sctp_pcb.c revision 281955
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 281955 2015-04-24 23:26:44Z hiren $");
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	net->flowtype = M_HASHTYPE_OPAQUE;
4057	if (netp) {
4058		*netp = net;
4059	}
4060	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4061	if (net->ro.ro_rt == NULL) {
4062		/* Since we have no route put it at the back */
4063		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4064	} else if (netfirst == NULL) {
4065		/* We are the first one in the pool. */
4066		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4067	} else if (netfirst->ro.ro_rt == NULL) {
4068		/*
4069		 * First one has NO route. Place this one ahead of the first
4070		 * one.
4071		 */
4072		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4073	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4074		/*
4075		 * This one has a different interface than the one at the
4076		 * top of the list. Place it ahead.
4077		 */
4078		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4079	} else {
4080		/*
4081		 * Ok we have the same interface as the first one. Move
4082		 * forward until we find either a) one with a NULL route...
4083		 * insert ahead of that b) one with a different ifp.. insert
4084		 * after that. c) end of the list.. insert at the tail.
4085		 */
4086		struct sctp_nets *netlook;
4087
4088		do {
4089			netlook = TAILQ_NEXT(netfirst, sctp_next);
4090			if (netlook == NULL) {
4091				/* End of the list */
4092				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4093				break;
4094			} else if (netlook->ro.ro_rt == NULL) {
4095				/* next one has NO route */
4096				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4097				break;
4098			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4099				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4100				    net, sctp_next);
4101				break;
4102			}
4103			/* Shift forward */
4104			netfirst = netlook;
4105		} while (netlook != NULL);
4106	}
4107
4108	/* got to have a primary set */
4109	if (stcb->asoc.primary_destination == 0) {
4110		stcb->asoc.primary_destination = net;
4111	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4112		    (net->ro.ro_rt) &&
4113	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4114		/* No route to current primary adopt new primary */
4115		stcb->asoc.primary_destination = net;
4116	}
4117	/* Validate primary is first */
4118	net = TAILQ_FIRST(&stcb->asoc.nets);
4119	if ((net != stcb->asoc.primary_destination) &&
4120	    (stcb->asoc.primary_destination)) {
4121		/*
4122		 * first one on the list is NOT the primary sctp_cmpaddr()
4123		 * is much more efficient if the primary is the first on the
4124		 * list, make it so.
4125		 */
4126		TAILQ_REMOVE(&stcb->asoc.nets,
4127		    stcb->asoc.primary_destination, sctp_next);
4128		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4129		    stcb->asoc.primary_destination, sctp_next);
4130	}
4131	return (0);
4132}
4133
4134
4135static uint32_t
4136sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4137{
4138	uint32_t id;
4139	struct sctpasochead *head;
4140	struct sctp_tcb *lstcb;
4141
4142	SCTP_INP_WLOCK(inp);
4143try_again:
4144	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4145		/* TSNH */
4146		SCTP_INP_WUNLOCK(inp);
4147		return (0);
4148	}
4149	/*
4150	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4151	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4152	 */
4153	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4154		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4155	}
4156	id = inp->sctp_associd_counter;
4157	inp->sctp_associd_counter++;
4158	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4159	if (lstcb) {
4160		goto try_again;
4161	}
4162	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4163	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4164	stcb->asoc.in_asocid_hash = 1;
4165	SCTP_INP_WUNLOCK(inp);
4166	return id;
4167}
4168
4169/*
4170 * allocate an association and add it to the endpoint. The caller must be
4171 * careful to add all additional addresses once they are know right away or
4172 * else the assoc will be may experience a blackout scenario.
4173 */
4174struct sctp_tcb *
4175sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4176    int *error, uint32_t override_tag, uint32_t vrf_id,
4177    struct thread *p
4178)
4179{
4180	/* note the p argument is only valid in unbound sockets */
4181
4182	struct sctp_tcb *stcb;
4183	struct sctp_association *asoc;
4184	struct sctpasochead *head;
4185	uint16_t rport;
4186	int err;
4187
4188	/*
4189	 * Assumption made here: Caller has done a
4190	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4191	 * address does not exist already.
4192	 */
4193	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4194		/* Hit max assoc, sorry no more */
4195		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4196		*error = ENOBUFS;
4197		return (NULL);
4198	}
4199	if (firstaddr == NULL) {
4200		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4201		*error = EINVAL;
4202		return (NULL);
4203	}
4204	SCTP_INP_RLOCK(inp);
4205	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4206	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4207	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4208		/*
4209		 * If its in the TCP pool, its NOT allowed to create an
4210		 * association. The parent listener needs to call
4211		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4212		 * off, or connected one does this.. its an error.
4213		 */
4214		SCTP_INP_RUNLOCK(inp);
4215		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4216		*error = EINVAL;
4217		return (NULL);
4218	}
4219	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4220	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4221		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4222		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4223			SCTP_INP_RUNLOCK(inp);
4224			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4225			*error = EINVAL;
4226			return (NULL);
4227		}
4228	}
4229	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4230#ifdef SCTP_DEBUG
4231	if (firstaddr) {
4232		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4233		switch (firstaddr->sa_family) {
4234#ifdef INET
4235		case AF_INET:
4236			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4237			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4238			break;
4239#endif
4240#ifdef INET6
4241		case AF_INET6:
4242			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4243			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4244			break;
4245#endif
4246		default:
4247			break;
4248		}
4249	} else {
4250		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4251	}
4252#endif				/* SCTP_DEBUG */
4253	switch (firstaddr->sa_family) {
4254#ifdef INET
4255	case AF_INET:
4256		{
4257			struct sockaddr_in *sin;
4258
4259			sin = (struct sockaddr_in *)firstaddr;
4260			if ((ntohs(sin->sin_port) == 0) ||
4261			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4262			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4263			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4264				/* Invalid address */
4265				SCTP_INP_RUNLOCK(inp);
4266				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4267				*error = EINVAL;
4268				return (NULL);
4269			}
4270			rport = sin->sin_port;
4271			break;
4272		}
4273#endif
4274#ifdef INET6
4275	case AF_INET6:
4276		{
4277			struct sockaddr_in6 *sin6;
4278
4279			sin6 = (struct sockaddr_in6 *)firstaddr;
4280			if ((ntohs(sin6->sin6_port) == 0) ||
4281			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4282			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4283				/* Invalid address */
4284				SCTP_INP_RUNLOCK(inp);
4285				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4286				*error = EINVAL;
4287				return (NULL);
4288			}
4289			rport = sin6->sin6_port;
4290			break;
4291		}
4292#endif
4293	default:
4294		/* not supported family type */
4295		SCTP_INP_RUNLOCK(inp);
4296		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4297		*error = EINVAL;
4298		return (NULL);
4299	}
4300	SCTP_INP_RUNLOCK(inp);
4301	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4302		/*
4303		 * If you have not performed a bind, then we need to do the
4304		 * ephemeral bind for you.
4305		 */
4306		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4307		    (struct sockaddr *)NULL,
4308		    (struct sctp_ifa *)NULL,
4309		    p
4310		    ))) {
4311			/* bind error, probably perm */
4312			*error = err;
4313			return (NULL);
4314		}
4315	}
4316	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4317	if (stcb == NULL) {
4318		/* out of memory? */
4319		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4320		*error = ENOMEM;
4321		return (NULL);
4322	}
4323	SCTP_INCR_ASOC_COUNT();
4324
4325	bzero(stcb, sizeof(*stcb));
4326	asoc = &stcb->asoc;
4327
4328	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4329	SCTP_TCB_LOCK_INIT(stcb);
4330	SCTP_TCB_SEND_LOCK_INIT(stcb);
4331	stcb->rport = rport;
4332	/* setup back pointer's */
4333	stcb->sctp_ep = inp;
4334	stcb->sctp_socket = inp->sctp_socket;
4335	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4336		/* failed */
4337		SCTP_TCB_LOCK_DESTROY(stcb);
4338		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4339		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4340		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4341		SCTP_DECR_ASOC_COUNT();
4342		*error = err;
4343		return (NULL);
4344	}
4345	/* and the port */
4346	SCTP_INP_INFO_WLOCK();
4347	SCTP_INP_WLOCK(inp);
4348	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4349		/* inpcb freed while alloc going on */
4350		SCTP_TCB_LOCK_DESTROY(stcb);
4351		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4352		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4353		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4354		SCTP_INP_WUNLOCK(inp);
4355		SCTP_INP_INFO_WUNLOCK();
4356		SCTP_DECR_ASOC_COUNT();
4357		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4358		*error = EINVAL;
4359		return (NULL);
4360	}
4361	SCTP_TCB_LOCK(stcb);
4362
4363	/* now that my_vtag is set, add it to the hash */
4364	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4365	/* put it in the bucket in the vtag hash of assoc's for the system */
4366	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4367	SCTP_INP_INFO_WUNLOCK();
4368
4369	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4370		/* failure.. memory error? */
4371		if (asoc->strmout) {
4372			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4373			asoc->strmout = NULL;
4374		}
4375		if (asoc->mapping_array) {
4376			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4377			asoc->mapping_array = NULL;
4378		}
4379		if (asoc->nr_mapping_array) {
4380			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4381			asoc->nr_mapping_array = NULL;
4382		}
4383		SCTP_DECR_ASOC_COUNT();
4384		SCTP_TCB_UNLOCK(stcb);
4385		SCTP_TCB_LOCK_DESTROY(stcb);
4386		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4387		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4388		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4389		SCTP_INP_WUNLOCK(inp);
4390		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4391		*error = ENOBUFS;
4392		return (NULL);
4393	}
4394	/* Init all the timers */
4395	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4396	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4397	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4398	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4399	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4400	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4401	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4402
4403	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4404	/* now file the port under the hash as well */
4405	if (inp->sctp_tcbhash != NULL) {
4406		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4407		    inp->sctp_hashmark)];
4408		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4409	}
4410	SCTP_INP_WUNLOCK(inp);
4411	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4412	return (stcb);
4413}
4414
4415
4416void
4417sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4418{
4419	struct sctp_association *asoc;
4420
4421	asoc = &stcb->asoc;
4422	asoc->numnets--;
4423	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4424	if (net == asoc->primary_destination) {
4425		/* Reset primary */
4426		struct sctp_nets *lnet;
4427
4428		lnet = TAILQ_FIRST(&asoc->nets);
4429		/*
4430		 * Mobility adaptation Ideally, if deleted destination is
4431		 * the primary, it becomes a fast retransmission trigger by
4432		 * the subsequent SET PRIMARY. (by micchie)
4433		 */
4434		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4435		    SCTP_MOBILITY_BASE) ||
4436		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4437		    SCTP_MOBILITY_FASTHANDOFF)) {
4438			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4439			if (asoc->deleted_primary != NULL) {
4440				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4441				goto out;
4442			}
4443			asoc->deleted_primary = net;
4444			atomic_add_int(&net->ref_count, 1);
4445			memset(&net->lastsa, 0, sizeof(net->lastsa));
4446			memset(&net->lastsv, 0, sizeof(net->lastsv));
4447			sctp_mobility_feature_on(stcb->sctp_ep,
4448			    SCTP_MOBILITY_PRIM_DELETED);
4449			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4450			    stcb->sctp_ep, stcb, NULL);
4451		}
4452out:
4453		/* Try to find a confirmed primary */
4454		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4455	}
4456	if (net == asoc->last_data_chunk_from) {
4457		/* Reset primary */
4458		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4459	}
4460	if (net == asoc->last_control_chunk_from) {
4461		/* Clear net */
4462		asoc->last_control_chunk_from = NULL;
4463	}
4464	if (net == stcb->asoc.alternate) {
4465		sctp_free_remote_addr(stcb->asoc.alternate);
4466		stcb->asoc.alternate = NULL;
4467	}
4468	sctp_free_remote_addr(net);
4469}
4470
4471/*
4472 * remove a remote endpoint address from an association, it will fail if the
4473 * address does not exist.
