sctp_pcb.c revision 270359
1/*-
2 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * a) Redistributions of source code must retain the above copyright notice,
10 *    this list of conditions and the following disclaimer.
11 *
12 * b) Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *    the documentation and/or other materials provided with the distribution.
15 *
16 * c) Neither the name of Cisco Systems, Inc. nor the names of its
17 *    contributors may be used to endorse or promote products derived
18 *    from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33#include <sys/cdefs.h>
34__FBSDID("$FreeBSD: stable/10/sys/netinet/sctp_pcb.c 270359 2014-08-22 20:05:09Z tuexen $");
35
36#include <netinet/sctp_os.h>
37#include <sys/proc.h>
38#include <netinet/sctp_var.h>
39#include <netinet/sctp_sysctl.h>
40#include <netinet/sctp_pcb.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp.h>
43#include <netinet/sctp_header.h>
44#include <netinet/sctp_asconf.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_timer.h>
47#include <netinet/sctp_bsd_addr.h>
48#include <netinet/sctp_dtrace_define.h>
49#if defined(INET) || defined(INET6)
50#include <netinet/udp.h>
51#endif
52#ifdef INET6
53#include <netinet6/ip6_var.h>
54#endif
55#include <sys/sched.h>
56#include <sys/smp.h>
57#include <sys/unistd.h>
58
59
60VNET_DEFINE(struct sctp_base_info, system_base_info);
61
62/* FIX: we don't handle multiple link local scopes */
63/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
64#ifdef INET6
65int
66SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
67{
68	struct sockaddr_in6 tmp_a, tmp_b;
69
70	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
71	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
72		return (0);
73	}
74	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
75	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
76		return (0);
77	}
78	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
79}
80
81#endif
82
83void
84sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
85{
86	/*
87	 * We really don't need to lock this, but I will just because it
88	 * does not hurt.
89	 */
90	SCTP_INP_INFO_RLOCK();
91	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
92	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
93	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
94	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
95	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
96	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
97	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
98	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
99	SCTP_INP_INFO_RUNLOCK();
100}
101
102/*-
103 * Addresses are added to VRF's (Virtual Router's). For BSD we
104 * have only the default VRF 0. We maintain a hash list of
105 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
106 * these has a list of addresses. When we add a new address
107 * to a VRF we lookup the ifn/ifn_index, if the ifn does
108 * not exist we create it and add it to the list of IFN's
109 * within the VRF. Once we have the sctp_ifn, we add the
110 * address to the list. So we look something like:
111 *
112 * hash-vrf-table
113 *   vrf-> ifn-> ifn -> ifn
114 *   vrf    |
115 *    ...   +--ifa-> ifa -> ifa
116 *   vrf
117 *
118 * We keep these separate lists since the SCTP subsystem will
119 * point to these from its source address selection nets structure.
120 * When an address is deleted it does not happen right away on
121 * the SCTP side, it gets scheduled. What we do when a
122 * delete happens is immediately remove the address from
123 * the master list and decrement the refcount. As our
124 * addip iterator works through and frees the src address
125 * selection pointing to the sctp_ifa, eventually the refcount
126 * will reach 0 and we will delete it. Note that it is assumed
127 * that any locking on system level ifn/ifa is done at the
128 * caller of these functions and these routines will only
129 * lock the SCTP structures as they add or delete things.
130 *
131 * Other notes on VRF concepts.
132 *  - An endpoint can be in multiple VRF's
133 *  - An association lives within a VRF and only one VRF.
134 *  - Any incoming packet we can deduce the VRF for by
135 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
136 *  - Any downward send call or connect call must supply the
137 *    VRF via ancillary data or via some sort of set default
138 *    VRF socket option call (again for BSD no brainer since
139 *    the VRF is always 0).
140 *  - An endpoint may add multiple VRF's to it.
141 *  - Listening sockets can accept associations in any
142 *    of the VRF's they are in but the assoc will end up
143 *    in only one VRF (gotten from the packet or connect/send).
144 *
145 */
146
147struct sctp_vrf *
148sctp_allocate_vrf(int vrf_id)
149{
150	struct sctp_vrf *vrf = NULL;
151	struct sctp_vrflist *bucket;
152
153	/* First allocate the VRF structure */
154	vrf = sctp_find_vrf(vrf_id);
155	if (vrf) {
156		/* Already allocated */
157		return (vrf);
158	}
159	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
160	    SCTP_M_VRF);
161	if (vrf == NULL) {
162		/* No memory */
163#ifdef INVARIANTS
164		panic("No memory for VRF:%d", vrf_id);
165#endif
166		return (NULL);
167	}
168	/* setup the VRF */
169	memset(vrf, 0, sizeof(struct sctp_vrf));
170	vrf->vrf_id = vrf_id;
171	LIST_INIT(&vrf->ifnlist);
172	vrf->total_ifa_count = 0;
173	vrf->refcount = 0;
174	/* now also setup table ids */
175	SCTP_INIT_VRF_TABLEID(vrf);
176	/* Init the HASH of addresses */
177	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
178	    &vrf->vrf_addr_hashmark);
179	if (vrf->vrf_addr_hash == NULL) {
180		/* No memory */
181#ifdef INVARIANTS
182		panic("No memory for VRF:%d", vrf_id);
183#endif
184		SCTP_FREE(vrf, SCTP_M_VRF);
185		return (NULL);
186	}
187	/* Add it to the hash table */
188	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
189	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
190	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
191	return (vrf);
192}
193
194
195struct sctp_ifn *
196sctp_find_ifn(void *ifn, uint32_t ifn_index)
197{
198	struct sctp_ifn *sctp_ifnp;
199	struct sctp_ifnlist *hash_ifn_head;
200
201	/*
202	 * We assume the lock is held for the addresses if that's wrong
203	 * problems could occur :-)
204	 */
205	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
206	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
207		if (sctp_ifnp->ifn_index == ifn_index) {
208			return (sctp_ifnp);
209		}
210		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
211			return (sctp_ifnp);
212		}
213	}
214	return (NULL);
215}
216
217
218struct sctp_vrf *
219sctp_find_vrf(uint32_t vrf_id)
220{
221	struct sctp_vrflist *bucket;
222	struct sctp_vrf *liste;
223
224	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
225	LIST_FOREACH(liste, bucket, next_vrf) {
226		if (vrf_id == liste->vrf_id) {
227			return (liste);
228		}
229	}
230	return (NULL);
231}
232
233
234void
235sctp_free_vrf(struct sctp_vrf *vrf)
236{
237	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
238		if (vrf->vrf_addr_hash) {
239			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
240			vrf->vrf_addr_hash = NULL;
241		}
242		/* We zero'd the count */
243		LIST_REMOVE(vrf, next_vrf);
244		SCTP_FREE(vrf, SCTP_M_VRF);
245		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
246	}
247}
248
249
250void
251sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
252{
253	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
254		/* We zero'd the count */
255		if (sctp_ifnp->vrf) {
256			sctp_free_vrf(sctp_ifnp->vrf);
257		}
258		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
259		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
260	}
261}
262
263
264void
265sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
266{
267	struct sctp_ifn *sctp_ifnp;
268
269	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
270	if (sctp_ifnp != NULL) {
271		sctp_ifnp->ifn_mtu = mtu;
272	}
273}
274
275
276void
277sctp_free_ifa(struct sctp_ifa *sctp_ifap)
278{
279	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
280		/* We zero'd the count */
281		if (sctp_ifap->ifn_p) {
282			sctp_free_ifn(sctp_ifap->ifn_p);
283		}
284		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
285		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
286	}
287}
288
289
290static void
291sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
292{
293	struct sctp_ifn *found;
294
295	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
296	if (found == NULL) {
297		/* Not in the list.. sorry */
298		return;
299	}
300	if (hold_addr_lock == 0)
301		SCTP_IPI_ADDR_WLOCK();
302	LIST_REMOVE(sctp_ifnp, next_bucket);
303	LIST_REMOVE(sctp_ifnp, next_ifn);
304	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
305	    sctp_ifnp->registered_af);
306	if (hold_addr_lock == 0)
307		SCTP_IPI_ADDR_WUNLOCK();
308	/* Take away the reference, and possibly free it */
309	sctp_free_ifn(sctp_ifnp);
310}
311
312
313void
314sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
315    const char *if_name, uint32_t ifn_index)
316{
317	struct sctp_vrf *vrf;
318	struct sctp_ifa *sctp_ifap;
319
320	SCTP_IPI_ADDR_RLOCK();
321	vrf = sctp_find_vrf(vrf_id);
322	if (vrf == NULL) {
323		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
324		goto out;
325
326	}
327	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
328	if (sctp_ifap == NULL) {
329		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
330		goto out;
331	}
332	if (sctp_ifap->ifn_p == NULL) {
333		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
334		goto out;
335	}
336	if (if_name) {
337		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
338			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
339			    sctp_ifap->ifn_p->ifn_name, if_name);
340			goto out;
341		}
342	} else {
343		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
344			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
345			    sctp_ifap->ifn_p->ifn_index, ifn_index);
346			goto out;
347		}
348	}
349
350	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
351	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
352out:
353	SCTP_IPI_ADDR_RUNLOCK();
354}
355
356
357void
358sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
359    const char *if_name, uint32_t ifn_index)
360{
361	struct sctp_vrf *vrf;
362	struct sctp_ifa *sctp_ifap;
363
364	SCTP_IPI_ADDR_RLOCK();
365	vrf = sctp_find_vrf(vrf_id);
366	if (vrf == NULL) {
367		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
368		goto out;
369
370	}
371	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
372	if (sctp_ifap == NULL) {
373		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
374		goto out;
375	}
376	if (sctp_ifap->ifn_p == NULL) {
377		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
378		goto out;
379	}
380	if (if_name) {
381		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
382			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
383			    sctp_ifap->ifn_p->ifn_name, if_name);
384			goto out;
385		}
386	} else {
387		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
388			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
389			    sctp_ifap->ifn_p->ifn_index, ifn_index);
390			goto out;
391		}
392	}
393
394	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
395	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
396out:
397	SCTP_IPI_ADDR_RUNLOCK();
398}
399
400
401/*-
402 * Add an ifa to an ifn.
403 * Register the interface as necessary.
404 * NOTE: ADDR write lock MUST be held.
405 */
406static void
407sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
408{
409	int ifa_af;
410
411	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
412	sctp_ifap->ifn_p = sctp_ifnp;
413	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
414	/* update address counts */
415	sctp_ifnp->ifa_count++;
416	ifa_af = sctp_ifap->address.sa.sa_family;
417	switch (ifa_af) {
418#ifdef INET
419	case AF_INET:
420		sctp_ifnp->num_v4++;
421		break;
422#endif
423#ifdef INET6
424	case AF_INET6:
425		sctp_ifnp->num_v6++;
426		break;
427#endif
428	default:
429		break;
430	}
431	if (sctp_ifnp->ifa_count == 1) {
432		/* register the new interface */
433		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
434		sctp_ifnp->registered_af = ifa_af;
435	}
436}
437
438
439/*-
440 * Remove an ifa from its ifn.
441 * If no more addresses exist, remove the ifn too. Otherwise, re-register
442 * the interface based on the remaining address families left.
443 * NOTE: ADDR write lock MUST be held.
444 */
445static void
446sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
447{
448	LIST_REMOVE(sctp_ifap, next_ifa);
449	if (sctp_ifap->ifn_p) {
450		/* update address counts */
451		sctp_ifap->ifn_p->ifa_count--;
452		switch (sctp_ifap->address.sa.sa_family) {
453#ifdef INET
454		case AF_INET:
455			sctp_ifap->ifn_p->num_v4--;
456			break;
457#endif
458#ifdef INET6
459		case AF_INET6:
460			sctp_ifap->ifn_p->num_v6--;
461			break;
462#endif
463		default:
464			break;
465		}
466
467		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
468			/* remove the ifn, possibly freeing it */
469			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
470		} else {
471			/* re-register address family type, if needed */
472			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
473			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
474				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
475				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
476				sctp_ifap->ifn_p->registered_af = AF_INET;
477			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
478			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
479				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
480				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
481				sctp_ifap->ifn_p->registered_af = AF_INET6;
482			}
483			/* free the ifn refcount */
484			sctp_free_ifn(sctp_ifap->ifn_p);
485		}
486		sctp_ifap->ifn_p = NULL;
487	}
488}
489
490
491struct sctp_ifa *
492sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
493    uint32_t ifn_type, const char *if_name, void *ifa,
494    struct sockaddr *addr, uint32_t ifa_flags,
495    int dynamic_add)
496{
497	struct sctp_vrf *vrf;
498	struct sctp_ifn *sctp_ifnp = NULL;
499	struct sctp_ifa *sctp_ifap = NULL;
500	struct sctp_ifalist *hash_addr_head;
501	struct sctp_ifnlist *hash_ifn_head;
502	uint32_t hash_of_addr;
503	int new_ifn_af = 0;
504
505#ifdef SCTP_DEBUG
506	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
507	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
508#endif
509	SCTP_IPI_ADDR_WLOCK();
510	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
511	if (sctp_ifnp) {
512		vrf = sctp_ifnp->vrf;
513	} else {
514		vrf = sctp_find_vrf(vrf_id);
515		if (vrf == NULL) {
516			vrf = sctp_allocate_vrf(vrf_id);
517			if (vrf == NULL) {
518				SCTP_IPI_ADDR_WUNLOCK();
519				return (NULL);
520			}
521		}
522	}
523	if (sctp_ifnp == NULL) {
524		/*
525		 * build one and add it, can't hold lock until after malloc
526		 * done though.
527		 */
528		SCTP_IPI_ADDR_WUNLOCK();
529		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
530		    sizeof(struct sctp_ifn), SCTP_M_IFN);
531		if (sctp_ifnp == NULL) {
532#ifdef INVARIANTS
533			panic("No memory for IFN");
534#endif
535			return (NULL);
536		}
537		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
538		sctp_ifnp->ifn_index = ifn_index;
539		sctp_ifnp->ifn_p = ifn;
540		sctp_ifnp->ifn_type = ifn_type;
541		sctp_ifnp->refcount = 0;
542		sctp_ifnp->vrf = vrf;
543		atomic_add_int(&vrf->refcount, 1);
544		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
545		if (if_name != NULL) {
546			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
547		} else {
548			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
549		}
550		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
551		LIST_INIT(&sctp_ifnp->ifalist);
552		SCTP_IPI_ADDR_WLOCK();
553		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
554		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
555		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
556		new_ifn_af = 1;
557	}
558	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
559	if (sctp_ifap) {
560		/* Hmm, it already exists? */
561		if ((sctp_ifap->ifn_p) &&
562		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
563			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
564			    sctp_ifap->ifn_p->ifn_name, ifn_index,
565			    (void *)sctp_ifap);
566			if (new_ifn_af) {
567				/* Remove the created one that we don't want */
568				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
569			}
570			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
571				/* easy to solve, just switch back to active */
572				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
573				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
574				sctp_ifap->ifn_p = sctp_ifnp;
575				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
576			}
577	exit_stage_left:
578			SCTP_IPI_ADDR_WUNLOCK();
579			return (sctp_ifap);
580		} else {
581			if (sctp_ifap->ifn_p) {
582				/*
583				 * The last IFN gets the address, remove the
584				 * old one
585				 */
586				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
587				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
588				    sctp_ifap->ifn_p->ifn_index, if_name,
589				    ifn_index);
590				/* remove the address from the old ifn */
591				sctp_remove_ifa_from_ifn(sctp_ifap);
592				/* move the address over to the new ifn */
593				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
594				goto exit_stage_left;
595			} else {
596				/* repair ifnp which was NULL ? */
597				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
598				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
599				    (void *)sctp_ifnp, (void *)sctp_ifap);
600				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
601			}
602			goto exit_stage_left;
603		}
604	}
605	SCTP_IPI_ADDR_WUNLOCK();
606	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
607	if (sctp_ifap == NULL) {
608#ifdef INVARIANTS
609		panic("No memory for IFA");
610#endif
611		return (NULL);
612	}
613	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
614	sctp_ifap->ifn_p = sctp_ifnp;
615	atomic_add_int(&sctp_ifnp->refcount, 1);
616	sctp_ifap->vrf_id = vrf_id;
617	sctp_ifap->ifa = ifa;
618	memcpy(&sctp_ifap->address, addr, addr->sa_len);
619	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
620	sctp_ifap->flags = ifa_flags;
621	/* Set scope */
622	switch (sctp_ifap->address.sa.sa_family) {
623#ifdef INET
624	case AF_INET:
625		{
626			struct sockaddr_in *sin;
627
628			sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
629			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
630			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
631				sctp_ifap->src_is_loop = 1;
632			}
633			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
634				sctp_ifap->src_is_priv = 1;
635			}
636			sctp_ifnp->num_v4++;
637			if (new_ifn_af)
638				new_ifn_af = AF_INET;
639			break;
640		}
641#endif
642#ifdef INET6
643	case AF_INET6:
644		{
645			/* ok to use deprecated addresses? */
646			struct sockaddr_in6 *sin6;
647
648			sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
649			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
650			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
651				sctp_ifap->src_is_loop = 1;
652			}
653			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
654				sctp_ifap->src_is_priv = 1;
655			}
656			sctp_ifnp->num_v6++;
657			if (new_ifn_af)
658				new_ifn_af = AF_INET6;
659			break;
660		}
661#endif
662	default:
663		new_ifn_af = 0;
664		break;
665	}
666	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
667
668	if ((sctp_ifap->src_is_priv == 0) &&
669	    (sctp_ifap->src_is_loop == 0)) {
670		sctp_ifap->src_is_glob = 1;
671	}
672	SCTP_IPI_ADDR_WLOCK();
673	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
674	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
675	sctp_ifap->refcount = 1;
676	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
677	sctp_ifnp->ifa_count++;
678	vrf->total_ifa_count++;
679	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
680	if (new_ifn_af) {
681		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
682		sctp_ifnp->registered_af = new_ifn_af;
683	}
684	SCTP_IPI_ADDR_WUNLOCK();
685	if (dynamic_add) {
686		/*
687		 * Bump up the refcount so that when the timer completes it
688		 * will drop back down.
689		 */
690		struct sctp_laddr *wi;
691
692		atomic_add_int(&sctp_ifap->refcount, 1);
693		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
694		if (wi == NULL) {
695			/*
696			 * Gak, what can we do? We have lost an address
697			 * change can you say HOSED?
698			 */
699			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
700			/* Opps, must decrement the count */
701			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
702			    if_name);
703			return (NULL);
704		}
705		SCTP_INCR_LADDR_COUNT();
706		bzero(wi, sizeof(*wi));
707		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
708		wi->ifa = sctp_ifap;
709		wi->action = SCTP_ADD_IP_ADDRESS;
710
711		SCTP_WQ_ADDR_LOCK();
712		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
713		SCTP_WQ_ADDR_UNLOCK();
714
715		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
716		    (struct sctp_inpcb *)NULL,
717		    (struct sctp_tcb *)NULL,
718		    (struct sctp_nets *)NULL);
719	} else {
720		/* it's ready for use */
721		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
722	}
723	return (sctp_ifap);
724}
725
726void
727sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
728    uint32_t ifn_index, const char *if_name)
729{
730	struct sctp_vrf *vrf;
731	struct sctp_ifa *sctp_ifap = NULL;
732
733	SCTP_IPI_ADDR_WLOCK();
734	vrf = sctp_find_vrf(vrf_id);
735	if (vrf == NULL) {
736		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
737		goto out_now;
738	}
739#ifdef SCTP_DEBUG
740	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
741	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
742#endif
743	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
744	if (sctp_ifap) {
745		/* Validate the delete */
746		if (sctp_ifap->ifn_p) {
747			int valid = 0;
748
749			/*-
750			 * The name has priority over the ifn_index
751			 * if its given. We do this especially for
752			 * panda who might recycle indexes fast.
753			 */
754			if (if_name) {
755				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
756					/* They match its a correct delete */
757					valid = 1;
758				}
759			}
760			if (!valid) {
761				/* last ditch check ifn_index */
762				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
763					valid = 1;
764				}
765			}
766			if (!valid) {
767				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
768				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
769				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
770				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
771				SCTP_IPI_ADDR_WUNLOCK();
772				return;
773			}
774		}
775		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
776		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
777		/*
778		 * We don't set the flag. This means that the structure will
779		 * hang around in EP's that have bound specific to it until
780		 * they close. This gives us TCP like behavior if someone
781		 * removes an address (or for that matter adds it right
782		 * back).
783		 */
784		/* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */
785		vrf->total_ifa_count--;
786		LIST_REMOVE(sctp_ifap, next_bucket);
787		sctp_remove_ifa_from_ifn(sctp_ifap);
788	}
789#ifdef SCTP_DEBUG
790	else {
791		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
792		    ifn_index);
793		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
794	}
795#endif
796
797out_now:
798	SCTP_IPI_ADDR_WUNLOCK();
799	if (sctp_ifap) {
800		struct sctp_laddr *wi;
801
802		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
803		if (wi == NULL) {
804			/*
805			 * Gak, what can we do? We have lost an address
806			 * change can you say HOSED?
807			 */
808			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
809
810			/* Oops, must decrement the count */
811			sctp_free_ifa(sctp_ifap);
812			return;
813		}
814		SCTP_INCR_LADDR_COUNT();
815		bzero(wi, sizeof(*wi));
816		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
817		wi->ifa = sctp_ifap;
818		wi->action = SCTP_DEL_IP_ADDRESS;
819		SCTP_WQ_ADDR_LOCK();
820		/*
821		 * Should this really be a tailq? As it is we will process
822		 * the newest first :-0
823		 */
824		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
825		SCTP_WQ_ADDR_UNLOCK();
826
827		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
828		    (struct sctp_inpcb *)NULL,
829		    (struct sctp_tcb *)NULL,
830		    (struct sctp_nets *)NULL);
831	}
832	return;
833}
834
835
836static int
837sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
838{
839	int loopback_scope;
840
841#if defined(INET)
842	int ipv4_local_scope, ipv4_addr_legal;
843
844#endif
845#if defined(INET6)
846	int local_scope, site_scope, ipv6_addr_legal;
847
848#endif
849	struct sctp_vrf *vrf;
850	struct sctp_ifn *sctp_ifn;
851	struct sctp_ifa *sctp_ifa;
852
853	loopback_scope = stcb->asoc.scope.loopback_scope;
854#if defined(INET)
855	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
856	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
857#endif
858#if defined(INET6)
859	local_scope = stcb->asoc.scope.local_scope;
860	site_scope = stcb->asoc.scope.site_scope;
861	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
862#endif
863
864	SCTP_IPI_ADDR_RLOCK();
865	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
866	if (vrf == NULL) {
867		/* no vrf, no addresses */
868		SCTP_IPI_ADDR_RUNLOCK();
869		return (0);
870	}
871	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
872		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
873			if ((loopback_scope == 0) &&
874			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
875				continue;
876			}
877			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
878				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
879				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
880					/*
881					 * We allow pending addresses, where
882					 * we have sent an asconf-add to be
883					 * considered valid.
884					 */
885					continue;
886				}
887				if (sctp_ifa->address.sa.sa_family != to->sa_family) {
888					continue;
889				}
890				switch (sctp_ifa->address.sa.sa_family) {
891#ifdef INET
892				case AF_INET:
893					if (ipv4_addr_legal) {
894						struct sockaddr_in *sin,
895						           *rsin;
896
897						sin = &sctp_ifa->address.sin;
898						rsin = (struct sockaddr_in *)to;
899						if ((ipv4_local_scope == 0) &&
900						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
901							continue;
902						}
903						if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred,
904						    &sin->sin_addr) != 0) {
905							continue;
906						}
907						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
908							SCTP_IPI_ADDR_RUNLOCK();
909							return (1);
910						}
911					}
912					break;
913#endif
914#ifdef INET6
915				case AF_INET6:
916					if (ipv6_addr_legal) {
917						struct sockaddr_in6 *sin6,
918						            *rsin6;
919
920						sin6 = &sctp_ifa->address.sin6;
921						rsin6 = (struct sockaddr_in6 *)to;
922						if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred,
923						    &sin6->sin6_addr) != 0) {
924							continue;
925						}
926						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
927							if (local_scope == 0)
928								continue;
929							if (sin6->sin6_scope_id == 0) {
930								if (sa6_recoverscope(sin6) != 0)
931									continue;
932							}
933						}
934						if ((site_scope == 0) &&
935						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
936							continue;
937						}
938						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
939							SCTP_IPI_ADDR_RUNLOCK();
940							return (1);
941						}
942					}
943					break;
944#endif
945				default:
946					/* TSNH */
947					break;
948				}
949			}
950		}
951	} else {
952		struct sctp_laddr *laddr;
953
954		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
955			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
956				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
957				continue;
958			}
959			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
960			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
961				/*
962				 * We allow pending addresses, where we have
963				 * sent an asconf-add to be considered
964				 * valid.
