sctp_pcb.c revision 218269
1/*-
2 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2011, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2011, by Michael Tuexen. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * a) Redistributions of source code must retain the above copyright notice,
10 *   this list of conditions and the following disclaimer.
11 *
12 * b) Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *   the documentation and/or other materials provided with the distribution.
15 *
16 * c) Neither the name of Cisco Systems, Inc. nor the names of its
17 *    contributors may be used to endorse or promote products derived
18 *    from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33/* $KAME: sctp_pcb.c,v 1.38 2005/03/06 16:04:18 itojun Exp $	 */
34
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: head/sys/netinet/sctp_pcb.c 218269 2011-02-04 13:50:30Z rrs $");
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#include <netinet/sctp_dtrace_define.h>
51#include <netinet/udp.h>
52#include <sys/sched.h>
53#include <sys/smp.h>
54#include <sys/unistd.h>
55
56
57VNET_DEFINE(struct sctp_base_info, system_base_info);
58
59/* FIX: we don't handle multiple link local scopes */
60/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
61#ifdef INET6
62int
63SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
64{
65	struct sockaddr_in6 tmp_a, tmp_b;
66
67	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
68	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
69		return 0;
70	}
71	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
72	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
73		return 0;
74	}
75	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
76}
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
97	SCTP_INP_INFO_RUNLOCK();
98}
99
100/*
101 * Addresses are added to VRF's (Virtual Router's). For BSD we
102 * have only the default VRF 0. We maintain a hash list of
103 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
104 * these has a list of addresses. When we add a new address
105 * to a VRF we lookup the ifn/ifn_index, if the ifn does
106 * not exist we create it and add it to the list of IFN's
107 * within the VRF. Once we have the sctp_ifn, we add the
108 * address to the list. So we look something like:
109 *
110 * hash-vrf-table
111 *   vrf-> ifn-> ifn -> ifn
112 *   vrf    |
113 *    ...   +--ifa-> ifa -> ifa
114 *   vrf
115 *
116 * We keep these separate lists since the SCTP subsystem will
117 * point to these from its source address selection nets structure.
118 * When an address is deleted it does not happen right away on
119 * the SCTP side, it gets scheduled. What we do when a
120 * delete happens is immediately remove the address from
121 * the master list and decrement the refcount. As our
122 * addip iterator works through and frees the src address
123 * selection pointing to the sctp_ifa, eventually the refcount
124 * will reach 0 and we will delete it. Note that it is assumed
125 * that any locking on system level ifn/ifa is done at the
126 * caller of these functions and these routines will only
127 * lock the SCTP structures as they add or delete things.
128 *
129 * Other notes on VRF concepts.
130 *  - An endpoint can be in multiple VRF's
131 *  - An association lives within a VRF and only one VRF.
132 *  - Any incoming packet we can deduce the VRF for by
133 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
134 *  - Any downward send call or connect call must supply the
135 *    VRF via ancillary data or via some sort of set default
136 *    VRF socket option call (again for BSD no brainer since
137 *    the VRF is always 0).
138 *  - An endpoint may add multiple VRF's to it.
139 *  - Listening sockets can accept associations in any
140 *    of the VRF's they are in but the assoc will end up
141 *    in only one VRF (gotten from the packet or connect/send).
142 *
143 */
144
145struct sctp_vrf *
146sctp_allocate_vrf(int vrf_id)
147{
148	struct sctp_vrf *vrf = NULL;
149	struct sctp_vrflist *bucket;
150
151	/* First allocate the VRF structure */
152	vrf = sctp_find_vrf(vrf_id);
153	if (vrf) {
154		/* Already allocated */
155		return (vrf);
156	}
157	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
158	    SCTP_M_VRF);
159	if (vrf == NULL) {
160		/* No memory */
161#ifdef INVARIANTS
162		panic("No memory for VRF:%d", vrf_id);
163#endif
164		return (NULL);
165	}
166	/* setup the VRF */
167	memset(vrf, 0, sizeof(struct sctp_vrf));
168	vrf->vrf_id = vrf_id;
169	LIST_INIT(&vrf->ifnlist);
170	vrf->total_ifa_count = 0;
171	vrf->refcount = 0;
172	/* now also setup table ids */
173	SCTP_INIT_VRF_TABLEID(vrf);
174	/* Init the HASH of addresses */
175	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
176	    &vrf->vrf_addr_hashmark);
177	if (vrf->vrf_addr_hash == NULL) {
178		/* No memory */
179#ifdef INVARIANTS
180		panic("No memory for VRF:%d", vrf_id);
181#endif
182		SCTP_FREE(vrf, SCTP_M_VRF);
183		return (NULL);
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
192
193struct sctp_ifn *
194sctp_find_ifn(void *ifn, uint32_t ifn_index)
195{
196	struct sctp_ifn *sctp_ifnp;
197	struct sctp_ifnlist *hash_ifn_head;
198
199	/*
200	 * We assume the lock is held for the addresses if that's wrong
201	 * problems could occur :-)
202	 */
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
215
216
217struct sctp_vrf *
218sctp_find_vrf(uint32_t vrf_id)
219{
220	struct sctp_vrflist *bucket;
221	struct sctp_vrf *liste;
222
223	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
224	LIST_FOREACH(liste, bucket, next_vrf) {
225		if (vrf_id == liste->vrf_id) {
226			return (liste);
227		}
228	}
229	return (NULL);
230}
231
232void
233sctp_free_vrf(struct sctp_vrf *vrf)
234{
235	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
236		if (vrf->vrf_addr_hash) {
237			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
238			vrf->vrf_addr_hash = NULL;
239		}
240		/* We zero'd the count */
241		LIST_REMOVE(vrf, next_vrf);
242		SCTP_FREE(vrf, SCTP_M_VRF);
243		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
244	}
245}
246
247void
248sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
249{
250	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
251		/* We zero'd the count */
252		if (sctp_ifnp->vrf) {
253			sctp_free_vrf(sctp_ifnp->vrf);
254		}
255		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
256		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
257	}
258}
259
260void
261sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
262{
263	struct sctp_ifn *sctp_ifnp;
264
265	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
266	if (sctp_ifnp != NULL) {
267		sctp_ifnp->ifn_mtu = mtu;
268	}
269}
270
271
272void
273sctp_free_ifa(struct sctp_ifa *sctp_ifap)
274{
275	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
276		/* We zero'd the count */
277		if (sctp_ifap->ifn_p) {
278			sctp_free_ifn(sctp_ifap->ifn_p);
279		}
280		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
281		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
282	}
283}
284
285static void
286sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
287{
288	struct sctp_ifn *found;
289
290	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
291	if (found == NULL) {
292		/* Not in the list.. sorry */
293		return;
294	}
295	if (hold_addr_lock == 0)
296		SCTP_IPI_ADDR_WLOCK();
297	LIST_REMOVE(sctp_ifnp, next_bucket);
298	LIST_REMOVE(sctp_ifnp, next_ifn);
299	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
300	    sctp_ifnp->registered_af);
301	if (hold_addr_lock == 0)
302		SCTP_IPI_ADDR_WUNLOCK();
303	/* Take away the reference, and possibly free it */
304	sctp_free_ifn(sctp_ifnp);
305}
306
307void
308sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
309    const char *if_name, uint32_t ifn_index)
310{
311	struct sctp_vrf *vrf;
312	struct sctp_ifa *sctp_ifap = NULL;
313
314	SCTP_IPI_ADDR_RLOCK();
315	vrf = sctp_find_vrf(vrf_id);
316	if (vrf == NULL) {
317		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
318		goto out;
319
320	}
321	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
322	if (sctp_ifap == NULL) {
323		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
324		goto out;
325	}
326	if (sctp_ifap->ifn_p == NULL) {
327		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
328		goto out;
329	}
330	if (if_name) {
331		int len1, len2;
332
333		len1 = strlen(if_name);
334		len2 = strlen(sctp_ifap->ifn_p->ifn_name);
335		if (len1 != len2) {
336			SCTPDBG(SCTP_DEBUG_PCB4, "IFN of ifa names different length %d vs %d - ignored\n",
337			    len1, len2);
338			goto out;
339		}
340		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) != 0) {
341			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
342			    sctp_ifap->ifn_p->ifn_name,
343			    if_name);
344			goto out;
345		}
346	} else {
347		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
348			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
349			    sctp_ifap->ifn_p->ifn_index, ifn_index);
350			goto out;
351		}
352	}
353
354	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
355	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
356out:
357	SCTP_IPI_ADDR_RUNLOCK();
358}
359
360void
361sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
362    const char *if_name, uint32_t ifn_index)
363{
364	struct sctp_vrf *vrf;
365	struct sctp_ifa *sctp_ifap = NULL;
366
367	SCTP_IPI_ADDR_RLOCK();
368	vrf = sctp_find_vrf(vrf_id);
369	if (vrf == NULL) {
370		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
371		goto out;
372
373	}
374	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
375	if (sctp_ifap == NULL) {
376		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
377		goto out;
378	}
379	if (sctp_ifap->ifn_p == NULL) {
380		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
381		goto out;
382	}
383	if (if_name) {
384		int len1, len2;
385
386		len1 = strlen(if_name);
387		len2 = strlen(sctp_ifap->ifn_p->ifn_name);
388		if (len1 != len2) {
389			SCTPDBG(SCTP_DEBUG_PCB4, "IFN of ifa names different length %d vs %d - ignored\n",
390			    len1, len2);
391			goto out;
392		}
393		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) != 0) {
394			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
395			    sctp_ifap->ifn_p->ifn_name,
396			    if_name);
397			goto out;
398		}
399	} else {
400		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
401			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
402			    sctp_ifap->ifn_p->ifn_index, ifn_index);
403			goto out;
404		}
405	}
406
407	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
408	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
409out:
410	SCTP_IPI_ADDR_RUNLOCK();
411}
412
413/*-
414 * Add an ifa to an ifn.
415 * Register the interface as necessary.
416 * NOTE: ADDR write lock MUST be held.
417 */
418static void
419sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
420{
421	int ifa_af;
422
423	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
424	sctp_ifap->ifn_p = sctp_ifnp;
425	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
426	/* update address counts */
427	sctp_ifnp->ifa_count++;
428	ifa_af = sctp_ifap->address.sa.sa_family;
429	if (ifa_af == AF_INET)
430		sctp_ifnp->num_v4++;
431	else
432		sctp_ifnp->num_v6++;
433	if (sctp_ifnp->ifa_count == 1) {
434		/* register the new interface */
435		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
436		sctp_ifnp->registered_af = ifa_af;
437	}
438}
439
440/*-
441 * Remove an ifa from its ifn.
442 * If no more addresses exist, remove the ifn too. Otherwise, re-register
443 * the interface based on the remaining address families left.
444 * NOTE: ADDR write lock MUST be held.
445 */
446static void
447sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
448{
449	uint32_t ifn_index;
450
451	LIST_REMOVE(sctp_ifap, next_ifa);
452	if (sctp_ifap->ifn_p) {
453		/* update address counts */
454		sctp_ifap->ifn_p->ifa_count--;
455		if (sctp_ifap->address.sa.sa_family == AF_INET6)
456			sctp_ifap->ifn_p->num_v6--;
457		else if (sctp_ifap->address.sa.sa_family == AF_INET)
458			sctp_ifap->ifn_p->num_v4--;
459
460		ifn_index = sctp_ifap->ifn_p->ifn_index;
461		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
462			/* remove the ifn, possibly freeing it */
463			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
464		} else {
465			/* re-register address family type, if needed */
466			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
467			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
468				SCTP_DEREGISTER_INTERFACE(ifn_index, AF_INET6);
469				SCTP_REGISTER_INTERFACE(ifn_index, AF_INET);
470				sctp_ifap->ifn_p->registered_af = AF_INET;
471			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
472			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
473				SCTP_DEREGISTER_INTERFACE(ifn_index, AF_INET);
474				SCTP_REGISTER_INTERFACE(ifn_index, AF_INET6);
475				sctp_ifap->ifn_p->registered_af = AF_INET6;
476			}
477			/* free the ifn refcount */
478			sctp_free_ifn(sctp_ifap->ifn_p);
479		}
480		sctp_ifap->ifn_p = NULL;
481	}
482}
483
484struct sctp_ifa *
485sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
486    uint32_t ifn_type, const char *if_name, void *ifa,
487    struct sockaddr *addr, uint32_t ifa_flags,
488    int dynamic_add)
489{
490	struct sctp_vrf *vrf;
491	struct sctp_ifn *sctp_ifnp = NULL;
492	struct sctp_ifa *sctp_ifap = NULL;
493	struct sctp_ifalist *hash_addr_head;
494	struct sctp_ifnlist *hash_ifn_head;
495	uint32_t hash_of_addr;
496	int new_ifn_af = 0;
497
498#ifdef SCTP_DEBUG
499	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
500	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
501#endif
502	SCTP_IPI_ADDR_WLOCK();
503	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
504	if (sctp_ifnp) {
505		vrf = sctp_ifnp->vrf;
506	} else {
507		vrf = sctp_find_vrf(vrf_id);
508		if (vrf == NULL) {
509			vrf = sctp_allocate_vrf(vrf_id);
510			if (vrf == NULL) {
511				SCTP_IPI_ADDR_WUNLOCK();
512				return (NULL);
513			}
514		}
515	}
516	if (sctp_ifnp == NULL) {
517		/*
518		 * build one and add it, can't hold lock until after malloc
519		 * done though.
520		 */
521		SCTP_IPI_ADDR_WUNLOCK();
522		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
523		    sizeof(struct sctp_ifn), SCTP_M_IFN);
524		if (sctp_ifnp == NULL) {
525#ifdef INVARIANTS
526			panic("No memory for IFN");
527#endif
528			return (NULL);
529		}
530		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
531		sctp_ifnp->ifn_index = ifn_index;
532		sctp_ifnp->ifn_p = ifn;
533		sctp_ifnp->ifn_type = ifn_type;
534		sctp_ifnp->refcount = 0;
535		sctp_ifnp->vrf = vrf;
536		atomic_add_int(&vrf->refcount, 1);
537		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
538		if (if_name != NULL) {
539			memcpy(sctp_ifnp->ifn_name, if_name, SCTP_IFNAMSIZ);
540		} else {
541			memcpy(sctp_ifnp->ifn_name, "unknown", min(7, SCTP_IFNAMSIZ));
542		}
543		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
544		LIST_INIT(&sctp_ifnp->ifalist);
545		SCTP_IPI_ADDR_WLOCK();
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			    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			return (sctp_ifap);
573		} else {
574			if (sctp_ifap->ifn_p) {
575				/*
576				 * The last IFN gets the address, remove the
577				 * old one
578				 */
579				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
580				    sctp_ifap, sctp_ifap->ifn_p->ifn_name,
581				    sctp_ifap->ifn_p->ifn_index, if_name,
582				    ifn_index);
583				/* remove the address from the old ifn */
584				sctp_remove_ifa_from_ifn(sctp_ifap);
585				/* move the address over to the new ifn */
586				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
587				goto exit_stage_left;
588			} else {
589				/* repair ifnp which was NULL ? */
590				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
591				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
592				    sctp_ifnp, sctp_ifap);
593				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
594			}
595			goto exit_stage_left;
596		}
597	}
598	SCTP_IPI_ADDR_WUNLOCK();
599	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
600	if (sctp_ifap == NULL) {
601#ifdef INVARIANTS
602		panic("No memory for IFA");
603#endif
604		return (NULL);
605	}
606	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
607	sctp_ifap->ifn_p = sctp_ifnp;
608	atomic_add_int(&sctp_ifnp->refcount, 1);
609	sctp_ifap->vrf_id = vrf_id;
610	sctp_ifap->ifa = ifa;
611	memcpy(&sctp_ifap->address, addr, addr->sa_len);
612	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
613	sctp_ifap->flags = ifa_flags;
614	/* Set scope */
615	switch (sctp_ifap->address.sa.sa_family) {
616	case AF_INET:
617		{
618			struct sockaddr_in *sin;
619
620			sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
621			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
622			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
623				sctp_ifap->src_is_loop = 1;
624			}
625			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
626				sctp_ifap->src_is_priv = 1;
627			}
628			sctp_ifnp->num_v4++;
629			if (new_ifn_af)
630				new_ifn_af = AF_INET;
631			break;
632		}
633#ifdef INET6
634	case AF_INET6:
635		{
636			/* ok to use deprecated addresses? */
637			struct sockaddr_in6 *sin6;
638
639			sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
640			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
641			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
642				sctp_ifap->src_is_loop = 1;
643			}
644			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
645				sctp_ifap->src_is_priv = 1;
646			}
647			sctp_ifnp->num_v6++;
648			if (new_ifn_af)
649				new_ifn_af = AF_INET6;
650			break;
651		}
652#endif
653	default:
654		new_ifn_af = 0;
655		break;
656	}
657	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
658
659	if ((sctp_ifap->src_is_priv == 0) &&
660	    (sctp_ifap->src_is_loop == 0)) {
661		sctp_ifap->src_is_glob = 1;
662	}
663	SCTP_IPI_ADDR_WLOCK();
664	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
665	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
666	sctp_ifap->refcount = 1;
667	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
668	sctp_ifnp->ifa_count++;
669	vrf->total_ifa_count++;
670	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
671	if (new_ifn_af) {
672		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
673		sctp_ifnp->registered_af = new_ifn_af;
674	}
675	SCTP_IPI_ADDR_WUNLOCK();
676	if (dynamic_add) {
677		/*
678		 * Bump up the refcount so that when the timer completes it
679		 * will drop back down.
680		 */
681		struct sctp_laddr *wi;
682
683		atomic_add_int(&sctp_ifap->refcount, 1);
684		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
685		if (wi == NULL) {
686			/*
687			 * Gak, what can we do? We have lost an address
688			 * change can you say HOSED?
689			 */
690			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
691			/* Opps, must decrement the count */
692			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
693			    if_name);
694			return (NULL);
695		}
696		SCTP_INCR_LADDR_COUNT();
697		bzero(wi, sizeof(*wi));
698		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
699		wi->ifa = sctp_ifap;
700		wi->action = SCTP_ADD_IP_ADDRESS;
701
702		SCTP_WQ_ADDR_LOCK();
703		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
704		SCTP_WQ_ADDR_UNLOCK();
705
706		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
707		    (struct sctp_inpcb *)NULL,
708		    (struct sctp_tcb *)NULL,
709		    (struct sctp_nets *)NULL);
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#ifdef SCTP_DEBUG
731	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
732	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
733#endif
734	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
735	if (sctp_ifap) {
736		/* Validate the delete */
737		if (sctp_ifap->ifn_p) {
738			int valid = 0;
739
740			/*-
741			 * The name has priority over the ifn_index
742			 * if its given. We do this especially for
743			 * panda who might recycle indexes fast.
744			 */
745			if (if_name) {
746				int len1, len2;
747
748				len1 = min(SCTP_IFNAMSIZ, strlen(if_name));
749				len2 = min(SCTP_IFNAMSIZ, strlen(sctp_ifap->ifn_p->ifn_name));
750				if (len1 && len2 && (len1 == len2)) {
751					/* we can compare them */
752					if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) == 0) {
753						/*
754						 * They match its a correct
755						 * delete
756						 */
757						valid = 1;
758					}
759				}
760			}
761			if (!valid) {
762				/* last ditch check ifn_index */
763				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
764					valid = 1;
765				}
766			}
767			if (!valid) {
768				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
769				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
770				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
771				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
772				SCTP_IPI_ADDR_WUNLOCK();
773				return;
774			}
775		}
776		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", sctp_ifap);
777		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
778		sctp_ifap->localifa_flags |= SCTP_BEING_DELETED;
779		vrf->total_ifa_count--;
780		LIST_REMOVE(sctp_ifap, next_bucket);
781		sctp_remove_ifa_from_ifn(sctp_ifap);
782	}
783#ifdef SCTP_DEBUG
784	else {
785		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
786		    ifn_index);
787		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
788	}
789#endif
790
791out_now:
792	SCTP_IPI_ADDR_WUNLOCK();
793	if (sctp_ifap) {
794		struct sctp_laddr *wi;
795
796		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
797		if (wi == NULL) {
798			/*
799			 * Gak, what can we do? We have lost an address
800			 * change can you say HOSED?
801			 */
802			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
803
804			/* Oops, must decrement the count */
805			sctp_free_ifa(sctp_ifap);
806			return;
807		}
808		SCTP_INCR_LADDR_COUNT();
809		bzero(wi, sizeof(*wi));
810		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
811		wi->ifa = sctp_ifap;
812		wi->action = SCTP_DEL_IP_ADDRESS;
813		SCTP_WQ_ADDR_LOCK();
814		/*
815		 * Should this really be a tailq? As it is we will process
816		 * the newest first :-0
817		 */
818		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
819		SCTP_WQ_ADDR_UNLOCK();
820
821		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
822		    (struct sctp_inpcb *)NULL,
823		    (struct sctp_tcb *)NULL,
824		    (struct sctp_nets *)NULL);
825	}
826	return;
827}
828
829
830static struct sctp_tcb *
831sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
832    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
833{
834	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
835	/*
836	 * If we support the TCP model, then we must now dig through to see
837	 * if we can find our endpoint in the list of tcp ep's.
838	 */
839	uint16_t lport, rport;
840	struct sctppcbhead *ephead;
841	struct sctp_inpcb *inp;
842	struct sctp_laddr *laddr;
843	struct sctp_tcb *stcb;
844	struct sctp_nets *net;
845
846	if ((to == NULL) || (from == NULL)) {
847		return (NULL);
848	}
849	if (to->sa_family == AF_INET && from->sa_family == AF_INET) {
850		lport = ((struct sockaddr_in *)to)->sin_port;
851		rport = ((struct sockaddr_in *)from)->sin_port;
852	} else if (to->sa_family == AF_INET6 && from->sa_family == AF_INET6) {
853		lport = ((struct sockaddr_in6 *)to)->sin6_port;
854		rport = ((struct sockaddr_in6 *)from)->sin6_port;
855	} else {
856		return NULL;
857	}
858	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
859	/*
860	 * Ok now for each of the guys in this bucket we must look and see:
861	 * - Does the remote port match. - Does there single association's
862	 * addresses match this address (to). If so we update p_ep to point
863	 * to this ep and return the tcb from it.
864	 */
865	LIST_FOREACH(inp, ephead, sctp_hash) {
866		SCTP_INP_RLOCK(inp);
867		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
868			SCTP_INP_RUNLOCK(inp);
869			continue;
870		}
871		if (lport != inp->sctp_lport) {
872			SCTP_INP_RUNLOCK(inp);
873			continue;
874		}
875		if (inp->def_vrf_id != vrf_id) {
876			SCTP_INP_RUNLOCK(inp);
877			continue;
878		}
879		/* check to see if the ep has one of the addresses */
880		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
881			/* We are NOT bound all, so look further */
882			int match = 0;
883
884			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
885
886				if (laddr->ifa == NULL) {
887					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
888					continue;
889				}
890				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
891					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
892					continue;
893				}
894				if (laddr->ifa->address.sa.sa_family ==
895				    to->sa_family) {
896					/* see if it matches */
897					struct sockaddr_in *intf_addr, *sin;
898
899					intf_addr = &laddr->ifa->address.sin;
900					sin = (struct sockaddr_in *)to;
901					if (from->sa_family == AF_INET) {
902						if (sin->sin_addr.s_addr ==
903						    intf_addr->sin_addr.s_addr) {
904							match = 1;
905							break;
906						}
907					}
908#ifdef INET6
909					if (from->sa_family == AF_INET6) {
910						struct sockaddr_in6 *intf_addr6;
911						struct sockaddr_in6 *sin6;
912
913						sin6 = (struct sockaddr_in6 *)
914						    to;
915						intf_addr6 = &laddr->ifa->address.sin6;
916
917						if (SCTP6_ARE_ADDR_EQUAL(sin6,
918						    intf_addr6)) {
919							match = 1;
920							break;
921						}
922					}
923#endif
924				}
925			}
926			if (match == 0) {
927				/* This endpoint does not have this address */
928				SCTP_INP_RUNLOCK(inp);
929				continue;
930			}
931		}
932		/*
933		 * Ok if we hit here the ep has the address, does it hold
934		 * the tcb?
