sctp_pcb.c revision 271750
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 271750 2014-09-18 09:49:49Z 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		if (head == NULL) {
1445			goto null_return;
1446		}
1447		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1448			if (stcb->rport != rport) {
1449				/* remote port does not match */
1450				continue;
1451			}
1452			SCTP_TCB_LOCK(stcb);
1453			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1454				SCTP_TCB_UNLOCK(stcb);
1455				continue;
1456			}
1457			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1458				SCTP_TCB_UNLOCK(stcb);
1459				continue;
1460			}
1461			/* now look at the list of remote addresses */
1462			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1463#ifdef INVARIANTS
1464				if (net == (TAILQ_NEXT(net, sctp_next))) {
1465					panic("Corrupt net list");
1466				}
1467#endif
1468				if (net->ro._l_addr.sa.sa_family !=
1469				    remote->sa_family) {
1470					/* not the same family */
1471					continue;
1472				}
1473				switch (remote->sa_family) {
1474#ifdef INET
1475				case AF_INET:
1476					{
1477						struct sockaddr_in *sin,
1478						           *rsin;
1479
1480						sin = (struct sockaddr_in *)
1481						    &net->ro._l_addr;
1482						rsin = (struct sockaddr_in *)remote;
1483						if (sin->sin_addr.s_addr ==
1484						    rsin->sin_addr.s_addr) {
1485							/* found it */
1486							if (netp != NULL) {
1487								*netp = net;
1488							}
1489							if (locked_tcb == NULL) {
1490								SCTP_INP_DECR_REF(inp);
1491							} else if (locked_tcb != stcb) {
1492								SCTP_TCB_LOCK(locked_tcb);
1493							}
1494							if (locked_tcb) {
1495								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1496							}
1497							SCTP_INP_WUNLOCK(inp);
1498							SCTP_INP_INFO_RUNLOCK();
1499							return (stcb);
1500						}
1501						break;
1502					}
1503#endif
1504#ifdef INET6
1505				case AF_INET6:
1506					{
1507						struct sockaddr_in6 *sin6,
1508						            *rsin6;
1509
1510						sin6 = (struct sockaddr_in6 *)
1511						    &net->ro._l_addr;
1512						rsin6 = (struct sockaddr_in6 *)remote;
1513						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1514						    rsin6)) {
1515							/* found it */
1516							if (netp != NULL) {
1517								*netp = net;
1518							}
1519							if (locked_tcb == NULL) {
1520								SCTP_INP_DECR_REF(inp);
1521							} else if (locked_tcb != stcb) {
1522								SCTP_TCB_LOCK(locked_tcb);
1523							}
1524							if (locked_tcb) {
1525								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1526							}
1527							SCTP_INP_WUNLOCK(inp);
1528							SCTP_INP_INFO_RUNLOCK();
1529							return (stcb);
1530						}
1531						break;
1532					}
1533#endif
1534				default:
1535					/* TSNH */
1536					break;
1537				}
1538			}
1539			SCTP_TCB_UNLOCK(stcb);
1540		}
1541	}
1542null_return:
1543	/* clean up for returning null */
1544	if (locked_tcb) {
1545		SCTP_TCB_LOCK(locked_tcb);
1546		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1547	}
1548	SCTP_INP_WUNLOCK(inp);
1549	SCTP_INP_INFO_RUNLOCK();
1550	/* not found */
1551	return (NULL);
1552}
1553
1554
1555/*
1556 * Find an association for a specific endpoint using the association id given
1557 * out in the COMM_UP notification
1558 */
1559struct sctp_tcb *
1560sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1561{
1562	/*
1563	 * Use my the assoc_id to find a endpoint
1564	 */
1565	struct sctpasochead *head;
1566	struct sctp_tcb *stcb;
1567	uint32_t id;
1568
1569	if (inp == NULL) {
1570		SCTP_PRINTF("TSNH ep_associd\n");
1571		return (NULL);
1572	}
1573	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1574		SCTP_PRINTF("TSNH ep_associd0\n");
1575		return (NULL);
1576	}
1577	id = (uint32_t) asoc_id;
1578	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1579	if (head == NULL) {
1580		/* invalid id TSNH */
1581		SCTP_PRINTF("TSNH ep_associd1\n");
1582		return (NULL);
1583	}
1584	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1585		if (stcb->asoc.assoc_id == id) {
1586			if (inp != stcb->sctp_ep) {
1587				/*
1588				 * some other guy has the same id active (id
1589				 * collision ??).
1590				 */
1591				SCTP_PRINTF("TSNH ep_associd2\n");
1592				continue;
1593			}
1594			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1595				continue;
1596			}
1597			if (want_lock) {
1598				SCTP_TCB_LOCK(stcb);
1599			}
1600			return (stcb);
1601		}
1602	}
1603	return (NULL);
1604}
1605
1606
1607struct sctp_tcb *
1608sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1609{
1610	struct sctp_tcb *stcb;
1611
1612	SCTP_INP_RLOCK(inp);
1613	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1614	SCTP_INP_RUNLOCK(inp);
1615	return (stcb);
1616}
1617
1618
1619/*
1620 * Endpoint probe expects that the INP_INFO is locked.
1621 */
1622static struct sctp_inpcb *
1623sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1624    uint16_t lport, uint32_t vrf_id)
1625{
1626	struct sctp_inpcb *inp;
1627	struct sctp_laddr *laddr;
1628
1629#ifdef INET
1630	struct sockaddr_in *sin;
1631
1632#endif
1633#ifdef INET6
1634	struct sockaddr_in6 *sin6;
1635	struct sockaddr_in6 *intf_addr6;
1636
1637#endif
1638	int fnd;
1639
1640#ifdef INET
1641	sin = NULL;
1642#endif
1643#ifdef INET6
1644	sin6 = NULL;
1645#endif
1646	switch (nam->sa_family) {
1647#ifdef INET
1648	case AF_INET:
1649		sin = (struct sockaddr_in *)nam;
1650		break;
1651#endif
1652#ifdef INET6
1653	case AF_INET6:
1654		sin6 = (struct sockaddr_in6 *)nam;
1655		break;
1656#endif
1657	default:
1658		/* unsupported family */
1659		return (NULL);
1660	}
1661
1662	if (head == NULL)
1663		return (NULL);
1664
1665	LIST_FOREACH(inp, head, sctp_hash) {
1666		SCTP_INP_RLOCK(inp);
1667		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1668			SCTP_INP_RUNLOCK(inp);
1669			continue;
1670		}
1671		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1672		    (inp->sctp_lport == lport)) {
1673			/* got it */
1674			switch (nam->sa_family) {
1675#ifdef INET
1676			case AF_INET:
1677				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1678				    SCTP_IPV6_V6ONLY(inp)) {
1679					/*
1680					 * IPv4 on a IPv6 socket with ONLY
1681					 * IPv6 set
1682					 */
1683					SCTP_INP_RUNLOCK(inp);
1684					continue;
1685				}
1686				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1687				    &sin->sin_addr) != 0) {
1688					SCTP_INP_RUNLOCK(inp);
1689					continue;
1690				}
1691				break;
1692#endif
1693#ifdef INET6
1694			case AF_INET6:
1695				/*
1696				 * A V6 address and the endpoint is NOT
1697				 * bound V6
1698				 */
1699				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1700					SCTP_INP_RUNLOCK(inp);
1701					continue;
1702				}
1703				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1704				    &sin6->sin6_addr) != 0) {
1705					SCTP_INP_RUNLOCK(inp);
1706					continue;
1707				}
1708				break;
1709#endif
1710			default:
1711				break;
1712			}
1713			/* does a VRF id match? */
1714			fnd = 0;
1715			if (inp->def_vrf_id == vrf_id)
1716				fnd = 1;
1717
1718			SCTP_INP_RUNLOCK(inp);
1719			if (!fnd)
1720				continue;
1721			return (inp);
1722		}
1723		SCTP_INP_RUNLOCK(inp);
1724	}
1725	switch (nam->sa_family) {
1726#ifdef INET
1727	case AF_INET:
1728		if (sin->sin_addr.s_addr == INADDR_ANY) {
1729			/* Can't hunt for one that has no address specified */
1730			return (NULL);
1731		}
1732		break;
1733#endif
1734#ifdef INET6
1735	case AF_INET6:
1736		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1737			/* Can't hunt for one that has no address specified */
1738			return (NULL);
1739		}
1740		break;
1741#endif
1742	default:
1743		break;
1744	}
1745	/*
1746	 * ok, not bound to all so see if we can find a EP bound to this
1747	 * address.
1748	 */
1749	LIST_FOREACH(inp, head, sctp_hash) {
1750		SCTP_INP_RLOCK(inp);
1751		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1752			SCTP_INP_RUNLOCK(inp);
1753			continue;
1754		}
1755		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1756			SCTP_INP_RUNLOCK(inp);
1757			continue;
1758		}
1759		/*
1760		 * Ok this could be a likely candidate, look at all of its
1761		 * addresses
1762		 */
1763		if (inp->sctp_lport != lport) {
1764			SCTP_INP_RUNLOCK(inp);
1765			continue;
1766		}
1767		/* does a VRF id match? */
1768		fnd = 0;
1769		if (inp->def_vrf_id == vrf_id)
1770			fnd = 1;
1771
1772		if (!fnd) {
1773			SCTP_INP_RUNLOCK(inp);
1774			continue;
1775		}
1776		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1777			if (laddr->ifa == NULL) {
1778				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1779				    __FUNCTION__);
1780				continue;
1781			}
1782			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1783			    (void *)laddr->ifa);
1784			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1785				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1786				continue;
1787			}
1788			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1789				/* possible, see if it matches */
1790				switch (nam->sa_family) {
1791#ifdef INET
1792				case AF_INET:
1793					if (sin->sin_addr.s_addr ==
1794					    laddr->ifa->address.sin.sin_addr.s_addr) {
1795						SCTP_INP_RUNLOCK(inp);
1796						return (inp);
1797					}
1798					break;
1799#endif
1800#ifdef INET6
1801				case AF_INET6:
1802					intf_addr6 = &laddr->ifa->address.sin6;
1803					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1804					    intf_addr6)) {
1805						SCTP_INP_RUNLOCK(inp);
1806						return (inp);
1807					}
1808					break;
1809#endif
1810				}
1811			}
1812		}
1813		SCTP_INP_RUNLOCK(inp);
1814	}
1815	return (NULL);
1816}
1817
1818
1819static struct sctp_inpcb *
1820sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1821{
1822	struct sctppcbhead *head;
1823	struct sctp_inpcb *t_inp;
1824	int fnd;
1825
1826	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1827	    SCTP_BASE_INFO(hashmark))];
1828	LIST_FOREACH(t_inp, head, sctp_hash) {
1829		if (t_inp->sctp_lport != lport) {
1830			continue;
1831		}
1832		/* is it in the VRF in question */
1833		fnd = 0;
1834		if (t_inp->def_vrf_id == vrf_id)
1835			fnd = 1;
1836		if (!fnd)
1837			continue;
1838
1839		/* This one is in use. */
1840		/* check the v6/v4 binding issue */
1841		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1842		    SCTP_IPV6_V6ONLY(t_inp)) {
1843			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1844				/* collision in V6 space */
1845				return (t_inp);
1846			} else {
1847				/* inp is BOUND_V4 no conflict */
1848				continue;
1849			}
1850		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1851			/* t_inp is bound v4 and v6, conflict always */
1852			return (t_inp);
1853		} else {
1854			/* t_inp is bound only V4 */
1855			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1856			    SCTP_IPV6_V6ONLY(inp)) {
1857				/* no conflict */
1858				continue;
1859			}
1860			/* else fall through to conflict */
1861		}
1862		return (t_inp);
1863	}
1864	return (NULL);
1865}
1866
1867
1868int
1869sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1870{
1871	/* For 1-2-1 with port reuse */
1872	struct sctppcbhead *head;
1873	struct sctp_inpcb *tinp;
1874
1875	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1876		/* only works with port reuse on */
1877		return (-1);
1878	}
1879	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1880		return (0);
1881	}
1882	SCTP_INP_RUNLOCK(inp);
1883	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1884	    SCTP_BASE_INFO(hashmark))];
1885	/* Kick out all non-listeners to the TCP hash */
1886	LIST_FOREACH(tinp, head, sctp_hash) {
1887		if (tinp->sctp_lport != inp->sctp_lport) {
1888			continue;
1889		}
1890		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1891			continue;
1892		}
1893		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1894			continue;
1895		}
1896		if (tinp->sctp_socket->so_qlimit) {
1897			continue;
1898		}
1899		SCTP_INP_WLOCK(tinp);
1900		LIST_REMOVE(tinp, sctp_hash);
1901		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1902		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1903		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1904		SCTP_INP_WUNLOCK(tinp);
1905	}
1906	SCTP_INP_WLOCK(inp);
1907	/* Pull from where he was */
1908	LIST_REMOVE(inp, sctp_hash);
1909	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1910	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1911	LIST_INSERT_HEAD(head, inp, sctp_hash);
1912	SCTP_INP_WUNLOCK(inp);
1913	SCTP_INP_RLOCK(inp);
1914	return (0);
1915}
1916
1917
1918struct sctp_inpcb *
1919sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1920    uint32_t vrf_id)
1921{
1922	/*
1923	 * First we check the hash table to see if someone has this port
1924	 * bound with just the port.
1925	 */
1926	struct sctp_inpcb *inp;
1927	struct sctppcbhead *head;
1928	int lport;
1929	unsigned int i;
1930
1931#ifdef INET
1932	struct sockaddr_in *sin;
1933
1934#endif
1935#ifdef INET6
1936	struct sockaddr_in6 *sin6;
1937
1938#endif
1939
1940	switch (nam->sa_family) {
1941#ifdef INET
1942	case AF_INET:
1943		sin = (struct sockaddr_in *)nam;
1944		lport = sin->sin_port;
1945		break;
1946#endif
1947#ifdef INET6
1948	case AF_INET6:
1949		sin6 = (struct sockaddr_in6 *)nam;
1950		lport = sin6->sin6_port;
1951		break;
1952#endif
1953	default:
1954		return (NULL);
1955	}
1956	/*
1957	 * I could cheat here and just cast to one of the types but we will
1958	 * do it right. It also provides the check against an Unsupported
1959	 * type too.
1960	 */
1961	/* Find the head of the ALLADDR chain */
1962	if (have_lock == 0) {
1963		SCTP_INP_INFO_RLOCK();
1964	}
1965	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1966	    SCTP_BASE_INFO(hashmark))];
1967	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1968
1969	/*
1970	 * If the TCP model exists it could be that the main listening
1971	 * endpoint is gone but there still exists a connected socket for
1972	 * this guy. If so we can return the first one that we find. This
1973	 * may NOT be the correct one so the caller should be wary on the
1974	 * returned INP. Currently the only caller that sets find_tcp_pool
1975	 * is in bindx where we are verifying that a user CAN bind the
1976	 * address. He either has bound it already, or someone else has, or
1977	 * its open to bind, so this is good enough.
1978	 */
1979	if (inp == NULL && find_tcp_pool) {
1980		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1981			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1982			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1983			if (inp) {
1984				break;
1985			}
1986		}
1987	}
1988	if (inp) {
1989		SCTP_INP_INCR_REF(inp);
1990	}
1991	if (have_lock == 0) {
1992		SCTP_INP_INFO_RUNLOCK();
1993	}
1994	return (inp);
1995}
1996
1997
1998/*
1999 * Find an association for an endpoint with the pointer to whom you want to
2000 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
2001 * need to change the *to to some other struct like a mbuf...
2002 */
2003struct sctp_tcb *
2004sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
2005    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
2006    uint32_t vrf_id)
2007{
2008	struct sctp_inpcb *inp = NULL;
2009	struct sctp_tcb *stcb;
2010
2011	SCTP_INP_INFO_RLOCK();
2012	if (find_tcp_pool) {
2013		if (inp_p != NULL) {
2014			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
2015			    vrf_id);
2016		} else {
2017			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
2018			    vrf_id);
2019		}
2020		if (stcb != NULL) {
2021			SCTP_INP_INFO_RUNLOCK();
2022			return (stcb);
2023		}
2024	}
2025	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
2026	if (inp_p != NULL) {
2027		*inp_p = inp;
2028	}
2029	SCTP_INP_INFO_RUNLOCK();
2030	if (inp == NULL) {
2031		return (NULL);
2032	}
2033	/*
2034	 * ok, we have an endpoint, now lets find the assoc for it (if any)
2035	 * we now place the source address or from in the to of the find
2036	 * endpoint call. Since in reality this chain is used from the
2037	 * inbound packet side.
2038	 */
2039	if (inp_p != NULL) {
2040		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
2041		    NULL);
2042	} else {
2043		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
2044		    NULL);
2045	}
2046	return (stcb);
2047}
2048
2049
2050/*
2051 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
2052 * find all addresses that the sender has specified in any address list. Each
2053 * address will be used to lookup the TCB and see if one exits.
2054 */
2055static struct sctp_tcb *
2056sctp_findassociation_special_addr(struct mbuf *m, int offset,
2057    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
2058    struct sockaddr *dst)
2059{
2060	struct sctp_paramhdr *phdr, parm_buf;
2061
2062#if defined(INET) || defined(INET6)
2063	struct sctp_tcb *stcb;
2064	uint16_t ptype;
2065
2066#endif
2067	uint16_t plen;
2068
2069#ifdef INET
2070	struct sockaddr_in sin4;
2071
2072#endif
2073#ifdef INET6
2074	struct sockaddr_in6 sin6;
2075
2076#endif
2077
2078#ifdef INET
2079	memset(&sin4, 0, sizeof(sin4));
2080	sin4.sin_len = sizeof(sin4);
2081	sin4.sin_family = AF_INET;
2082	sin4.sin_port = sh->src_port;
2083#endif
2084#ifdef INET6
2085	memset(&sin6, 0, sizeof(sin6));
2086	sin6.sin6_len = sizeof(sin6);
2087	sin6.sin6_family = AF_INET6;
2088	sin6.sin6_port = sh->src_port;
2089#endif
2090
2091	offset += sizeof(struct sctp_init_chunk);
2092
2093	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2094	while (phdr != NULL) {
2095		/* now we must see if we want the parameter */
2096#if defined(INET) || defined(INET6)
2097		ptype = ntohs(phdr->param_type);
2098#endif
2099		plen = ntohs(phdr->param_length);
2100		if (plen == 0) {
2101			break;
2102		}
2103#ifdef INET
2104		if (ptype == SCTP_IPV4_ADDRESS &&
2105		    plen == sizeof(struct sctp_ipv4addr_param)) {
2106			/* Get the rest of the address */
2107			struct sctp_ipv4addr_param ip4_parm, *p4;
2108
2109			phdr = sctp_get_next_param(m, offset,
2110			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2111			if (phdr == NULL) {
2112				return (NULL);
2113			}
2114			p4 = (struct sctp_ipv4addr_param *)phdr;
2115			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2116			/* look it up */
2117			stcb = sctp_findassociation_ep_addr(inp_p,
2118			    (struct sockaddr *)&sin4, netp, dst, NULL);
2119			if (stcb != NULL) {
2120				return (stcb);
2121			}
2122		}
2123#endif
2124#ifdef INET6
2125		if (ptype == SCTP_IPV6_ADDRESS &&
2126		    plen == sizeof(struct sctp_ipv6addr_param)) {
2127			/* Get the rest of the address */
2128			struct sctp_ipv6addr_param ip6_parm, *p6;
2129
2130			phdr = sctp_get_next_param(m, offset,
2131			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2132			if (phdr == NULL) {
2133				return (NULL);
2134			}
2135			p6 = (struct sctp_ipv6addr_param *)phdr;
2136			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2137			/* look it up */
2138			stcb = sctp_findassociation_ep_addr(inp_p,
2139			    (struct sockaddr *)&sin6, netp, dst, NULL);
2140			if (stcb != NULL) {
2141				return (stcb);
2142			}
2143		}
2144#endif
2145		offset += SCTP_SIZE32(plen);
2146		phdr = sctp_get_next_param(m, offset, &parm_buf,
2147		    sizeof(parm_buf));
2148	}
2149	return (NULL);
2150}
2151
2152static struct sctp_tcb *
2153sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2154    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2155    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2156{
2157	/*
2158	 * Use my vtag to hash. If we find it we then verify the source addr
2159	 * is in the assoc. If all goes well we save a bit on rec of a
2160	 * packet.
