sctp_pcb.c revision 283732
1/*-
2 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * a) Redistributions of source code must retain the above copyright notice,
10 *    this list of conditions and the following disclaimer.
11 *
12 * b) Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *    the documentation and/or other materials provided with the distribution.
15 *
16 * c) Neither the name of Cisco Systems, Inc. nor the names of its
17 *    contributors may be used to endorse or promote products derived
18 *    from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33#include <sys/cdefs.h>
34__FBSDID("$FreeBSD: stable/10/sys/netinet/sctp_pcb.c 283732 2015-05-29 13:18:10Z tuexen $");
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#include <netinet/sctp_dtrace_define.h>
49#if defined(INET) || defined(INET6)
50#include <netinet/udp.h>
51#endif
52#ifdef INET6
53#include <netinet6/ip6_var.h>
54#endif
55#include <sys/sched.h>
56#include <sys/smp.h>
57#include <sys/unistd.h>
58
59
60VNET_DEFINE(struct sctp_base_info, system_base_info);
61
62/* FIX: we don't handle multiple link local scopes */
63/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
64#ifdef INET6
65int
66SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
67{
68	struct sockaddr_in6 tmp_a, tmp_b;
69
70	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
71	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
72		return (0);
73	}
74	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
75	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
76		return (0);
77	}
78	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
79}
80
81#endif
82
83void
84sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
85{
86	/*
87	 * We really don't need to lock this, but I will just because it
88	 * does not hurt.
89	 */
90	SCTP_INP_INFO_RLOCK();
91	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
92	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
93	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
94	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
95	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
96	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
97	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
98	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
99	SCTP_INP_INFO_RUNLOCK();
100}
101
102/*-
103 * Addresses are added to VRF's (Virtual Router's). For BSD we
104 * have only the default VRF 0. We maintain a hash list of
105 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
106 * these has a list of addresses. When we add a new address
107 * to a VRF we lookup the ifn/ifn_index, if the ifn does
108 * not exist we create it and add it to the list of IFN's
109 * within the VRF. Once we have the sctp_ifn, we add the
110 * address to the list. So we look something like:
111 *
112 * hash-vrf-table
113 *   vrf-> ifn-> ifn -> ifn
114 *   vrf    |
115 *    ...   +--ifa-> ifa -> ifa
116 *   vrf
117 *
118 * We keep these separate lists since the SCTP subsystem will
119 * point to these from its source address selection nets structure.
120 * When an address is deleted it does not happen right away on
121 * the SCTP side, it gets scheduled. What we do when a
122 * delete happens is immediately remove the address from
123 * the master list and decrement the refcount. As our
124 * addip iterator works through and frees the src address
125 * selection pointing to the sctp_ifa, eventually the refcount
126 * will reach 0 and we will delete it. Note that it is assumed
127 * that any locking on system level ifn/ifa is done at the
128 * caller of these functions and these routines will only
129 * lock the SCTP structures as they add or delete things.
130 *
131 * Other notes on VRF concepts.
132 *  - An endpoint can be in multiple VRF's
133 *  - An association lives within a VRF and only one VRF.
134 *  - Any incoming packet we can deduce the VRF for by
135 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
136 *  - Any downward send call or connect call must supply the
137 *    VRF via ancillary data or via some sort of set default
138 *    VRF socket option call (again for BSD no brainer since
139 *    the VRF is always 0).
140 *  - An endpoint may add multiple VRF's to it.
141 *  - Listening sockets can accept associations in any
142 *    of the VRF's they are in but the assoc will end up
143 *    in only one VRF (gotten from the packet or connect/send).
144 *
145 */
146
147struct sctp_vrf *
148sctp_allocate_vrf(int vrf_id)
149{
150	struct sctp_vrf *vrf = NULL;
151	struct sctp_vrflist *bucket;
152
153	/* First allocate the VRF structure */
154	vrf = sctp_find_vrf(vrf_id);
155	if (vrf) {
156		/* Already allocated */
157		return (vrf);
158	}
159	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
160	    SCTP_M_VRF);
161	if (vrf == NULL) {
162		/* No memory */
163#ifdef INVARIANTS
164		panic("No memory for VRF:%d", vrf_id);
165#endif
166		return (NULL);
167	}
168	/* setup the VRF */
169	memset(vrf, 0, sizeof(struct sctp_vrf));
170	vrf->vrf_id = vrf_id;
171	LIST_INIT(&vrf->ifnlist);
172	vrf->total_ifa_count = 0;
173	vrf->refcount = 0;
174	/* now also setup table ids */
175	SCTP_INIT_VRF_TABLEID(vrf);
176	/* Init the HASH of addresses */
177	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
178	    &vrf->vrf_addr_hashmark);
179	if (vrf->vrf_addr_hash == NULL) {
180		/* No memory */
181#ifdef INVARIANTS
182		panic("No memory for VRF:%d", vrf_id);
183#endif
184		SCTP_FREE(vrf, SCTP_M_VRF);
185		return (NULL);
186	}
187	/* Add it to the hash table */
188	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
189	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
190	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
191	return (vrf);
192}
193
194
195struct sctp_ifn *
196sctp_find_ifn(void *ifn, uint32_t ifn_index)
197{
198	struct sctp_ifn *sctp_ifnp;
199	struct sctp_ifnlist *hash_ifn_head;
200
201	/*
202	 * We assume the lock is held for the addresses if that's wrong
203	 * problems could occur :-)
204	 */
205	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
206	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
207		if (sctp_ifnp->ifn_index == ifn_index) {
208			return (sctp_ifnp);
209		}
210		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
211			return (sctp_ifnp);
212		}
213	}
214	return (NULL);
215}
216
217
218struct sctp_vrf *
219sctp_find_vrf(uint32_t vrf_id)
220{
221	struct sctp_vrflist *bucket;
222	struct sctp_vrf *liste;
223
224	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
225	LIST_FOREACH(liste, bucket, next_vrf) {
226		if (vrf_id == liste->vrf_id) {
227			return (liste);
228		}
229	}
230	return (NULL);
231}
232
233
234void
235sctp_free_vrf(struct sctp_vrf *vrf)
236{
237	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
238		if (vrf->vrf_addr_hash) {
239			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
240			vrf->vrf_addr_hash = NULL;
241		}
242		/* We zero'd the count */
243		LIST_REMOVE(vrf, next_vrf);
244		SCTP_FREE(vrf, SCTP_M_VRF);
245		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
246	}
247}
248
249
250void
251sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
252{
253	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
254		/* We zero'd the count */
255		if (sctp_ifnp->vrf) {
256			sctp_free_vrf(sctp_ifnp->vrf);
257		}
258		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
259		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
260	}
261}
262
263
264void
265sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
266{
267	struct sctp_ifn *sctp_ifnp;
268
269	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
270	if (sctp_ifnp != NULL) {
271		sctp_ifnp->ifn_mtu = mtu;
272	}
273}
274
275
276void
277sctp_free_ifa(struct sctp_ifa *sctp_ifap)
278{
279	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
280		/* We zero'd the count */
281		if (sctp_ifap->ifn_p) {
282			sctp_free_ifn(sctp_ifap->ifn_p);
283		}
284		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
285		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
286	}
287}
288
289
290static void
291sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
292{
293	struct sctp_ifn *found;
294
295	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
296	if (found == NULL) {
297		/* Not in the list.. sorry */
298		return;
299	}
300	if (hold_addr_lock == 0)
301		SCTP_IPI_ADDR_WLOCK();
302	LIST_REMOVE(sctp_ifnp, next_bucket);
303	LIST_REMOVE(sctp_ifnp, next_ifn);
304	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
305	    sctp_ifnp->registered_af);
306	if (hold_addr_lock == 0)
307		SCTP_IPI_ADDR_WUNLOCK();
308	/* Take away the reference, and possibly free it */
309	sctp_free_ifn(sctp_ifnp);
310}
311
312
313void
314sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
315    const char *if_name, uint32_t ifn_index)
316{
317	struct sctp_vrf *vrf;
318	struct sctp_ifa *sctp_ifap;
319
320	SCTP_IPI_ADDR_RLOCK();
321	vrf = sctp_find_vrf(vrf_id);
322	if (vrf == NULL) {
323		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
324		goto out;
325
326	}
327	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
328	if (sctp_ifap == NULL) {
329		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
330		goto out;
331	}
332	if (sctp_ifap->ifn_p == NULL) {
333		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
334		goto out;
335	}
336	if (if_name) {
337		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
338			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
339			    sctp_ifap->ifn_p->ifn_name, if_name);
340			goto out;
341		}
342	} else {
343		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
344			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
345			    sctp_ifap->ifn_p->ifn_index, ifn_index);
346			goto out;
347		}
348	}
349
350	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
351	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
352out:
353	SCTP_IPI_ADDR_RUNLOCK();
354}
355
356
357void
358sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
359    const char *if_name, uint32_t ifn_index)
360{
361	struct sctp_vrf *vrf;
362	struct sctp_ifa *sctp_ifap;
363
364	SCTP_IPI_ADDR_RLOCK();
365	vrf = sctp_find_vrf(vrf_id);
366	if (vrf == NULL) {
367		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
368		goto out;
369
370	}
371	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
372	if (sctp_ifap == NULL) {
373		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
374		goto out;
375	}
376	if (sctp_ifap->ifn_p == NULL) {
377		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
378		goto out;
379	}
380	if (if_name) {
381		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
382			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
383			    sctp_ifap->ifn_p->ifn_name, if_name);
384			goto out;
385		}
386	} else {
387		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
388			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
389			    sctp_ifap->ifn_p->ifn_index, ifn_index);
390			goto out;
391		}
392	}
393
394	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
395	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
396out:
397	SCTP_IPI_ADDR_RUNLOCK();
398}
399
400
401/*-
402 * Add an ifa to an ifn.
403 * Register the interface as necessary.
404 * NOTE: ADDR write lock MUST be held.
405 */
406static void
407sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
408{
409	int ifa_af;
410
411	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
412	sctp_ifap->ifn_p = sctp_ifnp;
413	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
414	/* update address counts */
415	sctp_ifnp->ifa_count++;
416	ifa_af = sctp_ifap->address.sa.sa_family;
417	switch (ifa_af) {
418#ifdef INET
419	case AF_INET:
420		sctp_ifnp->num_v4++;
421		break;
422#endif
423#ifdef INET6
424	case AF_INET6:
425		sctp_ifnp->num_v6++;
426		break;
427#endif
428	default:
429		break;
430	}
431	if (sctp_ifnp->ifa_count == 1) {
432		/* register the new interface */
433		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
434		sctp_ifnp->registered_af = ifa_af;
435	}
436}
437
438
439/*-
440 * Remove an ifa from its ifn.
441 * If no more addresses exist, remove the ifn too. Otherwise, re-register
442 * the interface based on the remaining address families left.
443 * NOTE: ADDR write lock MUST be held.
444 */
445static void
446sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
447{
448	LIST_REMOVE(sctp_ifap, next_ifa);
449	if (sctp_ifap->ifn_p) {
450		/* update address counts */
451		sctp_ifap->ifn_p->ifa_count--;
452		switch (sctp_ifap->address.sa.sa_family) {
453#ifdef INET
454		case AF_INET:
455			sctp_ifap->ifn_p->num_v4--;
456			break;
457#endif
458#ifdef INET6
459		case AF_INET6:
460			sctp_ifap->ifn_p->num_v6--;
461			break;
462#endif
463		default:
464			break;
465		}
466
467		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
468			/* remove the ifn, possibly freeing it */
469			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
470		} else {
471			/* re-register address family type, if needed */
472			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
473			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
474				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
475				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
476				sctp_ifap->ifn_p->registered_af = AF_INET;
477			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
478			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
479				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
480				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
481				sctp_ifap->ifn_p->registered_af = AF_INET6;
482			}
483			/* free the ifn refcount */
484			sctp_free_ifn(sctp_ifap->ifn_p);
485		}
486		sctp_ifap->ifn_p = NULL;
487	}
488}
489
490
491struct sctp_ifa *
492sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
493    uint32_t ifn_type, const char *if_name, void *ifa,
494    struct sockaddr *addr, uint32_t ifa_flags,
495    int dynamic_add)
496{
497	struct sctp_vrf *vrf;
498	struct sctp_ifn *sctp_ifnp = NULL;
499	struct sctp_ifa *sctp_ifap = NULL;
500	struct sctp_ifalist *hash_addr_head;
501	struct sctp_ifnlist *hash_ifn_head;
502	uint32_t hash_of_addr;
503	int new_ifn_af = 0;
504
505#ifdef SCTP_DEBUG
506	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
507	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
508#endif
509	SCTP_IPI_ADDR_WLOCK();
510	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
511	if (sctp_ifnp) {
512		vrf = sctp_ifnp->vrf;
513	} else {
514		vrf = sctp_find_vrf(vrf_id);
515		if (vrf == NULL) {
516			vrf = sctp_allocate_vrf(vrf_id);
517			if (vrf == NULL) {
518				SCTP_IPI_ADDR_WUNLOCK();
519				return (NULL);
520			}
521		}
522	}
523	if (sctp_ifnp == NULL) {
524		/*
525		 * build one and add it, can't hold lock until after malloc
526		 * done though.
527		 */
528		SCTP_IPI_ADDR_WUNLOCK();
529		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
530		    sizeof(struct sctp_ifn), SCTP_M_IFN);
531		if (sctp_ifnp == NULL) {
532#ifdef INVARIANTS
533			panic("No memory for IFN");
534#endif
535			return (NULL);
536		}
537		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
538		sctp_ifnp->ifn_index = ifn_index;
539		sctp_ifnp->ifn_p = ifn;
540		sctp_ifnp->ifn_type = ifn_type;
541		sctp_ifnp->refcount = 0;
542		sctp_ifnp->vrf = vrf;
543		atomic_add_int(&vrf->refcount, 1);
544		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
545		if (if_name != NULL) {
546			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
547		} else {
548			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
549		}
550		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
551		LIST_INIT(&sctp_ifnp->ifalist);
552		SCTP_IPI_ADDR_WLOCK();
553		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
554		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
555		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
556		new_ifn_af = 1;
557	}
558	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
559	if (sctp_ifap) {
560		/* Hmm, it already exists? */
561		if ((sctp_ifap->ifn_p) &&
562		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
563			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
564			    sctp_ifap->ifn_p->ifn_name, ifn_index,
565			    (void *)sctp_ifap);
566			if (new_ifn_af) {
567				/* Remove the created one that we don't want */
568				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
569			}
570			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
571				/* easy to solve, just switch back to active */
572				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
573				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
574				sctp_ifap->ifn_p = sctp_ifnp;
575				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
576			}
577	exit_stage_left:
578			SCTP_IPI_ADDR_WUNLOCK();
579			return (sctp_ifap);
580		} else {
581			if (sctp_ifap->ifn_p) {
582				/*
583				 * The last IFN gets the address, remove the
584				 * old one
585				 */
586				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
587				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
588				    sctp_ifap->ifn_p->ifn_index, if_name,
589				    ifn_index);
590				/* remove the address from the old ifn */
591				sctp_remove_ifa_from_ifn(sctp_ifap);
592				/* move the address over to the new ifn */
593				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
594				goto exit_stage_left;
595			} else {
596				/* repair ifnp which was NULL ? */
597				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
598				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
599				    (void *)sctp_ifnp, (void *)sctp_ifap);
600				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
601			}
602			goto exit_stage_left;
603		}
604	}
605	SCTP_IPI_ADDR_WUNLOCK();
606	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
607	if (sctp_ifap == NULL) {
608#ifdef INVARIANTS
609		panic("No memory for IFA");
610#endif
611		return (NULL);
612	}
613	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
614	sctp_ifap->ifn_p = sctp_ifnp;
615	atomic_add_int(&sctp_ifnp->refcount, 1);
616	sctp_ifap->vrf_id = vrf_id;
617	sctp_ifap->ifa = ifa;
618	memcpy(&sctp_ifap->address, addr, addr->sa_len);
619	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
620	sctp_ifap->flags = ifa_flags;
621	/* Set scope */
622	switch (sctp_ifap->address.sa.sa_family) {
623#ifdef INET
624	case AF_INET:
625		{
626			struct sockaddr_in *sin;
627
628			sin = &sctp_ifap->address.sin;
629			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
630			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
631				sctp_ifap->src_is_loop = 1;
632			}
633			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
634				sctp_ifap->src_is_priv = 1;
635			}
636			sctp_ifnp->num_v4++;
637			if (new_ifn_af)
638				new_ifn_af = AF_INET;
639			break;
640		}
641#endif
642#ifdef INET6
643	case AF_INET6:
644		{
645			/* ok to use deprecated addresses? */
646			struct sockaddr_in6 *sin6;
647
648			sin6 = &sctp_ifap->address.sin6;
649			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
650			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
651				sctp_ifap->src_is_loop = 1;
652			}
653			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
654				sctp_ifap->src_is_priv = 1;
655			}
656			sctp_ifnp->num_v6++;
657			if (new_ifn_af)
658				new_ifn_af = AF_INET6;
659			break;
660		}
661#endif
662	default:
663		new_ifn_af = 0;
664		break;
665	}
666	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
667
668	if ((sctp_ifap->src_is_priv == 0) &&
669	    (sctp_ifap->src_is_loop == 0)) {
670		sctp_ifap->src_is_glob = 1;
671	}
672	SCTP_IPI_ADDR_WLOCK();
673	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
674	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
675	sctp_ifap->refcount = 1;
676	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
677	sctp_ifnp->ifa_count++;
678	vrf->total_ifa_count++;
679	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
680	if (new_ifn_af) {
681		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
682		sctp_ifnp->registered_af = new_ifn_af;
683	}
684	SCTP_IPI_ADDR_WUNLOCK();
685	if (dynamic_add) {
686		/*
687		 * Bump up the refcount so that when the timer completes it
688		 * will drop back down.
689		 */
690		struct sctp_laddr *wi;
691
692		atomic_add_int(&sctp_ifap->refcount, 1);
693		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
694		if (wi == NULL) {
695			/*
696			 * Gak, what can we do? We have lost an address
697			 * change can you say HOSED?
698			 */
699			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
700			/* Opps, must decrement the count */
701			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
702			    if_name);
703			return (NULL);
704		}
705		SCTP_INCR_LADDR_COUNT();
706		bzero(wi, sizeof(*wi));
707		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
708		wi->ifa = sctp_ifap;
709		wi->action = SCTP_ADD_IP_ADDRESS;
710
711		SCTP_WQ_ADDR_LOCK();
712		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
713		SCTP_WQ_ADDR_UNLOCK();
714
715		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
716		    (struct sctp_inpcb *)NULL,
717		    (struct sctp_tcb *)NULL,
718		    (struct sctp_nets *)NULL);
719	} else {
720		/* it's ready for use */
721		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
722	}
723	return (sctp_ifap);
724}
725
726void
727sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
728    uint32_t ifn_index, const char *if_name)
729{
730	struct sctp_vrf *vrf;
731	struct sctp_ifa *sctp_ifap = NULL;
732
733	SCTP_IPI_ADDR_WLOCK();
734	vrf = sctp_find_vrf(vrf_id);
735	if (vrf == NULL) {
736		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
737		goto out_now;
738	}
739#ifdef SCTP_DEBUG
740	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
741	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
742#endif
743	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
744	if (sctp_ifap) {
745		/* Validate the delete */
746		if (sctp_ifap->ifn_p) {
747			int valid = 0;
748
749			/*-
750			 * The name has priority over the ifn_index
751			 * if its given. We do this especially for
752			 * panda who might recycle indexes fast.
753			 */
754			if (if_name) {
755				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
756					/* They match its a correct delete */
757					valid = 1;
758				}
759			}
760			if (!valid) {
761				/* last ditch check ifn_index */
762				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
763					valid = 1;
764				}
765			}
766			if (!valid) {
767				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
768				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
769				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
770				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
771				SCTP_IPI_ADDR_WUNLOCK();
772				return;
773			}
774		}
775		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
776		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
777		/*
778		 * We don't set the flag. This means that the structure will
779		 * hang around in EP's that have bound specific to it until
780		 * they close. This gives us TCP like behavior if someone
781		 * removes an address (or for that matter adds it right
782		 * back).
783		 */
784		/* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */
785		vrf->total_ifa_count--;
786		LIST_REMOVE(sctp_ifap, next_bucket);
787		sctp_remove_ifa_from_ifn(sctp_ifap);
788	}
789#ifdef SCTP_DEBUG
790	else {
791		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
792		    ifn_index);
793		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
794	}
795#endif
796
797out_now:
798	SCTP_IPI_ADDR_WUNLOCK();
799	if (sctp_ifap) {
800		struct sctp_laddr *wi;
801
802		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
803		if (wi == NULL) {
804			/*
805			 * Gak, what can we do? We have lost an address
806			 * change can you say HOSED?
807			 */
808			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
809
810			/* Oops, must decrement the count */
811			sctp_free_ifa(sctp_ifap);
812			return;
813		}
814		SCTP_INCR_LADDR_COUNT();
815		bzero(wi, sizeof(*wi));
816		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
817		wi->ifa = sctp_ifap;
818		wi->action = SCTP_DEL_IP_ADDRESS;
819		SCTP_WQ_ADDR_LOCK();
820		/*
821		 * Should this really be a tailq? As it is we will process
822		 * the newest first :-0
823		 */
824		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
825		SCTP_WQ_ADDR_UNLOCK();
826
827		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
828		    (struct sctp_inpcb *)NULL,
829		    (struct sctp_tcb *)NULL,
830		    (struct sctp_nets *)NULL);
831	}
832	return;
833}
834
835
836static int
837sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
838{
839	int loopback_scope;
840
841#if defined(INET)
842	int ipv4_local_scope, ipv4_addr_legal;
843
844#endif
845#if defined(INET6)
846	int local_scope, site_scope, ipv6_addr_legal;
847
848#endif
849	struct sctp_vrf *vrf;
850	struct sctp_ifn *sctp_ifn;
851	struct sctp_ifa *sctp_ifa;
852
853	loopback_scope = stcb->asoc.scope.loopback_scope;
854#if defined(INET)
855	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
856	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
857#endif
858#if defined(INET6)
859	local_scope = stcb->asoc.scope.local_scope;
860	site_scope = stcb->asoc.scope.site_scope;
861	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
862#endif
863
864	SCTP_IPI_ADDR_RLOCK();
865	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
866	if (vrf == NULL) {
867		/* no vrf, no addresses */
868		SCTP_IPI_ADDR_RUNLOCK();
869		return (0);
870	}
871	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
872		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
873			if ((loopback_scope == 0) &&
874			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
875				continue;
876			}
877			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
878				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
879				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
880					/*
881					 * We allow pending addresses, where
882					 * we have sent an asconf-add to be
883					 * considered valid.
884					 */
885					continue;
886				}
887				if (sctp_ifa->address.sa.sa_family != to->sa_family) {
888					continue;
889				}
890				switch (sctp_ifa->address.sa.sa_family) {
891#ifdef INET
892				case AF_INET:
893					if (ipv4_addr_legal) {
894						struct sockaddr_in *sin,
895						           *rsin;
896
897						sin = &sctp_ifa->address.sin;
898						rsin = (struct sockaddr_in *)to;
899						if ((ipv4_local_scope == 0) &&
900						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
901							continue;
902						}
903						if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred,
904						    &sin->sin_addr) != 0) {
905							continue;
906						}
907						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
908							SCTP_IPI_ADDR_RUNLOCK();
909							return (1);
910						}
911					}
912					break;
913#endif
914#ifdef INET6
915				case AF_INET6:
916					if (ipv6_addr_legal) {
917						struct sockaddr_in6 *sin6,
918						            *rsin6;
919
920						sin6 = &sctp_ifa->address.sin6;
921						rsin6 = (struct sockaddr_in6 *)to;
922						if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred,
923						    &sin6->sin6_addr) != 0) {
924							continue;
925						}
926						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
927							if (local_scope == 0)
928								continue;
929							if (sin6->sin6_scope_id == 0) {
930								if (sa6_recoverscope(sin6) != 0)
931									continue;
932							}
933						}
934						if ((site_scope == 0) &&
935						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
936							continue;
937						}
938						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
939							SCTP_IPI_ADDR_RUNLOCK();
940							return (1);
941						}
942					}
943					break;
944#endif
945				default:
946					/* TSNH */
947					break;
948				}
949			}
950		}
951	} else {
952		struct sctp_laddr *laddr;
953
954		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
955			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
956				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
957				continue;
958			}
959			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
960			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
961				/*
962				 * We allow pending addresses, where we have
963				 * sent an asconf-add to be considered
964				 * valid.
965				 */
966				continue;
967			}
968			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
969				continue;
970			}
971			switch (to->sa_family) {
972#ifdef INET
973			case AF_INET:
974				{
975					struct sockaddr_in *sin, *rsin;
976
977					sin = &laddr->ifa->address.sin;
978					rsin = (struct sockaddr_in *)to;
979					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
980						SCTP_IPI_ADDR_RUNLOCK();
981						return (1);
982					}
983					break;
984				}
985#endif
986#ifdef INET6
987			case AF_INET6:
988				{
989					struct sockaddr_in6 *sin6, *rsin6;
990
991					sin6 = &laddr->ifa->address.sin6;
992					rsin6 = (struct sockaddr_in6 *)to;
993					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
994						SCTP_IPI_ADDR_RUNLOCK();
995						return (1);
996					}
997					break;
998				}
999
1000#endif
1001			default:
1002				/* TSNH */
1003				break;
1004			}
1005
1006		}
1007	}
1008	SCTP_IPI_ADDR_RUNLOCK();
1009	return (0);
1010}
1011
1012
1013static struct sctp_tcb *
1014sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
1015    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
1016{
1017	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
1018	/*
1019	 * If we support the TCP model, then we must now dig through to see
1020	 * if we can find our endpoint in the list of tcp ep's.