935		 */
936
937		stcb = LIST_FIRST(&inp->sctp_asoc_list);
938		if (stcb == NULL) {
939			SCTP_INP_RUNLOCK(inp);
940			continue;
941		}
942		SCTP_TCB_LOCK(stcb);
943		if (stcb->rport != rport) {
944			/* remote port does not match. */
945			SCTP_TCB_UNLOCK(stcb);
946			SCTP_INP_RUNLOCK(inp);
947			continue;
948		}
949		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
950			SCTP_TCB_UNLOCK(stcb);
951			SCTP_INP_RUNLOCK(inp);
952			continue;
953		}
954		/* Does this TCB have a matching address? */
955		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
956
957			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
958				/* not the same family, can't be a match */
959				continue;
960			}
961			switch (from->sa_family) {
962			case AF_INET:
963				{
964					struct sockaddr_in *sin, *rsin;
965
966					sin = (struct sockaddr_in *)&net->ro._l_addr;
967					rsin = (struct sockaddr_in *)from;
968					if (sin->sin_addr.s_addr ==
969					    rsin->sin_addr.s_addr) {
970						/* found it */
971						if (netp != NULL) {
972							*netp = net;
973						}
974						/*
975						 * Update the endpoint
976						 * pointer
977						 */
978						*inp_p = inp;
979						SCTP_INP_RUNLOCK(inp);
980						return (stcb);
981					}
982					break;
983				}
984#ifdef INET6
985			case AF_INET6:
986				{
987					struct sockaddr_in6 *sin6, *rsin6;
988
989					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
990					rsin6 = (struct sockaddr_in6 *)from;
991					if (SCTP6_ARE_ADDR_EQUAL(sin6,
992					    rsin6)) {
993						/* found it */
994						if (netp != NULL) {
995							*netp = net;
996						}
997						/*
998						 * Update the endpoint
999						 * pointer
1000						 */
1001						*inp_p = inp;
1002						SCTP_INP_RUNLOCK(inp);
1003						return (stcb);
1004					}
1005					break;
1006				}
1007#endif
1008			default:
1009				/* TSNH */
1010				break;
1011			}
1012		}
1013		SCTP_TCB_UNLOCK(stcb);
1014		SCTP_INP_RUNLOCK(inp);
1015	}
1016	return (NULL);
1017}
1018
1019static int
1020sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
1021{
1022	int loopback_scope, ipv4_local_scope, local_scope, site_scope;
1023	int ipv4_addr_legal, ipv6_addr_legal;
1024	struct sctp_vrf *vrf;
1025	struct sctp_ifn *sctp_ifn;
1026	struct sctp_ifa *sctp_ifa;
1027
1028	loopback_scope = stcb->asoc.loopback_scope;
1029	ipv4_local_scope = stcb->asoc.ipv4_local_scope;
1030	local_scope = stcb->asoc.local_scope;
1031	site_scope = stcb->asoc.site_scope;
1032	ipv4_addr_legal = ipv6_addr_legal = 0;
1033	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1034		ipv6_addr_legal = 1;
1035		if (SCTP_IPV6_V6ONLY(stcb->sctp_ep) == 0) {
1036			ipv4_addr_legal = 1;
1037		}
1038	} else {
1039		ipv4_addr_legal = 1;
1040	}
1041
1042	SCTP_IPI_ADDR_RLOCK();
1043	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
1044	if (vrf == NULL) {
1045		/* no vrf, no addresses */
1046		SCTP_IPI_ADDR_RUNLOCK();
1047		return (0);
1048	}
1049	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
1050		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
1051			if ((loopback_scope == 0) &&
1052			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
1053				continue;
1054			}
1055			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
1056				if (sctp_is_addr_restricted(stcb, sctp_ifa))
1057					continue;
1058				switch (sctp_ifa->address.sa.sa_family) {
1059#ifdef INET
1060				case AF_INET:
1061					if (ipv4_addr_legal) {
1062						struct sockaddr_in *sin,
1063						           *rsin;
1064
1065						sin = &sctp_ifa->address.sin;
1066						rsin = (struct sockaddr_in *)to;
1067						if ((ipv4_local_scope == 0) &&
1068						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
1069							continue;
1070						}
1071						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
1072							SCTP_IPI_ADDR_RUNLOCK();
1073							return (1);
1074						}
1075					}
1076					break;
1077#endif
1078#ifdef INET6
1079				case AF_INET6:
1080					if (ipv6_addr_legal) {
1081						struct sockaddr_in6 *sin6,
1082						            *rsin6;
1083
1084						sin6 = &sctp_ifa->address.sin6;
1085						rsin6 = (struct sockaddr_in6 *)to;
1086						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
1087							if (local_scope == 0)
1088								continue;
1089							if (sin6->sin6_scope_id == 0) {
1090								if (sa6_recoverscope(sin6) != 0)
1091									continue;
1092							}
1093						}
1094						if ((site_scope == 0) &&
1095						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
1096							continue;
1097						}
1098						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
1099							SCTP_IPI_ADDR_RUNLOCK();
1100							return (1);
1101						}
1102					}
1103					break;
1104#endif
1105				default:
1106					/* TSNH */
1107					break;
1108				}
1109			}
1110		}
1111	} else {
1112		struct sctp_laddr *laddr;
1113
1114		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
1115			if (sctp_is_addr_restricted(stcb, laddr->ifa)) {
1116				continue;
1117			}
1118			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
1119				continue;
1120			}
1121			switch (to->sa_family) {
1122#ifdef INET
1123			case AF_INET:
1124				{
1125					struct sockaddr_in *sin, *rsin;
1126
1127					sin = (struct sockaddr_in *)&laddr->ifa->address.sin;
1128					rsin = (struct sockaddr_in *)to;
1129					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
1130						SCTP_IPI_ADDR_RUNLOCK();
1131						return (1);
1132					}
1133					break;
1134				}
1135#endif
1136#ifdef INET6
1137			case AF_INET6:
1138				{
1139					struct sockaddr_in6 *sin6, *rsin6;
1140
1141					sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
1142					rsin6 = (struct sockaddr_in6 *)to;
1143					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
1144						SCTP_IPI_ADDR_RUNLOCK();
1145						return (1);
1146					}
1147					break;
1148				}
1149
1150#endif
1151			default:
1152				/* TSNH */
1153				break;
1154			}
1155
1156		}
1157	}
1158	SCTP_IPI_ADDR_RUNLOCK();
1159	return (0);
1160}
1161
1162/*
1163 * rules for use
1164 *
1165 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1166 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1167 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1168 * NULL.
1169 */
1170
1171struct sctp_tcb *
1172sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1173    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1174{
1175	struct sctpasochead *head;
1176	struct sctp_inpcb *inp;
1177	struct sctp_tcb *stcb = NULL;
1178	struct sctp_nets *net;
1179	uint16_t rport;
1180
1181	inp = *inp_p;
1182	if (remote->sa_family == AF_INET) {
1183		rport = (((struct sockaddr_in *)remote)->sin_port);
1184	} else if (remote->sa_family == AF_INET6) {
1185		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1186	} else {
1187		return (NULL);
1188	}
1189	if (locked_tcb) {
1190		/*
1191		 * UN-lock so we can do proper locking here this occurs when
1192		 * called from load_addresses_from_init.
1193		 */
1194		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1195		SCTP_TCB_UNLOCK(locked_tcb);
1196	}
1197	SCTP_INP_INFO_RLOCK();
1198	if (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
1199		/*-
1200		 * Now either this guy is our listener or it's the
1201		 * connector. If it is the one that issued the connect, then
1202		 * it's only chance is to be the first TCB in the list. If
1203		 * it is the acceptor, then do the special_lookup to hash
1204		 * and find the real inp.
1205		 */
1206		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1207			/* to is peer addr, from is my addr */
1208			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1209			    netp, inp->def_vrf_id);
1210			if ((stcb != NULL) && (locked_tcb == NULL)) {
1211				/* we have a locked tcb, lower refcount */
1212				SCTP_INP_DECR_REF(inp);
1213			}
1214			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1215				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1216				SCTP_TCB_LOCK(locked_tcb);
1217				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1218				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1219			}
1220			SCTP_INP_INFO_RUNLOCK();
1221			return (stcb);
1222		} else {
1223			SCTP_INP_WLOCK(inp);
1224			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1225				goto null_return;
1226			}
1227			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1228			if (stcb == NULL) {
1229				goto null_return;
1230			}
1231			SCTP_TCB_LOCK(stcb);
1232
1233			if (stcb->rport != rport) {
1234				/* remote port does not match. */
1235				SCTP_TCB_UNLOCK(stcb);
1236				goto null_return;
1237			}
1238			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1239				SCTP_TCB_UNLOCK(stcb);
1240				goto null_return;
1241			}
1242			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1243				SCTP_TCB_UNLOCK(stcb);
1244				goto null_return;
1245			}
1246			/* now look at the list of remote addresses */
1247			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1248#ifdef INVARIANTS
1249				if (net == (TAILQ_NEXT(net, sctp_next))) {
1250					panic("Corrupt net list");
1251				}
1252#endif
1253				if (net->ro._l_addr.sa.sa_family !=
1254				    remote->sa_family) {
1255					/* not the same family */
1256					continue;
1257				}
1258				switch (remote->sa_family) {
1259				case AF_INET:
1260					{
1261						struct sockaddr_in *sin,
1262						           *rsin;
1263
1264						sin = (struct sockaddr_in *)
1265						    &net->ro._l_addr;
1266						rsin = (struct sockaddr_in *)remote;
1267						if (sin->sin_addr.s_addr ==
1268						    rsin->sin_addr.s_addr) {
1269							/* found it */
1270							if (netp != NULL) {
1271								*netp = net;
1272							}
1273							if (locked_tcb == NULL) {
1274								SCTP_INP_DECR_REF(inp);
1275							} else if (locked_tcb != stcb) {
1276								SCTP_TCB_LOCK(locked_tcb);
1277							}
1278							if (locked_tcb) {
1279								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1280							}
1281							SCTP_INP_WUNLOCK(inp);
1282							SCTP_INP_INFO_RUNLOCK();
1283							return (stcb);
1284						}
1285						break;
1286					}
1287#ifdef INET6
1288				case AF_INET6:
1289					{
1290						struct sockaddr_in6 *sin6,
1291						            *rsin6;
1292
1293						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1294						rsin6 = (struct sockaddr_in6 *)remote;
1295						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1296						    rsin6)) {
1297							/* found it */
1298							if (netp != NULL) {
1299								*netp = net;
1300							}
1301							if (locked_tcb == NULL) {
1302								SCTP_INP_DECR_REF(inp);
1303							} else if (locked_tcb != stcb) {
1304								SCTP_TCB_LOCK(locked_tcb);
1305							}
1306							if (locked_tcb) {
1307								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1308							}
1309							SCTP_INP_WUNLOCK(inp);
1310							SCTP_INP_INFO_RUNLOCK();
1311							return (stcb);
1312						}
1313						break;
1314					}
1315#endif
1316				default:
1317					/* TSNH */
1318					break;
1319				}
1320			}
1321			SCTP_TCB_UNLOCK(stcb);
1322		}
1323	} else {
1324		SCTP_INP_WLOCK(inp);
1325		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1326			goto null_return;
1327		}
1328		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1329		    inp->sctp_hashmark)];
1330		if (head == NULL) {
1331			goto null_return;
1332		}
1333		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1334			if (stcb->rport != rport) {
1335				/* remote port does not match */
1336				continue;
1337			}
1338			SCTP_TCB_LOCK(stcb);
1339			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1340				SCTP_TCB_UNLOCK(stcb);
1341				continue;
1342			}
1343			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1344				SCTP_TCB_UNLOCK(stcb);
1345				continue;
1346			}
1347			/* now look at the list of remote addresses */
1348			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1349#ifdef INVARIANTS
1350				if (net == (TAILQ_NEXT(net, sctp_next))) {
1351					panic("Corrupt net list");
1352				}
1353#endif
1354				if (net->ro._l_addr.sa.sa_family !=
1355				    remote->sa_family) {
1356					/* not the same family */
1357					continue;
1358				}
1359				switch (remote->sa_family) {
1360				case AF_INET:
1361					{
1362						struct sockaddr_in *sin,
1363						           *rsin;
1364
1365						sin = (struct sockaddr_in *)
1366						    &net->ro._l_addr;
1367						rsin = (struct sockaddr_in *)remote;
1368						if (sin->sin_addr.s_addr ==
1369						    rsin->sin_addr.s_addr) {
1370							/* found it */
1371							if (netp != NULL) {
1372								*netp = net;
1373							}
1374							if (locked_tcb == NULL) {
1375								SCTP_INP_DECR_REF(inp);
1376							} else if (locked_tcb != stcb) {
1377								SCTP_TCB_LOCK(locked_tcb);
1378							}
1379							if (locked_tcb) {
1380								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1381							}
1382							SCTP_INP_WUNLOCK(inp);
1383							SCTP_INP_INFO_RUNLOCK();
1384							return (stcb);
1385						}
1386						break;
1387					}
1388#ifdef INET6
1389				case AF_INET6:
1390					{
1391						struct sockaddr_in6 *sin6,
1392						            *rsin6;
1393
1394						sin6 = (struct sockaddr_in6 *)
1395						    &net->ro._l_addr;
1396						rsin6 = (struct sockaddr_in6 *)remote;
1397						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1398						    rsin6)) {
1399							/* found it */
1400							if (netp != NULL) {
1401								*netp = net;
1402							}
1403							if (locked_tcb == NULL) {
1404								SCTP_INP_DECR_REF(inp);
1405							} else if (locked_tcb != stcb) {
1406								SCTP_TCB_LOCK(locked_tcb);
1407							}
1408							if (locked_tcb) {
1409								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1410							}
1411							SCTP_INP_WUNLOCK(inp);
1412							SCTP_INP_INFO_RUNLOCK();
1413							return (stcb);
1414						}
1415						break;
1416					}
1417#endif
1418				default:
1419					/* TSNH */
1420					break;
1421				}
1422			}
1423			SCTP_TCB_UNLOCK(stcb);
1424		}
1425	}
1426null_return:
1427	/* clean up for returning null */
1428	if (locked_tcb) {
1429		SCTP_TCB_LOCK(locked_tcb);
1430		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1431	}
1432	SCTP_INP_WUNLOCK(inp);
1433	SCTP_INP_INFO_RUNLOCK();
1434	/* not found */
1435	return (NULL);
1436}
1437
1438/*
1439 * Find an association for a specific endpoint using the association id given
1440 * out in the COMM_UP notification
1441 */
1442
1443struct sctp_tcb *
1444sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1445{
1446	/*
1447	 * Use my the assoc_id to find a endpoint
1448	 */
1449	struct sctpasochead *head;
1450	struct sctp_tcb *stcb;
1451	uint32_t id;
1452
1453	if (inp == NULL) {
1454		SCTP_PRINTF("TSNH ep_associd\n");
1455		return (NULL);
1456	}
1457	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1458		SCTP_PRINTF("TSNH ep_associd0\n");
1459		return (NULL);
1460	}
1461	id = (uint32_t) asoc_id;
1462	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1463	if (head == NULL) {
1464		/* invalid id TSNH */
1465		SCTP_PRINTF("TSNH ep_associd1\n");
1466		return (NULL);
1467	}
1468	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1469		if (stcb->asoc.assoc_id == id) {
1470			if (inp != stcb->sctp_ep) {
1471				/*
1472				 * some other guy has the same id active (id
1473				 * collision ??).
1474				 */
1475				SCTP_PRINTF("TSNH ep_associd2\n");
1476				continue;
1477			}
1478			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1479				continue;
1480			}
1481			if (want_lock) {
1482				SCTP_TCB_LOCK(stcb);
1483			}
1484			return (stcb);
1485		}
1486	}
1487	return (NULL);
1488}
1489
1490
1491struct sctp_tcb *
1492sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1493{
1494	struct sctp_tcb *stcb;
1495
1496	SCTP_INP_RLOCK(inp);
1497	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1498	SCTP_INP_RUNLOCK(inp);
1499	return (stcb);
1500}
1501
1502
1503static struct sctp_inpcb *
1504sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1505    uint16_t lport, uint32_t vrf_id)
1506{
1507	struct sctp_inpcb *inp;
1508	struct sockaddr_in *sin;
1509
1510#ifdef INET6
1511	struct sockaddr_in6 *sin6;
1512
1513#endif
1514	struct sctp_laddr *laddr;
1515
1516#ifdef INET6
1517	struct sockaddr_in6 *intf_addr6;
1518
1519#endif
1520
1521	int fnd;
1522
1523	/*
1524	 * Endpoint probe expects that the INP_INFO is locked.
1525	 */
1526	sin = NULL;
1527#ifdef INET6
1528	sin6 = NULL;
1529#endif
1530	switch (nam->sa_family) {
1531	case AF_INET:
1532		sin = (struct sockaddr_in *)nam;
1533		break;
1534#ifdef INET6
1535	case AF_INET6:
1536		sin6 = (struct sockaddr_in6 *)nam;
1537		break;
1538#endif
1539	default:
1540		/* unsupported family */
1541		return (NULL);
1542	}
1543
1544	if (head == NULL)
1545		return (NULL);
1546
1547	LIST_FOREACH(inp, head, sctp_hash) {
1548		SCTP_INP_RLOCK(inp);
1549		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1550			SCTP_INP_RUNLOCK(inp);
1551			continue;
1552		}
1553		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1554		    (inp->sctp_lport == lport)) {
1555			/* got it */
1556			if ((nam->sa_family == AF_INET) &&
1557			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1558			    SCTP_IPV6_V6ONLY(inp)) {
1559				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1560				SCTP_INP_RUNLOCK(inp);
1561				continue;
1562			}
1563			/* A V6 address and the endpoint is NOT bound V6 */
1564			if (nam->sa_family == AF_INET6 &&
1565			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1566				SCTP_INP_RUNLOCK(inp);
1567				continue;
1568			}
1569			/* does a VRF id match? */
1570			fnd = 0;
1571			if (inp->def_vrf_id == vrf_id)
1572				fnd = 1;
1573
1574			SCTP_INP_RUNLOCK(inp);
1575			if (!fnd)
1576				continue;
1577			return (inp);
1578		}
1579		SCTP_INP_RUNLOCK(inp);
1580	}
1581	if ((nam->sa_family == AF_INET) &&
1582	    (sin->sin_addr.s_addr == INADDR_ANY)) {
1583		/* Can't hunt for one that has no address specified */
1584		return (NULL);
1585	}
1586#ifdef INET6
1587	if ((nam->sa_family == AF_INET6) &&
1588	    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
1589		/* Can't hunt for one that has no address specified */
1590		return (NULL);
1591	}
1592#endif
1593	/*
1594	 * ok, not bound to all so see if we can find a EP bound to this
1595	 * address.
1596	 */
1597	LIST_FOREACH(inp, head, sctp_hash) {
1598		SCTP_INP_RLOCK(inp);
1599		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1600			SCTP_INP_RUNLOCK(inp);
1601			continue;
1602		}
1603		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1604			SCTP_INP_RUNLOCK(inp);
1605			continue;
1606		}
1607		/*
1608		 * Ok this could be a likely candidate, look at all of its
1609		 * addresses
1610		 */
1611		if (inp->sctp_lport != lport) {
1612			SCTP_INP_RUNLOCK(inp);
1613			continue;
1614		}
1615		/* does a VRF id match? */
1616		fnd = 0;
1617		if (inp->def_vrf_id == vrf_id)
1618			fnd = 1;
1619
1620		if (!fnd) {
1621			SCTP_INP_RUNLOCK(inp);
1622			continue;
1623		}
1624		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1625			if (laddr->ifa == NULL) {
1626				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1627				    __FUNCTION__);
1628				continue;
1629			}
1630			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1631			    laddr->ifa);
1632			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1633				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1634				continue;
1635			}
1636			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1637				/* possible, see if it matches */
1638				struct sockaddr_in *intf_addr;
1639
1640				intf_addr = &laddr->ifa->address.sin;
1641				switch (nam->sa_family) {
1642				case AF_INET:
1643					if (sin->sin_addr.s_addr ==
1644					    intf_addr->sin_addr.s_addr) {
1645						SCTP_INP_RUNLOCK(inp);
1646						return (inp);
1647					}
1648					break;
1649#ifdef INET6
1650				case AF_INET6:
1651					intf_addr6 = &laddr->ifa->address.sin6;
1652					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1653					    intf_addr6)) {
1654						SCTP_INP_RUNLOCK(inp);
1655						return (inp);
1656					}
1657					break;
1658#endif
1659				}
1660			}
1661		}
1662		SCTP_INP_RUNLOCK(inp);
1663	}
1664	return (NULL);
1665}
1666
1667
1668static struct sctp_inpcb *
1669sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1670{
1671	struct sctppcbhead *head;
1672	struct sctp_inpcb *t_inp;
1673	int fnd;
1674
1675	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1676	    SCTP_BASE_INFO(hashmark))];
1677	LIST_FOREACH(t_inp, head, sctp_hash) {
1678		if (t_inp->sctp_lport != lport) {
1679			continue;
1680		}
1681		/* is it in the VRF in question */
1682		fnd = 0;
1683		if (t_inp->def_vrf_id == vrf_id)
1684			fnd = 1;
1685		if (!fnd)
1686			continue;
1687
1688		/* This one is in use. */
1689		/* check the v6/v4 binding issue */
1690		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1691		    SCTP_IPV6_V6ONLY(t_inp)) {
1692			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1693				/* collision in V6 space */
1694				return (t_inp);
1695			} else {
1696				/* inp is BOUND_V4 no conflict */
1697				continue;
1698			}
1699		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1700			/* t_inp is bound v4 and v6, conflict always */
1701			return (t_inp);
1702		} else {
1703			/* t_inp is bound only V4 */
1704			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1705			    SCTP_IPV6_V6ONLY(inp)) {
1706				/* no conflict */
1707				continue;
1708			}
1709			/* else fall through to conflict */
1710		}
1711		return (t_inp);
1712	}
1713	return (NULL);
1714}
1715
1716
1717int
1718sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1719{
1720	/* For 1-2-1 with port reuse */
1721	struct sctppcbhead *head;
1722	struct sctp_inpcb *tinp;
1723
1724	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1725		/* only works with port reuse on */
1726		return (-1);
1727	}
1728	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1729		return (0);
1730	}
1731	SCTP_INP_RUNLOCK(inp);
1732	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1733	    SCTP_BASE_INFO(hashmark))];
1734	/* Kick out all non-listeners to the TCP hash */
1735	LIST_FOREACH(tinp, head, sctp_hash) {
1736		if (tinp->sctp_lport != inp->sctp_lport) {
1737			continue;
1738		}
1739		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1740			continue;
1741		}
1742		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1743			continue;
1744		}
1745		if (tinp->sctp_socket->so_qlimit) {
1746			continue;
1747		}
1748		SCTP_INP_WLOCK(tinp);
1749		LIST_REMOVE(tinp, sctp_hash);
1750		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1751		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1752		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1753		SCTP_INP_WUNLOCK(tinp);
1754	}
1755	SCTP_INP_WLOCK(inp);
1756	/* Pull from where he was */
1757	LIST_REMOVE(inp, sctp_hash);
1758	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1759	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1760	LIST_INSERT_HEAD(head, inp, sctp_hash);
1761	SCTP_INP_WUNLOCK(inp);
1762	SCTP_INP_RLOCK(inp);
1763	return (0);
1764}
1765
1766
1767struct sctp_inpcb *
1768sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1769    uint32_t vrf_id)
1770{
1771	/*
1772	 * First we check the hash table to see if someone has this port
1773	 * bound with just the port.
1774	 */
1775	struct sctp_inpcb *inp;
1776	struct sctppcbhead *head;
1777	struct sockaddr_in *sin;
1778	struct sockaddr_in6 *sin6;
1779	int lport;
1780	unsigned int i;
1781
1782	if (nam->sa_family == AF_INET) {
1783		sin = (struct sockaddr_in *)nam;
1784		lport = ((struct sockaddr_in *)nam)->sin_port;
1785	} else if (nam->sa_family == AF_INET6) {
1786		sin6 = (struct sockaddr_in6 *)nam;
1787		lport = ((struct sockaddr_in6 *)nam)->sin6_port;
1788	} else {
1789		/* unsupported family */
1790		return (NULL);
1791	}
1792	/*
1793	 * I could cheat here and just cast to one of the types but we will
1794	 * do it right. It also provides the check against an Unsupported
1795	 * type too.
1796	 */
1797	/* Find the head of the ALLADDR chain */
1798	if (have_lock == 0) {
1799		SCTP_INP_INFO_RLOCK();
1800	}
1801	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1802	    SCTP_BASE_INFO(hashmark))];
1803	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1804
1805	/*
1806	 * If the TCP model exists it could be that the main listening
1807	 * endpoint is gone but there still exists a connected socket for
1808	 * this guy. If so we can return the first one that we find. This
1809	 * may NOT be the correct one so the caller should be wary on the
1810	 * returned INP. Currently the only caller that sets find_tcp_pool
1811	 * is in bindx where we are verifying that a user CAN bind the
1812	 * address. He either has bound it already, or someone else has, or
1813	 * its open to bind, so this is good enough.
1814	 */
1815	if (inp == NULL && find_tcp_pool) {
1816		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1817			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1818			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1819			if (inp) {
1820				break;
1821			}
1822		}
1823	}
1824	if (inp) {
1825		SCTP_INP_INCR_REF(inp);
1826	}
1827	if (have_lock == 0) {
1828		SCTP_INP_INFO_RUNLOCK();
1829	}
1830	return (inp);
1831}
1832
1833/*
1834 * Find an association for an endpoint with the pointer to whom you want to
1835 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1836 * need to change the *to to some other struct like a mbuf...
1837 */
1838struct sctp_tcb *
1839sctp_findassociation_addr_sa(struct sockaddr *to, struct sockaddr *from,
1840    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1841    uint32_t vrf_id)
1842{
1843	struct sctp_inpcb *inp = NULL;
1844	struct sctp_tcb *retval;
1845
1846	SCTP_INP_INFO_RLOCK();
1847	if (find_tcp_pool) {
1848		if (inp_p != NULL) {
1849			retval = sctp_tcb_special_locate(inp_p, from, to, netp,
1850			    vrf_id);
1851		} else {
1852			retval = sctp_tcb_special_locate(&inp, from, to, netp,
1853			    vrf_id);
1854		}
1855		if (retval != NULL) {
1856			SCTP_INP_INFO_RUNLOCK();
1857			return (retval);
1858		}
1859	}
1860	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1861	if (inp_p != NULL) {
1862		*inp_p = inp;
1863	}
1864	SCTP_INP_INFO_RUNLOCK();
1865
1866	if (inp == NULL) {
1867		return (NULL);
1868	}
1869	/*
1870	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1871	 * we now place the source address or from in the to of the find
1872	 * endpoint call. Since in reality this chain is used from the
1873	 * inbound packet side.
