1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
5 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
6 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
10 *
11 * a) Redistributions of source code must retain the above copyright notice,
12 *    this list of conditions and the following disclaimer.
13 *
14 * b) Redistributions in binary form must reproduce the above copyright
15 *    notice, this list of conditions and the following disclaimer in
16 *    the documentation and/or other materials provided with the distribution.
17 *
18 * c) Neither the name of Cisco Systems, Inc. nor the names of its
19 *    contributors may be used to endorse or promote products derived
20 *    from this software without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
24 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
26 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
29 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
30 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
31 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
32 * THE POSSIBILITY OF SUCH DAMAGE.
33 */
34
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD$");
37
38#include <netinet/sctp_os.h>
39#include <sys/proc.h>
40#include <netinet/sctp_var.h>
41#include <netinet/sctp_sysctl.h>
42#include <netinet/sctp_pcb.h>
43#include <netinet/sctputil.h>
44#include <netinet/sctp.h>
45#include <netinet/sctp_header.h>
46#include <netinet/sctp_asconf.h>
47#include <netinet/sctp_output.h>
48#include <netinet/sctp_timer.h>
49#include <netinet/sctp_bsd_addr.h>
50#if defined(INET) || defined(INET6)
51#include <netinet/udp.h>
52#endif
53#ifdef INET6
54#include <netinet6/ip6_var.h>
55#endif
56#include <sys/sched.h>
57#include <sys/smp.h>
58#include <sys/unistd.h>
59
60/* FIX: we don't handle multiple link local scopes */
61/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
62#ifdef INET6
63int
64SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
65{
66	struct sockaddr_in6 tmp_a, tmp_b;
67
68	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
69	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
70		return (0);
71	}
72	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
73	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
74		return (0);
75	}
76	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
77}
78#endif
79
80void
81sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
82{
83	/*
84	 * We really don't need to lock this, but I will just because it
85	 * does not hurt.
86	 */
87	SCTP_INP_INFO_RLOCK();
88	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
89	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
90	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
91	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
92	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
93	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
94	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
95	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
96	SCTP_INP_INFO_RUNLOCK();
97}
98
99/*-
100 * Addresses are added to VRF's (Virtual Router's). For BSD we
101 * have only the default VRF 0. We maintain a hash list of
102 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
103 * these has a list of addresses. When we add a new address
104 * to a VRF we lookup the ifn/ifn_index, if the ifn does
105 * not exist we create it and add it to the list of IFN's
106 * within the VRF. Once we have the sctp_ifn, we add the
107 * address to the list. So we look something like:
108 *
109 * hash-vrf-table
110 *   vrf-> ifn-> ifn -> ifn
111 *   vrf    |
112 *    ...   +--ifa-> ifa -> ifa
113 *   vrf
114 *
115 * We keep these separate lists since the SCTP subsystem will
116 * point to these from its source address selection nets structure.
117 * When an address is deleted it does not happen right away on
118 * the SCTP side, it gets scheduled. What we do when a
119 * delete happens is immediately remove the address from
120 * the master list and decrement the refcount. As our
121 * addip iterator works through and frees the src address
122 * selection pointing to the sctp_ifa, eventually the refcount
123 * will reach 0 and we will delete it. Note that it is assumed
124 * that any locking on system level ifn/ifa is done at the
125 * caller of these functions and these routines will only
126 * lock the SCTP structures as they add or delete things.
127 *
128 * Other notes on VRF concepts.
129 *  - An endpoint can be in multiple VRF's
130 *  - An association lives within a VRF and only one VRF.
131 *  - Any incoming packet we can deduce the VRF for by
132 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
133 *  - Any downward send call or connect call must supply the
134 *    VRF via ancillary data or via some sort of set default
135 *    VRF socket option call (again for BSD no brainer since
136 *    the VRF is always 0).
137 *  - An endpoint may add multiple VRF's to it.
138 *  - Listening sockets can accept associations in any
139 *    of the VRF's they are in but the assoc will end up
140 *    in only one VRF (gotten from the packet or connect/send).
141 *
142 */
143
144struct sctp_vrf *
145sctp_allocate_vrf(int vrf_id)
146{
147	struct sctp_vrf *vrf = NULL;
148	struct sctp_vrflist *bucket;
149
150	/* First allocate the VRF structure */
151	vrf = sctp_find_vrf(vrf_id);
152	if (vrf) {
153		/* Already allocated */
154		return (vrf);
155	}
156	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
157	    SCTP_M_VRF);
158	if (vrf == NULL) {
159		/* No memory */
160#ifdef INVARIANTS
161		panic("No memory for VRF:%d", vrf_id);
162#endif
163		return (NULL);
164	}
165	/* setup the VRF */
166	memset(vrf, 0, sizeof(struct sctp_vrf));
167	vrf->vrf_id = vrf_id;
168	LIST_INIT(&vrf->ifnlist);
169	vrf->total_ifa_count = 0;
170	vrf->refcount = 0;
171	/* now also setup table ids */
172	SCTP_INIT_VRF_TABLEID(vrf);
173	/* Init the HASH of addresses */
174	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
175	    &vrf->vrf_addr_hashmark);
176	if (vrf->vrf_addr_hash == NULL) {
177		/* No memory */
178#ifdef INVARIANTS
179		panic("No memory for VRF:%d", vrf_id);
180#endif
181		SCTP_FREE(vrf, SCTP_M_VRF);
182		return (NULL);
183	}
184
185	/* Add it to the hash table */
186	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
187	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
188	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
189	return (vrf);
190}
191
192struct sctp_ifn *
193sctp_find_ifn(void *ifn, uint32_t ifn_index)
194{
195	struct sctp_ifn *sctp_ifnp;
196	struct sctp_ifnlist *hash_ifn_head;
197
198	/*
199	 * We assume the lock is held for the addresses if that's wrong
200	 * problems could occur :-)
201	 */
202	SCTP_IPI_ADDR_LOCK_ASSERT();
203	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
204	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
205		if (sctp_ifnp->ifn_index == ifn_index) {
206			return (sctp_ifnp);
207		}
208		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
209			return (sctp_ifnp);
210		}
211	}
212	return (NULL);
213}
214
215struct sctp_vrf *
216sctp_find_vrf(uint32_t vrf_id)
217{
218	struct sctp_vrflist *bucket;
219	struct sctp_vrf *liste;
220
221	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
222	LIST_FOREACH(liste, bucket, next_vrf) {
223		if (vrf_id == liste->vrf_id) {
224			return (liste);
225		}
226	}
227	return (NULL);
228}
229
230void
231sctp_free_vrf(struct sctp_vrf *vrf)
232{
233	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
234		if (vrf->vrf_addr_hash) {
235			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
236			vrf->vrf_addr_hash = NULL;
237		}
238		/* We zero'd the count */
239		LIST_REMOVE(vrf, next_vrf);
240		SCTP_FREE(vrf, SCTP_M_VRF);
241		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
242	}
243}
244
245void
246sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
247{
248	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
249		/* We zero'd the count */
250		if (sctp_ifnp->vrf) {
251			sctp_free_vrf(sctp_ifnp->vrf);
252		}
253		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
254		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
255	}
256}
257
258void
259sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
260{
261	struct sctp_ifn *sctp_ifnp;
262
263	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
264	if (sctp_ifnp != NULL) {
265		sctp_ifnp->ifn_mtu = mtu;
266	}
267}
268
269void
270sctp_free_ifa(struct sctp_ifa *sctp_ifap)
271{
272	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
273		/* We zero'd the count */
274		if (sctp_ifap->ifn_p) {
275			sctp_free_ifn(sctp_ifap->ifn_p);
276		}
277		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
278		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
279	}
280}
281
282static void
283sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
284{
285	struct sctp_ifn *found;
286
287	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
288	if (found == NULL) {
289		/* Not in the list.. sorry */
290		return;
291	}
292	if (hold_addr_lock == 0) {
293		SCTP_IPI_ADDR_WLOCK();
294	} else {
295		SCTP_IPI_ADDR_WLOCK_ASSERT();
296	}
297	LIST_REMOVE(sctp_ifnp, next_bucket);
298	LIST_REMOVE(sctp_ifnp, next_ifn);
299	if (hold_addr_lock == 0) {
300		SCTP_IPI_ADDR_WUNLOCK();
301	}
302	/* Take away the reference, and possibly free it */
303	sctp_free_ifn(sctp_ifnp);
304}
305
306void
307sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
308    const char *if_name, uint32_t ifn_index)
309{
310	struct sctp_vrf *vrf;
311	struct sctp_ifa *sctp_ifap;
312
313	SCTP_IPI_ADDR_RLOCK();
314	vrf = sctp_find_vrf(vrf_id);
315	if (vrf == NULL) {
316		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
317		goto out;
318	}
319	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
320	if (sctp_ifap == NULL) {
321		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
322		goto out;
323	}
324	if (sctp_ifap->ifn_p == NULL) {
325		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unusable\n");
326		goto out;
327	}
328	if (if_name) {
329		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
330			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
331			    sctp_ifap->ifn_p->ifn_name, if_name);
332			goto out;
333		}
334	} else {
335		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
336			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
337			    sctp_ifap->ifn_p->ifn_index, ifn_index);
338			goto out;
339		}
340	}
341
342	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
343	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
344out:
345	SCTP_IPI_ADDR_RUNLOCK();
346}
347
348void
349sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
350    const char *if_name, uint32_t ifn_index)
351{
352	struct sctp_vrf *vrf;
353	struct sctp_ifa *sctp_ifap;
354
355	SCTP_IPI_ADDR_RLOCK();
356	vrf = sctp_find_vrf(vrf_id);
357	if (vrf == NULL) {
358		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
359		goto out;
360	}
361	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
362	if (sctp_ifap == NULL) {
363		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
364		goto out;
365	}
366	if (sctp_ifap->ifn_p == NULL) {
367		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unusable\n");
368		goto out;
369	}
370	if (if_name) {
371		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
372			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
373			    sctp_ifap->ifn_p->ifn_name, if_name);
374			goto out;
375		}
376	} else {
377		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
378			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
379			    sctp_ifap->ifn_p->ifn_index, ifn_index);
380			goto out;
381		}
382	}
383
384	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
385	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
386out:
387	SCTP_IPI_ADDR_RUNLOCK();
388}
389
390/*-
391 * Add an ifa to an ifn.
392 * Register the interface as necessary.
393 * NOTE: ADDR write lock MUST be held.
394 */
395static void
396sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
397{
398	int ifa_af;
399
400	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
401	sctp_ifap->ifn_p = sctp_ifnp;
402	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
403	/* update address counts */
404	sctp_ifnp->ifa_count++;
405	ifa_af = sctp_ifap->address.sa.sa_family;
406	switch (ifa_af) {
407#ifdef INET
408	case AF_INET:
409		sctp_ifnp->num_v4++;
410		break;
411#endif
412#ifdef INET6
413	case AF_INET6:
414		sctp_ifnp->num_v6++;
415		break;
416#endif
417	default:
418		break;
419	}
420	if (sctp_ifnp->ifa_count == 1) {
421		/* register the new interface */
422		sctp_ifnp->registered_af = ifa_af;
423	}
424}
425
426/*-
427 * Remove an ifa from its ifn.
428 * If no more addresses exist, remove the ifn too. Otherwise, re-register
429 * the interface based on the remaining address families left.
430 * NOTE: ADDR write lock MUST be held.
431 */
432static void
433sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
434{
435	LIST_REMOVE(sctp_ifap, next_ifa);
436	if (sctp_ifap->ifn_p) {
437		/* update address counts */
438		sctp_ifap->ifn_p->ifa_count--;
439		switch (sctp_ifap->address.sa.sa_family) {
440#ifdef INET
441		case AF_INET:
442			sctp_ifap->ifn_p->num_v4--;
443			break;
444#endif
445#ifdef INET6
446		case AF_INET6:
447			sctp_ifap->ifn_p->num_v6--;
448			break;
449#endif
450		default:
451			break;
452		}
453
454		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
455			/* remove the ifn, possibly freeing it */
456			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
457		} else {
458			/* re-register address family type, if needed */
459			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
460			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
461				sctp_ifap->ifn_p->registered_af = AF_INET;
462			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
463			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
464				sctp_ifap->ifn_p->registered_af = AF_INET6;
465			}
466			/* free the ifn refcount */
467			sctp_free_ifn(sctp_ifap->ifn_p);
468		}
469		sctp_ifap->ifn_p = NULL;
470	}
471}
472
473struct sctp_ifa *
474sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
475    uint32_t ifn_type, const char *if_name, void *ifa,
476    struct sockaddr *addr, uint32_t ifa_flags,
477    int dynamic_add)
478{
479	struct sctp_vrf *vrf;
480	struct sctp_ifn *sctp_ifnp, *new_sctp_ifnp;
481	struct sctp_ifa *sctp_ifap, *new_sctp_ifap;
482	struct sctp_ifalist *hash_addr_head;
483	struct sctp_ifnlist *hash_ifn_head;
484	uint32_t hash_of_addr;
485	int new_ifn_af = 0;
486
487#ifdef SCTP_DEBUG
488	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
489	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
490#endif
491	SCTP_MALLOC(new_sctp_ifnp, struct sctp_ifn *,
492	    sizeof(struct sctp_ifn), SCTP_M_IFN);
493	if (new_sctp_ifnp == NULL) {
494#ifdef INVARIANTS
495		panic("No memory for IFN");
496#endif
497		return (NULL);
498	}
499	SCTP_MALLOC(new_sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
500	if (new_sctp_ifap == NULL) {
501#ifdef INVARIANTS
502		panic("No memory for IFA");
503#endif
504		SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN);
505		return (NULL);
506	}
507
508	SCTP_IPI_ADDR_WLOCK();
509	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
510	if (sctp_ifnp) {
511		vrf = sctp_ifnp->vrf;
512	} else {
513		vrf = sctp_find_vrf(vrf_id);
514		if (vrf == NULL) {
515			vrf = sctp_allocate_vrf(vrf_id);
516			if (vrf == NULL) {
517				SCTP_IPI_ADDR_WUNLOCK();
518				SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN);
519				SCTP_FREE(new_sctp_ifap, SCTP_M_IFA);
520				return (NULL);
521			}
522		}
523	}
524	if (sctp_ifnp == NULL) {
525		/*
526		 * build one and add it, can't hold lock until after malloc
527		 * done though.
528		 */
529		sctp_ifnp = new_sctp_ifnp;
530		new_sctp_ifnp = NULL;
531		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
532		sctp_ifnp->ifn_index = ifn_index;
533		sctp_ifnp->ifn_p = ifn;
534		sctp_ifnp->ifn_type = ifn_type;
535		sctp_ifnp->refcount = 0;
536		sctp_ifnp->vrf = vrf;
537		atomic_add_int(&vrf->refcount, 1);
538		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
539		if (if_name != NULL) {
540			SCTP_SNPRINTF(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
541		} else {
542			SCTP_SNPRINTF(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
543		}
544		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
545		LIST_INIT(&sctp_ifnp->ifalist);
546		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
547		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
548		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
549		new_ifn_af = 1;
550	}
551	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
552	if (sctp_ifap) {
553		/* Hmm, it already exists? */
554		if ((sctp_ifap->ifn_p) &&
555		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
556			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
557			    sctp_ifap->ifn_p->ifn_name, ifn_index,
558			    (void *)sctp_ifap);
559			if (new_ifn_af) {
560				/* Remove the created one that we don't want */
561				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
562			}
563			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
564				/* easy to solve, just switch back to active */
565				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
566				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
567				sctp_ifap->ifn_p = sctp_ifnp;
568				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
569			}
570	exit_stage_left:
571			SCTP_IPI_ADDR_WUNLOCK();
572			if (new_sctp_ifnp != NULL) {
573				SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN);
574			}
575			SCTP_FREE(new_sctp_ifap, SCTP_M_IFA);
576			return (sctp_ifap);
577		} else {
578			if (sctp_ifap->ifn_p) {
579				/*
580				 * The last IFN gets the address, remove the
581				 * old one
582				 */
583				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
584				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
585				    sctp_ifap->ifn_p->ifn_index, if_name,
586				    ifn_index);
587				/* remove the address from the old ifn */
588				sctp_remove_ifa_from_ifn(sctp_ifap);
589				/* move the address over to the new ifn */
590				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
591				goto exit_stage_left;
592			} else {
593				/* repair ifnp which was NULL ? */
594				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
595				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
596				    (void *)sctp_ifnp, (void *)sctp_ifap);
597				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
598			}
599			goto exit_stage_left;
600		}
601	}
602	sctp_ifap = new_sctp_ifap;
603	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
604	sctp_ifap->ifn_p = sctp_ifnp;
605	atomic_add_int(&sctp_ifnp->refcount, 1);
606	sctp_ifap->vrf_id = vrf_id;
607	sctp_ifap->ifa = ifa;
608	memcpy(&sctp_ifap->address, addr, addr->sa_len);
609	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
610	sctp_ifap->flags = ifa_flags;
611	/* Set scope */
612	switch (sctp_ifap->address.sa.sa_family) {
613#ifdef INET
614	case AF_INET:
615		{
616			struct sockaddr_in *sin;
617
618			sin = &sctp_ifap->address.sin;
619			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
620			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
621				sctp_ifap->src_is_loop = 1;
622			}
623			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
624				sctp_ifap->src_is_priv = 1;
625			}
626			sctp_ifnp->num_v4++;
627			if (new_ifn_af)
628				new_ifn_af = AF_INET;
629			break;
630		}
631#endif
632#ifdef INET6
633	case AF_INET6:
634		{
635			/* ok to use deprecated addresses? */
636			struct sockaddr_in6 *sin6;
637
638			sin6 = &sctp_ifap->address.sin6;
639			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
640			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
641				sctp_ifap->src_is_loop = 1;
642			}
643			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
644				sctp_ifap->src_is_priv = 1;
645			}
646			sctp_ifnp->num_v6++;
647			if (new_ifn_af)
648				new_ifn_af = AF_INET6;
649			break;
650		}
651#endif
652	default:
653		new_ifn_af = 0;
654		break;
655	}
656	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
657
658	if ((sctp_ifap->src_is_priv == 0) &&
659	    (sctp_ifap->src_is_loop == 0)) {
660		sctp_ifap->src_is_glob = 1;
661	}
662	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
663	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
664	sctp_ifap->refcount = 1;
665	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
666	sctp_ifnp->ifa_count++;
667	vrf->total_ifa_count++;
668	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
669	if (new_ifn_af) {
670		sctp_ifnp->registered_af = new_ifn_af;
671	}
672	SCTP_IPI_ADDR_WUNLOCK();
673	if (new_sctp_ifnp != NULL) {
674		SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN);
675	}
676
677	if (dynamic_add) {
678		/*
679		 * Bump up the refcount so that when the timer completes it
680		 * will drop back down.
681		 */
682		struct sctp_laddr *wi;
683
684		atomic_add_int(&sctp_ifap->refcount, 1);
685		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
686		if (wi == NULL) {
687			/*
688			 * Gak, what can we do? We have lost an address
689			 * change can you say HOSED?
690			 */
691			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
692			/* Opps, must decrement the count */
693			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
694			    if_name);
695			return (NULL);
696		}
697		SCTP_INCR_LADDR_COUNT();
698		memset(wi, 0, sizeof(*wi));
699		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
700		wi->ifa = sctp_ifap;
701		wi->action = SCTP_ADD_IP_ADDRESS;
702
703		SCTP_WQ_ADDR_LOCK();
704		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
705		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
706		    (struct sctp_inpcb *)NULL,
707		    (struct sctp_tcb *)NULL,
708		    (struct sctp_nets *)NULL);
709		SCTP_WQ_ADDR_UNLOCK();
710	} else {
711		/* it's ready for use */
712		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
713	}
714	return (sctp_ifap);
715}
716
717void
718sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
719    uint32_t ifn_index, const char *if_name)
720{
721	struct sctp_vrf *vrf;
722	struct sctp_ifa *sctp_ifap = NULL;
723
724	SCTP_IPI_ADDR_WLOCK();
725	vrf = sctp_find_vrf(vrf_id);
726	if (vrf == NULL) {
727		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
728		goto out_now;
729	}
730
731#ifdef SCTP_DEBUG
732	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
733	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
734#endif
735	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
736	if (sctp_ifap) {
737		/* Validate the delete */
738		if (sctp_ifap->ifn_p) {
739			int valid = 0;
740
741			/*-
742			 * The name has priority over the ifn_index
743			 * if its given.
744			 */
745			if (if_name) {
746				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
747					/* They match its a correct delete */
748					valid = 1;
749				}
750			}
751			if (!valid) {
752				/* last ditch check ifn_index */
753				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
754					valid = 1;
755				}
756			}
757			if (!valid) {
758				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
759				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
760				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
761				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
762				SCTP_IPI_ADDR_WUNLOCK();
763				return;
764			}
765		}
766		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
767		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
768		/*
769		 * We don't set the flag. This means that the structure will
770		 * hang around in EP's that have bound specific to it until
771		 * they close. This gives us TCP like behavior if someone
772		 * removes an address (or for that matter adds it right
773		 * back).
774		 */
775		/* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */
776		vrf->total_ifa_count--;
777		LIST_REMOVE(sctp_ifap, next_bucket);
778		sctp_remove_ifa_from_ifn(sctp_ifap);
779	}
780#ifdef SCTP_DEBUG
781	else {
782		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
783		    ifn_index);
784		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
785	}
786#endif
787
788out_now:
789	SCTP_IPI_ADDR_WUNLOCK();
790	if (sctp_ifap) {
791		struct sctp_laddr *wi;
792
793		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
794		if (wi == NULL) {
795			/*
796			 * Gak, what can we do? We have lost an address
797			 * change can you say HOSED?
798			 */
799			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
800
801			/* Oops, must decrement the count */
802			sctp_free_ifa(sctp_ifap);
803			return;
804		}
805		SCTP_INCR_LADDR_COUNT();
806		memset(wi, 0, sizeof(*wi));
807		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
808		wi->ifa = sctp_ifap;
809		wi->action = SCTP_DEL_IP_ADDRESS;
810		SCTP_WQ_ADDR_LOCK();
811		/*
812		 * Should this really be a tailq? As it is we will process
813		 * the newest first :-0
814		 */
815		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
816		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
817		    (struct sctp_inpcb *)NULL,
818		    (struct sctp_tcb *)NULL,
819		    (struct sctp_nets *)NULL);
820		SCTP_WQ_ADDR_UNLOCK();
821	}
822	return;
823}
824
825static int
826sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
827{
828	int loopback_scope;
829#if defined(INET)
830	int ipv4_local_scope, ipv4_addr_legal;
831#endif
832#if defined(INET6)
833	int local_scope, site_scope, ipv6_addr_legal;
834#endif
835	struct sctp_vrf *vrf;
836	struct sctp_ifn *sctp_ifn;
837	struct sctp_ifa *sctp_ifa;
838
839	loopback_scope = stcb->asoc.scope.loopback_scope;
840#if defined(INET)
841	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
842	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
843#endif
844#if defined(INET6)
845	local_scope = stcb->asoc.scope.local_scope;
846	site_scope = stcb->asoc.scope.site_scope;
847	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
848#endif
849
850	SCTP_IPI_ADDR_RLOCK();
851	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
852	if (vrf == NULL) {
853		/* no vrf, no addresses */
854		SCTP_IPI_ADDR_RUNLOCK();
855		return (0);
856	}
857
858	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
859		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
860			if ((loopback_scope == 0) &&
861			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
862				continue;
863			}
864			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
865				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
866				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
867					/*
868					 * We allow pending addresses, where
869					 * we have sent an asconf-add to be
870					 * considered valid.
871					 */
872					continue;
873				}
874				if (sctp_ifa->address.sa.sa_family != to->sa_family) {
875					continue;
876				}
877				switch (sctp_ifa->address.sa.sa_family) {
878#ifdef INET
879				case AF_INET:
880					if (ipv4_addr_legal) {
881						struct sockaddr_in *sin,
882						           *rsin;
883
884						sin = &sctp_ifa->address.sin;
885						rsin = (struct sockaddr_in *)to;
886						if ((ipv4_local_scope == 0) &&
887						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
888							continue;
889						}
890						if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred,
891						    &sin->sin_addr) != 0) {
892							continue;
893						}
894						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
895							SCTP_IPI_ADDR_RUNLOCK();
896							return (1);
897						}
898					}
899					break;
900#endif
901#ifdef INET6
902				case AF_INET6:
903					if (ipv6_addr_legal) {
904						struct sockaddr_in6 *sin6,
905						            *rsin6;
906
907						sin6 = &sctp_ifa->address.sin6;
908						rsin6 = (struct sockaddr_in6 *)to;
909						if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred,
910						    &sin6->sin6_addr) != 0) {
911							continue;
912						}
913						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
914							if (local_scope == 0)
915								continue;
916							if (sin6->sin6_scope_id == 0) {
917								if (sa6_recoverscope(sin6) != 0)
918									continue;
919							}
920						}
921						if ((site_scope == 0) &&
922						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
923							continue;
924						}
925						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
926							SCTP_IPI_ADDR_RUNLOCK();
927							return (1);
928						}
929					}
930					break;
931#endif
932				default:
933					/* TSNH */
934					break;
935				}
936			}
937		}
938	} else {
939		struct sctp_laddr *laddr;
940
941		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
942			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
943				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
944				continue;
945			}
946			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
947			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
948				/*
949				 * We allow pending addresses, where we have
950				 * sent an asconf-add to be considered
951				 * valid.
952				 */
953				continue;
954			}
955			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
956				continue;
957			}
958			switch (to->sa_family) {
959#ifdef INET
960			case AF_INET:
961				{
962					struct sockaddr_in *sin, *rsin;
963
964					sin = &laddr->ifa->address.sin;
965					rsin = (struct sockaddr_in *)to;
966					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
967						SCTP_IPI_ADDR_RUNLOCK();
968						return (1);
969					}
970					break;
971				}
972#endif
973#ifdef INET6
974			case AF_INET6:
975				{
976					struct sockaddr_in6 *sin6, *rsin6;
977
978					sin6 = &laddr->ifa->address.sin6;
979					rsin6 = (struct sockaddr_in6 *)to;
980					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
981						SCTP_IPI_ADDR_RUNLOCK();
982						return (1);
983					}
984					break;
985				}
986
987#endif
988			default:
989				/* TSNH */
990				break;
991			}
992		}
993	}
994	SCTP_IPI_ADDR_RUNLOCK();
995	return (0);
996}
997
998static struct sctp_tcb *
999sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
1000    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
1001{
1002	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
1003	/*
1004	 * If we support the TCP model, then we must now dig through to see
1005	 * if we can find our endpoint in the list of tcp ep's.
