 1;
3034				stcb->asoc.local_scope = 0;
3035				stcb->asoc.ipv4_local_scope = 1;
3036				stcb->asoc.site_scope = 1;
3037			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3038				/*
3039				 * If the new destination is a LINK_LOCAL we
3040				 * must have common site scope. Don't set
3041				 * the local scope since we may not share
3042				 * all links, only loopback can do this.
3043				 * Links on the local network would also be
3044				 * on our private network for v4 too.
3045				 */
3046				stcb->asoc.ipv4_local_scope = 1;
3047				stcb->asoc.site_scope = 1;
3048			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3049				/*
3050				 * If the new destination is SITE_LOCAL then
3051				 * we must have site scope in common.
3052				 */
3053				stcb->asoc.site_scope = 1;
3054			}
3055		} else {
3056			/* Validate the address is in scope */
3057			if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3058			    (stcb->asoc.loopback_scope == 0)) {
3059				addr_inscope = 0;
3060			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3061			    (stcb->asoc.local_scope == 0)) {
3062				addr_inscope = 0;
3063			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3064			    (stcb->asoc.site_scope == 0)) {
3065				addr_inscope = 0;
3066			}
3067		}
3068	} else {
3069		/* not supported family type */
3070		return (-1);
3071	}
3072	net = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_net, struct sctp_nets);
3073	if (net == NULL) {
3074		return (-1);
3075	}
3076	SCTP_INCR_RADDR_COUNT();
3077	bzero(net, sizeof(*net));
3078	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3079	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3080	if (newaddr->sa_family == AF_INET) {
3081		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3082	} else if (newaddr->sa_family == AF_INET6) {
3083		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3084	}
3085	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3086	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3087		stcb->asoc.loopback_scope = 1;
3088		stcb->asoc.ipv4_local_scope = 1;
3089		stcb->asoc.local_scope = 0;
3090		stcb->asoc.site_scope = 1;
3091		addr_inscope = 1;
3092	}
3093	net->failure_threshold = stcb->asoc.def_net_failure;
3094	if (addr_inscope == 0) {
3095		net->dest_state = (SCTP_ADDR_REACHABLE |
3096		    SCTP_ADDR_OUT_OF_SCOPE);
3097	} else {
3098		if (from == SCTP_ADDR_IS_CONFIRMED)
3099			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3100			net->dest_state = SCTP_ADDR_REACHABLE;
3101		else
3102			net->dest_state = SCTP_ADDR_REACHABLE |
3103			    SCTP_ADDR_UNCONFIRMED;
3104	}
3105	/*
3106	 * We set this to 0, the timer code knows that this means its an
3107	 * initial value
3108	 */
3109	net->RTO = 0;
3110	net->RTO_measured = 0;
3111	stcb->asoc.numnets++;
3112	*(&net->ref_count) = 1;
3113	net->tos_flowlabel = 0;
3114#ifdef INET
3115	if (newaddr->sa_family == AF_INET)
3116		net->tos_flowlabel = stcb->asoc.default_tos;
3117#endif
3118#ifdef INET6
3119	if (newaddr->sa_family == AF_INET6)
3120		net->tos_flowlabel = stcb->asoc.default_flowlabel;
3121#endif
3122	/* Init the timer structure */
3123	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3124	SCTP_OS_TIMER_INIT(&net->fr_timer.timer);
3125	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3126
3127	/* Now generate a route for this guy */
3128	/* KAME hack: embed scopeid */
3129	if (newaddr->sa_family == AF_INET6) {
3130		struct sockaddr_in6 *sin6;
3131
3132		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3133		(void)sa6_embedscope(sin6, ip6_use_defzone);
3134		sin6->sin6_scope_id = 0;
3135	}
3136	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3137
3138	if (newaddr->sa_family == AF_INET6) {
3139		struct sockaddr_in6 *sin6;
3140
3141		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3142		(void)sa6_recoverscope(sin6);
3143	}
3144	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3145		/* Get source address */
3146		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3147		    stcb,
3148		    (sctp_route_t *) & net->ro,
3149		    net,
3150		    0,
3151		    stcb->asoc.vrf_id);
3152		/* Now get the interface MTU */
3153		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3154			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3155		} else {
3156			net->mtu = 0;
3157		}
3158#ifdef SCTP_PRINT_FOR_B_AND_M
3159		SCTP_PRINTF("We have found an interface mtu of %d\n", net->mtu);
3160#endif
3161		if (net->mtu == 0) {
3162			/* Huh ?? */
3163			net->mtu = SCTP_DEFAULT_MTU;
3164		} else {
3165			uint32_t rmtu;
3166
3167			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3168#ifdef SCTP_PRINT_FOR_B_AND_M
3169			SCTP_PRINTF("The route mtu is %d\n", rmtu);
3170#endif
3171			if (rmtu == 0) {
3172				/*
3173				 * Start things off to match mtu of
3174				 * interface please.
3175				 */
3176				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3177				    net->ro.ro_rt, net->mtu);
3178			} else {
3179				/*
3180				 * we take the route mtu over the interface,
3181				 * since the route may be leading out the
3182				 * loopback, or a different interface.
3183				 */
3184				net->mtu = rmtu;
3185			}
3186		}
3187		if (from == SCTP_ALLOC_ASOC) {
3188#ifdef SCTP_PRINT_FOR_B_AND_M
3189			SCTP_PRINTF("New assoc sets mtu to :%d\n", net->mtu);
3190#endif
3191			stcb->asoc.smallest_mtu = net->mtu;
3192		}
3193	} else {
3194		net->mtu = stcb->asoc.smallest_mtu;
3195	}
3196	if (stcb->asoc.smallest_mtu > net->mtu) {
3197#ifdef SCTP_PRINT_FOR_B_AND_M
3198		SCTP_PRINTF("new address mtu:%d smaller than smallest:%d\n",
3199		    net->mtu, stcb->asoc.smallest_mtu);
3200#endif
3201		stcb->asoc.smallest_mtu = net->mtu;
3202	}
3203	/*
3204	 * We take the max of the burst limit times a MTU or the
3205	 * INITIAL_CWND. We then limit this to 4 MTU's of sending.
3206	 */
3207	sctp_set_initial_cc_param(stcb, net);
3208
3209
3210	if (sctp_logging_level & (SCTP_CWND_MONITOR_ENABLE | SCTP_CWND_LOGGING_ENABLE)) {
3211		sctp_log_cwnd(stcb, net, 0, SCTP_CWND_INITIALIZATION);
3212	}
3213	/*
3214	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3215	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3216	 */
3217	net->find_pseudo_cumack = 1;
3218	net->find_rtx_pseudo_cumack = 1;
3219	net->src_addr_selected = 0;
3220	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3221	if (net->ro.ro_rt == NULL) {
3222		/* Since we have no route put it at the back */
3223		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3224	} else if (netfirst == NULL) {
3225		/* We are the first one in the pool. */
3226		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3227	} else if (netfirst->ro.ro_rt == NULL) {
3228		/*
3229		 * First one has NO route. Place this one ahead of the first
3230		 * one.
3231		 */
3232		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3233	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
3234		/*
3235		 * This one has a different interface than the one at the
3236		 * top of the list. Place it ahead.
3237		 */
3238		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3239	} else {
3240		/*
3241		 * Ok we have the same interface as the first one. Move
3242		 * forward until we find either a) one with a NULL route...
3243		 * insert ahead of that b) one with a different ifp.. insert
3244		 * after that. c) end of the list.. insert at the tail.
3245		 */
3246		struct sctp_nets *netlook;
3247
3248		do {
3249			netlook = TAILQ_NEXT(netfirst, sctp_next);
3250			if (netlook == NULL) {
3251				/* End of the list */
3252				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3253				break;
3254			} else if (netlook->ro.ro_rt == NULL) {
3255				/* next one has NO route */
3256				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
3257				break;
3258			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
3259				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
3260				    net, sctp_next);
3261				break;
3262			}
3263			/* Shift forward */
3264			netfirst = netlook;
3265		} while (netlook != NULL);
3266	}
3267
3268	/* got to have a primary set */
3269	if (stcb->asoc.primary_destination == 0) {
3270		stcb->asoc.primary_destination = net;
3271	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
3272		    (net->ro.ro_rt) &&
3273	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
3274		/* No route to current primary adopt new primary */
3275		stcb->asoc.primary_destination = net;
3276	}
3277	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb,
3278	    net);
3279	/* Validate primary is first */
3280	net = TAILQ_FIRST(&stcb->asoc.nets);
3281	if ((net != stcb->asoc.primary_destination) &&
3282	    (stcb->asoc.primary_destination)) {
3283		/*
3284		 * first one on the list is NOT the primary sctp_cmpaddr()
3285		 * is much more efficent if the primary is the first on the
3286		 * list, make it so.
3287		 */
3288		TAILQ_REMOVE(&stcb->asoc.nets,
3289		    stcb->asoc.primary_destination, sctp_next);
3290		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
3291		    stcb->asoc.primary_destination, sctp_next);
3292	}
3293	return (0);
3294}
3295
3296
3297/*
3298 * allocate an association and add it to the endpoint. The caller must be
3299 * careful to add all additional addresses once they are know right away or
3300 * else the assoc will be may experience a blackout scenario.
3301 */
3302struct sctp_tcb *
3303sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
3304    int for_a_init, int *error, uint32_t override_tag, uint32_t vrf_id)
3305{
3306	struct sctp_tcb *stcb;
3307	struct sctp_association *asoc;
3308	struct sctpasochead *head;
3309	uint16_t rport;
3310	int err;
3311
3312	/*
3313	 * Assumption made here: Caller has done a
3314	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
3315	 * address does not exist already.
3316	 */
3317	if (sctppcbinfo.ipi_count_asoc >= SCTP_MAX_NUM_OF_ASOC) {
3318		/* Hit max assoc, sorry no more */
3319		*error = ENOBUFS;
3320		return (NULL);
3321	}
3322	if (firstaddr == NULL) {
3323		*error = EINVAL;
3324		return (NULL);
3325	}
3326	SCTP_INP_RLOCK(inp);
3327	if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
3328		/*
3329		 * If its in the TCP pool, its NOT allowed to create an
3330		 * association. The parent listener needs to call
3331		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
3332		 * off, or connected one does this.. its an error.
3333		 */
3334		SCTP_INP_RUNLOCK(inp);
3335		*error = EINVAL;
3336		return (NULL);
3337	}
3338	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
3339#ifdef SCTP_DEBUG
3340	if (firstaddr) {
3341		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
3342		SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
3343		    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
3344	} else {
3345		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
3346	}
3347#endif				/* SCTP_DEBUG */
3348	if (firstaddr->sa_family == AF_INET) {
3349		struct sockaddr_in *sin;
3350
3351		sin = (struct sockaddr_in *)firstaddr;
3352		if ((sin->sin_port == 0) || (sin->sin_addr.s_addr == 0)) {
3353			/* Invalid address */
3354			SCTP_INP_RUNLOCK(inp);
3355			*error = EINVAL;
3356			return (NULL);
3357		}
3358		rport = sin->sin_port;
3359	} else if (firstaddr->sa_family == AF_INET6) {
3360		struct sockaddr_in6 *sin6;
3361
3362		sin6 = (struct sockaddr_in6 *)firstaddr;
3363		if ((sin6->sin6_port == 0) ||
3364		    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
3365			/* Invalid address */
3366			SCTP_INP_RUNLOCK(inp);
3367			*error = EINVAL;
3368			return (NULL);
3369		}
3370		rport = sin6->sin6_port;
3371	} else {
3372		/* not supported family type */
3373		SCTP_INP_RUNLOCK(inp);
3374		*error = EINVAL;
3375		return (NULL);
3376	}
3377	SCTP_INP_RUNLOCK(inp);
3378	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
3379		/*
3380		 * If you have not performed a bind, then we need to do the
3381		 * ephemerial bind for you.