1874	 */
1875	if (inp_p != NULL) {
1876		retval = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1877		    NULL);
1878	} else {
1879		retval = sctp_findassociation_ep_addr(&inp, from, netp, to,
1880		    NULL);
1881	}
1882	return retval;
1883}
1884
1885
1886/*
1887 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1888 * find all addresses that the sender has specified in any address list. Each
1889 * address will be used to lookup the TCB and see if one exits.
1890 */
1891static struct sctp_tcb *
1892sctp_findassociation_special_addr(struct mbuf *m, int iphlen, int offset,
1893    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1894    struct sockaddr *dest)
1895{
1896	struct sockaddr_in sin4;
1897	struct sockaddr_in6 sin6;
1898	struct sctp_paramhdr *phdr, parm_buf;
1899	struct sctp_tcb *retval;
1900	uint32_t ptype, plen;
1901
1902	memset(&sin4, 0, sizeof(sin4));
1903	memset(&sin6, 0, sizeof(sin6));
1904	sin4.sin_len = sizeof(sin4);
1905	sin4.sin_family = AF_INET;
1906	sin4.sin_port = sh->src_port;
1907	sin6.sin6_len = sizeof(sin6);
1908	sin6.sin6_family = AF_INET6;
1909	sin6.sin6_port = sh->src_port;
1910
1911	retval = NULL;
1912	offset += sizeof(struct sctp_init_chunk);
1913
1914	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
1915	while (phdr != NULL) {
1916		/* now we must see if we want the parameter */
1917		ptype = ntohs(phdr->param_type);
1918		plen = ntohs(phdr->param_length);
1919		if (plen == 0) {
1920			break;
1921		}
1922		if (ptype == SCTP_IPV4_ADDRESS &&
1923		    plen == sizeof(struct sctp_ipv4addr_param)) {
1924			/* Get the rest of the address */
1925			struct sctp_ipv4addr_param ip4_parm, *p4;
1926
1927			phdr = sctp_get_next_param(m, offset,
1928			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
1929			if (phdr == NULL) {
1930				return (NULL);
1931			}
1932			p4 = (struct sctp_ipv4addr_param *)phdr;
1933			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
1934			/* look it up */
1935			retval = sctp_findassociation_ep_addr(inp_p,
1936			    (struct sockaddr *)&sin4, netp, dest, NULL);
1937			if (retval != NULL) {
1938				return (retval);
1939			}
1940		} else if (ptype == SCTP_IPV6_ADDRESS &&
1941		    plen == sizeof(struct sctp_ipv6addr_param)) {
1942			/* Get the rest of the address */
1943			struct sctp_ipv6addr_param ip6_parm, *p6;
1944
1945			phdr = sctp_get_next_param(m, offset,
1946			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
1947			if (phdr == NULL) {
1948				return (NULL);
1949			}
1950			p6 = (struct sctp_ipv6addr_param *)phdr;
1951			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
1952			/* look it up */
1953			retval = sctp_findassociation_ep_addr(inp_p,
1954			    (struct sockaddr *)&sin6, netp, dest, NULL);
1955			if (retval != NULL) {
1956				return (retval);
1957			}
1958		}
1959		offset += SCTP_SIZE32(plen);
1960		phdr = sctp_get_next_param(m, offset, &parm_buf,
1961		    sizeof(parm_buf));
1962	}
1963	return (NULL);
1964}
1965
1966static struct sctp_tcb *
1967sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
1968    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
1969    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
1970{
1971	/*
1972	 * Use my vtag to hash. If we find it we then verify the source addr
1973	 * is in the assoc. If all goes well we save a bit on rec of a
1974	 * packet.
1975	 */
1976	struct sctpasochead *head;
1977	struct sctp_nets *net;
1978	struct sctp_tcb *stcb;
1979
1980	*netp = NULL;
1981	*inp_p = NULL;
1982	SCTP_INP_INFO_RLOCK();
1983	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
1984	    SCTP_BASE_INFO(hashasocmark))];
1985	if (head == NULL) {
1986		/* invalid vtag */
1987		SCTP_INP_INFO_RUNLOCK();
1988		return (NULL);
1989	}
1990	LIST_FOREACH(stcb, head, sctp_asocs) {
1991		SCTP_INP_RLOCK(stcb->sctp_ep);
1992		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1993			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1994			continue;
1995		}
1996		SCTP_TCB_LOCK(stcb);
1997		SCTP_INP_RUNLOCK(stcb->sctp_ep);
1998		if (stcb->asoc.my_vtag == vtag) {
1999			/* candidate */
2000			if (stcb->rport != rport) {
2001				SCTP_TCB_UNLOCK(stcb);
2002				continue;
2003			}
2004			if (stcb->sctp_ep->sctp_lport != lport) {
2005				SCTP_TCB_UNLOCK(stcb);
2006				continue;
2007			}
2008			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2009				SCTP_TCB_UNLOCK(stcb);
2010				continue;
2011			}
2012			/* RRS:Need toaddr check here */
2013			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2014				/* Endpoint does not own this address */
2015				SCTP_TCB_UNLOCK(stcb);
2016				continue;
2017			}
2018			if (remote_tag) {
2019				/*
2020				 * If we have both vtags that's all we match
2021				 * on
2022				 */
2023				if (stcb->asoc.peer_vtag == remote_tag) {
2024					/*
2025					 * If both tags match we consider it
2026					 * conclusive and check NO
2027					 * source/destination addresses
2028					 */
2029					goto conclusive;
2030				}
2031			}
2032			if (skip_src_check) {
2033		conclusive:
2034				if (from) {
2035					net = sctp_findnet(stcb, from);
2036				} else {
2037					*netp = NULL;	/* unknown */
2038				}
2039				if (inp_p)
2040					*inp_p = stcb->sctp_ep;
2041				SCTP_INP_INFO_RUNLOCK();
2042				return (stcb);
2043			}
2044			net = sctp_findnet(stcb, from);
2045			if (net) {
2046				/* yep its him. */
2047				*netp = net;
2048				SCTP_STAT_INCR(sctps_vtagexpress);
2049				*inp_p = stcb->sctp_ep;
2050				SCTP_INP_INFO_RUNLOCK();
2051				return (stcb);
2052			} else {
2053				/*
2054				 * not him, this should only happen in rare
2055				 * cases so I peg it.
2056				 */
2057				SCTP_STAT_INCR(sctps_vtagbogus);
2058			}
2059		}
2060		SCTP_TCB_UNLOCK(stcb);
2061	}
2062	SCTP_INP_INFO_RUNLOCK();
2063	return (NULL);
2064}
2065
2066/*
2067 * Find an association with the pointer to the inbound IP packet. This can be
2068 * a IPv4 or IPv6 packet.
2069 */
2070struct sctp_tcb *
2071sctp_findassociation_addr(struct mbuf *m, int iphlen, int offset,
2072    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2073    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2074{
2075	int find_tcp_pool;
2076	struct ip *iph;
2077	struct sctp_tcb *retval;
2078	struct sockaddr_storage to_store, from_store;
2079	struct sockaddr *to = (struct sockaddr *)&to_store;
2080	struct sockaddr *from = (struct sockaddr *)&from_store;
2081	struct sctp_inpcb *inp;
2082
2083	iph = mtod(m, struct ip *);
2084	switch (iph->ip_v) {
2085	case IPVERSION:
2086		{
2087			/* its IPv4 */
2088			struct sockaddr_in *from4;
2089
2090			from4 = (struct sockaddr_in *)&from_store;
2091			bzero(from4, sizeof(*from4));
2092			from4->sin_family = AF_INET;
2093			from4->sin_len = sizeof(struct sockaddr_in);
2094			from4->sin_addr.s_addr = iph->ip_src.s_addr;
2095			from4->sin_port = sh->src_port;
2096			break;
2097		}
2098#ifdef INET6
2099	case IPV6_VERSION >> 4:
2100		{
2101			/* its IPv6 */
2102			struct ip6_hdr *ip6;
2103			struct sockaddr_in6 *from6;
2104
2105			ip6 = mtod(m, struct ip6_hdr *);
2106			from6 = (struct sockaddr_in6 *)&from_store;
2107			bzero(from6, sizeof(*from6));
2108			from6->sin6_family = AF_INET6;
2109			from6->sin6_len = sizeof(struct sockaddr_in6);
2110			from6->sin6_addr = ip6->ip6_src;
2111			from6->sin6_port = sh->src_port;
2112			/* Get the scopes in properly to the sin6 addr's */
2113			/* we probably don't need these operations */
2114			(void)sa6_recoverscope(from6);
2115			sa6_embedscope(from6, MODULE_GLOBAL(ip6_use_defzone));
2116			break;
2117		}
2118#endif
2119	default:
2120		/* Currently not supported. */
2121		return (NULL);
2122	}
2123
2124
2125	switch (iph->ip_v) {
2126	case IPVERSION:
2127		{
2128			/* its IPv4 */
2129			struct sockaddr_in *to4;
2130
2131			to4 = (struct sockaddr_in *)&to_store;
2132			bzero(to4, sizeof(*to4));
2133			to4->sin_family = AF_INET;
2134			to4->sin_len = sizeof(struct sockaddr_in);
2135			to4->sin_addr.s_addr = iph->ip_dst.s_addr;
2136			to4->sin_port = sh->dest_port;
2137			break;
2138		}
2139#ifdef INET6
2140	case IPV6_VERSION >> 4:
2141		{
2142			/* its IPv6 */
2143			struct ip6_hdr *ip6;
2144			struct sockaddr_in6 *to6;
2145
2146			ip6 = mtod(m, struct ip6_hdr *);
2147			to6 = (struct sockaddr_in6 *)&to_store;
2148			bzero(to6, sizeof(*to6));
2149			to6->sin6_family = AF_INET6;
2150			to6->sin6_len = sizeof(struct sockaddr_in6);
2151			to6->sin6_addr = ip6->ip6_dst;
2152			to6->sin6_port = sh->dest_port;
2153			/* Get the scopes in properly to the sin6 addr's */
2154			/* we probably don't need these operations */
2155			(void)sa6_recoverscope(to6);
2156			sa6_embedscope(to6, MODULE_GLOBAL(ip6_use_defzone));
2157			break;
2158		}
2159#endif
2160	default:
2161		/* TSNH */
2162		break;
2163	}
2164	if (sh->v_tag) {
2165		/* we only go down this path if vtag is non-zero */
2166		retval = sctp_findassoc_by_vtag(from, to, ntohl(sh->v_tag),
2167		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2168		if (retval) {
2169			return (retval);
2170		}
2171	}
2172	find_tcp_pool = 0;
2173	if ((ch->chunk_type != SCTP_INITIATION) &&
2174	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2175	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2176	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2177		/* Other chunk types go to the tcp pool. */
2178		find_tcp_pool = 1;
2179	}
2180	if (inp_p) {
2181		retval = sctp_findassociation_addr_sa(to, from, inp_p, netp,
2182		    find_tcp_pool, vrf_id);
2183		inp = *inp_p;
2184	} else {
2185		retval = sctp_findassociation_addr_sa(to, from, &inp, netp,
2186		    find_tcp_pool, vrf_id);
2187	}
2188	SCTPDBG(SCTP_DEBUG_PCB1, "retval:%p inp:%p\n", retval, inp);
2189	if (retval == NULL && inp) {
2190		/* Found a EP but not this address */
2191		if ((ch->chunk_type == SCTP_INITIATION) ||
2192		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2193			/*-
2194			 * special hook, we do NOT return linp or an
2195			 * association that is linked to an existing
2196			 * association that is under the TCP pool (i.e. no
2197			 * listener exists). The endpoint finding routine
2198			 * will always find a listener before examining the
2199			 * TCP pool.
2200			 */
2201			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2202				if (inp_p) {
2203					*inp_p = NULL;
2204				}
2205				return (NULL);
2206			}
2207			retval = sctp_findassociation_special_addr(m, iphlen,
2208			    offset, sh, &inp, netp, to);
2209			if (inp_p != NULL) {
2210				*inp_p = inp;
2211			}
2212		}
2213	}
2214	SCTPDBG(SCTP_DEBUG_PCB1, "retval is %p\n", retval);
2215	return (retval);
2216}
2217
2218/*
2219 * lookup an association by an ASCONF lookup address.
2220 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2221 */
2222struct sctp_tcb *
2223sctp_findassociation_ep_asconf(struct mbuf *m, int iphlen, int offset,
2224    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2225{
2226	struct sctp_tcb *stcb;
2227	struct sockaddr_in *sin;
2228
2229#ifdef INET6
2230	struct sockaddr_in6 *sin6;
2231
2232#endif
2233	struct sockaddr_storage local_store, remote_store;
2234	struct sockaddr *to;
2235	struct ip *iph;
2236
2237#ifdef INET6
2238	struct ip6_hdr *ip6;
2239
2240#endif
2241	struct sctp_paramhdr parm_buf, *phdr;
2242	int ptype;
2243	int zero_address = 0;
2244
2245
2246	memset(&local_store, 0, sizeof(local_store));
2247	memset(&remote_store, 0, sizeof(remote_store));
2248	to = (struct sockaddr *)&local_store;
2249	/* First get the destination address setup too. */
2250	iph = mtod(m, struct ip *);
2251	switch (iph->ip_v) {
2252	case IPVERSION:
2253		/* its IPv4 */
2254		sin = (struct sockaddr_in *)&local_store;
2255		sin->sin_family = AF_INET;
2256		sin->sin_len = sizeof(*sin);
2257		sin->sin_port = sh->dest_port;
2258		sin->sin_addr.s_addr = iph->ip_dst.s_addr;
2259		break;
2260#ifdef INET6
2261	case IPV6_VERSION >> 4:
2262		/* its IPv6 */
2263		ip6 = mtod(m, struct ip6_hdr *);
2264		sin6 = (struct sockaddr_in6 *)&local_store;
2265		sin6->sin6_family = AF_INET6;
2266		sin6->sin6_len = sizeof(*sin6);
2267		sin6->sin6_port = sh->dest_port;
2268		sin6->sin6_addr = ip6->ip6_dst;
2269		break;
2270#endif
2271	default:
2272		return NULL;
2273	}
2274
2275	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2276	    &parm_buf, sizeof(struct sctp_paramhdr));
2277	if (phdr == NULL) {
2278		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2279		    __FUNCTION__);
2280		return NULL;
2281	}
2282	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2283	/* get the correlation address */
2284	switch (ptype) {
2285#ifdef INET6
2286	case SCTP_IPV6_ADDRESS:
2287		{
2288			/* ipv6 address param */
2289			struct sctp_ipv6addr_param *p6, p6_buf;
2290
2291			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2292				return NULL;
2293			}
2294			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2295			    offset + sizeof(struct sctp_asconf_chunk),
2296			    &p6_buf.ph, sizeof(*p6));
2297			if (p6 == NULL) {
2298				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2299				    __FUNCTION__);
2300				return (NULL);
2301			}
2302			sin6 = (struct sockaddr_in6 *)&remote_store;
2303			sin6->sin6_family = AF_INET6;
2304			sin6->sin6_len = sizeof(*sin6);
2305			sin6->sin6_port = sh->src_port;
2306			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2307			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2308				zero_address = 1;
2309			break;
2310		}
2311#endif
2312	case SCTP_IPV4_ADDRESS:
2313		{
2314			/* ipv4 address param */
2315			struct sctp_ipv4addr_param *p4, p4_buf;
2316
2317			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2318				return NULL;
2319			}
2320			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2321			    offset + sizeof(struct sctp_asconf_chunk),
2322			    &p4_buf.ph, sizeof(*p4));
2323			if (p4 == NULL) {
2324				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2325				    __FUNCTION__);
2326				return (NULL);
2327			}
2328			sin = (struct sockaddr_in *)&remote_store;
2329			sin->sin_family = AF_INET;
2330			sin->sin_len = sizeof(*sin);
2331			sin->sin_port = sh->src_port;
2332			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2333			if (sin->sin_addr.s_addr == INADDR_ANY)
2334				zero_address = 1;
2335			break;
2336		}
2337	default:
2338		/* invalid address param type */
2339		return NULL;
2340	}
2341
2342	if (zero_address) {
2343		stcb = sctp_findassoc_by_vtag(NULL, to, ntohl(sh->v_tag), inp_p,
2344		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2345		/*
2346		 * printf("findassociation_ep_asconf: zero lookup address
2347		 * finds stcb 0x%x\n", (uint32_t)stcb);
2348		 */
2349	} else {
2350		stcb = sctp_findassociation_ep_addr(inp_p,
2351		    (struct sockaddr *)&remote_store, netp,
2352		    to, NULL);
2353	}
2354	return (stcb);
2355}
2356
2357
2358/*
2359 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2360 * addresses. This way if we don't get a bind we by default pick a ephemeral
2361 * port with all addresses bound.
2362 */
2363int
2364sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2365{
2366	/*
2367	 * we get called when a new endpoint starts up. We need to allocate
2368	 * the sctp_inpcb structure from the zone and init it. Mark it as
2369	 * unbound and find a port that we can use as an ephemeral with
2370	 * INADDR_ANY. If the user binds later no problem we can then add in
2371	 * the specific addresses. And setup the default parameters for the
2372	 * EP.
2373	 */
2374	int i, error;
2375	struct sctp_inpcb *inp;
2376	struct sctp_pcb *m;
2377	struct timeval time;
2378	sctp_sharedkey_t *null_key;
2379
2380	error = 0;
2381
2382	SCTP_INP_INFO_WLOCK();
2383	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2384	if (inp == NULL) {
2385		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2386		SCTP_INP_INFO_WUNLOCK();
2387		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2388		return (ENOBUFS);
2389	}
2390	/* zap it */
2391	bzero(inp, sizeof(*inp));
2392
2393	/* bump generations */
2394	/* setup socket pointers */
2395	inp->sctp_socket = so;
2396	inp->ip_inp.inp.inp_socket = so;
2397	inp->sctp_associd_counter = 1;
2398	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2399	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2400	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2401	inp->sctp_ecn_enable = SCTP_BASE_SYSCTL(sctp_ecn_enable);
2402	/* init the small hash table we use to track asocid <-> tcb */
2403	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2404	if (inp->sctp_asocidhash == NULL) {
2405		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2406		SCTP_INP_INFO_WUNLOCK();
2407		return (ENOBUFS);
2408	}
2409#ifdef IPSEC
2410	{
2411		struct inpcbpolicy *pcb_sp = NULL;
2412
2413		error = ipsec_init_policy(so, &pcb_sp);
2414		/* Arrange to share the policy */
2415		inp->ip_inp.inp.inp_sp = pcb_sp;
2416		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2417	}
2418	if (error != 0) {
2419		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2420		SCTP_INP_INFO_WUNLOCK();
2421		return error;
2422	}
2423#endif				/* IPSEC */
2424	SCTP_INCR_EP_COUNT();
2425	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2426	SCTP_INP_INFO_WUNLOCK();
2427
2428	so->so_pcb = (caddr_t)inp;
2429
2430	if ((SCTP_SO_TYPE(so) == SOCK_DGRAM) ||
2431	    (SCTP_SO_TYPE(so) == SOCK_SEQPACKET)) {
2432		/* UDP style socket */
2433		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2434		    SCTP_PCB_FLAGS_UNBOUND);
2435		/* Be sure it is NON-BLOCKING IO for UDP */
2436		/* SCTP_SET_SO_NBIO(so); */
2437	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2438		/* TCP style socket */
2439		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2440		    SCTP_PCB_FLAGS_UNBOUND);
2441		/* Be sure we have blocking IO by default */
2442		SCTP_CLEAR_SO_NBIO(so);
2443	} else {
2444		/*
2445		 * unsupported socket type (RAW, etc)- in case we missed it
2446		 * in protosw
2447		 */
2448		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2449		so->so_pcb = NULL;
2450		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2451		return (EOPNOTSUPP);
2452	}
2453	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2454		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2455		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2456	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2457		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2458		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2459	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2460		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2461		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2462	}
2463	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2464	    &inp->sctp_hashmark);
2465	if (inp->sctp_tcbhash == NULL) {
2466		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2467		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2468		so->so_pcb = NULL;
2469		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2470		return (ENOBUFS);
2471	}
2472	inp->def_vrf_id = vrf_id;
2473
2474	SCTP_INP_INFO_WLOCK();
2475	SCTP_INP_LOCK_INIT(inp);
2476	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2477	SCTP_INP_READ_INIT(inp);
2478	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2479	/* lock the new ep */
2480	SCTP_INP_WLOCK(inp);
2481
2482	/* add it to the info area */
2483	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2484	SCTP_INP_INFO_WUNLOCK();
2485
2486	TAILQ_INIT(&inp->read_queue);
2487	LIST_INIT(&inp->sctp_addr_list);
2488
2489	LIST_INIT(&inp->sctp_asoc_list);
2490
2491#ifdef SCTP_TRACK_FREED_ASOCS
2492	/* TEMP CODE */
2493	LIST_INIT(&inp->sctp_asoc_free_list);
2494#endif
2495	/* Init the timer structure for signature change */
2496	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2497	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2498
2499	/* now init the actual endpoint default data */
2500	m = &inp->sctp_ep;
2501
2502	/* setup the base timeout information */
2503	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2504	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2505	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2506	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2507	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2508	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2509	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2510	/* all max/min max are in ms */
2511	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2512	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2513	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2514	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2515	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2516
2517	m->max_open_streams_intome = MAX_SCTP_STREAMS;
2518
2519	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2520	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2521	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2522	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2523	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2524	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2525	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2526
2527	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2528	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2529	/* number of streams to pre-open on a association */
2530	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2531
2532	/* Add adaptation cookie */
2533	m->adaptation_layer_indicator = 0x504C5253;
2534
2535	/* seed random number generator */
2536	m->random_counter = 1;
2537	m->store_at = SCTP_SIGNATURE_SIZE;
2538	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2539	sctp_fill_random_store(m);
2540
2541	/* Minimum cookie size */
2542	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2543	    sizeof(struct sctp_state_cookie);
2544	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2545
2546	/* Setup the initial secret */
2547	(void)SCTP_GETTIME_TIMEVAL(&time);
2548	m->time_of_secret_change = time.tv_sec;
2549
2550	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2551		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2552	}
2553	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2554
2555	/* How long is a cookie good for ? */
2556	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2557	/*
2558	 * Initialize authentication parameters
2559	 */
2560	m->local_hmacs = sctp_default_supported_hmaclist();
2561	m->local_auth_chunks = sctp_alloc_chunklist();
2562	sctp_auth_set_default_chunks(m->local_auth_chunks);
2563	LIST_INIT(&m->shared_keys);
2564	/* add default NULL key as key id 0 */
2565	null_key = sctp_alloc_sharedkey();
2566	sctp_insert_sharedkey(&m->shared_keys, null_key);
2567	SCTP_INP_WUNLOCK(inp);
2568#ifdef SCTP_LOG_CLOSING
2569	sctp_log_closing(inp, NULL, 12);
2570#endif
2571	return (error);
2572}
2573
2574
2575void
2576sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2577    struct sctp_tcb *stcb)
2578{
2579	struct sctp_nets *net;
2580	uint16_t lport, rport;
2581	struct sctppcbhead *head;
2582	struct sctp_laddr *laddr, *oladdr;
2583
2584	atomic_add_int(&stcb->asoc.refcnt, 1);
2585	SCTP_TCB_UNLOCK(stcb);
2586	SCTP_INP_INFO_WLOCK();
2587	SCTP_INP_WLOCK(old_inp);
2588	SCTP_INP_WLOCK(new_inp);
2589	SCTP_TCB_LOCK(stcb);
2590	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2591
2592	new_inp->sctp_ep.time_of_secret_change =
2593	    old_inp->sctp_ep.time_of_secret_change;
2594	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2595	    sizeof(old_inp->sctp_ep.secret_key));
2596	new_inp->sctp_ep.current_secret_number =
2597	    old_inp->sctp_ep.current_secret_number;
2598	new_inp->sctp_ep.last_secret_number =
2599	    old_inp->sctp_ep.last_secret_number;
2600	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2601
2602	/* make it so new data pours into the new socket */
2603	stcb->sctp_socket = new_inp->sctp_socket;
2604	stcb->sctp_ep = new_inp;
2605
2606	/* Copy the port across */
2607	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2608	rport = stcb->rport;
2609	/* Pull the tcb from the old association */
2610	LIST_REMOVE(stcb, sctp_tcbhash);
2611	LIST_REMOVE(stcb, sctp_tcblist);
2612	if (stcb->asoc.in_asocid_hash) {
2613		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2614	}
2615	/* Now insert the new_inp into the TCP connected hash */
2616	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2617
2618	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2619	/* Its safe to access */
2620	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2621
2622	/* Now move the tcb into the endpoint list */
2623	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2624	/*
2625	 * Question, do we even need to worry about the ep-hash since we
2626	 * only have one connection? Probably not :> so lets get rid of it
2627	 * and not suck up any kernel memory in that.
2628	 */
2629	if (stcb->asoc.in_asocid_hash) {
2630		struct sctpasochead *lhd;
2631
2632		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2633		    new_inp->hashasocidmark)];
2634		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2635	}
2636	/* Ok. Let's restart timer. */
2637	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2638		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2639		    stcb, net);
2640	}
2641
2642	SCTP_INP_INFO_WUNLOCK();
2643	if (new_inp->sctp_tcbhash != NULL) {
2644		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2645		new_inp->sctp_tcbhash = NULL;
2646	}
2647	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2648		/* Subset bound, so copy in the laddr list from the old_inp */
2649		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2650			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2651			if (laddr == NULL) {
2652				/*
2653				 * Gak, what can we do? This assoc is really
2654				 * HOSED. We probably should send an abort
2655				 * here.
