sctp_pcb.c revision 296052
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 296052 2016-02-25 18:46:06Z tuexen $");
35
36#include <netinet/sctp_os.h>
37#include <sys/proc.h>
38#include <netinet/sctp_var.h>
39#include <netinet/sctp_sysctl.h>
40#include <netinet/sctp_pcb.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp.h>
43#include <netinet/sctp_header.h>
44#include <netinet/sctp_asconf.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_timer.h>
47#include <netinet/sctp_bsd_addr.h>
48#include <netinet/sctp_dtrace_define.h>
49#if defined(INET) || defined(INET6)
50#include <netinet/udp.h>
51#endif
52#ifdef INET6
53#include <netinet6/ip6_var.h>
54#endif
55#include <sys/sched.h>
56#include <sys/smp.h>
57#include <sys/unistd.h>
58
59
60VNET_DEFINE(struct sctp_base_info, system_base_info);
61
62/* FIX: we don't handle multiple link local scopes */
63/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
64#ifdef INET6
65int
66SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
67{
68	struct sockaddr_in6 tmp_a, tmp_b;
69
70	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
71	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
72		return (0);
73	}
74	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
75	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
76		return (0);
77	}
78	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
79}
80
81#endif
82
83void
84sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
85{
86	/*
87	 * We really don't need to lock this, but I will just because it
88	 * does not hurt.
89	 */
90	SCTP_INP_INFO_RLOCK();
91	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
92	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
93	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
94	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
95	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
96	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
97	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
98	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
99	SCTP_INP_INFO_RUNLOCK();
100}
101
102/*-
103 * Addresses are added to VRF's (Virtual Router's). For BSD we
104 * have only the default VRF 0. We maintain a hash list of
105 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
106 * these has a list of addresses. When we add a new address
107 * to a VRF we lookup the ifn/ifn_index, if the ifn does
108 * not exist we create it and add it to the list of IFN's
109 * within the VRF. Once we have the sctp_ifn, we add the
110 * address to the list. So we look something like:
111 *
112 * hash-vrf-table
113 *   vrf-> ifn-> ifn -> ifn
114 *   vrf    |
115 *    ...   +--ifa-> ifa -> ifa
116 *   vrf
117 *
118 * We keep these separate lists since the SCTP subsystem will
119 * point to these from its source address selection nets structure.
120 * When an address is deleted it does not happen right away on
121 * the SCTP side, it gets scheduled. What we do when a
122 * delete happens is immediately remove the address from
123 * the master list and decrement the refcount. As our
124 * addip iterator works through and frees the src address
125 * selection pointing to the sctp_ifa, eventually the refcount
126 * will reach 0 and we will delete it. Note that it is assumed
127 * that any locking on system level ifn/ifa is done at the
128 * caller of these functions and these routines will only
129 * lock the SCTP structures as they add or delete things.
130 *
131 * Other notes on VRF concepts.
132 *  - An endpoint can be in multiple VRF's
133 *  - An association lives within a VRF and only one VRF.
134 *  - Any incoming packet we can deduce the VRF for by
135 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
136 *  - Any downward send call or connect call must supply the
137 *    VRF via ancillary data or via some sort of set default
138 *    VRF socket option call (again for BSD no brainer since
139 *    the VRF is always 0).
140 *  - An endpoint may add multiple VRF's to it.
141 *  - Listening sockets can accept associations in any
142 *    of the VRF's they are in but the assoc will end up
143 *    in only one VRF (gotten from the packet or connect/send).
144 *
145 */
146
147struct sctp_vrf *
148sctp_allocate_vrf(int vrf_id)
149{
150	struct sctp_vrf *vrf = NULL;
151	struct sctp_vrflist *bucket;
152
153	/* First allocate the VRF structure */
154	vrf = sctp_find_vrf(vrf_id);
155	if (vrf) {
156		/* Already allocated */
157		return (vrf);
158	}
159	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
160	    SCTP_M_VRF);
161	if (vrf == NULL) {
162		/* No memory */
163#ifdef INVARIANTS
164		panic("No memory for VRF:%d", vrf_id);
165#endif
166		return (NULL);
167	}
168	/* setup the VRF */
169	memset(vrf, 0, sizeof(struct sctp_vrf));
170	vrf->vrf_id = vrf_id;
171	LIST_INIT(&vrf->ifnlist);
172	vrf->total_ifa_count = 0;
173	vrf->refcount = 0;
174	/* now also setup table ids */
175	SCTP_INIT_VRF_TABLEID(vrf);
176	/* Init the HASH of addresses */
177	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
178	    &vrf->vrf_addr_hashmark);
179	if (vrf->vrf_addr_hash == NULL) {
180		/* No memory */
181#ifdef INVARIANTS
182		panic("No memory for VRF:%d", vrf_id);
183#endif
184		SCTP_FREE(vrf, SCTP_M_VRF);
185		return (NULL);
186	}
187	/* Add it to the hash table */
188	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
189	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
190	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
191	return (vrf);
192}
193
194
195struct sctp_ifn *
196sctp_find_ifn(void *ifn, uint32_t ifn_index)
197{
198	struct sctp_ifn *sctp_ifnp;
199	struct sctp_ifnlist *hash_ifn_head;
200
201	/*
202	 * We assume the lock is held for the addresses if that's wrong
203	 * problems could occur :-)
204	 */
205	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
206	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
207		if (sctp_ifnp->ifn_index == ifn_index) {
208			return (sctp_ifnp);
209		}
210		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
211			return (sctp_ifnp);
212		}
213	}
214	return (NULL);
215}
216
217
218struct sctp_vrf *
219sctp_find_vrf(uint32_t vrf_id)
220{
221	struct sctp_vrflist *bucket;
222	struct sctp_vrf *liste;
223
224	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
225	LIST_FOREACH(liste, bucket, next_vrf) {
226		if (vrf_id == liste->vrf_id) {
227			return (liste);
228		}
229	}
230	return (NULL);
231}
232
233
234void
235sctp_free_vrf(struct sctp_vrf *vrf)
236{
237	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
238		if (vrf->vrf_addr_hash) {
239			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
240			vrf->vrf_addr_hash = NULL;
241		}
242		/* We zero'd the count */
243		LIST_REMOVE(vrf, next_vrf);
244		SCTP_FREE(vrf, SCTP_M_VRF);
245		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
246	}
247}
248
249
250void
251sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
252{
253	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
254		/* We zero'd the count */
255		if (sctp_ifnp->vrf) {
256			sctp_free_vrf(sctp_ifnp->vrf);
257		}
258		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
259		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
260	}
261}
262
263
264void
265sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
266{
267	struct sctp_ifn *sctp_ifnp;
268
269	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
270	if (sctp_ifnp != NULL) {
271		sctp_ifnp->ifn_mtu = mtu;
272	}
273}
274
275
276void
277sctp_free_ifa(struct sctp_ifa *sctp_ifap)
278{
279	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
280		/* We zero'd the count */
281		if (sctp_ifap->ifn_p) {
282			sctp_free_ifn(sctp_ifap->ifn_p);
283		}
284		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
285		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
286	}
287}
288
289
290static void
291sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
292{
293	struct sctp_ifn *found;
294
295	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
296	if (found == NULL) {
297		/* Not in the list.. sorry */
298		return;
299	}
300	if (hold_addr_lock == 0)
301		SCTP_IPI_ADDR_WLOCK();
302	LIST_REMOVE(sctp_ifnp, next_bucket);
303	LIST_REMOVE(sctp_ifnp, next_ifn);
304	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
305	    sctp_ifnp->registered_af);
306	if (hold_addr_lock == 0)
307		SCTP_IPI_ADDR_WUNLOCK();
308	/* Take away the reference, and possibly free it */
309	sctp_free_ifn(sctp_ifnp);
310}
311
312
313void
314sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
315    const char *if_name, uint32_t ifn_index)
316{
317	struct sctp_vrf *vrf;
318	struct sctp_ifa *sctp_ifap;
319
320	SCTP_IPI_ADDR_RLOCK();
321	vrf = sctp_find_vrf(vrf_id);
322	if (vrf == NULL) {
323		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
324		goto out;
325
326	}
327	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
328	if (sctp_ifap == NULL) {
329		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
330		goto out;
331	}
332	if (sctp_ifap->ifn_p == NULL) {
333		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
334		goto out;
335	}
336	if (if_name) {
337		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
338			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
339			    sctp_ifap->ifn_p->ifn_name, if_name);
340			goto out;
341		}
342	} else {
343		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
344			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
345			    sctp_ifap->ifn_p->ifn_index, ifn_index);
346			goto out;
347		}
348	}
349
350	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
351	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
352out:
353	SCTP_IPI_ADDR_RUNLOCK();
354}
355
356
357void
358sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
359    const char *if_name, uint32_t ifn_index)
360{
361	struct sctp_vrf *vrf;
362	struct sctp_ifa *sctp_ifap;
363
364	SCTP_IPI_ADDR_RLOCK();
365	vrf = sctp_find_vrf(vrf_id);
366	if (vrf == NULL) {
367		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
368		goto out;
369
370	}
371	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
372	if (sctp_ifap == NULL) {
373		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
374		goto out;
375	}
376	if (sctp_ifap->ifn_p == NULL) {
377		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
378		goto out;
379	}
380	if (if_name) {
381		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
382			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
383			    sctp_ifap->ifn_p->ifn_name, if_name);
384			goto out;
385		}
386	} else {
387		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
388			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
389			    sctp_ifap->ifn_p->ifn_index, ifn_index);
390			goto out;
391		}
392	}
393
394	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
395	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
396out:
397	SCTP_IPI_ADDR_RUNLOCK();
398}
399
400
401/*-
402 * Add an ifa to an ifn.
403 * Register the interface as necessary.
404 * NOTE: ADDR write lock MUST be held.
405 */
406static void
407sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
408{
409	int ifa_af;
410
411	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
412	sctp_ifap->ifn_p = sctp_ifnp;
413	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
414	/* update address counts */
415	sctp_ifnp->ifa_count++;
416	ifa_af = sctp_ifap->address.sa.sa_family;
417	switch (ifa_af) {
418#ifdef INET
419	case AF_INET:
420		sctp_ifnp->num_v4++;
421		break;
422#endif
423#ifdef INET6
424	case AF_INET6:
425		sctp_ifnp->num_v6++;
426		break;
427#endif
428	default:
429		break;
430	}
431	if (sctp_ifnp->ifa_count == 1) {
432		/* register the new interface */
433		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
434		sctp_ifnp->registered_af = ifa_af;
435	}
436}
437
438
439/*-
440 * Remove an ifa from its ifn.
441 * If no more addresses exist, remove the ifn too. Otherwise, re-register
442 * the interface based on the remaining address families left.
443 * NOTE: ADDR write lock MUST be held.
444 */
445static void
446sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
447{
448	LIST_REMOVE(sctp_ifap, next_ifa);
449	if (sctp_ifap->ifn_p) {
450		/* update address counts */
451		sctp_ifap->ifn_p->ifa_count--;
452		switch (sctp_ifap->address.sa.sa_family) {
453#ifdef INET
454		case AF_INET:
455			sctp_ifap->ifn_p->num_v4--;
456			break;
457#endif
458#ifdef INET6
459		case AF_INET6:
460			sctp_ifap->ifn_p->num_v6--;
461			break;
462#endif
463		default:
464			break;
465		}
466
467		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
468			/* remove the ifn, possibly freeing it */
469			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
470		} else {
471			/* re-register address family type, if needed */
472			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
473			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
474				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
475				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
476				sctp_ifap->ifn_p->registered_af = AF_INET;
477			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
478			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
479				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
480				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
481				sctp_ifap->ifn_p->registered_af = AF_INET6;
482			}
483			/* free the ifn refcount */
484			sctp_free_ifn(sctp_ifap->ifn_p);
485		}
486		sctp_ifap->ifn_p = NULL;
487	}
488}
489
490
491struct sctp_ifa *
492sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
493    uint32_t ifn_type, const char *if_name, void *ifa,
494    struct sockaddr *addr, uint32_t ifa_flags,
495    int dynamic_add)
496{
497	struct sctp_vrf *vrf;
498	struct sctp_ifn *sctp_ifnp = NULL;
499	struct sctp_ifa *sctp_ifap = NULL;
500	struct sctp_ifalist *hash_addr_head;
501	struct sctp_ifnlist *hash_ifn_head;
502	uint32_t hash_of_addr;
503	int new_ifn_af = 0;
504
505#ifdef SCTP_DEBUG
506	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
507	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
508#endif
509	SCTP_IPI_ADDR_WLOCK();
510	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
511	if (sctp_ifnp) {
512		vrf = sctp_ifnp->vrf;
513	} else {
514		vrf = sctp_find_vrf(vrf_id);
515		if (vrf == NULL) {
516			vrf = sctp_allocate_vrf(vrf_id);
517			if (vrf == NULL) {
518				SCTP_IPI_ADDR_WUNLOCK();
519				return (NULL);
520			}
521		}
522	}
523	if (sctp_ifnp == NULL) {
524		/*
525		 * build one and add it, can't hold lock until after malloc
526		 * done though.
527		 */
528		SCTP_IPI_ADDR_WUNLOCK();
529		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
530		    sizeof(struct sctp_ifn), SCTP_M_IFN);
531		if (sctp_ifnp == NULL) {
532#ifdef INVARIANTS
533			panic("No memory for IFN");
534#endif
535			return (NULL);
536		}
537		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
538		sctp_ifnp->ifn_index = ifn_index;
539		sctp_ifnp->ifn_p = ifn;
540		sctp_ifnp->ifn_type = ifn_type;
541		sctp_ifnp->refcount = 0;
542		sctp_ifnp->vrf = vrf;
543		atomic_add_int(&vrf->refcount, 1);
544		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
545		if (if_name != NULL) {
546			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
547		} else {
548			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
549		}
550		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
551		LIST_INIT(&sctp_ifnp->ifalist);
552		SCTP_IPI_ADDR_WLOCK();
553		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
554		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
555		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
556		new_ifn_af = 1;
557	}
558	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
559	if (sctp_ifap) {
560		/* Hmm, it already exists? */
561		if ((sctp_ifap->ifn_p) &&
562		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
563			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
564			    sctp_ifap->ifn_p->ifn_name, ifn_index,
565			    (void *)sctp_ifap);
566			if (new_ifn_af) {
567				/* Remove the created one that we don't want */
568				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
569			}
570			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
571				/* easy to solve, just switch back to active */
572				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
573				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
574				sctp_ifap->ifn_p = sctp_ifnp;
575				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
576			}
577	exit_stage_left:
578			SCTP_IPI_ADDR_WUNLOCK();
579			return (sctp_ifap);
580		} else {
581			if (sctp_ifap->ifn_p) {
582				/*
583				 * The last IFN gets the address, remove the
584				 * old one
585				 */
586				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
587				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
588				    sctp_ifap->ifn_p->ifn_index, if_name,
589				    ifn_index);
590				/* remove the address from the old ifn */
591				sctp_remove_ifa_from_ifn(sctp_ifap);
592				/* move the address over to the new ifn */
593				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
594				goto exit_stage_left;
595			} else {
596				/* repair ifnp which was NULL ? */
597				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
598				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
599				    (void *)sctp_ifnp, (void *)sctp_ifap);
600				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
601			}
602			goto exit_stage_left;
603		}
604	}
605	SCTP_IPI_ADDR_WUNLOCK();
606	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
607	if (sctp_ifap == NULL) {
608#ifdef INVARIANTS
609		panic("No memory for IFA");
610#endif
611		return (NULL);
612	}
613	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
614	sctp_ifap->ifn_p = sctp_ifnp;
615	atomic_add_int(&sctp_ifnp->refcount, 1);
616	sctp_ifap->vrf_id = vrf_id;
617	sctp_ifap->ifa = ifa;
618	memcpy(&sctp_ifap->address, addr, addr->sa_len);
619	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
620	sctp_ifap->flags = ifa_flags;
621	/* Set scope */
622	switch (sctp_ifap->address.sa.sa_family) {
623#ifdef INET
624	case AF_INET:
625		{
626			struct sockaddr_in *sin;
627
628			sin = &sctp_ifap->address.sin;
629			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
630			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
631				sctp_ifap->src_is_loop = 1;
632			}
633			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
634				sctp_ifap->src_is_priv = 1;
635			}
636			sctp_ifnp->num_v4++;
637			if (new_ifn_af)
638				new_ifn_af = AF_INET;
639			break;
640		}
641#endif
642#ifdef INET6
643	case AF_INET6:
644		{
645			/* ok to use deprecated addresses? */
646			struct sockaddr_in6 *sin6;
647
648			sin6 = &sctp_ifap->address.sin6;
649			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
650			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
651				sctp_ifap->src_is_loop = 1;
652			}
653			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
654				sctp_ifap->src_is_priv = 1;
655			}
656			sctp_ifnp->num_v6++;
657			if (new_ifn_af)
658				new_ifn_af = AF_INET6;
659			break;
660		}
661#endif
662	default:
663		new_ifn_af = 0;
664		break;
665	}
666	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
667
668	if ((sctp_ifap->src_is_priv == 0) &&
669	    (sctp_ifap->src_is_loop == 0)) {
670		sctp_ifap->src_is_glob = 1;
671	}
672	SCTP_IPI_ADDR_WLOCK();
673	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
674	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
675	sctp_ifap->refcount = 1;
676	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
677	sctp_ifnp->ifa_count++;
678	vrf->total_ifa_count++;
679	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
680	if (new_ifn_af) {
681		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
682		sctp_ifnp->registered_af = new_ifn_af;
683	}
684	SCTP_IPI_ADDR_WUNLOCK();
685	if (dynamic_add) {
686		/*
687		 * Bump up the refcount so that when the timer completes it
688		 * will drop back down.
689		 */
690		struct sctp_laddr *wi;
691
692		atomic_add_int(&sctp_ifap->refcount, 1);
693		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
694		if (wi == NULL) {
695			/*
696			 * Gak, what can we do? We have lost an address
697			 * change can you say HOSED?
698			 */
699			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
700			/* Opps, must decrement the count */
701			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
702			    if_name);
703			return (NULL);
704		}
705		SCTP_INCR_LADDR_COUNT();
706		bzero(wi, sizeof(*wi));
707		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
708		wi->ifa = sctp_ifap;
709		wi->action = SCTP_ADD_IP_ADDRESS;
710
711		SCTP_WQ_ADDR_LOCK();
712		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
713		SCTP_WQ_ADDR_UNLOCK();
714
715		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
716		    (struct sctp_inpcb *)NULL,
717		    (struct sctp_tcb *)NULL,
718		    (struct sctp_nets *)NULL);
719	} else {
720		/* it's ready for use */
721		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
722	}
723	return (sctp_ifap);
724}
725
726void
727sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
728    uint32_t ifn_index, const char *if_name)
729{
730	struct sctp_vrf *vrf;
731	struct sctp_ifa *sctp_ifap = NULL;
732
733	SCTP_IPI_ADDR_WLOCK();
734	vrf = sctp_find_vrf(vrf_id);
735	if (vrf == NULL) {
736		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
737		goto out_now;
738	}
739#ifdef SCTP_DEBUG
740	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
741	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
742#endif
743	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
744	if (sctp_ifap) {
745		/* Validate the delete */
746		if (sctp_ifap->ifn_p) {
747			int valid = 0;
748
749			/*-
750			 * The name has priority over the ifn_index
751			 * if its given. We do this especially for
752			 * panda who might recycle indexes fast.
753			 */
754			if (if_name) {
755				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
756					/* They match its a correct delete */
757					valid = 1;
758				}
759			}
760			if (!valid) {
761				/* last ditch check ifn_index */
762				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
763					valid = 1;
764				}
765			}
766			if (!valid) {
767				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
768				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
769				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
770				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
771				SCTP_IPI_ADDR_WUNLOCK();
772				return;
773			}
774		}
775		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
776		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
777		/*
778		 * We don't set the flag. This means that the structure will
779		 * hang around in EP's that have bound specific to it until
780		 * they close. This gives us TCP like behavior if someone
781		 * removes an address (or for that matter adds it right
782		 * back).
783		 */
784		/* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */
785		vrf->total_ifa_count--;
786		LIST_REMOVE(sctp_ifap, next_bucket);
787		sctp_remove_ifa_from_ifn(sctp_ifap);
788	}
789#ifdef SCTP_DEBUG
790	else {
791		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
792		    ifn_index);
793		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
794	}
795#endif
796
797out_now:
798	SCTP_IPI_ADDR_WUNLOCK();
799	if (sctp_ifap) {
800		struct sctp_laddr *wi;
801
802		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
803		if (wi == NULL) {
804			/*
805			 * Gak, what can we do? We have lost an address
806			 * change can you say HOSED?
807			 */
808			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
809
810			/* Oops, must decrement the count */
811			sctp_free_ifa(sctp_ifap);
812			return;
813		}
814		SCTP_INCR_LADDR_COUNT();
815		bzero(wi, sizeof(*wi));
816		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
817		wi->ifa = sctp_ifap;
818		wi->action = SCTP_DEL_IP_ADDRESS;
819		SCTP_WQ_ADDR_LOCK();
820		/*
821		 * Should this really be a tailq? As it is we will process
822		 * the newest first :-0
823		 */
824		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
825		SCTP_WQ_ADDR_UNLOCK();
826
827		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
828		    (struct sctp_inpcb *)NULL,
829		    (struct sctp_tcb *)NULL,
830		    (struct sctp_nets *)NULL);
831	}
832	return;
833}
834
835
836static int
837sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
838{
839	int loopback_scope;
840
841#if defined(INET)
842	int ipv4_local_scope, ipv4_addr_legal;
843
844#endif
845#if defined(INET6)
846	int local_scope, site_scope, ipv6_addr_legal;
847
848#endif
849	struct sctp_vrf *vrf;
850	struct sctp_ifn *sctp_ifn;
851	struct sctp_ifa *sctp_ifa;
852
853	loopback_scope = stcb->asoc.scope.loopback_scope;
854#if defined(INET)
855	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
856	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
857#endif
858#if defined(INET6)
859	local_scope = stcb->asoc.scope.local_scope;
860	site_scope = stcb->asoc.scope.site_scope;
861	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
862#endif
863
864	SCTP_IPI_ADDR_RLOCK();
865	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
866	if (vrf == NULL) {
867		/* no vrf, no addresses */
868		SCTP_IPI_ADDR_RUNLOCK();
869		return (0);
870	}
871	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
872		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
873			if ((loopback_scope == 0) &&
874			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
875				continue;
876			}
877			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
878				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
879				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
880					/*
881					 * We allow pending addresses, where
882					 * we have sent an asconf-add to be
883					 * considered valid.
884					 */
885					continue;
886				}
887				if (sctp_ifa->address.sa.sa_family != to->sa_family) {
888					continue;
889				}
890				switch (sctp_ifa->address.sa.sa_family) {
891#ifdef INET
892				case AF_INET:
893					if (ipv4_addr_legal) {
894						struct sockaddr_in *sin,
895						           *rsin;
896
897						sin = &sctp_ifa->address.sin;
898						rsin = (struct sockaddr_in *)to;
899						if ((ipv4_local_scope == 0) &&
900						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
901							continue;
902						}
903						if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred,
904						    &sin->sin_addr) != 0) {
905							continue;
906						}
907						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
908							SCTP_IPI_ADDR_RUNLOCK();
909							return (1);
910						}
911					}
912					break;
913#endif
914#ifdef INET6
915				case AF_INET6:
916					if (ipv6_addr_legal) {
917						struct sockaddr_in6 *sin6,
918						            *rsin6;
919
920						sin6 = &sctp_ifa->address.sin6;
921						rsin6 = (struct sockaddr_in6 *)to;
922						if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred,
923						    &sin6->sin6_addr) != 0) {
924							continue;
925						}
926						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
927							if (local_scope == 0)
928								continue;
929							if (sin6->sin6_scope_id == 0) {
930								if (sa6_recoverscope(sin6) != 0)
931									continue;
932							}
933						}
934						if ((site_scope == 0) &&
935						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
936							continue;
937						}
938						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
939							SCTP_IPI_ADDR_RUNLOCK();
940							return (1);
941						}
942					}
943					break;
944#endif
945				default:
946					/* TSNH */
947					break;
948				}
949			}
950		}
951	} else {
952		struct sctp_laddr *laddr;
953
954		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
955			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
956				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
957				continue;
958			}
959			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
960			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
961				/*
962				 * We allow pending addresses, where we have
963				 * sent an asconf-add to be considered
964				 * valid.
965				 */
966				continue;
967			}
968			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
969				continue;
970			}
971			switch (to->sa_family) {
972#ifdef INET
973			case AF_INET:
974				{
975					struct sockaddr_in *sin, *rsin;
976
977					sin = &laddr->ifa->address.sin;
978					rsin = (struct sockaddr_in *)to;
979					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
980						SCTP_IPI_ADDR_RUNLOCK();
981						return (1);
982					}
983					break;
984				}
985#endif
986#ifdef INET6
987			case AF_INET6:
988				{
989					struct sockaddr_in6 *sin6, *rsin6;
990
991					sin6 = &laddr->ifa->address.sin6;
992					rsin6 = (struct sockaddr_in6 *)to;
993					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
994						SCTP_IPI_ADDR_RUNLOCK();
995						return (1);
996					}
997					break;
998				}
999
1000#endif
1001			default:
1002				/* TSNH */
1003				break;
1004			}
1005
1006		}
1007	}
1008	SCTP_IPI_ADDR_RUNLOCK();
1009	return (0);
1010}
1011
1012
1013static struct sctp_tcb *
1014sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
1015    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
1016{
1017	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
1018	/*
1019	 * If we support the TCP model, then we must now dig through to see
1020	 * if we can find our endpoint in the list of tcp ep's.