1021	 */
1022	uint16_t lport, rport;
1023	struct sctppcbhead *ephead;
1024	struct sctp_inpcb *inp;
1025	struct sctp_laddr *laddr;
1026	struct sctp_tcb *stcb;
1027	struct sctp_nets *net;
1028
1029	if ((to == NULL) || (from == NULL)) {
1030		return (NULL);
1031	}
1032	switch (to->sa_family) {
1033#ifdef INET
1034	case AF_INET:
1035		if (from->sa_family == AF_INET) {
1036			lport = ((struct sockaddr_in *)to)->sin_port;
1037			rport = ((struct sockaddr_in *)from)->sin_port;
1038		} else {
1039			return (NULL);
1040		}
1041		break;
1042#endif
1043#ifdef INET6
1044	case AF_INET6:
1045		if (from->sa_family == AF_INET6) {
1046			lport = ((struct sockaddr_in6 *)to)->sin6_port;
1047			rport = ((struct sockaddr_in6 *)from)->sin6_port;
1048		} else {
1049			return (NULL);
1050		}
1051		break;
1052#endif
1053	default:
1054		return (NULL);
1055	}
1056	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
1057	/*
1058	 * Ok now for each of the guys in this bucket we must look and see:
1059	 * - Does the remote port match. - Does there single association's
1060	 * addresses match this address (to). If so we update p_ep to point
1061	 * to this ep and return the tcb from it.
1062	 */
1063	LIST_FOREACH(inp, ephead, sctp_hash) {
1064		SCTP_INP_RLOCK(inp);
1065		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1066			SCTP_INP_RUNLOCK(inp);
1067			continue;
1068		}
1069		if (lport != inp->sctp_lport) {
1070			SCTP_INP_RUNLOCK(inp);
1071			continue;
1072		}
1073		switch (to->sa_family) {
1074#ifdef INET
1075		case AF_INET:
1076			{
1077				struct sockaddr_in *sin;
1078
1079				sin = (struct sockaddr_in *)to;
1080				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1081				    &sin->sin_addr) != 0) {
1082					SCTP_INP_RUNLOCK(inp);
1083					continue;
1084				}
1085				break;
1086			}
1087#endif
1088#ifdef INET6
1089		case AF_INET6:
1090			{
1091				struct sockaddr_in6 *sin6;
1092
1093				sin6 = (struct sockaddr_in6 *)to;
1094				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1095				    &sin6->sin6_addr) != 0) {
1096					SCTP_INP_RUNLOCK(inp);
1097					continue;
1098				}
1099				break;
1100			}
1101#endif
1102		default:
1103			SCTP_INP_RUNLOCK(inp);
1104			continue;
1105		}
1106		if (inp->def_vrf_id != vrf_id) {
1107			SCTP_INP_RUNLOCK(inp);
1108			continue;
1109		}
1110		/* check to see if the ep has one of the addresses */
1111		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1112			/* We are NOT bound all, so look further */
1113			int match = 0;
1114
1115			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1116
1117				if (laddr->ifa == NULL) {
1118					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
1119					continue;
1120				}
1121				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1122					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1123					continue;
1124				}
1125				if (laddr->ifa->address.sa.sa_family ==
1126				    to->sa_family) {
1127					/* see if it matches */
1128#ifdef INET
1129					if (from->sa_family == AF_INET) {
1130						struct sockaddr_in *intf_addr,
1131						           *sin;
1132
1133						intf_addr = &laddr->ifa->address.sin;
1134						sin = (struct sockaddr_in *)to;
1135						if (sin->sin_addr.s_addr ==
1136						    intf_addr->sin_addr.s_addr) {
1137							match = 1;
1138							break;
1139						}
1140					}
1141#endif
1142#ifdef INET6
1143					if (from->sa_family == AF_INET6) {
1144						struct sockaddr_in6 *intf_addr6;
1145						struct sockaddr_in6 *sin6;
1146
1147						sin6 = (struct sockaddr_in6 *)
1148						    to;
1149						intf_addr6 = &laddr->ifa->address.sin6;
1150
1151						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1152						    intf_addr6)) {
1153							match = 1;
1154							break;
1155						}
1156					}
1157#endif
1158				}
1159			}
1160			if (match == 0) {
1161				/* This endpoint does not have this address */
1162				SCTP_INP_RUNLOCK(inp);
1163				continue;
1164			}
1165		}
1166		/*
1167		 * Ok if we hit here the ep has the address, does it hold
1168		 * the tcb?
1169		 */
1170		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1171		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1172		if (stcb == NULL) {
1173			SCTP_INP_RUNLOCK(inp);
1174			continue;
1175		}
1176		SCTP_TCB_LOCK(stcb);
1177		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1178			SCTP_TCB_UNLOCK(stcb);
1179			SCTP_INP_RUNLOCK(inp);
1180			continue;
1181		}
1182		if (stcb->rport != rport) {
1183			/* remote port does not match. */
1184			SCTP_TCB_UNLOCK(stcb);
1185			SCTP_INP_RUNLOCK(inp);
1186			continue;
1187		}
1188		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1189			SCTP_TCB_UNLOCK(stcb);
1190			SCTP_INP_RUNLOCK(inp);
1191			continue;
1192		}
1193		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1194			SCTP_TCB_UNLOCK(stcb);
1195			SCTP_INP_RUNLOCK(inp);
1196			continue;
1197		}
1198		/* Does this TCB have a matching address? */
1199		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1200
1201			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1202				/* not the same family, can't be a match */
1203				continue;
1204			}
1205			switch (from->sa_family) {
1206#ifdef INET
1207			case AF_INET:
1208				{
1209					struct sockaddr_in *sin, *rsin;
1210
1211					sin = (struct sockaddr_in *)&net->ro._l_addr;
1212					rsin = (struct sockaddr_in *)from;
1213					if (sin->sin_addr.s_addr ==
1214					    rsin->sin_addr.s_addr) {
1215						/* found it */
1216						if (netp != NULL) {
1217							*netp = net;
1218						}
1219						/*
1220						 * Update the endpoint
1221						 * pointer
1222						 */
1223						*inp_p = inp;
1224						SCTP_INP_RUNLOCK(inp);
1225						return (stcb);
1226					}
1227					break;
1228				}
1229#endif
1230#ifdef INET6
1231			case AF_INET6:
1232				{
1233					struct sockaddr_in6 *sin6, *rsin6;
1234
1235					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1236					rsin6 = (struct sockaddr_in6 *)from;
1237					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1238					    rsin6)) {
1239						/* found it */
1240						if (netp != NULL) {
1241							*netp = net;
1242						}
1243						/*
1244						 * Update the endpoint
1245						 * pointer
1246						 */
1247						*inp_p = inp;
1248						SCTP_INP_RUNLOCK(inp);
1249						return (stcb);
1250					}
1251					break;
1252				}
1253#endif
1254			default:
1255				/* TSNH */
1256				break;
1257			}
1258		}
1259		SCTP_TCB_UNLOCK(stcb);
1260		SCTP_INP_RUNLOCK(inp);
1261	}
1262	return (NULL);
1263}
1264
1265
1266/*
1267 * rules for use
1268 *
1269 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1270 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1271 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1272 * NULL.
1273 */
1274
1275struct sctp_tcb *
1276sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1277    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1278{
1279	struct sctpasochead *head;
1280	struct sctp_inpcb *inp;
1281	struct sctp_tcb *stcb = NULL;
1282	struct sctp_nets *net;
1283	uint16_t rport;
1284
1285	inp = *inp_p;
1286	switch (remote->sa_family) {
1287#ifdef INET
1288	case AF_INET:
1289		rport = (((struct sockaddr_in *)remote)->sin_port);
1290		break;
1291#endif
1292#ifdef INET6
1293	case AF_INET6:
1294		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1295		break;
1296#endif
1297	default:
1298		return (NULL);
1299	}
1300	if (locked_tcb) {
1301		/*
1302		 * UN-lock so we can do proper locking here this occurs when
1303		 * called from load_addresses_from_init.
1304		 */
1305		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1306		SCTP_TCB_UNLOCK(locked_tcb);
1307	}
1308	SCTP_INP_INFO_RLOCK();
1309	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1310	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1311		/*-
1312		 * Now either this guy is our listener or it's the
1313		 * connector. If it is the one that issued the connect, then
1314		 * it's only chance is to be the first TCB in the list. If
1315		 * it is the acceptor, then do the special_lookup to hash
1316		 * and find the real inp.
1317		 */
1318		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1319			/* to is peer addr, from is my addr */
1320			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1321			    netp, inp->def_vrf_id);
1322			if ((stcb != NULL) && (locked_tcb == NULL)) {
1323				/* we have a locked tcb, lower refcount */
1324				SCTP_INP_DECR_REF(inp);
1325			}
1326			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1327				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1328				SCTP_TCB_LOCK(locked_tcb);
1329				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1330				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1331			}
1332			SCTP_INP_INFO_RUNLOCK();
1333			return (stcb);
1334		} else {
1335			SCTP_INP_WLOCK(inp);
1336			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1337				goto null_return;
1338			}
1339			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1340			if (stcb == NULL) {
1341				goto null_return;
1342			}
1343			SCTP_TCB_LOCK(stcb);
1344
1345			if (stcb->rport != rport) {
1346				/* remote port does not match. */
1347				SCTP_TCB_UNLOCK(stcb);
1348				goto null_return;
1349			}
1350			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1351				SCTP_TCB_UNLOCK(stcb);
1352				goto null_return;
1353			}
1354			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1355				SCTP_TCB_UNLOCK(stcb);
1356				goto null_return;
1357			}
1358			/* now look at the list of remote addresses */
1359			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1360#ifdef INVARIANTS
1361				if (net == (TAILQ_NEXT(net, sctp_next))) {
1362					panic("Corrupt net list");
1363				}
1364#endif
1365				if (net->ro._l_addr.sa.sa_family !=
1366				    remote->sa_family) {
1367					/* not the same family */
1368					continue;
1369				}
1370				switch (remote->sa_family) {
1371#ifdef INET
1372				case AF_INET:
1373					{
1374						struct sockaddr_in *sin,
1375						           *rsin;
1376
1377						sin = (struct sockaddr_in *)
1378						    &net->ro._l_addr;
1379						rsin = (struct sockaddr_in *)remote;
1380						if (sin->sin_addr.s_addr ==
1381						    rsin->sin_addr.s_addr) {
1382							/* found it */
1383							if (netp != NULL) {
1384								*netp = net;
1385							}
1386							if (locked_tcb == NULL) {
1387								SCTP_INP_DECR_REF(inp);
1388							} else if (locked_tcb != stcb) {
1389								SCTP_TCB_LOCK(locked_tcb);
1390							}
1391							if (locked_tcb) {
1392								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1393							}
1394							SCTP_INP_WUNLOCK(inp);
1395							SCTP_INP_INFO_RUNLOCK();
1396							return (stcb);
1397						}
1398						break;
1399					}
1400#endif
1401#ifdef INET6
1402				case AF_INET6:
1403					{
1404						struct sockaddr_in6 *sin6,
1405						            *rsin6;
1406
1407						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1408						rsin6 = (struct sockaddr_in6 *)remote;
1409						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1410						    rsin6)) {
1411							/* found it */
1412							if (netp != NULL) {
1413								*netp = net;
1414							}
1415							if (locked_tcb == NULL) {
1416								SCTP_INP_DECR_REF(inp);
1417							} else if (locked_tcb != stcb) {
1418								SCTP_TCB_LOCK(locked_tcb);
1419							}
1420							if (locked_tcb) {
1421								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1422							}
1423							SCTP_INP_WUNLOCK(inp);
1424							SCTP_INP_INFO_RUNLOCK();
1425							return (stcb);
1426						}
1427						break;
1428					}
1429#endif
1430				default:
1431					/* TSNH */
1432					break;
1433				}
1434			}
1435			SCTP_TCB_UNLOCK(stcb);
1436		}
1437	} else {
1438		SCTP_INP_WLOCK(inp);
1439		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1440			goto null_return;
1441		}
1442		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1443		    inp->sctp_hashmark)];
1444		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1445			if (stcb->rport != rport) {
1446				/* remote port does not match */
1447				continue;
1448			}
1449			SCTP_TCB_LOCK(stcb);
1450			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1451				SCTP_TCB_UNLOCK(stcb);
1452				continue;
1453			}
1454			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1455				SCTP_TCB_UNLOCK(stcb);
1456				continue;
1457			}
1458			/* now look at the list of remote addresses */
1459			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1460#ifdef INVARIANTS
1461				if (net == (TAILQ_NEXT(net, sctp_next))) {
1462					panic("Corrupt net list");
1463				}
1464#endif
1465				if (net->ro._l_addr.sa.sa_family !=
1466				    remote->sa_family) {
1467					/* not the same family */
1468					continue;
1469				}
1470				switch (remote->sa_family) {
1471#ifdef INET
1472				case AF_INET:
1473					{
1474						struct sockaddr_in *sin,
1475						           *rsin;
1476
1477						sin = (struct sockaddr_in *)
1478						    &net->ro._l_addr;
1479						rsin = (struct sockaddr_in *)remote;
1480						if (sin->sin_addr.s_addr ==
1481						    rsin->sin_addr.s_addr) {
1482							/* found it */
1483							if (netp != NULL) {
1484								*netp = net;
1485							}
1486							if (locked_tcb == NULL) {
1487								SCTP_INP_DECR_REF(inp);
1488							} else if (locked_tcb != stcb) {
1489								SCTP_TCB_LOCK(locked_tcb);
1490							}
1491							if (locked_tcb) {
1492								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1493							}
1494							SCTP_INP_WUNLOCK(inp);
1495							SCTP_INP_INFO_RUNLOCK();
1496							return (stcb);
1497						}
1498						break;
1499					}
1500#endif
1501#ifdef INET6
1502				case AF_INET6:
1503					{
1504						struct sockaddr_in6 *sin6,
1505						            *rsin6;
1506
1507						sin6 = (struct sockaddr_in6 *)
1508						    &net->ro._l_addr;
1509						rsin6 = (struct sockaddr_in6 *)remote;
1510						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1511						    rsin6)) {
1512							/* found it */
1513							if (netp != NULL) {
1514								*netp = net;
1515							}
1516							if (locked_tcb == NULL) {
1517								SCTP_INP_DECR_REF(inp);
1518							} else if (locked_tcb != stcb) {
1519								SCTP_TCB_LOCK(locked_tcb);
1520							}
1521							if (locked_tcb) {
1522								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1523							}
1524							SCTP_INP_WUNLOCK(inp);
1525							SCTP_INP_INFO_RUNLOCK();
1526							return (stcb);
1527						}
1528						break;
1529					}
1530#endif
1531				default:
1532					/* TSNH */
1533					break;
1534				}
1535			}
1536			SCTP_TCB_UNLOCK(stcb);
1537		}
1538	}
1539null_return:
1540	/* clean up for returning null */
1541	if (locked_tcb) {
1542		SCTP_TCB_LOCK(locked_tcb);
1543		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1544	}
1545	SCTP_INP_WUNLOCK(inp);
1546	SCTP_INP_INFO_RUNLOCK();
1547	/* not found */
1548	return (NULL);
1549}
1550
1551
1552/*
1553 * Find an association for a specific endpoint using the association id given
1554 * out in the COMM_UP notification
1555 */
1556struct sctp_tcb *
1557sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1558{
1559	/*
1560	 * Use my the assoc_id to find a endpoint
1561	 */
1562	struct sctpasochead *head;
1563	struct sctp_tcb *stcb;
1564	uint32_t id;
1565
1566	if (inp == NULL) {
1567		SCTP_PRINTF("TSNH ep_associd\n");
1568		return (NULL);
1569	}
1570	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1571		SCTP_PRINTF("TSNH ep_associd0\n");
1572		return (NULL);
1573	}
1574	id = (uint32_t) asoc_id;
1575	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1576	if (head == NULL) {
1577		/* invalid id TSNH */
1578		SCTP_PRINTF("TSNH ep_associd1\n");
1579		return (NULL);
1580	}
1581	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1582		if (stcb->asoc.assoc_id == id) {
1583			if (inp != stcb->sctp_ep) {
1584				/*
1585				 * some other guy has the same id active (id
1586				 * collision ??).
1587				 */
1588				SCTP_PRINTF("TSNH ep_associd2\n");
1589				continue;
1590			}
1591			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1592				continue;
1593			}
1594			if (want_lock) {
1595				SCTP_TCB_LOCK(stcb);
1596			}
1597			return (stcb);
1598		}
1599	}
1600	return (NULL);
1601}
1602
1603
1604struct sctp_tcb *
1605sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1606{
1607	struct sctp_tcb *stcb;
1608
1609	SCTP_INP_RLOCK(inp);
1610	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1611	SCTP_INP_RUNLOCK(inp);
1612	return (stcb);
1613}
1614
1615
1616/*
1617 * Endpoint probe expects that the INP_INFO is locked.
1618 */
1619static struct sctp_inpcb *
1620sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1621    uint16_t lport, uint32_t vrf_id)
1622{
1623	struct sctp_inpcb *inp;
1624	struct sctp_laddr *laddr;
1625
1626#ifdef INET
1627	struct sockaddr_in *sin;
1628
1629#endif
1630#ifdef INET6
1631	struct sockaddr_in6 *sin6;
1632	struct sockaddr_in6 *intf_addr6;
1633
1634#endif
1635	int fnd;
1636
1637#ifdef INET
1638	sin = NULL;
1639#endif
1640#ifdef INET6
1641	sin6 = NULL;
1642#endif
1643	switch (nam->sa_family) {
1644#ifdef INET
1645	case AF_INET:
1646		sin = (struct sockaddr_in *)nam;
1647		break;
1648#endif
1649#ifdef INET6
1650	case AF_INET6:
1651		sin6 = (struct sockaddr_in6 *)nam;
1652		break;
1653#endif
1654	default:
1655		/* unsupported family */
1656		return (NULL);
1657	}
1658
1659	if (head == NULL)
1660		return (NULL);
1661
1662	LIST_FOREACH(inp, head, sctp_hash) {
1663		SCTP_INP_RLOCK(inp);
1664		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1665			SCTP_INP_RUNLOCK(inp);
1666			continue;
1667		}
1668		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1669		    (inp->sctp_lport == lport)) {
1670			/* got it */
1671			switch (nam->sa_family) {
1672#ifdef INET
1673			case AF_INET:
1674				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1675				    SCTP_IPV6_V6ONLY(inp)) {
1676					/*
1677					 * IPv4 on a IPv6 socket with ONLY
1678					 * IPv6 set
1679					 */
1680					SCTP_INP_RUNLOCK(inp);
1681					continue;
1682				}
1683				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1684				    &sin->sin_addr) != 0) {
1685					SCTP_INP_RUNLOCK(inp);
1686					continue;
1687				}
1688				break;
1689#endif
1690#ifdef INET6
1691			case AF_INET6:
1692				/*
1693				 * A V6 address and the endpoint is NOT
1694				 * bound V6
1695				 */
1696				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1697					SCTP_INP_RUNLOCK(inp);
1698					continue;
1699				}
1700				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1701				    &sin6->sin6_addr) != 0) {
1702					SCTP_INP_RUNLOCK(inp);
1703					continue;
1704				}
1705				break;
1706#endif
1707			default:
1708				break;
1709			}
1710			/* does a VRF id match? */
1711			fnd = 0;
1712			if (inp->def_vrf_id == vrf_id)
1713				fnd = 1;
1714
1715			SCTP_INP_RUNLOCK(inp);
1716			if (!fnd)
1717				continue;
1718			return (inp);
1719		}
1720		SCTP_INP_RUNLOCK(inp);
1721	}
1722	switch (nam->sa_family) {
1723#ifdef INET
1724	case AF_INET:
1725		if (sin->sin_addr.s_addr == INADDR_ANY) {
1726			/* Can't hunt for one that has no address specified */
1727			return (NULL);
1728		}
1729		break;
1730#endif
1731#ifdef INET6
1732	case AF_INET6:
1733		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1734			/* Can't hunt for one that has no address specified */
1735			return (NULL);
1736		}
1737		break;
1738#endif
1739	default:
1740		break;
1741	}
1742	/*
1743	 * ok, not bound to all so see if we can find a EP bound to this
1744	 * address.
1745	 */
1746	LIST_FOREACH(inp, head, sctp_hash) {
1747		SCTP_INP_RLOCK(inp);
1748		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1749			SCTP_INP_RUNLOCK(inp);
1750			continue;
1751		}
1752		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1753			SCTP_INP_RUNLOCK(inp);
1754			continue;
1755		}
1756		/*
1757		 * Ok this could be a likely candidate, look at all of its
1758		 * addresses
1759		 */
1760		if (inp->sctp_lport != lport) {
1761			SCTP_INP_RUNLOCK(inp);
1762			continue;
1763		}
1764		/* does a VRF id match? */
1765		fnd = 0;
1766		if (inp->def_vrf_id == vrf_id)
1767			fnd = 1;
1768
1769		if (!fnd) {
1770			SCTP_INP_RUNLOCK(inp);
1771			continue;
1772		}
1773		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1774			if (laddr->ifa == NULL) {
1775				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1776				    __FUNCTION__);
1777				continue;
1778			}
1779			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1780			    (void *)laddr->ifa);
1781			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1782				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1783				continue;
1784			}
1785			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1786				/* possible, see if it matches */
1787				switch (nam->sa_family) {
1788#ifdef INET
1789				case AF_INET:
1790					if (sin->sin_addr.s_addr ==
1791					    laddr->ifa->address.sin.sin_addr.s_addr) {
1792						SCTP_INP_RUNLOCK(inp);
1793						return (inp);
1794					}
1795					break;
1796#endif
1797#ifdef INET6
1798				case AF_INET6:
1799					intf_addr6 = &laddr->ifa->address.sin6;
1800					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1801					    intf_addr6)) {
1802						SCTP_INP_RUNLOCK(inp);
1803						return (inp);
1804					}
1805					break;
1806#endif
1807				}
1808			}
1809		}
1810		SCTP_INP_RUNLOCK(inp);
1811	}
1812	return (NULL);
1813}
1814
1815
1816static struct sctp_inpcb *
1817sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1818{
1819	struct sctppcbhead *head;
1820	struct sctp_inpcb *t_inp;
1821	int fnd;
1822
1823	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1824	    SCTP_BASE_INFO(hashmark))];
1825	LIST_FOREACH(t_inp, head, sctp_hash) {
1826		if (t_inp->sctp_lport != lport) {
1827			continue;
1828		}
1829		/* is it in the VRF in question */
1830		fnd = 0;
1831		if (t_inp->def_vrf_id == vrf_id)
1832			fnd = 1;
1833		if (!fnd)
1834			continue;
1835
1836		/* This one is in use. */
1837		/* check the v6/v4 binding issue */
1838		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1839		    SCTP_IPV6_V6ONLY(t_inp)) {
1840			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1841				/* collision in V6 space */
1842				return (t_inp);
1843			} else {
1844				/* inp is BOUND_V4 no conflict */
1845				continue;
1846			}
1847		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1848			/* t_inp is bound v4 and v6, conflict always */
1849			return (t_inp);
1850		} else {
1851			/* t_inp is bound only V4 */
1852			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1853			    SCTP_IPV6_V6ONLY(inp)) {
1854				/* no conflict */
1855				continue;
1856			}
1857			/* else fall through to conflict */
1858		}
1859		return (t_inp);
1860	}
1861	return (NULL);
1862}
1863
1864
1865int
1866sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1867{
1868	/* For 1-2-1 with port reuse */
1869	struct sctppcbhead *head;
1870	struct sctp_inpcb *tinp, *ninp;
1871
1872	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1873		/* only works with port reuse on */
1874		return (-1);
1875	}
1876	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1877		return (0);
1878	}
1879	SCTP_INP_RUNLOCK(inp);
1880	SCTP_INP_INFO_WLOCK();
1881	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1882	    SCTP_BASE_INFO(hashmark))];
1883	/* Kick out all non-listeners to the TCP hash */
1884	LIST_FOREACH_SAFE(tinp, head, sctp_hash, ninp) {
1885		if (tinp->sctp_lport != inp->sctp_lport) {
1886			continue;
1887		}
1888		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1889			continue;
1890		}
1891		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1892			continue;
1893		}
1894		if (tinp->sctp_socket->so_qlimit) {
1895			continue;
1896		}
1897		SCTP_INP_WLOCK(tinp);
1898		LIST_REMOVE(tinp, sctp_hash);
1899		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1900		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1901		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1902		SCTP_INP_WUNLOCK(tinp);
1903	}
1904	SCTP_INP_WLOCK(inp);
1905	/* Pull from where he was */
1906	LIST_REMOVE(inp, sctp_hash);
1907	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1908	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1909	LIST_INSERT_HEAD(head, inp, sctp_hash);
1910	SCTP_INP_WUNLOCK(inp);
1911	SCTP_INP_RLOCK(inp);
1912	SCTP_INP_INFO_WUNLOCK();
1913	return (0);
1914}
1915
1916
1917struct sctp_inpcb *
1918sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1919    uint32_t vrf_id)
1920{
1921	/*
1922	 * First we check the hash table to see if someone has this port
1923	 * bound with just the port.
