sctp_pcb.c revision 227655
1/*-
2 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2011, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2011, by Michael Tuexen. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * a) Redistributions of source code must retain the above copyright notice,
10 *   this list of conditions and the following disclaimer.
11 *
12 * b) Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *   the documentation and/or other materials provided with the distribution.
15 *
16 * c) Neither the name of Cisco Systems, Inc. nor the names of its
17 *    contributors may be used to endorse or promote products derived
18 *    from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33/* $KAME: sctp_pcb.c,v 1.38 2005/03/06 16:04:18 itojun Exp $	 */
34
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: head/sys/netinet/sctp_pcb.c 227655 2011-11-18 09:01:08Z tuexen $");
37
38#include <netinet/sctp_os.h>
39#include <sys/proc.h>
40#include <netinet/sctp_var.h>
41#include <netinet/sctp_sysctl.h>
42#include <netinet/sctp_pcb.h>
43#include <netinet/sctputil.h>
44#include <netinet/sctp.h>
45#include <netinet/sctp_header.h>
46#include <netinet/sctp_asconf.h>
47#include <netinet/sctp_output.h>
48#include <netinet/sctp_timer.h>
49#include <netinet/sctp_bsd_addr.h>
50#include <netinet/sctp_dtrace_define.h>
51#include <netinet/udp.h>
52#ifdef INET6
53#include <netinet6/ip6_var.h>
54#endif
55#include <sys/sched.h>
56#include <sys/smp.h>
57#include <sys/unistd.h>
58
59
60VNET_DEFINE(struct sctp_base_info, system_base_info);
61
62/* FIX: we don't handle multiple link local scopes */
63/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
64#ifdef INET6
65int
66SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
67{
68	struct sockaddr_in6 tmp_a, tmp_b;
69
70	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
71	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
72		return 0;
73	}
74	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
75	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
76		return 0;
77	}
78	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
79}
80
81#endif
82
83void
84sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
85{
86	/*
87	 * We really don't need to lock this, but I will just because it
88	 * does not hurt.
89	 */
90	SCTP_INP_INFO_RLOCK();
91	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
92	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
93	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
94	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
95	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
96	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
97	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
98	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
99
100	SCTP_INP_INFO_RUNLOCK();
101}
102
103/*
104 * Addresses are added to VRF's (Virtual Router's). For BSD we
105 * have only the default VRF 0. We maintain a hash list of
106 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
107 * these has a list of addresses. When we add a new address
108 * to a VRF we lookup the ifn/ifn_index, if the ifn does
109 * not exist we create it and add it to the list of IFN's
110 * within the VRF. Once we have the sctp_ifn, we add the
111 * address to the list. So we look something like:
112 *
113 * hash-vrf-table
114 *   vrf-> ifn-> ifn -> ifn
115 *   vrf    |
116 *    ...   +--ifa-> ifa -> ifa
117 *   vrf
118 *
119 * We keep these separate lists since the SCTP subsystem will
120 * point to these from its source address selection nets structure.
121 * When an address is deleted it does not happen right away on
122 * the SCTP side, it gets scheduled. What we do when a
123 * delete happens is immediately remove the address from
124 * the master list and decrement the refcount. As our
125 * addip iterator works through and frees the src address
126 * selection pointing to the sctp_ifa, eventually the refcount
127 * will reach 0 and we will delete it. Note that it is assumed
128 * that any locking on system level ifn/ifa is done at the
129 * caller of these functions and these routines will only
130 * lock the SCTP structures as they add or delete things.
131 *
132 * Other notes on VRF concepts.
133 *  - An endpoint can be in multiple VRF's
134 *  - An association lives within a VRF and only one VRF.
135 *  - Any incoming packet we can deduce the VRF for by
136 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
137 *  - Any downward send call or connect call must supply the
138 *    VRF via ancillary data or via some sort of set default
139 *    VRF socket option call (again for BSD no brainer since
140 *    the VRF is always 0).
141 *  - An endpoint may add multiple VRF's to it.
142 *  - Listening sockets can accept associations in any
143 *    of the VRF's they are in but the assoc will end up
144 *    in only one VRF (gotten from the packet or connect/send).
145 *
146 */
147
148struct sctp_vrf *
149sctp_allocate_vrf(int vrf_id)
150{
151	struct sctp_vrf *vrf = NULL;
152	struct sctp_vrflist *bucket;
153
154	/* First allocate the VRF structure */
155	vrf = sctp_find_vrf(vrf_id);
156	if (vrf) {
157		/* Already allocated */
158		return (vrf);
159	}
160	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
161	    SCTP_M_VRF);
162	if (vrf == NULL) {
163		/* No memory */
164#ifdef INVARIANTS
165		panic("No memory for VRF:%d", vrf_id);
166#endif
167		return (NULL);
168	}
169	/* setup the VRF */
170	memset(vrf, 0, sizeof(struct sctp_vrf));
171	vrf->vrf_id = vrf_id;
172	LIST_INIT(&vrf->ifnlist);
173	vrf->total_ifa_count = 0;
174	vrf->refcount = 0;
175	/* now also setup table ids */
176	SCTP_INIT_VRF_TABLEID(vrf);
177	/* Init the HASH of addresses */
178	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
179	    &vrf->vrf_addr_hashmark);
180	if (vrf->vrf_addr_hash == NULL) {
181		/* No memory */
182#ifdef INVARIANTS
183		panic("No memory for VRF:%d", vrf_id);
184#endif
185		SCTP_FREE(vrf, SCTP_M_VRF);
186		return (NULL);
187	}
188	/* Add it to the hash table */
189	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
190	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
191	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
192	return (vrf);
193}
194
195
196struct sctp_ifn *
197sctp_find_ifn(void *ifn, uint32_t ifn_index)
198{
199	struct sctp_ifn *sctp_ifnp;
200	struct sctp_ifnlist *hash_ifn_head;
201
202	/*
203	 * We assume the lock is held for the addresses if that's wrong
204	 * problems could occur :-)
205	 */
206	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
207	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
208		if (sctp_ifnp->ifn_index == ifn_index) {
209			return (sctp_ifnp);
210		}
211		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
212			return (sctp_ifnp);
213		}
214	}
215	return (NULL);
216}
217
218
219
220struct sctp_vrf *
221sctp_find_vrf(uint32_t vrf_id)
222{
223	struct sctp_vrflist *bucket;
224	struct sctp_vrf *liste;
225
226	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
227	LIST_FOREACH(liste, bucket, next_vrf) {
228		if (vrf_id == liste->vrf_id) {
229			return (liste);
230		}
231	}
232	return (NULL);
233}
234
235void
236sctp_free_vrf(struct sctp_vrf *vrf)
237{
238	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
239		if (vrf->vrf_addr_hash) {
240			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
241			vrf->vrf_addr_hash = NULL;
242		}
243		/* We zero'd the count */
244		LIST_REMOVE(vrf, next_vrf);
245		SCTP_FREE(vrf, SCTP_M_VRF);
246		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
247	}
248}
249
250void
251sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
252{
253	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
254		/* We zero'd the count */
255		if (sctp_ifnp->vrf) {
256			sctp_free_vrf(sctp_ifnp->vrf);
257		}
258		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
259		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
260	}
261}
262
263void
264sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
265{
266	struct sctp_ifn *sctp_ifnp;
267
268	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
269	if (sctp_ifnp != NULL) {
270		sctp_ifnp->ifn_mtu = mtu;
271	}
272}
273
274
275void
276sctp_free_ifa(struct sctp_ifa *sctp_ifap)
277{
278	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
279		/* We zero'd the count */
280		if (sctp_ifap->ifn_p) {
281			sctp_free_ifn(sctp_ifap->ifn_p);
282		}
283		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
284		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
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
310void
311sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
312    const char *if_name, uint32_t ifn_index)
313{
314	struct sctp_vrf *vrf;
315	struct sctp_ifa *sctp_ifap = NULL;
316
317	SCTP_IPI_ADDR_RLOCK();
318	vrf = sctp_find_vrf(vrf_id);
319	if (vrf == NULL) {
320		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
321		goto out;
322
323	}
324	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
325	if (sctp_ifap == NULL) {
326		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
327		goto out;
328	}
329	if (sctp_ifap->ifn_p == NULL) {
330		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
331		goto out;
332	}
333	if (if_name) {
334		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
335			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
336			    sctp_ifap->ifn_p->ifn_name, if_name);
337			goto out;
338		}
339	} else {
340		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
341			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
342			    sctp_ifap->ifn_p->ifn_index, ifn_index);
343			goto out;
344		}
345	}
346
347	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
348	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
349out:
350	SCTP_IPI_ADDR_RUNLOCK();
351}
352
353void
354sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
355    const char *if_name, uint32_t ifn_index)
356{
357	struct sctp_vrf *vrf;
358	struct sctp_ifa *sctp_ifap = NULL;
359
360	SCTP_IPI_ADDR_RLOCK();
361	vrf = sctp_find_vrf(vrf_id);
362	if (vrf == NULL) {
363		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
364		goto out;
365
366	}
367	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
368	if (sctp_ifap == NULL) {
369		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
370		goto out;
371	}
372	if (sctp_ifap->ifn_p == NULL) {
373		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
374		goto out;
375	}
376	if (if_name) {
377		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
378			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
379			    sctp_ifap->ifn_p->ifn_name, if_name);
380			goto out;
381		}
382	} else {
383		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
384			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
385			    sctp_ifap->ifn_p->ifn_index, ifn_index);
386			goto out;
387		}
388	}
389
390	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
391	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
392out:
393	SCTP_IPI_ADDR_RUNLOCK();
394}
395
396/*-
397 * Add an ifa to an ifn.
398 * Register the interface as necessary.
399 * NOTE: ADDR write lock MUST be held.
400 */
401static void
402sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
403{
404	int ifa_af;
405
406	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
407	sctp_ifap->ifn_p = sctp_ifnp;
408	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
409	/* update address counts */
410	sctp_ifnp->ifa_count++;
411	ifa_af = sctp_ifap->address.sa.sa_family;
412	switch (ifa_af) {
413#ifdef INET
414	case AF_INET:
415		sctp_ifnp->num_v4++;
416		break;
417#endif
418#ifdef INET6
419	case AF_INET6:
420		sctp_ifnp->num_v6++;
421		break;
422#endif
423	default:
424		break;
425	}
426	if (sctp_ifnp->ifa_count == 1) {
427		/* register the new interface */
428		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
429		sctp_ifnp->registered_af = ifa_af;
430	}
431}
432
433/*-
434 * Remove an ifa from its ifn.
435 * If no more addresses exist, remove the ifn too. Otherwise, re-register
436 * the interface based on the remaining address families left.
437 * NOTE: ADDR write lock MUST be held.
438 */
439static void
440sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
441{
442	uint32_t ifn_index;
443
444	LIST_REMOVE(sctp_ifap, next_ifa);
445	if (sctp_ifap->ifn_p) {
446		/* update address counts */
447		sctp_ifap->ifn_p->ifa_count--;
448		switch (sctp_ifap->address.sa.sa_family) {
449#ifdef INET
450		case AF_INET:
451			sctp_ifap->ifn_p->num_v4--;
452			break;
453#endif
454#ifdef INET6
455		case AF_INET6:
456			sctp_ifap->ifn_p->num_v6--;
457			break;
458#endif
459		default:
460			break;
461		}
462
463		ifn_index = sctp_ifap->ifn_p->ifn_index;
464		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
465			/* remove the ifn, possibly freeing it */
466			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
467		} else {
468			/* re-register address family type, if needed */
469			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
470			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
471				SCTP_DEREGISTER_INTERFACE(ifn_index, AF_INET6);
472				SCTP_REGISTER_INTERFACE(ifn_index, AF_INET);
473				sctp_ifap->ifn_p->registered_af = AF_INET;
474			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
475			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
476				SCTP_DEREGISTER_INTERFACE(ifn_index, AF_INET);
477				SCTP_REGISTER_INTERFACE(ifn_index, AF_INET6);
478				sctp_ifap->ifn_p->registered_af = AF_INET6;
479			}
480			/* free the ifn refcount */
481			sctp_free_ifn(sctp_ifap->ifn_p);
482		}
483		sctp_ifap->ifn_p = NULL;
484	}
485}
486
487struct sctp_ifa *
488sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
489    uint32_t ifn_type, const char *if_name, void *ifa,
490    struct sockaddr *addr, uint32_t ifa_flags,
491    int dynamic_add)
492{
493	struct sctp_vrf *vrf;
494	struct sctp_ifn *sctp_ifnp = NULL;
495	struct sctp_ifa *sctp_ifap = NULL;
496	struct sctp_ifalist *hash_addr_head;
497	struct sctp_ifnlist *hash_ifn_head;
498	uint32_t hash_of_addr;
499	int new_ifn_af = 0;
500
501#ifdef SCTP_DEBUG
502	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
503	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
504#endif
505	SCTP_IPI_ADDR_WLOCK();
506	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
507	if (sctp_ifnp) {
508		vrf = sctp_ifnp->vrf;
509	} else {
510		vrf = sctp_find_vrf(vrf_id);
511		if (vrf == NULL) {
512			vrf = sctp_allocate_vrf(vrf_id);
513			if (vrf == NULL) {
514				SCTP_IPI_ADDR_WUNLOCK();
515				return (NULL);
516			}
517		}
518	}
519	if (sctp_ifnp == NULL) {
520		/*
521		 * build one and add it, can't hold lock until after malloc
522		 * done though.
523		 */
524		SCTP_IPI_ADDR_WUNLOCK();
525		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
526		    sizeof(struct sctp_ifn), SCTP_M_IFN);
527		if (sctp_ifnp == NULL) {
528#ifdef INVARIANTS
529			panic("No memory for IFN");
530#endif
531			return (NULL);
532		}
533		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
534		sctp_ifnp->ifn_index = ifn_index;
535		sctp_ifnp->ifn_p = ifn;
536		sctp_ifnp->ifn_type = ifn_type;
537		sctp_ifnp->refcount = 0;
538		sctp_ifnp->vrf = vrf;
539		atomic_add_int(&vrf->refcount, 1);
540		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
541		if (if_name != NULL) {
542			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
543		} else {
544			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
545		}
546		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
547		LIST_INIT(&sctp_ifnp->ifalist);
548		SCTP_IPI_ADDR_WLOCK();
549		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
550		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
551		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
552		new_ifn_af = 1;
553	}
554	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
555	if (sctp_ifap) {
556		/* Hmm, it already exists? */
557		if ((sctp_ifap->ifn_p) &&
558		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
559			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
560			    sctp_ifap->ifn_p->ifn_name, ifn_index,
561			    sctp_ifap);
562			if (new_ifn_af) {
563				/* Remove the created one that we don't want */
564				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
565			}
566			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
567				/* easy to solve, just switch back to active */
568				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
569				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
570				sctp_ifap->ifn_p = sctp_ifnp;
571				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
572			}
573	exit_stage_left:
574			SCTP_IPI_ADDR_WUNLOCK();
575			return (sctp_ifap);
576		} else {
577			if (sctp_ifap->ifn_p) {
578				/*
579				 * The last IFN gets the address, remove the
580				 * old one
581				 */
582				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
583				    sctp_ifap, sctp_ifap->ifn_p->ifn_name,
584				    sctp_ifap->ifn_p->ifn_index, if_name,
585				    ifn_index);
586				/* remove the address from the old ifn */
587				sctp_remove_ifa_from_ifn(sctp_ifap);
588				/* move the address over to the new ifn */
589				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
590				goto exit_stage_left;
591			} else {
592				/* repair ifnp which was NULL ? */
593				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
594				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
595				    sctp_ifnp, sctp_ifap);
596				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
597			}
598			goto exit_stage_left;
599		}
600	}
601	SCTP_IPI_ADDR_WUNLOCK();
602	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
603	if (sctp_ifap == NULL) {
604#ifdef INVARIANTS
605		panic("No memory for IFA");
606#endif
607		return (NULL);
608	}
609	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
610	sctp_ifap->ifn_p = sctp_ifnp;
611	atomic_add_int(&sctp_ifnp->refcount, 1);
612	sctp_ifap->vrf_id = vrf_id;
613	sctp_ifap->ifa = ifa;
614	memcpy(&sctp_ifap->address, addr, addr->sa_len);
615	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
616	sctp_ifap->flags = ifa_flags;
617	/* Set scope */
618	switch (sctp_ifap->address.sa.sa_family) {
619#ifdef INET
620	case AF_INET:
621		{
622			struct sockaddr_in *sin;
623
624			sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
625			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
626			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
627				sctp_ifap->src_is_loop = 1;
628			}
629			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
630				sctp_ifap->src_is_priv = 1;
631			}
632			sctp_ifnp->num_v4++;
633			if (new_ifn_af)
634				new_ifn_af = AF_INET;
635			break;
636		}
637#endif
638#ifdef INET6
639	case AF_INET6:
640		{
641			/* ok to use deprecated addresses? */
642			struct sockaddr_in6 *sin6;
643
644			sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
645			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
646			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
647				sctp_ifap->src_is_loop = 1;
648			}
649			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
650				sctp_ifap->src_is_priv = 1;
651			}
652			sctp_ifnp->num_v6++;
653			if (new_ifn_af)
654				new_ifn_af = AF_INET6;
655			break;
656		}
657#endif
658	default:
659		new_ifn_af = 0;
660		break;
661	}
662	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
663
664	if ((sctp_ifap->src_is_priv == 0) &&
665	    (sctp_ifap->src_is_loop == 0)) {
666		sctp_ifap->src_is_glob = 1;
667	}
668	SCTP_IPI_ADDR_WLOCK();
669	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
670	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
671	sctp_ifap->refcount = 1;
672	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
673	sctp_ifnp->ifa_count++;
674	vrf->total_ifa_count++;
675	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
676	if (new_ifn_af) {
677		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
678		sctp_ifnp->registered_af = new_ifn_af;
679	}
680	SCTP_IPI_ADDR_WUNLOCK();
681	if (dynamic_add) {
682		/*
683		 * Bump up the refcount so that when the timer completes it
684		 * will drop back down.
685		 */
686		struct sctp_laddr *wi;
687
688		atomic_add_int(&sctp_ifap->refcount, 1);
689		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
690		if (wi == NULL) {
691			/*
692			 * Gak, what can we do? We have lost an address
693			 * change can you say HOSED?
694			 */
695			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
696			/* Opps, must decrement the count */
697			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
698			    if_name);
699			return (NULL);
700		}
701		SCTP_INCR_LADDR_COUNT();
702		bzero(wi, sizeof(*wi));
703		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
704		wi->ifa = sctp_ifap;
705		wi->action = SCTP_ADD_IP_ADDRESS;
706
707		SCTP_WQ_ADDR_LOCK();
708		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
709		SCTP_WQ_ADDR_UNLOCK();
710
711		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
712		    (struct sctp_inpcb *)NULL,
713		    (struct sctp_tcb *)NULL,
714		    (struct sctp_nets *)NULL);
715	} else {
716		/* it's ready for use */
717		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
718	}
719	return (sctp_ifap);
720}
721
722void
723sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
724    uint32_t ifn_index, const char *if_name)
725{
726	struct sctp_vrf *vrf;
727	struct sctp_ifa *sctp_ifap = NULL;
728
729	SCTP_IPI_ADDR_WLOCK();
730	vrf = sctp_find_vrf(vrf_id);
731	if (vrf == NULL) {
732		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
733		goto out_now;
734	}
735#ifdef SCTP_DEBUG
736	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
737	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
738#endif
739	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
740	if (sctp_ifap) {
741		/* Validate the delete */
742		if (sctp_ifap->ifn_p) {
743			int valid = 0;
744
745			/*-
746			 * The name has priority over the ifn_index
747			 * if its given. We do this especially for
748			 * panda who might recycle indexes fast.
749			 */
750			if (if_name) {
751				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
752					/* They match its a correct delete */
753					valid = 1;
754				}
755			}
756			if (!valid) {
757				/* last ditch check ifn_index */
758				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
759					valid = 1;
760				}
761			}
762			if (!valid) {
763				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
764				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
765				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
766				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
767				SCTP_IPI_ADDR_WUNLOCK();
768				return;
769			}
770		}
771		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", sctp_ifap);
772		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
773		sctp_ifap->localifa_flags |= SCTP_BEING_DELETED;
774		vrf->total_ifa_count--;
775		LIST_REMOVE(sctp_ifap, next_bucket);
776		sctp_remove_ifa_from_ifn(sctp_ifap);
777	}
778#ifdef SCTP_DEBUG
779	else {
780		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
781		    ifn_index);
782		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
783	}
784#endif
785
786out_now:
787	SCTP_IPI_ADDR_WUNLOCK();
788	if (sctp_ifap) {
789		struct sctp_laddr *wi;
790
791		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
792		if (wi == NULL) {
793			/*
794			 * Gak, what can we do? We have lost an address
795			 * change can you say HOSED?
796			 */
797			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
798
799			/* Oops, must decrement the count */
800			sctp_free_ifa(sctp_ifap);
801			return;
802		}
803		SCTP_INCR_LADDR_COUNT();
804		bzero(wi, sizeof(*wi));
805		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
806		wi->ifa = sctp_ifap;
807		wi->action = SCTP_DEL_IP_ADDRESS;
808		SCTP_WQ_ADDR_LOCK();
809		/*
810		 * Should this really be a tailq? As it is we will process
811		 * the newest first :-0
812		 */
813		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
814		SCTP_WQ_ADDR_UNLOCK();
815
816		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
817		    (struct sctp_inpcb *)NULL,
818		    (struct sctp_tcb *)NULL,
819		    (struct sctp_nets *)NULL);
820	}
821	return;
822}
823
824
825static struct sctp_tcb *
826sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
827    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
828{
829	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
830	/*
831	 * If we support the TCP model, then we must now dig through to see
832	 * if we can find our endpoint in the list of tcp ep's.
833	 */
834	uint16_t lport, rport;
835	struct sctppcbhead *ephead;
836	struct sctp_inpcb *inp;
837	struct sctp_laddr *laddr;
838	struct sctp_tcb *stcb;
839	struct sctp_nets *net;
840
841	if ((to == NULL) || (from == NULL)) {
842		return (NULL);
843	}
844	switch (to->sa_family) {
845#ifdef INET
846	case AF_INET:
847		if (from->sa_family == AF_INET) {
848			lport = ((struct sockaddr_in *)to)->sin_port;
849			rport = ((struct sockaddr_in *)from)->sin_port;
850		} else {
851			return (NULL);
852		}
853		break;
854#endif
855#ifdef INET6
856	case AF_INET6:
857		if (from->sa_family == AF_INET6) {
858			lport = ((struct sockaddr_in6 *)to)->sin6_port;
859			rport = ((struct sockaddr_in6 *)from)->sin6_port;
860		} else {
861			return (NULL);
862		}
863		break;
864#endif
865	default:
866		return (NULL);
867	}
868	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
869	/*
870	 * Ok now for each of the guys in this bucket we must look and see:
871	 * - Does the remote port match. - Does there single association's
872	 * addresses match this address (to). If so we update p_ep to point
873	 * to this ep and return the tcb from it.
874	 */
875	LIST_FOREACH(inp, ephead, sctp_hash) {
876		SCTP_INP_RLOCK(inp);
877		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
878			SCTP_INP_RUNLOCK(inp);
879			continue;
880		}
881		if (lport != inp->sctp_lport) {
882			SCTP_INP_RUNLOCK(inp);
883			continue;
884		}
885		if (inp->def_vrf_id != vrf_id) {
886			SCTP_INP_RUNLOCK(inp);
887			continue;
888		}
889		/* check to see if the ep has one of the addresses */
890		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
891			/* We are NOT bound all, so look further */
892			int match = 0;
893
894			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
895
896				if (laddr->ifa == NULL) {
897					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
898					continue;
899				}
900				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
901					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
902					continue;
903				}
904				if (laddr->ifa->address.sa.sa_family ==
905				    to->sa_family) {
906					/* see if it matches */
907
908#ifdef INET
909					if (from->sa_family == AF_INET) {
910						struct sockaddr_in *intf_addr,
911						           *sin;
912
913						intf_addr = &laddr->ifa->address.sin;
914						sin = (struct sockaddr_in *)to;
915						if (sin->sin_addr.s_addr ==
916						    intf_addr->sin_addr.s_addr) {
917							match = 1;
918							break;
919						}
920					}
921#endif
922#ifdef INET6
923					if (from->sa_family == AF_INET6) {
924						struct sockaddr_in6 *intf_addr6;
925						struct sockaddr_in6 *sin6;
926
927						sin6 = (struct sockaddr_in6 *)
928						    to;
929						intf_addr6 = &laddr->ifa->address.sin6;
930
931						if (SCTP6_ARE_ADDR_EQUAL(sin6,
932						    intf_addr6)) {
933							match = 1;
934							break;
935						}
936					}
937#endif
938				}
939			}
940			if (match == 0) {
941				/* This endpoint does not have this address */
942				SCTP_INP_RUNLOCK(inp);
943				continue;
944			}
945		}
946		/*
947		 * Ok if we hit here the ep has the address, does it hold
948		 * the tcb?
