xref: /freebsd/sys/netinet6/ip6_mroute.c (revision a812392203d7c4c3f0db9d8a0f3391374c49c71f)
1 /*-
2  * Copyright (C) 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1989 Stephen Deering
34  * Copyright (c) 1992, 1993
35  *      The Regents of the University of California.  All rights reserved.
36  *
37  * This code is derived from software contributed to Berkeley by
38  * Stephen Deering of Stanford University.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 4. Neither the name of the University nor the names of its contributors
49  *    may be used to endorse or promote products derived from this software
50  *    without specific prior written permission.
51  *
52  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62  * SUCH DAMAGE.
63  *
64  *	@(#)ip_mroute.c	8.2 (Berkeley) 11/15/93
65  *	BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp
66  */
67 
68 /*
69  * IP multicast forwarding procedures
70  *
71  * Written by David Waitzman, BBN Labs, August 1988.
72  * Modified by Steve Deering, Stanford, February 1989.
73  * Modified by Mark J. Steiglitz, Stanford, May, 1991
74  * Modified by Van Jacobson, LBL, January 1993
75  * Modified by Ajit Thyagarajan, PARC, August 1993
76  * Modified by Bill Fenner, PARC, April 1994
77  *
78  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
79  */
80 
81 #include <sys/cdefs.h>
82 __FBSDID("$FreeBSD$");
83 
84 #include "opt_inet6.h"
85 
86 #include <sys/param.h>
87 #include <sys/callout.h>
88 #include <sys/errno.h>
89 #include <sys/kernel.h>
90 #include <sys/lock.h>
91 #include <sys/malloc.h>
92 #include <sys/mbuf.h>
93 #include <sys/module.h>
94 #include <sys/domain.h>
95 #include <sys/protosw.h>
96 #include <sys/sdt.h>
97 #include <sys/signalvar.h>
98 #include <sys/socket.h>
99 #include <sys/socketvar.h>
100 #include <sys/sockio.h>
101 #include <sys/sx.h>
102 #include <sys/sysctl.h>
103 #include <sys/syslog.h>
104 #include <sys/systm.h>
105 #include <sys/time.h>
106 
107 #include <net/if.h>
108 #include <net/if_var.h>
109 #include <net/if_types.h>
110 #include <net/raw_cb.h>
111 #include <net/vnet.h>
112 
113 #include <netinet/in.h>
114 #include <netinet/in_var.h>
115 #include <netinet/icmp6.h>
116 #include <netinet/ip_encap.h>
117 
118 #include <netinet/ip6.h>
119 #include <netinet/in_kdtrace.h>
120 #include <netinet6/ip6_var.h>
121 #include <netinet6/scope6_var.h>
122 #include <netinet6/nd6.h>
123 #include <netinet6/ip6_mroute.h>
124 #include <netinet6/pim6.h>
125 #include <netinet6/pim6_var.h>
126 
127 static MALLOC_DEFINE(M_MRTABLE6, "mf6c", "multicast forwarding cache entry");
128 
129 static int	ip6_mdq(struct mbuf *, struct ifnet *, struct mf6c *);
130 static void	phyint_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
131 static int	register_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
132 static int	set_pim6(int *);
133 static int	socket_send(struct socket *, struct mbuf *,
134 		    struct sockaddr_in6 *);
135 
136 extern int in6_mcast_loop;
137 extern struct domain inet6domain;
138 
139 static const struct encaptab *pim6_encap_cookie;
140 static const struct protosw in6_pim_protosw = {
141 	.pr_type =		SOCK_RAW,
142 	.pr_domain =		&inet6domain,
143 	.pr_protocol =		IPPROTO_PIM,
144 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
145 	.pr_input =		pim6_input,
146 	.pr_output =		rip6_output,
147 	.pr_ctloutput =		rip6_ctloutput,
148 	.pr_usrreqs =		&rip6_usrreqs
149 };
150 static int pim6_encapcheck(const struct mbuf *, int, int, void *);
151 
152 static VNET_DEFINE(int, ip6_mrouter_ver) = 0;
153 #define	V_ip6_mrouter_ver	VNET(ip6_mrouter_ver)
154 
155 SYSCTL_DECL(_net_inet6);
156 SYSCTL_DECL(_net_inet6_ip6);
157 static SYSCTL_NODE(_net_inet6, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
158 
159 static struct mrt6stat mrt6stat;
160 SYSCTL_STRUCT(_net_inet6_ip6, OID_AUTO, mrt6stat, CTLFLAG_RW,
161     &mrt6stat, mrt6stat,
162     "Multicast Routing Statistics (struct mrt6stat, netinet6/ip6_mroute.h)");
163 
164 #define	MRT6STAT_INC(name)	mrt6stat.name += 1
165 #define NO_RTE_FOUND	0x1
166 #define RTE_FOUND	0x2
167 
168 static struct mtx mrouter6_mtx;
169 #define	MROUTER6_LOCK()		mtx_lock(&mrouter6_mtx)
170 #define	MROUTER6_UNLOCK()	mtx_unlock(&mrouter6_mtx)
171 #define	MROUTER6_LOCK_ASSERT()	do {					\
172 	mtx_assert(&mrouter6_mtx, MA_OWNED);				\
173 	NET_ASSERT_GIANT();						\
174 } while (0)
175 #define	MROUTER6_LOCK_INIT()	\
176 	mtx_init(&mrouter6_mtx, "IPv6 multicast forwarding", NULL, MTX_DEF)
177 #define	MROUTER6_LOCK_DESTROY()	mtx_destroy(&mrouter6_mtx)
178 
179 static struct mf6c *mf6ctable[MF6CTBLSIZ];
180 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mf6ctable, CTLFLAG_RD,
181     &mf6ctable, sizeof(mf6ctable), "S,*mf6ctable[MF6CTBLSIZ]",
182     "IPv6 Multicast Forwarding Table (struct *mf6ctable[MF6CTBLSIZ], "
183     "netinet6/ip6_mroute.h)");
184 
185 static struct mtx mfc6_mtx;
186 #define	MFC6_LOCK()		mtx_lock(&mfc6_mtx)
187 #define	MFC6_UNLOCK()		mtx_unlock(&mfc6_mtx)
188 #define	MFC6_LOCK_ASSERT()	do {					\
189 	mtx_assert(&mfc6_mtx, MA_OWNED);				\
190 	NET_ASSERT_GIANT();						\
191 } while (0)
192 #define	MFC6_LOCK_INIT()		\
193 	mtx_init(&mfc6_mtx, "IPv6 multicast forwarding cache", NULL, MTX_DEF)
194 #define	MFC6_LOCK_DESTROY()	mtx_destroy(&mfc6_mtx)
195 
196 static u_char n6expire[MF6CTBLSIZ];
197 
198 static struct mif6 mif6table[MAXMIFS];
199 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mif6table, CTLFLAG_RD,
200     &mif6table, sizeof(mif6table), "S,mif6[MAXMIFS]",
201     "IPv6 Multicast Interfaces (struct mif6[MAXMIFS], netinet6/ip6_mroute.h)");
202 
203 static struct mtx mif6_mtx;
204 #define	MIF6_LOCK()		mtx_lock(&mif6_mtx)
205 #define	MIF6_UNLOCK()		mtx_unlock(&mif6_mtx)
206 #define	MIF6_LOCK_ASSERT()	mtx_assert(&mif6_mtx, MA_OWNED)
207 #define	MIF6_LOCK_INIT()	\
208 	mtx_init(&mif6_mtx, "IPv6 multicast interfaces", NULL, MTX_DEF)
209 #define	MIF6_LOCK_DESTROY()	mtx_destroy(&mif6_mtx)
210 
211 #ifdef MRT6DEBUG
212 static VNET_DEFINE(u_int, mrt6debug) = 0;	/* debug level */
213 #define	V_mrt6debug		VNET(mrt6debug)
214 #define DEBUG_MFC	0x02
215 #define DEBUG_FORWARD	0x04
216 #define DEBUG_EXPIRE	0x08
217 #define DEBUG_XMIT	0x10
218 #define DEBUG_REG	0x20
219 #define DEBUG_PIM	0x40
220 #define	DEBUG_ERR	0x80
221 #define	DEBUG_ANY	0x7f
222 #define	MRT6_DLOG(m, fmt, ...)	\
223 	if (V_mrt6debug & (m))	\
224 		log(((m) & DEBUG_ERR) ? LOG_ERR: LOG_DEBUG, \
225 		    "%s: " fmt "\n", __func__, ##__VA_ARGS__)
226 #else
227 #define	MRT6_DLOG(m, fmt, ...)