4474 */
4475int
4476sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4477{
4478	/*
4479	 * Here we need to remove a remote address. This is quite simple, we
4480	 * first find it in the list of address for the association
4481	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4482	 * on that item. Note we do not allow it to be removed if there are
4483	 * no other addresses.
4484	 */
4485	struct sctp_association *asoc;
4486	struct sctp_nets *net, *nnet;
4487
4488	asoc = &stcb->asoc;
4489
4490	/* locate the address */
4491	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4492		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4493			continue;
4494		}
4495		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4496		    remaddr)) {
4497			/* we found the guy */
4498			if (asoc->numnets < 2) {
4499				/* Must have at LEAST two remote addresses */
4500				return (-1);
4501			} else {
4502				sctp_remove_net(stcb, net);
4503				return (0);
4504			}
4505		}
4506	}
4507	/* not found. */
4508	return (-2);
4509}
4510
4511void
4512sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4513{
4514	struct sctpvtaghead *chain;
4515	struct sctp_tagblock *twait_block;
4516	int found = 0;
4517	int i;
4518
4519	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4520	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4521		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4522			if ((twait_block->vtag_block[i].v_tag == tag) &&
4523			    (twait_block->vtag_block[i].lport == lport) &&
4524			    (twait_block->vtag_block[i].rport == rport)) {
4525				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4526				twait_block->vtag_block[i].v_tag = 0;
4527				twait_block->vtag_block[i].lport = 0;
4528				twait_block->vtag_block[i].rport = 0;
4529				found = 1;
4530				break;
4531			}
4532		}
4533		if (found)
4534			break;
4535	}
4536}
4537
4538int
4539sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4540{
4541	struct sctpvtaghead *chain;
4542	struct sctp_tagblock *twait_block;
4543	int found = 0;
4544	int i;
4545
4546	SCTP_INP_INFO_WLOCK();
4547	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4548	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4549		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4550			if ((twait_block->vtag_block[i].v_tag == tag) &&
4551			    (twait_block->vtag_block[i].lport == lport) &&
4552			    (twait_block->vtag_block[i].rport == rport)) {
4553				found = 1;
4554				break;
4555			}
4556		}
4557		if (found)
4558			break;
4559	}
4560	SCTP_INP_INFO_WUNLOCK();
4561	return (found);
4562}
4563
4564
4565void
4566sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4567{
4568	struct sctpvtaghead *chain;
4569	struct sctp_tagblock *twait_block;
4570	struct timeval now;
4571	int set, i;
4572
4573	if (time == 0) {
4574		/* Its disabled */
4575		return;
4576	}
4577	(void)SCTP_GETTIME_TIMEVAL(&now);
4578	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4579	set = 0;
4580	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4581		/* Block(s) present, lets find space, and expire on the fly */
4582		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4583			if ((twait_block->vtag_block[i].v_tag == 0) &&
4584			    !set) {
4585				twait_block->vtag_block[i].tv_sec_at_expire =
4586				    now.tv_sec + time;
4587				twait_block->vtag_block[i].v_tag = tag;
4588				twait_block->vtag_block[i].lport = lport;
4589				twait_block->vtag_block[i].rport = rport;
4590				set = 1;
4591			} else if ((twait_block->vtag_block[i].v_tag) &&
4592			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4593				/* Audit expires this guy */
4594				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4595				twait_block->vtag_block[i].v_tag = 0;
4596				twait_block->vtag_block[i].lport = 0;
4597				twait_block->vtag_block[i].rport = 0;
4598				if (set == 0) {
4599					/* Reuse it for my new tag */
4600					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4601					twait_block->vtag_block[i].v_tag = tag;
4602					twait_block->vtag_block[i].lport = lport;
4603					twait_block->vtag_block[i].rport = rport;
4604					set = 1;
4605				}
4606			}
4607		}
4608		if (set) {
4609			/*
4610			 * We only do up to the block where we can place our
4611			 * tag for audits
4612			 */
4613			break;
4614		}
4615	}
4616	/* Need to add a new block to chain */
4617	if (!set) {
4618		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4619		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4620		if (twait_block == NULL) {
4621#ifdef INVARIANTS
4622			panic("Can not alloc tagblock");
4623#endif
4624			return;
4625		}
4626		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4627		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4628		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4629		twait_block->vtag_block[0].v_tag = tag;
4630		twait_block->vtag_block[0].lport = lport;
4631		twait_block->vtag_block[0].rport = rport;
4632	}
4633}
4634
4635
4636
4637/*-
4638 * Free the association after un-hashing the remote port. This
4639 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4640 * expect that the input to this function IS a locked TCB.
4641 * It will return 0, if it did NOT destroy the association (instead
4642 * it unlocks it. It will return NON-zero if it either destroyed the
4643 * association OR the association is already destroyed.
4644 */
4645int
4646sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4647{
4648	int i;
4649	struct sctp_association *asoc;
4650	struct sctp_nets *net, *nnet;
4651	struct sctp_laddr *laddr, *naddr;
4652	struct sctp_tmit_chunk *chk, *nchk;
4653	struct sctp_asconf_addr *aparam, *naparam;
4654	struct sctp_asconf_ack *aack, *naack;
4655	struct sctp_stream_reset_list *strrst, *nstrrst;
4656	struct sctp_queued_to_read *sq, *nsq;
4657	struct sctp_stream_queue_pending *sp, *nsp;
4658	sctp_sharedkey_t *shared_key, *nshared_key;
4659	struct socket *so;
4660
4661	/* first, lets purge the entry from the hash table. */
4662
4663#ifdef SCTP_LOG_CLOSING
4664	sctp_log_closing(inp, stcb, 6);
4665#endif
4666	if (stcb->asoc.state == 0) {
4667#ifdef SCTP_LOG_CLOSING
4668		sctp_log_closing(inp, NULL, 7);
4669#endif
4670		/* there is no asoc, really TSNH :-0 */
4671		return (1);
4672	}
4673	if (stcb->asoc.alternate) {
4674		sctp_free_remote_addr(stcb->asoc.alternate);
4675		stcb->asoc.alternate = NULL;
4676	}
4677	/* TEMP CODE */
4678	if (stcb->freed_from_where == 0) {
4679		/* Only record the first place free happened from */
4680		stcb->freed_from_where = from_location;
4681	}
4682	/* TEMP CODE */
4683
4684	asoc = &stcb->asoc;
4685	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4686	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4687		/* nothing around */
4688		so = NULL;
4689	else
4690		so = inp->sctp_socket;
4691
4692	/*
4693	 * We used timer based freeing if a reader or writer is in the way.
4694	 * So we first check if we are actually being called from a timer,
4695	 * if so we abort early if a reader or writer is still in the way.
4696	 */
4697	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4698	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4699		/*
4700		 * is it the timer driving us? if so are the reader/writers
4701		 * gone?
4702		 */
4703		if (stcb->asoc.refcnt) {
4704			/* nope, reader or writer in the way */
4705			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4706			/* no asoc destroyed */
4707			SCTP_TCB_UNLOCK(stcb);
4708#ifdef SCTP_LOG_CLOSING
4709			sctp_log_closing(inp, stcb, 8);
4710#endif
4711			return (0);
4712		}
4713	}
4714	/* now clean up any other timers */
4715	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4716	asoc->dack_timer.self = NULL;
4717	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4718	/*-
4719	 * For stream reset we don't blast this unless
4720	 * it is a str-reset timer, it might be the
4721	 * free-asoc timer which we DON'T want to
4722	 * disturb.
4723	 */
4724	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4725		asoc->strreset_timer.self = NULL;
4726	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4727	asoc->asconf_timer.self = NULL;
4728	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4729	asoc->autoclose_timer.self = NULL;
4730	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4731	asoc->shut_guard_timer.self = NULL;
4732	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4733	asoc->delayed_event_timer.self = NULL;
4734	/* Mobility adaptation */
4735	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4736	asoc->delete_prim_timer.self = NULL;
4737	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4738		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4739		net->rxt_timer.self = NULL;
4740		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4741		net->pmtu_timer.self = NULL;
4742		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4743		net->hb_timer.self = NULL;
4744	}
4745	/* Now the read queue needs to be cleaned up (only once) */
4746	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4747		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4748		SCTP_INP_READ_LOCK(inp);
4749		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4750			if (sq->stcb == stcb) {
4751				sq->do_not_ref_stcb = 1;
4752				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4753				/*
4754				 * If there is no end, there never will be
4755				 * now.
4756				 */
4757				if (sq->end_added == 0) {
4758					/* Held for PD-API clear that. */
4759					sq->pdapi_aborted = 1;
4760					sq->held_length = 0;
4761					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4762						/*
4763						 * Need to add a PD-API
4764						 * aborted indication.
4765						 * Setting the control_pdapi
4766						 * assures that it will be
4767						 * added right after this
4768						 * msg.
4769						 */
4770						uint32_t strseq;
4771
4772						stcb->asoc.control_pdapi = sq;
4773						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4774						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4775						    stcb,
4776						    SCTP_PARTIAL_DELIVERY_ABORTED,
4777						    (void *)&strseq,
4778						    SCTP_SO_LOCKED);
4779						stcb->asoc.control_pdapi = NULL;
4780					}
4781				}
4782				/* Add an end to wake them */
4783				sq->end_added = 1;
4784			}
4785		}
4786		SCTP_INP_READ_UNLOCK(inp);
4787		if (stcb->block_entry) {
4788			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4789			stcb->block_entry->error = ECONNRESET;
4790			stcb->block_entry = NULL;
4791		}
4792	}
4793	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4794		/*
4795		 * Someone holds a reference OR the socket is unaccepted
4796		 * yet.
