1/*-
2 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * a) Redistributions of source code must retain the above copyright notice,
10 *    this list of conditions and the following disclaimer.
11 *
12 * b) Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *    the documentation and/or other materials provided with the distribution.
15 *
16 * c) Neither the name of Cisco Systems, Inc. nor the names of its
17 *    contributors may be used to endorse or promote products derived
18 *    from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33#include <sys/cdefs.h>
34__FBSDID("$FreeBSD$");
35
36#include <netinet/sctp_os.h>
37#include <sys/proc.h>
38#include <netinet/sctp_var.h>
39#include <netinet/sctp_sysctl.h>
40#include <netinet/sctp_pcb.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp.h>
43#include <netinet/sctp_header.h>
44#include <netinet/sctp_asconf.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_timer.h>
47#include <netinet/sctp_bsd_addr.h>
48#include <netinet/sctp_dtrace_define.h>
49#include <netinet/udp.h>
50#ifdef INET6
51#include <netinet6/ip6_var.h>
52#endif
53#include <sys/sched.h>
54#include <sys/smp.h>
55#include <sys/unistd.h>
56
57
58VNET_DEFINE(struct sctp_base_info, system_base_info);
59
60/* FIX: we don't handle multiple link local scopes */
61/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
62#ifdef INET6
63int
64SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
65{
66	struct sockaddr_in6 tmp_a, tmp_b;
67
68	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
69	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
70		return (0);
71	}
72	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
73	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
74		return (0);
75	}
76	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
77}
78
79#endif
80
81void
82sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
83{
84	/*
85	 * We really don't need to lock this, but I will just because it
86	 * does not hurt.
87	 */
88	SCTP_INP_INFO_RLOCK();
89	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
90	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
91	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
92	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
93	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
94	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
95	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
96	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
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
216struct sctp_vrf *
217sctp_find_vrf(uint32_t vrf_id)
218{
219	struct sctp_vrflist *bucket;
220	struct sctp_vrf *liste;
221
222	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
223	LIST_FOREACH(liste, bucket, next_vrf) {
224		if (vrf_id == liste->vrf_id) {
225			return (liste);
226		}
227	}
228	return (NULL);
229}
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
247
248void
249sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
250{
251	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
252		/* We zero'd the count */
253		if (sctp_ifnp->vrf) {
254			sctp_free_vrf(sctp_ifnp->vrf);
255		}
256		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
257		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
258	}
259}
260
261
262void
263sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
264{
265	struct sctp_ifn *sctp_ifnp;
266
267	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
268	if (sctp_ifnp != NULL) {
269		sctp_ifnp->ifn_mtu = mtu;
270	}
271}
272
273
274void
275sctp_free_ifa(struct sctp_ifa *sctp_ifap)
276{
277	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
278		/* We zero'd the count */
279		if (sctp_ifap->ifn_p) {
280			sctp_free_ifn(sctp_ifap->ifn_p);
281		}
282		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
283		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
284	}
285}
286
287
288static void
289sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
290{
291	struct sctp_ifn *found;
292
293	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
294	if (found == NULL) {
295		/* Not in the list.. sorry */
296		return;
297	}
298	if (hold_addr_lock == 0)
299		SCTP_IPI_ADDR_WLOCK();
300	LIST_REMOVE(sctp_ifnp, next_bucket);
301	LIST_REMOVE(sctp_ifnp, next_ifn);
302	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
303	    sctp_ifnp->registered_af);
304	if (hold_addr_lock == 0)
305		SCTP_IPI_ADDR_WUNLOCK();
306	/* Take away the reference, and possibly free it */
307	sctp_free_ifn(sctp_ifnp);
308}
309
310
311void
312sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
313    const char *if_name, uint32_t ifn_index)
314{
315	struct sctp_vrf *vrf;
316	struct sctp_ifa *sctp_ifap;
317
318	SCTP_IPI_ADDR_RLOCK();
319	vrf = sctp_find_vrf(vrf_id);
320	if (vrf == NULL) {
321		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
322		goto out;
323
324	}
325	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
326	if (sctp_ifap == NULL) {
327		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
328		goto out;
329	}
330	if (sctp_ifap->ifn_p == NULL) {
331		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
332		goto out;
333	}
334	if (if_name) {
335		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
336			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
337			    sctp_ifap->ifn_p->ifn_name, if_name);
338			goto out;
339		}
340	} else {
341		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
342			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
343			    sctp_ifap->ifn_p->ifn_index, ifn_index);
344			goto out;
345		}
346	}
347
348	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
349	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
350out:
351	SCTP_IPI_ADDR_RUNLOCK();
352}
353
354
355void
356sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
357    const char *if_name, uint32_t ifn_index)
358{
359	struct sctp_vrf *vrf;
360	struct sctp_ifa *sctp_ifap;
361
362	SCTP_IPI_ADDR_RLOCK();
363	vrf = sctp_find_vrf(vrf_id);
364	if (vrf == NULL) {
365		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
366		goto out;
367
368	}
369	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
370	if (sctp_ifap == NULL) {
371		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
372		goto out;
373	}
374	if (sctp_ifap->ifn_p == NULL) {
375		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
376		goto out;
377	}
378	if (if_name) {
379		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
380			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
381			    sctp_ifap->ifn_p->ifn_name, if_name);
382			goto out;
383		}
384	} else {
385		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
386			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
387			    sctp_ifap->ifn_p->ifn_index, ifn_index);
388			goto out;
389		}
390	}
391
392	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
393	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
394out:
395	SCTP_IPI_ADDR_RUNLOCK();
396}
397
398
399/*-
400 * Add an ifa to an ifn.
401 * Register the interface as necessary.
402 * NOTE: ADDR write lock MUST be held.
403 */
404static void
405sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
406{
407	int ifa_af;
408
409	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
410	sctp_ifap->ifn_p = sctp_ifnp;
411	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
412	/* update address counts */
413	sctp_ifnp->ifa_count++;
414	ifa_af = sctp_ifap->address.sa.sa_family;
415	switch (ifa_af) {
416#ifdef INET
417	case AF_INET:
418		sctp_ifnp->num_v4++;
419		break;
420#endif
421#ifdef INET6
422	case AF_INET6:
423		sctp_ifnp->num_v6++;
424		break;
425#endif
426	default:
427		break;
428	}
429	if (sctp_ifnp->ifa_count == 1) {
430		/* register the new interface */
431		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
432		sctp_ifnp->registered_af = ifa_af;
433	}
434}
435
436
437/*-
438 * Remove an ifa from its ifn.
439 * If no more addresses exist, remove the ifn too. Otherwise, re-register
440 * the interface based on the remaining address families left.
441 * NOTE: ADDR write lock MUST be held.
442 */
443static void
444sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
445{
446	LIST_REMOVE(sctp_ifap, next_ifa);
447	if (sctp_ifap->ifn_p) {
448		/* update address counts */
449		sctp_ifap->ifn_p->ifa_count--;
450		switch (sctp_ifap->address.sa.sa_family) {
451#ifdef INET
452		case AF_INET:
453			sctp_ifap->ifn_p->num_v4--;
454			break;
455#endif
456#ifdef INET6
457		case AF_INET6:
458			sctp_ifap->ifn_p->num_v6--;
459			break;
460#endif
461		default:
462			break;
463		}
464
465		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
466			/* remove the ifn, possibly freeing it */
467			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
468		} else {
469			/* re-register address family type, if needed */
470			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
471			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
472				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
473				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
474				sctp_ifap->ifn_p->registered_af = AF_INET;
475			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
476			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
477				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
478				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
479				sctp_ifap->ifn_p->registered_af = AF_INET6;
480			}
481			/* free the ifn refcount */
482			sctp_free_ifn(sctp_ifap->ifn_p);
483		}
484		sctp_ifap->ifn_p = NULL;
485	}
486}
487
488
489struct sctp_ifa *
490sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
491    uint32_t ifn_type, const char *if_name, void *ifa,
492    struct sockaddr *addr, uint32_t ifa_flags,
493    int dynamic_add)
494{
495	struct sctp_vrf *vrf;
496	struct sctp_ifn *sctp_ifnp = NULL;
497	struct sctp_ifa *sctp_ifap = NULL;
498	struct sctp_ifalist *hash_addr_head;
499	struct sctp_ifnlist *hash_ifn_head;
500	uint32_t hash_of_addr;
501	int new_ifn_af = 0;
502
503#ifdef SCTP_DEBUG
504	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
505	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
506#endif
507	SCTP_IPI_ADDR_WLOCK();
508	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
509	if (sctp_ifnp) {
510		vrf = sctp_ifnp->vrf;
511	} else {
512		vrf = sctp_find_vrf(vrf_id);
513		if (vrf == NULL) {
514			vrf = sctp_allocate_vrf(vrf_id);
515			if (vrf == NULL) {
516				SCTP_IPI_ADDR_WUNLOCK();
517				return (NULL);
518			}
519		}
520	}
521	if (sctp_ifnp == NULL) {
522		/*
523		 * build one and add it, can't hold lock until after malloc
524		 * done though.
525		 */
526		SCTP_IPI_ADDR_WUNLOCK();
527		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
528		    sizeof(struct sctp_ifn), SCTP_M_IFN);
529		if (sctp_ifnp == NULL) {
530#ifdef INVARIANTS
531			panic("No memory for IFN");
532#endif
533			return (NULL);
534		}
535		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
536		sctp_ifnp->ifn_index = ifn_index;
537		sctp_ifnp->ifn_p = ifn;
538		sctp_ifnp->ifn_type = ifn_type;
539		sctp_ifnp->refcount = 0;
540		sctp_ifnp->vrf = vrf;
541		atomic_add_int(&vrf->refcount, 1);
542		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
543		if (if_name != NULL) {
544			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
545		} else {
546			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
547		}
548		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
549		LIST_INIT(&sctp_ifnp->ifalist);
550		SCTP_IPI_ADDR_WLOCK();
551		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
552		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
553		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
554		new_ifn_af = 1;
555	}
556	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
557	if (sctp_ifap) {
558		/* Hmm, it already exists? */
559		if ((sctp_ifap->ifn_p) &&
560		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
561			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
562			    sctp_ifap->ifn_p->ifn_name, ifn_index,
563			    (void *)sctp_ifap);
564			if (new_ifn_af) {
565				/* Remove the created one that we don't want */
566				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
567			}
568			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
569				/* easy to solve, just switch back to active */
570				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
571				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
572				sctp_ifap->ifn_p = sctp_ifnp;
573				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
574			}
575	exit_stage_left:
576			SCTP_IPI_ADDR_WUNLOCK();
577			return (sctp_ifap);
578		} else {
579			if (sctp_ifap->ifn_p) {
580				/*
581				 * The last IFN gets the address, remove the
582				 * old one
583				 */
584				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
585				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
586				    sctp_ifap->ifn_p->ifn_index, if_name,
587				    ifn_index);
588				/* remove the address from the old ifn */
589				sctp_remove_ifa_from_ifn(sctp_ifap);
590				/* move the address over to the new ifn */
591				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
592				goto exit_stage_left;
593			} else {
594				/* repair ifnp which was NULL ? */
595				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
596				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
597				    (void *)sctp_ifnp, (void *)sctp_ifap);
598				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
599			}
600			goto exit_stage_left;
601		}
602	}
603	SCTP_IPI_ADDR_WUNLOCK();
604	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
605	if (sctp_ifap == NULL) {
606#ifdef INVARIANTS
607		panic("No memory for IFA");
608#endif
609		return (NULL);
610	}
611	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
612	sctp_ifap->ifn_p = sctp_ifnp;
613	atomic_add_int(&sctp_ifnp->refcount, 1);
614	sctp_ifap->vrf_id = vrf_id;
615	sctp_ifap->ifa = ifa;
616	memcpy(&sctp_ifap->address, addr, addr->sa_len);
617	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
618	sctp_ifap->flags = ifa_flags;
619	/* Set scope */
620	switch (sctp_ifap->address.sa.sa_family) {
621#ifdef INET
622	case AF_INET:
623		{
624			struct sockaddr_in *sin;
625
626			sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
627			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
628			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
629				sctp_ifap->src_is_loop = 1;
630			}
631			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
632				sctp_ifap->src_is_priv = 1;
633			}
634			sctp_ifnp->num_v4++;
635			if (new_ifn_af)
636				new_ifn_af = AF_INET;
637			break;
638		}
639#endif
640#ifdef INET6
641	case AF_INET6:
642		{
643			/* ok to use deprecated addresses? */
644			struct sockaddr_in6 *sin6;
645
646			sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
647			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
648			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
649				sctp_ifap->src_is_loop = 1;
650			}
651			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
652				sctp_ifap->src_is_priv = 1;
653			}
654			sctp_ifnp->num_v6++;
655			if (new_ifn_af)
656				new_ifn_af = AF_INET6;
657			break;
658		}
659#endif
660	default:
661		new_ifn_af = 0;
662		break;
663	}
664	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
665
666	if ((sctp_ifap->src_is_priv == 0) &&
667	    (sctp_ifap->src_is_loop == 0)) {
668		sctp_ifap->src_is_glob = 1;
669	}
670	SCTP_IPI_ADDR_WLOCK();
671	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
672	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
673	sctp_ifap->refcount = 1;
674	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
675	sctp_ifnp->ifa_count++;
676	vrf->total_ifa_count++;
677	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
678	if (new_ifn_af) {
679		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
680		sctp_ifnp->registered_af = new_ifn_af;
681	}
682	SCTP_IPI_ADDR_WUNLOCK();
683	if (dynamic_add) {
684		/*
685		 * Bump up the refcount so that when the timer completes it
686		 * will drop back down.
687		 */
688		struct sctp_laddr *wi;
689
690		atomic_add_int(&sctp_ifap->refcount, 1);
691		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
692		if (wi == NULL) {
693			/*
694			 * Gak, what can we do? We have lost an address
695			 * change can you say HOSED?
696			 */
697			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
698			/* Opps, must decrement the count */
699			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
700			    if_name);
701			return (NULL);
702		}
703		SCTP_INCR_LADDR_COUNT();
704		bzero(wi, sizeof(*wi));
705		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
706		wi->ifa = sctp_ifap;
707		wi->action = SCTP_ADD_IP_ADDRESS;
708
709		SCTP_WQ_ADDR_LOCK();
710		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
711		SCTP_WQ_ADDR_UNLOCK();
712
713		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
714		    (struct sctp_inpcb *)NULL,
715		    (struct sctp_tcb *)NULL,
716		    (struct sctp_nets *)NULL);
717	} else {
718		/* it's ready for use */
719		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
720	}
721	return (sctp_ifap);
722}
723
724void
725sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
726    uint32_t ifn_index, const char *if_name)
727{
728	struct sctp_vrf *vrf;
729	struct sctp_ifa *sctp_ifap = NULL;
730
731	SCTP_IPI_ADDR_WLOCK();
732	vrf = sctp_find_vrf(vrf_id);
733	if (vrf == NULL) {
734		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
735		goto out_now;
736	}
737#ifdef SCTP_DEBUG
738	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
739	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
740#endif
741	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
742	if (sctp_ifap) {
743		/* Validate the delete */
744		if (sctp_ifap->ifn_p) {
745			int valid = 0;
746
747			/*-
748			 * The name has priority over the ifn_index
749			 * if its given. We do this especially for
750			 * panda who might recycle indexes fast.
751			 */
752			if (if_name) {
753				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
754					/* They match its a correct delete */
755					valid = 1;
756				}
757			}
758			if (!valid) {
759				/* last ditch check ifn_index */
760				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
761					valid = 1;
762				}
763			}
764			if (!valid) {
765				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
766				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
767				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
768				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
769				SCTP_IPI_ADDR_WUNLOCK();
770				return;
771			}
772		}
773		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
774		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
775		/*
776		 * We don't set the flag. This means that the structure will
777		 * hang around in EP's that have bound specific to it until
778		 * they close. This gives us TCP like behavior if someone
779		 * removes an address (or for that matter adds it right
780		 * back).
781		 */
782		/* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */
783		vrf->total_ifa_count--;
784		LIST_REMOVE(sctp_ifap, next_bucket);
785		sctp_remove_ifa_from_ifn(sctp_ifap);
786	}
787#ifdef SCTP_DEBUG
788	else {
789		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
790		    ifn_index);
791		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
792	}
793#endif
794
795out_now:
796	SCTP_IPI_ADDR_WUNLOCK();
797	if (sctp_ifap) {
798		struct sctp_laddr *wi;
799
800		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
801		if (wi == NULL) {
802			/*
803			 * Gak, what can we do? We have lost an address
804			 * change can you say HOSED?
805			 */
806			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
807
808			/* Oops, must decrement the count */
809			sctp_free_ifa(sctp_ifap);
810			return;
811		}
812		SCTP_INCR_LADDR_COUNT();
813		bzero(wi, sizeof(*wi));
814		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
815		wi->ifa = sctp_ifap;
816		wi->action = SCTP_DEL_IP_ADDRESS;
817		SCTP_WQ_ADDR_LOCK();
818		/*
819		 * Should this really be a tailq? As it is we will process
820		 * the newest first :-0
821		 */
822		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
823		SCTP_WQ_ADDR_UNLOCK();
824
825		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
826		    (struct sctp_inpcb *)NULL,
827		    (struct sctp_tcb *)NULL,
828		    (struct sctp_nets *)NULL);
829	}
830	return;
831}
832
833
834static int
835sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
836{
837	int loopback_scope;
838
839#if defined(INET)
840	int ipv4_local_scope, ipv4_addr_legal;
841
842#endif
843#if defined(INET6)
844	int local_scope, site_scope, ipv6_addr_legal;
845
846#endif
847	struct sctp_vrf *vrf;
848	struct sctp_ifn *sctp_ifn;
849	struct sctp_ifa *sctp_ifa;
850
851	loopback_scope = stcb->asoc.scope.loopback_scope;
852#if defined(INET)
853	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
854	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
855#endif
856#if defined(INET6)
857	local_scope = stcb->asoc.scope.local_scope;
858	site_scope = stcb->asoc.scope.site_scope;
859	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
860#endif
861
862	SCTP_IPI_ADDR_RLOCK();
863	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
864	if (vrf == NULL) {
865		/* no vrf, no addresses */
866		SCTP_IPI_ADDR_RUNLOCK();
867		return (0);
868	}
869	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
870		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
871			if ((loopback_scope == 0) &&
872			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
873				continue;
874			}
875			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
876				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
877				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
878					/*
879					 * We allow pending addresses, where
880					 * we have sent an asconf-add to be
881					 * considered valid.
882					 */
883					continue;
884				}
885				if (sctp_ifa->address.sa.sa_family != to->sa_family) {
886					continue;
887				}
888				switch (sctp_ifa->address.sa.sa_family) {
889#ifdef INET
890				case AF_INET:
891					if (ipv4_addr_legal) {
892						struct sockaddr_in *sin,
893						           *rsin;
894
895						sin = &sctp_ifa->address.sin;
896						rsin = (struct sockaddr_in *)to;
897						if ((ipv4_local_scope == 0) &&
898						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
899							continue;
900						}
901						if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred,
902						    &sin->sin_addr) != 0) {
903							continue;
904						}
905						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
906							SCTP_IPI_ADDR_RUNLOCK();
907							return (1);
908						}
909					}
910					break;
911#endif
912#ifdef INET6
913				case AF_INET6:
914					if (ipv6_addr_legal) {
915						struct sockaddr_in6 *sin6,
916						            *rsin6;
917
918						sin6 = &sctp_ifa->address.sin6;
919						rsin6 = (struct sockaddr_in6 *)to;
920						if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred,
921						    &sin6->sin6_addr) != 0) {
922							continue;
923						}
924						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
925							if (local_scope == 0)
926								continue;
927							if (sin6->sin6_scope_id == 0) {
928								if (sa6_recoverscope(sin6) != 0)
929									continue;
930							}
931						}
932						if ((site_scope == 0) &&
933						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
934							continue;
935						}
936						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
937							SCTP_IPI_ADDR_RUNLOCK();
938							return (1);
939						}
940					}
941					break;
942#endif
943				default:
944					/* TSNH */
945					break;
946				}
947			}
948		}
949	} else {
950		struct sctp_laddr *laddr;
951
952		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
953			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
954				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
955				continue;
956			}
957			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
958			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
959				/*
960				 * We allow pending addresses, where we have
961				 * sent an asconf-add to be considered
962				 * valid.
963				 */
964				continue;
965			}
966			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
967				continue;
968			}
969			switch (to->sa_family) {
970#ifdef INET
971			case AF_INET:
972				{
973					struct sockaddr_in *sin, *rsin;
974
975					sin = (struct sockaddr_in *)&laddr->ifa->address.sin;
976					rsin = (struct sockaddr_in *)to;
977					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
978						SCTP_IPI_ADDR_RUNLOCK();
979						return (1);
980					}
981					break;
982				}
983#endif
984#ifdef INET6
985			case AF_INET6:
986				{
987					struct sockaddr_in6 *sin6, *rsin6;
988
989					sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
990					rsin6 = (struct sockaddr_in6 *)to;
991					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
992						SCTP_IPI_ADDR_RUNLOCK();
993						return (1);
994					}
995					break;
996				}
997
998#endif
999			default:
1000				/* TSNH */
1001				break;
1002			}
1003
1004		}
1005	}
1006	SCTP_IPI_ADDR_RUNLOCK();
1007	return (0);
1008}
1009
1010
1011static struct sctp_tcb *
1012sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
1013    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
1014{
1015	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
1016	/*
1017	 * If we support the TCP model, then we must now dig through to see
1018	 * if we can find our endpoint in the list of tcp ep's.
1019	 */
1020	uint16_t lport, rport;
1021	struct sctppcbhead *ephead;
1022	struct sctp_inpcb *inp;
1023	struct sctp_laddr *laddr;
1024	struct sctp_tcb *stcb;
1025	struct sctp_nets *net;
1026
1027	if ((to == NULL) || (from == NULL)) {
1028		return (NULL);
1029	}
1030	switch (to->sa_family) {
1031#ifdef INET
1032	case AF_INET:
1033		if (from->sa_family == AF_INET) {
1034			lport = ((struct sockaddr_in *)to)->sin_port;
1035			rport = ((struct sockaddr_in *)from)->sin_port;
1036		} else {
1037			return (NULL);
1038		}
1039		break;
1040#endif
1041#ifdef INET6
1042	case AF_INET6:
1043		if (from->sa_family == AF_INET6) {
1044			lport = ((struct sockaddr_in6 *)to)->sin6_port;
1045			rport = ((struct sockaddr_in6 *)from)->sin6_port;
1046		} else {
1047			return (NULL);
1048		}
1049		break;
1050#endif
1051	default:
1052		return (NULL);
1053	}
1054	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
1055	/*
1056	 * Ok now for each of the guys in this bucket we must look and see:
1057	 * - Does the remote port match. - Does there single association's
1058	 * addresses match this address (to). If so we update p_ep to point
1059	 * to this ep and return the tcb from it.
1060	 */
1061	LIST_FOREACH(inp, ephead, sctp_hash) {
1062		SCTP_INP_RLOCK(inp);
1063		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1064			SCTP_INP_RUNLOCK(inp);
1065			continue;
1066		}
1067		if (lport != inp->sctp_lport) {
1068			SCTP_INP_RUNLOCK(inp);
1069			continue;
1070		}
1071		switch (to->sa_family) {
1072#ifdef INET
1073		case AF_INET:
1074			{
1075				struct sockaddr_in *sin;
1076
1077				sin = (struct sockaddr_in *)to;
1078				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1079				    &sin->sin_addr) != 0) {
1080					SCTP_INP_RUNLOCK(inp);
1081					continue;
1082				}
1083				break;
1084			}
1085#endif
1086#ifdef INET6
1087		case AF_INET6:
1088			{
1089				struct sockaddr_in6 *sin6;
1090
1091				sin6 = (struct sockaddr_in6 *)to;
1092				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1093				    &sin6->sin6_addr) != 0) {
1094					SCTP_INP_RUNLOCK(inp);
1095					continue;
1096				}
1097				break;
1098			}
1099#endif
1100		default:
1101			SCTP_INP_RUNLOCK(inp);
1102			continue;
1103		}
1104		if (inp->def_vrf_id != vrf_id) {
1105			SCTP_INP_RUNLOCK(inp);
1106			continue;
1107		}
1108		/* check to see if the ep has one of the addresses */
1109		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1110			/* We are NOT bound all, so look further */
1111			int match = 0;
1112
1113			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1114
1115				if (laddr->ifa == NULL) {
1116					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
1117					continue;
1118				}
1119				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1120					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1121					continue;
1122				}
1123				if (laddr->ifa->address.sa.sa_family ==
1124				    to->sa_family) {
1125					/* see if it matches */
1126#ifdef INET
1127					if (from->sa_family == AF_INET) {
1128						struct sockaddr_in *intf_addr,
1129						           *sin;
1130
1131						intf_addr = &laddr->ifa->address.sin;
1132						sin = (struct sockaddr_in *)to;
1133						if (sin->sin_addr.s_addr ==
1134						    intf_addr->sin_addr.s_addr) {
1135							match = 1;
1136							break;
1137						}
1138					}
1139#endif
1140#ifdef INET6
1141					if (from->sa_family == AF_INET6) {
1142						struct sockaddr_in6 *intf_addr6;
1143						struct sockaddr_in6 *sin6;
1144
1145						sin6 = (struct sockaddr_in6 *)
1146						    to;
1147						intf_addr6 = &laddr->ifa->address.sin6;
1148
1149						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1150						    intf_addr6)) {
1151							match = 1;
1152							break;
1153						}
1154					}
1155#endif
1156				}
1157			}
1158			if (match == 0) {
1159				/* This endpoint does not have this address */
1160				SCTP_INP_RUNLOCK(inp);
1161				continue;
1162			}
1163		}
1164		/*
1165		 * Ok if we hit here the ep has the address, does it hold
1166		 * the tcb?
