xref: /freebsd/sys/netinet6/ip6_mroute.c (revision f4f8f02054f3abb6ceb84aefcdecc78d5c8b462f)
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_inet.h"
85 #include "opt_inet6.h"
86 #include "opt_route.h"
87 
88 #include <sys/param.h>
89 #include <sys/callout.h>
90 #include <sys/errno.h>
91 #include <sys/kernel.h>
92 #include <sys/lock.h>
93 #include <sys/malloc.h>
94 #include <sys/mbuf.h>
95 #include <sys/module.h>
96 #include <sys/domain.h>
97 #include <sys/protosw.h>
98 #include <sys/signalvar.h>
99 #include <sys/socket.h>
100 #include <sys/socketvar.h>
101 #include <sys/sockio.h>
102 #include <sys/sx.h>
103 #include <sys/sysctl.h>
104 #include <sys/syslog.h>
105 #include <sys/systm.h>
106 #include <sys/time.h>
107 #include <sys/vimage.h>
108 
109 #include <net/if.h>
110 #include <net/if_types.h>
111 #include <net/raw_cb.h>
112 #include <net/route.h>
113 #include <net/vnet.h>
114 
115 #include <netinet/in.h>
116 #include <netinet/in_var.h>
117 #include <netinet/icmp6.h>
118 #include <netinet/vinet.h>
119 #include <netinet/ip_encap.h>
120 
121 #include <netinet/ip6.h>
122 #include <netinet6/ip6_var.h>
123 #include <netinet6/scope6_var.h>
124 #include <netinet6/nd6.h>
125 #include <netinet6/ip6_mroute.h>
126 #include <netinet6/ip6protosw.h>
127 #include <netinet6/pim6.h>
128 #include <netinet6/pim6_var.h>
129 #include <netinet6/vinet6.h>
130 
131 static MALLOC_DEFINE(M_MRTABLE6, "mf6c", "multicast forwarding cache entry");
132 
133 /* XXX: this is a very common idiom; move to <sys/mbuf.h> ? */
134 #define M_HASCL(m) ((m)->m_flags & M_EXT)
135 
136 static int	ip6_mdq(struct mbuf *, struct ifnet *, struct mf6c *);
137 static void	phyint_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
138 static void	pim6_init(void);
139 static int	register_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
140 static int	set_pim6(int *);
141 static int	socket_send(struct socket *, struct mbuf *,
142 		    struct sockaddr_in6 *);
143 
144 extern int in6_mcast_loop;
145 extern struct domain inet6domain;
146 
147 static const struct encaptab *pim6_encap_cookie;
148 static const struct ip6protosw in6_pim_protosw = {
149 	.pr_type =		SOCK_RAW,
150 	.pr_domain =		&inet6domain,
151 	.pr_protocol =		IPPROTO_PIM,
152 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
153 	.pr_input =		pim6_input,
154 	.pr_output =		rip6_output,
155 	.pr_ctloutput =		rip6_ctloutput,
156 	.pr_init =		pim6_init,
157 	.pr_usrreqs =		&rip6_usrreqs
158 };
159 static int pim6_encapcheck(const struct mbuf *, int, int, void *);
160 
161 #ifdef VIMAGE_GLOBALS
162 static int ip6_mrouter_ver;
163 #endif
164 
165 SYSCTL_DECL(_net_inet6);
166 SYSCTL_DECL(_net_inet6_ip6);
167 SYSCTL_NODE(_net_inet6, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
168 
169 static struct mrt6stat mrt6stat;
170 SYSCTL_STRUCT(_net_inet6_ip6, OID_AUTO, mrt6stat, CTLFLAG_RW,
171     &mrt6stat, mrt6stat,
172     "Multicast Routing Statistics (struct mrt6stat, netinet6/ip6_mroute.h)");
173 
174 #define NO_RTE_FOUND	0x1
175 #define RTE_FOUND	0x2
176 
