sctp_pcb.c revision 270350
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 270350 2014-08-22 19:37:50Z 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 = (struct sockaddr_in *)&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 = (struct sockaddr_in6 *)&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 = (struct sockaddr_in *)&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 = (struct sockaddr_in6 *)&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		if (head == NULL) {
1445			goto null_return;
1446		}
1447		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1448			if (stcb->rport != rport) {
1449				/* remote port does not match */
1450				continue;
1451			}
1452			SCTP_TCB_LOCK(stcb);
1453			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1454				SCTP_TCB_UNLOCK(stcb);
1455				continue;
1456			}
1457			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1458				SCTP_TCB_UNLOCK(stcb);
1459				continue;
1460			}
1461			/* now look at the list of remote addresses */
1462			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1463#ifdef INVARIANTS
1464				if (net == (TAILQ_NEXT(net, sctp_next))) {
1465					panic("Corrupt net list");
1466				}
1467#endif
1468				if (net->ro._l_addr.sa.sa_family !=
1469				    remote->sa_family) {
1470					/* not the same family */
1471					continue;
1472				}
1473				switch (remote->sa_family) {
1474#ifdef INET
1475				case AF_INET:
1476					{
1477						struct sockaddr_in *sin,
1478						           *rsin;
1479
1480						sin = (struct sockaddr_in *)
1481						    &net->ro._l_addr;
1482						rsin = (struct sockaddr_in *)remote;
1483						if (sin->sin_addr.s_addr ==
1484						    rsin->sin_addr.s_addr) {
1485							/* found it */
1486							if (netp != NULL) {
1487								*netp = net;
1488							}
1489							if (locked_tcb == NULL) {
1490								SCTP_INP_DECR_REF(inp);
1491							} else if (locked_tcb != stcb) {
1492								SCTP_TCB_LOCK(locked_tcb);
1493							}
1494							if (locked_tcb) {
1495								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1496							}
1497							SCTP_INP_WUNLOCK(inp);
1498							SCTP_INP_INFO_RUNLOCK();
1499							return (stcb);
1500						}
1501						break;
1502					}
1503#endif
1504#ifdef INET6
1505				case AF_INET6:
1506					{
1507						struct sockaddr_in6 *sin6,
1508						            *rsin6;
1509
1510						sin6 = (struct sockaddr_in6 *)
1511						    &net->ro._l_addr;
1512						rsin6 = (struct sockaddr_in6 *)remote;
1513						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1514						    rsin6)) {
1515							/* found it */
1516							if (netp != NULL) {
1517								*netp = net;
1518							}
1519							if (locked_tcb == NULL) {
1520								SCTP_INP_DECR_REF(inp);
1521							} else if (locked_tcb != stcb) {
1522								SCTP_TCB_LOCK(locked_tcb);
1523							}
1524							if (locked_tcb) {
1525								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1526							}
1527							SCTP_INP_WUNLOCK(inp);
1528							SCTP_INP_INFO_RUNLOCK();
1529							return (stcb);
1530						}
1531						break;
1532					}
1533#endif
1534				default:
1535					/* TSNH */
1536					break;
1537				}
1538			}
1539			SCTP_TCB_UNLOCK(stcb);
1540		}
1541	}
1542null_return:
1543	/* clean up for returning null */
1544	if (locked_tcb) {
1545		SCTP_TCB_LOCK(locked_tcb);
1546		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1547	}
1548	SCTP_INP_WUNLOCK(inp);
1549	SCTP_INP_INFO_RUNLOCK();
1550	/* not found */
1551	return (NULL);
1552}
1553
1554
1555/*
1556 * Find an association for a specific endpoint using the association id given
1557 * out in the COMM_UP notification
1558 */
1559struct sctp_tcb *
1560sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1561{
1562	/*
1563	 * Use my the assoc_id to find a endpoint
1564	 */
1565	struct sctpasochead *head;
1566	struct sctp_tcb *stcb;
1567	uint32_t id;
1568
1569	if (inp == NULL) {
1570		SCTP_PRINTF("TSNH ep_associd\n");
1571		return (NULL);
1572	}
1573	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1574		SCTP_PRINTF("TSNH ep_associd0\n");
1575		return (NULL);
1576	}
1577	id = (uint32_t) asoc_id;
1578	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1579	if (head == NULL) {
1580		/* invalid id TSNH */
1581		SCTP_PRINTF("TSNH ep_associd1\n");
1582		return (NULL);
1583	}
1584	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1585		if (stcb->asoc.assoc_id == id) {
1586			if (inp != stcb->sctp_ep) {
1587				/*
1588				 * some other guy has the same id active (id
1589				 * collision ??).
1590				 */
1591				SCTP_PRINTF("TSNH ep_associd2\n");
1592				continue;
1593			}
1594			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1595				continue;
1596			}
1597			if (want_lock) {
1598				SCTP_TCB_LOCK(stcb);
1599			}
1600			return (stcb);
1601		}
1602	}
1603	return (NULL);
1604}
1605
1606
1607struct sctp_tcb *
1608sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1609{
1610	struct sctp_tcb *stcb;
1611
1612	SCTP_INP_RLOCK(inp);
1613	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1614	SCTP_INP_RUNLOCK(inp);
1615	return (stcb);
1616}
1617
1618
1619/*
1620 * Endpoint probe expects that the INP_INFO is locked.
1621 */
1622static struct sctp_inpcb *
1623sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1624    uint16_t lport, uint32_t vrf_id)
1625{
1626	struct sctp_inpcb *inp;
1627	struct sctp_laddr *laddr;
1628
1629#ifdef INET
1630	struct sockaddr_in *sin;
1631
1632#endif
1633#ifdef INET6
1634	struct sockaddr_in6 *sin6;
1635	struct sockaddr_in6 *intf_addr6;
1636
1637#endif
1638	int fnd;
1639
1640#ifdef INET
1641	sin = NULL;
1642#endif
1643#ifdef INET6
1644	sin6 = NULL;
1645#endif
1646	switch (nam->sa_family) {
1647#ifdef INET
1648	case AF_INET:
1649		sin = (struct sockaddr_in *)nam;
1650		break;
1651#endif
1652#ifdef INET6
1653	case AF_INET6:
1654		sin6 = (struct sockaddr_in6 *)nam;
1655		break;
1656#endif
1657	default:
1658		/* unsupported family */
1659		return (NULL);
1660	}
1661
1662	if (head == NULL)
1663		return (NULL);
1664
1665	LIST_FOREACH(inp, head, sctp_hash) {
1666		SCTP_INP_RLOCK(inp);
1667		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1668			SCTP_INP_RUNLOCK(inp);
1669			continue;
1670		}
1671		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1672		    (inp->sctp_lport == lport)) {
1673			/* got it */
1674			switch (nam->sa_family) {
1675#ifdef INET
1676			case AF_INET:
1677				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1678				    SCTP_IPV6_V6ONLY(inp)) {
1679					/*
1680					 * IPv4 on a IPv6 socket with ONLY
1681					 * IPv6 set
1682					 */
1683					SCTP_INP_RUNLOCK(inp);
1684					continue;
1685				}
1686				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1687				    &sin->sin_addr) != 0) {
1688					SCTP_INP_RUNLOCK(inp);
1689					continue;
1690				}
1691				break;
1692#endif
1693#ifdef INET6
1694			case AF_INET6:
1695				/*
1696				 * A V6 address and the endpoint is NOT
1697				 * bound V6
1698				 */
1699				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1700					SCTP_INP_RUNLOCK(inp);
1701					continue;
1702				}
1703				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1704				    &sin6->sin6_addr) != 0) {
1705					SCTP_INP_RUNLOCK(inp);
1706					continue;
1707				}
1708				break;
1709#endif
1710			default:
1711				break;
1712			}
1713			/* does a VRF id match? */
1714			fnd = 0;
1715			if (inp->def_vrf_id == vrf_id)
1716				fnd = 1;
1717
1718			SCTP_INP_RUNLOCK(inp);
1719			if (!fnd)
1720				continue;
1721			return (inp);
1722		}
1723		SCTP_INP_RUNLOCK(inp);
1724	}
1725	switch (nam->sa_family) {
1726#ifdef INET
1727	case AF_INET:
1728		if (sin->sin_addr.s_addr == INADDR_ANY) {
1729			/* Can't hunt for one that has no address specified */
1730			return (NULL);
1731		}
1732		break;
1733#endif
1734#ifdef INET6
1735	case AF_INET6:
1736		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1737			/* Can't hunt for one that has no address specified */
1738			return (NULL);
1739		}
1740		break;
1741#endif
1742	default:
1743		break;
1744	}
1745	/*
1746	 * ok, not bound to all so see if we can find a EP bound to this
1747	 * address.
1748	 */
1749	LIST_FOREACH(inp, head, sctp_hash) {
1750		SCTP_INP_RLOCK(inp);
1751		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1752			SCTP_INP_RUNLOCK(inp);
1753			continue;
1754		}
1755		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1756			SCTP_INP_RUNLOCK(inp);
1757			continue;
1758		}
1759		/*
1760		 * Ok this could be a likely candidate, look at all of its
1761		 * addresses
1762		 */
1763		if (inp->sctp_lport != lport) {
1764			SCTP_INP_RUNLOCK(inp);
1765			continue;
1766		}
1767		/* does a VRF id match? */
1768		fnd = 0;
1769		if (inp->def_vrf_id == vrf_id)
1770			fnd = 1;
1771
1772		if (!fnd) {
1773			SCTP_INP_RUNLOCK(inp);
1774			continue;
1775		}
1776		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1777			if (laddr->ifa == NULL) {
1778				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1779				    __FUNCTION__);
1780				continue;
1781			}
1782			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1783			    (void *)laddr->ifa);
1784			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1785				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1786				continue;
1787			}
1788			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1789				/* possible, see if it matches */
1790				switch (nam->sa_family) {
1791#ifdef INET
1792				case AF_INET:
1793					if (sin->sin_addr.s_addr ==
1794					    laddr->ifa->address.sin.sin_addr.s_addr) {
1795						SCTP_INP_RUNLOCK(inp);
1796						return (inp);
1797					}
1798					break;
1799#endif
1800#ifdef INET6
1801				case AF_INET6:
1802					intf_addr6 = &laddr->ifa->address.sin6;
1803					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1804					    intf_addr6)) {
1805						SCTP_INP_RUNLOCK(inp);
1806						return (inp);
1807					}
1808					break;
1809#endif
1810				}
1811			}
1812		}
1813		SCTP_INP_RUNLOCK(inp);
1814	}
1815	return (NULL);
1816}
1817
1818
1819static struct sctp_inpcb *
1820sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1821{
1822	struct sctppcbhead *head;
1823	struct sctp_inpcb *t_inp;
1824	int fnd;
1825
1826	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1827	    SCTP_BASE_INFO(hashmark))];
1828	LIST_FOREACH(t_inp, head, sctp_hash) {
1829		if (t_inp->sctp_lport != lport) {
1830			continue;
1831		}
1832		/* is it in the VRF in question */
1833		fnd = 0;
1834		if (t_inp->def_vrf_id == vrf_id)
1835			fnd = 1;
1836		if (!fnd)
1837			continue;
1838
1839		/* This one is in use. */
1840		/* check the v6/v4 binding issue */
1841		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1842		    SCTP_IPV6_V6ONLY(t_inp)) {
1843			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1844				/* collision in V6 space */
1845				return (t_inp);
1846			} else {
1847				/* inp is BOUND_V4 no conflict */
1848				continue;
1849			}
1850		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1851			/* t_inp is bound v4 and v6, conflict always */
1852			return (t_inp);
1853		} else {
1854			/* t_inp is bound only V4 */
1855			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1856			    SCTP_IPV6_V6ONLY(inp)) {
1857				/* no conflict */
1858				continue;
1859			}
1860			/* else fall through to conflict */
1861		}
1862		return (t_inp);
1863	}
1864	return (NULL);
1865}
1866
1867
1868int
1869sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1870{
1871	/* For 1-2-1 with port reuse */
1872	struct sctppcbhead *head;
1873	struct sctp_inpcb *tinp;
1874
1875	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1876		/* only works with port reuse on */
1877		return (-1);
1878	}
1879	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1880		return (0);
1881	}
1882	SCTP_INP_RUNLOCK(inp);
1883	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1884	    SCTP_BASE_INFO(hashmark))];
1885	/* Kick out all non-listeners to the TCP hash */
1886	LIST_FOREACH(tinp, head, sctp_hash) {
1887		if (tinp->sctp_lport != inp->sctp_lport) {
1888			continue;
1889		}
1890		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1891			continue;
1892		}
1893		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1894			continue;
1895		}
1896		if (tinp->sctp_socket->so_qlimit) {
1897			continue;
1898		}
1899		SCTP_INP_WLOCK(tinp);
1900		LIST_REMOVE(tinp, sctp_hash);
1901		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1902		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1903		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1904		SCTP_INP_WUNLOCK(tinp);
1905	}
1906	SCTP_INP_WLOCK(inp);
1907	/* Pull from where he was */
1908	LIST_REMOVE(inp, sctp_hash);
1909	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1910	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1911	LIST_INSERT_HEAD(head, inp, sctp_hash);
1912	SCTP_INP_WUNLOCK(inp);
1913	SCTP_INP_RLOCK(inp);
1914	return (0);
1915}
1916
1917
1918struct sctp_inpcb *
1919sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1920    uint32_t vrf_id)
1921{
1922	/*
1923	 * First we check the hash table to see if someone has this port
1924	 * bound with just the port.
1925	 */
1926	struct sctp_inpcb *inp;
1927	struct sctppcbhead *head;
1928	int lport;
1929	unsigned int i;
1930
1931#ifdef INET
1932	struct sockaddr_in *sin;
1933
1934#endif
1935#ifdef INET6
1936	struct sockaddr_in6 *sin6;
1937
1938#endif
1939
1940	switch (nam->sa_family) {
1941#ifdef INET
1942	case AF_INET:
1943		sin = (struct sockaddr_in *)nam;
1944		lport = sin->sin_port;
1945		break;
1946#endif
1947#ifdef INET6
1948	case AF_INET6:
1949		sin6 = (struct sockaddr_in6 *)nam;
1950		lport = sin6->sin6_port;
1951		break;
1952#endif
1953	default:
1954		return (NULL);
1955	}
1956	/*
1957	 * I could cheat here and just cast to one of the types but we will
1958	 * do it right. It also provides the check against an Unsupported
1959	 * type too.
1960	 */
1961	/* Find the head of the ALLADDR chain */
1962	if (have_lock == 0) {
1963		SCTP_INP_INFO_RLOCK();
1964	}
1965	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1966	    SCTP_BASE_INFO(hashmark))];
1967	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1968
1969	/*
1970	 * If the TCP model exists it could be that the main listening
1971	 * endpoint is gone but there still exists a connected socket for
1972	 * this guy. If so we can return the first one that we find. This
1973	 * may NOT be the correct one so the caller should be wary on the
1974	 * returned INP. Currently the only caller that sets find_tcp_pool
1975	 * is in bindx where we are verifying that a user CAN bind the
1976	 * address. He either has bound it already, or someone else has, or
1977	 * its open to bind, so this is good enough.
1978	 */
1979	if (inp == NULL && find_tcp_pool) {
1980		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1981			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1982			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1983			if (inp) {
1984				break;
1985			}
1986		}
1987	}
1988	if (inp) {
1989		SCTP_INP_INCR_REF(inp);
1990	}
1991	if (have_lock == 0) {
1992		SCTP_INP_INFO_RUNLOCK();
1993	}
1994	return (inp);
1995}
1996
1997
1998/*
1999 * Find an association for an endpoint with the pointer to whom you want to
2000 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
2001 * need to change the *to to some other struct like a mbuf...
2002 */
2003struct sctp_tcb *
2004sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
2005    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
2006    uint32_t vrf_id)
2007{
2008	struct sctp_inpcb *inp = NULL;
2009	struct sctp_tcb *stcb;
2010
2011	SCTP_INP_INFO_RLOCK();
2012	if (find_tcp_pool) {
2013		if (inp_p != NULL) {
2014			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
2015			    vrf_id);
2016		} else {
2017			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
2018			    vrf_id);
2019		}
2020		if (stcb != NULL) {
2021			SCTP_INP_INFO_RUNLOCK();
2022			return (stcb);
2023		}
2024	}
2025	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
2026	if (inp_p != NULL) {
2027		*inp_p = inp;
2028	}
2029	SCTP_INP_INFO_RUNLOCK();
2030	if (inp == NULL) {
2031		return (NULL);
2032	}
2033	/*
2034	 * ok, we have an endpoint, now lets find the assoc for it (if any)
2035	 * we now place the source address or from in the to of the find
2036	 * endpoint call. Since in reality this chain is used from the
2037	 * inbound packet side.
2038	 */
2039	if (inp_p != NULL) {
2040		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
2041		    NULL);
2042	} else {
2043		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
2044		    NULL);
2045	}
2046	return (stcb);
2047}
2048
2049
2050/*
2051 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
2052 * find all addresses that the sender has specified in any address list. Each
2053 * address will be used to lookup the TCB and see if one exits.
2054 */
2055static struct sctp_tcb *
2056sctp_findassociation_special_addr(struct mbuf *m, int offset,
2057    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
2058    struct sockaddr *dst)
2059{
2060	struct sctp_paramhdr *phdr, parm_buf;
2061
2062#if defined(INET) || defined(INET6)
2063	struct sctp_tcb *stcb;
2064	uint16_t ptype;
2065
2066#endif
2067	uint16_t plen;
2068
2069#ifdef INET
2070	struct sockaddr_in sin4;
2071
2072#endif
2073#ifdef INET6
2074	struct sockaddr_in6 sin6;
2075
2076#endif
2077
2078#ifdef INET
2079	memset(&sin4, 0, sizeof(sin4));
2080	sin4.sin_len = sizeof(sin4);
2081	sin4.sin_family = AF_INET;
2082	sin4.sin_port = sh->src_port;
2083#endif
2084#ifdef INET6
2085	memset(&sin6, 0, sizeof(sin6));
2086	sin6.sin6_len = sizeof(sin6);
2087	sin6.sin6_family = AF_INET6;
2088	sin6.sin6_port = sh->src_port;
2089#endif
2090
2091	offset += sizeof(struct sctp_init_chunk);
2092
2093	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2094	while (phdr != NULL) {
2095		/* now we must see if we want the parameter */
2096#if defined(INET) || defined(INET6)
2097		ptype = ntohs(phdr->param_type);
2098#endif
2099		plen = ntohs(phdr->param_length);
2100		if (plen == 0) {
2101			break;
2102		}
2103#ifdef INET
2104		if (ptype == SCTP_IPV4_ADDRESS &&
2105		    plen == sizeof(struct sctp_ipv4addr_param)) {
2106			/* Get the rest of the address */
2107			struct sctp_ipv4addr_param ip4_parm, *p4;
2108
2109			phdr = sctp_get_next_param(m, offset,
2110			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2111			if (phdr == NULL) {
2112				return (NULL);
2113			}
2114			p4 = (struct sctp_ipv4addr_param *)phdr;
2115			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2116			/* look it up */
2117			stcb = sctp_findassociation_ep_addr(inp_p,
2118			    (struct sockaddr *)&sin4, netp, dst, NULL);
2119			if (stcb != NULL) {
2120				return (stcb);
2121			}
2122		}
2123#endif
2124#ifdef INET6
2125		if (ptype == SCTP_IPV6_ADDRESS &&
2126		    plen == sizeof(struct sctp_ipv6addr_param)) {
2127			/* Get the rest of the address */
2128			struct sctp_ipv6addr_param ip6_parm, *p6;
2129
2130			phdr = sctp_get_next_param(m, offset,
2131			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2132			if (phdr == NULL) {
2133				return (NULL);
2134			}
2135			p6 = (struct sctp_ipv6addr_param *)phdr;
2136			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2137			/* look it up */
2138			stcb = sctp_findassociation_ep_addr(inp_p,
2139			    (struct sockaddr *)&sin6, netp, dst, NULL);
2140			if (stcb != NULL) {
2141				return (stcb);
2142			}
2143		}
2144#endif
2145		offset += SCTP_SIZE32(plen);
2146		phdr = sctp_get_next_param(m, offset, &parm_buf,
2147		    sizeof(parm_buf));
2148	}
2149	return (NULL);
2150}
2151
2152static struct sctp_tcb *
2153sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2154    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2155    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2156{
2157	/*
2158	 * Use my vtag to hash. If we find it we then verify the source addr
2159	 * is in the assoc. If all goes well we save a bit on rec of a
2160	 * packet.