4797		 */
4798		if ((stcb->asoc.refcnt) ||
4799		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4800		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4801			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4802			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4803		}
4804		SCTP_TCB_UNLOCK(stcb);
4805		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4806		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4807			/* nothing around */
4808			so = NULL;
4809		if (so) {
4810			/* Wake any reader/writers */
4811			sctp_sorwakeup(inp, so);
4812			sctp_sowwakeup(inp, so);
4813		}
4814#ifdef SCTP_LOG_CLOSING
4815		sctp_log_closing(inp, stcb, 9);
4816#endif
4817		/* no asoc destroyed */
4818		return (0);
4819	}
4820#ifdef SCTP_LOG_CLOSING
4821	sctp_log_closing(inp, stcb, 10);
4822#endif
4823	/*
4824	 * When I reach here, no others want to kill the assoc yet.. and I
4825	 * own the lock. Now its possible an abort comes in when I do the
4826	 * lock exchange below to grab all the locks to do the final take
4827	 * out. to prevent this we increment the count, which will start a
4828	 * timer and blow out above thus assuring us that we hold exclusive
4829	 * killing of the asoc. Note that after getting back the TCB lock we
4830	 * will go ahead and increment the counter back up and stop any
4831	 * timer a passing stranger may have started :-S
4832	 */
4833	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4834		atomic_add_int(&stcb->asoc.refcnt, 1);
4835
4836		SCTP_TCB_UNLOCK(stcb);
4837		SCTP_INP_INFO_WLOCK();
4838		SCTP_INP_WLOCK(inp);
4839		SCTP_TCB_LOCK(stcb);
4840	}
4841	/* Double check the GONE flag */
4842	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4843	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4844		/* nothing around */
4845		so = NULL;
4846
4847	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4848	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4849		/*
4850		 * For TCP type we need special handling when we are
4851		 * connected. We also include the peel'ed off ones to.
4852		 */
4853		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4854			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4855			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4856			if (so) {
4857				SOCK_LOCK(so);
4858				if (so->so_rcv.sb_cc == 0) {
4859					so->so_state &= ~(SS_ISCONNECTING |
4860					    SS_ISDISCONNECTING |
4861					    SS_ISCONFIRMING |
4862					    SS_ISCONNECTED);
4863				}
4864				socantrcvmore_locked(so);
4865				sctp_sowwakeup(inp, so);
4866				sctp_sorwakeup(inp, so);
4867				SCTP_SOWAKEUP(so);
4868			}
4869		}
4870	}
4871	/*
4872	 * Make it invalid too, that way if its about to run it will abort
4873	 * and return.
4874	 */
4875	/* re-increment the lock */
4876	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4877		atomic_add_int(&stcb->asoc.refcnt, -1);
4878	}
4879	if (stcb->asoc.refcnt) {
4880		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4881		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4882		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4883			SCTP_INP_INFO_WUNLOCK();
4884			SCTP_INP_WUNLOCK(inp);
4885		}
4886		SCTP_TCB_UNLOCK(stcb);
4887		return (0);
4888	}
4889	asoc->state = 0;
4890	if (inp->sctp_tcbhash) {
4891		LIST_REMOVE(stcb, sctp_tcbhash);
4892	}
4893	if (stcb->asoc.in_asocid_hash) {
4894		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4895	}
4896	/* Now lets remove it from the list of ALL associations in the EP */
4897	LIST_REMOVE(stcb, sctp_tcblist);
4898	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4899		SCTP_INP_INCR_REF(inp);
4900		SCTP_INP_WUNLOCK(inp);
4901	}
4902	/* pull from vtag hash */
4903	LIST_REMOVE(stcb, sctp_asocs);
4904	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4905	    inp->sctp_lport, stcb->rport);
4906
4907	/*
4908	 * Now restop the timers to be sure this is paranoia at is finest!
4909	 */
4910	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4911	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4912	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4913	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4914	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4915	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4916	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4917	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4918		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4919		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4920		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4921	}
4922
4923	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4924	/*
4925	 * The chunk lists and such SHOULD be empty but we check them just
4926	 * in case.
4927	 */
4928	/* anything on the wheel needs to be removed */
4929	for (i = 0; i < asoc->streamoutcnt; i++) {
4930		struct sctp_stream_out *outs;
4931
4932		outs = &asoc->strmout[i];
4933		/* now clean up any chunks here */
4934		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4935			TAILQ_REMOVE(&outs->outqueue, sp, next);
4936			sctp_free_spbufspace(stcb, asoc, sp);
4937			if (sp->data) {
4938				if (so) {
4939					/* Still an open socket - report */
4940					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4941					    0, (void *)sp, SCTP_SO_LOCKED);
4942				}
4943				if (sp->data) {
4944					sctp_m_freem(sp->data);
4945					sp->data = NULL;
4946					sp->tail_mbuf = NULL;
4947					sp->length = 0;
4948				}
4949			}
4950			if (sp->net) {
4951				sctp_free_remote_addr(sp->net);
4952				sp->net = NULL;
4953			}
4954			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4955		}
4956	}
4957	/* sa_ignore FREED_MEMORY */
4958	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4959		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4960		SCTP_FREE(strrst, SCTP_M_STRESET);
4961	}
4962	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4963		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4964		if (sq->data) {
4965			sctp_m_freem(sq->data);
4966			sq->data = NULL;
4967		}
4968		sctp_free_remote_addr(sq->whoFrom);
4969		sq->whoFrom = NULL;
4970		sq->stcb = NULL;
4971		/* Free the ctl entry */
4972		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4973		SCTP_DECR_READQ_COUNT();
4974		/* sa_ignore FREED_MEMORY */
4975	}
4976	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4977		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4978		if (chk->data) {
4979			sctp_m_freem(chk->data);
4980			chk->data = NULL;
4981		}
4982		if (chk->holds_key_ref)
4983			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4984		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4985		SCTP_DECR_CHK_COUNT();
4986		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4987		asoc->free_chunk_cnt--;
4988		/* sa_ignore FREED_MEMORY */
4989	}
4990	/* pending send queue SHOULD be empty */
4991	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4992		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4993			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4994#ifdef INVARIANTS
4995		} else {
4996			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4997#endif
4998		}
4999		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
5000		if (chk->data) {
5001			if (so) {
5002				/* Still a socket? */
5003				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
5004				    0, chk, SCTP_SO_LOCKED);
5005			}
5006			if (chk->data) {
5007				sctp_m_freem(chk->data);
5008				chk->data = NULL;
5009			}
5010		}
5011		if (chk->holds_key_ref)
5012			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5013		if (chk->whoTo) {
5014			sctp_free_remote_addr(chk->whoTo);
5015			chk->whoTo = NULL;
5016		}
5017		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5018		SCTP_DECR_CHK_COUNT();
5019		/* sa_ignore FREED_MEMORY */
5020	}
5021	/* sent queue SHOULD be empty */
5022	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5023		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5024			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
5025				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
5026#ifdef INVARIANTS
5027			} else {
5028				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
5029#endif
5030			}
5031		}
5032		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5033		if (chk->data) {
5034			if (so) {
5035				/* Still a socket? */
5036				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5037				    0, chk, SCTP_SO_LOCKED);
5038			}
5039			if (chk->data) {
5040				sctp_m_freem(chk->data);
5041				chk->data = NULL;
5042			}
5043		}
5044		if (chk->holds_key_ref)
5045			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5046		sctp_free_remote_addr(chk->whoTo);
5047		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5048		SCTP_DECR_CHK_COUNT();
5049		/* sa_ignore FREED_MEMORY */
5050	}
5051#ifdef INVARIANTS
5052	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5053		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5054			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5055		}
5056	}
5057#endif
5058	/* control queue MAY not be empty */
5059	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5060		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5061		if (chk->data) {
5062			sctp_m_freem(chk->data);
5063			chk->data = NULL;
5064		}
5065		if (chk->holds_key_ref)
5066			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5067		sctp_free_remote_addr(chk->whoTo);
5068		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5069		SCTP_DECR_CHK_COUNT();
5070		/* sa_ignore FREED_MEMORY */
5071	}
5072	/* ASCONF queue MAY not be empty */
5073	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5074		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5075		if (chk->data) {
5076			sctp_m_freem(chk->data);
5077			chk->data = NULL;
5078		}
5079		if (chk->holds_key_ref)
5080			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5081		sctp_free_remote_addr(chk->whoTo);
5082		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5083		SCTP_DECR_CHK_COUNT();
5084		/* sa_ignore FREED_MEMORY */
5085	}
5086	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5087		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5088		if (chk->data) {
5089			sctp_m_freem(chk->data);
5090			chk->data = NULL;
5091		}
5092		if (chk->holds_key_ref)
5093			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5094		sctp_free_remote_addr(chk->whoTo);
5095		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5096		SCTP_DECR_CHK_COUNT();
5097		/* sa_ignore FREED_MEMORY */
5098	}
5099
5100	if (asoc->mapping_array) {
5101		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5102		asoc->mapping_array = NULL;
5103	}
5104	if (asoc->nr_mapping_array) {
5105		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5106		asoc->nr_mapping_array = NULL;
5107	}
5108	/* the stream outs */
5109	if (asoc->strmout) {
5110		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5111		asoc->strmout = NULL;
5112	}
5113	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5114	if (asoc->strmin) {
5115		struct sctp_queued_to_read *ctl, *nctl;
5116
5117		for (i = 0; i < asoc->streamincnt; i++) {
5118			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5119				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5120				sctp_free_remote_addr(ctl->whoFrom);
5121				if (ctl->data) {
5122					sctp_m_freem(ctl->data);
5123					ctl->data = NULL;
5124				}
5125				/*
5126				 * We don't free the address here since all
5127				 * the net's were freed above.
5128				 */
5129				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5130				SCTP_DECR_READQ_COUNT();
5131			}
5132		}
5133		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5134		asoc->strmin = NULL;
5135	}
5136	asoc->streamincnt = 0;
5137	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5138#ifdef INVARIANTS
5139		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5140			panic("no net's left alloc'ed, or list points to itself");
5141		}
5142#endif
5143		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5144		sctp_free_remote_addr(net);
5145	}
5146	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5147		/* sa_ignore FREED_MEMORY */
5148		sctp_remove_laddr(laddr);
5149	}
5150
5151	/* pending asconf (address) parameters */
5152	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5153		/* sa_ignore FREED_MEMORY */
5154		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5155		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5156	}
5157	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5158		/* sa_ignore FREED_MEMORY */
5159		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5160		if (aack->data != NULL) {
5161			sctp_m_freem(aack->data);
5162		}
5163		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5164	}
5165	/* clean up auth stuff */
5166	if (asoc->local_hmacs)
5167		sctp_free_hmaclist(asoc->local_hmacs);
5168	if (asoc->peer_hmacs)
5169		sctp_free_hmaclist(asoc->peer_hmacs);
5170
5171	if (asoc->local_auth_chunks)
5172		sctp_free_chunklist(asoc->local_auth_chunks);
5173	if (asoc->peer_auth_chunks)
5174		sctp_free_chunklist(asoc->peer_auth_chunks);
5175
5176	sctp_free_authinfo(&asoc->authinfo);
5177
5178	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5179		LIST_REMOVE(shared_key, next);
5180		sctp_free_sharedkey(shared_key);
5181		/* sa_ignore FREED_MEMORY */
5182	}
5183
5184	/* Insert new items here :> */
5185
5186	/* Get rid of LOCK */
5187	SCTP_TCB_UNLOCK(stcb);
5188	SCTP_TCB_LOCK_DESTROY(stcb);
5189	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5190	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5191		SCTP_INP_INFO_WUNLOCK();
5192		SCTP_INP_RLOCK(inp);
5193	}
5194#ifdef SCTP_TRACK_FREED_ASOCS
5195	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5196		/* now clean up the tasoc itself */
5197		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5198		SCTP_DECR_ASOC_COUNT();
5199	} else {
5200		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5201	}
5202#else
5203	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5204	SCTP_DECR_ASOC_COUNT();
5205#endif
5206	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5207		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5208			/*
5209			 * If its NOT the inp_free calling us AND sctp_close
5210			 * as been called, we call back...