177 static struct mtx mrouter6_mtx;
178 #define	MROUTER6_LOCK()		mtx_lock(&mrouter6_mtx)
179 #define	MROUTER6_UNLOCK()	mtx_unlock(&mrouter6_mtx)
180 #define	MROUTER6_LOCK_ASSERT()	do {					\
181 	mtx_assert(&mrouter6_mtx, MA_OWNED);				\
182 	NET_ASSERT_GIANT();						\
183 } while (0)
184 #define	MROUTER6_LOCK_INIT()	\
185 	mtx_init(&mrouter6_mtx, "IPv6 multicast forwarding", NULL, MTX_DEF)
186 #define	MROUTER6_LOCK_DESTROY()	mtx_destroy(&mrouter6_mtx)
187 
188 static struct mf6c *mf6ctable[MF6CTBLSIZ];
189 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mf6ctable, CTLFLAG_RD,
190     &mf6ctable, sizeof(mf6ctable), "S,*mf6ctable[MF6CTBLSIZ]",
191     "IPv6 Multicast Forwarding Table (struct *mf6ctable[MF6CTBLSIZ], "
192     "netinet6/ip6_mroute.h)");
193 
194 static struct mtx mfc6_mtx;
195 #define	MFC6_LOCK()		mtx_lock(&mfc6_mtx)
196 #define	MFC6_UNLOCK()		mtx_unlock(&mfc6_mtx)
197 #define	MFC6_LOCK_ASSERT()	do {					\
198 	mtx_assert(&mfc6_mtx, MA_OWNED);				\
199 	NET_ASSERT_GIANT();						\
200 } while (0)
201 #define	MFC6_LOCK_INIT()		\
202 	mtx_init(&mfc6_mtx, "IPv6 multicast forwarding cache", NULL, MTX_DEF)
203 #define	MFC6_LOCK_DESTROY()	mtx_destroy(&mfc6_mtx)
204 
205 static u_char n6expire[MF6CTBLSIZ];
206 
207 static struct mif6 mif6table[MAXMIFS];
208 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mif6table, CTLFLAG_RD,
209     &mif6table, sizeof(mif6table), "S,mif6[MAXMIFS]",
210     "IPv6 Multicast Interfaces (struct mif6[MAXMIFS], netinet6/ip6_mroute.h)");
211 
212 static struct mtx mif6_mtx;
213 #define	MIF6_LOCK()		mtx_lock(&mif6_mtx)
214 #define	MIF6_UNLOCK()		mtx_unlock(&mif6_mtx)
215 #define	MIF6_LOCK_ASSERT()	mtx_assert(&mif6_mtx, MA_OWNED)
216 #define	MIF6_LOCK_INIT()	\
217 	mtx_init(&mif6_mtx, "IPv6 multicast interfaces", NULL, MTX_DEF)
218 #define	MIF6_LOCK_DESTROY()	mtx_destroy(&mif6_mtx)
219 
220 #ifdef MRT6DEBUG
221 #ifdef VIMAGE_GLOBALS
222 static u_int mrt6debug;		/* debug level */
223 #endif
224 #define DEBUG_MFC	0x02
225 #define DEBUG_FORWARD	0x04
226 #define DEBUG_EXPIRE	0x08
227 #define DEBUG_XMIT	0x10
228 #define DEBUG_REG	0x20
229 #define DEBUG_PIM	0x40
230 #endif
231 
232 static void	expire_upcalls(void *);
233 #define	EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
234 #define	UPCALL_EXPIRE	6		/* number of timeouts */
235 
236 /*
237  * XXX TODO: maintain a count to if_allmulti() calls in struct ifnet.
238  */
239 
240 /*
241  * 'Interfaces' associated with decapsulator (so we can tell
242  * packets that went through it from ones that get reflected
243  * by a broken gateway).  Different from IPv4 register_if,
244  * these interfaces are linked into the system ifnet list,
245  * because per-interface IPv6 statistics are maintained in
246  * ifp->if_afdata.  But it does not have any routes point
247  * to them.  I.e., packets can't be sent this way.  They
248  * only exist as a placeholder for multicast source
249  * verification.
250  */
251 static struct ifnet *multicast_register_if6;
252 
253 #define ENCAP_HOPS 64
254 
255 /*
256  * Private variables.
