sctp_pcb.c revision 171158
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 171158 2007-07-02 19:22:22Z 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		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
343		sctp_ifnp->ifn_index = ifn_index;
344		sctp_ifnp->ifn_p = ifn;
345		sctp_ifnp->ifn_type = ifn_type;
346		sctp_ifnp->refcount = 1;
347		sctp_ifnp->vrf = vrf;
348
349		atomic_add_int(&vrf->refcount, 1);
350		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
351		if (if_name != NULL) {
352			memcpy(sctp_ifnp->ifn_name, if_name, SCTP_IFNAMSIZ);
353		} else {
354			memcpy(sctp_ifnp->ifn_name, "unknown", min(7, SCTP_IFNAMSIZ));
355		}
356		hash_ifn_head = &sctppcbinfo.vrf_ifn_hash[(ifn_index & sctppcbinfo.vrf_ifn_hashmark)];
357		LIST_INIT(&sctp_ifnp->ifalist);
358		SCTP_IPI_ADDR_LOCK();
359		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
360		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
361		atomic_add_int(&sctppcbinfo.ipi_count_ifns, 1);
362		new_ifn_af = 1;
363	}
364	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, 1);
365	if (sctp_ifap) {
366		/* Hmm, it already exists? */
367		if ((sctp_ifap->ifn_p) &&
368		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
369			if (new_ifn_af) {
370				/* Remove the created one that we don't want */
371				sctp_delete_ifn(sctp_ifap->ifn_p, 1);
372			}
373			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
374				/* easy to solve, just switch back to active */
375				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
376				sctp_ifap->ifn_p = sctp_ifnp;
377				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
378		exit_stage_left:
379				SCTP_IPI_ADDR_UNLOCK();
380				return (sctp_ifap);
381			} else {
382				goto exit_stage_left;
383			}
384		} else {
385			if (sctp_ifap->ifn_p) {
386				/*
387				 * The first IFN gets the address,
388				 * duplicates are ignored.
389				 */
390				if (new_ifn_af) {
391					/*
392					 * Remove the created one that we
393					 * don't want
394					 */
395					sctp_delete_ifn(sctp_ifap->ifn_p, 1);
396				}
397				goto exit_stage_left;
398			} else {
399				/* repair ifnp which was NULL ? */
400				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
401				sctp_ifap->ifn_p = sctp_ifnp;
402				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
403			}
404			goto exit_stage_left;
405		}
406	}
407	SCTP_IPI_ADDR_UNLOCK();
408	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
409	if (sctp_ifap == NULL) {
410#ifdef INVARIANTS
411		panic("No memory for IFA");
412#endif
413		return (NULL);
414	}
415	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
416	sctp_ifap->ifn_p = sctp_ifnp;
417	atomic_add_int(&sctp_ifnp->refcount, 1);
418	sctp_ifap->vrf_id = vrf_id;
419	sctp_ifap->ifa = ifa;
420	memcpy(&sctp_ifap->address, addr, addr->sa_len);
421	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
422	sctp_ifap->flags = ifa_flags;
423	/* Set scope */
424	if (sctp_ifap->address.sa.sa_family == AF_INET) {
425		struct sockaddr_in *sin;
426
427		sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
428		if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
429		    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
430			sctp_ifap->src_is_loop = 1;
431		}
432		if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
433			sctp_ifap->src_is_priv = 1;
434		}
435		sctp_ifnp->num_v4++;
436		if (new_ifn_af)
437			new_ifn_af = AF_INET;
438	} else if (sctp_ifap->address.sa.sa_family == AF_INET6) {
439		/* ok to use deprecated addresses? */
440		struct sockaddr_in6 *sin6;
441
442		sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
443		if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
444		    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
445			sctp_ifap->src_is_loop = 1;
446		}
447		if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
448			sctp_ifap->src_is_priv = 1;
449		}
450		sctp_ifnp->num_v6++;
451		if (new_ifn_af)
452			new_ifn_af = AF_INET6;
453	} else {
454		new_ifn_af = 0;
455	}
456	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
457
458	if ((sctp_ifap->src_is_priv == 0) &&
459	    (sctp_ifap->src_is_loop == 0)) {
460		sctp_ifap->src_is_glob = 1;
461	}
462	SCTP_IPI_ADDR_LOCK();
463	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
464	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
465	sctp_ifap->refcount = 1;
466	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
467	sctp_ifnp->ifa_count++;
468	vrf->total_ifa_count++;
469	atomic_add_int(&sctppcbinfo.ipi_count_ifas, 1);
470	if (new_ifn_af) {
471		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
472		sctp_ifnp->registered_af = new_ifn_af;
473	}
474	SCTP_IPI_ADDR_UNLOCK();
475	if (dynamic_add) {
476		/*
477		 * Bump up the refcount so that when the timer completes it
478		 * will drop back down.
479		 */
480		struct sctp_laddr *wi;
481
482		atomic_add_int(&sctp_ifap->refcount, 1);
483		wi = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
484		if (wi == NULL) {
485			/*
486			 * Gak, what can we do? We have lost an address
487			 * change can you say HOSED?
488			 */
489			SCTPDBG(SCTP_DEBUG_PCB1, "Lost and address change ???\n");
490			/* Opps, must decrement the count */
491			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index);
492			return (NULL);
493		}
494		SCTP_INCR_LADDR_COUNT();
495		bzero(wi, sizeof(*wi));
496		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
497		wi->ifa = sctp_ifap;
498		wi->action = SCTP_ADD_IP_ADDRESS;
499		SCTP_IPI_ITERATOR_WQ_LOCK();
500		/*
501		 * Should this really be a tailq? As it is we will process
502		 * the newest first :-0
503		 */
504		LIST_INSERT_HEAD(&sctppcbinfo.addr_wq, wi, sctp_nxt_addr);
505		SCTP_IPI_ITERATOR_WQ_UNLOCK();
506		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
507		    (struct sctp_inpcb *)NULL,
508		    (struct sctp_tcb *)NULL,
509		    (struct sctp_nets *)NULL);
510	} else {
511		/* it's ready for use */
512		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
513	}
514	return (sctp_ifap);
515}
516
517void
518sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
519    uint32_t ifn_index)
520{
521	struct sctp_vrf *vrf;
522	struct sctp_ifa *sctp_ifap = NULL;
523
524	SCTP_IPI_ADDR_LOCK();
525
526	vrf = sctp_find_vrf(vrf_id);
527	if (vrf == NULL) {
528		SCTP_PRINTF("Can't find vrf_id:%d\n", vrf_id);
529		goto out_now;
530	}
531	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, 1);
532	if (sctp_ifap) {
533		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
534		sctp_ifap->localifa_flags |= SCTP_BEING_DELETED;
535		vrf->total_ifa_count--;
536		LIST_REMOVE(sctp_ifap, next_bucket);
537		LIST_REMOVE(sctp_ifap, next_ifa);
538		if (sctp_ifap->ifn_p) {
539			sctp_ifap->ifn_p->ifa_count--;
540			if (sctp_ifap->address.sa.sa_family == AF_INET6)
541				sctp_ifap->ifn_p->num_v6--;
542			else if (sctp_ifap->address.sa.sa_family == AF_INET)
543				sctp_ifap->ifn_p->num_v4--;
544			if (SCTP_LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
545				sctp_delete_ifn(sctp_ifap->ifn_p, 1);
546			} else {
547				if ((sctp_ifap->ifn_p->num_v6 == 0) &&
548				    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
549					SCTP_DEREGISTER_INTERFACE(ifn_index,
550					    AF_INET6);
551					SCTP_REGISTER_INTERFACE(ifn_index,
552					    AF_INET);
553					sctp_ifap->ifn_p->registered_af = AF_INET;
554				} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
555				    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
556					SCTP_DEREGISTER_INTERFACE(ifn_index,
557					    AF_INET);
558					SCTP_REGISTER_INTERFACE(ifn_index,
559					    AF_INET6);
560					sctp_ifap->ifn_p->registered_af = AF_INET6;
561				}
562			}
563			sctp_free_ifn(sctp_ifap->ifn_p);
564			sctp_ifap->ifn_p = NULL;
565		}
566	}
567#ifdef SCTP_DEBUG
568	else {
569		SCTPDBG(SCTP_DEBUG_PCB1, "Del Addr-ifn:%d Could not find address:",
570		    ifn_index);
571		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
572	}
573#endif
574
575out_now:
576	SCTP_IPI_ADDR_UNLOCK();
577	if (sctp_ifap) {
578		struct sctp_laddr *wi;
579
580		wi = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
581		if (wi == NULL) {
582			/*
583			 * Gak, what can we do? We have lost an address
584			 * change can you say HOSED?
585			 */
586			SCTPDBG(SCTP_DEBUG_PCB1, "Lost and address change ???\n");
587
588			/* Opps, must decrement the count */
589			sctp_free_ifa(sctp_ifap);
590			return;
591		}
592		SCTP_INCR_LADDR_COUNT();
593		bzero(wi, sizeof(*wi));
594		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
595		wi->ifa = sctp_ifap;
596		wi->action = SCTP_DEL_IP_ADDRESS;
597		SCTP_IPI_ITERATOR_WQ_LOCK();
598		/*
599		 * Should this really be a tailq? As it is we will process
600		 * the newest first :-0
601		 */
602		LIST_INSERT_HEAD(&sctppcbinfo.addr_wq, wi, sctp_nxt_addr);
603		SCTP_IPI_ITERATOR_WQ_UNLOCK();
604
605		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
606		    (struct sctp_inpcb *)NULL,
607		    (struct sctp_tcb *)NULL,
608		    (struct sctp_nets *)NULL);
609	}
610	return;
611}
612
613
614
615static struct sctp_tcb *
616sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
617    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
618{
619	/**** ASSUMSES THE CALLER holds the INP_INFO_RLOCK */
620	/*
621	 * If we support the TCP model, then we must now dig through to see
622	 * if we can find our endpoint in the list of tcp ep's.
623	 */
624	uint16_t lport, rport;
625	struct sctppcbhead *ephead;
626	struct sctp_inpcb *inp;
627	struct sctp_laddr *laddr;
628	struct sctp_tcb *stcb;
629	struct sctp_nets *net;
630
631	if ((to == NULL) || (from == NULL)) {
632		return (NULL);
633	}
634	if (to->sa_family == AF_INET && from->sa_family == AF_INET) {
635		lport = ((struct sockaddr_in *)to)->sin_port;
636		rport = ((struct sockaddr_in *)from)->sin_port;
637	} else if (to->sa_family == AF_INET6 && from->sa_family == AF_INET6) {
638		lport = ((struct sockaddr_in6 *)to)->sin6_port;
639		rport = ((struct sockaddr_in6 *)from)->sin6_port;
640	} else {
641		return NULL;
642	}
643	ephead = &sctppcbinfo.sctp_tcpephash[SCTP_PCBHASH_ALLADDR(
644	    (lport + rport), sctppcbinfo.hashtcpmark)];
645	/*
646	 * Ok now for each of the guys in this bucket we must look and see:
647	 * - Does the remote port match. - Does there single association's
648	 * addresses match this address (to). If so we update p_ep to point
649	 * to this ep and return the tcb from it.
650	 */
651	LIST_FOREACH(inp, ephead, sctp_hash) {
652		SCTP_INP_RLOCK(inp);
653		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
654			SCTP_INP_RUNLOCK(inp);
655			continue;
656		}
657		if (lport != inp->sctp_lport) {
658			SCTP_INP_RUNLOCK(inp);
659			continue;
660		}
661		if (inp->def_vrf_id != vrf_id) {
662			SCTP_INP_RUNLOCK(inp);
663			continue;
664		}
665		/* check to see if the ep has one of the addresses */
666		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
667			/* We are NOT bound all, so look further */
668			int match = 0;
669
670			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
671
672				if (laddr->ifa == NULL) {
673					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
674					continue;
675				}
676				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
677					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
678					continue;
679				}
680				if (laddr->ifa->address.sa.sa_family ==
681				    to->sa_family) {
682					/* see if it matches */
683					struct sockaddr_in *intf_addr, *sin;
684
685					intf_addr = &laddr->ifa->address.sin;
686					sin = (struct sockaddr_in *)to;
687					if (from->sa_family == AF_INET) {
688						if (sin->sin_addr.s_addr ==
689						    intf_addr->sin_addr.s_addr) {
690							match = 1;
691							break;
692						}
693					} else {
694						struct sockaddr_in6 *intf_addr6;
695						struct sockaddr_in6 *sin6;
696
697						sin6 = (struct sockaddr_in6 *)
698						    to;
699						intf_addr6 = &laddr->ifa->address.sin6;
700
701						if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
702						    &intf_addr6->sin6_addr)) {
703							match = 1;
704							break;
705						}
706					}
707				}
708			}
709			if (match == 0) {
710				/* This endpoint does not have this address */
711				SCTP_INP_RUNLOCK(inp);
712				continue;
713			}
714		}
715		/*
716		 * Ok if we hit here the ep has the address, does it hold
717		 * the tcb?
718		 */
719
720		stcb = LIST_FIRST(&inp->sctp_asoc_list);
721		if (stcb == NULL) {
722			SCTP_INP_RUNLOCK(inp);
723			continue;
724		}
725		SCTP_TCB_LOCK(stcb);
726		if (stcb->rport != rport) {
727			/* remote port does not match. */
728			SCTP_TCB_UNLOCK(stcb);
729			SCTP_INP_RUNLOCK(inp);
730			continue;
731		}
732		/* Does this TCB have a matching address? */
733		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
734
735			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
736				/* not the same family, can't be a match */
737				continue;
738			}
739			if (from->sa_family == AF_INET) {
740				struct sockaddr_in *sin, *rsin;
741
742				sin = (struct sockaddr_in *)&net->ro._l_addr;
743				rsin = (struct sockaddr_in *)from;
744				if (sin->sin_addr.s_addr ==
745				    rsin->sin_addr.s_addr) {
746					/* found it */
747					if (netp != NULL) {
748						*netp = net;
749					}
750					/* Update the endpoint pointer */
751					*inp_p = inp;
752					SCTP_INP_RUNLOCK(inp);
753					return (stcb);
754				}
755			} else {
756				struct sockaddr_in6 *sin6, *rsin6;
757
758				sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
759				rsin6 = (struct sockaddr_in6 *)from;
760				if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
761				    &rsin6->sin6_addr)) {
762					/* found it */
763					if (netp != NULL) {
764						*netp = net;
765					}
766					/* Update the endpoint pointer */
767					*inp_p = inp;
768					SCTP_INP_RUNLOCK(inp);
769					return (stcb);
770				}
771			}
772		}
773		SCTP_TCB_UNLOCK(stcb);
774		SCTP_INP_RUNLOCK(inp);
775	}
776	return (NULL);
777}
778
779/*
780 * rules for use
781 *
782 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
783 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
784 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
785 * NULL.
786 */
787
788struct sctp_tcb *
789sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
790    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
791{
792	struct sctpasochead *head;
793	struct sctp_inpcb *inp;
794	struct sctp_tcb *stcb = NULL;
795	struct sctp_nets *net;
796	uint16_t rport;
797
798	inp = *inp_p;
799	if (remote->sa_family == AF_INET) {
800		rport = (((struct sockaddr_in *)remote)->sin_port);
801	} else if (remote->sa_family == AF_INET6) {
802		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
803	} else {
804		return (NULL);
805	}
806	if (locked_tcb) {
807		/*
808		 * UN-lock so we can do proper locking here this occurs when
809		 * called from load_addresses_from_init.
810		 */
811		SCTP_TCB_UNLOCK(locked_tcb);
812	}
813	SCTP_INP_INFO_RLOCK();
814	if (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
815		/*-
816		 * Now either this guy is our listener or it's the
817		 * connector. If it is the one that issued the connect, then
818		 * it's only chance is to be the first TCB in the list. If
819		 * it is the acceptor, then do the special_lookup to hash
820		 * and find the real inp.
821		 */
822		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
823			/* to is peer addr, from is my addr */
824			stcb = sctp_tcb_special_locate(inp_p, remote, local,
825			    netp, inp->def_vrf_id);
826			if ((stcb != NULL) && (locked_tcb == NULL)) {
827				/* we have a locked tcb, lower refcount */
828				SCTP_INP_WLOCK(inp);
829				SCTP_INP_DECR_REF(inp);
830				SCTP_INP_WUNLOCK(inp);
831			}
832			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
833				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
834				SCTP_TCB_LOCK(locked_tcb);
835				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
836			}
837			SCTP_INP_INFO_RUNLOCK();
838			return (stcb);
839		} else {
840			SCTP_INP_WLOCK(inp);
841			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
842				goto null_return;
843			}
844			stcb = LIST_FIRST(&inp->sctp_asoc_list);
845			if (stcb == NULL) {
846				goto null_return;
847			}
848			SCTP_TCB_LOCK(stcb);
849			if (stcb->rport != rport) {
850				/* remote port does not match. */
851				SCTP_TCB_UNLOCK(stcb);
852				goto null_return;
853			}
854			/* now look at the list of remote addresses */
855			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
856#ifdef INVARIANTS
857				if (net == (TAILQ_NEXT(net, sctp_next))) {
858					panic("Corrupt net list");
859				}
860#endif
861				if (net->ro._l_addr.sa.sa_family !=
862				    remote->sa_family) {
863					/* not the same family */
864					continue;
865				}
866				if (remote->sa_family == AF_INET) {
867					struct sockaddr_in *sin, *rsin;
868
869					sin = (struct sockaddr_in *)
870					    &net->ro._l_addr;
871					rsin = (struct sockaddr_in *)remote;
872					if (sin->sin_addr.s_addr ==
873					    rsin->sin_addr.s_addr) {
874						/* found it */
875						if (netp != NULL) {
876							*netp = net;
877						}
878						if (locked_tcb == NULL) {
879							SCTP_INP_DECR_REF(inp);
880						} else if (locked_tcb != stcb) {
881							SCTP_TCB_LOCK(locked_tcb);
882						}
883						SCTP_INP_WUNLOCK(inp);
884						SCTP_INP_INFO_RUNLOCK();
885						return (stcb);
886					}
887				} else if (remote->sa_family == AF_INET6) {
888					struct sockaddr_in6 *sin6, *rsin6;
889
890					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
891					rsin6 = (struct sockaddr_in6 *)remote;
892					if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
893					    &rsin6->sin6_addr)) {
894						/* found it */
895						if (netp != NULL) {
896							*netp = net;
897						}
898						if (locked_tcb == NULL) {
899							SCTP_INP_DECR_REF(inp);
900						} else if (locked_tcb != stcb) {
901							SCTP_TCB_LOCK(locked_tcb);
902						}
903						SCTP_INP_WUNLOCK(inp);
904						SCTP_INP_INFO_RUNLOCK();
905						return (stcb);
906					}
907				}
908			}
909			SCTP_TCB_UNLOCK(stcb);
910		}
911	} else {
912		SCTP_INP_WLOCK(inp);
913		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
914			goto null_return;
915		}
916		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
917		    inp->sctp_hashmark)];
918		if (head == NULL) {
919			goto null_return;
920		}
921		LIST_FOREACH(stcb, head, sctp_tcbhash) {
922			if (stcb->rport != rport) {
923				/* remote port does not match */
924				continue;
925			}
926			/* now look at the list of remote addresses */
927			SCTP_TCB_LOCK(stcb);
928			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
929#ifdef INVARIANTS
930				if (net == (TAILQ_NEXT(net, sctp_next))) {
931					panic("Corrupt net list");
932				}
933#endif
934				if (net->ro._l_addr.sa.sa_family !=
935				    remote->sa_family) {
936					/* not the same family */
937					continue;
938				}
939				if (remote->sa_family == AF_INET) {
940					struct sockaddr_in *sin, *rsin;
941
942					sin = (struct sockaddr_in *)
943					    &net->ro._l_addr;
944					rsin = (struct sockaddr_in *)remote;
945					if (sin->sin_addr.s_addr ==
946					    rsin->sin_addr.s_addr) {
947						/* found it */
948						if (netp != NULL) {
949							*netp = net;
950						}
951						if (locked_tcb == NULL) {
952							SCTP_INP_DECR_REF(inp);
953						} else if (locked_tcb != stcb) {
954							SCTP_TCB_LOCK(locked_tcb);
955						}
956						SCTP_INP_WUNLOCK(inp);
957						SCTP_INP_INFO_RUNLOCK();
958						return (stcb);
959					}
960				} else if (remote->sa_family == AF_INET6) {
961					struct sockaddr_in6 *sin6, *rsin6;
962
963					sin6 = (struct sockaddr_in6 *)
964					    &net->ro._l_addr;
965					rsin6 = (struct sockaddr_in6 *)remote;
966					if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
967					    &rsin6->sin6_addr)) {
968						/* found it */
969						if (netp != NULL) {
970							*netp = net;
971						}
972						if (locked_tcb == NULL) {
973							SCTP_INP_DECR_REF(inp);
974						} else if (locked_tcb != stcb) {
975							SCTP_TCB_LOCK(locked_tcb);
976						}
977						SCTP_INP_WUNLOCK(inp);
978						SCTP_INP_INFO_RUNLOCK();
979						return (stcb);
980					}
981				}
982			}
983			SCTP_TCB_UNLOCK(stcb);
984		}
985	}
986null_return:
987	/* clean up for returning null */
988	if (locked_tcb) {
989		SCTP_TCB_LOCK(locked_tcb);
990	}
991	SCTP_INP_WUNLOCK(inp);
992	SCTP_INP_INFO_RUNLOCK();
993	/* not found */
994	return (NULL);
995}
996
997/*
998 * Find an association for a specific endpoint using the association id given
999 * out in the COMM_UP notification
1000 */
1001
1002struct sctp_tcb *
1003sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1004{
1005	/*
1006	 * Use my the assoc_id to find a endpoint
1007	 */
1008	struct sctpasochead *head;
1009	struct sctp_tcb *stcb;
1010	uint32_t id;
1011
1012	if (asoc_id == 0 || inp == NULL) {
1013		return (NULL);
1014	}
1015	SCTP_INP_INFO_RLOCK();
1016	id = (uint32_t) asoc_id;
1017	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(id,
1018	    sctppcbinfo.hashasocmark)];
1019	if (head == NULL) {
1020		/* invalid id TSNH */
1021		SCTP_INP_INFO_RUNLOCK();
1022		return (NULL);
1023	}
1024	LIST_FOREACH(stcb, head, sctp_asocs) {
1025		SCTP_INP_RLOCK(stcb->sctp_ep);
1026		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1027			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1028			SCTP_INP_INFO_RUNLOCK();
1029			return (NULL);
1030		}
1031		if (stcb->asoc.assoc_id == id) {
1032			/* candidate */
1033			if (inp != stcb->sctp_ep) {
1034				/*
1035				 * some other guy has the same id active (id
1036				 * collision ??).