1167		 */
1168		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1169		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1170		if (stcb == NULL) {
1171			SCTP_INP_RUNLOCK(inp);
1172			continue;
1173		}
1174		SCTP_TCB_LOCK(stcb);
1175		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1176			SCTP_TCB_UNLOCK(stcb);
1177			SCTP_INP_RUNLOCK(inp);
1178			continue;
1179		}
1180		if (stcb->rport != rport) {
1181			/* remote port does not match. */
1182			SCTP_TCB_UNLOCK(stcb);
1183			SCTP_INP_RUNLOCK(inp);
1184			continue;
1185		}
1186		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1187			SCTP_TCB_UNLOCK(stcb);
1188			SCTP_INP_RUNLOCK(inp);
1189			continue;
1190		}
1191		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1192			SCTP_TCB_UNLOCK(stcb);
1193			SCTP_INP_RUNLOCK(inp);
1194			continue;
1195		}
1196		/* Does this TCB have a matching address? */
1197		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1198
1199			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1200				/* not the same family, can't be a match */
1201				continue;
1202			}
1203			switch (from->sa_family) {
1204#ifdef INET
1205			case AF_INET:
1206				{
1207					struct sockaddr_in *sin, *rsin;
1208
1209					sin = (struct sockaddr_in *)&net->ro._l_addr;
1210					rsin = (struct sockaddr_in *)from;
1211					if (sin->sin_addr.s_addr ==
1212					    rsin->sin_addr.s_addr) {
1213						/* found it */
1214						if (netp != NULL) {
1215							*netp = net;
1216						}
1217						/*
1218						 * Update the endpoint
1219						 * pointer
1220						 */
1221						*inp_p = inp;
1222						SCTP_INP_RUNLOCK(inp);
1223						return (stcb);
1224					}
1225					break;
1226				}
1227#endif
1228#ifdef INET6
1229			case AF_INET6:
1230				{
1231					struct sockaddr_in6 *sin6, *rsin6;
1232
1233					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1234					rsin6 = (struct sockaddr_in6 *)from;
1235					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1236					    rsin6)) {
1237						/* found it */
1238						if (netp != NULL) {
1239							*netp = net;
1240						}
1241						/*
1242						 * Update the endpoint
1243						 * pointer
1244						 */
1245						*inp_p = inp;
1246						SCTP_INP_RUNLOCK(inp);
1247						return (stcb);
1248					}
1249					break;
1250				}
1251#endif
1252			default:
1253				/* TSNH */
1254				break;
1255			}
1256		}
1257		SCTP_TCB_UNLOCK(stcb);
1258		SCTP_INP_RUNLOCK(inp);
1259	}
1260	return (NULL);
1261}
1262
1263
1264/*
1265 * rules for use
1266 *
1267 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1268 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1269 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1270 * NULL.
1271 */
1272
1273struct sctp_tcb *
1274sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1275    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1276{
1277	struct sctpasochead *head;
1278	struct sctp_inpcb *inp;
1279	struct sctp_tcb *stcb = NULL;
1280	struct sctp_nets *net;
1281	uint16_t rport;
1282
1283	inp = *inp_p;
1284	switch (remote->sa_family) {
1285#ifdef INET
1286	case AF_INET:
1287		rport = (((struct sockaddr_in *)remote)->sin_port);
1288		break;
1289#endif
1290#ifdef INET6
1291	case AF_INET6:
1292		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1293		break;
1294#endif
1295	default:
1296		return (NULL);
1297	}
1298	if (locked_tcb) {
1299		/*
1300		 * UN-lock so we can do proper locking here this occurs when
1301		 * called from load_addresses_from_init.
1302		 */
1303		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1304		SCTP_TCB_UNLOCK(locked_tcb);
1305	}
1306	SCTP_INP_INFO_RLOCK();
1307	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1308	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1309		/*-
1310		 * Now either this guy is our listener or it's the
1311		 * connector. If it is the one that issued the connect, then
1312		 * it's only chance is to be the first TCB in the list. If
1313		 * it is the acceptor, then do the special_lookup to hash
1314		 * and find the real inp.
1315		 */
1316		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1317			/* to is peer addr, from is my addr */
1318			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1319			    netp, inp->def_vrf_id);
1320			if ((stcb != NULL) && (locked_tcb == NULL)) {
1321				/* we have a locked tcb, lower refcount */
1322				SCTP_INP_DECR_REF(inp);
1323			}
1324			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1325				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1326				SCTP_TCB_LOCK(locked_tcb);
1327				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1328				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1329			}
1330			SCTP_INP_INFO_RUNLOCK();
1331			return (stcb);
1332		} else {
1333			SCTP_INP_WLOCK(inp);
1334			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1335				goto null_return;
1336			}
1337			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1338			if (stcb == NULL) {
1339				goto null_return;
1340			}
1341			SCTP_TCB_LOCK(stcb);
1342
1343			if (stcb->rport != rport) {
1344				/* remote port does not match. */
1345				SCTP_TCB_UNLOCK(stcb);
1346				goto null_return;
1347			}
1348			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1349				SCTP_TCB_UNLOCK(stcb);
1350				goto null_return;
1351			}
1352			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1353				SCTP_TCB_UNLOCK(stcb);
1354				goto null_return;
1355			}
1356			/* now look at the list of remote addresses */
1357			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1358#ifdef INVARIANTS
1359				if (net == (TAILQ_NEXT(net, sctp_next))) {
1360					panic("Corrupt net list");
1361				}
1362#endif
1363				if (net->ro._l_addr.sa.sa_family !=
1364				    remote->sa_family) {
1365					/* not the same family */
1366					continue;
1367				}
1368				switch (remote->sa_family) {
1369#ifdef INET
1370				case AF_INET:
1371					{
1372						struct sockaddr_in *sin,
1373						           *rsin;
1374
1375						sin = (struct sockaddr_in *)
1376						    &net->ro._l_addr;
1377						rsin = (struct sockaddr_in *)remote;
1378						if (sin->sin_addr.s_addr ==
1379						    rsin->sin_addr.s_addr) {
1380							/* found it */
1381							if (netp != NULL) {
1382								*netp = net;
1383							}
1384							if (locked_tcb == NULL) {
1385								SCTP_INP_DECR_REF(inp);
1386							} else if (locked_tcb != stcb) {
1387								SCTP_TCB_LOCK(locked_tcb);
1388							}
1389							if (locked_tcb) {
1390								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1391							}
1392							SCTP_INP_WUNLOCK(inp);
1393							SCTP_INP_INFO_RUNLOCK();
1394							return (stcb);
1395						}
1396						break;
1397					}
1398#endif
1399#ifdef INET6
1400				case AF_INET6:
1401					{
1402						struct sockaddr_in6 *sin6,
1403						            *rsin6;
1404
1405						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1406						rsin6 = (struct sockaddr_in6 *)remote;
1407						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1408						    rsin6)) {
1409							/* found it */
1410							if (netp != NULL) {
1411								*netp = net;
1412							}
1413							if (locked_tcb == NULL) {
1414								SCTP_INP_DECR_REF(inp);
1415							} else if (locked_tcb != stcb) {
1416								SCTP_TCB_LOCK(locked_tcb);
1417							}
1418							if (locked_tcb) {
1419								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1420							}
1421							SCTP_INP_WUNLOCK(inp);
1422							SCTP_INP_INFO_RUNLOCK();
1423							return (stcb);
1424						}
1425						break;
1426					}
1427#endif
1428				default:
1429					/* TSNH */
1430					break;
1431				}
1432			}
1433			SCTP_TCB_UNLOCK(stcb);
1434		}
1435	} else {
1436		SCTP_INP_WLOCK(inp);
1437		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1438			goto null_return;
1439		}
1440		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1441		    inp->sctp_hashmark)];
1442		if (head == NULL) {
1443			goto null_return;
1444		}
1445		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1446			if (stcb->rport != rport) {
1447				/* remote port does not match */
1448				continue;
1449			}
1450			SCTP_TCB_LOCK(stcb);
1451			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1452				SCTP_TCB_UNLOCK(stcb);
1453				continue;
1454			}
1455			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1456				SCTP_TCB_UNLOCK(stcb);
1457				continue;
1458			}
1459			/* now look at the list of remote addresses */
1460			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1461#ifdef INVARIANTS
1462				if (net == (TAILQ_NEXT(net, sctp_next))) {
1463					panic("Corrupt net list");
1464				}
1465#endif
1466				if (net->ro._l_addr.sa.sa_family !=
1467				    remote->sa_family) {
1468					/* not the same family */
1469					continue;
1470				}
1471				switch (remote->sa_family) {
1472#ifdef INET
1473				case AF_INET:
1474					{
1475						struct sockaddr_in *sin,
1476						           *rsin;
1477
1478						sin = (struct sockaddr_in *)
1479						    &net->ro._l_addr;
1480						rsin = (struct sockaddr_in *)remote;
1481						if (sin->sin_addr.s_addr ==
1482						    rsin->sin_addr.s_addr) {
1483							/* found it */
1484							if (netp != NULL) {
1485								*netp = net;
1486							}
1487							if (locked_tcb == NULL) {
1488								SCTP_INP_DECR_REF(inp);
1489							} else if (locked_tcb != stcb) {
1490								SCTP_TCB_LOCK(locked_tcb);
1491							}
1492							if (locked_tcb) {
1493								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1494							}
1495							SCTP_INP_WUNLOCK(inp);
1496							SCTP_INP_INFO_RUNLOCK();
1497							return (stcb);
1498						}
1499						break;
1500					}
1501#endif
1502#ifdef INET6
1503				case AF_INET6:
1504					{
1505						struct sockaddr_in6 *sin6,
1506						            *rsin6;
1507
1508						sin6 = (struct sockaddr_in6 *)
1509						    &net->ro._l_addr;
1510						rsin6 = (struct sockaddr_in6 *)remote;
1511						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1512						    rsin6)) {
1513							/* found it */
1514							if (netp != NULL) {
1515								*netp = net;
1516							}
1517							if (locked_tcb == NULL) {
1518								SCTP_INP_DECR_REF(inp);
1519							} else if (locked_tcb != stcb) {
1520								SCTP_TCB_LOCK(locked_tcb);
1521							}
1522							if (locked_tcb) {
1523								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1524							}
1525							SCTP_INP_WUNLOCK(inp);
1526							SCTP_INP_INFO_RUNLOCK();
1527							return (stcb);
1528						}
1529						break;
1530					}
1531#endif
1532				default:
1533					/* TSNH */
1534					break;
1535				}
1536			}
1537			SCTP_TCB_UNLOCK(stcb);
1538		}
1539	}
1540null_return:
1541	/* clean up for returning null */
1542	if (locked_tcb) {
1543		SCTP_TCB_LOCK(locked_tcb);
1544		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1545	}
1546	SCTP_INP_WUNLOCK(inp);
1547	SCTP_INP_INFO_RUNLOCK();
1548	/* not found */
1549	return (NULL);
1550}
1551
1552
1553/*
1554 * Find an association for a specific endpoint using the association id given
1555 * out in the COMM_UP notification
1556 */
1557struct sctp_tcb *
1558sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1559{
1560	/*
1561	 * Use my the assoc_id to find a endpoint
1562	 */
1563	struct sctpasochead *head;
1564	struct sctp_tcb *stcb;
1565	uint32_t id;
1566
1567	if (inp == NULL) {
1568		SCTP_PRINTF("TSNH ep_associd\n");
1569		return (NULL);
1570	}
1571	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1572		SCTP_PRINTF("TSNH ep_associd0\n");
1573		return (NULL);
1574	}
1575	id = (uint32_t) asoc_id;
1576	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1577	if (head == NULL) {
1578		/* invalid id TSNH */
1579		SCTP_PRINTF("TSNH ep_associd1\n");
1580		return (NULL);
1581	}
1582	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1583		if (stcb->asoc.assoc_id == id) {
1584			if (inp != stcb->sctp_ep) {
1585				/*
1586				 * some other guy has the same id active (id
1587				 * collision ??).
1588				 */
1589				SCTP_PRINTF("TSNH ep_associd2\n");
1590				continue;
1591			}
1592			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1593				continue;
1594			}
1595			if (want_lock) {
1596				SCTP_TCB_LOCK(stcb);
1597			}
1598			return (stcb);
1599		}
1600	}
1601	return (NULL);
1602}
1603
1604
1605struct sctp_tcb *
1606sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1607{
1608	struct sctp_tcb *stcb;
1609
1610	SCTP_INP_RLOCK(inp);
1611	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1612	SCTP_INP_RUNLOCK(inp);
1613	return (stcb);
1614}
1615
1616
1617/*
1618 * Endpoint probe expects that the INP_INFO is locked.
1619 */
1620static struct sctp_inpcb *
1621sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1622    uint16_t lport, uint32_t vrf_id)
1623{
1624	struct sctp_inpcb *inp;
1625	struct sctp_laddr *laddr;
1626
1627#ifdef INET
1628	struct sockaddr_in *sin;
1629
1630#endif
1631#ifdef INET6
1632	struct sockaddr_in6 *sin6;
1633	struct sockaddr_in6 *intf_addr6;
1634
1635#endif
1636	int fnd;
1637
1638#ifdef INET
1639	sin = NULL;
1640#endif
1641#ifdef INET6
1642	sin6 = NULL;
1643#endif
1644	switch (nam->sa_family) {
1645#ifdef INET
1646	case AF_INET:
1647		sin = (struct sockaddr_in *)nam;
1648		break;
1649#endif
1650#ifdef INET6
1651	case AF_INET6:
1652		sin6 = (struct sockaddr_in6 *)nam;
1653		break;
1654#endif
1655	default:
1656		/* unsupported family */
1657		return (NULL);
1658	}
1659
1660	if (head == NULL)
1661		return (NULL);
1662
1663	LIST_FOREACH(inp, head, sctp_hash) {
1664		SCTP_INP_RLOCK(inp);
1665		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1666			SCTP_INP_RUNLOCK(inp);
1667			continue;
1668		}
1669		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1670		    (inp->sctp_lport == lport)) {
1671			/* got it */
1672			switch (nam->sa_family) {
1673#ifdef INET
1674			case AF_INET:
1675				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1676				    SCTP_IPV6_V6ONLY(inp)) {
1677					/*
1678					 * IPv4 on a IPv6 socket with ONLY
1679					 * IPv6 set
1680					 */
1681					SCTP_INP_RUNLOCK(inp);
1682					continue;
1683				}
1684				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1685				    &sin->sin_addr) != 0) {
1686					SCTP_INP_RUNLOCK(inp);
1687					continue;
1688				}
1689				break;
1690#endif
1691#ifdef INET6
1692			case AF_INET6:
1693				/*
1694				 * A V6 address and the endpoint is NOT
1695				 * bound V6
1696				 */
1697				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1698					SCTP_INP_RUNLOCK(inp);
1699					continue;
1700				}
1701				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1702				    &sin6->sin6_addr) != 0) {
1703					SCTP_INP_RUNLOCK(inp);
1704					continue;
1705				}
1706				break;
1707#endif
1708			default:
1709				break;
1710			}
1711			/* does a VRF id match? */
1712			fnd = 0;
1713			if (inp->def_vrf_id == vrf_id)
1714				fnd = 1;
1715
1716			SCTP_INP_RUNLOCK(inp);
1717			if (!fnd)
1718				continue;
1719			return (inp);
1720		}
1721		SCTP_INP_RUNLOCK(inp);
1722	}
1723	switch (nam->sa_family) {
1724#ifdef INET
1725	case AF_INET:
1726		if (sin->sin_addr.s_addr == INADDR_ANY) {
1727			/* Can't hunt for one that has no address specified */
1728			return (NULL);
1729		}
1730		break;
1731#endif
1732#ifdef INET6
1733	case AF_INET6:
1734		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1735			/* Can't hunt for one that has no address specified */
1736			return (NULL);
1737		}
1738		break;
1739#endif
1740	default:
1741		break;
1742	}
1743	/*
1744	 * ok, not bound to all so see if we can find a EP bound to this
1745	 * address.
1746	 */
1747	LIST_FOREACH(inp, head, sctp_hash) {
1748		SCTP_INP_RLOCK(inp);
1749		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1750			SCTP_INP_RUNLOCK(inp);
1751			continue;
1752		}
1753		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1754			SCTP_INP_RUNLOCK(inp);
1755			continue;
1756		}
1757		/*
1758		 * Ok this could be a likely candidate, look at all of its
1759		 * addresses
1760		 */
1761		if (inp->sctp_lport != lport) {
1762			SCTP_INP_RUNLOCK(inp);
1763			continue;
1764		}
1765		/* does a VRF id match? */
1766		fnd = 0;
1767		if (inp->def_vrf_id == vrf_id)
1768			fnd = 1;
1769
1770		if (!fnd) {
1771			SCTP_INP_RUNLOCK(inp);
1772			continue;
1773		}
1774		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1775			if (laddr->ifa == NULL) {
1776				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1777				    __FUNCTION__);
1778				continue;
1779			}
1780			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1781			    (void *)laddr->ifa);
1782			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1783				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1784				continue;
1785			}
1786			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1787				/* possible, see if it matches */
1788				switch (nam->sa_family) {
1789#ifdef INET
1790				case AF_INET:
1791					if (sin->sin_addr.s_addr ==
1792					    laddr->ifa->address.sin.sin_addr.s_addr) {
1793						SCTP_INP_RUNLOCK(inp);
1794						return (inp);
1795					}
1796					break;
1797#endif
1798#ifdef INET6
1799				case AF_INET6:
1800					intf_addr6 = &laddr->ifa->address.sin6;
1801					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1802					    intf_addr6)) {
1803						SCTP_INP_RUNLOCK(inp);
1804						return (inp);
1805					}
1806					break;
1807#endif
1808				}
1809			}
1810		}
1811		SCTP_INP_RUNLOCK(inp);
1812	}
1813	return (NULL);
1814}
1815
1816
1817static struct sctp_inpcb *
1818sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1819{
1820	struct sctppcbhead *head;
1821	struct sctp_inpcb *t_inp;
1822	int fnd;
1823
1824	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1825	    SCTP_BASE_INFO(hashmark))];
1826	LIST_FOREACH(t_inp, head, sctp_hash) {
1827		if (t_inp->sctp_lport != lport) {
1828			continue;
1829		}
1830		/* is it in the VRF in question */
1831		fnd = 0;
1832		if (t_inp->def_vrf_id == vrf_id)
1833			fnd = 1;
1834		if (!fnd)
1835			continue;
1836
1837		/* This one is in use. */
1838		/* check the v6/v4 binding issue */
1839		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1840		    SCTP_IPV6_V6ONLY(t_inp)) {
1841			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1842				/* collision in V6 space */
1843				return (t_inp);
1844			} else {
1845				/* inp is BOUND_V4 no conflict */
1846				continue;
1847			}
1848		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1849			/* t_inp is bound v4 and v6, conflict always */
1850			return (t_inp);
1851		} else {
1852			/* t_inp is bound only V4 */
1853			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1854			    SCTP_IPV6_V6ONLY(inp)) {
1855				/* no conflict */
1856				continue;
1857			}
1858			/* else fall through to conflict */
1859		}
1860		return (t_inp);
1861	}
1862	return (NULL);
1863}
1864
1865
1866int
1867sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1868{
1869	/* For 1-2-1 with port reuse */
1870	struct sctppcbhead *head;
1871	struct sctp_inpcb *tinp;
1872
1873	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1874		/* only works with port reuse on */
1875		return (-1);
1876	}
1877	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1878		return (0);
1879	}
1880	SCTP_INP_RUNLOCK(inp);
1881	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1882	    SCTP_BASE_INFO(hashmark))];
1883	/* Kick out all non-listeners to the TCP hash */
1884	LIST_FOREACH(tinp, head, sctp_hash) {
1885		if (tinp->sctp_lport != inp->sctp_lport) {
1886			continue;
1887		}
1888		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1889			continue;
1890		}
1891		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1892			continue;
1893		}
1894		if (tinp->sctp_socket->so_qlimit) {
1895			continue;
1896		}
1897		SCTP_INP_WLOCK(tinp);
1898		LIST_REMOVE(tinp, sctp_hash);
1899		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1900		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1901		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1902		SCTP_INP_WUNLOCK(tinp);
1903	}
1904	SCTP_INP_WLOCK(inp);
1905	/* Pull from where he was */
1906	LIST_REMOVE(inp, sctp_hash);
1907	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1908	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1909	LIST_INSERT_HEAD(head, inp, sctp_hash);
1910	SCTP_INP_WUNLOCK(inp);
1911	SCTP_INP_RLOCK(inp);
1912	return (0);
1913}
1914
1915
1916struct sctp_inpcb *
1917sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1918    uint32_t vrf_id)
1919{
1920	/*
1921	 * First we check the hash table to see if someone has this port
1922	 * bound with just the port.
1923	 */
1924	struct sctp_inpcb *inp;
1925	struct sctppcbhead *head;
1926	int lport;
1927	unsigned int i;
1928
1929#ifdef INET
1930	struct sockaddr_in *sin;
1931
1932#endif
1933#ifdef INET6
1934	struct sockaddr_in6 *sin6;
1935
1936#endif
1937
1938	switch (nam->sa_family) {
1939#ifdef INET
1940	case AF_INET:
1941		sin = (struct sockaddr_in *)nam;
1942		lport = sin->sin_port;
1943		break;
1944#endif
1945#ifdef INET6
1946	case AF_INET6:
1947		sin6 = (struct sockaddr_in6 *)nam;
1948		lport = sin6->sin6_port;
1949		break;
1950#endif
1951	default:
1952		return (NULL);
1953	}
1954	/*
1955	 * I could cheat here and just cast to one of the types but we will
1956	 * do it right. It also provides the check against an Unsupported
1957	 * type too.
1958	 */
1959	/* Find the head of the ALLADDR chain */
1960	if (have_lock == 0) {
1961		SCTP_INP_INFO_RLOCK();
1962	}
1963	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1964	    SCTP_BASE_INFO(hashmark))];
1965	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1966
1967	/*
1968	 * If the TCP model exists it could be that the main listening
1969	 * endpoint is gone but there still exists a connected socket for
1970	 * this guy. If so we can return the first one that we find. This
1971	 * may NOT be the correct one so the caller should be wary on the
1972	 * returned INP. Currently the only caller that sets find_tcp_pool
1973	 * is in bindx where we are verifying that a user CAN bind the
1974	 * address. He either has bound it already, or someone else has, or
1975	 * its open to bind, so this is good enough.
1976	 */
1977	if (inp == NULL && find_tcp_pool) {
1978		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1979			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1980			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1981			if (inp) {
1982				break;
1983			}
1984		}
1985	}
1986	if (inp) {
1987		SCTP_INP_INCR_REF(inp);
1988	}
1989	if (have_lock == 0) {
1990		SCTP_INP_INFO_RUNLOCK();
1991	}
1992	return (inp);
1993}
1994
1995
1996/*
1997 * Find an association for an endpoint with the pointer to whom you want to
1998 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1999 * need to change the *to to some other struct like a mbuf...