2161	 */
2162	struct sctpasochead *head;
2163	struct sctp_nets *net;
2164	struct sctp_tcb *stcb;
2165
2166	SCTP_INP_INFO_RLOCK();
2167	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2168	    SCTP_BASE_INFO(hashasocmark))];
2169	if (head == NULL) {
2170		/* invalid vtag */
2171		SCTP_INP_INFO_RUNLOCK();
2172		return (NULL);
2173	}
2174	LIST_FOREACH(stcb, head, sctp_asocs) {
2175		SCTP_INP_RLOCK(stcb->sctp_ep);
2176		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2177			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2178			continue;
2179		}
2180		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2181			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2182			continue;
2183		}
2184		SCTP_TCB_LOCK(stcb);
2185		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2186		if (stcb->asoc.my_vtag == vtag) {
2187			/* candidate */
2188			if (stcb->rport != rport) {
2189				SCTP_TCB_UNLOCK(stcb);
2190				continue;
2191			}
2192			if (stcb->sctp_ep->sctp_lport != lport) {
2193				SCTP_TCB_UNLOCK(stcb);
2194				continue;
2195			}
2196			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2197				SCTP_TCB_UNLOCK(stcb);
2198				continue;
2199			}
2200			/* RRS:Need toaddr check here */
2201			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2202				/* Endpoint does not own this address */
2203				SCTP_TCB_UNLOCK(stcb);
2204				continue;
2205			}
2206			if (remote_tag) {
2207				/*
2208				 * If we have both vtags that's all we match
2209				 * on
2210				 */
2211				if (stcb->asoc.peer_vtag == remote_tag) {
2212					/*
2213					 * If both tags match we consider it
2214					 * conclusive and check NO
2215					 * source/destination addresses
2216					 */
2217					goto conclusive;
2218				}
2219			}
2220			if (skip_src_check) {
2221		conclusive:
2222				if (from) {
2223					*netp = sctp_findnet(stcb, from);
2224				} else {
2225					*netp = NULL;	/* unknown */
2226				}
2227				if (inp_p)
2228					*inp_p = stcb->sctp_ep;
2229				SCTP_INP_INFO_RUNLOCK();
2230				return (stcb);
2231			}
2232			net = sctp_findnet(stcb, from);
2233			if (net) {
2234				/* yep its him. */
2235				*netp = net;
2236				SCTP_STAT_INCR(sctps_vtagexpress);
2237				*inp_p = stcb->sctp_ep;
2238				SCTP_INP_INFO_RUNLOCK();
2239				return (stcb);
2240			} else {
2241				/*
2242				 * not him, this should only happen in rare
2243				 * cases so I peg it.
2244				 */
2245				SCTP_STAT_INCR(sctps_vtagbogus);
2246			}
2247		}
2248		SCTP_TCB_UNLOCK(stcb);
2249	}
2250	SCTP_INP_INFO_RUNLOCK();
2251	return (NULL);
2252}
2253
2254
2255/*
2256 * Find an association with the pointer to the inbound IP packet. This can be
2257 * a IPv4 or IPv6 packet.
2258 */
2259struct sctp_tcb *
2260sctp_findassociation_addr(struct mbuf *m, int offset,
2261    struct sockaddr *src, struct sockaddr *dst,
2262    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2263    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2264{
2265	int find_tcp_pool;
2266	struct sctp_tcb *stcb;
2267	struct sctp_inpcb *inp;
2268
2269	if (sh->v_tag) {
2270		/* we only go down this path if vtag is non-zero */
2271		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2272		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2273		if (stcb) {
2274			return (stcb);
2275		}
2276	}
2277	find_tcp_pool = 0;
2278	if ((ch->chunk_type != SCTP_INITIATION) &&
2279	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2280	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2281	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2282		/* Other chunk types go to the tcp pool. */
2283		find_tcp_pool = 1;
2284	}
2285	if (inp_p) {
2286		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2287		    find_tcp_pool, vrf_id);
2288		inp = *inp_p;
2289	} else {
2290		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2291		    find_tcp_pool, vrf_id);
2292	}
2293	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2294	if (stcb == NULL && inp) {
2295		/* Found a EP but not this address */
2296		if ((ch->chunk_type == SCTP_INITIATION) ||
2297		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2298			/*-
2299			 * special hook, we do NOT return linp or an
2300			 * association that is linked to an existing
2301			 * association that is under the TCP pool (i.e. no
2302			 * listener exists). The endpoint finding routine
2303			 * will always find a listener before examining the
2304			 * TCP pool.
2305			 */
2306			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2307				if (inp_p) {
2308					*inp_p = NULL;
2309				}
2310				return (NULL);
2311			}
2312			stcb = sctp_findassociation_special_addr(m,
2313			    offset, sh, &inp, netp, dst);
2314			if (inp_p != NULL) {
2315				*inp_p = inp;
2316			}
2317		}
2318	}
2319	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2320	return (stcb);
2321}
2322
2323/*
2324 * lookup an association by an ASCONF lookup address.
2325 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2326 */
2327struct sctp_tcb *
2328sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2329    struct sockaddr *dst, struct sctphdr *sh,
2330    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2331{
2332	struct sctp_tcb *stcb;
2333	struct sockaddr_storage remote_store;
2334	struct sctp_paramhdr parm_buf, *phdr;
2335	int ptype;
2336	int zero_address = 0;
2337
2338#ifdef INET
2339	struct sockaddr_in *sin;
2340
2341#endif
2342#ifdef INET6
2343	struct sockaddr_in6 *sin6;
2344
2345#endif
2346
2347	memset(&remote_store, 0, sizeof(remote_store));
2348	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2349	    &parm_buf, sizeof(struct sctp_paramhdr));
2350	if (phdr == NULL) {
2351		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2352		    __FUNCTION__);
2353		return NULL;
2354	}
2355	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2356	/* get the correlation address */
2357	switch (ptype) {
2358#ifdef INET6
2359	case SCTP_IPV6_ADDRESS:
2360		{
2361			/* ipv6 address param */
2362			struct sctp_ipv6addr_param *p6, p6_buf;
2363
2364			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2365				return NULL;
2366			}
2367			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2368			    offset + sizeof(struct sctp_asconf_chunk),
2369			    &p6_buf.ph, sizeof(*p6));
2370			if (p6 == NULL) {
2371				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2372				    __FUNCTION__);
2373				return (NULL);
2374			}
2375			sin6 = (struct sockaddr_in6 *)&remote_store;
2376			sin6->sin6_family = AF_INET6;
2377			sin6->sin6_len = sizeof(*sin6);
2378			sin6->sin6_port = sh->src_port;
2379			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2380			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2381				zero_address = 1;
2382			break;
2383		}
2384#endif
2385#ifdef INET
2386	case SCTP_IPV4_ADDRESS:
2387		{
2388			/* ipv4 address param */
2389			struct sctp_ipv4addr_param *p4, p4_buf;
2390
2391			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2392				return NULL;
2393			}
2394			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2395			    offset + sizeof(struct sctp_asconf_chunk),
2396			    &p4_buf.ph, sizeof(*p4));
2397			if (p4 == NULL) {
2398				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2399				    __FUNCTION__);
2400				return (NULL);
2401			}
2402			sin = (struct sockaddr_in *)&remote_store;
2403			sin->sin_family = AF_INET;
2404			sin->sin_len = sizeof(*sin);
2405			sin->sin_port = sh->src_port;
2406			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2407			if (sin->sin_addr.s_addr == INADDR_ANY)
2408				zero_address = 1;
2409			break;
2410		}
2411#endif
2412	default:
2413		/* invalid address param type */
2414		return NULL;
2415	}
2416
2417	if (zero_address) {
2418		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2419		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2420		if (stcb != NULL) {
2421			SCTP_INP_DECR_REF(*inp_p);
2422		}
2423	} else {
2424		stcb = sctp_findassociation_ep_addr(inp_p,
2425		    (struct sockaddr *)&remote_store, netp,
2426		    dst, NULL);
2427	}
2428	return (stcb);
2429}
2430
2431
2432/*
2433 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2434 * addresses. This way if we don't get a bind we by default pick a ephemeral
2435 * port with all addresses bound.
2436 */
2437int
2438sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2439{
2440	/*
2441	 * we get called when a new endpoint starts up. We need to allocate
2442	 * the sctp_inpcb structure from the zone and init it. Mark it as
2443	 * unbound and find a port that we can use as an ephemeral with
2444	 * INADDR_ANY. If the user binds later no problem we can then add in
2445	 * the specific addresses. And setup the default parameters for the
2446	 * EP.
2447	 */
2448	int i, error;
2449	struct sctp_inpcb *inp;
2450	struct sctp_pcb *m;
2451	struct timeval time;
2452	sctp_sharedkey_t *null_key;
2453
2454	error = 0;
2455
2456	SCTP_INP_INFO_WLOCK();
2457	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2458	if (inp == NULL) {
2459		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2460		SCTP_INP_INFO_WUNLOCK();
2461		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2462		return (ENOBUFS);
2463	}
2464	/* zap it */
2465	bzero(inp, sizeof(*inp));
2466
2467	/* bump generations */
2468	/* setup socket pointers */
2469	inp->sctp_socket = so;
2470	inp->ip_inp.inp.inp_socket = so;
2471	inp->ip_inp.inp.inp_cred = crhold(so->so_cred);
2472#ifdef INET6
2473	if (INP_SOCKAF(so) == AF_INET6) {
2474		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2475			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2476		}
2477		if (MODULE_GLOBAL(ip6_v6only)) {
2478			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2479		}
2480	}
2481#endif
2482	inp->sctp_associd_counter = 1;
2483	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2484	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2485	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2486	inp->sctp_ecn_enable = SCTP_BASE_SYSCTL(sctp_ecn_enable);
2487	/* init the small hash table we use to track asocid <-> tcb */
2488	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2489	if (inp->sctp_asocidhash == NULL) {
2490		crfree(inp->ip_inp.inp.inp_cred);
2491		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2492		SCTP_INP_INFO_WUNLOCK();
2493		return (ENOBUFS);
2494	}
2495#ifdef IPSEC
2496	{
2497		struct inpcbpolicy *pcb_sp = NULL;
2498
2499		error = ipsec_init_policy(so, &pcb_sp);
2500		/* Arrange to share the policy */
2501		inp->ip_inp.inp.inp_sp = pcb_sp;
2502		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2503	}
2504	if (error != 0) {
2505		crfree(inp->ip_inp.inp.inp_cred);
2506		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2507		SCTP_INP_INFO_WUNLOCK();
2508		return error;
2509	}
2510#endif				/* IPSEC */
2511	SCTP_INCR_EP_COUNT();
2512	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2513	SCTP_INP_INFO_WUNLOCK();
2514
2515	so->so_pcb = (caddr_t)inp;
2516
2517	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2518		/* UDP style socket */
2519		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2520		    SCTP_PCB_FLAGS_UNBOUND);
2521		/* Be sure it is NON-BLOCKING IO for UDP */
2522		/* SCTP_SET_SO_NBIO(so); */
2523	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2524		/* TCP style socket */
2525		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2526		    SCTP_PCB_FLAGS_UNBOUND);
2527		/* Be sure we have blocking IO by default */
2528		SCTP_CLEAR_SO_NBIO(so);
2529	} else {
2530		/*
2531		 * unsupported socket type (RAW, etc)- in case we missed it
2532		 * in protosw
2533		 */
2534		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2535		so->so_pcb = NULL;
2536		crfree(inp->ip_inp.inp.inp_cred);
2537		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2538		return (EOPNOTSUPP);
2539	}
2540	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2541		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2542		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2543	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2544		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2545		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2546	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2547		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2548		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2549	}
2550	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2551	    &inp->sctp_hashmark);
2552	if (inp->sctp_tcbhash == NULL) {
2553		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2554		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2555		so->so_pcb = NULL;
2556		crfree(inp->ip_inp.inp.inp_cred);
2557		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2558		return (ENOBUFS);
2559	}
2560	inp->def_vrf_id = vrf_id;
2561
2562	SCTP_INP_INFO_WLOCK();
2563	SCTP_INP_LOCK_INIT(inp);
2564	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2565	SCTP_INP_READ_INIT(inp);
2566	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2567	/* lock the new ep */
2568	SCTP_INP_WLOCK(inp);
2569
2570	/* add it to the info area */
2571	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2572	SCTP_INP_INFO_WUNLOCK();
2573
2574	TAILQ_INIT(&inp->read_queue);
2575	LIST_INIT(&inp->sctp_addr_list);
2576
2577	LIST_INIT(&inp->sctp_asoc_list);
2578
2579#ifdef SCTP_TRACK_FREED_ASOCS
2580	/* TEMP CODE */
2581	LIST_INIT(&inp->sctp_asoc_free_list);
2582#endif
2583	/* Init the timer structure for signature change */
2584	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2585	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2586
2587	/* now init the actual endpoint default data */
2588	m = &inp->sctp_ep;
2589
2590	/* setup the base timeout information */
2591	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2592	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2593	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2594	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2595	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2596	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2597	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2598	/* all max/min max are in ms */
2599	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2600	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2601	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2602	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2603	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2604	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2605	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2606	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2607	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2608	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2609	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2610	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2611	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2612
2613	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2614	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2615	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2616	/* number of streams to pre-open on a association */
2617	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2618
2619	/* Add adaptation cookie */
2620	m->adaptation_layer_indicator = 0;
2621	m->adaptation_layer_indicator_provided = 0;
2622
2623	/* seed random number generator */
2624	m->random_counter = 1;
2625	m->store_at = SCTP_SIGNATURE_SIZE;
2626	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2627	sctp_fill_random_store(m);
2628
2629	/* Minimum cookie size */
2630	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2631	    sizeof(struct sctp_state_cookie);
2632	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2633
2634	/* Setup the initial secret */
2635	(void)SCTP_GETTIME_TIMEVAL(&time);
2636	m->time_of_secret_change = time.tv_sec;
2637
2638	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2639		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2640	}
2641	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2642
2643	/* How long is a cookie good for ? */
2644	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2645	/*
2646	 * Initialize authentication parameters
2647	 */
2648	m->local_hmacs = sctp_default_supported_hmaclist();
2649	m->local_auth_chunks = sctp_alloc_chunklist();
2650	m->default_dscp = 0;
2651#ifdef INET6
2652	m->default_flowlabel = 0;
2653#endif
2654	m->port = 0;		/* encapsulation disabled by default */
2655	sctp_auth_set_default_chunks(m->local_auth_chunks);
2656	LIST_INIT(&m->shared_keys);
2657	/* add default NULL key as key id 0 */
2658	null_key = sctp_alloc_sharedkey();
2659	sctp_insert_sharedkey(&m->shared_keys, null_key);
2660	SCTP_INP_WUNLOCK(inp);
2661#ifdef SCTP_LOG_CLOSING
2662	sctp_log_closing(inp, NULL, 12);
2663#endif
2664	return (error);
2665}
2666
2667
2668void
2669sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2670    struct sctp_tcb *stcb)
2671{
2672	struct sctp_nets *net;
2673	uint16_t lport, rport;
2674	struct sctppcbhead *head;
2675	struct sctp_laddr *laddr, *oladdr;
2676
2677	atomic_add_int(&stcb->asoc.refcnt, 1);
2678	SCTP_TCB_UNLOCK(stcb);
2679	SCTP_INP_INFO_WLOCK();
2680	SCTP_INP_WLOCK(old_inp);
2681	SCTP_INP_WLOCK(new_inp);
2682	SCTP_TCB_LOCK(stcb);
2683	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2684
2685	new_inp->sctp_ep.time_of_secret_change =
2686	    old_inp->sctp_ep.time_of_secret_change;
2687	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2688	    sizeof(old_inp->sctp_ep.secret_key));
2689	new_inp->sctp_ep.current_secret_number =
2690	    old_inp->sctp_ep.current_secret_number;
2691	new_inp->sctp_ep.last_secret_number =
2692	    old_inp->sctp_ep.last_secret_number;
2693	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2694
2695	/* make it so new data pours into the new socket */
2696	stcb->sctp_socket = new_inp->sctp_socket;
2697	stcb->sctp_ep = new_inp;
2698
2699	/* Copy the port across */
2700	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2701	rport = stcb->rport;
2702	/* Pull the tcb from the old association */
2703	LIST_REMOVE(stcb, sctp_tcbhash);
2704	LIST_REMOVE(stcb, sctp_tcblist);
2705	if (stcb->asoc.in_asocid_hash) {
2706		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2707	}
2708	/* Now insert the new_inp into the TCP connected hash */
2709	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2710
2711	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2712	/* Its safe to access */
2713	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2714
2715	/* Now move the tcb into the endpoint list */
2716	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2717	/*
2718	 * Question, do we even need to worry about the ep-hash since we
2719	 * only have one connection? Probably not :> so lets get rid of it
2720	 * and not suck up any kernel memory in that.
2721	 */
2722	if (stcb->asoc.in_asocid_hash) {
2723		struct sctpasochead *lhd;
2724
2725		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2726		    new_inp->hashasocidmark)];
2727		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2728	}
2729	/* Ok. Let's restart timer. */
2730	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2731		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2732		    stcb, net);
2733	}
2734
2735	SCTP_INP_INFO_WUNLOCK();
2736	if (new_inp->sctp_tcbhash != NULL) {
2737		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2738		new_inp->sctp_tcbhash = NULL;
2739	}
2740	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2741		/* Subset bound, so copy in the laddr list from the old_inp */
2742		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2743			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2744			if (laddr == NULL) {
2745				/*
2746				 * Gak, what can we do? This assoc is really
2747				 * HOSED. We probably should send an abort
2748				 * here.
2749				 */
2750				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2751				continue;
2752			}
2753			SCTP_INCR_LADDR_COUNT();
2754			bzero(laddr, sizeof(*laddr));
2755			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2756			laddr->ifa = oladdr->ifa;
2757			atomic_add_int(&laddr->ifa->refcount, 1);
2758			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2759			    sctp_nxt_addr);
2760			new_inp->laddr_count++;
2761			if (oladdr == stcb->asoc.last_used_address) {
2762				stcb->asoc.last_used_address = laddr;
2763			}
2764		}
2765	}
2766	/*
2767	 * Now any running timers need to be adjusted since we really don't
2768	 * care if they are running or not just blast in the new_inp into
2769	 * all of them.
2770	 */
2771
2772	stcb->asoc.dack_timer.ep = (void *)new_inp;
2773	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2774	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2775	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2776	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2777	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2778	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2779	/* now what about the nets? */
2780	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2781		net->pmtu_timer.ep = (void *)new_inp;
2782		net->hb_timer.ep = (void *)new_inp;
2783		net->rxt_timer.ep = (void *)new_inp;
2784	}
2785	SCTP_INP_WUNLOCK(new_inp);
2786	SCTP_INP_WUNLOCK(old_inp);
2787}
2788
2789
2790
2791
2792/* sctp_ifap is used to bypass normal local address validation checks */
2793int
2794sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2795    struct sctp_ifa *sctp_ifap, struct thread *p)
2796{
2797	/* bind a ep to a socket address */
2798	struct sctppcbhead *head;
2799	struct sctp_inpcb *inp, *inp_tmp;
2800	struct inpcb *ip_inp;
2801	int port_reuse_active = 0;
2802	int bindall;
2803	uint16_t lport;
2804	int error;
2805	uint32_t vrf_id;
2806
2807	lport = 0;
2808	bindall = 1;
2809	inp = (struct sctp_inpcb *)so->so_pcb;
2810	ip_inp = (struct inpcb *)so->so_pcb;
2811#ifdef SCTP_DEBUG
2812	if (addr) {
2813		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2814		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2815		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2816		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2817	}
2818#endif
2819	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2820		/* already did a bind, subsequent binds NOT allowed ! */
2821		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2822		return (EINVAL);
2823	}
2824#ifdef INVARIANTS
2825	if (p == NULL)
2826		panic("null proc/thread");
2827#endif
2828	if (addr != NULL) {
2829		switch (addr->sa_family) {
2830#ifdef INET
2831		case AF_INET:
2832			{
2833				struct sockaddr_in *sin;
2834
2835				/* IPV6_V6ONLY socket? */
2836				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2837					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2838					return (EINVAL);
2839				}
2840				if (addr->sa_len != sizeof(*sin)) {
2841					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2842					return (EINVAL);
2843				}
2844				sin = (struct sockaddr_in *)addr;
2845				lport = sin->sin_port;
2846				/*
2847				 * For LOOPBACK the prison_local_ip4() call
2848				 * will transmute the ip address to the
2849				 * proper value.
