sctp_pcb.c revision 171572
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 171572 2007-07-24 20:06:02Z 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 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				/* 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	if ((sctp_default_cc_module >= SCTP_CC_RFC2581) &&
1915	    (sctp_default_cc_module <= SCTP_CC_HTCP)) {
1916		m->sctp_default_cc_module = sctp_default_cc_module;
1917	} else {
1918		/* sysctl done with invalid value, set to 2581 */
1919		m->sctp_default_cc_module = SCTP_CC_RFC2581;
1920	}
1921	/* number of streams to pre-open on a association */
1922	m->pre_open_stream_count = sctp_nr_outgoing_streams_default;
1923
1924	/* Add adaptation cookie */
1925	m->adaptation_layer_indicator = 0x504C5253;
1926
1927	/* seed random number generator */
1928	m->random_counter = 1;
1929	m->store_at = SCTP_SIGNATURE_SIZE;
1930	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
1931	sctp_fill_random_store(m);
1932
1933	/* Minimum cookie size */
1934	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
1935	    sizeof(struct sctp_state_cookie);
1936	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
1937
1938	/* Setup the initial secret */
1939	(void)SCTP_GETTIME_TIMEVAL(&time);
1940	m->time_of_secret_change = time.tv_sec;
1941
1942	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
1943		m->secret_key[0][i] = sctp_select_initial_TSN(m);
1944	}
1945	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
1946
1947	/* How long is a cookie good for ? */
1948	m->def_cookie_life = MSEC_TO_TICKS(sctp_valid_cookie_life_default);
1949	/*
1950	 * Initialize authentication parameters
1951	 */
1952	m->local_hmacs = sctp_default_supported_hmaclist();
1953	m->local_auth_chunks = sctp_alloc_chunklist();
1954	sctp_auth_set_default_chunks(m->local_auth_chunks);
1955	LIST_INIT(&m->shared_keys);
1956	/* add default NULL key as key id 0 */
1957	null_key = sctp_alloc_sharedkey();
1958	sctp_insert_sharedkey(&m->shared_keys, null_key);
1959	SCTP_INP_WUNLOCK(inp);
1960#ifdef SCTP_LOG_CLOSING
1961	sctp_log_closing(inp, NULL, 12);
1962#endif
1963	return (error);
1964}
1965
1966
1967void
1968sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
1969    struct sctp_tcb *stcb)
1970{
1971	struct sctp_nets *net;
1972	uint16_t lport, rport;
1973	struct sctppcbhead *head;
1974	struct sctp_laddr *laddr, *oladdr;
1975
1976	SCTP_TCB_UNLOCK(stcb);
1977	SCTP_INP_INFO_WLOCK();
1978	SCTP_INP_WLOCK(old_inp);
1979	SCTP_INP_WLOCK(new_inp);
1980	SCTP_TCB_LOCK(stcb);
1981
1982	new_inp->sctp_ep.time_of_secret_change =
1983	    old_inp->sctp_ep.time_of_secret_change;
1984	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
1985	    sizeof(old_inp->sctp_ep.secret_key));
1986	new_inp->sctp_ep.current_secret_number =
1987	    old_inp->sctp_ep.current_secret_number;
1988	new_inp->sctp_ep.last_secret_number =
1989	    old_inp->sctp_ep.last_secret_number;
1990	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
1991
1992	/* make it so new data pours into the new socket */
1993	stcb->sctp_socket = new_inp->sctp_socket;
1994	stcb->sctp_ep = new_inp;
1995
1996	/* Copy the port across */
1997	lport = new_inp->sctp_lport = old_inp->sctp_lport;
1998	rport = stcb->rport;
1999	/* Pull the tcb from the old association */
2000	LIST_REMOVE(stcb, sctp_tcbhash);
2001	LIST_REMOVE(stcb, sctp_tcblist);
2002
2003	/* Now insert the new_inp into the TCP connected hash */
2004	head = &sctppcbinfo.sctp_tcpephash[SCTP_PCBHASH_ALLADDR((lport + rport),
2005	    sctppcbinfo.hashtcpmark)];
2006
2007	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2008	/* Its safe to access */
2009	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2010
2011	/* Now move the tcb into the endpoint list */
2012	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2013	/*
2014	 * Question, do we even need to worry about the ep-hash since we
2015	 * only have one connection? Probably not :> so lets get rid of it
2016	 * and not suck up any kernel memory in that.
2017	 */
2018
2019	/* Ok. Let's restart timer. */
2020	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2021		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2022		    stcb, net);
2023	}
2024
2025	SCTP_INP_INFO_WUNLOCK();
2026	if (new_inp->sctp_tcbhash != NULL) {
2027		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2028		new_inp->sctp_tcbhash = NULL;
2029	}
2030	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2031		/* Subset bound, so copy in the laddr list from the old_inp */
2032		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2033			laddr = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
2034			if (laddr == NULL) {
2035				/*
2036				 * Gak, what can we do? This assoc is really
2037				 * HOSED. We probably should send an abort
2038				 * here.
2039				 */
2040				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2041				continue;
2042			}
2043			SCTP_INCR_LADDR_COUNT();
2044			bzero(laddr, sizeof(*laddr));
2045			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2046			laddr->ifa = oladdr->ifa;
2047			atomic_add_int(&laddr->ifa->refcount, 1);
2048			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2049			    sctp_nxt_addr);
2050			new_inp->laddr_count++;
2051		}
2052	}
2053	/*
2054	 * Now any running timers need to be adjusted since we really don't
2055	 * care if they are running or not just blast in the new_inp into
2056	 * all of them.
2057	 */
2058
2059	stcb->asoc.hb_timer.ep = (void *)new_inp;
2060	stcb->asoc.dack_timer.ep = (void *)new_inp;
2061	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2062	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2063	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2064	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2065	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2066	/* now what about the nets? */
2067	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2068		net->pmtu_timer.ep = (void *)new_inp;
2069		net->rxt_timer.ep = (void *)new_inp;
2070		net->fr_timer.ep = (void *)new_inp;
2071	}
2072	SCTP_INP_WUNLOCK(new_inp);
2073	SCTP_INP_WUNLOCK(old_inp);
2074}
2075
2076static int
2077sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
2078{
2079	struct sctppcbhead *head;
2080	struct sctp_inpcb *t_inp;
2081	int fnd;
2082
2083	head = &sctppcbinfo.sctp_ephash[SCTP_PCBHASH_ALLADDR(lport,
2084	    sctppcbinfo.hashmark)];
2085	LIST_FOREACH(t_inp, head, sctp_hash) {
2086		if (t_inp->sctp_lport != lport) {
2087			continue;
2088		}
2089		/* is it in the VRF in question */
2090		fnd = 0;
2091		if (t_inp->def_vrf_id == vrf_id)
2092			fnd = 1;
2093		if (!fnd)
2094			continue;
2095
2096		/* This one is in use. */
2097		/* check the v6/v4 binding issue */
2098		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
2099		    SCTP_IPV6_V6ONLY(t_inp)) {
2100			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
2101				/* collision in V6 space */
2102				return (1);
2103			} else {
2104				/* inp is BOUND_V4 no conflict */
2105				continue;
2106			}
2107		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
2108			/* t_inp is bound v4 and v6, conflict always */
2109			return (1);
2110		} else {
2111			/* t_inp is bound only V4 */
2112			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
2113			    SCTP_IPV6_V6ONLY(inp)) {
2114				/* no conflict */
2115				continue;
2116			}
2117			/* else fall through to conflict */
2118		}
2119		return (1);
2120	}
2121	return (0);
2122}
2123
2124
2125
2126/* sctp_ifap is used to bypass normal local address validation checks */
2127int
2128sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2129    struct sctp_ifa *sctp_ifap, struct thread *p)
2130{
2131	/* bind a ep to a socket address */
2132	struct sctppcbhead *head;
2133	struct sctp_inpcb *inp, *inp_tmp;
2134	struct inpcb *ip_inp;
2135	int bindall;
2136	int prison = 0;
2137	uint16_t lport;
2138	int error;
2139	uint32_t vrf_id;
2140
2141	lport = 0;
2142	error = 0;
2143	bindall = 1;
2144	inp = (struct sctp_inpcb *)so->so_pcb;
2145	ip_inp = (struct inpcb *)so->so_pcb;
2146#ifdef SCTP_DEBUG
2147	if (addr) {
2148		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n",
2149		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2150		SCTPDBG(SCTP_DEBUG_PCB1, "Addr :");
2151		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2152	}
2153#endif
2154	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2155		/* already did a bind, subsequent binds NOT allowed ! */
2156		return (EINVAL);
2157	}
2158#ifdef INVARIANTS
2159	if (p == NULL)
2160		panic("null proc/thread");
2161#endif
2162	if (p && jailed(p->td_ucred)) {
2163		prison = 1;
2164	}
2165	if (addr != NULL) {
2166		if (addr->sa_family == AF_INET) {
2167			struct sockaddr_in *sin;
2168
2169			/* IPV6_V6ONLY socket? */
2170			if (SCTP_IPV6_V6ONLY(ip_inp)) {
2171				return (EINVAL);
2172			}
2173			if (addr->sa_len != sizeof(*sin))
2174				return (EINVAL);
2175
2176			sin = (struct sockaddr_in *)addr;
2177			lport = sin->sin_port;
2178			if (prison) {
2179				/*
2180				 * For INADDR_ANY and  LOOPBACK the
2181				 * prison_ip() call will tranmute the ip
2182				 * address to the proper valie.
2183				 */
2184				if (prison_ip(p->td_ucred, 0, &sin->sin_addr.s_addr))
2185					return (EINVAL);
2186			}
2187			if (sin->sin_addr.s_addr != INADDR_ANY) {
2188				bindall = 0;
2189			}
2190		} else if (addr->sa_family == AF_INET6) {
2191			/* Only for pure IPv6 Address. (No IPv4 Mapped!) */
2192			struct sockaddr_in6 *sin6;
2193
2194			sin6 = (struct sockaddr_in6 *)addr;
2195
2196			if (addr->sa_len != sizeof(*sin6))
2197				return (EINVAL);
2198
2199			lport = sin6->sin6_port;
2200			/*
2201			 * Jail checks for IPv6 should go HERE! i.e. add the
2202			 * prison_ip() equivilant in this postion to
2203			 * transmute the addresses to the proper one jailed.
2204			 */
2205			if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2206				bindall = 0;
2207				/* KAME hack: embed scopeid */
2208				if (sa6_embedscope(sin6, ip6_use_defzone) != 0)
2209					return (EINVAL);
2210			}
2211			/* this must be cleared for ifa_ifwithaddr() */
2212			sin6->sin6_scope_id = 0;
2213		} else {
2214			return (EAFNOSUPPORT);
2215		}
2216	}
2217	/* Setup a vrf_id to be the default for the non-bind-all case. */
2218	vrf_id = inp->def_vrf_id;
2219
2220	SCTP_INP_INFO_WLOCK();
2221	SCTP_INP_WLOCK(inp);
2222	/* increase our count due to the unlock we do */
2223	SCTP_INP_INCR_REF(inp);
2224	if (lport) {
2225		/*
2226		 * Did the caller specify a port? if so we must see if a ep
2227		 * already has this one bound.
2228		 */
2229		/* got to be root to get at low ports */
2230		if (ntohs(lport) < IPPORT_RESERVED) {
2231			if (p && (error =
2232			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2233			    )) {
2234				SCTP_INP_DECR_REF(inp);
2235				SCTP_INP_WUNLOCK(inp);
2236				SCTP_INP_INFO_WUNLOCK();
2237				return (error);
2238			}
2239		}
2240		if (p == NULL) {
2241			SCTP_INP_DECR_REF(inp);
2242			SCTP_INP_WUNLOCK(inp);
2243			SCTP_INP_INFO_WUNLOCK();
2244			return (error);
2245		}
2246		SCTP_INP_WUNLOCK(inp);
2247		if (bindall) {
2248			vrf_id = inp->def_vrf_id;
2249			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2250			if (inp_tmp != NULL) {
2251				/*
2252				 * lock guy returned and lower count note
2253				 * that we are not bound so inp_tmp should
2254				 * NEVER be inp. And it is this inp
2255				 * (inp_tmp) that gets the reference bump,
2256				 * so we must lower it.
2257				 */
2258				SCTP_INP_DECR_REF(inp_tmp);
2259				SCTP_INP_DECR_REF(inp);
2260				/* unlock info */
2261				SCTP_INP_INFO_WUNLOCK();
2262				return (EADDRINUSE);
2263			}
2264		} else {
2265			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2266			if (inp_tmp != NULL) {
2267				/*
2268				 * lock guy returned and lower count note
2269				 * that we are not bound so inp_tmp should
2270				 * NEVER be inp. And it is this inp
2271				 * (inp_tmp) that gets the reference bump,
2272				 * so we must lower it.
2273				 */
2274				SCTP_INP_DECR_REF(inp_tmp);
2275				SCTP_INP_DECR_REF(inp);
2276				/* unlock info */
2277				SCTP_INP_INFO_WUNLOCK();
2278				return (EADDRINUSE);
2279			}
2280		}
2281		SCTP_INP_WLOCK(inp);
2282		if (bindall) {
2283			/* verify that no lport is not used by a singleton */
2284			if (sctp_isport_inuse(inp, lport, vrf_id)) {
2285				/* Sorry someone already has this one bound */
2286				SCTP_INP_DECR_REF(inp);
2287				SCTP_INP_WUNLOCK(inp);
2288				SCTP_INP_INFO_WUNLOCK();
2289				return (EADDRINUSE);
2290			}
2291		}
2292	} else {
2293		uint16_t first, last, candidate;
2294		uint16_t count;
2295		int done;
2296
2297		if (ip_inp->inp_flags & INP_HIGHPORT) {
2298			first = ipport_hifirstauto;
2299			last = ipport_hilastauto;
2300		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2301			if (p && (error =
2302			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2303			    )) {
2304				SCTP_INP_DECR_REF(inp);
2305				SCTP_INP_WUNLOCK(inp);
2306				SCTP_INP_INFO_WUNLOCK();
2307				return (error);
2308			}
2309			first = ipport_lowfirstauto;
2310			last = ipport_lowlastauto;
2311		} else {
2312			first = ipport_firstauto;
2313			last = ipport_lastauto;
2314		}
2315		if (first > last) {
2316			uint16_t temp;
2317
2318			temp = first;
2319			first = last;
2320			last = temp;
2321		}
2322		count = last - first + 1;	/* number of candidates */
2323		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2324
2325		done = 0;
2326		while (!done) {
2327			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == 0) {
2328				done = 1;
2329			}
2330			if (!done) {
2331				if (--count == 0) {
2332					SCTP_INP_DECR_REF(inp);
2333					SCTP_INP_WUNLOCK(inp);
2334					SCTP_INP_INFO_WUNLOCK();
2335					return (EADDRINUSE);
2336				}
2337				if (candidate == last)
2338					candidate = first;
2339				else
2340					candidate = candidate + 1;
2341			}
2342		}
2343		lport = htons(candidate);
2344	}
2345	SCTP_INP_DECR_REF(inp);
2346	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2347	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2348		/*
2349		 * this really should not happen. The guy did a non-blocking
2350		 * bind and then did a close at the same time.
2351		 */
2352		SCTP_INP_WUNLOCK(inp);
2353		SCTP_INP_INFO_WUNLOCK();
2354		return (EINVAL);
2355	}
2356	/* ok we look clear to give out this port, so lets setup the binding */
2357	if (bindall) {
2358		/* binding to all addresses, so just set in the proper flags */
2359		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
2360		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2361		/* set the automatic addr changes from kernel flag */
2362		if (sctp_auto_asconf == 0) {
2363			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2364		} else {
2365			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2366		}
2367	} else {
2368		/*
2369		 * bind specific, make sure flags is off and add a new
2370		 * address structure to the sctp_addr_list inside the ep
2371		 * structure.
2372		 *
2373		 * We will need to allocate one and insert it at the head. The
2374		 * socketopt call can just insert new addresses in there as
2375		 * well. It will also have to do the embed scope kame hack
2376		 * too (before adding).