1037				 */
1038				SCTP_INP_RUNLOCK(stcb->sctp_ep);
1039				continue;
1040			}
1041			if (want_lock) {
1042				SCTP_TCB_LOCK(stcb);
1043			}
1044			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1045			SCTP_INP_INFO_RUNLOCK();
1046			return (stcb);
1047		}
1048		SCTP_INP_RUNLOCK(stcb->sctp_ep);
1049	}
1050	/* Ok if we missed here, lets try the restart hash */
1051	head = &sctppcbinfo.sctp_restarthash[SCTP_PCBHASH_ASOC(id, sctppcbinfo.hashrestartmark)];
1052	if (head == NULL) {
1053		/* invalid id TSNH */
1054		SCTP_INP_INFO_RUNLOCK();
1055		return (NULL);
1056	}
1057	LIST_FOREACH(stcb, head, sctp_tcbrestarhash) {
1058		SCTP_INP_RLOCK(stcb->sctp_ep);
1059		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1060			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1061			continue;
1062		}
1063		if (want_lock) {
1064			SCTP_TCB_LOCK(stcb);
1065		}
1066		if (stcb->asoc.assoc_id == id) {
1067			/* candidate */
1068			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1069			if (inp != stcb->sctp_ep) {
1070				/*
1071				 * some other guy has the same id active (id
1072				 * collision ??).
1073				 */
1074				if (want_lock) {
1075					SCTP_TCB_UNLOCK(stcb);
1076				}
1077				continue;
1078			}
1079			SCTP_INP_INFO_RUNLOCK();
1080			return (stcb);
1081		} else {
1082			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1083		}
1084		if (want_lock) {
1085			SCTP_TCB_UNLOCK(stcb);
1086		}
1087	}
1088	SCTP_INP_INFO_RUNLOCK();
1089	return (NULL);
1090}
1091
1092
1093static struct sctp_inpcb *
1094sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1095    uint16_t lport, uint32_t vrf_id)
1096{
1097	struct sctp_inpcb *inp;
1098	struct sockaddr_in *sin;
1099	struct sockaddr_in6 *sin6;
1100	struct sctp_laddr *laddr;
1101	int fnd;
1102
1103	/*
1104	 * Endpoing probe expects that the INP_INFO is locked.
1105	 */
1106	if (nam->sa_family == AF_INET) {
1107		sin = (struct sockaddr_in *)nam;
1108		sin6 = NULL;
1109	} else if (nam->sa_family == AF_INET6) {
1110		sin6 = (struct sockaddr_in6 *)nam;
1111		sin = NULL;
1112	} else {
1113		/* unsupported family */
1114		return (NULL);
1115	}
1116	if (head == NULL)
1117		return (NULL);
1118	LIST_FOREACH(inp, head, sctp_hash) {
1119		SCTP_INP_RLOCK(inp);
1120		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1121			SCTP_INP_RUNLOCK(inp);
1122			continue;
1123		}
1124		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1125		    (inp->sctp_lport == lport)) {
1126			/* got it */
1127			if ((nam->sa_family == AF_INET) &&
1128			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1129			    SCTP_IPV6_V6ONLY(inp)) {
1130				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1131				SCTP_INP_RUNLOCK(inp);
1132				continue;
1133			}
1134			/* A V6 address and the endpoint is NOT bound V6 */
1135			if (nam->sa_family == AF_INET6 &&
1136			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1137				SCTP_INP_RUNLOCK(inp);
1138				continue;
1139			}
1140			/* does a VRF id match? */
1141			fnd = 0;
1142			if (inp->def_vrf_id == vrf_id)
1143				fnd = 1;
1144
1145			SCTP_INP_RUNLOCK(inp);
1146			if (!fnd)
1147				continue;
1148			return (inp);
1149		}
1150		SCTP_INP_RUNLOCK(inp);
1151	}
1152
1153	if ((nam->sa_family == AF_INET) &&
1154	    (sin->sin_addr.s_addr == INADDR_ANY)) {
1155		/* Can't hunt for one that has no address specified */
1156		return (NULL);
1157	} else if ((nam->sa_family == AF_INET6) &&
1158	    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
1159		/* Can't hunt for one that has no address specified */
1160		return (NULL);
1161	}
1162	/*
1163	 * ok, not bound to all so see if we can find a EP bound to this
1164	 * address.
1165	 */
1166	LIST_FOREACH(inp, head, sctp_hash) {
1167		SCTP_INP_RLOCK(inp);
1168		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1169			SCTP_INP_RUNLOCK(inp);
1170			continue;
1171		}
1172		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1173			SCTP_INP_RUNLOCK(inp);
1174			continue;
1175		}
1176		/*
1177		 * Ok this could be a likely candidate, look at all of its
1178		 * addresses
1179		 */
1180		if (inp->sctp_lport != lport) {
1181			SCTP_INP_RUNLOCK(inp);
1182			continue;
1183		}
1184		/* does a VRF id match? */
1185		fnd = 0;
1186		if (inp->def_vrf_id == vrf_id)
1187			fnd = 1;
1188
1189		if (!fnd) {
1190			SCTP_INP_RUNLOCK(inp);
1191			continue;
1192		}
1193		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1194			if (laddr->ifa == NULL) {
1195				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1196				    __FUNCTION__);
1197				continue;
1198			}
1199			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1200			    laddr->ifa);
1201			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1202				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1203				continue;
1204			}
1205			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1206				/* possible, see if it matches */
1207				struct sockaddr_in *intf_addr;
1208
1209				intf_addr = &laddr->ifa->address.sin;
1210				if (nam->sa_family == AF_INET) {
1211					if (sin->sin_addr.s_addr ==
1212					    intf_addr->sin_addr.s_addr) {
1213						SCTP_INP_RUNLOCK(inp);
1214						return (inp);
1215					}
1216				} else if (nam->sa_family == AF_INET6) {
1217					struct sockaddr_in6 *intf_addr6;
1218
1219					intf_addr6 = &laddr->ifa->address.sin6;
1220					if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
1221					    &intf_addr6->sin6_addr)) {
1222						SCTP_INP_RUNLOCK(inp);
1223						return (inp);
1224					}
1225				}
1226			}
1227		}
1228		SCTP_INP_RUNLOCK(inp);
1229	}
1230	return (NULL);
1231}
1232
1233
1234struct sctp_inpcb *
1235sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1236    uint32_t vrf_id)
1237{
1238	/*
1239	 * First we check the hash table to see if someone has this port
1240	 * bound with just the port.
1241	 */
1242	struct sctp_inpcb *inp;
1243	struct sctppcbhead *head;
1244	struct sockaddr_in *sin;
1245	struct sockaddr_in6 *sin6;
1246	int lport;
1247
1248	if (nam->sa_family == AF_INET) {
1249		sin = (struct sockaddr_in *)nam;
1250		lport = ((struct sockaddr_in *)nam)->sin_port;
1251	} else if (nam->sa_family == AF_INET6) {
1252		sin6 = (struct sockaddr_in6 *)nam;
1253		lport = ((struct sockaddr_in6 *)nam)->sin6_port;
1254	} else {
1255		/* unsupported family */
1256		return (NULL);
1257	}
1258	/*
1259	 * I could cheat here and just cast to one of the types but we will
1260	 * do it right. It also provides the check against an Unsupported
1261	 * type too.
1262	 */
1263	/* Find the head of the ALLADDR chain */
1264	if (have_lock == 0) {
1265		SCTP_INP_INFO_RLOCK();
1266
1267	}
1268	head = &sctppcbinfo.sctp_ephash[SCTP_PCBHASH_ALLADDR(lport,
1269	    sctppcbinfo.hashmark)];
1270	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1271
1272	/*
1273	 * If the TCP model exists it could be that the main listening
1274	 * endpoint is gone but there exists a connected socket for this guy
1275	 * yet. If so we can return the first one that we find. This may NOT
1276	 * be the correct one but the sctp_findassociation_ep_addr has
1277	 * further code to look at all TCP models.
1278	 */
1279	if (inp == NULL && find_tcp_pool) {
1280		unsigned int i;
1281
1282		for (i = 0; i < sctppcbinfo.hashtblsize; i++) {
1283			/*
1284			 * This is real gross, but we do NOT have a remote
1285			 * port at this point depending on who is calling.
1286			 * We must therefore look for ANY one that matches
1287			 * our local port :/
1288			 */
1289			head = &sctppcbinfo.sctp_tcpephash[i];
1290			if (LIST_FIRST(head)) {
1291				inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1292				if (inp) {
1293					/* Found one */
1294					break;
1295				}
1296			}
1297		}
1298	}
1299	if (inp) {
1300		SCTP_INP_INCR_REF(inp);
1301	}
1302	if (have_lock == 0) {
1303		SCTP_INP_INFO_RUNLOCK();
1304	}
1305	return (inp);
1306}
1307
1308/*
1309 * Find an association for an endpoint with the pointer to whom you want to
1310 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1311 * need to change the *to to some other struct like a mbuf...
1312 */
1313struct sctp_tcb *
1314sctp_findassociation_addr_sa(struct sockaddr *to, struct sockaddr *from,
1315    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1316    uint32_t vrf_id)
1317{
1318	struct sctp_inpcb *inp = NULL;
1319	struct sctp_tcb *retval;
1320
1321	SCTP_INP_INFO_RLOCK();
1322	if (find_tcp_pool) {
1323		if (inp_p != NULL) {
1324			retval = sctp_tcb_special_locate(inp_p, from, to, netp,
1325			    vrf_id);
1326		} else {
1327			retval = sctp_tcb_special_locate(&inp, from, to, netp,
1328			    vrf_id);
1329		}
1330		if (retval != NULL) {
1331			SCTP_INP_INFO_RUNLOCK();
1332			return (retval);
1333		}
1334	}
1335	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1336	if (inp_p != NULL) {
1337		*inp_p = inp;
1338	}
1339	SCTP_INP_INFO_RUNLOCK();
1340
1341	if (inp == NULL) {
1342		return (NULL);
1343	}
1344	/*
1345	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1346	 * we now place the source address or from in the to of the find
1347	 * endpoint call. Since in reality this chain is used from the
1348	 * inbound packet side.
1349	 */
1350	if (inp_p != NULL) {
1351		retval = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1352		    NULL);
1353	} else {
1354		retval = sctp_findassociation_ep_addr(&inp, from, netp, to,
1355		    NULL);
1356	}
1357	return retval;
1358}
1359
1360
1361/*
1362 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1363 * find all addresses that the sender has specified in any address list. Each
1364 * address will be used to lookup the TCB and see if one exits.
1365 */
1366static struct sctp_tcb *
1367sctp_findassociation_special_addr(struct mbuf *m, int iphlen, int offset,
1368    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1369    struct sockaddr *dest)
1370{
1371	struct sockaddr_in sin4;
1372	struct sockaddr_in6 sin6;
1373	struct sctp_paramhdr *phdr, parm_buf;
1374	struct sctp_tcb *retval;
1375	uint32_t ptype, plen;
1376
1377	memset(&sin4, 0, sizeof(sin4));
1378	memset(&sin6, 0, sizeof(sin6));
1379	sin4.sin_len = sizeof(sin4);
1380	sin4.sin_family = AF_INET;
1381	sin4.sin_port = sh->src_port;
1382	sin6.sin6_len = sizeof(sin6);
1383	sin6.sin6_family = AF_INET6;
1384	sin6.sin6_port = sh->src_port;
1385
1386	retval = NULL;
1387	offset += sizeof(struct sctp_init_chunk);
1388
1389	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
1390	while (phdr != NULL) {
1391		/* now we must see if we want the parameter */
1392		ptype = ntohs(phdr->param_type);
1393		plen = ntohs(phdr->param_length);
1394		if (plen == 0) {
1395			break;
1396		}
1397		if (ptype == SCTP_IPV4_ADDRESS &&
1398		    plen == sizeof(struct sctp_ipv4addr_param)) {
1399			/* Get the rest of the address */
1400			struct sctp_ipv4addr_param ip4_parm, *p4;
1401
1402			phdr = sctp_get_next_param(m, offset,
1403			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
1404			if (phdr == NULL) {
1405				return (NULL);
1406			}
1407			p4 = (struct sctp_ipv4addr_param *)phdr;
1408			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
1409			/* look it up */
1410			retval = sctp_findassociation_ep_addr(inp_p,
1411			    (struct sockaddr *)&sin4, netp, dest, NULL);
1412			if (retval != NULL) {
1413				return (retval);
1414			}
1415		} else if (ptype == SCTP_IPV6_ADDRESS &&
1416		    plen == sizeof(struct sctp_ipv6addr_param)) {
1417			/* Get the rest of the address */
1418			struct sctp_ipv6addr_param ip6_parm, *p6;
1419
1420			phdr = sctp_get_next_param(m, offset,
1421			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
1422			if (phdr == NULL) {
1423				return (NULL);
1424			}
1425			p6 = (struct sctp_ipv6addr_param *)phdr;
1426			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
1427			/* look it up */
1428			retval = sctp_findassociation_ep_addr(inp_p,
1429			    (struct sockaddr *)&sin6, netp, dest, NULL);
1430			if (retval != NULL) {
1431				return (retval);
1432			}
1433		}
1434		offset += SCTP_SIZE32(plen);
1435		phdr = sctp_get_next_param(m, offset, &parm_buf,
1436		    sizeof(parm_buf));
1437	}
1438	return (NULL);
1439}
1440
1441
1442static struct sctp_tcb *
1443sctp_findassoc_by_vtag(struct sockaddr *from, uint32_t vtag,
1444    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
1445    uint16_t lport, int skip_src_check)
1446{
1447	/*
1448	 * Use my vtag to hash. If we find it we then verify the source addr
1449	 * is in the assoc. If all goes well we save a bit on rec of a
1450	 * packet.
1451	 */
1452	struct sctpasochead *head;
1453	struct sctp_nets *net;
1454	struct sctp_tcb *stcb;
1455
1456	*netp = NULL;
1457	*inp_p = NULL;
1458	SCTP_INP_INFO_RLOCK();
1459	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(vtag,
1460	    sctppcbinfo.hashasocmark)];
1461	if (head == NULL) {
1462		/* invalid vtag */
1463		SCTP_INP_INFO_RUNLOCK();
1464		return (NULL);
1465	}
1466	LIST_FOREACH(stcb, head, sctp_asocs) {
1467		SCTP_INP_RLOCK(stcb->sctp_ep);
1468		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1469			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1470			continue;
1471		}
1472		SCTP_TCB_LOCK(stcb);
1473		SCTP_INP_RUNLOCK(stcb->sctp_ep);
1474		if (stcb->asoc.my_vtag == vtag) {
1475			/* candidate */
1476			if (stcb->rport != rport) {
1477				/*
1478				 * we could remove this if vtags are unique
1479				 * across the system.
1480				 */
1481				SCTP_TCB_UNLOCK(stcb);
1482				continue;
1483			}
1484			if (stcb->sctp_ep->sctp_lport != lport) {
1485				/*
1486				 * we could remove this if vtags are unique
1487				 * across the system.
1488				 */
1489				SCTP_TCB_UNLOCK(stcb);
1490				continue;
1491			}
1492			if (skip_src_check) {
1493				*netp = NULL;	/* unknown */
1494				if (inp_p)
1495					*inp_p = stcb->sctp_ep;
1496				SCTP_INP_INFO_RUNLOCK();
1497				return (stcb);
1498			}
1499			net = sctp_findnet(stcb, from);
1500			if (net) {
1501				/* yep its him. */
1502				*netp = net;
1503				SCTP_STAT_INCR(sctps_vtagexpress);
1504				*inp_p = stcb->sctp_ep;
1505				SCTP_INP_INFO_RUNLOCK();
1506				return (stcb);
1507			} else {
1508				/*
1509				 * not him, this should only happen in rare
1510				 * cases so I peg it.
1511				 */
1512				SCTP_STAT_INCR(sctps_vtagbogus);
1513			}
1514		}
1515		SCTP_TCB_UNLOCK(stcb);
1516	}
1517	SCTP_INP_INFO_RUNLOCK();
1518	return (NULL);
1519}
1520
1521/*
1522 * Find an association with the pointer to the inbound IP packet. This can be
1523 * a IPv4 or IPv6 packet.
1524 */
1525struct sctp_tcb *
1526sctp_findassociation_addr(struct mbuf *m, int iphlen, int offset,
1527    struct sctphdr *sh, struct sctp_chunkhdr *ch,
1528    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
1529{
1530	int find_tcp_pool;
1531	struct ip *iph;
1532	struct sctp_tcb *retval;
1533	struct sockaddr_storage to_store, from_store;
1534	struct sockaddr *to = (struct sockaddr *)&to_store;
1535	struct sockaddr *from = (struct sockaddr *)&from_store;
1536	struct sctp_inpcb *inp;
1537
1538	iph = mtod(m, struct ip *);
1539	if (iph->ip_v == IPVERSION) {
1540		/* its IPv4 */
1541		struct sockaddr_in *from4;
1542
1543		from4 = (struct sockaddr_in *)&from_store;
1544		bzero(from4, sizeof(*from4));
1545		from4->sin_family = AF_INET;
1546		from4->sin_len = sizeof(struct sockaddr_in);
1547		from4->sin_addr.s_addr = iph->ip_src.s_addr;
1548		from4->sin_port = sh->src_port;
1549	} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
1550		/* its IPv6 */
1551		struct ip6_hdr *ip6;
1552		struct sockaddr_in6 *from6;
1553
1554		ip6 = mtod(m, struct ip6_hdr *);
1555		from6 = (struct sockaddr_in6 *)&from_store;
1556		bzero(from6, sizeof(*from6));
1557		from6->sin6_family = AF_INET6;
1558		from6->sin6_len = sizeof(struct sockaddr_in6);
1559		from6->sin6_addr = ip6->ip6_src;
1560		from6->sin6_port = sh->src_port;
1561		/* Get the scopes in properly to the sin6 addr's */
1562		/* we probably don't need these operations */
1563		(void)sa6_recoverscope(from6);
1564		sa6_embedscope(from6, ip6_use_defzone);
1565	} else {
1566		/* Currently not supported. */
1567		return (NULL);
1568	}
1569	if (sh->v_tag) {
1570		/* we only go down this path if vtag is non-zero */
1571		retval = sctp_findassoc_by_vtag(from, ntohl(sh->v_tag),
1572		    inp_p, netp, sh->src_port, sh->dest_port, 0);
1573		if (retval) {
1574			return (retval);
1575		}
1576	}
1577	if (iph->ip_v == IPVERSION) {
1578		/* its IPv4 */
1579		struct sockaddr_in *to4;
1580
1581		to4 = (struct sockaddr_in *)&to_store;
1582		bzero(to4, sizeof(*to4));
1583		to4->sin_family = AF_INET;
1584		to4->sin_len = sizeof(struct sockaddr_in);
1585		to4->sin_addr.s_addr = iph->ip_dst.s_addr;
1586		to4->sin_port = sh->dest_port;
1587	} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
1588		/* its IPv6 */
1589		struct ip6_hdr *ip6;
1590		struct sockaddr_in6 *to6;
1591
1592		ip6 = mtod(m, struct ip6_hdr *);
1593		to6 = (struct sockaddr_in6 *)&to_store;
1594		bzero(to6, sizeof(*to6));
1595		to6->sin6_family = AF_INET6;
1596		to6->sin6_len = sizeof(struct sockaddr_in6);
1597		to6->sin6_addr = ip6->ip6_dst;
1598		to6->sin6_port = sh->dest_port;
1599		/* Get the scopes in properly to the sin6 addr's */
1600		/* we probably don't need these operations */
1601		(void)sa6_recoverscope(to6);
1602		sa6_embedscope(to6, ip6_use_defzone);
1603	}
1604	find_tcp_pool = 0;
1605	if ((ch->chunk_type != SCTP_INITIATION) &&
1606	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
1607	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
1608	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
1609		/* Other chunk types go to the tcp pool. */
1610		find_tcp_pool = 1;
1611	}
1612	if (inp_p) {
1613		retval = sctp_findassociation_addr_sa(to, from, inp_p, netp,
1614		    find_tcp_pool, vrf_id);
1615		inp = *inp_p;
1616	} else {
1617		retval = sctp_findassociation_addr_sa(to, from, &inp, netp,
1618		    find_tcp_pool, vrf_id);
1619	}
1620	SCTPDBG(SCTP_DEBUG_PCB1, "retval:%p inp:%p\n", retval, inp);
1621	if (retval == NULL && inp) {
1622		/* Found a EP but not this address */
1623		if ((ch->chunk_type == SCTP_INITIATION) ||
1624		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
1625			/*-
1626			 * special hook, we do NOT return linp or an
1627			 * association that is linked to an existing
1628			 * association that is under the TCP pool (i.e. no
1629			 * listener exists). The endpoint finding routine
1630			 * will always find a listner before examining the
1631			 * TCP pool.
1632			 */
1633			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
1634				if (inp_p) {
1635					*inp_p = NULL;
1636				}
1637				return (NULL);
1638			}
1639			retval = sctp_findassociation_special_addr(m, iphlen,
1640			    offset, sh, &inp, netp, to);
1641			if (inp_p != NULL) {
1642				*inp_p = inp;
1643			}
1644		}
1645	}
1646	SCTPDBG(SCTP_DEBUG_PCB1, "retval is %p\n", retval);
1647	return (retval);
1648}
1649
1650/*
1651 * lookup an association by an ASCONF lookup address.