2000 */
2001struct sctp_tcb *
2002sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
2003    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
2004    uint32_t vrf_id)
2005{
2006	struct sctp_inpcb *inp = NULL;
2007	struct sctp_tcb *stcb;
2008
2009	SCTP_INP_INFO_RLOCK();
2010	if (find_tcp_pool) {
2011		if (inp_p != NULL) {
2012			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
2013			    vrf_id);
2014		} else {
2015			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
2016			    vrf_id);
2017		}
2018		if (stcb != NULL) {
2019			SCTP_INP_INFO_RUNLOCK();
2020			return (stcb);
2021		}
2022	}
2023	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
2024	if (inp_p != NULL) {
2025		*inp_p = inp;
2026	}
2027	SCTP_INP_INFO_RUNLOCK();
2028	if (inp == NULL) {
2029		return (NULL);
2030	}
2031	/*
2032	 * ok, we have an endpoint, now lets find the assoc for it (if any)
2033	 * we now place the source address or from in the to of the find
2034	 * endpoint call. Since in reality this chain is used from the
2035	 * inbound packet side.
2036	 */
2037	if (inp_p != NULL) {
2038		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
2039		    NULL);
2040	} else {
2041		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
2042		    NULL);
2043	}
2044	return (stcb);
2045}
2046
2047
2048/*
2049 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
2050 * find all addresses that the sender has specified in any address list. Each
2051 * address will be used to lookup the TCB and see if one exits.
2052 */
2053static struct sctp_tcb *
2054sctp_findassociation_special_addr(struct mbuf *m, int offset,
2055    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
2056    struct sockaddr *dst)
2057{
2058	struct sctp_paramhdr *phdr, parm_buf;
2059
2060#if defined(INET) || defined(INET6)
2061	struct sctp_tcb *stcb;
2062	uint16_t ptype;
2063
2064#endif
2065	uint16_t plen;
2066
2067#ifdef INET
2068	struct sockaddr_in sin4;
2069
2070#endif
2071#ifdef INET6
2072	struct sockaddr_in6 sin6;
2073
2074#endif
2075
2076#ifdef INET
2077	memset(&sin4, 0, sizeof(sin4));
2078	sin4.sin_len = sizeof(sin4);
2079	sin4.sin_family = AF_INET;
2080	sin4.sin_port = sh->src_port;
2081#endif
2082#ifdef INET6
2083	memset(&sin6, 0, sizeof(sin6));
2084	sin6.sin6_len = sizeof(sin6);
2085	sin6.sin6_family = AF_INET6;
2086	sin6.sin6_port = sh->src_port;
2087#endif
2088
2089	offset += sizeof(struct sctp_init_chunk);
2090
2091	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2092	while (phdr != NULL) {
2093		/* now we must see if we want the parameter */
2094#if defined(INET) || defined(INET6)
2095		ptype = ntohs(phdr->param_type);
2096#endif
2097		plen = ntohs(phdr->param_length);
2098		if (plen == 0) {
2099			break;
2100		}
2101#ifdef INET
2102		if (ptype == SCTP_IPV4_ADDRESS &&
2103		    plen == sizeof(struct sctp_ipv4addr_param)) {
2104			/* Get the rest of the address */
2105			struct sctp_ipv4addr_param ip4_parm, *p4;
2106
2107			phdr = sctp_get_next_param(m, offset,
2108			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2109			if (phdr == NULL) {
2110				return (NULL);
2111			}
2112			p4 = (struct sctp_ipv4addr_param *)phdr;
2113			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2114			/* look it up */
2115			stcb = sctp_findassociation_ep_addr(inp_p,
2116			    (struct sockaddr *)&sin4, netp, dst, NULL);
2117			if (stcb != NULL) {
2118				return (stcb);
2119			}
2120		}
2121#endif
2122#ifdef INET6
2123		if (ptype == SCTP_IPV6_ADDRESS &&
2124		    plen == sizeof(struct sctp_ipv6addr_param)) {
2125			/* Get the rest of the address */
2126			struct sctp_ipv6addr_param ip6_parm, *p6;
2127
2128			phdr = sctp_get_next_param(m, offset,
2129			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2130			if (phdr == NULL) {
2131				return (NULL);
2132			}
2133			p6 = (struct sctp_ipv6addr_param *)phdr;
2134			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2135			/* look it up */
2136			stcb = sctp_findassociation_ep_addr(inp_p,
2137			    (struct sockaddr *)&sin6, netp, dst, NULL);
2138			if (stcb != NULL) {
2139				return (stcb);
2140			}
2141		}
2142#endif
2143		offset += SCTP_SIZE32(plen);
2144		phdr = sctp_get_next_param(m, offset, &parm_buf,
2145		    sizeof(parm_buf));
2146	}
2147	return (NULL);
2148}
2149
2150static struct sctp_tcb *
2151sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2152    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2153    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2154{
2155	/*
2156	 * Use my vtag to hash. If we find it we then verify the source addr
2157	 * is in the assoc. If all goes well we save a bit on rec of a
2158	 * packet.
2159	 */
2160	struct sctpasochead *head;
2161	struct sctp_nets *net;
2162	struct sctp_tcb *stcb;
2163
2164	SCTP_INP_INFO_RLOCK();
2165	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2166	    SCTP_BASE_INFO(hashasocmark))];
2167	if (head == NULL) {
2168		/* invalid vtag */
2169		SCTP_INP_INFO_RUNLOCK();
2170		return (NULL);
2171	}
2172	LIST_FOREACH(stcb, head, sctp_asocs) {
2173		SCTP_INP_RLOCK(stcb->sctp_ep);
2174		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2175			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2176			continue;
2177		}
2178		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2179			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2180			continue;
2181		}
2182		SCTP_TCB_LOCK(stcb);
2183		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2184		if (stcb->asoc.my_vtag == vtag) {
2185			/* candidate */
2186			if (stcb->rport != rport) {
2187				SCTP_TCB_UNLOCK(stcb);
2188				continue;
2189			}
2190			if (stcb->sctp_ep->sctp_lport != lport) {
2191				SCTP_TCB_UNLOCK(stcb);
2192				continue;
2193			}
2194			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2195				SCTP_TCB_UNLOCK(stcb);
2196				continue;
2197			}
2198			/* RRS:Need toaddr check here */
2199			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2200				/* Endpoint does not own this address */
2201				SCTP_TCB_UNLOCK(stcb);
2202				continue;
2203			}
2204			if (remote_tag) {
2205				/*
2206				 * If we have both vtags that's all we match
2207				 * on
2208				 */
2209				if (stcb->asoc.peer_vtag == remote_tag) {
2210					/*
2211					 * If both tags match we consider it
2212					 * conclusive and check NO
2213					 * source/destination addresses
2214					 */
2215					goto conclusive;
2216				}
2217			}
2218			if (skip_src_check) {
2219		conclusive:
2220				if (from) {
2221					*netp = sctp_findnet(stcb, from);
2222				} else {
2223					*netp = NULL;	/* unknown */
2224				}
2225				if (inp_p)
2226					*inp_p = stcb->sctp_ep;
2227				SCTP_INP_INFO_RUNLOCK();
2228				return (stcb);
2229			}
2230			net = sctp_findnet(stcb, from);
2231			if (net) {
2232				/* yep its him. */
2233				*netp = net;
2234				SCTP_STAT_INCR(sctps_vtagexpress);
2235				*inp_p = stcb->sctp_ep;
2236				SCTP_INP_INFO_RUNLOCK();
2237				return (stcb);
2238			} else {
2239				/*
2240				 * not him, this should only happen in rare
2241				 * cases so I peg it.
2242				 */
2243				SCTP_STAT_INCR(sctps_vtagbogus);
2244			}
2245		}
2246		SCTP_TCB_UNLOCK(stcb);
2247	}
2248	SCTP_INP_INFO_RUNLOCK();
2249	return (NULL);
2250}
2251
2252
2253/*
2254 * Find an association with the pointer to the inbound IP packet. This can be
2255 * a IPv4 or IPv6 packet.
2256 */
2257struct sctp_tcb *
2258sctp_findassociation_addr(struct mbuf *m, int offset,
2259    struct sockaddr *src, struct sockaddr *dst,
2260    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2261    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2262{
2263	int find_tcp_pool;
2264	struct sctp_tcb *stcb;
2265	struct sctp_inpcb *inp;
2266
2267	if (sh->v_tag) {
2268		/* we only go down this path if vtag is non-zero */
2269		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2270		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2271		if (stcb) {
2272			return (stcb);
2273		}
2274	}
2275	find_tcp_pool = 0;
2276	if ((ch->chunk_type != SCTP_INITIATION) &&
2277	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2278	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2279	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2280		/* Other chunk types go to the tcp pool. */
2281		find_tcp_pool = 1;
2282	}
2283	if (inp_p) {
2284		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2285		    find_tcp_pool, vrf_id);
2286		inp = *inp_p;
2287	} else {
2288		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2289		    find_tcp_pool, vrf_id);
2290	}
2291	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2292	if (stcb == NULL && inp) {
2293		/* Found a EP but not this address */
2294		if ((ch->chunk_type == SCTP_INITIATION) ||
2295		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2296			/*-
2297			 * special hook, we do NOT return linp or an
2298			 * association that is linked to an existing
2299			 * association that is under the TCP pool (i.e. no
2300			 * listener exists). The endpoint finding routine
2301			 * will always find a listener before examining the
2302			 * TCP pool.
2303			 */
2304			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2305				if (inp_p) {
2306					*inp_p = NULL;
2307				}
2308				return (NULL);
2309			}
2310			stcb = sctp_findassociation_special_addr(m,
2311			    offset, sh, &inp, netp, dst);
2312			if (inp_p != NULL) {
2313				*inp_p = inp;
2314			}
2315		}
2316	}
2317	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2318	return (stcb);
2319}
2320
2321/*
2322 * lookup an association by an ASCONF lookup address.
2323 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2324 */
2325struct sctp_tcb *
2326sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2327    struct sockaddr *dst, struct sctphdr *sh,
2328    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2329{
2330	struct sctp_tcb *stcb;
2331	struct sockaddr_storage remote_store;
2332	struct sctp_paramhdr parm_buf, *phdr;
2333	int ptype;
2334	int zero_address = 0;
2335
2336#ifdef INET
2337	struct sockaddr_in *sin;
2338
2339#endif
2340#ifdef INET6
2341	struct sockaddr_in6 *sin6;
2342
2343#endif
2344
2345	memset(&remote_store, 0, sizeof(remote_store));
2346	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2347	    &parm_buf, sizeof(struct sctp_paramhdr));
2348	if (phdr == NULL) {
2349		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2350		    __FUNCTION__);
2351		return NULL;
2352	}
2353	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2354	/* get the correlation address */
2355	switch (ptype) {
2356#ifdef INET6
2357	case SCTP_IPV6_ADDRESS:
2358		{
2359			/* ipv6 address param */
2360			struct sctp_ipv6addr_param *p6, p6_buf;
2361
2362			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2363				return NULL;
2364			}
2365			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2366			    offset + sizeof(struct sctp_asconf_chunk),
2367			    &p6_buf.ph, sizeof(*p6));
2368			if (p6 == NULL) {
2369				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2370				    __FUNCTION__);
2371				return (NULL);
2372			}
2373			sin6 = (struct sockaddr_in6 *)&remote_store;
2374			sin6->sin6_family = AF_INET6;
2375			sin6->sin6_len = sizeof(*sin6);
2376			sin6->sin6_port = sh->src_port;
2377			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2378			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2379				zero_address = 1;
2380			break;
2381		}
2382#endif
2383#ifdef INET
2384	case SCTP_IPV4_ADDRESS:
2385		{
2386			/* ipv4 address param */
2387			struct sctp_ipv4addr_param *p4, p4_buf;
2388
2389			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2390				return NULL;
2391			}
2392			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2393			    offset + sizeof(struct sctp_asconf_chunk),
2394			    &p4_buf.ph, sizeof(*p4));
2395			if (p4 == NULL) {
2396				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2397				    __FUNCTION__);
2398				return (NULL);
2399			}
2400			sin = (struct sockaddr_in *)&remote_store;
2401			sin->sin_family = AF_INET;
2402			sin->sin_len = sizeof(*sin);
2403			sin->sin_port = sh->src_port;
2404			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2405			if (sin->sin_addr.s_addr == INADDR_ANY)
2406				zero_address = 1;
2407			break;
2408		}
2409#endif
2410	default:
2411		/* invalid address param type */
2412		return NULL;
2413	}
2414
2415	if (zero_address) {
2416		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2417		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2418		if (stcb != NULL) {
2419			SCTP_INP_DECR_REF(*inp_p);
2420		}
2421	} else {
2422		stcb = sctp_findassociation_ep_addr(inp_p,
2423		    (struct sockaddr *)&remote_store, netp,
2424		    dst, NULL);
2425	}
2426	return (stcb);
2427}
2428
2429
2430/*
2431 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2432 * addresses. This way if we don't get a bind we by default pick a ephemeral
2433 * port with all addresses bound.
2434 */
2435int
2436sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2437{
2438	/*
2439	 * we get called when a new endpoint starts up. We need to allocate
2440	 * the sctp_inpcb structure from the zone and init it. Mark it as
2441	 * unbound and find a port that we can use as an ephemeral with
2442	 * INADDR_ANY. If the user binds later no problem we can then add in
2443	 * the specific addresses. And setup the default parameters for the
2444	 * EP.
2445	 */
2446	int i, error;
2447	struct sctp_inpcb *inp;
2448	struct sctp_pcb *m;
2449	struct timeval time;
2450	sctp_sharedkey_t *null_key;
2451
2452	error = 0;
2453
2454	SCTP_INP_INFO_WLOCK();
2455	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2456	if (inp == NULL) {
2457		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2458		SCTP_INP_INFO_WUNLOCK();
2459		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2460		return (ENOBUFS);
2461	}
2462	/* zap it */
2463	bzero(inp, sizeof(*inp));
2464
2465	/* bump generations */
2466	/* setup socket pointers */
2467	inp->sctp_socket = so;
2468	inp->ip_inp.inp.inp_socket = so;
2469	inp->ip_inp.inp.inp_cred = crhold(so->so_cred);
2470#ifdef INET6
2471	if (INP_SOCKAF(so) == AF_INET6) {
2472		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2473			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2474		}
2475		if (MODULE_GLOBAL(ip6_v6only)) {
2476			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2477		}
2478	}
2479#endif
2480	inp->sctp_associd_counter = 1;
2481	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2482	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2483	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2484	inp->sctp_ecn_enable = SCTP_BASE_SYSCTL(sctp_ecn_enable);
2485	/* init the small hash table we use to track asocid <-> tcb */
2486	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2487	if (inp->sctp_asocidhash == NULL) {
2488		crfree(inp->ip_inp.inp.inp_cred);
2489		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2490		SCTP_INP_INFO_WUNLOCK();
2491		return (ENOBUFS);
2492	}
2493#ifdef IPSEC
2494	{
2495		struct inpcbpolicy *pcb_sp = NULL;
2496
2497		error = ipsec_init_policy(so, &pcb_sp);
2498		/* Arrange to share the policy */
2499		inp->ip_inp.inp.inp_sp = pcb_sp;
2500		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2501	}
2502	if (error != 0) {
2503		crfree(inp->ip_inp.inp.inp_cred);
2504		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2505		SCTP_INP_INFO_WUNLOCK();
2506		return error;
2507	}
2508#endif				/* IPSEC */
2509	SCTP_INCR_EP_COUNT();
2510	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2511	SCTP_INP_INFO_WUNLOCK();
2512
2513	so->so_pcb = (caddr_t)inp;
2514
2515	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2516		/* UDP style socket */
2517		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2518		    SCTP_PCB_FLAGS_UNBOUND);
2519		/* Be sure it is NON-BLOCKING IO for UDP */
2520		/* SCTP_SET_SO_NBIO(so); */
2521	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2522		/* TCP style socket */
2523		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2524		    SCTP_PCB_FLAGS_UNBOUND);
2525		/* Be sure we have blocking IO by default */
2526		SCTP_CLEAR_SO_NBIO(so);
2527	} else {
2528		/*
2529		 * unsupported socket type (RAW, etc)- in case we missed it
2530		 * in protosw
2531		 */
2532		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2533		so->so_pcb = NULL;
2534		crfree(inp->ip_inp.inp.inp_cred);
2535		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2536		return (EOPNOTSUPP);
2537	}
2538	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2539		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2540		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2541	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2542		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2543		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2544	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2545		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2546		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2547	}
2548	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2549	    &inp->sctp_hashmark);
2550	if (inp->sctp_tcbhash == NULL) {
2551		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2552		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2553		so->so_pcb = NULL;
2554		crfree(inp->ip_inp.inp.inp_cred);
2555		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2556		return (ENOBUFS);
2557	}
2558	inp->def_vrf_id = vrf_id;
2559
2560	SCTP_INP_INFO_WLOCK();
2561	SCTP_INP_LOCK_INIT(inp);
2562	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2563	SCTP_INP_READ_INIT(inp);
2564	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2565	/* lock the new ep */
2566	SCTP_INP_WLOCK(inp);
2567
2568	/* add it to the info area */
2569	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2570	SCTP_INP_INFO_WUNLOCK();
2571
2572	TAILQ_INIT(&inp->read_queue);
2573	LIST_INIT(&inp->sctp_addr_list);
2574
2575	LIST_INIT(&inp->sctp_asoc_list);
2576
2577#ifdef SCTP_TRACK_FREED_ASOCS
2578	/* TEMP CODE */
2579	LIST_INIT(&inp->sctp_asoc_free_list);
2580#endif
2581	/* Init the timer structure for signature change */
2582	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2583	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2584
2585	/* now init the actual endpoint default data */
2586	m = &inp->sctp_ep;
2587
2588	/* setup the base timeout information */
2589	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2590	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2591	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2592	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2593	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2594	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2595	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2596	/* all max/min max are in ms */
2597	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2598	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2599	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2600	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2601	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2602	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2603	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2604	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2605	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2606	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2607	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2608	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2609	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2610
2611	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2612	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2613	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2614	/* number of streams to pre-open on a association */
2615	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2616
2617	/* Add adaptation cookie */
2618	m->adaptation_layer_indicator = 0;
2619	m->adaptation_layer_indicator_provided = 0;
2620
2621	/* seed random number generator */
2622	m->random_counter = 1;
2623	m->store_at = SCTP_SIGNATURE_SIZE;
2624	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2625	sctp_fill_random_store(m);
2626
2627	/* Minimum cookie size */
2628	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2629	    sizeof(struct sctp_state_cookie);
2630	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2631
2632	/* Setup the initial secret */
2633	(void)SCTP_GETTIME_TIMEVAL(&time);
2634	m->time_of_secret_change = time.tv_sec;
2635
2636	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2637		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2638	}
2639	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2640
2641	/* How long is a cookie good for ? */
2642	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2643	/*
2644	 * Initialize authentication parameters
2645	 */
2646	m->local_hmacs = sctp_default_supported_hmaclist();
2647	m->local_auth_chunks = sctp_alloc_chunklist();
2648	m->default_dscp = 0;
2649#ifdef INET6
2650	m->default_flowlabel = 0;
2651#endif
2652	m->port = 0;		/* encapsulation disabled by default */
2653	sctp_auth_set_default_chunks(m->local_auth_chunks);
2654	LIST_INIT(&m->shared_keys);
2655	/* add default NULL key as key id 0 */
2656	null_key = sctp_alloc_sharedkey();
2657	sctp_insert_sharedkey(&m->shared_keys, null_key);
2658	SCTP_INP_WUNLOCK(inp);
2659#ifdef SCTP_LOG_CLOSING
2660	sctp_log_closing(inp, NULL, 12);
2661#endif
2662	return (error);
2663}
2664
2665
2666void
2667sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2668    struct sctp_tcb *stcb)
2669{
2670	struct sctp_nets *net;
2671	uint16_t lport, rport;
2672	struct sctppcbhead *head;
2673	struct sctp_laddr *laddr, *oladdr;
2674
2675	atomic_add_int(&stcb->asoc.refcnt, 1);
2676	SCTP_TCB_UNLOCK(stcb);
2677	SCTP_INP_INFO_WLOCK();
2678	SCTP_INP_WLOCK(old_inp);
2679	SCTP_INP_WLOCK(new_inp);
2680	SCTP_TCB_LOCK(stcb);
2681	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2682
2683	new_inp->sctp_ep.time_of_secret_change =
2684	    old_inp->sctp_ep.time_of_secret_change;
2685	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2686	    sizeof(old_inp->sctp_ep.secret_key));
2687	new_inp->sctp_ep.current_secret_number =
2688	    old_inp->sctp_ep.current_secret_number;
2689	new_inp->sctp_ep.last_secret_number =
2690	    old_inp->sctp_ep.last_secret_number;
2691	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2692
2693	/* make it so new data pours into the new socket */
2694	stcb->sctp_socket = new_inp->sctp_socket;
2695	stcb->sctp_ep = new_inp;
2696
2697	/* Copy the port across */
2698	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2699	rport = stcb->rport;
2700	/* Pull the tcb from the old association */
2701	LIST_REMOVE(stcb, sctp_tcbhash);
2702	LIST_REMOVE(stcb, sctp_tcblist);
2703	if (stcb->asoc.in_asocid_hash) {
2704		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2705	}
2706	/* Now insert the new_inp into the TCP connected hash */
2707	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2708
2709	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2710	/* Its safe to access */
2711	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2712
2713	/* Now move the tcb into the endpoint list */
2714	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2715	/*
2716	 * Question, do we even need to worry about the ep-hash since we
2717	 * only have one connection? Probably not :> so lets get rid of it
2718	 * and not suck up any kernel memory in that.
2719	 */
2720	if (stcb->asoc.in_asocid_hash) {
2721		struct sctpasochead *lhd;
2722
2723		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2724		    new_inp->hashasocidmark)];
2725		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2726	}
2727	/* Ok. Let's restart timer. */
2728	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2729		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2730		    stcb, net);
2731	}
2732
2733	SCTP_INP_INFO_WUNLOCK();
2734	if (new_inp->sctp_tcbhash != NULL) {
2735		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2736		new_inp->sctp_tcbhash = NULL;
2737	}
2738	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2739		/* Subset bound, so copy in the laddr list from the old_inp */
2740		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2741			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2742			if (laddr == NULL) {
2743				/*
2744				 * Gak, what can we do? This assoc is really
2745				 * HOSED. We probably should send an abort
2746				 * here.
2747				 */
2748				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2749				continue;
2750			}
2751			SCTP_INCR_LADDR_COUNT();
2752			bzero(laddr, sizeof(*laddr));
2753			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2754			laddr->ifa = oladdr->ifa;
2755			atomic_add_int(&laddr->ifa->refcount, 1);
2756			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2757			    sctp_nxt_addr);
2758			new_inp->laddr_count++;
2759			if (oladdr == stcb->asoc.last_used_address) {
2760				stcb->asoc.last_used_address = laddr;
2761			}
2762		}
2763	}
2764	/*
2765	 * Now any running timers need to be adjusted since we really don't
2766	 * care if they are running or not just blast in the new_inp into
2767	 * all of them.
2768	 */
2769
2770	stcb->asoc.dack_timer.ep = (void *)new_inp;
2771	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2772	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2773	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2774	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2775	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2776	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2777	/* now what about the nets? */
2778	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2779		net->pmtu_timer.ep = (void *)new_inp;
2780		net->hb_timer.ep = (void *)new_inp;
2781		net->rxt_timer.ep = (void *)new_inp;
2782	}
2783	SCTP_INP_WUNLOCK(new_inp);
2784	SCTP_INP_WUNLOCK(old_inp);
2785}
2786
2787
2788
2789
2790/* sctp_ifap is used to bypass normal local address validation checks */
2791int
2792sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2793    struct sctp_ifa *sctp_ifap, struct thread *p)
2794{
2795	/* bind a ep to a socket address */
2796	struct sctppcbhead *head;
2797	struct sctp_inpcb *inp, *inp_tmp;
2798	struct inpcb *ip_inp;
2799	int port_reuse_active = 0;
2800	int bindall;
2801	uint16_t lport;
2802	int error;
2803	uint32_t vrf_id;
2804
2805	lport = 0;
2806	bindall = 1;
2807	inp = (struct sctp_inpcb *)so->so_pcb;
2808	ip_inp = (struct inpcb *)so->so_pcb;
2809#ifdef SCTP_DEBUG
2810	if (addr) {
2811		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2812		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2813		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2814		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2815	}
2816#endif
2817	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2818		/* already did a bind, subsequent binds NOT allowed ! */
2819		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2820		return (EINVAL);
2821	}
2822#ifdef INVARIANTS
2823	if (p == NULL)
2824		panic("null proc/thread");
2825#endif
2826	if (addr != NULL) {
2827		switch (addr->sa_family) {
2828#ifdef INET
2829		case AF_INET:
2830			{
2831				struct sockaddr_in *sin;
2832
2833				/* IPV6_V6ONLY socket? */
2834				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2835					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2836					return (EINVAL);
2837				}
2838				if (addr->sa_len != sizeof(*sin)) {
2839					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2840					return (EINVAL);
2841				}
2842				sin = (struct sockaddr_in *)addr;
2843				lport = sin->sin_port;
2844				/*
2845				 * For LOOPBACK the prison_local_ip4() call
2846				 * will transmute the ip address to the
2847				 * proper value.
