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: releng/11.0/sys/netinet/sctp_pcb.c 303956 2016-08-11 10:14:03Z tuexen $");
35
36#include <netinet/sctp_os.h>
37#include <sys/proc.h>
38#include <netinet/sctp_var.h>
39#include <netinet/sctp_sysctl.h>
40#include <netinet/sctp_pcb.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp.h>
43#include <netinet/sctp_header.h>
44#include <netinet/sctp_asconf.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_timer.h>
47#include <netinet/sctp_bsd_addr.h>
48#include <netinet/sctp_dtrace_define.h>
49#if defined(INET) || defined(INET6)
50#include <netinet/udp.h>
51#endif
52#ifdef INET6
53#include <netinet6/ip6_var.h>
54#endif
55#include <sys/sched.h>
56#include <sys/smp.h>
57#include <sys/unistd.h>
58
59
60VNET_DEFINE(struct sctp_base_info, system_base_info);
61
62/* FIX: we don't handle multiple link local scopes */
63/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
64#ifdef INET6
65int
66SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
67{
68	struct sockaddr_in6 tmp_a, tmp_b;
69
70	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
71	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
72		return (0);
73	}
74	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
75	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
76		return (0);
77	}
78	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
79}
80
81#endif
82
83void
84sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
85{
86	/*
87	 * We really don't need to lock this, but I will just because it
88	 * does not hurt.
89	 */
90	SCTP_INP_INFO_RLOCK();
91	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
92	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
93	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
94	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
95	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
96	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
97	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
98	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
99	SCTP_INP_INFO_RUNLOCK();
100}
101
102/*-
103 * Addresses are added to VRF's (Virtual Router's). For BSD we
104 * have only the default VRF 0. We maintain a hash list of
105 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
106 * these has a list of addresses. When we add a new address
107 * to a VRF we lookup the ifn/ifn_index, if the ifn does
108 * not exist we create it and add it to the list of IFN's
109 * within the VRF. Once we have the sctp_ifn, we add the
110 * address to the list. So we look something like:
111 *
112 * hash-vrf-table
113 *   vrf-> ifn-> ifn -> ifn
114 *   vrf    |
115 *    ...   +--ifa-> ifa -> ifa
116 *   vrf
117 *
118 * We keep these separate lists since the SCTP subsystem will
119 * point to these from its source address selection nets structure.
120 * When an address is deleted it does not happen right away on
121 * the SCTP side, it gets scheduled. What we do when a
122 * delete happens is immediately remove the address from
123 * the master list and decrement the refcount. As our
124 * addip iterator works through and frees the src address
125 * selection pointing to the sctp_ifa, eventually the refcount
126 * will reach 0 and we will delete it. Note that it is assumed
127 * that any locking on system level ifn/ifa is done at the
128 * caller of these functions and these routines will only
129 * lock the SCTP structures as they add or delete things.
130 *
131 * Other notes on VRF concepts.
132 *  - An endpoint can be in multiple VRF's
133 *  - An association lives within a VRF and only one VRF.
134 *  - Any incoming packet we can deduce the VRF for by
135 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
136 *  - Any downward send call or connect call must supply the
137 *    VRF via ancillary data or via some sort of set default
138 *    VRF socket option call (again for BSD no brainer since
139 *    the VRF is always 0).
140 *  - An endpoint may add multiple VRF's to it.
141 *  - Listening sockets can accept associations in any
142 *    of the VRF's they are in but the assoc will end up
143 *    in only one VRF (gotten from the packet or connect/send).
144 *
145 */
146
147struct sctp_vrf *
148sctp_allocate_vrf(int vrf_id)
149{
150	struct sctp_vrf *vrf = NULL;
151	struct sctp_vrflist *bucket;
152
153	/* First allocate the VRF structure */
154	vrf = sctp_find_vrf(vrf_id);
155	if (vrf) {
156		/* Already allocated */
157		return (vrf);
158	}
159	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
160	    SCTP_M_VRF);
161	if (vrf == NULL) {
162		/* No memory */
163#ifdef INVARIANTS
164		panic("No memory for VRF:%d", vrf_id);
165#endif
166		return (NULL);
167	}
168	/* setup the VRF */
169	memset(vrf, 0, sizeof(struct sctp_vrf));
170	vrf->vrf_id = vrf_id;
171	LIST_INIT(&vrf->ifnlist);
172	vrf->total_ifa_count = 0;
173	vrf->refcount = 0;
174	/* now also setup table ids */
175	SCTP_INIT_VRF_TABLEID(vrf);
176	/* Init the HASH of addresses */
177	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
178	    &vrf->vrf_addr_hashmark);
179	if (vrf->vrf_addr_hash == NULL) {
180		/* No memory */
181#ifdef INVARIANTS
182		panic("No memory for VRF:%d", vrf_id);
183#endif
184		SCTP_FREE(vrf, SCTP_M_VRF);
185		return (NULL);
186	}
187	/* Add it to the hash table */
188	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
189	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
190	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
191	return (vrf);
192}
193
194
195struct sctp_ifn *
196sctp_find_ifn(void *ifn, uint32_t ifn_index)
197{
198	struct sctp_ifn *sctp_ifnp;
199	struct sctp_ifnlist *hash_ifn_head;
200
201	/*
202	 * We assume the lock is held for the addresses if that's wrong
203	 * problems could occur :-)
204	 */
205	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
206	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
207		if (sctp_ifnp->ifn_index == ifn_index) {
208			return (sctp_ifnp);
209		}
210		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
211			return (sctp_ifnp);
212		}
213	}
214	return (NULL);
215}
216
217
218struct sctp_vrf *
219sctp_find_vrf(uint32_t vrf_id)
220{
221	struct sctp_vrflist *bucket;
222	struct sctp_vrf *liste;
223
224	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
225	LIST_FOREACH(liste, bucket, next_vrf) {
226		if (vrf_id == liste->vrf_id) {
227			return (liste);
228		}
229	}
230	return (NULL);
231}
232
233
234void
235sctp_free_vrf(struct sctp_vrf *vrf)
236{
237	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
238		if (vrf->vrf_addr_hash) {
239			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
240			vrf->vrf_addr_hash = NULL;
241		}
242		/* We zero'd the count */
243		LIST_REMOVE(vrf, next_vrf);
244		SCTP_FREE(vrf, SCTP_M_VRF);
245		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
246	}
247}
248
249
250void
251sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
252{
253	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
254		/* We zero'd the count */
255		if (sctp_ifnp->vrf) {
256			sctp_free_vrf(sctp_ifnp->vrf);
257		}
258		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
259		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
260	}
261}
262
263
264void
265sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
266{
267	struct sctp_ifn *sctp_ifnp;
268
269	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
270	if (sctp_ifnp != NULL) {
271		sctp_ifnp->ifn_mtu = mtu;
272	}
273}
274
275
276void
277sctp_free_ifa(struct sctp_ifa *sctp_ifap)
278{
279	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
280		/* We zero'd the count */
281		if (sctp_ifap->ifn_p) {
282			sctp_free_ifn(sctp_ifap->ifn_p);
283		}
284		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
285		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
286	}
287}
288
289
290static void
291sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
292{
293	struct sctp_ifn *found;
294
295	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
296	if (found == NULL) {
297		/* Not in the list.. sorry */
298		return;
299	}
300	if (hold_addr_lock == 0)
301		SCTP_IPI_ADDR_WLOCK();
302	LIST_REMOVE(sctp_ifnp, next_bucket);
303	LIST_REMOVE(sctp_ifnp, next_ifn);
304	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
305	    sctp_ifnp->registered_af);
306	if (hold_addr_lock == 0)
307		SCTP_IPI_ADDR_WUNLOCK();
308	/* Take away the reference, and possibly free it */
309	sctp_free_ifn(sctp_ifnp);
310}
311
312
313void
314sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
315    const char *if_name, uint32_t ifn_index)
316{
317	struct sctp_vrf *vrf;
318	struct sctp_ifa *sctp_ifap;
319
320	SCTP_IPI_ADDR_RLOCK();
321	vrf = sctp_find_vrf(vrf_id);
322	if (vrf == NULL) {
323		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
324		goto out;
325
326	}
327	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
328	if (sctp_ifap == NULL) {
329		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
330		goto out;
331	}
332	if (sctp_ifap->ifn_p == NULL) {
333		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unusable\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 unusable\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", __func__);
1119					continue;
1120				}
1121				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1122					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1123					continue;
1124				}
1125				if (laddr->ifa->address.sa.sa_family ==
1126				    to->sa_family) {
1127					/* see if it matches */
1128#ifdef INET
1129					if (from->sa_family == AF_INET) {
1130						struct sockaddr_in *intf_addr,
1131						           *sin;
1132
1133						intf_addr = &laddr->ifa->address.sin;
1134						sin = (struct sockaddr_in *)to;
1135						if (sin->sin_addr.s_addr ==
1136						    intf_addr->sin_addr.s_addr) {
1137							match = 1;
1138							break;
1139						}
1140					}
1141#endif
1142#ifdef INET6
1143					if (from->sa_family == AF_INET6) {
1144						struct sockaddr_in6 *intf_addr6;
1145						struct sockaddr_in6 *sin6;
1146
1147						sin6 = (struct sockaddr_in6 *)
1148						    to;
1149						intf_addr6 = &laddr->ifa->address.sin6;
1150
1151						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1152						    intf_addr6)) {
1153							match = 1;
1154							break;
1155						}
1156					}
1157#endif
1158				}
1159			}
1160			if (match == 0) {
1161				/* This endpoint does not have this address */
1162				SCTP_INP_RUNLOCK(inp);
1163				continue;
1164			}
1165		}
1166		/*
1167		 * Ok if we hit here the ep has the address, does it hold
1168		 * the tcb?
1169		 */
1170		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1171		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1172		if (stcb == NULL) {
1173			SCTP_INP_RUNLOCK(inp);
1174			continue;
1175		}
1176		SCTP_TCB_LOCK(stcb);
1177		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1178			SCTP_TCB_UNLOCK(stcb);
1179			SCTP_INP_RUNLOCK(inp);
1180			continue;
1181		}
1182		if (stcb->rport != rport) {
1183			/* remote port does not match. */
1184			SCTP_TCB_UNLOCK(stcb);
1185			SCTP_INP_RUNLOCK(inp);
1186			continue;
1187		}
1188		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1189			SCTP_TCB_UNLOCK(stcb);
1190			SCTP_INP_RUNLOCK(inp);
1191			continue;
1192		}
1193		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1194			SCTP_TCB_UNLOCK(stcb);
1195			SCTP_INP_RUNLOCK(inp);
1196			continue;
1197		}
1198		/* Does this TCB have a matching address? */
1199		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1200
1201			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1202				/* not the same family, can't be a match */
1203				continue;
1204			}
1205			switch (from->sa_family) {
1206#ifdef INET
1207			case AF_INET:
1208				{
1209					struct sockaddr_in *sin, *rsin;
1210
1211					sin = (struct sockaddr_in *)&net->ro._l_addr;
1212					rsin = (struct sockaddr_in *)from;
1213					if (sin->sin_addr.s_addr ==
1214					    rsin->sin_addr.s_addr) {
1215						/* found it */
1216						if (netp != NULL) {
1217							*netp = net;
1218						}
1219						/*
1220						 * Update the endpoint
1221						 * pointer
1222						 */
1223						*inp_p = inp;
1224						SCTP_INP_RUNLOCK(inp);
1225						return (stcb);
1226					}
1227					break;
1228				}
1229#endif
1230#ifdef INET6
1231			case AF_INET6:
1232				{
1233					struct sockaddr_in6 *sin6, *rsin6;
1234
1235					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1236					rsin6 = (struct sockaddr_in6 *)from;
1237					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1238					    rsin6)) {
1239						/* found it */
1240						if (netp != NULL) {
1241							*netp = net;
1242						}
1243						/*
1244						 * Update the endpoint
1245						 * pointer
1246						 */
1247						*inp_p = inp;
1248						SCTP_INP_RUNLOCK(inp);
1249						return (stcb);
1250					}
1251					break;
1252				}
1253#endif
1254			default:
1255				/* TSNH */
1256				break;
1257			}
1258		}
1259		SCTP_TCB_UNLOCK(stcb);
1260		SCTP_INP_RUNLOCK(inp);
1261	}
1262	return (NULL);
1263}
1264
1265
1266/*
1267 * rules for use
1268 *
1269 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1270 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1271 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1272 * NULL.
1273 */
1274
1275struct sctp_tcb *
1276sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1277    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1278{
1279	struct sctpasochead *head;
1280	struct sctp_inpcb *inp;
1281	struct sctp_tcb *stcb = NULL;
1282	struct sctp_nets *net;
1283	uint16_t rport;
1284
1285	inp = *inp_p;
1286	switch (remote->sa_family) {
1287#ifdef INET
1288	case AF_INET:
1289		rport = (((struct sockaddr_in *)remote)->sin_port);
1290		break;
1291#endif
1292#ifdef INET6
1293	case AF_INET6:
1294		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1295		break;
1296#endif
1297	default:
1298		return (NULL);
1299	}
1300	if (locked_tcb) {
1301		/*
1302		 * UN-lock so we can do proper locking here this occurs when
1303		 * called from load_addresses_from_init.
1304		 */
1305		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1306		SCTP_TCB_UNLOCK(locked_tcb);
1307	}
1308	SCTP_INP_INFO_RLOCK();
1309	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1310	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1311		/*-
1312		 * Now either this guy is our listener or it's the
1313		 * connector. If it is the one that issued the connect, then
1314		 * it's only chance is to be the first TCB in the list. If
1315		 * it is the acceptor, then do the special_lookup to hash
1316		 * and find the real inp.
1317		 */
1318		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1319			/* to is peer addr, from is my addr */
1320			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1321			    netp, inp->def_vrf_id);
1322			if ((stcb != NULL) && (locked_tcb == NULL)) {
1323				/* we have a locked tcb, lower refcount */
1324				SCTP_INP_DECR_REF(inp);
1325			}
1326			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1327				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1328				SCTP_TCB_LOCK(locked_tcb);
1329				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1330				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1331			}
1332			SCTP_INP_INFO_RUNLOCK();
1333			return (stcb);
1334		} else {
1335			SCTP_INP_WLOCK(inp);
1336			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1337				goto null_return;
1338			}
1339			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1340			if (stcb == NULL) {
1341				goto null_return;
1342			}
1343			SCTP_TCB_LOCK(stcb);
1344
1345			if (stcb->rport != rport) {
1346				/* remote port does not match. */
1347				SCTP_TCB_UNLOCK(stcb);
1348				goto null_return;
1349			}
1350			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1351				SCTP_TCB_UNLOCK(stcb);
1352				goto null_return;
1353			}
1354			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1355				SCTP_TCB_UNLOCK(stcb);
1356				goto null_return;
1357			}
1358			/* now look at the list of remote addresses */
1359			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1360#ifdef INVARIANTS
1361				if (net == (TAILQ_NEXT(net, sctp_next))) {
1362					panic("Corrupt net list");
1363				}
1364#endif
1365				if (net->ro._l_addr.sa.sa_family !=
1366				    remote->sa_family) {
1367					/* not the same family */
1368					continue;
1369				}
1370				switch (remote->sa_family) {
1371#ifdef INET
1372				case AF_INET:
1373					{
1374						struct sockaddr_in *sin,
1375						           *rsin;
1376
1377						sin = (struct sockaddr_in *)
1378						    &net->ro._l_addr;
1379						rsin = (struct sockaddr_in *)remote;
1380						if (sin->sin_addr.s_addr ==
1381						    rsin->sin_addr.s_addr) {
1382							/* found it */
1383							if (netp != NULL) {
1384								*netp = net;
1385							}
1386							if (locked_tcb == NULL) {
1387								SCTP_INP_DECR_REF(inp);
1388							} else if (locked_tcb != stcb) {
1389								SCTP_TCB_LOCK(locked_tcb);
1390							}
1391							if (locked_tcb) {
1392								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1393							}
1394							SCTP_INP_WUNLOCK(inp);
1395							SCTP_INP_INFO_RUNLOCK();
1396							return (stcb);
1397						}
1398						break;
1399					}
1400#endif
1401#ifdef INET6
1402				case AF_INET6:
1403					{
1404						struct sockaddr_in6 *sin6,
1405						            *rsin6;
1406
1407						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1408						rsin6 = (struct sockaddr_in6 *)remote;
1409						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1410						    rsin6)) {
1411							/* found it */
1412							if (netp != NULL) {
1413								*netp = net;
1414							}
1415							if (locked_tcb == NULL) {
1416								SCTP_INP_DECR_REF(inp);
1417							} else if (locked_tcb != stcb) {
1418								SCTP_TCB_LOCK(locked_tcb);
1419							}
1420							if (locked_tcb) {
1421								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1422							}
1423							SCTP_INP_WUNLOCK(inp);
1424							SCTP_INP_INFO_RUNLOCK();
1425							return (stcb);
1426						}
1427						break;
1428					}
1429#endif
1430				default:
1431					/* TSNH */
1432					break;
1433				}
1434			}
1435			SCTP_TCB_UNLOCK(stcb);
1436		}
1437	} else {
1438		SCTP_INP_WLOCK(inp);
1439		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1440			goto null_return;
1441		}
1442		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1443		    inp->sctp_hashmark)];
1444		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1445			if (stcb->rport != rport) {
1446				/* remote port does not match */
1447				continue;
1448			}
1449			SCTP_TCB_LOCK(stcb);
1450			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1451				SCTP_TCB_UNLOCK(stcb);
1452				continue;
1453			}
1454			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1455				SCTP_TCB_UNLOCK(stcb);
1456				continue;
1457			}
1458			/* now look at the list of remote addresses */
1459			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1460#ifdef INVARIANTS
1461				if (net == (TAILQ_NEXT(net, sctp_next))) {
1462					panic("Corrupt net list");
1463				}
1464#endif
1465				if (net->ro._l_addr.sa.sa_family !=
1466				    remote->sa_family) {
1467					/* not the same family */
1468					continue;
1469				}
1470				switch (remote->sa_family) {
1471#ifdef INET
1472				case AF_INET:
1473					{
1474						struct sockaddr_in *sin,
1475						           *rsin;
1476
1477						sin = (struct sockaddr_in *)
1478						    &net->ro._l_addr;
1479						rsin = (struct sockaddr_in *)remote;
1480						if (sin->sin_addr.s_addr ==
1481						    rsin->sin_addr.s_addr) {
1482							/* found it */
1483							if (netp != NULL) {
1484								*netp = net;
1485							}
1486							if (locked_tcb == NULL) {
1487								SCTP_INP_DECR_REF(inp);
1488							} else if (locked_tcb != stcb) {
1489								SCTP_TCB_LOCK(locked_tcb);
1490							}
1491							if (locked_tcb) {
1492								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1493							}
1494							SCTP_INP_WUNLOCK(inp);
1495							SCTP_INP_INFO_RUNLOCK();
1496							return (stcb);
1497						}
1498						break;
1499					}
1500#endif
1501#ifdef INET6
1502				case AF_INET6:
1503					{
1504						struct sockaddr_in6 *sin6,
1505						            *rsin6;
1506
1507						sin6 = (struct sockaddr_in6 *)
1508						    &net->ro._l_addr;
1509						rsin6 = (struct sockaddr_in6 *)remote;
1510						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1511						    rsin6)) {
1512							/* found it */
1513							if (netp != NULL) {
1514								*netp = net;
1515							}
1516							if (locked_tcb == NULL) {
1517								SCTP_INP_DECR_REF(inp);
1518							} else if (locked_tcb != stcb) {
1519								SCTP_TCB_LOCK(locked_tcb);
1520							}
1521							if (locked_tcb) {
1522								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1523							}
1524							SCTP_INP_WUNLOCK(inp);
1525							SCTP_INP_INFO_RUNLOCK();
1526							return (stcb);
1527						}
1528						break;
1529					}
1530#endif
1531				default:
1532					/* TSNH */
1533					break;
1534				}
1535			}
1536			SCTP_TCB_UNLOCK(stcb);
1537		}
1538	}
1539null_return:
1540	/* clean up for returning null */
1541	if (locked_tcb) {
1542		SCTP_TCB_LOCK(locked_tcb);
1543		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1544	}
1545	SCTP_INP_WUNLOCK(inp);
1546	SCTP_INP_INFO_RUNLOCK();
1547	/* not found */
1548	return (NULL);
1549}
1550
1551
1552/*
1553 * Find an association for a specific endpoint using the association id given
1554 * out in the COMM_UP notification
1555 */
1556struct sctp_tcb *
1557sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1558{
1559	/*
1560	 * Use my the assoc_id to find a endpoint
1561	 */
1562	struct sctpasochead *head;
1563	struct sctp_tcb *stcb;
1564	uint32_t id;
1565
1566	if (inp == NULL) {
1567		SCTP_PRINTF("TSNH ep_associd\n");
1568		return (NULL);
1569	}
1570	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1571		SCTP_PRINTF("TSNH ep_associd0\n");
1572		return (NULL);
1573	}
1574	id = (uint32_t) asoc_id;
1575	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1576	if (head == NULL) {
1577		/* invalid id TSNH */
1578		SCTP_PRINTF("TSNH ep_associd1\n");
1579		return (NULL);
1580	}
1581	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1582		if (stcb->asoc.assoc_id == id) {
1583			if (inp != stcb->sctp_ep) {
1584				/*
1585				 * some other guy has the same id active (id
1586				 * collision ??).
1587				 */
1588				SCTP_PRINTF("TSNH ep_associd2\n");
1589				continue;
1590			}
1591			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1592				continue;
1593			}
1594			if (want_lock) {
1595				SCTP_TCB_LOCK(stcb);
1596			}
1597			return (stcb);
1598		}
1599	}
1600	return (NULL);
1601}
1602
1603
1604struct sctp_tcb *
1605sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1606{
1607	struct sctp_tcb *stcb;
1608
1609	SCTP_INP_RLOCK(inp);
1610	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1611	SCTP_INP_RUNLOCK(inp);
1612	return (stcb);
1613}
1614
1615
1616/*
1617 * Endpoint probe expects that the INP_INFO is locked.
