sctp_pcb.c revision 267725
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 267725 2014-06-22 16:32:41Z tuexen $");
35
36#include <netinet/sctp_os.h>
37#include <sys/proc.h>
38#include <netinet/sctp_var.h>
39#include <netinet/sctp_sysctl.h>
40#include <netinet/sctp_pcb.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp.h>
43#include <netinet/sctp_header.h>
44#include <netinet/sctp_asconf.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_timer.h>
47#include <netinet/sctp_bsd_addr.h>
48#include <netinet/sctp_dtrace_define.h>
49#include <netinet/udp.h>
50#ifdef INET6
51#include <netinet6/ip6_var.h>
52#endif
53#include <sys/sched.h>
54#include <sys/smp.h>
55#include <sys/unistd.h>
56
57
58VNET_DEFINE(struct sctp_base_info, system_base_info);
59
60/* FIX: we don't handle multiple link local scopes */
61/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
62#ifdef INET6
63int
64SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
65{
66	struct sockaddr_in6 tmp_a, tmp_b;
67
68	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
69	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
70		return (0);
71	}
72	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
73	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
74		return (0);
75	}
76	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
77}
78
79#endif
80
81void
82sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
83{
84	/*
85	 * We really don't need to lock this, but I will just because it
86	 * does not hurt.
87	 */
88	SCTP_INP_INFO_RLOCK();
89	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
90	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
91	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
92	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
93	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
94	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
95	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
96	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
97	SCTP_INP_INFO_RUNLOCK();
98}
99
100/*-
101 * Addresses are added to VRF's (Virtual Router's). For BSD we
102 * have only the default VRF 0. We maintain a hash list of
103 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
104 * these has a list of addresses. When we add a new address
105 * to a VRF we lookup the ifn/ifn_index, if the ifn does
106 * not exist we create it and add it to the list of IFN's
107 * within the VRF. Once we have the sctp_ifn, we add the
108 * address to the list. So we look something like:
109 *
110 * hash-vrf-table
111 *   vrf-> ifn-> ifn -> ifn
112 *   vrf    |
113 *    ...   +--ifa-> ifa -> ifa
114 *   vrf
115 *
116 * We keep these separate lists since the SCTP subsystem will
117 * point to these from its source address selection nets structure.
118 * When an address is deleted it does not happen right away on
119 * the SCTP side, it gets scheduled. What we do when a
120 * delete happens is immediately remove the address from
121 * the master list and decrement the refcount. As our
122 * addip iterator works through and frees the src address
123 * selection pointing to the sctp_ifa, eventually the refcount
124 * will reach 0 and we will delete it. Note that it is assumed
125 * that any locking on system level ifn/ifa is done at the
126 * caller of these functions and these routines will only
127 * lock the SCTP structures as they add or delete things.
128 *
129 * Other notes on VRF concepts.
130 *  - An endpoint can be in multiple VRF's
131 *  - An association lives within a VRF and only one VRF.
132 *  - Any incoming packet we can deduce the VRF for by
133 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
134 *  - Any downward send call or connect call must supply the
135 *    VRF via ancillary data or via some sort of set default
136 *    VRF socket option call (again for BSD no brainer since
137 *    the VRF is always 0).
138 *  - An endpoint may add multiple VRF's to it.
139 *  - Listening sockets can accept associations in any
140 *    of the VRF's they are in but the assoc will end up
141 *    in only one VRF (gotten from the packet or connect/send).
142 *
143 */
144
145struct sctp_vrf *
146sctp_allocate_vrf(int vrf_id)
147{
148	struct sctp_vrf *vrf = NULL;
149	struct sctp_vrflist *bucket;
150
151	/* First allocate the VRF structure */
152	vrf = sctp_find_vrf(vrf_id);
153	if (vrf) {
154		/* Already allocated */
155		return (vrf);
156	}
157	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
158	    SCTP_M_VRF);
159	if (vrf == NULL) {
160		/* No memory */
161#ifdef INVARIANTS
162		panic("No memory for VRF:%d", vrf_id);
163#endif
164		return (NULL);
165	}
166	/* setup the VRF */
167	memset(vrf, 0, sizeof(struct sctp_vrf));
168	vrf->vrf_id = vrf_id;
169	LIST_INIT(&vrf->ifnlist);
170	vrf->total_ifa_count = 0;
171	vrf->refcount = 0;
172	/* now also setup table ids */
173	SCTP_INIT_VRF_TABLEID(vrf);
174	/* Init the HASH of addresses */
175	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
176	    &vrf->vrf_addr_hashmark);
177	if (vrf->vrf_addr_hash == NULL) {
178		/* No memory */
179#ifdef INVARIANTS
180		panic("No memory for VRF:%d", vrf_id);
181#endif
182		SCTP_FREE(vrf, SCTP_M_VRF);
183		return (NULL);
184	}
185	/* Add it to the hash table */
186	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
187	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
188	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
189	return (vrf);
190}
191
192
193struct sctp_ifn *
194sctp_find_ifn(void *ifn, uint32_t ifn_index)
195{
196	struct sctp_ifn *sctp_ifnp;
197	struct sctp_ifnlist *hash_ifn_head;
198
199	/*
200	 * We assume the lock is held for the addresses if that's wrong
201	 * problems could occur :-)
202	 */
203	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
204	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
205		if (sctp_ifnp->ifn_index == ifn_index) {
206			return (sctp_ifnp);
207		}
208		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
209			return (sctp_ifnp);
210		}
211	}
212	return (NULL);
213}
214
215
216struct sctp_vrf *
217sctp_find_vrf(uint32_t vrf_id)
218{
219	struct sctp_vrflist *bucket;
220	struct sctp_vrf *liste;
221
222	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
223	LIST_FOREACH(liste, bucket, next_vrf) {
224		if (vrf_id == liste->vrf_id) {
225			return (liste);
226		}
227	}
228	return (NULL);
229}
230
231
232void
233sctp_free_vrf(struct sctp_vrf *vrf)
234{
235	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
236		if (vrf->vrf_addr_hash) {
237			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
238			vrf->vrf_addr_hash = NULL;
239		}
240		/* We zero'd the count */
241		LIST_REMOVE(vrf, next_vrf);
242		SCTP_FREE(vrf, SCTP_M_VRF);
243		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
244	}
245}
246
247
248void
249sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
250{
251	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
252		/* We zero'd the count */
253		if (sctp_ifnp->vrf) {
254			sctp_free_vrf(sctp_ifnp->vrf);
255		}
256		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
257		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
258	}
259}
260
261
262void
263sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
264{
265	struct sctp_ifn *sctp_ifnp;
266
267	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
268	if (sctp_ifnp != NULL) {
269		sctp_ifnp->ifn_mtu = mtu;
270	}
271}
272
273
274void
275sctp_free_ifa(struct sctp_ifa *sctp_ifap)
276{
277	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
278		/* We zero'd the count */
279		if (sctp_ifap->ifn_p) {
280			sctp_free_ifn(sctp_ifap->ifn_p);
281		}
282		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
283		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
284	}
285}
286
287
288static void
289sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
290{
291	struct sctp_ifn *found;
292
293	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
294	if (found == NULL) {
295		/* Not in the list.. sorry */
296		return;
297	}
298	if (hold_addr_lock == 0)
299		SCTP_IPI_ADDR_WLOCK();
300	LIST_REMOVE(sctp_ifnp, next_bucket);
301	LIST_REMOVE(sctp_ifnp, next_ifn);
302	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
303	    sctp_ifnp->registered_af);
304	if (hold_addr_lock == 0)
305		SCTP_IPI_ADDR_WUNLOCK();
306	/* Take away the reference, and possibly free it */
307	sctp_free_ifn(sctp_ifnp);
308}
309
310
311void
312sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
313    const char *if_name, uint32_t ifn_index)
314{
315	struct sctp_vrf *vrf;
316	struct sctp_ifa *sctp_ifap;
317
318	SCTP_IPI_ADDR_RLOCK();
319	vrf = sctp_find_vrf(vrf_id);
320	if (vrf == NULL) {
321		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
322		goto out;
323
324	}
325	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
326	if (sctp_ifap == NULL) {
327		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
328		goto out;
329	}
330	if (sctp_ifap->ifn_p == NULL) {
331		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
332		goto out;
333	}
334	if (if_name) {
335		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
336			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
337			    sctp_ifap->ifn_p->ifn_name, if_name);
338			goto out;
339		}
340	} else {
341		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
342			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
343			    sctp_ifap->ifn_p->ifn_index, ifn_index);
344			goto out;
345		}
346	}
347
348	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
349	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
350out:
351	SCTP_IPI_ADDR_RUNLOCK();
352}
353
354
355void
356sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
357    const char *if_name, uint32_t ifn_index)
358{
359	struct sctp_vrf *vrf;
360	struct sctp_ifa *sctp_ifap;
361
362	SCTP_IPI_ADDR_RLOCK();
363	vrf = sctp_find_vrf(vrf_id);
364	if (vrf == NULL) {
365		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
366		goto out;
367
368	}
369	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
370	if (sctp_ifap == NULL) {
371		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
372		goto out;
373	}
374	if (sctp_ifap->ifn_p == NULL) {
375		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
376		goto out;
377	}
378	if (if_name) {
379		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
380			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
381			    sctp_ifap->ifn_p->ifn_name, if_name);
382			goto out;
383		}
384	} else {
385		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
386			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
387			    sctp_ifap->ifn_p->ifn_index, ifn_index);
388			goto out;
389		}
390	}
391
392	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
393	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
394out:
395	SCTP_IPI_ADDR_RUNLOCK();
396}
397
398
399/*-
400 * Add an ifa to an ifn.
401 * Register the interface as necessary.
402 * NOTE: ADDR write lock MUST be held.
403 */
404static void
405sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
406{
407	int ifa_af;
408
409	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
410	sctp_ifap->ifn_p = sctp_ifnp;
411	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
412	/* update address counts */
413	sctp_ifnp->ifa_count++;
414	ifa_af = sctp_ifap->address.sa.sa_family;
415	switch (ifa_af) {
416#ifdef INET
417	case AF_INET:
418		sctp_ifnp->num_v4++;
419		break;
420#endif
421#ifdef INET6
422	case AF_INET6:
423		sctp_ifnp->num_v6++;
424		break;
425#endif
426	default:
427		break;
428	}
429	if (sctp_ifnp->ifa_count == 1) {
430		/* register the new interface */
431		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
432		sctp_ifnp->registered_af = ifa_af;
433	}
434}
435
436
437/*-
438 * Remove an ifa from its ifn.
439 * If no more addresses exist, remove the ifn too. Otherwise, re-register
440 * the interface based on the remaining address families left.
441 * NOTE: ADDR write lock MUST be held.
442 */
443static void
444sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
445{
446	LIST_REMOVE(sctp_ifap, next_ifa);
447	if (sctp_ifap->ifn_p) {
448		/* update address counts */
449		sctp_ifap->ifn_p->ifa_count--;
450		switch (sctp_ifap->address.sa.sa_family) {
451#ifdef INET
452		case AF_INET:
453			sctp_ifap->ifn_p->num_v4--;
454			break;
455#endif
456#ifdef INET6
457		case AF_INET6:
458			sctp_ifap->ifn_p->num_v6--;
459			break;
460#endif
461		default:
462			break;
463		}
464
465		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
466			/* remove the ifn, possibly freeing it */
467			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
468		} else {
469			/* re-register address family type, if needed */
470			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
471			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
472				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
473				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
474				sctp_ifap->ifn_p->registered_af = AF_INET;
475			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
476			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
477				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
478				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
479				sctp_ifap->ifn_p->registered_af = AF_INET6;
480			}
481			/* free the ifn refcount */
482			sctp_free_ifn(sctp_ifap->ifn_p);
483		}
484		sctp_ifap->ifn_p = NULL;
485	}
486}
487
488
489struct sctp_ifa *
490sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
491    uint32_t ifn_type, const char *if_name, void *ifa,
492    struct sockaddr *addr, uint32_t ifa_flags,
493    int dynamic_add)
494{
495	struct sctp_vrf *vrf;
496	struct sctp_ifn *sctp_ifnp = NULL;
497	struct sctp_ifa *sctp_ifap = NULL;
498	struct sctp_ifalist *hash_addr_head;
499	struct sctp_ifnlist *hash_ifn_head;
500	uint32_t hash_of_addr;
501	int new_ifn_af = 0;
502
503#ifdef SCTP_DEBUG
504	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
505	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
506#endif
507	SCTP_IPI_ADDR_WLOCK();
508	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
509	if (sctp_ifnp) {
510		vrf = sctp_ifnp->vrf;
511	} else {
512		vrf = sctp_find_vrf(vrf_id);
513		if (vrf == NULL) {
514			vrf = sctp_allocate_vrf(vrf_id);
515			if (vrf == NULL) {
516				SCTP_IPI_ADDR_WUNLOCK();
517				return (NULL);
518			}
519		}
520	}
521	if (sctp_ifnp == NULL) {
522		/*
523		 * build one and add it, can't hold lock until after malloc
524		 * done though.
525		 */
526		SCTP_IPI_ADDR_WUNLOCK();
527		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
528		    sizeof(struct sctp_ifn), SCTP_M_IFN);
529		if (sctp_ifnp == NULL) {
530#ifdef INVARIANTS
531			panic("No memory for IFN");
532#endif
533			return (NULL);
534		}
535		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
536		sctp_ifnp->ifn_index = ifn_index;
537		sctp_ifnp->ifn_p = ifn;
538		sctp_ifnp->ifn_type = ifn_type;
539		sctp_ifnp->refcount = 0;
540		sctp_ifnp->vrf = vrf;
541		atomic_add_int(&vrf->refcount, 1);
542		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
543		if (if_name != NULL) {
544			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
545		} else {
546			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
547		}
548		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
549		LIST_INIT(&sctp_ifnp->ifalist);
550		SCTP_IPI_ADDR_WLOCK();
551		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
552		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
553		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
554		new_ifn_af = 1;
555	}
556	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
557	if (sctp_ifap) {
558		/* Hmm, it already exists? */
559		if ((sctp_ifap->ifn_p) &&
560		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
561			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
562			    sctp_ifap->ifn_p->ifn_name, ifn_index,
563			    (void *)sctp_ifap);
564			if (new_ifn_af) {
565				/* Remove the created one that we don't want */
566				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
567			}
568			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
569				/* easy to solve, just switch back to active */
570				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
571				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
572				sctp_ifap->ifn_p = sctp_ifnp;
573				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
574			}
575	exit_stage_left:
576			SCTP_IPI_ADDR_WUNLOCK();
577			return (sctp_ifap);
578		} else {
579			if (sctp_ifap->ifn_p) {
580				/*
581				 * The last IFN gets the address, remove the
582				 * old one
583				 */
584				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
585				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
586				    sctp_ifap->ifn_p->ifn_index, if_name,
587				    ifn_index);
588				/* remove the address from the old ifn */
589				sctp_remove_ifa_from_ifn(sctp_ifap);
590				/* move the address over to the new ifn */
591				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
592				goto exit_stage_left;
593			} else {
594				/* repair ifnp which was NULL ? */
595				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
596				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
597				    (void *)sctp_ifnp, (void *)sctp_ifap);
598				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
599			}
600			goto exit_stage_left;
601		}
602	}
603	SCTP_IPI_ADDR_WUNLOCK();
604	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
605	if (sctp_ifap == NULL) {
606#ifdef INVARIANTS
607		panic("No memory for IFA");
608#endif
609		return (NULL);
610	}
611	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
612	sctp_ifap->ifn_p = sctp_ifnp;
613	atomic_add_int(&sctp_ifnp->refcount, 1);
614	sctp_ifap->vrf_id = vrf_id;
615	sctp_ifap->ifa = ifa;
616	memcpy(&sctp_ifap->address, addr, addr->sa_len);
617	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
618	sctp_ifap->flags = ifa_flags;
619	/* Set scope */
620	switch (sctp_ifap->address.sa.sa_family) {
621#ifdef INET
622	case AF_INET:
623		{
624			struct sockaddr_in *sin;
625
626			sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
627			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
628			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
629				sctp_ifap->src_is_loop = 1;
630			}
631			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
632				sctp_ifap->src_is_priv = 1;
633			}
634			sctp_ifnp->num_v4++;
635			if (new_ifn_af)
636				new_ifn_af = AF_INET;
637			break;
638		}
639#endif
640#ifdef INET6
641	case AF_INET6:
642		{
643			/* ok to use deprecated addresses? */
644			struct sockaddr_in6 *sin6;
645
646			sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
647			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
648			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
649				sctp_ifap->src_is_loop = 1;
650			}
651			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
652				sctp_ifap->src_is_priv = 1;
653			}
654			sctp_ifnp->num_v6++;
655			if (new_ifn_af)
656				new_ifn_af = AF_INET6;
657			break;
658		}
659#endif
660	default:
661		new_ifn_af = 0;
662		break;
663	}
664	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
665
666	if ((sctp_ifap->src_is_priv == 0) &&
667	    (sctp_ifap->src_is_loop == 0)) {
668		sctp_ifap->src_is_glob = 1;
669	}
670	SCTP_IPI_ADDR_WLOCK();
671	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
672	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
673	sctp_ifap->refcount = 1;
674	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
675	sctp_ifnp->ifa_count++;
676	vrf->total_ifa_count++;
677	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
678	if (new_ifn_af) {
679		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
680		sctp_ifnp->registered_af = new_ifn_af;
681	}
682	SCTP_IPI_ADDR_WUNLOCK();
683	if (dynamic_add) {
684		/*
685		 * Bump up the refcount so that when the timer completes it
686		 * will drop back down.
687		 */
688		struct sctp_laddr *wi;
689
690		atomic_add_int(&sctp_ifap->refcount, 1);
691		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
692		if (wi == NULL) {
693			/*
694			 * Gak, what can we do? We have lost an address
695			 * change can you say HOSED?
696			 */
697			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
698			/* Opps, must decrement the count */
699			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
700			    if_name);
701			return (NULL);
702		}
703		SCTP_INCR_LADDR_COUNT();
704		bzero(wi, sizeof(*wi));
705		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
706		wi->ifa = sctp_ifap;
707		wi->action = SCTP_ADD_IP_ADDRESS;
708
709		SCTP_WQ_ADDR_LOCK();
710		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
711		SCTP_WQ_ADDR_UNLOCK();
712
713		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
714		    (struct sctp_inpcb *)NULL,
715		    (struct sctp_tcb *)NULL,
716		    (struct sctp_nets *)NULL);
717	} else {
718		/* it's ready for use */
719		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
720	}
721	return (sctp_ifap);
722}
723
724void
725sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
726    uint32_t ifn_index, const char *if_name)
727{
728	struct sctp_vrf *vrf;
729	struct sctp_ifa *sctp_ifap = NULL;
730
731	SCTP_IPI_ADDR_WLOCK();
732	vrf = sctp_find_vrf(vrf_id);
733	if (vrf == NULL) {
734		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
735		goto out_now;
736	}
737#ifdef SCTP_DEBUG
738	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
739	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
740#endif
741	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
742	if (sctp_ifap) {
743		/* Validate the delete */
744		if (sctp_ifap->ifn_p) {
745			int valid = 0;
746
747			/*-
748			 * The name has priority over the ifn_index
749			 * if its given. We do this especially for
750			 * panda who might recycle indexes fast.
751			 */
752			if (if_name) {
753				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
754					/* They match its a correct delete */
755					valid = 1;
756				}
757			}
758			if (!valid) {
759				/* last ditch check ifn_index */
760				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
761					valid = 1;
762				}
763			}
764			if (!valid) {
765				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
766				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
767				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
768				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
769				SCTP_IPI_ADDR_WUNLOCK();
770				return;
771			}
772		}
773		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
774		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
775		/*
776		 * We don't set the flag. This means that the structure will
777		 * hang around in EP's that have bound specific to it until
778		 * they close. This gives us TCP like behavior if someone
779		 * removes an address (or for that matter adds it right
780		 * back).
781		 */
782		/* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */
783		vrf->total_ifa_count--;
784		LIST_REMOVE(sctp_ifap, next_bucket);
785		sctp_remove_ifa_from_ifn(sctp_ifap);
786	}
787#ifdef SCTP_DEBUG
788	else {
789		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
790		    ifn_index);
791		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
792	}
793#endif
794
795out_now:
796	SCTP_IPI_ADDR_WUNLOCK();
797	if (sctp_ifap) {
798		struct sctp_laddr *wi;
799
800		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
801		if (wi == NULL) {
802			/*
803			 * Gak, what can we do? We have lost an address
804			 * change can you say HOSED?
805			 */
806			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
807
808			/* Oops, must decrement the count */
809			sctp_free_ifa(sctp_ifap);
810			return;
811		}
812		SCTP_INCR_LADDR_COUNT();
813		bzero(wi, sizeof(*wi));
814		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
815		wi->ifa = sctp_ifap;
816		wi->action = SCTP_DEL_IP_ADDRESS;
817		SCTP_WQ_ADDR_LOCK();
818		/*
819		 * Should this really be a tailq? As it is we will process
820		 * the newest first :-0
821		 */
822		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
823		SCTP_WQ_ADDR_UNLOCK();
824
825		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
826		    (struct sctp_inpcb *)NULL,
827		    (struct sctp_tcb *)NULL,
828		    (struct sctp_nets *)NULL);
829	}
830	return;
831}
832
833
834static int
835sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
836{
837	int loopback_scope;
838
839#if defined(INET)
840	int ipv4_local_scope, ipv4_addr_legal;
841
842#endif
843#if defined(INET6)
844	int local_scope, site_scope, ipv6_addr_legal;
845
846#endif
847	struct sctp_vrf *vrf;
848	struct sctp_ifn *sctp_ifn;
849	struct sctp_ifa *sctp_ifa;
850
851	loopback_scope = stcb->asoc.scope.loopback_scope;
852#if defined(INET)
853	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
854	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
855#endif
856#if defined(INET6)
857	local_scope = stcb->asoc.scope.local_scope;
858	site_scope = stcb->asoc.scope.site_scope;
859	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
860#endif
861
862	SCTP_IPI_ADDR_RLOCK();
863	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
864	if (vrf == NULL) {
865		/* no vrf, no addresses */
866		SCTP_IPI_ADDR_RUNLOCK();
867		return (0);
868	}
869	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
870		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
871			if ((loopback_scope == 0) &&
872			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
873				continue;
874			}
875			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
876				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
877				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
878					/*
879					 * We allow pending addresses, where
880					 * we have sent an asconf-add to be
881					 * considered valid.
882					 */
883					continue;
884				}
885				if (sctp_ifa->address.sa.sa_family != to->sa_family) {
886					continue;
887				}
888				switch (sctp_ifa->address.sa.sa_family) {
889#ifdef INET
890				case AF_INET:
891					if (ipv4_addr_legal) {
892						struct sockaddr_in *sin,
893						           *rsin;
894
895						sin = &sctp_ifa->address.sin;
896						rsin = (struct sockaddr_in *)to;
897						if ((ipv4_local_scope == 0) &&
898						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
899							continue;
900						}
901						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
902							SCTP_IPI_ADDR_RUNLOCK();
903							return (1);
904						}
905					}
906					break;
907#endif
908#ifdef INET6
909				case AF_INET6:
910					if (ipv6_addr_legal) {
911						struct sockaddr_in6 *sin6,
912						            *rsin6;
913
914						sin6 = &sctp_ifa->address.sin6;
915						rsin6 = (struct sockaddr_in6 *)to;
916						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
917							if (local_scope == 0)
918								continue;
919							if (sin6->sin6_scope_id == 0) {
920								if (sa6_recoverscope(sin6) != 0)
921									continue;
922							}
923						}
924						if ((site_scope == 0) &&
925						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
926							continue;
927						}
928						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
929							SCTP_IPI_ADDR_RUNLOCK();
930							return (1);
931						}
932					}
933					break;
934#endif
935				default:
936					/* TSNH */
937					break;
938				}
939			}
940		}
941	} else {
942		struct sctp_laddr *laddr;
943
944		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
945			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
946				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
947				continue;
948			}
949			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
950			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
951				/*
952				 * We allow pending addresses, where we have
953				 * sent an asconf-add to be considered
954				 * valid.
955				 */
956				continue;
957			}
958			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
959				continue;
960			}
961			switch (to->sa_family) {
962#ifdef INET
963			case AF_INET:
964				{
965					struct sockaddr_in *sin, *rsin;
966
967					sin = (struct sockaddr_in *)&laddr->ifa->address.sin;
968					rsin = (struct sockaddr_in *)to;
969					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
970						SCTP_IPI_ADDR_RUNLOCK();
971						return (1);
972					}
973					break;
974				}
975#endif
976#ifdef INET6
977			case AF_INET6:
978				{
979					struct sockaddr_in6 *sin6, *rsin6;
980
981					sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
982					rsin6 = (struct sockaddr_in6 *)to;
983					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
984						SCTP_IPI_ADDR_RUNLOCK();
985						return (1);
986					}
987					break;
988				}
989
990#endif
991			default:
992				/* TSNH */
993				break;
994			}
995
996		}
997	}
998	SCTP_IPI_ADDR_RUNLOCK();
999	return (0);
1000}
1001
1002
1003static struct sctp_tcb *
1004sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
1005    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
1006{
1007	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
1008	/*
1009	 * If we support the TCP model, then we must now dig through to see
1010	 * if we can find our endpoint in the list of tcp ep's.