2656				 */
2657				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2658				continue;
2659			}
2660			SCTP_INCR_LADDR_COUNT();
2661			bzero(laddr, sizeof(*laddr));
2662			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2663			laddr->ifa = oladdr->ifa;
2664			atomic_add_int(&laddr->ifa->refcount, 1);
2665			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2666			    sctp_nxt_addr);
2667			new_inp->laddr_count++;
2668		}
2669	}
2670	/*
2671	 * Now any running timers need to be adjusted since we really don't
2672	 * care if they are running or not just blast in the new_inp into
2673	 * all of them.
2674	 */
2675
2676	stcb->asoc.hb_timer.ep = (void *)new_inp;
2677	stcb->asoc.dack_timer.ep = (void *)new_inp;
2678	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2679	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2680	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2681	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2682	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2683	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2684	/* now what about the nets? */
2685	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2686		net->pmtu_timer.ep = (void *)new_inp;
2687		net->rxt_timer.ep = (void *)new_inp;
2688		net->fr_timer.ep = (void *)new_inp;
2689	}
2690	SCTP_INP_WUNLOCK(new_inp);
2691	SCTP_INP_WUNLOCK(old_inp);
2692}
2693
2694
2695
2696
2697/* sctp_ifap is used to bypass normal local address validation checks */
2698int
2699sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2700    struct sctp_ifa *sctp_ifap, struct thread *p)
2701{
2702	/* bind a ep to a socket address */
2703	struct sctppcbhead *head;
2704	struct sctp_inpcb *inp, *inp_tmp;
2705	struct inpcb *ip_inp;
2706	int port_reuse_active = 0;
2707	int bindall;
2708	uint16_t lport;
2709	int error;
2710	uint32_t vrf_id;
2711
2712	lport = 0;
2713	error = 0;
2714	bindall = 1;
2715	inp = (struct sctp_inpcb *)so->so_pcb;
2716	ip_inp = (struct inpcb *)so->so_pcb;
2717#ifdef SCTP_DEBUG
2718	if (addr) {
2719		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n",
2720		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2721		SCTPDBG(SCTP_DEBUG_PCB1, "Addr :");
2722		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2723	}
2724#endif
2725	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2726		/* already did a bind, subsequent binds NOT allowed ! */
2727		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2728		return (EINVAL);
2729	}
2730#ifdef INVARIANTS
2731	if (p == NULL)
2732		panic("null proc/thread");
2733#endif
2734	if (addr != NULL) {
2735		switch (addr->sa_family) {
2736		case AF_INET:
2737			{
2738				struct sockaddr_in *sin;
2739
2740				/* IPV6_V6ONLY socket? */
2741				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2742					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2743					return (EINVAL);
2744				}
2745				if (addr->sa_len != sizeof(*sin)) {
2746					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2747					return (EINVAL);
2748				}
2749				sin = (struct sockaddr_in *)addr;
2750				lport = sin->sin_port;
2751				/*
2752				 * For LOOPBACK the prison_local_ip4() call
2753				 * will transmute the ip address to the
2754				 * proper value.
2755				 */
2756				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2757					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2758					return (error);
2759				}
2760				if (sin->sin_addr.s_addr != INADDR_ANY) {
2761					bindall = 0;
2762				}
2763				break;
2764			}
2765#ifdef INET6
2766		case AF_INET6:
2767			{
2768				/*
2769				 * Only for pure IPv6 Address. (No IPv4
2770				 * Mapped!)
2771				 */
2772				struct sockaddr_in6 *sin6;
2773
2774				sin6 = (struct sockaddr_in6 *)addr;
2775
2776				if (addr->sa_len != sizeof(*sin6)) {
2777					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2778					return (EINVAL);
2779				}
2780				lport = sin6->sin6_port;
2781
2782				/*
2783				 * For LOOPBACK the prison_local_ip6() call
2784				 * will transmute the ipv6 address to the
2785				 * proper value.
2786				 */
2787				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2788				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2789					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2790					return (error);
2791				}
2792				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2793					bindall = 0;
2794					/* KAME hack: embed scopeid */
2795					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2796						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2797						return (EINVAL);
2798					}
2799				}
2800				/* this must be cleared for ifa_ifwithaddr() */
2801				sin6->sin6_scope_id = 0;
2802				break;
2803			}
2804#endif
2805		default:
2806			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2807			return (EAFNOSUPPORT);
2808		}
2809	}
2810	SCTP_INP_INFO_WLOCK();
2811	SCTP_INP_WLOCK(inp);
2812	/* Setup a vrf_id to be the default for the non-bind-all case. */
2813	vrf_id = inp->def_vrf_id;
2814
2815	/* increase our count due to the unlock we do */
2816	SCTP_INP_INCR_REF(inp);
2817	if (lport) {
2818		/*
2819		 * Did the caller specify a port? if so we must see if a ep
2820		 * already has this one bound.
2821		 */
2822		/* got to be root to get at low ports */
2823		if (ntohs(lport) < IPPORT_RESERVED) {
2824			if (p && (error =
2825			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2826			    )) {
2827				SCTP_INP_DECR_REF(inp);
2828				SCTP_INP_WUNLOCK(inp);
2829				SCTP_INP_INFO_WUNLOCK();
2830				return (error);
2831			}
2832		}
2833		if (p == NULL) {
2834			SCTP_INP_DECR_REF(inp);
2835			SCTP_INP_WUNLOCK(inp);
2836			SCTP_INP_INFO_WUNLOCK();
2837			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2838			return (error);
2839		}
2840		SCTP_INP_WUNLOCK(inp);
2841		if (bindall) {
2842			vrf_id = inp->def_vrf_id;
2843			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2844			if (inp_tmp != NULL) {
2845				/*
2846				 * lock guy returned and lower count note
2847				 * that we are not bound so inp_tmp should
2848				 * NEVER be inp. And it is this inp
2849				 * (inp_tmp) that gets the reference bump,
2850				 * so we must lower it.
2851				 */
2852				SCTP_INP_DECR_REF(inp_tmp);
2853				/* unlock info */
2854				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2855				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2856					/*
2857					 * Ok, must be one-2-one and
2858					 * allowing port re-use
2859					 */
2860					port_reuse_active = 1;
2861					goto continue_anyway;
2862				}
2863				SCTP_INP_DECR_REF(inp);
2864				SCTP_INP_INFO_WUNLOCK();
2865				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2866				return (EADDRINUSE);
2867			}
2868		} else {
2869			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2870			if (inp_tmp != NULL) {
2871				/*
2872				 * lock guy returned and lower count note
2873				 * that we are not bound so inp_tmp should
2874				 * NEVER be inp. And it is this inp
2875				 * (inp_tmp) that gets the reference bump,
2876				 * so we must lower it.
2877				 */
2878				SCTP_INP_DECR_REF(inp_tmp);
2879				/* unlock info */
2880				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2881				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2882					/*
2883					 * Ok, must be one-2-one and
2884					 * allowing port re-use
2885					 */
2886					port_reuse_active = 1;
2887					goto continue_anyway;
2888				}
2889				SCTP_INP_DECR_REF(inp);
2890				SCTP_INP_INFO_WUNLOCK();
2891				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2892				return (EADDRINUSE);
2893			}
2894		}
2895continue_anyway:
2896		SCTP_INP_WLOCK(inp);
2897		if (bindall) {
2898			/* verify that no lport is not used by a singleton */
2899			if ((port_reuse_active == 0) &&
2900			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))
2901			    ) {
2902				/* Sorry someone already has this one bound */
2903				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2904				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2905					port_reuse_active = 1;
2906				} else {
2907					SCTP_INP_DECR_REF(inp);
2908					SCTP_INP_WUNLOCK(inp);
2909					SCTP_INP_INFO_WUNLOCK();
2910					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2911					return (EADDRINUSE);
2912				}
2913			}
2914		}
2915	} else {
2916		uint16_t first, last, candidate;
2917		uint16_t count;
2918		int done;
2919
2920		if (ip_inp->inp_flags & INP_HIGHPORT) {
2921			first = MODULE_GLOBAL(ipport_hifirstauto);
2922			last = MODULE_GLOBAL(ipport_hilastauto);
2923		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2924			if (p && (error =
2925			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2926			    )) {
2927				SCTP_INP_DECR_REF(inp);
2928				SCTP_INP_WUNLOCK(inp);
2929				SCTP_INP_INFO_WUNLOCK();
2930				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2931				return (error);
2932			}
2933			first = MODULE_GLOBAL(ipport_lowfirstauto);
2934			last = MODULE_GLOBAL(ipport_lowlastauto);
2935		} else {
2936			first = MODULE_GLOBAL(ipport_firstauto);
2937			last = MODULE_GLOBAL(ipport_lastauto);
2938		}
2939		if (first > last) {
2940			uint16_t temp;
2941
2942			temp = first;
2943			first = last;
2944			last = temp;
2945		}
2946		count = last - first + 1;	/* number of candidates */
2947		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2948
2949		done = 0;
2950		while (!done) {
2951			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
2952				done = 1;
2953			}
2954			if (!done) {
2955				if (--count == 0) {
2956					SCTP_INP_DECR_REF(inp);
2957					SCTP_INP_WUNLOCK(inp);
2958					SCTP_INP_INFO_WUNLOCK();
2959					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2960					return (EADDRINUSE);
2961				}
2962				if (candidate == last)
2963					candidate = first;
2964				else
2965					candidate = candidate + 1;
2966			}
2967		}
2968		lport = htons(candidate);
2969	}
2970	SCTP_INP_DECR_REF(inp);
2971	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2972	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2973		/*
2974		 * this really should not happen. The guy did a non-blocking
2975		 * bind and then did a close at the same time.
2976		 */
2977		SCTP_INP_WUNLOCK(inp);
2978		SCTP_INP_INFO_WUNLOCK();
2979		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2980		return (EINVAL);
2981	}
2982	/* ok we look clear to give out this port, so lets setup the binding */
2983	if (bindall) {
2984		/* binding to all addresses, so just set in the proper flags */
2985		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
2986		/* set the automatic addr changes from kernel flag */
2987		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
2988			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2989			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2990		} else {
2991			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2992			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2993		}
2994		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
2995			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2996		} else {
2997			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2998		}
2999		/*
3000		 * set the automatic mobility_base from kernel flag (by
3001		 * micchie)
3002		 */
3003		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3004			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3005			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3006		} else {
3007			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3008			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3009		}
3010		/*
3011		 * set the automatic mobility_fasthandoff from kernel flag
3012		 * (by micchie)
3013		 */
3014		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3015			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3016			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3017		} else {
3018			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3019			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3020		}
3021	} else {
3022		/*
3023		 * bind specific, make sure flags is off and add a new
3024		 * address structure to the sctp_addr_list inside the ep
3025		 * structure.
3026		 *
3027		 * We will need to allocate one and insert it at the head. The
3028		 * socketopt call can just insert new addresses in there as
3029		 * well. It will also have to do the embed scope kame hack
3030		 * too (before adding).
3031		 */
3032		struct sctp_ifa *ifa;
3033		struct sockaddr_storage store_sa;
3034
3035		memset(&store_sa, 0, sizeof(store_sa));
3036		if (addr->sa_family == AF_INET) {
3037			struct sockaddr_in *sin;
3038
3039			sin = (struct sockaddr_in *)&store_sa;
3040			memcpy(sin, addr, sizeof(struct sockaddr_in));
3041			sin->sin_port = 0;
3042		} else if (addr->sa_family == AF_INET6) {
3043			struct sockaddr_in6 *sin6;
3044
3045			sin6 = (struct sockaddr_in6 *)&store_sa;
3046			memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3047			sin6->sin6_port = 0;
3048		}
3049		/*
3050		 * first find the interface with the bound address need to
3051		 * zero out the port to find the address! yuck! can't do
3052		 * this earlier since need port for sctp_pcb_findep()
3053		 */
3054		if (sctp_ifap != NULL)
3055			ifa = sctp_ifap;
3056		else {
3057			/*
3058			 * Note for BSD we hit here always other O/S's will
3059			 * pass things in via the sctp_ifap argument
3060			 * (Panda).
3061			 */
3062			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3063			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3064		}
3065		if (ifa == NULL) {
3066			/* Can't find an interface with that address */
3067			SCTP_INP_WUNLOCK(inp);
3068			SCTP_INP_INFO_WUNLOCK();
3069			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3070			return (EADDRNOTAVAIL);
3071		}
3072		if (addr->sa_family == AF_INET6) {
3073			/* GAK, more FIXME IFA lock? */
3074			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3075				/* Can't bind a non-existent addr. */
3076				SCTP_INP_WUNLOCK(inp);
3077				SCTP_INP_INFO_WUNLOCK();
3078				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3079				return (EINVAL);
3080			}
3081		}
3082		/* we're not bound all */
3083		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3084		/* allow bindx() to send ASCONF's for binding changes */
3085		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3086		/* clear automatic addr changes from kernel flag */
3087		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3088
3089		/* add this address to the endpoint list */
3090		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3091		if (error != 0) {
3092			SCTP_INP_WUNLOCK(inp);
3093			SCTP_INP_INFO_WUNLOCK();
3094			return (error);
3095		}
3096		inp->laddr_count++;
3097	}
3098	/* find the bucket */
3099	if (port_reuse_active) {
3100		/* Put it into tcp 1-2-1 hash */
3101		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3102		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3103	} else {
3104		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3105	}
3106	/* put it in the bucket */
3107	LIST_INSERT_HEAD(head, inp, sctp_hash);
3108	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3109	    head, ntohs(lport), port_reuse_active);
3110	/* set in the port */
3111	inp->sctp_lport = lport;
3112
3113	/* turn off just the unbound flag */
3114	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3115	SCTP_INP_WUNLOCK(inp);
3116	SCTP_INP_INFO_WUNLOCK();
3117	return (0);
3118}
3119
3120
3121static void
3122sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3123{
3124	struct sctp_iterator *it, *nit;
3125
3126	/*
3127	 * We enter with the only the ITERATOR_LOCK in place and a write
3128	 * lock on the inp_info stuff.
3129	 */
3130	it = sctp_it_ctl.cur_it;
3131	if (it && (it->vn != curvnet)) {
3132		/* Its not looking at our VNET */
3133		return;
3134	}
3135	if (it && (it->inp == inp)) {
3136		/*
3137		 * This is tricky and we hold the iterator lock, but when it
3138		 * returns and gets the lock (when we release it) the
3139		 * iterator will try to operate on inp. We need to stop that
3140		 * from happening. But of course the iterator has a
3141		 * reference on the stcb and inp. We can mark it and it will
3142		 * stop.
3143		 *
3144		 * If its a single iterator situation, we set the end iterator
3145		 * flag. Otherwise we set the iterator to go to the next
3146		 * inp.
3147		 *
3148		 */
3149		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3150			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3151		} else {
3152			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3153		}
3154	}
3155	/*
3156	 * Now go through and remove any single reference to our inp that
3157	 * may be still pending on the list
3158	 */
3159	SCTP_IPI_ITERATOR_WQ_LOCK();
3160	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3161		if (it->vn != curvnet) {
3162			continue;
3163		}
3164		if (it->inp == inp) {
3165			/* This one points to me is it inp specific? */
3166			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3167				/* Remove and free this one */
3168				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3169				    it, sctp_nxt_itr);
3170				if (it->function_atend != NULL) {
3171					(*it->function_atend) (it->pointer, it->val);
3172				}
3173				SCTP_FREE(it, SCTP_M_ITER);
3174			} else {
3175				it->inp = LIST_NEXT(it->inp, sctp_list);
3176				if (it->inp) {
3177					SCTP_INP_INCR_REF(it->inp);
3178				}
3179			}
3180			/*
3181			 * When its put in the refcnt is incremented so decr
3182			 * it
3183			 */
3184			SCTP_INP_DECR_REF(inp);
3185		}
3186	}
3187	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3188}
3189
3190/* release sctp_inpcb unbind the port */
3191void
3192sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3193{
3194	/*
3195	 * Here we free a endpoint. We must find it (if it is in the Hash
3196	 * table) and remove it from there. Then we must also find it in the
3197	 * overall list and remove it from there. After all removals are
3198	 * complete then any timer has to be stopped. Then start the actual
3199	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3200	 * all associations. d) finally the ep itself.
3201	 */
3202	struct sctp_pcb *m;
3203	struct sctp_tcb *asoc, *nasoc;
3204	struct sctp_laddr *laddr, *nladdr;
3205	struct inpcb *ip_pcb;
3206	struct socket *so;
3207	int being_refed = 0;
3208	struct sctp_queued_to_read *sq, *nsq;
3209	int cnt;
3210	sctp_sharedkey_t *shared_key, *nshared_key;
3211
3212
3213#ifdef SCTP_LOG_CLOSING
3214	sctp_log_closing(inp, NULL, 0);
3215#endif
3216	SCTP_ITERATOR_LOCK();
3217	/* mark any iterators on the list or being processed */
3218	sctp_iterator_inp_being_freed(inp);
3219	SCTP_ITERATOR_UNLOCK();
3220	so = inp->sctp_socket;
3221	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3222		/* been here before.. eeks.. get out of here */
3223		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3224#ifdef SCTP_LOG_CLOSING
3225		sctp_log_closing(inp, NULL, 1);
3226#endif
3227		return;
3228	}
3229	SCTP_ASOC_CREATE_LOCK(inp);
3230	SCTP_INP_INFO_WLOCK();
3231
3232	SCTP_INP_WLOCK(inp);
3233	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3234		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3235		/* socket is gone, so no more wakeups allowed */
3236		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3237		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3238		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3239
3240	}
3241	/* First time through we have the socket lock, after that no more. */
3242	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3243	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3244
3245	if (inp->control) {
3246		sctp_m_freem(inp->control);
3247		inp->control = NULL;
3248	}
3249	if (inp->pkt) {
3250		sctp_m_freem(inp->pkt);
3251		inp->pkt = NULL;
3252	}
3253	m = &inp->sctp_ep;
3254	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3255					 * here but I will be nice :> (i.e.
3256					 * ip_pcb = ep;) */
3257	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3258		int cnt_in_sd;
3259
3260		cnt_in_sd = 0;
3261		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3262			SCTP_TCB_LOCK(asoc);
3263			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3264				/* Skip guys being freed */
3265				cnt_in_sd++;
3266				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3267					/*
3268					 * Special case - we did not start a
3269					 * kill timer on the asoc due to it
3270					 * was not closed. So go ahead and
3271					 * start it now.
3272					 */
3273					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3274					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3275				}
3276				SCTP_TCB_UNLOCK(asoc);
3277				continue;
3278			}
3279			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3280			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3281			    (asoc->asoc.total_output_queue_size == 0)) {
3282				/*
3283				 * If we have data in queue, we don't want
3284				 * to just free since the app may have done,
3285				 * send()/close or connect/send/close. And
3286				 * it wants the data to get across first.
3287				 */
3288				/* Just abandon things in the front states */
3289				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3290				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3291					cnt_in_sd++;
3292				}
3293				continue;
3294			}
3295			/* Disconnect the socket please */
3296			asoc->sctp_socket = NULL;
3297			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3298			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3299			    (asoc->asoc.control_pdapi) ||
3300			    (asoc->asoc.size_on_all_streams > 0) ||
3301			    (so && (so->so_rcv.sb_cc > 0))
3302			    ) {
3303				/* Left with Data unread */
3304				struct mbuf *op_err;
3305
3306				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3307				    0, M_DONTWAIT, 1, MT_DATA);
3308				if (op_err) {
3309					/* Fill in the user initiated abort */
3310					struct sctp_paramhdr *ph;
3311					uint32_t *ippp;
3312
3313					SCTP_BUF_LEN(op_err) =
3314					    sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
3315					ph = mtod(op_err,
3316					    struct sctp_paramhdr *);
3317					ph->param_type = htons(
3318					    SCTP_CAUSE_USER_INITIATED_ABT);
3319					ph->param_length = htons(SCTP_BUF_LEN(op_err));
3320					ippp = (uint32_t *) (ph + 1);
3321					*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3);
3322				}
3323				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3324#if defined(SCTP_PANIC_ON_ABORT)
3325				panic("inpcb_free does an abort");
3326#endif
3327				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3328				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3329				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3330				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3331					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3332				}
3333				if (sctp_free_assoc(inp, asoc,
3334				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3335					cnt_in_sd++;
3336				}
3337				continue;
3338			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3339				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3340				    (asoc->asoc.stream_queue_cnt == 0)
3341			    ) {
3342				if (asoc->asoc.locked_on_sending) {
3343					goto abort_anyway;
3344				}
3345				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3346				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3347					/*
3348					 * there is nothing queued to send,
3349					 * so I send shutdown
3350					 */
3351					sctp_send_shutdown(asoc, asoc->asoc.primary_destination);
3352					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3353					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3354						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3355					}
3356					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3357					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3358					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3359					    asoc->asoc.primary_destination);
3360					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3361					    asoc->asoc.primary_destination);
3362					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3363				}
3364			} else {
3365				/* mark into shutdown pending */
3366				struct sctp_stream_queue_pending *sp;
3367
3368				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3369				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3370				    asoc->asoc.primary_destination);
3371				if (asoc->asoc.locked_on_sending) {
3372					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3373					    sctp_streamhead);
3374					if (sp == NULL) {
3375						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3376						    asoc->asoc.locked_on_sending,
3377						    asoc->asoc.locked_on_sending->stream_no);
3378					} else {
3379						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3380							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3381					}
3382				}
3383				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3384				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3385				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3386					struct mbuf *op_err;
3387
3388			abort_anyway:
3389					op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3390					    0, M_DONTWAIT, 1, MT_DATA);
3391					if (op_err) {
3392						/*
3393						 * Fill in the user
3394						 * initiated abort
3395						 */
3396						struct sctp_paramhdr *ph;
3397						uint32_t *ippp;
3398
3399						SCTP_BUF_LEN(op_err) =
3400						    (sizeof(struct sctp_paramhdr) +
3401						    sizeof(uint32_t));
3402						ph = mtod(op_err,
3403						    struct sctp_paramhdr *);
3404						ph->param_type = htons(
3405						    SCTP_CAUSE_USER_INITIATED_ABT);
3406						ph->param_length = htons(SCTP_BUF_LEN(op_err));
3407						ippp = (uint32_t *) (ph + 1);
3408						*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5);
3409					}
3410					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3411#if defined(SCTP_PANIC_ON_ABORT)
3412					panic("inpcb_free does an abort");
3413#endif
3414
3415					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3416					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3417					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3418					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3419						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3420					}
3421					if (sctp_free_assoc(inp, asoc,
3422					    SCTP_PCBFREE_NOFORCE,
3423					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3424						cnt_in_sd++;
3425					}
3426					continue;
3427				} else {
3428					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3429				}
3430			}
3431			cnt_in_sd++;
3432			SCTP_TCB_UNLOCK(asoc);
3433		}
3434		/* now is there some left in our SHUTDOWN state? */
3435		if (cnt_in_sd) {
3436#ifdef SCTP_LOG_CLOSING
3437			sctp_log_closing(inp, NULL, 2);
3438#endif
3439			inp->sctp_socket = NULL;
3440			SCTP_INP_WUNLOCK(inp);
3441			SCTP_ASOC_CREATE_UNLOCK(inp);
3442			SCTP_INP_INFO_WUNLOCK();
3443			return;
3444		}
3445	}
3446	inp->sctp_socket = NULL;
3447	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3448	    SCTP_PCB_FLAGS_UNBOUND) {
3449		/*
3450		 * ok, this guy has been bound. It's port is somewhere in
3451		 * the SCTP_BASE_INFO(hash table). Remove it!
3452		 */
3453		LIST_REMOVE(inp, sctp_hash);
3454		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3455	}
3456	/*
3457	 * If there is a timer running to kill us, forget it, since it may
3458	 * have a contest on the INP lock.. which would cause us to die ...
3459	 */
3460	cnt = 0;
3461	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3462		SCTP_TCB_LOCK(asoc);
3463		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3464			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3465				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3466				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3467			}
3468			cnt++;
3469			SCTP_TCB_UNLOCK(asoc);
3470			continue;
3471		}
3472		/* Free associations that are NOT killing us */
3473		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3474		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3475			struct mbuf *op_err;
3476			uint32_t *ippp;
3477
3478			op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3479			    0, M_DONTWAIT, 1, MT_DATA);
3480			if (op_err) {
3481				/* Fill in the user initiated abort */
3482				struct sctp_paramhdr *ph;
3483
3484				SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) +
3485				    sizeof(uint32_t));
3486				ph = mtod(op_err, struct sctp_paramhdr *);
3487				ph->param_type = htons(
3488				    SCTP_CAUSE_USER_INITIATED_ABT);
3489				ph->param_length = htons(SCTP_BUF_LEN(op_err));
3490				ippp = (uint32_t *) (ph + 1);
3491				*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7);
3492
3493			}
3494			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3495#if defined(SCTP_PANIC_ON_ABORT)
3496			panic("inpcb_free does an abort");
3497#endif
3498			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3499			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3500		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3501			cnt++;
3502			SCTP_TCB_UNLOCK(asoc);
3503			continue;
3504		}
3505		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3506		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3507			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3508		}
3509		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3510			cnt++;
3511		}
3512	}
3513	if (cnt) {
3514		/* Ok we have someone out there that will kill us */
3515		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3516#ifdef SCTP_LOG_CLOSING
3517		sctp_log_closing(inp, NULL, 3);
3518#endif
3519		SCTP_INP_WUNLOCK(inp);
3520		SCTP_ASOC_CREATE_UNLOCK(inp);
3521		SCTP_INP_INFO_WUNLOCK();
3522		return;
3523	}
3524	if (SCTP_INP_LOCK_CONTENDED(inp))
3525		being_refed++;
3526	if (SCTP_INP_READ_CONTENDED(inp))
3527		being_refed++;
3528	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3529		being_refed++;
3530
3531	if ((inp->refcount) ||
3532	    (being_refed) ||
3533	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3534		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3535#ifdef SCTP_LOG_CLOSING
3536		sctp_log_closing(inp, NULL, 4);
3537#endif
3538		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3539		SCTP_INP_WUNLOCK(inp);
3540		SCTP_ASOC_CREATE_UNLOCK(inp);
3541		SCTP_INP_INFO_WUNLOCK();
3542		return;
3543	}
3544	inp->sctp_ep.signature_change.type = 0;
3545	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3546	/*
3547	 * Remove it from the list .. last thing we need a lock for.