2377		 */
2378		struct sctp_ifa *ifa;
2379		struct sockaddr_storage store_sa;
2380
2381		memset(&store_sa, 0, sizeof(store_sa));
2382		if (addr->sa_family == AF_INET) {
2383			struct sockaddr_in *sin;
2384
2385			sin = (struct sockaddr_in *)&store_sa;
2386			memcpy(sin, addr, sizeof(struct sockaddr_in));
2387			sin->sin_port = 0;
2388		} else if (addr->sa_family == AF_INET6) {
2389			struct sockaddr_in6 *sin6;
2390
2391			sin6 = (struct sockaddr_in6 *)&store_sa;
2392			memcpy(sin6, addr, sizeof(struct sockaddr_in6));
2393			sin6->sin6_port = 0;
2394		}
2395		/*
2396		 * first find the interface with the bound address need to
2397		 * zero out the port to find the address! yuck! can't do
2398		 * this earlier since need port for sctp_pcb_findep()
2399		 */
2400		if (sctp_ifap != NULL)
2401			ifa = sctp_ifap;
2402		else {
2403			/*
2404			 * Note for BSD we hit here always other O/S's will
2405			 * pass things in via the sctp_ifap argument
2406			 * (Panda).
2407			 */
2408			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
2409			    vrf_id, 0);
2410		}
2411		if (ifa == NULL) {
2412			/* Can't find an interface with that address */
2413			SCTP_INP_WUNLOCK(inp);
2414			SCTP_INP_INFO_WUNLOCK();
2415			return (EADDRNOTAVAIL);
2416		}
2417		if (addr->sa_family == AF_INET6) {
2418			/* GAK, more FIXME IFA lock? */
2419			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
2420				/* Can't bind a non-existent addr. */
2421				SCTP_INP_WUNLOCK(inp);
2422				SCTP_INP_INFO_WUNLOCK();
2423				return (EINVAL);
2424			}
2425		}
2426		/* we're not bound all */
2427		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
2428		/* allow bindx() to send ASCONF's for binding changes */
2429		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2430		/* clear automatic addr changes from kernel flag */
2431		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2432
2433		/* add this address to the endpoint list */
2434		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
2435		if (error != 0) {
2436			SCTP_INP_WUNLOCK(inp);
2437			SCTP_INP_INFO_WUNLOCK();
2438			return (error);
2439		}
2440		inp->laddr_count++;
2441	}
2442	/* find the bucket */
2443	head = &sctppcbinfo.sctp_ephash[SCTP_PCBHASH_ALLADDR(lport,
2444	    sctppcbinfo.hashmark)];
2445	/* put it in the bucket */
2446	LIST_INSERT_HEAD(head, inp, sctp_hash);
2447	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d\n",
2448	    head, ntohs(lport));
2449	/* set in the port */
2450	inp->sctp_lport = lport;
2451
2452	/* turn off just the unbound flag */
2453	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2454	SCTP_INP_WUNLOCK(inp);
2455	SCTP_INP_INFO_WUNLOCK();
2456	return (0);
2457}
2458
2459
2460static void
2461sctp_iterator_inp_being_freed(struct sctp_inpcb *inp, struct sctp_inpcb *inp_next)
2462{
2463	struct sctp_iterator *it;
2464
2465	/*
2466	 * We enter with the only the ITERATOR_LOCK in place and a write
2467	 * lock on the inp_info stuff.
2468	 */
2469
2470	/*
2471	 * Go through all iterators, we must do this since it is possible
2472	 * that some iterator does NOT have the lock, but is waiting for it.
2473	 * And the one that had the lock has either moved in the last
2474	 * iteration or we just cleared it above. We need to find all of
2475	 * those guys. The list of iterators should never be very big
2476	 * though.
2477	 */
2478	TAILQ_FOREACH(it, &sctppcbinfo.iteratorhead, sctp_nxt_itr) {
2479		if (it == inp->inp_starting_point_for_iterator)
2480			/* skip this guy, he's special */
2481			continue;
2482		if (it->inp == inp) {
2483			/*
2484			 * This is tricky and we DON'T lock the iterator.
2485			 * Reason is he's running but waiting for me since
2486			 * inp->inp_starting_point_for_iterator has the lock
2487			 * on me (the guy above we skipped). This tells us
2488			 * its is not running but waiting for
2489			 * inp->inp_starting_point_for_iterator to be
2490			 * released by the guy that does have our INP in a
2491			 * lock.
2492			 */
2493			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
2494				it->inp = NULL;
2495				it->stcb = NULL;
2496			} else {
2497				/* set him up to do the next guy not me */
2498				it->inp = inp_next;
2499				it->stcb = NULL;
2500			}
2501		}
2502	}
2503	it = inp->inp_starting_point_for_iterator;
2504	if (it) {
2505		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
2506			it->inp = NULL;
2507		} else {
2508			it->inp = inp_next;
2509		}
2510		it->stcb = NULL;
2511	}
2512}
2513
2514/* release sctp_inpcb unbind the port */
2515void
2516sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
2517{
2518	/*
2519	 * Here we free a endpoint. We must find it (if it is in the Hash
2520	 * table) and remove it from there. Then we must also find it in the
2521	 * overall list and remove it from there. After all removals are
2522	 * complete then any timer has to be stopped. Then start the actual
2523	 * freeing. a) Any local lists. b) Any associations. c) The hash of
2524	 * all associations. d) finally the ep itself.
2525	 */
2526	struct sctp_pcb *m;
2527	struct sctp_inpcb *inp_save;
2528	struct sctp_tcb *asoc, *nasoc;
2529	struct sctp_laddr *laddr, *nladdr;
2530	struct inpcb *ip_pcb;
2531	struct socket *so;
2532
2533	struct sctp_queued_to_read *sq;
2534
2535
2536	int cnt;
2537	sctp_sharedkey_t *shared_key;
2538
2539
2540#ifdef SCTP_LOG_CLOSING
2541	sctp_log_closing(inp, NULL, 0);
2542#endif
2543
2544	SCTP_ITERATOR_LOCK();
2545	so = inp->sctp_socket;
2546	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2547		/* been here before.. eeks.. get out of here */
2548		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
2549		SCTP_ITERATOR_UNLOCK();
2550#ifdef SCTP_LOG_CLOSING
2551		sctp_log_closing(inp, NULL, 1);
2552#endif
2553		return;
2554	}
2555	SCTP_ASOC_CREATE_LOCK(inp);
2556	SCTP_INP_INFO_WLOCK();
2557
2558	SCTP_INP_WLOCK(inp);
2559	/* First time through we have the socket lock, after that no more. */
2560	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
2561		/*
2562		 * Once we are in we can remove the flag from = 1 is only
2563		 * passed from the actual closing routines that are called
2564		 * via the sockets layer.
2565		 */
2566		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
2567	}
2568	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
2569	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
2570
2571	if (inp->control) {
2572		sctp_m_freem(inp->control);
2573		inp->control = NULL;
2574	}
2575	if (inp->pkt) {
2576		sctp_m_freem(inp->pkt);
2577		inp->pkt = NULL;
2578	}
2579	m = &inp->sctp_ep;
2580	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
2581					 * here but I will be nice :> (i.e.
2582					 * ip_pcb = ep;) */
2583	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
2584		int cnt_in_sd;
2585
2586		cnt_in_sd = 0;
2587		for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
2588		    asoc = nasoc) {
2589			nasoc = LIST_NEXT(asoc, sctp_tcblist);
2590			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2591				/* Skip guys being freed */
2592				asoc->sctp_socket = NULL;
2593				cnt_in_sd++;
2594				continue;
2595			}
2596			if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
2597			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) {
2598				/*
2599				 * If we have data in queue, we don't want
2600				 * to just free since the app may have done,
2601				 * send()/close or connect/send/close. And
2602				 * it wants the data to get across first.
2603				 */
2604				if (asoc->asoc.total_output_queue_size == 0) {
2605					/*
2606					 * Just abandon things in the front
2607					 * states
2608					 */
2609					sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_2);
2610					continue;
2611				}
2612			}
2613			SCTP_TCB_LOCK(asoc);
2614			/* Disconnect the socket please */
2615			asoc->sctp_socket = NULL;
2616			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
2617			if ((asoc->asoc.size_on_reasm_queue > 0) ||
2618			    (asoc->asoc.control_pdapi) ||
2619			    (asoc->asoc.size_on_all_streams > 0) ||
2620			    (so && (so->so_rcv.sb_cc > 0))
2621			    ) {
2622				/* Left with Data unread */
2623				struct mbuf *op_err;
2624
2625				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2626				    0, M_DONTWAIT, 1, MT_DATA);
2627				if (op_err) {
2628					/* Fill in the user initiated abort */
2629					struct sctp_paramhdr *ph;
2630					uint32_t *ippp;
2631
2632					SCTP_BUF_LEN(op_err) =
2633					    sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
2634					ph = mtod(op_err,
2635					    struct sctp_paramhdr *);
2636					ph->param_type = htons(
2637					    SCTP_CAUSE_USER_INITIATED_ABT);
2638					ph->param_length = htons(SCTP_BUF_LEN(op_err));
2639					ippp = (uint32_t *) (ph + 1);
2640					*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3);
2641				}
2642				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
2643				sctp_send_abort_tcb(asoc, op_err);
2644				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2645				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2646				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2647					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2648				}
2649				sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4);
2650				continue;
2651			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
2652				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
2653				    (asoc->asoc.stream_queue_cnt == 0)
2654			    ) {
2655				if (asoc->asoc.locked_on_sending) {
2656					goto abort_anyway;
2657				}
2658				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
2659				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
2660					/*
2661					 * there is nothing queued to send,
2662					 * so I send shutdown
2663					 */
2664					sctp_send_shutdown(asoc, asoc->asoc.primary_destination);
2665					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2666					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2667						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2668					}
2669					asoc->asoc.state = SCTP_STATE_SHUTDOWN_SENT;
2670					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
2671					    asoc->asoc.primary_destination);
2672					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
2673					    asoc->asoc.primary_destination);
2674					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR);
2675				}
2676			} else {
2677				/* mark into shutdown pending */
2678				struct sctp_stream_queue_pending *sp;
2679
2680				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
2681				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
2682				    asoc->asoc.primary_destination);
2683				if (asoc->asoc.locked_on_sending) {
2684					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
2685					    sctp_streamhead);
2686					if (sp == NULL) {
2687						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
2688						    asoc->asoc.locked_on_sending,
2689						    asoc->asoc.locked_on_sending->stream_no);
2690					} else {
2691						if ((sp->length == 0) && (sp->msg_is_complete == 0))
2692							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
2693					}
2694				}
2695				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
2696				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
2697				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
2698					struct mbuf *op_err;
2699
2700			abort_anyway:
2701					op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2702					    0, M_DONTWAIT, 1, MT_DATA);
2703					if (op_err) {
2704						/*
2705						 * Fill in the user
2706						 * initiated abort
2707						 */
2708						struct sctp_paramhdr *ph;
2709						uint32_t *ippp;
2710
2711						SCTP_BUF_LEN(op_err) =
2712						    (sizeof(struct sctp_paramhdr) +
2713						    sizeof(uint32_t));
2714						ph = mtod(op_err,
2715						    struct sctp_paramhdr *);
2716						ph->param_type = htons(
2717						    SCTP_CAUSE_USER_INITIATED_ABT);
2718						ph->param_length = htons(SCTP_BUF_LEN(op_err));
2719						ippp = (uint32_t *) (ph + 1);
2720						*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5);
2721					}
2722					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
2723					sctp_send_abort_tcb(asoc, op_err);
2724					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2725					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2726					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2727						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2728					}
2729					sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_6);
2730					continue;
2731				}
2732			}
2733			cnt_in_sd++;
2734			SCTP_TCB_UNLOCK(asoc);
2735		}
2736		/* now is there some left in our SHUTDOWN state? */
2737		if (cnt_in_sd) {
2738			SCTP_INP_WUNLOCK(inp);
2739			SCTP_ASOC_CREATE_UNLOCK(inp);
2740			SCTP_INP_INFO_WUNLOCK();
2741			SCTP_ITERATOR_UNLOCK();
2742#ifdef SCTP_LOG_CLOSING
2743			sctp_log_closing(inp, NULL, 2);
2744#endif
2745			return;
2746		}
2747	}
2748	inp->sctp_socket = NULL;
2749	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
2750	    SCTP_PCB_FLAGS_UNBOUND) {
2751		/*
2752		 * ok, this guy has been bound. It's port is somewhere in
2753		 * the sctppcbinfo hash table. Remove it!
2754		 */
2755		LIST_REMOVE(inp, sctp_hash);
2756		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
2757	}
2758	/*
2759	 * If there is a timer running to kill us, forget it, since it may
2760	 * have a contest on the INP lock.. which would cause us to die ...
2761	 */
2762	cnt = 0;
2763	for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
2764	    asoc = nasoc) {
2765		nasoc = LIST_NEXT(asoc, sctp_tcblist);
2766		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2767			cnt++;
2768			continue;
2769		}
2770		/* Free associations that are NOT killing us */
2771		SCTP_TCB_LOCK(asoc);
2772		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
2773		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
2774			struct mbuf *op_err;
2775			uint32_t *ippp;
2776
2777			op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2778			    0, M_DONTWAIT, 1, MT_DATA);
2779			if (op_err) {
2780				/* Fill in the user initiated abort */
2781				struct sctp_paramhdr *ph;
2782
2783				SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) +
2784				    sizeof(uint32_t));
2785				ph = mtod(op_err, struct sctp_paramhdr *);
2786				ph->param_type = htons(
2787				    SCTP_CAUSE_USER_INITIATED_ABT);
2788				ph->param_length = htons(SCTP_BUF_LEN(op_err));
2789				ippp = (uint32_t *) (ph + 1);
2790				*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7);
2791
2792			}
2793			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
2794			sctp_send_abort_tcb(asoc, op_err);
2795			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2796		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2797			cnt++;
2798			SCTP_TCB_UNLOCK(asoc);
2799			continue;
2800		}
2801		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2802		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2803			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2804		}
2805		sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8);
2806	}
2807	if (cnt) {
2808		/* Ok we have someone out there that will kill us */
2809		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2810		SCTP_INP_WUNLOCK(inp);
2811		SCTP_ASOC_CREATE_UNLOCK(inp);
2812		SCTP_INP_INFO_WUNLOCK();
2813		SCTP_ITERATOR_UNLOCK();
2814#ifdef SCTP_LOG_CLOSING
2815		sctp_log_closing(inp, NULL, 3);
2816#endif
2817		return;
2818	}
2819	if ((inp->refcount) || (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
2820		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2821		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
2822		SCTP_INP_WUNLOCK(inp);
2823		SCTP_ASOC_CREATE_UNLOCK(inp);
2824		SCTP_INP_INFO_WUNLOCK();
2825		SCTP_ITERATOR_UNLOCK();
2826#ifdef SCTP_LOG_CLOSING
2827		sctp_log_closing(inp, NULL, 4);
2828#endif
2829		return;
2830	}
2831	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2832	inp->sctp_ep.signature_change.type = 0;
2833	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
2834
2835#ifdef SCTP_LOG_CLOSING
2836	sctp_log_closing(inp, NULL, 5);
2837#endif
2838
2839	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2840	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NONE;
2841	/* Clear the read queue */
2842	/* sa_ignore FREED_MEMORY */
2843	while ((sq = TAILQ_FIRST(&inp->read_queue)) != NULL) {
2844		/* Its only abandoned if it had data left */
2845		if (sq->length)
2846			SCTP_STAT_INCR(sctps_left_abandon);
2847
2848		TAILQ_REMOVE(&inp->read_queue, sq, next);
2849		sctp_free_remote_addr(sq->whoFrom);
2850		if (so)
2851			so->so_rcv.sb_cc -= sq->length;
2852		if (sq->data) {
2853			sctp_m_freem(sq->data);
2854			sq->data = NULL;
2855		}
2856		/*
2857		 * no need to free the net count, since at this point all
2858		 * assoc's are gone.
2859		 */
2860		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, sq);
2861		SCTP_DECR_READQ_COUNT();
2862	}
2863	/* Now the sctp_pcb things */
2864	/*
2865	 * free each asoc if it is not already closed/free. we can't use the
2866	 * macro here since le_next will get freed as part of the
2867	 * sctp_free_assoc() call.