1011	 */
1012	uint16_t lport, rport;
1013	struct sctppcbhead *ephead;
1014	struct sctp_inpcb *inp;
1015	struct sctp_laddr *laddr;
1016	struct sctp_tcb *stcb;
1017	struct sctp_nets *net;
1018
1019	if ((to == NULL) || (from == NULL)) {
1020		return (NULL);
1021	}
1022	switch (to->sa_family) {
1023#ifdef INET
1024	case AF_INET:
1025		if (from->sa_family == AF_INET) {
1026			lport = ((struct sockaddr_in *)to)->sin_port;
1027			rport = ((struct sockaddr_in *)from)->sin_port;
1028		} else {
1029			return (NULL);
1030		}
1031		break;
1032#endif
1033#ifdef INET6
1034	case AF_INET6:
1035		if (from->sa_family == AF_INET6) {
1036			lport = ((struct sockaddr_in6 *)to)->sin6_port;
1037			rport = ((struct sockaddr_in6 *)from)->sin6_port;
1038		} else {
1039			return (NULL);
1040		}
1041		break;
1042#endif
1043	default:
1044		return (NULL);
1045	}
1046	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
1047	/*
1048	 * Ok now for each of the guys in this bucket we must look and see:
1049	 * - Does the remote port match. - Does there single association's
1050	 * addresses match this address (to). If so we update p_ep to point
1051	 * to this ep and return the tcb from it.
1052	 */
1053	LIST_FOREACH(inp, ephead, sctp_hash) {
1054		SCTP_INP_RLOCK(inp);
1055		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1056			SCTP_INP_RUNLOCK(inp);
1057			continue;
1058		}
1059		if (lport != inp->sctp_lport) {
1060			SCTP_INP_RUNLOCK(inp);
1061			continue;
1062		}
1063		if (inp->def_vrf_id != vrf_id) {
1064			SCTP_INP_RUNLOCK(inp);
1065			continue;
1066		}
1067		/* check to see if the ep has one of the addresses */
1068		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1069			/* We are NOT bound all, so look further */
1070			int match = 0;
1071
1072			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1073
1074				if (laddr->ifa == NULL) {
1075					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
1076					continue;
1077				}
1078				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1079					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1080					continue;
1081				}
1082				if (laddr->ifa->address.sa.sa_family ==
1083				    to->sa_family) {
1084					/* see if it matches */
1085#ifdef INET
1086					if (from->sa_family == AF_INET) {
1087						struct sockaddr_in *intf_addr,
1088						           *sin;
1089
1090						intf_addr = &laddr->ifa->address.sin;
1091						sin = (struct sockaddr_in *)to;
1092						if (sin->sin_addr.s_addr ==
1093						    intf_addr->sin_addr.s_addr) {
1094							match = 1;
1095							break;
1096						}
1097					}
1098#endif
1099#ifdef INET6
1100					if (from->sa_family == AF_INET6) {
1101						struct sockaddr_in6 *intf_addr6;
1102						struct sockaddr_in6 *sin6;
1103
1104						sin6 = (struct sockaddr_in6 *)
1105						    to;
1106						intf_addr6 = &laddr->ifa->address.sin6;
1107
1108						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1109						    intf_addr6)) {
1110							match = 1;
1111							break;
1112						}
1113					}
1114#endif
1115				}
1116			}
1117			if (match == 0) {
1118				/* This endpoint does not have this address */
1119				SCTP_INP_RUNLOCK(inp);
1120				continue;
1121			}
1122		}
1123		/*
1124		 * Ok if we hit here the ep has the address, does it hold
1125		 * the tcb?
1126		 */
1127		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1128		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1129		if (stcb == NULL) {
1130			SCTP_INP_RUNLOCK(inp);
1131			continue;
1132		}
1133		SCTP_TCB_LOCK(stcb);
1134		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1135			SCTP_TCB_UNLOCK(stcb);
1136			SCTP_INP_RUNLOCK(inp);
1137			continue;
1138		}
1139		if (stcb->rport != rport) {
1140			/* remote port does not match. */
1141			SCTP_TCB_UNLOCK(stcb);
1142			SCTP_INP_RUNLOCK(inp);
1143			continue;
1144		}
1145		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1146			SCTP_TCB_UNLOCK(stcb);
1147			SCTP_INP_RUNLOCK(inp);
1148			continue;
1149		}
1150		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1151			SCTP_TCB_UNLOCK(stcb);
1152			SCTP_INP_RUNLOCK(inp);
1153			continue;
1154		}
1155		/* Does this TCB have a matching address? */
1156		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1157
1158			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1159				/* not the same family, can't be a match */
1160				continue;
1161			}
1162			switch (from->sa_family) {
1163#ifdef INET
1164			case AF_INET:
1165				{
1166					struct sockaddr_in *sin, *rsin;
1167
1168					sin = (struct sockaddr_in *)&net->ro._l_addr;
1169					rsin = (struct sockaddr_in *)from;
1170					if (sin->sin_addr.s_addr ==
1171					    rsin->sin_addr.s_addr) {
1172						/* found it */
1173						if (netp != NULL) {
1174							*netp = net;
1175						}
1176						/*
1177						 * Update the endpoint
1178						 * pointer
1179						 */
1180						*inp_p = inp;
1181						SCTP_INP_RUNLOCK(inp);
1182						return (stcb);
1183					}
1184					break;
1185				}
1186#endif
1187#ifdef INET6
1188			case AF_INET6:
1189				{
1190					struct sockaddr_in6 *sin6, *rsin6;
1191
1192					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1193					rsin6 = (struct sockaddr_in6 *)from;
1194					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1195					    rsin6)) {
1196						/* found it */
1197						if (netp != NULL) {
1198							*netp = net;
1199						}
1200						/*
1201						 * Update the endpoint
1202						 * pointer
1203						 */
1204						*inp_p = inp;
1205						SCTP_INP_RUNLOCK(inp);
1206						return (stcb);
1207					}
1208					break;
1209				}
1210#endif
1211			default:
1212				/* TSNH */
1213				break;
1214			}
1215		}
1216		SCTP_TCB_UNLOCK(stcb);
1217		SCTP_INP_RUNLOCK(inp);
1218	}
1219	return (NULL);
1220}
1221
1222
1223/*
1224 * rules for use
1225 *
1226 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1227 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1228 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1229 * NULL.
1230 */
1231
1232struct sctp_tcb *
1233sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1234    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1235{
1236	struct sctpasochead *head;
1237	struct sctp_inpcb *inp;
1238	struct sctp_tcb *stcb = NULL;
1239	struct sctp_nets *net;
1240	uint16_t rport;
1241
1242	inp = *inp_p;
1243	switch (remote->sa_family) {
1244#ifdef INET
1245	case AF_INET:
1246		rport = (((struct sockaddr_in *)remote)->sin_port);
1247		break;
1248#endif
1249#ifdef INET6
1250	case AF_INET6:
1251		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1252		break;
1253#endif
1254	default:
1255		return (NULL);
1256	}
1257	if (locked_tcb) {
1258		/*
1259		 * UN-lock so we can do proper locking here this occurs when
1260		 * called from load_addresses_from_init.
1261		 */
1262		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1263		SCTP_TCB_UNLOCK(locked_tcb);
1264	}
1265	SCTP_INP_INFO_RLOCK();
1266	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1267	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1268		/*-
1269		 * Now either this guy is our listener or it's the
1270		 * connector. If it is the one that issued the connect, then
1271		 * it's only chance is to be the first TCB in the list. If
1272		 * it is the acceptor, then do the special_lookup to hash
1273		 * and find the real inp.
1274		 */
1275		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1276			/* to is peer addr, from is my addr */
1277			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1278			    netp, inp->def_vrf_id);
1279			if ((stcb != NULL) && (locked_tcb == NULL)) {
1280				/* we have a locked tcb, lower refcount */
1281				SCTP_INP_DECR_REF(inp);
1282			}
1283			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1284				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1285				SCTP_TCB_LOCK(locked_tcb);
1286				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1287				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1288			}
1289			SCTP_INP_INFO_RUNLOCK();
1290			return (stcb);
1291		} else {
1292			SCTP_INP_WLOCK(inp);
1293			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1294				goto null_return;
1295			}
1296			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1297			if (stcb == NULL) {
1298				goto null_return;
1299			}
1300			SCTP_TCB_LOCK(stcb);
1301
1302			if (stcb->rport != rport) {
1303				/* remote port does not match. */
1304				SCTP_TCB_UNLOCK(stcb);
1305				goto null_return;
1306			}
1307			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1308				SCTP_TCB_UNLOCK(stcb);
1309				goto null_return;
1310			}
1311			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1312				SCTP_TCB_UNLOCK(stcb);
1313				goto null_return;
1314			}
1315			/* now look at the list of remote addresses */
1316			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1317#ifdef INVARIANTS
1318				if (net == (TAILQ_NEXT(net, sctp_next))) {
1319					panic("Corrupt net list");
1320				}
1321#endif
1322				if (net->ro._l_addr.sa.sa_family !=
1323				    remote->sa_family) {
1324					/* not the same family */
1325					continue;
1326				}
1327				switch (remote->sa_family) {
1328#ifdef INET
1329				case AF_INET:
1330					{
1331						struct sockaddr_in *sin,
1332						           *rsin;
1333
1334						sin = (struct sockaddr_in *)
1335						    &net->ro._l_addr;
1336						rsin = (struct sockaddr_in *)remote;
1337						if (sin->sin_addr.s_addr ==
1338						    rsin->sin_addr.s_addr) {
1339							/* found it */
1340							if (netp != NULL) {
1341								*netp = net;
1342							}
1343							if (locked_tcb == NULL) {
1344								SCTP_INP_DECR_REF(inp);
1345							} else if (locked_tcb != stcb) {
1346								SCTP_TCB_LOCK(locked_tcb);
1347							}
1348							if (locked_tcb) {
1349								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1350							}
1351							SCTP_INP_WUNLOCK(inp);
1352							SCTP_INP_INFO_RUNLOCK();
1353							return (stcb);
1354						}
1355						break;
1356					}
1357#endif
1358#ifdef INET6
1359				case AF_INET6:
1360					{
1361						struct sockaddr_in6 *sin6,
1362						            *rsin6;
1363
1364						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1365						rsin6 = (struct sockaddr_in6 *)remote;
1366						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1367						    rsin6)) {
1368							/* found it */
1369							if (netp != NULL) {
1370								*netp = net;
1371							}
1372							if (locked_tcb == NULL) {
1373								SCTP_INP_DECR_REF(inp);
1374							} else if (locked_tcb != stcb) {
1375								SCTP_TCB_LOCK(locked_tcb);
1376							}
1377							if (locked_tcb) {
1378								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1379							}
1380							SCTP_INP_WUNLOCK(inp);
1381							SCTP_INP_INFO_RUNLOCK();
1382							return (stcb);
1383						}
1384						break;
1385					}
1386#endif
1387				default:
1388					/* TSNH */
1389					break;
1390				}
1391			}
1392			SCTP_TCB_UNLOCK(stcb);
1393		}
1394	} else {
1395		SCTP_INP_WLOCK(inp);
1396		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1397			goto null_return;
1398		}
1399		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1400		    inp->sctp_hashmark)];
1401		if (head == NULL) {
1402			goto null_return;
1403		}
1404		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1405			if (stcb->rport != rport) {
1406				/* remote port does not match */
1407				continue;
1408			}
1409			SCTP_TCB_LOCK(stcb);
1410			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1411				SCTP_TCB_UNLOCK(stcb);
1412				continue;
1413			}
1414			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1415				SCTP_TCB_UNLOCK(stcb);
1416				continue;
1417			}
1418			/* now look at the list of remote addresses */
1419			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1420#ifdef INVARIANTS
1421				if (net == (TAILQ_NEXT(net, sctp_next))) {
1422					panic("Corrupt net list");
1423				}
1424#endif
1425				if (net->ro._l_addr.sa.sa_family !=
1426				    remote->sa_family) {
1427					/* not the same family */
1428					continue;
1429				}
1430				switch (remote->sa_family) {
1431#ifdef INET
1432				case AF_INET:
1433					{
1434						struct sockaddr_in *sin,
1435						           *rsin;
1436
1437						sin = (struct sockaddr_in *)
1438						    &net->ro._l_addr;
1439						rsin = (struct sockaddr_in *)remote;
1440						if (sin->sin_addr.s_addr ==
1441						    rsin->sin_addr.s_addr) {
1442							/* found it */
1443							if (netp != NULL) {
1444								*netp = net;
1445							}
1446							if (locked_tcb == NULL) {
1447								SCTP_INP_DECR_REF(inp);
1448							} else if (locked_tcb != stcb) {
1449								SCTP_TCB_LOCK(locked_tcb);
1450							}
1451							if (locked_tcb) {
1452								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1453							}
1454							SCTP_INP_WUNLOCK(inp);
1455							SCTP_INP_INFO_RUNLOCK();
1456							return (stcb);
1457						}
1458						break;
1459					}
1460#endif
1461#ifdef INET6
1462				case AF_INET6:
1463					{
1464						struct sockaddr_in6 *sin6,
1465						            *rsin6;
1466
1467						sin6 = (struct sockaddr_in6 *)
1468						    &net->ro._l_addr;
1469						rsin6 = (struct sockaddr_in6 *)remote;
1470						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1471						    rsin6)) {
1472							/* found it */
1473							if (netp != NULL) {
1474								*netp = net;
1475							}
1476							if (locked_tcb == NULL) {
1477								SCTP_INP_DECR_REF(inp);
1478							} else if (locked_tcb != stcb) {
1479								SCTP_TCB_LOCK(locked_tcb);
1480							}
1481							if (locked_tcb) {
1482								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1483							}
1484							SCTP_INP_WUNLOCK(inp);
1485							SCTP_INP_INFO_RUNLOCK();
1486							return (stcb);
1487						}
1488						break;
1489					}
1490#endif
1491				default:
1492					/* TSNH */
1493					break;
1494				}
1495			}
1496			SCTP_TCB_UNLOCK(stcb);
1497		}
1498	}
1499null_return:
1500	/* clean up for returning null */
1501	if (locked_tcb) {
1502		SCTP_TCB_LOCK(locked_tcb);
1503		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1504	}
1505	SCTP_INP_WUNLOCK(inp);
1506	SCTP_INP_INFO_RUNLOCK();
1507	/* not found */
1508	return (NULL);
1509}
1510
1511
1512/*
1513 * Find an association for a specific endpoint using the association id given
1514 * out in the COMM_UP notification
1515 */
1516struct sctp_tcb *
1517sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1518{
1519	/*
1520	 * Use my the assoc_id to find a endpoint
1521	 */
1522	struct sctpasochead *head;
1523	struct sctp_tcb *stcb;
1524	uint32_t id;
1525
1526	if (inp == NULL) {
1527		SCTP_PRINTF("TSNH ep_associd\n");
1528		return (NULL);
1529	}
1530	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1531		SCTP_PRINTF("TSNH ep_associd0\n");
1532		return (NULL);
1533	}
1534	id = (uint32_t) asoc_id;
1535	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1536	if (head == NULL) {
1537		/* invalid id TSNH */
1538		SCTP_PRINTF("TSNH ep_associd1\n");
1539		return (NULL);
1540	}
1541	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1542		if (stcb->asoc.assoc_id == id) {
1543			if (inp != stcb->sctp_ep) {
1544				/*
1545				 * some other guy has the same id active (id
1546				 * collision ??).
1547				 */
1548				SCTP_PRINTF("TSNH ep_associd2\n");
1549				continue;
1550			}
1551			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1552				continue;
1553			}
1554			if (want_lock) {
1555				SCTP_TCB_LOCK(stcb);
1556			}
1557			return (stcb);
1558		}
1559	}
1560	return (NULL);
1561}
1562
1563
1564struct sctp_tcb *
1565sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1566{
1567	struct sctp_tcb *stcb;
1568
1569	SCTP_INP_RLOCK(inp);
1570	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1571	SCTP_INP_RUNLOCK(inp);
1572	return (stcb);
1573}
1574
1575
1576/*
1577 * Endpoint probe expects that the INP_INFO is locked.
1578 */
1579static struct sctp_inpcb *
1580sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1581    uint16_t lport, uint32_t vrf_id)
1582{
1583	struct sctp_inpcb *inp;
1584	struct sctp_laddr *laddr;
1585
1586#ifdef INET
1587	struct sockaddr_in *sin;
1588
1589#endif
1590#ifdef INET6
1591	struct sockaddr_in6 *sin6;
1592	struct sockaddr_in6 *intf_addr6;
1593
1594#endif
1595	int fnd;
1596
1597#ifdef INET
1598	sin = NULL;
1599#endif
1600#ifdef INET6
1601	sin6 = NULL;
1602#endif
1603	switch (nam->sa_family) {
1604#ifdef INET
1605	case AF_INET:
1606		sin = (struct sockaddr_in *)nam;
1607		break;
1608#endif
1609#ifdef INET6
1610	case AF_INET6:
1611		sin6 = (struct sockaddr_in6 *)nam;
1612		break;
1613#endif
1614	default:
1615		/* unsupported family */
1616		return (NULL);
1617	}
1618
1619	if (head == NULL)
1620		return (NULL);
1621
1622	LIST_FOREACH(inp, head, sctp_hash) {
1623		SCTP_INP_RLOCK(inp);
1624		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1625			SCTP_INP_RUNLOCK(inp);
1626			continue;
1627		}
1628		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1629		    (inp->sctp_lport == lport)) {
1630			/* got it */
1631#ifdef INET
1632			if ((nam->sa_family == AF_INET) &&
1633			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1634			    SCTP_IPV6_V6ONLY(inp)) {
1635				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1636				SCTP_INP_RUNLOCK(inp);
1637				continue;
1638			}
1639#endif
1640#ifdef INET6
1641			/* A V6 address and the endpoint is NOT bound V6 */
1642			if (nam->sa_family == AF_INET6 &&
1643			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1644				SCTP_INP_RUNLOCK(inp);
1645				continue;
1646			}
1647#endif
1648			/* does a VRF id match? */
1649			fnd = 0;
1650			if (inp->def_vrf_id == vrf_id)
1651				fnd = 1;
1652
1653			SCTP_INP_RUNLOCK(inp);
1654			if (!fnd)
1655				continue;
1656			return (inp);
1657		}
1658		SCTP_INP_RUNLOCK(inp);
1659	}
1660	switch (nam->sa_family) {
1661#ifdef INET
1662	case AF_INET:
1663		if (sin->sin_addr.s_addr == INADDR_ANY) {
1664			/* Can't hunt for one that has no address specified */
1665			return (NULL);
1666		}
1667		break;
1668#endif
1669#ifdef INET6
1670	case AF_INET6:
1671		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1672			/* Can't hunt for one that has no address specified */
1673			return (NULL);
1674		}
1675		break;
1676#endif
1677	default:
1678		break;
1679	}
1680	/*
1681	 * ok, not bound to all so see if we can find a EP bound to this
1682	 * address.
1683	 */
1684	LIST_FOREACH(inp, head, sctp_hash) {
1685		SCTP_INP_RLOCK(inp);
1686		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1687			SCTP_INP_RUNLOCK(inp);
1688			continue;
1689		}
1690		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1691			SCTP_INP_RUNLOCK(inp);
1692			continue;
1693		}
1694		/*
1695		 * Ok this could be a likely candidate, look at all of its
1696		 * addresses
1697		 */
1698		if (inp->sctp_lport != lport) {
1699			SCTP_INP_RUNLOCK(inp);
1700			continue;
1701		}
1702		/* does a VRF id match? */
1703		fnd = 0;
1704		if (inp->def_vrf_id == vrf_id)
1705			fnd = 1;
1706
1707		if (!fnd) {
1708			SCTP_INP_RUNLOCK(inp);
1709			continue;
1710		}
1711		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1712			if (laddr->ifa == NULL) {
1713				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1714				    __FUNCTION__);
1715				continue;
1716			}
1717			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1718			    (void *)laddr->ifa);
1719			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1720				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1721				continue;
1722			}
1723			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1724				/* possible, see if it matches */
1725				switch (nam->sa_family) {
1726#ifdef INET
1727				case AF_INET:
1728					if (sin->sin_addr.s_addr ==
1729					    laddr->ifa->address.sin.sin_addr.s_addr) {
1730						SCTP_INP_RUNLOCK(inp);
1731						return (inp);
1732					}
1733					break;
1734#endif
1735#ifdef INET6
1736				case AF_INET6:
1737					intf_addr6 = &laddr->ifa->address.sin6;
1738					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1739					    intf_addr6)) {
1740						SCTP_INP_RUNLOCK(inp);
1741						return (inp);
1742					}
1743					break;
1744#endif
1745				}
1746			}
1747		}
1748		SCTP_INP_RUNLOCK(inp);
1749	}
1750	return (NULL);
1751}
1752
1753
1754static struct sctp_inpcb *
1755sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1756{
1757	struct sctppcbhead *head;
1758	struct sctp_inpcb *t_inp;
1759	int fnd;
1760
1761	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1762	    SCTP_BASE_INFO(hashmark))];
1763	LIST_FOREACH(t_inp, head, sctp_hash) {
1764		if (t_inp->sctp_lport != lport) {
1765			continue;
1766		}
1767		/* is it in the VRF in question */
1768		fnd = 0;
1769		if (t_inp->def_vrf_id == vrf_id)
1770			fnd = 1;
1771		if (!fnd)
1772			continue;
1773
1774		/* This one is in use. */
1775		/* check the v6/v4 binding issue */
1776		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1777		    SCTP_IPV6_V6ONLY(t_inp)) {
1778			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1779				/* collision in V6 space */
1780				return (t_inp);
1781			} else {
1782				/* inp is BOUND_V4 no conflict */
1783				continue;
1784			}
1785		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1786			/* t_inp is bound v4 and v6, conflict always */
1787			return (t_inp);
1788		} else {
1789			/* t_inp is bound only V4 */
1790			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1791			    SCTP_IPV6_V6ONLY(inp)) {
1792				/* no conflict */
1793				continue;
1794			}
1795			/* else fall through to conflict */
1796		}
1797		return (t_inp);
1798	}
1799	return (NULL);
1800}
1801
1802
1803int
1804sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1805{
1806	/* For 1-2-1 with port reuse */
1807	struct sctppcbhead *head;
1808	struct sctp_inpcb *tinp;
1809
1810	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1811		/* only works with port reuse on */
1812		return (-1);
1813	}
1814	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1815		return (0);
1816	}
1817	SCTP_INP_RUNLOCK(inp);
1818	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1819	    SCTP_BASE_INFO(hashmark))];
1820	/* Kick out all non-listeners to the TCP hash */
1821	LIST_FOREACH(tinp, head, sctp_hash) {
1822		if (tinp->sctp_lport != inp->sctp_lport) {
1823			continue;
1824		}
1825		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1826			continue;
1827		}
1828		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1829			continue;
1830		}
1831		if (tinp->sctp_socket->so_qlimit) {
1832			continue;
1833		}
1834		SCTP_INP_WLOCK(tinp);
1835		LIST_REMOVE(tinp, sctp_hash);
1836		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1837		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1838		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1839		SCTP_INP_WUNLOCK(tinp);
1840	}
1841	SCTP_INP_WLOCK(inp);
1842	/* Pull from where he was */
1843	LIST_REMOVE(inp, sctp_hash);
1844	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1845	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1846	LIST_INSERT_HEAD(head, inp, sctp_hash);
1847	SCTP_INP_WUNLOCK(inp);
1848	SCTP_INP_RLOCK(inp);
1849	return (0);
1850}
1851
1852
1853struct sctp_inpcb *
1854sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1855    uint32_t vrf_id)
1856{
1857	/*
1858	 * First we check the hash table to see if someone has this port
1859	 * bound with just the port.