1618 */
1619static struct sctp_inpcb *
1620sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1621    uint16_t lport, uint32_t vrf_id)
1622{
1623	struct sctp_inpcb *inp;
1624	struct sctp_laddr *laddr;
1625
1626#ifdef INET
1627	struct sockaddr_in *sin;
1628
1629#endif
1630#ifdef INET6
1631	struct sockaddr_in6 *sin6;
1632	struct sockaddr_in6 *intf_addr6;
1633
1634#endif
1635	int fnd;
1636
1637#ifdef INET
1638	sin = NULL;
1639#endif
1640#ifdef INET6
1641	sin6 = NULL;
1642#endif
1643	switch (nam->sa_family) {
1644#ifdef INET
1645	case AF_INET:
1646		sin = (struct sockaddr_in *)nam;
1647		break;
1648#endif
1649#ifdef INET6
1650	case AF_INET6:
1651		sin6 = (struct sockaddr_in6 *)nam;
1652		break;
1653#endif
1654	default:
1655		/* unsupported family */
1656		return (NULL);
1657	}
1658
1659	if (head == NULL)
1660		return (NULL);
1661
1662	LIST_FOREACH(inp, head, sctp_hash) {
1663		SCTP_INP_RLOCK(inp);
1664		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1665			SCTP_INP_RUNLOCK(inp);
1666			continue;
1667		}
1668		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1669		    (inp->sctp_lport == lport)) {
1670			/* got it */
1671			switch (nam->sa_family) {
1672#ifdef INET
1673			case AF_INET:
1674				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1675				    SCTP_IPV6_V6ONLY(inp)) {
1676					/*
1677					 * IPv4 on a IPv6 socket with ONLY
1678					 * IPv6 set
1679					 */
1680					SCTP_INP_RUNLOCK(inp);
1681					continue;
1682				}
1683				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1684				    &sin->sin_addr) != 0) {
1685					SCTP_INP_RUNLOCK(inp);
1686					continue;
1687				}
1688				break;
1689#endif
1690#ifdef INET6
1691			case AF_INET6:
1692				/*
1693				 * A V6 address and the endpoint is NOT
1694				 * bound V6
1695				 */
1696				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1697					SCTP_INP_RUNLOCK(inp);
1698					continue;
1699				}
1700				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1701				    &sin6->sin6_addr) != 0) {
1702					SCTP_INP_RUNLOCK(inp);
1703					continue;
1704				}
1705				break;
1706#endif
1707			default:
1708				break;
1709			}
1710			/* does a VRF id match? */
1711			fnd = 0;
1712			if (inp->def_vrf_id == vrf_id)
1713				fnd = 1;
1714
1715			SCTP_INP_RUNLOCK(inp);
1716			if (!fnd)
1717				continue;
1718			return (inp);
1719		}
1720		SCTP_INP_RUNLOCK(inp);
1721	}
1722	switch (nam->sa_family) {
1723#ifdef INET
1724	case AF_INET:
1725		if (sin->sin_addr.s_addr == INADDR_ANY) {
1726			/* Can't hunt for one that has no address specified */
1727			return (NULL);
1728		}
1729		break;
1730#endif
1731#ifdef INET6
1732	case AF_INET6:
1733		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1734			/* Can't hunt for one that has no address specified */
1735			return (NULL);
1736		}
1737		break;
1738#endif
1739	default:
1740		break;
1741	}
1742	/*
1743	 * ok, not bound to all so see if we can find a EP bound to this
1744	 * address.
1745	 */
1746	LIST_FOREACH(inp, head, sctp_hash) {
1747		SCTP_INP_RLOCK(inp);
1748		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1749			SCTP_INP_RUNLOCK(inp);
1750			continue;
1751		}
1752		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1753			SCTP_INP_RUNLOCK(inp);
1754			continue;
1755		}
1756		/*
1757		 * Ok this could be a likely candidate, look at all of its
1758		 * addresses
1759		 */
1760		if (inp->sctp_lport != lport) {
1761			SCTP_INP_RUNLOCK(inp);
1762			continue;
1763		}
1764		/* does a VRF id match? */
1765		fnd = 0;
1766		if (inp->def_vrf_id == vrf_id)
1767			fnd = 1;
1768
1769		if (!fnd) {
1770			SCTP_INP_RUNLOCK(inp);
1771			continue;
1772		}
1773		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1774			if (laddr->ifa == NULL) {
1775				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1776				    __func__);
1777				continue;
1778			}
1779			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1780			    (void *)laddr->ifa);
1781			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1782				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1783				continue;
1784			}
1785			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1786				/* possible, see if it matches */
1787				switch (nam->sa_family) {
1788#ifdef INET
1789				case AF_INET:
1790					if (sin->sin_addr.s_addr ==
1791					    laddr->ifa->address.sin.sin_addr.s_addr) {
1792						SCTP_INP_RUNLOCK(inp);
1793						return (inp);
1794					}
1795					break;
1796#endif
1797#ifdef INET6
1798				case AF_INET6:
1799					intf_addr6 = &laddr->ifa->address.sin6;
1800					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1801					    intf_addr6)) {
1802						SCTP_INP_RUNLOCK(inp);
1803						return (inp);
1804					}
1805					break;
1806#endif
1807				}
1808			}
1809		}
1810		SCTP_INP_RUNLOCK(inp);
1811	}
1812	return (NULL);
1813}
1814
1815
1816static struct sctp_inpcb *
1817sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1818{
1819	struct sctppcbhead *head;
1820	struct sctp_inpcb *t_inp;
1821	int fnd;
1822
1823	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1824	    SCTP_BASE_INFO(hashmark))];
1825	LIST_FOREACH(t_inp, head, sctp_hash) {
1826		if (t_inp->sctp_lport != lport) {
1827			continue;
1828		}
1829		/* is it in the VRF in question */
1830		fnd = 0;
1831		if (t_inp->def_vrf_id == vrf_id)
1832			fnd = 1;
1833		if (!fnd)
1834			continue;
1835
1836		/* This one is in use. */
1837		/* check the v6/v4 binding issue */
1838		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1839		    SCTP_IPV6_V6ONLY(t_inp)) {
1840			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1841				/* collision in V6 space */
1842				return (t_inp);
1843			} else {
1844				/* inp is BOUND_V4 no conflict */
1845				continue;
1846			}
1847		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1848			/* t_inp is bound v4 and v6, conflict always */
1849			return (t_inp);
1850		} else {
1851			/* t_inp is bound only V4 */
1852			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1853			    SCTP_IPV6_V6ONLY(inp)) {
1854				/* no conflict */
1855				continue;
1856			}
1857			/* else fall through to conflict */
1858		}
1859		return (t_inp);
1860	}
1861	return (NULL);
1862}
1863
1864
1865int
1866sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1867{
1868	/* For 1-2-1 with port reuse */
1869	struct sctppcbhead *head;
1870	struct sctp_inpcb *tinp, *ninp;
1871
1872	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1873		/* only works with port reuse on */
1874		return (-1);
1875	}
1876	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1877		return (0);
1878	}
1879	SCTP_INP_RUNLOCK(inp);
1880	SCTP_INP_INFO_WLOCK();
1881	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1882	    SCTP_BASE_INFO(hashmark))];
1883	/* Kick out all non-listeners to the TCP hash */
1884	LIST_FOREACH_SAFE(tinp, head, sctp_hash, ninp) {
1885		if (tinp->sctp_lport != inp->sctp_lport) {
1886			continue;
1887		}
1888		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1889			continue;
1890		}
1891		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1892			continue;
1893		}
1894		if (tinp->sctp_socket->so_qlimit) {
1895			continue;
1896		}
1897		SCTP_INP_WLOCK(tinp);
1898		LIST_REMOVE(tinp, sctp_hash);
1899		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1900		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1901		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1902		SCTP_INP_WUNLOCK(tinp);
1903	}
1904	SCTP_INP_WLOCK(inp);
1905	/* Pull from where he was */
1906	LIST_REMOVE(inp, sctp_hash);
1907	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1908	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1909	LIST_INSERT_HEAD(head, inp, sctp_hash);
1910	SCTP_INP_WUNLOCK(inp);
1911	SCTP_INP_RLOCK(inp);
1912	SCTP_INP_INFO_WUNLOCK();
1913	return (0);
1914}
1915
1916
1917struct sctp_inpcb *
1918sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1919    uint32_t vrf_id)
1920{
1921	/*
1922	 * First we check the hash table to see if someone has this port
1923	 * bound with just the port.
1924	 */
1925	struct sctp_inpcb *inp;
1926	struct sctppcbhead *head;
1927	int lport;
1928	unsigned int i;
1929
1930#ifdef INET
1931	struct sockaddr_in *sin;
1932
1933#endif
1934#ifdef INET6
1935	struct sockaddr_in6 *sin6;
1936
1937#endif
1938
1939	switch (nam->sa_family) {
1940#ifdef INET
1941	case AF_INET:
1942		sin = (struct sockaddr_in *)nam;
1943		lport = sin->sin_port;
1944		break;
1945#endif
1946#ifdef INET6
1947	case AF_INET6:
1948		sin6 = (struct sockaddr_in6 *)nam;
1949		lport = sin6->sin6_port;
1950		break;
1951#endif
1952	default:
1953		return (NULL);
1954	}
1955	/*
1956	 * I could cheat here and just cast to one of the types but we will
1957	 * do it right. It also provides the check against an Unsupported
1958	 * type too.
1959	 */
1960	/* Find the head of the ALLADDR chain */
1961	if (have_lock == 0) {
1962		SCTP_INP_INFO_RLOCK();
1963	}
1964	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1965	    SCTP_BASE_INFO(hashmark))];
1966	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1967
1968	/*
1969	 * If the TCP model exists it could be that the main listening
1970	 * endpoint is gone but there still exists a connected socket for
1971	 * this guy. If so we can return the first one that we find. This
1972	 * may NOT be the correct one so the caller should be wary on the
1973	 * returned INP. Currently the only caller that sets find_tcp_pool
1974	 * is in bindx where we are verifying that a user CAN bind the
1975	 * address. He either has bound it already, or someone else has, or
1976	 * its open to bind, so this is good enough.
1977	 */
1978	if (inp == NULL && find_tcp_pool) {
1979		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1980			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1981			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1982			if (inp) {
1983				break;
1984			}
1985		}
1986	}
1987	if (inp) {
1988		SCTP_INP_INCR_REF(inp);
1989	}
1990	if (have_lock == 0) {
1991		SCTP_INP_INFO_RUNLOCK();
1992	}
1993	return (inp);
1994}
1995
1996
1997/*
1998 * Find an association for an endpoint with the pointer to whom you want to
1999 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
2000 * need to change the *to to some other struct like a mbuf...
2001 */
2002struct sctp_tcb *
2003sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
2004    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
2005    uint32_t vrf_id)
2006{
2007	struct sctp_inpcb *inp = NULL;
2008	struct sctp_tcb *stcb;
2009
2010	SCTP_INP_INFO_RLOCK();
2011	if (find_tcp_pool) {
2012		if (inp_p != NULL) {
2013			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
2014			    vrf_id);
2015		} else {
2016			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
2017			    vrf_id);
2018		}
2019		if (stcb != NULL) {
2020			SCTP_INP_INFO_RUNLOCK();
2021			return (stcb);
2022		}
2023	}
2024	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
2025	if (inp_p != NULL) {
2026		*inp_p = inp;
2027	}
2028	SCTP_INP_INFO_RUNLOCK();
2029	if (inp == NULL) {
2030		return (NULL);
2031	}
2032	/*
2033	 * ok, we have an endpoint, now lets find the assoc for it (if any)
2034	 * we now place the source address or from in the to of the find
2035	 * endpoint call. Since in reality this chain is used from the
2036	 * inbound packet side.
2037	 */
2038	if (inp_p != NULL) {
2039		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
2040		    NULL);
2041	} else {
2042		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
2043		    NULL);
2044	}
2045	return (stcb);
2046}
2047
2048
2049/*
2050 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
2051 * find all addresses that the sender has specified in any address list. Each
2052 * address will be used to lookup the TCB and see if one exits.
2053 */
2054static struct sctp_tcb *
2055sctp_findassociation_special_addr(struct mbuf *m, int offset,
2056    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
2057    struct sockaddr *dst)
2058{
2059	struct sctp_paramhdr *phdr, parm_buf;
2060
2061#if defined(INET) || defined(INET6)
2062	struct sctp_tcb *stcb;
2063	uint16_t ptype;
2064
2065#endif
2066	uint16_t plen;
2067
2068#ifdef INET
2069	struct sockaddr_in sin4;
2070
2071#endif
2072#ifdef INET6
2073	struct sockaddr_in6 sin6;
2074
2075#endif
2076
2077#ifdef INET
2078	memset(&sin4, 0, sizeof(sin4));
2079	sin4.sin_len = sizeof(sin4);
2080	sin4.sin_family = AF_INET;
2081	sin4.sin_port = sh->src_port;
2082#endif
2083#ifdef INET6
2084	memset(&sin6, 0, sizeof(sin6));
2085	sin6.sin6_len = sizeof(sin6);
2086	sin6.sin6_family = AF_INET6;
2087	sin6.sin6_port = sh->src_port;
2088#endif
2089
2090	offset += sizeof(struct sctp_init_chunk);
2091
2092	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2093	while (phdr != NULL) {
2094		/* now we must see if we want the parameter */
2095#if defined(INET) || defined(INET6)
2096		ptype = ntohs(phdr->param_type);
2097#endif
2098		plen = ntohs(phdr->param_length);
2099		if (plen == 0) {
2100			break;
2101		}
2102#ifdef INET
2103		if (ptype == SCTP_IPV4_ADDRESS &&
2104		    plen == sizeof(struct sctp_ipv4addr_param)) {
2105			/* Get the rest of the address */
2106			struct sctp_ipv4addr_param ip4_parm, *p4;
2107
2108			phdr = sctp_get_next_param(m, offset,
2109			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2110			if (phdr == NULL) {
2111				return (NULL);
2112			}
2113			p4 = (struct sctp_ipv4addr_param *)phdr;
2114			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2115			/* look it up */
2116			stcb = sctp_findassociation_ep_addr(inp_p,
2117			    (struct sockaddr *)&sin4, netp, dst, NULL);
2118			if (stcb != NULL) {
2119				return (stcb);
2120			}
2121		}
2122#endif
2123#ifdef INET6
2124		if (ptype == SCTP_IPV6_ADDRESS &&
2125		    plen == sizeof(struct sctp_ipv6addr_param)) {
2126			/* Get the rest of the address */
2127			struct sctp_ipv6addr_param ip6_parm, *p6;
2128
2129			phdr = sctp_get_next_param(m, offset,
2130			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2131			if (phdr == NULL) {
2132				return (NULL);
2133			}
2134			p6 = (struct sctp_ipv6addr_param *)phdr;
2135			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2136			/* look it up */
2137			stcb = sctp_findassociation_ep_addr(inp_p,
2138			    (struct sockaddr *)&sin6, netp, dst, NULL);
2139			if (stcb != NULL) {
2140				return (stcb);
2141			}
2142		}
2143#endif
2144		offset += SCTP_SIZE32(plen);
2145		phdr = sctp_get_next_param(m, offset, &parm_buf,
2146		    sizeof(parm_buf));
2147	}
2148	return (NULL);
2149}
2150
2151static struct sctp_tcb *
2152sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2153    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2154    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2155{
2156	/*
2157	 * Use my vtag to hash. If we find it we then verify the source addr
2158	 * is in the assoc. If all goes well we save a bit on rec of a
2159	 * packet.
2160	 */
2161	struct sctpasochead *head;
2162	struct sctp_nets *net;
2163	struct sctp_tcb *stcb;
2164
2165	SCTP_INP_INFO_RLOCK();
2166	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2167	    SCTP_BASE_INFO(hashasocmark))];
2168	LIST_FOREACH(stcb, head, sctp_asocs) {
2169		SCTP_INP_RLOCK(stcb->sctp_ep);
2170		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2171			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2172			continue;
2173		}
2174		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2175			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2176			continue;
2177		}
2178		SCTP_TCB_LOCK(stcb);
2179		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2180		if (stcb->asoc.my_vtag == vtag) {
2181			/* candidate */
2182			if (stcb->rport != rport) {
2183				SCTP_TCB_UNLOCK(stcb);
2184				continue;
2185			}
2186			if (stcb->sctp_ep->sctp_lport != lport) {
2187				SCTP_TCB_UNLOCK(stcb);
2188				continue;
2189			}
2190			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2191				SCTP_TCB_UNLOCK(stcb);
2192				continue;
2193			}
2194			/* RRS:Need toaddr check here */
2195			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2196				/* Endpoint does not own this address */
2197				SCTP_TCB_UNLOCK(stcb);
2198				continue;
2199			}
2200			if (remote_tag) {
2201				/*
2202				 * If we have both vtags that's all we match
2203				 * on
2204				 */
2205				if (stcb->asoc.peer_vtag == remote_tag) {
2206					/*
2207					 * If both tags match we consider it
2208					 * conclusive and check NO
2209					 * source/destination addresses
2210					 */
2211					goto conclusive;
2212				}
2213			}
2214			if (skip_src_check) {
2215		conclusive:
2216				if (from) {
2217					*netp = sctp_findnet(stcb, from);
2218				} else {
2219					*netp = NULL;	/* unknown */
2220				}
2221				if (inp_p)
2222					*inp_p = stcb->sctp_ep;
2223				SCTP_INP_INFO_RUNLOCK();
2224				return (stcb);
2225			}
2226			net = sctp_findnet(stcb, from);
2227			if (net) {
2228				/* yep its him. */
2229				*netp = net;
2230				SCTP_STAT_INCR(sctps_vtagexpress);
2231				*inp_p = stcb->sctp_ep;
2232				SCTP_INP_INFO_RUNLOCK();
2233				return (stcb);
2234			} else {
2235				/*
2236				 * not him, this should only happen in rare
2237				 * cases so I peg it.
2238				 */
2239				SCTP_STAT_INCR(sctps_vtagbogus);
2240			}
2241		}
2242		SCTP_TCB_UNLOCK(stcb);
2243	}
2244	SCTP_INP_INFO_RUNLOCK();
2245	return (NULL);
2246}
2247
2248
2249/*
2250 * Find an association with the pointer to the inbound IP packet. This can be
2251 * a IPv4 or IPv6 packet.
2252 */
2253struct sctp_tcb *
2254sctp_findassociation_addr(struct mbuf *m, int offset,
2255    struct sockaddr *src, struct sockaddr *dst,
2256    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2257    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2258{
2259	struct sctp_tcb *stcb;
2260	struct sctp_inpcb *inp;
2261
2262	if (sh->v_tag) {
2263		/* we only go down this path if vtag is non-zero */
2264		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2265		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2266		if (stcb) {
2267			return (stcb);
2268		}
2269	}
2270	if (inp_p) {
2271		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2272		    1, vrf_id);
2273		inp = *inp_p;
2274	} else {
2275		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2276		    1, vrf_id);
2277	}
2278	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2279	if (stcb == NULL && inp) {
2280		/* Found a EP but not this address */
2281		if ((ch->chunk_type == SCTP_INITIATION) ||
2282		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2283			/*-
2284			 * special hook, we do NOT return linp or an
2285			 * association that is linked to an existing
2286			 * association that is under the TCP pool (i.e. no
2287			 * listener exists). The endpoint finding routine
2288			 * will always find a listener before examining the
2289			 * TCP pool.
2290			 */
2291			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2292				if (inp_p) {
2293					*inp_p = NULL;
2294				}
2295				return (NULL);
2296			}
2297			stcb = sctp_findassociation_special_addr(m,
2298			    offset, sh, &inp, netp, dst);
2299			if (inp_p != NULL) {
2300				*inp_p = inp;
2301			}
2302		}
2303	}
2304	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2305	return (stcb);
2306}
2307
2308/*
2309 * lookup an association by an ASCONF lookup address.
2310 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2311 */
2312struct sctp_tcb *
2313sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2314    struct sockaddr *dst, struct sctphdr *sh,
2315    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2316{
2317	struct sctp_tcb *stcb;
2318	union sctp_sockstore remote_store;
2319	struct sctp_paramhdr parm_buf, *phdr;
2320	int ptype;
2321	int zero_address = 0;
2322
2323#ifdef INET
2324	struct sockaddr_in *sin;
2325
2326#endif
2327#ifdef INET6
2328	struct sockaddr_in6 *sin6;
2329
2330#endif
2331
2332	memset(&remote_store, 0, sizeof(remote_store));
2333	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2334	    &parm_buf, sizeof(struct sctp_paramhdr));
2335	if (phdr == NULL) {
2336		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2337		    __func__);
2338		return NULL;
2339	}
2340	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2341	/* get the correlation address */
2342	switch (ptype) {
2343#ifdef INET6
2344	case SCTP_IPV6_ADDRESS:
2345		{
2346			/* ipv6 address param */
2347			struct sctp_ipv6addr_param *p6, p6_buf;
2348
2349			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2350				return NULL;
2351			}
2352			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2353			    offset + sizeof(struct sctp_asconf_chunk),
2354			    &p6_buf.ph, sizeof(*p6));
2355			if (p6 == NULL) {
2356				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2357				    __func__);
2358				return (NULL);
2359			}
2360			sin6 = &remote_store.sin6;
2361			sin6->sin6_family = AF_INET6;
2362			sin6->sin6_len = sizeof(*sin6);
2363			sin6->sin6_port = sh->src_port;
2364			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2365			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2366				zero_address = 1;
2367			break;
2368		}
2369#endif
2370#ifdef INET
2371	case SCTP_IPV4_ADDRESS:
2372		{
2373			/* ipv4 address param */
2374			struct sctp_ipv4addr_param *p4, p4_buf;
2375
2376			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2377				return NULL;
2378			}
2379			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2380			    offset + sizeof(struct sctp_asconf_chunk),
2381			    &p4_buf.ph, sizeof(*p4));
2382			if (p4 == NULL) {
2383				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2384				    __func__);
2385				return (NULL);
2386			}
2387			sin = &remote_store.sin;
2388			sin->sin_family = AF_INET;
2389			sin->sin_len = sizeof(*sin);
2390			sin->sin_port = sh->src_port;
2391			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2392			if (sin->sin_addr.s_addr == INADDR_ANY)
2393				zero_address = 1;
2394			break;
2395		}
2396#endif
2397	default:
2398		/* invalid address param type */
2399		return NULL;
2400	}
2401
2402	if (zero_address) {
2403		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2404		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2405		if (stcb != NULL) {
2406			SCTP_INP_DECR_REF(*inp_p);
2407		}
2408	} else {
2409		stcb = sctp_findassociation_ep_addr(inp_p,
2410		    &remote_store.sa, netp,
2411		    dst, NULL);
2412	}
2413	return (stcb);
2414}
2415
2416
2417/*
2418 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2419 * addresses. This way if we don't get a bind we by default pick a ephemeral
2420 * port with all addresses bound.
2421 */
2422int
2423sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2424{
2425	/*
2426	 * we get called when a new endpoint starts up. We need to allocate
2427	 * the sctp_inpcb structure from the zone and init it. Mark it as
2428	 * unbound and find a port that we can use as an ephemeral with
2429	 * INADDR_ANY. If the user binds later no problem we can then add in
2430	 * the specific addresses. And setup the default parameters for the
2431	 * EP.