949		 */
950
951		stcb = LIST_FIRST(&inp->sctp_asoc_list);
952		if (stcb == NULL) {
953			SCTP_INP_RUNLOCK(inp);
954			continue;
955		}
956		SCTP_TCB_LOCK(stcb);
957		if (stcb->rport != rport) {
958			/* remote port does not match. */
959			SCTP_TCB_UNLOCK(stcb);
960			SCTP_INP_RUNLOCK(inp);
961			continue;
962		}
963		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
964			SCTP_TCB_UNLOCK(stcb);
965			SCTP_INP_RUNLOCK(inp);
966			continue;
967		}
968		/* Does this TCB have a matching address? */
969		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
970
971			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
972				/* not the same family, can't be a match */
973				continue;
974			}
975			switch (from->sa_family) {
976#ifdef INET
977			case AF_INET:
978				{
979					struct sockaddr_in *sin, *rsin;
980
981					sin = (struct sockaddr_in *)&net->ro._l_addr;
982					rsin = (struct sockaddr_in *)from;
983					if (sin->sin_addr.s_addr ==
984					    rsin->sin_addr.s_addr) {
985						/* found it */
986						if (netp != NULL) {
987							*netp = net;
988						}
989						/*
990						 * Update the endpoint
991						 * pointer
992						 */
993						*inp_p = inp;
994						SCTP_INP_RUNLOCK(inp);
995						return (stcb);
996					}
997					break;
998				}
999#endif
1000#ifdef INET6
1001			case AF_INET6:
1002				{
1003					struct sockaddr_in6 *sin6, *rsin6;
1004
1005					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1006					rsin6 = (struct sockaddr_in6 *)from;
1007					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1008					    rsin6)) {
1009						/* found it */
1010						if (netp != NULL) {
1011							*netp = net;
1012						}
1013						/*
1014						 * Update the endpoint
1015						 * pointer
1016						 */
1017						*inp_p = inp;
1018						SCTP_INP_RUNLOCK(inp);
1019						return (stcb);
1020					}
1021					break;
1022				}
1023#endif
1024			default:
1025				/* TSNH */
1026				break;
1027			}
1028		}
1029		SCTP_TCB_UNLOCK(stcb);
1030		SCTP_INP_RUNLOCK(inp);
1031	}
1032	return (NULL);
1033}
1034
1035static int
1036sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
1037{
1038	int loopback_scope, ipv4_local_scope, local_scope, site_scope;
1039	int ipv4_addr_legal, ipv6_addr_legal;
1040	struct sctp_vrf *vrf;
1041	struct sctp_ifn *sctp_ifn;
1042	struct sctp_ifa *sctp_ifa;
1043
1044	loopback_scope = stcb->asoc.loopback_scope;
1045	ipv4_local_scope = stcb->asoc.ipv4_local_scope;
1046	local_scope = stcb->asoc.local_scope;
1047	site_scope = stcb->asoc.site_scope;
1048	ipv4_addr_legal = ipv6_addr_legal = 0;
1049	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1050		ipv6_addr_legal = 1;
1051		if (SCTP_IPV6_V6ONLY(stcb->sctp_ep) == 0) {
1052			ipv4_addr_legal = 1;
1053		}
1054	} else {
1055		ipv4_addr_legal = 1;
1056	}
1057
1058	SCTP_IPI_ADDR_RLOCK();
1059	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
1060	if (vrf == NULL) {
1061		/* no vrf, no addresses */
1062		SCTP_IPI_ADDR_RUNLOCK();
1063		return (0);
1064	}
1065	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
1066		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
1067			if ((loopback_scope == 0) &&
1068			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
1069				continue;
1070			}
1071			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
1072				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
1073				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
1074					/*
1075					 * We allow pending addresses, where
1076					 * we have sent an asconf-add to be
1077					 * considered valid.
1078					 */
1079					continue;
1080				}
1081				switch (sctp_ifa->address.sa.sa_family) {
1082#ifdef INET
1083				case AF_INET:
1084					if (ipv4_addr_legal) {
1085						struct sockaddr_in *sin,
1086						           *rsin;
1087
1088						sin = &sctp_ifa->address.sin;
1089						rsin = (struct sockaddr_in *)to;
1090						if ((ipv4_local_scope == 0) &&
1091						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
1092							continue;
1093						}
1094						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
1095							SCTP_IPI_ADDR_RUNLOCK();
1096							return (1);
1097						}
1098					}
1099					break;
1100#endif
1101#ifdef INET6
1102				case AF_INET6:
1103					if (ipv6_addr_legal) {
1104						struct sockaddr_in6 *sin6,
1105						            *rsin6;
1106
1107						sin6 = &sctp_ifa->address.sin6;
1108						rsin6 = (struct sockaddr_in6 *)to;
1109						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
1110							if (local_scope == 0)
1111								continue;
1112							if (sin6->sin6_scope_id == 0) {
1113								if (sa6_recoverscope(sin6) != 0)
1114									continue;
1115							}
1116						}
1117						if ((site_scope == 0) &&
1118						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
1119							continue;
1120						}
1121						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
1122							SCTP_IPI_ADDR_RUNLOCK();
1123							return (1);
1124						}
1125					}
1126					break;
1127#endif
1128				default:
1129					/* TSNH */
1130					break;
1131				}
1132			}
1133		}
1134	} else {
1135		struct sctp_laddr *laddr;
1136
1137		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
1138			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
1139			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
1140				/*
1141				 * We allow pending addresses, where we have
1142				 * sent an asconf-add to be considered
1143				 * valid.
1144				 */
1145				continue;
1146			}
1147			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
1148				continue;
1149			}
1150			switch (to->sa_family) {
1151#ifdef INET
1152			case AF_INET:
1153				{
1154					struct sockaddr_in *sin, *rsin;
1155
1156					sin = (struct sockaddr_in *)&laddr->ifa->address.sin;
1157					rsin = (struct sockaddr_in *)to;
1158					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
1159						SCTP_IPI_ADDR_RUNLOCK();
1160						return (1);
1161					}
1162					break;
1163				}
1164#endif
1165#ifdef INET6
1166			case AF_INET6:
1167				{
1168					struct sockaddr_in6 *sin6, *rsin6;
1169
1170					sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
1171					rsin6 = (struct sockaddr_in6 *)to;
1172					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
1173						SCTP_IPI_ADDR_RUNLOCK();
1174						return (1);
1175					}
1176					break;
1177				}
1178
1179#endif
1180			default:
1181				/* TSNH */
1182				break;
1183			}
1184
1185		}
1186	}
1187	SCTP_IPI_ADDR_RUNLOCK();
1188	return (0);
1189}
1190
1191/*
1192 * rules for use
1193 *
1194 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1195 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1196 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1197 * NULL.
1198 */
1199
1200struct sctp_tcb *
1201sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1202    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1203{
1204	struct sctpasochead *head;
1205	struct sctp_inpcb *inp;
1206	struct sctp_tcb *stcb = NULL;
1207	struct sctp_nets *net;
1208	uint16_t rport;
1209
1210	inp = *inp_p;
1211	if (remote->sa_family == AF_INET) {
1212		rport = (((struct sockaddr_in *)remote)->sin_port);
1213	} else if (remote->sa_family == AF_INET6) {
1214		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1215	} else {
1216		return (NULL);
1217	}
1218	if (locked_tcb) {
1219		/*
1220		 * UN-lock so we can do proper locking here this occurs when
1221		 * called from load_addresses_from_init.
1222		 */
1223		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1224		SCTP_TCB_UNLOCK(locked_tcb);
1225	}
1226	SCTP_INP_INFO_RLOCK();
1227	if (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
1228		/*-
1229		 * Now either this guy is our listener or it's the
1230		 * connector. If it is the one that issued the connect, then
1231		 * it's only chance is to be the first TCB in the list. If
1232		 * it is the acceptor, then do the special_lookup to hash
1233		 * and find the real inp.
1234		 */
1235		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1236			/* to is peer addr, from is my addr */
1237			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1238			    netp, inp->def_vrf_id);
1239			if ((stcb != NULL) && (locked_tcb == NULL)) {
1240				/* we have a locked tcb, lower refcount */
1241				SCTP_INP_DECR_REF(inp);
1242			}
1243			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1244				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1245				SCTP_TCB_LOCK(locked_tcb);
1246				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1247				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1248			}
1249			SCTP_INP_INFO_RUNLOCK();
1250			return (stcb);
1251		} else {
1252			SCTP_INP_WLOCK(inp);
1253			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1254				goto null_return;
1255			}
1256			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1257			if (stcb == NULL) {
1258				goto null_return;
1259			}
1260			SCTP_TCB_LOCK(stcb);
1261
1262			if (stcb->rport != rport) {
1263				/* remote port does not match. */
1264				SCTP_TCB_UNLOCK(stcb);
1265				goto null_return;
1266			}
1267			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1268				SCTP_TCB_UNLOCK(stcb);
1269				goto null_return;
1270			}
1271			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1272				SCTP_TCB_UNLOCK(stcb);
1273				goto null_return;
1274			}
1275			/* now look at the list of remote addresses */
1276			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1277#ifdef INVARIANTS
1278				if (net == (TAILQ_NEXT(net, sctp_next))) {
1279					panic("Corrupt net list");
1280				}
1281#endif
1282				if (net->ro._l_addr.sa.sa_family !=
1283				    remote->sa_family) {
1284					/* not the same family */
1285					continue;
1286				}
1287				switch (remote->sa_family) {
1288#ifdef INET
1289				case AF_INET:
1290					{
1291						struct sockaddr_in *sin,
1292						           *rsin;
1293
1294						sin = (struct sockaddr_in *)
1295						    &net->ro._l_addr;
1296						rsin = (struct sockaddr_in *)remote;
1297						if (sin->sin_addr.s_addr ==
1298						    rsin->sin_addr.s_addr) {
1299							/* found it */
1300							if (netp != NULL) {
1301								*netp = net;
1302							}
1303							if (locked_tcb == NULL) {
1304								SCTP_INP_DECR_REF(inp);
1305							} else if (locked_tcb != stcb) {
1306								SCTP_TCB_LOCK(locked_tcb);
1307							}
1308							if (locked_tcb) {
1309								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1310							}
1311							SCTP_INP_WUNLOCK(inp);
1312							SCTP_INP_INFO_RUNLOCK();
1313							return (stcb);
1314						}
1315						break;
1316					}
1317#endif
1318#ifdef INET6
1319				case AF_INET6:
1320					{
1321						struct sockaddr_in6 *sin6,
1322						            *rsin6;
1323
1324						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1325						rsin6 = (struct sockaddr_in6 *)remote;
1326						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1327						    rsin6)) {
1328							/* found it */
1329							if (netp != NULL) {
1330								*netp = net;
1331							}
1332							if (locked_tcb == NULL) {
1333								SCTP_INP_DECR_REF(inp);
1334							} else if (locked_tcb != stcb) {
1335								SCTP_TCB_LOCK(locked_tcb);
1336							}
1337							if (locked_tcb) {
1338								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1339							}
1340							SCTP_INP_WUNLOCK(inp);
1341							SCTP_INP_INFO_RUNLOCK();
1342							return (stcb);
1343						}
1344						break;
1345					}
1346#endif
1347				default:
1348					/* TSNH */
1349					break;
1350				}
1351			}
1352			SCTP_TCB_UNLOCK(stcb);
1353		}
1354	} else {
1355		SCTP_INP_WLOCK(inp);
1356		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1357			goto null_return;
1358		}
1359		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1360		    inp->sctp_hashmark)];
1361		if (head == NULL) {
1362			goto null_return;
1363		}
1364		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1365			if (stcb->rport != rport) {
1366				/* remote port does not match */
1367				continue;
1368			}
1369			SCTP_TCB_LOCK(stcb);
1370			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1371				SCTP_TCB_UNLOCK(stcb);
1372				continue;
1373			}
1374			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1375				SCTP_TCB_UNLOCK(stcb);
1376				continue;
1377			}
1378			/* now look at the list of remote addresses */
1379			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1380#ifdef INVARIANTS
1381				if (net == (TAILQ_NEXT(net, sctp_next))) {
1382					panic("Corrupt net list");
1383				}
1384#endif
1385				if (net->ro._l_addr.sa.sa_family !=
1386				    remote->sa_family) {
1387					/* not the same family */
1388					continue;
1389				}
1390				switch (remote->sa_family) {
1391#ifdef INET
1392				case AF_INET:
1393					{
1394						struct sockaddr_in *sin,
1395						           *rsin;
1396
1397						sin = (struct sockaddr_in *)
1398						    &net->ro._l_addr;
1399						rsin = (struct sockaddr_in *)remote;
1400						if (sin->sin_addr.s_addr ==
1401						    rsin->sin_addr.s_addr) {
1402							/* found it */
1403							if (netp != NULL) {
1404								*netp = net;
1405							}
1406							if (locked_tcb == NULL) {
1407								SCTP_INP_DECR_REF(inp);
1408							} else if (locked_tcb != stcb) {
1409								SCTP_TCB_LOCK(locked_tcb);
1410							}
1411							if (locked_tcb) {
1412								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1413							}
1414							SCTP_INP_WUNLOCK(inp);
1415							SCTP_INP_INFO_RUNLOCK();
1416							return (stcb);
1417						}
1418						break;
1419					}
1420#endif
1421#ifdef INET6
1422				case AF_INET6:
1423					{
1424						struct sockaddr_in6 *sin6,
1425						            *rsin6;
1426
1427						sin6 = (struct sockaddr_in6 *)
1428						    &net->ro._l_addr;
1429						rsin6 = (struct sockaddr_in6 *)remote;
1430						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1431						    rsin6)) {
1432							/* found it */
1433							if (netp != NULL) {
1434								*netp = net;
1435							}
1436							if (locked_tcb == NULL) {
1437								SCTP_INP_DECR_REF(inp);
1438							} else if (locked_tcb != stcb) {
1439								SCTP_TCB_LOCK(locked_tcb);
1440							}
1441							if (locked_tcb) {
1442								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1443							}
1444							SCTP_INP_WUNLOCK(inp);
1445							SCTP_INP_INFO_RUNLOCK();
1446							return (stcb);
1447						}
1448						break;
1449					}
1450#endif
1451				default:
1452					/* TSNH */
1453					break;
1454				}
1455			}
1456			SCTP_TCB_UNLOCK(stcb);
1457		}
1458	}
1459null_return:
1460	/* clean up for returning null */
1461	if (locked_tcb) {
1462		SCTP_TCB_LOCK(locked_tcb);
1463		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1464	}
1465	SCTP_INP_WUNLOCK(inp);
1466	SCTP_INP_INFO_RUNLOCK();
1467	/* not found */
1468	return (NULL);
1469}
1470
1471/*
1472 * Find an association for a specific endpoint using the association id given
1473 * out in the COMM_UP notification
1474 */
1475
1476struct sctp_tcb *
1477sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1478{
1479	/*
1480	 * Use my the assoc_id to find a endpoint
1481	 */
1482	struct sctpasochead *head;
1483	struct sctp_tcb *stcb;
1484	uint32_t id;
1485
1486	if (inp == NULL) {
1487		SCTP_PRINTF("TSNH ep_associd\n");
1488		return (NULL);
1489	}
1490	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1491		SCTP_PRINTF("TSNH ep_associd0\n");
1492		return (NULL);
1493	}
1494	id = (uint32_t) asoc_id;
1495	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1496	if (head == NULL) {
1497		/* invalid id TSNH */
1498		SCTP_PRINTF("TSNH ep_associd1\n");
1499		return (NULL);
1500	}
1501	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1502		if (stcb->asoc.assoc_id == id) {
1503			if (inp != stcb->sctp_ep) {
1504				/*
1505				 * some other guy has the same id active (id
1506				 * collision ??).
1507				 */
1508				SCTP_PRINTF("TSNH ep_associd2\n");
1509				continue;
1510			}
1511			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1512				continue;
1513			}
1514			if (want_lock) {
1515				SCTP_TCB_LOCK(stcb);
1516			}
1517			return (stcb);
1518		}
1519	}
1520	return (NULL);
1521}
1522
1523
1524struct sctp_tcb *
1525sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1526{
1527	struct sctp_tcb *stcb;
1528
1529	SCTP_INP_RLOCK(inp);
1530	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1531	SCTP_INP_RUNLOCK(inp);
1532	return (stcb);
1533}
1534
1535
1536static struct sctp_inpcb *
1537sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1538    uint16_t lport, uint32_t vrf_id)
1539{
1540	struct sctp_inpcb *inp;
1541	struct sctp_laddr *laddr;
1542
1543#ifdef INET
1544	struct sockaddr_in *sin;
1545
1546#endif
1547#ifdef INET6
1548	struct sockaddr_in6 *sin6;
1549	struct sockaddr_in6 *intf_addr6;
1550
1551#endif
1552
1553	int fnd;
1554
1555	/*
1556	 * Endpoint probe expects that the INP_INFO is locked.
1557	 */
1558#ifdef INET
1559	sin = NULL;
1560#endif
1561#ifdef INET6
1562	sin6 = NULL;
1563#endif
1564	switch (nam->sa_family) {
1565#ifdef INET
1566	case AF_INET:
1567		sin = (struct sockaddr_in *)nam;
1568		break;
1569#endif
1570#ifdef INET6
1571	case AF_INET6:
1572		sin6 = (struct sockaddr_in6 *)nam;
1573		break;
1574#endif
1575	default:
1576		/* unsupported family */
1577		return (NULL);
1578	}
1579
1580	if (head == NULL)
1581		return (NULL);
1582
1583	LIST_FOREACH(inp, head, sctp_hash) {
1584		SCTP_INP_RLOCK(inp);
1585		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1586			SCTP_INP_RUNLOCK(inp);
1587			continue;
1588		}
1589		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1590		    (inp->sctp_lport == lport)) {
1591			/* got it */
1592#ifdef INET
1593			if ((nam->sa_family == AF_INET) &&
1594			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1595			    SCTP_IPV6_V6ONLY(inp)) {
1596				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1597				SCTP_INP_RUNLOCK(inp);
1598				continue;
1599			}
1600#endif
1601#ifdef INET6
1602			/* A V6 address and the endpoint is NOT bound V6 */
1603			if (nam->sa_family == AF_INET6 &&
1604			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1605				SCTP_INP_RUNLOCK(inp);
1606				continue;
1607			}
1608#endif
1609			/* does a VRF id match? */
1610			fnd = 0;
1611			if (inp->def_vrf_id == vrf_id)
1612				fnd = 1;
1613
1614			SCTP_INP_RUNLOCK(inp);
1615			if (!fnd)
1616				continue;
1617			return (inp);
1618		}
1619		SCTP_INP_RUNLOCK(inp);
1620	}
1621	switch (nam->sa_family) {
1622#ifdef INET
1623	case AF_INET:
1624		if (sin->sin_addr.s_addr == INADDR_ANY) {
1625			/* Can't hunt for one that has no address specified */
1626			return (NULL);
1627		}
1628		break;
1629#endif
1630#ifdef INET6
1631	case AF_INET6:
1632		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1633			/* Can't hunt for one that has no address specified */
1634			return (NULL);
1635		}
1636		break;
1637#endif
1638	default:
1639		break;
1640	}
1641	/*
1642	 * ok, not bound to all so see if we can find a EP bound to this
1643	 * address.
1644	 */
1645	LIST_FOREACH(inp, head, sctp_hash) {
1646		SCTP_INP_RLOCK(inp);
1647		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1648			SCTP_INP_RUNLOCK(inp);
1649			continue;
1650		}
1651		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1652			SCTP_INP_RUNLOCK(inp);
1653			continue;
1654		}
1655		/*
1656		 * Ok this could be a likely candidate, look at all of its
1657		 * addresses
1658		 */
1659		if (inp->sctp_lport != lport) {
1660			SCTP_INP_RUNLOCK(inp);
1661			continue;
1662		}
1663		/* does a VRF id match? */
1664		fnd = 0;
1665		if (inp->def_vrf_id == vrf_id)
1666			fnd = 1;
1667
1668		if (!fnd) {
1669			SCTP_INP_RUNLOCK(inp);
1670			continue;
1671		}
1672		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1673			if (laddr->ifa == NULL) {
1674				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1675				    __FUNCTION__);
1676				continue;
1677			}
1678			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1679			    laddr->ifa);
1680			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1681				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1682				continue;
1683			}
1684			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1685				/* possible, see if it matches */
1686				switch (nam->sa_family) {
1687#ifdef INET
1688				case AF_INET:
1689					if (sin->sin_addr.s_addr ==
1690					    laddr->ifa->address.sin.sin_addr.s_addr) {
1691						SCTP_INP_RUNLOCK(inp);
1692						return (inp);
1693					}
1694					break;
1695#endif
1696#ifdef INET6
1697				case AF_INET6:
1698					intf_addr6 = &laddr->ifa->address.sin6;
1699					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1700					    intf_addr6)) {
1701						SCTP_INP_RUNLOCK(inp);
1702						return (inp);
1703					}
1704					break;
1705#endif
1706				}
1707			}
1708		}
1709		SCTP_INP_RUNLOCK(inp);
1710	}
1711	return (NULL);
1712}
1713
1714
1715static struct sctp_inpcb *
1716sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1717{
1718	struct sctppcbhead *head;
1719	struct sctp_inpcb *t_inp;
1720	int fnd;
1721
1722	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1723	    SCTP_BASE_INFO(hashmark))];
1724	LIST_FOREACH(t_inp, head, sctp_hash) {
1725		if (t_inp->sctp_lport != lport) {
1726			continue;
1727		}
1728		/* is it in the VRF in question */
1729		fnd = 0;
1730		if (t_inp->def_vrf_id == vrf_id)
1731			fnd = 1;
1732		if (!fnd)
1733			continue;
1734
1735		/* This one is in use. */
1736		/* check the v6/v4 binding issue */
1737		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1738		    SCTP_IPV6_V6ONLY(t_inp)) {
1739			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1740				/* collision in V6 space */
1741				return (t_inp);
1742			} else {
1743				/* inp is BOUND_V4 no conflict */
1744				continue;
1745			}
1746		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1747			/* t_inp is bound v4 and v6, conflict always */
1748			return (t_inp);
1749		} else {
1750			/* t_inp is bound only V4 */
1751			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1752			    SCTP_IPV6_V6ONLY(inp)) {
1753				/* no conflict */
1754				continue;
1755			}
1756			/* else fall through to conflict */
1757		}
1758		return (t_inp);
1759	}
1760	return (NULL);
1761}
1762
1763
1764int
1765sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1766{
1767	/* For 1-2-1 with port reuse */
1768	struct sctppcbhead *head;
1769	struct sctp_inpcb *tinp;
1770
1771	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1772		/* only works with port reuse on */
1773		return (-1);
1774	}
1775	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1776		return (0);
1777	}
1778	SCTP_INP_RUNLOCK(inp);
1779	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1780	    SCTP_BASE_INFO(hashmark))];
1781	/* Kick out all non-listeners to the TCP hash */
1782	LIST_FOREACH(tinp, head, sctp_hash) {
1783		if (tinp->sctp_lport != inp->sctp_lport) {
1784			continue;
1785		}
1786		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1787			continue;
1788		}
1789		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1790			continue;
1791		}
1792		if (tinp->sctp_socket->so_qlimit) {
1793			continue;
1794		}
1795		SCTP_INP_WLOCK(tinp);
1796		LIST_REMOVE(tinp, sctp_hash);
1797		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1798		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1799		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1800		SCTP_INP_WUNLOCK(tinp);
1801	}
1802	SCTP_INP_WLOCK(inp);
1803	/* Pull from where he was */
1804	LIST_REMOVE(inp, sctp_hash);
1805	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1806	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1807	LIST_INSERT_HEAD(head, inp, sctp_hash);
1808	SCTP_INP_WUNLOCK(inp);
1809	SCTP_INP_RLOCK(inp);
1810	return (0);
1811}
1812
1813
1814struct sctp_inpcb *
1815sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1816    uint32_t vrf_id)
1817{
1818	/*
1819	 * First we check the hash table to see if someone has this port
1820	 * bound with just the port.
1821	 */
1822	struct sctp_inpcb *inp;
1823	struct sctppcbhead *head;
1824	int lport;
1825	unsigned int i;
1826
1827#ifdef INET
1828	struct sockaddr_in *sin;
1829
1830#endif
1831#ifdef INET6
1832	struct sockaddr_in6 *sin6;
1833
1834#endif
1835
1836	switch (nam->sa_family) {
1837#ifdef INET
1838	case AF_INET:
1839		sin = (struct sockaddr_in *)nam;
1840		lport = ((struct sockaddr_in *)nam)->sin_port;
1841		break;
1842#endif
1843#ifdef INET6
1844	case AF_INET6:
1845		sin6 = (struct sockaddr_in6 *)nam;
1846		lport = ((struct sockaddr_in6 *)nam)->sin6_port;
1847		break;
1848#endif
1849	default:
1850		return (NULL);
1851	}
1852	/*
1853	 * I could cheat here and just cast to one of the types but we will
1854	 * do it right. It also provides the check against an Unsupported
1855	 * type too.
1856	 */
1857	/* Find the head of the ALLADDR chain */
1858	if (have_lock == 0) {
1859		SCTP_INP_INFO_RLOCK();
1860	}
1861	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1862	    SCTP_BASE_INFO(hashmark))];
1863	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1864
1865	/*
1866	 * If the TCP model exists it could be that the main listening
1867	 * endpoint is gone but there still exists a connected socket for
1868	 * this guy. If so we can return the first one that we find. This
1869	 * may NOT be the correct one so the caller should be wary on the
1870	 * returned INP. Currently the only caller that sets find_tcp_pool
1871	 * is in bindx where we are verifying that a user CAN bind the
1872	 * address. He either has bound it already, or someone else has, or
1873	 * its open to bind, so this is good enough.