1860	 */
1861	struct sctp_inpcb *inp;
1862	struct sctppcbhead *head;
1863	int lport;
1864	unsigned int i;
1865
1866#ifdef INET
1867	struct sockaddr_in *sin;
1868
1869#endif
1870#ifdef INET6
1871	struct sockaddr_in6 *sin6;
1872
1873#endif
1874
1875	switch (nam->sa_family) {
1876#ifdef INET
1877	case AF_INET:
1878		sin = (struct sockaddr_in *)nam;
1879		lport = sin->sin_port;
1880		break;
1881#endif
1882#ifdef INET6
1883	case AF_INET6:
1884		sin6 = (struct sockaddr_in6 *)nam;
1885		lport = sin6->sin6_port;
1886		break;
1887#endif
1888	default:
1889		return (NULL);
1890	}
1891	/*
1892	 * I could cheat here and just cast to one of the types but we will
1893	 * do it right. It also provides the check against an Unsupported
1894	 * type too.
1895	 */
1896	/* Find the head of the ALLADDR chain */
1897	if (have_lock == 0) {
1898		SCTP_INP_INFO_RLOCK();
1899	}
1900	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1901	    SCTP_BASE_INFO(hashmark))];
1902	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1903
1904	/*
1905	 * If the TCP model exists it could be that the main listening
1906	 * endpoint is gone but there still exists a connected socket for
1907	 * this guy. If so we can return the first one that we find. This
1908	 * may NOT be the correct one so the caller should be wary on the
1909	 * returned INP. Currently the only caller that sets find_tcp_pool
1910	 * is in bindx where we are verifying that a user CAN bind the
1911	 * address. He either has bound it already, or someone else has, or
1912	 * its open to bind, so this is good enough.
1913	 */
1914	if (inp == NULL && find_tcp_pool) {
1915		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1916			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1917			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1918			if (inp) {
1919				break;
1920			}
1921		}
1922	}
1923	if (inp) {
1924		SCTP_INP_INCR_REF(inp);
1925	}
1926	if (have_lock == 0) {
1927		SCTP_INP_INFO_RUNLOCK();
1928	}
1929	return (inp);
1930}
1931
1932
1933/*
1934 * Find an association for an endpoint with the pointer to whom you want to
1935 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1936 * need to change the *to to some other struct like a mbuf...
1937 */
1938struct sctp_tcb *
1939sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
1940    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1941    uint32_t vrf_id)
1942{
1943	struct sctp_inpcb *inp = NULL;
1944	struct sctp_tcb *stcb;
1945
1946	SCTP_INP_INFO_RLOCK();
1947	if (find_tcp_pool) {
1948		if (inp_p != NULL) {
1949			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
1950			    vrf_id);
1951		} else {
1952			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
1953			    vrf_id);
1954		}
1955		if (stcb != NULL) {
1956			SCTP_INP_INFO_RUNLOCK();
1957			return (stcb);
1958		}
1959	}
1960	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1961	if (inp_p != NULL) {
1962		*inp_p = inp;
1963	}
1964	SCTP_INP_INFO_RUNLOCK();
1965	if (inp == NULL) {
1966		return (NULL);
1967	}
1968	/*
1969	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1970	 * we now place the source address or from in the to of the find
1971	 * endpoint call. Since in reality this chain is used from the
1972	 * inbound packet side.
1973	 */
1974	if (inp_p != NULL) {
1975		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1976		    NULL);
1977	} else {
1978		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
1979		    NULL);
1980	}
1981	return (stcb);
1982}
1983
1984
1985/*
1986 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1987 * find all addresses that the sender has specified in any address list. Each
1988 * address will be used to lookup the TCB and see if one exits.
1989 */
1990static struct sctp_tcb *
1991sctp_findassociation_special_addr(struct mbuf *m, int offset,
1992    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1993    struct sockaddr *dst)
1994{
1995	struct sctp_paramhdr *phdr, parm_buf;
1996
1997#if defined(INET) || defined(INET6)
1998	struct sctp_tcb *stcb;
1999	uint16_t ptype;
2000
2001#endif
2002	uint16_t plen;
2003
2004#ifdef INET
2005	struct sockaddr_in sin4;
2006
2007#endif
2008#ifdef INET6
2009	struct sockaddr_in6 sin6;
2010
2011#endif
2012
2013#ifdef INET
2014	memset(&sin4, 0, sizeof(sin4));
2015	sin4.sin_len = sizeof(sin4);
2016	sin4.sin_family = AF_INET;
2017	sin4.sin_port = sh->src_port;
2018#endif
2019#ifdef INET6
2020	memset(&sin6, 0, sizeof(sin6));
2021	sin6.sin6_len = sizeof(sin6);
2022	sin6.sin6_family = AF_INET6;
2023	sin6.sin6_port = sh->src_port;
2024#endif
2025
2026	offset += sizeof(struct sctp_init_chunk);
2027
2028	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2029	while (phdr != NULL) {
2030		/* now we must see if we want the parameter */
2031#if defined(INET) || defined(INET6)
2032		ptype = ntohs(phdr->param_type);
2033#endif
2034		plen = ntohs(phdr->param_length);
2035		if (plen == 0) {
2036			break;
2037		}
2038#ifdef INET
2039		if (ptype == SCTP_IPV4_ADDRESS &&
2040		    plen == sizeof(struct sctp_ipv4addr_param)) {
2041			/* Get the rest of the address */
2042			struct sctp_ipv4addr_param ip4_parm, *p4;
2043
2044			phdr = sctp_get_next_param(m, offset,
2045			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2046			if (phdr == NULL) {
2047				return (NULL);
2048			}
2049			p4 = (struct sctp_ipv4addr_param *)phdr;
2050			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2051			/* look it up */
2052			stcb = sctp_findassociation_ep_addr(inp_p,
2053			    (struct sockaddr *)&sin4, netp, dst, NULL);
2054			if (stcb != NULL) {
2055				return (stcb);
2056			}
2057		}
2058#endif
2059#ifdef INET6
2060		if (ptype == SCTP_IPV6_ADDRESS &&
2061		    plen == sizeof(struct sctp_ipv6addr_param)) {
2062			/* Get the rest of the address */
2063			struct sctp_ipv6addr_param ip6_parm, *p6;
2064
2065			phdr = sctp_get_next_param(m, offset,
2066			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2067			if (phdr == NULL) {
2068				return (NULL);
2069			}
2070			p6 = (struct sctp_ipv6addr_param *)phdr;
2071			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2072			/* look it up */
2073			stcb = sctp_findassociation_ep_addr(inp_p,
2074			    (struct sockaddr *)&sin6, netp, dst, NULL);
2075			if (stcb != NULL) {
2076				return (stcb);
2077			}
2078		}
2079#endif
2080		offset += SCTP_SIZE32(plen);
2081		phdr = sctp_get_next_param(m, offset, &parm_buf,
2082		    sizeof(parm_buf));
2083	}
2084	return (NULL);
2085}
2086
2087static struct sctp_tcb *
2088sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2089    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2090    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2091{
2092	/*
2093	 * Use my vtag to hash. If we find it we then verify the source addr
2094	 * is in the assoc. If all goes well we save a bit on rec of a
2095	 * packet.
2096	 */
2097	struct sctpasochead *head;
2098	struct sctp_nets *net;
2099	struct sctp_tcb *stcb;
2100
2101	SCTP_INP_INFO_RLOCK();
2102	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2103	    SCTP_BASE_INFO(hashasocmark))];
2104	if (head == NULL) {
2105		/* invalid vtag */
2106		SCTP_INP_INFO_RUNLOCK();
2107		return (NULL);
2108	}
2109	LIST_FOREACH(stcb, head, sctp_asocs) {
2110		SCTP_INP_RLOCK(stcb->sctp_ep);
2111		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2112			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2113			continue;
2114		}
2115		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2116			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2117			continue;
2118		}
2119		SCTP_TCB_LOCK(stcb);
2120		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2121		if (stcb->asoc.my_vtag == vtag) {
2122			/* candidate */
2123			if (stcb->rport != rport) {
2124				SCTP_TCB_UNLOCK(stcb);
2125				continue;
2126			}
2127			if (stcb->sctp_ep->sctp_lport != lport) {
2128				SCTP_TCB_UNLOCK(stcb);
2129				continue;
2130			}
2131			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2132				SCTP_TCB_UNLOCK(stcb);
2133				continue;
2134			}
2135			/* RRS:Need toaddr check here */
2136			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2137				/* Endpoint does not own this address */
2138				SCTP_TCB_UNLOCK(stcb);
2139				continue;
2140			}
2141			if (remote_tag) {
2142				/*
2143				 * If we have both vtags that's all we match
2144				 * on
2145				 */
2146				if (stcb->asoc.peer_vtag == remote_tag) {
2147					/*
2148					 * If both tags match we consider it
2149					 * conclusive and check NO
2150					 * source/destination addresses
2151					 */
2152					goto conclusive;
2153				}
2154			}
2155			if (skip_src_check) {
2156		conclusive:
2157				if (from) {
2158					*netp = sctp_findnet(stcb, from);
2159				} else {
2160					*netp = NULL;	/* unknown */
2161				}
2162				if (inp_p)
2163					*inp_p = stcb->sctp_ep;
2164				SCTP_INP_INFO_RUNLOCK();
2165				return (stcb);
2166			}
2167			net = sctp_findnet(stcb, from);
2168			if (net) {
2169				/* yep its him. */
2170				*netp = net;
2171				SCTP_STAT_INCR(sctps_vtagexpress);
2172				*inp_p = stcb->sctp_ep;
2173				SCTP_INP_INFO_RUNLOCK();
2174				return (stcb);
2175			} else {
2176				/*
2177				 * not him, this should only happen in rare
2178				 * cases so I peg it.
2179				 */
2180				SCTP_STAT_INCR(sctps_vtagbogus);
2181			}
2182		}
2183		SCTP_TCB_UNLOCK(stcb);
2184	}
2185	SCTP_INP_INFO_RUNLOCK();
2186	return (NULL);
2187}
2188
2189
2190/*
2191 * Find an association with the pointer to the inbound IP packet. This can be
2192 * a IPv4 or IPv6 packet.
2193 */
2194struct sctp_tcb *
2195sctp_findassociation_addr(struct mbuf *m, int offset,
2196    struct sockaddr *src, struct sockaddr *dst,
2197    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2198    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2199{
2200	int find_tcp_pool;
2201	struct sctp_tcb *stcb;
2202	struct sctp_inpcb *inp;
2203
2204	if (sh->v_tag) {
2205		/* we only go down this path if vtag is non-zero */
2206		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2207		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2208		if (stcb) {
2209			return (stcb);
2210		}
2211	}
2212	find_tcp_pool = 0;
2213	if ((ch->chunk_type != SCTP_INITIATION) &&
2214	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2215	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2216	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2217		/* Other chunk types go to the tcp pool. */
2218		find_tcp_pool = 1;
2219	}
2220	if (inp_p) {
2221		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2222		    find_tcp_pool, vrf_id);
2223		inp = *inp_p;
2224	} else {
2225		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2226		    find_tcp_pool, vrf_id);
2227	}
2228	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2229	if (stcb == NULL && inp) {
2230		/* Found a EP but not this address */
2231		if ((ch->chunk_type == SCTP_INITIATION) ||
2232		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2233			/*-
2234			 * special hook, we do NOT return linp or an
2235			 * association that is linked to an existing
2236			 * association that is under the TCP pool (i.e. no
2237			 * listener exists). The endpoint finding routine
2238			 * will always find a listener before examining the
2239			 * TCP pool.
2240			 */
2241			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2242				if (inp_p) {
2243					*inp_p = NULL;
2244				}
2245				return (NULL);
2246			}
2247			stcb = sctp_findassociation_special_addr(m,
2248			    offset, sh, &inp, netp, dst);
2249			if (inp_p != NULL) {
2250				*inp_p = inp;
2251			}
2252		}
2253	}
2254	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2255	return (stcb);
2256}
2257
2258/*
2259 * lookup an association by an ASCONF lookup address.
2260 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2261 */
2262struct sctp_tcb *
2263sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2264    struct sockaddr *dst, struct sctphdr *sh,
2265    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2266{
2267	struct sctp_tcb *stcb;
2268	struct sockaddr_storage remote_store;
2269	struct sctp_paramhdr parm_buf, *phdr;
2270	int ptype;
2271	int zero_address = 0;
2272
2273#ifdef INET
2274	struct sockaddr_in *sin;
2275
2276#endif
2277#ifdef INET6
2278	struct sockaddr_in6 *sin6;
2279
2280#endif
2281
2282	memset(&remote_store, 0, sizeof(remote_store));
2283	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2284	    &parm_buf, sizeof(struct sctp_paramhdr));
2285	if (phdr == NULL) {
2286		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2287		    __FUNCTION__);
2288		return NULL;
2289	}
2290	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2291	/* get the correlation address */
2292	switch (ptype) {
2293#ifdef INET6
2294	case SCTP_IPV6_ADDRESS:
2295		{
2296			/* ipv6 address param */
2297			struct sctp_ipv6addr_param *p6, p6_buf;
2298
2299			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2300				return NULL;
2301			}
2302			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2303			    offset + sizeof(struct sctp_asconf_chunk),
2304			    &p6_buf.ph, sizeof(*p6));
2305			if (p6 == NULL) {
2306				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2307				    __FUNCTION__);
2308				return (NULL);
2309			}
2310			sin6 = (struct sockaddr_in6 *)&remote_store;
2311			sin6->sin6_family = AF_INET6;
2312			sin6->sin6_len = sizeof(*sin6);
2313			sin6->sin6_port = sh->src_port;
2314			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2315			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2316				zero_address = 1;
2317			break;
2318		}
2319#endif
2320#ifdef INET
2321	case SCTP_IPV4_ADDRESS:
2322		{
2323			/* ipv4 address param */
2324			struct sctp_ipv4addr_param *p4, p4_buf;
2325
2326			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2327				return NULL;
2328			}
2329			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2330			    offset + sizeof(struct sctp_asconf_chunk),
2331			    &p4_buf.ph, sizeof(*p4));
2332			if (p4 == NULL) {
2333				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2334				    __FUNCTION__);
2335				return (NULL);
2336			}
2337			sin = (struct sockaddr_in *)&remote_store;
2338			sin->sin_family = AF_INET;
2339			sin->sin_len = sizeof(*sin);
2340			sin->sin_port = sh->src_port;
2341			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2342			if (sin->sin_addr.s_addr == INADDR_ANY)
2343				zero_address = 1;
2344			break;
2345		}
2346#endif
2347	default:
2348		/* invalid address param type */
2349		return NULL;
2350	}
2351
2352	if (zero_address) {
2353		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2354		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2355		if (stcb != NULL) {
2356			SCTP_INP_DECR_REF(*inp_p);
2357		}
2358	} else {
2359		stcb = sctp_findassociation_ep_addr(inp_p,
2360		    (struct sockaddr *)&remote_store, netp,
2361		    dst, NULL);
2362	}
2363	return (stcb);
2364}
2365
2366
2367/*
2368 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2369 * addresses. This way if we don't get a bind we by default pick a ephemeral
2370 * port with all addresses bound.
2371 */
2372int
2373sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2374{
2375	/*
2376	 * we get called when a new endpoint starts up. We need to allocate
2377	 * the sctp_inpcb structure from the zone and init it. Mark it as
2378	 * unbound and find a port that we can use as an ephemeral with
2379	 * INADDR_ANY. If the user binds later no problem we can then add in
2380	 * the specific addresses. And setup the default parameters for the
2381	 * EP.
2382	 */
2383	int i, error;
2384	struct sctp_inpcb *inp;
2385	struct sctp_pcb *m;
2386	struct timeval time;
2387	sctp_sharedkey_t *null_key;
2388
2389	error = 0;
2390
2391	SCTP_INP_INFO_WLOCK();
2392	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2393	if (inp == NULL) {
2394		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2395		SCTP_INP_INFO_WUNLOCK();
2396		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2397		return (ENOBUFS);
2398	}
2399	/* zap it */
2400	bzero(inp, sizeof(*inp));
2401
2402	/* bump generations */
2403	/* setup socket pointers */
2404	inp->sctp_socket = so;
2405	inp->ip_inp.inp.inp_socket = so;
2406#ifdef INET6
2407	if (INP_SOCKAF(so) == AF_INET6) {
2408		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2409			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2410		}
2411		if (MODULE_GLOBAL(ip6_v6only)) {
2412			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2413		}
2414	}
2415#endif
2416	inp->sctp_associd_counter = 1;
2417	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2418	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2419	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2420	inp->sctp_ecn_enable = SCTP_BASE_SYSCTL(sctp_ecn_enable);
2421	/* init the small hash table we use to track asocid <-> tcb */
2422	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2423	if (inp->sctp_asocidhash == NULL) {
2424		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2425		SCTP_INP_INFO_WUNLOCK();
2426		return (ENOBUFS);
2427	}
2428#ifdef IPSEC
2429	{
2430		struct inpcbpolicy *pcb_sp = NULL;
2431
2432		error = ipsec_init_policy(so, &pcb_sp);
2433		/* Arrange to share the policy */
2434		inp->ip_inp.inp.inp_sp = pcb_sp;
2435		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2436	}
2437	if (error != 0) {
2438		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2439		SCTP_INP_INFO_WUNLOCK();
2440		return error;
2441	}
2442#endif				/* IPSEC */
2443	SCTP_INCR_EP_COUNT();
2444	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2445	SCTP_INP_INFO_WUNLOCK();
2446
2447	so->so_pcb = (caddr_t)inp;
2448
2449	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2450		/* UDP style socket */
2451		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2452		    SCTP_PCB_FLAGS_UNBOUND);
2453		/* Be sure it is NON-BLOCKING IO for UDP */
2454		/* SCTP_SET_SO_NBIO(so); */
2455	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2456		/* TCP style socket */
2457		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2458		    SCTP_PCB_FLAGS_UNBOUND);
2459		/* Be sure we have blocking IO by default */
2460		SCTP_CLEAR_SO_NBIO(so);
2461	} else {
2462		/*
2463		 * unsupported socket type (RAW, etc)- in case we missed it
2464		 * in protosw
2465		 */
2466		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2467		so->so_pcb = NULL;
2468		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2469		return (EOPNOTSUPP);
2470	}
2471	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2472		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2473		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2474	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2475		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2476		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2477	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2478		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2479		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2480	}
2481	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2482	    &inp->sctp_hashmark);
2483	if (inp->sctp_tcbhash == NULL) {
2484		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2485		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2486		so->so_pcb = NULL;
2487		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2488		return (ENOBUFS);
2489	}
2490	inp->def_vrf_id = vrf_id;
2491
2492	SCTP_INP_INFO_WLOCK();
2493	SCTP_INP_LOCK_INIT(inp);
2494	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2495	SCTP_INP_READ_INIT(inp);
2496	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2497	/* lock the new ep */
2498	SCTP_INP_WLOCK(inp);
2499
2500	/* add it to the info area */
2501	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2502	SCTP_INP_INFO_WUNLOCK();
2503
2504	TAILQ_INIT(&inp->read_queue);
2505	LIST_INIT(&inp->sctp_addr_list);
2506
2507	LIST_INIT(&inp->sctp_asoc_list);
2508
2509#ifdef SCTP_TRACK_FREED_ASOCS
2510	/* TEMP CODE */
2511	LIST_INIT(&inp->sctp_asoc_free_list);
2512#endif
2513	/* Init the timer structure for signature change */
2514	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2515	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2516
2517	/* now init the actual endpoint default data */
2518	m = &inp->sctp_ep;
2519
2520	/* setup the base timeout information */
2521	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2522	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2523	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2524	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2525	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2526	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2527	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2528	/* all max/min max are in ms */
2529	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2530	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2531	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2532	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2533	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2534	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2535	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2536	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2537	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2538	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2539	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2540	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2541	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2542
2543	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2544	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2545	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2546	/* number of streams to pre-open on a association */
2547	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2548
2549	/* Add adaptation cookie */
2550	m->adaptation_layer_indicator = 0;
2551	m->adaptation_layer_indicator_provided = 0;
2552
2553	/* seed random number generator */
2554	m->random_counter = 1;
2555	m->store_at = SCTP_SIGNATURE_SIZE;
2556	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2557	sctp_fill_random_store(m);
2558
2559	/* Minimum cookie size */
2560	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2561	    sizeof(struct sctp_state_cookie);
2562	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2563
2564	/* Setup the initial secret */
2565	(void)SCTP_GETTIME_TIMEVAL(&time);
2566	m->time_of_secret_change = time.tv_sec;
2567
2568	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2569		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2570	}
2571	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2572
2573	/* How long is a cookie good for ? */
2574	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2575	/*
2576	 * Initialize authentication parameters
2577	 */
2578	m->local_hmacs = sctp_default_supported_hmaclist();
2579	m->local_auth_chunks = sctp_alloc_chunklist();
2580	m->default_dscp = 0;
2581#ifdef INET6
2582	m->default_flowlabel = 0;
2583#endif
2584	m->port = 0;		/* encapsulation disabled by default */
2585	sctp_auth_set_default_chunks(m->local_auth_chunks);
2586	LIST_INIT(&m->shared_keys);
2587	/* add default NULL key as key id 0 */
2588	null_key = sctp_alloc_sharedkey();
2589	sctp_insert_sharedkey(&m->shared_keys, null_key);
2590	SCTP_INP_WUNLOCK(inp);
2591#ifdef SCTP_LOG_CLOSING
2592	sctp_log_closing(inp, NULL, 12);
2593#endif
2594	return (error);
2595}
2596
2597
2598void
2599sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2600    struct sctp_tcb *stcb)
2601{
2602	struct sctp_nets *net;
2603	uint16_t lport, rport;
2604	struct sctppcbhead *head;
2605	struct sctp_laddr *laddr, *oladdr;
2606
2607	atomic_add_int(&stcb->asoc.refcnt, 1);
2608	SCTP_TCB_UNLOCK(stcb);
2609	SCTP_INP_INFO_WLOCK();
2610	SCTP_INP_WLOCK(old_inp);
2611	SCTP_INP_WLOCK(new_inp);
2612	SCTP_TCB_LOCK(stcb);
2613	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2614
2615	new_inp->sctp_ep.time_of_secret_change =
2616	    old_inp->sctp_ep.time_of_secret_change;
2617	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2618	    sizeof(old_inp->sctp_ep.secret_key));
2619	new_inp->sctp_ep.current_secret_number =
2620	    old_inp->sctp_ep.current_secret_number;
2621	new_inp->sctp_ep.last_secret_number =
2622	    old_inp->sctp_ep.last_secret_number;
2623	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2624
2625	/* make it so new data pours into the new socket */
2626	stcb->sctp_socket = new_inp->sctp_socket;
2627	stcb->sctp_ep = new_inp;
2628
2629	/* Copy the port across */
2630	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2631	rport = stcb->rport;
2632	/* Pull the tcb from the old association */
2633	LIST_REMOVE(stcb, sctp_tcbhash);
2634	LIST_REMOVE(stcb, sctp_tcblist);
2635	if (stcb->asoc.in_asocid_hash) {
2636		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2637	}
2638	/* Now insert the new_inp into the TCP connected hash */
2639	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2640
2641	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2642	/* Its safe to access */
2643	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2644
2645	/* Now move the tcb into the endpoint list */
2646	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2647	/*
2648	 * Question, do we even need to worry about the ep-hash since we
2649	 * only have one connection? Probably not :> so lets get rid of it
2650	 * and not suck up any kernel memory in that.
2651	 */
2652	if (stcb->asoc.in_asocid_hash) {
2653		struct sctpasochead *lhd;
2654
2655		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2656		    new_inp->hashasocidmark)];
2657		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2658	}
2659	/* Ok. Let's restart timer. */
2660	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2661		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2662		    stcb, net);
2663	}
2664
2665	SCTP_INP_INFO_WUNLOCK();
2666	if (new_inp->sctp_tcbhash != NULL) {
2667		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2668		new_inp->sctp_tcbhash = NULL;
2669	}
2670	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2671		/* Subset bound, so copy in the laddr list from the old_inp */
2672		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2673			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2674			if (laddr == NULL) {
2675				/*
2676				 * Gak, what can we do? This assoc is really
2677				 * HOSED. We probably should send an abort
2678				 * here.
2679				 */
2680				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2681				continue;
2682			}
2683			SCTP_INCR_LADDR_COUNT();
2684			bzero(laddr, sizeof(*laddr));
2685			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2686			laddr->ifa = oladdr->ifa;
2687			atomic_add_int(&laddr->ifa->refcount, 1);
2688			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2689			    sctp_nxt_addr);
2690			new_inp->laddr_count++;
2691			if (oladdr == stcb->asoc.last_used_address) {
2692				stcb->asoc.last_used_address = laddr;
2693			}
2694		}
2695	}
2696	/*
2697	 * Now any running timers need to be adjusted since we really don't
2698	 * care if they are running or not just blast in the new_inp into
2699	 * all of them.
