sctp_pcb.c revision 235828
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: head/sys/netinet/sctp_pcb.c 235828 2012-05-23 11:26:28Z 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#include <netinet/udp.h>
50#ifdef INET6
51#include <netinet6/ip6_var.h>
52#endif
53#include <sys/sched.h>
54#include <sys/smp.h>
55#include <sys/unistd.h>
56
57
58VNET_DEFINE(struct sctp_base_info, system_base_info);
59
60/* FIX: we don't handle multiple link local scopes */
61/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
62#ifdef INET6
63int
64SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
65{
66	struct sockaddr_in6 tmp_a, tmp_b;
67
68	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
69	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
70		return (0);
71	}
72	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
73	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
74		return (0);
75	}
76	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
77}
78
79#endif
80
81void
82sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
83{
84	/*
85	 * We really don't need to lock this, but I will just because it
86	 * does not hurt.
87	 */
88	SCTP_INP_INFO_RLOCK();
89	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
90	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
91	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
92	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
93	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
94	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
95	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
96	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
97
98	SCTP_INP_INFO_RUNLOCK();
99}
100
101/*
102 * Addresses are added to VRF's (Virtual Router's). For BSD we
103 * have only the default VRF 0. We maintain a hash list of
104 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
105 * these has a list of addresses. When we add a new address
106 * to a VRF we lookup the ifn/ifn_index, if the ifn does
107 * not exist we create it and add it to the list of IFN's
108 * within the VRF. Once we have the sctp_ifn, we add the
109 * address to the list. So we look something like:
110 *
111 * hash-vrf-table
112 *   vrf-> ifn-> ifn -> ifn
113 *   vrf    |
114 *    ...   +--ifa-> ifa -> ifa
115 *   vrf
116 *
117 * We keep these separate lists since the SCTP subsystem will
118 * point to these from its source address selection nets structure.
119 * When an address is deleted it does not happen right away on
120 * the SCTP side, it gets scheduled. What we do when a
121 * delete happens is immediately remove the address from
122 * the master list and decrement the refcount. As our
123 * addip iterator works through and frees the src address
124 * selection pointing to the sctp_ifa, eventually the refcount
125 * will reach 0 and we will delete it. Note that it is assumed
126 * that any locking on system level ifn/ifa is done at the
127 * caller of these functions and these routines will only
128 * lock the SCTP structures as they add or delete things.
129 *
130 * Other notes on VRF concepts.
131 *  - An endpoint can be in multiple VRF's
132 *  - An association lives within a VRF and only one VRF.
133 *  - Any incoming packet we can deduce the VRF for by
134 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
135 *  - Any downward send call or connect call must supply the
136 *    VRF via ancillary data or via some sort of set default
137 *    VRF socket option call (again for BSD no brainer since
138 *    the VRF is always 0).
139 *  - An endpoint may add multiple VRF's to it.
140 *  - Listening sockets can accept associations in any
141 *    of the VRF's they are in but the assoc will end up
142 *    in only one VRF (gotten from the packet or connect/send).
143 *
144 */
145
146struct sctp_vrf *
147sctp_allocate_vrf(int vrf_id)
148{
149	struct sctp_vrf *vrf = NULL;
150	struct sctp_vrflist *bucket;
151
152	/* First allocate the VRF structure */
153	vrf = sctp_find_vrf(vrf_id);
154	if (vrf) {
155		/* Already allocated */
156		return (vrf);
157	}
158	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
159	    SCTP_M_VRF);
160	if (vrf == NULL) {
161		/* No memory */
162#ifdef INVARIANTS
163		panic("No memory for VRF:%d", vrf_id);
164#endif
165		return (NULL);
166	}
167	/* setup the VRF */
168	memset(vrf, 0, sizeof(struct sctp_vrf));
169	vrf->vrf_id = vrf_id;
170	LIST_INIT(&vrf->ifnlist);
171	vrf->total_ifa_count = 0;
172	vrf->refcount = 0;
173	/* now also setup table ids */
174	SCTP_INIT_VRF_TABLEID(vrf);
175	/* Init the HASH of addresses */
176	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
177	    &vrf->vrf_addr_hashmark);
178	if (vrf->vrf_addr_hash == NULL) {
179		/* No memory */
180#ifdef INVARIANTS
181		panic("No memory for VRF:%d", vrf_id);
182#endif
183		SCTP_FREE(vrf, SCTP_M_VRF);
184		return (NULL);
185	}
186	/* Add it to the hash table */
187	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
188	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
189	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
190	return (vrf);
191}
192
193
194struct sctp_ifn *
195sctp_find_ifn(void *ifn, uint32_t ifn_index)
196{
197	struct sctp_ifn *sctp_ifnp;
198	struct sctp_ifnlist *hash_ifn_head;
199
200	/*
201	 * We assume the lock is held for the addresses if that's wrong
202	 * problems could occur :-)
203	 */
204	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
205	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
206		if (sctp_ifnp->ifn_index == ifn_index) {
207			return (sctp_ifnp);
208		}
209		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
210			return (sctp_ifnp);
211		}
212	}
213	return (NULL);
214}
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
233void
234sctp_free_vrf(struct sctp_vrf *vrf)
235{
236	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
237		if (vrf->vrf_addr_hash) {
238			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
239			vrf->vrf_addr_hash = NULL;
240		}
241		/* We zero'd the count */
242		LIST_REMOVE(vrf, next_vrf);
243		SCTP_FREE(vrf, SCTP_M_VRF);
244		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
245	}
246}
247
248void
249sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
250{
251	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
252		/* We zero'd the count */
253		if (sctp_ifnp->vrf) {
254			sctp_free_vrf(sctp_ifnp->vrf);
255		}
256		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
257		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
258	}
259}
260
261void
262sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
263{
264	struct sctp_ifn *sctp_ifnp;
265
266	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
267	if (sctp_ifnp != NULL) {
268		sctp_ifnp->ifn_mtu = mtu;
269	}
270}
271
272
273void
274sctp_free_ifa(struct sctp_ifa *sctp_ifap)
275{
276	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
277		/* We zero'd the count */
278		if (sctp_ifap->ifn_p) {
279			sctp_free_ifn(sctp_ifap->ifn_p);
280		}
281		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
282		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
283	}
284}
285
286static void
287sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
288{
289	struct sctp_ifn *found;
290
291	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
292	if (found == NULL) {
293		/* Not in the list.. sorry */
294		return;
295	}
296	if (hold_addr_lock == 0)
297		SCTP_IPI_ADDR_WLOCK();
298	LIST_REMOVE(sctp_ifnp, next_bucket);
299	LIST_REMOVE(sctp_ifnp, next_ifn);
300	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
301	    sctp_ifnp->registered_af);
302	if (hold_addr_lock == 0)
303		SCTP_IPI_ADDR_WUNLOCK();
304	/* Take away the reference, and possibly free it */
305	sctp_free_ifn(sctp_ifnp);
306}
307
308void
309sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
310    const char *if_name, uint32_t ifn_index)
311{
312	struct sctp_vrf *vrf;
313	struct sctp_ifa *sctp_ifap = NULL;
314
315	SCTP_IPI_ADDR_RLOCK();
316	vrf = sctp_find_vrf(vrf_id);
317	if (vrf == NULL) {
318		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
319		goto out;
320
321	}
322	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
323	if (sctp_ifap == NULL) {
324		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
325		goto out;
326	}
327	if (sctp_ifap->ifn_p == NULL) {
328		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
329		goto out;
330	}
331	if (if_name) {
332		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
333			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
334			    sctp_ifap->ifn_p->ifn_name, if_name);
335			goto out;
336		}
337	} else {
338		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
339			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
340			    sctp_ifap->ifn_p->ifn_index, ifn_index);
341			goto out;
342		}
343	}
344
345	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
346	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
347out:
348	SCTP_IPI_ADDR_RUNLOCK();
349}
350
351void
352sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
353    const char *if_name, uint32_t ifn_index)
354{
355	struct sctp_vrf *vrf;
356	struct sctp_ifa *sctp_ifap = NULL;
357
358	SCTP_IPI_ADDR_RLOCK();
359	vrf = sctp_find_vrf(vrf_id);
360	if (vrf == NULL) {
361		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
362		goto out;
363
364	}
365	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
366	if (sctp_ifap == NULL) {
367		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
368		goto out;
369	}
370	if (sctp_ifap->ifn_p == NULL) {
371		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
372		goto out;
373	}
374	if (if_name) {
375		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
376			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
377			    sctp_ifap->ifn_p->ifn_name, if_name);
378			goto out;
379		}
380	} else {
381		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
382			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
383			    sctp_ifap->ifn_p->ifn_index, ifn_index);
384			goto out;
385		}
386	}
387
388	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
389	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
390out:
391	SCTP_IPI_ADDR_RUNLOCK();
392}
393
394/*-
395 * Add an ifa to an ifn.
396 * Register the interface as necessary.
397 * NOTE: ADDR write lock MUST be held.
398 */
399static void
400sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
401{
402	int ifa_af;
403
404	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
405	sctp_ifap->ifn_p = sctp_ifnp;
406	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
407	/* update address counts */
408	sctp_ifnp->ifa_count++;
409	ifa_af = sctp_ifap->address.sa.sa_family;
410	switch (ifa_af) {
411#ifdef INET
412	case AF_INET:
413		sctp_ifnp->num_v4++;
414		break;
415#endif
416#ifdef INET6
417	case AF_INET6:
418		sctp_ifnp->num_v6++;
419		break;
420#endif
421	default:
422		break;
423	}
424	if (sctp_ifnp->ifa_count == 1) {
425		/* register the new interface */
426		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
427		sctp_ifnp->registered_af = ifa_af;
428	}
429}
430
431/*-
432 * Remove an ifa from its ifn.
433 * If no more addresses exist, remove the ifn too. Otherwise, re-register
434 * the interface based on the remaining address families left.
435 * NOTE: ADDR write lock MUST be held.
436 */
437static void
438sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
439{
440	LIST_REMOVE(sctp_ifap, next_ifa);
441	if (sctp_ifap->ifn_p) {
442		/* update address counts */
443		sctp_ifap->ifn_p->ifa_count--;
444		switch (sctp_ifap->address.sa.sa_family) {
445#ifdef INET
446		case AF_INET:
447			sctp_ifap->ifn_p->num_v4--;
448			break;
449#endif
450#ifdef INET6
451		case AF_INET6:
452			sctp_ifap->ifn_p->num_v6--;
453			break;
454#endif
455		default:
456			break;
457		}
458
459		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
460			/* remove the ifn, possibly freeing it */
461			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
462		} else {
463			/* re-register address family type, if needed */
464			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
465			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
466				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
467				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
468				sctp_ifap->ifn_p->registered_af = AF_INET;
469			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
470			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
471				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
472				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
473				sctp_ifap->ifn_p->registered_af = AF_INET6;
474			}
475			/* free the ifn refcount */
476			sctp_free_ifn(sctp_ifap->ifn_p);
477		}
478		sctp_ifap->ifn_p = NULL;
479	}
480}
481
482struct sctp_ifa *
483sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
484    uint32_t ifn_type, const char *if_name, void *ifa,
485    struct sockaddr *addr, uint32_t ifa_flags,
486    int dynamic_add)
487{
488	struct sctp_vrf *vrf;
489	struct sctp_ifn *sctp_ifnp = NULL;
490	struct sctp_ifa *sctp_ifap = NULL;
491	struct sctp_ifalist *hash_addr_head;
492	struct sctp_ifnlist *hash_ifn_head;
493	uint32_t hash_of_addr;
494	int new_ifn_af = 0;
495
496#ifdef SCTP_DEBUG
497	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
498	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
499#endif
500	SCTP_IPI_ADDR_WLOCK();
501	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
502	if (sctp_ifnp) {
503		vrf = sctp_ifnp->vrf;
504	} else {
505		vrf = sctp_find_vrf(vrf_id);
506		if (vrf == NULL) {
507			vrf = sctp_allocate_vrf(vrf_id);
508			if (vrf == NULL) {
509				SCTP_IPI_ADDR_WUNLOCK();
510				return (NULL);
511			}
512		}
513	}
514	if (sctp_ifnp == NULL) {
515		/*
516		 * build one and add it, can't hold lock until after malloc
517		 * done though.
518		 */
519		SCTP_IPI_ADDR_WUNLOCK();
520		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
521		    sizeof(struct sctp_ifn), SCTP_M_IFN);
522		if (sctp_ifnp == NULL) {
523#ifdef INVARIANTS
524			panic("No memory for IFN");
525#endif
526			return (NULL);
527		}
528		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
529		sctp_ifnp->ifn_index = ifn_index;
530		sctp_ifnp->ifn_p = ifn;
531		sctp_ifnp->ifn_type = ifn_type;
532		sctp_ifnp->refcount = 0;
533		sctp_ifnp->vrf = vrf;
534		atomic_add_int(&vrf->refcount, 1);
535		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
536		if (if_name != NULL) {
537			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
538		} else {
539			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
540		}
541		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
542		LIST_INIT(&sctp_ifnp->ifalist);
543		SCTP_IPI_ADDR_WLOCK();
544		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
545		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
546		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
547		new_ifn_af = 1;
548	}
549	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
550	if (sctp_ifap) {
551		/* Hmm, it already exists? */
552		if ((sctp_ifap->ifn_p) &&
553		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
554			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
555			    sctp_ifap->ifn_p->ifn_name, ifn_index,
556			    sctp_ifap);
557			if (new_ifn_af) {
558				/* Remove the created one that we don't want */
559				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
560			}
561			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
562				/* easy to solve, just switch back to active */
563				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
564				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
565				sctp_ifap->ifn_p = sctp_ifnp;
566				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
567			}
568	exit_stage_left:
569			SCTP_IPI_ADDR_WUNLOCK();
570			return (sctp_ifap);
571		} else {
572			if (sctp_ifap->ifn_p) {
573				/*
574				 * The last IFN gets the address, remove the
575				 * old one
576				 */
577				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
578				    sctp_ifap, sctp_ifap->ifn_p->ifn_name,
579				    sctp_ifap->ifn_p->ifn_index, if_name,
580				    ifn_index);
581				/* remove the address from the old ifn */
582				sctp_remove_ifa_from_ifn(sctp_ifap);
583				/* move the address over to the new ifn */
584				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
585				goto exit_stage_left;
586			} else {
587				/* repair ifnp which was NULL ? */
588				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
589				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
590				    sctp_ifnp, sctp_ifap);
591				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
592			}
593			goto exit_stage_left;
594		}
595	}
596	SCTP_IPI_ADDR_WUNLOCK();
597	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
598	if (sctp_ifap == NULL) {
599#ifdef INVARIANTS
600		panic("No memory for IFA");
601#endif
602		return (NULL);
603	}
604	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
605	sctp_ifap->ifn_p = sctp_ifnp;
606	atomic_add_int(&sctp_ifnp->refcount, 1);
607	sctp_ifap->vrf_id = vrf_id;
608	sctp_ifap->ifa = ifa;
609	memcpy(&sctp_ifap->address, addr, addr->sa_len);
610	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
611	sctp_ifap->flags = ifa_flags;
612	/* Set scope */
613	switch (sctp_ifap->address.sa.sa_family) {
614#ifdef INET
615	case AF_INET:
616		{
617			struct sockaddr_in *sin;
618
619			sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
620			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
621			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
622				sctp_ifap->src_is_loop = 1;
623			}
624			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
625				sctp_ifap->src_is_priv = 1;
626			}
627			sctp_ifnp->num_v4++;
628			if (new_ifn_af)
629				new_ifn_af = AF_INET;
630			break;
631		}
632#endif
633#ifdef INET6
634	case AF_INET6:
635		{
636			/* ok to use deprecated addresses? */
637			struct sockaddr_in6 *sin6;
638
639			sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
640			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
641			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
642				sctp_ifap->src_is_loop = 1;
643			}
644			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
645				sctp_ifap->src_is_priv = 1;
646			}
647			sctp_ifnp->num_v6++;
648			if (new_ifn_af)
649				new_ifn_af = AF_INET6;
650			break;
651		}
652#endif
653	default:
654		new_ifn_af = 0;
655		break;
656	}
657	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
658
659	if ((sctp_ifap->src_is_priv == 0) &&
660	    (sctp_ifap->src_is_loop == 0)) {
661		sctp_ifap->src_is_glob = 1;
662	}
663	SCTP_IPI_ADDR_WLOCK();
664	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
665	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
666	sctp_ifap->refcount = 1;
667	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
668	sctp_ifnp->ifa_count++;
669	vrf->total_ifa_count++;
670	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
671	if (new_ifn_af) {
672		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
673		sctp_ifnp->registered_af = new_ifn_af;
674	}
675	SCTP_IPI_ADDR_WUNLOCK();
676	if (dynamic_add) {
677		/*
678		 * Bump up the refcount so that when the timer completes it
679		 * will drop back down.
680		 */
681		struct sctp_laddr *wi;
682
683		atomic_add_int(&sctp_ifap->refcount, 1);
684		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
685		if (wi == NULL) {
686			/*
687			 * Gak, what can we do? We have lost an address
688			 * change can you say HOSED?
689			 */
690			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
691			/* Opps, must decrement the count */
692			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
693			    if_name);
694			return (NULL);
695		}
696		SCTP_INCR_LADDR_COUNT();
697		bzero(wi, sizeof(*wi));
698		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
699		wi->ifa = sctp_ifap;
700		wi->action = SCTP_ADD_IP_ADDRESS;
701
702		SCTP_WQ_ADDR_LOCK();
703		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
704		SCTP_WQ_ADDR_UNLOCK();
705
706		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
707		    (struct sctp_inpcb *)NULL,
708		    (struct sctp_tcb *)NULL,
709		    (struct sctp_nets *)NULL);
710	} else {
711		/* it's ready for use */
712		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
713	}
714	return (sctp_ifap);
715}
716
717void
718sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
719    uint32_t ifn_index, const char *if_name)
720{
721	struct sctp_vrf *vrf;
722	struct sctp_ifa *sctp_ifap = NULL;
723
724	SCTP_IPI_ADDR_WLOCK();
725	vrf = sctp_find_vrf(vrf_id);
726	if (vrf == NULL) {
727		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
728		goto out_now;
729	}
730#ifdef SCTP_DEBUG
731	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
732	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
733#endif
734	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
735	if (sctp_ifap) {
736		/* Validate the delete */
737		if (sctp_ifap->ifn_p) {
738			int valid = 0;
739
740			/*-
741			 * The name has priority over the ifn_index
742			 * if its given. We do this especially for
743			 * panda who might recycle indexes fast.
744			 */
745			if (if_name) {
746				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
747					/* They match its a correct delete */
748					valid = 1;
749				}
750			}
751			if (!valid) {
752				/* last ditch check ifn_index */
753				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
754					valid = 1;
755				}
756			}
757			if (!valid) {
758				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
759				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
760				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
761				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
762				SCTP_IPI_ADDR_WUNLOCK();
763				return;
764			}
765		}
766		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", sctp_ifap);
767		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
768		sctp_ifap->localifa_flags |= SCTP_BEING_DELETED;
769		vrf->total_ifa_count--;
770		LIST_REMOVE(sctp_ifap, next_bucket);
771		sctp_remove_ifa_from_ifn(sctp_ifap);
772	}
773#ifdef SCTP_DEBUG
774	else {
775		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
776		    ifn_index);
777		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
778	}
779#endif
780
781out_now:
782	SCTP_IPI_ADDR_WUNLOCK();
783	if (sctp_ifap) {
784		struct sctp_laddr *wi;
785
786		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
787		if (wi == NULL) {
788			/*
789			 * Gak, what can we do? We have lost an address
790			 * change can you say HOSED?
791			 */
792			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
793
794			/* Oops, must decrement the count */
795			sctp_free_ifa(sctp_ifap);
796			return;
797		}
798		SCTP_INCR_LADDR_COUNT();
799		bzero(wi, sizeof(*wi));
800		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
801		wi->ifa = sctp_ifap;
802		wi->action = SCTP_DEL_IP_ADDRESS;
803		SCTP_WQ_ADDR_LOCK();
804		/*
805		 * Should this really be a tailq? As it is we will process
806		 * the newest first :-0
807		 */
808		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
809		SCTP_WQ_ADDR_UNLOCK();
810
811		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
812		    (struct sctp_inpcb *)NULL,
813		    (struct sctp_tcb *)NULL,
814		    (struct sctp_nets *)NULL);
815	}
816	return;
817}
818
819
820static struct sctp_tcb *
821sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
822    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
823{
824	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
825	/*
826	 * If we support the TCP model, then we must now dig through to see
827	 * if we can find our endpoint in the list of tcp ep's.
828	 */
829	uint16_t lport, rport;
830	struct sctppcbhead *ephead;
831	struct sctp_inpcb *inp;
832	struct sctp_laddr *laddr;
833	struct sctp_tcb *stcb;
834	struct sctp_nets *net;
835
836	if ((to == NULL) || (from == NULL)) {
837		return (NULL);
838	}
839	switch (to->sa_family) {
840#ifdef INET
841	case AF_INET:
842		if (from->sa_family == AF_INET) {
843			lport = ((struct sockaddr_in *)to)->sin_port;
844			rport = ((struct sockaddr_in *)from)->sin_port;
845		} else {
846			return (NULL);
847		}
848		break;
849#endif
850#ifdef INET6
851	case AF_INET6:
852		if (from->sa_family == AF_INET6) {
853			lport = ((struct sockaddr_in6 *)to)->sin6_port;
854			rport = ((struct sockaddr_in6 *)from)->sin6_port;
855		} else {
856			return (NULL);
857		}
858		break;
859#endif
860	default:
861		return (NULL);
862	}
863	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
864	/*
865	 * Ok now for each of the guys in this bucket we must look and see:
866	 * - Does the remote port match. - Does there single association's
867	 * addresses match this address (to). If so we update p_ep to point
868	 * to this ep and return the tcb from it.
869	 */
870	LIST_FOREACH(inp, ephead, sctp_hash) {
871		SCTP_INP_RLOCK(inp);
872		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
873			SCTP_INP_RUNLOCK(inp);
874			continue;
875		}
876		if (lport != inp->sctp_lport) {
877			SCTP_INP_RUNLOCK(inp);
878			continue;
879		}
880		if (inp->def_vrf_id != vrf_id) {
881			SCTP_INP_RUNLOCK(inp);
882			continue;
883		}
884		/* check to see if the ep has one of the addresses */
885		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
886			/* We are NOT bound all, so look further */
887			int match = 0;
888
889			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
890
891				if (laddr->ifa == NULL) {
892					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
893					continue;
894				}
895				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
896					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
897					continue;
898				}
899				if (laddr->ifa->address.sa.sa_family ==
900				    to->sa_family) {
901					/* see if it matches */
902
903#ifdef INET
904					if (from->sa_family == AF_INET) {
905						struct sockaddr_in *intf_addr,
906						           *sin;
907
908						intf_addr = &laddr->ifa->address.sin;
909						sin = (struct sockaddr_in *)to;
910						if (sin->sin_addr.s_addr ==
911						    intf_addr->sin_addr.s_addr) {
912							match = 1;
913							break;
914						}
915					}
916#endif
917#ifdef INET6
918					if (from->sa_family == AF_INET6) {
919						struct sockaddr_in6 *intf_addr6;
920						struct sockaddr_in6 *sin6;
921
922						sin6 = (struct sockaddr_in6 *)
923						    to;
924						intf_addr6 = &laddr->ifa->address.sin6;
925
926						if (SCTP6_ARE_ADDR_EQUAL(sin6,
927						    intf_addr6)) {
928							match = 1;
929							break;
930						}
931					}
932#endif
933				}
934			}
935			if (match == 0) {
936				/* This endpoint does not have this address */
937				SCTP_INP_RUNLOCK(inp);
938				continue;
939			}
940		}
941		/*
942		 * Ok if we hit here the ep has the address, does it hold
943		 * the tcb?
944		 */
945
946		stcb = LIST_FIRST(&inp->sctp_asoc_list);
947		if (stcb == NULL) {
948			SCTP_INP_RUNLOCK(inp);
949			continue;
950		}
951		SCTP_TCB_LOCK(stcb);
952		if (stcb->rport != rport) {
953			/* remote port does not match. */
954			SCTP_TCB_UNLOCK(stcb);
955			SCTP_INP_RUNLOCK(inp);
956			continue;
957		}
958		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
959			SCTP_TCB_UNLOCK(stcb);
960			SCTP_INP_RUNLOCK(inp);
961			continue;
962		}
963		/* Does this TCB have a matching address? */
964		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
965
966			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
967				/* not the same family, can't be a match */
968				continue;
969			}
970			switch (from->sa_family) {
971#ifdef INET
972			case AF_INET:
973				{
974					struct sockaddr_in *sin, *rsin;
975
976					sin = (struct sockaddr_in *)&net->ro._l_addr;
977					rsin = (struct sockaddr_in *)from;
978					if (sin->sin_addr.s_addr ==
979					    rsin->sin_addr.s_addr) {
980						/* found it */
981						if (netp != NULL) {
982							*netp = net;
983						}
984						/*
985						 * Update the endpoint
986						 * pointer
987						 */
988						*inp_p = inp;
989						SCTP_INP_RUNLOCK(inp);
990						return (stcb);
991					}
992					break;
993				}
994#endif
995#ifdef INET6
996			case AF_INET6:
997				{
998					struct sockaddr_in6 *sin6, *rsin6;
999
1000					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1001					rsin6 = (struct sockaddr_in6 *)from;
1002					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1003					    rsin6)) {
1004						/* found it */
1005						if (netp != NULL) {
1006							*netp = net;
1007						}
1008						/*
1009						 * Update the endpoint
1010						 * pointer
1011						 */
1012						*inp_p = inp;
1013						SCTP_INP_RUNLOCK(inp);
1014						return (stcb);
1015					}
1016					break;
1017				}
1018#endif
1019			default:
1020				/* TSNH */
1021				break;
1022			}
1023		}
1024		SCTP_TCB_UNLOCK(stcb);
1025		SCTP_INP_RUNLOCK(inp);
1026	}
1027	return (NULL);
1028}
1029
1030static int
1031sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
1032{
1033	int loopback_scope, ipv4_local_scope, local_scope, site_scope;
1034	int ipv4_addr_legal, ipv6_addr_legal;
1035	struct sctp_vrf *vrf;
1036	struct sctp_ifn *sctp_ifn;
1037	struct sctp_ifa *sctp_ifa;
1038
1039	loopback_scope = stcb->asoc.loopback_scope;
1040	ipv4_local_scope = stcb->asoc.ipv4_local_scope;
1041	local_scope = stcb->asoc.local_scope;
1042	site_scope = stcb->asoc.site_scope;
1043	ipv4_addr_legal = ipv6_addr_legal = 0;
1044	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1045		ipv6_addr_legal = 1;
1046		if (SCTP_IPV6_V6ONLY(stcb->sctp_ep) == 0) {
1047			ipv4_addr_legal = 1;
1048		}
1049	} else {
1050		ipv4_addr_legal = 1;
1051	}
1052
1053	SCTP_IPI_ADDR_RLOCK();
1054	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
1055	if (vrf == NULL) {
1056		/* no vrf, no addresses */
1057		SCTP_IPI_ADDR_RUNLOCK();
1058		return (0);
1059	}
1060	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
1061		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
1062			if ((loopback_scope == 0) &&
1063			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
1064				continue;
1065			}
1066			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
1067				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
1068				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
1069					/*
1070					 * We allow pending addresses, where
1071					 * we have sent an asconf-add to be
1072					 * considered valid.