3382		 */
3383		if ((err = sctp_inpcb_bind(inp->sctp_socket,
3384		    (struct sockaddr *)NULL, (struct sctp_ifa *)NULL,
3385		    (struct thread *)NULL
3386		    ))) {
3387			/* bind error, probably perm */
3388			*error = err;
3389			return (NULL);
3390		}
3391	}
3392	stcb = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_asoc, struct sctp_tcb);
3393	if (stcb == NULL) {
3394		/* out of memory? */
3395		*error = ENOMEM;
3396		return (NULL);
3397	}
3398	SCTP_INCR_ASOC_COUNT();
3399
3400	bzero(stcb, sizeof(*stcb));
3401	asoc = &stcb->asoc;
3402	SCTP_TCB_LOCK_INIT(stcb);
3403	SCTP_TCB_SEND_LOCK_INIT(stcb);
3404	/* setup back pointer's */
3405	stcb->sctp_ep = inp;
3406	stcb->sctp_socket = inp->sctp_socket;
3407	if ((err = sctp_init_asoc(inp, stcb, for_a_init, override_tag, vrf_id))) {
3408		/* failed */
3409		SCTP_TCB_LOCK_DESTROY(stcb);
3410		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3411		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3412		SCTP_DECR_ASOC_COUNT();
3413		*error = err;
3414		return (NULL);
3415	}
3416	/* and the port */
3417	stcb->rport = rport;
3418	SCTP_INP_INFO_WLOCK();
3419	SCTP_INP_WLOCK(inp);
3420	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3421		/* inpcb freed while alloc going on */
3422		SCTP_TCB_LOCK_DESTROY(stcb);
3423		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3424		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3425		SCTP_INP_WUNLOCK(inp);
3426		SCTP_INP_INFO_WUNLOCK();
3427		SCTP_DECR_ASOC_COUNT();
3428		*error = EINVAL;
3429		return (NULL);
3430	}
3431	SCTP_TCB_LOCK(stcb);
3432
3433	/* now that my_vtag is set, add it to the hash */
3434	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag,
3435	    sctppcbinfo.hashasocmark)];
3436	/* put it in the bucket in the vtag hash of assoc's for the system */
3437	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
3438	SCTP_INP_INFO_WUNLOCK();
3439
3440	if ((err = sctp_add_remote_addr(stcb, firstaddr, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
3441		/* failure.. memory error? */
3442		if (asoc->strmout) {
3443			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
3444			asoc->strmout = NULL;
3445		}
3446		if (asoc->mapping_array) {
3447			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
3448			asoc->mapping_array = NULL;
3449		}
3450		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3451		SCTP_DECR_ASOC_COUNT();
3452		SCTP_TCB_LOCK_DESTROY(stcb);
3453		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3454		SCTP_INP_WUNLOCK(inp);
3455		*error = ENOBUFS;
3456		return (NULL);
3457	}
3458	/* Init all the timers */
3459	SCTP_OS_TIMER_INIT(&asoc->hb_timer.timer);
3460	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
3461	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
3462	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
3463	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
3464	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
3465	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
3466
3467	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
3468	/* now file the port under the hash as well */
3469	if (inp->sctp_tcbhash != NULL) {
3470		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
3471		    inp->sctp_hashmark)];
3472		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
3473	}
3474	SCTP_INP_WUNLOCK(inp);
3475	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", stcb);
3476	return (stcb);
3477}
3478
3479
3480void
3481sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
3482{
3483	struct sctp_association *asoc;
3484
3485	asoc = &stcb->asoc;
3486	asoc->numnets--;
3487	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
3488	if (net == asoc->primary_destination) {
3489		/* Reset primary */
3490		struct sctp_nets *lnet;
3491
3492		lnet = TAILQ_FIRST(&asoc->nets);
3493		/* Try to find a confirmed primary */
3494		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
3495	}
3496	if (net == asoc->last_data_chunk_from) {
3497		/* Reset primary */
3498		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
3499	}
3500	if (net == asoc->last_control_chunk_from) {
3501		/* Clear net */
3502		asoc->last_control_chunk_from = NULL;
3503	}
3504	sctp_free_remote_addr(net);
3505}
3506
3507/*
3508 * remove a remote endpoint address from an association, it will fail if the
3509 * address does not exist.
3510 */
3511int
3512sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
3513{
3514	/*
3515	 * Here we need to remove a remote address. This is quite simple, we
3516	 * first find it in the list of address for the association
3517	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
3518	 * on that item. Note we do not allow it to be removed if there are
3519	 * no other addresses.
3520	 */
3521	struct sctp_association *asoc;
3522	struct sctp_nets *net, *net_tmp;
3523
3524	asoc = &stcb->asoc;
3525
3526	/* locate the address */
3527	for (net = TAILQ_FIRST(&asoc->nets); net != NULL; net = net_tmp) {
3528		net_tmp = TAILQ_NEXT(net, sctp_next);
3529		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
3530			continue;
3531		}
3532		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
3533		    remaddr)) {
3534			/* we found the guy */
3535			if (asoc->numnets < 2) {
3536				/* Must have at LEAST two remote addresses */
3537				return (-1);
3538			} else {
3539				sctp_remove_net(stcb, net);
3540				return (0);
3541			}
3542		}
3543	}
3544	/* not found. */
3545	return (-2);
3546}
3547
3548
3549void
3550sctp_add_vtag_to_timewait(struct sctp_inpcb *inp, uint32_t tag, uint32_t time)
3551{
3552	struct sctpvtaghead *chain;
3553	struct sctp_tagblock *twait_block;
3554	struct timeval now;
3555	int set, i;
3556
3557	(void)SCTP_GETTIME_TIMEVAL(&now);
3558	chain = &sctppcbinfo.vtag_timewait[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
3559	set = 0;
3560	if (!SCTP_LIST_EMPTY(chain)) {
3561		/* Block(s) present, lets find space, and expire on the fly */
3562		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
3563			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
3564				if ((twait_block->vtag_block[i].v_tag == 0) &&
3565				    !set) {
3566					twait_block->vtag_block[i].tv_sec_at_expire =
3567					    now.tv_sec + time;
3568					twait_block->vtag_block[i].v_tag = tag;
3569					set = 1;
3570				} else if ((twait_block->vtag_block[i].v_tag) &&
3571					    ((long)twait_block->vtag_block[i].tv_sec_at_expire >
3572				    now.tv_sec)) {
3573					/* Audit expires this guy */
3574					twait_block->vtag_block[i].tv_sec_at_expire = 0;
3575					twait_block->vtag_block[i].v_tag = 0;
3576					if (set == 0) {
3577						/* Reuse it for my new tag */
3578						twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + SCTP_TIME_WAIT;
3579						twait_block->vtag_block[0].v_tag = tag;
3580						set = 1;
3581					}
3582				}
3583			}
3584			if (set) {
3585				/*
3586				 * We only do up to the block where we can
3587				 * place our tag for audits
3588				 */
3589				break;
3590			}
3591		}
3592	}
3593	/* Need to add a new block to chain */
3594	if (!set) {
3595		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
3596		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
3597		if (twait_block == NULL) {
3598			return;
3599		}
3600		memset(twait_block, 0, sizeof(struct sctp_tagblock));
3601		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
3602		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec +
3603		    SCTP_TIME_WAIT;
3604		twait_block->vtag_block[0].v_tag = tag;
3605	}
3606}
3607
3608
3609static void
3610sctp_iterator_asoc_being_freed(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
3611{
3612	struct sctp_iterator *it;
3613
3614	/*
3615	 * Unlock the tcb lock we do this so we avoid a dead lock scenario
3616	 * where the iterator is waiting on the TCB lock and the TCB lock is
3617	 * waiting on the iterator lock.
3618	 */
3619	it = stcb->asoc.stcb_starting_point_for_iterator;
3620	if (it == NULL) {
3621		return;
3622	}
3623	if (it->inp != stcb->sctp_ep) {
3624		/* hmm, focused on the wrong one? */
3625		return;
3626	}
3627	if (it->stcb != stcb) {
3628		return;
3629	}
3630	it->stcb = LIST_NEXT(stcb, sctp_tcblist);
3631	if (it->stcb == NULL) {
3632		/* done with all asoc's in this assoc */
3633		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3634			it->inp = NULL;
3635		} else {
3636			it->inp = LIST_NEXT(inp, sctp_list);
3637		}
3638	}
3639}
3640
3641
3642/*
3643 * Free the association after un-hashing the remote port.
3644 */
3645int
3646sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
3647{
3648	int i;
3649	struct sctp_association *asoc;
3650	struct sctp_nets *net, *prev;
3651	struct sctp_laddr *laddr;
3652	struct sctp_tmit_chunk *chk;
3653	struct sctp_asconf_addr *aparam;
3654	struct sctp_stream_reset_list *liste;
3655	struct sctp_queued_to_read *sq;
3656	struct sctp_stream_queue_pending *sp;
3657	sctp_sharedkey_t *shared_key;
3658	struct socket *so;
3659	int ccnt = 0;
3660	int cnt = 0;
3661
3662	/* first, lets purge the entry from the hash table. */
3663
3664#ifdef SCTP_LOG_CLOSING
3665	sctp_log_closing(inp, stcb, 6);
3666#endif
3667	if (stcb->asoc.state == 0) {
3668#ifdef SCTP_LOG_CLOSING
3669		sctp_log_closing(inp, NULL, 7);
3670#endif
3671		/* there is no asoc, really TSNH :-0 */
3672		return (1);
3673	}
3674	/* TEMP CODE */
3675	if (stcb->freed_from_where == 0) {
3676		/* Only record the first place free happened from */
3677		stcb->freed_from_where = from_location;
3678	}
3679	/* TEMP CODE */
3680
3681	asoc = &stcb->asoc;
3682	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3683	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3684		/* nothing around */
3685		so = NULL;
3686	else
3687		so = inp->sctp_socket;
3688
3689	/*
3690	 * We used timer based freeing if a reader or writer is in the way.
3691	 * So we first check if we are actually being called from a timer,
3692	 * if so we abort early if a reader or writer is still in the way.
3693	 */
3694	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
3695	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
3696		/*
3697		 * is it the timer driving us? if so are the reader/writers
3698		 * gone?
3699		 */
3700		if (stcb->asoc.refcnt) {
3701			/* nope, reader or writer in the way */
3702			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
3703			/* no asoc destroyed */
3704			SCTP_TCB_UNLOCK(stcb);
3705#ifdef SCTP_LOG_CLOSING
3706			sctp_log_closing(inp, stcb, 8);
3707#endif
3708			return (0);
3709		}
3710	}
3711	/* now clean up any other timers */
3712	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
3713	asoc->hb_timer.self = NULL;
3714	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
3715	asoc->dack_timer.self = NULL;
3716	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3717	/*-
3718	 * For stream reset we don't blast this unless
3719	 * it is a str-reset timer, it might be the
3720	 * free-asoc timer which we DON'T want to
3721	 * disturb.