2850				 */
2851				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2852					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2853					return (error);
2854				}
2855				if (sin->sin_addr.s_addr != INADDR_ANY) {
2856					bindall = 0;
2857				}
2858				break;
2859			}
2860#endif
2861#ifdef INET6
2862		case AF_INET6:
2863			{
2864				/*
2865				 * Only for pure IPv6 Address. (No IPv4
2866				 * Mapped!)
2867				 */
2868				struct sockaddr_in6 *sin6;
2869
2870				sin6 = (struct sockaddr_in6 *)addr;
2871
2872				if (addr->sa_len != sizeof(*sin6)) {
2873					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2874					return (EINVAL);
2875				}
2876				lport = sin6->sin6_port;
2877				/*
2878				 * For LOOPBACK the prison_local_ip6() call
2879				 * will transmute the ipv6 address to the
2880				 * proper value.
2881				 */
2882				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2883				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2884					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2885					return (error);
2886				}
2887				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2888					bindall = 0;
2889					/* KAME hack: embed scopeid */
2890					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2891						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2892						return (EINVAL);
2893					}
2894				}
2895				/* this must be cleared for ifa_ifwithaddr() */
2896				sin6->sin6_scope_id = 0;
2897				break;
2898			}
2899#endif
2900		default:
2901			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2902			return (EAFNOSUPPORT);
2903		}
2904	}
2905	SCTP_INP_INFO_WLOCK();
2906	SCTP_INP_WLOCK(inp);
2907	/* Setup a vrf_id to be the default for the non-bind-all case. */
2908	vrf_id = inp->def_vrf_id;
2909
2910	/* increase our count due to the unlock we do */
2911	SCTP_INP_INCR_REF(inp);
2912	if (lport) {
2913		/*
2914		 * Did the caller specify a port? if so we must see if an ep
2915		 * already has this one bound.
2916		 */
2917		/* got to be root to get at low ports */
2918		if (ntohs(lport) < IPPORT_RESERVED) {
2919			if (p && (error =
2920			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2921			    )) {
2922				SCTP_INP_DECR_REF(inp);
2923				SCTP_INP_WUNLOCK(inp);
2924				SCTP_INP_INFO_WUNLOCK();
2925				return (error);
2926			}
2927		}
2928		SCTP_INP_WUNLOCK(inp);
2929		if (bindall) {
2930			vrf_id = inp->def_vrf_id;
2931			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2932			if (inp_tmp != NULL) {
2933				/*
2934				 * lock guy returned and lower count note
2935				 * that we are not bound so inp_tmp should
2936				 * NEVER be inp. And it is this inp
2937				 * (inp_tmp) that gets the reference bump,
2938				 * so we must lower it.
2939				 */
2940				SCTP_INP_DECR_REF(inp_tmp);
2941				/* unlock info */
2942				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2943				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2944					/*
2945					 * Ok, must be one-2-one and
2946					 * allowing port re-use
2947					 */
2948					port_reuse_active = 1;
2949					goto continue_anyway;
2950				}
2951				SCTP_INP_DECR_REF(inp);
2952				SCTP_INP_INFO_WUNLOCK();
2953				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2954				return (EADDRINUSE);
2955			}
2956		} else {
2957			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2958			if (inp_tmp != NULL) {
2959				/*
2960				 * lock guy returned and lower count note
2961				 * that we are not bound so inp_tmp should
2962				 * NEVER be inp. And it is this inp
2963				 * (inp_tmp) that gets the reference bump,
2964				 * so we must lower it.
2965				 */
2966				SCTP_INP_DECR_REF(inp_tmp);
2967				/* unlock info */
2968				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2969				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2970					/*
2971					 * Ok, must be one-2-one and
2972					 * allowing port re-use
2973					 */
2974					port_reuse_active = 1;
2975					goto continue_anyway;
2976				}
2977				SCTP_INP_DECR_REF(inp);
2978				SCTP_INP_INFO_WUNLOCK();
2979				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2980				return (EADDRINUSE);
2981			}
2982		}
2983continue_anyway:
2984		SCTP_INP_WLOCK(inp);
2985		if (bindall) {
2986			/* verify that no lport is not used by a singleton */
2987			if ((port_reuse_active == 0) &&
2988			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
2989				/* Sorry someone already has this one bound */
2990				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2991				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2992					port_reuse_active = 1;
2993				} else {
2994					SCTP_INP_DECR_REF(inp);
2995					SCTP_INP_WUNLOCK(inp);
2996					SCTP_INP_INFO_WUNLOCK();
2997					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2998					return (EADDRINUSE);
2999				}
3000			}
3001		}
3002	} else {
3003		uint16_t first, last, candidate;
3004		uint16_t count;
3005		int done;
3006
3007		if (ip_inp->inp_flags & INP_HIGHPORT) {
3008			first = MODULE_GLOBAL(ipport_hifirstauto);
3009			last = MODULE_GLOBAL(ipport_hilastauto);
3010		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3011			if (p && (error =
3012			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3013			    )) {
3014				SCTP_INP_DECR_REF(inp);
3015				SCTP_INP_WUNLOCK(inp);
3016				SCTP_INP_INFO_WUNLOCK();
3017				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3018				return (error);
3019			}
3020			first = MODULE_GLOBAL(ipport_lowfirstauto);
3021			last = MODULE_GLOBAL(ipport_lowlastauto);
3022		} else {
3023			first = MODULE_GLOBAL(ipport_firstauto);
3024			last = MODULE_GLOBAL(ipport_lastauto);
3025		}
3026		if (first > last) {
3027			uint16_t temp;
3028
3029			temp = first;
3030			first = last;
3031			last = temp;
3032		}
3033		count = last - first + 1;	/* number of candidates */
3034		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3035
3036		done = 0;
3037		while (!done) {
3038			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3039				done = 1;
3040			}
3041			if (!done) {
3042				if (--count == 0) {
3043					SCTP_INP_DECR_REF(inp);
3044					SCTP_INP_WUNLOCK(inp);
3045					SCTP_INP_INFO_WUNLOCK();
3046					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3047					return (EADDRINUSE);
3048				}
3049				if (candidate == last)
3050					candidate = first;
3051				else
3052					candidate = candidate + 1;
3053			}
3054		}
3055		lport = htons(candidate);
3056	}
3057	SCTP_INP_DECR_REF(inp);
3058	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3059	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3060		/*
3061		 * this really should not happen. The guy did a non-blocking
3062		 * bind and then did a close at the same time.
3063		 */
3064		SCTP_INP_WUNLOCK(inp);
3065		SCTP_INP_INFO_WUNLOCK();
3066		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3067		return (EINVAL);
3068	}
3069	/* ok we look clear to give out this port, so lets setup the binding */
3070	if (bindall) {
3071		/* binding to all addresses, so just set in the proper flags */
3072		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3073		/* set the automatic addr changes from kernel flag */
3074		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3075			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3076			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3077		} else {
3078			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3079			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3080		}
3081		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3082			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3083		} else {
3084			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3085		}
3086		/*
3087		 * set the automatic mobility_base from kernel flag (by
3088		 * micchie)
3089		 */
3090		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3091			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3092			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3093		} else {
3094			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3095			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3096		}
3097		/*
3098		 * set the automatic mobility_fasthandoff from kernel flag
3099		 * (by micchie)
3100		 */
3101		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3102			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3103			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3104		} else {
3105			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3106			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3107		}
3108	} else {
3109		/*
3110		 * bind specific, make sure flags is off and add a new
3111		 * address structure to the sctp_addr_list inside the ep
3112		 * structure.
3113		 *
3114		 * We will need to allocate one and insert it at the head. The
3115		 * socketopt call can just insert new addresses in there as
3116		 * well. It will also have to do the embed scope kame hack
3117		 * too (before adding).
3118		 */
3119		struct sctp_ifa *ifa;
3120		struct sockaddr_storage store_sa;
3121
3122		memset(&store_sa, 0, sizeof(store_sa));
3123		switch (addr->sa_family) {
3124#ifdef INET
3125		case AF_INET:
3126			{
3127				struct sockaddr_in *sin;
3128
3129				sin = (struct sockaddr_in *)&store_sa;
3130				memcpy(sin, addr, sizeof(struct sockaddr_in));
3131				sin->sin_port = 0;
3132				break;
3133			}
3134#endif
3135#ifdef INET6
3136		case AF_INET6:
3137			{
3138				struct sockaddr_in6 *sin6;
3139
3140				sin6 = (struct sockaddr_in6 *)&store_sa;
3141				memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3142				sin6->sin6_port = 0;
3143				break;
3144			}
3145#endif
3146		default:
3147			break;
3148		}
3149		/*
3150		 * first find the interface with the bound address need to
3151		 * zero out the port to find the address! yuck! can't do
3152		 * this earlier since need port for sctp_pcb_findep()
3153		 */
3154		if (sctp_ifap != NULL) {
3155			ifa = sctp_ifap;
3156		} else {
3157			/*
3158			 * Note for BSD we hit here always other O/S's will
3159			 * pass things in via the sctp_ifap argument
3160			 * (Panda).
3161			 */
3162			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3163			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3164		}
3165		if (ifa == NULL) {
3166			/* Can't find an interface with that address */
3167			SCTP_INP_WUNLOCK(inp);
3168			SCTP_INP_INFO_WUNLOCK();
3169			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3170			return (EADDRNOTAVAIL);
3171		}
3172#ifdef INET6
3173		if (addr->sa_family == AF_INET6) {
3174			/* GAK, more FIXME IFA lock? */
3175			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3176				/* Can't bind a non-existent addr. */
3177				SCTP_INP_WUNLOCK(inp);
3178				SCTP_INP_INFO_WUNLOCK();
3179				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3180				return (EINVAL);
3181			}
3182		}
3183#endif
3184		/* we're not bound all */
3185		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3186		/* allow bindx() to send ASCONF's for binding changes */
3187		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3188		/* clear automatic addr changes from kernel flag */
3189		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3190
3191		/* add this address to the endpoint list */
3192		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3193		if (error != 0) {
3194			SCTP_INP_WUNLOCK(inp);
3195			SCTP_INP_INFO_WUNLOCK();
3196			return (error);
3197		}
3198		inp->laddr_count++;
3199	}
3200	/* find the bucket */
3201	if (port_reuse_active) {
3202		/* Put it into tcp 1-2-1 hash */
3203		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3204		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3205	} else {
3206		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3207	}
3208	/* put it in the bucket */
3209	LIST_INSERT_HEAD(head, inp, sctp_hash);
3210	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3211	    (void *)head, ntohs(lport), port_reuse_active);
3212	/* set in the port */
3213	inp->sctp_lport = lport;
3214
3215	/* turn off just the unbound flag */
3216	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3217	SCTP_INP_WUNLOCK(inp);
3218	SCTP_INP_INFO_WUNLOCK();
3219	return (0);
3220}
3221
3222
3223static void
3224sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3225{
3226	struct sctp_iterator *it, *nit;
3227
3228	/*
3229	 * We enter with the only the ITERATOR_LOCK in place and a write
3230	 * lock on the inp_info stuff.
3231	 */
3232	it = sctp_it_ctl.cur_it;
3233	if (it && (it->vn != curvnet)) {
3234		/* Its not looking at our VNET */
3235		return;
3236	}
3237	if (it && (it->inp == inp)) {
3238		/*
3239		 * This is tricky and we hold the iterator lock, but when it
3240		 * returns and gets the lock (when we release it) the
3241		 * iterator will try to operate on inp. We need to stop that
3242		 * from happening. But of course the iterator has a
3243		 * reference on the stcb and inp. We can mark it and it will
3244		 * stop.
3245		 *
3246		 * If its a single iterator situation, we set the end iterator
3247		 * flag. Otherwise we set the iterator to go to the next
3248		 * inp.
3249		 *
3250		 */
3251		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3252			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3253		} else {
3254			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3255		}
3256	}
3257	/*
3258	 * Now go through and remove any single reference to our inp that
3259	 * may be still pending on the list
3260	 */
3261	SCTP_IPI_ITERATOR_WQ_LOCK();
3262	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3263		if (it->vn != curvnet) {
3264			continue;
3265		}
3266		if (it->inp == inp) {
3267			/* This one points to me is it inp specific? */
3268			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3269				/* Remove and free this one */
3270				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3271				    it, sctp_nxt_itr);
3272				if (it->function_atend != NULL) {
3273					(*it->function_atend) (it->pointer, it->val);
3274				}
3275				SCTP_FREE(it, SCTP_M_ITER);
3276			} else {
3277				it->inp = LIST_NEXT(it->inp, sctp_list);
3278				if (it->inp) {
3279					SCTP_INP_INCR_REF(it->inp);
3280				}
3281			}
3282			/*
3283			 * When its put in the refcnt is incremented so decr
3284			 * it
3285			 */
3286			SCTP_INP_DECR_REF(inp);
3287		}
3288	}
3289	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3290}
3291
3292/* release sctp_inpcb unbind the port */
3293void
3294sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3295{
3296	/*
3297	 * Here we free a endpoint. We must find it (if it is in the Hash
3298	 * table) and remove it from there. Then we must also find it in the
3299	 * overall list and remove it from there. After all removals are
3300	 * complete then any timer has to be stopped. Then start the actual
3301	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3302	 * all associations. d) finally the ep itself.
3303	 */
3304	struct sctp_tcb *asoc, *nasoc;
3305	struct sctp_laddr *laddr, *nladdr;
3306	struct inpcb *ip_pcb;
3307	struct socket *so;
3308	int being_refed = 0;
3309	struct sctp_queued_to_read *sq, *nsq;
3310	int cnt;
3311	sctp_sharedkey_t *shared_key, *nshared_key;
3312
3313
3314#ifdef SCTP_LOG_CLOSING
3315	sctp_log_closing(inp, NULL, 0);
3316#endif
3317	SCTP_ITERATOR_LOCK();
3318	/* mark any iterators on the list or being processed */
3319	sctp_iterator_inp_being_freed(inp);
3320	SCTP_ITERATOR_UNLOCK();
3321	so = inp->sctp_socket;
3322	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3323		/* been here before.. eeks.. get out of here */
3324		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3325#ifdef SCTP_LOG_CLOSING
3326		sctp_log_closing(inp, NULL, 1);
3327#endif
3328		return;
3329	}
3330	SCTP_ASOC_CREATE_LOCK(inp);
3331	SCTP_INP_INFO_WLOCK();
3332
3333	SCTP_INP_WLOCK(inp);
3334	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3335		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3336		/* socket is gone, so no more wakeups allowed */
3337		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3338		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3339		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3340
3341	}
3342	/* First time through we have the socket lock, after that no more. */
3343	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3344	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3345
3346	if (inp->control) {
3347		sctp_m_freem(inp->control);
3348		inp->control = NULL;
3349	}
3350	if (inp->pkt) {
3351		sctp_m_freem(inp->pkt);
3352		inp->pkt = NULL;
3353	}
3354	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3355					 * here but I will be nice :> (i.e.
3356					 * ip_pcb = ep;) */
3357	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3358		int cnt_in_sd;
3359
3360		cnt_in_sd = 0;
3361		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3362			SCTP_TCB_LOCK(asoc);
3363			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3364				/* Skip guys being freed */
3365				cnt_in_sd++;
3366				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3367					/*
3368					 * Special case - we did not start a
3369					 * kill timer on the asoc due to it
3370					 * was not closed. So go ahead and
3371					 * start it now.
3372					 */
3373					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3374					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3375				}
3376				SCTP_TCB_UNLOCK(asoc);
3377				continue;
3378			}
3379			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3380			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3381			    (asoc->asoc.total_output_queue_size == 0)) {
3382				/*
3383				 * If we have data in queue, we don't want
3384				 * to just free since the app may have done,
3385				 * send()/close or connect/send/close. And
3386				 * it wants the data to get across first.
3387				 */
3388				/* Just abandon things in the front states */
3389				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3390				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3391					cnt_in_sd++;
3392				}
3393				continue;
3394			}
3395			/* Disconnect the socket please */
3396			asoc->sctp_socket = NULL;
3397			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3398			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3399			    (asoc->asoc.control_pdapi) ||
3400			    (asoc->asoc.size_on_all_streams > 0) ||
3401			    (so && (so->so_rcv.sb_cc > 0))) {
3402				/* Left with Data unread */
3403				struct mbuf *op_err;
3404
3405				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3406				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3407				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3408				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3409				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3410				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3411					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3412				}
3413				if (sctp_free_assoc(inp, asoc,
3414				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3415					cnt_in_sd++;
3416				}
3417				continue;
3418			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3419				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3420			    (asoc->asoc.stream_queue_cnt == 0)) {
3421				if (asoc->asoc.locked_on_sending) {
3422					goto abort_anyway;
3423				}
3424				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3425				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3426					struct sctp_nets *netp;
3427
3428					/*
3429					 * there is nothing queued to send,
3430					 * so I send shutdown
3431					 */
3432					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3433					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3434						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3435					}
3436					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3437					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3438					sctp_stop_timers_for_shutdown(asoc);
3439					if (asoc->asoc.alternate) {
3440						netp = asoc->asoc.alternate;
3441					} else {
3442						netp = asoc->asoc.primary_destination;
3443					}
3444					sctp_send_shutdown(asoc, netp);
3445					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3446					    netp);
3447					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3448					    asoc->asoc.primary_destination);
3449					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3450				}
3451			} else {
3452				/* mark into shutdown pending */
3453				struct sctp_stream_queue_pending *sp;
3454
3455				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3456				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3457				    asoc->asoc.primary_destination);
3458				if (asoc->asoc.locked_on_sending) {
3459					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3460					    sctp_streamhead);
3461					if (sp == NULL) {
3462						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3463						    (void *)asoc->asoc.locked_on_sending,
3464						    asoc->asoc.locked_on_sending->stream_no);
3465					} else {
3466						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3467							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3468					}
3469				}
3470				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3471				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3472				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3473					struct mbuf *op_err;
3474
3475			abort_anyway:
3476					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3477					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3478					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3479					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3480					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3481					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3482						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3483					}
3484					if (sctp_free_assoc(inp, asoc,
3485					    SCTP_PCBFREE_NOFORCE,
3486					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3487						cnt_in_sd++;
3488					}
3489					continue;
3490				} else {
3491					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3492				}
3493			}
3494			cnt_in_sd++;
3495			SCTP_TCB_UNLOCK(asoc);
3496		}
3497		/* now is there some left in our SHUTDOWN state? */
3498		if (cnt_in_sd) {
3499#ifdef SCTP_LOG_CLOSING
3500			sctp_log_closing(inp, NULL, 2);
3501#endif
3502			inp->sctp_socket = NULL;
3503			SCTP_INP_WUNLOCK(inp);
3504			SCTP_ASOC_CREATE_UNLOCK(inp);
3505			SCTP_INP_INFO_WUNLOCK();
3506			return;
3507		}
3508	}
3509	inp->sctp_socket = NULL;
3510	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3511	    SCTP_PCB_FLAGS_UNBOUND) {
3512		/*
3513		 * ok, this guy has been bound. It's port is somewhere in
3514		 * the SCTP_BASE_INFO(hash table). Remove it!
3515		 */
3516		LIST_REMOVE(inp, sctp_hash);
3517		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3518	}
3519	/*
3520	 * If there is a timer running to kill us, forget it, since it may
3521	 * have a contest on the INP lock.. which would cause us to die ...