1652 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
1653 */
1654struct sctp_tcb *
1655sctp_findassociation_ep_asconf(struct mbuf *m, int iphlen, int offset,
1656    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp)
1657{
1658	struct sctp_tcb *stcb;
1659	struct sockaddr_in *sin;
1660	struct sockaddr_in6 *sin6;
1661	struct sockaddr_storage local_store, remote_store;
1662	struct ip *iph;
1663	struct sctp_paramhdr parm_buf, *phdr;
1664	int ptype;
1665	int zero_address = 0;
1666
1667
1668	memset(&local_store, 0, sizeof(local_store));
1669	memset(&remote_store, 0, sizeof(remote_store));
1670
1671	/* First get the destination address setup too. */
1672	iph = mtod(m, struct ip *);
1673	if (iph->ip_v == IPVERSION) {
1674		/* its IPv4 */
1675		sin = (struct sockaddr_in *)&local_store;
1676		sin->sin_family = AF_INET;
1677		sin->sin_len = sizeof(*sin);
1678		sin->sin_port = sh->dest_port;
1679		sin->sin_addr.s_addr = iph->ip_dst.s_addr;
1680	} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
1681		/* its IPv6 */
1682		struct ip6_hdr *ip6;
1683
1684		ip6 = mtod(m, struct ip6_hdr *);
1685		sin6 = (struct sockaddr_in6 *)&local_store;
1686		sin6->sin6_family = AF_INET6;
1687		sin6->sin6_len = sizeof(*sin6);
1688		sin6->sin6_port = sh->dest_port;
1689		sin6->sin6_addr = ip6->ip6_dst;
1690	} else {
1691		return NULL;
1692	}
1693
1694	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
1695	    &parm_buf, sizeof(struct sctp_paramhdr));
1696	if (phdr == NULL) {
1697		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
1698		    __FUNCTION__);
1699		return NULL;
1700	}
1701	ptype = (int)((uint32_t) ntohs(phdr->param_type));
1702	/* get the correlation address */
1703	if (ptype == SCTP_IPV6_ADDRESS) {
1704		/* ipv6 address param */
1705		struct sctp_ipv6addr_param *p6, p6_buf;
1706
1707		if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
1708			return NULL;
1709		}
1710		p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
1711		    offset + sizeof(struct sctp_asconf_chunk),
1712		    &p6_buf.ph, sizeof(*p6));
1713		if (p6 == NULL) {
1714			SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
1715			    __FUNCTION__);
1716			return (NULL);
1717		}
1718		sin6 = (struct sockaddr_in6 *)&remote_store;
1719		sin6->sin6_family = AF_INET6;
1720		sin6->sin6_len = sizeof(*sin6);
1721		sin6->sin6_port = sh->src_port;
1722		memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
1723		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
1724			zero_address = 1;
1725	} else if (ptype == SCTP_IPV4_ADDRESS) {
1726		/* ipv4 address param */
1727		struct sctp_ipv4addr_param *p4, p4_buf;
1728
1729		if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
1730			return NULL;
1731		}
1732		p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
1733		    offset + sizeof(struct sctp_asconf_chunk),
1734		    &p4_buf.ph, sizeof(*p4));
1735		if (p4 == NULL) {
1736			SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
1737			    __FUNCTION__);
1738			return (NULL);
1739		}
1740		sin = (struct sockaddr_in *)&remote_store;
1741		sin->sin_family = AF_INET;
1742		sin->sin_len = sizeof(*sin);
1743		sin->sin_port = sh->src_port;
1744		memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
1745		if (sin->sin_addr.s_addr == INADDR_ANY)
1746			zero_address = 1;
1747	} else {
1748		/* invalid address param type */
1749		return NULL;
1750	}
1751
1752	if (zero_address) {
1753		stcb = sctp_findassoc_by_vtag(NULL, ntohl(sh->v_tag), inp_p,
1754		    netp, sh->src_port, sh->dest_port, 1);
1755		/*
1756		 * printf("findassociation_ep_asconf: zero lookup address
1757		 * finds stcb 0x%x\n", (uint32_t)stcb);
1758		 */
1759	} else {
1760		stcb = sctp_findassociation_ep_addr(inp_p,
1761		    (struct sockaddr *)&remote_store, netp,
1762		    (struct sockaddr *)&local_store, NULL);
1763	}
1764	return (stcb);
1765}
1766
1767
1768/*
1769 * allocate a sctp_inpcb and setup a temporary binding to a port/all
1770 * addresses. This way if we don't get a bind we by default pick a ephemeral
1771 * port with all addresses bound.
1772 */
1773int
1774sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
1775{
1776	/*
1777	 * we get called when a new endpoint starts up. We need to allocate
1778	 * the sctp_inpcb structure from the zone and init it. Mark it as
1779	 * unbound and find a port that we can use as an ephemeral with
1780	 * INADDR_ANY. If the user binds later no problem we can then add in
1781	 * the specific addresses. And setup the default parameters for the
1782	 * EP.
1783	 */
1784	int i, error;
1785	struct sctp_inpcb *inp;
1786	struct sctp_pcb *m;
1787	struct timeval time;
1788	sctp_sharedkey_t *null_key;
1789
1790	error = 0;
1791
1792	SCTP_INP_INFO_WLOCK();
1793	inp = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_ep, struct sctp_inpcb);
1794	if (inp == NULL) {
1795		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
1796		SCTP_INP_INFO_WUNLOCK();
1797		return (ENOBUFS);
1798	}
1799	/* zap it */
1800	bzero(inp, sizeof(*inp));
1801
1802	/* bump generations */
1803	/* setup socket pointers */
1804	inp->sctp_socket = so;
1805	inp->ip_inp.inp.inp_socket = so;
1806
1807	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
1808	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
1809
1810#ifdef FAST_IPSEC
1811	{
1812		struct inpcbpolicy *pcb_sp = NULL;
1813
1814		error = ipsec_init_policy(so, &pcb_sp);
1815		/* Arrange to share the policy */
1816		inp->ip_inp.inp.inp_sp = pcb_sp;
1817		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
1818	}
1819	if (error != 0) {
1820		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
1821		SCTP_INP_INFO_WUNLOCK();
1822		return error;
1823	}
1824#endif				/* FAST_IPSEC */
1825	SCTP_INCR_EP_COUNT();
1826	inp->ip_inp.inp.inp_ip_ttl = ip_defttl;
1827	SCTP_INP_INFO_WUNLOCK();
1828
1829	so->so_pcb = (caddr_t)inp;
1830
1831	if ((SCTP_SO_TYPE(so) == SOCK_DGRAM) ||
1832	    (SCTP_SO_TYPE(so) == SOCK_SEQPACKET)) {
1833		/* UDP style socket */
1834		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
1835		    SCTP_PCB_FLAGS_UNBOUND);
1836		/* Be sure it is NON-BLOCKING IO for UDP */
1837		/* SCTP_SET_SO_NBIO(so); */
1838	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
1839		/* TCP style socket */
1840		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
1841		    SCTP_PCB_FLAGS_UNBOUND);
1842		/* Be sure we have blocking IO by default */
1843		SCTP_CLEAR_SO_NBIO(so);
1844	} else {
1845		/*
1846		 * unsupported socket type (RAW, etc)- in case we missed it
1847		 * in protosw
1848		 */
1849		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
1850		return (EOPNOTSUPP);
1851	}
1852	sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
1853
1854	inp->sctp_tcbhash = SCTP_HASH_INIT(sctp_pcbtblsize,
1855	    &inp->sctp_hashmark);
1856	if (inp->sctp_tcbhash == NULL) {
1857		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
1858		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
1859		return (ENOBUFS);
1860	}
1861	inp->def_vrf_id = vrf_id;
1862
1863	SCTP_INP_INFO_WLOCK();
1864	SCTP_INP_LOCK_INIT(inp);
1865	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
1866	SCTP_INP_READ_INIT(inp);
1867	SCTP_ASOC_CREATE_LOCK_INIT(inp);
1868	/* lock the new ep */
1869	SCTP_INP_WLOCK(inp);
1870
1871	/* add it to the info area */
1872	LIST_INSERT_HEAD(&sctppcbinfo.listhead, inp, sctp_list);
1873	SCTP_INP_INFO_WUNLOCK();
1874
1875	TAILQ_INIT(&inp->read_queue);
1876	LIST_INIT(&inp->sctp_addr_list);
1877
1878	LIST_INIT(&inp->sctp_asoc_list);
1879
1880#ifdef SCTP_TRACK_FREED_ASOCS
1881	/* TEMP CODE */
1882	LIST_INIT(&inp->sctp_asoc_free_list);
1883#endif
1884	/* Init the timer structure for signature change */
1885	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
1886	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
1887
1888	/* now init the actual endpoint default data */
1889	m = &inp->sctp_ep;
1890
1891	/* setup the base timeout information */
1892	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
1893	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
1894	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(sctp_delayed_sack_time_default);
1895	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(sctp_heartbeat_interval_default);
1896	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(sctp_pmtu_raise_time_default);
1897	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(sctp_shutdown_guard_time_default);
1898	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(sctp_secret_lifetime_default);
1899	/* all max/min max are in ms */
1900	m->sctp_maxrto = sctp_rto_max_default;
1901	m->sctp_minrto = sctp_rto_min_default;
1902	m->initial_rto = sctp_rto_initial_default;
1903	m->initial_init_rto_max = sctp_init_rto_max_default;
1904	m->sctp_sack_freq = sctp_sack_freq_default;
1905
1906	m->max_open_streams_intome = MAX_SCTP_STREAMS;
1907
1908	m->max_init_times = sctp_init_rtx_max_default;
1909	m->max_send_times = sctp_assoc_rtx_max_default;
1910	m->def_net_failure = sctp_path_rtx_max_default;
1911	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
1912	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
1913	m->max_burst = sctp_max_burst_default;
1914	/* number of streams to pre-open on a association */
1915	m->pre_open_stream_count = sctp_nr_outgoing_streams_default;
1916
1917	/* Add adaptation cookie */
1918	m->adaptation_layer_indicator = 0x504C5253;
1919
1920	/* seed random number generator */
1921	m->random_counter = 1;
1922	m->store_at = SCTP_SIGNATURE_SIZE;
1923	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
1924	sctp_fill_random_store(m);
1925
1926	/* Minimum cookie size */
1927	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
1928	    sizeof(struct sctp_state_cookie);
1929	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
1930
1931	/* Setup the initial secret */
1932	(void)SCTP_GETTIME_TIMEVAL(&time);
1933	m->time_of_secret_change = time.tv_sec;
1934
1935	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
1936		m->secret_key[0][i] = sctp_select_initial_TSN(m);
1937	}
1938	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
1939
1940	/* How long is a cookie good for ? */
1941	m->def_cookie_life = MSEC_TO_TICKS(sctp_valid_cookie_life_default);
1942	/*
1943	 * Initialize authentication parameters
1944	 */
1945	m->local_hmacs = sctp_default_supported_hmaclist();
1946	m->local_auth_chunks = sctp_alloc_chunklist();
1947	sctp_auth_set_default_chunks(m->local_auth_chunks);
1948	LIST_INIT(&m->shared_keys);
1949	/* add default NULL key as key id 0 */
1950	null_key = sctp_alloc_sharedkey();
1951	sctp_insert_sharedkey(&m->shared_keys, null_key);
1952	SCTP_INP_WUNLOCK(inp);
1953#ifdef SCTP_LOG_CLOSING
1954	sctp_log_closing(inp, NULL, 12);
1955#endif
1956	return (error);
1957}
1958
1959
1960void
1961sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
1962    struct sctp_tcb *stcb)
1963{
1964	struct sctp_nets *net;
1965	uint16_t lport, rport;
1966	struct sctppcbhead *head;
1967	struct sctp_laddr *laddr, *oladdr;
1968
1969	SCTP_TCB_UNLOCK(stcb);
1970	SCTP_INP_INFO_WLOCK();
1971	SCTP_INP_WLOCK(old_inp);
1972	SCTP_INP_WLOCK(new_inp);
1973	SCTP_TCB_LOCK(stcb);
1974
1975	new_inp->sctp_ep.time_of_secret_change =
1976	    old_inp->sctp_ep.time_of_secret_change;
1977	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
1978	    sizeof(old_inp->sctp_ep.secret_key));
1979	new_inp->sctp_ep.current_secret_number =
1980	    old_inp->sctp_ep.current_secret_number;
1981	new_inp->sctp_ep.last_secret_number =
1982	    old_inp->sctp_ep.last_secret_number;
1983	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
1984
1985	/* make it so new data pours into the new socket */
1986	stcb->sctp_socket = new_inp->sctp_socket;
1987	stcb->sctp_ep = new_inp;
1988
1989	/* Copy the port across */
1990	lport = new_inp->sctp_lport = old_inp->sctp_lport;
1991	rport = stcb->rport;
1992	/* Pull the tcb from the old association */
1993	LIST_REMOVE(stcb, sctp_tcbhash);
1994	LIST_REMOVE(stcb, sctp_tcblist);
1995
1996	/* Now insert the new_inp into the TCP connected hash */
1997	head = &sctppcbinfo.sctp_tcpephash[SCTP_PCBHASH_ALLADDR((lport + rport),
1998	    sctppcbinfo.hashtcpmark)];
1999
2000	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2001	/* Its safe to access */
2002	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2003
2004	/* Now move the tcb into the endpoint list */
2005	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2006	/*
2007	 * Question, do we even need to worry about the ep-hash since we
2008	 * only have one connection? Probably not :> so lets get rid of it
2009	 * and not suck up any kernel memory in that.
2010	 */
2011
2012	/* Ok. Let's restart timer. */
2013	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2014		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2015		    stcb, net);
2016	}
2017
2018	SCTP_INP_INFO_WUNLOCK();
2019	if (new_inp->sctp_tcbhash != NULL) {
2020		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2021		new_inp->sctp_tcbhash = NULL;
2022	}
2023	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2024		/* Subset bound, so copy in the laddr list from the old_inp */
2025		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2026			laddr = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
2027			if (laddr == NULL) {
2028				/*
2029				 * Gak, what can we do? This assoc is really
2030				 * HOSED. We probably should send an abort
2031				 * here.
2032				 */
2033				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2034				continue;
2035			}
2036			SCTP_INCR_LADDR_COUNT();
2037			bzero(laddr, sizeof(*laddr));
2038			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2039			laddr->ifa = oladdr->ifa;
2040			atomic_add_int(&laddr->ifa->refcount, 1);
2041			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2042			    sctp_nxt_addr);
2043			new_inp->laddr_count++;
2044		}
2045	}
2046	/*
2047	 * Now any running timers need to be adjusted since we really don't
2048	 * care if they are running or not just blast in the new_inp into
2049	 * all of them.
2050	 */
2051
2052	stcb->asoc.hb_timer.ep = (void *)new_inp;
2053	stcb->asoc.dack_timer.ep = (void *)new_inp;
2054	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2055	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2056	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2057	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2058	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2059	/* now what about the nets? */
2060	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2061		net->pmtu_timer.ep = (void *)new_inp;
2062		net->rxt_timer.ep = (void *)new_inp;
2063		net->fr_timer.ep = (void *)new_inp;
2064	}
2065	SCTP_INP_WUNLOCK(new_inp);
2066	SCTP_INP_WUNLOCK(old_inp);
2067}
2068
2069static int
2070sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
2071{
2072	struct sctppcbhead *head;
2073	struct sctp_inpcb *t_inp;
2074	int fnd;
2075
2076	head = &sctppcbinfo.sctp_ephash[SCTP_PCBHASH_ALLADDR(lport,
2077	    sctppcbinfo.hashmark)];
2078	LIST_FOREACH(t_inp, head, sctp_hash) {
2079		if (t_inp->sctp_lport != lport) {
2080			continue;
2081		}
2082		/* is it in the VRF in question */
2083		fnd = 0;
2084		if (t_inp->def_vrf_id == vrf_id)
2085			fnd = 1;
2086		if (!fnd)
2087			continue;
2088
2089		/* This one is in use. */
2090		/* check the v6/v4 binding issue */
2091		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
2092		    SCTP_IPV6_V6ONLY(t_inp)) {
2093			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
2094				/* collision in V6 space */
2095				return (1);
2096			} else {
2097				/* inp is BOUND_V4 no conflict */
2098				continue;
2099			}
2100		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
2101			/* t_inp is bound v4 and v6, conflict always */
2102			return (1);
2103		} else {
2104			/* t_inp is bound only V4 */
2105			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
2106			    SCTP_IPV6_V6ONLY(inp)) {
2107				/* no conflict */
2108				continue;
2109			}
2110			/* else fall through to conflict */
2111		}
2112		return (1);
2113	}
2114	return (0);
2115}
2116
2117
2118
2119int
2120sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2121    struct sctp_ifa *sctp_ifap, struct thread *p)
2122{
2123	/* bind a ep to a socket address */
2124	struct sctppcbhead *head;
2125	struct sctp_inpcb *inp, *inp_tmp;
2126	struct inpcb *ip_inp;
2127	int bindall;
2128	uint16_t lport;
2129	int error;
2130	uint32_t vrf_id;
2131
2132	lport = 0;
2133	error = 0;
2134	bindall = 1;
2135	inp = (struct sctp_inpcb *)so->so_pcb;
2136	ip_inp = (struct inpcb *)so->so_pcb;
2137#ifdef SCTP_DEBUG
2138	if (addr) {
2139		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n",
2140		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2141		SCTPDBG(SCTP_DEBUG_PCB1, "Addr :");
2142		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2143	}
2144#endif
2145	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2146		/* already did a bind, subsequent binds NOT allowed ! */
2147		return (EINVAL);
2148	}
2149	if (addr != NULL) {
2150		if (addr->sa_family == AF_INET) {
2151			struct sockaddr_in *sin;
2152
2153			/* IPV6_V6ONLY socket? */
2154			if (SCTP_IPV6_V6ONLY(ip_inp)) {
2155				return (EINVAL);
2156			}
2157			if (addr->sa_len != sizeof(*sin))
2158				return (EINVAL);
2159
2160			sin = (struct sockaddr_in *)addr;
2161			lport = sin->sin_port;
2162
2163			if (sin->sin_addr.s_addr != INADDR_ANY) {
2164				bindall = 0;
2165			}
2166		} else if (addr->sa_family == AF_INET6) {
2167			/* Only for pure IPv6 Address. (No IPv4 Mapped!) */
2168			struct sockaddr_in6 *sin6;
2169
2170			sin6 = (struct sockaddr_in6 *)addr;
2171
2172			if (addr->sa_len != sizeof(*sin6))
2173				return (EINVAL);
2174
2175			lport = sin6->sin6_port;
2176			if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2177				bindall = 0;
2178				/* KAME hack: embed scopeid */
2179				if (sa6_embedscope(sin6, ip6_use_defzone) != 0)
2180					return (EINVAL);
2181			}
2182			/* this must be cleared for ifa_ifwithaddr() */
2183			sin6->sin6_scope_id = 0;
2184		} else {
2185			return (EAFNOSUPPORT);
2186		}
2187	}
2188	/* Setup a vrf_id to be the default for the non-bind-all case. */
2189	vrf_id = inp->def_vrf_id;
2190
2191	SCTP_INP_INFO_WLOCK();
2192	SCTP_INP_WLOCK(inp);
2193	/* increase our count due to the unlock we do */
2194	SCTP_INP_INCR_REF(inp);
2195	if (lport) {
2196		/*
2197		 * Did the caller specify a port? if so we must see if a ep
2198		 * already has this one bound.
2199		 */
2200		/* got to be root to get at low ports */
2201		if (ntohs(lport) < IPPORT_RESERVED) {
2202			if (p && (error =
2203			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2204			    )) {
2205				SCTP_INP_DECR_REF(inp);
2206				SCTP_INP_WUNLOCK(inp);
2207				SCTP_INP_INFO_WUNLOCK();
2208				return (error);
2209			}
2210		}
2211		if (p == NULL) {
2212			SCTP_INP_DECR_REF(inp);
2213			SCTP_INP_WUNLOCK(inp);
2214			SCTP_INP_INFO_WUNLOCK();
2215			return (error);
2216		}
2217		SCTP_INP_WUNLOCK(inp);
2218		if (bindall) {
2219			vrf_id = inp->def_vrf_id;
2220			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2221			if (inp_tmp != NULL) {
2222				/*
2223				 * lock guy returned and lower count note
2224				 * that we are not bound so inp_tmp should
2225				 * NEVER be inp. And it is this inp
2226				 * (inp_tmp) that gets the reference bump,
2227				 * so we must lower it.
2228				 */
2229				SCTP_INP_DECR_REF(inp_tmp);
2230				SCTP_INP_DECR_REF(inp);
2231				/* unlock info */
2232				SCTP_INP_INFO_WUNLOCK();
2233				return (EADDRINUSE);
2234			}
2235		} else {
2236			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2237			if (inp_tmp != NULL) {
2238				/*
2239				 * lock guy returned and lower count note
2240				 * that we are not bound so inp_tmp should
2241				 * NEVER be inp. And it is this inp
2242				 * (inp_tmp) that gets the reference bump,
2243				 * so we must lower it.
2244				 */
2245				SCTP_INP_DECR_REF(inp_tmp);
2246				SCTP_INP_DECR_REF(inp);
2247				/* unlock info */
2248				SCTP_INP_INFO_WUNLOCK();
2249				return (EADDRINUSE);
2250			}
2251		}
2252		SCTP_INP_WLOCK(inp);
2253		if (bindall) {
2254			/* verify that no lport is not used by a singleton */
2255			if (sctp_isport_inuse(inp, lport, vrf_id)) {
2256				/* Sorry someone already has this one bound */
2257				SCTP_INP_DECR_REF(inp);
2258				SCTP_INP_WUNLOCK(inp);
2259				SCTP_INP_INFO_WUNLOCK();
2260				return (EADDRINUSE);
2261			}
2262		}
2263	} else {
2264		uint16_t first, last, candidate;
2265		uint16_t count;
2266		int done;
2267
2268		if (ip_inp->inp_flags & INP_HIGHPORT) {
2269			first = ipport_hifirstauto;
2270			last = ipport_hilastauto;
2271		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2272			if (p && (error =
2273			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2274			    )) {
2275				SCTP_INP_DECR_REF(inp);
2276				SCTP_INP_WUNLOCK(inp);
2277				SCTP_INP_INFO_WUNLOCK();
2278				return (error);
2279			}
2280			first = ipport_lowfirstauto;
2281			last = ipport_lowlastauto;
2282		} else {
2283			first = ipport_firstauto;
2284			last = ipport_lastauto;
2285		}
2286		if (first > last) {
2287			uint16_t temp;
2288
2289			temp = first;
2290			first = last;
2291			last = temp;
2292		}
2293		count = last - first + 1;	/* number of candidates */
2294		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2295
2296		done = 0;
2297		while (!done) {
2298			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == 0) {
2299				done = 1;
2300			}
2301			if (!done) {
2302				if (--count == 0) {
2303					SCTP_INP_DECR_REF(inp);
2304					SCTP_INP_WUNLOCK(inp);
2305					SCTP_INP_INFO_WUNLOCK();
2306					return (EADDRINUSE);
2307				}
2308				if (candidate == last)
2309					candidate = first;
2310				else
2311					candidate = candidate + 1;
2312			}
2313		}
2314		lport = htons(candidate);
2315	}
2316	SCTP_INP_DECR_REF(inp);
2317	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2318	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2319		/*
2320		 * this really should not happen. The guy did a non-blocking
2321		 * bind and then did a close at the same time.
2322		 */
2323		SCTP_INP_WUNLOCK(inp);
2324		SCTP_INP_INFO_WUNLOCK();
2325		return (EINVAL);
2326	}
2327	/* ok we look clear to give out this port, so lets setup the binding */
2328	if (bindall) {
2329		/* binding to all addresses, so just set in the proper flags */
2330		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
2331		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2332		/* set the automatic addr changes from kernel flag */
2333		if (sctp_auto_asconf == 0) {
2334			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2335		} else {
2336			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2337		}
2338	} else {
2339		/*
2340		 * bind specific, make sure flags is off and add a new
2341		 * address structure to the sctp_addr_list inside the ep
2342		 * structure.
2343		 *
2344		 * We will need to allocate one and insert it at the head. The
2345		 * socketopt call can just insert new addresses in there as
2346		 * well. It will also have to do the embed scope kame hack
2347		 * too (before adding).