1874	 */
1875	if (inp == NULL && find_tcp_pool) {
1876		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1877			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1878			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1879			if (inp) {
1880				break;
1881			}
1882		}
1883	}
1884	if (inp) {
1885		SCTP_INP_INCR_REF(inp);
1886	}
1887	if (have_lock == 0) {
1888		SCTP_INP_INFO_RUNLOCK();
1889	}
1890	return (inp);
1891}
1892
1893/*
1894 * Find an association for an endpoint with the pointer to whom you want to
1895 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1896 * need to change the *to to some other struct like a mbuf...
1897 */
1898struct sctp_tcb *
1899sctp_findassociation_addr_sa(struct sockaddr *to, struct sockaddr *from,
1900    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1901    uint32_t vrf_id)
1902{
1903	struct sctp_inpcb *inp = NULL;
1904	struct sctp_tcb *retval;
1905
1906	SCTP_INP_INFO_RLOCK();
1907	if (find_tcp_pool) {
1908		if (inp_p != NULL) {
1909			retval = sctp_tcb_special_locate(inp_p, from, to, netp,
1910			    vrf_id);
1911		} else {
1912			retval = sctp_tcb_special_locate(&inp, from, to, netp,
1913			    vrf_id);
1914		}
1915		if (retval != NULL) {
1916			SCTP_INP_INFO_RUNLOCK();
1917			return (retval);
1918		}
1919	}
1920	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1921	if (inp_p != NULL) {
1922		*inp_p = inp;
1923	}
1924	SCTP_INP_INFO_RUNLOCK();
1925
1926	if (inp == NULL) {
1927		return (NULL);
1928	}
1929	/*
1930	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1931	 * we now place the source address or from in the to of the find
1932	 * endpoint call. Since in reality this chain is used from the
1933	 * inbound packet side.
1934	 */
1935	if (inp_p != NULL) {
1936		retval = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1937		    NULL);
1938	} else {
1939		retval = sctp_findassociation_ep_addr(&inp, from, netp, to,
1940		    NULL);
1941	}
1942	return retval;
1943}
1944
1945
1946/*
1947 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1948 * find all addresses that the sender has specified in any address list. Each
1949 * address will be used to lookup the TCB and see if one exits.
1950 */
1951static struct sctp_tcb *
1952sctp_findassociation_special_addr(struct mbuf *m, int iphlen, int offset,
1953    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1954    struct sockaddr *dest)
1955{
1956	struct sctp_paramhdr *phdr, parm_buf;
1957	struct sctp_tcb *retval;
1958	uint32_t ptype, plen;
1959
1960#ifdef INET
1961	struct sockaddr_in sin4;
1962
1963#endif
1964#ifdef INET6
1965	struct sockaddr_in6 sin6;
1966
1967#endif
1968
1969#ifdef INET
1970	memset(&sin4, 0, sizeof(sin4));
1971	sin4.sin_len = sizeof(sin4);
1972	sin4.sin_family = AF_INET;
1973	sin4.sin_port = sh->src_port;
1974#endif
1975#ifdef INET6
1976	memset(&sin6, 0, sizeof(sin6));
1977	sin6.sin6_len = sizeof(sin6);
1978	sin6.sin6_family = AF_INET6;
1979	sin6.sin6_port = sh->src_port;
1980#endif
1981
1982	retval = NULL;
1983	offset += sizeof(struct sctp_init_chunk);
1984
1985	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
1986	while (phdr != NULL) {
1987		/* now we must see if we want the parameter */
1988		ptype = ntohs(phdr->param_type);
1989		plen = ntohs(phdr->param_length);
1990		if (plen == 0) {
1991			break;
1992		}
1993#ifdef INET
1994		if (ptype == SCTP_IPV4_ADDRESS &&
1995		    plen == sizeof(struct sctp_ipv4addr_param)) {
1996			/* Get the rest of the address */
1997			struct sctp_ipv4addr_param ip4_parm, *p4;
1998
1999			phdr = sctp_get_next_param(m, offset,
2000			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2001			if (phdr == NULL) {
2002				return (NULL);
2003			}
2004			p4 = (struct sctp_ipv4addr_param *)phdr;
2005			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2006			/* look it up */
2007			retval = sctp_findassociation_ep_addr(inp_p,
2008			    (struct sockaddr *)&sin4, netp, dest, NULL);
2009			if (retval != NULL) {
2010				return (retval);
2011			}
2012		}
2013#endif
2014#ifdef INET6
2015		if (ptype == SCTP_IPV6_ADDRESS &&
2016		    plen == sizeof(struct sctp_ipv6addr_param)) {
2017			/* Get the rest of the address */
2018			struct sctp_ipv6addr_param ip6_parm, *p6;
2019
2020			phdr = sctp_get_next_param(m, offset,
2021			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2022			if (phdr == NULL) {
2023				return (NULL);
2024			}
2025			p6 = (struct sctp_ipv6addr_param *)phdr;
2026			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2027			/* look it up */
2028			retval = sctp_findassociation_ep_addr(inp_p,
2029			    (struct sockaddr *)&sin6, netp, dest, NULL);
2030			if (retval != NULL) {
2031				return (retval);
2032			}
2033		}
2034#endif
2035		offset += SCTP_SIZE32(plen);
2036		phdr = sctp_get_next_param(m, offset, &parm_buf,
2037		    sizeof(parm_buf));
2038	}
2039	return (NULL);
2040}
2041
2042static struct sctp_tcb *
2043sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2044    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2045    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2046{
2047	/*
2048	 * Use my vtag to hash. If we find it we then verify the source addr
2049	 * is in the assoc. If all goes well we save a bit on rec of a
2050	 * packet.
2051	 */
2052	struct sctpasochead *head;
2053	struct sctp_nets *net;
2054	struct sctp_tcb *stcb;
2055
2056	*netp = NULL;
2057	*inp_p = NULL;
2058	SCTP_INP_INFO_RLOCK();
2059	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2060	    SCTP_BASE_INFO(hashasocmark))];
2061	if (head == NULL) {
2062		/* invalid vtag */
2063		SCTP_INP_INFO_RUNLOCK();
2064		return (NULL);
2065	}
2066	LIST_FOREACH(stcb, head, sctp_asocs) {
2067		SCTP_INP_RLOCK(stcb->sctp_ep);
2068		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2069			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2070			continue;
2071		}
2072		SCTP_TCB_LOCK(stcb);
2073		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2074		if (stcb->asoc.my_vtag == vtag) {
2075			/* candidate */
2076			if (stcb->rport != rport) {
2077				SCTP_TCB_UNLOCK(stcb);
2078				continue;
2079			}
2080			if (stcb->sctp_ep->sctp_lport != lport) {
2081				SCTP_TCB_UNLOCK(stcb);
2082				continue;
2083			}
2084			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2085				SCTP_TCB_UNLOCK(stcb);
2086				continue;
2087			}
2088			/* RRS:Need toaddr check here */
2089			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2090				/* Endpoint does not own this address */
2091				SCTP_TCB_UNLOCK(stcb);
2092				continue;
2093			}
2094			if (remote_tag) {
2095				/*
2096				 * If we have both vtags that's all we match
2097				 * on
2098				 */
2099				if (stcb->asoc.peer_vtag == remote_tag) {
2100					/*
2101					 * If both tags match we consider it
2102					 * conclusive and check NO
2103					 * source/destination addresses
2104					 */
2105					goto conclusive;
2106				}
2107			}
2108			if (skip_src_check) {
2109		conclusive:
2110				if (from) {
2111					net = sctp_findnet(stcb, from);
2112				} else {
2113					*netp = NULL;	/* unknown */
2114				}
2115				if (inp_p)
2116					*inp_p = stcb->sctp_ep;
2117				SCTP_INP_INFO_RUNLOCK();
2118				return (stcb);
2119			}
2120			net = sctp_findnet(stcb, from);
2121			if (net) {
2122				/* yep its him. */
2123				*netp = net;
2124				SCTP_STAT_INCR(sctps_vtagexpress);
2125				*inp_p = stcb->sctp_ep;
2126				SCTP_INP_INFO_RUNLOCK();
2127				return (stcb);
2128			} else {
2129				/*
2130				 * not him, this should only happen in rare
2131				 * cases so I peg it.
2132				 */
2133				SCTP_STAT_INCR(sctps_vtagbogus);
2134			}
2135		}
2136		SCTP_TCB_UNLOCK(stcb);
2137	}
2138	SCTP_INP_INFO_RUNLOCK();
2139	return (NULL);
2140}
2141
2142/*
2143 * Find an association with the pointer to the inbound IP packet. This can be
2144 * a IPv4 or IPv6 packet.
2145 */
2146struct sctp_tcb *
2147sctp_findassociation_addr(struct mbuf *m, int iphlen, int offset,
2148    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2149    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2150{
2151	int find_tcp_pool;
2152	struct ip *iph;
2153	struct sctp_tcb *retval;
2154	struct sockaddr_storage to_store, from_store;
2155	struct sockaddr *to = (struct sockaddr *)&to_store;
2156	struct sockaddr *from = (struct sockaddr *)&from_store;
2157	struct sctp_inpcb *inp;
2158
2159	iph = mtod(m, struct ip *);
2160	switch (iph->ip_v) {
2161#ifdef INET
2162	case IPVERSION:
2163		{
2164			/* its IPv4 */
2165			struct sockaddr_in *from4;
2166
2167			from4 = (struct sockaddr_in *)&from_store;
2168			bzero(from4, sizeof(*from4));
2169			from4->sin_family = AF_INET;
2170			from4->sin_len = sizeof(struct sockaddr_in);
2171			from4->sin_addr.s_addr = iph->ip_src.s_addr;
2172			from4->sin_port = sh->src_port;
2173			break;
2174		}
2175#endif
2176#ifdef INET6
2177	case IPV6_VERSION >> 4:
2178		{
2179			/* its IPv6 */
2180			struct ip6_hdr *ip6;
2181			struct sockaddr_in6 *from6;
2182
2183			ip6 = mtod(m, struct ip6_hdr *);
2184			from6 = (struct sockaddr_in6 *)&from_store;
2185			bzero(from6, sizeof(*from6));
2186			from6->sin6_family = AF_INET6;
2187			from6->sin6_len = sizeof(struct sockaddr_in6);
2188			from6->sin6_addr = ip6->ip6_src;
2189			from6->sin6_port = sh->src_port;
2190			/* Get the scopes in properly to the sin6 addr's */
2191			/* we probably don't need these operations */
2192			(void)sa6_recoverscope(from6);
2193			sa6_embedscope(from6, MODULE_GLOBAL(ip6_use_defzone));
2194			break;
2195		}
2196#endif
2197	default:
2198		/* Currently not supported. */
2199		return (NULL);
2200	}
2201
2202
2203	switch (iph->ip_v) {
2204#ifdef INET
2205	case IPVERSION:
2206		{
2207			/* its IPv4 */
2208			struct sockaddr_in *to4;
2209
2210			to4 = (struct sockaddr_in *)&to_store;
2211			bzero(to4, sizeof(*to4));
2212			to4->sin_family = AF_INET;
2213			to4->sin_len = sizeof(struct sockaddr_in);
2214			to4->sin_addr.s_addr = iph->ip_dst.s_addr;
2215			to4->sin_port = sh->dest_port;
2216			break;
2217		}
2218#endif
2219#ifdef INET6
2220	case IPV6_VERSION >> 4:
2221		{
2222			/* its IPv6 */
2223			struct ip6_hdr *ip6;
2224			struct sockaddr_in6 *to6;
2225
2226			ip6 = mtod(m, struct ip6_hdr *);
2227			to6 = (struct sockaddr_in6 *)&to_store;
2228			bzero(to6, sizeof(*to6));
2229			to6->sin6_family = AF_INET6;
2230			to6->sin6_len = sizeof(struct sockaddr_in6);
2231			to6->sin6_addr = ip6->ip6_dst;
2232			to6->sin6_port = sh->dest_port;
2233			/* Get the scopes in properly to the sin6 addr's */
2234			/* we probably don't need these operations */
2235			(void)sa6_recoverscope(to6);
2236			sa6_embedscope(to6, MODULE_GLOBAL(ip6_use_defzone));
2237			break;
2238		}
2239#endif
2240	default:
2241		/* TSNH */
2242		break;
2243	}
2244	if (sh->v_tag) {
2245		/* we only go down this path if vtag is non-zero */
2246		retval = sctp_findassoc_by_vtag(from, to, ntohl(sh->v_tag),
2247		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2248		if (retval) {
2249			return (retval);
2250		}
2251	}
2252	find_tcp_pool = 0;
2253	if ((ch->chunk_type != SCTP_INITIATION) &&
2254	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2255	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2256	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2257		/* Other chunk types go to the tcp pool. */
2258		find_tcp_pool = 1;
2259	}
2260	if (inp_p) {
2261		retval = sctp_findassociation_addr_sa(to, from, inp_p, netp,
2262		    find_tcp_pool, vrf_id);
2263		inp = *inp_p;
2264	} else {
2265		retval = sctp_findassociation_addr_sa(to, from, &inp, netp,
2266		    find_tcp_pool, vrf_id);
2267	}
2268	SCTPDBG(SCTP_DEBUG_PCB1, "retval:%p inp:%p\n", retval, inp);
2269	if (retval == NULL && inp) {
2270		/* Found a EP but not this address */
2271		if ((ch->chunk_type == SCTP_INITIATION) ||
2272		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2273			/*-
2274			 * special hook, we do NOT return linp or an
2275			 * association that is linked to an existing
2276			 * association that is under the TCP pool (i.e. no
2277			 * listener exists). The endpoint finding routine
2278			 * will always find a listener before examining the
2279			 * TCP pool.
2280			 */
2281			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2282				if (inp_p) {
2283					*inp_p = NULL;
2284				}
2285				return (NULL);
2286			}
2287			retval = sctp_findassociation_special_addr(m, iphlen,
2288			    offset, sh, &inp, netp, to);
2289			if (inp_p != NULL) {
2290				*inp_p = inp;
2291			}
2292		}
2293	}
2294	SCTPDBG(SCTP_DEBUG_PCB1, "retval is %p\n", retval);
2295	return (retval);
2296}
2297
2298/*
2299 * lookup an association by an ASCONF lookup address.
2300 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2301 */
2302struct sctp_tcb *
2303sctp_findassociation_ep_asconf(struct mbuf *m, int iphlen, int offset,
2304    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2305{
2306	struct sctp_tcb *stcb;
2307	struct sockaddr_storage local_store, remote_store;
2308	struct sockaddr *to;
2309	struct ip *iph;
2310	struct sctp_paramhdr parm_buf, *phdr;
2311	int ptype;
2312	int zero_address = 0;
2313
2314#ifdef INET
2315	struct sockaddr_in *sin;
2316
2317#endif
2318#ifdef INET6
2319	struct ip6_hdr *ip6;
2320	struct sockaddr_in6 *sin6;
2321
2322#endif
2323
2324	memset(&local_store, 0, sizeof(local_store));
2325	memset(&remote_store, 0, sizeof(remote_store));
2326	to = (struct sockaddr *)&local_store;
2327	/* First get the destination address setup too. */
2328	iph = mtod(m, struct ip *);
2329	switch (iph->ip_v) {
2330#ifdef INET
2331	case IPVERSION:
2332		/* its IPv4 */
2333		sin = (struct sockaddr_in *)&local_store;
2334		sin->sin_family = AF_INET;
2335		sin->sin_len = sizeof(*sin);
2336		sin->sin_port = sh->dest_port;
2337		sin->sin_addr.s_addr = iph->ip_dst.s_addr;
2338		break;
2339#endif
2340#ifdef INET6
2341	case IPV6_VERSION >> 4:
2342		/* its IPv6 */
2343		ip6 = mtod(m, struct ip6_hdr *);
2344		sin6 = (struct sockaddr_in6 *)&local_store;
2345		sin6->sin6_family = AF_INET6;
2346		sin6->sin6_len = sizeof(*sin6);
2347		sin6->sin6_port = sh->dest_port;
2348		sin6->sin6_addr = ip6->ip6_dst;
2349		break;
2350#endif
2351	default:
2352		return NULL;
2353	}
2354
2355	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2356	    &parm_buf, sizeof(struct sctp_paramhdr));
2357	if (phdr == NULL) {
2358		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2359		    __FUNCTION__);
2360		return NULL;
2361	}
2362	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2363	/* get the correlation address */
2364	switch (ptype) {
2365#ifdef INET6
2366	case SCTP_IPV6_ADDRESS:
2367		{
2368			/* ipv6 address param */
2369			struct sctp_ipv6addr_param *p6, p6_buf;
2370
2371			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2372				return NULL;
2373			}
2374			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2375			    offset + sizeof(struct sctp_asconf_chunk),
2376			    &p6_buf.ph, sizeof(*p6));
2377			if (p6 == NULL) {
2378				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2379				    __FUNCTION__);
2380				return (NULL);
2381			}
2382			sin6 = (struct sockaddr_in6 *)&remote_store;
2383			sin6->sin6_family = AF_INET6;
2384			sin6->sin6_len = sizeof(*sin6);
2385			sin6->sin6_port = sh->src_port;
2386			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2387			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2388				zero_address = 1;
2389			break;
2390		}
2391#endif
2392#ifdef INET
2393	case SCTP_IPV4_ADDRESS:
2394		{
2395			/* ipv4 address param */
2396			struct sctp_ipv4addr_param *p4, p4_buf;
2397
2398			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2399				return NULL;
2400			}
2401			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2402			    offset + sizeof(struct sctp_asconf_chunk),
2403			    &p4_buf.ph, sizeof(*p4));
2404			if (p4 == NULL) {
2405				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2406				    __FUNCTION__);
2407				return (NULL);
2408			}
2409			sin = (struct sockaddr_in *)&remote_store;
2410			sin->sin_family = AF_INET;
2411			sin->sin_len = sizeof(*sin);
2412			sin->sin_port = sh->src_port;
2413			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2414			if (sin->sin_addr.s_addr == INADDR_ANY)
2415				zero_address = 1;
2416			break;
2417		}
2418#endif
2419	default:
2420		/* invalid address param type */
2421		return NULL;
2422	}
2423
2424	if (zero_address) {
2425		stcb = sctp_findassoc_by_vtag(NULL, to, ntohl(sh->v_tag), inp_p,
2426		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2427		/*
2428		 * printf("findassociation_ep_asconf: zero lookup address
2429		 * finds stcb 0x%x\n", (uint32_t)stcb);
2430		 */
2431	} else {
2432		stcb = sctp_findassociation_ep_addr(inp_p,
2433		    (struct sockaddr *)&remote_store, netp,
2434		    to, NULL);
2435	}
2436	return (stcb);
2437}
2438
2439
2440/*
2441 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2442 * addresses. This way if we don't get a bind we by default pick a ephemeral
2443 * port with all addresses bound.
2444 */
2445int
2446sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2447{
2448	/*
2449	 * we get called when a new endpoint starts up. We need to allocate
2450	 * the sctp_inpcb structure from the zone and init it. Mark it as
2451	 * unbound and find a port that we can use as an ephemeral with
2452	 * INADDR_ANY. If the user binds later no problem we can then add in
2453	 * the specific addresses. And setup the default parameters for the
2454	 * EP.
2455	 */
2456	int i, error;
2457	struct sctp_inpcb *inp;
2458	struct sctp_pcb *m;
2459	struct timeval time;
2460	sctp_sharedkey_t *null_key;
2461
2462	error = 0;
2463
2464	SCTP_INP_INFO_WLOCK();
2465	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2466	if (inp == NULL) {
2467		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2468		SCTP_INP_INFO_WUNLOCK();
2469		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2470		return (ENOBUFS);
2471	}
2472	/* zap it */
2473	bzero(inp, sizeof(*inp));
2474
2475	/* bump generations */
2476	/* setup socket pointers */
2477	inp->sctp_socket = so;
2478	inp->ip_inp.inp.inp_socket = so;
2479#ifdef INET6
2480	if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2481		inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2482	}
2483#endif
2484	inp->sctp_associd_counter = 1;
2485	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2486	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2487	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2488	inp->sctp_ecn_enable = SCTP_BASE_SYSCTL(sctp_ecn_enable);
2489	/* init the small hash table we use to track asocid <-> tcb */
2490	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2491	if (inp->sctp_asocidhash == NULL) {
2492		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2493		SCTP_INP_INFO_WUNLOCK();
2494		return (ENOBUFS);
2495	}
2496#ifdef IPSEC
2497	{
2498		struct inpcbpolicy *pcb_sp = NULL;
2499
2500		error = ipsec_init_policy(so, &pcb_sp);
2501		/* Arrange to share the policy */
2502		inp->ip_inp.inp.inp_sp = pcb_sp;
2503		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2504	}
2505	if (error != 0) {
2506		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2507		SCTP_INP_INFO_WUNLOCK();
2508		return error;
2509	}
2510#endif				/* IPSEC */
2511	SCTP_INCR_EP_COUNT();
2512	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2513	SCTP_INP_INFO_WUNLOCK();
2514
2515	so->so_pcb = (caddr_t)inp;
2516
2517	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2518		/* UDP style socket */
2519		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2520		    SCTP_PCB_FLAGS_UNBOUND);
2521		/* Be sure it is NON-BLOCKING IO for UDP */
2522		/* SCTP_SET_SO_NBIO(so); */
2523	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2524		/* TCP style socket */
2525		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2526		    SCTP_PCB_FLAGS_UNBOUND);
2527		/* Be sure we have blocking IO by default */
2528		SCTP_CLEAR_SO_NBIO(so);
2529	} else {
2530		/*
2531		 * unsupported socket type (RAW, etc)- in case we missed it
2532		 * in protosw
2533		 */
2534		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2535		so->so_pcb = NULL;
2536		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2537		return (EOPNOTSUPP);
2538	}
2539	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2540		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2541		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2542	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2543		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2544		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2545	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2546		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2547		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2548	}
2549	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2550	    &inp->sctp_hashmark);
2551	if (inp->sctp_tcbhash == NULL) {
2552		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2553		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2554		so->so_pcb = NULL;
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
2603	m->max_open_streams_intome = MAX_SCTP_STREAMS;
2604
2605	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2606	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2607	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2608	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2609	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2610	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2611	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2612	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2613
2614	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2615	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2616	/* number of streams to pre-open on a association */
2617	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2618
2619	/* Add adaptation cookie */
2620	m->adaptation_layer_indicator = 0x504C5253;
2621
2622	/* seed random number generator */
2623	m->random_counter = 1;
2624	m->store_at = SCTP_SIGNATURE_SIZE;
2625	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2626	sctp_fill_random_store(m);
2627
2628	/* Minimum cookie size */
2629	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2630	    sizeof(struct sctp_state_cookie);
2631	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2632
2633	/* Setup the initial secret */
2634	(void)SCTP_GETTIME_TIMEVAL(&time);
2635	m->time_of_secret_change = time.tv_sec;
2636
2637	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2638		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2639	}
2640	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2641
2642	/* How long is a cookie good for ? */
2643	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2644	/*
2645	 * Initialize authentication parameters
2646	 */
2647	m->local_hmacs = sctp_default_supported_hmaclist();
2648	m->local_auth_chunks = sctp_alloc_chunklist();
2649	m->default_dscp = 0;
2650#ifdef INET6
2651	m->default_flowlabel = 0;
2652#endif
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	error = 0;
2807	bindall = 1;
2808	inp = (struct sctp_inpcb *)so->so_pcb;
2809	ip_inp = (struct inpcb *)so->so_pcb;
2810#ifdef SCTP_DEBUG
2811	if (addr) {
2812		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n",
2813		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2814		SCTPDBG(SCTP_DEBUG_PCB1, "Addr :");
2815		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2816	}
2817#endif
2818	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2819		/* already did a bind, subsequent binds NOT allowed ! */
2820		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2821		return (EINVAL);
2822	}
2823#ifdef INVARIANTS
2824	if (p == NULL)
2825		panic("null proc/thread");
2826#endif
2827	if (addr != NULL) {
2828		switch (addr->sa_family) {
2829#ifdef INET
2830		case AF_INET:
2831			{
2832				struct sockaddr_in *sin;
2833
2834				/* IPV6_V6ONLY socket? */
2835				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2836					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2837					return (EINVAL);
2838				}
2839				if (addr->sa_len != sizeof(*sin)) {
2840					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2841					return (EINVAL);
2842				}
2843				sin = (struct sockaddr_in *)addr;
2844				lport = sin->sin_port;
2845				/*
2846				 * For LOOPBACK the prison_local_ip4() call
2847				 * will transmute the ip address to the
2848				 * proper value.
2849				 */
2850				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2851					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2852					return (error);
2853				}
2854				if (sin->sin_addr.s_addr != INADDR_ANY) {
2855					bindall = 0;
2856				}
2857				break;
2858			}
2859#endif
2860#ifdef INET6
2861		case AF_INET6:
2862			{
2863				/*
2864				 * Only for pure IPv6 Address. (No IPv4
2865				 * Mapped!)
2866				 */
2867				struct sockaddr_in6 *sin6;
2868
2869				sin6 = (struct sockaddr_in6 *)addr;
2870
2871				if (addr->sa_len != sizeof(*sin6)) {
2872					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2873					return (EINVAL);
2874				}
2875				lport = sin6->sin6_port;
2876
2877				/*
2878				 * For LOOPBACK the prison_local_ip6() call
2879				 * will transmute the ipv6 address to the
2880				 * proper value.