965				 */
966				continue;
967			}
968			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
969				continue;
970			}
971			switch (to->sa_family) {
972#ifdef INET
973			case AF_INET:
974				{
975					struct sockaddr_in *sin, *rsin;
976
977					sin = (struct sockaddr_in *)&laddr->ifa->address.sin;
978					rsin = (struct sockaddr_in *)to;
979					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
980						SCTP_IPI_ADDR_RUNLOCK();
981						return (1);
982					}
983					break;
984				}
985#endif
986#ifdef INET6
987			case AF_INET6:
988				{
989					struct sockaddr_in6 *sin6, *rsin6;
990
991					sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
992					rsin6 = (struct sockaddr_in6 *)to;
993					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
994						SCTP_IPI_ADDR_RUNLOCK();
995						return (1);
996					}
997					break;
998				}
999
1000#endif
1001			default:
1002				/* TSNH */
1003				break;
1004			}
1005
1006		}
1007	}
1008	SCTP_IPI_ADDR_RUNLOCK();
1009	return (0);
1010}
1011
1012
1013static struct sctp_tcb *
1014sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
1015    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
1016{
1017	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
1018	/*
1019	 * If we support the TCP model, then we must now dig through to see
1020	 * if we can find our endpoint in the list of tcp ep's.
1021	 */
1022	uint16_t lport, rport;
1023	struct sctppcbhead *ephead;
1024	struct sctp_inpcb *inp;
1025	struct sctp_laddr *laddr;
1026	struct sctp_tcb *stcb;
1027	struct sctp_nets *net;
1028
1029	if ((to == NULL) || (from == NULL)) {
1030		return (NULL);
1031	}
1032	switch (to->sa_family) {
1033#ifdef INET
1034	case AF_INET:
1035		if (from->sa_family == AF_INET) {
1036			lport = ((struct sockaddr_in *)to)->sin_port;
1037			rport = ((struct sockaddr_in *)from)->sin_port;
1038		} else {
1039			return (NULL);
1040		}
1041		break;
1042#endif
1043#ifdef INET6
1044	case AF_INET6:
1045		if (from->sa_family == AF_INET6) {
1046			lport = ((struct sockaddr_in6 *)to)->sin6_port;
1047			rport = ((struct sockaddr_in6 *)from)->sin6_port;
1048		} else {
1049			return (NULL);
1050		}
1051		break;
1052#endif
1053	default:
1054		return (NULL);
1055	}
1056	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
1057	/*
1058	 * Ok now for each of the guys in this bucket we must look and see:
1059	 * - Does the remote port match. - Does there single association's
1060	 * addresses match this address (to). If so we update p_ep to point
1061	 * to this ep and return the tcb from it.
1062	 */
1063	LIST_FOREACH(inp, ephead, sctp_hash) {
1064		SCTP_INP_RLOCK(inp);
1065		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1066			SCTP_INP_RUNLOCK(inp);
1067			continue;
1068		}
1069		if (lport != inp->sctp_lport) {
1070			SCTP_INP_RUNLOCK(inp);
1071			continue;
1072		}
1073		switch (to->sa_family) {
1074#ifdef INET
1075		case AF_INET:
1076			{
1077				struct sockaddr_in *sin;
1078
1079				sin = (struct sockaddr_in *)to;
1080				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1081				    &sin->sin_addr) != 0) {
1082					SCTP_INP_RUNLOCK(inp);
1083					continue;
1084				}
1085				break;
1086			}
1087#endif
1088#ifdef INET6
1089		case AF_INET6:
1090			{
1091				struct sockaddr_in6 *sin6;
1092
1093				sin6 = (struct sockaddr_in6 *)to;
1094				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1095				    &sin6->sin6_addr) != 0) {
1096					SCTP_INP_RUNLOCK(inp);
1097					continue;
1098				}
1099				break;
1100			}
1101#endif
1102		default:
1103			SCTP_INP_RUNLOCK(inp);
1104			continue;
1105		}
1106		if (inp->def_vrf_id != vrf_id) {
1107			SCTP_INP_RUNLOCK(inp);
1108			continue;
1109		}
1110		/* check to see if the ep has one of the addresses */
1111		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1112			/* We are NOT bound all, so look further */
1113			int match = 0;
1114
1115			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1116
1117				if (laddr->ifa == NULL) {
1118					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
1119					continue;
1120				}
1121				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1122					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1123					continue;
1124				}
1125				if (laddr->ifa->address.sa.sa_family ==
1126				    to->sa_family) {
1127					/* see if it matches */
1128#ifdef INET
1129					if (from->sa_family == AF_INET) {
1130						struct sockaddr_in *intf_addr,
1131						           *sin;
1132
1133						intf_addr = &laddr->ifa->address.sin;
1134						sin = (struct sockaddr_in *)to;
1135						if (sin->sin_addr.s_addr ==
1136						    intf_addr->sin_addr.s_addr) {
1137							match = 1;
1138							break;
1139						}
1140					}
1141#endif
1142#ifdef INET6
1143					if (from->sa_family == AF_INET6) {
1144						struct sockaddr_in6 *intf_addr6;
1145						struct sockaddr_in6 *sin6;
1146
1147						sin6 = (struct sockaddr_in6 *)
1148						    to;
1149						intf_addr6 = &laddr->ifa->address.sin6;
1150
1151						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1152						    intf_addr6)) {
1153							match = 1;
1154							break;
1155						}
1156					}
1157#endif
1158				}
1159			}
1160			if (match == 0) {
1161				/* This endpoint does not have this address */
1162				SCTP_INP_RUNLOCK(inp);
1163				continue;
1164			}
1165		}
1166		/*
1167		 * Ok if we hit here the ep has the address, does it hold
1168		 * the tcb?
1169		 */
1170		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1171		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1172		if (stcb == NULL) {
1173			SCTP_INP_RUNLOCK(inp);
1174			continue;
1175		}
1176		SCTP_TCB_LOCK(stcb);
1177		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1178			SCTP_TCB_UNLOCK(stcb);
1179			SCTP_INP_RUNLOCK(inp);
1180			continue;
1181		}
1182		if (stcb->rport != rport) {
1183			/* remote port does not match. */
1184			SCTP_TCB_UNLOCK(stcb);
1185			SCTP_INP_RUNLOCK(inp);
1186			continue;
1187		}
1188		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1189			SCTP_TCB_UNLOCK(stcb);
1190			SCTP_INP_RUNLOCK(inp);
1191			continue;
1192		}
1193		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1194			SCTP_TCB_UNLOCK(stcb);
1195			SCTP_INP_RUNLOCK(inp);
1196			continue;
1197		}
1198		/* Does this TCB have a matching address? */
1199		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1200
1201			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1202				/* not the same family, can't be a match */
1203				continue;
1204			}
1205			switch (from->sa_family) {
1206#ifdef INET
1207			case AF_INET:
1208				{
1209					struct sockaddr_in *sin, *rsin;
1210
1211					sin = (struct sockaddr_in *)&net->ro._l_addr;
1212					rsin = (struct sockaddr_in *)from;
1213					if (sin->sin_addr.s_addr ==
1214					    rsin->sin_addr.s_addr) {
1215						/* found it */
1216						if (netp != NULL) {
1217							*netp = net;
1218						}
1219						/*
1220						 * Update the endpoint
1221						 * pointer
1222						 */
1223						*inp_p = inp;
1224						SCTP_INP_RUNLOCK(inp);
1225						return (stcb);
1226					}
1227					break;
1228				}
1229#endif
1230#ifdef INET6
1231			case AF_INET6:
1232				{
1233					struct sockaddr_in6 *sin6, *rsin6;
1234
1235					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1236					rsin6 = (struct sockaddr_in6 *)from;
1237					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1238					    rsin6)) {
1239						/* found it */
1240						if (netp != NULL) {
1241							*netp = net;
1242						}
1243						/*
1244						 * Update the endpoint
1245						 * pointer
1246						 */
1247						*inp_p = inp;
1248						SCTP_INP_RUNLOCK(inp);
1249						return (stcb);
1250					}
1251					break;
1252				}
1253#endif
1254			default:
1255				/* TSNH */
1256				break;
1257			}
1258		}
1259		SCTP_TCB_UNLOCK(stcb);
1260		SCTP_INP_RUNLOCK(inp);
1261	}
1262	return (NULL);
1263}
1264
1265
1266/*
1267 * rules for use
1268 *
1269 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1270 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1271 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1272 * NULL.
1273 */
1274
1275struct sctp_tcb *
1276sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1277    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1278{
1279	struct sctpasochead *head;
1280	struct sctp_inpcb *inp;
1281	struct sctp_tcb *stcb = NULL;
1282	struct sctp_nets *net;
1283	uint16_t rport;
1284
1285	inp = *inp_p;
1286	switch (remote->sa_family) {
1287#ifdef INET
1288	case AF_INET:
1289		rport = (((struct sockaddr_in *)remote)->sin_port);
1290		break;
1291#endif
1292#ifdef INET6
1293	case AF_INET6:
1294		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1295		break;
1296#endif
1297	default:
1298		return (NULL);
1299	}
1300	if (locked_tcb) {
1301		/*
1302		 * UN-lock so we can do proper locking here this occurs when
1303		 * called from load_addresses_from_init.
1304		 */
1305		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1306		SCTP_TCB_UNLOCK(locked_tcb);
1307	}
1308	SCTP_INP_INFO_RLOCK();
1309	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1310	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1311		/*-
1312		 * Now either this guy is our listener or it's the
1313		 * connector. If it is the one that issued the connect, then
1314		 * it's only chance is to be the first TCB in the list. If
1315		 * it is the acceptor, then do the special_lookup to hash
1316		 * and find the real inp.
1317		 */
1318		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1319			/* to is peer addr, from is my addr */
1320			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1321			    netp, inp->def_vrf_id);
1322			if ((stcb != NULL) && (locked_tcb == NULL)) {
1323				/* we have a locked tcb, lower refcount */
1324				SCTP_INP_DECR_REF(inp);
1325			}
1326			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1327				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1328				SCTP_TCB_LOCK(locked_tcb);
1329				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1330				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1331			}
1332			SCTP_INP_INFO_RUNLOCK();
1333			return (stcb);
1334		} else {
1335			SCTP_INP_WLOCK(inp);
1336			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1337				goto null_return;
1338			}
1339			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1340			if (stcb == NULL) {
1341				goto null_return;
1342			}
1343			SCTP_TCB_LOCK(stcb);
1344
1345			if (stcb->rport != rport) {
1346				/* remote port does not match. */
1347				SCTP_TCB_UNLOCK(stcb);
1348				goto null_return;
1349			}
1350			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1351				SCTP_TCB_UNLOCK(stcb);
1352				goto null_return;
1353			}
1354			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1355				SCTP_TCB_UNLOCK(stcb);
1356				goto null_return;
1357			}
1358			/* now look at the list of remote addresses */
1359			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1360#ifdef INVARIANTS
1361				if (net == (TAILQ_NEXT(net, sctp_next))) {
1362					panic("Corrupt net list");
1363				}
1364#endif
1365				if (net->ro._l_addr.sa.sa_family !=
1366				    remote->sa_family) {
1367					/* not the same family */
1368					continue;
1369				}
1370				switch (remote->sa_family) {
1371#ifdef INET
1372				case AF_INET:
1373					{
1374						struct sockaddr_in *sin,
1375						           *rsin;
1376
1377						sin = (struct sockaddr_in *)
1378						    &net->ro._l_addr;
1379						rsin = (struct sockaddr_in *)remote;
1380						if (sin->sin_addr.s_addr ==
1381						    rsin->sin_addr.s_addr) {
1382							/* found it */
1383							if (netp != NULL) {
1384								*netp = net;
1385							}
1386							if (locked_tcb == NULL) {
1387								SCTP_INP_DECR_REF(inp);
1388							} else if (locked_tcb != stcb) {
1389								SCTP_TCB_LOCK(locked_tcb);
1390							}
1391							if (locked_tcb) {
1392								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1393							}
1394							SCTP_INP_WUNLOCK(inp);
1395							SCTP_INP_INFO_RUNLOCK();
1396							return (stcb);
1397						}
1398						break;
1399					}
1400#endif
1401#ifdef INET6
1402				case AF_INET6:
1403					{
1404						struct sockaddr_in6 *sin6,
1405						            *rsin6;
1406
1407						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1408						rsin6 = (struct sockaddr_in6 *)remote;
1409						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1410						    rsin6)) {
1411							/* found it */
1412							if (netp != NULL) {
1413								*netp = net;
1414							}
1415							if (locked_tcb == NULL) {
1416								SCTP_INP_DECR_REF(inp);
1417							} else if (locked_tcb != stcb) {
1418								SCTP_TCB_LOCK(locked_tcb);
1419							}
1420							if (locked_tcb) {
1421								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1422							}
1423							SCTP_INP_WUNLOCK(inp);
1424							SCTP_INP_INFO_RUNLOCK();
1425							return (stcb);
1426						}
1427						break;
1428					}
1429#endif
1430				default:
1431					/* TSNH */
1432					break;
1433				}
1434			}
1435			SCTP_TCB_UNLOCK(stcb);
1436		}
1437	} else {
1438		SCTP_INP_WLOCK(inp);
1439		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1440			goto null_return;
1441		}
1442		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1443		    inp->sctp_hashmark)];
1444		if (head == NULL) {
1445			goto null_return;
1446		}
1447		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1448			if (stcb->rport != rport) {
1449				/* remote port does not match */
1450				continue;
1451			}
1452			SCTP_TCB_LOCK(stcb);
1453			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1454				SCTP_TCB_UNLOCK(stcb);
1455				continue;
1456			}
1457			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1458				SCTP_TCB_UNLOCK(stcb);
1459				continue;
1460			}
1461			/* now look at the list of remote addresses */
1462			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1463#ifdef INVARIANTS
1464				if (net == (TAILQ_NEXT(net, sctp_next))) {
1465					panic("Corrupt net list");
1466				}
1467#endif
1468				if (net->ro._l_addr.sa.sa_family !=
1469				    remote->sa_family) {
1470					/* not the same family */
1471					continue;
1472				}
1473				switch (remote->sa_family) {
1474#ifdef INET
1475				case AF_INET:
1476					{
1477						struct sockaddr_in *sin,
1478						           *rsin;
1479
1480						sin = (struct sockaddr_in *)
1481						    &net->ro._l_addr;
1482						rsin = (struct sockaddr_in *)remote;
1483						if (sin->sin_addr.s_addr ==
1484						    rsin->sin_addr.s_addr) {
1485							/* found it */
1486							if (netp != NULL) {
1487								*netp = net;
1488							}
1489							if (locked_tcb == NULL) {
1490								SCTP_INP_DECR_REF(inp);
1491							} else if (locked_tcb != stcb) {
1492								SCTP_TCB_LOCK(locked_tcb);
1493							}
1494							if (locked_tcb) {
1495								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1496							}
1497							SCTP_INP_WUNLOCK(inp);
1498							SCTP_INP_INFO_RUNLOCK();
1499							return (stcb);
1500						}
1501						break;
1502					}
1503#endif
1504#ifdef INET6
1505				case AF_INET6:
1506					{
1507						struct sockaddr_in6 *sin6,
1508						            *rsin6;
1509
1510						sin6 = (struct sockaddr_in6 *)
1511						    &net->ro._l_addr;
1512						rsin6 = (struct sockaddr_in6 *)remote;
1513						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1514						    rsin6)) {
1515							/* found it */
1516							if (netp != NULL) {
1517								*netp = net;
1518							}
1519							if (locked_tcb == NULL) {
1520								SCTP_INP_DECR_REF(inp);
1521							} else if (locked_tcb != stcb) {
1522								SCTP_TCB_LOCK(locked_tcb);
1523							}
1524							if (locked_tcb) {
1525								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1526							}
1527							SCTP_INP_WUNLOCK(inp);
1528							SCTP_INP_INFO_RUNLOCK();
1529							return (stcb);
1530						}
1531						break;
1532					}
1533#endif
1534				default:
1535					/* TSNH */
1536					break;
1537				}
1538			}
1539			SCTP_TCB_UNLOCK(stcb);
1540		}
1541	}
1542null_return:
1543	/* clean up for returning null */
1544	if (locked_tcb) {
1545		SCTP_TCB_LOCK(locked_tcb);
1546		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1547	}
1548	SCTP_INP_WUNLOCK(inp);
1549	SCTP_INP_INFO_RUNLOCK();
1550	/* not found */
1551	return (NULL);
1552}
1553
1554
1555/*
1556 * Find an association for a specific endpoint using the association id given
1557 * out in the COMM_UP notification
1558 */
1559struct sctp_tcb *
1560sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1561{
1562	/*
1563	 * Use my the assoc_id to find a endpoint
1564	 */
1565	struct sctpasochead *head;
1566	struct sctp_tcb *stcb;
1567	uint32_t id;
1568
1569	if (inp == NULL) {
1570		SCTP_PRINTF("TSNH ep_associd\n");
1571		return (NULL);
1572	}
1573	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1574		SCTP_PRINTF("TSNH ep_associd0\n");
1575		return (NULL);
1576	}
1577	id = (uint32_t) asoc_id;
1578	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1579	if (head == NULL) {
1580		/* invalid id TSNH */
1581		SCTP_PRINTF("TSNH ep_associd1\n");
1582		return (NULL);
1583	}
1584	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1585		if (stcb->asoc.assoc_id == id) {
1586			if (inp != stcb->sctp_ep) {
1587				/*
1588				 * some other guy has the same id active (id
1589				 * collision ??).
1590				 */
1591				SCTP_PRINTF("TSNH ep_associd2\n");
1592				continue;
1593			}
1594			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1595				continue;
1596			}
1597			if (want_lock) {
1598				SCTP_TCB_LOCK(stcb);
1599			}
1600			return (stcb);
1601		}
1602	}
1603	return (NULL);
1604}
1605
1606
1607struct sctp_tcb *
1608sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1609{
1610	struct sctp_tcb *stcb;
1611
1612	SCTP_INP_RLOCK(inp);
1613	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1614	SCTP_INP_RUNLOCK(inp);
1615	return (stcb);
1616}
1617
1618
1619/*
1620 * Endpoint probe expects that the INP_INFO is locked.
1621 */
1622static struct sctp_inpcb *
1623sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1624    uint16_t lport, uint32_t vrf_id)
1625{
1626	struct sctp_inpcb *inp;
1627	struct sctp_laddr *laddr;
1628
1629#ifdef INET
1630	struct sockaddr_in *sin;
1631
1632#endif
1633#ifdef INET6
1634	struct sockaddr_in6 *sin6;
1635	struct sockaddr_in6 *intf_addr6;
1636
1637#endif
1638	int fnd;
1639
1640#ifdef INET
1641	sin = NULL;
1642#endif
1643#ifdef INET6
1644	sin6 = NULL;
1645#endif
1646	switch (nam->sa_family) {
1647#ifdef INET
1648	case AF_INET:
1649		sin = (struct sockaddr_in *)nam;
1650		break;
1651#endif
1652#ifdef INET6
1653	case AF_INET6:
1654		sin6 = (struct sockaddr_in6 *)nam;
1655		break;
1656#endif
1657	default:
1658		/* unsupported family */
1659		return (NULL);
1660	}
1661
1662	if (head == NULL)
1663		return (NULL);
1664
1665	LIST_FOREACH(inp, head, sctp_hash) {
1666		SCTP_INP_RLOCK(inp);
1667		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1668			SCTP_INP_RUNLOCK(inp);
1669			continue;
1670		}
1671		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1672		    (inp->sctp_lport == lport)) {
1673			/* got it */
1674			switch (nam->sa_family) {
1675#ifdef INET
1676			case AF_INET:
1677				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1678				    SCTP_IPV6_V6ONLY(inp)) {
1679					/*
1680					 * IPv4 on a IPv6 socket with ONLY
1681					 * IPv6 set
1682					 */
1683					SCTP_INP_RUNLOCK(inp);
1684					continue;
1685				}
1686				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1687				    &sin->sin_addr) != 0) {
1688					SCTP_INP_RUNLOCK(inp);
1689					continue;
1690				}
1691				break;
1692#endif
1693#ifdef INET6
1694			case AF_INET6:
1695				/*
1696				 * A V6 address and the endpoint is NOT
1697				 * bound V6
1698				 */
1699				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1700					SCTP_INP_RUNLOCK(inp);
1701					continue;
1702				}
1703				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1704				    &sin6->sin6_addr) != 0) {
1705					SCTP_INP_RUNLOCK(inp);
1706					continue;
1707				}
1708				break;
1709#endif
1710			default:
1711				break;
1712			}
1713			/* does a VRF id match? */
1714			fnd = 0;
1715			if (inp->def_vrf_id == vrf_id)
1716				fnd = 1;
1717
1718			SCTP_INP_RUNLOCK(inp);
1719			if (!fnd)
1720				continue;
1721			return (inp);
1722		}
1723		SCTP_INP_RUNLOCK(inp);
1724	}
1725	switch (nam->sa_family) {
1726#ifdef INET
1727	case AF_INET:
1728		if (sin->sin_addr.s_addr == INADDR_ANY) {
1729			/* Can't hunt for one that has no address specified */
1730			return (NULL);
1731		}
1732		break;
1733#endif
1734#ifdef INET6
1735	case AF_INET6:
1736		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1737			/* Can't hunt for one that has no address specified */
1738			return (NULL);
1739		}
1740		break;
1741#endif
1742	default:
1743		break;
1744	}
1745	/*
1746	 * ok, not bound to all so see if we can find a EP bound to this
1747	 * address.
1748	 */
1749	LIST_FOREACH(inp, head, sctp_hash) {
1750		SCTP_INP_RLOCK(inp);
1751		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1752			SCTP_INP_RUNLOCK(inp);
1753			continue;
1754		}
1755		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1756			SCTP_INP_RUNLOCK(inp);
1757			continue;
1758		}
1759		/*
1760		 * Ok this could be a likely candidate, look at all of its
1761		 * addresses
1762		 */
1763		if (inp->sctp_lport != lport) {
1764			SCTP_INP_RUNLOCK(inp);
1765			continue;
1766		}
1767		/* does a VRF id match? */
1768		fnd = 0;
1769		if (inp->def_vrf_id == vrf_id)
1770			fnd = 1;
1771
1772		if (!fnd) {
1773			SCTP_INP_RUNLOCK(inp);
1774			continue;
1775		}
1776		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1777			if (laddr->ifa == NULL) {
1778				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1779				    __FUNCTION__);
1780				continue;
1781			}
1782			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1783			    (void *)laddr->ifa);
1784			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1785				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1786				continue;
1787			}
1788			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1789				/* possible, see if it matches */
1790				switch (nam->sa_family) {
1791#ifdef INET
1792				case AF_INET:
1793					if (sin->sin_addr.s_addr ==
1794					    laddr->ifa->address.sin.sin_addr.s_addr) {
1795						SCTP_INP_RUNLOCK(inp);
1796						return (inp);
1797					}
1798					break;
1799#endif
1800#ifdef INET6
1801				case AF_INET6:
1802					intf_addr6 = &laddr->ifa->address.sin6;
1803					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1804					    intf_addr6)) {
1805						SCTP_INP_RUNLOCK(inp);
1806						return (inp);
1807					}
1808					break;
1809#endif
1810				}
1811			}
1812		}
1813		SCTP_INP_RUNLOCK(inp);
1814	}
1815	return (NULL);
1816}
1817
1818
1819static struct sctp_inpcb *
1820sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1821{
1822	struct sctppcbhead *head;
1823	struct sctp_inpcb *t_inp;
1824	int fnd;
1825
1826	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1827	    SCTP_BASE_INFO(hashmark))];
1828	LIST_FOREACH(t_inp, head, sctp_hash) {
1829		if (t_inp->sctp_lport != lport) {
1830			continue;
1831		}
1832		/* is it in the VRF in question */
1833		fnd = 0;
1834		if (t_inp->def_vrf_id == vrf_id)
1835			fnd = 1;
1836		if (!fnd)
1837			continue;
1838
1839		/* This one is in use. */
1840		/* check the v6/v4 binding issue */
1841		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1842		    SCTP_IPV6_V6ONLY(t_inp)) {
1843			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1844				/* collision in V6 space */
1845				return (t_inp);
1846			} else {
1847				/* inp is BOUND_V4 no conflict */
1848				continue;
1849			}
1850		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1851			/* t_inp is bound v4 and v6, conflict always */
1852			return (t_inp);
1853		} else {
1854			/* t_inp is bound only V4 */
1855			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1856			    SCTP_IPV6_V6ONLY(inp)) {
1857				/* no conflict */
1858				continue;
1859			}
1860			/* else fall through to conflict */
1861		}
1862		return (t_inp);
1863	}
1864	return (NULL);
1865}
1866
1867
1868int
1869sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1870{
1871	/* For 1-2-1 with port reuse */
1872	struct sctppcbhead *head;
1873	struct sctp_inpcb *tinp;
1874
1875	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1876		/* only works with port reuse on */
1877		return (-1);
1878	}
1879	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1880		return (0);
1881	}
1882	SCTP_INP_RUNLOCK(inp);
1883	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1884	    SCTP_BASE_INFO(hashmark))];
1885	/* Kick out all non-listeners to the TCP hash */
1886	LIST_FOREACH(tinp, head, sctp_hash) {
1887		if (tinp->sctp_lport != inp->sctp_lport) {
1888			continue;
1889		}
1890		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1891			continue;
1892		}
1893		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1894			continue;
1895		}
1896		if (tinp->sctp_socket->so_qlimit) {
1897			continue;
1898		}
1899		SCTP_INP_WLOCK(tinp);
1900		LIST_REMOVE(tinp, sctp_hash);
1901		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1902		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1903		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1904		SCTP_INP_WUNLOCK(tinp);
1905	}
1906	SCTP_INP_WLOCK(inp);
1907	/* Pull from where he was */
1908	LIST_REMOVE(inp, sctp_hash);
1909	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1910	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1911	LIST_INSERT_HEAD(head, inp, sctp_hash);
1912	SCTP_INP_WUNLOCK(inp);
1913	SCTP_INP_RLOCK(inp);
1914	return (0);
1915}
1916
1917
1918struct sctp_inpcb *
1919sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1920    uint32_t vrf_id)
1921{
1922	/*
1923	 * First we check the hash table to see if someone has this port
1924	 * bound with just the port.