1073					 */
1074					continue;
1075				}
1076				switch (sctp_ifa->address.sa.sa_family) {
1077#ifdef INET
1078				case AF_INET:
1079					if (ipv4_addr_legal) {
1080						struct sockaddr_in *sin,
1081						           *rsin;
1082
1083						sin = &sctp_ifa->address.sin;
1084						rsin = (struct sockaddr_in *)to;
1085						if ((ipv4_local_scope == 0) &&
1086						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
1087							continue;
1088						}
1089						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
1090							SCTP_IPI_ADDR_RUNLOCK();
1091							return (1);
1092						}
1093					}
1094					break;
1095#endif
1096#ifdef INET6
1097				case AF_INET6:
1098					if (ipv6_addr_legal) {
1099						struct sockaddr_in6 *sin6,
1100						            *rsin6;
1101
1102						sin6 = &sctp_ifa->address.sin6;
1103						rsin6 = (struct sockaddr_in6 *)to;
1104						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
1105							if (local_scope == 0)
1106								continue;
1107							if (sin6->sin6_scope_id == 0) {
1108								if (sa6_recoverscope(sin6) != 0)
1109									continue;
1110							}
1111						}
1112						if ((site_scope == 0) &&
1113						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
1114							continue;
1115						}
1116						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
1117							SCTP_IPI_ADDR_RUNLOCK();
1118							return (1);
1119						}
1120					}
1121					break;
1122#endif
1123				default:
1124					/* TSNH */
1125					break;
1126				}
1127			}
1128		}
1129	} else {
1130		struct sctp_laddr *laddr;
1131
1132		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
1133			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
1134			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
1135				/*
1136				 * We allow pending addresses, where we have
1137				 * sent an asconf-add to be considered
1138				 * valid.
1139				 */
1140				continue;
1141			}
1142			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
1143				continue;
1144			}
1145			switch (to->sa_family) {
1146#ifdef INET
1147			case AF_INET:
1148				{
1149					struct sockaddr_in *sin, *rsin;
1150
1151					sin = (struct sockaddr_in *)&laddr->ifa->address.sin;
1152					rsin = (struct sockaddr_in *)to;
1153					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
1154						SCTP_IPI_ADDR_RUNLOCK();
1155						return (1);
1156					}
1157					break;
1158				}
1159#endif
1160#ifdef INET6
1161			case AF_INET6:
1162				{
1163					struct sockaddr_in6 *sin6, *rsin6;
1164
1165					sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
1166					rsin6 = (struct sockaddr_in6 *)to;
1167					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
1168						SCTP_IPI_ADDR_RUNLOCK();
1169						return (1);
1170					}
1171					break;
1172				}
1173
1174#endif
1175			default:
1176				/* TSNH */
1177				break;
1178			}
1179
1180		}
1181	}
1182	SCTP_IPI_ADDR_RUNLOCK();
1183	return (0);
1184}
1185
1186/*
1187 * rules for use
1188 *
1189 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1190 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1191 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1192 * NULL.
1193 */
1194
1195struct sctp_tcb *
1196sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1197    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1198{
1199	struct sctpasochead *head;
1200	struct sctp_inpcb *inp;
1201	struct sctp_tcb *stcb = NULL;
1202	struct sctp_nets *net;
1203	uint16_t rport;
1204
1205	inp = *inp_p;
1206	if (remote->sa_family == AF_INET) {
1207		rport = (((struct sockaddr_in *)remote)->sin_port);
1208	} else if (remote->sa_family == AF_INET6) {
1209		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1210	} else {
1211		return (NULL);
1212	}
1213	if (locked_tcb) {
1214		/*
1215		 * UN-lock so we can do proper locking here this occurs when
1216		 * called from load_addresses_from_init.
1217		 */
1218		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1219		SCTP_TCB_UNLOCK(locked_tcb);
1220	}
1221	SCTP_INP_INFO_RLOCK();
1222	if (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
1223		/*-
1224		 * Now either this guy is our listener or it's the
1225		 * connector. If it is the one that issued the connect, then
1226		 * it's only chance is to be the first TCB in the list. If
1227		 * it is the acceptor, then do the special_lookup to hash
1228		 * and find the real inp.
1229		 */
1230		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1231			/* to is peer addr, from is my addr */
1232			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1233			    netp, inp->def_vrf_id);
1234			if ((stcb != NULL) && (locked_tcb == NULL)) {
1235				/* we have a locked tcb, lower refcount */
1236				SCTP_INP_DECR_REF(inp);
1237			}
1238			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1239				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1240				SCTP_TCB_LOCK(locked_tcb);
1241				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1242				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1243			}
1244			SCTP_INP_INFO_RUNLOCK();
1245			return (stcb);
1246		} else {
1247			SCTP_INP_WLOCK(inp);
1248			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1249				goto null_return;
1250			}
1251			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1252			if (stcb == NULL) {
1253				goto null_return;
1254			}
1255			SCTP_TCB_LOCK(stcb);
1256
1257			if (stcb->rport != rport) {
1258				/* remote port does not match. */
1259				SCTP_TCB_UNLOCK(stcb);
1260				goto null_return;
1261			}
1262			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1263				SCTP_TCB_UNLOCK(stcb);
1264				goto null_return;
1265			}
1266			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1267				SCTP_TCB_UNLOCK(stcb);
1268				goto null_return;
1269			}
1270			/* now look at the list of remote addresses */
1271			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1272#ifdef INVARIANTS
1273				if (net == (TAILQ_NEXT(net, sctp_next))) {
1274					panic("Corrupt net list");
1275				}
1276#endif
1277				if (net->ro._l_addr.sa.sa_family !=
1278				    remote->sa_family) {
1279					/* not the same family */
1280					continue;
1281				}
1282				switch (remote->sa_family) {
1283#ifdef INET
1284				case AF_INET:
1285					{
1286						struct sockaddr_in *sin,
1287						           *rsin;
1288
1289						sin = (struct sockaddr_in *)
1290						    &net->ro._l_addr;
1291						rsin = (struct sockaddr_in *)remote;
1292						if (sin->sin_addr.s_addr ==
1293						    rsin->sin_addr.s_addr) {
1294							/* found it */
1295							if (netp != NULL) {
1296								*netp = net;
1297							}
1298							if (locked_tcb == NULL) {
1299								SCTP_INP_DECR_REF(inp);
1300							} else if (locked_tcb != stcb) {
1301								SCTP_TCB_LOCK(locked_tcb);
1302							}
1303							if (locked_tcb) {
1304								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1305							}
1306							SCTP_INP_WUNLOCK(inp);
1307							SCTP_INP_INFO_RUNLOCK();
1308							return (stcb);
1309						}
1310						break;
1311					}
1312#endif
1313#ifdef INET6
1314				case AF_INET6:
1315					{
1316						struct sockaddr_in6 *sin6,
1317						            *rsin6;
1318
1319						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1320						rsin6 = (struct sockaddr_in6 *)remote;
1321						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1322						    rsin6)) {
1323							/* found it */
1324							if (netp != NULL) {
1325								*netp = net;
1326							}
1327							if (locked_tcb == NULL) {
1328								SCTP_INP_DECR_REF(inp);
1329							} else if (locked_tcb != stcb) {
1330								SCTP_TCB_LOCK(locked_tcb);
1331							}
1332							if (locked_tcb) {
1333								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1334							}
1335							SCTP_INP_WUNLOCK(inp);
1336							SCTP_INP_INFO_RUNLOCK();
1337							return (stcb);
1338						}
1339						break;
1340					}
1341#endif
1342				default:
1343					/* TSNH */
1344					break;
1345				}
1346			}
1347			SCTP_TCB_UNLOCK(stcb);
1348		}
1349	} else {
1350		SCTP_INP_WLOCK(inp);
1351		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1352			goto null_return;
1353		}
1354		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1355		    inp->sctp_hashmark)];
1356		if (head == NULL) {
1357			goto null_return;
1358		}
1359		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1360			if (stcb->rport != rport) {
1361				/* remote port does not match */
1362				continue;
1363			}
1364			SCTP_TCB_LOCK(stcb);
1365			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1366				SCTP_TCB_UNLOCK(stcb);
1367				continue;
1368			}
1369			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1370				SCTP_TCB_UNLOCK(stcb);
1371				continue;
1372			}
1373			/* now look at the list of remote addresses */
1374			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1375#ifdef INVARIANTS
1376				if (net == (TAILQ_NEXT(net, sctp_next))) {
1377					panic("Corrupt net list");
1378				}
1379#endif
1380				if (net->ro._l_addr.sa.sa_family !=
1381				    remote->sa_family) {
1382					/* not the same family */
1383					continue;
1384				}
1385				switch (remote->sa_family) {
1386#ifdef INET
1387				case AF_INET:
1388					{
1389						struct sockaddr_in *sin,
1390						           *rsin;
1391
1392						sin = (struct sockaddr_in *)
1393						    &net->ro._l_addr;
1394						rsin = (struct sockaddr_in *)remote;
1395						if (sin->sin_addr.s_addr ==
1396						    rsin->sin_addr.s_addr) {
1397							/* found it */
1398							if (netp != NULL) {
1399								*netp = net;
1400							}
1401							if (locked_tcb == NULL) {
1402								SCTP_INP_DECR_REF(inp);
1403							} else if (locked_tcb != stcb) {
1404								SCTP_TCB_LOCK(locked_tcb);
1405							}
1406							if (locked_tcb) {
1407								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1408							}
1409							SCTP_INP_WUNLOCK(inp);
1410							SCTP_INP_INFO_RUNLOCK();
1411							return (stcb);
1412						}
1413						break;
1414					}
1415#endif
1416#ifdef INET6
1417				case AF_INET6:
1418					{
1419						struct sockaddr_in6 *sin6,
1420						            *rsin6;
1421
1422						sin6 = (struct sockaddr_in6 *)
1423						    &net->ro._l_addr;
1424						rsin6 = (struct sockaddr_in6 *)remote;
1425						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1426						    rsin6)) {
1427							/* found it */
1428							if (netp != NULL) {
1429								*netp = net;
1430							}
1431							if (locked_tcb == NULL) {
1432								SCTP_INP_DECR_REF(inp);
1433							} else if (locked_tcb != stcb) {
1434								SCTP_TCB_LOCK(locked_tcb);
1435							}
1436							if (locked_tcb) {
1437								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1438							}
1439							SCTP_INP_WUNLOCK(inp);
1440							SCTP_INP_INFO_RUNLOCK();
1441							return (stcb);
1442						}
1443						break;
1444					}
1445#endif
1446				default:
1447					/* TSNH */
1448					break;
1449				}
1450			}
1451			SCTP_TCB_UNLOCK(stcb);
1452		}
1453	}
1454null_return:
1455	/* clean up for returning null */
1456	if (locked_tcb) {
1457		SCTP_TCB_LOCK(locked_tcb);
1458		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1459	}
1460	SCTP_INP_WUNLOCK(inp);
1461	SCTP_INP_INFO_RUNLOCK();
1462	/* not found */
1463	return (NULL);
1464}
1465
1466/*
1467 * Find an association for a specific endpoint using the association id given
1468 * out in the COMM_UP notification
1469 */
1470
1471struct sctp_tcb *
1472sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1473{
1474	/*
1475	 * Use my the assoc_id to find a endpoint
1476	 */
1477	struct sctpasochead *head;
1478	struct sctp_tcb *stcb;
1479	uint32_t id;
1480
1481	if (inp == NULL) {
1482		SCTP_PRINTF("TSNH ep_associd\n");
1483		return (NULL);
1484	}
1485	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1486		SCTP_PRINTF("TSNH ep_associd0\n");
1487		return (NULL);
1488	}
1489	id = (uint32_t) asoc_id;
1490	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1491	if (head == NULL) {
1492		/* invalid id TSNH */
1493		SCTP_PRINTF("TSNH ep_associd1\n");
1494		return (NULL);
1495	}
1496	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1497		if (stcb->asoc.assoc_id == id) {
1498			if (inp != stcb->sctp_ep) {
1499				/*
1500				 * some other guy has the same id active (id
1501				 * collision ??).
1502				 */
1503				SCTP_PRINTF("TSNH ep_associd2\n");
1504				continue;
1505			}
1506			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1507				continue;
1508			}
1509			if (want_lock) {
1510				SCTP_TCB_LOCK(stcb);
1511			}
1512			return (stcb);
1513		}
1514	}
1515	return (NULL);
1516}
1517
1518
1519struct sctp_tcb *
1520sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1521{
1522	struct sctp_tcb *stcb;
1523
1524	SCTP_INP_RLOCK(inp);
1525	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1526	SCTP_INP_RUNLOCK(inp);
1527	return (stcb);
1528}
1529
1530
1531static struct sctp_inpcb *
1532sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1533    uint16_t lport, uint32_t vrf_id)
1534{
1535	struct sctp_inpcb *inp;
1536	struct sctp_laddr *laddr;
1537
1538#ifdef INET
1539	struct sockaddr_in *sin;
1540
1541#endif
1542#ifdef INET6
1543	struct sockaddr_in6 *sin6;
1544	struct sockaddr_in6 *intf_addr6;
1545
1546#endif
1547
1548	int fnd;
1549
1550	/*
1551	 * Endpoint probe expects that the INP_INFO is locked.
1552	 */
1553#ifdef INET
1554	sin = NULL;
1555#endif
1556#ifdef INET6
1557	sin6 = NULL;
1558#endif
1559	switch (nam->sa_family) {
1560#ifdef INET
1561	case AF_INET:
1562		sin = (struct sockaddr_in *)nam;
1563		break;
1564#endif
1565#ifdef INET6
1566	case AF_INET6:
1567		sin6 = (struct sockaddr_in6 *)nam;
1568		break;
1569#endif
1570	default:
1571		/* unsupported family */
1572		return (NULL);
1573	}
1574
1575	if (head == NULL)
1576		return (NULL);
1577
1578	LIST_FOREACH(inp, head, sctp_hash) {
1579		SCTP_INP_RLOCK(inp);
1580		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1581			SCTP_INP_RUNLOCK(inp);
1582			continue;
1583		}
1584		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1585		    (inp->sctp_lport == lport)) {
1586			/* got it */
1587#ifdef INET
1588			if ((nam->sa_family == AF_INET) &&
1589			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1590			    SCTP_IPV6_V6ONLY(inp)) {
1591				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1592				SCTP_INP_RUNLOCK(inp);
1593				continue;
1594			}
1595#endif
1596#ifdef INET6
1597			/* A V6 address and the endpoint is NOT bound V6 */
1598			if (nam->sa_family == AF_INET6 &&
1599			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1600				SCTP_INP_RUNLOCK(inp);
1601				continue;
1602			}
1603#endif
1604			/* does a VRF id match? */
1605			fnd = 0;
1606			if (inp->def_vrf_id == vrf_id)
1607				fnd = 1;
1608
1609			SCTP_INP_RUNLOCK(inp);
1610			if (!fnd)
1611				continue;
1612			return (inp);
1613		}
1614		SCTP_INP_RUNLOCK(inp);
1615	}
1616	switch (nam->sa_family) {
1617#ifdef INET
1618	case AF_INET:
1619		if (sin->sin_addr.s_addr == INADDR_ANY) {
1620			/* Can't hunt for one that has no address specified */
1621			return (NULL);
1622		}
1623		break;
1624#endif
1625#ifdef INET6
1626	case AF_INET6:
1627		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1628			/* Can't hunt for one that has no address specified */
1629			return (NULL);
1630		}
1631		break;
1632#endif
1633	default:
1634		break;
1635	}
1636	/*
1637	 * ok, not bound to all so see if we can find a EP bound to this
1638	 * address.
1639	 */
1640	LIST_FOREACH(inp, head, sctp_hash) {
1641		SCTP_INP_RLOCK(inp);
1642		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1643			SCTP_INP_RUNLOCK(inp);
1644			continue;
1645		}
1646		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1647			SCTP_INP_RUNLOCK(inp);
1648			continue;
1649		}
1650		/*
1651		 * Ok this could be a likely candidate, look at all of its
1652		 * addresses
1653		 */
1654		if (inp->sctp_lport != lport) {
1655			SCTP_INP_RUNLOCK(inp);
1656			continue;
1657		}
1658		/* does a VRF id match? */
1659		fnd = 0;
1660		if (inp->def_vrf_id == vrf_id)
1661			fnd = 1;
1662
1663		if (!fnd) {
1664			SCTP_INP_RUNLOCK(inp);
1665			continue;
1666		}
1667		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1668			if (laddr->ifa == NULL) {
1669				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1670				    __FUNCTION__);
1671				continue;
1672			}
1673			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1674			    laddr->ifa);
1675			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1676				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1677				continue;
1678			}
1679			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1680				/* possible, see if it matches */
1681				switch (nam->sa_family) {
1682#ifdef INET
1683				case AF_INET:
1684					if (sin->sin_addr.s_addr ==
1685					    laddr->ifa->address.sin.sin_addr.s_addr) {
1686						SCTP_INP_RUNLOCK(inp);
1687						return (inp);
1688					}
1689					break;
1690#endif
1691#ifdef INET6
1692				case AF_INET6:
1693					intf_addr6 = &laddr->ifa->address.sin6;
1694					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1695					    intf_addr6)) {
1696						SCTP_INP_RUNLOCK(inp);
1697						return (inp);
1698					}
1699					break;
1700#endif
1701				}
1702			}
1703		}
1704		SCTP_INP_RUNLOCK(inp);
1705	}
1706	return (NULL);
1707}
1708
1709
1710static struct sctp_inpcb *
1711sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1712{
1713	struct sctppcbhead *head;
1714	struct sctp_inpcb *t_inp;
1715	int fnd;
1716
1717	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1718	    SCTP_BASE_INFO(hashmark))];
1719	LIST_FOREACH(t_inp, head, sctp_hash) {
1720		if (t_inp->sctp_lport != lport) {
1721			continue;
1722		}
1723		/* is it in the VRF in question */
1724		fnd = 0;
1725		if (t_inp->def_vrf_id == vrf_id)
1726			fnd = 1;
1727		if (!fnd)
1728			continue;
1729
1730		/* This one is in use. */
1731		/* check the v6/v4 binding issue */
1732		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1733		    SCTP_IPV6_V6ONLY(t_inp)) {
1734			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1735				/* collision in V6 space */
1736				return (t_inp);
1737			} else {
1738				/* inp is BOUND_V4 no conflict */
1739				continue;
1740			}
1741		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1742			/* t_inp is bound v4 and v6, conflict always */
1743			return (t_inp);
1744		} else {
1745			/* t_inp is bound only V4 */
1746			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1747			    SCTP_IPV6_V6ONLY(inp)) {
1748				/* no conflict */
1749				continue;
1750			}
1751			/* else fall through to conflict */
1752		}
1753		return (t_inp);
1754	}
1755	return (NULL);
1756}
1757
1758
1759int
1760sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1761{
1762	/* For 1-2-1 with port reuse */
1763	struct sctppcbhead *head;
1764	struct sctp_inpcb *tinp;
1765
1766	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1767		/* only works with port reuse on */
1768		return (-1);
1769	}
1770	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1771		return (0);
1772	}
1773	SCTP_INP_RUNLOCK(inp);
1774	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1775	    SCTP_BASE_INFO(hashmark))];
1776	/* Kick out all non-listeners to the TCP hash */
1777	LIST_FOREACH(tinp, head, sctp_hash) {
1778		if (tinp->sctp_lport != inp->sctp_lport) {
1779			continue;
1780		}
1781		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1782			continue;
1783		}
1784		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1785			continue;
1786		}
1787		if (tinp->sctp_socket->so_qlimit) {
1788			continue;
1789		}
1790		SCTP_INP_WLOCK(tinp);
1791		LIST_REMOVE(tinp, sctp_hash);
1792		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1793		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1794		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1795		SCTP_INP_WUNLOCK(tinp);
1796	}
1797	SCTP_INP_WLOCK(inp);
1798	/* Pull from where he was */
1799	LIST_REMOVE(inp, sctp_hash);
1800	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1801	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1802	LIST_INSERT_HEAD(head, inp, sctp_hash);
1803	SCTP_INP_WUNLOCK(inp);
1804	SCTP_INP_RLOCK(inp);
1805	return (0);
1806}
1807
1808
1809struct sctp_inpcb *
1810sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1811    uint32_t vrf_id)
1812{
1813	/*
1814	 * First we check the hash table to see if someone has this port
1815	 * bound with just the port.
1816	 */
1817	struct sctp_inpcb *inp;
1818	struct sctppcbhead *head;
1819	int lport;
1820	unsigned int i;
1821
1822#ifdef INET
1823	struct sockaddr_in *sin;
1824
1825#endif
1826#ifdef INET6
1827	struct sockaddr_in6 *sin6;
1828
1829#endif
1830
1831	switch (nam->sa_family) {
1832#ifdef INET
1833	case AF_INET:
1834		sin = (struct sockaddr_in *)nam;
1835		lport = sin->sin_port;
1836		break;
1837#endif
1838#ifdef INET6
1839	case AF_INET6:
1840		sin6 = (struct sockaddr_in6 *)nam;
1841		lport = sin6->sin6_port;
1842		break;
1843#endif
1844	default:
1845		return (NULL);
1846	}
1847	/*
1848	 * I could cheat here and just cast to one of the types but we will
1849	 * do it right. It also provides the check against an Unsupported
1850	 * type too.
1851	 */
1852	/* Find the head of the ALLADDR chain */
1853	if (have_lock == 0) {
1854		SCTP_INP_INFO_RLOCK();
1855	}
1856	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1857	    SCTP_BASE_INFO(hashmark))];
1858	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1859
1860	/*
1861	 * If the TCP model exists it could be that the main listening
1862	 * endpoint is gone but there still exists a connected socket for
1863	 * this guy. If so we can return the first one that we find. This
1864	 * may NOT be the correct one so the caller should be wary on the
1865	 * returned INP. Currently the only caller that sets find_tcp_pool
1866	 * is in bindx where we are verifying that a user CAN bind the
1867	 * address. He either has bound it already, or someone else has, or
1868	 * its open to bind, so this is good enough.
1869	 */
1870	if (inp == NULL && find_tcp_pool) {
1871		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1872			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1873			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1874			if (inp) {
1875				break;
1876			}
1877		}
1878	}
1879	if (inp) {
1880		SCTP_INP_INCR_REF(inp);
1881	}
1882	if (have_lock == 0) {
1883		SCTP_INP_INFO_RUNLOCK();
1884	}
1885	return (inp);
1886}
1887
1888/*
1889 * Find an association for an endpoint with the pointer to whom you want to
1890 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1891 * need to change the *to to some other struct like a mbuf...
1892 */
1893struct sctp_tcb *
1894sctp_findassociation_addr_sa(struct sockaddr *to, struct sockaddr *from,
1895    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1896    uint32_t vrf_id)
1897{
1898	struct sctp_inpcb *inp = NULL;
1899	struct sctp_tcb *retval;
1900
1901	SCTP_INP_INFO_RLOCK();
1902	if (find_tcp_pool) {
1903		if (inp_p != NULL) {
1904			retval = sctp_tcb_special_locate(inp_p, from, to, netp,
1905			    vrf_id);
1906		} else {
1907			retval = sctp_tcb_special_locate(&inp, from, to, netp,
1908			    vrf_id);
1909		}
1910		if (retval != NULL) {
1911			SCTP_INP_INFO_RUNLOCK();
1912			return (retval);
1913		}
1914	}
1915	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1916	if (inp_p != NULL) {
1917		*inp_p = inp;
1918	}
1919	SCTP_INP_INFO_RUNLOCK();
1920
1921	if (inp == NULL) {
1922		return (NULL);
1923	}
1924	/*
1925	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1926	 * we now place the source address or from in the to of the find
1927	 * endpoint call. Since in reality this chain is used from the
1928	 * inbound packet side.