1924	 */
1925	struct sctp_inpcb *inp;
1926	struct sctppcbhead *head;
1927	int lport;
1928	unsigned int i;
1929
1930#ifdef INET
1931	struct sockaddr_in *sin;
1932
1933#endif
1934#ifdef INET6
1935	struct sockaddr_in6 *sin6;
1936
1937#endif
1938
1939	switch (nam->sa_family) {
1940#ifdef INET
1941	case AF_INET:
1942		sin = (struct sockaddr_in *)nam;
1943		lport = sin->sin_port;
1944		break;
1945#endif
1946#ifdef INET6
1947	case AF_INET6:
1948		sin6 = (struct sockaddr_in6 *)nam;
1949		lport = sin6->sin6_port;
1950		break;
1951#endif
1952	default:
1953		return (NULL);
1954	}
1955	/*
1956	 * I could cheat here and just cast to one of the types but we will
1957	 * do it right. It also provides the check against an Unsupported
1958	 * type too.
1959	 */
1960	/* Find the head of the ALLADDR chain */
1961	if (have_lock == 0) {
1962		SCTP_INP_INFO_RLOCK();
1963	}
1964	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1965	    SCTP_BASE_INFO(hashmark))];
1966	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1967
1968	/*
1969	 * If the TCP model exists it could be that the main listening
1970	 * endpoint is gone but there still exists a connected socket for
1971	 * this guy. If so we can return the first one that we find. This
1972	 * may NOT be the correct one so the caller should be wary on the
1973	 * returned INP. Currently the only caller that sets find_tcp_pool
1974	 * is in bindx where we are verifying that a user CAN bind the
1975	 * address. He either has bound it already, or someone else has, or
1976	 * its open to bind, so this is good enough.
1977	 */
1978	if (inp == NULL && find_tcp_pool) {
1979		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1980			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1981			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1982			if (inp) {
1983				break;
1984			}
1985		}
1986	}
1987	if (inp) {
1988		SCTP_INP_INCR_REF(inp);
1989	}
1990	if (have_lock == 0) {
1991		SCTP_INP_INFO_RUNLOCK();
1992	}
1993	return (inp);
1994}
1995
1996
1997/*
1998 * Find an association for an endpoint with the pointer to whom you want to
1999 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
2000 * need to change the *to to some other struct like a mbuf...
2001 */
2002struct sctp_tcb *
2003sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
2004    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
2005    uint32_t vrf_id)
2006{
2007	struct sctp_inpcb *inp = NULL;
2008	struct sctp_tcb *stcb;
2009
2010	SCTP_INP_INFO_RLOCK();
2011	if (find_tcp_pool) {
2012		if (inp_p != NULL) {
2013			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
2014			    vrf_id);
2015		} else {
2016			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
2017			    vrf_id);
2018		}
2019		if (stcb != NULL) {
2020			SCTP_INP_INFO_RUNLOCK();
2021			return (stcb);
2022		}
2023	}
2024	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
2025	if (inp_p != NULL) {
2026		*inp_p = inp;
2027	}
2028	SCTP_INP_INFO_RUNLOCK();
2029	if (inp == NULL) {
2030		return (NULL);
2031	}
2032	/*
2033	 * ok, we have an endpoint, now lets find the assoc for it (if any)
2034	 * we now place the source address or from in the to of the find
2035	 * endpoint call. Since in reality this chain is used from the
2036	 * inbound packet side.
2037	 */
2038	if (inp_p != NULL) {
2039		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
2040		    NULL);
2041	} else {
2042		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
2043		    NULL);
2044	}
2045	return (stcb);
2046}
2047
2048
2049/*
2050 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
2051 * find all addresses that the sender has specified in any address list. Each
2052 * address will be used to lookup the TCB and see if one exits.
2053 */
2054static struct sctp_tcb *
2055sctp_findassociation_special_addr(struct mbuf *m, int offset,
2056    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
2057    struct sockaddr *dst)
2058{
2059	struct sctp_paramhdr *phdr, parm_buf;
2060
2061#if defined(INET) || defined(INET6)
2062	struct sctp_tcb *stcb;
2063	uint16_t ptype;
2064
2065#endif
2066	uint16_t plen;
2067
2068#ifdef INET
2069	struct sockaddr_in sin4;
2070
2071#endif
2072#ifdef INET6
2073	struct sockaddr_in6 sin6;
2074
2075#endif
2076
2077#ifdef INET
2078	memset(&sin4, 0, sizeof(sin4));
2079	sin4.sin_len = sizeof(sin4);
2080	sin4.sin_family = AF_INET;
2081	sin4.sin_port = sh->src_port;
2082#endif
2083#ifdef INET6
2084	memset(&sin6, 0, sizeof(sin6));
2085	sin6.sin6_len = sizeof(sin6);
2086	sin6.sin6_family = AF_INET6;
2087	sin6.sin6_port = sh->src_port;
2088#endif
2089
2090	offset += sizeof(struct sctp_init_chunk);
2091
2092	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2093	while (phdr != NULL) {
2094		/* now we must see if we want the parameter */
2095#if defined(INET) || defined(INET6)
2096		ptype = ntohs(phdr->param_type);
2097#endif
2098		plen = ntohs(phdr->param_length);
2099		if (plen == 0) {
2100			break;
2101		}
2102#ifdef INET
2103		if (ptype == SCTP_IPV4_ADDRESS &&
2104		    plen == sizeof(struct sctp_ipv4addr_param)) {
2105			/* Get the rest of the address */
2106			struct sctp_ipv4addr_param ip4_parm, *p4;
2107
2108			phdr = sctp_get_next_param(m, offset,
2109			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2110			if (phdr == NULL) {
2111				return (NULL);
2112			}
2113			p4 = (struct sctp_ipv4addr_param *)phdr;
2114			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2115			/* look it up */
2116			stcb = sctp_findassociation_ep_addr(inp_p,
2117			    (struct sockaddr *)&sin4, netp, dst, NULL);
2118			if (stcb != NULL) {
2119				return (stcb);
2120			}
2121		}
2122#endif
2123#ifdef INET6
2124		if (ptype == SCTP_IPV6_ADDRESS &&
2125		    plen == sizeof(struct sctp_ipv6addr_param)) {
2126			/* Get the rest of the address */
2127			struct sctp_ipv6addr_param ip6_parm, *p6;
2128
2129			phdr = sctp_get_next_param(m, offset,
2130			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2131			if (phdr == NULL) {
2132				return (NULL);
2133			}
2134			p6 = (struct sctp_ipv6addr_param *)phdr;
2135			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2136			/* look it up */
2137			stcb = sctp_findassociation_ep_addr(inp_p,
2138			    (struct sockaddr *)&sin6, netp, dst, NULL);
2139			if (stcb != NULL) {
2140				return (stcb);
2141			}
2142		}
2143#endif
2144		offset += SCTP_SIZE32(plen);
2145		phdr = sctp_get_next_param(m, offset, &parm_buf,
2146		    sizeof(parm_buf));
2147	}
2148	return (NULL);
2149}
2150
2151static struct sctp_tcb *
2152sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2153    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2154    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2155{
2156	/*
2157	 * Use my vtag to hash. If we find it we then verify the source addr
2158	 * is in the assoc. If all goes well we save a bit on rec of a
2159	 * packet.
2160	 */
2161	struct sctpasochead *head;
2162	struct sctp_nets *net;
2163	struct sctp_tcb *stcb;
2164
2165	SCTP_INP_INFO_RLOCK();
2166	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2167	    SCTP_BASE_INFO(hashasocmark))];
2168	LIST_FOREACH(stcb, head, sctp_asocs) {
2169		SCTP_INP_RLOCK(stcb->sctp_ep);
2170		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2171			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2172			continue;
2173		}
2174		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2175			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2176			continue;
2177		}
2178		SCTP_TCB_LOCK(stcb);
2179		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2180		if (stcb->asoc.my_vtag == vtag) {
2181			/* candidate */
2182			if (stcb->rport != rport) {
2183				SCTP_TCB_UNLOCK(stcb);
2184				continue;
2185			}
2186			if (stcb->sctp_ep->sctp_lport != lport) {
2187				SCTP_TCB_UNLOCK(stcb);
2188				continue;
2189			}
2190			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2191				SCTP_TCB_UNLOCK(stcb);
2192				continue;
2193			}
2194			/* RRS:Need toaddr check here */
2195			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2196				/* Endpoint does not own this address */
2197				SCTP_TCB_UNLOCK(stcb);
2198				continue;
2199			}
2200			if (remote_tag) {
2201				/*
2202				 * If we have both vtags that's all we match
2203				 * on
2204				 */
2205				if (stcb->asoc.peer_vtag == remote_tag) {
2206					/*
2207					 * If both tags match we consider it
2208					 * conclusive and check NO
2209					 * source/destination addresses
2210					 */
2211					goto conclusive;
2212				}
2213			}
2214			if (skip_src_check) {
2215		conclusive:
2216				if (from) {
2217					*netp = sctp_findnet(stcb, from);
2218				} else {
2219					*netp = NULL;	/* unknown */
2220				}
2221				if (inp_p)
2222					*inp_p = stcb->sctp_ep;
2223				SCTP_INP_INFO_RUNLOCK();
2224				return (stcb);
2225			}
2226			net = sctp_findnet(stcb, from);
2227			if (net) {
2228				/* yep its him. */
2229				*netp = net;
2230				SCTP_STAT_INCR(sctps_vtagexpress);
2231				*inp_p = stcb->sctp_ep;
2232				SCTP_INP_INFO_RUNLOCK();
2233				return (stcb);
2234			} else {
2235				/*
2236				 * not him, this should only happen in rare
2237				 * cases so I peg it.
2238				 */
2239				SCTP_STAT_INCR(sctps_vtagbogus);
2240			}
2241		}
2242		SCTP_TCB_UNLOCK(stcb);
2243	}
2244	SCTP_INP_INFO_RUNLOCK();
2245	return (NULL);
2246}
2247
2248
2249/*
2250 * Find an association with the pointer to the inbound IP packet. This can be
2251 * a IPv4 or IPv6 packet.
2252 */
2253struct sctp_tcb *
2254sctp_findassociation_addr(struct mbuf *m, int offset,
2255    struct sockaddr *src, struct sockaddr *dst,
2256    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2257    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2258{
2259	int find_tcp_pool;
2260	struct sctp_tcb *stcb;
2261	struct sctp_inpcb *inp;
2262
2263	if (sh->v_tag) {
2264		/* we only go down this path if vtag is non-zero */
2265		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2266		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2267		if (stcb) {
2268			return (stcb);
2269		}
2270	}
2271	find_tcp_pool = 0;
2272	if ((ch->chunk_type != SCTP_INITIATION) &&
2273	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2274	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2275	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2276		/* Other chunk types go to the tcp pool. */
2277		find_tcp_pool = 1;
2278	}
2279	if (inp_p) {
2280		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2281		    find_tcp_pool, vrf_id);
2282		inp = *inp_p;
2283	} else {
2284		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2285		    find_tcp_pool, vrf_id);
2286	}
2287	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2288	if (stcb == NULL && inp) {
2289		/* Found a EP but not this address */
2290		if ((ch->chunk_type == SCTP_INITIATION) ||
2291		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2292			/*-
2293			 * special hook, we do NOT return linp or an
2294			 * association that is linked to an existing
2295			 * association that is under the TCP pool (i.e. no
2296			 * listener exists). The endpoint finding routine
2297			 * will always find a listener before examining the
2298			 * TCP pool.
2299			 */
2300			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2301				if (inp_p) {
2302					*inp_p = NULL;
2303				}
2304				return (NULL);
2305			}
2306			stcb = sctp_findassociation_special_addr(m,
2307			    offset, sh, &inp, netp, dst);
2308			if (inp_p != NULL) {
2309				*inp_p = inp;
2310			}
2311		}
2312	}
2313	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2314	return (stcb);
2315}
2316
2317/*
2318 * lookup an association by an ASCONF lookup address.
2319 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2320 */
2321struct sctp_tcb *
2322sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2323    struct sockaddr *dst, struct sctphdr *sh,
2324    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2325{
2326	struct sctp_tcb *stcb;
2327	union sctp_sockstore remote_store;
2328	struct sctp_paramhdr parm_buf, *phdr;
2329	int ptype;
2330	int zero_address = 0;
2331
2332#ifdef INET
2333	struct sockaddr_in *sin;
2334
2335#endif
2336#ifdef INET6
2337	struct sockaddr_in6 *sin6;
2338
2339#endif
2340
2341	memset(&remote_store, 0, sizeof(remote_store));
2342	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2343	    &parm_buf, sizeof(struct sctp_paramhdr));
2344	if (phdr == NULL) {
2345		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2346		    __FUNCTION__);
2347		return NULL;
2348	}
2349	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2350	/* get the correlation address */
2351	switch (ptype) {
2352#ifdef INET6
2353	case SCTP_IPV6_ADDRESS:
2354		{
2355			/* ipv6 address param */
2356			struct sctp_ipv6addr_param *p6, p6_buf;
2357
2358			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2359				return NULL;
2360			}
2361			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2362			    offset + sizeof(struct sctp_asconf_chunk),
2363			    &p6_buf.ph, sizeof(*p6));
2364			if (p6 == NULL) {
2365				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2366				    __FUNCTION__);
2367				return (NULL);
2368			}
2369			sin6 = &remote_store.sin6;
2370			sin6->sin6_family = AF_INET6;
2371			sin6->sin6_len = sizeof(*sin6);
2372			sin6->sin6_port = sh->src_port;
2373			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2374			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2375				zero_address = 1;
2376			break;
2377		}
2378#endif
2379#ifdef INET
2380	case SCTP_IPV4_ADDRESS:
2381		{
2382			/* ipv4 address param */
2383			struct sctp_ipv4addr_param *p4, p4_buf;
2384
2385			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2386				return NULL;
2387			}
2388			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2389			    offset + sizeof(struct sctp_asconf_chunk),
2390			    &p4_buf.ph, sizeof(*p4));
2391			if (p4 == NULL) {
2392				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2393				    __FUNCTION__);
2394				return (NULL);
2395			}
2396			sin = &remote_store.sin;
2397			sin->sin_family = AF_INET;
2398			sin->sin_len = sizeof(*sin);
2399			sin->sin_port = sh->src_port;
2400			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2401			if (sin->sin_addr.s_addr == INADDR_ANY)
2402				zero_address = 1;
2403			break;
2404		}
2405#endif
2406	default:
2407		/* invalid address param type */
2408		return NULL;
2409	}
2410
2411	if (zero_address) {
2412		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2413		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2414		if (stcb != NULL) {
2415			SCTP_INP_DECR_REF(*inp_p);
2416		}
2417	} else {
2418		stcb = sctp_findassociation_ep_addr(inp_p,
2419		    &remote_store.sa, netp,
2420		    dst, NULL);
2421	}
2422	return (stcb);
2423}
2424
2425
2426/*
2427 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2428 * addresses. This way if we don't get a bind we by default pick a ephemeral
2429 * port with all addresses bound.
2430 */
2431int
2432sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2433{
2434	/*
2435	 * we get called when a new endpoint starts up. We need to allocate
2436	 * the sctp_inpcb structure from the zone and init it. Mark it as
2437	 * unbound and find a port that we can use as an ephemeral with
2438	 * INADDR_ANY. If the user binds later no problem we can then add in
2439	 * the specific addresses. And setup the default parameters for the
2440	 * EP.
2441	 */
2442	int i, error;
2443	struct sctp_inpcb *inp;
2444	struct sctp_pcb *m;
2445	struct timeval time;
2446	sctp_sharedkey_t *null_key;
2447
2448	error = 0;
2449
2450	SCTP_INP_INFO_WLOCK();
2451	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2452	if (inp == NULL) {
2453		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2454		SCTP_INP_INFO_WUNLOCK();
2455		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2456		return (ENOBUFS);
2457	}
2458	/* zap it */
2459	bzero(inp, sizeof(*inp));
2460
2461	/* bump generations */
2462	/* setup socket pointers */
2463	inp->sctp_socket = so;
2464	inp->ip_inp.inp.inp_socket = so;
2465	inp->ip_inp.inp.inp_cred = crhold(so->so_cred);
2466#ifdef INET6
2467	if (INP_SOCKAF(so) == AF_INET6) {
2468		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2469			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2470		}
2471		if (MODULE_GLOBAL(ip6_v6only)) {
2472			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2473		}
2474	}
2475#endif
2476	inp->sctp_associd_counter = 1;
2477	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2478	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2479	inp->max_cwnd = 0;
2480	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2481	inp->ecn_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_ecn_enable);
2482	inp->prsctp_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pr_enable);
2483	if (SCTP_BASE_SYSCTL(sctp_auth_disable)) {
2484		inp->auth_supported = 0;
2485	} else {
2486		inp->auth_supported = 1;
2487	}
2488	inp->asconf_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_asconf_enable);
2489	inp->reconfig_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_reconfig_enable);
2490	inp->nrsack_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_nrsack_enable);
2491	inp->pktdrop_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pktdrop_enable);
2492	/* init the small hash table we use to track asocid <-> tcb */
2493	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2494	if (inp->sctp_asocidhash == NULL) {
2495		crfree(inp->ip_inp.inp.inp_cred);
2496		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2497		SCTP_INP_INFO_WUNLOCK();
2498		return (ENOBUFS);
2499	}
2500#ifdef IPSEC
2501	{
2502		struct inpcbpolicy *pcb_sp = NULL;
2503
2504		error = ipsec_init_policy(so, &pcb_sp);
2505		/* Arrange to share the policy */
2506		inp->ip_inp.inp.inp_sp = pcb_sp;
2507		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2508	}
2509	if (error != 0) {
2510		crfree(inp->ip_inp.inp.inp_cred);
2511		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2512		SCTP_INP_INFO_WUNLOCK();
2513		return error;
2514	}
2515#endif				/* IPSEC */
2516	SCTP_INCR_EP_COUNT();
2517	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2518	SCTP_INP_INFO_WUNLOCK();
2519
2520	so->so_pcb = (caddr_t)inp;
2521
2522	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2523		/* UDP style socket */
2524		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2525		    SCTP_PCB_FLAGS_UNBOUND);
2526		/* Be sure it is NON-BLOCKING IO for UDP */
2527		/* SCTP_SET_SO_NBIO(so); */
2528	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2529		/* TCP style socket */
2530		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2531		    SCTP_PCB_FLAGS_UNBOUND);
2532		/* Be sure we have blocking IO by default */
2533		SCTP_CLEAR_SO_NBIO(so);
2534	} else {
2535		/*
2536		 * unsupported socket type (RAW, etc)- in case we missed it
2537		 * in protosw
2538		 */
2539		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2540		so->so_pcb = NULL;
2541		crfree(inp->ip_inp.inp.inp_cred);
2542		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2543		return (EOPNOTSUPP);
2544	}
2545	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2546		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2547		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2548	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2549		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2550		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2551	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2552		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2553		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2554	}
2555	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2556	    &inp->sctp_hashmark);
2557	if (inp->sctp_tcbhash == NULL) {
2558		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2559		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2560		so->so_pcb = NULL;
2561		crfree(inp->ip_inp.inp.inp_cred);
2562		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2563		return (ENOBUFS);
2564	}
2565	inp->def_vrf_id = vrf_id;
2566
2567	SCTP_INP_INFO_WLOCK();
2568	SCTP_INP_LOCK_INIT(inp);
2569	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2570	SCTP_INP_READ_INIT(inp);
2571	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2572	/* lock the new ep */
2573	SCTP_INP_WLOCK(inp);
2574
2575	/* add it to the info area */
2576	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2577	SCTP_INP_INFO_WUNLOCK();
2578
2579	TAILQ_INIT(&inp->read_queue);
2580	LIST_INIT(&inp->sctp_addr_list);
2581
2582	LIST_INIT(&inp->sctp_asoc_list);
2583
2584#ifdef SCTP_TRACK_FREED_ASOCS
2585	/* TEMP CODE */
2586	LIST_INIT(&inp->sctp_asoc_free_list);
2587#endif
2588	/* Init the timer structure for signature change */
2589	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2590	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2591
2592	/* now init the actual endpoint default data */
2593	m = &inp->sctp_ep;
2594
2595	/* setup the base timeout information */
2596	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2597	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2598	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2599	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2600	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2601	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2602	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2603	/* all max/min max are in ms */
2604	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2605	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2606	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2607	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2608	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2609	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2610	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2611	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2612	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2613	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2614	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2615	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2616	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2617
2618	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2619	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2620	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2621	/* number of streams to pre-open on a association */
2622	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2623
2624	/* Add adaptation cookie */
2625	m->adaptation_layer_indicator = 0;
2626	m->adaptation_layer_indicator_provided = 0;
2627
2628	/* seed random number generator */
2629	m->random_counter = 1;
2630	m->store_at = SCTP_SIGNATURE_SIZE;
2631	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2632	sctp_fill_random_store(m);
2633
2634	/* Minimum cookie size */
2635	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2636	    sizeof(struct sctp_state_cookie);
2637	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2638
2639	/* Setup the initial secret */
2640	(void)SCTP_GETTIME_TIMEVAL(&time);
2641	m->time_of_secret_change = time.tv_sec;
2642
2643	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2644		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2645	}
2646	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2647
2648	/* How long is a cookie good for ? */
2649	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2650	/*
2651	 * Initialize authentication parameters
2652	 */
2653	m->local_hmacs = sctp_default_supported_hmaclist();
2654	m->local_auth_chunks = sctp_alloc_chunklist();
2655	if (inp->asconf_supported) {
2656		sctp_auth_add_chunk(SCTP_ASCONF, m->local_auth_chunks);
2657		sctp_auth_add_chunk(SCTP_ASCONF_ACK, m->local_auth_chunks);
2658	}
2659	m->default_dscp = 0;
2660#ifdef INET6
2661	m->default_flowlabel = 0;
2662#endif
2663	m->port = 0;		/* encapsulation disabled by default */
2664	LIST_INIT(&m->shared_keys);
2665	/* add default NULL key as key id 0 */
2666	null_key = sctp_alloc_sharedkey();
2667	sctp_insert_sharedkey(&m->shared_keys, null_key);
2668	SCTP_INP_WUNLOCK(inp);
2669#ifdef SCTP_LOG_CLOSING
2670	sctp_log_closing(inp, NULL, 12);
2671#endif
2672	return (error);
2673}
2674
2675
2676void
2677sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2678    struct sctp_tcb *stcb)
2679{
2680	struct sctp_nets *net;
2681	uint16_t lport, rport;
2682	struct sctppcbhead *head;
2683	struct sctp_laddr *laddr, *oladdr;
2684
2685	atomic_add_int(&stcb->asoc.refcnt, 1);
2686	SCTP_TCB_UNLOCK(stcb);
2687	SCTP_INP_INFO_WLOCK();
2688	SCTP_INP_WLOCK(old_inp);
2689	SCTP_INP_WLOCK(new_inp);
2690	SCTP_TCB_LOCK(stcb);
2691	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2692
2693	new_inp->sctp_ep.time_of_secret_change =
2694	    old_inp->sctp_ep.time_of_secret_change;
2695	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2696	    sizeof(old_inp->sctp_ep.secret_key));
2697	new_inp->sctp_ep.current_secret_number =
2698	    old_inp->sctp_ep.current_secret_number;
2699	new_inp->sctp_ep.last_secret_number =
2700	    old_inp->sctp_ep.last_secret_number;
2701	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2702
2703	/* make it so new data pours into the new socket */
2704	stcb->sctp_socket = new_inp->sctp_socket;
2705	stcb->sctp_ep = new_inp;
2706
2707	/* Copy the port across */
2708	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2709	rport = stcb->rport;
2710	/* Pull the tcb from the old association */
2711	LIST_REMOVE(stcb, sctp_tcbhash);
2712	LIST_REMOVE(stcb, sctp_tcblist);
2713	if (stcb->asoc.in_asocid_hash) {
2714		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2715	}
2716	/* Now insert the new_inp into the TCP connected hash */
2717	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2718
2719	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2720	/* Its safe to access */
2721	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2722
2723	/* Now move the tcb into the endpoint list */
2724	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2725	/*
2726	 * Question, do we even need to worry about the ep-hash since we
2727	 * only have one connection? Probably not :> so lets get rid of it
2728	 * and not suck up any kernel memory in that.
2729	 */
2730	if (stcb->asoc.in_asocid_hash) {
2731		struct sctpasochead *lhd;
2732
2733		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2734		    new_inp->hashasocidmark)];
2735		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2736	}
2737	/* Ok. Let's restart timer. */
2738	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2739		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2740		    stcb, net);
2741	}
2742
2743	SCTP_INP_INFO_WUNLOCK();
2744	if (new_inp->sctp_tcbhash != NULL) {
2745		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2746		new_inp->sctp_tcbhash = NULL;
2747	}
2748	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2749		/* Subset bound, so copy in the laddr list from the old_inp */
2750		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2751			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2752			if (laddr == NULL) {
2753				/*
2754				 * Gak, what can we do? This assoc is really
2755				 * HOSED. We probably should send an abort
2756				 * here.