2348		 */
2349		struct sctp_ifa *ifa;
2350		struct sockaddr_storage store_sa;
2351
2352		memset(&store_sa, 0, sizeof(store_sa));
2353		if (addr->sa_family == AF_INET) {
2354			struct sockaddr_in *sin;
2355
2356			sin = (struct sockaddr_in *)&store_sa;
2357			memcpy(sin, addr, sizeof(struct sockaddr_in));
2358			sin->sin_port = 0;
2359		} else if (addr->sa_family == AF_INET6) {
2360			struct sockaddr_in6 *sin6;
2361
2362			sin6 = (struct sockaddr_in6 *)&store_sa;
2363			memcpy(sin6, addr, sizeof(struct sockaddr_in6));
2364			sin6->sin6_port = 0;
2365		}
2366		/*
2367		 * first find the interface with the bound address need to
2368		 * zero out the port to find the address! yuck! can't do
2369		 * this earlier since need port for sctp_pcb_findep()
2370		 */
2371		if (sctp_ifap)
2372			ifa = sctp_ifap;
2373		else
2374			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
2375			    vrf_id, 0);
2376		if (ifa == NULL) {
2377			/* Can't find an interface with that address */
2378			SCTP_INP_WUNLOCK(inp);
2379			SCTP_INP_INFO_WUNLOCK();
2380			return (EADDRNOTAVAIL);
2381		}
2382		if (addr->sa_family == AF_INET6) {
2383			/* GAK, more FIXME IFA lock? */
2384			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
2385				/* Can't bind a non-existent addr. */
2386				SCTP_INP_WUNLOCK(inp);
2387				SCTP_INP_INFO_WUNLOCK();
2388				return (EINVAL);
2389			}
2390		}
2391		/* we're not bound all */
2392		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
2393		/* allow bindx() to send ASCONF's for binding changes */
2394		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2395		/* clear automatic addr changes from kernel flag */
2396		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2397
2398		/* add this address to the endpoint list */
2399		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
2400		if (error != 0) {
2401			SCTP_INP_WUNLOCK(inp);
2402			SCTP_INP_INFO_WUNLOCK();
2403			return (error);
2404		}
2405		inp->laddr_count++;
2406	}
2407	/* find the bucket */
2408	head = &sctppcbinfo.sctp_ephash[SCTP_PCBHASH_ALLADDR(lport,
2409	    sctppcbinfo.hashmark)];
2410	/* put it in the bucket */
2411	LIST_INSERT_HEAD(head, inp, sctp_hash);
2412	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d\n",
2413	    head, ntohs(lport));
2414	/* set in the port */
2415	inp->sctp_lport = lport;
2416
2417	/* turn off just the unbound flag */
2418	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2419	SCTP_INP_WUNLOCK(inp);
2420	SCTP_INP_INFO_WUNLOCK();
2421	return (0);
2422}
2423
2424
2425static void
2426sctp_iterator_inp_being_freed(struct sctp_inpcb *inp, struct sctp_inpcb *inp_next)
2427{
2428	struct sctp_iterator *it;
2429
2430	/*
2431	 * We enter with the only the ITERATOR_LOCK in place and a write
2432	 * lock on the inp_info stuff.
2433	 */
2434
2435	/*
2436	 * Go through all iterators, we must do this since it is possible
2437	 * that some iterator does NOT have the lock, but is waiting for it.
2438	 * And the one that had the lock has either moved in the last
2439	 * iteration or we just cleared it above. We need to find all of
2440	 * those guys. The list of iterators should never be very big
2441	 * though.
2442	 */
2443	TAILQ_FOREACH(it, &sctppcbinfo.iteratorhead, sctp_nxt_itr) {
2444		if (it == inp->inp_starting_point_for_iterator)
2445			/* skip this guy, he's special */
2446			continue;
2447		if (it->inp == inp) {
2448			/*
2449			 * This is tricky and we DON'T lock the iterator.
2450			 * Reason is he's running but waiting for me since
2451			 * inp->inp_starting_point_for_iterator has the lock
2452			 * on me (the guy above we skipped). This tells us
2453			 * its is not running but waiting for
2454			 * inp->inp_starting_point_for_iterator to be
2455			 * released by the guy that does have our INP in a
2456			 * lock.
2457			 */
2458			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
2459				it->inp = NULL;
2460				it->stcb = NULL;
2461			} else {
2462				/* set him up to do the next guy not me */
2463				it->inp = inp_next;
2464				it->stcb = NULL;
2465			}
2466		}
2467	}
2468	it = inp->inp_starting_point_for_iterator;
2469	if (it) {
2470		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
2471			it->inp = NULL;
2472		} else {
2473			it->inp = inp_next;
2474		}
2475		it->stcb = NULL;
2476	}
2477}
2478
2479/* release sctp_inpcb unbind the port */
2480void
2481sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
2482{
2483	/*
2484	 * Here we free a endpoint. We must find it (if it is in the Hash
2485	 * table) and remove it from there. Then we must also find it in the
2486	 * overall list and remove it from there. After all removals are
2487	 * complete then any timer has to be stopped. Then start the actual
2488	 * freeing. a) Any local lists. b) Any associations. c) The hash of
2489	 * all associations. d) finally the ep itself.
2490	 */
2491	struct sctp_pcb *m;
2492	struct sctp_inpcb *inp_save;
2493	struct sctp_tcb *asoc, *nasoc;
2494	struct sctp_laddr *laddr, *nladdr;
2495	struct inpcb *ip_pcb;
2496	struct socket *so;
2497
2498	struct sctp_queued_to_read *sq;
2499
2500
2501	int cnt;
2502	sctp_sharedkey_t *shared_key;
2503
2504
2505#ifdef SCTP_LOG_CLOSING
2506	sctp_log_closing(inp, NULL, 0);
2507#endif
2508
2509	SCTP_ITERATOR_LOCK();
2510	so = inp->sctp_socket;
2511	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2512		/* been here before.. eeks.. get out of here */
2513		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
2514		SCTP_ITERATOR_UNLOCK();
2515#ifdef SCTP_LOG_CLOSING
2516		sctp_log_closing(inp, NULL, 1);
2517#endif
2518		return;
2519	}
2520	SCTP_ASOC_CREATE_LOCK(inp);
2521	SCTP_INP_INFO_WLOCK();
2522
2523	SCTP_INP_WLOCK(inp);
2524	/* First time through we have the socket lock, after that no more. */
2525	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
2526		/*
2527		 * Once we are in we can remove the flag from = 1 is only
2528		 * passed from the actual closing routines that are called
2529		 * via the sockets layer.
2530		 */
2531		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
2532	}
2533	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
2534	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
2535
2536	if (inp->control) {
2537		sctp_m_freem(inp->control);
2538		inp->control = NULL;
2539	}
2540	if (inp->pkt) {
2541		sctp_m_freem(inp->pkt);
2542		inp->pkt = NULL;
2543	}
2544	m = &inp->sctp_ep;
2545	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
2546					 * here but I will be nice :> (i.e.
2547					 * ip_pcb = ep;) */
2548	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
2549		int cnt_in_sd;
2550
2551		cnt_in_sd = 0;
2552		for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
2553		    asoc = nasoc) {
2554			nasoc = LIST_NEXT(asoc, sctp_tcblist);
2555			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2556				/* Skip guys being freed */
2557				asoc->sctp_socket = NULL;
2558				cnt_in_sd++;
2559				continue;
2560			}
2561			if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
2562			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) {
2563				/*
2564				 * If we have data in queue, we don't want
2565				 * to just free since the app may have done,
2566				 * send()/close or connect/send/close. And
2567				 * it wants the data to get across first.
2568				 */
2569				if (asoc->asoc.total_output_queue_size == 0) {
2570					/*
2571					 * Just abandon things in the front
2572					 * states
2573					 */
2574					sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_2);
2575					continue;
2576				}
2577			}
2578			SCTP_TCB_LOCK(asoc);
2579			/* Disconnect the socket please */
2580			asoc->sctp_socket = NULL;
2581			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
2582			if ((asoc->asoc.size_on_reasm_queue > 0) ||
2583			    (asoc->asoc.control_pdapi) ||
2584			    (asoc->asoc.size_on_all_streams > 0) ||
2585			    (so && (so->so_rcv.sb_cc > 0))
2586			    ) {
2587				/* Left with Data unread */
2588				struct mbuf *op_err;
2589
2590				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2591				    0, M_DONTWAIT, 1, MT_DATA);
2592				if (op_err) {
2593					/* Fill in the user initiated abort */
2594					struct sctp_paramhdr *ph;
2595					uint32_t *ippp;
2596
2597					SCTP_BUF_LEN(op_err) =
2598					    sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
2599					ph = mtod(op_err,
2600					    struct sctp_paramhdr *);
2601					ph->param_type = htons(
2602					    SCTP_CAUSE_USER_INITIATED_ABT);
2603					ph->param_length = htons(SCTP_BUF_LEN(op_err));
2604					ippp = (uint32_t *) (ph + 1);
2605					*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3);
2606				}
2607				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
2608				sctp_send_abort_tcb(asoc, op_err);
2609				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2610				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2611				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2612					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2613				}
2614				sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4);
2615				continue;
2616			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
2617				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
2618				    (asoc->asoc.stream_queue_cnt == 0)
2619			    ) {
2620				if (asoc->asoc.locked_on_sending) {
2621					goto abort_anyway;
2622				}
2623				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
2624				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
2625					/*
2626					 * there is nothing queued to send,
2627					 * so I send shutdown
2628					 */
2629					sctp_send_shutdown(asoc, asoc->asoc.primary_destination);
2630					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2631					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2632						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2633					}
2634					asoc->asoc.state = SCTP_STATE_SHUTDOWN_SENT;
2635					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
2636					    asoc->asoc.primary_destination);
2637					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
2638					    asoc->asoc.primary_destination);
2639					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR);
2640				}
2641			} else {
2642				/* mark into shutdown pending */
2643				struct sctp_stream_queue_pending *sp;
2644
2645				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
2646				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
2647				    asoc->asoc.primary_destination);
2648				if (asoc->asoc.locked_on_sending) {
2649					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
2650					    sctp_streamhead);
2651					if (sp == NULL) {
2652						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
2653						    asoc->asoc.locked_on_sending,
2654						    asoc->asoc.locked_on_sending->stream_no);
2655					} else {
2656						if ((sp->length == 0) && (sp->msg_is_complete == 0))
2657							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
2658					}
2659				}
2660				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
2661				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
2662				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
2663					struct mbuf *op_err;
2664
2665			abort_anyway:
2666					op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2667					    0, M_DONTWAIT, 1, MT_DATA);
2668					if (op_err) {
2669						/*
2670						 * Fill in the user
2671						 * initiated abort
2672						 */
2673						struct sctp_paramhdr *ph;
2674						uint32_t *ippp;
2675
2676						SCTP_BUF_LEN(op_err) =
2677						    (sizeof(struct sctp_paramhdr) +
2678						    sizeof(uint32_t));
2679						ph = mtod(op_err,
2680						    struct sctp_paramhdr *);
2681						ph->param_type = htons(
2682						    SCTP_CAUSE_USER_INITIATED_ABT);
2683						ph->param_length = htons(SCTP_BUF_LEN(op_err));
2684						ippp = (uint32_t *) (ph + 1);
2685						*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5);
2686					}
2687					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
2688					sctp_send_abort_tcb(asoc, op_err);
2689					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2690					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2691					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2692						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2693					}
2694					sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_6);
2695					continue;
2696				}
2697			}
2698			cnt_in_sd++;
2699			SCTP_TCB_UNLOCK(asoc);
2700		}
2701		/* now is there some left in our SHUTDOWN state? */
2702		if (cnt_in_sd) {
2703			SCTP_INP_WUNLOCK(inp);
2704			SCTP_ASOC_CREATE_UNLOCK(inp);
2705			SCTP_INP_INFO_WUNLOCK();
2706			SCTP_ITERATOR_UNLOCK();
2707#ifdef SCTP_LOG_CLOSING
2708			sctp_log_closing(inp, NULL, 2);
2709#endif
2710			return;
2711		}
2712	}
2713	inp->sctp_socket = NULL;
2714	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
2715	    SCTP_PCB_FLAGS_UNBOUND) {
2716		/*
2717		 * ok, this guy has been bound. It's port is somewhere in
2718		 * the sctppcbinfo hash table. Remove it!
2719		 */
2720		LIST_REMOVE(inp, sctp_hash);
2721		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
2722	}
2723	/*
2724	 * If there is a timer running to kill us, forget it, since it may
2725	 * have a contest on the INP lock.. which would cause us to die ...
2726	 */
2727	cnt = 0;
2728	for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
2729	    asoc = nasoc) {
2730		nasoc = LIST_NEXT(asoc, sctp_tcblist);
2731		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2732			cnt++;
2733			continue;
2734		}
2735		/* Free associations that are NOT killing us */
2736		SCTP_TCB_LOCK(asoc);
2737		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
2738		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
2739			struct mbuf *op_err;
2740			uint32_t *ippp;
2741
2742			op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2743			    0, M_DONTWAIT, 1, MT_DATA);
2744			if (op_err) {
2745				/* Fill in the user initiated abort */
2746				struct sctp_paramhdr *ph;
2747
2748				SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) +
2749				    sizeof(uint32_t));
2750				ph = mtod(op_err, struct sctp_paramhdr *);
2751				ph->param_type = htons(
2752				    SCTP_CAUSE_USER_INITIATED_ABT);
2753				ph->param_length = htons(SCTP_BUF_LEN(op_err));
2754				ippp = (uint32_t *) (ph + 1);
2755				*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7);
2756
2757			}
2758			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
2759			sctp_send_abort_tcb(asoc, op_err);
2760			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2761		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2762			cnt++;
2763			SCTP_TCB_UNLOCK(asoc);
2764			continue;
2765		}
2766		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2767		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2768			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2769		}
2770		sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8);
2771	}
2772	if (cnt) {
2773		/* Ok we have someone out there that will kill us */
2774		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2775		SCTP_INP_WUNLOCK(inp);
2776		SCTP_ASOC_CREATE_UNLOCK(inp);
2777		SCTP_INP_INFO_WUNLOCK();
2778		SCTP_ITERATOR_UNLOCK();
2779#ifdef SCTP_LOG_CLOSING
2780		sctp_log_closing(inp, NULL, 3);
2781#endif
2782		return;
2783	}
2784	if ((inp->refcount) || (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
2785		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2786		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
2787		SCTP_INP_WUNLOCK(inp);
2788		SCTP_ASOC_CREATE_UNLOCK(inp);
2789		SCTP_INP_INFO_WUNLOCK();
2790		SCTP_ITERATOR_UNLOCK();
2791#ifdef SCTP_LOG_CLOSING
2792		sctp_log_closing(inp, NULL, 4);
2793#endif
2794		return;
2795	}
2796	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2797	inp->sctp_ep.signature_change.type = 0;
2798	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
2799
2800#ifdef SCTP_LOG_CLOSING
2801	sctp_log_closing(inp, NULL, 5);
2802#endif
2803
2804	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2805	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NONE;
2806	/* Clear the read queue */
2807	/* sa_ignore FREED_MEMORY */
2808	while ((sq = TAILQ_FIRST(&inp->read_queue)) != NULL) {
2809		/* Its only abandoned if it had data left */
2810		if (sq->length)
2811			SCTP_STAT_INCR(sctps_left_abandon);
2812
2813		TAILQ_REMOVE(&inp->read_queue, sq, next);
2814		sctp_free_remote_addr(sq->whoFrom);
2815		if (so)
2816			so->so_rcv.sb_cc -= sq->length;
2817		if (sq->data) {
2818			sctp_m_freem(sq->data);
2819			sq->data = NULL;
2820		}
2821		/*
2822		 * no need to free the net count, since at this point all
2823		 * assoc's are gone.
2824		 */
2825		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, sq);
2826		SCTP_DECR_READQ_COUNT();
2827	}
2828	/* Now the sctp_pcb things */
2829	/*
2830	 * free each asoc if it is not already closed/free. we can't use the
2831	 * macro here since le_next will get freed as part of the
2832	 * sctp_free_assoc() call.
2833	 */
2834	cnt = 0;
2835	if (so) {
2836#ifdef FAST_IPSEC
2837		ipsec4_delete_pcbpolicy(ip_pcb);
2838#endif				/* FAST_IPSEC */
2839
2840		/* Unlocks not needed since the socket is gone now */
2841	}
2842	if (ip_pcb->inp_options) {
2843		(void)sctp_m_free(ip_pcb->inp_options);
2844		ip_pcb->inp_options = 0;
2845	}
2846	if (ip_pcb->inp_moptions) {
2847		inp_freemoptions(ip_pcb->inp_moptions);
2848		ip_pcb->inp_moptions = 0;
2849	}
2850#ifdef INET6
2851	if (ip_pcb->inp_vflag & INP_IPV6) {
2852		struct in6pcb *in6p;
2853
2854		in6p = (struct in6pcb *)inp;
2855		ip6_freepcbopts(in6p->in6p_outputopts);
2856	}
2857#endif				/* INET6 */
2858	ip_pcb->inp_vflag = 0;
2859	/* free up authentication fields */
2860	if (inp->sctp_ep.local_auth_chunks != NULL)
2861		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
2862	if (inp->sctp_ep.local_hmacs != NULL)
2863		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
2864
2865	shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
2866	while (shared_key) {
2867		LIST_REMOVE(shared_key, next);
2868		sctp_free_sharedkey(shared_key);
2869		/* sa_ignore FREED_MEMORY */
2870		shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
2871	}
2872
2873	inp_save = LIST_NEXT(inp, sctp_list);
2874	LIST_REMOVE(inp, sctp_list);
2875
2876	/* fix any iterators only after out of the list */
2877	sctp_iterator_inp_being_freed(inp, inp_save);
2878	/*
2879	 * if we have an address list the following will free the list of
2880	 * ifaddr's that are set into this ep. Again macro limitations here,
2881	 * since the LIST_FOREACH could be a bad idea.
2882	 */
2883	for ((laddr = LIST_FIRST(&inp->sctp_addr_list)); laddr != NULL;
2884	    laddr = nladdr) {
2885		nladdr = LIST_NEXT(laddr, sctp_nxt_addr);
2886		sctp_remove_laddr(laddr);
2887	}
2888
2889#ifdef SCTP_TRACK_FREED_ASOCS
2890	/* TEMP CODE */
2891	for ((asoc = LIST_FIRST(&inp->sctp_asoc_free_list)); asoc != NULL;
2892	    asoc = nasoc) {
2893		nasoc = LIST_NEXT(asoc, sctp_tcblist);
2894		LIST_REMOVE(asoc, sctp_tcblist);
2895		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, asoc);
2896		SCTP_DECR_ASOC_COUNT();
2897	}
2898	/* *** END TEMP CODE *** */
2899#endif
2900	/* Now lets see about freeing the EP hash table. */
2901	if (inp->sctp_tcbhash != NULL) {
2902		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
2903		inp->sctp_tcbhash = NULL;
2904	}
2905	/* Now we must put the ep memory back into the zone pool */
2906	INP_LOCK_DESTROY(&inp->ip_inp.inp);
2907	SCTP_INP_LOCK_DESTROY(inp);
2908	SCTP_INP_READ_DESTROY(inp);
2909	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
2910	SCTP_INP_INFO_WUNLOCK();
2911
2912	SCTP_ITERATOR_UNLOCK();
2913
2914	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
2915	SCTP_DECR_EP_COUNT();
2916
2917}
2918
2919
2920struct sctp_nets *
2921sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
2922{
2923	struct sctp_nets *net;
2924
2925	/* locate the address */
2926	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2927		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
2928			return (net);
2929	}
2930	return (NULL);
2931}
2932
2933
2934int
2935sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
2936{
2937	struct sctp_ifa *sctp_ifa;
2938
2939	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, 0);
2940	if (sctp_ifa) {
2941		return (1);
2942	} else {
2943		return (0);
2944	}
2945}
2946
2947void
2948sctp_set_initial_cc_param(struct sctp_tcb *stcb, struct sctp_nets *net)
2949{
2950	net->cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND));
2951	/* we always get at LEAST 2 MTU's */
2952	if (net->cwnd < (2 * net->mtu)) {
2953		net->cwnd = 2 * net->mtu;
2954	}
2955	net->ssthresh = stcb->asoc.peers_rwnd;
2956}
2957
2958
2959
2960/*
2961 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
2962 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
2963 * stats of stuff.
2964 */
2965int
2966sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
2967    int set_scope, int from)
2968{
2969	/*
2970	 * The following is redundant to the same lines in the
2971	 * sctp_aloc_assoc() but is needed since other's call the add
2972	 * address function
2973	 */
2974	struct sctp_nets *net, *netfirst;
2975	int addr_inscope;
2976
2977	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
2978	    from);
2979	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
2980
2981	netfirst = sctp_findnet(stcb, newaddr);
2982	if (netfirst) {
2983		/*
2984		 * Lie and return ok, we don't want to make the association
2985		 * go away for this behavior. It will happen in the TCP
2986		 * model in a connected socket. It does not reach the hash
2987		 * table until after the association is built so it can't be
2988		 * found. Mark as reachable, since the initial creation will
2989		 * have been cleared and the NOT_IN_ASSOC flag will have
2990		 * been added... and we don't want to end up removing it
2991		 * back out.
2992		 */
2993		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
2994			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
2995			    SCTP_ADDR_UNCONFIRMED);
2996		} else {
2997			netfirst->dest_state = SCTP_ADDR_REACHABLE;
2998		}
2999
3000		return (0);
3001	}
3002	addr_inscope = 1;
3003	if (newaddr->sa_family == AF_INET) {
3004		struct sockaddr_in *sin;
3005
3006		sin = (struct sockaddr_in *)newaddr;
3007		if (sin->sin_addr.s_addr == 0) {
3008			/* Invalid address */
3009			return (-1);
3010		}
3011		/* zero out the bzero area */
3012		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3013
3014		/* assure len is set */
3015		sin->sin_len = sizeof(struct sockaddr_in);
3016		if (set_scope) {
3017#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3018			stcb->ipv4_local_scope = 1;
3019#else
3020			if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3021				stcb->asoc.ipv4_local_scope = 1;
3022			}
3023#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3024		} else {
3025			/* Validate the address is in scope */
3026			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3027			    (stcb->asoc.ipv4_local_scope == 0)) {
3028				addr_inscope = 0;
3029			}
3030		}
3031	} else if (newaddr->sa_family == AF_INET6) {
3032		struct sockaddr_in6 *sin6;
3033
3034		sin6 = (struct sockaddr_in6 *)newaddr;
3035		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3036			/* Invalid address */
3037			return (-1);
3038		}
3039		/* assure len is set */
3040		sin6->sin6_len = sizeof(struct sockaddr_in6);
3041		if (set_scope) {
3042			if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3043				stcb->asoc.loopback_scope = 1;
3044				stcb->asoc.local_scope = 0;
3045				stcb->asoc.ipv4_local_scope = 1;
3046				stcb->asoc.site_scope = 1;
3047			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3048				/*
3049				 * If the new destination is a LINK_LOCAL we
3050				 * must have common site scope. Don't set
3051				 * the local scope since we may not share
3052				 * all links, only loopback can do this.