2432	 */
2433	int i, error;
2434	struct sctp_inpcb *inp;
2435	struct sctp_pcb *m;
2436	struct timeval time;
2437	sctp_sharedkey_t *null_key;
2438
2439	error = 0;
2440
2441	SCTP_INP_INFO_WLOCK();
2442	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2443	if (inp == NULL) {
2444		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2445		SCTP_INP_INFO_WUNLOCK();
2446		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2447		return (ENOBUFS);
2448	}
2449	/* zap it */
2450	bzero(inp, sizeof(*inp));
2451
2452	/* bump generations */
2453	/* setup socket pointers */
2454	inp->sctp_socket = so;
2455	inp->ip_inp.inp.inp_socket = so;
2456	inp->ip_inp.inp.inp_cred = crhold(so->so_cred);
2457#ifdef INET6
2458	if (INP_SOCKAF(so) == AF_INET6) {
2459		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2460			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2461		}
2462		if (MODULE_GLOBAL(ip6_v6only)) {
2463			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2464		}
2465	}
2466#endif
2467	inp->sctp_associd_counter = 1;
2468	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2469	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2470	inp->max_cwnd = 0;
2471	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2472	inp->ecn_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_ecn_enable);
2473	inp->prsctp_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pr_enable);
2474	inp->auth_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_auth_enable);
2475	inp->asconf_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_asconf_enable);
2476	inp->reconfig_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_reconfig_enable);
2477	inp->nrsack_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_nrsack_enable);
2478	inp->pktdrop_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pktdrop_enable);
2479	inp->idata_supported = 0;
2480
2481	inp->fibnum = so->so_fibnum;
2482	/* init the small hash table we use to track asocid <-> tcb */
2483	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2484	if (inp->sctp_asocidhash == NULL) {
2485		crfree(inp->ip_inp.inp.inp_cred);
2486		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2487		SCTP_INP_INFO_WUNLOCK();
2488		return (ENOBUFS);
2489	}
2490#ifdef IPSEC
2491	error = ipsec_init_policy(so, &inp->ip_inp.inp.inp_sp);
2492	if (error != 0) {
2493		crfree(inp->ip_inp.inp.inp_cred);
2494		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2495		SCTP_INP_INFO_WUNLOCK();
2496		return error;
2497	}
2498#endif				/* IPSEC */
2499	SCTP_INCR_EP_COUNT();
2500	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2501	SCTP_INP_INFO_WUNLOCK();
2502
2503	so->so_pcb = (caddr_t)inp;
2504
2505	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2506		/* UDP style socket */
2507		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2508		    SCTP_PCB_FLAGS_UNBOUND);
2509		/* Be sure it is NON-BLOCKING IO for UDP */
2510		/* SCTP_SET_SO_NBIO(so); */
2511	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2512		/* TCP style socket */
2513		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2514		    SCTP_PCB_FLAGS_UNBOUND);
2515		/* Be sure we have blocking IO by default */
2516		SCTP_CLEAR_SO_NBIO(so);
2517	} else {
2518		/*
2519		 * unsupported socket type (RAW, etc)- in case we missed it
2520		 * in protosw
2521		 */
2522		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2523		so->so_pcb = NULL;
2524		crfree(inp->ip_inp.inp.inp_cred);
2525#ifdef IPSEC
2526		ipsec_delete_pcbpolicy(&inp->ip_inp.inp);
2527#endif
2528		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2529		return (EOPNOTSUPP);
2530	}
2531	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2532		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2533		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2534	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2535		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2536		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2537	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2538		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2539		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2540	}
2541	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2542	    &inp->sctp_hashmark);
2543	if (inp->sctp_tcbhash == NULL) {
2544		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2545		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2546		so->so_pcb = NULL;
2547		crfree(inp->ip_inp.inp.inp_cred);
2548#ifdef IPSEC
2549		ipsec_delete_pcbpolicy(&inp->ip_inp.inp);
2550#endif
2551		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2552		return (ENOBUFS);
2553	}
2554	inp->def_vrf_id = vrf_id;
2555
2556	SCTP_INP_INFO_WLOCK();
2557	SCTP_INP_LOCK_INIT(inp);
2558	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2559	SCTP_INP_READ_INIT(inp);
2560	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2561	/* lock the new ep */
2562	SCTP_INP_WLOCK(inp);
2563
2564	/* add it to the info area */
2565	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2566	SCTP_INP_INFO_WUNLOCK();
2567
2568	TAILQ_INIT(&inp->read_queue);
2569	LIST_INIT(&inp->sctp_addr_list);
2570
2571	LIST_INIT(&inp->sctp_asoc_list);
2572
2573#ifdef SCTP_TRACK_FREED_ASOCS
2574	/* TEMP CODE */
2575	LIST_INIT(&inp->sctp_asoc_free_list);
2576#endif
2577	/* Init the timer structure for signature change */
2578	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2579	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2580
2581	/* now init the actual endpoint default data */
2582	m = &inp->sctp_ep;
2583
2584	/* setup the base timeout information */
2585	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2586	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2587	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2588	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2589	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2590	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2591	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2592	/* all max/min max are in ms */
2593	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2594	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2595	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2596	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2597	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2598	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2599	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2600	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2601	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2602	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2603	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2604	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2605	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2606
2607	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2608	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2609	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2610	/* number of streams to pre-open on a association */
2611	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2612
2613	/* Add adaptation cookie */
2614	m->adaptation_layer_indicator = 0;
2615	m->adaptation_layer_indicator_provided = 0;
2616
2617	/* seed random number generator */
2618	m->random_counter = 1;
2619	m->store_at = SCTP_SIGNATURE_SIZE;
2620	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2621	sctp_fill_random_store(m);
2622
2623	/* Minimum cookie size */
2624	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2625	    sizeof(struct sctp_state_cookie);
2626	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2627
2628	/* Setup the initial secret */
2629	(void)SCTP_GETTIME_TIMEVAL(&time);
2630	m->time_of_secret_change = time.tv_sec;
2631
2632	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2633		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2634	}
2635	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2636
2637	/* How long is a cookie good for ? */
2638	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2639	/*
2640	 * Initialize authentication parameters
2641	 */
2642	m->local_hmacs = sctp_default_supported_hmaclist();
2643	m->local_auth_chunks = sctp_alloc_chunklist();
2644	if (inp->asconf_supported) {
2645		sctp_auth_add_chunk(SCTP_ASCONF, m->local_auth_chunks);
2646		sctp_auth_add_chunk(SCTP_ASCONF_ACK, m->local_auth_chunks);
2647	}
2648	m->default_dscp = 0;
2649#ifdef INET6
2650	m->default_flowlabel = 0;
2651#endif
2652	m->port = 0;		/* encapsulation disabled by default */
2653	LIST_INIT(&m->shared_keys);
2654	/* add default NULL key as key id 0 */
2655	null_key = sctp_alloc_sharedkey();
2656	sctp_insert_sharedkey(&m->shared_keys, null_key);
2657	SCTP_INP_WUNLOCK(inp);
2658#ifdef SCTP_LOG_CLOSING
2659	sctp_log_closing(inp, NULL, 12);
2660#endif
2661	return (error);
2662}
2663
2664
2665void
2666sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2667    struct sctp_tcb *stcb)
2668{
2669	struct sctp_nets *net;
2670	uint16_t lport, rport;
2671	struct sctppcbhead *head;
2672	struct sctp_laddr *laddr, *oladdr;
2673
2674	atomic_add_int(&stcb->asoc.refcnt, 1);
2675	SCTP_TCB_UNLOCK(stcb);
2676	SCTP_INP_INFO_WLOCK();
2677	SCTP_INP_WLOCK(old_inp);
2678	SCTP_INP_WLOCK(new_inp);
2679	SCTP_TCB_LOCK(stcb);
2680	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2681
2682	new_inp->sctp_ep.time_of_secret_change =
2683	    old_inp->sctp_ep.time_of_secret_change;
2684	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2685	    sizeof(old_inp->sctp_ep.secret_key));
2686	new_inp->sctp_ep.current_secret_number =
2687	    old_inp->sctp_ep.current_secret_number;
2688	new_inp->sctp_ep.last_secret_number =
2689	    old_inp->sctp_ep.last_secret_number;
2690	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2691
2692	/* make it so new data pours into the new socket */
2693	stcb->sctp_socket = new_inp->sctp_socket;
2694	stcb->sctp_ep = new_inp;
2695
2696	/* Copy the port across */
2697	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2698	rport = stcb->rport;
2699	/* Pull the tcb from the old association */
2700	LIST_REMOVE(stcb, sctp_tcbhash);
2701	LIST_REMOVE(stcb, sctp_tcblist);
2702	if (stcb->asoc.in_asocid_hash) {
2703		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2704	}
2705	/* Now insert the new_inp into the TCP connected hash */
2706	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2707
2708	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2709	/* Its safe to access */
2710	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2711
2712	/* Now move the tcb into the endpoint list */
2713	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2714	/*
2715	 * Question, do we even need to worry about the ep-hash since we
2716	 * only have one connection? Probably not :> so lets get rid of it
2717	 * and not suck up any kernel memory in that.
2718	 */
2719	if (stcb->asoc.in_asocid_hash) {
2720		struct sctpasochead *lhd;
2721
2722		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2723		    new_inp->hashasocidmark)];
2724		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2725	}
2726	/* Ok. Let's restart timer. */
2727	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2728		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2729		    stcb, net);
2730	}
2731
2732	SCTP_INP_INFO_WUNLOCK();
2733	if (new_inp->sctp_tcbhash != NULL) {
2734		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2735		new_inp->sctp_tcbhash = NULL;
2736	}
2737	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2738		/* Subset bound, so copy in the laddr list from the old_inp */
2739		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2740			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2741			if (laddr == NULL) {
2742				/*
2743				 * Gak, what can we do? This assoc is really
2744				 * HOSED. We probably should send an abort
2745				 * here.
2746				 */
2747				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2748				continue;
2749			}
2750			SCTP_INCR_LADDR_COUNT();
2751			bzero(laddr, sizeof(*laddr));
2752			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2753			laddr->ifa = oladdr->ifa;
2754			atomic_add_int(&laddr->ifa->refcount, 1);
2755			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2756			    sctp_nxt_addr);
2757			new_inp->laddr_count++;
2758			if (oladdr == stcb->asoc.last_used_address) {
2759				stcb->asoc.last_used_address = laddr;
2760			}
2761		}
2762	}
2763	/*
2764	 * Now any running timers need to be adjusted since we really don't
2765	 * care if they are running or not just blast in the new_inp into
2766	 * all of them.
2767	 */
2768
2769	stcb->asoc.dack_timer.ep = (void *)new_inp;
2770	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2771	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2772	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2773	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2774	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2775	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2776	/* now what about the nets? */
2777	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2778		net->pmtu_timer.ep = (void *)new_inp;
2779		net->hb_timer.ep = (void *)new_inp;
2780		net->rxt_timer.ep = (void *)new_inp;
2781	}
2782	SCTP_INP_WUNLOCK(new_inp);
2783	SCTP_INP_WUNLOCK(old_inp);
2784}
2785
2786/*
2787 * insert an laddr entry with the given ifa for the desired list
2788 */
2789static int
2790sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
2791{
2792	struct sctp_laddr *laddr;
2793
2794	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2795	if (laddr == NULL) {
2796		/* out of memory? */
2797		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2798		return (EINVAL);
2799	}
2800	SCTP_INCR_LADDR_COUNT();
2801	bzero(laddr, sizeof(*laddr));
2802	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2803	laddr->ifa = ifa;
2804	laddr->action = act;
2805	atomic_add_int(&ifa->refcount, 1);
2806	/* insert it */
2807	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
2808
2809	return (0);
2810}
2811
2812/*
2813 * Remove an laddr entry from the local address list (on an assoc)
2814 */
2815static void
2816sctp_remove_laddr(struct sctp_laddr *laddr)
2817{
2818
2819	/* remove from the list */
2820	LIST_REMOVE(laddr, sctp_nxt_addr);
2821	sctp_free_ifa(laddr->ifa);
2822	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
2823	SCTP_DECR_LADDR_COUNT();
2824}
2825
2826
2827
2828/* sctp_ifap is used to bypass normal local address validation checks */
2829int
2830sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2831    struct sctp_ifa *sctp_ifap, struct thread *p)
2832{
2833	/* bind a ep to a socket address */
2834	struct sctppcbhead *head;
2835	struct sctp_inpcb *inp, *inp_tmp;
2836	struct inpcb *ip_inp;
2837	int port_reuse_active = 0;
2838	int bindall;
2839	uint16_t lport;
2840	int error;
2841	uint32_t vrf_id;
2842
2843	lport = 0;
2844	bindall = 1;
2845	inp = (struct sctp_inpcb *)so->so_pcb;
2846	ip_inp = (struct inpcb *)so->so_pcb;
2847#ifdef SCTP_DEBUG
2848	if (addr) {
2849		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2850		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2851		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2852		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2853	}
2854#endif
2855	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2856		/* already did a bind, subsequent binds NOT allowed ! */
2857		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2858		return (EINVAL);
2859	}
2860#ifdef INVARIANTS
2861	if (p == NULL)
2862		panic("null proc/thread");
2863#endif
2864	if (addr != NULL) {
2865		switch (addr->sa_family) {
2866#ifdef INET
2867		case AF_INET:
2868			{
2869				struct sockaddr_in *sin;
2870
2871				/* IPV6_V6ONLY socket? */
2872				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2873					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2874					return (EINVAL);
2875				}
2876				if (addr->sa_len != sizeof(*sin)) {
2877					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2878					return (EINVAL);
2879				}
2880				sin = (struct sockaddr_in *)addr;
2881				lport = sin->sin_port;
2882				/*
2883				 * For LOOPBACK the prison_local_ip4() call
2884				 * will transmute the ip address to the
2885				 * proper value.
2886				 */
2887				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2888					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2889					return (error);
2890				}
2891				if (sin->sin_addr.s_addr != INADDR_ANY) {
2892					bindall = 0;
2893				}
2894				break;
2895			}
2896#endif
2897#ifdef INET6
2898		case AF_INET6:
2899			{
2900				/*
2901				 * Only for pure IPv6 Address. (No IPv4
2902				 * Mapped!)
2903				 */
2904				struct sockaddr_in6 *sin6;
2905
2906				sin6 = (struct sockaddr_in6 *)addr;
2907
2908				if (addr->sa_len != sizeof(*sin6)) {
2909					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2910					return (EINVAL);
2911				}
2912				lport = sin6->sin6_port;
2913				/*
2914				 * For LOOPBACK the prison_local_ip6() call
2915				 * will transmute the ipv6 address to the
2916				 * proper value.
2917				 */
2918				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2919				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2920					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2921					return (error);
2922				}
2923				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2924					bindall = 0;
2925					/* KAME hack: embed scopeid */
2926					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2927						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2928						return (EINVAL);
2929					}
2930				}
2931				/* this must be cleared for ifa_ifwithaddr() */
2932				sin6->sin6_scope_id = 0;
2933				break;
2934			}
2935#endif
2936		default:
2937			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2938			return (EAFNOSUPPORT);
2939		}
2940	}
2941	SCTP_INP_INFO_WLOCK();
2942	SCTP_INP_WLOCK(inp);
2943	/* Setup a vrf_id to be the default for the non-bind-all case. */
2944	vrf_id = inp->def_vrf_id;
2945
2946	/* increase our count due to the unlock we do */
2947	SCTP_INP_INCR_REF(inp);
2948	if (lport) {
2949		/*
2950		 * Did the caller specify a port? if so we must see if an ep
2951		 * already has this one bound.
2952		 */
2953		/* got to be root to get at low ports */
2954		if (ntohs(lport) < IPPORT_RESERVED) {
2955			if (p && (error =
2956			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2957			    )) {
2958				SCTP_INP_DECR_REF(inp);
2959				SCTP_INP_WUNLOCK(inp);
2960				SCTP_INP_INFO_WUNLOCK();
2961				return (error);
2962			}
2963		}
2964		SCTP_INP_WUNLOCK(inp);
2965		if (bindall) {
2966			vrf_id = inp->def_vrf_id;
2967			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2968			if (inp_tmp != NULL) {
2969				/*
2970				 * lock guy returned and lower count note
2971				 * that we are not bound so inp_tmp should
2972				 * NEVER be inp. And it is this inp
2973				 * (inp_tmp) that gets the reference bump,
2974				 * so we must lower it.
2975				 */
2976				SCTP_INP_DECR_REF(inp_tmp);
2977				/* unlock info */
2978				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2979				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2980					/*
2981					 * Ok, must be one-2-one and
2982					 * allowing port re-use
2983					 */
2984					port_reuse_active = 1;
2985					goto continue_anyway;
2986				}
2987				SCTP_INP_DECR_REF(inp);
2988				SCTP_INP_INFO_WUNLOCK();
2989				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2990				return (EADDRINUSE);
2991			}
2992		} else {
2993			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2994			if (inp_tmp != NULL) {
2995				/*
2996				 * lock guy returned and lower count note
2997				 * that we are not bound so inp_tmp should
2998				 * NEVER be inp. And it is this inp
2999				 * (inp_tmp) that gets the reference bump,
3000				 * so we must lower it.
3001				 */
3002				SCTP_INP_DECR_REF(inp_tmp);
3003				/* unlock info */
3004				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
3005				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
3006					/*
3007					 * Ok, must be one-2-one and
3008					 * allowing port re-use
3009					 */
3010					port_reuse_active = 1;
3011					goto continue_anyway;
3012				}
3013				SCTP_INP_DECR_REF(inp);
3014				SCTP_INP_INFO_WUNLOCK();
3015				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3016				return (EADDRINUSE);
3017			}
3018		}
3019continue_anyway:
3020		SCTP_INP_WLOCK(inp);
3021		if (bindall) {
3022			/* verify that no lport is not used by a singleton */
3023			if ((port_reuse_active == 0) &&
3024			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
3025				/* Sorry someone already has this one bound */
3026				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
3027				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
3028					port_reuse_active = 1;
3029				} else {
3030					SCTP_INP_DECR_REF(inp);
3031					SCTP_INP_WUNLOCK(inp);
3032					SCTP_INP_INFO_WUNLOCK();
3033					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3034					return (EADDRINUSE);
3035				}
3036			}
3037		}
3038	} else {
3039		uint16_t first, last, candidate;
3040		uint16_t count;
3041		int done;
3042
3043		if (ip_inp->inp_flags & INP_HIGHPORT) {
3044			first = MODULE_GLOBAL(ipport_hifirstauto);
3045			last = MODULE_GLOBAL(ipport_hilastauto);
3046		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3047			if (p && (error =
3048			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3049			    )) {
3050				SCTP_INP_DECR_REF(inp);
3051				SCTP_INP_WUNLOCK(inp);
3052				SCTP_INP_INFO_WUNLOCK();
3053				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3054				return (error);
3055			}
3056			first = MODULE_GLOBAL(ipport_lowfirstauto);
3057			last = MODULE_GLOBAL(ipport_lowlastauto);
3058		} else {
3059			first = MODULE_GLOBAL(ipport_firstauto);
3060			last = MODULE_GLOBAL(ipport_lastauto);
3061		}
3062		if (first > last) {
3063			uint16_t temp;
3064
3065			temp = first;
3066			first = last;
3067			last = temp;
3068		}
3069		count = last - first + 1;	/* number of candidates */
3070		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3071
3072		done = 0;
3073		while (!done) {
3074			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3075				done = 1;
3076			}
3077			if (!done) {
3078				if (--count == 0) {
3079					SCTP_INP_DECR_REF(inp);
3080					SCTP_INP_WUNLOCK(inp);
3081					SCTP_INP_INFO_WUNLOCK();
3082					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3083					return (EADDRINUSE);
3084				}
3085				if (candidate == last)
3086					candidate = first;
3087				else
3088					candidate = candidate + 1;
3089			}
3090		}
3091		lport = htons(candidate);
3092	}
3093	SCTP_INP_DECR_REF(inp);
3094	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3095	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3096		/*
3097		 * this really should not happen. The guy did a non-blocking
3098		 * bind and then did a close at the same time.
3099		 */
3100		SCTP_INP_WUNLOCK(inp);
3101		SCTP_INP_INFO_WUNLOCK();
3102		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3103		return (EINVAL);
3104	}
3105	/* ok we look clear to give out this port, so lets setup the binding */
3106	if (bindall) {
3107		/* binding to all addresses, so just set in the proper flags */
3108		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3109		/* set the automatic addr changes from kernel flag */
3110		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3111			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3112			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3113		} else {
3114			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3115			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3116		}
3117		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3118			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3119		} else {
3120			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3121		}
3122		/*
3123		 * set the automatic mobility_base from kernel flag (by
3124		 * micchie)
3125		 */
3126		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3127			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3128			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3129		} else {
3130			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3131			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3132		}
3133		/*
3134		 * set the automatic mobility_fasthandoff from kernel flag
3135		 * (by micchie)
3136		 */
3137		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3138			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3139			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3140		} else {
3141			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3142			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3143		}
3144	} else {
3145		/*
3146		 * bind specific, make sure flags is off and add a new
3147		 * address structure to the sctp_addr_list inside the ep
3148		 * structure.
3149		 *
3150		 * We will need to allocate one and insert it at the head. The
3151		 * socketopt call can just insert new addresses in there as
3152		 * well. It will also have to do the embed scope kame hack
3153		 * too (before adding).
3154		 */
3155		struct sctp_ifa *ifa;
3156		union sctp_sockstore store;
3157
3158		memset(&store, 0, sizeof(store));
3159		switch (addr->sa_family) {
3160#ifdef INET
3161		case AF_INET:
3162			memcpy(&store.sin, addr, sizeof(struct sockaddr_in));
3163			store.sin.sin_port = 0;
3164			break;
3165#endif
3166#ifdef INET6
3167		case AF_INET6:
3168			memcpy(&store.sin6, addr, sizeof(struct sockaddr_in6));
3169			store.sin6.sin6_port = 0;
3170			break;
3171#endif
3172		default:
3173			break;
3174		}
3175		/*
3176		 * first find the interface with the bound address need to
3177		 * zero out the port to find the address! yuck! can't do
3178		 * this earlier since need port for sctp_pcb_findep()
3179		 */
3180		if (sctp_ifap != NULL) {
3181			ifa = sctp_ifap;
3182		} else {
3183			/*
3184			 * Note for BSD we hit here always other O/S's will
3185			 * pass things in via the sctp_ifap argument
3186			 * (Panda).
3187			 */
3188			ifa = sctp_find_ifa_by_addr(&store.sa,
3189			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3190		}
3191		if (ifa == NULL) {
3192			/* Can't find an interface with that address */
3193			SCTP_INP_WUNLOCK(inp);
3194			SCTP_INP_INFO_WUNLOCK();
3195			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3196			return (EADDRNOTAVAIL);
3197		}
3198#ifdef INET6
3199		if (addr->sa_family == AF_INET6) {
3200			/* GAK, more FIXME IFA lock? */
3201			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3202				/* Can't bind a non-existent addr. */
3203				SCTP_INP_WUNLOCK(inp);
3204				SCTP_INP_INFO_WUNLOCK();
3205				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3206				return (EINVAL);
3207			}
3208		}
3209#endif
3210		/* we're not bound all */
3211		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3212		/* allow bindx() to send ASCONF's for binding changes */
3213		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3214		/* clear automatic addr changes from kernel flag */
3215		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3216
3217		/* add this address to the endpoint list */
3218		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3219		if (error != 0) {
3220			SCTP_INP_WUNLOCK(inp);
3221			SCTP_INP_INFO_WUNLOCK();
3222			return (error);
3223		}
3224		inp->laddr_count++;
3225	}
3226	/* find the bucket */
3227	if (port_reuse_active) {
3228		/* Put it into tcp 1-2-1 hash */
3229		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3230		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3231	} else {
3232		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3233	}
3234	/* put it in the bucket */
3235	LIST_INSERT_HEAD(head, inp, sctp_hash);
3236	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3237	    (void *)head, ntohs(lport), port_reuse_active);
3238	/* set in the port */
3239	inp->sctp_lport = lport;
3240
3241	/* turn off just the unbound flag */
3242	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3243	SCTP_INP_WUNLOCK(inp);
3244	SCTP_INP_INFO_WUNLOCK();
3245	return (0);
3246}
3247
3248
3249static void
3250sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3251{
3252	struct sctp_iterator *it, *nit;
3253
3254	/*
3255	 * We enter with the only the ITERATOR_LOCK in place and a write
3256	 * lock on the inp_info stuff.