1929	 */
1930	if (inp_p != NULL) {
1931		retval = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1932		    NULL);
1933	} else {
1934		retval = sctp_findassociation_ep_addr(&inp, from, netp, to,
1935		    NULL);
1936	}
1937	return retval;
1938}
1939
1940
1941/*
1942 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1943 * find all addresses that the sender has specified in any address list. Each
1944 * address will be used to lookup the TCB and see if one exits.
1945 */
1946static struct sctp_tcb *
1947sctp_findassociation_special_addr(struct mbuf *m, int offset,
1948    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1949    struct sockaddr *dest)
1950{
1951	struct sctp_paramhdr *phdr, parm_buf;
1952	struct sctp_tcb *retval;
1953	uint32_t ptype, plen;
1954
1955#ifdef INET
1956	struct sockaddr_in sin4;
1957
1958#endif
1959#ifdef INET6
1960	struct sockaddr_in6 sin6;
1961
1962#endif
1963
1964#ifdef INET
1965	memset(&sin4, 0, sizeof(sin4));
1966	sin4.sin_len = sizeof(sin4);
1967	sin4.sin_family = AF_INET;
1968	sin4.sin_port = sh->src_port;
1969#endif
1970#ifdef INET6
1971	memset(&sin6, 0, sizeof(sin6));
1972	sin6.sin6_len = sizeof(sin6);
1973	sin6.sin6_family = AF_INET6;
1974	sin6.sin6_port = sh->src_port;
1975#endif
1976
1977	retval = NULL;
1978	offset += sizeof(struct sctp_init_chunk);
1979
1980	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
1981	while (phdr != NULL) {
1982		/* now we must see if we want the parameter */
1983		ptype = ntohs(phdr->param_type);
1984		plen = ntohs(phdr->param_length);
1985		if (plen == 0) {
1986			break;
1987		}
1988#ifdef INET
1989		if (ptype == SCTP_IPV4_ADDRESS &&
1990		    plen == sizeof(struct sctp_ipv4addr_param)) {
1991			/* Get the rest of the address */
1992			struct sctp_ipv4addr_param ip4_parm, *p4;
1993
1994			phdr = sctp_get_next_param(m, offset,
1995			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
1996			if (phdr == NULL) {
1997				return (NULL);
1998			}
1999			p4 = (struct sctp_ipv4addr_param *)phdr;
2000			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2001			/* look it up */
2002			retval = sctp_findassociation_ep_addr(inp_p,
2003			    (struct sockaddr *)&sin4, netp, dest, NULL);
2004			if (retval != NULL) {
2005				return (retval);
2006			}
2007		}
2008#endif
2009#ifdef INET6
2010		if (ptype == SCTP_IPV6_ADDRESS &&
2011		    plen == sizeof(struct sctp_ipv6addr_param)) {
2012			/* Get the rest of the address */
2013			struct sctp_ipv6addr_param ip6_parm, *p6;
2014
2015			phdr = sctp_get_next_param(m, offset,
2016			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2017			if (phdr == NULL) {
2018				return (NULL);
2019			}
2020			p6 = (struct sctp_ipv6addr_param *)phdr;
2021			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2022			/* look it up */
2023			retval = sctp_findassociation_ep_addr(inp_p,
2024			    (struct sockaddr *)&sin6, netp, dest, NULL);
2025			if (retval != NULL) {
2026				return (retval);
2027			}
2028		}
2029#endif
2030		offset += SCTP_SIZE32(plen);
2031		phdr = sctp_get_next_param(m, offset, &parm_buf,
2032		    sizeof(parm_buf));
2033	}
2034	return (NULL);
2035}
2036
2037static struct sctp_tcb *
2038sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2039    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2040    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2041{
2042	/*
2043	 * Use my vtag to hash. If we find it we then verify the source addr
2044	 * is in the assoc. If all goes well we save a bit on rec of a
2045	 * packet.
2046	 */
2047	struct sctpasochead *head;
2048	struct sctp_nets *net;
2049	struct sctp_tcb *stcb;
2050
2051	*netp = NULL;
2052	*inp_p = NULL;
2053	SCTP_INP_INFO_RLOCK();
2054	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2055	    SCTP_BASE_INFO(hashasocmark))];
2056	if (head == NULL) {
2057		/* invalid vtag */
2058		SCTP_INP_INFO_RUNLOCK();
2059		return (NULL);
2060	}
2061	LIST_FOREACH(stcb, head, sctp_asocs) {
2062		SCTP_INP_RLOCK(stcb->sctp_ep);
2063		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2064			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2065			continue;
2066		}
2067		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2068			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2069			continue;
2070		}
2071		SCTP_TCB_LOCK(stcb);
2072		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2073		if (stcb->asoc.my_vtag == vtag) {
2074			/* candidate */
2075			if (stcb->rport != rport) {
2076				SCTP_TCB_UNLOCK(stcb);
2077				continue;
2078			}
2079			if (stcb->sctp_ep->sctp_lport != lport) {
2080				SCTP_TCB_UNLOCK(stcb);
2081				continue;
2082			}
2083			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2084				SCTP_TCB_UNLOCK(stcb);
2085				continue;
2086			}
2087			/* RRS:Need toaddr check here */
2088			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2089				/* Endpoint does not own this address */
2090				SCTP_TCB_UNLOCK(stcb);
2091				continue;
2092			}
2093			if (remote_tag) {
2094				/*
2095				 * If we have both vtags that's all we match
2096				 * on
2097				 */
2098				if (stcb->asoc.peer_vtag == remote_tag) {
2099					/*
2100					 * If both tags match we consider it
2101					 * conclusive and check NO
2102					 * source/destination addresses
2103					 */
2104					goto conclusive;
2105				}
2106			}
2107			if (skip_src_check) {
2108		conclusive:
2109				if (from) {
2110					*netp = sctp_findnet(stcb, from);
2111				} else {
2112					*netp = NULL;	/* unknown */
2113				}
2114				if (inp_p)
2115					*inp_p = stcb->sctp_ep;
2116				SCTP_INP_INFO_RUNLOCK();
2117				return (stcb);
2118			}
2119			net = sctp_findnet(stcb, from);
2120			if (net) {
2121				/* yep its him. */
2122				*netp = net;
2123				SCTP_STAT_INCR(sctps_vtagexpress);
2124				*inp_p = stcb->sctp_ep;
2125				SCTP_INP_INFO_RUNLOCK();
2126				return (stcb);
2127			} else {
2128				/*
2129				 * not him, this should only happen in rare
2130				 * cases so I peg it.
2131				 */
2132				SCTP_STAT_INCR(sctps_vtagbogus);
2133			}
2134		}
2135		SCTP_TCB_UNLOCK(stcb);
2136	}
2137	SCTP_INP_INFO_RUNLOCK();
2138	return (NULL);
2139}
2140
2141/*
2142 * Find an association with the pointer to the inbound IP packet. This can be
2143 * a IPv4 or IPv6 packet.
2144 */
2145struct sctp_tcb *
2146sctp_findassociation_addr(struct mbuf *m, int offset,
2147    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2148    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2149{
2150	int find_tcp_pool;
2151	struct ip *iph;
2152	struct sctp_tcb *retval;
2153	struct sockaddr_storage to_store, from_store;
2154	struct sockaddr *to = (struct sockaddr *)&to_store;
2155	struct sockaddr *from = (struct sockaddr *)&from_store;
2156	struct sctp_inpcb *inp;
2157
2158	iph = mtod(m, struct ip *);
2159	switch (iph->ip_v) {
2160#ifdef INET
2161	case IPVERSION:
2162		{
2163			/* its IPv4 */
2164			struct sockaddr_in *from4;
2165
2166			from4 = (struct sockaddr_in *)&from_store;
2167			bzero(from4, sizeof(*from4));
2168			from4->sin_family = AF_INET;
2169			from4->sin_len = sizeof(struct sockaddr_in);
2170			from4->sin_addr.s_addr = iph->ip_src.s_addr;
2171			from4->sin_port = sh->src_port;
2172			break;
2173		}
2174#endif
2175#ifdef INET6
2176	case IPV6_VERSION >> 4:
2177		{
2178			/* its IPv6 */
2179			struct ip6_hdr *ip6;
2180			struct sockaddr_in6 *from6;
2181
2182			ip6 = mtod(m, struct ip6_hdr *);
2183			from6 = (struct sockaddr_in6 *)&from_store;
2184			bzero(from6, sizeof(*from6));
2185			from6->sin6_family = AF_INET6;
2186			from6->sin6_len = sizeof(struct sockaddr_in6);
2187			from6->sin6_addr = ip6->ip6_src;
2188			from6->sin6_port = sh->src_port;
2189			/* Get the scopes in properly to the sin6 addr's */
2190			/* we probably don't need these operations */
2191			(void)sa6_recoverscope(from6);
2192			sa6_embedscope(from6, MODULE_GLOBAL(ip6_use_defzone));
2193			break;
2194		}
2195#endif
2196	default:
2197		/* Currently not supported. */
2198		return (NULL);
2199	}
2200
2201
2202	switch (iph->ip_v) {
2203#ifdef INET
2204	case IPVERSION:
2205		{
2206			/* its IPv4 */
2207			struct sockaddr_in *to4;
2208
2209			to4 = (struct sockaddr_in *)&to_store;
2210			bzero(to4, sizeof(*to4));
2211			to4->sin_family = AF_INET;
2212			to4->sin_len = sizeof(struct sockaddr_in);
2213			to4->sin_addr.s_addr = iph->ip_dst.s_addr;
2214			to4->sin_port = sh->dest_port;
2215			break;
2216		}
2217#endif
2218#ifdef INET6
2219	case IPV6_VERSION >> 4:
2220		{
2221			/* its IPv6 */
2222			struct ip6_hdr *ip6;
2223			struct sockaddr_in6 *to6;
2224
2225			ip6 = mtod(m, struct ip6_hdr *);
2226			to6 = (struct sockaddr_in6 *)&to_store;
2227			bzero(to6, sizeof(*to6));
2228			to6->sin6_family = AF_INET6;
2229			to6->sin6_len = sizeof(struct sockaddr_in6);
2230			to6->sin6_addr = ip6->ip6_dst;
2231			to6->sin6_port = sh->dest_port;
2232			/* Get the scopes in properly to the sin6 addr's */
2233			/* we probably don't need these operations */
2234			(void)sa6_recoverscope(to6);
2235			sa6_embedscope(to6, MODULE_GLOBAL(ip6_use_defzone));
2236			break;
2237		}
2238#endif
2239	default:
2240		/* TSNH */
2241		break;
2242	}
2243	if (sh->v_tag) {
2244		/* we only go down this path if vtag is non-zero */
2245		retval = sctp_findassoc_by_vtag(from, to, ntohl(sh->v_tag),
2246		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2247		if (retval) {
2248			return (retval);
2249		}
2250	}
2251	find_tcp_pool = 0;
2252	if ((ch->chunk_type != SCTP_INITIATION) &&
2253	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2254	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2255	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2256		/* Other chunk types go to the tcp pool. */
2257		find_tcp_pool = 1;
2258	}
2259	if (inp_p) {
2260		retval = sctp_findassociation_addr_sa(to, from, inp_p, netp,
2261		    find_tcp_pool, vrf_id);
2262		inp = *inp_p;
2263	} else {
2264		retval = sctp_findassociation_addr_sa(to, from, &inp, netp,
2265		    find_tcp_pool, vrf_id);
2266	}
2267	SCTPDBG(SCTP_DEBUG_PCB1, "retval:%p inp:%p\n", retval, inp);
2268	if (retval == NULL && inp) {
2269		/* Found a EP but not this address */
2270		if ((ch->chunk_type == SCTP_INITIATION) ||
2271		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2272			/*-
2273			 * special hook, we do NOT return linp or an
2274			 * association that is linked to an existing
2275			 * association that is under the TCP pool (i.e. no
2276			 * listener exists). The endpoint finding routine
2277			 * will always find a listener before examining the
2278			 * TCP pool.
2279			 */
2280			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2281				if (inp_p) {
2282					*inp_p = NULL;
2283				}
2284				return (NULL);
2285			}
2286			retval = sctp_findassociation_special_addr(m,
2287			    offset, sh, &inp, netp, to);
2288			if (inp_p != NULL) {
2289				*inp_p = inp;
2290			}
2291		}
2292	}
2293	SCTPDBG(SCTP_DEBUG_PCB1, "retval is %p\n", retval);
2294	return (retval);
2295}
2296
2297/*
2298 * lookup an association by an ASCONF lookup address.
2299 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2300 */
2301struct sctp_tcb *
2302sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2303    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2304{
2305	struct sctp_tcb *stcb;
2306	struct sockaddr_storage local_store, remote_store;
2307	struct sockaddr *to;
2308	struct ip *iph;
2309	struct sctp_paramhdr parm_buf, *phdr;
2310	int ptype;
2311	int zero_address = 0;
2312
2313#ifdef INET
2314	struct sockaddr_in *sin;
2315
2316#endif
2317#ifdef INET6
2318	struct ip6_hdr *ip6;
2319	struct sockaddr_in6 *sin6;
2320
2321#endif
2322
2323	memset(&local_store, 0, sizeof(local_store));
2324	memset(&remote_store, 0, sizeof(remote_store));
2325	to = (struct sockaddr *)&local_store;
2326	/* First get the destination address setup too. */
2327	iph = mtod(m, struct ip *);
2328	switch (iph->ip_v) {
2329#ifdef INET
2330	case IPVERSION:
2331		/* its IPv4 */
2332		sin = (struct sockaddr_in *)&local_store;
2333		sin->sin_family = AF_INET;
2334		sin->sin_len = sizeof(*sin);
2335		sin->sin_port = sh->dest_port;
2336		sin->sin_addr.s_addr = iph->ip_dst.s_addr;
2337		break;
2338#endif
2339#ifdef INET6
2340	case IPV6_VERSION >> 4:
2341		/* its IPv6 */
2342		ip6 = mtod(m, struct ip6_hdr *);
2343		sin6 = (struct sockaddr_in6 *)&local_store;
2344		sin6->sin6_family = AF_INET6;
2345		sin6->sin6_len = sizeof(*sin6);
2346		sin6->sin6_port = sh->dest_port;
2347		sin6->sin6_addr = ip6->ip6_dst;
2348		break;
2349#endif
2350	default:
2351		return NULL;
2352	}
2353
2354	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2355	    &parm_buf, sizeof(struct sctp_paramhdr));
2356	if (phdr == NULL) {
2357		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2358		    __FUNCTION__);
2359		return NULL;
2360	}
2361	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2362	/* get the correlation address */
2363	switch (ptype) {
2364#ifdef INET6
2365	case SCTP_IPV6_ADDRESS:
2366		{
2367			/* ipv6 address param */
2368			struct sctp_ipv6addr_param *p6, p6_buf;
2369
2370			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2371				return NULL;
2372			}
2373			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2374			    offset + sizeof(struct sctp_asconf_chunk),
2375			    &p6_buf.ph, sizeof(*p6));
2376			if (p6 == NULL) {
2377				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2378				    __FUNCTION__);
2379				return (NULL);
2380			}
2381			sin6 = (struct sockaddr_in6 *)&remote_store;
2382			sin6->sin6_family = AF_INET6;
2383			sin6->sin6_len = sizeof(*sin6);
2384			sin6->sin6_port = sh->src_port;
2385			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2386			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2387				zero_address = 1;
2388			break;
2389		}
2390#endif
2391#ifdef INET
2392	case SCTP_IPV4_ADDRESS:
2393		{
2394			/* ipv4 address param */
2395			struct sctp_ipv4addr_param *p4, p4_buf;
2396
2397			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2398				return NULL;
2399			}
2400			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2401			    offset + sizeof(struct sctp_asconf_chunk),
2402			    &p4_buf.ph, sizeof(*p4));
2403			if (p4 == NULL) {
2404				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2405				    __FUNCTION__);
2406				return (NULL);
2407			}
2408			sin = (struct sockaddr_in *)&remote_store;
2409			sin->sin_family = AF_INET;
2410			sin->sin_len = sizeof(*sin);
2411			sin->sin_port = sh->src_port;
2412			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2413			if (sin->sin_addr.s_addr == INADDR_ANY)
2414				zero_address = 1;
2415			break;
2416		}
2417#endif
2418	default:
2419		/* invalid address param type */
2420		return NULL;
2421	}
2422
2423	if (zero_address) {
2424		stcb = sctp_findassoc_by_vtag(NULL, to, ntohl(sh->v_tag), inp_p,
2425		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2426		/*
2427		 * SCTP_PRINTF("findassociation_ep_asconf: zero lookup
2428		 * address finds stcb 0x%x\n", (uint32_t)stcb);
2429		 */
2430	} else {
2431		stcb = sctp_findassociation_ep_addr(inp_p,
2432		    (struct sockaddr *)&remote_store, netp,
2433		    to, NULL);
2434	}
2435	return (stcb);
2436}
2437
2438
2439/*
2440 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2441 * addresses. This way if we don't get a bind we by default pick a ephemeral
2442 * port with all addresses bound.
2443 */
2444int
2445sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2446{
2447	/*
2448	 * we get called when a new endpoint starts up. We need to allocate
2449	 * the sctp_inpcb structure from the zone and init it. Mark it as
2450	 * unbound and find a port that we can use as an ephemeral with
2451	 * INADDR_ANY. If the user binds later no problem we can then add in
2452	 * the specific addresses. And setup the default parameters for the
2453	 * EP.
2454	 */
2455	int i, error;
2456	struct sctp_inpcb *inp;
2457	struct sctp_pcb *m;
2458	struct timeval time;
2459	sctp_sharedkey_t *null_key;
2460
2461	error = 0;
2462
2463	SCTP_INP_INFO_WLOCK();
2464	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2465	if (inp == NULL) {
2466		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2467		SCTP_INP_INFO_WUNLOCK();
2468		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2469		return (ENOBUFS);
2470	}
2471	/* zap it */
2472	bzero(inp, sizeof(*inp));
2473
2474	/* bump generations */
2475	/* setup socket pointers */
2476	inp->sctp_socket = so;
2477	inp->ip_inp.inp.inp_socket = so;
2478#ifdef INET6
2479	if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2480		inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2481	}
2482#endif
2483	inp->sctp_associd_counter = 1;
2484	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2485	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2486	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2487	inp->sctp_ecn_enable = SCTP_BASE_SYSCTL(sctp_ecn_enable);
2488	/* init the small hash table we use to track asocid <-> tcb */
2489	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2490	if (inp->sctp_asocidhash == NULL) {
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		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2506		SCTP_INP_INFO_WUNLOCK();
2507		return error;
2508	}
2509#endif				/* IPSEC */
2510	SCTP_INCR_EP_COUNT();
2511	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2512	SCTP_INP_INFO_WUNLOCK();
2513
2514	so->so_pcb = (caddr_t)inp;
2515
2516	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2517		/* UDP style socket */
2518		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2519		    SCTP_PCB_FLAGS_UNBOUND);
2520		/* Be sure it is NON-BLOCKING IO for UDP */
2521		/* SCTP_SET_SO_NBIO(so); */
2522	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2523		/* TCP style socket */
2524		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2525		    SCTP_PCB_FLAGS_UNBOUND);
2526		/* Be sure we have blocking IO by default */
2527		SCTP_CLEAR_SO_NBIO(so);
2528	} else {
2529		/*
2530		 * unsupported socket type (RAW, etc)- in case we missed it
2531		 * in protosw
2532		 */
2533		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2534		so->so_pcb = NULL;
2535		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2536		return (EOPNOTSUPP);
2537	}
2538	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2539		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2540		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2541	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2542		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2543		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2544	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2545		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2546		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2547	}
2548	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2549	    &inp->sctp_hashmark);
2550	if (inp->sctp_tcbhash == NULL) {
2551		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2552		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2553		so->so_pcb = NULL;
2554		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2555		return (ENOBUFS);
2556	}
2557	inp->def_vrf_id = vrf_id;
2558
2559	SCTP_INP_INFO_WLOCK();
2560	SCTP_INP_LOCK_INIT(inp);
2561	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2562	SCTP_INP_READ_INIT(inp);
2563	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2564	/* lock the new ep */
2565	SCTP_INP_WLOCK(inp);
2566
2567	/* add it to the info area */
2568	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2569	SCTP_INP_INFO_WUNLOCK();
2570
2571	TAILQ_INIT(&inp->read_queue);
2572	LIST_INIT(&inp->sctp_addr_list);
2573
2574	LIST_INIT(&inp->sctp_asoc_list);
2575
2576#ifdef SCTP_TRACK_FREED_ASOCS
2577	/* TEMP CODE */
2578	LIST_INIT(&inp->sctp_asoc_free_list);
2579#endif
2580	/* Init the timer structure for signature change */
2581	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2582	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2583
2584	/* now init the actual endpoint default data */
2585	m = &inp->sctp_ep;
2586
2587	/* setup the base timeout information */
2588	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2589	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2590	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2591	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2592	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2593	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2594	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2595	/* all max/min max are in ms */
2596	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2597	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2598	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2599	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2600	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2601
2602	m->max_open_streams_intome = MAX_SCTP_STREAMS;
2603
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	/* number of streams to pre-open on a association */
2616	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2617
2618	/* Add adaptation cookie */
2619	m->adaptation_layer_indicator = 0x504C5253;
2620
2621	/* seed random number generator */
2622	m->random_counter = 1;
2623	m->store_at = SCTP_SIGNATURE_SIZE;
2624	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2625	sctp_fill_random_store(m);
2626
2627	/* Minimum cookie size */
2628	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2629	    sizeof(struct sctp_state_cookie);
2630	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2631
2632	/* Setup the initial secret */
2633	(void)SCTP_GETTIME_TIMEVAL(&time);
2634	m->time_of_secret_change = time.tv_sec;
2635
2636	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2637		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2638	}
2639	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2640
2641	/* How long is a cookie good for ? */
2642	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2643	/*
2644	 * Initialize authentication parameters
2645	 */
2646	m->local_hmacs = sctp_default_supported_hmaclist();
2647	m->local_auth_chunks = sctp_alloc_chunklist();
2648	m->default_dscp = 0;
2649#ifdef INET6
2650	m->default_flowlabel = 0;
2651#endif
2652	m->port = 0;		/* encapsulation disabled by default */
2653	sctp_auth_set_default_chunks(m->local_auth_chunks);
2654	LIST_INIT(&m->shared_keys);
2655	/* add default NULL key as key id 0 */
2656	null_key = sctp_alloc_sharedkey();
2657	sctp_insert_sharedkey(&m->shared_keys, null_key);
2658	SCTP_INP_WUNLOCK(inp);
2659#ifdef SCTP_LOG_CLOSING
2660	sctp_log_closing(inp, NULL, 12);
2661#endif
2662	return (error);
2663}
2664
2665
2666void
2667sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2668    struct sctp_tcb *stcb)
2669{
2670	struct sctp_nets *net;
2671	uint16_t lport, rport;
2672	struct sctppcbhead *head;
2673	struct sctp_laddr *laddr, *oladdr;
2674
2675	atomic_add_int(&stcb->asoc.refcnt, 1);
2676	SCTP_TCB_UNLOCK(stcb);
2677	SCTP_INP_INFO_WLOCK();
2678	SCTP_INP_WLOCK(old_inp);
2679	SCTP_INP_WLOCK(new_inp);
2680	SCTP_TCB_LOCK(stcb);
2681	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2682
2683	new_inp->sctp_ep.time_of_secret_change =
2684	    old_inp->sctp_ep.time_of_secret_change;
2685	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2686	    sizeof(old_inp->sctp_ep.secret_key));
2687	new_inp->sctp_ep.current_secret_number =
2688	    old_inp->sctp_ep.current_secret_number;
2689	new_inp->sctp_ep.last_secret_number =
2690	    old_inp->sctp_ep.last_secret_number;
2691	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2692
2693	/* make it so new data pours into the new socket */
2694	stcb->sctp_socket = new_inp->sctp_socket;
2695	stcb->sctp_ep = new_inp;
2696
2697	/* Copy the port across */
2698	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2699	rport = stcb->rport;
2700	/* Pull the tcb from the old association */
2701	LIST_REMOVE(stcb, sctp_tcbhash);
2702	LIST_REMOVE(stcb, sctp_tcblist);
2703	if (stcb->asoc.in_asocid_hash) {
2704		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2705	}
2706	/* Now insert the new_inp into the TCP connected hash */
2707	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2708
2709	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2710	/* Its safe to access */
2711	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2712
2713	/* Now move the tcb into the endpoint list */
2714	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2715	/*
2716	 * Question, do we even need to worry about the ep-hash since we
2717	 * only have one connection? Probably not :> so lets get rid of it
2718	 * and not suck up any kernel memory in that.
2719	 */
2720	if (stcb->asoc.in_asocid_hash) {
2721		struct sctpasochead *lhd;
2722
2723		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2724		    new_inp->hashasocidmark)];
2725		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2726	}
2727	/* Ok. Let's restart timer. */
2728	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2729		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2730		    stcb, net);
2731	}
2732
2733	SCTP_INP_INFO_WUNLOCK();
2734	if (new_inp->sctp_tcbhash != NULL) {
2735		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2736		new_inp->sctp_tcbhash = NULL;
2737	}
2738	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2739		/* Subset bound, so copy in the laddr list from the old_inp */
2740		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2741			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2742			if (laddr == NULL) {
2743				/*
2744				 * Gak, what can we do? This assoc is really
2745				 * HOSED. We probably should send an abort
2746				 * here.
2747				 */
2748				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2749				continue;
2750			}
2751			SCTP_INCR_LADDR_COUNT();
2752			bzero(laddr, sizeof(*laddr));
2753			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2754			laddr->ifa = oladdr->ifa;
2755			atomic_add_int(&laddr->ifa->refcount, 1);
2756			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2757			    sctp_nxt_addr);
2758			new_inp->laddr_count++;
2759			if (oladdr == stcb->asoc.last_used_address) {
2760				stcb->asoc.last_used_address = laddr;
2761			}
2762		}
2763	}
2764	/*
2765	 * Now any running timers need to be adjusted since we really don't
2766	 * care if they are running or not just blast in the new_inp into
2767	 * all of them.
