sctp_pcb.c revision 283710
1/*-
2 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * a) Redistributions of source code must retain the above copyright notice,
10 *    this list of conditions and the following disclaimer.
11 *
12 * b) Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *    the documentation and/or other materials provided with the distribution.
15 *
16 * c) Neither the name of Cisco Systems, Inc. nor the names of its
17 *    contributors may be used to endorse or promote products derived
18 *    from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33#include <sys/cdefs.h>
34__FBSDID("$FreeBSD: stable/10/sys/netinet/sctp_pcb.c 283710 2015-05-29 12:08:49Z tuexen $");
35
36#include <netinet/sctp_os.h>
37#include <sys/proc.h>
38#include <netinet/sctp_var.h>
39#include <netinet/sctp_sysctl.h>
40#include <netinet/sctp_pcb.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp.h>
43#include <netinet/sctp_header.h>
44#include <netinet/sctp_asconf.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_timer.h>
47#include <netinet/sctp_bsd_addr.h>
48#include <netinet/sctp_dtrace_define.h>
49#if defined(INET) || defined(INET6)
50#include <netinet/udp.h>
51#endif
52#ifdef INET6
53#include <netinet6/ip6_var.h>
54#endif
55#include <sys/sched.h>
56#include <sys/smp.h>
57#include <sys/unistd.h>
58
59
60VNET_DEFINE(struct sctp_base_info, system_base_info);
61
62/* FIX: we don't handle multiple link local scopes */
63/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
64#ifdef INET6
65int
66SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
67{
68	struct sockaddr_in6 tmp_a, tmp_b;
69
70	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
71	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
72		return (0);
73	}
74	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
75	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
76		return (0);
77	}
78	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
79}
80
81#endif
82
83void
84sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
85{
86	/*
87	 * We really don't need to lock this, but I will just because it
88	 * does not hurt.
89	 */
90	SCTP_INP_INFO_RLOCK();
91	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
92	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
93	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
94	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
95	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
96	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
97	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
98	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
99	SCTP_INP_INFO_RUNLOCK();
100}
101
102/*-
103 * Addresses are added to VRF's (Virtual Router's). For BSD we
104 * have only the default VRF 0. We maintain a hash list of
105 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
106 * these has a list of addresses. When we add a new address
107 * to a VRF we lookup the ifn/ifn_index, if the ifn does
108 * not exist we create it and add it to the list of IFN's
109 * within the VRF. Once we have the sctp_ifn, we add the
110 * address to the list. So we look something like:
111 *
112 * hash-vrf-table
113 *   vrf-> ifn-> ifn -> ifn
114 *   vrf    |
115 *    ...   +--ifa-> ifa -> ifa
116 *   vrf
117 *
118 * We keep these separate lists since the SCTP subsystem will
119 * point to these from its source address selection nets structure.
120 * When an address is deleted it does not happen right away on
121 * the SCTP side, it gets scheduled. What we do when a
122 * delete happens is immediately remove the address from
123 * the master list and decrement the refcount. As our
124 * addip iterator works through and frees the src address
125 * selection pointing to the sctp_ifa, eventually the refcount
126 * will reach 0 and we will delete it. Note that it is assumed
127 * that any locking on system level ifn/ifa is done at the
128 * caller of these functions and these routines will only
129 * lock the SCTP structures as they add or delete things.
130 *
131 * Other notes on VRF concepts.
132 *  - An endpoint can be in multiple VRF's
133 *  - An association lives within a VRF and only one VRF.
134 *  - Any incoming packet we can deduce the VRF for by
135 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
136 *  - Any downward send call or connect call must supply the
137 *    VRF via ancillary data or via some sort of set default
138 *    VRF socket option call (again for BSD no brainer since
139 *    the VRF is always 0).
140 *  - An endpoint may add multiple VRF's to it.
141 *  - Listening sockets can accept associations in any
142 *    of the VRF's they are in but the assoc will end up
143 *    in only one VRF (gotten from the packet or connect/send).
144 *
145 */
146
147struct sctp_vrf *
148sctp_allocate_vrf(int vrf_id)
149{
150	struct sctp_vrf *vrf = NULL;
151	struct sctp_vrflist *bucket;
152
153	/* First allocate the VRF structure */
154	vrf = sctp_find_vrf(vrf_id);
155	if (vrf) {
156		/* Already allocated */
157		return (vrf);
158	}
159	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
160	    SCTP_M_VRF);
161	if (vrf == NULL) {
162		/* No memory */
163#ifdef INVARIANTS
164		panic("No memory for VRF:%d", vrf_id);
165#endif
166		return (NULL);
167	}
168	/* setup the VRF */
169	memset(vrf, 0, sizeof(struct sctp_vrf));
170	vrf->vrf_id = vrf_id;
171	LIST_INIT(&vrf->ifnlist);
172	vrf->total_ifa_count = 0;
173	vrf->refcount = 0;
174	/* now also setup table ids */
175	SCTP_INIT_VRF_TABLEID(vrf);
176	/* Init the HASH of addresses */
177	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
178	    &vrf->vrf_addr_hashmark);
179	if (vrf->vrf_addr_hash == NULL) {
180		/* No memory */
181#ifdef INVARIANTS
182		panic("No memory for VRF:%d", vrf_id);
183#endif
184		SCTP_FREE(vrf, SCTP_M_VRF);
185		return (NULL);
186	}
187	/* Add it to the hash table */
188	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
189	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
190	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
191	return (vrf);
192}
193
194
195struct sctp_ifn *
196sctp_find_ifn(void *ifn, uint32_t ifn_index)
197{
198	struct sctp_ifn *sctp_ifnp;
199	struct sctp_ifnlist *hash_ifn_head;
200
201	/*
202	 * We assume the lock is held for the addresses if that's wrong
203	 * problems could occur :-)
204	 */
205	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
206	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
207		if (sctp_ifnp->ifn_index == ifn_index) {
208			return (sctp_ifnp);
209		}
210		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
211			return (sctp_ifnp);
212		}
213	}
214	return (NULL);
215}
216
217
218struct sctp_vrf *
219sctp_find_vrf(uint32_t vrf_id)
220{
221	struct sctp_vrflist *bucket;
222	struct sctp_vrf *liste;
223
224	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
225	LIST_FOREACH(liste, bucket, next_vrf) {
226		if (vrf_id == liste->vrf_id) {
227			return (liste);
228		}
229	}
230	return (NULL);
231}
232
233
234void
235sctp_free_vrf(struct sctp_vrf *vrf)
236{
237	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
238		if (vrf->vrf_addr_hash) {
239			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
240			vrf->vrf_addr_hash = NULL;
241		}
242		/* We zero'd the count */
243		LIST_REMOVE(vrf, next_vrf);
244		SCTP_FREE(vrf, SCTP_M_VRF);
245		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
246	}
247}
248
249
250void
251sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
252{
253	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
254		/* We zero'd the count */
255		if (sctp_ifnp->vrf) {
256			sctp_free_vrf(sctp_ifnp->vrf);
257		}
258		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
259		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
260	}
261}
262
263
264void
265sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
266{
267	struct sctp_ifn *sctp_ifnp;
268
269	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
270	if (sctp_ifnp != NULL) {
271		sctp_ifnp->ifn_mtu = mtu;
272	}
273}
274
275
276void
277sctp_free_ifa(struct sctp_ifa *sctp_ifap)
278{
279	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
280		/* We zero'd the count */
281		if (sctp_ifap->ifn_p) {
282			sctp_free_ifn(sctp_ifap->ifn_p);
283		}
284		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
285		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
286	}
287}
288
289
290static void
291sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
292{
293	struct sctp_ifn *found;
294
295	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
296	if (found == NULL) {
297		/* Not in the list.. sorry */
298		return;
299	}
300	if (hold_addr_lock == 0)
301		SCTP_IPI_ADDR_WLOCK();
302	LIST_REMOVE(sctp_ifnp, next_bucket);
303	LIST_REMOVE(sctp_ifnp, next_ifn);
304	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
305	    sctp_ifnp->registered_af);
306	if (hold_addr_lock == 0)
307		SCTP_IPI_ADDR_WUNLOCK();
308	/* Take away the reference, and possibly free it */
309	sctp_free_ifn(sctp_ifnp);
310}
311
312
313void
314sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
315    const char *if_name, uint32_t ifn_index)
316{
317	struct sctp_vrf *vrf;
318	struct sctp_ifa *sctp_ifap;
319
320	SCTP_IPI_ADDR_RLOCK();
321	vrf = sctp_find_vrf(vrf_id);
322	if (vrf == NULL) {
323		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
324		goto out;
325
326	}
327	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
328	if (sctp_ifap == NULL) {
329		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
330		goto out;
331	}
332	if (sctp_ifap->ifn_p == NULL) {
333		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
334		goto out;
335	}
336	if (if_name) {
337		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
338			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
339			    sctp_ifap->ifn_p->ifn_name, if_name);
340			goto out;
341		}
342	} else {
343		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
344			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
345			    sctp_ifap->ifn_p->ifn_index, ifn_index);
346			goto out;
347		}
348	}
349
350	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
351	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
352out:
353	SCTP_IPI_ADDR_RUNLOCK();
354}
355
356
357void
358sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
359    const char *if_name, uint32_t ifn_index)
360{
361	struct sctp_vrf *vrf;
362	struct sctp_ifa *sctp_ifap;
363
364	SCTP_IPI_ADDR_RLOCK();
365	vrf = sctp_find_vrf(vrf_id);
366	if (vrf == NULL) {
367		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
368		goto out;
369
370	}
371	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
372	if (sctp_ifap == NULL) {
373		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
374		goto out;
375	}
376	if (sctp_ifap->ifn_p == NULL) {
377		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
378		goto out;
379	}
380	if (if_name) {
381		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
382			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
383			    sctp_ifap->ifn_p->ifn_name, if_name);
384			goto out;
385		}
386	} else {
387		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
388			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
389			    sctp_ifap->ifn_p->ifn_index, ifn_index);
390			goto out;
391		}
392	}
393
394	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
395	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
396out:
397	SCTP_IPI_ADDR_RUNLOCK();
398}
399
400
401/*-
402 * Add an ifa to an ifn.
403 * Register the interface as necessary.
404 * NOTE: ADDR write lock MUST be held.
405 */
406static void
407sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
408{
409	int ifa_af;
410
411	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
412	sctp_ifap->ifn_p = sctp_ifnp;
413	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
414	/* update address counts */
415	sctp_ifnp->ifa_count++;
416	ifa_af = sctp_ifap->address.sa.sa_family;
417	switch (ifa_af) {
418#ifdef INET
419	case AF_INET:
420		sctp_ifnp->num_v4++;
421		break;
422#endif
423#ifdef INET6
424	case AF_INET6:
425		sctp_ifnp->num_v6++;
426		break;
427#endif
428	default:
429		break;
430	}
431	if (sctp_ifnp->ifa_count == 1) {
432		/* register the new interface */
433		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
434		sctp_ifnp->registered_af = ifa_af;
435	}
436}
437
438
439/*-
440 * Remove an ifa from its ifn.
441 * If no more addresses exist, remove the ifn too. Otherwise, re-register
442 * the interface based on the remaining address families left.
443 * NOTE: ADDR write lock MUST be held.
444 */
445static void
446sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
447{
448	LIST_REMOVE(sctp_ifap, next_ifa);
449	if (sctp_ifap->ifn_p) {
450		/* update address counts */
451		sctp_ifap->ifn_p->ifa_count--;
452		switch (sctp_ifap->address.sa.sa_family) {
453#ifdef INET
454		case AF_INET:
455			sctp_ifap->ifn_p->num_v4--;
456			break;
457#endif
458#ifdef INET6
459		case AF_INET6:
460			sctp_ifap->ifn_p->num_v6--;
461			break;
462#endif
463		default:
464			break;
465		}
466
467		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
468			/* remove the ifn, possibly freeing it */
469			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
470		} else {
471			/* re-register address family type, if needed */
472			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
473			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
474				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
475				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
476				sctp_ifap->ifn_p->registered_af = AF_INET;
477			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
478			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
479				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
480				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
481				sctp_ifap->ifn_p->registered_af = AF_INET6;
482			}
483			/* free the ifn refcount */
484			sctp_free_ifn(sctp_ifap->ifn_p);
485		}
486		sctp_ifap->ifn_p = NULL;
487	}
488}
489
490
491struct sctp_ifa *
492sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
493    uint32_t ifn_type, const char *if_name, void *ifa,
494    struct sockaddr *addr, uint32_t ifa_flags,
495    int dynamic_add)
496{
497	struct sctp_vrf *vrf;
498	struct sctp_ifn *sctp_ifnp = NULL;
499	struct sctp_ifa *sctp_ifap = NULL;
500	struct sctp_ifalist *hash_addr_head;
501	struct sctp_ifnlist *hash_ifn_head;
502	uint32_t hash_of_addr;
503	int new_ifn_af = 0;
504
505#ifdef SCTP_DEBUG
506	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
507	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
508#endif
509	SCTP_IPI_ADDR_WLOCK();
510	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
511	if (sctp_ifnp) {
512		vrf = sctp_ifnp->vrf;
513	} else {
514		vrf = sctp_find_vrf(vrf_id);
515		if (vrf == NULL) {
516			vrf = sctp_allocate_vrf(vrf_id);
517			if (vrf == NULL) {
518				SCTP_IPI_ADDR_WUNLOCK();
519				return (NULL);
520			}
521		}
522	}
523	if (sctp_ifnp == NULL) {
524		/*
525		 * build one and add it, can't hold lock until after malloc
526		 * done though.
527		 */
528		SCTP_IPI_ADDR_WUNLOCK();
529		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
530		    sizeof(struct sctp_ifn), SCTP_M_IFN);
531		if (sctp_ifnp == NULL) {
532#ifdef INVARIANTS
533			panic("No memory for IFN");
534#endif
535			return (NULL);
536		}
537		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
538		sctp_ifnp->ifn_index = ifn_index;
539		sctp_ifnp->ifn_p = ifn;
540		sctp_ifnp->ifn_type = ifn_type;
541		sctp_ifnp->refcount = 0;
542		sctp_ifnp->vrf = vrf;
543		atomic_add_int(&vrf->refcount, 1);
544		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
545		if (if_name != NULL) {
546			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
547		} else {
548			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
549		}
550		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
551		LIST_INIT(&sctp_ifnp->ifalist);
552		SCTP_IPI_ADDR_WLOCK();
553		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
554		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
555		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
556		new_ifn_af = 1;
557	}
558	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
559	if (sctp_ifap) {
560		/* Hmm, it already exists? */
561		if ((sctp_ifap->ifn_p) &&
562		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
563			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
564			    sctp_ifap->ifn_p->ifn_name, ifn_index,
565			    (void *)sctp_ifap);
566			if (new_ifn_af) {
567				/* Remove the created one that we don't want */
568				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
569			}
570			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
571				/* easy to solve, just switch back to active */
572				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
573				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
574				sctp_ifap->ifn_p = sctp_ifnp;
575				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
576			}
577	exit_stage_left:
578			SCTP_IPI_ADDR_WUNLOCK();
579			return (sctp_ifap);
580		} else {
581			if (sctp_ifap->ifn_p) {
582				/*
583				 * The last IFN gets the address, remove the
584				 * old one
585				 */
586				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
587				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
588				    sctp_ifap->ifn_p->ifn_index, if_name,
589				    ifn_index);
590				/* remove the address from the old ifn */
591				sctp_remove_ifa_from_ifn(sctp_ifap);
592				/* move the address over to the new ifn */
593				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
594				goto exit_stage_left;
595			} else {
596				/* repair ifnp which was NULL ? */
597				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
598				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
599				    (void *)sctp_ifnp, (void *)sctp_ifap);
600				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
601			}
602			goto exit_stage_left;
603		}
604	}
605	SCTP_IPI_ADDR_WUNLOCK();
606	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
607	if (sctp_ifap == NULL) {
608#ifdef INVARIANTS
609		panic("No memory for IFA");
610#endif
611		return (NULL);
612	}
613	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
614	sctp_ifap->ifn_p = sctp_ifnp;
615	atomic_add_int(&sctp_ifnp->refcount, 1);
616	sctp_ifap->vrf_id = vrf_id;
617	sctp_ifap->ifa = ifa;
618	memcpy(&sctp_ifap->address, addr, addr->sa_len);
619	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
620	sctp_ifap->flags = ifa_flags;
621	/* Set scope */
622	switch (sctp_ifap->address.sa.sa_family) {
623#ifdef INET
624	case AF_INET:
625		{
626			struct sockaddr_in *sin;
627
628			sin = &sctp_ifap->address.sin;
629			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
630			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
631				sctp_ifap->src_is_loop = 1;
632			}
633			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
634				sctp_ifap->src_is_priv = 1;
635			}
636			sctp_ifnp->num_v4++;
637			if (new_ifn_af)
638				new_ifn_af = AF_INET;
639			break;
640		}
641#endif
642#ifdef INET6
643	case AF_INET6:
644		{
645			/* ok to use deprecated addresses? */
646			struct sockaddr_in6 *sin6;
647
648			sin6 = &sctp_ifap->address.sin6;
649			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
650			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
651				sctp_ifap->src_is_loop = 1;
652			}
653			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
654				sctp_ifap->src_is_priv = 1;
655			}
656			sctp_ifnp->num_v6++;
657			if (new_ifn_af)
658				new_ifn_af = AF_INET6;
659			break;
660		}
661#endif
662	default:
663		new_ifn_af = 0;
664		break;
665	}
666	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
667
668	if ((sctp_ifap->src_is_priv == 0) &&
669	    (sctp_ifap->src_is_loop == 0)) {
670		sctp_ifap->src_is_glob = 1;
671	}
672	SCTP_IPI_ADDR_WLOCK();
673	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
674	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
675	sctp_ifap->refcount = 1;
676	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
677	sctp_ifnp->ifa_count++;
678	vrf->total_ifa_count++;
679	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
680	if (new_ifn_af) {
681		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
682		sctp_ifnp->registered_af = new_ifn_af;
683	}
684	SCTP_IPI_ADDR_WUNLOCK();
685	if (dynamic_add) {
686		/*
687		 * Bump up the refcount so that when the timer completes it
688		 * will drop back down.
689		 */
690		struct sctp_laddr *wi;
691
692		atomic_add_int(&sctp_ifap->refcount, 1);
693		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
694		if (wi == NULL) {
695			/*
696			 * Gak, what can we do? We have lost an address
697			 * change can you say HOSED?
698			 */
699			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
700			/* Opps, must decrement the count */
701			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
702			    if_name);
703			return (NULL);
704		}
705		SCTP_INCR_LADDR_COUNT();
706		bzero(wi, sizeof(*wi));
707		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
708		wi->ifa = sctp_ifap;
709		wi->action = SCTP_ADD_IP_ADDRESS;
710
711		SCTP_WQ_ADDR_LOCK();
712		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
713		SCTP_WQ_ADDR_UNLOCK();
714
715		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
716		    (struct sctp_inpcb *)NULL,
717		    (struct sctp_tcb *)NULL,
718		    (struct sctp_nets *)NULL);
719	} else {
720		/* it's ready for use */
721		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
722	}
723	return (sctp_ifap);
724}
725
726void
727sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
728    uint32_t ifn_index, const char *if_name)
729{
730	struct sctp_vrf *vrf;
731	struct sctp_ifa *sctp_ifap = NULL;
732
733	SCTP_IPI_ADDR_WLOCK();
734	vrf = sctp_find_vrf(vrf_id);
735	if (vrf == NULL) {
736		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
737		goto out_now;
738	}
739#ifdef SCTP_DEBUG
740	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
741	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
742#endif
743	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
744	if (sctp_ifap) {
745		/* Validate the delete */
746		if (sctp_ifap->ifn_p) {
747			int valid = 0;
748
749			/*-
750			 * The name has priority over the ifn_index
751			 * if its given. We do this especially for
752			 * panda who might recycle indexes fast.
753			 */
754			if (if_name) {
755				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
756					/* They match its a correct delete */
757					valid = 1;
758				}
759			}
760			if (!valid) {
761				/* last ditch check ifn_index */
762				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
763					valid = 1;
764				}
765			}
766			if (!valid) {
767				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
768				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
769				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
770				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
771				SCTP_IPI_ADDR_WUNLOCK();
772				return;
773			}
774		}
775		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
776		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
777		/*
778		 * We don't set the flag. This means that the structure will
779		 * hang around in EP's that have bound specific to it until
780		 * they close. This gives us TCP like behavior if someone
781		 * removes an address (or for that matter adds it right
782		 * back).
783		 */
784		/* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */
785		vrf->total_ifa_count--;
786		LIST_REMOVE(sctp_ifap, next_bucket);
787		sctp_remove_ifa_from_ifn(sctp_ifap);
788	}
789#ifdef SCTP_DEBUG
790	else {
791		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
792		    ifn_index);
793		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
794	}
795#endif
796
797out_now:
798	SCTP_IPI_ADDR_WUNLOCK();
799	if (sctp_ifap) {
800		struct sctp_laddr *wi;
801
802		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
803		if (wi == NULL) {
804			/*
805			 * Gak, what can we do? We have lost an address
806			 * change can you say HOSED?
807			 */
808			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
809
810			/* Oops, must decrement the count */
811			sctp_free_ifa(sctp_ifap);
812			return;
813		}
814		SCTP_INCR_LADDR_COUNT();
815		bzero(wi, sizeof(*wi));
816		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
817		wi->ifa = sctp_ifap;
818		wi->action = SCTP_DEL_IP_ADDRESS;
819		SCTP_WQ_ADDR_LOCK();
820		/*
821		 * Should this really be a tailq? As it is we will process
822		 * the newest first :-0
823		 */
824		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
825		SCTP_WQ_ADDR_UNLOCK();
826
827		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
828		    (struct sctp_inpcb *)NULL,
829		    (struct sctp_tcb *)NULL,
830		    (struct sctp_nets *)NULL);
831	}
832	return;
833}
834
835
836static int
837sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
838{
839	int loopback_scope;
840
841#if defined(INET)
842	int ipv4_local_scope, ipv4_addr_legal;
843
844#endif
845#if defined(INET6)
846	int local_scope, site_scope, ipv6_addr_legal;
847
848#endif
849	struct sctp_vrf *vrf;
850	struct sctp_ifn *sctp_ifn;
851	struct sctp_ifa *sctp_ifa;
852
853	loopback_scope = stcb->asoc.scope.loopback_scope;
854#if defined(INET)
855	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
856	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
857#endif
858#if defined(INET6)
859	local_scope = stcb->asoc.scope.local_scope;
860	site_scope = stcb->asoc.scope.site_scope;
861	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
862#endif
863
864	SCTP_IPI_ADDR_RLOCK();
865	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
866	if (vrf == NULL) {
867		/* no vrf, no addresses */
868		SCTP_IPI_ADDR_RUNLOCK();
869		return (0);
870	}
871	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
872		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
873			if ((loopback_scope == 0) &&
874			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
875				continue;
876			}
877			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
878				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
879				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
880					/*
881					 * We allow pending addresses, where
882					 * we have sent an asconf-add to be
883					 * considered valid.
884					 */
885					continue;
886				}
887				if (sctp_ifa->address.sa.sa_family != to->sa_family) {
888					continue;
889				}
890				switch (sctp_ifa->address.sa.sa_family) {
891#ifdef INET
892				case AF_INET:
893					if (ipv4_addr_legal) {
894						struct sockaddr_in *sin,
895						           *rsin;
896
897						sin = &sctp_ifa->address.sin;
898						rsin = (struct sockaddr_in *)to;
899						if ((ipv4_local_scope == 0) &&
900						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
901							continue;
902						}
903						if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred,
904						    &sin->sin_addr) != 0) {
905							continue;
906						}
907						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
908							SCTP_IPI_ADDR_RUNLOCK();
909							return (1);
910						}
911					}
912					break;
913#endif
914#ifdef INET6
915				case AF_INET6:
916					if (ipv6_addr_legal) {
917						struct sockaddr_in6 *sin6,
918						            *rsin6;
919
920						sin6 = &sctp_ifa->address.sin6;
921						rsin6 = (struct sockaddr_in6 *)to;
922						if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred,
923						    &sin6->sin6_addr) != 0) {
924							continue;
925						}
926						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
927							if (local_scope == 0)
928								continue;
929							if (sin6->sin6_scope_id == 0) {
930								if (sa6_recoverscope(sin6) != 0)
931									continue;
932							}
933						}
934						if ((site_scope == 0) &&
935						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
936							continue;
937						}
938						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
939							SCTP_IPI_ADDR_RUNLOCK();
940							return (1);
941						}
942					}
943					break;
944#endif
945				default:
946					/* TSNH */
947					break;
948				}
949			}
950		}
951	} else {
952		struct sctp_laddr *laddr;
953
954		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
955			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
956				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
957				continue;
958			}
959			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
960			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
961				/*
962				 * We allow pending addresses, where we have
963				 * sent an asconf-add to be considered
964				 * valid.