3257	 */
3258	it = sctp_it_ctl.cur_it;
3259	if (it && (it->vn != curvnet)) {
3260		/* Its not looking at our VNET */
3261		return;
3262	}
3263	if (it && (it->inp == inp)) {
3264		/*
3265		 * This is tricky and we hold the iterator lock, but when it
3266		 * returns and gets the lock (when we release it) the
3267		 * iterator will try to operate on inp. We need to stop that
3268		 * from happening. But of course the iterator has a
3269		 * reference on the stcb and inp. We can mark it and it will
3270		 * stop.
3271		 *
3272		 * If its a single iterator situation, we set the end iterator
3273		 * flag. Otherwise we set the iterator to go to the next
3274		 * inp.
3275		 *
3276		 */
3277		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3278			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3279		} else {
3280			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3281		}
3282	}
3283	/*
3284	 * Now go through and remove any single reference to our inp that
3285	 * may be still pending on the list
3286	 */
3287	SCTP_IPI_ITERATOR_WQ_LOCK();
3288	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3289		if (it->vn != curvnet) {
3290			continue;
3291		}
3292		if (it->inp == inp) {
3293			/* This one points to me is it inp specific? */
3294			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3295				/* Remove and free this one */
3296				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3297				    it, sctp_nxt_itr);
3298				if (it->function_atend != NULL) {
3299					(*it->function_atend) (it->pointer, it->val);
3300				}
3301				SCTP_FREE(it, SCTP_M_ITER);
3302			} else {
3303				it->inp = LIST_NEXT(it->inp, sctp_list);
3304				if (it->inp) {
3305					SCTP_INP_INCR_REF(it->inp);
3306				}
3307			}
3308			/*
3309			 * When its put in the refcnt is incremented so decr
3310			 * it
3311			 */
3312			SCTP_INP_DECR_REF(inp);
3313		}
3314	}
3315	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3316}
3317
3318/* release sctp_inpcb unbind the port */
3319void
3320sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3321{
3322	/*
3323	 * Here we free a endpoint. We must find it (if it is in the Hash
3324	 * table) and remove it from there. Then we must also find it in the
3325	 * overall list and remove it from there. After all removals are
3326	 * complete then any timer has to be stopped. Then start the actual
3327	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3328	 * all associations. d) finally the ep itself.
3329	 */
3330	struct sctp_tcb *asoc, *nasoc;
3331	struct sctp_laddr *laddr, *nladdr;
3332	struct inpcb *ip_pcb;
3333	struct socket *so;
3334	int being_refed = 0;
3335	struct sctp_queued_to_read *sq, *nsq;
3336	int cnt;
3337	sctp_sharedkey_t *shared_key, *nshared_key;
3338
3339
3340#ifdef SCTP_LOG_CLOSING
3341	sctp_log_closing(inp, NULL, 0);
3342#endif
3343	SCTP_ITERATOR_LOCK();
3344	/* mark any iterators on the list or being processed */
3345	sctp_iterator_inp_being_freed(inp);
3346	SCTP_ITERATOR_UNLOCK();
3347	so = inp->sctp_socket;
3348	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3349		/* been here before.. eeks.. get out of here */
3350		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3351#ifdef SCTP_LOG_CLOSING
3352		sctp_log_closing(inp, NULL, 1);
3353#endif
3354		return;
3355	}
3356	SCTP_ASOC_CREATE_LOCK(inp);
3357	SCTP_INP_INFO_WLOCK();
3358
3359	SCTP_INP_WLOCK(inp);
3360	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3361		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3362		/* socket is gone, so no more wakeups allowed */
3363		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3364		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3365		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3366
3367	}
3368	/* First time through we have the socket lock, after that no more. */
3369	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3370	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3371
3372	if (inp->control) {
3373		sctp_m_freem(inp->control);
3374		inp->control = NULL;
3375	}
3376	if (inp->pkt) {
3377		sctp_m_freem(inp->pkt);
3378		inp->pkt = NULL;
3379	}
3380	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3381					 * here but I will be nice :> (i.e.
3382					 * ip_pcb = ep;) */
3383	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3384		int cnt_in_sd;
3385
3386		cnt_in_sd = 0;
3387		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3388			SCTP_TCB_LOCK(asoc);
3389			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3390				/* Skip guys being freed */
3391				cnt_in_sd++;
3392				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3393					/*
3394					 * Special case - we did not start a
3395					 * kill timer on the asoc due to it
3396					 * was not closed. So go ahead and
3397					 * start it now.
3398					 */
3399					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3400					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3401				}
3402				SCTP_TCB_UNLOCK(asoc);
3403				continue;
3404			}
3405			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3406			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3407			    (asoc->asoc.total_output_queue_size == 0)) {
3408				/*
3409				 * If we have data in queue, we don't want
3410				 * to just free since the app may have done,
3411				 * send()/close or connect/send/close. And
3412				 * it wants the data to get across first.
3413				 */
3414				/* Just abandon things in the front states */
3415				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3416				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3417					cnt_in_sd++;
3418				}
3419				continue;
3420			}
3421			/* Disconnect the socket please */
3422			asoc->sctp_socket = NULL;
3423			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3424			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3425			    (asoc->asoc.control_pdapi) ||
3426			    (asoc->asoc.size_on_all_streams > 0) ||
3427			    (so && (so->so_rcv.sb_cc > 0))) {
3428				/* Left with Data unread */
3429				struct mbuf *op_err;
3430
3431				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3432				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3433				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3434				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
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				if (sctp_free_assoc(inp, asoc,
3440				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3441					cnt_in_sd++;
3442				}
3443				continue;
3444			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3445				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3446			    (asoc->asoc.stream_queue_cnt == 0)) {
3447				if ((*asoc->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete) (asoc, &asoc->asoc)) {
3448					goto abort_anyway;
3449				}
3450				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3451				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3452					struct sctp_nets *netp;
3453
3454					/*
3455					 * there is nothing queued to send,
3456					 * so I send shutdown
3457					 */
3458					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3459					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3460						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3461					}
3462					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3463					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3464					sctp_stop_timers_for_shutdown(asoc);
3465					if (asoc->asoc.alternate) {
3466						netp = asoc->asoc.alternate;
3467					} else {
3468						netp = asoc->asoc.primary_destination;
3469					}
3470					sctp_send_shutdown(asoc, netp);
3471					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3472					    netp);
3473					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3474					    asoc->asoc.primary_destination);
3475					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3476				}
3477			} else {
3478				/* mark into shutdown pending */
3479				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3480				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3481				    asoc->asoc.primary_destination);
3482				if ((*asoc->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete) (asoc, &asoc->asoc)) {
3483					asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3484				}
3485				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3486				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3487				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3488					struct mbuf *op_err;
3489
3490			abort_anyway:
3491					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3492					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3493					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3494					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3495					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3496					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3497						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3498					}
3499					if (sctp_free_assoc(inp, asoc,
3500					    SCTP_PCBFREE_NOFORCE,
3501					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3502						cnt_in_sd++;
3503					}
3504					continue;
3505				} else {
3506					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3507				}
3508			}
3509			cnt_in_sd++;
3510			SCTP_TCB_UNLOCK(asoc);
3511		}
3512		/* now is there some left in our SHUTDOWN state? */
3513		if (cnt_in_sd) {
3514#ifdef SCTP_LOG_CLOSING
3515			sctp_log_closing(inp, NULL, 2);
3516#endif
3517			inp->sctp_socket = NULL;
3518			SCTP_INP_WUNLOCK(inp);
3519			SCTP_ASOC_CREATE_UNLOCK(inp);
3520			SCTP_INP_INFO_WUNLOCK();
3521			return;
3522		}
3523	}
3524	inp->sctp_socket = NULL;
3525	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3526	    SCTP_PCB_FLAGS_UNBOUND) {
3527		/*
3528		 * ok, this guy has been bound. It's port is somewhere in
3529		 * the SCTP_BASE_INFO(hash table). Remove it!
3530		 */
3531		LIST_REMOVE(inp, sctp_hash);
3532		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3533	}
3534	/*
3535	 * If there is a timer running to kill us, forget it, since it may
3536	 * have a contest on the INP lock.. which would cause us to die ...
3537	 */
3538	cnt = 0;
3539	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3540		SCTP_TCB_LOCK(asoc);
3541		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3542			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3543				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3544				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3545			}
3546			cnt++;
3547			SCTP_TCB_UNLOCK(asoc);
3548			continue;
3549		}
3550		/* Free associations that are NOT killing us */
3551		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3552		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3553			struct mbuf *op_err;
3554
3555			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3556			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3557			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3558			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3559		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3560			cnt++;
3561			SCTP_TCB_UNLOCK(asoc);
3562			continue;
3563		}
3564		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3565		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3566			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3567		}
3568		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE,
3569		    SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3570			cnt++;
3571		}
3572	}
3573	if (cnt) {
3574		/* Ok we have someone out there that will kill us */
3575		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3576#ifdef SCTP_LOG_CLOSING
3577		sctp_log_closing(inp, NULL, 3);
3578#endif
3579		SCTP_INP_WUNLOCK(inp);
3580		SCTP_ASOC_CREATE_UNLOCK(inp);
3581		SCTP_INP_INFO_WUNLOCK();
3582		return;
3583	}
3584	if (SCTP_INP_LOCK_CONTENDED(inp))
3585		being_refed++;
3586	if (SCTP_INP_READ_CONTENDED(inp))
3587		being_refed++;
3588	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3589		being_refed++;
3590
3591	if ((inp->refcount) ||
3592	    (being_refed) ||
3593	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3594		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3595#ifdef SCTP_LOG_CLOSING
3596		sctp_log_closing(inp, NULL, 4);
3597#endif
3598		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3599		SCTP_INP_WUNLOCK(inp);
3600		SCTP_ASOC_CREATE_UNLOCK(inp);
3601		SCTP_INP_INFO_WUNLOCK();
3602		return;
3603	}
3604	inp->sctp_ep.signature_change.type = 0;
3605	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3606	/*
3607	 * Remove it from the list .. last thing we need a lock for.
3608	 */
3609	LIST_REMOVE(inp, sctp_list);
3610	SCTP_INP_WUNLOCK(inp);
3611	SCTP_ASOC_CREATE_UNLOCK(inp);
3612	SCTP_INP_INFO_WUNLOCK();
3613	/*
3614	 * Now we release all locks. Since this INP cannot be found anymore
3615	 * except possibly by the kill timer that might be running. We call
3616	 * the drain function here. It should hit the case were it sees the
3617	 * ACTIVE flag cleared and exit out freeing us to proceed and
3618	 * destroy everything.
3619	 */
3620	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3621		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3622	} else {
3623		/* Probably un-needed */
3624		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3625	}
3626
3627#ifdef SCTP_LOG_CLOSING
3628	sctp_log_closing(inp, NULL, 5);
3629#endif
3630
3631
3632	if ((inp->sctp_asocidhash) != NULL) {
3633		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3634		inp->sctp_asocidhash = NULL;
3635	}
3636	/* sa_ignore FREED_MEMORY */
3637	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3638		/* Its only abandoned if it had data left */
3639		if (sq->length)
3640			SCTP_STAT_INCR(sctps_left_abandon);
3641
3642		TAILQ_REMOVE(&inp->read_queue, sq, next);
3643		sctp_free_remote_addr(sq->whoFrom);
3644		if (so)
3645			so->so_rcv.sb_cc -= sq->length;
3646		if (sq->data) {
3647			sctp_m_freem(sq->data);
3648			sq->data = NULL;
3649		}
3650		/*
3651		 * no need to free the net count, since at this point all
3652		 * assoc's are gone.
3653		 */
3654		sctp_free_a_readq(NULL, sq);
3655	}
3656	/* Now the sctp_pcb things */
3657	/*
3658	 * free each asoc if it is not already closed/free. we can't use the
3659	 * macro here since le_next will get freed as part of the
3660	 * sctp_free_assoc() call.
3661	 */
3662#ifdef IPSEC
3663	ipsec_delete_pcbpolicy(ip_pcb);
3664#endif
3665	if (ip_pcb->inp_options) {
3666		(void)sctp_m_free(ip_pcb->inp_options);
3667		ip_pcb->inp_options = 0;
3668	}
3669#ifdef INET6
3670	if (ip_pcb->inp_vflag & INP_IPV6) {
3671		struct in6pcb *in6p;
3672
3673		in6p = (struct in6pcb *)inp;
3674		ip6_freepcbopts(in6p->in6p_outputopts);
3675	}
3676#endif				/* INET6 */
3677	ip_pcb->inp_vflag = 0;
3678	/* free up authentication fields */
3679	if (inp->sctp_ep.local_auth_chunks != NULL)
3680		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3681	if (inp->sctp_ep.local_hmacs != NULL)
3682		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3683
3684	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3685		LIST_REMOVE(shared_key, next);
3686		sctp_free_sharedkey(shared_key);
3687		/* sa_ignore FREED_MEMORY */
3688	}
3689
3690	/*
3691	 * if we have an address list the following will free the list of
3692	 * ifaddr's that are set into this ep. Again macro limitations here,
3693	 * since the LIST_FOREACH could be a bad idea.
3694	 */
3695	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3696		sctp_remove_laddr(laddr);
3697	}
3698
3699#ifdef SCTP_TRACK_FREED_ASOCS
3700	/* TEMP CODE */
3701	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3702		LIST_REMOVE(asoc, sctp_tcblist);
3703		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3704		SCTP_DECR_ASOC_COUNT();
3705	}
3706	/* *** END TEMP CODE *** */
3707#endif
3708	/* Now lets see about freeing the EP hash table. */
3709	if (inp->sctp_tcbhash != NULL) {
3710		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3711		inp->sctp_tcbhash = NULL;
3712	}
3713	/* Now we must put the ep memory back into the zone pool */
3714	crfree(inp->ip_inp.inp.inp_cred);
3715	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3716	SCTP_INP_LOCK_DESTROY(inp);
3717	SCTP_INP_READ_DESTROY(inp);
3718	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3719	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3720	SCTP_DECR_EP_COUNT();
3721}
3722
3723
3724struct sctp_nets *
3725sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3726{
3727	struct sctp_nets *net;
3728
3729	/* locate the address */
3730	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3731		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3732			return (net);
3733	}
3734	return (NULL);
3735}
3736
3737
3738int
3739sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3740{
3741	struct sctp_ifa *sctp_ifa;
3742
3743	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3744	if (sctp_ifa) {
3745		return (1);
3746	} else {
3747		return (0);
3748	}
3749}
3750
3751/*
3752 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3753 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3754 * stats of stuff.
3755 */
3756int
3757sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3758    struct sctp_nets **netp, uint16_t port, int set_scope, int from)
3759{
3760	/*
3761	 * The following is redundant to the same lines in the
3762	 * sctp_aloc_assoc() but is needed since others call the add address
3763	 * function
3764	 */
3765	struct sctp_nets *net, *netfirst;
3766	int addr_inscope;
3767
3768	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3769	    from);
3770	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3771
3772	netfirst = sctp_findnet(stcb, newaddr);
3773	if (netfirst) {
3774		/*
3775		 * Lie and return ok, we don't want to make the association
3776		 * go away for this behavior. It will happen in the TCP
3777		 * model in a connected socket. It does not reach the hash
3778		 * table until after the association is built so it can't be
3779		 * found. Mark as reachable, since the initial creation will
3780		 * have been cleared and the NOT_IN_ASSOC flag will have
3781		 * been added... and we don't want to end up removing it
3782		 * back out.
3783		 */
3784		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3785			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3786			    SCTP_ADDR_UNCONFIRMED);
3787		} else {
3788			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3789		}
3790
3791		return (0);
3792	}
3793	addr_inscope = 1;
3794	switch (newaddr->sa_family) {
3795#ifdef INET
3796	case AF_INET:
3797		{
3798			struct sockaddr_in *sin;
3799
3800			sin = (struct sockaddr_in *)newaddr;
3801			if (sin->sin_addr.s_addr == 0) {
3802				/* Invalid address */
3803				return (-1);
3804			}
3805			/* zero out the bzero area */
3806			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3807
3808			/* assure len is set */
3809			sin->sin_len = sizeof(struct sockaddr_in);
3810			if (set_scope) {
3811				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3812					stcb->asoc.scope.ipv4_local_scope = 1;
3813				}
3814			} else {
3815				/* Validate the address is in scope */
3816				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3817				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3818					addr_inscope = 0;
3819				}
3820			}
3821			break;
3822		}
3823#endif
3824#ifdef INET6
3825	case AF_INET6:
3826		{
3827			struct sockaddr_in6 *sin6;
3828
3829			sin6 = (struct sockaddr_in6 *)newaddr;
3830			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3831				/* Invalid address */
3832				return (-1);
3833			}
3834			/* assure len is set */
3835			sin6->sin6_len = sizeof(struct sockaddr_in6);
3836			if (set_scope) {
3837				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3838					stcb->asoc.scope.loopback_scope = 1;
3839					stcb->asoc.scope.local_scope = 0;
3840					stcb->asoc.scope.ipv4_local_scope = 1;
3841					stcb->asoc.scope.site_scope = 1;
3842				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3843					/*
3844					 * If the new destination is a
3845					 * LINK_LOCAL we must have common
3846					 * site scope. Don't set the local
3847					 * scope since we may not share all
3848					 * links, only loopback can do this.
3849					 * Links on the local network would
3850					 * also be on our private network
3851					 * for v4 too.
3852					 */
3853					stcb->asoc.scope.ipv4_local_scope = 1;
3854					stcb->asoc.scope.site_scope = 1;
3855				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3856					/*
3857					 * If the new destination is
3858					 * SITE_LOCAL then we must have site
3859					 * scope in common.
3860					 */
3861					stcb->asoc.scope.site_scope = 1;
3862				}
3863			} else {
3864				/* Validate the address is in scope */
3865				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3866				    (stcb->asoc.scope.loopback_scope == 0)) {
3867					addr_inscope = 0;
3868				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3869				    (stcb->asoc.scope.local_scope == 0)) {
3870					addr_inscope = 0;
3871				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3872				    (stcb->asoc.scope.site_scope == 0)) {
3873					addr_inscope = 0;
3874				}
3875			}
3876			break;
3877		}
3878#endif
3879	default:
3880		/* not supported family type */
3881		return (-1);
3882	}
3883	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3884	if (net == NULL) {
3885		return (-1);
3886	}
3887	SCTP_INCR_RADDR_COUNT();
3888	bzero(net, sizeof(struct sctp_nets));
3889	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3890	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3891	switch (newaddr->sa_family) {
3892#ifdef INET
3893	case AF_INET:
3894		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3895		break;
3896#endif
3897#ifdef INET6
3898	case AF_INET6:
3899		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3900		break;
3901#endif
3902	default:
3903		break;
3904	}
3905	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3906	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3907		stcb->asoc.scope.loopback_scope = 1;
3908		stcb->asoc.scope.ipv4_local_scope = 1;
3909		stcb->asoc.scope.local_scope = 0;
3910		stcb->asoc.scope.site_scope = 1;
3911		addr_inscope = 1;
3912	}
3913	net->failure_threshold = stcb->asoc.def_net_failure;
3914	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3915	if (addr_inscope == 0) {
3916		net->dest_state = (SCTP_ADDR_REACHABLE |
3917		    SCTP_ADDR_OUT_OF_SCOPE);
3918	} else {
3919		if (from == SCTP_ADDR_IS_CONFIRMED)
3920			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3921			net->dest_state = SCTP_ADDR_REACHABLE;
3922		else
3923			net->dest_state = SCTP_ADDR_REACHABLE |
3924			    SCTP_ADDR_UNCONFIRMED;
3925	}
3926	/*
3927	 * We set this to 0, the timer code knows that this means its an
3928	 * initial value
3929	 */
3930	net->rto_needed = 1;
3931	net->RTO = 0;
3932	net->RTO_measured = 0;
3933	stcb->asoc.numnets++;
3934	net->ref_count = 1;
3935	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3936	net->port = port;
3937	net->dscp = stcb->asoc.default_dscp;
3938#ifdef INET6
3939	net->flowlabel = stcb->asoc.default_flowlabel;
3940#endif
3941	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3942		net->dest_state |= SCTP_ADDR_NOHB;
3943	} else {
3944		net->dest_state &= ~SCTP_ADDR_NOHB;
3945	}
3946	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3947		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3948	} else {
3949		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3950	}
3951	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3952	/* Init the timer structure */
3953	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3954	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3955	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3956
3957	/* Now generate a route for this guy */
3958#ifdef INET6
3959	/* KAME hack: embed scopeid */
3960	if (newaddr->sa_family == AF_INET6) {
3961		struct sockaddr_in6 *sin6;
3962
3963		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3964		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3965		sin6->sin6_scope_id = 0;
3966	}
3967#endif
3968	SCTP_RTALLOC((sctp_route_t *) & net->ro,
3969	    stcb->asoc.vrf_id,
3970	    stcb->sctp_ep->fibnum);
3971
3972	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3973		/* Get source address */
3974		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3975		    stcb,
3976		    (sctp_route_t *) & net->ro,
3977		    net,
3978		    0,
3979		    stcb->asoc.vrf_id);
3980		if (net->ro._s_addr != NULL) {
3981			net->src_addr_selected = 1;
3982			/* Now get the interface MTU */
3983			if (net->ro._s_addr->ifn_p != NULL) {
3984				net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3985			}
3986		} else {
3987			net->src_addr_selected = 0;
3988		}
3989		if (net->mtu > 0) {
3990			uint32_t rmtu;
3991
3992			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3993			if (rmtu == 0) {
3994				/*
3995				 * Start things off to match mtu of
3996				 * interface please.
3997				 */
3998				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3999				    net->ro.ro_rt, net->mtu);
4000			} else {
4001				/*
4002				 * we take the route mtu over the interface,
4003				 * since the route may be leading out the
4004				 * loopback, or a different interface.
4005				 */
4006				net->mtu = rmtu;
4007			}
4008		}
4009	} else {
4010		net->src_addr_selected = 0;
4011	}
4012	if (net->mtu == 0) {
4013		switch (newaddr->sa_family) {
4014#ifdef INET
4015		case AF_INET:
4016			net->mtu = SCTP_DEFAULT_MTU;
4017			break;
4018#endif
4019#ifdef INET6
4020		case AF_INET6:
4021			net->mtu = 1280;
4022			break;
4023#endif
4024		default:
4025			break;
4026		}
4027	}
4028#if defined(INET) || defined(INET6)
4029	if (net->port) {
4030		net->mtu -= (uint32_t) sizeof(struct udphdr);
4031	}
4032#endif
4033	if (from == SCTP_ALLOC_ASOC) {
4034		stcb->asoc.smallest_mtu = net->mtu;
4035	}
4036	if (stcb->asoc.smallest_mtu > net->mtu) {
4037		sctp_pathmtu_adjustment(stcb, net->mtu);
4038	}
4039#ifdef INET6
4040	if (newaddr->sa_family == AF_INET6) {
4041		struct sockaddr_in6 *sin6;
4042
4043		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4044		(void)sa6_recoverscope(sin6);
4045	}
4046#endif
4047
4048	/* JRS - Use the congestion control given in the CC module */
4049	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4050		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4051
4052	/*
4053	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4054	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4055	 */
4056	net->find_pseudo_cumack = 1;
4057	net->find_rtx_pseudo_cumack = 1;
4058	/* Choose an initial flowid. */
4059	net->flowid = stcb->asoc.my_vtag ^
4060	    ntohs(stcb->rport) ^
4061	    ntohs(stcb->sctp_ep->sctp_lport);
4062	net->flowtype = M_HASHTYPE_OPAQUE_HASH;
4063	if (netp) {
4064		*netp = net;
4065	}
4066	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4067	if (net->ro.ro_rt == NULL) {
4068		/* Since we have no route put it at the back */
4069		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4070	} else if (netfirst == NULL) {
4071		/* We are the first one in the pool. */
4072		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4073	} else if (netfirst->ro.ro_rt == NULL) {
4074		/*
4075		 * First one has NO route. Place this one ahead of the first
4076		 * one.