2700	 */
2701
2702	stcb->asoc.dack_timer.ep = (void *)new_inp;
2703	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2704	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2705	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2706	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2707	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2708	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2709	/* now what about the nets? */
2710	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2711		net->pmtu_timer.ep = (void *)new_inp;
2712		net->hb_timer.ep = (void *)new_inp;
2713		net->rxt_timer.ep = (void *)new_inp;
2714	}
2715	SCTP_INP_WUNLOCK(new_inp);
2716	SCTP_INP_WUNLOCK(old_inp);
2717}
2718
2719
2720
2721
2722/* sctp_ifap is used to bypass normal local address validation checks */
2723int
2724sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2725    struct sctp_ifa *sctp_ifap, struct thread *p)
2726{
2727	/* bind a ep to a socket address */
2728	struct sctppcbhead *head;
2729	struct sctp_inpcb *inp, *inp_tmp;
2730	struct inpcb *ip_inp;
2731	int port_reuse_active = 0;
2732	int bindall;
2733	uint16_t lport;
2734	int error;
2735	uint32_t vrf_id;
2736
2737	lport = 0;
2738	error = 0;
2739	bindall = 1;
2740	inp = (struct sctp_inpcb *)so->so_pcb;
2741	ip_inp = (struct inpcb *)so->so_pcb;
2742#ifdef SCTP_DEBUG
2743	if (addr) {
2744		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2745		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2746		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2747		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2748	}
2749#endif
2750	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2751		/* already did a bind, subsequent binds NOT allowed ! */
2752		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2753		return (EINVAL);
2754	}
2755#ifdef INVARIANTS
2756	if (p == NULL)
2757		panic("null proc/thread");
2758#endif
2759	if (addr != NULL) {
2760		switch (addr->sa_family) {
2761#ifdef INET
2762		case AF_INET:
2763			{
2764				struct sockaddr_in *sin;
2765
2766				/* IPV6_V6ONLY socket? */
2767				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2768					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2769					return (EINVAL);
2770				}
2771				if (addr->sa_len != sizeof(*sin)) {
2772					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2773					return (EINVAL);
2774				}
2775				sin = (struct sockaddr_in *)addr;
2776				lport = sin->sin_port;
2777				/*
2778				 * For LOOPBACK the prison_local_ip4() call
2779				 * will transmute the ip address to the
2780				 * proper value.
2781				 */
2782				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2783					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2784					return (error);
2785				}
2786				if (sin->sin_addr.s_addr != INADDR_ANY) {
2787					bindall = 0;
2788				}
2789				break;
2790			}
2791#endif
2792#ifdef INET6
2793		case AF_INET6:
2794			{
2795				/*
2796				 * Only for pure IPv6 Address. (No IPv4
2797				 * Mapped!)
2798				 */
2799				struct sockaddr_in6 *sin6;
2800
2801				sin6 = (struct sockaddr_in6 *)addr;
2802
2803				if (addr->sa_len != sizeof(*sin6)) {
2804					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2805					return (EINVAL);
2806				}
2807				lport = sin6->sin6_port;
2808				/*
2809				 * For LOOPBACK the prison_local_ip6() call
2810				 * will transmute the ipv6 address to the
2811				 * proper value.
2812				 */
2813				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2814				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2815					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2816					return (error);
2817				}
2818				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2819					bindall = 0;
2820					/* KAME hack: embed scopeid */
2821					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2822						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2823						return (EINVAL);
2824					}
2825				}
2826				/* this must be cleared for ifa_ifwithaddr() */
2827				sin6->sin6_scope_id = 0;
2828				break;
2829			}
2830#endif
2831		default:
2832			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2833			return (EAFNOSUPPORT);
2834		}
2835	}
2836	SCTP_INP_INFO_WLOCK();
2837	SCTP_INP_WLOCK(inp);
2838	/* Setup a vrf_id to be the default for the non-bind-all case. */
2839	vrf_id = inp->def_vrf_id;
2840
2841	/* increase our count due to the unlock we do */
2842	SCTP_INP_INCR_REF(inp);
2843	if (lport) {
2844		/*
2845		 * Did the caller specify a port? if so we must see if an ep
2846		 * already has this one bound.
2847		 */
2848		/* got to be root to get at low ports */
2849		if (ntohs(lport) < IPPORT_RESERVED) {
2850			if (p && (error =
2851			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2852			    )) {
2853				SCTP_INP_DECR_REF(inp);
2854				SCTP_INP_WUNLOCK(inp);
2855				SCTP_INP_INFO_WUNLOCK();
2856				return (error);
2857			}
2858		}
2859		if (p == NULL) {
2860			SCTP_INP_DECR_REF(inp);
2861			SCTP_INP_WUNLOCK(inp);
2862			SCTP_INP_INFO_WUNLOCK();
2863			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2864			return (error);
2865		}
2866		SCTP_INP_WUNLOCK(inp);
2867		if (bindall) {
2868			vrf_id = inp->def_vrf_id;
2869			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2870			if (inp_tmp != NULL) {
2871				/*
2872				 * lock guy returned and lower count note
2873				 * that we are not bound so inp_tmp should
2874				 * NEVER be inp. And it is this inp
2875				 * (inp_tmp) that gets the reference bump,
2876				 * so we must lower it.
2877				 */
2878				SCTP_INP_DECR_REF(inp_tmp);
2879				/* unlock info */
2880				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2881				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2882					/*
2883					 * Ok, must be one-2-one and
2884					 * allowing port re-use
2885					 */
2886					port_reuse_active = 1;
2887					goto continue_anyway;
2888				}
2889				SCTP_INP_DECR_REF(inp);
2890				SCTP_INP_INFO_WUNLOCK();
2891				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2892				return (EADDRINUSE);
2893			}
2894		} else {
2895			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2896			if (inp_tmp != NULL) {
2897				/*
2898				 * lock guy returned and lower count note
2899				 * that we are not bound so inp_tmp should
2900				 * NEVER be inp. And it is this inp
2901				 * (inp_tmp) that gets the reference bump,
2902				 * so we must lower it.
2903				 */
2904				SCTP_INP_DECR_REF(inp_tmp);
2905				/* unlock info */
2906				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2907				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2908					/*
2909					 * Ok, must be one-2-one and
2910					 * allowing port re-use
2911					 */
2912					port_reuse_active = 1;
2913					goto continue_anyway;
2914				}
2915				SCTP_INP_DECR_REF(inp);
2916				SCTP_INP_INFO_WUNLOCK();
2917				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2918				return (EADDRINUSE);
2919			}
2920		}
2921continue_anyway:
2922		SCTP_INP_WLOCK(inp);
2923		if (bindall) {
2924			/* verify that no lport is not used by a singleton */
2925			if ((port_reuse_active == 0) &&
2926			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
2927				/* Sorry someone already has this one bound */
2928				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2929				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2930					port_reuse_active = 1;
2931				} else {
2932					SCTP_INP_DECR_REF(inp);
2933					SCTP_INP_WUNLOCK(inp);
2934					SCTP_INP_INFO_WUNLOCK();
2935					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2936					return (EADDRINUSE);
2937				}
2938			}
2939		}
2940	} else {
2941		uint16_t first, last, candidate;
2942		uint16_t count;
2943		int done;
2944
2945		if (ip_inp->inp_flags & INP_HIGHPORT) {
2946			first = MODULE_GLOBAL(ipport_hifirstauto);
2947			last = MODULE_GLOBAL(ipport_hilastauto);
2948		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2949			if (p && (error =
2950			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2951			    )) {
2952				SCTP_INP_DECR_REF(inp);
2953				SCTP_INP_WUNLOCK(inp);
2954				SCTP_INP_INFO_WUNLOCK();
2955				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2956				return (error);
2957			}
2958			first = MODULE_GLOBAL(ipport_lowfirstauto);
2959			last = MODULE_GLOBAL(ipport_lowlastauto);
2960		} else {
2961			first = MODULE_GLOBAL(ipport_firstauto);
2962			last = MODULE_GLOBAL(ipport_lastauto);
2963		}
2964		if (first > last) {
2965			uint16_t temp;
2966
2967			temp = first;
2968			first = last;
2969			last = temp;
2970		}
2971		count = last - first + 1;	/* number of candidates */
2972		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2973
2974		done = 0;
2975		while (!done) {
2976			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
2977				done = 1;
2978			}
2979			if (!done) {
2980				if (--count == 0) {
2981					SCTP_INP_DECR_REF(inp);
2982					SCTP_INP_WUNLOCK(inp);
2983					SCTP_INP_INFO_WUNLOCK();
2984					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2985					return (EADDRINUSE);
2986				}
2987				if (candidate == last)
2988					candidate = first;
2989				else
2990					candidate = candidate + 1;
2991			}
2992		}
2993		lport = htons(candidate);
2994	}
2995	SCTP_INP_DECR_REF(inp);
2996	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2997	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2998		/*
2999		 * this really should not happen. The guy did a non-blocking
3000		 * bind and then did a close at the same time.
3001		 */
3002		SCTP_INP_WUNLOCK(inp);
3003		SCTP_INP_INFO_WUNLOCK();
3004		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3005		return (EINVAL);
3006	}
3007	/* ok we look clear to give out this port, so lets setup the binding */
3008	if (bindall) {
3009		/* binding to all addresses, so just set in the proper flags */
3010		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3011		/* set the automatic addr changes from kernel flag */
3012		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3013			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3014			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3015		} else {
3016			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3017			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3018		}
3019		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3020			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3021		} else {
3022			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3023		}
3024		/*
3025		 * set the automatic mobility_base from kernel flag (by
3026		 * micchie)
3027		 */
3028		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3029			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3030			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3031		} else {
3032			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3033			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3034		}
3035		/*
3036		 * set the automatic mobility_fasthandoff from kernel flag
3037		 * (by micchie)
3038		 */
3039		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3040			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3041			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3042		} else {
3043			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3044			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3045		}
3046	} else {
3047		/*
3048		 * bind specific, make sure flags is off and add a new
3049		 * address structure to the sctp_addr_list inside the ep
3050		 * structure.
3051		 *
3052		 * We will need to allocate one and insert it at the head. The
3053		 * socketopt call can just insert new addresses in there as
3054		 * well. It will also have to do the embed scope kame hack
3055		 * too (before adding).
3056		 */
3057		struct sctp_ifa *ifa;
3058		struct sockaddr_storage store_sa;
3059
3060		memset(&store_sa, 0, sizeof(store_sa));
3061		switch (addr->sa_family) {
3062#ifdef INET
3063		case AF_INET:
3064			{
3065				struct sockaddr_in *sin;
3066
3067				sin = (struct sockaddr_in *)&store_sa;
3068				memcpy(sin, addr, sizeof(struct sockaddr_in));
3069				sin->sin_port = 0;
3070				break;
3071			}
3072#endif
3073#ifdef INET6
3074		case AF_INET6:
3075			{
3076				struct sockaddr_in6 *sin6;
3077
3078				sin6 = (struct sockaddr_in6 *)&store_sa;
3079				memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3080				sin6->sin6_port = 0;
3081				break;
3082			}
3083#endif
3084		default:
3085			break;
3086		}
3087		/*
3088		 * first find the interface with the bound address need to
3089		 * zero out the port to find the address! yuck! can't do
3090		 * this earlier since need port for sctp_pcb_findep()
3091		 */
3092		if (sctp_ifap != NULL) {
3093			ifa = sctp_ifap;
3094		} else {
3095			/*
3096			 * Note for BSD we hit here always other O/S's will
3097			 * pass things in via the sctp_ifap argument
3098			 * (Panda).
3099			 */
3100			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3101			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3102		}
3103		if (ifa == NULL) {
3104			/* Can't find an interface with that address */
3105			SCTP_INP_WUNLOCK(inp);
3106			SCTP_INP_INFO_WUNLOCK();
3107			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3108			return (EADDRNOTAVAIL);
3109		}
3110#ifdef INET6
3111		if (addr->sa_family == AF_INET6) {
3112			/* GAK, more FIXME IFA lock? */
3113			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3114				/* Can't bind a non-existent addr. */
3115				SCTP_INP_WUNLOCK(inp);
3116				SCTP_INP_INFO_WUNLOCK();
3117				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3118				return (EINVAL);
3119			}
3120		}
3121#endif
3122		/* we're not bound all */
3123		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3124		/* allow bindx() to send ASCONF's for binding changes */
3125		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3126		/* clear automatic addr changes from kernel flag */
3127		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3128
3129		/* add this address to the endpoint list */
3130		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3131		if (error != 0) {
3132			SCTP_INP_WUNLOCK(inp);
3133			SCTP_INP_INFO_WUNLOCK();
3134			return (error);
3135		}
3136		inp->laddr_count++;
3137	}
3138	/* find the bucket */
3139	if (port_reuse_active) {
3140		/* Put it into tcp 1-2-1 hash */
3141		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3142		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3143	} else {
3144		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3145	}
3146	/* put it in the bucket */
3147	LIST_INSERT_HEAD(head, inp, sctp_hash);
3148	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3149	    (void *)head, ntohs(lport), port_reuse_active);
3150	/* set in the port */
3151	inp->sctp_lport = lport;
3152
3153	/* turn off just the unbound flag */
3154	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3155	SCTP_INP_WUNLOCK(inp);
3156	SCTP_INP_INFO_WUNLOCK();
3157	return (0);
3158}
3159
3160
3161static void
3162sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3163{
3164	struct sctp_iterator *it, *nit;
3165
3166	/*
3167	 * We enter with the only the ITERATOR_LOCK in place and a write
3168	 * lock on the inp_info stuff.
3169	 */
3170	it = sctp_it_ctl.cur_it;
3171	if (it && (it->vn != curvnet)) {
3172		/* Its not looking at our VNET */
3173		return;
3174	}
3175	if (it && (it->inp == inp)) {
3176		/*
3177		 * This is tricky and we hold the iterator lock, but when it
3178		 * returns and gets the lock (when we release it) the
3179		 * iterator will try to operate on inp. We need to stop that
3180		 * from happening. But of course the iterator has a
3181		 * reference on the stcb and inp. We can mark it and it will
3182		 * stop.
3183		 *
3184		 * If its a single iterator situation, we set the end iterator
3185		 * flag. Otherwise we set the iterator to go to the next
3186		 * inp.
3187		 *
3188		 */
3189		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3190			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3191		} else {
3192			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3193		}
3194	}
3195	/*
3196	 * Now go through and remove any single reference to our inp that
3197	 * may be still pending on the list
3198	 */
3199	SCTP_IPI_ITERATOR_WQ_LOCK();
3200	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3201		if (it->vn != curvnet) {
3202			continue;
3203		}
3204		if (it->inp == inp) {
3205			/* This one points to me is it inp specific? */
3206			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3207				/* Remove and free this one */
3208				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3209				    it, sctp_nxt_itr);
3210				if (it->function_atend != NULL) {
3211					(*it->function_atend) (it->pointer, it->val);
3212				}
3213				SCTP_FREE(it, SCTP_M_ITER);
3214			} else {
3215				it->inp = LIST_NEXT(it->inp, sctp_list);
3216				if (it->inp) {
3217					SCTP_INP_INCR_REF(it->inp);
3218				}
3219			}
3220			/*
3221			 * When its put in the refcnt is incremented so decr
3222			 * it
3223			 */
3224			SCTP_INP_DECR_REF(inp);
3225		}
3226	}
3227	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3228}
3229
3230/* release sctp_inpcb unbind the port */
3231void
3232sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3233{
3234	/*
3235	 * Here we free a endpoint. We must find it (if it is in the Hash
3236	 * table) and remove it from there. Then we must also find it in the
3237	 * overall list and remove it from there. After all removals are
3238	 * complete then any timer has to be stopped. Then start the actual
3239	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3240	 * all associations. d) finally the ep itself.
3241	 */
3242	struct sctp_tcb *asoc, *nasoc;
3243	struct sctp_laddr *laddr, *nladdr;
3244	struct inpcb *ip_pcb;
3245	struct socket *so;
3246	int being_refed = 0;
3247	struct sctp_queued_to_read *sq, *nsq;
3248	int cnt;
3249	sctp_sharedkey_t *shared_key, *nshared_key;
3250
3251
3252#ifdef SCTP_LOG_CLOSING
3253	sctp_log_closing(inp, NULL, 0);
3254#endif
3255	SCTP_ITERATOR_LOCK();
3256	/* mark any iterators on the list or being processed */
3257	sctp_iterator_inp_being_freed(inp);
3258	SCTP_ITERATOR_UNLOCK();
3259	so = inp->sctp_socket;
3260	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3261		/* been here before.. eeks.. get out of here */
3262		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3263#ifdef SCTP_LOG_CLOSING
3264		sctp_log_closing(inp, NULL, 1);
3265#endif
3266		return;
3267	}
3268	SCTP_ASOC_CREATE_LOCK(inp);
3269	SCTP_INP_INFO_WLOCK();
3270
3271	SCTP_INP_WLOCK(inp);
3272	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3273		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3274		/* socket is gone, so no more wakeups allowed */
3275		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3276		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3277		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3278
3279	}
3280	/* First time through we have the socket lock, after that no more. */
3281	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3282	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3283
3284	if (inp->control) {
3285		sctp_m_freem(inp->control);
3286		inp->control = NULL;
3287	}
3288	if (inp->pkt) {
3289		sctp_m_freem(inp->pkt);
3290		inp->pkt = NULL;
3291	}
3292	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3293					 * here but I will be nice :> (i.e.
3294					 * ip_pcb = ep;) */
3295	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3296		int cnt_in_sd;
3297
3298		cnt_in_sd = 0;
3299		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3300			SCTP_TCB_LOCK(asoc);
3301			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3302				/* Skip guys being freed */
3303				cnt_in_sd++;
3304				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3305					/*
3306					 * Special case - we did not start a
3307					 * kill timer on the asoc due to it
3308					 * was not closed. So go ahead and
3309					 * start it now.
3310					 */
3311					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3312					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3313				}
3314				SCTP_TCB_UNLOCK(asoc);
3315				continue;
3316			}
3317			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3318			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3319			    (asoc->asoc.total_output_queue_size == 0)) {
3320				/*
3321				 * If we have data in queue, we don't want
3322				 * to just free since the app may have done,
3323				 * send()/close or connect/send/close. And
3324				 * it wants the data to get across first.
3325				 */
3326				/* Just abandon things in the front states */
3327				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3328				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3329					cnt_in_sd++;
3330				}
3331				continue;
3332			}
3333			/* Disconnect the socket please */
3334			asoc->sctp_socket = NULL;
3335			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3336			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3337			    (asoc->asoc.control_pdapi) ||
3338			    (asoc->asoc.size_on_all_streams > 0) ||
3339			    (so && (so->so_rcv.sb_cc > 0))) {
3340				/* Left with Data unread */
3341				struct mbuf *op_err;
3342
3343				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3344				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3345				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3346				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3347				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3348				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3349					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3350				}
3351				if (sctp_free_assoc(inp, asoc,
3352				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3353					cnt_in_sd++;
3354				}
3355				continue;
3356			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3357				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3358			    (asoc->asoc.stream_queue_cnt == 0)) {
3359				if (asoc->asoc.locked_on_sending) {
3360					goto abort_anyway;
3361				}
3362				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3363				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3364					struct sctp_nets *netp;
3365
3366					/*
3367					 * there is nothing queued to send,
3368					 * so I send shutdown
3369					 */
3370					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3371					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3372						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3373					}
3374					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3375					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3376					sctp_stop_timers_for_shutdown(asoc);
3377					if (asoc->asoc.alternate) {
3378						netp = asoc->asoc.alternate;
3379					} else {
3380						netp = asoc->asoc.primary_destination;
3381					}
3382					sctp_send_shutdown(asoc, netp);
3383					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3384					    netp);
3385					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3386					    asoc->asoc.primary_destination);
3387					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3388				}
3389			} else {
3390				/* mark into shutdown pending */
3391				struct sctp_stream_queue_pending *sp;
3392
3393				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3394				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3395				    asoc->asoc.primary_destination);
3396				if (asoc->asoc.locked_on_sending) {
3397					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3398					    sctp_streamhead);
3399					if (sp == NULL) {
3400						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3401						    (void *)asoc->asoc.locked_on_sending,
3402						    asoc->asoc.locked_on_sending->stream_no);
3403					} else {
3404						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3405							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3406					}
3407				}
3408				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3409				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3410				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3411					struct mbuf *op_err;
3412
3413			abort_anyway:
3414					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3415					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3416					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3417					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3418					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3419					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3420						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3421					}
3422					if (sctp_free_assoc(inp, asoc,
3423					    SCTP_PCBFREE_NOFORCE,
3424					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3425						cnt_in_sd++;
3426					}
3427					continue;
3428				} else {
3429					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3430				}
3431			}
3432			cnt_in_sd++;
3433			SCTP_TCB_UNLOCK(asoc);
3434		}
3435		/* now is there some left in our SHUTDOWN state? */
3436		if (cnt_in_sd) {
3437#ifdef SCTP_LOG_CLOSING
3438			sctp_log_closing(inp, NULL, 2);
3439#endif
3440			inp->sctp_socket = NULL;
3441			SCTP_INP_WUNLOCK(inp);
3442			SCTP_ASOC_CREATE_UNLOCK(inp);
3443			SCTP_INP_INFO_WUNLOCK();
3444			return;
3445		}
3446	}
3447	inp->sctp_socket = NULL;
3448	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3449	    SCTP_PCB_FLAGS_UNBOUND) {
3450		/*
3451		 * ok, this guy has been bound. It's port is somewhere in
3452		 * the SCTP_BASE_INFO(hash table). Remove it!
3453		 */
3454		LIST_REMOVE(inp, sctp_hash);
3455		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3456	}
3457	/*
3458	 * If there is a timer running to kill us, forget it, since it may
3459	 * have a contest on the INP lock.. which would cause us to die ...
3460	 */
3461	cnt = 0;
3462	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3463		SCTP_TCB_LOCK(asoc);
3464		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3465			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3466				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3467				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3468			}
3469			cnt++;
3470			SCTP_TCB_UNLOCK(asoc);
3471			continue;
3472		}
3473		/* Free associations that are NOT killing us */
3474		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3475		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3476			struct mbuf *op_err;
3477
3478			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3479			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3480			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3481			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3482		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3483			cnt++;
3484			SCTP_TCB_UNLOCK(asoc);
3485			continue;
3486		}
3487		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3488		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3489			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3490		}
3491		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3492			cnt++;
3493		}
3494	}
3495	if (cnt) {
3496		/* Ok we have someone out there that will kill us */
3497		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3498#ifdef SCTP_LOG_CLOSING
3499		sctp_log_closing(inp, NULL, 3);
3500#endif
3501		SCTP_INP_WUNLOCK(inp);
3502		SCTP_ASOC_CREATE_UNLOCK(inp);
3503		SCTP_INP_INFO_WUNLOCK();
3504		return;
3505	}
3506	if (SCTP_INP_LOCK_CONTENDED(inp))
3507		being_refed++;
3508	if (SCTP_INP_READ_CONTENDED(inp))
3509		being_refed++;
3510	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3511		being_refed++;
3512
3513	if ((inp->refcount) ||
3514	    (being_refed) ||
3515	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3516		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3517#ifdef SCTP_LOG_CLOSING
3518		sctp_log_closing(inp, NULL, 4);
3519#endif
3520		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3521		SCTP_INP_WUNLOCK(inp);
3522		SCTP_ASOC_CREATE_UNLOCK(inp);
3523		SCTP_INP_INFO_WUNLOCK();
3524		return;
3525	}
3526	inp->sctp_ep.signature_change.type = 0;
3527	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3528	/*
3529	 * Remove it from the list .. last thing we need a lock for.
3530	 */
3531	LIST_REMOVE(inp, sctp_list);
3532	SCTP_INP_WUNLOCK(inp);
3533	SCTP_ASOC_CREATE_UNLOCK(inp);
3534	SCTP_INP_INFO_WUNLOCK();
3535	/*
3536	 * Now we release all locks. Since this INP cannot be found anymore
3537	 * except possibly by the kill timer that might be running. We call
3538	 * the drain function here. It should hit the case were it sees the
3539	 * ACTIVE flag cleared and exit out freeing us to proceed and
3540	 * destroy everything.
3541	 */
3542	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3543		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3544	} else {
3545		/* Probably un-needed */
3546		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3547	}
3548
3549#ifdef SCTP_LOG_CLOSING
3550	sctp_log_closing(inp, NULL, 5);
3551#endif
3552
3553
3554	if ((inp->sctp_asocidhash) != NULL) {
3555		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3556		inp->sctp_asocidhash = NULL;
3557	}
3558	/* sa_ignore FREED_MEMORY */
3559	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3560		/* Its only abandoned if it had data left */
3561		if (sq->length)
3562			SCTP_STAT_INCR(sctps_left_abandon);
3563
3564		TAILQ_REMOVE(&inp->read_queue, sq, next);
3565		sctp_free_remote_addr(sq->whoFrom);
3566		if (so)
3567			so->so_rcv.sb_cc -= sq->length;
3568		if (sq->data) {
3569			sctp_m_freem(sq->data);
3570			sq->data = NULL;
3571		}
3572		/*
3573		 * no need to free the net count, since at this point all
3574		 * assoc's are gone.