1925	 */
1926	struct sctp_inpcb *inp;
1927	struct sctppcbhead *head;
1928	int lport;
1929	unsigned int i;
1930
1931#ifdef INET
1932	struct sockaddr_in *sin;
1933
1934#endif
1935#ifdef INET6
1936	struct sockaddr_in6 *sin6;
1937
1938#endif
1939
1940	switch (nam->sa_family) {
1941#ifdef INET
1942	case AF_INET:
1943		sin = (struct sockaddr_in *)nam;
1944		lport = sin->sin_port;
1945		break;
1946#endif
1947#ifdef INET6
1948	case AF_INET6:
1949		sin6 = (struct sockaddr_in6 *)nam;
1950		lport = sin6->sin6_port;
1951		break;
1952#endif
1953	default:
1954		return (NULL);
1955	}
1956	/*
1957	 * I could cheat here and just cast to one of the types but we will
1958	 * do it right. It also provides the check against an Unsupported
1959	 * type too.
1960	 */
1961	/* Find the head of the ALLADDR chain */
1962	if (have_lock == 0) {
1963		SCTP_INP_INFO_RLOCK();
1964	}
1965	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1966	    SCTP_BASE_INFO(hashmark))];
1967	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1968
1969	/*
1970	 * If the TCP model exists it could be that the main listening
1971	 * endpoint is gone but there still exists a connected socket for
1972	 * this guy. If so we can return the first one that we find. This
1973	 * may NOT be the correct one so the caller should be wary on the
1974	 * returned INP. Currently the only caller that sets find_tcp_pool
1975	 * is in bindx where we are verifying that a user CAN bind the
1976	 * address. He either has bound it already, or someone else has, or
1977	 * its open to bind, so this is good enough.
1978	 */
1979	if (inp == NULL && find_tcp_pool) {
1980		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1981			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1982			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1983			if (inp) {
1984				break;
1985			}
1986		}
1987	}
1988	if (inp) {
1989		SCTP_INP_INCR_REF(inp);
1990	}
1991	if (have_lock == 0) {
1992		SCTP_INP_INFO_RUNLOCK();
1993	}
1994	return (inp);
1995}
1996
1997
1998/*
1999 * Find an association for an endpoint with the pointer to whom you want to
2000 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
2001 * need to change the *to to some other struct like a mbuf...
2002 */
2003struct sctp_tcb *
2004sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
2005    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
2006    uint32_t vrf_id)
2007{
2008	struct sctp_inpcb *inp = NULL;
2009	struct sctp_tcb *stcb;
2010
2011	SCTP_INP_INFO_RLOCK();
2012	if (find_tcp_pool) {
2013		if (inp_p != NULL) {
2014			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
2015			    vrf_id);
2016		} else {
2017			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
2018			    vrf_id);
2019		}
2020		if (stcb != NULL) {
2021			SCTP_INP_INFO_RUNLOCK();
2022			return (stcb);
2023		}
2024	}
2025	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
2026	if (inp_p != NULL) {
2027		*inp_p = inp;
2028	}
2029	SCTP_INP_INFO_RUNLOCK();
2030	if (inp == NULL) {
2031		return (NULL);
2032	}
2033	/*
2034	 * ok, we have an endpoint, now lets find the assoc for it (if any)
2035	 * we now place the source address or from in the to of the find
2036	 * endpoint call. Since in reality this chain is used from the
2037	 * inbound packet side.
2038	 */
2039	if (inp_p != NULL) {
2040		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
2041		    NULL);
2042	} else {
2043		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
2044		    NULL);
2045	}
2046	return (stcb);
2047}
2048
2049
2050/*
2051 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
2052 * find all addresses that the sender has specified in any address list. Each
2053 * address will be used to lookup the TCB and see if one exits.
2054 */
2055static struct sctp_tcb *
2056sctp_findassociation_special_addr(struct mbuf *m, int offset,
2057    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
2058    struct sockaddr *dst)
2059{
2060	struct sctp_paramhdr *phdr, parm_buf;
2061
2062#if defined(INET) || defined(INET6)
2063	struct sctp_tcb *stcb;
2064	uint16_t ptype;
2065
2066#endif
2067	uint16_t plen;
2068
2069#ifdef INET
2070	struct sockaddr_in sin4;
2071
2072#endif
2073#ifdef INET6
2074	struct sockaddr_in6 sin6;
2075
2076#endif
2077
2078#ifdef INET
2079	memset(&sin4, 0, sizeof(sin4));
2080	sin4.sin_len = sizeof(sin4);
2081	sin4.sin_family = AF_INET;
2082	sin4.sin_port = sh->src_port;
2083#endif
2084#ifdef INET6
2085	memset(&sin6, 0, sizeof(sin6));
2086	sin6.sin6_len = sizeof(sin6);
2087	sin6.sin6_family = AF_INET6;
2088	sin6.sin6_port = sh->src_port;
2089#endif
2090
2091	offset += sizeof(struct sctp_init_chunk);
2092
2093	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2094	while (phdr != NULL) {
2095		/* now we must see if we want the parameter */
2096#if defined(INET) || defined(INET6)
2097		ptype = ntohs(phdr->param_type);
2098#endif
2099		plen = ntohs(phdr->param_length);
2100		if (plen == 0) {
2101			break;
2102		}
2103#ifdef INET
2104		if (ptype == SCTP_IPV4_ADDRESS &&
2105		    plen == sizeof(struct sctp_ipv4addr_param)) {
2106			/* Get the rest of the address */
2107			struct sctp_ipv4addr_param ip4_parm, *p4;
2108
2109			phdr = sctp_get_next_param(m, offset,
2110			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2111			if (phdr == NULL) {
2112				return (NULL);
2113			}
2114			p4 = (struct sctp_ipv4addr_param *)phdr;
2115			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2116			/* look it up */
2117			stcb = sctp_findassociation_ep_addr(inp_p,
2118			    (struct sockaddr *)&sin4, netp, dst, NULL);
2119			if (stcb != NULL) {
2120				return (stcb);
2121			}
2122		}
2123#endif
2124#ifdef INET6
2125		if (ptype == SCTP_IPV6_ADDRESS &&
2126		    plen == sizeof(struct sctp_ipv6addr_param)) {
2127			/* Get the rest of the address */
2128			struct sctp_ipv6addr_param ip6_parm, *p6;
2129
2130			phdr = sctp_get_next_param(m, offset,
2131			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2132			if (phdr == NULL) {
2133				return (NULL);
2134			}
2135			p6 = (struct sctp_ipv6addr_param *)phdr;
2136			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2137			/* look it up */
2138			stcb = sctp_findassociation_ep_addr(inp_p,
2139			    (struct sockaddr *)&sin6, netp, dst, NULL);
2140			if (stcb != NULL) {
2141				return (stcb);
2142			}
2143		}
2144#endif
2145		offset += SCTP_SIZE32(plen);
2146		phdr = sctp_get_next_param(m, offset, &parm_buf,
2147		    sizeof(parm_buf));
2148	}
2149	return (NULL);
2150}
2151
2152static struct sctp_tcb *
2153sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2154    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2155    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2156{
2157	/*
2158	 * Use my vtag to hash. If we find it we then verify the source addr
2159	 * is in the assoc. If all goes well we save a bit on rec of a
2160	 * packet.
2161	 */
2162	struct sctpasochead *head;
2163	struct sctp_nets *net;
2164	struct sctp_tcb *stcb;
2165
2166	SCTP_INP_INFO_RLOCK();
2167	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2168	    SCTP_BASE_INFO(hashasocmark))];
2169	if (head == NULL) {
2170		/* invalid vtag */
2171		SCTP_INP_INFO_RUNLOCK();
2172		return (NULL);
2173	}
2174	LIST_FOREACH(stcb, head, sctp_asocs) {
2175		SCTP_INP_RLOCK(stcb->sctp_ep);
2176		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2177			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2178			continue;
2179		}
2180		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2181			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2182			continue;
2183		}
2184		SCTP_TCB_LOCK(stcb);
2185		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2186		if (stcb->asoc.my_vtag == vtag) {
2187			/* candidate */
2188			if (stcb->rport != rport) {
2189				SCTP_TCB_UNLOCK(stcb);
2190				continue;
2191			}
2192			if (stcb->sctp_ep->sctp_lport != lport) {
2193				SCTP_TCB_UNLOCK(stcb);
2194				continue;
2195			}
2196			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2197				SCTP_TCB_UNLOCK(stcb);
2198				continue;
2199			}
2200			/* RRS:Need toaddr check here */
2201			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2202				/* Endpoint does not own this address */
2203				SCTP_TCB_UNLOCK(stcb);
2204				continue;
2205			}
2206			if (remote_tag) {
2207				/*
2208				 * If we have both vtags that's all we match
2209				 * on
2210				 */
2211				if (stcb->asoc.peer_vtag == remote_tag) {
2212					/*
2213					 * If both tags match we consider it
2214					 * conclusive and check NO
2215					 * source/destination addresses
2216					 */
2217					goto conclusive;
2218				}
2219			}
2220			if (skip_src_check) {
2221		conclusive:
2222				if (from) {
2223					*netp = sctp_findnet(stcb, from);
2224				} else {
2225					*netp = NULL;	/* unknown */
2226				}
2227				if (inp_p)
2228					*inp_p = stcb->sctp_ep;
2229				SCTP_INP_INFO_RUNLOCK();
2230				return (stcb);
2231			}
2232			net = sctp_findnet(stcb, from);
2233			if (net) {
2234				/* yep its him. */
2235				*netp = net;
2236				SCTP_STAT_INCR(sctps_vtagexpress);
2237				*inp_p = stcb->sctp_ep;
2238				SCTP_INP_INFO_RUNLOCK();
2239				return (stcb);
2240			} else {
2241				/*
2242				 * not him, this should only happen in rare
2243				 * cases so I peg it.
2244				 */
2245				SCTP_STAT_INCR(sctps_vtagbogus);
2246			}
2247		}
2248		SCTP_TCB_UNLOCK(stcb);
2249	}
2250	SCTP_INP_INFO_RUNLOCK();
2251	return (NULL);
2252}
2253
2254
2255/*
2256 * Find an association with the pointer to the inbound IP packet. This can be
2257 * a IPv4 or IPv6 packet.
2258 */
2259struct sctp_tcb *
2260sctp_findassociation_addr(struct mbuf *m, int offset,
2261    struct sockaddr *src, struct sockaddr *dst,
2262    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2263    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2264{
2265	int find_tcp_pool;
2266	struct sctp_tcb *stcb;
2267	struct sctp_inpcb *inp;
2268
2269	if (sh->v_tag) {
2270		/* we only go down this path if vtag is non-zero */
2271		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2272		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2273		if (stcb) {
2274			return (stcb);
2275		}
2276	}
2277	find_tcp_pool = 0;
2278	if ((ch->chunk_type != SCTP_INITIATION) &&
2279	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2280	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2281	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2282		/* Other chunk types go to the tcp pool. */
2283		find_tcp_pool = 1;
2284	}
2285	if (inp_p) {
2286		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2287		    find_tcp_pool, vrf_id);
2288		inp = *inp_p;
2289	} else {
2290		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2291		    find_tcp_pool, vrf_id);
2292	}
2293	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2294	if (stcb == NULL && inp) {
2295		/* Found a EP but not this address */
2296		if ((ch->chunk_type == SCTP_INITIATION) ||
2297		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2298			/*-
2299			 * special hook, we do NOT return linp or an
2300			 * association that is linked to an existing
2301			 * association that is under the TCP pool (i.e. no
2302			 * listener exists). The endpoint finding routine
2303			 * will always find a listener before examining the
2304			 * TCP pool.
2305			 */
2306			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2307				if (inp_p) {
2308					*inp_p = NULL;
2309				}
2310				return (NULL);
2311			}
2312			stcb = sctp_findassociation_special_addr(m,
2313			    offset, sh, &inp, netp, dst);
2314			if (inp_p != NULL) {
2315				*inp_p = inp;
2316			}
2317		}
2318	}
2319	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2320	return (stcb);
2321}
2322
2323/*
2324 * lookup an association by an ASCONF lookup address.
2325 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2326 */
2327struct sctp_tcb *
2328sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2329    struct sockaddr *dst, struct sctphdr *sh,
2330    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2331{
2332	struct sctp_tcb *stcb;
2333	struct sockaddr_storage remote_store;
2334	struct sctp_paramhdr parm_buf, *phdr;
2335	int ptype;
2336	int zero_address = 0;
2337
2338#ifdef INET
2339	struct sockaddr_in *sin;
2340
2341#endif
2342#ifdef INET6
2343	struct sockaddr_in6 *sin6;
2344
2345#endif
2346
2347	memset(&remote_store, 0, sizeof(remote_store));
2348	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2349	    &parm_buf, sizeof(struct sctp_paramhdr));
2350	if (phdr == NULL) {
2351		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2352		    __FUNCTION__);
2353		return NULL;
2354	}
2355	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2356	/* get the correlation address */
2357	switch (ptype) {
2358#ifdef INET6
2359	case SCTP_IPV6_ADDRESS:
2360		{
2361			/* ipv6 address param */
2362			struct sctp_ipv6addr_param *p6, p6_buf;
2363
2364			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2365				return NULL;
2366			}
2367			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2368			    offset + sizeof(struct sctp_asconf_chunk),
2369			    &p6_buf.ph, sizeof(*p6));
2370			if (p6 == NULL) {
2371				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2372				    __FUNCTION__);
2373				return (NULL);
2374			}
2375			sin6 = (struct sockaddr_in6 *)&remote_store;
2376			sin6->sin6_family = AF_INET6;
2377			sin6->sin6_len = sizeof(*sin6);
2378			sin6->sin6_port = sh->src_port;
2379			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2380			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2381				zero_address = 1;
2382			break;
2383		}
2384#endif
2385#ifdef INET
2386	case SCTP_IPV4_ADDRESS:
2387		{
2388			/* ipv4 address param */
2389			struct sctp_ipv4addr_param *p4, p4_buf;
2390
2391			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2392				return NULL;
2393			}
2394			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2395			    offset + sizeof(struct sctp_asconf_chunk),
2396			    &p4_buf.ph, sizeof(*p4));
2397			if (p4 == NULL) {
2398				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2399				    __FUNCTION__);
2400				return (NULL);
2401			}
2402			sin = (struct sockaddr_in *)&remote_store;
2403			sin->sin_family = AF_INET;
2404			sin->sin_len = sizeof(*sin);
2405			sin->sin_port = sh->src_port;
2406			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2407			if (sin->sin_addr.s_addr == INADDR_ANY)
2408				zero_address = 1;
2409			break;
2410		}
2411#endif
2412	default:
2413		/* invalid address param type */
2414		return NULL;
2415	}
2416
2417	if (zero_address) {
2418		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2419		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2420		if (stcb != NULL) {
2421			SCTP_INP_DECR_REF(*inp_p);
2422		}
2423	} else {
2424		stcb = sctp_findassociation_ep_addr(inp_p,
2425		    (struct sockaddr *)&remote_store, netp,
2426		    dst, NULL);
2427	}
2428	return (stcb);
2429}
2430
2431
2432/*
2433 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2434 * addresses. This way if we don't get a bind we by default pick a ephemeral
2435 * port with all addresses bound.
2436 */
2437int
2438sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2439{
2440	/*
2441	 * we get called when a new endpoint starts up. We need to allocate
2442	 * the sctp_inpcb structure from the zone and init it. Mark it as
2443	 * unbound and find a port that we can use as an ephemeral with
2444	 * INADDR_ANY. If the user binds later no problem we can then add in
2445	 * the specific addresses. And setup the default parameters for the
2446	 * EP.
2447	 */
2448	int i, error;
2449	struct sctp_inpcb *inp;
2450	struct sctp_pcb *m;
2451	struct timeval time;
2452	sctp_sharedkey_t *null_key;
2453
2454	error = 0;
2455
2456	SCTP_INP_INFO_WLOCK();
2457	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2458	if (inp == NULL) {
2459		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2460		SCTP_INP_INFO_WUNLOCK();
2461		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2462		return (ENOBUFS);
2463	}
2464	/* zap it */
2465	bzero(inp, sizeof(*inp));
2466
2467	/* bump generations */
2468	/* setup socket pointers */
2469	inp->sctp_socket = so;
2470	inp->ip_inp.inp.inp_socket = so;
2471	inp->ip_inp.inp.inp_cred = crhold(so->so_cred);
2472#ifdef INET6
2473	if (INP_SOCKAF(so) == AF_INET6) {
2474		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2475			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2476		}
2477		if (MODULE_GLOBAL(ip6_v6only)) {
2478			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2479		}
2480	}
2481#endif
2482	inp->sctp_associd_counter = 1;
2483	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2484	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2485	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2486	inp->ecn_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_ecn_enable);
2487	inp->prsctp_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pr_enable);
2488	inp->nrsack_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_nrsack_enable);
2489	/* init the small hash table we use to track asocid <-> tcb */
2490	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2491	if (inp->sctp_asocidhash == NULL) {
2492		crfree(inp->ip_inp.inp.inp_cred);
2493		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2494		SCTP_INP_INFO_WUNLOCK();
2495		return (ENOBUFS);
2496	}
2497#ifdef IPSEC
2498	{
2499		struct inpcbpolicy *pcb_sp = NULL;
2500
2501		error = ipsec_init_policy(so, &pcb_sp);
2502		/* Arrange to share the policy */
2503		inp->ip_inp.inp.inp_sp = pcb_sp;
2504		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2505	}
2506	if (error != 0) {
2507		crfree(inp->ip_inp.inp.inp_cred);
2508		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2509		SCTP_INP_INFO_WUNLOCK();
2510		return error;
2511	}
2512#endif				/* IPSEC */
2513	SCTP_INCR_EP_COUNT();
2514	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2515	SCTP_INP_INFO_WUNLOCK();
2516
2517	so->so_pcb = (caddr_t)inp;
2518
2519	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2520		/* UDP style socket */
2521		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2522		    SCTP_PCB_FLAGS_UNBOUND);
2523		/* Be sure it is NON-BLOCKING IO for UDP */
2524		/* SCTP_SET_SO_NBIO(so); */
2525	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2526		/* TCP style socket */
2527		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2528		    SCTP_PCB_FLAGS_UNBOUND);
2529		/* Be sure we have blocking IO by default */
2530		SCTP_CLEAR_SO_NBIO(so);
2531	} else {
2532		/*
2533		 * unsupported socket type (RAW, etc)- in case we missed it
2534		 * in protosw
2535		 */
2536		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2537		so->so_pcb = NULL;
2538		crfree(inp->ip_inp.inp.inp_cred);
2539		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2540		return (EOPNOTSUPP);
2541	}
2542	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2543		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2544		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2545	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2546		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2547		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2548	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2549		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2550		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2551	}
2552	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2553	    &inp->sctp_hashmark);
2554	if (inp->sctp_tcbhash == NULL) {
2555		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2556		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2557		so->so_pcb = NULL;
2558		crfree(inp->ip_inp.inp.inp_cred);
2559		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2560		return (ENOBUFS);
2561	}
2562	inp->def_vrf_id = vrf_id;
2563
2564	SCTP_INP_INFO_WLOCK();
2565	SCTP_INP_LOCK_INIT(inp);
2566	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2567	SCTP_INP_READ_INIT(inp);
2568	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2569	/* lock the new ep */
2570	SCTP_INP_WLOCK(inp);
2571
2572	/* add it to the info area */
2573	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2574	SCTP_INP_INFO_WUNLOCK();
2575
2576	TAILQ_INIT(&inp->read_queue);
2577	LIST_INIT(&inp->sctp_addr_list);
2578
2579	LIST_INIT(&inp->sctp_asoc_list);
2580
2581#ifdef SCTP_TRACK_FREED_ASOCS
2582	/* TEMP CODE */
2583	LIST_INIT(&inp->sctp_asoc_free_list);
2584#endif
2585	/* Init the timer structure for signature change */
2586	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2587	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2588
2589	/* now init the actual endpoint default data */
2590	m = &inp->sctp_ep;
2591
2592	/* setup the base timeout information */
2593	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2594	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2595	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2596	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2597	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2598	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2599	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2600	/* all max/min max are in ms */
2601	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2602	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2603	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2604	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2605	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2606	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2607	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2608	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2609	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2610	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2611	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2612	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2613	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2614
2615	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2616	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2617	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2618	/* number of streams to pre-open on a association */
2619	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2620
2621	/* Add adaptation cookie */
2622	m->adaptation_layer_indicator = 0;
2623	m->adaptation_layer_indicator_provided = 0;
2624
2625	/* seed random number generator */
2626	m->random_counter = 1;
2627	m->store_at = SCTP_SIGNATURE_SIZE;
2628	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2629	sctp_fill_random_store(m);
2630
2631	/* Minimum cookie size */
2632	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2633	    sizeof(struct sctp_state_cookie);
2634	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2635
2636	/* Setup the initial secret */
2637	(void)SCTP_GETTIME_TIMEVAL(&time);
2638	m->time_of_secret_change = time.tv_sec;
2639
2640	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2641		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2642	}
2643	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2644
2645	/* How long is a cookie good for ? */
2646	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2647	/*
2648	 * Initialize authentication parameters
2649	 */
2650	m->local_hmacs = sctp_default_supported_hmaclist();
2651	m->local_auth_chunks = sctp_alloc_chunklist();
2652	m->default_dscp = 0;
2653#ifdef INET6
2654	m->default_flowlabel = 0;
2655#endif
2656	m->port = 0;		/* encapsulation disabled by default */
2657	sctp_auth_set_default_chunks(m->local_auth_chunks);
2658	LIST_INIT(&m->shared_keys);
2659	/* add default NULL key as key id 0 */
2660	null_key = sctp_alloc_sharedkey();
2661	sctp_insert_sharedkey(&m->shared_keys, null_key);
2662	SCTP_INP_WUNLOCK(inp);
2663#ifdef SCTP_LOG_CLOSING
2664	sctp_log_closing(inp, NULL, 12);
2665#endif
2666	return (error);
2667}
2668
2669
2670void
2671sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2672    struct sctp_tcb *stcb)
2673{
2674	struct sctp_nets *net;
2675	uint16_t lport, rport;
2676	struct sctppcbhead *head;
2677	struct sctp_laddr *laddr, *oladdr;
2678
2679	atomic_add_int(&stcb->asoc.refcnt, 1);
2680	SCTP_TCB_UNLOCK(stcb);
2681	SCTP_INP_INFO_WLOCK();
2682	SCTP_INP_WLOCK(old_inp);
2683	SCTP_INP_WLOCK(new_inp);
2684	SCTP_TCB_LOCK(stcb);
2685	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2686
2687	new_inp->sctp_ep.time_of_secret_change =
2688	    old_inp->sctp_ep.time_of_secret_change;
2689	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2690	    sizeof(old_inp->sctp_ep.secret_key));
2691	new_inp->sctp_ep.current_secret_number =
2692	    old_inp->sctp_ep.current_secret_number;
2693	new_inp->sctp_ep.last_secret_number =
2694	    old_inp->sctp_ep.last_secret_number;
2695	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2696
2697	/* make it so new data pours into the new socket */
2698	stcb->sctp_socket = new_inp->sctp_socket;
2699	stcb->sctp_ep = new_inp;
2700
2701	/* Copy the port across */
2702	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2703	rport = stcb->rport;
2704	/* Pull the tcb from the old association */
2705	LIST_REMOVE(stcb, sctp_tcbhash);
2706	LIST_REMOVE(stcb, sctp_tcblist);
2707	if (stcb->asoc.in_asocid_hash) {
2708		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2709	}
2710	/* Now insert the new_inp into the TCP connected hash */
2711	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2712
2713	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2714	/* Its safe to access */
2715	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2716
2717	/* Now move the tcb into the endpoint list */
2718	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2719	/*
2720	 * Question, do we even need to worry about the ep-hash since we
2721	 * only have one connection? Probably not :> so lets get rid of it
2722	 * and not suck up any kernel memory in that.