965				 */
966				continue;
967			}
968			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
969				continue;
970			}
971			switch (to->sa_family) {
972#ifdef INET
973			case AF_INET:
974				{
975					struct sockaddr_in *sin, *rsin;
976
977					sin = &laddr->ifa->address.sin;
978					rsin = (struct sockaddr_in *)to;
979					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
980						SCTP_IPI_ADDR_RUNLOCK();
981						return (1);
982					}
983					break;
984				}
985#endif
986#ifdef INET6
987			case AF_INET6:
988				{
989					struct sockaddr_in6 *sin6, *rsin6;
990
991					sin6 = &laddr->ifa->address.sin6;
992					rsin6 = (struct sockaddr_in6 *)to;
993					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
994						SCTP_IPI_ADDR_RUNLOCK();
995						return (1);
996					}
997					break;
998				}
999
1000#endif
1001			default:
1002				/* TSNH */
1003				break;
1004			}
1005
1006		}
1007	}
1008	SCTP_IPI_ADDR_RUNLOCK();
1009	return (0);
1010}
1011
1012
1013static struct sctp_tcb *
1014sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
1015    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
1016{
1017	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
1018	/*
1019	 * If we support the TCP model, then we must now dig through to see
1020	 * if we can find our endpoint in the list of tcp ep's.
1021	 */
1022	uint16_t lport, rport;
1023	struct sctppcbhead *ephead;
1024	struct sctp_inpcb *inp;
1025	struct sctp_laddr *laddr;
1026	struct sctp_tcb *stcb;
1027	struct sctp_nets *net;
1028
1029	if ((to == NULL) || (from == NULL)) {
1030		return (NULL);
1031	}
1032	switch (to->sa_family) {
1033#ifdef INET
1034	case AF_INET:
1035		if (from->sa_family == AF_INET) {
1036			lport = ((struct sockaddr_in *)to)->sin_port;
1037			rport = ((struct sockaddr_in *)from)->sin_port;
1038		} else {
1039			return (NULL);
1040		}
1041		break;
1042#endif
1043#ifdef INET6
1044	case AF_INET6:
1045		if (from->sa_family == AF_INET6) {
1046			lport = ((struct sockaddr_in6 *)to)->sin6_port;
1047			rport = ((struct sockaddr_in6 *)from)->sin6_port;
1048		} else {
1049			return (NULL);
1050		}
1051		break;
1052#endif
1053	default:
1054		return (NULL);
1055	}
1056	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
1057	/*
1058	 * Ok now for each of the guys in this bucket we must look and see:
1059	 * - Does the remote port match. - Does there single association's
1060	 * addresses match this address (to). If so we update p_ep to point
1061	 * to this ep and return the tcb from it.
1062	 */
1063	LIST_FOREACH(inp, ephead, sctp_hash) {
1064		SCTP_INP_RLOCK(inp);
1065		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1066			SCTP_INP_RUNLOCK(inp);
1067			continue;
1068		}
1069		if (lport != inp->sctp_lport) {
1070			SCTP_INP_RUNLOCK(inp);
1071			continue;
1072		}
1073		switch (to->sa_family) {
1074#ifdef INET
1075		case AF_INET:
1076			{
1077				struct sockaddr_in *sin;
1078
1079				sin = (struct sockaddr_in *)to;
1080				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1081				    &sin->sin_addr) != 0) {
1082					SCTP_INP_RUNLOCK(inp);
1083					continue;
1084				}
1085				break;
1086			}
1087#endif
1088#ifdef INET6
1089		case AF_INET6:
1090			{
1091				struct sockaddr_in6 *sin6;
1092
1093				sin6 = (struct sockaddr_in6 *)to;
1094				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1095				    &sin6->sin6_addr) != 0) {
1096					SCTP_INP_RUNLOCK(inp);
1097					continue;
1098				}
1099				break;
1100			}
1101#endif
1102		default:
1103			SCTP_INP_RUNLOCK(inp);
1104			continue;
1105		}
1106		if (inp->def_vrf_id != vrf_id) {
1107			SCTP_INP_RUNLOCK(inp);
1108			continue;
1109		}
1110		/* check to see if the ep has one of the addresses */
1111		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1112			/* We are NOT bound all, so look further */
1113			int match = 0;
1114
1115			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1116
1117				if (laddr->ifa == NULL) {
1118					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
1119					continue;
1120				}
1121				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1122					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1123					continue;
1124				}
1125				if (laddr->ifa->address.sa.sa_family ==
1126				    to->sa_family) {
1127					/* see if it matches */
1128#ifdef INET
1129					if (from->sa_family == AF_INET) {
1130						struct sockaddr_in *intf_addr,
1131						           *sin;
1132
1133						intf_addr = &laddr->ifa->address.sin;
1134						sin = (struct sockaddr_in *)to;
1135						if (sin->sin_addr.s_addr ==
1136						    intf_addr->sin_addr.s_addr) {
1137							match = 1;
1138							break;
1139						}
1140					}
1141#endif
1142#ifdef INET6
1143					if (from->sa_family == AF_INET6) {
1144						struct sockaddr_in6 *intf_addr6;
1145						struct sockaddr_in6 *sin6;
1146
1147						sin6 = (struct sockaddr_in6 *)
1148						    to;
1149						intf_addr6 = &laddr->ifa->address.sin6;
1150
1151						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1152						    intf_addr6)) {
1153							match = 1;
1154							break;
1155						}
1156					}
1157#endif
1158				}
1159			}
1160			if (match == 0) {
1161				/* This endpoint does not have this address */
1162				SCTP_INP_RUNLOCK(inp);
1163				continue;
1164			}
1165		}
1166		/*
1167		 * Ok if we hit here the ep has the address, does it hold
1168		 * the tcb?
1169		 */
1170		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1171		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1172		if (stcb == NULL) {
1173			SCTP_INP_RUNLOCK(inp);
1174			continue;
1175		}
1176		SCTP_TCB_LOCK(stcb);
1177		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1178			SCTP_TCB_UNLOCK(stcb);
1179			SCTP_INP_RUNLOCK(inp);
1180			continue;
1181		}
1182		if (stcb->rport != rport) {
1183			/* remote port does not match. */
1184			SCTP_TCB_UNLOCK(stcb);
1185			SCTP_INP_RUNLOCK(inp);
1186			continue;
1187		}
1188		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1189			SCTP_TCB_UNLOCK(stcb);
1190			SCTP_INP_RUNLOCK(inp);
1191			continue;
1192		}
1193		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1194			SCTP_TCB_UNLOCK(stcb);
1195			SCTP_INP_RUNLOCK(inp);
1196			continue;
1197		}
1198		/* Does this TCB have a matching address? */
1199		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1200
1201			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1202				/* not the same family, can't be a match */
1203				continue;
1204			}
1205			switch (from->sa_family) {
1206#ifdef INET
1207			case AF_INET:
1208				{
1209					struct sockaddr_in *sin, *rsin;
1210
1211					sin = (struct sockaddr_in *)&net->ro._l_addr;
1212					rsin = (struct sockaddr_in *)from;
1213					if (sin->sin_addr.s_addr ==
1214					    rsin->sin_addr.s_addr) {
1215						/* found it */
1216						if (netp != NULL) {
1217							*netp = net;
1218						}
1219						/*
1220						 * Update the endpoint
1221						 * pointer
1222						 */
1223						*inp_p = inp;
1224						SCTP_INP_RUNLOCK(inp);
1225						return (stcb);
1226					}
1227					break;
1228				}
1229#endif
1230#ifdef INET6
1231			case AF_INET6:
1232				{
1233					struct sockaddr_in6 *sin6, *rsin6;
1234
1235					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1236					rsin6 = (struct sockaddr_in6 *)from;
1237					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1238					    rsin6)) {
1239						/* found it */
1240						if (netp != NULL) {
1241							*netp = net;
1242						}
1243						/*
1244						 * Update the endpoint
1245						 * pointer
1246						 */
1247						*inp_p = inp;
1248						SCTP_INP_RUNLOCK(inp);
1249						return (stcb);
1250					}
1251					break;
1252				}
1253#endif
1254			default:
1255				/* TSNH */
1256				break;
1257			}
1258		}
1259		SCTP_TCB_UNLOCK(stcb);
1260		SCTP_INP_RUNLOCK(inp);
1261	}
1262	return (NULL);
1263}
1264
1265
1266/*
1267 * rules for use
1268 *
1269 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1270 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1271 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1272 * NULL.
1273 */
1274
1275struct sctp_tcb *
1276sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1277    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1278{
1279	struct sctpasochead *head;
1280	struct sctp_inpcb *inp;
1281	struct sctp_tcb *stcb = NULL;
1282	struct sctp_nets *net;
1283	uint16_t rport;
1284
1285	inp = *inp_p;
1286	switch (remote->sa_family) {
1287#ifdef INET
1288	case AF_INET:
1289		rport = (((struct sockaddr_in *)remote)->sin_port);
1290		break;
1291#endif
1292#ifdef INET6
1293	case AF_INET6:
1294		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1295		break;
1296#endif
1297	default:
1298		return (NULL);
1299	}
1300	if (locked_tcb) {
1301		/*
1302		 * UN-lock so we can do proper locking here this occurs when
1303		 * called from load_addresses_from_init.
1304		 */
1305		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1306		SCTP_TCB_UNLOCK(locked_tcb);
1307	}
1308	SCTP_INP_INFO_RLOCK();
1309	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1310	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1311		/*-
1312		 * Now either this guy is our listener or it's the
1313		 * connector. If it is the one that issued the connect, then
1314		 * it's only chance is to be the first TCB in the list. If
1315		 * it is the acceptor, then do the special_lookup to hash
1316		 * and find the real inp.
1317		 */
1318		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1319			/* to is peer addr, from is my addr */
1320			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1321			    netp, inp->def_vrf_id);
1322			if ((stcb != NULL) && (locked_tcb == NULL)) {
1323				/* we have a locked tcb, lower refcount */
1324				SCTP_INP_DECR_REF(inp);
1325			}
1326			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1327				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1328				SCTP_TCB_LOCK(locked_tcb);
1329				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1330				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1331			}
1332			SCTP_INP_INFO_RUNLOCK();
1333			return (stcb);
1334		} else {
1335			SCTP_INP_WLOCK(inp);
1336			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1337				goto null_return;
1338			}
1339			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1340			if (stcb == NULL) {
1341				goto null_return;
1342			}
1343			SCTP_TCB_LOCK(stcb);
1344
1345			if (stcb->rport != rport) {
1346				/* remote port does not match. */
1347				SCTP_TCB_UNLOCK(stcb);
1348				goto null_return;
1349			}
1350			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1351				SCTP_TCB_UNLOCK(stcb);
1352				goto null_return;
1353			}
1354			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1355				SCTP_TCB_UNLOCK(stcb);
1356				goto null_return;
1357			}
1358			/* now look at the list of remote addresses */
1359			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1360#ifdef INVARIANTS
1361				if (net == (TAILQ_NEXT(net, sctp_next))) {
1362					panic("Corrupt net list");
1363				}
1364#endif
1365				if (net->ro._l_addr.sa.sa_family !=
1366				    remote->sa_family) {
1367					/* not the same family */
1368					continue;
1369				}
1370				switch (remote->sa_family) {
1371#ifdef INET
1372				case AF_INET:
1373					{
1374						struct sockaddr_in *sin,
1375						           *rsin;
1376
1377						sin = (struct sockaddr_in *)
1378						    &net->ro._l_addr;
1379						rsin = (struct sockaddr_in *)remote;
1380						if (sin->sin_addr.s_addr ==
1381						    rsin->sin_addr.s_addr) {
1382							/* found it */
1383							if (netp != NULL) {
1384								*netp = net;
1385							}
1386							if (locked_tcb == NULL) {
1387								SCTP_INP_DECR_REF(inp);
1388							} else if (locked_tcb != stcb) {
1389								SCTP_TCB_LOCK(locked_tcb);
1390							}
1391							if (locked_tcb) {
1392								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1393							}
1394							SCTP_INP_WUNLOCK(inp);
1395							SCTP_INP_INFO_RUNLOCK();
1396							return (stcb);
1397						}
1398						break;
1399					}
1400#endif
1401#ifdef INET6
1402				case AF_INET6:
1403					{
1404						struct sockaddr_in6 *sin6,
1405						            *rsin6;
1406
1407						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1408						rsin6 = (struct sockaddr_in6 *)remote;
1409						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1410						    rsin6)) {
1411							/* found it */
1412							if (netp != NULL) {
1413								*netp = net;
1414							}
1415							if (locked_tcb == NULL) {
1416								SCTP_INP_DECR_REF(inp);
1417							} else if (locked_tcb != stcb) {
1418								SCTP_TCB_LOCK(locked_tcb);
1419							}
1420							if (locked_tcb) {
1421								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1422							}
1423							SCTP_INP_WUNLOCK(inp);
1424							SCTP_INP_INFO_RUNLOCK();
1425							return (stcb);
1426						}
1427						break;
1428					}
1429#endif
1430				default:
1431					/* TSNH */
1432					break;
1433				}
1434			}
1435			SCTP_TCB_UNLOCK(stcb);
1436		}
1437	} else {
1438		SCTP_INP_WLOCK(inp);
1439		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1440			goto null_return;
1441		}
1442		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1443		    inp->sctp_hashmark)];
1444		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1445			if (stcb->rport != rport) {
1446				/* remote port does not match */
1447				continue;
1448			}
1449			SCTP_TCB_LOCK(stcb);
1450			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1451				SCTP_TCB_UNLOCK(stcb);
1452				continue;
1453			}
1454			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1455				SCTP_TCB_UNLOCK(stcb);
1456				continue;
1457			}
1458			/* now look at the list of remote addresses */
1459			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1460#ifdef INVARIANTS
1461				if (net == (TAILQ_NEXT(net, sctp_next))) {
1462					panic("Corrupt net list");
1463				}
1464#endif
1465				if (net->ro._l_addr.sa.sa_family !=
1466				    remote->sa_family) {
1467					/* not the same family */
1468					continue;
1469				}
1470				switch (remote->sa_family) {
1471#ifdef INET
1472				case AF_INET:
1473					{
1474						struct sockaddr_in *sin,
1475						           *rsin;
1476
1477						sin = (struct sockaddr_in *)
1478						    &net->ro._l_addr;
1479						rsin = (struct sockaddr_in *)remote;
1480						if (sin->sin_addr.s_addr ==
1481						    rsin->sin_addr.s_addr) {
1482							/* found it */
1483							if (netp != NULL) {
1484								*netp = net;
1485							}
1486							if (locked_tcb == NULL) {
1487								SCTP_INP_DECR_REF(inp);
1488							} else if (locked_tcb != stcb) {
1489								SCTP_TCB_LOCK(locked_tcb);
1490							}
1491							if (locked_tcb) {
1492								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1493							}
1494							SCTP_INP_WUNLOCK(inp);
1495							SCTP_INP_INFO_RUNLOCK();
1496							return (stcb);
1497						}
1498						break;
1499					}
1500#endif
1501#ifdef INET6
1502				case AF_INET6:
1503					{
1504						struct sockaddr_in6 *sin6,
1505						            *rsin6;
1506
1507						sin6 = (struct sockaddr_in6 *)
1508						    &net->ro._l_addr;
1509						rsin6 = (struct sockaddr_in6 *)remote;
1510						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1511						    rsin6)) {
1512							/* found it */
1513							if (netp != NULL) {
1514								*netp = net;
1515							}
1516							if (locked_tcb == NULL) {
1517								SCTP_INP_DECR_REF(inp);
1518							} else if (locked_tcb != stcb) {
1519								SCTP_TCB_LOCK(locked_tcb);
1520							}
1521							if (locked_tcb) {
1522								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1523							}
1524							SCTP_INP_WUNLOCK(inp);
1525							SCTP_INP_INFO_RUNLOCK();
1526							return (stcb);
1527						}
1528						break;
1529					}
1530#endif
1531				default:
1532					/* TSNH */
1533					break;
1534				}
1535			}
1536			SCTP_TCB_UNLOCK(stcb);
1537		}
1538	}
1539null_return:
1540	/* clean up for returning null */
1541	if (locked_tcb) {
1542		SCTP_TCB_LOCK(locked_tcb);
1543		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1544	}
1545	SCTP_INP_WUNLOCK(inp);
1546	SCTP_INP_INFO_RUNLOCK();
1547	/* not found */
1548	return (NULL);
1549}
1550
1551
1552/*
1553 * Find an association for a specific endpoint using the association id given
1554 * out in the COMM_UP notification
1555 */
1556struct sctp_tcb *
1557sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1558{
1559	/*
1560	 * Use my the assoc_id to find a endpoint
1561	 */
1562	struct sctpasochead *head;
1563	struct sctp_tcb *stcb;
1564	uint32_t id;
1565
1566	if (inp == NULL) {
1567		SCTP_PRINTF("TSNH ep_associd\n");
1568		return (NULL);
1569	}
1570	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1571		SCTP_PRINTF("TSNH ep_associd0\n");
1572		return (NULL);
1573	}
1574	id = (uint32_t) asoc_id;
1575	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1576	if (head == NULL) {
1577		/* invalid id TSNH */
1578		SCTP_PRINTF("TSNH ep_associd1\n");
1579		return (NULL);
1580	}
1581	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1582		if (stcb->asoc.assoc_id == id) {
1583			if (inp != stcb->sctp_ep) {
1584				/*
1585				 * some other guy has the same id active (id
1586				 * collision ??).
1587				 */
1588				SCTP_PRINTF("TSNH ep_associd2\n");
1589				continue;
1590			}
1591			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1592				continue;
1593			}
1594			if (want_lock) {
1595				SCTP_TCB_LOCK(stcb);
1596			}
1597			return (stcb);
1598		}
1599	}
1600	return (NULL);
1601}
1602
1603
1604struct sctp_tcb *
1605sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1606{
1607	struct sctp_tcb *stcb;
1608
1609	SCTP_INP_RLOCK(inp);
1610	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1611	SCTP_INP_RUNLOCK(inp);
1612	return (stcb);
1613}
1614
1615
1616/*
1617 * Endpoint probe expects that the INP_INFO is locked.
1618 */
1619static struct sctp_inpcb *
1620sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1621    uint16_t lport, uint32_t vrf_id)
1622{
1623	struct sctp_inpcb *inp;
1624	struct sctp_laddr *laddr;
1625
1626#ifdef INET
1627	struct sockaddr_in *sin;
1628
1629#endif
1630#ifdef INET6
1631	struct sockaddr_in6 *sin6;
1632	struct sockaddr_in6 *intf_addr6;
1633
1634#endif
1635	int fnd;
1636
1637#ifdef INET
1638	sin = NULL;
1639#endif
1640#ifdef INET6
1641	sin6 = NULL;
1642#endif
1643	switch (nam->sa_family) {
1644#ifdef INET
1645	case AF_INET:
1646		sin = (struct sockaddr_in *)nam;
1647		break;
1648#endif
1649#ifdef INET6
1650	case AF_INET6:
1651		sin6 = (struct sockaddr_in6 *)nam;
1652		break;
1653#endif
1654	default:
1655		/* unsupported family */
1656		return (NULL);
1657	}
1658
1659	if (head == NULL)
1660		return (NULL);
1661
1662	LIST_FOREACH(inp, head, sctp_hash) {
1663		SCTP_INP_RLOCK(inp);
1664		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1665			SCTP_INP_RUNLOCK(inp);
1666			continue;
1667		}
1668		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1669		    (inp->sctp_lport == lport)) {
1670			/* got it */
1671			switch (nam->sa_family) {
1672#ifdef INET
1673			case AF_INET:
1674				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1675				    SCTP_IPV6_V6ONLY(inp)) {
1676					/*
1677					 * IPv4 on a IPv6 socket with ONLY
1678					 * IPv6 set
1679					 */
1680					SCTP_INP_RUNLOCK(inp);
1681					continue;
1682				}
1683				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1684				    &sin->sin_addr) != 0) {
1685					SCTP_INP_RUNLOCK(inp);
1686					continue;
1687				}
1688				break;
1689#endif
1690#ifdef INET6
1691			case AF_INET6:
1692				/*
1693				 * A V6 address and the endpoint is NOT
1694				 * bound V6
1695				 */
1696				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1697					SCTP_INP_RUNLOCK(inp);
1698					continue;
1699				}
1700				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1701				    &sin6->sin6_addr) != 0) {
1702					SCTP_INP_RUNLOCK(inp);
1703					continue;
1704				}
1705				break;
1706#endif
1707			default:
1708				break;
1709			}
1710			/* does a VRF id match? */
1711			fnd = 0;
1712			if (inp->def_vrf_id == vrf_id)
1713				fnd = 1;
1714
1715			SCTP_INP_RUNLOCK(inp);
1716			if (!fnd)
1717				continue;
1718			return (inp);
1719		}
1720		SCTP_INP_RUNLOCK(inp);
1721	}
1722	switch (nam->sa_family) {
1723#ifdef INET
1724	case AF_INET:
1725		if (sin->sin_addr.s_addr == INADDR_ANY) {
1726			/* Can't hunt for one that has no address specified */
1727			return (NULL);
1728		}
1729		break;
1730#endif
1731#ifdef INET6
1732	case AF_INET6:
1733		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1734			/* Can't hunt for one that has no address specified */
1735			return (NULL);
1736		}
1737		break;
1738#endif
1739	default:
1740		break;
1741	}
1742	/*
1743	 * ok, not bound to all so see if we can find a EP bound to this
1744	 * address.
1745	 */
1746	LIST_FOREACH(inp, head, sctp_hash) {
1747		SCTP_INP_RLOCK(inp);
1748		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1749			SCTP_INP_RUNLOCK(inp);
1750			continue;
1751		}
1752		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1753			SCTP_INP_RUNLOCK(inp);
1754			continue;
1755		}
1756		/*
1757		 * Ok this could be a likely candidate, look at all of its
1758		 * addresses
1759		 */
1760		if (inp->sctp_lport != lport) {
1761			SCTP_INP_RUNLOCK(inp);
1762			continue;
1763		}
1764		/* does a VRF id match? */
1765		fnd = 0;
1766		if (inp->def_vrf_id == vrf_id)
1767			fnd = 1;
1768
1769		if (!fnd) {
1770			SCTP_INP_RUNLOCK(inp);
1771			continue;
1772		}
1773		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1774			if (laddr->ifa == NULL) {
1775				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1776				    __FUNCTION__);
1777				continue;
1778			}
1779			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1780			    (void *)laddr->ifa);
1781			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1782				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1783				continue;
1784			}
1785			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1786				/* possible, see if it matches */
1787				switch (nam->sa_family) {
1788#ifdef INET
1789				case AF_INET:
1790					if (sin->sin_addr.s_addr ==
1791					    laddr->ifa->address.sin.sin_addr.s_addr) {
1792						SCTP_INP_RUNLOCK(inp);
1793						return (inp);
1794					}
1795					break;
1796#endif
1797#ifdef INET6
1798				case AF_INET6:
1799					intf_addr6 = &laddr->ifa->address.sin6;
1800					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1801					    intf_addr6)) {
1802						SCTP_INP_RUNLOCK(inp);
1803						return (inp);
1804					}
1805					break;
1806#endif
1807				}
1808			}
1809		}
1810		SCTP_INP_RUNLOCK(inp);
1811	}
1812	return (NULL);
1813}
1814
1815
1816static struct sctp_inpcb *
1817sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1818{
1819	struct sctppcbhead *head;
1820	struct sctp_inpcb *t_inp;
1821	int fnd;
1822
1823	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1824	    SCTP_BASE_INFO(hashmark))];
1825	LIST_FOREACH(t_inp, head, sctp_hash) {
1826		if (t_inp->sctp_lport != lport) {
1827			continue;
1828		}
1829		/* is it in the VRF in question */
1830		fnd = 0;
1831		if (t_inp->def_vrf_id == vrf_id)
1832			fnd = 1;
1833		if (!fnd)
1834			continue;
1835
1836		/* This one is in use. */
1837		/* check the v6/v4 binding issue */
1838		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1839		    SCTP_IPV6_V6ONLY(t_inp)) {
1840			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1841				/* collision in V6 space */
1842				return (t_inp);
1843			} else {
1844				/* inp is BOUND_V4 no conflict */
1845				continue;
1846			}
1847		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1848			/* t_inp is bound v4 and v6, conflict always */
1849			return (t_inp);
1850		} else {
1851			/* t_inp is bound only V4 */
1852			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1853			    SCTP_IPV6_V6ONLY(inp)) {
1854				/* no conflict */
1855				continue;
1856			}
1857			/* else fall through to conflict */
1858		}
1859		return (t_inp);
1860	}
1861	return (NULL);
1862}
1863
1864
1865int
1866sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1867{
1868	/* For 1-2-1 with port reuse */
1869	struct sctppcbhead *head;
1870	struct sctp_inpcb *tinp;
1871
1872	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1873		/* only works with port reuse on */
1874		return (-1);
1875	}
1876	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1877		return (0);
1878	}
1879	SCTP_INP_RUNLOCK(inp);
1880	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1881	    SCTP_BASE_INFO(hashmark))];
1882	/* Kick out all non-listeners to the TCP hash */
1883	LIST_FOREACH(tinp, head, sctp_hash) {
1884		if (tinp->sctp_lport != inp->sctp_lport) {
1885			continue;
1886		}
1887		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1888			continue;
1889		}
1890		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1891			continue;
1892		}
1893		if (tinp->sctp_socket->so_qlimit) {
1894			continue;
1895		}
1896		SCTP_INP_WLOCK(tinp);
1897		LIST_REMOVE(tinp, sctp_hash);
1898		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1899		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1900		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1901		SCTP_INP_WUNLOCK(tinp);
1902	}
1903	SCTP_INP_WLOCK(inp);
1904	/* Pull from where he was */
1905	LIST_REMOVE(inp, sctp_hash);
1906	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1907	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1908	LIST_INSERT_HEAD(head, inp, sctp_hash);
1909	SCTP_INP_WUNLOCK(inp);
1910	SCTP_INP_RLOCK(inp);
1911	return (0);
1912}
1913
1914
1915struct sctp_inpcb *
1916sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1917    uint32_t vrf_id)
1918{
1919	/*
1920	 * First we check the hash table to see if someone has this port
1921	 * bound with just the port.