228 #endif
229 
230 static void	expire_upcalls(void *);
231 #define	EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
232 #define	UPCALL_EXPIRE	6		/* number of timeouts */
233 
234 /*
235  * XXX TODO: maintain a count to if_allmulti() calls in struct ifnet.
236  */
237 
238 /*
239  * 'Interfaces' associated with decapsulator (so we can tell
240  * packets that went through it from ones that get reflected
241  * by a broken gateway).  Different from IPv4 register_if,
242  * these interfaces are linked into the system ifnet list,
243  * because per-interface IPv6 statistics are maintained in
244  * ifp->if_afdata.  But it does not have any routes point
245  * to them.  I.e., packets can't be sent this way.  They
246  * only exist as a placeholder for multicast source
247  * verification.
248  */
249 static struct ifnet *multicast_register_if6;
250 
251 #define ENCAP_HOPS 64
252 
253 /*
254  * Private variables.
255  */
256 static mifi_t nummifs = 0;
257 static mifi_t reg_mif_num = (mifi_t)-1;
258 
259 static struct pim6stat pim6stat;
260 SYSCTL_STRUCT(_net_inet6_pim, PIM6CTL_STATS, stats, CTLFLAG_RW,
261     &pim6stat, pim6stat,
262     "PIM Statistics (struct pim6stat, netinet6/pim6_var.h)");
263 
264 #define	PIM6STAT_INC(name)	pim6stat.name += 1
265 static VNET_DEFINE(int, pim6);
266 #define	V_pim6		VNET(pim6)
267 
268 /*
269  * Hash function for a source, group entry
270  */
271 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
272 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
273 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
274 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
275 
276 /*
277  * Find a route for a given origin IPv6 address and Multicast group address.
278  */
279 #define MF6CFIND(o, g, rt) do { \
280 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
281 	rt = NULL; \
282 	while (_rt) { \
283 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
284 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
285 		    (_rt->mf6c_stall == NULL)) { \
286 			rt = _rt; \
287 			break; \
288 		} \
289 		_rt = _rt->mf6c_next; \
290 	} \
291 	if (rt == NULL) { \
292 		MRT6STAT_INC(mrt6s_mfc_misses); \
293 	} \
294 } while (/*CONSTCOND*/ 0)
295 
296 /*
297  * Macros to compute elapsed time efficiently
298  * Borrowed from Van Jacobson's scheduling code
299  * XXX: replace with timersub() ?
300  */
301 #define TV_DELTA(a, b, delta) do { \
302 	    int xxs; \
303 		\
304 	    delta = (a).tv_usec - (b).tv_usec; \
305 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
306 	       switch (xxs) { \
307 		      case 2: \
308 			  delta += 1000000; \
309 			      /* FALLTHROUGH */ \
310 		      case 1: \
311 			  delta += 1000000; \
312 			  break; \
313 		      default: \
314 			  delta += (1000000 * xxs); \
315 	       } \
316 	    } \
317 } while (/*CONSTCOND*/ 0)
318 
319 /* XXX: replace with timercmp(a, b, <) ? */
320 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
321 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
322 
323 #ifdef UPCALL_TIMING
324 #define UPCALL_MAX	50
325 static u_long upcall_data[UPCALL_MAX + 1];
326 static void collate();
327 #endif /* UPCALL_TIMING */
328 
329 static int ip6_mrouter_init(struct socket *, int, int);
330 static int add_m6fc(struct mf6cctl *);
331 static int add_m6if(struct mif6ctl *);
332 static int del_m6fc(struct mf6cctl *);
333 static int del_m6if(mifi_t *);
334 static int del_m6if_locked(mifi_t *);
335 static int get_mif6_cnt(struct sioc_mif_req6 *);
336 static int get_sg_cnt(struct sioc_sg_req6 *);
337 
338 static struct callout expire_upcalls_ch;
339 
340 int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *);
341 int X_ip6_mrouter_done(void);
342 int X_ip6_mrouter_set(struct socket *, struct sockopt *);
343 int X_ip6_mrouter_get(struct socket *, struct sockopt *);
344 int X_mrt6_ioctl(u_long, caddr_t);
345 
346 /*
347  * Handle MRT setsockopt commands to modify the multicast routing tables.
348  */
349 int
350 X_ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
351 {
352 	int error = 0;
353 	int optval;
354 	struct mif6ctl mifc;
355 	struct mf6cctl mfcc;
356 	mifi_t mifi;
357 
358 	if (so != V_ip6_mrouter && sopt->sopt_name != MRT6_INIT)
359 		return (EPERM);
360 
361 	switch (sopt->sopt_name) {
362 	case MRT6_INIT:
363 #ifdef MRT6_OINIT
364 	case MRT6_OINIT:
365 #endif
366 		error = sooptcopyin(sopt, &optval, sizeof(optval),
367 		    sizeof(optval));
368 		if (error)
369 			break;
370 		error = ip6_mrouter_init(so, optval, sopt->sopt_name);
371 		break;
372 	case MRT6_DONE:
373 		error = X_ip6_mrouter_done();
374 		break;
375 	case MRT6_ADD_MIF:
376 		error = sooptcopyin(sopt, &mifc, sizeof(mifc), sizeof(mifc));
377 		if (error)
378 			break;
379 		error = add_m6if(&mifc);
380 		break;
381 	case MRT6_ADD_MFC:
382 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
383 		if (error)
384 			break;
385 		error = add_m6fc(&mfcc);
386 		break;
387 	case MRT6_DEL_MFC:
388 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
389 		if (error)
390 			break;
391 		error = del_m6fc(&mfcc);
392 		break;
393 	case MRT6_DEL_MIF:
394 		error = sooptcopyin(sopt, &mifi, sizeof(mifi), sizeof(mifi));
395 		if (error)
396 			break;
397 		error = del_m6if(&mifi);
398 		break;
399 	case MRT6_PIM:
400 		error = sooptcopyin(sopt, &optval, sizeof(optval),
401 		    sizeof(optval));
402 		if (error)
403 			break;
404 		error = set_pim6(&optval);
405 		break;
406 	default:
407 		error = EOPNOTSUPP;
408 		break;
409 	}
410 
411 	return (error);
412 }
413 
414 /*
415  * Handle MRT getsockopt commands
416  */
417 int
418 X_ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
419 {
420 	int error = 0;
421 
422 	if (so != V_ip6_mrouter)
423 		return (EACCES);
424 
425 	switch (sopt->sopt_name) {
426 		case MRT6_PIM:
427 			error = sooptcopyout(sopt, &V_pim6, sizeof(V_pim6));
428 			break;
429 	}
430 	return (error);
431 }
432 
433 /*
434  * Handle ioctl commands to obtain information from the cache
435  */
436 int
437 X_mrt6_ioctl(u_long cmd, caddr_t data)
438 {
439 	int ret;
440 
441 	ret = EINVAL;
442 
443 	switch (cmd) {
444 	case SIOCGETSGCNT_IN6:
445 		ret = get_sg_cnt((struct sioc_sg_req6 *)data);
446 		break;
447 
448 	case SIOCGETMIFCNT_IN6:
449 		ret = get_mif6_cnt((struct sioc_mif_req6 *)data);
450 		break;
451 
452 	default:
453 		break;
454 	}
455 
456 	return (ret);
457 }
458 
459 /*
460  * returns the packet, byte, rpf-failure count for the source group provided
461  */
462 static int
463 get_sg_cnt(struct sioc_sg_req6 *req)
464 {
465 	struct mf6c *rt;
466 	int ret;
467 
468 	ret = 0;
469 
470 	MFC6_LOCK();
471 
472 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
473 	if (rt == NULL) {
474 		ret = ESRCH;
475 	} else {
476 		req->pktcnt = rt->mf6c_pkt_cnt;
477 		req->bytecnt = rt->mf6c_byte_cnt;
478 		req->wrong_if = rt->mf6c_wrong_if;
479 	}
480 
481 	MFC6_UNLOCK();
482 
483 	return (ret);
484 }
485 
486 /*
487  * returns the input and output packet and byte counts on the mif provided
488  */
489 static int
490 get_mif6_cnt(struct sioc_mif_req6 *req)
491 {
492 	mifi_t mifi;
493 	int ret;
494 
495 	ret = 0;
496 	mifi = req->mifi;
497 
498 	MIF6_LOCK();
499 
500 	if (mifi >= nummifs) {
501 		ret = EINVAL;
502 	} else {
503 		req->icount = mif6table[mifi].m6_pkt_in;
504 		req->ocount = mif6table[mifi].m6_pkt_out;
505 		req->ibytes = mif6table[mifi].m6_bytes_in;
506 		req->obytes = mif6table[mifi].m6_bytes_out;
507 	}
508 
509 	MIF6_UNLOCK();
510 
511 	return (ret);
512 }
513 
514 static int
515 set_pim6(int *i)
516 {
517 	if ((*i != 1) && (*i != 0))
518 		return (EINVAL);
519 
520 	V_pim6 = *i;
521 
522 	return (0);
523 }
524 
525 /*
526  * Enable multicast routing
527  */
528 static int
529 ip6_mrouter_init(struct socket *so, int v, int cmd)
530 {
531 
532 	MRT6_DLOG(DEBUG_ANY, "so_type = %d, pr_protocol = %d",
533 	    so->so_type, so->so_proto->pr_protocol);
534 
535 	if (so->so_type != SOCK_RAW ||
536 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
537 		return (EOPNOTSUPP);
538 
539 	if (v != 1)
540 		return (ENOPROTOOPT);
541 
542 	MROUTER6_LOCK();
543 
544 	if (V_ip6_mrouter != NULL) {
545 		MROUTER6_UNLOCK();
546 		return (EADDRINUSE);
547 	}
548 
549 	V_ip6_mrouter = so;
550 	V_ip6_mrouter_ver = cmd;
551 
552 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
553 	bzero((caddr_t)n6expire, sizeof(n6expire));
554 
555 	V_pim6 = 0;/* used for stubbing out/in pim stuff */
556 
557 	callout_init(&expire_upcalls_ch, 0);
558 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
559 	    expire_upcalls, NULL);
560 
561 	MROUTER6_UNLOCK();
562 	MRT6_DLOG(DEBUG_ANY, "finished");
563 
564 	return (0);
565 }
566 
567 /*
568  * Disable IPv6 multicast forwarding.