2848				 */
2849				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2850					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2851					return (error);
2852				}
2853				if (sin->sin_addr.s_addr != INADDR_ANY) {
2854					bindall = 0;
2855				}
2856				break;
2857			}
2858#endif
2859#ifdef INET6
2860		case AF_INET6:
2861			{
2862				/*
2863				 * Only for pure IPv6 Address. (No IPv4
2864				 * Mapped!)
2865				 */
2866				struct sockaddr_in6 *sin6;
2867
2868				sin6 = (struct sockaddr_in6 *)addr;
2869
2870				if (addr->sa_len != sizeof(*sin6)) {
2871					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2872					return (EINVAL);
2873				}
2874				lport = sin6->sin6_port;
2875				/*
2876				 * For LOOPBACK the prison_local_ip6() call
2877				 * will transmute the ipv6 address to the
2878				 * proper value.
2879				 */
2880				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2881				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2882					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2883					return (error);
2884				}
2885				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2886					bindall = 0;
2887					/* KAME hack: embed scopeid */
2888					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2889						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2890						return (EINVAL);
2891					}
2892				}
2893				/* this must be cleared for ifa_ifwithaddr() */
2894				sin6->sin6_scope_id = 0;
2895				break;
2896			}
2897#endif
2898		default:
2899			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2900			return (EAFNOSUPPORT);
2901		}
2902	}
2903	SCTP_INP_INFO_WLOCK();
2904	SCTP_INP_WLOCK(inp);
2905	/* Setup a vrf_id to be the default for the non-bind-all case. */
2906	vrf_id = inp->def_vrf_id;
2907
2908	/* increase our count due to the unlock we do */
2909	SCTP_INP_INCR_REF(inp);
2910	if (lport) {
2911		/*
2912		 * Did the caller specify a port? if so we must see if an ep
2913		 * already has this one bound.
2914		 */
2915		/* got to be root to get at low ports */
2916		if (ntohs(lport) < IPPORT_RESERVED) {
2917			if (p && (error =
2918			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2919			    )) {
2920				SCTP_INP_DECR_REF(inp);
2921				SCTP_INP_WUNLOCK(inp);
2922				SCTP_INP_INFO_WUNLOCK();
2923				return (error);
2924			}
2925		}
2926		SCTP_INP_WUNLOCK(inp);
2927		if (bindall) {
2928			vrf_id = inp->def_vrf_id;
2929			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2930			if (inp_tmp != NULL) {
2931				/*
2932				 * lock guy returned and lower count note
2933				 * that we are not bound so inp_tmp should
2934				 * NEVER be inp. And it is this inp
2935				 * (inp_tmp) that gets the reference bump,
2936				 * so we must lower it.
2937				 */
2938				SCTP_INP_DECR_REF(inp_tmp);
2939				/* unlock info */
2940				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2941				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2942					/*
2943					 * Ok, must be one-2-one and
2944					 * allowing port re-use
2945					 */
2946					port_reuse_active = 1;
2947					goto continue_anyway;
2948				}
2949				SCTP_INP_DECR_REF(inp);
2950				SCTP_INP_INFO_WUNLOCK();
2951				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2952				return (EADDRINUSE);
2953			}
2954		} else {
2955			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2956			if (inp_tmp != NULL) {
2957				/*
2958				 * lock guy returned and lower count note
2959				 * that we are not bound so inp_tmp should
2960				 * NEVER be inp. And it is this inp
2961				 * (inp_tmp) that gets the reference bump,
2962				 * so we must lower it.
2963				 */
2964				SCTP_INP_DECR_REF(inp_tmp);
2965				/* unlock info */
2966				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2967				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2968					/*
2969					 * Ok, must be one-2-one and
2970					 * allowing port re-use
2971					 */
2972					port_reuse_active = 1;
2973					goto continue_anyway;
2974				}
2975				SCTP_INP_DECR_REF(inp);
2976				SCTP_INP_INFO_WUNLOCK();
2977				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2978				return (EADDRINUSE);
2979			}
2980		}
2981continue_anyway:
2982		SCTP_INP_WLOCK(inp);
2983		if (bindall) {
2984			/* verify that no lport is not used by a singleton */
2985			if ((port_reuse_active == 0) &&
2986			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
2987				/* Sorry someone already has this one bound */
2988				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2989				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2990					port_reuse_active = 1;
2991				} else {
2992					SCTP_INP_DECR_REF(inp);
2993					SCTP_INP_WUNLOCK(inp);
2994					SCTP_INP_INFO_WUNLOCK();
2995					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2996					return (EADDRINUSE);
2997				}
2998			}
2999		}
3000	} else {
3001		uint16_t first, last, candidate;
3002		uint16_t count;
3003		int done;
3004
3005		if (ip_inp->inp_flags & INP_HIGHPORT) {
3006			first = MODULE_GLOBAL(ipport_hifirstauto);
3007			last = MODULE_GLOBAL(ipport_hilastauto);
3008		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3009			if (p && (error =
3010			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3011			    )) {
3012				SCTP_INP_DECR_REF(inp);
3013				SCTP_INP_WUNLOCK(inp);
3014				SCTP_INP_INFO_WUNLOCK();
3015				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3016				return (error);
3017			}
3018			first = MODULE_GLOBAL(ipport_lowfirstauto);
3019			last = MODULE_GLOBAL(ipport_lowlastauto);
3020		} else {
3021			first = MODULE_GLOBAL(ipport_firstauto);
3022			last = MODULE_GLOBAL(ipport_lastauto);
3023		}
3024		if (first > last) {
3025			uint16_t temp;
3026
3027			temp = first;
3028			first = last;
3029			last = temp;
3030		}
3031		count = last - first + 1;	/* number of candidates */
3032		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3033
3034		done = 0;
3035		while (!done) {
3036			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3037				done = 1;
3038			}
3039			if (!done) {
3040				if (--count == 0) {
3041					SCTP_INP_DECR_REF(inp);
3042					SCTP_INP_WUNLOCK(inp);
3043					SCTP_INP_INFO_WUNLOCK();
3044					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3045					return (EADDRINUSE);
3046				}
3047				if (candidate == last)
3048					candidate = first;
3049				else
3050					candidate = candidate + 1;
3051			}
3052		}
3053		lport = htons(candidate);
3054	}
3055	SCTP_INP_DECR_REF(inp);
3056	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3057	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3058		/*
3059		 * this really should not happen. The guy did a non-blocking
3060		 * bind and then did a close at the same time.
3061		 */
3062		SCTP_INP_WUNLOCK(inp);
3063		SCTP_INP_INFO_WUNLOCK();
3064		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3065		return (EINVAL);
3066	}
3067	/* ok we look clear to give out this port, so lets setup the binding */
3068	if (bindall) {
3069		/* binding to all addresses, so just set in the proper flags */
3070		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3071		/* set the automatic addr changes from kernel flag */
3072		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3073			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3074			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3075		} else {
3076			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3077			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3078		}
3079		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3080			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3081		} else {
3082			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3083		}
3084		/*
3085		 * set the automatic mobility_base from kernel flag (by
3086		 * micchie)
3087		 */
3088		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3089			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3090			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3091		} else {
3092			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3093			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3094		}
3095		/*
3096		 * set the automatic mobility_fasthandoff from kernel flag
3097		 * (by micchie)
3098		 */
3099		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3100			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3101			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3102		} else {
3103			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3104			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3105		}
3106	} else {
3107		/*
3108		 * bind specific, make sure flags is off and add a new
3109		 * address structure to the sctp_addr_list inside the ep
3110		 * structure.
3111		 *
3112		 * We will need to allocate one and insert it at the head. The
3113		 * socketopt call can just insert new addresses in there as
3114		 * well. It will also have to do the embed scope kame hack
3115		 * too (before adding).
3116		 */
3117		struct sctp_ifa *ifa;
3118		struct sockaddr_storage store_sa;
3119
3120		memset(&store_sa, 0, sizeof(store_sa));
3121		switch (addr->sa_family) {
3122#ifdef INET
3123		case AF_INET:
3124			{
3125				struct sockaddr_in *sin;
3126
3127				sin = (struct sockaddr_in *)&store_sa;
3128				memcpy(sin, addr, sizeof(struct sockaddr_in));
3129				sin->sin_port = 0;
3130				break;
3131			}
3132#endif
3133#ifdef INET6
3134		case AF_INET6:
3135			{
3136				struct sockaddr_in6 *sin6;
3137
3138				sin6 = (struct sockaddr_in6 *)&store_sa;
3139				memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3140				sin6->sin6_port = 0;
3141				break;
3142			}
3143#endif
3144		default:
3145			break;
3146		}
3147		/*
3148		 * first find the interface with the bound address need to
3149		 * zero out the port to find the address! yuck! can't do
3150		 * this earlier since need port for sctp_pcb_findep()
3151		 */
3152		if (sctp_ifap != NULL) {
3153			ifa = sctp_ifap;
3154		} else {
3155			/*
3156			 * Note for BSD we hit here always other O/S's will
3157			 * pass things in via the sctp_ifap argument
3158			 * (Panda).
3159			 */
3160			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3161			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3162		}
3163		if (ifa == NULL) {
3164			/* Can't find an interface with that address */
3165			SCTP_INP_WUNLOCK(inp);
3166			SCTP_INP_INFO_WUNLOCK();
3167			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3168			return (EADDRNOTAVAIL);
3169		}
3170#ifdef INET6
3171		if (addr->sa_family == AF_INET6) {
3172			/* GAK, more FIXME IFA lock? */
3173			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3174				/* Can't bind a non-existent addr. */
3175				SCTP_INP_WUNLOCK(inp);
3176				SCTP_INP_INFO_WUNLOCK();
3177				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3178				return (EINVAL);
3179			}
3180		}
3181#endif
3182		/* we're not bound all */
3183		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3184		/* allow bindx() to send ASCONF's for binding changes */
3185		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3186		/* clear automatic addr changes from kernel flag */
3187		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3188
3189		/* add this address to the endpoint list */
3190		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3191		if (error != 0) {
3192			SCTP_INP_WUNLOCK(inp);
3193			SCTP_INP_INFO_WUNLOCK();
3194			return (error);
3195		}
3196		inp->laddr_count++;
3197	}
3198	/* find the bucket */
3199	if (port_reuse_active) {
3200		/* Put it into tcp 1-2-1 hash */
3201		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3202		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3203	} else {
3204		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3205	}
3206	/* put it in the bucket */
3207	LIST_INSERT_HEAD(head, inp, sctp_hash);
3208	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3209	    (void *)head, ntohs(lport), port_reuse_active);
3210	/* set in the port */
3211	inp->sctp_lport = lport;
3212
3213	/* turn off just the unbound flag */
3214	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3215	SCTP_INP_WUNLOCK(inp);
3216	SCTP_INP_INFO_WUNLOCK();
3217	return (0);
3218}
3219
3220
3221static void
3222sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3223{
3224	struct sctp_iterator *it, *nit;
3225
3226	/*
3227	 * We enter with the only the ITERATOR_LOCK in place and a write
3228	 * lock on the inp_info stuff.
3229	 */
3230	it = sctp_it_ctl.cur_it;
3231	if (it && (it->vn != curvnet)) {
3232		/* Its not looking at our VNET */
3233		return;
3234	}
3235	if (it && (it->inp == inp)) {
3236		/*
3237		 * This is tricky and we hold the iterator lock, but when it
3238		 * returns and gets the lock (when we release it) the
3239		 * iterator will try to operate on inp. We need to stop that
3240		 * from happening. But of course the iterator has a
3241		 * reference on the stcb and inp. We can mark it and it will
3242		 * stop.
3243		 *
3244		 * If its a single iterator situation, we set the end iterator
3245		 * flag. Otherwise we set the iterator to go to the next
3246		 * inp.
3247		 *
3248		 */
3249		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3250			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3251		} else {
3252			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3253		}
3254	}
3255	/*
3256	 * Now go through and remove any single reference to our inp that
3257	 * may be still pending on the list
3258	 */
3259	SCTP_IPI_ITERATOR_WQ_LOCK();
3260	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3261		if (it->vn != curvnet) {
3262			continue;
3263		}
3264		if (it->inp == inp) {
3265			/* This one points to me is it inp specific? */
3266			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3267				/* Remove and free this one */
3268				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3269				    it, sctp_nxt_itr);
3270				if (it->function_atend != NULL) {
3271					(*it->function_atend) (it->pointer, it->val);
3272				}
3273				SCTP_FREE(it, SCTP_M_ITER);
3274			} else {
3275				it->inp = LIST_NEXT(it->inp, sctp_list);
3276				if (it->inp) {
3277					SCTP_INP_INCR_REF(it->inp);
3278				}
3279			}
3280			/*
3281			 * When its put in the refcnt is incremented so decr
3282			 * it
3283			 */
3284			SCTP_INP_DECR_REF(inp);
3285		}
3286	}
3287	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3288}
3289
3290/* release sctp_inpcb unbind the port */
3291void
3292sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3293{
3294	/*
3295	 * Here we free a endpoint. We must find it (if it is in the Hash
3296	 * table) and remove it from there. Then we must also find it in the
3297	 * overall list and remove it from there. After all removals are
3298	 * complete then any timer has to be stopped. Then start the actual
3299	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3300	 * all associations. d) finally the ep itself.
3301	 */
3302	struct sctp_tcb *asoc, *nasoc;
3303	struct sctp_laddr *laddr, *nladdr;
3304	struct inpcb *ip_pcb;
3305	struct socket *so;
3306	int being_refed = 0;
3307	struct sctp_queued_to_read *sq, *nsq;
3308	int cnt;
3309	sctp_sharedkey_t *shared_key, *nshared_key;
3310
3311
3312#ifdef SCTP_LOG_CLOSING
3313	sctp_log_closing(inp, NULL, 0);
3314#endif
3315	SCTP_ITERATOR_LOCK();
3316	/* mark any iterators on the list or being processed */
3317	sctp_iterator_inp_being_freed(inp);
3318	SCTP_ITERATOR_UNLOCK();
3319	so = inp->sctp_socket;
3320	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3321		/* been here before.. eeks.. get out of here */
3322		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3323#ifdef SCTP_LOG_CLOSING
3324		sctp_log_closing(inp, NULL, 1);
3325#endif
3326		return;
3327	}
3328	SCTP_ASOC_CREATE_LOCK(inp);
3329	SCTP_INP_INFO_WLOCK();
3330
3331	SCTP_INP_WLOCK(inp);
3332	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3333		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3334		/* socket is gone, so no more wakeups allowed */
3335		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3336		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3337		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3338
3339	}
3340	/* First time through we have the socket lock, after that no more. */
3341	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3342	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3343
3344	if (inp->control) {
3345		sctp_m_freem(inp->control);
3346		inp->control = NULL;
3347	}
3348	if (inp->pkt) {
3349		sctp_m_freem(inp->pkt);
3350		inp->pkt = NULL;
3351	}
3352	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3353					 * here but I will be nice :> (i.e.
3354					 * ip_pcb = ep;) */
3355	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3356		int cnt_in_sd;
3357
3358		cnt_in_sd = 0;
3359		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3360			SCTP_TCB_LOCK(asoc);
3361			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3362				/* Skip guys being freed */
3363				cnt_in_sd++;
3364				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3365					/*
3366					 * Special case - we did not start a
3367					 * kill timer on the asoc due to it
3368					 * was not closed. So go ahead and
3369					 * start it now.
3370					 */
3371					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3372					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3373				}
3374				SCTP_TCB_UNLOCK(asoc);
3375				continue;
3376			}
3377			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3378			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3379			    (asoc->asoc.total_output_queue_size == 0)) {
3380				/*
3381				 * If we have data in queue, we don't want
3382				 * to just free since the app may have done,
3383				 * send()/close or connect/send/close. And
3384				 * it wants the data to get across first.
3385				 */
3386				/* Just abandon things in the front states */
3387				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3388				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3389					cnt_in_sd++;
3390				}
3391				continue;
3392			}
3393			/* Disconnect the socket please */
3394			asoc->sctp_socket = NULL;
3395			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3396			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3397			    (asoc->asoc.control_pdapi) ||
3398			    (asoc->asoc.size_on_all_streams > 0) ||
3399			    (so && (so->so_rcv.sb_cc > 0))) {
3400				/* Left with Data unread */
3401				struct mbuf *op_err;
3402
3403				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3404				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3405				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3406				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3407				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3408				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3409					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3410				}
3411				if (sctp_free_assoc(inp, asoc,
3412				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3413					cnt_in_sd++;
3414				}
3415				continue;
3416			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3417				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3418			    (asoc->asoc.stream_queue_cnt == 0)) {
3419				if (asoc->asoc.locked_on_sending) {
3420					goto abort_anyway;
3421				}
3422				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3423				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3424					struct sctp_nets *netp;
3425
3426					/*
3427					 * there is nothing queued to send,
3428					 * so I send shutdown
3429					 */
3430					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3431					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3432						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3433					}
3434					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3435					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3436					sctp_stop_timers_for_shutdown(asoc);
3437					if (asoc->asoc.alternate) {
3438						netp = asoc->asoc.alternate;
3439					} else {
3440						netp = asoc->asoc.primary_destination;
3441					}
3442					sctp_send_shutdown(asoc, netp);
3443					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3444					    netp);
3445					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3446					    asoc->asoc.primary_destination);
3447					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3448				}
3449			} else {
3450				/* mark into shutdown pending */
3451				struct sctp_stream_queue_pending *sp;
3452
3453				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3454				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3455				    asoc->asoc.primary_destination);
3456				if (asoc->asoc.locked_on_sending) {
3457					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3458					    sctp_streamhead);
3459					if (sp == NULL) {
3460						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3461						    (void *)asoc->asoc.locked_on_sending,
3462						    asoc->asoc.locked_on_sending->stream_no);
3463					} else {
3464						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3465							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3466					}
3467				}
3468				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3469				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3470				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3471					struct mbuf *op_err;
3472
3473			abort_anyway:
3474					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3475					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3476					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3477					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3478					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3479					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3480						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3481					}
3482					if (sctp_free_assoc(inp, asoc,
3483					    SCTP_PCBFREE_NOFORCE,
3484					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3485						cnt_in_sd++;
3486					}
3487					continue;
3488				} else {
3489					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3490				}
3491			}
3492			cnt_in_sd++;
3493			SCTP_TCB_UNLOCK(asoc);
3494		}
3495		/* now is there some left in our SHUTDOWN state? */
3496		if (cnt_in_sd) {
3497#ifdef SCTP_LOG_CLOSING
3498			sctp_log_closing(inp, NULL, 2);
3499#endif
3500			inp->sctp_socket = NULL;
3501			SCTP_INP_WUNLOCK(inp);
3502			SCTP_ASOC_CREATE_UNLOCK(inp);
3503			SCTP_INP_INFO_WUNLOCK();
3504			return;
3505		}
3506	}
3507	inp->sctp_socket = NULL;
3508	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3509	    SCTP_PCB_FLAGS_UNBOUND) {
3510		/*
3511		 * ok, this guy has been bound. It's port is somewhere in
3512		 * the SCTP_BASE_INFO(hash table). Remove it!
3513		 */
3514		LIST_REMOVE(inp, sctp_hash);
3515		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3516	}
3517	/*
3518	 * If there is a timer running to kill us, forget it, since it may
3519	 * have a contest on the INP lock.. which would cause us to die ...
3520	 */
3521	cnt = 0;
3522	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3523		SCTP_TCB_LOCK(asoc);
3524		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3525			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3526				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3527				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3528			}
3529			cnt++;
3530			SCTP_TCB_UNLOCK(asoc);
3531			continue;
3532		}
3533		/* Free associations that are NOT killing us */
3534		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3535		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3536			struct mbuf *op_err;
3537
3538			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3539			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3540			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3541			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3542		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3543			cnt++;
3544			SCTP_TCB_UNLOCK(asoc);
3545			continue;
3546		}
3547		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3548		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3549			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3550		}
3551		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3552			cnt++;
3553		}
3554	}
3555	if (cnt) {
3556		/* Ok we have someone out there that will kill us */
3557		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3558#ifdef SCTP_LOG_CLOSING
3559		sctp_log_closing(inp, NULL, 3);
3560#endif
3561		SCTP_INP_WUNLOCK(inp);
3562		SCTP_ASOC_CREATE_UNLOCK(inp);
3563		SCTP_INP_INFO_WUNLOCK();
3564		return;
3565	}
3566	if (SCTP_INP_LOCK_CONTENDED(inp))
3567		being_refed++;
3568	if (SCTP_INP_READ_CONTENDED(inp))
3569		being_refed++;
3570	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3571		being_refed++;
3572
3573	if ((inp->refcount) ||
3574	    (being_refed) ||
3575	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3576		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3577#ifdef SCTP_LOG_CLOSING
3578		sctp_log_closing(inp, NULL, 4);
3579#endif
3580		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3581		SCTP_INP_WUNLOCK(inp);
3582		SCTP_ASOC_CREATE_UNLOCK(inp);
3583		SCTP_INP_INFO_WUNLOCK();
3584		return;
3585	}
3586	inp->sctp_ep.signature_change.type = 0;
3587	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3588	/*
3589	 * Remove it from the list .. last thing we need a lock for.
3590	 */
3591	LIST_REMOVE(inp, sctp_list);
3592	SCTP_INP_WUNLOCK(inp);
3593	SCTP_ASOC_CREATE_UNLOCK(inp);
3594	SCTP_INP_INFO_WUNLOCK();
3595	/*
3596	 * Now we release all locks. Since this INP cannot be found anymore
3597	 * except possibly by the kill timer that might be running. We call
3598	 * the drain function here. It should hit the case were it sees the
3599	 * ACTIVE flag cleared and exit out freeing us to proceed and
3600	 * destroy everything.
3601	 */
3602	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3603		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3604	} else {
3605		/* Probably un-needed */
3606		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3607	}
3608
3609#ifdef SCTP_LOG_CLOSING
3610	sctp_log_closing(inp, NULL, 5);
3611#endif
3612
3613
3614	if ((inp->sctp_asocidhash) != NULL) {
3615		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3616		inp->sctp_asocidhash = NULL;
3617	}
3618	/* sa_ignore FREED_MEMORY */
3619	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3620		/* Its only abandoned if it had data left */
3621		if (sq->length)
3622			SCTP_STAT_INCR(sctps_left_abandon);
3623
3624		TAILQ_REMOVE(&inp->read_queue, sq, next);
3625		sctp_free_remote_addr(sq->whoFrom);
3626		if (so)
3627			so->so_rcv.sb_cc -= sq->length;
3628		if (sq->data) {
3629			sctp_m_freem(sq->data);
3630			sq->data = NULL;
3631		}
3632		/*
3633		 * no need to free the net count, since at this point all
3634		 * assoc's are gone.
3635		 */
3636		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3637		SCTP_DECR_READQ_COUNT();
3638	}
3639	/* Now the sctp_pcb things */
3640	/*
3641	 * free each asoc if it is not already closed/free. we can't use the
3642	 * macro here since le_next will get freed as part of the
3643	 * sctp_free_assoc() call.
3644	 */
3645	if (so) {
3646#ifdef IPSEC
3647		ipsec_delete_pcbpolicy(ip_pcb);
3648#endif				/* IPSEC */
3649
3650		/* Unlocks not needed since the socket is gone now */
3651	}
3652	if (ip_pcb->inp_options) {
3653		(void)sctp_m_free(ip_pcb->inp_options);
3654		ip_pcb->inp_options = 0;
3655	}
3656#ifdef INET6
3657	if (ip_pcb->inp_vflag & INP_IPV6) {
3658		struct in6pcb *in6p;
3659
3660		in6p = (struct in6pcb *)inp;
3661		ip6_freepcbopts(in6p->in6p_outputopts);
3662	}
3663#endif				/* INET6 */
3664	ip_pcb->inp_vflag = 0;
3665	/* free up authentication fields */
3666	if (inp->sctp_ep.local_auth_chunks != NULL)
3667		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3668	if (inp->sctp_ep.local_hmacs != NULL)
3669		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3670
3671	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3672		LIST_REMOVE(shared_key, next);
3673		sctp_free_sharedkey(shared_key);
3674		/* sa_ignore FREED_MEMORY */
3675	}
3676
3677	/*
3678	 * if we have an address list the following will free the list of
3679	 * ifaddr's that are set into this ep. Again macro limitations here,
3680	 * since the LIST_FOREACH could be a bad idea.