2757				 */
2758				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2759				continue;
2760			}
2761			SCTP_INCR_LADDR_COUNT();
2762			bzero(laddr, sizeof(*laddr));
2763			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2764			laddr->ifa = oladdr->ifa;
2765			atomic_add_int(&laddr->ifa->refcount, 1);
2766			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2767			    sctp_nxt_addr);
2768			new_inp->laddr_count++;
2769			if (oladdr == stcb->asoc.last_used_address) {
2770				stcb->asoc.last_used_address = laddr;
2771			}
2772		}
2773	}
2774	/*
2775	 * Now any running timers need to be adjusted since we really don't
2776	 * care if they are running or not just blast in the new_inp into
2777	 * all of them.
2778	 */
2779
2780	stcb->asoc.dack_timer.ep = (void *)new_inp;
2781	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2782	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2783	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2784	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2785	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2786	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2787	/* now what about the nets? */
2788	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2789		net->pmtu_timer.ep = (void *)new_inp;
2790		net->hb_timer.ep = (void *)new_inp;
2791		net->rxt_timer.ep = (void *)new_inp;
2792	}
2793	SCTP_INP_WUNLOCK(new_inp);
2794	SCTP_INP_WUNLOCK(old_inp);
2795}
2796
2797
2798
2799
2800/* sctp_ifap is used to bypass normal local address validation checks */
2801int
2802sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2803    struct sctp_ifa *sctp_ifap, struct thread *p)
2804{
2805	/* bind a ep to a socket address */
2806	struct sctppcbhead *head;
2807	struct sctp_inpcb *inp, *inp_tmp;
2808	struct inpcb *ip_inp;
2809	int port_reuse_active = 0;
2810	int bindall;
2811	uint16_t lport;
2812	int error;
2813	uint32_t vrf_id;
2814
2815	lport = 0;
2816	bindall = 1;
2817	inp = (struct sctp_inpcb *)so->so_pcb;
2818	ip_inp = (struct inpcb *)so->so_pcb;
2819#ifdef SCTP_DEBUG
2820	if (addr) {
2821		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2822		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2823		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2824		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2825	}
2826#endif
2827	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2828		/* already did a bind, subsequent binds NOT allowed ! */
2829		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2830		return (EINVAL);
2831	}
2832#ifdef INVARIANTS
2833	if (p == NULL)
2834		panic("null proc/thread");
2835#endif
2836	if (addr != NULL) {
2837		switch (addr->sa_family) {
2838#ifdef INET
2839		case AF_INET:
2840			{
2841				struct sockaddr_in *sin;
2842
2843				/* IPV6_V6ONLY socket? */
2844				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2845					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2846					return (EINVAL);
2847				}
2848				if (addr->sa_len != sizeof(*sin)) {
2849					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2850					return (EINVAL);
2851				}
2852				sin = (struct sockaddr_in *)addr;
2853				lport = sin->sin_port;
2854				/*
2855				 * For LOOPBACK the prison_local_ip4() call
2856				 * will transmute the ip address to the
2857				 * proper value.
2858				 */
2859				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2860					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2861					return (error);
2862				}
2863				if (sin->sin_addr.s_addr != INADDR_ANY) {
2864					bindall = 0;
2865				}
2866				break;
2867			}
2868#endif
2869#ifdef INET6
2870		case AF_INET6:
2871			{
2872				/*
2873				 * Only for pure IPv6 Address. (No IPv4
2874				 * Mapped!)
2875				 */
2876				struct sockaddr_in6 *sin6;
2877
2878				sin6 = (struct sockaddr_in6 *)addr;
2879
2880				if (addr->sa_len != sizeof(*sin6)) {
2881					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2882					return (EINVAL);
2883				}
2884				lport = sin6->sin6_port;
2885				/*
2886				 * For LOOPBACK the prison_local_ip6() call
2887				 * will transmute the ipv6 address to the
2888				 * proper value.
2889				 */
2890				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2891				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2892					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2893					return (error);
2894				}
2895				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2896					bindall = 0;
2897					/* KAME hack: embed scopeid */
2898					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2899						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2900						return (EINVAL);
2901					}
2902				}
2903				/* this must be cleared for ifa_ifwithaddr() */
2904				sin6->sin6_scope_id = 0;
2905				break;
2906			}
2907#endif
2908		default:
2909			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2910			return (EAFNOSUPPORT);
2911		}
2912	}
2913	SCTP_INP_INFO_WLOCK();
2914	SCTP_INP_WLOCK(inp);
2915	/* Setup a vrf_id to be the default for the non-bind-all case. */
2916	vrf_id = inp->def_vrf_id;
2917
2918	/* increase our count due to the unlock we do */
2919	SCTP_INP_INCR_REF(inp);
2920	if (lport) {
2921		/*
2922		 * Did the caller specify a port? if so we must see if an ep
2923		 * already has this one bound.
2924		 */
2925		/* got to be root to get at low ports */
2926		if (ntohs(lport) < IPPORT_RESERVED) {
2927			if (p && (error =
2928			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2929			    )) {
2930				SCTP_INP_DECR_REF(inp);
2931				SCTP_INP_WUNLOCK(inp);
2932				SCTP_INP_INFO_WUNLOCK();
2933				return (error);
2934			}
2935		}
2936		SCTP_INP_WUNLOCK(inp);
2937		if (bindall) {
2938			vrf_id = inp->def_vrf_id;
2939			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2940			if (inp_tmp != NULL) {
2941				/*
2942				 * lock guy returned and lower count note
2943				 * that we are not bound so inp_tmp should
2944				 * NEVER be inp. And it is this inp
2945				 * (inp_tmp) that gets the reference bump,
2946				 * so we must lower it.
2947				 */
2948				SCTP_INP_DECR_REF(inp_tmp);
2949				/* unlock info */
2950				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2951				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2952					/*
2953					 * Ok, must be one-2-one and
2954					 * allowing port re-use
2955					 */
2956					port_reuse_active = 1;
2957					goto continue_anyway;
2958				}
2959				SCTP_INP_DECR_REF(inp);
2960				SCTP_INP_INFO_WUNLOCK();
2961				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2962				return (EADDRINUSE);
2963			}
2964		} else {
2965			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2966			if (inp_tmp != NULL) {
2967				/*
2968				 * lock guy returned and lower count note
2969				 * that we are not bound so inp_tmp should
2970				 * NEVER be inp. And it is this inp
2971				 * (inp_tmp) that gets the reference bump,
2972				 * so we must lower it.
2973				 */
2974				SCTP_INP_DECR_REF(inp_tmp);
2975				/* unlock info */
2976				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2977				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2978					/*
2979					 * Ok, must be one-2-one and
2980					 * allowing port re-use
2981					 */
2982					port_reuse_active = 1;
2983					goto continue_anyway;
2984				}
2985				SCTP_INP_DECR_REF(inp);
2986				SCTP_INP_INFO_WUNLOCK();
2987				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2988				return (EADDRINUSE);
2989			}
2990		}
2991continue_anyway:
2992		SCTP_INP_WLOCK(inp);
2993		if (bindall) {
2994			/* verify that no lport is not used by a singleton */
2995			if ((port_reuse_active == 0) &&
2996			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
2997				/* Sorry someone already has this one bound */
2998				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2999				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
3000					port_reuse_active = 1;
3001				} else {
3002					SCTP_INP_DECR_REF(inp);
3003					SCTP_INP_WUNLOCK(inp);
3004					SCTP_INP_INFO_WUNLOCK();
3005					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3006					return (EADDRINUSE);
3007				}
3008			}
3009		}
3010	} else {
3011		uint16_t first, last, candidate;
3012		uint16_t count;
3013		int done;
3014
3015		if (ip_inp->inp_flags & INP_HIGHPORT) {
3016			first = MODULE_GLOBAL(ipport_hifirstauto);
3017			last = MODULE_GLOBAL(ipport_hilastauto);
3018		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3019			if (p && (error =
3020			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3021			    )) {
3022				SCTP_INP_DECR_REF(inp);
3023				SCTP_INP_WUNLOCK(inp);
3024				SCTP_INP_INFO_WUNLOCK();
3025				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3026				return (error);
3027			}
3028			first = MODULE_GLOBAL(ipport_lowfirstauto);
3029			last = MODULE_GLOBAL(ipport_lowlastauto);
3030		} else {
3031			first = MODULE_GLOBAL(ipport_firstauto);
3032			last = MODULE_GLOBAL(ipport_lastauto);
3033		}
3034		if (first > last) {
3035			uint16_t temp;
3036
3037			temp = first;
3038			first = last;
3039			last = temp;
3040		}
3041		count = last - first + 1;	/* number of candidates */
3042		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3043
3044		done = 0;
3045		while (!done) {
3046			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3047				done = 1;
3048			}
3049			if (!done) {
3050				if (--count == 0) {
3051					SCTP_INP_DECR_REF(inp);
3052					SCTP_INP_WUNLOCK(inp);
3053					SCTP_INP_INFO_WUNLOCK();
3054					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3055					return (EADDRINUSE);
3056				}
3057				if (candidate == last)
3058					candidate = first;
3059				else
3060					candidate = candidate + 1;
3061			}
3062		}
3063		lport = htons(candidate);
3064	}
3065	SCTP_INP_DECR_REF(inp);
3066	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3067	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3068		/*
3069		 * this really should not happen. The guy did a non-blocking
3070		 * bind and then did a close at the same time.
3071		 */
3072		SCTP_INP_WUNLOCK(inp);
3073		SCTP_INP_INFO_WUNLOCK();
3074		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3075		return (EINVAL);
3076	}
3077	/* ok we look clear to give out this port, so lets setup the binding */
3078	if (bindall) {
3079		/* binding to all addresses, so just set in the proper flags */
3080		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3081		/* set the automatic addr changes from kernel flag */
3082		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3083			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3084			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3085		} else {
3086			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3087			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3088		}
3089		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3090			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3091		} else {
3092			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3093		}
3094		/*
3095		 * set the automatic mobility_base from kernel flag (by
3096		 * micchie)
3097		 */
3098		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3099			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3100			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3101		} else {
3102			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3103			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3104		}
3105		/*
3106		 * set the automatic mobility_fasthandoff from kernel flag
3107		 * (by micchie)
3108		 */
3109		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3110			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3111			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3112		} else {
3113			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3114			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3115		}
3116	} else {
3117		/*
3118		 * bind specific, make sure flags is off and add a new
3119		 * address structure to the sctp_addr_list inside the ep
3120		 * structure.
3121		 *
3122		 * We will need to allocate one and insert it at the head. The
3123		 * socketopt call can just insert new addresses in there as
3124		 * well. It will also have to do the embed scope kame hack
3125		 * too (before adding).
3126		 */
3127		struct sctp_ifa *ifa;
3128		union sctp_sockstore store;
3129
3130		memset(&store, 0, sizeof(store));
3131		switch (addr->sa_family) {
3132#ifdef INET
3133		case AF_INET:
3134			memcpy(&store.sin, addr, sizeof(struct sockaddr_in));
3135			store.sin.sin_port = 0;
3136			break;
3137#endif
3138#ifdef INET6
3139		case AF_INET6:
3140			memcpy(&store.sin6, addr, sizeof(struct sockaddr_in6));
3141			store.sin6.sin6_port = 0;
3142			break;
3143#endif
3144		default:
3145			break;
3146		}
3147		/*
3148		 * first find the interface with the bound address need to
3149		 * zero out the port to find the address! yuck! can't do
3150		 * this earlier since need port for sctp_pcb_findep()
3151		 */
3152		if (sctp_ifap != NULL) {
3153			ifa = sctp_ifap;
3154		} else {
3155			/*
3156			 * Note for BSD we hit here always other O/S's will
3157			 * pass things in via the sctp_ifap argument
3158			 * (Panda).
3159			 */
3160			ifa = sctp_find_ifa_by_addr(&store.sa,
3161			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3162		}
3163		if (ifa == NULL) {
3164			/* Can't find an interface with that address */
3165			SCTP_INP_WUNLOCK(inp);
3166			SCTP_INP_INFO_WUNLOCK();
3167			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3168			return (EADDRNOTAVAIL);
3169		}
3170#ifdef INET6
3171		if (addr->sa_family == AF_INET6) {
3172			/* GAK, more FIXME IFA lock? */
3173			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3174				/* Can't bind a non-existent addr. */
3175				SCTP_INP_WUNLOCK(inp);
3176				SCTP_INP_INFO_WUNLOCK();
3177				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3178				return (EINVAL);
3179			}
3180		}
3181#endif
3182		/* we're not bound all */
3183		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3184		/* allow bindx() to send ASCONF's for binding changes */
3185		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3186		/* clear automatic addr changes from kernel flag */
3187		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3188
3189		/* add this address to the endpoint list */
3190		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3191		if (error != 0) {
3192			SCTP_INP_WUNLOCK(inp);
3193			SCTP_INP_INFO_WUNLOCK();
3194			return (error);
3195		}
3196		inp->laddr_count++;
3197	}
3198	/* find the bucket */
3199	if (port_reuse_active) {
3200		/* Put it into tcp 1-2-1 hash */
3201		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3202		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3203	} else {
3204		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3205	}
3206	/* put it in the bucket */
3207	LIST_INSERT_HEAD(head, inp, sctp_hash);
3208	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3209	    (void *)head, ntohs(lport), port_reuse_active);
3210	/* set in the port */
3211	inp->sctp_lport = lport;
3212
3213	/* turn off just the unbound flag */
3214	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3215	SCTP_INP_WUNLOCK(inp);
3216	SCTP_INP_INFO_WUNLOCK();
3217	return (0);
3218}
3219
3220
3221static void
3222sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3223{
3224	struct sctp_iterator *it, *nit;
3225
3226	/*
3227	 * We enter with the only the ITERATOR_LOCK in place and a write
3228	 * lock on the inp_info stuff.
3229	 */
3230	it = sctp_it_ctl.cur_it;
3231	if (it && (it->vn != curvnet)) {
3232		/* Its not looking at our VNET */
3233		return;
3234	}
3235	if (it && (it->inp == inp)) {
3236		/*
3237		 * This is tricky and we hold the iterator lock, but when it
3238		 * returns and gets the lock (when we release it) the
3239		 * iterator will try to operate on inp. We need to stop that
3240		 * from happening. But of course the iterator has a
3241		 * reference on the stcb and inp. We can mark it and it will
3242		 * stop.
3243		 *
3244		 * If its a single iterator situation, we set the end iterator
3245		 * flag. Otherwise we set the iterator to go to the next
3246		 * inp.
3247		 *
3248		 */
3249		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3250			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3251		} else {
3252			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3253		}
3254	}
3255	/*
3256	 * Now go through and remove any single reference to our inp that
3257	 * may be still pending on the list
3258	 */
3259	SCTP_IPI_ITERATOR_WQ_LOCK();
3260	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3261		if (it->vn != curvnet) {
3262			continue;
3263		}
3264		if (it->inp == inp) {
3265			/* This one points to me is it inp specific? */
3266			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3267				/* Remove and free this one */
3268				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3269				    it, sctp_nxt_itr);
3270				if (it->function_atend != NULL) {
3271					(*it->function_atend) (it->pointer, it->val);
3272				}
3273				SCTP_FREE(it, SCTP_M_ITER);
3274			} else {
3275				it->inp = LIST_NEXT(it->inp, sctp_list);
3276				if (it->inp) {
3277					SCTP_INP_INCR_REF(it->inp);
3278				}
3279			}
3280			/*
3281			 * When its put in the refcnt is incremented so decr
3282			 * it
3283			 */
3284			SCTP_INP_DECR_REF(inp);
3285		}
3286	}
3287	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3288}
3289
3290/* release sctp_inpcb unbind the port */
3291void
3292sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3293{
3294	/*
3295	 * Here we free a endpoint. We must find it (if it is in the Hash
3296	 * table) and remove it from there. Then we must also find it in the
3297	 * overall list and remove it from there. After all removals are
3298	 * complete then any timer has to be stopped. Then start the actual
3299	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3300	 * all associations. d) finally the ep itself.
3301	 */
3302	struct sctp_tcb *asoc, *nasoc;
3303	struct sctp_laddr *laddr, *nladdr;
3304	struct inpcb *ip_pcb;
3305	struct socket *so;
3306	int being_refed = 0;
3307	struct sctp_queued_to_read *sq, *nsq;
3308	int cnt;
3309	sctp_sharedkey_t *shared_key, *nshared_key;
3310
3311
3312#ifdef SCTP_LOG_CLOSING
3313	sctp_log_closing(inp, NULL, 0);
3314#endif
3315	SCTP_ITERATOR_LOCK();
3316	/* mark any iterators on the list or being processed */
3317	sctp_iterator_inp_being_freed(inp);
3318	SCTP_ITERATOR_UNLOCK();
3319	so = inp->sctp_socket;
3320	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3321		/* been here before.. eeks.. get out of here */
3322		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3323#ifdef SCTP_LOG_CLOSING
3324		sctp_log_closing(inp, NULL, 1);
3325#endif
3326		return;
3327	}
3328	SCTP_ASOC_CREATE_LOCK(inp);
3329	SCTP_INP_INFO_WLOCK();
3330
3331	SCTP_INP_WLOCK(inp);
3332	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3333		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3334		/* socket is gone, so no more wakeups allowed */
3335		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3336		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3337		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3338
3339	}
3340	/* First time through we have the socket lock, after that no more. */
3341	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3342	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3343
3344	if (inp->control) {
3345		sctp_m_freem(inp->control);
3346		inp->control = NULL;
3347	}
3348	if (inp->pkt) {
3349		sctp_m_freem(inp->pkt);
3350		inp->pkt = NULL;
3351	}
3352	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3353					 * here but I will be nice :> (i.e.
3354					 * ip_pcb = ep;) */
3355	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3356		int cnt_in_sd;
3357
3358		cnt_in_sd = 0;
3359		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3360			SCTP_TCB_LOCK(asoc);
3361			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3362				/* Skip guys being freed */
3363				cnt_in_sd++;
3364				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3365					/*
3366					 * Special case - we did not start a
3367					 * kill timer on the asoc due to it
3368					 * was not closed. So go ahead and
3369					 * start it now.
3370					 */
3371					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3372					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3373				}
3374				SCTP_TCB_UNLOCK(asoc);
3375				continue;
3376			}
3377			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3378			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3379			    (asoc->asoc.total_output_queue_size == 0)) {
3380				/*
3381				 * If we have data in queue, we don't want
3382				 * to just free since the app may have done,
3383				 * send()/close or connect/send/close. And
3384				 * it wants the data to get across first.
3385				 */
3386				/* Just abandon things in the front states */
3387				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3388				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3389					cnt_in_sd++;
3390				}
3391				continue;
3392			}
3393			/* Disconnect the socket please */
3394			asoc->sctp_socket = NULL;
3395			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3396			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3397			    (asoc->asoc.control_pdapi) ||
3398			    (asoc->asoc.size_on_all_streams > 0) ||
3399			    (so && (so->so_rcv.sb_cc > 0))) {
3400				/* Left with Data unread */
3401				struct mbuf *op_err;
3402
3403				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3404				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3405				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3406				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3407				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3408				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3409					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3410				}
3411				if (sctp_free_assoc(inp, asoc,
3412				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3413					cnt_in_sd++;
3414				}
3415				continue;
3416			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3417				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3418			    (asoc->asoc.stream_queue_cnt == 0)) {
3419				if (asoc->asoc.locked_on_sending) {
3420					goto abort_anyway;
3421				}
3422				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3423				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3424					struct sctp_nets *netp;
3425
3426					/*
3427					 * there is nothing queued to send,
3428					 * so I send shutdown
3429					 */
3430					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3431					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3432						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3433					}
3434					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3435					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3436					sctp_stop_timers_for_shutdown(asoc);
3437					if (asoc->asoc.alternate) {
3438						netp = asoc->asoc.alternate;
3439					} else {
3440						netp = asoc->asoc.primary_destination;
3441					}
3442					sctp_send_shutdown(asoc, netp);
3443					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3444					    netp);
3445					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3446					    asoc->asoc.primary_destination);
3447					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3448				}
3449			} else {
3450				/* mark into shutdown pending */
3451				struct sctp_stream_queue_pending *sp;
3452
3453				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3454				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3455				    asoc->asoc.primary_destination);
3456				if (asoc->asoc.locked_on_sending) {
3457					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3458					    sctp_streamhead);
3459					if (sp == NULL) {
3460						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3461						    (void *)asoc->asoc.locked_on_sending,
3462						    asoc->asoc.locked_on_sending->stream_no);
3463					} else {
3464						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3465							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3466					}
3467				}
3468				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3469				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3470				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3471					struct mbuf *op_err;
3472
3473			abort_anyway:
3474					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3475					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3476					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3477					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3478					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3479					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3480						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3481					}
3482					if (sctp_free_assoc(inp, asoc,
3483					    SCTP_PCBFREE_NOFORCE,
3484					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3485						cnt_in_sd++;
3486					}
3487					continue;
3488				} else {
3489					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3490				}
3491			}
3492			cnt_in_sd++;
3493			SCTP_TCB_UNLOCK(asoc);
3494		}
3495		/* now is there some left in our SHUTDOWN state? */
3496		if (cnt_in_sd) {
3497#ifdef SCTP_LOG_CLOSING
3498			sctp_log_closing(inp, NULL, 2);
3499#endif
3500			inp->sctp_socket = NULL;
3501			SCTP_INP_WUNLOCK(inp);
3502			SCTP_ASOC_CREATE_UNLOCK(inp);
3503			SCTP_INP_INFO_WUNLOCK();
3504			return;
3505		}
3506	}
3507	inp->sctp_socket = NULL;
3508	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3509	    SCTP_PCB_FLAGS_UNBOUND) {
3510		/*
3511		 * ok, this guy has been bound. It's port is somewhere in
3512		 * the SCTP_BASE_INFO(hash table). Remove it!
3513		 */
3514		LIST_REMOVE(inp, sctp_hash);
3515		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3516	}
3517	/*
3518	 * If there is a timer running to kill us, forget it, since it may
3519	 * have a contest on the INP lock.. which would cause us to die ...
3520	 */
3521	cnt = 0;
3522	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3523		SCTP_TCB_LOCK(asoc);
3524		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3525			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3526				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3527				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3528			}
3529			cnt++;
3530			SCTP_TCB_UNLOCK(asoc);
3531			continue;
3532		}
3533		/* Free associations that are NOT killing us */
3534		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3535		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3536			struct mbuf *op_err;
3537
3538			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3539			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3540			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3541			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3542		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3543			cnt++;
3544			SCTP_TCB_UNLOCK(asoc);
3545			continue;
3546		}
3547		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3548		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3549			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3550		}
3551		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3552			cnt++;
3553		}
3554	}
3555	if (cnt) {
3556		/* Ok we have someone out there that will kill us */
3557		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3558#ifdef SCTP_LOG_CLOSING
3559		sctp_log_closing(inp, NULL, 3);
3560#endif
3561		SCTP_INP_WUNLOCK(inp);
3562		SCTP_ASOC_CREATE_UNLOCK(inp);
3563		SCTP_INP_INFO_WUNLOCK();
3564		return;
3565	}
3566	if (SCTP_INP_LOCK_CONTENDED(inp))
3567		being_refed++;
3568	if (SCTP_INP_READ_CONTENDED(inp))
3569		being_refed++;
3570	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3571		being_refed++;
3572
3573	if ((inp->refcount) ||
3574	    (being_refed) ||
3575	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3576		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3577#ifdef SCTP_LOG_CLOSING
3578		sctp_log_closing(inp, NULL, 4);
3579#endif
3580		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3581		SCTP_INP_WUNLOCK(inp);
3582		SCTP_ASOC_CREATE_UNLOCK(inp);
3583		SCTP_INP_INFO_WUNLOCK();
3584		return;
3585	}
3586	inp->sctp_ep.signature_change.type = 0;
3587	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3588	/*
3589	 * Remove it from the list .. last thing we need a lock for.
3590	 */
3591	LIST_REMOVE(inp, sctp_list);
3592	SCTP_INP_WUNLOCK(inp);
3593	SCTP_ASOC_CREATE_UNLOCK(inp);
3594	SCTP_INP_INFO_WUNLOCK();
3595	/*
3596	 * Now we release all locks. Since this INP cannot be found anymore
3597	 * except possibly by the kill timer that might be running. We call
3598	 * the drain function here. It should hit the case were it sees the
3599	 * ACTIVE flag cleared and exit out freeing us to proceed and
3600	 * destroy everything.