2161	 */
2162	struct sctpasochead *head;
2163	struct sctp_nets *net;
2164	struct sctp_tcb *stcb;
2165
2166	SCTP_INP_INFO_RLOCK();
2167	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2168	    SCTP_BASE_INFO(hashasocmark))];
2169	if (head == NULL) {
2170		/* invalid vtag */
2171		SCTP_INP_INFO_RUNLOCK();
2172		return (NULL);
2173	}
2174	LIST_FOREACH(stcb, head, sctp_asocs) {
2175		SCTP_INP_RLOCK(stcb->sctp_ep);
2176		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2177			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2178			continue;
2179		}
2180		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2181			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2182			continue;
2183		}
2184		SCTP_TCB_LOCK(stcb);
2185		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2186		if (stcb->asoc.my_vtag == vtag) {
2187			/* candidate */
2188			if (stcb->rport != rport) {
2189				SCTP_TCB_UNLOCK(stcb);
2190				continue;
2191			}
2192			if (stcb->sctp_ep->sctp_lport != lport) {
2193				SCTP_TCB_UNLOCK(stcb);
2194				continue;
2195			}
2196			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2197				SCTP_TCB_UNLOCK(stcb);
2198				continue;
2199			}
2200			/* RRS:Need toaddr check here */
2201			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2202				/* Endpoint does not own this address */
2203				SCTP_TCB_UNLOCK(stcb);
2204				continue;
2205			}
2206			if (remote_tag) {
2207				/*
2208				 * If we have both vtags that's all we match
2209				 * on
2210				 */
2211				if (stcb->asoc.peer_vtag == remote_tag) {
2212					/*
2213					 * If both tags match we consider it
2214					 * conclusive and check NO
2215					 * source/destination addresses
2216					 */
2217					goto conclusive;
2218				}
2219			}
2220			if (skip_src_check) {
2221		conclusive:
2222				if (from) {
2223					*netp = sctp_findnet(stcb, from);
2224				} else {
2225					*netp = NULL;	/* unknown */
2226				}
2227				if (inp_p)
2228					*inp_p = stcb->sctp_ep;
2229				SCTP_INP_INFO_RUNLOCK();
2230				return (stcb);
2231			}
2232			net = sctp_findnet(stcb, from);
2233			if (net) {
2234				/* yep its him. */
2235				*netp = net;
2236				SCTP_STAT_INCR(sctps_vtagexpress);
2237				*inp_p = stcb->sctp_ep;
2238				SCTP_INP_INFO_RUNLOCK();
2239				return (stcb);
2240			} else {
2241				/*
2242				 * not him, this should only happen in rare
2243				 * cases so I peg it.
2244				 */
2245				SCTP_STAT_INCR(sctps_vtagbogus);
2246			}
2247		}
2248		SCTP_TCB_UNLOCK(stcb);
2249	}
2250	SCTP_INP_INFO_RUNLOCK();
2251	return (NULL);
2252}
2253
2254
2255/*
2256 * Find an association with the pointer to the inbound IP packet. This can be
2257 * a IPv4 or IPv6 packet.
2258 */
2259struct sctp_tcb *
2260sctp_findassociation_addr(struct mbuf *m, int offset,
2261    struct sockaddr *src, struct sockaddr *dst,
2262    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2263    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2264{
2265	int find_tcp_pool;
2266	struct sctp_tcb *stcb;
2267	struct sctp_inpcb *inp;
2268
2269	if (sh->v_tag) {
2270		/* we only go down this path if vtag is non-zero */
2271		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2272		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2273		if (stcb) {
2274			return (stcb);
2275		}
2276	}
2277	find_tcp_pool = 0;
2278	if ((ch->chunk_type != SCTP_INITIATION) &&
2279	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2280	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2281	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2282		/* Other chunk types go to the tcp pool. */
2283		find_tcp_pool = 1;
2284	}
2285	if (inp_p) {
2286		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2287		    find_tcp_pool, vrf_id);
2288		inp = *inp_p;
2289	} else {
2290		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2291		    find_tcp_pool, vrf_id);
2292	}
2293	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2294	if (stcb == NULL && inp) {
2295		/* Found a EP but not this address */
2296		if ((ch->chunk_type == SCTP_INITIATION) ||
2297		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2298			/*-
2299			 * special hook, we do NOT return linp or an
2300			 * association that is linked to an existing
2301			 * association that is under the TCP pool (i.e. no
2302			 * listener exists). The endpoint finding routine
2303			 * will always find a listener before examining the
2304			 * TCP pool.
2305			 */
2306			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2307				if (inp_p) {
2308					*inp_p = NULL;
2309				}
2310				return (NULL);
2311			}
2312			stcb = sctp_findassociation_special_addr(m,
2313			    offset, sh, &inp, netp, dst);
2314			if (inp_p != NULL) {
2315				*inp_p = inp;
2316			}
2317		}
2318	}
2319	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2320	return (stcb);
2321}
2322
2323/*
2324 * lookup an association by an ASCONF lookup address.
2325 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2326 */
2327struct sctp_tcb *
2328sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2329    struct sockaddr *dst, struct sctphdr *sh,
2330    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2331{
2332	struct sctp_tcb *stcb;
2333	union sctp_sockstore remote_store;
2334	struct sctp_paramhdr parm_buf, *phdr;
2335	int ptype;
2336	int zero_address = 0;
2337
2338#ifdef INET
2339	struct sockaddr_in *sin;
2340
2341#endif
2342#ifdef INET6
2343	struct sockaddr_in6 *sin6;
2344
2345#endif
2346
2347	memset(&remote_store, 0, sizeof(remote_store));
2348	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2349	    &parm_buf, sizeof(struct sctp_paramhdr));
2350	if (phdr == NULL) {
2351		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2352		    __FUNCTION__);
2353		return NULL;
2354	}
2355	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2356	/* get the correlation address */
2357	switch (ptype) {
2358#ifdef INET6
2359	case SCTP_IPV6_ADDRESS:
2360		{
2361			/* ipv6 address param */
2362			struct sctp_ipv6addr_param *p6, p6_buf;
2363
2364			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2365				return NULL;
2366			}
2367			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2368			    offset + sizeof(struct sctp_asconf_chunk),
2369			    &p6_buf.ph, sizeof(*p6));
2370			if (p6 == NULL) {
2371				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2372				    __FUNCTION__);
2373				return (NULL);
2374			}
2375			sin6 = &remote_store.sin6;
2376			sin6->sin6_family = AF_INET6;
2377			sin6->sin6_len = sizeof(*sin6);
2378			sin6->sin6_port = sh->src_port;
2379			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2380			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2381				zero_address = 1;
2382			break;
2383		}
2384#endif
2385#ifdef INET
2386	case SCTP_IPV4_ADDRESS:
2387		{
2388			/* ipv4 address param */
2389			struct sctp_ipv4addr_param *p4, p4_buf;
2390
2391			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2392				return NULL;
2393			}
2394			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2395			    offset + sizeof(struct sctp_asconf_chunk),
2396			    &p4_buf.ph, sizeof(*p4));
2397			if (p4 == NULL) {
2398				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2399				    __FUNCTION__);
2400				return (NULL);
2401			}
2402			sin = &remote_store.sin;
2403			sin->sin_family = AF_INET;
2404			sin->sin_len = sizeof(*sin);
2405			sin->sin_port = sh->src_port;
2406			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2407			if (sin->sin_addr.s_addr == INADDR_ANY)
2408				zero_address = 1;
2409			break;
2410		}
2411#endif
2412	default:
2413		/* invalid address param type */
2414		return NULL;
2415	}
2416
2417	if (zero_address) {
2418		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2419		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2420		if (stcb != NULL) {
2421			SCTP_INP_DECR_REF(*inp_p);
2422		}
2423	} else {
2424		stcb = sctp_findassociation_ep_addr(inp_p,
2425		    &remote_store.sa, netp,
2426		    dst, NULL);
2427	}
2428	return (stcb);
2429}
2430
2431
2432/*
2433 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2434 * addresses. This way if we don't get a bind we by default pick a ephemeral
2435 * port with all addresses bound.
2436 */
2437int
2438sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2439{
2440	/*
2441	 * we get called when a new endpoint starts up. We need to allocate
2442	 * the sctp_inpcb structure from the zone and init it. Mark it as
2443	 * unbound and find a port that we can use as an ephemeral with
2444	 * INADDR_ANY. If the user binds later no problem we can then add in
2445	 * the specific addresses. And setup the default parameters for the
2446	 * EP.
2447	 */
2448	int i, error;
2449	struct sctp_inpcb *inp;
2450	struct sctp_pcb *m;
2451	struct timeval time;
2452	sctp_sharedkey_t *null_key;
2453
2454	error = 0;
2455
2456	SCTP_INP_INFO_WLOCK();
2457	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2458	if (inp == NULL) {
2459		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2460		SCTP_INP_INFO_WUNLOCK();
2461		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2462		return (ENOBUFS);
2463	}
2464	/* zap it */
2465	bzero(inp, sizeof(*inp));
2466
2467	/* bump generations */
2468	/* setup socket pointers */
2469	inp->sctp_socket = so;
2470	inp->ip_inp.inp.inp_socket = so;
2471	inp->ip_inp.inp.inp_cred = crhold(so->so_cred);
2472#ifdef INET6
2473	if (INP_SOCKAF(so) == AF_INET6) {
2474		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2475			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2476		}
2477		if (MODULE_GLOBAL(ip6_v6only)) {
2478			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2479		}
2480	}
2481#endif
2482	inp->sctp_associd_counter = 1;
2483	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2484	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2485	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2486	inp->ecn_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_ecn_enable);
2487	inp->prsctp_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pr_enable);
2488	if (SCTP_BASE_SYSCTL(sctp_auth_disable)) {
2489		inp->auth_supported = 0;
2490	} else {
2491		inp->auth_supported = 1;
2492	}
2493	inp->asconf_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_asconf_enable);
2494	inp->reconfig_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_reconfig_enable);
2495	inp->nrsack_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_nrsack_enable);
2496	inp->pktdrop_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pktdrop_enable);
2497	/* init the small hash table we use to track asocid <-> tcb */
2498	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2499	if (inp->sctp_asocidhash == NULL) {
2500		crfree(inp->ip_inp.inp.inp_cred);
2501		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2502		SCTP_INP_INFO_WUNLOCK();
2503		return (ENOBUFS);
2504	}
2505#ifdef IPSEC
2506	{
2507		struct inpcbpolicy *pcb_sp = NULL;
2508
2509		error = ipsec_init_policy(so, &pcb_sp);
2510		/* Arrange to share the policy */
2511		inp->ip_inp.inp.inp_sp = pcb_sp;
2512		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2513	}
2514	if (error != 0) {
2515		crfree(inp->ip_inp.inp.inp_cred);
2516		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2517		SCTP_INP_INFO_WUNLOCK();
2518		return error;
2519	}
2520#endif				/* IPSEC */
2521	SCTP_INCR_EP_COUNT();
2522	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2523	SCTP_INP_INFO_WUNLOCK();
2524
2525	so->so_pcb = (caddr_t)inp;
2526
2527	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2528		/* UDP style socket */
2529		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2530		    SCTP_PCB_FLAGS_UNBOUND);
2531		/* Be sure it is NON-BLOCKING IO for UDP */
2532		/* SCTP_SET_SO_NBIO(so); */
2533	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2534		/* TCP style socket */
2535		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2536		    SCTP_PCB_FLAGS_UNBOUND);
2537		/* Be sure we have blocking IO by default */
2538		SCTP_CLEAR_SO_NBIO(so);
2539	} else {
2540		/*
2541		 * unsupported socket type (RAW, etc)- in case we missed it
2542		 * in protosw
2543		 */
2544		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2545		so->so_pcb = NULL;
2546		crfree(inp->ip_inp.inp.inp_cred);
2547		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2548		return (EOPNOTSUPP);
2549	}
2550	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2551		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2552		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2553	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2554		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2555		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2556	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2557		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2558		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2559	}
2560	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2561	    &inp->sctp_hashmark);
2562	if (inp->sctp_tcbhash == NULL) {
2563		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2564		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2565		so->so_pcb = NULL;
2566		crfree(inp->ip_inp.inp.inp_cred);
2567		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2568		return (ENOBUFS);
2569	}
2570	inp->def_vrf_id = vrf_id;
2571
2572	SCTP_INP_INFO_WLOCK();
2573	SCTP_INP_LOCK_INIT(inp);
2574	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2575	SCTP_INP_READ_INIT(inp);
2576	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2577	/* lock the new ep */
2578	SCTP_INP_WLOCK(inp);
2579
2580	/* add it to the info area */
2581	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2582	SCTP_INP_INFO_WUNLOCK();
2583
2584	TAILQ_INIT(&inp->read_queue);
2585	LIST_INIT(&inp->sctp_addr_list);
2586
2587	LIST_INIT(&inp->sctp_asoc_list);
2588
2589#ifdef SCTP_TRACK_FREED_ASOCS
2590	/* TEMP CODE */
2591	LIST_INIT(&inp->sctp_asoc_free_list);
2592#endif
2593	/* Init the timer structure for signature change */
2594	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2595	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2596
2597	/* now init the actual endpoint default data */
2598	m = &inp->sctp_ep;
2599
2600	/* setup the base timeout information */
2601	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2602	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2603	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2604	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2605	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2606	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2607	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2608	/* all max/min max are in ms */
2609	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2610	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2611	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2612	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2613	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2614	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2615	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2616	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2617	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2618	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2619	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2620	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2621	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2622
2623	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2624	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2625	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2626	/* number of streams to pre-open on a association */
2627	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2628
2629	/* Add adaptation cookie */
2630	m->adaptation_layer_indicator = 0;
2631	m->adaptation_layer_indicator_provided = 0;
2632
2633	/* seed random number generator */
2634	m->random_counter = 1;
2635	m->store_at = SCTP_SIGNATURE_SIZE;
2636	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2637	sctp_fill_random_store(m);
2638
2639	/* Minimum cookie size */
2640	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2641	    sizeof(struct sctp_state_cookie);
2642	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2643
2644	/* Setup the initial secret */
2645	(void)SCTP_GETTIME_TIMEVAL(&time);
2646	m->time_of_secret_change = time.tv_sec;
2647
2648	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2649		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2650	}
2651	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2652
2653	/* How long is a cookie good for ? */
2654	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2655	/*
2656	 * Initialize authentication parameters
2657	 */
2658	m->local_hmacs = sctp_default_supported_hmaclist();
2659	m->local_auth_chunks = sctp_alloc_chunklist();
2660	if (inp->asconf_supported) {
2661		sctp_auth_add_chunk(SCTP_ASCONF, m->local_auth_chunks);
2662		sctp_auth_add_chunk(SCTP_ASCONF_ACK, m->local_auth_chunks);
2663	}
2664	m->default_dscp = 0;
2665#ifdef INET6
2666	m->default_flowlabel = 0;
2667#endif
2668	m->port = 0;		/* encapsulation disabled by default */
2669	LIST_INIT(&m->shared_keys);
2670	/* add default NULL key as key id 0 */
2671	null_key = sctp_alloc_sharedkey();
2672	sctp_insert_sharedkey(&m->shared_keys, null_key);
2673	SCTP_INP_WUNLOCK(inp);
2674#ifdef SCTP_LOG_CLOSING
2675	sctp_log_closing(inp, NULL, 12);
2676#endif
2677	return (error);
2678}
2679
2680
2681void
2682sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2683    struct sctp_tcb *stcb)
2684{
2685	struct sctp_nets *net;
2686	uint16_t lport, rport;
2687	struct sctppcbhead *head;
2688	struct sctp_laddr *laddr, *oladdr;
2689
2690	atomic_add_int(&stcb->asoc.refcnt, 1);
2691	SCTP_TCB_UNLOCK(stcb);
2692	SCTP_INP_INFO_WLOCK();
2693	SCTP_INP_WLOCK(old_inp);
2694	SCTP_INP_WLOCK(new_inp);
2695	SCTP_TCB_LOCK(stcb);
2696	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2697
2698	new_inp->sctp_ep.time_of_secret_change =
2699	    old_inp->sctp_ep.time_of_secret_change;
2700	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2701	    sizeof(old_inp->sctp_ep.secret_key));
2702	new_inp->sctp_ep.current_secret_number =
2703	    old_inp->sctp_ep.current_secret_number;
2704	new_inp->sctp_ep.last_secret_number =
2705	    old_inp->sctp_ep.last_secret_number;
2706	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2707
2708	/* make it so new data pours into the new socket */
2709	stcb->sctp_socket = new_inp->sctp_socket;
2710	stcb->sctp_ep = new_inp;
2711
2712	/* Copy the port across */
2713	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2714	rport = stcb->rport;
2715	/* Pull the tcb from the old association */
2716	LIST_REMOVE(stcb, sctp_tcbhash);
2717	LIST_REMOVE(stcb, sctp_tcblist);
2718	if (stcb->asoc.in_asocid_hash) {
2719		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2720	}
2721	/* Now insert the new_inp into the TCP connected hash */
2722	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2723
2724	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2725	/* Its safe to access */
2726	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2727
2728	/* Now move the tcb into the endpoint list */
2729	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2730	/*
2731	 * Question, do we even need to worry about the ep-hash since we
2732	 * only have one connection? Probably not :> so lets get rid of it
2733	 * and not suck up any kernel memory in that.
2734	 */
2735	if (stcb->asoc.in_asocid_hash) {
2736		struct sctpasochead *lhd;
2737
2738		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2739		    new_inp->hashasocidmark)];
2740		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2741	}
2742	/* Ok. Let's restart timer. */
2743	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2744		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2745		    stcb, net);
2746	}
2747
2748	SCTP_INP_INFO_WUNLOCK();
2749	if (new_inp->sctp_tcbhash != NULL) {
2750		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2751		new_inp->sctp_tcbhash = NULL;
2752	}
2753	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2754		/* Subset bound, so copy in the laddr list from the old_inp */
2755		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2756			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2757			if (laddr == NULL) {
2758				/*
2759				 * Gak, what can we do? This assoc is really
2760				 * HOSED. We probably should send an abort
2761				 * here.
2762				 */
2763				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2764				continue;
2765			}
2766			SCTP_INCR_LADDR_COUNT();
2767			bzero(laddr, sizeof(*laddr));
2768			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2769			laddr->ifa = oladdr->ifa;
2770			atomic_add_int(&laddr->ifa->refcount, 1);
2771			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2772			    sctp_nxt_addr);
2773			new_inp->laddr_count++;
2774			if (oladdr == stcb->asoc.last_used_address) {
2775				stcb->asoc.last_used_address = laddr;
2776			}
2777		}
2778	}
2779	/*
2780	 * Now any running timers need to be adjusted since we really don't
2781	 * care if they are running or not just blast in the new_inp into
2782	 * all of them.
2783	 */
2784
2785	stcb->asoc.dack_timer.ep = (void *)new_inp;
2786	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2787	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2788	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2789	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2790	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2791	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2792	/* now what about the nets? */
2793	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2794		net->pmtu_timer.ep = (void *)new_inp;
2795		net->hb_timer.ep = (void *)new_inp;
2796		net->rxt_timer.ep = (void *)new_inp;
2797	}
2798	SCTP_INP_WUNLOCK(new_inp);
2799	SCTP_INP_WUNLOCK(old_inp);
2800}
2801
2802
2803
2804
2805/* sctp_ifap is used to bypass normal local address validation checks */
2806int
2807sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2808    struct sctp_ifa *sctp_ifap, struct thread *p)
2809{
2810	/* bind a ep to a socket address */
2811	struct sctppcbhead *head;
2812	struct sctp_inpcb *inp, *inp_tmp;
2813	struct inpcb *ip_inp;
2814	int port_reuse_active = 0;
2815	int bindall;
2816	uint16_t lport;
2817	int error;
2818	uint32_t vrf_id;
2819
2820	lport = 0;
2821	bindall = 1;
2822	inp = (struct sctp_inpcb *)so->so_pcb;
2823	ip_inp = (struct inpcb *)so->so_pcb;
2824#ifdef SCTP_DEBUG
2825	if (addr) {
2826		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2827		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2828		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2829		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2830	}
2831#endif
2832	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2833		/* already did a bind, subsequent binds NOT allowed ! */
2834		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2835		return (EINVAL);
2836	}
2837#ifdef INVARIANTS
2838	if (p == NULL)
2839		panic("null proc/thread");
2840#endif
2841	if (addr != NULL) {
2842		switch (addr->sa_family) {
2843#ifdef INET
2844		case AF_INET:
2845			{
2846				struct sockaddr_in *sin;
2847
2848				/* IPV6_V6ONLY socket? */
2849				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2850					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2851					return (EINVAL);
2852				}
2853				if (addr->sa_len != sizeof(*sin)) {
2854					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2855					return (EINVAL);
2856				}
2857				sin = (struct sockaddr_in *)addr;
2858				lport = sin->sin_port;
2859				/*
2860				 * For LOOPBACK the prison_local_ip4() call
2861				 * will transmute the ip address to the
2862				 * proper value.