1006	 */
1007	uint16_t lport, rport;
1008	struct sctppcbhead *ephead;
1009	struct sctp_inpcb *inp;
1010	struct sctp_laddr *laddr;
1011	struct sctp_tcb *stcb;
1012	struct sctp_nets *net;
1013
1014	if ((to == NULL) || (from == NULL)) {
1015		return (NULL);
1016	}
1017
1018	switch (to->sa_family) {
1019#ifdef INET
1020	case AF_INET:
1021		if (from->sa_family == AF_INET) {
1022			lport = ((struct sockaddr_in *)to)->sin_port;
1023			rport = ((struct sockaddr_in *)from)->sin_port;
1024		} else {
1025			return (NULL);
1026		}
1027		break;
1028#endif
1029#ifdef INET6
1030	case AF_INET6:
1031		if (from->sa_family == AF_INET6) {
1032			lport = ((struct sockaddr_in6 *)to)->sin6_port;
1033			rport = ((struct sockaddr_in6 *)from)->sin6_port;
1034		} else {
1035			return (NULL);
1036		}
1037		break;
1038#endif
1039	default:
1040		return (NULL);
1041	}
1042	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
1043	/*
1044	 * Ok now for each of the guys in this bucket we must look and see:
1045	 * - Does the remote port match. - Does there single association's
1046	 * addresses match this address (to). If so we update p_ep to point
1047	 * to this ep and return the tcb from it.
1048	 */
1049	LIST_FOREACH(inp, ephead, sctp_hash) {
1050		SCTP_INP_RLOCK(inp);
1051		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1052			SCTP_INP_RUNLOCK(inp);
1053			continue;
1054		}
1055		if (lport != inp->sctp_lport) {
1056			SCTP_INP_RUNLOCK(inp);
1057			continue;
1058		}
1059		switch (to->sa_family) {
1060#ifdef INET
1061		case AF_INET:
1062			{
1063				struct sockaddr_in *sin;
1064
1065				sin = (struct sockaddr_in *)to;
1066				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1067				    &sin->sin_addr) != 0) {
1068					SCTP_INP_RUNLOCK(inp);
1069					continue;
1070				}
1071				break;
1072			}
1073#endif
1074#ifdef INET6
1075		case AF_INET6:
1076			{
1077				struct sockaddr_in6 *sin6;
1078
1079				sin6 = (struct sockaddr_in6 *)to;
1080				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1081				    &sin6->sin6_addr) != 0) {
1082					SCTP_INP_RUNLOCK(inp);
1083					continue;
1084				}
1085				break;
1086			}
1087#endif
1088		default:
1089			SCTP_INP_RUNLOCK(inp);
1090			continue;
1091		}
1092		if (inp->def_vrf_id != vrf_id) {
1093			SCTP_INP_RUNLOCK(inp);
1094			continue;
1095		}
1096		/* check to see if the ep has one of the addresses */
1097		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1098			/* We are NOT bound all, so look further */
1099			int match = 0;
1100
1101			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1102				if (laddr->ifa == NULL) {
1103					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __func__);
1104					continue;
1105				}
1106				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1107					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1108					continue;
1109				}
1110				if (laddr->ifa->address.sa.sa_family ==
1111				    to->sa_family) {
1112					/* see if it matches */
1113#ifdef INET
1114					if (from->sa_family == AF_INET) {
1115						struct sockaddr_in *intf_addr,
1116						           *sin;
1117
1118						intf_addr = &laddr->ifa->address.sin;
1119						sin = (struct sockaddr_in *)to;
1120						if (sin->sin_addr.s_addr ==
1121						    intf_addr->sin_addr.s_addr) {
1122							match = 1;
1123							break;
1124						}
1125					}
1126#endif
1127#ifdef INET6
1128					if (from->sa_family == AF_INET6) {
1129						struct sockaddr_in6 *intf_addr6;
1130						struct sockaddr_in6 *sin6;
1131
1132						sin6 = (struct sockaddr_in6 *)
1133						    to;
1134						intf_addr6 = &laddr->ifa->address.sin6;
1135
1136						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1137						    intf_addr6)) {
1138							match = 1;
1139							break;
1140						}
1141					}
1142#endif
1143				}
1144			}
1145			if (match == 0) {
1146				/* This endpoint does not have this address */
1147				SCTP_INP_RUNLOCK(inp);
1148				continue;
1149			}
1150		}
1151		/*
1152		 * Ok if we hit here the ep has the address, does it hold
1153		 * the tcb?
1154		 */
1155		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1156		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1157		if (stcb == NULL) {
1158			SCTP_INP_RUNLOCK(inp);
1159			continue;
1160		}
1161		SCTP_TCB_LOCK(stcb);
1162		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1163			SCTP_TCB_UNLOCK(stcb);
1164			SCTP_INP_RUNLOCK(inp);
1165			continue;
1166		}
1167		if (stcb->rport != rport) {
1168			/* remote port does not match. */
1169			SCTP_TCB_UNLOCK(stcb);
1170			SCTP_INP_RUNLOCK(inp);
1171			continue;
1172		}
1173		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1174			SCTP_TCB_UNLOCK(stcb);
1175			SCTP_INP_RUNLOCK(inp);
1176			continue;
1177		}
1178		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1179			SCTP_TCB_UNLOCK(stcb);
1180			SCTP_INP_RUNLOCK(inp);
1181			continue;
1182		}
1183		/* Does this TCB have a matching address? */
1184		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1185			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1186				/* not the same family, can't be a match */
1187				continue;
1188			}
1189			switch (from->sa_family) {
1190#ifdef INET
1191			case AF_INET:
1192				{
1193					struct sockaddr_in *sin, *rsin;
1194
1195					sin = (struct sockaddr_in *)&net->ro._l_addr;
1196					rsin = (struct sockaddr_in *)from;
1197					if (sin->sin_addr.s_addr ==
1198					    rsin->sin_addr.s_addr) {
1199						/* found it */
1200						if (netp != NULL) {
1201							*netp = net;
1202						}
1203						/*
1204						 * Update the endpoint
1205						 * pointer
1206						 */
1207						*inp_p = inp;
1208						SCTP_INP_RUNLOCK(inp);
1209						return (stcb);
1210					}
1211					break;
1212				}
1213#endif
1214#ifdef INET6
1215			case AF_INET6:
1216				{
1217					struct sockaddr_in6 *sin6, *rsin6;
1218
1219					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1220					rsin6 = (struct sockaddr_in6 *)from;
1221					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1222					    rsin6)) {
1223						/* found it */
1224						if (netp != NULL) {
1225							*netp = net;
1226						}
1227						/*
1228						 * Update the endpoint
1229						 * pointer
1230						 */
1231						*inp_p = inp;
1232						SCTP_INP_RUNLOCK(inp);
1233						return (stcb);
1234					}
1235					break;
1236				}
1237#endif
1238			default:
1239				/* TSNH */
1240				break;
1241			}
1242		}
1243		SCTP_TCB_UNLOCK(stcb);
1244		SCTP_INP_RUNLOCK(inp);
1245	}
1246	return (NULL);
1247}
1248
1249/*
1250 * rules for use
1251 *
1252 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1253 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1254 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1255 * NULL.
1256 */
1257
1258struct sctp_tcb *
1259sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1260    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1261{
1262	struct sctpasochead *head;
1263	struct sctp_inpcb *inp;
1264	struct sctp_tcb *stcb = NULL;
1265	struct sctp_nets *net;
1266	uint16_t rport;
1267
1268	inp = *inp_p;
1269	switch (remote->sa_family) {
1270#ifdef INET
1271	case AF_INET:
1272		rport = (((struct sockaddr_in *)remote)->sin_port);
1273		break;
1274#endif
1275#ifdef INET6
1276	case AF_INET6:
1277		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1278		break;
1279#endif
1280	default:
1281		return (NULL);
1282	}
1283	if (locked_tcb) {
1284		/*
1285		 * UN-lock so we can do proper locking here this occurs when
1286		 * called from load_addresses_from_init.
1287		 */
1288		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1289		SCTP_TCB_UNLOCK(locked_tcb);
1290	}
1291	SCTP_INP_INFO_RLOCK();
1292	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1293	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1294		/*-
1295		 * Now either this guy is our listener or it's the
1296		 * connector. If it is the one that issued the connect, then
1297		 * it's only chance is to be the first TCB in the list. If
1298		 * it is the acceptor, then do the special_lookup to hash
1299		 * and find the real inp.
1300		 */
1301		if ((inp->sctp_socket) && SCTP_IS_LISTENING(inp)) {
1302			/* to is peer addr, from is my addr */
1303			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1304			    netp, inp->def_vrf_id);
1305			if ((stcb != NULL) && (locked_tcb == NULL)) {
1306				/* we have a locked tcb, lower refcount */
1307				SCTP_INP_DECR_REF(inp);
1308			}
1309			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1310				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1311				SCTP_TCB_LOCK(locked_tcb);
1312				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1313				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1314			}
1315			SCTP_INP_INFO_RUNLOCK();
1316			return (stcb);
1317		} else {
1318			SCTP_INP_WLOCK(inp);
1319			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1320				goto null_return;
1321			}
1322			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1323			if (stcb == NULL) {
1324				goto null_return;
1325			}
1326			SCTP_TCB_LOCK(stcb);
1327
1328			if (stcb->rport != rport) {
1329				/* remote port does not match. */
1330				SCTP_TCB_UNLOCK(stcb);
1331				goto null_return;
1332			}
1333			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1334				SCTP_TCB_UNLOCK(stcb);
1335				goto null_return;
1336			}
1337			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1338				SCTP_TCB_UNLOCK(stcb);
1339				goto null_return;
1340			}
1341			/* now look at the list of remote addresses */
1342			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1343#ifdef INVARIANTS
1344				if (net == (TAILQ_NEXT(net, sctp_next))) {
1345					panic("Corrupt net list");
1346				}
1347#endif
1348				if (net->ro._l_addr.sa.sa_family !=
1349				    remote->sa_family) {
1350					/* not the same family */
1351					continue;
1352				}
1353				switch (remote->sa_family) {
1354#ifdef INET
1355				case AF_INET:
1356					{
1357						struct sockaddr_in *sin,
1358						           *rsin;
1359
1360						sin = (struct sockaddr_in *)
1361						    &net->ro._l_addr;
1362						rsin = (struct sockaddr_in *)remote;
1363						if (sin->sin_addr.s_addr ==
1364						    rsin->sin_addr.s_addr) {
1365							/* found it */
1366							if (netp != NULL) {
1367								*netp = net;
1368							}
1369							if (locked_tcb == NULL) {
1370								SCTP_INP_DECR_REF(inp);
1371							} else if (locked_tcb != stcb) {
1372								SCTP_TCB_LOCK(locked_tcb);
1373							}
1374							if (locked_tcb) {
1375								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1376							}
1377
1378							SCTP_INP_WUNLOCK(inp);
1379							SCTP_INP_INFO_RUNLOCK();
1380							return (stcb);
1381						}
1382						break;
1383					}
1384#endif
1385#ifdef INET6
1386				case AF_INET6:
1387					{
1388						struct sockaddr_in6 *sin6,
1389						            *rsin6;
1390
1391						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1392						rsin6 = (struct sockaddr_in6 *)remote;
1393						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1394						    rsin6)) {
1395							/* found it */
1396							if (netp != NULL) {
1397								*netp = net;
1398							}
1399							if (locked_tcb == NULL) {
1400								SCTP_INP_DECR_REF(inp);
1401							} else if (locked_tcb != stcb) {
1402								SCTP_TCB_LOCK(locked_tcb);
1403							}
1404							if (locked_tcb) {
1405								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1406							}
1407							SCTP_INP_WUNLOCK(inp);
1408							SCTP_INP_INFO_RUNLOCK();
1409							return (stcb);
1410						}
1411						break;
1412					}
1413#endif
1414				default:
1415					/* TSNH */
1416					break;
1417				}
1418			}
1419			SCTP_TCB_UNLOCK(stcb);
1420		}
1421	} else {
1422		SCTP_INP_WLOCK(inp);
1423		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1424			goto null_return;
1425		}
1426		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1427		    inp->sctp_hashmark)];
1428		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1429			if (stcb->rport != rport) {
1430				/* remote port does not match */
1431				continue;
1432			}
1433			SCTP_TCB_LOCK(stcb);
1434			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1435				SCTP_TCB_UNLOCK(stcb);
1436				continue;
1437			}
1438			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1439				SCTP_TCB_UNLOCK(stcb);
1440				continue;
1441			}
1442			/* now look at the list of remote addresses */
1443			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1444#ifdef INVARIANTS
1445				if (net == (TAILQ_NEXT(net, sctp_next))) {
1446					panic("Corrupt net list");
1447				}
1448#endif
1449				if (net->ro._l_addr.sa.sa_family !=
1450				    remote->sa_family) {
1451					/* not the same family */
1452					continue;
1453				}
1454				switch (remote->sa_family) {
1455#ifdef INET
1456				case AF_INET:
1457					{
1458						struct sockaddr_in *sin,
1459						           *rsin;
1460
1461						sin = (struct sockaddr_in *)
1462						    &net->ro._l_addr;
1463						rsin = (struct sockaddr_in *)remote;
1464						if (sin->sin_addr.s_addr ==
1465						    rsin->sin_addr.s_addr) {
1466							/* found it */
1467							if (netp != NULL) {
1468								*netp = net;
1469							}
1470							if (locked_tcb == NULL) {
1471								SCTP_INP_DECR_REF(inp);
1472							} else if (locked_tcb != stcb) {
1473								SCTP_TCB_LOCK(locked_tcb);
1474							}
1475							if (locked_tcb) {
1476								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1477							}
1478							SCTP_INP_WUNLOCK(inp);
1479							SCTP_INP_INFO_RUNLOCK();
1480							return (stcb);
1481						}
1482						break;
1483					}
1484#endif
1485#ifdef INET6
1486				case AF_INET6:
1487					{
1488						struct sockaddr_in6 *sin6,
1489						            *rsin6;
1490
1491						sin6 = (struct sockaddr_in6 *)
1492						    &net->ro._l_addr;
1493						rsin6 = (struct sockaddr_in6 *)remote;
1494						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1495						    rsin6)) {
1496							/* found it */
1497							if (netp != NULL) {
1498								*netp = net;
1499							}
1500							if (locked_tcb == NULL) {
1501								SCTP_INP_DECR_REF(inp);
1502							} else if (locked_tcb != stcb) {
1503								SCTP_TCB_LOCK(locked_tcb);
1504							}
1505							if (locked_tcb) {
1506								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1507							}
1508							SCTP_INP_WUNLOCK(inp);
1509							SCTP_INP_INFO_RUNLOCK();
1510							return (stcb);
1511						}
1512						break;
1513					}
1514#endif
1515				default:
1516					/* TSNH */
1517					break;
1518				}
1519			}
1520			SCTP_TCB_UNLOCK(stcb);
1521		}
1522	}
1523null_return:
1524	/* clean up for returning null */
1525	if (locked_tcb) {
1526		SCTP_TCB_LOCK(locked_tcb);
1527		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1528	}
1529	SCTP_INP_WUNLOCK(inp);
1530	SCTP_INP_INFO_RUNLOCK();
1531	/* not found */
1532	return (NULL);
1533}
1534
1535/*
1536 * Find an association for a specific endpoint using the association id given
1537 * out in the COMM_UP notification
1538 */
1539struct sctp_tcb *
1540sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1541{
1542	/*
1543	 * Use my the assoc_id to find a endpoint
1544	 */
1545	struct sctpasochead *head;
1546	struct sctp_tcb *stcb;
1547	uint32_t id;
1548
1549	if (inp == NULL) {
1550		SCTP_PRINTF("TSNH ep_associd\n");
1551		return (NULL);
1552	}
1553	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1554		SCTP_PRINTF("TSNH ep_associd0\n");
1555		return (NULL);
1556	}
1557	id = (uint32_t)asoc_id;
1558	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1559	if (head == NULL) {
1560		/* invalid id TSNH */
1561		SCTP_PRINTF("TSNH ep_associd1\n");
1562		return (NULL);
1563	}
1564	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1565		if (stcb->asoc.assoc_id == id) {
1566			if (inp != stcb->sctp_ep) {
1567				/*
1568				 * some other guy has the same id active (id
1569				 * collision ??).
1570				 */
1571				SCTP_PRINTF("TSNH ep_associd2\n");
1572				continue;
1573			}
1574			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1575				continue;
1576			}
1577			if (want_lock) {
1578				SCTP_TCB_LOCK(stcb);
1579			}
1580			return (stcb);
1581		}
1582	}
1583	return (NULL);
1584}
1585
1586struct sctp_tcb *
1587sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1588{
1589	struct sctp_tcb *stcb;
1590
1591	SCTP_INP_RLOCK(inp);
1592	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1593	SCTP_INP_RUNLOCK(inp);
1594	return (stcb);
1595}
1596
1597/*
1598 * Endpoint probe expects that the INP_INFO is locked.
1599 */
1600static struct sctp_inpcb *
1601sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1602    uint16_t lport, uint32_t vrf_id)
1603{
1604	struct sctp_inpcb *inp;
1605	struct sctp_laddr *laddr;
1606#ifdef INET
1607	struct sockaddr_in *sin;
1608#endif
1609#ifdef INET6
1610	struct sockaddr_in6 *sin6;
1611	struct sockaddr_in6 *intf_addr6;
1612#endif
1613	int fnd;
1614
1615#ifdef INET
1616	sin = NULL;
1617#endif
1618#ifdef INET6
1619	sin6 = NULL;
1620#endif
1621	switch (nam->sa_family) {
1622#ifdef INET
1623	case AF_INET:
1624		sin = (struct sockaddr_in *)nam;
1625		break;
1626#endif
1627#ifdef INET6
1628	case AF_INET6:
1629		sin6 = (struct sockaddr_in6 *)nam;
1630		break;
1631#endif
1632	default:
1633		/* unsupported family */
1634		return (NULL);
1635	}
1636
1637	if (head == NULL)
1638		return (NULL);
1639
1640	LIST_FOREACH(inp, head, sctp_hash) {
1641		SCTP_INP_RLOCK(inp);
1642		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1643			SCTP_INP_RUNLOCK(inp);
1644			continue;
1645		}
1646		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1647		    (inp->sctp_lport == lport)) {
1648			/* got it */
1649			switch (nam->sa_family) {
1650#ifdef INET
1651			case AF_INET:
1652				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1653				    SCTP_IPV6_V6ONLY(inp)) {
1654					/*
1655					 * IPv4 on a IPv6 socket with ONLY
1656					 * IPv6 set
1657					 */
1658					SCTP_INP_RUNLOCK(inp);
1659					continue;
1660				}
1661				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1662				    &sin->sin_addr) != 0) {
1663					SCTP_INP_RUNLOCK(inp);
1664					continue;
1665				}
1666				break;
1667#endif
1668#ifdef INET6
1669			case AF_INET6:
1670				/*
1671				 * A V6 address and the endpoint is NOT
1672				 * bound V6
1673				 */
1674				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1675					SCTP_INP_RUNLOCK(inp);
1676					continue;
1677				}
1678				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1679				    &sin6->sin6_addr) != 0) {
1680					SCTP_INP_RUNLOCK(inp);
1681					continue;
1682				}
1683				break;
1684#endif
1685			default:
1686				break;
1687			}
1688			/* does a VRF id match? */
1689			fnd = 0;
1690			if (inp->def_vrf_id == vrf_id)
1691				fnd = 1;
1692
1693			SCTP_INP_RUNLOCK(inp);
1694			if (!fnd)
1695				continue;
1696			return (inp);
1697		}
1698		SCTP_INP_RUNLOCK(inp);
1699	}
1700	switch (nam->sa_family) {
1701#ifdef INET
1702	case AF_INET:
1703		if (sin->sin_addr.s_addr == INADDR_ANY) {
1704			/* Can't hunt for one that has no address specified */
1705			return (NULL);
1706		}
1707		break;
1708#endif
1709#ifdef INET6
1710	case AF_INET6:
1711		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1712			/* Can't hunt for one that has no address specified */
1713			return (NULL);
1714		}
1715		break;
1716#endif
1717	default:
1718		break;
1719	}
1720	/*
1721	 * ok, not bound to all so see if we can find a EP bound to this
1722	 * address.
1723	 */
1724	LIST_FOREACH(inp, head, sctp_hash) {
1725		SCTP_INP_RLOCK(inp);
1726		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1727			SCTP_INP_RUNLOCK(inp);
1728			continue;
1729		}
1730		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1731			SCTP_INP_RUNLOCK(inp);
1732			continue;
1733		}
1734		/*
1735		 * Ok this could be a likely candidate, look at all of its
1736		 * addresses
1737		 */
1738		if (inp->sctp_lport != lport) {
1739			SCTP_INP_RUNLOCK(inp);
1740			continue;
1741		}
1742		/* does a VRF id match? */
1743		fnd = 0;
1744		if (inp->def_vrf_id == vrf_id)
1745			fnd = 1;
1746
1747		if (!fnd) {
1748			SCTP_INP_RUNLOCK(inp);
1749			continue;
1750		}
1751		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1752			if (laddr->ifa == NULL) {
1753				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1754				    __func__);
1755				continue;
1756			}
1757			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1758			    (void *)laddr->ifa);
1759			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1760				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1761				continue;
1762			}
1763			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1764				/* possible, see if it matches */
1765				switch (nam->sa_family) {
1766#ifdef INET
1767				case AF_INET:
1768					if (sin->sin_addr.s_addr ==
1769					    laddr->ifa->address.sin.sin_addr.s_addr) {
1770						SCTP_INP_RUNLOCK(inp);
1771						return (inp);
1772					}
1773					break;
1774#endif
1775#ifdef INET6
1776				case AF_INET6:
1777					intf_addr6 = &laddr->ifa->address.sin6;
1778					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1779					    intf_addr6)) {
1780						SCTP_INP_RUNLOCK(inp);
1781						return (inp);
1782					}
1783					break;
1784#endif
1785				}
1786			}
1787		}
1788		SCTP_INP_RUNLOCK(inp);
1789	}
1790	return (NULL);
1791}
1792
1793static struct sctp_inpcb *
1794sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1795{
1796	struct sctppcbhead *head;
1797	struct sctp_inpcb *t_inp;
1798	int fnd;
1799
1800	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1801	    SCTP_BASE_INFO(hashmark))];
1802	LIST_FOREACH(t_inp, head, sctp_hash) {
1803		if (t_inp->sctp_lport != lport) {
1804			continue;
1805		}
1806		/* is it in the VRF in question */
1807		fnd = 0;
1808		if (t_inp->def_vrf_id == vrf_id)
1809			fnd = 1;
1810		if (!fnd)
1811			continue;
1812
1813		/* This one is in use. */
1814		/* check the v6/v4 binding issue */
1815		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1816		    SCTP_IPV6_V6ONLY(t_inp)) {
1817			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1818				/* collision in V6 space */
1819				return (t_inp);
1820			} else {
1821				/* inp is BOUND_V4 no conflict */
1822				continue;
1823			}
1824		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1825			/* t_inp is bound v4 and v6, conflict always */
1826			return (t_inp);
1827		} else {
1828			/* t_inp is bound only V4 */
1829			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1830			    SCTP_IPV6_V6ONLY(inp)) {
1831				/* no conflict */
1832				continue;
1833			}
1834			/* else fall through to conflict */
1835		}
1836		return (t_inp);
1837	}
1838	return (NULL);
1839}
1840
1841int
1842sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1843{
1844	/* For 1-2-1 with port reuse */
1845	struct sctppcbhead *head;
1846	struct sctp_inpcb *tinp, *ninp;
1847
1848	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1849		/* only works with port reuse on */
1850		return (-1);
1851	}
1852	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1853		return (0);
1854	}
1855	SCTP_INP_RUNLOCK(inp);
1856	SCTP_INP_INFO_WLOCK();
1857	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1858	    SCTP_BASE_INFO(hashmark))];
1859	/* Kick out all non-listeners to the TCP hash */
1860	LIST_FOREACH_SAFE(tinp, head, sctp_hash, ninp) {
1861		if (tinp->sctp_lport != inp->sctp_lport) {
1862			continue;
1863		}
1864		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1865			continue;
1866		}
1867		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1868			continue;
1869		}
1870		if (SCTP_IS_LISTENING(tinp)) {
1871			continue;
1872		}
1873		SCTP_INP_WLOCK(tinp);
1874		LIST_REMOVE(tinp, sctp_hash);
1875		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1876		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1877		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1878		SCTP_INP_WUNLOCK(tinp);
1879	}
1880	SCTP_INP_WLOCK(inp);
1881	/* Pull from where he was */
1882	LIST_REMOVE(inp, sctp_hash);
1883	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1884	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1885	LIST_INSERT_HEAD(head, inp, sctp_hash);
1886	SCTP_INP_WUNLOCK(inp);
1887	SCTP_INP_RLOCK(inp);
1888	SCTP_INP_INFO_WUNLOCK();
1889	return (0);
1890}
1891
1892struct sctp_inpcb *
1893sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1894    uint32_t vrf_id)
1895{
1896	/*
1897	 * First we check the hash table to see if someone has this port
1898	 * bound with just the port.
1899	 */
1900	struct sctp_inpcb *inp;
1901	struct sctppcbhead *head;
1902	int lport;
1903	unsigned int i;
1904#ifdef INET
1905	struct sockaddr_in *sin;
1906#endif
1907#ifdef INET6
1908	struct sockaddr_in6 *sin6;
1909#endif
1910
1911	switch (nam->sa_family) {
1912#ifdef INET
1913	case AF_INET:
1914		sin = (struct sockaddr_in *)nam;
1915		lport = sin->sin_port;
1916		break;
1917#endif
1918#ifdef INET6
1919	case AF_INET6:
1920		sin6 = (struct sockaddr_in6 *)nam;
1921		lport = sin6->sin6_port;
1922		break;
1923#endif
1924	default:
1925		return (NULL);
1926	}
1927	/*
1928	 * I could cheat here and just cast to one of the types but we will
1929	 * do it right. It also provides the check against an Unsupported
1930	 * type too.
1931	 */
1932	/* Find the head of the ALLADDR chain */
1933	if (have_lock == 0) {
1934		SCTP_INP_INFO_RLOCK();
1935	}
1936	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1937	    SCTP_BASE_INFO(hashmark))];
1938	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1939
1940	/*
1941	 * If the TCP model exists it could be that the main listening
1942	 * endpoint is gone but there still exists a connected socket for
1943	 * this guy. If so we can return the first one that we find. This
1944	 * may NOT be the correct one so the caller should be wary on the
1945	 * returned INP. Currently the only caller that sets find_tcp_pool
1946	 * is in bindx where we are verifying that a user CAN bind the
1947	 * address. He either has bound it already, or someone else has, or
1948	 * its open to bind, so this is good enough.
1949	 */
1950	if (inp == NULL && find_tcp_pool) {
1951		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1952			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1953			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1954			if (inp) {
1955				break;
1956			}
1957		}
1958	}
1959	if (inp) {
1960		SCTP_INP_INCR_REF(inp);
1961	}
1962	if (have_lock == 0) {
1963		SCTP_INP_INFO_RUNLOCK();
1964	}
1965	return (inp);
1966}
1967
1968/*
1969 * Find an association for an endpoint with the pointer to whom you want to
1970 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1971 * need to change the *to to some other struct like a mbuf...