257  */
258 static mifi_t nummifs = 0;
259 static mifi_t reg_mif_num = (mifi_t)-1;
260 
261 static struct pim6stat pim6stat;
262 SYSCTL_STRUCT(_net_inet6_pim, PIM6CTL_STATS, stats, CTLFLAG_RD,
263     &pim6stat, pim6stat,
264     "PIM Statistics (struct pim6stat, netinet6/pim_var.h)");
265 
266 #ifdef VIMAGE_GLOBALS
267 static int pim6;
268 #endif
269 
270 /*
271  * Hash function for a source, group entry
272  */
273 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
274 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
275 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
276 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
277 
278 /*
279  * Find a route for a given origin IPv6 address and Multicast group address.
280  */
281 #define MF6CFIND(o, g, rt) do { \
282 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
283 	rt = NULL; \
284 	mrt6stat.mrt6s_mfc_lookups++; \
285 	while (_rt) { \
286 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
287 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
288 		    (_rt->mf6c_stall == NULL)) { \
289 			rt = _rt; \
290 			break; \
291 		} \
292 		_rt = _rt->mf6c_next; \
293 	} \
294 	if (rt == NULL) { \
295 		mrt6stat.mrt6s_mfc_misses++; \
296 	} \
297 } while (/*CONSTCOND*/ 0)
298 
299 /*
300  * Macros to compute elapsed time efficiently
301  * Borrowed from Van Jacobson's scheduling code
302  * XXX: replace with timersub() ?
303  */
304 #define TV_DELTA(a, b, delta) do { \
305 	    int xxs; \
306 		\
307 	    delta = (a).tv_usec - (b).tv_usec; \
308 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
309 	       switch (xxs) { \
310 		      case 2: \
311 			  delta += 1000000; \
312 			      /* FALLTHROUGH */ \
313 		      case 1: \
314 			  delta += 1000000; \
315 			  break; \
316 		      default: \
317 			  delta += (1000000 * xxs); \
318 	       } \
319 	    } \
320 } while (/*CONSTCOND*/ 0)
321 
322 /* XXX: replace with timercmp(a, b, <) ? */
323 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
324 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
325 
326 #ifdef UPCALL_TIMING
327 #define UPCALL_MAX	50
328 static u_long upcall_data[UPCALL_MAX + 1];
329 static void collate();
330 #endif /* UPCALL_TIMING */
331 
332 static int ip6_mrouter_init(struct socket *, int, int);
333 static int add_m6fc(struct mf6cctl *);
334 static int add_m6if(struct mif6ctl *);
335 static int del_m6fc(struct mf6cctl *);
336 static int del_m6if(mifi_t *);
337 static int del_m6if_locked(mifi_t *);
338 static int get_mif6_cnt(struct sioc_mif_req6 *);
339 static int get_sg_cnt(struct sioc_sg_req6 *);
340 
341 static struct callout expire_upcalls_ch;
342 
343 int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *);
344 int X_ip6_mrouter_done(void);
345 int X_ip6_mrouter_set(struct socket *, struct sockopt *);
346 int X_ip6_mrouter_get(struct socket *, struct sockopt *);
347 int X_mrt6_ioctl(int, caddr_t);
348 
349 static void
350 pim6_init(void)
351 {
352 	INIT_VNET_INET6(curvnet);
353 
354 	V_ip6_mrouter_ver = 0;
355 #ifdef MRT6DEBUG
356 	V_mrt6debug = 0;	/* debug level */
357 #endif
358 }
359 
360 /*
361  * Handle MRT setsockopt commands to modify the multicast routing tables.