2768	 */
2769
2770	stcb->asoc.dack_timer.ep = (void *)new_inp;
2771	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2772	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2773	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2774	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2775	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2776	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2777	/* now what about the nets? */
2778	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2779		net->pmtu_timer.ep = (void *)new_inp;
2780		net->hb_timer.ep = (void *)new_inp;
2781		net->rxt_timer.ep = (void *)new_inp;
2782	}
2783	SCTP_INP_WUNLOCK(new_inp);
2784	SCTP_INP_WUNLOCK(old_inp);
2785}
2786
2787
2788
2789
2790/* sctp_ifap is used to bypass normal local address validation checks */
2791int
2792sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2793    struct sctp_ifa *sctp_ifap, struct thread *p)
2794{
2795	/* bind a ep to a socket address */
2796	struct sctppcbhead *head;
2797	struct sctp_inpcb *inp, *inp_tmp;
2798	struct inpcb *ip_inp;
2799	int port_reuse_active = 0;
2800	int bindall;
2801	uint16_t lport;
2802	int error;
2803	uint32_t vrf_id;
2804
2805	lport = 0;
2806	error = 0;
2807	bindall = 1;
2808	inp = (struct sctp_inpcb *)so->so_pcb;
2809	ip_inp = (struct inpcb *)so->so_pcb;
2810#ifdef SCTP_DEBUG
2811	if (addr) {
2812		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n",
2813		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2814		SCTPDBG(SCTP_DEBUG_PCB1, "Addr :");
2815		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2816	}
2817#endif
2818	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2819		/* already did a bind, subsequent binds NOT allowed ! */
2820		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2821		return (EINVAL);
2822	}
2823#ifdef INVARIANTS
2824	if (p == NULL)
2825		panic("null proc/thread");
2826#endif
2827	if (addr != NULL) {
2828		switch (addr->sa_family) {
2829#ifdef INET
2830		case AF_INET:
2831			{
2832				struct sockaddr_in *sin;
2833
2834				/* IPV6_V6ONLY socket? */
2835				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2836					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2837					return (EINVAL);
2838				}
2839				if (addr->sa_len != sizeof(*sin)) {
2840					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2841					return (EINVAL);
2842				}
2843				sin = (struct sockaddr_in *)addr;
2844				lport = sin->sin_port;
2845				/*
2846				 * For LOOPBACK the prison_local_ip4() call
2847				 * will transmute the ip address to the
2848				 * proper value.
2849				 */
2850				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2851					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2852					return (error);
2853				}
2854				if (sin->sin_addr.s_addr != INADDR_ANY) {
2855					bindall = 0;
2856				}
2857				break;
2858			}
2859#endif
2860#ifdef INET6
2861		case AF_INET6:
2862			{
2863				/*
2864				 * Only for pure IPv6 Address. (No IPv4
2865				 * Mapped!)
2866				 */
2867				struct sockaddr_in6 *sin6;
2868
2869				sin6 = (struct sockaddr_in6 *)addr;
2870
2871				if (addr->sa_len != sizeof(*sin6)) {
2872					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2873					return (EINVAL);
2874				}
2875				lport = sin6->sin6_port;
2876
2877				/*
2878				 * For LOOPBACK the prison_local_ip6() call
2879				 * will transmute the ipv6 address to the
2880				 * proper value.
2881				 */
2882				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2883				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2884					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2885					return (error);
2886				}
2887				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2888					bindall = 0;
2889					/* KAME hack: embed scopeid */
2890					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2891						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2892						return (EINVAL);
2893					}
2894				}
2895				/* this must be cleared for ifa_ifwithaddr() */
2896				sin6->sin6_scope_id = 0;
2897				break;
2898			}
2899#endif
2900		default:
2901			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2902			return (EAFNOSUPPORT);
2903		}
2904	}
2905	SCTP_INP_INFO_WLOCK();
2906	SCTP_INP_WLOCK(inp);
2907	/* Setup a vrf_id to be the default for the non-bind-all case. */
2908	vrf_id = inp->def_vrf_id;
2909
2910	/* increase our count due to the unlock we do */
2911	SCTP_INP_INCR_REF(inp);
2912	if (lport) {
2913		/*
2914		 * Did the caller specify a port? if so we must see if a ep
2915		 * already has this one bound.
2916		 */
2917		/* got to be root to get at low ports */
2918		if (ntohs(lport) < IPPORT_RESERVED) {
2919			if (p && (error =
2920			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2921			    )) {
2922				SCTP_INP_DECR_REF(inp);
2923				SCTP_INP_WUNLOCK(inp);
2924				SCTP_INP_INFO_WUNLOCK();
2925				return (error);
2926			}
2927		}
2928		if (p == NULL) {
2929			SCTP_INP_DECR_REF(inp);
2930			SCTP_INP_WUNLOCK(inp);
2931			SCTP_INP_INFO_WUNLOCK();
2932			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2933			return (error);
2934		}
2935		SCTP_INP_WUNLOCK(inp);
2936		if (bindall) {
2937			vrf_id = inp->def_vrf_id;
2938			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2939			if (inp_tmp != NULL) {
2940				/*
2941				 * lock guy returned and lower count note
2942				 * that we are not bound so inp_tmp should
2943				 * NEVER be inp. And it is this inp
2944				 * (inp_tmp) that gets the reference bump,
2945				 * so we must lower it.
2946				 */
2947				SCTP_INP_DECR_REF(inp_tmp);
2948				/* unlock info */
2949				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2950				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2951					/*
2952					 * Ok, must be one-2-one and
2953					 * allowing port re-use
2954					 */
2955					port_reuse_active = 1;
2956					goto continue_anyway;
2957				}
2958				SCTP_INP_DECR_REF(inp);
2959				SCTP_INP_INFO_WUNLOCK();
2960				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2961				return (EADDRINUSE);
2962			}
2963		} else {
2964			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2965			if (inp_tmp != NULL) {
2966				/*
2967				 * lock guy returned and lower count note
2968				 * that we are not bound so inp_tmp should
2969				 * NEVER be inp. And it is this inp
2970				 * (inp_tmp) that gets the reference bump,
2971				 * so we must lower it.
2972				 */
2973				SCTP_INP_DECR_REF(inp_tmp);
2974				/* unlock info */
2975				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2976				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2977					/*
2978					 * Ok, must be one-2-one and
2979					 * allowing port re-use
2980					 */
2981					port_reuse_active = 1;
2982					goto continue_anyway;
2983				}
2984				SCTP_INP_DECR_REF(inp);
2985				SCTP_INP_INFO_WUNLOCK();
2986				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2987				return (EADDRINUSE);
2988			}
2989		}
2990continue_anyway:
2991		SCTP_INP_WLOCK(inp);
2992		if (bindall) {
2993			/* verify that no lport is not used by a singleton */
2994			if ((port_reuse_active == 0) &&
2995			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))
2996			    ) {
2997				/* Sorry someone already has this one bound */
2998				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2999				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
3000					port_reuse_active = 1;
3001				} else {
3002					SCTP_INP_DECR_REF(inp);
3003					SCTP_INP_WUNLOCK(inp);
3004					SCTP_INP_INFO_WUNLOCK();
3005					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3006					return (EADDRINUSE);
3007				}
3008			}
3009		}
3010	} else {
3011		uint16_t first, last, candidate;
3012		uint16_t count;
3013		int done;
3014
3015		if (ip_inp->inp_flags & INP_HIGHPORT) {
3016			first = MODULE_GLOBAL(ipport_hifirstauto);
3017			last = MODULE_GLOBAL(ipport_hilastauto);
3018		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3019			if (p && (error =
3020			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3021			    )) {
3022				SCTP_INP_DECR_REF(inp);
3023				SCTP_INP_WUNLOCK(inp);
3024				SCTP_INP_INFO_WUNLOCK();
3025				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3026				return (error);
3027			}
3028			first = MODULE_GLOBAL(ipport_lowfirstauto);
3029			last = MODULE_GLOBAL(ipport_lowlastauto);
3030		} else {
3031			first = MODULE_GLOBAL(ipport_firstauto);
3032			last = MODULE_GLOBAL(ipport_lastauto);
3033		}
3034		if (first > last) {
3035			uint16_t temp;
3036
3037			temp = first;
3038			first = last;
3039			last = temp;
3040		}
3041		count = last - first + 1;	/* number of candidates */
3042		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3043
3044		done = 0;
3045		while (!done) {
3046			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3047				done = 1;
3048			}
3049			if (!done) {
3050				if (--count == 0) {
3051					SCTP_INP_DECR_REF(inp);
3052					SCTP_INP_WUNLOCK(inp);
3053					SCTP_INP_INFO_WUNLOCK();
3054					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3055					return (EADDRINUSE);
3056				}
3057				if (candidate == last)
3058					candidate = first;
3059				else
3060					candidate = candidate + 1;
3061			}
3062		}
3063		lport = htons(candidate);
3064	}
3065	SCTP_INP_DECR_REF(inp);
3066	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3067	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3068		/*
3069		 * this really should not happen. The guy did a non-blocking
3070		 * bind and then did a close at the same time.
3071		 */
3072		SCTP_INP_WUNLOCK(inp);
3073		SCTP_INP_INFO_WUNLOCK();
3074		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3075		return (EINVAL);
3076	}
3077	/* ok we look clear to give out this port, so lets setup the binding */
3078	if (bindall) {
3079		/* binding to all addresses, so just set in the proper flags */
3080		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3081		/* set the automatic addr changes from kernel flag */
3082		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3083			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3084			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3085		} else {
3086			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3087			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3088		}
3089		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3090			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3091		} else {
3092			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3093		}
3094		/*
3095		 * set the automatic mobility_base from kernel flag (by
3096		 * micchie)
3097		 */
3098		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3099			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3100			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3101		} else {
3102			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3103			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3104		}
3105		/*
3106		 * set the automatic mobility_fasthandoff from kernel flag
3107		 * (by micchie)
3108		 */
3109		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3110			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3111			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3112		} else {
3113			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3114			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3115		}
3116	} else {
3117		/*
3118		 * bind specific, make sure flags is off and add a new
3119		 * address structure to the sctp_addr_list inside the ep
3120		 * structure.
3121		 *
3122		 * We will need to allocate one and insert it at the head. The
3123		 * socketopt call can just insert new addresses in there as
3124		 * well. It will also have to do the embed scope kame hack
3125		 * too (before adding).
3126		 */
3127		struct sctp_ifa *ifa;
3128		struct sockaddr_storage store_sa;
3129
3130		memset(&store_sa, 0, sizeof(store_sa));
3131		switch (addr->sa_family) {
3132		case AF_INET:
3133			{
3134				struct sockaddr_in *sin;
3135
3136				sin = (struct sockaddr_in *)&store_sa;
3137				memcpy(sin, addr, sizeof(struct sockaddr_in));
3138				sin->sin_port = 0;
3139				break;
3140			}
3141		case AF_INET6:
3142			{
3143				struct sockaddr_in6 *sin6;
3144
3145				sin6 = (struct sockaddr_in6 *)&store_sa;
3146				memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3147				sin6->sin6_port = 0;
3148				break;
3149			}
3150		default:
3151			break;
3152		}
3153		/*
3154		 * first find the interface with the bound address need to
3155		 * zero out the port to find the address! yuck! can't do
3156		 * this earlier since need port for sctp_pcb_findep()
3157		 */
3158		if (sctp_ifap != NULL)
3159			ifa = sctp_ifap;
3160		else {
3161			/*
3162			 * Note for BSD we hit here always other O/S's will
3163			 * pass things in via the sctp_ifap argument
3164			 * (Panda).
3165			 */
3166			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3167			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3168		}
3169		if (ifa == NULL) {
3170			/* Can't find an interface with that address */
3171			SCTP_INP_WUNLOCK(inp);
3172			SCTP_INP_INFO_WUNLOCK();
3173			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3174			return (EADDRNOTAVAIL);
3175		}
3176#ifdef INET6
3177		if (addr->sa_family == AF_INET6) {
3178			/* GAK, more FIXME IFA lock? */
3179			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3180				/* Can't bind a non-existent addr. */
3181				SCTP_INP_WUNLOCK(inp);
3182				SCTP_INP_INFO_WUNLOCK();
3183				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3184				return (EINVAL);
3185			}
3186		}
3187#endif
3188		/* we're not bound all */
3189		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3190		/* allow bindx() to send ASCONF's for binding changes */
3191		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3192		/* clear automatic addr changes from kernel flag */
3193		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3194
3195		/* add this address to the endpoint list */
3196		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3197		if (error != 0) {
3198			SCTP_INP_WUNLOCK(inp);
3199			SCTP_INP_INFO_WUNLOCK();
3200			return (error);
3201		}
3202		inp->laddr_count++;
3203	}
3204	/* find the bucket */
3205	if (port_reuse_active) {
3206		/* Put it into tcp 1-2-1 hash */
3207		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3208		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3209	} else {
3210		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3211	}
3212	/* put it in the bucket */
3213	LIST_INSERT_HEAD(head, inp, sctp_hash);
3214	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3215	    head, ntohs(lport), port_reuse_active);
3216	/* set in the port */
3217	inp->sctp_lport = lport;
3218
3219	/* turn off just the unbound flag */
3220	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3221	SCTP_INP_WUNLOCK(inp);
3222	SCTP_INP_INFO_WUNLOCK();
3223	return (0);
3224}
3225
3226
3227static void
3228sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3229{
3230	struct sctp_iterator *it, *nit;
3231
3232	/*
3233	 * We enter with the only the ITERATOR_LOCK in place and a write
3234	 * lock on the inp_info stuff.
3235	 */
3236	it = sctp_it_ctl.cur_it;
3237	if (it && (it->vn != curvnet)) {
3238		/* Its not looking at our VNET */
3239		return;
3240	}
3241	if (it && (it->inp == inp)) {
3242		/*
3243		 * This is tricky and we hold the iterator lock, but when it
3244		 * returns and gets the lock (when we release it) the
3245		 * iterator will try to operate on inp. We need to stop that
3246		 * from happening. But of course the iterator has a
3247		 * reference on the stcb and inp. We can mark it and it will
3248		 * stop.
3249		 *
3250		 * If its a single iterator situation, we set the end iterator
3251		 * flag. Otherwise we set the iterator to go to the next
3252		 * inp.
3253		 *
3254		 */
3255		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3256			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3257		} else {
3258			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3259		}
3260	}
3261	/*
3262	 * Now go through and remove any single reference to our inp that
3263	 * may be still pending on the list
3264	 */
3265	SCTP_IPI_ITERATOR_WQ_LOCK();
3266	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3267		if (it->vn != curvnet) {
3268			continue;
3269		}
3270		if (it->inp == inp) {
3271			/* This one points to me is it inp specific? */
3272			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3273				/* Remove and free this one */
3274				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3275				    it, sctp_nxt_itr);
3276				if (it->function_atend != NULL) {
3277					(*it->function_atend) (it->pointer, it->val);
3278				}
3279				SCTP_FREE(it, SCTP_M_ITER);
3280			} else {
3281				it->inp = LIST_NEXT(it->inp, sctp_list);
3282				if (it->inp) {
3283					SCTP_INP_INCR_REF(it->inp);
3284				}
3285			}
3286			/*
3287			 * When its put in the refcnt is incremented so decr
3288			 * it
3289			 */
3290			SCTP_INP_DECR_REF(inp);
3291		}
3292	}
3293	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3294}
3295
3296/* release sctp_inpcb unbind the port */
3297void
3298sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3299{
3300	/*
3301	 * Here we free a endpoint. We must find it (if it is in the Hash
3302	 * table) and remove it from there. Then we must also find it in the
3303	 * overall list and remove it from there. After all removals are
3304	 * complete then any timer has to be stopped. Then start the actual
3305	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3306	 * all associations. d) finally the ep itself.
3307	 */
3308	struct sctp_tcb *asoc, *nasoc;
3309	struct sctp_laddr *laddr, *nladdr;
3310	struct inpcb *ip_pcb;
3311	struct socket *so;
3312	int being_refed = 0;
3313	struct sctp_queued_to_read *sq, *nsq;
3314	int cnt;
3315	sctp_sharedkey_t *shared_key, *nshared_key;
3316
3317
3318#ifdef SCTP_LOG_CLOSING
3319	sctp_log_closing(inp, NULL, 0);
3320#endif
3321	SCTP_ITERATOR_LOCK();
3322	/* mark any iterators on the list or being processed */
3323	sctp_iterator_inp_being_freed(inp);
3324	SCTP_ITERATOR_UNLOCK();
3325	so = inp->sctp_socket;
3326	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3327		/* been here before.. eeks.. get out of here */
3328		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3329#ifdef SCTP_LOG_CLOSING
3330		sctp_log_closing(inp, NULL, 1);
3331#endif
3332		return;
3333	}
3334	SCTP_ASOC_CREATE_LOCK(inp);
3335	SCTP_INP_INFO_WLOCK();
3336
3337	SCTP_INP_WLOCK(inp);
3338	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3339		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3340		/* socket is gone, so no more wakeups allowed */
3341		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3342		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3343		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3344
3345	}
3346	/* First time through we have the socket lock, after that no more. */
3347	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3348	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3349
3350	if (inp->control) {
3351		sctp_m_freem(inp->control);
3352		inp->control = NULL;
3353	}
3354	if (inp->pkt) {
3355		sctp_m_freem(inp->pkt);
3356		inp->pkt = NULL;
3357	}
3358	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3359					 * here but I will be nice :> (i.e.
3360					 * ip_pcb = ep;) */
3361	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3362		int cnt_in_sd;
3363
3364		cnt_in_sd = 0;
3365		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3366			SCTP_TCB_LOCK(asoc);
3367			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3368				/* Skip guys being freed */
3369				cnt_in_sd++;
3370				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3371					/*
3372					 * Special case - we did not start a
3373					 * kill timer on the asoc due to it
3374					 * was not closed. So go ahead and
3375					 * start it now.
3376					 */
3377					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3378					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3379				}
3380				SCTP_TCB_UNLOCK(asoc);
3381				continue;
3382			}
3383			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3384			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3385			    (asoc->asoc.total_output_queue_size == 0)) {
3386				/*
3387				 * If we have data in queue, we don't want
3388				 * to just free since the app may have done,
3389				 * send()/close or connect/send/close. And
3390				 * it wants the data to get across first.
3391				 */
3392				/* Just abandon things in the front states */
3393				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3394				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3395					cnt_in_sd++;
3396				}
3397				continue;
3398			}
3399			/* Disconnect the socket please */
3400			asoc->sctp_socket = NULL;
3401			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3402			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3403			    (asoc->asoc.control_pdapi) ||
3404			    (asoc->asoc.size_on_all_streams > 0) ||
3405			    (so && (so->so_rcv.sb_cc > 0))) {
3406				/* Left with Data unread */
3407				struct mbuf *op_err;
3408
3409				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3410				    0, M_DONTWAIT, 1, MT_DATA);
3411				if (op_err) {
3412					/* Fill in the user initiated abort */
3413					struct sctp_paramhdr *ph;
3414					uint32_t *ippp;
3415
3416					SCTP_BUF_LEN(op_err) =
3417					    sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
3418					ph = mtod(op_err,
3419					    struct sctp_paramhdr *);
3420					ph->param_type = htons(
3421					    SCTP_CAUSE_USER_INITIATED_ABT);
3422					ph->param_length = htons(SCTP_BUF_LEN(op_err));
3423					ippp = (uint32_t *) (ph + 1);
3424					*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3);
3425				}
3426				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3427				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3428				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3429				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3430				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3431					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3432				}
3433				if (sctp_free_assoc(inp, asoc,
3434				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3435					cnt_in_sd++;
3436				}
3437				continue;
3438			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3439				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3440				    (asoc->asoc.stream_queue_cnt == 0)
3441			    ) {
3442				if (asoc->asoc.locked_on_sending) {
3443					goto abort_anyway;
3444				}
3445				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3446				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3447					struct sctp_nets *netp;
3448
3449					if (asoc->asoc.alternate) {
3450						netp = asoc->asoc.alternate;
3451					} else {
3452						netp = asoc->asoc.primary_destination;
3453					}
3454					/*
3455					 * there is nothing queued to send,
3456					 * so I send shutdown
3457					 */
3458					sctp_send_shutdown(asoc, netp);
3459					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3460					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3461						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3462					}
3463					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3464					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3465					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3466					    netp);
3467					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3468					    asoc->asoc.primary_destination);
3469					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3470				}
3471			} else {
3472				/* mark into shutdown pending */
3473				struct sctp_stream_queue_pending *sp;
3474
3475				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3476				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3477				    asoc->asoc.primary_destination);
3478				if (asoc->asoc.locked_on_sending) {
3479					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3480					    sctp_streamhead);
3481					if (sp == NULL) {
3482						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3483						    asoc->asoc.locked_on_sending,
3484						    asoc->asoc.locked_on_sending->stream_no);
3485					} else {
3486						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3487							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3488					}
3489				}
3490				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3491				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3492				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3493					struct mbuf *op_err;
3494
3495			abort_anyway:
3496					op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3497					    0, M_DONTWAIT, 1, MT_DATA);
3498					if (op_err) {
3499						/*
3500						 * Fill in the user
3501						 * initiated abort
3502						 */
3503						struct sctp_paramhdr *ph;
3504						uint32_t *ippp;
3505
3506						SCTP_BUF_LEN(op_err) =
3507						    (sizeof(struct sctp_paramhdr) +
3508						    sizeof(uint32_t));
3509						ph = mtod(op_err,
3510						    struct sctp_paramhdr *);
3511						ph->param_type = htons(
3512						    SCTP_CAUSE_USER_INITIATED_ABT);
3513						ph->param_length = htons(SCTP_BUF_LEN(op_err));
3514						ippp = (uint32_t *) (ph + 1);
3515						*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5);
3516					}
3517					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3518					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3519					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3520					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3521					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3522						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3523					}
3524					if (sctp_free_assoc(inp, asoc,
3525					    SCTP_PCBFREE_NOFORCE,
3526					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3527						cnt_in_sd++;
3528					}
3529					continue;
3530				} else {
3531					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3532				}
3533			}
3534			cnt_in_sd++;
3535			SCTP_TCB_UNLOCK(asoc);
3536		}
3537		/* now is there some left in our SHUTDOWN state? */
3538		if (cnt_in_sd) {
3539#ifdef SCTP_LOG_CLOSING
3540			sctp_log_closing(inp, NULL, 2);
3541#endif
3542			inp->sctp_socket = NULL;
3543			SCTP_INP_WUNLOCK(inp);
3544			SCTP_ASOC_CREATE_UNLOCK(inp);
3545			SCTP_INP_INFO_WUNLOCK();
3546			return;
3547		}
3548	}
3549	inp->sctp_socket = NULL;
3550	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3551	    SCTP_PCB_FLAGS_UNBOUND) {
3552		/*
3553		 * ok, this guy has been bound. It's port is somewhere in
3554		 * the SCTP_BASE_INFO(hash table). Remove it!
3555		 */
3556		LIST_REMOVE(inp, sctp_hash);
3557		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3558	}
3559	/*
3560	 * If there is a timer running to kill us, forget it, since it may
3561	 * have a contest on the INP lock.. which would cause us to die ...
3562	 */
3563	cnt = 0;
3564	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3565		SCTP_TCB_LOCK(asoc);
3566		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3567			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3568				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3569				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3570			}
3571			cnt++;
3572			SCTP_TCB_UNLOCK(asoc);
3573			continue;
3574		}
3575		/* Free associations that are NOT killing us */
3576		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3577		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3578			struct mbuf *op_err;
3579			uint32_t *ippp;
3580
3581			op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3582			    0, M_DONTWAIT, 1, MT_DATA);
3583			if (op_err) {
3584				/* Fill in the user initiated abort */
3585				struct sctp_paramhdr *ph;
3586
3587				SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) +
3588				    sizeof(uint32_t));
3589				ph = mtod(op_err, struct sctp_paramhdr *);
3590				ph->param_type = htons(
3591				    SCTP_CAUSE_USER_INITIATED_ABT);
3592				ph->param_length = htons(SCTP_BUF_LEN(op_err));
3593				ippp = (uint32_t *) (ph + 1);
3594				*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7);
3595
3596			}
3597			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3598			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3599			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3600		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3601			cnt++;
3602			SCTP_TCB_UNLOCK(asoc);
3603			continue;
3604		}
3605		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3606		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3607			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3608		}
3609		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3610			cnt++;
3611		}
3612	}
3613	if (cnt) {
3614		/* Ok we have someone out there that will kill us */
3615		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3616#ifdef SCTP_LOG_CLOSING
3617		sctp_log_closing(inp, NULL, 3);
3618#endif
3619		SCTP_INP_WUNLOCK(inp);
3620		SCTP_ASOC_CREATE_UNLOCK(inp);
3621		SCTP_INP_INFO_WUNLOCK();
3622		return;
3623	}
3624	if (SCTP_INP_LOCK_CONTENDED(inp))
3625		being_refed++;
3626	if (SCTP_INP_READ_CONTENDED(inp))
3627		being_refed++;
3628	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3629		being_refed++;
3630
3631	if ((inp->refcount) ||
3632	    (being_refed) ||
3633	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3634		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3635#ifdef SCTP_LOG_CLOSING
3636		sctp_log_closing(inp, NULL, 4);
3637#endif
3638		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3639		SCTP_INP_WUNLOCK(inp);
3640		SCTP_ASOC_CREATE_UNLOCK(inp);
3641		SCTP_INP_INFO_WUNLOCK();
3642		return;
3643	}
3644	inp->sctp_ep.signature_change.type = 0;
3645	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3646	/*
3647	 * Remove it from the list .. last thing we need a lock for.