3681	 */
3682	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3683		sctp_remove_laddr(laddr);
3684	}
3685
3686#ifdef SCTP_TRACK_FREED_ASOCS
3687	/* TEMP CODE */
3688	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3689		LIST_REMOVE(asoc, sctp_tcblist);
3690		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3691		SCTP_DECR_ASOC_COUNT();
3692	}
3693	/* *** END TEMP CODE *** */
3694#endif
3695	/* Now lets see about freeing the EP hash table. */
3696	if (inp->sctp_tcbhash != NULL) {
3697		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3698		inp->sctp_tcbhash = NULL;
3699	}
3700	/* Now we must put the ep memory back into the zone pool */
3701	crfree(inp->ip_inp.inp.inp_cred);
3702	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3703	SCTP_INP_LOCK_DESTROY(inp);
3704	SCTP_INP_READ_DESTROY(inp);
3705	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3706	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3707	SCTP_DECR_EP_COUNT();
3708}
3709
3710
3711struct sctp_nets *
3712sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3713{
3714	struct sctp_nets *net;
3715
3716	/* locate the address */
3717	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3718		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3719			return (net);
3720	}
3721	return (NULL);
3722}
3723
3724
3725int
3726sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3727{
3728	struct sctp_ifa *sctp_ifa;
3729
3730	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3731	if (sctp_ifa) {
3732		return (1);
3733	} else {
3734		return (0);
3735	}
3736}
3737
3738/*
3739 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3740 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3741 * stats of stuff.
3742 */
3743int
3744sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3745    struct sctp_nets **netp, int set_scope, int from)
3746{
3747	/*
3748	 * The following is redundant to the same lines in the
3749	 * sctp_aloc_assoc() but is needed since others call the add address
3750	 * function
3751	 */
3752	struct sctp_nets *net, *netfirst;
3753	int addr_inscope;
3754
3755	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3756	    from);
3757	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3758
3759	netfirst = sctp_findnet(stcb, newaddr);
3760	if (netfirst) {
3761		/*
3762		 * Lie and return ok, we don't want to make the association
3763		 * go away for this behavior. It will happen in the TCP
3764		 * model in a connected socket. It does not reach the hash
3765		 * table until after the association is built so it can't be
3766		 * found. Mark as reachable, since the initial creation will
3767		 * have been cleared and the NOT_IN_ASSOC flag will have
3768		 * been added... and we don't want to end up removing it
3769		 * back out.
3770		 */
3771		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3772			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3773			    SCTP_ADDR_UNCONFIRMED);
3774		} else {
3775			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3776		}
3777
3778		return (0);
3779	}
3780	addr_inscope = 1;
3781	switch (newaddr->sa_family) {
3782#ifdef INET
3783	case AF_INET:
3784		{
3785			struct sockaddr_in *sin;
3786
3787			sin = (struct sockaddr_in *)newaddr;
3788			if (sin->sin_addr.s_addr == 0) {
3789				/* Invalid address */
3790				return (-1);
3791			}
3792			/* zero out the bzero area */
3793			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3794
3795			/* assure len is set */
3796			sin->sin_len = sizeof(struct sockaddr_in);
3797			if (set_scope) {
3798#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3799				stcb->asoc.scope.ipv4_local_scope = 1;
3800#else
3801				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3802					stcb->asoc.scope.ipv4_local_scope = 1;
3803				}
3804#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3805			} else {
3806				/* Validate the address is in scope */
3807				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3808				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3809					addr_inscope = 0;
3810				}
3811			}
3812			break;
3813		}
3814#endif
3815#ifdef INET6
3816	case AF_INET6:
3817		{
3818			struct sockaddr_in6 *sin6;
3819
3820			sin6 = (struct sockaddr_in6 *)newaddr;
3821			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3822				/* Invalid address */
3823				return (-1);
3824			}
3825			/* assure len is set */
3826			sin6->sin6_len = sizeof(struct sockaddr_in6);
3827			if (set_scope) {
3828				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3829					stcb->asoc.scope.loopback_scope = 1;
3830					stcb->asoc.scope.local_scope = 0;
3831					stcb->asoc.scope.ipv4_local_scope = 1;
3832					stcb->asoc.scope.site_scope = 1;
3833				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3834					/*
3835					 * If the new destination is a
3836					 * LINK_LOCAL we must have common
3837					 * site scope. Don't set the local
3838					 * scope since we may not share all
3839					 * links, only loopback can do this.
3840					 * Links on the local network would
3841					 * also be on our private network
3842					 * for v4 too.
3843					 */
3844					stcb->asoc.scope.ipv4_local_scope = 1;
3845					stcb->asoc.scope.site_scope = 1;
3846				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3847					/*
3848					 * If the new destination is
3849					 * SITE_LOCAL then we must have site
3850					 * scope in common.
3851					 */
3852					stcb->asoc.scope.site_scope = 1;
3853				}
3854			} else {
3855				/* Validate the address is in scope */
3856				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3857				    (stcb->asoc.scope.loopback_scope == 0)) {
3858					addr_inscope = 0;
3859				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3860				    (stcb->asoc.scope.local_scope == 0)) {
3861					addr_inscope = 0;
3862				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3863				    (stcb->asoc.scope.site_scope == 0)) {
3864					addr_inscope = 0;
3865				}
3866			}
3867			break;
3868		}
3869#endif
3870	default:
3871		/* not supported family type */
3872		return (-1);
3873	}
3874	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3875	if (net == NULL) {
3876		return (-1);
3877	}
3878	SCTP_INCR_RADDR_COUNT();
3879	bzero(net, sizeof(struct sctp_nets));
3880	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3881	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3882	switch (newaddr->sa_family) {
3883#ifdef INET
3884	case AF_INET:
3885		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3886		break;
3887#endif
3888#ifdef INET6
3889	case AF_INET6:
3890		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3891		break;
3892#endif
3893	default:
3894		break;
3895	}
3896	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3897	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3898		stcb->asoc.scope.loopback_scope = 1;
3899		stcb->asoc.scope.ipv4_local_scope = 1;
3900		stcb->asoc.scope.local_scope = 0;
3901		stcb->asoc.scope.site_scope = 1;
3902		addr_inscope = 1;
3903	}
3904	net->failure_threshold = stcb->asoc.def_net_failure;
3905	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3906	if (addr_inscope == 0) {
3907		net->dest_state = (SCTP_ADDR_REACHABLE |
3908		    SCTP_ADDR_OUT_OF_SCOPE);
3909	} else {
3910		if (from == SCTP_ADDR_IS_CONFIRMED)
3911			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3912			net->dest_state = SCTP_ADDR_REACHABLE;
3913		else
3914			net->dest_state = SCTP_ADDR_REACHABLE |
3915			    SCTP_ADDR_UNCONFIRMED;
3916	}
3917	/*
3918	 * We set this to 0, the timer code knows that this means its an
3919	 * initial value
3920	 */
3921	net->rto_needed = 1;
3922	net->RTO = 0;
3923	net->RTO_measured = 0;
3924	stcb->asoc.numnets++;
3925	net->ref_count = 1;
3926	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3927	net->port = stcb->asoc.port;
3928	net->dscp = stcb->asoc.default_dscp;
3929#ifdef INET6
3930	net->flowlabel = stcb->asoc.default_flowlabel;
3931#endif
3932	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3933		net->dest_state |= SCTP_ADDR_NOHB;
3934	} else {
3935		net->dest_state &= ~SCTP_ADDR_NOHB;
3936	}
3937	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3938		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3939	} else {
3940		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3941	}
3942	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3943	/* Init the timer structure */
3944	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3945	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3946	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3947
3948	/* Now generate a route for this guy */
3949#ifdef INET6
3950	/* KAME hack: embed scopeid */
3951	if (newaddr->sa_family == AF_INET6) {
3952		struct sockaddr_in6 *sin6;
3953
3954		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3955		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3956		sin6->sin6_scope_id = 0;
3957	}
3958#endif
3959	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3960
3961	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3962		/* Get source address */
3963		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3964		    stcb,
3965		    (sctp_route_t *) & net->ro,
3966		    net,
3967		    0,
3968		    stcb->asoc.vrf_id);
3969		/* Now get the interface MTU */
3970		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3971			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3972		}
3973		if (net->mtu > 0) {
3974			uint32_t rmtu;
3975
3976			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3977			if (rmtu == 0) {
3978				/*
3979				 * Start things off to match mtu of
3980				 * interface please.
3981				 */
3982				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3983				    net->ro.ro_rt, net->mtu);
3984			} else {
3985				/*
3986				 * we take the route mtu over the interface,
3987				 * since the route may be leading out the
3988				 * loopback, or a different interface.
3989				 */
3990				net->mtu = rmtu;
3991			}
3992		}
3993	}
3994	if (net->mtu == 0) {
3995		switch (newaddr->sa_family) {
3996#ifdef INET
3997		case AF_INET:
3998			net->mtu = SCTP_DEFAULT_MTU;
3999			break;
4000#endif
4001#ifdef INET6
4002		case AF_INET6:
4003			net->mtu = 1280;
4004			break;
4005#endif
4006		default:
4007			break;
4008		}
4009	}
4010	if (net->port) {
4011		net->mtu -= (uint32_t) sizeof(struct udphdr);
4012	}
4013	if (from == SCTP_ALLOC_ASOC) {
4014		stcb->asoc.smallest_mtu = net->mtu;
4015	}
4016	if (stcb->asoc.smallest_mtu > net->mtu) {
4017		stcb->asoc.smallest_mtu = net->mtu;
4018	}
4019#ifdef INET6
4020	if (newaddr->sa_family == AF_INET6) {
4021		struct sockaddr_in6 *sin6;
4022
4023		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4024		(void)sa6_recoverscope(sin6);
4025	}
4026#endif
4027
4028	/* JRS - Use the congestion control given in the CC module */
4029	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4030		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4031
4032	/*
4033	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4034	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4035	 */
4036	net->find_pseudo_cumack = 1;
4037	net->find_rtx_pseudo_cumack = 1;
4038	net->src_addr_selected = 0;
4039	/* Choose an initial flowid. */
4040	net->flowid = stcb->asoc.my_vtag ^
4041	    ntohs(stcb->rport) ^
4042	    ntohs(stcb->sctp_ep->sctp_lport);
4043#ifdef INVARIANTS
4044	net->flowidset = 1;
4045#endif
4046	if (netp) {
4047		*netp = net;
4048	}
4049	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4050	if (net->ro.ro_rt == NULL) {
4051		/* Since we have no route put it at the back */
4052		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4053	} else if (netfirst == NULL) {
4054		/* We are the first one in the pool. */
4055		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4056	} else if (netfirst->ro.ro_rt == NULL) {
4057		/*
4058		 * First one has NO route. Place this one ahead of the first
4059		 * one.
4060		 */
4061		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4062	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4063		/*
4064		 * This one has a different interface than the one at the
4065		 * top of the list. Place it ahead.
4066		 */
4067		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4068	} else {
4069		/*
4070		 * Ok we have the same interface as the first one. Move
4071		 * forward until we find either a) one with a NULL route...
4072		 * insert ahead of that b) one with a different ifp.. insert
4073		 * after that. c) end of the list.. insert at the tail.
4074		 */
4075		struct sctp_nets *netlook;
4076
4077		do {
4078			netlook = TAILQ_NEXT(netfirst, sctp_next);
4079			if (netlook == NULL) {
4080				/* End of the list */
4081				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4082				break;
4083			} else if (netlook->ro.ro_rt == NULL) {
4084				/* next one has NO route */
4085				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4086				break;
4087			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4088				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4089				    net, sctp_next);
4090				break;
4091			}
4092			/* Shift forward */
4093			netfirst = netlook;
4094		} while (netlook != NULL);
4095	}
4096
4097	/* got to have a primary set */
4098	if (stcb->asoc.primary_destination == 0) {
4099		stcb->asoc.primary_destination = net;
4100	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4101		    (net->ro.ro_rt) &&
4102	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4103		/* No route to current primary adopt new primary */
4104		stcb->asoc.primary_destination = net;
4105	}
4106	/* Validate primary is first */
4107	net = TAILQ_FIRST(&stcb->asoc.nets);
4108	if ((net != stcb->asoc.primary_destination) &&
4109	    (stcb->asoc.primary_destination)) {
4110		/*
4111		 * first one on the list is NOT the primary sctp_cmpaddr()
4112		 * is much more efficient if the primary is the first on the
4113		 * list, make it so.
4114		 */
4115		TAILQ_REMOVE(&stcb->asoc.nets,
4116		    stcb->asoc.primary_destination, sctp_next);
4117		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4118		    stcb->asoc.primary_destination, sctp_next);
4119	}
4120	return (0);
4121}
4122
4123
4124static uint32_t
4125sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4126{
4127	uint32_t id;
4128	struct sctpasochead *head;
4129	struct sctp_tcb *lstcb;
4130
4131	SCTP_INP_WLOCK(inp);
4132try_again:
4133	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4134		/* TSNH */
4135		SCTP_INP_WUNLOCK(inp);
4136		return (0);
4137	}
4138	/*
4139	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4140	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4141	 */
4142	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4143		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4144	}
4145	id = inp->sctp_associd_counter;
4146	inp->sctp_associd_counter++;
4147	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4148	if (lstcb) {
4149		goto try_again;
4150	}
4151	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4152	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4153	stcb->asoc.in_asocid_hash = 1;
4154	SCTP_INP_WUNLOCK(inp);
4155	return id;
4156}
4157
4158/*
4159 * allocate an association and add it to the endpoint. The caller must be
4160 * careful to add all additional addresses once they are know right away or
4161 * else the assoc will be may experience a blackout scenario.
4162 */
4163struct sctp_tcb *
4164sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4165    int *error, uint32_t override_tag, uint32_t vrf_id,
4166    struct thread *p
4167)
4168{
4169	/* note the p argument is only valid in unbound sockets */
4170
4171	struct sctp_tcb *stcb;
4172	struct sctp_association *asoc;
4173	struct sctpasochead *head;
4174	uint16_t rport;
4175	int err;
4176
4177	/*
4178	 * Assumption made here: Caller has done a
4179	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4180	 * address does not exist already.
4181	 */
4182	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4183		/* Hit max assoc, sorry no more */
4184		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4185		*error = ENOBUFS;
4186		return (NULL);
4187	}
4188	if (firstaddr == NULL) {
4189		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4190		*error = EINVAL;
4191		return (NULL);
4192	}
4193	SCTP_INP_RLOCK(inp);
4194	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4195	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4196	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4197		/*
4198		 * If its in the TCP pool, its NOT allowed to create an
4199		 * association. The parent listener needs to call
4200		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4201		 * off, or connected one does this.. its an error.
4202		 */
4203		SCTP_INP_RUNLOCK(inp);
4204		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4205		*error = EINVAL;
4206		return (NULL);
4207	}
4208	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4209	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4210		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4211		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4212			SCTP_INP_RUNLOCK(inp);
4213			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4214			*error = EINVAL;
4215			return (NULL);
4216		}
4217	}
4218	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4219#ifdef SCTP_DEBUG
4220	if (firstaddr) {
4221		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4222		switch (firstaddr->sa_family) {
4223#ifdef INET
4224		case AF_INET:
4225			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4226			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4227			break;
4228#endif
4229#ifdef INET6
4230		case AF_INET6:
4231			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4232			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4233			break;
4234#endif
4235		default:
4236			break;
4237		}
4238	} else {
4239		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4240	}
4241#endif				/* SCTP_DEBUG */
4242	switch (firstaddr->sa_family) {
4243#ifdef INET
4244	case AF_INET:
4245		{
4246			struct sockaddr_in *sin;
4247
4248			sin = (struct sockaddr_in *)firstaddr;
4249			if ((ntohs(sin->sin_port) == 0) ||
4250			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4251			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4252			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4253				/* Invalid address */
4254				SCTP_INP_RUNLOCK(inp);
4255				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4256				*error = EINVAL;
4257				return (NULL);
4258			}
4259			rport = sin->sin_port;
4260			break;
4261		}
4262#endif
4263#ifdef INET6
4264	case AF_INET6:
4265		{
4266			struct sockaddr_in6 *sin6;
4267
4268			sin6 = (struct sockaddr_in6 *)firstaddr;
4269			if ((ntohs(sin6->sin6_port) == 0) ||
4270			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4271			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4272				/* Invalid address */
4273				SCTP_INP_RUNLOCK(inp);
4274				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4275				*error = EINVAL;
4276				return (NULL);
4277			}
4278			rport = sin6->sin6_port;
4279			break;
4280		}
4281#endif
4282	default:
4283		/* not supported family type */
4284		SCTP_INP_RUNLOCK(inp);
4285		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4286		*error = EINVAL;
4287		return (NULL);
4288	}
4289	SCTP_INP_RUNLOCK(inp);
4290	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4291		/*
4292		 * If you have not performed a bind, then we need to do the
4293		 * ephemeral bind for you.
4294		 */
4295		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4296		    (struct sockaddr *)NULL,
4297		    (struct sctp_ifa *)NULL,
4298		    p
4299		    ))) {
4300			/* bind error, probably perm */
4301			*error = err;
4302			return (NULL);
4303		}
4304	}
4305	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4306	if (stcb == NULL) {
4307		/* out of memory? */
4308		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4309		*error = ENOMEM;
4310		return (NULL);
4311	}
4312	SCTP_INCR_ASOC_COUNT();
4313
4314	bzero(stcb, sizeof(*stcb));
4315	asoc = &stcb->asoc;
4316
4317	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4318	SCTP_TCB_LOCK_INIT(stcb);
4319	SCTP_TCB_SEND_LOCK_INIT(stcb);
4320	stcb->rport = rport;
4321	/* setup back pointer's */
4322	stcb->sctp_ep = inp;
4323	stcb->sctp_socket = inp->sctp_socket;
4324	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4325		/* failed */
4326		SCTP_TCB_LOCK_DESTROY(stcb);
4327		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4328		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4329		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4330		SCTP_DECR_ASOC_COUNT();
4331		*error = err;
4332		return (NULL);
4333	}
4334	/* and the port */
4335	SCTP_INP_INFO_WLOCK();
4336	SCTP_INP_WLOCK(inp);
4337	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4338		/* inpcb freed while alloc going on */
4339		SCTP_TCB_LOCK_DESTROY(stcb);
4340		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4341		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4342		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4343		SCTP_INP_WUNLOCK(inp);
4344		SCTP_INP_INFO_WUNLOCK();
4345		SCTP_DECR_ASOC_COUNT();
4346		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4347		*error = EINVAL;
4348		return (NULL);
4349	}
4350	SCTP_TCB_LOCK(stcb);
4351
4352	/* now that my_vtag is set, add it to the hash */
4353	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4354	/* put it in the bucket in the vtag hash of assoc's for the system */
4355	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4356	SCTP_INP_INFO_WUNLOCK();
4357
4358	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4359		/* failure.. memory error? */
4360		if (asoc->strmout) {
4361			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4362			asoc->strmout = NULL;
4363		}
4364		if (asoc->mapping_array) {
4365			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4366			asoc->mapping_array = NULL;
4367		}
4368		if (asoc->nr_mapping_array) {
4369			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4370			asoc->nr_mapping_array = NULL;
4371		}
4372		SCTP_DECR_ASOC_COUNT();
4373		SCTP_TCB_UNLOCK(stcb);
4374		SCTP_TCB_LOCK_DESTROY(stcb);
4375		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4376		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4377		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4378		SCTP_INP_WUNLOCK(inp);
4379		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4380		*error = ENOBUFS;
4381		return (NULL);
4382	}
4383	/* Init all the timers */
4384	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4385	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4386	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4387	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4388	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4389	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4390	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4391
4392	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4393	/* now file the port under the hash as well */
4394	if (inp->sctp_tcbhash != NULL) {
4395		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4396		    inp->sctp_hashmark)];
4397		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4398	}
4399	SCTP_INP_WUNLOCK(inp);
4400	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4401	return (stcb);
4402}
4403
4404
4405void
4406sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4407{
4408	struct sctp_association *asoc;
4409
4410	asoc = &stcb->asoc;
4411	asoc->numnets--;
4412	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4413	if (net == asoc->primary_destination) {
4414		/* Reset primary */
4415		struct sctp_nets *lnet;
4416
4417		lnet = TAILQ_FIRST(&asoc->nets);
4418		/*
4419		 * Mobility adaptation Ideally, if deleted destination is
4420		 * the primary, it becomes a fast retransmission trigger by
4421		 * the subsequent SET PRIMARY. (by micchie)
4422		 */
4423		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4424		    SCTP_MOBILITY_BASE) ||
4425		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4426		    SCTP_MOBILITY_FASTHANDOFF)) {
4427			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4428			if (asoc->deleted_primary != NULL) {
4429				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4430				goto out;
4431			}
4432			asoc->deleted_primary = net;
4433			atomic_add_int(&net->ref_count, 1);
4434			memset(&net->lastsa, 0, sizeof(net->lastsa));
4435			memset(&net->lastsv, 0, sizeof(net->lastsv));
4436			sctp_mobility_feature_on(stcb->sctp_ep,
4437			    SCTP_MOBILITY_PRIM_DELETED);
4438			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4439			    stcb->sctp_ep, stcb, NULL);
4440		}
4441out:
4442		/* Try to find a confirmed primary */
4443		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4444	}
4445	if (net == asoc->last_data_chunk_from) {
4446		/* Reset primary */
4447		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4448	}
4449	if (net == asoc->last_control_chunk_from) {
4450		/* Clear net */
4451		asoc->last_control_chunk_from = NULL;
4452	}
4453	if (net == stcb->asoc.alternate) {
4454		sctp_free_remote_addr(stcb->asoc.alternate);
4455		stcb->asoc.alternate = NULL;
4456	}
4457	sctp_free_remote_addr(net);
4458}
4459
4460/*
4461 * remove a remote endpoint address from an association, it will fail if the
4462 * address does not exist.
4463 */
4464int
4465sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4466{
4467	/*
4468	 * Here we need to remove a remote address. This is quite simple, we
4469	 * first find it in the list of address for the association
4470	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4471	 * on that item. Note we do not allow it to be removed if there are
4472	 * no other addresses.