2863				 */
2864				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2865					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2866					return (error);
2867				}
2868				if (sin->sin_addr.s_addr != INADDR_ANY) {
2869					bindall = 0;
2870				}
2871				break;
2872			}
2873#endif
2874#ifdef INET6
2875		case AF_INET6:
2876			{
2877				/*
2878				 * Only for pure IPv6 Address. (No IPv4
2879				 * Mapped!)
2880				 */
2881				struct sockaddr_in6 *sin6;
2882
2883				sin6 = (struct sockaddr_in6 *)addr;
2884
2885				if (addr->sa_len != sizeof(*sin6)) {
2886					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2887					return (EINVAL);
2888				}
2889				lport = sin6->sin6_port;
2890				/*
2891				 * For LOOPBACK the prison_local_ip6() call
2892				 * will transmute the ipv6 address to the
2893				 * proper value.
2894				 */
2895				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2896				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2897					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2898					return (error);
2899				}
2900				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2901					bindall = 0;
2902					/* KAME hack: embed scopeid */
2903					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2904						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2905						return (EINVAL);
2906					}
2907				}
2908				/* this must be cleared for ifa_ifwithaddr() */
2909				sin6->sin6_scope_id = 0;
2910				break;
2911			}
2912#endif
2913		default:
2914			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2915			return (EAFNOSUPPORT);
2916		}
2917	}
2918	SCTP_INP_INFO_WLOCK();
2919	SCTP_INP_WLOCK(inp);
2920	/* Setup a vrf_id to be the default for the non-bind-all case. */
2921	vrf_id = inp->def_vrf_id;
2922
2923	/* increase our count due to the unlock we do */
2924	SCTP_INP_INCR_REF(inp);
2925	if (lport) {
2926		/*
2927		 * Did the caller specify a port? if so we must see if an ep
2928		 * already has this one bound.
2929		 */
2930		/* got to be root to get at low ports */
2931		if (ntohs(lport) < IPPORT_RESERVED) {
2932			if (p && (error =
2933			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2934			    )) {
2935				SCTP_INP_DECR_REF(inp);
2936				SCTP_INP_WUNLOCK(inp);
2937				SCTP_INP_INFO_WUNLOCK();
2938				return (error);
2939			}
2940		}
2941		SCTP_INP_WUNLOCK(inp);
2942		if (bindall) {
2943			vrf_id = inp->def_vrf_id;
2944			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2945			if (inp_tmp != NULL) {
2946				/*
2947				 * lock guy returned and lower count note
2948				 * that we are not bound so inp_tmp should
2949				 * NEVER be inp. And it is this inp
2950				 * (inp_tmp) that gets the reference bump,
2951				 * so we must lower it.
2952				 */
2953				SCTP_INP_DECR_REF(inp_tmp);
2954				/* unlock info */
2955				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2956				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2957					/*
2958					 * Ok, must be one-2-one and
2959					 * allowing port re-use
2960					 */
2961					port_reuse_active = 1;
2962					goto continue_anyway;
2963				}
2964				SCTP_INP_DECR_REF(inp);
2965				SCTP_INP_INFO_WUNLOCK();
2966				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2967				return (EADDRINUSE);
2968			}
2969		} else {
2970			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2971			if (inp_tmp != NULL) {
2972				/*
2973				 * lock guy returned and lower count note
2974				 * that we are not bound so inp_tmp should
2975				 * NEVER be inp. And it is this inp
2976				 * (inp_tmp) that gets the reference bump,
2977				 * so we must lower it.
2978				 */
2979				SCTP_INP_DECR_REF(inp_tmp);
2980				/* unlock info */
2981				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2982				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2983					/*
2984					 * Ok, must be one-2-one and
2985					 * allowing port re-use
2986					 */
2987					port_reuse_active = 1;
2988					goto continue_anyway;
2989				}
2990				SCTP_INP_DECR_REF(inp);
2991				SCTP_INP_INFO_WUNLOCK();
2992				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2993				return (EADDRINUSE);
2994			}
2995		}
2996continue_anyway:
2997		SCTP_INP_WLOCK(inp);
2998		if (bindall) {
2999			/* verify that no lport is not used by a singleton */
3000			if ((port_reuse_active == 0) &&
3001			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
3002				/* Sorry someone already has this one bound */
3003				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
3004				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
3005					port_reuse_active = 1;
3006				} else {
3007					SCTP_INP_DECR_REF(inp);
3008					SCTP_INP_WUNLOCK(inp);
3009					SCTP_INP_INFO_WUNLOCK();
3010					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3011					return (EADDRINUSE);
3012				}
3013			}
3014		}
3015	} else {
3016		uint16_t first, last, candidate;
3017		uint16_t count;
3018		int done;
3019
3020		if (ip_inp->inp_flags & INP_HIGHPORT) {
3021			first = MODULE_GLOBAL(ipport_hifirstauto);
3022			last = MODULE_GLOBAL(ipport_hilastauto);
3023		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3024			if (p && (error =
3025			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3026			    )) {
3027				SCTP_INP_DECR_REF(inp);
3028				SCTP_INP_WUNLOCK(inp);
3029				SCTP_INP_INFO_WUNLOCK();
3030				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3031				return (error);
3032			}
3033			first = MODULE_GLOBAL(ipport_lowfirstauto);
3034			last = MODULE_GLOBAL(ipport_lowlastauto);
3035		} else {
3036			first = MODULE_GLOBAL(ipport_firstauto);
3037			last = MODULE_GLOBAL(ipport_lastauto);
3038		}
3039		if (first > last) {
3040			uint16_t temp;
3041
3042			temp = first;
3043			first = last;
3044			last = temp;
3045		}
3046		count = last - first + 1;	/* number of candidates */
3047		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3048
3049		done = 0;
3050		while (!done) {
3051			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3052				done = 1;
3053			}
3054			if (!done) {
3055				if (--count == 0) {
3056					SCTP_INP_DECR_REF(inp);
3057					SCTP_INP_WUNLOCK(inp);
3058					SCTP_INP_INFO_WUNLOCK();
3059					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3060					return (EADDRINUSE);
3061				}
3062				if (candidate == last)
3063					candidate = first;
3064				else
3065					candidate = candidate + 1;
3066			}
3067		}
3068		lport = htons(candidate);
3069	}
3070	SCTP_INP_DECR_REF(inp);
3071	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3072	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3073		/*
3074		 * this really should not happen. The guy did a non-blocking
3075		 * bind and then did a close at the same time.
3076		 */
3077		SCTP_INP_WUNLOCK(inp);
3078		SCTP_INP_INFO_WUNLOCK();
3079		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3080		return (EINVAL);
3081	}
3082	/* ok we look clear to give out this port, so lets setup the binding */
3083	if (bindall) {
3084		/* binding to all addresses, so just set in the proper flags */
3085		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3086		/* set the automatic addr changes from kernel flag */
3087		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3088			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3089			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3090		} else {
3091			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3092			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3093		}
3094		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3095			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3096		} else {
3097			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3098		}
3099		/*
3100		 * set the automatic mobility_base from kernel flag (by
3101		 * micchie)
3102		 */
3103		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3104			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3105			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3106		} else {
3107			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3108			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3109		}
3110		/*
3111		 * set the automatic mobility_fasthandoff from kernel flag
3112		 * (by micchie)
3113		 */
3114		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3115			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3116			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3117		} else {
3118			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3119			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3120		}
3121	} else {
3122		/*
3123		 * bind specific, make sure flags is off and add a new
3124		 * address structure to the sctp_addr_list inside the ep
3125		 * structure.
3126		 *
3127		 * We will need to allocate one and insert it at the head. The
3128		 * socketopt call can just insert new addresses in there as
3129		 * well. It will also have to do the embed scope kame hack
3130		 * too (before adding).
3131		 */
3132		struct sctp_ifa *ifa;
3133		union sctp_sockstore store;
3134
3135		memset(&store, 0, sizeof(store));
3136		switch (addr->sa_family) {
3137#ifdef INET
3138		case AF_INET:
3139			memcpy(&store.sin, addr, sizeof(struct sockaddr_in));
3140			store.sin.sin_port = 0;
3141			break;
3142#endif
3143#ifdef INET6
3144		case AF_INET6:
3145			memcpy(&store.sin6, addr, sizeof(struct sockaddr_in6));
3146			store.sin6.sin6_port = 0;
3147			break;
3148#endif
3149		default:
3150			break;
3151		}
3152		/*
3153		 * first find the interface with the bound address need to
3154		 * zero out the port to find the address! yuck! can't do
3155		 * this earlier since need port for sctp_pcb_findep()
3156		 */
3157		if (sctp_ifap != NULL) {
3158			ifa = sctp_ifap;
3159		} else {
3160			/*
3161			 * Note for BSD we hit here always other O/S's will
3162			 * pass things in via the sctp_ifap argument
3163			 * (Panda).
3164			 */
3165			ifa = sctp_find_ifa_by_addr(&store.sa,
3166			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3167		}
3168		if (ifa == NULL) {
3169			/* Can't find an interface with that address */
3170			SCTP_INP_WUNLOCK(inp);
3171			SCTP_INP_INFO_WUNLOCK();
3172			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3173			return (EADDRNOTAVAIL);
3174		}
3175#ifdef INET6
3176		if (addr->sa_family == AF_INET6) {
3177			/* GAK, more FIXME IFA lock? */
3178			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3179				/* Can't bind a non-existent addr. */
3180				SCTP_INP_WUNLOCK(inp);
3181				SCTP_INP_INFO_WUNLOCK();
3182				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3183				return (EINVAL);
3184			}
3185		}
3186#endif
3187		/* we're not bound all */
3188		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3189		/* allow bindx() to send ASCONF's for binding changes */
3190		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3191		/* clear automatic addr changes from kernel flag */
3192		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3193
3194		/* add this address to the endpoint list */
3195		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3196		if (error != 0) {
3197			SCTP_INP_WUNLOCK(inp);
3198			SCTP_INP_INFO_WUNLOCK();
3199			return (error);
3200		}
3201		inp->laddr_count++;
3202	}
3203	/* find the bucket */
3204	if (port_reuse_active) {
3205		/* Put it into tcp 1-2-1 hash */
3206		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3207		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3208	} else {
3209		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3210	}
3211	/* put it in the bucket */
3212	LIST_INSERT_HEAD(head, inp, sctp_hash);
3213	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3214	    (void *)head, ntohs(lport), port_reuse_active);
3215	/* set in the port */
3216	inp->sctp_lport = lport;
3217
3218	/* turn off just the unbound flag */
3219	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3220	SCTP_INP_WUNLOCK(inp);
3221	SCTP_INP_INFO_WUNLOCK();
3222	return (0);
3223}
3224
3225
3226static void
3227sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3228{
3229	struct sctp_iterator *it, *nit;
3230
3231	/*
3232	 * We enter with the only the ITERATOR_LOCK in place and a write
3233	 * lock on the inp_info stuff.
3234	 */
3235	it = sctp_it_ctl.cur_it;
3236	if (it && (it->vn != curvnet)) {
3237		/* Its not looking at our VNET */
3238		return;
3239	}
3240	if (it && (it->inp == inp)) {
3241		/*
3242		 * This is tricky and we hold the iterator lock, but when it
3243		 * returns and gets the lock (when we release it) the
3244		 * iterator will try to operate on inp. We need to stop that
3245		 * from happening. But of course the iterator has a
3246		 * reference on the stcb and inp. We can mark it and it will
3247		 * stop.
3248		 *
3249		 * If its a single iterator situation, we set the end iterator
3250		 * flag. Otherwise we set the iterator to go to the next
3251		 * inp.
3252		 *
3253		 */
3254		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3255			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3256		} else {
3257			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3258		}
3259	}
3260	/*
3261	 * Now go through and remove any single reference to our inp that
3262	 * may be still pending on the list
3263	 */
3264	SCTP_IPI_ITERATOR_WQ_LOCK();
3265	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3266		if (it->vn != curvnet) {
3267			continue;
3268		}
3269		if (it->inp == inp) {
3270			/* This one points to me is it inp specific? */
3271			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3272				/* Remove and free this one */
3273				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3274				    it, sctp_nxt_itr);
3275				if (it->function_atend != NULL) {
3276					(*it->function_atend) (it->pointer, it->val);
3277				}
3278				SCTP_FREE(it, SCTP_M_ITER);
3279			} else {
3280				it->inp = LIST_NEXT(it->inp, sctp_list);
3281				if (it->inp) {
3282					SCTP_INP_INCR_REF(it->inp);
3283				}
3284			}
3285			/*
3286			 * When its put in the refcnt is incremented so decr
3287			 * it
3288			 */
3289			SCTP_INP_DECR_REF(inp);
3290		}
3291	}
3292	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3293}
3294
3295/* release sctp_inpcb unbind the port */
3296void
3297sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3298{
3299	/*
3300	 * Here we free a endpoint. We must find it (if it is in the Hash
3301	 * table) and remove it from there. Then we must also find it in the
3302	 * overall list and remove it from there. After all removals are
3303	 * complete then any timer has to be stopped. Then start the actual
3304	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3305	 * all associations. d) finally the ep itself.
3306	 */
3307	struct sctp_tcb *asoc, *nasoc;
3308	struct sctp_laddr *laddr, *nladdr;
3309	struct inpcb *ip_pcb;
3310	struct socket *so;
3311	int being_refed = 0;
3312	struct sctp_queued_to_read *sq, *nsq;
3313	int cnt;
3314	sctp_sharedkey_t *shared_key, *nshared_key;
3315
3316
3317#ifdef SCTP_LOG_CLOSING
3318	sctp_log_closing(inp, NULL, 0);
3319#endif
3320	SCTP_ITERATOR_LOCK();
3321	/* mark any iterators on the list or being processed */
3322	sctp_iterator_inp_being_freed(inp);
3323	SCTP_ITERATOR_UNLOCK();
3324	so = inp->sctp_socket;
3325	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3326		/* been here before.. eeks.. get out of here */
3327		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3328#ifdef SCTP_LOG_CLOSING
3329		sctp_log_closing(inp, NULL, 1);
3330#endif
3331		return;
3332	}
3333	SCTP_ASOC_CREATE_LOCK(inp);
3334	SCTP_INP_INFO_WLOCK();
3335
3336	SCTP_INP_WLOCK(inp);
3337	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3338		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3339		/* socket is gone, so no more wakeups allowed */
3340		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3341		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3342		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3343
3344	}
3345	/* First time through we have the socket lock, after that no more. */
3346	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3347	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3348
3349	if (inp->control) {
3350		sctp_m_freem(inp->control);
3351		inp->control = NULL;
3352	}
3353	if (inp->pkt) {
3354		sctp_m_freem(inp->pkt);
3355		inp->pkt = NULL;
3356	}
3357	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3358					 * here but I will be nice :> (i.e.
3359					 * ip_pcb = ep;) */
3360	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3361		int cnt_in_sd;
3362
3363		cnt_in_sd = 0;
3364		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3365			SCTP_TCB_LOCK(asoc);
3366			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3367				/* Skip guys being freed */
3368				cnt_in_sd++;
3369				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3370					/*
3371					 * Special case - we did not start a
3372					 * kill timer on the asoc due to it
3373					 * was not closed. So go ahead and
3374					 * start it now.
3375					 */
3376					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3377					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3378				}
3379				SCTP_TCB_UNLOCK(asoc);
3380				continue;
3381			}
3382			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3383			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3384			    (asoc->asoc.total_output_queue_size == 0)) {
3385				/*
3386				 * If we have data in queue, we don't want
3387				 * to just free since the app may have done,
3388				 * send()/close or connect/send/close. And
3389				 * it wants the data to get across first.
3390				 */
3391				/* Just abandon things in the front states */
3392				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3393				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3394					cnt_in_sd++;
3395				}
3396				continue;
3397			}
3398			/* Disconnect the socket please */
3399			asoc->sctp_socket = NULL;
3400			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3401			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3402			    (asoc->asoc.control_pdapi) ||
3403			    (asoc->asoc.size_on_all_streams > 0) ||
3404			    (so && (so->so_rcv.sb_cc > 0))) {
3405				/* Left with Data unread */
3406				struct mbuf *op_err;
3407
3408				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3409				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3410				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3411				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3412				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3413				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3414					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3415				}
3416				if (sctp_free_assoc(inp, asoc,
3417				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3418					cnt_in_sd++;
3419				}
3420				continue;
3421			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3422				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3423			    (asoc->asoc.stream_queue_cnt == 0)) {
3424				if (asoc->asoc.locked_on_sending) {
3425					goto abort_anyway;
3426				}
3427				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3428				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3429					struct sctp_nets *netp;
3430
3431					/*
3432					 * there is nothing queued to send,
3433					 * so I send shutdown
3434					 */
3435					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3436					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3437						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3438					}
3439					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3440					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3441					sctp_stop_timers_for_shutdown(asoc);
3442					if (asoc->asoc.alternate) {
3443						netp = asoc->asoc.alternate;
3444					} else {
3445						netp = asoc->asoc.primary_destination;
3446					}
3447					sctp_send_shutdown(asoc, netp);
3448					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3449					    netp);
3450					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3451					    asoc->asoc.primary_destination);
3452					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3453				}
3454			} else {
3455				/* mark into shutdown pending */
3456				struct sctp_stream_queue_pending *sp;
3457
3458				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3459				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3460				    asoc->asoc.primary_destination);
3461				if (asoc->asoc.locked_on_sending) {
3462					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3463					    sctp_streamhead);
3464					if (sp == NULL) {
3465						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3466						    (void *)asoc->asoc.locked_on_sending,
3467						    asoc->asoc.locked_on_sending->stream_no);
3468					} else {
3469						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3470							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3471					}
3472				}
3473				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3474				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3475				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3476					struct mbuf *op_err;
3477
3478			abort_anyway:
3479					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3480					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3481					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3482					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3483					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3484					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3485						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3486					}
3487					if (sctp_free_assoc(inp, asoc,
3488					    SCTP_PCBFREE_NOFORCE,
3489					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3490						cnt_in_sd++;
3491					}
3492					continue;
3493				} else {
3494					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3495				}
3496			}
3497			cnt_in_sd++;
3498			SCTP_TCB_UNLOCK(asoc);
3499		}
3500		/* now is there some left in our SHUTDOWN state? */
3501		if (cnt_in_sd) {
3502#ifdef SCTP_LOG_CLOSING
3503			sctp_log_closing(inp, NULL, 2);
3504#endif
3505			inp->sctp_socket = NULL;
3506			SCTP_INP_WUNLOCK(inp);
3507			SCTP_ASOC_CREATE_UNLOCK(inp);
3508			SCTP_INP_INFO_WUNLOCK();
3509			return;
3510		}
3511	}
3512	inp->sctp_socket = NULL;
3513	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3514	    SCTP_PCB_FLAGS_UNBOUND) {
3515		/*
3516		 * ok, this guy has been bound. It's port is somewhere in
3517		 * the SCTP_BASE_INFO(hash table). Remove it!
3518		 */
3519		LIST_REMOVE(inp, sctp_hash);
3520		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3521	}
3522	/*
3523	 * If there is a timer running to kill us, forget it, since it may
3524	 * have a contest on the INP lock.. which would cause us to die ...
3525	 */
3526	cnt = 0;
3527	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3528		SCTP_TCB_LOCK(asoc);
3529		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3530			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3531				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3532				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3533			}
3534			cnt++;
3535			SCTP_TCB_UNLOCK(asoc);
3536			continue;
3537		}
3538		/* Free associations that are NOT killing us */
3539		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3540		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3541			struct mbuf *op_err;
3542
3543			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3544			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3545			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3546			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3547		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3548			cnt++;
3549			SCTP_TCB_UNLOCK(asoc);
3550			continue;
3551		}
3552		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3553		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3554			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3555		}
3556		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3557			cnt++;
3558		}
3559	}
3560	if (cnt) {
3561		/* Ok we have someone out there that will kill us */
3562		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3563#ifdef SCTP_LOG_CLOSING
3564		sctp_log_closing(inp, NULL, 3);
3565#endif
3566		SCTP_INP_WUNLOCK(inp);
3567		SCTP_ASOC_CREATE_UNLOCK(inp);
3568		SCTP_INP_INFO_WUNLOCK();
3569		return;
3570	}
3571	if (SCTP_INP_LOCK_CONTENDED(inp))
3572		being_refed++;
3573	if (SCTP_INP_READ_CONTENDED(inp))
3574		being_refed++;
3575	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3576		being_refed++;
3577
3578	if ((inp->refcount) ||
3579	    (being_refed) ||
3580	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3581		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3582#ifdef SCTP_LOG_CLOSING
3583		sctp_log_closing(inp, NULL, 4);
3584#endif
3585		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3586		SCTP_INP_WUNLOCK(inp);
3587		SCTP_ASOC_CREATE_UNLOCK(inp);
3588		SCTP_INP_INFO_WUNLOCK();
3589		return;
3590	}
3591	inp->sctp_ep.signature_change.type = 0;
3592	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3593	/*
3594	 * Remove it from the list .. last thing we need a lock for.