3575		 */
3576		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3577		SCTP_DECR_READQ_COUNT();
3578	}
3579	/* Now the sctp_pcb things */
3580	/*
3581	 * free each asoc if it is not already closed/free. we can't use the
3582	 * macro here since le_next will get freed as part of the
3583	 * sctp_free_assoc() call.
3584	 */
3585	if (so) {
3586#ifdef IPSEC
3587		ipsec_delete_pcbpolicy(ip_pcb);
3588#endif				/* IPSEC */
3589
3590		/* Unlocks not needed since the socket is gone now */
3591	}
3592	if (ip_pcb->inp_options) {
3593		(void)sctp_m_free(ip_pcb->inp_options);
3594		ip_pcb->inp_options = 0;
3595	}
3596#ifdef INET6
3597	if (ip_pcb->inp_vflag & INP_IPV6) {
3598		struct in6pcb *in6p;
3599
3600		in6p = (struct in6pcb *)inp;
3601		ip6_freepcbopts(in6p->in6p_outputopts);
3602	}
3603#endif				/* INET6 */
3604	ip_pcb->inp_vflag = 0;
3605	/* free up authentication fields */
3606	if (inp->sctp_ep.local_auth_chunks != NULL)
3607		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3608	if (inp->sctp_ep.local_hmacs != NULL)
3609		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3610
3611	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3612		LIST_REMOVE(shared_key, next);
3613		sctp_free_sharedkey(shared_key);
3614		/* sa_ignore FREED_MEMORY */
3615	}
3616
3617	/*
3618	 * if we have an address list the following will free the list of
3619	 * ifaddr's that are set into this ep. Again macro limitations here,
3620	 * since the LIST_FOREACH could be a bad idea.
3621	 */
3622	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3623		sctp_remove_laddr(laddr);
3624	}
3625
3626#ifdef SCTP_TRACK_FREED_ASOCS
3627	/* TEMP CODE */
3628	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3629		LIST_REMOVE(asoc, sctp_tcblist);
3630		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3631		SCTP_DECR_ASOC_COUNT();
3632	}
3633	/* *** END TEMP CODE *** */
3634#endif
3635	/* Now lets see about freeing the EP hash table. */
3636	if (inp->sctp_tcbhash != NULL) {
3637		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3638		inp->sctp_tcbhash = NULL;
3639	}
3640	/* Now we must put the ep memory back into the zone pool */
3641	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3642	SCTP_INP_LOCK_DESTROY(inp);
3643	SCTP_INP_READ_DESTROY(inp);
3644	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3645	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3646	SCTP_DECR_EP_COUNT();
3647}
3648
3649
3650struct sctp_nets *
3651sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3652{
3653	struct sctp_nets *net;
3654
3655	/* locate the address */
3656	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3657		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3658			return (net);
3659	}
3660	return (NULL);
3661}
3662
3663
3664int
3665sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3666{
3667	struct sctp_ifa *sctp_ifa;
3668
3669	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3670	if (sctp_ifa) {
3671		return (1);
3672	} else {
3673		return (0);
3674	}
3675}
3676
3677/*
3678 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3679 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3680 * stats of stuff.
3681 */
3682int
3683sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3684    struct sctp_nets **netp, int set_scope, int from)
3685{
3686	/*
3687	 * The following is redundant to the same lines in the
3688	 * sctp_aloc_assoc() but is needed since others call the add address
3689	 * function
3690	 */
3691	struct sctp_nets *net, *netfirst;
3692	int addr_inscope;
3693
3694	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3695	    from);
3696	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3697
3698	netfirst = sctp_findnet(stcb, newaddr);
3699	if (netfirst) {
3700		/*
3701		 * Lie and return ok, we don't want to make the association
3702		 * go away for this behavior. It will happen in the TCP
3703		 * model in a connected socket. It does not reach the hash
3704		 * table until after the association is built so it can't be
3705		 * found. Mark as reachable, since the initial creation will
3706		 * have been cleared and the NOT_IN_ASSOC flag will have
3707		 * been added... and we don't want to end up removing it
3708		 * back out.
3709		 */
3710		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3711			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3712			    SCTP_ADDR_UNCONFIRMED);
3713		} else {
3714			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3715		}
3716
3717		return (0);
3718	}
3719	addr_inscope = 1;
3720	switch (newaddr->sa_family) {
3721#ifdef INET
3722	case AF_INET:
3723		{
3724			struct sockaddr_in *sin;
3725
3726			sin = (struct sockaddr_in *)newaddr;
3727			if (sin->sin_addr.s_addr == 0) {
3728				/* Invalid address */
3729				return (-1);
3730			}
3731			/* zero out the bzero area */
3732			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3733
3734			/* assure len is set */
3735			sin->sin_len = sizeof(struct sockaddr_in);
3736			if (set_scope) {
3737#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3738				stcb->asoc.scope.ipv4_local_scope = 1;
3739#else
3740				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3741					stcb->asoc.scope.ipv4_local_scope = 1;
3742				}
3743#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3744			} else {
3745				/* Validate the address is in scope */
3746				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3747				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3748					addr_inscope = 0;
3749				}
3750			}
3751			break;
3752		}
3753#endif
3754#ifdef INET6
3755	case AF_INET6:
3756		{
3757			struct sockaddr_in6 *sin6;
3758
3759			sin6 = (struct sockaddr_in6 *)newaddr;
3760			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3761				/* Invalid address */
3762				return (-1);
3763			}
3764			/* assure len is set */
3765			sin6->sin6_len = sizeof(struct sockaddr_in6);
3766			if (set_scope) {
3767				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3768					stcb->asoc.scope.loopback_scope = 1;
3769					stcb->asoc.scope.local_scope = 0;
3770					stcb->asoc.scope.ipv4_local_scope = 1;
3771					stcb->asoc.scope.site_scope = 1;
3772				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3773					/*
3774					 * If the new destination is a
3775					 * LINK_LOCAL we must have common
3776					 * site scope. Don't set the local
3777					 * scope since we may not share all
3778					 * links, only loopback can do this.
3779					 * Links on the local network would
3780					 * also be on our private network
3781					 * for v4 too.
3782					 */
3783					stcb->asoc.scope.ipv4_local_scope = 1;
3784					stcb->asoc.scope.site_scope = 1;
3785				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3786					/*
3787					 * If the new destination is
3788					 * SITE_LOCAL then we must have site
3789					 * scope in common.
3790					 */
3791					stcb->asoc.scope.site_scope = 1;
3792				}
3793			} else {
3794				/* Validate the address is in scope */
3795				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3796				    (stcb->asoc.scope.loopback_scope == 0)) {
3797					addr_inscope = 0;
3798				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3799				    (stcb->asoc.scope.local_scope == 0)) {
3800					addr_inscope = 0;
3801				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3802				    (stcb->asoc.scope.site_scope == 0)) {
3803					addr_inscope = 0;
3804				}
3805			}
3806			break;
3807		}
3808#endif
3809	default:
3810		/* not supported family type */
3811		return (-1);
3812	}
3813	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3814	if (net == NULL) {
3815		return (-1);
3816	}
3817	SCTP_INCR_RADDR_COUNT();
3818	bzero(net, sizeof(struct sctp_nets));
3819	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3820	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3821	switch (newaddr->sa_family) {
3822#ifdef INET
3823	case AF_INET:
3824		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3825		break;
3826#endif
3827#ifdef INET6
3828	case AF_INET6:
3829		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3830		break;
3831#endif
3832	default:
3833		break;
3834	}
3835	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3836	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3837		stcb->asoc.scope.loopback_scope = 1;
3838		stcb->asoc.scope.ipv4_local_scope = 1;
3839		stcb->asoc.scope.local_scope = 0;
3840		stcb->asoc.scope.site_scope = 1;
3841		addr_inscope = 1;
3842	}
3843	net->failure_threshold = stcb->asoc.def_net_failure;
3844	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3845	if (addr_inscope == 0) {
3846		net->dest_state = (SCTP_ADDR_REACHABLE |
3847		    SCTP_ADDR_OUT_OF_SCOPE);
3848	} else {
3849		if (from == SCTP_ADDR_IS_CONFIRMED)
3850			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3851			net->dest_state = SCTP_ADDR_REACHABLE;
3852		else
3853			net->dest_state = SCTP_ADDR_REACHABLE |
3854			    SCTP_ADDR_UNCONFIRMED;
3855	}
3856	/*
3857	 * We set this to 0, the timer code knows that this means its an
3858	 * initial value
3859	 */
3860	net->rto_needed = 1;
3861	net->RTO = 0;
3862	net->RTO_measured = 0;
3863	stcb->asoc.numnets++;
3864	net->ref_count = 1;
3865	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3866	net->port = stcb->asoc.port;
3867	net->dscp = stcb->asoc.default_dscp;
3868#ifdef INET6
3869	net->flowlabel = stcb->asoc.default_flowlabel;
3870#endif
3871	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3872		net->dest_state |= SCTP_ADDR_NOHB;
3873	} else {
3874		net->dest_state &= ~SCTP_ADDR_NOHB;
3875	}
3876	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3877		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3878	} else {
3879		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3880	}
3881	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3882	/* Init the timer structure */
3883	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3884	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3885	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3886
3887	/* Now generate a route for this guy */
3888#ifdef INET6
3889	/* KAME hack: embed scopeid */
3890	if (newaddr->sa_family == AF_INET6) {
3891		struct sockaddr_in6 *sin6;
3892
3893		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3894		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3895		sin6->sin6_scope_id = 0;
3896	}
3897#endif
3898	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3899
3900	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3901		/* Get source address */
3902		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3903		    stcb,
3904		    (sctp_route_t *) & net->ro,
3905		    net,
3906		    0,
3907		    stcb->asoc.vrf_id);
3908		/* Now get the interface MTU */
3909		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3910			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3911		}
3912		if (net->mtu > 0) {
3913			uint32_t rmtu;
3914
3915			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3916			if (rmtu == 0) {
3917				/*
3918				 * Start things off to match mtu of
3919				 * interface please.
3920				 */
3921				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3922				    net->ro.ro_rt, net->mtu);
3923			} else {
3924				/*
3925				 * we take the route mtu over the interface,
3926				 * since the route may be leading out the
3927				 * loopback, or a different interface.
3928				 */
3929				net->mtu = rmtu;
3930			}
3931		}
3932	}
3933	if (net->mtu == 0) {
3934		switch (newaddr->sa_family) {
3935#ifdef INET
3936		case AF_INET:
3937			net->mtu = SCTP_DEFAULT_MTU;
3938			break;
3939#endif
3940#ifdef INET6
3941		case AF_INET6:
3942			net->mtu = 1280;
3943			break;
3944#endif
3945		default:
3946			break;
3947		}
3948	}
3949	if (net->port) {
3950		net->mtu -= (uint32_t) sizeof(struct udphdr);
3951	}
3952	if (from == SCTP_ALLOC_ASOC) {
3953		stcb->asoc.smallest_mtu = net->mtu;
3954	}
3955	if (stcb->asoc.smallest_mtu > net->mtu) {
3956		stcb->asoc.smallest_mtu = net->mtu;
3957	}
3958#ifdef INET6
3959	if (newaddr->sa_family == AF_INET6) {
3960		struct sockaddr_in6 *sin6;
3961
3962		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3963		(void)sa6_recoverscope(sin6);
3964	}
3965#endif
3966
3967	/* JRS - Use the congestion control given in the CC module */
3968	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
3969		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
3970
3971	/*
3972	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3973	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3974	 */
3975	net->find_pseudo_cumack = 1;
3976	net->find_rtx_pseudo_cumack = 1;
3977	net->src_addr_selected = 0;
3978	/* Choose an initial flowid. */
3979	net->flowid = stcb->asoc.my_vtag ^
3980	    ntohs(stcb->rport) ^
3981	    ntohs(stcb->sctp_ep->sctp_lport);
3982#ifdef INVARIANTS
3983	net->flowidset = 1;
3984#endif
3985	if (netp) {
3986		*netp = net;
3987	}
3988	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3989	if (net->ro.ro_rt == NULL) {
3990		/* Since we have no route put it at the back */
3991		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3992	} else if (netfirst == NULL) {
3993		/* We are the first one in the pool. */
3994		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3995	} else if (netfirst->ro.ro_rt == NULL) {
3996		/*
3997		 * First one has NO route. Place this one ahead of the first
3998		 * one.
3999		 */
4000		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4001	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4002		/*
4003		 * This one has a different interface than the one at the
4004		 * top of the list. Place it ahead.
4005		 */
4006		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4007	} else {
4008		/*
4009		 * Ok we have the same interface as the first one. Move
4010		 * forward until we find either a) one with a NULL route...
4011		 * insert ahead of that b) one with a different ifp.. insert
4012		 * after that. c) end of the list.. insert at the tail.
4013		 */
4014		struct sctp_nets *netlook;
4015
4016		do {
4017			netlook = TAILQ_NEXT(netfirst, sctp_next);
4018			if (netlook == NULL) {
4019				/* End of the list */
4020				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4021				break;
4022			} else if (netlook->ro.ro_rt == NULL) {
4023				/* next one has NO route */
4024				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4025				break;
4026			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4027				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4028				    net, sctp_next);
4029				break;
4030			}
4031			/* Shift forward */
4032			netfirst = netlook;
4033		} while (netlook != NULL);
4034	}
4035
4036	/* got to have a primary set */
4037	if (stcb->asoc.primary_destination == 0) {
4038		stcb->asoc.primary_destination = net;
4039	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4040		    (net->ro.ro_rt) &&
4041	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4042		/* No route to current primary adopt new primary */
4043		stcb->asoc.primary_destination = net;
4044	}
4045	/* Validate primary is first */
4046	net = TAILQ_FIRST(&stcb->asoc.nets);
4047	if ((net != stcb->asoc.primary_destination) &&
4048	    (stcb->asoc.primary_destination)) {
4049		/*
4050		 * first one on the list is NOT the primary sctp_cmpaddr()
4051		 * is much more efficient if the primary is the first on the
4052		 * list, make it so.
4053		 */
4054		TAILQ_REMOVE(&stcb->asoc.nets,
4055		    stcb->asoc.primary_destination, sctp_next);
4056		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4057		    stcb->asoc.primary_destination, sctp_next);
4058	}
4059	return (0);
4060}
4061
4062
4063static uint32_t
4064sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4065{
4066	uint32_t id;
4067	struct sctpasochead *head;
4068	struct sctp_tcb *lstcb;
4069
4070	SCTP_INP_WLOCK(inp);
4071try_again:
4072	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4073		/* TSNH */
4074		SCTP_INP_WUNLOCK(inp);
4075		return (0);
4076	}
4077	/*
4078	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4079	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4080	 */
4081	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4082		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4083	}
4084	id = inp->sctp_associd_counter;
4085	inp->sctp_associd_counter++;
4086	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4087	if (lstcb) {
4088		goto try_again;
4089	}
4090	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4091	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4092	stcb->asoc.in_asocid_hash = 1;
4093	SCTP_INP_WUNLOCK(inp);
4094	return id;
4095}
4096
4097/*
4098 * allocate an association and add it to the endpoint. The caller must be
4099 * careful to add all additional addresses once they are know right away or
4100 * else the assoc will be may experience a blackout scenario.
4101 */
4102struct sctp_tcb *
4103sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4104    int *error, uint32_t override_tag, uint32_t vrf_id,
4105    struct thread *p
4106)
4107{
4108	/* note the p argument is only valid in unbound sockets */
4109
4110	struct sctp_tcb *stcb;
4111	struct sctp_association *asoc;
4112	struct sctpasochead *head;
4113	uint16_t rport;
4114	int err;
4115
4116	/*
4117	 * Assumption made here: Caller has done a
4118	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4119	 * address does not exist already.
4120	 */
4121	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4122		/* Hit max assoc, sorry no more */
4123		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4124		*error = ENOBUFS;
4125		return (NULL);
4126	}
4127	if (firstaddr == NULL) {
4128		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4129		*error = EINVAL;
4130		return (NULL);
4131	}
4132	SCTP_INP_RLOCK(inp);
4133	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4134	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4135	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4136		/*
4137		 * If its in the TCP pool, its NOT allowed to create an
4138		 * association. The parent listener needs to call
4139		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4140		 * off, or connected one does this.. its an error.
4141		 */
4142		SCTP_INP_RUNLOCK(inp);
4143		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4144		*error = EINVAL;
4145		return (NULL);
4146	}
4147	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4148	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4149		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4150		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4151			SCTP_INP_RUNLOCK(inp);
4152			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4153			*error = EINVAL;
4154			return (NULL);
4155		}
4156	}
4157	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4158#ifdef SCTP_DEBUG
4159	if (firstaddr) {
4160		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4161		switch (firstaddr->sa_family) {
4162#ifdef INET
4163		case AF_INET:
4164			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4165			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4166			break;
4167#endif
4168#ifdef INET6
4169		case AF_INET6:
4170			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4171			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4172			break;
4173#endif
4174		default:
4175			break;
4176		}
4177	} else {
4178		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4179	}
4180#endif				/* SCTP_DEBUG */
4181	switch (firstaddr->sa_family) {
4182#ifdef INET
4183	case AF_INET:
4184		{
4185			struct sockaddr_in *sin;
4186
4187			sin = (struct sockaddr_in *)firstaddr;
4188			if ((ntohs(sin->sin_port) == 0) ||
4189			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4190			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4191			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4192				/* Invalid address */
4193				SCTP_INP_RUNLOCK(inp);
4194				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4195				*error = EINVAL;
4196				return (NULL);
4197			}
4198			rport = sin->sin_port;
4199			break;
4200		}
4201#endif
4202#ifdef INET6
4203	case AF_INET6:
4204		{
4205			struct sockaddr_in6 *sin6;
4206
4207			sin6 = (struct sockaddr_in6 *)firstaddr;
4208			if ((ntohs(sin6->sin6_port) == 0) ||
4209			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4210			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4211				/* Invalid address */
4212				SCTP_INP_RUNLOCK(inp);
4213				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4214				*error = EINVAL;
4215				return (NULL);
4216			}
4217			rport = sin6->sin6_port;
4218			break;
4219		}
4220#endif
4221	default:
4222		/* not supported family type */
4223		SCTP_INP_RUNLOCK(inp);
4224		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4225		*error = EINVAL;
4226		return (NULL);
4227	}
4228	SCTP_INP_RUNLOCK(inp);
4229	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4230		/*
4231		 * If you have not performed a bind, then we need to do the
4232		 * ephemeral bind for you.
4233		 */
4234		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4235		    (struct sockaddr *)NULL,
4236		    (struct sctp_ifa *)NULL,
4237		    p
4238		    ))) {
4239			/* bind error, probably perm */
4240			*error = err;
4241			return (NULL);
4242		}
4243	}
4244	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4245	if (stcb == NULL) {
4246		/* out of memory? */
4247		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4248		*error = ENOMEM;
4249		return (NULL);
4250	}
4251	SCTP_INCR_ASOC_COUNT();
4252
4253	bzero(stcb, sizeof(*stcb));
4254	asoc = &stcb->asoc;
4255
4256	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4257	SCTP_TCB_LOCK_INIT(stcb);
4258	SCTP_TCB_SEND_LOCK_INIT(stcb);
4259	stcb->rport = rport;
4260	/* setup back pointer's */
4261	stcb->sctp_ep = inp;
4262	stcb->sctp_socket = inp->sctp_socket;
4263	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4264		/* failed */
4265		SCTP_TCB_LOCK_DESTROY(stcb);
4266		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4267		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4268		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4269		SCTP_DECR_ASOC_COUNT();
4270		*error = err;
4271		return (NULL);
4272	}
4273	/* and the port */
4274	SCTP_INP_INFO_WLOCK();
4275	SCTP_INP_WLOCK(inp);
4276	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4277		/* inpcb freed while alloc going on */
4278		SCTP_TCB_LOCK_DESTROY(stcb);
4279		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4280		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4281		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4282		SCTP_INP_WUNLOCK(inp);
4283		SCTP_INP_INFO_WUNLOCK();
4284		SCTP_DECR_ASOC_COUNT();
4285		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4286		*error = EINVAL;
4287		return (NULL);
4288	}
4289	SCTP_TCB_LOCK(stcb);
4290
4291	/* now that my_vtag is set, add it to the hash */
4292	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4293	/* put it in the bucket in the vtag hash of assoc's for the system */
4294	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4295	SCTP_INP_INFO_WUNLOCK();
4296
4297	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4298		/* failure.. memory error? */
4299		if (asoc->strmout) {
4300			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4301			asoc->strmout = NULL;
4302		}
4303		if (asoc->mapping_array) {
4304			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4305			asoc->mapping_array = NULL;
4306		}
4307		if (asoc->nr_mapping_array) {
4308			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4309			asoc->nr_mapping_array = NULL;
4310		}
4311		SCTP_DECR_ASOC_COUNT();
4312		SCTP_TCB_UNLOCK(stcb);
4313		SCTP_TCB_LOCK_DESTROY(stcb);
4314		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4315		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4316		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4317		SCTP_INP_WUNLOCK(inp);
4318		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4319		*error = ENOBUFS;
4320		return (NULL);
4321	}
4322	/* Init all the timers */
4323	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4324	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4325	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4326	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4327	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4328	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4329	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4330
4331	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4332	/* now file the port under the hash as well */
4333	if (inp->sctp_tcbhash != NULL) {
4334		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4335		    inp->sctp_hashmark)];
4336		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4337	}
4338	SCTP_INP_WUNLOCK(inp);
4339	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4340	return (stcb);
4341}
4342
4343
4344void
4345sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4346{
4347	struct sctp_association *asoc;
4348
4349	asoc = &stcb->asoc;
4350	asoc->numnets--;
4351	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4352	if (net == asoc->primary_destination) {
4353		/* Reset primary */
4354		struct sctp_nets *lnet;
4355
4356		lnet = TAILQ_FIRST(&asoc->nets);
4357		/*
4358		 * Mobility adaptation Ideally, if deleted destination is
4359		 * the primary, it becomes a fast retransmission trigger by
4360		 * the subsequent SET PRIMARY. (by micchie)
4361		 */
4362		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4363		    SCTP_MOBILITY_BASE) ||
4364		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4365		    SCTP_MOBILITY_FASTHANDOFF)) {
4366			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4367			if (asoc->deleted_primary != NULL) {
4368				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4369				goto out;
4370			}
4371			asoc->deleted_primary = net;
4372			atomic_add_int(&net->ref_count, 1);
4373			memset(&net->lastsa, 0, sizeof(net->lastsa));
4374			memset(&net->lastsv, 0, sizeof(net->lastsv));
4375			sctp_mobility_feature_on(stcb->sctp_ep,
4376			    SCTP_MOBILITY_PRIM_DELETED);
4377			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4378			    stcb->sctp_ep, stcb, NULL);
4379		}
4380out:
4381		/* Try to find a confirmed primary */
4382		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4383	}
4384	if (net == asoc->last_data_chunk_from) {
4385		/* Reset primary */
4386		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4387	}
4388	if (net == asoc->last_control_chunk_from) {
4389		/* Clear net */
4390		asoc->last_control_chunk_from = NULL;
4391	}
4392	if (net == stcb->asoc.alternate) {
4393		sctp_free_remote_addr(stcb->asoc.alternate);
4394		stcb->asoc.alternate = NULL;
4395	}
4396	sctp_free_remote_addr(net);
4397}
4398
4399/*
4400 * remove a remote endpoint address from an association, it will fail if the
4401 * address does not exist.
4402 */
4403int
4404sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4405{
4406	/*
4407	 * Here we need to remove a remote address. This is quite simple, we
4408	 * first find it in the list of address for the association
4409	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4410	 * on that item. Note we do not allow it to be removed if there are
4411	 * no other addresses.