3601	 */
3602	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3603		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3604	} else {
3605		/* Probably un-needed */
3606		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3607	}
3608
3609#ifdef SCTP_LOG_CLOSING
3610	sctp_log_closing(inp, NULL, 5);
3611#endif
3612
3613
3614	if ((inp->sctp_asocidhash) != NULL) {
3615		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3616		inp->sctp_asocidhash = NULL;
3617	}
3618	/* sa_ignore FREED_MEMORY */
3619	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3620		/* Its only abandoned if it had data left */
3621		if (sq->length)
3622			SCTP_STAT_INCR(sctps_left_abandon);
3623
3624		TAILQ_REMOVE(&inp->read_queue, sq, next);
3625		sctp_free_remote_addr(sq->whoFrom);
3626		if (so)
3627			so->so_rcv.sb_cc -= sq->length;
3628		if (sq->data) {
3629			sctp_m_freem(sq->data);
3630			sq->data = NULL;
3631		}
3632		/*
3633		 * no need to free the net count, since at this point all
3634		 * assoc's are gone.
3635		 */
3636		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3637		SCTP_DECR_READQ_COUNT();
3638	}
3639	/* Now the sctp_pcb things */
3640	/*
3641	 * free each asoc if it is not already closed/free. we can't use the
3642	 * macro here since le_next will get freed as part of the
3643	 * sctp_free_assoc() call.
3644	 */
3645	if (so) {
3646#ifdef IPSEC
3647		ipsec_delete_pcbpolicy(ip_pcb);
3648#endif				/* IPSEC */
3649
3650		/* Unlocks not needed since the socket is gone now */
3651	}
3652	if (ip_pcb->inp_options) {
3653		(void)sctp_m_free(ip_pcb->inp_options);
3654		ip_pcb->inp_options = 0;
3655	}
3656#ifdef INET6
3657	if (ip_pcb->inp_vflag & INP_IPV6) {
3658		struct in6pcb *in6p;
3659
3660		in6p = (struct in6pcb *)inp;
3661		ip6_freepcbopts(in6p->in6p_outputopts);
3662	}
3663#endif				/* INET6 */
3664	ip_pcb->inp_vflag = 0;
3665	/* free up authentication fields */
3666	if (inp->sctp_ep.local_auth_chunks != NULL)
3667		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3668	if (inp->sctp_ep.local_hmacs != NULL)
3669		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3670
3671	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3672		LIST_REMOVE(shared_key, next);
3673		sctp_free_sharedkey(shared_key);
3674		/* sa_ignore FREED_MEMORY */
3675	}
3676
3677	/*
3678	 * if we have an address list the following will free the list of
3679	 * ifaddr's that are set into this ep. Again macro limitations here,
3680	 * since the LIST_FOREACH could be a bad idea.
3681	 */
3682	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3683		sctp_remove_laddr(laddr);
3684	}
3685
3686#ifdef SCTP_TRACK_FREED_ASOCS
3687	/* TEMP CODE */
3688	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3689		LIST_REMOVE(asoc, sctp_tcblist);
3690		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3691		SCTP_DECR_ASOC_COUNT();
3692	}
3693	/* *** END TEMP CODE *** */
3694#endif
3695	/* Now lets see about freeing the EP hash table. */
3696	if (inp->sctp_tcbhash != NULL) {
3697		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3698		inp->sctp_tcbhash = NULL;
3699	}
3700	/* Now we must put the ep memory back into the zone pool */
3701	crfree(inp->ip_inp.inp.inp_cred);
3702	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3703	SCTP_INP_LOCK_DESTROY(inp);
3704	SCTP_INP_READ_DESTROY(inp);
3705	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3706	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3707	SCTP_DECR_EP_COUNT();
3708}
3709
3710
3711struct sctp_nets *
3712sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3713{
3714	struct sctp_nets *net;
3715
3716	/* locate the address */
3717	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3718		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3719			return (net);
3720	}
3721	return (NULL);
3722}
3723
3724
3725int
3726sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3727{
3728	struct sctp_ifa *sctp_ifa;
3729
3730	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3731	if (sctp_ifa) {
3732		return (1);
3733	} else {
3734		return (0);
3735	}
3736}
3737
3738/*
3739 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3740 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3741 * stats of stuff.
3742 */
3743int
3744sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3745    struct sctp_nets **netp, int set_scope, int from)
3746{
3747	/*
3748	 * The following is redundant to the same lines in the
3749	 * sctp_aloc_assoc() but is needed since others call the add address
3750	 * function
3751	 */
3752	struct sctp_nets *net, *netfirst;
3753	int addr_inscope;
3754
3755	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3756	    from);
3757	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3758
3759	netfirst = sctp_findnet(stcb, newaddr);
3760	if (netfirst) {
3761		/*
3762		 * Lie and return ok, we don't want to make the association
3763		 * go away for this behavior. It will happen in the TCP
3764		 * model in a connected socket. It does not reach the hash
3765		 * table until after the association is built so it can't be
3766		 * found. Mark as reachable, since the initial creation will
3767		 * have been cleared and the NOT_IN_ASSOC flag will have
3768		 * been added... and we don't want to end up removing it
3769		 * back out.
3770		 */
3771		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3772			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3773			    SCTP_ADDR_UNCONFIRMED);
3774		} else {
3775			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3776		}
3777
3778		return (0);
3779	}
3780	addr_inscope = 1;
3781	switch (newaddr->sa_family) {
3782#ifdef INET
3783	case AF_INET:
3784		{
3785			struct sockaddr_in *sin;
3786
3787			sin = (struct sockaddr_in *)newaddr;
3788			if (sin->sin_addr.s_addr == 0) {
3789				/* Invalid address */
3790				return (-1);
3791			}
3792			/* zero out the bzero area */
3793			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3794
3795			/* assure len is set */
3796			sin->sin_len = sizeof(struct sockaddr_in);
3797			if (set_scope) {
3798#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3799				stcb->asoc.scope.ipv4_local_scope = 1;
3800#else
3801				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3802					stcb->asoc.scope.ipv4_local_scope = 1;
3803				}
3804#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3805			} else {
3806				/* Validate the address is in scope */
3807				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3808				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3809					addr_inscope = 0;
3810				}
3811			}
3812			break;
3813		}
3814#endif
3815#ifdef INET6
3816	case AF_INET6:
3817		{
3818			struct sockaddr_in6 *sin6;
3819
3820			sin6 = (struct sockaddr_in6 *)newaddr;
3821			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3822				/* Invalid address */
3823				return (-1);
3824			}
3825			/* assure len is set */
3826			sin6->sin6_len = sizeof(struct sockaddr_in6);
3827			if (set_scope) {
3828				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3829					stcb->asoc.scope.loopback_scope = 1;
3830					stcb->asoc.scope.local_scope = 0;
3831					stcb->asoc.scope.ipv4_local_scope = 1;
3832					stcb->asoc.scope.site_scope = 1;
3833				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3834					/*
3835					 * If the new destination is a
3836					 * LINK_LOCAL we must have common
3837					 * site scope. Don't set the local
3838					 * scope since we may not share all
3839					 * links, only loopback can do this.
3840					 * Links on the local network would
3841					 * also be on our private network
3842					 * for v4 too.
3843					 */
3844					stcb->asoc.scope.ipv4_local_scope = 1;
3845					stcb->asoc.scope.site_scope = 1;
3846				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3847					/*
3848					 * If the new destination is
3849					 * SITE_LOCAL then we must have site
3850					 * scope in common.
3851					 */
3852					stcb->asoc.scope.site_scope = 1;
3853				}
3854			} else {
3855				/* Validate the address is in scope */
3856				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3857				    (stcb->asoc.scope.loopback_scope == 0)) {
3858					addr_inscope = 0;
3859				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3860				    (stcb->asoc.scope.local_scope == 0)) {
3861					addr_inscope = 0;
3862				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3863				    (stcb->asoc.scope.site_scope == 0)) {
3864					addr_inscope = 0;
3865				}
3866			}
3867			break;
3868		}
3869#endif
3870	default:
3871		/* not supported family type */
3872		return (-1);
3873	}
3874	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3875	if (net == NULL) {
3876		return (-1);
3877	}
3878	SCTP_INCR_RADDR_COUNT();
3879	bzero(net, sizeof(struct sctp_nets));
3880	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3881	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3882	switch (newaddr->sa_family) {
3883#ifdef INET
3884	case AF_INET:
3885		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3886		break;
3887#endif
3888#ifdef INET6
3889	case AF_INET6:
3890		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3891		break;
3892#endif
3893	default:
3894		break;
3895	}
3896	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3897	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3898		stcb->asoc.scope.loopback_scope = 1;
3899		stcb->asoc.scope.ipv4_local_scope = 1;
3900		stcb->asoc.scope.local_scope = 0;
3901		stcb->asoc.scope.site_scope = 1;
3902		addr_inscope = 1;
3903	}
3904	net->failure_threshold = stcb->asoc.def_net_failure;
3905	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3906	if (addr_inscope == 0) {
3907		net->dest_state = (SCTP_ADDR_REACHABLE |
3908		    SCTP_ADDR_OUT_OF_SCOPE);
3909	} else {
3910		if (from == SCTP_ADDR_IS_CONFIRMED)
3911			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3912			net->dest_state = SCTP_ADDR_REACHABLE;
3913		else
3914			net->dest_state = SCTP_ADDR_REACHABLE |
3915			    SCTP_ADDR_UNCONFIRMED;
3916	}
3917	/*
3918	 * We set this to 0, the timer code knows that this means its an
3919	 * initial value
3920	 */
3921	net->rto_needed = 1;
3922	net->RTO = 0;
3923	net->RTO_measured = 0;
3924	stcb->asoc.numnets++;
3925	net->ref_count = 1;
3926	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3927	net->port = stcb->asoc.port;
3928	net->dscp = stcb->asoc.default_dscp;
3929#ifdef INET6
3930	net->flowlabel = stcb->asoc.default_flowlabel;
3931#endif
3932	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3933		net->dest_state |= SCTP_ADDR_NOHB;
3934	} else {
3935		net->dest_state &= ~SCTP_ADDR_NOHB;
3936	}
3937	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3938		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3939	} else {
3940		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3941	}
3942	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3943	/* Init the timer structure */
3944	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3945	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3946	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3947
3948	/* Now generate a route for this guy */
3949#ifdef INET6
3950	/* KAME hack: embed scopeid */
3951	if (newaddr->sa_family == AF_INET6) {
3952		struct sockaddr_in6 *sin6;
3953
3954		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3955		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3956		sin6->sin6_scope_id = 0;
3957	}
3958#endif
3959	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3960
3961	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3962		/* Get source address */
3963		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3964		    stcb,
3965		    (sctp_route_t *) & net->ro,
3966		    net,
3967		    0,
3968		    stcb->asoc.vrf_id);
3969		if (net->ro._s_addr != NULL) {
3970			net->src_addr_selected = 1;
3971			/* Now get the interface MTU */
3972			if (net->ro._s_addr->ifn_p != NULL) {
3973				net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3974			}
3975		} else {
3976			net->src_addr_selected = 0;
3977		}
3978		if (net->mtu > 0) {
3979			uint32_t rmtu;
3980
3981			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3982			if (rmtu == 0) {
3983				/*
3984				 * Start things off to match mtu of
3985				 * interface please.
3986				 */
3987				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3988				    net->ro.ro_rt, net->mtu);
3989			} else {
3990				/*
3991				 * we take the route mtu over the interface,
3992				 * since the route may be leading out the
3993				 * loopback, or a different interface.
3994				 */
3995				net->mtu = rmtu;
3996			}
3997		}
3998	} else {
3999		net->src_addr_selected = 0;
4000	}
4001	if (net->mtu == 0) {
4002		switch (newaddr->sa_family) {
4003#ifdef INET
4004		case AF_INET:
4005			net->mtu = SCTP_DEFAULT_MTU;
4006			break;
4007#endif
4008#ifdef INET6
4009		case AF_INET6:
4010			net->mtu = 1280;
4011			break;
4012#endif
4013		default:
4014			break;
4015		}
4016	}
4017#if defined(INET) || defined(INET6)
4018	if (net->port) {
4019		net->mtu -= (uint32_t) sizeof(struct udphdr);
4020	}
4021#endif
4022	if (from == SCTP_ALLOC_ASOC) {
4023		stcb->asoc.smallest_mtu = net->mtu;
4024	}
4025	if (stcb->asoc.smallest_mtu > net->mtu) {
4026		stcb->asoc.smallest_mtu = net->mtu;
4027	}
4028#ifdef INET6
4029	if (newaddr->sa_family == AF_INET6) {
4030		struct sockaddr_in6 *sin6;
4031
4032		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4033		(void)sa6_recoverscope(sin6);
4034	}
4035#endif
4036
4037	/* JRS - Use the congestion control given in the CC module */
4038	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4039		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4040
4041	/*
4042	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4043	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4044	 */
4045	net->find_pseudo_cumack = 1;
4046	net->find_rtx_pseudo_cumack = 1;
4047	/* Choose an initial flowid. */
4048	net->flowid = stcb->asoc.my_vtag ^
4049	    ntohs(stcb->rport) ^
4050	    ntohs(stcb->sctp_ep->sctp_lport);
4051	net->flowtype = M_HASHTYPE_OPAQUE;
4052	if (netp) {
4053		*netp = net;
4054	}
4055	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4056	if (net->ro.ro_rt == NULL) {
4057		/* Since we have no route put it at the back */
4058		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4059	} else if (netfirst == NULL) {
4060		/* We are the first one in the pool. */
4061		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4062	} else if (netfirst->ro.ro_rt == NULL) {
4063		/*
4064		 * First one has NO route. Place this one ahead of the first
4065		 * one.
4066		 */
4067		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4068	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4069		/*
4070		 * This one has a different interface than the one at the
4071		 * top of the list. Place it ahead.
4072		 */
4073		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4074	} else {
4075		/*
4076		 * Ok we have the same interface as the first one. Move
4077		 * forward until we find either a) one with a NULL route...
4078		 * insert ahead of that b) one with a different ifp.. insert
4079		 * after that. c) end of the list.. insert at the tail.
4080		 */
4081		struct sctp_nets *netlook;
4082
4083		do {
4084			netlook = TAILQ_NEXT(netfirst, sctp_next);
4085			if (netlook == NULL) {
4086				/* End of the list */
4087				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4088				break;
4089			} else if (netlook->ro.ro_rt == NULL) {
4090				/* next one has NO route */
4091				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4092				break;
4093			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4094				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4095				    net, sctp_next);
4096				break;
4097			}
4098			/* Shift forward */
4099			netfirst = netlook;
4100		} while (netlook != NULL);
4101	}
4102
4103	/* got to have a primary set */
4104	if (stcb->asoc.primary_destination == 0) {
4105		stcb->asoc.primary_destination = net;
4106	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4107		    (net->ro.ro_rt) &&
4108	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4109		/* No route to current primary adopt new primary */
4110		stcb->asoc.primary_destination = net;
4111	}
4112	/* Validate primary is first */
4113	net = TAILQ_FIRST(&stcb->asoc.nets);
4114	if ((net != stcb->asoc.primary_destination) &&
4115	    (stcb->asoc.primary_destination)) {
4116		/*
4117		 * first one on the list is NOT the primary sctp_cmpaddr()
4118		 * is much more efficient if the primary is the first on the
4119		 * list, make it so.
4120		 */
4121		TAILQ_REMOVE(&stcb->asoc.nets,
4122		    stcb->asoc.primary_destination, sctp_next);
4123		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4124		    stcb->asoc.primary_destination, sctp_next);
4125	}
4126	return (0);
4127}
4128
4129
4130static uint32_t
4131sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4132{
4133	uint32_t id;
4134	struct sctpasochead *head;
4135	struct sctp_tcb *lstcb;
4136
4137	SCTP_INP_WLOCK(inp);
4138try_again:
4139	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4140		/* TSNH */
4141		SCTP_INP_WUNLOCK(inp);
4142		return (0);
4143	}
4144	/*
4145	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4146	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4147	 */
4148	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4149		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4150	}
4151	id = inp->sctp_associd_counter;
4152	inp->sctp_associd_counter++;
4153	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4154	if (lstcb) {
4155		goto try_again;
4156	}
4157	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4158	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4159	stcb->asoc.in_asocid_hash = 1;
4160	SCTP_INP_WUNLOCK(inp);
4161	return id;
4162}
4163
4164/*
4165 * allocate an association and add it to the endpoint. The caller must be
4166 * careful to add all additional addresses once they are know right away or
4167 * else the assoc will be may experience a blackout scenario.
4168 */
4169struct sctp_tcb *
4170sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4171    int *error, uint32_t override_tag, uint32_t vrf_id,
4172    struct thread *p
4173)
4174{
4175	/* note the p argument is only valid in unbound sockets */
4176
4177	struct sctp_tcb *stcb;
4178	struct sctp_association *asoc;
4179	struct sctpasochead *head;
4180	uint16_t rport;
4181	int err;
4182
4183	/*
4184	 * Assumption made here: Caller has done a
4185	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4186	 * address does not exist already.
4187	 */
4188	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4189		/* Hit max assoc, sorry no more */
4190		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4191		*error = ENOBUFS;
4192		return (NULL);
4193	}
4194	if (firstaddr == NULL) {
4195		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4196		*error = EINVAL;
4197		return (NULL);
4198	}
4199	SCTP_INP_RLOCK(inp);
4200	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4201	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4202	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4203		/*
4204		 * If its in the TCP pool, its NOT allowed to create an
4205		 * association. The parent listener needs to call
4206		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4207		 * off, or connected one does this.. its an error.
4208		 */
4209		SCTP_INP_RUNLOCK(inp);
4210		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4211		*error = EINVAL;
4212		return (NULL);
4213	}
4214	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4215	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4216		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4217		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4218			SCTP_INP_RUNLOCK(inp);
4219			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4220			*error = EINVAL;
4221			return (NULL);
4222		}
4223	}
4224	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4225#ifdef SCTP_DEBUG
4226	if (firstaddr) {
4227		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4228		switch (firstaddr->sa_family) {
4229#ifdef INET
4230		case AF_INET:
4231			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4232			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4233			break;
4234#endif
4235#ifdef INET6
4236		case AF_INET6:
4237			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4238			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4239			break;
4240#endif
4241		default:
4242			break;
4243		}
4244	} else {
4245		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4246	}
4247#endif				/* SCTP_DEBUG */
4248	switch (firstaddr->sa_family) {
4249#ifdef INET
4250	case AF_INET:
4251		{
4252			struct sockaddr_in *sin;
4253
4254			sin = (struct sockaddr_in *)firstaddr;
4255			if ((ntohs(sin->sin_port) == 0) ||
4256			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4257			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4258			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4259				/* Invalid address */
4260				SCTP_INP_RUNLOCK(inp);
4261				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4262				*error = EINVAL;
4263				return (NULL);
4264			}
4265			rport = sin->sin_port;
4266			break;
4267		}
4268#endif
4269#ifdef INET6
4270	case AF_INET6:
4271		{
4272			struct sockaddr_in6 *sin6;
4273
4274			sin6 = (struct sockaddr_in6 *)firstaddr;
4275			if ((ntohs(sin6->sin6_port) == 0) ||
4276			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4277			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4278				/* Invalid address */
4279				SCTP_INP_RUNLOCK(inp);
4280				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4281				*error = EINVAL;
4282				return (NULL);
4283			}
4284			rport = sin6->sin6_port;
4285			break;
4286		}
4287#endif
4288	default:
4289		/* not supported family type */
4290		SCTP_INP_RUNLOCK(inp);
4291		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4292		*error = EINVAL;
4293		return (NULL);
4294	}
4295	SCTP_INP_RUNLOCK(inp);
4296	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4297		/*
4298		 * If you have not performed a bind, then we need to do the
4299		 * ephemeral bind for you.
4300		 */
4301		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4302		    (struct sockaddr *)NULL,
4303		    (struct sctp_ifa *)NULL,
4304		    p
4305		    ))) {
4306			/* bind error, probably perm */
4307			*error = err;
4308			return (NULL);
4309		}
4310	}
4311	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4312	if (stcb == NULL) {
4313		/* out of memory? */
4314		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4315		*error = ENOMEM;
4316		return (NULL);
4317	}
4318	SCTP_INCR_ASOC_COUNT();
4319
4320	bzero(stcb, sizeof(*stcb));
4321	asoc = &stcb->asoc;
4322
4323	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4324	SCTP_TCB_LOCK_INIT(stcb);
4325	SCTP_TCB_SEND_LOCK_INIT(stcb);
4326	stcb->rport = rport;
4327	/* setup back pointer's */
4328	stcb->sctp_ep = inp;
4329	stcb->sctp_socket = inp->sctp_socket;
4330	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4331		/* failed */
4332		SCTP_TCB_LOCK_DESTROY(stcb);
4333		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4334		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4335		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4336		SCTP_DECR_ASOC_COUNT();
4337		*error = err;
4338		return (NULL);
4339	}
4340	/* and the port */
4341	SCTP_INP_INFO_WLOCK();
4342	SCTP_INP_WLOCK(inp);
4343	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4344		/* inpcb freed while alloc going on */
4345		SCTP_TCB_LOCK_DESTROY(stcb);
4346		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4347		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4348		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4349		SCTP_INP_WUNLOCK(inp);
4350		SCTP_INP_INFO_WUNLOCK();
4351		SCTP_DECR_ASOC_COUNT();
4352		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4353		*error = EINVAL;
4354		return (NULL);
4355	}
4356	SCTP_TCB_LOCK(stcb);
4357
4358	/* now that my_vtag is set, add it to the hash */
4359	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4360	/* put it in the bucket in the vtag hash of assoc's for the system */
4361	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4362	SCTP_INP_INFO_WUNLOCK();
4363
4364	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4365		/* failure.. memory error? */
4366		if (asoc->strmout) {
4367			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4368			asoc->strmout = NULL;
4369		}
4370		if (asoc->mapping_array) {
4371			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4372			asoc->mapping_array = NULL;
4373		}
4374		if (asoc->nr_mapping_array) {
4375			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4376			asoc->nr_mapping_array = NULL;
4377		}
4378		SCTP_DECR_ASOC_COUNT();
4379		SCTP_TCB_UNLOCK(stcb);
4380		SCTP_TCB_LOCK_DESTROY(stcb);
4381		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4382		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4383		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4384		SCTP_INP_WUNLOCK(inp);
4385		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4386		*error = ENOBUFS;
4387		return (NULL);
4388	}
4389	/* Init all the timers */
4390	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4391	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4392	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4393	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4394	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4395	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4396	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4397
4398	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4399	/* now file the port under the hash as well */
4400	if (inp->sctp_tcbhash != NULL) {
4401		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4402		    inp->sctp_hashmark)];
4403		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4404	}
4405	SCTP_INP_WUNLOCK(inp);
4406	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4407	return (stcb);
4408}
4409
4410
4411void
4412sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4413{
4414	struct sctp_association *asoc;
4415
4416	asoc = &stcb->asoc;
4417	asoc->numnets--;
4418	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4419	if (net == asoc->primary_destination) {
4420		/* Reset primary */
4421		struct sctp_nets *lnet;
4422
4423		lnet = TAILQ_FIRST(&asoc->nets);
4424		/*
4425		 * Mobility adaptation Ideally, if deleted destination is
4426		 * the primary, it becomes a fast retransmission trigger by
4427		 * the subsequent SET PRIMARY. (by micchie)
4428		 */
4429		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4430		    SCTP_MOBILITY_BASE) ||
4431		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4432		    SCTP_MOBILITY_FASTHANDOFF)) {
4433			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4434			if (asoc->deleted_primary != NULL) {
4435				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4436				goto out;
4437			}
4438			asoc->deleted_primary = net;
4439			atomic_add_int(&net->ref_count, 1);
4440			memset(&net->lastsa, 0, sizeof(net->lastsa));
4441			memset(&net->lastsv, 0, sizeof(net->lastsv));
4442			sctp_mobility_feature_on(stcb->sctp_ep,
4443			    SCTP_MOBILITY_PRIM_DELETED);
4444			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4445			    stcb->sctp_ep, stcb, NULL);
4446		}
4447out:
4448		/* Try to find a confirmed primary */
4449		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4450	}
4451	if (net == asoc->last_data_chunk_from) {
4452		/* Reset primary */
4453		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4454	}
4455	if (net == asoc->last_control_chunk_from) {
4456		/* Clear net */
4457		asoc->last_control_chunk_from = NULL;
4458	}
4459	if (net == stcb->asoc.alternate) {
4460		sctp_free_remote_addr(stcb->asoc.alternate);
4461		stcb->asoc.alternate = NULL;
4462	}
4463	sctp_free_remote_addr(net);
4464}
4465
4466/*
4467 * remove a remote endpoint address from an association, it will fail if the
4468 * address does not exist.
4469 */
4470int
4471sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4472{
4473	/*
4474	 * Here we need to remove a remote address. This is quite simple, we
4475	 * first find it in the list of address for the association
4476	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4477	 * on that item. Note we do not allow it to be removed if there are
4478	 * no other addresses.