4473	 */
4474	struct sctp_association *asoc;
4475	struct sctp_nets *net, *nnet;
4476
4477	asoc = &stcb->asoc;
4478
4479	/* locate the address */
4480	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4481		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4482			continue;
4483		}
4484		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4485		    remaddr)) {
4486			/* we found the guy */
4487			if (asoc->numnets < 2) {
4488				/* Must have at LEAST two remote addresses */
4489				return (-1);
4490			} else {
4491				sctp_remove_net(stcb, net);
4492				return (0);
4493			}
4494		}
4495	}
4496	/* not found. */
4497	return (-2);
4498}
4499
4500void
4501sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4502{
4503	struct sctpvtaghead *chain;
4504	struct sctp_tagblock *twait_block;
4505	int found = 0;
4506	int i;
4507
4508	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4509	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4510		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4511			if ((twait_block->vtag_block[i].v_tag == tag) &&
4512			    (twait_block->vtag_block[i].lport == lport) &&
4513			    (twait_block->vtag_block[i].rport == rport)) {
4514				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4515				twait_block->vtag_block[i].v_tag = 0;
4516				twait_block->vtag_block[i].lport = 0;
4517				twait_block->vtag_block[i].rport = 0;
4518				found = 1;
4519				break;
4520			}
4521		}
4522		if (found)
4523			break;
4524	}
4525}
4526
4527int
4528sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4529{
4530	struct sctpvtaghead *chain;
4531	struct sctp_tagblock *twait_block;
4532	int found = 0;
4533	int i;
4534
4535	SCTP_INP_INFO_WLOCK();
4536	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4537	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4538		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4539			if ((twait_block->vtag_block[i].v_tag == tag) &&
4540			    (twait_block->vtag_block[i].lport == lport) &&
4541			    (twait_block->vtag_block[i].rport == rport)) {
4542				found = 1;
4543				break;
4544			}
4545		}
4546		if (found)
4547			break;
4548	}
4549	SCTP_INP_INFO_WUNLOCK();
4550	return (found);
4551}
4552
4553
4554void
4555sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4556{
4557	struct sctpvtaghead *chain;
4558	struct sctp_tagblock *twait_block;
4559	struct timeval now;
4560	int set, i;
4561
4562	if (time == 0) {
4563		/* Its disabled */
4564		return;
4565	}
4566	(void)SCTP_GETTIME_TIMEVAL(&now);
4567	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4568	set = 0;
4569	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4570		/* Block(s) present, lets find space, and expire on the fly */
4571		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4572			if ((twait_block->vtag_block[i].v_tag == 0) &&
4573			    !set) {
4574				twait_block->vtag_block[i].tv_sec_at_expire =
4575				    now.tv_sec + time;
4576				twait_block->vtag_block[i].v_tag = tag;
4577				twait_block->vtag_block[i].lport = lport;
4578				twait_block->vtag_block[i].rport = rport;
4579				set = 1;
4580			} else if ((twait_block->vtag_block[i].v_tag) &&
4581			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4582				/* Audit expires this guy */
4583				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4584				twait_block->vtag_block[i].v_tag = 0;
4585				twait_block->vtag_block[i].lport = 0;
4586				twait_block->vtag_block[i].rport = 0;
4587				if (set == 0) {
4588					/* Reuse it for my new tag */
4589					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4590					twait_block->vtag_block[i].v_tag = tag;
4591					twait_block->vtag_block[i].lport = lport;
4592					twait_block->vtag_block[i].rport = rport;
4593					set = 1;
4594				}
4595			}
4596		}
4597		if (set) {
4598			/*
4599			 * We only do up to the block where we can place our
4600			 * tag for audits
4601			 */
4602			break;
4603		}
4604	}
4605	/* Need to add a new block to chain */
4606	if (!set) {
4607		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4608		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4609		if (twait_block == NULL) {
4610#ifdef INVARIANTS
4611			panic("Can not alloc tagblock");
4612#endif
4613			return;
4614		}
4615		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4616		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4617		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4618		twait_block->vtag_block[0].v_tag = tag;
4619		twait_block->vtag_block[0].lport = lport;
4620		twait_block->vtag_block[0].rport = rport;
4621	}
4622}
4623
4624
4625
4626/*-
4627 * Free the association after un-hashing the remote port. This
4628 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4629 * expect that the input to this function IS a locked TCB.
4630 * It will return 0, if it did NOT destroy the association (instead
4631 * it unlocks it. It will return NON-zero if it either destroyed the
4632 * association OR the association is already destroyed.
4633 */
4634int
4635sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4636{
4637	int i;
4638	struct sctp_association *asoc;
4639	struct sctp_nets *net, *nnet;
4640	struct sctp_laddr *laddr, *naddr;
4641	struct sctp_tmit_chunk *chk, *nchk;
4642	struct sctp_asconf_addr *aparam, *naparam;
4643	struct sctp_asconf_ack *aack, *naack;
4644	struct sctp_stream_reset_list *strrst, *nstrrst;
4645	struct sctp_queued_to_read *sq, *nsq;
4646	struct sctp_stream_queue_pending *sp, *nsp;
4647	sctp_sharedkey_t *shared_key, *nshared_key;
4648	struct socket *so;
4649
4650	/* first, lets purge the entry from the hash table. */
4651
4652#ifdef SCTP_LOG_CLOSING
4653	sctp_log_closing(inp, stcb, 6);
4654#endif
4655	if (stcb->asoc.state == 0) {
4656#ifdef SCTP_LOG_CLOSING
4657		sctp_log_closing(inp, NULL, 7);
4658#endif
4659		/* there is no asoc, really TSNH :-0 */
4660		return (1);
4661	}
4662	if (stcb->asoc.alternate) {
4663		sctp_free_remote_addr(stcb->asoc.alternate);
4664		stcb->asoc.alternate = NULL;
4665	}
4666	/* TEMP CODE */
4667	if (stcb->freed_from_where == 0) {
4668		/* Only record the first place free happened from */
4669		stcb->freed_from_where = from_location;
4670	}
4671	/* TEMP CODE */
4672
4673	asoc = &stcb->asoc;
4674	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4675	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4676		/* nothing around */
4677		so = NULL;
4678	else
4679		so = inp->sctp_socket;
4680
4681	/*
4682	 * We used timer based freeing if a reader or writer is in the way.
4683	 * So we first check if we are actually being called from a timer,
4684	 * if so we abort early if a reader or writer is still in the way.
4685	 */
4686	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4687	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4688		/*
4689		 * is it the timer driving us? if so are the reader/writers
4690		 * gone?
4691		 */
4692		if (stcb->asoc.refcnt) {
4693			/* nope, reader or writer in the way */
4694			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4695			/* no asoc destroyed */
4696			SCTP_TCB_UNLOCK(stcb);
4697#ifdef SCTP_LOG_CLOSING
4698			sctp_log_closing(inp, stcb, 8);
4699#endif
4700			return (0);
4701		}
4702	}
4703	/* now clean up any other timers */
4704	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4705	asoc->dack_timer.self = NULL;
4706	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4707	/*-
4708	 * For stream reset we don't blast this unless
4709	 * it is a str-reset timer, it might be the
4710	 * free-asoc timer which we DON'T want to
4711	 * disturb.
4712	 */
4713	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4714		asoc->strreset_timer.self = NULL;
4715	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4716	asoc->asconf_timer.self = NULL;
4717	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4718	asoc->autoclose_timer.self = NULL;
4719	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4720	asoc->shut_guard_timer.self = NULL;
4721	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4722	asoc->delayed_event_timer.self = NULL;
4723	/* Mobility adaptation */
4724	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4725	asoc->delete_prim_timer.self = NULL;
4726	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4727		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4728		net->rxt_timer.self = NULL;
4729		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4730		net->pmtu_timer.self = NULL;
4731		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4732		net->hb_timer.self = NULL;
4733	}
4734	/* Now the read queue needs to be cleaned up (only once) */
4735	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4736		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4737		SCTP_INP_READ_LOCK(inp);
4738		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4739			if (sq->stcb == stcb) {
4740				sq->do_not_ref_stcb = 1;
4741				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4742				/*
4743				 * If there is no end, there never will be
4744				 * now.
4745				 */
4746				if (sq->end_added == 0) {
4747					/* Held for PD-API clear that. */
4748					sq->pdapi_aborted = 1;
4749					sq->held_length = 0;
4750					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4751						/*
4752						 * Need to add a PD-API
4753						 * aborted indication.
4754						 * Setting the control_pdapi
4755						 * assures that it will be
4756						 * added right after this
4757						 * msg.
4758						 */
4759						uint32_t strseq;
4760
4761						stcb->asoc.control_pdapi = sq;
4762						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4763						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4764						    stcb,
4765						    SCTP_PARTIAL_DELIVERY_ABORTED,
4766						    (void *)&strseq,
4767						    SCTP_SO_LOCKED);
4768						stcb->asoc.control_pdapi = NULL;
4769					}
4770				}
4771				/* Add an end to wake them */
4772				sq->end_added = 1;
4773			}
4774		}
4775		SCTP_INP_READ_UNLOCK(inp);
4776		if (stcb->block_entry) {
4777			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4778			stcb->block_entry->error = ECONNRESET;
4779			stcb->block_entry = NULL;
4780		}
4781	}
4782	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4783		/*
4784		 * Someone holds a reference OR the socket is unaccepted
4785		 * yet.
4786		 */
4787		if ((stcb->asoc.refcnt) ||
4788		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4789		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4790			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4791			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4792		}
4793		SCTP_TCB_UNLOCK(stcb);
4794		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4795		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4796			/* nothing around */
4797			so = NULL;
4798		if (so) {
4799			/* Wake any reader/writers */
4800			sctp_sorwakeup(inp, so);
4801			sctp_sowwakeup(inp, so);
4802		}
4803#ifdef SCTP_LOG_CLOSING
4804		sctp_log_closing(inp, stcb, 9);
4805#endif
4806		/* no asoc destroyed */
4807		return (0);
4808	}
4809#ifdef SCTP_LOG_CLOSING
4810	sctp_log_closing(inp, stcb, 10);
4811#endif
4812	/*
4813	 * When I reach here, no others want to kill the assoc yet.. and I
4814	 * own the lock. Now its possible an abort comes in when I do the
4815	 * lock exchange below to grab all the locks to do the final take
4816	 * out. to prevent this we increment the count, which will start a
4817	 * timer and blow out above thus assuring us that we hold exclusive
4818	 * killing of the asoc. Note that after getting back the TCB lock we
4819	 * will go ahead and increment the counter back up and stop any
4820	 * timer a passing stranger may have started :-S
4821	 */
4822	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4823		atomic_add_int(&stcb->asoc.refcnt, 1);
4824
4825		SCTP_TCB_UNLOCK(stcb);
4826		SCTP_INP_INFO_WLOCK();
4827		SCTP_INP_WLOCK(inp);
4828		SCTP_TCB_LOCK(stcb);
4829	}
4830	/* Double check the GONE flag */
4831	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4832	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4833		/* nothing around */
4834		so = NULL;
4835
4836	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4837	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4838		/*
4839		 * For TCP type we need special handling when we are
4840		 * connected. We also include the peel'ed off ones to.
4841		 */
4842		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4843			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4844			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4845			if (so) {
4846				SOCK_LOCK(so);
4847				if (so->so_rcv.sb_cc == 0) {
4848					so->so_state &= ~(SS_ISCONNECTING |
4849					    SS_ISDISCONNECTING |
4850					    SS_ISCONFIRMING |
4851					    SS_ISCONNECTED);
4852				}
4853				socantrcvmore_locked(so);
4854				sctp_sowwakeup(inp, so);
4855				sctp_sorwakeup(inp, so);
4856				SCTP_SOWAKEUP(so);
4857			}
4858		}
4859	}
4860	/*
4861	 * Make it invalid too, that way if its about to run it will abort
4862	 * and return.
4863	 */
4864	/* re-increment the lock */
4865	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4866		atomic_add_int(&stcb->asoc.refcnt, -1);
4867	}
4868	if (stcb->asoc.refcnt) {
4869		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4870		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4871		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4872			SCTP_INP_INFO_WUNLOCK();
4873			SCTP_INP_WUNLOCK(inp);
4874		}
4875		SCTP_TCB_UNLOCK(stcb);
4876		return (0);
4877	}
4878	asoc->state = 0;
4879	if (inp->sctp_tcbhash) {
4880		LIST_REMOVE(stcb, sctp_tcbhash);
4881	}
4882	if (stcb->asoc.in_asocid_hash) {
4883		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4884	}
4885	/* Now lets remove it from the list of ALL associations in the EP */
4886	LIST_REMOVE(stcb, sctp_tcblist);
4887	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4888		SCTP_INP_INCR_REF(inp);
4889		SCTP_INP_WUNLOCK(inp);
4890	}
4891	/* pull from vtag hash */
4892	LIST_REMOVE(stcb, sctp_asocs);
4893	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4894	    inp->sctp_lport, stcb->rport);
4895
4896	/*
4897	 * Now restop the timers to be sure this is paranoia at is finest!
4898	 */
4899	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4900	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4901	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4902	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4903	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4904	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4905	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4906	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4907		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4908		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4909		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4910	}
4911
4912	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4913	/*
4914	 * The chunk lists and such SHOULD be empty but we check them just
4915	 * in case.
4916	 */
4917	/* anything on the wheel needs to be removed */
4918	for (i = 0; i < asoc->streamoutcnt; i++) {
4919		struct sctp_stream_out *outs;
4920
4921		outs = &asoc->strmout[i];
4922		/* now clean up any chunks here */
4923		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4924			TAILQ_REMOVE(&outs->outqueue, sp, next);
4925			sctp_free_spbufspace(stcb, asoc, sp);
4926			if (sp->data) {
4927				if (so) {
4928					/* Still an open socket - report */
4929					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4930					    0, (void *)sp, SCTP_SO_LOCKED);
4931				}
4932				if (sp->data) {
4933					sctp_m_freem(sp->data);
4934					sp->data = NULL;
4935					sp->tail_mbuf = NULL;
4936					sp->length = 0;
4937				}
4938			}
4939			if (sp->net) {
4940				sctp_free_remote_addr(sp->net);
4941				sp->net = NULL;
4942			}
4943			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4944		}
4945	}
4946	/* sa_ignore FREED_MEMORY */
4947	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4948		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4949		SCTP_FREE(strrst, SCTP_M_STRESET);
4950	}
4951	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4952		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4953		if (sq->data) {
4954			sctp_m_freem(sq->data);
4955			sq->data = NULL;
4956		}
4957		sctp_free_remote_addr(sq->whoFrom);
4958		sq->whoFrom = NULL;
4959		sq->stcb = NULL;
4960		/* Free the ctl entry */
4961		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4962		SCTP_DECR_READQ_COUNT();
4963		/* sa_ignore FREED_MEMORY */
4964	}
4965	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4966		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4967		if (chk->data) {
4968			sctp_m_freem(chk->data);
4969			chk->data = NULL;
4970		}
4971		if (chk->holds_key_ref)
4972			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4973		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4974		SCTP_DECR_CHK_COUNT();
4975		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4976		asoc->free_chunk_cnt--;
4977		/* sa_ignore FREED_MEMORY */
4978	}
4979	/* pending send queue SHOULD be empty */
4980	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4981		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4982			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4983#ifdef INVARIANTS
4984		} else {
4985			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4986#endif
4987		}
4988		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4989		if (chk->data) {
4990			if (so) {
4991				/* Still a socket? */
4992				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
4993				    0, chk, SCTP_SO_LOCKED);
4994			}
4995			if (chk->data) {
4996				sctp_m_freem(chk->data);
4997				chk->data = NULL;
4998			}
4999		}
5000		if (chk->holds_key_ref)
5001			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5002		if (chk->whoTo) {
5003			sctp_free_remote_addr(chk->whoTo);
5004			chk->whoTo = NULL;
5005		}
5006		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5007		SCTP_DECR_CHK_COUNT();
5008		/* sa_ignore FREED_MEMORY */
5009	}
5010	/* sent queue SHOULD be empty */
5011	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5012		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5013			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
5014				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
5015#ifdef INVARIANTS
5016			} else {
5017				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
5018#endif
5019			}
5020		}
5021		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5022		if (chk->data) {
5023			if (so) {
5024				/* Still a socket? */
5025				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5026				    0, chk, SCTP_SO_LOCKED);
5027			}
5028			if (chk->data) {
5029				sctp_m_freem(chk->data);
5030				chk->data = NULL;
5031			}
5032		}
5033		if (chk->holds_key_ref)
5034			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5035		sctp_free_remote_addr(chk->whoTo);
5036		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5037		SCTP_DECR_CHK_COUNT();
5038		/* sa_ignore FREED_MEMORY */
5039	}
5040#ifdef INVARIANTS
5041	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5042		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5043			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5044		}
5045	}
5046#endif
5047	/* control queue MAY not be empty */
5048	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5049		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5050		if (chk->data) {
5051			sctp_m_freem(chk->data);
5052			chk->data = NULL;
5053		}
5054		if (chk->holds_key_ref)
5055			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5056		sctp_free_remote_addr(chk->whoTo);
5057		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5058		SCTP_DECR_CHK_COUNT();
5059		/* sa_ignore FREED_MEMORY */
5060	}
5061	/* ASCONF queue MAY not be empty */
5062	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5063		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5064		if (chk->data) {
5065			sctp_m_freem(chk->data);
5066			chk->data = NULL;
5067		}
5068		if (chk->holds_key_ref)
5069			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5070		sctp_free_remote_addr(chk->whoTo);
5071		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5072		SCTP_DECR_CHK_COUNT();
5073		/* sa_ignore FREED_MEMORY */
5074	}
5075	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5076		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5077		if (chk->data) {
5078			sctp_m_freem(chk->data);
5079			chk->data = NULL;
5080		}
5081		if (chk->holds_key_ref)
5082			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5083		sctp_free_remote_addr(chk->whoTo);
5084		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5085		SCTP_DECR_CHK_COUNT();
5086		/* sa_ignore FREED_MEMORY */
5087	}
5088
5089	if (asoc->mapping_array) {
5090		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5091		asoc->mapping_array = NULL;
5092	}
5093	if (asoc->nr_mapping_array) {
5094		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5095		asoc->nr_mapping_array = NULL;
5096	}
5097	/* the stream outs */
5098	if (asoc->strmout) {
5099		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5100		asoc->strmout = NULL;
5101	}
5102	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5103	if (asoc->strmin) {
5104		struct sctp_queued_to_read *ctl, *nctl;
5105
5106		for (i = 0; i < asoc->streamincnt; i++) {
5107			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5108				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5109				sctp_free_remote_addr(ctl->whoFrom);
5110				if (ctl->data) {
5111					sctp_m_freem(ctl->data);
5112					ctl->data = NULL;
5113				}
5114				/*
5115				 * We don't free the address here since all
5116				 * the net's were freed above.
5117				 */
5118				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5119				SCTP_DECR_READQ_COUNT();
5120			}
5121		}
5122		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5123		asoc->strmin = NULL;
5124	}
5125	asoc->streamincnt = 0;
5126	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5127#ifdef INVARIANTS
5128		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5129			panic("no net's left alloc'ed, or list points to itself");
5130		}
5131#endif
5132		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5133		sctp_free_remote_addr(net);
5134	}
5135	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5136		/* sa_ignore FREED_MEMORY */
5137		sctp_remove_laddr(laddr);
5138	}
5139
5140	/* pending asconf (address) parameters */
5141	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5142		/* sa_ignore FREED_MEMORY */
5143		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5144		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5145	}
5146	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5147		/* sa_ignore FREED_MEMORY */
5148		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5149		if (aack->data != NULL) {
5150			sctp_m_freem(aack->data);
5151		}
5152		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5153	}
5154	/* clean up auth stuff */
5155	if (asoc->local_hmacs)
5156		sctp_free_hmaclist(asoc->local_hmacs);
5157	if (asoc->peer_hmacs)
5158		sctp_free_hmaclist(asoc->peer_hmacs);
5159
5160	if (asoc->local_auth_chunks)
5161		sctp_free_chunklist(asoc->local_auth_chunks);
5162	if (asoc->peer_auth_chunks)
5163		sctp_free_chunklist(asoc->peer_auth_chunks);
5164
5165	sctp_free_authinfo(&asoc->authinfo);
5166
5167	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5168		LIST_REMOVE(shared_key, next);
5169		sctp_free_sharedkey(shared_key);
5170		/* sa_ignore FREED_MEMORY */
5171	}
5172
5173	/* Insert new items here :> */
5174
5175	/* Get rid of LOCK */
5176	SCTP_TCB_UNLOCK(stcb);
5177	SCTP_TCB_LOCK_DESTROY(stcb);
5178	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5179	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5180		SCTP_INP_INFO_WUNLOCK();
5181		SCTP_INP_RLOCK(inp);
5182	}
5183#ifdef SCTP_TRACK_FREED_ASOCS
5184	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5185		/* now clean up the tasoc itself */
5186		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5187		SCTP_DECR_ASOC_COUNT();
5188	} else {
5189		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5190	}
5191#else
5192	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5193	SCTP_DECR_ASOC_COUNT();
5194#endif
5195	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5196		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5197			/*
5198			 * If its NOT the inp_free calling us AND sctp_close
5199			 * as been called, we call back...
5200			 */
5201			SCTP_INP_RUNLOCK(inp);
5202			/*
5203			 * This will start the kill timer (if we are the
5204			 * last one) since we hold an increment yet. But
5205			 * this is the only safe way to do this since
5206			 * otherwise if the socket closes at the same time
5207			 * we are here we might collide in the cleanup.
5208			 */
5209			sctp_inpcb_free(inp,
5210			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5211			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5212			SCTP_INP_DECR_REF(inp);
5213			goto out_of;
5214		} else {
5215			/* The socket is still open. */
5216			SCTP_INP_DECR_REF(inp);
5217		}
5218	}
5219	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5220		SCTP_INP_RUNLOCK(inp);
5221	}
5222out_of:
5223	/* destroyed the asoc */
5224#ifdef SCTP_LOG_CLOSING
5225	sctp_log_closing(inp, NULL, 11);
5226#endif
5227	return (1);
5228}
5229
5230
5231
5232/*
5233 * determine if a destination is "reachable" based upon the addresses bound
5234 * to the current endpoint (e.g. only v4 or v6 currently bound)
5235 */
5236/*
5237 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5238 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5239 * types bound as its endpoint
5240 */
5241int
5242sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5243{
5244	struct sctp_inpcb *inp;
5245	int answer;
5246
5247	/*
5248	 * No locks here, the TCB, in all cases is already locked and an
5249	 * assoc is up. There is either a INP lock by the caller applied (in
5250	 * asconf case when deleting an address) or NOT in the HB case,
5251	 * however if HB then the INP increment is up and the INP will not
5252	 * be removed (on top of the fact that we have a TCB lock). So we
5253	 * only want to read the sctp_flags, which is either bound-all or
5254	 * not.. no protection needed since once an assoc is up you can't be
5255	 * changing your binding.
5256	 */
5257	inp = stcb->sctp_ep;
5258	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5259		/* if bound all, destination is not restricted */
5260		/*
5261		 * RRS: Question during lock work: Is this correct? If you
5262		 * are bound-all you still might need to obey the V4--V6
5263		 * flags??? IMO this bound-all stuff needs to be removed!
5264		 */
5265		return (1);
5266	}
5267	/* NOTE: all "scope" checks are done when local addresses are added */
5268	switch (destaddr->sa_family) {
5269#ifdef INET6
5270	case AF_INET6:
5271		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5272		break;
5273#endif
5274#ifdef INET
5275	case AF_INET:
5276		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5277		break;
5278#endif
5279	default:
5280		/* invalid family, so it's unreachable */
5281		answer = 0;
5282		break;
5283	}
5284	return (answer);
5285}
5286
5287/*
5288 * update the inp_vflags on an endpoint
5289 */
5290static void
5291sctp_update_ep_vflag(struct sctp_inpcb *inp)
5292{
5293	struct sctp_laddr *laddr;
5294
5295	/* first clear the flag */
5296	inp->ip_inp.inp.inp_vflag = 0;
5297	/* set the flag based on addresses on the ep list */
5298	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5299		if (laddr->ifa == NULL) {
5300			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5301			    __FUNCTION__);
5302			continue;
5303		}
5304		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5305			continue;
5306		}
5307		switch (laddr->ifa->address.sa.sa_family) {
5308#ifdef INET6
5309		case AF_INET6:
5310			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5311			break;
5312#endif
5313#ifdef INET
5314		case AF_INET:
5315			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5316			break;
5317#endif
5318		default:
5319			break;
5320		}
5321	}
5322}
5323
5324/*
5325 * Add the address to the endpoint local address list There is nothing to be
5326 * done if we are bound to all addresses
5327 */
5328void
5329sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5330{
5331	struct sctp_laddr *laddr;
5332	int fnd, error = 0;
5333
5334	fnd = 0;
5335
5336	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5337		/* You are already bound to all. You have it already */
5338		return;
5339	}
5340#ifdef INET6
5341	if (ifa->address.sa.sa_family == AF_INET6) {
5342		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5343			/* Can't bind a non-useable addr. */
5344			return;
5345		}
5346	}
5347#endif
5348	/* first, is it already present? */
5349	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5350		if (laddr->ifa == ifa) {
5351			fnd = 1;
5352			break;
5353		}
5354	}
5355
5356	if (fnd == 0) {
5357		/* Not in the ep list */
5358		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5359		if (error != 0)
5360			return;
5361		inp->laddr_count++;
5362		/* update inp_vflag flags */
5363		switch (ifa->address.sa.sa_family) {
5364#ifdef INET6
5365		case AF_INET6:
5366			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5367			break;
5368#endif
5369#ifdef INET
5370		case AF_INET:
5371			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5372			break;
5373#endif
5374		default:
5375			break;
5376		}
5377	}
5378	return;
5379}
5380
5381
5382/*
5383 * select a new (hopefully reachable) destination net (should only be used
5384 * when we deleted an ep addr that is the only usable source address to reach
5385 * the destination net)
5386 */
5387static void
5388sctp_select_primary_destination(struct sctp_tcb *stcb)
5389{
5390	struct sctp_nets *net;
5391
5392	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5393		/* for now, we'll just pick the first reachable one we find */
5394		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5395			continue;
5396		if (sctp_destination_is_reachable(stcb,
5397		    (struct sockaddr *)&net->ro._l_addr)) {
5398			/* found a reachable destination */
5399			stcb->asoc.primary_destination = net;
5400		}
5401	}
5402	/* I can't there from here! ...we're gonna die shortly... */
5403}
5404
5405
5406/*
5407 * Delete the address from the endpoint local address list There is nothing
5408 * to be done if we are bound to all addresses
5409 */
5410void
5411sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5412{
5413	struct sctp_laddr *laddr;
5414	int fnd;
5415
5416	fnd = 0;
5417	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5418		/* You are already bound to all. You have it already */
5419		return;
5420	}
5421	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5422		if (laddr->ifa == ifa) {
5423			fnd = 1;
5424			break;
5425		}
5426	}
5427	if (fnd && (inp->laddr_count < 2)) {
5428		/* can't delete unless there are at LEAST 2 addresses */
5429		return;
5430	}
5431	if (fnd) {
5432		/*
5433		 * clean up any use of this address go through our
5434		 * associations and clear any last_used_address that match
5435		 * this one for each assoc, see if a new primary_destination
5436		 * is needed
5437		 */
5438		struct sctp_tcb *stcb;
5439
5440		/* clean up "next_addr_touse" */
5441		if (inp->next_addr_touse == laddr)
5442			/* delete this address */
5443			inp->next_addr_touse = NULL;
5444
5445		/* clean up "last_used_address" */
5446		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5447			struct sctp_nets *net;
5448
5449			SCTP_TCB_LOCK(stcb);
5450			if (stcb->asoc.last_used_address == laddr)
5451				/* delete this address */
5452				stcb->asoc.last_used_address = NULL;
5453			/*
5454			 * Now spin through all the nets and purge any ref
5455			 * to laddr
5456			 */
5457			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5458				if (net->ro._s_addr &&
5459				    (net->ro._s_addr->ifa == laddr->ifa)) {
5460					/* Yep, purge src address selected */
5461					sctp_rtentry_t *rt;
5462
5463					/* delete this address if cached */
5464					rt = net->ro.ro_rt;
5465					if (rt != NULL) {
5466						RTFREE(rt);
5467						net->ro.ro_rt = NULL;
5468					}
5469					sctp_free_ifa(net->ro._s_addr);
5470					net->ro._s_addr = NULL;
5471					net->src_addr_selected = 0;
5472				}
5473			}
5474			SCTP_TCB_UNLOCK(stcb);
5475		}		/* for each tcb */
5476		/* remove it from the ep list */
5477		sctp_remove_laddr(laddr);
5478		inp->laddr_count--;
5479		/* update inp_vflag flags */
5480		sctp_update_ep_vflag(inp);
5481	}
5482	return;
5483}
5484
5485/*
5486 * Add the address to the TCB local address restricted list.