4412	 */
4413	struct sctp_association *asoc;
4414	struct sctp_nets *net, *nnet;
4415
4416	asoc = &stcb->asoc;
4417
4418	/* locate the address */
4419	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4420		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4421			continue;
4422		}
4423		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4424		    remaddr)) {
4425			/* we found the guy */
4426			if (asoc->numnets < 2) {
4427				/* Must have at LEAST two remote addresses */
4428				return (-1);
4429			} else {
4430				sctp_remove_net(stcb, net);
4431				return (0);
4432			}
4433		}
4434	}
4435	/* not found. */
4436	return (-2);
4437}
4438
4439void
4440sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4441{
4442	struct sctpvtaghead *chain;
4443	struct sctp_tagblock *twait_block;
4444	int found = 0;
4445	int i;
4446
4447	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4448	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4449		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4450			if ((twait_block->vtag_block[i].v_tag == tag) &&
4451			    (twait_block->vtag_block[i].lport == lport) &&
4452			    (twait_block->vtag_block[i].rport == rport)) {
4453				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4454				twait_block->vtag_block[i].v_tag = 0;
4455				twait_block->vtag_block[i].lport = 0;
4456				twait_block->vtag_block[i].rport = 0;
4457				found = 1;
4458				break;
4459			}
4460		}
4461		if (found)
4462			break;
4463	}
4464}
4465
4466int
4467sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4468{
4469	struct sctpvtaghead *chain;
4470	struct sctp_tagblock *twait_block;
4471	int found = 0;
4472	int i;
4473
4474	SCTP_INP_INFO_WLOCK();
4475	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4476	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4477		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4478			if ((twait_block->vtag_block[i].v_tag == tag) &&
4479			    (twait_block->vtag_block[i].lport == lport) &&
4480			    (twait_block->vtag_block[i].rport == rport)) {
4481				found = 1;
4482				break;
4483			}
4484		}
4485		if (found)
4486			break;
4487	}
4488	SCTP_INP_INFO_WUNLOCK();
4489	return (found);
4490}
4491
4492
4493void
4494sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4495{
4496	struct sctpvtaghead *chain;
4497	struct sctp_tagblock *twait_block;
4498	struct timeval now;
4499	int set, i;
4500
4501	if (time == 0) {
4502		/* Its disabled */
4503		return;
4504	}
4505	(void)SCTP_GETTIME_TIMEVAL(&now);
4506	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4507	set = 0;
4508	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4509		/* Block(s) present, lets find space, and expire on the fly */
4510		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4511			if ((twait_block->vtag_block[i].v_tag == 0) &&
4512			    !set) {
4513				twait_block->vtag_block[i].tv_sec_at_expire =
4514				    now.tv_sec + time;
4515				twait_block->vtag_block[i].v_tag = tag;
4516				twait_block->vtag_block[i].lport = lport;
4517				twait_block->vtag_block[i].rport = rport;
4518				set = 1;
4519			} else if ((twait_block->vtag_block[i].v_tag) &&
4520			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4521				/* Audit expires this guy */
4522				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4523				twait_block->vtag_block[i].v_tag = 0;
4524				twait_block->vtag_block[i].lport = 0;
4525				twait_block->vtag_block[i].rport = 0;
4526				if (set == 0) {
4527					/* Reuse it for my new tag */
4528					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4529					twait_block->vtag_block[i].v_tag = tag;
4530					twait_block->vtag_block[i].lport = lport;
4531					twait_block->vtag_block[i].rport = rport;
4532					set = 1;
4533				}
4534			}
4535		}
4536		if (set) {
4537			/*
4538			 * We only do up to the block where we can place our
4539			 * tag for audits
4540			 */
4541			break;
4542		}
4543	}
4544	/* Need to add a new block to chain */
4545	if (!set) {
4546		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4547		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4548		if (twait_block == NULL) {
4549#ifdef INVARIANTS
4550			panic("Can not alloc tagblock");
4551#endif
4552			return;
4553		}
4554		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4555		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4556		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4557		twait_block->vtag_block[0].v_tag = tag;
4558		twait_block->vtag_block[0].lport = lport;
4559		twait_block->vtag_block[0].rport = rport;
4560	}
4561}
4562
4563
4564
4565/*-
4566 * Free the association after un-hashing the remote port. This
4567 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4568 * expect that the input to this function IS a locked TCB.
4569 * It will return 0, if it did NOT destroy the association (instead
4570 * it unlocks it. It will return NON-zero if it either destroyed the
4571 * association OR the association is already destroyed.
4572 */
4573int
4574sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4575{
4576	int i;
4577	struct sctp_association *asoc;
4578	struct sctp_nets *net, *nnet;
4579	struct sctp_laddr *laddr, *naddr;
4580	struct sctp_tmit_chunk *chk, *nchk;
4581	struct sctp_asconf_addr *aparam, *naparam;
4582	struct sctp_asconf_ack *aack, *naack;
4583	struct sctp_stream_reset_list *strrst, *nstrrst;
4584	struct sctp_queued_to_read *sq, *nsq;
4585	struct sctp_stream_queue_pending *sp, *nsp;
4586	sctp_sharedkey_t *shared_key, *nshared_key;
4587	struct socket *so;
4588
4589	/* first, lets purge the entry from the hash table. */
4590
4591#ifdef SCTP_LOG_CLOSING
4592	sctp_log_closing(inp, stcb, 6);
4593#endif
4594	if (stcb->asoc.state == 0) {
4595#ifdef SCTP_LOG_CLOSING
4596		sctp_log_closing(inp, NULL, 7);
4597#endif
4598		/* there is no asoc, really TSNH :-0 */
4599		return (1);
4600	}
4601	if (stcb->asoc.alternate) {
4602		sctp_free_remote_addr(stcb->asoc.alternate);
4603		stcb->asoc.alternate = NULL;
4604	}
4605	/* TEMP CODE */
4606	if (stcb->freed_from_where == 0) {
4607		/* Only record the first place free happened from */
4608		stcb->freed_from_where = from_location;
4609	}
4610	/* TEMP CODE */
4611
4612	asoc = &stcb->asoc;
4613	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4614	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4615		/* nothing around */
4616		so = NULL;
4617	else
4618		so = inp->sctp_socket;
4619
4620	/*
4621	 * We used timer based freeing if a reader or writer is in the way.
4622	 * So we first check if we are actually being called from a timer,
4623	 * if so we abort early if a reader or writer is still in the way.
4624	 */
4625	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4626	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4627		/*
4628		 * is it the timer driving us? if so are the reader/writers
4629		 * gone?
4630		 */
4631		if (stcb->asoc.refcnt) {
4632			/* nope, reader or writer in the way */
4633			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4634			/* no asoc destroyed */
4635			SCTP_TCB_UNLOCK(stcb);
4636#ifdef SCTP_LOG_CLOSING
4637			sctp_log_closing(inp, stcb, 8);
4638#endif
4639			return (0);
4640		}
4641	}
4642	/* now clean up any other timers */
4643	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4644	asoc->dack_timer.self = NULL;
4645	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4646	/*-
4647	 * For stream reset we don't blast this unless
4648	 * it is a str-reset timer, it might be the
4649	 * free-asoc timer which we DON'T want to
4650	 * disturb.
4651	 */
4652	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4653		asoc->strreset_timer.self = NULL;
4654	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4655	asoc->asconf_timer.self = NULL;
4656	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4657	asoc->autoclose_timer.self = NULL;
4658	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4659	asoc->shut_guard_timer.self = NULL;
4660	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4661	asoc->delayed_event_timer.self = NULL;
4662	/* Mobility adaptation */
4663	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4664	asoc->delete_prim_timer.self = NULL;
4665	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4666		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4667		net->rxt_timer.self = NULL;
4668		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4669		net->pmtu_timer.self = NULL;
4670		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4671		net->hb_timer.self = NULL;
4672	}
4673	/* Now the read queue needs to be cleaned up (only once) */
4674	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4675		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4676		SCTP_INP_READ_LOCK(inp);
4677		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4678			if (sq->stcb == stcb) {
4679				sq->do_not_ref_stcb = 1;
4680				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4681				/*
4682				 * If there is no end, there never will be
4683				 * now.
4684				 */
4685				if (sq->end_added == 0) {
4686					/* Held for PD-API clear that. */
4687					sq->pdapi_aborted = 1;
4688					sq->held_length = 0;
4689					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4690						/*
4691						 * Need to add a PD-API
4692						 * aborted indication.
4693						 * Setting the control_pdapi
4694						 * assures that it will be
4695						 * added right after this
4696						 * msg.
4697						 */
4698						uint32_t strseq;
4699
4700						stcb->asoc.control_pdapi = sq;
4701						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4702						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4703						    stcb,
4704						    SCTP_PARTIAL_DELIVERY_ABORTED,
4705						    (void *)&strseq,
4706						    SCTP_SO_LOCKED);
4707						stcb->asoc.control_pdapi = NULL;
4708					}
4709				}
4710				/* Add an end to wake them */
4711				sq->end_added = 1;
4712			}
4713		}
4714		SCTP_INP_READ_UNLOCK(inp);
4715		if (stcb->block_entry) {
4716			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4717			stcb->block_entry->error = ECONNRESET;
4718			stcb->block_entry = NULL;
4719		}
4720	}
4721	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4722		/*
4723		 * Someone holds a reference OR the socket is unaccepted
4724		 * yet.
4725		 */
4726		if ((stcb->asoc.refcnt) ||
4727		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4728		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4729			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4730			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4731		}
4732		SCTP_TCB_UNLOCK(stcb);
4733		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4734		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4735			/* nothing around */
4736			so = NULL;
4737		if (so) {
4738			/* Wake any reader/writers */
4739			sctp_sorwakeup(inp, so);
4740			sctp_sowwakeup(inp, so);
4741		}
4742#ifdef SCTP_LOG_CLOSING
4743		sctp_log_closing(inp, stcb, 9);
4744#endif
4745		/* no asoc destroyed */
4746		return (0);
4747	}
4748#ifdef SCTP_LOG_CLOSING
4749	sctp_log_closing(inp, stcb, 10);
4750#endif
4751	/*
4752	 * When I reach here, no others want to kill the assoc yet.. and I
4753	 * own the lock. Now its possible an abort comes in when I do the
4754	 * lock exchange below to grab all the locks to do the final take
4755	 * out. to prevent this we increment the count, which will start a
4756	 * timer and blow out above thus assuring us that we hold exclusive
4757	 * killing of the asoc. Note that after getting back the TCB lock we
4758	 * will go ahead and increment the counter back up and stop any
4759	 * timer a passing stranger may have started :-S
4760	 */
4761	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4762		atomic_add_int(&stcb->asoc.refcnt, 1);
4763
4764		SCTP_TCB_UNLOCK(stcb);
4765		SCTP_INP_INFO_WLOCK();
4766		SCTP_INP_WLOCK(inp);
4767		SCTP_TCB_LOCK(stcb);
4768	}
4769	/* Double check the GONE flag */
4770	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4771	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4772		/* nothing around */
4773		so = NULL;
4774
4775	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4776	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4777		/*
4778		 * For TCP type we need special handling when we are
4779		 * connected. We also include the peel'ed off ones to.
4780		 */
4781		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4782			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4783			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4784			if (so) {
4785				SOCK_LOCK(so);
4786				if (so->so_rcv.sb_cc == 0) {
4787					so->so_state &= ~(SS_ISCONNECTING |
4788					    SS_ISDISCONNECTING |
4789					    SS_ISCONFIRMING |
4790					    SS_ISCONNECTED);
4791				}
4792				socantrcvmore_locked(so);
4793				sctp_sowwakeup(inp, so);
4794				sctp_sorwakeup(inp, so);
4795				SCTP_SOWAKEUP(so);
4796			}
4797		}
4798	}
4799	/*
4800	 * Make it invalid too, that way if its about to run it will abort
4801	 * and return.
4802	 */
4803	/* re-increment the lock */
4804	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4805		atomic_add_int(&stcb->asoc.refcnt, -1);
4806	}
4807	if (stcb->asoc.refcnt) {
4808		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4809		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4810		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4811			SCTP_INP_INFO_WUNLOCK();
4812			SCTP_INP_WUNLOCK(inp);
4813		}
4814		SCTP_TCB_UNLOCK(stcb);
4815		return (0);
4816	}
4817	asoc->state = 0;
4818	if (inp->sctp_tcbhash) {
4819		LIST_REMOVE(stcb, sctp_tcbhash);
4820	}
4821	if (stcb->asoc.in_asocid_hash) {
4822		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4823	}
4824	/* Now lets remove it from the list of ALL associations in the EP */
4825	LIST_REMOVE(stcb, sctp_tcblist);
4826	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4827		SCTP_INP_INCR_REF(inp);
4828		SCTP_INP_WUNLOCK(inp);
4829	}
4830	/* pull from vtag hash */
4831	LIST_REMOVE(stcb, sctp_asocs);
4832	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4833	    inp->sctp_lport, stcb->rport);
4834
4835	/*
4836	 * Now restop the timers to be sure this is paranoia at is finest!
4837	 */
4838	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4839	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4840	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4841	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4842	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4843	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4844	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4845	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4846		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4847		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4848		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4849	}
4850
4851	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4852	/*
4853	 * The chunk lists and such SHOULD be empty but we check them just
4854	 * in case.
4855	 */
4856	/* anything on the wheel needs to be removed */
4857	for (i = 0; i < asoc->streamoutcnt; i++) {
4858		struct sctp_stream_out *outs;
4859
4860		outs = &asoc->strmout[i];
4861		/* now clean up any chunks here */
4862		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4863			TAILQ_REMOVE(&outs->outqueue, sp, next);
4864			sctp_free_spbufspace(stcb, asoc, sp);
4865			if (sp->data) {
4866				if (so) {
4867					/* Still an open socket - report */
4868					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4869					    0, (void *)sp, SCTP_SO_LOCKED);
4870				}
4871				if (sp->data) {
4872					sctp_m_freem(sp->data);
4873					sp->data = NULL;
4874					sp->tail_mbuf = NULL;
4875					sp->length = 0;
4876				}
4877			}
4878			if (sp->net) {
4879				sctp_free_remote_addr(sp->net);
4880				sp->net = NULL;
4881			}
4882			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4883		}
4884	}
4885	/* sa_ignore FREED_MEMORY */
4886	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4887		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4888		SCTP_FREE(strrst, SCTP_M_STRESET);
4889	}
4890	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4891		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4892		if (sq->data) {
4893			sctp_m_freem(sq->data);
4894			sq->data = NULL;
4895		}
4896		sctp_free_remote_addr(sq->whoFrom);
4897		sq->whoFrom = NULL;
4898		sq->stcb = NULL;
4899		/* Free the ctl entry */
4900		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4901		SCTP_DECR_READQ_COUNT();
4902		/* sa_ignore FREED_MEMORY */
4903	}
4904	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4905		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4906		if (chk->data) {
4907			sctp_m_freem(chk->data);
4908			chk->data = NULL;
4909		}
4910		if (chk->holds_key_ref)
4911			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4912		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4913		SCTP_DECR_CHK_COUNT();
4914		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4915		asoc->free_chunk_cnt--;
4916		/* sa_ignore FREED_MEMORY */
4917	}
4918	/* pending send queue SHOULD be empty */
4919	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4920		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4921			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4922#ifdef INVARIANTS
4923		} else {
4924			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4925#endif
4926		}
4927		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4928		if (chk->data) {
4929			if (so) {
4930				/* Still a socket? */
4931				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
4932				    0, chk, SCTP_SO_LOCKED);
4933			}
4934			if (chk->data) {
4935				sctp_m_freem(chk->data);
4936				chk->data = NULL;
4937			}
4938		}
4939		if (chk->holds_key_ref)
4940			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4941		if (chk->whoTo) {
4942			sctp_free_remote_addr(chk->whoTo);
4943			chk->whoTo = NULL;
4944		}
4945		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4946		SCTP_DECR_CHK_COUNT();
4947		/* sa_ignore FREED_MEMORY */
4948	}
4949	/* sent queue SHOULD be empty */
4950	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
4951		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
4952			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4953				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4954#ifdef INVARIANTS
4955			} else {
4956				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4957#endif
4958			}
4959		}
4960		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4961		if (chk->data) {
4962			if (so) {
4963				/* Still a socket? */
4964				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
4965				    0, chk, SCTP_SO_LOCKED);
4966			}
4967			if (chk->data) {
4968				sctp_m_freem(chk->data);
4969				chk->data = NULL;
4970			}
4971		}
4972		if (chk->holds_key_ref)
4973			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4974		sctp_free_remote_addr(chk->whoTo);
4975		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4976		SCTP_DECR_CHK_COUNT();
4977		/* sa_ignore FREED_MEMORY */
4978	}
4979#ifdef INVARIANTS
4980	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
4981		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
4982			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
4983		}
4984	}
4985#endif
4986	/* control queue MAY not be empty */
4987	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
4988		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4989		if (chk->data) {
4990			sctp_m_freem(chk->data);
4991			chk->data = NULL;
4992		}
4993		if (chk->holds_key_ref)
4994			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4995		sctp_free_remote_addr(chk->whoTo);
4996		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4997		SCTP_DECR_CHK_COUNT();
4998		/* sa_ignore FREED_MEMORY */
4999	}
5000	/* ASCONF queue MAY not be empty */
5001	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5002		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5003		if (chk->data) {
5004			sctp_m_freem(chk->data);
5005			chk->data = NULL;
5006		}
5007		if (chk->holds_key_ref)
5008			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5009		sctp_free_remote_addr(chk->whoTo);
5010		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5011		SCTP_DECR_CHK_COUNT();
5012		/* sa_ignore FREED_MEMORY */
5013	}
5014	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5015		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5016		if (chk->data) {
5017			sctp_m_freem(chk->data);
5018			chk->data = NULL;
5019		}
5020		if (chk->holds_key_ref)
5021			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5022		sctp_free_remote_addr(chk->whoTo);
5023		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5024		SCTP_DECR_CHK_COUNT();
5025		/* sa_ignore FREED_MEMORY */
5026	}
5027
5028	if (asoc->mapping_array) {
5029		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5030		asoc->mapping_array = NULL;
5031	}
5032	if (asoc->nr_mapping_array) {
5033		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5034		asoc->nr_mapping_array = NULL;
5035	}
5036	/* the stream outs */
5037	if (asoc->strmout) {
5038		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5039		asoc->strmout = NULL;
5040	}
5041	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5042	if (asoc->strmin) {
5043		struct sctp_queued_to_read *ctl, *nctl;
5044
5045		for (i = 0; i < asoc->streamincnt; i++) {
5046			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5047				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5048				sctp_free_remote_addr(ctl->whoFrom);
5049				if (ctl->data) {
5050					sctp_m_freem(ctl->data);
5051					ctl->data = NULL;
5052				}
5053				/*
5054				 * We don't free the address here since all
5055				 * the net's were freed above.
5056				 */
5057				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5058				SCTP_DECR_READQ_COUNT();
5059			}
5060		}
5061		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5062		asoc->strmin = NULL;
5063	}
5064	asoc->streamincnt = 0;
5065	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5066#ifdef INVARIANTS
5067		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5068			panic("no net's left alloc'ed, or list points to itself");
5069		}
5070#endif
5071		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5072		sctp_free_remote_addr(net);
5073	}
5074	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5075		/* sa_ignore FREED_MEMORY */
5076		sctp_remove_laddr(laddr);
5077	}
5078
5079	/* pending asconf (address) parameters */
5080	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5081		/* sa_ignore FREED_MEMORY */
5082		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5083		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5084	}
5085	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5086		/* sa_ignore FREED_MEMORY */
5087		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5088		if (aack->data != NULL) {
5089			sctp_m_freem(aack->data);
5090		}
5091		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5092	}
5093	/* clean up auth stuff */
5094	if (asoc->local_hmacs)
5095		sctp_free_hmaclist(asoc->local_hmacs);
5096	if (asoc->peer_hmacs)
5097		sctp_free_hmaclist(asoc->peer_hmacs);
5098
5099	if (asoc->local_auth_chunks)
5100		sctp_free_chunklist(asoc->local_auth_chunks);
5101	if (asoc->peer_auth_chunks)
5102		sctp_free_chunklist(asoc->peer_auth_chunks);
5103
5104	sctp_free_authinfo(&asoc->authinfo);
5105
5106	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5107		LIST_REMOVE(shared_key, next);
5108		sctp_free_sharedkey(shared_key);
5109		/* sa_ignore FREED_MEMORY */
5110	}
5111
5112	/* Insert new items here :> */
5113
5114	/* Get rid of LOCK */
5115	SCTP_TCB_UNLOCK(stcb);
5116	SCTP_TCB_LOCK_DESTROY(stcb);
5117	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5118	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5119		SCTP_INP_INFO_WUNLOCK();
5120		SCTP_INP_RLOCK(inp);
5121	}
5122#ifdef SCTP_TRACK_FREED_ASOCS
5123	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5124		/* now clean up the tasoc itself */
5125		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5126		SCTP_DECR_ASOC_COUNT();
5127	} else {
5128		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5129	}
5130#else
5131	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5132	SCTP_DECR_ASOC_COUNT();
5133#endif
5134	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5135		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5136			/*
5137			 * If its NOT the inp_free calling us AND sctp_close
5138			 * as been called, we call back...
5139			 */
5140			SCTP_INP_RUNLOCK(inp);
5141			/*
5142			 * This will start the kill timer (if we are the
5143			 * last one) since we hold an increment yet. But
5144			 * this is the only safe way to do this since
5145			 * otherwise if the socket closes at the same time
5146			 * we are here we might collide in the cleanup.
5147			 */
5148			sctp_inpcb_free(inp,
5149			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5150			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5151			SCTP_INP_DECR_REF(inp);
5152			goto out_of;
5153		} else {
5154			/* The socket is still open. */
5155			SCTP_INP_DECR_REF(inp);
5156		}
5157	}
5158	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5159		SCTP_INP_RUNLOCK(inp);
5160	}
5161out_of:
5162	/* destroyed the asoc */
5163#ifdef SCTP_LOG_CLOSING
5164	sctp_log_closing(inp, NULL, 11);
5165#endif
5166	return (1);
5167}
5168
5169
5170
5171/*
5172 * determine if a destination is "reachable" based upon the addresses bound
5173 * to the current endpoint (e.g. only v4 or v6 currently bound)
5174 */
5175/*
5176 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5177 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5178 * types bound as its endpoint
5179 */
5180int
5181sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5182{
5183	struct sctp_inpcb *inp;
5184	int answer;
5185
5186	/*
5187	 * No locks here, the TCB, in all cases is already locked and an
5188	 * assoc is up. There is either a INP lock by the caller applied (in
5189	 * asconf case when deleting an address) or NOT in the HB case,
5190	 * however if HB then the INP increment is up and the INP will not
5191	 * be removed (on top of the fact that we have a TCB lock). So we
5192	 * only want to read the sctp_flags, which is either bound-all or
5193	 * not.. no protection needed since once an assoc is up you can't be
5194	 * changing your binding.
5195	 */
5196	inp = stcb->sctp_ep;
5197	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5198		/* if bound all, destination is not restricted */
5199		/*
5200		 * RRS: Question during lock work: Is this correct? If you
5201		 * are bound-all you still might need to obey the V4--V6
5202		 * flags??? IMO this bound-all stuff needs to be removed!