1972 */
1973struct sctp_tcb *
1974sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
1975    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1976    uint32_t vrf_id)
1977{
1978	struct sctp_inpcb *inp = NULL;
1979	struct sctp_tcb *stcb;
1980
1981	SCTP_INP_INFO_RLOCK();
1982	if (find_tcp_pool) {
1983		if (inp_p != NULL) {
1984			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
1985			    vrf_id);
1986		} else {
1987			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
1988			    vrf_id);
1989		}
1990		if (stcb != NULL) {
1991			SCTP_INP_INFO_RUNLOCK();
1992			return (stcb);
1993		}
1994	}
1995	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1996	if (inp_p != NULL) {
1997		*inp_p = inp;
1998	}
1999	SCTP_INP_INFO_RUNLOCK();
2000	if (inp == NULL) {
2001		return (NULL);
2002	}
2003	/*
2004	 * ok, we have an endpoint, now lets find the assoc for it (if any)
2005	 * we now place the source address or from in the to of the find
2006	 * endpoint call. Since in reality this chain is used from the
2007	 * inbound packet side.
2008	 */
2009	if (inp_p != NULL) {
2010		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
2011		    NULL);
2012	} else {
2013		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
2014		    NULL);
2015	}
2016	return (stcb);
2017}
2018
2019/*
2020 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
2021 * find all addresses that the sender has specified in any address list. Each
2022 * address will be used to lookup the TCB and see if one exits.
2023 */
2024static struct sctp_tcb *
2025sctp_findassociation_special_addr(struct mbuf *m, int offset,
2026    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
2027    struct sockaddr *dst)
2028{
2029	struct sctp_paramhdr *phdr, param_buf;
2030#if defined(INET) || defined(INET6)
2031	struct sctp_tcb *stcb;
2032	uint16_t ptype;
2033#endif
2034	uint16_t plen;
2035#ifdef INET
2036	struct sockaddr_in sin4;
2037#endif
2038#ifdef INET6
2039	struct sockaddr_in6 sin6;
2040#endif
2041
2042#ifdef INET
2043	memset(&sin4, 0, sizeof(sin4));
2044	sin4.sin_len = sizeof(sin4);
2045	sin4.sin_family = AF_INET;
2046	sin4.sin_port = sh->src_port;
2047#endif
2048#ifdef INET6
2049	memset(&sin6, 0, sizeof(sin6));
2050	sin6.sin6_len = sizeof(sin6);
2051	sin6.sin6_family = AF_INET6;
2052	sin6.sin6_port = sh->src_port;
2053#endif
2054
2055	offset += sizeof(struct sctp_init_chunk);
2056
2057	phdr = sctp_get_next_param(m, offset, &param_buf, sizeof(param_buf));
2058	while (phdr != NULL) {
2059		/* now we must see if we want the parameter */
2060#if defined(INET) || defined(INET6)
2061		ptype = ntohs(phdr->param_type);
2062#endif
2063		plen = ntohs(phdr->param_length);
2064		if (plen == 0) {
2065			break;
2066		}
2067#ifdef INET
2068		if (ptype == SCTP_IPV4_ADDRESS &&
2069		    plen == sizeof(struct sctp_ipv4addr_param)) {
2070			/* Get the rest of the address */
2071			struct sctp_ipv4addr_param ip4_param, *p4;
2072
2073			phdr = sctp_get_next_param(m, offset,
2074			    (struct sctp_paramhdr *)&ip4_param, sizeof(ip4_param));
2075			if (phdr == NULL) {
2076				return (NULL);
2077			}
2078			p4 = (struct sctp_ipv4addr_param *)phdr;
2079			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2080			/* look it up */
2081			stcb = sctp_findassociation_ep_addr(inp_p,
2082			    (struct sockaddr *)&sin4, netp, dst, NULL);
2083			if (stcb != NULL) {
2084				return (stcb);
2085			}
2086		}
2087#endif
2088#ifdef INET6
2089		if (ptype == SCTP_IPV6_ADDRESS &&
2090		    plen == sizeof(struct sctp_ipv6addr_param)) {
2091			/* Get the rest of the address */
2092			struct sctp_ipv6addr_param ip6_param, *p6;
2093
2094			phdr = sctp_get_next_param(m, offset,
2095			    (struct sctp_paramhdr *)&ip6_param, sizeof(ip6_param));
2096			if (phdr == NULL) {
2097				return (NULL);
2098			}
2099			p6 = (struct sctp_ipv6addr_param *)phdr;
2100			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2101			/* look it up */
2102			stcb = sctp_findassociation_ep_addr(inp_p,
2103			    (struct sockaddr *)&sin6, netp, dst, NULL);
2104			if (stcb != NULL) {
2105				return (stcb);
2106			}
2107		}
2108#endif
2109		offset += SCTP_SIZE32(plen);
2110		phdr = sctp_get_next_param(m, offset, &param_buf,
2111		    sizeof(param_buf));
2112	}
2113	return (NULL);
2114}
2115
2116static struct sctp_tcb *
2117sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2118    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2119    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2120{
2121	/*
2122	 * Use my vtag to hash. If we find it we then verify the source addr
2123	 * is in the assoc. If all goes well we save a bit on rec of a
2124	 * packet.
2125	 */
2126	struct sctpasochead *head;
2127	struct sctp_nets *net;
2128	struct sctp_tcb *stcb;
2129
2130	SCTP_INP_INFO_RLOCK();
2131	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2132	    SCTP_BASE_INFO(hashasocmark))];
2133	LIST_FOREACH(stcb, head, sctp_asocs) {
2134		SCTP_INP_RLOCK(stcb->sctp_ep);
2135		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2136			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2137			continue;
2138		}
2139		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2140			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2141			continue;
2142		}
2143		SCTP_TCB_LOCK(stcb);
2144		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2145		if (stcb->asoc.my_vtag == vtag) {
2146			/* candidate */
2147			if (stcb->rport != rport) {
2148				SCTP_TCB_UNLOCK(stcb);
2149				continue;
2150			}
2151			if (stcb->sctp_ep->sctp_lport != lport) {
2152				SCTP_TCB_UNLOCK(stcb);
2153				continue;
2154			}
2155			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2156				SCTP_TCB_UNLOCK(stcb);
2157				continue;
2158			}
2159			/* RRS:Need toaddr check here */
2160			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2161				/* Endpoint does not own this address */
2162				SCTP_TCB_UNLOCK(stcb);
2163				continue;
2164			}
2165			if (remote_tag) {
2166				/*
2167				 * If we have both vtags that's all we match
2168				 * on
2169				 */
2170				if (stcb->asoc.peer_vtag == remote_tag) {
2171					/*
2172					 * If both tags match we consider it
2173					 * conclusive and check NO
2174					 * source/destination addresses
2175					 */
2176					goto conclusive;
2177				}
2178			}
2179			if (skip_src_check) {
2180		conclusive:
2181				if (from) {
2182					*netp = sctp_findnet(stcb, from);
2183				} else {
2184					*netp = NULL;	/* unknown */
2185				}
2186				if (inp_p)
2187					*inp_p = stcb->sctp_ep;
2188				SCTP_INP_INFO_RUNLOCK();
2189				return (stcb);
2190			}
2191			net = sctp_findnet(stcb, from);
2192			if (net) {
2193				/* yep its him. */
2194				*netp = net;
2195				SCTP_STAT_INCR(sctps_vtagexpress);
2196				*inp_p = stcb->sctp_ep;
2197				SCTP_INP_INFO_RUNLOCK();
2198				return (stcb);
2199			} else {
2200				/*
2201				 * not him, this should only happen in rare
2202				 * cases so I peg it.
2203				 */
2204				SCTP_STAT_INCR(sctps_vtagbogus);
2205			}
2206		}
2207		SCTP_TCB_UNLOCK(stcb);
2208	}
2209	SCTP_INP_INFO_RUNLOCK();
2210	return (NULL);
2211}
2212
2213/*
2214 * Find an association with the pointer to the inbound IP packet. This can be
2215 * a IPv4 or IPv6 packet.
2216 */
2217struct sctp_tcb *
2218sctp_findassociation_addr(struct mbuf *m, int offset,
2219    struct sockaddr *src, struct sockaddr *dst,
2220    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2221    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2222{
2223	struct sctp_tcb *stcb;
2224	struct sctp_inpcb *inp;
2225
2226	if (sh->v_tag) {
2227		/* we only go down this path if vtag is non-zero */
2228		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2229		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2230		if (stcb) {
2231			return (stcb);
2232		}
2233	}
2234
2235	if (inp_p) {
2236		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2237		    1, vrf_id);
2238		inp = *inp_p;
2239	} else {
2240		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2241		    1, vrf_id);
2242	}
2243	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2244	if (stcb == NULL && inp) {
2245		/* Found a EP but not this address */
2246		if ((ch->chunk_type == SCTP_INITIATION) ||
2247		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2248			/*-
2249			 * special hook, we do NOT return linp or an
2250			 * association that is linked to an existing
2251			 * association that is under the TCP pool (i.e. no
2252			 * listener exists). The endpoint finding routine
2253			 * will always find a listener before examining the
2254			 * TCP pool.
2255			 */
2256			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2257				if (inp_p) {
2258					*inp_p = NULL;
2259				}
2260				return (NULL);
2261			}
2262			stcb = sctp_findassociation_special_addr(m,
2263			    offset, sh, &inp, netp, dst);
2264			if (inp_p != NULL) {
2265				*inp_p = inp;
2266			}
2267		}
2268	}
2269	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2270	return (stcb);
2271}
2272
2273/*
2274 * lookup an association by an ASCONF lookup address.
2275 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2276 */
2277struct sctp_tcb *
2278sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2279    struct sockaddr *dst, struct sctphdr *sh,
2280    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2281{
2282	struct sctp_tcb *stcb;
2283	union sctp_sockstore remote_store;
2284	struct sctp_paramhdr param_buf, *phdr;
2285	int ptype;
2286	int zero_address = 0;
2287#ifdef INET
2288	struct sockaddr_in *sin;
2289#endif
2290#ifdef INET6
2291	struct sockaddr_in6 *sin6;
2292#endif
2293
2294	memset(&remote_store, 0, sizeof(remote_store));
2295	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2296	    &param_buf, sizeof(struct sctp_paramhdr));
2297	if (phdr == NULL) {
2298		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2299		    __func__);
2300		return NULL;
2301	}
2302	ptype = (int)((uint32_t)ntohs(phdr->param_type));
2303	/* get the correlation address */
2304	switch (ptype) {
2305#ifdef INET6
2306	case SCTP_IPV6_ADDRESS:
2307		{
2308			/* ipv6 address param */
2309			struct sctp_ipv6addr_param *p6, p6_buf;
2310
2311			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2312				return NULL;
2313			}
2314			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2315			    offset + sizeof(struct sctp_asconf_chunk),
2316			    &p6_buf.ph, sizeof(p6_buf));
2317			if (p6 == NULL) {
2318				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2319				    __func__);
2320				return (NULL);
2321			}
2322			sin6 = &remote_store.sin6;
2323			sin6->sin6_family = AF_INET6;
2324			sin6->sin6_len = sizeof(*sin6);
2325			sin6->sin6_port = sh->src_port;
2326			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2327			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2328				zero_address = 1;
2329			break;
2330		}
2331#endif
2332#ifdef INET
2333	case SCTP_IPV4_ADDRESS:
2334		{
2335			/* ipv4 address param */
2336			struct sctp_ipv4addr_param *p4, p4_buf;
2337
2338			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2339				return NULL;
2340			}
2341			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2342			    offset + sizeof(struct sctp_asconf_chunk),
2343			    &p4_buf.ph, sizeof(p4_buf));
2344			if (p4 == NULL) {
2345				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2346				    __func__);
2347				return (NULL);
2348			}
2349			sin = &remote_store.sin;
2350			sin->sin_family = AF_INET;
2351			sin->sin_len = sizeof(*sin);
2352			sin->sin_port = sh->src_port;
2353			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2354			if (sin->sin_addr.s_addr == INADDR_ANY)
2355				zero_address = 1;
2356			break;
2357		}
2358#endif
2359	default:
2360		/* invalid address param type */
2361		return NULL;
2362	}
2363
2364	if (zero_address) {
2365		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2366		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2367		if (stcb != NULL) {
2368			SCTP_INP_DECR_REF(*inp_p);
2369		}
2370	} else {
2371		stcb = sctp_findassociation_ep_addr(inp_p,
2372		    &remote_store.sa, netp,
2373		    dst, NULL);
2374	}
2375	return (stcb);
2376}
2377
2378/*
2379 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2380 * addresses. This way if we don't get a bind we by default pick a ephemeral
2381 * port with all addresses bound.
2382 */
2383int
2384sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2385{
2386	/*
2387	 * we get called when a new endpoint starts up. We need to allocate
2388	 * the sctp_inpcb structure from the zone and init it. Mark it as
2389	 * unbound and find a port that we can use as an ephemeral with
2390	 * INADDR_ANY. If the user binds later no problem we can then add in
2391	 * the specific addresses. And setup the default parameters for the
2392	 * EP.
2393	 */
2394	int i, error;
2395	struct sctp_inpcb *inp;
2396	struct sctp_pcb *m;
2397	struct timeval time;
2398	sctp_sharedkey_t *null_key;
2399
2400	error = 0;
2401
2402	SCTP_INP_INFO_WLOCK();
2403	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2404	if (inp == NULL) {
2405		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2406		SCTP_INP_INFO_WUNLOCK();
2407		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2408		return (ENOBUFS);
2409	}
2410	/* zap it */
2411	memset(inp, 0, sizeof(*inp));
2412
2413	/* bump generations */
2414	/* setup socket pointers */
2415	inp->sctp_socket = so;
2416	inp->ip_inp.inp.inp_socket = so;
2417	inp->ip_inp.inp.inp_cred = crhold(so->so_cred);
2418#ifdef INET6
2419	if (INP_SOCKAF(so) == AF_INET6) {
2420		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2421			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2422		}
2423		if (MODULE_GLOBAL(ip6_v6only)) {
2424			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2425		}
2426	}
2427#endif
2428	inp->sctp_associd_counter = 1;
2429	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2430	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2431	inp->max_cwnd = 0;
2432	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2433	inp->ecn_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_ecn_enable);
2434	inp->prsctp_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_pr_enable);
2435	inp->auth_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_auth_enable);
2436	inp->asconf_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_asconf_enable);
2437	inp->reconfig_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_reconfig_enable);
2438	inp->nrsack_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_nrsack_enable);
2439	inp->pktdrop_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_pktdrop_enable);
2440	inp->idata_supported = 0;
2441
2442	inp->fibnum = so->so_fibnum;
2443	/* init the small hash table we use to track asocid <-> tcb */
2444	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2445	if (inp->sctp_asocidhash == NULL) {
2446		crfree(inp->ip_inp.inp.inp_cred);
2447		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2448		SCTP_INP_INFO_WUNLOCK();
2449		return (ENOBUFS);
2450	}
2451	SCTP_INCR_EP_COUNT();
2452	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2453	SCTP_INP_INFO_WUNLOCK();
2454
2455	so->so_pcb = (caddr_t)inp;
2456
2457	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2458		/* UDP style socket */
2459		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2460		    SCTP_PCB_FLAGS_UNBOUND);
2461		/* Be sure it is NON-BLOCKING IO for UDP */
2462		/* SCTP_SET_SO_NBIO(so); */
2463	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2464		/* TCP style socket */
2465		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2466		    SCTP_PCB_FLAGS_UNBOUND);
2467		/* Be sure we have blocking IO by default */
2468		SOCK_LOCK(so);
2469		SCTP_CLEAR_SO_NBIO(so);
2470		SOCK_UNLOCK(so);
2471	} else {
2472		/*
2473		 * unsupported socket type (RAW, etc)- in case we missed it
2474		 * in protosw
2475		 */
2476		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2477		so->so_pcb = NULL;
2478		crfree(inp->ip_inp.inp.inp_cred);
2479		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2480		return (EOPNOTSUPP);
2481	}
2482	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2483		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2484		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2485	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2486		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2487		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2488	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2489		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2490		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2491	}
2492	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2493	    &inp->sctp_hashmark);
2494	if (inp->sctp_tcbhash == NULL) {
2495		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2496		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2497		so->so_pcb = NULL;
2498		crfree(inp->ip_inp.inp.inp_cred);
2499		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2500		return (ENOBUFS);
2501	}
2502	inp->def_vrf_id = vrf_id;
2503
2504	SCTP_INP_INFO_WLOCK();
2505	SCTP_INP_LOCK_INIT(inp);
2506	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2507	SCTP_INP_READ_INIT(inp);
2508	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2509	/* lock the new ep */
2510	SCTP_INP_WLOCK(inp);
2511
2512	/* add it to the info area */
2513	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2514	SCTP_INP_INFO_WUNLOCK();
2515
2516	TAILQ_INIT(&inp->read_queue);
2517	LIST_INIT(&inp->sctp_addr_list);
2518
2519	LIST_INIT(&inp->sctp_asoc_list);
2520
2521#ifdef SCTP_TRACK_FREED_ASOCS
2522	/* TEMP CODE */
2523	LIST_INIT(&inp->sctp_asoc_free_list);
2524#endif
2525	/* Init the timer structure for signature change */
2526	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2527	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2528
2529	/* now init the actual endpoint default data */
2530	m = &inp->sctp_ep;
2531
2532	/* setup the base timeout information */
2533	m->sctp_timeoutticks[SCTP_TIMER_SEND] = sctp_secs_to_ticks(SCTP_SEND_SEC);	/* needed ? */
2534	m->sctp_timeoutticks[SCTP_TIMER_INIT] = sctp_secs_to_ticks(SCTP_INIT_SEC);	/* needed ? */
2535	m->sctp_timeoutticks[SCTP_TIMER_RECV] = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2536	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2537	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2538	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2539	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2540	/* all max/min max are in ms */
2541	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2542	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2543	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2544	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2545	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2546	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2547	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2548	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2549	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2550	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2551	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2552	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2553	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2554
2555	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2556	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2557	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2558	/* number of streams to pre-open on a association */
2559	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2560
2561	m->default_mtu = 0;
2562	/* Add adaptation cookie */
2563	m->adaptation_layer_indicator = 0;
2564	m->adaptation_layer_indicator_provided = 0;
2565
2566	/* seed random number generator */
2567	m->random_counter = 1;
2568	m->store_at = SCTP_SIGNATURE_SIZE;
2569	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2570	sctp_fill_random_store(m);
2571
2572	/* Minimum cookie size */
2573	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2574	    sizeof(struct sctp_state_cookie);
2575	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2576
2577	/* Setup the initial secret */
2578	(void)SCTP_GETTIME_TIMEVAL(&time);
2579	m->time_of_secret_change = time.tv_sec;
2580
2581	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2582		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2583	}
2584	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2585
2586	/* How long is a cookie good for ? */
2587	m->def_cookie_life = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2588	/*
2589	 * Initialize authentication parameters
2590	 */
2591	m->local_hmacs = sctp_default_supported_hmaclist();
2592	m->local_auth_chunks = sctp_alloc_chunklist();
2593	if (inp->asconf_supported) {
2594		sctp_auth_add_chunk(SCTP_ASCONF, m->local_auth_chunks);
2595		sctp_auth_add_chunk(SCTP_ASCONF_ACK, m->local_auth_chunks);
2596	}
2597	m->default_dscp = 0;
2598#ifdef INET6
2599	m->default_flowlabel = 0;
2600#endif
2601	m->port = 0;		/* encapsulation disabled by default */
2602	LIST_INIT(&m->shared_keys);
2603	/* add default NULL key as key id 0 */
2604	null_key = sctp_alloc_sharedkey();
2605	sctp_insert_sharedkey(&m->shared_keys, null_key);
2606	SCTP_INP_WUNLOCK(inp);
2607#ifdef SCTP_LOG_CLOSING
2608	sctp_log_closing(inp, NULL, 12);
2609#endif
2610	return (error);
2611}
2612
2613void
2614sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2615    struct sctp_tcb *stcb)
2616{
2617	struct sctp_nets *net;
2618	uint16_t lport, rport;
2619	struct sctppcbhead *head;
2620	struct sctp_laddr *laddr, *oladdr;
2621
2622	atomic_add_int(&stcb->asoc.refcnt, 1);
2623	SCTP_TCB_UNLOCK(stcb);
2624	SCTP_INP_INFO_WLOCK();
2625	SCTP_INP_WLOCK(old_inp);
2626	SCTP_INP_WLOCK(new_inp);
2627	SCTP_TCB_LOCK(stcb);
2628	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2629
2630	new_inp->sctp_ep.time_of_secret_change =
2631	    old_inp->sctp_ep.time_of_secret_change;
2632	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2633	    sizeof(old_inp->sctp_ep.secret_key));
2634	new_inp->sctp_ep.current_secret_number =
2635	    old_inp->sctp_ep.current_secret_number;
2636	new_inp->sctp_ep.last_secret_number =
2637	    old_inp->sctp_ep.last_secret_number;
2638	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2639
2640	/* make it so new data pours into the new socket */
2641	stcb->sctp_socket = new_inp->sctp_socket;
2642	stcb->sctp_ep = new_inp;
2643
2644	/* Copy the port across */
2645	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2646	rport = stcb->rport;
2647	/* Pull the tcb from the old association */
2648	LIST_REMOVE(stcb, sctp_tcbhash);
2649	LIST_REMOVE(stcb, sctp_tcblist);
2650	if (stcb->asoc.in_asocid_hash) {
2651		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2652	}
2653	/* Now insert the new_inp into the TCP connected hash */
2654	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2655
2656	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2657	/* Its safe to access */
2658	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2659
2660	/* Now move the tcb into the endpoint list */
2661	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2662	/*
2663	 * Question, do we even need to worry about the ep-hash since we
2664	 * only have one connection? Probably not :> so lets get rid of it
2665	 * and not suck up any kernel memory in that.
2666	 */
2667	if (stcb->asoc.in_asocid_hash) {
2668		struct sctpasochead *lhd;
2669
2670		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2671		    new_inp->hashasocidmark)];
2672		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2673	}
2674	/* Ok. Let's restart timer. */
2675	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2676		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2677		    stcb, net);
2678	}
2679
2680	SCTP_INP_INFO_WUNLOCK();
2681	if (new_inp->sctp_tcbhash != NULL) {
2682		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2683		new_inp->sctp_tcbhash = NULL;
2684	}
2685	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2686		/* Subset bound, so copy in the laddr list from the old_inp */
2687		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2688			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2689			if (laddr == NULL) {
2690				/*
2691				 * Gak, what can we do? This assoc is really
2692				 * HOSED. We probably should send an abort
2693				 * here.
2694				 */
2695				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2696				continue;
2697			}
2698			SCTP_INCR_LADDR_COUNT();
2699			memset(laddr, 0, sizeof(*laddr));
2700			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2701			laddr->ifa = oladdr->ifa;
2702			atomic_add_int(&laddr->ifa->refcount, 1);
2703			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2704			    sctp_nxt_addr);
2705			new_inp->laddr_count++;
2706			if (oladdr == stcb->asoc.last_used_address) {
2707				stcb->asoc.last_used_address = laddr;
2708			}
2709		}
2710	}
2711	/* Now any running timers need to be adjusted. */
2712	if (stcb->asoc.dack_timer.ep == old_inp) {
2713		SCTP_INP_DECR_REF(old_inp);
2714		stcb->asoc.dack_timer.ep = new_inp;
2715		SCTP_INP_INCR_REF(new_inp);
2716	}
2717	if (stcb->asoc.asconf_timer.ep == old_inp) {
2718		SCTP_INP_DECR_REF(old_inp);
2719		stcb->asoc.asconf_timer.ep = new_inp;
2720		SCTP_INP_INCR_REF(new_inp);
2721	}
2722	if (stcb->asoc.strreset_timer.ep == old_inp) {
2723		SCTP_INP_DECR_REF(old_inp);
2724		stcb->asoc.strreset_timer.ep = new_inp;
2725		SCTP_INP_INCR_REF(new_inp);
2726	}
2727	if (stcb->asoc.shut_guard_timer.ep == old_inp) {
2728		SCTP_INP_DECR_REF(old_inp);
2729		stcb->asoc.shut_guard_timer.ep = new_inp;
2730		SCTP_INP_INCR_REF(new_inp);
2731	}
2732	if (stcb->asoc.autoclose_timer.ep == old_inp) {
2733		SCTP_INP_DECR_REF(old_inp);
2734		stcb->asoc.autoclose_timer.ep = new_inp;
2735		SCTP_INP_INCR_REF(new_inp);
2736	}
2737	if (stcb->asoc.delete_prim_timer.ep == old_inp) {
2738		SCTP_INP_DECR_REF(old_inp);
2739		stcb->asoc.delete_prim_timer.ep = new_inp;
2740		SCTP_INP_INCR_REF(new_inp);
2741	}
2742	/* now what about the nets? */
2743	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2744		if (net->pmtu_timer.ep == old_inp) {
2745			SCTP_INP_DECR_REF(old_inp);
2746			net->pmtu_timer.ep = new_inp;
2747			SCTP_INP_INCR_REF(new_inp);
2748		}
2749		if (net->hb_timer.ep == old_inp) {
2750			SCTP_INP_DECR_REF(old_inp);
2751			net->hb_timer.ep = new_inp;
2752			SCTP_INP_INCR_REF(new_inp);
2753		}
2754		if (net->rxt_timer.ep == old_inp) {
2755			SCTP_INP_DECR_REF(old_inp);
2756			net->rxt_timer.ep = new_inp;
2757			SCTP_INP_INCR_REF(new_inp);
2758		}
2759	}
2760	SCTP_INP_WUNLOCK(new_inp);
2761	SCTP_INP_WUNLOCK(old_inp);
2762}
2763
2764/*
2765 * insert an laddr entry with the given ifa for the desired list
2766 */
2767static int
2768sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
2769{
2770	struct sctp_laddr *laddr;
2771
2772	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2773	if (laddr == NULL) {
2774		/* out of memory? */
2775		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2776		return (EINVAL);
2777	}
2778	SCTP_INCR_LADDR_COUNT();
2779	memset(laddr, 0, sizeof(*laddr));
2780	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2781	laddr->ifa = ifa;
2782	laddr->action = act;
2783	atomic_add_int(&ifa->refcount, 1);
2784	/* insert it */
2785	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
2786
2787	return (0);
2788}
2789
2790/*
2791 * Remove an laddr entry from the local address list (on an assoc)
2792 */
2793static void
2794sctp_remove_laddr(struct sctp_laddr *laddr)
2795{
2796
2797	/* remove from the list */
2798	LIST_REMOVE(laddr, sctp_nxt_addr);
2799	sctp_free_ifa(laddr->ifa);
2800	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
2801	SCTP_DECR_LADDR_COUNT();
2802}
2803
2804/* sctp_ifap is used to bypass normal local address validation checks */
2805int
2806sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2807    struct sctp_ifa *sctp_ifap, struct thread *p)
2808{
2809	/* bind a ep to a socket address */
2810	struct sctppcbhead *head;
2811	struct sctp_inpcb *inp, *inp_tmp;
2812	struct inpcb *ip_inp;
2813	int port_reuse_active = 0;
2814	int bindall;
2815	uint16_t lport;
2816	int error;
2817	uint32_t vrf_id;
2818
2819	lport = 0;
2820	bindall = 1;
2821	inp = (struct sctp_inpcb *)so->so_pcb;
2822	ip_inp = (struct inpcb *)so->so_pcb;
2823#ifdef SCTP_DEBUG
2824	if (addr) {
2825		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2826		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2827		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2828		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2829	}
2830#endif
2831	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2832		/* already did a bind, subsequent binds NOT allowed ! */
2833		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2834		return (EINVAL);
2835	}
2836#ifdef INVARIANTS
2837	if (p == NULL)
2838		panic("null proc/thread");
2839#endif
2840	if (addr != NULL) {
2841		switch (addr->sa_family) {
2842#ifdef INET
2843		case AF_INET:
2844			{
2845				struct sockaddr_in *sin;
2846
2847				/* IPV6_V6ONLY socket? */
2848				if (SCTP_IPV6_V6ONLY(inp)) {
2849					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2850					return (EINVAL);
2851				}
2852				if (addr->sa_len != sizeof(*sin)) {
2853					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2854					return (EINVAL);
2855				}
2856
2857				sin = (struct sockaddr_in *)addr;
2858				lport = sin->sin_port;
2859				/*
2860				 * For LOOPBACK the prison_local_ip4() call
2861				 * will transmute the ip address to the
2862				 * proper value.