3648	 */
3649	LIST_REMOVE(inp, sctp_list);
3650	SCTP_INP_WUNLOCK(inp);
3651	SCTP_ASOC_CREATE_UNLOCK(inp);
3652	SCTP_INP_INFO_WUNLOCK();
3653	/*
3654	 * Now we release all locks. Since this INP cannot be found anymore
3655	 * except possibly by the kill timer that might be running. We call
3656	 * the drain function here. It should hit the case were it sees the
3657	 * ACTIVE flag cleared and exit out freeing us to proceed and
3658	 * destroy everything.
3659	 */
3660	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3661		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3662	} else {
3663		/* Probably un-needed */
3664		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3665	}
3666
3667#ifdef SCTP_LOG_CLOSING
3668	sctp_log_closing(inp, NULL, 5);
3669#endif
3670
3671
3672	if ((inp->sctp_asocidhash) != NULL) {
3673		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3674		inp->sctp_asocidhash = NULL;
3675	}
3676	/* sa_ignore FREED_MEMORY */
3677	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3678		/* Its only abandoned if it had data left */
3679		if (sq->length)
3680			SCTP_STAT_INCR(sctps_left_abandon);
3681
3682		TAILQ_REMOVE(&inp->read_queue, sq, next);
3683		sctp_free_remote_addr(sq->whoFrom);
3684		if (so)
3685			so->so_rcv.sb_cc -= sq->length;
3686		if (sq->data) {
3687			sctp_m_freem(sq->data);
3688			sq->data = NULL;
3689		}
3690		/*
3691		 * no need to free the net count, since at this point all
3692		 * assoc's are gone.
3693		 */
3694		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3695		SCTP_DECR_READQ_COUNT();
3696	}
3697	/* Now the sctp_pcb things */
3698	/*
3699	 * free each asoc if it is not already closed/free. we can't use the
3700	 * macro here since le_next will get freed as part of the
3701	 * sctp_free_assoc() call.
3702	 */
3703	if (so) {
3704#ifdef IPSEC
3705		ipsec_delete_pcbpolicy(ip_pcb);
3706#endif				/* IPSEC */
3707
3708		/* Unlocks not needed since the socket is gone now */
3709	}
3710	if (ip_pcb->inp_options) {
3711		(void)sctp_m_free(ip_pcb->inp_options);
3712		ip_pcb->inp_options = 0;
3713	}
3714#ifdef INET6
3715	if (ip_pcb->inp_vflag & INP_IPV6) {
3716		struct in6pcb *in6p;
3717
3718		in6p = (struct in6pcb *)inp;
3719		ip6_freepcbopts(in6p->in6p_outputopts);
3720	}
3721#endif				/* INET6 */
3722	ip_pcb->inp_vflag = 0;
3723	/* free up authentication fields */
3724	if (inp->sctp_ep.local_auth_chunks != NULL)
3725		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3726	if (inp->sctp_ep.local_hmacs != NULL)
3727		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3728
3729	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3730		LIST_REMOVE(shared_key, next);
3731		sctp_free_sharedkey(shared_key);
3732		/* sa_ignore FREED_MEMORY */
3733	}
3734
3735	/*
3736	 * if we have an address list the following will free the list of
3737	 * ifaddr's that are set into this ep. Again macro limitations here,
3738	 * since the LIST_FOREACH could be a bad idea.
3739	 */
3740	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3741		sctp_remove_laddr(laddr);
3742	}
3743
3744#ifdef SCTP_TRACK_FREED_ASOCS
3745	/* TEMP CODE */
3746	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3747		LIST_REMOVE(asoc, sctp_tcblist);
3748		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3749		SCTP_DECR_ASOC_COUNT();
3750	}
3751	/* *** END TEMP CODE *** */
3752#endif
3753	/* Now lets see about freeing the EP hash table. */
3754	if (inp->sctp_tcbhash != NULL) {
3755		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3756		inp->sctp_tcbhash = NULL;
3757	}
3758	/* Now we must put the ep memory back into the zone pool */
3759	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3760	SCTP_INP_LOCK_DESTROY(inp);
3761	SCTP_INP_READ_DESTROY(inp);
3762	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3763	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3764	SCTP_DECR_EP_COUNT();
3765}
3766
3767
3768struct sctp_nets *
3769sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3770{
3771	struct sctp_nets *net;
3772
3773	/* locate the address */
3774	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3775		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3776			return (net);
3777	}
3778	return (NULL);
3779}
3780
3781
3782int
3783sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3784{
3785	struct sctp_ifa *sctp_ifa;
3786
3787	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3788	if (sctp_ifa) {
3789		return (1);
3790	} else {
3791		return (0);
3792	}
3793}
3794
3795/*
3796 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3797 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3798 * stats of stuff.
3799 */
3800int
3801sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3802    struct sctp_nets **netp, int set_scope, int from)
3803{
3804	/*
3805	 * The following is redundant to the same lines in the
3806	 * sctp_aloc_assoc() but is needed since others call the add address
3807	 * function
3808	 */
3809	struct sctp_nets *net, *netfirst;
3810	int addr_inscope;
3811
3812	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3813	    from);
3814	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3815
3816	netfirst = sctp_findnet(stcb, newaddr);
3817	if (netfirst) {
3818		/*
3819		 * Lie and return ok, we don't want to make the association
3820		 * go away for this behavior. It will happen in the TCP
3821		 * model in a connected socket. It does not reach the hash
3822		 * table until after the association is built so it can't be
3823		 * found. Mark as reachable, since the initial creation will
3824		 * have been cleared and the NOT_IN_ASSOC flag will have
3825		 * been added... and we don't want to end up removing it
3826		 * back out.
3827		 */
3828		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3829			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3830			    SCTP_ADDR_UNCONFIRMED);
3831		} else {
3832			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3833		}
3834
3835		return (0);
3836	}
3837	addr_inscope = 1;
3838	switch (newaddr->sa_family) {
3839#ifdef INET
3840	case AF_INET:
3841		{
3842			struct sockaddr_in *sin;
3843
3844			sin = (struct sockaddr_in *)newaddr;
3845			if (sin->sin_addr.s_addr == 0) {
3846				/* Invalid address */
3847				return (-1);
3848			}
3849			/* zero out the bzero area */
3850			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3851
3852			/* assure len is set */
3853			sin->sin_len = sizeof(struct sockaddr_in);
3854			if (set_scope) {
3855#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3856				stcb->ipv4_local_scope = 1;
3857#else
3858				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3859					stcb->asoc.ipv4_local_scope = 1;
3860				}
3861#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3862			} else {
3863				/* Validate the address is in scope */
3864				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3865				    (stcb->asoc.ipv4_local_scope == 0)) {
3866					addr_inscope = 0;
3867				}
3868			}
3869			break;
3870		}
3871#endif
3872#ifdef INET6
3873	case AF_INET6:
3874		{
3875			struct sockaddr_in6 *sin6;
3876
3877			sin6 = (struct sockaddr_in6 *)newaddr;
3878			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3879				/* Invalid address */
3880				return (-1);
3881			}
3882			/* assure len is set */
3883			sin6->sin6_len = sizeof(struct sockaddr_in6);
3884			if (set_scope) {
3885				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3886					stcb->asoc.loopback_scope = 1;
3887					stcb->asoc.local_scope = 0;
3888					stcb->asoc.ipv4_local_scope = 1;
3889					stcb->asoc.site_scope = 1;
3890				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3891					/*
3892					 * If the new destination is a
3893					 * LINK_LOCAL we must have common
3894					 * site scope. Don't set the local
3895					 * scope since we may not share all
3896					 * links, only loopback can do this.
3897					 * Links on the local network would
3898					 * also be on our private network
3899					 * for v4 too.
3900					 */
3901					stcb->asoc.ipv4_local_scope = 1;
3902					stcb->asoc.site_scope = 1;
3903				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3904					/*
3905					 * If the new destination is
3906					 * SITE_LOCAL then we must have site
3907					 * scope in common.
3908					 */
3909					stcb->asoc.site_scope = 1;
3910				}
3911			} else {
3912				/* Validate the address is in scope */
3913				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3914				    (stcb->asoc.loopback_scope == 0)) {
3915					addr_inscope = 0;
3916				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3917				    (stcb->asoc.local_scope == 0)) {
3918					addr_inscope = 0;
3919				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3920				    (stcb->asoc.site_scope == 0)) {
3921					addr_inscope = 0;
3922				}
3923			}
3924			break;
3925		}
3926#endif
3927	default:
3928		/* not supported family type */
3929		return (-1);
3930	}
3931	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3932	if (net == NULL) {
3933		return (-1);
3934	}
3935	SCTP_INCR_RADDR_COUNT();
3936	bzero(net, sizeof(struct sctp_nets));
3937	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3938	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3939	switch (newaddr->sa_family) {
3940#ifdef INET
3941	case AF_INET:
3942		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3943		break;
3944#endif
3945#ifdef INET6
3946	case AF_INET6:
3947		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3948		break;
3949#endif
3950	default:
3951		break;
3952	}
3953	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3954	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3955		stcb->asoc.loopback_scope = 1;
3956		stcb->asoc.ipv4_local_scope = 1;
3957		stcb->asoc.local_scope = 0;
3958		stcb->asoc.site_scope = 1;
3959		addr_inscope = 1;
3960	}
3961	net->failure_threshold = stcb->asoc.def_net_failure;
3962	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3963	if (addr_inscope == 0) {
3964		net->dest_state = (SCTP_ADDR_REACHABLE |
3965		    SCTP_ADDR_OUT_OF_SCOPE);
3966	} else {
3967		if (from == SCTP_ADDR_IS_CONFIRMED)
3968			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3969			net->dest_state = SCTP_ADDR_REACHABLE;
3970		else
3971			net->dest_state = SCTP_ADDR_REACHABLE |
3972			    SCTP_ADDR_UNCONFIRMED;
3973	}
3974	/*
3975	 * We set this to 0, the timer code knows that this means its an
3976	 * initial value
3977	 */
3978	net->rto_needed = 1;
3979	net->RTO = 0;
3980	net->RTO_measured = 0;
3981	stcb->asoc.numnets++;
3982	net->ref_count = 1;
3983	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3984	net->port = stcb->asoc.port;
3985	net->dscp = stcb->asoc.default_dscp;
3986#ifdef INET6
3987	net->flowlabel = stcb->asoc.default_flowlabel;
3988#endif
3989	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3990		net->dest_state |= SCTP_ADDR_NOHB;
3991	} else {
3992		net->dest_state &= ~SCTP_ADDR_NOHB;
3993	}
3994	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3995		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3996	} else {
3997		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3998	}
3999	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
4000	/* Init the timer structure */
4001	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
4002	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
4003	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
4004
4005	/* Now generate a route for this guy */
4006#ifdef INET6
4007	/* KAME hack: embed scopeid */
4008	if (newaddr->sa_family == AF_INET6) {
4009		struct sockaddr_in6 *sin6;
4010
4011		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4012		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
4013		sin6->sin6_scope_id = 0;
4014	}
4015#endif
4016	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
4017
4018	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
4019		/* Get source address */
4020		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
4021		    stcb,
4022		    (sctp_route_t *) & net->ro,
4023		    net,
4024		    0,
4025		    stcb->asoc.vrf_id);
4026		/* Now get the interface MTU */
4027		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
4028			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
4029		}
4030		if (net->mtu > 0) {
4031			uint32_t rmtu;
4032
4033			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
4034			if (rmtu == 0) {
4035				/*
4036				 * Start things off to match mtu of
4037				 * interface please.
4038				 */
4039				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
4040				    net->ro.ro_rt, net->mtu);
4041			} else {
4042				/*
4043				 * we take the route mtu over the interface,
4044				 * since the route may be leading out the
4045				 * loopback, or a different interface.
4046				 */
4047				net->mtu = rmtu;
4048			}
4049		}
4050	}
4051	if (net->mtu == 0) {
4052		switch (newaddr->sa_family) {
4053#ifdef INET
4054		case AF_INET:
4055			net->mtu = SCTP_DEFAULT_MTU;
4056			break;
4057#endif
4058#ifdef INET6
4059		case AF_INET6:
4060			net->mtu = 1280;
4061			break;
4062#endif
4063		default:
4064			break;
4065		}
4066	}
4067	if (net->port) {
4068		net->mtu -= (uint32_t) sizeof(struct udphdr);
4069	}
4070	if (from == SCTP_ALLOC_ASOC) {
4071		stcb->asoc.smallest_mtu = net->mtu;
4072	}
4073	if (stcb->asoc.smallest_mtu > net->mtu) {
4074		stcb->asoc.smallest_mtu = net->mtu;
4075	}
4076#ifdef INET6
4077	if (newaddr->sa_family == AF_INET6) {
4078		struct sockaddr_in6 *sin6;
4079
4080		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4081		(void)sa6_recoverscope(sin6);
4082	}
4083#endif
4084
4085	/* JRS - Use the congestion control given in the CC module */
4086	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4087		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4088
4089	/*
4090	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4091	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4092	 */
4093	net->find_pseudo_cumack = 1;
4094	net->find_rtx_pseudo_cumack = 1;
4095	net->src_addr_selected = 0;
4096	/* Choose an initial flowid. */
4097	net->flowid = stcb->asoc.my_vtag ^
4098	    ntohs(stcb->rport) ^
4099	    ntohs(stcb->sctp_ep->sctp_lport);
4100#ifdef INVARIANTS
4101	net->flowidset = 1;
4102#endif
4103	if (netp) {
4104		*netp = net;
4105	}
4106	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4107	if (net->ro.ro_rt == NULL) {
4108		/* Since we have no route put it at the back */
4109		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4110	} else if (netfirst == NULL) {
4111		/* We are the first one in the pool. */
4112		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4113	} else if (netfirst->ro.ro_rt == NULL) {
4114		/*
4115		 * First one has NO route. Place this one ahead of the first
4116		 * one.
4117		 */
4118		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4119	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4120		/*
4121		 * This one has a different interface than the one at the
4122		 * top of the list. Place it ahead.
4123		 */
4124		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4125	} else {
4126		/*
4127		 * Ok we have the same interface as the first one. Move
4128		 * forward until we find either a) one with a NULL route...
4129		 * insert ahead of that b) one with a different ifp.. insert
4130		 * after that. c) end of the list.. insert at the tail.
4131		 */
4132		struct sctp_nets *netlook;
4133
4134		do {
4135			netlook = TAILQ_NEXT(netfirst, sctp_next);
4136			if (netlook == NULL) {
4137				/* End of the list */
4138				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4139				break;
4140			} else if (netlook->ro.ro_rt == NULL) {
4141				/* next one has NO route */
4142				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4143				break;
4144			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4145				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4146				    net, sctp_next);
4147				break;
4148			}
4149			/* Shift forward */
4150			netfirst = netlook;
4151		} while (netlook != NULL);
4152	}
4153
4154	/* got to have a primary set */
4155	if (stcb->asoc.primary_destination == 0) {
4156		stcb->asoc.primary_destination = net;
4157	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4158		    (net->ro.ro_rt) &&
4159	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4160		/* No route to current primary adopt new primary */
4161		stcb->asoc.primary_destination = net;
4162	}
4163	/* Validate primary is first */
4164	net = TAILQ_FIRST(&stcb->asoc.nets);
4165	if ((net != stcb->asoc.primary_destination) &&
4166	    (stcb->asoc.primary_destination)) {
4167		/*
4168		 * first one on the list is NOT the primary sctp_cmpaddr()
4169		 * is much more efficient if the primary is the first on the
4170		 * list, make it so.
4171		 */
4172		TAILQ_REMOVE(&stcb->asoc.nets,
4173		    stcb->asoc.primary_destination, sctp_next);
4174		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4175		    stcb->asoc.primary_destination, sctp_next);
4176	}
4177	return (0);
4178}
4179
4180
4181static uint32_t
4182sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4183{
4184	uint32_t id;
4185	struct sctpasochead *head;
4186	struct sctp_tcb *lstcb;
4187
4188	SCTP_INP_WLOCK(inp);
4189try_again:
4190	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4191		/* TSNH */
4192		SCTP_INP_WUNLOCK(inp);
4193		return (0);
4194	}
4195	/*
4196	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4197	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4198	 */
4199	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4200		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4201	}
4202	id = inp->sctp_associd_counter;
4203	inp->sctp_associd_counter++;
4204	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4205	if (lstcb) {
4206		goto try_again;
4207	}
4208	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4209	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4210	stcb->asoc.in_asocid_hash = 1;
4211	SCTP_INP_WUNLOCK(inp);
4212	return id;
4213}
4214
4215/*
4216 * allocate an association and add it to the endpoint. The caller must be
4217 * careful to add all additional addresses once they are know right away or
4218 * else the assoc will be may experience a blackout scenario.
4219 */
4220struct sctp_tcb *
4221sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4222    int *error, uint32_t override_tag, uint32_t vrf_id,
4223    struct thread *p
4224)
4225{
4226	/* note the p argument is only valid in unbound sockets */
4227
4228	struct sctp_tcb *stcb;
4229	struct sctp_association *asoc;
4230	struct sctpasochead *head;
4231	uint16_t rport;
4232	int err;
4233
4234	/*
4235	 * Assumption made here: Caller has done a
4236	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4237	 * address does not exist already.
4238	 */
4239	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4240		/* Hit max assoc, sorry no more */
4241		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4242		*error = ENOBUFS;
4243		return (NULL);
4244	}
4245	if (firstaddr == NULL) {
4246		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4247		*error = EINVAL;
4248		return (NULL);
4249	}
4250	SCTP_INP_RLOCK(inp);
4251	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4252	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4253	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4254		/*
4255		 * If its in the TCP pool, its NOT allowed to create an
4256		 * association. The parent listener needs to call
4257		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4258		 * off, or connected one does this.. its an error.
4259		 */
4260		SCTP_INP_RUNLOCK(inp);
4261		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4262		*error = EINVAL;
4263		return (NULL);
4264	}
4265	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4266	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4267		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4268		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4269			SCTP_INP_RUNLOCK(inp);
4270			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4271			*error = EINVAL;
4272			return (NULL);
4273		}
4274	}
4275	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4276#ifdef SCTP_DEBUG
4277	if (firstaddr) {
4278		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4279		switch (firstaddr->sa_family) {
4280#ifdef INET
4281		case AF_INET:
4282			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4283			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4284			break;
4285#endif
4286#ifdef INET6
4287		case AF_INET6:
4288			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4289			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4290			break;
4291#endif
4292		default:
4293			break;
4294		}
4295	} else {
4296		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4297	}
4298#endif				/* SCTP_DEBUG */
4299	switch (firstaddr->sa_family) {
4300#ifdef INET
4301	case AF_INET:
4302		{
4303			struct sockaddr_in *sin;
4304
4305			sin = (struct sockaddr_in *)firstaddr;
4306			if ((ntohs(sin->sin_port) == 0) ||
4307			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4308			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4309			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4310				/* Invalid address */
4311				SCTP_INP_RUNLOCK(inp);
4312				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4313				*error = EINVAL;
4314				return (NULL);
4315			}
4316			rport = sin->sin_port;
4317			break;
4318		}
4319#endif
4320#ifdef INET6
4321	case AF_INET6:
4322		{
4323			struct sockaddr_in6 *sin6;
4324
4325			sin6 = (struct sockaddr_in6 *)firstaddr;
4326			if ((ntohs(sin6->sin6_port) == 0) ||
4327			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4328			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4329				/* Invalid address */
4330				SCTP_INP_RUNLOCK(inp);
4331				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4332				*error = EINVAL;
4333				return (NULL);
4334			}
4335			rport = sin6->sin6_port;
4336			break;
4337		}
4338#endif
4339	default:
4340		/* not supported family type */
4341		SCTP_INP_RUNLOCK(inp);
4342		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4343		*error = EINVAL;
4344		return (NULL);
4345	}
4346	SCTP_INP_RUNLOCK(inp);
4347	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4348		/*
4349		 * If you have not performed a bind, then we need to do the
4350		 * ephemeral bind for you.
4351		 */
4352		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4353		    (struct sockaddr *)NULL,
4354		    (struct sctp_ifa *)NULL,
4355		    p
4356		    ))) {
4357			/* bind error, probably perm */
4358			*error = err;
4359			return (NULL);
4360		}
4361	}
4362	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4363	if (stcb == NULL) {
4364		/* out of memory? */
4365		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4366		*error = ENOMEM;
4367		return (NULL);
4368	}
4369	SCTP_INCR_ASOC_COUNT();
4370
4371	bzero(stcb, sizeof(*stcb));
4372	asoc = &stcb->asoc;
4373
4374	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4375	SCTP_TCB_LOCK_INIT(stcb);
4376	SCTP_TCB_SEND_LOCK_INIT(stcb);
4377	stcb->rport = rport;
4378	/* setup back pointer's */
4379	stcb->sctp_ep = inp;
4380	stcb->sctp_socket = inp->sctp_socket;
4381	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4382		/* failed */
4383		SCTP_TCB_LOCK_DESTROY(stcb);
4384		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4385		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4386		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4387		SCTP_DECR_ASOC_COUNT();
4388		*error = err;
4389		return (NULL);
4390	}
4391	/* and the port */
4392	SCTP_INP_INFO_WLOCK();
4393	SCTP_INP_WLOCK(inp);
4394	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4395		/* inpcb freed while alloc going on */
4396		SCTP_TCB_LOCK_DESTROY(stcb);
4397		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4398		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4399		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4400		SCTP_INP_WUNLOCK(inp);
4401		SCTP_INP_INFO_WUNLOCK();
4402		SCTP_DECR_ASOC_COUNT();
4403		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4404		*error = EINVAL;
4405		return (NULL);
4406	}
4407	SCTP_TCB_LOCK(stcb);
4408
4409	/* now that my_vtag is set, add it to the hash */
4410	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4411	/* put it in the bucket in the vtag hash of assoc's for the system */
4412	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4413	SCTP_INP_INFO_WUNLOCK();
4414
4415	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4416		/* failure.. memory error? */
4417		if (asoc->strmout) {
4418			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4419			asoc->strmout = NULL;
4420		}
4421		if (asoc->mapping_array) {
4422			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4423			asoc->mapping_array = NULL;
4424		}
4425		if (asoc->nr_mapping_array) {
4426			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4427			asoc->nr_mapping_array = NULL;
4428		}
4429		SCTP_DECR_ASOC_COUNT();
4430		SCTP_TCB_LOCK_DESTROY(stcb);
4431		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4432		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4433		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4434		SCTP_INP_WUNLOCK(inp);
4435		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4436		*error = ENOBUFS;
4437		return (NULL);
4438	}
4439	/* Init all the timers */
4440	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4441	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4442	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4443	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4444	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4445	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4446	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4447
4448	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4449	/* now file the port under the hash as well */
4450	if (inp->sctp_tcbhash != NULL) {
4451		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4452		    inp->sctp_hashmark)];
4453		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4454	}
4455	SCTP_INP_WUNLOCK(inp);
4456	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", stcb);
4457	return (stcb);
4458}
4459
4460
4461void
4462sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4463{
4464	struct sctp_association *asoc;
4465
4466	asoc = &stcb->asoc;
4467	asoc->numnets--;
4468	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4469	if (net == asoc->primary_destination) {
4470		/* Reset primary */
4471		struct sctp_nets *lnet;
4472
4473		lnet = TAILQ_FIRST(&asoc->nets);
4474		/*
4475		 * Mobility adaptation Ideally, if deleted destination is
4476		 * the primary, it becomes a fast retransmission trigger by
4477		 * the subsequent SET PRIMARY. (by micchie)
4478		 */
4479		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4480		    SCTP_MOBILITY_BASE) ||
4481		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4482		    SCTP_MOBILITY_FASTHANDOFF)) {
4483			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4484			if (asoc->deleted_primary != NULL) {
4485				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4486				goto out;
4487			}
4488			asoc->deleted_primary = net;
4489			atomic_add_int(&net->ref_count, 1);
4490			memset(&net->lastsa, 0, sizeof(net->lastsa));
4491			memset(&net->lastsv, 0, sizeof(net->lastsv));
4492			sctp_mobility_feature_on(stcb->sctp_ep,
4493			    SCTP_MOBILITY_PRIM_DELETED);
4494			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4495			    stcb->sctp_ep, stcb, NULL);
4496		}
4497out:
4498		/* Try to find a confirmed primary */
4499		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4500	}
4501	if (net == asoc->last_data_chunk_from) {
4502		/* Reset primary */
4503		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4504	}
4505	if (net == asoc->last_control_chunk_from) {
4506		/* Clear net */
4507		asoc->last_control_chunk_from = NULL;
4508	}
4509	if (net == stcb->asoc.alternate) {
4510		sctp_free_remote_addr(stcb->asoc.alternate);
4511		stcb->asoc.alternate = NULL;
4512	}
4513	sctp_free_remote_addr(net);
4514}
4515
4516/*
4517 * remove a remote endpoint address from an association, it will fail if the
4518 * address does not exist.
4519 */
4520int
4521sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4522{
4523	/*
4524	 * Here we need to remove a remote address. This is quite simple, we
4525	 * first find it in the list of address for the association
4526	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4527	 * on that item. Note we do not allow it to be removed if there are
4528	 * no other addresses.