5203		 */
5204		return (1);
5205	}
5206	/* NOTE: all "scope" checks are done when local addresses are added */
5207	switch (destaddr->sa_family) {
5208#ifdef INET6
5209	case AF_INET6:
5210		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5211		break;
5212#endif
5213#ifdef INET
5214	case AF_INET:
5215		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5216		break;
5217#endif
5218	default:
5219		/* invalid family, so it's unreachable */
5220		answer = 0;
5221		break;
5222	}
5223	return (answer);
5224}
5225
5226/*
5227 * update the inp_vflags on an endpoint
5228 */
5229static void
5230sctp_update_ep_vflag(struct sctp_inpcb *inp)
5231{
5232	struct sctp_laddr *laddr;
5233
5234	/* first clear the flag */
5235	inp->ip_inp.inp.inp_vflag = 0;
5236	/* set the flag based on addresses on the ep list */
5237	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5238		if (laddr->ifa == NULL) {
5239			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5240			    __FUNCTION__);
5241			continue;
5242		}
5243		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5244			continue;
5245		}
5246		switch (laddr->ifa->address.sa.sa_family) {
5247#ifdef INET6
5248		case AF_INET6:
5249			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5250			break;
5251#endif
5252#ifdef INET
5253		case AF_INET:
5254			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5255			break;
5256#endif
5257		default:
5258			break;
5259		}
5260	}
5261}
5262
5263/*
5264 * Add the address to the endpoint local address list There is nothing to be
5265 * done if we are bound to all addresses
5266 */
5267void
5268sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5269{
5270	struct sctp_laddr *laddr;
5271	int fnd, error = 0;
5272
5273	fnd = 0;
5274
5275	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5276		/* You are already bound to all. You have it already */
5277		return;
5278	}
5279#ifdef INET6
5280	if (ifa->address.sa.sa_family == AF_INET6) {
5281		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5282			/* Can't bind a non-useable addr. */
5283			return;
5284		}
5285	}
5286#endif
5287	/* first, is it already present? */
5288	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5289		if (laddr->ifa == ifa) {
5290			fnd = 1;
5291			break;
5292		}
5293	}
5294
5295	if (fnd == 0) {
5296		/* Not in the ep list */
5297		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5298		if (error != 0)
5299			return;
5300		inp->laddr_count++;
5301		/* update inp_vflag flags */
5302		switch (ifa->address.sa.sa_family) {
5303#ifdef INET6
5304		case AF_INET6:
5305			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5306			break;
5307#endif
5308#ifdef INET
5309		case AF_INET:
5310			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5311			break;
5312#endif
5313		default:
5314			break;
5315		}
5316	}
5317	return;
5318}
5319
5320
5321/*
5322 * select a new (hopefully reachable) destination net (should only be used
5323 * when we deleted an ep addr that is the only usable source address to reach
5324 * the destination net)
5325 */
5326static void
5327sctp_select_primary_destination(struct sctp_tcb *stcb)
5328{
5329	struct sctp_nets *net;
5330
5331	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5332		/* for now, we'll just pick the first reachable one we find */
5333		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5334			continue;
5335		if (sctp_destination_is_reachable(stcb,
5336		    (struct sockaddr *)&net->ro._l_addr)) {
5337			/* found a reachable destination */
5338			stcb->asoc.primary_destination = net;
5339		}
5340	}
5341	/* I can't there from here! ...we're gonna die shortly... */
5342}
5343
5344
5345/*
5346 * Delete the address from the endpoint local address list There is nothing
5347 * to be done if we are bound to all addresses
5348 */
5349void
5350sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5351{
5352	struct sctp_laddr *laddr;
5353	int fnd;
5354
5355	fnd = 0;
5356	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5357		/* You are already bound to all. You have it already */
5358		return;
5359	}
5360	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5361		if (laddr->ifa == ifa) {
5362			fnd = 1;
5363			break;
5364		}
5365	}
5366	if (fnd && (inp->laddr_count < 2)) {
5367		/* can't delete unless there are at LEAST 2 addresses */
5368		return;
5369	}
5370	if (fnd) {
5371		/*
5372		 * clean up any use of this address go through our
5373		 * associations and clear any last_used_address that match
5374		 * this one for each assoc, see if a new primary_destination
5375		 * is needed
5376		 */
5377		struct sctp_tcb *stcb;
5378
5379		/* clean up "next_addr_touse" */
5380		if (inp->next_addr_touse == laddr)
5381			/* delete this address */
5382			inp->next_addr_touse = NULL;
5383
5384		/* clean up "last_used_address" */
5385		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5386			struct sctp_nets *net;
5387
5388			SCTP_TCB_LOCK(stcb);
5389			if (stcb->asoc.last_used_address == laddr)
5390				/* delete this address */
5391				stcb->asoc.last_used_address = NULL;
5392			/*
5393			 * Now spin through all the nets and purge any ref
5394			 * to laddr
5395			 */
5396			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5397				if (net->ro._s_addr &&
5398				    (net->ro._s_addr->ifa == laddr->ifa)) {
5399					/* Yep, purge src address selected */
5400					sctp_rtentry_t *rt;
5401
5402					/* delete this address if cached */
5403					rt = net->ro.ro_rt;
5404					if (rt != NULL) {
5405						RTFREE(rt);
5406						net->ro.ro_rt = NULL;
5407					}
5408					sctp_free_ifa(net->ro._s_addr);
5409					net->ro._s_addr = NULL;
5410					net->src_addr_selected = 0;
5411				}
5412			}
5413			SCTP_TCB_UNLOCK(stcb);
5414		}		/* for each tcb */
5415		/* remove it from the ep list */
5416		sctp_remove_laddr(laddr);
5417		inp->laddr_count--;
5418		/* update inp_vflag flags */
5419		sctp_update_ep_vflag(inp);
5420	}
5421	return;
5422}
5423
5424/*
5425 * Add the address to the TCB local address restricted list.
5426 * This is a "pending" address list (eg. addresses waiting for an
5427 * ASCONF-ACK response) and cannot be used as a valid source address.
5428 */
5429void
5430sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5431{
5432	struct sctp_laddr *laddr;
5433	struct sctpladdr *list;
5434
5435	/*
5436	 * Assumes TCB is locked.. and possibly the INP. May need to
5437	 * confirm/fix that if we need it and is not the case.
5438	 */
5439	list = &stcb->asoc.sctp_restricted_addrs;
5440
5441#ifdef INET6
5442	if (ifa->address.sa.sa_family == AF_INET6) {
5443		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5444			/* Can't bind a non-existent addr. */
5445			return;
5446		}
5447	}
5448#endif
5449	/* does the address already exist? */
5450	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5451		if (laddr->ifa == ifa) {
5452			return;
5453		}
5454	}
5455
5456	/* add to the list */
5457	(void)sctp_insert_laddr(list, ifa, 0);
5458	return;
5459}
5460
5461/*
5462 * insert an laddr entry with the given ifa for the desired list
5463 */
5464int
5465sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5466{
5467	struct sctp_laddr *laddr;
5468
5469	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5470	if (laddr == NULL) {
5471		/* out of memory? */
5472		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5473		return (EINVAL);
5474	}
5475	SCTP_INCR_LADDR_COUNT();
5476	bzero(laddr, sizeof(*laddr));
5477	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5478	laddr->ifa = ifa;
5479	laddr->action = act;
5480	atomic_add_int(&ifa->refcount, 1);
5481	/* insert it */
5482	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5483
5484	return (0);
5485}
5486
5487/*
5488 * Remove an laddr entry from the local address list (on an assoc)
5489 */
5490void
5491sctp_remove_laddr(struct sctp_laddr *laddr)
5492{
5493
5494	/* remove from the list */
5495	LIST_REMOVE(laddr, sctp_nxt_addr);
5496	sctp_free_ifa(laddr->ifa);
5497	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5498	SCTP_DECR_LADDR_COUNT();
5499}
5500
5501/*
5502 * Remove a local address from the TCB local address restricted list
5503 */
5504void
5505sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5506{
5507	struct sctp_inpcb *inp;
5508	struct sctp_laddr *laddr;
5509
5510	/*
5511	 * This is called by asconf work. It is assumed that a) The TCB is
5512	 * locked and b) The INP is locked. This is true in as much as I can
5513	 * trace through the entry asconf code where I did these locks.
5514	 * Again, the ASCONF code is a bit different in that it does lock
5515	 * the INP during its work often times. This must be since we don't
5516	 * want other proc's looking up things while what they are looking
5517	 * up is changing :-D
5518	 */
5519
5520	inp = stcb->sctp_ep;
5521	/* if subset bound and don't allow ASCONF's, can't delete last */
5522	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5523	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5524		if (stcb->sctp_ep->laddr_count < 2) {
5525			/* can't delete last address */
5526			return;
5527		}
5528	}
5529	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5530		/* remove the address if it exists */
5531		if (laddr->ifa == NULL)
5532			continue;
5533		if (laddr->ifa == ifa) {
5534			sctp_remove_laddr(laddr);
5535			return;
5536		}
5537	}
5538
5539	/* address not found! */
5540	return;
5541}
5542
5543/*
5544 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5545 */
5546/* sysctl */
5547static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5548static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5549
5550
5551
5552#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5553struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5554int *sctp_cpuarry = NULL;
5555void
5556sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5557{
5558	/* Queue a packet to a processor for the specified core */
5559	struct sctp_mcore_queue *qent;
5560	struct sctp_mcore_ctrl *wkq;
5561	int need_wake = 0;
5562
5563	if (sctp_mcore_workers == NULL) {
5564		/* Something went way bad during setup */
5565		sctp_input_with_port(m, off, 0);
5566		return;
5567	}
5568	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5569	    (sizeof(struct sctp_mcore_queue)),
5570	    SCTP_M_MCORE);
5571	if (qent == NULL) {
5572		/* This is trouble  */
5573		sctp_input_with_port(m, off, 0);
5574		return;
5575	}
5576	qent->vn = curvnet;
5577	qent->m = m;
5578	qent->off = off;
5579	qent->v6 = 0;
5580	wkq = &sctp_mcore_workers[cpu_to_use];
5581	SCTP_MCORE_QLOCK(wkq);
5582
5583	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5584	if (wkq->running == 0) {
5585		need_wake = 1;
5586	}
5587	SCTP_MCORE_QUNLOCK(wkq);
5588	if (need_wake) {
5589		wakeup(&wkq->running);
5590	}
5591}
5592
5593static void
5594sctp_mcore_thread(void *arg)
5595{
5596
5597	struct sctp_mcore_ctrl *wkq;
5598	struct sctp_mcore_queue *qent;
5599
5600	wkq = (struct sctp_mcore_ctrl *)arg;
5601	struct mbuf *m;
5602	int off, v6;
5603
5604	/* Wait for first tickle */
5605	SCTP_MCORE_LOCK(wkq);
5606	wkq->running = 0;
5607	msleep(&wkq->running,
5608	    &wkq->core_mtx,
5609	    0, "wait for pkt", 0);
5610	SCTP_MCORE_UNLOCK(wkq);
5611
5612	/* Bind to our cpu */
5613	thread_lock(curthread);
5614	sched_bind(curthread, wkq->cpuid);
5615	thread_unlock(curthread);
5616
5617	/* Now lets start working */
5618	SCTP_MCORE_LOCK(wkq);
5619	/* Now grab lock and go */
5620	for (;;) {
5621		SCTP_MCORE_QLOCK(wkq);
5622skip_sleep:
5623		wkq->running = 1;
5624		qent = TAILQ_FIRST(&wkq->que);
5625		if (qent) {
5626			TAILQ_REMOVE(&wkq->que, qent, next);
5627			SCTP_MCORE_QUNLOCK(wkq);
5628			CURVNET_SET(qent->vn);
5629			m = qent->m;
5630			off = qent->off;
5631			v6 = qent->v6;
5632			SCTP_FREE(qent, SCTP_M_MCORE);
5633			if (v6 == 0) {
5634				sctp_input_with_port(m, off, 0);
5635			} else {
5636				SCTP_PRINTF("V6 not yet supported\n");
5637				sctp_m_freem(m);
5638			}
5639			CURVNET_RESTORE();
5640			SCTP_MCORE_QLOCK(wkq);
5641		}
5642		wkq->running = 0;
5643		if (!TAILQ_EMPTY(&wkq->que)) {
5644			goto skip_sleep;
5645		}
5646		SCTP_MCORE_QUNLOCK(wkq);
5647		msleep(&wkq->running,
5648		    &wkq->core_mtx,
5649		    0, "wait for pkt", 0);
5650	}
5651}
5652
5653static void
5654sctp_startup_mcore_threads(void)
5655{
5656	int i, cpu;
5657
5658	if (mp_ncpus == 1)
5659		return;
5660
5661	if (sctp_mcore_workers != NULL) {
5662		/*
5663		 * Already been here in some previous vnet?
5664		 */
5665		return;
5666	}
5667	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5668	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5669	    SCTP_M_MCORE);
5670	if (sctp_mcore_workers == NULL) {
5671		/* TSNH I hope */
5672		return;
5673	}
5674	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5675	    sizeof(struct sctp_mcore_ctrl)));
5676	/* Init the structures */
5677	for (i = 0; i <= mp_maxid; i++) {
5678		TAILQ_INIT(&sctp_mcore_workers[i].que);
5679		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5680		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5681		sctp_mcore_workers[i].cpuid = i;
5682	}
5683	if (sctp_cpuarry == NULL) {
5684		SCTP_MALLOC(sctp_cpuarry, int *,
5685		    (mp_ncpus * sizeof(int)),
5686		    SCTP_M_MCORE);
5687		i = 0;
5688		CPU_FOREACH(cpu) {
5689			sctp_cpuarry[i] = cpu;
5690			i++;
5691		}
5692	}
5693	/* Now start them all */
5694	CPU_FOREACH(cpu) {
5695		(void)kproc_create(sctp_mcore_thread,
5696		    (void *)&sctp_mcore_workers[cpu],
5697		    &sctp_mcore_workers[cpu].thread_proc,
5698		    RFPROC,
5699		    SCTP_KTHREAD_PAGES,
5700		    SCTP_MCORE_NAME);
5701
5702	}
5703}
5704
5705#endif
5706
5707void
5708sctp_pcb_init()
5709{
5710	/*
5711	 * SCTP initialization for the PCB structures should be called by
5712	 * the sctp_init() funciton.
5713	 */
5714	int i;
5715	struct timeval tv;
5716
5717	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5718		/* error I was called twice */
5719		return;
5720	}
5721	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5722
5723#if defined(SCTP_LOCAL_TRACE_BUF)
5724	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5725#endif
5726#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5727	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5728	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5729	    SCTP_M_MCORE);
5730#endif
5731	(void)SCTP_GETTIME_TIMEVAL(&tv);
5732#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5733	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5734	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5735	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5736#else
5737	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5738	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5739	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5740#endif
5741	/* init the empty list of (All) Endpoints */
5742	LIST_INIT(&SCTP_BASE_INFO(listhead));
5743
5744
5745	/* init the hash table of endpoints */
5746	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5747	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5748	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5749	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5750	    &SCTP_BASE_INFO(hashasocmark));
5751	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5752	    &SCTP_BASE_INFO(hashmark));
5753	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5754	    &SCTP_BASE_INFO(hashtcpmark));
5755	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5756
5757
5758	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5759	    &SCTP_BASE_INFO(hashvrfmark));
5760
5761	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5762	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5763	/* init the zones */
5764	/*
5765	 * FIX ME: Should check for NULL returns, but if it does fail we are
5766	 * doomed to panic anyways... add later maybe.
5767	 */
5768	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5769	    sizeof(struct sctp_inpcb), maxsockets);
5770
5771	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5772	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5773
5774	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5775	    sizeof(struct sctp_laddr),
5776	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5777
5778	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5779	    sizeof(struct sctp_nets),
5780	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5781
5782	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5783	    sizeof(struct sctp_tmit_chunk),
5784	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5785
5786	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5787	    sizeof(struct sctp_queued_to_read),
5788	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5789
5790	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5791	    sizeof(struct sctp_stream_queue_pending),
5792	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5793
5794	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5795	    sizeof(struct sctp_asconf),
5796	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5797
5798	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5799	    sizeof(struct sctp_asconf_ack),
5800	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5801
5802
5803	/* Master Lock INIT for info structure */
5804	SCTP_INP_INFO_LOCK_INIT();
5805	SCTP_STATLOG_INIT_LOCK();
5806
5807	SCTP_IPI_COUNT_INIT();
5808	SCTP_IPI_ADDR_INIT();
5809#ifdef SCTP_PACKET_LOGGING
5810	SCTP_IP_PKTLOG_INIT();
5811#endif
5812	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5813
5814	SCTP_WQ_ADDR_INIT();
5815	/* not sure if we need all the counts */
5816	SCTP_BASE_INFO(ipi_count_ep) = 0;
5817	/* assoc/tcb zone info */
5818	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5819	/* local addrlist zone info */
5820	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5821	/* remote addrlist zone info */
5822	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5823	/* chunk info */
5824	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5825
5826	/* socket queue zone info */
5827	SCTP_BASE_INFO(ipi_count_readq) = 0;
5828
5829	/* stream out queue cont */
5830	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5831
5832	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5833	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5834
5835	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5836
5837	/* Init the TIMEWAIT list */
5838	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5839		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5840	}
5841	sctp_startup_iterator();
5842
5843#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5844	sctp_startup_mcore_threads();
5845#endif
5846
5847	/*
5848	 * INIT the default VRF which for BSD is the only one, other O/S's
5849	 * may have more. But initially they must start with one and then
5850	 * add the VRF's as addresses are added.
5851	 */
5852	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5853}
5854
5855/*
5856 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5857 */
5858void
5859sctp_pcb_finish(void)
5860{
5861	struct sctp_vrflist *vrf_bucket;
5862	struct sctp_vrf *vrf, *nvrf;
5863	struct sctp_ifn *ifn, *nifn;
5864	struct sctp_ifa *ifa, *nifa;
5865	struct sctpvtaghead *chain;
5866	struct sctp_tagblock *twait_block, *prev_twait_block;
5867	struct sctp_laddr *wi, *nwi;
5868	int i;
5869	struct sctp_iterator *it, *nit;
5870
5871	/*
5872	 * In FreeBSD the iterator thread never exits but we do clean up.
5873	 * The only way FreeBSD reaches here is if we have VRF's but we
5874	 * still add the ifdef to make it compile on old versions.
5875	 */
5876	SCTP_IPI_ITERATOR_WQ_LOCK();
5877	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5878		if (it->vn != curvnet) {
5879			continue;
5880		}
5881		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5882		if (it->function_atend != NULL) {
5883			(*it->function_atend) (it->pointer, it->val);
5884		}
5885		SCTP_FREE(it, SCTP_M_ITER);
5886	}
5887	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5888	SCTP_ITERATOR_LOCK();
5889	if ((sctp_it_ctl.cur_it) &&
5890	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5891		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5892	}
5893	SCTP_ITERATOR_UNLOCK();
5894	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5895	SCTP_WQ_ADDR_LOCK();
5896	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5897		LIST_REMOVE(wi, sctp_nxt_addr);
5898		SCTP_DECR_LADDR_COUNT();
5899		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5900	}
5901	SCTP_WQ_ADDR_UNLOCK();
5902
5903	/*
5904	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5905	 * destroyed first).
5906	 */
5907	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5908	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5909		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5910			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5911				/* free the ifa */
5912				LIST_REMOVE(ifa, next_bucket);
5913				LIST_REMOVE(ifa, next_ifa);
5914				SCTP_FREE(ifa, SCTP_M_IFA);
5915			}
5916			/* free the ifn */
5917			LIST_REMOVE(ifn, next_bucket);
5918			LIST_REMOVE(ifn, next_ifn);
5919			SCTP_FREE(ifn, SCTP_M_IFN);
5920		}
5921		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5922		/* free the vrf */
5923		LIST_REMOVE(vrf, next_vrf);
5924		SCTP_FREE(vrf, SCTP_M_VRF);
5925	}
5926	/* free the vrf hashes */
5927	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5928	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5929
5930	/*
5931	 * free the TIMEWAIT list elements malloc'd in the function
5932	 * sctp_add_vtag_to_timewait()...
5933	 */
5934	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5935		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5936		if (!LIST_EMPTY(chain)) {
5937			prev_twait_block = NULL;
5938			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5939				if (prev_twait_block) {
5940					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5941				}
5942				prev_twait_block = twait_block;
5943			}
5944			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5945		}
5946	}
5947
5948	/* free the locks and mutexes */
5949#ifdef SCTP_PACKET_LOGGING
5950	SCTP_IP_PKTLOG_DESTROY();
5951#endif
5952	SCTP_IPI_ADDR_DESTROY();
5953	SCTP_STATLOG_DESTROY();
5954	SCTP_INP_INFO_LOCK_DESTROY();
5955
5956	SCTP_WQ_ADDR_DESTROY();
5957
5958	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
5959	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
5960	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
5961	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
5962	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
5963	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
5964	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
5965	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
5966	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
5967	/* Get rid of other stuff to */
5968	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5969		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5970	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5971		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
5972	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
5973		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
5974#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5975	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
5976#endif
5977}
5978
5979
5980int
5981sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
5982    int offset, int limit,
5983    struct sockaddr *src, struct sockaddr *dst,
5984    struct sockaddr *altsa)
5985{
5986	/*
5987	 * grub through the INIT pulling addresses and loading them to the
5988	 * nets structure in the asoc. The from address in the mbuf should
5989	 * also be loaded (if it is not already). This routine can be called
5990	 * with either INIT or INIT-ACK's as long as the m points to the IP
5991	 * packet and the offset points to the beginning of the parameters.
5992	 */
5993	struct sctp_inpcb *inp;
5994	struct sctp_nets *net, *nnet, *net_tmp;
5995	struct sctp_paramhdr *phdr, parm_buf;
5996	struct sctp_tcb *stcb_tmp;
5997	uint16_t ptype, plen;
5998	struct sockaddr *sa;
5999	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6000	struct sctp_auth_random *p_random = NULL;
6001	uint16_t random_len = 0;
6002	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6003	struct sctp_auth_hmac_algo *hmacs = NULL;
6004	uint16_t hmacs_len = 0;
6005	uint8_t saw_asconf = 0;
6006	uint8_t saw_asconf_ack = 0;
6007	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6008	struct sctp_auth_chunk_list *chunks = NULL;
6009	uint16_t num_chunks = 0;
6010	sctp_key_t *new_key;
6011	uint32_t keylen;
6012	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6013	uint8_t ecn_allowed;
6014
6015#ifdef INET
6016	struct sockaddr_in sin;
6017
6018#endif
6019#ifdef INET6
6020	struct sockaddr_in6 sin6;
6021
6022#endif
6023
6024	/* First get the destination address setup too. */
6025#ifdef INET
6026	memset(&sin, 0, sizeof(sin));
6027	sin.sin_family = AF_INET;
6028	sin.sin_len = sizeof(sin);
6029	sin.sin_port = stcb->rport;
6030#endif
6031#ifdef INET6
6032	memset(&sin6, 0, sizeof(sin6));
6033	sin6.sin6_family = AF_INET6;
6034	sin6.sin6_len = sizeof(struct sockaddr_in6);
6035	sin6.sin6_port = stcb->rport;
6036#endif
6037	if (altsa) {
6038		sa = altsa;
6039	} else {
6040		sa = src;
6041	}
6042	/* Turn off ECN until we get through all params */
6043	ecn_allowed = 0;
6044	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6045		/* mark all addresses that we have currently on the list */
6046		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6047	}
6048	/* does the source address already exist? if so skip it */
6049	inp = stcb->sctp_ep;
6050	atomic_add_int(&stcb->asoc.refcnt, 1);
6051	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6052	atomic_add_int(&stcb->asoc.refcnt, -1);
6053
6054	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6055		/* we must add the source address */
6056		/* no scope set here since we have a tcb already. */
6057		switch (sa->sa_family) {
6058#ifdef INET
6059		case AF_INET:
6060			if (stcb->asoc.scope.ipv4_addr_legal) {
6061				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6062					return (-1);
6063				}
6064			}
6065			break;
6066#endif
6067#ifdef INET6
6068		case AF_INET6:
6069			if (stcb->asoc.scope.ipv6_addr_legal) {
6070				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6071					return (-2);
6072				}
6073			}
6074			break;
6075#endif
6076		default:
6077			break;
6078		}
6079	} else {
6080		if (net_tmp != NULL && stcb_tmp == stcb) {
6081			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6082		} else if (stcb_tmp != stcb) {
6083			/* It belongs to another association? */
6084			if (stcb_tmp)
6085				SCTP_TCB_UNLOCK(stcb_tmp);
6086			return (-3);
6087		}
6088	}
6089	if (stcb->asoc.state == 0) {
6090		/* the assoc was freed? */
6091		return (-4);
6092	}
6093	/*
6094	 * peer must explicitly turn this on. This may have been initialized
6095	 * to be "on" in order to allow local addr changes while INIT's are
6096	 * in flight.
6097	 */
6098	stcb->asoc.peer_supports_asconf = 0;
6099	/* now we must go through each of the params. */
6100	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6101	while (phdr) {
6102		ptype = ntohs(phdr->param_type);
6103		plen = ntohs(phdr->param_length);
6104		/*
6105		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6106		 * (uint32_t)ptype, (int)plen);
6107		 */
6108		if (offset + plen > limit) {
6109			break;
6110		}
6111		if (plen == 0) {
6112			break;
6113		}
6114#ifdef INET
6115		if (ptype == SCTP_IPV4_ADDRESS) {
6116			if (stcb->asoc.scope.ipv4_addr_legal) {
6117				struct sctp_ipv4addr_param *p4, p4_buf;
6118
6119				/* ok get the v4 address and check/add */
6120				phdr = sctp_get_next_param(m, offset,
6121				    (struct sctp_paramhdr *)&p4_buf,
6122				    sizeof(p4_buf));
6123				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6124				    phdr == NULL) {
6125					return (-5);
6126				}
6127				p4 = (struct sctp_ipv4addr_param *)phdr;
6128				sin.sin_addr.s_addr = p4->addr;
6129				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6130					/* Skip multi-cast addresses */
6131					goto next_param;
6132				}
6133				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6134				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6135					goto next_param;
6136				}
6137				sa = (struct sockaddr *)&sin;
6138				inp = stcb->sctp_ep;
6139				atomic_add_int(&stcb->asoc.refcnt, 1);
6140				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6141				    dst, stcb);
6142				atomic_add_int(&stcb->asoc.refcnt, -1);
6143
6144				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6145				    inp == NULL) {
6146					/* we must add the source address */
6147					/*
6148					 * no scope set since we have a tcb
6149					 * already
6150					 */
6151
6152					/*
6153					 * we must validate the state again
6154					 * here
6155					 */
6156			add_it_now:
6157					if (stcb->asoc.state == 0) {
6158						/* the assoc was freed? */
6159						return (-7);
6160					}
6161					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6162						return (-8);
6163					}
6164				} else if (stcb_tmp == stcb) {
6165					if (stcb->asoc.state == 0) {
6166						/* the assoc was freed? */
6167						return (-10);
6168					}
6169					if (net != NULL) {
6170						/* clear flag */
6171						net->dest_state &=
6172						    ~SCTP_ADDR_NOT_IN_ASSOC;
6173					}
6174				} else {
6175					/*
6176					 * strange, address is in another
6177					 * assoc? straighten out locks.