569  */
570 int
571 X_ip6_mrouter_done(void)
572 {
573 	mifi_t mifi;
574 	u_long i;
575 	struct mf6c *rt;
576 	struct rtdetq *rte;
577 
578 	MROUTER6_LOCK();
579 
580 	if (V_ip6_mrouter == NULL) {
581 		MROUTER6_UNLOCK();
582 		return (EINVAL);
583 	}
584 
585 	/*
586 	 * For each phyint in use, disable promiscuous reception of all IPv6
587 	 * multicasts.
588 	 */
589 	for (mifi = 0; mifi < nummifs; mifi++) {
590 		if (mif6table[mifi].m6_ifp &&
591 		    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
592 			if_allmulti(mif6table[mifi].m6_ifp, 0);
593 		}
594 	}
595 	bzero((caddr_t)mif6table, sizeof(mif6table));
596 	nummifs = 0;
597 
598 	V_pim6 = 0; /* used to stub out/in pim specific code */
599 
600 	callout_stop(&expire_upcalls_ch);
601 
602 	/*
603 	 * Free all multicast forwarding cache entries.
604 	 */
605 	MFC6_LOCK();
606 	for (i = 0; i < MF6CTBLSIZ; i++) {
607 		rt = mf6ctable[i];
608 		while (rt) {
609 			struct mf6c *frt;
610 
611 			for (rte = rt->mf6c_stall; rte != NULL; ) {
612 				struct rtdetq *n = rte->next;
613 
614 				m_freem(rte->m);
615 				free(rte, M_MRTABLE6);
616 				rte = n;
617 			}
618 			frt = rt;
619 			rt = rt->mf6c_next;
620 			free(frt, M_MRTABLE6);
621 		}
622 	}
623 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
624 	MFC6_UNLOCK();
625 
626 	/*
627 	 * Reset register interface
628 	 */
629 	if (reg_mif_num != (mifi_t)-1 && multicast_register_if6 != NULL) {
630 		if_detach(multicast_register_if6);
631 		if_free(multicast_register_if6);
632 		reg_mif_num = (mifi_t)-1;
633 		multicast_register_if6 = NULL;
634 	}
635 
636 	V_ip6_mrouter = NULL;
637 	V_ip6_mrouter_ver = 0;
638 
639 	MROUTER6_UNLOCK();
640 	MRT6_DLOG(DEBUG_ANY, "finished");
641 
642 	return (0);
643 }
644 
645 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
646 
647 /*
648  * Add a mif to the mif table
649  */
650 static int
651 add_m6if(struct mif6ctl *mifcp)
652 {
653 	struct mif6 *mifp;
654 	struct ifnet *ifp;
655 	int error;
656 
657 	MIF6_LOCK();
658 
659 	if (mifcp->mif6c_mifi >= MAXMIFS) {
660 		MIF6_UNLOCK();
661 		return (EINVAL);
662 	}
663 	mifp = mif6table + mifcp->mif6c_mifi;
664 	if (mifp->m6_ifp != NULL) {
665 		MIF6_UNLOCK();
666 		return (EADDRINUSE); /* XXX: is it appropriate? */
667 	}
668 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > V_if_index) {
669 		MIF6_UNLOCK();
670 		return (ENXIO);
671 	}
672 
673 	ifp = ifnet_byindex(mifcp->mif6c_pifi);
674 
675 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
676 		if (reg_mif_num == (mifi_t)-1) {
677 			ifp = if_alloc(IFT_OTHER);
678 
679 			if_initname(ifp, "register_mif", 0);
680 			ifp->if_flags |= IFF_LOOPBACK;
681 			if_attach(ifp);
682 			multicast_register_if6 = ifp;
683 			reg_mif_num = mifcp->mif6c_mifi;
684 			/*
685 			 * it is impossible to guess the ifindex of the
686 			 * register interface.  So mif6c_pifi is automatically
687 			 * calculated.