2723	 */
2724	if (stcb->asoc.in_asocid_hash) {
2725		struct sctpasochead *lhd;
2726
2727		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2728		    new_inp->hashasocidmark)];
2729		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2730	}
2731	/* Ok. Let's restart timer. */
2732	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2733		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2734		    stcb, net);
2735	}
2736
2737	SCTP_INP_INFO_WUNLOCK();
2738	if (new_inp->sctp_tcbhash != NULL) {
2739		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2740		new_inp->sctp_tcbhash = NULL;
2741	}
2742	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2743		/* Subset bound, so copy in the laddr list from the old_inp */
2744		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2745			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2746			if (laddr == NULL) {
2747				/*
2748				 * Gak, what can we do? This assoc is really
2749				 * HOSED. We probably should send an abort
2750				 * here.
2751				 */
2752				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2753				continue;
2754			}
2755			SCTP_INCR_LADDR_COUNT();
2756			bzero(laddr, sizeof(*laddr));
2757			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2758			laddr->ifa = oladdr->ifa;
2759			atomic_add_int(&laddr->ifa->refcount, 1);
2760			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2761			    sctp_nxt_addr);
2762			new_inp->laddr_count++;
2763			if (oladdr == stcb->asoc.last_used_address) {
2764				stcb->asoc.last_used_address = laddr;
2765			}
2766		}
2767	}
2768	/*
2769	 * Now any running timers need to be adjusted since we really don't
2770	 * care if they are running or not just blast in the new_inp into
2771	 * all of them.
2772	 */
2773
2774	stcb->asoc.dack_timer.ep = (void *)new_inp;
2775	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2776	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2777	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2778	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2779	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2780	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2781	/* now what about the nets? */
2782	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2783		net->pmtu_timer.ep = (void *)new_inp;
2784		net->hb_timer.ep = (void *)new_inp;
2785		net->rxt_timer.ep = (void *)new_inp;
2786	}
2787	SCTP_INP_WUNLOCK(new_inp);
2788	SCTP_INP_WUNLOCK(old_inp);
2789}
2790
2791
2792
2793
2794/* sctp_ifap is used to bypass normal local address validation checks */
2795int
2796sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2797    struct sctp_ifa *sctp_ifap, struct thread *p)
2798{
2799	/* bind a ep to a socket address */
2800	struct sctppcbhead *head;
2801	struct sctp_inpcb *inp, *inp_tmp;
2802	struct inpcb *ip_inp;
2803	int port_reuse_active = 0;
2804	int bindall;
2805	uint16_t lport;
2806	int error;
2807	uint32_t vrf_id;
2808
2809	lport = 0;
2810	bindall = 1;
2811	inp = (struct sctp_inpcb *)so->so_pcb;
2812	ip_inp = (struct inpcb *)so->so_pcb;
2813#ifdef SCTP_DEBUG
2814	if (addr) {
2815		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2816		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2817		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2818		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2819	}
2820#endif
2821	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2822		/* already did a bind, subsequent binds NOT allowed ! */
2823		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2824		return (EINVAL);
2825	}
2826#ifdef INVARIANTS
2827	if (p == NULL)
2828		panic("null proc/thread");
2829#endif
2830	if (addr != NULL) {
2831		switch (addr->sa_family) {
2832#ifdef INET
2833		case AF_INET:
2834			{
2835				struct sockaddr_in *sin;
2836
2837				/* IPV6_V6ONLY socket? */
2838				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2839					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2840					return (EINVAL);
2841				}
2842				if (addr->sa_len != sizeof(*sin)) {
2843					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2844					return (EINVAL);
2845				}
2846				sin = (struct sockaddr_in *)addr;
2847				lport = sin->sin_port;
2848				/*
2849				 * For LOOPBACK the prison_local_ip4() call
2850				 * will transmute the ip address to the
2851				 * proper value.
2852				 */
2853				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2854					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2855					return (error);
2856				}
2857				if (sin->sin_addr.s_addr != INADDR_ANY) {
2858					bindall = 0;
2859				}
2860				break;
2861			}
2862#endif
2863#ifdef INET6
2864		case AF_INET6:
2865			{
2866				/*
2867				 * Only for pure IPv6 Address. (No IPv4
2868				 * Mapped!)
2869				 */
2870				struct sockaddr_in6 *sin6;
2871
2872				sin6 = (struct sockaddr_in6 *)addr;
2873
2874				if (addr->sa_len != sizeof(*sin6)) {
2875					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2876					return (EINVAL);
2877				}
2878				lport = sin6->sin6_port;
2879				/*
2880				 * For LOOPBACK the prison_local_ip6() call
2881				 * will transmute the ipv6 address to the
2882				 * proper value.
2883				 */
2884				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2885				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2886					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2887					return (error);
2888				}
2889				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2890					bindall = 0;
2891					/* KAME hack: embed scopeid */
2892					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2893						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2894						return (EINVAL);
2895					}
2896				}
2897				/* this must be cleared for ifa_ifwithaddr() */
2898				sin6->sin6_scope_id = 0;
2899				break;
2900			}
2901#endif
2902		default:
2903			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2904			return (EAFNOSUPPORT);
2905		}
2906	}
2907	SCTP_INP_INFO_WLOCK();
2908	SCTP_INP_WLOCK(inp);
2909	/* Setup a vrf_id to be the default for the non-bind-all case. */
2910	vrf_id = inp->def_vrf_id;
2911
2912	/* increase our count due to the unlock we do */
2913	SCTP_INP_INCR_REF(inp);
2914	if (lport) {
2915		/*
2916		 * Did the caller specify a port? if so we must see if an ep
2917		 * already has this one bound.
2918		 */
2919		/* got to be root to get at low ports */
2920		if (ntohs(lport) < IPPORT_RESERVED) {
2921			if (p && (error =
2922			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2923			    )) {
2924				SCTP_INP_DECR_REF(inp);
2925				SCTP_INP_WUNLOCK(inp);
2926				SCTP_INP_INFO_WUNLOCK();
2927				return (error);
2928			}
2929		}
2930		SCTP_INP_WUNLOCK(inp);
2931		if (bindall) {
2932			vrf_id = inp->def_vrf_id;
2933			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2934			if (inp_tmp != NULL) {
2935				/*
2936				 * lock guy returned and lower count note
2937				 * that we are not bound so inp_tmp should
2938				 * NEVER be inp. And it is this inp
2939				 * (inp_tmp) that gets the reference bump,
2940				 * so we must lower it.
2941				 */
2942				SCTP_INP_DECR_REF(inp_tmp);
2943				/* unlock info */
2944				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2945				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2946					/*
2947					 * Ok, must be one-2-one and
2948					 * allowing port re-use
2949					 */
2950					port_reuse_active = 1;
2951					goto continue_anyway;
2952				}
2953				SCTP_INP_DECR_REF(inp);
2954				SCTP_INP_INFO_WUNLOCK();
2955				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2956				return (EADDRINUSE);
2957			}
2958		} else {
2959			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2960			if (inp_tmp != NULL) {
2961				/*
2962				 * lock guy returned and lower count note
2963				 * that we are not bound so inp_tmp should
2964				 * NEVER be inp. And it is this inp
2965				 * (inp_tmp) that gets the reference bump,
2966				 * so we must lower it.
2967				 */
2968				SCTP_INP_DECR_REF(inp_tmp);
2969				/* unlock info */
2970				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2971				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2972					/*
2973					 * Ok, must be one-2-one and
2974					 * allowing port re-use
2975					 */
2976					port_reuse_active = 1;
2977					goto continue_anyway;
2978				}
2979				SCTP_INP_DECR_REF(inp);
2980				SCTP_INP_INFO_WUNLOCK();
2981				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2982				return (EADDRINUSE);
2983			}
2984		}
2985continue_anyway:
2986		SCTP_INP_WLOCK(inp);
2987		if (bindall) {
2988			/* verify that no lport is not used by a singleton */
2989			if ((port_reuse_active == 0) &&
2990			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
2991				/* Sorry someone already has this one bound */
2992				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2993				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2994					port_reuse_active = 1;
2995				} else {
2996					SCTP_INP_DECR_REF(inp);
2997					SCTP_INP_WUNLOCK(inp);
2998					SCTP_INP_INFO_WUNLOCK();
2999					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3000					return (EADDRINUSE);
3001				}
3002			}
3003		}
3004	} else {
3005		uint16_t first, last, candidate;
3006		uint16_t count;
3007		int done;
3008
3009		if (ip_inp->inp_flags & INP_HIGHPORT) {
3010			first = MODULE_GLOBAL(ipport_hifirstauto);
3011			last = MODULE_GLOBAL(ipport_hilastauto);
3012		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3013			if (p && (error =
3014			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3015			    )) {
3016				SCTP_INP_DECR_REF(inp);
3017				SCTP_INP_WUNLOCK(inp);
3018				SCTP_INP_INFO_WUNLOCK();
3019				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3020				return (error);
3021			}
3022			first = MODULE_GLOBAL(ipport_lowfirstauto);
3023			last = MODULE_GLOBAL(ipport_lowlastauto);
3024		} else {
3025			first = MODULE_GLOBAL(ipport_firstauto);
3026			last = MODULE_GLOBAL(ipport_lastauto);
3027		}
3028		if (first > last) {
3029			uint16_t temp;
3030
3031			temp = first;
3032			first = last;
3033			last = temp;
3034		}
3035		count = last - first + 1;	/* number of candidates */
3036		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3037
3038		done = 0;
3039		while (!done) {
3040			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3041				done = 1;
3042			}
3043			if (!done) {
3044				if (--count == 0) {
3045					SCTP_INP_DECR_REF(inp);
3046					SCTP_INP_WUNLOCK(inp);
3047					SCTP_INP_INFO_WUNLOCK();
3048					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3049					return (EADDRINUSE);
3050				}
3051				if (candidate == last)
3052					candidate = first;
3053				else
3054					candidate = candidate + 1;
3055			}
3056		}
3057		lport = htons(candidate);
3058	}
3059	SCTP_INP_DECR_REF(inp);
3060	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3061	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3062		/*
3063		 * this really should not happen. The guy did a non-blocking
3064		 * bind and then did a close at the same time.
3065		 */
3066		SCTP_INP_WUNLOCK(inp);
3067		SCTP_INP_INFO_WUNLOCK();
3068		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3069		return (EINVAL);
3070	}
3071	/* ok we look clear to give out this port, so lets setup the binding */
3072	if (bindall) {
3073		/* binding to all addresses, so just set in the proper flags */
3074		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3075		/* set the automatic addr changes from kernel flag */
3076		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3077			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3078			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3079		} else {
3080			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3081			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3082		}
3083		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3084			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3085		} else {
3086			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3087		}
3088		/*
3089		 * set the automatic mobility_base from kernel flag (by
3090		 * micchie)
3091		 */
3092		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3093			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3094			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3095		} else {
3096			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3097			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3098		}
3099		/*
3100		 * set the automatic mobility_fasthandoff from kernel flag
3101		 * (by micchie)
3102		 */
3103		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3104			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3105			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3106		} else {
3107			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3108			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3109		}
3110	} else {
3111		/*
3112		 * bind specific, make sure flags is off and add a new
3113		 * address structure to the sctp_addr_list inside the ep
3114		 * structure.
3115		 *
3116		 * We will need to allocate one and insert it at the head. The
3117		 * socketopt call can just insert new addresses in there as
3118		 * well. It will also have to do the embed scope kame hack
3119		 * too (before adding).
3120		 */
3121		struct sctp_ifa *ifa;
3122		struct sockaddr_storage store_sa;
3123
3124		memset(&store_sa, 0, sizeof(store_sa));
3125		switch (addr->sa_family) {
3126#ifdef INET
3127		case AF_INET:
3128			{
3129				struct sockaddr_in *sin;
3130
3131				sin = (struct sockaddr_in *)&store_sa;
3132				memcpy(sin, addr, sizeof(struct sockaddr_in));
3133				sin->sin_port = 0;
3134				break;
3135			}
3136#endif
3137#ifdef INET6
3138		case AF_INET6:
3139			{
3140				struct sockaddr_in6 *sin6;
3141
3142				sin6 = (struct sockaddr_in6 *)&store_sa;
3143				memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3144				sin6->sin6_port = 0;
3145				break;
3146			}
3147#endif
3148		default:
3149			break;
3150		}
3151		/*
3152		 * first find the interface with the bound address need to
3153		 * zero out the port to find the address! yuck! can't do
3154		 * this earlier since need port for sctp_pcb_findep()
3155		 */
3156		if (sctp_ifap != NULL) {
3157			ifa = sctp_ifap;
3158		} else {
3159			/*
3160			 * Note for BSD we hit here always other O/S's will
3161			 * pass things in via the sctp_ifap argument
3162			 * (Panda).
3163			 */
3164			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3165			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3166		}
3167		if (ifa == NULL) {
3168			/* Can't find an interface with that address */
3169			SCTP_INP_WUNLOCK(inp);
3170			SCTP_INP_INFO_WUNLOCK();
3171			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3172			return (EADDRNOTAVAIL);
3173		}
3174#ifdef INET6
3175		if (addr->sa_family == AF_INET6) {
3176			/* GAK, more FIXME IFA lock? */
3177			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3178				/* Can't bind a non-existent addr. */
3179				SCTP_INP_WUNLOCK(inp);
3180				SCTP_INP_INFO_WUNLOCK();
3181				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3182				return (EINVAL);
3183			}
3184		}
3185#endif
3186		/* we're not bound all */
3187		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3188		/* allow bindx() to send ASCONF's for binding changes */
3189		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3190		/* clear automatic addr changes from kernel flag */
3191		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3192
3193		/* add this address to the endpoint list */
3194		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3195		if (error != 0) {
3196			SCTP_INP_WUNLOCK(inp);
3197			SCTP_INP_INFO_WUNLOCK();
3198			return (error);
3199		}
3200		inp->laddr_count++;
3201	}
3202	/* find the bucket */
3203	if (port_reuse_active) {
3204		/* Put it into tcp 1-2-1 hash */
3205		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3206		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3207	} else {
3208		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3209	}
3210	/* put it in the bucket */
3211	LIST_INSERT_HEAD(head, inp, sctp_hash);
3212	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3213	    (void *)head, ntohs(lport), port_reuse_active);
3214	/* set in the port */
3215	inp->sctp_lport = lport;
3216
3217	/* turn off just the unbound flag */
3218	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3219	SCTP_INP_WUNLOCK(inp);
3220	SCTP_INP_INFO_WUNLOCK();
3221	return (0);
3222}
3223
3224
3225static void
3226sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3227{
3228	struct sctp_iterator *it, *nit;
3229
3230	/*
3231	 * We enter with the only the ITERATOR_LOCK in place and a write
3232	 * lock on the inp_info stuff.
3233	 */
3234	it = sctp_it_ctl.cur_it;
3235	if (it && (it->vn != curvnet)) {
3236		/* Its not looking at our VNET */
3237		return;
3238	}
3239	if (it && (it->inp == inp)) {
3240		/*
3241		 * This is tricky and we hold the iterator lock, but when it
3242		 * returns and gets the lock (when we release it) the
3243		 * iterator will try to operate on inp. We need to stop that
3244		 * from happening. But of course the iterator has a
3245		 * reference on the stcb and inp. We can mark it and it will
3246		 * stop.
3247		 *
3248		 * If its a single iterator situation, we set the end iterator
3249		 * flag. Otherwise we set the iterator to go to the next
3250		 * inp.
3251		 *
3252		 */
3253		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3254			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3255		} else {
3256			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3257		}
3258	}
3259	/*
3260	 * Now go through and remove any single reference to our inp that
3261	 * may be still pending on the list
3262	 */
3263	SCTP_IPI_ITERATOR_WQ_LOCK();
3264	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3265		if (it->vn != curvnet) {
3266			continue;
3267		}
3268		if (it->inp == inp) {
3269			/* This one points to me is it inp specific? */
3270			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3271				/* Remove and free this one */
3272				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3273				    it, sctp_nxt_itr);
3274				if (it->function_atend != NULL) {
3275					(*it->function_atend) (it->pointer, it->val);
3276				}
3277				SCTP_FREE(it, SCTP_M_ITER);
3278			} else {
3279				it->inp = LIST_NEXT(it->inp, sctp_list);
3280				if (it->inp) {
3281					SCTP_INP_INCR_REF(it->inp);
3282				}
3283			}
3284			/*
3285			 * When its put in the refcnt is incremented so decr
3286			 * it
3287			 */
3288			SCTP_INP_DECR_REF(inp);
3289		}
3290	}
3291	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3292}
3293
3294/* release sctp_inpcb unbind the port */
3295void
3296sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3297{
3298	/*
3299	 * Here we free a endpoint. We must find it (if it is in the Hash
3300	 * table) and remove it from there. Then we must also find it in the
3301	 * overall list and remove it from there. After all removals are
3302	 * complete then any timer has to be stopped. Then start the actual
3303	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3304	 * all associations. d) finally the ep itself.
3305	 */
3306	struct sctp_tcb *asoc, *nasoc;
3307	struct sctp_laddr *laddr, *nladdr;
3308	struct inpcb *ip_pcb;
3309	struct socket *so;
3310	int being_refed = 0;
3311	struct sctp_queued_to_read *sq, *nsq;
3312	int cnt;
3313	sctp_sharedkey_t *shared_key, *nshared_key;
3314
3315
3316#ifdef SCTP_LOG_CLOSING
3317	sctp_log_closing(inp, NULL, 0);
3318#endif
3319	SCTP_ITERATOR_LOCK();
3320	/* mark any iterators on the list or being processed */
3321	sctp_iterator_inp_being_freed(inp);
3322	SCTP_ITERATOR_UNLOCK();
3323	so = inp->sctp_socket;
3324	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3325		/* been here before.. eeks.. get out of here */
3326		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3327#ifdef SCTP_LOG_CLOSING
3328		sctp_log_closing(inp, NULL, 1);
3329#endif
3330		return;
3331	}
3332	SCTP_ASOC_CREATE_LOCK(inp);
3333	SCTP_INP_INFO_WLOCK();
3334
3335	SCTP_INP_WLOCK(inp);
3336	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3337		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3338		/* socket is gone, so no more wakeups allowed */
3339		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3340		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3341		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3342
3343	}
3344	/* First time through we have the socket lock, after that no more. */
3345	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3346	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3347
3348	if (inp->control) {
3349		sctp_m_freem(inp->control);
3350		inp->control = NULL;
3351	}
3352	if (inp->pkt) {
3353		sctp_m_freem(inp->pkt);
3354		inp->pkt = NULL;
3355	}
3356	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3357					 * here but I will be nice :> (i.e.
3358					 * ip_pcb = ep;) */
3359	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3360		int cnt_in_sd;
3361
3362		cnt_in_sd = 0;
3363		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3364			SCTP_TCB_LOCK(asoc);
3365			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3366				/* Skip guys being freed */
3367				cnt_in_sd++;
3368				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3369					/*
3370					 * Special case - we did not start a
3371					 * kill timer on the asoc due to it
3372					 * was not closed. So go ahead and
3373					 * start it now.
3374					 */
3375					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3376					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3377				}
3378				SCTP_TCB_UNLOCK(asoc);
3379				continue;
3380			}
3381			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3382			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3383			    (asoc->asoc.total_output_queue_size == 0)) {
3384				/*
3385				 * If we have data in queue, we don't want
3386				 * to just free since the app may have done,
3387				 * send()/close or connect/send/close. And
3388				 * it wants the data to get across first.
3389				 */
3390				/* Just abandon things in the front states */
3391				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3392				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3393					cnt_in_sd++;
3394				}
3395				continue;
3396			}
3397			/* Disconnect the socket please */
3398			asoc->sctp_socket = NULL;
3399			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3400			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3401			    (asoc->asoc.control_pdapi) ||
3402			    (asoc->asoc.size_on_all_streams > 0) ||
3403			    (so && (so->so_rcv.sb_cc > 0))) {
3404				/* Left with Data unread */
3405				struct mbuf *op_err;
3406
3407				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3408				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3409				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3410				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3411				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3412				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3413					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3414				}
3415				if (sctp_free_assoc(inp, asoc,
3416				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3417					cnt_in_sd++;
3418				}
3419				continue;
3420			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3421				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3422			    (asoc->asoc.stream_queue_cnt == 0)) {
3423				if (asoc->asoc.locked_on_sending) {
3424					goto abort_anyway;
3425				}
3426				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3427				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3428					struct sctp_nets *netp;
3429
3430					/*
3431					 * there is nothing queued to send,
3432					 * so I send shutdown
3433					 */
3434					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3435					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3436						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3437					}
3438					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3439					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3440					sctp_stop_timers_for_shutdown(asoc);
3441					if (asoc->asoc.alternate) {
3442						netp = asoc->asoc.alternate;
3443					} else {
3444						netp = asoc->asoc.primary_destination;
3445					}
3446					sctp_send_shutdown(asoc, netp);
3447					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3448					    netp);
3449					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3450					    asoc->asoc.primary_destination);
3451					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3452				}
3453			} else {
3454				/* mark into shutdown pending */
3455				struct sctp_stream_queue_pending *sp;
3456
3457				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3458				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3459				    asoc->asoc.primary_destination);
3460				if (asoc->asoc.locked_on_sending) {
3461					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3462					    sctp_streamhead);
3463					if (sp == NULL) {
3464						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3465						    (void *)asoc->asoc.locked_on_sending,
3466						    asoc->asoc.locked_on_sending->stream_no);
3467					} else {
3468						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3469							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3470					}
3471				}
3472				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3473				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3474				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3475					struct mbuf *op_err;
3476
3477			abort_anyway:
3478					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3479					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3480					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3481					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3482					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3483					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3484						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3485					}
3486					if (sctp_free_assoc(inp, asoc,
3487					    SCTP_PCBFREE_NOFORCE,
3488					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3489						cnt_in_sd++;
3490					}
3491					continue;
3492				} else {
3493					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3494				}
3495			}
3496			cnt_in_sd++;
3497			SCTP_TCB_UNLOCK(asoc);
3498		}
3499		/* now is there some left in our SHUTDOWN state? */
3500		if (cnt_in_sd) {
3501#ifdef SCTP_LOG_CLOSING
3502			sctp_log_closing(inp, NULL, 2);
3503#endif
3504			inp->sctp_socket = NULL;
3505			SCTP_INP_WUNLOCK(inp);
3506			SCTP_ASOC_CREATE_UNLOCK(inp);
3507			SCTP_INP_INFO_WUNLOCK();
3508			return;
3509		}
3510	}
3511	inp->sctp_socket = NULL;
3512	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3513	    SCTP_PCB_FLAGS_UNBOUND) {
3514		/*
3515		 * ok, this guy has been bound. It's port is somewhere in
3516		 * the SCTP_BASE_INFO(hash table). Remove it!
3517		 */
3518		LIST_REMOVE(inp, sctp_hash);
3519		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3520	}
3521	/*
3522	 * If there is a timer running to kill us, forget it, since it may
3523	 * have a contest on the INP lock.. which would cause us to die ...
3524	 */
3525	cnt = 0;
3526	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3527		SCTP_TCB_LOCK(asoc);
3528		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3529			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3530				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3531				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3532			}
3533			cnt++;
3534			SCTP_TCB_UNLOCK(asoc);
3535			continue;
3536		}
3537		/* Free associations that are NOT killing us */
3538		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3539		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3540			struct mbuf *op_err;
3541
3542			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3543			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3544			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3545			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3546		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3547			cnt++;
3548			SCTP_TCB_UNLOCK(asoc);
3549			continue;
3550		}
3551		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3552		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3553			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3554		}
3555		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3556			cnt++;
3557		}
3558	}
3559	if (cnt) {
3560		/* Ok we have someone out there that will kill us */
3561		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3562#ifdef SCTP_LOG_CLOSING
3563		sctp_log_closing(inp, NULL, 3);
3564#endif
3565		SCTP_INP_WUNLOCK(inp);
3566		SCTP_ASOC_CREATE_UNLOCK(inp);
3567		SCTP_INP_INFO_WUNLOCK();
3568		return;
3569	}
3570	if (SCTP_INP_LOCK_CONTENDED(inp))
3571		being_refed++;
3572	if (SCTP_INP_READ_CONTENDED(inp))
3573		being_refed++;
3574	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3575		being_refed++;
3576
3577	if ((inp->refcount) ||
3578	    (being_refed) ||
3579	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3580		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3581#ifdef SCTP_LOG_CLOSING
3582		sctp_log_closing(inp, NULL, 4);
3583#endif
3584		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3585		SCTP_INP_WUNLOCK(inp);
3586		SCTP_ASOC_CREATE_UNLOCK(inp);
3587		SCTP_INP_INFO_WUNLOCK();
3588		return;
3589	}
3590	inp->sctp_ep.signature_change.type = 0;
3591	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3592	/*
3593	 * Remove it from the list .. last thing we need a lock for.
3594	 */
3595	LIST_REMOVE(inp, sctp_list);
3596	SCTP_INP_WUNLOCK(inp);
3597	SCTP_ASOC_CREATE_UNLOCK(inp);
3598	SCTP_INP_INFO_WUNLOCK();
3599	/*
3600	 * Now we release all locks. Since this INP cannot be found anymore
3601	 * except possibly by the kill timer that might be running. We call
3602	 * the drain function here. It should hit the case were it sees the
3603	 * ACTIVE flag cleared and exit out freeing us to proceed and
3604	 * destroy everything.