3522	 */
3523	cnt = 0;
3524	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3525		SCTP_TCB_LOCK(asoc);
3526		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3527			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3528				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3529				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3530			}
3531			cnt++;
3532			SCTP_TCB_UNLOCK(asoc);
3533			continue;
3534		}
3535		/* Free associations that are NOT killing us */
3536		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3537		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3538			struct mbuf *op_err;
3539
3540			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3541			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3542			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3543			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3544		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3545			cnt++;
3546			SCTP_TCB_UNLOCK(asoc);
3547			continue;
3548		}
3549		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3550		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3551			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3552		}
3553		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3554			cnt++;
3555		}
3556	}
3557	if (cnt) {
3558		/* Ok we have someone out there that will kill us */
3559		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3560#ifdef SCTP_LOG_CLOSING
3561		sctp_log_closing(inp, NULL, 3);
3562#endif
3563		SCTP_INP_WUNLOCK(inp);
3564		SCTP_ASOC_CREATE_UNLOCK(inp);
3565		SCTP_INP_INFO_WUNLOCK();
3566		return;
3567	}
3568	if (SCTP_INP_LOCK_CONTENDED(inp))
3569		being_refed++;
3570	if (SCTP_INP_READ_CONTENDED(inp))
3571		being_refed++;
3572	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3573		being_refed++;
3574
3575	if ((inp->refcount) ||
3576	    (being_refed) ||
3577	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3578		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3579#ifdef SCTP_LOG_CLOSING
3580		sctp_log_closing(inp, NULL, 4);
3581#endif
3582		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3583		SCTP_INP_WUNLOCK(inp);
3584		SCTP_ASOC_CREATE_UNLOCK(inp);
3585		SCTP_INP_INFO_WUNLOCK();
3586		return;
3587	}
3588	inp->sctp_ep.signature_change.type = 0;
3589	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3590	/*
3591	 * Remove it from the list .. last thing we need a lock for.
3592	 */
3593	LIST_REMOVE(inp, sctp_list);
3594	SCTP_INP_WUNLOCK(inp);
3595	SCTP_ASOC_CREATE_UNLOCK(inp);
3596	SCTP_INP_INFO_WUNLOCK();
3597	/*
3598	 * Now we release all locks. Since this INP cannot be found anymore
3599	 * except possibly by the kill timer that might be running. We call
3600	 * the drain function here. It should hit the case were it sees the
3601	 * ACTIVE flag cleared and exit out freeing us to proceed and
3602	 * destroy everything.
3603	 */
3604	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3605		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3606	} else {
3607		/* Probably un-needed */
3608		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3609	}
3610
3611#ifdef SCTP_LOG_CLOSING
3612	sctp_log_closing(inp, NULL, 5);
3613#endif
3614
3615
3616	if ((inp->sctp_asocidhash) != NULL) {
3617		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3618		inp->sctp_asocidhash = NULL;
3619	}
3620	/* sa_ignore FREED_MEMORY */
3621	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3622		/* Its only abandoned if it had data left */
3623		if (sq->length)
3624			SCTP_STAT_INCR(sctps_left_abandon);
3625
3626		TAILQ_REMOVE(&inp->read_queue, sq, next);
3627		sctp_free_remote_addr(sq->whoFrom);
3628		if (so)
3629			so->so_rcv.sb_cc -= sq->length;
3630		if (sq->data) {
3631			sctp_m_freem(sq->data);
3632			sq->data = NULL;
3633		}
3634		/*
3635		 * no need to free the net count, since at this point all
3636		 * assoc's are gone.
3637		 */
3638		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3639		SCTP_DECR_READQ_COUNT();
3640	}
3641	/* Now the sctp_pcb things */
3642	/*
3643	 * free each asoc if it is not already closed/free. we can't use the
3644	 * macro here since le_next will get freed as part of the
3645	 * sctp_free_assoc() call.
3646	 */
3647	if (so) {
3648#ifdef IPSEC
3649		ipsec_delete_pcbpolicy(ip_pcb);
3650#endif				/* IPSEC */
3651
3652		/* Unlocks not needed since the socket is gone now */
3653	}
3654	if (ip_pcb->inp_options) {
3655		(void)sctp_m_free(ip_pcb->inp_options);
3656		ip_pcb->inp_options = 0;
3657	}
3658#ifdef INET6
3659	if (ip_pcb->inp_vflag & INP_IPV6) {
3660		struct in6pcb *in6p;
3661
3662		in6p = (struct in6pcb *)inp;
3663		ip6_freepcbopts(in6p->in6p_outputopts);
3664	}
3665#endif				/* INET6 */
3666	ip_pcb->inp_vflag = 0;
3667	/* free up authentication fields */
3668	if (inp->sctp_ep.local_auth_chunks != NULL)
3669		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3670	if (inp->sctp_ep.local_hmacs != NULL)
3671		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3672
3673	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3674		LIST_REMOVE(shared_key, next);
3675		sctp_free_sharedkey(shared_key);
3676		/* sa_ignore FREED_MEMORY */
3677	}
3678
3679	/*
3680	 * if we have an address list the following will free the list of
3681	 * ifaddr's that are set into this ep. Again macro limitations here,
3682	 * since the LIST_FOREACH could be a bad idea.
3683	 */
3684	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3685		sctp_remove_laddr(laddr);
3686	}
3687
3688#ifdef SCTP_TRACK_FREED_ASOCS
3689	/* TEMP CODE */
3690	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3691		LIST_REMOVE(asoc, sctp_tcblist);
3692		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3693		SCTP_DECR_ASOC_COUNT();
3694	}
3695	/* *** END TEMP CODE *** */
3696#endif
3697	/* Now lets see about freeing the EP hash table. */
3698	if (inp->sctp_tcbhash != NULL) {
3699		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3700		inp->sctp_tcbhash = NULL;
3701	}
3702	/* Now we must put the ep memory back into the zone pool */
3703	crfree(inp->ip_inp.inp.inp_cred);
3704	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3705	SCTP_INP_LOCK_DESTROY(inp);
3706	SCTP_INP_READ_DESTROY(inp);
3707	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3708	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3709	SCTP_DECR_EP_COUNT();
3710}
3711
3712
3713struct sctp_nets *
3714sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3715{
3716	struct sctp_nets *net;
3717
3718	/* locate the address */
3719	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3720		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3721			return (net);
3722	}
3723	return (NULL);
3724}
3725
3726
3727int
3728sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3729{
3730	struct sctp_ifa *sctp_ifa;
3731
3732	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3733	if (sctp_ifa) {
3734		return (1);
3735	} else {
3736		return (0);
3737	}
3738}
3739
3740/*
3741 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3742 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3743 * stats of stuff.
3744 */
3745int
3746sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3747    struct sctp_nets **netp, int set_scope, int from)
3748{
3749	/*
3750	 * The following is redundant to the same lines in the
3751	 * sctp_aloc_assoc() but is needed since others call the add address
3752	 * function
3753	 */
3754	struct sctp_nets *net, *netfirst;
3755	int addr_inscope;
3756
3757	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3758	    from);
3759	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3760
3761	netfirst = sctp_findnet(stcb, newaddr);
3762	if (netfirst) {
3763		/*
3764		 * Lie and return ok, we don't want to make the association
3765		 * go away for this behavior. It will happen in the TCP
3766		 * model in a connected socket. It does not reach the hash
3767		 * table until after the association is built so it can't be
3768		 * found. Mark as reachable, since the initial creation will
3769		 * have been cleared and the NOT_IN_ASSOC flag will have
3770		 * been added... and we don't want to end up removing it
3771		 * back out.
3772		 */
3773		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3774			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3775			    SCTP_ADDR_UNCONFIRMED);
3776		} else {
3777			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3778		}
3779
3780		return (0);
3781	}
3782	addr_inscope = 1;
3783	switch (newaddr->sa_family) {
3784#ifdef INET
3785	case AF_INET:
3786		{
3787			struct sockaddr_in *sin;
3788
3789			sin = (struct sockaddr_in *)newaddr;
3790			if (sin->sin_addr.s_addr == 0) {
3791				/* Invalid address */
3792				return (-1);
3793			}
3794			/* zero out the bzero area */
3795			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3796
3797			/* assure len is set */
3798			sin->sin_len = sizeof(struct sockaddr_in);
3799			if (set_scope) {
3800#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3801				stcb->asoc.scope.ipv4_local_scope = 1;
3802#else
3803				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3804					stcb->asoc.scope.ipv4_local_scope = 1;
3805				}
3806#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3807			} else {
3808				/* Validate the address is in scope */
3809				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3810				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3811					addr_inscope = 0;
3812				}
3813			}
3814			break;
3815		}
3816#endif
3817#ifdef INET6
3818	case AF_INET6:
3819		{
3820			struct sockaddr_in6 *sin6;
3821
3822			sin6 = (struct sockaddr_in6 *)newaddr;
3823			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3824				/* Invalid address */
3825				return (-1);
3826			}
3827			/* assure len is set */
3828			sin6->sin6_len = sizeof(struct sockaddr_in6);
3829			if (set_scope) {
3830				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3831					stcb->asoc.scope.loopback_scope = 1;
3832					stcb->asoc.scope.local_scope = 0;
3833					stcb->asoc.scope.ipv4_local_scope = 1;
3834					stcb->asoc.scope.site_scope = 1;
3835				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3836					/*
3837					 * If the new destination is a
3838					 * LINK_LOCAL we must have common
3839					 * site scope. Don't set the local
3840					 * scope since we may not share all
3841					 * links, only loopback can do this.
3842					 * Links on the local network would
3843					 * also be on our private network
3844					 * for v4 too.
3845					 */
3846					stcb->asoc.scope.ipv4_local_scope = 1;
3847					stcb->asoc.scope.site_scope = 1;
3848				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3849					/*
3850					 * If the new destination is
3851					 * SITE_LOCAL then we must have site
3852					 * scope in common.
3853					 */
3854					stcb->asoc.scope.site_scope = 1;
3855				}
3856			} else {
3857				/* Validate the address is in scope */
3858				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3859				    (stcb->asoc.scope.loopback_scope == 0)) {
3860					addr_inscope = 0;
3861				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3862				    (stcb->asoc.scope.local_scope == 0)) {
3863					addr_inscope = 0;
3864				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3865				    (stcb->asoc.scope.site_scope == 0)) {
3866					addr_inscope = 0;
3867				}
3868			}
3869			break;
3870		}
3871#endif
3872	default:
3873		/* not supported family type */
3874		return (-1);
3875	}
3876	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3877	if (net == NULL) {
3878		return (-1);
3879	}
3880	SCTP_INCR_RADDR_COUNT();
3881	bzero(net, sizeof(struct sctp_nets));
3882	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3883	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3884	switch (newaddr->sa_family) {
3885#ifdef INET
3886	case AF_INET:
3887		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3888		break;
3889#endif
3890#ifdef INET6
3891	case AF_INET6:
3892		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3893		break;
3894#endif
3895	default:
3896		break;
3897	}
3898	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3899	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3900		stcb->asoc.scope.loopback_scope = 1;
3901		stcb->asoc.scope.ipv4_local_scope = 1;
3902		stcb->asoc.scope.local_scope = 0;
3903		stcb->asoc.scope.site_scope = 1;
3904		addr_inscope = 1;
3905	}
3906	net->failure_threshold = stcb->asoc.def_net_failure;
3907	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3908	if (addr_inscope == 0) {
3909		net->dest_state = (SCTP_ADDR_REACHABLE |
3910		    SCTP_ADDR_OUT_OF_SCOPE);
3911	} else {
3912		if (from == SCTP_ADDR_IS_CONFIRMED)
3913			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3914			net->dest_state = SCTP_ADDR_REACHABLE;
3915		else
3916			net->dest_state = SCTP_ADDR_REACHABLE |
3917			    SCTP_ADDR_UNCONFIRMED;
3918	}
3919	/*
3920	 * We set this to 0, the timer code knows that this means its an
3921	 * initial value
3922	 */
3923	net->rto_needed = 1;
3924	net->RTO = 0;
3925	net->RTO_measured = 0;
3926	stcb->asoc.numnets++;
3927	net->ref_count = 1;
3928	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3929	net->port = stcb->asoc.port;
3930	net->dscp = stcb->asoc.default_dscp;
3931#ifdef INET6
3932	net->flowlabel = stcb->asoc.default_flowlabel;
3933#endif
3934	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3935		net->dest_state |= SCTP_ADDR_NOHB;
3936	} else {
3937		net->dest_state &= ~SCTP_ADDR_NOHB;
3938	}
3939	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3940		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3941	} else {
3942		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3943	}
3944	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3945	/* Init the timer structure */
3946	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3947	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3948	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3949
3950	/* Now generate a route for this guy */
3951#ifdef INET6
3952	/* KAME hack: embed scopeid */
3953	if (newaddr->sa_family == AF_INET6) {
3954		struct sockaddr_in6 *sin6;
3955
3956		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3957		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3958		sin6->sin6_scope_id = 0;
3959	}
3960#endif
3961	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3962
3963	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3964		/* Get source address */
3965		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3966		    stcb,
3967		    (sctp_route_t *) & net->ro,
3968		    net,
3969		    0,
3970		    stcb->asoc.vrf_id);
3971		/* Now get the interface MTU */
3972		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3973			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
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	}
3996	if (net->mtu == 0) {
3997		switch (newaddr->sa_family) {
3998#ifdef INET
3999		case AF_INET:
4000			net->mtu = SCTP_DEFAULT_MTU;
4001			break;
4002#endif
4003#ifdef INET6
4004		case AF_INET6:
4005			net->mtu = 1280;
4006			break;
4007#endif
4008		default:
4009			break;
4010		}
4011	}
4012#if defined(INET) || defined(INET6)
4013	if (net->port) {
4014		net->mtu -= (uint32_t) sizeof(struct udphdr);
4015	}
4016#endif
4017	if (from == SCTP_ALLOC_ASOC) {
4018		stcb->asoc.smallest_mtu = net->mtu;
4019	}
4020	if (stcb->asoc.smallest_mtu > net->mtu) {
4021		stcb->asoc.smallest_mtu = net->mtu;
4022	}
4023#ifdef INET6
4024	if (newaddr->sa_family == AF_INET6) {
4025		struct sockaddr_in6 *sin6;
4026
4027		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4028		(void)sa6_recoverscope(sin6);
4029	}
4030#endif
4031
4032	/* JRS - Use the congestion control given in the CC module */
4033	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4034		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4035
4036	/*
4037	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4038	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4039	 */
4040	net->find_pseudo_cumack = 1;
4041	net->find_rtx_pseudo_cumack = 1;
4042	net->src_addr_selected = 0;
4043	/* Choose an initial flowid. */
4044	net->flowid = stcb->asoc.my_vtag ^
4045	    ntohs(stcb->rport) ^
4046	    ntohs(stcb->sctp_ep->sctp_lport);
4047#ifdef INVARIANTS
4048	net->flowidset = 1;
4049#endif
4050	if (netp) {
4051		*netp = net;
4052	}
4053	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4054	if (net->ro.ro_rt == NULL) {
4055		/* Since we have no route put it at the back */
4056		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4057	} else if (netfirst == NULL) {
4058		/* We are the first one in the pool. */
4059		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4060	} else if (netfirst->ro.ro_rt == NULL) {
4061		/*
4062		 * First one has NO route. Place this one ahead of the first
4063		 * one.
4064		 */
4065		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4066	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4067		/*
4068		 * This one has a different interface than the one at the
4069		 * top of the list. Place it ahead.
4070		 */
4071		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4072	} else {
4073		/*
4074		 * Ok we have the same interface as the first one. Move
4075		 * forward until we find either a) one with a NULL route...
4076		 * insert ahead of that b) one with a different ifp.. insert
4077		 * after that. c) end of the list.. insert at the tail.
4078		 */
4079		struct sctp_nets *netlook;
4080
4081		do {
4082			netlook = TAILQ_NEXT(netfirst, sctp_next);
4083			if (netlook == NULL) {
4084				/* End of the list */
4085				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4086				break;
4087			} else if (netlook->ro.ro_rt == NULL) {
4088				/* next one has NO route */
4089				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4090				break;
4091			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4092				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4093				    net, sctp_next);
4094				break;
4095			}
4096			/* Shift forward */
4097			netfirst = netlook;
4098		} while (netlook != NULL);
4099	}
4100
4101	/* got to have a primary set */
4102	if (stcb->asoc.primary_destination == 0) {
4103		stcb->asoc.primary_destination = net;
4104	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4105		    (net->ro.ro_rt) &&
4106	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4107		/* No route to current primary adopt new primary */
4108		stcb->asoc.primary_destination = net;
4109	}
4110	/* Validate primary is first */
4111	net = TAILQ_FIRST(&stcb->asoc.nets);
4112	if ((net != stcb->asoc.primary_destination) &&
4113	    (stcb->asoc.primary_destination)) {
4114		/*
4115		 * first one on the list is NOT the primary sctp_cmpaddr()
4116		 * is much more efficient if the primary is the first on the
4117		 * list, make it so.
4118		 */
4119		TAILQ_REMOVE(&stcb->asoc.nets,
4120		    stcb->asoc.primary_destination, sctp_next);
4121		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4122		    stcb->asoc.primary_destination, sctp_next);
4123	}
4124	return (0);
4125}
4126
4127
4128static uint32_t
4129sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4130{
4131	uint32_t id;
4132	struct sctpasochead *head;
4133	struct sctp_tcb *lstcb;
4134
4135	SCTP_INP_WLOCK(inp);
4136try_again:
4137	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4138		/* TSNH */
4139		SCTP_INP_WUNLOCK(inp);
4140		return (0);
4141	}
4142	/*
4143	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4144	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4145	 */
4146	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4147		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4148	}
4149	id = inp->sctp_associd_counter;
4150	inp->sctp_associd_counter++;
4151	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4152	if (lstcb) {
4153		goto try_again;
4154	}
4155	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4156	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4157	stcb->asoc.in_asocid_hash = 1;
4158	SCTP_INP_WUNLOCK(inp);
4159	return id;
4160}
4161
4162/*
4163 * allocate an association and add it to the endpoint. The caller must be
4164 * careful to add all additional addresses once they are know right away or
4165 * else the assoc will be may experience a blackout scenario.
4166 */
4167struct sctp_tcb *
4168sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4169    int *error, uint32_t override_tag, uint32_t vrf_id,
4170    struct thread *p
4171)
4172{
4173	/* note the p argument is only valid in unbound sockets */
4174
4175	struct sctp_tcb *stcb;
4176	struct sctp_association *asoc;
4177	struct sctpasochead *head;
4178	uint16_t rport;
4179	int err;
4180
4181	/*
4182	 * Assumption made here: Caller has done a
4183	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4184	 * address does not exist already.
4185	 */
4186	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4187		/* Hit max assoc, sorry no more */
4188		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4189		*error = ENOBUFS;
4190		return (NULL);
4191	}
4192	if (firstaddr == NULL) {
4193		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4194		*error = EINVAL;
4195		return (NULL);
4196	}
4197	SCTP_INP_RLOCK(inp);
4198	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4199	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4200	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4201		/*
4202		 * If its in the TCP pool, its NOT allowed to create an
4203		 * association. The parent listener needs to call
4204		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4205		 * off, or connected one does this.. its an error.
4206		 */
4207		SCTP_INP_RUNLOCK(inp);
4208		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4209		*error = EINVAL;
4210		return (NULL);
4211	}
4212	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4213	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4214		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4215		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4216			SCTP_INP_RUNLOCK(inp);
4217			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4218			*error = EINVAL;
4219			return (NULL);
4220		}
4221	}
4222	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4223#ifdef SCTP_DEBUG
4224	if (firstaddr) {
4225		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4226		switch (firstaddr->sa_family) {
4227#ifdef INET
4228		case AF_INET:
4229			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4230			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4231			break;
4232#endif
4233#ifdef INET6
4234		case AF_INET6:
4235			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4236			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4237			break;
4238#endif
4239		default:
4240			break;
4241		}
4242	} else {
4243		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4244	}
4245#endif				/* SCTP_DEBUG */
4246	switch (firstaddr->sa_family) {
4247#ifdef INET
4248	case AF_INET:
4249		{
4250			struct sockaddr_in *sin;
4251
4252			sin = (struct sockaddr_in *)firstaddr;
4253			if ((ntohs(sin->sin_port) == 0) ||
4254			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4255			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4256			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4257				/* Invalid address */
4258				SCTP_INP_RUNLOCK(inp);
4259				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4260				*error = EINVAL;
4261				return (NULL);
4262			}
4263			rport = sin->sin_port;
4264			break;
4265		}
4266#endif
4267#ifdef INET6
4268	case AF_INET6:
4269		{
4270			struct sockaddr_in6 *sin6;
4271
4272			sin6 = (struct sockaddr_in6 *)firstaddr;
4273			if ((ntohs(sin6->sin6_port) == 0) ||
4274			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4275			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4276				/* Invalid address */
4277				SCTP_INP_RUNLOCK(inp);
4278				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4279				*error = EINVAL;
4280				return (NULL);
4281			}
4282			rport = sin6->sin6_port;
4283			break;
4284		}
4285#endif
4286	default:
4287		/* not supported family type */
4288		SCTP_INP_RUNLOCK(inp);
4289		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4290		*error = EINVAL;
4291		return (NULL);
4292	}
4293	SCTP_INP_RUNLOCK(inp);
4294	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4295		/*
4296		 * If you have not performed a bind, then we need to do the
4297		 * ephemeral bind for you.