2881				 */
2882				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2883				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2884					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2885					return (error);
2886				}
2887				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2888					bindall = 0;
2889					/* KAME hack: embed scopeid */
2890					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2891						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2892						return (EINVAL);
2893					}
2894				}
2895				/* this must be cleared for ifa_ifwithaddr() */
2896				sin6->sin6_scope_id = 0;
2897				break;
2898			}
2899#endif
2900		default:
2901			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2902			return (EAFNOSUPPORT);
2903		}
2904	}
2905	SCTP_INP_INFO_WLOCK();
2906	SCTP_INP_WLOCK(inp);
2907	/* Setup a vrf_id to be the default for the non-bind-all case. */
2908	vrf_id = inp->def_vrf_id;
2909
2910	/* increase our count due to the unlock we do */
2911	SCTP_INP_INCR_REF(inp);
2912	if (lport) {
2913		/*
2914		 * Did the caller specify a port? if so we must see if a ep
2915		 * already has this one bound.
2916		 */
2917		/* got to be root to get at low ports */
2918		if (ntohs(lport) < IPPORT_RESERVED) {
2919			if (p && (error =
2920			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2921			    )) {
2922				SCTP_INP_DECR_REF(inp);
2923				SCTP_INP_WUNLOCK(inp);
2924				SCTP_INP_INFO_WUNLOCK();
2925				return (error);
2926			}
2927		}
2928		if (p == NULL) {
2929			SCTP_INP_DECR_REF(inp);
2930			SCTP_INP_WUNLOCK(inp);
2931			SCTP_INP_INFO_WUNLOCK();
2932			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2933			return (error);
2934		}
2935		SCTP_INP_WUNLOCK(inp);
2936		if (bindall) {
2937			vrf_id = inp->def_vrf_id;
2938			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2939			if (inp_tmp != NULL) {
2940				/*
2941				 * lock guy returned and lower count note
2942				 * that we are not bound so inp_tmp should
2943				 * NEVER be inp. And it is this inp
2944				 * (inp_tmp) that gets the reference bump,
2945				 * so we must lower it.
2946				 */
2947				SCTP_INP_DECR_REF(inp_tmp);
2948				/* unlock info */
2949				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2950				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2951					/*
2952					 * Ok, must be one-2-one and
2953					 * allowing port re-use
2954					 */
2955					port_reuse_active = 1;
2956					goto continue_anyway;
2957				}
2958				SCTP_INP_DECR_REF(inp);
2959				SCTP_INP_INFO_WUNLOCK();
2960				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2961				return (EADDRINUSE);
2962			}
2963		} else {
2964			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2965			if (inp_tmp != NULL) {
2966				/*
2967				 * lock guy returned and lower count note
2968				 * that we are not bound so inp_tmp should
2969				 * NEVER be inp. And it is this inp
2970				 * (inp_tmp) that gets the reference bump,
2971				 * so we must lower it.
2972				 */
2973				SCTP_INP_DECR_REF(inp_tmp);
2974				/* unlock info */
2975				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2976				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2977					/*
2978					 * Ok, must be one-2-one and
2979					 * allowing port re-use
2980					 */
2981					port_reuse_active = 1;
2982					goto continue_anyway;
2983				}
2984				SCTP_INP_DECR_REF(inp);
2985				SCTP_INP_INFO_WUNLOCK();
2986				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2987				return (EADDRINUSE);
2988			}
2989		}
2990continue_anyway:
2991		SCTP_INP_WLOCK(inp);
2992		if (bindall) {
2993			/* verify that no lport is not used by a singleton */
2994			if ((port_reuse_active == 0) &&
2995			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))
2996			    ) {
2997				/* Sorry someone already has this one bound */
2998				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2999				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
3000					port_reuse_active = 1;
3001				} else {
3002					SCTP_INP_DECR_REF(inp);
3003					SCTP_INP_WUNLOCK(inp);
3004					SCTP_INP_INFO_WUNLOCK();
3005					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3006					return (EADDRINUSE);
3007				}
3008			}
3009		}
3010	} else {
3011		uint16_t first, last, candidate;
3012		uint16_t count;
3013		int done;
3014
3015		if (ip_inp->inp_flags & INP_HIGHPORT) {
3016			first = MODULE_GLOBAL(ipport_hifirstauto);
3017			last = MODULE_GLOBAL(ipport_hilastauto);
3018		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3019			if (p && (error =
3020			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3021			    )) {
3022				SCTP_INP_DECR_REF(inp);
3023				SCTP_INP_WUNLOCK(inp);
3024				SCTP_INP_INFO_WUNLOCK();
3025				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3026				return (error);
3027			}
3028			first = MODULE_GLOBAL(ipport_lowfirstauto);
3029			last = MODULE_GLOBAL(ipport_lowlastauto);
3030		} else {
3031			first = MODULE_GLOBAL(ipport_firstauto);
3032			last = MODULE_GLOBAL(ipport_lastauto);
3033		}
3034		if (first > last) {
3035			uint16_t temp;
3036
3037			temp = first;
3038			first = last;
3039			last = temp;
3040		}
3041		count = last - first + 1;	/* number of candidates */
3042		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3043
3044		done = 0;
3045		while (!done) {
3046			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3047				done = 1;
3048			}
3049			if (!done) {
3050				if (--count == 0) {
3051					SCTP_INP_DECR_REF(inp);
3052					SCTP_INP_WUNLOCK(inp);
3053					SCTP_INP_INFO_WUNLOCK();
3054					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3055					return (EADDRINUSE);
3056				}
3057				if (candidate == last)
3058					candidate = first;
3059				else
3060					candidate = candidate + 1;
3061			}
3062		}
3063		lport = htons(candidate);
3064	}
3065	SCTP_INP_DECR_REF(inp);
3066	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3067	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3068		/*
3069		 * this really should not happen. The guy did a non-blocking
3070		 * bind and then did a close at the same time.
3071		 */
3072		SCTP_INP_WUNLOCK(inp);
3073		SCTP_INP_INFO_WUNLOCK();
3074		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3075		return (EINVAL);
3076	}
3077	/* ok we look clear to give out this port, so lets setup the binding */
3078	if (bindall) {
3079		/* binding to all addresses, so just set in the proper flags */
3080		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3081		/* set the automatic addr changes from kernel flag */
3082		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3083			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3084			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3085		} else {
3086			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3087			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3088		}
3089		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3090			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3091		} else {
3092			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3093		}
3094		/*
3095		 * set the automatic mobility_base from kernel flag (by
3096		 * micchie)
3097		 */
3098		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3099			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3100			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3101		} else {
3102			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3103			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3104		}
3105		/*
3106		 * set the automatic mobility_fasthandoff from kernel flag
3107		 * (by micchie)
3108		 */
3109		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3110			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3111			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3112		} else {
3113			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3114			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3115		}
3116	} else {
3117		/*
3118		 * bind specific, make sure flags is off and add a new
3119		 * address structure to the sctp_addr_list inside the ep
3120		 * structure.
3121		 *
3122		 * We will need to allocate one and insert it at the head. The
3123		 * socketopt call can just insert new addresses in there as
3124		 * well. It will also have to do the embed scope kame hack
3125		 * too (before adding).
3126		 */
3127		struct sctp_ifa *ifa;
3128		struct sockaddr_storage store_sa;
3129
3130		memset(&store_sa, 0, sizeof(store_sa));
3131		switch (addr->sa_family) {
3132		case AF_INET:
3133			{
3134				struct sockaddr_in *sin;
3135
3136				sin = (struct sockaddr_in *)&store_sa;
3137				memcpy(sin, addr, sizeof(struct sockaddr_in));
3138				sin->sin_port = 0;
3139				break;
3140			}
3141		case AF_INET6:
3142			{
3143				struct sockaddr_in6 *sin6;
3144
3145				sin6 = (struct sockaddr_in6 *)&store_sa;
3146				memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3147				sin6->sin6_port = 0;
3148				break;
3149			}
3150		default:
3151			break;
3152		}
3153		/*
3154		 * first find the interface with the bound address need to
3155		 * zero out the port to find the address! yuck! can't do
3156		 * this earlier since need port for sctp_pcb_findep()
3157		 */
3158		if (sctp_ifap != NULL)
3159			ifa = sctp_ifap;
3160		else {
3161			/*
3162			 * Note for BSD we hit here always other O/S's will
3163			 * pass things in via the sctp_ifap argument
3164			 * (Panda).
3165			 */
3166			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3167			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3168		}
3169		if (ifa == NULL) {
3170			/* Can't find an interface with that address */
3171			SCTP_INP_WUNLOCK(inp);
3172			SCTP_INP_INFO_WUNLOCK();
3173			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3174			return (EADDRNOTAVAIL);
3175		}
3176#ifdef INET6
3177		if (addr->sa_family == AF_INET6) {
3178			/* GAK, more FIXME IFA lock? */
3179			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3180				/* Can't bind a non-existent addr. */
3181				SCTP_INP_WUNLOCK(inp);
3182				SCTP_INP_INFO_WUNLOCK();
3183				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3184				return (EINVAL);
3185			}
3186		}
3187#endif
3188		/* we're not bound all */
3189		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3190		/* allow bindx() to send ASCONF's for binding changes */
3191		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3192		/* clear automatic addr changes from kernel flag */
3193		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3194
3195		/* add this address to the endpoint list */
3196		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3197		if (error != 0) {
3198			SCTP_INP_WUNLOCK(inp);
3199			SCTP_INP_INFO_WUNLOCK();
3200			return (error);
3201		}
3202		inp->laddr_count++;
3203	}
3204	/* find the bucket */
3205	if (port_reuse_active) {
3206		/* Put it into tcp 1-2-1 hash */
3207		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3208		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3209	} else {
3210		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3211	}
3212	/* put it in the bucket */
3213	LIST_INSERT_HEAD(head, inp, sctp_hash);
3214	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3215	    head, ntohs(lport), port_reuse_active);
3216	/* set in the port */
3217	inp->sctp_lport = lport;
3218
3219	/* turn off just the unbound flag */
3220	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3221	SCTP_INP_WUNLOCK(inp);
3222	SCTP_INP_INFO_WUNLOCK();
3223	return (0);
3224}
3225
3226
3227static void
3228sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3229{
3230	struct sctp_iterator *it, *nit;
3231
3232	/*
3233	 * We enter with the only the ITERATOR_LOCK in place and a write
3234	 * lock on the inp_info stuff.
3235	 */
3236	it = sctp_it_ctl.cur_it;
3237	if (it && (it->vn != curvnet)) {
3238		/* Its not looking at our VNET */
3239		return;
3240	}
3241	if (it && (it->inp == inp)) {
3242		/*
3243		 * This is tricky and we hold the iterator lock, but when it
3244		 * returns and gets the lock (when we release it) the
3245		 * iterator will try to operate on inp. We need to stop that
3246		 * from happening. But of course the iterator has a
3247		 * reference on the stcb and inp. We can mark it and it will
3248		 * stop.
3249		 *
3250		 * If its a single iterator situation, we set the end iterator
3251		 * flag. Otherwise we set the iterator to go to the next
3252		 * inp.
3253		 *
3254		 */
3255		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3256			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3257		} else {
3258			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3259		}
3260	}
3261	/*
3262	 * Now go through and remove any single reference to our inp that
3263	 * may be still pending on the list
3264	 */
3265	SCTP_IPI_ITERATOR_WQ_LOCK();
3266	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3267		if (it->vn != curvnet) {
3268			continue;
3269		}
3270		if (it->inp == inp) {
3271			/* This one points to me is it inp specific? */
3272			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3273				/* Remove and free this one */
3274				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3275				    it, sctp_nxt_itr);
3276				if (it->function_atend != NULL) {
3277					(*it->function_atend) (it->pointer, it->val);
3278				}
3279				SCTP_FREE(it, SCTP_M_ITER);
3280			} else {
3281				it->inp = LIST_NEXT(it->inp, sctp_list);
3282				if (it->inp) {
3283					SCTP_INP_INCR_REF(it->inp);
3284				}
3285			}
3286			/*
3287			 * When its put in the refcnt is incremented so decr
3288			 * it
3289			 */
3290			SCTP_INP_DECR_REF(inp);
3291		}
3292	}
3293	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3294}
3295
3296/* release sctp_inpcb unbind the port */
3297void
3298sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3299{
3300	/*
3301	 * Here we free a endpoint. We must find it (if it is in the Hash
3302	 * table) and remove it from there. Then we must also find it in the
3303	 * overall list and remove it from there. After all removals are
3304	 * complete then any timer has to be stopped. Then start the actual
3305	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3306	 * all associations. d) finally the ep itself.
3307	 */
3308	struct sctp_pcb *m;
3309	struct sctp_tcb *asoc, *nasoc;
3310	struct sctp_laddr *laddr, *nladdr;
3311	struct inpcb *ip_pcb;
3312	struct socket *so;
3313	int being_refed = 0;
3314	struct sctp_queued_to_read *sq, *nsq;
3315	int cnt;
3316	sctp_sharedkey_t *shared_key, *nshared_key;
3317
3318
3319#ifdef SCTP_LOG_CLOSING
3320	sctp_log_closing(inp, NULL, 0);
3321#endif
3322	SCTP_ITERATOR_LOCK();
3323	/* mark any iterators on the list or being processed */
3324	sctp_iterator_inp_being_freed(inp);
3325	SCTP_ITERATOR_UNLOCK();
3326	so = inp->sctp_socket;
3327	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3328		/* been here before.. eeks.. get out of here */
3329		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3330#ifdef SCTP_LOG_CLOSING
3331		sctp_log_closing(inp, NULL, 1);
3332#endif
3333		return;
3334	}
3335	SCTP_ASOC_CREATE_LOCK(inp);
3336	SCTP_INP_INFO_WLOCK();
3337
3338	SCTP_INP_WLOCK(inp);
3339	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3340		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3341		/* socket is gone, so no more wakeups allowed */
3342		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3343		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3344		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3345
3346	}
3347	/* First time through we have the socket lock, after that no more. */
3348	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3349	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3350
3351	if (inp->control) {
3352		sctp_m_freem(inp->control);
3353		inp->control = NULL;
3354	}
3355	if (inp->pkt) {
3356		sctp_m_freem(inp->pkt);
3357		inp->pkt = NULL;
3358	}
3359	m = &inp->sctp_ep;
3360	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3361					 * here but I will be nice :> (i.e.
3362					 * ip_pcb = ep;) */
3363	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3364		int cnt_in_sd;
3365
3366		cnt_in_sd = 0;
3367		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3368			SCTP_TCB_LOCK(asoc);
3369			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3370				/* Skip guys being freed */
3371				cnt_in_sd++;
3372				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3373					/*
3374					 * Special case - we did not start a
3375					 * kill timer on the asoc due to it
3376					 * was not closed. So go ahead and
3377					 * start it now.
3378					 */
3379					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3380					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3381				}
3382				SCTP_TCB_UNLOCK(asoc);
3383				continue;
3384			}
3385			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3386			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3387			    (asoc->asoc.total_output_queue_size == 0)) {
3388				/*
3389				 * If we have data in queue, we don't want
3390				 * to just free since the app may have done,
3391				 * send()/close or connect/send/close. And
3392				 * it wants the data to get across first.
3393				 */
3394				/* Just abandon things in the front states */
3395				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3396				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3397					cnt_in_sd++;
3398				}
3399				continue;
3400			}
3401			/* Disconnect the socket please */
3402			asoc->sctp_socket = NULL;
3403			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3404			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3405			    (asoc->asoc.control_pdapi) ||
3406			    (asoc->asoc.size_on_all_streams > 0) ||
3407			    (so && (so->so_rcv.sb_cc > 0))
3408			    ) {
3409				/* Left with Data unread */
3410				struct mbuf *op_err;
3411
3412				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3413				    0, M_DONTWAIT, 1, MT_DATA);
3414				if (op_err) {
3415					/* Fill in the user initiated abort */
3416					struct sctp_paramhdr *ph;
3417					uint32_t *ippp;
3418
3419					SCTP_BUF_LEN(op_err) =
3420					    sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
3421					ph = mtod(op_err,
3422					    struct sctp_paramhdr *);
3423					ph->param_type = htons(
3424					    SCTP_CAUSE_USER_INITIATED_ABT);
3425					ph->param_length = htons(SCTP_BUF_LEN(op_err));
3426					ippp = (uint32_t *) (ph + 1);
3427					*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3);
3428				}
3429				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3430#if defined(SCTP_PANIC_ON_ABORT)
3431				panic("inpcb_free does an abort");
3432#endif
3433				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3434				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3435				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3436				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3437					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3438				}
3439				if (sctp_free_assoc(inp, asoc,
3440				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3441					cnt_in_sd++;
3442				}
3443				continue;
3444			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3445				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3446				    (asoc->asoc.stream_queue_cnt == 0)
3447			    ) {
3448				if (asoc->asoc.locked_on_sending) {
3449					goto abort_anyway;
3450				}
3451				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3452				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3453					struct sctp_nets *netp;
3454
3455					if (asoc->asoc.alternate) {
3456						netp = asoc->asoc.alternate;
3457					} else {
3458						netp = asoc->asoc.primary_destination;
3459					}
3460					/*
3461					 * there is nothing queued to send,
3462					 * so I send shutdown
3463					 */
3464					sctp_send_shutdown(asoc, netp);
3465					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3466					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3467						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3468					}
3469					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3470					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3471					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3472					    netp);
3473					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3474					    asoc->asoc.primary_destination);
3475					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3476				}
3477			} else {
3478				/* mark into shutdown pending */
3479				struct sctp_stream_queue_pending *sp;
3480
3481				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3482				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3483				    asoc->asoc.primary_destination);
3484				if (asoc->asoc.locked_on_sending) {
3485					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3486					    sctp_streamhead);
3487					if (sp == NULL) {
3488						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3489						    asoc->asoc.locked_on_sending,
3490						    asoc->asoc.locked_on_sending->stream_no);
3491					} else {
3492						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3493							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3494					}
3495				}
3496				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3497				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3498				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3499					struct mbuf *op_err;
3500
3501			abort_anyway:
3502					op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3503					    0, M_DONTWAIT, 1, MT_DATA);
3504					if (op_err) {
3505						/*
3506						 * Fill in the user
3507						 * initiated abort
3508						 */
3509						struct sctp_paramhdr *ph;
3510						uint32_t *ippp;
3511
3512						SCTP_BUF_LEN(op_err) =
3513						    (sizeof(struct sctp_paramhdr) +
3514						    sizeof(uint32_t));
3515						ph = mtod(op_err,
3516						    struct sctp_paramhdr *);
3517						ph->param_type = htons(
3518						    SCTP_CAUSE_USER_INITIATED_ABT);
3519						ph->param_length = htons(SCTP_BUF_LEN(op_err));
3520						ippp = (uint32_t *) (ph + 1);
3521						*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5);
3522					}
3523					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3524#if defined(SCTP_PANIC_ON_ABORT)
3525					panic("inpcb_free does an abort");
3526#endif
3527
3528					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3529					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3530					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3531					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3532						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3533					}
3534					if (sctp_free_assoc(inp, asoc,
3535					    SCTP_PCBFREE_NOFORCE,
3536					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3537						cnt_in_sd++;
3538					}
3539					continue;
3540				} else {
3541					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3542				}
3543			}
3544			cnt_in_sd++;
3545			SCTP_TCB_UNLOCK(asoc);
3546		}
3547		/* now is there some left in our SHUTDOWN state? */
3548		if (cnt_in_sd) {
3549#ifdef SCTP_LOG_CLOSING
3550			sctp_log_closing(inp, NULL, 2);
3551#endif
3552			inp->sctp_socket = NULL;
3553			SCTP_INP_WUNLOCK(inp);
3554			SCTP_ASOC_CREATE_UNLOCK(inp);
3555			SCTP_INP_INFO_WUNLOCK();
3556			return;
3557		}
3558	}
3559	inp->sctp_socket = NULL;
3560	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3561	    SCTP_PCB_FLAGS_UNBOUND) {
3562		/*
3563		 * ok, this guy has been bound. It's port is somewhere in
3564		 * the SCTP_BASE_INFO(hash table). Remove it!
3565		 */
3566		LIST_REMOVE(inp, sctp_hash);
3567		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3568	}
3569	/*
3570	 * If there is a timer running to kill us, forget it, since it may
3571	 * have a contest on the INP lock.. which would cause us to die ...
3572	 */
3573	cnt = 0;
3574	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3575		SCTP_TCB_LOCK(asoc);
3576		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3577			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3578				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3579				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3580			}
3581			cnt++;
3582			SCTP_TCB_UNLOCK(asoc);
3583			continue;
3584		}
3585		/* Free associations that are NOT killing us */
3586		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3587		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3588			struct mbuf *op_err;
3589			uint32_t *ippp;
3590
3591			op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3592			    0, M_DONTWAIT, 1, MT_DATA);
3593			if (op_err) {
3594				/* Fill in the user initiated abort */
3595				struct sctp_paramhdr *ph;
3596
3597				SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) +
3598				    sizeof(uint32_t));
3599				ph = mtod(op_err, struct sctp_paramhdr *);
3600				ph->param_type = htons(
3601				    SCTP_CAUSE_USER_INITIATED_ABT);
3602				ph->param_length = htons(SCTP_BUF_LEN(op_err));
3603				ippp = (uint32_t *) (ph + 1);
3604				*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7);
3605
3606			}
3607			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3608#if defined(SCTP_PANIC_ON_ABORT)
3609			panic("inpcb_free does an abort");
3610#endif
3611			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3612			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3613		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3614			cnt++;
3615			SCTP_TCB_UNLOCK(asoc);
3616			continue;
3617		}
3618		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3619		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3620			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3621		}
3622		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3623			cnt++;
3624		}
3625	}
3626	if (cnt) {
3627		/* Ok we have someone out there that will kill us */
3628		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3629#ifdef SCTP_LOG_CLOSING
3630		sctp_log_closing(inp, NULL, 3);
3631#endif
3632		SCTP_INP_WUNLOCK(inp);
3633		SCTP_ASOC_CREATE_UNLOCK(inp);
3634		SCTP_INP_INFO_WUNLOCK();
3635		return;
3636	}
3637	if (SCTP_INP_LOCK_CONTENDED(inp))
3638		being_refed++;
3639	if (SCTP_INP_READ_CONTENDED(inp))
3640		being_refed++;
3641	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3642		being_refed++;
3643
3644	if ((inp->refcount) ||
3645	    (being_refed) ||
3646	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3647		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3648#ifdef SCTP_LOG_CLOSING
3649		sctp_log_closing(inp, NULL, 4);
3650#endif
3651		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3652		SCTP_INP_WUNLOCK(inp);
3653		SCTP_ASOC_CREATE_UNLOCK(inp);
3654		SCTP_INP_INFO_WUNLOCK();
3655		return;
3656	}
3657	inp->sctp_ep.signature_change.type = 0;
3658	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3659	/*
3660	 * Remove it from the list .. last thing we need a lock for.
3661	 */
3662	LIST_REMOVE(inp, sctp_list);
3663	SCTP_INP_WUNLOCK(inp);
3664	SCTP_ASOC_CREATE_UNLOCK(inp);
3665	SCTP_INP_INFO_WUNLOCK();
3666	/*
3667	 * Now we release all locks. Since this INP cannot be found anymore
3668	 * except possibly by the kill timer that might be running. We call
3669	 * the drain function here. It should hit the case were it sees the
3670	 * ACTIVE flag cleared and exit out freeing us to proceed and
3671	 * destroy everything.
3672	 */
3673	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3674		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3675	} else {
3676		/* Probably un-needed */
3677		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3678	}
3679
3680#ifdef SCTP_LOG_CLOSING
3681	sctp_log_closing(inp, NULL, 5);
3682#endif
3683
3684
3685	if ((inp->sctp_asocidhash) != NULL) {
3686		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3687		inp->sctp_asocidhash = NULL;
3688	}
3689	/* sa_ignore FREED_MEMORY */
3690	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3691		/* Its only abandoned if it had data left */
3692		if (sq->length)
3693			SCTP_STAT_INCR(sctps_left_abandon);
3694
3695		TAILQ_REMOVE(&inp->read_queue, sq, next);
3696		sctp_free_remote_addr(sq->whoFrom);
3697		if (so)
3698			so->so_rcv.sb_cc -= sq->length;
3699		if (sq->data) {
3700			sctp_m_freem(sq->data);
3701			sq->data = NULL;
3702		}
3703		/*
3704		 * no need to free the net count, since at this point all
3705		 * assoc's are gone.
3706		 */
3707		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3708		SCTP_DECR_READQ_COUNT();
3709	}
3710	/* Now the sctp_pcb things */
3711	/*
3712	 * free each asoc if it is not already closed/free. we can't use the
3713	 * macro here since le_next will get freed as part of the
3714	 * sctp_free_assoc() call.