688 			 */
689 			mifcp->mif6c_pifi = ifp->if_index;
690 		} else {
691 			ifp = multicast_register_if6;
692 		}
693 	} else {
694 		/* Make sure the interface supports multicast */
695 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
696 			MIF6_UNLOCK();
697 			return (EOPNOTSUPP);
698 		}
699 
700 		error = if_allmulti(ifp, 1);
701 		if (error) {
702 			MIF6_UNLOCK();
703 			return (error);
704 		}
705 	}
706 
707 	mifp->m6_flags     = mifcp->mif6c_flags;
708 	mifp->m6_ifp       = ifp;
709 
710 	/* initialize per mif pkt counters */
711 	mifp->m6_pkt_in    = 0;
712 	mifp->m6_pkt_out   = 0;
713 	mifp->m6_bytes_in  = 0;
714 	mifp->m6_bytes_out = 0;
715 
716 	/* Adjust nummifs up if the mifi is higher than nummifs */
717 	if (nummifs <= mifcp->mif6c_mifi)
718 		nummifs = mifcp->mif6c_mifi + 1;
719 
720 	MIF6_UNLOCK();
721 	MRT6_DLOG(DEBUG_ANY, "mif #%d, phyint %s", mifcp->mif6c_mifi,
722 	    if_name(ifp));
723 
724 	return (0);
725 }
726 
727 /*
728  * Delete a mif from the mif table
729  */
730 static int
731 del_m6if_locked(mifi_t *mifip)
732 {
733 	struct mif6 *mifp = mif6table + *mifip;
734 	mifi_t mifi;
735 	struct ifnet *ifp;
736 
737 	MIF6_LOCK_ASSERT();
738 
739 	if (*mifip >= nummifs)
740 		return (EINVAL);
741 	if (mifp->m6_ifp == NULL)
742 		return (EINVAL);
743 
744 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
745 		/* XXX: TODO: Maintain an ALLMULTI refcount in struct ifnet. */
746 		ifp = mifp->m6_ifp;
747 		if_allmulti(ifp, 0);
748 	} else {
749 		if (reg_mif_num != (mifi_t)-1 &&
750 		    multicast_register_if6 != NULL) {
751 			if_detach(multicast_register_if6);
752 			if_free(multicast_register_if6);
753 			reg_mif_num = (mifi_t)-1;
754 			multicast_register_if6 = NULL;
755 		}
756 	}
757 
758 	bzero((caddr_t)mifp, sizeof(*mifp));
759 
760 	/* Adjust nummifs down */
761 	for (mifi = nummifs; mifi > 0; mifi--)
762 		if (mif6table[mifi - 1].m6_ifp)
763 			break;
764 	nummifs = mifi;
765 	MRT6_DLOG(DEBUG_ANY, "mif %d, nummifs %d", *mifip, nummifs);
766 
767 	return (0);
768 }
769 
770 static int
771 del_m6if(mifi_t *mifip)
772 {
773 	int cc;
774 
775 	MIF6_LOCK();
776 	cc = del_m6if_locked(mifip);
777 	MIF6_UNLOCK();
778 
779 	return (cc);
780 }
781 
782 /*
783  * Add an mfc entry
784  */
785 static int
786 add_m6fc(struct mf6cctl *mfccp)
787 {
788 	struct mf6c *rt;
789 	u_long hash;
790 	struct rtdetq *rte;
791 	u_short nstl;
792 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
793 
794 	MFC6_LOCK();
795 
796 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
797 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
798 
799 	/* If an entry already exists, just update the fields */
800 	if (rt) {
801 		MRT6_DLOG(DEBUG_MFC, "no upcall o %s g %s p %x",
802 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
803 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
804 		    mfccp->mf6cc_parent);
805 
806 		rt->mf6c_parent = mfccp->mf6cc_parent;
807 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
808 
809 		MFC6_UNLOCK();
810 		return (0);
811 	}
812 
813 	/*
814 	 * Find the entry for which the upcall was made and update
815 	 */
816 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
817 			mfccp->mf6cc_mcastgrp.sin6_addr);
818 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
819 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
820 				       &mfccp->mf6cc_origin.sin6_addr) &&
821 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
822 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
823 		    (rt->mf6c_stall != NULL)) {
824 
825 			if (nstl++)
826 				log(LOG_ERR,
827 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
828 				    "multiple kernel entries",
829 				    ip6_sprintf(ip6bufo,
830 					    &mfccp->mf6cc_origin.sin6_addr),
831 				    ip6_sprintf(ip6bufg,
832 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
833 				    mfccp->mf6cc_parent, rt->mf6c_stall);
834 
835 			MRT6_DLOG(DEBUG_MFC, "o %s g %s p %x dbg %p",
836 			    ip6_sprintf(ip6bufo,
837 			    &mfccp->mf6cc_origin.sin6_addr),
838 			    ip6_sprintf(ip6bufg,
839 				&mfccp->mf6cc_mcastgrp.sin6_addr),
840 			    mfccp->mf6cc_parent, rt->mf6c_stall);
841 
842 			rt->mf6c_origin     = mfccp->mf6cc_origin;
843 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
844 			rt->mf6c_parent     = mfccp->mf6cc_parent;
845 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
846 			/* initialize pkt counters per src-grp */
847 			rt->mf6c_pkt_cnt    = 0;
848 			rt->mf6c_byte_cnt   = 0;
849 			rt->mf6c_wrong_if   = 0;
850 
851 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
852 			n6expire[hash]--;
853 
854 			/* free packets Qed at the end of this entry */
855 			for (rte = rt->mf6c_stall; rte != NULL; ) {
856 				struct rtdetq *n = rte->next;
857 				ip6_mdq(rte->m, rte->ifp, rt);
858 				m_freem(rte->m);
859 #ifdef UPCALL_TIMING
860 				collate(&(rte->t));
861 #endif /* UPCALL_TIMING */
862 				free(rte, M_MRTABLE6);
863 				rte = n;
864 			}
865 			rt->mf6c_stall = NULL;
866 		}
867 	}
868 
869 	/*
870 	 * It is possible that an entry is being inserted without an upcall
871 	 */
872 	if (nstl == 0) {
873 		MRT6_DLOG(DEBUG_MFC, "no upcall h %lu o %s g %s p %x", hash,
874 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
875 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
876 		    mfccp->mf6cc_parent);
877 
878 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
879 
880 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
881 					       &mfccp->mf6cc_origin.sin6_addr)&&
882 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
883 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
884 
885 				rt->mf6c_origin     = mfccp->mf6cc_origin;
886 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
887 				rt->mf6c_parent     = mfccp->mf6cc_parent;
888 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
889 				/* initialize pkt counters per src-grp */
890 				rt->mf6c_pkt_cnt    = 0;
891 				rt->mf6c_byte_cnt   = 0;
892 				rt->mf6c_wrong_if   = 0;
893 
894 				if (rt->mf6c_expire)
895 					n6expire[hash]--;
896 				rt->mf6c_expire	   = 0;
897 			}
898 		}
899 		if (rt == NULL) {
900 			/* no upcall, so make a new entry */
901 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
902 						  M_NOWAIT);
903 			if (rt == NULL) {
904 				MFC6_UNLOCK();
905 				return (ENOBUFS);
906 			}
907 
908 			/* insert new entry at head of hash chain */
909 			rt->mf6c_origin     = mfccp->mf6cc_origin;
910 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
911 			rt->mf6c_parent     = mfccp->mf6cc_parent;
912 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
913 			/* initialize pkt counters per src-grp */
914 			rt->mf6c_pkt_cnt    = 0;
915 			rt->mf6c_byte_cnt   = 0;
916 			rt->mf6c_wrong_if   = 0;
917 			rt->mf6c_expire     = 0;
918 			rt->mf6c_stall = NULL;
919 
920 			/* link into table */
921 			rt->mf6c_next  = mf6ctable[hash];
922 			mf6ctable[hash] = rt;
923 		}
924 	}
925 
926 	MFC6_UNLOCK();
927 	return (0);
928 }
929 
930 #ifdef UPCALL_TIMING
931 /*
932  * collect delay statistics on the upcalls
933  */
934 static void
935 collate(struct timeval *t)
936 {
937 	u_long d;
938 	struct timeval tp;
939 	u_long delta;
940 
941 	GET_TIME(tp);
942 
943 	if (TV_LT(*t, tp))
944 	{
945 		TV_DELTA(tp, *t, delta);
946 
947 		d = delta >> 10;
948 		if (d > UPCALL_MAX)
949 			d = UPCALL_MAX;
950 
951 		++upcall_data[d];
952 	}
953 }
954 #endif /* UPCALL_TIMING */
955 
956 /*
957  * Delete an mfc entry
958  */
959 static int
960 del_m6fc(struct mf6cctl *mfccp)
961 {
962 #ifdef MRT6DEBUG
963 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
964 #endif
965 	struct sockaddr_in6	origin;
966 	struct sockaddr_in6	mcastgrp;
967 	struct mf6c		*rt;
968 	struct mf6c		**nptr;
969 	u_long		hash;
970 
971 	origin = mfccp->mf6cc_origin;
972 	mcastgrp = mfccp->mf6cc_mcastgrp;
973 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
974 
975 	MRT6_DLOG(DEBUG_MFC, "orig %s mcastgrp %s",
976 	    ip6_sprintf(ip6bufo, &origin.sin6_addr),
977 	    ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr));
978 
979 	MFC6_LOCK();
980 
981 	nptr = &mf6ctable[hash];
982 	while ((rt = *nptr) != NULL) {
983 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
984 				       &rt->mf6c_origin.sin6_addr) &&
985 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
986 				       &rt->mf6c_mcastgrp.sin6_addr) &&
987 		    rt->mf6c_stall == NULL)
988 			break;
989 
990 		nptr = &rt->mf6c_next;
991 	}
992 	if (rt == NULL) {
993 		MFC6_UNLOCK();
994 		return (EADDRNOTAVAIL);
995 	}
996 
997 	*nptr = rt->mf6c_next;
998 	free(rt, M_MRTABLE6);
999 
1000 	MFC6_UNLOCK();
1001 
1002 	return (0);
1003 }
1004 
1005 static int
1006 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
1007 {
1008 
1009 	if (s) {
1010 		if (sbappendaddr(&s->so_rcv,
1011 				 (struct sockaddr *)src,
1012 				 mm, (struct mbuf *)0) != 0) {
1013 			sorwakeup(s);
1014 			return (0);
1015 		}
1016 	}
1017 	m_freem(mm);
1018 	return (-1);
1019 }
1020 
1021 /*
1022  * IPv6 multicast forwarding function. This function assumes that the packet
1023  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
1024  * pointed to by "ifp", and the packet is to be relayed to other networks
1025  * that have members of the packet's destination IPv6 multicast group.