3053				 * Links on the local network would also be
3054				 * on our private network for v4 too.
3055				 */
3056				stcb->asoc.ipv4_local_scope = 1;
3057				stcb->asoc.site_scope = 1;
3058			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3059				/*
3060				 * If the new destination is SITE_LOCAL then
3061				 * we must have site scope in common.
3062				 */
3063				stcb->asoc.site_scope = 1;
3064			}
3065		} else {
3066			/* Validate the address is in scope */
3067			if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3068			    (stcb->asoc.loopback_scope == 0)) {
3069				addr_inscope = 0;
3070			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3071			    (stcb->asoc.local_scope == 0)) {
3072				addr_inscope = 0;
3073			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3074			    (stcb->asoc.site_scope == 0)) {
3075				addr_inscope = 0;
3076			}
3077		}
3078	} else {
3079		/* not supported family type */
3080		return (-1);
3081	}
3082	net = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_net, struct sctp_nets);
3083	if (net == NULL) {
3084		return (-1);
3085	}
3086	SCTP_INCR_RADDR_COUNT();
3087	bzero(net, sizeof(*net));
3088	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3089	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3090	if (newaddr->sa_family == AF_INET) {
3091		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3092	} else if (newaddr->sa_family == AF_INET6) {
3093		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3094	}
3095	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3096	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3097		stcb->asoc.loopback_scope = 1;
3098		stcb->asoc.ipv4_local_scope = 1;
3099		stcb->asoc.local_scope = 0;
3100		stcb->asoc.site_scope = 1;
3101		addr_inscope = 1;
3102	}
3103	net->failure_threshold = stcb->asoc.def_net_failure;
3104	if (addr_inscope == 0) {
3105		net->dest_state = (SCTP_ADDR_REACHABLE |
3106		    SCTP_ADDR_OUT_OF_SCOPE);
3107	} else {
3108		if (from == SCTP_ADDR_IS_CONFIRMED)
3109			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3110			net->dest_state = SCTP_ADDR_REACHABLE;
3111		else
3112			net->dest_state = SCTP_ADDR_REACHABLE |
3113			    SCTP_ADDR_UNCONFIRMED;
3114	}
3115	/*
3116	 * We set this to 0, the timer code knows that this means its an
3117	 * initial value
3118	 */
3119	net->RTO = 0;
3120	net->RTO_measured = 0;
3121	stcb->asoc.numnets++;
3122	*(&net->ref_count) = 1;
3123	net->tos_flowlabel = 0;
3124#ifdef INET
3125	if (newaddr->sa_family == AF_INET)
3126		net->tos_flowlabel = stcb->asoc.default_tos;
3127#endif
3128#ifdef INET6
3129	if (newaddr->sa_family == AF_INET6)
3130		net->tos_flowlabel = stcb->asoc.default_flowlabel;
3131#endif
3132	/* Init the timer structure */
3133	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3134	SCTP_OS_TIMER_INIT(&net->fr_timer.timer);
3135	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3136
3137	/* Now generate a route for this guy */
3138	/* KAME hack: embed scopeid */
3139	if (newaddr->sa_family == AF_INET6) {
3140		struct sockaddr_in6 *sin6;
3141
3142		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3143		(void)sa6_embedscope(sin6, ip6_use_defzone);
3144		sin6->sin6_scope_id = 0;
3145	}
3146	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3147
3148	if (newaddr->sa_family == AF_INET6) {
3149		struct sockaddr_in6 *sin6;
3150
3151		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3152		(void)sa6_recoverscope(sin6);
3153	}
3154	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3155		/* Get source address */
3156		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3157		    stcb,
3158		    (sctp_route_t *) & net->ro,
3159		    net,
3160		    0,
3161		    stcb->asoc.vrf_id);
3162		/* Now get the interface MTU */
3163		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3164			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3165		} else {
3166			net->mtu = 0;
3167		}
3168#ifdef SCTP_PRINT_FOR_B_AND_M
3169		SCTP_PRINTF("We have found an interface mtu of %d\n", net->mtu);
3170#endif
3171		if (net->mtu == 0) {
3172			/* Huh ?? */
3173			net->mtu = SCTP_DEFAULT_MTU;
3174		} else {
3175			uint32_t rmtu;
3176
3177			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3178#ifdef SCTP_PRINT_FOR_B_AND_M
3179			SCTP_PRINTF("The route mtu is %d\n", rmtu);
3180#endif
3181			if (rmtu == 0) {
3182				/*
3183				 * Start things off to match mtu of
3184				 * interface please.
3185				 */
3186				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3187				    net->ro.ro_rt, net->mtu);
3188			} else {
3189				/*
3190				 * we take the route mtu over the interface,
3191				 * since the route may be leading out the
3192				 * loopback, or a different interface.
3193				 */
3194				net->mtu = rmtu;
3195			}
3196		}
3197		if (from == SCTP_ALLOC_ASOC) {
3198#ifdef SCTP_PRINT_FOR_B_AND_M
3199			SCTP_PRINTF("New assoc sets mtu to :%d\n", net->mtu);
3200#endif
3201			stcb->asoc.smallest_mtu = net->mtu;
3202		}
3203	} else {
3204		net->mtu = stcb->asoc.smallest_mtu;
3205	}
3206	if (stcb->asoc.smallest_mtu > net->mtu) {
3207#ifdef SCTP_PRINT_FOR_B_AND_M
3208		SCTP_PRINTF("new address mtu:%d smaller than smallest:%d\n",
3209		    net->mtu, stcb->asoc.smallest_mtu);
3210#endif
3211		stcb->asoc.smallest_mtu = net->mtu;
3212	}
3213	/*
3214	 * We take the max of the burst limit times a MTU or the
3215	 * INITIAL_CWND. We then limit this to 4 MTU's of sending.
3216	 */
3217	sctp_set_initial_cc_param(stcb, net);
3218
3219
3220	if (sctp_logging_level & (SCTP_CWND_MONITOR_ENABLE | SCTP_CWND_LOGGING_ENABLE)) {
3221		sctp_log_cwnd(stcb, net, 0, SCTP_CWND_INITIALIZATION);
3222	}
3223	/*
3224	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3225	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3226	 */
3227	net->find_pseudo_cumack = 1;
3228	net->find_rtx_pseudo_cumack = 1;
3229	net->src_addr_selected = 0;
3230	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3231	if (net->ro.ro_rt == NULL) {
3232		/* Since we have no route put it at the back */
3233		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3234	} else if (netfirst == NULL) {
3235		/* We are the first one in the pool. */
3236		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3237	} else if (netfirst->ro.ro_rt == NULL) {
3238		/*
3239		 * First one has NO route. Place this one ahead of the first
3240		 * one.
3241		 */
3242		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3243	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
3244		/*
3245		 * This one has a different interface than the one at the
3246		 * top of the list. Place it ahead.
3247		 */
3248		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3249	} else {
3250		/*
3251		 * Ok we have the same interface as the first one. Move
3252		 * forward until we find either a) one with a NULL route...
3253		 * insert ahead of that b) one with a different ifp.. insert
3254		 * after that. c) end of the list.. insert at the tail.
3255		 */
3256		struct sctp_nets *netlook;
3257
3258		do {
3259			netlook = TAILQ_NEXT(netfirst, sctp_next);
3260			if (netlook == NULL) {
3261				/* End of the list */
3262				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3263				break;
3264			} else if (netlook->ro.ro_rt == NULL) {
3265				/* next one has NO route */
3266				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
3267				break;
3268			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
3269				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
3270				    net, sctp_next);
3271				break;
3272			}
3273			/* Shift forward */
3274			netfirst = netlook;
3275		} while (netlook != NULL);
3276	}
3277
3278	/* got to have a primary set */
3279	if (stcb->asoc.primary_destination == 0) {
3280		stcb->asoc.primary_destination = net;
3281	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
3282		    (net->ro.ro_rt) &&
3283	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
3284		/* No route to current primary adopt new primary */
3285		stcb->asoc.primary_destination = net;
3286	}
3287	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb,
3288	    net);
3289	/* Validate primary is first */
3290	net = TAILQ_FIRST(&stcb->asoc.nets);
3291	if ((net != stcb->asoc.primary_destination) &&
3292	    (stcb->asoc.primary_destination)) {
3293		/*
3294		 * first one on the list is NOT the primary sctp_cmpaddr()
3295		 * is much more efficent if the primary is the first on the
3296		 * list, make it so.
3297		 */
3298		TAILQ_REMOVE(&stcb->asoc.nets,
3299		    stcb->asoc.primary_destination, sctp_next);
3300		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
3301		    stcb->asoc.primary_destination, sctp_next);
3302	}
3303	return (0);
3304}
3305
3306
3307/*
3308 * allocate an association and add it to the endpoint. The caller must be
3309 * careful to add all additional addresses once they are know right away or
3310 * else the assoc will be may experience a blackout scenario.
3311 */
3312struct sctp_tcb *
3313sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
3314    int for_a_init, int *error, uint32_t override_tag, uint32_t vrf_id)
3315{
3316	struct sctp_tcb *stcb;
3317	struct sctp_association *asoc;
3318	struct sctpasochead *head;
3319	uint16_t rport;
3320	int err;
3321
3322	/*
3323	 * Assumption made here: Caller has done a
3324	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
3325	 * address does not exist already.
3326	 */
3327	if (sctppcbinfo.ipi_count_asoc >= SCTP_MAX_NUM_OF_ASOC) {
3328		/* Hit max assoc, sorry no more */
3329		*error = ENOBUFS;
3330		return (NULL);
3331	}
3332	if (firstaddr == NULL) {
3333		*error = EINVAL;
3334		return (NULL);
3335	}
3336	SCTP_INP_RLOCK(inp);
3337	if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
3338		/*
3339		 * If its in the TCP pool, its NOT allowed to create an
3340		 * association. The parent listener needs to call
3341		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
3342		 * off, or connected one does this.. its an error.
3343		 */
3344		SCTP_INP_RUNLOCK(inp);
3345		*error = EINVAL;
3346		return (NULL);
3347	}
3348	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
3349#ifdef SCTP_DEBUG
3350	if (firstaddr) {
3351		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
3352		SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
3353		    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
3354	} else {
3355		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
3356	}
3357#endif				/* SCTP_DEBUG */
3358	if (firstaddr->sa_family == AF_INET) {
3359		struct sockaddr_in *sin;
3360
3361		sin = (struct sockaddr_in *)firstaddr;
3362		if ((sin->sin_port == 0) || (sin->sin_addr.s_addr == 0)) {
3363			/* Invalid address */
3364			SCTP_INP_RUNLOCK(inp);
3365			*error = EINVAL;
3366			return (NULL);
3367		}
3368		rport = sin->sin_port;
3369	} else if (firstaddr->sa_family == AF_INET6) {
3370		struct sockaddr_in6 *sin6;
3371
3372		sin6 = (struct sockaddr_in6 *)firstaddr;
3373		if ((sin6->sin6_port == 0) ||
3374		    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
3375			/* Invalid address */
3376			SCTP_INP_RUNLOCK(inp);
3377			*error = EINVAL;
3378			return (NULL);
3379		}
3380		rport = sin6->sin6_port;
3381	} else {
3382		/* not supported family type */
3383		SCTP_INP_RUNLOCK(inp);
3384		*error = EINVAL;
3385		return (NULL);
3386	}
3387	SCTP_INP_RUNLOCK(inp);
3388	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
3389		/*
3390		 * If you have not performed a bind, then we need to do the
3391		 * ephemerial bind for you.
3392		 */
3393		if ((err = sctp_inpcb_bind(inp->sctp_socket,
3394		    (struct sockaddr *)NULL, (struct sctp_ifa *)NULL,
3395		    (struct thread *)NULL
3396		    ))) {
3397			/* bind error, probably perm */
3398			*error = err;
3399			return (NULL);
3400		}
3401	}
3402	stcb = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_asoc, struct sctp_tcb);
3403	if (stcb == NULL) {
3404		/* out of memory? */
3405		*error = ENOMEM;
3406		return (NULL);
3407	}
3408	SCTP_INCR_ASOC_COUNT();
3409
3410	bzero(stcb, sizeof(*stcb));
3411	asoc = &stcb->asoc;
3412	SCTP_TCB_LOCK_INIT(stcb);
3413	SCTP_TCB_SEND_LOCK_INIT(stcb);
3414	/* setup back pointer's */
3415	stcb->sctp_ep = inp;
3416	stcb->sctp_socket = inp->sctp_socket;
3417	if ((err = sctp_init_asoc(inp, stcb, for_a_init, override_tag, vrf_id))) {
3418		/* failed */
3419		SCTP_TCB_LOCK_DESTROY(stcb);
3420		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3421		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3422		SCTP_DECR_ASOC_COUNT();
3423		*error = err;
3424		return (NULL);
3425	}
3426	/* and the port */
3427	stcb->rport = rport;
3428	SCTP_INP_INFO_WLOCK();
3429	SCTP_INP_WLOCK(inp);
3430	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3431		/* inpcb freed while alloc going on */
3432		SCTP_TCB_LOCK_DESTROY(stcb);
3433		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3434		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3435		SCTP_INP_WUNLOCK(inp);
3436		SCTP_INP_INFO_WUNLOCK();
3437		SCTP_DECR_ASOC_COUNT();
3438		*error = EINVAL;
3439		return (NULL);
3440	}
3441	SCTP_TCB_LOCK(stcb);
3442
3443	/* now that my_vtag is set, add it to the hash */
3444	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag,
3445	    sctppcbinfo.hashasocmark)];
3446	/* put it in the bucket in the vtag hash of assoc's for the system */
3447	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
3448	SCTP_INP_INFO_WUNLOCK();
3449
3450	if ((err = sctp_add_remote_addr(stcb, firstaddr, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
3451		/* failure.. memory error? */
3452		if (asoc->strmout) {
3453			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
3454			asoc->strmout = NULL;
3455		}
3456		if (asoc->mapping_array) {
3457			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
3458			asoc->mapping_array = NULL;
3459		}
3460		SCTP_DECR_ASOC_COUNT();
3461		SCTP_TCB_LOCK_DESTROY(stcb);
3462		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3463		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3464		SCTP_INP_WUNLOCK(inp);
3465		*error = ENOBUFS;
3466		return (NULL);
3467	}
3468	/* Init all the timers */
3469	SCTP_OS_TIMER_INIT(&asoc->hb_timer.timer);
3470	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
3471	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
3472	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
3473	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
3474	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
3475	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
3476
3477	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
3478	/* now file the port under the hash as well */
3479	if (inp->sctp_tcbhash != NULL) {
3480		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
3481		    inp->sctp_hashmark)];
3482		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
3483	}
3484	SCTP_INP_WUNLOCK(inp);
3485	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", stcb);
3486	return (stcb);
3487}
3488
3489
3490void
3491sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
3492{
3493	struct sctp_association *asoc;
3494
3495	asoc = &stcb->asoc;
3496	asoc->numnets--;
3497	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
3498	if (net == asoc->primary_destination) {
3499		/* Reset primary */
3500		struct sctp_nets *lnet;
3501
3502		lnet = TAILQ_FIRST(&asoc->nets);
3503		/* Try to find a confirmed primary */
3504		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
3505	}
3506	if (net == asoc->last_data_chunk_from) {
3507		/* Reset primary */
3508		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
3509	}
3510	if (net == asoc->last_control_chunk_from) {
3511		/* Clear net */
3512		asoc->last_control_chunk_from = NULL;
3513	}
3514	sctp_free_remote_addr(net);
3515}
3516
3517/*
3518 * remove a remote endpoint address from an association, it will fail if the
3519 * address does not exist.
3520 */
3521int
3522sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
3523{
3524	/*
3525	 * Here we need to remove a remote address. This is quite simple, we
3526	 * first find it in the list of address for the association
3527	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
3528	 * on that item. Note we do not allow it to be removed if there are
3529	 * no other addresses.
3530	 */
3531	struct sctp_association *asoc;
3532	struct sctp_nets *net, *net_tmp;
3533
3534	asoc = &stcb->asoc;
3535
3536	/* locate the address */
3537	for (net = TAILQ_FIRST(&asoc->nets); net != NULL; net = net_tmp) {
3538		net_tmp = TAILQ_NEXT(net, sctp_next);
3539		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
3540			continue;
3541		}
3542		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
3543		    remaddr)) {
3544			/* we found the guy */
3545			if (asoc->numnets < 2) {
3546				/* Must have at LEAST two remote addresses */
3547				return (-1);
3548			} else {
3549				sctp_remove_net(stcb, net);
3550				return (0);
3551			}
3552		}
3553	}
3554	/* not found. */
3555	return (-2);
3556}
3557
3558
3559void
3560sctp_add_vtag_to_timewait(struct sctp_inpcb *inp, uint32_t tag, uint32_t time)
3561{
3562	struct sctpvtaghead *chain;
3563	struct sctp_tagblock *twait_block;
3564	struct timeval now;
3565	int set, i;
3566
3567	(void)SCTP_GETTIME_TIMEVAL(&now);
3568	chain = &sctppcbinfo.vtag_timewait[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
3569	set = 0;
3570	if (!SCTP_LIST_EMPTY(chain)) {
3571		/* Block(s) present, lets find space, and expire on the fly */
3572		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
3573			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
3574				if ((twait_block->vtag_block[i].v_tag == 0) &&
3575				    !set) {
3576					twait_block->vtag_block[i].tv_sec_at_expire =
3577					    now.tv_sec + time;
3578					twait_block->vtag_block[i].v_tag = tag;
3579					set = 1;
3580				} else if ((twait_block->vtag_block[i].v_tag) &&
3581					    ((long)twait_block->vtag_block[i].tv_sec_at_expire >
3582				    now.tv_sec)) {
3583					/* Audit expires this guy */
3584					twait_block->vtag_block[i].tv_sec_at_expire = 0;
3585					twait_block->vtag_block[i].v_tag = 0;
3586					if (set == 0) {
3587						/* Reuse it for my new tag */
3588						twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + SCTP_TIME_WAIT;
3589						twait_block->vtag_block[0].v_tag = tag;
3590						set = 1;
3591					}
3592				}
3593			}
3594			if (set) {
3595				/*
3596				 * We only do up to the block where we can
3597				 * place our tag for audits
3598				 */
3599				break;
3600			}
3601		}
3602	}
3603	/* Need to add a new block to chain */
3604	if (!set) {
3605		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
3606		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
3607		if (twait_block == NULL) {
3608			return;
3609		}
3610		memset(twait_block, 0, sizeof(struct sctp_tagblock));
3611		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
3612		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec +
3613		    SCTP_TIME_WAIT;
3614		twait_block->vtag_block[0].v_tag = tag;
3615	}
3616}
3617
3618
3619static void
3620sctp_iterator_asoc_being_freed(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
3621{
3622	struct sctp_iterator *it;
3623
3624	/*
3625	 * Unlock the tcb lock we do this so we avoid a dead lock scenario
3626	 * where the iterator is waiting on the TCB lock and the TCB lock is
3627	 * waiting on the iterator lock.
3628	 */
3629	it = stcb->asoc.stcb_starting_point_for_iterator;
3630	if (it == NULL) {
3631		return;
3632	}
3633	if (it->inp != stcb->sctp_ep) {
3634		/* hmm, focused on the wrong one? */
3635		return;
3636	}
3637	if (it->stcb != stcb) {
3638		return;
3639	}
3640	it->stcb = LIST_NEXT(stcb, sctp_tcblist);
3641	if (it->stcb == NULL) {
3642		/* done with all asoc's in this assoc */
3643		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3644			it->inp = NULL;
3645		} else {
3646			it->inp = LIST_NEXT(inp, sctp_list);
3647		}
3648	}
3649}
3650
3651
3652/*
3653 * Free the association after un-hashing the remote port.
3654 */
3655int
3656sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
3657{
3658	int i;
3659	struct sctp_association *asoc;
3660	struct sctp_nets *net, *prev;
3661	struct sctp_laddr *laddr;
3662	struct sctp_tmit_chunk *chk;
3663	struct sctp_asconf_addr *aparam;
3664	struct sctp_stream_reset_list *liste;
3665	struct sctp_queued_to_read *sq;
3666	struct sctp_stream_queue_pending *sp;
3667	sctp_sharedkey_t *shared_key;
3668	struct socket *so;
3669	int ccnt = 0;
3670	int cnt = 0;
3671
3672	/* first, lets purge the entry from the hash table. */
3673
3674#ifdef SCTP_LOG_CLOSING
3675	sctp_log_closing(inp, stcb, 6);
3676#endif
3677	if (stcb->asoc.state == 0) {
3678#ifdef SCTP_LOG_CLOSING
3679		sctp_log_closing(inp, NULL, 7);
3680#endif
3681		/* there is no asoc, really TSNH :-0 */
3682		return (1);
3683	}
3684	/* TEMP CODE */
3685	if (stcb->freed_from_where == 0) {
3686		/* Only record the first place free happened from */
3687		stcb->freed_from_where = from_location;
3688	}
3689	/* TEMP CODE */
3690
3691	asoc = &stcb->asoc;
3692	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3693	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3694		/* nothing around */
3695		so = NULL;
3696	else
3697		so = inp->sctp_socket;
3698
3699	/*
3700	 * We used timer based freeing if a reader or writer is in the way.
3701	 * So we first check if we are actually being called from a timer,
3702	 * if so we abort early if a reader or writer is still in the way.
3703	 */
3704	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
3705	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
3706		/*
3707		 * is it the timer driving us? if so are the reader/writers
3708		 * gone?
3709		 */
3710		if (stcb->asoc.refcnt) {
3711			/* nope, reader or writer in the way */
3712			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
3713			/* no asoc destroyed */
3714			SCTP_TCB_UNLOCK(stcb);
3715#ifdef SCTP_LOG_CLOSING
3716			sctp_log_closing(inp, stcb, 8);
3717#endif
3718			return (0);
3719		}
3720	}
3721	/* now clean up any other timers */
3722	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
3723	asoc->hb_timer.self = NULL;
3724	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
3725	asoc->dack_timer.self = NULL;
3726	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3727	/*-
3728	 * For stream reset we don't blast this unless
3729	 * it is a str-reset timer, it might be the
3730	 * free-asoc timer which we DON'T want to
3731	 * disturb.