3722	 */
3723	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
3724		asoc->strreset_timer.self = NULL;
3725	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
3726	asoc->asconf_timer.self = NULL;
3727	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
3728	asoc->autoclose_timer.self = NULL;
3729	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
3730	asoc->shut_guard_timer.self = NULL;
3731	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
3732	asoc->delayed_event_timer.self = NULL;
3733	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
3734		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
3735		net->fr_timer.self = NULL;
3736		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
3737		net->rxt_timer.self = NULL;
3738		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
3739		net->pmtu_timer.self = NULL;
3740	}
3741	/* Now the read queue needs to be cleaned up (only once) */
3742	cnt = 0;
3743	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
3744		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
3745		SCTP_INP_READ_LOCK(inp);
3746		TAILQ_FOREACH(sq, &inp->read_queue, next) {
3747			if (sq->stcb == stcb) {
3748				sq->do_not_ref_stcb = 1;
3749				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
3750				/*
3751				 * If there is no end, there never will be
3752				 * now.
3753				 */
3754				if (sq->end_added == 0) {
3755					/* Held for PD-API clear that. */
3756					sq->pdapi_aborted = 1;
3757					sq->held_length = 0;
3758					if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT)) {
3759						/*
3760						 * Need to add a PD-API
3761						 * aborted indication.
3762						 * Setting the control_pdapi
3763						 * assures that it will be
3764						 * added right after this
3765						 * msg.
3766						 */
3767						uint32_t strseq;
3768
3769						stcb->asoc.control_pdapi = sq;
3770						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
3771						sctp_notify_partial_delivery_indication(stcb,
3772						    SCTP_PARTIAL_DELIVERY_ABORTED, 1, strseq);
3773						stcb->asoc.control_pdapi = NULL;
3774					}
3775				}
3776				/* Add an end to wake them */
3777				sq->end_added = 1;
3778				cnt++;
3779			}
3780		}
3781		SCTP_INP_READ_UNLOCK(inp);
3782		if (stcb->block_entry) {
3783			cnt++;
3784			stcb->block_entry->error = ECONNRESET;
3785			stcb->block_entry = NULL;
3786		}
3787	}
3788	if ((from_inpcbfree != SCTP_PCBFREE_FORCE) && (stcb->asoc.refcnt)) {
3789		/*
3790		 * reader or writer in the way, we have hopefully given him
3791		 * something to chew on above.
3792		 */
3793		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
3794		SCTP_TCB_UNLOCK(stcb);
3795		if (so) {
3796			SCTP_INP_RLOCK(inp);
3797			if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3798			    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3799				/* nothing around */
3800				so = NULL;
3801			if (so) {
3802				/* Wake any reader/writers */
3803				sctp_sorwakeup(inp, so);
3804				sctp_sowwakeup(inp, so);
3805			}
3806			SCTP_INP_RUNLOCK(inp);
3807
3808		}
3809#ifdef SCTP_LOG_CLOSING
3810		sctp_log_closing(inp, stcb, 9);
3811#endif
3812		/* no asoc destroyed */
3813		return (0);
3814	}
3815#ifdef SCTP_LOG_CLOSING
3816	sctp_log_closing(inp, stcb, 10);
3817#endif
3818	/*
3819	 * When I reach here, no others want to kill the assoc yet.. and I
3820	 * own the lock. Now its possible an abort comes in when I do the
3821	 * lock exchange below to grab all the locks to do the final take
3822	 * out. to prevent this we increment the count, which will start a
3823	 * timer and blow out above thus assuring us that we hold exclusive
3824	 * killing of the asoc. Note that after getting back the TCB lock we
3825	 * will go ahead and increment the counter back up and stop any
3826	 * timer a passing stranger may have started :-S
3827	 */
3828	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3829		atomic_add_int(&stcb->asoc.refcnt, 1);
3830
3831		SCTP_TCB_UNLOCK(stcb);
3832
3833		SCTP_ITERATOR_LOCK();
3834		SCTP_INP_INFO_WLOCK();
3835		SCTP_INP_WLOCK(inp);
3836		SCTP_TCB_LOCK(stcb);
3837	}
3838	/* Double check the GONE flag */
3839	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3840	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3841		/* nothing around */
3842		so = NULL;
3843
3844	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
3845	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
3846		/*
3847		 * For TCP type we need special handling when we are
3848		 * connected. We also include the peel'ed off ones to.
3849		 */
3850		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
3851			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
3852			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
3853			if (so) {
3854				SOCK_LOCK(so);
3855				if (so->so_rcv.sb_cc == 0) {
3856					so->so_state &= ~(SS_ISCONNECTING |
3857					    SS_ISDISCONNECTING |
3858					    SS_ISCONFIRMING |
3859					    SS_ISCONNECTED);
3860				}
3861				SOCK_UNLOCK(so);
3862				sctp_sowwakeup(inp, so);
3863				sctp_sorwakeup(inp, so);
3864				SCTP_SOWAKEUP(so);
3865			}
3866		}
3867	}
3868	/*
3869	 * Make it invalid too, that way if its about to run it will abort
3870	 * and return.
3871	 */
3872	sctp_iterator_asoc_being_freed(inp, stcb);
3873	/* re-increment the lock */
3874	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3875		atomic_add_int(&stcb->asoc.refcnt, -1);
3876	}
3877	asoc->state = 0;
3878	if (inp->sctp_tcbhash) {
3879		LIST_REMOVE(stcb, sctp_tcbhash);
3880	}
3881	if (stcb->asoc.in_restart_hash) {
3882		LIST_REMOVE(stcb, sctp_tcbrestarhash);
3883	}
3884	/* Now lets remove it from the list of ALL associations in the EP */
3885	LIST_REMOVE(stcb, sctp_tcblist);
3886	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3887		SCTP_INP_INCR_REF(inp);
3888		SCTP_INP_WUNLOCK(inp);
3889		SCTP_ITERATOR_UNLOCK();
3890	}
3891	/* pull from vtag hash */
3892	LIST_REMOVE(stcb, sctp_asocs);
3893	sctp_add_vtag_to_timewait(inp, asoc->my_vtag, SCTP_TIME_WAIT);
3894
3895
3896	/*
3897	 * Now restop the timers to be sure - this is paranoia at is finest!
3898	 */
3899	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3900	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
3901	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
3902	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3903	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
3904	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
3905	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
3906	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
3907
3908	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
3909		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
3910		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
3911		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
3912	}
3913
3914	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
3915	prev = NULL;
3916	/*
3917	 * The chunk lists and such SHOULD be empty but we check them just
3918	 * in case.
3919	 */
3920	/* anything on the wheel needs to be removed */
3921	for (i = 0; i < asoc->streamoutcnt; i++) {
3922		struct sctp_stream_out *outs;
3923
3924		outs = &asoc->strmout[i];
3925		/* now clean up any chunks here */
3926		sp = TAILQ_FIRST(&outs->outqueue);
3927		while (sp) {
3928			TAILQ_REMOVE(&outs->outqueue, sp, next);
3929			if (sp->data) {
3930				sctp_m_freem(sp->data);
3931				sp->data = NULL;
3932				sp->tail_mbuf = NULL;
3933			}
3934			sctp_free_remote_addr(sp->net);
3935			sctp_free_spbufspace(stcb, asoc, sp);
3936			/* Free the zone stuff  */
3937			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_strmoq, sp);
3938			SCTP_DECR_STRMOQ_COUNT();
3939			/* sa_ignore FREED_MEMORY */
3940			sp = TAILQ_FIRST(&outs->outqueue);
3941		}
3942	}
3943
3944	/* sa_ignore FREED_MEMORY */
3945	while ((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) {
3946		TAILQ_REMOVE(&asoc->resetHead, liste, next_resp);
3947		SCTP_FREE(liste, SCTP_M_STRESET);
3948	}
3949
3950	sq = TAILQ_FIRST(&asoc->pending_reply_queue);
3951	while (sq) {
3952		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
3953		if (sq->data) {
3954			sctp_m_freem(sq->data);
3955			sq->data = NULL;
3956		}
3957		sctp_free_remote_addr(sq->whoFrom);
3958		sq->whoFrom = NULL;
3959		sq->stcb = NULL;
3960		/* Free the ctl entry */
3961		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, sq);
3962		SCTP_DECR_READQ_COUNT();
3963		/* sa_ignore FREED_MEMORY */
3964		sq = TAILQ_FIRST(&asoc->pending_reply_queue);
3965	}
3966
3967	chk = TAILQ_FIRST(&asoc->free_chunks);
3968	while (chk) {
3969		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
3970		if (chk->data) {
3971			sctp_m_freem(chk->data);
3972			chk->data = NULL;
3973		}
3974		ccnt++;
3975		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
3976		SCTP_DECR_CHK_COUNT();
3977		atomic_subtract_int(&sctppcbinfo.ipi_free_chunks, 1);
3978		asoc->free_chunk_cnt--;
3979		/* sa_ignore FREED_MEMORY */
3980		chk = TAILQ_FIRST(&asoc->free_chunks);
3981	}
3982	/* pending send queue SHOULD be empty */
3983	if (!TAILQ_EMPTY(&asoc->send_queue)) {
3984		chk = TAILQ_FIRST(&asoc->send_queue);
3985		while (chk) {
3986			TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
3987			if (chk->data) {
3988				sctp_m_freem(chk->data);
3989				chk->data = NULL;
3990			}
3991			ccnt++;
3992			sctp_free_remote_addr(chk->whoTo);
3993			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
3994			SCTP_DECR_CHK_COUNT();
3995			/* sa_ignore FREED_MEMORY */
3996			chk = TAILQ_FIRST(&asoc->send_queue);
3997		}
3998	}
3999/*
4000  if(ccnt) {
4001  printf("Freed %d from send_queue\n", ccnt);
4002  ccnt = 0;
4003  }
4004*/
4005	/* sent queue SHOULD be empty */
4006	if (!TAILQ_EMPTY(&asoc->sent_queue)) {
4007		chk = TAILQ_FIRST(&asoc->sent_queue);
4008		while (chk) {
4009			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4010			if (chk->data) {
4011				sctp_m_freem(chk->data);
4012				chk->data = NULL;
4013			}
4014			ccnt++;
4015			sctp_free_remote_addr(chk->whoTo);
4016			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4017			SCTP_DECR_CHK_COUNT();
4018			/* sa_ignore FREED_MEMORY */
4019			chk = TAILQ_FIRST(&asoc->sent_queue);
4020		}
4021	}
4022/*
4023  if(ccnt) {
4024  printf("Freed %d from sent_queue\n", ccnt);
4025  ccnt = 0;
4026  }
4027*/
4028	/* control queue MAY not be empty */
4029	if (!TAILQ_EMPTY(&asoc->control_send_queue)) {
4030		chk = TAILQ_FIRST(&asoc->control_send_queue);
4031		while (chk) {
4032			TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4033			if (chk->data) {
4034				sctp_m_freem(chk->data);
4035				chk->data = NULL;
4036			}
4037			ccnt++;
4038			sctp_free_remote_addr(chk->whoTo);
4039			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4040			SCTP_DECR_CHK_COUNT();
4041			/* sa_ignore FREED_MEMORY */
4042			chk = TAILQ_FIRST(&asoc->control_send_queue);
4043		}
4044	}
4045/*
4046  if(ccnt) {
4047  printf("Freed %d from ctrl_queue\n", ccnt);
4048  ccnt = 0;
4049  }
4050*/
4051	if (!TAILQ_EMPTY(&asoc->reasmqueue)) {
4052		chk = TAILQ_FIRST(&asoc->reasmqueue);
4053		while (chk) {
4054			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
4055			if (chk->data) {
4056				sctp_m_freem(chk->data);
4057				chk->data = NULL;
4058			}
4059			sctp_free_remote_addr(chk->whoTo);
4060			ccnt++;
4061			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4062			SCTP_DECR_CHK_COUNT();
4063			/* sa_ignore FREED_MEMORY */
4064			chk = TAILQ_FIRST(&asoc->reasmqueue);
4065		}
4066	}
4067/*
4068  if(ccnt) {
4069  printf("Freed %d from reasm_queue\n", ccnt);
4070  ccnt = 0;
4071  }
4072*/
4073	if (asoc->mapping_array) {
4074		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4075		asoc->mapping_array = NULL;
4076	}
4077	/* the stream outs */
4078	if (asoc->strmout) {
4079		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4080		asoc->strmout = NULL;
4081	}
4082	asoc->streamoutcnt = 0;
4083	if (asoc->strmin) {
4084		struct sctp_queued_to_read *ctl;
4085
4086		for (i = 0; i < asoc->streamincnt; i++) {
4087			if (!TAILQ_EMPTY(&asoc->strmin[i].inqueue)) {
4088				/* We have somethings on the streamin queue */
4089				ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4090				while (ctl) {
4091					TAILQ_REMOVE(&asoc->strmin[i].inqueue,
4092					    ctl, next);
4093					sctp_free_remote_addr(ctl->whoFrom);
4094					if (ctl->data) {
4095						sctp_m_freem(ctl->data);
4096						ctl->data = NULL;
4097					}
4098					/*
4099					 * We don't free the address here
4100					 * since all the net's were freed
4101					 * above.