3715	 */
3716	cnt = 0;
3717	if (so) {
3718#ifdef IPSEC
3719		ipsec_delete_pcbpolicy(ip_pcb);
3720#endif				/* IPSEC */
3721
3722		/* Unlocks not needed since the socket is gone now */
3723	}
3724	if (ip_pcb->inp_options) {
3725		(void)sctp_m_free(ip_pcb->inp_options);
3726		ip_pcb->inp_options = 0;
3727	}
3728#ifdef INET6
3729	if (ip_pcb->inp_vflag & INP_IPV6) {
3730		struct in6pcb *in6p;
3731
3732		in6p = (struct in6pcb *)inp;
3733		ip6_freepcbopts(in6p->in6p_outputopts);
3734	}
3735#endif				/* INET6 */
3736	ip_pcb->inp_vflag = 0;
3737	/* free up authentication fields */
3738	if (inp->sctp_ep.local_auth_chunks != NULL)
3739		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3740	if (inp->sctp_ep.local_hmacs != NULL)
3741		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3742
3743	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3744		LIST_REMOVE(shared_key, next);
3745		sctp_free_sharedkey(shared_key);
3746		/* sa_ignore FREED_MEMORY */
3747	}
3748
3749	/*
3750	 * if we have an address list the following will free the list of
3751	 * ifaddr's that are set into this ep. Again macro limitations here,
3752	 * since the LIST_FOREACH could be a bad idea.
3753	 */
3754	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3755		sctp_remove_laddr(laddr);
3756	}
3757
3758#ifdef SCTP_TRACK_FREED_ASOCS
3759	/* TEMP CODE */
3760	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3761		LIST_REMOVE(asoc, sctp_tcblist);
3762		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3763		SCTP_DECR_ASOC_COUNT();
3764	}
3765	/* *** END TEMP CODE *** */
3766#endif
3767	/* Now lets see about freeing the EP hash table. */
3768	if (inp->sctp_tcbhash != NULL) {
3769		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3770		inp->sctp_tcbhash = NULL;
3771	}
3772	/* Now we must put the ep memory back into the zone pool */
3773	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3774	SCTP_INP_LOCK_DESTROY(inp);
3775	SCTP_INP_READ_DESTROY(inp);
3776	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3777	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3778	SCTP_DECR_EP_COUNT();
3779}
3780
3781
3782struct sctp_nets *
3783sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3784{
3785	struct sctp_nets *net;
3786
3787	/* locate the address */
3788	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3789		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3790			return (net);
3791	}
3792	return (NULL);
3793}
3794
3795
3796int
3797sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3798{
3799	struct sctp_ifa *sctp_ifa;
3800
3801	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3802	if (sctp_ifa) {
3803		return (1);
3804	} else {
3805		return (0);
3806	}
3807}
3808
3809/*
3810 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3811 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3812 * stats of stuff.
3813 */
3814int
3815sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3816    struct sctp_nets **netp, int set_scope, int from)
3817{
3818	/*
3819	 * The following is redundant to the same lines in the
3820	 * sctp_aloc_assoc() but is needed since others call the add address
3821	 * function
3822	 */
3823	struct sctp_nets *net, *netfirst;
3824	int addr_inscope;
3825
3826	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3827	    from);
3828	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3829
3830	netfirst = sctp_findnet(stcb, newaddr);
3831	if (netfirst) {
3832		/*
3833		 * Lie and return ok, we don't want to make the association
3834		 * go away for this behavior. It will happen in the TCP
3835		 * model in a connected socket. It does not reach the hash
3836		 * table until after the association is built so it can't be
3837		 * found. Mark as reachable, since the initial creation will
3838		 * have been cleared and the NOT_IN_ASSOC flag will have
3839		 * been added... and we don't want to end up removing it
3840		 * back out.
3841		 */
3842		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3843			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3844			    SCTP_ADDR_UNCONFIRMED);
3845		} else {
3846			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3847		}
3848
3849		return (0);
3850	}
3851	addr_inscope = 1;
3852	switch (newaddr->sa_family) {
3853#ifdef INET
3854	case AF_INET:
3855		{
3856			struct sockaddr_in *sin;
3857
3858			sin = (struct sockaddr_in *)newaddr;
3859			if (sin->sin_addr.s_addr == 0) {
3860				/* Invalid address */
3861				return (-1);
3862			}
3863			/* zero out the bzero area */
3864			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3865
3866			/* assure len is set */
3867			sin->sin_len = sizeof(struct sockaddr_in);
3868			if (set_scope) {
3869#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3870				stcb->ipv4_local_scope = 1;
3871#else
3872				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3873					stcb->asoc.ipv4_local_scope = 1;
3874				}
3875#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3876			} else {
3877				/* Validate the address is in scope */
3878				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3879				    (stcb->asoc.ipv4_local_scope == 0)) {
3880					addr_inscope = 0;
3881				}
3882			}
3883			break;
3884		}
3885#endif
3886#ifdef INET6
3887	case AF_INET6:
3888		{
3889			struct sockaddr_in6 *sin6;
3890
3891			sin6 = (struct sockaddr_in6 *)newaddr;
3892			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3893				/* Invalid address */
3894				return (-1);
3895			}
3896			/* assure len is set */
3897			sin6->sin6_len = sizeof(struct sockaddr_in6);
3898			if (set_scope) {
3899				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3900					stcb->asoc.loopback_scope = 1;
3901					stcb->asoc.local_scope = 0;
3902					stcb->asoc.ipv4_local_scope = 1;
3903					stcb->asoc.site_scope = 1;
3904				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3905					/*
3906					 * If the new destination is a
3907					 * LINK_LOCAL we must have common
3908					 * site scope. Don't set the local
3909					 * scope since we may not share all
3910					 * links, only loopback can do this.
3911					 * Links on the local network would
3912					 * also be on our private network
3913					 * for v4 too.
3914					 */
3915					stcb->asoc.ipv4_local_scope = 1;
3916					stcb->asoc.site_scope = 1;
3917				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3918					/*
3919					 * If the new destination is
3920					 * SITE_LOCAL then we must have site
3921					 * scope in common.
3922					 */
3923					stcb->asoc.site_scope = 1;
3924				}
3925			} else {
3926				/* Validate the address is in scope */
3927				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3928				    (stcb->asoc.loopback_scope == 0)) {
3929					addr_inscope = 0;
3930				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3931				    (stcb->asoc.local_scope == 0)) {
3932					addr_inscope = 0;
3933				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3934				    (stcb->asoc.site_scope == 0)) {
3935					addr_inscope = 0;
3936				}
3937			}
3938			break;
3939		}
3940#endif
3941	default:
3942		/* not supported family type */
3943		return (-1);
3944	}
3945	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3946	if (net == NULL) {
3947		return (-1);
3948	}
3949	SCTP_INCR_RADDR_COUNT();
3950	bzero(net, sizeof(struct sctp_nets));
3951	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3952	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3953	switch (newaddr->sa_family) {
3954#ifdef INET
3955	case AF_INET:
3956		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3957		break;
3958#endif
3959#ifdef INET6
3960	case AF_INET6:
3961		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3962		break;
3963#endif
3964	default:
3965		break;
3966	}
3967	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3968	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3969		stcb->asoc.loopback_scope = 1;
3970		stcb->asoc.ipv4_local_scope = 1;
3971		stcb->asoc.local_scope = 0;
3972		stcb->asoc.site_scope = 1;
3973		addr_inscope = 1;
3974	}
3975	net->failure_threshold = stcb->asoc.def_net_failure;
3976	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3977	if (addr_inscope == 0) {
3978		net->dest_state = (SCTP_ADDR_REACHABLE |
3979		    SCTP_ADDR_OUT_OF_SCOPE);
3980	} else {
3981		if (from == SCTP_ADDR_IS_CONFIRMED)
3982			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3983			net->dest_state = SCTP_ADDR_REACHABLE;
3984		else
3985			net->dest_state = SCTP_ADDR_REACHABLE |
3986			    SCTP_ADDR_UNCONFIRMED;
3987	}
3988	/*
3989	 * We set this to 0, the timer code knows that this means its an
3990	 * initial value
3991	 */
3992	net->rto_needed = 1;
3993	net->RTO = 0;
3994	net->RTO_measured = 0;
3995	stcb->asoc.numnets++;
3996	*(&net->ref_count) = 1;
3997	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3998	if (SCTP_BASE_SYSCTL(sctp_udp_tunneling_for_client_enable)) {
3999		net->port = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port));
4000	} else {
4001		net->port = 0;
4002	}
4003	net->dscp = stcb->asoc.default_dscp;
4004#ifdef INET6
4005	net->flowlabel = stcb->asoc.default_flowlabel;
4006#endif
4007	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
4008		net->dest_state |= SCTP_ADDR_NOHB;
4009	} else {
4010		net->dest_state &= ~SCTP_ADDR_NOHB;
4011	}
4012	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
4013		net->dest_state |= SCTP_ADDR_NO_PMTUD;
4014	} else {
4015		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
4016	}
4017	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
4018	/* Init the timer structure */
4019	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
4020	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
4021	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
4022
4023	/* Now generate a route for this guy */
4024#ifdef INET6
4025	/* KAME hack: embed scopeid */
4026	if (newaddr->sa_family == AF_INET6) {
4027		struct sockaddr_in6 *sin6;
4028
4029		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4030		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
4031		sin6->sin6_scope_id = 0;
4032	}
4033#endif
4034	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
4035
4036	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
4037		/* Get source address */
4038		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
4039		    stcb,
4040		    (sctp_route_t *) & net->ro,
4041		    net,
4042		    0,
4043		    stcb->asoc.vrf_id);
4044		/* Now get the interface MTU */
4045		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
4046			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
4047		}
4048		if (net->mtu > 0) {
4049			uint32_t rmtu;
4050
4051			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
4052			if (rmtu == 0) {
4053				/*
4054				 * Start things off to match mtu of
4055				 * interface please.
4056				 */
4057				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
4058				    net->ro.ro_rt, net->mtu);
4059			} else {
4060				/*
4061				 * we take the route mtu over the interface,
4062				 * since the route may be leading out the
4063				 * loopback, or a different interface.
4064				 */
4065				net->mtu = rmtu;
4066			}
4067		}
4068	}
4069	if (net->mtu == 0) {
4070		switch (newaddr->sa_family) {
4071#ifdef INET
4072		case AF_INET:
4073			net->mtu = SCTP_DEFAULT_MTU;
4074			break;
4075#endif
4076#ifdef INET6
4077		case AF_INET6:
4078			net->mtu = 1280;
4079			break;
4080#endif
4081		default:
4082			break;
4083		}
4084	}
4085	if (net->port) {
4086		net->mtu -= (uint32_t) sizeof(struct udphdr);
4087	}
4088	if (from == SCTP_ALLOC_ASOC) {
4089		stcb->asoc.smallest_mtu = net->mtu;
4090	}
4091	if (stcb->asoc.smallest_mtu > net->mtu) {
4092		stcb->asoc.smallest_mtu = net->mtu;
4093	}
4094#ifdef INET6
4095	if (newaddr->sa_family == AF_INET6) {
4096		struct sockaddr_in6 *sin6;
4097
4098		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4099		(void)sa6_recoverscope(sin6);
4100	}
4101#endif
4102
4103	/* JRS - Use the congestion control given in the CC module */
4104	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4105		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4106
4107	/*
4108	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4109	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4110	 */
4111	net->find_pseudo_cumack = 1;
4112	net->find_rtx_pseudo_cumack = 1;
4113	net->src_addr_selected = 0;
4114	/* Choose an initial flowid. */
4115	net->flowid = stcb->asoc.my_vtag ^
4116	    ntohs(stcb->rport) ^
4117	    ntohs(stcb->sctp_ep->sctp_lport);
4118#ifdef INVARIANTS
4119	net->flowidset = 1;
4120#endif
4121	if (netp) {
4122		*netp = net;
4123	}
4124	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4125	if (net->ro.ro_rt == NULL) {
4126		/* Since we have no route put it at the back */
4127		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4128	} else if (netfirst == NULL) {
4129		/* We are the first one in the pool. */
4130		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4131	} else if (netfirst->ro.ro_rt == NULL) {
4132		/*
4133		 * First one has NO route. Place this one ahead of the first
4134		 * one.
4135		 */
4136		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4137	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4138		/*
4139		 * This one has a different interface than the one at the
4140		 * top of the list. Place it ahead.
4141		 */
4142		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4143	} else {
4144		/*
4145		 * Ok we have the same interface as the first one. Move
4146		 * forward until we find either a) one with a NULL route...
4147		 * insert ahead of that b) one with a different ifp.. insert
4148		 * after that. c) end of the list.. insert at the tail.
4149		 */
4150		struct sctp_nets *netlook;
4151
4152		do {
4153			netlook = TAILQ_NEXT(netfirst, sctp_next);
4154			if (netlook == NULL) {
4155				/* End of the list */
4156				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4157				break;
4158			} else if (netlook->ro.ro_rt == NULL) {
4159				/* next one has NO route */
4160				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4161				break;
4162			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4163				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4164				    net, sctp_next);
4165				break;
4166			}
4167			/* Shift forward */
4168			netfirst = netlook;
4169		} while (netlook != NULL);
4170	}
4171
4172	/* got to have a primary set */
4173	if (stcb->asoc.primary_destination == 0) {
4174		stcb->asoc.primary_destination = net;
4175	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4176		    (net->ro.ro_rt) &&
4177	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4178		/* No route to current primary adopt new primary */
4179		stcb->asoc.primary_destination = net;
4180	}
4181	/* Validate primary is first */
4182	net = TAILQ_FIRST(&stcb->asoc.nets);
4183	if ((net != stcb->asoc.primary_destination) &&
4184	    (stcb->asoc.primary_destination)) {
4185		/*
4186		 * first one on the list is NOT the primary sctp_cmpaddr()
4187		 * is much more efficient if the primary is the first on the
4188		 * list, make it so.
4189		 */
4190		TAILQ_REMOVE(&stcb->asoc.nets,
4191		    stcb->asoc.primary_destination, sctp_next);
4192		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4193		    stcb->asoc.primary_destination, sctp_next);
4194	}
4195	return (0);
4196}
4197
4198
4199static uint32_t
4200sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4201{
4202	uint32_t id;
4203	struct sctpasochead *head;
4204	struct sctp_tcb *lstcb;
4205
4206	SCTP_INP_WLOCK(inp);
4207try_again:
4208	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4209		/* TSNH */
4210		SCTP_INP_WUNLOCK(inp);
4211		return (0);
4212	}
4213	/*
4214	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4215	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4216	 */
4217	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4218		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4219	}
4220	id = inp->sctp_associd_counter;
4221	inp->sctp_associd_counter++;
4222	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4223	if (lstcb) {
4224		goto try_again;
4225	}
4226	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4227	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4228	stcb->asoc.in_asocid_hash = 1;
4229	SCTP_INP_WUNLOCK(inp);
4230	return id;
4231}
4232
4233/*
4234 * allocate an association and add it to the endpoint. The caller must be
4235 * careful to add all additional addresses once they are know right away or
4236 * else the assoc will be may experience a blackout scenario.
4237 */
4238struct sctp_tcb *
4239sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4240    int *error, uint32_t override_tag, uint32_t vrf_id,
4241    struct thread *p
4242)
4243{
4244	/* note the p argument is only valid in unbound sockets */
4245
4246	struct sctp_tcb *stcb;
4247	struct sctp_association *asoc;
4248	struct sctpasochead *head;
4249	uint16_t rport;
4250	int err;
4251
4252	/*
4253	 * Assumption made here: Caller has done a
4254	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4255	 * address does not exist already.
4256	 */
4257	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4258		/* Hit max assoc, sorry no more */
4259		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4260		*error = ENOBUFS;
4261		return (NULL);
4262	}
4263	if (firstaddr == NULL) {
4264		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4265		*error = EINVAL;
4266		return (NULL);
4267	}
4268	SCTP_INP_RLOCK(inp);
4269	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4270	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4271	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4272		/*
4273		 * If its in the TCP pool, its NOT allowed to create an
4274		 * association. The parent listener needs to call
4275		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4276		 * off, or connected one does this.. its an error.
4277		 */
4278		SCTP_INP_RUNLOCK(inp);
4279		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4280		*error = EINVAL;
4281		return (NULL);
4282	}
4283	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4284	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4285		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4286		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4287			SCTP_INP_RUNLOCK(inp);
4288			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4289			*error = EINVAL;
4290			return (NULL);
4291		}
4292	}
4293	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4294#ifdef SCTP_DEBUG
4295	if (firstaddr) {
4296		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4297		switch (firstaddr->sa_family) {
4298#ifdef INET
4299		case AF_INET:
4300			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4301			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4302			break;
4303#endif
4304#ifdef INET6
4305		case AF_INET6:
4306			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4307			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4308			break;
4309#endif
4310		default:
4311			break;
4312		}
4313	} else {
4314		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4315	}
4316#endif				/* SCTP_DEBUG */
4317	switch (firstaddr->sa_family) {
4318#ifdef INET
4319	case AF_INET:
4320		{
4321			struct sockaddr_in *sin;
4322
4323			sin = (struct sockaddr_in *)firstaddr;
4324			if ((ntohs(sin->sin_port) == 0) ||
4325			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4326			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4327			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4328				/* Invalid address */
4329				SCTP_INP_RUNLOCK(inp);
4330				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4331				*error = EINVAL;
4332				return (NULL);
4333			}
4334			rport = sin->sin_port;
4335			break;
4336		}
4337#endif
4338#ifdef INET6
4339	case AF_INET6:
4340		{
4341			struct sockaddr_in6 *sin6;
4342
4343			sin6 = (struct sockaddr_in6 *)firstaddr;
4344			if ((ntohs(sin6->sin6_port) == 0) ||
4345			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4346			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4347				/* Invalid address */
4348				SCTP_INP_RUNLOCK(inp);
4349				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4350				*error = EINVAL;
4351				return (NULL);
4352			}
4353			rport = sin6->sin6_port;
4354			break;
4355		}
4356#endif
4357	default:
4358		/* not supported family type */
4359		SCTP_INP_RUNLOCK(inp);
4360		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4361		*error = EINVAL;
4362		return (NULL);
4363	}
4364	SCTP_INP_RUNLOCK(inp);
4365	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4366		/*
4367		 * If you have not performed a bind, then we need to do the
4368		 * ephemeral bind for you.
4369		 */
4370		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4371		    (struct sockaddr *)NULL,
4372		    (struct sctp_ifa *)NULL,
4373		    p
4374		    ))) {
4375			/* bind error, probably perm */
4376			*error = err;
4377			return (NULL);
4378		}
4379	}
4380	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4381	if (stcb == NULL) {
4382		/* out of memory? */
4383		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4384		*error = ENOMEM;
4385		return (NULL);
4386	}
4387	SCTP_INCR_ASOC_COUNT();
4388
4389	bzero(stcb, sizeof(*stcb));
4390	asoc = &stcb->asoc;
4391
4392	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4393	SCTP_TCB_LOCK_INIT(stcb);
4394	SCTP_TCB_SEND_LOCK_INIT(stcb);
4395	stcb->rport = rport;
4396	/* setup back pointer's */
4397	stcb->sctp_ep = inp;
4398	stcb->sctp_socket = inp->sctp_socket;
4399	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4400		/* failed */
4401		SCTP_TCB_LOCK_DESTROY(stcb);
4402		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4403		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4404		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4405		SCTP_DECR_ASOC_COUNT();
4406		*error = err;
4407		return (NULL);
4408	}
4409	/* and the port */
4410	SCTP_INP_INFO_WLOCK();
4411	SCTP_INP_WLOCK(inp);
4412	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4413		/* inpcb freed while alloc going on */
4414		SCTP_TCB_LOCK_DESTROY(stcb);
4415		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4416		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4417		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4418		SCTP_INP_WUNLOCK(inp);
4419		SCTP_INP_INFO_WUNLOCK();
4420		SCTP_DECR_ASOC_COUNT();
4421		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4422		*error = EINVAL;
4423		return (NULL);
4424	}
4425	SCTP_TCB_LOCK(stcb);
4426
4427	/* now that my_vtag is set, add it to the hash */
4428	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4429	/* put it in the bucket in the vtag hash of assoc's for the system */
4430	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4431	SCTP_INP_INFO_WUNLOCK();
4432
4433	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4434		/* failure.. memory error? */
4435		if (asoc->strmout) {
4436			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4437			asoc->strmout = NULL;
4438		}
4439		if (asoc->mapping_array) {
4440			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4441			asoc->mapping_array = NULL;
4442		}
4443		if (asoc->nr_mapping_array) {
4444			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4445			asoc->nr_mapping_array = NULL;
4446		}
4447		SCTP_DECR_ASOC_COUNT();
4448		SCTP_TCB_LOCK_DESTROY(stcb);
4449		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4450		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4451		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4452		SCTP_INP_WUNLOCK(inp);
4453		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4454		*error = ENOBUFS;
4455		return (NULL);
4456	}
4457	/* Init all the timers */
4458	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4459	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4460	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4461	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4462	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4463	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4464	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4465
4466	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4467	/* now file the port under the hash as well */
4468	if (inp->sctp_tcbhash != NULL) {
4469		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4470		    inp->sctp_hashmark)];
4471		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4472	}
4473	SCTP_INP_WUNLOCK(inp);
4474	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", stcb);
4475	return (stcb);
4476}
4477
4478
4479void
4480sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4481{
4482	struct sctp_association *asoc;
4483
4484	asoc = &stcb->asoc;
4485	asoc->numnets--;
4486	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4487	if (net == asoc->primary_destination) {
4488		/* Reset primary */
4489		struct sctp_nets *lnet;
4490
4491		lnet = TAILQ_FIRST(&asoc->nets);
4492		/*
4493		 * Mobility adaptation Ideally, if deleted destination is
4494		 * the primary, it becomes a fast retransmission trigger by
4495		 * the subsequent SET PRIMARY. (by micchie)
4496		 */
4497		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4498		    SCTP_MOBILITY_BASE) ||
4499		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4500		    SCTP_MOBILITY_FASTHANDOFF)) {
4501			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4502			if (asoc->deleted_primary != NULL) {
4503				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4504				goto out;
4505			}
4506			asoc->deleted_primary = net;
4507			atomic_add_int(&net->ref_count, 1);
4508			memset(&net->lastsa, 0, sizeof(net->lastsa));
4509			memset(&net->lastsv, 0, sizeof(net->lastsv));
4510			sctp_mobility_feature_on(stcb->sctp_ep,
4511			    SCTP_MOBILITY_PRIM_DELETED);
4512			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4513			    stcb->sctp_ep, stcb, NULL);
4514		}
4515out:
4516		/* Try to find a confirmed primary */
4517		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4518	}
4519	if (net == asoc->last_data_chunk_from) {
4520		/* Reset primary */
4521		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4522	}
4523	if (net == asoc->last_control_chunk_from) {
4524		/* Clear net */
4525		asoc->last_control_chunk_from = NULL;
4526	}
4527	if (net == stcb->asoc.alternate) {
4528		sctp_free_remote_addr(stcb->asoc.alternate);
4529		stcb->asoc.alternate = NULL;
4530	}
4531	sctp_free_remote_addr(net);
4532}
4533
4534/*
4535 * remove a remote endpoint address from an association, it will fail if the
4536 * address does not exist.
4537 */
4538int
4539sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4540{
4541	/*
4542	 * Here we need to remove a remote address. This is quite simple, we
4543	 * first find it in the list of address for the association
4544	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4545	 * on that item. Note we do not allow it to be removed if there are
4546	 * no other addresses.
4547	 */
4548	struct sctp_association *asoc;
4549	struct sctp_nets *net, *nnet;
4550
4551	asoc = &stcb->asoc;
4552
4553	/* locate the address */
4554	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4555		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4556			continue;
4557		}
4558		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4559		    remaddr)) {
4560			/* we found the guy */
4561			if (asoc->numnets < 2) {
4562				/* Must have at LEAST two remote addresses */
4563				return (-1);
4564			} else {
4565				sctp_remove_net(stcb, net);
4566				return (0);
4567			}
4568		}
4569	}
4570	/* not found. */
4571	return (-2);
4572}
4573
4574void
4575sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4576{
4577	struct sctpvtaghead *chain;
4578	struct sctp_tagblock *twait_block;
4579	int found = 0;
4580	int i;
4581
4582	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4583	if (!LIST_EMPTY(chain)) {
4584		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4585			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4586				if ((twait_block->vtag_block[i].v_tag == tag) &&
4587				    (twait_block->vtag_block[i].lport == lport) &&
4588				    (twait_block->vtag_block[i].rport == rport)) {
4589					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4590					twait_block->vtag_block[i].v_tag = 0;
4591					twait_block->vtag_block[i].lport = 0;
4592					twait_block->vtag_block[i].rport = 0;
4593					found = 1;
4594					break;
4595				}
4596			}
4597			if (found)
4598				break;
4599		}
4600	}
4601}
4602
4603int
4604sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4605{
4606	struct sctpvtaghead *chain;
4607	struct sctp_tagblock *twait_block;
4608	int found = 0;
4609	int i;
4610
4611	SCTP_INP_INFO_WLOCK();
4612	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4613	if (!LIST_EMPTY(chain)) {
4614		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4615			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4616				if ((twait_block->vtag_block[i].v_tag == tag) &&
4617				    (twait_block->vtag_block[i].lport == lport) &&
4618				    (twait_block->vtag_block[i].rport == rport)) {
4619					found = 1;
4620					break;
4621				}
4622			}
4623			if (found)
4624				break;
4625		}
4626	}
4627	SCTP_INP_INFO_WUNLOCK();
4628	return (found);
4629}
4630
4631
4632void
4633sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4634{
4635	struct sctpvtaghead *chain;
4636	struct sctp_tagblock *twait_block;
4637	struct timeval now;
4638	int set, i;
4639
4640	if (time == 0) {
4641		/* Its disabled */
4642		return;
4643	}
4644	(void)SCTP_GETTIME_TIMEVAL(&now);
4645	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4646	set = 0;
4647	if (!LIST_EMPTY(chain)) {
4648		/* Block(s) present, lets find space, and expire on the fly */
4649		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4650			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4651				if ((twait_block->vtag_block[i].v_tag == 0) &&
4652				    !set) {
4653					twait_block->vtag_block[i].tv_sec_at_expire =
4654					    now.tv_sec + time;
4655					twait_block->vtag_block[i].v_tag = tag;
4656					twait_block->vtag_block[i].lport = lport;
4657					twait_block->vtag_block[i].rport = rport;
4658					set = 1;
4659				} else if ((twait_block->vtag_block[i].v_tag) &&
4660				    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4661					/* Audit expires this guy */
4662					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4663					twait_block->vtag_block[i].v_tag = 0;
4664					twait_block->vtag_block[i].lport = 0;
4665					twait_block->vtag_block[i].rport = 0;
4666					if (set == 0) {
4667						/* Reuse it for my new tag */
4668						twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4669						twait_block->vtag_block[i].v_tag = tag;
4670						twait_block->vtag_block[i].lport = lport;
4671						twait_block->vtag_block[i].rport = rport;
4672						set = 1;
4673					}
4674				}
4675			}
4676			if (set) {
4677				/*
4678				 * We only do up to the block where we can
4679				 * place our tag for audits
4680				 */
4681				break;
4682			}
4683		}
4684	}
4685	/* Need to add a new block to chain */
4686	if (!set) {
4687		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4688		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4689		if (twait_block == NULL) {
4690#ifdef INVARIANTS
4691			panic("Can not alloc tagblock");
4692#endif
4693			return;
4694		}
4695		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4696		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4697		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4698		twait_block->vtag_block[0].v_tag = tag;
4699		twait_block->vtag_block[0].lport = lport;
4700		twait_block->vtag_block[0].rport = rport;
4701	}
4702}
4703
4704
4705
4706/*-
4707 * Free the association after un-hashing the remote port. This
4708 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4709 * expect that the input to this function IS a locked TCB.