4298		 */
4299		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4300		    (struct sockaddr *)NULL,
4301		    (struct sctp_ifa *)NULL,
4302		    p
4303		    ))) {
4304			/* bind error, probably perm */
4305			*error = err;
4306			return (NULL);
4307		}
4308	}
4309	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4310	if (stcb == NULL) {
4311		/* out of memory? */
4312		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4313		*error = ENOMEM;
4314		return (NULL);
4315	}
4316	SCTP_INCR_ASOC_COUNT();
4317
4318	bzero(stcb, sizeof(*stcb));
4319	asoc = &stcb->asoc;
4320
4321	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4322	SCTP_TCB_LOCK_INIT(stcb);
4323	SCTP_TCB_SEND_LOCK_INIT(stcb);
4324	stcb->rport = rport;
4325	/* setup back pointer's */
4326	stcb->sctp_ep = inp;
4327	stcb->sctp_socket = inp->sctp_socket;
4328	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4329		/* failed */
4330		SCTP_TCB_LOCK_DESTROY(stcb);
4331		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4332		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4333		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4334		SCTP_DECR_ASOC_COUNT();
4335		*error = err;
4336		return (NULL);
4337	}
4338	/* and the port */
4339	SCTP_INP_INFO_WLOCK();
4340	SCTP_INP_WLOCK(inp);
4341	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4342		/* inpcb freed while alloc going on */
4343		SCTP_TCB_LOCK_DESTROY(stcb);
4344		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4345		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4346		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4347		SCTP_INP_WUNLOCK(inp);
4348		SCTP_INP_INFO_WUNLOCK();
4349		SCTP_DECR_ASOC_COUNT();
4350		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4351		*error = EINVAL;
4352		return (NULL);
4353	}
4354	SCTP_TCB_LOCK(stcb);
4355
4356	/* now that my_vtag is set, add it to the hash */
4357	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4358	/* put it in the bucket in the vtag hash of assoc's for the system */
4359	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4360	SCTP_INP_INFO_WUNLOCK();
4361
4362	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4363		/* failure.. memory error? */
4364		if (asoc->strmout) {
4365			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4366			asoc->strmout = NULL;
4367		}
4368		if (asoc->mapping_array) {
4369			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4370			asoc->mapping_array = NULL;
4371		}
4372		if (asoc->nr_mapping_array) {
4373			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4374			asoc->nr_mapping_array = NULL;
4375		}
4376		SCTP_DECR_ASOC_COUNT();
4377		SCTP_TCB_UNLOCK(stcb);
4378		SCTP_TCB_LOCK_DESTROY(stcb);
4379		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4380		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4381		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4382		SCTP_INP_WUNLOCK(inp);
4383		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4384		*error = ENOBUFS;
4385		return (NULL);
4386	}
4387	/* Init all the timers */
4388	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4389	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4390	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4391	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4392	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4393	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4394	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4395
4396	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4397	/* now file the port under the hash as well */
4398	if (inp->sctp_tcbhash != NULL) {
4399		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4400		    inp->sctp_hashmark)];
4401		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4402	}
4403	SCTP_INP_WUNLOCK(inp);
4404	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4405	return (stcb);
4406}
4407
4408
4409void
4410sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4411{
4412	struct sctp_association *asoc;
4413
4414	asoc = &stcb->asoc;
4415	asoc->numnets--;
4416	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4417	if (net == asoc->primary_destination) {
4418		/* Reset primary */
4419		struct sctp_nets *lnet;
4420
4421		lnet = TAILQ_FIRST(&asoc->nets);
4422		/*
4423		 * Mobility adaptation Ideally, if deleted destination is
4424		 * the primary, it becomes a fast retransmission trigger by
4425		 * the subsequent SET PRIMARY. (by micchie)
4426		 */
4427		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4428		    SCTP_MOBILITY_BASE) ||
4429		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4430		    SCTP_MOBILITY_FASTHANDOFF)) {
4431			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4432			if (asoc->deleted_primary != NULL) {
4433				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4434				goto out;
4435			}
4436			asoc->deleted_primary = net;
4437			atomic_add_int(&net->ref_count, 1);
4438			memset(&net->lastsa, 0, sizeof(net->lastsa));
4439			memset(&net->lastsv, 0, sizeof(net->lastsv));
4440			sctp_mobility_feature_on(stcb->sctp_ep,
4441			    SCTP_MOBILITY_PRIM_DELETED);
4442			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4443			    stcb->sctp_ep, stcb, NULL);
4444		}
4445out:
4446		/* Try to find a confirmed primary */
4447		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4448	}
4449	if (net == asoc->last_data_chunk_from) {
4450		/* Reset primary */
4451		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4452	}
4453	if (net == asoc->last_control_chunk_from) {
4454		/* Clear net */
4455		asoc->last_control_chunk_from = NULL;
4456	}
4457	if (net == stcb->asoc.alternate) {
4458		sctp_free_remote_addr(stcb->asoc.alternate);
4459		stcb->asoc.alternate = NULL;
4460	}
4461	sctp_free_remote_addr(net);
4462}
4463
4464/*
4465 * remove a remote endpoint address from an association, it will fail if the
4466 * address does not exist.
4467 */
4468int
4469sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4470{
4471	/*
4472	 * Here we need to remove a remote address. This is quite simple, we
4473	 * first find it in the list of address for the association
4474	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4475	 * on that item. Note we do not allow it to be removed if there are
4476	 * no other addresses.
4477	 */
4478	struct sctp_association *asoc;
4479	struct sctp_nets *net, *nnet;
4480
4481	asoc = &stcb->asoc;
4482
4483	/* locate the address */
4484	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4485		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4486			continue;
4487		}
4488		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4489		    remaddr)) {
4490			/* we found the guy */
4491			if (asoc->numnets < 2) {
4492				/* Must have at LEAST two remote addresses */
4493				return (-1);
4494			} else {
4495				sctp_remove_net(stcb, net);
4496				return (0);
4497			}
4498		}
4499	}
4500	/* not found. */
4501	return (-2);
4502}
4503
4504void
4505sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4506{
4507	struct sctpvtaghead *chain;
4508	struct sctp_tagblock *twait_block;
4509	int found = 0;
4510	int i;
4511
4512	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4513	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4514		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4515			if ((twait_block->vtag_block[i].v_tag == tag) &&
4516			    (twait_block->vtag_block[i].lport == lport) &&
4517			    (twait_block->vtag_block[i].rport == rport)) {
4518				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4519				twait_block->vtag_block[i].v_tag = 0;
4520				twait_block->vtag_block[i].lport = 0;
4521				twait_block->vtag_block[i].rport = 0;
4522				found = 1;
4523				break;
4524			}
4525		}
4526		if (found)
4527			break;
4528	}
4529}
4530
4531int
4532sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4533{
4534	struct sctpvtaghead *chain;
4535	struct sctp_tagblock *twait_block;
4536	int found = 0;
4537	int i;
4538
4539	SCTP_INP_INFO_WLOCK();
4540	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4541	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4542		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4543			if ((twait_block->vtag_block[i].v_tag == tag) &&
4544			    (twait_block->vtag_block[i].lport == lport) &&
4545			    (twait_block->vtag_block[i].rport == rport)) {
4546				found = 1;
4547				break;
4548			}
4549		}
4550		if (found)
4551			break;
4552	}
4553	SCTP_INP_INFO_WUNLOCK();
4554	return (found);
4555}
4556
4557
4558void
4559sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4560{
4561	struct sctpvtaghead *chain;
4562	struct sctp_tagblock *twait_block;
4563	struct timeval now;
4564	int set, i;
4565
4566	if (time == 0) {
4567		/* Its disabled */
4568		return;
4569	}
4570	(void)SCTP_GETTIME_TIMEVAL(&now);
4571	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4572	set = 0;
4573	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4574		/* Block(s) present, lets find space, and expire on the fly */
4575		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4576			if ((twait_block->vtag_block[i].v_tag == 0) &&
4577			    !set) {
4578				twait_block->vtag_block[i].tv_sec_at_expire =
4579				    now.tv_sec + time;
4580				twait_block->vtag_block[i].v_tag = tag;
4581				twait_block->vtag_block[i].lport = lport;
4582				twait_block->vtag_block[i].rport = rport;
4583				set = 1;
4584			} else if ((twait_block->vtag_block[i].v_tag) &&
4585			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4586				/* Audit expires this guy */
4587				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4588				twait_block->vtag_block[i].v_tag = 0;
4589				twait_block->vtag_block[i].lport = 0;
4590				twait_block->vtag_block[i].rport = 0;
4591				if (set == 0) {
4592					/* Reuse it for my new tag */
4593					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4594					twait_block->vtag_block[i].v_tag = tag;
4595					twait_block->vtag_block[i].lport = lport;
4596					twait_block->vtag_block[i].rport = rport;
4597					set = 1;
4598				}
4599			}
4600		}
4601		if (set) {
4602			/*
4603			 * We only do up to the block where we can place our
4604			 * tag for audits
4605			 */
4606			break;
4607		}
4608	}
4609	/* Need to add a new block to chain */
4610	if (!set) {
4611		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4612		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4613		if (twait_block == NULL) {
4614#ifdef INVARIANTS
4615			panic("Can not alloc tagblock");
4616#endif
4617			return;
4618		}
4619		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4620		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4621		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4622		twait_block->vtag_block[0].v_tag = tag;
4623		twait_block->vtag_block[0].lport = lport;
4624		twait_block->vtag_block[0].rport = rport;
4625	}
4626}
4627
4628
4629
4630/*-
4631 * Free the association after un-hashing the remote port. This
4632 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4633 * expect that the input to this function IS a locked TCB.
4634 * It will return 0, if it did NOT destroy the association (instead
4635 * it unlocks it. It will return NON-zero if it either destroyed the
4636 * association OR the association is already destroyed.
4637 */
4638int
4639sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4640{
4641	int i;
4642	struct sctp_association *asoc;
4643	struct sctp_nets *net, *nnet;
4644	struct sctp_laddr *laddr, *naddr;
4645	struct sctp_tmit_chunk *chk, *nchk;
4646	struct sctp_asconf_addr *aparam, *naparam;
4647	struct sctp_asconf_ack *aack, *naack;
4648	struct sctp_stream_reset_list *strrst, *nstrrst;
4649	struct sctp_queued_to_read *sq, *nsq;
4650	struct sctp_stream_queue_pending *sp, *nsp;
4651	sctp_sharedkey_t *shared_key, *nshared_key;
4652	struct socket *so;
4653
4654	/* first, lets purge the entry from the hash table. */
4655
4656#ifdef SCTP_LOG_CLOSING
4657	sctp_log_closing(inp, stcb, 6);
4658#endif
4659	if (stcb->asoc.state == 0) {
4660#ifdef SCTP_LOG_CLOSING
4661		sctp_log_closing(inp, NULL, 7);
4662#endif
4663		/* there is no asoc, really TSNH :-0 */
4664		return (1);
4665	}
4666	if (stcb->asoc.alternate) {
4667		sctp_free_remote_addr(stcb->asoc.alternate);
4668		stcb->asoc.alternate = NULL;
4669	}
4670	/* TEMP CODE */
4671	if (stcb->freed_from_where == 0) {
4672		/* Only record the first place free happened from */
4673		stcb->freed_from_where = from_location;
4674	}
4675	/* TEMP CODE */
4676
4677	asoc = &stcb->asoc;
4678	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4679	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4680		/* nothing around */
4681		so = NULL;
4682	else
4683		so = inp->sctp_socket;
4684
4685	/*
4686	 * We used timer based freeing if a reader or writer is in the way.
4687	 * So we first check if we are actually being called from a timer,
4688	 * if so we abort early if a reader or writer is still in the way.
4689	 */
4690	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4691	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4692		/*
4693		 * is it the timer driving us? if so are the reader/writers
4694		 * gone?
4695		 */
4696		if (stcb->asoc.refcnt) {
4697			/* nope, reader or writer in the way */
4698			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4699			/* no asoc destroyed */
4700			SCTP_TCB_UNLOCK(stcb);
4701#ifdef SCTP_LOG_CLOSING
4702			sctp_log_closing(inp, stcb, 8);
4703#endif
4704			return (0);
4705		}
4706	}
4707	/* now clean up any other timers */
4708	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4709	asoc->dack_timer.self = NULL;
4710	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4711	/*-
4712	 * For stream reset we don't blast this unless
4713	 * it is a str-reset timer, it might be the
4714	 * free-asoc timer which we DON'T want to
4715	 * disturb.
4716	 */
4717	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4718		asoc->strreset_timer.self = NULL;
4719	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4720	asoc->asconf_timer.self = NULL;
4721	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4722	asoc->autoclose_timer.self = NULL;
4723	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4724	asoc->shut_guard_timer.self = NULL;
4725	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4726	asoc->delayed_event_timer.self = NULL;
4727	/* Mobility adaptation */
4728	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4729	asoc->delete_prim_timer.self = NULL;
4730	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4731		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4732		net->rxt_timer.self = NULL;
4733		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4734		net->pmtu_timer.self = NULL;
4735		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4736		net->hb_timer.self = NULL;
4737	}
4738	/* Now the read queue needs to be cleaned up (only once) */
4739	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4740		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4741		SCTP_INP_READ_LOCK(inp);
4742		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4743			if (sq->stcb == stcb) {
4744				sq->do_not_ref_stcb = 1;
4745				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4746				/*
4747				 * If there is no end, there never will be
4748				 * now.
4749				 */
4750				if (sq->end_added == 0) {
4751					/* Held for PD-API clear that. */
4752					sq->pdapi_aborted = 1;
4753					sq->held_length = 0;
4754					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4755						/*
4756						 * Need to add a PD-API
4757						 * aborted indication.
4758						 * Setting the control_pdapi
4759						 * assures that it will be
4760						 * added right after this
4761						 * msg.
4762						 */
4763						uint32_t strseq;
4764
4765						stcb->asoc.control_pdapi = sq;
4766						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4767						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4768						    stcb,
4769						    SCTP_PARTIAL_DELIVERY_ABORTED,
4770						    (void *)&strseq,
4771						    SCTP_SO_LOCKED);
4772						stcb->asoc.control_pdapi = NULL;
4773					}
4774				}
4775				/* Add an end to wake them */
4776				sq->end_added = 1;
4777			}
4778		}
4779		SCTP_INP_READ_UNLOCK(inp);
4780		if (stcb->block_entry) {
4781			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4782			stcb->block_entry->error = ECONNRESET;
4783			stcb->block_entry = NULL;
4784		}
4785	}
4786	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4787		/*
4788		 * Someone holds a reference OR the socket is unaccepted
4789		 * yet.
4790		 */
4791		if ((stcb->asoc.refcnt) ||
4792		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4793		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4794			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4795			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4796		}
4797		SCTP_TCB_UNLOCK(stcb);
4798		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4799		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4800			/* nothing around */
4801			so = NULL;
4802		if (so) {
4803			/* Wake any reader/writers */
4804			sctp_sorwakeup(inp, so);
4805			sctp_sowwakeup(inp, so);
4806		}
4807#ifdef SCTP_LOG_CLOSING
4808		sctp_log_closing(inp, stcb, 9);
4809#endif
4810		/* no asoc destroyed */
4811		return (0);
4812	}
4813#ifdef SCTP_LOG_CLOSING
4814	sctp_log_closing(inp, stcb, 10);
4815#endif
4816	/*
4817	 * When I reach here, no others want to kill the assoc yet.. and I
4818	 * own the lock. Now its possible an abort comes in when I do the
4819	 * lock exchange below to grab all the locks to do the final take
4820	 * out. to prevent this we increment the count, which will start a
4821	 * timer and blow out above thus assuring us that we hold exclusive
4822	 * killing of the asoc. Note that after getting back the TCB lock we
4823	 * will go ahead and increment the counter back up and stop any
4824	 * timer a passing stranger may have started :-S
4825	 */
4826	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4827		atomic_add_int(&stcb->asoc.refcnt, 1);
4828
4829		SCTP_TCB_UNLOCK(stcb);
4830		SCTP_INP_INFO_WLOCK();
4831		SCTP_INP_WLOCK(inp);
4832		SCTP_TCB_LOCK(stcb);
4833	}
4834	/* Double check the GONE flag */
4835	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4836	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4837		/* nothing around */
4838		so = NULL;
4839
4840	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4841	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4842		/*
4843		 * For TCP type we need special handling when we are
4844		 * connected. We also include the peel'ed off ones to.
4845		 */
4846		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4847			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4848			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4849			if (so) {
4850				SOCK_LOCK(so);
4851				if (so->so_rcv.sb_cc == 0) {
4852					so->so_state &= ~(SS_ISCONNECTING |
4853					    SS_ISDISCONNECTING |
4854					    SS_ISCONFIRMING |
4855					    SS_ISCONNECTED);
4856				}
4857				socantrcvmore_locked(so);
4858				sctp_sowwakeup(inp, so);
4859				sctp_sorwakeup(inp, so);
4860				SCTP_SOWAKEUP(so);
4861			}
4862		}
4863	}
4864	/*
4865	 * Make it invalid too, that way if its about to run it will abort
4866	 * and return.
4867	 */
4868	/* re-increment the lock */
4869	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4870		atomic_add_int(&stcb->asoc.refcnt, -1);
4871	}
4872	if (stcb->asoc.refcnt) {
4873		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4874		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4875		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4876			SCTP_INP_INFO_WUNLOCK();
4877			SCTP_INP_WUNLOCK(inp);
4878		}
4879		SCTP_TCB_UNLOCK(stcb);
4880		return (0);
4881	}
4882	asoc->state = 0;
4883	if (inp->sctp_tcbhash) {
4884		LIST_REMOVE(stcb, sctp_tcbhash);
4885	}
4886	if (stcb->asoc.in_asocid_hash) {
4887		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4888	}
4889	/* Now lets remove it from the list of ALL associations in the EP */
4890	LIST_REMOVE(stcb, sctp_tcblist);
4891	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4892		SCTP_INP_INCR_REF(inp);
4893		SCTP_INP_WUNLOCK(inp);
4894	}
4895	/* pull from vtag hash */
4896	LIST_REMOVE(stcb, sctp_asocs);
4897	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4898	    inp->sctp_lport, stcb->rport);
4899
4900	/*
4901	 * Now restop the timers to be sure this is paranoia at is finest!
4902	 */
4903	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4904	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4905	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4906	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4907	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4908	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4909	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4910	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4911		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4912		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4913		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4914	}
4915
4916	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4917	/*
4918	 * The chunk lists and such SHOULD be empty but we check them just
4919	 * in case.