4479	 */
4480	struct sctp_association *asoc;
4481	struct sctp_nets *net, *nnet;
4482
4483	asoc = &stcb->asoc;
4484
4485	/* locate the address */
4486	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4487		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4488			continue;
4489		}
4490		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4491		    remaddr)) {
4492			/* we found the guy */
4493			if (asoc->numnets < 2) {
4494				/* Must have at LEAST two remote addresses */
4495				return (-1);
4496			} else {
4497				sctp_remove_net(stcb, net);
4498				return (0);
4499			}
4500		}
4501	}
4502	/* not found. */
4503	return (-2);
4504}
4505
4506void
4507sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4508{
4509	struct sctpvtaghead *chain;
4510	struct sctp_tagblock *twait_block;
4511	int found = 0;
4512	int i;
4513
4514	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4515	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4516		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4517			if ((twait_block->vtag_block[i].v_tag == tag) &&
4518			    (twait_block->vtag_block[i].lport == lport) &&
4519			    (twait_block->vtag_block[i].rport == rport)) {
4520				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4521				twait_block->vtag_block[i].v_tag = 0;
4522				twait_block->vtag_block[i].lport = 0;
4523				twait_block->vtag_block[i].rport = 0;
4524				found = 1;
4525				break;
4526			}
4527		}
4528		if (found)
4529			break;
4530	}
4531}
4532
4533int
4534sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4535{
4536	struct sctpvtaghead *chain;
4537	struct sctp_tagblock *twait_block;
4538	int found = 0;
4539	int i;
4540
4541	SCTP_INP_INFO_WLOCK();
4542	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4543	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4544		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4545			if ((twait_block->vtag_block[i].v_tag == tag) &&
4546			    (twait_block->vtag_block[i].lport == lport) &&
4547			    (twait_block->vtag_block[i].rport == rport)) {
4548				found = 1;
4549				break;
4550			}
4551		}
4552		if (found)
4553			break;
4554	}
4555	SCTP_INP_INFO_WUNLOCK();
4556	return (found);
4557}
4558
4559
4560void
4561sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4562{
4563	struct sctpvtaghead *chain;
4564	struct sctp_tagblock *twait_block;
4565	struct timeval now;
4566	int set, i;
4567
4568	if (time == 0) {
4569		/* Its disabled */
4570		return;
4571	}
4572	(void)SCTP_GETTIME_TIMEVAL(&now);
4573	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4574	set = 0;
4575	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4576		/* Block(s) present, lets find space, and expire on the fly */
4577		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4578			if ((twait_block->vtag_block[i].v_tag == 0) &&
4579			    !set) {
4580				twait_block->vtag_block[i].tv_sec_at_expire =
4581				    now.tv_sec + time;
4582				twait_block->vtag_block[i].v_tag = tag;
4583				twait_block->vtag_block[i].lport = lport;
4584				twait_block->vtag_block[i].rport = rport;
4585				set = 1;
4586			} else if ((twait_block->vtag_block[i].v_tag) &&
4587			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4588				/* Audit expires this guy */
4589				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4590				twait_block->vtag_block[i].v_tag = 0;
4591				twait_block->vtag_block[i].lport = 0;
4592				twait_block->vtag_block[i].rport = 0;
4593				if (set == 0) {
4594					/* Reuse it for my new tag */
4595					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4596					twait_block->vtag_block[i].v_tag = tag;
4597					twait_block->vtag_block[i].lport = lport;
4598					twait_block->vtag_block[i].rport = rport;
4599					set = 1;
4600				}
4601			}
4602		}
4603		if (set) {
4604			/*
4605			 * We only do up to the block where we can place our
4606			 * tag for audits
4607			 */
4608			break;
4609		}
4610	}
4611	/* Need to add a new block to chain */
4612	if (!set) {
4613		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4614		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4615		if (twait_block == NULL) {
4616#ifdef INVARIANTS
4617			panic("Can not alloc tagblock");
4618#endif
4619			return;
4620		}
4621		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4622		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4623		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4624		twait_block->vtag_block[0].v_tag = tag;
4625		twait_block->vtag_block[0].lport = lport;
4626		twait_block->vtag_block[0].rport = rport;
4627	}
4628}
4629
4630
4631
4632/*-
4633 * Free the association after un-hashing the remote port. This
4634 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4635 * expect that the input to this function IS a locked TCB.
4636 * It will return 0, if it did NOT destroy the association (instead
4637 * it unlocks it. It will return NON-zero if it either destroyed the
4638 * association OR the association is already destroyed.
4639 */
4640int
4641sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4642{
4643	int i;
4644	struct sctp_association *asoc;
4645	struct sctp_nets *net, *nnet;
4646	struct sctp_laddr *laddr, *naddr;
4647	struct sctp_tmit_chunk *chk, *nchk;
4648	struct sctp_asconf_addr *aparam, *naparam;
4649	struct sctp_asconf_ack *aack, *naack;
4650	struct sctp_stream_reset_list *strrst, *nstrrst;
4651	struct sctp_queued_to_read *sq, *nsq;
4652	struct sctp_stream_queue_pending *sp, *nsp;
4653	sctp_sharedkey_t *shared_key, *nshared_key;
4654	struct socket *so;
4655
4656	/* first, lets purge the entry from the hash table. */
4657
4658#ifdef SCTP_LOG_CLOSING
4659	sctp_log_closing(inp, stcb, 6);
4660#endif
4661	if (stcb->asoc.state == 0) {
4662#ifdef SCTP_LOG_CLOSING
4663		sctp_log_closing(inp, NULL, 7);
4664#endif
4665		/* there is no asoc, really TSNH :-0 */
4666		return (1);
4667	}
4668	if (stcb->asoc.alternate) {
4669		sctp_free_remote_addr(stcb->asoc.alternate);
4670		stcb->asoc.alternate = NULL;
4671	}
4672	/* TEMP CODE */
4673	if (stcb->freed_from_where == 0) {
4674		/* Only record the first place free happened from */
4675		stcb->freed_from_where = from_location;
4676	}
4677	/* TEMP CODE */
4678
4679	asoc = &stcb->asoc;
4680	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4681	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4682		/* nothing around */
4683		so = NULL;
4684	else
4685		so = inp->sctp_socket;
4686
4687	/*
4688	 * We used timer based freeing if a reader or writer is in the way.
4689	 * So we first check if we are actually being called from a timer,
4690	 * if so we abort early if a reader or writer is still in the way.
4691	 */
4692	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4693	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4694		/*
4695		 * is it the timer driving us? if so are the reader/writers
4696		 * gone?
4697		 */
4698		if (stcb->asoc.refcnt) {
4699			/* nope, reader or writer in the way */
4700			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4701			/* no asoc destroyed */
4702			SCTP_TCB_UNLOCK(stcb);
4703#ifdef SCTP_LOG_CLOSING
4704			sctp_log_closing(inp, stcb, 8);
4705#endif
4706			return (0);
4707		}
4708	}
4709	/* now clean up any other timers */
4710	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4711	asoc->dack_timer.self = NULL;
4712	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4713	/*-
4714	 * For stream reset we don't blast this unless
4715	 * it is a str-reset timer, it might be the
4716	 * free-asoc timer which we DON'T want to
4717	 * disturb.
4718	 */
4719	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4720		asoc->strreset_timer.self = NULL;
4721	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4722	asoc->asconf_timer.self = NULL;
4723	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4724	asoc->autoclose_timer.self = NULL;
4725	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4726	asoc->shut_guard_timer.self = NULL;
4727	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4728	asoc->delayed_event_timer.self = NULL;
4729	/* Mobility adaptation */
4730	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4731	asoc->delete_prim_timer.self = NULL;
4732	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4733		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4734		net->rxt_timer.self = NULL;
4735		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4736		net->pmtu_timer.self = NULL;
4737		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4738		net->hb_timer.self = NULL;
4739	}
4740	/* Now the read queue needs to be cleaned up (only once) */
4741	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4742		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4743		SCTP_INP_READ_LOCK(inp);
4744		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4745			if (sq->stcb == stcb) {
4746				sq->do_not_ref_stcb = 1;
4747				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4748				/*
4749				 * If there is no end, there never will be
4750				 * now.
4751				 */
4752				if (sq->end_added == 0) {
4753					/* Held for PD-API clear that. */
4754					sq->pdapi_aborted = 1;
4755					sq->held_length = 0;
4756					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4757						/*
4758						 * Need to add a PD-API
4759						 * aborted indication.
4760						 * Setting the control_pdapi
4761						 * assures that it will be
4762						 * added right after this
4763						 * msg.
4764						 */
4765						uint32_t strseq;
4766
4767						stcb->asoc.control_pdapi = sq;
4768						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4769						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4770						    stcb,
4771						    SCTP_PARTIAL_DELIVERY_ABORTED,
4772						    (void *)&strseq,
4773						    SCTP_SO_LOCKED);
4774						stcb->asoc.control_pdapi = NULL;
4775					}
4776				}
4777				/* Add an end to wake them */
4778				sq->end_added = 1;
4779			}
4780		}
4781		SCTP_INP_READ_UNLOCK(inp);
4782		if (stcb->block_entry) {
4783			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4784			stcb->block_entry->error = ECONNRESET;
4785			stcb->block_entry = NULL;
4786		}
4787	}
4788	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4789		/*
4790		 * Someone holds a reference OR the socket is unaccepted
4791		 * yet.
4792		 */
4793		if ((stcb->asoc.refcnt) ||
4794		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4795		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4796			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4797			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4798		}
4799		SCTP_TCB_UNLOCK(stcb);
4800		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4801		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4802			/* nothing around */
4803			so = NULL;
4804		if (so) {
4805			/* Wake any reader/writers */
4806			sctp_sorwakeup(inp, so);
4807			sctp_sowwakeup(inp, so);
4808		}
4809#ifdef SCTP_LOG_CLOSING
4810		sctp_log_closing(inp, stcb, 9);
4811#endif
4812		/* no asoc destroyed */
4813		return (0);
4814	}
4815#ifdef SCTP_LOG_CLOSING
4816	sctp_log_closing(inp, stcb, 10);
4817#endif
4818	/*
4819	 * When I reach here, no others want to kill the assoc yet.. and I
4820	 * own the lock. Now its possible an abort comes in when I do the
4821	 * lock exchange below to grab all the locks to do the final take
4822	 * out. to prevent this we increment the count, which will start a
4823	 * timer and blow out above thus assuring us that we hold exclusive
4824	 * killing of the asoc. Note that after getting back the TCB lock we
4825	 * will go ahead and increment the counter back up and stop any
4826	 * timer a passing stranger may have started :-S
4827	 */
4828	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4829		atomic_add_int(&stcb->asoc.refcnt, 1);
4830
4831		SCTP_TCB_UNLOCK(stcb);
4832		SCTP_INP_INFO_WLOCK();
4833		SCTP_INP_WLOCK(inp);
4834		SCTP_TCB_LOCK(stcb);
4835	}
4836	/* Double check the GONE flag */
4837	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4838	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4839		/* nothing around */
4840		so = NULL;
4841
4842	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4843	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4844		/*
4845		 * For TCP type we need special handling when we are
4846		 * connected. We also include the peel'ed off ones to.
4847		 */
4848		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4849			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4850			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4851			if (so) {
4852				SOCK_LOCK(so);
4853				if (so->so_rcv.sb_cc == 0) {
4854					so->so_state &= ~(SS_ISCONNECTING |
4855					    SS_ISDISCONNECTING |
4856					    SS_ISCONFIRMING |
4857					    SS_ISCONNECTED);
4858				}
4859				socantrcvmore_locked(so);
4860				sctp_sowwakeup(inp, so);
4861				sctp_sorwakeup(inp, so);
4862				SCTP_SOWAKEUP(so);
4863			}
4864		}
4865	}
4866	/*
4867	 * Make it invalid too, that way if its about to run it will abort
4868	 * and return.
4869	 */
4870	/* re-increment the lock */
4871	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4872		atomic_add_int(&stcb->asoc.refcnt, -1);
4873	}
4874	if (stcb->asoc.refcnt) {
4875		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4876		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4877		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4878			SCTP_INP_INFO_WUNLOCK();
4879			SCTP_INP_WUNLOCK(inp);
4880		}
4881		SCTP_TCB_UNLOCK(stcb);
4882		return (0);
4883	}
4884	asoc->state = 0;
4885	if (inp->sctp_tcbhash) {
4886		LIST_REMOVE(stcb, sctp_tcbhash);
4887	}
4888	if (stcb->asoc.in_asocid_hash) {
4889		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4890	}
4891	/* Now lets remove it from the list of ALL associations in the EP */
4892	LIST_REMOVE(stcb, sctp_tcblist);
4893	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4894		SCTP_INP_INCR_REF(inp);
4895		SCTP_INP_WUNLOCK(inp);
4896	}
4897	/* pull from vtag hash */
4898	LIST_REMOVE(stcb, sctp_asocs);
4899	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4900	    inp->sctp_lport, stcb->rport);
4901
4902	/*
4903	 * Now restop the timers to be sure this is paranoia at is finest!
4904	 */
4905	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4906	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4907	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4908	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4909	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4910	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4911	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4912	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4913		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4914		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4915		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4916	}
4917
4918	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4919	/*
4920	 * The chunk lists and such SHOULD be empty but we check them just
4921	 * in case.
4922	 */
4923	/* anything on the wheel needs to be removed */
4924	for (i = 0; i < asoc->streamoutcnt; i++) {
4925		struct sctp_stream_out *outs;
4926
4927		outs = &asoc->strmout[i];
4928		/* now clean up any chunks here */
4929		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4930			TAILQ_REMOVE(&outs->outqueue, sp, next);
4931			sctp_free_spbufspace(stcb, asoc, sp);
4932			if (sp->data) {
4933				if (so) {
4934					/* Still an open socket - report */
4935					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4936					    0, (void *)sp, SCTP_SO_LOCKED);
4937				}
4938				if (sp->data) {
4939					sctp_m_freem(sp->data);
4940					sp->data = NULL;
4941					sp->tail_mbuf = NULL;
4942					sp->length = 0;
4943				}
4944			}
4945			if (sp->net) {
4946				sctp_free_remote_addr(sp->net);
4947				sp->net = NULL;
4948			}
4949			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4950		}
4951	}
4952	/* sa_ignore FREED_MEMORY */
4953	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4954		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4955		SCTP_FREE(strrst, SCTP_M_STRESET);
4956	}
4957	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4958		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4959		if (sq->data) {
4960			sctp_m_freem(sq->data);
4961			sq->data = NULL;
4962		}
4963		sctp_free_remote_addr(sq->whoFrom);
4964		sq->whoFrom = NULL;
4965		sq->stcb = NULL;
4966		/* Free the ctl entry */
4967		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4968		SCTP_DECR_READQ_COUNT();
4969		/* sa_ignore FREED_MEMORY */
4970	}
4971	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4972		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4973		if (chk->data) {
4974			sctp_m_freem(chk->data);
4975			chk->data = NULL;
4976		}
4977		if (chk->holds_key_ref)
4978			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4979		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4980		SCTP_DECR_CHK_COUNT();
4981		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4982		asoc->free_chunk_cnt--;
4983		/* sa_ignore FREED_MEMORY */
4984	}
4985	/* pending send queue SHOULD be empty */
4986	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4987		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4988			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4989#ifdef INVARIANTS
4990		} else {
4991			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4992#endif
4993		}
4994		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4995		if (chk->data) {
4996			if (so) {
4997				/* Still a socket? */
4998				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
4999				    0, chk, SCTP_SO_LOCKED);
5000			}
5001			if (chk->data) {
5002				sctp_m_freem(chk->data);
5003				chk->data = NULL;
5004			}
5005		}
5006		if (chk->holds_key_ref)
5007			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5008		if (chk->whoTo) {
5009			sctp_free_remote_addr(chk->whoTo);
5010			chk->whoTo = NULL;
5011		}
5012		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5013		SCTP_DECR_CHK_COUNT();
5014		/* sa_ignore FREED_MEMORY */
5015	}
5016	/* sent queue SHOULD be empty */
5017	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5018		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5019			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
5020				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
5021#ifdef INVARIANTS
5022			} else {
5023				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
5024#endif
5025			}
5026		}
5027		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5028		if (chk->data) {
5029			if (so) {
5030				/* Still a socket? */
5031				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5032				    0, chk, SCTP_SO_LOCKED);
5033			}
5034			if (chk->data) {
5035				sctp_m_freem(chk->data);
5036				chk->data = NULL;
5037			}
5038		}
5039		if (chk->holds_key_ref)
5040			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5041		sctp_free_remote_addr(chk->whoTo);
5042		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5043		SCTP_DECR_CHK_COUNT();
5044		/* sa_ignore FREED_MEMORY */
5045	}
5046#ifdef INVARIANTS
5047	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5048		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5049			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5050		}
5051	}
5052#endif
5053	/* control queue MAY not be empty */
5054	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5055		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5056		if (chk->data) {
5057			sctp_m_freem(chk->data);
5058			chk->data = NULL;
5059		}
5060		if (chk->holds_key_ref)
5061			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5062		sctp_free_remote_addr(chk->whoTo);
5063		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5064		SCTP_DECR_CHK_COUNT();
5065		/* sa_ignore FREED_MEMORY */
5066	}
5067	/* ASCONF queue MAY not be empty */
5068	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5069		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5070		if (chk->data) {
5071			sctp_m_freem(chk->data);
5072			chk->data = NULL;
5073		}
5074		if (chk->holds_key_ref)
5075			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5076		sctp_free_remote_addr(chk->whoTo);
5077		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5078		SCTP_DECR_CHK_COUNT();
5079		/* sa_ignore FREED_MEMORY */
5080	}
5081	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5082		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5083		if (chk->data) {
5084			sctp_m_freem(chk->data);
5085			chk->data = NULL;
5086		}
5087		if (chk->holds_key_ref)
5088			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5089		sctp_free_remote_addr(chk->whoTo);
5090		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5091		SCTP_DECR_CHK_COUNT();
5092		/* sa_ignore FREED_MEMORY */
5093	}
5094
5095	if (asoc->mapping_array) {
5096		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5097		asoc->mapping_array = NULL;
5098	}
5099	if (asoc->nr_mapping_array) {
5100		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5101		asoc->nr_mapping_array = NULL;
5102	}
5103	/* the stream outs */
5104	if (asoc->strmout) {
5105		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5106		asoc->strmout = NULL;
5107	}
5108	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5109	if (asoc->strmin) {
5110		struct sctp_queued_to_read *ctl, *nctl;
5111
5112		for (i = 0; i < asoc->streamincnt; i++) {
5113			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5114				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5115				sctp_free_remote_addr(ctl->whoFrom);
5116				if (ctl->data) {
5117					sctp_m_freem(ctl->data);
5118					ctl->data = NULL;
5119				}
5120				/*
5121				 * We don't free the address here since all
5122				 * the net's were freed above.
5123				 */
5124				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5125				SCTP_DECR_READQ_COUNT();
5126			}
5127		}
5128		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5129		asoc->strmin = NULL;
5130	}
5131	asoc->streamincnt = 0;
5132	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5133#ifdef INVARIANTS
5134		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5135			panic("no net's left alloc'ed, or list points to itself");
5136		}
5137#endif
5138		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5139		sctp_free_remote_addr(net);
5140	}
5141	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5142		/* sa_ignore FREED_MEMORY */
5143		sctp_remove_laddr(laddr);
5144	}
5145
5146	/* pending asconf (address) parameters */
5147	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5148		/* sa_ignore FREED_MEMORY */
5149		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5150		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5151	}
5152	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5153		/* sa_ignore FREED_MEMORY */
5154		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5155		if (aack->data != NULL) {
5156			sctp_m_freem(aack->data);
5157		}
5158		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5159	}
5160	/* clean up auth stuff */
5161	if (asoc->local_hmacs)
5162		sctp_free_hmaclist(asoc->local_hmacs);
5163	if (asoc->peer_hmacs)
5164		sctp_free_hmaclist(asoc->peer_hmacs);
5165
5166	if (asoc->local_auth_chunks)
5167		sctp_free_chunklist(asoc->local_auth_chunks);
5168	if (asoc->peer_auth_chunks)
5169		sctp_free_chunklist(asoc->peer_auth_chunks);
5170
5171	sctp_free_authinfo(&asoc->authinfo);
5172
5173	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5174		LIST_REMOVE(shared_key, next);
5175		sctp_free_sharedkey(shared_key);
5176		/* sa_ignore FREED_MEMORY */
5177	}
5178
5179	/* Insert new items here :> */
5180
5181	/* Get rid of LOCK */
5182	SCTP_TCB_UNLOCK(stcb);
5183	SCTP_TCB_LOCK_DESTROY(stcb);
5184	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5185	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5186		SCTP_INP_INFO_WUNLOCK();
5187		SCTP_INP_RLOCK(inp);
5188	}
5189#ifdef SCTP_TRACK_FREED_ASOCS
5190	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5191		/* now clean up the tasoc itself */
5192		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5193		SCTP_DECR_ASOC_COUNT();
5194	} else {
5195		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5196	}
5197#else
5198	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5199	SCTP_DECR_ASOC_COUNT();
5200#endif
5201	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5202		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5203			/*
5204			 * If its NOT the inp_free calling us AND sctp_close
5205			 * as been called, we call back...
5206			 */
5207			SCTP_INP_RUNLOCK(inp);
5208			/*
5209			 * This will start the kill timer (if we are the
5210			 * last one) since we hold an increment yet. But
5211			 * this is the only safe way to do this since
5212			 * otherwise if the socket closes at the same time
5213			 * we are here we might collide in the cleanup.
5214			 */
5215			sctp_inpcb_free(inp,
5216			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5217			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5218			SCTP_INP_DECR_REF(inp);
5219			goto out_of;
5220		} else {
5221			/* The socket is still open. */
5222			SCTP_INP_DECR_REF(inp);
5223		}
5224	}
5225	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5226		SCTP_INP_RUNLOCK(inp);
5227	}
5228out_of:
5229	/* destroyed the asoc */
5230#ifdef SCTP_LOG_CLOSING
5231	sctp_log_closing(inp, NULL, 11);
5232#endif
5233	return (1);
5234}
5235
5236
5237
5238/*
5239 * determine if a destination is "reachable" based upon the addresses bound
5240 * to the current endpoint (e.g. only v4 or v6 currently bound)
5241 */
5242/*
5243 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5244 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5245 * types bound as its endpoint
5246 */
5247int
5248sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5249{
5250	struct sctp_inpcb *inp;
5251	int answer;
5252
5253	/*
5254	 * No locks here, the TCB, in all cases is already locked and an
5255	 * assoc is up. There is either a INP lock by the caller applied (in
5256	 * asconf case when deleting an address) or NOT in the HB case,
5257	 * however if HB then the INP increment is up and the INP will not
5258	 * be removed (on top of the fact that we have a TCB lock). So we
5259	 * only want to read the sctp_flags, which is either bound-all or
5260	 * not.. no protection needed since once an assoc is up you can't be
5261	 * changing your binding.
5262	 */
5263	inp = stcb->sctp_ep;
5264	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5265		/* if bound all, destination is not restricted */
5266		/*
5267		 * RRS: Question during lock work: Is this correct? If you
5268		 * are bound-all you still might need to obey the V4--V6
5269		 * flags??? IMO this bound-all stuff needs to be removed!