1922	 */
1923	struct sctp_inpcb *inp;
1924	struct sctppcbhead *head;
1925	int lport;
1926	unsigned int i;
1927
1928#ifdef INET
1929	struct sockaddr_in *sin;
1930
1931#endif
1932#ifdef INET6
1933	struct sockaddr_in6 *sin6;
1934
1935#endif
1936
1937	switch (nam->sa_family) {
1938#ifdef INET
1939	case AF_INET:
1940		sin = (struct sockaddr_in *)nam;
1941		lport = sin->sin_port;
1942		break;
1943#endif
1944#ifdef INET6
1945	case AF_INET6:
1946		sin6 = (struct sockaddr_in6 *)nam;
1947		lport = sin6->sin6_port;
1948		break;
1949#endif
1950	default:
1951		return (NULL);
1952	}
1953	/*
1954	 * I could cheat here and just cast to one of the types but we will
1955	 * do it right. It also provides the check against an Unsupported
1956	 * type too.
1957	 */
1958	/* Find the head of the ALLADDR chain */
1959	if (have_lock == 0) {
1960		SCTP_INP_INFO_RLOCK();
1961	}
1962	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1963	    SCTP_BASE_INFO(hashmark))];
1964	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1965
1966	/*
1967	 * If the TCP model exists it could be that the main listening
1968	 * endpoint is gone but there still exists a connected socket for
1969	 * this guy. If so we can return the first one that we find. This
1970	 * may NOT be the correct one so the caller should be wary on the
1971	 * returned INP. Currently the only caller that sets find_tcp_pool
1972	 * is in bindx where we are verifying that a user CAN bind the
1973	 * address. He either has bound it already, or someone else has, or
1974	 * its open to bind, so this is good enough.
1975	 */
1976	if (inp == NULL && find_tcp_pool) {
1977		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1978			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1979			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1980			if (inp) {
1981				break;
1982			}
1983		}
1984	}
1985	if (inp) {
1986		SCTP_INP_INCR_REF(inp);
1987	}
1988	if (have_lock == 0) {
1989		SCTP_INP_INFO_RUNLOCK();
1990	}
1991	return (inp);
1992}
1993
1994
1995/*
1996 * Find an association for an endpoint with the pointer to whom you want to
1997 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1998 * need to change the *to to some other struct like a mbuf...
1999 */
2000struct sctp_tcb *
2001sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
2002    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
2003    uint32_t vrf_id)
2004{
2005	struct sctp_inpcb *inp = NULL;
2006	struct sctp_tcb *stcb;
2007
2008	SCTP_INP_INFO_RLOCK();
2009	if (find_tcp_pool) {
2010		if (inp_p != NULL) {
2011			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
2012			    vrf_id);
2013		} else {
2014			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
2015			    vrf_id);
2016		}
2017		if (stcb != NULL) {
2018			SCTP_INP_INFO_RUNLOCK();
2019			return (stcb);
2020		}
2021	}
2022	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
2023	if (inp_p != NULL) {
2024		*inp_p = inp;
2025	}
2026	SCTP_INP_INFO_RUNLOCK();
2027	if (inp == NULL) {
2028		return (NULL);
2029	}
2030	/*
2031	 * ok, we have an endpoint, now lets find the assoc for it (if any)
2032	 * we now place the source address or from in the to of the find
2033	 * endpoint call. Since in reality this chain is used from the
2034	 * inbound packet side.
2035	 */
2036	if (inp_p != NULL) {
2037		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
2038		    NULL);
2039	} else {
2040		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
2041		    NULL);
2042	}
2043	return (stcb);
2044}
2045
2046
2047/*
2048 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
2049 * find all addresses that the sender has specified in any address list. Each
2050 * address will be used to lookup the TCB and see if one exits.
2051 */
2052static struct sctp_tcb *
2053sctp_findassociation_special_addr(struct mbuf *m, int offset,
2054    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
2055    struct sockaddr *dst)
2056{
2057	struct sctp_paramhdr *phdr, parm_buf;
2058
2059#if defined(INET) || defined(INET6)
2060	struct sctp_tcb *stcb;
2061	uint16_t ptype;
2062
2063#endif
2064	uint16_t plen;
2065
2066#ifdef INET
2067	struct sockaddr_in sin4;
2068
2069#endif
2070#ifdef INET6
2071	struct sockaddr_in6 sin6;
2072
2073#endif
2074
2075#ifdef INET
2076	memset(&sin4, 0, sizeof(sin4));
2077	sin4.sin_len = sizeof(sin4);
2078	sin4.sin_family = AF_INET;
2079	sin4.sin_port = sh->src_port;
2080#endif
2081#ifdef INET6
2082	memset(&sin6, 0, sizeof(sin6));
2083	sin6.sin6_len = sizeof(sin6);
2084	sin6.sin6_family = AF_INET6;
2085	sin6.sin6_port = sh->src_port;
2086#endif
2087
2088	offset += sizeof(struct sctp_init_chunk);
2089
2090	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2091	while (phdr != NULL) {
2092		/* now we must see if we want the parameter */
2093#if defined(INET) || defined(INET6)
2094		ptype = ntohs(phdr->param_type);
2095#endif
2096		plen = ntohs(phdr->param_length);
2097		if (plen == 0) {
2098			break;
2099		}
2100#ifdef INET
2101		if (ptype == SCTP_IPV4_ADDRESS &&
2102		    plen == sizeof(struct sctp_ipv4addr_param)) {
2103			/* Get the rest of the address */
2104			struct sctp_ipv4addr_param ip4_parm, *p4;
2105
2106			phdr = sctp_get_next_param(m, offset,
2107			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2108			if (phdr == NULL) {
2109				return (NULL);
2110			}
2111			p4 = (struct sctp_ipv4addr_param *)phdr;
2112			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2113			/* look it up */
2114			stcb = sctp_findassociation_ep_addr(inp_p,
2115			    (struct sockaddr *)&sin4, netp, dst, NULL);
2116			if (stcb != NULL) {
2117				return (stcb);
2118			}
2119		}
2120#endif
2121#ifdef INET6
2122		if (ptype == SCTP_IPV6_ADDRESS &&
2123		    plen == sizeof(struct sctp_ipv6addr_param)) {
2124			/* Get the rest of the address */
2125			struct sctp_ipv6addr_param ip6_parm, *p6;
2126
2127			phdr = sctp_get_next_param(m, offset,
2128			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2129			if (phdr == NULL) {
2130				return (NULL);
2131			}
2132			p6 = (struct sctp_ipv6addr_param *)phdr;
2133			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2134			/* look it up */
2135			stcb = sctp_findassociation_ep_addr(inp_p,
2136			    (struct sockaddr *)&sin6, netp, dst, NULL);
2137			if (stcb != NULL) {
2138				return (stcb);
2139			}
2140		}
2141#endif
2142		offset += SCTP_SIZE32(plen);
2143		phdr = sctp_get_next_param(m, offset, &parm_buf,
2144		    sizeof(parm_buf));
2145	}
2146	return (NULL);
2147}
2148
2149static struct sctp_tcb *
2150sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2151    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2152    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2153{
2154	/*
2155	 * Use my vtag to hash. If we find it we then verify the source addr
2156	 * is in the assoc. If all goes well we save a bit on rec of a
2157	 * packet.
2158	 */
2159	struct sctpasochead *head;
2160	struct sctp_nets *net;
2161	struct sctp_tcb *stcb;
2162
2163	SCTP_INP_INFO_RLOCK();
2164	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2165	    SCTP_BASE_INFO(hashasocmark))];
2166	LIST_FOREACH(stcb, head, sctp_asocs) {
2167		SCTP_INP_RLOCK(stcb->sctp_ep);
2168		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2169			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2170			continue;
2171		}
2172		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2173			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2174			continue;
2175		}
2176		SCTP_TCB_LOCK(stcb);
2177		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2178		if (stcb->asoc.my_vtag == vtag) {
2179			/* candidate */
2180			if (stcb->rport != rport) {
2181				SCTP_TCB_UNLOCK(stcb);
2182				continue;
2183			}
2184			if (stcb->sctp_ep->sctp_lport != lport) {
2185				SCTP_TCB_UNLOCK(stcb);
2186				continue;
2187			}
2188			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2189				SCTP_TCB_UNLOCK(stcb);
2190				continue;
2191			}
2192			/* RRS:Need toaddr check here */
2193			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2194				/* Endpoint does not own this address */
2195				SCTP_TCB_UNLOCK(stcb);
2196				continue;
2197			}
2198			if (remote_tag) {
2199				/*
2200				 * If we have both vtags that's all we match
2201				 * on
2202				 */
2203				if (stcb->asoc.peer_vtag == remote_tag) {
2204					/*
2205					 * If both tags match we consider it
2206					 * conclusive and check NO
2207					 * source/destination addresses
2208					 */
2209					goto conclusive;
2210				}
2211			}
2212			if (skip_src_check) {
2213		conclusive:
2214				if (from) {
2215					*netp = sctp_findnet(stcb, from);
2216				} else {
2217					*netp = NULL;	/* unknown */
2218				}
2219				if (inp_p)
2220					*inp_p = stcb->sctp_ep;
2221				SCTP_INP_INFO_RUNLOCK();
2222				return (stcb);
2223			}
2224			net = sctp_findnet(stcb, from);
2225			if (net) {
2226				/* yep its him. */
2227				*netp = net;
2228				SCTP_STAT_INCR(sctps_vtagexpress);
2229				*inp_p = stcb->sctp_ep;
2230				SCTP_INP_INFO_RUNLOCK();
2231				return (stcb);
2232			} else {
2233				/*
2234				 * not him, this should only happen in rare
2235				 * cases so I peg it.
2236				 */
2237				SCTP_STAT_INCR(sctps_vtagbogus);
2238			}
2239		}
2240		SCTP_TCB_UNLOCK(stcb);
2241	}
2242	SCTP_INP_INFO_RUNLOCK();
2243	return (NULL);
2244}
2245
2246
2247/*
2248 * Find an association with the pointer to the inbound IP packet. This can be
2249 * a IPv4 or IPv6 packet.
2250 */
2251struct sctp_tcb *
2252sctp_findassociation_addr(struct mbuf *m, int offset,
2253    struct sockaddr *src, struct sockaddr *dst,
2254    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2255    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2256{
2257	int find_tcp_pool;
2258	struct sctp_tcb *stcb;
2259	struct sctp_inpcb *inp;
2260
2261	if (sh->v_tag) {
2262		/* we only go down this path if vtag is non-zero */
2263		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2264		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2265		if (stcb) {
2266			return (stcb);
2267		}
2268	}
2269	find_tcp_pool = 0;
2270	if ((ch->chunk_type != SCTP_INITIATION) &&
2271	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2272	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2273	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2274		/* Other chunk types go to the tcp pool. */
2275		find_tcp_pool = 1;
2276	}
2277	if (inp_p) {
2278		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2279		    find_tcp_pool, vrf_id);
2280		inp = *inp_p;
2281	} else {
2282		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2283		    find_tcp_pool, vrf_id);
2284	}
2285	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2286	if (stcb == NULL && inp) {
2287		/* Found a EP but not this address */
2288		if ((ch->chunk_type == SCTP_INITIATION) ||
2289		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2290			/*-
2291			 * special hook, we do NOT return linp or an
2292			 * association that is linked to an existing
2293			 * association that is under the TCP pool (i.e. no
2294			 * listener exists). The endpoint finding routine
2295			 * will always find a listener before examining the
2296			 * TCP pool.
2297			 */
2298			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2299				if (inp_p) {
2300					*inp_p = NULL;
2301				}
2302				return (NULL);
2303			}
2304			stcb = sctp_findassociation_special_addr(m,
2305			    offset, sh, &inp, netp, dst);
2306			if (inp_p != NULL) {
2307				*inp_p = inp;
2308			}
2309		}
2310	}
2311	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2312	return (stcb);
2313}
2314
2315/*
2316 * lookup an association by an ASCONF lookup address.
2317 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2318 */
2319struct sctp_tcb *
2320sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2321    struct sockaddr *dst, struct sctphdr *sh,
2322    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2323{
2324	struct sctp_tcb *stcb;
2325	union sctp_sockstore remote_store;
2326	struct sctp_paramhdr parm_buf, *phdr;
2327	int ptype;
2328	int zero_address = 0;
2329
2330#ifdef INET
2331	struct sockaddr_in *sin;
2332
2333#endif
2334#ifdef INET6
2335	struct sockaddr_in6 *sin6;
2336
2337#endif
2338
2339	memset(&remote_store, 0, sizeof(remote_store));
2340	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2341	    &parm_buf, sizeof(struct sctp_paramhdr));
2342	if (phdr == NULL) {
2343		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2344		    __FUNCTION__);
2345		return NULL;
2346	}
2347	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2348	/* get the correlation address */
2349	switch (ptype) {
2350#ifdef INET6
2351	case SCTP_IPV6_ADDRESS:
2352		{
2353			/* ipv6 address param */
2354			struct sctp_ipv6addr_param *p6, p6_buf;
2355
2356			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2357				return NULL;
2358			}
2359			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2360			    offset + sizeof(struct sctp_asconf_chunk),
2361			    &p6_buf.ph, sizeof(*p6));
2362			if (p6 == NULL) {
2363				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2364				    __FUNCTION__);
2365				return (NULL);
2366			}
2367			sin6 = &remote_store.sin6;
2368			sin6->sin6_family = AF_INET6;
2369			sin6->sin6_len = sizeof(*sin6);
2370			sin6->sin6_port = sh->src_port;
2371			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2372			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2373				zero_address = 1;
2374			break;
2375		}
2376#endif
2377#ifdef INET
2378	case SCTP_IPV4_ADDRESS:
2379		{
2380			/* ipv4 address param */
2381			struct sctp_ipv4addr_param *p4, p4_buf;
2382
2383			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2384				return NULL;
2385			}
2386			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2387			    offset + sizeof(struct sctp_asconf_chunk),
2388			    &p4_buf.ph, sizeof(*p4));
2389			if (p4 == NULL) {
2390				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2391				    __FUNCTION__);
2392				return (NULL);
2393			}
2394			sin = &remote_store.sin;
2395			sin->sin_family = AF_INET;
2396			sin->sin_len = sizeof(*sin);
2397			sin->sin_port = sh->src_port;
2398			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2399			if (sin->sin_addr.s_addr == INADDR_ANY)
2400				zero_address = 1;
2401			break;
2402		}
2403#endif
2404	default:
2405		/* invalid address param type */
2406		return NULL;
2407	}
2408
2409	if (zero_address) {
2410		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2411		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2412		if (stcb != NULL) {
2413			SCTP_INP_DECR_REF(*inp_p);
2414		}
2415	} else {
2416		stcb = sctp_findassociation_ep_addr(inp_p,
2417		    &remote_store.sa, netp,
2418		    dst, NULL);
2419	}
2420	return (stcb);
2421}
2422
2423
2424/*
2425 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2426 * addresses. This way if we don't get a bind we by default pick a ephemeral
2427 * port with all addresses bound.
2428 */
2429int
2430sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2431{
2432	/*
2433	 * we get called when a new endpoint starts up. We need to allocate
2434	 * the sctp_inpcb structure from the zone and init it. Mark it as
2435	 * unbound and find a port that we can use as an ephemeral with
2436	 * INADDR_ANY. If the user binds later no problem we can then add in
2437	 * the specific addresses. And setup the default parameters for the
2438	 * EP.
2439	 */
2440	int i, error;
2441	struct sctp_inpcb *inp;
2442	struct sctp_pcb *m;
2443	struct timeval time;
2444	sctp_sharedkey_t *null_key;
2445
2446	error = 0;
2447
2448	SCTP_INP_INFO_WLOCK();
2449	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2450	if (inp == NULL) {
2451		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2452		SCTP_INP_INFO_WUNLOCK();
2453		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2454		return (ENOBUFS);
2455	}
2456	/* zap it */
2457	bzero(inp, sizeof(*inp));
2458
2459	/* bump generations */
2460	/* setup socket pointers */
2461	inp->sctp_socket = so;
2462	inp->ip_inp.inp.inp_socket = so;
2463	inp->ip_inp.inp.inp_cred = crhold(so->so_cred);
2464#ifdef INET6
2465	if (INP_SOCKAF(so) == AF_INET6) {
2466		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2467			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2468		}
2469		if (MODULE_GLOBAL(ip6_v6only)) {
2470			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2471		}
2472	}
2473#endif
2474	inp->sctp_associd_counter = 1;
2475	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2476	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2477	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2478	inp->ecn_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_ecn_enable);
2479	inp->prsctp_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pr_enable);
2480	if (SCTP_BASE_SYSCTL(sctp_auth_disable)) {
2481		inp->auth_supported = 0;
2482	} else {
2483		inp->auth_supported = 1;
2484	}
2485	inp->asconf_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_asconf_enable);
2486	inp->reconfig_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_reconfig_enable);
2487	inp->nrsack_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_nrsack_enable);
2488	inp->pktdrop_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pktdrop_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		crfree(inp->ip_inp.inp.inp_cred);
2493		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2494		SCTP_INP_INFO_WUNLOCK();
2495		return (ENOBUFS);
2496	}
2497#ifdef IPSEC
2498	{
2499		struct inpcbpolicy *pcb_sp = NULL;
2500
2501		error = ipsec_init_policy(so, &pcb_sp);
2502		/* Arrange to share the policy */
2503		inp->ip_inp.inp.inp_sp = pcb_sp;
2504		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2505	}
2506	if (error != 0) {
2507		crfree(inp->ip_inp.inp.inp_cred);
2508		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2509		SCTP_INP_INFO_WUNLOCK();
2510		return error;
2511	}
2512#endif				/* IPSEC */
2513	SCTP_INCR_EP_COUNT();
2514	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2515	SCTP_INP_INFO_WUNLOCK();
2516
2517	so->so_pcb = (caddr_t)inp;
2518
2519	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2520		/* UDP style socket */
2521		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2522		    SCTP_PCB_FLAGS_UNBOUND);
2523		/* Be sure it is NON-BLOCKING IO for UDP */
2524		/* SCTP_SET_SO_NBIO(so); */
2525	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2526		/* TCP style socket */
2527		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2528		    SCTP_PCB_FLAGS_UNBOUND);
2529		/* Be sure we have blocking IO by default */
2530		SCTP_CLEAR_SO_NBIO(so);
2531	} else {
2532		/*
2533		 * unsupported socket type (RAW, etc)- in case we missed it
2534		 * in protosw
2535		 */
2536		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2537		so->so_pcb = NULL;
2538		crfree(inp->ip_inp.inp.inp_cred);
2539		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2540		return (EOPNOTSUPP);
2541	}
2542	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2543		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2544		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2545	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2546		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2547		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2548	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2549		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2550		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2551	}
2552	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2553	    &inp->sctp_hashmark);
2554	if (inp->sctp_tcbhash == NULL) {
2555		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2556		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2557		so->so_pcb = NULL;
2558		crfree(inp->ip_inp.inp.inp_cred);
2559		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2560		return (ENOBUFS);
2561	}
2562	inp->def_vrf_id = vrf_id;
2563
2564	SCTP_INP_INFO_WLOCK();
2565	SCTP_INP_LOCK_INIT(inp);
2566	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2567	SCTP_INP_READ_INIT(inp);
2568	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2569	/* lock the new ep */
2570	SCTP_INP_WLOCK(inp);
2571
2572	/* add it to the info area */
2573	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2574	SCTP_INP_INFO_WUNLOCK();
2575
2576	TAILQ_INIT(&inp->read_queue);
2577	LIST_INIT(&inp->sctp_addr_list);
2578
2579	LIST_INIT(&inp->sctp_asoc_list);
2580
2581#ifdef SCTP_TRACK_FREED_ASOCS
2582	/* TEMP CODE */
2583	LIST_INIT(&inp->sctp_asoc_free_list);
2584#endif
2585	/* Init the timer structure for signature change */
2586	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2587	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2588
2589	/* now init the actual endpoint default data */
2590	m = &inp->sctp_ep;
2591
2592	/* setup the base timeout information */
2593	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2594	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2595	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2596	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2597	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2598	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2599	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2600	/* all max/min max are in ms */
2601	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2602	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2603	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2604	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2605	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2606	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2607	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2608	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2609	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2610	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2611	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2612	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2613	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2614
2615	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2616	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2617	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2618	/* number of streams to pre-open on a association */
2619	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2620
2621	/* Add adaptation cookie */
2622	m->adaptation_layer_indicator = 0;
2623	m->adaptation_layer_indicator_provided = 0;
2624
2625	/* seed random number generator */
2626	m->random_counter = 1;
2627	m->store_at = SCTP_SIGNATURE_SIZE;
2628	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2629	sctp_fill_random_store(m);
2630
2631	/* Minimum cookie size */
2632	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2633	    sizeof(struct sctp_state_cookie);
2634	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2635
2636	/* Setup the initial secret */
2637	(void)SCTP_GETTIME_TIMEVAL(&time);
2638	m->time_of_secret_change = time.tv_sec;
2639
2640	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2641		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2642	}
2643	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2644
2645	/* How long is a cookie good for ? */
2646	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2647	/*
2648	 * Initialize authentication parameters
2649	 */
2650	m->local_hmacs = sctp_default_supported_hmaclist();
2651	m->local_auth_chunks = sctp_alloc_chunklist();
2652	if (inp->asconf_supported) {
2653		sctp_auth_add_chunk(SCTP_ASCONF, m->local_auth_chunks);
2654		sctp_auth_add_chunk(SCTP_ASCONF_ACK, m->local_auth_chunks);
2655	}
2656	m->default_dscp = 0;
2657#ifdef INET6
2658	m->default_flowlabel = 0;
2659#endif
2660	m->port = 0;		/* encapsulation disabled by default */
2661	LIST_INIT(&m->shared_keys);
2662	/* add default NULL key as key id 0 */
2663	null_key = sctp_alloc_sharedkey();
2664	sctp_insert_sharedkey(&m->shared_keys, null_key);
2665	SCTP_INP_WUNLOCK(inp);
2666#ifdef SCTP_LOG_CLOSING
2667	sctp_log_closing(inp, NULL, 12);
2668#endif
2669	return (error);
2670}
2671
2672
2673void
2674sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2675    struct sctp_tcb *stcb)
2676{
2677	struct sctp_nets *net;
2678	uint16_t lport, rport;
2679	struct sctppcbhead *head;
2680	struct sctp_laddr *laddr, *oladdr;
2681
2682	atomic_add_int(&stcb->asoc.refcnt, 1);
2683	SCTP_TCB_UNLOCK(stcb);
2684	SCTP_INP_INFO_WLOCK();
2685	SCTP_INP_WLOCK(old_inp);
2686	SCTP_INP_WLOCK(new_inp);
2687	SCTP_TCB_LOCK(stcb);
2688	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2689
2690	new_inp->sctp_ep.time_of_secret_change =
2691	    old_inp->sctp_ep.time_of_secret_change;
2692	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2693	    sizeof(old_inp->sctp_ep.secret_key));
2694	new_inp->sctp_ep.current_secret_number =
2695	    old_inp->sctp_ep.current_secret_number;
2696	new_inp->sctp_ep.last_secret_number =
2697	    old_inp->sctp_ep.last_secret_number;
2698	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2699
2700	/* make it so new data pours into the new socket */
2701	stcb->sctp_socket = new_inp->sctp_socket;
2702	stcb->sctp_ep = new_inp;
2703
2704	/* Copy the port across */
2705	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2706	rport = stcb->rport;
2707	/* Pull the tcb from the old association */
2708	LIST_REMOVE(stcb, sctp_tcbhash);
2709	LIST_REMOVE(stcb, sctp_tcblist);
2710	if (stcb->asoc.in_asocid_hash) {
2711		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2712	}
2713	/* Now insert the new_inp into the TCP connected hash */
2714	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2715
2716	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2717	/* Its safe to access */
2718	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2719
2720	/* Now move the tcb into the endpoint list */
2721	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2722	/*
2723	 * Question, do we even need to worry about the ep-hash since we
2724	 * only have one connection? Probably not :> so lets get rid of it
2725	 * and not suck up any kernel memory in that.