3548	 */
3549	LIST_REMOVE(inp, sctp_list);
3550	SCTP_INP_WUNLOCK(inp);
3551	SCTP_ASOC_CREATE_UNLOCK(inp);
3552	SCTP_INP_INFO_WUNLOCK();
3553	/*
3554	 * Now we release all locks. Since this INP cannot be found anymore
3555	 * except possibly by the kill timer that might be running. We call
3556	 * the drain function here. It should hit the case were it sees the
3557	 * ACTIVE flag cleared and exit out freeing us to proceed and
3558	 * destroy everything.
3559	 */
3560	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3561		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3562	} else {
3563		/* Probably un-needed */
3564		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3565	}
3566
3567#ifdef SCTP_LOG_CLOSING
3568	sctp_log_closing(inp, NULL, 5);
3569#endif
3570
3571
3572	if ((inp->sctp_asocidhash) != NULL) {
3573		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3574		inp->sctp_asocidhash = NULL;
3575	}
3576	/* sa_ignore FREED_MEMORY */
3577	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3578		/* Its only abandoned if it had data left */
3579		if (sq->length)
3580			SCTP_STAT_INCR(sctps_left_abandon);
3581
3582		TAILQ_REMOVE(&inp->read_queue, sq, next);
3583		sctp_free_remote_addr(sq->whoFrom);
3584		if (so)
3585			so->so_rcv.sb_cc -= sq->length;
3586		if (sq->data) {
3587			sctp_m_freem(sq->data);
3588			sq->data = NULL;
3589		}
3590		/*
3591		 * no need to free the net count, since at this point all
3592		 * assoc's are gone.
3593		 */
3594		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3595		SCTP_DECR_READQ_COUNT();
3596	}
3597	/* Now the sctp_pcb things */
3598	/*
3599	 * free each asoc if it is not already closed/free. we can't use the
3600	 * macro here since le_next will get freed as part of the
3601	 * sctp_free_assoc() call.
3602	 */
3603	cnt = 0;
3604	if (so) {
3605#ifdef IPSEC
3606		ipsec_delete_pcbpolicy(ip_pcb);
3607#endif				/* IPSEC */
3608
3609		/* Unlocks not needed since the socket is gone now */
3610	}
3611	if (ip_pcb->inp_options) {
3612		(void)sctp_m_free(ip_pcb->inp_options);
3613		ip_pcb->inp_options = 0;
3614	}
3615	if (ip_pcb->inp_moptions) {
3616		inp_freemoptions(ip_pcb->inp_moptions);
3617		ip_pcb->inp_moptions = 0;
3618	}
3619#ifdef INET6
3620	if (ip_pcb->inp_vflag & INP_IPV6) {
3621		struct in6pcb *in6p;
3622
3623		in6p = (struct in6pcb *)inp;
3624		ip6_freepcbopts(in6p->in6p_outputopts);
3625	}
3626#endif				/* INET6 */
3627	ip_pcb->inp_vflag = 0;
3628	/* free up authentication fields */
3629	if (inp->sctp_ep.local_auth_chunks != NULL)
3630		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3631	if (inp->sctp_ep.local_hmacs != NULL)
3632		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3633
3634	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3635		LIST_REMOVE(shared_key, next);
3636		sctp_free_sharedkey(shared_key);
3637		/* sa_ignore FREED_MEMORY */
3638	}
3639
3640	/*
3641	 * if we have an address list the following will free the list of
3642	 * ifaddr's that are set into this ep. Again macro limitations here,
3643	 * since the LIST_FOREACH could be a bad idea.
3644	 */
3645	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3646		sctp_remove_laddr(laddr);
3647	}
3648
3649#ifdef SCTP_TRACK_FREED_ASOCS
3650	/* TEMP CODE */
3651	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3652		LIST_REMOVE(asoc, sctp_tcblist);
3653		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3654		SCTP_DECR_ASOC_COUNT();
3655	}
3656	/* *** END TEMP CODE *** */
3657#endif
3658	/* Now lets see about freeing the EP hash table. */
3659	if (inp->sctp_tcbhash != NULL) {
3660		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3661		inp->sctp_tcbhash = NULL;
3662	}
3663	/* Now we must put the ep memory back into the zone pool */
3664	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3665	SCTP_INP_LOCK_DESTROY(inp);
3666	SCTP_INP_READ_DESTROY(inp);
3667	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3668	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3669	SCTP_DECR_EP_COUNT();
3670}
3671
3672
3673struct sctp_nets *
3674sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3675{
3676	struct sctp_nets *net;
3677
3678	/* locate the address */
3679	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3680		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3681			return (net);
3682	}
3683	return (NULL);
3684}
3685
3686
3687int
3688sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3689{
3690	struct sctp_ifa *sctp_ifa;
3691
3692	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3693	if (sctp_ifa) {
3694		return (1);
3695	} else {
3696		return (0);
3697	}
3698}
3699
3700/*
3701 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3702 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3703 * stats of stuff.
3704 */
3705int
3706sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3707    int set_scope, int from)
3708{
3709	/*
3710	 * The following is redundant to the same lines in the
3711	 * sctp_aloc_assoc() but is needed since others call the add address
3712	 * function
3713	 */
3714	struct sctp_nets *net, *netfirst;
3715	int addr_inscope;
3716
3717	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3718	    from);
3719	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3720
3721	netfirst = sctp_findnet(stcb, newaddr);
3722	if (netfirst) {
3723		/*
3724		 * Lie and return ok, we don't want to make the association
3725		 * go away for this behavior. It will happen in the TCP
3726		 * model in a connected socket. It does not reach the hash
3727		 * table until after the association is built so it can't be
3728		 * found. Mark as reachable, since the initial creation will
3729		 * have been cleared and the NOT_IN_ASSOC flag will have
3730		 * been added... and we don't want to end up removing it
3731		 * back out.
3732		 */
3733		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3734			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3735			    SCTP_ADDR_UNCONFIRMED);
3736		} else {
3737			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3738		}
3739
3740		return (0);
3741	}
3742	addr_inscope = 1;
3743	if (newaddr->sa_family == AF_INET) {
3744		struct sockaddr_in *sin;
3745
3746		sin = (struct sockaddr_in *)newaddr;
3747		if (sin->sin_addr.s_addr == 0) {
3748			/* Invalid address */
3749			return (-1);
3750		}
3751		/* zero out the bzero area */
3752		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3753
3754		/* assure len is set */
3755		sin->sin_len = sizeof(struct sockaddr_in);
3756		if (set_scope) {
3757#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3758			stcb->ipv4_local_scope = 1;
3759#else
3760			if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3761				stcb->asoc.ipv4_local_scope = 1;
3762			}
3763#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3764		} else {
3765			/* Validate the address is in scope */
3766			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3767			    (stcb->asoc.ipv4_local_scope == 0)) {
3768				addr_inscope = 0;
3769			}
3770		}
3771#ifdef INET6
3772	} else if (newaddr->sa_family == AF_INET6) {
3773		struct sockaddr_in6 *sin6;
3774
3775		sin6 = (struct sockaddr_in6 *)newaddr;
3776		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3777			/* Invalid address */
3778			return (-1);
3779		}
3780		/* assure len is set */
3781		sin6->sin6_len = sizeof(struct sockaddr_in6);
3782		if (set_scope) {
3783			if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3784				stcb->asoc.loopback_scope = 1;
3785				stcb->asoc.local_scope = 0;
3786				stcb->asoc.ipv4_local_scope = 1;
3787				stcb->asoc.site_scope = 1;
3788			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3789				/*
3790				 * If the new destination is a LINK_LOCAL we
3791				 * must have common site scope. Don't set
3792				 * the local scope since we may not share
3793				 * all links, only loopback can do this.
3794				 * Links on the local network would also be
3795				 * on our private network for v4 too.
3796				 */
3797				stcb->asoc.ipv4_local_scope = 1;
3798				stcb->asoc.site_scope = 1;
3799			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3800				/*
3801				 * If the new destination is SITE_LOCAL then
3802				 * we must have site scope in common.
3803				 */
3804				stcb->asoc.site_scope = 1;
3805			}
3806		} else {
3807			/* Validate the address is in scope */
3808			if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3809			    (stcb->asoc.loopback_scope == 0)) {
3810				addr_inscope = 0;
3811			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3812			    (stcb->asoc.local_scope == 0)) {
3813				addr_inscope = 0;
3814			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3815			    (stcb->asoc.site_scope == 0)) {
3816				addr_inscope = 0;
3817			}
3818		}
3819#endif
3820	} else {
3821		/* not supported family type */
3822		return (-1);
3823	}
3824	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3825	if (net == NULL) {
3826		return (-1);
3827	}
3828	SCTP_INCR_RADDR_COUNT();
3829	bzero(net, sizeof(*net));
3830	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3831	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3832	if (newaddr->sa_family == AF_INET) {
3833		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3834	} else if (newaddr->sa_family == AF_INET6) {
3835		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3836	}
3837	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3838	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3839		stcb->asoc.loopback_scope = 1;
3840		stcb->asoc.ipv4_local_scope = 1;
3841		stcb->asoc.local_scope = 0;
3842		stcb->asoc.site_scope = 1;
3843		addr_inscope = 1;
3844	}
3845	net->failure_threshold = stcb->asoc.def_net_failure;
3846	if (addr_inscope == 0) {
3847		net->dest_state = (SCTP_ADDR_REACHABLE |
3848		    SCTP_ADDR_OUT_OF_SCOPE);
3849	} else {
3850		if (from == SCTP_ADDR_IS_CONFIRMED)
3851			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3852			net->dest_state = SCTP_ADDR_REACHABLE;
3853		else
3854			net->dest_state = SCTP_ADDR_REACHABLE |
3855			    SCTP_ADDR_UNCONFIRMED;
3856	}
3857	/*
3858	 * We set this to 0, the timer code knows that this means its an
3859	 * initial value
3860	 */
3861	net->RTO = 0;
3862	net->RTO_measured = 0;
3863	stcb->asoc.numnets++;
3864	*(&net->ref_count) = 1;
3865	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3866	net->tos_flowlabel = 0;
3867	if (SCTP_BASE_SYSCTL(sctp_udp_tunneling_for_client_enable)) {
3868		net->port = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port));
3869	} else {
3870		net->port = 0;
3871	}
3872#ifdef INET
3873	if (newaddr->sa_family == AF_INET)
3874		net->tos_flowlabel = stcb->asoc.default_tos;
3875#endif
3876#ifdef INET6
3877	if (newaddr->sa_family == AF_INET6)
3878		net->tos_flowlabel = stcb->asoc.default_flowlabel;
3879#endif
3880	/* Init the timer structure */
3881	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3882	SCTP_OS_TIMER_INIT(&net->fr_timer.timer);
3883	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3884
3885	/* Now generate a route for this guy */
3886#ifdef INET6
3887	/* KAME hack: embed scopeid */
3888	if (newaddr->sa_family == AF_INET6) {
3889		struct sockaddr_in6 *sin6;
3890
3891		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3892		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3893		sin6->sin6_scope_id = 0;
3894	}
3895#endif
3896	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3897
3898	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3899		/* Get source address */
3900		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3901		    stcb,
3902		    (sctp_route_t *) & net->ro,
3903		    net,
3904		    0,
3905		    stcb->asoc.vrf_id);
3906		/* Now get the interface MTU */
3907		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3908			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3909		} else {
3910			net->mtu = 0;
3911		}
3912		if (net->mtu == 0) {
3913			/* Huh ?? */
3914			net->mtu = SCTP_DEFAULT_MTU;
3915		} else {
3916			uint32_t rmtu;
3917
3918			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3919			if (rmtu == 0) {
3920				/*
3921				 * Start things off to match mtu of
3922				 * interface please.
3923				 */
3924				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3925				    net->ro.ro_rt, net->mtu);
3926			} else {
3927				/*
3928				 * we take the route mtu over the interface,
3929				 * since the route may be leading out the
3930				 * loopback, or a different interface.
3931				 */
3932				net->mtu = rmtu;
3933			}
3934		}
3935		if (from == SCTP_ALLOC_ASOC) {
3936			stcb->asoc.smallest_mtu = net->mtu;
3937		}
3938	} else {
3939		net->mtu = stcb->asoc.smallest_mtu;
3940	}
3941#ifdef INET6
3942	if (newaddr->sa_family == AF_INET6) {
3943		struct sockaddr_in6 *sin6;
3944
3945		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3946		(void)sa6_recoverscope(sin6);
3947	}
3948#endif
3949	if (net->port) {
3950		net->mtu -= sizeof(struct udphdr);
3951	}
3952	if (stcb->asoc.smallest_mtu > net->mtu) {
3953		stcb->asoc.smallest_mtu = net->mtu;
3954	}
3955	/* JRS - Use the congestion control given in the CC module */
3956	stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net);
3957
3958	/*
3959	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3960	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3961	 */
3962	net->find_pseudo_cumack = 1;
3963	net->find_rtx_pseudo_cumack = 1;
3964	net->src_addr_selected = 0;
3965	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3966	if (net->ro.ro_rt == NULL) {
3967		/* Since we have no route put it at the back */
3968		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3969	} else if (netfirst == NULL) {
3970		/* We are the first one in the pool. */
3971		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3972	} else if (netfirst->ro.ro_rt == NULL) {
3973		/*
3974		 * First one has NO route. Place this one ahead of the first
3975		 * one.
3976		 */
3977		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3978	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
3979		/*
3980		 * This one has a different interface than the one at the
3981		 * top of the list. Place it ahead.
3982		 */
3983		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3984	} else {
3985		/*
3986		 * Ok we have the same interface as the first one. Move
3987		 * forward until we find either a) one with a NULL route...
3988		 * insert ahead of that b) one with a different ifp.. insert
3989		 * after that. c) end of the list.. insert at the tail.
3990		 */
3991		struct sctp_nets *netlook;
3992
3993		do {
3994			netlook = TAILQ_NEXT(netfirst, sctp_next);
3995			if (netlook == NULL) {
3996				/* End of the list */
3997				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3998				break;
3999			} else if (netlook->ro.ro_rt == NULL) {
4000				/* next one has NO route */
4001				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4002				break;
4003			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4004				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4005				    net, sctp_next);
4006				break;
4007			}
4008			/* Shift forward */
4009			netfirst = netlook;
4010		} while (netlook != NULL);
4011	}
4012
4013	/* got to have a primary set */
4014	if (stcb->asoc.primary_destination == 0) {
4015		stcb->asoc.primary_destination = net;
4016	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4017		    (net->ro.ro_rt) &&
4018	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4019		/* No route to current primary adopt new primary */
4020		stcb->asoc.primary_destination = net;
4021	}
4022	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb,
4023	    net);
4024	/* Validate primary is first */
4025	net = TAILQ_FIRST(&stcb->asoc.nets);
4026	if ((net != stcb->asoc.primary_destination) &&
4027	    (stcb->asoc.primary_destination)) {
4028		/*
4029		 * first one on the list is NOT the primary sctp_cmpaddr()
4030		 * is much more efficient if the primary is the first on the
4031		 * list, make it so.
4032		 */
4033		TAILQ_REMOVE(&stcb->asoc.nets,
4034		    stcb->asoc.primary_destination, sctp_next);
4035		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4036		    stcb->asoc.primary_destination, sctp_next);
4037	}
4038	return (0);
4039}
4040
4041
4042static uint32_t
4043sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4044{
4045	uint32_t id;
4046	struct sctpasochead *head;
4047	struct sctp_tcb *lstcb;
4048
4049	SCTP_INP_WLOCK(inp);
4050try_again:
4051	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4052		/* TSNH */
4053		SCTP_INP_WUNLOCK(inp);
4054		return (0);
4055	}
4056	/*
4057	 * We don't allow assoc id to be 0, this is needed otherwise if the
4058	 * id were to wrap we would have issues with some socket options.
4059	 */
4060	if (inp->sctp_associd_counter == 0) {
4061		inp->sctp_associd_counter++;
4062	}
4063	id = inp->sctp_associd_counter;
4064	inp->sctp_associd_counter++;
4065	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4066	if (lstcb) {
4067		goto try_again;
4068	}
4069	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4070	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4071	stcb->asoc.in_asocid_hash = 1;
4072	SCTP_INP_WUNLOCK(inp);
4073	return id;
4074}
4075
4076/*
4077 * allocate an association and add it to the endpoint. The caller must be
4078 * careful to add all additional addresses once they are know right away or
4079 * else the assoc will be may experience a blackout scenario.
4080 */
4081struct sctp_tcb *
4082sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4083    int *error, uint32_t override_tag, uint32_t vrf_id,
4084    struct thread *p
4085)
4086{
4087	/* note the p argument is only valid in unbound sockets */
4088
4089	struct sctp_tcb *stcb;
4090	struct sctp_association *asoc;
4091	struct sctpasochead *head;
4092	uint16_t rport;
4093	int err;
4094
4095	/*
4096	 * Assumption made here: Caller has done a
4097	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4098	 * address does not exist already.
4099	 */
4100	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4101		/* Hit max assoc, sorry no more */
4102		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4103		*error = ENOBUFS;
4104		return (NULL);
4105	}
4106	if (firstaddr == NULL) {
4107		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4108		*error = EINVAL;
4109		return (NULL);
4110	}
4111	SCTP_INP_RLOCK(inp);
4112	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4113	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4114	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4115		/*
4116		 * If its in the TCP pool, its NOT allowed to create an
4117		 * association. The parent listener needs to call
4118		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4119		 * off, or connected one does this.. its an error.
4120		 */
4121		SCTP_INP_RUNLOCK(inp);
4122		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4123		*error = EINVAL;
4124		return (NULL);
4125	}
4126	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4127#ifdef SCTP_DEBUG
4128	if (firstaddr) {
4129		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4130		SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4131		    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4132	} else {
4133		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4134	}
4135#endif				/* SCTP_DEBUG */
4136	if (firstaddr->sa_family == AF_INET) {
4137		struct sockaddr_in *sin;
4138
4139		sin = (struct sockaddr_in *)firstaddr;
4140		if ((sin->sin_port == 0) || (sin->sin_addr.s_addr == 0)) {
4141			/* Invalid address */
4142			SCTP_INP_RUNLOCK(inp);
4143			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4144			*error = EINVAL;
4145			return (NULL);
4146		}
4147		rport = sin->sin_port;
4148	} else if (firstaddr->sa_family == AF_INET6) {
4149		struct sockaddr_in6 *sin6;
4150
4151		sin6 = (struct sockaddr_in6 *)firstaddr;
4152		if ((sin6->sin6_port == 0) ||
4153		    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
4154			/* Invalid address */
4155			SCTP_INP_RUNLOCK(inp);
4156			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4157			*error = EINVAL;
4158			return (NULL);
4159		}
4160		rport = sin6->sin6_port;
4161	} else {
4162		/* not supported family type */
4163		SCTP_INP_RUNLOCK(inp);
4164		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4165		*error = EINVAL;
4166		return (NULL);
4167	}
4168	SCTP_INP_RUNLOCK(inp);
4169	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4170		/*
4171		 * If you have not performed a bind, then we need to do the
4172		 * ephemeral bind for you.
4173		 */
4174		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4175		    (struct sockaddr *)NULL,
4176		    (struct sctp_ifa *)NULL,
4177		    p
4178		    ))) {
4179			/* bind error, probably perm */
4180			*error = err;
4181			return (NULL);
4182		}
4183	}
4184	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4185	if (stcb == NULL) {
4186		/* out of memory? */
4187		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4188		*error = ENOMEM;
4189		return (NULL);
4190	}
4191	SCTP_INCR_ASOC_COUNT();
4192
4193	bzero(stcb, sizeof(*stcb));
4194	asoc = &stcb->asoc;
4195
4196	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4197	SCTP_TCB_LOCK_INIT(stcb);
4198	SCTP_TCB_SEND_LOCK_INIT(stcb);
4199	stcb->rport = rport;
4200	/* setup back pointer's */
4201	stcb->sctp_ep = inp;
4202	stcb->sctp_socket = inp->sctp_socket;
4203	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4204		/* failed */
4205		SCTP_TCB_LOCK_DESTROY(stcb);
4206		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4207		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4208		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4209		SCTP_DECR_ASOC_COUNT();
4210		*error = err;
4211		return (NULL);
4212	}
4213	/* and the port */
4214	SCTP_INP_INFO_WLOCK();
4215	SCTP_INP_WLOCK(inp);
4216	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4217		/* inpcb freed while alloc going on */
4218		SCTP_TCB_LOCK_DESTROY(stcb);
4219		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4220		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4221		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4222		SCTP_INP_WUNLOCK(inp);
4223		SCTP_INP_INFO_WUNLOCK();
4224		SCTP_DECR_ASOC_COUNT();
4225		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4226		*error = EINVAL;
4227		return (NULL);
4228	}
4229	SCTP_TCB_LOCK(stcb);
4230
4231	/* now that my_vtag is set, add it to the hash */
4232	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4233	/* put it in the bucket in the vtag hash of assoc's for the system */
4234	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4235	SCTP_INP_INFO_WUNLOCK();
4236
4237	if ((err = sctp_add_remote_addr(stcb, firstaddr, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4238		/* failure.. memory error? */
4239		if (asoc->strmout) {
4240			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4241			asoc->strmout = NULL;
4242		}
4243		if (asoc->mapping_array) {
4244			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4245			asoc->mapping_array = NULL;
4246		}
4247		if (asoc->nr_mapping_array) {
4248			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4249			asoc->nr_mapping_array = NULL;
4250		}
4251		SCTP_DECR_ASOC_COUNT();
4252		SCTP_TCB_LOCK_DESTROY(stcb);
4253		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4254		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4255		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4256		SCTP_INP_WUNLOCK(inp);
4257		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4258		*error = ENOBUFS;
4259		return (NULL);
4260	}
4261	/* Init all the timers */
4262	SCTP_OS_TIMER_INIT(&asoc->hb_timer.timer);
4263	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4264	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4265	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4266	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4267	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4268	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4269	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4270
4271	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4272	/* now file the port under the hash as well */
4273	if (inp->sctp_tcbhash != NULL) {
4274		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4275		    inp->sctp_hashmark)];
4276		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4277	}
4278	SCTP_INP_WUNLOCK(inp);
4279	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", stcb);
4280	return (stcb);
4281}
4282
4283
4284void
4285sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4286{
4287	struct sctp_association *asoc;
4288
4289	asoc = &stcb->asoc;
4290	asoc->numnets--;
4291	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4292	if (net == asoc->primary_destination) {
4293		/* Reset primary */
4294		struct sctp_nets *lnet;
4295
4296		lnet = TAILQ_FIRST(&asoc->nets);
4297		/*
4298		 * Mobility adaptation Ideally, if deleted destination is
4299		 * the primary, it becomes a fast retransmission trigger by
4300		 * the subsequent SET PRIMARY. (by micchie)
4301		 */
4302		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4303		    SCTP_MOBILITY_BASE) ||
4304		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4305		    SCTP_MOBILITY_FASTHANDOFF)) {
4306			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4307			if (asoc->deleted_primary != NULL) {
4308				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4309				goto out;
4310			}
4311			asoc->deleted_primary = net;
4312			atomic_add_int(&net->ref_count, 1);
4313			memset(&net->lastsa, 0, sizeof(net->lastsa));
4314			memset(&net->lastsv, 0, sizeof(net->lastsv));
4315			sctp_mobility_feature_on(stcb->sctp_ep,
4316			    SCTP_MOBILITY_PRIM_DELETED);
4317			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4318			    stcb->sctp_ep, stcb, NULL);
4319		}
4320out:
4321		/* Try to find a confirmed primary */
4322		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4323	}
4324	if (net == asoc->last_data_chunk_from) {
4325		/* Reset primary */
4326		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4327	}
4328	if (net == asoc->last_control_chunk_from) {
4329		/* Clear net */
4330		asoc->last_control_chunk_from = NULL;
4331	}
4332	sctp_free_remote_addr(net);
4333}
4334
4335/*
4336 * remove a remote endpoint address from an association, it will fail if the
4337 * address does not exist.
4338 */
4339int
4340sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4341{
4342	/*
4343	 * Here we need to remove a remote address. This is quite simple, we
4344	 * first find it in the list of address for the association
4345	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4346	 * on that item. Note we do not allow it to be removed if there are
4347	 * no other addresses.