4077		 */
4078		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4079	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4080		/*
4081		 * This one has a different interface than the one at the
4082		 * top of the list. Place it ahead.
4083		 */
4084		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4085	} else {
4086		/*
4087		 * Ok we have the same interface as the first one. Move
4088		 * forward until we find either a) one with a NULL route...
4089		 * insert ahead of that b) one with a different ifp.. insert
4090		 * after that. c) end of the list.. insert at the tail.
4091		 */
4092		struct sctp_nets *netlook;
4093
4094		do {
4095			netlook = TAILQ_NEXT(netfirst, sctp_next);
4096			if (netlook == NULL) {
4097				/* End of the list */
4098				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4099				break;
4100			} else if (netlook->ro.ro_rt == NULL) {
4101				/* next one has NO route */
4102				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4103				break;
4104			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4105				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4106				    net, sctp_next);
4107				break;
4108			}
4109			/* Shift forward */
4110			netfirst = netlook;
4111		} while (netlook != NULL);
4112	}
4113
4114	/* got to have a primary set */
4115	if (stcb->asoc.primary_destination == 0) {
4116		stcb->asoc.primary_destination = net;
4117	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4118		    (net->ro.ro_rt) &&
4119	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4120		/* No route to current primary adopt new primary */
4121		stcb->asoc.primary_destination = net;
4122	}
4123	/* Validate primary is first */
4124	net = TAILQ_FIRST(&stcb->asoc.nets);
4125	if ((net != stcb->asoc.primary_destination) &&
4126	    (stcb->asoc.primary_destination)) {
4127		/*
4128		 * first one on the list is NOT the primary sctp_cmpaddr()
4129		 * is much more efficient if the primary is the first on the
4130		 * list, make it so.
4131		 */
4132		TAILQ_REMOVE(&stcb->asoc.nets,
4133		    stcb->asoc.primary_destination, sctp_next);
4134		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4135		    stcb->asoc.primary_destination, sctp_next);
4136	}
4137	return (0);
4138}
4139
4140
4141static uint32_t
4142sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4143{
4144	uint32_t id;
4145	struct sctpasochead *head;
4146	struct sctp_tcb *lstcb;
4147
4148	SCTP_INP_WLOCK(inp);
4149try_again:
4150	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4151		/* TSNH */
4152		SCTP_INP_WUNLOCK(inp);
4153		return (0);
4154	}
4155	/*
4156	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4157	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4158	 */
4159	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4160		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4161	}
4162	id = inp->sctp_associd_counter;
4163	inp->sctp_associd_counter++;
4164	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4165	if (lstcb) {
4166		goto try_again;
4167	}
4168	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4169	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4170	stcb->asoc.in_asocid_hash = 1;
4171	SCTP_INP_WUNLOCK(inp);
4172	return id;
4173}
4174
4175/*
4176 * allocate an association and add it to the endpoint. The caller must be
4177 * careful to add all additional addresses once they are know right away or
4178 * else the assoc will be may experience a blackout scenario.
4179 */
4180struct sctp_tcb *
4181sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4182    int *error, uint32_t override_tag, uint32_t vrf_id,
4183    uint16_t o_streams, uint16_t port,
4184    struct thread *p
4185)
4186{
4187	/* note the p argument is only valid in unbound sockets */
4188
4189	struct sctp_tcb *stcb;
4190	struct sctp_association *asoc;
4191	struct sctpasochead *head;
4192	uint16_t rport;
4193	int err;
4194
4195	/*
4196	 * Assumption made here: Caller has done a
4197	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4198	 * address does not exist already.
4199	 */
4200	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4201		/* Hit max assoc, sorry no more */
4202		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4203		*error = ENOBUFS;
4204		return (NULL);
4205	}
4206	if (firstaddr == NULL) {
4207		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4208		*error = EINVAL;
4209		return (NULL);
4210	}
4211	SCTP_INP_RLOCK(inp);
4212	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4213	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4214	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4215		/*
4216		 * If its in the TCP pool, its NOT allowed to create an
4217		 * association. The parent listener needs to call
4218		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4219		 * off, or connected one does this.. its an error.
4220		 */
4221		SCTP_INP_RUNLOCK(inp);
4222		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4223		*error = EINVAL;
4224		return (NULL);
4225	}
4226	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4227	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4228		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4229		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4230			SCTP_INP_RUNLOCK(inp);
4231			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4232			*error = EINVAL;
4233			return (NULL);
4234		}
4235	}
4236	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4237#ifdef SCTP_DEBUG
4238	if (firstaddr) {
4239		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4240		switch (firstaddr->sa_family) {
4241#ifdef INET
4242		case AF_INET:
4243			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4244			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4245			break;
4246#endif
4247#ifdef INET6
4248		case AF_INET6:
4249			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4250			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4251			break;
4252#endif
4253		default:
4254			break;
4255		}
4256	} else {
4257		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4258	}
4259#endif				/* SCTP_DEBUG */
4260	switch (firstaddr->sa_family) {
4261#ifdef INET
4262	case AF_INET:
4263		{
4264			struct sockaddr_in *sin;
4265
4266			sin = (struct sockaddr_in *)firstaddr;
4267			if ((ntohs(sin->sin_port) == 0) ||
4268			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4269			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4270			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4271				/* Invalid address */
4272				SCTP_INP_RUNLOCK(inp);
4273				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4274				*error = EINVAL;
4275				return (NULL);
4276			}
4277			rport = sin->sin_port;
4278			break;
4279		}
4280#endif
4281#ifdef INET6
4282	case AF_INET6:
4283		{
4284			struct sockaddr_in6 *sin6;
4285
4286			sin6 = (struct sockaddr_in6 *)firstaddr;
4287			if ((ntohs(sin6->sin6_port) == 0) ||
4288			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4289			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4290				/* Invalid address */
4291				SCTP_INP_RUNLOCK(inp);
4292				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4293				*error = EINVAL;
4294				return (NULL);
4295			}
4296			rport = sin6->sin6_port;
4297			break;
4298		}
4299#endif
4300	default:
4301		/* not supported family type */
4302		SCTP_INP_RUNLOCK(inp);
4303		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4304		*error = EINVAL;
4305		return (NULL);
4306	}
4307	SCTP_INP_RUNLOCK(inp);
4308	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4309		/*
4310		 * If you have not performed a bind, then we need to do the
4311		 * ephemeral bind for you.
4312		 */
4313		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4314		    (struct sockaddr *)NULL,
4315		    (struct sctp_ifa *)NULL,
4316		    p
4317		    ))) {
4318			/* bind error, probably perm */
4319			*error = err;
4320			return (NULL);
4321		}
4322	}
4323	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4324	if (stcb == NULL) {
4325		/* out of memory? */
4326		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4327		*error = ENOMEM;
4328		return (NULL);
4329	}
4330	SCTP_INCR_ASOC_COUNT();
4331
4332	bzero(stcb, sizeof(*stcb));
4333	asoc = &stcb->asoc;
4334
4335	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4336	SCTP_TCB_LOCK_INIT(stcb);
4337	SCTP_TCB_SEND_LOCK_INIT(stcb);
4338	stcb->rport = rport;
4339	/* setup back pointer's */
4340	stcb->sctp_ep = inp;
4341	stcb->sctp_socket = inp->sctp_socket;
4342	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id, o_streams))) {
4343		/* failed */
4344		SCTP_TCB_LOCK_DESTROY(stcb);
4345		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4346		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4347		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4348		SCTP_DECR_ASOC_COUNT();
4349		*error = err;
4350		return (NULL);
4351	}
4352	/* and the port */
4353	SCTP_INP_INFO_WLOCK();
4354	SCTP_INP_WLOCK(inp);
4355	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4356		/* inpcb freed while alloc going on */
4357		SCTP_TCB_LOCK_DESTROY(stcb);
4358		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4359		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4360		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4361		SCTP_INP_WUNLOCK(inp);
4362		SCTP_INP_INFO_WUNLOCK();
4363		SCTP_DECR_ASOC_COUNT();
4364		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4365		*error = EINVAL;
4366		return (NULL);
4367	}
4368	SCTP_TCB_LOCK(stcb);
4369
4370	/* now that my_vtag is set, add it to the hash */
4371	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4372	/* put it in the bucket in the vtag hash of assoc's for the system */
4373	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4374	SCTP_INP_INFO_WUNLOCK();
4375
4376	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, port, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4377		/* failure.. memory error? */
4378		if (asoc->strmout) {
4379			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4380			asoc->strmout = NULL;
4381		}
4382		if (asoc->mapping_array) {
4383			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4384			asoc->mapping_array = NULL;
4385		}
4386		if (asoc->nr_mapping_array) {
4387			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4388			asoc->nr_mapping_array = NULL;
4389		}
4390		SCTP_DECR_ASOC_COUNT();
4391		SCTP_TCB_UNLOCK(stcb);
4392		SCTP_TCB_LOCK_DESTROY(stcb);
4393		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4394		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4395		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4396		SCTP_INP_WUNLOCK(inp);
4397		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4398		*error = ENOBUFS;
4399		return (NULL);
4400	}
4401	/* Init all the timers */
4402	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4403	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4404	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4405	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4406	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4407	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4408	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4409
4410	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4411	/* now file the port under the hash as well */
4412	if (inp->sctp_tcbhash != NULL) {
4413		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4414		    inp->sctp_hashmark)];
4415		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4416	}
4417	SCTP_INP_WUNLOCK(inp);
4418	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4419	return (stcb);
4420}
4421
4422
4423void
4424sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4425{
4426	struct sctp_association *asoc;
4427
4428	asoc = &stcb->asoc;
4429	asoc->numnets--;
4430	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4431	if (net == asoc->primary_destination) {
4432		/* Reset primary */
4433		struct sctp_nets *lnet;
4434
4435		lnet = TAILQ_FIRST(&asoc->nets);
4436		/*
4437		 * Mobility adaptation Ideally, if deleted destination is
4438		 * the primary, it becomes a fast retransmission trigger by
4439		 * the subsequent SET PRIMARY. (by micchie)
4440		 */
4441		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4442		    SCTP_MOBILITY_BASE) ||
4443		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4444		    SCTP_MOBILITY_FASTHANDOFF)) {
4445			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4446			if (asoc->deleted_primary != NULL) {
4447				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4448				goto out;
4449			}
4450			asoc->deleted_primary = net;
4451			atomic_add_int(&net->ref_count, 1);
4452			memset(&net->lastsa, 0, sizeof(net->lastsa));
4453			memset(&net->lastsv, 0, sizeof(net->lastsv));
4454			sctp_mobility_feature_on(stcb->sctp_ep,
4455			    SCTP_MOBILITY_PRIM_DELETED);
4456			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4457			    stcb->sctp_ep, stcb, NULL);
4458		}
4459out:
4460		/* Try to find a confirmed primary */
4461		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4462	}
4463	if (net == asoc->last_data_chunk_from) {
4464		/* Reset primary */
4465		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4466	}
4467	if (net == asoc->last_control_chunk_from) {
4468		/* Clear net */
4469		asoc->last_control_chunk_from = NULL;
4470	}
4471	if (net == stcb->asoc.alternate) {
4472		sctp_free_remote_addr(stcb->asoc.alternate);
4473		stcb->asoc.alternate = NULL;
4474	}
4475	sctp_free_remote_addr(net);
4476}
4477
4478/*
4479 * remove a remote endpoint address from an association, it will fail if the
4480 * address does not exist.
4481 */
4482int
4483sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4484{
4485	/*
4486	 * Here we need to remove a remote address. This is quite simple, we
4487	 * first find it in the list of address for the association
4488	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4489	 * on that item. Note we do not allow it to be removed if there are
4490	 * no other addresses.
4491	 */
4492	struct sctp_association *asoc;
4493	struct sctp_nets *net, *nnet;
4494
4495	asoc = &stcb->asoc;
4496
4497	/* locate the address */
4498	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4499		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4500			continue;
4501		}
4502		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4503		    remaddr)) {
4504			/* we found the guy */
4505			if (asoc->numnets < 2) {
4506				/* Must have at LEAST two remote addresses */
4507				return (-1);
4508			} else {
4509				sctp_remove_net(stcb, net);
4510				return (0);
4511			}
4512		}
4513	}
4514	/* not found. */
4515	return (-2);
4516}
4517
4518void
4519sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4520{
4521	struct sctpvtaghead *chain;
4522	struct sctp_tagblock *twait_block;
4523	int found = 0;
4524	int i;
4525
4526	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4527	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4528		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4529			if ((twait_block->vtag_block[i].v_tag == tag) &&
4530			    (twait_block->vtag_block[i].lport == lport) &&
4531			    (twait_block->vtag_block[i].rport == rport)) {
4532				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4533				twait_block->vtag_block[i].v_tag = 0;
4534				twait_block->vtag_block[i].lport = 0;
4535				twait_block->vtag_block[i].rport = 0;
4536				found = 1;
4537				break;
4538			}
4539		}
4540		if (found)
4541			break;
4542	}
4543}
4544
4545int
4546sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4547{
4548	struct sctpvtaghead *chain;
4549	struct sctp_tagblock *twait_block;
4550	int found = 0;
4551	int i;
4552
4553	SCTP_INP_INFO_WLOCK();
4554	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4555	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4556		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4557			if ((twait_block->vtag_block[i].v_tag == tag) &&
4558			    (twait_block->vtag_block[i].lport == lport) &&
4559			    (twait_block->vtag_block[i].rport == rport)) {
4560				found = 1;
4561				break;
4562			}
4563		}
4564		if (found)
4565			break;
4566	}
4567	SCTP_INP_INFO_WUNLOCK();
4568	return (found);
4569}
4570
4571
4572void
4573sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4574{
4575	struct sctpvtaghead *chain;
4576	struct sctp_tagblock *twait_block;
4577	struct timeval now;
4578	int set, i;
4579
4580	if (time == 0) {
4581		/* Its disabled */
4582		return;
4583	}
4584	(void)SCTP_GETTIME_TIMEVAL(&now);
4585	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4586	set = 0;
4587	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4588		/* Block(s) present, lets find space, and expire on the fly */
4589		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4590			if ((twait_block->vtag_block[i].v_tag == 0) &&
4591			    !set) {
4592				twait_block->vtag_block[i].tv_sec_at_expire =
4593				    now.tv_sec + time;
4594				twait_block->vtag_block[i].v_tag = tag;
4595				twait_block->vtag_block[i].lport = lport;
4596				twait_block->vtag_block[i].rport = rport;
4597				set = 1;
4598			} else if ((twait_block->vtag_block[i].v_tag) &&
4599			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4600				/* Audit expires this guy */
4601				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4602				twait_block->vtag_block[i].v_tag = 0;
4603				twait_block->vtag_block[i].lport = 0;
4604				twait_block->vtag_block[i].rport = 0;
4605				if (set == 0) {
4606					/* Reuse it for my new tag */
4607					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4608					twait_block->vtag_block[i].v_tag = tag;
4609					twait_block->vtag_block[i].lport = lport;
4610					twait_block->vtag_block[i].rport = rport;
4611					set = 1;
4612				}
4613			}
4614		}
4615		if (set) {
4616			/*
4617			 * We only do up to the block where we can place our
4618			 * tag for audits
4619			 */
4620			break;
4621		}
4622	}
4623	/* Need to add a new block to chain */
4624	if (!set) {
4625		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4626		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4627		if (twait_block == NULL) {
4628#ifdef INVARIANTS
4629			panic("Can not alloc tagblock");
4630#endif
4631			return;
4632		}
4633		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4634		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4635		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4636		twait_block->vtag_block[0].v_tag = tag;
4637		twait_block->vtag_block[0].lport = lport;
4638		twait_block->vtag_block[0].rport = rport;
4639	}
4640}
4641
4642void
4643sctp_clean_up_stream(struct sctp_tcb *stcb, struct sctp_readhead *rh)
4644{
4645	struct sctp_tmit_chunk *chk, *nchk;
4646	struct sctp_queued_to_read *ctl, *nctl;
4647
4648	TAILQ_FOREACH_SAFE(ctl, rh, next_instrm, nctl) {
4649		TAILQ_REMOVE(rh, ctl, next_instrm);
4650		ctl->on_strm_q = 0;
4651		if (ctl->on_read_q == 0) {
4652			sctp_free_remote_addr(ctl->whoFrom);
4653			if (ctl->data) {
4654				sctp_m_freem(ctl->data);
4655				ctl->data = NULL;
4656			}
4657		}
4658		/* Reassembly free? */
4659		TAILQ_FOREACH_SAFE(chk, &ctl->reasm, sctp_next, nchk) {
4660			TAILQ_REMOVE(&ctl->reasm, chk, sctp_next);
4661			if (chk->data) {
4662				sctp_m_freem(chk->data);
4663				chk->data = NULL;
4664			}
4665			if (chk->holds_key_ref)
4666				sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4667			sctp_free_remote_addr(chk->whoTo);
4668			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4669			SCTP_DECR_CHK_COUNT();
4670			/* sa_ignore FREED_MEMORY */
4671		}
4672		/*
4673		 * We don't free the address here since all the net's were
4674		 * freed above.
4675		 */
4676		if (ctl->on_read_q == 0) {
4677			sctp_free_a_readq(stcb, ctl);
4678		}
4679	}
4680}
4681
4682
4683/*-
4684 * Free the association after un-hashing the remote port. This
4685 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4686 * expect that the input to this function IS a locked TCB.
4687 * It will return 0, if it did NOT destroy the association (instead
4688 * it unlocks it. It will return NON-zero if it either destroyed the
4689 * association OR the association is already destroyed.
4690 */
4691int
4692sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4693{
4694	int i;
4695	struct sctp_association *asoc;
4696	struct sctp_nets *net, *nnet;
4697	struct sctp_laddr *laddr, *naddr;
4698	struct sctp_tmit_chunk *chk, *nchk;
4699	struct sctp_asconf_addr *aparam, *naparam;
4700	struct sctp_asconf_ack *aack, *naack;
4701	struct sctp_stream_reset_list *strrst, *nstrrst;
4702	struct sctp_queued_to_read *sq, *nsq;
4703	struct sctp_stream_queue_pending *sp, *nsp;
4704	sctp_sharedkey_t *shared_key, *nshared_key;
4705	struct socket *so;
4706
4707	/* first, lets purge the entry from the hash table. */
4708
4709#ifdef SCTP_LOG_CLOSING
4710	sctp_log_closing(inp, stcb, 6);
4711#endif
4712	if (stcb->asoc.state == 0) {
4713#ifdef SCTP_LOG_CLOSING
4714		sctp_log_closing(inp, NULL, 7);
4715#endif
4716		/* there is no asoc, really TSNH :-0 */
4717		return (1);
4718	}
4719	if (stcb->asoc.alternate) {
4720		sctp_free_remote_addr(stcb->asoc.alternate);
4721		stcb->asoc.alternate = NULL;
4722	}
4723	/* TEMP CODE */
4724	if (stcb->freed_from_where == 0) {
4725		/* Only record the first place free happened from */
4726		stcb->freed_from_where = from_location;
4727	}
4728	/* TEMP CODE */
4729
4730	asoc = &stcb->asoc;
4731	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4732	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4733		/* nothing around */
4734		so = NULL;
4735	else
4736		so = inp->sctp_socket;
4737
4738	/*
4739	 * We used timer based freeing if a reader or writer is in the way.
4740	 * So we first check if we are actually being called from a timer,
4741	 * if so we abort early if a reader or writer is still in the way.
4742	 */
4743	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4744	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4745		/*
4746		 * is it the timer driving us? if so are the reader/writers
4747		 * gone?
4748		 */
4749		if (stcb->asoc.refcnt) {
4750			/* nope, reader or writer in the way */
4751			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4752			/* no asoc destroyed */
4753			SCTP_TCB_UNLOCK(stcb);
4754#ifdef SCTP_LOG_CLOSING
4755			sctp_log_closing(inp, stcb, 8);
4756#endif
4757			return (0);
4758		}
4759	}
4760	/* now clean up any other timers */
4761	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4762	asoc->dack_timer.self = NULL;
4763	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4764	/*-
4765	 * For stream reset we don't blast this unless
4766	 * it is a str-reset timer, it might be the
4767	 * free-asoc timer which we DON'T want to
4768	 * disturb.
4769	 */
4770	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4771		asoc->strreset_timer.self = NULL;
4772	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4773	asoc->asconf_timer.self = NULL;
4774	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4775	asoc->autoclose_timer.self = NULL;
4776	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4777	asoc->shut_guard_timer.self = NULL;
4778	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4779	asoc->delayed_event_timer.self = NULL;
4780	/* Mobility adaptation */
4781	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4782	asoc->delete_prim_timer.self = NULL;
4783	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4784		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4785		net->rxt_timer.self = NULL;
4786		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4787		net->pmtu_timer.self = NULL;
4788		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4789		net->hb_timer.self = NULL;
4790	}
4791	/* Now the read queue needs to be cleaned up (only once) */
4792	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4793		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4794		SCTP_INP_READ_LOCK(inp);
4795		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4796			if (sq->stcb == stcb) {
4797				sq->do_not_ref_stcb = 1;
4798				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4799				/*
4800				 * If there is no end, there never will be
4801				 * now.
4802				 */
4803				if (sq->end_added == 0) {
4804					/* Held for PD-API clear that. */
4805					sq->pdapi_aborted = 1;
4806					sq->held_length = 0;
4807					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4808						/*
4809						 * Need to add a PD-API
4810						 * aborted indication.
4811						 * Setting the control_pdapi
4812						 * assures that it will be
4813						 * added right after this
4814						 * msg.
4815						 */
4816						uint32_t strseq;
4817
4818						stcb->asoc.control_pdapi = sq;
4819						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4820						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4821						    stcb,
4822						    SCTP_PARTIAL_DELIVERY_ABORTED,
4823						    (void *)&strseq,
4824						    SCTP_SO_LOCKED);
4825						stcb->asoc.control_pdapi = NULL;
4826					}
4827				}
4828				/* Add an end to wake them */
4829				sq->end_added = 1;
4830			}
4831		}
4832		SCTP_INP_READ_UNLOCK(inp);
4833		if (stcb->block_entry) {
4834			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4835			stcb->block_entry->error = ECONNRESET;
4836			stcb->block_entry = NULL;
4837		}
4838	}
4839	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4840		/*
4841		 * Someone holds a reference OR the socket is unaccepted
4842		 * yet.