3605	 */
3606	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3607		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3608	} else {
3609		/* Probably un-needed */
3610		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3611	}
3612
3613#ifdef SCTP_LOG_CLOSING
3614	sctp_log_closing(inp, NULL, 5);
3615#endif
3616
3617
3618	if ((inp->sctp_asocidhash) != NULL) {
3619		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3620		inp->sctp_asocidhash = NULL;
3621	}
3622	/* sa_ignore FREED_MEMORY */
3623	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3624		/* Its only abandoned if it had data left */
3625		if (sq->length)
3626			SCTP_STAT_INCR(sctps_left_abandon);
3627
3628		TAILQ_REMOVE(&inp->read_queue, sq, next);
3629		sctp_free_remote_addr(sq->whoFrom);
3630		if (so)
3631			so->so_rcv.sb_cc -= sq->length;
3632		if (sq->data) {
3633			sctp_m_freem(sq->data);
3634			sq->data = NULL;
3635		}
3636		/*
3637		 * no need to free the net count, since at this point all
3638		 * assoc's are gone.
3639		 */
3640		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3641		SCTP_DECR_READQ_COUNT();
3642	}
3643	/* Now the sctp_pcb things */
3644	/*
3645	 * free each asoc if it is not already closed/free. we can't use the
3646	 * macro here since le_next will get freed as part of the
3647	 * sctp_free_assoc() call.
3648	 */
3649	if (so) {
3650#ifdef IPSEC
3651		ipsec_delete_pcbpolicy(ip_pcb);
3652#endif				/* IPSEC */
3653
3654		/* Unlocks not needed since the socket is gone now */
3655	}
3656	if (ip_pcb->inp_options) {
3657		(void)sctp_m_free(ip_pcb->inp_options);
3658		ip_pcb->inp_options = 0;
3659	}
3660#ifdef INET6
3661	if (ip_pcb->inp_vflag & INP_IPV6) {
3662		struct in6pcb *in6p;
3663
3664		in6p = (struct in6pcb *)inp;
3665		ip6_freepcbopts(in6p->in6p_outputopts);
3666	}
3667#endif				/* INET6 */
3668	ip_pcb->inp_vflag = 0;
3669	/* free up authentication fields */
3670	if (inp->sctp_ep.local_auth_chunks != NULL)
3671		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3672	if (inp->sctp_ep.local_hmacs != NULL)
3673		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3674
3675	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3676		LIST_REMOVE(shared_key, next);
3677		sctp_free_sharedkey(shared_key);
3678		/* sa_ignore FREED_MEMORY */
3679	}
3680
3681	/*
3682	 * if we have an address list the following will free the list of
3683	 * ifaddr's that are set into this ep. Again macro limitations here,
3684	 * since the LIST_FOREACH could be a bad idea.
3685	 */
3686	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3687		sctp_remove_laddr(laddr);
3688	}
3689
3690#ifdef SCTP_TRACK_FREED_ASOCS
3691	/* TEMP CODE */
3692	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3693		LIST_REMOVE(asoc, sctp_tcblist);
3694		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3695		SCTP_DECR_ASOC_COUNT();
3696	}
3697	/* *** END TEMP CODE *** */
3698#endif
3699	/* Now lets see about freeing the EP hash table. */
3700	if (inp->sctp_tcbhash != NULL) {
3701		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3702		inp->sctp_tcbhash = NULL;
3703	}
3704	/* Now we must put the ep memory back into the zone pool */
3705	crfree(inp->ip_inp.inp.inp_cred);
3706	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3707	SCTP_INP_LOCK_DESTROY(inp);
3708	SCTP_INP_READ_DESTROY(inp);
3709	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3710	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3711	SCTP_DECR_EP_COUNT();
3712}
3713
3714
3715struct sctp_nets *
3716sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3717{
3718	struct sctp_nets *net;
3719
3720	/* locate the address */
3721	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3722		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3723			return (net);
3724	}
3725	return (NULL);
3726}
3727
3728
3729int
3730sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3731{
3732	struct sctp_ifa *sctp_ifa;
3733
3734	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3735	if (sctp_ifa) {
3736		return (1);
3737	} else {
3738		return (0);
3739	}
3740}
3741
3742/*
3743 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3744 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3745 * stats of stuff.
3746 */
3747int
3748sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3749    struct sctp_nets **netp, int set_scope, int from)
3750{
3751	/*
3752	 * The following is redundant to the same lines in the
3753	 * sctp_aloc_assoc() but is needed since others call the add address
3754	 * function
3755	 */
3756	struct sctp_nets *net, *netfirst;
3757	int addr_inscope;
3758
3759	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3760	    from);
3761	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3762
3763	netfirst = sctp_findnet(stcb, newaddr);
3764	if (netfirst) {
3765		/*
3766		 * Lie and return ok, we don't want to make the association
3767		 * go away for this behavior. It will happen in the TCP
3768		 * model in a connected socket. It does not reach the hash
3769		 * table until after the association is built so it can't be
3770		 * found. Mark as reachable, since the initial creation will
3771		 * have been cleared and the NOT_IN_ASSOC flag will have
3772		 * been added... and we don't want to end up removing it
3773		 * back out.
3774		 */
3775		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3776			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3777			    SCTP_ADDR_UNCONFIRMED);
3778		} else {
3779			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3780		}
3781
3782		return (0);
3783	}
3784	addr_inscope = 1;
3785	switch (newaddr->sa_family) {
3786#ifdef INET
3787	case AF_INET:
3788		{
3789			struct sockaddr_in *sin;
3790
3791			sin = (struct sockaddr_in *)newaddr;
3792			if (sin->sin_addr.s_addr == 0) {
3793				/* Invalid address */
3794				return (-1);
3795			}
3796			/* zero out the bzero area */
3797			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3798
3799			/* assure len is set */
3800			sin->sin_len = sizeof(struct sockaddr_in);
3801			if (set_scope) {
3802#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3803				stcb->asoc.scope.ipv4_local_scope = 1;
3804#else
3805				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3806					stcb->asoc.scope.ipv4_local_scope = 1;
3807				}
3808#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3809			} else {
3810				/* Validate the address is in scope */
3811				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3812				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3813					addr_inscope = 0;
3814				}
3815			}
3816			break;
3817		}
3818#endif
3819#ifdef INET6
3820	case AF_INET6:
3821		{
3822			struct sockaddr_in6 *sin6;
3823
3824			sin6 = (struct sockaddr_in6 *)newaddr;
3825			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3826				/* Invalid address */
3827				return (-1);
3828			}
3829			/* assure len is set */
3830			sin6->sin6_len = sizeof(struct sockaddr_in6);
3831			if (set_scope) {
3832				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3833					stcb->asoc.scope.loopback_scope = 1;
3834					stcb->asoc.scope.local_scope = 0;
3835					stcb->asoc.scope.ipv4_local_scope = 1;
3836					stcb->asoc.scope.site_scope = 1;
3837				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3838					/*
3839					 * If the new destination is a
3840					 * LINK_LOCAL we must have common
3841					 * site scope. Don't set the local
3842					 * scope since we may not share all
3843					 * links, only loopback can do this.
3844					 * Links on the local network would
3845					 * also be on our private network
3846					 * for v4 too.
3847					 */
3848					stcb->asoc.scope.ipv4_local_scope = 1;
3849					stcb->asoc.scope.site_scope = 1;
3850				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3851					/*
3852					 * If the new destination is
3853					 * SITE_LOCAL then we must have site
3854					 * scope in common.
3855					 */
3856					stcb->asoc.scope.site_scope = 1;
3857				}
3858			} else {
3859				/* Validate the address is in scope */
3860				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3861				    (stcb->asoc.scope.loopback_scope == 0)) {
3862					addr_inscope = 0;
3863				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3864				    (stcb->asoc.scope.local_scope == 0)) {
3865					addr_inscope = 0;
3866				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3867				    (stcb->asoc.scope.site_scope == 0)) {
3868					addr_inscope = 0;
3869				}
3870			}
3871			break;
3872		}
3873#endif
3874	default:
3875		/* not supported family type */
3876		return (-1);
3877	}
3878	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3879	if (net == NULL) {
3880		return (-1);
3881	}
3882	SCTP_INCR_RADDR_COUNT();
3883	bzero(net, sizeof(struct sctp_nets));
3884	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3885	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3886	switch (newaddr->sa_family) {
3887#ifdef INET
3888	case AF_INET:
3889		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3890		break;
3891#endif
3892#ifdef INET6
3893	case AF_INET6:
3894		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3895		break;
3896#endif
3897	default:
3898		break;
3899	}
3900	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3901	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3902		stcb->asoc.scope.loopback_scope = 1;
3903		stcb->asoc.scope.ipv4_local_scope = 1;
3904		stcb->asoc.scope.local_scope = 0;
3905		stcb->asoc.scope.site_scope = 1;
3906		addr_inscope = 1;
3907	}
3908	net->failure_threshold = stcb->asoc.def_net_failure;
3909	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3910	if (addr_inscope == 0) {
3911		net->dest_state = (SCTP_ADDR_REACHABLE |
3912		    SCTP_ADDR_OUT_OF_SCOPE);
3913	} else {
3914		if (from == SCTP_ADDR_IS_CONFIRMED)
3915			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3916			net->dest_state = SCTP_ADDR_REACHABLE;
3917		else
3918			net->dest_state = SCTP_ADDR_REACHABLE |
3919			    SCTP_ADDR_UNCONFIRMED;
3920	}
3921	/*
3922	 * We set this to 0, the timer code knows that this means its an
3923	 * initial value
3924	 */
3925	net->rto_needed = 1;
3926	net->RTO = 0;
3927	net->RTO_measured = 0;
3928	stcb->asoc.numnets++;
3929	net->ref_count = 1;
3930	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3931	net->port = stcb->asoc.port;
3932	net->dscp = stcb->asoc.default_dscp;
3933#ifdef INET6
3934	net->flowlabel = stcb->asoc.default_flowlabel;
3935#endif
3936	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3937		net->dest_state |= SCTP_ADDR_NOHB;
3938	} else {
3939		net->dest_state &= ~SCTP_ADDR_NOHB;
3940	}
3941	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3942		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3943	} else {
3944		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3945	}
3946	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3947	/* Init the timer structure */
3948	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3949	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3950	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3951
3952	/* Now generate a route for this guy */
3953#ifdef INET6
3954	/* KAME hack: embed scopeid */
3955	if (newaddr->sa_family == AF_INET6) {
3956		struct sockaddr_in6 *sin6;
3957
3958		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3959		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3960		sin6->sin6_scope_id = 0;
3961	}
3962#endif
3963	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3964
3965	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3966		/* Get source address */
3967		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3968		    stcb,
3969		    (sctp_route_t *) & net->ro,
3970		    net,
3971		    0,
3972		    stcb->asoc.vrf_id);
3973		if (net->ro._s_addr != NULL) {
3974			net->src_addr_selected = 1;
3975			/* Now get the interface MTU */
3976			if (net->ro._s_addr->ifn_p != NULL) {
3977				net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3978			}
3979		} else {
3980			net->src_addr_selected = 0;
3981		}
3982		if (net->mtu > 0) {
3983			uint32_t rmtu;
3984
3985			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3986			if (rmtu == 0) {
3987				/*
3988				 * Start things off to match mtu of
3989				 * interface please.
3990				 */
3991				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3992				    net->ro.ro_rt, net->mtu);
3993			} else {
3994				/*
3995				 * we take the route mtu over the interface,
3996				 * since the route may be leading out the
3997				 * loopback, or a different interface.
3998				 */
3999				net->mtu = rmtu;
4000			}
4001		}
4002	} else {
4003		net->src_addr_selected = 0;
4004	}
4005	if (net->mtu == 0) {
4006		switch (newaddr->sa_family) {
4007#ifdef INET
4008		case AF_INET:
4009			net->mtu = SCTP_DEFAULT_MTU;
4010			break;
4011#endif
4012#ifdef INET6
4013		case AF_INET6:
4014			net->mtu = 1280;
4015			break;
4016#endif
4017		default:
4018			break;
4019		}
4020	}
4021#if defined(INET) || defined(INET6)
4022	if (net->port) {
4023		net->mtu -= (uint32_t) sizeof(struct udphdr);
4024	}
4025#endif
4026	if (from == SCTP_ALLOC_ASOC) {
4027		stcb->asoc.smallest_mtu = net->mtu;
4028	}
4029	if (stcb->asoc.smallest_mtu > net->mtu) {
4030		stcb->asoc.smallest_mtu = net->mtu;
4031	}
4032#ifdef INET6
4033	if (newaddr->sa_family == AF_INET6) {
4034		struct sockaddr_in6 *sin6;
4035
4036		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4037		(void)sa6_recoverscope(sin6);
4038	}
4039#endif
4040
4041	/* JRS - Use the congestion control given in the CC module */
4042	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4043		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4044
4045	/*
4046	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4047	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4048	 */
4049	net->find_pseudo_cumack = 1;
4050	net->find_rtx_pseudo_cumack = 1;
4051	/* Choose an initial flowid. */
4052	net->flowid = stcb->asoc.my_vtag ^
4053	    ntohs(stcb->rport) ^
4054	    ntohs(stcb->sctp_ep->sctp_lport);
4055#ifdef INVARIANTS
4056	net->flowidset = 1;
4057#endif
4058	if (netp) {
4059		*netp = net;
4060	}
4061	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4062	if (net->ro.ro_rt == NULL) {
4063		/* Since we have no route put it at the back */
4064		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4065	} else if (netfirst == NULL) {
4066		/* We are the first one in the pool. */
4067		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4068	} else if (netfirst->ro.ro_rt == NULL) {
4069		/*
4070		 * First one has NO route. Place this one ahead of the first
4071		 * one.
4072		 */
4073		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4074	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4075		/*
4076		 * This one has a different interface than the one at the
4077		 * top of the list. Place it ahead.
4078		 */
4079		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4080	} else {
4081		/*
4082		 * Ok we have the same interface as the first one. Move
4083		 * forward until we find either a) one with a NULL route...
4084		 * insert ahead of that b) one with a different ifp.. insert
4085		 * after that. c) end of the list.. insert at the tail.
4086		 */
4087		struct sctp_nets *netlook;
4088
4089		do {
4090			netlook = TAILQ_NEXT(netfirst, sctp_next);
4091			if (netlook == NULL) {
4092				/* End of the list */
4093				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4094				break;
4095			} else if (netlook->ro.ro_rt == NULL) {
4096				/* next one has NO route */
4097				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4098				break;
4099			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4100				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4101				    net, sctp_next);
4102				break;
4103			}
4104			/* Shift forward */
4105			netfirst = netlook;
4106		} while (netlook != NULL);
4107	}
4108
4109	/* got to have a primary set */
4110	if (stcb->asoc.primary_destination == 0) {
4111		stcb->asoc.primary_destination = net;
4112	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4113		    (net->ro.ro_rt) &&
4114	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4115		/* No route to current primary adopt new primary */
4116		stcb->asoc.primary_destination = net;
4117	}
4118	/* Validate primary is first */
4119	net = TAILQ_FIRST(&stcb->asoc.nets);
4120	if ((net != stcb->asoc.primary_destination) &&
4121	    (stcb->asoc.primary_destination)) {
4122		/*
4123		 * first one on the list is NOT the primary sctp_cmpaddr()
4124		 * is much more efficient if the primary is the first on the
4125		 * list, make it so.
4126		 */
4127		TAILQ_REMOVE(&stcb->asoc.nets,
4128		    stcb->asoc.primary_destination, sctp_next);
4129		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4130		    stcb->asoc.primary_destination, sctp_next);
4131	}
4132	return (0);
4133}
4134
4135
4136static uint32_t
4137sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4138{
4139	uint32_t id;
4140	struct sctpasochead *head;
4141	struct sctp_tcb *lstcb;
4142
4143	SCTP_INP_WLOCK(inp);
4144try_again:
4145	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4146		/* TSNH */
4147		SCTP_INP_WUNLOCK(inp);
4148		return (0);
4149	}
4150	/*
4151	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4152	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4153	 */
4154	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4155		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4156	}
4157	id = inp->sctp_associd_counter;
4158	inp->sctp_associd_counter++;
4159	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4160	if (lstcb) {
4161		goto try_again;
4162	}
4163	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4164	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4165	stcb->asoc.in_asocid_hash = 1;
4166	SCTP_INP_WUNLOCK(inp);
4167	return id;
4168}
4169
4170/*
4171 * allocate an association and add it to the endpoint. The caller must be
4172 * careful to add all additional addresses once they are know right away or
4173 * else the assoc will be may experience a blackout scenario.
4174 */
4175struct sctp_tcb *
4176sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4177    int *error, uint32_t override_tag, uint32_t vrf_id,
4178    struct thread *p
4179)
4180{
4181	/* note the p argument is only valid in unbound sockets */
4182
4183	struct sctp_tcb *stcb;
4184	struct sctp_association *asoc;
4185	struct sctpasochead *head;
4186	uint16_t rport;
4187	int err;
4188
4189	/*
4190	 * Assumption made here: Caller has done a
4191	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4192	 * address does not exist already.
4193	 */
4194	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4195		/* Hit max assoc, sorry no more */
4196		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4197		*error = ENOBUFS;
4198		return (NULL);
4199	}
4200	if (firstaddr == NULL) {
4201		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4202		*error = EINVAL;
4203		return (NULL);
4204	}
4205	SCTP_INP_RLOCK(inp);
4206	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4207	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4208	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4209		/*
4210		 * If its in the TCP pool, its NOT allowed to create an
4211		 * association. The parent listener needs to call
4212		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4213		 * off, or connected one does this.. its an error.
4214		 */
4215		SCTP_INP_RUNLOCK(inp);
4216		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4217		*error = EINVAL;
4218		return (NULL);
4219	}
4220	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4221	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4222		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4223		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4224			SCTP_INP_RUNLOCK(inp);
4225			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4226			*error = EINVAL;
4227			return (NULL);
4228		}
4229	}
4230	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4231#ifdef SCTP_DEBUG
4232	if (firstaddr) {
4233		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4234		switch (firstaddr->sa_family) {
4235#ifdef INET
4236		case AF_INET:
4237			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4238			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4239			break;
4240#endif
4241#ifdef INET6
4242		case AF_INET6:
4243			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4244			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4245			break;
4246#endif
4247		default:
4248			break;
4249		}
4250	} else {
4251		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4252	}
4253#endif				/* SCTP_DEBUG */
4254	switch (firstaddr->sa_family) {
4255#ifdef INET
4256	case AF_INET:
4257		{
4258			struct sockaddr_in *sin;
4259
4260			sin = (struct sockaddr_in *)firstaddr;
4261			if ((ntohs(sin->sin_port) == 0) ||
4262			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4263			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4264			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4265				/* Invalid address */
4266				SCTP_INP_RUNLOCK(inp);
4267				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4268				*error = EINVAL;
4269				return (NULL);
4270			}
4271			rport = sin->sin_port;
4272			break;
4273		}
4274#endif
4275#ifdef INET6
4276	case AF_INET6:
4277		{
4278			struct sockaddr_in6 *sin6;
4279
4280			sin6 = (struct sockaddr_in6 *)firstaddr;
4281			if ((ntohs(sin6->sin6_port) == 0) ||
4282			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4283			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4284				/* Invalid address */
4285				SCTP_INP_RUNLOCK(inp);
4286				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4287				*error = EINVAL;
4288				return (NULL);
4289			}
4290			rport = sin6->sin6_port;
4291			break;
4292		}
4293#endif
4294	default:
4295		/* not supported family type */
4296		SCTP_INP_RUNLOCK(inp);
4297		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4298		*error = EINVAL;
4299		return (NULL);
4300	}
4301	SCTP_INP_RUNLOCK(inp);
4302	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4303		/*
4304		 * If you have not performed a bind, then we need to do the
4305		 * ephemeral bind for you.
4306		 */
4307		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4308		    (struct sockaddr *)NULL,
4309		    (struct sctp_ifa *)NULL,
4310		    p
4311		    ))) {
4312			/* bind error, probably perm */
4313			*error = err;
4314			return (NULL);
4315		}
4316	}
4317	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4318	if (stcb == NULL) {
4319		/* out of memory? */
4320		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4321		*error = ENOMEM;
4322		return (NULL);
4323	}
4324	SCTP_INCR_ASOC_COUNT();
4325
4326	bzero(stcb, sizeof(*stcb));
4327	asoc = &stcb->asoc;
4328
4329	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4330	SCTP_TCB_LOCK_INIT(stcb);
4331	SCTP_TCB_SEND_LOCK_INIT(stcb);
4332	stcb->rport = rport;
4333	/* setup back pointer's */
4334	stcb->sctp_ep = inp;
4335	stcb->sctp_socket = inp->sctp_socket;
4336	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4337		/* failed */
4338		SCTP_TCB_LOCK_DESTROY(stcb);
4339		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4340		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4341		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4342		SCTP_DECR_ASOC_COUNT();
4343		*error = err;
4344		return (NULL);
4345	}
4346	/* and the port */
4347	SCTP_INP_INFO_WLOCK();
4348	SCTP_INP_WLOCK(inp);
4349	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4350		/* inpcb freed while alloc going on */
4351		SCTP_TCB_LOCK_DESTROY(stcb);
4352		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4353		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4354		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4355		SCTP_INP_WUNLOCK(inp);
4356		SCTP_INP_INFO_WUNLOCK();
4357		SCTP_DECR_ASOC_COUNT();
4358		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4359		*error = EINVAL;
4360		return (NULL);
4361	}
4362	SCTP_TCB_LOCK(stcb);
4363
4364	/* now that my_vtag is set, add it to the hash */
4365	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4366	/* put it in the bucket in the vtag hash of assoc's for the system */
4367	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4368	SCTP_INP_INFO_WUNLOCK();
4369
4370	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4371		/* failure.. memory error? */
4372		if (asoc->strmout) {
4373			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4374			asoc->strmout = NULL;
4375		}
4376		if (asoc->mapping_array) {
4377			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4378			asoc->mapping_array = NULL;
4379		}
4380		if (asoc->nr_mapping_array) {
4381			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4382			asoc->nr_mapping_array = NULL;
4383		}
4384		SCTP_DECR_ASOC_COUNT();
4385		SCTP_TCB_UNLOCK(stcb);
4386		SCTP_TCB_LOCK_DESTROY(stcb);
4387		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4388		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4389		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4390		SCTP_INP_WUNLOCK(inp);
4391		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4392		*error = ENOBUFS;
4393		return (NULL);
4394	}
4395	/* Init all the timers */
4396	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4397	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4398	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4399	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4400	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4401	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4402	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4403
4404	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4405	/* now file the port under the hash as well */
4406	if (inp->sctp_tcbhash != NULL) {
4407		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4408		    inp->sctp_hashmark)];
4409		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4410	}
4411	SCTP_INP_WUNLOCK(inp);
4412	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4413	return (stcb);
4414}
4415
4416
4417void
4418sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4419{
4420	struct sctp_association *asoc;
4421
4422	asoc = &stcb->asoc;
4423	asoc->numnets--;
4424	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4425	if (net == asoc->primary_destination) {
4426		/* Reset primary */
4427		struct sctp_nets *lnet;
4428
4429		lnet = TAILQ_FIRST(&asoc->nets);
4430		/*
4431		 * Mobility adaptation Ideally, if deleted destination is
4432		 * the primary, it becomes a fast retransmission trigger by
4433		 * the subsequent SET PRIMARY. (by micchie)
4434		 */
4435		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4436		    SCTP_MOBILITY_BASE) ||
4437		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4438		    SCTP_MOBILITY_FASTHANDOFF)) {
4439			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4440			if (asoc->deleted_primary != NULL) {
4441				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4442				goto out;
4443			}
4444			asoc->deleted_primary = net;
4445			atomic_add_int(&net->ref_count, 1);
4446			memset(&net->lastsa, 0, sizeof(net->lastsa));
4447			memset(&net->lastsv, 0, sizeof(net->lastsv));
4448			sctp_mobility_feature_on(stcb->sctp_ep,
4449			    SCTP_MOBILITY_PRIM_DELETED);
4450			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4451			    stcb->sctp_ep, stcb, NULL);
4452		}
4453out:
4454		/* Try to find a confirmed primary */
4455		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4456	}
4457	if (net == asoc->last_data_chunk_from) {
4458		/* Reset primary */
4459		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4460	}
4461	if (net == asoc->last_control_chunk_from) {
4462		/* Clear net */
4463		asoc->last_control_chunk_from = NULL;
4464	}
4465	if (net == stcb->asoc.alternate) {
4466		sctp_free_remote_addr(stcb->asoc.alternate);
4467		stcb->asoc.alternate = NULL;
4468	}
4469	sctp_free_remote_addr(net);
4470}
4471
4472/*
4473 * remove a remote endpoint address from an association, it will fail if the
4474 * address does not exist.
4475 */
4476int
4477sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4478{
4479	/*
4480	 * Here we need to remove a remote address. This is quite simple, we
4481	 * first find it in the list of address for the association
4482	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4483	 * on that item. Note we do not allow it to be removed if there are
4484	 * no other addresses.