4348	 */
4349	struct sctp_association *asoc;
4350	struct sctp_nets *net, *nnet;
4351
4352	asoc = &stcb->asoc;
4353
4354	/* locate the address */
4355	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4356		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4357			continue;
4358		}
4359		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4360		    remaddr)) {
4361			/* we found the guy */
4362			if (asoc->numnets < 2) {
4363				/* Must have at LEAST two remote addresses */
4364				return (-1);
4365			} else {
4366				sctp_remove_net(stcb, net);
4367				return (0);
4368			}
4369		}
4370	}
4371	/* not found. */
4372	return (-2);
4373}
4374
4375void
4376sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4377{
4378	struct sctpvtaghead *chain;
4379	struct sctp_tagblock *twait_block;
4380	int found = 0;
4381	int i;
4382
4383	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4384	if (!LIST_EMPTY(chain)) {
4385		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4386			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4387				if ((twait_block->vtag_block[i].v_tag == tag) &&
4388				    (twait_block->vtag_block[i].lport == lport) &&
4389				    (twait_block->vtag_block[i].rport == rport)) {
4390					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4391					twait_block->vtag_block[i].v_tag = 0;
4392					twait_block->vtag_block[i].lport = 0;
4393					twait_block->vtag_block[i].rport = 0;
4394					found = 1;
4395					break;
4396				}
4397			}
4398			if (found)
4399				break;
4400		}
4401	}
4402}
4403
4404int
4405sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4406{
4407	struct sctpvtaghead *chain;
4408	struct sctp_tagblock *twait_block;
4409	int found = 0;
4410	int i;
4411
4412	SCTP_INP_INFO_WLOCK();
4413	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4414	if (!LIST_EMPTY(chain)) {
4415		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4416			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4417				if ((twait_block->vtag_block[i].v_tag == tag) &&
4418				    (twait_block->vtag_block[i].lport == lport) &&
4419				    (twait_block->vtag_block[i].rport == rport)) {
4420					found = 1;
4421					break;
4422				}
4423			}
4424			if (found)
4425				break;
4426		}
4427	}
4428	SCTP_INP_INFO_WUNLOCK();
4429	return (found);
4430}
4431
4432
4433void
4434sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4435{
4436	struct sctpvtaghead *chain;
4437	struct sctp_tagblock *twait_block;
4438	struct timeval now;
4439	int set, i;
4440
4441	if (time == 0) {
4442		/* Its disabled */
4443		return;
4444	}
4445	(void)SCTP_GETTIME_TIMEVAL(&now);
4446	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4447	set = 0;
4448	if (!LIST_EMPTY(chain)) {
4449		/* Block(s) present, lets find space, and expire on the fly */
4450		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4451			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4452				if ((twait_block->vtag_block[i].v_tag == 0) &&
4453				    !set) {
4454					twait_block->vtag_block[i].tv_sec_at_expire =
4455					    now.tv_sec + time;
4456					twait_block->vtag_block[i].v_tag = tag;
4457					twait_block->vtag_block[i].lport = lport;
4458					twait_block->vtag_block[i].rport = rport;
4459					set = 1;
4460				} else if ((twait_block->vtag_block[i].v_tag) &&
4461				    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4462					/* Audit expires this guy */
4463					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4464					twait_block->vtag_block[i].v_tag = 0;
4465					twait_block->vtag_block[i].lport = 0;
4466					twait_block->vtag_block[i].rport = 0;
4467					if (set == 0) {
4468						/* Reuse it for my new tag */
4469						twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4470						twait_block->vtag_block[i].v_tag = tag;
4471						twait_block->vtag_block[i].lport = lport;
4472						twait_block->vtag_block[i].rport = rport;
4473						set = 1;
4474					}
4475				}
4476			}
4477			if (set) {
4478				/*
4479				 * We only do up to the block where we can
4480				 * place our tag for audits
4481				 */
4482				break;
4483			}
4484		}
4485	}
4486	/* Need to add a new block to chain */
4487	if (!set) {
4488		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4489		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4490		if (twait_block == NULL) {
4491#ifdef INVARIANTS
4492			panic("Can not alloc tagblock");
4493#endif
4494			return;
4495		}
4496		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4497		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4498		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4499		twait_block->vtag_block[0].v_tag = tag;
4500		twait_block->vtag_block[0].lport = lport;
4501		twait_block->vtag_block[0].rport = rport;
4502	}
4503}
4504
4505
4506
4507/*-
4508 * Free the association after un-hashing the remote port. This
4509 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4510 * expect that the input to this function IS a locked TCB.
4511 * It will return 0, if it did NOT destroy the association (instead
4512 * it unlocks it. It will return NON-zero if it either destroyed the
4513 * association OR the association is already destroyed.
4514 */
4515int
4516sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4517{
4518	int i;
4519	struct sctp_association *asoc;
4520	struct sctp_nets *net, *nnet;
4521	struct sctp_laddr *laddr, *naddr;
4522	struct sctp_tmit_chunk *chk, *nchk;
4523	struct sctp_asconf_addr *aparam, *naparam;
4524	struct sctp_asconf_ack *aack, *naack;
4525	struct sctp_stream_reset_list *strrst, *nstrrst;
4526	struct sctp_queued_to_read *sq, *nsq;
4527	struct sctp_stream_queue_pending *sp, *nsp;
4528	sctp_sharedkey_t *shared_key, *nshared_key;
4529	struct socket *so;
4530
4531	/* first, lets purge the entry from the hash table. */
4532
4533#ifdef SCTP_LOG_CLOSING
4534	sctp_log_closing(inp, stcb, 6);
4535#endif
4536	if (stcb->asoc.state == 0) {
4537#ifdef SCTP_LOG_CLOSING
4538		sctp_log_closing(inp, NULL, 7);
4539#endif
4540		/* there is no asoc, really TSNH :-0 */
4541		return (1);
4542	}
4543	/* TEMP CODE */
4544	if (stcb->freed_from_where == 0) {
4545		/* Only record the first place free happened from */
4546		stcb->freed_from_where = from_location;
4547	}
4548	/* TEMP CODE */
4549
4550	asoc = &stcb->asoc;
4551	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4552	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4553		/* nothing around */
4554		so = NULL;
4555	else
4556		so = inp->sctp_socket;
4557
4558	/*
4559	 * We used timer based freeing if a reader or writer is in the way.
4560	 * So we first check if we are actually being called from a timer,
4561	 * if so we abort early if a reader or writer is still in the way.
4562	 */
4563	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4564	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4565		/*
4566		 * is it the timer driving us? if so are the reader/writers
4567		 * gone?
4568		 */
4569		if (stcb->asoc.refcnt) {
4570			/* nope, reader or writer in the way */
4571			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4572			/* no asoc destroyed */
4573			SCTP_TCB_UNLOCK(stcb);
4574#ifdef SCTP_LOG_CLOSING
4575			sctp_log_closing(inp, stcb, 8);
4576#endif
4577			return (0);
4578		}
4579	}
4580	/* now clean up any other timers */
4581	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
4582	asoc->hb_timer.self = NULL;
4583	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4584	asoc->dack_timer.self = NULL;
4585	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4586	/*-
4587	 * For stream reset we don't blast this unless
4588	 * it is a str-reset timer, it might be the
4589	 * free-asoc timer which we DON'T want to
4590	 * disturb.
4591	 */
4592	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4593		asoc->strreset_timer.self = NULL;
4594	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4595	asoc->asconf_timer.self = NULL;
4596	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4597	asoc->autoclose_timer.self = NULL;
4598	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4599	asoc->shut_guard_timer.self = NULL;
4600	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4601	asoc->delayed_event_timer.self = NULL;
4602	/* Mobility adaptation */
4603	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4604	asoc->delete_prim_timer.self = NULL;
4605	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4606		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
4607		net->fr_timer.self = NULL;
4608		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4609		net->rxt_timer.self = NULL;
4610		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4611		net->pmtu_timer.self = NULL;
4612	}
4613	/* Now the read queue needs to be cleaned up (only once) */
4614	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4615		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4616		SCTP_INP_READ_LOCK(inp);
4617		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4618			if (sq->stcb == stcb) {
4619				sq->do_not_ref_stcb = 1;
4620				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4621				/*
4622				 * If there is no end, there never will be
4623				 * now.
4624				 */
4625				if (sq->end_added == 0) {
4626					/* Held for PD-API clear that. */
4627					sq->pdapi_aborted = 1;
4628					sq->held_length = 0;
4629					if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4630						/*
4631						 * Need to add a PD-API
4632						 * aborted indication.
4633						 * Setting the control_pdapi
4634						 * assures that it will be
4635						 * added right after this
4636						 * msg.
4637						 */
4638						uint32_t strseq;
4639
4640						stcb->asoc.control_pdapi = sq;
4641						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4642						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4643						    stcb,
4644						    SCTP_PARTIAL_DELIVERY_ABORTED,
4645						    (void *)&strseq,
4646						    SCTP_SO_LOCKED);
4647						stcb->asoc.control_pdapi = NULL;
4648					}
4649				}
4650				/* Add an end to wake them */
4651				sq->end_added = 1;
4652			}
4653		}
4654		SCTP_INP_READ_UNLOCK(inp);
4655		if (stcb->block_entry) {
4656			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4657			stcb->block_entry->error = ECONNRESET;
4658			stcb->block_entry = NULL;
4659		}
4660	}
4661	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4662		/*
4663		 * Someone holds a reference OR the socket is unaccepted
4664		 * yet.
4665		 */
4666		if ((stcb->asoc.refcnt) ||
4667		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4668		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4669			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4670			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4671		}
4672		SCTP_TCB_UNLOCK(stcb);
4673		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4674		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4675			/* nothing around */
4676			so = NULL;
4677		if (so) {
4678			/* Wake any reader/writers */
4679			sctp_sorwakeup(inp, so);
4680			sctp_sowwakeup(inp, so);
4681		}
4682#ifdef SCTP_LOG_CLOSING
4683		sctp_log_closing(inp, stcb, 9);
4684#endif
4685		/* no asoc destroyed */
4686		return (0);
4687	}
4688#ifdef SCTP_LOG_CLOSING
4689	sctp_log_closing(inp, stcb, 10);
4690#endif
4691	/*
4692	 * When I reach here, no others want to kill the assoc yet.. and I
4693	 * own the lock. Now its possible an abort comes in when I do the
4694	 * lock exchange below to grab all the locks to do the final take
4695	 * out. to prevent this we increment the count, which will start a
4696	 * timer and blow out above thus assuring us that we hold exclusive
4697	 * killing of the asoc. Note that after getting back the TCB lock we
4698	 * will go ahead and increment the counter back up and stop any
4699	 * timer a passing stranger may have started :-S
4700	 */
4701	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4702		atomic_add_int(&stcb->asoc.refcnt, 1);
4703
4704		SCTP_TCB_UNLOCK(stcb);
4705		SCTP_INP_INFO_WLOCK();
4706		SCTP_INP_WLOCK(inp);
4707		SCTP_TCB_LOCK(stcb);
4708	}
4709	/* Double check the GONE flag */
4710	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4711	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4712		/* nothing around */
4713		so = NULL;
4714
4715	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4716	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4717		/*
4718		 * For TCP type we need special handling when we are
4719		 * connected. We also include the peel'ed off ones to.
4720		 */
4721		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4722			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4723			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4724			if (so) {
4725				SOCK_LOCK(so);
4726				if (so->so_rcv.sb_cc == 0) {
4727					so->so_state &= ~(SS_ISCONNECTING |
4728					    SS_ISDISCONNECTING |
4729					    SS_ISCONFIRMING |
4730					    SS_ISCONNECTED);
4731				}
4732				socantrcvmore_locked(so);
4733				sctp_sowwakeup(inp, so);
4734				sctp_sorwakeup(inp, so);
4735				SCTP_SOWAKEUP(so);
4736			}
4737		}
4738	}
4739	/*
4740	 * Make it invalid too, that way if its about to run it will abort
4741	 * and return.
4742	 */
4743	/* re-increment the lock */
4744	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4745		atomic_add_int(&stcb->asoc.refcnt, -1);
4746	}
4747	if (stcb->asoc.refcnt) {
4748		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4749		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4750		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4751			SCTP_INP_INFO_WUNLOCK();
4752			SCTP_INP_WUNLOCK(inp);
4753		}
4754		SCTP_TCB_UNLOCK(stcb);
4755		return (0);
4756	}
4757	asoc->state = 0;
4758	if (inp->sctp_tcbhash) {
4759		LIST_REMOVE(stcb, sctp_tcbhash);
4760	}
4761	if (stcb->asoc.in_asocid_hash) {
4762		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4763	}
4764	/* Now lets remove it from the list of ALL associations in the EP */
4765	LIST_REMOVE(stcb, sctp_tcblist);
4766	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4767		SCTP_INP_INCR_REF(inp);
4768		SCTP_INP_WUNLOCK(inp);
4769	}
4770	/* pull from vtag hash */
4771	LIST_REMOVE(stcb, sctp_asocs);
4772	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4773	    inp->sctp_lport, stcb->rport);
4774
4775	/*
4776	 * Now restop the timers to be sure this is paranoia at is finest!
4777	 */
4778	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4779	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
4780	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4781	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4782	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4783	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4784	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4785	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4786	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4787		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
4788		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4789		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4790	}
4791
4792	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4793	/*
4794	 * The chunk lists and such SHOULD be empty but we check them just
4795	 * in case.
4796	 */
4797	/* anything on the wheel needs to be removed */
4798	for (i = 0; i < asoc->streamoutcnt; i++) {
4799		struct sctp_stream_out *outs;
4800
4801		outs = &asoc->strmout[i];
4802		/* now clean up any chunks here */
4803		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4804			TAILQ_REMOVE(&outs->outqueue, sp, next);
4805			if (sp->data) {
4806				if (so) {
4807					/* Still an open socket - report */
4808					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4809					    SCTP_NOTIFY_DATAGRAM_UNSENT,
4810					    (void *)sp, SCTP_SO_LOCKED);
4811				}
4812				if (sp->data) {
4813					sctp_m_freem(sp->data);
4814					sp->data = NULL;
4815					sp->tail_mbuf = NULL;
4816				}
4817			}
4818			if (sp->net) {
4819				sctp_free_remote_addr(sp->net);
4820				sp->net = NULL;
4821			}
4822			sctp_free_spbufspace(stcb, asoc, sp);
4823			if (sp->holds_key_ref)
4824				sctp_auth_key_release(stcb, sp->auth_keyid);
4825			/* Free the zone stuff  */
4826			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_strmoq), sp);
4827			SCTP_DECR_STRMOQ_COUNT();
4828			/* sa_ignore FREED_MEMORY */
4829		}
4830	}
4831	/* sa_ignore FREED_MEMORY */
4832	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4833		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4834		SCTP_FREE(strrst, SCTP_M_STRESET);
4835	}
4836	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4837		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4838		if (sq->data) {
4839			sctp_m_freem(sq->data);
4840			sq->data = NULL;
4841		}
4842		sctp_free_remote_addr(sq->whoFrom);
4843		sq->whoFrom = NULL;
4844		sq->stcb = NULL;
4845		/* Free the ctl entry */
4846		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4847		SCTP_DECR_READQ_COUNT();
4848		/* sa_ignore FREED_MEMORY */
4849	}
4850	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4851		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4852		if (chk->data) {
4853			sctp_m_freem(chk->data);
4854			chk->data = NULL;
4855		}
4856		if (chk->holds_key_ref)
4857			sctp_auth_key_release(stcb, chk->auth_keyid);
4858		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4859		SCTP_DECR_CHK_COUNT();
4860		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4861		asoc->free_chunk_cnt--;
4862		/* sa_ignore FREED_MEMORY */
4863	}
4864	/* pending send queue SHOULD be empty */
4865	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4866		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4867		if (chk->data) {
4868			if (so) {
4869				/* Still a socket? */
4870				sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb,
4871				    SCTP_NOTIFY_DATAGRAM_UNSENT, chk, SCTP_SO_LOCKED);
4872			}
4873			if (chk->data) {
4874				sctp_m_freem(chk->data);
4875				chk->data = NULL;
4876			}
4877		}
4878		if (chk->holds_key_ref)
4879			sctp_auth_key_release(stcb, chk->auth_keyid);
4880		if (chk->whoTo) {
4881			sctp_free_remote_addr(chk->whoTo);
4882			chk->whoTo = NULL;
4883		}
4884		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4885		SCTP_DECR_CHK_COUNT();
4886		/* sa_ignore FREED_MEMORY */
4887	}
4888	/* sent queue SHOULD be empty */
4889	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
4890		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4891		if (chk->data) {
4892			if (so) {
4893				/* Still a socket? */
4894				sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb,
4895				    SCTP_NOTIFY_DATAGRAM_SENT, chk, SCTP_SO_LOCKED);
4896			}
4897			if (chk->data) {
4898				sctp_m_freem(chk->data);
4899				chk->data = NULL;
4900			}
4901		}
4902		if (chk->holds_key_ref)
4903			sctp_auth_key_release(stcb, chk->auth_keyid);
4904		sctp_free_remote_addr(chk->whoTo);
4905		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4906		SCTP_DECR_CHK_COUNT();
4907		/* sa_ignore FREED_MEMORY */
4908	}
4909	/* control queue MAY not be empty */
4910	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
4911		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4912		if (chk->data) {
4913			sctp_m_freem(chk->data);
4914			chk->data = NULL;
4915		}
4916		if (chk->holds_key_ref)
4917			sctp_auth_key_release(stcb, chk->auth_keyid);
4918		sctp_free_remote_addr(chk->whoTo);
4919		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4920		SCTP_DECR_CHK_COUNT();
4921		/* sa_ignore FREED_MEMORY */
4922	}
4923	/* ASCONF queue MAY not be empty */
4924	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
4925		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
4926		if (chk->data) {
4927			sctp_m_freem(chk->data);
4928			chk->data = NULL;
4929		}
4930		if (chk->holds_key_ref)
4931			sctp_auth_key_release(stcb, chk->auth_keyid);
4932		sctp_free_remote_addr(chk->whoTo);
4933		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4934		SCTP_DECR_CHK_COUNT();
4935		/* sa_ignore FREED_MEMORY */
4936	}
4937	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
4938		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
4939		if (chk->data) {
4940			sctp_m_freem(chk->data);
4941			chk->data = NULL;
4942		}
4943		if (chk->holds_key_ref)
4944			sctp_auth_key_release(stcb, chk->auth_keyid);
4945		sctp_free_remote_addr(chk->whoTo);
4946		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4947		SCTP_DECR_CHK_COUNT();
4948		/* sa_ignore FREED_MEMORY */
4949	}
4950
4951	if (asoc->mapping_array) {
4952		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4953		asoc->mapping_array = NULL;
4954	}
4955	if (asoc->nr_mapping_array) {
4956		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4957		asoc->nr_mapping_array = NULL;
4958	}
4959	/* the stream outs */
4960	if (asoc->strmout) {
4961		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4962		asoc->strmout = NULL;
4963	}
4964	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
4965	if (asoc->strmin) {
4966		struct sctp_queued_to_read *ctl, *nctl;
4967
4968		for (i = 0; i < asoc->streamincnt; i++) {
4969			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
4970				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
4971				sctp_free_remote_addr(ctl->whoFrom);
4972				if (ctl->data) {
4973					sctp_m_freem(ctl->data);
4974					ctl->data = NULL;
4975				}
4976				/*
4977				 * We don't free the address here since all
4978				 * the net's were freed above.
4979				 */
4980				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
4981				SCTP_DECR_READQ_COUNT();
4982			}
4983		}
4984		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
4985		asoc->strmin = NULL;
4986	}
4987	asoc->streamincnt = 0;
4988	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4989#ifdef INVARIANTS
4990		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
4991			panic("no net's left alloc'ed, or list points to itself");
4992		}
4993#endif
4994		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4995		sctp_free_remote_addr(net);
4996	}
4997	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
4998		/* sa_ignore FREED_MEMORY */
4999		sctp_remove_laddr(laddr);
5000	}
5001
5002	/* pending asconf (address) parameters */
5003	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5004		/* sa_ignore FREED_MEMORY */
5005		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5006		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5007	}
5008	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5009		/* sa_ignore FREED_MEMORY */
5010		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5011		if (aack->data != NULL) {
5012			sctp_m_freem(aack->data);
5013		}
5014		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5015	}
5016	/* clean up auth stuff */
5017	if (asoc->local_hmacs)
5018		sctp_free_hmaclist(asoc->local_hmacs);
5019	if (asoc->peer_hmacs)
5020		sctp_free_hmaclist(asoc->peer_hmacs);
5021
5022	if (asoc->local_auth_chunks)
5023		sctp_free_chunklist(asoc->local_auth_chunks);
5024	if (asoc->peer_auth_chunks)
5025		sctp_free_chunklist(asoc->peer_auth_chunks);
5026
5027	sctp_free_authinfo(&asoc->authinfo);
5028
5029	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5030		LIST_REMOVE(shared_key, next);
5031		sctp_free_sharedkey(shared_key);
5032		/* sa_ignore FREED_MEMORY */
5033	}
5034
5035	/* Insert new items here :> */
5036
5037	/* Get rid of LOCK */
5038	SCTP_TCB_LOCK_DESTROY(stcb);
5039	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5040	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5041		SCTP_INP_INFO_WUNLOCK();
5042		SCTP_INP_RLOCK(inp);
5043	}
5044#ifdef SCTP_TRACK_FREED_ASOCS
5045	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5046		/* now clean up the tasoc itself */
5047		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5048		SCTP_DECR_ASOC_COUNT();
5049	} else {
5050		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5051	}
5052#else
5053	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5054	SCTP_DECR_ASOC_COUNT();
5055#endif
5056	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5057		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5058			/*
5059			 * If its NOT the inp_free calling us AND sctp_close
5060			 * as been called, we call back...
5061			 */
5062			SCTP_INP_RUNLOCK(inp);
5063			/*
5064			 * This will start the kill timer (if we are the
5065			 * last one) since we hold an increment yet. But
5066			 * this is the only safe way to do this since
5067			 * otherwise if the socket closes at the same time
5068			 * we are here we might collide in the cleanup.
5069			 */
5070			sctp_inpcb_free(inp,
5071			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5072			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5073			SCTP_INP_DECR_REF(inp);
5074			goto out_of;
5075		} else {
5076			/* The socket is still open. */
5077			SCTP_INP_DECR_REF(inp);
5078		}
5079	}
5080	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5081		SCTP_INP_RUNLOCK(inp);
5082	}
5083out_of:
5084	/* destroyed the asoc */
5085#ifdef SCTP_LOG_CLOSING
5086	sctp_log_closing(inp, NULL, 11);
5087#endif
5088	return (1);
5089}
5090
5091
5092
5093/*
5094 * determine if a destination is "reachable" based upon the addresses bound
5095 * to the current endpoint (e.g. only v4 or v6 currently bound)
5096 */
5097/*
5098 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5099 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5100 * types bound as its endpoint
5101 */
5102int
5103sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5104{
5105	struct sctp_inpcb *inp;
5106	int answer;
5107
5108	/*
5109	 * No locks here, the TCB, in all cases is already locked and an
5110	 * assoc is up. There is either a INP lock by the caller applied (in
5111	 * asconf case when deleting an address) or NOT in the HB case,
5112	 * however if HB then the INP increment is up and the INP will not
5113	 * be removed (on top of the fact that we have a TCB lock). So we
5114	 * only want to read the sctp_flags, which is either bound-all or
5115	 * not.. no protection needed since once an assoc is up you can't be
5116	 * changing your binding.
5117	 */
5118	inp = stcb->sctp_ep;
5119	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5120		/* if bound all, destination is not restricted */
5121		/*
5122		 * RRS: Question during lock work: Is this correct? If you
5123		 * are bound-all you still might need to obey the V4--V6
5124		 * flags??? IMO this bound-all stuff needs to be removed!