5487 * This is a "pending" address list (eg. addresses waiting for an
5488 * ASCONF-ACK response) and cannot be used as a valid source address.
5489 */
5490void
5491sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5492{
5493	struct sctp_laddr *laddr;
5494	struct sctpladdr *list;
5495
5496	/*
5497	 * Assumes TCB is locked.. and possibly the INP. May need to
5498	 * confirm/fix that if we need it and is not the case.
5499	 */
5500	list = &stcb->asoc.sctp_restricted_addrs;
5501
5502#ifdef INET6
5503	if (ifa->address.sa.sa_family == AF_INET6) {
5504		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5505			/* Can't bind a non-existent addr. */
5506			return;
5507		}
5508	}
5509#endif
5510	/* does the address already exist? */
5511	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5512		if (laddr->ifa == ifa) {
5513			return;
5514		}
5515	}
5516
5517	/* add to the list */
5518	(void)sctp_insert_laddr(list, ifa, 0);
5519	return;
5520}
5521
5522/*
5523 * insert an laddr entry with the given ifa for the desired list
5524 */
5525int
5526sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5527{
5528	struct sctp_laddr *laddr;
5529
5530	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5531	if (laddr == NULL) {
5532		/* out of memory? */
5533		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5534		return (EINVAL);
5535	}
5536	SCTP_INCR_LADDR_COUNT();
5537	bzero(laddr, sizeof(*laddr));
5538	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5539	laddr->ifa = ifa;
5540	laddr->action = act;
5541	atomic_add_int(&ifa->refcount, 1);
5542	/* insert it */
5543	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5544
5545	return (0);
5546}
5547
5548/*
5549 * Remove an laddr entry from the local address list (on an assoc)
5550 */
5551void
5552sctp_remove_laddr(struct sctp_laddr *laddr)
5553{
5554
5555	/* remove from the list */
5556	LIST_REMOVE(laddr, sctp_nxt_addr);
5557	sctp_free_ifa(laddr->ifa);
5558	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5559	SCTP_DECR_LADDR_COUNT();
5560}
5561
5562/*
5563 * Remove a local address from the TCB local address restricted list
5564 */
5565void
5566sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5567{
5568	struct sctp_inpcb *inp;
5569	struct sctp_laddr *laddr;
5570
5571	/*
5572	 * This is called by asconf work. It is assumed that a) The TCB is
5573	 * locked and b) The INP is locked. This is true in as much as I can
5574	 * trace through the entry asconf code where I did these locks.
5575	 * Again, the ASCONF code is a bit different in that it does lock
5576	 * the INP during its work often times. This must be since we don't
5577	 * want other proc's looking up things while what they are looking
5578	 * up is changing :-D
5579	 */
5580
5581	inp = stcb->sctp_ep;
5582	/* if subset bound and don't allow ASCONF's, can't delete last */
5583	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5584	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5585		if (stcb->sctp_ep->laddr_count < 2) {
5586			/* can't delete last address */
5587			return;
5588		}
5589	}
5590	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5591		/* remove the address if it exists */
5592		if (laddr->ifa == NULL)
5593			continue;
5594		if (laddr->ifa == ifa) {
5595			sctp_remove_laddr(laddr);
5596			return;
5597		}
5598	}
5599
5600	/* address not found! */
5601	return;
5602}
5603
5604/*
5605 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5606 */
5607/* sysctl */
5608static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5609static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5610
5611
5612
5613#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5614struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5615int *sctp_cpuarry = NULL;
5616void
5617sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5618{
5619	/* Queue a packet to a processor for the specified core */
5620	struct sctp_mcore_queue *qent;
5621	struct sctp_mcore_ctrl *wkq;
5622	int need_wake = 0;
5623
5624	if (sctp_mcore_workers == NULL) {
5625		/* Something went way bad during setup */
5626		sctp_input_with_port(m, off, 0);
5627		return;
5628	}
5629	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5630	    (sizeof(struct sctp_mcore_queue)),
5631	    SCTP_M_MCORE);
5632	if (qent == NULL) {
5633		/* This is trouble  */
5634		sctp_input_with_port(m, off, 0);
5635		return;
5636	}
5637	qent->vn = curvnet;
5638	qent->m = m;
5639	qent->off = off;
5640	qent->v6 = 0;
5641	wkq = &sctp_mcore_workers[cpu_to_use];
5642	SCTP_MCORE_QLOCK(wkq);
5643
5644	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5645	if (wkq->running == 0) {
5646		need_wake = 1;
5647	}
5648	SCTP_MCORE_QUNLOCK(wkq);
5649	if (need_wake) {
5650		wakeup(&wkq->running);
5651	}
5652}
5653
5654static void
5655sctp_mcore_thread(void *arg)
5656{
5657
5658	struct sctp_mcore_ctrl *wkq;
5659	struct sctp_mcore_queue *qent;
5660
5661	wkq = (struct sctp_mcore_ctrl *)arg;
5662	struct mbuf *m;
5663	int off, v6;
5664
5665	/* Wait for first tickle */
5666	SCTP_MCORE_LOCK(wkq);
5667	wkq->running = 0;
5668	msleep(&wkq->running,
5669	    &wkq->core_mtx,
5670	    0, "wait for pkt", 0);
5671	SCTP_MCORE_UNLOCK(wkq);
5672
5673	/* Bind to our cpu */
5674	thread_lock(curthread);
5675	sched_bind(curthread, wkq->cpuid);
5676	thread_unlock(curthread);
5677
5678	/* Now lets start working */
5679	SCTP_MCORE_LOCK(wkq);
5680	/* Now grab lock and go */
5681	for (;;) {
5682		SCTP_MCORE_QLOCK(wkq);
5683skip_sleep:
5684		wkq->running = 1;
5685		qent = TAILQ_FIRST(&wkq->que);
5686		if (qent) {
5687			TAILQ_REMOVE(&wkq->que, qent, next);
5688			SCTP_MCORE_QUNLOCK(wkq);
5689			CURVNET_SET(qent->vn);
5690			m = qent->m;
5691			off = qent->off;
5692			v6 = qent->v6;
5693			SCTP_FREE(qent, SCTP_M_MCORE);
5694			if (v6 == 0) {
5695				sctp_input_with_port(m, off, 0);
5696			} else {
5697				SCTP_PRINTF("V6 not yet supported\n");
5698				sctp_m_freem(m);
5699			}
5700			CURVNET_RESTORE();
5701			SCTP_MCORE_QLOCK(wkq);
5702		}
5703		wkq->running = 0;
5704		if (!TAILQ_EMPTY(&wkq->que)) {
5705			goto skip_sleep;
5706		}
5707		SCTP_MCORE_QUNLOCK(wkq);
5708		msleep(&wkq->running,
5709		    &wkq->core_mtx,
5710		    0, "wait for pkt", 0);
5711	}
5712}
5713
5714static void
5715sctp_startup_mcore_threads(void)
5716{
5717	int i, cpu;
5718
5719	if (mp_ncpus == 1)
5720		return;
5721
5722	if (sctp_mcore_workers != NULL) {
5723		/*
5724		 * Already been here in some previous vnet?
5725		 */
5726		return;
5727	}
5728	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5729	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5730	    SCTP_M_MCORE);
5731	if (sctp_mcore_workers == NULL) {
5732		/* TSNH I hope */
5733		return;
5734	}
5735	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5736	    sizeof(struct sctp_mcore_ctrl)));
5737	/* Init the structures */
5738	for (i = 0; i <= mp_maxid; i++) {
5739		TAILQ_INIT(&sctp_mcore_workers[i].que);
5740		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5741		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5742		sctp_mcore_workers[i].cpuid = i;
5743	}
5744	if (sctp_cpuarry == NULL) {
5745		SCTP_MALLOC(sctp_cpuarry, int *,
5746		    (mp_ncpus * sizeof(int)),
5747		    SCTP_M_MCORE);
5748		i = 0;
5749		CPU_FOREACH(cpu) {
5750			sctp_cpuarry[i] = cpu;
5751			i++;
5752		}
5753	}
5754	/* Now start them all */
5755	CPU_FOREACH(cpu) {
5756		(void)kproc_create(sctp_mcore_thread,
5757		    (void *)&sctp_mcore_workers[cpu],
5758		    &sctp_mcore_workers[cpu].thread_proc,
5759		    RFPROC,
5760		    SCTP_KTHREAD_PAGES,
5761		    SCTP_MCORE_NAME);
5762
5763	}
5764}
5765
5766#endif
5767
5768void
5769sctp_pcb_init()
5770{
5771	/*
5772	 * SCTP initialization for the PCB structures should be called by
5773	 * the sctp_init() funciton.
5774	 */
5775	int i;
5776	struct timeval tv;
5777
5778	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5779		/* error I was called twice */
5780		return;
5781	}
5782	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5783
5784#if defined(SCTP_LOCAL_TRACE_BUF)
5785	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5786#endif
5787#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5788	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5789	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5790	    SCTP_M_MCORE);
5791#endif
5792	(void)SCTP_GETTIME_TIMEVAL(&tv);
5793#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5794	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5795	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5796	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5797#else
5798	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5799	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5800	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5801#endif
5802	/* init the empty list of (All) Endpoints */
5803	LIST_INIT(&SCTP_BASE_INFO(listhead));
5804
5805
5806	/* init the hash table of endpoints */
5807	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5808	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5809	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5810	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5811	    &SCTP_BASE_INFO(hashasocmark));
5812	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5813	    &SCTP_BASE_INFO(hashmark));
5814	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5815	    &SCTP_BASE_INFO(hashtcpmark));
5816	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5817
5818
5819	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5820	    &SCTP_BASE_INFO(hashvrfmark));
5821
5822	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5823	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5824	/* init the zones */
5825	/*
5826	 * FIX ME: Should check for NULL returns, but if it does fail we are
5827	 * doomed to panic anyways... add later maybe.
5828	 */
5829	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5830	    sizeof(struct sctp_inpcb), maxsockets);
5831
5832	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5833	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5834
5835	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5836	    sizeof(struct sctp_laddr),
5837	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5838
5839	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5840	    sizeof(struct sctp_nets),
5841	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5842
5843	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5844	    sizeof(struct sctp_tmit_chunk),
5845	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5846
5847	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5848	    sizeof(struct sctp_queued_to_read),
5849	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5850
5851	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5852	    sizeof(struct sctp_stream_queue_pending),
5853	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5854
5855	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5856	    sizeof(struct sctp_asconf),
5857	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5858
5859	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5860	    sizeof(struct sctp_asconf_ack),
5861	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5862
5863
5864	/* Master Lock INIT for info structure */
5865	SCTP_INP_INFO_LOCK_INIT();
5866	SCTP_STATLOG_INIT_LOCK();
5867
5868	SCTP_IPI_COUNT_INIT();
5869	SCTP_IPI_ADDR_INIT();
5870#ifdef SCTP_PACKET_LOGGING
5871	SCTP_IP_PKTLOG_INIT();
5872#endif
5873	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5874
5875	SCTP_WQ_ADDR_INIT();
5876	/* not sure if we need all the counts */
5877	SCTP_BASE_INFO(ipi_count_ep) = 0;
5878	/* assoc/tcb zone info */
5879	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5880	/* local addrlist zone info */
5881	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5882	/* remote addrlist zone info */
5883	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5884	/* chunk info */
5885	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5886
5887	/* socket queue zone info */
5888	SCTP_BASE_INFO(ipi_count_readq) = 0;
5889
5890	/* stream out queue cont */
5891	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5892
5893	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5894	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5895
5896	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5897
5898	/* Init the TIMEWAIT list */
5899	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5900		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5901	}
5902	sctp_startup_iterator();
5903
5904#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5905	sctp_startup_mcore_threads();
5906#endif
5907
5908	/*
5909	 * INIT the default VRF which for BSD is the only one, other O/S's
5910	 * may have more. But initially they must start with one and then
5911	 * add the VRF's as addresses are added.
5912	 */
5913	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5914}
5915
5916/*
5917 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5918 */
5919void
5920sctp_pcb_finish(void)
5921{
5922	struct sctp_vrflist *vrf_bucket;
5923	struct sctp_vrf *vrf, *nvrf;
5924	struct sctp_ifn *ifn, *nifn;
5925	struct sctp_ifa *ifa, *nifa;
5926	struct sctpvtaghead *chain;
5927	struct sctp_tagblock *twait_block, *prev_twait_block;
5928	struct sctp_laddr *wi, *nwi;
5929	int i;
5930	struct sctp_iterator *it, *nit;
5931
5932	/*
5933	 * In FreeBSD the iterator thread never exits but we do clean up.
5934	 * The only way FreeBSD reaches here is if we have VRF's but we
5935	 * still add the ifdef to make it compile on old versions.
5936	 */
5937	SCTP_IPI_ITERATOR_WQ_LOCK();
5938	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5939		if (it->vn != curvnet) {
5940			continue;
5941		}
5942		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5943		if (it->function_atend != NULL) {
5944			(*it->function_atend) (it->pointer, it->val);
5945		}
5946		SCTP_FREE(it, SCTP_M_ITER);
5947	}
5948	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5949	SCTP_ITERATOR_LOCK();
5950	if ((sctp_it_ctl.cur_it) &&
5951	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5952		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5953	}
5954	SCTP_ITERATOR_UNLOCK();
5955	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5956	SCTP_WQ_ADDR_LOCK();
5957	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5958		LIST_REMOVE(wi, sctp_nxt_addr);
5959		SCTP_DECR_LADDR_COUNT();
5960		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5961	}
5962	SCTP_WQ_ADDR_UNLOCK();
5963
5964	/*
5965	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5966	 * destroyed first).
5967	 */
5968	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5969	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5970		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5971			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5972				/* free the ifa */
5973				LIST_REMOVE(ifa, next_bucket);
5974				LIST_REMOVE(ifa, next_ifa);
5975				SCTP_FREE(ifa, SCTP_M_IFA);
5976			}
5977			/* free the ifn */
5978			LIST_REMOVE(ifn, next_bucket);
5979			LIST_REMOVE(ifn, next_ifn);
5980			SCTP_FREE(ifn, SCTP_M_IFN);
5981		}
5982		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5983		/* free the vrf */
5984		LIST_REMOVE(vrf, next_vrf);
5985		SCTP_FREE(vrf, SCTP_M_VRF);
5986	}
5987	/* free the vrf hashes */
5988	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5989	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5990
5991	/*
5992	 * free the TIMEWAIT list elements malloc'd in the function
5993	 * sctp_add_vtag_to_timewait()...
5994	 */
5995	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5996		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5997		if (!LIST_EMPTY(chain)) {
5998			prev_twait_block = NULL;
5999			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6000				if (prev_twait_block) {
6001					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6002				}
6003				prev_twait_block = twait_block;
6004			}
6005			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6006		}
6007	}
6008
6009	/* free the locks and mutexes */
6010#ifdef SCTP_PACKET_LOGGING
6011	SCTP_IP_PKTLOG_DESTROY();
6012#endif
6013	SCTP_IPI_ADDR_DESTROY();
6014	SCTP_STATLOG_DESTROY();
6015	SCTP_INP_INFO_LOCK_DESTROY();
6016
6017	SCTP_WQ_ADDR_DESTROY();
6018
6019	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6020	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6021	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6022	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6023	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6024	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6025	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6026	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6027	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6028	/* Get rid of other stuff to */
6029	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
6030		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
6031	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
6032		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6033	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6034		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6035#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6036	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6037#endif
6038}
6039
6040
6041int
6042sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6043    int offset, int limit,
6044    struct sockaddr *src, struct sockaddr *dst,
6045    struct sockaddr *altsa)
6046{
6047	/*
6048	 * grub through the INIT pulling addresses and loading them to the
6049	 * nets structure in the asoc. The from address in the mbuf should
6050	 * also be loaded (if it is not already). This routine can be called
6051	 * with either INIT or INIT-ACK's as long as the m points to the IP
6052	 * packet and the offset points to the beginning of the parameters.
6053	 */
6054	struct sctp_inpcb *inp;
6055	struct sctp_nets *net, *nnet, *net_tmp;
6056	struct sctp_paramhdr *phdr, parm_buf;
6057	struct sctp_tcb *stcb_tmp;
6058	uint16_t ptype, plen;
6059	struct sockaddr *sa;
6060	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6061	struct sctp_auth_random *p_random = NULL;
6062	uint16_t random_len = 0;
6063	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6064	struct sctp_auth_hmac_algo *hmacs = NULL;
6065	uint16_t hmacs_len = 0;
6066	uint8_t saw_asconf = 0;
6067	uint8_t saw_asconf_ack = 0;
6068	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6069	struct sctp_auth_chunk_list *chunks = NULL;
6070	uint16_t num_chunks = 0;
6071	sctp_key_t *new_key;
6072	uint32_t keylen;
6073	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6074	uint8_t ecn_allowed;
6075
6076#ifdef INET
6077	struct sockaddr_in sin;
6078
6079#endif
6080#ifdef INET6
6081	struct sockaddr_in6 sin6;
6082
6083#endif
6084
6085	/* First get the destination address setup too. */
6086#ifdef INET
6087	memset(&sin, 0, sizeof(sin));
6088	sin.sin_family = AF_INET;
6089	sin.sin_len = sizeof(sin);
6090	sin.sin_port = stcb->rport;
6091#endif
6092#ifdef INET6
6093	memset(&sin6, 0, sizeof(sin6));
6094	sin6.sin6_family = AF_INET6;
6095	sin6.sin6_len = sizeof(struct sockaddr_in6);
6096	sin6.sin6_port = stcb->rport;
6097#endif
6098	if (altsa) {
6099		sa = altsa;
6100	} else {
6101		sa = src;
6102	}
6103	/* Turn off ECN until we get through all params */
6104	ecn_allowed = 0;
6105	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6106		/* mark all addresses that we have currently on the list */
6107		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6108	}
6109	/* does the source address already exist? if so skip it */
6110	inp = stcb->sctp_ep;
6111	atomic_add_int(&stcb->asoc.refcnt, 1);
6112	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6113	atomic_add_int(&stcb->asoc.refcnt, -1);
6114
6115	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6116		/* we must add the source address */
6117		/* no scope set here since we have a tcb already. */
6118		switch (sa->sa_family) {
6119#ifdef INET
6120		case AF_INET:
6121			if (stcb->asoc.scope.ipv4_addr_legal) {
6122				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6123					return (-1);
6124				}
6125			}
6126			break;
6127#endif
6128#ifdef INET6
6129		case AF_INET6:
6130			if (stcb->asoc.scope.ipv6_addr_legal) {
6131				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6132					return (-2);
6133				}
6134			}
6135			break;
6136#endif
6137		default:
6138			break;
6139		}
6140	} else {
6141		if (net_tmp != NULL && stcb_tmp == stcb) {
6142			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6143		} else if (stcb_tmp != stcb) {
6144			/* It belongs to another association? */
6145			if (stcb_tmp)
6146				SCTP_TCB_UNLOCK(stcb_tmp);
6147			return (-3);
6148		}
6149	}
6150	if (stcb->asoc.state == 0) {
6151		/* the assoc was freed? */
6152		return (-4);
6153	}
6154	/*
6155	 * peer must explicitly turn this on. This may have been initialized
6156	 * to be "on" in order to allow local addr changes while INIT's are
6157	 * in flight.
6158	 */
6159	stcb->asoc.peer_supports_asconf = 0;
6160	/* now we must go through each of the params. */
6161	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6162	while (phdr) {
6163		ptype = ntohs(phdr->param_type);
6164		plen = ntohs(phdr->param_length);
6165		/*
6166		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6167		 * (uint32_t)ptype, (int)plen);
6168		 */
6169		if (offset + plen > limit) {
6170			break;
6171		}
6172		if (plen == 0) {
6173			break;
6174		}
6175#ifdef INET
6176		if (ptype == SCTP_IPV4_ADDRESS) {
6177			if (stcb->asoc.scope.ipv4_addr_legal) {
6178				struct sctp_ipv4addr_param *p4, p4_buf;
6179
6180				/* ok get the v4 address and check/add */
6181				phdr = sctp_get_next_param(m, offset,
6182				    (struct sctp_paramhdr *)&p4_buf,
6183				    sizeof(p4_buf));
6184				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6185				    phdr == NULL) {
6186					return (-5);
6187				}
6188				p4 = (struct sctp_ipv4addr_param *)phdr;
6189				sin.sin_addr.s_addr = p4->addr;
6190				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6191					/* Skip multi-cast addresses */
6192					goto next_param;
6193				}
6194				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6195				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6196					goto next_param;
6197				}
6198				sa = (struct sockaddr *)&sin;
6199				inp = stcb->sctp_ep;
6200				atomic_add_int(&stcb->asoc.refcnt, 1);
6201				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6202				    dst, stcb);
6203				atomic_add_int(&stcb->asoc.refcnt, -1);
6204
6205				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6206				    inp == NULL) {
6207					/* we must add the source address */
6208					/*
6209					 * no scope set since we have a tcb
6210					 * already
6211					 */
6212
6213					/*
6214					 * we must validate the state again
6215					 * here
6216					 */
6217			add_it_now:
6218					if (stcb->asoc.state == 0) {
6219						/* the assoc was freed? */
6220						return (-7);
6221					}
6222					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6223						return (-8);
6224					}
6225				} else if (stcb_tmp == stcb) {
6226					if (stcb->asoc.state == 0) {
6227						/* the assoc was freed? */
6228						return (-10);
6229					}
6230					if (net != NULL) {
6231						/* clear flag */
6232						net->dest_state &=
6233						    ~SCTP_ADDR_NOT_IN_ASSOC;
6234					}
6235				} else {
6236					/*
6237					 * strange, address is in another
6238					 * assoc? straighten out locks.