4843		 */
4844		if ((stcb->asoc.refcnt) ||
4845		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4846		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4847			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4848			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4849		}
4850		SCTP_TCB_UNLOCK(stcb);
4851		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4852		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4853			/* nothing around */
4854			so = NULL;
4855		if (so) {
4856			/* Wake any reader/writers */
4857			sctp_sorwakeup(inp, so);
4858			sctp_sowwakeup(inp, so);
4859		}
4860#ifdef SCTP_LOG_CLOSING
4861		sctp_log_closing(inp, stcb, 9);
4862#endif
4863		/* no asoc destroyed */
4864		return (0);
4865	}
4866#ifdef SCTP_LOG_CLOSING
4867	sctp_log_closing(inp, stcb, 10);
4868#endif
4869	/*
4870	 * When I reach here, no others want to kill the assoc yet.. and I
4871	 * own the lock. Now its possible an abort comes in when I do the
4872	 * lock exchange below to grab all the locks to do the final take
4873	 * out. to prevent this we increment the count, which will start a
4874	 * timer and blow out above thus assuring us that we hold exclusive
4875	 * killing of the asoc. Note that after getting back the TCB lock we
4876	 * will go ahead and increment the counter back up and stop any
4877	 * timer a passing stranger may have started :-S
4878	 */
4879	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4880		atomic_add_int(&stcb->asoc.refcnt, 1);
4881
4882		SCTP_TCB_UNLOCK(stcb);
4883		SCTP_INP_INFO_WLOCK();
4884		SCTP_INP_WLOCK(inp);
4885		SCTP_TCB_LOCK(stcb);
4886	}
4887	/* Double check the GONE flag */
4888	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4889	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4890		/* nothing around */
4891		so = NULL;
4892
4893	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4894	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4895		/*
4896		 * For TCP type we need special handling when we are
4897		 * connected. We also include the peel'ed off ones to.
4898		 */
4899		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4900			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4901			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4902			if (so) {
4903				SOCK_LOCK(so);
4904				if (so->so_rcv.sb_cc == 0) {
4905					so->so_state &= ~(SS_ISCONNECTING |
4906					    SS_ISDISCONNECTING |
4907					    SS_ISCONFIRMING |
4908					    SS_ISCONNECTED);
4909				}
4910				socantrcvmore_locked(so);
4911				sctp_sowwakeup(inp, so);
4912				sctp_sorwakeup(inp, so);
4913				SCTP_SOWAKEUP(so);
4914			}
4915		}
4916	}
4917	/*
4918	 * Make it invalid too, that way if its about to run it will abort
4919	 * and return.
4920	 */
4921	/* re-increment the lock */
4922	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4923		atomic_add_int(&stcb->asoc.refcnt, -1);
4924	}
4925	if (stcb->asoc.refcnt) {
4926		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4927		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4928		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4929			SCTP_INP_INFO_WUNLOCK();
4930			SCTP_INP_WUNLOCK(inp);
4931		}
4932		SCTP_TCB_UNLOCK(stcb);
4933		return (0);
4934	}
4935	asoc->state = 0;
4936	if (inp->sctp_tcbhash) {
4937		LIST_REMOVE(stcb, sctp_tcbhash);
4938	}
4939	if (stcb->asoc.in_asocid_hash) {
4940		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4941	}
4942	/* Now lets remove it from the list of ALL associations in the EP */
4943	LIST_REMOVE(stcb, sctp_tcblist);
4944	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4945		SCTP_INP_INCR_REF(inp);
4946		SCTP_INP_WUNLOCK(inp);
4947	}
4948	/* pull from vtag hash */
4949	LIST_REMOVE(stcb, sctp_asocs);
4950	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4951	    inp->sctp_lport, stcb->rport);
4952
4953	/*
4954	 * Now restop the timers to be sure this is paranoia at is finest!
4955	 */
4956	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4957	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4958	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4959	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4960	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4961	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4962	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4963	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4964		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4965		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4966		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4967	}
4968
4969	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4970	/*
4971	 * The chunk lists and such SHOULD be empty but we check them just
4972	 * in case.
4973	 */
4974	/* anything on the wheel needs to be removed */
4975	for (i = 0; i < asoc->streamoutcnt; i++) {
4976		struct sctp_stream_out *outs;
4977
4978		outs = &asoc->strmout[i];
4979		/* now clean up any chunks here */
4980		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4981			atomic_subtract_int(&asoc->stream_queue_cnt, 1);
4982			TAILQ_REMOVE(&outs->outqueue, sp, next);
4983			stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, outs, sp, 0);
4984			sctp_free_spbufspace(stcb, asoc, sp);
4985			if (sp->data) {
4986				if (so) {
4987					/* Still an open socket - report */
4988					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4989					    0, (void *)sp, SCTP_SO_LOCKED);
4990				}
4991				if (sp->data) {
4992					sctp_m_freem(sp->data);
4993					sp->data = NULL;
4994					sp->tail_mbuf = NULL;
4995					sp->length = 0;
4996				}
4997			}
4998			if (sp->net) {
4999				sctp_free_remote_addr(sp->net);
5000				sp->net = NULL;
5001			}
5002			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
5003		}
5004	}
5005	/* sa_ignore FREED_MEMORY */
5006	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
5007		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
5008		SCTP_FREE(strrst, SCTP_M_STRESET);
5009	}
5010	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
5011		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
5012		if (sq->data) {
5013			sctp_m_freem(sq->data);
5014			sq->data = NULL;
5015		}
5016		sctp_free_remote_addr(sq->whoFrom);
5017		sq->whoFrom = NULL;
5018		sq->stcb = NULL;
5019		/* Free the ctl entry */
5020		sctp_free_a_readq(stcb, sq);
5021		/* sa_ignore FREED_MEMORY */
5022	}
5023	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
5024		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
5025		if (chk->data) {
5026			sctp_m_freem(chk->data);
5027			chk->data = NULL;
5028		}
5029		if (chk->holds_key_ref)
5030			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5031		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5032		SCTP_DECR_CHK_COUNT();
5033		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
5034		asoc->free_chunk_cnt--;
5035		/* sa_ignore FREED_MEMORY */
5036	}
5037	/* pending send queue SHOULD be empty */
5038	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
5039		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
5040			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
5041#ifdef INVARIANTS
5042		} else {
5043			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
5044#endif
5045		}
5046		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
5047		if (chk->data) {
5048			if (so) {
5049				/* Still a socket? */
5050				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
5051				    0, chk, SCTP_SO_LOCKED);
5052			}
5053			if (chk->data) {
5054				sctp_m_freem(chk->data);
5055				chk->data = NULL;
5056			}
5057		}
5058		if (chk->holds_key_ref)
5059			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5060		if (chk->whoTo) {
5061			sctp_free_remote_addr(chk->whoTo);
5062			chk->whoTo = NULL;
5063		}
5064		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5065		SCTP_DECR_CHK_COUNT();
5066		/* sa_ignore FREED_MEMORY */
5067	}
5068	/* sent queue SHOULD be empty */
5069	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5070		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5071			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
5072				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
5073#ifdef INVARIANTS
5074			} else {
5075				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
5076#endif
5077			}
5078		}
5079		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5080		if (chk->data) {
5081			if (so) {
5082				/* Still a socket? */
5083				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5084				    0, chk, SCTP_SO_LOCKED);
5085			}
5086			if (chk->data) {
5087				sctp_m_freem(chk->data);
5088				chk->data = NULL;
5089			}
5090		}
5091		if (chk->holds_key_ref)
5092			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5093		sctp_free_remote_addr(chk->whoTo);
5094		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5095		SCTP_DECR_CHK_COUNT();
5096		/* sa_ignore FREED_MEMORY */
5097	}
5098#ifdef INVARIANTS
5099	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5100		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5101			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5102		}
5103	}
5104#endif
5105	/* control queue MAY not be empty */
5106	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5107		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5108		if (chk->data) {
5109			sctp_m_freem(chk->data);
5110			chk->data = NULL;
5111		}
5112		if (chk->holds_key_ref)
5113			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5114		sctp_free_remote_addr(chk->whoTo);
5115		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5116		SCTP_DECR_CHK_COUNT();
5117		/* sa_ignore FREED_MEMORY */
5118	}
5119	/* ASCONF queue MAY not be empty */
5120	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5121		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5122		if (chk->data) {
5123			sctp_m_freem(chk->data);
5124			chk->data = NULL;
5125		}
5126		if (chk->holds_key_ref)
5127			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5128		sctp_free_remote_addr(chk->whoTo);
5129		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5130		SCTP_DECR_CHK_COUNT();
5131		/* sa_ignore FREED_MEMORY */
5132	}
5133	if (asoc->mapping_array) {
5134		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5135		asoc->mapping_array = NULL;
5136	}
5137	if (asoc->nr_mapping_array) {
5138		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5139		asoc->nr_mapping_array = NULL;
5140	}
5141	/* the stream outs */
5142	if (asoc->strmout) {
5143		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5144		asoc->strmout = NULL;
5145	}
5146	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5147	if (asoc->strmin) {
5148		for (i = 0; i < asoc->streamincnt; i++) {
5149			sctp_clean_up_stream(stcb, &asoc->strmin[i].inqueue);
5150			sctp_clean_up_stream(stcb, &asoc->strmin[i].uno_inqueue);
5151		}
5152		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5153		asoc->strmin = NULL;
5154	}
5155	asoc->streamincnt = 0;
5156	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5157#ifdef INVARIANTS
5158		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5159			panic("no net's left alloc'ed, or list points to itself");
5160		}
5161#endif
5162		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5163		sctp_free_remote_addr(net);
5164	}
5165	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5166		/* sa_ignore FREED_MEMORY */
5167		sctp_remove_laddr(laddr);
5168	}
5169
5170	/* pending asconf (address) parameters */
5171	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5172		/* sa_ignore FREED_MEMORY */
5173		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5174		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5175	}
5176	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5177		/* sa_ignore FREED_MEMORY */
5178		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5179		if (aack->data != NULL) {
5180			sctp_m_freem(aack->data);
5181		}
5182		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5183	}
5184	/* clean up auth stuff */
5185	if (asoc->local_hmacs)
5186		sctp_free_hmaclist(asoc->local_hmacs);
5187	if (asoc->peer_hmacs)
5188		sctp_free_hmaclist(asoc->peer_hmacs);
5189
5190	if (asoc->local_auth_chunks)
5191		sctp_free_chunklist(asoc->local_auth_chunks);
5192	if (asoc->peer_auth_chunks)
5193		sctp_free_chunklist(asoc->peer_auth_chunks);
5194
5195	sctp_free_authinfo(&asoc->authinfo);
5196
5197	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5198		LIST_REMOVE(shared_key, next);
5199		sctp_free_sharedkey(shared_key);
5200		/* sa_ignore FREED_MEMORY */
5201	}
5202
5203	/* Insert new items here :> */
5204
5205	/* Get rid of LOCK */
5206	SCTP_TCB_UNLOCK(stcb);
5207	SCTP_TCB_LOCK_DESTROY(stcb);
5208	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5209	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5210		SCTP_INP_INFO_WUNLOCK();
5211		SCTP_INP_RLOCK(inp);
5212	}
5213#ifdef SCTP_TRACK_FREED_ASOCS
5214	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5215		/* now clean up the tasoc itself */
5216		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5217		SCTP_DECR_ASOC_COUNT();
5218	} else {
5219		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5220	}
5221#else
5222	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5223	SCTP_DECR_ASOC_COUNT();
5224#endif
5225	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5226		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5227			/*
5228			 * If its NOT the inp_free calling us AND sctp_close
5229			 * as been called, we call back...
5230			 */
5231			SCTP_INP_RUNLOCK(inp);
5232			/*
5233			 * This will start the kill timer (if we are the
5234			 * last one) since we hold an increment yet. But
5235			 * this is the only safe way to do this since
5236			 * otherwise if the socket closes at the same time
5237			 * we are here we might collide in the cleanup.
5238			 */
5239			sctp_inpcb_free(inp,
5240			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5241			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5242			SCTP_INP_DECR_REF(inp);
5243			goto out_of;
5244		} else {
5245			/* The socket is still open. */
5246			SCTP_INP_DECR_REF(inp);
5247		}
5248	}
5249	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5250		SCTP_INP_RUNLOCK(inp);
5251	}
5252out_of:
5253	/* destroyed the asoc */
5254#ifdef SCTP_LOG_CLOSING
5255	sctp_log_closing(inp, NULL, 11);
5256#endif
5257	return (1);
5258}
5259
5260
5261
5262/*
5263 * determine if a destination is "reachable" based upon the addresses bound
5264 * to the current endpoint (e.g. only v4 or v6 currently bound)
5265 */
5266/*
5267 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5268 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5269 * types bound as its endpoint
5270 */
5271int
5272sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5273{
5274	struct sctp_inpcb *inp;
5275	int answer;
5276
5277	/*
5278	 * No locks here, the TCB, in all cases is already locked and an
5279	 * assoc is up. There is either a INP lock by the caller applied (in
5280	 * asconf case when deleting an address) or NOT in the HB case,
5281	 * however if HB then the INP increment is up and the INP will not
5282	 * be removed (on top of the fact that we have a TCB lock). So we
5283	 * only want to read the sctp_flags, which is either bound-all or
5284	 * not.. no protection needed since once an assoc is up you can't be
5285	 * changing your binding.
5286	 */
5287	inp = stcb->sctp_ep;
5288	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5289		/* if bound all, destination is not restricted */
5290		/*
5291		 * RRS: Question during lock work: Is this correct? If you
5292		 * are bound-all you still might need to obey the V4--V6
5293		 * flags??? IMO this bound-all stuff needs to be removed!
5294		 */
5295		return (1);
5296	}
5297	/* NOTE: all "scope" checks are done when local addresses are added */
5298	switch (destaddr->sa_family) {
5299#ifdef INET6
5300	case AF_INET6:
5301		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5302		break;
5303#endif
5304#ifdef INET
5305	case AF_INET:
5306		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5307		break;
5308#endif
5309	default:
5310		/* invalid family, so it's unreachable */
5311		answer = 0;
5312		break;
5313	}
5314	return (answer);
5315}
5316
5317/*
5318 * update the inp_vflags on an endpoint
5319 */
5320static void
5321sctp_update_ep_vflag(struct sctp_inpcb *inp)
5322{
5323	struct sctp_laddr *laddr;
5324
5325	/* first clear the flag */
5326	inp->ip_inp.inp.inp_vflag = 0;
5327	/* set the flag based on addresses on the ep list */
5328	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5329		if (laddr->ifa == NULL) {
5330			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5331			    __func__);
5332			continue;
5333		}
5334		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5335			continue;
5336		}
5337		switch (laddr->ifa->address.sa.sa_family) {
5338#ifdef INET6
5339		case AF_INET6:
5340			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5341			break;
5342#endif
5343#ifdef INET
5344		case AF_INET:
5345			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5346			break;
5347#endif
5348		default:
5349			break;
5350		}
5351	}
5352}
5353
5354/*
5355 * Add the address to the endpoint local address list There is nothing to be
5356 * done if we are bound to all addresses
5357 */
5358void
5359sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5360{
5361	struct sctp_laddr *laddr;
5362	struct sctp_tcb *stcb;
5363	int fnd, error = 0;
5364
5365	fnd = 0;
5366
5367	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5368		/* You are already bound to all. You have it already */
5369		return;
5370	}
5371#ifdef INET6
5372	if (ifa->address.sa.sa_family == AF_INET6) {
5373		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5374			/* Can't bind a non-useable addr. */
5375			return;
5376		}
5377	}
5378#endif
5379	/* first, is it already present? */
5380	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5381		if (laddr->ifa == ifa) {
5382			fnd = 1;
5383			break;
5384		}
5385	}
5386
5387	if (fnd == 0) {
5388		/* Not in the ep list */
5389		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5390		if (error != 0)
5391			return;
5392		inp->laddr_count++;
5393		/* update inp_vflag flags */
5394		switch (ifa->address.sa.sa_family) {
5395#ifdef INET6
5396		case AF_INET6:
5397			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5398			break;
5399#endif
5400#ifdef INET
5401		case AF_INET:
5402			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5403			break;
5404#endif
5405		default:
5406			break;
5407		}
5408		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5409			sctp_add_local_addr_restricted(stcb, ifa);
5410		}
5411	}
5412	return;
5413}
5414
5415
5416/*
5417 * select a new (hopefully reachable) destination net (should only be used
5418 * when we deleted an ep addr that is the only usable source address to reach
5419 * the destination net)
5420 */
5421static void
5422sctp_select_primary_destination(struct sctp_tcb *stcb)
5423{
5424	struct sctp_nets *net;
5425
5426	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5427		/* for now, we'll just pick the first reachable one we find */
5428		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5429			continue;
5430		if (sctp_destination_is_reachable(stcb,
5431		    (struct sockaddr *)&net->ro._l_addr)) {
5432			/* found a reachable destination */
5433			stcb->asoc.primary_destination = net;
5434		}
5435	}
5436	/* I can't there from here! ...we're gonna die shortly... */
5437}
5438
5439
5440/*
5441 * Delete the address from the endpoint local address list. There is nothing
5442 * to be done if we are bound to all addresses
5443 */
5444void
5445sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5446{
5447	struct sctp_laddr *laddr;
5448	int fnd;
5449
5450	fnd = 0;
5451	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5452		/* You are already bound to all. You have it already */
5453		return;
5454	}
5455	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5456		if (laddr->ifa == ifa) {
5457			fnd = 1;
5458			break;
5459		}
5460	}
5461	if (fnd && (inp->laddr_count < 2)) {
5462		/* can't delete unless there are at LEAST 2 addresses */
5463		return;
5464	}
5465	if (fnd) {
5466		/*
5467		 * clean up any use of this address go through our
5468		 * associations and clear any last_used_address that match
5469		 * this one for each assoc, see if a new primary_destination
5470		 * is needed
5471		 */
5472		struct sctp_tcb *stcb;
5473
5474		/* clean up "next_addr_touse" */
5475		if (inp->next_addr_touse == laddr)
5476			/* delete this address */
5477			inp->next_addr_touse = NULL;
5478
5479		/* clean up "last_used_address" */
5480		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5481			struct sctp_nets *net;
5482
5483			SCTP_TCB_LOCK(stcb);
5484			if (stcb->asoc.last_used_address == laddr)
5485				/* delete this address */
5486				stcb->asoc.last_used_address = NULL;
5487			/*
5488			 * Now spin through all the nets and purge any ref
5489			 * to laddr
5490			 */
5491			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5492				if (net->ro._s_addr == laddr->ifa) {
5493					/* Yep, purge src address selected */
5494					sctp_rtentry_t *rt;
5495
5496					/* delete this address if cached */
5497					rt = net->ro.ro_rt;
5498					if (rt != NULL) {
5499						RTFREE(rt);
5500						net->ro.ro_rt = NULL;
5501					}
5502					sctp_free_ifa(net->ro._s_addr);
5503					net->ro._s_addr = NULL;
5504					net->src_addr_selected = 0;
5505				}
5506			}
5507			SCTP_TCB_UNLOCK(stcb);
5508		}		/* for each tcb */
5509		/* remove it from the ep list */
5510		sctp_remove_laddr(laddr);
5511		inp->laddr_count--;
5512		/* update inp_vflag flags */
5513		sctp_update_ep_vflag(inp);
5514	}
5515	return;
5516}
5517
5518/*
5519 * Add the address to the TCB local address restricted list.
5520 * This is a "pending" address list (eg. addresses waiting for an
5521 * ASCONF-ACK response) and cannot be used as a valid source address.
5522 */
5523void
5524sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5525{
5526	struct sctp_laddr *laddr;
5527	struct sctpladdr *list;
5528
5529	/*
5530	 * Assumes TCB is locked.. and possibly the INP. May need to
5531	 * confirm/fix that if we need it and is not the case.
5532	 */
5533	list = &stcb->asoc.sctp_restricted_addrs;
5534
5535#ifdef INET6
5536	if (ifa->address.sa.sa_family == AF_INET6) {
5537		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5538			/* Can't bind a non-existent addr. */
5539			return;
5540		}
5541	}
5542#endif
5543	/* does the address already exist? */
5544	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5545		if (laddr->ifa == ifa) {
5546			return;
5547		}
5548	}
5549
5550	/* add to the list */
5551	(void)sctp_insert_laddr(list, ifa, 0);
5552	return;
5553}
5554
5555/*
5556 * Remove a local address from the TCB local address restricted list
5557 */
5558void
5559sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5560{
5561	struct sctp_inpcb *inp;
5562	struct sctp_laddr *laddr;
5563
5564	/*
5565	 * This is called by asconf work. It is assumed that a) The TCB is
5566	 * locked and b) The INP is locked. This is true in as much as I can
5567	 * trace through the entry asconf code where I did these locks.
5568	 * Again, the ASCONF code is a bit different in that it does lock
5569	 * the INP during its work often times. This must be since we don't
5570	 * want other proc's looking up things while what they are looking
5571	 * up is changing :-D
5572	 */
5573
5574	inp = stcb->sctp_ep;
5575	/* if subset bound and don't allow ASCONF's, can't delete last */
5576	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5577	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5578		if (stcb->sctp_ep->laddr_count < 2) {
5579			/* can't delete last address */
5580			return;
5581		}
5582	}
5583	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5584		/* remove the address if it exists */
5585		if (laddr->ifa == NULL)
5586			continue;
5587		if (laddr->ifa == ifa) {
5588			sctp_remove_laddr(laddr);
5589			return;
5590		}
5591	}
5592
5593	/* address not found! */
5594	return;
5595}
5596
5597/*
5598 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5599 */
5600/* sysctl */
5601static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5602static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5603
5604
5605
5606#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5607struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5608int *sctp_cpuarry = NULL;
5609void
5610sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5611{
5612	/* Queue a packet to a processor for the specified core */
5613	struct sctp_mcore_queue *qent;
5614	struct sctp_mcore_ctrl *wkq;
5615	int need_wake = 0;
5616
5617	if (sctp_mcore_workers == NULL) {
5618		/* Something went way bad during setup */
5619		sctp_input_with_port(m, off, 0);
5620		return;
5621	}
5622	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5623	    (sizeof(struct sctp_mcore_queue)),
5624	    SCTP_M_MCORE);
5625	if (qent == NULL) {
5626		/* This is trouble  */
5627		sctp_input_with_port(m, off, 0);
5628		return;
5629	}
5630	qent->vn = curvnet;
5631	qent->m = m;
5632	qent->off = off;
5633	qent->v6 = 0;
5634	wkq = &sctp_mcore_workers[cpu_to_use];
5635	SCTP_MCORE_QLOCK(wkq);
5636
5637	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5638	if (wkq->running == 0) {
5639		need_wake = 1;
5640	}
5641	SCTP_MCORE_QUNLOCK(wkq);
5642	if (need_wake) {
5643		wakeup(&wkq->running);
5644	}
5645}
5646
5647static void
5648sctp_mcore_thread(void *arg)
5649{
5650
5651	struct sctp_mcore_ctrl *wkq;
5652	struct sctp_mcore_queue *qent;
5653
5654	wkq = (struct sctp_mcore_ctrl *)arg;
5655	struct mbuf *m;
5656	int off, v6;
5657
5658	/* Wait for first tickle */
5659	SCTP_MCORE_LOCK(wkq);
5660	wkq->running = 0;
5661	msleep(&wkq->running,
5662	    &wkq->core_mtx,
5663	    0, "wait for pkt", 0);
5664	SCTP_MCORE_UNLOCK(wkq);
5665
5666	/* Bind to our cpu */
5667	thread_lock(curthread);
5668	sched_bind(curthread, wkq->cpuid);
5669	thread_unlock(curthread);
5670
5671	/* Now lets start working */
5672	SCTP_MCORE_LOCK(wkq);
5673	/* Now grab lock and go */
5674	for (;;) {
5675		SCTP_MCORE_QLOCK(wkq);
5676skip_sleep:
5677		wkq->running = 1;
5678		qent = TAILQ_FIRST(&wkq->que);
5679		if (qent) {
5680			TAILQ_REMOVE(&wkq->que, qent, next);
5681			SCTP_MCORE_QUNLOCK(wkq);
5682			CURVNET_SET(qent->vn);
5683			m = qent->m;
5684			off = qent->off;
5685			v6 = qent->v6;
5686			SCTP_FREE(qent, SCTP_M_MCORE);
5687			if (v6 == 0) {
5688				sctp_input_with_port(m, off, 0);
5689			} else {
5690				SCTP_PRINTF("V6 not yet supported\n");
5691				sctp_m_freem(m);
5692			}
5693			CURVNET_RESTORE();
5694			SCTP_MCORE_QLOCK(wkq);
5695		}
5696		wkq->running = 0;
5697		if (!TAILQ_EMPTY(&wkq->que)) {
5698			goto skip_sleep;
5699		}
5700		SCTP_MCORE_QUNLOCK(wkq);
5701		msleep(&wkq->running,
5702		    &wkq->core_mtx,
5703		    0, "wait for pkt", 0);
5704	}
5705}
5706
5707static void
5708sctp_startup_mcore_threads(void)
5709{
5710	int i, cpu;
5711
5712	if (mp_ncpus == 1)
5713		return;
5714
5715	if (sctp_mcore_workers != NULL) {
5716		/*
5717		 * Already been here in some previous vnet?