4710 * It will return 0, if it did NOT destroy the association (instead
4711 * it unlocks it. It will return NON-zero if it either destroyed the
4712 * association OR the association is already destroyed.
4713 */
4714int
4715sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4716{
4717	int i;
4718	struct sctp_association *asoc;
4719	struct sctp_nets *net, *nnet;
4720	struct sctp_laddr *laddr, *naddr;
4721	struct sctp_tmit_chunk *chk, *nchk;
4722	struct sctp_asconf_addr *aparam, *naparam;
4723	struct sctp_asconf_ack *aack, *naack;
4724	struct sctp_stream_reset_list *strrst, *nstrrst;
4725	struct sctp_queued_to_read *sq, *nsq;
4726	struct sctp_stream_queue_pending *sp, *nsp;
4727	sctp_sharedkey_t *shared_key, *nshared_key;
4728	struct socket *so;
4729
4730	/* first, lets purge the entry from the hash table. */
4731
4732#ifdef SCTP_LOG_CLOSING
4733	sctp_log_closing(inp, stcb, 6);
4734#endif
4735	if (stcb->asoc.state == 0) {
4736#ifdef SCTP_LOG_CLOSING
4737		sctp_log_closing(inp, NULL, 7);
4738#endif
4739		/* there is no asoc, really TSNH :-0 */
4740		return (1);
4741	}
4742	if (stcb->asoc.alternate) {
4743		sctp_free_remote_addr(stcb->asoc.alternate);
4744		stcb->asoc.alternate = NULL;
4745	}
4746	/* TEMP CODE */
4747	if (stcb->freed_from_where == 0) {
4748		/* Only record the first place free happened from */
4749		stcb->freed_from_where = from_location;
4750	}
4751	/* TEMP CODE */
4752
4753	asoc = &stcb->asoc;
4754	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4755	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4756		/* nothing around */
4757		so = NULL;
4758	else
4759		so = inp->sctp_socket;
4760
4761	/*
4762	 * We used timer based freeing if a reader or writer is in the way.
4763	 * So we first check if we are actually being called from a timer,
4764	 * if so we abort early if a reader or writer is still in the way.
4765	 */
4766	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4767	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4768		/*
4769		 * is it the timer driving us? if so are the reader/writers
4770		 * gone?
4771		 */
4772		if (stcb->asoc.refcnt) {
4773			/* nope, reader or writer in the way */
4774			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4775			/* no asoc destroyed */
4776			SCTP_TCB_UNLOCK(stcb);
4777#ifdef SCTP_LOG_CLOSING
4778			sctp_log_closing(inp, stcb, 8);
4779#endif
4780			return (0);
4781		}
4782	}
4783	/* now clean up any other timers */
4784	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4785	asoc->dack_timer.self = NULL;
4786	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4787	/*-
4788	 * For stream reset we don't blast this unless
4789	 * it is a str-reset timer, it might be the
4790	 * free-asoc timer which we DON'T want to
4791	 * disturb.
4792	 */
4793	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4794		asoc->strreset_timer.self = NULL;
4795	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4796	asoc->asconf_timer.self = NULL;
4797	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4798	asoc->autoclose_timer.self = NULL;
4799	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4800	asoc->shut_guard_timer.self = NULL;
4801	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4802	asoc->delayed_event_timer.self = NULL;
4803	/* Mobility adaptation */
4804	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4805	asoc->delete_prim_timer.self = NULL;
4806	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4807		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4808		net->rxt_timer.self = NULL;
4809		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4810		net->pmtu_timer.self = NULL;
4811		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4812		net->hb_timer.self = NULL;
4813	}
4814	/* Now the read queue needs to be cleaned up (only once) */
4815	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4816		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4817		SCTP_INP_READ_LOCK(inp);
4818		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4819			if (sq->stcb == stcb) {
4820				sq->do_not_ref_stcb = 1;
4821				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4822				/*
4823				 * If there is no end, there never will be
4824				 * now.
4825				 */
4826				if (sq->end_added == 0) {
4827					/* Held for PD-API clear that. */
4828					sq->pdapi_aborted = 1;
4829					sq->held_length = 0;
4830					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4831						/*
4832						 * Need to add a PD-API
4833						 * aborted indication.
4834						 * Setting the control_pdapi
4835						 * assures that it will be
4836						 * added right after this
4837						 * msg.
4838						 */
4839						uint32_t strseq;
4840
4841						stcb->asoc.control_pdapi = sq;
4842						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4843						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4844						    stcb,
4845						    SCTP_PARTIAL_DELIVERY_ABORTED,
4846						    (void *)&strseq,
4847						    SCTP_SO_LOCKED);
4848						stcb->asoc.control_pdapi = NULL;
4849					}
4850				}
4851				/* Add an end to wake them */
4852				sq->end_added = 1;
4853			}
4854		}
4855		SCTP_INP_READ_UNLOCK(inp);
4856		if (stcb->block_entry) {
4857			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4858			stcb->block_entry->error = ECONNRESET;
4859			stcb->block_entry = NULL;
4860		}
4861	}
4862	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4863		/*
4864		 * Someone holds a reference OR the socket is unaccepted
4865		 * yet.
4866		 */
4867		if ((stcb->asoc.refcnt) ||
4868		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4869		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4870			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4871			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4872		}
4873		SCTP_TCB_UNLOCK(stcb);
4874		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4875		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4876			/* nothing around */
4877			so = NULL;
4878		if (so) {
4879			/* Wake any reader/writers */
4880			sctp_sorwakeup(inp, so);
4881			sctp_sowwakeup(inp, so);
4882		}
4883#ifdef SCTP_LOG_CLOSING
4884		sctp_log_closing(inp, stcb, 9);
4885#endif
4886		/* no asoc destroyed */
4887		return (0);
4888	}
4889#ifdef SCTP_LOG_CLOSING
4890	sctp_log_closing(inp, stcb, 10);
4891#endif
4892	/*
4893	 * When I reach here, no others want to kill the assoc yet.. and I
4894	 * own the lock. Now its possible an abort comes in when I do the
4895	 * lock exchange below to grab all the locks to do the final take
4896	 * out. to prevent this we increment the count, which will start a
4897	 * timer and blow out above thus assuring us that we hold exclusive
4898	 * killing of the asoc. Note that after getting back the TCB lock we
4899	 * will go ahead and increment the counter back up and stop any
4900	 * timer a passing stranger may have started :-S
4901	 */
4902	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4903		atomic_add_int(&stcb->asoc.refcnt, 1);
4904
4905		SCTP_TCB_UNLOCK(stcb);
4906		SCTP_INP_INFO_WLOCK();
4907		SCTP_INP_WLOCK(inp);
4908		SCTP_TCB_LOCK(stcb);
4909	}
4910	/* Double check the GONE flag */
4911	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4912	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4913		/* nothing around */
4914		so = NULL;
4915
4916	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4917	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4918		/*
4919		 * For TCP type we need special handling when we are
4920		 * connected. We also include the peel'ed off ones to.
4921		 */
4922		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4923			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4924			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4925			if (so) {
4926				SOCK_LOCK(so);
4927				if (so->so_rcv.sb_cc == 0) {
4928					so->so_state &= ~(SS_ISCONNECTING |
4929					    SS_ISDISCONNECTING |
4930					    SS_ISCONFIRMING |
4931					    SS_ISCONNECTED);
4932				}
4933				socantrcvmore_locked(so);
4934				sctp_sowwakeup(inp, so);
4935				sctp_sorwakeup(inp, so);
4936				SCTP_SOWAKEUP(so);
4937			}
4938		}
4939	}
4940	/*
4941	 * Make it invalid too, that way if its about to run it will abort
4942	 * and return.
4943	 */
4944	/* re-increment the lock */
4945	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4946		atomic_add_int(&stcb->asoc.refcnt, -1);
4947	}
4948	if (stcb->asoc.refcnt) {
4949		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4950		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4951		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4952			SCTP_INP_INFO_WUNLOCK();
4953			SCTP_INP_WUNLOCK(inp);
4954		}
4955		SCTP_TCB_UNLOCK(stcb);
4956		return (0);
4957	}
4958	asoc->state = 0;
4959	if (inp->sctp_tcbhash) {
4960		LIST_REMOVE(stcb, sctp_tcbhash);
4961	}
4962	if (stcb->asoc.in_asocid_hash) {
4963		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4964	}
4965	/* Now lets remove it from the list of ALL associations in the EP */
4966	LIST_REMOVE(stcb, sctp_tcblist);
4967	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4968		SCTP_INP_INCR_REF(inp);
4969		SCTP_INP_WUNLOCK(inp);
4970	}
4971	/* pull from vtag hash */
4972	LIST_REMOVE(stcb, sctp_asocs);
4973	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4974	    inp->sctp_lport, stcb->rport);
4975
4976	/*
4977	 * Now restop the timers to be sure this is paranoia at is finest!
4978	 */
4979	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4980	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4981	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4982	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4983	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4984	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4985	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4986	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4987		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4988		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4989		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4990	}
4991
4992	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4993	/*
4994	 * The chunk lists and such SHOULD be empty but we check them just
4995	 * in case.
4996	 */
4997	/* anything on the wheel needs to be removed */
4998	for (i = 0; i < asoc->streamoutcnt; i++) {
4999		struct sctp_stream_out *outs;
5000
5001		outs = &asoc->strmout[i];
5002		/* now clean up any chunks here */
5003		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
5004			TAILQ_REMOVE(&outs->outqueue, sp, next);
5005			if (sp->data) {
5006				if (so) {
5007					/* Still an open socket - report */
5008					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
5009					    SCTP_NOTIFY_DATAGRAM_UNSENT,
5010					    (void *)sp, SCTP_SO_LOCKED);
5011				}
5012				if (sp->data) {
5013					sctp_m_freem(sp->data);
5014					sp->data = NULL;
5015					sp->tail_mbuf = NULL;
5016				}
5017			}
5018			if (sp->net) {
5019				sctp_free_remote_addr(sp->net);
5020				sp->net = NULL;
5021			}
5022			sctp_free_spbufspace(stcb, asoc, sp);
5023			if (sp->holds_key_ref)
5024				sctp_auth_key_release(stcb, sp->auth_keyid, SCTP_SO_LOCKED);
5025			/* Free the zone stuff  */
5026			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_strmoq), sp);
5027			SCTP_DECR_STRMOQ_COUNT();
5028			/* sa_ignore FREED_MEMORY */
5029		}
5030	}
5031	/* sa_ignore FREED_MEMORY */
5032	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
5033		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
5034		SCTP_FREE(strrst, SCTP_M_STRESET);
5035	}
5036	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
5037		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
5038		if (sq->data) {
5039			sctp_m_freem(sq->data);
5040			sq->data = NULL;
5041		}
5042		sctp_free_remote_addr(sq->whoFrom);
5043		sq->whoFrom = NULL;
5044		sq->stcb = NULL;
5045		/* Free the ctl entry */
5046		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
5047		SCTP_DECR_READQ_COUNT();
5048		/* sa_ignore FREED_MEMORY */
5049	}
5050	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
5051		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
5052		if (chk->data) {
5053			sctp_m_freem(chk->data);
5054			chk->data = NULL;
5055		}
5056		if (chk->holds_key_ref)
5057			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5058		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5059		SCTP_DECR_CHK_COUNT();
5060		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
5061		asoc->free_chunk_cnt--;
5062		/* sa_ignore FREED_MEMORY */
5063	}
5064	/* pending send queue SHOULD be empty */
5065	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
5066		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
5067		if (chk->data) {
5068			if (so) {
5069				/* Still a socket? */
5070				sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb,
5071				    SCTP_NOTIFY_DATAGRAM_UNSENT, chk, SCTP_SO_LOCKED);
5072			}
5073			if (chk->data) {
5074				sctp_m_freem(chk->data);
5075				chk->data = NULL;
5076			}
5077		}
5078		if (chk->holds_key_ref)
5079			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5080		if (chk->whoTo) {
5081			sctp_free_remote_addr(chk->whoTo);
5082			chk->whoTo = NULL;
5083		}
5084		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5085		SCTP_DECR_CHK_COUNT();
5086		/* sa_ignore FREED_MEMORY */
5087	}
5088	/* sent queue SHOULD be empty */
5089	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5090		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5091		if (chk->data) {
5092			if (so) {
5093				/* Still a socket? */
5094				sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb,
5095				    SCTP_NOTIFY_DATAGRAM_SENT, chk, SCTP_SO_LOCKED);
5096			}
5097			if (chk->data) {
5098				sctp_m_freem(chk->data);
5099				chk->data = NULL;
5100			}
5101		}
5102		if (chk->holds_key_ref)
5103			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5104		sctp_free_remote_addr(chk->whoTo);
5105		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5106		SCTP_DECR_CHK_COUNT();
5107		/* sa_ignore FREED_MEMORY */
5108	}
5109	/* control queue MAY not be empty */
5110	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5111		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5112		if (chk->data) {
5113			sctp_m_freem(chk->data);
5114			chk->data = NULL;
5115		}
5116		if (chk->holds_key_ref)
5117			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5118		sctp_free_remote_addr(chk->whoTo);
5119		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5120		SCTP_DECR_CHK_COUNT();
5121		/* sa_ignore FREED_MEMORY */
5122	}
5123	/* ASCONF queue MAY not be empty */
5124	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5125		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5126		if (chk->data) {
5127			sctp_m_freem(chk->data);
5128			chk->data = NULL;
5129		}
5130		if (chk->holds_key_ref)
5131			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5132		sctp_free_remote_addr(chk->whoTo);
5133		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5134		SCTP_DECR_CHK_COUNT();
5135		/* sa_ignore FREED_MEMORY */
5136	}
5137	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5138		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5139		if (chk->data) {
5140			sctp_m_freem(chk->data);
5141			chk->data = NULL;
5142		}
5143		if (chk->holds_key_ref)
5144			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5145		sctp_free_remote_addr(chk->whoTo);
5146		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5147		SCTP_DECR_CHK_COUNT();
5148		/* sa_ignore FREED_MEMORY */
5149	}
5150
5151	if (asoc->mapping_array) {
5152		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5153		asoc->mapping_array = NULL;
5154	}
5155	if (asoc->nr_mapping_array) {
5156		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5157		asoc->nr_mapping_array = NULL;
5158	}
5159	/* the stream outs */
5160	if (asoc->strmout) {
5161		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5162		asoc->strmout = NULL;
5163	}
5164	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5165	if (asoc->strmin) {
5166		struct sctp_queued_to_read *ctl, *nctl;
5167
5168		for (i = 0; i < asoc->streamincnt; i++) {
5169			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5170				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5171				sctp_free_remote_addr(ctl->whoFrom);
5172				if (ctl->data) {
5173					sctp_m_freem(ctl->data);
5174					ctl->data = NULL;
5175				}
5176				/*
5177				 * We don't free the address here since all
5178				 * the net's were freed above.
5179				 */
5180				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5181				SCTP_DECR_READQ_COUNT();
5182			}
5183		}
5184		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5185		asoc->strmin = NULL;
5186	}
5187	asoc->streamincnt = 0;
5188	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5189#ifdef INVARIANTS
5190		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5191			panic("no net's left alloc'ed, or list points to itself");
5192		}
5193#endif
5194		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5195		sctp_free_remote_addr(net);
5196	}
5197	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5198		/* sa_ignore FREED_MEMORY */
5199		sctp_remove_laddr(laddr);
5200	}
5201
5202	/* pending asconf (address) parameters */
5203	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5204		/* sa_ignore FREED_MEMORY */
5205		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5206		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5207	}
5208	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5209		/* sa_ignore FREED_MEMORY */
5210		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5211		if (aack->data != NULL) {
5212			sctp_m_freem(aack->data);
5213		}
5214		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5215	}
5216	/* clean up auth stuff */
5217	if (asoc->local_hmacs)
5218		sctp_free_hmaclist(asoc->local_hmacs);
5219	if (asoc->peer_hmacs)
5220		sctp_free_hmaclist(asoc->peer_hmacs);
5221
5222	if (asoc->local_auth_chunks)
5223		sctp_free_chunklist(asoc->local_auth_chunks);
5224	if (asoc->peer_auth_chunks)
5225		sctp_free_chunklist(asoc->peer_auth_chunks);
5226
5227	sctp_free_authinfo(&asoc->authinfo);
5228
5229	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5230		LIST_REMOVE(shared_key, next);
5231		sctp_free_sharedkey(shared_key);
5232		/* sa_ignore FREED_MEMORY */
5233	}
5234
5235	/* Insert new items here :> */
5236
5237	/* Get rid of LOCK */
5238	SCTP_TCB_LOCK_DESTROY(stcb);
5239	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5240	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5241		SCTP_INP_INFO_WUNLOCK();
5242		SCTP_INP_RLOCK(inp);
5243	}
5244#ifdef SCTP_TRACK_FREED_ASOCS
5245	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5246		/* now clean up the tasoc itself */
5247		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5248		SCTP_DECR_ASOC_COUNT();
5249	} else {
5250		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5251	}
5252#else
5253	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5254	SCTP_DECR_ASOC_COUNT();
5255#endif
5256	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5257		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5258			/*
5259			 * If its NOT the inp_free calling us AND sctp_close
5260			 * as been called, we call back...
5261			 */
5262			SCTP_INP_RUNLOCK(inp);
5263			/*
5264			 * This will start the kill timer (if we are the
5265			 * last one) since we hold an increment yet. But
5266			 * this is the only safe way to do this since
5267			 * otherwise if the socket closes at the same time
5268			 * we are here we might collide in the cleanup.
5269			 */
5270			sctp_inpcb_free(inp,
5271			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5272			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5273			SCTP_INP_DECR_REF(inp);
5274			goto out_of;
5275		} else {
5276			/* The socket is still open. */
5277			SCTP_INP_DECR_REF(inp);
5278		}
5279	}
5280	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5281		SCTP_INP_RUNLOCK(inp);
5282	}
5283out_of:
5284	/* destroyed the asoc */
5285#ifdef SCTP_LOG_CLOSING
5286	sctp_log_closing(inp, NULL, 11);
5287#endif
5288	return (1);
5289}
5290
5291
5292
5293/*
5294 * determine if a destination is "reachable" based upon the addresses bound
5295 * to the current endpoint (e.g. only v4 or v6 currently bound)
5296 */
5297/*
5298 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5299 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5300 * types bound as its endpoint
5301 */
5302int
5303sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5304{
5305	struct sctp_inpcb *inp;
5306	int answer;
5307
5308	/*
5309	 * No locks here, the TCB, in all cases is already locked and an
5310	 * assoc is up. There is either a INP lock by the caller applied (in
5311	 * asconf case when deleting an address) or NOT in the HB case,
5312	 * however if HB then the INP increment is up and the INP will not
5313	 * be removed (on top of the fact that we have a TCB lock). So we
5314	 * only want to read the sctp_flags, which is either bound-all or
5315	 * not.. no protection needed since once an assoc is up you can't be
5316	 * changing your binding.
5317	 */
5318	inp = stcb->sctp_ep;
5319	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5320		/* if bound all, destination is not restricted */
5321		/*
5322		 * RRS: Question during lock work: Is this correct? If you
5323		 * are bound-all you still might need to obey the V4--V6
5324		 * flags??? IMO this bound-all stuff needs to be removed!
5325		 */
5326		return (1);
5327	}
5328	/* NOTE: all "scope" checks are done when local addresses are added */
5329	switch (destaddr->sa_family) {
5330	case AF_INET6:
5331		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5332		break;
5333	case AF_INET:
5334		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5335		break;
5336	default:
5337		/* invalid family, so it's unreachable */
5338		answer = 0;
5339		break;
5340	}
5341	return (answer);
5342}
5343
5344/*
5345 * update the inp_vflags on an endpoint
5346 */
5347static void
5348sctp_update_ep_vflag(struct sctp_inpcb *inp)
5349{
5350	struct sctp_laddr *laddr;
5351
5352	/* first clear the flag */
5353	inp->ip_inp.inp.inp_vflag = 0;
5354	/* set the flag based on addresses on the ep list */
5355	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5356		if (laddr->ifa == NULL) {
5357			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5358			    __FUNCTION__);
5359			continue;
5360		}
5361		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5362			continue;
5363		}
5364		switch (laddr->ifa->address.sa.sa_family) {
5365#ifdef INET6
5366		case AF_INET6:
5367			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5368			break;
5369#endif
5370#ifdef INET
5371		case AF_INET:
5372			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5373			break;
5374#endif
5375		default:
5376			break;
5377		}
5378	}
5379}
5380
5381/*
5382 * Add the address to the endpoint local address list There is nothing to be
5383 * done if we are bound to all addresses
5384 */
5385void
5386sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5387{
5388	struct sctp_laddr *laddr;
5389	int fnd, error = 0;
5390
5391	fnd = 0;
5392
5393	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5394		/* You are already bound to all. You have it already */
5395		return;
5396	}
5397#ifdef INET6
5398	if (ifa->address.sa.sa_family == AF_INET6) {
5399		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5400			/* Can't bind a non-useable addr. */
5401			return;
5402		}
5403	}
5404#endif
5405	/* first, is it already present? */
5406	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5407		if (laddr->ifa == ifa) {
5408			fnd = 1;
5409			break;
5410		}
5411	}
5412
5413	if (fnd == 0) {
5414		/* Not in the ep list */
5415		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5416		if (error != 0)
5417			return;
5418		inp->laddr_count++;
5419		/* update inp_vflag flags */
5420		switch (ifa->address.sa.sa_family) {
5421#ifdef INET6
5422		case AF_INET6:
5423			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5424			break;
5425#endif
5426#ifdef INET6
5427		case AF_INET:
5428			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5429			break;
5430#endif
5431		default:
5432			break;
5433		}
5434	}
5435	return;
5436}
5437
5438
5439/*
5440 * select a new (hopefully reachable) destination net (should only be used
5441 * when we deleted an ep addr that is the only usable source address to reach
5442 * the destination net)
5443 */
5444static void
5445sctp_select_primary_destination(struct sctp_tcb *stcb)
5446{
5447	struct sctp_nets *net;
5448
5449	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5450		/* for now, we'll just pick the first reachable one we find */
5451		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5452			continue;
5453		if (sctp_destination_is_reachable(stcb,
5454		    (struct sockaddr *)&net->ro._l_addr)) {
5455			/* found a reachable destination */
5456			stcb->asoc.primary_destination = net;
5457		}
5458	}
5459	/* I can't there from here! ...we're gonna die shortly... */
5460}
5461
5462
5463/*
5464 * Delete the address from the endpoint local address list There is nothing
5465 * to be done if we are bound to all addresses
5466 */
5467void
5468sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5469{
5470	struct sctp_laddr *laddr;
5471	int fnd;
5472
5473	fnd = 0;
5474	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5475		/* You are already bound to all. You have it already */
5476		return;
5477	}
5478	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5479		if (laddr->ifa == ifa) {
5480			fnd = 1;
5481			break;
5482		}
5483	}
5484	if (fnd && (inp->laddr_count < 2)) {
5485		/* can't delete unless there are at LEAST 2 addresses */
5486		return;
5487	}
5488	if (fnd) {
5489		/*
5490		 * clean up any use of this address go through our
5491		 * associations and clear any last_used_address that match
5492		 * this one for each assoc, see if a new primary_destination
5493		 * is needed
5494		 */
5495		struct sctp_tcb *stcb;
5496
5497		/* clean up "next_addr_touse" */
5498		if (inp->next_addr_touse == laddr)
5499			/* delete this address */
5500			inp->next_addr_touse = NULL;
5501
5502		/* clean up "last_used_address" */
5503		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5504			struct sctp_nets *net;
5505
5506			SCTP_TCB_LOCK(stcb);
5507			if (stcb->asoc.last_used_address == laddr)
5508				/* delete this address */
5509				stcb->asoc.last_used_address = NULL;
5510			/*
5511			 * Now spin through all the nets and purge any ref
5512			 * to laddr
5513			 */
5514			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5515				if (net->ro._s_addr &&
5516				    (net->ro._s_addr->ifa == laddr->ifa)) {
5517					/* Yep, purge src address selected */
5518					sctp_rtentry_t *rt;
5519
5520					/* delete this address if cached */
5521					rt = net->ro.ro_rt;
5522					if (rt != NULL) {
5523						RTFREE(rt);
5524						net->ro.ro_rt = NULL;
5525					}
5526					sctp_free_ifa(net->ro._s_addr);
5527					net->ro._s_addr = NULL;
5528					net->src_addr_selected = 0;
5529				}
5530			}
5531			SCTP_TCB_UNLOCK(stcb);
5532		}		/* for each tcb */
5533		/* remove it from the ep list */
5534		sctp_remove_laddr(laddr);
5535		inp->laddr_count--;
5536		/* update inp_vflag flags */
5537		sctp_update_ep_vflag(inp);
5538	}
5539	return;
5540}
5541
5542/*
5543 * Add the address to the TCB local address restricted list.