4529	 */
4530	struct sctp_association *asoc;
4531	struct sctp_nets *net, *nnet;
4532
4533	asoc = &stcb->asoc;
4534
4535	/* locate the address */
4536	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4537		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4538			continue;
4539		}
4540		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4541		    remaddr)) {
4542			/* we found the guy */
4543			if (asoc->numnets < 2) {
4544				/* Must have at LEAST two remote addresses */
4545				return (-1);
4546			} else {
4547				sctp_remove_net(stcb, net);
4548				return (0);
4549			}
4550		}
4551	}
4552	/* not found. */
4553	return (-2);
4554}
4555
4556void
4557sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4558{
4559	struct sctpvtaghead *chain;
4560	struct sctp_tagblock *twait_block;
4561	int found = 0;
4562	int i;
4563
4564	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4565	if (!LIST_EMPTY(chain)) {
4566		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4567			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4568				if ((twait_block->vtag_block[i].v_tag == tag) &&
4569				    (twait_block->vtag_block[i].lport == lport) &&
4570				    (twait_block->vtag_block[i].rport == rport)) {
4571					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4572					twait_block->vtag_block[i].v_tag = 0;
4573					twait_block->vtag_block[i].lport = 0;
4574					twait_block->vtag_block[i].rport = 0;
4575					found = 1;
4576					break;
4577				}
4578			}
4579			if (found)
4580				break;
4581		}
4582	}
4583}
4584
4585int
4586sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4587{
4588	struct sctpvtaghead *chain;
4589	struct sctp_tagblock *twait_block;
4590	int found = 0;
4591	int i;
4592
4593	SCTP_INP_INFO_WLOCK();
4594	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4595	if (!LIST_EMPTY(chain)) {
4596		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4597			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4598				if ((twait_block->vtag_block[i].v_tag == tag) &&
4599				    (twait_block->vtag_block[i].lport == lport) &&
4600				    (twait_block->vtag_block[i].rport == rport)) {
4601					found = 1;
4602					break;
4603				}
4604			}
4605			if (found)
4606				break;
4607		}
4608	}
4609	SCTP_INP_INFO_WUNLOCK();
4610	return (found);
4611}
4612
4613
4614void
4615sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4616{
4617	struct sctpvtaghead *chain;
4618	struct sctp_tagblock *twait_block;
4619	struct timeval now;
4620	int set, i;
4621
4622	if (time == 0) {
4623		/* Its disabled */
4624		return;
4625	}
4626	(void)SCTP_GETTIME_TIMEVAL(&now);
4627	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4628	set = 0;
4629	if (!LIST_EMPTY(chain)) {
4630		/* Block(s) present, lets find space, and expire on the fly */
4631		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4632			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4633				if ((twait_block->vtag_block[i].v_tag == 0) &&
4634				    !set) {
4635					twait_block->vtag_block[i].tv_sec_at_expire =
4636					    now.tv_sec + time;
4637					twait_block->vtag_block[i].v_tag = tag;
4638					twait_block->vtag_block[i].lport = lport;
4639					twait_block->vtag_block[i].rport = rport;
4640					set = 1;
4641				} else if ((twait_block->vtag_block[i].v_tag) &&
4642				    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4643					/* Audit expires this guy */
4644					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4645					twait_block->vtag_block[i].v_tag = 0;
4646					twait_block->vtag_block[i].lport = 0;
4647					twait_block->vtag_block[i].rport = 0;
4648					if (set == 0) {
4649						/* Reuse it for my new tag */
4650						twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4651						twait_block->vtag_block[i].v_tag = tag;
4652						twait_block->vtag_block[i].lport = lport;
4653						twait_block->vtag_block[i].rport = rport;
4654						set = 1;
4655					}
4656				}
4657			}
4658			if (set) {
4659				/*
4660				 * We only do up to the block where we can
4661				 * place our tag for audits
4662				 */
4663				break;
4664			}
4665		}
4666	}
4667	/* Need to add a new block to chain */
4668	if (!set) {
4669		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4670		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4671		if (twait_block == NULL) {
4672#ifdef INVARIANTS
4673			panic("Can not alloc tagblock");
4674#endif
4675			return;
4676		}
4677		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4678		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4679		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4680		twait_block->vtag_block[0].v_tag = tag;
4681		twait_block->vtag_block[0].lport = lport;
4682		twait_block->vtag_block[0].rport = rport;
4683	}
4684}
4685
4686
4687
4688/*-
4689 * Free the association after un-hashing the remote port. This
4690 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4691 * expect that the input to this function IS a locked TCB.
4692 * It will return 0, if it did NOT destroy the association (instead
4693 * it unlocks it. It will return NON-zero if it either destroyed the
4694 * association OR the association is already destroyed.
4695 */
4696int
4697sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4698{
4699	int i;
4700	struct sctp_association *asoc;
4701	struct sctp_nets *net, *nnet;
4702	struct sctp_laddr *laddr, *naddr;
4703	struct sctp_tmit_chunk *chk, *nchk;
4704	struct sctp_asconf_addr *aparam, *naparam;
4705	struct sctp_asconf_ack *aack, *naack;
4706	struct sctp_stream_reset_list *strrst, *nstrrst;
4707	struct sctp_queued_to_read *sq, *nsq;
4708	struct sctp_stream_queue_pending *sp, *nsp;
4709	sctp_sharedkey_t *shared_key, *nshared_key;
4710	struct socket *so;
4711
4712	/* first, lets purge the entry from the hash table. */
4713
4714#ifdef SCTP_LOG_CLOSING
4715	sctp_log_closing(inp, stcb, 6);
4716#endif
4717	if (stcb->asoc.state == 0) {
4718#ifdef SCTP_LOG_CLOSING
4719		sctp_log_closing(inp, NULL, 7);
4720#endif
4721		/* there is no asoc, really TSNH :-0 */
4722		return (1);
4723	}
4724	if (stcb->asoc.alternate) {
4725		sctp_free_remote_addr(stcb->asoc.alternate);
4726		stcb->asoc.alternate = NULL;
4727	}
4728	/* TEMP CODE */
4729	if (stcb->freed_from_where == 0) {
4730		/* Only record the first place free happened from */
4731		stcb->freed_from_where = from_location;
4732	}
4733	/* TEMP CODE */
4734
4735	asoc = &stcb->asoc;
4736	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4737	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4738		/* nothing around */
4739		so = NULL;
4740	else
4741		so = inp->sctp_socket;
4742
4743	/*
4744	 * We used timer based freeing if a reader or writer is in the way.
4745	 * So we first check if we are actually being called from a timer,
4746	 * if so we abort early if a reader or writer is still in the way.
4747	 */
4748	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4749	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4750		/*
4751		 * is it the timer driving us? if so are the reader/writers
4752		 * gone?
4753		 */
4754		if (stcb->asoc.refcnt) {
4755			/* nope, reader or writer in the way */
4756			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4757			/* no asoc destroyed */
4758			SCTP_TCB_UNLOCK(stcb);
4759#ifdef SCTP_LOG_CLOSING
4760			sctp_log_closing(inp, stcb, 8);
4761#endif
4762			return (0);
4763		}
4764	}
4765	/* now clean up any other timers */
4766	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4767	asoc->dack_timer.self = NULL;
4768	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4769	/*-
4770	 * For stream reset we don't blast this unless
4771	 * it is a str-reset timer, it might be the
4772	 * free-asoc timer which we DON'T want to
4773	 * disturb.
4774	 */
4775	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4776		asoc->strreset_timer.self = NULL;
4777	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4778	asoc->asconf_timer.self = NULL;
4779	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4780	asoc->autoclose_timer.self = NULL;
4781	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4782	asoc->shut_guard_timer.self = NULL;
4783	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4784	asoc->delayed_event_timer.self = NULL;
4785	/* Mobility adaptation */
4786	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4787	asoc->delete_prim_timer.self = NULL;
4788	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4789		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4790		net->rxt_timer.self = NULL;
4791		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4792		net->pmtu_timer.self = NULL;
4793		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4794		net->hb_timer.self = NULL;
4795	}
4796	/* Now the read queue needs to be cleaned up (only once) */
4797	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4798		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4799		SCTP_INP_READ_LOCK(inp);
4800		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4801			if (sq->stcb == stcb) {
4802				sq->do_not_ref_stcb = 1;
4803				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4804				/*
4805				 * If there is no end, there never will be
4806				 * now.
4807				 */
4808				if (sq->end_added == 0) {
4809					/* Held for PD-API clear that. */
4810					sq->pdapi_aborted = 1;
4811					sq->held_length = 0;
4812					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4813						/*
4814						 * Need to add a PD-API
4815						 * aborted indication.
4816						 * Setting the control_pdapi
4817						 * assures that it will be
4818						 * added right after this
4819						 * msg.
4820						 */
4821						uint32_t strseq;
4822
4823						stcb->asoc.control_pdapi = sq;
4824						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4825						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4826						    stcb,
4827						    SCTP_PARTIAL_DELIVERY_ABORTED,
4828						    (void *)&strseq,
4829						    SCTP_SO_LOCKED);
4830						stcb->asoc.control_pdapi = NULL;
4831					}
4832				}
4833				/* Add an end to wake them */
4834				sq->end_added = 1;
4835			}
4836		}
4837		SCTP_INP_READ_UNLOCK(inp);
4838		if (stcb->block_entry) {
4839			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4840			stcb->block_entry->error = ECONNRESET;
4841			stcb->block_entry = NULL;
4842		}
4843	}
4844	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4845		/*
4846		 * Someone holds a reference OR the socket is unaccepted
4847		 * yet.
4848		 */
4849		if ((stcb->asoc.refcnt) ||
4850		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4851		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4852			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4853			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4854		}
4855		SCTP_TCB_UNLOCK(stcb);
4856		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4857		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4858			/* nothing around */
4859			so = NULL;
4860		if (so) {
4861			/* Wake any reader/writers */
4862			sctp_sorwakeup(inp, so);
4863			sctp_sowwakeup(inp, so);
4864		}
4865#ifdef SCTP_LOG_CLOSING
4866		sctp_log_closing(inp, stcb, 9);
4867#endif
4868		/* no asoc destroyed */
4869		return (0);
4870	}
4871#ifdef SCTP_LOG_CLOSING
4872	sctp_log_closing(inp, stcb, 10);
4873#endif
4874	/*
4875	 * When I reach here, no others want to kill the assoc yet.. and I
4876	 * own the lock. Now its possible an abort comes in when I do the
4877	 * lock exchange below to grab all the locks to do the final take
4878	 * out. to prevent this we increment the count, which will start a
4879	 * timer and blow out above thus assuring us that we hold exclusive
4880	 * killing of the asoc. Note that after getting back the TCB lock we
4881	 * will go ahead and increment the counter back up and stop any
4882	 * timer a passing stranger may have started :-S
4883	 */
4884	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4885		atomic_add_int(&stcb->asoc.refcnt, 1);
4886
4887		SCTP_TCB_UNLOCK(stcb);
4888		SCTP_INP_INFO_WLOCK();
4889		SCTP_INP_WLOCK(inp);
4890		SCTP_TCB_LOCK(stcb);
4891	}
4892	/* Double check the GONE flag */
4893	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4894	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4895		/* nothing around */
4896		so = NULL;
4897
4898	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4899	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4900		/*
4901		 * For TCP type we need special handling when we are
4902		 * connected. We also include the peel'ed off ones to.
4903		 */
4904		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4905			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4906			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4907			if (so) {
4908				SOCK_LOCK(so);
4909				if (so->so_rcv.sb_cc == 0) {
4910					so->so_state &= ~(SS_ISCONNECTING |
4911					    SS_ISDISCONNECTING |
4912					    SS_ISCONFIRMING |
4913					    SS_ISCONNECTED);
4914				}
4915				socantrcvmore_locked(so);
4916				sctp_sowwakeup(inp, so);
4917				sctp_sorwakeup(inp, so);
4918				SCTP_SOWAKEUP(so);
4919			}
4920		}
4921	}
4922	/*
4923	 * Make it invalid too, that way if its about to run it will abort
4924	 * and return.
4925	 */
4926	/* re-increment the lock */
4927	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4928		atomic_add_int(&stcb->asoc.refcnt, -1);
4929	}
4930	if (stcb->asoc.refcnt) {
4931		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4932		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4933		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4934			SCTP_INP_INFO_WUNLOCK();
4935			SCTP_INP_WUNLOCK(inp);
4936		}
4937		SCTP_TCB_UNLOCK(stcb);
4938		return (0);
4939	}
4940	asoc->state = 0;
4941	if (inp->sctp_tcbhash) {
4942		LIST_REMOVE(stcb, sctp_tcbhash);
4943	}
4944	if (stcb->asoc.in_asocid_hash) {
4945		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4946	}
4947	/* Now lets remove it from the list of ALL associations in the EP */
4948	LIST_REMOVE(stcb, sctp_tcblist);
4949	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4950		SCTP_INP_INCR_REF(inp);
4951		SCTP_INP_WUNLOCK(inp);
4952	}
4953	/* pull from vtag hash */
4954	LIST_REMOVE(stcb, sctp_asocs);
4955	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4956	    inp->sctp_lport, stcb->rport);
4957
4958	/*
4959	 * Now restop the timers to be sure this is paranoia at is finest!
4960	 */
4961	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4962	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4963	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4964	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4965	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4966	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4967	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4968	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4969		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4970		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4971		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4972	}
4973
4974	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4975	/*
4976	 * The chunk lists and such SHOULD be empty but we check them just
4977	 * in case.
4978	 */
4979	/* anything on the wheel needs to be removed */
4980	for (i = 0; i < asoc->streamoutcnt; i++) {
4981		struct sctp_stream_out *outs;
4982
4983		outs = &asoc->strmout[i];
4984		/* now clean up any chunks here */
4985		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4986			TAILQ_REMOVE(&outs->outqueue, sp, next);
4987			if (sp->data) {
4988				if (so) {
4989					/* Still an open socket - report */
4990					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4991					    0, (void *)sp, SCTP_SO_LOCKED);
4992				}
4993				if (sp->data) {
4994					sctp_m_freem(sp->data);
4995					sp->data = NULL;
4996					sp->tail_mbuf = NULL;
4997				}
4998			}
4999			if (sp->net) {
5000				sctp_free_remote_addr(sp->net);
5001				sp->net = NULL;
5002			}
5003			sctp_free_spbufspace(stcb, asoc, sp);
5004			if (sp->holds_key_ref)
5005				sctp_auth_key_release(stcb, sp->auth_keyid, SCTP_SO_LOCKED);
5006			/* Free the zone stuff  */
5007			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_strmoq), sp);
5008			SCTP_DECR_STRMOQ_COUNT();
5009			/* sa_ignore FREED_MEMORY */
5010		}
5011	}
5012	/* sa_ignore FREED_MEMORY */
5013	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
5014		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
5015		SCTP_FREE(strrst, SCTP_M_STRESET);
5016	}
5017	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
5018		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
5019		if (sq->data) {
5020			sctp_m_freem(sq->data);
5021			sq->data = NULL;
5022		}
5023		sctp_free_remote_addr(sq->whoFrom);
5024		sq->whoFrom = NULL;
5025		sq->stcb = NULL;
5026		/* Free the ctl entry */
5027		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
5028		SCTP_DECR_READQ_COUNT();
5029		/* sa_ignore FREED_MEMORY */
5030	}
5031	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
5032		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
5033		if (chk->data) {
5034			sctp_m_freem(chk->data);
5035			chk->data = NULL;
5036		}
5037		if (chk->holds_key_ref)
5038			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5039		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5040		SCTP_DECR_CHK_COUNT();
5041		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
5042		asoc->free_chunk_cnt--;
5043		/* sa_ignore FREED_MEMORY */
5044	}
5045	/* pending send queue SHOULD be empty */
5046	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
5047		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
5048		if (chk->data) {
5049			if (so) {
5050				/* Still a socket? */
5051				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
5052				    0, chk, SCTP_SO_LOCKED);
5053			}
5054			if (chk->data) {
5055				sctp_m_freem(chk->data);
5056				chk->data = NULL;
5057			}
5058		}
5059		if (chk->holds_key_ref)
5060			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5061		if (chk->whoTo) {
5062			sctp_free_remote_addr(chk->whoTo);
5063			chk->whoTo = NULL;
5064		}
5065		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5066		SCTP_DECR_CHK_COUNT();
5067		/* sa_ignore FREED_MEMORY */
5068	}
5069	/* sent queue SHOULD be empty */
5070	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5071		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5072		if (chk->data) {
5073			if (so) {
5074				/* Still a socket? */
5075				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5076				    0, chk, SCTP_SO_LOCKED);
5077			}
5078			if (chk->data) {
5079				sctp_m_freem(chk->data);
5080				chk->data = NULL;
5081			}
5082		}
5083		if (chk->holds_key_ref)
5084			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5085		sctp_free_remote_addr(chk->whoTo);
5086		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5087		SCTP_DECR_CHK_COUNT();
5088		/* sa_ignore FREED_MEMORY */
5089	}
5090	/* control queue MAY not be empty */
5091	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5092		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5093		if (chk->data) {
5094			sctp_m_freem(chk->data);
5095			chk->data = NULL;
5096		}
5097		if (chk->holds_key_ref)
5098			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5099		sctp_free_remote_addr(chk->whoTo);
5100		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5101		SCTP_DECR_CHK_COUNT();
5102		/* sa_ignore FREED_MEMORY */
5103	}
5104	/* ASCONF queue MAY not be empty */
5105	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5106		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5107		if (chk->data) {
5108			sctp_m_freem(chk->data);
5109			chk->data = NULL;
5110		}
5111		if (chk->holds_key_ref)
5112			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5113		sctp_free_remote_addr(chk->whoTo);
5114		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5115		SCTP_DECR_CHK_COUNT();
5116		/* sa_ignore FREED_MEMORY */
5117	}
5118	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5119		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5120		if (chk->data) {
5121			sctp_m_freem(chk->data);
5122			chk->data = NULL;
5123		}
5124		if (chk->holds_key_ref)
5125			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5126		sctp_free_remote_addr(chk->whoTo);
5127		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5128		SCTP_DECR_CHK_COUNT();
5129		/* sa_ignore FREED_MEMORY */
5130	}
5131
5132	if (asoc->mapping_array) {
5133		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5134		asoc->mapping_array = NULL;
5135	}
5136	if (asoc->nr_mapping_array) {
5137		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5138		asoc->nr_mapping_array = NULL;
5139	}
5140	/* the stream outs */
5141	if (asoc->strmout) {
5142		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5143		asoc->strmout = NULL;
5144	}
5145	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5146	if (asoc->strmin) {
5147		struct sctp_queued_to_read *ctl, *nctl;
5148
5149		for (i = 0; i < asoc->streamincnt; i++) {
5150			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5151				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5152				sctp_free_remote_addr(ctl->whoFrom);
5153				if (ctl->data) {
5154					sctp_m_freem(ctl->data);
5155					ctl->data = NULL;
5156				}
5157				/*
5158				 * We don't free the address here since all
5159				 * the net's were freed above.
5160				 */
5161				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5162				SCTP_DECR_READQ_COUNT();
5163			}
5164		}
5165		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5166		asoc->strmin = NULL;
5167	}
5168	asoc->streamincnt = 0;
5169	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5170#ifdef INVARIANTS
5171		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5172			panic("no net's left alloc'ed, or list points to itself");
5173		}
5174#endif
5175		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5176		sctp_free_remote_addr(net);
5177	}
5178	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5179		/* sa_ignore FREED_MEMORY */
5180		sctp_remove_laddr(laddr);
5181	}
5182
5183	/* pending asconf (address) parameters */
5184	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5185		/* sa_ignore FREED_MEMORY */
5186		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5187		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5188	}
5189	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5190		/* sa_ignore FREED_MEMORY */
5191		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5192		if (aack->data != NULL) {
5193			sctp_m_freem(aack->data);
5194		}
5195		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5196	}
5197	/* clean up auth stuff */
5198	if (asoc->local_hmacs)
5199		sctp_free_hmaclist(asoc->local_hmacs);
5200	if (asoc->peer_hmacs)
5201		sctp_free_hmaclist(asoc->peer_hmacs);
5202
5203	if (asoc->local_auth_chunks)
5204		sctp_free_chunklist(asoc->local_auth_chunks);
5205	if (asoc->peer_auth_chunks)
5206		sctp_free_chunklist(asoc->peer_auth_chunks);
5207
5208	sctp_free_authinfo(&asoc->authinfo);
5209
5210	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5211		LIST_REMOVE(shared_key, next);
5212		sctp_free_sharedkey(shared_key);
5213		/* sa_ignore FREED_MEMORY */
5214	}
5215
5216	/* Insert new items here :> */
5217
5218	/* Get rid of LOCK */
5219	SCTP_TCB_LOCK_DESTROY(stcb);
5220	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5221	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5222		SCTP_INP_INFO_WUNLOCK();
5223		SCTP_INP_RLOCK(inp);
5224	}
5225#ifdef SCTP_TRACK_FREED_ASOCS
5226	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5227		/* now clean up the tasoc itself */
5228		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5229		SCTP_DECR_ASOC_COUNT();
5230	} else {
5231		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5232	}
5233#else
5234	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5235	SCTP_DECR_ASOC_COUNT();
5236#endif
5237	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5238		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5239			/*
5240			 * If its NOT the inp_free calling us AND sctp_close
5241			 * as been called, we call back...
5242			 */
5243			SCTP_INP_RUNLOCK(inp);
5244			/*
5245			 * This will start the kill timer (if we are the
5246			 * last one) since we hold an increment yet. But
5247			 * this is the only safe way to do this since
5248			 * otherwise if the socket closes at the same time
5249			 * we are here we might collide in the cleanup.
5250			 */
5251			sctp_inpcb_free(inp,
5252			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5253			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5254			SCTP_INP_DECR_REF(inp);
5255			goto out_of;
5256		} else {
5257			/* The socket is still open. */
5258			SCTP_INP_DECR_REF(inp);
5259		}
5260	}
5261	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5262		SCTP_INP_RUNLOCK(inp);
5263	}
5264out_of:
5265	/* destroyed the asoc */
5266#ifdef SCTP_LOG_CLOSING
5267	sctp_log_closing(inp, NULL, 11);
5268#endif
5269	return (1);
5270}
5271
5272
5273
5274/*
5275 * determine if a destination is "reachable" based upon the addresses bound
5276 * to the current endpoint (e.g. only v4 or v6 currently bound)
5277 */
5278/*
5279 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5280 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5281 * types bound as its endpoint
5282 */
5283int
5284sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5285{
5286	struct sctp_inpcb *inp;
5287	int answer;
5288
5289	/*
5290	 * No locks here, the TCB, in all cases is already locked and an
5291	 * assoc is up. There is either a INP lock by the caller applied (in
5292	 * asconf case when deleting an address) or NOT in the HB case,
5293	 * however if HB then the INP increment is up and the INP will not
5294	 * be removed (on top of the fact that we have a TCB lock). So we
5295	 * only want to read the sctp_flags, which is either bound-all or
5296	 * not.. no protection needed since once an assoc is up you can't be
5297	 * changing your binding.
5298	 */
5299	inp = stcb->sctp_ep;
5300	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5301		/* if bound all, destination is not restricted */
5302		/*
5303		 * RRS: Question during lock work: Is this correct? If you
5304		 * are bound-all you still might need to obey the V4--V6
5305		 * flags??? IMO this bound-all stuff needs to be removed!
5306		 */
5307		return (1);
5308	}
5309	/* NOTE: all "scope" checks are done when local addresses are added */
5310	switch (destaddr->sa_family) {
5311	case AF_INET6:
5312		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5313		break;
5314	case AF_INET:
5315		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5316		break;
5317	default:
5318		/* invalid family, so it's unreachable */
5319		answer = 0;
5320		break;
5321	}
5322	return (answer);
5323}
5324
5325/*
5326 * update the inp_vflags on an endpoint
5327 */
5328static void
5329sctp_update_ep_vflag(struct sctp_inpcb *inp)
5330{
5331	struct sctp_laddr *laddr;
5332
5333	/* first clear the flag */
5334	inp->ip_inp.inp.inp_vflag = 0;
5335	/* set the flag based on addresses on the ep list */
5336	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5337		if (laddr->ifa == NULL) {
5338			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5339			    __FUNCTION__);
5340			continue;
5341		}
5342		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5343			continue;
5344		}
5345		switch (laddr->ifa->address.sa.sa_family) {
5346#ifdef INET6
5347		case AF_INET6:
5348			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5349			break;
5350#endif
5351#ifdef INET
5352		case AF_INET:
5353			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5354			break;
5355#endif
5356		default:
5357			break;
5358		}
5359	}
5360}
5361
5362/*
5363 * Add the address to the endpoint local address list There is nothing to be
5364 * done if we are bound to all addresses
5365 */
5366void
5367sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5368{
5369	struct sctp_laddr *laddr;
5370	int fnd, error = 0;
5371
5372	fnd = 0;
5373
5374	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5375		/* You are already bound to all. You have it already */
5376		return;
5377	}
5378#ifdef INET6
5379	if (ifa->address.sa.sa_family == AF_INET6) {
5380		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5381			/* Can't bind a non-useable addr. */
5382			return;
5383		}
5384	}
5385#endif
5386	/* first, is it already present? */
5387	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5388		if (laddr->ifa == ifa) {
5389			fnd = 1;
5390			break;
5391		}
5392	}
5393
5394	if (fnd == 0) {
5395		/* Not in the ep list */
5396		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5397		if (error != 0)
5398			return;
5399		inp->laddr_count++;
5400		/* update inp_vflag flags */
5401		switch (ifa->address.sa.sa_family) {
5402#ifdef INET6
5403		case AF_INET6:
5404			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5405			break;
5406#endif
5407#ifdef INET6
5408		case AF_INET:
5409			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5410			break;
5411#endif
5412		default:
5413			break;
5414		}
5415	}
5416	return;
5417}
5418
5419
5420/*
5421 * select a new (hopefully reachable) destination net (should only be used
5422 * when we deleted an ep addr that is the only usable source address to reach
5423 * the destination net)
5424 */
5425static void
5426sctp_select_primary_destination(struct sctp_tcb *stcb)
5427{
5428	struct sctp_nets *net;
5429
5430	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5431		/* for now, we'll just pick the first reachable one we find */
5432		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5433			continue;
5434		if (sctp_destination_is_reachable(stcb,
5435		    (struct sockaddr *)&net->ro._l_addr)) {
5436			/* found a reachable destination */
5437			stcb->asoc.primary_destination = net;
5438		}
5439	}
5440	/* I can't there from here! ...we're gonna die shortly... */
5441}
5442
5443
5444/*
5445 * Delete the address from the endpoint local address list There is nothing
5446 * to be done if we are bound to all addresses
5447 */
5448void
5449sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5450{
5451	struct sctp_laddr *laddr;
5452	int fnd;
5453
5454	fnd = 0;
5455	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5456		/* You are already bound to all. You have it already */
5457		return;
5458	}
5459	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5460		if (laddr->ifa == ifa) {
5461			fnd = 1;
5462			break;
5463		}
5464	}
5465	if (fnd && (inp->laddr_count < 2)) {
5466		/* can't delete unless there are at LEAST 2 addresses */
5467		return;
5468	}
5469	if (fnd) {
5470		/*
5471		 * clean up any use of this address go through our
5472		 * associations and clear any last_used_address that match
5473		 * this one for each assoc, see if a new primary_destination
5474		 * is needed
5475		 */
5476		struct sctp_tcb *stcb;
5477
5478		/* clean up "next_addr_touse" */
5479		if (inp->next_addr_touse == laddr)
5480			/* delete this address */
5481			inp->next_addr_touse = NULL;
5482
5483		/* clean up "last_used_address" */
5484		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5485			struct sctp_nets *net;
5486
5487			SCTP_TCB_LOCK(stcb);
5488			if (stcb->asoc.last_used_address == laddr)
5489				/* delete this address */
5490				stcb->asoc.last_used_address = NULL;
5491			/*
5492			 * Now spin through all the nets and purge any ref
5493			 * to laddr
5494			 */
5495			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5496				if (net->ro._s_addr &&
5497				    (net->ro._s_addr->ifa == laddr->ifa)) {
5498					/* Yep, purge src address selected */
5499					sctp_rtentry_t *rt;
5500
5501					/* delete this address if cached */
5502					rt = net->ro.ro_rt;
5503					if (rt != NULL) {
5504						RTFREE(rt);
5505						net->ro.ro_rt = NULL;
5506					}
5507					sctp_free_ifa(net->ro._s_addr);
5508					net->ro._s_addr = NULL;
5509					net->src_addr_selected = 0;
5510				}
5511			}
5512			SCTP_TCB_UNLOCK(stcb);
5513		}		/* for each tcb */
5514		/* remove it from the ep list */
5515		sctp_remove_laddr(laddr);
5516		inp->laddr_count--;
5517		/* update inp_vflag flags */
5518		sctp_update_ep_vflag(inp);
5519	}
5520	return;
5521}
5522
5523/*
5524 * Add the address to the TCB local address restricted list.