4485	 */
4486	struct sctp_association *asoc;
4487	struct sctp_nets *net, *nnet;
4488
4489	asoc = &stcb->asoc;
4490
4491	/* locate the address */
4492	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4493		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4494			continue;
4495		}
4496		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4497		    remaddr)) {
4498			/* we found the guy */
4499			if (asoc->numnets < 2) {
4500				/* Must have at LEAST two remote addresses */
4501				return (-1);
4502			} else {
4503				sctp_remove_net(stcb, net);
4504				return (0);
4505			}
4506		}
4507	}
4508	/* not found. */
4509	return (-2);
4510}
4511
4512void
4513sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4514{
4515	struct sctpvtaghead *chain;
4516	struct sctp_tagblock *twait_block;
4517	int found = 0;
4518	int i;
4519
4520	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4521	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4522		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4523			if ((twait_block->vtag_block[i].v_tag == tag) &&
4524			    (twait_block->vtag_block[i].lport == lport) &&
4525			    (twait_block->vtag_block[i].rport == rport)) {
4526				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4527				twait_block->vtag_block[i].v_tag = 0;
4528				twait_block->vtag_block[i].lport = 0;
4529				twait_block->vtag_block[i].rport = 0;
4530				found = 1;
4531				break;
4532			}
4533		}
4534		if (found)
4535			break;
4536	}
4537}
4538
4539int
4540sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4541{
4542	struct sctpvtaghead *chain;
4543	struct sctp_tagblock *twait_block;
4544	int found = 0;
4545	int i;
4546
4547	SCTP_INP_INFO_WLOCK();
4548	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4549	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4550		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4551			if ((twait_block->vtag_block[i].v_tag == tag) &&
4552			    (twait_block->vtag_block[i].lport == lport) &&
4553			    (twait_block->vtag_block[i].rport == rport)) {
4554				found = 1;
4555				break;
4556			}
4557		}
4558		if (found)
4559			break;
4560	}
4561	SCTP_INP_INFO_WUNLOCK();
4562	return (found);
4563}
4564
4565
4566void
4567sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4568{
4569	struct sctpvtaghead *chain;
4570	struct sctp_tagblock *twait_block;
4571	struct timeval now;
4572	int set, i;
4573
4574	if (time == 0) {
4575		/* Its disabled */
4576		return;
4577	}
4578	(void)SCTP_GETTIME_TIMEVAL(&now);
4579	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4580	set = 0;
4581	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4582		/* Block(s) present, lets find space, and expire on the fly */
4583		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4584			if ((twait_block->vtag_block[i].v_tag == 0) &&
4585			    !set) {
4586				twait_block->vtag_block[i].tv_sec_at_expire =
4587				    now.tv_sec + time;
4588				twait_block->vtag_block[i].v_tag = tag;
4589				twait_block->vtag_block[i].lport = lport;
4590				twait_block->vtag_block[i].rport = rport;
4591				set = 1;
4592			} else if ((twait_block->vtag_block[i].v_tag) &&
4593			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4594				/* Audit expires this guy */
4595				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4596				twait_block->vtag_block[i].v_tag = 0;
4597				twait_block->vtag_block[i].lport = 0;
4598				twait_block->vtag_block[i].rport = 0;
4599				if (set == 0) {
4600					/* Reuse it for my new tag */
4601					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4602					twait_block->vtag_block[i].v_tag = tag;
4603					twait_block->vtag_block[i].lport = lport;
4604					twait_block->vtag_block[i].rport = rport;
4605					set = 1;
4606				}
4607			}
4608		}
4609		if (set) {
4610			/*
4611			 * We only do up to the block where we can place our
4612			 * tag for audits
4613			 */
4614			break;
4615		}
4616	}
4617	/* Need to add a new block to chain */
4618	if (!set) {
4619		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4620		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4621		if (twait_block == NULL) {
4622#ifdef INVARIANTS
4623			panic("Can not alloc tagblock");
4624#endif
4625			return;
4626		}
4627		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4628		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4629		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4630		twait_block->vtag_block[0].v_tag = tag;
4631		twait_block->vtag_block[0].lport = lport;
4632		twait_block->vtag_block[0].rport = rport;
4633	}
4634}
4635
4636
4637
4638/*-
4639 * Free the association after un-hashing the remote port. This
4640 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4641 * expect that the input to this function IS a locked TCB.
4642 * It will return 0, if it did NOT destroy the association (instead
4643 * it unlocks it. It will return NON-zero if it either destroyed the
4644 * association OR the association is already destroyed.
4645 */
4646int
4647sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4648{
4649	int i;
4650	struct sctp_association *asoc;
4651	struct sctp_nets *net, *nnet;
4652	struct sctp_laddr *laddr, *naddr;
4653	struct sctp_tmit_chunk *chk, *nchk;
4654	struct sctp_asconf_addr *aparam, *naparam;
4655	struct sctp_asconf_ack *aack, *naack;
4656	struct sctp_stream_reset_list *strrst, *nstrrst;
4657	struct sctp_queued_to_read *sq, *nsq;
4658	struct sctp_stream_queue_pending *sp, *nsp;
4659	sctp_sharedkey_t *shared_key, *nshared_key;
4660	struct socket *so;
4661
4662	/* first, lets purge the entry from the hash table. */
4663
4664#ifdef SCTP_LOG_CLOSING
4665	sctp_log_closing(inp, stcb, 6);
4666#endif
4667	if (stcb->asoc.state == 0) {
4668#ifdef SCTP_LOG_CLOSING
4669		sctp_log_closing(inp, NULL, 7);
4670#endif
4671		/* there is no asoc, really TSNH :-0 */
4672		return (1);
4673	}
4674	if (stcb->asoc.alternate) {
4675		sctp_free_remote_addr(stcb->asoc.alternate);
4676		stcb->asoc.alternate = NULL;
4677	}
4678	/* TEMP CODE */
4679	if (stcb->freed_from_where == 0) {
4680		/* Only record the first place free happened from */
4681		stcb->freed_from_where = from_location;
4682	}
4683	/* TEMP CODE */
4684
4685	asoc = &stcb->asoc;
4686	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4687	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4688		/* nothing around */
4689		so = NULL;
4690	else
4691		so = inp->sctp_socket;
4692
4693	/*
4694	 * We used timer based freeing if a reader or writer is in the way.
4695	 * So we first check if we are actually being called from a timer,
4696	 * if so we abort early if a reader or writer is still in the way.
4697	 */
4698	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4699	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4700		/*
4701		 * is it the timer driving us? if so are the reader/writers
4702		 * gone?
4703		 */
4704		if (stcb->asoc.refcnt) {
4705			/* nope, reader or writer in the way */
4706			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4707			/* no asoc destroyed */
4708			SCTP_TCB_UNLOCK(stcb);
4709#ifdef SCTP_LOG_CLOSING
4710			sctp_log_closing(inp, stcb, 8);
4711#endif
4712			return (0);
4713		}
4714	}
4715	/* now clean up any other timers */
4716	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4717	asoc->dack_timer.self = NULL;
4718	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4719	/*-
4720	 * For stream reset we don't blast this unless
4721	 * it is a str-reset timer, it might be the
4722	 * free-asoc timer which we DON'T want to
4723	 * disturb.
4724	 */
4725	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4726		asoc->strreset_timer.self = NULL;
4727	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4728	asoc->asconf_timer.self = NULL;
4729	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4730	asoc->autoclose_timer.self = NULL;
4731	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4732	asoc->shut_guard_timer.self = NULL;
4733	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4734	asoc->delayed_event_timer.self = NULL;
4735	/* Mobility adaptation */
4736	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4737	asoc->delete_prim_timer.self = NULL;
4738	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4739		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4740		net->rxt_timer.self = NULL;
4741		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4742		net->pmtu_timer.self = NULL;
4743		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4744		net->hb_timer.self = NULL;
4745	}
4746	/* Now the read queue needs to be cleaned up (only once) */
4747	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4748		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4749		SCTP_INP_READ_LOCK(inp);
4750		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4751			if (sq->stcb == stcb) {
4752				sq->do_not_ref_stcb = 1;
4753				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4754				/*
4755				 * If there is no end, there never will be
4756				 * now.
4757				 */
4758				if (sq->end_added == 0) {
4759					/* Held for PD-API clear that. */
4760					sq->pdapi_aborted = 1;
4761					sq->held_length = 0;
4762					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4763						/*
4764						 * Need to add a PD-API
4765						 * aborted indication.
4766						 * Setting the control_pdapi
4767						 * assures that it will be
4768						 * added right after this
4769						 * msg.
4770						 */
4771						uint32_t strseq;
4772
4773						stcb->asoc.control_pdapi = sq;
4774						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4775						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4776						    stcb,
4777						    SCTP_PARTIAL_DELIVERY_ABORTED,
4778						    (void *)&strseq,
4779						    SCTP_SO_LOCKED);
4780						stcb->asoc.control_pdapi = NULL;
4781					}
4782				}
4783				/* Add an end to wake them */
4784				sq->end_added = 1;
4785			}
4786		}
4787		SCTP_INP_READ_UNLOCK(inp);
4788		if (stcb->block_entry) {
4789			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4790			stcb->block_entry->error = ECONNRESET;
4791			stcb->block_entry = NULL;
4792		}
4793	}
4794	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4795		/*
4796		 * Someone holds a reference OR the socket is unaccepted
4797		 * yet.
4798		 */
4799		if ((stcb->asoc.refcnt) ||
4800		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4801		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4802			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4803			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4804		}
4805		SCTP_TCB_UNLOCK(stcb);
4806		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4807		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4808			/* nothing around */
4809			so = NULL;
4810		if (so) {
4811			/* Wake any reader/writers */
4812			sctp_sorwakeup(inp, so);
4813			sctp_sowwakeup(inp, so);
4814		}
4815#ifdef SCTP_LOG_CLOSING
4816		sctp_log_closing(inp, stcb, 9);
4817#endif
4818		/* no asoc destroyed */
4819		return (0);
4820	}
4821#ifdef SCTP_LOG_CLOSING
4822	sctp_log_closing(inp, stcb, 10);
4823#endif
4824	/*
4825	 * When I reach here, no others want to kill the assoc yet.. and I
4826	 * own the lock. Now its possible an abort comes in when I do the
4827	 * lock exchange below to grab all the locks to do the final take
4828	 * out. to prevent this we increment the count, which will start a
4829	 * timer and blow out above thus assuring us that we hold exclusive
4830	 * killing of the asoc. Note that after getting back the TCB lock we
4831	 * will go ahead and increment the counter back up and stop any
4832	 * timer a passing stranger may have started :-S
4833	 */
4834	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4835		atomic_add_int(&stcb->asoc.refcnt, 1);
4836
4837		SCTP_TCB_UNLOCK(stcb);
4838		SCTP_INP_INFO_WLOCK();
4839		SCTP_INP_WLOCK(inp);
4840		SCTP_TCB_LOCK(stcb);
4841	}
4842	/* Double check the GONE flag */
4843	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4844	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4845		/* nothing around */
4846		so = NULL;
4847
4848	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4849	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4850		/*
4851		 * For TCP type we need special handling when we are
4852		 * connected. We also include the peel'ed off ones to.
4853		 */
4854		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4855			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4856			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4857			if (so) {
4858				SOCK_LOCK(so);
4859				if (so->so_rcv.sb_cc == 0) {
4860					so->so_state &= ~(SS_ISCONNECTING |
4861					    SS_ISDISCONNECTING |
4862					    SS_ISCONFIRMING |
4863					    SS_ISCONNECTED);
4864				}
4865				socantrcvmore_locked(so);
4866				sctp_sowwakeup(inp, so);
4867				sctp_sorwakeup(inp, so);
4868				SCTP_SOWAKEUP(so);
4869			}
4870		}
4871	}
4872	/*
4873	 * Make it invalid too, that way if its about to run it will abort
4874	 * and return.
4875	 */
4876	/* re-increment the lock */
4877	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4878		atomic_add_int(&stcb->asoc.refcnt, -1);
4879	}
4880	if (stcb->asoc.refcnt) {
4881		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4882		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4883		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4884			SCTP_INP_INFO_WUNLOCK();
4885			SCTP_INP_WUNLOCK(inp);
4886		}
4887		SCTP_TCB_UNLOCK(stcb);
4888		return (0);
4889	}
4890	asoc->state = 0;
4891	if (inp->sctp_tcbhash) {
4892		LIST_REMOVE(stcb, sctp_tcbhash);
4893	}
4894	if (stcb->asoc.in_asocid_hash) {
4895		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4896	}
4897	/* Now lets remove it from the list of ALL associations in the EP */
4898	LIST_REMOVE(stcb, sctp_tcblist);
4899	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4900		SCTP_INP_INCR_REF(inp);
4901		SCTP_INP_WUNLOCK(inp);
4902	}
4903	/* pull from vtag hash */
4904	LIST_REMOVE(stcb, sctp_asocs);
4905	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4906	    inp->sctp_lport, stcb->rport);
4907
4908	/*
4909	 * Now restop the timers to be sure this is paranoia at is finest!
4910	 */
4911	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4912	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4913	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4914	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4915	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4916	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4917	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4918	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4919		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4920		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4921		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4922	}
4923
4924	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4925	/*
4926	 * The chunk lists and such SHOULD be empty but we check them just
4927	 * in case.
4928	 */
4929	/* anything on the wheel needs to be removed */
4930	for (i = 0; i < asoc->streamoutcnt; i++) {
4931		struct sctp_stream_out *outs;
4932
4933		outs = &asoc->strmout[i];
4934		/* now clean up any chunks here */
4935		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4936			TAILQ_REMOVE(&outs->outqueue, sp, next);
4937			sctp_free_spbufspace(stcb, asoc, sp);
4938			if (sp->data) {
4939				if (so) {
4940					/* Still an open socket - report */
4941					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4942					    0, (void *)sp, SCTP_SO_LOCKED);
4943				}
4944				if (sp->data) {
4945					sctp_m_freem(sp->data);
4946					sp->data = NULL;
4947					sp->tail_mbuf = NULL;
4948					sp->length = 0;
4949				}
4950			}
4951			if (sp->net) {
4952				sctp_free_remote_addr(sp->net);
4953				sp->net = NULL;
4954			}
4955			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4956		}
4957	}
4958	/* sa_ignore FREED_MEMORY */
4959	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4960		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4961		SCTP_FREE(strrst, SCTP_M_STRESET);
4962	}
4963	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4964		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4965		if (sq->data) {
4966			sctp_m_freem(sq->data);
4967			sq->data = NULL;
4968		}
4969		sctp_free_remote_addr(sq->whoFrom);
4970		sq->whoFrom = NULL;
4971		sq->stcb = NULL;
4972		/* Free the ctl entry */
4973		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4974		SCTP_DECR_READQ_COUNT();
4975		/* sa_ignore FREED_MEMORY */
4976	}
4977	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4978		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4979		if (chk->data) {
4980			sctp_m_freem(chk->data);
4981			chk->data = NULL;
4982		}
4983		if (chk->holds_key_ref)
4984			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4985		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4986		SCTP_DECR_CHK_COUNT();
4987		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4988		asoc->free_chunk_cnt--;
4989		/* sa_ignore FREED_MEMORY */
4990	}
4991	/* pending send queue SHOULD be empty */
4992	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4993		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4994			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4995#ifdef INVARIANTS
4996		} else {
4997			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4998#endif
4999		}
5000		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
5001		if (chk->data) {
5002			if (so) {
5003				/* Still a socket? */
5004				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
5005				    0, chk, SCTP_SO_LOCKED);
5006			}
5007			if (chk->data) {
5008				sctp_m_freem(chk->data);
5009				chk->data = NULL;
5010			}
5011		}
5012		if (chk->holds_key_ref)
5013			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5014		if (chk->whoTo) {
5015			sctp_free_remote_addr(chk->whoTo);
5016			chk->whoTo = NULL;
5017		}
5018		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5019		SCTP_DECR_CHK_COUNT();
5020		/* sa_ignore FREED_MEMORY */
5021	}
5022	/* sent queue SHOULD be empty */
5023	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5024		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5025			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
5026				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
5027#ifdef INVARIANTS
5028			} else {
5029				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
5030#endif
5031			}
5032		}
5033		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5034		if (chk->data) {
5035			if (so) {
5036				/* Still a socket? */
5037				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5038				    0, chk, SCTP_SO_LOCKED);
5039			}
5040			if (chk->data) {
5041				sctp_m_freem(chk->data);
5042				chk->data = NULL;
5043			}
5044		}
5045		if (chk->holds_key_ref)
5046			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5047		sctp_free_remote_addr(chk->whoTo);
5048		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5049		SCTP_DECR_CHK_COUNT();
5050		/* sa_ignore FREED_MEMORY */
5051	}
5052#ifdef INVARIANTS
5053	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5054		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5055			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5056		}
5057	}
5058#endif
5059	/* control queue MAY not be empty */
5060	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5061		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5062		if (chk->data) {
5063			sctp_m_freem(chk->data);
5064			chk->data = NULL;
5065		}
5066		if (chk->holds_key_ref)
5067			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5068		sctp_free_remote_addr(chk->whoTo);
5069		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5070		SCTP_DECR_CHK_COUNT();
5071		/* sa_ignore FREED_MEMORY */
5072	}
5073	/* ASCONF queue MAY not be empty */
5074	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5075		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5076		if (chk->data) {
5077			sctp_m_freem(chk->data);
5078			chk->data = NULL;
5079		}
5080		if (chk->holds_key_ref)
5081			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5082		sctp_free_remote_addr(chk->whoTo);
5083		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5084		SCTP_DECR_CHK_COUNT();
5085		/* sa_ignore FREED_MEMORY */
5086	}
5087	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5088		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5089		if (chk->data) {
5090			sctp_m_freem(chk->data);
5091			chk->data = NULL;
5092		}
5093		if (chk->holds_key_ref)
5094			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5095		sctp_free_remote_addr(chk->whoTo);
5096		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5097		SCTP_DECR_CHK_COUNT();
5098		/* sa_ignore FREED_MEMORY */
5099	}
5100
5101	if (asoc->mapping_array) {
5102		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5103		asoc->mapping_array = NULL;
5104	}
5105	if (asoc->nr_mapping_array) {
5106		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5107		asoc->nr_mapping_array = NULL;
5108	}
5109	/* the stream outs */
5110	if (asoc->strmout) {
5111		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5112		asoc->strmout = NULL;
5113	}
5114	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5115	if (asoc->strmin) {
5116		struct sctp_queued_to_read *ctl, *nctl;
5117
5118		for (i = 0; i < asoc->streamincnt; i++) {
5119			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5120				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5121				sctp_free_remote_addr(ctl->whoFrom);
5122				if (ctl->data) {
5123					sctp_m_freem(ctl->data);
5124					ctl->data = NULL;
5125				}
5126				/*
5127				 * We don't free the address here since all
5128				 * the net's were freed above.
5129				 */
5130				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5131				SCTP_DECR_READQ_COUNT();
5132			}
5133		}
5134		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5135		asoc->strmin = NULL;
5136	}
5137	asoc->streamincnt = 0;
5138	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5139#ifdef INVARIANTS
5140		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5141			panic("no net's left alloc'ed, or list points to itself");
5142		}
5143#endif
5144		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5145		sctp_free_remote_addr(net);
5146	}
5147	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5148		/* sa_ignore FREED_MEMORY */
5149		sctp_remove_laddr(laddr);
5150	}
5151
5152	/* pending asconf (address) parameters */
5153	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5154		/* sa_ignore FREED_MEMORY */
5155		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5156		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5157	}
5158	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5159		/* sa_ignore FREED_MEMORY */
5160		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5161		if (aack->data != NULL) {
5162			sctp_m_freem(aack->data);
5163		}
5164		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5165	}
5166	/* clean up auth stuff */
5167	if (asoc->local_hmacs)
5168		sctp_free_hmaclist(asoc->local_hmacs);
5169	if (asoc->peer_hmacs)
5170		sctp_free_hmaclist(asoc->peer_hmacs);
5171
5172	if (asoc->local_auth_chunks)
5173		sctp_free_chunklist(asoc->local_auth_chunks);
5174	if (asoc->peer_auth_chunks)
5175		sctp_free_chunklist(asoc->peer_auth_chunks);
5176
5177	sctp_free_authinfo(&asoc->authinfo);
5178
5179	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5180		LIST_REMOVE(shared_key, next);
5181		sctp_free_sharedkey(shared_key);
5182		/* sa_ignore FREED_MEMORY */
5183	}
5184
5185	/* Insert new items here :> */
5186
5187	/* Get rid of LOCK */
5188	SCTP_TCB_UNLOCK(stcb);
5189	SCTP_TCB_LOCK_DESTROY(stcb);
5190	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5191	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5192		SCTP_INP_INFO_WUNLOCK();
5193		SCTP_INP_RLOCK(inp);
5194	}
5195#ifdef SCTP_TRACK_FREED_ASOCS
5196	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5197		/* now clean up the tasoc itself */
5198		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5199		SCTP_DECR_ASOC_COUNT();
5200	} else {
5201		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5202	}
5203#else
5204	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5205	SCTP_DECR_ASOC_COUNT();
5206#endif
5207	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5208		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5209			/*
5210			 * If its NOT the inp_free calling us AND sctp_close
5211			 * as been called, we call back...
5212			 */
5213			SCTP_INP_RUNLOCK(inp);
5214			/*
5215			 * This will start the kill timer (if we are the
5216			 * last one) since we hold an increment yet. But
5217			 * this is the only safe way to do this since
5218			 * otherwise if the socket closes at the same time
5219			 * we are here we might collide in the cleanup.
5220			 */
5221			sctp_inpcb_free(inp,
5222			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5223			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5224			SCTP_INP_DECR_REF(inp);
5225			goto out_of;
5226		} else {
5227			/* The socket is still open. */
5228			SCTP_INP_DECR_REF(inp);
5229		}
5230	}
5231	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5232		SCTP_INP_RUNLOCK(inp);
5233	}
5234out_of:
5235	/* destroyed the asoc */
5236#ifdef SCTP_LOG_CLOSING
5237	sctp_log_closing(inp, NULL, 11);
5238#endif
5239	return (1);
5240}
5241
5242
5243
5244/*
5245 * determine if a destination is "reachable" based upon the addresses bound
5246 * to the current endpoint (e.g. only v4 or v6 currently bound)
5247 */
5248/*
5249 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5250 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5251 * types bound as its endpoint
5252 */
5253int
5254sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5255{
5256	struct sctp_inpcb *inp;
5257	int answer;
5258
5259	/*
5260	 * No locks here, the TCB, in all cases is already locked and an
5261	 * assoc is up. There is either a INP lock by the caller applied (in
5262	 * asconf case when deleting an address) or NOT in the HB case,
5263	 * however if HB then the INP increment is up and the INP will not
5264	 * be removed (on top of the fact that we have a TCB lock). So we
5265	 * only want to read the sctp_flags, which is either bound-all or
5266	 * not.. no protection needed since once an assoc is up you can't be
5267	 * changing your binding.
5268	 */
5269	inp = stcb->sctp_ep;
5270	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5271		/* if bound all, destination is not restricted */
5272		/*
5273		 * RRS: Question during lock work: Is this correct? If you
5274		 * are bound-all you still might need to obey the V4--V6
5275		 * flags??? IMO this bound-all stuff needs to be removed!