2726	 */
2727	if (stcb->asoc.in_asocid_hash) {
2728		struct sctpasochead *lhd;
2729
2730		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2731		    new_inp->hashasocidmark)];
2732		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2733	}
2734	/* Ok. Let's restart timer. */
2735	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2736		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2737		    stcb, net);
2738	}
2739
2740	SCTP_INP_INFO_WUNLOCK();
2741	if (new_inp->sctp_tcbhash != NULL) {
2742		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2743		new_inp->sctp_tcbhash = NULL;
2744	}
2745	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2746		/* Subset bound, so copy in the laddr list from the old_inp */
2747		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2748			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2749			if (laddr == NULL) {
2750				/*
2751				 * Gak, what can we do? This assoc is really
2752				 * HOSED. We probably should send an abort
2753				 * here.
2754				 */
2755				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2756				continue;
2757			}
2758			SCTP_INCR_LADDR_COUNT();
2759			bzero(laddr, sizeof(*laddr));
2760			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2761			laddr->ifa = oladdr->ifa;
2762			atomic_add_int(&laddr->ifa->refcount, 1);
2763			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2764			    sctp_nxt_addr);
2765			new_inp->laddr_count++;
2766			if (oladdr == stcb->asoc.last_used_address) {
2767				stcb->asoc.last_used_address = laddr;
2768			}
2769		}
2770	}
2771	/*
2772	 * Now any running timers need to be adjusted since we really don't
2773	 * care if they are running or not just blast in the new_inp into
2774	 * all of them.
2775	 */
2776
2777	stcb->asoc.dack_timer.ep = (void *)new_inp;
2778	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2779	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2780	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2781	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2782	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2783	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2784	/* now what about the nets? */
2785	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2786		net->pmtu_timer.ep = (void *)new_inp;
2787		net->hb_timer.ep = (void *)new_inp;
2788		net->rxt_timer.ep = (void *)new_inp;
2789	}
2790	SCTP_INP_WUNLOCK(new_inp);
2791	SCTP_INP_WUNLOCK(old_inp);
2792}
2793
2794
2795
2796
2797/* sctp_ifap is used to bypass normal local address validation checks */
2798int
2799sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2800    struct sctp_ifa *sctp_ifap, struct thread *p)
2801{
2802	/* bind a ep to a socket address */
2803	struct sctppcbhead *head;
2804	struct sctp_inpcb *inp, *inp_tmp;
2805	struct inpcb *ip_inp;
2806	int port_reuse_active = 0;
2807	int bindall;
2808	uint16_t lport;
2809	int error;
2810	uint32_t vrf_id;
2811
2812	lport = 0;
2813	bindall = 1;
2814	inp = (struct sctp_inpcb *)so->so_pcb;
2815	ip_inp = (struct inpcb *)so->so_pcb;
2816#ifdef SCTP_DEBUG
2817	if (addr) {
2818		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2819		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2820		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2821		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2822	}
2823#endif
2824	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2825		/* already did a bind, subsequent binds NOT allowed ! */
2826		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2827		return (EINVAL);
2828	}
2829#ifdef INVARIANTS
2830	if (p == NULL)
2831		panic("null proc/thread");
2832#endif
2833	if (addr != NULL) {
2834		switch (addr->sa_family) {
2835#ifdef INET
2836		case AF_INET:
2837			{
2838				struct sockaddr_in *sin;
2839
2840				/* IPV6_V6ONLY socket? */
2841				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2842					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2843					return (EINVAL);
2844				}
2845				if (addr->sa_len != sizeof(*sin)) {
2846					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2847					return (EINVAL);
2848				}
2849				sin = (struct sockaddr_in *)addr;
2850				lport = sin->sin_port;
2851				/*
2852				 * For LOOPBACK the prison_local_ip4() call
2853				 * will transmute the ip address to the
2854				 * proper value.
2855				 */
2856				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2857					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2858					return (error);
2859				}
2860				if (sin->sin_addr.s_addr != INADDR_ANY) {
2861					bindall = 0;
2862				}
2863				break;
2864			}
2865#endif
2866#ifdef INET6
2867		case AF_INET6:
2868			{
2869				/*
2870				 * Only for pure IPv6 Address. (No IPv4
2871				 * Mapped!)
2872				 */
2873				struct sockaddr_in6 *sin6;
2874
2875				sin6 = (struct sockaddr_in6 *)addr;
2876
2877				if (addr->sa_len != sizeof(*sin6)) {
2878					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2879					return (EINVAL);
2880				}
2881				lport = sin6->sin6_port;
2882				/*
2883				 * For LOOPBACK the prison_local_ip6() call
2884				 * will transmute the ipv6 address to the
2885				 * proper value.
2886				 */
2887				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2888				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2889					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2890					return (error);
2891				}
2892				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2893					bindall = 0;
2894					/* KAME hack: embed scopeid */
2895					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2896						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2897						return (EINVAL);
2898					}
2899				}
2900				/* this must be cleared for ifa_ifwithaddr() */
2901				sin6->sin6_scope_id = 0;
2902				break;
2903			}
2904#endif
2905		default:
2906			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2907			return (EAFNOSUPPORT);
2908		}
2909	}
2910	SCTP_INP_INFO_WLOCK();
2911	SCTP_INP_WLOCK(inp);
2912	/* Setup a vrf_id to be the default for the non-bind-all case. */
2913	vrf_id = inp->def_vrf_id;
2914
2915	/* increase our count due to the unlock we do */
2916	SCTP_INP_INCR_REF(inp);
2917	if (lport) {
2918		/*
2919		 * Did the caller specify a port? if so we must see if an ep
2920		 * already has this one bound.
2921		 */
2922		/* got to be root to get at low ports */
2923		if (ntohs(lport) < IPPORT_RESERVED) {
2924			if (p && (error =
2925			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2926			    )) {
2927				SCTP_INP_DECR_REF(inp);
2928				SCTP_INP_WUNLOCK(inp);
2929				SCTP_INP_INFO_WUNLOCK();
2930				return (error);
2931			}
2932		}
2933		SCTP_INP_WUNLOCK(inp);
2934		if (bindall) {
2935			vrf_id = inp->def_vrf_id;
2936			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2937			if (inp_tmp != NULL) {
2938				/*
2939				 * lock guy returned and lower count note
2940				 * that we are not bound so inp_tmp should
2941				 * NEVER be inp. And it is this inp
2942				 * (inp_tmp) that gets the reference bump,
2943				 * so we must lower it.
2944				 */
2945				SCTP_INP_DECR_REF(inp_tmp);
2946				/* unlock info */
2947				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2948				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2949					/*
2950					 * Ok, must be one-2-one and
2951					 * allowing port re-use
2952					 */
2953					port_reuse_active = 1;
2954					goto continue_anyway;
2955				}
2956				SCTP_INP_DECR_REF(inp);
2957				SCTP_INP_INFO_WUNLOCK();
2958				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2959				return (EADDRINUSE);
2960			}
2961		} else {
2962			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2963			if (inp_tmp != NULL) {
2964				/*
2965				 * lock guy returned and lower count note
2966				 * that we are not bound so inp_tmp should
2967				 * NEVER be inp. And it is this inp
2968				 * (inp_tmp) that gets the reference bump,
2969				 * so we must lower it.
2970				 */
2971				SCTP_INP_DECR_REF(inp_tmp);
2972				/* unlock info */
2973				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2974				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2975					/*
2976					 * Ok, must be one-2-one and
2977					 * allowing port re-use
2978					 */
2979					port_reuse_active = 1;
2980					goto continue_anyway;
2981				}
2982				SCTP_INP_DECR_REF(inp);
2983				SCTP_INP_INFO_WUNLOCK();
2984				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2985				return (EADDRINUSE);
2986			}
2987		}
2988continue_anyway:
2989		SCTP_INP_WLOCK(inp);
2990		if (bindall) {
2991			/* verify that no lport is not used by a singleton */
2992			if ((port_reuse_active == 0) &&
2993			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
2994				/* Sorry someone already has this one bound */
2995				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2996				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2997					port_reuse_active = 1;
2998				} else {
2999					SCTP_INP_DECR_REF(inp);
3000					SCTP_INP_WUNLOCK(inp);
3001					SCTP_INP_INFO_WUNLOCK();
3002					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3003					return (EADDRINUSE);
3004				}
3005			}
3006		}
3007	} else {
3008		uint16_t first, last, candidate;
3009		uint16_t count;
3010		int done;
3011
3012		if (ip_inp->inp_flags & INP_HIGHPORT) {
3013			first = MODULE_GLOBAL(ipport_hifirstauto);
3014			last = MODULE_GLOBAL(ipport_hilastauto);
3015		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3016			if (p && (error =
3017			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3018			    )) {
3019				SCTP_INP_DECR_REF(inp);
3020				SCTP_INP_WUNLOCK(inp);
3021				SCTP_INP_INFO_WUNLOCK();
3022				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3023				return (error);
3024			}
3025			first = MODULE_GLOBAL(ipport_lowfirstauto);
3026			last = MODULE_GLOBAL(ipport_lowlastauto);
3027		} else {
3028			first = MODULE_GLOBAL(ipport_firstauto);
3029			last = MODULE_GLOBAL(ipport_lastauto);
3030		}
3031		if (first > last) {
3032			uint16_t temp;
3033
3034			temp = first;
3035			first = last;
3036			last = temp;
3037		}
3038		count = last - first + 1;	/* number of candidates */
3039		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3040
3041		done = 0;
3042		while (!done) {
3043			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3044				done = 1;
3045			}
3046			if (!done) {
3047				if (--count == 0) {
3048					SCTP_INP_DECR_REF(inp);
3049					SCTP_INP_WUNLOCK(inp);
3050					SCTP_INP_INFO_WUNLOCK();
3051					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3052					return (EADDRINUSE);
3053				}
3054				if (candidate == last)
3055					candidate = first;
3056				else
3057					candidate = candidate + 1;
3058			}
3059		}
3060		lport = htons(candidate);
3061	}
3062	SCTP_INP_DECR_REF(inp);
3063	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3064	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3065		/*
3066		 * this really should not happen. The guy did a non-blocking
3067		 * bind and then did a close at the same time.
3068		 */
3069		SCTP_INP_WUNLOCK(inp);
3070		SCTP_INP_INFO_WUNLOCK();
3071		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3072		return (EINVAL);
3073	}
3074	/* ok we look clear to give out this port, so lets setup the binding */
3075	if (bindall) {
3076		/* binding to all addresses, so just set in the proper flags */
3077		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3078		/* set the automatic addr changes from kernel flag */
3079		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3080			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3081			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3082		} else {
3083			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3084			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3085		}
3086		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3087			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3088		} else {
3089			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3090		}
3091		/*
3092		 * set the automatic mobility_base from kernel flag (by
3093		 * micchie)
3094		 */
3095		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3096			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3097			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3098		} else {
3099			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3100			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3101		}
3102		/*
3103		 * set the automatic mobility_fasthandoff from kernel flag
3104		 * (by micchie)
3105		 */
3106		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3107			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3108			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3109		} else {
3110			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3111			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3112		}
3113	} else {
3114		/*
3115		 * bind specific, make sure flags is off and add a new
3116		 * address structure to the sctp_addr_list inside the ep
3117		 * structure.
3118		 *
3119		 * We will need to allocate one and insert it at the head. The
3120		 * socketopt call can just insert new addresses in there as
3121		 * well. It will also have to do the embed scope kame hack
3122		 * too (before adding).
3123		 */
3124		struct sctp_ifa *ifa;
3125		union sctp_sockstore store;
3126
3127		memset(&store, 0, sizeof(store));
3128		switch (addr->sa_family) {
3129#ifdef INET
3130		case AF_INET:
3131			memcpy(&store.sin, addr, sizeof(struct sockaddr_in));
3132			store.sin.sin_port = 0;
3133			break;
3134#endif
3135#ifdef INET6
3136		case AF_INET6:
3137			memcpy(&store.sin6, addr, sizeof(struct sockaddr_in6));
3138			store.sin6.sin6_port = 0;
3139			break;
3140#endif
3141		default:
3142			break;
3143		}
3144		/*
3145		 * first find the interface with the bound address need to
3146		 * zero out the port to find the address! yuck! can't do
3147		 * this earlier since need port for sctp_pcb_findep()
3148		 */
3149		if (sctp_ifap != NULL) {
3150			ifa = sctp_ifap;
3151		} else {
3152			/*
3153			 * Note for BSD we hit here always other O/S's will
3154			 * pass things in via the sctp_ifap argument
3155			 * (Panda).
3156			 */
3157			ifa = sctp_find_ifa_by_addr(&store.sa,
3158			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3159		}
3160		if (ifa == NULL) {
3161			/* Can't find an interface with that address */
3162			SCTP_INP_WUNLOCK(inp);
3163			SCTP_INP_INFO_WUNLOCK();
3164			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3165			return (EADDRNOTAVAIL);
3166		}
3167#ifdef INET6
3168		if (addr->sa_family == AF_INET6) {
3169			/* GAK, more FIXME IFA lock? */
3170			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3171				/* Can't bind a non-existent addr. */
3172				SCTP_INP_WUNLOCK(inp);
3173				SCTP_INP_INFO_WUNLOCK();
3174				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3175				return (EINVAL);
3176			}
3177		}
3178#endif
3179		/* we're not bound all */
3180		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3181		/* allow bindx() to send ASCONF's for binding changes */
3182		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3183		/* clear automatic addr changes from kernel flag */
3184		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3185
3186		/* add this address to the endpoint list */
3187		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3188		if (error != 0) {
3189			SCTP_INP_WUNLOCK(inp);
3190			SCTP_INP_INFO_WUNLOCK();
3191			return (error);
3192		}
3193		inp->laddr_count++;
3194	}
3195	/* find the bucket */
3196	if (port_reuse_active) {
3197		/* Put it into tcp 1-2-1 hash */
3198		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3199		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3200	} else {
3201		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3202	}
3203	/* put it in the bucket */
3204	LIST_INSERT_HEAD(head, inp, sctp_hash);
3205	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3206	    (void *)head, ntohs(lport), port_reuse_active);
3207	/* set in the port */
3208	inp->sctp_lport = lport;
3209
3210	/* turn off just the unbound flag */
3211	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3212	SCTP_INP_WUNLOCK(inp);
3213	SCTP_INP_INFO_WUNLOCK();
3214	return (0);
3215}
3216
3217
3218static void
3219sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3220{
3221	struct sctp_iterator *it, *nit;
3222
3223	/*
3224	 * We enter with the only the ITERATOR_LOCK in place and a write
3225	 * lock on the inp_info stuff.
3226	 */
3227	it = sctp_it_ctl.cur_it;
3228	if (it && (it->vn != curvnet)) {
3229		/* Its not looking at our VNET */
3230		return;
3231	}
3232	if (it && (it->inp == inp)) {
3233		/*
3234		 * This is tricky and we hold the iterator lock, but when it
3235		 * returns and gets the lock (when we release it) the
3236		 * iterator will try to operate on inp. We need to stop that
3237		 * from happening. But of course the iterator has a
3238		 * reference on the stcb and inp. We can mark it and it will
3239		 * stop.
3240		 *
3241		 * If its a single iterator situation, we set the end iterator
3242		 * flag. Otherwise we set the iterator to go to the next
3243		 * inp.
3244		 *
3245		 */
3246		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3247			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3248		} else {
3249			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3250		}
3251	}
3252	/*
3253	 * Now go through and remove any single reference to our inp that
3254	 * may be still pending on the list
3255	 */
3256	SCTP_IPI_ITERATOR_WQ_LOCK();
3257	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3258		if (it->vn != curvnet) {
3259			continue;
3260		}
3261		if (it->inp == inp) {
3262			/* This one points to me is it inp specific? */
3263			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3264				/* Remove and free this one */
3265				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3266				    it, sctp_nxt_itr);
3267				if (it->function_atend != NULL) {
3268					(*it->function_atend) (it->pointer, it->val);
3269				}
3270				SCTP_FREE(it, SCTP_M_ITER);
3271			} else {
3272				it->inp = LIST_NEXT(it->inp, sctp_list);
3273				if (it->inp) {
3274					SCTP_INP_INCR_REF(it->inp);
3275				}
3276			}
3277			/*
3278			 * When its put in the refcnt is incremented so decr
3279			 * it
3280			 */
3281			SCTP_INP_DECR_REF(inp);
3282		}
3283	}
3284	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3285}
3286
3287/* release sctp_inpcb unbind the port */
3288void
3289sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3290{
3291	/*
3292	 * Here we free a endpoint. We must find it (if it is in the Hash
3293	 * table) and remove it from there. Then we must also find it in the
3294	 * overall list and remove it from there. After all removals are
3295	 * complete then any timer has to be stopped. Then start the actual
3296	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3297	 * all associations. d) finally the ep itself.
3298	 */
3299	struct sctp_tcb *asoc, *nasoc;
3300	struct sctp_laddr *laddr, *nladdr;
3301	struct inpcb *ip_pcb;
3302	struct socket *so;
3303	int being_refed = 0;
3304	struct sctp_queued_to_read *sq, *nsq;
3305	int cnt;
3306	sctp_sharedkey_t *shared_key, *nshared_key;
3307
3308
3309#ifdef SCTP_LOG_CLOSING
3310	sctp_log_closing(inp, NULL, 0);
3311#endif
3312	SCTP_ITERATOR_LOCK();
3313	/* mark any iterators on the list or being processed */
3314	sctp_iterator_inp_being_freed(inp);
3315	SCTP_ITERATOR_UNLOCK();
3316	so = inp->sctp_socket;
3317	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3318		/* been here before.. eeks.. get out of here */
3319		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3320#ifdef SCTP_LOG_CLOSING
3321		sctp_log_closing(inp, NULL, 1);
3322#endif
3323		return;
3324	}
3325	SCTP_ASOC_CREATE_LOCK(inp);
3326	SCTP_INP_INFO_WLOCK();
3327
3328	SCTP_INP_WLOCK(inp);
3329	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3330		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3331		/* socket is gone, so no more wakeups allowed */
3332		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3333		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3334		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3335
3336	}
3337	/* First time through we have the socket lock, after that no more. */
3338	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3339	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3340
3341	if (inp->control) {
3342		sctp_m_freem(inp->control);
3343		inp->control = NULL;
3344	}
3345	if (inp->pkt) {
3346		sctp_m_freem(inp->pkt);
3347		inp->pkt = NULL;
3348	}
3349	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3350					 * here but I will be nice :> (i.e.
3351					 * ip_pcb = ep;) */
3352	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3353		int cnt_in_sd;
3354
3355		cnt_in_sd = 0;
3356		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3357			SCTP_TCB_LOCK(asoc);
3358			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3359				/* Skip guys being freed */
3360				cnt_in_sd++;
3361				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3362					/*
3363					 * Special case - we did not start a
3364					 * kill timer on the asoc due to it
3365					 * was not closed. So go ahead and
3366					 * start it now.
3367					 */
3368					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3369					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3370				}
3371				SCTP_TCB_UNLOCK(asoc);
3372				continue;
3373			}
3374			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3375			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3376			    (asoc->asoc.total_output_queue_size == 0)) {
3377				/*
3378				 * If we have data in queue, we don't want
3379				 * to just free since the app may have done,
3380				 * send()/close or connect/send/close. And
3381				 * it wants the data to get across first.
3382				 */
3383				/* Just abandon things in the front states */
3384				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3385				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3386					cnt_in_sd++;
3387				}
3388				continue;
3389			}
3390			/* Disconnect the socket please */
3391			asoc->sctp_socket = NULL;
3392			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3393			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3394			    (asoc->asoc.control_pdapi) ||
3395			    (asoc->asoc.size_on_all_streams > 0) ||
3396			    (so && (so->so_rcv.sb_cc > 0))) {
3397				/* Left with Data unread */
3398				struct mbuf *op_err;
3399
3400				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3401				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3402				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3403				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3404				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3405				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3406					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3407				}
3408				if (sctp_free_assoc(inp, asoc,
3409				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3410					cnt_in_sd++;
3411				}
3412				continue;
3413			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3414				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3415			    (asoc->asoc.stream_queue_cnt == 0)) {
3416				if (asoc->asoc.locked_on_sending) {
3417					goto abort_anyway;
3418				}
3419				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3420				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3421					struct sctp_nets *netp;
3422
3423					/*
3424					 * there is nothing queued to send,
3425					 * so I send shutdown
3426					 */
3427					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3428					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3429						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3430					}
3431					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3432					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3433					sctp_stop_timers_for_shutdown(asoc);
3434					if (asoc->asoc.alternate) {
3435						netp = asoc->asoc.alternate;
3436					} else {
3437						netp = asoc->asoc.primary_destination;
3438					}
3439					sctp_send_shutdown(asoc, netp);
3440					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3441					    netp);
3442					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3443					    asoc->asoc.primary_destination);
3444					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3445				}
3446			} else {
3447				/* mark into shutdown pending */
3448				struct sctp_stream_queue_pending *sp;
3449
3450				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3451				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3452				    asoc->asoc.primary_destination);
3453				if (asoc->asoc.locked_on_sending) {
3454					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3455					    sctp_streamhead);
3456					if (sp == NULL) {
3457						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3458						    (void *)asoc->asoc.locked_on_sending,
3459						    asoc->asoc.locked_on_sending->stream_no);
3460					} else {
3461						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3462							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3463					}
3464				}
3465				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3466				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3467				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3468					struct mbuf *op_err;
3469
3470			abort_anyway:
3471					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3472					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3473					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3474					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3475					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3476					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3477						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3478					}
3479					if (sctp_free_assoc(inp, asoc,
3480					    SCTP_PCBFREE_NOFORCE,
3481					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3482						cnt_in_sd++;
3483					}
3484					continue;
3485				} else {
3486					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3487				}
3488			}
3489			cnt_in_sd++;
3490			SCTP_TCB_UNLOCK(asoc);
3491		}
3492		/* now is there some left in our SHUTDOWN state? */
3493		if (cnt_in_sd) {
3494#ifdef SCTP_LOG_CLOSING
3495			sctp_log_closing(inp, NULL, 2);
3496#endif
3497			inp->sctp_socket = NULL;
3498			SCTP_INP_WUNLOCK(inp);
3499			SCTP_ASOC_CREATE_UNLOCK(inp);
3500			SCTP_INP_INFO_WUNLOCK();
3501			return;
3502		}
3503	}
3504	inp->sctp_socket = NULL;
3505	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3506	    SCTP_PCB_FLAGS_UNBOUND) {
3507		/*
3508		 * ok, this guy has been bound. It's port is somewhere in
3509		 * the SCTP_BASE_INFO(hash table). Remove it!
3510		 */
3511		LIST_REMOVE(inp, sctp_hash);
3512		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3513	}
3514	/*
3515	 * If there is a timer running to kill us, forget it, since it may
3516	 * have a contest on the INP lock.. which would cause us to die ...
3517	 */
3518	cnt = 0;
3519	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3520		SCTP_TCB_LOCK(asoc);
3521		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3522			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3523				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3524				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3525			}
3526			cnt++;
3527			SCTP_TCB_UNLOCK(asoc);
3528			continue;
3529		}
3530		/* Free associations that are NOT killing us */
3531		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3532		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3533			struct mbuf *op_err;
3534
3535			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3536			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3537			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3538			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3539		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3540			cnt++;
3541			SCTP_TCB_UNLOCK(asoc);
3542			continue;
3543		}
3544		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3545		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3546			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3547		}
3548		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3549			cnt++;
3550		}
3551	}
3552	if (cnt) {
3553		/* Ok we have someone out there that will kill us */
3554		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3555#ifdef SCTP_LOG_CLOSING
3556		sctp_log_closing(inp, NULL, 3);
3557#endif
3558		SCTP_INP_WUNLOCK(inp);
3559		SCTP_ASOC_CREATE_UNLOCK(inp);
3560		SCTP_INP_INFO_WUNLOCK();
3561		return;
3562	}
3563	if (SCTP_INP_LOCK_CONTENDED(inp))
3564		being_refed++;
3565	if (SCTP_INP_READ_CONTENDED(inp))
3566		being_refed++;
3567	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3568		being_refed++;
3569
3570	if ((inp->refcount) ||
3571	    (being_refed) ||
3572	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3573		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3574#ifdef SCTP_LOG_CLOSING
3575		sctp_log_closing(inp, NULL, 4);
3576#endif
3577		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3578		SCTP_INP_WUNLOCK(inp);
3579		SCTP_ASOC_CREATE_UNLOCK(inp);
3580		SCTP_INP_INFO_WUNLOCK();
3581		return;
3582	}
3583	inp->sctp_ep.signature_change.type = 0;
3584	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3585	/*
3586	 * Remove it from the list .. last thing we need a lock for.
3587	 */
3588	LIST_REMOVE(inp, sctp_list);
3589	SCTP_INP_WUNLOCK(inp);
3590	SCTP_ASOC_CREATE_UNLOCK(inp);
3591	SCTP_INP_INFO_WUNLOCK();
3592	/*
3593	 * Now we release all locks. Since this INP cannot be found anymore
3594	 * except possibly by the kill timer that might be running. We call
3595	 * the drain function here. It should hit the case were it sees the
3596	 * ACTIVE flag cleared and exit out freeing us to proceed and
3597	 * destroy everything.