5525 * This is a "pending" address list (eg. addresses waiting for an
5526 * ASCONF-ACK response) and cannot be used as a valid source address.
5527 */
5528void
5529sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5530{
5531	struct sctp_laddr *laddr;
5532	struct sctpladdr *list;
5533
5534	/*
5535	 * Assumes TCB is locked.. and possibly the INP. May need to
5536	 * confirm/fix that if we need it and is not the case.
5537	 */
5538	list = &stcb->asoc.sctp_restricted_addrs;
5539
5540#ifdef INET6
5541	if (ifa->address.sa.sa_family == AF_INET6) {
5542		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5543			/* Can't bind a non-existent addr. */
5544			return;
5545		}
5546	}
5547#endif
5548	/* does the address already exist? */
5549	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5550		if (laddr->ifa == ifa) {
5551			return;
5552		}
5553	}
5554
5555	/* add to the list */
5556	(void)sctp_insert_laddr(list, ifa, 0);
5557	return;
5558}
5559
5560/*
5561 * insert an laddr entry with the given ifa for the desired list
5562 */
5563int
5564sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5565{
5566	struct sctp_laddr *laddr;
5567
5568	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5569	if (laddr == NULL) {
5570		/* out of memory? */
5571		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5572		return (EINVAL);
5573	}
5574	SCTP_INCR_LADDR_COUNT();
5575	bzero(laddr, sizeof(*laddr));
5576	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5577	laddr->ifa = ifa;
5578	laddr->action = act;
5579	atomic_add_int(&ifa->refcount, 1);
5580	/* insert it */
5581	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5582
5583	return (0);
5584}
5585
5586/*
5587 * Remove an laddr entry from the local address list (on an assoc)
5588 */
5589void
5590sctp_remove_laddr(struct sctp_laddr *laddr)
5591{
5592
5593	/* remove from the list */
5594	LIST_REMOVE(laddr, sctp_nxt_addr);
5595	sctp_free_ifa(laddr->ifa);
5596	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5597	SCTP_DECR_LADDR_COUNT();
5598}
5599
5600/*
5601 * Remove a local address from the TCB local address restricted list
5602 */
5603void
5604sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5605{
5606	struct sctp_inpcb *inp;
5607	struct sctp_laddr *laddr;
5608
5609	/*
5610	 * This is called by asconf work. It is assumed that a) The TCB is
5611	 * locked and b) The INP is locked. This is true in as much as I can
5612	 * trace through the entry asconf code where I did these locks.
5613	 * Again, the ASCONF code is a bit different in that it does lock
5614	 * the INP during its work often times. This must be since we don't
5615	 * want other proc's looking up things while what they are looking
5616	 * up is changing :-D
5617	 */
5618
5619	inp = stcb->sctp_ep;
5620	/* if subset bound and don't allow ASCONF's, can't delete last */
5621	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5622	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5623		if (stcb->sctp_ep->laddr_count < 2) {
5624			/* can't delete last address */
5625			return;
5626		}
5627	}
5628	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5629		/* remove the address if it exists */
5630		if (laddr->ifa == NULL)
5631			continue;
5632		if (laddr->ifa == ifa) {
5633			sctp_remove_laddr(laddr);
5634			return;
5635		}
5636	}
5637
5638	/* address not found! */
5639	return;
5640}
5641
5642/*
5643 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5644 */
5645/* sysctl */
5646static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5647static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5648
5649
5650
5651#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5652struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5653int *sctp_cpuarry = NULL;
5654void
5655sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5656{
5657	/* Queue a packet to a processor for the specified core */
5658	struct sctp_mcore_queue *qent;
5659	struct sctp_mcore_ctrl *wkq;
5660	int need_wake = 0;
5661
5662	if (sctp_mcore_workers == NULL) {
5663		/* Something went way bad during setup */
5664		sctp_input_with_port(m, off, 0);
5665		return;
5666	}
5667	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5668	    (sizeof(struct sctp_mcore_queue)),
5669	    SCTP_M_MCORE);
5670	if (qent == NULL) {
5671		/* This is trouble  */
5672		sctp_input_with_port(m, off, 0);
5673		return;
5674	}
5675	qent->vn = curvnet;
5676	qent->m = m;
5677	qent->off = off;
5678	qent->v6 = 0;
5679	wkq = &sctp_mcore_workers[cpu_to_use];
5680	SCTP_MCORE_QLOCK(wkq);
5681
5682	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5683	if (wkq->running == 0) {
5684		need_wake = 1;
5685	}
5686	SCTP_MCORE_QUNLOCK(wkq);
5687	if (need_wake) {
5688		wakeup(&wkq->running);
5689	}
5690}
5691
5692static void
5693sctp_mcore_thread(void *arg)
5694{
5695
5696	struct sctp_mcore_ctrl *wkq;
5697	struct sctp_mcore_queue *qent;
5698
5699	wkq = (struct sctp_mcore_ctrl *)arg;
5700	struct mbuf *m;
5701	int off, v6;
5702
5703	/* Wait for first tickle */
5704	SCTP_MCORE_LOCK(wkq);
5705	wkq->running = 0;
5706	msleep(&wkq->running,
5707	    &wkq->core_mtx,
5708	    0, "wait for pkt", 0);
5709	SCTP_MCORE_UNLOCK(wkq);
5710
5711	/* Bind to our cpu */
5712	thread_lock(curthread);
5713	sched_bind(curthread, wkq->cpuid);
5714	thread_unlock(curthread);
5715
5716	/* Now lets start working */
5717	SCTP_MCORE_LOCK(wkq);
5718	/* Now grab lock and go */
5719	for (;;) {
5720		SCTP_MCORE_QLOCK(wkq);
5721skip_sleep:
5722		wkq->running = 1;
5723		qent = TAILQ_FIRST(&wkq->que);
5724		if (qent) {
5725			TAILQ_REMOVE(&wkq->que, qent, next);
5726			SCTP_MCORE_QUNLOCK(wkq);
5727			CURVNET_SET(qent->vn);
5728			m = qent->m;
5729			off = qent->off;
5730			v6 = qent->v6;
5731			SCTP_FREE(qent, SCTP_M_MCORE);
5732			if (v6 == 0) {
5733				sctp_input_with_port(m, off, 0);
5734			} else {
5735				SCTP_PRINTF("V6 not yet supported\n");
5736				sctp_m_freem(m);
5737			}
5738			CURVNET_RESTORE();
5739			SCTP_MCORE_QLOCK(wkq);
5740		}
5741		wkq->running = 0;
5742		if (!TAILQ_EMPTY(&wkq->que)) {
5743			goto skip_sleep;
5744		}
5745		SCTP_MCORE_QUNLOCK(wkq);
5746		msleep(&wkq->running,
5747		    &wkq->core_mtx,
5748		    0, "wait for pkt", 0);
5749	}
5750}
5751
5752static void
5753sctp_startup_mcore_threads(void)
5754{
5755	int i, cpu;
5756
5757	if (mp_ncpus == 1)
5758		return;
5759
5760	if (sctp_mcore_workers != NULL) {
5761		/*
5762		 * Already been here in some previous vnet?
5763		 */
5764		return;
5765	}
5766	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5767	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5768	    SCTP_M_MCORE);
5769	if (sctp_mcore_workers == NULL) {
5770		/* TSNH I hope */
5771		return;
5772	}
5773	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5774	    sizeof(struct sctp_mcore_ctrl)));
5775	/* Init the structures */
5776	for (i = 0; i <= mp_maxid; i++) {
5777		TAILQ_INIT(&sctp_mcore_workers[i].que);
5778		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5779		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5780		sctp_mcore_workers[i].cpuid = i;
5781	}
5782	if (sctp_cpuarry == NULL) {
5783		SCTP_MALLOC(sctp_cpuarry, int *,
5784		    (mp_ncpus * sizeof(int)),
5785		    SCTP_M_MCORE);
5786		i = 0;
5787		CPU_FOREACH(cpu) {
5788			sctp_cpuarry[i] = cpu;
5789			i++;
5790		}
5791	}
5792	/* Now start them all */
5793	CPU_FOREACH(cpu) {
5794		(void)kproc_create(sctp_mcore_thread,
5795		    (void *)&sctp_mcore_workers[cpu],
5796		    &sctp_mcore_workers[cpu].thread_proc,
5797		    RFPROC,
5798		    SCTP_KTHREAD_PAGES,
5799		    SCTP_MCORE_NAME);
5800
5801	}
5802}
5803
5804#endif
5805
5806void
5807sctp_pcb_init()
5808{
5809	/*
5810	 * SCTP initialization for the PCB structures should be called by
5811	 * the sctp_init() funciton.
5812	 */
5813	int i;
5814	struct timeval tv;
5815
5816	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5817		/* error I was called twice */
5818		return;
5819	}
5820	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5821
5822#if defined(SCTP_LOCAL_TRACE_BUF)
5823	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5824#endif
5825#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5826	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5827	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5828	    SCTP_M_MCORE);
5829#endif
5830	(void)SCTP_GETTIME_TIMEVAL(&tv);
5831#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5832	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5833	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5834	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5835#else
5836	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5837	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5838	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5839#endif
5840	/* init the empty list of (All) Endpoints */
5841	LIST_INIT(&SCTP_BASE_INFO(listhead));
5842
5843
5844	/* init the hash table of endpoints */
5845	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5846	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5847	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5848	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5849	    &SCTP_BASE_INFO(hashasocmark));
5850	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5851	    &SCTP_BASE_INFO(hashmark));
5852	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5853	    &SCTP_BASE_INFO(hashtcpmark));
5854	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5855
5856
5857	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5858	    &SCTP_BASE_INFO(hashvrfmark));
5859
5860	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5861	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5862	/* init the zones */
5863	/*
5864	 * FIX ME: Should check for NULL returns, but if it does fail we are
5865	 * doomed to panic anyways... add later maybe.
5866	 */
5867	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5868	    sizeof(struct sctp_inpcb), maxsockets);
5869
5870	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5871	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5872
5873	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5874	    sizeof(struct sctp_laddr),
5875	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5876
5877	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5878	    sizeof(struct sctp_nets),
5879	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5880
5881	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5882	    sizeof(struct sctp_tmit_chunk),
5883	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5884
5885	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5886	    sizeof(struct sctp_queued_to_read),
5887	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5888
5889	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5890	    sizeof(struct sctp_stream_queue_pending),
5891	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5892
5893	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5894	    sizeof(struct sctp_asconf),
5895	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5896
5897	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5898	    sizeof(struct sctp_asconf_ack),
5899	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5900
5901
5902	/* Master Lock INIT for info structure */
5903	SCTP_INP_INFO_LOCK_INIT();
5904	SCTP_STATLOG_INIT_LOCK();
5905
5906	SCTP_IPI_COUNT_INIT();
5907	SCTP_IPI_ADDR_INIT();
5908#ifdef SCTP_PACKET_LOGGING
5909	SCTP_IP_PKTLOG_INIT();
5910#endif
5911	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5912
5913	SCTP_WQ_ADDR_INIT();
5914	/* not sure if we need all the counts */
5915	SCTP_BASE_INFO(ipi_count_ep) = 0;
5916	/* assoc/tcb zone info */
5917	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5918	/* local addrlist zone info */
5919	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5920	/* remote addrlist zone info */
5921	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5922	/* chunk info */
5923	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5924
5925	/* socket queue zone info */
5926	SCTP_BASE_INFO(ipi_count_readq) = 0;
5927
5928	/* stream out queue cont */
5929	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5930
5931	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5932	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5933
5934	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5935
5936	/* Init the TIMEWAIT list */
5937	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5938		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5939	}
5940
5941	sctp_startup_iterator();
5942
5943#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5944	sctp_startup_mcore_threads();
5945#endif
5946
5947	/*
5948	 * INIT the default VRF which for BSD is the only one, other O/S's
5949	 * may have more. But initially they must start with one and then
5950	 * add the VRF's as addresses are added.
5951	 */
5952	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5953}
5954
5955/*
5956 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5957 */
5958void
5959sctp_pcb_finish(void)
5960{
5961	struct sctp_vrflist *vrf_bucket;
5962	struct sctp_vrf *vrf, *nvrf;
5963	struct sctp_ifn *ifn, *nifn;
5964	struct sctp_ifa *ifa, *nifa;
5965	struct sctpvtaghead *chain;
5966	struct sctp_tagblock *twait_block, *prev_twait_block;
5967	struct sctp_laddr *wi, *nwi;
5968	int i;
5969
5970	/*
5971	 * Free BSD the it thread never exits but we do clean up. The only
5972	 * way freebsd reaches here if we have VRF's but we still add the
5973	 * ifdef to make it compile on old versions.
5974	 */
5975	{
5976		struct sctp_iterator *it, *nit;
5977
5978		SCTP_IPI_ITERATOR_WQ_LOCK();
5979		TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5980			if (it->vn != curvnet) {
5981				continue;
5982			}
5983			TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5984			if (it->function_atend != NULL) {
5985				(*it->function_atend) (it->pointer, it->val);
5986			}
5987			SCTP_FREE(it, SCTP_M_ITER);
5988		}
5989		SCTP_IPI_ITERATOR_WQ_UNLOCK();
5990		SCTP_ITERATOR_LOCK();
5991		if ((sctp_it_ctl.cur_it) &&
5992		    (sctp_it_ctl.cur_it->vn == curvnet)) {
5993			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5994		}
5995		SCTP_ITERATOR_UNLOCK();
5996	}
5997
5998	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5999	SCTP_WQ_ADDR_LOCK();
6000	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
6001		LIST_REMOVE(wi, sctp_nxt_addr);
6002		SCTP_DECR_LADDR_COUNT();
6003		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
6004	}
6005	SCTP_WQ_ADDR_UNLOCK();
6006
6007	/*
6008	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
6009	 * destroyed first).
6010	 */
6011	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
6012	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
6013		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
6014			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
6015				/* free the ifa */
6016				LIST_REMOVE(ifa, next_bucket);
6017				LIST_REMOVE(ifa, next_ifa);
6018				SCTP_FREE(ifa, SCTP_M_IFA);
6019			}
6020			/* free the ifn */
6021			LIST_REMOVE(ifn, next_bucket);
6022			LIST_REMOVE(ifn, next_ifn);
6023			SCTP_FREE(ifn, SCTP_M_IFN);
6024		}
6025		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
6026		/* free the vrf */
6027		LIST_REMOVE(vrf, next_vrf);
6028		SCTP_FREE(vrf, SCTP_M_VRF);
6029	}
6030	/* free the vrf hashes */
6031	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
6032	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
6033
6034	/*
6035	 * free the TIMEWAIT list elements malloc'd in the function
6036	 * sctp_add_vtag_to_timewait()...
6037	 */
6038	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
6039		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
6040		if (!LIST_EMPTY(chain)) {
6041			prev_twait_block = NULL;
6042			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6043				if (prev_twait_block) {
6044					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6045				}
6046				prev_twait_block = twait_block;
6047			}
6048			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6049		}
6050	}
6051
6052	/* free the locks and mutexes */
6053#ifdef SCTP_PACKET_LOGGING
6054	SCTP_IP_PKTLOG_DESTROY();
6055#endif
6056	SCTP_IPI_ADDR_DESTROY();
6057	SCTP_STATLOG_DESTROY();
6058	SCTP_INP_INFO_LOCK_DESTROY();
6059
6060	SCTP_WQ_ADDR_DESTROY();
6061
6062	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6063	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6064	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6065	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6066	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6067	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6068	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6069	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6070	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6071	/* Get rid of other stuff to */
6072	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
6073		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
6074	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
6075		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6076	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6077		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6078#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6079	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6080#endif
6081}
6082
6083
6084int
6085sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6086    int offset, int limit, struct sctphdr *sh,
6087    struct sockaddr *altsa)
6088{
6089	/*
6090	 * grub through the INIT pulling addresses and loading them to the
6091	 * nets structure in the asoc. The from address in the mbuf should
6092	 * also be loaded (if it is not already). This routine can be called
6093	 * with either INIT or INIT-ACK's as long as the m points to the IP
6094	 * packet and the offset points to the beginning of the parameters.
6095	 */
6096	struct sctp_inpcb *inp;
6097	struct sctp_nets *net, *nnet, *net_tmp;
6098	struct ip *iph;
6099	struct sctp_paramhdr *phdr, parm_buf;
6100	struct sctp_tcb *stcb_tmp;
6101	uint16_t ptype, plen;
6102	struct sockaddr *sa;
6103	struct sockaddr_storage dest_store;
6104	struct sockaddr *local_sa = (struct sockaddr *)&dest_store;
6105	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6106	struct sctp_auth_random *p_random = NULL;
6107	uint16_t random_len = 0;
6108	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6109	struct sctp_auth_hmac_algo *hmacs = NULL;
6110	uint16_t hmacs_len = 0;
6111	uint8_t saw_asconf = 0;
6112	uint8_t saw_asconf_ack = 0;
6113	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6114	struct sctp_auth_chunk_list *chunks = NULL;
6115	uint16_t num_chunks = 0;
6116	sctp_key_t *new_key;
6117	uint32_t keylen;
6118	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6119	uint8_t ecn_allowed;
6120
6121#ifdef INET
6122	struct sockaddr_in sin;
6123
6124#endif
6125#ifdef INET6
6126	struct sockaddr_in6 sin6;
6127
6128#endif
6129
6130	/* First get the destination address setup too. */
6131#ifdef INET
6132	memset(&sin, 0, sizeof(sin));
6133	sin.sin_family = AF_INET;
6134	sin.sin_len = sizeof(sin);
6135	sin.sin_port = stcb->rport;
6136#endif
6137#ifdef INET6
6138	memset(&sin6, 0, sizeof(sin6));
6139	sin6.sin6_family = AF_INET6;
6140	sin6.sin6_len = sizeof(struct sockaddr_in6);
6141	sin6.sin6_port = stcb->rport;
6142#endif
6143	iph = mtod(m, struct ip *);
6144	switch (iph->ip_v) {
6145#ifdef INET
6146	case IPVERSION:
6147		{
6148			/* its IPv4 */
6149			struct sockaddr_in *sin_2;
6150
6151			sin_2 = (struct sockaddr_in *)(local_sa);
6152			memset(sin_2, 0, sizeof(sin));
6153			sin_2->sin_family = AF_INET;
6154			sin_2->sin_len = sizeof(sin);
6155			sin_2->sin_port = sh->dest_port;
6156			sin_2->sin_addr.s_addr = iph->ip_dst.s_addr;
6157			if (altsa) {
6158				/*
6159				 * For cookies we use the src address NOT
6160				 * from the packet but from the original
6161				 * INIT.
6162				 */
6163				sa = altsa;
6164			} else {
6165				sin.sin_addr = iph->ip_src;
6166				sa = (struct sockaddr *)&sin;
6167			}
6168			break;
6169		}
6170#endif
6171#ifdef INET6
6172	case IPV6_VERSION >> 4:
6173		{
6174			/* its IPv6 */
6175			struct ip6_hdr *ip6;
6176			struct sockaddr_in6 *sin6_2;
6177
6178			ip6 = mtod(m, struct ip6_hdr *);
6179			sin6_2 = (struct sockaddr_in6 *)(local_sa);
6180			memset(sin6_2, 0, sizeof(sin6));
6181			sin6_2->sin6_family = AF_INET6;
6182			sin6_2->sin6_len = sizeof(struct sockaddr_in6);
6183			sin6_2->sin6_port = sh->dest_port;
6184			sin6_2->sin6_addr = ip6->ip6_dst;
6185			if (altsa) {
6186				/*
6187				 * For cookies we use the src address NOT
6188				 * from the packet but from the original
6189				 * INIT.
6190				 */
6191				sa = altsa;
6192			} else {
6193				sin6.sin6_addr = ip6->ip6_src;
6194				sa = (struct sockaddr *)&sin6;
6195			}
6196			break;
6197		}
6198#endif
6199	default:
6200		return (-1);
6201		break;
6202	}
6203	/* Turn off ECN until we get through all params */
6204	ecn_allowed = 0;
6205	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6206		/* mark all addresses that we have currently on the list */
6207		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6208	}
6209	/* does the source address already exist? if so skip it */
6210	inp = stcb->sctp_ep;
6211	atomic_add_int(&stcb->asoc.refcnt, 1);
6212	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb);
6213	atomic_add_int(&stcb->asoc.refcnt, -1);
6214
6215	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6216		/* we must add the source address */
6217		/* no scope set here since we have a tcb already. */
6218		switch (sa->sa_family) {
6219#ifdef INET
6220		case AF_INET:
6221			if (stcb->asoc.ipv4_addr_legal) {
6222				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6223					return (-1);
6224				}
6225			}
6226			break;
6227#endif
6228#ifdef INET6
6229		case AF_INET6:
6230			if (stcb->asoc.ipv6_addr_legal) {
6231				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6232					return (-2);
6233				}
6234			}
6235			break;
6236#endif
6237		default:
6238			break;
6239		}
6240	} else {
6241		if (net_tmp != NULL && stcb_tmp == stcb) {
6242			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6243		} else if (stcb_tmp != stcb) {
6244			/* It belongs to another association? */
6245			if (stcb_tmp)
6246				SCTP_TCB_UNLOCK(stcb_tmp);
6247			return (-3);
6248		}
6249	}
6250	if (stcb->asoc.state == 0) {
6251		/* the assoc was freed? */
6252		return (-4);
6253	}
6254	/*
6255	 * peer must explicitly turn this on. This may have been initialized
6256	 * to be "on" in order to allow local addr changes while INIT's are
6257	 * in flight.
6258	 */
6259	stcb->asoc.peer_supports_asconf = 0;
6260	/* now we must go through each of the params. */
6261	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6262	while (phdr) {
6263		ptype = ntohs(phdr->param_type);
6264		plen = ntohs(phdr->param_length);
6265		/*
6266		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6267		 * (uint32_t)ptype, (int)plen);
6268		 */
6269		if (offset + plen > limit) {
6270			break;
6271		}
6272		if (plen == 0) {
6273			break;
6274		}
6275#ifdef INET
6276		if (ptype == SCTP_IPV4_ADDRESS) {
6277			if (stcb->asoc.ipv4_addr_legal) {
6278				struct sctp_ipv4addr_param *p4, p4_buf;
6279
6280				/* ok get the v4 address and check/add */
6281				phdr = sctp_get_next_param(m, offset,
6282				    (struct sctp_paramhdr *)&p4_buf,
6283				    sizeof(p4_buf));
6284				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6285				    phdr == NULL) {
6286					return (-5);
6287				}
6288				p4 = (struct sctp_ipv4addr_param *)phdr;
6289				sin.sin_addr.s_addr = p4->addr;
6290				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6291					/* Skip multi-cast addresses */
6292					goto next_param;
6293				}
6294				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6295				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6296					goto next_param;
6297				}
6298				sa = (struct sockaddr *)&sin;
6299				inp = stcb->sctp_ep;
6300				atomic_add_int(&stcb->asoc.refcnt, 1);
6301				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6302				    local_sa, stcb);
6303				atomic_add_int(&stcb->asoc.refcnt, -1);
6304
6305				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6306				    inp == NULL) {
6307					/* we must add the source address */
6308					/*
6309					 * no scope set since we have a tcb
6310					 * already
6311					 */
6312
6313					/*
6314					 * we must validate the state again
6315					 * here
6316					 */
6317			add_it_now:
6318					if (stcb->asoc.state == 0) {
6319						/* the assoc was freed? */
6320						return (-7);
6321					}
6322					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6323						return (-8);
6324					}
6325				} else if (stcb_tmp == stcb) {
6326					if (stcb->asoc.state == 0) {
6327						/* the assoc was freed? */
6328						return (-10);
6329					}
6330					if (net != NULL) {
6331						/* clear flag */
6332						net->dest_state &=
6333						    ~SCTP_ADDR_NOT_IN_ASSOC;
6334					}
6335				} else {
6336					/*
6337					 * strange, address is in another
6338					 * assoc? straighten out locks.