1021	 */
1022	uint16_t lport, rport;
1023	struct sctppcbhead *ephead;
1024	struct sctp_inpcb *inp;
1025	struct sctp_laddr *laddr;
1026	struct sctp_tcb *stcb;
1027	struct sctp_nets *net;
1028
1029	if ((to == NULL) || (from == NULL)) {
1030		return (NULL);
1031	}
1032	switch (to->sa_family) {
1033#ifdef INET
1034	case AF_INET:
1035		if (from->sa_family == AF_INET) {
1036			lport = ((struct sockaddr_in *)to)->sin_port;
1037			rport = ((struct sockaddr_in *)from)->sin_port;
1038		} else {
1039			return (NULL);
1040		}
1041		break;
1042#endif
1043#ifdef INET6
1044	case AF_INET6:
1045		if (from->sa_family == AF_INET6) {
1046			lport = ((struct sockaddr_in6 *)to)->sin6_port;
1047			rport = ((struct sockaddr_in6 *)from)->sin6_port;
1048		} else {
1049			return (NULL);
1050		}
1051		break;
1052#endif
1053	default:
1054		return (NULL);
1055	}
1056	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
1057	/*
1058	 * Ok now for each of the guys in this bucket we must look and see:
1059	 * - Does the remote port match. - Does there single association's
1060	 * addresses match this address (to). If so we update p_ep to point
1061	 * to this ep and return the tcb from it.
1062	 */
1063	LIST_FOREACH(inp, ephead, sctp_hash) {
1064		SCTP_INP_RLOCK(inp);
1065		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1066			SCTP_INP_RUNLOCK(inp);
1067			continue;
1068		}
1069		if (lport != inp->sctp_lport) {
1070			SCTP_INP_RUNLOCK(inp);
1071			continue;
1072		}
1073		switch (to->sa_family) {
1074#ifdef INET
1075		case AF_INET:
1076			{
1077				struct sockaddr_in *sin;
1078
1079				sin = (struct sockaddr_in *)to;
1080				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1081				    &sin->sin_addr) != 0) {
1082					SCTP_INP_RUNLOCK(inp);
1083					continue;
1084				}
1085				break;
1086			}
1087#endif
1088#ifdef INET6
1089		case AF_INET6:
1090			{
1091				struct sockaddr_in6 *sin6;
1092
1093				sin6 = (struct sockaddr_in6 *)to;
1094				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1095				    &sin6->sin6_addr) != 0) {
1096					SCTP_INP_RUNLOCK(inp);
1097					continue;
1098				}
1099				break;
1100			}
1101#endif
1102		default:
1103			SCTP_INP_RUNLOCK(inp);
1104			continue;
1105		}
1106		if (inp->def_vrf_id != vrf_id) {
1107			SCTP_INP_RUNLOCK(inp);
1108			continue;
1109		}
1110		/* check to see if the ep has one of the addresses */
1111		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1112			/* We are NOT bound all, so look further */
1113			int match = 0;
1114
1115			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1116
1117				if (laddr->ifa == NULL) {
1118					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __func__);
1119					continue;
1120				}
1121				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1122					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1123					continue;
1124				}
1125				if (laddr->ifa->address.sa.sa_family ==
1126				    to->sa_family) {
1127					/* see if it matches */
1128#ifdef INET
1129					if (from->sa_family == AF_INET) {
1130						struct sockaddr_in *intf_addr,
1131						           *sin;
1132
1133						intf_addr = &laddr->ifa->address.sin;
1134						sin = (struct sockaddr_in *)to;
1135						if (sin->sin_addr.s_addr ==
1136						    intf_addr->sin_addr.s_addr) {
1137							match = 1;
1138							break;
1139						}
1140					}
1141#endif
1142#ifdef INET6
1143					if (from->sa_family == AF_INET6) {
1144						struct sockaddr_in6 *intf_addr6;
1145						struct sockaddr_in6 *sin6;
1146
1147						sin6 = (struct sockaddr_in6 *)
1148						    to;
1149						intf_addr6 = &laddr->ifa->address.sin6;
1150
1151						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1152						    intf_addr6)) {
1153							match = 1;
1154							break;
1155						}
1156					}
1157#endif
1158				}
1159			}
1160			if (match == 0) {
1161				/* This endpoint does not have this address */
1162				SCTP_INP_RUNLOCK(inp);
1163				continue;
1164			}
1165		}
1166		/*
1167		 * Ok if we hit here the ep has the address, does it hold
1168		 * the tcb?
1169		 */
1170		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1171		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1172		if (stcb == NULL) {
1173			SCTP_INP_RUNLOCK(inp);
1174			continue;
1175		}
1176		SCTP_TCB_LOCK(stcb);
1177		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1178			SCTP_TCB_UNLOCK(stcb);
1179			SCTP_INP_RUNLOCK(inp);
1180			continue;
1181		}
1182		if (stcb->rport != rport) {
1183			/* remote port does not match. */
1184			SCTP_TCB_UNLOCK(stcb);
1185			SCTP_INP_RUNLOCK(inp);
1186			continue;
1187		}
1188		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1189			SCTP_TCB_UNLOCK(stcb);
1190			SCTP_INP_RUNLOCK(inp);
1191			continue;
1192		}
1193		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1194			SCTP_TCB_UNLOCK(stcb);
1195			SCTP_INP_RUNLOCK(inp);
1196			continue;
1197		}
1198		/* Does this TCB have a matching address? */
1199		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1200
1201			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1202				/* not the same family, can't be a match */
1203				continue;
1204			}
1205			switch (from->sa_family) {
1206#ifdef INET
1207			case AF_INET:
1208				{
1209					struct sockaddr_in *sin, *rsin;
1210
1211					sin = (struct sockaddr_in *)&net->ro._l_addr;
1212					rsin = (struct sockaddr_in *)from;
1213					if (sin->sin_addr.s_addr ==
1214					    rsin->sin_addr.s_addr) {
1215						/* found it */
1216						if (netp != NULL) {
1217							*netp = net;
1218						}
1219						/*
1220						 * Update the endpoint
1221						 * pointer
1222						 */
1223						*inp_p = inp;
1224						SCTP_INP_RUNLOCK(inp);
1225						return (stcb);
1226					}
1227					break;
1228				}
1229#endif
1230#ifdef INET6
1231			case AF_INET6:
1232				{
1233					struct sockaddr_in6 *sin6, *rsin6;
1234
1235					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1236					rsin6 = (struct sockaddr_in6 *)from;
1237					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1238					    rsin6)) {
1239						/* found it */
1240						if (netp != NULL) {
1241							*netp = net;
1242						}
1243						/*
1244						 * Update the endpoint
1245						 * pointer
1246						 */
1247						*inp_p = inp;
1248						SCTP_INP_RUNLOCK(inp);
1249						return (stcb);
1250					}
1251					break;
1252				}
1253#endif
1254			default:
1255				/* TSNH */
1256				break;
1257			}
1258		}
1259		SCTP_TCB_UNLOCK(stcb);
1260		SCTP_INP_RUNLOCK(inp);
1261	}
1262	return (NULL);
1263}
1264
1265
1266/*
1267 * rules for use
1268 *
1269 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1270 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1271 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1272 * NULL.
1273 */
1274
1275struct sctp_tcb *
1276sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1277    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1278{
1279	struct sctpasochead *head;
1280	struct sctp_inpcb *inp;
1281	struct sctp_tcb *stcb = NULL;
1282	struct sctp_nets *net;
1283	uint16_t rport;
1284
1285	inp = *inp_p;
1286	switch (remote->sa_family) {
1287#ifdef INET
1288	case AF_INET:
1289		rport = (((struct sockaddr_in *)remote)->sin_port);
1290		break;
1291#endif
1292#ifdef INET6
1293	case AF_INET6:
1294		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1295		break;
1296#endif
1297	default:
1298		return (NULL);
1299	}
1300	if (locked_tcb) {
1301		/*
1302		 * UN-lock so we can do proper locking here this occurs when
1303		 * called from load_addresses_from_init.
1304		 */
1305		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1306		SCTP_TCB_UNLOCK(locked_tcb);
1307	}
1308	SCTP_INP_INFO_RLOCK();
1309	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1310	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1311		/*-
1312		 * Now either this guy is our listener or it's the
1313		 * connector. If it is the one that issued the connect, then
1314		 * it's only chance is to be the first TCB in the list. If
1315		 * it is the acceptor, then do the special_lookup to hash
1316		 * and find the real inp.
1317		 */
1318		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1319			/* to is peer addr, from is my addr */
1320			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1321			    netp, inp->def_vrf_id);
1322			if ((stcb != NULL) && (locked_tcb == NULL)) {
1323				/* we have a locked tcb, lower refcount */
1324				SCTP_INP_DECR_REF(inp);
1325			}
1326			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1327				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1328				SCTP_TCB_LOCK(locked_tcb);
1329				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1330				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1331			}
1332			SCTP_INP_INFO_RUNLOCK();
1333			return (stcb);
1334		} else {
1335			SCTP_INP_WLOCK(inp);
1336			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1337				goto null_return;
1338			}
1339			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1340			if (stcb == NULL) {
1341				goto null_return;
1342			}
1343			SCTP_TCB_LOCK(stcb);
1344
1345			if (stcb->rport != rport) {
1346				/* remote port does not match. */
1347				SCTP_TCB_UNLOCK(stcb);
1348				goto null_return;
1349			}
1350			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1351				SCTP_TCB_UNLOCK(stcb);
1352				goto null_return;
1353			}
1354			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1355				SCTP_TCB_UNLOCK(stcb);
1356				goto null_return;
1357			}
1358			/* now look at the list of remote addresses */
1359			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1360#ifdef INVARIANTS
1361				if (net == (TAILQ_NEXT(net, sctp_next))) {
1362					panic("Corrupt net list");
1363				}
1364#endif
1365				if (net->ro._l_addr.sa.sa_family !=
1366				    remote->sa_family) {
1367					/* not the same family */
1368					continue;
1369				}
1370				switch (remote->sa_family) {
1371#ifdef INET
1372				case AF_INET:
1373					{
1374						struct sockaddr_in *sin,
1375						           *rsin;
1376
1377						sin = (struct sockaddr_in *)
1378						    &net->ro._l_addr;
1379						rsin = (struct sockaddr_in *)remote;
1380						if (sin->sin_addr.s_addr ==
1381						    rsin->sin_addr.s_addr) {
1382							/* found it */
1383							if (netp != NULL) {
1384								*netp = net;
1385							}
1386							if (locked_tcb == NULL) {
1387								SCTP_INP_DECR_REF(inp);
1388							} else if (locked_tcb != stcb) {
1389								SCTP_TCB_LOCK(locked_tcb);
1390							}
1391							if (locked_tcb) {
1392								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1393							}
1394							SCTP_INP_WUNLOCK(inp);
1395							SCTP_INP_INFO_RUNLOCK();
1396							return (stcb);
1397						}
1398						break;
1399					}
1400#endif
1401#ifdef INET6
1402				case AF_INET6:
1403					{
1404						struct sockaddr_in6 *sin6,
1405						            *rsin6;
1406
1407						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1408						rsin6 = (struct sockaddr_in6 *)remote;
1409						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1410						    rsin6)) {
1411							/* found it */
1412							if (netp != NULL) {
1413								*netp = net;
1414							}
1415							if (locked_tcb == NULL) {
1416								SCTP_INP_DECR_REF(inp);
1417							} else if (locked_tcb != stcb) {
1418								SCTP_TCB_LOCK(locked_tcb);
1419							}
1420							if (locked_tcb) {
1421								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1422							}
1423							SCTP_INP_WUNLOCK(inp);
1424							SCTP_INP_INFO_RUNLOCK();
1425							return (stcb);
1426						}
1427						break;
1428					}
1429#endif
1430				default:
1431					/* TSNH */
1432					break;
1433				}
1434			}
1435			SCTP_TCB_UNLOCK(stcb);
1436		}
1437	} else {
1438		SCTP_INP_WLOCK(inp);
1439		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1440			goto null_return;
1441		}
1442		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1443		    inp->sctp_hashmark)];
1444		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1445			if (stcb->rport != rport) {
1446				/* remote port does not match */
1447				continue;
1448			}
1449			SCTP_TCB_LOCK(stcb);
1450			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1451				SCTP_TCB_UNLOCK(stcb);
1452				continue;
1453			}
1454			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1455				SCTP_TCB_UNLOCK(stcb);
1456				continue;
1457			}
1458			/* now look at the list of remote addresses */
1459			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1460#ifdef INVARIANTS
1461				if (net == (TAILQ_NEXT(net, sctp_next))) {
1462					panic("Corrupt net list");
1463				}
1464#endif
1465				if (net->ro._l_addr.sa.sa_family !=
1466				    remote->sa_family) {
1467					/* not the same family */
1468					continue;
1469				}
1470				switch (remote->sa_family) {
1471#ifdef INET
1472				case AF_INET:
1473					{
1474						struct sockaddr_in *sin,
1475						           *rsin;
1476
1477						sin = (struct sockaddr_in *)
1478						    &net->ro._l_addr;
1479						rsin = (struct sockaddr_in *)remote;
1480						if (sin->sin_addr.s_addr ==
1481						    rsin->sin_addr.s_addr) {
1482							/* found it */
1483							if (netp != NULL) {
1484								*netp = net;
1485							}
1486							if (locked_tcb == NULL) {
1487								SCTP_INP_DECR_REF(inp);
1488							} else if (locked_tcb != stcb) {
1489								SCTP_TCB_LOCK(locked_tcb);
1490							}
1491							if (locked_tcb) {
1492								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1493							}
1494							SCTP_INP_WUNLOCK(inp);
1495							SCTP_INP_INFO_RUNLOCK();
1496							return (stcb);
1497						}
1498						break;
1499					}
1500#endif
1501#ifdef INET6
1502				case AF_INET6:
1503					{
1504						struct sockaddr_in6 *sin6,
1505						            *rsin6;
1506
1507						sin6 = (struct sockaddr_in6 *)
1508						    &net->ro._l_addr;
1509						rsin6 = (struct sockaddr_in6 *)remote;
1510						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1511						    rsin6)) {
1512							/* found it */
1513							if (netp != NULL) {
1514								*netp = net;
1515							}
1516							if (locked_tcb == NULL) {
1517								SCTP_INP_DECR_REF(inp);
1518							} else if (locked_tcb != stcb) {
1519								SCTP_TCB_LOCK(locked_tcb);
1520							}
1521							if (locked_tcb) {
1522								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1523							}
1524							SCTP_INP_WUNLOCK(inp);
1525							SCTP_INP_INFO_RUNLOCK();
1526							return (stcb);
1527						}
1528						break;
1529					}
1530#endif
1531				default:
1532					/* TSNH */
1533					break;
1534				}
1535			}
1536			SCTP_TCB_UNLOCK(stcb);
1537		}
1538	}
1539null_return:
1540	/* clean up for returning null */
1541	if (locked_tcb) {
1542		SCTP_TCB_LOCK(locked_tcb);
1543		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1544	}
1545	SCTP_INP_WUNLOCK(inp);
1546	SCTP_INP_INFO_RUNLOCK();
1547	/* not found */
1548	return (NULL);
1549}
1550
1551
1552/*
1553 * Find an association for a specific endpoint using the association id given
1554 * out in the COMM_UP notification
1555 */
1556struct sctp_tcb *
1557sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1558{
1559	/*
1560	 * Use my the assoc_id to find a endpoint
1561	 */
1562	struct sctpasochead *head;
1563	struct sctp_tcb *stcb;
1564	uint32_t id;
1565
1566	if (inp == NULL) {
1567		SCTP_PRINTF("TSNH ep_associd\n");
1568		return (NULL);
1569	}
1570	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1571		SCTP_PRINTF("TSNH ep_associd0\n");
1572		return (NULL);
1573	}
1574	id = (uint32_t) asoc_id;
1575	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1576	if (head == NULL) {
1577		/* invalid id TSNH */
1578		SCTP_PRINTF("TSNH ep_associd1\n");
1579		return (NULL);
1580	}
1581	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1582		if (stcb->asoc.assoc_id == id) {
1583			if (inp != stcb->sctp_ep) {
1584				/*
1585				 * some other guy has the same id active (id
1586				 * collision ??).
1587				 */
1588				SCTP_PRINTF("TSNH ep_associd2\n");
1589				continue;
1590			}
1591			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1592				continue;
1593			}
1594			if (want_lock) {
1595				SCTP_TCB_LOCK(stcb);
1596			}
1597			return (stcb);
1598		}
1599	}
1600	return (NULL);
1601}
1602
1603
1604struct sctp_tcb *
1605sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1606{
1607	struct sctp_tcb *stcb;
1608
1609	SCTP_INP_RLOCK(inp);
1610	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1611	SCTP_INP_RUNLOCK(inp);
1612	return (stcb);
1613}
1614
1615
1616/*
1617 * Endpoint probe expects that the INP_INFO is locked.
1618 */
1619static struct sctp_inpcb *
1620sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1621    uint16_t lport, uint32_t vrf_id)
1622{
1623	struct sctp_inpcb *inp;
1624	struct sctp_laddr *laddr;
1625
1626#ifdef INET
1627	struct sockaddr_in *sin;
1628
1629#endif
1630#ifdef INET6
1631	struct sockaddr_in6 *sin6;
1632	struct sockaddr_in6 *intf_addr6;
1633
1634#endif
1635	int fnd;
1636
1637#ifdef INET
1638	sin = NULL;
1639#endif
1640#ifdef INET6
1641	sin6 = NULL;
1642#endif
1643	switch (nam->sa_family) {
1644#ifdef INET
1645	case AF_INET:
1646		sin = (struct sockaddr_in *)nam;
1647		break;
1648#endif
1649#ifdef INET6
1650	case AF_INET6:
1651		sin6 = (struct sockaddr_in6 *)nam;
1652		break;
1653#endif
1654	default:
1655		/* unsupported family */
1656		return (NULL);
1657	}
1658
1659	if (head == NULL)
1660		return (NULL);
1661
1662	LIST_FOREACH(inp, head, sctp_hash) {
1663		SCTP_INP_RLOCK(inp);
1664		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1665			SCTP_INP_RUNLOCK(inp);
1666			continue;
1667		}
1668		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1669		    (inp->sctp_lport == lport)) {
1670			/* got it */
1671			switch (nam->sa_family) {
1672#ifdef INET
1673			case AF_INET:
1674				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1675				    SCTP_IPV6_V6ONLY(inp)) {
1676					/*
1677					 * IPv4 on a IPv6 socket with ONLY
1678					 * IPv6 set
1679					 */
1680					SCTP_INP_RUNLOCK(inp);
1681					continue;
1682				}
1683				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1684				    &sin->sin_addr) != 0) {
1685					SCTP_INP_RUNLOCK(inp);
1686					continue;
1687				}
1688				break;
1689#endif
1690#ifdef INET6
1691			case AF_INET6:
1692				/*
1693				 * A V6 address and the endpoint is NOT
1694				 * bound V6
1695				 */
1696				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1697					SCTP_INP_RUNLOCK(inp);
1698					continue;
1699				}
1700				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1701				    &sin6->sin6_addr) != 0) {
1702					SCTP_INP_RUNLOCK(inp);
1703					continue;
1704				}
1705				break;
1706#endif
1707			default:
1708				break;
1709			}
1710			/* does a VRF id match? */
1711			fnd = 0;
1712			if (inp->def_vrf_id == vrf_id)
1713				fnd = 1;
1714
1715			SCTP_INP_RUNLOCK(inp);
1716			if (!fnd)
1717				continue;
1718			return (inp);
1719		}
1720		SCTP_INP_RUNLOCK(inp);
1721	}
1722	switch (nam->sa_family) {
1723#ifdef INET
1724	case AF_INET:
1725		if (sin->sin_addr.s_addr == INADDR_ANY) {
1726			/* Can't hunt for one that has no address specified */
1727			return (NULL);
1728		}
1729		break;
1730#endif
1731#ifdef INET6
1732	case AF_INET6:
1733		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1734			/* Can't hunt for one that has no address specified */
1735			return (NULL);
1736		}
1737		break;
1738#endif
1739	default:
1740		break;
1741	}
1742	/*
1743	 * ok, not bound to all so see if we can find a EP bound to this
1744	 * address.
1745	 */
1746	LIST_FOREACH(inp, head, sctp_hash) {
1747		SCTP_INP_RLOCK(inp);
1748		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1749			SCTP_INP_RUNLOCK(inp);
1750			continue;
1751		}
1752		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1753			SCTP_INP_RUNLOCK(inp);
1754			continue;
1755		}
1756		/*
1757		 * Ok this could be a likely candidate, look at all of its
1758		 * addresses
1759		 */
1760		if (inp->sctp_lport != lport) {
1761			SCTP_INP_RUNLOCK(inp);
1762			continue;
1763		}
1764		/* does a VRF id match? */
1765		fnd = 0;
1766		if (inp->def_vrf_id == vrf_id)
1767			fnd = 1;
1768
1769		if (!fnd) {
1770			SCTP_INP_RUNLOCK(inp);
1771			continue;
1772		}
1773		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1774			if (laddr->ifa == NULL) {
1775				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1776				    __func__);
1777				continue;
1778			}
1779			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1780			    (void *)laddr->ifa);
1781			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1782				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1783				continue;
1784			}
1785			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1786				/* possible, see if it matches */
1787				switch (nam->sa_family) {
1788#ifdef INET
1789				case AF_INET:
1790					if (sin->sin_addr.s_addr ==
1791					    laddr->ifa->address.sin.sin_addr.s_addr) {
1792						SCTP_INP_RUNLOCK(inp);
1793						return (inp);
1794					}
1795					break;
1796#endif
1797#ifdef INET6
1798				case AF_INET6:
1799					intf_addr6 = &laddr->ifa->address.sin6;
1800					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1801					    intf_addr6)) {
1802						SCTP_INP_RUNLOCK(inp);
1803						return (inp);
1804					}
1805					break;
1806#endif
1807				}
1808			}
1809		}
1810		SCTP_INP_RUNLOCK(inp);
1811	}
1812	return (NULL);
1813}
1814
1815
1816static struct sctp_inpcb *
1817sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1818{
1819	struct sctppcbhead *head;
1820	struct sctp_inpcb *t_inp;
1821	int fnd;
1822
1823	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1824	    SCTP_BASE_INFO(hashmark))];
1825	LIST_FOREACH(t_inp, head, sctp_hash) {
1826		if (t_inp->sctp_lport != lport) {
1827			continue;
1828		}
1829		/* is it in the VRF in question */
1830		fnd = 0;
1831		if (t_inp->def_vrf_id == vrf_id)
1832			fnd = 1;
1833		if (!fnd)
1834			continue;
1835
1836		/* This one is in use. */
1837		/* check the v6/v4 binding issue */
1838		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1839		    SCTP_IPV6_V6ONLY(t_inp)) {
1840			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1841				/* collision in V6 space */
1842				return (t_inp);
1843			} else {
1844				/* inp is BOUND_V4 no conflict */
1845				continue;
1846			}
1847		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1848			/* t_inp is bound v4 and v6, conflict always */
1849			return (t_inp);
1850		} else {
1851			/* t_inp is bound only V4 */
1852			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1853			    SCTP_IPV6_V6ONLY(inp)) {
1854				/* no conflict */
1855				continue;
1856			}
1857			/* else fall through to conflict */
1858		}
1859		return (t_inp);
1860	}
1861	return (NULL);
1862}
1863
1864
1865int
1866sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1867{
1868	/* For 1-2-1 with port reuse */
1869	struct sctppcbhead *head;
1870	struct sctp_inpcb *tinp, *ninp;
1871
1872	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1873		/* only works with port reuse on */
1874		return (-1);
1875	}
1876	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1877		return (0);
1878	}
1879	SCTP_INP_RUNLOCK(inp);
1880	SCTP_INP_INFO_WLOCK();
1881	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1882	    SCTP_BASE_INFO(hashmark))];
1883	/* Kick out all non-listeners to the TCP hash */
1884	LIST_FOREACH_SAFE(tinp, head, sctp_hash, ninp) {
1885		if (tinp->sctp_lport != inp->sctp_lport) {
1886			continue;
1887		}
1888		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1889			continue;
1890		}
1891		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1892			continue;
1893		}
1894		if (tinp->sctp_socket->so_qlimit) {
1895			continue;
1896		}
1897		SCTP_INP_WLOCK(tinp);
1898		LIST_REMOVE(tinp, sctp_hash);
1899		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1900		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1901		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1902		SCTP_INP_WUNLOCK(tinp);
1903	}
1904	SCTP_INP_WLOCK(inp);
1905	/* Pull from where he was */
1906	LIST_REMOVE(inp, sctp_hash);
1907	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1908	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1909	LIST_INSERT_HEAD(head, inp, sctp_hash);
1910	SCTP_INP_WUNLOCK(inp);
1911	SCTP_INP_RLOCK(inp);
1912	SCTP_INP_INFO_WUNLOCK();
1913	return (0);
1914}
1915
1916
1917struct sctp_inpcb *
1918sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1919    uint32_t vrf_id)
1920{
1921	/*
1922	 * First we check the hash table to see if someone has this port
1923	 * bound with just the port.