4920	 */
4921	/* anything on the wheel needs to be removed */
4922	for (i = 0; i < asoc->streamoutcnt; i++) {
4923		struct sctp_stream_out *outs;
4924
4925		outs = &asoc->strmout[i];
4926		/* now clean up any chunks here */
4927		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4928			TAILQ_REMOVE(&outs->outqueue, sp, next);
4929			sctp_free_spbufspace(stcb, asoc, sp);
4930			if (sp->data) {
4931				if (so) {
4932					/* Still an open socket - report */
4933					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4934					    0, (void *)sp, SCTP_SO_LOCKED);
4935				}
4936				if (sp->data) {
4937					sctp_m_freem(sp->data);
4938					sp->data = NULL;
4939					sp->tail_mbuf = NULL;
4940					sp->length = 0;
4941				}
4942			}
4943			if (sp->net) {
4944				sctp_free_remote_addr(sp->net);
4945				sp->net = NULL;
4946			}
4947			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4948		}
4949	}
4950	/* sa_ignore FREED_MEMORY */
4951	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4952		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4953		SCTP_FREE(strrst, SCTP_M_STRESET);
4954	}
4955	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4956		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4957		if (sq->data) {
4958			sctp_m_freem(sq->data);
4959			sq->data = NULL;
4960		}
4961		sctp_free_remote_addr(sq->whoFrom);
4962		sq->whoFrom = NULL;
4963		sq->stcb = NULL;
4964		/* Free the ctl entry */
4965		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4966		SCTP_DECR_READQ_COUNT();
4967		/* sa_ignore FREED_MEMORY */
4968	}
4969	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4970		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4971		if (chk->data) {
4972			sctp_m_freem(chk->data);
4973			chk->data = NULL;
4974		}
4975		if (chk->holds_key_ref)
4976			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4977		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4978		SCTP_DECR_CHK_COUNT();
4979		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4980		asoc->free_chunk_cnt--;
4981		/* sa_ignore FREED_MEMORY */
4982	}
4983	/* pending send queue SHOULD be empty */
4984	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4985		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4986			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4987#ifdef INVARIANTS
4988		} else {
4989			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4990#endif
4991		}
4992		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4993		if (chk->data) {
4994			if (so) {
4995				/* Still a socket? */
4996				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
4997				    0, chk, SCTP_SO_LOCKED);
4998			}
4999			if (chk->data) {
5000				sctp_m_freem(chk->data);
5001				chk->data = NULL;
5002			}
5003		}
5004		if (chk->holds_key_ref)
5005			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5006		if (chk->whoTo) {
5007			sctp_free_remote_addr(chk->whoTo);
5008			chk->whoTo = NULL;
5009		}
5010		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5011		SCTP_DECR_CHK_COUNT();
5012		/* sa_ignore FREED_MEMORY */
5013	}
5014	/* sent queue SHOULD be empty */
5015	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5016		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5017			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
5018				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
5019#ifdef INVARIANTS
5020			} else {
5021				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
5022#endif
5023			}
5024		}
5025		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5026		if (chk->data) {
5027			if (so) {
5028				/* Still a socket? */
5029				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5030				    0, chk, SCTP_SO_LOCKED);
5031			}
5032			if (chk->data) {
5033				sctp_m_freem(chk->data);
5034				chk->data = NULL;
5035			}
5036		}
5037		if (chk->holds_key_ref)
5038			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5039		sctp_free_remote_addr(chk->whoTo);
5040		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5041		SCTP_DECR_CHK_COUNT();
5042		/* sa_ignore FREED_MEMORY */
5043	}
5044#ifdef INVARIANTS
5045	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5046		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5047			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5048		}
5049	}
5050#endif
5051	/* control queue MAY not be empty */
5052	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5053		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5054		if (chk->data) {
5055			sctp_m_freem(chk->data);
5056			chk->data = NULL;
5057		}
5058		if (chk->holds_key_ref)
5059			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5060		sctp_free_remote_addr(chk->whoTo);
5061		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5062		SCTP_DECR_CHK_COUNT();
5063		/* sa_ignore FREED_MEMORY */
5064	}
5065	/* ASCONF queue MAY not be empty */
5066	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5067		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5068		if (chk->data) {
5069			sctp_m_freem(chk->data);
5070			chk->data = NULL;
5071		}
5072		if (chk->holds_key_ref)
5073			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5074		sctp_free_remote_addr(chk->whoTo);
5075		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5076		SCTP_DECR_CHK_COUNT();
5077		/* sa_ignore FREED_MEMORY */
5078	}
5079	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5080		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5081		if (chk->data) {
5082			sctp_m_freem(chk->data);
5083			chk->data = NULL;
5084		}
5085		if (chk->holds_key_ref)
5086			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5087		sctp_free_remote_addr(chk->whoTo);
5088		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5089		SCTP_DECR_CHK_COUNT();
5090		/* sa_ignore FREED_MEMORY */
5091	}
5092
5093	if (asoc->mapping_array) {
5094		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5095		asoc->mapping_array = NULL;
5096	}
5097	if (asoc->nr_mapping_array) {
5098		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5099		asoc->nr_mapping_array = NULL;
5100	}
5101	/* the stream outs */
5102	if (asoc->strmout) {
5103		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5104		asoc->strmout = NULL;
5105	}
5106	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5107	if (asoc->strmin) {
5108		struct sctp_queued_to_read *ctl, *nctl;
5109
5110		for (i = 0; i < asoc->streamincnt; i++) {
5111			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5112				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5113				sctp_free_remote_addr(ctl->whoFrom);
5114				if (ctl->data) {
5115					sctp_m_freem(ctl->data);
5116					ctl->data = NULL;
5117				}
5118				/*
5119				 * We don't free the address here since all
5120				 * the net's were freed above.
5121				 */
5122				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5123				SCTP_DECR_READQ_COUNT();
5124			}
5125		}
5126		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5127		asoc->strmin = NULL;
5128	}
5129	asoc->streamincnt = 0;
5130	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5131#ifdef INVARIANTS
5132		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5133			panic("no net's left alloc'ed, or list points to itself");
5134		}
5135#endif
5136		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5137		sctp_free_remote_addr(net);
5138	}
5139	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5140		/* sa_ignore FREED_MEMORY */
5141		sctp_remove_laddr(laddr);
5142	}
5143
5144	/* pending asconf (address) parameters */
5145	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5146		/* sa_ignore FREED_MEMORY */
5147		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5148		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5149	}
5150	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5151		/* sa_ignore FREED_MEMORY */
5152		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5153		if (aack->data != NULL) {
5154			sctp_m_freem(aack->data);
5155		}
5156		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5157	}
5158	/* clean up auth stuff */
5159	if (asoc->local_hmacs)
5160		sctp_free_hmaclist(asoc->local_hmacs);
5161	if (asoc->peer_hmacs)
5162		sctp_free_hmaclist(asoc->peer_hmacs);
5163
5164	if (asoc->local_auth_chunks)
5165		sctp_free_chunklist(asoc->local_auth_chunks);
5166	if (asoc->peer_auth_chunks)
5167		sctp_free_chunklist(asoc->peer_auth_chunks);
5168
5169	sctp_free_authinfo(&asoc->authinfo);
5170
5171	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5172		LIST_REMOVE(shared_key, next);
5173		sctp_free_sharedkey(shared_key);
5174		/* sa_ignore FREED_MEMORY */
5175	}
5176
5177	/* Insert new items here :> */
5178
5179	/* Get rid of LOCK */
5180	SCTP_TCB_UNLOCK(stcb);
5181	SCTP_TCB_LOCK_DESTROY(stcb);
5182	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5183	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5184		SCTP_INP_INFO_WUNLOCK();
5185		SCTP_INP_RLOCK(inp);
5186	}
5187#ifdef SCTP_TRACK_FREED_ASOCS
5188	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5189		/* now clean up the tasoc itself */
5190		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5191		SCTP_DECR_ASOC_COUNT();
5192	} else {
5193		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5194	}
5195#else
5196	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5197	SCTP_DECR_ASOC_COUNT();
5198#endif
5199	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5200		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5201			/*
5202			 * If its NOT the inp_free calling us AND sctp_close
5203			 * as been called, we call back...
5204			 */
5205			SCTP_INP_RUNLOCK(inp);
5206			/*
5207			 * This will start the kill timer (if we are the
5208			 * last one) since we hold an increment yet. But
5209			 * this is the only safe way to do this since
5210			 * otherwise if the socket closes at the same time
5211			 * we are here we might collide in the cleanup.
5212			 */
5213			sctp_inpcb_free(inp,
5214			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5215			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5216			SCTP_INP_DECR_REF(inp);
5217			goto out_of;
5218		} else {
5219			/* The socket is still open. */
5220			SCTP_INP_DECR_REF(inp);
5221		}
5222	}
5223	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5224		SCTP_INP_RUNLOCK(inp);
5225	}
5226out_of:
5227	/* destroyed the asoc */
5228#ifdef SCTP_LOG_CLOSING
5229	sctp_log_closing(inp, NULL, 11);
5230#endif
5231	return (1);
5232}
5233
5234
5235
5236/*
5237 * determine if a destination is "reachable" based upon the addresses bound
5238 * to the current endpoint (e.g. only v4 or v6 currently bound)
5239 */
5240/*
5241 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5242 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5243 * types bound as its endpoint
5244 */
5245int
5246sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5247{
5248	struct sctp_inpcb *inp;
5249	int answer;
5250
5251	/*
5252	 * No locks here, the TCB, in all cases is already locked and an
5253	 * assoc is up. There is either a INP lock by the caller applied (in
5254	 * asconf case when deleting an address) or NOT in the HB case,
5255	 * however if HB then the INP increment is up and the INP will not
5256	 * be removed (on top of the fact that we have a TCB lock). So we
5257	 * only want to read the sctp_flags, which is either bound-all or
5258	 * not.. no protection needed since once an assoc is up you can't be
5259	 * changing your binding.
5260	 */
5261	inp = stcb->sctp_ep;
5262	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5263		/* if bound all, destination is not restricted */
5264		/*
5265		 * RRS: Question during lock work: Is this correct? If you
5266		 * are bound-all you still might need to obey the V4--V6
5267		 * flags??? IMO this bound-all stuff needs to be removed!
5268		 */
5269		return (1);
5270	}
5271	/* NOTE: all "scope" checks are done when local addresses are added */
5272	switch (destaddr->sa_family) {
5273#ifdef INET6
5274	case AF_INET6:
5275		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5276		break;
5277#endif
5278#ifdef INET
5279	case AF_INET:
5280		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5281		break;
5282#endif
5283	default:
5284		/* invalid family, so it's unreachable */
5285		answer = 0;
5286		break;
5287	}
5288	return (answer);
5289}
5290
5291/*
5292 * update the inp_vflags on an endpoint
5293 */
5294static void
5295sctp_update_ep_vflag(struct sctp_inpcb *inp)
5296{
5297	struct sctp_laddr *laddr;
5298
5299	/* first clear the flag */
5300	inp->ip_inp.inp.inp_vflag = 0;
5301	/* set the flag based on addresses on the ep list */
5302	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5303		if (laddr->ifa == NULL) {
5304			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5305			    __FUNCTION__);
5306			continue;
5307		}
5308		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5309			continue;
5310		}
5311		switch (laddr->ifa->address.sa.sa_family) {
5312#ifdef INET6
5313		case AF_INET6:
5314			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5315			break;
5316#endif
5317#ifdef INET
5318		case AF_INET:
5319			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5320			break;
5321#endif
5322		default:
5323			break;
5324		}
5325	}
5326}
5327
5328/*
5329 * Add the address to the endpoint local address list There is nothing to be
5330 * done if we are bound to all addresses
5331 */
5332void
5333sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5334{
5335	struct sctp_laddr *laddr;
5336	int fnd, error = 0;
5337
5338	fnd = 0;
5339
5340	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5341		/* You are already bound to all. You have it already */
5342		return;
5343	}
5344#ifdef INET6
5345	if (ifa->address.sa.sa_family == AF_INET6) {
5346		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5347			/* Can't bind a non-useable addr. */
5348			return;
5349		}
5350	}
5351#endif
5352	/* first, is it already present? */
5353	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5354		if (laddr->ifa == ifa) {
5355			fnd = 1;
5356			break;
5357		}
5358	}
5359
5360	if (fnd == 0) {
5361		/* Not in the ep list */
5362		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5363		if (error != 0)
5364			return;
5365		inp->laddr_count++;
5366		/* update inp_vflag flags */
5367		switch (ifa->address.sa.sa_family) {
5368#ifdef INET6
5369		case AF_INET6:
5370			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5371			break;
5372#endif
5373#ifdef INET
5374		case AF_INET:
5375			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5376			break;
5377#endif
5378		default:
5379			break;
5380		}
5381	}
5382	return;
5383}
5384
5385
5386/*
5387 * select a new (hopefully reachable) destination net (should only be used
5388 * when we deleted an ep addr that is the only usable source address to reach
5389 * the destination net)
5390 */
5391static void
5392sctp_select_primary_destination(struct sctp_tcb *stcb)
5393{
5394	struct sctp_nets *net;
5395
5396	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5397		/* for now, we'll just pick the first reachable one we find */
5398		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5399			continue;
5400		if (sctp_destination_is_reachable(stcb,
5401		    (struct sockaddr *)&net->ro._l_addr)) {
5402			/* found a reachable destination */
5403			stcb->asoc.primary_destination = net;
5404		}
5405	}
5406	/* I can't there from here! ...we're gonna die shortly... */
5407}
5408
5409
5410/*
5411 * Delete the address from the endpoint local address list There is nothing
5412 * to be done if we are bound to all addresses
5413 */
5414void
5415sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5416{
5417	struct sctp_laddr *laddr;
5418	int fnd;
5419
5420	fnd = 0;
5421	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5422		/* You are already bound to all. You have it already */
5423		return;
5424	}
5425	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5426		if (laddr->ifa == ifa) {
5427			fnd = 1;
5428			break;
5429		}
5430	}
5431	if (fnd && (inp->laddr_count < 2)) {
5432		/* can't delete unless there are at LEAST 2 addresses */
5433		return;
5434	}
5435	if (fnd) {
5436		/*
5437		 * clean up any use of this address go through our
5438		 * associations and clear any last_used_address that match
5439		 * this one for each assoc, see if a new primary_destination
5440		 * is needed
5441		 */
5442		struct sctp_tcb *stcb;
5443
5444		/* clean up "next_addr_touse" */
5445		if (inp->next_addr_touse == laddr)
5446			/* delete this address */
5447			inp->next_addr_touse = NULL;
5448
5449		/* clean up "last_used_address" */
5450		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5451			struct sctp_nets *net;
5452
5453			SCTP_TCB_LOCK(stcb);
5454			if (stcb->asoc.last_used_address == laddr)
5455				/* delete this address */
5456				stcb->asoc.last_used_address = NULL;
5457			/*
5458			 * Now spin through all the nets and purge any ref
5459			 * to laddr
5460			 */
5461			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5462				if (net->ro._s_addr &&
5463				    (net->ro._s_addr->ifa == laddr->ifa)) {
5464					/* Yep, purge src address selected */
5465					sctp_rtentry_t *rt;
5466
5467					/* delete this address if cached */
5468					rt = net->ro.ro_rt;
5469					if (rt != NULL) {
5470						RTFREE(rt);
5471						net->ro.ro_rt = NULL;
5472					}
5473					sctp_free_ifa(net->ro._s_addr);
5474					net->ro._s_addr = NULL;
5475					net->src_addr_selected = 0;
5476				}
5477			}
5478			SCTP_TCB_UNLOCK(stcb);
5479		}		/* for each tcb */
5480		/* remove it from the ep list */
5481		sctp_remove_laddr(laddr);
5482		inp->laddr_count--;
5483		/* update inp_vflag flags */
5484		sctp_update_ep_vflag(inp);
5485	}
5486	return;
5487}
5488
5489/*
5490 * Add the address to the TCB local address restricted list.
5491 * This is a "pending" address list (eg. addresses waiting for an
5492 * ASCONF-ACK response) and cannot be used as a valid source address.
5493 */
5494void
5495sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5496{
5497	struct sctp_laddr *laddr;
5498	struct sctpladdr *list;
5499
5500	/*
5501	 * Assumes TCB is locked.. and possibly the INP. May need to
5502	 * confirm/fix that if we need it and is not the case.
5503	 */
5504	list = &stcb->asoc.sctp_restricted_addrs;
5505
5506#ifdef INET6
5507	if (ifa->address.sa.sa_family == AF_INET6) {
5508		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5509			/* Can't bind a non-existent addr. */
5510			return;
5511		}
5512	}
5513#endif
5514	/* does the address already exist? */
5515	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5516		if (laddr->ifa == ifa) {
5517			return;
5518		}
5519	}
5520
5521	/* add to the list */
5522	(void)sctp_insert_laddr(list, ifa, 0);
5523	return;
5524}
5525
5526/*
5527 * insert an laddr entry with the given ifa for the desired list
5528 */
5529int
5530sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5531{
5532	struct sctp_laddr *laddr;
5533
5534	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5535	if (laddr == NULL) {
5536		/* out of memory? */
5537		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5538		return (EINVAL);
5539	}
5540	SCTP_INCR_LADDR_COUNT();
5541	bzero(laddr, sizeof(*laddr));
5542	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5543	laddr->ifa = ifa;
5544	laddr->action = act;
5545	atomic_add_int(&ifa->refcount, 1);
5546	/* insert it */
5547	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5548
5549	return (0);
5550}
5551
5552/*
5553 * Remove an laddr entry from the local address list (on an assoc)
5554 */
5555void
5556sctp_remove_laddr(struct sctp_laddr *laddr)
5557{
5558
5559	/* remove from the list */
5560	LIST_REMOVE(laddr, sctp_nxt_addr);
5561	sctp_free_ifa(laddr->ifa);
5562	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5563	SCTP_DECR_LADDR_COUNT();
5564}
5565
5566/*
5567 * Remove a local address from the TCB local address restricted list
5568 */
5569void
5570sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5571{
5572	struct sctp_inpcb *inp;
5573	struct sctp_laddr *laddr;
5574
5575	/*
5576	 * This is called by asconf work. It is assumed that a) The TCB is
5577	 * locked and b) The INP is locked. This is true in as much as I can
5578	 * trace through the entry asconf code where I did these locks.
5579	 * Again, the ASCONF code is a bit different in that it does lock
5580	 * the INP during its work often times. This must be since we don't
5581	 * want other proc's looking up things while what they are looking
5582	 * up is changing :-D
5583	 */
5584
5585	inp = stcb->sctp_ep;
5586	/* if subset bound and don't allow ASCONF's, can't delete last */
5587	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5588	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5589		if (stcb->sctp_ep->laddr_count < 2) {
5590			/* can't delete last address */
5591			return;
5592		}
5593	}
5594	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5595		/* remove the address if it exists */
5596		if (laddr->ifa == NULL)
5597			continue;
5598		if (laddr->ifa == ifa) {
5599			sctp_remove_laddr(laddr);
5600			return;
5601		}
5602	}
5603
5604	/* address not found! */
5605	return;
5606}
5607
5608/*
5609 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5610 */
5611/* sysctl */
5612static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5613static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5614
5615
5616
5617#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5618struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5619int *sctp_cpuarry = NULL;
5620void
5621sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5622{
5623	/* Queue a packet to a processor for the specified core */
5624	struct sctp_mcore_queue *qent;
5625	struct sctp_mcore_ctrl *wkq;
5626	int need_wake = 0;
5627
5628	if (sctp_mcore_workers == NULL) {
5629		/* Something went way bad during setup */
5630		sctp_input_with_port(m, off, 0);
5631		return;
5632	}
5633	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5634	    (sizeof(struct sctp_mcore_queue)),
5635	    SCTP_M_MCORE);
5636	if (qent == NULL) {
5637		/* This is trouble  */
5638		sctp_input_with_port(m, off, 0);
5639		return;
5640	}
5641	qent->vn = curvnet;
5642	qent->m = m;
5643	qent->off = off;
5644	qent->v6 = 0;
5645	wkq = &sctp_mcore_workers[cpu_to_use];
5646	SCTP_MCORE_QLOCK(wkq);
5647
5648	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5649	if (wkq->running == 0) {
5650		need_wake = 1;
5651	}
5652	SCTP_MCORE_QUNLOCK(wkq);
5653	if (need_wake) {
5654		wakeup(&wkq->running);
5655	}
5656}
5657
5658static void
5659sctp_mcore_thread(void *arg)
5660{
5661
5662	struct sctp_mcore_ctrl *wkq;
5663	struct sctp_mcore_queue *qent;
5664
5665	wkq = (struct sctp_mcore_ctrl *)arg;
5666	struct mbuf *m;
5667	int off, v6;
5668
5669	/* Wait for first tickle */
5670	SCTP_MCORE_LOCK(wkq);
5671	wkq->running = 0;
5672	msleep(&wkq->running,
5673	    &wkq->core_mtx,
5674	    0, "wait for pkt", 0);
5675	SCTP_MCORE_UNLOCK(wkq);
5676
5677	/* Bind to our cpu */
5678	thread_lock(curthread);
5679	sched_bind(curthread, wkq->cpuid);
5680	thread_unlock(curthread);
5681
5682	/* Now lets start working */
5683	SCTP_MCORE_LOCK(wkq);
5684	/* Now grab lock and go */
5685	for (;;) {
5686		SCTP_MCORE_QLOCK(wkq);
5687skip_sleep:
5688		wkq->running = 1;
5689		qent = TAILQ_FIRST(&wkq->que);
5690		if (qent) {
5691			TAILQ_REMOVE(&wkq->que, qent, next);
5692			SCTP_MCORE_QUNLOCK(wkq);
5693			CURVNET_SET(qent->vn);
5694			m = qent->m;
5695			off = qent->off;
5696			v6 = qent->v6;
5697			SCTP_FREE(qent, SCTP_M_MCORE);
5698			if (v6 == 0) {
5699				sctp_input_with_port(m, off, 0);
5700			} else {
5701				SCTP_PRINTF("V6 not yet supported\n");
5702				sctp_m_freem(m);
5703			}
5704			CURVNET_RESTORE();
5705			SCTP_MCORE_QLOCK(wkq);
5706		}
5707		wkq->running = 0;
5708		if (!TAILQ_EMPTY(&wkq->que)) {
5709			goto skip_sleep;
5710		}
5711		SCTP_MCORE_QUNLOCK(wkq);
5712		msleep(&wkq->running,
5713		    &wkq->core_mtx,
5714		    0, "wait for pkt", 0);
5715	}
5716}
5717
5718static void
5719sctp_startup_mcore_threads(void)
5720{
5721	int i, cpu;
5722
5723	if (mp_ncpus == 1)
5724		return;
5725
5726	if (sctp_mcore_workers != NULL) {
5727		/*
5728		 * Already been here in some previous vnet?