5211			 */
5212			SCTP_INP_RUNLOCK(inp);
5213			/*
5214			 * This will start the kill timer (if we are the
5215			 * last one) since we hold an increment yet. But
5216			 * this is the only safe way to do this since
5217			 * otherwise if the socket closes at the same time
5218			 * we are here we might collide in the cleanup.
5219			 */
5220			sctp_inpcb_free(inp,
5221			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5222			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5223			SCTP_INP_DECR_REF(inp);
5224			goto out_of;
5225		} else {
5226			/* The socket is still open. */
5227			SCTP_INP_DECR_REF(inp);
5228		}
5229	}
5230	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5231		SCTP_INP_RUNLOCK(inp);
5232	}
5233out_of:
5234	/* destroyed the asoc */
5235#ifdef SCTP_LOG_CLOSING
5236	sctp_log_closing(inp, NULL, 11);
5237#endif
5238	return (1);
5239}
5240
5241
5242
5243/*
5244 * determine if a destination is "reachable" based upon the addresses bound
5245 * to the current endpoint (e.g. only v4 or v6 currently bound)
5246 */
5247/*
5248 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5249 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5250 * types bound as its endpoint
5251 */
5252int
5253sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5254{
5255	struct sctp_inpcb *inp;
5256	int answer;
5257
5258	/*
5259	 * No locks here, the TCB, in all cases is already locked and an
5260	 * assoc is up. There is either a INP lock by the caller applied (in
5261	 * asconf case when deleting an address) or NOT in the HB case,
5262	 * however if HB then the INP increment is up and the INP will not
5263	 * be removed (on top of the fact that we have a TCB lock). So we
5264	 * only want to read the sctp_flags, which is either bound-all or
5265	 * not.. no protection needed since once an assoc is up you can't be
5266	 * changing your binding.
5267	 */
5268	inp = stcb->sctp_ep;
5269	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5270		/* if bound all, destination is not restricted */
5271		/*
5272		 * RRS: Question during lock work: Is this correct? If you
5273		 * are bound-all you still might need to obey the V4--V6
5274		 * flags??? IMO this bound-all stuff needs to be removed!
5275		 */
5276		return (1);
5277	}
5278	/* NOTE: all "scope" checks are done when local addresses are added */
5279	switch (destaddr->sa_family) {
5280#ifdef INET6
5281	case AF_INET6:
5282		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5283		break;
5284#endif
5285#ifdef INET
5286	case AF_INET:
5287		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5288		break;
5289#endif
5290	default:
5291		/* invalid family, so it's unreachable */
5292		answer = 0;
5293		break;
5294	}
5295	return (answer);
5296}
5297
5298/*
5299 * update the inp_vflags on an endpoint
5300 */
5301static void
5302sctp_update_ep_vflag(struct sctp_inpcb *inp)
5303{
5304	struct sctp_laddr *laddr;
5305
5306	/* first clear the flag */
5307	inp->ip_inp.inp.inp_vflag = 0;
5308	/* set the flag based on addresses on the ep list */
5309	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5310		if (laddr->ifa == NULL) {
5311			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5312			    __FUNCTION__);
5313			continue;
5314		}
5315		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5316			continue;
5317		}
5318		switch (laddr->ifa->address.sa.sa_family) {
5319#ifdef INET6
5320		case AF_INET6:
5321			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5322			break;
5323#endif
5324#ifdef INET
5325		case AF_INET:
5326			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5327			break;
5328#endif
5329		default:
5330			break;
5331		}
5332	}
5333}
5334
5335/*
5336 * Add the address to the endpoint local address list There is nothing to be
5337 * done if we are bound to all addresses
5338 */
5339void
5340sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5341{
5342	struct sctp_laddr *laddr;
5343	int fnd, error = 0;
5344
5345	fnd = 0;
5346
5347	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5348		/* You are already bound to all. You have it already */
5349		return;
5350	}
5351#ifdef INET6
5352	if (ifa->address.sa.sa_family == AF_INET6) {
5353		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5354			/* Can't bind a non-useable addr. */
5355			return;
5356		}
5357	}
5358#endif
5359	/* first, is it already present? */
5360	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5361		if (laddr->ifa == ifa) {
5362			fnd = 1;
5363			break;
5364		}
5365	}
5366
5367	if (fnd == 0) {
5368		/* Not in the ep list */
5369		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5370		if (error != 0)
5371			return;
5372		inp->laddr_count++;
5373		/* update inp_vflag flags */
5374		switch (ifa->address.sa.sa_family) {
5375#ifdef INET6
5376		case AF_INET6:
5377			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5378			break;
5379#endif
5380#ifdef INET
5381		case AF_INET:
5382			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5383			break;
5384#endif
5385		default:
5386			break;
5387		}
5388	}
5389	return;
5390}
5391
5392
5393/*
5394 * select a new (hopefully reachable) destination net (should only be used
5395 * when we deleted an ep addr that is the only usable source address to reach
5396 * the destination net)
5397 */
5398static void
5399sctp_select_primary_destination(struct sctp_tcb *stcb)
5400{
5401	struct sctp_nets *net;
5402
5403	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5404		/* for now, we'll just pick the first reachable one we find */
5405		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5406			continue;
5407		if (sctp_destination_is_reachable(stcb,
5408		    (struct sockaddr *)&net->ro._l_addr)) {
5409			/* found a reachable destination */
5410			stcb->asoc.primary_destination = net;
5411		}
5412	}
5413	/* I can't there from here! ...we're gonna die shortly... */
5414}
5415
5416
5417/*
5418 * Delete the address from the endpoint local address list There is nothing
5419 * to be done if we are bound to all addresses
5420 */
5421void
5422sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5423{
5424	struct sctp_laddr *laddr;
5425	int fnd;
5426
5427	fnd = 0;
5428	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5429		/* You are already bound to all. You have it already */
5430		return;
5431	}
5432	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5433		if (laddr->ifa == ifa) {
5434			fnd = 1;
5435			break;
5436		}
5437	}
5438	if (fnd && (inp->laddr_count < 2)) {
5439		/* can't delete unless there are at LEAST 2 addresses */
5440		return;
5441	}
5442	if (fnd) {
5443		/*
5444		 * clean up any use of this address go through our
5445		 * associations and clear any last_used_address that match
5446		 * this one for each assoc, see if a new primary_destination
5447		 * is needed
5448		 */
5449		struct sctp_tcb *stcb;
5450
5451		/* clean up "next_addr_touse" */
5452		if (inp->next_addr_touse == laddr)
5453			/* delete this address */
5454			inp->next_addr_touse = NULL;
5455
5456		/* clean up "last_used_address" */
5457		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5458			struct sctp_nets *net;
5459
5460			SCTP_TCB_LOCK(stcb);
5461			if (stcb->asoc.last_used_address == laddr)
5462				/* delete this address */
5463				stcb->asoc.last_used_address = NULL;
5464			/*
5465			 * Now spin through all the nets and purge any ref
5466			 * to laddr
5467			 */
5468			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5469				if (net->ro._s_addr &&
5470				    (net->ro._s_addr->ifa == laddr->ifa)) {
5471					/* Yep, purge src address selected */
5472					sctp_rtentry_t *rt;
5473
5474					/* delete this address if cached */
5475					rt = net->ro.ro_rt;
5476					if (rt != NULL) {
5477						RTFREE(rt);
5478						net->ro.ro_rt = NULL;
5479					}
5480					sctp_free_ifa(net->ro._s_addr);
5481					net->ro._s_addr = NULL;
5482					net->src_addr_selected = 0;
5483				}
5484			}
5485			SCTP_TCB_UNLOCK(stcb);
5486		}		/* for each tcb */
5487		/* remove it from the ep list */
5488		sctp_remove_laddr(laddr);
5489		inp->laddr_count--;
5490		/* update inp_vflag flags */
5491		sctp_update_ep_vflag(inp);
5492	}
5493	return;
5494}
5495
5496/*
5497 * Add the address to the TCB local address restricted list.
5498 * This is a "pending" address list (eg. addresses waiting for an
5499 * ASCONF-ACK response) and cannot be used as a valid source address.
5500 */
5501void
5502sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5503{
5504	struct sctp_laddr *laddr;
5505	struct sctpladdr *list;
5506
5507	/*
5508	 * Assumes TCB is locked.. and possibly the INP. May need to
5509	 * confirm/fix that if we need it and is not the case.
5510	 */
5511	list = &stcb->asoc.sctp_restricted_addrs;
5512
5513#ifdef INET6
5514	if (ifa->address.sa.sa_family == AF_INET6) {
5515		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5516			/* Can't bind a non-existent addr. */
5517			return;
5518		}
5519	}
5520#endif
5521	/* does the address already exist? */
5522	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5523		if (laddr->ifa == ifa) {
5524			return;
5525		}
5526	}
5527
5528	/* add to the list */
5529	(void)sctp_insert_laddr(list, ifa, 0);
5530	return;
5531}
5532
5533/*
5534 * insert an laddr entry with the given ifa for the desired list
5535 */
5536int
5537sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5538{
5539	struct sctp_laddr *laddr;
5540
5541	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5542	if (laddr == NULL) {
5543		/* out of memory? */
5544		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5545		return (EINVAL);
5546	}
5547	SCTP_INCR_LADDR_COUNT();
5548	bzero(laddr, sizeof(*laddr));
5549	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5550	laddr->ifa = ifa;
5551	laddr->action = act;
5552	atomic_add_int(&ifa->refcount, 1);
5553	/* insert it */
5554	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5555
5556	return (0);
5557}
5558
5559/*
5560 * Remove an laddr entry from the local address list (on an assoc)
5561 */
5562void
5563sctp_remove_laddr(struct sctp_laddr *laddr)
5564{
5565
5566	/* remove from the list */
5567	LIST_REMOVE(laddr, sctp_nxt_addr);
5568	sctp_free_ifa(laddr->ifa);
5569	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5570	SCTP_DECR_LADDR_COUNT();
5571}
5572
5573/*
5574 * Remove a local address from the TCB local address restricted list
5575 */
5576void
5577sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5578{
5579	struct sctp_inpcb *inp;
5580	struct sctp_laddr *laddr;
5581
5582	/*
5583	 * This is called by asconf work. It is assumed that a) The TCB is
5584	 * locked and b) The INP is locked. This is true in as much as I can
5585	 * trace through the entry asconf code where I did these locks.