1924	 */
1925	struct sctp_inpcb *inp;
1926	struct sctppcbhead *head;
1927	int lport;
1928	unsigned int i;
1929
1930#ifdef INET
1931	struct sockaddr_in *sin;
1932
1933#endif
1934#ifdef INET6
1935	struct sockaddr_in6 *sin6;
1936
1937#endif
1938
1939	switch (nam->sa_family) {
1940#ifdef INET
1941	case AF_INET:
1942		sin = (struct sockaddr_in *)nam;
1943		lport = sin->sin_port;
1944		break;
1945#endif
1946#ifdef INET6
1947	case AF_INET6:
1948		sin6 = (struct sockaddr_in6 *)nam;
1949		lport = sin6->sin6_port;
1950		break;
1951#endif
1952	default:
1953		return (NULL);
1954	}
1955	/*
1956	 * I could cheat here and just cast to one of the types but we will
1957	 * do it right. It also provides the check against an Unsupported
1958	 * type too.
1959	 */
1960	/* Find the head of the ALLADDR chain */
1961	if (have_lock == 0) {
1962		SCTP_INP_INFO_RLOCK();
1963	}
1964	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1965	    SCTP_BASE_INFO(hashmark))];
1966	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1967
1968	/*
1969	 * If the TCP model exists it could be that the main listening
1970	 * endpoint is gone but there still exists a connected socket for
1971	 * this guy. If so we can return the first one that we find. This
1972	 * may NOT be the correct one so the caller should be wary on the
1973	 * returned INP. Currently the only caller that sets find_tcp_pool
1974	 * is in bindx where we are verifying that a user CAN bind the
1975	 * address. He either has bound it already, or someone else has, or
1976	 * its open to bind, so this is good enough.
1977	 */
1978	if (inp == NULL && find_tcp_pool) {
1979		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1980			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1981			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1982			if (inp) {
1983				break;
1984			}
1985		}
1986	}
1987	if (inp) {
1988		SCTP_INP_INCR_REF(inp);
1989	}
1990	if (have_lock == 0) {
1991		SCTP_INP_INFO_RUNLOCK();
1992	}
1993	return (inp);
1994}
1995
1996
1997/*
1998 * Find an association for an endpoint with the pointer to whom you want to
1999 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
2000 * need to change the *to to some other struct like a mbuf...
2001 */
2002struct sctp_tcb *
2003sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
2004    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
2005    uint32_t vrf_id)
2006{
2007	struct sctp_inpcb *inp = NULL;
2008	struct sctp_tcb *stcb;
2009
2010	SCTP_INP_INFO_RLOCK();
2011	if (find_tcp_pool) {
2012		if (inp_p != NULL) {
2013			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
2014			    vrf_id);
2015		} else {
2016			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
2017			    vrf_id);
2018		}
2019		if (stcb != NULL) {
2020			SCTP_INP_INFO_RUNLOCK();
2021			return (stcb);
2022		}
2023	}
2024	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
2025	if (inp_p != NULL) {
2026		*inp_p = inp;
2027	}
2028	SCTP_INP_INFO_RUNLOCK();
2029	if (inp == NULL) {
2030		return (NULL);
2031	}
2032	/*
2033	 * ok, we have an endpoint, now lets find the assoc for it (if any)
2034	 * we now place the source address or from in the to of the find
2035	 * endpoint call. Since in reality this chain is used from the
2036	 * inbound packet side.
2037	 */
2038	if (inp_p != NULL) {
2039		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
2040		    NULL);
2041	} else {
2042		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
2043		    NULL);
2044	}
2045	return (stcb);
2046}
2047
2048
2049/*
2050 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
2051 * find all addresses that the sender has specified in any address list. Each
2052 * address will be used to lookup the TCB and see if one exits.
2053 */
2054static struct sctp_tcb *
2055sctp_findassociation_special_addr(struct mbuf *m, int offset,
2056    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
2057    struct sockaddr *dst)
2058{
2059	struct sctp_paramhdr *phdr, parm_buf;
2060
2061#if defined(INET) || defined(INET6)
2062	struct sctp_tcb *stcb;
2063	uint16_t ptype;
2064
2065#endif
2066	uint16_t plen;
2067
2068#ifdef INET
2069	struct sockaddr_in sin4;
2070
2071#endif
2072#ifdef INET6
2073	struct sockaddr_in6 sin6;
2074
2075#endif
2076
2077#ifdef INET
2078	memset(&sin4, 0, sizeof(sin4));
2079	sin4.sin_len = sizeof(sin4);
2080	sin4.sin_family = AF_INET;
2081	sin4.sin_port = sh->src_port;
2082#endif
2083#ifdef INET6
2084	memset(&sin6, 0, sizeof(sin6));
2085	sin6.sin6_len = sizeof(sin6);
2086	sin6.sin6_family = AF_INET6;
2087	sin6.sin6_port = sh->src_port;
2088#endif
2089
2090	offset += sizeof(struct sctp_init_chunk);
2091
2092	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2093	while (phdr != NULL) {
2094		/* now we must see if we want the parameter */
2095#if defined(INET) || defined(INET6)
2096		ptype = ntohs(phdr->param_type);
2097#endif
2098		plen = ntohs(phdr->param_length);
2099		if (plen == 0) {
2100			break;
2101		}
2102#ifdef INET
2103		if (ptype == SCTP_IPV4_ADDRESS &&
2104		    plen == sizeof(struct sctp_ipv4addr_param)) {
2105			/* Get the rest of the address */
2106			struct sctp_ipv4addr_param ip4_parm, *p4;
2107
2108			phdr = sctp_get_next_param(m, offset,
2109			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2110			if (phdr == NULL) {
2111				return (NULL);
2112			}
2113			p4 = (struct sctp_ipv4addr_param *)phdr;
2114			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2115			/* look it up */
2116			stcb = sctp_findassociation_ep_addr(inp_p,
2117			    (struct sockaddr *)&sin4, netp, dst, NULL);
2118			if (stcb != NULL) {
2119				return (stcb);
2120			}
2121		}
2122#endif
2123#ifdef INET6
2124		if (ptype == SCTP_IPV6_ADDRESS &&
2125		    plen == sizeof(struct sctp_ipv6addr_param)) {
2126			/* Get the rest of the address */
2127			struct sctp_ipv6addr_param ip6_parm, *p6;
2128
2129			phdr = sctp_get_next_param(m, offset,
2130			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2131			if (phdr == NULL) {
2132				return (NULL);
2133			}
2134			p6 = (struct sctp_ipv6addr_param *)phdr;
2135			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2136			/* look it up */
2137			stcb = sctp_findassociation_ep_addr(inp_p,
2138			    (struct sockaddr *)&sin6, netp, dst, NULL);
2139			if (stcb != NULL) {
2140				return (stcb);
2141			}
2142		}
2143#endif
2144		offset += SCTP_SIZE32(plen);
2145		phdr = sctp_get_next_param(m, offset, &parm_buf,
2146		    sizeof(parm_buf));
2147	}
2148	return (NULL);
2149}
2150
2151static struct sctp_tcb *
2152sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2153    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2154    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2155{
2156	/*
2157	 * Use my vtag to hash. If we find it we then verify the source addr
2158	 * is in the assoc. If all goes well we save a bit on rec of a
2159	 * packet.
2160	 */
2161	struct sctpasochead *head;
2162	struct sctp_nets *net;
2163	struct sctp_tcb *stcb;
2164
2165	SCTP_INP_INFO_RLOCK();
2166	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2167	    SCTP_BASE_INFO(hashasocmark))];
2168	LIST_FOREACH(stcb, head, sctp_asocs) {
2169		SCTP_INP_RLOCK(stcb->sctp_ep);
2170		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2171			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2172			continue;
2173		}
2174		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2175			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2176			continue;
2177		}
2178		SCTP_TCB_LOCK(stcb);
2179		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2180		if (stcb->asoc.my_vtag == vtag) {
2181			/* candidate */
2182			if (stcb->rport != rport) {
2183				SCTP_TCB_UNLOCK(stcb);
2184				continue;
2185			}
2186			if (stcb->sctp_ep->sctp_lport != lport) {
2187				SCTP_TCB_UNLOCK(stcb);
2188				continue;
2189			}
2190			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2191				SCTP_TCB_UNLOCK(stcb);
2192				continue;
2193			}
2194			/* RRS:Need toaddr check here */
2195			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2196				/* Endpoint does not own this address */
2197				SCTP_TCB_UNLOCK(stcb);
2198				continue;
2199			}
2200			if (remote_tag) {
2201				/*
2202				 * If we have both vtags that's all we match
2203				 * on
2204				 */
2205				if (stcb->asoc.peer_vtag == remote_tag) {
2206					/*
2207					 * If both tags match we consider it
2208					 * conclusive and check NO
2209					 * source/destination addresses
2210					 */
2211					goto conclusive;
2212				}
2213			}
2214			if (skip_src_check) {
2215		conclusive:
2216				if (from) {
2217					*netp = sctp_findnet(stcb, from);
2218				} else {
2219					*netp = NULL;	/* unknown */
2220				}
2221				if (inp_p)
2222					*inp_p = stcb->sctp_ep;
2223				SCTP_INP_INFO_RUNLOCK();
2224				return (stcb);
2225			}
2226			net = sctp_findnet(stcb, from);
2227			if (net) {
2228				/* yep its him. */
2229				*netp = net;
2230				SCTP_STAT_INCR(sctps_vtagexpress);
2231				*inp_p = stcb->sctp_ep;
2232				SCTP_INP_INFO_RUNLOCK();
2233				return (stcb);
2234			} else {
2235				/*
2236				 * not him, this should only happen in rare
2237				 * cases so I peg it.
2238				 */
2239				SCTP_STAT_INCR(sctps_vtagbogus);
2240			}
2241		}
2242		SCTP_TCB_UNLOCK(stcb);
2243	}
2244	SCTP_INP_INFO_RUNLOCK();
2245	return (NULL);
2246}
2247
2248
2249/*
2250 * Find an association with the pointer to the inbound IP packet. This can be
2251 * a IPv4 or IPv6 packet.
2252 */
2253struct sctp_tcb *
2254sctp_findassociation_addr(struct mbuf *m, int offset,
2255    struct sockaddr *src, struct sockaddr *dst,
2256    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2257    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2258{
2259	struct sctp_tcb *stcb;
2260	struct sctp_inpcb *inp;
2261
2262	if (sh->v_tag) {
2263		/* we only go down this path if vtag is non-zero */
2264		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2265		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2266		if (stcb) {
2267			return (stcb);
2268		}
2269	}
2270	if (inp_p) {
2271		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2272		    1, vrf_id);
2273		inp = *inp_p;
2274	} else {
2275		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2276		    1, vrf_id);
2277	}
2278	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2279	if (stcb == NULL && inp) {
2280		/* Found a EP but not this address */
2281		if ((ch->chunk_type == SCTP_INITIATION) ||
2282		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2283			/*-
2284			 * special hook, we do NOT return linp or an
2285			 * association that is linked to an existing
2286			 * association that is under the TCP pool (i.e. no
2287			 * listener exists). The endpoint finding routine
2288			 * will always find a listener before examining the
2289			 * TCP pool.
2290			 */
2291			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2292				if (inp_p) {
2293					*inp_p = NULL;
2294				}
2295				return (NULL);
2296			}
2297			stcb = sctp_findassociation_special_addr(m,
2298			    offset, sh, &inp, netp, dst);
2299			if (inp_p != NULL) {
2300				*inp_p = inp;
2301			}
2302		}
2303	}
2304	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2305	return (stcb);
2306}
2307
2308/*
2309 * lookup an association by an ASCONF lookup address.
2310 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2311 */
2312struct sctp_tcb *
2313sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2314    struct sockaddr *dst, struct sctphdr *sh,
2315    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2316{
2317	struct sctp_tcb *stcb;
2318	union sctp_sockstore remote_store;
2319	struct sctp_paramhdr parm_buf, *phdr;
2320	int ptype;
2321	int zero_address = 0;
2322
2323#ifdef INET
2324	struct sockaddr_in *sin;
2325
2326#endif
2327#ifdef INET6
2328	struct sockaddr_in6 *sin6;
2329
2330#endif
2331
2332	memset(&remote_store, 0, sizeof(remote_store));
2333	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2334	    &parm_buf, sizeof(struct sctp_paramhdr));
2335	if (phdr == NULL) {
2336		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2337		    __func__);
2338		return NULL;
2339	}
2340	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2341	/* get the correlation address */
2342	switch (ptype) {
2343#ifdef INET6
2344	case SCTP_IPV6_ADDRESS:
2345		{
2346			/* ipv6 address param */
2347			struct sctp_ipv6addr_param *p6, p6_buf;
2348
2349			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2350				return NULL;
2351			}
2352			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2353			    offset + sizeof(struct sctp_asconf_chunk),
2354			    &p6_buf.ph, sizeof(*p6));
2355			if (p6 == NULL) {
2356				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2357				    __func__);
2358				return (NULL);
2359			}
2360			sin6 = &remote_store.sin6;
2361			sin6->sin6_family = AF_INET6;
2362			sin6->sin6_len = sizeof(*sin6);
2363			sin6->sin6_port = sh->src_port;
2364			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2365			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2366				zero_address = 1;
2367			break;
2368		}
2369#endif
2370#ifdef INET
2371	case SCTP_IPV4_ADDRESS:
2372		{
2373			/* ipv4 address param */
2374			struct sctp_ipv4addr_param *p4, p4_buf;
2375
2376			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2377				return NULL;
2378			}
2379			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2380			    offset + sizeof(struct sctp_asconf_chunk),
2381			    &p4_buf.ph, sizeof(*p4));
2382			if (p4 == NULL) {
2383				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2384				    __func__);
2385				return (NULL);
2386			}
2387			sin = &remote_store.sin;
2388			sin->sin_family = AF_INET;
2389			sin->sin_len = sizeof(*sin);
2390			sin->sin_port = sh->src_port;
2391			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2392			if (sin->sin_addr.s_addr == INADDR_ANY)
2393				zero_address = 1;
2394			break;
2395		}
2396#endif
2397	default:
2398		/* invalid address param type */
2399		return NULL;
2400	}
2401
2402	if (zero_address) {
2403		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2404		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2405		if (stcb != NULL) {
2406			SCTP_INP_DECR_REF(*inp_p);
2407		}
2408	} else {
2409		stcb = sctp_findassociation_ep_addr(inp_p,
2410		    &remote_store.sa, netp,
2411		    dst, NULL);
2412	}
2413	return (stcb);
2414}
2415
2416
2417/*
2418 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2419 * addresses. This way if we don't get a bind we by default pick a ephemeral
2420 * port with all addresses bound.
2421 */
2422int
2423sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2424{
2425	/*
2426	 * we get called when a new endpoint starts up. We need to allocate
2427	 * the sctp_inpcb structure from the zone and init it. Mark it as
2428	 * unbound and find a port that we can use as an ephemeral with
2429	 * INADDR_ANY. If the user binds later no problem we can then add in
2430	 * the specific addresses. And setup the default parameters for the
2431	 * EP.
2432	 */
2433	int i, error;
2434	struct sctp_inpcb *inp;
2435	struct sctp_pcb *m;
2436	struct timeval time;
2437	sctp_sharedkey_t *null_key;
2438
2439	error = 0;
2440
2441	SCTP_INP_INFO_WLOCK();
2442	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2443	if (inp == NULL) {
2444		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2445		SCTP_INP_INFO_WUNLOCK();
2446		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2447		return (ENOBUFS);
2448	}
2449	/* zap it */
2450	bzero(inp, sizeof(*inp));
2451
2452	/* bump generations */
2453	/* setup socket pointers */
2454	inp->sctp_socket = so;
2455	inp->ip_inp.inp.inp_socket = so;
2456	inp->ip_inp.inp.inp_cred = crhold(so->so_cred);
2457#ifdef INET6
2458	if (INP_SOCKAF(so) == AF_INET6) {
2459		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2460			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2461		}
2462		if (MODULE_GLOBAL(ip6_v6only)) {
2463			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2464		}
2465	}
2466#endif
2467	inp->sctp_associd_counter = 1;
2468	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2469	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2470	inp->max_cwnd = 0;
2471	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2472	inp->ecn_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_ecn_enable);
2473	inp->prsctp_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pr_enable);
2474	if (SCTP_BASE_SYSCTL(sctp_auth_disable)) {
2475		inp->auth_supported = 0;
2476	} else {
2477		inp->auth_supported = 1;
2478	}
2479	inp->asconf_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_asconf_enable);
2480	inp->reconfig_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_reconfig_enable);
2481	inp->nrsack_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_nrsack_enable);
2482	inp->pktdrop_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pktdrop_enable);
2483	inp->fibnum = so->so_fibnum;
2484	/* init the small hash table we use to track asocid <-> tcb */
2485	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2486	if (inp->sctp_asocidhash == NULL) {
2487		crfree(inp->ip_inp.inp.inp_cred);
2488		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2489		SCTP_INP_INFO_WUNLOCK();
2490		return (ENOBUFS);
2491	}
2492#ifdef IPSEC
2493	error = ipsec_init_policy(so, &inp->ip_inp.inp.inp_sp);
2494	if (error != 0) {
2495		crfree(inp->ip_inp.inp.inp_cred);
2496		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2497		SCTP_INP_INFO_WUNLOCK();
2498		return error;
2499	}
2500#endif				/* IPSEC */
2501	SCTP_INCR_EP_COUNT();
2502	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2503	SCTP_INP_INFO_WUNLOCK();
2504
2505	so->so_pcb = (caddr_t)inp;
2506
2507	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2508		/* UDP style socket */
2509		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2510		    SCTP_PCB_FLAGS_UNBOUND);
2511		/* Be sure it is NON-BLOCKING IO for UDP */
2512		/* SCTP_SET_SO_NBIO(so); */
2513	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2514		/* TCP style socket */
2515		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2516		    SCTP_PCB_FLAGS_UNBOUND);
2517		/* Be sure we have blocking IO by default */
2518		SCTP_CLEAR_SO_NBIO(so);
2519	} else {
2520		/*
2521		 * unsupported socket type (RAW, etc)- in case we missed it
2522		 * in protosw
2523		 */
2524		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2525		so->so_pcb = NULL;
2526		crfree(inp->ip_inp.inp.inp_cred);
2527#ifdef IPSEC
2528		ipsec_delete_pcbpolicy(&inp->ip_inp.inp);
2529#endif
2530		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2531		return (EOPNOTSUPP);
2532	}
2533	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2534		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2535		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2536	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2537		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2538		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2539	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2540		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2541		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2542	}
2543	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2544	    &inp->sctp_hashmark);
2545	if (inp->sctp_tcbhash == NULL) {
2546		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2547		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2548		so->so_pcb = NULL;
2549		crfree(inp->ip_inp.inp.inp_cred);
2550#ifdef IPSEC
2551		ipsec_delete_pcbpolicy(&inp->ip_inp.inp);
2552#endif
2553		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2554		return (ENOBUFS);
2555	}
2556	inp->def_vrf_id = vrf_id;
2557
2558	SCTP_INP_INFO_WLOCK();
2559	SCTP_INP_LOCK_INIT(inp);
2560	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2561	SCTP_INP_READ_INIT(inp);
2562	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2563	/* lock the new ep */
2564	SCTP_INP_WLOCK(inp);
2565
2566	/* add it to the info area */
2567	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2568	SCTP_INP_INFO_WUNLOCK();
2569
2570	TAILQ_INIT(&inp->read_queue);
2571	LIST_INIT(&inp->sctp_addr_list);
2572
2573	LIST_INIT(&inp->sctp_asoc_list);
2574
2575#ifdef SCTP_TRACK_FREED_ASOCS
2576	/* TEMP CODE */
2577	LIST_INIT(&inp->sctp_asoc_free_list);
2578#endif
2579	/* Init the timer structure for signature change */
2580	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2581	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2582
2583	/* now init the actual endpoint default data */
2584	m = &inp->sctp_ep;
2585
2586	/* setup the base timeout information */
2587	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2588	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2589	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2590	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2591	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2592	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2593	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2594	/* all max/min max are in ms */
2595	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2596	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2597	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2598	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2599	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2600	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2601	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2602	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2603	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2604	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2605	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2606	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2607	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2608
2609	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2610	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2611	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2612	/* number of streams to pre-open on a association */
2613	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2614
2615	/* Add adaptation cookie */
2616	m->adaptation_layer_indicator = 0;
2617	m->adaptation_layer_indicator_provided = 0;
2618
2619	/* seed random number generator */
2620	m->random_counter = 1;
2621	m->store_at = SCTP_SIGNATURE_SIZE;
2622	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2623	sctp_fill_random_store(m);
2624
2625	/* Minimum cookie size */
2626	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2627	    sizeof(struct sctp_state_cookie);
2628	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2629
2630	/* Setup the initial secret */
2631	(void)SCTP_GETTIME_TIMEVAL(&time);
2632	m->time_of_secret_change = time.tv_sec;
2633
2634	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2635		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2636	}
2637	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2638
2639	/* How long is a cookie good for ? */
2640	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2641	/*
2642	 * Initialize authentication parameters
2643	 */
2644	m->local_hmacs = sctp_default_supported_hmaclist();
2645	m->local_auth_chunks = sctp_alloc_chunklist();
2646	if (inp->asconf_supported) {
2647		sctp_auth_add_chunk(SCTP_ASCONF, m->local_auth_chunks);
2648		sctp_auth_add_chunk(SCTP_ASCONF_ACK, m->local_auth_chunks);
2649	}
2650	m->default_dscp = 0;
2651#ifdef INET6
2652	m->default_flowlabel = 0;
2653#endif
2654	m->port = 0;		/* encapsulation disabled by default */
2655	LIST_INIT(&m->shared_keys);
2656	/* add default NULL key as key id 0 */
2657	null_key = sctp_alloc_sharedkey();
2658	sctp_insert_sharedkey(&m->shared_keys, null_key);
2659	SCTP_INP_WUNLOCK(inp);
2660#ifdef SCTP_LOG_CLOSING
2661	sctp_log_closing(inp, NULL, 12);
2662#endif
2663	return (error);
2664}
2665
2666
2667void
2668sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2669    struct sctp_tcb *stcb)
2670{
2671	struct sctp_nets *net;
2672	uint16_t lport, rport;
2673	struct sctppcbhead *head;
2674	struct sctp_laddr *laddr, *oladdr;
2675
2676	atomic_add_int(&stcb->asoc.refcnt, 1);
2677	SCTP_TCB_UNLOCK(stcb);
2678	SCTP_INP_INFO_WLOCK();
2679	SCTP_INP_WLOCK(old_inp);
2680	SCTP_INP_WLOCK(new_inp);
2681	SCTP_TCB_LOCK(stcb);
2682	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2683
2684	new_inp->sctp_ep.time_of_secret_change =
2685	    old_inp->sctp_ep.time_of_secret_change;
2686	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2687	    sizeof(old_inp->sctp_ep.secret_key));
2688	new_inp->sctp_ep.current_secret_number =
2689	    old_inp->sctp_ep.current_secret_number;
2690	new_inp->sctp_ep.last_secret_number =
2691	    old_inp->sctp_ep.last_secret_number;
2692	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2693
2694	/* make it so new data pours into the new socket */
2695	stcb->sctp_socket = new_inp->sctp_socket;
2696	stcb->sctp_ep = new_inp;
2697
2698	/* Copy the port across */
2699	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2700	rport = stcb->rport;
2701	/* Pull the tcb from the old association */
2702	LIST_REMOVE(stcb, sctp_tcbhash);
2703	LIST_REMOVE(stcb, sctp_tcblist);
2704	if (stcb->asoc.in_asocid_hash) {
2705		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2706	}
2707	/* Now insert the new_inp into the TCP connected hash */
2708	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2709
2710	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2711	/* Its safe to access */
2712	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2713
2714	/* Now move the tcb into the endpoint list */
2715	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2716	/*
2717	 * Question, do we even need to worry about the ep-hash since we
2718	 * only have one connection? Probably not :> so lets get rid of it
2719	 * and not suck up any kernel memory in that.
2720	 */
2721	if (stcb->asoc.in_asocid_hash) {
2722		struct sctpasochead *lhd;
2723
2724		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2725		    new_inp->hashasocidmark)];
2726		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2727	}
2728	/* Ok. Let's restart timer. */
2729	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2730		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2731		    stcb, net);
2732	}
2733
2734	SCTP_INP_INFO_WUNLOCK();
2735	if (new_inp->sctp_tcbhash != NULL) {
2736		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2737		new_inp->sctp_tcbhash = NULL;
2738	}
2739	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2740		/* Subset bound, so copy in the laddr list from the old_inp */
2741		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2742			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2743			if (laddr == NULL) {
2744				/*
2745				 * Gak, what can we do? This assoc is really
2746				 * HOSED. We probably should send an abort
2747				 * here.
2748				 */
2749				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2750				continue;
2751			}
2752			SCTP_INCR_LADDR_COUNT();
2753			bzero(laddr, sizeof(*laddr));
2754			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2755			laddr->ifa = oladdr->ifa;
2756			atomic_add_int(&laddr->ifa->refcount, 1);
2757			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2758			    sctp_nxt_addr);
2759			new_inp->laddr_count++;
2760			if (oladdr == stcb->asoc.last_used_address) {
2761				stcb->asoc.last_used_address = laddr;
2762			}
2763		}
2764	}
2765	/*
2766	 * Now any running timers need to be adjusted since we really don't
2767	 * care if they are running or not just blast in the new_inp into
2768	 * all of them.