3595	 */
3596	LIST_REMOVE(inp, sctp_list);
3597	SCTP_INP_WUNLOCK(inp);
3598	SCTP_ASOC_CREATE_UNLOCK(inp);
3599	SCTP_INP_INFO_WUNLOCK();
3600	/*
3601	 * Now we release all locks. Since this INP cannot be found anymore
3602	 * except possibly by the kill timer that might be running. We call
3603	 * the drain function here. It should hit the case were it sees the
3604	 * ACTIVE flag cleared and exit out freeing us to proceed and
3605	 * destroy everything.
3606	 */
3607	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3608		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3609	} else {
3610		/* Probably un-needed */
3611		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3612	}
3613
3614#ifdef SCTP_LOG_CLOSING
3615	sctp_log_closing(inp, NULL, 5);
3616#endif
3617
3618
3619	if ((inp->sctp_asocidhash) != NULL) {
3620		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3621		inp->sctp_asocidhash = NULL;
3622	}
3623	/* sa_ignore FREED_MEMORY */
3624	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3625		/* Its only abandoned if it had data left */
3626		if (sq->length)
3627			SCTP_STAT_INCR(sctps_left_abandon);
3628
3629		TAILQ_REMOVE(&inp->read_queue, sq, next);
3630		sctp_free_remote_addr(sq->whoFrom);
3631		if (so)
3632			so->so_rcv.sb_cc -= sq->length;
3633		if (sq->data) {
3634			sctp_m_freem(sq->data);
3635			sq->data = NULL;
3636		}
3637		/*
3638		 * no need to free the net count, since at this point all
3639		 * assoc's are gone.
3640		 */
3641		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3642		SCTP_DECR_READQ_COUNT();
3643	}
3644	/* Now the sctp_pcb things */
3645	/*
3646	 * free each asoc if it is not already closed/free. we can't use the
3647	 * macro here since le_next will get freed as part of the
3648	 * sctp_free_assoc() call.
3649	 */
3650	if (so) {
3651#ifdef IPSEC
3652		ipsec_delete_pcbpolicy(ip_pcb);
3653#endif				/* IPSEC */
3654
3655		/* Unlocks not needed since the socket is gone now */
3656	}
3657	if (ip_pcb->inp_options) {
3658		(void)sctp_m_free(ip_pcb->inp_options);
3659		ip_pcb->inp_options = 0;
3660	}
3661#ifdef INET6
3662	if (ip_pcb->inp_vflag & INP_IPV6) {
3663		struct in6pcb *in6p;
3664
3665		in6p = (struct in6pcb *)inp;
3666		ip6_freepcbopts(in6p->in6p_outputopts);
3667	}
3668#endif				/* INET6 */
3669	ip_pcb->inp_vflag = 0;
3670	/* free up authentication fields */
3671	if (inp->sctp_ep.local_auth_chunks != NULL)
3672		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3673	if (inp->sctp_ep.local_hmacs != NULL)
3674		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3675
3676	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3677		LIST_REMOVE(shared_key, next);
3678		sctp_free_sharedkey(shared_key);
3679		/* sa_ignore FREED_MEMORY */
3680	}
3681
3682	/*
3683	 * if we have an address list the following will free the list of
3684	 * ifaddr's that are set into this ep. Again macro limitations here,
3685	 * since the LIST_FOREACH could be a bad idea.
3686	 */
3687	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3688		sctp_remove_laddr(laddr);
3689	}
3690
3691#ifdef SCTP_TRACK_FREED_ASOCS
3692	/* TEMP CODE */
3693	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3694		LIST_REMOVE(asoc, sctp_tcblist);
3695		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3696		SCTP_DECR_ASOC_COUNT();
3697	}
3698	/* *** END TEMP CODE *** */
3699#endif
3700	/* Now lets see about freeing the EP hash table. */
3701	if (inp->sctp_tcbhash != NULL) {
3702		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3703		inp->sctp_tcbhash = NULL;
3704	}
3705	/* Now we must put the ep memory back into the zone pool */
3706	crfree(inp->ip_inp.inp.inp_cred);
3707	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3708	SCTP_INP_LOCK_DESTROY(inp);
3709	SCTP_INP_READ_DESTROY(inp);
3710	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3711	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3712	SCTP_DECR_EP_COUNT();
3713}
3714
3715
3716struct sctp_nets *
3717sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3718{
3719	struct sctp_nets *net;
3720
3721	/* locate the address */
3722	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3723		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3724			return (net);
3725	}
3726	return (NULL);
3727}
3728
3729
3730int
3731sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3732{
3733	struct sctp_ifa *sctp_ifa;
3734
3735	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3736	if (sctp_ifa) {
3737		return (1);
3738	} else {
3739		return (0);
3740	}
3741}
3742
3743/*
3744 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3745 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3746 * stats of stuff.
3747 */
3748int
3749sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3750    struct sctp_nets **netp, int set_scope, int from)
3751{
3752	/*
3753	 * The following is redundant to the same lines in the
3754	 * sctp_aloc_assoc() but is needed since others call the add address
3755	 * function
3756	 */
3757	struct sctp_nets *net, *netfirst;
3758	int addr_inscope;
3759
3760	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3761	    from);
3762	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3763
3764	netfirst = sctp_findnet(stcb, newaddr);
3765	if (netfirst) {
3766		/*
3767		 * Lie and return ok, we don't want to make the association
3768		 * go away for this behavior. It will happen in the TCP
3769		 * model in a connected socket. It does not reach the hash
3770		 * table until after the association is built so it can't be
3771		 * found. Mark as reachable, since the initial creation will
3772		 * have been cleared and the NOT_IN_ASSOC flag will have
3773		 * been added... and we don't want to end up removing it
3774		 * back out.
3775		 */
3776		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3777			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3778			    SCTP_ADDR_UNCONFIRMED);
3779		} else {
3780			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3781		}
3782
3783		return (0);
3784	}
3785	addr_inscope = 1;
3786	switch (newaddr->sa_family) {
3787#ifdef INET
3788	case AF_INET:
3789		{
3790			struct sockaddr_in *sin;
3791
3792			sin = (struct sockaddr_in *)newaddr;
3793			if (sin->sin_addr.s_addr == 0) {
3794				/* Invalid address */
3795				return (-1);
3796			}
3797			/* zero out the bzero area */
3798			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3799
3800			/* assure len is set */
3801			sin->sin_len = sizeof(struct sockaddr_in);
3802			if (set_scope) {
3803#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3804				stcb->asoc.scope.ipv4_local_scope = 1;
3805#else
3806				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3807					stcb->asoc.scope.ipv4_local_scope = 1;
3808				}
3809#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3810			} else {
3811				/* Validate the address is in scope */
3812				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3813				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3814					addr_inscope = 0;
3815				}
3816			}
3817			break;
3818		}
3819#endif
3820#ifdef INET6
3821	case AF_INET6:
3822		{
3823			struct sockaddr_in6 *sin6;
3824
3825			sin6 = (struct sockaddr_in6 *)newaddr;
3826			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3827				/* Invalid address */
3828				return (-1);
3829			}
3830			/* assure len is set */
3831			sin6->sin6_len = sizeof(struct sockaddr_in6);
3832			if (set_scope) {
3833				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3834					stcb->asoc.scope.loopback_scope = 1;
3835					stcb->asoc.scope.local_scope = 0;
3836					stcb->asoc.scope.ipv4_local_scope = 1;
3837					stcb->asoc.scope.site_scope = 1;
3838				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3839					/*
3840					 * If the new destination is a
3841					 * LINK_LOCAL we must have common
3842					 * site scope. Don't set the local
3843					 * scope since we may not share all
3844					 * links, only loopback can do this.
3845					 * Links on the local network would
3846					 * also be on our private network
3847					 * for v4 too.
3848					 */
3849					stcb->asoc.scope.ipv4_local_scope = 1;
3850					stcb->asoc.scope.site_scope = 1;
3851				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3852					/*
3853					 * If the new destination is
3854					 * SITE_LOCAL then we must have site
3855					 * scope in common.
3856					 */
3857					stcb->asoc.scope.site_scope = 1;
3858				}
3859			} else {
3860				/* Validate the address is in scope */
3861				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3862				    (stcb->asoc.scope.loopback_scope == 0)) {
3863					addr_inscope = 0;
3864				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3865				    (stcb->asoc.scope.local_scope == 0)) {
3866					addr_inscope = 0;
3867				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3868				    (stcb->asoc.scope.site_scope == 0)) {
3869					addr_inscope = 0;
3870				}
3871			}
3872			break;
3873		}
3874#endif
3875	default:
3876		/* not supported family type */
3877		return (-1);
3878	}
3879	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3880	if (net == NULL) {
3881		return (-1);
3882	}
3883	SCTP_INCR_RADDR_COUNT();
3884	bzero(net, sizeof(struct sctp_nets));
3885	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3886	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3887	switch (newaddr->sa_family) {
3888#ifdef INET
3889	case AF_INET:
3890		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3891		break;
3892#endif
3893#ifdef INET6
3894	case AF_INET6:
3895		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3896		break;
3897#endif
3898	default:
3899		break;
3900	}
3901	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3902	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3903		stcb->asoc.scope.loopback_scope = 1;
3904		stcb->asoc.scope.ipv4_local_scope = 1;
3905		stcb->asoc.scope.local_scope = 0;
3906		stcb->asoc.scope.site_scope = 1;
3907		addr_inscope = 1;
3908	}
3909	net->failure_threshold = stcb->asoc.def_net_failure;
3910	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3911	if (addr_inscope == 0) {
3912		net->dest_state = (SCTP_ADDR_REACHABLE |
3913		    SCTP_ADDR_OUT_OF_SCOPE);
3914	} else {
3915		if (from == SCTP_ADDR_IS_CONFIRMED)
3916			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3917			net->dest_state = SCTP_ADDR_REACHABLE;
3918		else
3919			net->dest_state = SCTP_ADDR_REACHABLE |
3920			    SCTP_ADDR_UNCONFIRMED;
3921	}
3922	/*
3923	 * We set this to 0, the timer code knows that this means its an
3924	 * initial value
3925	 */
3926	net->rto_needed = 1;
3927	net->RTO = 0;
3928	net->RTO_measured = 0;
3929	stcb->asoc.numnets++;
3930	net->ref_count = 1;
3931	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3932	net->port = stcb->asoc.port;
3933	net->dscp = stcb->asoc.default_dscp;
3934#ifdef INET6
3935	net->flowlabel = stcb->asoc.default_flowlabel;
3936#endif
3937	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3938		net->dest_state |= SCTP_ADDR_NOHB;
3939	} else {
3940		net->dest_state &= ~SCTP_ADDR_NOHB;
3941	}
3942	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3943		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3944	} else {
3945		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3946	}
3947	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3948	/* Init the timer structure */
3949	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3950	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3951	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3952
3953	/* Now generate a route for this guy */
3954#ifdef INET6
3955	/* KAME hack: embed scopeid */
3956	if (newaddr->sa_family == AF_INET6) {
3957		struct sockaddr_in6 *sin6;
3958
3959		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3960		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3961		sin6->sin6_scope_id = 0;
3962	}
3963#endif
3964	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3965
3966	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3967		/* Get source address */
3968		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3969		    stcb,
3970		    (sctp_route_t *) & net->ro,
3971		    net,
3972		    0,
3973		    stcb->asoc.vrf_id);
3974		if (net->ro._s_addr != NULL) {
3975			net->src_addr_selected = 1;
3976			/* Now get the interface MTU */
3977			if (net->ro._s_addr->ifn_p != NULL) {
3978				net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3979			}
3980		} else {
3981			net->src_addr_selected = 0;
3982		}
3983		if (net->mtu > 0) {
3984			uint32_t rmtu;
3985
3986			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3987			if (rmtu == 0) {
3988				/*
3989				 * Start things off to match mtu of
3990				 * interface please.
3991				 */
3992				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3993				    net->ro.ro_rt, net->mtu);
3994			} else {
3995				/*
3996				 * we take the route mtu over the interface,
3997				 * since the route may be leading out the
3998				 * loopback, or a different interface.
3999				 */
4000				net->mtu = rmtu;
4001			}
4002		}
4003	} else {
4004		net->src_addr_selected = 0;
4005	}
4006	if (net->mtu == 0) {
4007		switch (newaddr->sa_family) {
4008#ifdef INET
4009		case AF_INET:
4010			net->mtu = SCTP_DEFAULT_MTU;
4011			break;
4012#endif
4013#ifdef INET6
4014		case AF_INET6:
4015			net->mtu = 1280;
4016			break;
4017#endif
4018		default:
4019			break;
4020		}
4021	}
4022#if defined(INET) || defined(INET6)
4023	if (net->port) {
4024		net->mtu -= (uint32_t) sizeof(struct udphdr);
4025	}
4026#endif
4027	if (from == SCTP_ALLOC_ASOC) {
4028		stcb->asoc.smallest_mtu = net->mtu;
4029	}
4030	if (stcb->asoc.smallest_mtu > net->mtu) {
4031		stcb->asoc.smallest_mtu = net->mtu;
4032	}
4033#ifdef INET6
4034	if (newaddr->sa_family == AF_INET6) {
4035		struct sockaddr_in6 *sin6;
4036
4037		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4038		(void)sa6_recoverscope(sin6);
4039	}
4040#endif
4041
4042	/* JRS - Use the congestion control given in the CC module */
4043	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4044		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4045
4046	/*
4047	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4048	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4049	 */
4050	net->find_pseudo_cumack = 1;
4051	net->find_rtx_pseudo_cumack = 1;
4052	/* Choose an initial flowid. */
4053	net->flowid = stcb->asoc.my_vtag ^
4054	    ntohs(stcb->rport) ^
4055	    ntohs(stcb->sctp_ep->sctp_lport);
4056#ifdef INVARIANTS
4057	net->flowidset = 1;
4058#endif
4059	if (netp) {
4060		*netp = net;
4061	}
4062	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4063	if (net->ro.ro_rt == NULL) {
4064		/* Since we have no route put it at the back */
4065		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4066	} else if (netfirst == NULL) {
4067		/* We are the first one in the pool. */
4068		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4069	} else if (netfirst->ro.ro_rt == NULL) {
4070		/*
4071		 * First one has NO route. Place this one ahead of the first
4072		 * one.
4073		 */
4074		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4075	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4076		/*
4077		 * This one has a different interface than the one at the
4078		 * top of the list. Place it ahead.
4079		 */
4080		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4081	} else {
4082		/*
4083		 * Ok we have the same interface as the first one. Move
4084		 * forward until we find either a) one with a NULL route...
4085		 * insert ahead of that b) one with a different ifp.. insert
4086		 * after that. c) end of the list.. insert at the tail.
4087		 */
4088		struct sctp_nets *netlook;
4089
4090		do {
4091			netlook = TAILQ_NEXT(netfirst, sctp_next);
4092			if (netlook == NULL) {
4093				/* End of the list */
4094				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4095				break;
4096			} else if (netlook->ro.ro_rt == NULL) {
4097				/* next one has NO route */
4098				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4099				break;
4100			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4101				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4102				    net, sctp_next);
4103				break;
4104			}
4105			/* Shift forward */
4106			netfirst = netlook;
4107		} while (netlook != NULL);
4108	}
4109
4110	/* got to have a primary set */
4111	if (stcb->asoc.primary_destination == 0) {
4112		stcb->asoc.primary_destination = net;
4113	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4114		    (net->ro.ro_rt) &&
4115	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4116		/* No route to current primary adopt new primary */
4117		stcb->asoc.primary_destination = net;
4118	}
4119	/* Validate primary is first */
4120	net = TAILQ_FIRST(&stcb->asoc.nets);
4121	if ((net != stcb->asoc.primary_destination) &&
4122	    (stcb->asoc.primary_destination)) {
4123		/*
4124		 * first one on the list is NOT the primary sctp_cmpaddr()
4125		 * is much more efficient if the primary is the first on the
4126		 * list, make it so.
4127		 */
4128		TAILQ_REMOVE(&stcb->asoc.nets,
4129		    stcb->asoc.primary_destination, sctp_next);
4130		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4131		    stcb->asoc.primary_destination, sctp_next);
4132	}
4133	return (0);
4134}
4135
4136
4137static uint32_t
4138sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4139{
4140	uint32_t id;
4141	struct sctpasochead *head;
4142	struct sctp_tcb *lstcb;
4143
4144	SCTP_INP_WLOCK(inp);
4145try_again:
4146	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4147		/* TSNH */
4148		SCTP_INP_WUNLOCK(inp);
4149		return (0);
4150	}
4151	/*
4152	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4153	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4154	 */
4155	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4156		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4157	}
4158	id = inp->sctp_associd_counter;
4159	inp->sctp_associd_counter++;
4160	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4161	if (lstcb) {
4162		goto try_again;
4163	}
4164	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4165	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4166	stcb->asoc.in_asocid_hash = 1;
4167	SCTP_INP_WUNLOCK(inp);
4168	return id;
4169}
4170
4171/*
4172 * allocate an association and add it to the endpoint. The caller must be
4173 * careful to add all additional addresses once they are know right away or
4174 * else the assoc will be may experience a blackout scenario.
4175 */
4176struct sctp_tcb *
4177sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4178    int *error, uint32_t override_tag, uint32_t vrf_id,
4179    struct thread *p
4180)
4181{
4182	/* note the p argument is only valid in unbound sockets */
4183
4184	struct sctp_tcb *stcb;
4185	struct sctp_association *asoc;
4186	struct sctpasochead *head;
4187	uint16_t rport;
4188	int err;
4189
4190	/*
4191	 * Assumption made here: Caller has done a
4192	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4193	 * address does not exist already.
4194	 */
4195	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4196		/* Hit max assoc, sorry no more */
4197		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4198		*error = ENOBUFS;
4199		return (NULL);
4200	}
4201	if (firstaddr == NULL) {
4202		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4203		*error = EINVAL;
4204		return (NULL);
4205	}
4206	SCTP_INP_RLOCK(inp);
4207	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4208	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4209	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4210		/*
4211		 * If its in the TCP pool, its NOT allowed to create an
4212		 * association. The parent listener needs to call
4213		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4214		 * off, or connected one does this.. its an error.
4215		 */
4216		SCTP_INP_RUNLOCK(inp);
4217		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4218		*error = EINVAL;
4219		return (NULL);
4220	}
4221	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4222	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4223		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4224		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4225			SCTP_INP_RUNLOCK(inp);
4226			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4227			*error = EINVAL;
4228			return (NULL);
4229		}
4230	}
4231	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4232#ifdef SCTP_DEBUG
4233	if (firstaddr) {
4234		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4235		switch (firstaddr->sa_family) {
4236#ifdef INET
4237		case AF_INET:
4238			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4239			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4240			break;
4241#endif
4242#ifdef INET6
4243		case AF_INET6:
4244			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4245			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4246			break;
4247#endif
4248		default:
4249			break;
4250		}
4251	} else {
4252		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4253	}
4254#endif				/* SCTP_DEBUG */
4255	switch (firstaddr->sa_family) {
4256#ifdef INET
4257	case AF_INET:
4258		{
4259			struct sockaddr_in *sin;
4260
4261			sin = (struct sockaddr_in *)firstaddr;
4262			if ((ntohs(sin->sin_port) == 0) ||
4263			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4264			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4265			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4266				/* Invalid address */
4267				SCTP_INP_RUNLOCK(inp);
4268				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4269				*error = EINVAL;
4270				return (NULL);
4271			}
4272			rport = sin->sin_port;
4273			break;
4274		}
4275#endif
4276#ifdef INET6
4277	case AF_INET6:
4278		{
4279			struct sockaddr_in6 *sin6;
4280
4281			sin6 = (struct sockaddr_in6 *)firstaddr;
4282			if ((ntohs(sin6->sin6_port) == 0) ||
4283			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4284			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4285				/* Invalid address */
4286				SCTP_INP_RUNLOCK(inp);
4287				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4288				*error = EINVAL;
4289				return (NULL);
4290			}
4291			rport = sin6->sin6_port;
4292			break;
4293		}
4294#endif
4295	default:
4296		/* not supported family type */
4297		SCTP_INP_RUNLOCK(inp);
4298		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4299		*error = EINVAL;
4300		return (NULL);
4301	}
4302	SCTP_INP_RUNLOCK(inp);
4303	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4304		/*
4305		 * If you have not performed a bind, then we need to do the
4306		 * ephemeral bind for you.