2863				 */
2864				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2865					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2866					return (error);
2867				}
2868				if (sin->sin_addr.s_addr != INADDR_ANY) {
2869					bindall = 0;
2870				}
2871				break;
2872			}
2873#endif
2874#ifdef INET6
2875		case AF_INET6:
2876			{
2877				/*
2878				 * Only for pure IPv6 Address. (No IPv4
2879				 * Mapped!)
2880				 */
2881				struct sockaddr_in6 *sin6;
2882
2883				sin6 = (struct sockaddr_in6 *)addr;
2884
2885				if (addr->sa_len != sizeof(*sin6)) {
2886					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2887					return (EINVAL);
2888				}
2889				lport = sin6->sin6_port;
2890				/*
2891				 * For LOOPBACK the prison_local_ip6() call
2892				 * will transmute the ipv6 address to the
2893				 * proper value.
2894				 */
2895				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2896				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2897					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2898					return (error);
2899				}
2900				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2901					bindall = 0;
2902					/* KAME hack: embed scopeid */
2903					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2904						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2905						return (EINVAL);
2906					}
2907				}
2908				/* this must be cleared for ifa_ifwithaddr() */
2909				sin6->sin6_scope_id = 0;
2910				break;
2911			}
2912#endif
2913		default:
2914			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2915			return (EAFNOSUPPORT);
2916		}
2917	}
2918	SCTP_INP_INFO_WLOCK();
2919	SCTP_INP_WLOCK(inp);
2920	/* Setup a vrf_id to be the default for the non-bind-all case. */
2921	vrf_id = inp->def_vrf_id;
2922
2923	/* increase our count due to the unlock we do */
2924	SCTP_INP_INCR_REF(inp);
2925	if (lport) {
2926		/*
2927		 * Did the caller specify a port? if so we must see if an ep
2928		 * already has this one bound.
2929		 */
2930		/* got to be root to get at low ports */
2931		if (ntohs(lport) < IPPORT_RESERVED) {
2932			if ((p != NULL) && ((error =
2933			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2934			    ) != 0)) {
2935				SCTP_INP_DECR_REF(inp);
2936				SCTP_INP_WUNLOCK(inp);
2937				SCTP_INP_INFO_WUNLOCK();
2938				return (error);
2939			}
2940		}
2941		SCTP_INP_WUNLOCK(inp);
2942		if (bindall) {
2943			vrf_id = inp->def_vrf_id;
2944			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2945			if (inp_tmp != NULL) {
2946				/*
2947				 * lock guy returned and lower count note
2948				 * that we are not bound so inp_tmp should
2949				 * NEVER be inp. And it is this inp
2950				 * (inp_tmp) that gets the reference bump,
2951				 * so we must lower it.
2952				 */
2953				SCTP_INP_DECR_REF(inp_tmp);
2954				/* unlock info */
2955				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2956				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2957					/*
2958					 * Ok, must be one-2-one and
2959					 * allowing port re-use
2960					 */
2961					port_reuse_active = 1;
2962					goto continue_anyway;
2963				}
2964				SCTP_INP_DECR_REF(inp);
2965				SCTP_INP_INFO_WUNLOCK();
2966				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2967				return (EADDRINUSE);
2968			}
2969		} else {
2970			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2971			if (inp_tmp != NULL) {
2972				/*
2973				 * lock guy returned and lower count note
2974				 * that we are not bound so inp_tmp should
2975				 * NEVER be inp. And it is this inp
2976				 * (inp_tmp) that gets the reference bump,
2977				 * so we must lower it.
2978				 */
2979				SCTP_INP_DECR_REF(inp_tmp);
2980				/* unlock info */
2981				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2982				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2983					/*
2984					 * Ok, must be one-2-one and
2985					 * allowing port re-use
2986					 */
2987					port_reuse_active = 1;
2988					goto continue_anyway;
2989				}
2990				SCTP_INP_DECR_REF(inp);
2991				SCTP_INP_INFO_WUNLOCK();
2992				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2993				return (EADDRINUSE);
2994			}
2995		}
2996continue_anyway:
2997		SCTP_INP_WLOCK(inp);
2998		if (bindall) {
2999			/* verify that no lport is not used by a singleton */
3000			if ((port_reuse_active == 0) &&
3001			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
3002				/* Sorry someone already has this one bound */
3003				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
3004				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
3005					port_reuse_active = 1;
3006				} else {
3007					SCTP_INP_DECR_REF(inp);
3008					SCTP_INP_WUNLOCK(inp);
3009					SCTP_INP_INFO_WUNLOCK();
3010					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3011					return (EADDRINUSE);
3012				}
3013			}
3014		}
3015	} else {
3016		uint16_t first, last, candidate;
3017		uint16_t count;
3018		int done;
3019
3020		if (ip_inp->inp_flags & INP_HIGHPORT) {
3021			first = MODULE_GLOBAL(ipport_hifirstauto);
3022			last = MODULE_GLOBAL(ipport_hilastauto);
3023		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3024			if (p && (error =
3025			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3026			    )) {
3027				SCTP_INP_DECR_REF(inp);
3028				SCTP_INP_WUNLOCK(inp);
3029				SCTP_INP_INFO_WUNLOCK();
3030				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3031				return (error);
3032			}
3033			first = MODULE_GLOBAL(ipport_lowfirstauto);
3034			last = MODULE_GLOBAL(ipport_lowlastauto);
3035		} else {
3036			first = MODULE_GLOBAL(ipport_firstauto);
3037			last = MODULE_GLOBAL(ipport_lastauto);
3038		}
3039		if (first > last) {
3040			uint16_t temp;
3041
3042			temp = first;
3043			first = last;
3044			last = temp;
3045		}
3046		count = last - first + 1;	/* number of candidates */
3047		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3048
3049		done = 0;
3050		while (!done) {
3051			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3052				done = 1;
3053			}
3054			if (!done) {
3055				if (--count == 0) {
3056					SCTP_INP_DECR_REF(inp);
3057					SCTP_INP_WUNLOCK(inp);
3058					SCTP_INP_INFO_WUNLOCK();
3059					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3060					return (EADDRINUSE);
3061				}
3062				if (candidate == last)
3063					candidate = first;
3064				else
3065					candidate = candidate + 1;
3066			}
3067		}
3068		lport = htons(candidate);
3069	}
3070	SCTP_INP_DECR_REF(inp);
3071	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3072	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3073		/*
3074		 * this really should not happen. The guy did a non-blocking
3075		 * bind and then did a close at the same time.
3076		 */
3077		SCTP_INP_WUNLOCK(inp);
3078		SCTP_INP_INFO_WUNLOCK();
3079		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3080		return (EINVAL);
3081	}
3082	/* ok we look clear to give out this port, so lets setup the binding */
3083	if (bindall) {
3084		/* binding to all addresses, so just set in the proper flags */
3085		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3086		/* set the automatic addr changes from kernel flag */
3087		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3088			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3089			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3090		} else {
3091			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3092			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3093		}
3094		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3095			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3096		} else {
3097			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3098		}
3099		/*
3100		 * set the automatic mobility_base from kernel flag (by
3101		 * micchie)
3102		 */
3103		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3104			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3105			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3106		} else {
3107			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3108			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3109		}
3110		/*
3111		 * set the automatic mobility_fasthandoff from kernel flag
3112		 * (by micchie)
3113		 */
3114		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3115			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3116			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3117		} else {
3118			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3119			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3120		}
3121	} else {
3122		/*
3123		 * bind specific, make sure flags is off and add a new
3124		 * address structure to the sctp_addr_list inside the ep
3125		 * structure.
3126		 *
3127		 * We will need to allocate one and insert it at the head.
3128		 * The socketopt call can just insert new addresses in there
3129		 * as well. It will also have to do the embed scope kame
3130		 * hack too (before adding).
3131		 */
3132		struct sctp_ifa *ifa;
3133		union sctp_sockstore store;
3134
3135		memset(&store, 0, sizeof(store));
3136		switch (addr->sa_family) {
3137#ifdef INET
3138		case AF_INET:
3139			memcpy(&store.sin, addr, sizeof(struct sockaddr_in));
3140			store.sin.sin_port = 0;
3141			break;
3142#endif
3143#ifdef INET6
3144		case AF_INET6:
3145			memcpy(&store.sin6, addr, sizeof(struct sockaddr_in6));
3146			store.sin6.sin6_port = 0;
3147			break;
3148#endif
3149		default:
3150			break;
3151		}
3152		/*
3153		 * first find the interface with the bound address need to
3154		 * zero out the port to find the address! yuck! can't do
3155		 * this earlier since need port for sctp_pcb_findep()
3156		 */
3157		if (sctp_ifap != NULL) {
3158			ifa = sctp_ifap;
3159		} else {
3160			/*
3161			 * Note for BSD we hit here always other O/S's will
3162			 * pass things in via the sctp_ifap argument.
3163			 */
3164			ifa = sctp_find_ifa_by_addr(&store.sa,
3165			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3166		}
3167		if (ifa == NULL) {
3168			/* Can't find an interface with that address */
3169			SCTP_INP_WUNLOCK(inp);
3170			SCTP_INP_INFO_WUNLOCK();
3171			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3172			return (EADDRNOTAVAIL);
3173		}
3174#ifdef INET6
3175		if (addr->sa_family == AF_INET6) {
3176			/* GAK, more FIXME IFA lock? */
3177			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3178				/* Can't bind a non-existent addr. */
3179				SCTP_INP_WUNLOCK(inp);
3180				SCTP_INP_INFO_WUNLOCK();
3181				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3182				return (EINVAL);
3183			}
3184		}
3185#endif
3186		/* we're not bound all */
3187		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3188		/* allow bindx() to send ASCONF's for binding changes */
3189		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3190		/* clear automatic addr changes from kernel flag */
3191		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3192
3193		/* add this address to the endpoint list */
3194		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3195		if (error != 0) {
3196			SCTP_INP_WUNLOCK(inp);
3197			SCTP_INP_INFO_WUNLOCK();
3198			return (error);
3199		}
3200		inp->laddr_count++;
3201	}
3202	/* find the bucket */
3203	if (port_reuse_active) {
3204		/* Put it into tcp 1-2-1 hash */
3205		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3206		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3207	} else {
3208		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3209	}
3210	/* put it in the bucket */
3211	LIST_INSERT_HEAD(head, inp, sctp_hash);
3212	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3213	    (void *)head, ntohs(lport), port_reuse_active);
3214	/* set in the port */
3215	inp->sctp_lport = lport;
3216
3217	/* turn off just the unbound flag */
3218	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3219	SCTP_INP_WUNLOCK(inp);
3220	SCTP_INP_INFO_WUNLOCK();
3221	return (0);
3222}
3223
3224static void
3225sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3226{
3227	struct sctp_iterator *it, *nit;
3228
3229	/*
3230	 * We enter with the only the ITERATOR_LOCK in place and a write
3231	 * lock on the inp_info stuff.
3232	 */
3233	it = sctp_it_ctl.cur_it;
3234	if (it && (it->vn != curvnet)) {
3235		/* Its not looking at our VNET */
3236		return;
3237	}
3238	if (it && (it->inp == inp)) {
3239		/*
3240		 * This is tricky and we hold the iterator lock, but when it
3241		 * returns and gets the lock (when we release it) the
3242		 * iterator will try to operate on inp. We need to stop that
3243		 * from happening. But of course the iterator has a
3244		 * reference on the stcb and inp. We can mark it and it will
3245		 * stop.
3246		 *
3247		 * If its a single iterator situation, we set the end
3248		 * iterator flag. Otherwise we set the iterator to go to the
3249		 * next inp.
3250		 *
3251		 */
3252		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3253			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3254		} else {
3255			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3256		}
3257	}
3258	/*
3259	 * Now go through and remove any single reference to our inp that
3260	 * may be still pending on the list
3261	 */
3262	SCTP_IPI_ITERATOR_WQ_LOCK();
3263	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3264		if (it->vn != curvnet) {
3265			continue;
3266		}
3267		if (it->inp == inp) {
3268			/* This one points to me is it inp specific? */
3269			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3270				/* Remove and free this one */
3271				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3272				    it, sctp_nxt_itr);
3273				if (it->function_atend != NULL) {
3274					(*it->function_atend) (it->pointer, it->val);
3275				}
3276				SCTP_FREE(it, SCTP_M_ITER);
3277			} else {
3278				it->inp = LIST_NEXT(it->inp, sctp_list);
3279				if (it->inp) {
3280					SCTP_INP_INCR_REF(it->inp);
3281				}
3282			}
3283			/*
3284			 * When its put in the refcnt is incremented so decr
3285			 * it
3286			 */
3287			SCTP_INP_DECR_REF(inp);
3288		}
3289	}
3290	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3291}
3292
3293/* release sctp_inpcb unbind the port */
3294void
3295sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3296{
3297	/*
3298	 * Here we free a endpoint. We must find it (if it is in the Hash
3299	 * table) and remove it from there. Then we must also find it in the
3300	 * overall list and remove it from there. After all removals are
3301	 * complete then any timer has to be stopped. Then start the actual
3302	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3303	 * all associations. d) finally the ep itself.
3304	 */
3305	struct sctp_tcb *asoc, *nasoc;
3306	struct sctp_laddr *laddr, *nladdr;
3307	struct inpcb *ip_pcb;
3308	struct socket *so;
3309	int being_refed = 0;
3310	struct sctp_queued_to_read *sq, *nsq;
3311	int cnt;
3312	sctp_sharedkey_t *shared_key, *nshared_key;
3313
3314#ifdef SCTP_LOG_CLOSING
3315	sctp_log_closing(inp, NULL, 0);
3316#endif
3317	SCTP_ITERATOR_LOCK();
3318	/* mark any iterators on the list or being processed */
3319	sctp_iterator_inp_being_freed(inp);
3320	SCTP_ITERATOR_UNLOCK();
3321	so = inp->sctp_socket;
3322	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3323		/* been here before.. eeks.. get out of here */
3324		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3325#ifdef SCTP_LOG_CLOSING
3326		sctp_log_closing(inp, NULL, 1);
3327#endif
3328		return;
3329	}
3330	SCTP_ASOC_CREATE_LOCK(inp);
3331	SCTP_INP_INFO_WLOCK();
3332
3333	SCTP_INP_WLOCK(inp);
3334	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3335		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3336		/* socket is gone, so no more wakeups allowed */
3337		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3338		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3339		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3340	}
3341	/* First time through we have the socket lock, after that no more. */
3342	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3343	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3344
3345	if (inp->control) {
3346		sctp_m_freem(inp->control);
3347		inp->control = NULL;
3348	}
3349	if (inp->pkt) {
3350		sctp_m_freem(inp->pkt);
3351		inp->pkt = NULL;
3352	}
3353	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3354					 * here but I will be nice :> (i.e.
3355					 * ip_pcb = ep;) */
3356	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3357		int cnt_in_sd;
3358
3359		cnt_in_sd = 0;
3360		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3361			SCTP_TCB_LOCK(asoc);
3362			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3363				/* Skip guys being freed */
3364				cnt_in_sd++;
3365				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3366					/*
3367					 * Special case - we did not start a
3368					 * kill timer on the asoc due to it
3369					 * was not closed. So go ahead and
3370					 * start it now.
3371					 */
3372					SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_IN_ACCEPT_QUEUE);
3373					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3374				}
3375				SCTP_TCB_UNLOCK(asoc);
3376				continue;
3377			}
3378			if (((SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_WAIT) ||
3379			    (SCTP_GET_STATE(asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3380			    (asoc->asoc.total_output_queue_size == 0)) {
3381				/*
3382				 * If we have data in queue, we don't want
3383				 * to just free since the app may have done,
3384				 * send()/close or connect/send/close. And
3385				 * it wants the data to get across first.
3386				 */
3387				/* Just abandon things in the front states */
3388				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3389				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3390					cnt_in_sd++;
3391				}
3392				continue;
3393			}
3394			/* Disconnect the socket please */
3395			asoc->sctp_socket = NULL;
3396			SCTP_ADD_SUBSTATE(asoc, SCTP_STATE_CLOSED_SOCKET);
3397			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3398			    (asoc->asoc.control_pdapi) ||
3399			    (asoc->asoc.size_on_all_streams > 0) ||
3400			    (so && (so->so_rcv.sb_cc > 0))) {
3401				/* Left with Data unread */
3402				struct mbuf *op_err;
3403
3404				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3405				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3406				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3407				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3408				if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
3409				    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3410					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3411				}
3412				if (sctp_free_assoc(inp, asoc,
3413				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3414					cnt_in_sd++;
3415				}
3416				continue;
3417			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3418				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3419			    (asoc->asoc.stream_queue_cnt == 0)) {
3420				if ((*asoc->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete) (asoc, &asoc->asoc)) {
3421					goto abort_anyway;
3422				}
3423				if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3424				    (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3425					struct sctp_nets *netp;
3426
3427					/*
3428					 * there is nothing queued to send,
3429					 * so I send shutdown
3430					 */
3431					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
3432					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3433						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3434					}
3435					SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
3436					sctp_stop_timers_for_shutdown(asoc);
3437					if (asoc->asoc.alternate) {
3438						netp = asoc->asoc.alternate;
3439					} else {
3440						netp = asoc->asoc.primary_destination;
3441					}
3442					sctp_send_shutdown(asoc, netp);
3443					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3444					    netp);
3445					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc, NULL);
3446					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3447				}
3448			} else {
3449				/* mark into shutdown pending */
3450				SCTP_ADD_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
3451				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc, NULL);
3452				if ((*asoc->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete) (asoc, &asoc->asoc)) {
3453					SCTP_ADD_SUBSTATE(asoc, SCTP_STATE_PARTIAL_MSG_LEFT);
3454				}
3455				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3456				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3457				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3458					struct mbuf *op_err;
3459
3460			abort_anyway:
3461					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3462					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3463					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3464					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3465					if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
3466					    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3467						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3468					}
3469					if (sctp_free_assoc(inp, asoc,
3470					    SCTP_PCBFREE_NOFORCE,
3471					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3472						cnt_in_sd++;
3473					}
3474					continue;
3475				} else {
3476					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3477				}
3478			}
3479			cnt_in_sd++;
3480			SCTP_TCB_UNLOCK(asoc);
3481		}
3482		/* now is there some left in our SHUTDOWN state? */
3483		if (cnt_in_sd) {
3484#ifdef SCTP_LOG_CLOSING
3485			sctp_log_closing(inp, NULL, 2);
3486#endif
3487			inp->sctp_socket = NULL;
3488			SCTP_INP_WUNLOCK(inp);
3489			SCTP_ASOC_CREATE_UNLOCK(inp);
3490			SCTP_INP_INFO_WUNLOCK();
3491			return;
3492		}
3493	}
3494	inp->sctp_socket = NULL;
3495	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3496	    SCTP_PCB_FLAGS_UNBOUND) {
3497		/*
3498		 * ok, this guy has been bound. It's port is somewhere in
3499		 * the SCTP_BASE_INFO(hash table). Remove it!
3500		 */
3501		LIST_REMOVE(inp, sctp_hash);
3502		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3503	}
3504
3505	/*
3506	 * If there is a timer running to kill us, forget it, since it may
3507	 * have a contest on the INP lock.. which would cause us to die ...
3508	 */
3509	cnt = 0;
3510	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3511		SCTP_TCB_LOCK(asoc);
3512		if (immediate != SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3513			/* Disconnect the socket please */
3514			asoc->sctp_socket = NULL;
3515			SCTP_ADD_SUBSTATE(asoc, SCTP_STATE_CLOSED_SOCKET);
3516		}
3517		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3518			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3519				SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_IN_ACCEPT_QUEUE);
3520				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3521			}
3522			cnt++;
3523			SCTP_TCB_UNLOCK(asoc);
3524			continue;
3525		}
3526		/* Free associations that are NOT killing us */
3527		if ((SCTP_GET_STATE(asoc) != SCTP_STATE_COOKIE_WAIT) &&
3528		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3529			struct mbuf *op_err;
3530
3531			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3532			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3533			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3534			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3535		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3536			cnt++;
3537			SCTP_TCB_UNLOCK(asoc);
3538			continue;
3539		}
3540		if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
3541		    (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3542			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3543		}
3544		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE,
3545		    SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3546			cnt++;
3547		}
3548	}
3549	if (cnt) {
3550		/* Ok we have someone out there that will kill us */
3551#ifdef SCTP_LOG_CLOSING
3552		sctp_log_closing(inp, NULL, 3);
3553#endif
3554		SCTP_INP_WUNLOCK(inp);
3555		SCTP_ASOC_CREATE_UNLOCK(inp);
3556		SCTP_INP_INFO_WUNLOCK();
3557		return;
3558	}
3559	if (SCTP_INP_LOCK_CONTENDED(inp))
3560		being_refed++;
3561	if (SCTP_INP_READ_CONTENDED(inp))
3562		being_refed++;
3563	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3564		being_refed++;
3565	/* NOTE: 0 refcount also means no timers are referencing us. */
3566	if ((inp->refcount) ||
3567	    (being_refed) ||
3568	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3569#ifdef SCTP_LOG_CLOSING
3570		sctp_log_closing(inp, NULL, 4);
3571#endif
3572		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3573		SCTP_INP_WUNLOCK(inp);
3574		SCTP_ASOC_CREATE_UNLOCK(inp);
3575		SCTP_INP_INFO_WUNLOCK();
3576		return;
3577	}
3578	inp->sctp_ep.signature_change.type = 0;
3579	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3580	/*
3581	 * Remove it from the list .. last thing we need a lock for.
3582	 */
3583	LIST_REMOVE(inp, sctp_list);
3584	SCTP_INP_WUNLOCK(inp);
3585	SCTP_ASOC_CREATE_UNLOCK(inp);
3586	SCTP_INP_INFO_WUNLOCK();
3587
3588#ifdef SCTP_LOG_CLOSING
3589	sctp_log_closing(inp, NULL, 5);
3590#endif
3591	if ((inp->sctp_asocidhash) != NULL) {
3592		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3593		inp->sctp_asocidhash = NULL;
3594	}
3595	/* sa_ignore FREED_MEMORY */
3596	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3597		/* Its only abandoned if it had data left */
3598		if (sq->length)
3599			SCTP_STAT_INCR(sctps_left_abandon);
3600
3601		TAILQ_REMOVE(&inp->read_queue, sq, next);
3602		sctp_free_remote_addr(sq->whoFrom);
3603		if (so)
3604			so->so_rcv.sb_cc -= sq->length;
3605		if (sq->data) {
3606			sctp_m_freem(sq->data);
3607			sq->data = NULL;
3608		}
3609		/*
3610		 * no need to free the net count, since at this point all
3611		 * assoc's are gone.
3612		 */
3613		sctp_free_a_readq(NULL, sq);
3614	}
3615	/* Now the sctp_pcb things */
3616	/*
3617	 * free each asoc if it is not already closed/free. we can't use the
3618	 * macro here since le_next will get freed as part of the
3619	 * sctp_free_assoc() call.
3620	 */
3621	if (ip_pcb->inp_options) {
3622		(void)sctp_m_free(ip_pcb->inp_options);
3623		ip_pcb->inp_options = 0;
3624	}
3625#ifdef INET6
3626	if (ip_pcb->inp_vflag & INP_IPV6) {
3627		ip6_freepcbopts(ip_pcb->in6p_outputopts);
3628	}
3629#endif				/* INET6 */
3630	ip_pcb->inp_vflag = 0;
3631	/* free up authentication fields */
3632	if (inp->sctp_ep.local_auth_chunks != NULL)
3633		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3634	if (inp->sctp_ep.local_hmacs != NULL)
3635		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3636
3637	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3638		LIST_REMOVE(shared_key, next);
3639		sctp_free_sharedkey(shared_key);
3640		/* sa_ignore FREED_MEMORY */
3641	}
3642
3643	/*
3644	 * if we have an address list the following will free the list of
3645	 * ifaddr's that are set into this ep. Again macro limitations here,
3646	 * since the LIST_FOREACH could be a bad idea.
3647	 */
3648	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3649		sctp_remove_laddr(laddr);
3650	}
3651
3652#ifdef SCTP_TRACK_FREED_ASOCS
3653	/* TEMP CODE */
3654	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3655		LIST_REMOVE(asoc, sctp_tcblist);
3656		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3657		SCTP_DECR_ASOC_COUNT();
3658	}
3659	/* *** END TEMP CODE *** */
3660#endif
3661	/* Now lets see about freeing the EP hash table. */
3662	if (inp->sctp_tcbhash != NULL) {
3663		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3664		inp->sctp_tcbhash = NULL;
3665	}
3666	/* Now we must put the ep memory back into the zone pool */
3667	crfree(inp->ip_inp.inp.inp_cred);
3668	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3669	SCTP_INP_LOCK_DESTROY(inp);
3670	SCTP_INP_READ_DESTROY(inp);
3671	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3672	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3673	SCTP_DECR_EP_COUNT();
3674}
3675
3676struct sctp_nets *
3677sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3678{
3679	struct sctp_nets *net;
3680
3681	/* locate the address */
3682	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3683		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3684			return (net);
3685	}
3686	return (NULL);
3687}
3688
3689int
3690sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3691{
3692	struct sctp_ifa *sctp_ifa;
3693
3694	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3695	if (sctp_ifa) {
3696		return (1);
3697	} else {
3698		return (0);
3699	}
3700}
3701
3702/*
3703 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3704 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3705 * stats of stuff.