2769	 */
2770
2771	stcb->asoc.dack_timer.ep = (void *)new_inp;
2772	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2773	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2774	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2775	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2776	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2777	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2778	/* now what about the nets? */
2779	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2780		net->pmtu_timer.ep = (void *)new_inp;
2781		net->hb_timer.ep = (void *)new_inp;
2782		net->rxt_timer.ep = (void *)new_inp;
2783	}
2784	SCTP_INP_WUNLOCK(new_inp);
2785	SCTP_INP_WUNLOCK(old_inp);
2786}
2787
2788/*
2789 * insert an laddr entry with the given ifa for the desired list
2790 */
2791static int
2792sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
2793{
2794	struct sctp_laddr *laddr;
2795
2796	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2797	if (laddr == NULL) {
2798		/* out of memory? */
2799		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2800		return (EINVAL);
2801	}
2802	SCTP_INCR_LADDR_COUNT();
2803	bzero(laddr, sizeof(*laddr));
2804	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2805	laddr->ifa = ifa;
2806	laddr->action = act;
2807	atomic_add_int(&ifa->refcount, 1);
2808	/* insert it */
2809	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
2810
2811	return (0);
2812}
2813
2814/*
2815 * Remove an laddr entry from the local address list (on an assoc)
2816 */
2817static void
2818sctp_remove_laddr(struct sctp_laddr *laddr)
2819{
2820
2821	/* remove from the list */
2822	LIST_REMOVE(laddr, sctp_nxt_addr);
2823	sctp_free_ifa(laddr->ifa);
2824	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
2825	SCTP_DECR_LADDR_COUNT();
2826}
2827
2828
2829
2830/* sctp_ifap is used to bypass normal local address validation checks */
2831int
2832sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2833    struct sctp_ifa *sctp_ifap, struct thread *p)
2834{
2835	/* bind a ep to a socket address */
2836	struct sctppcbhead *head;
2837	struct sctp_inpcb *inp, *inp_tmp;
2838	struct inpcb *ip_inp;
2839	int port_reuse_active = 0;
2840	int bindall;
2841	uint16_t lport;
2842	int error;
2843	uint32_t vrf_id;
2844
2845	lport = 0;
2846	bindall = 1;
2847	inp = (struct sctp_inpcb *)so->so_pcb;
2848	ip_inp = (struct inpcb *)so->so_pcb;
2849#ifdef SCTP_DEBUG
2850	if (addr) {
2851		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2852		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2853		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2854		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2855	}
2856#endif
2857	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2858		/* already did a bind, subsequent binds NOT allowed ! */
2859		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2860		return (EINVAL);
2861	}
2862#ifdef INVARIANTS
2863	if (p == NULL)
2864		panic("null proc/thread");
2865#endif
2866	if (addr != NULL) {
2867		switch (addr->sa_family) {
2868#ifdef INET
2869		case AF_INET:
2870			{
2871				struct sockaddr_in *sin;
2872
2873				/* IPV6_V6ONLY socket? */
2874				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2875					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2876					return (EINVAL);
2877				}
2878				if (addr->sa_len != sizeof(*sin)) {
2879					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2880					return (EINVAL);
2881				}
2882				sin = (struct sockaddr_in *)addr;
2883				lport = sin->sin_port;
2884				/*
2885				 * For LOOPBACK the prison_local_ip4() call
2886				 * will transmute the ip address to the
2887				 * proper value.
2888				 */
2889				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2890					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2891					return (error);
2892				}
2893				if (sin->sin_addr.s_addr != INADDR_ANY) {
2894					bindall = 0;
2895				}
2896				break;
2897			}
2898#endif
2899#ifdef INET6
2900		case AF_INET6:
2901			{
2902				/*
2903				 * Only for pure IPv6 Address. (No IPv4
2904				 * Mapped!)
2905				 */
2906				struct sockaddr_in6 *sin6;
2907
2908				sin6 = (struct sockaddr_in6 *)addr;
2909
2910				if (addr->sa_len != sizeof(*sin6)) {
2911					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2912					return (EINVAL);
2913				}
2914				lport = sin6->sin6_port;
2915				/*
2916				 * For LOOPBACK the prison_local_ip6() call
2917				 * will transmute the ipv6 address to the
2918				 * proper value.
2919				 */
2920				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2921				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2922					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2923					return (error);
2924				}
2925				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2926					bindall = 0;
2927					/* KAME hack: embed scopeid */
2928					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2929						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2930						return (EINVAL);
2931					}
2932				}
2933				/* this must be cleared for ifa_ifwithaddr() */
2934				sin6->sin6_scope_id = 0;
2935				break;
2936			}
2937#endif
2938		default:
2939			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2940			return (EAFNOSUPPORT);
2941		}
2942	}
2943	SCTP_INP_INFO_WLOCK();
2944	SCTP_INP_WLOCK(inp);
2945	/* Setup a vrf_id to be the default for the non-bind-all case. */
2946	vrf_id = inp->def_vrf_id;
2947
2948	/* increase our count due to the unlock we do */
2949	SCTP_INP_INCR_REF(inp);
2950	if (lport) {
2951		/*
2952		 * Did the caller specify a port? if so we must see if an ep
2953		 * already has this one bound.
2954		 */
2955		/* got to be root to get at low ports */
2956		if (ntohs(lport) < IPPORT_RESERVED) {
2957			if (p && (error =
2958			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2959			    )) {
2960				SCTP_INP_DECR_REF(inp);
2961				SCTP_INP_WUNLOCK(inp);
2962				SCTP_INP_INFO_WUNLOCK();
2963				return (error);
2964			}
2965		}
2966		SCTP_INP_WUNLOCK(inp);
2967		if (bindall) {
2968			vrf_id = inp->def_vrf_id;
2969			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2970			if (inp_tmp != NULL) {
2971				/*
2972				 * lock guy returned and lower count note
2973				 * that we are not bound so inp_tmp should
2974				 * NEVER be inp. And it is this inp
2975				 * (inp_tmp) that gets the reference bump,
2976				 * so we must lower it.
2977				 */
2978				SCTP_INP_DECR_REF(inp_tmp);
2979				/* unlock info */
2980				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2981				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2982					/*
2983					 * Ok, must be one-2-one and
2984					 * allowing port re-use
2985					 */
2986					port_reuse_active = 1;
2987					goto continue_anyway;
2988				}
2989				SCTP_INP_DECR_REF(inp);
2990				SCTP_INP_INFO_WUNLOCK();
2991				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2992				return (EADDRINUSE);
2993			}
2994		} else {
2995			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2996			if (inp_tmp != NULL) {
2997				/*
2998				 * lock guy returned and lower count note
2999				 * that we are not bound so inp_tmp should
3000				 * NEVER be inp. And it is this inp
3001				 * (inp_tmp) that gets the reference bump,
3002				 * so we must lower it.
3003				 */
3004				SCTP_INP_DECR_REF(inp_tmp);
3005				/* unlock info */
3006				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
3007				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
3008					/*
3009					 * Ok, must be one-2-one and
3010					 * allowing port re-use
3011					 */
3012					port_reuse_active = 1;
3013					goto continue_anyway;
3014				}
3015				SCTP_INP_DECR_REF(inp);
3016				SCTP_INP_INFO_WUNLOCK();
3017				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3018				return (EADDRINUSE);
3019			}
3020		}
3021continue_anyway:
3022		SCTP_INP_WLOCK(inp);
3023		if (bindall) {
3024			/* verify that no lport is not used by a singleton */
3025			if ((port_reuse_active == 0) &&
3026			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
3027				/* Sorry someone already has this one bound */
3028				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
3029				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
3030					port_reuse_active = 1;
3031				} else {
3032					SCTP_INP_DECR_REF(inp);
3033					SCTP_INP_WUNLOCK(inp);
3034					SCTP_INP_INFO_WUNLOCK();
3035					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3036					return (EADDRINUSE);
3037				}
3038			}
3039		}
3040	} else {
3041		uint16_t first, last, candidate;
3042		uint16_t count;
3043		int done;
3044
3045		if (ip_inp->inp_flags & INP_HIGHPORT) {
3046			first = MODULE_GLOBAL(ipport_hifirstauto);
3047			last = MODULE_GLOBAL(ipport_hilastauto);
3048		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3049			if (p && (error =
3050			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3051			    )) {
3052				SCTP_INP_DECR_REF(inp);
3053				SCTP_INP_WUNLOCK(inp);
3054				SCTP_INP_INFO_WUNLOCK();
3055				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3056				return (error);
3057			}
3058			first = MODULE_GLOBAL(ipport_lowfirstauto);
3059			last = MODULE_GLOBAL(ipport_lowlastauto);
3060		} else {
3061			first = MODULE_GLOBAL(ipport_firstauto);
3062			last = MODULE_GLOBAL(ipport_lastauto);
3063		}
3064		if (first > last) {
3065			uint16_t temp;
3066
3067			temp = first;
3068			first = last;
3069			last = temp;
3070		}
3071		count = last - first + 1;	/* number of candidates */
3072		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3073
3074		done = 0;
3075		while (!done) {
3076			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3077				done = 1;
3078			}
3079			if (!done) {
3080				if (--count == 0) {
3081					SCTP_INP_DECR_REF(inp);
3082					SCTP_INP_WUNLOCK(inp);
3083					SCTP_INP_INFO_WUNLOCK();
3084					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3085					return (EADDRINUSE);
3086				}
3087				if (candidate == last)
3088					candidate = first;
3089				else
3090					candidate = candidate + 1;
3091			}
3092		}
3093		lport = htons(candidate);
3094	}
3095	SCTP_INP_DECR_REF(inp);
3096	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3097	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3098		/*
3099		 * this really should not happen. The guy did a non-blocking
3100		 * bind and then did a close at the same time.
3101		 */
3102		SCTP_INP_WUNLOCK(inp);
3103		SCTP_INP_INFO_WUNLOCK();
3104		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3105		return (EINVAL);
3106	}
3107	/* ok we look clear to give out this port, so lets setup the binding */
3108	if (bindall) {
3109		/* binding to all addresses, so just set in the proper flags */
3110		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3111		/* set the automatic addr changes from kernel flag */
3112		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3113			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3114			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3115		} else {
3116			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3117			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3118		}
3119		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3120			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3121		} else {
3122			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3123		}
3124		/*
3125		 * set the automatic mobility_base from kernel flag (by
3126		 * micchie)
3127		 */
3128		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3129			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3130			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3131		} else {
3132			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3133			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3134		}
3135		/*
3136		 * set the automatic mobility_fasthandoff from kernel flag
3137		 * (by micchie)
3138		 */
3139		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3140			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3141			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3142		} else {
3143			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3144			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3145		}
3146	} else {
3147		/*
3148		 * bind specific, make sure flags is off and add a new
3149		 * address structure to the sctp_addr_list inside the ep
3150		 * structure.
3151		 *
3152		 * We will need to allocate one and insert it at the head. The
3153		 * socketopt call can just insert new addresses in there as
3154		 * well. It will also have to do the embed scope kame hack
3155		 * too (before adding).
3156		 */
3157		struct sctp_ifa *ifa;
3158		union sctp_sockstore store;
3159
3160		memset(&store, 0, sizeof(store));
3161		switch (addr->sa_family) {
3162#ifdef INET
3163		case AF_INET:
3164			memcpy(&store.sin, addr, sizeof(struct sockaddr_in));
3165			store.sin.sin_port = 0;
3166			break;
3167#endif
3168#ifdef INET6
3169		case AF_INET6:
3170			memcpy(&store.sin6, addr, sizeof(struct sockaddr_in6));
3171			store.sin6.sin6_port = 0;
3172			break;
3173#endif
3174		default:
3175			break;
3176		}
3177		/*
3178		 * first find the interface with the bound address need to
3179		 * zero out the port to find the address! yuck! can't do
3180		 * this earlier since need port for sctp_pcb_findep()
3181		 */
3182		if (sctp_ifap != NULL) {
3183			ifa = sctp_ifap;
3184		} else {
3185			/*
3186			 * Note for BSD we hit here always other O/S's will
3187			 * pass things in via the sctp_ifap argument
3188			 * (Panda).
3189			 */
3190			ifa = sctp_find_ifa_by_addr(&store.sa,
3191			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3192		}
3193		if (ifa == NULL) {
3194			/* Can't find an interface with that address */
3195			SCTP_INP_WUNLOCK(inp);
3196			SCTP_INP_INFO_WUNLOCK();
3197			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3198			return (EADDRNOTAVAIL);
3199		}
3200#ifdef INET6
3201		if (addr->sa_family == AF_INET6) {
3202			/* GAK, more FIXME IFA lock? */
3203			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3204				/* Can't bind a non-existent addr. */
3205				SCTP_INP_WUNLOCK(inp);
3206				SCTP_INP_INFO_WUNLOCK();
3207				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3208				return (EINVAL);
3209			}
3210		}
3211#endif
3212		/* we're not bound all */
3213		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3214		/* allow bindx() to send ASCONF's for binding changes */
3215		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3216		/* clear automatic addr changes from kernel flag */
3217		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3218
3219		/* add this address to the endpoint list */
3220		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3221		if (error != 0) {
3222			SCTP_INP_WUNLOCK(inp);
3223			SCTP_INP_INFO_WUNLOCK();
3224			return (error);
3225		}
3226		inp->laddr_count++;
3227	}
3228	/* find the bucket */
3229	if (port_reuse_active) {
3230		/* Put it into tcp 1-2-1 hash */
3231		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3232		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3233	} else {
3234		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3235	}
3236	/* put it in the bucket */
3237	LIST_INSERT_HEAD(head, inp, sctp_hash);
3238	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3239	    (void *)head, ntohs(lport), port_reuse_active);
3240	/* set in the port */
3241	inp->sctp_lport = lport;
3242
3243	/* turn off just the unbound flag */
3244	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3245	SCTP_INP_WUNLOCK(inp);
3246	SCTP_INP_INFO_WUNLOCK();
3247	return (0);
3248}
3249
3250
3251static void
3252sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3253{
3254	struct sctp_iterator *it, *nit;
3255
3256	/*
3257	 * We enter with the only the ITERATOR_LOCK in place and a write
3258	 * lock on the inp_info stuff.
3259	 */
3260	it = sctp_it_ctl.cur_it;
3261	if (it && (it->vn != curvnet)) {
3262		/* Its not looking at our VNET */
3263		return;
3264	}
3265	if (it && (it->inp == inp)) {
3266		/*
3267		 * This is tricky and we hold the iterator lock, but when it
3268		 * returns and gets the lock (when we release it) the
3269		 * iterator will try to operate on inp. We need to stop that
3270		 * from happening. But of course the iterator has a
3271		 * reference on the stcb and inp. We can mark it and it will
3272		 * stop.
3273		 *
3274		 * If its a single iterator situation, we set the end iterator
3275		 * flag. Otherwise we set the iterator to go to the next
3276		 * inp.
3277		 *
3278		 */
3279		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3280			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3281		} else {
3282			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3283		}
3284	}
3285	/*
3286	 * Now go through and remove any single reference to our inp that
3287	 * may be still pending on the list
3288	 */
3289	SCTP_IPI_ITERATOR_WQ_LOCK();
3290	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3291		if (it->vn != curvnet) {
3292			continue;
3293		}
3294		if (it->inp == inp) {
3295			/* This one points to me is it inp specific? */
3296			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3297				/* Remove and free this one */
3298				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3299				    it, sctp_nxt_itr);
3300				if (it->function_atend != NULL) {
3301					(*it->function_atend) (it->pointer, it->val);
3302				}
3303				SCTP_FREE(it, SCTP_M_ITER);
3304			} else {
3305				it->inp = LIST_NEXT(it->inp, sctp_list);
3306				if (it->inp) {
3307					SCTP_INP_INCR_REF(it->inp);
3308				}
3309			}
3310			/*
3311			 * When its put in the refcnt is incremented so decr
3312			 * it
3313			 */
3314			SCTP_INP_DECR_REF(inp);
3315		}
3316	}
3317	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3318}
3319
3320/* release sctp_inpcb unbind the port */
3321void
3322sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3323{
3324	/*
3325	 * Here we free a endpoint. We must find it (if it is in the Hash
3326	 * table) and remove it from there. Then we must also find it in the
3327	 * overall list and remove it from there. After all removals are
3328	 * complete then any timer has to be stopped. Then start the actual
3329	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3330	 * all associations. d) finally the ep itself.
3331	 */
3332	struct sctp_tcb *asoc, *nasoc;
3333	struct sctp_laddr *laddr, *nladdr;
3334	struct inpcb *ip_pcb;
3335	struct socket *so;
3336	int being_refed = 0;
3337	struct sctp_queued_to_read *sq, *nsq;
3338	int cnt;
3339	sctp_sharedkey_t *shared_key, *nshared_key;
3340
3341
3342#ifdef SCTP_LOG_CLOSING
3343	sctp_log_closing(inp, NULL, 0);
3344#endif
3345	SCTP_ITERATOR_LOCK();
3346	/* mark any iterators on the list or being processed */
3347	sctp_iterator_inp_being_freed(inp);
3348	SCTP_ITERATOR_UNLOCK();
3349	so = inp->sctp_socket;
3350	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3351		/* been here before.. eeks.. get out of here */
3352		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3353#ifdef SCTP_LOG_CLOSING
3354		sctp_log_closing(inp, NULL, 1);
3355#endif
3356		return;
3357	}
3358	SCTP_ASOC_CREATE_LOCK(inp);
3359	SCTP_INP_INFO_WLOCK();
3360
3361	SCTP_INP_WLOCK(inp);
3362	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3363		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3364		/* socket is gone, so no more wakeups allowed */
3365		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3366		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3367		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3368
3369	}
3370	/* First time through we have the socket lock, after that no more. */
3371	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3372	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3373
3374	if (inp->control) {
3375		sctp_m_freem(inp->control);
3376		inp->control = NULL;
3377	}
3378	if (inp->pkt) {
3379		sctp_m_freem(inp->pkt);
3380		inp->pkt = NULL;
3381	}
3382	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3383					 * here but I will be nice :> (i.e.
3384					 * ip_pcb = ep;) */
3385	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3386		int cnt_in_sd;
3387
3388		cnt_in_sd = 0;
3389		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3390			SCTP_TCB_LOCK(asoc);
3391			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3392				/* Skip guys being freed */
3393				cnt_in_sd++;
3394				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3395					/*
3396					 * Special case - we did not start a
3397					 * kill timer on the asoc due to it
3398					 * was not closed. So go ahead and
3399					 * start it now.
3400					 */
3401					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3402					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3403				}
3404				SCTP_TCB_UNLOCK(asoc);
3405				continue;
3406			}
3407			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3408			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3409			    (asoc->asoc.total_output_queue_size == 0)) {
3410				/*
3411				 * If we have data in queue, we don't want
3412				 * to just free since the app may have done,
3413				 * send()/close or connect/send/close. And
3414				 * it wants the data to get across first.
3415				 */
3416				/* Just abandon things in the front states */
3417				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3418				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3419					cnt_in_sd++;
3420				}
3421				continue;
3422			}
3423			/* Disconnect the socket please */
3424			asoc->sctp_socket = NULL;
3425			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3426			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3427			    (asoc->asoc.control_pdapi) ||
3428			    (asoc->asoc.size_on_all_streams > 0) ||
3429			    (so && (so->so_rcv.sb_cc > 0))) {
3430				/* Left with Data unread */
3431				struct mbuf *op_err;
3432
3433				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3434				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3435				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3436				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3437				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3438				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3439					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3440				}
3441				if (sctp_free_assoc(inp, asoc,
3442				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3443					cnt_in_sd++;
3444				}
3445				continue;
3446			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3447				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3448			    (asoc->asoc.stream_queue_cnt == 0)) {
3449				if (asoc->asoc.locked_on_sending) {
3450					goto abort_anyway;
3451				}
3452				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3453				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3454					struct sctp_nets *netp;
3455
3456					/*
3457					 * there is nothing queued to send,
3458					 * so I send shutdown
3459					 */
3460					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3461					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3462						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3463					}
3464					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3465					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3466					sctp_stop_timers_for_shutdown(asoc);
3467					if (asoc->asoc.alternate) {
3468						netp = asoc->asoc.alternate;
3469					} else {
3470						netp = asoc->asoc.primary_destination;
3471					}
3472					sctp_send_shutdown(asoc, netp);
3473					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3474					    netp);
3475					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3476					    asoc->asoc.primary_destination);
3477					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3478				}
3479			} else {
3480				/* mark into shutdown pending */
3481				struct sctp_stream_queue_pending *sp;
3482
3483				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3484				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3485				    asoc->asoc.primary_destination);
3486				if (asoc->asoc.locked_on_sending) {
3487					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3488					    sctp_streamhead);
3489					if (sp == NULL) {
3490						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3491						    (void *)asoc->asoc.locked_on_sending,
3492						    asoc->asoc.locked_on_sending->stream_no);
3493					} else {
3494						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3495							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3496					}
3497				}
3498				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3499				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3500				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3501					struct mbuf *op_err;
3502
3503			abort_anyway:
3504					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3505					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3506					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3507					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3508					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3509					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3510						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3511					}
3512					if (sctp_free_assoc(inp, asoc,
3513					    SCTP_PCBFREE_NOFORCE,
3514					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3515						cnt_in_sd++;
3516					}
3517					continue;
3518				} else {
3519					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3520				}
3521			}
3522			cnt_in_sd++;
3523			SCTP_TCB_UNLOCK(asoc);
3524		}
3525		/* now is there some left in our SHUTDOWN state? */
3526		if (cnt_in_sd) {
3527#ifdef SCTP_LOG_CLOSING
3528			sctp_log_closing(inp, NULL, 2);
3529#endif
3530			inp->sctp_socket = NULL;
3531			SCTP_INP_WUNLOCK(inp);
3532			SCTP_ASOC_CREATE_UNLOCK(inp);
3533			SCTP_INP_INFO_WUNLOCK();
3534			return;
3535		}
3536	}
3537	inp->sctp_socket = NULL;
3538	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3539	    SCTP_PCB_FLAGS_UNBOUND) {
3540		/*
3541		 * ok, this guy has been bound. It's port is somewhere in
3542		 * the SCTP_BASE_INFO(hash table). Remove it!
3543		 */
3544		LIST_REMOVE(inp, sctp_hash);
3545		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3546	}
3547	/*
3548	 * If there is a timer running to kill us, forget it, since it may
3549	 * have a contest on the INP lock.. which would cause us to die ...
3550	 */
3551	cnt = 0;
3552	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3553		SCTP_TCB_LOCK(asoc);
3554		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3555			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3556				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3557				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3558			}
3559			cnt++;
3560			SCTP_TCB_UNLOCK(asoc);
3561			continue;
3562		}
3563		/* Free associations that are NOT killing us */
3564		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3565		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3566			struct mbuf *op_err;
3567
3568			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3569			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3570			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3571			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3572		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3573			cnt++;
3574			SCTP_TCB_UNLOCK(asoc);
3575			continue;
3576		}
3577		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3578		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3579			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3580		}
3581		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE,
3582		    SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3583			cnt++;
3584		}
3585	}
3586	if (cnt) {
3587		/* Ok we have someone out there that will kill us */
3588		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3589#ifdef SCTP_LOG_CLOSING
3590		sctp_log_closing(inp, NULL, 3);
3591#endif
3592		SCTP_INP_WUNLOCK(inp);
3593		SCTP_ASOC_CREATE_UNLOCK(inp);
3594		SCTP_INP_INFO_WUNLOCK();
3595		return;
3596	}
3597	if (SCTP_INP_LOCK_CONTENDED(inp))
3598		being_refed++;
3599	if (SCTP_INP_READ_CONTENDED(inp))
3600		being_refed++;
3601	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3602		being_refed++;
3603
3604	if ((inp->refcount) ||
3605	    (being_refed) ||
3606	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3607		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3608#ifdef SCTP_LOG_CLOSING
3609		sctp_log_closing(inp, NULL, 4);
3610#endif
3611		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3612		SCTP_INP_WUNLOCK(inp);
3613		SCTP_ASOC_CREATE_UNLOCK(inp);
3614		SCTP_INP_INFO_WUNLOCK();
3615		return;
3616	}
3617	inp->sctp_ep.signature_change.type = 0;
3618	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3619	/*
3620	 * Remove it from the list .. last thing we need a lock for.
3621	 */
3622	LIST_REMOVE(inp, sctp_list);
3623	SCTP_INP_WUNLOCK(inp);
3624	SCTP_ASOC_CREATE_UNLOCK(inp);
3625	SCTP_INP_INFO_WUNLOCK();
3626	/*
3627	 * Now we release all locks. Since this INP cannot be found anymore
3628	 * except possibly by the kill timer that might be running. We call
3629	 * the drain function here. It should hit the case were it sees the
3630	 * ACTIVE flag cleared and exit out freeing us to proceed and
3631	 * destroy everything.
3632	 */
3633	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3634		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3635	} else {
3636		/* Probably un-needed */
3637		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3638	}
3639
3640#ifdef SCTP_LOG_CLOSING
3641	sctp_log_closing(inp, NULL, 5);
3642#endif
3643
3644
3645	if ((inp->sctp_asocidhash) != NULL) {
3646		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3647		inp->sctp_asocidhash = NULL;
3648	}
3649	/* sa_ignore FREED_MEMORY */
3650	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3651		/* Its only abandoned if it had data left */
3652		if (sq->length)
3653			SCTP_STAT_INCR(sctps_left_abandon);
3654
3655		TAILQ_REMOVE(&inp->read_queue, sq, next);
3656		sctp_free_remote_addr(sq->whoFrom);
3657		if (so)
3658			so->so_rcv.sb_cc -= sq->length;
3659		if (sq->data) {
3660			sctp_m_freem(sq->data);
3661			sq->data = NULL;
3662		}
3663		/*
3664		 * no need to free the net count, since at this point all
3665		 * assoc's are gone.