5276		 */
5277		return (1);
5278	}
5279	/* NOTE: all "scope" checks are done when local addresses are added */
5280	switch (destaddr->sa_family) {
5281#ifdef INET6
5282	case AF_INET6:
5283		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5284		break;
5285#endif
5286#ifdef INET
5287	case AF_INET:
5288		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5289		break;
5290#endif
5291	default:
5292		/* invalid family, so it's unreachable */
5293		answer = 0;
5294		break;
5295	}
5296	return (answer);
5297}
5298
5299/*
5300 * update the inp_vflags on an endpoint
5301 */
5302static void
5303sctp_update_ep_vflag(struct sctp_inpcb *inp)
5304{
5305	struct sctp_laddr *laddr;
5306
5307	/* first clear the flag */
5308	inp->ip_inp.inp.inp_vflag = 0;
5309	/* set the flag based on addresses on the ep list */
5310	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5311		if (laddr->ifa == NULL) {
5312			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5313			    __FUNCTION__);
5314			continue;
5315		}
5316		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5317			continue;
5318		}
5319		switch (laddr->ifa->address.sa.sa_family) {
5320#ifdef INET6
5321		case AF_INET6:
5322			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5323			break;
5324#endif
5325#ifdef INET
5326		case AF_INET:
5327			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5328			break;
5329#endif
5330		default:
5331			break;
5332		}
5333	}
5334}
5335
5336/*
5337 * Add the address to the endpoint local address list There is nothing to be
5338 * done if we are bound to all addresses
5339 */
5340void
5341sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5342{
5343	struct sctp_laddr *laddr;
5344	int fnd, error = 0;
5345
5346	fnd = 0;
5347
5348	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5349		/* You are already bound to all. You have it already */
5350		return;
5351	}
5352#ifdef INET6
5353	if (ifa->address.sa.sa_family == AF_INET6) {
5354		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5355			/* Can't bind a non-useable addr. */
5356			return;
5357		}
5358	}
5359#endif
5360	/* first, is it already present? */
5361	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5362		if (laddr->ifa == ifa) {
5363			fnd = 1;
5364			break;
5365		}
5366	}
5367
5368	if (fnd == 0) {
5369		/* Not in the ep list */
5370		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5371		if (error != 0)
5372			return;
5373		inp->laddr_count++;
5374		/* update inp_vflag flags */
5375		switch (ifa->address.sa.sa_family) {
5376#ifdef INET6
5377		case AF_INET6:
5378			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5379			break;
5380#endif
5381#ifdef INET
5382		case AF_INET:
5383			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5384			break;
5385#endif
5386		default:
5387			break;
5388		}
5389	}
5390	return;
5391}
5392
5393
5394/*
5395 * select a new (hopefully reachable) destination net (should only be used
5396 * when we deleted an ep addr that is the only usable source address to reach
5397 * the destination net)
5398 */
5399static void
5400sctp_select_primary_destination(struct sctp_tcb *stcb)
5401{
5402	struct sctp_nets *net;
5403
5404	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5405		/* for now, we'll just pick the first reachable one we find */
5406		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5407			continue;
5408		if (sctp_destination_is_reachable(stcb,
5409		    (struct sockaddr *)&net->ro._l_addr)) {
5410			/* found a reachable destination */
5411			stcb->asoc.primary_destination = net;
5412		}
5413	}
5414	/* I can't there from here! ...we're gonna die shortly... */
5415}
5416
5417
5418/*
5419 * Delete the address from the endpoint local address list There is nothing
5420 * to be done if we are bound to all addresses
5421 */
5422void
5423sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5424{
5425	struct sctp_laddr *laddr;
5426	int fnd;
5427
5428	fnd = 0;
5429	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5430		/* You are already bound to all. You have it already */
5431		return;
5432	}
5433	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5434		if (laddr->ifa == ifa) {
5435			fnd = 1;
5436			break;
5437		}
5438	}
5439	if (fnd && (inp->laddr_count < 2)) {
5440		/* can't delete unless there are at LEAST 2 addresses */
5441		return;
5442	}
5443	if (fnd) {
5444		/*
5445		 * clean up any use of this address go through our
5446		 * associations and clear any last_used_address that match
5447		 * this one for each assoc, see if a new primary_destination
5448		 * is needed
5449		 */
5450		struct sctp_tcb *stcb;
5451
5452		/* clean up "next_addr_touse" */
5453		if (inp->next_addr_touse == laddr)
5454			/* delete this address */
5455			inp->next_addr_touse = NULL;
5456
5457		/* clean up "last_used_address" */
5458		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5459			struct sctp_nets *net;
5460
5461			SCTP_TCB_LOCK(stcb);
5462			if (stcb->asoc.last_used_address == laddr)
5463				/* delete this address */
5464				stcb->asoc.last_used_address = NULL;
5465			/*
5466			 * Now spin through all the nets and purge any ref
5467			 * to laddr
5468			 */
5469			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5470				if (net->ro._s_addr &&
5471				    (net->ro._s_addr->ifa == laddr->ifa)) {
5472					/* Yep, purge src address selected */
5473					sctp_rtentry_t *rt;
5474
5475					/* delete this address if cached */
5476					rt = net->ro.ro_rt;
5477					if (rt != NULL) {
5478						RTFREE(rt);
5479						net->ro.ro_rt = NULL;
5480					}
5481					sctp_free_ifa(net->ro._s_addr);
5482					net->ro._s_addr = NULL;
5483					net->src_addr_selected = 0;
5484				}
5485			}
5486			SCTP_TCB_UNLOCK(stcb);
5487		}		/* for each tcb */
5488		/* remove it from the ep list */
5489		sctp_remove_laddr(laddr);
5490		inp->laddr_count--;
5491		/* update inp_vflag flags */
5492		sctp_update_ep_vflag(inp);
5493	}
5494	return;
5495}
5496
5497/*
5498 * Add the address to the TCB local address restricted list.
5499 * This is a "pending" address list (eg. addresses waiting for an
5500 * ASCONF-ACK response) and cannot be used as a valid source address.
5501 */
5502void
5503sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5504{
5505	struct sctp_laddr *laddr;
5506	struct sctpladdr *list;
5507
5508	/*
5509	 * Assumes TCB is locked.. and possibly the INP. May need to
5510	 * confirm/fix that if we need it and is not the case.
5511	 */
5512	list = &stcb->asoc.sctp_restricted_addrs;
5513
5514#ifdef INET6
5515	if (ifa->address.sa.sa_family == AF_INET6) {
5516		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5517			/* Can't bind a non-existent addr. */
5518			return;
5519		}
5520	}
5521#endif
5522	/* does the address already exist? */
5523	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5524		if (laddr->ifa == ifa) {
5525			return;
5526		}
5527	}
5528
5529	/* add to the list */
5530	(void)sctp_insert_laddr(list, ifa, 0);
5531	return;
5532}
5533
5534/*
5535 * insert an laddr entry with the given ifa for the desired list
5536 */
5537int
5538sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5539{
5540	struct sctp_laddr *laddr;
5541
5542	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5543	if (laddr == NULL) {
5544		/* out of memory? */
5545		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5546		return (EINVAL);
5547	}
5548	SCTP_INCR_LADDR_COUNT();
5549	bzero(laddr, sizeof(*laddr));
5550	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5551	laddr->ifa = ifa;
5552	laddr->action = act;
5553	atomic_add_int(&ifa->refcount, 1);
5554	/* insert it */
5555	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5556
5557	return (0);
5558}
5559
5560/*
5561 * Remove an laddr entry from the local address list (on an assoc)
5562 */
5563void
5564sctp_remove_laddr(struct sctp_laddr *laddr)
5565{
5566
5567	/* remove from the list */
5568	LIST_REMOVE(laddr, sctp_nxt_addr);
5569	sctp_free_ifa(laddr->ifa);
5570	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5571	SCTP_DECR_LADDR_COUNT();
5572}
5573
5574/*
5575 * Remove a local address from the TCB local address restricted list
5576 */
5577void
5578sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5579{
5580	struct sctp_inpcb *inp;
5581	struct sctp_laddr *laddr;
5582
5583	/*
5584	 * This is called by asconf work. It is assumed that a) The TCB is
5585	 * locked and b) The INP is locked. This is true in as much as I can
5586	 * trace through the entry asconf code where I did these locks.
5587	 * Again, the ASCONF code is a bit different in that it does lock
5588	 * the INP during its work often times. This must be since we don't
5589	 * want other proc's looking up things while what they are looking
5590	 * up is changing :-D
5591	 */
5592
5593	inp = stcb->sctp_ep;
5594	/* if subset bound and don't allow ASCONF's, can't delete last */
5595	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5596	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5597		if (stcb->sctp_ep->laddr_count < 2) {
5598			/* can't delete last address */
5599			return;
5600		}
5601	}
5602	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5603		/* remove the address if it exists */
5604		if (laddr->ifa == NULL)
5605			continue;
5606		if (laddr->ifa == ifa) {
5607			sctp_remove_laddr(laddr);
5608			return;
5609		}
5610	}
5611
5612	/* address not found! */
5613	return;
5614}
5615
5616/*
5617 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5618 */
5619/* sysctl */
5620static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5621static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5622
5623
5624
5625#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5626struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5627int *sctp_cpuarry = NULL;
5628void
5629sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5630{
5631	/* Queue a packet to a processor for the specified core */
5632	struct sctp_mcore_queue *qent;
5633	struct sctp_mcore_ctrl *wkq;
5634	int need_wake = 0;
5635
5636	if (sctp_mcore_workers == NULL) {
5637		/* Something went way bad during setup */
5638		sctp_input_with_port(m, off, 0);
5639		return;
5640	}
5641	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5642	    (sizeof(struct sctp_mcore_queue)),
5643	    SCTP_M_MCORE);
5644	if (qent == NULL) {
5645		/* This is trouble  */
5646		sctp_input_with_port(m, off, 0);
5647		return;
5648	}
5649	qent->vn = curvnet;
5650	qent->m = m;
5651	qent->off = off;
5652	qent->v6 = 0;
5653	wkq = &sctp_mcore_workers[cpu_to_use];
5654	SCTP_MCORE_QLOCK(wkq);
5655
5656	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5657	if (wkq->running == 0) {
5658		need_wake = 1;
5659	}
5660	SCTP_MCORE_QUNLOCK(wkq);
5661	if (need_wake) {
5662		wakeup(&wkq->running);
5663	}
5664}
5665
5666static void
5667sctp_mcore_thread(void *arg)
5668{
5669
5670	struct sctp_mcore_ctrl *wkq;
5671	struct sctp_mcore_queue *qent;
5672
5673	wkq = (struct sctp_mcore_ctrl *)arg;
5674	struct mbuf *m;
5675	int off, v6;
5676
5677	/* Wait for first tickle */
5678	SCTP_MCORE_LOCK(wkq);
5679	wkq->running = 0;
5680	msleep(&wkq->running,
5681	    &wkq->core_mtx,
5682	    0, "wait for pkt", 0);
5683	SCTP_MCORE_UNLOCK(wkq);
5684
5685	/* Bind to our cpu */
5686	thread_lock(curthread);
5687	sched_bind(curthread, wkq->cpuid);
5688	thread_unlock(curthread);
5689
5690	/* Now lets start working */
5691	SCTP_MCORE_LOCK(wkq);
5692	/* Now grab lock and go */
5693	for (;;) {
5694		SCTP_MCORE_QLOCK(wkq);
5695skip_sleep:
5696		wkq->running = 1;
5697		qent = TAILQ_FIRST(&wkq->que);
5698		if (qent) {
5699			TAILQ_REMOVE(&wkq->que, qent, next);
5700			SCTP_MCORE_QUNLOCK(wkq);
5701			CURVNET_SET(qent->vn);
5702			m = qent->m;
5703			off = qent->off;
5704			v6 = qent->v6;
5705			SCTP_FREE(qent, SCTP_M_MCORE);
5706			if (v6 == 0) {
5707				sctp_input_with_port(m, off, 0);
5708			} else {
5709				SCTP_PRINTF("V6 not yet supported\n");
5710				sctp_m_freem(m);
5711			}
5712			CURVNET_RESTORE();
5713			SCTP_MCORE_QLOCK(wkq);
5714		}
5715		wkq->running = 0;
5716		if (!TAILQ_EMPTY(&wkq->que)) {
5717			goto skip_sleep;
5718		}
5719		SCTP_MCORE_QUNLOCK(wkq);
5720		msleep(&wkq->running,
5721		    &wkq->core_mtx,
5722		    0, "wait for pkt", 0);
5723	}
5724}
5725
5726static void
5727sctp_startup_mcore_threads(void)
5728{
5729	int i, cpu;
5730
5731	if (mp_ncpus == 1)
5732		return;
5733
5734	if (sctp_mcore_workers != NULL) {
5735		/*
5736		 * Already been here in some previous vnet?
5737		 */
5738		return;
5739	}
5740	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5741	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5742	    SCTP_M_MCORE);
5743	if (sctp_mcore_workers == NULL) {
5744		/* TSNH I hope */
5745		return;
5746	}
5747	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5748	    sizeof(struct sctp_mcore_ctrl)));
5749	/* Init the structures */
5750	for (i = 0; i <= mp_maxid; i++) {
5751		TAILQ_INIT(&sctp_mcore_workers[i].que);
5752		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5753		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5754		sctp_mcore_workers[i].cpuid = i;
5755	}
5756	if (sctp_cpuarry == NULL) {
5757		SCTP_MALLOC(sctp_cpuarry, int *,
5758		    (mp_ncpus * sizeof(int)),
5759		    SCTP_M_MCORE);
5760		i = 0;
5761		CPU_FOREACH(cpu) {
5762			sctp_cpuarry[i] = cpu;
5763			i++;
5764		}
5765	}
5766	/* Now start them all */
5767	CPU_FOREACH(cpu) {
5768		(void)kproc_create(sctp_mcore_thread,
5769		    (void *)&sctp_mcore_workers[cpu],
5770		    &sctp_mcore_workers[cpu].thread_proc,
5771		    RFPROC,
5772		    SCTP_KTHREAD_PAGES,
5773		    SCTP_MCORE_NAME);
5774
5775	}
5776}
5777
5778#endif
5779
5780void
5781sctp_pcb_init()
5782{
5783	/*
5784	 * SCTP initialization for the PCB structures should be called by
5785	 * the sctp_init() funciton.
5786	 */
5787	int i;
5788	struct timeval tv;
5789
5790	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5791		/* error I was called twice */
5792		return;
5793	}
5794	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5795
5796#if defined(SCTP_LOCAL_TRACE_BUF)
5797	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5798#endif
5799#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5800	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5801	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5802	    SCTP_M_MCORE);
5803#endif
5804	(void)SCTP_GETTIME_TIMEVAL(&tv);
5805#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5806	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5807	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5808	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5809#else
5810	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5811	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5812	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5813#endif
5814	/* init the empty list of (All) Endpoints */
5815	LIST_INIT(&SCTP_BASE_INFO(listhead));
5816
5817
5818	/* init the hash table of endpoints */
5819	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5820	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5821	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5822	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5823	    &SCTP_BASE_INFO(hashasocmark));
5824	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5825	    &SCTP_BASE_INFO(hashmark));
5826	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5827	    &SCTP_BASE_INFO(hashtcpmark));
5828	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5829
5830
5831	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5832	    &SCTP_BASE_INFO(hashvrfmark));
5833
5834	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5835	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5836	/* init the zones */
5837	/*
5838	 * FIX ME: Should check for NULL returns, but if it does fail we are
5839	 * doomed to panic anyways... add later maybe.
5840	 */
5841	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5842	    sizeof(struct sctp_inpcb), maxsockets);
5843
5844	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5845	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5846
5847	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5848	    sizeof(struct sctp_laddr),
5849	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5850
5851	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5852	    sizeof(struct sctp_nets),
5853	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5854
5855	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5856	    sizeof(struct sctp_tmit_chunk),
5857	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5858
5859	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5860	    sizeof(struct sctp_queued_to_read),
5861	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5862
5863	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5864	    sizeof(struct sctp_stream_queue_pending),
5865	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5866
5867	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5868	    sizeof(struct sctp_asconf),
5869	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5870
5871	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5872	    sizeof(struct sctp_asconf_ack),
5873	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5874
5875
5876	/* Master Lock INIT for info structure */
5877	SCTP_INP_INFO_LOCK_INIT();
5878	SCTP_STATLOG_INIT_LOCK();
5879
5880	SCTP_IPI_COUNT_INIT();
5881	SCTP_IPI_ADDR_INIT();
5882#ifdef SCTP_PACKET_LOGGING
5883	SCTP_IP_PKTLOG_INIT();
5884#endif
5885	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5886
5887	SCTP_WQ_ADDR_INIT();
5888	/* not sure if we need all the counts */
5889	SCTP_BASE_INFO(ipi_count_ep) = 0;
5890	/* assoc/tcb zone info */
5891	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5892	/* local addrlist zone info */
5893	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5894	/* remote addrlist zone info */
5895	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5896	/* chunk info */
5897	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5898
5899	/* socket queue zone info */
5900	SCTP_BASE_INFO(ipi_count_readq) = 0;
5901
5902	/* stream out queue cont */
5903	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5904
5905	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5906	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5907
5908	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5909
5910	/* Init the TIMEWAIT list */
5911	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5912		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5913	}
5914	sctp_startup_iterator();
5915
5916#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5917	sctp_startup_mcore_threads();
5918#endif
5919
5920	/*
5921	 * INIT the default VRF which for BSD is the only one, other O/S's
5922	 * may have more. But initially they must start with one and then
5923	 * add the VRF's as addresses are added.
5924	 */
5925	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5926}
5927
5928/*
5929 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5930 */
5931void
5932sctp_pcb_finish(void)
5933{
5934	struct sctp_vrflist *vrf_bucket;
5935	struct sctp_vrf *vrf, *nvrf;
5936	struct sctp_ifn *ifn, *nifn;
5937	struct sctp_ifa *ifa, *nifa;
5938	struct sctpvtaghead *chain;
5939	struct sctp_tagblock *twait_block, *prev_twait_block;
5940	struct sctp_laddr *wi, *nwi;
5941	int i;
5942	struct sctp_iterator *it, *nit;
5943
5944	/*
5945	 * In FreeBSD the iterator thread never exits but we do clean up.
5946	 * The only way FreeBSD reaches here is if we have VRF's but we
5947	 * still add the ifdef to make it compile on old versions.
5948	 */
5949	SCTP_IPI_ITERATOR_WQ_LOCK();
5950	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5951		if (it->vn != curvnet) {
5952			continue;
5953		}
5954		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5955		if (it->function_atend != NULL) {
5956			(*it->function_atend) (it->pointer, it->val);
5957		}
5958		SCTP_FREE(it, SCTP_M_ITER);
5959	}
5960	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5961	SCTP_ITERATOR_LOCK();
5962	if ((sctp_it_ctl.cur_it) &&
5963	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5964		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5965	}
5966	SCTP_ITERATOR_UNLOCK();
5967	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5968	SCTP_WQ_ADDR_LOCK();
5969	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5970		LIST_REMOVE(wi, sctp_nxt_addr);
5971		SCTP_DECR_LADDR_COUNT();
5972		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5973	}
5974	SCTP_WQ_ADDR_UNLOCK();
5975
5976	/*
5977	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5978	 * destroyed first).
5979	 */
5980	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5981	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5982		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5983			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5984				/* free the ifa */
5985				LIST_REMOVE(ifa, next_bucket);
5986				LIST_REMOVE(ifa, next_ifa);
5987				SCTP_FREE(ifa, SCTP_M_IFA);
5988			}
5989			/* free the ifn */
5990			LIST_REMOVE(ifn, next_bucket);
5991			LIST_REMOVE(ifn, next_ifn);
5992			SCTP_FREE(ifn, SCTP_M_IFN);
5993		}
5994		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5995		/* free the vrf */
5996		LIST_REMOVE(vrf, next_vrf);
5997		SCTP_FREE(vrf, SCTP_M_VRF);
5998	}
5999	/* free the vrf hashes */
6000	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
6001	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
6002
6003	/*
6004	 * free the TIMEWAIT list elements malloc'd in the function
6005	 * sctp_add_vtag_to_timewait()...
6006	 */
6007	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
6008		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
6009		if (!LIST_EMPTY(chain)) {
6010			prev_twait_block = NULL;
6011			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6012				if (prev_twait_block) {
6013					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6014				}
6015				prev_twait_block = twait_block;
6016			}
6017			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6018		}
6019	}
6020
6021	/* free the locks and mutexes */
6022#ifdef SCTP_PACKET_LOGGING
6023	SCTP_IP_PKTLOG_DESTROY();
6024#endif
6025	SCTP_IPI_ADDR_DESTROY();
6026	SCTP_STATLOG_DESTROY();
6027	SCTP_INP_INFO_LOCK_DESTROY();
6028
6029	SCTP_WQ_ADDR_DESTROY();
6030
6031	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6032	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6033	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6034	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6035	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6036	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6037	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6038	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6039	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6040	/* Get rid of other stuff to */
6041	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
6042		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
6043	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
6044		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6045	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6046		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6047#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6048	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6049#endif
6050}
6051
6052
6053int
6054sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6055    int offset, int limit,
6056    struct sockaddr *src, struct sockaddr *dst,
6057    struct sockaddr *altsa)
6058{
6059	/*
6060	 * grub through the INIT pulling addresses and loading them to the
6061	 * nets structure in the asoc. The from address in the mbuf should
6062	 * also be loaded (if it is not already). This routine can be called
6063	 * with either INIT or INIT-ACK's as long as the m points to the IP
6064	 * packet and the offset points to the beginning of the parameters.
6065	 */
6066	struct sctp_inpcb *inp;
6067	struct sctp_nets *net, *nnet, *net_tmp;
6068	struct sctp_paramhdr *phdr, parm_buf;
6069	struct sctp_tcb *stcb_tmp;
6070	uint16_t ptype, plen;
6071	struct sockaddr *sa;
6072	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6073	struct sctp_auth_random *p_random = NULL;
6074	uint16_t random_len = 0;
6075	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6076	struct sctp_auth_hmac_algo *hmacs = NULL;
6077	uint16_t hmacs_len = 0;
6078	uint8_t saw_asconf = 0;
6079	uint8_t saw_asconf_ack = 0;
6080	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6081	struct sctp_auth_chunk_list *chunks = NULL;
6082	uint16_t num_chunks = 0;
6083	sctp_key_t *new_key;
6084	uint32_t keylen;
6085	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6086	uint8_t ecn_supported;
6087	uint8_t prsctp_supported;
6088	uint8_t nrsack_supported;
6089
6090#ifdef INET
6091	struct sockaddr_in sin;
6092
6093#endif
6094#ifdef INET6
6095	struct sockaddr_in6 sin6;
6096
6097#endif
6098
6099	/* First get the destination address setup too. */
6100#ifdef INET
6101	memset(&sin, 0, sizeof(sin));
6102	sin.sin_family = AF_INET;
6103	sin.sin_len = sizeof(sin);
6104	sin.sin_port = stcb->rport;
6105#endif
6106#ifdef INET6
6107	memset(&sin6, 0, sizeof(sin6));
6108	sin6.sin6_family = AF_INET6;
6109	sin6.sin6_len = sizeof(struct sockaddr_in6);
6110	sin6.sin6_port = stcb->rport;
6111#endif
6112	if (altsa) {
6113		sa = altsa;
6114	} else {
6115		sa = src;
6116	}
6117	/* Turn off ECN until we get through all params */
6118	ecn_supported = 0;
6119	prsctp_supported = 0;
6120	nrsack_supported = 0;
6121	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6122		/* mark all addresses that we have currently on the list */
6123		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6124	}
6125	/* does the source address already exist? if so skip it */
6126	inp = stcb->sctp_ep;
6127	atomic_add_int(&stcb->asoc.refcnt, 1);
6128	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6129	atomic_add_int(&stcb->asoc.refcnt, -1);
6130
6131	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6132		/* we must add the source address */
6133		/* no scope set here since we have a tcb already. */
6134		switch (sa->sa_family) {
6135#ifdef INET
6136		case AF_INET:
6137			if (stcb->asoc.scope.ipv4_addr_legal) {
6138				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6139					return (-1);
6140				}
6141			}
6142			break;
6143#endif
6144#ifdef INET6
6145		case AF_INET6:
6146			if (stcb->asoc.scope.ipv6_addr_legal) {
6147				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6148					return (-2);
6149				}
6150			}
6151			break;
6152#endif
6153		default:
6154			break;
6155		}
6156	} else {
6157		if (net_tmp != NULL && stcb_tmp == stcb) {
6158			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6159		} else if (stcb_tmp != stcb) {
6160			/* It belongs to another association? */
6161			if (stcb_tmp)
6162				SCTP_TCB_UNLOCK(stcb_tmp);
6163			return (-3);
6164		}
6165	}
6166	if (stcb->asoc.state == 0) {
6167		/* the assoc was freed? */
6168		return (-4);
6169	}
6170	/*
6171	 * peer must explicitly turn this on. This may have been initialized
6172	 * to be "on" in order to allow local addr changes while INIT's are
6173	 * in flight.
6174	 */
6175	stcb->asoc.peer_supports_asconf = 0;
6176	/* now we must go through each of the params. */
6177	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6178	while (phdr) {
6179		ptype = ntohs(phdr->param_type);
6180		plen = ntohs(phdr->param_length);
6181		/*
6182		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6183		 * (uint32_t)ptype, (int)plen);
6184		 */
6185		if (offset + plen > limit) {
6186			break;
6187		}
6188		if (plen == 0) {
6189			break;
6190		}
6191#ifdef INET
6192		if (ptype == SCTP_IPV4_ADDRESS) {
6193			if (stcb->asoc.scope.ipv4_addr_legal) {
6194				struct sctp_ipv4addr_param *p4, p4_buf;
6195
6196				/* ok get the v4 address and check/add */
6197				phdr = sctp_get_next_param(m, offset,
6198				    (struct sctp_paramhdr *)&p4_buf,
6199				    sizeof(p4_buf));
6200				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6201				    phdr == NULL) {
6202					return (-5);
6203				}
6204				p4 = (struct sctp_ipv4addr_param *)phdr;
6205				sin.sin_addr.s_addr = p4->addr;
6206				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6207					/* Skip multi-cast addresses */
6208					goto next_param;
6209				}
6210				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6211				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6212					goto next_param;
6213				}
6214				sa = (struct sockaddr *)&sin;
6215				inp = stcb->sctp_ep;
6216				atomic_add_int(&stcb->asoc.refcnt, 1);
6217				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6218				    dst, stcb);
6219				atomic_add_int(&stcb->asoc.refcnt, -1);
6220
6221				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6222				    inp == NULL) {
6223					/* we must add the source address */
6224					/*
6225					 * no scope set since we have a tcb
6226					 * already
6227					 */
6228
6229					/*
6230					 * we must validate the state again
6231					 * here
6232					 */
6233			add_it_now:
6234					if (stcb->asoc.state == 0) {
6235						/* the assoc was freed? */
6236						return (-7);
6237					}
6238					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6239						return (-8);
6240					}
6241				} else if (stcb_tmp == stcb) {
6242					if (stcb->asoc.state == 0) {
6243						/* the assoc was freed? */
6244						return (-10);
6245					}
6246					if (net != NULL) {
6247						/* clear flag */
6248						net->dest_state &=
6249						    ~SCTP_ADDR_NOT_IN_ASSOC;
6250					}
6251				} else {
6252					/*
6253					 * strange, address is in another
6254					 * assoc? straighten out locks.