3706 */
3707int
3708sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3709    struct sctp_nets **netp, uint16_t port, int set_scope, int from)
3710{
3711	/*
3712	 * The following is redundant to the same lines in the
3713	 * sctp_aloc_assoc() but is needed since others call the add address
3714	 * function
3715	 */
3716	struct sctp_nets *net, *netfirst;
3717	int addr_inscope;
3718
3719	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3720	    from);
3721	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3722
3723	netfirst = sctp_findnet(stcb, newaddr);
3724	if (netfirst) {
3725		/*
3726		 * Lie and return ok, we don't want to make the association
3727		 * go away for this behavior. It will happen in the TCP
3728		 * model in a connected socket. It does not reach the hash
3729		 * table until after the association is built so it can't be
3730		 * found. Mark as reachable, since the initial creation will
3731		 * have been cleared and the NOT_IN_ASSOC flag will have
3732		 * been added... and we don't want to end up removing it
3733		 * back out.
3734		 */
3735		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3736			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3737			    SCTP_ADDR_UNCONFIRMED);
3738		} else {
3739			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3740		}
3741
3742		return (0);
3743	}
3744	addr_inscope = 1;
3745	switch (newaddr->sa_family) {
3746#ifdef INET
3747	case AF_INET:
3748		{
3749			struct sockaddr_in *sin;
3750
3751			sin = (struct sockaddr_in *)newaddr;
3752			if (sin->sin_addr.s_addr == 0) {
3753				/* Invalid address */
3754				return (-1);
3755			}
3756			/* zero out the zero area */
3757			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3758
3759			/* assure len is set */
3760			sin->sin_len = sizeof(struct sockaddr_in);
3761			if (set_scope) {
3762				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3763					stcb->asoc.scope.ipv4_local_scope = 1;
3764				}
3765			} else {
3766				/* Validate the address is in scope */
3767				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3768				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3769					addr_inscope = 0;
3770				}
3771			}
3772			break;
3773		}
3774#endif
3775#ifdef INET6
3776	case AF_INET6:
3777		{
3778			struct sockaddr_in6 *sin6;
3779
3780			sin6 = (struct sockaddr_in6 *)newaddr;
3781			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3782				/* Invalid address */
3783				return (-1);
3784			}
3785			/* assure len is set */
3786			sin6->sin6_len = sizeof(struct sockaddr_in6);
3787			if (set_scope) {
3788				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3789					stcb->asoc.scope.loopback_scope = 1;
3790					stcb->asoc.scope.local_scope = 0;
3791					stcb->asoc.scope.ipv4_local_scope = 1;
3792					stcb->asoc.scope.site_scope = 1;
3793				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3794					/*
3795					 * If the new destination is a
3796					 * LINK_LOCAL we must have common
3797					 * site scope. Don't set the local
3798					 * scope since we may not share all
3799					 * links, only loopback can do this.
3800					 * Links on the local network would
3801					 * also be on our private network
3802					 * for v4 too.
3803					 */
3804					stcb->asoc.scope.ipv4_local_scope = 1;
3805					stcb->asoc.scope.site_scope = 1;
3806				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3807					/*
3808					 * If the new destination is
3809					 * SITE_LOCAL then we must have site
3810					 * scope in common.
3811					 */
3812					stcb->asoc.scope.site_scope = 1;
3813				}
3814			} else {
3815				/* Validate the address is in scope */
3816				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3817				    (stcb->asoc.scope.loopback_scope == 0)) {
3818					addr_inscope = 0;
3819				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3820				    (stcb->asoc.scope.local_scope == 0)) {
3821					addr_inscope = 0;
3822				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3823				    (stcb->asoc.scope.site_scope == 0)) {
3824					addr_inscope = 0;
3825				}
3826			}
3827			break;
3828		}
3829#endif
3830	default:
3831		/* not supported family type */
3832		return (-1);
3833	}
3834	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3835	if (net == NULL) {
3836		return (-1);
3837	}
3838	SCTP_INCR_RADDR_COUNT();
3839	memset(net, 0, sizeof(struct sctp_nets));
3840	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3841	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3842	switch (newaddr->sa_family) {
3843#ifdef INET
3844	case AF_INET:
3845		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3846		break;
3847#endif
3848#ifdef INET6
3849	case AF_INET6:
3850		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3851		break;
3852#endif
3853	default:
3854		break;
3855	}
3856	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3857	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3858		stcb->asoc.scope.loopback_scope = 1;
3859		stcb->asoc.scope.ipv4_local_scope = 1;
3860		stcb->asoc.scope.local_scope = 0;
3861		stcb->asoc.scope.site_scope = 1;
3862		addr_inscope = 1;
3863	}
3864	net->failure_threshold = stcb->asoc.def_net_failure;
3865	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3866	if (addr_inscope == 0) {
3867		net->dest_state = (SCTP_ADDR_REACHABLE |
3868		    SCTP_ADDR_OUT_OF_SCOPE);
3869	} else {
3870		if (from == SCTP_ADDR_IS_CONFIRMED)
3871			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3872			net->dest_state = SCTP_ADDR_REACHABLE;
3873		else
3874			net->dest_state = SCTP_ADDR_REACHABLE |
3875			    SCTP_ADDR_UNCONFIRMED;
3876	}
3877	/*
3878	 * We set this to 0, the timer code knows that this means its an
3879	 * initial value
3880	 */
3881	net->rto_needed = 1;
3882	net->RTO = 0;
3883	net->RTO_measured = 0;
3884	stcb->asoc.numnets++;
3885	net->ref_count = 1;
3886	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3887	net->port = port;
3888	net->dscp = stcb->asoc.default_dscp;
3889#ifdef INET6
3890	net->flowlabel = stcb->asoc.default_flowlabel;
3891#endif
3892	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3893		net->dest_state |= SCTP_ADDR_NOHB;
3894	} else {
3895		net->dest_state &= ~SCTP_ADDR_NOHB;
3896	}
3897	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3898		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3899	} else {
3900		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3901	}
3902	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3903	/* Init the timer structure */
3904	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3905	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3906	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3907
3908	/* Now generate a route for this guy */
3909#ifdef INET6
3910	/* KAME hack: embed scopeid */
3911	if (newaddr->sa_family == AF_INET6) {
3912		struct sockaddr_in6 *sin6;
3913
3914		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3915		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3916		sin6->sin6_scope_id = 0;
3917	}
3918#endif
3919	SCTP_RTALLOC((sctp_route_t *)&net->ro,
3920	    stcb->asoc.vrf_id,
3921	    stcb->sctp_ep->fibnum);
3922
3923	net->src_addr_selected = 0;
3924	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3925		/* Get source address */
3926		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3927		    stcb,
3928		    (sctp_route_t *)&net->ro,
3929		    net,
3930		    0,
3931		    stcb->asoc.vrf_id);
3932		if (stcb->asoc.default_mtu > 0) {
3933			net->mtu = stcb->asoc.default_mtu;
3934			switch (net->ro._l_addr.sa.sa_family) {
3935#ifdef INET
3936			case AF_INET:
3937				net->mtu += SCTP_MIN_V4_OVERHEAD;
3938				break;
3939#endif
3940#ifdef INET6
3941			case AF_INET6:
3942				net->mtu += SCTP_MIN_OVERHEAD;
3943				break;
3944#endif
3945			default:
3946				break;
3947			}
3948#if defined(INET) || defined(INET6)
3949			if (net->port) {
3950				net->mtu += (uint32_t)sizeof(struct udphdr);
3951			}
3952#endif
3953		} else if (net->ro._s_addr != NULL) {
3954			uint32_t imtu, rmtu, hcmtu;
3955
3956			net->src_addr_selected = 1;
3957			/* Now get the interface MTU */
3958			if (net->ro._s_addr->ifn_p != NULL) {
3959				imtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3960			} else {
3961				imtu = 0;
3962			}
3963			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_nh);
3964			hcmtu = sctp_hc_get_mtu(&net->ro._l_addr, stcb->sctp_ep->fibnum);
3965			net->mtu = sctp_min_mtu(hcmtu, rmtu, imtu);
3966		}
3967	}
3968	if (net->mtu == 0) {
3969		if (stcb->asoc.default_mtu > 0) {
3970			net->mtu = stcb->asoc.default_mtu;
3971			switch (net->ro._l_addr.sa.sa_family) {
3972#ifdef INET
3973			case AF_INET:
3974				net->mtu += SCTP_MIN_V4_OVERHEAD;
3975				break;
3976#endif
3977#ifdef INET6
3978			case AF_INET6:
3979				net->mtu += SCTP_MIN_OVERHEAD;
3980				break;
3981#endif
3982			default:
3983				break;
3984			}
3985#if defined(INET) || defined(INET6)
3986			if (net->port) {
3987				net->mtu += (uint32_t)sizeof(struct udphdr);
3988			}
3989#endif
3990		} else {
3991			switch (newaddr->sa_family) {
3992#ifdef INET
3993			case AF_INET:
3994				net->mtu = SCTP_DEFAULT_MTU;
3995				break;
3996#endif
3997#ifdef INET6
3998			case AF_INET6:
3999				net->mtu = 1280;
4000				break;
4001#endif
4002			default:
4003				break;
4004			}
4005		}
4006	}
4007#if defined(INET) || defined(INET6)
4008	if (net->port) {
4009		net->mtu -= (uint32_t)sizeof(struct udphdr);
4010	}
4011#endif
4012	if (from == SCTP_ALLOC_ASOC) {
4013		stcb->asoc.smallest_mtu = net->mtu;
4014	}
4015	if (stcb->asoc.smallest_mtu > net->mtu) {
4016		sctp_pathmtu_adjustment(stcb, net->mtu);
4017	}
4018#ifdef INET6
4019	if (newaddr->sa_family == AF_INET6) {
4020		struct sockaddr_in6 *sin6;
4021
4022		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4023		(void)sa6_recoverscope(sin6);
4024	}
4025#endif
4026
4027	/* JRS - Use the congestion control given in the CC module */
4028	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4029		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4030
4031	/*
4032	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4033	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4034	 */
4035	net->find_pseudo_cumack = 1;
4036	net->find_rtx_pseudo_cumack = 1;
4037	/* Choose an initial flowid. */
4038	net->flowid = stcb->asoc.my_vtag ^
4039	    ntohs(stcb->rport) ^
4040	    ntohs(stcb->sctp_ep->sctp_lport);
4041	net->flowtype = M_HASHTYPE_OPAQUE_HASH;
4042	if (netp) {
4043		*netp = net;
4044	}
4045	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4046	if (net->ro.ro_nh == NULL) {
4047		/* Since we have no route put it at the back */
4048		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4049	} else if (netfirst == NULL) {
4050		/* We are the first one in the pool. */
4051		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4052	} else if (netfirst->ro.ro_nh == NULL) {
4053		/*
4054		 * First one has NO route. Place this one ahead of the first
4055		 * one.
4056		 */
4057		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4058	} else if (net->ro.ro_nh->nh_ifp != netfirst->ro.ro_nh->nh_ifp) {
4059		/*
4060		 * This one has a different interface than the one at the
4061		 * top of the list. Place it ahead.
4062		 */
4063		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4064	} else {
4065		/*
4066		 * Ok we have the same interface as the first one. Move
4067		 * forward until we find either a) one with a NULL route...
4068		 * insert ahead of that b) one with a different ifp.. insert
4069		 * after that. c) end of the list.. insert at the tail.
4070		 */
4071		struct sctp_nets *netlook;
4072
4073		do {
4074			netlook = TAILQ_NEXT(netfirst, sctp_next);
4075			if (netlook == NULL) {
4076				/* End of the list */
4077				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4078				break;
4079			} else if (netlook->ro.ro_nh == NULL) {
4080				/* next one has NO route */
4081				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4082				break;
4083			} else if (netlook->ro.ro_nh->nh_ifp != net->ro.ro_nh->nh_ifp) {
4084				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4085				    net, sctp_next);
4086				break;
4087			}
4088			/* Shift forward */
4089			netfirst = netlook;
4090		} while (netlook != NULL);
4091	}
4092
4093	/* got to have a primary set */
4094	if (stcb->asoc.primary_destination == 0) {
4095		stcb->asoc.primary_destination = net;
4096	} else if ((stcb->asoc.primary_destination->ro.ro_nh == NULL) &&
4097		    (net->ro.ro_nh) &&
4098	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4099		/* No route to current primary adopt new primary */
4100		stcb->asoc.primary_destination = net;
4101	}
4102	/* Validate primary is first */
4103	net = TAILQ_FIRST(&stcb->asoc.nets);
4104	if ((net != stcb->asoc.primary_destination) &&
4105	    (stcb->asoc.primary_destination)) {
4106		/*
4107		 * first one on the list is NOT the primary sctp_cmpaddr()
4108		 * is much more efficient if the primary is the first on the
4109		 * list, make it so.
4110		 */
4111		TAILQ_REMOVE(&stcb->asoc.nets,
4112		    stcb->asoc.primary_destination, sctp_next);
4113		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4114		    stcb->asoc.primary_destination, sctp_next);
4115	}
4116	return (0);
4117}
4118
4119static uint32_t
4120sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4121{
4122	uint32_t id;
4123	struct sctpasochead *head;
4124	struct sctp_tcb *lstcb;
4125
4126try_again:
4127	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4128		/* TSNH */
4129		return (0);
4130	}
4131	/*
4132	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4133	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4134	 */
4135	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4136		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4137	}
4138	id = inp->sctp_associd_counter;
4139	inp->sctp_associd_counter++;
4140	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t)id, 0);
4141	if (lstcb) {
4142		goto try_again;
4143	}
4144	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4145	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4146	stcb->asoc.in_asocid_hash = 1;
4147	return (id);
4148}
4149
4150/*
4151 * allocate an association and add it to the endpoint. The caller must be
4152 * careful to add all additional addresses once they are know right away or
4153 * else the assoc will be may experience a blackout scenario.
4154 */
4155struct sctp_tcb *
4156sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4157    int *error, uint32_t override_tag, uint32_t vrf_id,
4158    uint16_t o_streams, uint16_t port,
4159    struct thread *p,
4160    int initialize_auth_params)
4161{
4162	/* note the p argument is only valid in unbound sockets */
4163
4164	struct sctp_tcb *stcb;
4165	struct sctp_association *asoc;
4166	struct sctpasochead *head;
4167	uint16_t rport;
4168	int err;
4169
4170	/*
4171	 * Assumption made here: Caller has done a
4172	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4173	 * address does not exist already.
4174	 */
4175	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4176		/* Hit max assoc, sorry no more */
4177		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4178		*error = ENOBUFS;
4179		return (NULL);
4180	}
4181	if (firstaddr == NULL) {
4182		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4183		*error = EINVAL;
4184		return (NULL);
4185	}
4186	SCTP_INP_RLOCK(inp);
4187	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4188	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4189	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4190		/*
4191		 * If its in the TCP pool, its NOT allowed to create an
4192		 * association. The parent listener needs to call
4193		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4194		 * off, or connected one does this.. its an error.
4195		 */
4196		SCTP_INP_RUNLOCK(inp);
4197		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4198		*error = EINVAL;
4199		return (NULL);
4200	}
4201	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4202	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4203		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4204		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4205			SCTP_INP_RUNLOCK(inp);
4206			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4207			*error = EINVAL;
4208			return (NULL);
4209		}
4210	}
4211	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4212#ifdef SCTP_DEBUG
4213	if (firstaddr) {
4214		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4215		switch (firstaddr->sa_family) {
4216#ifdef INET
4217		case AF_INET:
4218			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4219			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4220			break;
4221#endif
4222#ifdef INET6
4223		case AF_INET6:
4224			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4225			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4226			break;
4227#endif
4228		default:
4229			break;
4230		}
4231	} else {
4232		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4233	}
4234#endif				/* SCTP_DEBUG */
4235	switch (firstaddr->sa_family) {
4236#ifdef INET
4237	case AF_INET:
4238		{
4239			struct sockaddr_in *sin;
4240
4241			sin = (struct sockaddr_in *)firstaddr;
4242			if ((ntohs(sin->sin_port) == 0) ||
4243			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4244			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4245			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) ||
4246			    (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) != 0) &&
4247			    (SCTP_IPV6_V6ONLY(inp) != 0))) {
4248				/* Invalid address */
4249				SCTP_INP_RUNLOCK(inp);
4250				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4251				*error = EINVAL;
4252				return (NULL);
4253			}
4254			rport = sin->sin_port;
4255			break;
4256		}
4257#endif
4258#ifdef INET6
4259	case AF_INET6:
4260		{
4261			struct sockaddr_in6 *sin6;
4262
4263			sin6 = (struct sockaddr_in6 *)firstaddr;
4264			if ((ntohs(sin6->sin6_port) == 0) ||
4265			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4266			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) ||
4267			    ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0)) {
4268				/* Invalid address */
4269				SCTP_INP_RUNLOCK(inp);
4270				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4271				*error = EINVAL;
4272				return (NULL);
4273			}
4274			rport = sin6->sin6_port;
4275			break;
4276		}
4277#endif
4278	default:
4279		/* not supported family type */
4280		SCTP_INP_RUNLOCK(inp);
4281		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4282		*error = EINVAL;
4283		return (NULL);
4284	}
4285	SCTP_INP_RUNLOCK(inp);
4286	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4287		/*
4288		 * If you have not performed a bind, then we need to do the
4289		 * ephemeral bind for you.
4290		 */
4291		if ((err = sctp_inpcb_bind(inp->sctp_socket, NULL, NULL, p))) {
4292			/* bind error, probably perm */
4293			*error = err;
4294			return (NULL);
4295		}
4296	}
4297	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4298	if (stcb == NULL) {
4299		/* out of memory? */
4300		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4301		*error = ENOMEM;
4302		return (NULL);
4303	}
4304	SCTP_INCR_ASOC_COUNT();
4305
4306	memset(stcb, 0, sizeof(*stcb));
4307	asoc = &stcb->asoc;
4308
4309	SCTP_TCB_LOCK_INIT(stcb);
4310	SCTP_TCB_SEND_LOCK_INIT(stcb);
4311	stcb->rport = rport;
4312	/* setup back pointer's */
4313	stcb->sctp_ep = inp;
4314	stcb->sctp_socket = inp->sctp_socket;
4315	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id, o_streams))) {
4316		/* failed */
4317		SCTP_TCB_LOCK_DESTROY(stcb);
4318		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4319		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4320		SCTP_DECR_ASOC_COUNT();
4321		*error = err;
4322		return (NULL);
4323	}
4324	/* and the port */
4325	SCTP_INP_INFO_WLOCK();
4326	SCTP_INP_WLOCK(inp);
4327	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4328		/* inpcb freed while alloc going on */
4329		SCTP_TCB_LOCK_DESTROY(stcb);
4330		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4331		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4332		SCTP_INP_WUNLOCK(inp);
4333		SCTP_INP_INFO_WUNLOCK();
4334		SCTP_DECR_ASOC_COUNT();
4335		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4336		*error = EINVAL;
4337		return (NULL);
4338	}
4339	SCTP_TCB_LOCK(stcb);
4340
4341	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4342	/* now that my_vtag is set, add it to the hash */
4343	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4344	/* put it in the bucket in the vtag hash of assoc's for the system */
4345	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4346	SCTP_INP_INFO_WUNLOCK();
4347
4348	if (sctp_add_remote_addr(stcb, firstaddr, NULL, port, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC)) {
4349		/* failure.. memory error? */
4350		if (asoc->strmout) {
4351			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4352			asoc->strmout = NULL;
4353		}
4354		if (asoc->mapping_array) {
4355			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4356			asoc->mapping_array = NULL;
4357		}
4358		if (asoc->nr_mapping_array) {
4359			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4360			asoc->nr_mapping_array = NULL;
4361		}
4362		SCTP_DECR_ASOC_COUNT();
4363		SCTP_TCB_UNLOCK(stcb);
4364		SCTP_TCB_LOCK_DESTROY(stcb);
4365		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4366		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4367		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4368		SCTP_INP_WUNLOCK(inp);
4369		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4370		*error = ENOBUFS;
4371		return (NULL);
4372	}
4373	/* Init all the timers */
4374	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4375	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4376	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4377	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4378	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4379	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4380
4381	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4382	/* now file the port under the hash as well */
4383	if (inp->sctp_tcbhash != NULL) {
4384		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4385		    inp->sctp_hashmark)];
4386		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4387	}
4388	if (initialize_auth_params == SCTP_INITIALIZE_AUTH_PARAMS) {
4389		sctp_initialize_auth_params(inp, stcb);
4390	}
4391	SCTP_INP_WUNLOCK(inp);
4392	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4393	return (stcb);
4394}
4395
4396void
4397sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4398{
4399	struct sctp_inpcb *inp;
4400	struct sctp_association *asoc;
4401
4402	inp = stcb->sctp_ep;
4403	asoc = &stcb->asoc;
4404	asoc->numnets--;
4405	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4406	if (net == asoc->primary_destination) {
4407		/* Reset primary */
4408		struct sctp_nets *lnet;
4409
4410		lnet = TAILQ_FIRST(&asoc->nets);
4411		/*
4412		 * Mobility adaptation Ideally, if deleted destination is
4413		 * the primary, it becomes a fast retransmission trigger by
4414		 * the subsequent SET PRIMARY. (by micchie)
4415		 */
4416		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4417		    SCTP_MOBILITY_BASE) ||
4418		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4419		    SCTP_MOBILITY_FASTHANDOFF)) {
4420			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4421			if (asoc->deleted_primary != NULL) {
4422				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4423				goto out;
4424			}
4425			asoc->deleted_primary = net;
4426			atomic_add_int(&net->ref_count, 1);
4427			memset(&net->lastsa, 0, sizeof(net->lastsa));
4428			memset(&net->lastsv, 0, sizeof(net->lastsv));
4429			sctp_mobility_feature_on(stcb->sctp_ep,
4430			    SCTP_MOBILITY_PRIM_DELETED);
4431			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4432			    stcb->sctp_ep, stcb, NULL);
4433		}
4434out:
4435		/* Try to find a confirmed primary */
4436		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4437	}
4438	if (net == asoc->last_data_chunk_from) {
4439		/* Reset primary */
4440		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4441	}
4442	if (net == asoc->last_control_chunk_from) {
4443		/* Clear net */
4444		asoc->last_control_chunk_from = NULL;
4445	}
4446	if (net == asoc->last_net_cmt_send_started) {
4447		/* Clear net */
4448		asoc->last_net_cmt_send_started = NULL;
4449	}
4450	if (net == stcb->asoc.alternate) {
4451		sctp_free_remote_addr(stcb->asoc.alternate);
4452		stcb->asoc.alternate = NULL;
4453	}
4454	sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net,
4455	    SCTP_FROM_SCTP_PCB + SCTP_LOC_9);
4456	sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net,
4457	    SCTP_FROM_SCTP_PCB + SCTP_LOC_10);
4458	net->dest_state |= SCTP_ADDR_BEING_DELETED;
4459	sctp_free_remote_addr(net);
4460}
4461
4462/*
4463 * remove a remote endpoint address from an association, it will fail if the
4464 * address does not exist.
4465 */
4466int
4467sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4468{
4469	/*
4470	 * Here we need to remove a remote address. This is quite simple, we
4471	 * first find it in the list of address for the association
4472	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4473	 * on that item. Note we do not allow it to be removed if there are
4474	 * no other addresses.
4475	 */
4476	struct sctp_association *asoc;
4477	struct sctp_nets *net, *nnet;
4478
4479	asoc = &stcb->asoc;
4480
4481	/* locate the address */
4482	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4483		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4484			continue;
4485		}
4486		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4487		    remaddr)) {
4488			/* we found the guy */
4489			if (asoc->numnets < 2) {
4490				/* Must have at LEAST two remote addresses */
4491				return (-1);
4492			} else {
4493				sctp_remove_net(stcb, net);
4494				return (0);
4495			}
4496		}
4497	}
4498	/* not found. */
4499	return (-2);
4500}
4501
4502void
4503sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4504{
4505	struct sctpvtaghead *chain;
4506	struct sctp_tagblock *twait_block;
4507	int found = 0;
4508	int i;
4509
4510	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4511	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4512		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4513			if ((twait_block->vtag_block[i].v_tag == tag) &&
4514			    (twait_block->vtag_block[i].lport == lport) &&
4515			    (twait_block->vtag_block[i].rport == rport)) {
4516				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4517				twait_block->vtag_block[i].v_tag = 0;
4518				twait_block->vtag_block[i].lport = 0;
4519				twait_block->vtag_block[i].rport = 0;
4520				found = 1;
4521				break;
4522			}
4523		}
4524		if (found)
4525			break;
4526	}
4527}
4528
4529int
4530sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4531{
4532	struct sctpvtaghead *chain;
4533	struct sctp_tagblock *twait_block;
4534	int found = 0;
4535	int i;
4536
4537	SCTP_INP_INFO_WLOCK();
4538	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4539	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4540		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4541			if ((twait_block->vtag_block[i].v_tag == tag) &&
4542			    (twait_block->vtag_block[i].lport == lport) &&
4543			    (twait_block->vtag_block[i].rport == rport)) {
4544				found = 1;
4545				break;
4546			}
4547		}
4548		if (found)
4549			break;
4550	}
4551	SCTP_INP_INFO_WUNLOCK();
4552	return (found);
4553}
4554
4555void
4556sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4557{
4558	struct sctpvtaghead *chain;
4559	struct sctp_tagblock *twait_block;
4560	struct timeval now;
4561	int set, i;
4562
4563	if (time == 0) {
4564		/* Its disabled */
4565		return;
4566	}
4567	(void)SCTP_GETTIME_TIMEVAL(&now);
4568	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4569	set = 0;
4570	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4571		/* Block(s) present, lets find space, and expire on the fly */
4572		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4573			if ((twait_block->vtag_block[i].v_tag == 0) &&
4574			    !set) {
4575				twait_block->vtag_block[i].tv_sec_at_expire =
4576				    now.tv_sec + time;
4577				twait_block->vtag_block[i].v_tag = tag;
4578				twait_block->vtag_block[i].lport = lport;
4579				twait_block->vtag_block[i].rport = rport;
4580				set = 1;
4581			} else if ((twait_block->vtag_block[i].v_tag) &&
4582			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4583				/* Audit expires this guy */
4584				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4585				twait_block->vtag_block[i].v_tag = 0;
4586				twait_block->vtag_block[i].lport = 0;
4587				twait_block->vtag_block[i].rport = 0;
4588				if (set == 0) {
4589					/* Reuse it for my new tag */
4590					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4591					twait_block->vtag_block[i].v_tag = tag;
4592					twait_block->vtag_block[i].lport = lport;
4593					twait_block->vtag_block[i].rport = rport;
4594					set = 1;
4595				}
4596			}
4597		}
4598		if (set) {
4599			/*
4600			 * We only do up to the block where we can place our
4601			 * tag for audits
4602			 */
4603			break;
4604		}
4605	}
4606	/* Need to add a new block to chain */
4607	if (!set) {
4608		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4609		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4610		if (twait_block == NULL) {
4611			return;
4612		}
4613		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4614		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4615		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4616		twait_block->vtag_block[0].v_tag = tag;
4617		twait_block->vtag_block[0].lport = lport;
4618		twait_block->vtag_block[0].rport = rport;
4619	}
4620}
4621
4622void
4623sctp_clean_up_stream(struct sctp_tcb *stcb, struct sctp_readhead *rh)
4624{
4625	struct sctp_tmit_chunk *chk, *nchk;
4626	struct sctp_queued_to_read *control, *ncontrol;
4627
4628	TAILQ_FOREACH_SAFE(control, rh, next_instrm, ncontrol) {
4629		TAILQ_REMOVE(rh, control, next_instrm);
4630		control->on_strm_q = 0;
4631		if (control->on_read_q == 0) {
4632			sctp_free_remote_addr(control->whoFrom);
4633			if (control->data) {
4634				sctp_m_freem(control->data);
4635				control->data = NULL;
4636			}
4637		}
4638		/* Reassembly free? */
4639		TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) {
4640			TAILQ_REMOVE(&control->reasm, chk, sctp_next);
4641			if (chk->data) {
4642				sctp_m_freem(chk->data);
4643				chk->data = NULL;
4644			}
4645			if (chk->holds_key_ref)
4646				sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4647			sctp_free_remote_addr(chk->whoTo);
4648			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4649			SCTP_DECR_CHK_COUNT();
4650			/* sa_ignore FREED_MEMORY */
4651		}
4652		/*
4653		 * We don't free the address here since all the net's were
4654		 * freed above.