3598	 */
3599	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3600		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3601	} else {
3602		/* Probably un-needed */
3603		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3604	}
3605
3606#ifdef SCTP_LOG_CLOSING
3607	sctp_log_closing(inp, NULL, 5);
3608#endif
3609
3610
3611	if ((inp->sctp_asocidhash) != NULL) {
3612		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3613		inp->sctp_asocidhash = NULL;
3614	}
3615	/* sa_ignore FREED_MEMORY */
3616	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3617		/* Its only abandoned if it had data left */
3618		if (sq->length)
3619			SCTP_STAT_INCR(sctps_left_abandon);
3620
3621		TAILQ_REMOVE(&inp->read_queue, sq, next);
3622		sctp_free_remote_addr(sq->whoFrom);
3623		if (so)
3624			so->so_rcv.sb_cc -= sq->length;
3625		if (sq->data) {
3626			sctp_m_freem(sq->data);
3627			sq->data = NULL;
3628		}
3629		/*
3630		 * no need to free the net count, since at this point all
3631		 * assoc's are gone.
3632		 */
3633		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3634		SCTP_DECR_READQ_COUNT();
3635	}
3636	/* Now the sctp_pcb things */
3637	/*
3638	 * free each asoc if it is not already closed/free. we can't use the
3639	 * macro here since le_next will get freed as part of the
3640	 * sctp_free_assoc() call.
3641	 */
3642	if (so) {
3643#ifdef IPSEC
3644		ipsec_delete_pcbpolicy(ip_pcb);
3645#endif				/* IPSEC */
3646
3647		/* Unlocks not needed since the socket is gone now */
3648	}
3649	if (ip_pcb->inp_options) {
3650		(void)sctp_m_free(ip_pcb->inp_options);
3651		ip_pcb->inp_options = 0;
3652	}
3653#ifdef INET6
3654	if (ip_pcb->inp_vflag & INP_IPV6) {
3655		struct in6pcb *in6p;
3656
3657		in6p = (struct in6pcb *)inp;
3658		ip6_freepcbopts(in6p->in6p_outputopts);
3659	}
3660#endif				/* INET6 */
3661	ip_pcb->inp_vflag = 0;
3662	/* free up authentication fields */
3663	if (inp->sctp_ep.local_auth_chunks != NULL)
3664		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3665	if (inp->sctp_ep.local_hmacs != NULL)
3666		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3667
3668	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3669		LIST_REMOVE(shared_key, next);
3670		sctp_free_sharedkey(shared_key);
3671		/* sa_ignore FREED_MEMORY */
3672	}
3673
3674	/*
3675	 * if we have an address list the following will free the list of
3676	 * ifaddr's that are set into this ep. Again macro limitations here,
3677	 * since the LIST_FOREACH could be a bad idea.
3678	 */
3679	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3680		sctp_remove_laddr(laddr);
3681	}
3682
3683#ifdef SCTP_TRACK_FREED_ASOCS
3684	/* TEMP CODE */
3685	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3686		LIST_REMOVE(asoc, sctp_tcblist);
3687		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3688		SCTP_DECR_ASOC_COUNT();
3689	}
3690	/* *** END TEMP CODE *** */
3691#endif
3692	/* Now lets see about freeing the EP hash table. */
3693	if (inp->sctp_tcbhash != NULL) {
3694		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3695		inp->sctp_tcbhash = NULL;
3696	}
3697	/* Now we must put the ep memory back into the zone pool */
3698	crfree(inp->ip_inp.inp.inp_cred);
3699	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3700	SCTP_INP_LOCK_DESTROY(inp);
3701	SCTP_INP_READ_DESTROY(inp);
3702	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3703	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3704	SCTP_DECR_EP_COUNT();
3705}
3706
3707
3708struct sctp_nets *
3709sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3710{
3711	struct sctp_nets *net;
3712
3713	/* locate the address */
3714	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3715		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3716			return (net);
3717	}
3718	return (NULL);
3719}
3720
3721
3722int
3723sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3724{
3725	struct sctp_ifa *sctp_ifa;
3726
3727	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3728	if (sctp_ifa) {
3729		return (1);
3730	} else {
3731		return (0);
3732	}
3733}
3734
3735/*
3736 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3737 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3738 * stats of stuff.
3739 */
3740int
3741sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3742    struct sctp_nets **netp, int set_scope, int from)
3743{
3744	/*
3745	 * The following is redundant to the same lines in the
3746	 * sctp_aloc_assoc() but is needed since others call the add address
3747	 * function
3748	 */
3749	struct sctp_nets *net, *netfirst;
3750	int addr_inscope;
3751
3752	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3753	    from);
3754	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3755
3756	netfirst = sctp_findnet(stcb, newaddr);
3757	if (netfirst) {
3758		/*
3759		 * Lie and return ok, we don't want to make the association
3760		 * go away for this behavior. It will happen in the TCP
3761		 * model in a connected socket. It does not reach the hash
3762		 * table until after the association is built so it can't be
3763		 * found. Mark as reachable, since the initial creation will
3764		 * have been cleared and the NOT_IN_ASSOC flag will have
3765		 * been added... and we don't want to end up removing it
3766		 * back out.
3767		 */
3768		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3769			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3770			    SCTP_ADDR_UNCONFIRMED);
3771		} else {
3772			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3773		}
3774
3775		return (0);
3776	}
3777	addr_inscope = 1;
3778	switch (newaddr->sa_family) {
3779#ifdef INET
3780	case AF_INET:
3781		{
3782			struct sockaddr_in *sin;
3783
3784			sin = (struct sockaddr_in *)newaddr;
3785			if (sin->sin_addr.s_addr == 0) {
3786				/* Invalid address */
3787				return (-1);
3788			}
3789			/* zero out the bzero area */
3790			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3791
3792			/* assure len is set */
3793			sin->sin_len = sizeof(struct sockaddr_in);
3794			if (set_scope) {
3795#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3796				stcb->asoc.scope.ipv4_local_scope = 1;
3797#else
3798				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3799					stcb->asoc.scope.ipv4_local_scope = 1;
3800				}
3801#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3802			} else {
3803				/* Validate the address is in scope */
3804				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3805				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3806					addr_inscope = 0;
3807				}
3808			}
3809			break;
3810		}
3811#endif
3812#ifdef INET6
3813	case AF_INET6:
3814		{
3815			struct sockaddr_in6 *sin6;
3816
3817			sin6 = (struct sockaddr_in6 *)newaddr;
3818			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3819				/* Invalid address */
3820				return (-1);
3821			}
3822			/* assure len is set */
3823			sin6->sin6_len = sizeof(struct sockaddr_in6);
3824			if (set_scope) {
3825				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3826					stcb->asoc.scope.loopback_scope = 1;
3827					stcb->asoc.scope.local_scope = 0;
3828					stcb->asoc.scope.ipv4_local_scope = 1;
3829					stcb->asoc.scope.site_scope = 1;
3830				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3831					/*
3832					 * If the new destination is a
3833					 * LINK_LOCAL we must have common
3834					 * site scope. Don't set the local
3835					 * scope since we may not share all
3836					 * links, only loopback can do this.
3837					 * Links on the local network would
3838					 * also be on our private network
3839					 * for v4 too.
3840					 */
3841					stcb->asoc.scope.ipv4_local_scope = 1;
3842					stcb->asoc.scope.site_scope = 1;
3843				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3844					/*
3845					 * If the new destination is
3846					 * SITE_LOCAL then we must have site
3847					 * scope in common.
3848					 */
3849					stcb->asoc.scope.site_scope = 1;
3850				}
3851			} else {
3852				/* Validate the address is in scope */
3853				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3854				    (stcb->asoc.scope.loopback_scope == 0)) {
3855					addr_inscope = 0;
3856				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3857				    (stcb->asoc.scope.local_scope == 0)) {
3858					addr_inscope = 0;
3859				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3860				    (stcb->asoc.scope.site_scope == 0)) {
3861					addr_inscope = 0;
3862				}
3863			}
3864			break;
3865		}
3866#endif
3867	default:
3868		/* not supported family type */
3869		return (-1);
3870	}
3871	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3872	if (net == NULL) {
3873		return (-1);
3874	}
3875	SCTP_INCR_RADDR_COUNT();
3876	bzero(net, sizeof(struct sctp_nets));
3877	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3878	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3879	switch (newaddr->sa_family) {
3880#ifdef INET
3881	case AF_INET:
3882		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3883		break;
3884#endif
3885#ifdef INET6
3886	case AF_INET6:
3887		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3888		break;
3889#endif
3890	default:
3891		break;
3892	}
3893	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3894	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3895		stcb->asoc.scope.loopback_scope = 1;
3896		stcb->asoc.scope.ipv4_local_scope = 1;
3897		stcb->asoc.scope.local_scope = 0;
3898		stcb->asoc.scope.site_scope = 1;
3899		addr_inscope = 1;
3900	}
3901	net->failure_threshold = stcb->asoc.def_net_failure;
3902	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3903	if (addr_inscope == 0) {
3904		net->dest_state = (SCTP_ADDR_REACHABLE |
3905		    SCTP_ADDR_OUT_OF_SCOPE);
3906	} else {
3907		if (from == SCTP_ADDR_IS_CONFIRMED)
3908			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3909			net->dest_state = SCTP_ADDR_REACHABLE;
3910		else
3911			net->dest_state = SCTP_ADDR_REACHABLE |
3912			    SCTP_ADDR_UNCONFIRMED;
3913	}
3914	/*
3915	 * We set this to 0, the timer code knows that this means its an
3916	 * initial value
3917	 */
3918	net->rto_needed = 1;
3919	net->RTO = 0;
3920	net->RTO_measured = 0;
3921	stcb->asoc.numnets++;
3922	net->ref_count = 1;
3923	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3924	net->port = stcb->asoc.port;
3925	net->dscp = stcb->asoc.default_dscp;
3926#ifdef INET6
3927	net->flowlabel = stcb->asoc.default_flowlabel;
3928#endif
3929	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3930		net->dest_state |= SCTP_ADDR_NOHB;
3931	} else {
3932		net->dest_state &= ~SCTP_ADDR_NOHB;
3933	}
3934	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3935		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3936	} else {
3937		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3938	}
3939	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3940	/* Init the timer structure */
3941	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3942	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3943	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3944
3945	/* Now generate a route for this guy */
3946#ifdef INET6
3947	/* KAME hack: embed scopeid */
3948	if (newaddr->sa_family == AF_INET6) {
3949		struct sockaddr_in6 *sin6;
3950
3951		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3952		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3953		sin6->sin6_scope_id = 0;
3954	}
3955#endif
3956	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3957
3958	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3959		/* Get source address */
3960		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3961		    stcb,
3962		    (sctp_route_t *) & net->ro,
3963		    net,
3964		    0,
3965		    stcb->asoc.vrf_id);
3966		if (net->ro._s_addr != NULL) {
3967			net->src_addr_selected = 1;
3968			/* Now get the interface MTU */
3969			if (net->ro._s_addr->ifn_p != NULL) {
3970				net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3971			}
3972		} else {
3973			net->src_addr_selected = 0;
3974		}
3975		if (net->mtu > 0) {
3976			uint32_t rmtu;
3977
3978			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3979			if (rmtu == 0) {
3980				/*
3981				 * Start things off to match mtu of
3982				 * interface please.
3983				 */
3984				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3985				    net->ro.ro_rt, net->mtu);
3986			} else {
3987				/*
3988				 * we take the route mtu over the interface,
3989				 * since the route may be leading out the
3990				 * loopback, or a different interface.
3991				 */
3992				net->mtu = rmtu;
3993			}
3994		}
3995	} else {
3996		net->src_addr_selected = 0;
3997	}
3998	if (net->mtu == 0) {
3999		switch (newaddr->sa_family) {
4000#ifdef INET
4001		case AF_INET:
4002			net->mtu = SCTP_DEFAULT_MTU;
4003			break;
4004#endif
4005#ifdef INET6
4006		case AF_INET6:
4007			net->mtu = 1280;
4008			break;
4009#endif
4010		default:
4011			break;
4012		}
4013	}
4014#if defined(INET) || defined(INET6)
4015	if (net->port) {
4016		net->mtu -= (uint32_t) sizeof(struct udphdr);
4017	}
4018#endif
4019	if (from == SCTP_ALLOC_ASOC) {
4020		stcb->asoc.smallest_mtu = net->mtu;
4021	}
4022	if (stcb->asoc.smallest_mtu > net->mtu) {
4023		stcb->asoc.smallest_mtu = net->mtu;
4024	}
4025#ifdef INET6
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_recoverscope(sin6);
4031	}
4032#endif
4033
4034	/* JRS - Use the congestion control given in the CC module */
4035	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4036		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4037
4038	/*
4039	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4040	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4041	 */
4042	net->find_pseudo_cumack = 1;
4043	net->find_rtx_pseudo_cumack = 1;
4044	/* Choose an initial flowid. */
4045	net->flowid = stcb->asoc.my_vtag ^
4046	    ntohs(stcb->rport) ^
4047	    ntohs(stcb->sctp_ep->sctp_lport);
4048	net->flowtype = M_HASHTYPE_OPAQUE;
4049	if (netp) {
4050		*netp = net;
4051	}
4052	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4053	if (net->ro.ro_rt == NULL) {
4054		/* Since we have no route put it at the back */
4055		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4056	} else if (netfirst == NULL) {
4057		/* We are the first one in the pool. */
4058		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4059	} else if (netfirst->ro.ro_rt == NULL) {
4060		/*
4061		 * First one has NO route. Place this one ahead of the first
4062		 * one.
4063		 */
4064		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4065	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4066		/*
4067		 * This one has a different interface than the one at the
4068		 * top of the list. Place it ahead.
4069		 */
4070		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4071	} else {
4072		/*
4073		 * Ok we have the same interface as the first one. Move
4074		 * forward until we find either a) one with a NULL route...
4075		 * insert ahead of that b) one with a different ifp.. insert
4076		 * after that. c) end of the list.. insert at the tail.
4077		 */
4078		struct sctp_nets *netlook;
4079
4080		do {
4081			netlook = TAILQ_NEXT(netfirst, sctp_next);
4082			if (netlook == NULL) {
4083				/* End of the list */
4084				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4085				break;
4086			} else if (netlook->ro.ro_rt == NULL) {
4087				/* next one has NO route */
4088				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4089				break;
4090			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4091				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4092				    net, sctp_next);
4093				break;
4094			}
4095			/* Shift forward */
4096			netfirst = netlook;
4097		} while (netlook != NULL);
4098	}
4099
4100	/* got to have a primary set */
4101	if (stcb->asoc.primary_destination == 0) {
4102		stcb->asoc.primary_destination = net;
4103	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4104		    (net->ro.ro_rt) &&
4105	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4106		/* No route to current primary adopt new primary */
4107		stcb->asoc.primary_destination = net;
4108	}
4109	/* Validate primary is first */
4110	net = TAILQ_FIRST(&stcb->asoc.nets);
4111	if ((net != stcb->asoc.primary_destination) &&
4112	    (stcb->asoc.primary_destination)) {
4113		/*
4114		 * first one on the list is NOT the primary sctp_cmpaddr()
4115		 * is much more efficient if the primary is the first on the
4116		 * list, make it so.
4117		 */
4118		TAILQ_REMOVE(&stcb->asoc.nets,
4119		    stcb->asoc.primary_destination, sctp_next);
4120		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4121		    stcb->asoc.primary_destination, sctp_next);
4122	}
4123	return (0);
4124}
4125
4126
4127static uint32_t
4128sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4129{
4130	uint32_t id;
4131	struct sctpasochead *head;
4132	struct sctp_tcb *lstcb;
4133
4134	SCTP_INP_WLOCK(inp);
4135try_again:
4136	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4137		/* TSNH */
4138		SCTP_INP_WUNLOCK(inp);
4139		return (0);
4140	}
4141	/*
4142	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4143	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4144	 */
4145	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4146		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4147	}
4148	id = inp->sctp_associd_counter;
4149	inp->sctp_associd_counter++;
4150	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4151	if (lstcb) {
4152		goto try_again;
4153	}
4154	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4155	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4156	stcb->asoc.in_asocid_hash = 1;
4157	SCTP_INP_WUNLOCK(inp);
4158	return id;
4159}
4160
4161/*
4162 * allocate an association and add it to the endpoint. The caller must be
4163 * careful to add all additional addresses once they are know right away or
4164 * else the assoc will be may experience a blackout scenario.
4165 */
4166struct sctp_tcb *
4167sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4168    int *error, uint32_t override_tag, uint32_t vrf_id,
4169    struct thread *p
4170)
4171{
4172	/* note the p argument is only valid in unbound sockets */
4173
4174	struct sctp_tcb *stcb;
4175	struct sctp_association *asoc;
4176	struct sctpasochead *head;
4177	uint16_t rport;
4178	int err;
4179
4180	/*
4181	 * Assumption made here: Caller has done a
4182	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4183	 * address does not exist already.
4184	 */
4185	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4186		/* Hit max assoc, sorry no more */
4187		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4188		*error = ENOBUFS;
4189		return (NULL);
4190	}
4191	if (firstaddr == NULL) {
4192		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4193		*error = EINVAL;
4194		return (NULL);
4195	}
4196	SCTP_INP_RLOCK(inp);
4197	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4198	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4199	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4200		/*
4201		 * If its in the TCP pool, its NOT allowed to create an
4202		 * association. The parent listener needs to call
4203		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4204		 * off, or connected one does this.. its an error.
4205		 */
4206		SCTP_INP_RUNLOCK(inp);
4207		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4208		*error = EINVAL;
4209		return (NULL);
4210	}
4211	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4212	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4213		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4214		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4215			SCTP_INP_RUNLOCK(inp);
4216			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4217			*error = EINVAL;
4218			return (NULL);
4219		}
4220	}
4221	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4222#ifdef SCTP_DEBUG
4223	if (firstaddr) {
4224		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4225		switch (firstaddr->sa_family) {
4226#ifdef INET
4227		case AF_INET:
4228			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4229			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4230			break;
4231#endif
4232#ifdef INET6
4233		case AF_INET6:
4234			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4235			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4236			break;
4237#endif
4238		default:
4239			break;
4240		}
4241	} else {
4242		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4243	}
4244#endif				/* SCTP_DEBUG */
4245	switch (firstaddr->sa_family) {
4246#ifdef INET
4247	case AF_INET:
4248		{
4249			struct sockaddr_in *sin;
4250
4251			sin = (struct sockaddr_in *)firstaddr;
4252			if ((ntohs(sin->sin_port) == 0) ||
4253			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4254			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4255			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4256				/* Invalid address */
4257				SCTP_INP_RUNLOCK(inp);
4258				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4259				*error = EINVAL;
4260				return (NULL);
4261			}
4262			rport = sin->sin_port;
4263			break;
4264		}
4265#endif
4266#ifdef INET6
4267	case AF_INET6:
4268		{
4269			struct sockaddr_in6 *sin6;
4270
4271			sin6 = (struct sockaddr_in6 *)firstaddr;
4272			if ((ntohs(sin6->sin6_port) == 0) ||
4273			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4274			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4275				/* Invalid address */
4276				SCTP_INP_RUNLOCK(inp);
4277				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4278				*error = EINVAL;
4279				return (NULL);
4280			}
4281			rport = sin6->sin6_port;
4282			break;
4283		}
4284#endif
4285	default:
4286		/* not supported family type */
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	SCTP_INP_RUNLOCK(inp);
4293	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4294		/*
4295		 * If you have not performed a bind, then we need to do the
4296		 * ephemeral bind for you.
4297		 */
4298		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4299		    (struct sockaddr *)NULL,
4300		    (struct sctp_ifa *)NULL,
4301		    p
4302		    ))) {
4303			/* bind error, probably perm */
4304			*error = err;
4305			return (NULL);
4306		}
4307	}
4308	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4309	if (stcb == NULL) {
4310		/* out of memory? */
4311		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4312		*error = ENOMEM;
4313		return (NULL);
4314	}
4315	SCTP_INCR_ASOC_COUNT();
4316
4317	bzero(stcb, sizeof(*stcb));
4318	asoc = &stcb->asoc;
4319
4320	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4321	SCTP_TCB_LOCK_INIT(stcb);
4322	SCTP_TCB_SEND_LOCK_INIT(stcb);
4323	stcb->rport = rport;
4324	/* setup back pointer's */
4325	stcb->sctp_ep = inp;
4326	stcb->sctp_socket = inp->sctp_socket;
4327	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4328		/* failed */
4329		SCTP_TCB_LOCK_DESTROY(stcb);
4330		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4331		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4332		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4333		SCTP_DECR_ASOC_COUNT();
4334		*error = err;
4335		return (NULL);
4336	}
4337	/* and the port */
4338	SCTP_INP_INFO_WLOCK();
4339	SCTP_INP_WLOCK(inp);
4340	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4341		/* inpcb freed while alloc going on */
4342		SCTP_TCB_LOCK_DESTROY(stcb);
4343		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4344		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4345		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4346		SCTP_INP_WUNLOCK(inp);
4347		SCTP_INP_INFO_WUNLOCK();
4348		SCTP_DECR_ASOC_COUNT();
4349		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4350		*error = EINVAL;
4351		return (NULL);
4352	}
4353	SCTP_TCB_LOCK(stcb);
4354
4355	/* now that my_vtag is set, add it to the hash */
4356	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4357	/* put it in the bucket in the vtag hash of assoc's for the system */
4358	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4359	SCTP_INP_INFO_WUNLOCK();
4360
4361	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4362		/* failure.. memory error? */
4363		if (asoc->strmout) {
4364			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4365			asoc->strmout = NULL;
4366		}
4367		if (asoc->mapping_array) {
4368			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4369			asoc->mapping_array = NULL;
4370		}
4371		if (asoc->nr_mapping_array) {
4372			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4373			asoc->nr_mapping_array = NULL;
4374		}
4375		SCTP_DECR_ASOC_COUNT();
4376		SCTP_TCB_UNLOCK(stcb);
4377		SCTP_TCB_LOCK_DESTROY(stcb);
4378		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4379		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4380		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4381		SCTP_INP_WUNLOCK(inp);
4382		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4383		*error = ENOBUFS;
4384		return (NULL);
4385	}
4386	/* Init all the timers */
4387	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4388	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4389	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4390	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4391	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4392	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4393	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4394
4395	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4396	/* now file the port under the hash as well */
4397	if (inp->sctp_tcbhash != NULL) {
4398		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4399		    inp->sctp_hashmark)];
4400		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4401	}
4402	SCTP_INP_WUNLOCK(inp);
4403	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4404	return (stcb);
4405}
4406
4407
4408void
4409sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4410{
4411	struct sctp_association *asoc;
4412
4413	asoc = &stcb->asoc;
4414	asoc->numnets--;
4415	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4416	if (net == asoc->primary_destination) {
4417		/* Reset primary */
4418		struct sctp_nets *lnet;
4419
4420		lnet = TAILQ_FIRST(&asoc->nets);
4421		/*
4422		 * Mobility adaptation Ideally, if deleted destination is
4423		 * the primary, it becomes a fast retransmission trigger by
4424		 * the subsequent SET PRIMARY. (by micchie)
4425		 */
4426		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4427		    SCTP_MOBILITY_BASE) ||
4428		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4429		    SCTP_MOBILITY_FASTHANDOFF)) {
4430			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4431			if (asoc->deleted_primary != NULL) {
4432				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4433				goto out;
4434			}
4435			asoc->deleted_primary = net;
4436			atomic_add_int(&net->ref_count, 1);
4437			memset(&net->lastsa, 0, sizeof(net->lastsa));
4438			memset(&net->lastsv, 0, sizeof(net->lastsv));
4439			sctp_mobility_feature_on(stcb->sctp_ep,
4440			    SCTP_MOBILITY_PRIM_DELETED);
4441			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4442			    stcb->sctp_ep, stcb, NULL);
4443		}
4444out:
4445		/* Try to find a confirmed primary */
4446		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4447	}
4448	if (net == asoc->last_data_chunk_from) {
4449		/* Reset primary */
4450		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4451	}
4452	if (net == asoc->last_control_chunk_from) {
4453		/* Clear net */
4454		asoc->last_control_chunk_from = NULL;
4455	}
4456	if (net == stcb->asoc.alternate) {
4457		sctp_free_remote_addr(stcb->asoc.alternate);
4458		stcb->asoc.alternate = NULL;
4459	}
4460	sctp_free_remote_addr(net);
4461}
4462
4463/*
4464 * remove a remote endpoint address from an association, it will fail if the
4465 * address does not exist.
4466 */
4467int
4468sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4469{
4470	/*
4471	 * Here we need to remove a remote address. This is quite simple, we
4472	 * first find it in the list of address for the association
4473	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4474	 * on that item. Note we do not allow it to be removed if there are
4475	 * no other addresses.