6178					 */
6179					if (stcb_tmp) {
6180						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6181							/*
6182							 * in setup state we
6183							 * abort this guy
6184							 */
6185							sctp_abort_an_association(stcb_tmp->sctp_ep,
6186							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6187							goto add_it_now;
6188						}
6189						SCTP_TCB_UNLOCK(stcb_tmp);
6190					}
6191					if (stcb->asoc.state == 0) {
6192						/* the assoc was freed? */
6193						return (-12);
6194					}
6195					return (-13);
6196				}
6197			}
6198		} else
6199#endif
6200#ifdef INET6
6201		if (ptype == SCTP_IPV6_ADDRESS) {
6202			if (stcb->asoc.scope.ipv6_addr_legal) {
6203				/* ok get the v6 address and check/add */
6204				struct sctp_ipv6addr_param *p6, p6_buf;
6205
6206				phdr = sctp_get_next_param(m, offset,
6207				    (struct sctp_paramhdr *)&p6_buf,
6208				    sizeof(p6_buf));
6209				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6210				    phdr == NULL) {
6211					return (-14);
6212				}
6213				p6 = (struct sctp_ipv6addr_param *)phdr;
6214				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6215				    sizeof(p6->addr));
6216				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6217					/* Skip multi-cast addresses */
6218					goto next_param;
6219				}
6220				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6221					/*
6222					 * Link local make no sense without
6223					 * scope
6224					 */
6225					goto next_param;
6226				}
6227				sa = (struct sockaddr *)&sin6;
6228				inp = stcb->sctp_ep;
6229				atomic_add_int(&stcb->asoc.refcnt, 1);
6230				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6231				    dst, stcb);
6232				atomic_add_int(&stcb->asoc.refcnt, -1);
6233				if (stcb_tmp == NULL &&
6234				    (inp == stcb->sctp_ep || inp == NULL)) {
6235					/*
6236					 * we must validate the state again
6237					 * here
6238					 */
6239			add_it_now6:
6240					if (stcb->asoc.state == 0) {
6241						/* the assoc was freed? */
6242						return (-16);
6243					}
6244					/*
6245					 * we must add the address, no scope
6246					 * set
6247					 */
6248					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6249						return (-17);
6250					}
6251				} else if (stcb_tmp == stcb) {
6252					/*
6253					 * we must validate the state again
6254					 * here
6255					 */
6256					if (stcb->asoc.state == 0) {
6257						/* the assoc was freed? */
6258						return (-19);
6259					}
6260					if (net != NULL) {
6261						/* clear flag */
6262						net->dest_state &=
6263						    ~SCTP_ADDR_NOT_IN_ASSOC;
6264					}
6265				} else {
6266					/*
6267					 * strange, address is in another
6268					 * assoc? straighten out locks.
6269					 */
6270					if (stcb_tmp)
6271						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6272							/*
6273							 * in setup state we
6274							 * abort this guy
6275							 */
6276							sctp_abort_an_association(stcb_tmp->sctp_ep,
6277							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6278							goto add_it_now6;
6279						}
6280					SCTP_TCB_UNLOCK(stcb_tmp);
6281
6282					if (stcb->asoc.state == 0) {
6283						/* the assoc was freed? */
6284						return (-21);
6285					}
6286					return (-22);
6287				}
6288			}
6289		} else
6290#endif
6291		if (ptype == SCTP_ECN_CAPABLE) {
6292			ecn_allowed = 1;
6293		} else if (ptype == SCTP_ULP_ADAPTATION) {
6294			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6295				struct sctp_adaptation_layer_indication ai,
6296				                                *aip;
6297
6298				phdr = sctp_get_next_param(m, offset,
6299				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6300				aip = (struct sctp_adaptation_layer_indication *)phdr;
6301				if (aip) {
6302					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6303					stcb->asoc.adaptation_needed = 1;
6304				}
6305			}
6306		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6307			struct sctp_asconf_addr_param lstore, *fee;
6308			int lptype;
6309			struct sockaddr *lsa = NULL;
6310
6311#ifdef INET
6312			struct sctp_asconf_addrv4_param *fii;
6313
6314#endif
6315
6316			stcb->asoc.peer_supports_asconf = 1;
6317			if (plen > sizeof(lstore)) {
6318				return (-23);
6319			}
6320			phdr = sctp_get_next_param(m, offset,
6321			    (struct sctp_paramhdr *)&lstore,
6322			    min(plen, sizeof(lstore)));
6323			if (phdr == NULL) {
6324				return (-24);
6325			}
6326			fee = (struct sctp_asconf_addr_param *)phdr;
6327			lptype = ntohs(fee->addrp.ph.param_type);
6328			switch (lptype) {
6329#ifdef INET
6330			case SCTP_IPV4_ADDRESS:
6331				if (plen !=
6332				    sizeof(struct sctp_asconf_addrv4_param)) {
6333					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6334					    (int)sizeof(struct sctp_asconf_addrv4_param),
6335					    plen);
6336				} else {
6337					fii = (struct sctp_asconf_addrv4_param *)fee;
6338					sin.sin_addr.s_addr = fii->addrp.addr;
6339					lsa = (struct sockaddr *)&sin;
6340				}
6341				break;
6342#endif
6343#ifdef INET6
6344			case SCTP_IPV6_ADDRESS:
6345				if (plen !=
6346				    sizeof(struct sctp_asconf_addr_param)) {
6347					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6348					    (int)sizeof(struct sctp_asconf_addr_param),
6349					    plen);
6350				} else {
6351					memcpy(sin6.sin6_addr.s6_addr,
6352					    fee->addrp.addr,
6353					    sizeof(fee->addrp.addr));
6354					lsa = (struct sockaddr *)&sin6;
6355				}
6356				break;
6357#endif
6358			default:
6359				break;
6360			}
6361			if (lsa) {
6362				(void)sctp_set_primary_addr(stcb, sa, NULL);
6363			}
6364		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6365			stcb->asoc.peer_supports_nat = 1;
6366		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6367			/* Peer supports pr-sctp */
6368			stcb->asoc.peer_supports_prsctp = 1;
6369		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6370			/* A supported extension chunk */
6371			struct sctp_supported_chunk_types_param *pr_supported;
6372			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6373			int num_ent, i;
6374
6375			phdr = sctp_get_next_param(m, offset,
6376			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6377			if (phdr == NULL) {
6378				return (-25);
6379			}
6380			stcb->asoc.peer_supports_asconf = 0;
6381			stcb->asoc.peer_supports_prsctp = 0;
6382			stcb->asoc.peer_supports_pktdrop = 0;
6383			stcb->asoc.peer_supports_strreset = 0;
6384			stcb->asoc.peer_supports_nr_sack = 0;
6385			stcb->asoc.peer_supports_auth = 0;
6386			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6387			num_ent = plen - sizeof(struct sctp_paramhdr);
6388			for (i = 0; i < num_ent; i++) {
6389				switch (pr_supported->chunk_types[i]) {
6390				case SCTP_ASCONF:
6391				case SCTP_ASCONF_ACK:
6392					stcb->asoc.peer_supports_asconf = 1;
6393					break;
6394				case SCTP_FORWARD_CUM_TSN:
6395					stcb->asoc.peer_supports_prsctp = 1;
6396					break;
6397				case SCTP_PACKET_DROPPED:
6398					stcb->asoc.peer_supports_pktdrop = 1;
6399					break;
6400				case SCTP_NR_SELECTIVE_ACK:
6401					stcb->asoc.peer_supports_nr_sack = 1;
6402					break;
6403				case SCTP_STREAM_RESET:
6404					stcb->asoc.peer_supports_strreset = 1;
6405					break;
6406				case SCTP_AUTHENTICATION:
6407					stcb->asoc.peer_supports_auth = 1;
6408					break;
6409				default:
6410					/* one I have not learned yet */
6411					break;
6412
6413				}
6414			}
6415		} else if (ptype == SCTP_RANDOM) {
6416			if (plen > sizeof(random_store))
6417				break;
6418			if (got_random) {
6419				/* already processed a RANDOM */
6420				goto next_param;
6421			}
6422			phdr = sctp_get_next_param(m, offset,
6423			    (struct sctp_paramhdr *)random_store,
6424			    min(sizeof(random_store), plen));
6425			if (phdr == NULL)
6426				return (-26);
6427			p_random = (struct sctp_auth_random *)phdr;
6428			random_len = plen - sizeof(*p_random);
6429			/* enforce the random length */
6430			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6431				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6432				return (-27);
6433			}
6434			got_random = 1;
6435		} else if (ptype == SCTP_HMAC_LIST) {
6436			int num_hmacs;
6437			int i;
6438
6439			if (plen > sizeof(hmacs_store))
6440				break;
6441			if (got_hmacs) {
6442				/* already processed a HMAC list */
6443				goto next_param;
6444			}
6445			phdr = sctp_get_next_param(m, offset,
6446			    (struct sctp_paramhdr *)hmacs_store,
6447			    min(plen, sizeof(hmacs_store)));
6448			if (phdr == NULL)
6449				return (-28);
6450			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6451			hmacs_len = plen - sizeof(*hmacs);
6452			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6453			/* validate the hmac list */
6454			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6455				return (-29);
6456			}
6457			if (stcb->asoc.peer_hmacs != NULL)
6458				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6459			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6460			if (stcb->asoc.peer_hmacs != NULL) {
6461				for (i = 0; i < num_hmacs; i++) {
6462					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6463					    ntohs(hmacs->hmac_ids[i]));
6464				}
6465			}
6466			got_hmacs = 1;
6467		} else if (ptype == SCTP_CHUNK_LIST) {
6468			int i;
6469
6470			if (plen > sizeof(chunks_store))
6471				break;
6472			if (got_chklist) {
6473				/* already processed a Chunks list */
6474				goto next_param;
6475			}
6476			phdr = sctp_get_next_param(m, offset,
6477			    (struct sctp_paramhdr *)chunks_store,
6478			    min(plen, sizeof(chunks_store)));
6479			if (phdr == NULL)
6480				return (-30);
6481			chunks = (struct sctp_auth_chunk_list *)phdr;
6482			num_chunks = plen - sizeof(*chunks);
6483			if (stcb->asoc.peer_auth_chunks != NULL)
6484				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6485			else
6486				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6487			for (i = 0; i < num_chunks; i++) {
6488				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6489				    stcb->asoc.peer_auth_chunks);
6490				/* record asconf/asconf-ack if listed */
6491				if (chunks->chunk_types[i] == SCTP_ASCONF)
6492					saw_asconf = 1;
6493				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6494					saw_asconf_ack = 1;
6495
6496			}
6497			got_chklist = 1;
6498		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6499			    (ptype == SCTP_STATE_COOKIE) ||
6500			    (ptype == SCTP_UNRECOG_PARAM) ||
6501			    (ptype == SCTP_COOKIE_PRESERVE) ||
6502			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6503			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6504			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6505			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6506		    (ptype == SCTP_SUCCESS_REPORT)) {
6507			 /* don't care */ ;
6508		} else {
6509			if ((ptype & 0x8000) == 0x0000) {
6510				/*
6511				 * must stop processing the rest of the
6512				 * param's. Any report bits were handled
6513				 * with the call to
6514				 * sctp_arethere_unrecognized_parameters()
6515				 * when the INIT or INIT-ACK was first seen.
6516				 */
6517				break;
6518			}
6519		}
6520
6521next_param:
6522		offset += SCTP_SIZE32(plen);
6523		if (offset >= limit) {
6524			break;
6525		}
6526		phdr = sctp_get_next_param(m, offset, &parm_buf,
6527		    sizeof(parm_buf));
6528	}
6529	/* Now check to see if we need to purge any addresses */
6530	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6531		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6532		    SCTP_ADDR_NOT_IN_ASSOC) {
6533			/* This address has been removed from the asoc */
6534			/* remove and free it */
6535			stcb->asoc.numnets--;
6536			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6537			sctp_free_remote_addr(net);
6538			if (net == stcb->asoc.primary_destination) {
6539				stcb->asoc.primary_destination = NULL;
6540				sctp_select_primary_destination(stcb);
6541			}
6542		}
6543	}
6544	if (ecn_allowed == 0) {
6545		stcb->asoc.ecn_allowed = 0;
6546	}
6547	/* validate authentication required parameters */
6548	if (got_random && got_hmacs) {
6549		stcb->asoc.peer_supports_auth = 1;
6550	} else {
6551		stcb->asoc.peer_supports_auth = 0;
6552	}
6553	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6554		/* peer does not support auth but sent a chunks list? */
6555		return (-31);
6556	}
6557	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6558	    !stcb->asoc.peer_supports_auth) {
6559		/* peer supports asconf but not auth? */
6560		return (-32);
6561	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6562	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6563		return (-33);
6564	}
6565	/* concatenate the full random key */
6566	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6567	if (chunks != NULL) {
6568		keylen += sizeof(*chunks) + num_chunks;
6569	}
6570	new_key = sctp_alloc_key(keylen);
6571	if (new_key != NULL) {
6572		/* copy in the RANDOM */
6573		if (p_random != NULL) {
6574			keylen = sizeof(*p_random) + random_len;
6575			bcopy(p_random, new_key->key, keylen);
6576		}
6577		/* append in the AUTH chunks */
6578		if (chunks != NULL) {
6579			bcopy(chunks, new_key->key + keylen,
6580			    sizeof(*chunks) + num_chunks);
6581			keylen += sizeof(*chunks) + num_chunks;
6582		}
6583		/* append in the HMACs */
6584		if (hmacs != NULL) {
6585			bcopy(hmacs, new_key->key + keylen,
6586			    sizeof(*hmacs) + hmacs_len);
6587		}
6588	} else {
6589		/* failed to get memory for the key */
6590		return (-34);
6591	}
6592	if (stcb->asoc.authinfo.peer_random != NULL)
6593		sctp_free_key(stcb->asoc.authinfo.peer_random);
6594	stcb->asoc.authinfo.peer_random = new_key;
6595	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6596	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6597
6598	return (0);
6599}
6600
6601int
6602sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6603    struct sctp_nets *net)
6604{
6605	/* make sure the requested primary address exists in the assoc */
6606	if (net == NULL && sa)
6607		net = sctp_findnet(stcb, sa);
6608
6609	if (net == NULL) {
6610		/* didn't find the requested primary address! */
6611		return (-1);
6612	} else {
6613		/* set the primary address */
6614		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6615			/* Must be confirmed, so queue to set */
6616			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6617			return (0);
6618		}
6619		stcb->asoc.primary_destination = net;
6620		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6621			sctp_free_remote_addr(stcb->asoc.alternate);
6622			stcb->asoc.alternate = NULL;
6623		}
6624		net = TAILQ_FIRST(&stcb->asoc.nets);
6625		if (net != stcb->asoc.primary_destination) {
6626			/*
6627			 * first one on the list is NOT the primary
6628			 * sctp_cmpaddr() is much more efficient if the
6629			 * primary is the first on the list, make it so.
6630			 */
6631			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6632			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6633		}
6634		return (0);
6635	}
6636}
6637
6638int
6639sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6640{
6641	/*
6642	 * This function serves two purposes. It will see if a TAG can be
6643	 * re-used and return 1 for yes it is ok and 0 for don't use that
6644	 * tag. A secondary function it will do is purge out old tags that
6645	 * can be removed.
6646	 */
6647	struct sctpvtaghead *chain;
6648	struct sctp_tagblock *twait_block;
6649	struct sctpasochead *head;
6650	struct sctp_tcb *stcb;
6651	int i;
6652
6653	SCTP_INP_INFO_RLOCK();
6654	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6655	    SCTP_BASE_INFO(hashasocmark))];
6656	if (head == NULL) {
6657		/* invalid vtag */
6658		goto skip_vtag_check;
6659	}
6660	LIST_FOREACH(stcb, head, sctp_asocs) {
6661		/*
6662		 * We choose not to lock anything here. TCB's can't be
6663		 * removed since we have the read lock, so they can't be
6664		 * freed on us, same thing for the INP. I may be wrong with
6665		 * this assumption, but we will go with it for now :-)
6666		 */
6667		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6668			continue;
6669		}
6670		if (stcb->asoc.my_vtag == tag) {
6671			/* candidate */
6672			if (stcb->rport != rport) {
6673				continue;
6674			}
6675			if (stcb->sctp_ep->sctp_lport != lport) {
6676				continue;
6677			}
6678			/* Its a used tag set */
6679			SCTP_INP_INFO_RUNLOCK();
6680			return (0);
6681		}
6682	}
6683skip_vtag_check:
6684
6685	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6686	/* Now what about timed wait ? */
6687	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6688		/*
6689		 * Block(s) are present, lets see if we have this tag in the
6690		 * list
6691		 */
6692		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6693			if (twait_block->vtag_block[i].v_tag == 0) {
6694				/* not used */
6695				continue;
6696			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6697			    now->tv_sec) {
6698				/* Audit expires this guy */
6699				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6700				twait_block->vtag_block[i].v_tag = 0;
6701				twait_block->vtag_block[i].lport = 0;
6702				twait_block->vtag_block[i].rport = 0;
6703			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6704				    (twait_block->vtag_block[i].lport == lport) &&
6705			    (twait_block->vtag_block[i].rport == rport)) {
6706				/* Bad tag, sorry :< */
6707				SCTP_INP_INFO_RUNLOCK();
6708				return (0);
6709			}
6710		}
6711	}
6712	SCTP_INP_INFO_RUNLOCK();
6713	return (1);
6714}
6715
6716static void
6717sctp_drain_mbufs(struct sctp_tcb *stcb)
6718{
6719	/*
6720	 * We must hunt this association for MBUF's past the cumack (i.e.
6721	 * out of order data that we can renege on).
6722	 */
6723	struct sctp_association *asoc;
6724	struct sctp_tmit_chunk *chk, *nchk;
6725	uint32_t cumulative_tsn_p1;
6726	struct sctp_queued_to_read *ctl, *nctl;
6727	int cnt, strmat;
6728	uint32_t gap, i;
6729	int fnd = 0;
6730
6731	/* We look for anything larger than the cum-ack + 1 */
6732
6733	asoc = &stcb->asoc;
6734	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6735		/* none we can reneg on. */
6736		return;
6737	}
6738	SCTP_STAT_INCR(sctps_protocol_drains_done);
6739	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6740	cnt = 0;
6741	/* First look in the re-assembly queue */
6742	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6743		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6744			/* Yep it is above cum-ack */
6745			cnt++;
6746			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6747			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6748			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6749			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6750			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6751			if (chk->data) {
6752				sctp_m_freem(chk->data);
6753				chk->data = NULL;
6754			}
6755			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6756		}
6757	}
6758	/* Ok that was fun, now we will drain all the inbound streams? */
6759	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6760		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6761			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6762				/* Yep it is above cum-ack */
6763				cnt++;
6764				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6765				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6766				sctp_ucount_decr(asoc->cnt_on_all_streams);
6767				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6768				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6769				if (ctl->data) {
6770					sctp_m_freem(ctl->data);
6771					ctl->data = NULL;
6772				}
6773				sctp_free_remote_addr(ctl->whoFrom);
6774				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6775				SCTP_DECR_READQ_COUNT();
6776			}
6777		}
6778	}
6779	if (cnt) {
6780		/* We must back down to see what the new highest is */
6781		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6782			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6783			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6784				asoc->highest_tsn_inside_map = i;
6785				fnd = 1;
6786				break;
6787			}
6788		}
6789		if (!fnd) {
6790			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6791		}
6792		/*
6793		 * Question, should we go through the delivery queue? The
6794		 * only reason things are on here is the app not reading OR
6795		 * a p-d-api up. An attacker COULD send enough in to
6796		 * initiate the PD-API and then send a bunch of stuff to
6797		 * other streams... these would wind up on the delivery
6798		 * queue.. and then we would not get to them. But in order
6799		 * to do this I then have to back-track and un-deliver
6800		 * sequence numbers in streams.. el-yucko. I think for now
6801		 * we will NOT look at the delivery queue and leave it to be
6802		 * something to consider later. An alternative would be to
6803		 * abort the P-D-API with a notification and then deliver
6804		 * the data.... Or another method might be to keep track of
6805		 * how many times the situation occurs and if we see a
6806		 * possible attack underway just abort the association.
6807		 */
6808#ifdef SCTP_DEBUG
6809		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6810#endif
6811		/*
6812		 * Now do we need to find a new
6813		 * asoc->highest_tsn_inside_map?
6814		 */
6815		asoc->last_revoke_count = cnt;
6816		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6817		/* sa_ignore NO_NULL_CHK */
6818		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6819		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6820	}
6821	/*
6822	 * Another issue, in un-setting the TSN's in the mapping array we
6823	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6824	 * two things to occur. It may cause us to do extra work in checking
6825	 * for our mapping array movement. More importantly it may cause us
6826	 * to SACK every datagram. This may not be a bad thing though since
6827	 * we will recover once we get our cum-ack above and all this stuff
6828	 * we dumped recovered.
6829	 */
6830}
6831
6832void
6833sctp_drain()
6834{
6835	/*
6836	 * We must walk the PCB lists for ALL associations here. The system
6837	 * is LOW on MBUF's and needs help. This is where reneging will
6838	 * occur. We really hope this does NOT happen!
6839	 */
6840	VNET_ITERATOR_DECL(vnet_iter);
6841	VNET_LIST_RLOCK_NOSLEEP();
6842	VNET_FOREACH(vnet_iter) {
6843		CURVNET_SET(vnet_iter);
6844		struct sctp_inpcb *inp;
6845		struct sctp_tcb *stcb;
6846
6847		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6848		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6849#ifdef VIMAGE
6850			continue;
6851#else
6852			return;
6853#endif
6854		}
6855		SCTP_INP_INFO_RLOCK();
6856		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6857			/* For each endpoint */
6858			SCTP_INP_RLOCK(inp);
6859			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6860				/* For each association */
6861				SCTP_TCB_LOCK(stcb);
6862				sctp_drain_mbufs(stcb);
6863				SCTP_TCB_UNLOCK(stcb);
6864			}
6865			SCTP_INP_RUNLOCK(inp);
6866		}
6867		SCTP_INP_INFO_RUNLOCK();
6868		CURVNET_RESTORE();
6869	}
6870	VNET_LIST_RUNLOCK_NOSLEEP();
6871}
6872
6873/*
6874 * start a new iterator
6875 * iterates through all endpoints and associations based on the pcb_state
6876 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6877 * assocs and "ef" (optional) is executed when the iterator completes.
6878 * "inpf" (optional) is executed for each new endpoint as it is being
6879 * iterated through. inpe (optional) is called when the inp completes
6880 * its way through all the stcbs.
6881 */
6882int
6883sctp_initiate_iterator(inp_func inpf,
6884    asoc_func af,
6885    inp_func inpe,
6886    uint32_t pcb_state,
6887    uint32_t pcb_features,
6888    uint32_t asoc_state,
6889    void *argp,
6890    uint32_t argi,
6891    end_func ef,
6892    struct sctp_inpcb *s_inp,
6893    uint8_t chunk_output_off)
6894{
6895	struct sctp_iterator *it = NULL;
6896
6897	if (af == NULL) {
6898		return (-1);
6899	}
6900	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6901	    SCTP_M_ITER);
6902	if (it == NULL) {
6903		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6904		return (ENOMEM);
6905	}
6906	memset(it, 0, sizeof(*it));
6907	it->function_assoc = af;
6908	it->function_inp = inpf;
6909	if (inpf)
6910		it->done_current_ep = 0;
6911	else
6912		it->done_current_ep = 1;
6913	it->function_atend = ef;
6914	it->pointer = argp;
6915	it->val = argi;
6916	it->pcb_flags = pcb_state;
6917	it->pcb_features = pcb_features;
6918	it->asoc_state = asoc_state;
6919	it->function_inp_end = inpe;
6920	it->no_chunk_output = chunk_output_off;
6921	it->vn = curvnet;
6922	if (s_inp) {
6923		/* Assume lock is held here */
6924		it->inp = s_inp;
6925		SCTP_INP_INCR_REF(it->inp);
6926		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
6927	} else {
6928		SCTP_INP_INFO_RLOCK();
6929		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
6930		if (it->inp) {
6931			SCTP_INP_INCR_REF(it->inp);
6932		}
6933		SCTP_INP_INFO_RUNLOCK();
6934		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
6935
6936	}
6937	SCTP_IPI_ITERATOR_WQ_LOCK();
6938
6939	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
6940	if (sctp_it_ctl.iterator_running == 0) {
6941		sctp_wakeup_iterator();
6942	}
6943	SCTP_IPI_ITERATOR_WQ_UNLOCK();
6944	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
6945	return (0);
6946}
6947