5586	 * Again, the ASCONF code is a bit different in that it does lock
5587	 * the INP during its work often times. This must be since we don't
5588	 * want other proc's looking up things while what they are looking
5589	 * up is changing :-D
5590	 */
5591
5592	inp = stcb->sctp_ep;
5593	/* if subset bound and don't allow ASCONF's, can't delete last */
5594	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5595	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5596		if (stcb->sctp_ep->laddr_count < 2) {
5597			/* can't delete last address */
5598			return;
5599		}
5600	}
5601	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5602		/* remove the address if it exists */
5603		if (laddr->ifa == NULL)
5604			continue;
5605		if (laddr->ifa == ifa) {
5606			sctp_remove_laddr(laddr);
5607			return;
5608		}
5609	}
5610
5611	/* address not found! */
5612	return;
5613}
5614
5615/*
5616 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5617 */
5618/* sysctl */
5619static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5620static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5621
5622
5623
5624#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5625struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5626int *sctp_cpuarry = NULL;
5627void
5628sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5629{
5630	/* Queue a packet to a processor for the specified core */
5631	struct sctp_mcore_queue *qent;
5632	struct sctp_mcore_ctrl *wkq;
5633	int need_wake = 0;
5634
5635	if (sctp_mcore_workers == NULL) {
5636		/* Something went way bad during setup */
5637		sctp_input_with_port(m, off, 0);
5638		return;
5639	}
5640	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5641	    (sizeof(struct sctp_mcore_queue)),
5642	    SCTP_M_MCORE);
5643	if (qent == NULL) {
5644		/* This is trouble  */
5645		sctp_input_with_port(m, off, 0);
5646		return;
5647	}
5648	qent->vn = curvnet;
5649	qent->m = m;
5650	qent->off = off;
5651	qent->v6 = 0;
5652	wkq = &sctp_mcore_workers[cpu_to_use];
5653	SCTP_MCORE_QLOCK(wkq);
5654
5655	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5656	if (wkq->running == 0) {
5657		need_wake = 1;
5658	}
5659	SCTP_MCORE_QUNLOCK(wkq);
5660	if (need_wake) {
5661		wakeup(&wkq->running);
5662	}
5663}
5664
5665static void
5666sctp_mcore_thread(void *arg)
5667{
5668
5669	struct sctp_mcore_ctrl *wkq;
5670	struct sctp_mcore_queue *qent;
5671
5672	wkq = (struct sctp_mcore_ctrl *)arg;
5673	struct mbuf *m;
5674	int off, v6;
5675
5676	/* Wait for first tickle */
5677	SCTP_MCORE_LOCK(wkq);
5678	wkq->running = 0;
5679	msleep(&wkq->running,
5680	    &wkq->core_mtx,
5681	    0, "wait for pkt", 0);
5682	SCTP_MCORE_UNLOCK(wkq);
5683
5684	/* Bind to our cpu */
5685	thread_lock(curthread);
5686	sched_bind(curthread, wkq->cpuid);
5687	thread_unlock(curthread);
5688
5689	/* Now lets start working */
5690	SCTP_MCORE_LOCK(wkq);
5691	/* Now grab lock and go */
5692	for (;;) {
5693		SCTP_MCORE_QLOCK(wkq);
5694skip_sleep:
5695		wkq->running = 1;
5696		qent = TAILQ_FIRST(&wkq->que);
5697		if (qent) {
5698			TAILQ_REMOVE(&wkq->que, qent, next);
5699			SCTP_MCORE_QUNLOCK(wkq);
5700			CURVNET_SET(qent->vn);
5701			m = qent->m;
5702			off = qent->off;
5703			v6 = qent->v6;
5704			SCTP_FREE(qent, SCTP_M_MCORE);
5705			if (v6 == 0) {
5706				sctp_input_with_port(m, off, 0);
5707			} else {
5708				SCTP_PRINTF("V6 not yet supported\n");
5709				sctp_m_freem(m);
5710			}
5711			CURVNET_RESTORE();
5712			SCTP_MCORE_QLOCK(wkq);
5713		}
5714		wkq->running = 0;
5715		if (!TAILQ_EMPTY(&wkq->que)) {
5716			goto skip_sleep;
5717		}
5718		SCTP_MCORE_QUNLOCK(wkq);
5719		msleep(&wkq->running,
5720		    &wkq->core_mtx,
5721		    0, "wait for pkt", 0);
5722	}
5723}
5724
5725static void
5726sctp_startup_mcore_threads(void)
5727{
5728	int i, cpu;
5729
5730	if (mp_ncpus == 1)
5731		return;
5732
5733	if (sctp_mcore_workers != NULL) {
5734		/*
5735		 * Already been here in some previous vnet?
5736		 */
5737		return;
5738	}
5739	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5740	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5741	    SCTP_M_MCORE);
5742	if (sctp_mcore_workers == NULL) {
5743		/* TSNH I hope */
5744		return;
5745	}
5746	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5747	    sizeof(struct sctp_mcore_ctrl)));
5748	/* Init the structures */
5749	for (i = 0; i <= mp_maxid; i++) {
5750		TAILQ_INIT(&sctp_mcore_workers[i].que);
5751		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5752		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5753		sctp_mcore_workers[i].cpuid = i;
5754	}
5755	if (sctp_cpuarry == NULL) {
5756		SCTP_MALLOC(sctp_cpuarry, int *,
5757		    (mp_ncpus * sizeof(int)),
5758		    SCTP_M_MCORE);
5759		i = 0;
5760		CPU_FOREACH(cpu) {
5761			sctp_cpuarry[i] = cpu;
5762			i++;
5763		}
5764	}
5765	/* Now start them all */
5766	CPU_FOREACH(cpu) {
5767		(void)kproc_create(sctp_mcore_thread,
5768		    (void *)&sctp_mcore_workers[cpu],
5769		    &sctp_mcore_workers[cpu].thread_proc,
5770		    RFPROC,
5771		    SCTP_KTHREAD_PAGES,
5772		    SCTP_MCORE_NAME);
5773
5774	}
5775}
5776
5777#endif
5778
5779void
5780sctp_pcb_init()
5781{
5782	/*
5783	 * SCTP initialization for the PCB structures should be called by
5784	 * the sctp_init() funciton.
5785	 */
5786	int i;
5787	struct timeval tv;
5788
5789	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5790		/* error I was called twice */
5791		return;
5792	}
5793	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5794
5795#if defined(SCTP_LOCAL_TRACE_BUF)
5796	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5797#endif
5798#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5799	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5800	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5801	    SCTP_M_MCORE);
5802#endif
5803	(void)SCTP_GETTIME_TIMEVAL(&tv);
5804#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5805	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5806	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5807	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5808#else
5809	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5810	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5811	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5812#endif
5813	/* init the empty list of (All) Endpoints */
5814	LIST_INIT(&SCTP_BASE_INFO(listhead));
5815
5816
5817	/* init the hash table of endpoints */
5818	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5819	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5820	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5821	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5822	    &SCTP_BASE_INFO(hashasocmark));
5823	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5824	    &SCTP_BASE_INFO(hashmark));
5825	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5826	    &SCTP_BASE_INFO(hashtcpmark));
5827	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5828
5829
5830	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5831	    &SCTP_BASE_INFO(hashvrfmark));
5832
5833	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5834	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5835	/* init the zones */
5836	/*
5837	 * FIX ME: Should check for NULL returns, but if it does fail we are
5838	 * doomed to panic anyways... add later maybe.
5839	 */
5840	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5841	    sizeof(struct sctp_inpcb), maxsockets);
5842
5843	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5844	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5845
5846	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5847	    sizeof(struct sctp_laddr),
5848	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5849
5850	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5851	    sizeof(struct sctp_nets),
5852	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5853
5854	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5855	    sizeof(struct sctp_tmit_chunk),
5856	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5857
5858	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5859	    sizeof(struct sctp_queued_to_read),
5860	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5861
5862	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5863	    sizeof(struct sctp_stream_queue_pending),
5864	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5865
5866	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5867	    sizeof(struct sctp_asconf),
5868	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5869
5870	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5871	    sizeof(struct sctp_asconf_ack),
5872	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5873
5874
5875	/* Master Lock INIT for info structure */
5876	SCTP_INP_INFO_LOCK_INIT();
5877	SCTP_STATLOG_INIT_LOCK();
5878
5879	SCTP_IPI_COUNT_INIT();
5880	SCTP_IPI_ADDR_INIT();
5881#ifdef SCTP_PACKET_LOGGING
5882	SCTP_IP_PKTLOG_INIT();
5883#endif
5884	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5885
5886	SCTP_WQ_ADDR_INIT();
5887	/* not sure if we need all the counts */
5888	SCTP_BASE_INFO(ipi_count_ep) = 0;
5889	/* assoc/tcb zone info */
5890	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5891	/* local addrlist zone info */
5892	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5893	/* remote addrlist zone info */
5894	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5895	/* chunk info */
5896	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5897
5898	/* socket queue zone info */
5899	SCTP_BASE_INFO(ipi_count_readq) = 0;
5900
5901	/* stream out queue cont */
5902	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5903
5904	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5905	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5906
5907	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5908
5909	/* Init the TIMEWAIT list */
5910	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5911		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5912	}
5913	sctp_startup_iterator();
5914
5915#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5916	sctp_startup_mcore_threads();
5917#endif
5918
5919	/*
5920	 * INIT the default VRF which for BSD is the only one, other O/S's
5921	 * may have more. But initially they must start with one and then
5922	 * add the VRF's as addresses are added.
5923	 */
5924	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5925}
5926
5927/*
5928 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5929 */
5930void
5931sctp_pcb_finish(void)
5932{
5933	struct sctp_vrflist *vrf_bucket;
5934	struct sctp_vrf *vrf, *nvrf;
5935	struct sctp_ifn *ifn, *nifn;
5936	struct sctp_ifa *ifa, *nifa;
5937	struct sctpvtaghead *chain;
5938	struct sctp_tagblock *twait_block, *prev_twait_block;
5939	struct sctp_laddr *wi, *nwi;
5940	int i;
5941	struct sctp_iterator *it, *nit;
5942
5943	/*
5944	 * In FreeBSD the iterator thread never exits but we do clean up.
5945	 * The only way FreeBSD reaches here is if we have VRF's but we
5946	 * still add the ifdef to make it compile on old versions.
5947	 */
5948	SCTP_IPI_ITERATOR_WQ_LOCK();
5949	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5950		if (it->vn != curvnet) {
5951			continue;
5952		}
5953		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5954		if (it->function_atend != NULL) {
5955			(*it->function_atend) (it->pointer, it->val);
5956		}
5957		SCTP_FREE(it, SCTP_M_ITER);
5958	}
5959	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5960	SCTP_ITERATOR_LOCK();
5961	if ((sctp_it_ctl.cur_it) &&
5962	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5963		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5964	}
5965	SCTP_ITERATOR_UNLOCK();
5966	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5967	SCTP_WQ_ADDR_LOCK();
5968	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5969		LIST_REMOVE(wi, sctp_nxt_addr);
5970		SCTP_DECR_LADDR_COUNT();
5971		if (wi->action == SCTP_DEL_IP_ADDRESS) {
5972			SCTP_FREE(wi->ifa, SCTP_M_IFA);
5973		}
5974		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5975	}
5976	SCTP_WQ_ADDR_UNLOCK();
5977
5978	/*
5979	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5980	 * destroyed first).
5981	 */
5982	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5983	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5984		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5985			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5986				/* free the ifa */
5987				LIST_REMOVE(ifa, next_bucket);
5988				LIST_REMOVE(ifa, next_ifa);
5989				SCTP_FREE(ifa, SCTP_M_IFA);
5990			}
5991			/* free the ifn */
5992			LIST_REMOVE(ifn, next_bucket);
5993			LIST_REMOVE(ifn, next_ifn);
5994			SCTP_FREE(ifn, SCTP_M_IFN);
5995		}
5996		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5997		/* free the vrf */
5998		LIST_REMOVE(vrf, next_vrf);
5999		SCTP_FREE(vrf, SCTP_M_VRF);
6000	}
6001	/* free the vrf hashes */
6002	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
6003	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
6004
6005	/*
6006	 * free the TIMEWAIT list elements malloc'd in the function
6007	 * sctp_add_vtag_to_timewait()...