5270		 */
5271		return (1);
5272	}
5273	/* NOTE: all "scope" checks are done when local addresses are added */
5274	switch (destaddr->sa_family) {
5275#ifdef INET6
5276	case AF_INET6:
5277		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5278		break;
5279#endif
5280#ifdef INET
5281	case AF_INET:
5282		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5283		break;
5284#endif
5285	default:
5286		/* invalid family, so it's unreachable */
5287		answer = 0;
5288		break;
5289	}
5290	return (answer);
5291}
5292
5293/*
5294 * update the inp_vflags on an endpoint
5295 */
5296static void
5297sctp_update_ep_vflag(struct sctp_inpcb *inp)
5298{
5299	struct sctp_laddr *laddr;
5300
5301	/* first clear the flag */
5302	inp->ip_inp.inp.inp_vflag = 0;
5303	/* set the flag based on addresses on the ep list */
5304	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5305		if (laddr->ifa == NULL) {
5306			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5307			    __FUNCTION__);
5308			continue;
5309		}
5310		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5311			continue;
5312		}
5313		switch (laddr->ifa->address.sa.sa_family) {
5314#ifdef INET6
5315		case AF_INET6:
5316			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5317			break;
5318#endif
5319#ifdef INET
5320		case AF_INET:
5321			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5322			break;
5323#endif
5324		default:
5325			break;
5326		}
5327	}
5328}
5329
5330/*
5331 * Add the address to the endpoint local address list There is nothing to be
5332 * done if we are bound to all addresses
5333 */
5334void
5335sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5336{
5337	struct sctp_laddr *laddr;
5338	int fnd, error = 0;
5339
5340	fnd = 0;
5341
5342	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5343		/* You are already bound to all. You have it already */
5344		return;
5345	}
5346#ifdef INET6
5347	if (ifa->address.sa.sa_family == AF_INET6) {
5348		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5349			/* Can't bind a non-useable addr. */
5350			return;
5351		}
5352	}
5353#endif
5354	/* first, is it already present? */
5355	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5356		if (laddr->ifa == ifa) {
5357			fnd = 1;
5358			break;
5359		}
5360	}
5361
5362	if (fnd == 0) {
5363		/* Not in the ep list */
5364		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5365		if (error != 0)
5366			return;
5367		inp->laddr_count++;
5368		/* update inp_vflag flags */
5369		switch (ifa->address.sa.sa_family) {
5370#ifdef INET6
5371		case AF_INET6:
5372			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5373			break;
5374#endif
5375#ifdef INET
5376		case AF_INET:
5377			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5378			break;
5379#endif
5380		default:
5381			break;
5382		}
5383	}
5384	return;
5385}
5386
5387
5388/*
5389 * select a new (hopefully reachable) destination net (should only be used
5390 * when we deleted an ep addr that is the only usable source address to reach
5391 * the destination net)
5392 */
5393static void
5394sctp_select_primary_destination(struct sctp_tcb *stcb)
5395{
5396	struct sctp_nets *net;
5397
5398	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5399		/* for now, we'll just pick the first reachable one we find */
5400		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5401			continue;
5402		if (sctp_destination_is_reachable(stcb,
5403		    (struct sockaddr *)&net->ro._l_addr)) {
5404			/* found a reachable destination */
5405			stcb->asoc.primary_destination = net;
5406		}
5407	}
5408	/* I can't there from here! ...we're gonna die shortly... */
5409}
5410
5411
5412/*
5413 * Delete the address from the endpoint local address list There is nothing
5414 * to be done if we are bound to all addresses
5415 */
5416void
5417sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5418{
5419	struct sctp_laddr *laddr;
5420	int fnd;
5421
5422	fnd = 0;
5423	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5424		/* You are already bound to all. You have it already */
5425		return;
5426	}
5427	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5428		if (laddr->ifa == ifa) {
5429			fnd = 1;
5430			break;
5431		}
5432	}
5433	if (fnd && (inp->laddr_count < 2)) {
5434		/* can't delete unless there are at LEAST 2 addresses */
5435		return;
5436	}
5437	if (fnd) {
5438		/*
5439		 * clean up any use of this address go through our
5440		 * associations and clear any last_used_address that match
5441		 * this one for each assoc, see if a new primary_destination
5442		 * is needed
5443		 */
5444		struct sctp_tcb *stcb;
5445
5446		/* clean up "next_addr_touse" */
5447		if (inp->next_addr_touse == laddr)
5448			/* delete this address */
5449			inp->next_addr_touse = NULL;
5450
5451		/* clean up "last_used_address" */
5452		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5453			struct sctp_nets *net;
5454
5455			SCTP_TCB_LOCK(stcb);
5456			if (stcb->asoc.last_used_address == laddr)
5457				/* delete this address */
5458				stcb->asoc.last_used_address = NULL;
5459			/*
5460			 * Now spin through all the nets and purge any ref
5461			 * to laddr
5462			 */
5463			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5464				if (net->ro._s_addr &&
5465				    (net->ro._s_addr->ifa == laddr->ifa)) {
5466					/* Yep, purge src address selected */
5467					sctp_rtentry_t *rt;
5468
5469					/* delete this address if cached */
5470					rt = net->ro.ro_rt;
5471					if (rt != NULL) {
5472						RTFREE(rt);
5473						net->ro.ro_rt = NULL;
5474					}
5475					sctp_free_ifa(net->ro._s_addr);
5476					net->ro._s_addr = NULL;
5477					net->src_addr_selected = 0;
5478				}
5479			}
5480			SCTP_TCB_UNLOCK(stcb);
5481		}		/* for each tcb */
5482		/* remove it from the ep list */
5483		sctp_remove_laddr(laddr);
5484		inp->laddr_count--;
5485		/* update inp_vflag flags */
5486		sctp_update_ep_vflag(inp);
5487	}
5488	return;
5489}
5490
5491/*
5492 * Add the address to the TCB local address restricted list.
5493 * This is a "pending" address list (eg. addresses waiting for an
5494 * ASCONF-ACK response) and cannot be used as a valid source address.
5495 */
5496void
5497sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5498{
5499	struct sctp_laddr *laddr;
5500	struct sctpladdr *list;
5501
5502	/*
5503	 * Assumes TCB is locked.. and possibly the INP. May need to
5504	 * confirm/fix that if we need it and is not the case.
5505	 */
5506	list = &stcb->asoc.sctp_restricted_addrs;
5507
5508#ifdef INET6
5509	if (ifa->address.sa.sa_family == AF_INET6) {
5510		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5511			/* Can't bind a non-existent addr. */
5512			return;
5513		}
5514	}
5515#endif
5516	/* does the address already exist? */
5517	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5518		if (laddr->ifa == ifa) {
5519			return;
5520		}
5521	}
5522
5523	/* add to the list */
5524	(void)sctp_insert_laddr(list, ifa, 0);
5525	return;
5526}
5527
5528/*
5529 * insert an laddr entry with the given ifa for the desired list
5530 */
5531int
5532sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5533{
5534	struct sctp_laddr *laddr;
5535
5536	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5537	if (laddr == NULL) {
5538		/* out of memory? */
5539		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5540		return (EINVAL);
5541	}
5542	SCTP_INCR_LADDR_COUNT();
5543	bzero(laddr, sizeof(*laddr));
5544	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5545	laddr->ifa = ifa;
5546	laddr->action = act;
5547	atomic_add_int(&ifa->refcount, 1);
5548	/* insert it */
5549	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5550
5551	return (0);
5552}
5553
5554/*
5555 * Remove an laddr entry from the local address list (on an assoc)
5556 */
5557void
5558sctp_remove_laddr(struct sctp_laddr *laddr)
5559{
5560
5561	/* remove from the list */
5562	LIST_REMOVE(laddr, sctp_nxt_addr);
5563	sctp_free_ifa(laddr->ifa);
5564	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5565	SCTP_DECR_LADDR_COUNT();
5566}
5567
5568/*
5569 * Remove a local address from the TCB local address restricted list
5570 */
5571void
5572sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5573{
5574	struct sctp_inpcb *inp;
5575	struct sctp_laddr *laddr;
5576
5577	/*
5578	 * This is called by asconf work. It is assumed that a) The TCB is
5579	 * locked and b) The INP is locked. This is true in as much as I can
5580	 * trace through the entry asconf code where I did these locks.
5581	 * Again, the ASCONF code is a bit different in that it does lock
5582	 * the INP during its work often times. This must be since we don't
5583	 * want other proc's looking up things while what they are looking
5584	 * up is changing :-D
5585	 */
5586
5587	inp = stcb->sctp_ep;
5588	/* if subset bound and don't allow ASCONF's, can't delete last */
5589	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5590	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5591		if (stcb->sctp_ep->laddr_count < 2) {
5592			/* can't delete last address */
5593			return;
5594		}
5595	}
5596	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5597		/* remove the address if it exists */
5598		if (laddr->ifa == NULL)
5599			continue;
5600		if (laddr->ifa == ifa) {
5601			sctp_remove_laddr(laddr);
5602			return;
5603		}
5604	}
5605
5606	/* address not found! */
5607	return;
5608}
5609
5610/*
5611 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5612 */
5613/* sysctl */
5614static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5615static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5616
5617
5618
5619#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5620struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5621int *sctp_cpuarry = NULL;
5622void
5623sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5624{
5625	/* Queue a packet to a processor for the specified core */
5626	struct sctp_mcore_queue *qent;
5627	struct sctp_mcore_ctrl *wkq;
5628	int need_wake = 0;
5629
5630	if (sctp_mcore_workers == NULL) {
5631		/* Something went way bad during setup */
5632		sctp_input_with_port(m, off, 0);
5633		return;
5634	}
5635	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5636	    (sizeof(struct sctp_mcore_queue)),
5637	    SCTP_M_MCORE);
5638	if (qent == NULL) {
5639		/* This is trouble  */
5640		sctp_input_with_port(m, off, 0);
5641		return;
5642	}
5643	qent->vn = curvnet;
5644	qent->m = m;
5645	qent->off = off;
5646	qent->v6 = 0;
5647	wkq = &sctp_mcore_workers[cpu_to_use];
5648	SCTP_MCORE_QLOCK(wkq);
5649
5650	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5651	if (wkq->running == 0) {
5652		need_wake = 1;
5653	}
5654	SCTP_MCORE_QUNLOCK(wkq);
5655	if (need_wake) {
5656		wakeup(&wkq->running);
5657	}
5658}
5659
5660static void
5661sctp_mcore_thread(void *arg)
5662{
5663
5664	struct sctp_mcore_ctrl *wkq;
5665	struct sctp_mcore_queue *qent;
5666
5667	wkq = (struct sctp_mcore_ctrl *)arg;
5668	struct mbuf *m;
5669	int off, v6;
5670
5671	/* Wait for first tickle */
5672	SCTP_MCORE_LOCK(wkq);
5673	wkq->running = 0;
5674	msleep(&wkq->running,
5675	    &wkq->core_mtx,
5676	    0, "wait for pkt", 0);
5677	SCTP_MCORE_UNLOCK(wkq);
5678
5679	/* Bind to our cpu */
5680	thread_lock(curthread);
5681	sched_bind(curthread, wkq->cpuid);
5682	thread_unlock(curthread);
5683
5684	/* Now lets start working */
5685	SCTP_MCORE_LOCK(wkq);
5686	/* Now grab lock and go */
5687	for (;;) {
5688		SCTP_MCORE_QLOCK(wkq);
5689skip_sleep:
5690		wkq->running = 1;
5691		qent = TAILQ_FIRST(&wkq->que);
5692		if (qent) {
5693			TAILQ_REMOVE(&wkq->que, qent, next);
5694			SCTP_MCORE_QUNLOCK(wkq);
5695			CURVNET_SET(qent->vn);
5696			m = qent->m;
5697			off = qent->off;
5698			v6 = qent->v6;
5699			SCTP_FREE(qent, SCTP_M_MCORE);
5700			if (v6 == 0) {
5701				sctp_input_with_port(m, off, 0);
5702			} else {
5703				SCTP_PRINTF("V6 not yet supported\n");
5704				sctp_m_freem(m);
5705			}
5706			CURVNET_RESTORE();
5707			SCTP_MCORE_QLOCK(wkq);
5708		}
5709		wkq->running = 0;
5710		if (!TAILQ_EMPTY(&wkq->que)) {
5711			goto skip_sleep;
5712		}
5713		SCTP_MCORE_QUNLOCK(wkq);
5714		msleep(&wkq->running,
5715		    &wkq->core_mtx,
5716		    0, "wait for pkt", 0);
5717	}
5718}
5719
5720static void
5721sctp_startup_mcore_threads(void)
5722{
5723	int i, cpu;
5724
5725	if (mp_ncpus == 1)
5726		return;
5727
5728	if (sctp_mcore_workers != NULL) {
5729		/*
5730		 * Already been here in some previous vnet?
5731		 */
5732		return;
5733	}
5734	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5735	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5736	    SCTP_M_MCORE);
5737	if (sctp_mcore_workers == NULL) {
5738		/* TSNH I hope */
5739		return;
5740	}
5741	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5742	    sizeof(struct sctp_mcore_ctrl)));
5743	/* Init the structures */
5744	for (i = 0; i <= mp_maxid; i++) {
5745		TAILQ_INIT(&sctp_mcore_workers[i].que);
5746		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5747		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5748		sctp_mcore_workers[i].cpuid = i;
5749	}
5750	if (sctp_cpuarry == NULL) {
5751		SCTP_MALLOC(sctp_cpuarry, int *,
5752		    (mp_ncpus * sizeof(int)),
5753		    SCTP_M_MCORE);
5754		i = 0;
5755		CPU_FOREACH(cpu) {
5756			sctp_cpuarry[i] = cpu;
5757			i++;
5758		}
5759	}
5760	/* Now start them all */
5761	CPU_FOREACH(cpu) {
5762		(void)kproc_create(sctp_mcore_thread,
5763		    (void *)&sctp_mcore_workers[cpu],
5764		    &sctp_mcore_workers[cpu].thread_proc,
5765		    RFPROC,
5766		    SCTP_KTHREAD_PAGES,
5767		    SCTP_MCORE_NAME);
5768
5769	}
5770}
5771
5772#endif
5773
5774void
5775sctp_pcb_init()
5776{
5777	/*
5778	 * SCTP initialization for the PCB structures should be called by
5779	 * the sctp_init() funciton.
5780	 */
5781	int i;
5782	struct timeval tv;
5783
5784	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5785		/* error I was called twice */
5786		return;
5787	}
5788	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5789
5790#if defined(SCTP_LOCAL_TRACE_BUF)
5791	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5792#endif
5793#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5794	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5795	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5796	    SCTP_M_MCORE);
5797#endif
5798	(void)SCTP_GETTIME_TIMEVAL(&tv);
5799#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5800	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5801	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5802	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5803#else
5804	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5805	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5806	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5807#endif
5808	/* init the empty list of (All) Endpoints */
5809	LIST_INIT(&SCTP_BASE_INFO(listhead));
5810
5811
5812	/* init the hash table of endpoints */
5813	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5814	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5815	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5816	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5817	    &SCTP_BASE_INFO(hashasocmark));
5818	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5819	    &SCTP_BASE_INFO(hashmark));
5820	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5821	    &SCTP_BASE_INFO(hashtcpmark));
5822	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5823
5824
5825	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5826	    &SCTP_BASE_INFO(hashvrfmark));
5827
5828	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5829	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5830	/* init the zones */
5831	/*
5832	 * FIX ME: Should check for NULL returns, but if it does fail we are
5833	 * doomed to panic anyways... add later maybe.
5834	 */
5835	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5836	    sizeof(struct sctp_inpcb), maxsockets);
5837
5838	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5839	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5840
5841	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5842	    sizeof(struct sctp_laddr),
5843	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5844
5845	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5846	    sizeof(struct sctp_nets),
5847	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5848
5849	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5850	    sizeof(struct sctp_tmit_chunk),
5851	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5852
5853	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5854	    sizeof(struct sctp_queued_to_read),
5855	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5856
5857	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5858	    sizeof(struct sctp_stream_queue_pending),
5859	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5860
5861	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5862	    sizeof(struct sctp_asconf),
5863	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5864
5865	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5866	    sizeof(struct sctp_asconf_ack),
5867	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5868
5869
5870	/* Master Lock INIT for info structure */
5871	SCTP_INP_INFO_LOCK_INIT();
5872	SCTP_STATLOG_INIT_LOCK();
5873
5874	SCTP_IPI_COUNT_INIT();
5875	SCTP_IPI_ADDR_INIT();
5876#ifdef SCTP_PACKET_LOGGING
5877	SCTP_IP_PKTLOG_INIT();
5878#endif
5879	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5880
5881	SCTP_WQ_ADDR_INIT();
5882	/* not sure if we need all the counts */
5883	SCTP_BASE_INFO(ipi_count_ep) = 0;
5884	/* assoc/tcb zone info */
5885	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5886	/* local addrlist zone info */
5887	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5888	/* remote addrlist zone info */
5889	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5890	/* chunk info */
5891	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5892
5893	/* socket queue zone info */
5894	SCTP_BASE_INFO(ipi_count_readq) = 0;
5895
5896	/* stream out queue cont */
5897	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5898
5899	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5900	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5901
5902	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5903
5904	/* Init the TIMEWAIT list */
5905	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5906		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5907	}
5908	sctp_startup_iterator();
5909
5910#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5911	sctp_startup_mcore_threads();
5912#endif
5913
5914	/*
5915	 * INIT the default VRF which for BSD is the only one, other O/S's
5916	 * may have more. But initially they must start with one and then
5917	 * add the VRF's as addresses are added.
5918	 */
5919	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5920}
5921
5922/*
5923 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5924 */
5925void
5926sctp_pcb_finish(void)
5927{
5928	struct sctp_vrflist *vrf_bucket;
5929	struct sctp_vrf *vrf, *nvrf;
5930	struct sctp_ifn *ifn, *nifn;
5931	struct sctp_ifa *ifa, *nifa;
5932	struct sctpvtaghead *chain;
5933	struct sctp_tagblock *twait_block, *prev_twait_block;
5934	struct sctp_laddr *wi, *nwi;
5935	int i;
5936	struct sctp_iterator *it, *nit;
5937
5938	/*
5939	 * In FreeBSD the iterator thread never exits but we do clean up.
5940	 * The only way FreeBSD reaches here is if we have VRF's but we
5941	 * still add the ifdef to make it compile on old versions.
5942	 */
5943	SCTP_IPI_ITERATOR_WQ_LOCK();
5944	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5945		if (it->vn != curvnet) {
5946			continue;
5947		}
5948		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5949		if (it->function_atend != NULL) {
5950			(*it->function_atend) (it->pointer, it->val);
5951		}
5952		SCTP_FREE(it, SCTP_M_ITER);
5953	}
5954	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5955	SCTP_ITERATOR_LOCK();
5956	if ((sctp_it_ctl.cur_it) &&
5957	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5958		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5959	}
5960	SCTP_ITERATOR_UNLOCK();
5961	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5962	SCTP_WQ_ADDR_LOCK();
5963	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5964		LIST_REMOVE(wi, sctp_nxt_addr);
5965		SCTP_DECR_LADDR_COUNT();
5966		if (wi->action == SCTP_DEL_IP_ADDRESS) {
5967			SCTP_FREE(wi->ifa, SCTP_M_IFA);
5968		}
5969		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5970	}
5971	SCTP_WQ_ADDR_UNLOCK();
5972
5973	/*
5974	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5975	 * destroyed first).
5976	 */
5977	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5978	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5979		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5980			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5981				/* free the ifa */
5982				LIST_REMOVE(ifa, next_bucket);
5983				LIST_REMOVE(ifa, next_ifa);
5984				SCTP_FREE(ifa, SCTP_M_IFA);
5985			}
5986			/* free the ifn */
5987			LIST_REMOVE(ifn, next_bucket);
5988			LIST_REMOVE(ifn, next_ifn);
5989			SCTP_FREE(ifn, SCTP_M_IFN);
5990		}
5991		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5992		/* free the vrf */
5993		LIST_REMOVE(vrf, next_vrf);
5994		SCTP_FREE(vrf, SCTP_M_VRF);
5995	}
5996	/* free the vrf hashes */
5997	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5998	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5999
6000	/*
6001	 * free the TIMEWAIT list elements malloc'd in the function
6002	 * sctp_add_vtag_to_timewait()...
6003	 */
6004	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
6005		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
6006		if (!LIST_EMPTY(chain)) {
6007			prev_twait_block = NULL;
6008			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6009				if (prev_twait_block) {
6010					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6011				}
6012				prev_twait_block = twait_block;
6013			}
6014			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6015		}
6016	}
6017
6018	/* free the locks and mutexes */
6019#ifdef SCTP_PACKET_LOGGING
6020	SCTP_IP_PKTLOG_DESTROY();
6021#endif
6022	SCTP_IPI_ADDR_DESTROY();
6023	SCTP_STATLOG_DESTROY();
6024	SCTP_INP_INFO_LOCK_DESTROY();
6025
6026	SCTP_WQ_ADDR_DESTROY();
6027
6028	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6029	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6030	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6031	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6032	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6033	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6034	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6035	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6036	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6037	/* Get rid of other stuff to */
6038	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
6039		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
6040	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
6041		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6042	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6043		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6044#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6045	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6046#endif
6047}
6048
6049
6050int
6051sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6052    int offset, int limit,
6053    struct sockaddr *src, struct sockaddr *dst,
6054    struct sockaddr *altsa)
6055{
6056	/*
6057	 * grub through the INIT pulling addresses and loading them to the
6058	 * nets structure in the asoc. The from address in the mbuf should
6059	 * also be loaded (if it is not already). This routine can be called
6060	 * with either INIT or INIT-ACK's as long as the m points to the IP
6061	 * packet and the offset points to the beginning of the parameters.
6062	 */
6063	struct sctp_inpcb *inp;
6064	struct sctp_nets *net, *nnet, *net_tmp;
6065	struct sctp_paramhdr *phdr, parm_buf;
6066	struct sctp_tcb *stcb_tmp;
6067	uint16_t ptype, plen;
6068	struct sockaddr *sa;
6069	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6070	struct sctp_auth_random *p_random = NULL;
6071	uint16_t random_len = 0;
6072	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6073	struct sctp_auth_hmac_algo *hmacs = NULL;
6074	uint16_t hmacs_len = 0;
6075	uint8_t saw_asconf = 0;
6076	uint8_t saw_asconf_ack = 0;
6077	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6078	struct sctp_auth_chunk_list *chunks = NULL;
6079	uint16_t num_chunks = 0;
6080	sctp_key_t *new_key;
6081	uint32_t keylen;
6082	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6083	uint8_t peer_supports_ecn;
6084	uint8_t peer_supports_prsctp;
6085	uint8_t peer_supports_auth;
6086	uint8_t peer_supports_asconf;
6087	uint8_t peer_supports_asconf_ack;
6088	uint8_t peer_supports_reconfig;
6089	uint8_t peer_supports_nrsack;
6090	uint8_t peer_supports_pktdrop;
6091
6092#ifdef INET
6093	struct sockaddr_in sin;
6094
6095#endif
6096#ifdef INET6
6097	struct sockaddr_in6 sin6;
6098
6099#endif
6100
6101	/* First get the destination address setup too. */
6102#ifdef INET
6103	memset(&sin, 0, sizeof(sin));
6104	sin.sin_family = AF_INET;
6105	sin.sin_len = sizeof(sin);
6106	sin.sin_port = stcb->rport;
6107#endif
6108#ifdef INET6
6109	memset(&sin6, 0, sizeof(sin6));
6110	sin6.sin6_family = AF_INET6;
6111	sin6.sin6_len = sizeof(struct sockaddr_in6);
6112	sin6.sin6_port = stcb->rport;
6113#endif
6114	if (altsa) {
6115		sa = altsa;
6116	} else {
6117		sa = src;
6118	}
6119	peer_supports_ecn = 0;
6120	peer_supports_prsctp = 0;
6121	peer_supports_auth = 0;
6122	peer_supports_asconf = 0;
6123	peer_supports_reconfig = 0;
6124	peer_supports_nrsack = 0;
6125	peer_supports_pktdrop = 0;
6126	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6127		/* mark all addresses that we have currently on the list */
6128		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6129	}
6130	/* does the source address already exist? if so skip it */
6131	inp = stcb->sctp_ep;
6132	atomic_add_int(&stcb->asoc.refcnt, 1);
6133	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6134	atomic_add_int(&stcb->asoc.refcnt, -1);
6135
6136	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6137		/* we must add the source address */
6138		/* no scope set here since we have a tcb already. */
6139		switch (sa->sa_family) {
6140#ifdef INET
6141		case AF_INET:
6142			if (stcb->asoc.scope.ipv4_addr_legal) {
6143				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6144					return (-1);
6145				}
6146			}
6147			break;
6148#endif
6149#ifdef INET6
6150		case AF_INET6:
6151			if (stcb->asoc.scope.ipv6_addr_legal) {
6152				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6153					return (-2);
6154				}
6155			}
6156			break;
6157#endif
6158		default:
6159			break;
6160		}
6161	} else {
6162		if (net_tmp != NULL && stcb_tmp == stcb) {
6163			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6164		} else if (stcb_tmp != stcb) {
6165			/* It belongs to another association? */
6166			if (stcb_tmp)
6167				SCTP_TCB_UNLOCK(stcb_tmp);
6168			return (-3);
6169		}
6170	}
6171	if (stcb->asoc.state == 0) {
6172		/* the assoc was freed? */
6173		return (-4);
6174	}
6175	/* now we must go through each of the params. */
6176	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6177	while (phdr) {
6178		ptype = ntohs(phdr->param_type);
6179		plen = ntohs(phdr->param_length);
6180		/*
6181		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6182		 * (uint32_t)ptype, (int)plen);
6183		 */
6184		if (offset + plen > limit) {
6185			break;
6186		}
6187		if (plen == 0) {
6188			break;
6189		}
6190#ifdef INET
6191		if (ptype == SCTP_IPV4_ADDRESS) {
6192			if (stcb->asoc.scope.ipv4_addr_legal) {
6193				struct sctp_ipv4addr_param *p4, p4_buf;
6194
6195				/* ok get the v4 address and check/add */
6196				phdr = sctp_get_next_param(m, offset,
6197				    (struct sctp_paramhdr *)&p4_buf,
6198				    sizeof(p4_buf));
6199				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6200				    phdr == NULL) {
6201					return (-5);
6202				}
6203				p4 = (struct sctp_ipv4addr_param *)phdr;
6204				sin.sin_addr.s_addr = p4->addr;
6205				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6206					/* Skip multi-cast addresses */
6207					goto next_param;
6208				}
6209				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6210				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6211					goto next_param;
6212				}
6213				sa = (struct sockaddr *)&sin;
6214				inp = stcb->sctp_ep;
6215				atomic_add_int(&stcb->asoc.refcnt, 1);
6216				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6217				    dst, stcb);
6218				atomic_add_int(&stcb->asoc.refcnt, -1);
6219
6220				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6221				    inp == NULL) {
6222					/* we must add the source address */
6223					/*
6224					 * no scope set since we have a tcb
6225					 * already
6226					 */
6227
6228					/*
6229					 * we must validate the state again
6230					 * here
6231					 */
6232			add_it_now:
6233					if (stcb->asoc.state == 0) {
6234						/* the assoc was freed? */
6235						return (-7);
6236					}
6237					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6238						return (-8);
6239					}
6240				} else if (stcb_tmp == stcb) {
6241					if (stcb->asoc.state == 0) {
6242						/* the assoc was freed? */
6243						return (-10);
6244					}
6245					if (net != NULL) {
6246						/* clear flag */
6247						net->dest_state &=
6248						    ~SCTP_ADDR_NOT_IN_ASSOC;
6249					}
6250				} else {
6251					/*
6252					 * strange, address is in another
6253					 * assoc? straighten out locks.