4655		 */
4656		if (control->on_read_q == 0) {
4657			sctp_free_a_readq(stcb, control);
4658		}
4659	}
4660}
4661
4662/*-
4663 * Free the association after un-hashing the remote port. This
4664 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4665 * expect that the input to this function IS a locked TCB.
4666 * It will return 0, if it did NOT destroy the association (instead
4667 * it unlocks it. It will return NON-zero if it either destroyed the
4668 * association OR the association is already destroyed.
4669 */
4670int
4671sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4672{
4673	int i;
4674	struct sctp_association *asoc;
4675	struct sctp_nets *net, *nnet;
4676	struct sctp_laddr *laddr, *naddr;
4677	struct sctp_tmit_chunk *chk, *nchk;
4678	struct sctp_asconf_addr *aparam, *naparam;
4679	struct sctp_asconf_ack *aack, *naack;
4680	struct sctp_stream_reset_list *strrst, *nstrrst;
4681	struct sctp_queued_to_read *sq, *nsq;
4682	struct sctp_stream_queue_pending *sp, *nsp;
4683	sctp_sharedkey_t *shared_key, *nshared_key;
4684	struct socket *so;
4685
4686	/* first, lets purge the entry from the hash table. */
4687
4688#ifdef SCTP_LOG_CLOSING
4689	sctp_log_closing(inp, stcb, 6);
4690#endif
4691	if (stcb->asoc.state == 0) {
4692#ifdef SCTP_LOG_CLOSING
4693		sctp_log_closing(inp, NULL, 7);
4694#endif
4695		/* there is no asoc, really TSNH :-0 */
4696		return (1);
4697	}
4698	SCTP_TCB_SEND_LOCK(stcb);
4699	if (stcb->asoc.alternate) {
4700		sctp_free_remote_addr(stcb->asoc.alternate);
4701		stcb->asoc.alternate = NULL;
4702	}
4703	/* TEMP CODE */
4704	if (stcb->freed_from_where == 0) {
4705		/* Only record the first place free happened from */
4706		stcb->freed_from_where = from_location;
4707	}
4708	/* TEMP CODE */
4709
4710	asoc = &stcb->asoc;
4711	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4712	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4713		/* nothing around */
4714		so = NULL;
4715	else
4716		so = inp->sctp_socket;
4717
4718	/*
4719	 * We used timer based freeing if a reader or writer is in the way.
4720	 * So we first check if we are actually being called from a timer,
4721	 * if so we abort early if a reader or writer is still in the way.
4722	 */
4723	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4724	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4725		/*
4726		 * is it the timer driving us? if so are the reader/writers
4727		 * gone?
4728		 */
4729		if (stcb->asoc.refcnt) {
4730			/* nope, reader or writer in the way */
4731			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4732			/* no asoc destroyed */
4733			SCTP_TCB_SEND_UNLOCK(stcb);
4734			SCTP_TCB_UNLOCK(stcb);
4735#ifdef SCTP_LOG_CLOSING
4736			sctp_log_closing(inp, stcb, 8);
4737#endif
4738			return (0);
4739		}
4740	}
4741	/* Now clean up any other timers */
4742	sctp_stop_association_timers(stcb, false);
4743	/* Now the read queue needs to be cleaned up (only once) */
4744	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4745		SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_ABOUT_TO_BE_FREED);
4746		SCTP_INP_READ_LOCK(inp);
4747		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4748			if (sq->stcb == stcb) {
4749				sq->do_not_ref_stcb = 1;
4750				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4751				/*
4752				 * If there is no end, there never will be
4753				 * now.
4754				 */
4755				if (sq->end_added == 0) {
4756					/* Held for PD-API clear that. */
4757					sq->pdapi_aborted = 1;
4758					sq->held_length = 0;
4759					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4760						/*
4761						 * Need to add a PD-API
4762						 * aborted indication.
4763						 * Setting the control_pdapi
4764						 * assures that it will be
4765						 * added right after this
4766						 * msg.
4767						 */
4768						uint32_t strseq;
4769
4770						stcb->asoc.control_pdapi = sq;
4771						strseq = (sq->sinfo_stream << 16) | (sq->mid & 0x0000ffff);
4772						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4773						    stcb,
4774						    SCTP_PARTIAL_DELIVERY_ABORTED,
4775						    (void *)&strseq,
4776						    SCTP_SO_LOCKED);
4777						stcb->asoc.control_pdapi = NULL;
4778					}
4779				}
4780				/* Add an end to wake them */
4781				sq->end_added = 1;
4782			}
4783		}
4784		SCTP_INP_READ_UNLOCK(inp);
4785		if (stcb->block_entry) {
4786			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4787			stcb->block_entry->error = ECONNRESET;
4788			stcb->block_entry = NULL;
4789		}
4790	}
4791	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4792		/*
4793		 * Someone holds a reference OR the socket is unaccepted
4794		 * yet.
4795		 */
4796		if ((stcb->asoc.refcnt) ||
4797		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4798		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4799			SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE);
4800			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4801		}
4802		SCTP_TCB_SEND_UNLOCK(stcb);
4803		SCTP_TCB_UNLOCK(stcb);
4804		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4805		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4806			/* nothing around */
4807			so = NULL;
4808		if (so) {
4809			/* Wake any reader/writers */
4810			sctp_sorwakeup(inp, so);
4811			sctp_sowwakeup(inp, so);
4812		}
4813
4814#ifdef SCTP_LOG_CLOSING
4815		sctp_log_closing(inp, stcb, 9);
4816#endif
4817		/* no asoc destroyed */
4818		return (0);
4819	}
4820#ifdef SCTP_LOG_CLOSING
4821	sctp_log_closing(inp, stcb, 10);
4822#endif
4823	/*
4824	 * When I reach here, no others want to kill the assoc yet.. and I
4825	 * own the lock. Now its possible an abort comes in when I do the
4826	 * lock exchange below to grab all the locks to do the final take
4827	 * out. to prevent this we increment the count, which will start a
4828	 * timer and blow out above thus assuring us that we hold exclusive
4829	 * killing of the asoc. Note that after getting back the TCB lock we
4830	 * will go ahead and increment the counter back up and stop any
4831	 * timer a passing stranger may have started :-S
4832	 */
4833	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4834		atomic_add_int(&stcb->asoc.refcnt, 1);
4835
4836		SCTP_TCB_SEND_UNLOCK(stcb);
4837		SCTP_TCB_UNLOCK(stcb);
4838		SCTP_INP_INFO_WLOCK();
4839		SCTP_INP_WLOCK(inp);
4840		SCTP_TCB_LOCK(stcb);
4841		SCTP_TCB_SEND_LOCK(stcb);
4842	}
4843	/* Double check the GONE flag */
4844	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4845	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4846		/* nothing around */
4847		so = NULL;
4848
4849	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4850	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4851		/*
4852		 * For TCP type we need special handling when we are
4853		 * connected. We also include the peel'ed off ones to.
4854		 */
4855		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4856			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4857			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4858			if (so) {
4859				SOCKBUF_LOCK(&so->so_rcv);
4860				so->so_state &= ~(SS_ISCONNECTING |
4861				    SS_ISDISCONNECTING |
4862				    SS_ISCONFIRMING |
4863				    SS_ISCONNECTED);
4864				so->so_state |= SS_ISDISCONNECTED;
4865				socantrcvmore_locked(so);
4866				socantsendmore(so);
4867				sctp_sowwakeup(inp, so);
4868				sctp_sorwakeup(inp, so);
4869				SCTP_SOWAKEUP(so);
4870			}
4871		}
4872	}
4873
4874	/*
4875	 * Make it invalid too, that way if its about to run it will abort
4876	 * and return.
4877	 */
4878	/* re-increment the lock */
4879	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4880		atomic_add_int(&stcb->asoc.refcnt, -1);
4881	}
4882	if (stcb->asoc.refcnt) {
4883		SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE);
4884		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4885		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4886			SCTP_INP_INFO_WUNLOCK();
4887			SCTP_INP_WUNLOCK(inp);
4888		}
4889		SCTP_TCB_SEND_UNLOCK(stcb);
4890		SCTP_TCB_UNLOCK(stcb);
4891		return (0);
4892	}
4893	asoc->state = 0;
4894	if (inp->sctp_tcbhash) {
4895		LIST_REMOVE(stcb, sctp_tcbhash);
4896	}
4897	if (stcb->asoc.in_asocid_hash) {
4898		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4899	}
4900	/* Now lets remove it from the list of ALL associations in the EP */
4901	LIST_REMOVE(stcb, sctp_tcblist);
4902	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4903		SCTP_INP_INCR_REF(inp);
4904		SCTP_INP_WUNLOCK(inp);
4905	}
4906	/* pull from vtag hash */
4907	LIST_REMOVE(stcb, sctp_asocs);
4908	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4909	    inp->sctp_lport, stcb->rport);
4910
4911	/*
4912	 * Now restop the timers to be sure this is paranoia at is finest!
4913	 */
4914	sctp_stop_association_timers(stcb, true);
4915
4916	/*
4917	 * The chunk lists and such SHOULD be empty but we check them just
4918	 * in case.
4919	 */
4920	/* anything on the wheel needs to be removed */
4921	for (i = 0; i < asoc->streamoutcnt; i++) {
4922		struct sctp_stream_out *outs;
4923
4924		outs = &asoc->strmout[i];
4925		/* now clean up any chunks here */
4926		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4927			atomic_subtract_int(&asoc->stream_queue_cnt, 1);
4928			TAILQ_REMOVE(&outs->outqueue, sp, next);
4929			stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, outs, sp, 1);
4930			sctp_free_spbufspace(stcb, asoc, sp);
4931			if (sp->data) {
4932				if (so) {
4933					/* Still an open socket - report */
4934					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4935					    0, (void *)sp, SCTP_SO_LOCKED);
4936				}
4937				if (sp->data) {
4938					sctp_m_freem(sp->data);
4939					sp->data = NULL;
4940					sp->tail_mbuf = NULL;
4941					sp->length = 0;
4942				}
4943			}
4944			if (sp->net) {
4945				sctp_free_remote_addr(sp->net);
4946				sp->net = NULL;
4947			}
4948			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4949		}
4950	}
4951	/* sa_ignore FREED_MEMORY */
4952	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4953		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4954		SCTP_FREE(strrst, SCTP_M_STRESET);
4955	}
4956	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4957		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4958		if (sq->data) {
4959			sctp_m_freem(sq->data);
4960			sq->data = NULL;
4961		}
4962		sctp_free_remote_addr(sq->whoFrom);
4963		sq->whoFrom = NULL;
4964		sq->stcb = NULL;
4965		/* Free the ctl entry */
4966		sctp_free_a_readq(stcb, sq);
4967		/* sa_ignore FREED_MEMORY */
4968	}
4969	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4970		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4971		if (chk->data) {
4972			sctp_m_freem(chk->data);
4973			chk->data = NULL;
4974		}
4975		if (chk->holds_key_ref)
4976			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4977		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4978		SCTP_DECR_CHK_COUNT();
4979		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4980		asoc->free_chunk_cnt--;
4981		/* sa_ignore FREED_MEMORY */
4982	}
4983	/* pending send queue SHOULD be empty */
4984	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4985		if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) {
4986			asoc->strmout[chk->rec.data.sid].chunks_on_queues--;
4987#ifdef INVARIANTS
4988		} else {
4989			panic("No chunks on the queues for sid %u.", chk->rec.data.sid);
4990#endif
4991		}
4992		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4993		if (chk->data) {
4994			if (so) {
4995				/* Still a socket? */
4996				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
4997				    0, chk, SCTP_SO_LOCKED);
4998			}
4999			if (chk->data) {
5000				sctp_m_freem(chk->data);
5001				chk->data = NULL;
5002			}
5003		}
5004		if (chk->holds_key_ref)
5005			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5006		if (chk->whoTo) {
5007			sctp_free_remote_addr(chk->whoTo);
5008			chk->whoTo = NULL;
5009		}
5010		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5011		SCTP_DECR_CHK_COUNT();
5012		/* sa_ignore FREED_MEMORY */
5013	}
5014	/* sent queue SHOULD be empty */
5015	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5016		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5017			if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) {
5018				asoc->strmout[chk->rec.data.sid].chunks_on_queues--;
5019#ifdef INVARIANTS
5020			} else {
5021				panic("No chunks on the queues for sid %u.", chk->rec.data.sid);
5022#endif
5023			}
5024		}
5025		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5026		if (chk->data) {
5027			if (so) {
5028				/* Still a socket? */
5029				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5030				    0, chk, SCTP_SO_LOCKED);
5031			}
5032			if (chk->data) {
5033				sctp_m_freem(chk->data);
5034				chk->data = NULL;
5035			}
5036		}
5037		if (chk->holds_key_ref)
5038			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5039		sctp_free_remote_addr(chk->whoTo);
5040		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5041		SCTP_DECR_CHK_COUNT();
5042		/* sa_ignore FREED_MEMORY */
5043	}
5044#ifdef INVARIANTS
5045	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5046		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5047			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5048		}
5049	}
5050#endif
5051	/* control queue MAY not be empty */
5052	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5053		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5054		if (chk->data) {
5055			sctp_m_freem(chk->data);
5056			chk->data = NULL;
5057		}
5058		if (chk->holds_key_ref)
5059			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5060		sctp_free_remote_addr(chk->whoTo);
5061		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5062		SCTP_DECR_CHK_COUNT();
5063		/* sa_ignore FREED_MEMORY */
5064	}
5065	/* ASCONF queue MAY not be empty */
5066	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5067		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5068		if (chk->data) {
5069			sctp_m_freem(chk->data);
5070			chk->data = NULL;
5071		}
5072		if (chk->holds_key_ref)
5073			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5074		sctp_free_remote_addr(chk->whoTo);
5075		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5076		SCTP_DECR_CHK_COUNT();
5077		/* sa_ignore FREED_MEMORY */
5078	}
5079	if (asoc->mapping_array) {
5080		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5081		asoc->mapping_array = NULL;
5082	}
5083	if (asoc->nr_mapping_array) {
5084		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5085		asoc->nr_mapping_array = NULL;
5086	}
5087	/* the stream outs */
5088	if (asoc->strmout) {
5089		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5090		asoc->strmout = NULL;
5091	}
5092	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5093	if (asoc->strmin) {
5094		for (i = 0; i < asoc->streamincnt; i++) {
5095			sctp_clean_up_stream(stcb, &asoc->strmin[i].inqueue);
5096			sctp_clean_up_stream(stcb, &asoc->strmin[i].uno_inqueue);
5097		}
5098		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5099		asoc->strmin = NULL;
5100	}
5101	asoc->streamincnt = 0;
5102	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5103#ifdef INVARIANTS
5104		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5105			panic("no net's left alloc'ed, or list points to itself");
5106		}
5107#endif
5108		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5109		sctp_free_remote_addr(net);
5110	}
5111	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5112		/* sa_ignore FREED_MEMORY */
5113		sctp_remove_laddr(laddr);
5114	}
5115
5116	/* pending asconf (address) parameters */
5117	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5118		/* sa_ignore FREED_MEMORY */
5119		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5120		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5121	}
5122	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5123		/* sa_ignore FREED_MEMORY */
5124		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5125		if (aack->data != NULL) {
5126			sctp_m_freem(aack->data);
5127		}
5128		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5129	}
5130	/* clean up auth stuff */
5131	if (asoc->local_hmacs)
5132		sctp_free_hmaclist(asoc->local_hmacs);
5133	if (asoc->peer_hmacs)
5134		sctp_free_hmaclist(asoc->peer_hmacs);
5135
5136	if (asoc->local_auth_chunks)
5137		sctp_free_chunklist(asoc->local_auth_chunks);
5138	if (asoc->peer_auth_chunks)
5139		sctp_free_chunklist(asoc->peer_auth_chunks);
5140
5141	sctp_free_authinfo(&asoc->authinfo);
5142
5143	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5144		LIST_REMOVE(shared_key, next);
5145		sctp_free_sharedkey(shared_key);
5146		/* sa_ignore FREED_MEMORY */
5147	}
5148
5149	/* Insert new items here :> */
5150
5151	/* Get rid of LOCK */
5152	SCTP_TCB_SEND_UNLOCK(stcb);
5153	SCTP_TCB_UNLOCK(stcb);
5154	SCTP_TCB_LOCK_DESTROY(stcb);
5155	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5156	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5157		SCTP_INP_INFO_WUNLOCK();
5158		SCTP_INP_RLOCK(inp);
5159	}
5160#ifdef SCTP_TRACK_FREED_ASOCS
5161	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5162		/* now clean up the tasoc itself */
5163		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5164		SCTP_DECR_ASOC_COUNT();
5165	} else {
5166		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5167	}
5168#else
5169	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5170	SCTP_DECR_ASOC_COUNT();
5171#endif
5172	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5173		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5174			/*
5175			 * If its NOT the inp_free calling us AND sctp_close
5176			 * as been called, we call back...
5177			 */
5178			SCTP_INP_RUNLOCK(inp);
5179			/*
5180			 * This will start the kill timer (if we are the
5181			 * last one) since we hold an increment yet. But
5182			 * this is the only safe way to do this since
5183			 * otherwise if the socket closes at the same time
5184			 * we are here we might collide in the cleanup.
5185			 */
5186			sctp_inpcb_free(inp,
5187			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5188			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5189			SCTP_INP_DECR_REF(inp);
5190		} else {
5191			/* The socket is still open. */
5192			SCTP_INP_DECR_REF(inp);
5193			SCTP_INP_RUNLOCK(inp);
5194		}
5195	}
5196	/* destroyed the asoc */
5197#ifdef SCTP_LOG_CLOSING
5198	sctp_log_closing(inp, NULL, 11);
5199#endif
5200	return (1);
5201}
5202
5203/*
5204 * determine if a destination is "reachable" based upon the addresses bound
5205 * to the current endpoint (e.g. only v4 or v6 currently bound)
5206 */
5207/*
5208 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5209 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5210 * types bound as its endpoint
5211 */
5212int
5213sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5214{
5215	struct sctp_inpcb *inp;
5216	int answer;
5217
5218	/*
5219	 * No locks here, the TCB, in all cases is already locked and an
5220	 * assoc is up. There is either a INP lock by the caller applied (in
5221	 * asconf case when deleting an address) or NOT in the HB case,
5222	 * however if HB then the INP increment is up and the INP will not
5223	 * be removed (on top of the fact that we have a TCB lock). So we
5224	 * only want to read the sctp_flags, which is either bound-all or
5225	 * not.. no protection needed since once an assoc is up you can't be
5226	 * changing your binding.
5227	 */
5228	inp = stcb->sctp_ep;
5229	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5230		/* if bound all, destination is not restricted */
5231		/*
5232		 * RRS: Question during lock work: Is this correct? If you
5233		 * are bound-all you still might need to obey the V4--V6
5234		 * flags??? IMO this bound-all stuff needs to be removed!
5235		 */
5236		return (1);
5237	}
5238	/* NOTE: all "scope" checks are done when local addresses are added */
5239	switch (destaddr->sa_family) {
5240#ifdef INET6
5241	case AF_INET6:
5242		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5243		break;
5244#endif
5245#ifdef INET
5246	case AF_INET:
5247		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5248		break;
5249#endif
5250	default:
5251		/* invalid family, so it's unreachable */
5252		answer = 0;
5253		break;
5254	}
5255	return (answer);
5256}
5257
5258/*
5259 * update the inp_vflags on an endpoint
5260 */
5261static void
5262sctp_update_ep_vflag(struct sctp_inpcb *inp)
5263{
5264	struct sctp_laddr *laddr;
5265
5266	/* first clear the flag */
5267	inp->ip_inp.inp.inp_vflag = 0;
5268	/* set the flag based on addresses on the ep list */
5269	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5270		if (laddr->ifa == NULL) {
5271			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5272			    __func__);
5273			continue;
5274		}
5275
5276		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5277			continue;
5278		}
5279		switch (laddr->ifa->address.sa.sa_family) {
5280#ifdef INET6
5281		case AF_INET6:
5282			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5283			break;
5284#endif
5285#ifdef INET
5286		case AF_INET:
5287			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5288			break;
5289#endif
5290		default:
5291			break;
5292		}
5293	}
5294}
5295
5296/*
5297 * Add the address to the endpoint local address list There is nothing to be
5298 * done if we are bound to all addresses
5299 */
5300void
5301sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5302{
5303	struct sctp_laddr *laddr;
5304	struct sctp_tcb *stcb;
5305	int fnd, error = 0;
5306
5307	fnd = 0;
5308
5309	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5310		/* You are already bound to all. You have it already */
5311		return;
5312	}
5313#ifdef INET6
5314	if (ifa->address.sa.sa_family == AF_INET6) {
5315		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5316			/* Can't bind a non-useable addr. */
5317			return;
5318		}
5319	}
5320#endif
5321	/* first, is it already present? */
5322	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5323		if (laddr->ifa == ifa) {
5324			fnd = 1;
5325			break;
5326		}
5327	}
5328
5329	if (fnd == 0) {
5330		/* Not in the ep list */
5331		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5332		if (error != 0)
5333			return;
5334		inp->laddr_count++;
5335		/* update inp_vflag flags */
5336		switch (ifa->address.sa.sa_family) {
5337#ifdef INET6
5338		case AF_INET6:
5339			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5340			break;
5341#endif
5342#ifdef INET
5343		case AF_INET:
5344			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5345			break;
5346#endif
5347		default:
5348			break;
5349		}
5350		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5351			sctp_add_local_addr_restricted(stcb, ifa);
5352		}
5353	}
5354	return;
5355}
5356
5357/*
5358 * select a new (hopefully reachable) destination net (should only be used
5359 * when we deleted an ep addr that is the only usable source address to reach
5360 * the destination net)
5361 */
5362static void
5363sctp_select_primary_destination(struct sctp_tcb *stcb)
5364{
5365	struct sctp_nets *net;
5366
5367	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5368		/* for now, we'll just pick the first reachable one we find */
5369		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5370			continue;
5371		if (sctp_destination_is_reachable(stcb,
5372		    (struct sockaddr *)&net->ro._l_addr)) {
5373			/* found a reachable destination */
5374			stcb->asoc.primary_destination = net;
5375		}
5376	}
5377	/* I can't there from here! ...we're gonna die shortly... */
5378}
5379
5380/*
5381 * Delete the address from the endpoint local address list. There is nothing
5382 * to be done if we are bound to all addresses
5383 */
5384void
5385sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5386{
5387	struct sctp_laddr *laddr;
5388	int fnd;
5389
5390	fnd = 0;
5391	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5392		/* You are already bound to all. You have it already */
5393		return;
5394	}
5395	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5396		if (laddr->ifa == ifa) {
5397			fnd = 1;
5398			break;
5399		}
5400	}
5401	if (fnd && (inp->laddr_count < 2)) {
5402		/* can't delete unless there are at LEAST 2 addresses */
5403		return;
5404	}
5405	if (fnd) {
5406		/*
5407		 * clean up any use of this address go through our
5408		 * associations and clear any last_used_address that match
5409		 * this one for each assoc, see if a new primary_destination
5410		 * is needed
5411		 */
5412		struct sctp_tcb *stcb;
5413
5414		/* clean up "next_addr_touse" */
5415		if (inp->next_addr_touse == laddr)
5416			/* delete this address */
5417			inp->next_addr_touse = NULL;
5418
5419		/* clean up "last_used_address" */
5420		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5421			struct sctp_nets *net;
5422
5423			SCTP_TCB_LOCK(stcb);
5424			if (stcb->asoc.last_used_address == laddr)
5425				/* delete this address */
5426				stcb->asoc.last_used_address = NULL;
5427			/*
5428			 * Now spin through all the nets and purge any ref
5429			 * to laddr
5430			 */
5431			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5432				if (net->ro._s_addr == laddr->ifa) {
5433					/* Yep, purge src address selected */
5434					RO_NHFREE(&net->ro);
5435					sctp_free_ifa(net->ro._s_addr);
5436					net->ro._s_addr = NULL;
5437					net->src_addr_selected = 0;
5438				}
5439			}
5440			SCTP_TCB_UNLOCK(stcb);
5441		}		/* for each tcb */
5442		/* remove it from the ep list */
5443		sctp_remove_laddr(laddr);
5444		inp->laddr_count--;
5445		/* update inp_vflag flags */
5446		sctp_update_ep_vflag(inp);
5447	}
5448	return;
5449}
5450
5451/*
5452 * Add the address to the TCB local address restricted list.
5453 * This is a "pending" address list (eg. addresses waiting for an
5454 * ASCONF-ACK response) and cannot be used as a valid source address.
5455 */
5456void
5457sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5458{
5459	struct sctp_laddr *laddr;
5460	struct sctpladdr *list;
5461
5462	/*
5463	 * Assumes TCB is locked.. and possibly the INP. May need to
5464	 * confirm/fix that if we need it and is not the case.