2868	 */
2869	cnt = 0;
2870	if (so) {
2871#ifdef IPSEC
2872		ipsec4_delete_pcbpolicy(ip_pcb);
2873#endif				/* IPSEC */
2874
2875		/* Unlocks not needed since the socket is gone now */
2876	}
2877	if (ip_pcb->inp_options) {
2878		(void)sctp_m_free(ip_pcb->inp_options);
2879		ip_pcb->inp_options = 0;
2880	}
2881	if (ip_pcb->inp_moptions) {
2882		inp_freemoptions(ip_pcb->inp_moptions);
2883		ip_pcb->inp_moptions = 0;
2884	}
2885#ifdef INET6
2886	if (ip_pcb->inp_vflag & INP_IPV6) {
2887		struct in6pcb *in6p;
2888
2889		in6p = (struct in6pcb *)inp;
2890		ip6_freepcbopts(in6p->in6p_outputopts);
2891	}
2892#endif				/* INET6 */
2893	ip_pcb->inp_vflag = 0;
2894	/* free up authentication fields */
2895	if (inp->sctp_ep.local_auth_chunks != NULL)
2896		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
2897	if (inp->sctp_ep.local_hmacs != NULL)
2898		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
2899
2900	shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
2901	while (shared_key) {
2902		LIST_REMOVE(shared_key, next);
2903		sctp_free_sharedkey(shared_key);
2904		/* sa_ignore FREED_MEMORY */
2905		shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
2906	}
2907
2908	inp_save = LIST_NEXT(inp, sctp_list);
2909	LIST_REMOVE(inp, sctp_list);
2910
2911	/* fix any iterators only after out of the list */
2912	sctp_iterator_inp_being_freed(inp, inp_save);
2913	/*
2914	 * if we have an address list the following will free the list of
2915	 * ifaddr's that are set into this ep. Again macro limitations here,
2916	 * since the LIST_FOREACH could be a bad idea.
2917	 */
2918	for ((laddr = LIST_FIRST(&inp->sctp_addr_list)); laddr != NULL;
2919	    laddr = nladdr) {
2920		nladdr = LIST_NEXT(laddr, sctp_nxt_addr);
2921		sctp_remove_laddr(laddr);
2922	}
2923
2924#ifdef SCTP_TRACK_FREED_ASOCS
2925	/* TEMP CODE */
2926	for ((asoc = LIST_FIRST(&inp->sctp_asoc_free_list)); asoc != NULL;
2927	    asoc = nasoc) {
2928		nasoc = LIST_NEXT(asoc, sctp_tcblist);
2929		LIST_REMOVE(asoc, sctp_tcblist);
2930		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, asoc);
2931		SCTP_DECR_ASOC_COUNT();
2932	}
2933	/* *** END TEMP CODE *** */
2934#endif
2935	/* Now lets see about freeing the EP hash table. */
2936	if (inp->sctp_tcbhash != NULL) {
2937		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
2938		inp->sctp_tcbhash = NULL;
2939	}
2940	/* Now we must put the ep memory back into the zone pool */
2941	INP_LOCK_DESTROY(&inp->ip_inp.inp);
2942	SCTP_INP_LOCK_DESTROY(inp);
2943	SCTP_INP_READ_DESTROY(inp);
2944	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
2945	SCTP_INP_INFO_WUNLOCK();
2946
2947	SCTP_ITERATOR_UNLOCK();
2948
2949	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
2950	SCTP_DECR_EP_COUNT();
2951
2952}
2953
2954
2955struct sctp_nets *
2956sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
2957{
2958	struct sctp_nets *net;
2959
2960	/* locate the address */
2961	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2962		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
2963			return (net);
2964	}
2965	return (NULL);
2966}
2967
2968
2969int
2970sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
2971{
2972	struct sctp_ifa *sctp_ifa;
2973
2974	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, 0);
2975	if (sctp_ifa) {
2976		return (1);
2977	} else {
2978		return (0);
2979	}
2980}
2981
2982/*
2983 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
2984 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
2985 * stats of stuff.
2986 */
2987int
2988sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
2989    int set_scope, int from)
2990{
2991	/*
2992	 * The following is redundant to the same lines in the
2993	 * sctp_aloc_assoc() but is needed since other's call the add
2994	 * address function
2995	 */
2996	struct sctp_nets *net, *netfirst;
2997	int addr_inscope;
2998
2999	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3000	    from);
3001	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3002
3003	netfirst = sctp_findnet(stcb, newaddr);
3004	if (netfirst) {
3005		/*
3006		 * Lie and return ok, we don't want to make the association
3007		 * go away for this behavior. It will happen in the TCP
3008		 * model in a connected socket. It does not reach the hash
3009		 * table until after the association is built so it can't be
3010		 * found. Mark as reachable, since the initial creation will
3011		 * have been cleared and the NOT_IN_ASSOC flag will have
3012		 * been added... and we don't want to end up removing it
3013		 * back out.
3014		 */
3015		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3016			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3017			    SCTP_ADDR_UNCONFIRMED);
3018		} else {
3019			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3020		}
3021
3022		return (0);
3023	}
3024	addr_inscope = 1;
3025	if (newaddr->sa_family == AF_INET) {
3026		struct sockaddr_in *sin;
3027
3028		sin = (struct sockaddr_in *)newaddr;
3029		if (sin->sin_addr.s_addr == 0) {
3030			/* Invalid address */
3031			return (-1);
3032		}
3033		/* zero out the bzero area */
3034		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3035
3036		/* assure len is set */
3037		sin->sin_len = sizeof(struct sockaddr_in);
3038		if (set_scope) {
3039#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3040			stcb->ipv4_local_scope = 1;
3041#else
3042			if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3043				stcb->asoc.ipv4_local_scope = 1;
3044			}
3045#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3046		} else {
3047			/* Validate the address is in scope */
3048			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3049			    (stcb->asoc.ipv4_local_scope == 0)) {
3050				addr_inscope = 0;
3051			}
3052		}
3053	} else if (newaddr->sa_family == AF_INET6) {
3054		struct sockaddr_in6 *sin6;
3055
3056		sin6 = (struct sockaddr_in6 *)newaddr;
3057		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3058			/* Invalid address */
3059			return (-1);
3060		}
3061		/* assure len is set */
3062		sin6->sin6_len = sizeof(struct sockaddr_in6);
3063		if (set_scope) {
3064			if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3065				stcb->asoc.loopback_scope = 1;
3066				stcb->asoc.local_scope = 0;
3067				stcb->asoc.ipv4_local_scope = 1;
3068				stcb->asoc.site_scope = 1;
3069			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3070				/*
3071				 * If the new destination is a LINK_LOCAL we
3072				 * must have common site scope. Don't set
3073				 * the local scope since we may not share
3074				 * all links, only loopback can do this.
3075				 * Links on the local network would also be
3076				 * on our private network for v4 too.
3077				 */
3078				stcb->asoc.ipv4_local_scope = 1;
3079				stcb->asoc.site_scope = 1;
3080			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3081				/*
3082				 * If the new destination is SITE_LOCAL then
3083				 * we must have site scope in common.
3084				 */
3085				stcb->asoc.site_scope = 1;
3086			}
3087		} else {
3088			/* Validate the address is in scope */
3089			if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3090			    (stcb->asoc.loopback_scope == 0)) {
3091				addr_inscope = 0;
3092			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3093			    (stcb->asoc.local_scope == 0)) {
3094				addr_inscope = 0;
3095			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3096			    (stcb->asoc.site_scope == 0)) {
3097				addr_inscope = 0;
3098			}
3099		}
3100	} else {
3101		/* not supported family type */
3102		return (-1);
3103	}
3104	net = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_net, struct sctp_nets);
3105	if (net == NULL) {
3106		return (-1);
3107	}
3108	SCTP_INCR_RADDR_COUNT();
3109	bzero(net, sizeof(*net));
3110	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3111	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3112	if (newaddr->sa_family == AF_INET) {
3113		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3114	} else if (newaddr->sa_family == AF_INET6) {
3115		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3116	}
3117	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3118	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3119		stcb->asoc.loopback_scope = 1;
3120		stcb->asoc.ipv4_local_scope = 1;
3121		stcb->asoc.local_scope = 0;
3122		stcb->asoc.site_scope = 1;
3123		addr_inscope = 1;
3124	}
3125	net->failure_threshold = stcb->asoc.def_net_failure;
3126	if (addr_inscope == 0) {
3127		net->dest_state = (SCTP_ADDR_REACHABLE |
3128		    SCTP_ADDR_OUT_OF_SCOPE);
3129	} else {
3130		if (from == SCTP_ADDR_IS_CONFIRMED)
3131			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3132			net->dest_state = SCTP_ADDR_REACHABLE;
3133		else
3134			net->dest_state = SCTP_ADDR_REACHABLE |
3135			    SCTP_ADDR_UNCONFIRMED;
3136	}
3137	/*
3138	 * We set this to 0, the timer code knows that this means its an
3139	 * initial value
3140	 */
3141	net->RTO = 0;
3142	net->RTO_measured = 0;
3143	stcb->asoc.numnets++;
3144	*(&net->ref_count) = 1;
3145	net->tos_flowlabel = 0;
3146#ifdef INET
3147	if (newaddr->sa_family == AF_INET)
3148		net->tos_flowlabel = stcb->asoc.default_tos;
3149#endif
3150#ifdef INET6
3151	if (newaddr->sa_family == AF_INET6)
3152		net->tos_flowlabel = stcb->asoc.default_flowlabel;
3153#endif
3154	/* Init the timer structure */
3155	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3156	SCTP_OS_TIMER_INIT(&net->fr_timer.timer);
3157	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3158
3159	/* Now generate a route for this guy */
3160	/* KAME hack: embed scopeid */
3161	if (newaddr->sa_family == AF_INET6) {
3162		struct sockaddr_in6 *sin6;
3163
3164		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3165		(void)sa6_embedscope(sin6, ip6_use_defzone);
3166		sin6->sin6_scope_id = 0;
3167	}
3168	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3169
3170	if (newaddr->sa_family == AF_INET6) {
3171		struct sockaddr_in6 *sin6;
3172
3173		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3174		(void)sa6_recoverscope(sin6);
3175	}
3176	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3177		/* Get source address */
3178		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3179		    stcb,
3180		    (sctp_route_t *) & net->ro,
3181		    net,
3182		    0,
3183		    stcb->asoc.vrf_id);
3184		/* Now get the interface MTU */
3185		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3186			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3187		} else {
3188			net->mtu = 0;
3189		}
3190#ifdef SCTP_PRINT_FOR_B_AND_M
3191		SCTP_PRINTF("We have found an interface mtu of %d\n", net->mtu);
3192#endif
3193		if (net->mtu == 0) {
3194			/* Huh ?? */
3195			net->mtu = SCTP_DEFAULT_MTU;
3196		} else {
3197			uint32_t rmtu;
3198
3199			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3200#ifdef SCTP_PRINT_FOR_B_AND_M
3201			SCTP_PRINTF("The route mtu is %d\n", rmtu);
3202#endif
3203			if (rmtu == 0) {
3204				/*
3205				 * Start things off to match mtu of
3206				 * interface please.
3207				 */
3208				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3209				    net->ro.ro_rt, net->mtu);
3210			} else {
3211				/*
3212				 * we take the route mtu over the interface,
3213				 * since the route may be leading out the
3214				 * loopback, or a different interface.
3215				 */
3216				net->mtu = rmtu;
3217			}
3218		}
3219		if (from == SCTP_ALLOC_ASOC) {
3220#ifdef SCTP_PRINT_FOR_B_AND_M
3221			SCTP_PRINTF("New assoc sets mtu to :%d\n", net->mtu);
3222#endif
3223			stcb->asoc.smallest_mtu = net->mtu;
3224		}
3225	} else {
3226		net->mtu = stcb->asoc.smallest_mtu;
3227	}
3228	if (stcb->asoc.smallest_mtu > net->mtu) {
3229#ifdef SCTP_PRINT_FOR_B_AND_M
3230		SCTP_PRINTF("new address mtu:%d smaller than smallest:%d\n",
3231		    net->mtu, stcb->asoc.smallest_mtu);
3232#endif
3233		stcb->asoc.smallest_mtu = net->mtu;
3234	}
3235	/* JRS - Use the congestion control given in the CC module */
3236	stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net);
3237
3238	/*
3239	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3240	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3241	 */
3242	net->find_pseudo_cumack = 1;
3243	net->find_rtx_pseudo_cumack = 1;
3244	net->src_addr_selected = 0;
3245	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3246	if (net->ro.ro_rt == NULL) {
3247		/* Since we have no route put it at the back */
3248		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3249	} else if (netfirst == NULL) {
3250		/* We are the first one in the pool. */
3251		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3252	} else if (netfirst->ro.ro_rt == NULL) {
3253		/*
3254		 * First one has NO route. Place this one ahead of the first
3255		 * one.
3256		 */
3257		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3258	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
3259		/*
3260		 * This one has a different interface than the one at the
3261		 * top of the list. Place it ahead.
3262		 */
3263		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3264	} else {
3265		/*
3266		 * Ok we have the same interface as the first one. Move
3267		 * forward until we find either a) one with a NULL route...
3268		 * insert ahead of that b) one with a different ifp.. insert
3269		 * after that. c) end of the list.. insert at the tail.
3270		 */
3271		struct sctp_nets *netlook;
3272
3273		do {
3274			netlook = TAILQ_NEXT(netfirst, sctp_next);
3275			if (netlook == NULL) {
3276				/* End of the list */
3277				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3278				break;
3279			} else if (netlook->ro.ro_rt == NULL) {
3280				/* next one has NO route */
3281				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
3282				break;
3283			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
3284				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
3285				    net, sctp_next);
3286				break;
3287			}
3288			/* Shift forward */
3289			netfirst = netlook;
3290		} while (netlook != NULL);
3291	}
3292
3293	/* got to have a primary set */
3294	if (stcb->asoc.primary_destination == 0) {
3295		stcb->asoc.primary_destination = net;
3296	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
3297		    (net->ro.ro_rt) &&
3298	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
3299		/* No route to current primary adopt new primary */
3300		stcb->asoc.primary_destination = net;
3301	}
3302	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb,
3303	    net);
3304	/* Validate primary is first */
3305	net = TAILQ_FIRST(&stcb->asoc.nets);
3306	if ((net != stcb->asoc.primary_destination) &&
3307	    (stcb->asoc.primary_destination)) {
3308		/*
3309		 * first one on the list is NOT the primary sctp_cmpaddr()
3310		 * is much more efficent if the primary is the first on the
3311		 * list, make it so.
3312		 */
3313		TAILQ_REMOVE(&stcb->asoc.nets,
3314		    stcb->asoc.primary_destination, sctp_next);
3315		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
3316		    stcb->asoc.primary_destination, sctp_next);
3317	}
3318	return (0);
3319}
3320
3321
3322/*
3323 * allocate an association and add it to the endpoint. The caller must be
3324 * careful to add all additional addresses once they are know right away or
3325 * else the assoc will be may experience a blackout scenario.
3326 */
3327struct sctp_tcb *
3328sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
3329    int for_a_init, int *error, uint32_t override_tag, uint32_t vrf_id,
3330    struct thread *p
3331)
3332{
3333	/* note the p argument is only valid in unbound sockets */
3334
3335	struct sctp_tcb *stcb;
3336	struct sctp_association *asoc;
3337	struct sctpasochead *head;
3338	uint16_t rport;
3339	int err;
3340
3341	/*
3342	 * Assumption made here: Caller has done a
3343	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
3344	 * address does not exist already.
3345	 */
3346	if (sctppcbinfo.ipi_count_asoc >= SCTP_MAX_NUM_OF_ASOC) {
3347		/* Hit max assoc, sorry no more */
3348		*error = ENOBUFS;
3349		return (NULL);
3350	}
3351	if (firstaddr == NULL) {
3352		*error = EINVAL;
3353		return (NULL);
3354	}
3355	SCTP_INP_RLOCK(inp);
3356	if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
3357		/*
3358		 * If its in the TCP pool, its NOT allowed to create an
3359		 * association. The parent listener needs to call
3360		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
3361		 * off, or connected one does this.. its an error.