5544 * This is a "pending" address list (eg. addresses waiting for an
5545 * ASCONF-ACK response) and cannot be used as a valid source address.
5546 */
5547void
5548sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5549{
5550	struct sctp_inpcb *inp;
5551	struct sctp_laddr *laddr;
5552	struct sctpladdr *list;
5553
5554	/*
5555	 * Assumes TCB is locked.. and possibly the INP. May need to
5556	 * confirm/fix that if we need it and is not the case.
5557	 */
5558	list = &stcb->asoc.sctp_restricted_addrs;
5559
5560	inp = stcb->sctp_ep;
5561#ifdef INET6
5562	if (ifa->address.sa.sa_family == AF_INET6) {
5563		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5564			/* Can't bind a non-existent addr. */
5565			return;
5566		}
5567	}
5568#endif
5569	/* does the address already exist? */
5570	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5571		if (laddr->ifa == ifa) {
5572			return;
5573		}
5574	}
5575
5576	/* add to the list */
5577	(void)sctp_insert_laddr(list, ifa, 0);
5578	return;
5579}
5580
5581/*
5582 * insert an laddr entry with the given ifa for the desired list
5583 */
5584int
5585sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5586{
5587	struct sctp_laddr *laddr;
5588
5589	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5590	if (laddr == NULL) {
5591		/* out of memory? */
5592		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5593		return (EINVAL);
5594	}
5595	SCTP_INCR_LADDR_COUNT();
5596	bzero(laddr, sizeof(*laddr));
5597	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5598	laddr->ifa = ifa;
5599	laddr->action = act;
5600	atomic_add_int(&ifa->refcount, 1);
5601	/* insert it */
5602	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5603
5604	return (0);
5605}
5606
5607/*
5608 * Remove an laddr entry from the local address list (on an assoc)
5609 */
5610void
5611sctp_remove_laddr(struct sctp_laddr *laddr)
5612{
5613
5614	/* remove from the list */
5615	LIST_REMOVE(laddr, sctp_nxt_addr);
5616	sctp_free_ifa(laddr->ifa);
5617	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5618	SCTP_DECR_LADDR_COUNT();
5619}
5620
5621/*
5622 * Remove a local address from the TCB local address restricted list
5623 */
5624void
5625sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5626{
5627	struct sctp_inpcb *inp;
5628	struct sctp_laddr *laddr;
5629
5630	/*
5631	 * This is called by asconf work. It is assumed that a) The TCB is
5632	 * locked and b) The INP is locked. This is true in as much as I can
5633	 * trace through the entry asconf code where I did these locks.
5634	 * Again, the ASCONF code is a bit different in that it does lock
5635	 * the INP during its work often times. This must be since we don't
5636	 * want other proc's looking up things while what they are looking
5637	 * up is changing :-D
5638	 */
5639
5640	inp = stcb->sctp_ep;
5641	/* if subset bound and don't allow ASCONF's, can't delete last */
5642	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5643	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5644		if (stcb->sctp_ep->laddr_count < 2) {
5645			/* can't delete last address */
5646			return;
5647		}
5648	}
5649	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5650		/* remove the address if it exists */
5651		if (laddr->ifa == NULL)
5652			continue;
5653		if (laddr->ifa == ifa) {
5654			sctp_remove_laddr(laddr);
5655			return;
5656		}
5657	}
5658
5659	/* address not found! */
5660	return;
5661}
5662
5663/*
5664 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5665 */
5666/* sysctl */
5667static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5668static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5669
5670
5671
5672#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5673struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5674int *sctp_cpuarry = NULL;
5675void
5676sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5677{
5678	/* Queue a packet to a processor for the specified core */
5679	struct sctp_mcore_queue *qent;
5680	struct sctp_mcore_ctrl *wkq;
5681	int need_wake = 0;
5682
5683	if (sctp_mcore_workers == NULL) {
5684		/* Something went way bad during setup */
5685		sctp_input_with_port(m, off, 0);
5686		return;
5687	}
5688	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5689	    (sizeof(struct sctp_mcore_queue)),
5690	    SCTP_M_MCORE);
5691	if (qent == NULL) {
5692		/* This is trouble  */
5693		sctp_input_with_port(m, off, 0);
5694		return;
5695	}
5696	qent->vn = curvnet;
5697	qent->m = m;
5698	qent->off = off;
5699	qent->v6 = 0;
5700	wkq = &sctp_mcore_workers[cpu_to_use];
5701	SCTP_MCORE_QLOCK(wkq);
5702
5703	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5704	if (wkq->running == 0) {
5705		need_wake = 1;
5706	}
5707	SCTP_MCORE_QUNLOCK(wkq);
5708	if (need_wake) {
5709		wakeup(&wkq->running);
5710	}
5711}
5712
5713static void
5714sctp_mcore_thread(void *arg)
5715{
5716
5717	struct sctp_mcore_ctrl *wkq;
5718	struct sctp_mcore_queue *qent;
5719
5720	wkq = (struct sctp_mcore_ctrl *)arg;
5721	struct mbuf *m;
5722	int off, v6;
5723
5724	/* Wait for first tickle */
5725	SCTP_MCORE_LOCK(wkq);
5726	wkq->running = 0;
5727	msleep(&wkq->running,
5728	    &wkq->core_mtx,
5729	    0, "wait for pkt", 0);
5730	SCTP_MCORE_UNLOCK(wkq);
5731
5732	/* Bind to our cpu */
5733	thread_lock(curthread);
5734	sched_bind(curthread, wkq->cpuid);
5735	thread_unlock(curthread);
5736
5737	/* Now lets start working */
5738	SCTP_MCORE_LOCK(wkq);
5739	/* Now grab lock and go */
5740	while (1) {
5741		SCTP_MCORE_QLOCK(wkq);
5742skip_sleep:
5743		wkq->running = 1;
5744		qent = TAILQ_FIRST(&wkq->que);
5745		if (qent) {
5746			TAILQ_REMOVE(&wkq->que, qent, next);
5747			SCTP_MCORE_QUNLOCK(wkq);
5748			CURVNET_SET(qent->vn);
5749			m = qent->m;
5750			off = qent->off;
5751			v6 = qent->v6;
5752			SCTP_FREE(qent, SCTP_M_MCORE);
5753			if (v6 == 0) {
5754				sctp_input_with_port(m, off, 0);
5755			} else {
5756				printf("V6 not yet supported\n");
5757				sctp_m_freem(m);
5758			}
5759			CURVNET_RESTORE();
5760			SCTP_MCORE_QLOCK(wkq);
5761		}
5762		wkq->running = 0;
5763		if (!TAILQ_EMPTY(&wkq->que)) {
5764			goto skip_sleep;
5765		}
5766		SCTP_MCORE_QUNLOCK(wkq);
5767		msleep(&wkq->running,
5768		    &wkq->core_mtx,
5769		    0, "wait for pkt", 0);
5770	};
5771}
5772
5773static void
5774sctp_startup_mcore_threads(void)
5775{
5776	int i, cpu;
5777
5778	if (mp_ncpus == 1)
5779		return;
5780
5781	if (sctp_mcore_workers != NULL) {
5782		/*
5783		 * Already been here in some previous vnet?
5784		 */
5785		return;
5786	}
5787	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5788	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5789	    SCTP_M_MCORE);
5790	if (sctp_mcore_workers == NULL) {
5791		/* TSNH I hope */
5792		return;
5793	}
5794	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5795	    sizeof(struct sctp_mcore_ctrl)));
5796	/* Init the structures */
5797	for (i = 0; i <= mp_maxid; i++) {
5798		TAILQ_INIT(&sctp_mcore_workers[i].que);
5799		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5800		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5801		sctp_mcore_workers[i].cpuid = i;
5802	}
5803	if (sctp_cpuarry == NULL) {
5804		SCTP_MALLOC(sctp_cpuarry, int *,
5805		    (mp_ncpus * sizeof(int)),
5806		    SCTP_M_MCORE);
5807		i = 0;
5808		CPU_FOREACH(cpu) {
5809			sctp_cpuarry[i] = cpu;
5810			i++;
5811		}
5812	}
5813	/* Now start them all */
5814	CPU_FOREACH(cpu) {
5815		(void)kproc_create(sctp_mcore_thread,
5816		    (void *)&sctp_mcore_workers[cpu],
5817		    &sctp_mcore_workers[cpu].thread_proc,
5818		    RFPROC,
5819		    SCTP_KTHREAD_PAGES,
5820		    SCTP_MCORE_NAME);
5821
5822	}
5823}
5824
5825#endif
5826
5827void
5828sctp_pcb_init()
5829{
5830	/*
5831	 * SCTP initialization for the PCB structures should be called by
5832	 * the sctp_init() funciton.
5833	 */
5834	int i;
5835	struct timeval tv;
5836
5837	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5838		/* error I was called twice */
5839		return;
5840	}
5841	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5842
5843#if defined(SCTP_LOCAL_TRACE_BUF)
5844	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5845#endif
5846#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5847	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5848	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5849	    SCTP_M_MCORE);
5850#endif
5851	(void)SCTP_GETTIME_TIMEVAL(&tv);
5852#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5853	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5854	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5855	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5856#else
5857	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5858	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5859	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5860#endif
5861	/* init the empty list of (All) Endpoints */
5862	LIST_INIT(&SCTP_BASE_INFO(listhead));
5863
5864
5865	/* init the hash table of endpoints */
5866	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5867	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5868	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5869	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5870	    &SCTP_BASE_INFO(hashasocmark));
5871	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5872	    &SCTP_BASE_INFO(hashmark));
5873	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5874	    &SCTP_BASE_INFO(hashtcpmark));
5875	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5876
5877
5878	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5879	    &SCTP_BASE_INFO(hashvrfmark));
5880
5881	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5882	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5883	/* init the zones */
5884	/*
5885	 * FIX ME: Should check for NULL returns, but if it does fail we are
5886	 * doomed to panic anyways... add later maybe.
5887	 */
5888	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5889	    sizeof(struct sctp_inpcb), maxsockets);
5890
5891	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5892	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5893
5894	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5895	    sizeof(struct sctp_laddr),
5896	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5897
5898	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5899	    sizeof(struct sctp_nets),
5900	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5901
5902	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5903	    sizeof(struct sctp_tmit_chunk),
5904	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5905
5906	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5907	    sizeof(struct sctp_queued_to_read),
5908	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5909
5910	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5911	    sizeof(struct sctp_stream_queue_pending),
5912	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5913
5914	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5915	    sizeof(struct sctp_asconf),
5916	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5917
5918	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5919	    sizeof(struct sctp_asconf_ack),
5920	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5921
5922
5923	/* Master Lock INIT for info structure */
5924	SCTP_INP_INFO_LOCK_INIT();
5925	SCTP_STATLOG_INIT_LOCK();
5926
5927	SCTP_IPI_COUNT_INIT();
5928	SCTP_IPI_ADDR_INIT();
5929#ifdef SCTP_PACKET_LOGGING
5930	SCTP_IP_PKTLOG_INIT();
5931#endif
5932	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5933
5934	SCTP_WQ_ADDR_INIT();
5935	/* not sure if we need all the counts */
5936	SCTP_BASE_INFO(ipi_count_ep) = 0;
5937	/* assoc/tcb zone info */
5938	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5939	/* local addrlist zone info */
5940	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5941	/* remote addrlist zone info */
5942	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5943	/* chunk info */
5944	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5945
5946	/* socket queue zone info */
5947	SCTP_BASE_INFO(ipi_count_readq) = 0;
5948
5949	/* stream out queue cont */
5950	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5951
5952	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5953	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5954
5955	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5956
5957	/* Init the TIMEWAIT list */
5958	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5959		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5960	}
5961
5962	sctp_startup_iterator();
5963
5964#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5965	sctp_startup_mcore_threads();
5966#endif
5967
5968	/*
5969	 * INIT the default VRF which for BSD is the only one, other O/S's
5970	 * may have more. But initially they must start with one and then
5971	 * add the VRF's as addresses are added.
5972	 */
5973	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5974}
5975
5976/*
5977 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5978 */
5979void
5980sctp_pcb_finish(void)
5981{
5982	struct sctp_vrflist *vrf_bucket;
5983	struct sctp_vrf *vrf, *nvrf;
5984	struct sctp_ifn *ifn, *nifn;
5985	struct sctp_ifa *ifa, *nifa;
5986	struct sctpvtaghead *chain;
5987	struct sctp_tagblock *twait_block, *prev_twait_block;
5988	struct sctp_laddr *wi, *nwi;
5989	int i;
5990
5991	/*
5992	 * Free BSD the it thread never exits but we do clean up. The only
5993	 * way freebsd reaches here if we have VRF's but we still add the
5994	 * ifdef to make it compile on old versions.
5995	 */
5996	{
5997		struct sctp_iterator *it, *nit;
5998
5999		SCTP_IPI_ITERATOR_WQ_LOCK();
6000		TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
6001			if (it->vn != curvnet) {
6002				continue;
6003			}
6004			TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
6005			if (it->function_atend != NULL) {
6006				(*it->function_atend) (it->pointer, it->val);
6007			}
6008			SCTP_FREE(it, SCTP_M_ITER);
6009		}
6010		SCTP_IPI_ITERATOR_WQ_UNLOCK();
6011		SCTP_ITERATOR_LOCK();
6012		if ((sctp_it_ctl.cur_it) &&
6013		    (sctp_it_ctl.cur_it->vn == curvnet)) {
6014			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
6015		}
6016		SCTP_ITERATOR_UNLOCK();
6017	}
6018
6019	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
6020	SCTP_WQ_ADDR_LOCK();
6021	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
6022		LIST_REMOVE(wi, sctp_nxt_addr);
6023		SCTP_DECR_LADDR_COUNT();
6024		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
6025	}
6026	SCTP_WQ_ADDR_UNLOCK();
6027
6028	/*
6029	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
6030	 * destroyed first).
6031	 */
6032	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
6033	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
6034		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
6035			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
6036				/* free the ifa */
6037				LIST_REMOVE(ifa, next_bucket);
6038				LIST_REMOVE(ifa, next_ifa);
6039				SCTP_FREE(ifa, SCTP_M_IFA);
6040			}
6041			/* free the ifn */
6042			LIST_REMOVE(ifn, next_bucket);
6043			LIST_REMOVE(ifn, next_ifn);
6044			SCTP_FREE(ifn, SCTP_M_IFN);
6045		}
6046		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
6047		/* free the vrf */
6048		LIST_REMOVE(vrf, next_vrf);
6049		SCTP_FREE(vrf, SCTP_M_VRF);
6050	}
6051	/* free the vrf hashes */
6052	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
6053	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
6054
6055	/*
6056	 * free the TIMEWAIT list elements malloc'd in the function
6057	 * sctp_add_vtag_to_timewait()...
6058	 */
6059	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
6060		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
6061		if (!LIST_EMPTY(chain)) {
6062			prev_twait_block = NULL;
6063			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6064				if (prev_twait_block) {
6065					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6066				}
6067				prev_twait_block = twait_block;
6068			}
6069			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6070		}
6071	}
6072
6073	/* free the locks and mutexes */
6074#ifdef SCTP_PACKET_LOGGING
6075	SCTP_IP_PKTLOG_DESTROY();
6076#endif
6077	SCTP_IPI_ADDR_DESTROY();
6078	SCTP_STATLOG_DESTROY();
6079	SCTP_INP_INFO_LOCK_DESTROY();
6080
6081	SCTP_WQ_ADDR_DESTROY();
6082
6083	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6084	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6085	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6086	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6087	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6088	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6089	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6090	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6091	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6092	/* Get rid of other stuff to */
6093	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
6094		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
6095	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
6096		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6097	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6098		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6099#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6100	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6101#endif
6102}
6103
6104
6105int
6106sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6107    int iphlen, int offset, int limit, struct sctphdr *sh,
6108    struct sockaddr *altsa)
6109{
6110	/*
6111	 * grub through the INIT pulling addresses and loading them to the
6112	 * nets structure in the asoc. The from address in the mbuf should
6113	 * also be loaded (if it is not already). This routine can be called
6114	 * with either INIT or INIT-ACK's as long as the m points to the IP
6115	 * packet and the offset points to the beginning of the parameters.
6116	 */
6117	struct sctp_inpcb *inp, *l_inp;
6118	struct sctp_nets *net, *nnet, *net_tmp;
6119	struct ip *iph;
6120	struct sctp_paramhdr *phdr, parm_buf;
6121	struct sctp_tcb *stcb_tmp;
6122	uint16_t ptype, plen;
6123	struct sockaddr *sa;
6124	struct sockaddr_storage dest_store;
6125	struct sockaddr *local_sa = (struct sockaddr *)&dest_store;
6126	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6127	struct sctp_auth_random *p_random = NULL;
6128	uint16_t random_len = 0;
6129	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6130	struct sctp_auth_hmac_algo *hmacs = NULL;
6131	uint16_t hmacs_len = 0;
6132	uint8_t saw_asconf = 0;
6133	uint8_t saw_asconf_ack = 0;
6134	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6135	struct sctp_auth_chunk_list *chunks = NULL;
6136	uint16_t num_chunks = 0;
6137	sctp_key_t *new_key;
6138	uint32_t keylen;
6139	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6140	uint8_t ecn_allowed;
6141
6142#ifdef INET
6143	struct sockaddr_in sin;
6144
6145#endif
6146#ifdef INET6
6147	struct sockaddr_in6 sin6;
6148
6149#endif
6150
6151	/* First get the destination address setup too. */
6152#ifdef INET
6153	memset(&sin, 0, sizeof(sin));
6154	sin.sin_family = AF_INET;
6155	sin.sin_len = sizeof(sin);
6156	sin.sin_port = stcb->rport;
6157#endif
6158#ifdef INET6
6159	memset(&sin6, 0, sizeof(sin6));
6160	sin6.sin6_family = AF_INET6;
6161	sin6.sin6_len = sizeof(struct sockaddr_in6);
6162	sin6.sin6_port = stcb->rport;
6163#endif
6164	iph = mtod(m, struct ip *);
6165	switch (iph->ip_v) {
6166#ifdef INET
6167	case IPVERSION:
6168		{
6169			/* its IPv4 */
6170			struct sockaddr_in *sin_2;
6171
6172			sin_2 = (struct sockaddr_in *)(local_sa);
6173			memset(sin_2, 0, sizeof(sin));
6174			sin_2->sin_family = AF_INET;
6175			sin_2->sin_len = sizeof(sin);
6176			sin_2->sin_port = sh->dest_port;
6177			sin_2->sin_addr.s_addr = iph->ip_dst.s_addr;
6178			if (altsa) {
6179				/*
6180				 * For cookies we use the src address NOT
6181				 * from the packet but from the original
6182				 * INIT.
6183				 */
6184				sa = altsa;
6185			} else {
6186				sin.sin_addr = iph->ip_src;
6187				sa = (struct sockaddr *)&sin;
6188			}
6189			break;
6190		}
6191#endif
6192#ifdef INET6
6193	case IPV6_VERSION >> 4:
6194		{
6195			/* its IPv6 */
6196			struct ip6_hdr *ip6;
6197			struct sockaddr_in6 *sin6_2;
6198
6199			ip6 = mtod(m, struct ip6_hdr *);
6200			sin6_2 = (struct sockaddr_in6 *)(local_sa);
6201			memset(sin6_2, 0, sizeof(sin6));
6202			sin6_2->sin6_family = AF_INET6;
6203			sin6_2->sin6_len = sizeof(struct sockaddr_in6);
6204			sin6_2->sin6_port = sh->dest_port;
6205			sin6_2->sin6_addr = ip6->ip6_dst;
6206			if (altsa) {
6207				/*
6208				 * For cookies we use the src address NOT
6209				 * from the packet but from the original
6210				 * INIT.
6211				 */
6212				sa = altsa;
6213			} else {
6214				sin6.sin6_addr = ip6->ip6_src;
6215				sa = (struct sockaddr *)&sin6;
6216			}
6217			break;
6218		}
6219#endif
6220	default:
6221		return (-1);
6222		break;
6223	}
6224	/* Turn off ECN until we get through all params */
6225	ecn_allowed = 0;
6226	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6227		/* mark all addresses that we have currently on the list */
6228		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6229	}
6230	/* does the source address already exist? if so skip it */
6231	l_inp = inp = stcb->sctp_ep;
6232
6233	atomic_add_int(&stcb->asoc.refcnt, 1);
6234	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb);
6235	atomic_add_int(&stcb->asoc.refcnt, -1);
6236
6237	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6238		/* we must add the source address */
6239		/* no scope set here since we have a tcb already. */
6240		switch (sa->sa_family) {
6241#ifdef INET
6242		case AF_INET:
6243			if (stcb->asoc.ipv4_addr_legal) {
6244				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6245					return (-1);
6246				}
6247			}
6248			break;
6249#endif
6250#ifdef INET6
6251		case AF_INET6:
6252			if (stcb->asoc.ipv6_addr_legal) {
6253				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6254					return (-2);
6255				}
6256			}
6257			break;
6258#endif
6259		default:
6260			break;
6261		}
6262	} else {
6263		if (net_tmp != NULL && stcb_tmp == stcb) {
6264			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6265		} else if (stcb_tmp != stcb) {
6266			/* It belongs to another association? */
6267			if (stcb_tmp)
6268				SCTP_TCB_UNLOCK(stcb_tmp);
6269			return (-3);
6270		}
6271	}
6272	if (stcb->asoc.state == 0) {
6273		/* the assoc was freed? */
6274		return (-4);
6275	}
6276	/*
6277	 * peer must explicitly turn this on. This may have been initialized
6278	 * to be "on" in order to allow local addr changes while INIT's are
6279	 * in flight.
6280	 */
6281	stcb->asoc.peer_supports_asconf = 0;
6282	/* now we must go through each of the params. */
6283	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6284	while (phdr) {
6285		ptype = ntohs(phdr->param_type);
6286		plen = ntohs(phdr->param_length);
6287		/*
6288		 * printf("ptype => %0x, plen => %d\n", (uint32_t)ptype,
6289		 * (int)plen);
6290		 */
6291		if (offset + plen > limit) {
6292			break;
6293		}
6294		if (plen == 0) {
6295			break;
6296		}
6297#ifdef INET
6298		if (ptype == SCTP_IPV4_ADDRESS) {
6299			if (stcb->asoc.ipv4_addr_legal) {
6300				struct sctp_ipv4addr_param *p4, p4_buf;
6301
6302				/* ok get the v4 address and check/add */
6303				phdr = sctp_get_next_param(m, offset,
6304				    (struct sctp_paramhdr *)&p4_buf,
6305				    sizeof(p4_buf));
6306				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6307				    phdr == NULL) {
6308					return (-5);
6309				}
6310				p4 = (struct sctp_ipv4addr_param *)phdr;
6311				sin.sin_addr.s_addr = p4->addr;
6312				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6313					/* Skip multi-cast addresses */
6314					goto next_param;
6315				}
6316				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6317				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6318					goto next_param;
6319				}
6320				sa = (struct sockaddr *)&sin;
6321				inp = stcb->sctp_ep;
6322				atomic_add_int(&stcb->asoc.refcnt, 1);
6323				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6324				    local_sa, stcb);
6325				atomic_add_int(&stcb->asoc.refcnt, -1);
6326
6327				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6328				    inp == NULL) {
6329					/* we must add the source address */
6330					/*
6331					 * no scope set since we have a tcb
6332					 * already
6333					 */
6334
6335					/*
6336					 * we must validate the state again
6337					 * here
6338					 */
6339			add_it_now:
6340					if (stcb->asoc.state == 0) {
6341						/* the assoc was freed? */
6342						return (-7);
6343					}
6344					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6345						return (-8);
6346					}
6347				} else if (stcb_tmp == stcb) {
6348					if (stcb->asoc.state == 0) {
6349						/* the assoc was freed? */
6350						return (-10);
6351					}
6352					if (net != NULL) {
6353						/* clear flag */
6354						net->dest_state &=
6355						    ~SCTP_ADDR_NOT_IN_ASSOC;
6356					}
6357				} else {
6358					/*
6359					 * strange, address is in another
6360					 * assoc? straighten out locks.