1026  *
1027  * The packet is returned unscathed to the caller, unless it is
1028  * erroneous, in which case a non-zero return value tells the caller to
1029  * discard it.
1030  *
1031  * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff
1032  * this function is called in the originating context (i.e., not when
1033  * forwarding a packet from other node).  ip6_output(), which is currently the
1034  * only function that calls this function is called in the originating context,
1035  * explicitly ensures this condition.  It is caller's responsibility to ensure
1036  * that if this function is called from somewhere else in the originating
1037  * context in the future.
1038  */
1039 int
1040 X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
1041 {
1042 	struct rtdetq *rte;
1043 	struct mbuf *mb0;
1044 	struct mf6c *rt;
1045 	struct mif6 *mifp;
1046 	struct mbuf *mm;
1047 	u_long hash;
1048 	mifi_t mifi;
1049 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1050 #ifdef UPCALL_TIMING
1051 	struct timeval tp;
1052 
1053 	GET_TIME(tp);
1054 #endif /* UPCALL_TIMING */
1055 
1056 	MRT6_DLOG(DEBUG_FORWARD, "src %s, dst %s, ifindex %d",
1057 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1058 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst), ifp->if_index);
1059 
1060 	/*
1061 	 * Don't forward a packet with Hop limit of zero or one,
1062 	 * or a packet destined to a local-only group.
1063 	 */
1064 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
1065 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1066 		return (0);
1067 	ip6->ip6_hlim--;
1068 
1069 	/*
1070 	 * Source address check: do not forward packets with unspecified
1071 	 * source. It was discussed in July 2000, on ipngwg mailing list.
1072 	 * This is rather more serious than unicast cases, because some
1073 	 * MLD packets can be sent with the unspecified source address
1074 	 * (although such packets must normally set 1 to the hop limit field).
1075 	 */
1076 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1077 		IP6STAT_INC(ip6s_cantforward);
1078 		if (V_ip6_log_time + V_ip6_log_interval < time_uptime) {
1079 			V_ip6_log_time = time_uptime;
1080 			log(LOG_DEBUG,
1081 			    "cannot forward "
1082 			    "from %s to %s nxt %d received on %s\n",
1083 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1084 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1085 			    ip6->ip6_nxt,
1086 			    if_name(m->m_pkthdr.rcvif));
1087 		}
1088 		return (0);
1089 	}
1090 
1091 	MFC6_LOCK();
1092 
1093 	/*
1094 	 * Determine forwarding mifs from the forwarding cache table
1095 	 */
1096 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
1097 	MRT6STAT_INC(mrt6s_mfc_lookups);
1098 
1099 	/* Entry exists, so forward if necessary */
1100 	if (rt) {
1101 		MFC6_UNLOCK();
1102 		return (ip6_mdq(m, ifp, rt));
1103 	}
1104 
1105 	/*
1106 	 * If we don't have a route for packet's origin,
1107 	 * Make a copy of the packet & send message to routing daemon.
1108 	 */
1109 	MRT6STAT_INC(mrt6s_no_route);
1110 	MRT6_DLOG(DEBUG_FORWARD | DEBUG_MFC, "no rte s %s g %s",
1111 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1112 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1113 
1114 	/*
1115 	 * Allocate mbufs early so that we don't do extra work if we
1116 	 * are just going to fail anyway.
1117 	 */
1118 	rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE6, M_NOWAIT);
1119 	if (rte == NULL) {
1120 		MFC6_UNLOCK();
1121 		return (ENOBUFS);
1122 	}
1123 	mb0 = m_copy(m, 0, M_COPYALL);
1124 	/*
1125 	 * Pullup packet header if needed before storing it,
1126 	 * as other references may modify it in the meantime.
1127 	 */
1128 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1129 		mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1130 	if (mb0 == NULL) {
1131 		free(rte, M_MRTABLE6);
1132 		MFC6_UNLOCK();
1133 		return (ENOBUFS);
1134 	}
1135 
1136 	/* is there an upcall waiting for this packet? */
1137 	hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1138 	for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1139 		if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1140 		    &rt->mf6c_origin.sin6_addr) &&
1141 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1142 		    &rt->mf6c_mcastgrp.sin6_addr) && (rt->mf6c_stall != NULL))
1143 			break;
1144 	}
1145 
1146 	if (rt == NULL) {
1147 		struct mrt6msg *im;
1148 #ifdef MRT6_OINIT
1149 		struct omrt6msg *oim;
1150 #endif
1151 		/* no upcall, so make a new entry */
1152 		rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6, M_NOWAIT);
1153 		if (rt == NULL) {
1154 			free(rte, M_MRTABLE6);
1155 			m_freem(mb0);
1156 			MFC6_UNLOCK();
1157 			return (ENOBUFS);
1158 		}
1159 		/*
1160 		 * Make a copy of the header to send to the user
1161 		 * level process
1162 		 */
1163 		mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1164 		if (mm == NULL) {
1165 			free(rte, M_MRTABLE6);
1166 			m_freem(mb0);
1167 			free(rt, M_MRTABLE6);
1168 			MFC6_UNLOCK();
1169 			return (ENOBUFS);
1170 		}
1171 
1172 		/*
1173 		 * Send message to routing daemon
1174 		 */
1175 		sin6.sin6_addr = ip6->ip6_src;
1176 		im = NULL;
1177 #ifdef MRT6_OINIT
1178 		oim = NULL;
1179 #endif
1180 		switch (V_ip6_mrouter_ver) {
1181 #ifdef MRT6_OINIT
1182 		case MRT6_OINIT:
1183 			oim = mtod(mm, struct omrt6msg *);
1184 			oim->im6_msgtype = MRT6MSG_NOCACHE;
1185 			oim->im6_mbz = 0;
1186 			break;
1187 #endif
1188 		case MRT6_INIT:
1189 			im = mtod(mm, struct mrt6msg *);
1190 			im->im6_msgtype = MRT6MSG_NOCACHE;
1191 			im->im6_mbz = 0;
1192 			break;
1193 		default:
1194 			free(rte, M_MRTABLE6);
1195 			m_freem(mb0);
1196 			free(rt, M_MRTABLE6);
1197 			MFC6_UNLOCK();
1198 			return (EINVAL);
1199 		}
1200 
1201 		MRT6_DLOG(DEBUG_FORWARD, "getting the iif info in the kernel");
1202 		for (mifp = mif6table, mifi = 0;
1203 		    mifi < nummifs && mifp->m6_ifp != ifp; mifp++, mifi++)
1204 				;
1205 
1206 		switch (V_ip6_mrouter_ver) {
1207 #ifdef MRT6_OINIT
1208 		case MRT6_OINIT:
1209 			oim->im6_mif = mifi;
1210 			break;
1211 #endif
1212 		case MRT6_INIT:
1213 			im->im6_mif = mifi;
1214 			break;
1215 		}
1216 
1217 		if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1218 			log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1219 			    "socket queue full\n");
1220 			MRT6STAT_INC(mrt6s_upq_sockfull);
1221 			free(rte, M_MRTABLE6);
1222 			m_freem(mb0);
1223 			free(rt, M_MRTABLE6);
1224 			MFC6_UNLOCK();
1225 			return (ENOBUFS);
1226 		}
1227 
1228 		MRT6STAT_INC(mrt6s_upcalls);
1229 
1230 		/* insert new entry at head of hash chain */
1231 		bzero(rt, sizeof(*rt));
1232 		rt->mf6c_origin.sin6_family = AF_INET6;
1233 		rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1234 		rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1235 		rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1236 		rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1237 		rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1238 		rt->mf6c_expire = UPCALL_EXPIRE;
1239 		n6expire[hash]++;
1240 		rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1241 
1242 		/* link into table */
1243 		rt->mf6c_next  = mf6ctable[hash];
1244 		mf6ctable[hash] = rt;
1245 		/* Add this entry to the end of the queue */
1246 		rt->mf6c_stall = rte;
1247 	} else {
1248 		/* determine if q has overflowed */
1249 		struct rtdetq **p;
1250 		int npkts = 0;
1251 
1252 		for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1253 			if (++npkts > MAX_UPQ6) {
1254 				MRT6STAT_INC(mrt6s_upq_ovflw);
1255 				free(rte, M_MRTABLE6);
1256 				m_freem(mb0);
1257 				MFC6_UNLOCK();
1258 				return (0);
1259 			}
1260 
1261 		/* Add this entry to the end of the queue */
1262 		*p = rte;
1263 	}
1264 
1265 	rte->next = NULL;
1266 	rte->m = mb0;
1267 	rte->ifp = ifp;
1268 #ifdef UPCALL_TIMING
1269 	rte->t = tp;
1270 #endif /* UPCALL_TIMING */
1271 
1272 	MFC6_UNLOCK();
1273 
1274 	return (0);
1275 }
1276 
1277 /*
1278  * Clean up cache entries if upcalls are not serviced
1279  * Call from the Slow Timeout mechanism, every half second.