5125		 */
5126		return (1);
5127	}
5128	/* NOTE: all "scope" checks are done when local addresses are added */
5129	if (destaddr->sa_family == AF_INET6) {
5130		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5131	} else if (destaddr->sa_family == AF_INET) {
5132		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5133	} else {
5134		/* invalid family, so it's unreachable */
5135		answer = 0;
5136	}
5137	return (answer);
5138}
5139
5140/*
5141 * update the inp_vflags on an endpoint
5142 */
5143static void
5144sctp_update_ep_vflag(struct sctp_inpcb *inp)
5145{
5146	struct sctp_laddr *laddr;
5147
5148	/* first clear the flag */
5149	inp->ip_inp.inp.inp_vflag = 0;
5150	/* set the flag based on addresses on the ep list */
5151	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5152		if (laddr->ifa == NULL) {
5153			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5154			    __FUNCTION__);
5155			continue;
5156		}
5157		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5158			continue;
5159		}
5160		if (laddr->ifa->address.sa.sa_family == AF_INET6) {
5161			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5162		} else if (laddr->ifa->address.sa.sa_family == AF_INET) {
5163			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5164		}
5165	}
5166}
5167
5168/*
5169 * Add the address to the endpoint local address list There is nothing to be
5170 * done if we are bound to all addresses
5171 */
5172void
5173sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5174{
5175	struct sctp_laddr *laddr;
5176	int fnd, error = 0;
5177
5178	fnd = 0;
5179
5180	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5181		/* You are already bound to all. You have it already */
5182		return;
5183	}
5184	if (ifa->address.sa.sa_family == AF_INET6) {
5185		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5186			/* Can't bind a non-useable addr. */
5187			return;
5188		}
5189	}
5190	/* first, is it already present? */
5191	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5192		if (laddr->ifa == ifa) {
5193			fnd = 1;
5194			break;
5195		}
5196	}
5197
5198	if (fnd == 0) {
5199		/* Not in the ep list */
5200		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5201		if (error != 0)
5202			return;
5203		inp->laddr_count++;
5204		/* update inp_vflag flags */
5205		if (ifa->address.sa.sa_family == AF_INET6) {
5206			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5207		} else if (ifa->address.sa.sa_family == AF_INET) {
5208			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5209		}
5210	}
5211	return;
5212}
5213
5214
5215/*
5216 * select a new (hopefully reachable) destination net (should only be used
5217 * when we deleted an ep addr that is the only usable source address to reach
5218 * the destination net)
5219 */
5220static void
5221sctp_select_primary_destination(struct sctp_tcb *stcb)
5222{
5223	struct sctp_nets *net;
5224
5225	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5226		/* for now, we'll just pick the first reachable one we find */
5227		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5228			continue;
5229		if (sctp_destination_is_reachable(stcb,
5230		    (struct sockaddr *)&net->ro._l_addr)) {
5231			/* found a reachable destination */
5232			stcb->asoc.primary_destination = net;
5233		}
5234	}
5235	/* I can't there from here! ...we're gonna die shortly... */
5236}
5237
5238
5239/*
5240 * Delete the address from the endpoint local address list There is nothing
5241 * to be done if we are bound to all addresses
5242 */
5243void
5244sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5245{
5246	struct sctp_laddr *laddr;
5247	int fnd;
5248
5249	fnd = 0;
5250	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5251		/* You are already bound to all. You have it already */
5252		return;
5253	}
5254	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5255		if (laddr->ifa == ifa) {
5256			fnd = 1;
5257			break;
5258		}
5259	}
5260	if (fnd && (inp->laddr_count < 2)) {
5261		/* can't delete unless there are at LEAST 2 addresses */
5262		return;
5263	}
5264	if (fnd) {
5265		/*
5266		 * clean up any use of this address go through our
5267		 * associations and clear any last_used_address that match
5268		 * this one for each assoc, see if a new primary_destination
5269		 * is needed
5270		 */
5271		struct sctp_tcb *stcb;
5272
5273		/* clean up "next_addr_touse" */
5274		if (inp->next_addr_touse == laddr)
5275			/* delete this address */
5276			inp->next_addr_touse = NULL;
5277
5278		/* clean up "last_used_address" */
5279		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5280			struct sctp_nets *net;
5281
5282			SCTP_TCB_LOCK(stcb);
5283			if (stcb->asoc.last_used_address == laddr)
5284				/* delete this address */
5285				stcb->asoc.last_used_address = NULL;
5286			/*
5287			 * Now spin through all the nets and purge any ref
5288			 * to laddr
5289			 */
5290			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5291				if (net->ro._s_addr &&
5292				    (net->ro._s_addr->ifa == laddr->ifa)) {
5293					/* Yep, purge src address selected */
5294					sctp_rtentry_t *rt;
5295
5296					/* delete this address if cached */
5297					rt = net->ro.ro_rt;
5298					if (rt != NULL) {
5299						RTFREE(rt);
5300						net->ro.ro_rt = NULL;
5301					}
5302					sctp_free_ifa(net->ro._s_addr);
5303					net->ro._s_addr = NULL;
5304					net->src_addr_selected = 0;
5305				}
5306			}
5307			SCTP_TCB_UNLOCK(stcb);
5308		}		/* for each tcb */
5309		/* remove it from the ep list */
5310		sctp_remove_laddr(laddr);
5311		inp->laddr_count--;
5312		/* update inp_vflag flags */
5313		sctp_update_ep_vflag(inp);
5314	}
5315	return;
5316}
5317
5318/*
5319 * Add the address to the TCB local address restricted list.
5320 * This is a "pending" address list (eg. addresses waiting for an
5321 * ASCONF-ACK response) and cannot be used as a valid source address.
5322 */
5323void
5324sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5325{
5326	struct sctp_inpcb *inp;
5327	struct sctp_laddr *laddr;
5328	struct sctpladdr *list;
5329
5330	/*
5331	 * Assumes TCB is locked.. and possibly the INP. May need to
5332	 * confirm/fix that if we need it and is not the case.
5333	 */
5334	list = &stcb->asoc.sctp_restricted_addrs;
5335
5336	inp = stcb->sctp_ep;
5337	if (ifa->address.sa.sa_family == AF_INET6) {
5338		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5339			/* Can't bind a non-existent addr. */
5340			return;
5341		}
5342	}
5343	/* does the address already exist? */
5344	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5345		if (laddr->ifa == ifa) {
5346			return;
5347		}
5348	}
5349
5350	/* add to the list */
5351	(void)sctp_insert_laddr(list, ifa, 0);
5352	return;
5353}
5354
5355/*
5356 * insert an laddr entry with the given ifa for the desired list
5357 */
5358int
5359sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5360{
5361	struct sctp_laddr *laddr;
5362
5363	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5364	if (laddr == NULL) {
5365		/* out of memory? */
5366		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5367		return (EINVAL);
5368	}
5369	SCTP_INCR_LADDR_COUNT();
5370	bzero(laddr, sizeof(*laddr));
5371	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5372	laddr->ifa = ifa;
5373	laddr->action = act;
5374	atomic_add_int(&ifa->refcount, 1);
5375	/* insert it */
5376	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5377
5378	return (0);
5379}
5380
5381/*
5382 * Remove an laddr entry from the local address list (on an assoc)
5383 */
5384void
5385sctp_remove_laddr(struct sctp_laddr *laddr)
5386{
5387
5388	/* remove from the list */
5389	LIST_REMOVE(laddr, sctp_nxt_addr);
5390	sctp_free_ifa(laddr->ifa);
5391	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5392	SCTP_DECR_LADDR_COUNT();
5393}
5394
5395/*
5396 * Remove a local address from the TCB local address restricted list
5397 */
5398void
5399sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5400{
5401	struct sctp_inpcb *inp;
5402	struct sctp_laddr *laddr;
5403
5404	/*
5405	 * This is called by asconf work. It is assumed that a) The TCB is
5406	 * locked and b) The INP is locked. This is true in as much as I can
5407	 * trace through the entry asconf code where I did these locks.
5408	 * Again, the ASCONF code is a bit different in that it does lock
5409	 * the INP during its work often times. This must be since we don't
5410	 * want other proc's looking up things while what they are looking
5411	 * up is changing :-D
5412	 */
5413
5414	inp = stcb->sctp_ep;
5415	/* if subset bound and don't allow ASCONF's, can't delete last */
5416	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5417	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5418		if (stcb->sctp_ep->laddr_count < 2) {
5419			/* can't delete last address */
5420			return;
5421		}
5422	}
5423	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5424		/* remove the address if it exists */
5425		if (laddr->ifa == NULL)
5426			continue;
5427		if (laddr->ifa == ifa) {
5428			sctp_remove_laddr(laddr);
5429			return;
5430		}
5431	}
5432
5433	/* address not found! */
5434	return;
5435}
5436
5437/*
5438 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5439 */
5440/* sysctl */
5441static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5442static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5443
5444
5445
5446#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5447struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5448int *sctp_cpuarry = NULL;
5449void
5450sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5451{
5452	/* Queue a packet to a processor for the specified core */
5453	struct sctp_mcore_queue *qent;
5454	struct sctp_mcore_ctrl *wkq;
5455	int need_wake = 0;
5456
5457	if (sctp_mcore_workers == NULL) {
5458		/* Something went way bad during setup */
5459		sctp_input_with_port(m, off, 0);
5460		return;
5461	}
5462	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5463	    (sizeof(struct sctp_mcore_queue)),
5464	    SCTP_M_MCORE);
5465	if (qent == NULL) {
5466		/* This is trouble  */
5467		sctp_input_with_port(m, off, 0);
5468		return;
5469	}
5470	qent->vn = curvnet;
5471	qent->m = m;
5472	qent->off = off;
5473	qent->v6 = 0;
5474	wkq = &sctp_mcore_workers[cpu_to_use];
5475	SCTP_MCORE_QLOCK(wkq);
5476
5477	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5478	if (wkq->running == 0) {
5479		need_wake = 1;
5480	}
5481	SCTP_MCORE_QUNLOCK(wkq);
5482	if (need_wake) {
5483		wakeup(&wkq->running);
5484	}
5485}
5486
5487static void
5488sctp_mcore_thread(void *arg)
5489{
5490
5491	struct sctp_mcore_ctrl *wkq;
5492	struct sctp_mcore_queue *qent;
5493
5494	wkq = (struct sctp_mcore_ctrl *)arg;
5495	struct mbuf *m;
5496	int off, v6;
5497
5498	/* Wait for first tickle */
5499	SCTP_MCORE_LOCK(wkq);
5500	wkq->running = 0;
5501	msleep(&wkq->running,
5502	    &wkq->core_mtx,
5503	    0, "wait for pkt", 0);
5504	SCTP_MCORE_UNLOCK(wkq);
5505
5506	/* Bind to our cpu */
5507	thread_lock(curthread);
5508	sched_bind(curthread, wkq->cpuid);
5509	thread_unlock(curthread);
5510
5511	/* Now lets start working */
5512	SCTP_MCORE_LOCK(wkq);
5513	/* Now grab lock and go */
5514	while (1) {
5515		SCTP_MCORE_QLOCK(wkq);
5516skip_sleep:
5517		wkq->running = 1;
5518		qent = TAILQ_FIRST(&wkq->que);
5519		if (qent) {
5520			TAILQ_REMOVE(&wkq->que, qent, next);
5521			SCTP_MCORE_QUNLOCK(wkq);
5522			CURVNET_SET(qent->vn);
5523			m = qent->m;
5524			off = qent->off;
5525			v6 = qent->v6;
5526			SCTP_FREE(qent, SCTP_M_MCORE);
5527			if (v6 == 0) {
5528				sctp_input_with_port(m, off, 0);
5529			} else {
5530				printf("V6 not yet supported\n");
5531				sctp_m_freem(m);
5532			}
5533			CURVNET_RESTORE();
5534			SCTP_MCORE_QLOCK(wkq);
5535		}
5536		wkq->running = 0;
5537		if (!TAILQ_EMPTY(&wkq->que)) {
5538			goto skip_sleep;
5539		}
5540		SCTP_MCORE_QUNLOCK(wkq);
5541		msleep(&wkq->running,
5542		    &wkq->core_mtx,
5543		    0, "wait for pkt", 0);
5544	};
5545}
5546
5547static void
5548sctp_startup_mcore_threads(void)
5549{
5550	int i, cpu;
5551
5552	if (mp_ncpus == 1)
5553		return;
5554
5555	if (sctp_mcore_workers != NULL) {
5556		/*
5557		 * Already been here in some previous vnet?
5558		 */
5559		return;
5560	}
5561	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5562	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5563	    SCTP_M_MCORE);
5564	if (sctp_mcore_workers == NULL) {
5565		/* TSNH I hope */
5566		return;
5567	}
5568	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5569	    sizeof(struct sctp_mcore_ctrl)));
5570	/* Init the structures */
5571	for (i = 0; i <= mp_maxid; i++) {
5572		TAILQ_INIT(&sctp_mcore_workers[i].que);
5573		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5574		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5575		sctp_mcore_workers[i].cpuid = i;
5576	}
5577	if (sctp_cpuarry == NULL) {
5578		SCTP_MALLOC(sctp_cpuarry, int *,
5579		    (mp_ncpus * sizeof(int)),
5580		    SCTP_M_MCORE);
5581		i = 0;
5582		CPU_FOREACH(cpu) {
5583			sctp_cpuarry[i] = cpu;
5584			i++;
5585		}
5586	}
5587	/* Now start them all */
5588	CPU_FOREACH(cpu) {
5589		(void)kproc_create(sctp_mcore_thread,
5590		    (void *)&sctp_mcore_workers[cpu],
5591		    &sctp_mcore_workers[cpu].thread_proc,
5592		    RFPROC,
5593		    SCTP_KTHREAD_PAGES,
5594		    SCTP_MCORE_NAME);
5595
5596	}
5597}
5598
5599#endif
5600
5601
5602void
5603sctp_pcb_init()
5604{
5605	/*
5606	 * SCTP initialization for the PCB structures should be called by
5607	 * the sctp_init() funciton.
5608	 */
5609	int i;
5610	struct timeval tv;
5611
5612	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5613		/* error I was called twice */
5614		return;
5615	}
5616	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5617
5618#if defined(SCTP_LOCAL_TRACE_BUF)
5619	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5620#endif
5621#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5622	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5623	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5624	    SCTP_M_MCORE);
5625#endif
5626	(void)SCTP_GETTIME_TIMEVAL(&tv);
5627#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5628	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5629	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5630	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5631#else
5632	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5633	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5634	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5635#endif
5636	/* init the empty list of (All) Endpoints */
5637	LIST_INIT(&SCTP_BASE_INFO(listhead));
5638
5639
5640	/* init the hash table of endpoints */
5641	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5642	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5643	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5644	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5645	    &SCTP_BASE_INFO(hashasocmark));
5646	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5647	    &SCTP_BASE_INFO(hashmark));
5648	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5649	    &SCTP_BASE_INFO(hashtcpmark));
5650	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5651
5652
5653	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5654	    &SCTP_BASE_INFO(hashvrfmark));
5655
5656	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5657	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5658	/* init the zones */
5659	/*
5660	 * FIX ME: Should check for NULL returns, but if it does fail we are
5661	 * doomed to panic anyways... add later maybe.
5662	 */
5663	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5664	    sizeof(struct sctp_inpcb), maxsockets);
5665
5666	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5667	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5668
5669	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5670	    sizeof(struct sctp_laddr),
5671	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5672
5673	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5674	    sizeof(struct sctp_nets),
5675	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5676
5677	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5678	    sizeof(struct sctp_tmit_chunk),
5679	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5680
5681	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5682	    sizeof(struct sctp_queued_to_read),
5683	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5684
5685	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5686	    sizeof(struct sctp_stream_queue_pending),
5687	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5688
5689	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5690	    sizeof(struct sctp_asconf),
5691	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5692
5693	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5694	    sizeof(struct sctp_asconf_ack),
5695	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5696
5697
5698	/* Master Lock INIT for info structure */
5699	SCTP_INP_INFO_LOCK_INIT();
5700	SCTP_STATLOG_INIT_LOCK();
5701
5702	SCTP_IPI_COUNT_INIT();
5703	SCTP_IPI_ADDR_INIT();
5704#ifdef SCTP_PACKET_LOGGING
5705	SCTP_IP_PKTLOG_INIT();
5706#endif
5707	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5708
5709	SCTP_WQ_ADDR_INIT();
5710	/* not sure if we need all the counts */
5711	SCTP_BASE_INFO(ipi_count_ep) = 0;
5712	/* assoc/tcb zone info */
5713	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5714	/* local addrlist zone info */
5715	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5716	/* remote addrlist zone info */
5717	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5718	/* chunk info */
5719	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5720
5721	/* socket queue zone info */
5722	SCTP_BASE_INFO(ipi_count_readq) = 0;
5723
5724	/* stream out queue cont */
5725	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5726
5727	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5728	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5729
5730	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5731
5732	/* Init the TIMEWAIT list */
5733	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5734		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5735	}
5736
5737	sctp_startup_iterator();
5738
5739#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5740	sctp_startup_mcore_threads();
5741#endif
5742
5743	/*
5744	 * INIT the default VRF which for BSD is the only one, other O/S's
5745	 * may have more. But initially they must start with one and then
5746	 * add the VRF's as addresses are added.
5747	 */
5748	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5749
5750}
5751
5752/*
5753 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5754 */
5755void
5756sctp_pcb_finish(void)
5757{
5758	struct sctp_vrflist *vrf_bucket;
5759	struct sctp_vrf *vrf, *nvrf;
5760	struct sctp_ifn *ifn, *nifn;
5761	struct sctp_ifa *ifa, *nifa;
5762	struct sctpvtaghead *chain;
5763	struct sctp_tagblock *twait_block, *prev_twait_block;
5764	struct sctp_laddr *wi, *nwi;
5765	int i;
5766
5767	/*
5768	 * Free BSD the it thread never exits but we do clean up. The only
5769	 * way freebsd reaches here if we have VRF's but we still add the
5770	 * ifdef to make it compile on old versions.
5771	 */
5772	{
5773		struct sctp_iterator *it, *nit;
5774
5775		SCTP_IPI_ITERATOR_WQ_LOCK();
5776		TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5777			if (it->vn != curvnet) {
5778				continue;
5779			}
5780			TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5781			if (it->function_atend != NULL) {
5782				(*it->function_atend) (it->pointer, it->val);
5783			}
5784			SCTP_FREE(it, SCTP_M_ITER);
5785		}
5786		SCTP_IPI_ITERATOR_WQ_UNLOCK();
5787		SCTP_ITERATOR_LOCK();
5788		if ((sctp_it_ctl.cur_it) &&
5789		    (sctp_it_ctl.cur_it->vn == curvnet)) {
5790			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5791		}
5792		SCTP_ITERATOR_UNLOCK();
5793	}
5794
5795	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5796	SCTP_WQ_ADDR_LOCK();
5797	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5798		LIST_REMOVE(wi, sctp_nxt_addr);
5799		SCTP_DECR_LADDR_COUNT();
5800		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5801	}
5802	SCTP_WQ_ADDR_UNLOCK();
5803
5804	/*
5805	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5806	 * destroyed first).
5807	 */
5808	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5809	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5810		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5811			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5812				/* free the ifa */
5813				LIST_REMOVE(ifa, next_bucket);
5814				LIST_REMOVE(ifa, next_ifa);
5815				SCTP_FREE(ifa, SCTP_M_IFA);
5816			}
5817			/* free the ifn */
5818			LIST_REMOVE(ifn, next_bucket);
5819			LIST_REMOVE(ifn, next_ifn);
5820			SCTP_FREE(ifn, SCTP_M_IFN);
5821		}
5822		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5823		/* free the vrf */
5824		LIST_REMOVE(vrf, next_vrf);
5825		SCTP_FREE(vrf, SCTP_M_VRF);
5826	}
5827	/* free the vrf hashes */
5828	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5829	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5830
5831	/*
5832	 * free the TIMEWAIT list elements malloc'd in the function
5833	 * sctp_add_vtag_to_timewait()...
5834	 */
5835	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5836		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5837		if (!LIST_EMPTY(chain)) {
5838			prev_twait_block = NULL;
5839			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5840				if (prev_twait_block) {
5841					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5842				}
5843				prev_twait_block = twait_block;
5844			}
5845			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5846		}
5847	}
5848
5849	/* free the locks and mutexes */
5850#ifdef SCTP_PACKET_LOGGING
5851	SCTP_IP_PKTLOG_DESTROY();
5852#endif
5853	SCTP_IPI_ADDR_DESTROY();
5854	SCTP_STATLOG_DESTROY();
5855	SCTP_INP_INFO_LOCK_DESTROY();
5856
5857	SCTP_WQ_ADDR_DESTROY();
5858
5859	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
5860	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
5861	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
5862	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
5863	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
5864	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
5865	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
5866	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
5867	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
5868	/* Get rid of other stuff to */
5869	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5870		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5871	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5872		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
5873	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
5874		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
5875#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5876	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
5877#endif
5878}
5879
5880
5881int
5882sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
5883    int iphlen, int offset, int limit, struct sctphdr *sh,
5884    struct sockaddr *altsa)
5885{
5886	/*
5887	 * grub through the INIT pulling addresses and loading them to the
5888	 * nets structure in the asoc. The from address in the mbuf should
5889	 * also be loaded (if it is not already). This routine can be called
5890	 * with either INIT or INIT-ACK's as long as the m points to the IP
5891	 * packet and the offset points to the beginning of the parameters.
5892	 */
5893	struct sctp_inpcb *inp, *l_inp;
5894	struct sctp_nets *net, *nnet, *net_tmp;
5895	struct ip *iph;
5896	struct sctp_paramhdr *phdr, parm_buf;
5897	struct sctp_tcb *stcb_tmp;
5898	uint16_t ptype, plen;
5899	struct sockaddr *sa;
5900	struct sockaddr_storage dest_store;
5901	struct sockaddr *local_sa = (struct sockaddr *)&dest_store;
5902	struct sockaddr_in sin;
5903	struct sockaddr_in6 sin6;
5904	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
5905	struct sctp_auth_random *p_random = NULL;
5906	uint16_t random_len = 0;
5907	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
5908	struct sctp_auth_hmac_algo *hmacs = NULL;
5909	uint16_t hmacs_len = 0;
5910	uint8_t saw_asconf = 0;
5911	uint8_t saw_asconf_ack = 0;
5912	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
5913	struct sctp_auth_chunk_list *chunks = NULL;
5914	uint16_t num_chunks = 0;
5915	sctp_key_t *new_key;
5916	uint32_t keylen;
5917	int got_random = 0, got_hmacs = 0, got_chklist = 0;
5918	uint8_t ecn_allowed;
5919
5920	/* First get the destination address setup too. */
5921	memset(&sin, 0, sizeof(sin));
5922	memset(&sin6, 0, sizeof(sin6));
5923
5924	sin.sin_family = AF_INET;
5925	sin.sin_len = sizeof(sin);
5926	sin.sin_port = stcb->rport;
5927
5928	sin6.sin6_family = AF_INET6;
5929	sin6.sin6_len = sizeof(struct sockaddr_in6);
5930	sin6.sin6_port = stcb->rport;
5931	if (altsa == NULL) {
5932		iph = mtod(m, struct ip *);
5933		switch (iph->ip_v) {
5934		case IPVERSION:
5935			{
5936				/* its IPv4 */
5937				struct sockaddr_in *sin_2;
5938
5939				sin_2 = (struct sockaddr_in *)(local_sa);
5940				memset(sin_2, 0, sizeof(sin));
5941				sin_2->sin_family = AF_INET;
5942				sin_2->sin_len = sizeof(sin);
5943				sin_2->sin_port = sh->dest_port;
5944				sin_2->sin_addr.s_addr = iph->ip_dst.s_addr;
5945				sin.sin_addr = iph->ip_src;
5946				sa = (struct sockaddr *)&sin;
5947				break;
5948			}
5949#ifdef INET6
5950		case IPV6_VERSION >> 4:
5951			{
5952				/* its IPv6 */
5953				struct ip6_hdr *ip6;
5954				struct sockaddr_in6 *sin6_2;
5955
5956				ip6 = mtod(m, struct ip6_hdr *);
5957				sin6_2 = (struct sockaddr_in6 *)(local_sa);
5958				memset(sin6_2, 0, sizeof(sin6));
5959				sin6_2->sin6_family = AF_INET6;
5960				sin6_2->sin6_len = sizeof(struct sockaddr_in6);
5961				sin6_2->sin6_port = sh->dest_port;
5962				sin6.sin6_addr = ip6->ip6_src;
5963				sa = (struct sockaddr *)&sin6;
5964				break;
5965			}
5966#endif
5967		default:
5968			return (-1);
5969			break;
5970		}
5971	} else {
5972		/*
5973		 * For cookies we use the src address NOT from the packet
5974		 * but from the original INIT
5975		 */
5976		sa = altsa;
5977	}
5978	/* Turn off ECN until we get through all params */
5979	ecn_allowed = 0;
5980	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5981		/* mark all addresses that we have currently on the list */
5982		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
5983	}
5984	/* does the source address already exist? if so skip it */
5985	l_inp = inp = stcb->sctp_ep;
5986
5987	atomic_add_int(&stcb->asoc.refcnt, 1);
5988	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb);
5989	atomic_add_int(&stcb->asoc.refcnt, -1);
5990
5991	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
5992		/* we must add the source address */
5993		/* no scope set here since we have a tcb already. */
5994		if ((sa->sa_family == AF_INET) &&
5995		    (stcb->asoc.ipv4_addr_legal)) {
5996			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
5997				return (-1);
5998			}
5999		} else if ((sa->sa_family == AF_INET6) &&
6000		    (stcb->asoc.ipv6_addr_legal)) {
6001			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6002				return (-2);
6003			}
6004		}
6005	} else {
6006		if (net_tmp != NULL && stcb_tmp == stcb) {
6007			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6008		} else if (stcb_tmp != stcb) {
6009			/* It belongs to another association? */
6010			if (stcb_tmp)
6011				SCTP_TCB_UNLOCK(stcb_tmp);
6012			return (-3);
6013		}
6014	}
6015	if (stcb->asoc.state == 0) {
6016		/* the assoc was freed? */
6017		return (-4);
6018	}
6019	/*
6020	 * peer must explicitly turn this on. This may have been initialized
6021	 * to be "on" in order to allow local addr changes while INIT's are
6022	 * in flight.
6023	 */
6024	stcb->asoc.peer_supports_asconf = 0;
6025	/* now we must go through each of the params. */
6026	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6027	while (phdr) {
6028		ptype = ntohs(phdr->param_type);
6029		plen = ntohs(phdr->param_length);
6030		/*
6031		 * printf("ptype => %0x, plen => %d\n", (uint32_t)ptype,
6032		 * (int)plen);
6033		 */
6034		if (offset + plen > limit) {
6035			break;
6036		}
6037		if (plen == 0) {
6038			break;
6039		}
6040		if (ptype == SCTP_IPV4_ADDRESS) {
6041			if (stcb->asoc.ipv4_addr_legal) {
6042				struct sctp_ipv4addr_param *p4, p4_buf;
6043
6044				/* ok get the v4 address and check/add */
6045				phdr = sctp_get_next_param(m, offset,
6046				    (struct sctp_paramhdr *)&p4_buf,
6047				    sizeof(p4_buf));
6048				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6049				    phdr == NULL) {
6050					return (-5);
6051				}
6052				p4 = (struct sctp_ipv4addr_param *)phdr;
6053				sin.sin_addr.s_addr = p4->addr;
6054				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6055					/* Skip multi-cast addresses */
6056					goto next_param;
6057				}
6058				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6059				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6060					goto next_param;
6061				}
6062				sa = (struct sockaddr *)&sin;
6063				inp = stcb->sctp_ep;
6064				atomic_add_int(&stcb->asoc.refcnt, 1);
6065				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6066				    local_sa, stcb);
6067				atomic_add_int(&stcb->asoc.refcnt, -1);
6068
6069				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6070				    inp == NULL) {
6071					/* we must add the source address */
6072					/*
6073					 * no scope set since we have a tcb
6074					 * already
6075					 */
6076
6077					/*
6078					 * we must validate the state again
6079					 * here
6080					 */
6081			add_it_now:
6082					if (stcb->asoc.state == 0) {
6083						/* the assoc was freed? */
6084						return (-7);
6085					}
6086					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6087						return (-8);
6088					}
6089				} else if (stcb_tmp == stcb) {
6090					if (stcb->asoc.state == 0) {
6091						/* the assoc was freed? */
6092						return (-10);
6093					}
6094					if (net != NULL) {
6095						/* clear flag */
6096						net->dest_state &=
6097						    ~SCTP_ADDR_NOT_IN_ASSOC;
6098					}
6099				} else {
6100					/*
6101					 * strange, address is in another
6102					 * assoc? straighten out locks.