4307		 */
4308		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4309		    (struct sockaddr *)NULL,
4310		    (struct sctp_ifa *)NULL,
4311		    p
4312		    ))) {
4313			/* bind error, probably perm */
4314			*error = err;
4315			return (NULL);
4316		}
4317	}
4318	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4319	if (stcb == NULL) {
4320		/* out of memory? */
4321		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4322		*error = ENOMEM;
4323		return (NULL);
4324	}
4325	SCTP_INCR_ASOC_COUNT();
4326
4327	bzero(stcb, sizeof(*stcb));
4328	asoc = &stcb->asoc;
4329
4330	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4331	SCTP_TCB_LOCK_INIT(stcb);
4332	SCTP_TCB_SEND_LOCK_INIT(stcb);
4333	stcb->rport = rport;
4334	/* setup back pointer's */
4335	stcb->sctp_ep = inp;
4336	stcb->sctp_socket = inp->sctp_socket;
4337	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4338		/* failed */
4339		SCTP_TCB_LOCK_DESTROY(stcb);
4340		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4341		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4342		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4343		SCTP_DECR_ASOC_COUNT();
4344		*error = err;
4345		return (NULL);
4346	}
4347	/* and the port */
4348	SCTP_INP_INFO_WLOCK();
4349	SCTP_INP_WLOCK(inp);
4350	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4351		/* inpcb freed while alloc going on */
4352		SCTP_TCB_LOCK_DESTROY(stcb);
4353		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4354		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4355		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4356		SCTP_INP_WUNLOCK(inp);
4357		SCTP_INP_INFO_WUNLOCK();
4358		SCTP_DECR_ASOC_COUNT();
4359		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4360		*error = EINVAL;
4361		return (NULL);
4362	}
4363	SCTP_TCB_LOCK(stcb);
4364
4365	/* now that my_vtag is set, add it to the hash */
4366	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4367	/* put it in the bucket in the vtag hash of assoc's for the system */
4368	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4369	SCTP_INP_INFO_WUNLOCK();
4370
4371	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4372		/* failure.. memory error? */
4373		if (asoc->strmout) {
4374			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4375			asoc->strmout = NULL;
4376		}
4377		if (asoc->mapping_array) {
4378			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4379			asoc->mapping_array = NULL;
4380		}
4381		if (asoc->nr_mapping_array) {
4382			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4383			asoc->nr_mapping_array = NULL;
4384		}
4385		SCTP_DECR_ASOC_COUNT();
4386		SCTP_TCB_UNLOCK(stcb);
4387		SCTP_TCB_LOCK_DESTROY(stcb);
4388		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4389		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4390		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4391		SCTP_INP_WUNLOCK(inp);
4392		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4393		*error = ENOBUFS;
4394		return (NULL);
4395	}
4396	/* Init all the timers */
4397	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4398	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4399	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4400	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4401	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4402	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4403	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4404
4405	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4406	/* now file the port under the hash as well */
4407	if (inp->sctp_tcbhash != NULL) {
4408		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4409		    inp->sctp_hashmark)];
4410		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4411	}
4412	SCTP_INP_WUNLOCK(inp);
4413	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4414	return (stcb);
4415}
4416
4417
4418void
4419sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4420{
4421	struct sctp_association *asoc;
4422
4423	asoc = &stcb->asoc;
4424	asoc->numnets--;
4425	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4426	if (net == asoc->primary_destination) {
4427		/* Reset primary */
4428		struct sctp_nets *lnet;
4429
4430		lnet = TAILQ_FIRST(&asoc->nets);
4431		/*
4432		 * Mobility adaptation Ideally, if deleted destination is
4433		 * the primary, it becomes a fast retransmission trigger by
4434		 * the subsequent SET PRIMARY. (by micchie)
4435		 */
4436		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4437		    SCTP_MOBILITY_BASE) ||
4438		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4439		    SCTP_MOBILITY_FASTHANDOFF)) {
4440			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4441			if (asoc->deleted_primary != NULL) {
4442				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4443				goto out;
4444			}
4445			asoc->deleted_primary = net;
4446			atomic_add_int(&net->ref_count, 1);
4447			memset(&net->lastsa, 0, sizeof(net->lastsa));
4448			memset(&net->lastsv, 0, sizeof(net->lastsv));
4449			sctp_mobility_feature_on(stcb->sctp_ep,
4450			    SCTP_MOBILITY_PRIM_DELETED);
4451			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4452			    stcb->sctp_ep, stcb, NULL);
4453		}
4454out:
4455		/* Try to find a confirmed primary */
4456		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4457	}
4458	if (net == asoc->last_data_chunk_from) {
4459		/* Reset primary */
4460		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4461	}
4462	if (net == asoc->last_control_chunk_from) {
4463		/* Clear net */
4464		asoc->last_control_chunk_from = NULL;
4465	}
4466	if (net == stcb->asoc.alternate) {
4467		sctp_free_remote_addr(stcb->asoc.alternate);
4468		stcb->asoc.alternate = NULL;
4469	}
4470	sctp_free_remote_addr(net);
4471}
4472
4473/*
4474 * remove a remote endpoint address from an association, it will fail if the
4475 * address does not exist.
4476 */
4477int
4478sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4479{
4480	/*
4481	 * Here we need to remove a remote address. This is quite simple, we
4482	 * first find it in the list of address for the association
4483	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4484	 * on that item. Note we do not allow it to be removed if there are
4485	 * no other addresses.
4486	 */
4487	struct sctp_association *asoc;
4488	struct sctp_nets *net, *nnet;
4489
4490	asoc = &stcb->asoc;
4491
4492	/* locate the address */
4493	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4494		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4495			continue;
4496		}
4497		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4498		    remaddr)) {
4499			/* we found the guy */
4500			if (asoc->numnets < 2) {
4501				/* Must have at LEAST two remote addresses */
4502				return (-1);
4503			} else {
4504				sctp_remove_net(stcb, net);
4505				return (0);
4506			}
4507		}
4508	}
4509	/* not found. */
4510	return (-2);
4511}
4512
4513void
4514sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4515{
4516	struct sctpvtaghead *chain;
4517	struct sctp_tagblock *twait_block;
4518	int found = 0;
4519	int i;
4520
4521	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4522	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4523		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4524			if ((twait_block->vtag_block[i].v_tag == tag) &&
4525			    (twait_block->vtag_block[i].lport == lport) &&
4526			    (twait_block->vtag_block[i].rport == rport)) {
4527				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4528				twait_block->vtag_block[i].v_tag = 0;
4529				twait_block->vtag_block[i].lport = 0;
4530				twait_block->vtag_block[i].rport = 0;
4531				found = 1;
4532				break;
4533			}
4534		}
4535		if (found)
4536			break;
4537	}
4538}
4539
4540int
4541sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4542{
4543	struct sctpvtaghead *chain;
4544	struct sctp_tagblock *twait_block;
4545	int found = 0;
4546	int i;
4547
4548	SCTP_INP_INFO_WLOCK();
4549	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4550	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4551		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4552			if ((twait_block->vtag_block[i].v_tag == tag) &&
4553			    (twait_block->vtag_block[i].lport == lport) &&
4554			    (twait_block->vtag_block[i].rport == rport)) {
4555				found = 1;
4556				break;
4557			}
4558		}
4559		if (found)
4560			break;
4561	}
4562	SCTP_INP_INFO_WUNLOCK();
4563	return (found);
4564}
4565
4566
4567void
4568sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4569{
4570	struct sctpvtaghead *chain;
4571	struct sctp_tagblock *twait_block;
4572	struct timeval now;
4573	int set, i;
4574
4575	if (time == 0) {
4576		/* Its disabled */
4577		return;
4578	}
4579	(void)SCTP_GETTIME_TIMEVAL(&now);
4580	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4581	set = 0;
4582	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4583		/* Block(s) present, lets find space, and expire on the fly */
4584		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4585			if ((twait_block->vtag_block[i].v_tag == 0) &&
4586			    !set) {
4587				twait_block->vtag_block[i].tv_sec_at_expire =
4588				    now.tv_sec + time;
4589				twait_block->vtag_block[i].v_tag = tag;
4590				twait_block->vtag_block[i].lport = lport;
4591				twait_block->vtag_block[i].rport = rport;
4592				set = 1;
4593			} else if ((twait_block->vtag_block[i].v_tag) &&
4594			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4595				/* Audit expires this guy */
4596				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4597				twait_block->vtag_block[i].v_tag = 0;
4598				twait_block->vtag_block[i].lport = 0;
4599				twait_block->vtag_block[i].rport = 0;
4600				if (set == 0) {
4601					/* Reuse it for my new tag */
4602					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4603					twait_block->vtag_block[i].v_tag = tag;
4604					twait_block->vtag_block[i].lport = lport;
4605					twait_block->vtag_block[i].rport = rport;
4606					set = 1;
4607				}
4608			}
4609		}
4610		if (set) {
4611			/*
4612			 * We only do up to the block where we can place our
4613			 * tag for audits
4614			 */
4615			break;
4616		}
4617	}
4618	/* Need to add a new block to chain */
4619	if (!set) {
4620		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4621		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4622		if (twait_block == NULL) {
4623#ifdef INVARIANTS
4624			panic("Can not alloc tagblock");
4625#endif
4626			return;
4627		}
4628		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4629		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4630		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4631		twait_block->vtag_block[0].v_tag = tag;
4632		twait_block->vtag_block[0].lport = lport;
4633		twait_block->vtag_block[0].rport = rport;
4634	}
4635}
4636
4637
4638
4639/*-
4640 * Free the association after un-hashing the remote port. This
4641 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4642 * expect that the input to this function IS a locked TCB.
4643 * It will return 0, if it did NOT destroy the association (instead
4644 * it unlocks it. It will return NON-zero if it either destroyed the
4645 * association OR the association is already destroyed.
4646 */
4647int
4648sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4649{
4650	int i;
4651	struct sctp_association *asoc;
4652	struct sctp_nets *net, *nnet;
4653	struct sctp_laddr *laddr, *naddr;
4654	struct sctp_tmit_chunk *chk, *nchk;
4655	struct sctp_asconf_addr *aparam, *naparam;
4656	struct sctp_asconf_ack *aack, *naack;
4657	struct sctp_stream_reset_list *strrst, *nstrrst;
4658	struct sctp_queued_to_read *sq, *nsq;
4659	struct sctp_stream_queue_pending *sp, *nsp;
4660	sctp_sharedkey_t *shared_key, *nshared_key;
4661	struct socket *so;
4662
4663	/* first, lets purge the entry from the hash table. */
4664
4665#ifdef SCTP_LOG_CLOSING
4666	sctp_log_closing(inp, stcb, 6);
4667#endif
4668	if (stcb->asoc.state == 0) {
4669#ifdef SCTP_LOG_CLOSING
4670		sctp_log_closing(inp, NULL, 7);
4671#endif
4672		/* there is no asoc, really TSNH :-0 */
4673		return (1);
4674	}
4675	if (stcb->asoc.alternate) {
4676		sctp_free_remote_addr(stcb->asoc.alternate);
4677		stcb->asoc.alternate = NULL;
4678	}
4679	/* TEMP CODE */
4680	if (stcb->freed_from_where == 0) {
4681		/* Only record the first place free happened from */
4682		stcb->freed_from_where = from_location;
4683	}
4684	/* TEMP CODE */
4685
4686	asoc = &stcb->asoc;
4687	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4688	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4689		/* nothing around */
4690		so = NULL;
4691	else
4692		so = inp->sctp_socket;
4693
4694	/*
4695	 * We used timer based freeing if a reader or writer is in the way.
4696	 * So we first check if we are actually being called from a timer,
4697	 * if so we abort early if a reader or writer is still in the way.
4698	 */
4699	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4700	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4701		/*
4702		 * is it the timer driving us? if so are the reader/writers
4703		 * gone?
4704		 */
4705		if (stcb->asoc.refcnt) {
4706			/* nope, reader or writer in the way */
4707			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4708			/* no asoc destroyed */
4709			SCTP_TCB_UNLOCK(stcb);
4710#ifdef SCTP_LOG_CLOSING
4711			sctp_log_closing(inp, stcb, 8);
4712#endif
4713			return (0);
4714		}
4715	}
4716	/* now clean up any other timers */
4717	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4718	asoc->dack_timer.self = NULL;
4719	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4720	/*-
4721	 * For stream reset we don't blast this unless
4722	 * it is a str-reset timer, it might be the
4723	 * free-asoc timer which we DON'T want to
4724	 * disturb.
4725	 */
4726	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4727		asoc->strreset_timer.self = NULL;
4728	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4729	asoc->asconf_timer.self = NULL;
4730	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4731	asoc->autoclose_timer.self = NULL;
4732	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4733	asoc->shut_guard_timer.self = NULL;
4734	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4735	asoc->delayed_event_timer.self = NULL;
4736	/* Mobility adaptation */
4737	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4738	asoc->delete_prim_timer.self = NULL;
4739	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4740		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4741		net->rxt_timer.self = NULL;
4742		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4743		net->pmtu_timer.self = NULL;
4744		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4745		net->hb_timer.self = NULL;
4746	}
4747	/* Now the read queue needs to be cleaned up (only once) */
4748	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4749		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4750		SCTP_INP_READ_LOCK(inp);
4751		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4752			if (sq->stcb == stcb) {
4753				sq->do_not_ref_stcb = 1;
4754				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4755				/*
4756				 * If there is no end, there never will be
4757				 * now.
4758				 */
4759				if (sq->end_added == 0) {
4760					/* Held for PD-API clear that. */
4761					sq->pdapi_aborted = 1;
4762					sq->held_length = 0;
4763					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4764						/*
4765						 * Need to add a PD-API
4766						 * aborted indication.
4767						 * Setting the control_pdapi
4768						 * assures that it will be
4769						 * added right after this
4770						 * msg.
4771						 */
4772						uint32_t strseq;
4773
4774						stcb->asoc.control_pdapi = sq;
4775						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4776						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4777						    stcb,
4778						    SCTP_PARTIAL_DELIVERY_ABORTED,
4779						    (void *)&strseq,
4780						    SCTP_SO_LOCKED);
4781						stcb->asoc.control_pdapi = NULL;
4782					}
4783				}
4784				/* Add an end to wake them */
4785				sq->end_added = 1;
4786			}
4787		}
4788		SCTP_INP_READ_UNLOCK(inp);
4789		if (stcb->block_entry) {
4790			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4791			stcb->block_entry->error = ECONNRESET;
4792			stcb->block_entry = NULL;
4793		}
4794	}
4795	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4796		/*
4797		 * Someone holds a reference OR the socket is unaccepted
4798		 * yet.
4799		 */
4800		if ((stcb->asoc.refcnt) ||
4801		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4802		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4803			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4804			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4805		}
4806		SCTP_TCB_UNLOCK(stcb);
4807		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4808		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4809			/* nothing around */
4810			so = NULL;
4811		if (so) {
4812			/* Wake any reader/writers */
4813			sctp_sorwakeup(inp, so);
4814			sctp_sowwakeup(inp, so);
4815		}
4816#ifdef SCTP_LOG_CLOSING
4817		sctp_log_closing(inp, stcb, 9);
4818#endif
4819		/* no asoc destroyed */
4820		return (0);
4821	}
4822#ifdef SCTP_LOG_CLOSING
4823	sctp_log_closing(inp, stcb, 10);
4824#endif
4825	/*
4826	 * When I reach here, no others want to kill the assoc yet.. and I
4827	 * own the lock. Now its possible an abort comes in when I do the
4828	 * lock exchange below to grab all the locks to do the final take
4829	 * out. to prevent this we increment the count, which will start a
4830	 * timer and blow out above thus assuring us that we hold exclusive
4831	 * killing of the asoc. Note that after getting back the TCB lock we
4832	 * will go ahead and increment the counter back up and stop any
4833	 * timer a passing stranger may have started :-S
4834	 */
4835	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4836		atomic_add_int(&stcb->asoc.refcnt, 1);
4837
4838		SCTP_TCB_UNLOCK(stcb);
4839		SCTP_INP_INFO_WLOCK();
4840		SCTP_INP_WLOCK(inp);
4841		SCTP_TCB_LOCK(stcb);
4842	}
4843	/* Double check the GONE flag */
4844	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4845	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4846		/* nothing around */
4847		so = NULL;
4848
4849	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4850	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4851		/*
4852		 * For TCP type we need special handling when we are
4853		 * connected. We also include the peel'ed off ones to.
4854		 */
4855		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4856			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4857			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4858			if (so) {
4859				SOCK_LOCK(so);
4860				if (so->so_rcv.sb_cc == 0) {
4861					so->so_state &= ~(SS_ISCONNECTING |
4862					    SS_ISDISCONNECTING |
4863					    SS_ISCONFIRMING |
4864					    SS_ISCONNECTED);
4865				}
4866				socantrcvmore_locked(so);
4867				sctp_sowwakeup(inp, so);
4868				sctp_sorwakeup(inp, so);
4869				SCTP_SOWAKEUP(so);
4870			}
4871		}
4872	}
4873	/*
4874	 * Make it invalid too, that way if its about to run it will abort
4875	 * and return.
4876	 */
4877	/* re-increment the lock */
4878	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4879		atomic_add_int(&stcb->asoc.refcnt, -1);
4880	}
4881	if (stcb->asoc.refcnt) {
4882		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4883		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4884		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4885			SCTP_INP_INFO_WUNLOCK();
4886			SCTP_INP_WUNLOCK(inp);
4887		}
4888		SCTP_TCB_UNLOCK(stcb);
4889		return (0);
4890	}
4891	asoc->state = 0;
4892	if (inp->sctp_tcbhash) {
4893		LIST_REMOVE(stcb, sctp_tcbhash);
4894	}
4895	if (stcb->asoc.in_asocid_hash) {
4896		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4897	}
4898	/* Now lets remove it from the list of ALL associations in the EP */
4899	LIST_REMOVE(stcb, sctp_tcblist);
4900	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4901		SCTP_INP_INCR_REF(inp);
4902		SCTP_INP_WUNLOCK(inp);
4903	}
4904	/* pull from vtag hash */
4905	LIST_REMOVE(stcb, sctp_asocs);
4906	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4907	    inp->sctp_lport, stcb->rport);
4908
4909	/*
4910	 * Now restop the timers to be sure this is paranoia at is finest!
4911	 */
4912	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4913	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4914	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4915	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4916	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4917	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4918	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4919	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4920		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4921		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4922		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4923	}
4924
4925	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4926	/*
4927	 * The chunk lists and such SHOULD be empty but we check them just
4928	 * in case.