5465	 */
5466	list = &stcb->asoc.sctp_restricted_addrs;
5467
5468#ifdef INET6
5469	if (ifa->address.sa.sa_family == AF_INET6) {
5470		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5471			/* Can't bind a non-existent addr. */
5472			return;
5473		}
5474	}
5475#endif
5476	/* does the address already exist? */
5477	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5478		if (laddr->ifa == ifa) {
5479			return;
5480		}
5481	}
5482
5483	/* add to the list */
5484	(void)sctp_insert_laddr(list, ifa, 0);
5485	return;
5486}
5487
5488/*
5489 * Remove a local address from the TCB local address restricted list
5490 */
5491void
5492sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5493{
5494	struct sctp_inpcb *inp;
5495	struct sctp_laddr *laddr;
5496
5497	/*
5498	 * This is called by asconf work. It is assumed that a) The TCB is
5499	 * locked and b) The INP is locked. This is true in as much as I can
5500	 * trace through the entry asconf code where I did these locks.
5501	 * Again, the ASCONF code is a bit different in that it does lock
5502	 * the INP during its work often times. This must be since we don't
5503	 * want other proc's looking up things while what they are looking
5504	 * up is changing :-D
5505	 */
5506
5507	inp = stcb->sctp_ep;
5508	/* if subset bound and don't allow ASCONF's, can't delete last */
5509	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5510	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5511		if (stcb->sctp_ep->laddr_count < 2) {
5512			/* can't delete last address */
5513			return;
5514		}
5515	}
5516	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5517		/* remove the address if it exists */
5518		if (laddr->ifa == NULL)
5519			continue;
5520		if (laddr->ifa == ifa) {
5521			sctp_remove_laddr(laddr);
5522			return;
5523		}
5524	}
5525
5526	/* address not found! */
5527	return;
5528}
5529
5530/* sysctl */
5531static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5532static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5533
5534#if defined(SCTP_MCORE_INPUT) && defined(SMP)
5535struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5536int *sctp_cpuarry = NULL;
5537
5538void
5539sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5540{
5541	/* Queue a packet to a processor for the specified core */
5542	struct sctp_mcore_queue *qent;
5543	struct sctp_mcore_ctrl *wkq;
5544	int need_wake = 0;
5545
5546	if (sctp_mcore_workers == NULL) {
5547		/* Something went way bad during setup */
5548		sctp_input_with_port(m, off, 0);
5549		return;
5550	}
5551	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5552	    (sizeof(struct sctp_mcore_queue)),
5553	    SCTP_M_MCORE);
5554	if (qent == NULL) {
5555		/* This is trouble  */
5556		sctp_input_with_port(m, off, 0);
5557		return;
5558	}
5559	qent->vn = curvnet;
5560	qent->m = m;
5561	qent->off = off;
5562	qent->v6 = 0;
5563	wkq = &sctp_mcore_workers[cpu_to_use];
5564	SCTP_MCORE_QLOCK(wkq);
5565
5566	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5567	if (wkq->running == 0) {
5568		need_wake = 1;
5569	}
5570	SCTP_MCORE_QUNLOCK(wkq);
5571	if (need_wake) {
5572		wakeup(&wkq->running);
5573	}
5574}
5575
5576static void
5577sctp_mcore_thread(void *arg)
5578{
5579
5580	struct sctp_mcore_ctrl *wkq;
5581	struct sctp_mcore_queue *qent;
5582
5583	wkq = (struct sctp_mcore_ctrl *)arg;
5584	struct mbuf *m;
5585	int off, v6;
5586
5587	/* Wait for first tickle */
5588	SCTP_MCORE_LOCK(wkq);
5589	wkq->running = 0;
5590	msleep(&wkq->running,
5591	    &wkq->core_mtx,
5592	    0, "wait for pkt", 0);
5593	SCTP_MCORE_UNLOCK(wkq);
5594
5595	/* Bind to our cpu */
5596	thread_lock(curthread);
5597	sched_bind(curthread, wkq->cpuid);
5598	thread_unlock(curthread);
5599
5600	/* Now lets start working */
5601	SCTP_MCORE_LOCK(wkq);
5602	/* Now grab lock and go */
5603	for (;;) {
5604		SCTP_MCORE_QLOCK(wkq);
5605skip_sleep:
5606		wkq->running = 1;
5607		qent = TAILQ_FIRST(&wkq->que);
5608		if (qent) {
5609			TAILQ_REMOVE(&wkq->que, qent, next);
5610			SCTP_MCORE_QUNLOCK(wkq);
5611			CURVNET_SET(qent->vn);
5612			m = qent->m;
5613			off = qent->off;
5614			v6 = qent->v6;
5615			SCTP_FREE(qent, SCTP_M_MCORE);
5616			if (v6 == 0) {
5617				sctp_input_with_port(m, off, 0);
5618			} else {
5619				SCTP_PRINTF("V6 not yet supported\n");
5620				sctp_m_freem(m);
5621			}
5622			CURVNET_RESTORE();
5623			SCTP_MCORE_QLOCK(wkq);
5624		}
5625		wkq->running = 0;
5626		if (!TAILQ_EMPTY(&wkq->que)) {
5627			goto skip_sleep;
5628		}
5629		SCTP_MCORE_QUNLOCK(wkq);
5630		msleep(&wkq->running,
5631		    &wkq->core_mtx,
5632		    0, "wait for pkt", 0);
5633	}
5634}
5635
5636static void
5637sctp_startup_mcore_threads(void)
5638{
5639	int i, cpu;
5640
5641	if (mp_ncpus == 1)
5642		return;
5643
5644	if (sctp_mcore_workers != NULL) {
5645		/*
5646		 * Already been here in some previous vnet?
5647		 */
5648		return;
5649	}
5650	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5651	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5652	    SCTP_M_MCORE);
5653	if (sctp_mcore_workers == NULL) {
5654		/* TSNH I hope */
5655		return;
5656	}
5657	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5658	    sizeof(struct sctp_mcore_ctrl)));
5659	/* Init the structures */
5660	for (i = 0; i <= mp_maxid; i++) {
5661		TAILQ_INIT(&sctp_mcore_workers[i].que);
5662		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5663		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5664		sctp_mcore_workers[i].cpuid = i;
5665	}
5666	if (sctp_cpuarry == NULL) {
5667		SCTP_MALLOC(sctp_cpuarry, int *,
5668		    (mp_ncpus * sizeof(int)),
5669		    SCTP_M_MCORE);
5670		i = 0;
5671		CPU_FOREACH(cpu) {
5672			sctp_cpuarry[i] = cpu;
5673			i++;
5674		}
5675	}
5676	/* Now start them all */
5677	CPU_FOREACH(cpu) {
5678		(void)kproc_create(sctp_mcore_thread,
5679		    (void *)&sctp_mcore_workers[cpu],
5680		    &sctp_mcore_workers[cpu].thread_proc,
5681		    0,
5682		    SCTP_KTHREAD_PAGES,
5683		    SCTP_MCORE_NAME);
5684	}
5685}
5686#endif
5687
5688void
5689sctp_pcb_init(void)
5690{
5691	/*
5692	 * SCTP initialization for the PCB structures should be called by
5693	 * the sctp_init() function.
5694	 */
5695	int i;
5696	struct timeval tv;
5697
5698	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5699		/* error I was called twice */
5700		return;
5701	}
5702	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5703
5704#if defined(SCTP_LOCAL_TRACE_BUF)
5705	memset(&SCTP_BASE_SYSCTL(sctp_log), 0, sizeof(struct sctp_log));
5706#endif
5707#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5708	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5709	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5710	    SCTP_M_MCORE);
5711#endif
5712	(void)SCTP_GETTIME_TIMEVAL(&tv);
5713#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5714	memset(SCTP_BASE_STATS, 0, sizeof(struct sctpstat) * (mp_maxid + 1));
5715	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t)tv.tv_sec;
5716	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t)tv.tv_usec;
5717#else
5718	memset(&SCTP_BASE_STATS, 0, sizeof(struct sctpstat));
5719	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t)tv.tv_sec;
5720	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t)tv.tv_usec;
5721#endif
5722	/* init the empty list of (All) Endpoints */
5723	LIST_INIT(&SCTP_BASE_INFO(listhead));
5724
5725	/* init the hash table of endpoints */
5726	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5727	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5728	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5729	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5730	    &SCTP_BASE_INFO(hashasocmark));
5731	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5732	    &SCTP_BASE_INFO(hashmark));
5733	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5734	    &SCTP_BASE_INFO(hashtcpmark));
5735	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5736	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5737	    &SCTP_BASE_INFO(hashvrfmark));
5738
5739	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5740	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5741	/* init the zones */
5742	/*
5743	 * FIX ME: Should check for NULL returns, but if it does fail we are
5744	 * doomed to panic anyways... add later maybe.
5745	 */
5746	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5747	    sizeof(struct sctp_inpcb), maxsockets);
5748
5749	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5750	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5751
5752	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5753	    sizeof(struct sctp_laddr),
5754	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5755
5756	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5757	    sizeof(struct sctp_nets),
5758	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5759
5760	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5761	    sizeof(struct sctp_tmit_chunk),
5762	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5763
5764	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5765	    sizeof(struct sctp_queued_to_read),
5766	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5767
5768	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5769	    sizeof(struct sctp_stream_queue_pending),
5770	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5771
5772	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5773	    sizeof(struct sctp_asconf),
5774	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5775
5776	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5777	    sizeof(struct sctp_asconf_ack),
5778	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5779
5780	/* Master Lock INIT for info structure */
5781	SCTP_INP_INFO_LOCK_INIT();
5782	SCTP_STATLOG_INIT_LOCK();
5783
5784	SCTP_IPI_COUNT_INIT();
5785	SCTP_IPI_ADDR_INIT();
5786#ifdef SCTP_PACKET_LOGGING
5787	SCTP_IP_PKTLOG_INIT();
5788#endif
5789	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5790
5791	SCTP_WQ_ADDR_INIT();
5792	/* not sure if we need all the counts */
5793	SCTP_BASE_INFO(ipi_count_ep) = 0;
5794	/* assoc/tcb zone info */
5795	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5796	/* local addrlist zone info */
5797	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5798	/* remote addrlist zone info */
5799	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5800	/* chunk info */
5801	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5802
5803	/* socket queue zone info */
5804	SCTP_BASE_INFO(ipi_count_readq) = 0;
5805
5806	/* stream out queue cont */
5807	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5808
5809	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5810	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5811
5812	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5813
5814	/* Init the TIMEWAIT list */
5815	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5816		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5817	}
5818	sctp_startup_iterator();
5819
5820#if defined(SCTP_MCORE_INPUT) && defined(SMP)
5821	sctp_startup_mcore_threads();
5822#endif
5823
5824	/*
5825	 * INIT the default VRF which for BSD is the only one, other O/S's
5826	 * may have more. But initially they must start with one and then
5827	 * add the VRF's as addresses are added.
5828	 */
5829	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5830}
5831
5832/*
5833 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5834 */
5835void
5836sctp_pcb_finish(void)
5837{
5838	struct sctp_vrflist *vrf_bucket;
5839	struct sctp_vrf *vrf, *nvrf;
5840	struct sctp_ifn *ifn, *nifn;
5841	struct sctp_ifa *ifa, *nifa;
5842	struct sctpvtaghead *chain;
5843	struct sctp_tagblock *twait_block, *prev_twait_block;
5844	struct sctp_laddr *wi, *nwi;
5845	int i;
5846	struct sctp_iterator *it, *nit;
5847
5848	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
5849		SCTP_PRINTF("%s: race condition on teardown.\n", __func__);
5850		return;
5851	}
5852	SCTP_BASE_VAR(sctp_pcb_initialized) = 0;
5853	/*
5854	 * In FreeBSD the iterator thread never exits but we do clean up.
5855	 * The only way FreeBSD reaches here is if we have VRF's but we
5856	 * still add the ifdef to make it compile on old versions.
5857	 */
5858retry:
5859	SCTP_IPI_ITERATOR_WQ_LOCK();
5860	/*
5861	 * sctp_iterator_worker() might be working on an it entry without
5862	 * holding the lock.  We won't find it on the list either and
5863	 * continue and free/destroy it.  While holding the lock, spin, to
5864	 * avoid the race condition as sctp_iterator_worker() will have to
5865	 * wait to re-acquire the lock.
5866	 */
5867	if (sctp_it_ctl.iterator_running != 0 || sctp_it_ctl.cur_it != NULL) {
5868		SCTP_IPI_ITERATOR_WQ_UNLOCK();
5869		SCTP_PRINTF("%s: Iterator running while we held the lock. Retry. "
5870		    "cur_it=%p\n", __func__, sctp_it_ctl.cur_it);
5871		DELAY(10);
5872		goto retry;
5873	}
5874	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5875		if (it->vn != curvnet) {
5876			continue;
5877		}
5878		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5879		if (it->function_atend != NULL) {
5880			(*it->function_atend) (it->pointer, it->val);
5881		}
5882		SCTP_FREE(it, SCTP_M_ITER);
5883	}
5884	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5885	SCTP_ITERATOR_LOCK();
5886	if ((sctp_it_ctl.cur_it) &&
5887	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5888		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5889	}
5890	SCTP_ITERATOR_UNLOCK();
5891	SCTP_OS_TIMER_STOP_DRAIN(&SCTP_BASE_INFO(addr_wq_timer.timer));
5892	SCTP_WQ_ADDR_LOCK();
5893	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5894		LIST_REMOVE(wi, sctp_nxt_addr);
5895		SCTP_DECR_LADDR_COUNT();
5896		if (wi->action == SCTP_DEL_IP_ADDRESS) {
5897			SCTP_FREE(wi->ifa, SCTP_M_IFA);
5898		}
5899		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5900	}
5901	SCTP_WQ_ADDR_UNLOCK();
5902
5903	/*
5904	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5905	 * destroyed first).
5906	 */
5907	SCTP_IPI_ADDR_WLOCK();
5908	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5909	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5910		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5911			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5912				/* free the ifa */
5913				LIST_REMOVE(ifa, next_bucket);
5914				LIST_REMOVE(ifa, next_ifa);
5915				SCTP_FREE(ifa, SCTP_M_IFA);
5916			}
5917			/* free the ifn */
5918			LIST_REMOVE(ifn, next_bucket);
5919			LIST_REMOVE(ifn, next_ifn);
5920			SCTP_FREE(ifn, SCTP_M_IFN);
5921		}
5922		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5923		/* free the vrf */
5924		LIST_REMOVE(vrf, next_vrf);
5925		SCTP_FREE(vrf, SCTP_M_VRF);
5926	}
5927	SCTP_IPI_ADDR_WUNLOCK();
5928	/* free the vrf hashes */
5929	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5930	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5931
5932	/*
5933	 * free the TIMEWAIT list elements malloc'd in the function
5934	 * sctp_add_vtag_to_timewait()...
5935	 */
5936	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5937		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5938		if (!LIST_EMPTY(chain)) {
5939			prev_twait_block = NULL;
5940			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5941				if (prev_twait_block) {
5942					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5943				}
5944				prev_twait_block = twait_block;
5945			}
5946			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5947		}
5948	}
5949
5950	/* free the locks and mutexes */
5951#ifdef SCTP_PACKET_LOGGING
5952	SCTP_IP_PKTLOG_DESTROY();
5953#endif
5954	SCTP_IPI_ADDR_DESTROY();
5955	SCTP_STATLOG_DESTROY();
5956	SCTP_INP_INFO_LOCK_DESTROY();
5957
5958	SCTP_WQ_ADDR_DESTROY();
5959
5960	/* Get rid of other stuff too. */
5961	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5962		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5963	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5964		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
5965	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
5966		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
5967
5968	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
5969	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
5970	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
5971	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
5972	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
5973	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
5974	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
5975	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
5976	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
5977#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5978	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
5979#endif
5980}
5981
5982int
5983sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
5984    int offset, int limit,
5985    struct sockaddr *src, struct sockaddr *dst,
5986    struct sockaddr *altsa, uint16_t port)
5987{
5988	/*
5989	 * grub through the INIT pulling addresses and loading them to the
5990	 * nets structure in the asoc. The from address in the mbuf should
5991	 * also be loaded (if it is not already). This routine can be called
5992	 * with either INIT or INIT-ACK's as long as the m points to the IP
5993	 * packet and the offset points to the beginning of the parameters.
5994	 */
5995	struct sctp_inpcb *inp;
5996	struct sctp_nets *net, *nnet, *net_tmp;
5997	struct sctp_paramhdr *phdr, param_buf;
5998	struct sctp_tcb *stcb_tmp;
5999	uint16_t ptype, plen;
6000	struct sockaddr *sa;
6001	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6002	struct sctp_auth_random *p_random = NULL;
6003	uint16_t random_len = 0;
6004	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6005	struct sctp_auth_hmac_algo *hmacs = NULL;
6006	uint16_t hmacs_len = 0;
6007	uint8_t saw_asconf = 0;
6008	uint8_t saw_asconf_ack = 0;
6009	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6010	struct sctp_auth_chunk_list *chunks = NULL;
6011	uint16_t num_chunks = 0;
6012	sctp_key_t *new_key;
6013	uint32_t keylen;
6014	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6015	uint8_t peer_supports_ecn;
6016	uint8_t peer_supports_prsctp;
6017	uint8_t peer_supports_auth;
6018	uint8_t peer_supports_asconf;
6019	uint8_t peer_supports_asconf_ack;
6020	uint8_t peer_supports_reconfig;
6021	uint8_t peer_supports_nrsack;
6022	uint8_t peer_supports_pktdrop;
6023	uint8_t peer_supports_idata;
6024#ifdef INET
6025	struct sockaddr_in sin;
6026#endif
6027#ifdef INET6
6028	struct sockaddr_in6 sin6;
6029#endif
6030
6031	/* First get the destination address setup too. */
6032#ifdef INET
6033	memset(&sin, 0, sizeof(sin));
6034	sin.sin_family = AF_INET;
6035	sin.sin_len = sizeof(sin);
6036	sin.sin_port = stcb->rport;
6037#endif
6038#ifdef INET6
6039	memset(&sin6, 0, sizeof(sin6));
6040	sin6.sin6_family = AF_INET6;
6041	sin6.sin6_len = sizeof(struct sockaddr_in6);
6042	sin6.sin6_port = stcb->rport;
6043#endif
6044	if (altsa) {
6045		sa = altsa;
6046	} else {
6047		sa = src;
6048	}
6049	peer_supports_idata = 0;
6050	peer_supports_ecn = 0;
6051	peer_supports_prsctp = 0;
6052	peer_supports_auth = 0;
6053	peer_supports_asconf = 0;
6054	peer_supports_asconf_ack = 0;
6055	peer_supports_reconfig = 0;
6056	peer_supports_nrsack = 0;
6057	peer_supports_pktdrop = 0;
6058	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6059		/* mark all addresses that we have currently on the list */
6060		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6061	}
6062	/* does the source address already exist? if so skip it */
6063	inp = stcb->sctp_ep;
6064	atomic_add_int(&stcb->asoc.refcnt, 1);
6065	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6066	atomic_add_int(&stcb->asoc.refcnt, -1);
6067
6068	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6069		/* we must add the source address */
6070		/* no scope set here since we have a tcb already. */
6071		switch (sa->sa_family) {
6072#ifdef INET
6073		case AF_INET:
6074			if (stcb->asoc.scope.ipv4_addr_legal) {
6075				if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6076					return (-1);
6077				}
6078			}
6079			break;
6080#endif
6081#ifdef INET6
6082		case AF_INET6:
6083			if (stcb->asoc.scope.ipv6_addr_legal) {
6084				if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6085					return (-2);
6086				}
6087			}
6088			break;
6089#endif
6090		default:
6091			break;
6092		}
6093	} else {
6094		if (net_tmp != NULL && stcb_tmp == stcb) {
6095			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6096		} else if (stcb_tmp != stcb) {
6097			/* It belongs to another association? */
6098			if (stcb_tmp)
6099				SCTP_TCB_UNLOCK(stcb_tmp);
6100			return (-3);
6101		}
6102	}
6103	if (stcb->asoc.state == 0) {
6104		/* the assoc was freed? */
6105		return (-4);
6106	}
6107	/* now we must go through each of the params. */
6108	phdr = sctp_get_next_param(m, offset, &param_buf, sizeof(param_buf));
6109	while (phdr) {
6110		ptype = ntohs(phdr->param_type);
6111		plen = ntohs(phdr->param_length);
6112		/*
6113		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6114		 * (uint32_t)ptype, (int)plen);
6115		 */
6116		if (offset + plen > limit) {
6117			break;
6118		}
6119		if (plen < sizeof(struct sctp_paramhdr)) {
6120			break;
6121		}
6122#ifdef INET
6123		if (ptype == SCTP_IPV4_ADDRESS) {
6124			if (stcb->asoc.scope.ipv4_addr_legal) {
6125				struct sctp_ipv4addr_param *p4, p4_buf;
6126
6127				/* ok get the v4 address and check/add */
6128				phdr = sctp_get_next_param(m, offset,
6129				    (struct sctp_paramhdr *)&p4_buf,
6130				    sizeof(p4_buf));
6131				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6132				    phdr == NULL) {
6133					return (-5);
6134				}
6135				p4 = (struct sctp_ipv4addr_param *)phdr;
6136				sin.sin_addr.s_addr = p4->addr;
6137				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6138					/* Skip multi-cast addresses */
6139					goto next_param;
6140				}
6141				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6142				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6143					goto next_param;
6144				}
6145				sa = (struct sockaddr *)&sin;
6146				inp = stcb->sctp_ep;
6147				atomic_add_int(&stcb->asoc.refcnt, 1);
6148				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6149				    dst, stcb);
6150				atomic_add_int(&stcb->asoc.refcnt, -1);
6151
6152				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6153				    inp == NULL) {
6154					/* we must add the source address */
6155					/*
6156					 * no scope set since we have a tcb
6157					 * already
6158					 */
6159
6160					/*
6161					 * we must validate the state again
6162					 * here
6163					 */
6164			add_it_now:
6165					if (stcb->asoc.state == 0) {
6166						/* the assoc was freed? */
6167						return (-7);
6168					}
6169					if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6170						return (-8);
6171					}
6172				} else if (stcb_tmp == stcb) {
6173					if (stcb->asoc.state == 0) {
6174						/* the assoc was freed? */
6175						return (-10);
6176					}
6177					if (net != NULL) {
6178						/* clear flag */
6179						net->dest_state &=
6180						    ~SCTP_ADDR_NOT_IN_ASSOC;
6181					}
6182				} else {
6183					/*
6184					 * strange, address is in another
6185					 * assoc? straighten out locks.
6186					 */
6187					if (stcb_tmp) {
6188						if (SCTP_GET_STATE(stcb_tmp) == SCTP_STATE_COOKIE_WAIT) {
6189							struct mbuf *op_err;
6190							char msg[SCTP_DIAG_INFO_LEN];
6191
6192							/*
6193							 * in setup state we
6194							 * abort this guy
6195							 */
6196							SCTP_SNPRINTF(msg, sizeof(msg),
6197							    "%s:%d at %s", __FILE__, __LINE__, __func__);
6198							op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
6199							    msg);
6200							sctp_abort_an_association(stcb_tmp->sctp_ep,
6201							    stcb_tmp, op_err,
6202							    SCTP_SO_NOT_LOCKED);
6203							goto add_it_now;
6204						}
6205						SCTP_TCB_UNLOCK(stcb_tmp);
6206					}
6207
6208					if (stcb->asoc.state == 0) {
6209						/* the assoc was freed? */
6210						return (-12);
6211					}
6212					return (-13);
6213				}
6214			}
6215		} else
6216#endif
6217#ifdef INET6
6218		if (ptype == SCTP_IPV6_ADDRESS) {
6219			if (stcb->asoc.scope.ipv6_addr_legal) {
6220				/* ok get the v6 address and check/add */
6221				struct sctp_ipv6addr_param *p6, p6_buf;
6222
6223				phdr = sctp_get_next_param(m, offset,
6224				    (struct sctp_paramhdr *)&p6_buf,
6225				    sizeof(p6_buf));
6226				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6227				    phdr == NULL) {
6228					return (-14);
6229				}
6230				p6 = (struct sctp_ipv6addr_param *)phdr;
6231				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6232				    sizeof(p6->addr));
6233				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6234					/* Skip multi-cast addresses */
6235					goto next_param;
6236				}
6237				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6238					/*
6239					 * Link local make no sense without
6240					 * scope
6241					 */
6242					goto next_param;
6243				}
6244				sa = (struct sockaddr *)&sin6;
6245				inp = stcb->sctp_ep;
6246				atomic_add_int(&stcb->asoc.refcnt, 1);
6247				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6248				    dst, stcb);
6249				atomic_add_int(&stcb->asoc.refcnt, -1);
6250				if (stcb_tmp == NULL &&
6251				    (inp == stcb->sctp_ep || inp == NULL)) {
6252					/*
6253					 * we must validate the state again
6254					 * here
6255					 */
6256			add_it_now6:
6257					if (stcb->asoc.state == 0) {
6258						/* the assoc was freed? */
6259						return (-16);
6260					}
6261					/*
6262					 * we must add the address, no scope
6263					 * set
6264					 */
6265					if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6266						return (-17);
6267					}
6268				} else if (stcb_tmp == stcb) {
6269					/*
6270					 * we must validate the state again
6271					 * here
6272					 */
6273					if (stcb->asoc.state == 0) {
6274						/* the assoc was freed? */
6275						return (-19);
6276					}
6277					if (net != NULL) {
6278						/* clear flag */
6279						net->dest_state &=
6280						    ~SCTP_ADDR_NOT_IN_ASSOC;
6281					}
6282				} else {
6283					/*
6284					 * strange, address is in another
6285					 * assoc? straighten out locks.