3362		 */
3363		SCTP_INP_RUNLOCK(inp);
3364		*error = EINVAL;
3365		return (NULL);
3366	}
3367	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
3368#ifdef SCTP_DEBUG
3369	if (firstaddr) {
3370		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
3371		SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
3372		    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
3373	} else {
3374		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
3375	}
3376#endif				/* SCTP_DEBUG */
3377	if (firstaddr->sa_family == AF_INET) {
3378		struct sockaddr_in *sin;
3379
3380		sin = (struct sockaddr_in *)firstaddr;
3381		if ((sin->sin_port == 0) || (sin->sin_addr.s_addr == 0)) {
3382			/* Invalid address */
3383			SCTP_INP_RUNLOCK(inp);
3384			*error = EINVAL;
3385			return (NULL);
3386		}
3387		rport = sin->sin_port;
3388	} else if (firstaddr->sa_family == AF_INET6) {
3389		struct sockaddr_in6 *sin6;
3390
3391		sin6 = (struct sockaddr_in6 *)firstaddr;
3392		if ((sin6->sin6_port == 0) ||
3393		    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
3394			/* Invalid address */
3395			SCTP_INP_RUNLOCK(inp);
3396			*error = EINVAL;
3397			return (NULL);
3398		}
3399		rport = sin6->sin6_port;
3400	} else {
3401		/* not supported family type */
3402		SCTP_INP_RUNLOCK(inp);
3403		*error = EINVAL;
3404		return (NULL);
3405	}
3406	SCTP_INP_RUNLOCK(inp);
3407	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
3408		/*
3409		 * If you have not performed a bind, then we need to do the
3410		 * ephemerial bind for you.
3411		 */
3412		if ((err = sctp_inpcb_bind(inp->sctp_socket,
3413		    (struct sockaddr *)NULL,
3414		    (struct sctp_ifa *)NULL,
3415		    p
3416		    ))) {
3417			/* bind error, probably perm */
3418			*error = err;
3419			return (NULL);
3420		}
3421	}
3422	stcb = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_asoc, struct sctp_tcb);
3423	if (stcb == NULL) {
3424		/* out of memory? */
3425		*error = ENOMEM;
3426		return (NULL);
3427	}
3428	SCTP_INCR_ASOC_COUNT();
3429
3430	bzero(stcb, sizeof(*stcb));
3431	asoc = &stcb->asoc;
3432	SCTP_TCB_LOCK_INIT(stcb);
3433	SCTP_TCB_SEND_LOCK_INIT(stcb);
3434	/* setup back pointer's */
3435	stcb->sctp_ep = inp;
3436	stcb->sctp_socket = inp->sctp_socket;
3437	if ((err = sctp_init_asoc(inp, stcb, for_a_init, override_tag, vrf_id))) {
3438		/* failed */
3439		SCTP_TCB_LOCK_DESTROY(stcb);
3440		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3441		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3442		SCTP_DECR_ASOC_COUNT();
3443		*error = err;
3444		return (NULL);
3445	}
3446	/* and the port */
3447	stcb->rport = rport;
3448	SCTP_INP_INFO_WLOCK();
3449	SCTP_INP_WLOCK(inp);
3450	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3451		/* inpcb freed while alloc going on */
3452		SCTP_TCB_LOCK_DESTROY(stcb);
3453		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3454		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3455		SCTP_INP_WUNLOCK(inp);
3456		SCTP_INP_INFO_WUNLOCK();
3457		SCTP_DECR_ASOC_COUNT();
3458		*error = EINVAL;
3459		return (NULL);
3460	}
3461	SCTP_TCB_LOCK(stcb);
3462
3463	/* now that my_vtag is set, add it to the hash */
3464	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag,
3465	    sctppcbinfo.hashasocmark)];
3466	/* put it in the bucket in the vtag hash of assoc's for the system */
3467	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
3468	SCTP_INP_INFO_WUNLOCK();
3469
3470	if ((err = sctp_add_remote_addr(stcb, firstaddr, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
3471		/* failure.. memory error? */
3472		if (asoc->strmout) {
3473			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
3474			asoc->strmout = NULL;
3475		}
3476		if (asoc->mapping_array) {
3477			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
3478			asoc->mapping_array = NULL;
3479		}
3480		SCTP_DECR_ASOC_COUNT();
3481		SCTP_TCB_LOCK_DESTROY(stcb);
3482		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3483		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3484		SCTP_INP_WUNLOCK(inp);
3485		*error = ENOBUFS;
3486		return (NULL);
3487	}
3488	/* Init all the timers */
3489	SCTP_OS_TIMER_INIT(&asoc->hb_timer.timer);
3490	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
3491	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
3492	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
3493	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
3494	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
3495	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
3496
3497	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
3498	/* now file the port under the hash as well */
3499	if (inp->sctp_tcbhash != NULL) {
3500		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
3501		    inp->sctp_hashmark)];
3502		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
3503	}
3504	SCTP_INP_WUNLOCK(inp);
3505	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", stcb);
3506	return (stcb);
3507}
3508
3509
3510void
3511sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
3512{
3513	struct sctp_association *asoc;
3514
3515	asoc = &stcb->asoc;
3516	asoc->numnets--;
3517	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
3518	if (net == asoc->primary_destination) {
3519		/* Reset primary */
3520		struct sctp_nets *lnet;
3521
3522		lnet = TAILQ_FIRST(&asoc->nets);
3523		/* Try to find a confirmed primary */
3524		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
3525	}
3526	if (net == asoc->last_data_chunk_from) {
3527		/* Reset primary */
3528		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
3529	}
3530	if (net == asoc->last_control_chunk_from) {
3531		/* Clear net */
3532		asoc->last_control_chunk_from = NULL;
3533	}
3534	sctp_free_remote_addr(net);
3535}
3536
3537/*
3538 * remove a remote endpoint address from an association, it will fail if the
3539 * address does not exist.
3540 */
3541int
3542sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
3543{
3544	/*
3545	 * Here we need to remove a remote address. This is quite simple, we
3546	 * first find it in the list of address for the association
3547	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
3548	 * on that item. Note we do not allow it to be removed if there are
3549	 * no other addresses.
3550	 */
3551	struct sctp_association *asoc;
3552	struct sctp_nets *net, *net_tmp;
3553
3554	asoc = &stcb->asoc;
3555
3556	/* locate the address */
3557	for (net = TAILQ_FIRST(&asoc->nets); net != NULL; net = net_tmp) {
3558		net_tmp = TAILQ_NEXT(net, sctp_next);
3559		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
3560			continue;
3561		}
3562		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
3563		    remaddr)) {
3564			/* we found the guy */
3565			if (asoc->numnets < 2) {
3566				/* Must have at LEAST two remote addresses */
3567				return (-1);
3568			} else {
3569				sctp_remove_net(stcb, net);
3570				return (0);
3571			}
3572		}
3573	}
3574	/* not found. */
3575	return (-2);
3576}
3577
3578
3579void
3580sctp_add_vtag_to_timewait(struct sctp_inpcb *inp, uint32_t tag, uint32_t time)
3581{
3582	struct sctpvtaghead *chain;
3583	struct sctp_tagblock *twait_block;
3584	struct timeval now;
3585	int set, i;
3586
3587	(void)SCTP_GETTIME_TIMEVAL(&now);
3588	chain = &sctppcbinfo.vtag_timewait[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
3589	set = 0;
3590	if (!SCTP_LIST_EMPTY(chain)) {
3591		/* Block(s) present, lets find space, and expire on the fly */
3592		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
3593			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
3594				if ((twait_block->vtag_block[i].v_tag == 0) &&
3595				    !set) {
3596					twait_block->vtag_block[i].tv_sec_at_expire =
3597					    now.tv_sec + time;
3598					twait_block->vtag_block[i].v_tag = tag;
3599					set = 1;
3600				} else if ((twait_block->vtag_block[i].v_tag) &&
3601					    ((long)twait_block->vtag_block[i].tv_sec_at_expire >
3602				    now.tv_sec)) {
3603					/* Audit expires this guy */
3604					twait_block->vtag_block[i].tv_sec_at_expire = 0;
3605					twait_block->vtag_block[i].v_tag = 0;
3606					if (set == 0) {
3607						/* Reuse it for my new tag */
3608						twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + SCTP_TIME_WAIT;
3609						twait_block->vtag_block[0].v_tag = tag;
3610						set = 1;
3611					}
3612				}
3613			}
3614			if (set) {
3615				/*
3616				 * We only do up to the block where we can
3617				 * place our tag for audits
3618				 */
3619				break;
3620			}
3621		}
3622	}
3623	/* Need to add a new block to chain */
3624	if (!set) {
3625		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
3626		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
3627		if (twait_block == NULL) {
3628			return;
3629		}
3630		memset(twait_block, 0, sizeof(struct sctp_tagblock));
3631		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
3632		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec +
3633		    SCTP_TIME_WAIT;
3634		twait_block->vtag_block[0].v_tag = tag;
3635	}
3636}
3637
3638
3639static void
3640sctp_iterator_asoc_being_freed(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
3641{
3642	struct sctp_iterator *it;
3643
3644	/*
3645	 * Unlock the tcb lock we do this so we avoid a dead lock scenario
3646	 * where the iterator is waiting on the TCB lock and the TCB lock is
3647	 * waiting on the iterator lock.
3648	 */
3649	it = stcb->asoc.stcb_starting_point_for_iterator;
3650	if (it == NULL) {
3651		return;
3652	}
3653	if (it->inp != stcb->sctp_ep) {
3654		/* hmm, focused on the wrong one? */
3655		return;
3656	}
3657	if (it->stcb != stcb) {
3658		return;
3659	}
3660	it->stcb = LIST_NEXT(stcb, sctp_tcblist);
3661	if (it->stcb == NULL) {
3662		/* done with all asoc's in this assoc */
3663		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3664			it->inp = NULL;
3665		} else {
3666			it->inp = LIST_NEXT(inp, sctp_list);
3667		}
3668	}
3669}
3670
3671
3672/*
3673 * Free the association after un-hashing the remote port.
3674 */
3675int
3676sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
3677{
3678	int i;
3679	struct sctp_association *asoc;
3680	struct sctp_nets *net, *prev;
3681	struct sctp_laddr *laddr;
3682	struct sctp_tmit_chunk *chk;
3683	struct sctp_asconf_addr *aparam;
3684	struct sctp_stream_reset_list *liste;
3685	struct sctp_queued_to_read *sq;
3686	struct sctp_stream_queue_pending *sp;
3687	sctp_sharedkey_t *shared_key;
3688	struct socket *so;
3689	int ccnt = 0;
3690	int cnt = 0;
3691
3692	/* first, lets purge the entry from the hash table. */
3693
3694#ifdef SCTP_LOG_CLOSING
3695	sctp_log_closing(inp, stcb, 6);
3696#endif
3697	if (stcb->asoc.state == 0) {
3698#ifdef SCTP_LOG_CLOSING
3699		sctp_log_closing(inp, NULL, 7);
3700#endif
3701		/* there is no asoc, really TSNH :-0 */
3702		return (1);
3703	}
3704	/* TEMP CODE */
3705	if (stcb->freed_from_where == 0) {
3706		/* Only record the first place free happened from */
3707		stcb->freed_from_where = from_location;
3708	}
3709	/* TEMP CODE */
3710
3711	asoc = &stcb->asoc;
3712	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3713	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3714		/* nothing around */
3715		so = NULL;
3716	else
3717		so = inp->sctp_socket;
3718
3719	/*
3720	 * We used timer based freeing if a reader or writer is in the way.
3721	 * So we first check if we are actually being called from a timer,
3722	 * if so we abort early if a reader or writer is still in the way.
3723	 */
3724	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
3725	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
3726		/*
3727		 * is it the timer driving us? if so are the reader/writers
3728		 * gone?
3729		 */
3730		if (stcb->asoc.refcnt) {
3731			/* nope, reader or writer in the way */
3732			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
3733			/* no asoc destroyed */
3734			SCTP_TCB_UNLOCK(stcb);
3735#ifdef SCTP_LOG_CLOSING
3736			sctp_log_closing(inp, stcb, 8);
3737#endif
3738			return (0);
3739		}
3740	}
3741	/* now clean up any other timers */
3742	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
3743	asoc->hb_timer.self = NULL;
3744	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
3745	asoc->dack_timer.self = NULL;
3746	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3747	/*-
3748	 * For stream reset we don't blast this unless
3749	 * it is a str-reset timer, it might be the
3750	 * free-asoc timer which we DON'T want to
3751	 * disturb.
3752	 */
3753	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
3754		asoc->strreset_timer.self = NULL;
3755	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
3756	asoc->asconf_timer.self = NULL;
3757	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
3758	asoc->autoclose_timer.self = NULL;
3759	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
3760	asoc->shut_guard_timer.self = NULL;
3761	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
3762	asoc->delayed_event_timer.self = NULL;
3763	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
3764		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
3765		net->fr_timer.self = NULL;
3766		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
3767		net->rxt_timer.self = NULL;
3768		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
3769		net->pmtu_timer.self = NULL;
3770	}
3771	/* Now the read queue needs to be cleaned up (only once) */
3772	cnt = 0;
3773	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
3774		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
3775		SCTP_INP_READ_LOCK(inp);
3776		TAILQ_FOREACH(sq, &inp->read_queue, next) {
3777			if (sq->stcb == stcb) {
3778				sq->do_not_ref_stcb = 1;
3779				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
3780				/*
3781				 * If there is no end, there never will be
3782				 * now.
3783				 */
3784				if (sq->end_added == 0) {
3785					/* Held for PD-API clear that. */
3786					sq->pdapi_aborted = 1;
3787					sq->held_length = 0;
3788					if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT)) {
3789						/*
3790						 * Need to add a PD-API
3791						 * aborted indication.
3792						 * Setting the control_pdapi
3793						 * assures that it will be
3794						 * added right after this
3795						 * msg.
3796						 */
3797						uint32_t strseq;
3798
3799						stcb->asoc.control_pdapi = sq;
3800						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
3801						sctp_notify_partial_delivery_indication(stcb,
3802						    SCTP_PARTIAL_DELIVERY_ABORTED, 1, strseq);
3803						stcb->asoc.control_pdapi = NULL;
3804					}
3805				}
3806				/* Add an end to wake them */
3807				sq->end_added = 1;
3808				cnt++;
3809			}
3810		}
3811		SCTP_INP_READ_UNLOCK(inp);
3812		if (stcb->block_entry) {
3813			cnt++;
3814			stcb->block_entry->error = ECONNRESET;
3815			stcb->block_entry = NULL;
3816		}
3817	}
3818	if ((from_inpcbfree != SCTP_PCBFREE_FORCE) && (stcb->asoc.refcnt)) {
3819		/*
3820		 * reader or writer in the way, we have hopefully given him
3821		 * something to chew on above.
3822		 */
3823		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
3824		SCTP_TCB_UNLOCK(stcb);
3825		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3826		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3827			/* nothing around */
3828			so = NULL;
3829		if (so) {
3830			/* Wake any reader/writers */
3831			sctp_sorwakeup(inp, so);
3832			sctp_sowwakeup(inp, so);
3833		}
3834#ifdef SCTP_LOG_CLOSING
3835		sctp_log_closing(inp, stcb, 9);
3836#endif
3837		/* no asoc destroyed */
3838		return (0);
3839	}
3840#ifdef SCTP_LOG_CLOSING
3841	sctp_log_closing(inp, stcb, 10);
3842#endif
3843	/*
3844	 * When I reach here, no others want to kill the assoc yet.. and I
3845	 * own the lock. Now its possible an abort comes in when I do the
3846	 * lock exchange below to grab all the locks to do the final take
3847	 * out. to prevent this we increment the count, which will start a
3848	 * timer and blow out above thus assuring us that we hold exclusive
3849	 * killing of the asoc. Note that after getting back the TCB lock we
3850	 * will go ahead and increment the counter back up and stop any
3851	 * timer a passing stranger may have started :-S
3852	 */
3853	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3854		atomic_add_int(&stcb->asoc.refcnt, 1);
3855
3856		SCTP_TCB_UNLOCK(stcb);
3857
3858		SCTP_ITERATOR_LOCK();
3859		SCTP_INP_INFO_WLOCK();
3860		SCTP_INP_WLOCK(inp);
3861		SCTP_TCB_LOCK(stcb);
3862	}
3863	/* Double check the GONE flag */
3864	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3865	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3866		/* nothing around */
3867		so = NULL;
3868
3869	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
3870	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
3871		/*
3872		 * For TCP type we need special handling when we are
3873		 * connected. We also include the peel'ed off ones to.