4102					 */
4103					SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, ctl);
4104					SCTP_DECR_READQ_COUNT();
4105					ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4106				}
4107			}
4108		}
4109		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
4110		asoc->strmin = NULL;
4111	}
4112	asoc->streamincnt = 0;
4113	while (!TAILQ_EMPTY(&asoc->nets)) {
4114		/* sa_ignore FREED_MEMORY */
4115		net = TAILQ_FIRST(&asoc->nets);
4116		/* pull from list */
4117		if ((sctppcbinfo.ipi_count_raddr == 0) || (prev == net)) {
4118#ifdef INVARIANTS
4119			panic("no net's left alloc'ed, or list points to itself");
4120#endif
4121			break;
4122		}
4123		prev = net;
4124		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4125		sctp_free_remote_addr(net);
4126	}
4127
4128	while (!SCTP_LIST_EMPTY(&asoc->sctp_restricted_addrs)) {
4129		/* sa_ignore FREED_MEMORY */
4130		laddr = LIST_FIRST(&asoc->sctp_restricted_addrs);
4131		sctp_remove_laddr(laddr);
4132	}
4133
4134	/* pending asconf (address) parameters */
4135	while (!TAILQ_EMPTY(&asoc->asconf_queue)) {
4136		/* sa_ignore FREED_MEMORY */
4137		aparam = TAILQ_FIRST(&asoc->asconf_queue);
4138		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
4139		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
4140	}
4141	if (asoc->last_asconf_ack_sent != NULL) {
4142		sctp_m_freem(asoc->last_asconf_ack_sent);
4143		asoc->last_asconf_ack_sent = NULL;
4144	}
4145	/* clean up auth stuff */
4146	if (asoc->local_hmacs)
4147		sctp_free_hmaclist(asoc->local_hmacs);
4148	if (asoc->peer_hmacs)
4149		sctp_free_hmaclist(asoc->peer_hmacs);
4150
4151	if (asoc->local_auth_chunks)
4152		sctp_free_chunklist(asoc->local_auth_chunks);
4153	if (asoc->peer_auth_chunks)
4154		sctp_free_chunklist(asoc->peer_auth_chunks);
4155
4156	sctp_free_authinfo(&asoc->authinfo);
4157
4158	shared_key = LIST_FIRST(&asoc->shared_keys);
4159	while (shared_key) {
4160		LIST_REMOVE(shared_key, next);
4161		sctp_free_sharedkey(shared_key);
4162		/* sa_ignore FREED_MEMORY */
4163		shared_key = LIST_FIRST(&asoc->shared_keys);
4164	}
4165
4166	/* Insert new items here :> */
4167
4168	/* Get rid of LOCK */
4169	SCTP_TCB_LOCK_DESTROY(stcb);
4170	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4171	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4172		SCTP_INP_INFO_WUNLOCK();
4173		SCTP_INP_RLOCK(inp);
4174	}
4175#ifdef SCTP_TRACK_FREED_ASOCS
4176	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
4177		/* now clean up the tasoc itself */
4178		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
4179		SCTP_DECR_ASOC_COUNT();
4180	} else {
4181		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
4182	}
4183#else
4184	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
4185	SCTP_DECR_ASOC_COUNT();
4186#endif
4187	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4188		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
4189			/*
4190			 * If its NOT the inp_free calling us AND sctp_close
4191			 * as been called, we call back...
4192			 */
4193			SCTP_INP_RUNLOCK(inp);
4194			/*
4195			 * This will start the kill timer (if we are the
4196			 * lastone) since we hold an increment yet. But this
4197			 * is the only safe way to do this since otherwise
4198			 * if the socket closes at the same time we are here
4199			 * we might collide in the cleanup.
4200			 */
4201			sctp_inpcb_free(inp,
4202			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
4203			    SCTP_CALLED_DIRECTLY_NOCMPSET);
4204			SCTP_INP_DECR_REF(inp);
4205			goto out_of;
4206		} else {
4207			/* The socket is still open. */
4208			SCTP_INP_DECR_REF(inp);
4209		}
4210	}
4211	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4212		SCTP_INP_RUNLOCK(inp);
4213	}
4214out_of:
4215	/* destroyed the asoc */
4216#ifdef SCTP_LOG_CLOSING
4217	sctp_log_closing(inp, NULL, 11);
4218#endif
4219	return (1);
4220}
4221
4222
4223
4224/*
4225 * determine if a destination is "reachable" based upon the addresses bound
4226 * to the current endpoint (e.g. only v4 or v6 currently bound)
4227 */
4228/*
4229 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
4230 * assoc level v4/v6 flags, as the assoc *may* not have the same address
4231 * types bound as its endpoint
4232 */
4233int
4234sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
4235{
4236	struct sctp_inpcb *inp;
4237	int answer;
4238
4239	/*
4240	 * No locks here, the TCB, in all cases is already locked and an
4241	 * assoc is up. There is either a INP lock by the caller applied (in
4242	 * asconf case when deleting an address) or NOT in the HB case,
4243	 * however if HB then the INP increment is up and the INP will not
4244	 * be removed (on top of the fact that we have a TCB lock). So we
4245	 * only want to read the sctp_flags, which is either bound-all or
4246	 * not.. no protection needed since once an assoc is up you can't be
4247	 * changing your binding.
4248	 */
4249	inp = stcb->sctp_ep;
4250	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4251		/* if bound all, destination is not restricted */
4252		/*
4253		 * RRS: Question during lock work: Is this correct? If you
4254		 * are bound-all you still might need to obey the V4--V6
4255		 * flags??? IMO this bound-all stuff needs to be removed!
4256		 */
4257		return (1);
4258	}
4259	/* NOTE: all "scope" checks are done when local addresses are added */
4260	if (destaddr->sa_family == AF_INET6) {
4261		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
4262	} else if (destaddr->sa_family == AF_INET) {
4263		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
4264	} else {
4265		/* invalid family, so it's unreachable */
4266		answer = 0;
4267	}
4268	return (answer);
4269}
4270
4271/*
4272 * update the inp_vflags on an endpoint
4273 */
4274static void
4275sctp_update_ep_vflag(struct sctp_inpcb *inp)
4276{
4277	struct sctp_laddr *laddr;
4278
4279	/* first clear the flag */
4280	inp->ip_inp.inp.inp_vflag = 0;
4281	/* set the flag based on addresses on the ep list */
4282	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4283		if (laddr->ifa == NULL) {
4284			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
4285			    __FUNCTION__);
4286			continue;
4287		}
4288		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
4289			continue;
4290		}
4291		if (laddr->ifa->address.sa.sa_family == AF_INET6) {
4292			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
4293		} else if (laddr->ifa->address.sa.sa_family == AF_INET) {
4294			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
4295		}
4296	}
4297}
4298
4299/*
4300 * Add the address to the endpoint local address list There is nothing to be
4301 * done if we are bound to all addresses
4302 */
4303void
4304sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
4305{
4306	struct sctp_laddr *laddr;
4307	int fnd, error = 0;
4308
4309	fnd = 0;
4310
4311	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4312		/* You are already bound to all. You have it already */
4313		return;
4314	}
4315	if (ifa->address.sa.sa_family == AF_INET6) {
4316		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
4317			/* Can't bind a non-useable addr. */
4318			return;
4319		}
4320	}
4321	/* first, is it already present? */
4322	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4323		if (laddr->ifa == ifa) {
4324			fnd = 1;
4325			break;
4326		}
4327	}
4328
4329	if (fnd == 0) {
4330		/* Not in the ep list */
4331		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
4332		if (error != 0)
4333			return;
4334		inp->laddr_count++;
4335		/* update inp_vflag flags */
4336		if (ifa->address.sa.sa_family == AF_INET6) {
4337			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
4338		} else if (ifa->address.sa.sa_family == AF_INET) {
4339			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
4340		}
4341	}
4342	return;
4343}
4344
4345
4346/*
4347 * select a new (hopefully reachable) destination net (should only be used
4348 * when we deleted an ep addr that is the only usable source address to reach
4349 * the destination net)
4350 */
4351static void
4352sctp_select_primary_destination(struct sctp_tcb *stcb)
4353{
4354	struct sctp_nets *net;
4355
4356	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4357		/* for now, we'll just pick the first reachable one we find */
4358		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
4359			continue;
4360		if (sctp_destination_is_reachable(stcb,
4361		    (struct sockaddr *)&net->ro._l_addr)) {
4362			/* found a reachable destination */
4363			stcb->asoc.primary_destination = net;
4364		}
4365	}
4366	/* I can't there from here! ...we're gonna die shortly... */
4367}
4368
4369
4370/*
4371 * Delete the address from the endpoint local address list There is nothing
4372 * to be done if we are bound to all addresses
4373 */
4374void
4375sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
4376{
4377	struct sctp_laddr *laddr;
4378	int fnd;
4379
4380	fnd = 0;
4381	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4382		/* You are already bound to all. You have it already */
4383		return;
4384	}
4385	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4386		if (laddr->ifa == ifa) {
4387			fnd = 1;
4388			break;
4389		}
4390	}
4391	if (fnd && (inp->laddr_count < 2)) {
4392		/* can't delete unless there are at LEAST 2 addresses */
4393		return;
4394	}
4395	if (fnd) {
4396		/*
4397		 * clean up any use of this address go through our
4398		 * associations and clear any last_used_address that match
4399		 * this one for each assoc, see if a new primary_destination
4400		 * is needed
4401		 */
4402		struct sctp_tcb *stcb;
4403
4404		/* clean up "next_addr_touse" */
4405		if (inp->next_addr_touse == laddr)
4406			/* delete this address */
4407			inp->next_addr_touse = NULL;
4408
4409		/* clean up "last_used_address" */
4410		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
4411			struct sctp_nets *net;
4412
4413			SCTP_TCB_LOCK(stcb);
4414			if (stcb->asoc.last_used_address == laddr)
4415				/* delete this address */
4416				stcb->asoc.last_used_address = NULL;
4417			/*
4418			 * Now spin through all the nets and purge any ref
4419			 * to laddr
4420			 */
4421			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4422				if (net->ro._s_addr &&
4423				    (net->ro._s_addr->ifa == laddr->ifa)) {
4424					/* Yep, purge src address selected */
4425					sctp_rtentry_t *rt;
4426
4427					/* delete this address if cached */
4428					rt = net->ro.ro_rt;
4429					if (rt != NULL) {
4430						RTFREE(rt);
4431						net->ro.ro_rt = NULL;
4432					}
4433					sctp_free_ifa(net->ro._s_addr);
4434					net->ro._s_addr = NULL;
4435					net->src_addr_selected = 0;
4436				}
4437			}
4438			SCTP_TCB_UNLOCK(stcb);
4439		}		/* for each tcb */
4440		/* remove it from the ep list */
4441		sctp_remove_laddr(laddr);
4442		inp->laddr_count--;
4443		/* update inp_vflag flags */
4444		sctp_update_ep_vflag(inp);
4445	}
4446	return;
4447}
4448
4449/*
4450 * Add the addr to the TCB local address list For the BOUNDALL or dynamic
4451 * case, this is a "pending" address list (eg. addresses waiting for an
4452 * ASCONF-ACK response) For the subset binding, static case, this is a
4453 * "valid" address list
4454 */
4455void
4456sctp_add_local_addr_assoc(struct sctp_tcb *stcb, struct sctp_ifa *ifa, int restricted_list)
4457{
4458	struct sctp_inpcb *inp;
4459	struct sctp_laddr *laddr;
4460	struct sctpladdr *list;
4461
4462	/*
4463	 * Assumes TCB is locked.. and possibly the INP. May need to
4464	 * confirm/fix that if we need it and is not the case.