3666		 */
3667		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3668		SCTP_DECR_READQ_COUNT();
3669	}
3670	/* Now the sctp_pcb things */
3671	/*
3672	 * free each asoc if it is not already closed/free. we can't use the
3673	 * macro here since le_next will get freed as part of the
3674	 * sctp_free_assoc() call.
3675	 */
3676#ifdef IPSEC
3677	ipsec_delete_pcbpolicy(ip_pcb);
3678#endif
3679	if (ip_pcb->inp_options) {
3680		(void)sctp_m_free(ip_pcb->inp_options);
3681		ip_pcb->inp_options = 0;
3682	}
3683#ifdef INET6
3684	if (ip_pcb->inp_vflag & INP_IPV6) {
3685		struct in6pcb *in6p;
3686
3687		in6p = (struct in6pcb *)inp;
3688		ip6_freepcbopts(in6p->in6p_outputopts);
3689	}
3690#endif				/* INET6 */
3691	ip_pcb->inp_vflag = 0;
3692	/* free up authentication fields */
3693	if (inp->sctp_ep.local_auth_chunks != NULL)
3694		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3695	if (inp->sctp_ep.local_hmacs != NULL)
3696		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3697
3698	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3699		LIST_REMOVE(shared_key, next);
3700		sctp_free_sharedkey(shared_key);
3701		/* sa_ignore FREED_MEMORY */
3702	}
3703
3704	/*
3705	 * if we have an address list the following will free the list of
3706	 * ifaddr's that are set into this ep. Again macro limitations here,
3707	 * since the LIST_FOREACH could be a bad idea.
3708	 */
3709	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3710		sctp_remove_laddr(laddr);
3711	}
3712
3713#ifdef SCTP_TRACK_FREED_ASOCS
3714	/* TEMP CODE */
3715	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3716		LIST_REMOVE(asoc, sctp_tcblist);
3717		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3718		SCTP_DECR_ASOC_COUNT();
3719	}
3720	/* *** END TEMP CODE *** */
3721#endif
3722	/* Now lets see about freeing the EP hash table. */
3723	if (inp->sctp_tcbhash != NULL) {
3724		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3725		inp->sctp_tcbhash = NULL;
3726	}
3727	/* Now we must put the ep memory back into the zone pool */
3728	crfree(inp->ip_inp.inp.inp_cred);
3729	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3730	SCTP_INP_LOCK_DESTROY(inp);
3731	SCTP_INP_READ_DESTROY(inp);
3732	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3733	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3734	SCTP_DECR_EP_COUNT();
3735}
3736
3737
3738struct sctp_nets *
3739sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3740{
3741	struct sctp_nets *net;
3742
3743	/* locate the address */
3744	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3745		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3746			return (net);
3747	}
3748	return (NULL);
3749}
3750
3751
3752int
3753sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3754{
3755	struct sctp_ifa *sctp_ifa;
3756
3757	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3758	if (sctp_ifa) {
3759		return (1);
3760	} else {
3761		return (0);
3762	}
3763}
3764
3765/*
3766 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3767 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3768 * stats of stuff.
3769 */
3770int
3771sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3772    struct sctp_nets **netp, int set_scope, int from)
3773{
3774	/*
3775	 * The following is redundant to the same lines in the
3776	 * sctp_aloc_assoc() but is needed since others call the add address
3777	 * function
3778	 */
3779	struct sctp_nets *net, *netfirst;
3780	int addr_inscope;
3781
3782	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3783	    from);
3784	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3785
3786	netfirst = sctp_findnet(stcb, newaddr);
3787	if (netfirst) {
3788		/*
3789		 * Lie and return ok, we don't want to make the association
3790		 * go away for this behavior. It will happen in the TCP
3791		 * model in a connected socket. It does not reach the hash
3792		 * table until after the association is built so it can't be
3793		 * found. Mark as reachable, since the initial creation will
3794		 * have been cleared and the NOT_IN_ASSOC flag will have
3795		 * been added... and we don't want to end up removing it
3796		 * back out.
3797		 */
3798		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3799			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3800			    SCTP_ADDR_UNCONFIRMED);
3801		} else {
3802			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3803		}
3804
3805		return (0);
3806	}
3807	addr_inscope = 1;
3808	switch (newaddr->sa_family) {
3809#ifdef INET
3810	case AF_INET:
3811		{
3812			struct sockaddr_in *sin;
3813
3814			sin = (struct sockaddr_in *)newaddr;
3815			if (sin->sin_addr.s_addr == 0) {
3816				/* Invalid address */
3817				return (-1);
3818			}
3819			/* zero out the bzero area */
3820			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3821
3822			/* assure len is set */
3823			sin->sin_len = sizeof(struct sockaddr_in);
3824			if (set_scope) {
3825				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3826					stcb->asoc.scope.ipv4_local_scope = 1;
3827				}
3828			} else {
3829				/* Validate the address is in scope */
3830				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3831				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3832					addr_inscope = 0;
3833				}
3834			}
3835			break;
3836		}
3837#endif
3838#ifdef INET6
3839	case AF_INET6:
3840		{
3841			struct sockaddr_in6 *sin6;
3842
3843			sin6 = (struct sockaddr_in6 *)newaddr;
3844			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3845				/* Invalid address */
3846				return (-1);
3847			}
3848			/* assure len is set */
3849			sin6->sin6_len = sizeof(struct sockaddr_in6);
3850			if (set_scope) {
3851				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3852					stcb->asoc.scope.loopback_scope = 1;
3853					stcb->asoc.scope.local_scope = 0;
3854					stcb->asoc.scope.ipv4_local_scope = 1;
3855					stcb->asoc.scope.site_scope = 1;
3856				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3857					/*
3858					 * If the new destination is a
3859					 * LINK_LOCAL we must have common
3860					 * site scope. Don't set the local
3861					 * scope since we may not share all
3862					 * links, only loopback can do this.
3863					 * Links on the local network would
3864					 * also be on our private network
3865					 * for v4 too.
3866					 */
3867					stcb->asoc.scope.ipv4_local_scope = 1;
3868					stcb->asoc.scope.site_scope = 1;
3869				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3870					/*
3871					 * If the new destination is
3872					 * SITE_LOCAL then we must have site
3873					 * scope in common.
3874					 */
3875					stcb->asoc.scope.site_scope = 1;
3876				}
3877			} else {
3878				/* Validate the address is in scope */
3879				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3880				    (stcb->asoc.scope.loopback_scope == 0)) {
3881					addr_inscope = 0;
3882				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3883				    (stcb->asoc.scope.local_scope == 0)) {
3884					addr_inscope = 0;
3885				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3886				    (stcb->asoc.scope.site_scope == 0)) {
3887					addr_inscope = 0;
3888				}
3889			}
3890			break;
3891		}
3892#endif
3893	default:
3894		/* not supported family type */
3895		return (-1);
3896	}
3897	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3898	if (net == NULL) {
3899		return (-1);
3900	}
3901	SCTP_INCR_RADDR_COUNT();
3902	bzero(net, sizeof(struct sctp_nets));
3903	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3904	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3905	switch (newaddr->sa_family) {
3906#ifdef INET
3907	case AF_INET:
3908		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3909		break;
3910#endif
3911#ifdef INET6
3912	case AF_INET6:
3913		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3914		break;
3915#endif
3916	default:
3917		break;
3918	}
3919	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3920	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3921		stcb->asoc.scope.loopback_scope = 1;
3922		stcb->asoc.scope.ipv4_local_scope = 1;
3923		stcb->asoc.scope.local_scope = 0;
3924		stcb->asoc.scope.site_scope = 1;
3925		addr_inscope = 1;
3926	}
3927	net->failure_threshold = stcb->asoc.def_net_failure;
3928	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3929	if (addr_inscope == 0) {
3930		net->dest_state = (SCTP_ADDR_REACHABLE |
3931		    SCTP_ADDR_OUT_OF_SCOPE);
3932	} else {
3933		if (from == SCTP_ADDR_IS_CONFIRMED)
3934			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3935			net->dest_state = SCTP_ADDR_REACHABLE;
3936		else
3937			net->dest_state = SCTP_ADDR_REACHABLE |
3938			    SCTP_ADDR_UNCONFIRMED;
3939	}
3940	/*
3941	 * We set this to 0, the timer code knows that this means its an
3942	 * initial value
3943	 */
3944	net->rto_needed = 1;
3945	net->RTO = 0;
3946	net->RTO_measured = 0;
3947	stcb->asoc.numnets++;
3948	net->ref_count = 1;
3949	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3950	net->port = stcb->asoc.port;
3951	net->dscp = stcb->asoc.default_dscp;
3952#ifdef INET6
3953	net->flowlabel = stcb->asoc.default_flowlabel;
3954#endif
3955	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3956		net->dest_state |= SCTP_ADDR_NOHB;
3957	} else {
3958		net->dest_state &= ~SCTP_ADDR_NOHB;
3959	}
3960	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3961		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3962	} else {
3963		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3964	}
3965	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3966	/* Init the timer structure */
3967	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3968	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3969	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3970
3971	/* Now generate a route for this guy */
3972#ifdef INET6
3973	/* KAME hack: embed scopeid */
3974	if (newaddr->sa_family == AF_INET6) {
3975		struct sockaddr_in6 *sin6;
3976
3977		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3978		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3979		sin6->sin6_scope_id = 0;
3980	}
3981#endif
3982	SCTP_RTALLOC((sctp_route_t *) & net->ro,
3983	    stcb->asoc.vrf_id,
3984	    stcb->sctp_ep->fibnum);
3985
3986	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3987		/* Get source address */
3988		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3989		    stcb,
3990		    (sctp_route_t *) & net->ro,
3991		    net,
3992		    0,
3993		    stcb->asoc.vrf_id);
3994		if (net->ro._s_addr != NULL) {
3995			net->src_addr_selected = 1;
3996			/* Now get the interface MTU */
3997			if (net->ro._s_addr->ifn_p != NULL) {
3998				net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3999			}
4000		} else {
4001			net->src_addr_selected = 0;
4002		}
4003		if (net->mtu > 0) {
4004			uint32_t rmtu;
4005
4006			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
4007			if (rmtu == 0) {
4008				/*
4009				 * Start things off to match mtu of
4010				 * interface please.
4011				 */
4012				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
4013				    net->ro.ro_rt, net->mtu);
4014			} else {
4015				/*
4016				 * we take the route mtu over the interface,
4017				 * since the route may be leading out the
4018				 * loopback, or a different interface.
4019				 */
4020				net->mtu = rmtu;
4021			}
4022		}
4023	} else {
4024		net->src_addr_selected = 0;
4025	}
4026	if (net->mtu == 0) {
4027		switch (newaddr->sa_family) {
4028#ifdef INET
4029		case AF_INET:
4030			net->mtu = SCTP_DEFAULT_MTU;
4031			break;
4032#endif
4033#ifdef INET6
4034		case AF_INET6:
4035			net->mtu = 1280;
4036			break;
4037#endif
4038		default:
4039			break;
4040		}
4041	}
4042#if defined(INET) || defined(INET6)
4043	if (net->port) {
4044		net->mtu -= (uint32_t) sizeof(struct udphdr);
4045	}
4046#endif
4047	if (from == SCTP_ALLOC_ASOC) {
4048		stcb->asoc.smallest_mtu = net->mtu;
4049	}
4050	if (stcb->asoc.smallest_mtu > net->mtu) {
4051		stcb->asoc.smallest_mtu = net->mtu;
4052	}
4053#ifdef INET6
4054	if (newaddr->sa_family == AF_INET6) {
4055		struct sockaddr_in6 *sin6;
4056
4057		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4058		(void)sa6_recoverscope(sin6);
4059	}
4060#endif
4061
4062	/* JRS - Use the congestion control given in the CC module */
4063	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4064		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4065
4066	/*
4067	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4068	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4069	 */
4070	net->find_pseudo_cumack = 1;
4071	net->find_rtx_pseudo_cumack = 1;
4072	/* Choose an initial flowid. */
4073	net->flowid = stcb->asoc.my_vtag ^
4074	    ntohs(stcb->rport) ^
4075	    ntohs(stcb->sctp_ep->sctp_lport);
4076	net->flowtype = M_HASHTYPE_OPAQUE;
4077	if (netp) {
4078		*netp = net;
4079	}
4080	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4081	if (net->ro.ro_rt == NULL) {
4082		/* Since we have no route put it at the back */
4083		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4084	} else if (netfirst == NULL) {
4085		/* We are the first one in the pool. */
4086		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4087	} else if (netfirst->ro.ro_rt == NULL) {
4088		/*
4089		 * First one has NO route. Place this one ahead of the first
4090		 * one.
4091		 */
4092		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4093	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4094		/*
4095		 * This one has a different interface than the one at the
4096		 * top of the list. Place it ahead.
4097		 */
4098		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4099	} else {
4100		/*
4101		 * Ok we have the same interface as the first one. Move
4102		 * forward until we find either a) one with a NULL route...
4103		 * insert ahead of that b) one with a different ifp.. insert
4104		 * after that. c) end of the list.. insert at the tail.
4105		 */
4106		struct sctp_nets *netlook;
4107
4108		do {
4109			netlook = TAILQ_NEXT(netfirst, sctp_next);
4110			if (netlook == NULL) {
4111				/* End of the list */
4112				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4113				break;
4114			} else if (netlook->ro.ro_rt == NULL) {
4115				/* next one has NO route */
4116				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4117				break;
4118			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4119				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4120				    net, sctp_next);
4121				break;
4122			}
4123			/* Shift forward */
4124			netfirst = netlook;
4125		} while (netlook != NULL);
4126	}
4127
4128	/* got to have a primary set */
4129	if (stcb->asoc.primary_destination == 0) {
4130		stcb->asoc.primary_destination = net;
4131	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4132		    (net->ro.ro_rt) &&
4133	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4134		/* No route to current primary adopt new primary */
4135		stcb->asoc.primary_destination = net;
4136	}
4137	/* Validate primary is first */
4138	net = TAILQ_FIRST(&stcb->asoc.nets);
4139	if ((net != stcb->asoc.primary_destination) &&
4140	    (stcb->asoc.primary_destination)) {
4141		/*
4142		 * first one on the list is NOT the primary sctp_cmpaddr()
4143		 * is much more efficient if the primary is the first on the
4144		 * list, make it so.
4145		 */
4146		TAILQ_REMOVE(&stcb->asoc.nets,
4147		    stcb->asoc.primary_destination, sctp_next);
4148		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4149		    stcb->asoc.primary_destination, sctp_next);
4150	}
4151	return (0);
4152}
4153
4154
4155static uint32_t
4156sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4157{
4158	uint32_t id;
4159	struct sctpasochead *head;
4160	struct sctp_tcb *lstcb;
4161
4162	SCTP_INP_WLOCK(inp);
4163try_again:
4164	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4165		/* TSNH */
4166		SCTP_INP_WUNLOCK(inp);
4167		return (0);
4168	}
4169	/*
4170	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4171	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4172	 */
4173	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4174		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4175	}
4176	id = inp->sctp_associd_counter;
4177	inp->sctp_associd_counter++;
4178	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4179	if (lstcb) {
4180		goto try_again;
4181	}
4182	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4183	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4184	stcb->asoc.in_asocid_hash = 1;
4185	SCTP_INP_WUNLOCK(inp);
4186	return id;
4187}
4188
4189/*
4190 * allocate an association and add it to the endpoint. The caller must be
4191 * careful to add all additional addresses once they are know right away or
4192 * else the assoc will be may experience a blackout scenario.
4193 */
4194struct sctp_tcb *
4195sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4196    int *error, uint32_t override_tag, uint32_t vrf_id,
4197    uint16_t o_streams,
4198    struct thread *p
4199)
4200{
4201	/* note the p argument is only valid in unbound sockets */
4202
4203	struct sctp_tcb *stcb;
4204	struct sctp_association *asoc;
4205	struct sctpasochead *head;
4206	uint16_t rport;
4207	int err;
4208
4209	/*
4210	 * Assumption made here: Caller has done a
4211	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4212	 * address does not exist already.
4213	 */
4214	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4215		/* Hit max assoc, sorry no more */
4216		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4217		*error = ENOBUFS;
4218		return (NULL);
4219	}
4220	if (firstaddr == NULL) {
4221		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4222		*error = EINVAL;
4223		return (NULL);
4224	}
4225	SCTP_INP_RLOCK(inp);
4226	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4227	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4228	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4229		/*
4230		 * If its in the TCP pool, its NOT allowed to create an
4231		 * association. The parent listener needs to call
4232		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4233		 * off, or connected one does this.. its an error.
4234		 */
4235		SCTP_INP_RUNLOCK(inp);
4236		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4237		*error = EINVAL;
4238		return (NULL);
4239	}
4240	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4241	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4242		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4243		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4244			SCTP_INP_RUNLOCK(inp);
4245			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4246			*error = EINVAL;
4247			return (NULL);
4248		}
4249	}
4250	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4251#ifdef SCTP_DEBUG
4252	if (firstaddr) {
4253		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4254		switch (firstaddr->sa_family) {
4255#ifdef INET
4256		case AF_INET:
4257			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4258			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4259			break;
4260#endif
4261#ifdef INET6
4262		case AF_INET6:
4263			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4264			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4265			break;
4266#endif
4267		default:
4268			break;
4269		}
4270	} else {
4271		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4272	}
4273#endif				/* SCTP_DEBUG */
4274	switch (firstaddr->sa_family) {
4275#ifdef INET
4276	case AF_INET:
4277		{
4278			struct sockaddr_in *sin;
4279
4280			sin = (struct sockaddr_in *)firstaddr;
4281			if ((ntohs(sin->sin_port) == 0) ||
4282			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4283			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4284			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4285				/* Invalid address */
4286				SCTP_INP_RUNLOCK(inp);
4287				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4288				*error = EINVAL;
4289				return (NULL);
4290			}
4291			rport = sin->sin_port;
4292			break;
4293		}
4294#endif
4295#ifdef INET6
4296	case AF_INET6:
4297		{
4298			struct sockaddr_in6 *sin6;
4299
4300			sin6 = (struct sockaddr_in6 *)firstaddr;
4301			if ((ntohs(sin6->sin6_port) == 0) ||
4302			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4303			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4304				/* Invalid address */
4305				SCTP_INP_RUNLOCK(inp);
4306				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4307				*error = EINVAL;
4308				return (NULL);
4309			}
4310			rport = sin6->sin6_port;
4311			break;
4312		}
4313#endif
4314	default:
4315		/* not supported family type */
4316		SCTP_INP_RUNLOCK(inp);
4317		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4318		*error = EINVAL;
4319		return (NULL);
4320	}
4321	SCTP_INP_RUNLOCK(inp);
4322	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4323		/*
4324		 * If you have not performed a bind, then we need to do the
4325		 * ephemeral bind for you.
4326		 */
4327		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4328		    (struct sockaddr *)NULL,
4329		    (struct sctp_ifa *)NULL,
4330		    p
4331		    ))) {
4332			/* bind error, probably perm */
4333			*error = err;
4334			return (NULL);
4335		}
4336	}
4337	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4338	if (stcb == NULL) {
4339		/* out of memory? */
4340		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4341		*error = ENOMEM;
4342		return (NULL);
4343	}
4344	SCTP_INCR_ASOC_COUNT();
4345
4346	bzero(stcb, sizeof(*stcb));
4347	asoc = &stcb->asoc;
4348
4349	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4350	SCTP_TCB_LOCK_INIT(stcb);
4351	SCTP_TCB_SEND_LOCK_INIT(stcb);
4352	stcb->rport = rport;
4353	/* setup back pointer's */
4354	stcb->sctp_ep = inp;
4355	stcb->sctp_socket = inp->sctp_socket;
4356	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id, o_streams))) {
4357		/* failed */
4358		SCTP_TCB_LOCK_DESTROY(stcb);
4359		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4360		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4361		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4362		SCTP_DECR_ASOC_COUNT();
4363		*error = err;
4364		return (NULL);
4365	}
4366	/* and the port */
4367	SCTP_INP_INFO_WLOCK();
4368	SCTP_INP_WLOCK(inp);
4369	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4370		/* inpcb freed while alloc going on */
4371		SCTP_TCB_LOCK_DESTROY(stcb);
4372		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4373		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4374		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4375		SCTP_INP_WUNLOCK(inp);
4376		SCTP_INP_INFO_WUNLOCK();
4377		SCTP_DECR_ASOC_COUNT();
4378		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4379		*error = EINVAL;
4380		return (NULL);
4381	}
4382	SCTP_TCB_LOCK(stcb);
4383
4384	/* now that my_vtag is set, add it to the hash */
4385	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4386	/* put it in the bucket in the vtag hash of assoc's for the system */
4387	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4388	SCTP_INP_INFO_WUNLOCK();
4389
4390	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4391		/* failure.. memory error? */
4392		if (asoc->strmout) {
4393			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4394			asoc->strmout = NULL;
4395		}
4396		if (asoc->mapping_array) {
4397			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4398			asoc->mapping_array = NULL;
4399		}
4400		if (asoc->nr_mapping_array) {
4401			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4402			asoc->nr_mapping_array = NULL;
4403		}
4404		SCTP_DECR_ASOC_COUNT();
4405		SCTP_TCB_UNLOCK(stcb);
4406		SCTP_TCB_LOCK_DESTROY(stcb);
4407		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4408		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4409		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4410		SCTP_INP_WUNLOCK(inp);
4411		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4412		*error = ENOBUFS;
4413		return (NULL);
4414	}
4415	/* Init all the timers */
4416	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4417	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4418	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4419	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4420	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4421	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4422	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4423
4424	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4425	/* now file the port under the hash as well */
4426	if (inp->sctp_tcbhash != NULL) {
4427		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4428		    inp->sctp_hashmark)];
4429		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4430	}
4431	SCTP_INP_WUNLOCK(inp);
4432	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4433	return (stcb);
4434}
4435
4436
4437void
4438sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4439{
4440	struct sctp_association *asoc;
4441
4442	asoc = &stcb->asoc;
4443	asoc->numnets--;
4444	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4445	if (net == asoc->primary_destination) {
4446		/* Reset primary */
4447		struct sctp_nets *lnet;
4448
4449		lnet = TAILQ_FIRST(&asoc->nets);
4450		/*
4451		 * Mobility adaptation Ideally, if deleted destination is
4452		 * the primary, it becomes a fast retransmission trigger by
4453		 * the subsequent SET PRIMARY. (by micchie)
4454		 */
4455		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4456		    SCTP_MOBILITY_BASE) ||
4457		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4458		    SCTP_MOBILITY_FASTHANDOFF)) {
4459			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4460			if (asoc->deleted_primary != NULL) {
4461				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4462				goto out;
4463			}
4464			asoc->deleted_primary = net;
4465			atomic_add_int(&net->ref_count, 1);
4466			memset(&net->lastsa, 0, sizeof(net->lastsa));
4467			memset(&net->lastsv, 0, sizeof(net->lastsv));
4468			sctp_mobility_feature_on(stcb->sctp_ep,
4469			    SCTP_MOBILITY_PRIM_DELETED);
4470			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4471			    stcb->sctp_ep, stcb, NULL);
4472		}
4473out:
4474		/* Try to find a confirmed primary */
4475		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4476	}
4477	if (net == asoc->last_data_chunk_from) {
4478		/* Reset primary */
4479		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4480	}
4481	if (net == asoc->last_control_chunk_from) {
4482		/* Clear net */
4483		asoc->last_control_chunk_from = NULL;
4484	}
4485	if (net == stcb->asoc.alternate) {
4486		sctp_free_remote_addr(stcb->asoc.alternate);
4487		stcb->asoc.alternate = NULL;
4488	}
4489	sctp_free_remote_addr(net);
4490}
4491
4492/*
4493 * remove a remote endpoint address from an association, it will fail if the
4494 * address does not exist.
4495 */
4496int
4497sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4498{
4499	/*
4500	 * Here we need to remove a remote address. This is quite simple, we
4501	 * first find it in the list of address for the association
4502	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4503	 * on that item. Note we do not allow it to be removed if there are
4504	 * no other addresses.