4929	 */
4930	/* anything on the wheel needs to be removed */
4931	for (i = 0; i < asoc->streamoutcnt; i++) {
4932		struct sctp_stream_out *outs;
4933
4934		outs = &asoc->strmout[i];
4935		/* now clean up any chunks here */
4936		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4937			TAILQ_REMOVE(&outs->outqueue, sp, next);
4938			sctp_free_spbufspace(stcb, asoc, sp);
4939			if (sp->data) {
4940				if (so) {
4941					/* Still an open socket - report */
4942					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4943					    0, (void *)sp, SCTP_SO_LOCKED);
4944				}
4945				if (sp->data) {
4946					sctp_m_freem(sp->data);
4947					sp->data = NULL;
4948					sp->tail_mbuf = NULL;
4949					sp->length = 0;
4950				}
4951			}
4952			if (sp->net) {
4953				sctp_free_remote_addr(sp->net);
4954				sp->net = NULL;
4955			}
4956			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4957		}
4958	}
4959	/* sa_ignore FREED_MEMORY */
4960	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4961		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4962		SCTP_FREE(strrst, SCTP_M_STRESET);
4963	}
4964	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4965		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4966		if (sq->data) {
4967			sctp_m_freem(sq->data);
4968			sq->data = NULL;
4969		}
4970		sctp_free_remote_addr(sq->whoFrom);
4971		sq->whoFrom = NULL;
4972		sq->stcb = NULL;
4973		/* Free the ctl entry */
4974		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4975		SCTP_DECR_READQ_COUNT();
4976		/* sa_ignore FREED_MEMORY */
4977	}
4978	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4979		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4980		if (chk->data) {
4981			sctp_m_freem(chk->data);
4982			chk->data = NULL;
4983		}
4984		if (chk->holds_key_ref)
4985			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4986		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4987		SCTP_DECR_CHK_COUNT();
4988		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4989		asoc->free_chunk_cnt--;
4990		/* sa_ignore FREED_MEMORY */
4991	}
4992	/* pending send queue SHOULD be empty */
4993	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4994		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4995			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4996#ifdef INVARIANTS
4997		} else {
4998			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4999#endif
5000		}
5001		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
5002		if (chk->data) {
5003			if (so) {
5004				/* Still a socket? */
5005				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
5006				    0, chk, SCTP_SO_LOCKED);
5007			}
5008			if (chk->data) {
5009				sctp_m_freem(chk->data);
5010				chk->data = NULL;
5011			}
5012		}
5013		if (chk->holds_key_ref)
5014			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5015		if (chk->whoTo) {
5016			sctp_free_remote_addr(chk->whoTo);
5017			chk->whoTo = NULL;
5018		}
5019		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5020		SCTP_DECR_CHK_COUNT();
5021		/* sa_ignore FREED_MEMORY */
5022	}
5023	/* sent queue SHOULD be empty */
5024	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5025		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5026			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
5027				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
5028#ifdef INVARIANTS
5029			} else {
5030				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
5031#endif
5032			}
5033		}
5034		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5035		if (chk->data) {
5036			if (so) {
5037				/* Still a socket? */
5038				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5039				    0, chk, SCTP_SO_LOCKED);
5040			}
5041			if (chk->data) {
5042				sctp_m_freem(chk->data);
5043				chk->data = NULL;
5044			}
5045		}
5046		if (chk->holds_key_ref)
5047			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5048		sctp_free_remote_addr(chk->whoTo);
5049		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5050		SCTP_DECR_CHK_COUNT();
5051		/* sa_ignore FREED_MEMORY */
5052	}
5053#ifdef INVARIANTS
5054	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5055		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5056			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5057		}
5058	}
5059#endif
5060	/* control queue MAY not be empty */
5061	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5062		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5063		if (chk->data) {
5064			sctp_m_freem(chk->data);
5065			chk->data = NULL;
5066		}
5067		if (chk->holds_key_ref)
5068			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5069		sctp_free_remote_addr(chk->whoTo);
5070		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5071		SCTP_DECR_CHK_COUNT();
5072		/* sa_ignore FREED_MEMORY */
5073	}
5074	/* ASCONF queue MAY not be empty */
5075	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5076		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5077		if (chk->data) {
5078			sctp_m_freem(chk->data);
5079			chk->data = NULL;
5080		}
5081		if (chk->holds_key_ref)
5082			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5083		sctp_free_remote_addr(chk->whoTo);
5084		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5085		SCTP_DECR_CHK_COUNT();
5086		/* sa_ignore FREED_MEMORY */
5087	}
5088	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5089		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5090		if (chk->data) {
5091			sctp_m_freem(chk->data);
5092			chk->data = NULL;
5093		}
5094		if (chk->holds_key_ref)
5095			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5096		sctp_free_remote_addr(chk->whoTo);
5097		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5098		SCTP_DECR_CHK_COUNT();
5099		/* sa_ignore FREED_MEMORY */
5100	}
5101
5102	if (asoc->mapping_array) {
5103		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5104		asoc->mapping_array = NULL;
5105	}
5106	if (asoc->nr_mapping_array) {
5107		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5108		asoc->nr_mapping_array = NULL;
5109	}
5110	/* the stream outs */
5111	if (asoc->strmout) {
5112		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5113		asoc->strmout = NULL;
5114	}
5115	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5116	if (asoc->strmin) {
5117		struct sctp_queued_to_read *ctl, *nctl;
5118
5119		for (i = 0; i < asoc->streamincnt; i++) {
5120			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5121				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5122				sctp_free_remote_addr(ctl->whoFrom);
5123				if (ctl->data) {
5124					sctp_m_freem(ctl->data);
5125					ctl->data = NULL;
5126				}
5127				/*
5128				 * We don't free the address here since all
5129				 * the net's were freed above.
5130				 */
5131				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5132				SCTP_DECR_READQ_COUNT();
5133			}
5134		}
5135		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5136		asoc->strmin = NULL;
5137	}
5138	asoc->streamincnt = 0;
5139	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5140#ifdef INVARIANTS
5141		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5142			panic("no net's left alloc'ed, or list points to itself");
5143		}
5144#endif
5145		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5146		sctp_free_remote_addr(net);
5147	}
5148	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5149		/* sa_ignore FREED_MEMORY */
5150		sctp_remove_laddr(laddr);
5151	}
5152
5153	/* pending asconf (address) parameters */
5154	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5155		/* sa_ignore FREED_MEMORY */
5156		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5157		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5158	}
5159	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5160		/* sa_ignore FREED_MEMORY */
5161		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5162		if (aack->data != NULL) {
5163			sctp_m_freem(aack->data);
5164		}
5165		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5166	}
5167	/* clean up auth stuff */
5168	if (asoc->local_hmacs)
5169		sctp_free_hmaclist(asoc->local_hmacs);
5170	if (asoc->peer_hmacs)
5171		sctp_free_hmaclist(asoc->peer_hmacs);
5172
5173	if (asoc->local_auth_chunks)
5174		sctp_free_chunklist(asoc->local_auth_chunks);
5175	if (asoc->peer_auth_chunks)
5176		sctp_free_chunklist(asoc->peer_auth_chunks);
5177
5178	sctp_free_authinfo(&asoc->authinfo);
5179
5180	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5181		LIST_REMOVE(shared_key, next);
5182		sctp_free_sharedkey(shared_key);
5183		/* sa_ignore FREED_MEMORY */
5184	}
5185
5186	/* Insert new items here :> */
5187
5188	/* Get rid of LOCK */
5189	SCTP_TCB_UNLOCK(stcb);
5190	SCTP_TCB_LOCK_DESTROY(stcb);
5191	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5192	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5193		SCTP_INP_INFO_WUNLOCK();
5194		SCTP_INP_RLOCK(inp);
5195	}
5196#ifdef SCTP_TRACK_FREED_ASOCS
5197	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5198		/* now clean up the tasoc itself */
5199		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5200		SCTP_DECR_ASOC_COUNT();
5201	} else {
5202		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5203	}
5204#else
5205	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5206	SCTP_DECR_ASOC_COUNT();
5207#endif
5208	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5209		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5210			/*
5211			 * If its NOT the inp_free calling us AND sctp_close
5212			 * as been called, we call back...
5213			 */
5214			SCTP_INP_RUNLOCK(inp);
5215			/*
5216			 * This will start the kill timer (if we are the
5217			 * last one) since we hold an increment yet. But
5218			 * this is the only safe way to do this since
5219			 * otherwise if the socket closes at the same time
5220			 * we are here we might collide in the cleanup.
5221			 */
5222			sctp_inpcb_free(inp,
5223			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5224			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5225			SCTP_INP_DECR_REF(inp);
5226			goto out_of;
5227		} else {
5228			/* The socket is still open. */
5229			SCTP_INP_DECR_REF(inp);
5230		}
5231	}
5232	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5233		SCTP_INP_RUNLOCK(inp);
5234	}
5235out_of:
5236	/* destroyed the asoc */
5237#ifdef SCTP_LOG_CLOSING
5238	sctp_log_closing(inp, NULL, 11);
5239#endif
5240	return (1);
5241}
5242
5243
5244
5245/*
5246 * determine if a destination is "reachable" based upon the addresses bound
5247 * to the current endpoint (e.g. only v4 or v6 currently bound)
5248 */
5249/*
5250 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5251 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5252 * types bound as its endpoint
5253 */
5254int
5255sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5256{
5257	struct sctp_inpcb *inp;
5258	int answer;
5259
5260	/*
5261	 * No locks here, the TCB, in all cases is already locked and an
5262	 * assoc is up. There is either a INP lock by the caller applied (in
5263	 * asconf case when deleting an address) or NOT in the HB case,
5264	 * however if HB then the INP increment is up and the INP will not
5265	 * be removed (on top of the fact that we have a TCB lock). So we
5266	 * only want to read the sctp_flags, which is either bound-all or
5267	 * not.. no protection needed since once an assoc is up you can't be
5268	 * changing your binding.
5269	 */
5270	inp = stcb->sctp_ep;
5271	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5272		/* if bound all, destination is not restricted */
5273		/*
5274		 * RRS: Question during lock work: Is this correct? If you
5275		 * are bound-all you still might need to obey the V4--V6
5276		 * flags??? IMO this bound-all stuff needs to be removed!
5277		 */
5278		return (1);
5279	}
5280	/* NOTE: all "scope" checks are done when local addresses are added */
5281	switch (destaddr->sa_family) {
5282#ifdef INET6
5283	case AF_INET6:
5284		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5285		break;
5286#endif
5287#ifdef INET
5288	case AF_INET:
5289		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5290		break;
5291#endif
5292	default:
5293		/* invalid family, so it's unreachable */
5294		answer = 0;
5295		break;
5296	}
5297	return (answer);
5298}
5299
5300/*
5301 * update the inp_vflags on an endpoint
5302 */
5303static void
5304sctp_update_ep_vflag(struct sctp_inpcb *inp)
5305{
5306	struct sctp_laddr *laddr;
5307
5308	/* first clear the flag */
5309	inp->ip_inp.inp.inp_vflag = 0;
5310	/* set the flag based on addresses on the ep list */
5311	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5312		if (laddr->ifa == NULL) {
5313			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5314			    __FUNCTION__);
5315			continue;
5316		}
5317		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5318			continue;
5319		}
5320		switch (laddr->ifa->address.sa.sa_family) {
5321#ifdef INET6
5322		case AF_INET6:
5323			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5324			break;
5325#endif
5326#ifdef INET
5327		case AF_INET:
5328			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5329			break;
5330#endif
5331		default:
5332			break;
5333		}
5334	}
5335}
5336
5337/*
5338 * Add the address to the endpoint local address list There is nothing to be
5339 * done if we are bound to all addresses
5340 */
5341void
5342sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5343{
5344	struct sctp_laddr *laddr;
5345	int fnd, error = 0;
5346
5347	fnd = 0;
5348
5349	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5350		/* You are already bound to all. You have it already */
5351		return;
5352	}
5353#ifdef INET6
5354	if (ifa->address.sa.sa_family == AF_INET6) {
5355		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5356			/* Can't bind a non-useable addr. */
5357			return;
5358		}
5359	}
5360#endif
5361	/* first, is it already present? */
5362	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5363		if (laddr->ifa == ifa) {
5364			fnd = 1;
5365			break;
5366		}
5367	}
5368
5369	if (fnd == 0) {
5370		/* Not in the ep list */
5371		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5372		if (error != 0)
5373			return;
5374		inp->laddr_count++;
5375		/* update inp_vflag flags */
5376		switch (ifa->address.sa.sa_family) {
5377#ifdef INET6
5378		case AF_INET6:
5379			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5380			break;
5381#endif
5382#ifdef INET
5383		case AF_INET:
5384			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5385			break;
5386#endif
5387		default:
5388			break;
5389		}
5390	}
5391	return;
5392}
5393
5394
5395/*
5396 * select a new (hopefully reachable) destination net (should only be used
5397 * when we deleted an ep addr that is the only usable source address to reach
5398 * the destination net)
5399 */
5400static void
5401sctp_select_primary_destination(struct sctp_tcb *stcb)
5402{
5403	struct sctp_nets *net;
5404
5405	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5406		/* for now, we'll just pick the first reachable one we find */
5407		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5408			continue;
5409		if (sctp_destination_is_reachable(stcb,
5410		    (struct sockaddr *)&net->ro._l_addr)) {
5411			/* found a reachable destination */
5412			stcb->asoc.primary_destination = net;
5413		}
5414	}
5415	/* I can't there from here! ...we're gonna die shortly... */
5416}
5417
5418
5419/*
5420 * Delete the address from the endpoint local address list There is nothing
5421 * to be done if we are bound to all addresses
5422 */
5423void
5424sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5425{
5426	struct sctp_laddr *laddr;
5427	int fnd;
5428
5429	fnd = 0;
5430	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5431		/* You are already bound to all. You have it already */
5432		return;
5433	}
5434	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5435		if (laddr->ifa == ifa) {
5436			fnd = 1;
5437			break;
5438		}
5439	}
5440	if (fnd && (inp->laddr_count < 2)) {
5441		/* can't delete unless there are at LEAST 2 addresses */
5442		return;
5443	}
5444	if (fnd) {
5445		/*
5446		 * clean up any use of this address go through our
5447		 * associations and clear any last_used_address that match
5448		 * this one for each assoc, see if a new primary_destination
5449		 * is needed
5450		 */
5451		struct sctp_tcb *stcb;
5452
5453		/* clean up "next_addr_touse" */
5454		if (inp->next_addr_touse == laddr)
5455			/* delete this address */
5456			inp->next_addr_touse = NULL;
5457
5458		/* clean up "last_used_address" */
5459		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5460			struct sctp_nets *net;
5461
5462			SCTP_TCB_LOCK(stcb);
5463			if (stcb->asoc.last_used_address == laddr)
5464				/* delete this address */
5465				stcb->asoc.last_used_address = NULL;
5466			/*
5467			 * Now spin through all the nets and purge any ref
5468			 * to laddr
5469			 */
5470			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5471				if (net->ro._s_addr &&
5472				    (net->ro._s_addr->ifa == laddr->ifa)) {
5473					/* Yep, purge src address selected */
5474					sctp_rtentry_t *rt;
5475
5476					/* delete this address if cached */
5477					rt = net->ro.ro_rt;
5478					if (rt != NULL) {
5479						RTFREE(rt);
5480						net->ro.ro_rt = NULL;
5481					}
5482					sctp_free_ifa(net->ro._s_addr);
5483					net->ro._s_addr = NULL;
5484					net->src_addr_selected = 0;
5485				}
5486			}
5487			SCTP_TCB_UNLOCK(stcb);
5488		}		/* for each tcb */
5489		/* remove it from the ep list */
5490		sctp_remove_laddr(laddr);
5491		inp->laddr_count--;
5492		/* update inp_vflag flags */
5493		sctp_update_ep_vflag(inp);
5494	}
5495	return;
5496}
5497
5498/*
5499 * Add the address to the TCB local address restricted list.
5500 * This is a "pending" address list (eg. addresses waiting for an
5501 * ASCONF-ACK response) and cannot be used as a valid source address.
5502 */
5503void
5504sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5505{
5506	struct sctp_laddr *laddr;
5507	struct sctpladdr *list;
5508
5509	/*
5510	 * Assumes TCB is locked.. and possibly the INP. May need to
5511	 * confirm/fix that if we need it and is not the case.
5512	 */
5513	list = &stcb->asoc.sctp_restricted_addrs;
5514
5515#ifdef INET6
5516	if (ifa->address.sa.sa_family == AF_INET6) {
5517		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5518			/* Can't bind a non-existent addr. */
5519			return;
5520		}
5521	}
5522#endif
5523	/* does the address already exist? */
5524	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5525		if (laddr->ifa == ifa) {
5526			return;
5527		}
5528	}
5529
5530	/* add to the list */
5531	(void)sctp_insert_laddr(list, ifa, 0);
5532	return;
5533}
5534
5535/*
5536 * insert an laddr entry with the given ifa for the desired list
5537 */
5538int
5539sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5540{
5541	struct sctp_laddr *laddr;
5542
5543	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5544	if (laddr == NULL) {
5545		/* out of memory? */
5546		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5547		return (EINVAL);
5548	}
5549	SCTP_INCR_LADDR_COUNT();
5550	bzero(laddr, sizeof(*laddr));
5551	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5552	laddr->ifa = ifa;
5553	laddr->action = act;
5554	atomic_add_int(&ifa->refcount, 1);
5555	/* insert it */
5556	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5557
5558	return (0);
5559}
5560
5561/*
5562 * Remove an laddr entry from the local address list (on an assoc)
5563 */
5564void
5565sctp_remove_laddr(struct sctp_laddr *laddr)
5566{
5567
5568	/* remove from the list */
5569	LIST_REMOVE(laddr, sctp_nxt_addr);
5570	sctp_free_ifa(laddr->ifa);
5571	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5572	SCTP_DECR_LADDR_COUNT();
5573}
5574
5575/*
5576 * Remove a local address from the TCB local address restricted list
5577 */
5578void
5579sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5580{
5581	struct sctp_inpcb *inp;
5582	struct sctp_laddr *laddr;
5583
5584	/*
5585	 * This is called by asconf work. It is assumed that a) The TCB is
5586	 * locked and b) The INP is locked. This is true in as much as I can
5587	 * trace through the entry asconf code where I did these locks.
5588	 * Again, the ASCONF code is a bit different in that it does lock
5589	 * the INP during its work often times. This must be since we don't
5590	 * want other proc's looking up things while what they are looking
5591	 * up is changing :-D
5592	 */
5593
5594	inp = stcb->sctp_ep;
5595	/* if subset bound and don't allow ASCONF's, can't delete last */
5596	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5597	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5598		if (stcb->sctp_ep->laddr_count < 2) {
5599			/* can't delete last address */
5600			return;
5601		}
5602	}
5603	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5604		/* remove the address if it exists */
5605		if (laddr->ifa == NULL)
5606			continue;
5607		if (laddr->ifa == ifa) {
5608			sctp_remove_laddr(laddr);
5609			return;
5610		}
5611	}
5612
5613	/* address not found! */
5614	return;
5615}
5616
5617/*
5618 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5619 */
5620/* sysctl */
5621static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5622static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5623
5624
5625
5626#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5627struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5628int *sctp_cpuarry = NULL;
5629void
5630sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5631{
5632	/* Queue a packet to a processor for the specified core */
5633	struct sctp_mcore_queue *qent;
5634	struct sctp_mcore_ctrl *wkq;
5635	int need_wake = 0;
5636
5637	if (sctp_mcore_workers == NULL) {
5638		/* Something went way bad during setup */
5639		sctp_input_with_port(m, off, 0);
5640		return;
5641	}
5642	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5643	    (sizeof(struct sctp_mcore_queue)),
5644	    SCTP_M_MCORE);
5645	if (qent == NULL) {
5646		/* This is trouble  */
5647		sctp_input_with_port(m, off, 0);
5648		return;
5649	}
5650	qent->vn = curvnet;
5651	qent->m = m;
5652	qent->off = off;
5653	qent->v6 = 0;
5654	wkq = &sctp_mcore_workers[cpu_to_use];
5655	SCTP_MCORE_QLOCK(wkq);
5656
5657	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5658	if (wkq->running == 0) {
5659		need_wake = 1;
5660	}
5661	SCTP_MCORE_QUNLOCK(wkq);
5662	if (need_wake) {
5663		wakeup(&wkq->running);
5664	}
5665}
5666
5667static void
5668sctp_mcore_thread(void *arg)
5669{
5670
5671	struct sctp_mcore_ctrl *wkq;
5672	struct sctp_mcore_queue *qent;
5673
5674	wkq = (struct sctp_mcore_ctrl *)arg;
5675	struct mbuf *m;
5676	int off, v6;
5677
5678	/* Wait for first tickle */
5679	SCTP_MCORE_LOCK(wkq);
5680	wkq->running = 0;
5681	msleep(&wkq->running,
5682	    &wkq->core_mtx,
5683	    0, "wait for pkt", 0);
5684	SCTP_MCORE_UNLOCK(wkq);
5685
5686	/* Bind to our cpu */
5687	thread_lock(curthread);
5688	sched_bind(curthread, wkq->cpuid);
5689	thread_unlock(curthread);
5690
5691	/* Now lets start working */
5692	SCTP_MCORE_LOCK(wkq);
5693	/* Now grab lock and go */
5694	for (;;) {
5695		SCTP_MCORE_QLOCK(wkq);
5696skip_sleep:
5697		wkq->running = 1;
5698		qent = TAILQ_FIRST(&wkq->que);
5699		if (qent) {
5700			TAILQ_REMOVE(&wkq->que, qent, next);
5701			SCTP_MCORE_QUNLOCK(wkq);
5702			CURVNET_SET(qent->vn);
5703			m = qent->m;
5704			off = qent->off;
5705			v6 = qent->v6;
5706			SCTP_FREE(qent, SCTP_M_MCORE);
5707			if (v6 == 0) {
5708				sctp_input_with_port(m, off, 0);
5709			} else {
5710				SCTP_PRINTF("V6 not yet supported\n");
5711				sctp_m_freem(m);
5712			}
5713			CURVNET_RESTORE();
5714			SCTP_MCORE_QLOCK(wkq);
5715		}
5716		wkq->running = 0;
5717		if (!TAILQ_EMPTY(&wkq->que)) {
5718			goto skip_sleep;
5719		}
5720		SCTP_MCORE_QUNLOCK(wkq);
5721		msleep(&wkq->running,
5722		    &wkq->core_mtx,
5723		    0, "wait for pkt", 0);
5724	}
5725}
5726
5727static void
5728sctp_startup_mcore_threads(void)
5729{
5730	int i, cpu;
5731
5732	if (mp_ncpus == 1)
5733		return;
5734
5735	if (sctp_mcore_workers != NULL) {
5736		/*
5737		 * Already been here in some previous vnet?