6008	 */
6009	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
6010		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
6011		if (!LIST_EMPTY(chain)) {
6012			prev_twait_block = NULL;
6013			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6014				if (prev_twait_block) {
6015					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6016				}
6017				prev_twait_block = twait_block;
6018			}
6019			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6020		}
6021	}
6022
6023	/* free the locks and mutexes */
6024#ifdef SCTP_PACKET_LOGGING
6025	SCTP_IP_PKTLOG_DESTROY();
6026#endif
6027	SCTP_IPI_ADDR_DESTROY();
6028	SCTP_STATLOG_DESTROY();
6029	SCTP_INP_INFO_LOCK_DESTROY();
6030
6031	SCTP_WQ_ADDR_DESTROY();
6032
6033	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6034	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6035	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6036	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6037	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6038	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6039	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6040	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6041	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6042	/* Get rid of other stuff to */
6043	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
6044		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
6045	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
6046		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6047	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6048		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6049#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6050	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6051#endif
6052}
6053
6054
6055int
6056sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6057    int offset, int limit,
6058    struct sockaddr *src, struct sockaddr *dst,
6059    struct sockaddr *altsa)
6060{
6061	/*
6062	 * grub through the INIT pulling addresses and loading them to the
6063	 * nets structure in the asoc. The from address in the mbuf should
6064	 * also be loaded (if it is not already). This routine can be called
6065	 * with either INIT or INIT-ACK's as long as the m points to the IP
6066	 * packet and the offset points to the beginning of the parameters.
6067	 */
6068	struct sctp_inpcb *inp;
6069	struct sctp_nets *net, *nnet, *net_tmp;
6070	struct sctp_paramhdr *phdr, parm_buf;
6071	struct sctp_tcb *stcb_tmp;
6072	uint16_t ptype, plen;
6073	struct sockaddr *sa;
6074	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6075	struct sctp_auth_random *p_random = NULL;
6076	uint16_t random_len = 0;
6077	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6078	struct sctp_auth_hmac_algo *hmacs = NULL;
6079	uint16_t hmacs_len = 0;
6080	uint8_t saw_asconf = 0;
6081	uint8_t saw_asconf_ack = 0;
6082	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6083	struct sctp_auth_chunk_list *chunks = NULL;
6084	uint16_t num_chunks = 0;
6085	sctp_key_t *new_key;
6086	uint32_t keylen;
6087	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6088	uint8_t peer_supports_ecn;
6089	uint8_t peer_supports_prsctp;
6090	uint8_t peer_supports_auth;
6091	uint8_t peer_supports_asconf;
6092	uint8_t peer_supports_asconf_ack;
6093	uint8_t peer_supports_reconfig;
6094	uint8_t peer_supports_nrsack;
6095	uint8_t peer_supports_pktdrop;
6096
6097#ifdef INET
6098	struct sockaddr_in sin;
6099
6100#endif
6101#ifdef INET6
6102	struct sockaddr_in6 sin6;
6103
6104#endif
6105
6106	/* First get the destination address setup too. */
6107#ifdef INET
6108	memset(&sin, 0, sizeof(sin));
6109	sin.sin_family = AF_INET;
6110	sin.sin_len = sizeof(sin);
6111	sin.sin_port = stcb->rport;
6112#endif
6113#ifdef INET6
6114	memset(&sin6, 0, sizeof(sin6));
6115	sin6.sin6_family = AF_INET6;
6116	sin6.sin6_len = sizeof(struct sockaddr_in6);
6117	sin6.sin6_port = stcb->rport;
6118#endif
6119	if (altsa) {
6120		sa = altsa;
6121	} else {
6122		sa = src;
6123	}
6124	peer_supports_ecn = 0;
6125	peer_supports_prsctp = 0;
6126	peer_supports_auth = 0;
6127	peer_supports_asconf = 0;
6128	peer_supports_reconfig = 0;
6129	peer_supports_nrsack = 0;
6130	peer_supports_pktdrop = 0;
6131	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6132		/* mark all addresses that we have currently on the list */
6133		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6134	}
6135	/* does the source address already exist? if so skip it */
6136	inp = stcb->sctp_ep;
6137	atomic_add_int(&stcb->asoc.refcnt, 1);
6138	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6139	atomic_add_int(&stcb->asoc.refcnt, -1);
6140
6141	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6142		/* we must add the source address */
6143		/* no scope set here since we have a tcb already. */
6144		switch (sa->sa_family) {
6145#ifdef INET
6146		case AF_INET:
6147			if (stcb->asoc.scope.ipv4_addr_legal) {
6148				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6149					return (-1);
6150				}
6151			}
6152			break;
6153#endif
6154#ifdef INET6
6155		case AF_INET6:
6156			if (stcb->asoc.scope.ipv6_addr_legal) {
6157				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6158					return (-2);
6159				}
6160			}
6161			break;
6162#endif
6163		default:
6164			break;
6165		}
6166	} else {
6167		if (net_tmp != NULL && stcb_tmp == stcb) {
6168			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6169		} else if (stcb_tmp != stcb) {
6170			/* It belongs to another association? */
6171			if (stcb_tmp)
6172				SCTP_TCB_UNLOCK(stcb_tmp);
6173			return (-3);
6174		}
6175	}
6176	if (stcb->asoc.state == 0) {
6177		/* the assoc was freed? */
6178		return (-4);
6179	}
6180	/* now we must go through each of the params. */
6181	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6182	while (phdr) {
6183		ptype = ntohs(phdr->param_type);
6184		plen = ntohs(phdr->param_length);
6185		/*
6186		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6187		 * (uint32_t)ptype, (int)plen);
6188		 */
6189		if (offset + plen > limit) {
6190			break;
6191		}
6192		if (plen == 0) {
6193			break;
6194		}
6195#ifdef INET
6196		if (ptype == SCTP_IPV4_ADDRESS) {
6197			if (stcb->asoc.scope.ipv4_addr_legal) {
6198				struct sctp_ipv4addr_param *p4, p4_buf;
6199
6200				/* ok get the v4 address and check/add */
6201				phdr = sctp_get_next_param(m, offset,
6202				    (struct sctp_paramhdr *)&p4_buf,
6203				    sizeof(p4_buf));
6204				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6205				    phdr == NULL) {
6206					return (-5);
6207				}
6208				p4 = (struct sctp_ipv4addr_param *)phdr;
6209				sin.sin_addr.s_addr = p4->addr;
6210				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6211					/* Skip multi-cast addresses */
6212					goto next_param;
6213				}
6214				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6215				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6216					goto next_param;
6217				}
6218				sa = (struct sockaddr *)&sin;
6219				inp = stcb->sctp_ep;
6220				atomic_add_int(&stcb->asoc.refcnt, 1);
6221				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6222				    dst, stcb);
6223				atomic_add_int(&stcb->asoc.refcnt, -1);
6224
6225				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6226				    inp == NULL) {
6227					/* we must add the source address */
6228					/*
6229					 * no scope set since we have a tcb
6230					 * already
6231					 */
6232
6233					/*
6234					 * we must validate the state again
6235					 * here
6236					 */
6237			add_it_now:
6238					if (stcb->asoc.state == 0) {
6239						/* the assoc was freed? */
6240						return (-7);
6241					}
6242					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6243						return (-8);
6244					}
6245				} else if (stcb_tmp == stcb) {
6246					if (stcb->asoc.state == 0) {
6247						/* the assoc was freed? */
6248						return (-10);
6249					}
6250					if (net != NULL) {
6251						/* clear flag */
6252						net->dest_state &=
6253						    ~SCTP_ADDR_NOT_IN_ASSOC;
6254					}
6255				} else {
6256					/*
6257					 * strange, address is in another
6258					 * assoc? straighten out locks.
6259					 */
6260					if (stcb_tmp) {
6261						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6262							/*
6263							 * in setup state we
6264							 * abort this guy
6265							 */
6266							sctp_abort_an_association(stcb_tmp->sctp_ep,
6267							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6268							goto add_it_now;
6269						}
6270						SCTP_TCB_UNLOCK(stcb_tmp);
6271					}
6272					if (stcb->asoc.state == 0) {
6273						/* the assoc was freed? */
6274						return (-12);
6275					}
6276					return (-13);
6277				}
6278			}
6279		} else
6280#endif
6281#ifdef INET6
6282		if (ptype == SCTP_IPV6_ADDRESS) {
6283			if (stcb->asoc.scope.ipv6_addr_legal) {
6284				/* ok get the v6 address and check/add */
6285				struct sctp_ipv6addr_param *p6, p6_buf;
6286
6287				phdr = sctp_get_next_param(m, offset,
6288				    (struct sctp_paramhdr *)&p6_buf,
6289				    sizeof(p6_buf));
6290				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6291				    phdr == NULL) {
6292					return (-14);
6293				}
6294				p6 = (struct sctp_ipv6addr_param *)phdr;
6295				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6296				    sizeof(p6->addr));
6297				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6298					/* Skip multi-cast addresses */
6299					goto next_param;
6300				}
6301				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6302					/*
6303					 * Link local make no sense without
6304					 * scope
6305					 */
6306					goto next_param;
6307				}
6308				sa = (struct sockaddr *)&sin6;
6309				inp = stcb->sctp_ep;
6310				atomic_add_int(&stcb->asoc.refcnt, 1);
6311				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6312				    dst, stcb);
6313				atomic_add_int(&stcb->asoc.refcnt, -1);
6314				if (stcb_tmp == NULL &&
6315				    (inp == stcb->sctp_ep || inp == NULL)) {
6316					/*
6317					 * we must validate the state again
6318					 * here
6319					 */
6320			add_it_now6:
6321					if (stcb->asoc.state == 0) {
6322						/* the assoc was freed? */
6323						return (-16);
6324					}
6325					/*
6326					 * we must add the address, no scope
6327					 * set
6328					 */
6329					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6330						return (-17);
6331					}
6332				} else if (stcb_tmp == stcb) {
6333					/*
6334					 * we must validate the state again
6335					 * here
6336					 */
6337					if (stcb->asoc.state == 0) {
6338						/* the assoc was freed? */
6339						return (-19);
6340					}
6341					if (net != NULL) {
6342						/* clear flag */
6343						net->dest_state &=
6344						    ~SCTP_ADDR_NOT_IN_ASSOC;
6345					}
6346				} else {
6347					/*
6348					 * strange, address is in another
6349					 * assoc? straighten out locks.