5729		 */
5730		return;
5731	}
5732	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5733	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5734	    SCTP_M_MCORE);
5735	if (sctp_mcore_workers == NULL) {
5736		/* TSNH I hope */
5737		return;
5738	}
5739	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5740	    sizeof(struct sctp_mcore_ctrl)));
5741	/* Init the structures */
5742	for (i = 0; i <= mp_maxid; i++) {
5743		TAILQ_INIT(&sctp_mcore_workers[i].que);
5744		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5745		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5746		sctp_mcore_workers[i].cpuid = i;
5747	}
5748	if (sctp_cpuarry == NULL) {
5749		SCTP_MALLOC(sctp_cpuarry, int *,
5750		    (mp_ncpus * sizeof(int)),
5751		    SCTP_M_MCORE);
5752		i = 0;
5753		CPU_FOREACH(cpu) {
5754			sctp_cpuarry[i] = cpu;
5755			i++;
5756		}
5757	}
5758	/* Now start them all */
5759	CPU_FOREACH(cpu) {
5760		(void)kproc_create(sctp_mcore_thread,
5761		    (void *)&sctp_mcore_workers[cpu],
5762		    &sctp_mcore_workers[cpu].thread_proc,
5763		    RFPROC,
5764		    SCTP_KTHREAD_PAGES,
5765		    SCTP_MCORE_NAME);
5766
5767	}
5768}
5769
5770#endif
5771
5772void
5773sctp_pcb_init()
5774{
5775	/*
5776	 * SCTP initialization for the PCB structures should be called by
5777	 * the sctp_init() funciton.
5778	 */
5779	int i;
5780	struct timeval tv;
5781
5782	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5783		/* error I was called twice */
5784		return;
5785	}
5786	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5787
5788#if defined(SCTP_LOCAL_TRACE_BUF)
5789	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5790#endif
5791#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5792	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5793	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5794	    SCTP_M_MCORE);
5795#endif
5796	(void)SCTP_GETTIME_TIMEVAL(&tv);
5797#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5798	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5799	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5800	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5801#else
5802	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5803	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5804	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5805#endif
5806	/* init the empty list of (All) Endpoints */
5807	LIST_INIT(&SCTP_BASE_INFO(listhead));
5808
5809
5810	/* init the hash table of endpoints */
5811	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5812	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5813	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5814	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5815	    &SCTP_BASE_INFO(hashasocmark));
5816	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5817	    &SCTP_BASE_INFO(hashmark));
5818	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5819	    &SCTP_BASE_INFO(hashtcpmark));
5820	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5821
5822
5823	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5824	    &SCTP_BASE_INFO(hashvrfmark));
5825
5826	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5827	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5828	/* init the zones */
5829	/*
5830	 * FIX ME: Should check for NULL returns, but if it does fail we are
5831	 * doomed to panic anyways... add later maybe.
5832	 */
5833	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5834	    sizeof(struct sctp_inpcb), maxsockets);
5835
5836	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5837	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5838
5839	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5840	    sizeof(struct sctp_laddr),
5841	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5842
5843	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5844	    sizeof(struct sctp_nets),
5845	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5846
5847	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5848	    sizeof(struct sctp_tmit_chunk),
5849	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5850
5851	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5852	    sizeof(struct sctp_queued_to_read),
5853	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5854
5855	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5856	    sizeof(struct sctp_stream_queue_pending),
5857	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5858
5859	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5860	    sizeof(struct sctp_asconf),
5861	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5862
5863	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5864	    sizeof(struct sctp_asconf_ack),
5865	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5866
5867
5868	/* Master Lock INIT for info structure */
5869	SCTP_INP_INFO_LOCK_INIT();
5870	SCTP_STATLOG_INIT_LOCK();
5871
5872	SCTP_IPI_COUNT_INIT();
5873	SCTP_IPI_ADDR_INIT();
5874#ifdef SCTP_PACKET_LOGGING
5875	SCTP_IP_PKTLOG_INIT();
5876#endif
5877	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5878
5879	SCTP_WQ_ADDR_INIT();
5880	/* not sure if we need all the counts */
5881	SCTP_BASE_INFO(ipi_count_ep) = 0;
5882	/* assoc/tcb zone info */
5883	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5884	/* local addrlist zone info */
5885	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5886	/* remote addrlist zone info */
5887	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5888	/* chunk info */
5889	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5890
5891	/* socket queue zone info */
5892	SCTP_BASE_INFO(ipi_count_readq) = 0;
5893
5894	/* stream out queue cont */
5895	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5896
5897	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5898	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5899
5900	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5901
5902	/* Init the TIMEWAIT list */
5903	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5904		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5905	}
5906	sctp_startup_iterator();
5907
5908#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5909	sctp_startup_mcore_threads();
5910#endif
5911
5912	/*
5913	 * INIT the default VRF which for BSD is the only one, other O/S's
5914	 * may have more. But initially they must start with one and then
5915	 * add the VRF's as addresses are added.
5916	 */
5917	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5918}
5919
5920/*
5921 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5922 */
5923void
5924sctp_pcb_finish(void)
5925{
5926	struct sctp_vrflist *vrf_bucket;
5927	struct sctp_vrf *vrf, *nvrf;
5928	struct sctp_ifn *ifn, *nifn;
5929	struct sctp_ifa *ifa, *nifa;
5930	struct sctpvtaghead *chain;
5931	struct sctp_tagblock *twait_block, *prev_twait_block;
5932	struct sctp_laddr *wi, *nwi;
5933	int i;
5934	struct sctp_iterator *it, *nit;
5935
5936	/*
5937	 * In FreeBSD the iterator thread never exits but we do clean up.
5938	 * The only way FreeBSD reaches here is if we have VRF's but we
5939	 * still add the ifdef to make it compile on old versions.
5940	 */
5941	SCTP_IPI_ITERATOR_WQ_LOCK();
5942	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5943		if (it->vn != curvnet) {
5944			continue;
5945		}
5946		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5947		if (it->function_atend != NULL) {
5948			(*it->function_atend) (it->pointer, it->val);
5949		}
5950		SCTP_FREE(it, SCTP_M_ITER);
5951	}
5952	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5953	SCTP_ITERATOR_LOCK();
5954	if ((sctp_it_ctl.cur_it) &&
5955	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5956		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5957	}
5958	SCTP_ITERATOR_UNLOCK();
5959	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5960	SCTP_WQ_ADDR_LOCK();
5961	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5962		LIST_REMOVE(wi, sctp_nxt_addr);
5963		SCTP_DECR_LADDR_COUNT();
5964		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5965	}
5966	SCTP_WQ_ADDR_UNLOCK();
5967
5968	/*
5969	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5970	 * destroyed first).
5971	 */
5972	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5973	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5974		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5975			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5976				/* free the ifa */
5977				LIST_REMOVE(ifa, next_bucket);
5978				LIST_REMOVE(ifa, next_ifa);
5979				SCTP_FREE(ifa, SCTP_M_IFA);
5980			}
5981			/* free the ifn */
5982			LIST_REMOVE(ifn, next_bucket);
5983			LIST_REMOVE(ifn, next_ifn);
5984			SCTP_FREE(ifn, SCTP_M_IFN);
5985		}
5986		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5987		/* free the vrf */
5988		LIST_REMOVE(vrf, next_vrf);
5989		SCTP_FREE(vrf, SCTP_M_VRF);
5990	}
5991	/* free the vrf hashes */
5992	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5993	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5994
5995	/*
5996	 * free the TIMEWAIT list elements malloc'd in the function
5997	 * sctp_add_vtag_to_timewait()...
5998	 */
5999	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
6000		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
6001		if (!LIST_EMPTY(chain)) {
6002			prev_twait_block = NULL;
6003			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6004				if (prev_twait_block) {
6005					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6006				}
6007				prev_twait_block = twait_block;
6008			}
6009			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6010		}
6011	}
6012
6013	/* free the locks and mutexes */
6014#ifdef SCTP_PACKET_LOGGING
6015	SCTP_IP_PKTLOG_DESTROY();
6016#endif
6017	SCTP_IPI_ADDR_DESTROY();
6018	SCTP_STATLOG_DESTROY();
6019	SCTP_INP_INFO_LOCK_DESTROY();
6020
6021	SCTP_WQ_ADDR_DESTROY();
6022
6023	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6024	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6025	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6026	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6027	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6028	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6029	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6030	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6031	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6032	/* Get rid of other stuff to */
6033	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
6034		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
6035	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
6036		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6037	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6038		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6039#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6040	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6041#endif
6042}
6043
6044
6045int
6046sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6047    int offset, int limit,
6048    struct sockaddr *src, struct sockaddr *dst,
6049    struct sockaddr *altsa)
6050{
6051	/*
6052	 * grub through the INIT pulling addresses and loading them to the
6053	 * nets structure in the asoc. The from address in the mbuf should
6054	 * also be loaded (if it is not already). This routine can be called
6055	 * with either INIT or INIT-ACK's as long as the m points to the IP
6056	 * packet and the offset points to the beginning of the parameters.
6057	 */
6058	struct sctp_inpcb *inp;
6059	struct sctp_nets *net, *nnet, *net_tmp;
6060	struct sctp_paramhdr *phdr, parm_buf;
6061	struct sctp_tcb *stcb_tmp;
6062	uint16_t ptype, plen;
6063	struct sockaddr *sa;
6064	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6065	struct sctp_auth_random *p_random = NULL;
6066	uint16_t random_len = 0;
6067	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6068	struct sctp_auth_hmac_algo *hmacs = NULL;
6069	uint16_t hmacs_len = 0;
6070	uint8_t saw_asconf = 0;
6071	uint8_t saw_asconf_ack = 0;
6072	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6073	struct sctp_auth_chunk_list *chunks = NULL;
6074	uint16_t num_chunks = 0;
6075	sctp_key_t *new_key;
6076	uint32_t keylen;
6077	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6078	uint8_t ecn_allowed;
6079
6080#ifdef INET
6081	struct sockaddr_in sin;
6082
6083#endif
6084#ifdef INET6
6085	struct sockaddr_in6 sin6;
6086
6087#endif
6088
6089	/* First get the destination address setup too. */
6090#ifdef INET
6091	memset(&sin, 0, sizeof(sin));
6092	sin.sin_family = AF_INET;
6093	sin.sin_len = sizeof(sin);
6094	sin.sin_port = stcb->rport;
6095#endif
6096#ifdef INET6
6097	memset(&sin6, 0, sizeof(sin6));
6098	sin6.sin6_family = AF_INET6;
6099	sin6.sin6_len = sizeof(struct sockaddr_in6);
6100	sin6.sin6_port = stcb->rport;
6101#endif
6102	if (altsa) {
6103		sa = altsa;
6104	} else {
6105		sa = src;
6106	}
6107	/* Turn off ECN until we get through all params */
6108	ecn_allowed = 0;
6109	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6110		/* mark all addresses that we have currently on the list */
6111		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6112	}
6113	/* does the source address already exist? if so skip it */
6114	inp = stcb->sctp_ep;
6115	atomic_add_int(&stcb->asoc.refcnt, 1);
6116	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6117	atomic_add_int(&stcb->asoc.refcnt, -1);
6118
6119	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6120		/* we must add the source address */
6121		/* no scope set here since we have a tcb already. */
6122		switch (sa->sa_family) {
6123#ifdef INET
6124		case AF_INET:
6125			if (stcb->asoc.scope.ipv4_addr_legal) {
6126				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6127					return (-1);
6128				}
6129			}
6130			break;
6131#endif
6132#ifdef INET6
6133		case AF_INET6:
6134			if (stcb->asoc.scope.ipv6_addr_legal) {
6135				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6136					return (-2);
6137				}
6138			}
6139			break;
6140#endif
6141		default:
6142			break;
6143		}
6144	} else {
6145		if (net_tmp != NULL && stcb_tmp == stcb) {
6146			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6147		} else if (stcb_tmp != stcb) {
6148			/* It belongs to another association? */
6149			if (stcb_tmp)
6150				SCTP_TCB_UNLOCK(stcb_tmp);
6151			return (-3);
6152		}
6153	}
6154	if (stcb->asoc.state == 0) {
6155		/* the assoc was freed? */
6156		return (-4);
6157	}
6158	/*
6159	 * peer must explicitly turn this on. This may have been initialized
6160	 * to be "on" in order to allow local addr changes while INIT's are
6161	 * in flight.
6162	 */
6163	stcb->asoc.peer_supports_asconf = 0;
6164	/* now we must go through each of the params. */
6165	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6166	while (phdr) {
6167		ptype = ntohs(phdr->param_type);
6168		plen = ntohs(phdr->param_length);
6169		/*
6170		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6171		 * (uint32_t)ptype, (int)plen);
6172		 */
6173		if (offset + plen > limit) {
6174			break;
6175		}
6176		if (plen == 0) {
6177			break;
6178		}
6179#ifdef INET
6180		if (ptype == SCTP_IPV4_ADDRESS) {
6181			if (stcb->asoc.scope.ipv4_addr_legal) {
6182				struct sctp_ipv4addr_param *p4, p4_buf;
6183
6184				/* ok get the v4 address and check/add */
6185				phdr = sctp_get_next_param(m, offset,
6186				    (struct sctp_paramhdr *)&p4_buf,
6187				    sizeof(p4_buf));
6188				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6189				    phdr == NULL) {
6190					return (-5);
6191				}
6192				p4 = (struct sctp_ipv4addr_param *)phdr;
6193				sin.sin_addr.s_addr = p4->addr;
6194				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6195					/* Skip multi-cast addresses */
6196					goto next_param;
6197				}
6198				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6199				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6200					goto next_param;
6201				}
6202				sa = (struct sockaddr *)&sin;
6203				inp = stcb->sctp_ep;
6204				atomic_add_int(&stcb->asoc.refcnt, 1);
6205				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6206				    dst, stcb);
6207				atomic_add_int(&stcb->asoc.refcnt, -1);
6208
6209				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6210				    inp == NULL) {
6211					/* we must add the source address */
6212					/*
6213					 * no scope set since we have a tcb
6214					 * already
6215					 */
6216
6217					/*
6218					 * we must validate the state again
6219					 * here
6220					 */
6221			add_it_now:
6222					if (stcb->asoc.state == 0) {
6223						/* the assoc was freed? */
6224						return (-7);
6225					}
6226					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6227						return (-8);
6228					}
6229				} else if (stcb_tmp == stcb) {
6230					if (stcb->asoc.state == 0) {
6231						/* the assoc was freed? */
6232						return (-10);
6233					}
6234					if (net != NULL) {
6235						/* clear flag */
6236						net->dest_state &=
6237						    ~SCTP_ADDR_NOT_IN_ASSOC;
6238					}
6239				} else {
6240					/*
6241					 * strange, address is in another
6242					 * assoc? straighten out locks.
6243					 */
6244					if (stcb_tmp) {
6245						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6246							/*
6247							 * in setup state we
6248							 * abort this guy
6249							 */
6250							sctp_abort_an_association(stcb_tmp->sctp_ep,
6251							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6252							goto add_it_now;
6253						}
6254						SCTP_TCB_UNLOCK(stcb_tmp);
6255					}
6256					if (stcb->asoc.state == 0) {
6257						/* the assoc was freed? */
6258						return (-12);
6259					}
6260					return (-13);
6261				}
6262			}
6263		} else
6264#endif
6265#ifdef INET6
6266		if (ptype == SCTP_IPV6_ADDRESS) {
6267			if (stcb->asoc.scope.ipv6_addr_legal) {
6268				/* ok get the v6 address and check/add */
6269				struct sctp_ipv6addr_param *p6, p6_buf;
6270
6271				phdr = sctp_get_next_param(m, offset,
6272				    (struct sctp_paramhdr *)&p6_buf,
6273				    sizeof(p6_buf));
6274				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6275				    phdr == NULL) {
6276					return (-14);
6277				}
6278				p6 = (struct sctp_ipv6addr_param *)phdr;
6279				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6280				    sizeof(p6->addr));
6281				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6282					/* Skip multi-cast addresses */
6283					goto next_param;
6284				}
6285				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6286					/*
6287					 * Link local make no sense without
6288					 * scope
6289					 */
6290					goto next_param;
6291				}
6292				sa = (struct sockaddr *)&sin6;
6293				inp = stcb->sctp_ep;
6294				atomic_add_int(&stcb->asoc.refcnt, 1);
6295				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6296				    dst, stcb);
6297				atomic_add_int(&stcb->asoc.refcnt, -1);
6298				if (stcb_tmp == NULL &&
6299				    (inp == stcb->sctp_ep || inp == NULL)) {
6300					/*
6301					 * we must validate the state again
6302					 * here
6303					 */
6304			add_it_now6:
6305					if (stcb->asoc.state == 0) {
6306						/* the assoc was freed? */
6307						return (-16);
6308					}
6309					/*
6310					 * we must add the address, no scope
6311					 * set
6312					 */
6313					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6314						return (-17);
6315					}
6316				} else if (stcb_tmp == stcb) {
6317					/*
6318					 * we must validate the state again
6319					 * here
6320					 */
6321					if (stcb->asoc.state == 0) {
6322						/* the assoc was freed? */
6323						return (-19);
6324					}
6325					if (net != NULL) {
6326						/* clear flag */
6327						net->dest_state &=
6328						    ~SCTP_ADDR_NOT_IN_ASSOC;
6329					}
6330				} else {
6331					/*
6332					 * strange, address is in another
6333					 * assoc? straighten out locks.