6361					 */
6362					if (stcb_tmp) {
6363						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6364							/*
6365							 * in setup state we
6366							 * abort this guy
6367							 */
6368							sctp_abort_an_association(stcb_tmp->sctp_ep,
6369							    stcb_tmp, 1, NULL, 0);
6370							goto add_it_now;
6371						}
6372						SCTP_TCB_UNLOCK(stcb_tmp);
6373					}
6374					if (stcb->asoc.state == 0) {
6375						/* the assoc was freed? */
6376						return (-12);
6377					}
6378					return (-13);
6379				}
6380			}
6381		} else
6382#endif
6383#ifdef INET6
6384		if (ptype == SCTP_IPV6_ADDRESS) {
6385			if (stcb->asoc.ipv6_addr_legal) {
6386				/* ok get the v6 address and check/add */
6387				struct sctp_ipv6addr_param *p6, p6_buf;
6388
6389				phdr = sctp_get_next_param(m, offset,
6390				    (struct sctp_paramhdr *)&p6_buf,
6391				    sizeof(p6_buf));
6392				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6393				    phdr == NULL) {
6394					return (-14);
6395				}
6396				p6 = (struct sctp_ipv6addr_param *)phdr;
6397				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6398				    sizeof(p6->addr));
6399				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6400					/* Skip multi-cast addresses */
6401					goto next_param;
6402				}
6403				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6404					/*
6405					 * Link local make no sense without
6406					 * scope
6407					 */
6408					goto next_param;
6409				}
6410				sa = (struct sockaddr *)&sin6;
6411				inp = stcb->sctp_ep;
6412				atomic_add_int(&stcb->asoc.refcnt, 1);
6413				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6414				    local_sa, stcb);
6415				atomic_add_int(&stcb->asoc.refcnt, -1);
6416				if (stcb_tmp == NULL &&
6417				    (inp == stcb->sctp_ep || inp == NULL)) {
6418					/*
6419					 * we must validate the state again
6420					 * here
6421					 */
6422			add_it_now6:
6423					if (stcb->asoc.state == 0) {
6424						/* the assoc was freed? */
6425						return (-16);
6426					}
6427					/*
6428					 * we must add the address, no scope
6429					 * set
6430					 */
6431					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6432						return (-17);
6433					}
6434				} else if (stcb_tmp == stcb) {
6435					/*
6436					 * we must validate the state again
6437					 * here
6438					 */
6439					if (stcb->asoc.state == 0) {
6440						/* the assoc was freed? */
6441						return (-19);
6442					}
6443					if (net != NULL) {
6444						/* clear flag */
6445						net->dest_state &=
6446						    ~SCTP_ADDR_NOT_IN_ASSOC;
6447					}
6448				} else {
6449					/*
6450					 * strange, address is in another
6451					 * assoc? straighten out locks.
6452					 */
6453					if (stcb_tmp)
6454						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6455							/*
6456							 * in setup state we
6457							 * abort this guy
6458							 */
6459							sctp_abort_an_association(stcb_tmp->sctp_ep,
6460							    stcb_tmp, 1, NULL, 0);
6461							goto add_it_now6;
6462						}
6463					SCTP_TCB_UNLOCK(stcb_tmp);
6464
6465					if (stcb->asoc.state == 0) {
6466						/* the assoc was freed? */
6467						return (-21);
6468					}
6469					return (-22);
6470				}
6471			}
6472		} else
6473#endif
6474		if (ptype == SCTP_ECN_CAPABLE) {
6475			ecn_allowed = 1;
6476		} else if (ptype == SCTP_ULP_ADAPTATION) {
6477			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6478				struct sctp_adaptation_layer_indication ai,
6479				                                *aip;
6480
6481				phdr = sctp_get_next_param(m, offset,
6482				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6483				aip = (struct sctp_adaptation_layer_indication *)phdr;
6484				if (aip) {
6485					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6486					stcb->asoc.adaptation_needed = 1;
6487				}
6488			}
6489		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6490			struct sctp_asconf_addr_param lstore, *fee;
6491			int lptype;
6492			struct sockaddr *lsa = NULL;
6493
6494#ifdef INET
6495			struct sctp_asconf_addrv4_param *fii;
6496
6497#endif
6498
6499			stcb->asoc.peer_supports_asconf = 1;
6500			if (plen > sizeof(lstore)) {
6501				return (-23);
6502			}
6503			phdr = sctp_get_next_param(m, offset,
6504			    (struct sctp_paramhdr *)&lstore,
6505			    min(plen, sizeof(lstore)));
6506			if (phdr == NULL) {
6507				return (-24);
6508			}
6509			fee = (struct sctp_asconf_addr_param *)phdr;
6510			lptype = ntohs(fee->addrp.ph.param_type);
6511			switch (lptype) {
6512#ifdef INET
6513			case SCTP_IPV4_ADDRESS:
6514				if (plen !=
6515				    sizeof(struct sctp_asconf_addrv4_param)) {
6516					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6517					    (int)sizeof(struct sctp_asconf_addrv4_param),
6518					    plen);
6519				} else {
6520					fii = (struct sctp_asconf_addrv4_param *)fee;
6521					sin.sin_addr.s_addr = fii->addrp.addr;
6522					lsa = (struct sockaddr *)&sin;
6523				}
6524				break;
6525#endif
6526#ifdef INET6
6527			case SCTP_IPV6_ADDRESS:
6528				if (plen !=
6529				    sizeof(struct sctp_asconf_addr_param)) {
6530					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6531					    (int)sizeof(struct sctp_asconf_addr_param),
6532					    plen);
6533				} else {
6534					memcpy(sin6.sin6_addr.s6_addr,
6535					    fee->addrp.addr,
6536					    sizeof(fee->addrp.addr));
6537					lsa = (struct sockaddr *)&sin6;
6538				}
6539				break;
6540#endif
6541			default:
6542				break;
6543			}
6544			if (lsa) {
6545				(void)sctp_set_primary_addr(stcb, sa, NULL);
6546			}
6547		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6548			stcb->asoc.peer_supports_nat = 1;
6549		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6550			/* Peer supports pr-sctp */
6551			stcb->asoc.peer_supports_prsctp = 1;
6552		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6553			/* A supported extension chunk */
6554			struct sctp_supported_chunk_types_param *pr_supported;
6555			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6556			int num_ent, i;
6557
6558			phdr = sctp_get_next_param(m, offset,
6559			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6560			if (phdr == NULL) {
6561				return (-25);
6562			}
6563			stcb->asoc.peer_supports_asconf = 0;
6564			stcb->asoc.peer_supports_prsctp = 0;
6565			stcb->asoc.peer_supports_pktdrop = 0;
6566			stcb->asoc.peer_supports_strreset = 0;
6567			stcb->asoc.peer_supports_nr_sack = 0;
6568			stcb->asoc.peer_supports_auth = 0;
6569			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6570			num_ent = plen - sizeof(struct sctp_paramhdr);
6571			for (i = 0; i < num_ent; i++) {
6572				switch (pr_supported->chunk_types[i]) {
6573				case SCTP_ASCONF:
6574				case SCTP_ASCONF_ACK:
6575					stcb->asoc.peer_supports_asconf = 1;
6576					break;
6577				case SCTP_FORWARD_CUM_TSN:
6578					stcb->asoc.peer_supports_prsctp = 1;
6579					break;
6580				case SCTP_PACKET_DROPPED:
6581					stcb->asoc.peer_supports_pktdrop = 1;
6582					break;
6583				case SCTP_NR_SELECTIVE_ACK:
6584					stcb->asoc.peer_supports_nr_sack = 1;
6585					break;
6586				case SCTP_STREAM_RESET:
6587					stcb->asoc.peer_supports_strreset = 1;
6588					break;
6589				case SCTP_AUTHENTICATION:
6590					stcb->asoc.peer_supports_auth = 1;
6591					break;
6592				default:
6593					/* one I have not learned yet */
6594					break;
6595
6596				}
6597			}
6598		} else if (ptype == SCTP_RANDOM) {
6599			if (plen > sizeof(random_store))
6600				break;
6601			if (got_random) {
6602				/* already processed a RANDOM */
6603				goto next_param;
6604			}
6605			phdr = sctp_get_next_param(m, offset,
6606			    (struct sctp_paramhdr *)random_store,
6607			    min(sizeof(random_store), plen));
6608			if (phdr == NULL)
6609				return (-26);
6610			p_random = (struct sctp_auth_random *)phdr;
6611			random_len = plen - sizeof(*p_random);
6612			/* enforce the random length */
6613			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6614				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6615				return (-27);
6616			}
6617			got_random = 1;
6618		} else if (ptype == SCTP_HMAC_LIST) {
6619			int num_hmacs;
6620			int i;
6621
6622			if (plen > sizeof(hmacs_store))
6623				break;
6624			if (got_hmacs) {
6625				/* already processed a HMAC list */
6626				goto next_param;
6627			}
6628			phdr = sctp_get_next_param(m, offset,
6629			    (struct sctp_paramhdr *)hmacs_store,
6630			    min(plen, sizeof(hmacs_store)));
6631			if (phdr == NULL)
6632				return (-28);
6633			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6634			hmacs_len = plen - sizeof(*hmacs);
6635			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6636			/* validate the hmac list */
6637			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6638				return (-29);
6639			}
6640			if (stcb->asoc.peer_hmacs != NULL)
6641				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6642			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6643			if (stcb->asoc.peer_hmacs != NULL) {
6644				for (i = 0; i < num_hmacs; i++) {
6645					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6646					    ntohs(hmacs->hmac_ids[i]));
6647				}
6648			}
6649			got_hmacs = 1;
6650		} else if (ptype == SCTP_CHUNK_LIST) {
6651			int i;
6652
6653			if (plen > sizeof(chunks_store))
6654				break;
6655			if (got_chklist) {
6656				/* already processed a Chunks list */
6657				goto next_param;
6658			}
6659			phdr = sctp_get_next_param(m, offset,
6660			    (struct sctp_paramhdr *)chunks_store,
6661			    min(plen, sizeof(chunks_store)));
6662			if (phdr == NULL)
6663				return (-30);
6664			chunks = (struct sctp_auth_chunk_list *)phdr;
6665			num_chunks = plen - sizeof(*chunks);
6666			if (stcb->asoc.peer_auth_chunks != NULL)
6667				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6668			else
6669				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6670			for (i = 0; i < num_chunks; i++) {
6671				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6672				    stcb->asoc.peer_auth_chunks);
6673				/* record asconf/asconf-ack if listed */
6674				if (chunks->chunk_types[i] == SCTP_ASCONF)
6675					saw_asconf = 1;
6676				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6677					saw_asconf_ack = 1;
6678
6679			}
6680			got_chklist = 1;
6681		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6682			    (ptype == SCTP_STATE_COOKIE) ||
6683			    (ptype == SCTP_UNRECOG_PARAM) ||
6684			    (ptype == SCTP_COOKIE_PRESERVE) ||
6685			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6686			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6687			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6688			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6689		    (ptype == SCTP_SUCCESS_REPORT)) {
6690			 /* don't care */ ;
6691		} else {
6692			if ((ptype & 0x8000) == 0x0000) {
6693				/*
6694				 * must stop processing the rest of the
6695				 * param's. Any report bits were handled
6696				 * with the call to
6697				 * sctp_arethere_unrecognized_parameters()
6698				 * when the INIT or INIT-ACK was first seen.
6699				 */
6700				break;
6701			}
6702		}
6703
6704next_param:
6705		offset += SCTP_SIZE32(plen);
6706		if (offset >= limit) {
6707			break;
6708		}
6709		phdr = sctp_get_next_param(m, offset, &parm_buf,
6710		    sizeof(parm_buf));
6711	}
6712	/* Now check to see if we need to purge any addresses */
6713	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6714		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6715		    SCTP_ADDR_NOT_IN_ASSOC) {
6716			/* This address has been removed from the asoc */
6717			/* remove and free it */
6718			stcb->asoc.numnets--;
6719			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6720			sctp_free_remote_addr(net);
6721			if (net == stcb->asoc.primary_destination) {
6722				stcb->asoc.primary_destination = NULL;
6723				sctp_select_primary_destination(stcb);
6724			}
6725		}
6726	}
6727	if (ecn_allowed == 0) {
6728		stcb->asoc.ecn_allowed = 0;
6729	}
6730	/* validate authentication required parameters */
6731	if (got_random && got_hmacs) {
6732		stcb->asoc.peer_supports_auth = 1;
6733	} else {
6734		stcb->asoc.peer_supports_auth = 0;
6735	}
6736	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6737		/* peer does not support auth but sent a chunks list? */
6738		return (-31);
6739	}
6740	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6741	    !stcb->asoc.peer_supports_auth) {
6742		/* peer supports asconf but not auth? */
6743		return (-32);
6744	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6745	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6746		return (-33);
6747	}
6748	/* concatenate the full random key */
6749	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6750	if (chunks != NULL) {
6751		keylen += sizeof(*chunks) + num_chunks;
6752	}
6753	new_key = sctp_alloc_key(keylen);
6754	if (new_key != NULL) {
6755		/* copy in the RANDOM */
6756		if (p_random != NULL) {
6757			keylen = sizeof(*p_random) + random_len;
6758			bcopy(p_random, new_key->key, keylen);
6759		}
6760		/* append in the AUTH chunks */
6761		if (chunks != NULL) {
6762			bcopy(chunks, new_key->key + keylen,
6763			    sizeof(*chunks) + num_chunks);
6764			keylen += sizeof(*chunks) + num_chunks;
6765		}
6766		/* append in the HMACs */
6767		if (hmacs != NULL) {
6768			bcopy(hmacs, new_key->key + keylen,
6769			    sizeof(*hmacs) + hmacs_len);
6770		}
6771	} else {
6772		/* failed to get memory for the key */
6773		return (-34);
6774	}
6775	if (stcb->asoc.authinfo.peer_random != NULL)
6776		sctp_free_key(stcb->asoc.authinfo.peer_random);
6777	stcb->asoc.authinfo.peer_random = new_key;
6778	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6779	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6780
6781	return (0);
6782}
6783
6784int
6785sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6786    struct sctp_nets *net)
6787{
6788	/* make sure the requested primary address exists in the assoc */
6789	if (net == NULL && sa)
6790		net = sctp_findnet(stcb, sa);
6791
6792	if (net == NULL) {
6793		/* didn't find the requested primary address! */
6794		return (-1);
6795	} else {
6796		/* set the primary address */
6797		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6798			/* Must be confirmed, so queue to set */
6799			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6800			return (0);
6801		}
6802		stcb->asoc.primary_destination = net;
6803		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6804			sctp_free_remote_addr(stcb->asoc.alternate);
6805			stcb->asoc.alternate = NULL;
6806		}
6807		net = TAILQ_FIRST(&stcb->asoc.nets);
6808		if (net != stcb->asoc.primary_destination) {
6809			/*
6810			 * first one on the list is NOT the primary
6811			 * sctp_cmpaddr() is much more efficient if the
6812			 * primary is the first on the list, make it so.
6813			 */
6814			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6815			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6816		}
6817		return (0);
6818	}
6819}
6820
6821int
6822sctp_is_vtag_good(struct sctp_inpcb *inp, uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now, int save_in_twait)
6823{
6824	/*
6825	 * This function serves two purposes. It will see if a TAG can be
6826	 * re-used and return 1 for yes it is ok and 0 for don't use that
6827	 * tag. A secondary function it will do is purge out old tags that
6828	 * can be removed.
6829	 */
6830	struct sctpvtaghead *chain;
6831	struct sctp_tagblock *twait_block;
6832	struct sctpasochead *head;
6833	struct sctp_tcb *stcb;
6834	int i;
6835
6836	SCTP_INP_INFO_RLOCK();
6837	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6838	    SCTP_BASE_INFO(hashasocmark))];
6839	if (head == NULL) {
6840		/* invalid vtag */
6841		goto skip_vtag_check;
6842	}
6843	LIST_FOREACH(stcb, head, sctp_asocs) {
6844		/*
6845		 * We choose not to lock anything here. TCB's can't be
6846		 * removed since we have the read lock, so they can't be
6847		 * freed on us, same thing for the INP. I may be wrong with
6848		 * this assumption, but we will go with it for now :-)
6849		 */
6850		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6851			continue;
6852		}
6853		if (stcb->asoc.my_vtag == tag) {
6854			/* candidate */
6855			if (stcb->rport != rport) {
6856				continue;
6857			}
6858			if (stcb->sctp_ep->sctp_lport != lport) {
6859				continue;
6860			}
6861			/* Its a used tag set */
6862			SCTP_INP_INFO_RUNLOCK();
6863			return (0);
6864		}
6865	}
6866skip_vtag_check:
6867
6868	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6869	/* Now what about timed wait ? */
6870	if (!LIST_EMPTY(chain)) {
6871		/*
6872		 * Block(s) are present, lets see if we have this tag in the
6873		 * list
6874		 */
6875		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6876			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6877				if (twait_block->vtag_block[i].v_tag == 0) {
6878					/* not used */
6879					continue;
6880				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6881				    now->tv_sec) {
6882					/* Audit expires this guy */
6883					twait_block->vtag_block[i].tv_sec_at_expire = 0;
6884					twait_block->vtag_block[i].v_tag = 0;
6885					twait_block->vtag_block[i].lport = 0;
6886					twait_block->vtag_block[i].rport = 0;
6887				} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6888					    (twait_block->vtag_block[i].lport == lport) &&
6889				    (twait_block->vtag_block[i].rport == rport)) {
6890					/* Bad tag, sorry :< */
6891					SCTP_INP_INFO_RUNLOCK();
6892					return (0);
6893				}
6894			}
6895		}
6896	}
6897	SCTP_INP_INFO_RUNLOCK();
6898	return (1);
6899}
6900
6901
6902static sctp_assoc_t reneged_asoc_ids[256];
6903static uint8_t reneged_at = 0;
6904
6905
6906static void
6907sctp_drain_mbufs(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
6908{
6909	/*
6910	 * We must hunt this association for MBUF's past the cumack (i.e.
6911	 * out of order data that we can renege on).
6912	 */
6913	struct sctp_association *asoc;
6914	struct sctp_tmit_chunk *chk, *nchk;
6915	uint32_t cumulative_tsn_p1;
6916	struct sctp_queued_to_read *ctl, *nctl;
6917	int cnt, strmat;
6918	uint32_t gap, i;
6919	int fnd = 0;
6920
6921	/* We look for anything larger than the cum-ack + 1 */
6922
6923	asoc = &stcb->asoc;
6924	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6925		/* none we can reneg on. */
6926		return;
6927	}
6928	SCTP_STAT_INCR(sctps_protocol_drains_done);
6929	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6930	cnt = 0;
6931	/* First look in the re-assembly queue */
6932	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6933		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6934			/* Yep it is above cum-ack */
6935			cnt++;
6936			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6937			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6938			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6939			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6940			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6941			if (chk->data) {
6942				sctp_m_freem(chk->data);
6943				chk->data = NULL;
6944			}
6945			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6946		}
6947	}
6948	/* Ok that was fun, now we will drain all the inbound streams? */
6949	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6950		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6951			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6952				/* Yep it is above cum-ack */
6953				cnt++;
6954				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6955				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6956				sctp_ucount_decr(asoc->cnt_on_all_streams);
6957				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6958				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6959				if (ctl->data) {
6960					sctp_m_freem(ctl->data);
6961					ctl->data = NULL;
6962				}
6963				sctp_free_remote_addr(ctl->whoFrom);
6964				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6965				SCTP_DECR_READQ_COUNT();
6966			}
6967		}
6968	}
6969	if (cnt) {
6970		/* We must back down to see what the new highest is */
6971		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6972			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6973			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6974				asoc->highest_tsn_inside_map = i;
6975				fnd = 1;
6976				break;
6977			}
6978		}
6979		if (!fnd) {
6980			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6981		}
6982		/*
6983		 * Question, should we go through the delivery queue? The
6984		 * only reason things are on here is the app not reading OR
6985		 * a p-d-api up. An attacker COULD send enough in to
6986		 * initiate the PD-API and then send a bunch of stuff to
6987		 * other streams... these would wind up on the delivery
6988		 * queue.. and then we would not get to them. But in order
6989		 * to do this I then have to back-track and un-deliver
6990		 * sequence numbers in streams.. el-yucko. I think for now
6991		 * we will NOT look at the delivery queue and leave it to be
6992		 * something to consider later. An alternative would be to
6993		 * abort the P-D-API with a notification and then deliver
6994		 * the data.... Or another method might be to keep track of
6995		 * how many times the situation occurs and if we see a
6996		 * possible attack underway just abort the association.
6997		 */
6998#ifdef SCTP_DEBUG
6999		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
7000#endif
7001		/*
7002		 * Now do we need to find a new
7003		 * asoc->highest_tsn_inside_map?
7004		 */
7005		asoc->last_revoke_count = cnt;
7006		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
7007		/* sa_ignore NO_NULL_CHK */
7008		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
7009		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
7010		reneged_asoc_ids[reneged_at] = sctp_get_associd(stcb);
7011		reneged_at++;
7012	}
7013	/*
7014	 * Another issue, in un-setting the TSN's in the mapping array we
7015	 * DID NOT adjust the highest_tsn marker.  This will cause one of
7016	 * two things to occur. It may cause us to do extra work in checking
7017	 * for our mapping array movement. More importantly it may cause us
7018	 * to SACK every datagram. This may not be a bad thing though since
7019	 * we will recover once we get our cum-ack above and all this stuff
7020	 * we dumped recovered.
7021	 */
7022}
7023
7024void
7025sctp_drain()
7026{
7027	/*
7028	 * We must walk the PCB lists for ALL associations here. The system
7029	 * is LOW on MBUF's and needs help. This is where reneging will
7030	 * occur. We really hope this does NOT happen!
7031	 */
7032	VNET_ITERATOR_DECL(vnet_iter);
7033	VNET_LIST_RLOCK_NOSLEEP();
7034	VNET_FOREACH(vnet_iter) {
7035		CURVNET_SET(vnet_iter);
7036		struct sctp_inpcb *inp;
7037		struct sctp_tcb *stcb;
7038
7039		SCTP_STAT_INCR(sctps_protocol_drain_calls);
7040		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
7041#ifdef VIMAGE
7042			continue;
7043#else
7044			return;
7045#endif
7046		}
7047		SCTP_INP_INFO_RLOCK();
7048		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
7049			/* For each endpoint */
7050			SCTP_INP_RLOCK(inp);
7051			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
7052				/* For each association */
7053				SCTP_TCB_LOCK(stcb);
7054				sctp_drain_mbufs(inp, stcb);
7055				SCTP_TCB_UNLOCK(stcb);
7056			}
7057			SCTP_INP_RUNLOCK(inp);
7058		}
7059		SCTP_INP_INFO_RUNLOCK();
7060		CURVNET_RESTORE();
7061	}
7062	VNET_LIST_RUNLOCK_NOSLEEP();
7063}
7064
7065/*
7066 * start a new iterator
7067 * iterates through all endpoints and associations based on the pcb_state
7068 * flags and asoc_state.  "af" (mandatory) is executed for all matching
7069 * assocs and "ef" (optional) is executed when the iterator completes.
7070 * "inpf" (optional) is executed for each new endpoint as it is being
7071 * iterated through. inpe (optional) is called when the inp completes
7072 * its way through all the stcbs.
7073 */
7074int
7075sctp_initiate_iterator(inp_func inpf,
7076    asoc_func af,
7077    inp_func inpe,
7078    uint32_t pcb_state,
7079    uint32_t pcb_features,
7080    uint32_t asoc_state,
7081    void *argp,
7082    uint32_t argi,
7083    end_func ef,
7084    struct sctp_inpcb *s_inp,
7085    uint8_t chunk_output_off)
7086{
7087	struct sctp_iterator *it = NULL;
7088
7089	if (af == NULL) {
7090		return (-1);
7091	}
7092	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
7093	    SCTP_M_ITER);
7094	if (it == NULL) {
7095		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
7096		return (ENOMEM);
7097	}
7098	memset(it, 0, sizeof(*it));
7099	it->function_assoc = af;
7100	it->function_inp = inpf;
7101	if (inpf)
7102		it->done_current_ep = 0;
7103	else
7104		it->done_current_ep = 1;
7105	it->function_atend = ef;
7106	it->pointer = argp;
7107	it->val = argi;
7108	it->pcb_flags = pcb_state;
7109	it->pcb_features = pcb_features;
7110	it->asoc_state = asoc_state;
7111	it->function_inp_end = inpe;
7112	it->no_chunk_output = chunk_output_off;
7113	it->vn = curvnet;
7114	if (s_inp) {
7115		/* Assume lock is held here */
7116		it->inp = s_inp;
7117		SCTP_INP_INCR_REF(it->inp);
7118		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7119	} else {
7120		SCTP_INP_INFO_RLOCK();
7121		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7122		if (it->inp) {
7123			SCTP_INP_INCR_REF(it->inp);
7124		}
7125		SCTP_INP_INFO_RUNLOCK();
7126		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7127
7128	}
7129	SCTP_IPI_ITERATOR_WQ_LOCK();
7130
7131	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7132	if (sctp_it_ctl.iterator_running == 0) {
7133		sctp_wakeup_iterator();
7134	}
7135	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7136	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7137	return (0);
7138}
7139