4476	 */
4477	struct sctp_association *asoc;
4478	struct sctp_nets *net, *nnet;
4479
4480	asoc = &stcb->asoc;
4481
4482	/* locate the address */
4483	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4484		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4485			continue;
4486		}
4487		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4488		    remaddr)) {
4489			/* we found the guy */
4490			if (asoc->numnets < 2) {
4491				/* Must have at LEAST two remote addresses */
4492				return (-1);
4493			} else {
4494				sctp_remove_net(stcb, net);
4495				return (0);
4496			}
4497		}
4498	}
4499	/* not found. */
4500	return (-2);
4501}
4502
4503void
4504sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4505{
4506	struct sctpvtaghead *chain;
4507	struct sctp_tagblock *twait_block;
4508	int found = 0;
4509	int i;
4510
4511	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4512	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4513		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4514			if ((twait_block->vtag_block[i].v_tag == tag) &&
4515			    (twait_block->vtag_block[i].lport == lport) &&
4516			    (twait_block->vtag_block[i].rport == rport)) {
4517				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4518				twait_block->vtag_block[i].v_tag = 0;
4519				twait_block->vtag_block[i].lport = 0;
4520				twait_block->vtag_block[i].rport = 0;
4521				found = 1;
4522				break;
4523			}
4524		}
4525		if (found)
4526			break;
4527	}
4528}
4529
4530int
4531sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4532{
4533	struct sctpvtaghead *chain;
4534	struct sctp_tagblock *twait_block;
4535	int found = 0;
4536	int i;
4537
4538	SCTP_INP_INFO_WLOCK();
4539	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4540	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4541		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4542			if ((twait_block->vtag_block[i].v_tag == tag) &&
4543			    (twait_block->vtag_block[i].lport == lport) &&
4544			    (twait_block->vtag_block[i].rport == rport)) {
4545				found = 1;
4546				break;
4547			}
4548		}
4549		if (found)
4550			break;
4551	}
4552	SCTP_INP_INFO_WUNLOCK();
4553	return (found);
4554}
4555
4556
4557void
4558sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4559{
4560	struct sctpvtaghead *chain;
4561	struct sctp_tagblock *twait_block;
4562	struct timeval now;
4563	int set, i;
4564
4565	if (time == 0) {
4566		/* Its disabled */
4567		return;
4568	}
4569	(void)SCTP_GETTIME_TIMEVAL(&now);
4570	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4571	set = 0;
4572	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4573		/* Block(s) present, lets find space, and expire on the fly */
4574		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4575			if ((twait_block->vtag_block[i].v_tag == 0) &&
4576			    !set) {
4577				twait_block->vtag_block[i].tv_sec_at_expire =
4578				    now.tv_sec + time;
4579				twait_block->vtag_block[i].v_tag = tag;
4580				twait_block->vtag_block[i].lport = lport;
4581				twait_block->vtag_block[i].rport = rport;
4582				set = 1;
4583			} else if ((twait_block->vtag_block[i].v_tag) &&
4584			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4585				/* Audit expires this guy */
4586				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4587				twait_block->vtag_block[i].v_tag = 0;
4588				twait_block->vtag_block[i].lport = 0;
4589				twait_block->vtag_block[i].rport = 0;
4590				if (set == 0) {
4591					/* Reuse it for my new tag */
4592					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4593					twait_block->vtag_block[i].v_tag = tag;
4594					twait_block->vtag_block[i].lport = lport;
4595					twait_block->vtag_block[i].rport = rport;
4596					set = 1;
4597				}
4598			}
4599		}
4600		if (set) {
4601			/*
4602			 * We only do up to the block where we can place our
4603			 * tag for audits
4604			 */
4605			break;
4606		}
4607	}
4608	/* Need to add a new block to chain */
4609	if (!set) {
4610		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4611		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4612		if (twait_block == NULL) {
4613#ifdef INVARIANTS
4614			panic("Can not alloc tagblock");
4615#endif
4616			return;
4617		}
4618		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4619		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4620		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4621		twait_block->vtag_block[0].v_tag = tag;
4622		twait_block->vtag_block[0].lport = lport;
4623		twait_block->vtag_block[0].rport = rport;
4624	}
4625}
4626
4627
4628
4629/*-
4630 * Free the association after un-hashing the remote port. This
4631 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4632 * expect that the input to this function IS a locked TCB.
4633 * It will return 0, if it did NOT destroy the association (instead
4634 * it unlocks it. It will return NON-zero if it either destroyed the
4635 * association OR the association is already destroyed.
4636 */
4637int
4638sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4639{
4640	int i;
4641	struct sctp_association *asoc;
4642	struct sctp_nets *net, *nnet;
4643	struct sctp_laddr *laddr, *naddr;
4644	struct sctp_tmit_chunk *chk, *nchk;
4645	struct sctp_asconf_addr *aparam, *naparam;
4646	struct sctp_asconf_ack *aack, *naack;
4647	struct sctp_stream_reset_list *strrst, *nstrrst;
4648	struct sctp_queued_to_read *sq, *nsq;
4649	struct sctp_stream_queue_pending *sp, *nsp;
4650	sctp_sharedkey_t *shared_key, *nshared_key;
4651	struct socket *so;
4652
4653	/* first, lets purge the entry from the hash table. */
4654
4655#ifdef SCTP_LOG_CLOSING
4656	sctp_log_closing(inp, stcb, 6);
4657#endif
4658	if (stcb->asoc.state == 0) {
4659#ifdef SCTP_LOG_CLOSING
4660		sctp_log_closing(inp, NULL, 7);
4661#endif
4662		/* there is no asoc, really TSNH :-0 */
4663		return (1);
4664	}
4665	if (stcb->asoc.alternate) {
4666		sctp_free_remote_addr(stcb->asoc.alternate);
4667		stcb->asoc.alternate = NULL;
4668	}
4669	/* TEMP CODE */
4670	if (stcb->freed_from_where == 0) {
4671		/* Only record the first place free happened from */
4672		stcb->freed_from_where = from_location;
4673	}
4674	/* TEMP CODE */
4675
4676	asoc = &stcb->asoc;
4677	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4678	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4679		/* nothing around */
4680		so = NULL;
4681	else
4682		so = inp->sctp_socket;
4683
4684	/*
4685	 * We used timer based freeing if a reader or writer is in the way.
4686	 * So we first check if we are actually being called from a timer,
4687	 * if so we abort early if a reader or writer is still in the way.
4688	 */
4689	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4690	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4691		/*
4692		 * is it the timer driving us? if so are the reader/writers
4693		 * gone?
4694		 */
4695		if (stcb->asoc.refcnt) {
4696			/* nope, reader or writer in the way */
4697			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4698			/* no asoc destroyed */
4699			SCTP_TCB_UNLOCK(stcb);
4700#ifdef SCTP_LOG_CLOSING
4701			sctp_log_closing(inp, stcb, 8);
4702#endif
4703			return (0);
4704		}
4705	}
4706	/* now clean up any other timers */
4707	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4708	asoc->dack_timer.self = NULL;
4709	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4710	/*-
4711	 * For stream reset we don't blast this unless
4712	 * it is a str-reset timer, it might be the
4713	 * free-asoc timer which we DON'T want to
4714	 * disturb.
4715	 */
4716	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4717		asoc->strreset_timer.self = NULL;
4718	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4719	asoc->asconf_timer.self = NULL;
4720	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4721	asoc->autoclose_timer.self = NULL;
4722	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4723	asoc->shut_guard_timer.self = NULL;
4724	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4725	asoc->delayed_event_timer.self = NULL;
4726	/* Mobility adaptation */
4727	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4728	asoc->delete_prim_timer.self = NULL;
4729	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4730		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4731		net->rxt_timer.self = NULL;
4732		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4733		net->pmtu_timer.self = NULL;
4734		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4735		net->hb_timer.self = NULL;
4736	}
4737	/* Now the read queue needs to be cleaned up (only once) */
4738	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4739		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4740		SCTP_INP_READ_LOCK(inp);
4741		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4742			if (sq->stcb == stcb) {
4743				sq->do_not_ref_stcb = 1;
4744				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4745				/*
4746				 * If there is no end, there never will be
4747				 * now.
4748				 */
4749				if (sq->end_added == 0) {
4750					/* Held for PD-API clear that. */
4751					sq->pdapi_aborted = 1;
4752					sq->held_length = 0;
4753					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4754						/*
4755						 * Need to add a PD-API
4756						 * aborted indication.
4757						 * Setting the control_pdapi
4758						 * assures that it will be
4759						 * added right after this
4760						 * msg.
4761						 */
4762						uint32_t strseq;
4763
4764						stcb->asoc.control_pdapi = sq;
4765						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4766						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4767						    stcb,
4768						    SCTP_PARTIAL_DELIVERY_ABORTED,
4769						    (void *)&strseq,
4770						    SCTP_SO_LOCKED);
4771						stcb->asoc.control_pdapi = NULL;
4772					}
4773				}
4774				/* Add an end to wake them */
4775				sq->end_added = 1;
4776			}
4777		}
4778		SCTP_INP_READ_UNLOCK(inp);
4779		if (stcb->block_entry) {
4780			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4781			stcb->block_entry->error = ECONNRESET;
4782			stcb->block_entry = NULL;
4783		}
4784	}
4785	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4786		/*
4787		 * Someone holds a reference OR the socket is unaccepted
4788		 * yet.
4789		 */
4790		if ((stcb->asoc.refcnt) ||
4791		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4792		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4793			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4794			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4795		}
4796		SCTP_TCB_UNLOCK(stcb);
4797		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4798		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4799			/* nothing around */
4800			so = NULL;
4801		if (so) {
4802			/* Wake any reader/writers */
4803			sctp_sorwakeup(inp, so);
4804			sctp_sowwakeup(inp, so);
4805		}
4806#ifdef SCTP_LOG_CLOSING
4807		sctp_log_closing(inp, stcb, 9);
4808#endif
4809		/* no asoc destroyed */
4810		return (0);
4811	}
4812#ifdef SCTP_LOG_CLOSING
4813	sctp_log_closing(inp, stcb, 10);
4814#endif
4815	/*
4816	 * When I reach here, no others want to kill the assoc yet.. and I
4817	 * own the lock. Now its possible an abort comes in when I do the
4818	 * lock exchange below to grab all the locks to do the final take
4819	 * out. to prevent this we increment the count, which will start a
4820	 * timer and blow out above thus assuring us that we hold exclusive
4821	 * killing of the asoc. Note that after getting back the TCB lock we
4822	 * will go ahead and increment the counter back up and stop any
4823	 * timer a passing stranger may have started :-S
4824	 */
4825	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4826		atomic_add_int(&stcb->asoc.refcnt, 1);
4827
4828		SCTP_TCB_UNLOCK(stcb);
4829		SCTP_INP_INFO_WLOCK();
4830		SCTP_INP_WLOCK(inp);
4831		SCTP_TCB_LOCK(stcb);
4832	}
4833	/* Double check the GONE flag */
4834	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4835	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4836		/* nothing around */
4837		so = NULL;
4838
4839	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4840	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4841		/*
4842		 * For TCP type we need special handling when we are
4843		 * connected. We also include the peel'ed off ones to.
4844		 */
4845		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4846			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4847			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4848			if (so) {
4849				SOCK_LOCK(so);
4850				if (so->so_rcv.sb_cc == 0) {
4851					so->so_state &= ~(SS_ISCONNECTING |
4852					    SS_ISDISCONNECTING |
4853					    SS_ISCONFIRMING |
4854					    SS_ISCONNECTED);
4855				}
4856				socantrcvmore_locked(so);
4857				sctp_sowwakeup(inp, so);
4858				sctp_sorwakeup(inp, so);
4859				SCTP_SOWAKEUP(so);
4860			}
4861		}
4862	}
4863	/*
4864	 * Make it invalid too, that way if its about to run it will abort
4865	 * and return.
4866	 */
4867	/* re-increment the lock */
4868	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4869		atomic_add_int(&stcb->asoc.refcnt, -1);
4870	}
4871	if (stcb->asoc.refcnt) {
4872		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4873		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4874		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4875			SCTP_INP_INFO_WUNLOCK();
4876			SCTP_INP_WUNLOCK(inp);
4877		}
4878		SCTP_TCB_UNLOCK(stcb);
4879		return (0);
4880	}
4881	asoc->state = 0;
4882	if (inp->sctp_tcbhash) {
4883		LIST_REMOVE(stcb, sctp_tcbhash);
4884	}
4885	if (stcb->asoc.in_asocid_hash) {
4886		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4887	}
4888	/* Now lets remove it from the list of ALL associations in the EP */
4889	LIST_REMOVE(stcb, sctp_tcblist);
4890	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4891		SCTP_INP_INCR_REF(inp);
4892		SCTP_INP_WUNLOCK(inp);
4893	}
4894	/* pull from vtag hash */
4895	LIST_REMOVE(stcb, sctp_asocs);
4896	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4897	    inp->sctp_lport, stcb->rport);
4898
4899	/*
4900	 * Now restop the timers to be sure this is paranoia at is finest!
4901	 */
4902	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4903	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4904	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4905	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4906	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4907	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4908	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4909	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4910		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4911		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4912		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4913	}
4914
4915	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4916	/*
4917	 * The chunk lists and such SHOULD be empty but we check them just
4918	 * in case.
4919	 */
4920	/* anything on the wheel needs to be removed */
4921	for (i = 0; i < asoc->streamoutcnt; i++) {
4922		struct sctp_stream_out *outs;
4923
4924		outs = &asoc->strmout[i];
4925		/* now clean up any chunks here */
4926		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4927			TAILQ_REMOVE(&outs->outqueue, sp, next);
4928			sctp_free_spbufspace(stcb, asoc, sp);
4929			if (sp->data) {
4930				if (so) {
4931					/* Still an open socket - report */
4932					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4933					    0, (void *)sp, SCTP_SO_LOCKED);
4934				}
4935				if (sp->data) {
4936					sctp_m_freem(sp->data);
4937					sp->data = NULL;
4938					sp->tail_mbuf = NULL;
4939					sp->length = 0;
4940				}
4941			}
4942			if (sp->net) {
4943				sctp_free_remote_addr(sp->net);
4944				sp->net = NULL;
4945			}
4946			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4947		}
4948	}
4949	/* sa_ignore FREED_MEMORY */
4950	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4951		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4952		SCTP_FREE(strrst, SCTP_M_STRESET);
4953	}
4954	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4955		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4956		if (sq->data) {
4957			sctp_m_freem(sq->data);
4958			sq->data = NULL;
4959		}
4960		sctp_free_remote_addr(sq->whoFrom);
4961		sq->whoFrom = NULL;
4962		sq->stcb = NULL;
4963		/* Free the ctl entry */
4964		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4965		SCTP_DECR_READQ_COUNT();
4966		/* sa_ignore FREED_MEMORY */
4967	}
4968	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4969		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4970		if (chk->data) {
4971			sctp_m_freem(chk->data);
4972			chk->data = NULL;
4973		}
4974		if (chk->holds_key_ref)
4975			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4976		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4977		SCTP_DECR_CHK_COUNT();
4978		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4979		asoc->free_chunk_cnt--;
4980		/* sa_ignore FREED_MEMORY */
4981	}
4982	/* pending send queue SHOULD be empty */
4983	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4984		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4985			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4986#ifdef INVARIANTS
4987		} else {
4988			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4989#endif
4990		}
4991		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4992		if (chk->data) {
4993			if (so) {
4994				/* Still a socket? */
4995				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
4996				    0, chk, SCTP_SO_LOCKED);
4997			}
4998			if (chk->data) {
4999				sctp_m_freem(chk->data);
5000				chk->data = NULL;
5001			}
5002		}
5003		if (chk->holds_key_ref)
5004			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5005		if (chk->whoTo) {
5006			sctp_free_remote_addr(chk->whoTo);
5007			chk->whoTo = NULL;
5008		}
5009		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5010		SCTP_DECR_CHK_COUNT();
5011		/* sa_ignore FREED_MEMORY */
5012	}
5013	/* sent queue SHOULD be empty */
5014	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5015		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5016			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
5017				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
5018#ifdef INVARIANTS
5019			} else {
5020				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
5021#endif
5022			}
5023		}
5024		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5025		if (chk->data) {
5026			if (so) {
5027				/* Still a socket? */
5028				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5029				    0, chk, SCTP_SO_LOCKED);
5030			}
5031			if (chk->data) {
5032				sctp_m_freem(chk->data);
5033				chk->data = NULL;
5034			}
5035		}
5036		if (chk->holds_key_ref)
5037			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5038		sctp_free_remote_addr(chk->whoTo);
5039		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5040		SCTP_DECR_CHK_COUNT();
5041		/* sa_ignore FREED_MEMORY */
5042	}
5043#ifdef INVARIANTS
5044	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5045		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5046			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5047		}
5048	}
5049#endif
5050	/* control queue MAY not be empty */
5051	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5052		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5053		if (chk->data) {
5054			sctp_m_freem(chk->data);
5055			chk->data = NULL;
5056		}
5057		if (chk->holds_key_ref)
5058			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5059		sctp_free_remote_addr(chk->whoTo);
5060		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5061		SCTP_DECR_CHK_COUNT();
5062		/* sa_ignore FREED_MEMORY */
5063	}
5064	/* ASCONF queue MAY not be empty */
5065	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5066		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5067		if (chk->data) {
5068			sctp_m_freem(chk->data);
5069			chk->data = NULL;
5070		}
5071		if (chk->holds_key_ref)
5072			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5073		sctp_free_remote_addr(chk->whoTo);
5074		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5075		SCTP_DECR_CHK_COUNT();
5076		/* sa_ignore FREED_MEMORY */
5077	}
5078	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5079		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5080		if (chk->data) {
5081			sctp_m_freem(chk->data);
5082			chk->data = NULL;
5083		}
5084		if (chk->holds_key_ref)
5085			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5086		sctp_free_remote_addr(chk->whoTo);
5087		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5088		SCTP_DECR_CHK_COUNT();
5089		/* sa_ignore FREED_MEMORY */
5090	}
5091
5092	if (asoc->mapping_array) {
5093		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5094		asoc->mapping_array = NULL;
5095	}
5096	if (asoc->nr_mapping_array) {
5097		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5098		asoc->nr_mapping_array = NULL;
5099	}
5100	/* the stream outs */
5101	if (asoc->strmout) {
5102		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5103		asoc->strmout = NULL;
5104	}
5105	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5106	if (asoc->strmin) {
5107		struct sctp_queued_to_read *ctl, *nctl;
5108
5109		for (i = 0; i < asoc->streamincnt; i++) {
5110			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5111				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5112				sctp_free_remote_addr(ctl->whoFrom);
5113				if (ctl->data) {
5114					sctp_m_freem(ctl->data);
5115					ctl->data = NULL;
5116				}
5117				/*
5118				 * We don't free the address here since all
5119				 * the net's were freed above.
5120				 */
5121				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5122				SCTP_DECR_READQ_COUNT();
5123			}
5124		}
5125		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5126		asoc->strmin = NULL;
5127	}
5128	asoc->streamincnt = 0;
5129	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5130#ifdef INVARIANTS
5131		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5132			panic("no net's left alloc'ed, or list points to itself");
5133		}
5134#endif
5135		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5136		sctp_free_remote_addr(net);
5137	}
5138	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5139		/* sa_ignore FREED_MEMORY */
5140		sctp_remove_laddr(laddr);
5141	}
5142
5143	/* pending asconf (address) parameters */
5144	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5145		/* sa_ignore FREED_MEMORY */
5146		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5147		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5148	}
5149	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5150		/* sa_ignore FREED_MEMORY */
5151		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5152		if (aack->data != NULL) {
5153			sctp_m_freem(aack->data);
5154		}
5155		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5156	}
5157	/* clean up auth stuff */
5158	if (asoc->local_hmacs)
5159		sctp_free_hmaclist(asoc->local_hmacs);
5160	if (asoc->peer_hmacs)
5161		sctp_free_hmaclist(asoc->peer_hmacs);
5162
5163	if (asoc->local_auth_chunks)
5164		sctp_free_chunklist(asoc->local_auth_chunks);
5165	if (asoc->peer_auth_chunks)
5166		sctp_free_chunklist(asoc->peer_auth_chunks);
5167
5168	sctp_free_authinfo(&asoc->authinfo);
5169
5170	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5171		LIST_REMOVE(shared_key, next);
5172		sctp_free_sharedkey(shared_key);
5173		/* sa_ignore FREED_MEMORY */
5174	}
5175
5176	/* Insert new items here :> */
5177
5178	/* Get rid of LOCK */
5179	SCTP_TCB_UNLOCK(stcb);
5180	SCTP_TCB_LOCK_DESTROY(stcb);
5181	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5182	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5183		SCTP_INP_INFO_WUNLOCK();
5184		SCTP_INP_RLOCK(inp);
5185	}
5186#ifdef SCTP_TRACK_FREED_ASOCS
5187	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5188		/* now clean up the tasoc itself */
5189		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5190		SCTP_DECR_ASOC_COUNT();
5191	} else {
5192		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5193	}
5194#else
5195	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5196	SCTP_DECR_ASOC_COUNT();
5197#endif
5198	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5199		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5200			/*
5201			 * If its NOT the inp_free calling us AND sctp_close
5202			 * as been called, we call back...
5203			 */
5204			SCTP_INP_RUNLOCK(inp);
5205			/*
5206			 * This will start the kill timer (if we are the
5207			 * last one) since we hold an increment yet. But
5208			 * this is the only safe way to do this since
5209			 * otherwise if the socket closes at the same time
5210			 * we are here we might collide in the cleanup.
5211			 */
5212			sctp_inpcb_free(inp,
5213			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5214			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5215			SCTP_INP_DECR_REF(inp);
5216			goto out_of;
5217		} else {
5218			/* The socket is still open. */
5219			SCTP_INP_DECR_REF(inp);
5220		}
5221	}
5222	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5223		SCTP_INP_RUNLOCK(inp);
5224	}
5225out_of:
5226	/* destroyed the asoc */
5227#ifdef SCTP_LOG_CLOSING
5228	sctp_log_closing(inp, NULL, 11);
5229#endif
5230	return (1);
5231}
5232
5233
5234
5235/*
5236 * determine if a destination is "reachable" based upon the addresses bound
5237 * to the current endpoint (e.g. only v4 or v6 currently bound)
5238 */
5239/*
5240 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5241 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5242 * types bound as its endpoint
5243 */
5244int
5245sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5246{
5247	struct sctp_inpcb *inp;
5248	int answer;
5249
5250	/*
5251	 * No locks here, the TCB, in all cases is already locked and an
5252	 * assoc is up. There is either a INP lock by the caller applied (in
5253	 * asconf case when deleting an address) or NOT in the HB case,
5254	 * however if HB then the INP increment is up and the INP will not
5255	 * be removed (on top of the fact that we have a TCB lock). So we
5256	 * only want to read the sctp_flags, which is either bound-all or
5257	 * not.. no protection needed since once an assoc is up you can't be
5258	 * changing your binding.
5259	 */
5260	inp = stcb->sctp_ep;
5261	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5262		/* if bound all, destination is not restricted */
5263		/*
5264		 * RRS: Question during lock work: Is this correct? If you
5265		 * are bound-all you still might need to obey the V4--V6
5266		 * flags??? IMO this bound-all stuff needs to be removed!