6334					 */
6335					if (stcb_tmp)
6336						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6337							/*
6338							 * in setup state we
6339							 * abort this guy
6340							 */
6341							sctp_abort_an_association(stcb_tmp->sctp_ep,
6342							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6343							goto add_it_now6;
6344						}
6345					SCTP_TCB_UNLOCK(stcb_tmp);
6346
6347					if (stcb->asoc.state == 0) {
6348						/* the assoc was freed? */
6349						return (-21);
6350					}
6351					return (-22);
6352				}
6353			}
6354		} else
6355#endif
6356		if (ptype == SCTP_ECN_CAPABLE) {
6357			ecn_allowed = 1;
6358		} else if (ptype == SCTP_ULP_ADAPTATION) {
6359			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6360				struct sctp_adaptation_layer_indication ai,
6361				                                *aip;
6362
6363				phdr = sctp_get_next_param(m, offset,
6364				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6365				aip = (struct sctp_adaptation_layer_indication *)phdr;
6366				if (aip) {
6367					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6368					stcb->asoc.adaptation_needed = 1;
6369				}
6370			}
6371		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6372			struct sctp_asconf_addr_param lstore, *fee;
6373			int lptype;
6374			struct sockaddr *lsa = NULL;
6375
6376#ifdef INET
6377			struct sctp_asconf_addrv4_param *fii;
6378
6379#endif
6380
6381			stcb->asoc.peer_supports_asconf = 1;
6382			if (plen > sizeof(lstore)) {
6383				return (-23);
6384			}
6385			phdr = sctp_get_next_param(m, offset,
6386			    (struct sctp_paramhdr *)&lstore,
6387			    min(plen, sizeof(lstore)));
6388			if (phdr == NULL) {
6389				return (-24);
6390			}
6391			fee = (struct sctp_asconf_addr_param *)phdr;
6392			lptype = ntohs(fee->addrp.ph.param_type);
6393			switch (lptype) {
6394#ifdef INET
6395			case SCTP_IPV4_ADDRESS:
6396				if (plen !=
6397				    sizeof(struct sctp_asconf_addrv4_param)) {
6398					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6399					    (int)sizeof(struct sctp_asconf_addrv4_param),
6400					    plen);
6401				} else {
6402					fii = (struct sctp_asconf_addrv4_param *)fee;
6403					sin.sin_addr.s_addr = fii->addrp.addr;
6404					lsa = (struct sockaddr *)&sin;
6405				}
6406				break;
6407#endif
6408#ifdef INET6
6409			case SCTP_IPV6_ADDRESS:
6410				if (plen !=
6411				    sizeof(struct sctp_asconf_addr_param)) {
6412					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6413					    (int)sizeof(struct sctp_asconf_addr_param),
6414					    plen);
6415				} else {
6416					memcpy(sin6.sin6_addr.s6_addr,
6417					    fee->addrp.addr,
6418					    sizeof(fee->addrp.addr));
6419					lsa = (struct sockaddr *)&sin6;
6420				}
6421				break;
6422#endif
6423			default:
6424				break;
6425			}
6426			if (lsa) {
6427				(void)sctp_set_primary_addr(stcb, sa, NULL);
6428			}
6429		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6430			stcb->asoc.peer_supports_nat = 1;
6431		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6432			/* Peer supports pr-sctp */
6433			stcb->asoc.peer_supports_prsctp = 1;
6434		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6435			/* A supported extension chunk */
6436			struct sctp_supported_chunk_types_param *pr_supported;
6437			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6438			int num_ent, i;
6439
6440			phdr = sctp_get_next_param(m, offset,
6441			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6442			if (phdr == NULL) {
6443				return (-25);
6444			}
6445			stcb->asoc.peer_supports_asconf = 0;
6446			stcb->asoc.peer_supports_prsctp = 0;
6447			stcb->asoc.peer_supports_pktdrop = 0;
6448			stcb->asoc.peer_supports_strreset = 0;
6449			stcb->asoc.peer_supports_nr_sack = 0;
6450			stcb->asoc.peer_supports_auth = 0;
6451			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6452			num_ent = plen - sizeof(struct sctp_paramhdr);
6453			for (i = 0; i < num_ent; i++) {
6454				switch (pr_supported->chunk_types[i]) {
6455				case SCTP_ASCONF:
6456				case SCTP_ASCONF_ACK:
6457					stcb->asoc.peer_supports_asconf = 1;
6458					break;
6459				case SCTP_FORWARD_CUM_TSN:
6460					stcb->asoc.peer_supports_prsctp = 1;
6461					break;
6462				case SCTP_PACKET_DROPPED:
6463					stcb->asoc.peer_supports_pktdrop = 1;
6464					break;
6465				case SCTP_NR_SELECTIVE_ACK:
6466					stcb->asoc.peer_supports_nr_sack = 1;
6467					break;
6468				case SCTP_STREAM_RESET:
6469					stcb->asoc.peer_supports_strreset = 1;
6470					break;
6471				case SCTP_AUTHENTICATION:
6472					stcb->asoc.peer_supports_auth = 1;
6473					break;
6474				default:
6475					/* one I have not learned yet */
6476					break;
6477
6478				}
6479			}
6480		} else if (ptype == SCTP_RANDOM) {
6481			if (plen > sizeof(random_store))
6482				break;
6483			if (got_random) {
6484				/* already processed a RANDOM */
6485				goto next_param;
6486			}
6487			phdr = sctp_get_next_param(m, offset,
6488			    (struct sctp_paramhdr *)random_store,
6489			    min(sizeof(random_store), plen));
6490			if (phdr == NULL)
6491				return (-26);
6492			p_random = (struct sctp_auth_random *)phdr;
6493			random_len = plen - sizeof(*p_random);
6494			/* enforce the random length */
6495			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6496				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6497				return (-27);
6498			}
6499			got_random = 1;
6500		} else if (ptype == SCTP_HMAC_LIST) {
6501			int num_hmacs;
6502			int i;
6503
6504			if (plen > sizeof(hmacs_store))
6505				break;
6506			if (got_hmacs) {
6507				/* already processed a HMAC list */
6508				goto next_param;
6509			}
6510			phdr = sctp_get_next_param(m, offset,
6511			    (struct sctp_paramhdr *)hmacs_store,
6512			    min(plen, sizeof(hmacs_store)));
6513			if (phdr == NULL)
6514				return (-28);
6515			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6516			hmacs_len = plen - sizeof(*hmacs);
6517			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6518			/* validate the hmac list */
6519			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6520				return (-29);
6521			}
6522			if (stcb->asoc.peer_hmacs != NULL)
6523				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6524			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6525			if (stcb->asoc.peer_hmacs != NULL) {
6526				for (i = 0; i < num_hmacs; i++) {
6527					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6528					    ntohs(hmacs->hmac_ids[i]));
6529				}
6530			}
6531			got_hmacs = 1;
6532		} else if (ptype == SCTP_CHUNK_LIST) {
6533			int i;
6534
6535			if (plen > sizeof(chunks_store))
6536				break;
6537			if (got_chklist) {
6538				/* already processed a Chunks list */
6539				goto next_param;
6540			}
6541			phdr = sctp_get_next_param(m, offset,
6542			    (struct sctp_paramhdr *)chunks_store,
6543			    min(plen, sizeof(chunks_store)));
6544			if (phdr == NULL)
6545				return (-30);
6546			chunks = (struct sctp_auth_chunk_list *)phdr;
6547			num_chunks = plen - sizeof(*chunks);
6548			if (stcb->asoc.peer_auth_chunks != NULL)
6549				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6550			else
6551				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6552			for (i = 0; i < num_chunks; i++) {
6553				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6554				    stcb->asoc.peer_auth_chunks);
6555				/* record asconf/asconf-ack if listed */
6556				if (chunks->chunk_types[i] == SCTP_ASCONF)
6557					saw_asconf = 1;
6558				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6559					saw_asconf_ack = 1;
6560
6561			}
6562			got_chklist = 1;
6563		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6564			    (ptype == SCTP_STATE_COOKIE) ||
6565			    (ptype == SCTP_UNRECOG_PARAM) ||
6566			    (ptype == SCTP_COOKIE_PRESERVE) ||
6567			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6568			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6569			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6570			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6571		    (ptype == SCTP_SUCCESS_REPORT)) {
6572			 /* don't care */ ;
6573		} else {
6574			if ((ptype & 0x8000) == 0x0000) {
6575				/*
6576				 * must stop processing the rest of the
6577				 * param's. Any report bits were handled
6578				 * with the call to
6579				 * sctp_arethere_unrecognized_parameters()
6580				 * when the INIT or INIT-ACK was first seen.
6581				 */
6582				break;
6583			}
6584		}
6585
6586next_param:
6587		offset += SCTP_SIZE32(plen);
6588		if (offset >= limit) {
6589			break;
6590		}
6591		phdr = sctp_get_next_param(m, offset, &parm_buf,
6592		    sizeof(parm_buf));
6593	}
6594	/* Now check to see if we need to purge any addresses */
6595	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6596		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6597		    SCTP_ADDR_NOT_IN_ASSOC) {
6598			/* This address has been removed from the asoc */
6599			/* remove and free it */
6600			stcb->asoc.numnets--;
6601			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6602			sctp_free_remote_addr(net);
6603			if (net == stcb->asoc.primary_destination) {
6604				stcb->asoc.primary_destination = NULL;
6605				sctp_select_primary_destination(stcb);
6606			}
6607		}
6608	}
6609	if (ecn_allowed == 0) {
6610		stcb->asoc.ecn_allowed = 0;
6611	}
6612	/* validate authentication required parameters */
6613	if (got_random && got_hmacs) {
6614		stcb->asoc.peer_supports_auth = 1;
6615	} else {
6616		stcb->asoc.peer_supports_auth = 0;
6617	}
6618	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6619		/* peer does not support auth but sent a chunks list? */
6620		return (-31);
6621	}
6622	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6623	    !stcb->asoc.peer_supports_auth) {
6624		/* peer supports asconf but not auth? */
6625		return (-32);
6626	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6627	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6628		return (-33);
6629	}
6630	/* concatenate the full random key */
6631	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6632	if (chunks != NULL) {
6633		keylen += sizeof(*chunks) + num_chunks;
6634	}
6635	new_key = sctp_alloc_key(keylen);
6636	if (new_key != NULL) {
6637		/* copy in the RANDOM */
6638		if (p_random != NULL) {
6639			keylen = sizeof(*p_random) + random_len;
6640			bcopy(p_random, new_key->key, keylen);
6641		}
6642		/* append in the AUTH chunks */
6643		if (chunks != NULL) {
6644			bcopy(chunks, new_key->key + keylen,
6645			    sizeof(*chunks) + num_chunks);
6646			keylen += sizeof(*chunks) + num_chunks;
6647		}
6648		/* append in the HMACs */
6649		if (hmacs != NULL) {
6650			bcopy(hmacs, new_key->key + keylen,
6651			    sizeof(*hmacs) + hmacs_len);
6652		}
6653	} else {
6654		/* failed to get memory for the key */
6655		return (-34);
6656	}
6657	if (stcb->asoc.authinfo.peer_random != NULL)
6658		sctp_free_key(stcb->asoc.authinfo.peer_random);
6659	stcb->asoc.authinfo.peer_random = new_key;
6660	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6661	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6662
6663	return (0);
6664}
6665
6666int
6667sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6668    struct sctp_nets *net)
6669{
6670	/* make sure the requested primary address exists in the assoc */
6671	if (net == NULL && sa)
6672		net = sctp_findnet(stcb, sa);
6673
6674	if (net == NULL) {
6675		/* didn't find the requested primary address! */
6676		return (-1);
6677	} else {
6678		/* set the primary address */
6679		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6680			/* Must be confirmed, so queue to set */
6681			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6682			return (0);
6683		}
6684		stcb->asoc.primary_destination = net;
6685		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6686			sctp_free_remote_addr(stcb->asoc.alternate);
6687			stcb->asoc.alternate = NULL;
6688		}
6689		net = TAILQ_FIRST(&stcb->asoc.nets);
6690		if (net != stcb->asoc.primary_destination) {
6691			/*
6692			 * first one on the list is NOT the primary
6693			 * sctp_cmpaddr() is much more efficient if the
6694			 * primary is the first on the list, make it so.
6695			 */
6696			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6697			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6698		}
6699		return (0);
6700	}
6701}
6702
6703int
6704sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6705{
6706	/*
6707	 * This function serves two purposes. It will see if a TAG can be
6708	 * re-used and return 1 for yes it is ok and 0 for don't use that
6709	 * tag. A secondary function it will do is purge out old tags that
6710	 * can be removed.
6711	 */
6712	struct sctpvtaghead *chain;
6713	struct sctp_tagblock *twait_block;
6714	struct sctpasochead *head;
6715	struct sctp_tcb *stcb;
6716	int i;
6717
6718	SCTP_INP_INFO_RLOCK();
6719	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6720	    SCTP_BASE_INFO(hashasocmark))];
6721	if (head == NULL) {
6722		/* invalid vtag */
6723		goto skip_vtag_check;
6724	}
6725	LIST_FOREACH(stcb, head, sctp_asocs) {
6726		/*
6727		 * We choose not to lock anything here. TCB's can't be
6728		 * removed since we have the read lock, so they can't be
6729		 * freed on us, same thing for the INP. I may be wrong with
6730		 * this assumption, but we will go with it for now :-)
6731		 */
6732		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6733			continue;
6734		}
6735		if (stcb->asoc.my_vtag == tag) {
6736			/* candidate */
6737			if (stcb->rport != rport) {
6738				continue;
6739			}
6740			if (stcb->sctp_ep->sctp_lport != lport) {
6741				continue;
6742			}
6743			/* Its a used tag set */
6744			SCTP_INP_INFO_RUNLOCK();
6745			return (0);
6746		}
6747	}
6748skip_vtag_check:
6749
6750	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6751	/* Now what about timed wait ? */
6752	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6753		/*
6754		 * Block(s) are present, lets see if we have this tag in the
6755		 * list
6756		 */
6757		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6758			if (twait_block->vtag_block[i].v_tag == 0) {
6759				/* not used */
6760				continue;
6761			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6762			    now->tv_sec) {
6763				/* Audit expires this guy */
6764				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6765				twait_block->vtag_block[i].v_tag = 0;
6766				twait_block->vtag_block[i].lport = 0;
6767				twait_block->vtag_block[i].rport = 0;
6768			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6769				    (twait_block->vtag_block[i].lport == lport) &&
6770			    (twait_block->vtag_block[i].rport == rport)) {
6771				/* Bad tag, sorry :< */
6772				SCTP_INP_INFO_RUNLOCK();
6773				return (0);
6774			}
6775		}
6776	}
6777	SCTP_INP_INFO_RUNLOCK();
6778	return (1);
6779}
6780
6781static void
6782sctp_drain_mbufs(struct sctp_tcb *stcb)
6783{
6784	/*
6785	 * We must hunt this association for MBUF's past the cumack (i.e.
6786	 * out of order data that we can renege on).
6787	 */
6788	struct sctp_association *asoc;
6789	struct sctp_tmit_chunk *chk, *nchk;
6790	uint32_t cumulative_tsn_p1;
6791	struct sctp_queued_to_read *ctl, *nctl;
6792	int cnt, strmat;
6793	uint32_t gap, i;
6794	int fnd = 0;
6795
6796	/* We look for anything larger than the cum-ack + 1 */
6797
6798	asoc = &stcb->asoc;
6799	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6800		/* none we can reneg on. */
6801		return;
6802	}
6803	SCTP_STAT_INCR(sctps_protocol_drains_done);
6804	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6805	cnt = 0;
6806	/* First look in the re-assembly queue */
6807	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6808		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6809			/* Yep it is above cum-ack */
6810			cnt++;
6811			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6812			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6813			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6814			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6815			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6816			if (chk->data) {
6817				sctp_m_freem(chk->data);
6818				chk->data = NULL;
6819			}
6820			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6821		}
6822	}
6823	/* Ok that was fun, now we will drain all the inbound streams? */
6824	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6825		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6826			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6827				/* Yep it is above cum-ack */
6828				cnt++;
6829				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6830				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6831				sctp_ucount_decr(asoc->cnt_on_all_streams);
6832				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6833				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6834				if (ctl->data) {
6835					sctp_m_freem(ctl->data);
6836					ctl->data = NULL;
6837				}
6838				sctp_free_remote_addr(ctl->whoFrom);
6839				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6840				SCTP_DECR_READQ_COUNT();
6841			}
6842		}
6843	}
6844	if (cnt) {
6845		/* We must back down to see what the new highest is */
6846		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6847			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6848			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6849				asoc->highest_tsn_inside_map = i;
6850				fnd = 1;
6851				break;
6852			}
6853		}
6854		if (!fnd) {
6855			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6856		}
6857		/*
6858		 * Question, should we go through the delivery queue? The
6859		 * only reason things are on here is the app not reading OR
6860		 * a p-d-api up. An attacker COULD send enough in to
6861		 * initiate the PD-API and then send a bunch of stuff to
6862		 * other streams... these would wind up on the delivery
6863		 * queue.. and then we would not get to them. But in order
6864		 * to do this I then have to back-track and un-deliver
6865		 * sequence numbers in streams.. el-yucko. I think for now
6866		 * we will NOT look at the delivery queue and leave it to be
6867		 * something to consider later. An alternative would be to
6868		 * abort the P-D-API with a notification and then deliver
6869		 * the data.... Or another method might be to keep track of
6870		 * how many times the situation occurs and if we see a
6871		 * possible attack underway just abort the association.
6872		 */
6873#ifdef SCTP_DEBUG
6874		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6875#endif
6876		/*
6877		 * Now do we need to find a new
6878		 * asoc->highest_tsn_inside_map?
6879		 */
6880		asoc->last_revoke_count = cnt;
6881		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6882		/* sa_ignore NO_NULL_CHK */
6883		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6884		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6885	}
6886	/*
6887	 * Another issue, in un-setting the TSN's in the mapping array we
6888	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6889	 * two things to occur. It may cause us to do extra work in checking
6890	 * for our mapping array movement. More importantly it may cause us
6891	 * to SACK every datagram. This may not be a bad thing though since
6892	 * we will recover once we get our cum-ack above and all this stuff
6893	 * we dumped recovered.
6894	 */
6895}
6896
6897void
6898sctp_drain()
6899{
6900	/*
6901	 * We must walk the PCB lists for ALL associations here. The system
6902	 * is LOW on MBUF's and needs help. This is where reneging will
6903	 * occur. We really hope this does NOT happen!
6904	 */
6905	VNET_ITERATOR_DECL(vnet_iter);
6906	VNET_LIST_RLOCK_NOSLEEP();
6907	VNET_FOREACH(vnet_iter) {
6908		CURVNET_SET(vnet_iter);
6909		struct sctp_inpcb *inp;
6910		struct sctp_tcb *stcb;
6911
6912		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6913		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6914#ifdef VIMAGE
6915			continue;
6916#else
6917			return;
6918#endif
6919		}
6920		SCTP_INP_INFO_RLOCK();
6921		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6922			/* For each endpoint */
6923			SCTP_INP_RLOCK(inp);
6924			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6925				/* For each association */
6926				SCTP_TCB_LOCK(stcb);
6927				sctp_drain_mbufs(stcb);
6928				SCTP_TCB_UNLOCK(stcb);
6929			}
6930			SCTP_INP_RUNLOCK(inp);
6931		}
6932		SCTP_INP_INFO_RUNLOCK();
6933		CURVNET_RESTORE();
6934	}
6935	VNET_LIST_RUNLOCK_NOSLEEP();
6936}
6937
6938/*
6939 * start a new iterator
6940 * iterates through all endpoints and associations based on the pcb_state
6941 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6942 * assocs and "ef" (optional) is executed when the iterator completes.
6943 * "inpf" (optional) is executed for each new endpoint as it is being
6944 * iterated through. inpe (optional) is called when the inp completes
6945 * its way through all the stcbs.
6946 */
6947int
6948sctp_initiate_iterator(inp_func inpf,
6949    asoc_func af,
6950    inp_func inpe,
6951    uint32_t pcb_state,
6952    uint32_t pcb_features,
6953    uint32_t asoc_state,
6954    void *argp,
6955    uint32_t argi,
6956    end_func ef,
6957    struct sctp_inpcb *s_inp,
6958    uint8_t chunk_output_off)
6959{
6960	struct sctp_iterator *it = NULL;
6961
6962	if (af == NULL) {
6963		return (-1);
6964	}
6965	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6966	    SCTP_M_ITER);
6967	if (it == NULL) {
6968		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6969		return (ENOMEM);
6970	}
6971	memset(it, 0, sizeof(*it));
6972	it->function_assoc = af;
6973	it->function_inp = inpf;
6974	if (inpf)
6975		it->done_current_ep = 0;
6976	else
6977		it->done_current_ep = 1;
6978	it->function_atend = ef;
6979	it->pointer = argp;
6980	it->val = argi;
6981	it->pcb_flags = pcb_state;
6982	it->pcb_features = pcb_features;
6983	it->asoc_state = asoc_state;
6984	it->function_inp_end = inpe;
6985	it->no_chunk_output = chunk_output_off;
6986	it->vn = curvnet;
6987	if (s_inp) {
6988		/* Assume lock is held here */
6989		it->inp = s_inp;
6990		SCTP_INP_INCR_REF(it->inp);
6991		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
6992	} else {
6993		SCTP_INP_INFO_RLOCK();
6994		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
6995		if (it->inp) {
6996			SCTP_INP_INCR_REF(it->inp);
6997		}
6998		SCTP_INP_INFO_RUNLOCK();
6999		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7000
7001	}
7002	SCTP_IPI_ITERATOR_WQ_LOCK();
7003
7004	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7005	if (sctp_it_ctl.iterator_running == 0) {
7006		sctp_wakeup_iterator();
7007	}
7008	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7009	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7010	return (0);
7011}
7012