5718		 */
5719		return;
5720	}
5721	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5722	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5723	    SCTP_M_MCORE);
5724	if (sctp_mcore_workers == NULL) {
5725		/* TSNH I hope */
5726		return;
5727	}
5728	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5729	    sizeof(struct sctp_mcore_ctrl)));
5730	/* Init the structures */
5731	for (i = 0; i <= mp_maxid; i++) {
5732		TAILQ_INIT(&sctp_mcore_workers[i].que);
5733		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5734		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5735		sctp_mcore_workers[i].cpuid = i;
5736	}
5737	if (sctp_cpuarry == NULL) {
5738		SCTP_MALLOC(sctp_cpuarry, int *,
5739		    (mp_ncpus * sizeof(int)),
5740		    SCTP_M_MCORE);
5741		i = 0;
5742		CPU_FOREACH(cpu) {
5743			sctp_cpuarry[i] = cpu;
5744			i++;
5745		}
5746	}
5747	/* Now start them all */
5748	CPU_FOREACH(cpu) {
5749		(void)kproc_create(sctp_mcore_thread,
5750		    (void *)&sctp_mcore_workers[cpu],
5751		    &sctp_mcore_workers[cpu].thread_proc,
5752		    RFPROC,
5753		    SCTP_KTHREAD_PAGES,
5754		    SCTP_MCORE_NAME);
5755
5756	}
5757}
5758
5759#endif
5760
5761void
5762sctp_pcb_init()
5763{
5764	/*
5765	 * SCTP initialization for the PCB structures should be called by
5766	 * the sctp_init() function.
5767	 */
5768	int i;
5769	struct timeval tv;
5770
5771	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5772		/* error I was called twice */
5773		return;
5774	}
5775	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5776
5777#if defined(SCTP_LOCAL_TRACE_BUF)
5778	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5779#endif
5780#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5781	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5782	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5783	    SCTP_M_MCORE);
5784#endif
5785	(void)SCTP_GETTIME_TIMEVAL(&tv);
5786#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5787	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5788	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5789	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5790#else
5791	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5792	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5793	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5794#endif
5795	/* init the empty list of (All) Endpoints */
5796	LIST_INIT(&SCTP_BASE_INFO(listhead));
5797
5798
5799	/* init the hash table of endpoints */
5800	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5801	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5802	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5803	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5804	    &SCTP_BASE_INFO(hashasocmark));
5805	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5806	    &SCTP_BASE_INFO(hashmark));
5807	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5808	    &SCTP_BASE_INFO(hashtcpmark));
5809	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5810
5811
5812	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5813	    &SCTP_BASE_INFO(hashvrfmark));
5814
5815	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5816	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5817	/* init the zones */
5818	/*
5819	 * FIX ME: Should check for NULL returns, but if it does fail we are
5820	 * doomed to panic anyways... add later maybe.
5821	 */
5822	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5823	    sizeof(struct sctp_inpcb), maxsockets);
5824
5825	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5826	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5827
5828	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5829	    sizeof(struct sctp_laddr),
5830	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5831
5832	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5833	    sizeof(struct sctp_nets),
5834	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5835
5836	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5837	    sizeof(struct sctp_tmit_chunk),
5838	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5839
5840	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5841	    sizeof(struct sctp_queued_to_read),
5842	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5843
5844	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5845	    sizeof(struct sctp_stream_queue_pending),
5846	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5847
5848	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5849	    sizeof(struct sctp_asconf),
5850	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5851
5852	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5853	    sizeof(struct sctp_asconf_ack),
5854	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5855
5856
5857	/* Master Lock INIT for info structure */
5858	SCTP_INP_INFO_LOCK_INIT();
5859	SCTP_STATLOG_INIT_LOCK();
5860
5861	SCTP_IPI_COUNT_INIT();
5862	SCTP_IPI_ADDR_INIT();
5863#ifdef SCTP_PACKET_LOGGING
5864	SCTP_IP_PKTLOG_INIT();
5865#endif
5866	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5867
5868	SCTP_WQ_ADDR_INIT();
5869	/* not sure if we need all the counts */
5870	SCTP_BASE_INFO(ipi_count_ep) = 0;
5871	/* assoc/tcb zone info */
5872	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5873	/* local addrlist zone info */
5874	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5875	/* remote addrlist zone info */
5876	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5877	/* chunk info */
5878	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5879
5880	/* socket queue zone info */
5881	SCTP_BASE_INFO(ipi_count_readq) = 0;
5882
5883	/* stream out queue cont */
5884	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5885
5886	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5887	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5888
5889	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5890
5891	/* Init the TIMEWAIT list */
5892	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5893		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5894	}
5895	sctp_startup_iterator();
5896
5897#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5898	sctp_startup_mcore_threads();
5899#endif
5900
5901	/*
5902	 * INIT the default VRF which for BSD is the only one, other O/S's
5903	 * may have more. But initially they must start with one and then
5904	 * add the VRF's as addresses are added.
5905	 */
5906	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5907}
5908
5909/*
5910 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5911 */
5912void
5913sctp_pcb_finish(void)
5914{
5915	struct sctp_vrflist *vrf_bucket;
5916	struct sctp_vrf *vrf, *nvrf;
5917	struct sctp_ifn *ifn, *nifn;
5918	struct sctp_ifa *ifa, *nifa;
5919	struct sctpvtaghead *chain;
5920	struct sctp_tagblock *twait_block, *prev_twait_block;
5921	struct sctp_laddr *wi, *nwi;
5922	int i;
5923	struct sctp_iterator *it, *nit;
5924
5925	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
5926		SCTP_PRINTF("%s: race condition on teardown.\n", __func__);
5927		return;
5928	}
5929	SCTP_BASE_VAR(sctp_pcb_initialized) = 0;
5930	/*
5931	 * In FreeBSD the iterator thread never exits but we do clean up.
5932	 * The only way FreeBSD reaches here is if we have VRF's but we
5933	 * still add the ifdef to make it compile on old versions.
5934	 */
5935retry:
5936	SCTP_IPI_ITERATOR_WQ_LOCK();
5937	/*
5938	 * sctp_iterator_worker() might be working on an it entry without
5939	 * holding the lock.  We won't find it on the list either and
5940	 * continue and free/destroy it.  While holding the lock, spin, to
5941	 * avoid the race condition as sctp_iterator_worker() will have to
5942	 * wait to re-aquire the lock.
5943	 */
5944	if (sctp_it_ctl.iterator_running != 0 || sctp_it_ctl.cur_it != NULL) {
5945		SCTP_IPI_ITERATOR_WQ_UNLOCK();
5946		SCTP_PRINTF("%s: Iterator running while we held the lock. Retry. "
5947		    "cur_it=%p\n", __func__, sctp_it_ctl.cur_it);
5948		DELAY(10);
5949		goto retry;
5950	}
5951	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5952		if (it->vn != curvnet) {
5953			continue;
5954		}
5955		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5956		if (it->function_atend != NULL) {
5957			(*it->function_atend) (it->pointer, it->val);
5958		}
5959		SCTP_FREE(it, SCTP_M_ITER);
5960	}
5961	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5962	SCTP_ITERATOR_LOCK();
5963	if ((sctp_it_ctl.cur_it) &&
5964	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5965		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5966	}
5967	SCTP_ITERATOR_UNLOCK();
5968	SCTP_OS_TIMER_STOP_DRAIN(&SCTP_BASE_INFO(addr_wq_timer.timer));
5969	SCTP_WQ_ADDR_LOCK();
5970	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5971		LIST_REMOVE(wi, sctp_nxt_addr);
5972		SCTP_DECR_LADDR_COUNT();
5973		if (wi->action == SCTP_DEL_IP_ADDRESS) {
5974			SCTP_FREE(wi->ifa, SCTP_M_IFA);
5975		}
5976		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5977	}
5978	SCTP_WQ_ADDR_UNLOCK();
5979
5980	/*
5981	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5982	 * destroyed first).
5983	 */
5984	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5985	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5986		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5987			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5988				/* free the ifa */
5989				LIST_REMOVE(ifa, next_bucket);
5990				LIST_REMOVE(ifa, next_ifa);
5991				SCTP_FREE(ifa, SCTP_M_IFA);
5992			}
5993			/* free the ifn */
5994			LIST_REMOVE(ifn, next_bucket);
5995			LIST_REMOVE(ifn, next_ifn);
5996			SCTP_FREE(ifn, SCTP_M_IFN);
5997		}
5998		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5999		/* free the vrf */
6000		LIST_REMOVE(vrf, next_vrf);
6001		SCTP_FREE(vrf, SCTP_M_VRF);
6002	}
6003	/* free the vrf hashes */
6004	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
6005	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
6006
6007	/*
6008	 * free the TIMEWAIT list elements malloc'd in the function
6009	 * sctp_add_vtag_to_timewait()...
6010	 */
6011	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
6012		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
6013		if (!LIST_EMPTY(chain)) {
6014			prev_twait_block = NULL;
6015			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6016				if (prev_twait_block) {
6017					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6018				}
6019				prev_twait_block = twait_block;
6020			}
6021			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6022		}
6023	}
6024
6025	/* free the locks and mutexes */
6026#ifdef SCTP_PACKET_LOGGING
6027	SCTP_IP_PKTLOG_DESTROY();
6028#endif
6029	SCTP_IPI_ADDR_DESTROY();
6030	SCTP_STATLOG_DESTROY();
6031	SCTP_INP_INFO_LOCK_DESTROY();
6032
6033	SCTP_WQ_ADDR_DESTROY();
6034
6035	/* Get rid of other stuff too. */
6036	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
6037		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
6038	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
6039		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6040	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6041		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6042
6043	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6044	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6045	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6046	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6047	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6048	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6049	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6050	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6051	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6052#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6053	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6054#endif
6055}
6056
6057
6058int
6059sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6060    int offset, int limit,
6061    struct sockaddr *src, struct sockaddr *dst,
6062    struct sockaddr *altsa, uint16_t port)
6063{
6064	/*
6065	 * grub through the INIT pulling addresses and loading them to the
6066	 * nets structure in the asoc. The from address in the mbuf should
6067	 * also be loaded (if it is not already). This routine can be called
6068	 * with either INIT or INIT-ACK's as long as the m points to the IP
6069	 * packet and the offset points to the beginning of the parameters.
6070	 */
6071	struct sctp_inpcb *inp;
6072	struct sctp_nets *net, *nnet, *net_tmp;
6073	struct sctp_paramhdr *phdr, parm_buf;
6074	struct sctp_tcb *stcb_tmp;
6075	uint16_t ptype, plen;
6076	struct sockaddr *sa;
6077	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6078	struct sctp_auth_random *p_random = NULL;
6079	uint16_t random_len = 0;
6080	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6081	struct sctp_auth_hmac_algo *hmacs = NULL;
6082	uint16_t hmacs_len = 0;
6083	uint8_t saw_asconf = 0;
6084	uint8_t saw_asconf_ack = 0;
6085	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6086	struct sctp_auth_chunk_list *chunks = NULL;
6087	uint16_t num_chunks = 0;
6088	sctp_key_t *new_key;
6089	uint32_t keylen;
6090	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6091	uint8_t peer_supports_ecn;
6092	uint8_t peer_supports_prsctp;
6093	uint8_t peer_supports_auth;
6094	uint8_t peer_supports_asconf;
6095	uint8_t peer_supports_asconf_ack;
6096	uint8_t peer_supports_reconfig;
6097	uint8_t peer_supports_nrsack;
6098	uint8_t peer_supports_pktdrop;
6099	uint8_t peer_supports_idata;
6100
6101#ifdef INET
6102	struct sockaddr_in sin;
6103
6104#endif
6105#ifdef INET6
6106	struct sockaddr_in6 sin6;
6107
6108#endif
6109
6110	/* First get the destination address setup too. */
6111#ifdef INET
6112	memset(&sin, 0, sizeof(sin));
6113	sin.sin_family = AF_INET;
6114	sin.sin_len = sizeof(sin);
6115	sin.sin_port = stcb->rport;
6116#endif
6117#ifdef INET6
6118	memset(&sin6, 0, sizeof(sin6));
6119	sin6.sin6_family = AF_INET6;
6120	sin6.sin6_len = sizeof(struct sockaddr_in6);
6121	sin6.sin6_port = stcb->rport;
6122#endif
6123	if (altsa) {
6124		sa = altsa;
6125	} else {
6126		sa = src;
6127	}
6128	peer_supports_idata = 0;
6129	peer_supports_ecn = 0;
6130	peer_supports_prsctp = 0;
6131	peer_supports_auth = 0;
6132	peer_supports_asconf = 0;
6133	peer_supports_reconfig = 0;
6134	peer_supports_nrsack = 0;
6135	peer_supports_pktdrop = 0;
6136	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6137		/* mark all addresses that we have currently on the list */
6138		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6139	}
6140	/* does the source address already exist? if so skip it */
6141	inp = stcb->sctp_ep;
6142	atomic_add_int(&stcb->asoc.refcnt, 1);
6143	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6144	atomic_add_int(&stcb->asoc.refcnt, -1);
6145
6146	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6147		/* we must add the source address */
6148		/* no scope set here since we have a tcb already. */
6149		switch (sa->sa_family) {
6150#ifdef INET
6151		case AF_INET:
6152			if (stcb->asoc.scope.ipv4_addr_legal) {
6153				if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6154					return (-1);
6155				}
6156			}
6157			break;
6158#endif
6159#ifdef INET6
6160		case AF_INET6:
6161			if (stcb->asoc.scope.ipv6_addr_legal) {
6162				if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6163					return (-2);
6164				}
6165			}
6166			break;
6167#endif
6168		default:
6169			break;
6170		}
6171	} else {
6172		if (net_tmp != NULL && stcb_tmp == stcb) {
6173			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6174		} else if (stcb_tmp != stcb) {
6175			/* It belongs to another association? */
6176			if (stcb_tmp)
6177				SCTP_TCB_UNLOCK(stcb_tmp);
6178			return (-3);
6179		}
6180	}
6181	if (stcb->asoc.state == 0) {
6182		/* the assoc was freed? */
6183		return (-4);
6184	}
6185	/* now we must go through each of the params. */
6186	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6187	while (phdr) {
6188		ptype = ntohs(phdr->param_type);
6189		plen = ntohs(phdr->param_length);
6190		/*
6191		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6192		 * (uint32_t)ptype, (int)plen);
6193		 */
6194		if (offset + plen > limit) {
6195			break;
6196		}
6197		if (plen == 0) {
6198			break;
6199		}
6200#ifdef INET
6201		if (ptype == SCTP_IPV4_ADDRESS) {
6202			if (stcb->asoc.scope.ipv4_addr_legal) {
6203				struct sctp_ipv4addr_param *p4, p4_buf;
6204
6205				/* ok get the v4 address and check/add */
6206				phdr = sctp_get_next_param(m, offset,
6207				    (struct sctp_paramhdr *)&p4_buf,
6208				    sizeof(p4_buf));
6209				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6210				    phdr == NULL) {
6211					return (-5);
6212				}
6213				p4 = (struct sctp_ipv4addr_param *)phdr;
6214				sin.sin_addr.s_addr = p4->addr;
6215				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6216					/* Skip multi-cast addresses */
6217					goto next_param;
6218				}
6219				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6220				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6221					goto next_param;
6222				}
6223				sa = (struct sockaddr *)&sin;
6224				inp = stcb->sctp_ep;
6225				atomic_add_int(&stcb->asoc.refcnt, 1);
6226				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6227				    dst, stcb);
6228				atomic_add_int(&stcb->asoc.refcnt, -1);
6229
6230				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6231				    inp == NULL) {
6232					/* we must add the source address */
6233					/*
6234					 * no scope set since we have a tcb
6235					 * already
6236					 */
6237
6238					/*
6239					 * we must validate the state again
6240					 * here
6241					 */
6242			add_it_now:
6243					if (stcb->asoc.state == 0) {
6244						/* the assoc was freed? */
6245						return (-7);
6246					}
6247					if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6248						return (-8);
6249					}
6250				} else if (stcb_tmp == stcb) {
6251					if (stcb->asoc.state == 0) {
6252						/* the assoc was freed? */
6253						return (-10);
6254					}
6255					if (net != NULL) {
6256						/* clear flag */
6257						net->dest_state &=
6258						    ~SCTP_ADDR_NOT_IN_ASSOC;
6259					}
6260				} else {
6261					/*
6262					 * strange, address is in another
6263					 * assoc? straighten out locks.
6264					 */
6265					if (stcb_tmp) {
6266						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6267							struct mbuf *op_err;
6268							char msg[SCTP_DIAG_INFO_LEN];
6269
6270							/*
6271							 * in setup state we
6272							 * abort this guy
6273							 */
6274							snprintf(msg, sizeof(msg),
6275							    "%s:%d at %s", __FILE__, __LINE__, __func__);
6276							op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
6277							    msg);
6278							sctp_abort_an_association(stcb_tmp->sctp_ep,
6279							    stcb_tmp, op_err,
6280							    SCTP_SO_NOT_LOCKED);
6281							goto add_it_now;
6282						}
6283						SCTP_TCB_UNLOCK(stcb_tmp);
6284					}
6285					if (stcb->asoc.state == 0) {
6286						/* the assoc was freed? */
6287						return (-12);
6288					}
6289					return (-13);
6290				}
6291			}
6292		} else
6293#endif
6294#ifdef INET6
6295		if (ptype == SCTP_IPV6_ADDRESS) {
6296			if (stcb->asoc.scope.ipv6_addr_legal) {
6297				/* ok get the v6 address and check/add */
6298				struct sctp_ipv6addr_param *p6, p6_buf;
6299
6300				phdr = sctp_get_next_param(m, offset,
6301				    (struct sctp_paramhdr *)&p6_buf,
6302				    sizeof(p6_buf));
6303				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6304				    phdr == NULL) {
6305					return (-14);
6306				}
6307				p6 = (struct sctp_ipv6addr_param *)phdr;
6308				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6309				    sizeof(p6->addr));
6310				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6311					/* Skip multi-cast addresses */
6312					goto next_param;
6313				}
6314				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6315					/*
6316					 * Link local make no sense without
6317					 * scope
6318					 */
6319					goto next_param;
6320				}
6321				sa = (struct sockaddr *)&sin6;
6322				inp = stcb->sctp_ep;
6323				atomic_add_int(&stcb->asoc.refcnt, 1);
6324				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6325				    dst, stcb);
6326				atomic_add_int(&stcb->asoc.refcnt, -1);
6327				if (stcb_tmp == NULL &&
6328				    (inp == stcb->sctp_ep || inp == NULL)) {
6329					/*
6330					 * we must validate the state again
6331					 * here
6332					 */
6333			add_it_now6:
6334					if (stcb->asoc.state == 0) {
6335						/* the assoc was freed? */
6336						return (-16);
6337					}
6338					/*
6339					 * we must add the address, no scope
6340					 * set
6341					 */
6342					if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6343						return (-17);
6344					}
6345				} else if (stcb_tmp == stcb) {
6346					/*
6347					 * we must validate the state again
6348					 * here
6349					 */
6350					if (stcb->asoc.state == 0) {
6351						/* the assoc was freed? */
6352						return (-19);
6353					}
6354					if (net != NULL) {
6355						/* clear flag */
6356						net->dest_state &=
6357						    ~SCTP_ADDR_NOT_IN_ASSOC;
6358					}
6359				} else {
6360					/*
6361					 * strange, address is in another
6362					 * assoc? straighten out locks.