6339					 */
6340					if (stcb_tmp) {
6341						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6342							/*
6343							 * in setup state we
6344							 * abort this guy
6345							 */
6346							sctp_abort_an_association(stcb_tmp->sctp_ep,
6347							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6348							goto add_it_now;
6349						}
6350						SCTP_TCB_UNLOCK(stcb_tmp);
6351					}
6352					if (stcb->asoc.state == 0) {
6353						/* the assoc was freed? */
6354						return (-12);
6355					}
6356					return (-13);
6357				}
6358			}
6359		} else
6360#endif
6361#ifdef INET6
6362		if (ptype == SCTP_IPV6_ADDRESS) {
6363			if (stcb->asoc.ipv6_addr_legal) {
6364				/* ok get the v6 address and check/add */
6365				struct sctp_ipv6addr_param *p6, p6_buf;
6366
6367				phdr = sctp_get_next_param(m, offset,
6368				    (struct sctp_paramhdr *)&p6_buf,
6369				    sizeof(p6_buf));
6370				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6371				    phdr == NULL) {
6372					return (-14);
6373				}
6374				p6 = (struct sctp_ipv6addr_param *)phdr;
6375				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6376				    sizeof(p6->addr));
6377				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6378					/* Skip multi-cast addresses */
6379					goto next_param;
6380				}
6381				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6382					/*
6383					 * Link local make no sense without
6384					 * scope
6385					 */
6386					goto next_param;
6387				}
6388				sa = (struct sockaddr *)&sin6;
6389				inp = stcb->sctp_ep;
6390				atomic_add_int(&stcb->asoc.refcnt, 1);
6391				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6392				    local_sa, stcb);
6393				atomic_add_int(&stcb->asoc.refcnt, -1);
6394				if (stcb_tmp == NULL &&
6395				    (inp == stcb->sctp_ep || inp == NULL)) {
6396					/*
6397					 * we must validate the state again
6398					 * here
6399					 */
6400			add_it_now6:
6401					if (stcb->asoc.state == 0) {
6402						/* the assoc was freed? */
6403						return (-16);
6404					}
6405					/*
6406					 * we must add the address, no scope
6407					 * set
6408					 */
6409					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6410						return (-17);
6411					}
6412				} else if (stcb_tmp == stcb) {
6413					/*
6414					 * we must validate the state again
6415					 * here
6416					 */
6417					if (stcb->asoc.state == 0) {
6418						/* the assoc was freed? */
6419						return (-19);
6420					}
6421					if (net != NULL) {
6422						/* clear flag */
6423						net->dest_state &=
6424						    ~SCTP_ADDR_NOT_IN_ASSOC;
6425					}
6426				} else {
6427					/*
6428					 * strange, address is in another
6429					 * assoc? straighten out locks.
6430					 */
6431					if (stcb_tmp)
6432						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6433							/*
6434							 * in setup state we
6435							 * abort this guy
6436							 */
6437							sctp_abort_an_association(stcb_tmp->sctp_ep,
6438							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6439							goto add_it_now6;
6440						}
6441					SCTP_TCB_UNLOCK(stcb_tmp);
6442
6443					if (stcb->asoc.state == 0) {
6444						/* the assoc was freed? */
6445						return (-21);
6446					}
6447					return (-22);
6448				}
6449			}
6450		} else
6451#endif
6452		if (ptype == SCTP_ECN_CAPABLE) {
6453			ecn_allowed = 1;
6454		} else if (ptype == SCTP_ULP_ADAPTATION) {
6455			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6456				struct sctp_adaptation_layer_indication ai,
6457				                                *aip;
6458
6459				phdr = sctp_get_next_param(m, offset,
6460				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6461				aip = (struct sctp_adaptation_layer_indication *)phdr;
6462				if (aip) {
6463					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6464					stcb->asoc.adaptation_needed = 1;
6465				}
6466			}
6467		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6468			struct sctp_asconf_addr_param lstore, *fee;
6469			int lptype;
6470			struct sockaddr *lsa = NULL;
6471
6472#ifdef INET
6473			struct sctp_asconf_addrv4_param *fii;
6474
6475#endif
6476
6477			stcb->asoc.peer_supports_asconf = 1;
6478			if (plen > sizeof(lstore)) {
6479				return (-23);
6480			}
6481			phdr = sctp_get_next_param(m, offset,
6482			    (struct sctp_paramhdr *)&lstore,
6483			    min(plen, sizeof(lstore)));
6484			if (phdr == NULL) {
6485				return (-24);
6486			}
6487			fee = (struct sctp_asconf_addr_param *)phdr;
6488			lptype = ntohs(fee->addrp.ph.param_type);
6489			switch (lptype) {
6490#ifdef INET
6491			case SCTP_IPV4_ADDRESS:
6492				if (plen !=
6493				    sizeof(struct sctp_asconf_addrv4_param)) {
6494					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6495					    (int)sizeof(struct sctp_asconf_addrv4_param),
6496					    plen);
6497				} else {
6498					fii = (struct sctp_asconf_addrv4_param *)fee;
6499					sin.sin_addr.s_addr = fii->addrp.addr;
6500					lsa = (struct sockaddr *)&sin;
6501				}
6502				break;
6503#endif
6504#ifdef INET6
6505			case SCTP_IPV6_ADDRESS:
6506				if (plen !=
6507				    sizeof(struct sctp_asconf_addr_param)) {
6508					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6509					    (int)sizeof(struct sctp_asconf_addr_param),
6510					    plen);
6511				} else {
6512					memcpy(sin6.sin6_addr.s6_addr,
6513					    fee->addrp.addr,
6514					    sizeof(fee->addrp.addr));
6515					lsa = (struct sockaddr *)&sin6;
6516				}
6517				break;
6518#endif
6519			default:
6520				break;
6521			}
6522			if (lsa) {
6523				(void)sctp_set_primary_addr(stcb, sa, NULL);
6524			}
6525		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6526			stcb->asoc.peer_supports_nat = 1;
6527		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6528			/* Peer supports pr-sctp */
6529			stcb->asoc.peer_supports_prsctp = 1;
6530		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6531			/* A supported extension chunk */
6532			struct sctp_supported_chunk_types_param *pr_supported;
6533			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6534			int num_ent, i;
6535
6536			phdr = sctp_get_next_param(m, offset,
6537			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6538			if (phdr == NULL) {
6539				return (-25);
6540			}
6541			stcb->asoc.peer_supports_asconf = 0;
6542			stcb->asoc.peer_supports_prsctp = 0;
6543			stcb->asoc.peer_supports_pktdrop = 0;
6544			stcb->asoc.peer_supports_strreset = 0;
6545			stcb->asoc.peer_supports_nr_sack = 0;
6546			stcb->asoc.peer_supports_auth = 0;
6547			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6548			num_ent = plen - sizeof(struct sctp_paramhdr);
6549			for (i = 0; i < num_ent; i++) {
6550				switch (pr_supported->chunk_types[i]) {
6551				case SCTP_ASCONF:
6552				case SCTP_ASCONF_ACK:
6553					stcb->asoc.peer_supports_asconf = 1;
6554					break;
6555				case SCTP_FORWARD_CUM_TSN:
6556					stcb->asoc.peer_supports_prsctp = 1;
6557					break;
6558				case SCTP_PACKET_DROPPED:
6559					stcb->asoc.peer_supports_pktdrop = 1;
6560					break;
6561				case SCTP_NR_SELECTIVE_ACK:
6562					stcb->asoc.peer_supports_nr_sack = 1;
6563					break;
6564				case SCTP_STREAM_RESET:
6565					stcb->asoc.peer_supports_strreset = 1;
6566					break;
6567				case SCTP_AUTHENTICATION:
6568					stcb->asoc.peer_supports_auth = 1;
6569					break;
6570				default:
6571					/* one I have not learned yet */
6572					break;
6573
6574				}
6575			}
6576		} else if (ptype == SCTP_RANDOM) {
6577			if (plen > sizeof(random_store))
6578				break;
6579			if (got_random) {
6580				/* already processed a RANDOM */
6581				goto next_param;
6582			}
6583			phdr = sctp_get_next_param(m, offset,
6584			    (struct sctp_paramhdr *)random_store,
6585			    min(sizeof(random_store), plen));
6586			if (phdr == NULL)
6587				return (-26);
6588			p_random = (struct sctp_auth_random *)phdr;
6589			random_len = plen - sizeof(*p_random);
6590			/* enforce the random length */
6591			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6592				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6593				return (-27);
6594			}
6595			got_random = 1;
6596		} else if (ptype == SCTP_HMAC_LIST) {
6597			int num_hmacs;
6598			int i;
6599
6600			if (plen > sizeof(hmacs_store))
6601				break;
6602			if (got_hmacs) {
6603				/* already processed a HMAC list */
6604				goto next_param;
6605			}
6606			phdr = sctp_get_next_param(m, offset,
6607			    (struct sctp_paramhdr *)hmacs_store,
6608			    min(plen, sizeof(hmacs_store)));
6609			if (phdr == NULL)
6610				return (-28);
6611			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6612			hmacs_len = plen - sizeof(*hmacs);
6613			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6614			/* validate the hmac list */
6615			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6616				return (-29);
6617			}
6618			if (stcb->asoc.peer_hmacs != NULL)
6619				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6620			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6621			if (stcb->asoc.peer_hmacs != NULL) {
6622				for (i = 0; i < num_hmacs; i++) {
6623					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6624					    ntohs(hmacs->hmac_ids[i]));
6625				}
6626			}
6627			got_hmacs = 1;
6628		} else if (ptype == SCTP_CHUNK_LIST) {
6629			int i;
6630
6631			if (plen > sizeof(chunks_store))
6632				break;
6633			if (got_chklist) {
6634				/* already processed a Chunks list */
6635				goto next_param;
6636			}
6637			phdr = sctp_get_next_param(m, offset,
6638			    (struct sctp_paramhdr *)chunks_store,
6639			    min(plen, sizeof(chunks_store)));
6640			if (phdr == NULL)
6641				return (-30);
6642			chunks = (struct sctp_auth_chunk_list *)phdr;
6643			num_chunks = plen - sizeof(*chunks);
6644			if (stcb->asoc.peer_auth_chunks != NULL)
6645				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6646			else
6647				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6648			for (i = 0; i < num_chunks; i++) {
6649				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6650				    stcb->asoc.peer_auth_chunks);
6651				/* record asconf/asconf-ack if listed */
6652				if (chunks->chunk_types[i] == SCTP_ASCONF)
6653					saw_asconf = 1;
6654				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6655					saw_asconf_ack = 1;
6656
6657			}
6658			got_chklist = 1;
6659		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6660			    (ptype == SCTP_STATE_COOKIE) ||
6661			    (ptype == SCTP_UNRECOG_PARAM) ||
6662			    (ptype == SCTP_COOKIE_PRESERVE) ||
6663			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6664			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6665			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6666			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6667		    (ptype == SCTP_SUCCESS_REPORT)) {
6668			 /* don't care */ ;
6669		} else {
6670			if ((ptype & 0x8000) == 0x0000) {
6671				/*
6672				 * must stop processing the rest of the
6673				 * param's. Any report bits were handled
6674				 * with the call to
6675				 * sctp_arethere_unrecognized_parameters()
6676				 * when the INIT or INIT-ACK was first seen.
6677				 */
6678				break;
6679			}
6680		}
6681
6682next_param:
6683		offset += SCTP_SIZE32(plen);
6684		if (offset >= limit) {
6685			break;
6686		}
6687		phdr = sctp_get_next_param(m, offset, &parm_buf,
6688		    sizeof(parm_buf));
6689	}
6690	/* Now check to see if we need to purge any addresses */
6691	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6692		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6693		    SCTP_ADDR_NOT_IN_ASSOC) {
6694			/* This address has been removed from the asoc */
6695			/* remove and free it */
6696			stcb->asoc.numnets--;
6697			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6698			sctp_free_remote_addr(net);
6699			if (net == stcb->asoc.primary_destination) {
6700				stcb->asoc.primary_destination = NULL;
6701				sctp_select_primary_destination(stcb);
6702			}
6703		}
6704	}
6705	if (ecn_allowed == 0) {
6706		stcb->asoc.ecn_allowed = 0;
6707	}
6708	/* validate authentication required parameters */
6709	if (got_random && got_hmacs) {
6710		stcb->asoc.peer_supports_auth = 1;
6711	} else {
6712		stcb->asoc.peer_supports_auth = 0;
6713	}
6714	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6715		/* peer does not support auth but sent a chunks list? */
6716		return (-31);
6717	}
6718	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6719	    !stcb->asoc.peer_supports_auth) {
6720		/* peer supports asconf but not auth? */
6721		return (-32);
6722	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6723	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6724		return (-33);
6725	}
6726	/* concatenate the full random key */
6727	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6728	if (chunks != NULL) {
6729		keylen += sizeof(*chunks) + num_chunks;
6730	}
6731	new_key = sctp_alloc_key(keylen);
6732	if (new_key != NULL) {
6733		/* copy in the RANDOM */
6734		if (p_random != NULL) {
6735			keylen = sizeof(*p_random) + random_len;
6736			bcopy(p_random, new_key->key, keylen);
6737		}
6738		/* append in the AUTH chunks */
6739		if (chunks != NULL) {
6740			bcopy(chunks, new_key->key + keylen,
6741			    sizeof(*chunks) + num_chunks);
6742			keylen += sizeof(*chunks) + num_chunks;
6743		}
6744		/* append in the HMACs */
6745		if (hmacs != NULL) {
6746			bcopy(hmacs, new_key->key + keylen,
6747			    sizeof(*hmacs) + hmacs_len);
6748		}
6749	} else {
6750		/* failed to get memory for the key */
6751		return (-34);
6752	}
6753	if (stcb->asoc.authinfo.peer_random != NULL)
6754		sctp_free_key(stcb->asoc.authinfo.peer_random);
6755	stcb->asoc.authinfo.peer_random = new_key;
6756	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6757	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6758
6759	return (0);
6760}
6761
6762int
6763sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6764    struct sctp_nets *net)
6765{
6766	/* make sure the requested primary address exists in the assoc */
6767	if (net == NULL && sa)
6768		net = sctp_findnet(stcb, sa);
6769
6770	if (net == NULL) {
6771		/* didn't find the requested primary address! */
6772		return (-1);
6773	} else {
6774		/* set the primary address */
6775		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6776			/* Must be confirmed, so queue to set */
6777			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6778			return (0);
6779		}
6780		stcb->asoc.primary_destination = net;
6781		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6782			sctp_free_remote_addr(stcb->asoc.alternate);
6783			stcb->asoc.alternate = NULL;
6784		}
6785		net = TAILQ_FIRST(&stcb->asoc.nets);
6786		if (net != stcb->asoc.primary_destination) {
6787			/*
6788			 * first one on the list is NOT the primary
6789			 * sctp_cmpaddr() is much more efficient if the
6790			 * primary is the first on the list, make it so.
6791			 */
6792			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6793			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6794		}
6795		return (0);
6796	}
6797}
6798
6799int
6800sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6801{
6802	/*
6803	 * This function serves two purposes. It will see if a TAG can be
6804	 * re-used and return 1 for yes it is ok and 0 for don't use that
6805	 * tag. A secondary function it will do is purge out old tags that
6806	 * can be removed.
6807	 */
6808	struct sctpvtaghead *chain;
6809	struct sctp_tagblock *twait_block;
6810	struct sctpasochead *head;
6811	struct sctp_tcb *stcb;
6812	int i;
6813
6814	SCTP_INP_INFO_RLOCK();
6815	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6816	    SCTP_BASE_INFO(hashasocmark))];
6817	if (head == NULL) {
6818		/* invalid vtag */
6819		goto skip_vtag_check;
6820	}
6821	LIST_FOREACH(stcb, head, sctp_asocs) {
6822		/*
6823		 * We choose not to lock anything here. TCB's can't be
6824		 * removed since we have the read lock, so they can't be
6825		 * freed on us, same thing for the INP. I may be wrong with
6826		 * this assumption, but we will go with it for now :-)
6827		 */
6828		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6829			continue;
6830		}
6831		if (stcb->asoc.my_vtag == tag) {
6832			/* candidate */
6833			if (stcb->rport != rport) {
6834				continue;
6835			}
6836			if (stcb->sctp_ep->sctp_lport != lport) {
6837				continue;
6838			}
6839			/* Its a used tag set */
6840			SCTP_INP_INFO_RUNLOCK();
6841			return (0);
6842		}
6843	}
6844skip_vtag_check:
6845
6846	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6847	/* Now what about timed wait ? */
6848	if (!LIST_EMPTY(chain)) {
6849		/*
6850		 * Block(s) are present, lets see if we have this tag in the
6851		 * list
6852		 */
6853		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6854			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6855				if (twait_block->vtag_block[i].v_tag == 0) {
6856					/* not used */
6857					continue;
6858				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6859				    now->tv_sec) {
6860					/* Audit expires this guy */
6861					twait_block->vtag_block[i].tv_sec_at_expire = 0;
6862					twait_block->vtag_block[i].v_tag = 0;
6863					twait_block->vtag_block[i].lport = 0;
6864					twait_block->vtag_block[i].rport = 0;
6865				} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6866					    (twait_block->vtag_block[i].lport == lport) &&
6867				    (twait_block->vtag_block[i].rport == rport)) {
6868					/* Bad tag, sorry :< */
6869					SCTP_INP_INFO_RUNLOCK();
6870					return (0);
6871				}
6872			}
6873		}
6874	}
6875	SCTP_INP_INFO_RUNLOCK();
6876	return (1);
6877}
6878
6879static void
6880sctp_drain_mbufs(struct sctp_tcb *stcb)
6881{
6882	/*
6883	 * We must hunt this association for MBUF's past the cumack (i.e.
6884	 * out of order data that we can renege on).
6885	 */
6886	struct sctp_association *asoc;
6887	struct sctp_tmit_chunk *chk, *nchk;
6888	uint32_t cumulative_tsn_p1;
6889	struct sctp_queued_to_read *ctl, *nctl;
6890	int cnt, strmat;
6891	uint32_t gap, i;
6892	int fnd = 0;
6893
6894	/* We look for anything larger than the cum-ack + 1 */
6895
6896	asoc = &stcb->asoc;
6897	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6898		/* none we can reneg on. */
6899		return;
6900	}
6901	SCTP_STAT_INCR(sctps_protocol_drains_done);
6902	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6903	cnt = 0;
6904	/* First look in the re-assembly queue */
6905	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6906		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6907			/* Yep it is above cum-ack */
6908			cnt++;
6909			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6910			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6911			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6912			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6913			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6914			if (chk->data) {
6915				sctp_m_freem(chk->data);
6916				chk->data = NULL;
6917			}
6918			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6919		}
6920	}
6921	/* Ok that was fun, now we will drain all the inbound streams? */
6922	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6923		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6924			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6925				/* Yep it is above cum-ack */
6926				cnt++;
6927				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6928				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6929				sctp_ucount_decr(asoc->cnt_on_all_streams);
6930				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6931				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6932				if (ctl->data) {
6933					sctp_m_freem(ctl->data);
6934					ctl->data = NULL;
6935				}
6936				sctp_free_remote_addr(ctl->whoFrom);
6937				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6938				SCTP_DECR_READQ_COUNT();
6939			}
6940		}
6941	}
6942	if (cnt) {
6943		/* We must back down to see what the new highest is */
6944		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6945			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6946			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6947				asoc->highest_tsn_inside_map = i;
6948				fnd = 1;
6949				break;
6950			}
6951		}
6952		if (!fnd) {
6953			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6954		}
6955		/*
6956		 * Question, should we go through the delivery queue? The
6957		 * only reason things are on here is the app not reading OR
6958		 * a p-d-api up. An attacker COULD send enough in to
6959		 * initiate the PD-API and then send a bunch of stuff to
6960		 * other streams... these would wind up on the delivery
6961		 * queue.. and then we would not get to them. But in order
6962		 * to do this I then have to back-track and un-deliver
6963		 * sequence numbers in streams.. el-yucko. I think for now
6964		 * we will NOT look at the delivery queue and leave it to be
6965		 * something to consider later. An alternative would be to
6966		 * abort the P-D-API with a notification and then deliver
6967		 * the data.... Or another method might be to keep track of
6968		 * how many times the situation occurs and if we see a
6969		 * possible attack underway just abort the association.
6970		 */
6971#ifdef SCTP_DEBUG
6972		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6973#endif
6974		/*
6975		 * Now do we need to find a new
6976		 * asoc->highest_tsn_inside_map?
6977		 */
6978		asoc->last_revoke_count = cnt;
6979		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6980		/* sa_ignore NO_NULL_CHK */
6981		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6982		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6983	}
6984	/*
6985	 * Another issue, in un-setting the TSN's in the mapping array we
6986	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6987	 * two things to occur. It may cause us to do extra work in checking
6988	 * for our mapping array movement. More importantly it may cause us
6989	 * to SACK every datagram. This may not be a bad thing though since
6990	 * we will recover once we get our cum-ack above and all this stuff
6991	 * we dumped recovered.
6992	 */
6993}
6994
6995void
6996sctp_drain()
6997{
6998	/*
6999	 * We must walk the PCB lists for ALL associations here. The system
7000	 * is LOW on MBUF's and needs help. This is where reneging will
7001	 * occur. We really hope this does NOT happen!
7002	 */
7003	VNET_ITERATOR_DECL(vnet_iter);
7004	VNET_LIST_RLOCK_NOSLEEP();
7005	VNET_FOREACH(vnet_iter) {
7006		CURVNET_SET(vnet_iter);
7007		struct sctp_inpcb *inp;
7008		struct sctp_tcb *stcb;
7009
7010		SCTP_STAT_INCR(sctps_protocol_drain_calls);
7011		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
7012#ifdef VIMAGE
7013			continue;
7014#else
7015			return;
7016#endif
7017		}
7018		SCTP_INP_INFO_RLOCK();
7019		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
7020			/* For each endpoint */
7021			SCTP_INP_RLOCK(inp);
7022			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
7023				/* For each association */
7024				SCTP_TCB_LOCK(stcb);
7025				sctp_drain_mbufs(stcb);
7026				SCTP_TCB_UNLOCK(stcb);
7027			}
7028			SCTP_INP_RUNLOCK(inp);
7029		}
7030		SCTP_INP_INFO_RUNLOCK();
7031		CURVNET_RESTORE();
7032	}
7033	VNET_LIST_RUNLOCK_NOSLEEP();
7034}
7035
7036/*
7037 * start a new iterator
7038 * iterates through all endpoints and associations based on the pcb_state
7039 * flags and asoc_state.  "af" (mandatory) is executed for all matching
7040 * assocs and "ef" (optional) is executed when the iterator completes.
7041 * "inpf" (optional) is executed for each new endpoint as it is being
7042 * iterated through. inpe (optional) is called when the inp completes
7043 * its way through all the stcbs.
7044 */
7045int
7046sctp_initiate_iterator(inp_func inpf,
7047    asoc_func af,
7048    inp_func inpe,
7049    uint32_t pcb_state,
7050    uint32_t pcb_features,
7051    uint32_t asoc_state,
7052    void *argp,
7053    uint32_t argi,
7054    end_func ef,
7055    struct sctp_inpcb *s_inp,
7056    uint8_t chunk_output_off)
7057{
7058	struct sctp_iterator *it = NULL;
7059
7060	if (af == NULL) {
7061		return (-1);
7062	}
7063	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
7064	    SCTP_M_ITER);
7065	if (it == NULL) {
7066		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
7067		return (ENOMEM);
7068	}
7069	memset(it, 0, sizeof(*it));
7070	it->function_assoc = af;
7071	it->function_inp = inpf;
7072	if (inpf)
7073		it->done_current_ep = 0;
7074	else
7075		it->done_current_ep = 1;
7076	it->function_atend = ef;
7077	it->pointer = argp;
7078	it->val = argi;
7079	it->pcb_flags = pcb_state;
7080	it->pcb_features = pcb_features;
7081	it->asoc_state = asoc_state;
7082	it->function_inp_end = inpe;
7083	it->no_chunk_output = chunk_output_off;
7084	it->vn = curvnet;
7085	if (s_inp) {
7086		/* Assume lock is held here */
7087		it->inp = s_inp;
7088		SCTP_INP_INCR_REF(it->inp);
7089		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7090	} else {
7091		SCTP_INP_INFO_RLOCK();
7092		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7093		if (it->inp) {
7094			SCTP_INP_INCR_REF(it->inp);
7095		}
7096		SCTP_INP_INFO_RUNLOCK();
7097		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7098
7099	}
7100	SCTP_IPI_ITERATOR_WQ_LOCK();
7101
7102	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7103	if (sctp_it_ctl.iterator_running == 0) {
7104		sctp_wakeup_iterator();
7105	}
7106	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7107	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7108	return (0);
7109}
7110