6239					 */
6240					if (stcb_tmp) {
6241						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6242							/*
6243							 * in setup state we
6244							 * abort this guy
6245							 */
6246							sctp_abort_an_association(stcb_tmp->sctp_ep,
6247							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6248							goto add_it_now;
6249						}
6250						SCTP_TCB_UNLOCK(stcb_tmp);
6251					}
6252					if (stcb->asoc.state == 0) {
6253						/* the assoc was freed? */
6254						return (-12);
6255					}
6256					return (-13);
6257				}
6258			}
6259		} else
6260#endif
6261#ifdef INET6
6262		if (ptype == SCTP_IPV6_ADDRESS) {
6263			if (stcb->asoc.scope.ipv6_addr_legal) {
6264				/* ok get the v6 address and check/add */
6265				struct sctp_ipv6addr_param *p6, p6_buf;
6266
6267				phdr = sctp_get_next_param(m, offset,
6268				    (struct sctp_paramhdr *)&p6_buf,
6269				    sizeof(p6_buf));
6270				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6271				    phdr == NULL) {
6272					return (-14);
6273				}
6274				p6 = (struct sctp_ipv6addr_param *)phdr;
6275				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6276				    sizeof(p6->addr));
6277				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6278					/* Skip multi-cast addresses */
6279					goto next_param;
6280				}
6281				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6282					/*
6283					 * Link local make no sense without
6284					 * scope
6285					 */
6286					goto next_param;
6287				}
6288				sa = (struct sockaddr *)&sin6;
6289				inp = stcb->sctp_ep;
6290				atomic_add_int(&stcb->asoc.refcnt, 1);
6291				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6292				    dst, stcb);
6293				atomic_add_int(&stcb->asoc.refcnt, -1);
6294				if (stcb_tmp == NULL &&
6295				    (inp == stcb->sctp_ep || inp == NULL)) {
6296					/*
6297					 * we must validate the state again
6298					 * here
6299					 */
6300			add_it_now6:
6301					if (stcb->asoc.state == 0) {
6302						/* the assoc was freed? */
6303						return (-16);
6304					}
6305					/*
6306					 * we must add the address, no scope
6307					 * set
6308					 */
6309					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6310						return (-17);
6311					}
6312				} else if (stcb_tmp == stcb) {
6313					/*
6314					 * we must validate the state again
6315					 * here
6316					 */
6317					if (stcb->asoc.state == 0) {
6318						/* the assoc was freed? */
6319						return (-19);
6320					}
6321					if (net != NULL) {
6322						/* clear flag */
6323						net->dest_state &=
6324						    ~SCTP_ADDR_NOT_IN_ASSOC;
6325					}
6326				} else {
6327					/*
6328					 * strange, address is in another
6329					 * assoc? straighten out locks.
6330					 */
6331					if (stcb_tmp)
6332						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6333							/*
6334							 * in setup state we
6335							 * abort this guy
6336							 */
6337							sctp_abort_an_association(stcb_tmp->sctp_ep,
6338							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6339							goto add_it_now6;
6340						}
6341					SCTP_TCB_UNLOCK(stcb_tmp);
6342
6343					if (stcb->asoc.state == 0) {
6344						/* the assoc was freed? */
6345						return (-21);
6346					}
6347					return (-22);
6348				}
6349			}
6350		} else
6351#endif
6352		if (ptype == SCTP_ECN_CAPABLE) {
6353			ecn_allowed = 1;
6354		} else if (ptype == SCTP_ULP_ADAPTATION) {
6355			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6356				struct sctp_adaptation_layer_indication ai,
6357				                                *aip;
6358
6359				phdr = sctp_get_next_param(m, offset,
6360				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6361				aip = (struct sctp_adaptation_layer_indication *)phdr;
6362				if (aip) {
6363					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6364					stcb->asoc.adaptation_needed = 1;
6365				}
6366			}
6367		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6368			struct sctp_asconf_addr_param lstore, *fee;
6369			int lptype;
6370			struct sockaddr *lsa = NULL;
6371
6372#ifdef INET
6373			struct sctp_asconf_addrv4_param *fii;
6374
6375#endif
6376
6377			stcb->asoc.peer_supports_asconf = 1;
6378			if (plen > sizeof(lstore)) {
6379				return (-23);
6380			}
6381			phdr = sctp_get_next_param(m, offset,
6382			    (struct sctp_paramhdr *)&lstore,
6383			    min(plen, sizeof(lstore)));
6384			if (phdr == NULL) {
6385				return (-24);
6386			}
6387			fee = (struct sctp_asconf_addr_param *)phdr;
6388			lptype = ntohs(fee->addrp.ph.param_type);
6389			switch (lptype) {
6390#ifdef INET
6391			case SCTP_IPV4_ADDRESS:
6392				if (plen !=
6393				    sizeof(struct sctp_asconf_addrv4_param)) {
6394					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6395					    (int)sizeof(struct sctp_asconf_addrv4_param),
6396					    plen);
6397				} else {
6398					fii = (struct sctp_asconf_addrv4_param *)fee;
6399					sin.sin_addr.s_addr = fii->addrp.addr;
6400					lsa = (struct sockaddr *)&sin;
6401				}
6402				break;
6403#endif
6404#ifdef INET6
6405			case SCTP_IPV6_ADDRESS:
6406				if (plen !=
6407				    sizeof(struct sctp_asconf_addr_param)) {
6408					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6409					    (int)sizeof(struct sctp_asconf_addr_param),
6410					    plen);
6411				} else {
6412					memcpy(sin6.sin6_addr.s6_addr,
6413					    fee->addrp.addr,
6414					    sizeof(fee->addrp.addr));
6415					lsa = (struct sockaddr *)&sin6;
6416				}
6417				break;
6418#endif
6419			default:
6420				break;
6421			}
6422			if (lsa) {
6423				(void)sctp_set_primary_addr(stcb, sa, NULL);
6424			}
6425		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6426			stcb->asoc.peer_supports_nat = 1;
6427		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6428			/* Peer supports pr-sctp */
6429			stcb->asoc.peer_supports_prsctp = 1;
6430		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6431			/* A supported extension chunk */
6432			struct sctp_supported_chunk_types_param *pr_supported;
6433			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6434			int num_ent, i;
6435
6436			phdr = sctp_get_next_param(m, offset,
6437			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6438			if (phdr == NULL) {
6439				return (-25);
6440			}
6441			stcb->asoc.peer_supports_asconf = 0;
6442			stcb->asoc.peer_supports_prsctp = 0;
6443			stcb->asoc.peer_supports_pktdrop = 0;
6444			stcb->asoc.peer_supports_strreset = 0;
6445			stcb->asoc.peer_supports_nr_sack = 0;
6446			stcb->asoc.peer_supports_auth = 0;
6447			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6448			num_ent = plen - sizeof(struct sctp_paramhdr);
6449			for (i = 0; i < num_ent; i++) {
6450				switch (pr_supported->chunk_types[i]) {
6451				case SCTP_ASCONF:
6452				case SCTP_ASCONF_ACK:
6453					stcb->asoc.peer_supports_asconf = 1;
6454					break;
6455				case SCTP_FORWARD_CUM_TSN:
6456					stcb->asoc.peer_supports_prsctp = 1;
6457					break;
6458				case SCTP_PACKET_DROPPED:
6459					stcb->asoc.peer_supports_pktdrop = 1;
6460					break;
6461				case SCTP_NR_SELECTIVE_ACK:
6462					stcb->asoc.peer_supports_nr_sack = 1;
6463					break;
6464				case SCTP_STREAM_RESET:
6465					stcb->asoc.peer_supports_strreset = 1;
6466					break;
6467				case SCTP_AUTHENTICATION:
6468					stcb->asoc.peer_supports_auth = 1;
6469					break;
6470				default:
6471					/* one I have not learned yet */
6472					break;
6473
6474				}
6475			}
6476		} else if (ptype == SCTP_RANDOM) {
6477			if (plen > sizeof(random_store))
6478				break;
6479			if (got_random) {
6480				/* already processed a RANDOM */
6481				goto next_param;
6482			}
6483			phdr = sctp_get_next_param(m, offset,
6484			    (struct sctp_paramhdr *)random_store,
6485			    min(sizeof(random_store), plen));
6486			if (phdr == NULL)
6487				return (-26);
6488			p_random = (struct sctp_auth_random *)phdr;
6489			random_len = plen - sizeof(*p_random);
6490			/* enforce the random length */
6491			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6492				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6493				return (-27);
6494			}
6495			got_random = 1;
6496		} else if (ptype == SCTP_HMAC_LIST) {
6497			int num_hmacs;
6498			int i;
6499
6500			if (plen > sizeof(hmacs_store))
6501				break;
6502			if (got_hmacs) {
6503				/* already processed a HMAC list */
6504				goto next_param;
6505			}
6506			phdr = sctp_get_next_param(m, offset,
6507			    (struct sctp_paramhdr *)hmacs_store,
6508			    min(plen, sizeof(hmacs_store)));
6509			if (phdr == NULL)
6510				return (-28);
6511			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6512			hmacs_len = plen - sizeof(*hmacs);
6513			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6514			/* validate the hmac list */
6515			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6516				return (-29);
6517			}
6518			if (stcb->asoc.peer_hmacs != NULL)
6519				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6520			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6521			if (stcb->asoc.peer_hmacs != NULL) {
6522				for (i = 0; i < num_hmacs; i++) {
6523					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6524					    ntohs(hmacs->hmac_ids[i]));
6525				}
6526			}
6527			got_hmacs = 1;
6528		} else if (ptype == SCTP_CHUNK_LIST) {
6529			int i;
6530
6531			if (plen > sizeof(chunks_store))
6532				break;
6533			if (got_chklist) {
6534				/* already processed a Chunks list */
6535				goto next_param;
6536			}
6537			phdr = sctp_get_next_param(m, offset,
6538			    (struct sctp_paramhdr *)chunks_store,
6539			    min(plen, sizeof(chunks_store)));
6540			if (phdr == NULL)
6541				return (-30);
6542			chunks = (struct sctp_auth_chunk_list *)phdr;
6543			num_chunks = plen - sizeof(*chunks);
6544			if (stcb->asoc.peer_auth_chunks != NULL)
6545				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6546			else
6547				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6548			for (i = 0; i < num_chunks; i++) {
6549				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6550				    stcb->asoc.peer_auth_chunks);
6551				/* record asconf/asconf-ack if listed */
6552				if (chunks->chunk_types[i] == SCTP_ASCONF)
6553					saw_asconf = 1;
6554				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6555					saw_asconf_ack = 1;
6556
6557			}
6558			got_chklist = 1;
6559		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6560			    (ptype == SCTP_STATE_COOKIE) ||
6561			    (ptype == SCTP_UNRECOG_PARAM) ||
6562			    (ptype == SCTP_COOKIE_PRESERVE) ||
6563			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6564			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6565			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6566			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6567		    (ptype == SCTP_SUCCESS_REPORT)) {
6568			 /* don't care */ ;
6569		} else {
6570			if ((ptype & 0x8000) == 0x0000) {
6571				/*
6572				 * must stop processing the rest of the
6573				 * param's. Any report bits were handled
6574				 * with the call to
6575				 * sctp_arethere_unrecognized_parameters()
6576				 * when the INIT or INIT-ACK was first seen.
6577				 */
6578				break;
6579			}
6580		}
6581
6582next_param:
6583		offset += SCTP_SIZE32(plen);
6584		if (offset >= limit) {
6585			break;
6586		}
6587		phdr = sctp_get_next_param(m, offset, &parm_buf,
6588		    sizeof(parm_buf));
6589	}
6590	/* Now check to see if we need to purge any addresses */
6591	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6592		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6593		    SCTP_ADDR_NOT_IN_ASSOC) {
6594			/* This address has been removed from the asoc */
6595			/* remove and free it */
6596			stcb->asoc.numnets--;
6597			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6598			sctp_free_remote_addr(net);
6599			if (net == stcb->asoc.primary_destination) {
6600				stcb->asoc.primary_destination = NULL;
6601				sctp_select_primary_destination(stcb);
6602			}
6603		}
6604	}
6605	if (ecn_allowed == 0) {
6606		stcb->asoc.ecn_allowed = 0;
6607	}
6608	/* validate authentication required parameters */
6609	if (got_random && got_hmacs) {
6610		stcb->asoc.peer_supports_auth = 1;
6611	} else {
6612		stcb->asoc.peer_supports_auth = 0;
6613	}
6614	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6615		/* peer does not support auth but sent a chunks list? */
6616		return (-31);
6617	}
6618	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6619	    !stcb->asoc.peer_supports_auth) {
6620		/* peer supports asconf but not auth? */
6621		return (-32);
6622	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6623	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6624		return (-33);
6625	}
6626	/* concatenate the full random key */
6627	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6628	if (chunks != NULL) {
6629		keylen += sizeof(*chunks) + num_chunks;
6630	}
6631	new_key = sctp_alloc_key(keylen);
6632	if (new_key != NULL) {
6633		/* copy in the RANDOM */
6634		if (p_random != NULL) {
6635			keylen = sizeof(*p_random) + random_len;
6636			bcopy(p_random, new_key->key, keylen);
6637		}
6638		/* append in the AUTH chunks */
6639		if (chunks != NULL) {
6640			bcopy(chunks, new_key->key + keylen,
6641			    sizeof(*chunks) + num_chunks);
6642			keylen += sizeof(*chunks) + num_chunks;
6643		}
6644		/* append in the HMACs */
6645		if (hmacs != NULL) {
6646			bcopy(hmacs, new_key->key + keylen,
6647			    sizeof(*hmacs) + hmacs_len);
6648		}
6649	} else {
6650		/* failed to get memory for the key */
6651		return (-34);
6652	}
6653	if (stcb->asoc.authinfo.peer_random != NULL)
6654		sctp_free_key(stcb->asoc.authinfo.peer_random);
6655	stcb->asoc.authinfo.peer_random = new_key;
6656	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6657	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6658
6659	return (0);
6660}
6661
6662int
6663sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6664    struct sctp_nets *net)
6665{
6666	/* make sure the requested primary address exists in the assoc */
6667	if (net == NULL && sa)
6668		net = sctp_findnet(stcb, sa);
6669
6670	if (net == NULL) {
6671		/* didn't find the requested primary address! */
6672		return (-1);
6673	} else {
6674		/* set the primary address */
6675		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6676			/* Must be confirmed, so queue to set */
6677			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6678			return (0);
6679		}
6680		stcb->asoc.primary_destination = net;
6681		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6682			sctp_free_remote_addr(stcb->asoc.alternate);
6683			stcb->asoc.alternate = NULL;
6684		}
6685		net = TAILQ_FIRST(&stcb->asoc.nets);
6686		if (net != stcb->asoc.primary_destination) {
6687			/*
6688			 * first one on the list is NOT the primary
6689			 * sctp_cmpaddr() is much more efficient if the
6690			 * primary is the first on the list, make it so.
6691			 */
6692			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6693			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6694		}
6695		return (0);
6696	}
6697}
6698
6699int
6700sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6701{
6702	/*
6703	 * This function serves two purposes. It will see if a TAG can be
6704	 * re-used and return 1 for yes it is ok and 0 for don't use that
6705	 * tag. A secondary function it will do is purge out old tags that
6706	 * can be removed.
6707	 */
6708	struct sctpvtaghead *chain;
6709	struct sctp_tagblock *twait_block;
6710	struct sctpasochead *head;
6711	struct sctp_tcb *stcb;
6712	int i;
6713
6714	SCTP_INP_INFO_RLOCK();
6715	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6716	    SCTP_BASE_INFO(hashasocmark))];
6717	if (head == NULL) {
6718		/* invalid vtag */
6719		goto skip_vtag_check;
6720	}
6721	LIST_FOREACH(stcb, head, sctp_asocs) {
6722		/*
6723		 * We choose not to lock anything here. TCB's can't be
6724		 * removed since we have the read lock, so they can't be
6725		 * freed on us, same thing for the INP. I may be wrong with
6726		 * this assumption, but we will go with it for now :-)
6727		 */
6728		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6729			continue;
6730		}
6731		if (stcb->asoc.my_vtag == tag) {
6732			/* candidate */
6733			if (stcb->rport != rport) {
6734				continue;
6735			}
6736			if (stcb->sctp_ep->sctp_lport != lport) {
6737				continue;
6738			}
6739			/* Its a used tag set */
6740			SCTP_INP_INFO_RUNLOCK();
6741			return (0);
6742		}
6743	}
6744skip_vtag_check:
6745
6746	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6747	/* Now what about timed wait ? */
6748	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6749		/*
6750		 * Block(s) are present, lets see if we have this tag in the
6751		 * list
6752		 */
6753		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6754			if (twait_block->vtag_block[i].v_tag == 0) {
6755				/* not used */
6756				continue;
6757			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6758			    now->tv_sec) {
6759				/* Audit expires this guy */
6760				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6761				twait_block->vtag_block[i].v_tag = 0;
6762				twait_block->vtag_block[i].lport = 0;
6763				twait_block->vtag_block[i].rport = 0;
6764			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6765				    (twait_block->vtag_block[i].lport == lport) &&
6766			    (twait_block->vtag_block[i].rport == rport)) {
6767				/* Bad tag, sorry :< */
6768				SCTP_INP_INFO_RUNLOCK();
6769				return (0);
6770			}
6771		}
6772	}
6773	SCTP_INP_INFO_RUNLOCK();
6774	return (1);
6775}
6776
6777static void
6778sctp_drain_mbufs(struct sctp_tcb *stcb)
6779{
6780	/*
6781	 * We must hunt this association for MBUF's past the cumack (i.e.
6782	 * out of order data that we can renege on).
6783	 */
6784	struct sctp_association *asoc;
6785	struct sctp_tmit_chunk *chk, *nchk;
6786	uint32_t cumulative_tsn_p1;
6787	struct sctp_queued_to_read *ctl, *nctl;
6788	int cnt, strmat;
6789	uint32_t gap, i;
6790	int fnd = 0;
6791
6792	/* We look for anything larger than the cum-ack + 1 */
6793
6794	asoc = &stcb->asoc;
6795	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6796		/* none we can reneg on. */
6797		return;
6798	}
6799	SCTP_STAT_INCR(sctps_protocol_drains_done);
6800	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6801	cnt = 0;
6802	/* First look in the re-assembly queue */
6803	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6804		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6805			/* Yep it is above cum-ack */
6806			cnt++;
6807			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6808			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6809			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6810			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6811			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6812			if (chk->data) {
6813				sctp_m_freem(chk->data);
6814				chk->data = NULL;
6815			}
6816			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6817		}
6818	}
6819	/* Ok that was fun, now we will drain all the inbound streams? */
6820	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6821		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6822			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6823				/* Yep it is above cum-ack */
6824				cnt++;
6825				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6826				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6827				sctp_ucount_decr(asoc->cnt_on_all_streams);
6828				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6829				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6830				if (ctl->data) {
6831					sctp_m_freem(ctl->data);
6832					ctl->data = NULL;
6833				}
6834				sctp_free_remote_addr(ctl->whoFrom);
6835				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6836				SCTP_DECR_READQ_COUNT();
6837			}
6838		}
6839	}
6840	if (cnt) {
6841		/* We must back down to see what the new highest is */
6842		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6843			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6844			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6845				asoc->highest_tsn_inside_map = i;
6846				fnd = 1;
6847				break;
6848			}
6849		}
6850		if (!fnd) {
6851			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6852		}
6853		/*
6854		 * Question, should we go through the delivery queue? The
6855		 * only reason things are on here is the app not reading OR
6856		 * a p-d-api up. An attacker COULD send enough in to
6857		 * initiate the PD-API and then send a bunch of stuff to
6858		 * other streams... these would wind up on the delivery
6859		 * queue.. and then we would not get to them. But in order
6860		 * to do this I then have to back-track and un-deliver
6861		 * sequence numbers in streams.. el-yucko. I think for now
6862		 * we will NOT look at the delivery queue and leave it to be
6863		 * something to consider later. An alternative would be to
6864		 * abort the P-D-API with a notification and then deliver
6865		 * the data.... Or another method might be to keep track of
6866		 * how many times the situation occurs and if we see a
6867		 * possible attack underway just abort the association.
6868		 */
6869#ifdef SCTP_DEBUG
6870		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6871#endif
6872		/*
6873		 * Now do we need to find a new
6874		 * asoc->highest_tsn_inside_map?
6875		 */
6876		asoc->last_revoke_count = cnt;
6877		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6878		/* sa_ignore NO_NULL_CHK */
6879		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6880		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6881	}
6882	/*
6883	 * Another issue, in un-setting the TSN's in the mapping array we
6884	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6885	 * two things to occur. It may cause us to do extra work in checking
6886	 * for our mapping array movement. More importantly it may cause us
6887	 * to SACK every datagram. This may not be a bad thing though since
6888	 * we will recover once we get our cum-ack above and all this stuff
6889	 * we dumped recovered.
6890	 */
6891}
6892
6893void
6894sctp_drain()
6895{
6896	/*
6897	 * We must walk the PCB lists for ALL associations here. The system
6898	 * is LOW on MBUF's and needs help. This is where reneging will
6899	 * occur. We really hope this does NOT happen!
6900	 */
6901	VNET_ITERATOR_DECL(vnet_iter);
6902	VNET_LIST_RLOCK_NOSLEEP();
6903	VNET_FOREACH(vnet_iter) {
6904		CURVNET_SET(vnet_iter);
6905		struct sctp_inpcb *inp;
6906		struct sctp_tcb *stcb;
6907
6908		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6909		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6910#ifdef VIMAGE
6911			continue;
6912#else
6913			return;
6914#endif
6915		}
6916		SCTP_INP_INFO_RLOCK();
6917		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6918			/* For each endpoint */
6919			SCTP_INP_RLOCK(inp);
6920			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6921				/* For each association */
6922				SCTP_TCB_LOCK(stcb);
6923				sctp_drain_mbufs(stcb);
6924				SCTP_TCB_UNLOCK(stcb);
6925			}
6926			SCTP_INP_RUNLOCK(inp);
6927		}
6928		SCTP_INP_INFO_RUNLOCK();
6929		CURVNET_RESTORE();
6930	}
6931	VNET_LIST_RUNLOCK_NOSLEEP();
6932}
6933
6934/*
6935 * start a new iterator
6936 * iterates through all endpoints and associations based on the pcb_state
6937 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6938 * assocs and "ef" (optional) is executed when the iterator completes.
6939 * "inpf" (optional) is executed for each new endpoint as it is being
6940 * iterated through. inpe (optional) is called when the inp completes
6941 * its way through all the stcbs.
6942 */
6943int
6944sctp_initiate_iterator(inp_func inpf,
6945    asoc_func af,
6946    inp_func inpe,
6947    uint32_t pcb_state,
6948    uint32_t pcb_features,
6949    uint32_t asoc_state,
6950    void *argp,
6951    uint32_t argi,
6952    end_func ef,
6953    struct sctp_inpcb *s_inp,
6954    uint8_t chunk_output_off)
6955{
6956	struct sctp_iterator *it = NULL;
6957
6958	if (af == NULL) {
6959		return (-1);
6960	}
6961	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6962	    SCTP_M_ITER);
6963	if (it == NULL) {
6964		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6965		return (ENOMEM);
6966	}
6967	memset(it, 0, sizeof(*it));
6968	it->function_assoc = af;
6969	it->function_inp = inpf;
6970	if (inpf)
6971		it->done_current_ep = 0;
6972	else
6973		it->done_current_ep = 1;
6974	it->function_atend = ef;
6975	it->pointer = argp;
6976	it->val = argi;
6977	it->pcb_flags = pcb_state;
6978	it->pcb_features = pcb_features;
6979	it->asoc_state = asoc_state;
6980	it->function_inp_end = inpe;
6981	it->no_chunk_output = chunk_output_off;
6982	it->vn = curvnet;
6983	if (s_inp) {
6984		/* Assume lock is held here */
6985		it->inp = s_inp;
6986		SCTP_INP_INCR_REF(it->inp);
6987		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
6988	} else {
6989		SCTP_INP_INFO_RLOCK();
6990		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
6991		if (it->inp) {
6992			SCTP_INP_INCR_REF(it->inp);
6993		}
6994		SCTP_INP_INFO_RUNLOCK();
6995		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
6996
6997	}
6998	SCTP_IPI_ITERATOR_WQ_LOCK();
6999
7000	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7001	if (sctp_it_ctl.iterator_running == 0) {
7002		sctp_wakeup_iterator();
7003	}
7004	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7005	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7006	return (0);
7007}
7008