3874		 */
3875		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
3876			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
3877			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
3878			if (so) {
3879				SOCK_LOCK(so);
3880				if (so->so_rcv.sb_cc == 0) {
3881					so->so_state &= ~(SS_ISCONNECTING |
3882					    SS_ISDISCONNECTING |
3883					    SS_ISCONFIRMING |
3884					    SS_ISCONNECTED);
3885				}
3886				SOCK_UNLOCK(so);
3887				sctp_sowwakeup(inp, so);
3888				sctp_sorwakeup(inp, so);
3889				SCTP_SOWAKEUP(so);
3890			}
3891		}
3892	}
3893	/*
3894	 * Make it invalid too, that way if its about to run it will abort
3895	 * and return.
3896	 */
3897	sctp_iterator_asoc_being_freed(inp, stcb);
3898	/* re-increment the lock */
3899	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3900		atomic_add_int(&stcb->asoc.refcnt, -1);
3901	}
3902	asoc->state = 0;
3903	if (inp->sctp_tcbhash) {
3904		LIST_REMOVE(stcb, sctp_tcbhash);
3905	}
3906	if (stcb->asoc.in_restart_hash) {
3907		LIST_REMOVE(stcb, sctp_tcbrestarhash);
3908	}
3909	/* Now lets remove it from the list of ALL associations in the EP */
3910	LIST_REMOVE(stcb, sctp_tcblist);
3911	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3912		SCTP_INP_INCR_REF(inp);
3913		SCTP_INP_WUNLOCK(inp);
3914		SCTP_ITERATOR_UNLOCK();
3915	}
3916	/* pull from vtag hash */
3917	LIST_REMOVE(stcb, sctp_asocs);
3918	sctp_add_vtag_to_timewait(inp, asoc->my_vtag, SCTP_TIME_WAIT);
3919
3920
3921	/*
3922	 * Now restop the timers to be sure - this is paranoia at is finest!
3923	 */
3924	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3925	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
3926	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
3927	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3928	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
3929	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
3930	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
3931	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
3932
3933	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
3934		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
3935		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
3936		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
3937	}
3938
3939	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
3940	prev = NULL;
3941	/*
3942	 * The chunk lists and such SHOULD be empty but we check them just
3943	 * in case.
3944	 */
3945	/* anything on the wheel needs to be removed */
3946	for (i = 0; i < asoc->streamoutcnt; i++) {
3947		struct sctp_stream_out *outs;
3948
3949		outs = &asoc->strmout[i];
3950		/* now clean up any chunks here */
3951		sp = TAILQ_FIRST(&outs->outqueue);
3952		while (sp) {
3953			TAILQ_REMOVE(&outs->outqueue, sp, next);
3954			if (sp->data) {
3955				sctp_m_freem(sp->data);
3956				sp->data = NULL;
3957				sp->tail_mbuf = NULL;
3958			}
3959			sctp_free_remote_addr(sp->net);
3960			sctp_free_spbufspace(stcb, asoc, sp);
3961			/* Free the zone stuff  */
3962			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_strmoq, sp);
3963			SCTP_DECR_STRMOQ_COUNT();
3964			/* sa_ignore FREED_MEMORY */
3965			sp = TAILQ_FIRST(&outs->outqueue);
3966		}
3967	}
3968
3969	/* sa_ignore FREED_MEMORY */
3970	while ((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) {
3971		TAILQ_REMOVE(&asoc->resetHead, liste, next_resp);
3972		SCTP_FREE(liste, SCTP_M_STRESET);
3973	}
3974
3975	sq = TAILQ_FIRST(&asoc->pending_reply_queue);
3976	while (sq) {
3977		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
3978		if (sq->data) {
3979			sctp_m_freem(sq->data);
3980			sq->data = NULL;
3981		}
3982		sctp_free_remote_addr(sq->whoFrom);
3983		sq->whoFrom = NULL;
3984		sq->stcb = NULL;
3985		/* Free the ctl entry */
3986		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, sq);
3987		SCTP_DECR_READQ_COUNT();
3988		/* sa_ignore FREED_MEMORY */
3989		sq = TAILQ_FIRST(&asoc->pending_reply_queue);
3990	}
3991
3992	chk = TAILQ_FIRST(&asoc->free_chunks);
3993	while (chk) {
3994		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
3995		if (chk->data) {
3996			sctp_m_freem(chk->data);
3997			chk->data = NULL;
3998		}
3999		ccnt++;
4000		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4001		SCTP_DECR_CHK_COUNT();
4002		atomic_subtract_int(&sctppcbinfo.ipi_free_chunks, 1);
4003		asoc->free_chunk_cnt--;
4004		/* sa_ignore FREED_MEMORY */
4005		chk = TAILQ_FIRST(&asoc->free_chunks);
4006	}
4007	/* pending send queue SHOULD be empty */
4008	if (!TAILQ_EMPTY(&asoc->send_queue)) {
4009		chk = TAILQ_FIRST(&asoc->send_queue);
4010		while (chk) {
4011			TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4012			if (chk->data) {
4013				sctp_m_freem(chk->data);
4014				chk->data = NULL;
4015			}
4016			ccnt++;
4017			sctp_free_remote_addr(chk->whoTo);
4018			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4019			SCTP_DECR_CHK_COUNT();
4020			/* sa_ignore FREED_MEMORY */
4021			chk = TAILQ_FIRST(&asoc->send_queue);
4022		}
4023	}
4024/*
4025  if(ccnt) {
4026  printf("Freed %d from send_queue\n", ccnt);
4027  ccnt = 0;
4028  }
4029*/
4030	/* sent queue SHOULD be empty */
4031	if (!TAILQ_EMPTY(&asoc->sent_queue)) {
4032		chk = TAILQ_FIRST(&asoc->sent_queue);
4033		while (chk) {
4034			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4035			if (chk->data) {
4036				sctp_m_freem(chk->data);
4037				chk->data = NULL;
4038			}
4039			ccnt++;
4040			sctp_free_remote_addr(chk->whoTo);
4041			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4042			SCTP_DECR_CHK_COUNT();
4043			/* sa_ignore FREED_MEMORY */
4044			chk = TAILQ_FIRST(&asoc->sent_queue);
4045		}
4046	}
4047/*
4048  if(ccnt) {
4049  printf("Freed %d from sent_queue\n", ccnt);
4050  ccnt = 0;
4051  }
4052*/
4053	/* control queue MAY not be empty */
4054	if (!TAILQ_EMPTY(&asoc->control_send_queue)) {
4055		chk = TAILQ_FIRST(&asoc->control_send_queue);
4056		while (chk) {
4057			TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4058			if (chk->data) {
4059				sctp_m_freem(chk->data);
4060				chk->data = NULL;
4061			}
4062			ccnt++;
4063			sctp_free_remote_addr(chk->whoTo);
4064			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4065			SCTP_DECR_CHK_COUNT();
4066			/* sa_ignore FREED_MEMORY */
4067			chk = TAILQ_FIRST(&asoc->control_send_queue);
4068		}
4069	}
4070/*
4071  if(ccnt) {
4072  printf("Freed %d from ctrl_queue\n", ccnt);
4073  ccnt = 0;
4074  }
4075*/
4076	if (!TAILQ_EMPTY(&asoc->reasmqueue)) {
4077		chk = TAILQ_FIRST(&asoc->reasmqueue);
4078		while (chk) {
4079			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
4080			if (chk->data) {
4081				sctp_m_freem(chk->data);
4082				chk->data = NULL;
4083			}
4084			sctp_free_remote_addr(chk->whoTo);
4085			ccnt++;
4086			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4087			SCTP_DECR_CHK_COUNT();
4088			/* sa_ignore FREED_MEMORY */
4089			chk = TAILQ_FIRST(&asoc->reasmqueue);
4090		}
4091	}
4092/*
4093  if(ccnt) {
4094  printf("Freed %d from reasm_queue\n", ccnt);
4095  ccnt = 0;
4096  }
4097*/
4098	if (asoc->mapping_array) {
4099		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4100		asoc->mapping_array = NULL;
4101	}
4102	/* the stream outs */
4103	if (asoc->strmout) {
4104		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4105		asoc->strmout = NULL;
4106	}
4107	asoc->streamoutcnt = 0;
4108	if (asoc->strmin) {
4109		struct sctp_queued_to_read *ctl;
4110
4111		for (i = 0; i < asoc->streamincnt; i++) {
4112			if (!TAILQ_EMPTY(&asoc->strmin[i].inqueue)) {
4113				/* We have somethings on the streamin queue */
4114				ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4115				while (ctl) {
4116					TAILQ_REMOVE(&asoc->strmin[i].inqueue,
4117					    ctl, next);
4118					sctp_free_remote_addr(ctl->whoFrom);
4119					if (ctl->data) {
4120						sctp_m_freem(ctl->data);
4121						ctl->data = NULL;
4122					}
4123					/*
4124					 * We don't free the address here
4125					 * since all the net's were freed
4126					 * above.
4127					 */
4128					SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, ctl);
4129					SCTP_DECR_READQ_COUNT();
4130					ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4131				}
4132			}
4133		}
4134		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
4135		asoc->strmin = NULL;
4136	}
4137	asoc->streamincnt = 0;
4138	while (!TAILQ_EMPTY(&asoc->nets)) {
4139		/* sa_ignore FREED_MEMORY */
4140		net = TAILQ_FIRST(&asoc->nets);
4141		/* pull from list */
4142		if ((sctppcbinfo.ipi_count_raddr == 0) || (prev == net)) {
4143#ifdef INVARIANTS
4144			panic("no net's left alloc'ed, or list points to itself");
4145#endif
4146			break;
4147		}
4148		prev = net;
4149		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4150		sctp_free_remote_addr(net);
4151	}
4152
4153	while (!SCTP_LIST_EMPTY(&asoc->sctp_restricted_addrs)) {
4154		/* sa_ignore FREED_MEMORY */
4155		laddr = LIST_FIRST(&asoc->sctp_restricted_addrs);
4156		sctp_remove_laddr(laddr);
4157	}
4158
4159	/* pending asconf (address) parameters */
4160	while (!TAILQ_EMPTY(&asoc->asconf_queue)) {
4161		/* sa_ignore FREED_MEMORY */
4162		aparam = TAILQ_FIRST(&asoc->asconf_queue);
4163		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
4164		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
4165	}
4166	if (asoc->last_asconf_ack_sent != NULL) {
4167		sctp_m_freem(asoc->last_asconf_ack_sent);
4168		asoc->last_asconf_ack_sent = NULL;
4169	}
4170	/* clean up auth stuff */
4171	if (asoc->local_hmacs)
4172		sctp_free_hmaclist(asoc->local_hmacs);
4173	if (asoc->peer_hmacs)
4174		sctp_free_hmaclist(asoc->peer_hmacs);
4175
4176	if (asoc->local_auth_chunks)
4177		sctp_free_chunklist(asoc->local_auth_chunks);
4178	if (asoc->peer_auth_chunks)
4179		sctp_free_chunklist(asoc->peer_auth_chunks);
4180
4181	sctp_free_authinfo(&asoc->authinfo);
4182
4183	shared_key = LIST_FIRST(&asoc->shared_keys);
4184	while (shared_key) {
4185		LIST_REMOVE(shared_key, next);
4186		sctp_free_sharedkey(shared_key);
4187		/* sa_ignore FREED_MEMORY */
4188		shared_key = LIST_FIRST(&asoc->shared_keys);
4189	}
4190
4191	/* Insert new items here :> */
4192
4193	/* Get rid of LOCK */
4194	SCTP_TCB_LOCK_DESTROY(stcb);
4195	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4196	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4197		SCTP_INP_INFO_WUNLOCK();
4198		SCTP_INP_RLOCK(inp);
4199	}
4200#ifdef SCTP_TRACK_FREED_ASOCS
4201	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
4202		/* now clean up the tasoc itself */
4203		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
4204		SCTP_DECR_ASOC_COUNT();
4205	} else {
4206		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
4207	}
4208#else
4209	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
4210	SCTP_DECR_ASOC_COUNT();
4211#endif
4212	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4213		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
4214			/*
4215			 * If its NOT the inp_free calling us AND sctp_close
4216			 * as been called, we call back...
4217			 */
4218			SCTP_INP_RUNLOCK(inp);
4219			/*
4220			 * This will start the kill timer (if we are the
4221			 * lastone) since we hold an increment yet. But this
4222			 * is the only safe way to do this since otherwise
4223			 * if the socket closes at the same time we are here
4224			 * we might collide in the cleanup.
4225			 */
4226			sctp_inpcb_free(inp,
4227			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
4228			    SCTP_CALLED_DIRECTLY_NOCMPSET);
4229			SCTP_INP_DECR_REF(inp);
4230			goto out_of;
4231		} else {
4232			/* The socket is still open. */
4233			SCTP_INP_DECR_REF(inp);
4234		}
4235	}
4236	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4237		SCTP_INP_RUNLOCK(inp);
4238	}
4239out_of:
4240	/* destroyed the asoc */
4241#ifdef SCTP_LOG_CLOSING
4242	sctp_log_closing(inp, NULL, 11);
4243#endif
4244	return (1);
4245}
4246
4247
4248
4249/*
4250 * determine if a destination is "reachable" based upon the addresses bound
4251 * to the current endpoint (e.g. only v4 or v6 currently bound)
4252 */
4253/*
4254 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
4255 * assoc level v4/v6 flags, as the assoc *may* not have the same address
4256 * types bound as its endpoint
4257 */
4258int
4259sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
4260{
4261	struct sctp_inpcb *inp;
4262	int answer;
4263
4264	/*
4265	 * No locks here, the TCB, in all cases is already locked and an
4266	 * assoc is up. There is either a INP lock by the caller applied (in
4267	 * asconf case when deleting an address) or NOT in the HB case,
4268	 * however if HB then the INP increment is up and the INP will not
4269	 * be removed (on top of the fact that we have a TCB lock). So we
4270	 * only want to read the sctp_flags, which is either bound-all or
4271	 * not.. no protection needed since once an assoc is up you can't be
4272	 * changing your binding.
4273	 */
4274	inp = stcb->sctp_ep;
4275	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4276		/* if bound all, destination is not restricted */
4277		/*
4278		 * RRS: Question during lock work: Is this correct? If you
4279		 * are bound-all you still might need to obey the V4--V6
4280		 * flags??? IMO this bound-all stuff needs to be removed!