5738		 */
5739		return;
5740	}
5741	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5742	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5743	    SCTP_M_MCORE);
5744	if (sctp_mcore_workers == NULL) {
5745		/* TSNH I hope */
5746		return;
5747	}
5748	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5749	    sizeof(struct sctp_mcore_ctrl)));
5750	/* Init the structures */
5751	for (i = 0; i <= mp_maxid; i++) {
5752		TAILQ_INIT(&sctp_mcore_workers[i].que);
5753		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5754		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5755		sctp_mcore_workers[i].cpuid = i;
5756	}
5757	if (sctp_cpuarry == NULL) {
5758		SCTP_MALLOC(sctp_cpuarry, int *,
5759		    (mp_ncpus * sizeof(int)),
5760		    SCTP_M_MCORE);
5761		i = 0;
5762		CPU_FOREACH(cpu) {
5763			sctp_cpuarry[i] = cpu;
5764			i++;
5765		}
5766	}
5767	/* Now start them all */
5768	CPU_FOREACH(cpu) {
5769		(void)kproc_create(sctp_mcore_thread,
5770		    (void *)&sctp_mcore_workers[cpu],
5771		    &sctp_mcore_workers[cpu].thread_proc,
5772		    RFPROC,
5773		    SCTP_KTHREAD_PAGES,
5774		    SCTP_MCORE_NAME);
5775
5776	}
5777}
5778
5779#endif
5780
5781void
5782sctp_pcb_init()
5783{
5784	/*
5785	 * SCTP initialization for the PCB structures should be called by
5786	 * the sctp_init() funciton.
5787	 */
5788	int i;
5789	struct timeval tv;
5790
5791	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5792		/* error I was called twice */
5793		return;
5794	}
5795	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5796
5797#if defined(SCTP_LOCAL_TRACE_BUF)
5798	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5799#endif
5800#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5801	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5802	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5803	    SCTP_M_MCORE);
5804#endif
5805	(void)SCTP_GETTIME_TIMEVAL(&tv);
5806#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5807	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5808	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5809	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5810#else
5811	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5812	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5813	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5814#endif
5815	/* init the empty list of (All) Endpoints */
5816	LIST_INIT(&SCTP_BASE_INFO(listhead));
5817
5818
5819	/* init the hash table of endpoints */
5820	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5821	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5822	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5823	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5824	    &SCTP_BASE_INFO(hashasocmark));
5825	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5826	    &SCTP_BASE_INFO(hashmark));
5827	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5828	    &SCTP_BASE_INFO(hashtcpmark));
5829	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5830
5831
5832	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5833	    &SCTP_BASE_INFO(hashvrfmark));
5834
5835	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5836	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5837	/* init the zones */
5838	/*
5839	 * FIX ME: Should check for NULL returns, but if it does fail we are
5840	 * doomed to panic anyways... add later maybe.
5841	 */
5842	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5843	    sizeof(struct sctp_inpcb), maxsockets);
5844
5845	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5846	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5847
5848	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5849	    sizeof(struct sctp_laddr),
5850	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5851
5852	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5853	    sizeof(struct sctp_nets),
5854	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5855
5856	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5857	    sizeof(struct sctp_tmit_chunk),
5858	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5859
5860	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5861	    sizeof(struct sctp_queued_to_read),
5862	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5863
5864	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5865	    sizeof(struct sctp_stream_queue_pending),
5866	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5867
5868	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5869	    sizeof(struct sctp_asconf),
5870	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5871
5872	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5873	    sizeof(struct sctp_asconf_ack),
5874	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5875
5876
5877	/* Master Lock INIT for info structure */
5878	SCTP_INP_INFO_LOCK_INIT();
5879	SCTP_STATLOG_INIT_LOCK();
5880
5881	SCTP_IPI_COUNT_INIT();
5882	SCTP_IPI_ADDR_INIT();
5883#ifdef SCTP_PACKET_LOGGING
5884	SCTP_IP_PKTLOG_INIT();
5885#endif
5886	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5887
5888	SCTP_WQ_ADDR_INIT();
5889	/* not sure if we need all the counts */
5890	SCTP_BASE_INFO(ipi_count_ep) = 0;
5891	/* assoc/tcb zone info */
5892	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5893	/* local addrlist zone info */
5894	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5895	/* remote addrlist zone info */
5896	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5897	/* chunk info */
5898	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5899
5900	/* socket queue zone info */
5901	SCTP_BASE_INFO(ipi_count_readq) = 0;
5902
5903	/* stream out queue cont */
5904	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5905
5906	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5907	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5908
5909	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5910
5911	/* Init the TIMEWAIT list */
5912	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5913		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5914	}
5915	sctp_startup_iterator();
5916
5917#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5918	sctp_startup_mcore_threads();
5919#endif
5920
5921	/*
5922	 * INIT the default VRF which for BSD is the only one, other O/S's
5923	 * may have more. But initially they must start with one and then
5924	 * add the VRF's as addresses are added.
5925	 */
5926	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5927}
5928
5929/*
5930 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5931 */
5932void
5933sctp_pcb_finish(void)
5934{
5935	struct sctp_vrflist *vrf_bucket;
5936	struct sctp_vrf *vrf, *nvrf;
5937	struct sctp_ifn *ifn, *nifn;
5938	struct sctp_ifa *ifa, *nifa;
5939	struct sctpvtaghead *chain;
5940	struct sctp_tagblock *twait_block, *prev_twait_block;
5941	struct sctp_laddr *wi, *nwi;
5942	int i;
5943	struct sctp_iterator *it, *nit;
5944
5945	/*
5946	 * In FreeBSD the iterator thread never exits but we do clean up.
5947	 * The only way FreeBSD reaches here is if we have VRF's but we
5948	 * still add the ifdef to make it compile on old versions.
5949	 */
5950	SCTP_IPI_ITERATOR_WQ_LOCK();
5951	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5952		if (it->vn != curvnet) {
5953			continue;
5954		}
5955		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5956		if (it->function_atend != NULL) {
5957			(*it->function_atend) (it->pointer, it->val);
5958		}
5959		SCTP_FREE(it, SCTP_M_ITER);
5960	}
5961	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5962	SCTP_ITERATOR_LOCK();
5963	if ((sctp_it_ctl.cur_it) &&
5964	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5965		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5966	}
5967	SCTP_ITERATOR_UNLOCK();
5968	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5969	SCTP_WQ_ADDR_LOCK();
5970	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5971		LIST_REMOVE(wi, sctp_nxt_addr);
5972		SCTP_DECR_LADDR_COUNT();
5973		if (wi->action == SCTP_DEL_IP_ADDRESS) {
5974			SCTP_FREE(wi->ifa, SCTP_M_IFA);
5975		}
5976		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5977	}
5978	SCTP_WQ_ADDR_UNLOCK();
5979
5980	/*
5981	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5982	 * destroyed first).
5983	 */
5984	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5985	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5986		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5987			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5988				/* free the ifa */
5989				LIST_REMOVE(ifa, next_bucket);
5990				LIST_REMOVE(ifa, next_ifa);
5991				SCTP_FREE(ifa, SCTP_M_IFA);
5992			}
5993			/* free the ifn */
5994			LIST_REMOVE(ifn, next_bucket);
5995			LIST_REMOVE(ifn, next_ifn);
5996			SCTP_FREE(ifn, SCTP_M_IFN);
5997		}
5998		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5999		/* free the vrf */
6000		LIST_REMOVE(vrf, next_vrf);
6001		SCTP_FREE(vrf, SCTP_M_VRF);
6002	}
6003	/* free the vrf hashes */
6004	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
6005	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
6006
6007	/*
6008	 * free the TIMEWAIT list elements malloc'd in the function
6009	 * sctp_add_vtag_to_timewait()...
6010	 */
6011	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
6012		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
6013		if (!LIST_EMPTY(chain)) {
6014			prev_twait_block = NULL;
6015			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6016				if (prev_twait_block) {
6017					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6018				}
6019				prev_twait_block = twait_block;
6020			}
6021			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6022		}
6023	}
6024
6025	/* free the locks and mutexes */
6026#ifdef SCTP_PACKET_LOGGING
6027	SCTP_IP_PKTLOG_DESTROY();
6028#endif
6029	SCTP_IPI_ADDR_DESTROY();
6030	SCTP_STATLOG_DESTROY();
6031	SCTP_INP_INFO_LOCK_DESTROY();
6032
6033	SCTP_WQ_ADDR_DESTROY();
6034
6035	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6036	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6037	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6038	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6039	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6040	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6041	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6042	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6043	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6044	/* Get rid of other stuff to */
6045	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
6046		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
6047	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
6048		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6049	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6050		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6051#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6052	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6053#endif
6054}
6055
6056
6057int
6058sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6059    int offset, int limit,
6060    struct sockaddr *src, struct sockaddr *dst,
6061    struct sockaddr *altsa)
6062{
6063	/*
6064	 * grub through the INIT pulling addresses and loading them to the
6065	 * nets structure in the asoc. The from address in the mbuf should
6066	 * also be loaded (if it is not already). This routine can be called
6067	 * with either INIT or INIT-ACK's as long as the m points to the IP
6068	 * packet and the offset points to the beginning of the parameters.
6069	 */
6070	struct sctp_inpcb *inp;
6071	struct sctp_nets *net, *nnet, *net_tmp;
6072	struct sctp_paramhdr *phdr, parm_buf;
6073	struct sctp_tcb *stcb_tmp;
6074	uint16_t ptype, plen;
6075	struct sockaddr *sa;
6076	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6077	struct sctp_auth_random *p_random = NULL;
6078	uint16_t random_len = 0;
6079	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6080	struct sctp_auth_hmac_algo *hmacs = NULL;
6081	uint16_t hmacs_len = 0;
6082	uint8_t saw_asconf = 0;
6083	uint8_t saw_asconf_ack = 0;
6084	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6085	struct sctp_auth_chunk_list *chunks = NULL;
6086	uint16_t num_chunks = 0;
6087	sctp_key_t *new_key;
6088	uint32_t keylen;
6089	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6090	uint8_t peer_supports_ecn;
6091	uint8_t peer_supports_prsctp;
6092	uint8_t peer_supports_auth;
6093	uint8_t peer_supports_asconf;
6094	uint8_t peer_supports_asconf_ack;
6095	uint8_t peer_supports_reconfig;
6096	uint8_t peer_supports_nrsack;
6097	uint8_t peer_supports_pktdrop;
6098
6099#ifdef INET
6100	struct sockaddr_in sin;
6101
6102#endif
6103#ifdef INET6
6104	struct sockaddr_in6 sin6;
6105
6106#endif
6107
6108	/* First get the destination address setup too. */
6109#ifdef INET
6110	memset(&sin, 0, sizeof(sin));
6111	sin.sin_family = AF_INET;
6112	sin.sin_len = sizeof(sin);
6113	sin.sin_port = stcb->rport;
6114#endif
6115#ifdef INET6
6116	memset(&sin6, 0, sizeof(sin6));
6117	sin6.sin6_family = AF_INET6;
6118	sin6.sin6_len = sizeof(struct sockaddr_in6);
6119	sin6.sin6_port = stcb->rport;
6120#endif
6121	if (altsa) {
6122		sa = altsa;
6123	} else {
6124		sa = src;
6125	}
6126	peer_supports_ecn = 0;
6127	peer_supports_prsctp = 0;
6128	peer_supports_auth = 0;
6129	peer_supports_asconf = 0;
6130	peer_supports_reconfig = 0;
6131	peer_supports_nrsack = 0;
6132	peer_supports_pktdrop = 0;
6133	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6134		/* mark all addresses that we have currently on the list */
6135		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6136	}
6137	/* does the source address already exist? if so skip it */
6138	inp = stcb->sctp_ep;
6139	atomic_add_int(&stcb->asoc.refcnt, 1);
6140	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6141	atomic_add_int(&stcb->asoc.refcnt, -1);
6142
6143	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6144		/* we must add the source address */
6145		/* no scope set here since we have a tcb already. */
6146		switch (sa->sa_family) {
6147#ifdef INET
6148		case AF_INET:
6149			if (stcb->asoc.scope.ipv4_addr_legal) {
6150				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6151					return (-1);
6152				}
6153			}
6154			break;
6155#endif
6156#ifdef INET6
6157		case AF_INET6:
6158			if (stcb->asoc.scope.ipv6_addr_legal) {
6159				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6160					return (-2);
6161				}
6162			}
6163			break;
6164#endif
6165		default:
6166			break;
6167		}
6168	} else {
6169		if (net_tmp != NULL && stcb_tmp == stcb) {
6170			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6171		} else if (stcb_tmp != stcb) {
6172			/* It belongs to another association? */
6173			if (stcb_tmp)
6174				SCTP_TCB_UNLOCK(stcb_tmp);
6175			return (-3);
6176		}
6177	}
6178	if (stcb->asoc.state == 0) {
6179		/* the assoc was freed? */
6180		return (-4);
6181	}
6182	/* now we must go through each of the params. */
6183	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6184	while (phdr) {
6185		ptype = ntohs(phdr->param_type);
6186		plen = ntohs(phdr->param_length);
6187		/*
6188		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6189		 * (uint32_t)ptype, (int)plen);
6190		 */
6191		if (offset + plen > limit) {
6192			break;
6193		}
6194		if (plen == 0) {
6195			break;
6196		}
6197#ifdef INET
6198		if (ptype == SCTP_IPV4_ADDRESS) {
6199			if (stcb->asoc.scope.ipv4_addr_legal) {
6200				struct sctp_ipv4addr_param *p4, p4_buf;
6201
6202				/* ok get the v4 address and check/add */
6203				phdr = sctp_get_next_param(m, offset,
6204				    (struct sctp_paramhdr *)&p4_buf,
6205				    sizeof(p4_buf));
6206				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6207				    phdr == NULL) {
6208					return (-5);
6209				}
6210				p4 = (struct sctp_ipv4addr_param *)phdr;
6211				sin.sin_addr.s_addr = p4->addr;
6212				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6213					/* Skip multi-cast addresses */
6214					goto next_param;
6215				}
6216				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6217				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6218					goto next_param;
6219				}
6220				sa = (struct sockaddr *)&sin;
6221				inp = stcb->sctp_ep;
6222				atomic_add_int(&stcb->asoc.refcnt, 1);
6223				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6224				    dst, stcb);
6225				atomic_add_int(&stcb->asoc.refcnt, -1);
6226
6227				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6228				    inp == NULL) {
6229					/* we must add the source address */
6230					/*
6231					 * no scope set since we have a tcb
6232					 * already
6233					 */
6234
6235					/*
6236					 * we must validate the state again
6237					 * here
6238					 */
6239			add_it_now:
6240					if (stcb->asoc.state == 0) {
6241						/* the assoc was freed? */
6242						return (-7);
6243					}
6244					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6245						return (-8);
6246					}
6247				} else if (stcb_tmp == stcb) {
6248					if (stcb->asoc.state == 0) {
6249						/* the assoc was freed? */
6250						return (-10);
6251					}
6252					if (net != NULL) {
6253						/* clear flag */
6254						net->dest_state &=
6255						    ~SCTP_ADDR_NOT_IN_ASSOC;
6256					}
6257				} else {
6258					/*
6259					 * strange, address is in another
6260					 * assoc? straighten out locks.
6261					 */
6262					if (stcb_tmp) {
6263						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6264							/*
6265							 * in setup state we
6266							 * abort this guy
6267							 */
6268							sctp_abort_an_association(stcb_tmp->sctp_ep,
6269							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6270							goto add_it_now;
6271						}
6272						SCTP_TCB_UNLOCK(stcb_tmp);
6273					}
6274					if (stcb->asoc.state == 0) {
6275						/* the assoc was freed? */
6276						return (-12);
6277					}
6278					return (-13);
6279				}
6280			}
6281		} else
6282#endif
6283#ifdef INET6
6284		if (ptype == SCTP_IPV6_ADDRESS) {
6285			if (stcb->asoc.scope.ipv6_addr_legal) {
6286				/* ok get the v6 address and check/add */
6287				struct sctp_ipv6addr_param *p6, p6_buf;
6288
6289				phdr = sctp_get_next_param(m, offset,
6290				    (struct sctp_paramhdr *)&p6_buf,
6291				    sizeof(p6_buf));
6292				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6293				    phdr == NULL) {
6294					return (-14);
6295				}
6296				p6 = (struct sctp_ipv6addr_param *)phdr;
6297				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6298				    sizeof(p6->addr));
6299				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6300					/* Skip multi-cast addresses */
6301					goto next_param;
6302				}
6303				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6304					/*
6305					 * Link local make no sense without
6306					 * scope
6307					 */
6308					goto next_param;
6309				}
6310				sa = (struct sockaddr *)&sin6;
6311				inp = stcb->sctp_ep;
6312				atomic_add_int(&stcb->asoc.refcnt, 1);
6313				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6314				    dst, stcb);
6315				atomic_add_int(&stcb->asoc.refcnt, -1);
6316				if (stcb_tmp == NULL &&
6317				    (inp == stcb->sctp_ep || inp == NULL)) {
6318					/*
6319					 * we must validate the state again
6320					 * here
6321					 */
6322			add_it_now6:
6323					if (stcb->asoc.state == 0) {
6324						/* the assoc was freed? */
6325						return (-16);
6326					}
6327					/*
6328					 * we must add the address, no scope
6329					 * set
6330					 */
6331					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6332						return (-17);
6333					}
6334				} else if (stcb_tmp == stcb) {
6335					/*
6336					 * we must validate the state again
6337					 * here
6338					 */
6339					if (stcb->asoc.state == 0) {
6340						/* the assoc was freed? */
6341						return (-19);
6342					}
6343					if (net != NULL) {
6344						/* clear flag */
6345						net->dest_state &=
6346						    ~SCTP_ADDR_NOT_IN_ASSOC;
6347					}
6348				} else {
6349					/*
6350					 * strange, address is in another
6351					 * assoc? straighten out locks.