1280  */
1281 static void
1282 expire_upcalls(void *unused)
1283 {
1284 #ifdef MRT6DEBUG
1285 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1286 #endif
1287 	struct rtdetq *rte;
1288 	struct mf6c *mfc, **nptr;
1289 	u_long i;
1290 
1291 	MFC6_LOCK();
1292 	for (i = 0; i < MF6CTBLSIZ; i++) {
1293 		if (n6expire[i] == 0)
1294 			continue;
1295 		nptr = &mf6ctable[i];
1296 		while ((mfc = *nptr) != NULL) {
1297 			rte = mfc->mf6c_stall;
1298 			/*
1299 			 * Skip real cache entries
1300 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1301 			 * If it expires now
1302 			 */
1303 			if (rte != NULL &&
1304 			    mfc->mf6c_expire != 0 &&
1305 			    --mfc->mf6c_expire == 0) {
1306 				MRT6_DLOG(DEBUG_EXPIRE, "expiring (%s %s)",
1307 				    ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr),
1308 				    ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr));
1309 				/*
1310 				 * drop all the packets
1311 				 * free the mbuf with the pkt, if, timing info
1312 				 */
1313 				do {
1314 					struct rtdetq *n = rte->next;
1315 					m_freem(rte->m);
1316 					free(rte, M_MRTABLE6);
1317 					rte = n;
1318 				} while (rte != NULL);
1319 				MRT6STAT_INC(mrt6s_cache_cleanups);
1320 				n6expire[i]--;
1321 
1322 				*nptr = mfc->mf6c_next;
1323 				free(mfc, M_MRTABLE6);
1324 			} else {
1325 				nptr = &mfc->mf6c_next;
1326 			}
1327 		}
1328 	}
1329 	MFC6_UNLOCK();
1330 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1331 	    expire_upcalls, NULL);
1332 }
1333 
1334 /*
1335  * Packet forwarding routine once entry in the cache is made
1336  */
1337 static int
1338 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1339 {
1340 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1341 	mifi_t mifi, iif;
1342 	struct mif6 *mifp;
1343 	int plen = m->m_pkthdr.len;
1344 	struct in6_addr src0, dst0; /* copies for local work */
1345 	u_int32_t iszone, idzone, oszone, odzone;
1346 	int error = 0;
1347 
1348 /*
1349  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1350  * input, they shouldn't get counted on output, so statistics keeping is
1351  * separate.
1352  */
1353 
1354 #define MC6_SEND(ip6, mifp, m) do {				\
1355 	if ((mifp)->m6_flags & MIFF_REGISTER)			\
1356 		register_send((ip6), (mifp), (m));		\
1357 	else							\
1358 		phyint_send((ip6), (mifp), (m));		\
1359 } while (/*CONSTCOND*/ 0)
1360 
1361 	/*
1362 	 * Don't forward if it didn't arrive from the parent mif
1363 	 * for its origin.
1364 	 */
1365 	mifi = rt->mf6c_parent;
1366 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1367 		/* came in the wrong interface */
1368 		MRT6_DLOG(DEBUG_FORWARD,
1369 		    "wrong if: ifid %d mifi %d mififid %x", ifp->if_index,
1370 		    mifi, mif6table[mifi].m6_ifp->if_index);
1371 		MRT6STAT_INC(mrt6s_wrong_if);
1372 		rt->mf6c_wrong_if++;
1373 		/*
1374 		 * If we are doing PIM processing, and we are forwarding
1375 		 * packets on this interface, send a message to the
1376 		 * routing daemon.
1377 		 */
1378 		/* have to make sure this is a valid mif */
1379 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1380 			if (V_pim6 && (m->m_flags & M_LOOP) == 0) {
1381 				/*
1382 				 * Check the M_LOOP flag to avoid an
1383 				 * unnecessary PIM assert.
1384 				 * XXX: M_LOOP is an ad-hoc hack...
1385 				 */
1386 				static struct sockaddr_in6 sin6 =
1387 				{ sizeof(sin6), AF_INET6 };
1388 
1389 				struct mbuf *mm;
1390 				struct mrt6msg *im;
1391 #ifdef MRT6_OINIT
1392 				struct omrt6msg *oim;
1393 #endif
1394 
1395 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1396 				if (mm &&
1397 				    (!M_WRITABLE(mm) ||
1398 				     mm->m_len < sizeof(struct ip6_hdr)))
1399 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1400 				if (mm == NULL)
1401 					return (ENOBUFS);
1402 
1403 #ifdef MRT6_OINIT
1404 				oim = NULL;
1405 #endif
1406 				im = NULL;
1407 				switch (V_ip6_mrouter_ver) {
1408 #ifdef MRT6_OINIT
1409 				case MRT6_OINIT:
1410 					oim = mtod(mm, struct omrt6msg *);
1411 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1412 					oim->im6_mbz = 0;
1413 					break;
1414 #endif
1415 				case MRT6_INIT:
1416 					im = mtod(mm, struct mrt6msg *);
1417 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1418 					im->im6_mbz = 0;
1419 					break;
1420 				default:
1421 					m_freem(mm);
1422 					return (EINVAL);
1423 				}
1424 
1425 				for (mifp = mif6table, iif = 0;
1426 				     iif < nummifs && mifp &&
1427 					     mifp->m6_ifp != ifp;
1428 				     mifp++, iif++)
1429 					;
1430 
1431 				switch (V_ip6_mrouter_ver) {
1432 #ifdef MRT6_OINIT
1433 				case MRT6_OINIT:
1434 					oim->im6_mif = iif;
1435 					sin6.sin6_addr = oim->im6_src;
1436 					break;
1437 #endif
1438 				case MRT6_INIT:
1439 					im->im6_mif = iif;
1440 					sin6.sin6_addr = im->im6_src;
1441 					break;
1442 				}
1443 
1444 				MRT6STAT_INC(mrt6s_upcalls);
1445 
1446 				if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1447 					MRT6_DLOG(DEBUG_ANY,
1448 					    "ip6_mrouter socket queue full");
1449 					MRT6STAT_INC(mrt6s_upq_sockfull);
1450 					return (ENOBUFS);
1451 				}	/* if socket Q full */
1452 			}		/* if PIM */
1453 		return (0);
1454 	}			/* if wrong iif */
1455 
1456 	/* If I sourced this packet, it counts as output, else it was input. */
1457 	if (m->m_pkthdr.rcvif == NULL) {
1458 		/* XXX: is rcvif really NULL when output?? */
1459 		mif6table[mifi].m6_pkt_out++;
1460 		mif6table[mifi].m6_bytes_out += plen;
1461 	} else {
1462 		mif6table[mifi].m6_pkt_in++;
1463 		mif6table[mifi].m6_bytes_in += plen;
1464 	}
1465 	rt->mf6c_pkt_cnt++;
1466 	rt->mf6c_byte_cnt += plen;
1467 
1468 	/*
1469 	 * For each mif, forward a copy of the packet if there are group
1470 	 * members downstream on the interface.
1471 	 */
1472 	src0 = ip6->ip6_src;
1473 	dst0 = ip6->ip6_dst;
1474 	if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 ||
1475 	    (error = in6_setscope(&dst0, ifp, &idzone)) != 0) {
1476 		IP6STAT_INC(ip6s_badscope);
1477 		return (error);
1478 	}
1479 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) {
1480 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1481 			/*
1482 			 * check if the outgoing packet is going to break
1483 			 * a scope boundary.