3732	 */
3733	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
3734		asoc->strreset_timer.self = NULL;
3735	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
3736	asoc->asconf_timer.self = NULL;
3737	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
3738	asoc->autoclose_timer.self = NULL;
3739	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
3740	asoc->shut_guard_timer.self = NULL;
3741	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
3742	asoc->delayed_event_timer.self = NULL;
3743	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
3744		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
3745		net->fr_timer.self = NULL;
3746		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
3747		net->rxt_timer.self = NULL;
3748		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
3749		net->pmtu_timer.self = NULL;
3750	}
3751	/* Now the read queue needs to be cleaned up (only once) */
3752	cnt = 0;
3753	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
3754		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
3755		SCTP_INP_READ_LOCK(inp);
3756		TAILQ_FOREACH(sq, &inp->read_queue, next) {
3757			if (sq->stcb == stcb) {
3758				sq->do_not_ref_stcb = 1;
3759				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
3760				/*
3761				 * If there is no end, there never will be
3762				 * now.
3763				 */
3764				if (sq->end_added == 0) {
3765					/* Held for PD-API clear that. */
3766					sq->pdapi_aborted = 1;
3767					sq->held_length = 0;
3768					if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT)) {
3769						/*
3770						 * Need to add a PD-API
3771						 * aborted indication.
3772						 * Setting the control_pdapi
3773						 * assures that it will be
3774						 * added right after this
3775						 * msg.
3776						 */
3777						uint32_t strseq;
3778
3779						stcb->asoc.control_pdapi = sq;
3780						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
3781						sctp_notify_partial_delivery_indication(stcb,
3782						    SCTP_PARTIAL_DELIVERY_ABORTED, 1, strseq);
3783						stcb->asoc.control_pdapi = NULL;
3784					}
3785				}
3786				/* Add an end to wake them */
3787				sq->end_added = 1;
3788				cnt++;
3789			}
3790		}
3791		SCTP_INP_READ_UNLOCK(inp);
3792		if (stcb->block_entry) {
3793			cnt++;
3794			stcb->block_entry->error = ECONNRESET;
3795			stcb->block_entry = NULL;
3796		}
3797	}
3798	if ((from_inpcbfree != SCTP_PCBFREE_FORCE) && (stcb->asoc.refcnt)) {
3799		/*
3800		 * reader or writer in the way, we have hopefully given him
3801		 * something to chew on above.
3802		 */
3803		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
3804		SCTP_TCB_UNLOCK(stcb);
3805		if (so) {
3806			SCTP_INP_RLOCK(inp);
3807			if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3808			    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3809				/* nothing around */
3810				so = NULL;
3811			if (so) {
3812				/* Wake any reader/writers */
3813				sctp_sorwakeup(inp, so);
3814				sctp_sowwakeup(inp, so);
3815			}
3816			SCTP_INP_RUNLOCK(inp);
3817
3818		}
3819#ifdef SCTP_LOG_CLOSING
3820		sctp_log_closing(inp, stcb, 9);
3821#endif
3822		/* no asoc destroyed */
3823		return (0);
3824	}
3825#ifdef SCTP_LOG_CLOSING
3826	sctp_log_closing(inp, stcb, 10);
3827#endif
3828	/*
3829	 * When I reach here, no others want to kill the assoc yet.. and I
3830	 * own the lock. Now its possible an abort comes in when I do the
3831	 * lock exchange below to grab all the locks to do the final take
3832	 * out. to prevent this we increment the count, which will start a
3833	 * timer and blow out above thus assuring us that we hold exclusive
3834	 * killing of the asoc. Note that after getting back the TCB lock we
3835	 * will go ahead and increment the counter back up and stop any
3836	 * timer a passing stranger may have started :-S
3837	 */
3838	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3839		atomic_add_int(&stcb->asoc.refcnt, 1);
3840
3841		SCTP_TCB_UNLOCK(stcb);
3842
3843		SCTP_ITERATOR_LOCK();
3844		SCTP_INP_INFO_WLOCK();
3845		SCTP_INP_WLOCK(inp);
3846		SCTP_TCB_LOCK(stcb);
3847	}
3848	/* Double check the GONE flag */
3849	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3850	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3851		/* nothing around */
3852		so = NULL;
3853
3854	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
3855	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
3856		/*
3857		 * For TCP type we need special handling when we are
3858		 * connected. We also include the peel'ed off ones to.
3859		 */
3860		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
3861			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
3862			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
3863			if (so) {
3864				SOCK_LOCK(so);
3865				if (so->so_rcv.sb_cc == 0) {
3866					so->so_state &= ~(SS_ISCONNECTING |
3867					    SS_ISDISCONNECTING |
3868					    SS_ISCONFIRMING |
3869					    SS_ISCONNECTED);
3870				}
3871				SOCK_UNLOCK(so);
3872				sctp_sowwakeup(inp, so);
3873				sctp_sorwakeup(inp, so);
3874				SCTP_SOWAKEUP(so);
3875			}
3876		}
3877	}
3878	/*
3879	 * Make it invalid too, that way if its about to run it will abort
3880	 * and return.
3881	 */
3882	sctp_iterator_asoc_being_freed(inp, stcb);
3883	/* re-increment the lock */
3884	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3885		atomic_add_int(&stcb->asoc.refcnt, -1);
3886	}
3887	asoc->state = 0;
3888	if (inp->sctp_tcbhash) {
3889		LIST_REMOVE(stcb, sctp_tcbhash);
3890	}
3891	if (stcb->asoc.in_restart_hash) {
3892		LIST_REMOVE(stcb, sctp_tcbrestarhash);
3893	}
3894	/* Now lets remove it from the list of ALL associations in the EP */
3895	LIST_REMOVE(stcb, sctp_tcblist);
3896	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3897		SCTP_INP_INCR_REF(inp);
3898		SCTP_INP_WUNLOCK(inp);
3899		SCTP_ITERATOR_UNLOCK();
3900	}
3901	/* pull from vtag hash */
3902	LIST_REMOVE(stcb, sctp_asocs);
3903	sctp_add_vtag_to_timewait(inp, asoc->my_vtag, SCTP_TIME_WAIT);
3904
3905
3906	/*
3907	 * Now restop the timers to be sure - this is paranoia at is finest!
3908	 */
3909	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3910	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
3911	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
3912	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3913	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
3914	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
3915	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
3916	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
3917
3918	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
3919		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
3920		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
3921		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
3922	}
3923
3924	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
3925	prev = NULL;
3926	/*
3927	 * The chunk lists and such SHOULD be empty but we check them just
3928	 * in case.
3929	 */
3930	/* anything on the wheel needs to be removed */
3931	for (i = 0; i < asoc->streamoutcnt; i++) {
3932		struct sctp_stream_out *outs;
3933
3934		outs = &asoc->strmout[i];
3935		/* now clean up any chunks here */
3936		sp = TAILQ_FIRST(&outs->outqueue);
3937		while (sp) {
3938			TAILQ_REMOVE(&outs->outqueue, sp, next);
3939			if (sp->data) {
3940				sctp_m_freem(sp->data);
3941				sp->data = NULL;
3942				sp->tail_mbuf = NULL;
3943			}
3944			sctp_free_remote_addr(sp->net);
3945			sctp_free_spbufspace(stcb, asoc, sp);
3946			/* Free the zone stuff  */
3947			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_strmoq, sp);
3948			SCTP_DECR_STRMOQ_COUNT();
3949			/* sa_ignore FREED_MEMORY */
3950			sp = TAILQ_FIRST(&outs->outqueue);
3951		}
3952	}
3953
3954	/* sa_ignore FREED_MEMORY */
3955	while ((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) {
3956		TAILQ_REMOVE(&asoc->resetHead, liste, next_resp);
3957		SCTP_FREE(liste, SCTP_M_STRESET);
3958	}
3959
3960	sq = TAILQ_FIRST(&asoc->pending_reply_queue);
3961	while (sq) {
3962		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
3963		if (sq->data) {
3964			sctp_m_freem(sq->data);
3965			sq->data = NULL;
3966		}
3967		sctp_free_remote_addr(sq->whoFrom);
3968		sq->whoFrom = NULL;
3969		sq->stcb = NULL;
3970		/* Free the ctl entry */
3971		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, sq);
3972		SCTP_DECR_READQ_COUNT();
3973		/* sa_ignore FREED_MEMORY */
3974		sq = TAILQ_FIRST(&asoc->pending_reply_queue);
3975	}
3976
3977	chk = TAILQ_FIRST(&asoc->free_chunks);
3978	while (chk) {
3979		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
3980		if (chk->data) {
3981			sctp_m_freem(chk->data);
3982			chk->data = NULL;
3983		}
3984		ccnt++;
3985		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
3986		SCTP_DECR_CHK_COUNT();
3987		atomic_subtract_int(&sctppcbinfo.ipi_free_chunks, 1);
3988		asoc->free_chunk_cnt--;
3989		/* sa_ignore FREED_MEMORY */
3990		chk = TAILQ_FIRST(&asoc->free_chunks);
3991	}
3992	/* pending send queue SHOULD be empty */
3993	if (!TAILQ_EMPTY(&asoc->send_queue)) {
3994		chk = TAILQ_FIRST(&asoc->send_queue);
3995		while (chk) {
3996			TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
3997			if (chk->data) {
3998				sctp_m_freem(chk->data);
3999				chk->data = NULL;
4000			}
4001			ccnt++;
4002			sctp_free_remote_addr(chk->whoTo);
4003			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4004			SCTP_DECR_CHK_COUNT();
4005			/* sa_ignore FREED_MEMORY */
4006			chk = TAILQ_FIRST(&asoc->send_queue);
4007		}
4008	}
4009/*
4010  if(ccnt) {
4011  printf("Freed %d from send_queue\n", ccnt);
4012  ccnt = 0;
4013  }
4014*/
4015	/* sent queue SHOULD be empty */
4016	if (!TAILQ_EMPTY(&asoc->sent_queue)) {
4017		chk = TAILQ_FIRST(&asoc->sent_queue);
4018		while (chk) {
4019			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4020			if (chk->data) {
4021				sctp_m_freem(chk->data);
4022				chk->data = NULL;
4023			}
4024			ccnt++;
4025			sctp_free_remote_addr(chk->whoTo);
4026			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4027			SCTP_DECR_CHK_COUNT();
4028			/* sa_ignore FREED_MEMORY */
4029			chk = TAILQ_FIRST(&asoc->sent_queue);
4030		}
4031	}
4032/*
4033  if(ccnt) {
4034  printf("Freed %d from sent_queue\n", ccnt);
4035  ccnt = 0;
4036  }
4037*/
4038	/* control queue MAY not be empty */
4039	if (!TAILQ_EMPTY(&asoc->control_send_queue)) {
4040		chk = TAILQ_FIRST(&asoc->control_send_queue);
4041		while (chk) {
4042			TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4043			if (chk->data) {
4044				sctp_m_freem(chk->data);
4045				chk->data = NULL;
4046			}
4047			ccnt++;
4048			sctp_free_remote_addr(chk->whoTo);
4049			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4050			SCTP_DECR_CHK_COUNT();
4051			/* sa_ignore FREED_MEMORY */
4052			chk = TAILQ_FIRST(&asoc->control_send_queue);
4053		}
4054	}
4055/*
4056  if(ccnt) {
4057  printf("Freed %d from ctrl_queue\n", ccnt);
4058  ccnt = 0;
4059  }
4060*/
4061	if (!TAILQ_EMPTY(&asoc->reasmqueue)) {
4062		chk = TAILQ_FIRST(&asoc->reasmqueue);
4063		while (chk) {
4064			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
4065			if (chk->data) {
4066				sctp_m_freem(chk->data);
4067				chk->data = NULL;
4068			}
4069			sctp_free_remote_addr(chk->whoTo);
4070			ccnt++;
4071			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4072			SCTP_DECR_CHK_COUNT();
4073			/* sa_ignore FREED_MEMORY */
4074			chk = TAILQ_FIRST(&asoc->reasmqueue);
4075		}
4076	}
4077/*
4078  if(ccnt) {
4079  printf("Freed %d from reasm_queue\n", ccnt);
4080  ccnt = 0;
4081  }
4082*/
4083	if (asoc->mapping_array) {
4084		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4085		asoc->mapping_array = NULL;
4086	}
4087	/* the stream outs */
4088	if (asoc->strmout) {
4089		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4090		asoc->strmout = NULL;
4091	}
4092	asoc->streamoutcnt = 0;
4093	if (asoc->strmin) {
4094		struct sctp_queued_to_read *ctl;
4095
4096		for (i = 0; i < asoc->streamincnt; i++) {
4097			if (!TAILQ_EMPTY(&asoc->strmin[i].inqueue)) {
4098				/* We have somethings on the streamin queue */
4099				ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4100				while (ctl) {
4101					TAILQ_REMOVE(&asoc->strmin[i].inqueue,
4102					    ctl, next);
4103					sctp_free_remote_addr(ctl->whoFrom);
4104					if (ctl->data) {
4105						sctp_m_freem(ctl->data);
4106						ctl->data = NULL;
4107					}
4108					/*
4109					 * We don't free the address here
4110					 * since all the net's were freed
4111					 * above.
4112					 */
4113					SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, ctl);
4114					SCTP_DECR_READQ_COUNT();
4115					ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4116				}
4117			}
4118		}
4119		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
4120		asoc->strmin = NULL;
4121	}
4122	asoc->streamincnt = 0;
4123	while (!TAILQ_EMPTY(&asoc->nets)) {
4124		/* sa_ignore FREED_MEMORY */
4125		net = TAILQ_FIRST(&asoc->nets);
4126		/* pull from list */
4127		if ((sctppcbinfo.ipi_count_raddr == 0) || (prev == net)) {
4128#ifdef INVARIANTS
4129			panic("no net's left alloc'ed, or list points to itself");
4130#endif
4131			break;
4132		}
4133		prev = net;
4134		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4135		sctp_free_remote_addr(net);
4136	}
4137
4138	while (!SCTP_LIST_EMPTY(&asoc->sctp_restricted_addrs)) {
4139		/* sa_ignore FREED_MEMORY */
4140		laddr = LIST_FIRST(&asoc->sctp_restricted_addrs);
4141		sctp_remove_laddr(laddr);
4142	}
4143
4144	/* pending asconf (address) parameters */
4145	while (!TAILQ_EMPTY(&asoc->asconf_queue)) {
4146		/* sa_ignore FREED_MEMORY */
4147		aparam = TAILQ_FIRST(&asoc->asconf_queue);
4148		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
4149		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
4150	}
4151	if (asoc->last_asconf_ack_sent != NULL) {
4152		sctp_m_freem(asoc->last_asconf_ack_sent);
4153		asoc->last_asconf_ack_sent = NULL;
4154	}
4155	/* clean up auth stuff */
4156	if (asoc->local_hmacs)
4157		sctp_free_hmaclist(asoc->local_hmacs);
4158	if (asoc->peer_hmacs)
4159		sctp_free_hmaclist(asoc->peer_hmacs);
4160
4161	if (asoc->local_auth_chunks)
4162		sctp_free_chunklist(asoc->local_auth_chunks);
4163	if (asoc->peer_auth_chunks)
4164		sctp_free_chunklist(asoc->peer_auth_chunks);
4165
4166	sctp_free_authinfo(&asoc->authinfo);
4167
4168	shared_key = LIST_FIRST(&asoc->shared_keys);
4169	while (shared_key) {
4170		LIST_REMOVE(shared_key, next);
4171		sctp_free_sharedkey(shared_key);
4172		/* sa_ignore FREED_MEMORY */
4173		shared_key = LIST_FIRST(&asoc->shared_keys);
4174	}
4175
4176	/* Insert new items here :> */
4177
4178	/* Get rid of LOCK */
4179	SCTP_TCB_LOCK_DESTROY(stcb);
4180	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4181	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4182		SCTP_INP_INFO_WUNLOCK();
4183		SCTP_INP_RLOCK(inp);
4184	}
4185#ifdef SCTP_TRACK_FREED_ASOCS
4186	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
4187		/* now clean up the tasoc itself */
4188		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
4189		SCTP_DECR_ASOC_COUNT();
4190	} else {
4191		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
4192	}
4193#else
4194	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
4195	SCTP_DECR_ASOC_COUNT();
4196#endif
4197	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4198		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
4199			/*
4200			 * If its NOT the inp_free calling us AND sctp_close
4201			 * as been called, we call back...
4202			 */
4203			SCTP_INP_RUNLOCK(inp);
4204			/*
4205			 * This will start the kill timer (if we are the
4206			 * lastone) since we hold an increment yet. But this
4207			 * is the only safe way to do this since otherwise
4208			 * if the socket closes at the same time we are here
4209			 * we might collide in the cleanup.
4210			 */
4211			sctp_inpcb_free(inp,
4212			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
4213			    SCTP_CALLED_DIRECTLY_NOCMPSET);
4214			SCTP_INP_DECR_REF(inp);
4215			goto out_of;
4216		} else {
4217			/* The socket is still open. */
4218			SCTP_INP_DECR_REF(inp);
4219		}
4220	}
4221	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4222		SCTP_INP_RUNLOCK(inp);
4223	}
4224out_of:
4225	/* destroyed the asoc */
4226#ifdef SCTP_LOG_CLOSING
4227	sctp_log_closing(inp, NULL, 11);
4228#endif
4229	return (1);
4230}
4231
4232
4233
4234/*
4235 * determine if a destination is "reachable" based upon the addresses bound
4236 * to the current endpoint (e.g. only v4 or v6 currently bound)
4237 */
4238/*
4239 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
4240 * assoc level v4/v6 flags, as the assoc *may* not have the same address
4241 * types bound as its endpoint
4242 */
4243int
4244sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
4245{
4246	struct sctp_inpcb *inp;
4247	int answer;
4248
4249	/*
4250	 * No locks here, the TCB, in all cases is already locked and an
4251	 * assoc is up. There is either a INP lock by the caller applied (in
4252	 * asconf case when deleting an address) or NOT in the HB case,
4253	 * however if HB then the INP increment is up and the INP will not
4254	 * be removed (on top of the fact that we have a TCB lock). So we
4255	 * only want to read the sctp_flags, which is either bound-all or
4256	 * not.. no protection needed since once an assoc is up you can't be
4257	 * changing your binding.
4258	 */
4259	inp = stcb->sctp_ep;
4260	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4261		/* if bound all, destination is not restricted */
4262		/*
4263		 * RRS: Question during lock work: Is this correct? If you
4264		 * are bound-all you still might need to obey the V4--V6
4265		 * flags??? IMO this bound-all stuff needs to be removed!
4266		 */
4267		return (1);
4268	}
4269	/* NOTE: all "scope" checks are done when local addresses are added */
4270	if (destaddr->sa_family == AF_INET6) {
4271		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
4272	} else if (destaddr->sa_family == AF_INET) {
4273		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
4274	} else {
4275		/* invalid family, so it's unreachable */
4276		answer = 0;
4277	}
4278	return (answer);
4279}
4280
4281/*
4282 * update the inp_vflags on an endpoint
4283 */
4284static void
4285sctp_update_ep_vflag(struct sctp_inpcb *inp)
4286{
4287	struct sctp_laddr *laddr;
4288
4289	/* first clear the flag */
4290	inp->ip_inp.inp.inp_vflag = 0;
4291	/* set the flag based on addresses on the ep list */
4292	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4293		if (laddr->ifa == NULL) {
4294			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
4295			    __FUNCTION__);
4296			continue;
4297		}
4298		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
4299			continue;
4300		}
4301		if (laddr->ifa->address.sa.sa_family == AF_INET6) {
4302			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
4303		} else if (laddr->ifa->address.sa.sa_family == AF_INET) {
4304			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
4305		}
4306	}
4307}
4308
4309/*
4310 * Add the address to the endpoint local address list There is nothing to be
4311 * done if we are bound to all addresses
4312 */
4313void
4314sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
4315{
4316	struct sctp_laddr *laddr;
4317	int fnd, error = 0;
4318
4319	fnd = 0;
4320
4321	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4322		/* You are already bound to all. You have it already */
4323		return;
4324	}
4325	if (ifa->address.sa.sa_family == AF_INET6) {
4326		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
4327			/* Can't bind a non-useable addr. */
4328			return;
4329		}
4330	}
4331	/* first, is it already present? */
4332	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4333		if (laddr->ifa == ifa) {
4334			fnd = 1;
4335			break;
4336		}
4337	}
4338
4339	if (fnd == 0) {
4340		/* Not in the ep list */
4341		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
4342		if (error != 0)
4343			return;
4344		inp->laddr_count++;
4345		/* update inp_vflag flags */
4346		if (ifa->address.sa.sa_family == AF_INET6) {
4347			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
4348		} else if (ifa->address.sa.sa_family == AF_INET) {
4349			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
4350		}
4351	}
4352	return;
4353}
4354
4355
4356/*
4357 * select a new (hopefully reachable) destination net (should only be used
4358 * when we deleted an ep addr that is the only usable source address to reach
4359 * the destination net)
4360 */
4361static void
4362sctp_select_primary_destination(struct sctp_tcb *stcb)
4363{
4364	struct sctp_nets *net;
4365
4366	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4367		/* for now, we'll just pick the first reachable one we find */
4368		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
4369			continue;
4370		if (sctp_destination_is_reachable(stcb,
4371		    (struct sockaddr *)&net->ro._l_addr)) {
4372			/* found a reachable destination */
4373			stcb->asoc.primary_destination = net;
4374		}
4375	}
4376	/* I can't there from here! ...we're gonna die shortly... */
4377}
4378
4379
4380/*
4381 * Delete the address from the endpoint local address list There is nothing
4382 * to be done if we are bound to all addresses
4383 */
4384void
4385sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
4386{
4387	struct sctp_laddr *laddr;
4388	int fnd;
4389
4390	fnd = 0;
4391	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4392		/* You are already bound to all. You have it already */
4393		return;
4394	}
4395	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4396		if (laddr->ifa == ifa) {
4397			fnd = 1;
4398			break;
4399		}
4400	}
4401	if (fnd && (inp->laddr_count < 2)) {
4402		/* can't delete unless there are at LEAST 2 addresses */
4403		return;
4404	}
4405	if (fnd) {
4406		/*
4407		 * clean up any use of this address go through our
4408		 * associations and clear any last_used_address that match
4409		 * this one for each assoc, see if a new primary_destination
4410		 * is needed
4411		 */
4412		struct sctp_tcb *stcb;
4413
4414		/* clean up "next_addr_touse" */
4415		if (inp->next_addr_touse == laddr)
4416			/* delete this address */
4417			inp->next_addr_touse = NULL;
4418
4419		/* clean up "last_used_address" */
4420		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
4421			struct sctp_nets *net;
4422
4423			SCTP_TCB_LOCK(stcb);
4424			if (stcb->asoc.last_used_address == laddr)
4425				/* delete this address */
4426				stcb->asoc.last_used_address = NULL;
4427			/*
4428			 * Now spin through all the nets and purge any ref
4429			 * to laddr
4430			 */
4431			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4432				if (net->ro._s_addr &&
4433				    (net->ro._s_addr->ifa == laddr->ifa)) {
4434					/* Yep, purge src address selected */
4435					sctp_rtentry_t *rt;
4436
4437					/* delete this address if cached */
4438					rt = net->ro.ro_rt;
4439					if (rt != NULL) {
4440						RTFREE(rt);
4441						net->ro.ro_rt = NULL;
4442					}
4443					sctp_free_ifa(net->ro._s_addr);
4444					net->ro._s_addr = NULL;
4445					net->src_addr_selected = 0;
4446				}
4447			}
4448			SCTP_TCB_UNLOCK(stcb);
4449		}		/* for each tcb */
4450		/* remove it from the ep list */
4451		sctp_remove_laddr(laddr);
4452		inp->laddr_count--;
4453		/* update inp_vflag flags */
4454		sctp_update_ep_vflag(inp);
4455	}
4456	return;
4457}
4458
4459/*
4460 * Add the addr to the TCB local address list For the BOUNDALL or dynamic
4461 * case, this is a "pending" address list (eg. addresses waiting for an
4462 * ASCONF-ACK response) For the subset binding, static case, this is a
4463 * "valid" address list
4464 */
4465void
4466sctp_add_local_addr_assoc(struct sctp_tcb *stcb, struct sctp_ifa *ifa, int restricted_list)
4467{
4468	struct sctp_inpcb *inp;
4469	struct sctp_laddr *laddr;
4470	struct sctpladdr *list;
4471
4472	/*
4473	 * Assumes TCB is locked.. and possibly the INP. May need to
4474	 * confirm/fix that if we need it and is not the case.