6363					 */
6364					if (stcb_tmp) {
6365						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6366							struct mbuf *op_err;
6367							char msg[SCTP_DIAG_INFO_LEN];
6368
6369							/*
6370							 * in setup state we
6371							 * abort this guy
6372							 */
6373							snprintf(msg, sizeof(msg),
6374							    "%s:%d at %s", __FILE__, __LINE__, __func__);
6375							op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
6376							    msg);
6377							sctp_abort_an_association(stcb_tmp->sctp_ep,
6378							    stcb_tmp, op_err,
6379							    SCTP_SO_NOT_LOCKED);
6380							goto add_it_now6;
6381						}
6382						SCTP_TCB_UNLOCK(stcb_tmp);
6383					}
6384					if (stcb->asoc.state == 0) {
6385						/* the assoc was freed? */
6386						return (-21);
6387					}
6388					return (-22);
6389				}
6390			}
6391		} else
6392#endif
6393		if (ptype == SCTP_ECN_CAPABLE) {
6394			peer_supports_ecn = 1;
6395		} else if (ptype == SCTP_ULP_ADAPTATION) {
6396			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6397				struct sctp_adaptation_layer_indication ai,
6398				                                *aip;
6399
6400				phdr = sctp_get_next_param(m, offset,
6401				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6402				aip = (struct sctp_adaptation_layer_indication *)phdr;
6403				if (aip) {
6404					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6405					stcb->asoc.adaptation_needed = 1;
6406				}
6407			}
6408		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6409			struct sctp_asconf_addr_param lstore, *fee;
6410			int lptype;
6411			struct sockaddr *lsa = NULL;
6412
6413#ifdef INET
6414			struct sctp_asconf_addrv4_param *fii;
6415
6416#endif
6417
6418			if (stcb->asoc.asconf_supported == 0) {
6419				return (-100);
6420			}
6421			if (plen > sizeof(lstore)) {
6422				return (-23);
6423			}
6424			phdr = sctp_get_next_param(m, offset,
6425			    (struct sctp_paramhdr *)&lstore,
6426			    min(plen, sizeof(lstore)));
6427			if (phdr == NULL) {
6428				return (-24);
6429			}
6430			fee = (struct sctp_asconf_addr_param *)phdr;
6431			lptype = ntohs(fee->addrp.ph.param_type);
6432			switch (lptype) {
6433#ifdef INET
6434			case SCTP_IPV4_ADDRESS:
6435				if (plen !=
6436				    sizeof(struct sctp_asconf_addrv4_param)) {
6437					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6438					    (int)sizeof(struct sctp_asconf_addrv4_param),
6439					    plen);
6440				} else {
6441					fii = (struct sctp_asconf_addrv4_param *)fee;
6442					sin.sin_addr.s_addr = fii->addrp.addr;
6443					lsa = (struct sockaddr *)&sin;
6444				}
6445				break;
6446#endif
6447#ifdef INET6
6448			case SCTP_IPV6_ADDRESS:
6449				if (plen !=
6450				    sizeof(struct sctp_asconf_addr_param)) {
6451					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6452					    (int)sizeof(struct sctp_asconf_addr_param),
6453					    plen);
6454				} else {
6455					memcpy(sin6.sin6_addr.s6_addr,
6456					    fee->addrp.addr,
6457					    sizeof(fee->addrp.addr));
6458					lsa = (struct sockaddr *)&sin6;
6459				}
6460				break;
6461#endif
6462			default:
6463				break;
6464			}
6465			if (lsa) {
6466				(void)sctp_set_primary_addr(stcb, sa, NULL);
6467			}
6468		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6469			stcb->asoc.peer_supports_nat = 1;
6470		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6471			/* Peer supports pr-sctp */
6472			peer_supports_prsctp = 1;
6473		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6474			/* A supported extension chunk */
6475			struct sctp_supported_chunk_types_param *pr_supported;
6476			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6477			int num_ent, i;
6478
6479			phdr = sctp_get_next_param(m, offset,
6480			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6481			if (phdr == NULL) {
6482				return (-25);
6483			}
6484			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6485			num_ent = plen - sizeof(struct sctp_paramhdr);
6486			for (i = 0; i < num_ent; i++) {
6487				switch (pr_supported->chunk_types[i]) {
6488				case SCTP_ASCONF:
6489					peer_supports_asconf = 1;
6490					break;
6491				case SCTP_ASCONF_ACK:
6492					peer_supports_asconf_ack = 1;
6493					break;
6494				case SCTP_FORWARD_CUM_TSN:
6495					peer_supports_prsctp = 1;
6496					break;
6497				case SCTP_PACKET_DROPPED:
6498					peer_supports_pktdrop = 1;
6499					break;
6500				case SCTP_NR_SELECTIVE_ACK:
6501					peer_supports_nrsack = 1;
6502					break;
6503				case SCTP_STREAM_RESET:
6504					peer_supports_reconfig = 1;
6505					break;
6506				case SCTP_AUTHENTICATION:
6507					peer_supports_auth = 1;
6508					break;
6509				case SCTP_IDATA:
6510					peer_supports_idata = 1;
6511					break;
6512				default:
6513					/* one I have not learned yet */
6514					break;
6515
6516				}
6517			}
6518		} else if (ptype == SCTP_RANDOM) {
6519			if (plen > sizeof(random_store))
6520				break;
6521			if (got_random) {
6522				/* already processed a RANDOM */
6523				goto next_param;
6524			}
6525			phdr = sctp_get_next_param(m, offset,
6526			    (struct sctp_paramhdr *)random_store,
6527			    min(sizeof(random_store), plen));
6528			if (phdr == NULL)
6529				return (-26);
6530			p_random = (struct sctp_auth_random *)phdr;
6531			random_len = plen - sizeof(*p_random);
6532			/* enforce the random length */
6533			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6534				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6535				return (-27);
6536			}
6537			got_random = 1;
6538		} else if (ptype == SCTP_HMAC_LIST) {
6539			uint16_t num_hmacs;
6540			uint16_t i;
6541
6542			if (plen > sizeof(hmacs_store))
6543				break;
6544			if (got_hmacs) {
6545				/* already processed a HMAC list */
6546				goto next_param;
6547			}
6548			phdr = sctp_get_next_param(m, offset,
6549			    (struct sctp_paramhdr *)hmacs_store,
6550			    min(plen, sizeof(hmacs_store)));
6551			if (phdr == NULL)
6552				return (-28);
6553			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6554			hmacs_len = plen - sizeof(*hmacs);
6555			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6556			/* validate the hmac list */
6557			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6558				return (-29);
6559			}
6560			if (stcb->asoc.peer_hmacs != NULL)
6561				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6562			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6563			if (stcb->asoc.peer_hmacs != NULL) {
6564				for (i = 0; i < num_hmacs; i++) {
6565					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6566					    ntohs(hmacs->hmac_ids[i]));
6567				}
6568			}
6569			got_hmacs = 1;
6570		} else if (ptype == SCTP_CHUNK_LIST) {
6571			int i;
6572
6573			if (plen > sizeof(chunks_store))
6574				break;
6575			if (got_chklist) {
6576				/* already processed a Chunks list */
6577				goto next_param;
6578			}
6579			phdr = sctp_get_next_param(m, offset,
6580			    (struct sctp_paramhdr *)chunks_store,
6581			    min(plen, sizeof(chunks_store)));
6582			if (phdr == NULL)
6583				return (-30);
6584			chunks = (struct sctp_auth_chunk_list *)phdr;
6585			num_chunks = plen - sizeof(*chunks);
6586			if (stcb->asoc.peer_auth_chunks != NULL)
6587				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6588			else
6589				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6590			for (i = 0; i < num_chunks; i++) {
6591				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6592				    stcb->asoc.peer_auth_chunks);
6593				/* record asconf/asconf-ack if listed */
6594				if (chunks->chunk_types[i] == SCTP_ASCONF)
6595					saw_asconf = 1;
6596				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6597					saw_asconf_ack = 1;
6598
6599			}
6600			got_chklist = 1;
6601		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6602			    (ptype == SCTP_STATE_COOKIE) ||
6603			    (ptype == SCTP_UNRECOG_PARAM) ||
6604			    (ptype == SCTP_COOKIE_PRESERVE) ||
6605			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6606			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6607			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6608			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6609		    (ptype == SCTP_SUCCESS_REPORT)) {
6610			 /* don't care */ ;
6611		} else {
6612			if ((ptype & 0x8000) == 0x0000) {
6613				/*
6614				 * must stop processing the rest of the
6615				 * param's. Any report bits were handled
6616				 * with the call to
6617				 * sctp_arethere_unrecognized_parameters()
6618				 * when the INIT or INIT-ACK was first seen.
6619				 */
6620				break;
6621			}
6622		}
6623
6624next_param:
6625		offset += SCTP_SIZE32(plen);
6626		if (offset >= limit) {
6627			break;
6628		}
6629		phdr = sctp_get_next_param(m, offset, &parm_buf,
6630		    sizeof(parm_buf));
6631	}
6632	/* Now check to see if we need to purge any addresses */
6633	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6634		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6635		    SCTP_ADDR_NOT_IN_ASSOC) {
6636			/* This address has been removed from the asoc */
6637			/* remove and free it */
6638			stcb->asoc.numnets--;
6639			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6640			sctp_free_remote_addr(net);
6641			if (net == stcb->asoc.primary_destination) {
6642				stcb->asoc.primary_destination = NULL;
6643				sctp_select_primary_destination(stcb);
6644			}
6645		}
6646	}
6647	if ((stcb->asoc.ecn_supported == 1) &&
6648	    (peer_supports_ecn == 0)) {
6649		stcb->asoc.ecn_supported = 0;
6650	}
6651	if ((stcb->asoc.prsctp_supported == 1) &&
6652	    (peer_supports_prsctp == 0)) {
6653		stcb->asoc.prsctp_supported = 0;
6654	}
6655	if ((stcb->asoc.auth_supported == 1) &&
6656	    ((peer_supports_auth == 0) ||
6657	    (got_random == 0) || (got_hmacs == 0))) {
6658		stcb->asoc.auth_supported = 0;
6659	}
6660	if ((stcb->asoc.asconf_supported == 1) &&
6661	    ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) ||
6662	    (stcb->asoc.auth_supported == 0) ||
6663	    (saw_asconf == 0) || (saw_asconf_ack == 0))) {
6664		stcb->asoc.asconf_supported = 0;
6665	}
6666	if ((stcb->asoc.reconfig_supported == 1) &&
6667	    (peer_supports_reconfig == 0)) {
6668		stcb->asoc.reconfig_supported = 0;
6669	}
6670	if ((stcb->asoc.idata_supported == 1) &&
6671	    (peer_supports_idata == 0)) {
6672		stcb->asoc.idata_supported = 0;
6673	}
6674	if ((stcb->asoc.nrsack_supported == 1) &&
6675	    (peer_supports_nrsack == 0)) {
6676		stcb->asoc.nrsack_supported = 0;
6677	}
6678	if ((stcb->asoc.pktdrop_supported == 1) &&
6679	    (peer_supports_pktdrop == 0)) {
6680		stcb->asoc.pktdrop_supported = 0;
6681	}
6682	/* validate authentication required parameters */
6683	if ((peer_supports_auth == 0) && (got_chklist == 1)) {
6684		/* peer does not support auth but sent a chunks list? */
6685		return (-31);
6686	}
6687	if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) {
6688		/* peer supports asconf but not auth? */
6689		return (-32);
6690	} else if ((peer_supports_asconf == 1) &&
6691		    (peer_supports_auth == 1) &&
6692	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6693		return (-33);
6694	}
6695	/* concatenate the full random key */
6696	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6697	if (chunks != NULL) {
6698		keylen += sizeof(*chunks) + num_chunks;
6699	}
6700	new_key = sctp_alloc_key(keylen);
6701	if (new_key != NULL) {
6702		/* copy in the RANDOM */
6703		if (p_random != NULL) {
6704			keylen = sizeof(*p_random) + random_len;
6705			bcopy(p_random, new_key->key, keylen);
6706		}
6707		/* append in the AUTH chunks */
6708		if (chunks != NULL) {
6709			bcopy(chunks, new_key->key + keylen,
6710			    sizeof(*chunks) + num_chunks);
6711			keylen += sizeof(*chunks) + num_chunks;
6712		}
6713		/* append in the HMACs */
6714		if (hmacs != NULL) {
6715			bcopy(hmacs, new_key->key + keylen,
6716			    sizeof(*hmacs) + hmacs_len);
6717		}
6718	} else {
6719		/* failed to get memory for the key */
6720		return (-34);
6721	}
6722	if (stcb->asoc.authinfo.peer_random != NULL)
6723		sctp_free_key(stcb->asoc.authinfo.peer_random);
6724	stcb->asoc.authinfo.peer_random = new_key;
6725	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6726	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6727
6728	return (0);
6729}
6730
6731int
6732sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6733    struct sctp_nets *net)
6734{
6735	/* make sure the requested primary address exists in the assoc */
6736	if (net == NULL && sa)
6737		net = sctp_findnet(stcb, sa);
6738
6739	if (net == NULL) {
6740		/* didn't find the requested primary address! */
6741		return (-1);
6742	} else {
6743		/* set the primary address */
6744		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6745			/* Must be confirmed, so queue to set */
6746			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6747			return (0);
6748		}
6749		stcb->asoc.primary_destination = net;
6750		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6751			sctp_free_remote_addr(stcb->asoc.alternate);
6752			stcb->asoc.alternate = NULL;
6753		}
6754		net = TAILQ_FIRST(&stcb->asoc.nets);
6755		if (net != stcb->asoc.primary_destination) {
6756			/*
6757			 * first one on the list is NOT the primary
6758			 * sctp_cmpaddr() is much more efficient if the
6759			 * primary is the first on the list, make it so.
6760			 */
6761			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6762			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6763		}
6764		return (0);
6765	}
6766}
6767
6768int
6769sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6770{
6771	/*
6772	 * This function serves two purposes. It will see if a TAG can be
6773	 * re-used and return 1 for yes it is ok and 0 for don't use that
6774	 * tag. A secondary function it will do is purge out old tags that
6775	 * can be removed.
6776	 */
6777	struct sctpvtaghead *chain;
6778	struct sctp_tagblock *twait_block;
6779	struct sctpasochead *head;
6780	struct sctp_tcb *stcb;
6781	int i;
6782
6783	SCTP_INP_INFO_RLOCK();
6784	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6785	    SCTP_BASE_INFO(hashasocmark))];
6786	LIST_FOREACH(stcb, head, sctp_asocs) {
6787		/*
6788		 * We choose not to lock anything here. TCB's can't be
6789		 * removed since we have the read lock, so they can't be
6790		 * freed on us, same thing for the INP. I may be wrong with
6791		 * this assumption, but we will go with it for now :-)
6792		 */
6793		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6794			continue;
6795		}
6796		if (stcb->asoc.my_vtag == tag) {
6797			/* candidate */
6798			if (stcb->rport != rport) {
6799				continue;
6800			}
6801			if (stcb->sctp_ep->sctp_lport != lport) {
6802				continue;
6803			}
6804			/* Its a used tag set */
6805			SCTP_INP_INFO_RUNLOCK();
6806			return (0);
6807		}
6808	}
6809	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6810	/* Now what about timed wait ? */
6811	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6812		/*
6813		 * Block(s) are present, lets see if we have this tag in the
6814		 * list
6815		 */
6816		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6817			if (twait_block->vtag_block[i].v_tag == 0) {
6818				/* not used */
6819				continue;
6820			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6821			    now->tv_sec) {
6822				/* Audit expires this guy */
6823				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6824				twait_block->vtag_block[i].v_tag = 0;
6825				twait_block->vtag_block[i].lport = 0;
6826				twait_block->vtag_block[i].rport = 0;
6827			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6828				    (twait_block->vtag_block[i].lport == lport) &&
6829			    (twait_block->vtag_block[i].rport == rport)) {
6830				/* Bad tag, sorry :< */
6831				SCTP_INP_INFO_RUNLOCK();
6832				return (0);
6833			}
6834		}
6835	}
6836	SCTP_INP_INFO_RUNLOCK();
6837	return (1);
6838}
6839
6840static void
6841sctp_drain_mbufs(struct sctp_tcb *stcb)
6842{
6843	/*
6844	 * We must hunt this association for MBUF's past the cumack (i.e.
6845	 * out of order data that we can renege on).
6846	 */
6847	struct sctp_association *asoc;
6848	struct sctp_tmit_chunk *chk, *nchk;
6849	uint32_t cumulative_tsn_p1;
6850	struct sctp_queued_to_read *ctl, *nctl;
6851	int cnt, strmat;
6852	uint32_t gap, i;
6853	int fnd = 0;
6854
6855	/* We look for anything larger than the cum-ack + 1 */
6856
6857	asoc = &stcb->asoc;
6858	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6859		/* none we can reneg on. */
6860		return;
6861	}
6862	SCTP_STAT_INCR(sctps_protocol_drains_done);
6863	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6864	cnt = 0;
6865	/* Ok that was fun, now we will drain all the inbound streams? */
6866	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6867		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next_instrm, nctl) {
6868#ifdef INVARIANTS
6869			if (ctl->on_strm_q != SCTP_ON_ORDERED) {
6870				panic("Huh control: %p on_q: %d -- not ordered?",
6871				    ctl, ctl->on_strm_q);
6872			}
6873#endif
6874			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6875				/* Yep it is above cum-ack */
6876				cnt++;
6877				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6878				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6879				sctp_ucount_decr(asoc->cnt_on_all_streams);
6880				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6881				if (ctl->on_read_q) {
6882					TAILQ_REMOVE(&stcb->sctp_ep->read_queue, ctl, next);
6883					ctl->on_read_q = 0;
6884				}
6885				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next_instrm);
6886				ctl->on_strm_q = 0;
6887				if (ctl->data) {
6888					sctp_m_freem(ctl->data);
6889					ctl->data = NULL;
6890				}
6891				sctp_free_remote_addr(ctl->whoFrom);
6892				/* Now its reasm? */
6893				TAILQ_FOREACH_SAFE(chk, &ctl->reasm, sctp_next, nchk) {
6894					cnt++;
6895					SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6896					asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6897					sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6898					SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6899					TAILQ_REMOVE(&ctl->reasm, chk, sctp_next);
6900					if (chk->data) {
6901						sctp_m_freem(chk->data);
6902						chk->data = NULL;
6903					}
6904					sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6905				}
6906				sctp_free_a_readq(stcb, ctl);
6907			}
6908		}
6909		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].uno_inqueue, next_instrm, nctl) {
6910#ifdef INVARIANTS
6911			if (ctl->on_strm_q != SCTP_ON_UNORDERED) {
6912				panic("Huh control: %p on_q: %d -- not unordered?",
6913				    ctl, ctl->on_strm_q);
6914			}
6915#endif
6916			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6917				/* Yep it is above cum-ack */
6918				cnt++;
6919				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6920				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6921				sctp_ucount_decr(asoc->cnt_on_all_streams);
6922				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6923				if (ctl->on_read_q) {
6924					TAILQ_REMOVE(&stcb->sctp_ep->read_queue, ctl, next);
6925					ctl->on_read_q = 0;
6926				}
6927				TAILQ_REMOVE(&asoc->strmin[strmat].uno_inqueue, ctl, next_instrm);
6928				ctl->on_strm_q = 0;
6929				if (ctl->data) {
6930					sctp_m_freem(ctl->data);
6931					ctl->data = NULL;
6932				}
6933				sctp_free_remote_addr(ctl->whoFrom);
6934				/* Now its reasm? */
6935				TAILQ_FOREACH_SAFE(chk, &ctl->reasm, sctp_next, nchk) {
6936					cnt++;
6937					SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6938					asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6939					sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6940					SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6941					TAILQ_REMOVE(&ctl->reasm, chk, sctp_next);
6942					if (chk->data) {
6943						sctp_m_freem(chk->data);
6944						chk->data = NULL;
6945					}
6946					sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6947				}
6948				sctp_free_a_readq(stcb, ctl);
6949			}
6950		}
6951	}
6952	if (cnt) {
6953		/* We must back down to see what the new highest is */
6954		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6955			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6956			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6957				asoc->highest_tsn_inside_map = i;
6958				fnd = 1;
6959				break;
6960			}
6961		}
6962		if (!fnd) {
6963			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6964		}
6965		/*
6966		 * Question, should we go through the delivery queue? The
6967		 * only reason things are on here is the app not reading OR
6968		 * a p-d-api up. An attacker COULD send enough in to
6969		 * initiate the PD-API and then send a bunch of stuff to
6970		 * other streams... these would wind up on the delivery
6971		 * queue.. and then we would not get to them. But in order
6972		 * to do this I then have to back-track and un-deliver
6973		 * sequence numbers in streams.. el-yucko. I think for now
6974		 * we will NOT look at the delivery queue and leave it to be
6975		 * something to consider later. An alternative would be to
6976		 * abort the P-D-API with a notification and then deliver
6977		 * the data.... Or another method might be to keep track of
6978		 * how many times the situation occurs and if we see a
6979		 * possible attack underway just abort the association.
6980		 */
6981#ifdef SCTP_DEBUG
6982		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6983#endif
6984		/*
6985		 * Now do we need to find a new
6986		 * asoc->highest_tsn_inside_map?
6987		 */
6988		asoc->last_revoke_count = cnt;
6989		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6990		/* sa_ignore NO_NULL_CHK */
6991		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6992		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6993	}
6994	/*
6995	 * Another issue, in un-setting the TSN's in the mapping array we
6996	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6997	 * two things to occur. It may cause us to do extra work in checking
6998	 * for our mapping array movement. More importantly it may cause us
6999	 * to SACK every datagram. This may not be a bad thing though since
7000	 * we will recover once we get our cum-ack above and all this stuff
7001	 * we dumped recovered.
7002	 */
7003}
7004
7005void
7006sctp_drain()
7007{
7008	/*
7009	 * We must walk the PCB lists for ALL associations here. The system
7010	 * is LOW on MBUF's and needs help. This is where reneging will
7011	 * occur. We really hope this does NOT happen!
7012	 */
7013	VNET_ITERATOR_DECL(vnet_iter);
7014	VNET_LIST_RLOCK_NOSLEEP();
7015	VNET_FOREACH(vnet_iter) {
7016		CURVNET_SET(vnet_iter);
7017		struct sctp_inpcb *inp;
7018		struct sctp_tcb *stcb;
7019
7020		SCTP_STAT_INCR(sctps_protocol_drain_calls);
7021		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
7022#ifdef VIMAGE
7023			continue;
7024#else
7025			return;
7026#endif
7027		}
7028		SCTP_INP_INFO_RLOCK();
7029		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
7030			/* For each endpoint */
7031			SCTP_INP_RLOCK(inp);
7032			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
7033				/* For each association */
7034				SCTP_TCB_LOCK(stcb);
7035				sctp_drain_mbufs(stcb);
7036				SCTP_TCB_UNLOCK(stcb);
7037			}
7038			SCTP_INP_RUNLOCK(inp);
7039		}
7040		SCTP_INP_INFO_RUNLOCK();
7041		CURVNET_RESTORE();
7042	}
7043	VNET_LIST_RUNLOCK_NOSLEEP();
7044}
7045
7046/*
7047 * start a new iterator
7048 * iterates through all endpoints and associations based on the pcb_state
7049 * flags and asoc_state.  "af" (mandatory) is executed for all matching
7050 * assocs and "ef" (optional) is executed when the iterator completes.
7051 * "inpf" (optional) is executed for each new endpoint as it is being
7052 * iterated through. inpe (optional) is called when the inp completes
7053 * its way through all the stcbs.
7054 */
7055int
7056sctp_initiate_iterator(inp_func inpf,
7057    asoc_func af,
7058    inp_func inpe,
7059    uint32_t pcb_state,
7060    uint32_t pcb_features,
7061    uint32_t asoc_state,
7062    void *argp,
7063    uint32_t argi,
7064    end_func ef,
7065    struct sctp_inpcb *s_inp,
7066    uint8_t chunk_output_off)
7067{
7068	struct sctp_iterator *it = NULL;
7069
7070	if (af == NULL) {
7071		return (-1);
7072	}
7073	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
7074		SCTP_PRINTF("%s: abort on initialize being %d\n", __func__,
7075		    SCTP_BASE_VAR(sctp_pcb_initialized));
7076		return (-1);
7077	}
7078	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
7079	    SCTP_M_ITER);
7080	if (it == NULL) {
7081		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
7082		return (ENOMEM);
7083	}
7084	memset(it, 0, sizeof(*it));
7085	it->function_assoc = af;
7086	it->function_inp = inpf;
7087	if (inpf)
7088		it->done_current_ep = 0;
7089	else
7090		it->done_current_ep = 1;
7091	it->function_atend = ef;
7092	it->pointer = argp;
7093	it->val = argi;
7094	it->pcb_flags = pcb_state;
7095	it->pcb_features = pcb_features;
7096	it->asoc_state = asoc_state;
7097	it->function_inp_end = inpe;
7098	it->no_chunk_output = chunk_output_off;
7099	it->vn = curvnet;
7100	if (s_inp) {
7101		/* Assume lock is held here */
7102		it->inp = s_inp;
7103		SCTP_INP_INCR_REF(it->inp);
7104		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7105	} else {
7106		SCTP_INP_INFO_RLOCK();
7107		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7108		if (it->inp) {
7109			SCTP_INP_INCR_REF(it->inp);
7110		}
7111		SCTP_INP_INFO_RUNLOCK();
7112		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7113
7114	}
7115	SCTP_IPI_ITERATOR_WQ_LOCK();
7116	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
7117		SCTP_IPI_ITERATOR_WQ_UNLOCK();
7118		SCTP_PRINTF("%s: rollback on initialize being %d it=%p\n", __func__,
7119		    SCTP_BASE_VAR(sctp_pcb_initialized), it);
7120		SCTP_FREE(it, SCTP_M_ITER);
7121		return (-1);
7122	}
7123	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7124	if (sctp_it_ctl.iterator_running == 0) {
7125		sctp_wakeup_iterator();
7126	}
7127	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7128	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7129	return (0);
7130}
7131