4465	 */
4466	list = &stcb->asoc.sctp_restricted_addrs;
4467
4468	inp = stcb->sctp_ep;
4469	if (ifa->address.sa.sa_family == AF_INET6) {
4470		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
4471			/* Can't bind a non-existent addr. */
4472			return;
4473		}
4474	}
4475	/* does the address already exist? */
4476	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
4477		if (laddr->ifa == ifa) {
4478			return;
4479		}
4480	}
4481
4482	/* add to the list */
4483	(void)sctp_insert_laddr(list, ifa, 0);
4484	return;
4485}
4486
4487/*
4488 * insert an laddr entry with the given ifa for the desired list
4489 */
4490int
4491sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
4492{
4493	struct sctp_laddr *laddr;
4494
4495	laddr = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
4496	if (laddr == NULL) {
4497		/* out of memory? */
4498		return (EINVAL);
4499	}
4500	SCTP_INCR_LADDR_COUNT();
4501	bzero(laddr, sizeof(*laddr));
4502	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
4503	laddr->ifa = ifa;
4504	laddr->action = act;
4505	atomic_add_int(&ifa->refcount, 1);
4506	/* insert it */
4507	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
4508
4509	return (0);
4510}
4511
4512/*
4513 * Remove an laddr entry from the local address list (on an assoc)
4514 */
4515void
4516sctp_remove_laddr(struct sctp_laddr *laddr)
4517{
4518
4519	/* remove from the list */
4520	LIST_REMOVE(laddr, sctp_nxt_addr);
4521	sctp_free_ifa(laddr->ifa);
4522	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_laddr, laddr);
4523	SCTP_DECR_LADDR_COUNT();
4524}
4525
4526/*
4527 * Remove an address from the TCB local address list
4528 */
4529void
4530sctp_del_local_addr_assoc(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
4531{
4532	struct sctp_inpcb *inp;
4533	struct sctp_laddr *laddr;
4534
4535	/*
4536	 * This is called by asconf work. It is assumed that a) The TCB is
4537	 * locked and b) The INP is locked. This is true in as much as I can
4538	 * trace through the entry asconf code where I did these locks.
4539	 * Again, the ASCONF code is a bit different in that it does lock
4540	 * the INP during its work often times. This must be since we don't
4541	 * want other proc's looking up things while what they are looking
4542	 * up is changing :-D
4543	 */
4544
4545	inp = stcb->sctp_ep;
4546	/* if subset bound and don't allow ASCONF's, can't delete last */
4547	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
4548	    (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF) == 0)) {
4549		if (stcb->asoc.numnets < 2) {
4550			/* can't delete last address */
4551			return;
4552		}
4553	}
4554	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
4555		/* remove the address if it exists */
4556		if (laddr->ifa == NULL)
4557			continue;
4558		if (laddr->ifa == ifa) {
4559			sctp_remove_laddr(laddr);
4560			return;
4561		}
4562	}
4563
4564	/* address not found! */
4565	return;
4566}
4567
4568static char sctp_pcb_initialized = 0;
4569
4570/*
4571 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
4572 */
4573/* sysctl */
4574static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
4575static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
4576
4577void
4578sctp_pcb_init()
4579{
4580	/*
4581	 * SCTP initialization for the PCB structures should be called by
4582	 * the sctp_init() funciton.
4583	 */
4584	int i;
4585
4586	if (sctp_pcb_initialized != 0) {
4587		/* error I was called twice */
4588		return;
4589	}
4590	sctp_pcb_initialized = 1;
4591
4592	bzero(&sctpstat, sizeof(struct sctpstat));
4593	(void)SCTP_GETTIME_TIMEVAL(&sctpstat.sctps_discontinuitytime);
4594	/* init the empty list of (All) Endpoints */
4595	LIST_INIT(&sctppcbinfo.listhead);
4596
4597	/* init the iterator head */
4598	TAILQ_INIT(&sctppcbinfo.iteratorhead);
4599
4600	/* init the hash table of endpoints */
4601	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &sctp_hashtblsize);
4602	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &sctp_pcbtblsize);
4603	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &sctp_chunkscale);
4604	sctppcbinfo.sctp_asochash = SCTP_HASH_INIT((sctp_hashtblsize * 31),
4605	    &sctppcbinfo.hashasocmark);
4606	sctppcbinfo.sctp_ephash = SCTP_HASH_INIT(sctp_hashtblsize,
4607	    &sctppcbinfo.hashmark);
4608	sctppcbinfo.sctp_tcpephash = SCTP_HASH_INIT(sctp_hashtblsize,
4609	    &sctppcbinfo.hashtcpmark);
4610	sctppcbinfo.hashtblsize = sctp_hashtblsize;
4611
4612	/* init the small hash table we use to track restarted asoc's */
4613	sctppcbinfo.sctp_restarthash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE,
4614	    &sctppcbinfo.hashrestartmark);
4615
4616
4617	sctppcbinfo.sctp_vrfhash = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
4618	    &sctppcbinfo.hashvrfmark);
4619
4620	sctppcbinfo.vrf_ifn_hash = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
4621	    &sctppcbinfo.vrf_ifn_hashmark);
4622
4623	/* init the zones */
4624	/*
4625	 * FIX ME: Should check for NULL returns, but if it does fail we are
4626	 * doomed to panic anyways... add later maybe.
4627	 */
4628	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_ep, "sctp_ep",
4629	    sizeof(struct sctp_inpcb), maxsockets);
4630
4631	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_asoc, "sctp_asoc",
4632	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
4633
4634	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_laddr, "sctp_laddr",
4635	    sizeof(struct sctp_laddr),
4636	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
4637
4638	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_net, "sctp_raddr",
4639	    sizeof(struct sctp_nets),
4640	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
4641
4642	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_chunk, "sctp_chunk",
4643	    sizeof(struct sctp_tmit_chunk),
4644	    (sctp_max_number_of_assoc * sctp_chunkscale));
4645
4646	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_readq, "sctp_readq",
4647	    sizeof(struct sctp_queued_to_read),
4648	    (sctp_max_number_of_assoc * sctp_chunkscale));
4649
4650	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_strmoq, "sctp_stream_msg_out",
4651	    sizeof(struct sctp_stream_queue_pending),
4652	    (sctp_max_number_of_assoc * sctp_chunkscale));
4653
4654	/* Master Lock INIT for info structure */
4655	SCTP_INP_INFO_LOCK_INIT();
4656	SCTP_STATLOG_INIT_LOCK();
4657	SCTP_ITERATOR_LOCK_INIT();
4658
4659	SCTP_IPI_COUNT_INIT();
4660	SCTP_IPI_ADDR_INIT();
4661	SCTP_IPI_ITERATOR_WQ_INIT();
4662#ifdef SCTP_PACKET_LOGGING
4663	SCTP_IP_PKTLOG_INIT();
4664#endif
4665	LIST_INIT(&sctppcbinfo.addr_wq);
4666
4667	/* not sure if we need all the counts */
4668	sctppcbinfo.ipi_count_ep = 0;
4669	/* assoc/tcb zone info */
4670	sctppcbinfo.ipi_count_asoc = 0;
4671	/* local addrlist zone info */
4672	sctppcbinfo.ipi_count_laddr = 0;
4673	/* remote addrlist zone info */
4674	sctppcbinfo.ipi_count_raddr = 0;
4675	/* chunk info */
4676	sctppcbinfo.ipi_count_chunk = 0;
4677
4678	/* socket queue zone info */
4679	sctppcbinfo.ipi_count_readq = 0;
4680
4681	/* stream out queue cont */
4682	sctppcbinfo.ipi_count_strmoq = 0;
4683
4684	sctppcbinfo.ipi_free_strmoq = 0;
4685	sctppcbinfo.ipi_free_chunks = 0;
4686
4687	SCTP_OS_TIMER_INIT(&sctppcbinfo.addr_wq_timer.timer);
4688
4689	/* Init the TIMEWAIT list */
4690	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE_A; i++) {
4691		LIST_INIT(&sctppcbinfo.vtag_timewait[i]);
4692	}
4693
4694#if defined(SCTP_USE_THREAD_BASED_ITERATOR)
4695	sctppcbinfo.iterator_running = 0;
4696	sctp_startup_iterator();
4697#endif
4698
4699	/*
4700	 * INIT the default VRF which for BSD is the only one, other O/S's
4701	 * may have more. But initially they must start with one and then
4702	 * add the VRF's as addresses are added.
4703	 */
4704	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
4705
4706}
4707
4708
4709int
4710sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
4711    int iphlen, int offset, int limit, struct sctphdr *sh,
4712    struct sockaddr *altsa)
4713{
4714	/*
4715	 * grub through the INIT pulling addresses and loading them to the
4716	 * nets structure in the asoc. The from address in the mbuf should
4717	 * also be loaded (if it is not already). This routine can be called
4718	 * with either INIT or INIT-ACK's as long as the m points to the IP
4719	 * packet and the offset points to the beginning of the parameters.