4281		 */
4282		return (1);
4283	}
4284	/* NOTE: all "scope" checks are done when local addresses are added */
4285	if (destaddr->sa_family == AF_INET6) {
4286		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
4287	} else if (destaddr->sa_family == AF_INET) {
4288		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
4289	} else {
4290		/* invalid family, so it's unreachable */
4291		answer = 0;
4292	}
4293	return (answer);
4294}
4295
4296/*
4297 * update the inp_vflags on an endpoint
4298 */
4299static void
4300sctp_update_ep_vflag(struct sctp_inpcb *inp)
4301{
4302	struct sctp_laddr *laddr;
4303
4304	/* first clear the flag */
4305	inp->ip_inp.inp.inp_vflag = 0;
4306	/* set the flag based on addresses on the ep list */
4307	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4308		if (laddr->ifa == NULL) {
4309			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
4310			    __FUNCTION__);
4311			continue;
4312		}
4313		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
4314			continue;
4315		}
4316		if (laddr->ifa->address.sa.sa_family == AF_INET6) {
4317			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
4318		} else if (laddr->ifa->address.sa.sa_family == AF_INET) {
4319			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
4320		}
4321	}
4322}
4323
4324/*
4325 * Add the address to the endpoint local address list There is nothing to be
4326 * done if we are bound to all addresses
4327 */
4328void
4329sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
4330{
4331	struct sctp_laddr *laddr;
4332	int fnd, error = 0;
4333
4334	fnd = 0;
4335
4336	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4337		/* You are already bound to all. You have it already */
4338		return;
4339	}
4340	if (ifa->address.sa.sa_family == AF_INET6) {
4341		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
4342			/* Can't bind a non-useable addr. */
4343			return;
4344		}
4345	}
4346	/* first, is it already present? */
4347	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4348		if (laddr->ifa == ifa) {
4349			fnd = 1;
4350			break;
4351		}
4352	}
4353
4354	if (fnd == 0) {
4355		/* Not in the ep list */
4356		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
4357		if (error != 0)
4358			return;
4359		inp->laddr_count++;
4360		/* update inp_vflag flags */
4361		if (ifa->address.sa.sa_family == AF_INET6) {
4362			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
4363		} else if (ifa->address.sa.sa_family == AF_INET) {
4364			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
4365		}
4366	}
4367	return;
4368}
4369
4370
4371/*
4372 * select a new (hopefully reachable) destination net (should only be used
4373 * when we deleted an ep addr that is the only usable source address to reach
4374 * the destination net)
4375 */
4376static void
4377sctp_select_primary_destination(struct sctp_tcb *stcb)
4378{
4379	struct sctp_nets *net;
4380
4381	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4382		/* for now, we'll just pick the first reachable one we find */
4383		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
4384			continue;
4385		if (sctp_destination_is_reachable(stcb,
4386		    (struct sockaddr *)&net->ro._l_addr)) {
4387			/* found a reachable destination */
4388			stcb->asoc.primary_destination = net;
4389		}
4390	}
4391	/* I can't there from here! ...we're gonna die shortly... */
4392}
4393
4394
4395/*
4396 * Delete the address from the endpoint local address list There is nothing
4397 * to be done if we are bound to all addresses
4398 */
4399void
4400sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
4401{
4402	struct sctp_laddr *laddr;
4403	int fnd;
4404
4405	fnd = 0;
4406	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4407		/* You are already bound to all. You have it already */
4408		return;
4409	}
4410	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4411		if (laddr->ifa == ifa) {
4412			fnd = 1;
4413			break;
4414		}
4415	}
4416	if (fnd && (inp->laddr_count < 2)) {
4417		/* can't delete unless there are at LEAST 2 addresses */
4418		return;
4419	}
4420	if (fnd) {
4421		/*
4422		 * clean up any use of this address go through our
4423		 * associations and clear any last_used_address that match
4424		 * this one for each assoc, see if a new primary_destination
4425		 * is needed
4426		 */
4427		struct sctp_tcb *stcb;
4428
4429		/* clean up "next_addr_touse" */
4430		if (inp->next_addr_touse == laddr)
4431			/* delete this address */
4432			inp->next_addr_touse = NULL;
4433
4434		/* clean up "last_used_address" */
4435		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
4436			struct sctp_nets *net;
4437
4438			SCTP_TCB_LOCK(stcb);
4439			if (stcb->asoc.last_used_address == laddr)
4440				/* delete this address */
4441				stcb->asoc.last_used_address = NULL;
4442			/*
4443			 * Now spin through all the nets and purge any ref
4444			 * to laddr
4445			 */
4446			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4447				if (net->ro._s_addr &&
4448				    (net->ro._s_addr->ifa == laddr->ifa)) {
4449					/* Yep, purge src address selected */
4450					sctp_rtentry_t *rt;
4451
4452					/* delete this address if cached */
4453					rt = net->ro.ro_rt;
4454					if (rt != NULL) {
4455						RTFREE(rt);
4456						net->ro.ro_rt = NULL;
4457					}
4458					sctp_free_ifa(net->ro._s_addr);
4459					net->ro._s_addr = NULL;
4460					net->src_addr_selected = 0;
4461				}
4462			}
4463			SCTP_TCB_UNLOCK(stcb);
4464		}		/* for each tcb */
4465		/* remove it from the ep list */
4466		sctp_remove_laddr(laddr);
4467		inp->laddr_count--;
4468		/* update inp_vflag flags */
4469		sctp_update_ep_vflag(inp);
4470	}
4471	return;
4472}
4473
4474/*
4475 * Add the address to the TCB local address restricted list.
4476 * This is a "pending" address list (eg. addresses waiting for an
4477 * ASCONF-ACK response) and cannot be used as a valid source address.
4478 */
4479void
4480sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
4481{
4482	struct sctp_inpcb *inp;
4483	struct sctp_laddr *laddr;
4484	struct sctpladdr *list;
4485
4486	/*
4487	 * Assumes TCB is locked.. and possibly the INP. May need to
4488	 * confirm/fix that if we need it and is not the case.
4489	 */
4490	list = &stcb->asoc.sctp_restricted_addrs;
4491
4492	inp = stcb->sctp_ep;
4493	if (ifa->address.sa.sa_family == AF_INET6) {
4494		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
4495			/* Can't bind a non-existent addr. */
4496			return;
4497		}
4498	}
4499	/* does the address already exist? */
4500	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
4501		if (laddr->ifa == ifa) {
4502			return;
4503		}
4504	}
4505
4506	/* add to the list */
4507	(void)sctp_insert_laddr(list, ifa, 0);
4508	return;
4509}
4510
4511/*
4512 * insert an laddr entry with the given ifa for the desired list
4513 */
4514int
4515sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
4516{
4517	struct sctp_laddr *laddr;
4518
4519	laddr = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
4520	if (laddr == NULL) {
4521		/* out of memory? */
4522		return (EINVAL);
4523	}
4524	SCTP_INCR_LADDR_COUNT();
4525	bzero(laddr, sizeof(*laddr));
4526	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
4527	laddr->ifa = ifa;
4528	laddr->action = act;
4529	atomic_add_int(&ifa->refcount, 1);
4530	/* insert it */
4531	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
4532
4533	return (0);
4534}
4535
4536/*
4537 * Remove an laddr entry from the local address list (on an assoc)
4538 */
4539void
4540sctp_remove_laddr(struct sctp_laddr *laddr)
4541{
4542
4543	/* remove from the list */
4544	LIST_REMOVE(laddr, sctp_nxt_addr);
4545	sctp_free_ifa(laddr->ifa);
4546	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_laddr, laddr);
4547	SCTP_DECR_LADDR_COUNT();
4548}
4549
4550/*
4551 * Remove a local address from the TCB local address restricted list
4552 */
4553void
4554sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
4555{
4556	struct sctp_inpcb *inp;
4557	struct sctp_laddr *laddr;
4558
4559	/*
4560	 * This is called by asconf work. It is assumed that a) The TCB is
4561	 * locked and b) The INP is locked. This is true in as much as I can
4562	 * trace through the entry asconf code where I did these locks.
4563	 * Again, the ASCONF code is a bit different in that it does lock
4564	 * the INP during its work often times. This must be since we don't
4565	 * want other proc's looking up things while what they are looking
4566	 * up is changing :-D
4567	 */
4568
4569	inp = stcb->sctp_ep;
4570	/* if subset bound and don't allow ASCONF's, can't delete last */
4571	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
4572	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
4573		if (stcb->sctp_ep->laddr_count < 2) {
4574			/* can't delete last address */
4575			return;
4576		}
4577	}
4578	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
4579		/* remove the address if it exists */
4580		if (laddr->ifa == NULL)
4581			continue;
4582		if (laddr->ifa == ifa) {
4583			sctp_remove_laddr(laddr);
4584			return;
4585		}
4586	}
4587
4588	/* address not found! */
4589	return;
4590}
4591
4592static char sctp_pcb_initialized = 0;
4593
4594/*
4595 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
4596 */
4597/* sysctl */
4598static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
4599static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
4600
4601void
4602sctp_pcb_init()
4603{
4604	/*
4605	 * SCTP initialization for the PCB structures should be called by
4606	 * the sctp_init() funciton.
4607	 */
4608	int i;
4609
4610	if (sctp_pcb_initialized != 0) {
4611		/* error I was called twice */
4612		return;
4613	}
4614	sctp_pcb_initialized = 1;
4615
4616	bzero(&sctpstat, sizeof(struct sctpstat));
4617	(void)SCTP_GETTIME_TIMEVAL(&sctpstat.sctps_discontinuitytime);
4618	/* init the empty list of (All) Endpoints */
4619	LIST_INIT(&sctppcbinfo.listhead);
4620
4621	/* init the iterator head */
4622	TAILQ_INIT(&sctppcbinfo.iteratorhead);
4623
4624	/* init the hash table of endpoints */
4625	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &sctp_hashtblsize);
4626	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &sctp_pcbtblsize);
4627	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &sctp_chunkscale);
4628	sctppcbinfo.sctp_asochash = SCTP_HASH_INIT((sctp_hashtblsize * 31),
4629	    &sctppcbinfo.hashasocmark);
4630	sctppcbinfo.sctp_ephash = SCTP_HASH_INIT(sctp_hashtblsize,
4631	    &sctppcbinfo.hashmark);
4632	sctppcbinfo.sctp_tcpephash = SCTP_HASH_INIT(sctp_hashtblsize,
4633	    &sctppcbinfo.hashtcpmark);
4634	sctppcbinfo.hashtblsize = sctp_hashtblsize;
4635
4636	/* init the small hash table we use to track restarted asoc's */
4637	sctppcbinfo.sctp_restarthash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE,
4638	    &sctppcbinfo.hashrestartmark);
4639
4640
4641	sctppcbinfo.sctp_vrfhash = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
4642	    &sctppcbinfo.hashvrfmark);
4643
4644	sctppcbinfo.vrf_ifn_hash = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
4645	    &sctppcbinfo.vrf_ifn_hashmark);
4646
4647	/* init the zones */
4648	/*
4649	 * FIX ME: Should check for NULL returns, but if it does fail we are
4650	 * doomed to panic anyways... add later maybe.
4651	 */
4652	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_ep, "sctp_ep",
4653	    sizeof(struct sctp_inpcb), maxsockets);
4654
4655	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_asoc, "sctp_asoc",
4656	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
4657
4658	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_laddr, "sctp_laddr",
4659	    sizeof(struct sctp_laddr),
4660	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
4661
4662	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_net, "sctp_raddr",
4663	    sizeof(struct sctp_nets),
4664	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
4665
4666	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_chunk, "sctp_chunk",
4667	    sizeof(struct sctp_tmit_chunk),
4668	    (sctp_max_number_of_assoc * sctp_chunkscale));
4669
4670	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_readq, "sctp_readq",
4671	    sizeof(struct sctp_queued_to_read),
4672	    (sctp_max_number_of_assoc * sctp_chunkscale));
4673
4674	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_strmoq, "sctp_stream_msg_out",
4675	    sizeof(struct sctp_stream_queue_pending),
4676	    (sctp_max_number_of_assoc * sctp_chunkscale));
4677
4678	/* Master Lock INIT for info structure */
4679	SCTP_INP_INFO_LOCK_INIT();
4680	SCTP_STATLOG_INIT_LOCK();
4681	SCTP_ITERATOR_LOCK_INIT();
4682
4683	SCTP_IPI_COUNT_INIT();
4684	SCTP_IPI_ADDR_INIT();
4685	SCTP_IPI_ITERATOR_WQ_INIT();
4686#ifdef SCTP_PACKET_LOGGING
4687	SCTP_IP_PKTLOG_INIT();
4688#endif
4689	LIST_INIT(&sctppcbinfo.addr_wq);
4690
4691	/* not sure if we need all the counts */
4692	sctppcbinfo.ipi_count_ep = 0;
4693	/* assoc/tcb zone info */
4694	sctppcbinfo.ipi_count_asoc = 0;
4695	/* local addrlist zone info */
4696	sctppcbinfo.ipi_count_laddr = 0;
4697	/* remote addrlist zone info */
4698	sctppcbinfo.ipi_count_raddr = 0;
4699	/* chunk info */
4700	sctppcbinfo.ipi_count_chunk = 0;
4701
4702	/* socket queue zone info */
4703	sctppcbinfo.ipi_count_readq = 0;
4704
4705	/* stream out queue cont */
4706	sctppcbinfo.ipi_count_strmoq = 0;
4707
4708	sctppcbinfo.ipi_free_strmoq = 0;
4709	sctppcbinfo.ipi_free_chunks = 0;
4710
4711	SCTP_OS_TIMER_INIT(&sctppcbinfo.addr_wq_timer.timer);
4712
4713	/* Init the TIMEWAIT list */
4714	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE_A; i++) {
4715		LIST_INIT(&sctppcbinfo.vtag_timewait[i]);
4716	}
4717
4718#if defined(SCTP_USE_THREAD_BASED_ITERATOR)
4719	sctppcbinfo.iterator_running = 0;
4720	sctp_startup_iterator();
4721#endif
4722
4723	/*
4724	 * INIT the default VRF which for BSD is the only one, other O/S's
4725	 * may have more. But initially they must start with one and then
4726	 * add the VRF's as addresses are added.
4727	 */
4728	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
4729
4730}
4731
4732
4733int
4734sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
4735    int iphlen, int offset, int limit, struct sctphdr *sh,
4736    struct sockaddr *altsa)
4737{
4738	/*
4739	 * grub through the INIT pulling addresses and loading them to the
4740	 * nets structure in the asoc. The from address in the mbuf should
4741	 * also be loaded (if it is not already). This routine can be called
4742	 * with either INIT or INIT-ACK's as long as the m points to the IP
4743	 * packet and the offset points to the beginning of the parameters.
4744	 */
4745	struct sctp_inpcb *inp, *l_inp;
4746	struct sctp_nets *net, *net_tmp;
4747	struct ip *iph;
4748	struct sctp_paramhdr *phdr, parm_buf;
4749	struct sctp_tcb *stcb_tmp;
4750	uint16_t ptype, plen;
4751	struct sockaddr *sa;
4752	struct sockaddr_storage dest_store;
4753	struct sockaddr *local_sa = (struct sockaddr *)&dest_store;
4754	struct sockaddr_in sin;
4755	struct sockaddr_in6 sin6;
4756	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
4757	struct sctp_auth_random *p_random = NULL;
4758	uint16_t random_len = 0;
4759	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
4760	struct sctp_auth_hmac_algo *hmacs = NULL;
4761	uint16_t hmacs_len = 0;
4762	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
4763	struct sctp_auth_chunk_list *chunks = NULL;
4764	uint16_t num_chunks = 0;
4765	sctp_key_t *new_key;
4766	uint32_t keylen;
4767	int got_random = 0, got_hmacs = 0, got_chklist = 0;
4768
4769	/* First get the destination address setup too. */
4770	memset(&sin, 0, sizeof(sin));
4771	memset(&sin6, 0, sizeof(sin6));
4772
4773	sin.sin_family = AF_INET;
4774	sin.sin_len = sizeof(sin);
4775	sin.sin_port = stcb->rport;
4776
4777	sin6.sin6_family = AF_INET6;
4778	sin6.sin6_len = sizeof(struct sockaddr_in6);
4779	sin6.sin6_port = stcb->rport;
4780	if (altsa == NULL) {
4781		iph = mtod(m, struct ip *);
4782		if (iph->ip_v == IPVERSION) {
4783			/* its IPv4 */
4784			struct sockaddr_in *sin_2;
4785
4786			sin_2 = (struct sockaddr_in *)(local_sa);
4787			memset(sin_2, 0, sizeof(sin));
4788			sin_2->sin_family = AF_INET;
4789			sin_2->sin_len = sizeof(sin);
4790			sin_2->sin_port = sh->dest_port;
4791			sin_2->sin_addr.s_addr = iph->ip_dst.s_addr;
4792			sin.sin_addr = iph->ip_src;
4793			sa = (struct sockaddr *)&sin;
4794		} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
4795			/* its IPv6 */
4796			struct ip6_hdr *ip6;
4797			struct sockaddr_in6 *sin6_2;
4798
4799			ip6 = mtod(m, struct ip6_hdr *);
4800			sin6_2 = (struct sockaddr_in6 *)(local_sa);
4801			memset(sin6_2, 0, sizeof(sin6));
4802			sin6_2->sin6_family = AF_INET6;
4803			sin6_2->sin6_len = sizeof(struct sockaddr_in6);
4804			sin6_2->sin6_port = sh->dest_port;
4805			sin6.sin6_addr = ip6->ip6_src;
4806			sa = (struct sockaddr *)&sin6;
4807		} else {
4808			sa = NULL;
4809		}
4810	} else {
4811		/*
4812		 * For cookies we use the src address NOT from the packet
4813		 * but from the original INIT
4814		 */
4815		sa = altsa;
4816	}
4817	/* Turn off ECN until we get through all params */
4818	stcb->asoc.ecn_allowed = 0;
4819	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4820		/* mark all addresses that we have currently on the list */
4821		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
4822	}
4823	/* does the source address already exist? if so skip it */
4824	l_inp = inp = stcb->sctp_ep;
4825
4826	atomic_add_int(&stcb->asoc.refcnt, 1);
4827	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb);
4828	atomic_add_int(&stcb->asoc.refcnt, -1);
4829
4830	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
4831		/* we must add the source address */
4832		/* no scope set here since we have a tcb already. */
4833		if ((sa->sa_family == AF_INET) &&
4834		    (stcb->asoc.ipv4_addr_legal)) {
4835			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
4836				return (-1);
4837			}
4838		} else if ((sa->sa_family == AF_INET6) &&
4839		    (stcb->asoc.ipv6_addr_legal)) {
4840			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
4841				return (-2);
4842			}
4843		}
4844	} else {
4845		if (net_tmp != NULL && stcb_tmp == stcb) {
4846			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
4847		} else if (stcb_tmp != stcb) {
4848			/* It belongs to another association? */
4849			if (stcb_tmp)
4850				SCTP_TCB_UNLOCK(stcb_tmp);
4851			return (-3);
4852		}
4853	}
4854	if (stcb->asoc.state == 0) {
4855		/* the assoc was freed? */
4856		return (-4);
4857	}
4858	/*
4859	 * peer must explicitly turn this on. This may have been initialized
4860	 * to be "on" in order to allow local addr changes while INIT's are
4861	 * in flight.