5267		 */
5268		return (1);
5269	}
5270	/* NOTE: all "scope" checks are done when local addresses are added */
5271	switch (destaddr->sa_family) {
5272#ifdef INET6
5273	case AF_INET6:
5274		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5275		break;
5276#endif
5277#ifdef INET
5278	case AF_INET:
5279		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5280		break;
5281#endif
5282	default:
5283		/* invalid family, so it's unreachable */
5284		answer = 0;
5285		break;
5286	}
5287	return (answer);
5288}
5289
5290/*
5291 * update the inp_vflags on an endpoint
5292 */
5293static void
5294sctp_update_ep_vflag(struct sctp_inpcb *inp)
5295{
5296	struct sctp_laddr *laddr;
5297
5298	/* first clear the flag */
5299	inp->ip_inp.inp.inp_vflag = 0;
5300	/* set the flag based on addresses on the ep list */
5301	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5302		if (laddr->ifa == NULL) {
5303			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5304			    __FUNCTION__);
5305			continue;
5306		}
5307		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5308			continue;
5309		}
5310		switch (laddr->ifa->address.sa.sa_family) {
5311#ifdef INET6
5312		case AF_INET6:
5313			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5314			break;
5315#endif
5316#ifdef INET
5317		case AF_INET:
5318			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5319			break;
5320#endif
5321		default:
5322			break;
5323		}
5324	}
5325}
5326
5327/*
5328 * Add the address to the endpoint local address list There is nothing to be
5329 * done if we are bound to all addresses
5330 */
5331void
5332sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5333{
5334	struct sctp_laddr *laddr;
5335	int fnd, error = 0;
5336
5337	fnd = 0;
5338
5339	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5340		/* You are already bound to all. You have it already */
5341		return;
5342	}
5343#ifdef INET6
5344	if (ifa->address.sa.sa_family == AF_INET6) {
5345		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5346			/* Can't bind a non-useable addr. */
5347			return;
5348		}
5349	}
5350#endif
5351	/* first, is it already present? */
5352	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5353		if (laddr->ifa == ifa) {
5354			fnd = 1;
5355			break;
5356		}
5357	}
5358
5359	if (fnd == 0) {
5360		/* Not in the ep list */
5361		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5362		if (error != 0)
5363			return;
5364		inp->laddr_count++;
5365		/* update inp_vflag flags */
5366		switch (ifa->address.sa.sa_family) {
5367#ifdef INET6
5368		case AF_INET6:
5369			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5370			break;
5371#endif
5372#ifdef INET
5373		case AF_INET:
5374			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5375			break;
5376#endif
5377		default:
5378			break;
5379		}
5380	}
5381	return;
5382}
5383
5384
5385/*
5386 * select a new (hopefully reachable) destination net (should only be used
5387 * when we deleted an ep addr that is the only usable source address to reach
5388 * the destination net)
5389 */
5390static void
5391sctp_select_primary_destination(struct sctp_tcb *stcb)
5392{
5393	struct sctp_nets *net;
5394
5395	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5396		/* for now, we'll just pick the first reachable one we find */
5397		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5398			continue;
5399		if (sctp_destination_is_reachable(stcb,
5400		    (struct sockaddr *)&net->ro._l_addr)) {
5401			/* found a reachable destination */
5402			stcb->asoc.primary_destination = net;
5403		}
5404	}
5405	/* I can't there from here! ...we're gonna die shortly... */
5406}
5407
5408
5409/*
5410 * Delete the address from the endpoint local address list There is nothing
5411 * to be done if we are bound to all addresses
5412 */
5413void
5414sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5415{
5416	struct sctp_laddr *laddr;
5417	int fnd;
5418
5419	fnd = 0;
5420	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5421		/* You are already bound to all. You have it already */
5422		return;
5423	}
5424	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5425		if (laddr->ifa == ifa) {
5426			fnd = 1;
5427			break;
5428		}
5429	}
5430	if (fnd && (inp->laddr_count < 2)) {
5431		/* can't delete unless there are at LEAST 2 addresses */
5432		return;
5433	}
5434	if (fnd) {
5435		/*
5436		 * clean up any use of this address go through our
5437		 * associations and clear any last_used_address that match
5438		 * this one for each assoc, see if a new primary_destination
5439		 * is needed
5440		 */
5441		struct sctp_tcb *stcb;
5442
5443		/* clean up "next_addr_touse" */
5444		if (inp->next_addr_touse == laddr)
5445			/* delete this address */
5446			inp->next_addr_touse = NULL;
5447
5448		/* clean up "last_used_address" */
5449		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5450			struct sctp_nets *net;
5451
5452			SCTP_TCB_LOCK(stcb);
5453			if (stcb->asoc.last_used_address == laddr)
5454				/* delete this address */
5455				stcb->asoc.last_used_address = NULL;
5456			/*
5457			 * Now spin through all the nets and purge any ref
5458			 * to laddr
5459			 */
5460			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5461				if (net->ro._s_addr &&
5462				    (net->ro._s_addr->ifa == laddr->ifa)) {
5463					/* Yep, purge src address selected */
5464					sctp_rtentry_t *rt;
5465
5466					/* delete this address if cached */
5467					rt = net->ro.ro_rt;
5468					if (rt != NULL) {
5469						RTFREE(rt);
5470						net->ro.ro_rt = NULL;
5471					}
5472					sctp_free_ifa(net->ro._s_addr);
5473					net->ro._s_addr = NULL;
5474					net->src_addr_selected = 0;
5475				}
5476			}
5477			SCTP_TCB_UNLOCK(stcb);
5478		}		/* for each tcb */
5479		/* remove it from the ep list */
5480		sctp_remove_laddr(laddr);
5481		inp->laddr_count--;
5482		/* update inp_vflag flags */
5483		sctp_update_ep_vflag(inp);
5484	}
5485	return;
5486}
5487
5488/*
5489 * Add the address to the TCB local address restricted list.
5490 * This is a "pending" address list (eg. addresses waiting for an
5491 * ASCONF-ACK response) and cannot be used as a valid source address.
5492 */
5493void
5494sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5495{
5496	struct sctp_laddr *laddr;
5497	struct sctpladdr *list;
5498
5499	/*
5500	 * Assumes TCB is locked.. and possibly the INP. May need to
5501	 * confirm/fix that if we need it and is not the case.
5502	 */
5503	list = &stcb->asoc.sctp_restricted_addrs;
5504
5505#ifdef INET6
5506	if (ifa->address.sa.sa_family == AF_INET6) {
5507		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5508			/* Can't bind a non-existent addr. */
5509			return;
5510		}
5511	}
5512#endif
5513	/* does the address already exist? */
5514	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5515		if (laddr->ifa == ifa) {
5516			return;
5517		}
5518	}
5519
5520	/* add to the list */
5521	(void)sctp_insert_laddr(list, ifa, 0);
5522	return;
5523}
5524
5525/*
5526 * insert an laddr entry with the given ifa for the desired list
5527 */
5528int
5529sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5530{
5531	struct sctp_laddr *laddr;
5532
5533	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5534	if (laddr == NULL) {
5535		/* out of memory? */
5536		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5537		return (EINVAL);
5538	}
5539	SCTP_INCR_LADDR_COUNT();
5540	bzero(laddr, sizeof(*laddr));
5541	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5542	laddr->ifa = ifa;
5543	laddr->action = act;
5544	atomic_add_int(&ifa->refcount, 1);
5545	/* insert it */
5546	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5547
5548	return (0);
5549}
5550
5551/*
5552 * Remove an laddr entry from the local address list (on an assoc)
5553 */
5554void
5555sctp_remove_laddr(struct sctp_laddr *laddr)
5556{
5557
5558	/* remove from the list */
5559	LIST_REMOVE(laddr, sctp_nxt_addr);
5560	sctp_free_ifa(laddr->ifa);
5561	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5562	SCTP_DECR_LADDR_COUNT();
5563}
5564
5565/*
5566 * Remove a local address from the TCB local address restricted list
5567 */
5568void
5569sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5570{
5571	struct sctp_inpcb *inp;
5572	struct sctp_laddr *laddr;
5573
5574	/*
5575	 * This is called by asconf work. It is assumed that a) The TCB is
5576	 * locked and b) The INP is locked. This is true in as much as I can
5577	 * trace through the entry asconf code where I did these locks.
5578	 * Again, the ASCONF code is a bit different in that it does lock
5579	 * the INP during its work often times. This must be since we don't
5580	 * want other proc's looking up things while what they are looking
5581	 * up is changing :-D
5582	 */
5583
5584	inp = stcb->sctp_ep;
5585	/* if subset bound and don't allow ASCONF's, can't delete last */
5586	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5587	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5588		if (stcb->sctp_ep->laddr_count < 2) {
5589			/* can't delete last address */
5590			return;
5591		}
5592	}
5593	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5594		/* remove the address if it exists */
5595		if (laddr->ifa == NULL)
5596			continue;
5597		if (laddr->ifa == ifa) {
5598			sctp_remove_laddr(laddr);
5599			return;
5600		}
5601	}
5602
5603	/* address not found! */
5604	return;
5605}
5606
5607/*
5608 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5609 */
5610/* sysctl */
5611static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5612static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5613
5614
5615
5616#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5617struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5618int *sctp_cpuarry = NULL;
5619void
5620sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5621{
5622	/* Queue a packet to a processor for the specified core */
5623	struct sctp_mcore_queue *qent;
5624	struct sctp_mcore_ctrl *wkq;
5625	int need_wake = 0;
5626
5627	if (sctp_mcore_workers == NULL) {
5628		/* Something went way bad during setup */
5629		sctp_input_with_port(m, off, 0);
5630		return;
5631	}
5632	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5633	    (sizeof(struct sctp_mcore_queue)),
5634	    SCTP_M_MCORE);
5635	if (qent == NULL) {
5636		/* This is trouble  */
5637		sctp_input_with_port(m, off, 0);
5638		return;
5639	}
5640	qent->vn = curvnet;
5641	qent->m = m;
5642	qent->off = off;
5643	qent->v6 = 0;
5644	wkq = &sctp_mcore_workers[cpu_to_use];
5645	SCTP_MCORE_QLOCK(wkq);
5646
5647	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5648	if (wkq->running == 0) {
5649		need_wake = 1;
5650	}
5651	SCTP_MCORE_QUNLOCK(wkq);
5652	if (need_wake) {
5653		wakeup(&wkq->running);
5654	}
5655}
5656
5657static void
5658sctp_mcore_thread(void *arg)
5659{
5660
5661	struct sctp_mcore_ctrl *wkq;
5662	struct sctp_mcore_queue *qent;
5663
5664	wkq = (struct sctp_mcore_ctrl *)arg;
5665	struct mbuf *m;
5666	int off, v6;
5667
5668	/* Wait for first tickle */
5669	SCTP_MCORE_LOCK(wkq);
5670	wkq->running = 0;
5671	msleep(&wkq->running,
5672	    &wkq->core_mtx,
5673	    0, "wait for pkt", 0);
5674	SCTP_MCORE_UNLOCK(wkq);
5675
5676	/* Bind to our cpu */
5677	thread_lock(curthread);
5678	sched_bind(curthread, wkq->cpuid);
5679	thread_unlock(curthread);
5680
5681	/* Now lets start working */
5682	SCTP_MCORE_LOCK(wkq);
5683	/* Now grab lock and go */
5684	for (;;) {
5685		SCTP_MCORE_QLOCK(wkq);
5686skip_sleep:
5687		wkq->running = 1;
5688		qent = TAILQ_FIRST(&wkq->que);
5689		if (qent) {
5690			TAILQ_REMOVE(&wkq->que, qent, next);
5691			SCTP_MCORE_QUNLOCK(wkq);
5692			CURVNET_SET(qent->vn);
5693			m = qent->m;
5694			off = qent->off;
5695			v6 = qent->v6;
5696			SCTP_FREE(qent, SCTP_M_MCORE);
5697			if (v6 == 0) {
5698				sctp_input_with_port(m, off, 0);
5699			} else {
5700				SCTP_PRINTF("V6 not yet supported\n");
5701				sctp_m_freem(m);
5702			}
5703			CURVNET_RESTORE();
5704			SCTP_MCORE_QLOCK(wkq);
5705		}
5706		wkq->running = 0;
5707		if (!TAILQ_EMPTY(&wkq->que)) {
5708			goto skip_sleep;
5709		}
5710		SCTP_MCORE_QUNLOCK(wkq);
5711		msleep(&wkq->running,
5712		    &wkq->core_mtx,
5713		    0, "wait for pkt", 0);
5714	}
5715}
5716
5717static void
5718sctp_startup_mcore_threads(void)
5719{
5720	int i, cpu;
5721
5722	if (mp_ncpus == 1)
5723		return;
5724
5725	if (sctp_mcore_workers != NULL) {
5726		/*
5727		 * Already been here in some previous vnet?
5728		 */
5729		return;
5730	}
5731	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5732	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5733	    SCTP_M_MCORE);
5734	if (sctp_mcore_workers == NULL) {
5735		/* TSNH I hope */
5736		return;
5737	}
5738	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5739	    sizeof(struct sctp_mcore_ctrl)));
5740	/* Init the structures */
5741	for (i = 0; i <= mp_maxid; i++) {
5742		TAILQ_INIT(&sctp_mcore_workers[i].que);
5743		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5744		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5745		sctp_mcore_workers[i].cpuid = i;
5746	}
5747	if (sctp_cpuarry == NULL) {
5748		SCTP_MALLOC(sctp_cpuarry, int *,
5749		    (mp_ncpus * sizeof(int)),
5750		    SCTP_M_MCORE);
5751		i = 0;
5752		CPU_FOREACH(cpu) {
5753			sctp_cpuarry[i] = cpu;
5754			i++;
5755		}
5756	}
5757	/* Now start them all */
5758	CPU_FOREACH(cpu) {
5759		(void)kproc_create(sctp_mcore_thread,
5760		    (void *)&sctp_mcore_workers[cpu],
5761		    &sctp_mcore_workers[cpu].thread_proc,
5762		    RFPROC,
5763		    SCTP_KTHREAD_PAGES,
5764		    SCTP_MCORE_NAME);
5765
5766	}
5767}
5768
5769#endif
5770
5771void
5772sctp_pcb_init()
5773{
5774	/*
5775	 * SCTP initialization for the PCB structures should be called by
5776	 * the sctp_init() funciton.
5777	 */
5778	int i;
5779	struct timeval tv;
5780
5781	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5782		/* error I was called twice */
5783		return;
5784	}
5785	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5786
5787#if defined(SCTP_LOCAL_TRACE_BUF)
5788	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5789#endif
5790#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5791	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5792	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5793	    SCTP_M_MCORE);
5794#endif
5795	(void)SCTP_GETTIME_TIMEVAL(&tv);
5796#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5797	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5798	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5799	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5800#else
5801	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5802	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5803	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5804#endif
5805	/* init the empty list of (All) Endpoints */
5806	LIST_INIT(&SCTP_BASE_INFO(listhead));
5807
5808
5809	/* init the hash table of endpoints */
5810	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5811	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5812	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5813	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5814	    &SCTP_BASE_INFO(hashasocmark));
5815	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5816	    &SCTP_BASE_INFO(hashmark));
5817	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5818	    &SCTP_BASE_INFO(hashtcpmark));
5819	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5820
5821
5822	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5823	    &SCTP_BASE_INFO(hashvrfmark));
5824
5825	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5826	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5827	/* init the zones */
5828	/*
5829	 * FIX ME: Should check for NULL returns, but if it does fail we are
5830	 * doomed to panic anyways... add later maybe.
5831	 */
5832	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5833	    sizeof(struct sctp_inpcb), maxsockets);
5834
5835	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5836	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5837
5838	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5839	    sizeof(struct sctp_laddr),
5840	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5841
5842	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5843	    sizeof(struct sctp_nets),
5844	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5845
5846	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5847	    sizeof(struct sctp_tmit_chunk),
5848	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5849
5850	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5851	    sizeof(struct sctp_queued_to_read),
5852	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5853
5854	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5855	    sizeof(struct sctp_stream_queue_pending),
5856	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5857
5858	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5859	    sizeof(struct sctp_asconf),
5860	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5861
5862	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5863	    sizeof(struct sctp_asconf_ack),
5864	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5865
5866
5867	/* Master Lock INIT for info structure */
5868	SCTP_INP_INFO_LOCK_INIT();
5869	SCTP_STATLOG_INIT_LOCK();
5870
5871	SCTP_IPI_COUNT_INIT();
5872	SCTP_IPI_ADDR_INIT();
5873#ifdef SCTP_PACKET_LOGGING
5874	SCTP_IP_PKTLOG_INIT();
5875#endif
5876	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5877
5878	SCTP_WQ_ADDR_INIT();
5879	/* not sure if we need all the counts */
5880	SCTP_BASE_INFO(ipi_count_ep) = 0;
5881	/* assoc/tcb zone info */
5882	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5883	/* local addrlist zone info */
5884	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5885	/* remote addrlist zone info */
5886	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5887	/* chunk info */
5888	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5889
5890	/* socket queue zone info */
5891	SCTP_BASE_INFO(ipi_count_readq) = 0;
5892
5893	/* stream out queue cont */
5894	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5895
5896	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5897	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5898
5899	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5900
5901	/* Init the TIMEWAIT list */
5902	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5903		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5904	}
5905	sctp_startup_iterator();
5906
5907#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5908	sctp_startup_mcore_threads();
5909#endif
5910
5911	/*
5912	 * INIT the default VRF which for BSD is the only one, other O/S's
5913	 * may have more. But initially they must start with one and then
5914	 * add the VRF's as addresses are added.
5915	 */
5916	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5917}
5918
5919/*
5920 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5921 */
5922void
5923sctp_pcb_finish(void)
5924{
5925	struct sctp_vrflist *vrf_bucket;
5926	struct sctp_vrf *vrf, *nvrf;
5927	struct sctp_ifn *ifn, *nifn;
5928	struct sctp_ifa *ifa, *nifa;
5929	struct sctpvtaghead *chain;
5930	struct sctp_tagblock *twait_block, *prev_twait_block;
5931	struct sctp_laddr *wi, *nwi;
5932	int i;
5933	struct sctp_iterator *it, *nit;
5934
5935	/*
5936	 * In FreeBSD the iterator thread never exits but we do clean up.
5937	 * The only way FreeBSD reaches here is if we have VRF's but we
5938	 * still add the ifdef to make it compile on old versions.
5939	 */
5940	SCTP_IPI_ITERATOR_WQ_LOCK();
5941	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5942		if (it->vn != curvnet) {
5943			continue;
5944		}
5945		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5946		if (it->function_atend != NULL) {
5947			(*it->function_atend) (it->pointer, it->val);
5948		}
5949		SCTP_FREE(it, SCTP_M_ITER);
5950	}
5951	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5952	SCTP_ITERATOR_LOCK();
5953	if ((sctp_it_ctl.cur_it) &&
5954	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5955		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5956	}
5957	SCTP_ITERATOR_UNLOCK();
5958	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5959	SCTP_WQ_ADDR_LOCK();
5960	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5961		LIST_REMOVE(wi, sctp_nxt_addr);
5962		SCTP_DECR_LADDR_COUNT();
5963		if (wi->action == SCTP_DEL_IP_ADDRESS) {
5964			SCTP_FREE(wi->ifa, SCTP_M_IFA);
5965		}
5966		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5967	}
5968	SCTP_WQ_ADDR_UNLOCK();
5969
5970	/*
5971	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5972	 * destroyed first).
5973	 */
5974	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5975	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5976		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5977			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5978				/* free the ifa */
5979				LIST_REMOVE(ifa, next_bucket);
5980				LIST_REMOVE(ifa, next_ifa);
5981				SCTP_FREE(ifa, SCTP_M_IFA);
5982			}
5983			/* free the ifn */
5984			LIST_REMOVE(ifn, next_bucket);
5985			LIST_REMOVE(ifn, next_ifn);
5986			SCTP_FREE(ifn, SCTP_M_IFN);
5987		}
5988		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5989		/* free the vrf */
5990		LIST_REMOVE(vrf, next_vrf);
5991		SCTP_FREE(vrf, SCTP_M_VRF);
5992	}
5993	/* free the vrf hashes */
5994	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5995	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5996
5997	/*
5998	 * free the TIMEWAIT list elements malloc'd in the function
5999	 * sctp_add_vtag_to_timewait()...
6000	 */
6001	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
6002		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
6003		if (!LIST_EMPTY(chain)) {
6004			prev_twait_block = NULL;
6005			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6006				if (prev_twait_block) {
6007					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6008				}
6009				prev_twait_block = twait_block;
6010			}
6011			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6012		}
6013	}
6014
6015	/* free the locks and mutexes */
6016#ifdef SCTP_PACKET_LOGGING
6017	SCTP_IP_PKTLOG_DESTROY();
6018#endif
6019	SCTP_IPI_ADDR_DESTROY();
6020	SCTP_STATLOG_DESTROY();
6021	SCTP_INP_INFO_LOCK_DESTROY();
6022
6023	SCTP_WQ_ADDR_DESTROY();
6024
6025	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6026	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6027	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6028	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6029	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6030	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6031	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6032	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6033	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6034	/* Get rid of other stuff to */
6035	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
6036		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
6037	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
6038		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6039	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6040		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6041#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6042	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6043#endif
6044}
6045
6046
6047int
6048sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6049    int offset, int limit,
6050    struct sockaddr *src, struct sockaddr *dst,
6051    struct sockaddr *altsa)
6052{
6053	/*
6054	 * grub through the INIT pulling addresses and loading them to the
6055	 * nets structure in the asoc. The from address in the mbuf should
6056	 * also be loaded (if it is not already). This routine can be called
6057	 * with either INIT or INIT-ACK's as long as the m points to the IP
6058	 * packet and the offset points to the beginning of the parameters.
6059	 */
6060	struct sctp_inpcb *inp;
6061	struct sctp_nets *net, *nnet, *net_tmp;
6062	struct sctp_paramhdr *phdr, parm_buf;
6063	struct sctp_tcb *stcb_tmp;
6064	uint16_t ptype, plen;
6065	struct sockaddr *sa;
6066	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6067	struct sctp_auth_random *p_random = NULL;
6068	uint16_t random_len = 0;
6069	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6070	struct sctp_auth_hmac_algo *hmacs = NULL;
6071	uint16_t hmacs_len = 0;
6072	uint8_t saw_asconf = 0;
6073	uint8_t saw_asconf_ack = 0;
6074	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6075	struct sctp_auth_chunk_list *chunks = NULL;
6076	uint16_t num_chunks = 0;
6077	sctp_key_t *new_key;
6078	uint32_t keylen;
6079	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6080	uint8_t peer_supports_ecn;
6081	uint8_t peer_supports_prsctp;
6082	uint8_t peer_supports_auth;
6083	uint8_t peer_supports_asconf;
6084	uint8_t peer_supports_asconf_ack;
6085	uint8_t peer_supports_reconfig;
6086	uint8_t peer_supports_nrsack;
6087	uint8_t peer_supports_pktdrop;
6088
6089#ifdef INET
6090	struct sockaddr_in sin;
6091
6092#endif
6093#ifdef INET6
6094	struct sockaddr_in6 sin6;
6095
6096#endif
6097
6098	/* First get the destination address setup too. */
6099#ifdef INET
6100	memset(&sin, 0, sizeof(sin));
6101	sin.sin_family = AF_INET;
6102	sin.sin_len = sizeof(sin);
6103	sin.sin_port = stcb->rport;
6104#endif
6105#ifdef INET6
6106	memset(&sin6, 0, sizeof(sin6));
6107	sin6.sin6_family = AF_INET6;
6108	sin6.sin6_len = sizeof(struct sockaddr_in6);
6109	sin6.sin6_port = stcb->rport;
6110#endif
6111	if (altsa) {
6112		sa = altsa;
6113	} else {
6114		sa = src;
6115	}
6116	peer_supports_ecn = 0;
6117	peer_supports_prsctp = 0;
6118	peer_supports_auth = 0;
6119	peer_supports_asconf = 0;
6120	peer_supports_reconfig = 0;
6121	peer_supports_nrsack = 0;
6122	peer_supports_pktdrop = 0;
6123	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6124		/* mark all addresses that we have currently on the list */
6125		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6126	}
6127	/* does the source address already exist? if so skip it */
6128	inp = stcb->sctp_ep;
6129	atomic_add_int(&stcb->asoc.refcnt, 1);
6130	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6131	atomic_add_int(&stcb->asoc.refcnt, -1);
6132
6133	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6134		/* we must add the source address */
6135		/* no scope set here since we have a tcb already. */
6136		switch (sa->sa_family) {
6137#ifdef INET
6138		case AF_INET:
6139			if (stcb->asoc.scope.ipv4_addr_legal) {
6140				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6141					return (-1);
6142				}
6143			}
6144			break;
6145#endif
6146#ifdef INET6
6147		case AF_INET6:
6148			if (stcb->asoc.scope.ipv6_addr_legal) {
6149				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6150					return (-2);
6151				}
6152			}
6153			break;
6154#endif
6155		default:
6156			break;
6157		}
6158	} else {
6159		if (net_tmp != NULL && stcb_tmp == stcb) {
6160			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6161		} else if (stcb_tmp != stcb) {
6162			/* It belongs to another association? */
6163			if (stcb_tmp)
6164				SCTP_TCB_UNLOCK(stcb_tmp);
6165			return (-3);
6166		}
6167	}
6168	if (stcb->asoc.state == 0) {
6169		/* the assoc was freed? */
6170		return (-4);
6171	}
6172	/* now we must go through each of the params. */
6173	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6174	while (phdr) {
6175		ptype = ntohs(phdr->param_type);
6176		plen = ntohs(phdr->param_length);
6177		/*
6178		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6179		 * (uint32_t)ptype, (int)plen);
6180		 */
6181		if (offset + plen > limit) {
6182			break;
6183		}
6184		if (plen == 0) {
6185			break;
6186		}
6187#ifdef INET
6188		if (ptype == SCTP_IPV4_ADDRESS) {
6189			if (stcb->asoc.scope.ipv4_addr_legal) {
6190				struct sctp_ipv4addr_param *p4, p4_buf;
6191
6192				/* ok get the v4 address and check/add */
6193				phdr = sctp_get_next_param(m, offset,
6194				    (struct sctp_paramhdr *)&p4_buf,
6195				    sizeof(p4_buf));
6196				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6197				    phdr == NULL) {
6198					return (-5);
6199				}
6200				p4 = (struct sctp_ipv4addr_param *)phdr;
6201				sin.sin_addr.s_addr = p4->addr;
6202				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6203					/* Skip multi-cast addresses */
6204					goto next_param;
6205				}
6206				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6207				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6208					goto next_param;
6209				}
6210				sa = (struct sockaddr *)&sin;
6211				inp = stcb->sctp_ep;
6212				atomic_add_int(&stcb->asoc.refcnt, 1);
6213				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6214				    dst, stcb);
6215				atomic_add_int(&stcb->asoc.refcnt, -1);
6216
6217				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6218				    inp == NULL) {
6219					/* we must add the source address */
6220					/*
6221					 * no scope set since we have a tcb
6222					 * already
6223					 */
6224
6225					/*
6226					 * we must validate the state again
6227					 * here
6228					 */
6229			add_it_now:
6230					if (stcb->asoc.state == 0) {
6231						/* the assoc was freed? */
6232						return (-7);
6233					}
6234					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6235						return (-8);
6236					}
6237				} else if (stcb_tmp == stcb) {
6238					if (stcb->asoc.state == 0) {
6239						/* the assoc was freed? */
6240						return (-10);
6241					}
6242					if (net != NULL) {
6243						/* clear flag */
6244						net->dest_state &=
6245						    ~SCTP_ADDR_NOT_IN_ASSOC;
6246					}
6247				} else {
6248					/*
6249					 * strange, address is in another
6250					 * assoc? straighten out locks.