6254					 */
6255					if (stcb_tmp) {
6256						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6257							/*
6258							 * in setup state we
6259							 * abort this guy
6260							 */
6261							sctp_abort_an_association(stcb_tmp->sctp_ep,
6262							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6263							goto add_it_now;
6264						}
6265						SCTP_TCB_UNLOCK(stcb_tmp);
6266					}
6267					if (stcb->asoc.state == 0) {
6268						/* the assoc was freed? */
6269						return (-12);
6270					}
6271					return (-13);
6272				}
6273			}
6274		} else
6275#endif
6276#ifdef INET6
6277		if (ptype == SCTP_IPV6_ADDRESS) {
6278			if (stcb->asoc.scope.ipv6_addr_legal) {
6279				/* ok get the v6 address and check/add */
6280				struct sctp_ipv6addr_param *p6, p6_buf;
6281
6282				phdr = sctp_get_next_param(m, offset,
6283				    (struct sctp_paramhdr *)&p6_buf,
6284				    sizeof(p6_buf));
6285				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6286				    phdr == NULL) {
6287					return (-14);
6288				}
6289				p6 = (struct sctp_ipv6addr_param *)phdr;
6290				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6291				    sizeof(p6->addr));
6292				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6293					/* Skip multi-cast addresses */
6294					goto next_param;
6295				}
6296				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6297					/*
6298					 * Link local make no sense without
6299					 * scope
6300					 */
6301					goto next_param;
6302				}
6303				sa = (struct sockaddr *)&sin6;
6304				inp = stcb->sctp_ep;
6305				atomic_add_int(&stcb->asoc.refcnt, 1);
6306				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6307				    dst, stcb);
6308				atomic_add_int(&stcb->asoc.refcnt, -1);
6309				if (stcb_tmp == NULL &&
6310				    (inp == stcb->sctp_ep || inp == NULL)) {
6311					/*
6312					 * we must validate the state again
6313					 * here
6314					 */
6315			add_it_now6:
6316					if (stcb->asoc.state == 0) {
6317						/* the assoc was freed? */
6318						return (-16);
6319					}
6320					/*
6321					 * we must add the address, no scope
6322					 * set
6323					 */
6324					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6325						return (-17);
6326					}
6327				} else if (stcb_tmp == stcb) {
6328					/*
6329					 * we must validate the state again
6330					 * here
6331					 */
6332					if (stcb->asoc.state == 0) {
6333						/* the assoc was freed? */
6334						return (-19);
6335					}
6336					if (net != NULL) {
6337						/* clear flag */
6338						net->dest_state &=
6339						    ~SCTP_ADDR_NOT_IN_ASSOC;
6340					}
6341				} else {
6342					/*
6343					 * strange, address is in another
6344					 * assoc? straighten out locks.
6345					 */
6346					if (stcb_tmp)
6347						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6348							/*
6349							 * in setup state we
6350							 * abort this guy
6351							 */
6352							sctp_abort_an_association(stcb_tmp->sctp_ep,
6353							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6354							goto add_it_now6;
6355						}
6356					SCTP_TCB_UNLOCK(stcb_tmp);
6357
6358					if (stcb->asoc.state == 0) {
6359						/* the assoc was freed? */
6360						return (-21);
6361					}
6362					return (-22);
6363				}
6364			}
6365		} else
6366#endif
6367		if (ptype == SCTP_ECN_CAPABLE) {
6368			peer_supports_ecn = 1;
6369		} else if (ptype == SCTP_ULP_ADAPTATION) {
6370			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6371				struct sctp_adaptation_layer_indication ai,
6372				                                *aip;
6373
6374				phdr = sctp_get_next_param(m, offset,
6375				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6376				aip = (struct sctp_adaptation_layer_indication *)phdr;
6377				if (aip) {
6378					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6379					stcb->asoc.adaptation_needed = 1;
6380				}
6381			}
6382		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6383			struct sctp_asconf_addr_param lstore, *fee;
6384			int lptype;
6385			struct sockaddr *lsa = NULL;
6386
6387#ifdef INET
6388			struct sctp_asconf_addrv4_param *fii;
6389
6390#endif
6391
6392			if (stcb->asoc.asconf_supported == 0) {
6393				return (-100);
6394			}
6395			if (plen > sizeof(lstore)) {
6396				return (-23);
6397			}
6398			phdr = sctp_get_next_param(m, offset,
6399			    (struct sctp_paramhdr *)&lstore,
6400			    min(plen, sizeof(lstore)));
6401			if (phdr == NULL) {
6402				return (-24);
6403			}
6404			fee = (struct sctp_asconf_addr_param *)phdr;
6405			lptype = ntohs(fee->addrp.ph.param_type);
6406			switch (lptype) {
6407#ifdef INET
6408			case SCTP_IPV4_ADDRESS:
6409				if (plen !=
6410				    sizeof(struct sctp_asconf_addrv4_param)) {
6411					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6412					    (int)sizeof(struct sctp_asconf_addrv4_param),
6413					    plen);
6414				} else {
6415					fii = (struct sctp_asconf_addrv4_param *)fee;
6416					sin.sin_addr.s_addr = fii->addrp.addr;
6417					lsa = (struct sockaddr *)&sin;
6418				}
6419				break;
6420#endif
6421#ifdef INET6
6422			case SCTP_IPV6_ADDRESS:
6423				if (plen !=
6424				    sizeof(struct sctp_asconf_addr_param)) {
6425					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6426					    (int)sizeof(struct sctp_asconf_addr_param),
6427					    plen);
6428				} else {
6429					memcpy(sin6.sin6_addr.s6_addr,
6430					    fee->addrp.addr,
6431					    sizeof(fee->addrp.addr));
6432					lsa = (struct sockaddr *)&sin6;
6433				}
6434				break;
6435#endif
6436			default:
6437				break;
6438			}
6439			if (lsa) {
6440				(void)sctp_set_primary_addr(stcb, sa, NULL);
6441			}
6442		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6443			stcb->asoc.peer_supports_nat = 1;
6444		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6445			/* Peer supports pr-sctp */
6446			peer_supports_prsctp = 1;
6447		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6448			/* A supported extension chunk */
6449			struct sctp_supported_chunk_types_param *pr_supported;
6450			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6451			int num_ent, i;
6452
6453			phdr = sctp_get_next_param(m, offset,
6454			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6455			if (phdr == NULL) {
6456				return (-25);
6457			}
6458			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6459			num_ent = plen - sizeof(struct sctp_paramhdr);
6460			for (i = 0; i < num_ent; i++) {
6461				switch (pr_supported->chunk_types[i]) {
6462				case SCTP_ASCONF:
6463					peer_supports_asconf = 1;
6464					break;
6465				case SCTP_ASCONF_ACK:
6466					peer_supports_asconf_ack = 1;
6467					break;
6468				case SCTP_FORWARD_CUM_TSN:
6469					peer_supports_prsctp = 1;
6470					break;
6471				case SCTP_PACKET_DROPPED:
6472					peer_supports_pktdrop = 1;
6473					break;
6474				case SCTP_NR_SELECTIVE_ACK:
6475					peer_supports_nrsack = 1;
6476					break;
6477				case SCTP_STREAM_RESET:
6478					peer_supports_reconfig = 1;
6479					break;
6480				case SCTP_AUTHENTICATION:
6481					peer_supports_auth = 1;
6482					break;
6483				default:
6484					/* one I have not learned yet */
6485					break;
6486
6487				}
6488			}
6489		} else if (ptype == SCTP_RANDOM) {
6490			if (plen > sizeof(random_store))
6491				break;
6492			if (got_random) {
6493				/* already processed a RANDOM */
6494				goto next_param;
6495			}
6496			phdr = sctp_get_next_param(m, offset,
6497			    (struct sctp_paramhdr *)random_store,
6498			    min(sizeof(random_store), plen));
6499			if (phdr == NULL)
6500				return (-26);
6501			p_random = (struct sctp_auth_random *)phdr;
6502			random_len = plen - sizeof(*p_random);
6503			/* enforce the random length */
6504			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6505				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6506				return (-27);
6507			}
6508			got_random = 1;
6509		} else if (ptype == SCTP_HMAC_LIST) {
6510			uint16_t num_hmacs;
6511			uint16_t i;
6512
6513			if (plen > sizeof(hmacs_store))
6514				break;
6515			if (got_hmacs) {
6516				/* already processed a HMAC list */
6517				goto next_param;
6518			}
6519			phdr = sctp_get_next_param(m, offset,
6520			    (struct sctp_paramhdr *)hmacs_store,
6521			    min(plen, sizeof(hmacs_store)));
6522			if (phdr == NULL)
6523				return (-28);
6524			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6525			hmacs_len = plen - sizeof(*hmacs);
6526			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6527			/* validate the hmac list */
6528			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6529				return (-29);
6530			}
6531			if (stcb->asoc.peer_hmacs != NULL)
6532				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6533			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6534			if (stcb->asoc.peer_hmacs != NULL) {
6535				for (i = 0; i < num_hmacs; i++) {
6536					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6537					    ntohs(hmacs->hmac_ids[i]));
6538				}
6539			}
6540			got_hmacs = 1;
6541		} else if (ptype == SCTP_CHUNK_LIST) {
6542			int i;
6543
6544			if (plen > sizeof(chunks_store))
6545				break;
6546			if (got_chklist) {
6547				/* already processed a Chunks list */
6548				goto next_param;
6549			}
6550			phdr = sctp_get_next_param(m, offset,
6551			    (struct sctp_paramhdr *)chunks_store,
6552			    min(plen, sizeof(chunks_store)));
6553			if (phdr == NULL)
6554				return (-30);
6555			chunks = (struct sctp_auth_chunk_list *)phdr;
6556			num_chunks = plen - sizeof(*chunks);
6557			if (stcb->asoc.peer_auth_chunks != NULL)
6558				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6559			else
6560				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6561			for (i = 0; i < num_chunks; i++) {
6562				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6563				    stcb->asoc.peer_auth_chunks);
6564				/* record asconf/asconf-ack if listed */
6565				if (chunks->chunk_types[i] == SCTP_ASCONF)
6566					saw_asconf = 1;
6567				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6568					saw_asconf_ack = 1;
6569
6570			}
6571			got_chklist = 1;
6572		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6573			    (ptype == SCTP_STATE_COOKIE) ||
6574			    (ptype == SCTP_UNRECOG_PARAM) ||
6575			    (ptype == SCTP_COOKIE_PRESERVE) ||
6576			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6577			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6578			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6579			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6580		    (ptype == SCTP_SUCCESS_REPORT)) {
6581			 /* don't care */ ;
6582		} else {
6583			if ((ptype & 0x8000) == 0x0000) {
6584				/*
6585				 * must stop processing the rest of the
6586				 * param's. Any report bits were handled
6587				 * with the call to
6588				 * sctp_arethere_unrecognized_parameters()
6589				 * when the INIT or INIT-ACK was first seen.
6590				 */
6591				break;
6592			}
6593		}
6594
6595next_param:
6596		offset += SCTP_SIZE32(plen);
6597		if (offset >= limit) {
6598			break;
6599		}
6600		phdr = sctp_get_next_param(m, offset, &parm_buf,
6601		    sizeof(parm_buf));
6602	}
6603	/* Now check to see if we need to purge any addresses */
6604	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6605		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6606		    SCTP_ADDR_NOT_IN_ASSOC) {
6607			/* This address has been removed from the asoc */
6608			/* remove and free it */
6609			stcb->asoc.numnets--;
6610			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6611			sctp_free_remote_addr(net);
6612			if (net == stcb->asoc.primary_destination) {
6613				stcb->asoc.primary_destination = NULL;
6614				sctp_select_primary_destination(stcb);
6615			}
6616		}
6617	}
6618	if ((stcb->asoc.ecn_supported == 1) &&
6619	    (peer_supports_ecn == 0)) {
6620		stcb->asoc.ecn_supported = 0;
6621	}
6622	if ((stcb->asoc.prsctp_supported == 1) &&
6623	    (peer_supports_prsctp == 0)) {
6624		stcb->asoc.prsctp_supported = 0;
6625	}
6626	if ((stcb->asoc.auth_supported == 1) &&
6627	    ((peer_supports_auth == 0) ||
6628	    (got_random == 0) || (got_hmacs == 0))) {
6629		stcb->asoc.auth_supported = 0;
6630	}
6631	if ((stcb->asoc.asconf_supported == 1) &&
6632	    ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) ||
6633	    (stcb->asoc.auth_supported == 0) ||
6634	    (saw_asconf == 0) || (saw_asconf_ack == 0))) {
6635		stcb->asoc.asconf_supported = 0;
6636	}
6637	if ((stcb->asoc.reconfig_supported == 1) &&
6638	    (peer_supports_reconfig == 0)) {
6639		stcb->asoc.reconfig_supported = 0;
6640	}
6641	if ((stcb->asoc.nrsack_supported == 1) &&
6642	    (peer_supports_nrsack == 0)) {
6643		stcb->asoc.nrsack_supported = 0;
6644	}
6645	if ((stcb->asoc.pktdrop_supported == 1) &&
6646	    (peer_supports_pktdrop == 0)) {
6647		stcb->asoc.pktdrop_supported = 0;
6648	}
6649	/* validate authentication required parameters */
6650	if ((peer_supports_auth == 0) && (got_chklist == 1)) {
6651		/* peer does not support auth but sent a chunks list? */
6652		return (-31);
6653	}
6654	if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) {
6655		/* peer supports asconf but not auth? */
6656		return (-32);
6657	} else if ((peer_supports_asconf == 1) &&
6658		    (peer_supports_auth == 1) &&
6659	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6660		return (-33);
6661	}
6662	/* concatenate the full random key */
6663	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6664	if (chunks != NULL) {
6665		keylen += sizeof(*chunks) + num_chunks;
6666	}
6667	new_key = sctp_alloc_key(keylen);
6668	if (new_key != NULL) {
6669		/* copy in the RANDOM */
6670		if (p_random != NULL) {
6671			keylen = sizeof(*p_random) + random_len;
6672			bcopy(p_random, new_key->key, keylen);
6673		}
6674		/* append in the AUTH chunks */
6675		if (chunks != NULL) {
6676			bcopy(chunks, new_key->key + keylen,
6677			    sizeof(*chunks) + num_chunks);
6678			keylen += sizeof(*chunks) + num_chunks;
6679		}
6680		/* append in the HMACs */
6681		if (hmacs != NULL) {
6682			bcopy(hmacs, new_key->key + keylen,
6683			    sizeof(*hmacs) + hmacs_len);
6684		}
6685	} else {
6686		/* failed to get memory for the key */
6687		return (-34);
6688	}
6689	if (stcb->asoc.authinfo.peer_random != NULL)
6690		sctp_free_key(stcb->asoc.authinfo.peer_random);
6691	stcb->asoc.authinfo.peer_random = new_key;
6692	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6693	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6694
6695	return (0);
6696}
6697
6698int
6699sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6700    struct sctp_nets *net)
6701{
6702	/* make sure the requested primary address exists in the assoc */
6703	if (net == NULL && sa)
6704		net = sctp_findnet(stcb, sa);
6705
6706	if (net == NULL) {
6707		/* didn't find the requested primary address! */
6708		return (-1);
6709	} else {
6710		/* set the primary address */
6711		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6712			/* Must be confirmed, so queue to set */
6713			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6714			return (0);
6715		}
6716		stcb->asoc.primary_destination = net;
6717		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6718			sctp_free_remote_addr(stcb->asoc.alternate);
6719			stcb->asoc.alternate = NULL;
6720		}
6721		net = TAILQ_FIRST(&stcb->asoc.nets);
6722		if (net != stcb->asoc.primary_destination) {
6723			/*
6724			 * first one on the list is NOT the primary
6725			 * sctp_cmpaddr() is much more efficient if the
6726			 * primary is the first on the list, make it so.
6727			 */
6728			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6729			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6730		}
6731		return (0);
6732	}
6733}
6734
6735int
6736sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6737{
6738	/*
6739	 * This function serves two purposes. It will see if a TAG can be
6740	 * re-used and return 1 for yes it is ok and 0 for don't use that
6741	 * tag. A secondary function it will do is purge out old tags that
6742	 * can be removed.
6743	 */
6744	struct sctpvtaghead *chain;
6745	struct sctp_tagblock *twait_block;
6746	struct sctpasochead *head;
6747	struct sctp_tcb *stcb;
6748	int i;
6749
6750	SCTP_INP_INFO_RLOCK();
6751	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6752	    SCTP_BASE_INFO(hashasocmark))];
6753	LIST_FOREACH(stcb, head, sctp_asocs) {
6754		/*
6755		 * We choose not to lock anything here. TCB's can't be
6756		 * removed since we have the read lock, so they can't be
6757		 * freed on us, same thing for the INP. I may be wrong with
6758		 * this assumption, but we will go with it for now :-)
6759		 */
6760		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6761			continue;
6762		}
6763		if (stcb->asoc.my_vtag == tag) {
6764			/* candidate */
6765			if (stcb->rport != rport) {
6766				continue;
6767			}
6768			if (stcb->sctp_ep->sctp_lport != lport) {
6769				continue;
6770			}
6771			/* Its a used tag set */
6772			SCTP_INP_INFO_RUNLOCK();
6773			return (0);
6774		}
6775	}
6776	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6777	/* Now what about timed wait ? */
6778	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6779		/*
6780		 * Block(s) are present, lets see if we have this tag in the
6781		 * list
6782		 */
6783		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6784			if (twait_block->vtag_block[i].v_tag == 0) {
6785				/* not used */
6786				continue;
6787			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6788			    now->tv_sec) {
6789				/* Audit expires this guy */
6790				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6791				twait_block->vtag_block[i].v_tag = 0;
6792				twait_block->vtag_block[i].lport = 0;
6793				twait_block->vtag_block[i].rport = 0;
6794			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6795				    (twait_block->vtag_block[i].lport == lport) &&
6796			    (twait_block->vtag_block[i].rport == rport)) {
6797				/* Bad tag, sorry :< */
6798				SCTP_INP_INFO_RUNLOCK();
6799				return (0);
6800			}
6801		}
6802	}
6803	SCTP_INP_INFO_RUNLOCK();
6804	return (1);
6805}
6806
6807static void
6808sctp_drain_mbufs(struct sctp_tcb *stcb)
6809{
6810	/*
6811	 * We must hunt this association for MBUF's past the cumack (i.e.
6812	 * out of order data that we can renege on).
6813	 */
6814	struct sctp_association *asoc;
6815	struct sctp_tmit_chunk *chk, *nchk;
6816	uint32_t cumulative_tsn_p1;
6817	struct sctp_queued_to_read *ctl, *nctl;
6818	int cnt, strmat;
6819	uint32_t gap, i;
6820	int fnd = 0;
6821
6822	/* We look for anything larger than the cum-ack + 1 */
6823
6824	asoc = &stcb->asoc;
6825	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6826		/* none we can reneg on. */
6827		return;
6828	}
6829	SCTP_STAT_INCR(sctps_protocol_drains_done);
6830	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6831	cnt = 0;
6832	/* First look in the re-assembly queue */
6833	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6834		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6835			/* Yep it is above cum-ack */
6836			cnt++;
6837			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6838			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6839			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6840			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6841			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6842			if (chk->data) {
6843				sctp_m_freem(chk->data);
6844				chk->data = NULL;
6845			}
6846			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6847		}
6848	}
6849	/* Ok that was fun, now we will drain all the inbound streams? */
6850	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6851		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6852			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6853				/* Yep it is above cum-ack */
6854				cnt++;
6855				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6856				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6857				sctp_ucount_decr(asoc->cnt_on_all_streams);
6858				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6859				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6860				if (ctl->data) {
6861					sctp_m_freem(ctl->data);
6862					ctl->data = NULL;
6863				}
6864				sctp_free_remote_addr(ctl->whoFrom);
6865				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6866				SCTP_DECR_READQ_COUNT();
6867			}
6868		}
6869	}
6870	if (cnt) {
6871		/* We must back down to see what the new highest is */
6872		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6873			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6874			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6875				asoc->highest_tsn_inside_map = i;
6876				fnd = 1;
6877				break;
6878			}
6879		}
6880		if (!fnd) {
6881			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6882		}
6883		/*
6884		 * Question, should we go through the delivery queue? The
6885		 * only reason things are on here is the app not reading OR
6886		 * a p-d-api up. An attacker COULD send enough in to
6887		 * initiate the PD-API and then send a bunch of stuff to
6888		 * other streams... these would wind up on the delivery
6889		 * queue.. and then we would not get to them. But in order
6890		 * to do this I then have to back-track and un-deliver
6891		 * sequence numbers in streams.. el-yucko. I think for now
6892		 * we will NOT look at the delivery queue and leave it to be
6893		 * something to consider later. An alternative would be to
6894		 * abort the P-D-API with a notification and then deliver
6895		 * the data.... Or another method might be to keep track of
6896		 * how many times the situation occurs and if we see a
6897		 * possible attack underway just abort the association.
6898		 */
6899#ifdef SCTP_DEBUG
6900		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6901#endif
6902		/*
6903		 * Now do we need to find a new
6904		 * asoc->highest_tsn_inside_map?
6905		 */
6906		asoc->last_revoke_count = cnt;
6907		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6908		/* sa_ignore NO_NULL_CHK */
6909		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6910		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6911	}
6912	/*
6913	 * Another issue, in un-setting the TSN's in the mapping array we
6914	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6915	 * two things to occur. It may cause us to do extra work in checking
6916	 * for our mapping array movement. More importantly it may cause us
6917	 * to SACK every datagram. This may not be a bad thing though since
6918	 * we will recover once we get our cum-ack above and all this stuff
6919	 * we dumped recovered.
6920	 */
6921}
6922
6923void
6924sctp_drain()
6925{
6926	/*
6927	 * We must walk the PCB lists for ALL associations here. The system
6928	 * is LOW on MBUF's and needs help. This is where reneging will
6929	 * occur. We really hope this does NOT happen!
6930	 */
6931	VNET_ITERATOR_DECL(vnet_iter);
6932	VNET_LIST_RLOCK_NOSLEEP();
6933	VNET_FOREACH(vnet_iter) {
6934		CURVNET_SET(vnet_iter);
6935		struct sctp_inpcb *inp;
6936		struct sctp_tcb *stcb;
6937
6938		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6939		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6940#ifdef VIMAGE
6941			continue;
6942#else
6943			return;
6944#endif
6945		}
6946		SCTP_INP_INFO_RLOCK();
6947		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6948			/* For each endpoint */
6949			SCTP_INP_RLOCK(inp);
6950			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6951				/* For each association */
6952				SCTP_TCB_LOCK(stcb);
6953				sctp_drain_mbufs(stcb);
6954				SCTP_TCB_UNLOCK(stcb);
6955			}
6956			SCTP_INP_RUNLOCK(inp);
6957		}
6958		SCTP_INP_INFO_RUNLOCK();
6959		CURVNET_RESTORE();
6960	}
6961	VNET_LIST_RUNLOCK_NOSLEEP();
6962}
6963
6964/*
6965 * start a new iterator
6966 * iterates through all endpoints and associations based on the pcb_state
6967 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6968 * assocs and "ef" (optional) is executed when the iterator completes.
6969 * "inpf" (optional) is executed for each new endpoint as it is being
6970 * iterated through. inpe (optional) is called when the inp completes
6971 * its way through all the stcbs.
6972 */
6973int
6974sctp_initiate_iterator(inp_func inpf,
6975    asoc_func af,
6976    inp_func inpe,
6977    uint32_t pcb_state,
6978    uint32_t pcb_features,
6979    uint32_t asoc_state,
6980    void *argp,
6981    uint32_t argi,
6982    end_func ef,
6983    struct sctp_inpcb *s_inp,
6984    uint8_t chunk_output_off)
6985{
6986	struct sctp_iterator *it = NULL;
6987
6988	if (af == NULL) {
6989		return (-1);
6990	}
6991	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6992	    SCTP_M_ITER);
6993	if (it == NULL) {
6994		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6995		return (ENOMEM);
6996	}
6997	memset(it, 0, sizeof(*it));
6998	it->function_assoc = af;
6999	it->function_inp = inpf;
7000	if (inpf)
7001		it->done_current_ep = 0;
7002	else
7003		it->done_current_ep = 1;
7004	it->function_atend = ef;
7005	it->pointer = argp;
7006	it->val = argi;
7007	it->pcb_flags = pcb_state;
7008	it->pcb_features = pcb_features;
7009	it->asoc_state = asoc_state;
7010	it->function_inp_end = inpe;
7011	it->no_chunk_output = chunk_output_off;
7012	it->vn = curvnet;
7013	if (s_inp) {
7014		/* Assume lock is held here */
7015		it->inp = s_inp;
7016		SCTP_INP_INCR_REF(it->inp);
7017		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7018	} else {
7019		SCTP_INP_INFO_RLOCK();
7020		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7021		if (it->inp) {
7022			SCTP_INP_INCR_REF(it->inp);
7023		}
7024		SCTP_INP_INFO_RUNLOCK();
7025		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7026
7027	}
7028	SCTP_IPI_ITERATOR_WQ_LOCK();
7029
7030	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7031	if (sctp_it_ctl.iterator_running == 0) {
7032		sctp_wakeup_iterator();
7033	}
7034	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7035	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7036	return (0);
7037}
7038