1484 			 * XXX For packets through PIM register tunnel
1485 			 * interface, we believe a routing daemon.
1486 			 */
1487 			if (!(mif6table[rt->mf6c_parent].m6_flags &
1488 			      MIFF_REGISTER) &&
1489 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
1490 				if (in6_setscope(&src0, mif6table[mifi].m6_ifp,
1491 				    &oszone) ||
1492 				    in6_setscope(&dst0, mif6table[mifi].m6_ifp,
1493 				    &odzone) ||
1494 				    iszone != oszone ||
1495 				    idzone != odzone) {
1496 					IP6STAT_INC(ip6s_badscope);
1497 					continue;
1498 				}
1499 			}
1500 
1501 			mifp->m6_pkt_out++;
1502 			mifp->m6_bytes_out += plen;
1503 			MC6_SEND(ip6, mifp, m);
1504 		}
1505 	}
1506 	return (0);
1507 }
1508 
1509 static void
1510 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1511 {
1512 #ifdef MRT6DEBUG
1513 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1514 #endif
1515 	struct mbuf *mb_copy;
1516 	struct ifnet *ifp = mifp->m6_ifp;
1517 	int error = 0;
1518 	u_long linkmtu;
1519 
1520 	/*
1521 	 * Make a new reference to the packet; make sure that
1522 	 * the IPv6 header is actually copied, not just referenced,
1523 	 * so that ip6_output() only scribbles on the copy.
1524 	 */
1525 	mb_copy = m_copy(m, 0, M_COPYALL);
1526 	if (mb_copy &&
1527 	    (!M_WRITABLE(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1528 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1529 	if (mb_copy == NULL) {
1530 		return;
1531 	}
1532 	/* set MCAST flag to the outgoing packet */
1533 	mb_copy->m_flags |= M_MCAST;
1534 
1535 	/*
1536 	 * If we sourced the packet, call ip6_output since we may devide
1537 	 * the packet into fragments when the packet is too big for the
1538 	 * outgoing interface.
1539 	 * Otherwise, we can simply send the packet to the interface
1540 	 * sending queue.
1541 	 */
1542 	if (m->m_pkthdr.rcvif == NULL) {
1543 		struct ip6_moptions im6o;
1544 
1545 		im6o.im6o_multicast_ifp = ifp;
1546 		/* XXX: ip6_output will override ip6->ip6_hlim */
1547 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1548 		im6o.im6o_multicast_loop = 1;
1549 		error = ip6_output(mb_copy, NULL, NULL, IPV6_FORWARDING, &im6o,
1550 		    NULL, NULL);
1551 
1552 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1553 		    (uint16_t)(mifp - mif6table), error);
1554 		return;
1555 	}
1556 
1557 	/*
1558 	 * If configured to loop back multicasts by default,
1559 	 * loop back a copy now.
1560 	 */
1561 	if (in6_mcast_loop) {
1562 		struct sockaddr_in6 dst6;
1563 
1564 		bzero(&dst6, sizeof(dst6));
1565 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1566 		dst6.sin6_family = AF_INET6;
1567 		dst6.sin6_addr = ip6->ip6_dst;
1568 		ip6_mloopback(ifp, m, &dst6);
1569 	}
1570 
1571 	/*
1572 	 * Put the packet into the sending queue of the outgoing interface
1573 	 * if it would fit in the MTU of the interface.
1574 	 */
1575 	linkmtu = IN6_LINKMTU(ifp);
1576 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1577 		struct sockaddr_in6 dst6;
1578 
1579 		bzero(&dst6, sizeof(dst6));
1580 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1581 		dst6.sin6_family = AF_INET6;
1582 		dst6.sin6_addr = ip6->ip6_dst;
1583 
1584 		IP_PROBE(send, NULL, NULL, ip6, ifp, NULL, ip6);
1585 		/*
1586 		 * We just call if_output instead of nd6_output here, since
1587 		 * we need no ND for a multicast forwarded packet...right?
1588 		 */
1589 		m_clrprotoflags(m);	/* Avoid confusing lower layers. */
1590 		error = (*ifp->if_output)(ifp, mb_copy,
1591 		    (struct sockaddr *)&dst6, NULL);
1592 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1593 		    (uint16_t)(mifp - mif6table), error);
1594 	} else {
1595 		/*
1596 		 * pMTU discovery is intentionally disabled by default, since
1597 		 * various router may notify pMTU in multicast, which can be
1598 		 * a DDoS to a router
1599 		 */
1600 		if (V_ip6_mcast_pmtu)
1601 			icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1602 		else {
1603 			MRT6_DLOG(DEBUG_XMIT, " packet too big on %s o %s "
1604 			    "g %s size %d (discarded)", if_name(ifp),
1605 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1606 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1607 			    mb_copy->m_pkthdr.len);
1608 			m_freem(mb_copy); /* simply discard the packet */
1609 		}
1610 	}
1611 }
1612 
1613 static int
1614 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1615 {
1616 #ifdef MRT6DEBUG
1617 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1618 #endif
1619 	struct mbuf *mm;
1620 	int i, len = m->m_pkthdr.len;
1621 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1622 	struct mrt6msg *im6;
1623 
1624 	MRT6_DLOG(DEBUG_ANY, "src %s dst %s",
1625 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1626 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1627 	PIM6STAT_INC(pim6s_snd_registers);
1628 
1629 	/* Make a copy of the packet to send to the user level process. */
1630 	mm = m_gethdr(M_NOWAIT, MT_DATA);
1631 	if (mm == NULL)
1632 		return (ENOBUFS);
1633 	mm->m_data += max_linkhdr;
1634 	mm->m_len = sizeof(struct ip6_hdr);
1635 
1636 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1637 		m_freem(mm);
1638 		return (ENOBUFS);
1639 	}
1640 	i = MHLEN - M_LEADINGSPACE(mm);
1641 	if (i > len)
1642 		i = len;
1643 	mm = m_pullup(mm, i);
1644 	if (mm == NULL)
1645 		return (ENOBUFS);
1646 /* TODO: check it! */
1647 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1648 
1649 	/*
1650 	 * Send message to routing daemon
1651 	 */
1652 	sin6.sin6_addr = ip6->ip6_src;
1653 
1654 	im6 = mtod(mm, struct mrt6msg *);
1655 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1656 	im6->im6_mbz          = 0;
1657 
1658 	im6->im6_mif = mif - mif6table;
1659 
1660 	/* iif info is not given for reg. encap.n */
1661 	MRT6STAT_INC(mrt6s_upcalls);
1662 
1663 	if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1664 		MRT6_DLOG(DEBUG_ANY, "ip6_mrouter socket queue full");
1665 		MRT6STAT_INC(mrt6s_upq_sockfull);
1666 		return (ENOBUFS);
1667 	}
1668 	return (0);
1669 }
1670 
1671 /*
1672  * pim6_encapcheck() is called by the encap6_input() path at runtime to
1673  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
1674  * into the kernel.
1675  */
1676 static int
1677 pim6_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
1678 {
1679 
1680 #ifdef DIAGNOSTIC
1681     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
1682 #endif
1683     if (proto != IPPROTO_PIM)
1684 	return 0;	/* not for us; reject the datagram. */
1685 
1686     return 64;		/* claim the datagram. */
1687 }
1688 
1689 /*
1690  * PIM sparse mode hook
1691  * Receives the pim control messages, and passes them up to the listening
1692  * socket, using rip6_input.
1693  * The only message processed is the REGISTER pim message; the pim header
1694  * is stripped off, and the inner packet is passed to register_mforward.
1695  */
1696 int
1697 pim6_input(struct mbuf **mp, int *offp, int proto)
1698 {
1699 	struct pim *pim; /* pointer to a pim struct */
1700 	struct ip6_hdr *ip6;
1701 	int pimlen;
1702 	struct mbuf *m = *mp;
1703 	int minlen;
1704 	int off = *offp;
1705 
1706 	PIM6STAT_INC(pim6s_rcv_total);
1707 
1708 	ip6 = mtod(m, struct ip6_hdr *);
1709 	pimlen = m->m_pkthdr.len - *offp;
1710 
1711 	/*
1712 	 * Validate lengths
1713 	 */
1714 	if (pimlen < PIM_MINLEN) {
1715 		PIM6STAT_INC(pim6s_rcv_tooshort);
1716 		MRT6_DLOG(DEBUG_PIM, "PIM packet too short");
1717 		m_freem(m);
1718 		return (IPPROTO_DONE);
1719 	}
1720 
1721 	/*
1722 	 * if the packet is at least as big as a REGISTER, go ahead
1723 	 * and grab the PIM REGISTER header size, to avoid another
1724 	 * possible m_pullup() later.