6251					 */
6252					if (stcb_tmp) {
6253						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6254							/*
6255							 * in setup state we
6256							 * abort this guy
6257							 */
6258							sctp_abort_an_association(stcb_tmp->sctp_ep,
6259							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6260							goto add_it_now;
6261						}
6262						SCTP_TCB_UNLOCK(stcb_tmp);
6263					}
6264					if (stcb->asoc.state == 0) {
6265						/* the assoc was freed? */
6266						return (-12);
6267					}
6268					return (-13);
6269				}
6270			}
6271		} else
6272#endif
6273#ifdef INET6
6274		if (ptype == SCTP_IPV6_ADDRESS) {
6275			if (stcb->asoc.scope.ipv6_addr_legal) {
6276				/* ok get the v6 address and check/add */
6277				struct sctp_ipv6addr_param *p6, p6_buf;
6278
6279				phdr = sctp_get_next_param(m, offset,
6280				    (struct sctp_paramhdr *)&p6_buf,
6281				    sizeof(p6_buf));
6282				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6283				    phdr == NULL) {
6284					return (-14);
6285				}
6286				p6 = (struct sctp_ipv6addr_param *)phdr;
6287				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6288				    sizeof(p6->addr));
6289				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6290					/* Skip multi-cast addresses */
6291					goto next_param;
6292				}
6293				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6294					/*
6295					 * Link local make no sense without
6296					 * scope
6297					 */
6298					goto next_param;
6299				}
6300				sa = (struct sockaddr *)&sin6;
6301				inp = stcb->sctp_ep;
6302				atomic_add_int(&stcb->asoc.refcnt, 1);
6303				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6304				    dst, stcb);
6305				atomic_add_int(&stcb->asoc.refcnt, -1);
6306				if (stcb_tmp == NULL &&
6307				    (inp == stcb->sctp_ep || inp == NULL)) {
6308					/*
6309					 * we must validate the state again
6310					 * here
6311					 */
6312			add_it_now6:
6313					if (stcb->asoc.state == 0) {
6314						/* the assoc was freed? */
6315						return (-16);
6316					}
6317					/*
6318					 * we must add the address, no scope
6319					 * set
6320					 */
6321					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6322						return (-17);
6323					}
6324				} else if (stcb_tmp == stcb) {
6325					/*
6326					 * we must validate the state again
6327					 * here
6328					 */
6329					if (stcb->asoc.state == 0) {
6330						/* the assoc was freed? */
6331						return (-19);
6332					}
6333					if (net != NULL) {
6334						/* clear flag */
6335						net->dest_state &=
6336						    ~SCTP_ADDR_NOT_IN_ASSOC;
6337					}
6338				} else {
6339					/*
6340					 * strange, address is in another
6341					 * assoc? straighten out locks.
6342					 */
6343					if (stcb_tmp)
6344						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6345							/*
6346							 * in setup state we
6347							 * abort this guy
6348							 */
6349							sctp_abort_an_association(stcb_tmp->sctp_ep,
6350							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6351							goto add_it_now6;
6352						}
6353					SCTP_TCB_UNLOCK(stcb_tmp);
6354
6355					if (stcb->asoc.state == 0) {
6356						/* the assoc was freed? */
6357						return (-21);
6358					}
6359					return (-22);
6360				}
6361			}
6362		} else
6363#endif
6364		if (ptype == SCTP_ECN_CAPABLE) {
6365			peer_supports_ecn = 1;
6366		} else if (ptype == SCTP_ULP_ADAPTATION) {
6367			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6368				struct sctp_adaptation_layer_indication ai,
6369				                                *aip;
6370
6371				phdr = sctp_get_next_param(m, offset,
6372				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6373				aip = (struct sctp_adaptation_layer_indication *)phdr;
6374				if (aip) {
6375					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6376					stcb->asoc.adaptation_needed = 1;
6377				}
6378			}
6379		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6380			struct sctp_asconf_addr_param lstore, *fee;
6381			int lptype;
6382			struct sockaddr *lsa = NULL;
6383
6384#ifdef INET
6385			struct sctp_asconf_addrv4_param *fii;
6386
6387#endif
6388
6389			if (stcb->asoc.asconf_supported == 0) {
6390				return (-100);
6391			}
6392			if (plen > sizeof(lstore)) {
6393				return (-23);
6394			}
6395			phdr = sctp_get_next_param(m, offset,
6396			    (struct sctp_paramhdr *)&lstore,
6397			    min(plen, sizeof(lstore)));
6398			if (phdr == NULL) {
6399				return (-24);
6400			}
6401			fee = (struct sctp_asconf_addr_param *)phdr;
6402			lptype = ntohs(fee->addrp.ph.param_type);
6403			switch (lptype) {
6404#ifdef INET
6405			case SCTP_IPV4_ADDRESS:
6406				if (plen !=
6407				    sizeof(struct sctp_asconf_addrv4_param)) {
6408					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6409					    (int)sizeof(struct sctp_asconf_addrv4_param),
6410					    plen);
6411				} else {
6412					fii = (struct sctp_asconf_addrv4_param *)fee;
6413					sin.sin_addr.s_addr = fii->addrp.addr;
6414					lsa = (struct sockaddr *)&sin;
6415				}
6416				break;
6417#endif
6418#ifdef INET6
6419			case SCTP_IPV6_ADDRESS:
6420				if (plen !=
6421				    sizeof(struct sctp_asconf_addr_param)) {
6422					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6423					    (int)sizeof(struct sctp_asconf_addr_param),
6424					    plen);
6425				} else {
6426					memcpy(sin6.sin6_addr.s6_addr,
6427					    fee->addrp.addr,
6428					    sizeof(fee->addrp.addr));
6429					lsa = (struct sockaddr *)&sin6;
6430				}
6431				break;
6432#endif
6433			default:
6434				break;
6435			}
6436			if (lsa) {
6437				(void)sctp_set_primary_addr(stcb, sa, NULL);
6438			}
6439		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6440			stcb->asoc.peer_supports_nat = 1;
6441		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6442			/* Peer supports pr-sctp */
6443			peer_supports_prsctp = 1;
6444		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6445			/* A supported extension chunk */
6446			struct sctp_supported_chunk_types_param *pr_supported;
6447			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6448			int num_ent, i;
6449
6450			phdr = sctp_get_next_param(m, offset,
6451			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6452			if (phdr == NULL) {
6453				return (-25);
6454			}
6455			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6456			num_ent = plen - sizeof(struct sctp_paramhdr);
6457			for (i = 0; i < num_ent; i++) {
6458				switch (pr_supported->chunk_types[i]) {
6459				case SCTP_ASCONF:
6460					peer_supports_asconf = 1;
6461					break;
6462				case SCTP_ASCONF_ACK:
6463					peer_supports_asconf_ack = 1;
6464					break;
6465				case SCTP_FORWARD_CUM_TSN:
6466					peer_supports_prsctp = 1;
6467					break;
6468				case SCTP_PACKET_DROPPED:
6469					peer_supports_pktdrop = 1;
6470					break;
6471				case SCTP_NR_SELECTIVE_ACK:
6472					peer_supports_nrsack = 1;
6473					break;
6474				case SCTP_STREAM_RESET:
6475					peer_supports_reconfig = 1;
6476					break;
6477				case SCTP_AUTHENTICATION:
6478					peer_supports_auth = 1;
6479					break;
6480				default:
6481					/* one I have not learned yet */
6482					break;
6483
6484				}
6485			}
6486		} else if (ptype == SCTP_RANDOM) {
6487			if (plen > sizeof(random_store))
6488				break;
6489			if (got_random) {
6490				/* already processed a RANDOM */
6491				goto next_param;
6492			}
6493			phdr = sctp_get_next_param(m, offset,
6494			    (struct sctp_paramhdr *)random_store,
6495			    min(sizeof(random_store), plen));
6496			if (phdr == NULL)
6497				return (-26);
6498			p_random = (struct sctp_auth_random *)phdr;
6499			random_len = plen - sizeof(*p_random);
6500			/* enforce the random length */
6501			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6502				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6503				return (-27);
6504			}
6505			got_random = 1;
6506		} else if (ptype == SCTP_HMAC_LIST) {
6507			uint16_t num_hmacs;
6508			uint16_t i;
6509
6510			if (plen > sizeof(hmacs_store))
6511				break;
6512			if (got_hmacs) {
6513				/* already processed a HMAC list */
6514				goto next_param;
6515			}
6516			phdr = sctp_get_next_param(m, offset,
6517			    (struct sctp_paramhdr *)hmacs_store,
6518			    min(plen, sizeof(hmacs_store)));
6519			if (phdr == NULL)
6520				return (-28);
6521			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6522			hmacs_len = plen - sizeof(*hmacs);
6523			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6524			/* validate the hmac list */
6525			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6526				return (-29);
6527			}
6528			if (stcb->asoc.peer_hmacs != NULL)
6529				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6530			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6531			if (stcb->asoc.peer_hmacs != NULL) {
6532				for (i = 0; i < num_hmacs; i++) {
6533					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6534					    ntohs(hmacs->hmac_ids[i]));
6535				}
6536			}
6537			got_hmacs = 1;
6538		} else if (ptype == SCTP_CHUNK_LIST) {
6539			int i;
6540
6541			if (plen > sizeof(chunks_store))
6542				break;
6543			if (got_chklist) {
6544				/* already processed a Chunks list */
6545				goto next_param;
6546			}
6547			phdr = sctp_get_next_param(m, offset,
6548			    (struct sctp_paramhdr *)chunks_store,
6549			    min(plen, sizeof(chunks_store)));
6550			if (phdr == NULL)
6551				return (-30);
6552			chunks = (struct sctp_auth_chunk_list *)phdr;
6553			num_chunks = plen - sizeof(*chunks);
6554			if (stcb->asoc.peer_auth_chunks != NULL)
6555				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6556			else
6557				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6558			for (i = 0; i < num_chunks; i++) {
6559				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6560				    stcb->asoc.peer_auth_chunks);
6561				/* record asconf/asconf-ack if listed */
6562				if (chunks->chunk_types[i] == SCTP_ASCONF)
6563					saw_asconf = 1;
6564				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6565					saw_asconf_ack = 1;
6566
6567			}
6568			got_chklist = 1;
6569		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6570			    (ptype == SCTP_STATE_COOKIE) ||
6571			    (ptype == SCTP_UNRECOG_PARAM) ||
6572			    (ptype == SCTP_COOKIE_PRESERVE) ||
6573			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6574			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6575			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6576			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6577		    (ptype == SCTP_SUCCESS_REPORT)) {
6578			 /* don't care */ ;
6579		} else {
6580			if ((ptype & 0x8000) == 0x0000) {
6581				/*
6582				 * must stop processing the rest of the
6583				 * param's. Any report bits were handled
6584				 * with the call to
6585				 * sctp_arethere_unrecognized_parameters()
6586				 * when the INIT or INIT-ACK was first seen.
6587				 */
6588				break;
6589			}
6590		}
6591
6592next_param:
6593		offset += SCTP_SIZE32(plen);
6594		if (offset >= limit) {
6595			break;
6596		}
6597		phdr = sctp_get_next_param(m, offset, &parm_buf,
6598		    sizeof(parm_buf));
6599	}
6600	/* Now check to see if we need to purge any addresses */
6601	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6602		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6603		    SCTP_ADDR_NOT_IN_ASSOC) {
6604			/* This address has been removed from the asoc */
6605			/* remove and free it */
6606			stcb->asoc.numnets--;
6607			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6608			sctp_free_remote_addr(net);
6609			if (net == stcb->asoc.primary_destination) {
6610				stcb->asoc.primary_destination = NULL;
6611				sctp_select_primary_destination(stcb);
6612			}
6613		}
6614	}
6615	if ((stcb->asoc.ecn_supported == 1) &&
6616	    (peer_supports_ecn == 0)) {
6617		stcb->asoc.ecn_supported = 0;
6618	}
6619	if ((stcb->asoc.prsctp_supported == 1) &&
6620	    (peer_supports_prsctp == 0)) {
6621		stcb->asoc.prsctp_supported = 0;
6622	}
6623	if ((stcb->asoc.auth_supported == 1) &&
6624	    ((peer_supports_auth == 0) ||
6625	    (got_random == 0) || (got_hmacs == 0))) {
6626		stcb->asoc.auth_supported = 0;
6627	}
6628	if ((stcb->asoc.asconf_supported == 1) &&
6629	    ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) ||
6630	    (stcb->asoc.auth_supported == 0) ||
6631	    (saw_asconf == 0) || (saw_asconf_ack == 0))) {
6632		stcb->asoc.asconf_supported = 0;
6633	}
6634	if ((stcb->asoc.reconfig_supported == 1) &&
6635	    (peer_supports_reconfig == 0)) {
6636		stcb->asoc.reconfig_supported = 0;
6637	}
6638	if ((stcb->asoc.nrsack_supported == 1) &&
6639	    (peer_supports_nrsack == 0)) {
6640		stcb->asoc.nrsack_supported = 0;
6641	}
6642	if ((stcb->asoc.pktdrop_supported == 1) &&
6643	    (peer_supports_pktdrop == 0)) {
6644		stcb->asoc.pktdrop_supported = 0;
6645	}
6646	/* validate authentication required parameters */
6647	if ((peer_supports_auth == 0) && (got_chklist == 1)) {
6648		/* peer does not support auth but sent a chunks list? */
6649		return (-31);
6650	}
6651	if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) {
6652		/* peer supports asconf but not auth? */
6653		return (-32);
6654	} else if ((peer_supports_asconf == 1) &&
6655		    (peer_supports_auth == 1) &&
6656	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6657		return (-33);
6658	}
6659	/* concatenate the full random key */
6660	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6661	if (chunks != NULL) {
6662		keylen += sizeof(*chunks) + num_chunks;
6663	}
6664	new_key = sctp_alloc_key(keylen);
6665	if (new_key != NULL) {
6666		/* copy in the RANDOM */
6667		if (p_random != NULL) {
6668			keylen = sizeof(*p_random) + random_len;
6669			bcopy(p_random, new_key->key, keylen);
6670		}
6671		/* append in the AUTH chunks */
6672		if (chunks != NULL) {
6673			bcopy(chunks, new_key->key + keylen,
6674			    sizeof(*chunks) + num_chunks);
6675			keylen += sizeof(*chunks) + num_chunks;
6676		}
6677		/* append in the HMACs */
6678		if (hmacs != NULL) {
6679			bcopy(hmacs, new_key->key + keylen,
6680			    sizeof(*hmacs) + hmacs_len);
6681		}
6682	} else {
6683		/* failed to get memory for the key */
6684		return (-34);
6685	}
6686	if (stcb->asoc.authinfo.peer_random != NULL)
6687		sctp_free_key(stcb->asoc.authinfo.peer_random);
6688	stcb->asoc.authinfo.peer_random = new_key;
6689	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6690	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6691
6692	return (0);
6693}
6694
6695int
6696sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6697    struct sctp_nets *net)
6698{
6699	/* make sure the requested primary address exists in the assoc */
6700	if (net == NULL && sa)
6701		net = sctp_findnet(stcb, sa);
6702
6703	if (net == NULL) {
6704		/* didn't find the requested primary address! */
6705		return (-1);
6706	} else {
6707		/* set the primary address */
6708		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6709			/* Must be confirmed, so queue to set */
6710			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6711			return (0);
6712		}
6713		stcb->asoc.primary_destination = net;
6714		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6715			sctp_free_remote_addr(stcb->asoc.alternate);
6716			stcb->asoc.alternate = NULL;
6717		}
6718		net = TAILQ_FIRST(&stcb->asoc.nets);
6719		if (net != stcb->asoc.primary_destination) {
6720			/*
6721			 * first one on the list is NOT the primary
6722			 * sctp_cmpaddr() is much more efficient if the
6723			 * primary is the first on the list, make it so.
6724			 */
6725			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6726			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6727		}
6728		return (0);
6729	}
6730}
6731
6732int
6733sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6734{
6735	/*
6736	 * This function serves two purposes. It will see if a TAG can be
6737	 * re-used and return 1 for yes it is ok and 0 for don't use that
6738	 * tag. A secondary function it will do is purge out old tags that
6739	 * can be removed.
6740	 */
6741	struct sctpvtaghead *chain;
6742	struct sctp_tagblock *twait_block;
6743	struct sctpasochead *head;
6744	struct sctp_tcb *stcb;
6745	int i;
6746
6747	SCTP_INP_INFO_RLOCK();
6748	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6749	    SCTP_BASE_INFO(hashasocmark))];
6750	LIST_FOREACH(stcb, head, sctp_asocs) {
6751		/*
6752		 * We choose not to lock anything here. TCB's can't be
6753		 * removed since we have the read lock, so they can't be
6754		 * freed on us, same thing for the INP. I may be wrong with
6755		 * this assumption, but we will go with it for now :-)
6756		 */
6757		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6758			continue;
6759		}
6760		if (stcb->asoc.my_vtag == tag) {
6761			/* candidate */
6762			if (stcb->rport != rport) {
6763				continue;
6764			}
6765			if (stcb->sctp_ep->sctp_lport != lport) {
6766				continue;
6767			}
6768			/* Its a used tag set */
6769			SCTP_INP_INFO_RUNLOCK();
6770			return (0);
6771		}
6772	}
6773	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6774	/* Now what about timed wait ? */
6775	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6776		/*
6777		 * Block(s) are present, lets see if we have this tag in the
6778		 * list
6779		 */
6780		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6781			if (twait_block->vtag_block[i].v_tag == 0) {
6782				/* not used */
6783				continue;
6784			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6785			    now->tv_sec) {
6786				/* Audit expires this guy */
6787				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6788				twait_block->vtag_block[i].v_tag = 0;
6789				twait_block->vtag_block[i].lport = 0;
6790				twait_block->vtag_block[i].rport = 0;
6791			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6792				    (twait_block->vtag_block[i].lport == lport) &&
6793			    (twait_block->vtag_block[i].rport == rport)) {
6794				/* Bad tag, sorry :< */
6795				SCTP_INP_INFO_RUNLOCK();
6796				return (0);
6797			}
6798		}
6799	}
6800	SCTP_INP_INFO_RUNLOCK();
6801	return (1);
6802}
6803
6804static void
6805sctp_drain_mbufs(struct sctp_tcb *stcb)
6806{
6807	/*
6808	 * We must hunt this association for MBUF's past the cumack (i.e.
6809	 * out of order data that we can renege on).
6810	 */
6811	struct sctp_association *asoc;
6812	struct sctp_tmit_chunk *chk, *nchk;
6813	uint32_t cumulative_tsn_p1;
6814	struct sctp_queued_to_read *ctl, *nctl;
6815	int cnt, strmat;
6816	uint32_t gap, i;
6817	int fnd = 0;
6818
6819	/* We look for anything larger than the cum-ack + 1 */
6820
6821	asoc = &stcb->asoc;
6822	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6823		/* none we can reneg on. */
6824		return;
6825	}
6826	SCTP_STAT_INCR(sctps_protocol_drains_done);
6827	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6828	cnt = 0;
6829	/* First look in the re-assembly queue */
6830	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6831		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6832			/* Yep it is above cum-ack */
6833			cnt++;
6834			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6835			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6836			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6837			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6838			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6839			if (chk->data) {
6840				sctp_m_freem(chk->data);
6841				chk->data = NULL;
6842			}
6843			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6844		}
6845	}
6846	/* Ok that was fun, now we will drain all the inbound streams? */
6847	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6848		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6849			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6850				/* Yep it is above cum-ack */
6851				cnt++;
6852				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6853				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6854				sctp_ucount_decr(asoc->cnt_on_all_streams);
6855				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6856				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6857				if (ctl->data) {
6858					sctp_m_freem(ctl->data);
6859					ctl->data = NULL;
6860				}
6861				sctp_free_remote_addr(ctl->whoFrom);
6862				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6863				SCTP_DECR_READQ_COUNT();
6864			}
6865		}
6866	}
6867	if (cnt) {
6868		/* We must back down to see what the new highest is */
6869		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6870			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6871			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6872				asoc->highest_tsn_inside_map = i;
6873				fnd = 1;
6874				break;
6875			}
6876		}
6877		if (!fnd) {
6878			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6879		}
6880		/*
6881		 * Question, should we go through the delivery queue? The
6882		 * only reason things are on here is the app not reading OR
6883		 * a p-d-api up. An attacker COULD send enough in to
6884		 * initiate the PD-API and then send a bunch of stuff to
6885		 * other streams... these would wind up on the delivery
6886		 * queue.. and then we would not get to them. But in order
6887		 * to do this I then have to back-track and un-deliver
6888		 * sequence numbers in streams.. el-yucko. I think for now
6889		 * we will NOT look at the delivery queue and leave it to be
6890		 * something to consider later. An alternative would be to
6891		 * abort the P-D-API with a notification and then deliver
6892		 * the data.... Or another method might be to keep track of
6893		 * how many times the situation occurs and if we see a
6894		 * possible attack underway just abort the association.
6895		 */
6896#ifdef SCTP_DEBUG
6897		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6898#endif
6899		/*
6900		 * Now do we need to find a new
6901		 * asoc->highest_tsn_inside_map?
6902		 */
6903		asoc->last_revoke_count = cnt;
6904		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6905		/* sa_ignore NO_NULL_CHK */
6906		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6907		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6908	}
6909	/*
6910	 * Another issue, in un-setting the TSN's in the mapping array we
6911	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6912	 * two things to occur. It may cause us to do extra work in checking
6913	 * for our mapping array movement. More importantly it may cause us
6914	 * to SACK every datagram. This may not be a bad thing though since
6915	 * we will recover once we get our cum-ack above and all this stuff
6916	 * we dumped recovered.
6917	 */
6918}
6919
6920void
6921sctp_drain()
6922{
6923	/*
6924	 * We must walk the PCB lists for ALL associations here. The system
6925	 * is LOW on MBUF's and needs help. This is where reneging will
6926	 * occur. We really hope this does NOT happen!
6927	 */
6928	VNET_ITERATOR_DECL(vnet_iter);
6929	VNET_LIST_RLOCK_NOSLEEP();
6930	VNET_FOREACH(vnet_iter) {
6931		CURVNET_SET(vnet_iter);
6932		struct sctp_inpcb *inp;
6933		struct sctp_tcb *stcb;
6934
6935		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6936		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6937#ifdef VIMAGE
6938			continue;
6939#else
6940			return;
6941#endif
6942		}
6943		SCTP_INP_INFO_RLOCK();
6944		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6945			/* For each endpoint */
6946			SCTP_INP_RLOCK(inp);
6947			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6948				/* For each association */
6949				SCTP_TCB_LOCK(stcb);
6950				sctp_drain_mbufs(stcb);
6951				SCTP_TCB_UNLOCK(stcb);
6952			}
6953			SCTP_INP_RUNLOCK(inp);
6954		}
6955		SCTP_INP_INFO_RUNLOCK();
6956		CURVNET_RESTORE();
6957	}
6958	VNET_LIST_RUNLOCK_NOSLEEP();
6959}
6960
6961/*
6962 * start a new iterator
6963 * iterates through all endpoints and associations based on the pcb_state
6964 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6965 * assocs and "ef" (optional) is executed when the iterator completes.
6966 * "inpf" (optional) is executed for each new endpoint as it is being
6967 * iterated through. inpe (optional) is called when the inp completes
6968 * its way through all the stcbs.
6969 */
6970int
6971sctp_initiate_iterator(inp_func inpf,
6972    asoc_func af,
6973    inp_func inpe,
6974    uint32_t pcb_state,
6975    uint32_t pcb_features,
6976    uint32_t asoc_state,
6977    void *argp,
6978    uint32_t argi,
6979    end_func ef,
6980    struct sctp_inpcb *s_inp,
6981    uint8_t chunk_output_off)
6982{
6983	struct sctp_iterator *it = NULL;
6984
6985	if (af == NULL) {
6986		return (-1);
6987	}
6988	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6989	    SCTP_M_ITER);
6990	if (it == NULL) {
6991		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6992		return (ENOMEM);
6993	}
6994	memset(it, 0, sizeof(*it));
6995	it->function_assoc = af;
6996	it->function_inp = inpf;
6997	if (inpf)
6998		it->done_current_ep = 0;
6999	else
7000		it->done_current_ep = 1;
7001	it->function_atend = ef;
7002	it->pointer = argp;
7003	it->val = argi;
7004	it->pcb_flags = pcb_state;
7005	it->pcb_features = pcb_features;
7006	it->asoc_state = asoc_state;
7007	it->function_inp_end = inpe;
7008	it->no_chunk_output = chunk_output_off;
7009	it->vn = curvnet;
7010	if (s_inp) {
7011		/* Assume lock is held here */
7012		it->inp = s_inp;
7013		SCTP_INP_INCR_REF(it->inp);
7014		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7015	} else {
7016		SCTP_INP_INFO_RLOCK();
7017		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7018		if (it->inp) {
7019			SCTP_INP_INCR_REF(it->inp);
7020		}
7021		SCTP_INP_INFO_RUNLOCK();
7022		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7023
7024	}
7025	SCTP_IPI_ITERATOR_WQ_LOCK();
7026
7027	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7028	if (sctp_it_ctl.iterator_running == 0) {
7029		sctp_wakeup_iterator();
7030	}
7031	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7032	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7033	return (0);
7034}
7035