4720	 */
4721	struct sctp_inpcb *inp, *l_inp;
4722	struct sctp_nets *net, *net_tmp;
4723	struct ip *iph;
4724	struct sctp_paramhdr *phdr, parm_buf;
4725	struct sctp_tcb *stcb_tmp;
4726	uint16_t ptype, plen;
4727	struct sockaddr *sa;
4728	struct sockaddr_storage dest_store;
4729	struct sockaddr *local_sa = (struct sockaddr *)&dest_store;
4730	struct sockaddr_in sin;
4731	struct sockaddr_in6 sin6;
4732	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
4733	struct sctp_auth_random *p_random = NULL;
4734	uint16_t random_len = 0;
4735	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
4736	struct sctp_auth_hmac_algo *hmacs = NULL;
4737	uint16_t hmacs_len = 0;
4738	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
4739	struct sctp_auth_chunk_list *chunks = NULL;
4740	uint16_t num_chunks = 0;
4741	sctp_key_t *new_key;
4742	uint32_t keylen;
4743	int got_random = 0, got_hmacs = 0, got_chklist = 0;
4744
4745	/* First get the destination address setup too. */
4746	memset(&sin, 0, sizeof(sin));
4747	memset(&sin6, 0, sizeof(sin6));
4748
4749	sin.sin_family = AF_INET;
4750	sin.sin_len = sizeof(sin);
4751	sin.sin_port = stcb->rport;
4752
4753	sin6.sin6_family = AF_INET6;
4754	sin6.sin6_len = sizeof(struct sockaddr_in6);
4755	sin6.sin6_port = stcb->rport;
4756	if (altsa == NULL) {
4757		iph = mtod(m, struct ip *);
4758		if (iph->ip_v == IPVERSION) {
4759			/* its IPv4 */
4760			struct sockaddr_in *sin_2;
4761
4762			sin_2 = (struct sockaddr_in *)(local_sa);
4763			memset(sin_2, 0, sizeof(sin));
4764			sin_2->sin_family = AF_INET;
4765			sin_2->sin_len = sizeof(sin);
4766			sin_2->sin_port = sh->dest_port;
4767			sin_2->sin_addr.s_addr = iph->ip_dst.s_addr;
4768			sin.sin_addr = iph->ip_src;
4769			sa = (struct sockaddr *)&sin;
4770		} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
4771			/* its IPv6 */
4772			struct ip6_hdr *ip6;
4773			struct sockaddr_in6 *sin6_2;
4774
4775			ip6 = mtod(m, struct ip6_hdr *);
4776			sin6_2 = (struct sockaddr_in6 *)(local_sa);
4777			memset(sin6_2, 0, sizeof(sin6));
4778			sin6_2->sin6_family = AF_INET6;
4779			sin6_2->sin6_len = sizeof(struct sockaddr_in6);
4780			sin6_2->sin6_port = sh->dest_port;
4781			sin6.sin6_addr = ip6->ip6_src;
4782			sa = (struct sockaddr *)&sin6;
4783		} else {
4784			sa = NULL;
4785		}
4786	} else {
4787		/*
4788		 * For cookies we use the src address NOT from the packet
4789		 * but from the original INIT
4790		 */
4791		sa = altsa;
4792	}
4793	/* Turn off ECN until we get through all params */
4794	stcb->asoc.ecn_allowed = 0;
4795	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4796		/* mark all addresses that we have currently on the list */
4797		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
4798	}
4799	/* does the source address already exist? if so skip it */
4800	l_inp = inp = stcb->sctp_ep;
4801
4802	atomic_add_int(&stcb->asoc.refcnt, 1);
4803	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb);
4804	atomic_add_int(&stcb->asoc.refcnt, -1);
4805
4806	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
4807		/* we must add the source address */
4808		/* no scope set here since we have a tcb already. */
4809		if ((sa->sa_family == AF_INET) &&
4810		    (stcb->asoc.ipv4_addr_legal)) {
4811			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
4812				return (-1);
4813			}
4814		} else if ((sa->sa_family == AF_INET6) &&
4815		    (stcb->asoc.ipv6_addr_legal)) {
4816			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
4817				return (-2);
4818			}
4819		}
4820	} else {
4821		if (net_tmp != NULL && stcb_tmp == stcb) {
4822			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
4823		} else if (stcb_tmp != stcb) {
4824			/* It belongs to another association? */
4825			SCTP_TCB_UNLOCK(stcb_tmp);
4826			return (-3);
4827		}
4828	}
4829	if (stcb->asoc.state == 0) {
4830		/* the assoc was freed? */
4831		return (-4);
4832	}
4833	/* now we must go through each of the params. */
4834	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
4835	while (phdr) {
4836		ptype = ntohs(phdr->param_type);
4837		plen = ntohs(phdr->param_length);
4838		/*
4839		 * printf("ptype => %0x, plen => %d\n", (uint32_t)ptype,
4840		 * (int)plen);
4841		 */
4842		if (offset + plen > limit) {
4843			break;
4844		}
4845		if (plen == 0) {
4846			break;
4847		}
4848		if (ptype == SCTP_IPV4_ADDRESS) {
4849			if (stcb->asoc.ipv4_addr_legal) {
4850				struct sctp_ipv4addr_param *p4, p4_buf;
4851
4852				/* ok get the v4 address and check/add */
4853				phdr = sctp_get_next_param(m, offset,
4854				    (struct sctp_paramhdr *)&p4_buf, sizeof(p4_buf));
4855				if (plen != sizeof(struct sctp_ipv4addr_param) ||
4856				    phdr == NULL) {
4857					return (-5);
4858				}
4859				p4 = (struct sctp_ipv4addr_param *)phdr;
4860				sin.sin_addr.s_addr = p4->addr;
4861				if (IN_MULTICAST(sin.sin_addr.s_addr)) {
4862					/* Skip multi-cast addresses */
4863					goto next_param;
4864				}
4865				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
4866				    (sin.sin_addr.s_addr == INADDR_ANY)) {
4867					goto next_param;
4868				}
4869				sa = (struct sockaddr *)&sin;
4870				inp = stcb->sctp_ep;
4871				atomic_add_int(&stcb->asoc.refcnt, 1);
4872				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
4873				    local_sa, stcb);
4874				atomic_add_int(&stcb->asoc.refcnt, -1);
4875
4876				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
4877				    inp == NULL) {
4878					/* we must add the source address */
4879					/*
4880					 * no scope set since we have a tcb
4881					 * already
4882					 */
4883
4884					/*
4885					 * we must validate the state again
4886					 * here
4887					 */
4888					if (stcb->asoc.state == 0) {
4889						/* the assoc was freed? */
4890						return (-7);
4891					}
4892					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
4893						return (-8);
4894					}
4895				} else if (stcb_tmp == stcb) {
4896					if (stcb->asoc.state == 0) {
4897						/* the assoc was freed? */
4898						return (-10);
4899					}
4900					if (net != NULL) {
4901						/* clear flag */
4902						net->dest_state &=
4903						    ~SCTP_ADDR_NOT_IN_ASSOC;
4904					}
4905				} else {
4906					/*
4907					 * strange, address is in another
4908					 * assoc? straighten out locks.
4909					 */
4910					if (stcb->asoc.state == 0) {
4911						/* the assoc was freed? */
4912						return (-12);
4913					}
4914					return (-13);
4915				}
4916			}
4917		} else if (ptype == SCTP_IPV6_ADDRESS) {
4918			if (stcb->asoc.ipv6_addr_legal) {
4919				/* ok get the v6 address and check/add */
4920				struct sctp_ipv6addr_param *p6, p6_buf;
4921
4922				phdr = sctp_get_next_param(m, offset,
4923				    (struct sctp_paramhdr *)&p6_buf, sizeof(p6_buf));
4924				if (plen != sizeof(struct sctp_ipv6addr_param) ||
4925				    phdr == NULL) {
4926					return (-14);
4927				}
4928				p6 = (struct sctp_ipv6addr_param *)phdr;
4929				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
4930				    sizeof(p6->addr));
4931				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
4932					/* Skip multi-cast addresses */
4933					goto next_param;
4934				}
4935				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
4936					/*
4937					 * Link local make no sense without
4938					 * scope
4939					 */
4940					goto next_param;
4941				}
4942				sa = (struct sockaddr *)&sin6;
4943				inp = stcb->sctp_ep;
4944				atomic_add_int(&stcb->asoc.refcnt, 1);
4945				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
4946				    local_sa, stcb);
4947				atomic_add_int(&stcb->asoc.refcnt, -1);
4948				if (stcb_tmp == NULL && (inp == stcb->sctp_ep ||
4949				    inp == NULL)) {
4950					/*
4951					 * we must validate the state again
4952					 * here
4953					 */
4954					if (stcb->asoc.state == 0) {
4955						/* the assoc was freed? */
4956						return (-16);
4957					}
4958					/*
4959					 * we must add the address, no scope
4960					 * set
4961					 */
4962					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
4963						return (-17);
4964					}
4965				} else if (stcb_tmp == stcb) {
4966					/*
4967					 * we must validate the state again
4968					 * here
4969					 */
4970					if (stcb->asoc.state == 0) {
4971						/* the assoc was freed? */
4972						return (-19);
4973					}
4974					if (net != NULL) {
4975						/* clear flag */
4976						net->dest_state &=
4977						    ~SCTP_ADDR_NOT_IN_ASSOC;
4978					}
4979				} else {
4980					/*
4981					 * strange, address is in another
4982					 * assoc? straighten out locks.
4983					 */
4984					if (stcb->asoc.state == 0) {
4985						/* the assoc was freed? */
4986						return (-21);
4987					}
4988					return (-22);
4989				}
4990			}
4991		} else if (ptype == SCTP_ECN_CAPABLE) {
4992			stcb->asoc.ecn_allowed = 1;
4993		} else if (ptype == SCTP_ULP_ADAPTATION) {
4994			if (stcb->asoc.state != SCTP_STATE_OPEN) {
4995				struct sctp_adaptation_layer_indication ai,
4996				                                *aip;
4997
4998				phdr = sctp_get_next_param(m, offset,
4999				    (struct sctp_paramhdr *)&ai, sizeof(ai));
5000				aip = (struct sctp_adaptation_layer_indication *)phdr;
5001				if (aip) {
5002					sctp_ulp_notify(SCTP_NOTIFY_ADAPTATION_INDICATION,
5003					    stcb, ntohl(aip->indication), NULL);
5004				}
5005			}
5006		} else if (ptype == SCTP_SET_PRIM_ADDR) {
5007			struct sctp_asconf_addr_param lstore, *fee;
5008			struct sctp_asconf_addrv4_param *fii;
5009			int lptype;
5010			struct sockaddr *lsa = NULL;
5011
5012			stcb->asoc.peer_supports_asconf = 1;
5013			if (plen > sizeof(lstore)) {
5014				return (-23);
5015			}
5016			phdr = sctp_get_next_param(m, offset,
5017			    (struct sctp_paramhdr *)&lstore, min(plen, sizeof(lstore)));
5018			if (phdr == NULL) {
5019				return (-24);
5020			}
5021			fee = (struct sctp_asconf_addr_param *)phdr;
5022			lptype = ntohs(fee->addrp.ph.param_type);
5023			if (lptype == SCTP_IPV4_ADDRESS) {
5024				if (plen !=
5025				    sizeof(struct sctp_asconf_addrv4_param)) {
5026					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
5027					    (int)sizeof(struct sctp_asconf_addrv4_param),
5028					    plen);
5029				} else {
5030					fii = (struct sctp_asconf_addrv4_param *)fee;
5031					sin.sin_addr.s_addr = fii->addrp.addr;
5032					lsa = (struct sockaddr *)&sin;
5033				}
5034			} else if (lptype == SCTP_IPV6_ADDRESS) {
5035				if (plen !=
5036				    sizeof(struct sctp_asconf_addr_param)) {
5037					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
5038					    (int)sizeof(struct sctp_asconf_addr_param),
5039					    plen);
5040				} else {
5041					memcpy(sin6.sin6_addr.s6_addr,