1725 	 *
1726 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1727 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1728 	 */
1729 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1730 
1731 	/*
1732 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1733 	 * possibly the PIM REGISTER header
1734 	 */
1735 #ifndef PULLDOWN_TEST
1736 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1737 	/* adjust pointer */
1738 	ip6 = mtod(m, struct ip6_hdr *);
1739 
1740 	/* adjust mbuf to point to the PIM header */
1741 	pim = (struct pim *)((caddr_t)ip6 + off);
1742 #else
1743 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1744 	if (pim == NULL) {
1745 		PIM6STAT_INC(pim6s_rcv_tooshort);
1746 		return (IPPROTO_DONE);
1747 	}
1748 #endif
1749 
1750 #define PIM6_CHECKSUM
1751 #ifdef PIM6_CHECKSUM
1752 	{
1753 		int cksumlen;
1754 
1755 		/*
1756 		 * Validate checksum.
1757 		 * If PIM REGISTER, exclude the data packet
1758 		 */
1759 		if (pim->pim_type == PIM_REGISTER)
1760 			cksumlen = PIM_MINLEN;
1761 		else
1762 			cksumlen = pimlen;
1763 
1764 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1765 			PIM6STAT_INC(pim6s_rcv_badsum);
1766 			MRT6_DLOG(DEBUG_PIM, "invalid checksum");
1767 			m_freem(m);
1768 			return (IPPROTO_DONE);
1769 		}
1770 	}
1771 #endif /* PIM_CHECKSUM */
1772 
1773 	/* PIM version check */
1774 	if (pim->pim_ver != PIM_VERSION) {
1775 		PIM6STAT_INC(pim6s_rcv_badversion);
1776 		MRT6_DLOG(DEBUG_ANY | DEBUG_ERR,
1777 		    "incorrect version %d, expecting %d",
1778 		    pim->pim_ver, PIM_VERSION);
1779 		m_freem(m);
1780 		return (IPPROTO_DONE);
1781 	}
1782 
1783 	if (pim->pim_type == PIM_REGISTER) {
1784 		/*
1785 		 * since this is a REGISTER, we'll make a copy of the register
1786 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1787 		 * routing daemon.
1788 		 */
1789 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1790 
1791 		struct mbuf *mcp;
1792 		struct ip6_hdr *eip6;
1793 		u_int32_t *reghdr;
1794 		int rc;
1795 #ifdef MRT6DEBUG
1796 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1797 #endif
1798 
1799 		PIM6STAT_INC(pim6s_rcv_registers);
1800 
1801 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1802 			MRT6_DLOG(DEBUG_PIM, "register mif not set: %d",
1803 			    reg_mif_num);
1804 			m_freem(m);
1805 			return (IPPROTO_DONE);
1806 		}
1807 
1808 		reghdr = (u_int32_t *)(pim + 1);
1809 
1810 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1811 			goto pim6_input_to_daemon;
1812 
1813 		/*
1814 		 * Validate length
1815 		 */
1816 		if (pimlen < PIM6_REG_MINLEN) {
1817 			PIM6STAT_INC(pim6s_rcv_tooshort);
1818 			PIM6STAT_INC(pim6s_rcv_badregisters);
1819 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "register packet "
1820 			    "size too small %d from %s",
1821 			    pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src));
1822 			m_freem(m);
1823 			return (IPPROTO_DONE);
1824 		}
1825 
1826 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1827 		MRT6_DLOG(DEBUG_PIM, "eip6: %s -> %s, eip6 plen %d",
1828 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1829 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1830 		    ntohs(eip6->ip6_plen));
1831 
1832 		/* verify the version number of the inner packet */
1833 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1834 			PIM6STAT_INC(pim6s_rcv_badregisters);
1835 			MRT6_DLOG(DEBUG_ANY, "invalid IP version (%d) "
1836 			    "of the inner packet",
1837 			    (eip6->ip6_vfc & IPV6_VERSION));
1838 			m_freem(m);
1839 			return (IPPROTO_NONE);
1840 		}
1841 
1842 		/* verify the inner packet is destined to a mcast group */
1843 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1844 			PIM6STAT_INC(pim6s_rcv_badregisters);
1845 			MRT6_DLOG(DEBUG_PIM, "inner packet of register "
1846 			    "is not multicast %s",
1847 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst));
1848 			m_freem(m);
1849 			return (IPPROTO_DONE);
1850 		}
1851 
1852 		/*
1853 		 * make a copy of the whole header to pass to the daemon later.
1854 		 */
1855 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1856 		if (mcp == NULL) {
1857 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "pim register: "
1858 			    "could not copy register head");
1859 			m_freem(m);
1860 			return (IPPROTO_DONE);
1861 		}
1862 
1863 		/*
1864 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1865 		 */
1866 		m_adj(m, off + PIM_MINLEN);
1867 		MRT6_DLOG(DEBUG_PIM, "forwarding decapsulated register: "
1868 		    "src %s, dst %s, mif %d",
1869 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1870 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst), reg_mif_num);
1871 
1872 		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1873 				dst.sin6_family, 0);
1874 
1875 		/* prepare the register head to send to the mrouting daemon */
1876 		m = mcp;
1877 	}
1878 
1879 	/*
1880 	 * Pass the PIM message up to the daemon; if it is a register message
1881 	 * pass the 'head' only up to the daemon. This includes the
1882 	 * encapsulator ip6 header, pim header, register header and the
1883 	 * encapsulated ip6 header.
1884 	 */
1885   pim6_input_to_daemon:
1886 	rip6_input(&m, offp, proto);
1887 	return (IPPROTO_DONE);
1888 }
1889 
1890 static int
1891 ip6_mroute_modevent(module_t mod, int type, void *unused)
1892 {
1893 
1894 	switch (type) {
1895 	case MOD_LOAD:
1896 		MROUTER6_LOCK_INIT();
1897 		MFC6_LOCK_INIT();
1898 		MIF6_LOCK_INIT();
1899 
1900 		pim6_encap_cookie = encap_attach_func(AF_INET6, IPPROTO_PIM,
1901 			pim6_encapcheck,
1902 			(const struct protosw *)&in6_pim_protosw, NULL);
1903 		if (pim6_encap_cookie == NULL) {
1904 			printf("ip6_mroute: unable to attach pim6 encap\n");
1905 			MIF6_LOCK_DESTROY();
1906 			MFC6_LOCK_DESTROY();
1907 			MROUTER6_LOCK_DESTROY();
1908 			return (EINVAL);
1909 		}
1910 
1911 		ip6_mforward = X_ip6_mforward;
1912 		ip6_mrouter_done = X_ip6_mrouter_done;
1913 		ip6_mrouter_get = X_ip6_mrouter_get;
1914 		ip6_mrouter_set = X_ip6_mrouter_set;
1915 		mrt6_ioctl = X_mrt6_ioctl;
1916 		break;
1917 
1918 	case MOD_UNLOAD:
1919 		if (V_ip6_mrouter != NULL)
1920 			return EINVAL;
1921 
1922 		if (pim6_encap_cookie) {
1923 			encap_detach(pim6_encap_cookie);
1924 			pim6_encap_cookie = NULL;
1925 		}
1926 		X_ip6_mrouter_done();
1927 		ip6_mforward = NULL;
1928 		ip6_mrouter_done = NULL;
1929 		ip6_mrouter_get = NULL;
1930 		ip6_mrouter_set = NULL;
1931 		mrt6_ioctl = NULL;
1932 
1933 		MIF6_LOCK_DESTROY();
1934 		MFC6_LOCK_DESTROY();
1935 		MROUTER6_LOCK_DESTROY();
1936 		break;
1937 
1938 	default:
1939 		return (EOPNOTSUPP);
1940 	}
1941 
1942 	return (0);
1943 }
1944 
1945 static moduledata_t ip6_mroutemod = {
1946 	"ip6_mroute",
1947 	ip6_mroute_modevent,
1948 	0
1949 };
1950 
1951 DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
1952