4862	 */
4863	stcb->asoc.peer_supports_asconf = 0;
4864	/* now we must go through each of the params. */
4865	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
4866	while (phdr) {
4867		ptype = ntohs(phdr->param_type);
4868		plen = ntohs(phdr->param_length);
4869		/*
4870		 * printf("ptype => %0x, plen => %d\n", (uint32_t)ptype,
4871		 * (int)plen);
4872		 */
4873		if (offset + plen > limit) {
4874			break;
4875		}
4876		if (plen == 0) {
4877			break;
4878		}
4879		if (ptype == SCTP_IPV4_ADDRESS) {
4880			if (stcb->asoc.ipv4_addr_legal) {
4881				struct sctp_ipv4addr_param *p4, p4_buf;
4882
4883				/* ok get the v4 address and check/add */
4884				phdr = sctp_get_next_param(m, offset,
4885				    (struct sctp_paramhdr *)&p4_buf, sizeof(p4_buf));
4886				if (plen != sizeof(struct sctp_ipv4addr_param) ||
4887				    phdr == NULL) {
4888					return (-5);
4889				}
4890				p4 = (struct sctp_ipv4addr_param *)phdr;
4891				sin.sin_addr.s_addr = p4->addr;
4892				if (IN_MULTICAST(sin.sin_addr.s_addr)) {
4893					/* Skip multi-cast addresses */
4894					goto next_param;
4895				}
4896				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
4897				    (sin.sin_addr.s_addr == INADDR_ANY)) {
4898					goto next_param;
4899				}
4900				sa = (struct sockaddr *)&sin;
4901				inp = stcb->sctp_ep;
4902				atomic_add_int(&stcb->asoc.refcnt, 1);
4903				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
4904				    local_sa, stcb);
4905				atomic_add_int(&stcb->asoc.refcnt, -1);
4906
4907				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
4908				    inp == NULL) {
4909					/* we must add the source address */
4910					/*
4911					 * no scope set since we have a tcb
4912					 * already
4913					 */
4914
4915					/*
4916					 * we must validate the state again
4917					 * here
4918					 */
4919					if (stcb->asoc.state == 0) {
4920						/* the assoc was freed? */
4921						return (-7);
4922					}
4923					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
4924						return (-8);
4925					}
4926				} else if (stcb_tmp == stcb) {
4927					if (stcb->asoc.state == 0) {
4928						/* the assoc was freed? */
4929						return (-10);
4930					}
4931					if (net != NULL) {
4932						/* clear flag */
4933						net->dest_state &=
4934						    ~SCTP_ADDR_NOT_IN_ASSOC;
4935					}
4936				} else {
4937					/*
4938					 * strange, address is in another
4939					 * assoc? straighten out locks.
4940					 */
4941					if (stcb_tmp)
4942						SCTP_TCB_UNLOCK(stcb_tmp);
4943
4944					if (stcb->asoc.state == 0) {
4945						/* the assoc was freed? */
4946						return (-12);
4947					}
4948					return (-13);
4949				}
4950			}
4951		} else if (ptype == SCTP_IPV6_ADDRESS) {
4952			if (stcb->asoc.ipv6_addr_legal) {
4953				/* ok get the v6 address and check/add */
4954				struct sctp_ipv6addr_param *p6, p6_buf;
4955
4956				phdr = sctp_get_next_param(m, offset,
4957				    (struct sctp_paramhdr *)&p6_buf, sizeof(p6_buf));
4958				if (plen != sizeof(struct sctp_ipv6addr_param) ||
4959				    phdr == NULL) {
4960					return (-14);
4961				}
4962				p6 = (struct sctp_ipv6addr_param *)phdr;
4963				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
4964				    sizeof(p6->addr));
4965				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
4966					/* Skip multi-cast addresses */
4967					goto next_param;
4968				}
4969				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
4970					/*
4971					 * Link local make no sense without
4972					 * scope
4973					 */
4974					goto next_param;
4975				}
4976				sa = (struct sockaddr *)&sin6;
4977				inp = stcb->sctp_ep;
4978				atomic_add_int(&stcb->asoc.refcnt, 1);
4979				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
4980				    local_sa, stcb);
4981				atomic_add_int(&stcb->asoc.refcnt, -1);
4982				if (stcb_tmp == NULL && (inp == stcb->sctp_ep ||
4983				    inp == NULL)) {
4984					/*
4985					 * we must validate the state again
4986					 * here
4987					 */
4988					if (stcb->asoc.state == 0) {
4989						/* the assoc was freed? */
4990						return (-16);
4991					}
4992					/*
4993					 * we must add the address, no scope
4994					 * set
4995					 */
4996					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
4997						return (-17);
4998					}
4999				} else if (stcb_tmp == stcb) {
5000					/*
5001					 * we must validate the state again
5002					 * here
5003					 */
5004					if (stcb->asoc.state == 0) {
5005						/* the assoc was freed? */
5006						return (-19);
5007					}
5008					if (net != NULL) {
5009						/* clear flag */
5010						net->dest_state &=
5011						    ~SCTP_ADDR_NOT_IN_ASSOC;
5012					}
5013				} else {
5014					/*
5015					 * strange, address is in another
5016					 * assoc? straighten out locks.
5017					 */
5018					if (stcb_tmp)
5019						SCTP_TCB_UNLOCK(stcb_tmp);
5020
5021					if (stcb->asoc.state == 0) {
5022						/* the assoc was freed? */
5023						return (-21);
5024					}
5025					return (-22);
5026				}
5027			}
5028		} else if (ptype == SCTP_ECN_CAPABLE) {
5029			stcb->asoc.ecn_allowed = 1;
5030		} else if (ptype == SCTP_ULP_ADAPTATION) {
5031			if (stcb->asoc.state != SCTP_STATE_OPEN) {
5032				struct sctp_adaptation_layer_indication ai,
5033				                                *aip;
5034
5035				phdr = sctp_get_next_param(m, offset,
5036				    (struct sctp_paramhdr *)&ai, sizeof(ai));
5037				aip = (struct sctp_adaptation_layer_indication *)phdr;
5038				if (aip) {
5039					sctp_ulp_notify(SCTP_NOTIFY_ADAPTATION_INDICATION,
5040					    stcb, ntohl(aip->indication), NULL);
5041				}
5042			}
5043		} else if (ptype == SCTP_SET_PRIM_ADDR) {
5044			struct sctp_asconf_addr_param lstore, *fee;
5045			struct sctp_asconf_addrv4_param *fii;
5046			int lptype;
5047			struct sockaddr *lsa = NULL;
5048
5049			stcb->asoc.peer_supports_asconf = 1;
5050			if (plen > sizeof(lstore)) {
5051				return (-23);
5052			}
5053			phdr = sctp_get_next_param(m, offset,
5054			    (struct sctp_paramhdr *)&lstore, min(plen, sizeof(lstore)));
5055			if (phdr == NULL) {
5056				return (-24);
5057			}
5058			fee = (struct sctp_asconf_addr_param *)phdr;
5059			lptype = ntohs(fee->addrp.ph.param_type);
5060			if (lptype == SCTP_IPV4_ADDRESS) {
5061				if (plen !=
5062				    sizeof(struct sctp_asconf_addrv4_param)) {
5063					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
5064					    (int)sizeof(struct sctp_asconf_addrv4_param),
5065					    plen);
5066				} else {
5067					fii = (struct sctp_asconf_addrv4_param *)fee;
5068					sin.sin_addr.s_addr = fii->addrp.addr;
5069					lsa = (struct sockaddr *)&sin;
5070				}
5071			} else if (lptype == SCTP_IPV6_ADDRESS) {
5072				if (plen !=
5073				    sizeof(struct sctp_asconf_addr_param)) {
5074					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
5075					    (int)sizeof(struct sctp_asconf_addr_param),
5076					    plen);
5077				} else {
5078					memcpy(sin6.sin6_addr.s6_addr,
5079					    fee->addrp.addr,
5080					    sizeof(fee->addrp.addr));
5081					lsa = (struct sockaddr *)&sin6;
5082				}
5083			}
5084			if (lsa) {
5085				(void)sctp_set_primary_addr(stcb, sa, NULL);
5086			}
5087		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
5088			/* Peer supports pr-sctp */
5089			stcb->asoc.peer_supports_prsctp = 1;
5090		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
5091			/* A supported extension chunk */
5092			struct sctp_supported_chunk_types_param *pr_supported;
5093			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
5094			int num_ent, i;
5095
5096			phdr = sctp_get_next_param(m, offset,
5097			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
5098			if (phdr == NULL) {
5099				return (-25);
5100			}
5101			stcb->asoc.peer_supports_asconf = 0;
5102			stcb->asoc.peer_supports_prsctp = 0;
5103			stcb->asoc.peer_supports_pktdrop = 0;
5104			stcb->asoc.peer_supports_strreset = 0;
5105			stcb->asoc.peer_supports_auth = 0;
5106			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
5107			num_ent = plen - sizeof(struct sctp_paramhdr);
5108			for (i = 0; i < num_ent; i++) {
5109				switch (pr_supported->chunk_types[i]) {
5110				case SCTP_ASCONF:
5111				case SCTP_ASCONF_ACK:
5112					stcb->asoc.peer_supports_asconf = 1;
5113					break;
5114				case SCTP_FORWARD_CUM_TSN:
5115					stcb->asoc.peer_supports_prsctp = 1;
5116					break;
5117				case SCTP_PACKET_DROPPED:
5118					stcb->asoc.peer_supports_pktdrop = 1;
5119					break;
5120				case SCTP_STREAM_RESET:
5121					stcb->asoc.peer_supports_strreset = 1;
5122					break;
5123				case SCTP_AUTHENTICATION:
5124					stcb->asoc.peer_supports_auth = 1;
5125					break;
5126				default:
5127					/* one I have not learned yet */
5128					break;
5129
5130				}
5131			}
5132		} else if (ptype == SCTP_ECN_NONCE_SUPPORTED) {
5133			/* Peer supports ECN-nonce */
5134			stcb->asoc.peer_supports_ecn_nonce = 1;
5135			stcb->asoc.ecn_nonce_allowed = 1;
5136		} else if (ptype == SCTP_RANDOM) {
5137			if (plen > sizeof(random_store))
5138				break;
5139			if (got_random) {
5140				/* already processed a RANDOM */
5141				goto next_param;
5142			}
5143			phdr = sctp_get_next_param(m, offset,
5144			    (struct sctp_paramhdr *)random_store,
5145			    min(sizeof(random_store), plen));
5146			if (phdr == NULL)
5147				return (-26);
5148			p_random = (struct sctp_auth_random *)phdr;
5149			random_len = plen - sizeof(*p_random);
5150			/* enforce the random length */
5151			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
5152				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
5153				return (-27);
5154			}
5155			got_random = 1;
5156		} else if (ptype == SCTP_HMAC_LIST) {
5157			int num_hmacs;
5158			int i;
5159
5160			if (plen > sizeof(hmacs_store))
5161				break;
5162			if (got_hmacs) {
5163				/* already processed a HMAC list */
5164				goto next_param;
5165			}
5166			phdr = sctp_get_next_param(m, offset,
5167			    (struct sctp_paramhdr *)hmacs_store,
5168			    min(plen, sizeof(hmacs_store)));
5169			if (phdr == NULL)
5170				return (-28);
5171			hmacs = (struct sctp_auth_hmac_algo *)phdr;
5172			hmacs_len = plen - sizeof(*hmacs);
5173			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
5174			/* validate the hmac list */
5175			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
5176				return (-29);
5177			}
5178			if (stcb->asoc.peer_hmacs != NULL)
5179				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
5180			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
5181			if (stcb->asoc.peer_hmacs != NULL) {
5182				for (i = 0; i < num_hmacs; i++) {
5183					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
5184					    ntohs(hmacs->hmac_ids[i]));
5185				}
5186			}
5187			got_hmacs = 1;
5188		} else if (ptype == SCTP_CHUNK_LIST) {
5189			int i;
5190
5191			if (plen > sizeof(chunks_store))
5192				break;
5193			if (got_chklist) {
5194				/* already processed a Chunks list */
5195				goto next_param;
5196			}
5197			phdr = sctp_get_next_param(m, offset,
5198			    (struct sctp_paramhdr *)chunks_store,
5199			    min(plen, sizeof(chunks_store)));
5200			if (phdr == NULL)
5201				return (-30);
5202			chunks = (struct sctp_auth_chunk_list *)phdr;
5203			num_chunks = plen - sizeof(*chunks);
5204			if (stcb->asoc.peer_auth_chunks != NULL)
5205				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
5206			else
5207				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
5208			for (i = 0; i < num_chunks; i++) {
5209				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
5210				    stcb->asoc.peer_auth_chunks);
5211			}
5212			got_chklist = 1;
5213		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
5214			    (ptype == SCTP_STATE_COOKIE) ||
5215			    (ptype == SCTP_UNRECOG_PARAM) ||
5216			    (ptype == SCTP_COOKIE_PRESERVE) ||
5217			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
5218			    (ptype == SCTP_ADD_IP_ADDRESS) ||
5219			    (ptype == SCTP_DEL_IP_ADDRESS) ||
5220			    (ptype == SCTP_ERROR_CAUSE_IND) ||
5221		    (ptype == SCTP_SUCCESS_REPORT)) {
5222			 /* don't care */ ;
5223		} else {
5224			if ((ptype & 0x8000) == 0x0000) {
5225				/*
5226				 * must stop processing the rest of the
5227				 * param's. Any report bits were handled
5228				 * with the call to
5229				 * sctp_arethere_unrecognized_parameters()
5230				 * when the INIT or INIT-ACK was first seen.