4505	 */
4506	struct sctp_association *asoc;
4507	struct sctp_nets *net, *nnet;
4508
4509	asoc = &stcb->asoc;
4510
4511	/* locate the address */
4512	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4513		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4514			continue;
4515		}
4516		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4517		    remaddr)) {
4518			/* we found the guy */
4519			if (asoc->numnets < 2) {
4520				/* Must have at LEAST two remote addresses */
4521				return (-1);
4522			} else {
4523				sctp_remove_net(stcb, net);
4524				return (0);
4525			}
4526		}
4527	}
4528	/* not found. */
4529	return (-2);
4530}
4531
4532void
4533sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4534{
4535	struct sctpvtaghead *chain;
4536	struct sctp_tagblock *twait_block;
4537	int found = 0;
4538	int i;
4539
4540	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4541	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4542		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4543			if ((twait_block->vtag_block[i].v_tag == tag) &&
4544			    (twait_block->vtag_block[i].lport == lport) &&
4545			    (twait_block->vtag_block[i].rport == rport)) {
4546				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4547				twait_block->vtag_block[i].v_tag = 0;
4548				twait_block->vtag_block[i].lport = 0;
4549				twait_block->vtag_block[i].rport = 0;
4550				found = 1;
4551				break;
4552			}
4553		}
4554		if (found)
4555			break;
4556	}
4557}
4558
4559int
4560sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4561{
4562	struct sctpvtaghead *chain;
4563	struct sctp_tagblock *twait_block;
4564	int found = 0;
4565	int i;
4566
4567	SCTP_INP_INFO_WLOCK();
4568	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4569	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4570		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4571			if ((twait_block->vtag_block[i].v_tag == tag) &&
4572			    (twait_block->vtag_block[i].lport == lport) &&
4573			    (twait_block->vtag_block[i].rport == rport)) {
4574				found = 1;
4575				break;
4576			}
4577		}
4578		if (found)
4579			break;
4580	}
4581	SCTP_INP_INFO_WUNLOCK();
4582	return (found);
4583}
4584
4585
4586void
4587sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4588{
4589	struct sctpvtaghead *chain;
4590	struct sctp_tagblock *twait_block;
4591	struct timeval now;
4592	int set, i;
4593
4594	if (time == 0) {
4595		/* Its disabled */
4596		return;
4597	}
4598	(void)SCTP_GETTIME_TIMEVAL(&now);
4599	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4600	set = 0;
4601	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4602		/* Block(s) present, lets find space, and expire on the fly */
4603		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4604			if ((twait_block->vtag_block[i].v_tag == 0) &&
4605			    !set) {
4606				twait_block->vtag_block[i].tv_sec_at_expire =
4607				    now.tv_sec + time;
4608				twait_block->vtag_block[i].v_tag = tag;
4609				twait_block->vtag_block[i].lport = lport;
4610				twait_block->vtag_block[i].rport = rport;
4611				set = 1;
4612			} else if ((twait_block->vtag_block[i].v_tag) &&
4613			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4614				/* Audit expires this guy */
4615				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4616				twait_block->vtag_block[i].v_tag = 0;
4617				twait_block->vtag_block[i].lport = 0;
4618				twait_block->vtag_block[i].rport = 0;
4619				if (set == 0) {
4620					/* Reuse it for my new tag */
4621					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4622					twait_block->vtag_block[i].v_tag = tag;
4623					twait_block->vtag_block[i].lport = lport;
4624					twait_block->vtag_block[i].rport = rport;
4625					set = 1;
4626				}
4627			}
4628		}
4629		if (set) {
4630			/*
4631			 * We only do up to the block where we can place our
4632			 * tag for audits
4633			 */
4634			break;
4635		}
4636	}
4637	/* Need to add a new block to chain */
4638	if (!set) {
4639		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4640		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4641		if (twait_block == NULL) {
4642#ifdef INVARIANTS
4643			panic("Can not alloc tagblock");
4644#endif
4645			return;
4646		}
4647		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4648		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4649		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4650		twait_block->vtag_block[0].v_tag = tag;
4651		twait_block->vtag_block[0].lport = lport;
4652		twait_block->vtag_block[0].rport = rport;
4653	}
4654}
4655
4656
4657
4658/*-
4659 * Free the association after un-hashing the remote port. This
4660 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4661 * expect that the input to this function IS a locked TCB.
4662 * It will return 0, if it did NOT destroy the association (instead
4663 * it unlocks it. It will return NON-zero if it either destroyed the
4664 * association OR the association is already destroyed.
4665 */
4666int
4667sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4668{
4669	int i;
4670	struct sctp_association *asoc;
4671	struct sctp_nets *net, *nnet;
4672	struct sctp_laddr *laddr, *naddr;
4673	struct sctp_tmit_chunk *chk, *nchk;
4674	struct sctp_asconf_addr *aparam, *naparam;
4675	struct sctp_asconf_ack *aack, *naack;
4676	struct sctp_stream_reset_list *strrst, *nstrrst;
4677	struct sctp_queued_to_read *sq, *nsq;
4678	struct sctp_stream_queue_pending *sp, *nsp;
4679	sctp_sharedkey_t *shared_key, *nshared_key;
4680	struct socket *so;
4681
4682	/* first, lets purge the entry from the hash table. */
4683
4684#ifdef SCTP_LOG_CLOSING
4685	sctp_log_closing(inp, stcb, 6);
4686#endif
4687	if (stcb->asoc.state == 0) {
4688#ifdef SCTP_LOG_CLOSING
4689		sctp_log_closing(inp, NULL, 7);
4690#endif
4691		/* there is no asoc, really TSNH :-0 */
4692		return (1);
4693	}
4694	if (stcb->asoc.alternate) {
4695		sctp_free_remote_addr(stcb->asoc.alternate);
4696		stcb->asoc.alternate = NULL;
4697	}
4698	/* TEMP CODE */
4699	if (stcb->freed_from_where == 0) {
4700		/* Only record the first place free happened from */
4701		stcb->freed_from_where = from_location;
4702	}
4703	/* TEMP CODE */
4704
4705	asoc = &stcb->asoc;
4706	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4707	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4708		/* nothing around */
4709		so = NULL;
4710	else
4711		so = inp->sctp_socket;
4712
4713	/*
4714	 * We used timer based freeing if a reader or writer is in the way.
4715	 * So we first check if we are actually being called from a timer,
4716	 * if so we abort early if a reader or writer is still in the way.
4717	 */
4718	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4719	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4720		/*
4721		 * is it the timer driving us? if so are the reader/writers
4722		 * gone?
4723		 */
4724		if (stcb->asoc.refcnt) {
4725			/* nope, reader or writer in the way */
4726			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4727			/* no asoc destroyed */
4728			SCTP_TCB_UNLOCK(stcb);
4729#ifdef SCTP_LOG_CLOSING
4730			sctp_log_closing(inp, stcb, 8);
4731#endif
4732			return (0);
4733		}
4734	}
4735	/* now clean up any other timers */
4736	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4737	asoc->dack_timer.self = NULL;
4738	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4739	/*-
4740	 * For stream reset we don't blast this unless
4741	 * it is a str-reset timer, it might be the
4742	 * free-asoc timer which we DON'T want to
4743	 * disturb.
4744	 */
4745	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4746		asoc->strreset_timer.self = NULL;
4747	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4748	asoc->asconf_timer.self = NULL;
4749	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4750	asoc->autoclose_timer.self = NULL;
4751	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4752	asoc->shut_guard_timer.self = NULL;
4753	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4754	asoc->delayed_event_timer.self = NULL;
4755	/* Mobility adaptation */
4756	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4757	asoc->delete_prim_timer.self = NULL;
4758	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4759		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4760		net->rxt_timer.self = NULL;
4761		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4762		net->pmtu_timer.self = NULL;
4763		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4764		net->hb_timer.self = NULL;
4765	}
4766	/* Now the read queue needs to be cleaned up (only once) */
4767	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4768		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4769		SCTP_INP_READ_LOCK(inp);
4770		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4771			if (sq->stcb == stcb) {
4772				sq->do_not_ref_stcb = 1;
4773				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4774				/*
4775				 * If there is no end, there never will be
4776				 * now.
4777				 */
4778				if (sq->end_added == 0) {
4779					/* Held for PD-API clear that. */
4780					sq->pdapi_aborted = 1;
4781					sq->held_length = 0;
4782					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4783						/*
4784						 * Need to add a PD-API
4785						 * aborted indication.
4786						 * Setting the control_pdapi
4787						 * assures that it will be
4788						 * added right after this
4789						 * msg.
4790						 */
4791						uint32_t strseq;
4792
4793						stcb->asoc.control_pdapi = sq;
4794						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4795						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4796						    stcb,
4797						    SCTP_PARTIAL_DELIVERY_ABORTED,
4798						    (void *)&strseq,
4799						    SCTP_SO_LOCKED);
4800						stcb->asoc.control_pdapi = NULL;
4801					}
4802				}
4803				/* Add an end to wake them */
4804				sq->end_added = 1;
4805			}
4806		}
4807		SCTP_INP_READ_UNLOCK(inp);
4808		if (stcb->block_entry) {
4809			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4810			stcb->block_entry->error = ECONNRESET;
4811			stcb->block_entry = NULL;
4812		}
4813	}
4814	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4815		/*
4816		 * Someone holds a reference OR the socket is unaccepted
4817		 * yet.
4818		 */
4819		if ((stcb->asoc.refcnt) ||
4820		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4821		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4822			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4823			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4824		}
4825		SCTP_TCB_UNLOCK(stcb);
4826		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4827		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4828			/* nothing around */
4829			so = NULL;
4830		if (so) {
4831			/* Wake any reader/writers */
4832			sctp_sorwakeup(inp, so);
4833			sctp_sowwakeup(inp, so);
4834		}
4835#ifdef SCTP_LOG_CLOSING
4836		sctp_log_closing(inp, stcb, 9);
4837#endif
4838		/* no asoc destroyed */
4839		return (0);
4840	}
4841#ifdef SCTP_LOG_CLOSING
4842	sctp_log_closing(inp, stcb, 10);
4843#endif
4844	/*
4845	 * When I reach here, no others want to kill the assoc yet.. and I
4846	 * own the lock. Now its possible an abort comes in when I do the
4847	 * lock exchange below to grab all the locks to do the final take
4848	 * out. to prevent this we increment the count, which will start a
4849	 * timer and blow out above thus assuring us that we hold exclusive
4850	 * killing of the asoc. Note that after getting back the TCB lock we
4851	 * will go ahead and increment the counter back up and stop any
4852	 * timer a passing stranger may have started :-S
4853	 */
4854	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4855		atomic_add_int(&stcb->asoc.refcnt, 1);
4856
4857		SCTP_TCB_UNLOCK(stcb);
4858		SCTP_INP_INFO_WLOCK();
4859		SCTP_INP_WLOCK(inp);
4860		SCTP_TCB_LOCK(stcb);
4861	}
4862	/* Double check the GONE flag */
4863	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4864	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4865		/* nothing around */
4866		so = NULL;
4867
4868	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4869	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4870		/*
4871		 * For TCP type we need special handling when we are
4872		 * connected. We also include the peel'ed off ones to.
4873		 */
4874		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4875			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4876			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4877			if (so) {
4878				SOCK_LOCK(so);
4879				if (so->so_rcv.sb_cc == 0) {
4880					so->so_state &= ~(SS_ISCONNECTING |
4881					    SS_ISDISCONNECTING |
4882					    SS_ISCONFIRMING |
4883					    SS_ISCONNECTED);
4884				}
4885				socantrcvmore_locked(so);
4886				sctp_sowwakeup(inp, so);
4887				sctp_sorwakeup(inp, so);
4888				SCTP_SOWAKEUP(so);
4889			}
4890		}
4891	}
4892	/*
4893	 * Make it invalid too, that way if its about to run it will abort
4894	 * and return.
4895	 */
4896	/* re-increment the lock */
4897	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4898		atomic_add_int(&stcb->asoc.refcnt, -1);
4899	}
4900	if (stcb->asoc.refcnt) {
4901		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4902		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4903		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4904			SCTP_INP_INFO_WUNLOCK();
4905			SCTP_INP_WUNLOCK(inp);
4906		}
4907		SCTP_TCB_UNLOCK(stcb);
4908		return (0);
4909	}
4910	asoc->state = 0;
4911	if (inp->sctp_tcbhash) {
4912		LIST_REMOVE(stcb, sctp_tcbhash);
4913	}
4914	if (stcb->asoc.in_asocid_hash) {
4915		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4916	}
4917	/* Now lets remove it from the list of ALL associations in the EP */
4918	LIST_REMOVE(stcb, sctp_tcblist);
4919	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4920		SCTP_INP_INCR_REF(inp);
4921		SCTP_INP_WUNLOCK(inp);
4922	}
4923	/* pull from vtag hash */
4924	LIST_REMOVE(stcb, sctp_asocs);
4925	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4926	    inp->sctp_lport, stcb->rport);
4927
4928	/*
4929	 * Now restop the timers to be sure this is paranoia at is finest!
4930	 */
4931	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4932	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4933	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4934	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4935	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4936	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4937	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4938	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4939		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4940		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4941		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4942	}
4943
4944	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4945	/*
4946	 * The chunk lists and such SHOULD be empty but we check them just
4947	 * in case.
4948	 */
4949	/* anything on the wheel needs to be removed */
4950	for (i = 0; i < asoc->streamoutcnt; i++) {
4951		struct sctp_stream_out *outs;
4952
4953		outs = &asoc->strmout[i];
4954		/* now clean up any chunks here */
4955		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4956			TAILQ_REMOVE(&outs->outqueue, sp, next);
4957			sctp_free_spbufspace(stcb, asoc, sp);
4958			if (sp->data) {
4959				if (so) {
4960					/* Still an open socket - report */
4961					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4962					    0, (void *)sp, SCTP_SO_LOCKED);
4963				}
4964				if (sp->data) {
4965					sctp_m_freem(sp->data);
4966					sp->data = NULL;
4967					sp->tail_mbuf = NULL;
4968					sp->length = 0;
4969				}
4970			}
4971			if (sp->net) {
4972				sctp_free_remote_addr(sp->net);
4973				sp->net = NULL;
4974			}
4975			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4976		}
4977	}
4978	/* sa_ignore FREED_MEMORY */
4979	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4980		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4981		SCTP_FREE(strrst, SCTP_M_STRESET);
4982	}
4983	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4984		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4985		if (sq->data) {
4986			sctp_m_freem(sq->data);
4987			sq->data = NULL;
4988		}
4989		sctp_free_remote_addr(sq->whoFrom);
4990		sq->whoFrom = NULL;
4991		sq->stcb = NULL;
4992		/* Free the ctl entry */
4993		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4994		SCTP_DECR_READQ_COUNT();
4995		/* sa_ignore FREED_MEMORY */
4996	}
4997	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4998		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4999		if (chk->data) {
5000			sctp_m_freem(chk->data);
5001			chk->data = NULL;
5002		}
5003		if (chk->holds_key_ref)
5004			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5005		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5006		SCTP_DECR_CHK_COUNT();
5007		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
5008		asoc->free_chunk_cnt--;
5009		/* sa_ignore FREED_MEMORY */
5010	}
5011	/* pending send queue SHOULD be empty */
5012	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
5013		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
5014			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
5015#ifdef INVARIANTS
5016		} else {
5017			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
5018#endif
5019		}
5020		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
5021		if (chk->data) {
5022			if (so) {
5023				/* Still a socket? */
5024				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
5025				    0, chk, SCTP_SO_LOCKED);
5026			}
5027			if (chk->data) {
5028				sctp_m_freem(chk->data);
5029				chk->data = NULL;
5030			}
5031		}
5032		if (chk->holds_key_ref)
5033			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5034		if (chk->whoTo) {
5035			sctp_free_remote_addr(chk->whoTo);
5036			chk->whoTo = NULL;
5037		}
5038		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5039		SCTP_DECR_CHK_COUNT();
5040		/* sa_ignore FREED_MEMORY */
5041	}
5042	/* sent queue SHOULD be empty */
5043	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5044		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5045			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
5046				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
5047#ifdef INVARIANTS
5048			} else {
5049				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
5050#endif
5051			}
5052		}
5053		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5054		if (chk->data) {
5055			if (so) {
5056				/* Still a socket? */
5057				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5058				    0, chk, SCTP_SO_LOCKED);
5059			}
5060			if (chk->data) {
5061				sctp_m_freem(chk->data);
5062				chk->data = NULL;
5063			}
5064		}
5065		if (chk->holds_key_ref)
5066			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5067		sctp_free_remote_addr(chk->whoTo);
5068		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5069		SCTP_DECR_CHK_COUNT();
5070		/* sa_ignore FREED_MEMORY */
5071	}
5072#ifdef INVARIANTS
5073	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5074		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5075			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5076		}
5077	}
5078#endif
5079	/* control queue MAY not be empty */
5080	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5081		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5082		if (chk->data) {
5083			sctp_m_freem(chk->data);
5084			chk->data = NULL;
5085		}
5086		if (chk->holds_key_ref)
5087			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5088		sctp_free_remote_addr(chk->whoTo);
5089		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5090		SCTP_DECR_CHK_COUNT();
5091		/* sa_ignore FREED_MEMORY */
5092	}
5093	/* ASCONF queue MAY not be empty */
5094	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5095		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5096		if (chk->data) {
5097			sctp_m_freem(chk->data);
5098			chk->data = NULL;
5099		}
5100		if (chk->holds_key_ref)
5101			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5102		sctp_free_remote_addr(chk->whoTo);
5103		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5104		SCTP_DECR_CHK_COUNT();
5105		/* sa_ignore FREED_MEMORY */
5106	}
5107	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5108		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5109		if (chk->data) {
5110			sctp_m_freem(chk->data);
5111			chk->data = NULL;
5112		}
5113		if (chk->holds_key_ref)
5114			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5115		sctp_free_remote_addr(chk->whoTo);
5116		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5117		SCTP_DECR_CHK_COUNT();
5118		/* sa_ignore FREED_MEMORY */
5119	}
5120
5121	if (asoc->mapping_array) {
5122		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5123		asoc->mapping_array = NULL;
5124	}
5125	if (asoc->nr_mapping_array) {
5126		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5127		asoc->nr_mapping_array = NULL;
5128	}
5129	/* the stream outs */
5130	if (asoc->strmout) {
5131		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5132		asoc->strmout = NULL;
5133	}
5134	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5135	if (asoc->strmin) {
5136		struct sctp_queued_to_read *ctl, *nctl;
5137
5138		for (i = 0; i < asoc->streamincnt; i++) {
5139			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5140				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5141				sctp_free_remote_addr(ctl->whoFrom);
5142				if (ctl->data) {
5143					sctp_m_freem(ctl->data);
5144					ctl->data = NULL;
5145				}
5146				/*
5147				 * We don't free the address here since all
5148				 * the net's were freed above.
5149				 */
5150				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5151				SCTP_DECR_READQ_COUNT();
5152			}
5153		}
5154		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5155		asoc->strmin = NULL;
5156	}
5157	asoc->streamincnt = 0;
5158	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5159#ifdef INVARIANTS
5160		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5161			panic("no net's left alloc'ed, or list points to itself");
5162		}
5163#endif
5164		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5165		sctp_free_remote_addr(net);
5166	}
5167	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5168		/* sa_ignore FREED_MEMORY */
5169		sctp_remove_laddr(laddr);
5170	}
5171
5172	/* pending asconf (address) parameters */
5173	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5174		/* sa_ignore FREED_MEMORY */
5175		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5176		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5177	}
5178	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5179		/* sa_ignore FREED_MEMORY */
5180		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5181		if (aack->data != NULL) {
5182			sctp_m_freem(aack->data);
5183		}
5184		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5185	}
5186	/* clean up auth stuff */
5187	if (asoc->local_hmacs)
5188		sctp_free_hmaclist(asoc->local_hmacs);
5189	if (asoc->peer_hmacs)
5190		sctp_free_hmaclist(asoc->peer_hmacs);
5191
5192	if (asoc->local_auth_chunks)
5193		sctp_free_chunklist(asoc->local_auth_chunks);
5194	if (asoc->peer_auth_chunks)
5195		sctp_free_chunklist(asoc->peer_auth_chunks);
5196
5197	sctp_free_authinfo(&asoc->authinfo);
5198
5199	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5200		LIST_REMOVE(shared_key, next);
5201		sctp_free_sharedkey(shared_key);
5202		/* sa_ignore FREED_MEMORY */
5203	}
5204
5205	/* Insert new items here :> */
5206
5207	/* Get rid of LOCK */
5208	SCTP_TCB_UNLOCK(stcb);
5209	SCTP_TCB_LOCK_DESTROY(stcb);
5210	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5211	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5212		SCTP_INP_INFO_WUNLOCK();
5213		SCTP_INP_RLOCK(inp);
5214	}
5215#ifdef SCTP_TRACK_FREED_ASOCS
5216	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5217		/* now clean up the tasoc itself */
5218		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5219		SCTP_DECR_ASOC_COUNT();
5220	} else {
5221		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5222	}
5223#else
5224	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5225	SCTP_DECR_ASOC_COUNT();
5226#endif
5227	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5228		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5229			/*
5230			 * If its NOT the inp_free calling us AND sctp_close
5231			 * as been called, we call back...
5232			 */
5233			SCTP_INP_RUNLOCK(inp);
5234			/*
5235			 * This will start the kill timer (if we are the
5236			 * last one) since we hold an increment yet. But
5237			 * this is the only safe way to do this since
5238			 * otherwise if the socket closes at the same time
5239			 * we are here we might collide in the cleanup.
5240			 */
5241			sctp_inpcb_free(inp,
5242			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5243			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5244			SCTP_INP_DECR_REF(inp);
5245			goto out_of;
5246		} else {
5247			/* The socket is still open. */
5248			SCTP_INP_DECR_REF(inp);
5249		}
5250	}
5251	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5252		SCTP_INP_RUNLOCK(inp);
5253	}
5254out_of:
5255	/* destroyed the asoc */
5256#ifdef SCTP_LOG_CLOSING
5257	sctp_log_closing(inp, NULL, 11);
5258#endif
5259	return (1);
5260}
5261
5262
5263
5264/*
5265 * determine if a destination is "reachable" based upon the addresses bound
5266 * to the current endpoint (e.g. only v4 or v6 currently bound)
5267 */
5268/*
5269 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5270 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5271 * types bound as its endpoint
5272 */
5273int
5274sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5275{
5276	struct sctp_inpcb *inp;
5277	int answer;
5278
5279	/*
5280	 * No locks here, the TCB, in all cases is already locked and an
5281	 * assoc is up. There is either a INP lock by the caller applied (in
5282	 * asconf case when deleting an address) or NOT in the HB case,
5283	 * however if HB then the INP increment is up and the INP will not
5284	 * be removed (on top of the fact that we have a TCB lock). So we
5285	 * only want to read the sctp_flags, which is either bound-all or
5286	 * not.. no protection needed since once an assoc is up you can't be
5287	 * changing your binding.
5288	 */
5289	inp = stcb->sctp_ep;
5290	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5291		/* if bound all, destination is not restricted */
5292		/*
5293		 * RRS: Question during lock work: Is this correct? If you
5294		 * are bound-all you still might need to obey the V4--V6
5295		 * flags??? IMO this bound-all stuff needs to be removed!
5296		 */
5297		return (1);
5298	}
5299	/* NOTE: all "scope" checks are done when local addresses are added */
5300	switch (destaddr->sa_family) {
5301#ifdef INET6
5302	case AF_INET6:
5303		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5304		break;
5305#endif
5306#ifdef INET
5307	case AF_INET:
5308		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5309		break;
5310#endif
5311	default:
5312		/* invalid family, so it's unreachable */
5313		answer = 0;
5314		break;
5315	}
5316	return (answer);
5317}
5318
5319/*
5320 * update the inp_vflags on an endpoint
5321 */
5322static void
5323sctp_update_ep_vflag(struct sctp_inpcb *inp)
5324{
5325	struct sctp_laddr *laddr;
5326
5327	/* first clear the flag */
5328	inp->ip_inp.inp.inp_vflag = 0;
5329	/* set the flag based on addresses on the ep list */
5330	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5331		if (laddr->ifa == NULL) {
5332			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5333			    __func__);
5334			continue;
5335		}
5336		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5337			continue;
5338		}
5339		switch (laddr->ifa->address.sa.sa_family) {
5340#ifdef INET6
5341		case AF_INET6:
5342			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5343			break;
5344#endif
5345#ifdef INET
5346		case AF_INET:
5347			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5348			break;
5349#endif
5350		default:
5351			break;
5352		}
5353	}
5354}
5355
5356/*
5357 * Add the address to the endpoint local address list There is nothing to be
5358 * done if we are bound to all addresses
5359 */
5360void
5361sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5362{
5363	struct sctp_laddr *laddr;
5364	int fnd, error = 0;
5365
5366	fnd = 0;
5367
5368	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5369		/* You are already bound to all. You have it already */
5370		return;
5371	}
5372#ifdef INET6
5373	if (ifa->address.sa.sa_family == AF_INET6) {
5374		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5375			/* Can't bind a non-useable addr. */
5376			return;
5377		}
5378	}
5379#endif
5380	/* first, is it already present? */
5381	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5382		if (laddr->ifa == ifa) {
5383			fnd = 1;
5384			break;
5385		}
5386	}
5387
5388	if (fnd == 0) {
5389		/* Not in the ep list */
5390		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5391		if (error != 0)
5392			return;
5393		inp->laddr_count++;
5394		/* update inp_vflag flags */
5395		switch (ifa->address.sa.sa_family) {
5396#ifdef INET6
5397		case AF_INET6:
5398			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5399			break;
5400#endif
5401#ifdef INET
5402		case AF_INET:
5403			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5404			break;
5405#endif
5406		default:
5407			break;
5408		}
5409	}
5410	return;
5411}
5412
5413
5414/*
5415 * select a new (hopefully reachable) destination net (should only be used
5416 * when we deleted an ep addr that is the only usable source address to reach
5417 * the destination net)
5418 */
5419static void
5420sctp_select_primary_destination(struct sctp_tcb *stcb)
5421{
5422	struct sctp_nets *net;
5423
5424	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5425		/* for now, we'll just pick the first reachable one we find */
5426		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5427			continue;
5428		if (sctp_destination_is_reachable(stcb,
5429		    (struct sockaddr *)&net->ro._l_addr)) {
5430			/* found a reachable destination */
5431			stcb->asoc.primary_destination = net;
5432		}
5433	}
5434	/* I can't there from here! ...we're gonna die shortly... */
5435}
5436
5437
5438/*
5439 * Delete the address from the endpoint local address list. There is nothing
5440 * to be done if we are bound to all addresses
5441 */
5442void
5443sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5444{
5445	struct sctp_laddr *laddr;
5446	int fnd;
5447
5448	fnd = 0;
5449	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5450		/* You are already bound to all. You have it already */
5451		return;
5452	}
5453	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5454		if (laddr->ifa == ifa) {
5455			fnd = 1;
5456			break;
5457		}
5458	}
5459	if (fnd && (inp->laddr_count < 2)) {
5460		/* can't delete unless there are at LEAST 2 addresses */
5461		return;
5462	}
5463	if (fnd) {
5464		/*
5465		 * clean up any use of this address go through our
5466		 * associations and clear any last_used_address that match
5467		 * this one for each assoc, see if a new primary_destination
5468		 * is needed
5469		 */
5470		struct sctp_tcb *stcb;
5471
5472		/* clean up "next_addr_touse" */
5473		if (inp->next_addr_touse == laddr)
5474			/* delete this address */
5475			inp->next_addr_touse = NULL;
5476
5477		/* clean up "last_used_address" */
5478		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5479			struct sctp_nets *net;
5480
5481			SCTP_TCB_LOCK(stcb);
5482			if (stcb->asoc.last_used_address == laddr)
5483				/* delete this address */
5484				stcb->asoc.last_used_address = NULL;
5485			/*
5486			 * Now spin through all the nets and purge any ref
5487			 * to laddr
5488			 */
5489			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5490				if (net->ro._s_addr == laddr->ifa) {
5491					/* Yep, purge src address selected */
5492					sctp_rtentry_t *rt;
5493
5494					/* delete this address if cached */
5495					rt = net->ro.ro_rt;
5496					if (rt != NULL) {
5497						RTFREE(rt);
5498						net->ro.ro_rt = NULL;
5499					}
5500					sctp_free_ifa(net->ro._s_addr);
5501					net->ro._s_addr = NULL;
5502					net->src_addr_selected = 0;
5503				}
5504			}
5505			SCTP_TCB_UNLOCK(stcb);
5506		}		/* for each tcb */
5507		/* remove it from the ep list */
5508		sctp_remove_laddr(laddr);
5509		inp->laddr_count--;
5510		/* update inp_vflag flags */
5511		sctp_update_ep_vflag(inp);
5512	}
5513	return;
5514}
5515
5516/*
5517 * Add the address to the TCB local address restricted list.