6255					 */
6256					if (stcb_tmp) {
6257						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6258							/*
6259							 * in setup state we
6260							 * abort this guy
6261							 */
6262							sctp_abort_an_association(stcb_tmp->sctp_ep,
6263							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6264							goto add_it_now;
6265						}
6266						SCTP_TCB_UNLOCK(stcb_tmp);
6267					}
6268					if (stcb->asoc.state == 0) {
6269						/* the assoc was freed? */
6270						return (-12);
6271					}
6272					return (-13);
6273				}
6274			}
6275		} else
6276#endif
6277#ifdef INET6
6278		if (ptype == SCTP_IPV6_ADDRESS) {
6279			if (stcb->asoc.scope.ipv6_addr_legal) {
6280				/* ok get the v6 address and check/add */
6281				struct sctp_ipv6addr_param *p6, p6_buf;
6282
6283				phdr = sctp_get_next_param(m, offset,
6284				    (struct sctp_paramhdr *)&p6_buf,
6285				    sizeof(p6_buf));
6286				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6287				    phdr == NULL) {
6288					return (-14);
6289				}
6290				p6 = (struct sctp_ipv6addr_param *)phdr;
6291				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6292				    sizeof(p6->addr));
6293				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6294					/* Skip multi-cast addresses */
6295					goto next_param;
6296				}
6297				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6298					/*
6299					 * Link local make no sense without
6300					 * scope
6301					 */
6302					goto next_param;
6303				}
6304				sa = (struct sockaddr *)&sin6;
6305				inp = stcb->sctp_ep;
6306				atomic_add_int(&stcb->asoc.refcnt, 1);
6307				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6308				    dst, stcb);
6309				atomic_add_int(&stcb->asoc.refcnt, -1);
6310				if (stcb_tmp == NULL &&
6311				    (inp == stcb->sctp_ep || inp == NULL)) {
6312					/*
6313					 * we must validate the state again
6314					 * here
6315					 */
6316			add_it_now6:
6317					if (stcb->asoc.state == 0) {
6318						/* the assoc was freed? */
6319						return (-16);
6320					}
6321					/*
6322					 * we must add the address, no scope
6323					 * set
6324					 */
6325					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6326						return (-17);
6327					}
6328				} else if (stcb_tmp == stcb) {
6329					/*
6330					 * we must validate the state again
6331					 * here
6332					 */
6333					if (stcb->asoc.state == 0) {
6334						/* the assoc was freed? */
6335						return (-19);
6336					}
6337					if (net != NULL) {
6338						/* clear flag */
6339						net->dest_state &=
6340						    ~SCTP_ADDR_NOT_IN_ASSOC;
6341					}
6342				} else {
6343					/*
6344					 * strange, address is in another
6345					 * assoc? straighten out locks.
6346					 */
6347					if (stcb_tmp)
6348						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6349							/*
6350							 * in setup state we
6351							 * abort this guy
6352							 */
6353							sctp_abort_an_association(stcb_tmp->sctp_ep,
6354							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6355							goto add_it_now6;
6356						}
6357					SCTP_TCB_UNLOCK(stcb_tmp);
6358
6359					if (stcb->asoc.state == 0) {
6360						/* the assoc was freed? */
6361						return (-21);
6362					}
6363					return (-22);
6364				}
6365			}
6366		} else
6367#endif
6368		if (ptype == SCTP_ECN_CAPABLE) {
6369			ecn_supported = 1;
6370		} else if (ptype == SCTP_ULP_ADAPTATION) {
6371			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6372				struct sctp_adaptation_layer_indication ai,
6373				                                *aip;
6374
6375				phdr = sctp_get_next_param(m, offset,
6376				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6377				aip = (struct sctp_adaptation_layer_indication *)phdr;
6378				if (aip) {
6379					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6380					stcb->asoc.adaptation_needed = 1;
6381				}
6382			}
6383		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6384			struct sctp_asconf_addr_param lstore, *fee;
6385			int lptype;
6386			struct sockaddr *lsa = NULL;
6387
6388#ifdef INET
6389			struct sctp_asconf_addrv4_param *fii;
6390
6391#endif
6392
6393			stcb->asoc.peer_supports_asconf = 1;
6394			if (plen > sizeof(lstore)) {
6395				return (-23);
6396			}
6397			phdr = sctp_get_next_param(m, offset,
6398			    (struct sctp_paramhdr *)&lstore,
6399			    min(plen, sizeof(lstore)));
6400			if (phdr == NULL) {
6401				return (-24);
6402			}
6403			fee = (struct sctp_asconf_addr_param *)phdr;
6404			lptype = ntohs(fee->addrp.ph.param_type);
6405			switch (lptype) {
6406#ifdef INET
6407			case SCTP_IPV4_ADDRESS:
6408				if (plen !=
6409				    sizeof(struct sctp_asconf_addrv4_param)) {
6410					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6411					    (int)sizeof(struct sctp_asconf_addrv4_param),
6412					    plen);
6413				} else {
6414					fii = (struct sctp_asconf_addrv4_param *)fee;
6415					sin.sin_addr.s_addr = fii->addrp.addr;
6416					lsa = (struct sockaddr *)&sin;
6417				}
6418				break;
6419#endif
6420#ifdef INET6
6421			case SCTP_IPV6_ADDRESS:
6422				if (plen !=
6423				    sizeof(struct sctp_asconf_addr_param)) {
6424					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6425					    (int)sizeof(struct sctp_asconf_addr_param),
6426					    plen);
6427				} else {
6428					memcpy(sin6.sin6_addr.s6_addr,
6429					    fee->addrp.addr,
6430					    sizeof(fee->addrp.addr));
6431					lsa = (struct sockaddr *)&sin6;
6432				}
6433				break;
6434#endif
6435			default:
6436				break;
6437			}
6438			if (lsa) {
6439				(void)sctp_set_primary_addr(stcb, sa, NULL);
6440			}
6441		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6442			stcb->asoc.peer_supports_nat = 1;
6443		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6444			/* Peer supports pr-sctp */
6445			prsctp_supported = 1;
6446		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6447			/* A supported extension chunk */
6448			struct sctp_supported_chunk_types_param *pr_supported;
6449			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6450			int num_ent, i;
6451
6452			phdr = sctp_get_next_param(m, offset,
6453			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6454			if (phdr == NULL) {
6455				return (-25);
6456			}
6457			stcb->asoc.peer_supports_asconf = 0;
6458			stcb->asoc.peer_supports_pktdrop = 0;
6459			stcb->asoc.peer_supports_strreset = 0;
6460			stcb->asoc.peer_supports_auth = 0;
6461			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6462			num_ent = plen - sizeof(struct sctp_paramhdr);
6463			for (i = 0; i < num_ent; i++) {
6464				switch (pr_supported->chunk_types[i]) {
6465				case SCTP_ASCONF:
6466				case SCTP_ASCONF_ACK:
6467					stcb->asoc.peer_supports_asconf = 1;
6468					break;
6469				case SCTP_FORWARD_CUM_TSN:
6470					prsctp_supported = 1;
6471					break;
6472				case SCTP_PACKET_DROPPED:
6473					stcb->asoc.peer_supports_pktdrop = 1;
6474					break;
6475				case SCTP_NR_SELECTIVE_ACK:
6476					nrsack_supported = 1;
6477					break;
6478				case SCTP_STREAM_RESET:
6479					stcb->asoc.peer_supports_strreset = 1;
6480					break;
6481				case SCTP_AUTHENTICATION:
6482					stcb->asoc.peer_supports_auth = 1;
6483					break;
6484				default:
6485					/* one I have not learned yet */
6486					break;
6487
6488				}
6489			}
6490		} else if (ptype == SCTP_RANDOM) {
6491			if (plen > sizeof(random_store))
6492				break;
6493			if (got_random) {
6494				/* already processed a RANDOM */
6495				goto next_param;
6496			}
6497			phdr = sctp_get_next_param(m, offset,
6498			    (struct sctp_paramhdr *)random_store,
6499			    min(sizeof(random_store), plen));
6500			if (phdr == NULL)
6501				return (-26);
6502			p_random = (struct sctp_auth_random *)phdr;
6503			random_len = plen - sizeof(*p_random);
6504			/* enforce the random length */
6505			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6506				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6507				return (-27);
6508			}
6509			got_random = 1;
6510		} else if (ptype == SCTP_HMAC_LIST) {
6511			int num_hmacs;
6512			int i;
6513
6514			if (plen > sizeof(hmacs_store))
6515				break;
6516			if (got_hmacs) {
6517				/* already processed a HMAC list */
6518				goto next_param;
6519			}
6520			phdr = sctp_get_next_param(m, offset,
6521			    (struct sctp_paramhdr *)hmacs_store,
6522			    min(plen, sizeof(hmacs_store)));
6523			if (phdr == NULL)
6524				return (-28);
6525			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6526			hmacs_len = plen - sizeof(*hmacs);
6527			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6528			/* validate the hmac list */
6529			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6530				return (-29);
6531			}
6532			if (stcb->asoc.peer_hmacs != NULL)
6533				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6534			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6535			if (stcb->asoc.peer_hmacs != NULL) {
6536				for (i = 0; i < num_hmacs; i++) {
6537					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6538					    ntohs(hmacs->hmac_ids[i]));
6539				}
6540			}
6541			got_hmacs = 1;
6542		} else if (ptype == SCTP_CHUNK_LIST) {
6543			int i;
6544
6545			if (plen > sizeof(chunks_store))
6546				break;
6547			if (got_chklist) {
6548				/* already processed a Chunks list */
6549				goto next_param;
6550			}
6551			phdr = sctp_get_next_param(m, offset,
6552			    (struct sctp_paramhdr *)chunks_store,
6553			    min(plen, sizeof(chunks_store)));
6554			if (phdr == NULL)
6555				return (-30);
6556			chunks = (struct sctp_auth_chunk_list *)phdr;
6557			num_chunks = plen - sizeof(*chunks);
6558			if (stcb->asoc.peer_auth_chunks != NULL)
6559				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6560			else
6561				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6562			for (i = 0; i < num_chunks; i++) {
6563				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6564				    stcb->asoc.peer_auth_chunks);
6565				/* record asconf/asconf-ack if listed */
6566				if (chunks->chunk_types[i] == SCTP_ASCONF)
6567					saw_asconf = 1;
6568				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6569					saw_asconf_ack = 1;
6570
6571			}
6572			got_chklist = 1;
6573		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6574			    (ptype == SCTP_STATE_COOKIE) ||
6575			    (ptype == SCTP_UNRECOG_PARAM) ||
6576			    (ptype == SCTP_COOKIE_PRESERVE) ||
6577			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6578			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6579			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6580			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6581		    (ptype == SCTP_SUCCESS_REPORT)) {
6582			 /* don't care */ ;
6583		} else {
6584			if ((ptype & 0x8000) == 0x0000) {
6585				/*
6586				 * must stop processing the rest of the
6587				 * param's. Any report bits were handled
6588				 * with the call to
6589				 * sctp_arethere_unrecognized_parameters()
6590				 * when the INIT or INIT-ACK was first seen.
6591				 */
6592				break;
6593			}
6594		}
6595
6596next_param:
6597		offset += SCTP_SIZE32(plen);
6598		if (offset >= limit) {
6599			break;
6600		}
6601		phdr = sctp_get_next_param(m, offset, &parm_buf,
6602		    sizeof(parm_buf));
6603	}
6604	/* Now check to see if we need to purge any addresses */
6605	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6606		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6607		    SCTP_ADDR_NOT_IN_ASSOC) {
6608			/* This address has been removed from the asoc */
6609			/* remove and free it */
6610			stcb->asoc.numnets--;
6611			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6612			sctp_free_remote_addr(net);
6613			if (net == stcb->asoc.primary_destination) {
6614				stcb->asoc.primary_destination = NULL;
6615				sctp_select_primary_destination(stcb);
6616			}
6617		}
6618	}
6619	stcb->asoc.ecn_supported &= ecn_supported;
6620	stcb->asoc.prsctp_supported &= prsctp_supported;
6621	stcb->asoc.nrsack_supported &= nrsack_supported;
6622	/* validate authentication required parameters */
6623	if (got_random && got_hmacs) {
6624		stcb->asoc.peer_supports_auth = 1;
6625	} else {
6626		stcb->asoc.peer_supports_auth = 0;
6627	}
6628	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6629		/* peer does not support auth but sent a chunks list? */
6630		return (-31);
6631	}
6632	if (stcb->asoc.peer_supports_asconf && !stcb->asoc.peer_supports_auth) {
6633		/* peer supports asconf but not auth? */
6634		return (-32);
6635	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6636	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6637		return (-33);
6638	}
6639	/* concatenate the full random key */
6640	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6641	if (chunks != NULL) {
6642		keylen += sizeof(*chunks) + num_chunks;
6643	}
6644	new_key = sctp_alloc_key(keylen);
6645	if (new_key != NULL) {
6646		/* copy in the RANDOM */
6647		if (p_random != NULL) {
6648			keylen = sizeof(*p_random) + random_len;
6649			bcopy(p_random, new_key->key, keylen);
6650		}
6651		/* append in the AUTH chunks */
6652		if (chunks != NULL) {
6653			bcopy(chunks, new_key->key + keylen,
6654			    sizeof(*chunks) + num_chunks);
6655			keylen += sizeof(*chunks) + num_chunks;
6656		}
6657		/* append in the HMACs */
6658		if (hmacs != NULL) {
6659			bcopy(hmacs, new_key->key + keylen,
6660			    sizeof(*hmacs) + hmacs_len);
6661		}
6662	} else {
6663		/* failed to get memory for the key */
6664		return (-34);
6665	}
6666	if (stcb->asoc.authinfo.peer_random != NULL)
6667		sctp_free_key(stcb->asoc.authinfo.peer_random);
6668	stcb->asoc.authinfo.peer_random = new_key;
6669	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6670	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6671
6672	return (0);
6673}
6674
6675int
6676sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6677    struct sctp_nets *net)
6678{
6679	/* make sure the requested primary address exists in the assoc */
6680	if (net == NULL && sa)
6681		net = sctp_findnet(stcb, sa);
6682
6683	if (net == NULL) {
6684		/* didn't find the requested primary address! */
6685		return (-1);
6686	} else {
6687		/* set the primary address */
6688		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6689			/* Must be confirmed, so queue to set */
6690			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6691			return (0);
6692		}
6693		stcb->asoc.primary_destination = net;
6694		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6695			sctp_free_remote_addr(stcb->asoc.alternate);
6696			stcb->asoc.alternate = NULL;
6697		}
6698		net = TAILQ_FIRST(&stcb->asoc.nets);
6699		if (net != stcb->asoc.primary_destination) {
6700			/*
6701			 * first one on the list is NOT the primary
6702			 * sctp_cmpaddr() is much more efficient if the
6703			 * primary is the first on the list, make it so.
6704			 */
6705			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6706			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6707		}
6708		return (0);
6709	}
6710}
6711
6712int
6713sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6714{
6715	/*
6716	 * This function serves two purposes. It will see if a TAG can be
6717	 * re-used and return 1 for yes it is ok and 0 for don't use that
6718	 * tag. A secondary function it will do is purge out old tags that
6719	 * can be removed.
6720	 */
6721	struct sctpvtaghead *chain;
6722	struct sctp_tagblock *twait_block;
6723	struct sctpasochead *head;
6724	struct sctp_tcb *stcb;
6725	int i;
6726
6727	SCTP_INP_INFO_RLOCK();
6728	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6729	    SCTP_BASE_INFO(hashasocmark))];
6730	if (head == NULL) {
6731		/* invalid vtag */
6732		goto skip_vtag_check;
6733	}
6734	LIST_FOREACH(stcb, head, sctp_asocs) {
6735		/*
6736		 * We choose not to lock anything here. TCB's can't be
6737		 * removed since we have the read lock, so they can't be
6738		 * freed on us, same thing for the INP. I may be wrong with
6739		 * this assumption, but we will go with it for now :-)
6740		 */
6741		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6742			continue;
6743		}
6744		if (stcb->asoc.my_vtag == tag) {
6745			/* candidate */
6746			if (stcb->rport != rport) {
6747				continue;
6748			}
6749			if (stcb->sctp_ep->sctp_lport != lport) {
6750				continue;
6751			}
6752			/* Its a used tag set */
6753			SCTP_INP_INFO_RUNLOCK();
6754			return (0);
6755		}
6756	}
6757skip_vtag_check:
6758
6759	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6760	/* Now what about timed wait ? */
6761	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6762		/*
6763		 * Block(s) are present, lets see if we have this tag in the
6764		 * list
6765		 */
6766		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6767			if (twait_block->vtag_block[i].v_tag == 0) {
6768				/* not used */
6769				continue;
6770			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6771			    now->tv_sec) {
6772				/* Audit expires this guy */
6773				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6774				twait_block->vtag_block[i].v_tag = 0;
6775				twait_block->vtag_block[i].lport = 0;
6776				twait_block->vtag_block[i].rport = 0;
6777			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6778				    (twait_block->vtag_block[i].lport == lport) &&
6779			    (twait_block->vtag_block[i].rport == rport)) {
6780				/* Bad tag, sorry :< */
6781				SCTP_INP_INFO_RUNLOCK();
6782				return (0);
6783			}
6784		}
6785	}
6786	SCTP_INP_INFO_RUNLOCK();
6787	return (1);
6788}
6789
6790static void
6791sctp_drain_mbufs(struct sctp_tcb *stcb)
6792{
6793	/*
6794	 * We must hunt this association for MBUF's past the cumack (i.e.
6795	 * out of order data that we can renege on).
6796	 */
6797	struct sctp_association *asoc;
6798	struct sctp_tmit_chunk *chk, *nchk;
6799	uint32_t cumulative_tsn_p1;
6800	struct sctp_queued_to_read *ctl, *nctl;
6801	int cnt, strmat;
6802	uint32_t gap, i;
6803	int fnd = 0;
6804
6805	/* We look for anything larger than the cum-ack + 1 */
6806
6807	asoc = &stcb->asoc;
6808	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6809		/* none we can reneg on. */
6810		return;
6811	}
6812	SCTP_STAT_INCR(sctps_protocol_drains_done);
6813	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6814	cnt = 0;
6815	/* First look in the re-assembly queue */
6816	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6817		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6818			/* Yep it is above cum-ack */
6819			cnt++;
6820			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6821			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6822			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6823			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6824			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6825			if (chk->data) {
6826				sctp_m_freem(chk->data);
6827				chk->data = NULL;
6828			}
6829			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6830		}
6831	}
6832	/* Ok that was fun, now we will drain all the inbound streams? */
6833	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6834		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6835			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6836				/* Yep it is above cum-ack */
6837				cnt++;
6838				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6839				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6840				sctp_ucount_decr(asoc->cnt_on_all_streams);
6841				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6842				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6843				if (ctl->data) {
6844					sctp_m_freem(ctl->data);
6845					ctl->data = NULL;
6846				}
6847				sctp_free_remote_addr(ctl->whoFrom);
6848				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6849				SCTP_DECR_READQ_COUNT();
6850			}
6851		}
6852	}
6853	if (cnt) {
6854		/* We must back down to see what the new highest is */
6855		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6856			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6857			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6858				asoc->highest_tsn_inside_map = i;
6859				fnd = 1;
6860				break;
6861			}
6862		}
6863		if (!fnd) {
6864			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6865		}
6866		/*
6867		 * Question, should we go through the delivery queue? The
6868		 * only reason things are on here is the app not reading OR
6869		 * a p-d-api up. An attacker COULD send enough in to
6870		 * initiate the PD-API and then send a bunch of stuff to
6871		 * other streams... these would wind up on the delivery
6872		 * queue.. and then we would not get to them. But in order
6873		 * to do this I then have to back-track and un-deliver
6874		 * sequence numbers in streams.. el-yucko. I think for now
6875		 * we will NOT look at the delivery queue and leave it to be
6876		 * something to consider later. An alternative would be to
6877		 * abort the P-D-API with a notification and then deliver
6878		 * the data.... Or another method might be to keep track of
6879		 * how many times the situation occurs and if we see a
6880		 * possible attack underway just abort the association.
6881		 */
6882#ifdef SCTP_DEBUG
6883		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6884#endif
6885		/*
6886		 * Now do we need to find a new
6887		 * asoc->highest_tsn_inside_map?
6888		 */
6889		asoc->last_revoke_count = cnt;
6890		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6891		/* sa_ignore NO_NULL_CHK */
6892		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6893		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6894	}
6895	/*
6896	 * Another issue, in un-setting the TSN's in the mapping array we
6897	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6898	 * two things to occur. It may cause us to do extra work in checking
6899	 * for our mapping array movement. More importantly it may cause us
6900	 * to SACK every datagram. This may not be a bad thing though since
6901	 * we will recover once we get our cum-ack above and all this stuff
6902	 * we dumped recovered.
6903	 */
6904}
6905
6906void
6907sctp_drain()
6908{
6909	/*
6910	 * We must walk the PCB lists for ALL associations here. The system
6911	 * is LOW on MBUF's and needs help. This is where reneging will
6912	 * occur. We really hope this does NOT happen!
6913	 */
6914	VNET_ITERATOR_DECL(vnet_iter);
6915	VNET_LIST_RLOCK_NOSLEEP();
6916	VNET_FOREACH(vnet_iter) {
6917		CURVNET_SET(vnet_iter);
6918		struct sctp_inpcb *inp;
6919		struct sctp_tcb *stcb;
6920
6921		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6922		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6923#ifdef VIMAGE
6924			continue;
6925#else
6926			return;
6927#endif
6928		}
6929		SCTP_INP_INFO_RLOCK();
6930		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6931			/* For each endpoint */
6932			SCTP_INP_RLOCK(inp);
6933			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6934				/* For each association */
6935				SCTP_TCB_LOCK(stcb);
6936				sctp_drain_mbufs(stcb);
6937				SCTP_TCB_UNLOCK(stcb);
6938			}
6939			SCTP_INP_RUNLOCK(inp);
6940		}
6941		SCTP_INP_INFO_RUNLOCK();
6942		CURVNET_RESTORE();
6943	}
6944	VNET_LIST_RUNLOCK_NOSLEEP();
6945}
6946
6947/*
6948 * start a new iterator
6949 * iterates through all endpoints and associations based on the pcb_state
6950 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6951 * assocs and "ef" (optional) is executed when the iterator completes.
6952 * "inpf" (optional) is executed for each new endpoint as it is being
6953 * iterated through. inpe (optional) is called when the inp completes
6954 * its way through all the stcbs.
6955 */
6956int
6957sctp_initiate_iterator(inp_func inpf,
6958    asoc_func af,
6959    inp_func inpe,
6960    uint32_t pcb_state,
6961    uint32_t pcb_features,
6962    uint32_t asoc_state,
6963    void *argp,
6964    uint32_t argi,
6965    end_func ef,
6966    struct sctp_inpcb *s_inp,
6967    uint8_t chunk_output_off)
6968{
6969	struct sctp_iterator *it = NULL;
6970
6971	if (af == NULL) {
6972		return (-1);
6973	}
6974	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6975	    SCTP_M_ITER);
6976	if (it == NULL) {
6977		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6978		return (ENOMEM);
6979	}
6980	memset(it, 0, sizeof(*it));
6981	it->function_assoc = af;
6982	it->function_inp = inpf;
6983	if (inpf)
6984		it->done_current_ep = 0;
6985	else
6986		it->done_current_ep = 1;
6987	it->function_atend = ef;
6988	it->pointer = argp;
6989	it->val = argi;
6990	it->pcb_flags = pcb_state;
6991	it->pcb_features = pcb_features;
6992	it->asoc_state = asoc_state;
6993	it->function_inp_end = inpe;
6994	it->no_chunk_output = chunk_output_off;
6995	it->vn = curvnet;
6996	if (s_inp) {
6997		/* Assume lock is held here */
6998		it->inp = s_inp;
6999		SCTP_INP_INCR_REF(it->inp);
7000		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7001	} else {
7002		SCTP_INP_INFO_RLOCK();
7003		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7004		if (it->inp) {
7005			SCTP_INP_INCR_REF(it->inp);
7006		}
7007		SCTP_INP_INFO_RUNLOCK();
7008		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7009
7010	}
7011	SCTP_IPI_ITERATOR_WQ_LOCK();
7012
7013	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7014	if (sctp_it_ctl.iterator_running == 0) {
7015		sctp_wakeup_iterator();
7016	}
7017	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7018	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7019	return (0);
7020}
7021