6350					 */
6351					if (stcb_tmp)
6352						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6353							/*
6354							 * in setup state we
6355							 * abort this guy
6356							 */
6357							sctp_abort_an_association(stcb_tmp->sctp_ep,
6358							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6359							goto add_it_now6;
6360						}
6361					SCTP_TCB_UNLOCK(stcb_tmp);
6362
6363					if (stcb->asoc.state == 0) {
6364						/* the assoc was freed? */
6365						return (-21);
6366					}
6367					return (-22);
6368				}
6369			}
6370		} else
6371#endif
6372		if (ptype == SCTP_ECN_CAPABLE) {
6373			peer_supports_ecn = 1;
6374		} else if (ptype == SCTP_ULP_ADAPTATION) {
6375			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6376				struct sctp_adaptation_layer_indication ai,
6377				                                *aip;
6378
6379				phdr = sctp_get_next_param(m, offset,
6380				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6381				aip = (struct sctp_adaptation_layer_indication *)phdr;
6382				if (aip) {
6383					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6384					stcb->asoc.adaptation_needed = 1;
6385				}
6386			}
6387		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6388			struct sctp_asconf_addr_param lstore, *fee;
6389			int lptype;
6390			struct sockaddr *lsa = NULL;
6391
6392#ifdef INET
6393			struct sctp_asconf_addrv4_param *fii;
6394
6395#endif
6396
6397			if (stcb->asoc.asconf_supported == 0) {
6398				return (-100);
6399			}
6400			if (plen > sizeof(lstore)) {
6401				return (-23);
6402			}
6403			phdr = sctp_get_next_param(m, offset,
6404			    (struct sctp_paramhdr *)&lstore,
6405			    min(plen, sizeof(lstore)));
6406			if (phdr == NULL) {
6407				return (-24);
6408			}
6409			fee = (struct sctp_asconf_addr_param *)phdr;
6410			lptype = ntohs(fee->addrp.ph.param_type);
6411			switch (lptype) {
6412#ifdef INET
6413			case SCTP_IPV4_ADDRESS:
6414				if (plen !=
6415				    sizeof(struct sctp_asconf_addrv4_param)) {
6416					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6417					    (int)sizeof(struct sctp_asconf_addrv4_param),
6418					    plen);
6419				} else {
6420					fii = (struct sctp_asconf_addrv4_param *)fee;
6421					sin.sin_addr.s_addr = fii->addrp.addr;
6422					lsa = (struct sockaddr *)&sin;
6423				}
6424				break;
6425#endif
6426#ifdef INET6
6427			case SCTP_IPV6_ADDRESS:
6428				if (plen !=
6429				    sizeof(struct sctp_asconf_addr_param)) {
6430					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6431					    (int)sizeof(struct sctp_asconf_addr_param),
6432					    plen);
6433				} else {
6434					memcpy(sin6.sin6_addr.s6_addr,
6435					    fee->addrp.addr,
6436					    sizeof(fee->addrp.addr));
6437					lsa = (struct sockaddr *)&sin6;
6438				}
6439				break;
6440#endif
6441			default:
6442				break;
6443			}
6444			if (lsa) {
6445				(void)sctp_set_primary_addr(stcb, sa, NULL);
6446			}
6447		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6448			stcb->asoc.peer_supports_nat = 1;
6449		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6450			/* Peer supports pr-sctp */
6451			peer_supports_prsctp = 1;
6452		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6453			/* A supported extension chunk */
6454			struct sctp_supported_chunk_types_param *pr_supported;
6455			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6456			int num_ent, i;
6457
6458			phdr = sctp_get_next_param(m, offset,
6459			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6460			if (phdr == NULL) {
6461				return (-25);
6462			}
6463			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6464			num_ent = plen - sizeof(struct sctp_paramhdr);
6465			for (i = 0; i < num_ent; i++) {
6466				switch (pr_supported->chunk_types[i]) {
6467				case SCTP_ASCONF:
6468					peer_supports_asconf = 1;
6469				case SCTP_ASCONF_ACK:
6470					peer_supports_asconf_ack = 1;
6471					break;
6472				case SCTP_FORWARD_CUM_TSN:
6473					peer_supports_prsctp = 1;
6474					break;
6475				case SCTP_PACKET_DROPPED:
6476					peer_supports_pktdrop = 1;
6477					break;
6478				case SCTP_NR_SELECTIVE_ACK:
6479					peer_supports_nrsack = 1;
6480					break;
6481				case SCTP_STREAM_RESET:
6482					peer_supports_reconfig = 1;
6483					break;
6484				case SCTP_AUTHENTICATION:
6485					peer_supports_auth = 1;
6486					break;
6487				default:
6488					/* one I have not learned yet */
6489					break;
6490
6491				}
6492			}
6493		} else if (ptype == SCTP_RANDOM) {
6494			if (plen > sizeof(random_store))
6495				break;
6496			if (got_random) {
6497				/* already processed a RANDOM */
6498				goto next_param;
6499			}
6500			phdr = sctp_get_next_param(m, offset,
6501			    (struct sctp_paramhdr *)random_store,
6502			    min(sizeof(random_store), plen));
6503			if (phdr == NULL)
6504				return (-26);
6505			p_random = (struct sctp_auth_random *)phdr;
6506			random_len = plen - sizeof(*p_random);
6507			/* enforce the random length */
6508			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6509				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6510				return (-27);
6511			}
6512			got_random = 1;
6513		} else if (ptype == SCTP_HMAC_LIST) {
6514			uint16_t num_hmacs;
6515			uint16_t i;
6516
6517			if (plen > sizeof(hmacs_store))
6518				break;
6519			if (got_hmacs) {
6520				/* already processed a HMAC list */
6521				goto next_param;
6522			}
6523			phdr = sctp_get_next_param(m, offset,
6524			    (struct sctp_paramhdr *)hmacs_store,
6525			    min(plen, sizeof(hmacs_store)));
6526			if (phdr == NULL)
6527				return (-28);
6528			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6529			hmacs_len = plen - sizeof(*hmacs);
6530			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6531			/* validate the hmac list */
6532			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6533				return (-29);
6534			}
6535			if (stcb->asoc.peer_hmacs != NULL)
6536				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6537			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6538			if (stcb->asoc.peer_hmacs != NULL) {
6539				for (i = 0; i < num_hmacs; i++) {
6540					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6541					    ntohs(hmacs->hmac_ids[i]));
6542				}
6543			}
6544			got_hmacs = 1;
6545		} else if (ptype == SCTP_CHUNK_LIST) {
6546			int i;
6547
6548			if (plen > sizeof(chunks_store))
6549				break;
6550			if (got_chklist) {
6551				/* already processed a Chunks list */
6552				goto next_param;
6553			}
6554			phdr = sctp_get_next_param(m, offset,
6555			    (struct sctp_paramhdr *)chunks_store,
6556			    min(plen, sizeof(chunks_store)));
6557			if (phdr == NULL)
6558				return (-30);
6559			chunks = (struct sctp_auth_chunk_list *)phdr;
6560			num_chunks = plen - sizeof(*chunks);
6561			if (stcb->asoc.peer_auth_chunks != NULL)
6562				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6563			else
6564				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6565			for (i = 0; i < num_chunks; i++) {
6566				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6567				    stcb->asoc.peer_auth_chunks);
6568				/* record asconf/asconf-ack if listed */
6569				if (chunks->chunk_types[i] == SCTP_ASCONF)
6570					saw_asconf = 1;
6571				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6572					saw_asconf_ack = 1;
6573
6574			}
6575			got_chklist = 1;
6576		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6577			    (ptype == SCTP_STATE_COOKIE) ||
6578			    (ptype == SCTP_UNRECOG_PARAM) ||
6579			    (ptype == SCTP_COOKIE_PRESERVE) ||
6580			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6581			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6582			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6583			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6584		    (ptype == SCTP_SUCCESS_REPORT)) {
6585			 /* don't care */ ;
6586		} else {
6587			if ((ptype & 0x8000) == 0x0000) {
6588				/*
6589				 * must stop processing the rest of the
6590				 * param's. Any report bits were handled
6591				 * with the call to
6592				 * sctp_arethere_unrecognized_parameters()
6593				 * when the INIT or INIT-ACK was first seen.
6594				 */
6595				break;
6596			}
6597		}
6598
6599next_param:
6600		offset += SCTP_SIZE32(plen);
6601		if (offset >= limit) {
6602			break;
6603		}
6604		phdr = sctp_get_next_param(m, offset, &parm_buf,
6605		    sizeof(parm_buf));
6606	}
6607	/* Now check to see if we need to purge any addresses */
6608	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6609		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6610		    SCTP_ADDR_NOT_IN_ASSOC) {
6611			/* This address has been removed from the asoc */
6612			/* remove and free it */
6613			stcb->asoc.numnets--;
6614			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6615			sctp_free_remote_addr(net);
6616			if (net == stcb->asoc.primary_destination) {
6617				stcb->asoc.primary_destination = NULL;
6618				sctp_select_primary_destination(stcb);
6619			}
6620		}
6621	}
6622	if ((stcb->asoc.ecn_supported == 1) &&
6623	    (peer_supports_ecn == 0)) {
6624		stcb->asoc.ecn_supported = 0;
6625	}
6626	if ((stcb->asoc.prsctp_supported == 1) &&
6627	    (peer_supports_prsctp == 0)) {
6628		stcb->asoc.prsctp_supported = 0;
6629	}
6630	if ((stcb->asoc.auth_supported == 1) &&
6631	    ((peer_supports_auth == 0) ||
6632	    (got_random == 0) || (got_hmacs == 0))) {
6633		stcb->asoc.auth_supported = 0;
6634	}
6635	if ((stcb->asoc.asconf_supported == 1) &&
6636	    ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) ||
6637	    (stcb->asoc.auth_supported == 0) ||
6638	    (saw_asconf == 0) || (saw_asconf_ack == 0))) {
6639		stcb->asoc.asconf_supported = 0;
6640	}
6641	if ((stcb->asoc.reconfig_supported == 1) &&
6642	    (peer_supports_reconfig == 0)) {
6643		stcb->asoc.reconfig_supported = 0;
6644	}
6645	if ((stcb->asoc.nrsack_supported == 1) &&
6646	    (peer_supports_nrsack == 0)) {
6647		stcb->asoc.nrsack_supported = 0;
6648	}
6649	if ((stcb->asoc.pktdrop_supported == 1) &&
6650	    (peer_supports_pktdrop == 0)) {
6651		stcb->asoc.pktdrop_supported = 0;
6652	}
6653	/* validate authentication required parameters */
6654	if ((peer_supports_auth == 0) && (got_chklist == 1)) {
6655		/* peer does not support auth but sent a chunks list? */
6656		return (-31);
6657	}
6658	if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) {
6659		/* peer supports asconf but not auth? */
6660		return (-32);
6661	} else if ((peer_supports_asconf == 1) &&
6662		    (peer_supports_auth == 1) &&
6663	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6664		return (-33);
6665	}
6666	/* concatenate the full random key */
6667	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6668	if (chunks != NULL) {
6669		keylen += sizeof(*chunks) + num_chunks;
6670	}
6671	new_key = sctp_alloc_key(keylen);
6672	if (new_key != NULL) {
6673		/* copy in the RANDOM */
6674		if (p_random != NULL) {
6675			keylen = sizeof(*p_random) + random_len;
6676			bcopy(p_random, new_key->key, keylen);
6677		}
6678		/* append in the AUTH chunks */
6679		if (chunks != NULL) {
6680			bcopy(chunks, new_key->key + keylen,
6681			    sizeof(*chunks) + num_chunks);
6682			keylen += sizeof(*chunks) + num_chunks;
6683		}
6684		/* append in the HMACs */
6685		if (hmacs != NULL) {
6686			bcopy(hmacs, new_key->key + keylen,
6687			    sizeof(*hmacs) + hmacs_len);
6688		}
6689	} else {
6690		/* failed to get memory for the key */
6691		return (-34);
6692	}
6693	if (stcb->asoc.authinfo.peer_random != NULL)
6694		sctp_free_key(stcb->asoc.authinfo.peer_random);
6695	stcb->asoc.authinfo.peer_random = new_key;
6696	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6697	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6698
6699	return (0);
6700}
6701
6702int
6703sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6704    struct sctp_nets *net)
6705{
6706	/* make sure the requested primary address exists in the assoc */
6707	if (net == NULL && sa)
6708		net = sctp_findnet(stcb, sa);
6709
6710	if (net == NULL) {
6711		/* didn't find the requested primary address! */
6712		return (-1);
6713	} else {
6714		/* set the primary address */
6715		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6716			/* Must be confirmed, so queue to set */
6717			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6718			return (0);
6719		}
6720		stcb->asoc.primary_destination = net;
6721		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6722			sctp_free_remote_addr(stcb->asoc.alternate);
6723			stcb->asoc.alternate = NULL;
6724		}
6725		net = TAILQ_FIRST(&stcb->asoc.nets);
6726		if (net != stcb->asoc.primary_destination) {
6727			/*
6728			 * first one on the list is NOT the primary
6729			 * sctp_cmpaddr() is much more efficient if the
6730			 * primary is the first on the list, make it so.