5518 * This is a "pending" address list (eg. addresses waiting for an
5519 * ASCONF-ACK response) and cannot be used as a valid source address.
5520 */
5521void
5522sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5523{
5524	struct sctp_laddr *laddr;
5525	struct sctpladdr *list;
5526
5527	/*
5528	 * Assumes TCB is locked.. and possibly the INP. May need to
5529	 * confirm/fix that if we need it and is not the case.
5530	 */
5531	list = &stcb->asoc.sctp_restricted_addrs;
5532
5533#ifdef INET6
5534	if (ifa->address.sa.sa_family == AF_INET6) {
5535		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5536			/* Can't bind a non-existent addr. */
5537			return;
5538		}
5539	}
5540#endif
5541	/* does the address already exist? */
5542	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5543		if (laddr->ifa == ifa) {
5544			return;
5545		}
5546	}
5547
5548	/* add to the list */
5549	(void)sctp_insert_laddr(list, ifa, 0);
5550	return;
5551}
5552
5553/*
5554 * Remove a local address from the TCB local address restricted list
5555 */
5556void
5557sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5558{
5559	struct sctp_inpcb *inp;
5560	struct sctp_laddr *laddr;
5561
5562	/*
5563	 * This is called by asconf work. It is assumed that a) The TCB is
5564	 * locked and b) The INP is locked. This is true in as much as I can
5565	 * trace through the entry asconf code where I did these locks.
5566	 * Again, the ASCONF code is a bit different in that it does lock
5567	 * the INP during its work often times. This must be since we don't
5568	 * want other proc's looking up things while what they are looking
5569	 * up is changing :-D
5570	 */
5571
5572	inp = stcb->sctp_ep;
5573	/* if subset bound and don't allow ASCONF's, can't delete last */
5574	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5575	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5576		if (stcb->sctp_ep->laddr_count < 2) {
5577			/* can't delete last address */
5578			return;
5579		}
5580	}
5581	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5582		/* remove the address if it exists */
5583		if (laddr->ifa == NULL)
5584			continue;
5585		if (laddr->ifa == ifa) {
5586			sctp_remove_laddr(laddr);
5587			return;
5588		}
5589	}
5590
5591	/* address not found! */
5592	return;
5593}
5594
5595/*
5596 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5597 */
5598/* sysctl */
5599static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5600static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5601
5602
5603
5604#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5605struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5606int *sctp_cpuarry = NULL;
5607void
5608sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5609{
5610	/* Queue a packet to a processor for the specified core */
5611	struct sctp_mcore_queue *qent;
5612	struct sctp_mcore_ctrl *wkq;
5613	int need_wake = 0;
5614
5615	if (sctp_mcore_workers == NULL) {
5616		/* Something went way bad during setup */
5617		sctp_input_with_port(m, off, 0);
5618		return;
5619	}
5620	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5621	    (sizeof(struct sctp_mcore_queue)),
5622	    SCTP_M_MCORE);
5623	if (qent == NULL) {
5624		/* This is trouble  */
5625		sctp_input_with_port(m, off, 0);
5626		return;
5627	}
5628	qent->vn = curvnet;
5629	qent->m = m;
5630	qent->off = off;
5631	qent->v6 = 0;
5632	wkq = &sctp_mcore_workers[cpu_to_use];
5633	SCTP_MCORE_QLOCK(wkq);
5634
5635	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5636	if (wkq->running == 0) {
5637		need_wake = 1;
5638	}
5639	SCTP_MCORE_QUNLOCK(wkq);
5640	if (need_wake) {
5641		wakeup(&wkq->running);
5642	}
5643}
5644
5645static void
5646sctp_mcore_thread(void *arg)
5647{
5648
5649	struct sctp_mcore_ctrl *wkq;
5650	struct sctp_mcore_queue *qent;
5651
5652	wkq = (struct sctp_mcore_ctrl *)arg;
5653	struct mbuf *m;
5654	int off, v6;
5655
5656	/* Wait for first tickle */
5657	SCTP_MCORE_LOCK(wkq);
5658	wkq->running = 0;
5659	msleep(&wkq->running,
5660	    &wkq->core_mtx,
5661	    0, "wait for pkt", 0);
5662	SCTP_MCORE_UNLOCK(wkq);
5663
5664	/* Bind to our cpu */
5665	thread_lock(curthread);
5666	sched_bind(curthread, wkq->cpuid);
5667	thread_unlock(curthread);
5668
5669	/* Now lets start working */
5670	SCTP_MCORE_LOCK(wkq);
5671	/* Now grab lock and go */
5672	for (;;) {
5673		SCTP_MCORE_QLOCK(wkq);
5674skip_sleep:
5675		wkq->running = 1;
5676		qent = TAILQ_FIRST(&wkq->que);
5677		if (qent) {
5678			TAILQ_REMOVE(&wkq->que, qent, next);
5679			SCTP_MCORE_QUNLOCK(wkq);
5680			CURVNET_SET(qent->vn);
5681			m = qent->m;
5682			off = qent->off;
5683			v6 = qent->v6;
5684			SCTP_FREE(qent, SCTP_M_MCORE);
5685			if (v6 == 0) {
5686				sctp_input_with_port(m, off, 0);
5687			} else {
5688				SCTP_PRINTF("V6 not yet supported\n");
5689				sctp_m_freem(m);
5690			}
5691			CURVNET_RESTORE();
5692			SCTP_MCORE_QLOCK(wkq);
5693		}
5694		wkq->running = 0;
5695		if (!TAILQ_EMPTY(&wkq->que)) {
5696			goto skip_sleep;
5697		}
5698		SCTP_MCORE_QUNLOCK(wkq);
5699		msleep(&wkq->running,
5700		    &wkq->core_mtx,
5701		    0, "wait for pkt", 0);
5702	}
5703}
5704
5705static void
5706sctp_startup_mcore_threads(void)
5707{
5708	int i, cpu;
5709
5710	if (mp_ncpus == 1)
5711		return;
5712
5713	if (sctp_mcore_workers != NULL) {
5714		/*
5715		 * Already been here in some previous vnet?
5716		 */
5717		return;
5718	}
5719	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5720	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5721	    SCTP_M_MCORE);
5722	if (sctp_mcore_workers == NULL) {
5723		/* TSNH I hope */
5724		return;
5725	}
5726	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5727	    sizeof(struct sctp_mcore_ctrl)));
5728	/* Init the structures */
5729	for (i = 0; i <= mp_maxid; i++) {
5730		TAILQ_INIT(&sctp_mcore_workers[i].que);
5731		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5732		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5733		sctp_mcore_workers[i].cpuid = i;
5734	}
5735	if (sctp_cpuarry == NULL) {
5736		SCTP_MALLOC(sctp_cpuarry, int *,
5737		    (mp_ncpus * sizeof(int)),
5738		    SCTP_M_MCORE);
5739		i = 0;
5740		CPU_FOREACH(cpu) {
5741			sctp_cpuarry[i] = cpu;
5742			i++;
5743		}
5744	}
5745	/* Now start them all */
5746	CPU_FOREACH(cpu) {
5747		(void)kproc_create(sctp_mcore_thread,
5748		    (void *)&sctp_mcore_workers[cpu],
5749		    &sctp_mcore_workers[cpu].thread_proc,
5750		    RFPROC,
5751		    SCTP_KTHREAD_PAGES,
5752		    SCTP_MCORE_NAME);
5753
5754	}
5755}
5756
5757#endif
5758
5759void
5760sctp_pcb_init()
5761{
5762	/*
5763	 * SCTP initialization for the PCB structures should be called by
5764	 * the sctp_init() funciton.
5765	 */
5766	int i;
5767	struct timeval tv;
5768
5769	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5770		/* error I was called twice */
5771		return;
5772	}
5773	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5774
5775#if defined(SCTP_LOCAL_TRACE_BUF)
5776	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5777#endif
5778#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5779	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5780	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5781	    SCTP_M_MCORE);
5782#endif
5783	(void)SCTP_GETTIME_TIMEVAL(&tv);
5784#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5785	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5786	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5787	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5788#else
5789	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5790	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5791	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5792#endif
5793	/* init the empty list of (All) Endpoints */
5794	LIST_INIT(&SCTP_BASE_INFO(listhead));
5795
5796
5797	/* init the hash table of endpoints */
5798	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5799	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5800	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5801	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5802	    &SCTP_BASE_INFO(hashasocmark));
5803	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5804	    &SCTP_BASE_INFO(hashmark));
5805	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5806	    &SCTP_BASE_INFO(hashtcpmark));
5807	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5808
5809
5810	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5811	    &SCTP_BASE_INFO(hashvrfmark));
5812
5813	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5814	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5815	/* init the zones */
5816	/*
5817	 * FIX ME: Should check for NULL returns, but if it does fail we are
5818	 * doomed to panic anyways... add later maybe.
5819	 */
5820	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5821	    sizeof(struct sctp_inpcb), maxsockets);
5822
5823	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5824	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5825
5826	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5827	    sizeof(struct sctp_laddr),
5828	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5829
5830	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5831	    sizeof(struct sctp_nets),
5832	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5833
5834	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5835	    sizeof(struct sctp_tmit_chunk),
5836	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5837
5838	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5839	    sizeof(struct sctp_queued_to_read),
5840	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5841
5842	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5843	    sizeof(struct sctp_stream_queue_pending),
5844	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5845
5846	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5847	    sizeof(struct sctp_asconf),
5848	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5849
5850	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5851	    sizeof(struct sctp_asconf_ack),
5852	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5853
5854
5855	/* Master Lock INIT for info structure */
5856	SCTP_INP_INFO_LOCK_INIT();
5857	SCTP_STATLOG_INIT_LOCK();
5858
5859	SCTP_IPI_COUNT_INIT();
5860	SCTP_IPI_ADDR_INIT();
5861#ifdef SCTP_PACKET_LOGGING
5862	SCTP_IP_PKTLOG_INIT();
5863#endif
5864	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5865
5866	SCTP_WQ_ADDR_INIT();
5867	/* not sure if we need all the counts */
5868	SCTP_BASE_INFO(ipi_count_ep) = 0;
5869	/* assoc/tcb zone info */
5870	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5871	/* local addrlist zone info */
5872	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5873	/* remote addrlist zone info */
5874	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5875	/* chunk info */
5876	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5877
5878	/* socket queue zone info */
5879	SCTP_BASE_INFO(ipi_count_readq) = 0;
5880
5881	/* stream out queue cont */
5882	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5883
5884	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5885	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5886
5887	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5888
5889	/* Init the TIMEWAIT list */
5890	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5891		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5892	}
5893	sctp_startup_iterator();
5894
5895#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5896	sctp_startup_mcore_threads();
5897#endif
5898
5899	/*
5900	 * INIT the default VRF which for BSD is the only one, other O/S's
5901	 * may have more. But initially they must start with one and then
5902	 * add the VRF's as addresses are added.
5903	 */
5904	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5905}
5906
5907/*
5908 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5909 */
5910void
5911sctp_pcb_finish(void)
5912{
5913	struct sctp_vrflist *vrf_bucket;
5914	struct sctp_vrf *vrf, *nvrf;
5915	struct sctp_ifn *ifn, *nifn;
5916	struct sctp_ifa *ifa, *nifa;
5917	struct sctpvtaghead *chain;
5918	struct sctp_tagblock *twait_block, *prev_twait_block;
5919	struct sctp_laddr *wi, *nwi;
5920	int i;
5921	struct sctp_iterator *it, *nit;
5922
5923	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
5924		SCTP_PRINTF("%s: race condition on teardown.\n", __func__);
5925		return;
5926	}
5927	SCTP_BASE_VAR(sctp_pcb_initialized) = 0;
5928	/*
5929	 * In FreeBSD the iterator thread never exits but we do clean up.
5930	 * The only way FreeBSD reaches here is if we have VRF's but we
5931	 * still add the ifdef to make it compile on old versions.
5932	 */
5933retry:
5934	SCTP_IPI_ITERATOR_WQ_LOCK();
5935	/*
5936	 * sctp_iterator_worker() might be working on an it entry without
5937	 * holding the lock.  We won't find it on the list either and
5938	 * continue and free/destroy it.  While holding the lock, spin, to
5939	 * avoid the race condition as sctp_iterator_worker() will have to
5940	 * wait to re-aquire the lock.
5941	 */
5942	if (sctp_it_ctl.iterator_running != 0 || sctp_it_ctl.cur_it != NULL) {
5943		SCTP_IPI_ITERATOR_WQ_UNLOCK();
5944		SCTP_PRINTF("%s: Iterator running while we held the lock. Retry. "
5945		    "cur_it=%p\n", __func__, sctp_it_ctl.cur_it);
5946		DELAY(10);
5947		goto retry;
5948	}
5949	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5950		if (it->vn != curvnet) {
5951			continue;
5952		}
5953		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5954		if (it->function_atend != NULL) {
5955			(*it->function_atend) (it->pointer, it->val);
5956		}
5957		SCTP_FREE(it, SCTP_M_ITER);
5958	}
5959	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5960	SCTP_ITERATOR_LOCK();
5961	if ((sctp_it_ctl.cur_it) &&
5962	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5963		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5964	}
5965	SCTP_ITERATOR_UNLOCK();
5966	SCTP_OS_TIMER_STOP_DRAIN(&SCTP_BASE_INFO(addr_wq_timer.timer));
5967	SCTP_WQ_ADDR_LOCK();
5968	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5969		LIST_REMOVE(wi, sctp_nxt_addr);
5970		SCTP_DECR_LADDR_COUNT();
5971		if (wi->action == SCTP_DEL_IP_ADDRESS) {
5972			SCTP_FREE(wi->ifa, SCTP_M_IFA);
5973		}
5974		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5975	}
5976	SCTP_WQ_ADDR_UNLOCK();
5977
5978	/*
5979	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5980	 * destroyed first).
5981	 */
5982	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5983	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5984		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5985			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5986				/* free the ifa */
5987				LIST_REMOVE(ifa, next_bucket);
5988				LIST_REMOVE(ifa, next_ifa);
5989				SCTP_FREE(ifa, SCTP_M_IFA);
5990			}
5991			/* free the ifn */
5992			LIST_REMOVE(ifn, next_bucket);
5993			LIST_REMOVE(ifn, next_ifn);
5994			SCTP_FREE(ifn, SCTP_M_IFN);
5995		}
5996		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5997		/* free the vrf */
5998		LIST_REMOVE(vrf, next_vrf);
5999		SCTP_FREE(vrf, SCTP_M_VRF);
6000	}
6001	/* free the vrf hashes */
6002	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
6003	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
6004
6005	/*
6006	 * free the TIMEWAIT list elements malloc'd in the function
6007	 * sctp_add_vtag_to_timewait()...
6008	 */
6009	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
6010		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
6011		if (!LIST_EMPTY(chain)) {
6012			prev_twait_block = NULL;
6013			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6014				if (prev_twait_block) {
6015					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6016				}
6017				prev_twait_block = twait_block;
6018			}
6019			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
6020		}
6021	}
6022
6023	/* free the locks and mutexes */
6024#ifdef SCTP_PACKET_LOGGING
6025	SCTP_IP_PKTLOG_DESTROY();
6026#endif
6027	SCTP_IPI_ADDR_DESTROY();
6028	SCTP_STATLOG_DESTROY();
6029	SCTP_INP_INFO_LOCK_DESTROY();
6030
6031	SCTP_WQ_ADDR_DESTROY();
6032
6033	/* Get rid of other stuff too. */
6034	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
6035		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
6036	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
6037		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6038	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6039		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6040
6041	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6042	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6043	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6044	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6045	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6046	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6047	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6048	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6049	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6050#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6051	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6052#endif
6053}
6054
6055
6056int
6057sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6058    int offset, int limit,
6059    struct sockaddr *src, struct sockaddr *dst,
6060    struct sockaddr *altsa)
6061{
6062	/*
6063	 * grub through the INIT pulling addresses and loading them to the
6064	 * nets structure in the asoc. The from address in the mbuf should
6065	 * also be loaded (if it is not already). This routine can be called
6066	 * with either INIT or INIT-ACK's as long as the m points to the IP
6067	 * packet and the offset points to the beginning of the parameters.
6068	 */
6069	struct sctp_inpcb *inp;
6070	struct sctp_nets *net, *nnet, *net_tmp;
6071	struct sctp_paramhdr *phdr, parm_buf;
6072	struct sctp_tcb *stcb_tmp;
6073	uint16_t ptype, plen;
6074	struct sockaddr *sa;
6075	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6076	struct sctp_auth_random *p_random = NULL;
6077	uint16_t random_len = 0;
6078	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6079	struct sctp_auth_hmac_algo *hmacs = NULL;
6080	uint16_t hmacs_len = 0;
6081	uint8_t saw_asconf = 0;
6082	uint8_t saw_asconf_ack = 0;
6083	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6084	struct sctp_auth_chunk_list *chunks = NULL;
6085	uint16_t num_chunks = 0;
6086	sctp_key_t *new_key;
6087	uint32_t keylen;
6088	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6089	uint8_t peer_supports_ecn;
6090	uint8_t peer_supports_prsctp;
6091	uint8_t peer_supports_auth;
6092	uint8_t peer_supports_asconf;
6093	uint8_t peer_supports_asconf_ack;
6094	uint8_t peer_supports_reconfig;
6095	uint8_t peer_supports_nrsack;
6096	uint8_t peer_supports_pktdrop;
6097
6098#ifdef INET
6099	struct sockaddr_in sin;
6100
6101#endif
6102#ifdef INET6
6103	struct sockaddr_in6 sin6;
6104
6105#endif
6106
6107	/* First get the destination address setup too. */
6108#ifdef INET
6109	memset(&sin, 0, sizeof(sin));
6110	sin.sin_family = AF_INET;
6111	sin.sin_len = sizeof(sin);
6112	sin.sin_port = stcb->rport;
6113#endif
6114#ifdef INET6
6115	memset(&sin6, 0, sizeof(sin6));
6116	sin6.sin6_family = AF_INET6;
6117	sin6.sin6_len = sizeof(struct sockaddr_in6);
6118	sin6.sin6_port = stcb->rport;
6119#endif
6120	if (altsa) {
6121		sa = altsa;
6122	} else {
6123		sa = src;
6124	}
6125	peer_supports_ecn = 0;
6126	peer_supports_prsctp = 0;
6127	peer_supports_auth = 0;
6128	peer_supports_asconf = 0;
6129	peer_supports_reconfig = 0;
6130	peer_supports_nrsack = 0;
6131	peer_supports_pktdrop = 0;
6132	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6133		/* mark all addresses that we have currently on the list */
6134		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6135	}
6136	/* does the source address already exist? if so skip it */
6137	inp = stcb->sctp_ep;
6138	atomic_add_int(&stcb->asoc.refcnt, 1);
6139	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6140	atomic_add_int(&stcb->asoc.refcnt, -1);
6141
6142	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6143		/* we must add the source address */
6144		/* no scope set here since we have a tcb already. */
6145		switch (sa->sa_family) {
6146#ifdef INET
6147		case AF_INET:
6148			if (stcb->asoc.scope.ipv4_addr_legal) {
6149				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6150					return (-1);
6151				}
6152			}
6153			break;
6154#endif
6155#ifdef INET6
6156		case AF_INET6:
6157			if (stcb->asoc.scope.ipv6_addr_legal) {
6158				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6159					return (-2);
6160				}
6161			}
6162			break;
6163#endif
6164		default:
6165			break;
6166		}
6167	} else {
6168		if (net_tmp != NULL && stcb_tmp == stcb) {
6169			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6170		} else if (stcb_tmp != stcb) {
6171			/* It belongs to another association? */
6172			if (stcb_tmp)
6173				SCTP_TCB_UNLOCK(stcb_tmp);
6174			return (-3);
6175		}
6176	}
6177	if (stcb->asoc.state == 0) {
6178		/* the assoc was freed? */
6179		return (-4);
6180	}
6181	/* now we must go through each of the params. */
6182	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6183	while (phdr) {
6184		ptype = ntohs(phdr->param_type);
6185		plen = ntohs(phdr->param_length);
6186		/*
6187		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6188		 * (uint32_t)ptype, (int)plen);
6189		 */
6190		if (offset + plen > limit) {
6191			break;
6192		}
6193		if (plen == 0) {
6194			break;
6195		}
6196#ifdef INET
6197		if (ptype == SCTP_IPV4_ADDRESS) {
6198			if (stcb->asoc.scope.ipv4_addr_legal) {
6199				struct sctp_ipv4addr_param *p4, p4_buf;
6200
6201				/* ok get the v4 address and check/add */
6202				phdr = sctp_get_next_param(m, offset,
6203				    (struct sctp_paramhdr *)&p4_buf,
6204				    sizeof(p4_buf));
6205				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6206				    phdr == NULL) {
6207					return (-5);
6208				}
6209				p4 = (struct sctp_ipv4addr_param *)phdr;
6210				sin.sin_addr.s_addr = p4->addr;
6211				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6212					/* Skip multi-cast addresses */
6213					goto next_param;
6214				}
6215				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6216				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6217					goto next_param;
6218				}
6219				sa = (struct sockaddr *)&sin;
6220				inp = stcb->sctp_ep;
6221				atomic_add_int(&stcb->asoc.refcnt, 1);
6222				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6223				    dst, stcb);
6224				atomic_add_int(&stcb->asoc.refcnt, -1);
6225
6226				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6227				    inp == NULL) {
6228					/* we must add the source address */
6229					/*
6230					 * no scope set since we have a tcb
6231					 * already
6232					 */
6233
6234					/*
6235					 * we must validate the state again
6236					 * here
6237					 */
6238			add_it_now:
6239					if (stcb->asoc.state == 0) {
6240						/* the assoc was freed? */
6241						return (-7);
6242					}
6243					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6244						return (-8);
6245					}
6246				} else if (stcb_tmp == stcb) {
6247					if (stcb->asoc.state == 0) {
6248						/* the assoc was freed? */
6249						return (-10);
6250					}
6251					if (net != NULL) {
6252						/* clear flag */
6253						net->dest_state &=
6254						    ~SCTP_ADDR_NOT_IN_ASSOC;
6255					}
6256				} else {
6257					/*
6258					 * strange, address is in another
6259					 * assoc? straighten out locks.