5231				 */
5232				break;
5233			}
5234		}
5235next_param:
5236		offset += SCTP_SIZE32(plen);
5237		if (offset >= limit) {
5238			break;
5239		}
5240		phdr = sctp_get_next_param(m, offset, &parm_buf,
5241		    sizeof(parm_buf));
5242	}
5243	/* Now check to see if we need to purge any addresses */
5244	for (net = TAILQ_FIRST(&stcb->asoc.nets); net != NULL; net = net_tmp) {
5245		net_tmp = TAILQ_NEXT(net, sctp_next);
5246		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
5247		    SCTP_ADDR_NOT_IN_ASSOC) {
5248			/* This address has been removed from the asoc */
5249			/* remove and free it */
5250			stcb->asoc.numnets--;
5251			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
5252			sctp_free_remote_addr(net);
5253			if (net == stcb->asoc.primary_destination) {
5254				stcb->asoc.primary_destination = NULL;
5255				sctp_select_primary_destination(stcb);
5256			}
5257		}
5258	}
5259	/* validate authentication required parameters */
5260	if (got_random && got_hmacs) {
5261		stcb->asoc.peer_supports_auth = 1;
5262	} else {
5263		stcb->asoc.peer_supports_auth = 0;
5264	}
5265	if (!stcb->asoc.peer_supports_auth && got_chklist) {
5266		/* peer does not support auth but sent a chunks list? */
5267		return (-31);
5268	}
5269	if (!sctp_asconf_auth_nochk && stcb->asoc.peer_supports_asconf &&
5270	    !stcb->asoc.peer_supports_auth) {
5271		/* peer supports asconf but not auth? */
5272		return (-32);
5273	}
5274	/* concatenate the full random key */
5275#ifdef SCTP_AUTH_DRAFT_04
5276	keylen = random_len;
5277	new_key = sctp_alloc_key(keylen);
5278	if (new_key != NULL) {
5279		/* copy in the RANDOM */
5280		if (p_random != NULL)
5281			bcopy(p_random->random_data, new_key->key, random_len);
5282	}
5283#else
5284	keylen = sizeof(*p_random) + random_len + sizeof(*chunks) + num_chunks +
5285	    sizeof(*hmacs) + hmacs_len;
5286	new_key = sctp_alloc_key(keylen);
5287	if (new_key != NULL) {
5288		/* copy in the RANDOM */
5289		if (p_random != NULL) {
5290			keylen = sizeof(*p_random) + random_len;
5291			bcopy(p_random, new_key->key, keylen);
5292		}
5293		/* append in the AUTH chunks */
5294		if (chunks != NULL) {
5295			bcopy(chunks, new_key->key + keylen,
5296			    sizeof(*chunks) + num_chunks);
5297			keylen += sizeof(*chunks) + num_chunks;
5298		}
5299		/* append in the HMACs */
5300		if (hmacs != NULL) {
5301			bcopy(hmacs, new_key->key + keylen,
5302			    sizeof(*hmacs) + hmacs_len);
5303		}
5304	}
5305#endif
5306	else {
5307		/* failed to get memory for the key */
5308		return (-33);
5309	}
5310	if (stcb->asoc.authinfo.peer_random != NULL)
5311		sctp_free_key(stcb->asoc.authinfo.peer_random);
5312	stcb->asoc.authinfo.peer_random = new_key;
5313#ifdef SCTP_AUTH_DRAFT_04
5314	/* don't include the chunks and hmacs for draft -04 */
5315	stcb->asoc.authinfo.peer_random->keylen = random_len;
5316#endif
5317	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
5318	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
5319
5320	return (0);
5321}
5322
5323int
5324sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
5325    struct sctp_nets *net)
5326{
5327	/* make sure the requested primary address exists in the assoc */
5328	if (net == NULL && sa)
5329		net = sctp_findnet(stcb, sa);
5330
5331	if (net == NULL) {
5332		/* didn't find the requested primary address! */
5333		return (-1);
5334	} else {
5335		/* set the primary address */
5336		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
5337			/* Must be confirmed, so queue to set */
5338			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
5339			return (0);
5340		}
5341		stcb->asoc.primary_destination = net;
5342		net->dest_state &= ~SCTP_ADDR_WAS_PRIMARY;
5343		net = TAILQ_FIRST(&stcb->asoc.nets);
5344		if (net != stcb->asoc.primary_destination) {
5345			/*
5346			 * first one on the list is NOT the primary
5347			 * sctp_cmpaddr() is much more efficent if the
5348			 * primary is the first on the list, make it so.
5349			 */
5350			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
5351			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
5352		}
5353		return (0);
5354	}
5355}
5356
5357
5358int
5359sctp_is_vtag_good(struct sctp_inpcb *inp, uint32_t tag, struct timeval *now)
5360{
5361	/*
5362	 * This function serves two purposes. It will see if a TAG can be
5363	 * re-used and return 1 for yes it is ok and 0 for don't use that
5364	 * tag. A secondary function it will do is purge out old tags that
5365	 * can be removed.
5366	 */
5367	struct sctpasochead *head;
5368	struct sctpvtaghead *chain;
5369	struct sctp_tagblock *twait_block;
5370	struct sctp_tcb *stcb;
5371	int i;
5372
5373	SCTP_INP_INFO_WLOCK();
5374	chain = &sctppcbinfo.vtag_timewait[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
5375	/* First is the vtag in use ? */
5376
5377	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(tag,
5378	    sctppcbinfo.hashasocmark)];
5379	if (head == NULL) {
5380		goto check_restart;
5381	}
5382	LIST_FOREACH(stcb, head, sctp_asocs) {
5383
5384		if (stcb->asoc.my_vtag == tag) {
5385			/*
5386			 * We should remove this if and return 0 always if
5387			 * we want vtags unique across all endpoints. For
5388			 * now within a endpoint is ok.
5389			 */
5390			if (inp == stcb->sctp_ep) {
5391				/* bad tag, in use */
5392				SCTP_INP_INFO_WUNLOCK();
5393				return (0);
5394			}
5395		}
5396	}
5397check_restart:
5398	/* Now lets check the restart hash */
5399	head = &sctppcbinfo.sctp_restarthash[SCTP_PCBHASH_ASOC(tag,
5400	    sctppcbinfo.hashrestartmark)];
5401	if (head == NULL) {
5402		goto check_time_wait;
5403	}
5404	LIST_FOREACH(stcb, head, sctp_tcbrestarhash) {
5405		if (stcb->asoc.assoc_id == tag) {
5406			/* candidate */
5407			if (inp == stcb->sctp_ep) {
5408				/* bad tag, in use */
5409				SCTP_INP_INFO_WUNLOCK();
5410				return (0);
5411			}
5412		}
5413	}
5414check_time_wait:
5415	/* Now what about timed wait ? */
5416	if (!SCTP_LIST_EMPTY(chain)) {
5417		/*
5418		 * Block(s) are present, lets see if we have this tag in the
5419		 * list
5420		 */
5421		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5422			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
5423				if (twait_block->vtag_block[i].v_tag == 0) {
5424					/* not used */
5425					continue;
5426				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire >
5427				    now->tv_sec) {
5428					/* Audit expires this guy */
5429					twait_block->vtag_block[i].tv_sec_at_expire = 0;
5430					twait_block->vtag_block[i].v_tag = 0;
5431				} else if (twait_block->vtag_block[i].v_tag ==
5432				    tag) {
5433					/* Bad tag, sorry :< */
5434					SCTP_INP_INFO_WUNLOCK();
5435					return (0);
5436				}
5437			}
5438		}
5439	}
5440	/* Not found, ok to use the tag */
5441	SCTP_INP_INFO_WUNLOCK();
5442	return (1);
5443}
5444
5445
5446static sctp_assoc_t reneged_asoc_ids[256];
5447static uint8_t reneged_at = 0;
5448
5449
5450static void
5451sctp_drain_mbufs(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
5452{
5453	/*
5454	 * We must hunt this association for MBUF's past the cumack (i.e.
5455	 * out of order data that we can renege on).
5456	 */
5457	struct sctp_association *asoc;
5458	struct sctp_tmit_chunk *chk, *nchk;
5459	uint32_t cumulative_tsn_p1, tsn;
5460	struct sctp_queued_to_read *ctl, *nctl;
5461	int cnt, strmat, gap;
5462
5463	/* We look for anything larger than the cum-ack + 1 */
5464
5465	SCTP_STAT_INCR(sctps_protocol_drain_calls);
5466	if (sctp_do_drain == 0) {
5467		return;
5468	}
5469	asoc = &stcb->asoc;
5470	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
5471		/* none we can reneg on. */
5472		return;
5473	}
5474	SCTP_STAT_INCR(sctps_protocol_drains_done);
5475	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
5476	cnt = 0;
5477	/* First look in the re-assembly queue */
5478	chk = TAILQ_FIRST(&asoc->reasmqueue);
5479	while (chk) {
5480		/* Get the next one */
5481		nchk = TAILQ_NEXT(chk, sctp_next);
5482		if (compare_with_wrap(chk->rec.data.TSN_seq,
5483		    cumulative_tsn_p1, MAX_TSN)) {
5484			/* Yep it is above cum-ack */
5485			cnt++;
5486			tsn = chk->rec.data.TSN_seq;
5487			if (tsn >= asoc->mapping_array_base_tsn) {
5488				gap = tsn - asoc->mapping_array_base_tsn;
5489			} else {
5490				gap = (MAX_TSN - asoc->mapping_array_base_tsn) +
5491				    tsn + 1;
5492			}
5493			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
5494			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
5495			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
5496			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5497			if (chk->data) {
5498				sctp_m_freem(chk->data);
5499				chk->data = NULL;
5500			}
5501			sctp_free_a_chunk(stcb, chk);
5502		}
5503		chk = nchk;
5504	}
5505	/* Ok that was fun, now we will drain all the inbound streams? */
5506	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
5507		ctl = TAILQ_FIRST(&asoc->strmin[strmat].inqueue);
5508		while (ctl) {
5509			nctl = TAILQ_NEXT(ctl, next);
5510			if (compare_with_wrap(ctl->sinfo_tsn,
5511			    cumulative_tsn_p1, MAX_TSN)) {
5512				/* Yep it is above cum-ack */
5513				cnt++;
5514				tsn = ctl->sinfo_tsn;
5515				if (tsn >= asoc->mapping_array_base_tsn) {
5516					gap = tsn -
5517					    asoc->mapping_array_base_tsn;
5518				} else {
5519					gap = (MAX_TSN -
5520					    asoc->mapping_array_base_tsn) +
5521					    tsn + 1;
5522				}
5523				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
5524				sctp_ucount_decr(asoc->cnt_on_all_streams);
5525
5526				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array,
5527				    gap);
5528				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue,
5529				    ctl, next);
5530				if (ctl->data) {
5531					sctp_m_freem(ctl->data);
5532					ctl->data = NULL;
5533				}
5534				sctp_free_remote_addr(ctl->whoFrom);
5535				SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, ctl);
5536				SCTP_DECR_READQ_COUNT();
5537			}
5538			ctl = nctl;
5539		}
5540	}
5541	/*
5542	 * Question, should we go through the delivery queue? The only
5543	 * reason things are on here is the app not reading OR a p-d-api up.
5544	 * An attacker COULD send enough in to initiate the PD-API and then
5545	 * send a bunch of stuff to other streams... these would wind up on
5546	 * the delivery queue.. and then we would not get to them. But in
5547	 * order to do this I then have to back-track and un-deliver
5548	 * sequence numbers in streams.. el-yucko. I think for now we will
5549	 * NOT look at the delivery queue and leave it to be something to
5550	 * consider later. An alternative would be to abort the P-D-API with
5551	 * a notification and then deliver the data.... Or another method
5552	 * might be to keep track of how many times the situation occurs and
5553	 * if we see a possible attack underway just abort the association.
5554	 */
5555#ifdef SCTP_DEBUG
5556	if (cnt) {
5557		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
5558	}
5559#endif
5560	if (cnt) {
5561		/*
5562		 * Now do we need to find a new
5563		 * asoc->highest_tsn_inside_map?
5564		 */
5565		if (asoc->highest_tsn_inside_map >= asoc->mapping_array_base_tsn) {
5566			gap = asoc->highest_tsn_inside_map - asoc->mapping_array_base_tsn;
5567		} else {
5568			gap = (MAX_TSN - asoc->mapping_array_base_tsn) +
5569			    asoc->highest_tsn_inside_map + 1;
5570		}
5571		if (gap >= (asoc->mapping_array_size << 3)) {
5572			/*
5573			 * Something bad happened or cum-ack and high were
5574			 * behind the base, but if so earlier checks should
5575			 * have found NO data... wierd... we will start at
5576			 * end of mapping array.
5577			 */
5578			SCTP_PRINTF("Gap was larger than array?? %d set to max:%d maparraymax:%x\n",
5579			    (int)gap,
5580			    (int)(asoc->mapping_array_size << 3),
5581			    (int)asoc->highest_tsn_inside_map);
5582			gap = asoc->mapping_array_size << 3;
5583		}
5584		while (gap > 0) {
5585			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
5586				/* found the new highest */
5587				asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn + gap;
5588				break;
5589			}
5590			gap--;
5591		}
5592		if (gap == 0) {
5593			/* Nothing left in map */
5594			memset(asoc->mapping_array, 0, asoc->mapping_array_size);
5595			asoc->mapping_array_base_tsn = asoc->cumulative_tsn + 1;
5596			asoc->highest_tsn_inside_map = asoc->cumulative_tsn;
5597		}
5598		asoc->last_revoke_count = cnt;
5599		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
5600		sctp_send_sack(stcb);
5601		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN);
5602		reneged_asoc_ids[reneged_at] = sctp_get_associd(stcb);
5603		reneged_at++;
5604	}
5605	/*
5606	 * Another issue, in un-setting the TSN's in the mapping array we
5607	 * DID NOT adjust the higest_tsn marker.  This will cause one of two
5608	 * things to occur. It may cause us to do extra work in checking for
5609	 * our mapping array movement. More importantly it may cause us to
5610	 * SACK every datagram. This may not be a bad thing though since we
5611	 * will recover once we get our cum-ack above and all this stuff we
5612	 * dumped recovered.
5613	 */
5614}
5615
5616void
5617sctp_drain()
5618{
5619	/*
5620	 * We must walk the PCB lists for ALL associations here. The system
5621	 * is LOW on MBUF's and needs help. This is where reneging will
5622	 * occur. We really hope this does NOT happen!
5623	 */
5624	struct sctp_inpcb *inp;
5625	struct sctp_tcb *stcb;
5626
5627	SCTP_INP_INFO_RLOCK();
5628	LIST_FOREACH(inp, &sctppcbinfo.listhead, sctp_list) {
5629		/* For each endpoint */
5630		SCTP_INP_RLOCK(inp);
5631		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5632			/* For each association */
5633			SCTP_TCB_LOCK(stcb);
5634			sctp_drain_mbufs(inp, stcb);
5635			SCTP_TCB_UNLOCK(stcb);
5636		}
5637		SCTP_INP_RUNLOCK(inp);
5638	}
5639	SCTP_INP_INFO_RUNLOCK();
5640}
5641
5642/*
5643 * start a new iterator
5644 * iterates through all endpoints and associations based on the pcb_state
5645 * flags and asoc_state.  "af" (mandatory) is executed for all matching
5646 * assocs and "ef" (optional) is executed when the iterator completes.
5647 * "inpf" (optional) is executed for each new endpoint as it is being
5648 * iterated through. inpe (optional) is called when the inp completes
5649 * its way through all the stcbs.
5650 */
5651int
5652sctp_initiate_iterator(inp_func inpf,
5653    asoc_func af,
5654    inp_func inpe,
5655    uint32_t pcb_state,
5656    uint32_t pcb_features,
5657    uint32_t asoc_state,
5658    void *argp,
5659    uint32_t argi,
5660    end_func ef,
5661    struct sctp_inpcb *s_inp,
5662    uint8_t chunk_output_off)
5663{
5664	struct sctp_iterator *it = NULL;
5665
5666	if (af == NULL) {
5667		return (-1);
5668	}
5669	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
5670	    SCTP_M_ITER);
5671	if (it == NULL) {
5672		return (ENOMEM);
5673	}
5674	memset(it, 0, sizeof(*it));
5675	it->function_assoc = af;
5676	it->function_inp = inpf;
5677	if (inpf)
5678		it->done_current_ep = 0;
5679	else
5680		it->done_current_ep = 1;
5681	it->function_atend = ef;
5682	it->pointer = argp;
5683	it->val = argi;
5684	it->pcb_flags = pcb_state;
5685	it->pcb_features = pcb_features;
5686	it->asoc_state = asoc_state;
5687	it->function_inp_end = inpe;
5688	it->no_chunk_output = chunk_output_off;
5689	if (s_inp) {
5690		it->inp = s_inp;
5691		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
5692	} else {
5693		SCTP_INP_INFO_RLOCK();
5694		it->inp = LIST_FIRST(&sctppcbinfo.listhead);
5695
5696		SCTP_INP_INFO_RUNLOCK();
5697		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
5698
5699	}
5700	SCTP_IPI_ITERATOR_WQ_LOCK();
5701	if (it->inp) {
5702		SCTP_INP_INCR_REF(it->inp);
5703	}
5704	TAILQ_INSERT_TAIL(&sctppcbinfo.iteratorhead, it, sctp_nxt_itr);
5705#if defined(SCTP_USE_THREAD_BASED_ITERATOR)
5706	if (sctppcbinfo.iterator_running == 0) {
5707		sctp_wakeup_iterator();
5708	}
5709	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5710#else
5711	if (it->inp)
5712		SCTP_INP_DECR_REF(it->inp);
5713	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5714	/* Init the timer */
5715	SCTP_OS_TIMER_INIT(&it->tmr.timer);
5716	/* add to the list of all iterators */
5717	sctp_timer_start(SCTP_TIMER_TYPE_ITERATOR, (struct sctp_inpcb *)it,
5718	    NULL, NULL);
5719#endif
5720	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
5721	return (0);
5722}
5723