6731			 */
6732			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6733			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6734		}
6735		return (0);
6736	}
6737}
6738
6739int
6740sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6741{
6742	/*
6743	 * This function serves two purposes. It will see if a TAG can be
6744	 * re-used and return 1 for yes it is ok and 0 for don't use that
6745	 * tag. A secondary function it will do is purge out old tags that
6746	 * can be removed.
6747	 */
6748	struct sctpvtaghead *chain;
6749	struct sctp_tagblock *twait_block;
6750	struct sctpasochead *head;
6751	struct sctp_tcb *stcb;
6752	int i;
6753
6754	SCTP_INP_INFO_RLOCK();
6755	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6756	    SCTP_BASE_INFO(hashasocmark))];
6757	if (head == NULL) {
6758		/* invalid vtag */
6759		goto skip_vtag_check;
6760	}
6761	LIST_FOREACH(stcb, head, sctp_asocs) {
6762		/*
6763		 * We choose not to lock anything here. TCB's can't be
6764		 * removed since we have the read lock, so they can't be
6765		 * freed on us, same thing for the INP. I may be wrong with
6766		 * this assumption, but we will go with it for now :-)
6767		 */
6768		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6769			continue;
6770		}
6771		if (stcb->asoc.my_vtag == tag) {
6772			/* candidate */
6773			if (stcb->rport != rport) {
6774				continue;
6775			}
6776			if (stcb->sctp_ep->sctp_lport != lport) {
6777				continue;
6778			}
6779			/* Its a used tag set */
6780			SCTP_INP_INFO_RUNLOCK();
6781			return (0);
6782		}
6783	}
6784skip_vtag_check:
6785
6786	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6787	/* Now what about timed wait ? */
6788	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6789		/*
6790		 * Block(s) are present, lets see if we have this tag in the
6791		 * list
6792		 */
6793		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6794			if (twait_block->vtag_block[i].v_tag == 0) {
6795				/* not used */
6796				continue;
6797			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6798			    now->tv_sec) {
6799				/* Audit expires this guy */
6800				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6801				twait_block->vtag_block[i].v_tag = 0;
6802				twait_block->vtag_block[i].lport = 0;
6803				twait_block->vtag_block[i].rport = 0;
6804			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6805				    (twait_block->vtag_block[i].lport == lport) &&
6806			    (twait_block->vtag_block[i].rport == rport)) {
6807				/* Bad tag, sorry :< */
6808				SCTP_INP_INFO_RUNLOCK();
6809				return (0);
6810			}
6811		}
6812	}
6813	SCTP_INP_INFO_RUNLOCK();
6814	return (1);
6815}
6816
6817static void
6818sctp_drain_mbufs(struct sctp_tcb *stcb)
6819{
6820	/*
6821	 * We must hunt this association for MBUF's past the cumack (i.e.
6822	 * out of order data that we can renege on).
6823	 */
6824	struct sctp_association *asoc;
6825	struct sctp_tmit_chunk *chk, *nchk;
6826	uint32_t cumulative_tsn_p1;
6827	struct sctp_queued_to_read *ctl, *nctl;
6828	int cnt, strmat;
6829	uint32_t gap, i;
6830	int fnd = 0;
6831
6832	/* We look for anything larger than the cum-ack + 1 */
6833
6834	asoc = &stcb->asoc;
6835	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6836		/* none we can reneg on. */
6837		return;
6838	}
6839	SCTP_STAT_INCR(sctps_protocol_drains_done);
6840	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6841	cnt = 0;
6842	/* First look in the re-assembly queue */
6843	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6844		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6845			/* Yep it is above cum-ack */
6846			cnt++;
6847			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6848			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6849			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6850			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6851			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6852			if (chk->data) {
6853				sctp_m_freem(chk->data);
6854				chk->data = NULL;
6855			}
6856			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6857		}
6858	}
6859	/* Ok that was fun, now we will drain all the inbound streams? */
6860	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6861		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6862			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6863				/* Yep it is above cum-ack */
6864				cnt++;
6865				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6866				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6867				sctp_ucount_decr(asoc->cnt_on_all_streams);
6868				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6869				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6870				if (ctl->data) {
6871					sctp_m_freem(ctl->data);
6872					ctl->data = NULL;
6873				}
6874				sctp_free_remote_addr(ctl->whoFrom);
6875				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6876				SCTP_DECR_READQ_COUNT();
6877			}
6878		}
6879	}
6880	if (cnt) {
6881		/* We must back down to see what the new highest is */
6882		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6883			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6884			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6885				asoc->highest_tsn_inside_map = i;
6886				fnd = 1;
6887				break;
6888			}
6889		}
6890		if (!fnd) {
6891			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6892		}
6893		/*
6894		 * Question, should we go through the delivery queue? The
6895		 * only reason things are on here is the app not reading OR
6896		 * a p-d-api up. An attacker COULD send enough in to
6897		 * initiate the PD-API and then send a bunch of stuff to
6898		 * other streams... these would wind up on the delivery
6899		 * queue.. and then we would not get to them. But in order
6900		 * to do this I then have to back-track and un-deliver
6901		 * sequence numbers in streams.. el-yucko. I think for now
6902		 * we will NOT look at the delivery queue and leave it to be
6903		 * something to consider later. An alternative would be to
6904		 * abort the P-D-API with a notification and then deliver
6905		 * the data.... Or another method might be to keep track of
6906		 * how many times the situation occurs and if we see a
6907		 * possible attack underway just abort the association.
6908		 */
6909#ifdef SCTP_DEBUG
6910		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6911#endif
6912		/*
6913		 * Now do we need to find a new
6914		 * asoc->highest_tsn_inside_map?
6915		 */
6916		asoc->last_revoke_count = cnt;
6917		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6918		/* sa_ignore NO_NULL_CHK */
6919		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6920		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6921	}
6922	/*
6923	 * Another issue, in un-setting the TSN's in the mapping array we
6924	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6925	 * two things to occur. It may cause us to do extra work in checking
6926	 * for our mapping array movement. More importantly it may cause us
6927	 * to SACK every datagram. This may not be a bad thing though since
6928	 * we will recover once we get our cum-ack above and all this stuff
6929	 * we dumped recovered.
6930	 */
6931}
6932
6933void
6934sctp_drain()
6935{
6936	/*
6937	 * We must walk the PCB lists for ALL associations here. The system
6938	 * is LOW on MBUF's and needs help. This is where reneging will
6939	 * occur. We really hope this does NOT happen!
6940	 */
6941	VNET_ITERATOR_DECL(vnet_iter);
6942	VNET_LIST_RLOCK_NOSLEEP();
6943	VNET_FOREACH(vnet_iter) {
6944		CURVNET_SET(vnet_iter);
6945		struct sctp_inpcb *inp;
6946		struct sctp_tcb *stcb;
6947
6948		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6949		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6950#ifdef VIMAGE
6951			continue;
6952#else
6953			return;
6954#endif
6955		}
6956		SCTP_INP_INFO_RLOCK();
6957		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6958			/* For each endpoint */
6959			SCTP_INP_RLOCK(inp);
6960			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6961				/* For each association */
6962				SCTP_TCB_LOCK(stcb);
6963				sctp_drain_mbufs(stcb);
6964				SCTP_TCB_UNLOCK(stcb);
6965			}
6966			SCTP_INP_RUNLOCK(inp);
6967		}
6968		SCTP_INP_INFO_RUNLOCK();
6969		CURVNET_RESTORE();
6970	}
6971	VNET_LIST_RUNLOCK_NOSLEEP();
6972}
6973
6974/*
6975 * start a new iterator
6976 * iterates through all endpoints and associations based on the pcb_state
6977 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6978 * assocs and "ef" (optional) is executed when the iterator completes.
6979 * "inpf" (optional) is executed for each new endpoint as it is being
6980 * iterated through. inpe (optional) is called when the inp completes
6981 * its way through all the stcbs.
6982 */
6983int
6984sctp_initiate_iterator(inp_func inpf,
6985    asoc_func af,
6986    inp_func inpe,
6987    uint32_t pcb_state,
6988    uint32_t pcb_features,
6989    uint32_t asoc_state,
6990    void *argp,
6991    uint32_t argi,
6992    end_func ef,
6993    struct sctp_inpcb *s_inp,
6994    uint8_t chunk_output_off)
6995{
6996	struct sctp_iterator *it = NULL;
6997
6998	if (af == NULL) {
6999		return (-1);
7000	}
7001	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
7002	    SCTP_M_ITER);
7003	if (it == NULL) {
7004		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
7005		return (ENOMEM);
7006	}
7007	memset(it, 0, sizeof(*it));
7008	it->function_assoc = af;
7009	it->function_inp = inpf;
7010	if (inpf)
7011		it->done_current_ep = 0;
7012	else
7013		it->done_current_ep = 1;
7014	it->function_atend = ef;
7015	it->pointer = argp;
7016	it->val = argi;
7017	it->pcb_flags = pcb_state;
7018	it->pcb_features = pcb_features;
7019	it->asoc_state = asoc_state;
7020	it->function_inp_end = inpe;
7021	it->no_chunk_output = chunk_output_off;
7022	it->vn = curvnet;
7023	if (s_inp) {
7024		/* Assume lock is held here */
7025		it->inp = s_inp;
7026		SCTP_INP_INCR_REF(it->inp);
7027		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7028	} else {
7029		SCTP_INP_INFO_RLOCK();
7030		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7031		if (it->inp) {
7032			SCTP_INP_INCR_REF(it->inp);
7033		}
7034		SCTP_INP_INFO_RUNLOCK();
7035		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7036
7037	}
7038	SCTP_IPI_ITERATOR_WQ_LOCK();
7039
7040	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7041	if (sctp_it_ctl.iterator_running == 0) {
7042		sctp_wakeup_iterator();
7043	}
7044	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7045	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7046	return (0);
7047}
7048