5042					    fee->addrp.addr,
5043					    sizeof(fee->addrp.addr));
5044					lsa = (struct sockaddr *)&sin6;
5045				}
5046			}
5047			if (lsa) {
5048				(void)sctp_set_primary_addr(stcb, sa, NULL);
5049			}
5050		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
5051			/* Peer supports pr-sctp */
5052			stcb->asoc.peer_supports_prsctp = 1;
5053		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
5054			/* A supported extension chunk */
5055			struct sctp_supported_chunk_types_param *pr_supported;
5056			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
5057			int num_ent, i;
5058
5059			phdr = sctp_get_next_param(m, offset,
5060			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
5061			if (phdr == NULL) {
5062				return (-25);
5063			}
5064			stcb->asoc.peer_supports_asconf = 0;
5065			stcb->asoc.peer_supports_prsctp = 0;
5066			stcb->asoc.peer_supports_pktdrop = 0;
5067			stcb->asoc.peer_supports_strreset = 0;
5068			stcb->asoc.peer_supports_auth = 0;
5069			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
5070			num_ent = plen - sizeof(struct sctp_paramhdr);
5071			for (i = 0; i < num_ent; i++) {
5072				switch (pr_supported->chunk_types[i]) {
5073				case SCTP_ASCONF:
5074				case SCTP_ASCONF_ACK:
5075					stcb->asoc.peer_supports_asconf = 1;
5076					break;
5077				case SCTP_FORWARD_CUM_TSN:
5078					stcb->asoc.peer_supports_prsctp = 1;
5079					break;
5080				case SCTP_PACKET_DROPPED:
5081					stcb->asoc.peer_supports_pktdrop = 1;
5082					break;
5083				case SCTP_STREAM_RESET:
5084					stcb->asoc.peer_supports_strreset = 1;
5085					break;
5086				case SCTP_AUTHENTICATION:
5087					stcb->asoc.peer_supports_auth = 1;
5088					break;
5089				default:
5090					/* one I have not learned yet */
5091					break;
5092
5093				}
5094			}
5095		} else if (ptype == SCTP_ECN_NONCE_SUPPORTED) {
5096			/* Peer supports ECN-nonce */
5097			stcb->asoc.peer_supports_ecn_nonce = 1;
5098			stcb->asoc.ecn_nonce_allowed = 1;
5099		} else if (ptype == SCTP_RANDOM) {
5100			if (plen > sizeof(random_store))
5101				break;
5102			if (got_random) {
5103				/* already processed a RANDOM */
5104				goto next_param;
5105			}
5106			phdr = sctp_get_next_param(m, offset,
5107			    (struct sctp_paramhdr *)random_store,
5108			    min(sizeof(random_store), plen));
5109			if (phdr == NULL)
5110				return (-26);
5111			p_random = (struct sctp_auth_random *)phdr;
5112			random_len = plen - sizeof(*p_random);
5113			/* enforce the random length */
5114			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
5115				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
5116				return (-27);
5117			}
5118			got_random = 1;
5119		} else if (ptype == SCTP_HMAC_LIST) {
5120			int num_hmacs;
5121			int i;
5122
5123			if (plen > sizeof(hmacs_store))
5124				break;
5125			if (got_hmacs) {
5126				/* already processed a HMAC list */
5127				goto next_param;
5128			}
5129			phdr = sctp_get_next_param(m, offset,
5130			    (struct sctp_paramhdr *)hmacs_store,
5131			    min(plen, sizeof(hmacs_store)));
5132			if (phdr == NULL)
5133				return (-28);
5134			hmacs = (struct sctp_auth_hmac_algo *)phdr;
5135			hmacs_len = plen - sizeof(*hmacs);
5136			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
5137			/* validate the hmac list */
5138			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
5139				return (-29);
5140			}
5141			if (stcb->asoc.peer_hmacs != NULL)
5142				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
5143			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
5144			if (stcb->asoc.peer_hmacs != NULL) {
5145				for (i = 0; i < num_hmacs; i++) {
5146					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
5147					    ntohs(hmacs->hmac_ids[i]));
5148				}
5149			}
5150			got_hmacs = 1;
5151		} else if (ptype == SCTP_CHUNK_LIST) {
5152			int i;
5153
5154			if (plen > sizeof(chunks_store))
5155				break;
5156			if (got_chklist) {
5157				/* already processed a Chunks list */
5158				goto next_param;
5159			}
5160			phdr = sctp_get_next_param(m, offset,
5161			    (struct sctp_paramhdr *)chunks_store,
5162			    min(plen, sizeof(chunks_store)));
5163			if (phdr == NULL)
5164				return (-30);
5165			chunks = (struct sctp_auth_chunk_list *)phdr;
5166			num_chunks = plen - sizeof(*chunks);
5167			if (stcb->asoc.peer_auth_chunks != NULL)
5168				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
5169			else
5170				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
5171			for (i = 0; i < num_chunks; i++) {
5172				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
5173				    stcb->asoc.peer_auth_chunks);
5174			}
5175			got_chklist = 1;
5176		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
5177			    (ptype == SCTP_STATE_COOKIE) ||
5178			    (ptype == SCTP_UNRECOG_PARAM) ||
5179			    (ptype == SCTP_COOKIE_PRESERVE) ||
5180			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
5181			    (ptype == SCTP_ADD_IP_ADDRESS) ||
5182			    (ptype == SCTP_DEL_IP_ADDRESS) ||
5183			    (ptype == SCTP_ERROR_CAUSE_IND) ||
5184		    (ptype == SCTP_SUCCESS_REPORT)) {
5185			 /* don't care */ ;
5186		} else {
5187			if ((ptype & 0x8000) == 0x0000) {
5188				/*
5189				 * must stop processing the rest of the
5190				 * param's. Any report bits were handled
5191				 * with the call to
5192				 * sctp_arethere_unrecognized_parameters()
5193				 * when the INIT or INIT-ACK was first seen.
5194				 */
5195				break;
5196			}
5197		}
5198next_param:
5199		offset += SCTP_SIZE32(plen);
5200		if (offset >= limit) {
5201			break;
5202		}
5203		phdr = sctp_get_next_param(m, offset, &parm_buf,
5204		    sizeof(parm_buf));
5205	}
5206	/* Now check to see if we need to purge any addresses */
5207	for (net = TAILQ_FIRST(&stcb->asoc.nets); net != NULL; net = net_tmp) {
5208		net_tmp = TAILQ_NEXT(net, sctp_next);
5209		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
5210		    SCTP_ADDR_NOT_IN_ASSOC) {
5211			/* This address has been removed from the asoc */
5212			/* remove and free it */
5213			stcb->asoc.numnets--;
5214			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
5215			sctp_free_remote_addr(net);
5216			if (net == stcb->asoc.primary_destination) {
5217				stcb->asoc.primary_destination = NULL;
5218				sctp_select_primary_destination(stcb);
5219			}
5220		}
5221	}
5222	/* validate authentication required parameters */
5223	if (got_random && got_hmacs) {
5224		stcb->asoc.peer_supports_auth = 1;
5225	} else {
5226		stcb->asoc.peer_supports_auth = 0;
5227	}
5228	if (!stcb->asoc.peer_supports_auth && got_chklist) {
5229		/* peer does not support auth but sent a chunks list? */
5230		return (-31);
5231	}
5232	if (!sctp_asconf_auth_nochk && stcb->asoc.peer_supports_asconf &&
5233	    !stcb->asoc.peer_supports_auth) {
5234		/* peer supports asconf but not auth? */
5235		return (-32);
5236	}
5237	/* concatenate the full random key */
5238#ifdef SCTP_AUTH_DRAFT_04
5239	keylen = random_len;
5240	new_key = sctp_alloc_key(keylen);
5241	if (new_key != NULL) {
5242		/* copy in the RANDOM */
5243		if (p_random != NULL)
5244			bcopy(p_random->random_data, new_key->key, random_len);
5245	}
5246#else
5247	keylen = sizeof(*p_random) + random_len + sizeof(*chunks) + num_chunks +
5248	    sizeof(*hmacs) + hmacs_len;
5249	new_key = sctp_alloc_key(keylen);
5250	if (new_key != NULL) {
5251		/* copy in the RANDOM */
5252		if (p_random != NULL) {
5253			keylen = sizeof(*p_random) + random_len;
5254			bcopy(p_random, new_key->key, keylen);
5255		}
5256		/* append in the AUTH chunks */
5257		if (chunks != NULL) {
5258			bcopy(chunks, new_key->key + keylen,
5259			    sizeof(*chunks) + num_chunks);
5260			keylen += sizeof(*chunks) + num_chunks;
5261		}
5262		/* append in the HMACs */
5263		if (hmacs != NULL) {
5264			bcopy(hmacs, new_key->key + keylen,
5265			    sizeof(*hmacs) + hmacs_len);
5266		}
5267	}
5268#endif
5269	else {
5270		/* failed to get memory for the key */
5271		return (-33);
5272	}
5273	if (stcb->asoc.authinfo.peer_random != NULL)
5274		sctp_free_key(stcb->asoc.authinfo.peer_random);
5275	stcb->asoc.authinfo.peer_random = new_key;
5276#ifdef SCTP_AUTH_DRAFT_04
5277	/* don't include the chunks and hmacs for draft -04 */
5278	stcb->asoc.authinfo.peer_random->keylen = random_len;
5279#endif
5280	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
5281	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
5282
5283	return (0);
5284}
5285
5286int
5287sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
5288    struct sctp_nets *net)
5289{
5290	/* make sure the requested primary address exists in the assoc */
5291	if (net == NULL && sa)
5292		net = sctp_findnet(stcb, sa);
5293
5294	if (net == NULL) {
5295		/* didn't find the requested primary address! */
5296		return (-1);
5297	} else {
5298		/* set the primary address */
5299		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
5300			/* Must be confirmed, so queue to set */
5301			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
5302			return (0);
5303		}
5304		stcb->asoc.primary_destination = net;
5305		net->dest_state &= ~SCTP_ADDR_WAS_PRIMARY;
5306		net = TAILQ_FIRST(&stcb->asoc.nets);
5307		if (net != stcb->asoc.primary_destination) {
5308			/*
5309			 * first one on the list is NOT the primary
5310			 * sctp_cmpaddr() is much more efficent if the
5311			 * primary is the first on the list, make it so.
5312			 */
5313			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
5314			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
5315		}
5316		return (0);
5317	}
5318}
5319
5320
5321int
5322sctp_is_vtag_good(struct sctp_inpcb *inp, uint32_t tag, struct timeval *now)
5323{
5324	/*
5325	 * This function serves two purposes. It will see if a TAG can be
5326	 * re-used and return 1 for yes it is ok and 0 for don't use that
5327	 * tag. A secondary function it will do is purge out old tags that
5328	 * can be removed.
5329	 */
5330	struct sctpasochead *head;
5331	struct sctpvtaghead *chain;
5332	struct sctp_tagblock *twait_block;
5333	struct sctp_tcb *stcb;
5334	int i;
5335
5336	SCTP_INP_INFO_WLOCK();
5337	chain = &sctppcbinfo.vtag_timewait[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
5338	/* First is the vtag in use ? */
5339
5340	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(tag,
5341	    sctppcbinfo.hashasocmark)];
5342	if (head == NULL) {
5343		goto check_restart;
5344	}
5345	LIST_FOREACH(stcb, head, sctp_asocs) {
5346
5347		if (stcb->asoc.my_vtag == tag) {
5348			/*
5349			 * We should remove this if and return 0 always if
5350			 * we want vtags unique across all endpoints. For
5351			 * now within a endpoint is ok.