6260					 */
6261					if (stcb_tmp) {
6262						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6263							struct mbuf *op_err;
6264							char msg[SCTP_DIAG_INFO_LEN];
6265
6266							/*
6267							 * in setup state we
6268							 * abort this guy
6269							 */
6270							snprintf(msg, sizeof(msg),
6271							    "%s:%d at %s", __FILE__, __LINE__, __func__);
6272							op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
6273							    msg);
6274							sctp_abort_an_association(stcb_tmp->sctp_ep,
6275							    stcb_tmp, op_err,
6276							    SCTP_SO_NOT_LOCKED);
6277							goto add_it_now;
6278						}
6279						SCTP_TCB_UNLOCK(stcb_tmp);
6280					}
6281					if (stcb->asoc.state == 0) {
6282						/* the assoc was freed? */
6283						return (-12);
6284					}
6285					return (-13);
6286				}
6287			}
6288		} else
6289#endif
6290#ifdef INET6
6291		if (ptype == SCTP_IPV6_ADDRESS) {
6292			if (stcb->asoc.scope.ipv6_addr_legal) {
6293				/* ok get the v6 address and check/add */
6294				struct sctp_ipv6addr_param *p6, p6_buf;
6295
6296				phdr = sctp_get_next_param(m, offset,
6297				    (struct sctp_paramhdr *)&p6_buf,
6298				    sizeof(p6_buf));
6299				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6300				    phdr == NULL) {
6301					return (-14);
6302				}
6303				p6 = (struct sctp_ipv6addr_param *)phdr;
6304				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6305				    sizeof(p6->addr));
6306				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6307					/* Skip multi-cast addresses */
6308					goto next_param;
6309				}
6310				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6311					/*
6312					 * Link local make no sense without
6313					 * scope
6314					 */
6315					goto next_param;
6316				}
6317				sa = (struct sockaddr *)&sin6;
6318				inp = stcb->sctp_ep;
6319				atomic_add_int(&stcb->asoc.refcnt, 1);
6320				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6321				    dst, stcb);
6322				atomic_add_int(&stcb->asoc.refcnt, -1);
6323				if (stcb_tmp == NULL &&
6324				    (inp == stcb->sctp_ep || inp == NULL)) {
6325					/*
6326					 * we must validate the state again
6327					 * here
6328					 */
6329			add_it_now6:
6330					if (stcb->asoc.state == 0) {
6331						/* the assoc was freed? */
6332						return (-16);
6333					}
6334					/*
6335					 * we must add the address, no scope
6336					 * set
6337					 */
6338					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6339						return (-17);
6340					}
6341				} else if (stcb_tmp == stcb) {
6342					/*
6343					 * we must validate the state again
6344					 * here
6345					 */
6346					if (stcb->asoc.state == 0) {
6347						/* the assoc was freed? */
6348						return (-19);
6349					}
6350					if (net != NULL) {
6351						/* clear flag */
6352						net->dest_state &=
6353						    ~SCTP_ADDR_NOT_IN_ASSOC;
6354					}
6355				} else {
6356					/*
6357					 * strange, address is in another
6358					 * assoc? straighten out locks.
6359					 */
6360					if (stcb_tmp) {
6361						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6362							struct mbuf *op_err;
6363							char msg[SCTP_DIAG_INFO_LEN];
6364
6365							/*
6366							 * in setup state we
6367							 * abort this guy
6368							 */
6369							snprintf(msg, sizeof(msg),
6370							    "%s:%d at %s", __FILE__, __LINE__, __func__);
6371							op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
6372							    msg);
6373							sctp_abort_an_association(stcb_tmp->sctp_ep,
6374							    stcb_tmp, op_err,
6375							    SCTP_SO_NOT_LOCKED);
6376							goto add_it_now6;
6377						}
6378						SCTP_TCB_UNLOCK(stcb_tmp);
6379					}
6380					if (stcb->asoc.state == 0) {
6381						/* the assoc was freed? */
6382						return (-21);
6383					}
6384					return (-22);
6385				}
6386			}
6387		} else
6388#endif
6389		if (ptype == SCTP_ECN_CAPABLE) {
6390			peer_supports_ecn = 1;
6391		} else if (ptype == SCTP_ULP_ADAPTATION) {
6392			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6393				struct sctp_adaptation_layer_indication ai,
6394				                                *aip;
6395
6396				phdr = sctp_get_next_param(m, offset,
6397				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6398				aip = (struct sctp_adaptation_layer_indication *)phdr;
6399				if (aip) {
6400					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6401					stcb->asoc.adaptation_needed = 1;
6402				}
6403			}
6404		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6405			struct sctp_asconf_addr_param lstore, *fee;
6406			int lptype;
6407			struct sockaddr *lsa = NULL;
6408
6409#ifdef INET
6410			struct sctp_asconf_addrv4_param *fii;
6411
6412#endif
6413
6414			if (stcb->asoc.asconf_supported == 0) {
6415				return (-100);
6416			}
6417			if (plen > sizeof(lstore)) {
6418				return (-23);
6419			}
6420			phdr = sctp_get_next_param(m, offset,
6421			    (struct sctp_paramhdr *)&lstore,
6422			    min(plen, sizeof(lstore)));
6423			if (phdr == NULL) {
6424				return (-24);
6425			}
6426			fee = (struct sctp_asconf_addr_param *)phdr;
6427			lptype = ntohs(fee->addrp.ph.param_type);
6428			switch (lptype) {
6429#ifdef INET
6430			case SCTP_IPV4_ADDRESS:
6431				if (plen !=
6432				    sizeof(struct sctp_asconf_addrv4_param)) {
6433					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6434					    (int)sizeof(struct sctp_asconf_addrv4_param),
6435					    plen);
6436				} else {
6437					fii = (struct sctp_asconf_addrv4_param *)fee;
6438					sin.sin_addr.s_addr = fii->addrp.addr;
6439					lsa = (struct sockaddr *)&sin;
6440				}
6441				break;
6442#endif
6443#ifdef INET6
6444			case SCTP_IPV6_ADDRESS:
6445				if (plen !=
6446				    sizeof(struct sctp_asconf_addr_param)) {
6447					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6448					    (int)sizeof(struct sctp_asconf_addr_param),
6449					    plen);
6450				} else {
6451					memcpy(sin6.sin6_addr.s6_addr,
6452					    fee->addrp.addr,
6453					    sizeof(fee->addrp.addr));
6454					lsa = (struct sockaddr *)&sin6;
6455				}
6456				break;
6457#endif
6458			default:
6459				break;
6460			}
6461			if (lsa) {
6462				(void)sctp_set_primary_addr(stcb, sa, NULL);
6463			}
6464		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6465			stcb->asoc.peer_supports_nat = 1;
6466		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6467			/* Peer supports pr-sctp */
6468			peer_supports_prsctp = 1;
6469		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6470			/* A supported extension chunk */
6471			struct sctp_supported_chunk_types_param *pr_supported;
6472			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6473			int num_ent, i;
6474
6475			phdr = sctp_get_next_param(m, offset,
6476			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6477			if (phdr == NULL) {
6478				return (-25);
6479			}
6480			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6481			num_ent = plen - sizeof(struct sctp_paramhdr);
6482			for (i = 0; i < num_ent; i++) {
6483				switch (pr_supported->chunk_types[i]) {
6484				case SCTP_ASCONF:
6485					peer_supports_asconf = 1;
6486					break;
6487				case SCTP_ASCONF_ACK:
6488					peer_supports_asconf_ack = 1;
6489					break;
6490				case SCTP_FORWARD_CUM_TSN:
6491					peer_supports_prsctp = 1;
6492					break;
6493				case SCTP_PACKET_DROPPED:
6494					peer_supports_pktdrop = 1;
6495					break;
6496				case SCTP_NR_SELECTIVE_ACK:
6497					peer_supports_nrsack = 1;
6498					break;
6499				case SCTP_STREAM_RESET:
6500					peer_supports_reconfig = 1;
6501					break;
6502				case SCTP_AUTHENTICATION:
6503					peer_supports_auth = 1;
6504					break;
6505				default:
6506					/* one I have not learned yet */
6507					break;
6508
6509				}
6510			}
6511		} else if (ptype == SCTP_RANDOM) {
6512			if (plen > sizeof(random_store))
6513				break;
6514			if (got_random) {
6515				/* already processed a RANDOM */
6516				goto next_param;
6517			}
6518			phdr = sctp_get_next_param(m, offset,
6519			    (struct sctp_paramhdr *)random_store,
6520			    min(sizeof(random_store), plen));
6521			if (phdr == NULL)
6522				return (-26);
6523			p_random = (struct sctp_auth_random *)phdr;
6524			random_len = plen - sizeof(*p_random);
6525			/* enforce the random length */
6526			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6527				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6528				return (-27);
6529			}
6530			got_random = 1;
6531		} else if (ptype == SCTP_HMAC_LIST) {
6532			uint16_t num_hmacs;
6533			uint16_t i;
6534
6535			if (plen > sizeof(hmacs_store))
6536				break;
6537			if (got_hmacs) {
6538				/* already processed a HMAC list */
6539				goto next_param;
6540			}
6541			phdr = sctp_get_next_param(m, offset,
6542			    (struct sctp_paramhdr *)hmacs_store,
6543			    min(plen, sizeof(hmacs_store)));
6544			if (phdr == NULL)
6545				return (-28);
6546			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6547			hmacs_len = plen - sizeof(*hmacs);
6548			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6549			/* validate the hmac list */
6550			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6551				return (-29);
6552			}
6553			if (stcb->asoc.peer_hmacs != NULL)
6554				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6555			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6556			if (stcb->asoc.peer_hmacs != NULL) {
6557				for (i = 0; i < num_hmacs; i++) {
6558					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6559					    ntohs(hmacs->hmac_ids[i]));
6560				}
6561			}
6562			got_hmacs = 1;
6563		} else if (ptype == SCTP_CHUNK_LIST) {
6564			int i;
6565
6566			if (plen > sizeof(chunks_store))
6567				break;
6568			if (got_chklist) {
6569				/* already processed a Chunks list */
6570				goto next_param;
6571			}
6572			phdr = sctp_get_next_param(m, offset,
6573			    (struct sctp_paramhdr *)chunks_store,
6574			    min(plen, sizeof(chunks_store)));
6575			if (phdr == NULL)
6576				return (-30);
6577			chunks = (struct sctp_auth_chunk_list *)phdr;
6578			num_chunks = plen - sizeof(*chunks);
6579			if (stcb->asoc.peer_auth_chunks != NULL)
6580				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6581			else
6582				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6583			for (i = 0; i < num_chunks; i++) {
6584				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6585				    stcb->asoc.peer_auth_chunks);
6586				/* record asconf/asconf-ack if listed */
6587				if (chunks->chunk_types[i] == SCTP_ASCONF)
6588					saw_asconf = 1;
6589				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6590					saw_asconf_ack = 1;
6591
6592			}
6593			got_chklist = 1;
6594		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6595			    (ptype == SCTP_STATE_COOKIE) ||
6596			    (ptype == SCTP_UNRECOG_PARAM) ||
6597			    (ptype == SCTP_COOKIE_PRESERVE) ||
6598			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6599			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6600			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6601			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6602		    (ptype == SCTP_SUCCESS_REPORT)) {
6603			 /* don't care */ ;
6604		} else {
6605			if ((ptype & 0x8000) == 0x0000) {
6606				/*
6607				 * must stop processing the rest of the
6608				 * param's. Any report bits were handled
6609				 * with the call to
6610				 * sctp_arethere_unrecognized_parameters()
6611				 * when the INIT or INIT-ACK was first seen.
6612				 */
6613				break;
6614			}
6615		}
6616
6617next_param:
6618		offset += SCTP_SIZE32(plen);
6619		if (offset >= limit) {
6620			break;
6621		}
6622		phdr = sctp_get_next_param(m, offset, &parm_buf,
6623		    sizeof(parm_buf));
6624	}
6625	/* Now check to see if we need to purge any addresses */
6626	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6627		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6628		    SCTP_ADDR_NOT_IN_ASSOC) {
6629			/* This address has been removed from the asoc */
6630			/* remove and free it */
6631			stcb->asoc.numnets--;
6632			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6633			sctp_free_remote_addr(net);
6634			if (net == stcb->asoc.primary_destination) {
6635				stcb->asoc.primary_destination = NULL;
6636				sctp_select_primary_destination(stcb);
6637			}
6638		}
6639	}
6640	if ((stcb->asoc.ecn_supported == 1) &&
6641	    (peer_supports_ecn == 0)) {
6642		stcb->asoc.ecn_supported = 0;
6643	}
6644	if ((stcb->asoc.prsctp_supported == 1) &&
6645	    (peer_supports_prsctp == 0)) {
6646		stcb->asoc.prsctp_supported = 0;
6647	}
6648	if ((stcb->asoc.auth_supported == 1) &&
6649	    ((peer_supports_auth == 0) ||
6650	    (got_random == 0) || (got_hmacs == 0))) {
6651		stcb->asoc.auth_supported = 0;
6652	}
6653	if ((stcb->asoc.asconf_supported == 1) &&
6654	    ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) ||
6655	    (stcb->asoc.auth_supported == 0) ||
6656	    (saw_asconf == 0) || (saw_asconf_ack == 0))) {
6657		stcb->asoc.asconf_supported = 0;
6658	}
6659	if ((stcb->asoc.reconfig_supported == 1) &&
6660	    (peer_supports_reconfig == 0)) {
6661		stcb->asoc.reconfig_supported = 0;
6662	}
6663	if ((stcb->asoc.nrsack_supported == 1) &&
6664	    (peer_supports_nrsack == 0)) {
6665		stcb->asoc.nrsack_supported = 0;
6666	}
6667	if ((stcb->asoc.pktdrop_supported == 1) &&
6668	    (peer_supports_pktdrop == 0)) {
6669		stcb->asoc.pktdrop_supported = 0;
6670	}
6671	/* validate authentication required parameters */
6672	if ((peer_supports_auth == 0) && (got_chklist == 1)) {
6673		/* peer does not support auth but sent a chunks list? */
6674		return (-31);
6675	}
6676	if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) {
6677		/* peer supports asconf but not auth? */
6678		return (-32);
6679	} else if ((peer_supports_asconf == 1) &&
6680		    (peer_supports_auth == 1) &&
6681	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6682		return (-33);
6683	}
6684	/* concatenate the full random key */
6685	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6686	if (chunks != NULL) {
6687		keylen += sizeof(*chunks) + num_chunks;
6688	}
6689	new_key = sctp_alloc_key(keylen);
6690	if (new_key != NULL) {
6691		/* copy in the RANDOM */
6692		if (p_random != NULL) {
6693			keylen = sizeof(*p_random) + random_len;
6694			bcopy(p_random, new_key->key, keylen);
6695		}
6696		/* append in the AUTH chunks */
6697		if (chunks != NULL) {
6698			bcopy(chunks, new_key->key + keylen,
6699			    sizeof(*chunks) + num_chunks);
6700			keylen += sizeof(*chunks) + num_chunks;
6701		}
6702		/* append in the HMACs */
6703		if (hmacs != NULL) {
6704			bcopy(hmacs, new_key->key + keylen,
6705			    sizeof(*hmacs) + hmacs_len);
6706		}
6707	} else {
6708		/* failed to get memory for the key */
6709		return (-34);
6710	}
6711	if (stcb->asoc.authinfo.peer_random != NULL)
6712		sctp_free_key(stcb->asoc.authinfo.peer_random);
6713	stcb->asoc.authinfo.peer_random = new_key;
6714	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6715	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6716
6717	return (0);
6718}
6719
6720int
6721sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6722    struct sctp_nets *net)
6723{
6724	/* make sure the requested primary address exists in the assoc */
6725	if (net == NULL && sa)
6726		net = sctp_findnet(stcb, sa);
6727
6728	if (net == NULL) {
6729		/* didn't find the requested primary address! */
6730		return (-1);
6731	} else {
6732		/* set the primary address */
6733		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6734			/* Must be confirmed, so queue to set */
6735			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6736			return (0);
6737		}
6738		stcb->asoc.primary_destination = net;
6739		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6740			sctp_free_remote_addr(stcb->asoc.alternate);
6741			stcb->asoc.alternate = NULL;
6742		}
6743		net = TAILQ_FIRST(&stcb->asoc.nets);
6744		if (net != stcb->asoc.primary_destination) {
6745			/*
6746			 * first one on the list is NOT the primary
6747			 * sctp_cmpaddr() is much more efficient if the
6748			 * primary is the first on the list, make it so.
6749			 */
6750			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6751			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6752		}
6753		return (0);
6754	}
6755}
6756
6757int
6758sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6759{
6760	/*
6761	 * This function serves two purposes. It will see if a TAG can be
6762	 * re-used and return 1 for yes it is ok and 0 for don't use that
6763	 * tag. A secondary function it will do is purge out old tags that
6764	 * can be removed.
6765	 */
6766	struct sctpvtaghead *chain;
6767	struct sctp_tagblock *twait_block;
6768	struct sctpasochead *head;
6769	struct sctp_tcb *stcb;
6770	int i;
6771
6772	SCTP_INP_INFO_RLOCK();
6773	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6774	    SCTP_BASE_INFO(hashasocmark))];
6775	LIST_FOREACH(stcb, head, sctp_asocs) {
6776		/*
6777		 * We choose not to lock anything here. TCB's can't be
6778		 * removed since we have the read lock, so they can't be
6779		 * freed on us, same thing for the INP. I may be wrong with
6780		 * this assumption, but we will go with it for now :-)
6781		 */
6782		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6783			continue;
6784		}
6785		if (stcb->asoc.my_vtag == tag) {
6786			/* candidate */
6787			if (stcb->rport != rport) {
6788				continue;
6789			}
6790			if (stcb->sctp_ep->sctp_lport != lport) {
6791				continue;
6792			}
6793			/* Its a used tag set */
6794			SCTP_INP_INFO_RUNLOCK();
6795			return (0);
6796		}
6797	}
6798	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6799	/* Now what about timed wait ? */
6800	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6801		/*
6802		 * Block(s) are present, lets see if we have this tag in the
6803		 * list
6804		 */
6805		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6806			if (twait_block->vtag_block[i].v_tag == 0) {
6807				/* not used */
6808				continue;
6809			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6810			    now->tv_sec) {
6811				/* Audit expires this guy */
6812				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6813				twait_block->vtag_block[i].v_tag = 0;
6814				twait_block->vtag_block[i].lport = 0;
6815				twait_block->vtag_block[i].rport = 0;
6816			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6817				    (twait_block->vtag_block[i].lport == lport) &&
6818			    (twait_block->vtag_block[i].rport == rport)) {
6819				/* Bad tag, sorry :< */
6820				SCTP_INP_INFO_RUNLOCK();
6821				return (0);
6822			}
6823		}
6824	}
6825	SCTP_INP_INFO_RUNLOCK();
6826	return (1);
6827}
6828
6829static void
6830sctp_drain_mbufs(struct sctp_tcb *stcb)
6831{
6832	/*
6833	 * We must hunt this association for MBUF's past the cumack (i.e.
6834	 * out of order data that we can renege on).
6835	 */
6836	struct sctp_association *asoc;
6837	struct sctp_tmit_chunk *chk, *nchk;
6838	uint32_t cumulative_tsn_p1;
6839	struct sctp_queued_to_read *ctl, *nctl;
6840	int cnt, strmat;
6841	uint32_t gap, i;
6842	int fnd = 0;
6843
6844	/* We look for anything larger than the cum-ack + 1 */
6845
6846	asoc = &stcb->asoc;
6847	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6848		/* none we can reneg on. */
6849		return;
6850	}
6851	SCTP_STAT_INCR(sctps_protocol_drains_done);
6852	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6853	cnt = 0;
6854	/* First look in the re-assembly queue */
6855	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6856		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6857			/* Yep it is above cum-ack */
6858			cnt++;
6859			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6860			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6861			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6862			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6863			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6864			if (chk->data) {
6865				sctp_m_freem(chk->data);
6866				chk->data = NULL;
6867			}
6868			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6869		}
6870	}
6871	/* Ok that was fun, now we will drain all the inbound streams? */
6872	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6873		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6874			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6875				/* Yep it is above cum-ack */
6876				cnt++;
6877				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6878				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6879				sctp_ucount_decr(asoc->cnt_on_all_streams);
6880				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6881				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6882				if (ctl->data) {
6883					sctp_m_freem(ctl->data);
6884					ctl->data = NULL;
6885				}
6886				sctp_free_remote_addr(ctl->whoFrom);
6887				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6888				SCTP_DECR_READQ_COUNT();
6889			}
6890		}
6891	}
6892	if (cnt) {
6893		/* We must back down to see what the new highest is */
6894		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6895			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6896			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6897				asoc->highest_tsn_inside_map = i;
6898				fnd = 1;
6899				break;
6900			}
6901		}
6902		if (!fnd) {
6903			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6904		}
6905		/*
6906		 * Question, should we go through the delivery queue? The
6907		 * only reason things are on here is the app not reading OR
6908		 * a p-d-api up. An attacker COULD send enough in to
6909		 * initiate the PD-API and then send a bunch of stuff to
6910		 * other streams... these would wind up on the delivery
6911		 * queue.. and then we would not get to them. But in order
6912		 * to do this I then have to back-track and un-deliver
6913		 * sequence numbers in streams.. el-yucko. I think for now
6914		 * we will NOT look at the delivery queue and leave it to be
6915		 * something to consider later. An alternative would be to
6916		 * abort the P-D-API with a notification and then deliver
6917		 * the data.... Or another method might be to keep track of
6918		 * how many times the situation occurs and if we see a
6919		 * possible attack underway just abort the association.
6920		 */
6921#ifdef SCTP_DEBUG
6922		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6923#endif
6924		/*
6925		 * Now do we need to find a new
6926		 * asoc->highest_tsn_inside_map?
6927		 */
6928		asoc->last_revoke_count = cnt;
6929		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6930		/* sa_ignore NO_NULL_CHK */
6931		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6932		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6933	}
6934	/*
6935	 * Another issue, in un-setting the TSN's in the mapping array we
6936	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6937	 * two things to occur. It may cause us to do extra work in checking
6938	 * for our mapping array movement. More importantly it may cause us
6939	 * to SACK every datagram. This may not be a bad thing though since
6940	 * we will recover once we get our cum-ack above and all this stuff
6941	 * we dumped recovered.
6942	 */
6943}
6944
6945void
6946sctp_drain()
6947{
6948	/*
6949	 * We must walk the PCB lists for ALL associations here. The system
6950	 * is LOW on MBUF's and needs help. This is where reneging will
6951	 * occur. We really hope this does NOT happen!
6952	 */
6953	VNET_ITERATOR_DECL(vnet_iter);
6954	VNET_LIST_RLOCK_NOSLEEP();
6955	VNET_FOREACH(vnet_iter) {
6956		CURVNET_SET(vnet_iter);
6957		struct sctp_inpcb *inp;
6958		struct sctp_tcb *stcb;
6959
6960		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6961		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6962#ifdef VIMAGE
6963			continue;
6964#else
6965			return;
6966#endif
6967		}
6968		SCTP_INP_INFO_RLOCK();
6969		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6970			/* For each endpoint */
6971			SCTP_INP_RLOCK(inp);
6972			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6973				/* For each association */
6974				SCTP_TCB_LOCK(stcb);
6975				sctp_drain_mbufs(stcb);
6976				SCTP_TCB_UNLOCK(stcb);
6977			}
6978			SCTP_INP_RUNLOCK(inp);
6979		}
6980		SCTP_INP_INFO_RUNLOCK();
6981		CURVNET_RESTORE();
6982	}
6983	VNET_LIST_RUNLOCK_NOSLEEP();
6984}
6985
6986/*
6987 * start a new iterator
6988 * iterates through all endpoints and associations based on the pcb_state
6989 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6990 * assocs and "ef" (optional) is executed when the iterator completes.
6991 * "inpf" (optional) is executed for each new endpoint as it is being
6992 * iterated through. inpe (optional) is called when the inp completes
6993 * its way through all the stcbs.
6994 */
6995int
6996sctp_initiate_iterator(inp_func inpf,
6997    asoc_func af,
6998    inp_func inpe,
6999    uint32_t pcb_state,
7000    uint32_t pcb_features,
7001    uint32_t asoc_state,
7002    void *argp,
7003    uint32_t argi,
7004    end_func ef,
7005    struct sctp_inpcb *s_inp,
7006    uint8_t chunk_output_off)
7007{
7008	struct sctp_iterator *it = NULL;
7009
7010	if (af == NULL) {
7011		return (-1);
7012	}
7013	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
7014		SCTP_PRINTF("%s: abort on initialize being %d\n", __func__,
7015		    SCTP_BASE_VAR(sctp_pcb_initialized));
7016		return (-1);
7017	}
7018	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
7019	    SCTP_M_ITER);
7020	if (it == NULL) {
7021		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
7022		return (ENOMEM);
7023	}
7024	memset(it, 0, sizeof(*it));
7025	it->function_assoc = af;
7026	it->function_inp = inpf;
7027	if (inpf)
7028		it->done_current_ep = 0;
7029	else
7030		it->done_current_ep = 1;
7031	it->function_atend = ef;
7032	it->pointer = argp;
7033	it->val = argi;
7034	it->pcb_flags = pcb_state;
7035	it->pcb_features = pcb_features;
7036	it->asoc_state = asoc_state;
7037	it->function_inp_end = inpe;
7038	it->no_chunk_output = chunk_output_off;
7039	it->vn = curvnet;
7040	if (s_inp) {
7041		/* Assume lock is held here */
7042		it->inp = s_inp;
7043		SCTP_INP_INCR_REF(it->inp);
7044		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7045	} else {
7046		SCTP_INP_INFO_RLOCK();
7047		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7048		if (it->inp) {
7049			SCTP_INP_INCR_REF(it->inp);
7050		}
7051		SCTP_INP_INFO_RUNLOCK();
7052		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7053
7054	}
7055	SCTP_IPI_ITERATOR_WQ_LOCK();
7056	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
7057		SCTP_IPI_ITERATOR_WQ_UNLOCK();
7058		SCTP_PRINTF("%s: rollback on initialize being %d it=%p\n", __func__,
7059		    SCTP_BASE_VAR(sctp_pcb_initialized), it);
7060		SCTP_FREE(it, SCTP_M_ITER);
7061		return (-1);
7062	}
7063	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7064	if (sctp_it_ctl.iterator_running == 0) {
7065		sctp_wakeup_iterator();
7066	}
7067	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7068	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7069	return (0);
7070}
7071