4475	 */
4476	list = &stcb->asoc.sctp_restricted_addrs;
4477
4478	inp = stcb->sctp_ep;
4479	if (ifa->address.sa.sa_family == AF_INET6) {
4480		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
4481			/* Can't bind a non-existent addr. */
4482			return;
4483		}
4484	}
4485	/* does the address already exist? */
4486	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
4487		if (laddr->ifa == ifa) {
4488			return;
4489		}
4490	}
4491
4492	/* add to the list */
4493	(void)sctp_insert_laddr(list, ifa, 0);
4494	return;
4495}
4496
4497/*
4498 * insert an laddr entry with the given ifa for the desired list
4499 */
4500int
4501sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
4502{
4503	struct sctp_laddr *laddr;
4504
4505	laddr = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
4506	if (laddr == NULL) {
4507		/* out of memory? */
4508		return (EINVAL);
4509	}
4510	SCTP_INCR_LADDR_COUNT();
4511	bzero(laddr, sizeof(*laddr));
4512	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
4513	laddr->ifa = ifa;
4514	laddr->action = act;
4515	atomic_add_int(&ifa->refcount, 1);
4516	/* insert it */
4517	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
4518
4519	return (0);
4520}
4521
4522/*
4523 * Remove an laddr entry from the local address list (on an assoc)
4524 */
4525void
4526sctp_remove_laddr(struct sctp_laddr *laddr)
4527{
4528
4529	/* remove from the list */
4530	LIST_REMOVE(laddr, sctp_nxt_addr);
4531	sctp_free_ifa(laddr->ifa);
4532	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_laddr, laddr);
4533	SCTP_DECR_LADDR_COUNT();
4534}
4535
4536/*
4537 * Remove an address from the TCB local address list
4538 */
4539void
4540sctp_del_local_addr_assoc(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
4541{
4542	struct sctp_inpcb *inp;
4543	struct sctp_laddr *laddr;
4544
4545	/*
4546	 * This is called by asconf work. It is assumed that a) The TCB is
4547	 * locked and b) The INP is locked. This is true in as much as I can
4548	 * trace through the entry asconf code where I did these locks.
4549	 * Again, the ASCONF code is a bit different in that it does lock
4550	 * the INP during its work often times. This must be since we don't
4551	 * want other proc's looking up things while what they are looking
4552	 * up is changing :-D
4553	 */
4554
4555	inp = stcb->sctp_ep;
4556	/* if subset bound and don't allow ASCONF's, can't delete last */
4557	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
4558	    (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF) == 0)) {
4559		if (stcb->asoc.numnets < 2) {
4560			/* can't delete last address */
4561			return;
4562		}
4563	}
4564	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
4565		/* remove the address if it exists */
4566		if (laddr->ifa == NULL)
4567			continue;
4568		if (laddr->ifa == ifa) {
4569			sctp_remove_laddr(laddr);
4570			return;
4571		}
4572	}
4573
4574	/* address not found! */
4575	return;
4576}
4577
4578static char sctp_pcb_initialized = 0;
4579
4580/*
4581 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
4582 */
4583/* sysctl */
4584static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
4585static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
4586
4587void
4588sctp_pcb_init()
4589{
4590	/*
4591	 * SCTP initialization for the PCB structures should be called by
4592	 * the sctp_init() funciton.
4593	 */
4594	int i;
4595
4596	if (sctp_pcb_initialized != 0) {
4597		/* error I was called twice */
4598		return;
4599	}
4600	sctp_pcb_initialized = 1;
4601
4602	bzero(&sctpstat, sizeof(struct sctpstat));
4603	(void)SCTP_GETTIME_TIMEVAL(&sctpstat.sctps_discontinuitytime);
4604	/* init the empty list of (All) Endpoints */
4605	LIST_INIT(&sctppcbinfo.listhead);
4606
4607	/* init the iterator head */
4608	TAILQ_INIT(&sctppcbinfo.iteratorhead);
4609
4610	/* init the hash table of endpoints */
4611	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &sctp_hashtblsize);
4612	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &sctp_pcbtblsize);
4613	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &sctp_chunkscale);
4614	sctppcbinfo.sctp_asochash = SCTP_HASH_INIT((sctp_hashtblsize * 31),
4615	    &sctppcbinfo.hashasocmark);
4616	sctppcbinfo.sctp_ephash = SCTP_HASH_INIT(sctp_hashtblsize,
4617	    &sctppcbinfo.hashmark);
4618	sctppcbinfo.sctp_tcpephash = SCTP_HASH_INIT(sctp_hashtblsize,
4619	    &sctppcbinfo.hashtcpmark);
4620	sctppcbinfo.hashtblsize = sctp_hashtblsize;
4621
4622	/* init the small hash table we use to track restarted asoc's */
4623	sctppcbinfo.sctp_restarthash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE,
4624	    &sctppcbinfo.hashrestartmark);
4625
4626
4627	sctppcbinfo.sctp_vrfhash = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
4628	    &sctppcbinfo.hashvrfmark);
4629
4630	sctppcbinfo.vrf_ifn_hash = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
4631	    &sctppcbinfo.vrf_ifn_hashmark);
4632
4633	/* init the zones */
4634	/*
4635	 * FIX ME: Should check for NULL returns, but if it does fail we are
4636	 * doomed to panic anyways... add later maybe.
4637	 */
4638	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_ep, "sctp_ep",
4639	    sizeof(struct sctp_inpcb), maxsockets);
4640
4641	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_asoc, "sctp_asoc",
4642	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
4643
4644	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_laddr, "sctp_laddr",
4645	    sizeof(struct sctp_laddr),
4646	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
4647
4648	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_net, "sctp_raddr",
4649	    sizeof(struct sctp_nets),
4650	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
4651
4652	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_chunk, "sctp_chunk",
4653	    sizeof(struct sctp_tmit_chunk),
4654	    (sctp_max_number_of_assoc * sctp_chunkscale));
4655
4656	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_readq, "sctp_readq",
4657	    sizeof(struct sctp_queued_to_read),
4658	    (sctp_max_number_of_assoc * sctp_chunkscale));
4659
4660	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_strmoq, "sctp_stream_msg_out",
4661	    sizeof(struct sctp_stream_queue_pending),
4662	    (sctp_max_number_of_assoc * sctp_chunkscale));
4663
4664	/* Master Lock INIT for info structure */
4665	SCTP_INP_INFO_LOCK_INIT();
4666	SCTP_STATLOG_INIT_LOCK();
4667	SCTP_ITERATOR_LOCK_INIT();
4668
4669	SCTP_IPI_COUNT_INIT();
4670	SCTP_IPI_ADDR_INIT();
4671	SCTP_IPI_ITERATOR_WQ_INIT();
4672#ifdef SCTP_PACKET_LOGGING
4673	SCTP_IP_PKTLOG_INIT();
4674#endif
4675	LIST_INIT(&sctppcbinfo.addr_wq);
4676
4677	/* not sure if we need all the counts */
4678	sctppcbinfo.ipi_count_ep = 0;
4679	/* assoc/tcb zone info */
4680	sctppcbinfo.ipi_count_asoc = 0;
4681	/* local addrlist zone info */
4682	sctppcbinfo.ipi_count_laddr = 0;
4683	/* remote addrlist zone info */
4684	sctppcbinfo.ipi_count_raddr = 0;
4685	/* chunk info */
4686	sctppcbinfo.ipi_count_chunk = 0;
4687
4688	/* socket queue zone info */
4689	sctppcbinfo.ipi_count_readq = 0;
4690
4691	/* stream out queue cont */
4692	sctppcbinfo.ipi_count_strmoq = 0;
4693
4694	sctppcbinfo.ipi_free_strmoq = 0;
4695	sctppcbinfo.ipi_free_chunks = 0;
4696
4697	SCTP_OS_TIMER_INIT(&sctppcbinfo.addr_wq_timer.timer);
4698
4699	/* Init the TIMEWAIT list */
4700	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE_A; i++) {
4701		LIST_INIT(&sctppcbinfo.vtag_timewait[i]);
4702	}
4703
4704#if defined(SCTP_USE_THREAD_BASED_ITERATOR)
4705	sctppcbinfo.iterator_running = 0;
4706	sctp_startup_iterator();
4707#endif
4708
4709	/*
4710	 * INIT the default VRF which for BSD is the only one, other O/S's
4711	 * may have more. But initially they must start with one and then
4712	 * add the VRF's as addresses are added.
4713	 */
4714	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
4715
4716}
4717
4718
4719int
4720sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
4721    int iphlen, int offset, int limit, struct sctphdr *sh,
4722    struct sockaddr *altsa)
4723{
4724	/*
4725	 * grub through the INIT pulling addresses and loading them to the
4726	 * nets structure in the asoc. The from address in the mbuf should
4727	 * also be loaded (if it is not already). This routine can be called
4728	 * with either INIT or INIT-ACK's as long as the m points to the IP
4729	 * packet and the offset points to the beginning of the parameters.
4730	 */
4731	struct sctp_inpcb *inp, *l_inp;
4732	struct sctp_nets *net, *net_tmp;
4733	struct ip *iph;
4734	struct sctp_paramhdr *phdr, parm_buf;
4735	struct sctp_tcb *stcb_tmp;
4736	uint16_t ptype, plen;
4737	struct sockaddr *sa;
4738	struct sockaddr_storage dest_store;
4739	struct sockaddr *local_sa = (struct sockaddr *)&dest_store;
4740	struct sockaddr_in sin;
4741	struct sockaddr_in6 sin6;
4742	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
4743	struct sctp_auth_random *p_random = NULL;
4744	uint16_t random_len = 0;
4745	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
4746	struct sctp_auth_hmac_algo *hmacs = NULL;
4747	uint16_t hmacs_len = 0;
4748	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
4749	struct sctp_auth_chunk_list *chunks = NULL;
4750	uint16_t num_chunks = 0;
4751	sctp_key_t *new_key;
4752	uint32_t keylen;
4753	int got_random = 0, got_hmacs = 0, got_chklist = 0;
4754
4755	/* First get the destination address setup too. */
4756	memset(&sin, 0, sizeof(sin));
4757	memset(&sin6, 0, sizeof(sin6));
4758
4759	sin.sin_family = AF_INET;
4760	sin.sin_len = sizeof(sin);
4761	sin.sin_port = stcb->rport;
4762
4763	sin6.sin6_family = AF_INET6;
4764	sin6.sin6_len = sizeof(struct sockaddr_in6);
4765	sin6.sin6_port = stcb->rport;
4766	if (altsa == NULL) {
4767		iph = mtod(m, struct ip *);
4768		if (iph->ip_v == IPVERSION) {
4769			/* its IPv4 */
4770			struct sockaddr_in *sin_2;
4771
4772			sin_2 = (struct sockaddr_in *)(local_sa);
4773			memset(sin_2, 0, sizeof(sin));
4774			sin_2->sin_family = AF_INET;
4775			sin_2->sin_len = sizeof(sin);
4776			sin_2->sin_port = sh->dest_port;
4777			sin_2->sin_addr.s_addr = iph->ip_dst.s_addr;
4778			sin.sin_addr = iph->ip_src;
4779			sa = (struct sockaddr *)&sin;
4780		} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
4781			/* its IPv6 */
4782			struct ip6_hdr *ip6;
4783			struct sockaddr_in6 *sin6_2;
4784
4785			ip6 = mtod(m, struct ip6_hdr *);
4786			sin6_2 = (struct sockaddr_in6 *)(local_sa);
4787			memset(sin6_2, 0, sizeof(sin6));
4788			sin6_2->sin6_family = AF_INET6;
4789			sin6_2->sin6_len = sizeof(struct sockaddr_in6);
4790			sin6_2->sin6_port = sh->dest_port;
4791			sin6.sin6_addr = ip6->ip6_src;
4792			sa = (struct sockaddr *)&sin6;
4793		} else {
4794			sa = NULL;
4795		}
4796	} else {
4797		/*
4798		 * For cookies we use the src address NOT from the packet
4799		 * but from the original INIT
4800		 */
4801		sa = altsa;
4802	}
4803	/* Turn off ECN until we get through all params */
4804	stcb->asoc.ecn_allowed = 0;
4805	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4806		/* mark all addresses that we have currently on the list */
4807		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
4808	}
4809	/* does the source address already exist? if so skip it */
4810	l_inp = inp = stcb->sctp_ep;
4811
4812	atomic_add_int(&stcb->asoc.refcnt, 1);
4813	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb);
4814	atomic_add_int(&stcb->asoc.refcnt, -1);
4815
4816	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
4817		/* we must add the source address */
4818		/* no scope set here since we have a tcb already. */
4819		if ((sa->sa_family == AF_INET) &&
4820		    (stcb->asoc.ipv4_addr_legal)) {
4821			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
4822				return (-1);
4823			}
4824		} else if ((sa->sa_family == AF_INET6) &&
4825		    (stcb->asoc.ipv6_addr_legal)) {
4826			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
4827				return (-2);
4828			}
4829		}
4830	} else {
4831		if (net_tmp != NULL && stcb_tmp == stcb) {
4832			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
4833		} else if (stcb_tmp != stcb) {
4834			/* It belongs to another association? */
4835			if (stcb_tmp)
4836				SCTP_TCB_UNLOCK(stcb_tmp);
4837			return (-3);
4838		}
4839	}
4840	if (stcb->asoc.state == 0) {
4841		/* the assoc was freed? */
4842		return (-4);
4843	}
4844	/* now we must go through each of the params. */
4845	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
4846	while (phdr) {
4847		ptype = ntohs(phdr->param_type);
4848		plen = ntohs(phdr->param_length);
4849		/*
4850		 * printf("ptype => %0x, plen => %d\n", (uint32_t)ptype,
4851		 * (int)plen);
4852		 */
4853		if (offset + plen > limit) {
4854			break;
4855		}
4856		if (plen == 0) {
4857			break;
4858		}
4859		if (ptype == SCTP_IPV4_ADDRESS) {
4860			if (stcb->asoc.ipv4_addr_legal) {
4861				struct sctp_ipv4addr_param *p4, p4_buf;
4862
4863				/* ok get the v4 address and check/add */
4864				phdr = sctp_get_next_param(m, offset,
4865				    (struct sctp_paramhdr *)&p4_buf, sizeof(p4_buf));
4866				if (plen != sizeof(struct sctp_ipv4addr_param) ||
4867				    phdr == NULL) {
4868					return (-5);
4869				}
4870				p4 = (struct sctp_ipv4addr_param *)phdr;
4871				sin.sin_addr.s_addr = p4->addr;
4872				if (IN_MULTICAST(sin.sin_addr.s_addr)) {
4873					/* Skip multi-cast addresses */
4874					goto next_param;
4875				}
4876				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
4877				    (sin.sin_addr.s_addr == INADDR_ANY)) {
4878					goto next_param;
4879				}
4880				sa = (struct sockaddr *)&sin;
4881				inp = stcb->sctp_ep;
4882				atomic_add_int(&stcb->asoc.refcnt, 1);
4883				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
4884				    local_sa, stcb);
4885				atomic_add_int(&stcb->asoc.refcnt, -1);
4886
4887				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
4888				    inp == NULL) {
4889					/* we must add the source address */
4890					/*
4891					 * no scope set since we have a tcb
4892					 * already
4893					 */
4894
4895					/*
4896					 * we must validate the state again
4897					 * here
4898					 */
4899					if (stcb->asoc.state == 0) {
4900						/* the assoc was freed? */
4901						return (-7);
4902					}
4903					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
4904						return (-8);
4905					}
4906				} else if (stcb_tmp == stcb) {
4907					if (stcb->asoc.state == 0) {
4908						/* the assoc was freed? */
4909						return (-10);
4910					}
4911					if (net != NULL) {
4912						/* clear flag */
4913						net->dest_state &=
4914						    ~SCTP_ADDR_NOT_IN_ASSOC;
4915					}
4916				} else {
4917					/*
4918					 * strange, address is in another
4919					 * assoc? straighten out locks.
4920					 */
4921					if (stcb->asoc.state == 0) {
4922						/* the assoc was freed? */
4923						return (-12);
4924					}
4925					return (-13);
4926				}
4927			}
4928		} else if (ptype == SCTP_IPV6_ADDRESS) {
4929			if (stcb->asoc.ipv6_addr_legal) {
4930				/* ok get the v6 address and check/add */
4931				struct sctp_ipv6addr_param *p6, p6_buf;
4932
4933				phdr = sctp_get_next_param(m, offset,
4934				    (struct sctp_paramhdr *)&p6_buf, sizeof(p6_buf));
4935				if (plen != sizeof(struct sctp_ipv6addr_param) ||
4936				    phdr == NULL) {
4937					return (-14);
4938				}
4939				p6 = (struct sctp_ipv6addr_param *)phdr;
4940				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
4941				    sizeof(p6->addr));
4942				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
4943					/* Skip multi-cast addresses */
4944					goto next_param;
4945				}
4946				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
4947					/*
4948					 * Link local make no sense without
4949					 * scope
4950					 */
4951					goto next_param;
4952				}
4953				sa = (struct sockaddr *)&sin6;
4954				inp = stcb->sctp_ep;
4955				atomic_add_int(&stcb->asoc.refcnt, 1);
4956				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
4957				    local_sa, stcb);
4958				atomic_add_int(&stcb->asoc.refcnt, -1);
4959				if (stcb_tmp == NULL && (inp == stcb->sctp_ep ||
4960				    inp == NULL)) {
4961					/*
4962					 * we must validate the state again
4963					 * here
4964					 */
4965					if (stcb->asoc.state == 0) {
4966						/* the assoc was freed? */
4967						return (-16);
4968					}
4969					/*
4970					 * we must add the address, no scope
4971					 * set
4972					 */
4973					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
4974						return (-17);
4975					}
4976				} else if (stcb_tmp == stcb) {
4977					/*
4978					 * we must validate the state again
4979					 * here
4980					 */
4981					if (stcb->asoc.state == 0) {
4982						/* the assoc was freed? */
4983						return (-19);
4984					}
4985					if (net != NULL) {
4986						/* clear flag */
4987						net->dest_state &=
4988						    ~SCTP_ADDR_NOT_IN_ASSOC;
4989					}
4990				} else {
4991					/*
4992					 * strange, address is in another
4993					 * assoc? straighten out locks.