5352			 */
5353			if (inp == stcb->sctp_ep) {
5354				/* bad tag, in use */
5355				SCTP_INP_INFO_WUNLOCK();
5356				return (0);
5357			}
5358		}
5359	}
5360check_restart:
5361	/* Now lets check the restart hash */
5362	head = &sctppcbinfo.sctp_restarthash[SCTP_PCBHASH_ASOC(tag,
5363	    sctppcbinfo.hashrestartmark)];
5364	if (head == NULL) {
5365		goto check_time_wait;
5366	}
5367	LIST_FOREACH(stcb, head, sctp_tcbrestarhash) {
5368		if (stcb->asoc.assoc_id == tag) {
5369			/* candidate */
5370			if (inp == stcb->sctp_ep) {
5371				/* bad tag, in use */
5372				SCTP_INP_INFO_WUNLOCK();
5373				return (0);
5374			}
5375		}
5376	}
5377check_time_wait:
5378	/* Now what about timed wait ? */
5379	if (!SCTP_LIST_EMPTY(chain)) {
5380		/*
5381		 * Block(s) are present, lets see if we have this tag in the
5382		 * list
5383		 */
5384		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5385			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
5386				if (twait_block->vtag_block[i].v_tag == 0) {
5387					/* not used */
5388					continue;
5389				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire >
5390				    now->tv_sec) {
5391					/* Audit expires this guy */
5392					twait_block->vtag_block[i].tv_sec_at_expire = 0;
5393					twait_block->vtag_block[i].v_tag = 0;
5394				} else if (twait_block->vtag_block[i].v_tag ==
5395				    tag) {
5396					/* Bad tag, sorry :< */
5397					SCTP_INP_INFO_WUNLOCK();
5398					return (0);
5399				}
5400			}
5401		}
5402	}
5403	/* Not found, ok to use the tag */
5404	SCTP_INP_INFO_WUNLOCK();
5405	return (1);
5406}
5407
5408
5409static sctp_assoc_t reneged_asoc_ids[256];
5410static uint8_t reneged_at = 0;
5411
5412
5413static void
5414sctp_drain_mbufs(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
5415{
5416	/*
5417	 * We must hunt this association for MBUF's past the cumack (i.e.
5418	 * out of order data that we can renege on).
5419	 */
5420	struct sctp_association *asoc;
5421	struct sctp_tmit_chunk *chk, *nchk;
5422	uint32_t cumulative_tsn_p1, tsn;
5423	struct sctp_queued_to_read *ctl, *nctl;
5424	int cnt, strmat, gap;
5425
5426	/* We look for anything larger than the cum-ack + 1 */
5427
5428	SCTP_STAT_INCR(sctps_protocol_drain_calls);
5429	if (sctp_do_drain == 0) {
5430		return;
5431	}
5432	asoc = &stcb->asoc;
5433	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
5434		/* none we can reneg on. */
5435		return;
5436	}
5437	SCTP_STAT_INCR(sctps_protocol_drains_done);
5438	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
5439	cnt = 0;
5440	/* First look in the re-assembly queue */
5441	chk = TAILQ_FIRST(&asoc->reasmqueue);
5442	while (chk) {
5443		/* Get the next one */
5444		nchk = TAILQ_NEXT(chk, sctp_next);
5445		if (compare_with_wrap(chk->rec.data.TSN_seq,
5446		    cumulative_tsn_p1, MAX_TSN)) {
5447			/* Yep it is above cum-ack */
5448			cnt++;
5449			tsn = chk->rec.data.TSN_seq;
5450			if (tsn >= asoc->mapping_array_base_tsn) {
5451				gap = tsn - asoc->mapping_array_base_tsn;
5452			} else {
5453				gap = (MAX_TSN - asoc->mapping_array_base_tsn) +
5454				    tsn + 1;
5455			}
5456			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
5457			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
5458			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
5459			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5460			if (chk->data) {
5461				sctp_m_freem(chk->data);
5462				chk->data = NULL;
5463			}
5464			sctp_free_remote_addr(chk->whoTo);
5465			sctp_free_a_chunk(stcb, chk);
5466		}
5467		chk = nchk;
5468	}
5469	/* Ok that was fun, now we will drain all the inbound streams? */
5470	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
5471		ctl = TAILQ_FIRST(&asoc->strmin[strmat].inqueue);
5472		while (ctl) {
5473			nctl = TAILQ_NEXT(ctl, next);
5474			if (compare_with_wrap(ctl->sinfo_tsn,
5475			    cumulative_tsn_p1, MAX_TSN)) {
5476				/* Yep it is above cum-ack */
5477				cnt++;
5478				tsn = ctl->sinfo_tsn;
5479				if (tsn >= asoc->mapping_array_base_tsn) {
5480					gap = tsn -
5481					    asoc->mapping_array_base_tsn;
5482				} else {
5483					gap = (MAX_TSN -
5484					    asoc->mapping_array_base_tsn) +
5485					    tsn + 1;
5486				}
5487				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
5488				sctp_ucount_decr(asoc->cnt_on_all_streams);
5489
5490				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array,
5491				    gap);
5492				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue,
5493				    ctl, next);
5494				if (ctl->data) {
5495					sctp_m_freem(ctl->data);
5496					ctl->data = NULL;
5497				}
5498				sctp_free_remote_addr(ctl->whoFrom);
5499				SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, ctl);
5500				SCTP_DECR_READQ_COUNT();
5501			}
5502			ctl = nctl;
5503		}
5504	}
5505	/*
5506	 * Question, should we go through the delivery queue? The only
5507	 * reason things are on here is the app not reading OR a p-d-api up.
5508	 * An attacker COULD send enough in to initiate the PD-API and then
5509	 * send a bunch of stuff to other streams... these would wind up on
5510	 * the delivery queue.. and then we would not get to them. But in
5511	 * order to do this I then have to back-track and un-deliver
5512	 * sequence numbers in streams.. el-yucko. I think for now we will
5513	 * NOT look at the delivery queue and leave it to be something to
5514	 * consider later. An alternative would be to abort the P-D-API with
5515	 * a notification and then deliver the data.... Or another method
5516	 * might be to keep track of how many times the situation occurs and
5517	 * if we see a possible attack underway just abort the association.
5518	 */
5519#ifdef SCTP_DEBUG
5520	if (cnt) {
5521		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
5522	}
5523#endif
5524	if (cnt) {
5525		/*
5526		 * Now do we need to find a new
5527		 * asoc->highest_tsn_inside_map?
5528		 */
5529		if (asoc->highest_tsn_inside_map >= asoc->mapping_array_base_tsn) {
5530			gap = asoc->highest_tsn_inside_map - asoc->mapping_array_base_tsn;
5531		} else {
5532			gap = (MAX_TSN - asoc->mapping_array_base_tsn) +
5533			    asoc->highest_tsn_inside_map + 1;
5534		}
5535		if (gap >= (asoc->mapping_array_size << 3)) {
5536			/*
5537			 * Something bad happened or cum-ack and high were
5538			 * behind the base, but if so earlier checks should
5539			 * have found NO data... wierd... we will start at
5540			 * end of mapping array.
5541			 */
5542			SCTP_PRINTF("Gap was larger than array?? %d set to max:%d maparraymax:%x\n",
5543			    (int)gap,
5544			    (int)(asoc->mapping_array_size << 3),
5545			    (int)asoc->highest_tsn_inside_map);
5546			gap = asoc->mapping_array_size << 3;
5547		}
5548		while (gap > 0) {
5549			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
5550				/* found the new highest */
5551				asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn + gap;
5552				break;
5553			}
5554			gap--;
5555		}
5556		if (gap == 0) {
5557			/* Nothing left in map */
5558			memset(asoc->mapping_array, 0, asoc->mapping_array_size);
5559			asoc->mapping_array_base_tsn = asoc->cumulative_tsn + 1;
5560			asoc->highest_tsn_inside_map = asoc->cumulative_tsn;
5561		}
5562		asoc->last_revoke_count = cnt;
5563		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
5564		sctp_send_sack(stcb);
5565		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN);
5566		reneged_asoc_ids[reneged_at] = sctp_get_associd(stcb);
5567		reneged_at++;
5568	}
5569	/*
5570	 * Another issue, in un-setting the TSN's in the mapping array we
5571	 * DID NOT adjust the higest_tsn marker.  This will cause one of two
5572	 * things to occur. It may cause us to do extra work in checking for
5573	 * our mapping array movement. More importantly it may cause us to
5574	 * SACK every datagram. This may not be a bad thing though since we
5575	 * will recover once we get our cum-ack above and all this stuff we
5576	 * dumped recovered.
5577	 */
5578}
5579
5580void
5581sctp_drain()
5582{
5583	/*
5584	 * We must walk the PCB lists for ALL associations here. The system
5585	 * is LOW on MBUF's and needs help. This is where reneging will
5586	 * occur. We really hope this does NOT happen!
5587	 */
5588	struct sctp_inpcb *inp;
5589	struct sctp_tcb *stcb;
5590
5591	SCTP_INP_INFO_RLOCK();
5592	LIST_FOREACH(inp, &sctppcbinfo.listhead, sctp_list) {
5593		/* For each endpoint */
5594		SCTP_INP_RLOCK(inp);
5595		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5596			/* For each association */
5597			SCTP_TCB_LOCK(stcb);
5598			sctp_drain_mbufs(inp, stcb);
5599			SCTP_TCB_UNLOCK(stcb);
5600		}
5601		SCTP_INP_RUNLOCK(inp);
5602	}
5603	SCTP_INP_INFO_RUNLOCK();
5604}
5605
5606/*
5607 * start a new iterator
5608 * iterates through all endpoints and associations based on the pcb_state
5609 * flags and asoc_state.  "af" (mandatory) is executed for all matching
5610 * assocs and "ef" (optional) is executed when the iterator completes.
5611 * "inpf" (optional) is executed for each new endpoint as it is being
5612 * iterated through. inpe (optional) is called when the inp completes
5613 * its way through all the stcbs.
5614 */
5615int
5616sctp_initiate_iterator(inp_func inpf,
5617    asoc_func af,
5618    inp_func inpe,
5619    uint32_t pcb_state,
5620    uint32_t pcb_features,
5621    uint32_t asoc_state,
5622    void *argp,
5623    uint32_t argi,
5624    end_func ef,
5625    struct sctp_inpcb *s_inp,
5626    uint8_t chunk_output_off)
5627{
5628	struct sctp_iterator *it = NULL;
5629
5630	if (af == NULL) {
5631		return (-1);
5632	}
5633	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
5634	    SCTP_M_ITER);
5635	if (it == NULL) {
5636		return (ENOMEM);
5637	}
5638	memset(it, 0, sizeof(*it));
5639	it->function_assoc = af;
5640	it->function_inp = inpf;
5641	if (inpf)
5642		it->done_current_ep = 0;
5643	else
5644		it->done_current_ep = 1;
5645	it->function_atend = ef;
5646	it->pointer = argp;
5647	it->val = argi;
5648	it->pcb_flags = pcb_state;
5649	it->pcb_features = pcb_features;
5650	it->asoc_state = asoc_state;
5651	it->function_inp_end = inpe;
5652	it->no_chunk_output = chunk_output_off;
5653	if (s_inp) {
5654		it->inp = s_inp;
5655		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
5656	} else {
5657		SCTP_INP_INFO_RLOCK();
5658		it->inp = LIST_FIRST(&sctppcbinfo.listhead);
5659
5660		SCTP_INP_INFO_RUNLOCK();
5661		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
5662
5663	}
5664	SCTP_IPI_ITERATOR_WQ_LOCK();
5665	if (it->inp) {
5666		SCTP_INP_INCR_REF(it->inp);
5667	}
5668	TAILQ_INSERT_TAIL(&sctppcbinfo.iteratorhead, it, sctp_nxt_itr);
5669#if defined(SCTP_USE_THREAD_BASED_ITERATOR)
5670	if (sctppcbinfo.iterator_running == 0) {
5671		sctp_wakeup_iterator();
5672	}
5673	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5674#else
5675	if (it->inp)
5676		SCTP_INP_DECR_REF(it->inp);
5677	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5678	/* Init the timer */
5679	SCTP_OS_TIMER_INIT(&it->tmr.timer);
5680	/* add to the list of all iterators */
5681	sctp_timer_start(SCTP_TIMER_TYPE_ITERATOR, (struct sctp_inpcb *)it,
5682	    NULL, NULL);
5683#endif
5684	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
5685	return (0);
5686}
5687