6352					 */
6353					if (stcb_tmp)
6354						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6355							/*
6356							 * in setup state we
6357							 * abort this guy
6358							 */
6359							sctp_abort_an_association(stcb_tmp->sctp_ep,
6360							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6361							goto add_it_now6;
6362						}
6363					SCTP_TCB_UNLOCK(stcb_tmp);
6364
6365					if (stcb->asoc.state == 0) {
6366						/* the assoc was freed? */
6367						return (-21);
6368					}
6369					return (-22);
6370				}
6371			}
6372		} else
6373#endif
6374		if (ptype == SCTP_ECN_CAPABLE) {
6375			peer_supports_ecn = 1;
6376		} else if (ptype == SCTP_ULP_ADAPTATION) {
6377			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6378				struct sctp_adaptation_layer_indication ai,
6379				                                *aip;
6380
6381				phdr = sctp_get_next_param(m, offset,
6382				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6383				aip = (struct sctp_adaptation_layer_indication *)phdr;
6384				if (aip) {
6385					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6386					stcb->asoc.adaptation_needed = 1;
6387				}
6388			}
6389		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6390			struct sctp_asconf_addr_param lstore, *fee;
6391			int lptype;
6392			struct sockaddr *lsa = NULL;
6393
6394#ifdef INET
6395			struct sctp_asconf_addrv4_param *fii;
6396
6397#endif
6398
6399			if (stcb->asoc.asconf_supported == 0) {
6400				return (-100);
6401			}
6402			if (plen > sizeof(lstore)) {
6403				return (-23);
6404			}
6405			phdr = sctp_get_next_param(m, offset,
6406			    (struct sctp_paramhdr *)&lstore,
6407			    min(plen, sizeof(lstore)));
6408			if (phdr == NULL) {
6409				return (-24);
6410			}
6411			fee = (struct sctp_asconf_addr_param *)phdr;
6412			lptype = ntohs(fee->addrp.ph.param_type);
6413			switch (lptype) {
6414#ifdef INET
6415			case SCTP_IPV4_ADDRESS:
6416				if (plen !=
6417				    sizeof(struct sctp_asconf_addrv4_param)) {
6418					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6419					    (int)sizeof(struct sctp_asconf_addrv4_param),
6420					    plen);
6421				} else {
6422					fii = (struct sctp_asconf_addrv4_param *)fee;
6423					sin.sin_addr.s_addr = fii->addrp.addr;
6424					lsa = (struct sockaddr *)&sin;
6425				}
6426				break;
6427#endif
6428#ifdef INET6
6429			case SCTP_IPV6_ADDRESS:
6430				if (plen !=
6431				    sizeof(struct sctp_asconf_addr_param)) {
6432					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6433					    (int)sizeof(struct sctp_asconf_addr_param),
6434					    plen);
6435				} else {
6436					memcpy(sin6.sin6_addr.s6_addr,
6437					    fee->addrp.addr,
6438					    sizeof(fee->addrp.addr));
6439					lsa = (struct sockaddr *)&sin6;
6440				}
6441				break;
6442#endif
6443			default:
6444				break;
6445			}
6446			if (lsa) {
6447				(void)sctp_set_primary_addr(stcb, sa, NULL);
6448			}
6449		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6450			stcb->asoc.peer_supports_nat = 1;
6451		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6452			/* Peer supports pr-sctp */
6453			peer_supports_prsctp = 1;
6454		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6455			/* A supported extension chunk */
6456			struct sctp_supported_chunk_types_param *pr_supported;
6457			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6458			int num_ent, i;
6459
6460			phdr = sctp_get_next_param(m, offset,
6461			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6462			if (phdr == NULL) {
6463				return (-25);
6464			}
6465			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6466			num_ent = plen - sizeof(struct sctp_paramhdr);
6467			for (i = 0; i < num_ent; i++) {
6468				switch (pr_supported->chunk_types[i]) {
6469				case SCTP_ASCONF:
6470					peer_supports_asconf = 1;
6471				case SCTP_ASCONF_ACK:
6472					peer_supports_asconf_ack = 1;
6473					break;
6474				case SCTP_FORWARD_CUM_TSN:
6475					peer_supports_prsctp = 1;
6476					break;
6477				case SCTP_PACKET_DROPPED:
6478					peer_supports_pktdrop = 1;
6479					break;
6480				case SCTP_NR_SELECTIVE_ACK:
6481					peer_supports_nrsack = 1;
6482					break;
6483				case SCTP_STREAM_RESET:
6484					peer_supports_reconfig = 1;
6485					break;
6486				case SCTP_AUTHENTICATION:
6487					peer_supports_auth = 1;
6488					break;
6489				default:
6490					/* one I have not learned yet */
6491					break;
6492
6493				}
6494			}
6495		} else if (ptype == SCTP_RANDOM) {
6496			if (plen > sizeof(random_store))
6497				break;
6498			if (got_random) {
6499				/* already processed a RANDOM */
6500				goto next_param;
6501			}
6502			phdr = sctp_get_next_param(m, offset,
6503			    (struct sctp_paramhdr *)random_store,
6504			    min(sizeof(random_store), plen));
6505			if (phdr == NULL)
6506				return (-26);
6507			p_random = (struct sctp_auth_random *)phdr;
6508			random_len = plen - sizeof(*p_random);
6509			/* enforce the random length */
6510			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6511				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6512				return (-27);
6513			}
6514			got_random = 1;
6515		} else if (ptype == SCTP_HMAC_LIST) {
6516			uint16_t num_hmacs;
6517			uint16_t i;
6518
6519			if (plen > sizeof(hmacs_store))
6520				break;
6521			if (got_hmacs) {
6522				/* already processed a HMAC list */
6523				goto next_param;
6524			}
6525			phdr = sctp_get_next_param(m, offset,
6526			    (struct sctp_paramhdr *)hmacs_store,
6527			    min(plen, sizeof(hmacs_store)));
6528			if (phdr == NULL)
6529				return (-28);
6530			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6531			hmacs_len = plen - sizeof(*hmacs);
6532			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6533			/* validate the hmac list */
6534			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6535				return (-29);
6536			}
6537			if (stcb->asoc.peer_hmacs != NULL)
6538				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6539			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6540			if (stcb->asoc.peer_hmacs != NULL) {
6541				for (i = 0; i < num_hmacs; i++) {
6542					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6543					    ntohs(hmacs->hmac_ids[i]));
6544				}
6545			}
6546			got_hmacs = 1;
6547		} else if (ptype == SCTP_CHUNK_LIST) {
6548			int i;
6549
6550			if (plen > sizeof(chunks_store))
6551				break;
6552			if (got_chklist) {
6553				/* already processed a Chunks list */
6554				goto next_param;
6555			}
6556			phdr = sctp_get_next_param(m, offset,
6557			    (struct sctp_paramhdr *)chunks_store,
6558			    min(plen, sizeof(chunks_store)));
6559			if (phdr == NULL)
6560				return (-30);
6561			chunks = (struct sctp_auth_chunk_list *)phdr;
6562			num_chunks = plen - sizeof(*chunks);
6563			if (stcb->asoc.peer_auth_chunks != NULL)
6564				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6565			else
6566				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6567			for (i = 0; i < num_chunks; i++) {
6568				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6569				    stcb->asoc.peer_auth_chunks);
6570				/* record asconf/asconf-ack if listed */
6571				if (chunks->chunk_types[i] == SCTP_ASCONF)
6572					saw_asconf = 1;
6573				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6574					saw_asconf_ack = 1;
6575
6576			}
6577			got_chklist = 1;
6578		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6579			    (ptype == SCTP_STATE_COOKIE) ||
6580			    (ptype == SCTP_UNRECOG_PARAM) ||
6581			    (ptype == SCTP_COOKIE_PRESERVE) ||
6582			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6583			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6584			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6585			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6586		    (ptype == SCTP_SUCCESS_REPORT)) {
6587			 /* don't care */ ;
6588		} else {
6589			if ((ptype & 0x8000) == 0x0000) {
6590				/*
6591				 * must stop processing the rest of the
6592				 * param's. Any report bits were handled
6593				 * with the call to
6594				 * sctp_arethere_unrecognized_parameters()
6595				 * when the INIT or INIT-ACK was first seen.
6596				 */
6597				break;
6598			}
6599		}
6600
6601next_param:
6602		offset += SCTP_SIZE32(plen);
6603		if (offset >= limit) {
6604			break;
6605		}
6606		phdr = sctp_get_next_param(m, offset, &parm_buf,
6607		    sizeof(parm_buf));
6608	}
6609	/* Now check to see if we need to purge any addresses */
6610	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6611		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6612		    SCTP_ADDR_NOT_IN_ASSOC) {
6613			/* This address has been removed from the asoc */
6614			/* remove and free it */
6615			stcb->asoc.numnets--;
6616			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6617			sctp_free_remote_addr(net);
6618			if (net == stcb->asoc.primary_destination) {
6619				stcb->asoc.primary_destination = NULL;
6620				sctp_select_primary_destination(stcb);
6621			}
6622		}
6623	}
6624	if ((stcb->asoc.ecn_supported == 1) &&
6625	    (peer_supports_ecn == 0)) {
6626		stcb->asoc.ecn_supported = 0;
6627	}
6628	if ((stcb->asoc.prsctp_supported == 1) &&
6629	    (peer_supports_prsctp == 0)) {
6630		stcb->asoc.prsctp_supported = 0;
6631	}
6632	if ((stcb->asoc.auth_supported == 1) &&
6633	    ((peer_supports_auth == 0) ||
6634	    (got_random == 0) || (got_hmacs == 0))) {
6635		stcb->asoc.auth_supported = 0;
6636	}
6637	if ((stcb->asoc.asconf_supported == 1) &&
6638	    ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) ||
6639	    (stcb->asoc.auth_supported == 0) ||
6640	    (saw_asconf == 0) || (saw_asconf_ack == 0))) {
6641		stcb->asoc.asconf_supported = 0;
6642	}
6643	if ((stcb->asoc.reconfig_supported == 1) &&
6644	    (peer_supports_reconfig == 0)) {
6645		stcb->asoc.reconfig_supported = 0;
6646	}
6647	if ((stcb->asoc.nrsack_supported == 1) &&
6648	    (peer_supports_nrsack == 0)) {
6649		stcb->asoc.nrsack_supported = 0;
6650	}
6651	if ((stcb->asoc.pktdrop_supported == 1) &&
6652	    (peer_supports_pktdrop == 0)) {
6653		stcb->asoc.pktdrop_supported = 0;
6654	}
6655	/* validate authentication required parameters */
6656	if ((peer_supports_auth == 0) && (got_chklist == 1)) {
6657		/* peer does not support auth but sent a chunks list? */
6658		return (-31);
6659	}
6660	if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) {
6661		/* peer supports asconf but not auth? */
6662		return (-32);
6663	} else if ((peer_supports_asconf == 1) &&
6664		    (peer_supports_auth == 1) &&
6665	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6666		return (-33);
6667	}
6668	/* concatenate the full random key */
6669	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6670	if (chunks != NULL) {
6671		keylen += sizeof(*chunks) + num_chunks;
6672	}
6673	new_key = sctp_alloc_key(keylen);
6674	if (new_key != NULL) {
6675		/* copy in the RANDOM */
6676		if (p_random != NULL) {
6677			keylen = sizeof(*p_random) + random_len;
6678			bcopy(p_random, new_key->key, keylen);
6679		}
6680		/* append in the AUTH chunks */
6681		if (chunks != NULL) {
6682			bcopy(chunks, new_key->key + keylen,
6683			    sizeof(*chunks) + num_chunks);
6684			keylen += sizeof(*chunks) + num_chunks;
6685		}
6686		/* append in the HMACs */
6687		if (hmacs != NULL) {
6688			bcopy(hmacs, new_key->key + keylen,
6689			    sizeof(*hmacs) + hmacs_len);
6690		}
6691	} else {
6692		/* failed to get memory for the key */
6693		return (-34);
6694	}
6695	if (stcb->asoc.authinfo.peer_random != NULL)
6696		sctp_free_key(stcb->asoc.authinfo.peer_random);
6697	stcb->asoc.authinfo.peer_random = new_key;
6698	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6699	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6700
6701	return (0);
6702}
6703
6704int
6705sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6706    struct sctp_nets *net)
6707{
6708	/* make sure the requested primary address exists in the assoc */
6709	if (net == NULL && sa)
6710		net = sctp_findnet(stcb, sa);
6711
6712	if (net == NULL) {
6713		/* didn't find the requested primary address! */
6714		return (-1);
6715	} else {
6716		/* set the primary address */
6717		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6718			/* Must be confirmed, so queue to set */
6719			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6720			return (0);
6721		}
6722		stcb->asoc.primary_destination = net;
6723		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6724			sctp_free_remote_addr(stcb->asoc.alternate);
6725			stcb->asoc.alternate = NULL;
6726		}
6727		net = TAILQ_FIRST(&stcb->asoc.nets);
6728		if (net != stcb->asoc.primary_destination) {
6729			/*
6730			 * first one on the list is NOT the primary
6731			 * sctp_cmpaddr() is much more efficient if the
6732			 * primary is the first on the list, make it so.
6733			 */
6734			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6735			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6736		}
6737		return (0);
6738	}
6739}
6740
6741int
6742sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6743{
6744	/*
6745	 * This function serves two purposes. It will see if a TAG can be
6746	 * re-used and return 1 for yes it is ok and 0 for don't use that
6747	 * tag. A secondary function it will do is purge out old tags that
6748	 * can be removed.
6749	 */
6750	struct sctpvtaghead *chain;
6751	struct sctp_tagblock *twait_block;
6752	struct sctpasochead *head;
6753	struct sctp_tcb *stcb;
6754	int i;
6755
6756	SCTP_INP_INFO_RLOCK();
6757	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6758	    SCTP_BASE_INFO(hashasocmark))];
6759	if (head == NULL) {
6760		/* invalid vtag */
6761		goto skip_vtag_check;
6762	}
6763	LIST_FOREACH(stcb, head, sctp_asocs) {
6764		/*
6765		 * We choose not to lock anything here. TCB's can't be
6766		 * removed since we have the read lock, so they can't be
6767		 * freed on us, same thing for the INP. I may be wrong with
6768		 * this assumption, but we will go with it for now :-)
6769		 */
6770		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6771			continue;
6772		}
6773		if (stcb->asoc.my_vtag == tag) {
6774			/* candidate */
6775			if (stcb->rport != rport) {
6776				continue;
6777			}
6778			if (stcb->sctp_ep->sctp_lport != lport) {
6779				continue;
6780			}
6781			/* Its a used tag set */
6782			SCTP_INP_INFO_RUNLOCK();
6783			return (0);
6784		}
6785	}
6786skip_vtag_check:
6787
6788	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6789	/* Now what about timed wait ? */
6790	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6791		/*
6792		 * Block(s) are present, lets see if we have this tag in the
6793		 * list
6794		 */
6795		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6796			if (twait_block->vtag_block[i].v_tag == 0) {
6797				/* not used */
6798				continue;
6799			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6800			    now->tv_sec) {
6801				/* Audit expires this guy */
6802				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6803				twait_block->vtag_block[i].v_tag = 0;
6804				twait_block->vtag_block[i].lport = 0;
6805				twait_block->vtag_block[i].rport = 0;
6806			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6807				    (twait_block->vtag_block[i].lport == lport) &&
6808			    (twait_block->vtag_block[i].rport == rport)) {
6809				/* Bad tag, sorry :< */
6810				SCTP_INP_INFO_RUNLOCK();
6811				return (0);
6812			}
6813		}
6814	}
6815	SCTP_INP_INFO_RUNLOCK();
6816	return (1);
6817}
6818
6819static void
6820sctp_drain_mbufs(struct sctp_tcb *stcb)
6821{
6822	/*
6823	 * We must hunt this association for MBUF's past the cumack (i.e.
6824	 * out of order data that we can renege on).
6825	 */
6826	struct sctp_association *asoc;
6827	struct sctp_tmit_chunk *chk, *nchk;
6828	uint32_t cumulative_tsn_p1;
6829	struct sctp_queued_to_read *ctl, *nctl;
6830	int cnt, strmat;
6831	uint32_t gap, i;
6832	int fnd = 0;
6833
6834	/* We look for anything larger than the cum-ack + 1 */
6835
6836	asoc = &stcb->asoc;
6837	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6838		/* none we can reneg on. */
6839		return;
6840	}
6841	SCTP_STAT_INCR(sctps_protocol_drains_done);
6842	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6843	cnt = 0;
6844	/* First look in the re-assembly queue */
6845	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6846		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6847			/* Yep it is above cum-ack */
6848			cnt++;
6849			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6850			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6851			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6852			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6853			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6854			if (chk->data) {
6855				sctp_m_freem(chk->data);
6856				chk->data = NULL;
6857			}
6858			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6859		}
6860	}
6861	/* Ok that was fun, now we will drain all the inbound streams? */
6862	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6863		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6864			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6865				/* Yep it is above cum-ack */
6866				cnt++;
6867				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6868				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6869				sctp_ucount_decr(asoc->cnt_on_all_streams);
6870				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6871				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6872				if (ctl->data) {
6873					sctp_m_freem(ctl->data);
6874					ctl->data = NULL;
6875				}
6876				sctp_free_remote_addr(ctl->whoFrom);
6877				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6878				SCTP_DECR_READQ_COUNT();
6879			}
6880		}
6881	}
6882	if (cnt) {
6883		/* We must back down to see what the new highest is */
6884		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6885			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6886			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6887				asoc->highest_tsn_inside_map = i;
6888				fnd = 1;
6889				break;
6890			}
6891		}
6892		if (!fnd) {
6893			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6894		}
6895		/*
6896		 * Question, should we go through the delivery queue? The
6897		 * only reason things are on here is the app not reading OR
6898		 * a p-d-api up. An attacker COULD send enough in to
6899		 * initiate the PD-API and then send a bunch of stuff to
6900		 * other streams... these would wind up on the delivery
6901		 * queue.. and then we would not get to them. But in order
6902		 * to do this I then have to back-track and un-deliver
6903		 * sequence numbers in streams.. el-yucko. I think for now
6904		 * we will NOT look at the delivery queue and leave it to be
6905		 * something to consider later. An alternative would be to
6906		 * abort the P-D-API with a notification and then deliver
6907		 * the data.... Or another method might be to keep track of
6908		 * how many times the situation occurs and if we see a
6909		 * possible attack underway just abort the association.
6910		 */
6911#ifdef SCTP_DEBUG
6912		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6913#endif
6914		/*
6915		 * Now do we need to find a new
6916		 * asoc->highest_tsn_inside_map?
6917		 */
6918		asoc->last_revoke_count = cnt;
6919		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6920		/* sa_ignore NO_NULL_CHK */
6921		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6922		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6923	}
6924	/*
6925	 * Another issue, in un-setting the TSN's in the mapping array we
6926	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6927	 * two things to occur. It may cause us to do extra work in checking
6928	 * for our mapping array movement. More importantly it may cause us
6929	 * to SACK every datagram. This may not be a bad thing though since
6930	 * we will recover once we get our cum-ack above and all this stuff
6931	 * we dumped recovered.
6932	 */
6933}
6934
6935void
6936sctp_drain()
6937{
6938	/*
6939	 * We must walk the PCB lists for ALL associations here. The system
6940	 * is LOW on MBUF's and needs help. This is where reneging will
6941	 * occur. We really hope this does NOT happen!
6942	 */
6943	VNET_ITERATOR_DECL(vnet_iter);
6944	VNET_LIST_RLOCK_NOSLEEP();
6945	VNET_FOREACH(vnet_iter) {
6946		CURVNET_SET(vnet_iter);
6947		struct sctp_inpcb *inp;
6948		struct sctp_tcb *stcb;
6949
6950		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6951		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6952#ifdef VIMAGE
6953			continue;
6954#else
6955			return;
6956#endif
6957		}
6958		SCTP_INP_INFO_RLOCK();
6959		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6960			/* For each endpoint */
6961			SCTP_INP_RLOCK(inp);
6962			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6963				/* For each association */
6964				SCTP_TCB_LOCK(stcb);
6965				sctp_drain_mbufs(stcb);
6966				SCTP_TCB_UNLOCK(stcb);
6967			}
6968			SCTP_INP_RUNLOCK(inp);
6969		}
6970		SCTP_INP_INFO_RUNLOCK();
6971		CURVNET_RESTORE();
6972	}
6973	VNET_LIST_RUNLOCK_NOSLEEP();
6974}
6975
6976/*
6977 * start a new iterator
6978 * iterates through all endpoints and associations based on the pcb_state
6979 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6980 * assocs and "ef" (optional) is executed when the iterator completes.
6981 * "inpf" (optional) is executed for each new endpoint as it is being
6982 * iterated through. inpe (optional) is called when the inp completes
6983 * its way through all the stcbs.
6984 */
6985int
6986sctp_initiate_iterator(inp_func inpf,
6987    asoc_func af,
6988    inp_func inpe,
6989    uint32_t pcb_state,
6990    uint32_t pcb_features,
6991    uint32_t asoc_state,
6992    void *argp,
6993    uint32_t argi,
6994    end_func ef,
6995    struct sctp_inpcb *s_inp,
6996    uint8_t chunk_output_off)
6997{
6998	struct sctp_iterator *it = NULL;
6999
7000	if (af == NULL) {
7001		return (-1);
7002	}
7003	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
7004	    SCTP_M_ITER);
7005	if (it == NULL) {
7006		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
7007		return (ENOMEM);
7008	}
7009	memset(it, 0, sizeof(*it));
7010	it->function_assoc = af;
7011	it->function_inp = inpf;
7012	if (inpf)
7013		it->done_current_ep = 0;
7014	else
7015		it->done_current_ep = 1;
7016	it->function_atend = ef;
7017	it->pointer = argp;
7018	it->val = argi;
7019	it->pcb_flags = pcb_state;
7020	it->pcb_features = pcb_features;
7021	it->asoc_state = asoc_state;
7022	it->function_inp_end = inpe;
7023	it->no_chunk_output = chunk_output_off;
7024	it->vn = curvnet;
7025	if (s_inp) {
7026		/* Assume lock is held here */
7027		it->inp = s_inp;
7028		SCTP_INP_INCR_REF(it->inp);
7029		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7030	} else {
7031		SCTP_INP_INFO_RLOCK();
7032		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7033		if (it->inp) {
7034			SCTP_INP_INCR_REF(it->inp);
7035		}
7036		SCTP_INP_INFO_RUNLOCK();
7037		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7038
7039	}
7040	SCTP_IPI_ITERATOR_WQ_LOCK();
7041
7042	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7043	if (sctp_it_ctl.iterator_running == 0) {
7044		sctp_wakeup_iterator();
7045	}
7046	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7047	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7048	return (0);
7049}
7050