4994					 */
4995					if (stcb->asoc.state == 0) {
4996						/* the assoc was freed? */
4997						return (-21);
4998					}
4999					return (-22);
5000				}
5001			}
5002		} else if (ptype == SCTP_ECN_CAPABLE) {
5003			stcb->asoc.ecn_allowed = 1;
5004		} else if (ptype == SCTP_ULP_ADAPTATION) {
5005			if (stcb->asoc.state != SCTP_STATE_OPEN) {
5006				struct sctp_adaptation_layer_indication ai,
5007				                                *aip;
5008
5009				phdr = sctp_get_next_param(m, offset,
5010				    (struct sctp_paramhdr *)&ai, sizeof(ai));
5011				aip = (struct sctp_adaptation_layer_indication *)phdr;
5012				if (aip) {
5013					sctp_ulp_notify(SCTP_NOTIFY_ADAPTATION_INDICATION,
5014					    stcb, ntohl(aip->indication), NULL);
5015				}
5016			}
5017		} else if (ptype == SCTP_SET_PRIM_ADDR) {
5018			struct sctp_asconf_addr_param lstore, *fee;
5019			struct sctp_asconf_addrv4_param *fii;
5020			int lptype;
5021			struct sockaddr *lsa = NULL;
5022
5023			stcb->asoc.peer_supports_asconf = 1;
5024			if (plen > sizeof(lstore)) {
5025				return (-23);
5026			}
5027			phdr = sctp_get_next_param(m, offset,
5028			    (struct sctp_paramhdr *)&lstore, min(plen, sizeof(lstore)));
5029			if (phdr == NULL) {
5030				return (-24);
5031			}
5032			fee = (struct sctp_asconf_addr_param *)phdr;
5033			lptype = ntohs(fee->addrp.ph.param_type);
5034			if (lptype == SCTP_IPV4_ADDRESS) {
5035				if (plen !=
5036				    sizeof(struct sctp_asconf_addrv4_param)) {
5037					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
5038					    (int)sizeof(struct sctp_asconf_addrv4_param),
5039					    plen);
5040				} else {
5041					fii = (struct sctp_asconf_addrv4_param *)fee;
5042					sin.sin_addr.s_addr = fii->addrp.addr;
5043					lsa = (struct sockaddr *)&sin;
5044				}
5045			} else if (lptype == SCTP_IPV6_ADDRESS) {
5046				if (plen !=
5047				    sizeof(struct sctp_asconf_addr_param)) {
5048					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
5049					    (int)sizeof(struct sctp_asconf_addr_param),
5050					    plen);
5051				} else {
5052					memcpy(sin6.sin6_addr.s6_addr,
5053					    fee->addrp.addr,
5054					    sizeof(fee->addrp.addr));
5055					lsa = (struct sockaddr *)&sin6;
5056				}
5057			}
5058			if (lsa) {
5059				(void)sctp_set_primary_addr(stcb, sa, NULL);
5060			}
5061		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
5062			/* Peer supports pr-sctp */
5063			stcb->asoc.peer_supports_prsctp = 1;
5064		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
5065			/* A supported extension chunk */
5066			struct sctp_supported_chunk_types_param *pr_supported;
5067			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
5068			int num_ent, i;
5069
5070			phdr = sctp_get_next_param(m, offset,
5071			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
5072			if (phdr == NULL) {
5073				return (-25);
5074			}
5075			stcb->asoc.peer_supports_asconf = 0;
5076			stcb->asoc.peer_supports_prsctp = 0;
5077			stcb->asoc.peer_supports_pktdrop = 0;
5078			stcb->asoc.peer_supports_strreset = 0;
5079			stcb->asoc.peer_supports_auth = 0;
5080			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
5081			num_ent = plen - sizeof(struct sctp_paramhdr);
5082			for (i = 0; i < num_ent; i++) {
5083				switch (pr_supported->chunk_types[i]) {
5084				case SCTP_ASCONF:
5085				case SCTP_ASCONF_ACK:
5086					stcb->asoc.peer_supports_asconf = 1;
5087					break;
5088				case SCTP_FORWARD_CUM_TSN:
5089					stcb->asoc.peer_supports_prsctp = 1;
5090					break;
5091				case SCTP_PACKET_DROPPED:
5092					stcb->asoc.peer_supports_pktdrop = 1;
5093					break;
5094				case SCTP_STREAM_RESET:
5095					stcb->asoc.peer_supports_strreset = 1;
5096					break;
5097				case SCTP_AUTHENTICATION:
5098					stcb->asoc.peer_supports_auth = 1;
5099					break;
5100				default:
5101					/* one I have not learned yet */
5102					break;
5103
5104				}
5105			}
5106		} else if (ptype == SCTP_ECN_NONCE_SUPPORTED) {
5107			/* Peer supports ECN-nonce */
5108			stcb->asoc.peer_supports_ecn_nonce = 1;
5109			stcb->asoc.ecn_nonce_allowed = 1;
5110		} else if (ptype == SCTP_RANDOM) {
5111			if (plen > sizeof(random_store))
5112				break;
5113			if (got_random) {
5114				/* already processed a RANDOM */
5115				goto next_param;
5116			}
5117			phdr = sctp_get_next_param(m, offset,
5118			    (struct sctp_paramhdr *)random_store,
5119			    min(sizeof(random_store), plen));
5120			if (phdr == NULL)
5121				return (-26);
5122			p_random = (struct sctp_auth_random *)phdr;
5123			random_len = plen - sizeof(*p_random);
5124			/* enforce the random length */
5125			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
5126				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
5127				return (-27);
5128			}
5129			got_random = 1;
5130		} else if (ptype == SCTP_HMAC_LIST) {
5131			int num_hmacs;
5132			int i;
5133
5134			if (plen > sizeof(hmacs_store))
5135				break;
5136			if (got_hmacs) {
5137				/* already processed a HMAC list */
5138				goto next_param;
5139			}
5140			phdr = sctp_get_next_param(m, offset,
5141			    (struct sctp_paramhdr *)hmacs_store,
5142			    min(plen, sizeof(hmacs_store)));
5143			if (phdr == NULL)
5144				return (-28);
5145			hmacs = (struct sctp_auth_hmac_algo *)phdr;
5146			hmacs_len = plen - sizeof(*hmacs);
5147			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
5148			/* validate the hmac list */
5149			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
5150				return (-29);
5151			}
5152			if (stcb->asoc.peer_hmacs != NULL)
5153				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
5154			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
5155			if (stcb->asoc.peer_hmacs != NULL) {
5156				for (i = 0; i < num_hmacs; i++) {
5157					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
5158					    ntohs(hmacs->hmac_ids[i]));
5159				}
5160			}
5161			got_hmacs = 1;
5162		} else if (ptype == SCTP_CHUNK_LIST) {
5163			int i;
5164
5165			if (plen > sizeof(chunks_store))
5166				break;
5167			if (got_chklist) {
5168				/* already processed a Chunks list */
5169				goto next_param;
5170			}
5171			phdr = sctp_get_next_param(m, offset,
5172			    (struct sctp_paramhdr *)chunks_store,
5173			    min(plen, sizeof(chunks_store)));
5174			if (phdr == NULL)
5175				return (-30);
5176			chunks = (struct sctp_auth_chunk_list *)phdr;
5177			num_chunks = plen - sizeof(*chunks);
5178			if (stcb->asoc.peer_auth_chunks != NULL)
5179				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
5180			else
5181				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
5182			for (i = 0; i < num_chunks; i++) {
5183				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
5184				    stcb->asoc.peer_auth_chunks);
5185			}
5186			got_chklist = 1;
5187		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
5188			    (ptype == SCTP_STATE_COOKIE) ||
5189			    (ptype == SCTP_UNRECOG_PARAM) ||
5190			    (ptype == SCTP_COOKIE_PRESERVE) ||
5191			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
5192			    (ptype == SCTP_ADD_IP_ADDRESS) ||
5193			    (ptype == SCTP_DEL_IP_ADDRESS) ||
5194			    (ptype == SCTP_ERROR_CAUSE_IND) ||
5195		    (ptype == SCTP_SUCCESS_REPORT)) {
5196			 /* don't care */ ;
5197		} else {
5198			if ((ptype & 0x8000) == 0x0000) {
5199				/*
5200				 * must stop processing the rest of the
5201				 * param's. Any report bits were handled
5202				 * with the call to
5203				 * sctp_arethere_unrecognized_parameters()
5204				 * when the INIT or INIT-ACK was first seen.
5205				 */
5206				break;
5207			}
5208		}
5209next_param:
5210		offset += SCTP_SIZE32(plen);
5211		if (offset >= limit) {
5212			break;
5213		}
5214		phdr = sctp_get_next_param(m, offset, &parm_buf,
5215		    sizeof(parm_buf));
5216	}
5217	/* Now check to see if we need to purge any addresses */
5218	for (net = TAILQ_FIRST(&stcb->asoc.nets); net != NULL; net = net_tmp) {
5219		net_tmp = TAILQ_NEXT(net, sctp_next);
5220		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
5221		    SCTP_ADDR_NOT_IN_ASSOC) {
5222			/* This address has been removed from the asoc */
5223			/* remove and free it */
5224			stcb->asoc.numnets--;
5225			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
5226			sctp_free_remote_addr(net);
5227			if (net == stcb->asoc.primary_destination) {
5228				stcb->asoc.primary_destination = NULL;
5229				sctp_select_primary_destination(stcb);
5230			}
5231		}
5232	}
5233	/* validate authentication required parameters */
5234	if (got_random && got_hmacs) {
5235		stcb->asoc.peer_supports_auth = 1;
5236	} else {
5237		stcb->asoc.peer_supports_auth = 0;
5238	}
5239	if (!stcb->asoc.peer_supports_auth && got_chklist) {
5240		/* peer does not support auth but sent a chunks list? */
5241		return (-31);
5242	}
5243	if (!sctp_asconf_auth_nochk && stcb->asoc.peer_supports_asconf &&
5244	    !stcb->asoc.peer_supports_auth) {
5245		/* peer supports asconf but not auth? */
5246		return (-32);
5247	}
5248	/* concatenate the full random key */
5249#ifdef SCTP_AUTH_DRAFT_04
5250	keylen = random_len;
5251	new_key = sctp_alloc_key(keylen);
5252	if (new_key != NULL) {
5253		/* copy in the RANDOM */
5254		if (p_random != NULL)
5255			bcopy(p_random->random_data, new_key->key, random_len);
5256	}
5257#else
5258	keylen = sizeof(*p_random) + random_len + sizeof(*chunks) + num_chunks +
5259	    sizeof(*hmacs) + hmacs_len;
5260	new_key = sctp_alloc_key(keylen);
5261	if (new_key != NULL) {
5262		/* copy in the RANDOM */
5263		if (p_random != NULL) {
5264			keylen = sizeof(*p_random) + random_len;
5265			bcopy(p_random, new_key->key, keylen);
5266		}
5267		/* append in the AUTH chunks */
5268		if (chunks != NULL) {
5269			bcopy(chunks, new_key->key + keylen,
5270			    sizeof(*chunks) + num_chunks);
5271			keylen += sizeof(*chunks) + num_chunks;
5272		}
5273		/* append in the HMACs */
5274		if (hmacs != NULL) {
5275			bcopy(hmacs, new_key->key + keylen,
5276			    sizeof(*hmacs) + hmacs_len);
5277		}
5278	}
5279#endif
5280	else {
5281		/* failed to get memory for the key */
5282		return (-33);
5283	}
5284	if (stcb->asoc.authinfo.peer_random != NULL)
5285		sctp_free_key(stcb->asoc.authinfo.peer_random);
5286	stcb->asoc.authinfo.peer_random = new_key;
5287#ifdef SCTP_AUTH_DRAFT_04
5288	/* don't include the chunks and hmacs for draft -04 */
5289	stcb->asoc.authinfo.peer_random->keylen = random_len;
5290#endif
5291	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
5292	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
5293
5294	return (0);
5295}
5296
5297int
5298sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
5299    struct sctp_nets *net)
5300{
5301	/* make sure the requested primary address exists in the assoc */
5302	if (net == NULL && sa)
5303		net = sctp_findnet(stcb, sa);
5304
5305	if (net == NULL) {
5306		/* didn't find the requested primary address! */
5307		return (-1);
5308	} else {
5309		/* set the primary address */
5310		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
5311			/* Must be confirmed, so queue to set */
5312			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
5313			return (0);
5314		}
5315		stcb->asoc.primary_destination = net;
5316		net->dest_state &= ~SCTP_ADDR_WAS_PRIMARY;
5317		net = TAILQ_FIRST(&stcb->asoc.nets);
5318		if (net != stcb->asoc.primary_destination) {
5319			/*
5320			 * first one on the list is NOT the primary
5321			 * sctp_cmpaddr() is much more efficent if the
5322			 * primary is the first on the list, make it so.
5323			 */
5324			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
5325			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
5326		}
5327		return (0);
5328	}
5329}
5330
5331
5332int
5333sctp_is_vtag_good(struct sctp_inpcb *inp, uint32_t tag, struct timeval *now)
5334{
5335	/*
5336	 * This function serves two purposes. It will see if a TAG can be
5337	 * re-used and return 1 for yes it is ok and 0 for don't use that
5338	 * tag. A secondary function it will do is purge out old tags that
5339	 * can be removed.
5340	 */
5341	struct sctpasochead *head;
5342	struct sctpvtaghead *chain;
5343	struct sctp_tagblock *twait_block;
5344	struct sctp_tcb *stcb;
5345	int i;
5346
5347	SCTP_INP_INFO_WLOCK();
5348	chain = &sctppcbinfo.vtag_timewait[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
5349	/* First is the vtag in use ? */
5350
5351	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(tag,
5352	    sctppcbinfo.hashasocmark)];
5353	if (head == NULL) {
5354		goto check_restart;
5355	}
5356	LIST_FOREACH(stcb, head, sctp_asocs) {
5357
5358		if (stcb->asoc.my_vtag == tag) {
5359			/*
5360			 * We should remove this if and return 0 always if
5361			 * we want vtags unique across all endpoints. For
5362			 * now within a endpoint is ok.
5363			 */
5364			if (inp == stcb->sctp_ep) {
5365				/* bad tag, in use */
5366				SCTP_INP_INFO_WUNLOCK();
5367				return (0);
5368			}
5369		}
5370	}
5371check_restart:
5372	/* Now lets check the restart hash */
5373	head = &sctppcbinfo.sctp_restarthash[SCTP_PCBHASH_ASOC(tag,
5374	    sctppcbinfo.hashrestartmark)];
5375	if (head == NULL) {
5376		goto check_time_wait;
5377	}
5378	LIST_FOREACH(stcb, head, sctp_tcbrestarhash) {
5379		if (stcb->asoc.assoc_id == tag) {
5380			/* candidate */
5381			if (inp == stcb->sctp_ep) {
5382				/* bad tag, in use */
5383				SCTP_INP_INFO_WUNLOCK();
5384				return (0);
5385			}
5386		}
5387	}
5388check_time_wait:
5389	/* Now what about timed wait ? */
5390	if (!SCTP_LIST_EMPTY(chain)) {
5391		/*
5392		 * Block(s) are present, lets see if we have this tag in the
5393		 * list
5394		 */
5395		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5396			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
5397				if (twait_block->vtag_block[i].v_tag == 0) {
5398					/* not used */
5399					continue;
5400				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire >
5401				    now->tv_sec) {
5402					/* Audit expires this guy */
5403					twait_block->vtag_block[i].tv_sec_at_expire = 0;
5404					twait_block->vtag_block[i].v_tag = 0;
5405				} else if (twait_block->vtag_block[i].v_tag ==
5406				    tag) {
5407					/* Bad tag, sorry :< */
5408					SCTP_INP_INFO_WUNLOCK();
5409					return (0);
5410				}
5411			}
5412		}
5413	}
5414	/* Not found, ok to use the tag */
5415	SCTP_INP_INFO_WUNLOCK();
5416	return (1);
5417}
5418
5419
5420static sctp_assoc_t reneged_asoc_ids[256];
5421static uint8_t reneged_at = 0;
5422
5423
5424static void
5425sctp_drain_mbufs(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
5426{
5427	/*
5428	 * We must hunt this association for MBUF's past the cumack (i.e.
5429	 * out of order data that we can renege on).
5430	 */
5431	struct sctp_association *asoc;
5432	struct sctp_tmit_chunk *chk, *nchk;
5433	uint32_t cumulative_tsn_p1, tsn;
5434	struct sctp_queued_to_read *ctl, *nctl;
5435	int cnt, strmat, gap;
5436
5437	/* We look for anything larger than the cum-ack + 1 */
5438
5439	SCTP_STAT_INCR(sctps_protocol_drain_calls);
5440	if (sctp_do_drain == 0) {
5441		return;
5442	}
5443	asoc = &stcb->asoc;
5444	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
5445		/* none we can reneg on. */
5446		return;
5447	}
5448	SCTP_STAT_INCR(sctps_protocol_drains_done);
5449	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
5450	cnt = 0;
5451	/* First look in the re-assembly queue */
5452	chk = TAILQ_FIRST(&asoc->reasmqueue);
5453	while (chk) {
5454		/* Get the next one */
5455		nchk = TAILQ_NEXT(chk, sctp_next);
5456		if (compare_with_wrap(chk->rec.data.TSN_seq,
5457		    cumulative_tsn_p1, MAX_TSN)) {
5458			/* Yep it is above cum-ack */
5459			cnt++;
5460			tsn = chk->rec.data.TSN_seq;
5461			if (tsn >= asoc->mapping_array_base_tsn) {
5462				gap = tsn - asoc->mapping_array_base_tsn;
5463			} else {
5464				gap = (MAX_TSN - asoc->mapping_array_base_tsn) +
5465				    tsn + 1;
5466			}
5467			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
5468			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
5469			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
5470			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5471			if (chk->data) {
5472				sctp_m_freem(chk->data);
5473				chk->data = NULL;
5474			}
5475			sctp_free_a_chunk(stcb, chk);
5476		}
5477		chk = nchk;
5478	}
5479	/* Ok that was fun, now we will drain all the inbound streams? */
5480	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
5481		ctl = TAILQ_FIRST(&asoc->strmin[strmat].inqueue);
5482		while (ctl) {
5483			nctl = TAILQ_NEXT(ctl, next);
5484			if (compare_with_wrap(ctl->sinfo_tsn,
5485			    cumulative_tsn_p1, MAX_TSN)) {
5486				/* Yep it is above cum-ack */
5487				cnt++;
5488				tsn = ctl->sinfo_tsn;
5489				if (tsn >= asoc->mapping_array_base_tsn) {
5490					gap = tsn -
5491					    asoc->mapping_array_base_tsn;
5492				} else {
5493					gap = (MAX_TSN -
5494					    asoc->mapping_array_base_tsn) +
5495					    tsn + 1;
5496				}
5497				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
5498				sctp_ucount_decr(asoc->cnt_on_all_streams);
5499
5500				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array,
5501				    gap);
5502				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue,
5503				    ctl, next);
5504				if (ctl->data) {
5505					sctp_m_freem(ctl->data);
5506					ctl->data = NULL;
5507				}
5508				sctp_free_remote_addr(ctl->whoFrom);
5509				SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, ctl);
5510				SCTP_DECR_READQ_COUNT();
5511			}
5512			ctl = nctl;
5513		}
5514	}
5515	/*
5516	 * Question, should we go through the delivery queue? The only
5517	 * reason things are on here is the app not reading OR a p-d-api up.
5518	 * An attacker COULD send enough in to initiate the PD-API and then
5519	 * send a bunch of stuff to other streams... these would wind up on
5520	 * the delivery queue.. and then we would not get to them. But in
5521	 * order to do this I then have to back-track and un-deliver
5522	 * sequence numbers in streams.. el-yucko. I think for now we will
5523	 * NOT look at the delivery queue and leave it to be something to
5524	 * consider later. An alternative would be to abort the P-D-API with
5525	 * a notification and then deliver the data.... Or another method
5526	 * might be to keep track of how many times the situation occurs and
5527	 * if we see a possible attack underway just abort the association.
5528	 */
5529#ifdef SCTP_DEBUG
5530	if (cnt) {
5531		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
5532	}
5533#endif
5534	if (cnt) {
5535		/*
5536		 * Now do we need to find a new
5537		 * asoc->highest_tsn_inside_map?
5538		 */
5539		if (asoc->highest_tsn_inside_map >= asoc->mapping_array_base_tsn) {
5540			gap = asoc->highest_tsn_inside_map - asoc->mapping_array_base_tsn;
5541		} else {
5542			gap = (MAX_TSN - asoc->mapping_array_base_tsn) +
5543			    asoc->highest_tsn_inside_map + 1;
5544		}
5545		if (gap >= (asoc->mapping_array_size << 3)) {
5546			/*
5547			 * Something bad happened or cum-ack and high were
5548			 * behind the base, but if so earlier checks should
5549			 * have found NO data... wierd... we will start at
5550			 * end of mapping array.
5551			 */
5552			SCTP_PRINTF("Gap was larger than array?? %d set to max:%d maparraymax:%x\n",
5553			    (int)gap,
5554			    (int)(asoc->mapping_array_size << 3),
5555			    (int)asoc->highest_tsn_inside_map);
5556			gap = asoc->mapping_array_size << 3;
5557		}
5558		while (gap > 0) {
5559			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
5560				/* found the new highest */
5561				asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn + gap;
5562				break;
5563			}
5564			gap--;
5565		}
5566		if (gap == 0) {
5567			/* Nothing left in map */
5568			memset(asoc->mapping_array, 0, asoc->mapping_array_size);
5569			asoc->mapping_array_base_tsn = asoc->cumulative_tsn + 1;
5570			asoc->highest_tsn_inside_map = asoc->cumulative_tsn;
5571		}
5572		asoc->last_revoke_count = cnt;
5573		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
5574		sctp_send_sack(stcb);
5575		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN);
5576		reneged_asoc_ids[reneged_at] = sctp_get_associd(stcb);
5577		reneged_at++;
5578	}
5579	/*
5580	 * Another issue, in un-setting the TSN's in the mapping array we
5581	 * DID NOT adjust the higest_tsn marker.  This will cause one of two
5582	 * things to occur. It may cause us to do extra work in checking for
5583	 * our mapping array movement. More importantly it may cause us to
5584	 * SACK every datagram. This may not be a bad thing though since we
5585	 * will recover once we get our cum-ack above and all this stuff we
5586	 * dumped recovered.
5587	 */
5588}
5589
5590void
5591sctp_drain()
5592{
5593	/*
5594	 * We must walk the PCB lists for ALL associations here. The system
5595	 * is LOW on MBUF's and needs help. This is where reneging will
5596	 * occur. We really hope this does NOT happen!
5597	 */
5598	struct sctp_inpcb *inp;
5599	struct sctp_tcb *stcb;
5600
5601	SCTP_INP_INFO_RLOCK();
5602	LIST_FOREACH(inp, &sctppcbinfo.listhead, sctp_list) {
5603		/* For each endpoint */
5604		SCTP_INP_RLOCK(inp);
5605		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5606			/* For each association */
5607			SCTP_TCB_LOCK(stcb);
5608			sctp_drain_mbufs(inp, stcb);
5609			SCTP_TCB_UNLOCK(stcb);
5610		}
5611		SCTP_INP_RUNLOCK(inp);
5612	}
5613	SCTP_INP_INFO_RUNLOCK();
5614}
5615
5616/*
5617 * start a new iterator
5618 * iterates through all endpoints and associations based on the pcb_state
5619 * flags and asoc_state.  "af" (mandatory) is executed for all matching
5620 * assocs and "ef" (optional) is executed when the iterator completes.
5621 * "inpf" (optional) is executed for each new endpoint as it is being
5622 * iterated through. inpe (optional) is called when the inp completes
5623 * its way through all the stcbs.
5624 */
5625int
5626sctp_initiate_iterator(inp_func inpf,
5627    asoc_func af,
5628    inp_func inpe,
5629    uint32_t pcb_state,
5630    uint32_t pcb_features,
5631    uint32_t asoc_state,
5632    void *argp,
5633    uint32_t argi,
5634    end_func ef,
5635    struct sctp_inpcb *s_inp,
5636    uint8_t chunk_output_off)
5637{
5638	struct sctp_iterator *it = NULL;
5639
5640	if (af == NULL) {
5641		return (-1);
5642	}
5643	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
5644	    SCTP_M_ITER);
5645	if (it == NULL) {
5646		return (ENOMEM);
5647	}
5648	memset(it, 0, sizeof(*it));
5649	it->function_assoc = af;
5650	it->function_inp = inpf;
5651	if (inpf)
5652		it->done_current_ep = 0;
5653	else
5654		it->done_current_ep = 1;
5655	it->function_atend = ef;
5656	it->pointer = argp;
5657	it->val = argi;
5658	it->pcb_flags = pcb_state;
5659	it->pcb_features = pcb_features;
5660	it->asoc_state = asoc_state;
5661	it->function_inp_end = inpe;
5662	it->no_chunk_output = chunk_output_off;
5663	if (s_inp) {
5664		it->inp = s_inp;
5665		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
5666	} else {
5667		SCTP_INP_INFO_RLOCK();
5668		it->inp = LIST_FIRST(&sctppcbinfo.listhead);
5669
5670		SCTP_INP_INFO_RUNLOCK();
5671		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
5672
5673	}
5674	SCTP_IPI_ITERATOR_WQ_LOCK();
5675	if (it->inp) {
5676		SCTP_INP_INCR_REF(it->inp);
5677	}
5678	TAILQ_INSERT_TAIL(&sctppcbinfo.iteratorhead, it, sctp_nxt_itr);
5679#if defined(SCTP_USE_THREAD_BASED_ITERATOR)
5680	if (sctppcbinfo.iterator_running == 0) {
5681		sctp_wakeup_iterator();
5682	}
5683	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5684#else
5685	if (it->inp)
5686		SCTP_INP_DECR_REF(it->inp);
5687	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5688	/* Init the timer */
5689	SCTP_OS_TIMER_INIT(&it->tmr.timer);
5690	/* add to the list of all iterators */
5691	sctp_timer_start(SCTP_TIMER_TYPE_ITERATOR, (struct sctp_inpcb *)it,
5692	    NULL, NULL);
5693#endif
5694	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
5695	return (0);
5696}
5697