6103					 */
6104					if (stcb_tmp) {
6105						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6106							/*
6107							 * in setup state we
6108							 * abort this guy
6109							 */
6110							sctp_abort_an_association(stcb_tmp->sctp_ep,
6111							    stcb_tmp, 1, NULL, 0);
6112							goto add_it_now;
6113						}
6114						SCTP_TCB_UNLOCK(stcb_tmp);
6115					}
6116					if (stcb->asoc.state == 0) {
6117						/* the assoc was freed? */
6118						return (-12);
6119					}
6120					return (-13);
6121				}
6122			}
6123		} else if (ptype == SCTP_IPV6_ADDRESS) {
6124			if (stcb->asoc.ipv6_addr_legal) {
6125				/* ok get the v6 address and check/add */
6126				struct sctp_ipv6addr_param *p6, p6_buf;
6127
6128				phdr = sctp_get_next_param(m, offset,
6129				    (struct sctp_paramhdr *)&p6_buf,
6130				    sizeof(p6_buf));
6131				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6132				    phdr == NULL) {
6133					return (-14);
6134				}
6135				p6 = (struct sctp_ipv6addr_param *)phdr;
6136				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6137				    sizeof(p6->addr));
6138				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6139					/* Skip multi-cast addresses */
6140					goto next_param;
6141				}
6142				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6143					/*
6144					 * Link local make no sense without
6145					 * scope
6146					 */
6147					goto next_param;
6148				}
6149				sa = (struct sockaddr *)&sin6;
6150				inp = stcb->sctp_ep;
6151				atomic_add_int(&stcb->asoc.refcnt, 1);
6152				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6153				    local_sa, stcb);
6154				atomic_add_int(&stcb->asoc.refcnt, -1);
6155				if (stcb_tmp == NULL &&
6156				    (inp == stcb->sctp_ep || inp == NULL)) {
6157					/*
6158					 * we must validate the state again
6159					 * here
6160					 */
6161			add_it_now6:
6162					if (stcb->asoc.state == 0) {
6163						/* the assoc was freed? */
6164						return (-16);
6165					}
6166					/*
6167					 * we must add the address, no scope
6168					 * set
6169					 */
6170					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6171						return (-17);
6172					}
6173				} else if (stcb_tmp == stcb) {
6174					/*
6175					 * we must validate the state again
6176					 * here
6177					 */
6178					if (stcb->asoc.state == 0) {
6179						/* the assoc was freed? */
6180						return (-19);
6181					}
6182					if (net != NULL) {
6183						/* clear flag */
6184						net->dest_state &=
6185						    ~SCTP_ADDR_NOT_IN_ASSOC;
6186					}
6187				} else {
6188					/*
6189					 * strange, address is in another
6190					 * assoc? straighten out locks.
6191					 */
6192					if (stcb_tmp)
6193						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6194							/*
6195							 * in setup state we
6196							 * abort this guy
6197							 */
6198							sctp_abort_an_association(stcb_tmp->sctp_ep,
6199							    stcb_tmp, 1, NULL, 0);
6200							goto add_it_now6;
6201						}
6202					SCTP_TCB_UNLOCK(stcb_tmp);
6203
6204					if (stcb->asoc.state == 0) {
6205						/* the assoc was freed? */
6206						return (-21);
6207					}
6208					return (-22);
6209				}
6210			}
6211		} else if (ptype == SCTP_ECN_CAPABLE) {
6212			ecn_allowed = 1;
6213		} else if (ptype == SCTP_ULP_ADAPTATION) {
6214			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6215				struct sctp_adaptation_layer_indication ai,
6216				                                *aip;
6217
6218				phdr = sctp_get_next_param(m, offset,
6219				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6220				aip = (struct sctp_adaptation_layer_indication *)phdr;
6221				if (aip) {
6222					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6223					stcb->asoc.adaptation_needed = 1;
6224				}
6225			}
6226		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6227			struct sctp_asconf_addr_param lstore, *fee;
6228			struct sctp_asconf_addrv4_param *fii;
6229			int lptype;
6230			struct sockaddr *lsa = NULL;
6231
6232			stcb->asoc.peer_supports_asconf = 1;
6233			if (plen > sizeof(lstore)) {
6234				return (-23);
6235			}
6236			phdr = sctp_get_next_param(m, offset,
6237			    (struct sctp_paramhdr *)&lstore,
6238			    min(plen, sizeof(lstore)));
6239			if (phdr == NULL) {
6240				return (-24);
6241			}
6242			fee = (struct sctp_asconf_addr_param *)phdr;
6243			lptype = ntohs(fee->addrp.ph.param_type);
6244			if (lptype == SCTP_IPV4_ADDRESS) {
6245				if (plen !=
6246				    sizeof(struct sctp_asconf_addrv4_param)) {
6247					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6248					    (int)sizeof(struct sctp_asconf_addrv4_param),
6249					    plen);
6250				} else {
6251					fii = (struct sctp_asconf_addrv4_param *)fee;
6252					sin.sin_addr.s_addr = fii->addrp.addr;
6253					lsa = (struct sockaddr *)&sin;
6254				}
6255			} else if (lptype == SCTP_IPV6_ADDRESS) {
6256				if (plen !=
6257				    sizeof(struct sctp_asconf_addr_param)) {
6258					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6259					    (int)sizeof(struct sctp_asconf_addr_param),
6260					    plen);
6261				} else {
6262					memcpy(sin6.sin6_addr.s6_addr,
6263					    fee->addrp.addr,
6264					    sizeof(fee->addrp.addr));
6265					lsa = (struct sockaddr *)&sin6;
6266				}
6267			}
6268			if (lsa) {
6269				(void)sctp_set_primary_addr(stcb, sa, NULL);
6270			}
6271		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6272			stcb->asoc.peer_supports_nat = 1;
6273		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6274			/* Peer supports pr-sctp */
6275			stcb->asoc.peer_supports_prsctp = 1;
6276		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6277			/* A supported extension chunk */
6278			struct sctp_supported_chunk_types_param *pr_supported;
6279			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6280			int num_ent, i;
6281
6282			phdr = sctp_get_next_param(m, offset,
6283			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6284			if (phdr == NULL) {
6285				return (-25);
6286			}
6287			stcb->asoc.peer_supports_asconf = 0;
6288			stcb->asoc.peer_supports_prsctp = 0;
6289			stcb->asoc.peer_supports_pktdrop = 0;
6290			stcb->asoc.peer_supports_strreset = 0;
6291			stcb->asoc.peer_supports_nr_sack = 0;
6292			stcb->asoc.peer_supports_auth = 0;
6293			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6294			num_ent = plen - sizeof(struct sctp_paramhdr);
6295			for (i = 0; i < num_ent; i++) {
6296				switch (pr_supported->chunk_types[i]) {
6297				case SCTP_ASCONF:
6298				case SCTP_ASCONF_ACK:
6299					stcb->asoc.peer_supports_asconf = 1;
6300					break;
6301				case SCTP_FORWARD_CUM_TSN:
6302					stcb->asoc.peer_supports_prsctp = 1;
6303					break;
6304				case SCTP_PACKET_DROPPED:
6305					stcb->asoc.peer_supports_pktdrop = 1;
6306					break;
6307				case SCTP_NR_SELECTIVE_ACK:
6308					stcb->asoc.peer_supports_nr_sack = 1;
6309					break;
6310				case SCTP_STREAM_RESET:
6311					stcb->asoc.peer_supports_strreset = 1;
6312					break;
6313				case SCTP_AUTHENTICATION:
6314					stcb->asoc.peer_supports_auth = 1;
6315					break;
6316				default:
6317					/* one I have not learned yet */
6318					break;
6319
6320				}
6321			}
6322		} else if (ptype == SCTP_RANDOM) {
6323			if (plen > sizeof(random_store))
6324				break;
6325			if (got_random) {
6326				/* already processed a RANDOM */
6327				goto next_param;
6328			}
6329			phdr = sctp_get_next_param(m, offset,
6330			    (struct sctp_paramhdr *)random_store,
6331			    min(sizeof(random_store), plen));
6332			if (phdr == NULL)
6333				return (-26);
6334			p_random = (struct sctp_auth_random *)phdr;
6335			random_len = plen - sizeof(*p_random);
6336			/* enforce the random length */
6337			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6338				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6339				return (-27);
6340			}
6341			got_random = 1;
6342		} else if (ptype == SCTP_HMAC_LIST) {
6343			int num_hmacs;
6344			int i;
6345
6346			if (plen > sizeof(hmacs_store))
6347				break;
6348			if (got_hmacs) {
6349				/* already processed a HMAC list */
6350				goto next_param;
6351			}
6352			phdr = sctp_get_next_param(m, offset,
6353			    (struct sctp_paramhdr *)hmacs_store,
6354			    min(plen, sizeof(hmacs_store)));
6355			if (phdr == NULL)
6356				return (-28);
6357			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6358			hmacs_len = plen - sizeof(*hmacs);
6359			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6360			/* validate the hmac list */
6361			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6362				return (-29);
6363			}
6364			if (stcb->asoc.peer_hmacs != NULL)
6365				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6366			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6367			if (stcb->asoc.peer_hmacs != NULL) {
6368				for (i = 0; i < num_hmacs; i++) {
6369					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6370					    ntohs(hmacs->hmac_ids[i]));
6371				}
6372			}
6373			got_hmacs = 1;
6374		} else if (ptype == SCTP_CHUNK_LIST) {
6375			int i;
6376
6377			if (plen > sizeof(chunks_store))
6378				break;
6379			if (got_chklist) {
6380				/* already processed a Chunks list */
6381				goto next_param;
6382			}
6383			phdr = sctp_get_next_param(m, offset,
6384			    (struct sctp_paramhdr *)chunks_store,
6385			    min(plen, sizeof(chunks_store)));
6386			if (phdr == NULL)
6387				return (-30);
6388			chunks = (struct sctp_auth_chunk_list *)phdr;
6389			num_chunks = plen - sizeof(*chunks);
6390			if (stcb->asoc.peer_auth_chunks != NULL)
6391				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6392			else
6393				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6394			for (i = 0; i < num_chunks; i++) {
6395				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6396				    stcb->asoc.peer_auth_chunks);
6397				/* record asconf/asconf-ack if listed */
6398				if (chunks->chunk_types[i] == SCTP_ASCONF)
6399					saw_asconf = 1;
6400				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6401					saw_asconf_ack = 1;
6402
6403			}
6404			got_chklist = 1;
6405		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6406			    (ptype == SCTP_STATE_COOKIE) ||
6407			    (ptype == SCTP_UNRECOG_PARAM) ||
6408			    (ptype == SCTP_COOKIE_PRESERVE) ||
6409			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6410			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6411			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6412			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6413		    (ptype == SCTP_SUCCESS_REPORT)) {
6414			 /* don't care */ ;
6415		} else {
6416			if ((ptype & 0x8000) == 0x0000) {
6417				/*
6418				 * must stop processing the rest of the
6419				 * param's. Any report bits were handled
6420				 * with the call to
6421				 * sctp_arethere_unrecognized_parameters()
6422				 * when the INIT or INIT-ACK was first seen.
6423				 */
6424				break;
6425			}
6426		}
6427
6428next_param:
6429		offset += SCTP_SIZE32(plen);
6430		if (offset >= limit) {
6431			break;
6432		}
6433		phdr = sctp_get_next_param(m, offset, &parm_buf,
6434		    sizeof(parm_buf));
6435	}
6436	/* Now check to see if we need to purge any addresses */
6437	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6438		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6439		    SCTP_ADDR_NOT_IN_ASSOC) {
6440			/* This address has been removed from the asoc */
6441			/* remove and free it */
6442			stcb->asoc.numnets--;
6443			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6444			sctp_free_remote_addr(net);
6445			if (net == stcb->asoc.primary_destination) {
6446				stcb->asoc.primary_destination = NULL;
6447				sctp_select_primary_destination(stcb);
6448			}
6449		}
6450	}
6451	if (ecn_allowed == 0) {
6452		stcb->asoc.ecn_allowed = 0;
6453	}
6454	/* validate authentication required parameters */
6455	if (got_random && got_hmacs) {
6456		stcb->asoc.peer_supports_auth = 1;
6457	} else {
6458		stcb->asoc.peer_supports_auth = 0;
6459	}
6460	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6461		/* peer does not support auth but sent a chunks list? */
6462		return (-31);
6463	}
6464	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6465	    !stcb->asoc.peer_supports_auth) {
6466		/* peer supports asconf but not auth? */
6467		return (-32);
6468	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6469	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6470		return (-33);
6471	}
6472	/* concatenate the full random key */
6473	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6474	if (chunks != NULL) {
6475		keylen += sizeof(*chunks) + num_chunks;
6476	}
6477	new_key = sctp_alloc_key(keylen);
6478	if (new_key != NULL) {
6479		/* copy in the RANDOM */
6480		if (p_random != NULL) {
6481			keylen = sizeof(*p_random) + random_len;
6482			bcopy(p_random, new_key->key, keylen);
6483		}
6484		/* append in the AUTH chunks */
6485		if (chunks != NULL) {
6486			bcopy(chunks, new_key->key + keylen,
6487			    sizeof(*chunks) + num_chunks);
6488			keylen += sizeof(*chunks) + num_chunks;
6489		}
6490		/* append in the HMACs */
6491		if (hmacs != NULL) {
6492			bcopy(hmacs, new_key->key + keylen,
6493			    sizeof(*hmacs) + hmacs_len);
6494		}
6495	} else {
6496		/* failed to get memory for the key */
6497		return (-34);
6498	}
6499	if (stcb->asoc.authinfo.peer_random != NULL)
6500		sctp_free_key(stcb->asoc.authinfo.peer_random);
6501	stcb->asoc.authinfo.peer_random = new_key;
6502	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6503	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6504
6505	return (0);
6506}
6507
6508int
6509sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6510    struct sctp_nets *net)
6511{
6512	/* make sure the requested primary address exists in the assoc */
6513	if (net == NULL && sa)
6514		net = sctp_findnet(stcb, sa);
6515
6516	if (net == NULL) {
6517		/* didn't find the requested primary address! */
6518		return (-1);
6519	} else {
6520		/* set the primary address */
6521		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6522			/* Must be confirmed, so queue to set */
6523			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6524			return (0);
6525		}
6526		stcb->asoc.primary_destination = net;
6527		net->dest_state &= ~SCTP_ADDR_WAS_PRIMARY;
6528		net = TAILQ_FIRST(&stcb->asoc.nets);
6529		if (net != stcb->asoc.primary_destination) {
6530			/*
6531			 * first one on the list is NOT the primary
6532			 * sctp_cmpaddr() is much more efficient if the
6533			 * primary is the first on the list, make it so.
6534			 */
6535			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6536			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6537		}
6538		return (0);
6539	}
6540}
6541
6542int
6543sctp_is_vtag_good(struct sctp_inpcb *inp, uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now, int save_in_twait)
6544{
6545	/*
6546	 * This function serves two purposes. It will see if a TAG can be
6547	 * re-used and return 1 for yes it is ok and 0 for don't use that
6548	 * tag. A secondary function it will do is purge out old tags that
6549	 * can be removed.
6550	 */
6551	struct sctpvtaghead *chain;
6552	struct sctp_tagblock *twait_block;
6553	struct sctpasochead *head;
6554	struct sctp_tcb *stcb;
6555	int i;
6556
6557	SCTP_INP_INFO_RLOCK();
6558	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6559	    SCTP_BASE_INFO(hashasocmark))];
6560	if (head == NULL) {
6561		/* invalid vtag */
6562		goto skip_vtag_check;
6563	}
6564	LIST_FOREACH(stcb, head, sctp_asocs) {
6565		/*
6566		 * We choose not to lock anything here. TCB's can't be
6567		 * removed since we have the read lock, so they can't be
6568		 * freed on us, same thing for the INP. I may be wrong with
6569		 * this assumption, but we will go with it for now :-)
6570		 */
6571		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6572			continue;
6573		}
6574		if (stcb->asoc.my_vtag == tag) {
6575			/* candidate */
6576			if (stcb->rport != rport) {
6577				continue;
6578			}
6579			if (stcb->sctp_ep->sctp_lport != lport) {
6580				continue;
6581			}
6582			/* Its a used tag set */
6583			SCTP_INP_INFO_RUNLOCK();
6584			return (0);
6585		}
6586	}
6587skip_vtag_check:
6588
6589	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6590	/* Now what about timed wait ? */
6591	if (!LIST_EMPTY(chain)) {
6592		/*
6593		 * Block(s) are present, lets see if we have this tag in the
6594		 * list
6595		 */
6596		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6597			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6598				if (twait_block->vtag_block[i].v_tag == 0) {
6599					/* not used */
6600					continue;
6601				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6602				    now->tv_sec) {
6603					/* Audit expires this guy */
6604					twait_block->vtag_block[i].tv_sec_at_expire = 0;
6605					twait_block->vtag_block[i].v_tag = 0;
6606					twait_block->vtag_block[i].lport = 0;
6607					twait_block->vtag_block[i].rport = 0;
6608				} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6609					    (twait_block->vtag_block[i].lport == lport) &&
6610				    (twait_block->vtag_block[i].rport == rport)) {
6611					/* Bad tag, sorry :< */
6612					SCTP_INP_INFO_RUNLOCK();
6613					return (0);
6614				}
6615			}
6616		}
6617	}
6618	SCTP_INP_INFO_RUNLOCK();
6619	return (1);
6620}
6621
6622
6623static sctp_assoc_t reneged_asoc_ids[256];
6624static uint8_t reneged_at = 0;
6625
6626
6627static void
6628sctp_drain_mbufs(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
6629{
6630	/*
6631	 * We must hunt this association for MBUF's past the cumack (i.e.
6632	 * out of order data that we can renege on).
6633	 */
6634	struct sctp_association *asoc;
6635	struct sctp_tmit_chunk *chk, *nchk;
6636	uint32_t cumulative_tsn_p1;
6637	struct sctp_queued_to_read *ctl, *nctl;
6638	int cnt, strmat;
6639	uint32_t gap, i;
6640	int fnd = 0;
6641
6642	/* We look for anything larger than the cum-ack + 1 */
6643
6644	asoc = &stcb->asoc;
6645	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6646		/* none we can reneg on. */
6647		return;
6648	}
6649	SCTP_STAT_INCR(sctps_protocol_drains_done);
6650	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6651	cnt = 0;
6652	/* First look in the re-assembly queue */
6653	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6654		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6655			/* Yep it is above cum-ack */
6656			cnt++;
6657			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6658			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6659			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6660			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6661			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6662			if (chk->data) {
6663				sctp_m_freem(chk->data);
6664				chk->data = NULL;
6665			}
6666			sctp_free_a_chunk(stcb, chk);
6667		}
6668	}
6669	/* Ok that was fun, now we will drain all the inbound streams? */
6670	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6671		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6672			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6673				/* Yep it is above cum-ack */
6674				cnt++;
6675				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6676				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6677				sctp_ucount_decr(asoc->cnt_on_all_streams);
6678				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6679				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6680				if (ctl->data) {
6681					sctp_m_freem(ctl->data);
6682					ctl->data = NULL;
6683				}
6684				sctp_free_remote_addr(ctl->whoFrom);
6685				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6686				SCTP_DECR_READQ_COUNT();
6687			}
6688		}
6689	}
6690	if (cnt) {
6691		/* We must back down to see what the new highest is */
6692		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6693			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6694			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6695				asoc->highest_tsn_inside_map = i;
6696				fnd = 1;
6697				break;
6698			}
6699		}
6700		if (!fnd) {
6701			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6702		}
6703		/*
6704		 * Question, should we go through the delivery queue? The
6705		 * only reason things are on here is the app not reading OR
6706		 * a p-d-api up. An attacker COULD send enough in to
6707		 * initiate the PD-API and then send a bunch of stuff to
6708		 * other streams... these would wind up on the delivery
6709		 * queue.. and then we would not get to them. But in order
6710		 * to do this I then have to back-track and un-deliver
6711		 * sequence numbers in streams.. el-yucko. I think for now
6712		 * we will NOT look at the delivery queue and leave it to be
6713		 * something to consider later. An alternative would be to
6714		 * abort the P-D-API with a notification and then deliver
6715		 * the data.... Or another method might be to keep track of
6716		 * how many times the situation occurs and if we see a
6717		 * possible attack underway just abort the association.
6718		 */
6719#ifdef SCTP_DEBUG
6720		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6721#endif
6722		/*
6723		 * Now do we need to find a new
6724		 * asoc->highest_tsn_inside_map?
6725		 */
6726		asoc->last_revoke_count = cnt;
6727		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6728		/* sa_ignore NO_NULL_CHK */
6729		sctp_send_sack(stcb);
6730		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6731		reneged_asoc_ids[reneged_at] = sctp_get_associd(stcb);
6732		reneged_at++;
6733	}
6734	/*
6735	 * Another issue, in un-setting the TSN's in the mapping array we
6736	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6737	 * two things to occur. It may cause us to do extra work in checking
6738	 * for our mapping array movement. More importantly it may cause us
6739	 * to SACK every datagram. This may not be a bad thing though since
6740	 * we will recover once we get our cum-ack above and all this stuff
6741	 * we dumped recovered.
6742	 */
6743}
6744
6745void
6746sctp_drain()
6747{
6748	/*
6749	 * We must walk the PCB lists for ALL associations here. The system
6750	 * is LOW on MBUF's and needs help. This is where reneging will
6751	 * occur. We really hope this does NOT happen!
6752	 */
6753	VNET_ITERATOR_DECL(vnet_iter);
6754	VNET_LIST_RLOCK_NOSLEEP();
6755	VNET_FOREACH(vnet_iter) {
6756		CURVNET_SET(vnet_iter);
6757		struct sctp_inpcb *inp;
6758		struct sctp_tcb *stcb;
6759
6760		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6761		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6762#ifdef VIMAGE
6763			continue;
6764#else
6765			return;
6766#endif
6767		}
6768		SCTP_INP_INFO_RLOCK();
6769		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6770			/* For each endpoint */
6771			SCTP_INP_RLOCK(inp);
6772			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6773				/* For each association */
6774				SCTP_TCB_LOCK(stcb);
6775				sctp_drain_mbufs(inp, stcb);
6776				SCTP_TCB_UNLOCK(stcb);
6777			}
6778			SCTP_INP_RUNLOCK(inp);
6779		}
6780		SCTP_INP_INFO_RUNLOCK();
6781		CURVNET_RESTORE();
6782	}
6783	VNET_LIST_RUNLOCK_NOSLEEP();
6784}
6785
6786/*
6787 * start a new iterator
6788 * iterates through all endpoints and associations based on the pcb_state
6789 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6790 * assocs and "ef" (optional) is executed when the iterator completes.
6791 * "inpf" (optional) is executed for each new endpoint as it is being
6792 * iterated through. inpe (optional) is called when the inp completes
6793 * its way through all the stcbs.
6794 */
6795int
6796sctp_initiate_iterator(inp_func inpf,
6797    asoc_func af,
6798    inp_func inpe,
6799    uint32_t pcb_state,
6800    uint32_t pcb_features,
6801    uint32_t asoc_state,
6802    void *argp,
6803    uint32_t argi,
6804    end_func ef,
6805    struct sctp_inpcb *s_inp,
6806    uint8_t chunk_output_off)
6807{
6808	struct sctp_iterator *it = NULL;
6809
6810	if (af == NULL) {
6811		return (-1);
6812	}
6813	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6814	    SCTP_M_ITER);
6815	if (it == NULL) {
6816		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6817		return (ENOMEM);
6818	}
6819	memset(it, 0, sizeof(*it));
6820	it->function_assoc = af;
6821	it->function_inp = inpf;
6822	if (inpf)
6823		it->done_current_ep = 0;
6824	else
6825		it->done_current_ep = 1;
6826	it->function_atend = ef;
6827	it->pointer = argp;
6828	it->val = argi;
6829	it->pcb_flags = pcb_state;
6830	it->pcb_features = pcb_features;
6831	it->asoc_state = asoc_state;
6832	it->function_inp_end = inpe;
6833	it->no_chunk_output = chunk_output_off;
6834	it->vn = curvnet;
6835	if (s_inp) {
6836		/* Assume lock is held here */
6837		it->inp = s_inp;
6838		SCTP_INP_INCR_REF(it->inp);
6839		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
6840	} else {
6841		SCTP_INP_INFO_RLOCK();
6842		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
6843		if (it->inp) {
6844			SCTP_INP_INCR_REF(it->inp);
6845		}
6846		SCTP_INP_INFO_RUNLOCK();
6847		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
6848
6849	}
6850	SCTP_IPI_ITERATOR_WQ_LOCK();
6851
6852	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
6853	if (sctp_it_ctl.iterator_running == 0) {
6854		sctp_wakeup_iterator();
6855	}
6856	SCTP_IPI_ITERATOR_WQ_UNLOCK();
6857	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
6858	return (0);
6859}
6860