6286					 */
6287					if (stcb_tmp) {
6288						if (SCTP_GET_STATE(stcb_tmp) == SCTP_STATE_COOKIE_WAIT) {
6289							struct mbuf *op_err;
6290							char msg[SCTP_DIAG_INFO_LEN];
6291
6292							/*
6293							 * in setup state we
6294							 * abort this guy
6295							 */
6296							SCTP_SNPRINTF(msg, sizeof(msg),
6297							    "%s:%d at %s", __FILE__, __LINE__, __func__);
6298							op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
6299							    msg);
6300							sctp_abort_an_association(stcb_tmp->sctp_ep,
6301							    stcb_tmp, op_err,
6302							    SCTP_SO_NOT_LOCKED);
6303							goto add_it_now6;
6304						}
6305						SCTP_TCB_UNLOCK(stcb_tmp);
6306					}
6307					if (stcb->asoc.state == 0) {
6308						/* the assoc was freed? */
6309						return (-21);
6310					}
6311					return (-22);
6312				}
6313			}
6314		} else
6315#endif
6316		if (ptype == SCTP_ECN_CAPABLE) {
6317			peer_supports_ecn = 1;
6318		} else if (ptype == SCTP_ULP_ADAPTATION) {
6319			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6320				struct sctp_adaptation_layer_indication ai,
6321				                                *aip;
6322
6323				phdr = sctp_get_next_param(m, offset,
6324				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6325				aip = (struct sctp_adaptation_layer_indication *)phdr;
6326				if (aip) {
6327					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6328					stcb->asoc.adaptation_needed = 1;
6329				}
6330			}
6331		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6332			struct sctp_asconf_addr_param lstore, *fee;
6333			int lptype;
6334			struct sockaddr *lsa = NULL;
6335#ifdef INET
6336			struct sctp_asconf_addrv4_param *fii;
6337#endif
6338
6339			if (stcb->asoc.asconf_supported == 0) {
6340				return (-100);
6341			}
6342			if (plen > sizeof(lstore)) {
6343				return (-23);
6344			}
6345			if (plen < sizeof(struct sctp_asconf_addrv4_param)) {
6346				return (-101);
6347			}
6348			phdr = sctp_get_next_param(m, offset,
6349			    (struct sctp_paramhdr *)&lstore,
6350			    plen);
6351			if (phdr == NULL) {
6352				return (-24);
6353			}
6354			fee = (struct sctp_asconf_addr_param *)phdr;
6355			lptype = ntohs(fee->addrp.ph.param_type);
6356			switch (lptype) {
6357#ifdef INET
6358			case SCTP_IPV4_ADDRESS:
6359				if (plen !=
6360				    sizeof(struct sctp_asconf_addrv4_param)) {
6361					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6362					    (int)sizeof(struct sctp_asconf_addrv4_param),
6363					    plen);
6364				} else {
6365					fii = (struct sctp_asconf_addrv4_param *)fee;
6366					sin.sin_addr.s_addr = fii->addrp.addr;
6367					lsa = (struct sockaddr *)&sin;
6368				}
6369				break;
6370#endif
6371#ifdef INET6
6372			case SCTP_IPV6_ADDRESS:
6373				if (plen !=
6374				    sizeof(struct sctp_asconf_addr_param)) {
6375					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6376					    (int)sizeof(struct sctp_asconf_addr_param),
6377					    plen);
6378				} else {
6379					memcpy(sin6.sin6_addr.s6_addr,
6380					    fee->addrp.addr,
6381					    sizeof(fee->addrp.addr));
6382					lsa = (struct sockaddr *)&sin6;
6383				}
6384				break;
6385#endif
6386			default:
6387				break;
6388			}
6389			if (lsa) {
6390				(void)sctp_set_primary_addr(stcb, sa, NULL);
6391			}
6392		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6393			stcb->asoc.peer_supports_nat = 1;
6394		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6395			/* Peer supports pr-sctp */
6396			peer_supports_prsctp = 1;
6397		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6398			/* A supported extension chunk */
6399			struct sctp_supported_chunk_types_param *pr_supported;
6400			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6401			int num_ent, i;
6402
6403			if (plen > sizeof(local_store)) {
6404				return (-35);
6405			}
6406			phdr = sctp_get_next_param(m, offset,
6407			    (struct sctp_paramhdr *)&local_store, plen);
6408			if (phdr == NULL) {
6409				return (-25);
6410			}
6411			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6412			num_ent = plen - sizeof(struct sctp_paramhdr);
6413			for (i = 0; i < num_ent; i++) {
6414				switch (pr_supported->chunk_types[i]) {
6415				case SCTP_ASCONF:
6416					peer_supports_asconf = 1;
6417					break;
6418				case SCTP_ASCONF_ACK:
6419					peer_supports_asconf_ack = 1;
6420					break;
6421				case SCTP_FORWARD_CUM_TSN:
6422					peer_supports_prsctp = 1;
6423					break;
6424				case SCTP_PACKET_DROPPED:
6425					peer_supports_pktdrop = 1;
6426					break;
6427				case SCTP_NR_SELECTIVE_ACK:
6428					peer_supports_nrsack = 1;
6429					break;
6430				case SCTP_STREAM_RESET:
6431					peer_supports_reconfig = 1;
6432					break;
6433				case SCTP_AUTHENTICATION:
6434					peer_supports_auth = 1;
6435					break;
6436				case SCTP_IDATA:
6437					peer_supports_idata = 1;
6438					break;
6439				default:
6440					/* one I have not learned yet */
6441					break;
6442				}
6443			}
6444		} else if (ptype == SCTP_RANDOM) {
6445			if (plen > sizeof(random_store))
6446				break;
6447			if (got_random) {
6448				/* already processed a RANDOM */
6449				goto next_param;
6450			}
6451			phdr = sctp_get_next_param(m, offset,
6452			    (struct sctp_paramhdr *)random_store,
6453			    plen);
6454			if (phdr == NULL)
6455				return (-26);
6456			p_random = (struct sctp_auth_random *)phdr;
6457			random_len = plen - sizeof(*p_random);
6458			/* enforce the random length */
6459			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6460				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6461				return (-27);
6462			}
6463			got_random = 1;
6464		} else if (ptype == SCTP_HMAC_LIST) {
6465			uint16_t num_hmacs;
6466			uint16_t i;
6467
6468			if (plen > sizeof(hmacs_store))
6469				break;
6470			if (got_hmacs) {
6471				/* already processed a HMAC list */
6472				goto next_param;
6473			}
6474			phdr = sctp_get_next_param(m, offset,
6475			    (struct sctp_paramhdr *)hmacs_store,
6476			    plen);
6477			if (phdr == NULL)
6478				return (-28);
6479			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6480			hmacs_len = plen - sizeof(*hmacs);
6481			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6482			/* validate the hmac list */
6483			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6484				return (-29);
6485			}
6486			if (stcb->asoc.peer_hmacs != NULL)
6487				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6488			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6489			if (stcb->asoc.peer_hmacs != NULL) {
6490				for (i = 0; i < num_hmacs; i++) {
6491					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6492					    ntohs(hmacs->hmac_ids[i]));
6493				}
6494			}
6495			got_hmacs = 1;
6496		} else if (ptype == SCTP_CHUNK_LIST) {
6497			int i;
6498
6499			if (plen > sizeof(chunks_store))
6500				break;
6501			if (got_chklist) {
6502				/* already processed a Chunks list */
6503				goto next_param;
6504			}
6505			phdr = sctp_get_next_param(m, offset,
6506			    (struct sctp_paramhdr *)chunks_store,
6507			    plen);
6508			if (phdr == NULL)
6509				return (-30);
6510			chunks = (struct sctp_auth_chunk_list *)phdr;
6511			num_chunks = plen - sizeof(*chunks);
6512			if (stcb->asoc.peer_auth_chunks != NULL)
6513				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6514			else
6515				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6516			for (i = 0; i < num_chunks; i++) {
6517				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6518				    stcb->asoc.peer_auth_chunks);
6519				/* record asconf/asconf-ack if listed */
6520				if (chunks->chunk_types[i] == SCTP_ASCONF)
6521					saw_asconf = 1;
6522				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6523					saw_asconf_ack = 1;
6524			}
6525			got_chklist = 1;
6526		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6527			    (ptype == SCTP_STATE_COOKIE) ||
6528			    (ptype == SCTP_UNRECOG_PARAM) ||
6529			    (ptype == SCTP_COOKIE_PRESERVE) ||
6530			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6531			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6532			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6533			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6534		    (ptype == SCTP_SUCCESS_REPORT)) {
6535			/* don't care */
6536		} else {
6537			if ((ptype & 0x8000) == 0x0000) {
6538				/*
6539				 * must stop processing the rest of the
6540				 * param's. Any report bits were handled
6541				 * with the call to
6542				 * sctp_arethere_unrecognized_parameters()
6543				 * when the INIT or INIT-ACK was first seen.
6544				 */
6545				break;
6546			}
6547		}
6548
6549next_param:
6550		offset += SCTP_SIZE32(plen);
6551		if (offset >= limit) {
6552			break;
6553		}
6554		phdr = sctp_get_next_param(m, offset, &param_buf,
6555		    sizeof(param_buf));
6556	}
6557	/* Now check to see if we need to purge any addresses */
6558	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6559		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6560		    SCTP_ADDR_NOT_IN_ASSOC) {
6561			/* This address has been removed from the asoc */
6562			/* remove and free it */
6563			stcb->asoc.numnets--;
6564			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6565			if (net == stcb->asoc.alternate) {
6566				sctp_free_remote_addr(stcb->asoc.alternate);
6567				stcb->asoc.alternate = NULL;
6568			}
6569			if (net == stcb->asoc.primary_destination) {
6570				stcb->asoc.primary_destination = NULL;
6571				sctp_select_primary_destination(stcb);
6572			}
6573			sctp_free_remote_addr(net);
6574		}
6575	}
6576	if ((stcb->asoc.ecn_supported == 1) &&
6577	    (peer_supports_ecn == 0)) {
6578		stcb->asoc.ecn_supported = 0;
6579	}
6580	if ((stcb->asoc.prsctp_supported == 1) &&
6581	    (peer_supports_prsctp == 0)) {
6582		stcb->asoc.prsctp_supported = 0;
6583	}
6584	if ((stcb->asoc.auth_supported == 1) &&
6585	    ((peer_supports_auth == 0) ||
6586	    (got_random == 0) || (got_hmacs == 0))) {
6587		stcb->asoc.auth_supported = 0;
6588	}
6589	if ((stcb->asoc.asconf_supported == 1) &&
6590	    ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) ||
6591	    (stcb->asoc.auth_supported == 0) ||
6592	    (saw_asconf == 0) || (saw_asconf_ack == 0))) {
6593		stcb->asoc.asconf_supported = 0;
6594	}
6595	if ((stcb->asoc.reconfig_supported == 1) &&
6596	    (peer_supports_reconfig == 0)) {
6597		stcb->asoc.reconfig_supported = 0;
6598	}
6599	if ((stcb->asoc.idata_supported == 1) &&
6600	    (peer_supports_idata == 0)) {
6601		stcb->asoc.idata_supported = 0;
6602	}
6603	if ((stcb->asoc.nrsack_supported == 1) &&
6604	    (peer_supports_nrsack == 0)) {
6605		stcb->asoc.nrsack_supported = 0;
6606	}
6607	if ((stcb->asoc.pktdrop_supported == 1) &&
6608	    (peer_supports_pktdrop == 0)) {
6609		stcb->asoc.pktdrop_supported = 0;
6610	}
6611	/* validate authentication required parameters */
6612	if ((peer_supports_auth == 0) && (got_chklist == 1)) {
6613		/* peer does not support auth but sent a chunks list? */
6614		return (-31);
6615	}
6616	if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) {
6617		/* peer supports asconf but not auth? */
6618		return (-32);
6619	} else if ((peer_supports_asconf == 1) &&
6620		    (peer_supports_auth == 1) &&
6621	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6622		return (-33);
6623	}
6624	/* concatenate the full random key */
6625	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6626	if (chunks != NULL) {
6627		keylen += sizeof(*chunks) + num_chunks;
6628	}
6629	new_key = sctp_alloc_key(keylen);
6630	if (new_key != NULL) {
6631		/* copy in the RANDOM */
6632		if (p_random != NULL) {
6633			keylen = sizeof(*p_random) + random_len;
6634			memcpy(new_key->key, p_random, keylen);
6635		} else {
6636			keylen = 0;
6637		}
6638		/* append in the AUTH chunks */
6639		if (chunks != NULL) {
6640			memcpy(new_key->key + keylen, chunks,
6641			    sizeof(*chunks) + num_chunks);
6642			keylen += sizeof(*chunks) + num_chunks;
6643		}
6644		/* append in the HMACs */
6645		if (hmacs != NULL) {
6646			memcpy(new_key->key + keylen, hmacs,
6647			    sizeof(*hmacs) + hmacs_len);
6648		}
6649	} else {
6650		/* failed to get memory for the key */
6651		return (-34);
6652	}
6653	if (stcb->asoc.authinfo.peer_random != NULL)
6654		sctp_free_key(stcb->asoc.authinfo.peer_random);
6655	stcb->asoc.authinfo.peer_random = new_key;
6656	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6657	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6658
6659	return (0);
6660}
6661
6662int
6663sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6664    struct sctp_nets *net)
6665{
6666	/* make sure the requested primary address exists in the assoc */
6667	if (net == NULL && sa)
6668		net = sctp_findnet(stcb, sa);
6669
6670	if (net == NULL) {
6671		/* didn't find the requested primary address! */
6672		return (-1);
6673	} else {
6674		/* set the primary address */
6675		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6676			/* Must be confirmed, so queue to set */
6677			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6678			return (0);
6679		}
6680		stcb->asoc.primary_destination = net;
6681		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6682			sctp_free_remote_addr(stcb->asoc.alternate);
6683			stcb->asoc.alternate = NULL;
6684		}
6685		net = TAILQ_FIRST(&stcb->asoc.nets);
6686		if (net != stcb->asoc.primary_destination) {
6687			/*
6688			 * first one on the list is NOT the primary
6689			 * sctp_cmpaddr() is much more efficient if the
6690			 * primary is the first on the list, make it so.
6691			 */
6692			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6693			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6694		}
6695		return (0);
6696	}
6697}
6698
6699int
6700sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6701{
6702	/*
6703	 * This function serves two purposes. It will see if a TAG can be
6704	 * re-used and return 1 for yes it is ok and 0 for don't use that
6705	 * tag. A secondary function it will do is purge out old tags that
6706	 * can be removed.
6707	 */
6708	struct sctpvtaghead *chain;
6709	struct sctp_tagblock *twait_block;
6710	struct sctpasochead *head;
6711	struct sctp_tcb *stcb;
6712	int i;
6713
6714	SCTP_INP_INFO_RLOCK();
6715	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6716	    SCTP_BASE_INFO(hashasocmark))];
6717	LIST_FOREACH(stcb, head, sctp_asocs) {
6718		/*
6719		 * We choose not to lock anything here. TCB's can't be
6720		 * removed since we have the read lock, so they can't be
6721		 * freed on us, same thing for the INP. I may be wrong with
6722		 * this assumption, but we will go with it for now :-)
6723		 */
6724		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6725			continue;
6726		}
6727		if (stcb->asoc.my_vtag == tag) {
6728			/* candidate */
6729			if (stcb->rport != rport) {
6730				continue;
6731			}
6732			if (stcb->sctp_ep->sctp_lport != lport) {
6733				continue;
6734			}
6735			/* Its a used tag set */
6736			SCTP_INP_INFO_RUNLOCK();
6737			return (0);
6738		}
6739	}
6740	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6741	/* Now what about timed wait ? */
6742	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6743		/*
6744		 * Block(s) are present, lets see if we have this tag in the
6745		 * list
6746		 */
6747		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6748			if (twait_block->vtag_block[i].v_tag == 0) {
6749				/* not used */
6750				continue;
6751			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6752			    now->tv_sec) {
6753				/* Audit expires this guy */
6754				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6755				twait_block->vtag_block[i].v_tag = 0;
6756				twait_block->vtag_block[i].lport = 0;
6757				twait_block->vtag_block[i].rport = 0;
6758			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6759				    (twait_block->vtag_block[i].lport == lport) &&
6760			    (twait_block->vtag_block[i].rport == rport)) {
6761				/* Bad tag, sorry :< */
6762				SCTP_INP_INFO_RUNLOCK();
6763				return (0);
6764			}
6765		}
6766	}
6767	SCTP_INP_INFO_RUNLOCK();
6768	return (1);
6769}
6770
6771static void
6772sctp_drain_mbufs(struct sctp_tcb *stcb)
6773{
6774	/*
6775	 * We must hunt this association for MBUF's past the cumack (i.e.
6776	 * out of order data that we can renege on).
6777	 */
6778	struct sctp_association *asoc;
6779	struct sctp_tmit_chunk *chk, *nchk;
6780	uint32_t cumulative_tsn_p1;
6781	struct sctp_queued_to_read *control, *ncontrol;
6782	int cnt, strmat;
6783	uint32_t gap, i;
6784	int fnd = 0;
6785
6786	/* We look for anything larger than the cum-ack + 1 */
6787
6788	asoc = &stcb->asoc;
6789	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6790		/* none we can reneg on. */
6791		return;
6792	}
6793	SCTP_STAT_INCR(sctps_protocol_drains_done);
6794	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6795	cnt = 0;
6796	/* Ok that was fun, now we will drain all the inbound streams? */
6797	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6798		TAILQ_FOREACH_SAFE(control, &asoc->strmin[strmat].inqueue, next_instrm, ncontrol) {
6799#ifdef INVARIANTS
6800			if (control->on_strm_q != SCTP_ON_ORDERED) {
6801				panic("Huh control: %p on_q: %d -- not ordered?",
6802				    control, control->on_strm_q);
6803			}
6804#endif
6805			if (SCTP_TSN_GT(control->sinfo_tsn, cumulative_tsn_p1)) {
6806				/* Yep it is above cum-ack */
6807				cnt++;
6808				SCTP_CALC_TSN_TO_GAP(gap, control->sinfo_tsn, asoc->mapping_array_base_tsn);
6809				KASSERT(control->length > 0, ("control has zero length"));
6810				if (asoc->size_on_all_streams >= control->length) {
6811					asoc->size_on_all_streams -= control->length;
6812				} else {
6813#ifdef INVARIANTS
6814					panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length);
6815#else
6816					asoc->size_on_all_streams = 0;
6817#endif
6818				}
6819				sctp_ucount_decr(asoc->cnt_on_all_streams);
6820				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6821				if (control->on_read_q) {
6822					TAILQ_REMOVE(&stcb->sctp_ep->read_queue, control, next);
6823					control->on_read_q = 0;
6824				}
6825				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, control, next_instrm);
6826				control->on_strm_q = 0;
6827				if (control->data) {
6828					sctp_m_freem(control->data);
6829					control->data = NULL;
6830				}
6831				sctp_free_remote_addr(control->whoFrom);
6832				/* Now its reasm? */
6833				TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) {
6834					cnt++;
6835					SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.tsn, asoc->mapping_array_base_tsn);
6836					KASSERT(chk->send_size > 0, ("chunk has zero length"));
6837					if (asoc->size_on_reasm_queue >= chk->send_size) {
6838						asoc->size_on_reasm_queue -= chk->send_size;
6839					} else {
6840#ifdef INVARIANTS
6841						panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, chk->send_size);
6842#else
6843						asoc->size_on_reasm_queue = 0;
6844#endif
6845					}
6846					sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6847					SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6848					TAILQ_REMOVE(&control->reasm, chk, sctp_next);
6849					if (chk->data) {
6850						sctp_m_freem(chk->data);
6851						chk->data = NULL;
6852					}
6853					sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6854				}
6855				sctp_free_a_readq(stcb, control);
6856			}
6857		}
6858		TAILQ_FOREACH_SAFE(control, &asoc->strmin[strmat].uno_inqueue, next_instrm, ncontrol) {
6859#ifdef INVARIANTS
6860			if (control->on_strm_q != SCTP_ON_UNORDERED) {
6861				panic("Huh control: %p on_q: %d -- not unordered?",
6862				    control, control->on_strm_q);
6863			}
6864#endif
6865			if (SCTP_TSN_GT(control->sinfo_tsn, cumulative_tsn_p1)) {
6866				/* Yep it is above cum-ack */
6867				cnt++;
6868				SCTP_CALC_TSN_TO_GAP(gap, control->sinfo_tsn, asoc->mapping_array_base_tsn);
6869				KASSERT(control->length > 0, ("control has zero length"));
6870				if (asoc->size_on_all_streams >= control->length) {
6871					asoc->size_on_all_streams -= control->length;
6872				} else {
6873#ifdef INVARIANTS
6874					panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length);
6875#else
6876					asoc->size_on_all_streams = 0;
6877#endif
6878				}
6879				sctp_ucount_decr(asoc->cnt_on_all_streams);
6880				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6881				if (control->on_read_q) {
6882					TAILQ_REMOVE(&stcb->sctp_ep->read_queue, control, next);
6883					control->on_read_q = 0;
6884				}
6885				TAILQ_REMOVE(&asoc->strmin[strmat].uno_inqueue, control, next_instrm);
6886				control->on_strm_q = 0;
6887				if (control->data) {
6888					sctp_m_freem(control->data);
6889					control->data = NULL;
6890				}
6891				sctp_free_remote_addr(control->whoFrom);
6892				/* Now its reasm? */
6893				TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) {
6894					cnt++;
6895					SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.tsn, asoc->mapping_array_base_tsn);
6896					KASSERT(chk->send_size > 0, ("chunk has zero length"));
6897					if (asoc->size_on_reasm_queue >= chk->send_size) {
6898						asoc->size_on_reasm_queue -= chk->send_size;
6899					} else {
6900#ifdef INVARIANTS
6901						panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, chk->send_size);
6902#else
6903						asoc->size_on_reasm_queue = 0;
6904#endif
6905					}
6906					sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6907					SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6908					TAILQ_REMOVE(&control->reasm, chk, sctp_next);
6909					if (chk->data) {
6910						sctp_m_freem(chk->data);
6911						chk->data = NULL;
6912					}
6913					sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6914				}
6915				sctp_free_a_readq(stcb, control);
6916			}
6917		}
6918	}
6919	if (cnt) {
6920		/* We must back down to see what the new highest is */
6921		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6922			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6923			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6924				asoc->highest_tsn_inside_map = i;
6925				fnd = 1;
6926				break;
6927			}
6928		}
6929		if (!fnd) {
6930			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6931		}
6932
6933		/*
6934		 * Question, should we go through the delivery queue? The
6935		 * only reason things are on here is the app not reading OR
6936		 * a p-d-api up. An attacker COULD send enough in to
6937		 * initiate the PD-API and then send a bunch of stuff to
6938		 * other streams... these would wind up on the delivery
6939		 * queue.. and then we would not get to them. But in order
6940		 * to do this I then have to back-track and un-deliver
6941		 * sequence numbers in streams.. el-yucko. I think for now
6942		 * we will NOT look at the delivery queue and leave it to be
6943		 * something to consider later. An alternative would be to
6944		 * abort the P-D-API with a notification and then deliver
6945		 * the data.... Or another method might be to keep track of
6946		 * how many times the situation occurs and if we see a
6947		 * possible attack underway just abort the association.
6948		 */
6949#ifdef SCTP_DEBUG
6950		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6951#endif
6952		/*
6953		 * Now do we need to find a new
6954		 * asoc->highest_tsn_inside_map?
6955		 */
6956		asoc->last_revoke_count = cnt;
6957		sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL,
6958		    SCTP_FROM_SCTP_PCB + SCTP_LOC_11);
6959		/* sa_ignore NO_NULL_CHK */
6960		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6961		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6962	}
6963	/*
6964	 * Another issue, in un-setting the TSN's in the mapping array we
6965	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6966	 * two things to occur. It may cause us to do extra work in checking
6967	 * for our mapping array movement. More importantly it may cause us
6968	 * to SACK every datagram. This may not be a bad thing though since
6969	 * we will recover once we get our cum-ack above and all this stuff
6970	 * we dumped recovered.
6971	 */
6972}
6973
6974void
6975sctp_drain()
6976{
6977	/*
6978	 * We must walk the PCB lists for ALL associations here. The system
6979	 * is LOW on MBUF's and needs help. This is where reneging will
6980	 * occur. We really hope this does NOT happen!
6981	 */
6982	VNET_ITERATOR_DECL(vnet_iter);
6983	VNET_LIST_RLOCK_NOSLEEP();
6984	VNET_FOREACH(vnet_iter) {
6985		CURVNET_SET(vnet_iter);
6986		struct sctp_inpcb *inp;
6987		struct sctp_tcb *stcb;
6988
6989		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6990		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6991#ifdef VIMAGE
6992			continue;
6993#else
6994			return;
6995#endif
6996		}
6997		SCTP_INP_INFO_RLOCK();
6998		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6999			/* For each endpoint */
7000			SCTP_INP_RLOCK(inp);
7001			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
7002				/* For each association */
7003				SCTP_TCB_LOCK(stcb);
7004				sctp_drain_mbufs(stcb);
7005				SCTP_TCB_UNLOCK(stcb);
7006			}
7007			SCTP_INP_RUNLOCK(inp);
7008		}
7009		SCTP_INP_INFO_RUNLOCK();
7010		CURVNET_RESTORE();
7011	}
7012	VNET_LIST_RUNLOCK_NOSLEEP();
7013}
7014
7015/*
7016 * start a new iterator
7017 * iterates through all endpoints and associations based on the pcb_state
7018 * flags and asoc_state.  "af" (mandatory) is executed for all matching
7019 * assocs and "ef" (optional) is executed when the iterator completes.
7020 * "inpf" (optional) is executed for each new endpoint as it is being
7021 * iterated through. inpe (optional) is called when the inp completes
7022 * its way through all the stcbs.
7023 */
7024int
7025sctp_initiate_iterator(inp_func inpf,
7026    asoc_func af,
7027    inp_func inpe,
7028    uint32_t pcb_state,
7029    uint32_t pcb_features,
7030    uint32_t asoc_state,
7031    void *argp,
7032    uint32_t argi,
7033    end_func ef,
7034    struct sctp_inpcb *s_inp,
7035    uint8_t chunk_output_off)
7036{
7037	struct sctp_iterator *it = NULL;
7038
7039	if (af == NULL) {
7040		return (-1);
7041	}
7042	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
7043		SCTP_PRINTF("%s: abort on initialize being %d\n", __func__,
7044		    SCTP_BASE_VAR(sctp_pcb_initialized));
7045		return (-1);
7046	}
7047	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
7048	    SCTP_M_ITER);
7049	if (it == NULL) {
7050		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
7051		return (-1);
7052	}
7053	memset(it, 0, sizeof(*it));
7054	it->function_assoc = af;
7055	it->function_inp = inpf;
7056	if (inpf)
7057		it->done_current_ep = 0;
7058	else
7059		it->done_current_ep = 1;
7060	it->function_atend = ef;
7061	it->pointer = argp;
7062	it->val = argi;
7063	it->pcb_flags = pcb_state;
7064	it->pcb_features = pcb_features;
7065	it->asoc_state = asoc_state;
7066	it->function_inp_end = inpe;
7067	it->no_chunk_output = chunk_output_off;
7068	it->vn = curvnet;
7069	if (s_inp) {
7070		/* Assume lock is held here */
7071		it->inp = s_inp;
7072		SCTP_INP_INCR_REF(it->inp);
7073		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7074	} else {
7075		SCTP_INP_INFO_RLOCK();
7076		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7077		if (it->inp) {
7078			SCTP_INP_INCR_REF(it->inp);
7079		}
7080		SCTP_INP_INFO_RUNLOCK();
7081		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7082	}
7083	SCTP_IPI_ITERATOR_WQ_LOCK();
7084	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
7085		SCTP_IPI_ITERATOR_WQ_UNLOCK();
7086		SCTP_PRINTF("%s: rollback on initialize being %d it=%p\n", __func__,
7087		    SCTP_BASE_VAR(sctp_pcb_initialized), it);
7088		SCTP_FREE(it, SCTP_M_ITER);
7089		return (-1);
7090	}
7091	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7092	if (sctp_it_ctl.iterator_running == 0) {
7093		sctp_wakeup_iterator();
7094	}
7095	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7096	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7097	return (0);
7098}
7099