362  */
363 int
364 X_ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
365 {
366 	int error = 0;
367 	int optval;
368 	struct mif6ctl mifc;
369 	struct mf6cctl mfcc;
370 	mifi_t mifi;
371 
372 	if (so != V_ip6_mrouter && sopt->sopt_name != MRT6_INIT)
373 		return (EACCES);
374 
375 	switch (sopt->sopt_name) {
376 	case MRT6_INIT:
377 #ifdef MRT6_OINIT
378 	case MRT6_OINIT:
379 #endif
380 		error = sooptcopyin(sopt, &optval, sizeof(optval),
381 		    sizeof(optval));
382 		if (error)
383 			break;
384 		error = ip6_mrouter_init(so, optval, sopt->sopt_name);
385 		break;
386 	case MRT6_DONE:
387 		error = X_ip6_mrouter_done();
388 		break;
389 	case MRT6_ADD_MIF:
390 		error = sooptcopyin(sopt, &mifc, sizeof(mifc), sizeof(mifc));
391 		if (error)
392 			break;
393 		error = add_m6if(&mifc);
394 		break;
395 	case MRT6_ADD_MFC:
396 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
397 		if (error)
398 			break;
399 		error = add_m6fc(&mfcc);
400 		break;
401 	case MRT6_DEL_MFC:
402 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
403 		if (error)
404 			break;
405 		error = del_m6fc(&mfcc);
406 		break;
407 	case MRT6_DEL_MIF:
408 		error = sooptcopyin(sopt, &mifi, sizeof(mifi), sizeof(mifi));
409 		if (error)
410 			break;
411 		error = del_m6if(&mifi);
412 		break;
413 	case MRT6_PIM:
414 		error = sooptcopyin(sopt, &optval, sizeof(optval),
415 		    sizeof(optval));
416 		if (error)
417 			break;
418 		error = set_pim6(&optval);
419 		break;
420 	default:
421 		error = EOPNOTSUPP;
422 		break;
423 	}
424 
425 	return (error);
426 }
427 
428 /*
429  * Handle MRT getsockopt commands
430  */
431 int
432 X_ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
433 {
434 	INIT_VNET_INET6(curvnet);
435 	int error = 0;
436 
437 	if (so != V_ip6_mrouter)
438 		return (EACCES);
439 
440 	switch (sopt->sopt_name) {
441 		case MRT6_PIM:
442 			error = sooptcopyout(sopt, &V_pim6, sizeof(V_pim6));
443 			break;
444 	}
445 	return (error);
446 }
447 
448 /*
449  * Handle ioctl commands to obtain information from the cache
450  */
451 int
452 X_mrt6_ioctl(int cmd, caddr_t data)
453 {
454 	int ret;
455 
456 	ret = EINVAL;
457 
458 	switch (cmd) {
459 	case SIOCGETSGCNT_IN6:
460 		ret = get_sg_cnt((struct sioc_sg_req6 *)data);
461 		break;
462 
463 	case SIOCGETMIFCNT_IN6:
464 		ret = get_mif6_cnt((struct sioc_mif_req6 *)data);
465 		break;
466 
467 	default:
468 		break;
469 	}
470 
471 	return (ret);
472 }
473 
474 /*
475  * returns the packet, byte, rpf-failure count for the source group provided
476  */
477 static int
478 get_sg_cnt(struct sioc_sg_req6 *req)
479 {
480 	struct mf6c *rt;
481 	int ret;
482 
483 	ret = 0;
484 
485 	MFC6_LOCK();
486 
487 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
488 	if (rt == NULL) {
489 		ret = ESRCH;
490 	} else {
491 		req->pktcnt = rt->mf6c_pkt_cnt;
492 		req->bytecnt = rt->mf6c_byte_cnt;
493 		req->wrong_if = rt->mf6c_wrong_if;
494 	}
495 
496 	MFC6_UNLOCK();
497 
498 	return (ret);
499 }
500 
501 /*
502  * returns the input and output packet and byte counts on the mif provided
503  */
504 static int
505 get_mif6_cnt(struct sioc_mif_req6 *req)
506 {
507 	mifi_t mifi;
508 	int ret;
509 
510 	ret = 0;
511 	mifi = req->mifi;
512 
513 	MIF6_LOCK();
514 
515 	if (mifi >= nummifs) {
516 		ret = EINVAL;
517 	} else {
518 		req->icount = mif6table[mifi].m6_pkt_in;
519 		req->ocount = mif6table[mifi].m6_pkt_out;
520 		req->ibytes = mif6table[mifi].m6_bytes_in;
521 		req->obytes = mif6table[mifi].m6_bytes_out;
522 	}
523 
524 	MIF6_UNLOCK();
525 
526 	return (ret);
527 }
528 
529 static int
530 set_pim6(int *i)
531 {
532 	INIT_VNET_INET6(curvnet);
533 	if ((*i != 1) && (*i != 0))
534 		return (EINVAL);
535 
536 	V_pim6 = *i;
537 
538 	return (0);
539 }
540 
541 /*
542  * Enable multicast routing
543  */
544 static int
545 ip6_mrouter_init(struct socket *so, int v, int cmd)
546 {
547 	INIT_VNET_INET6(curvnet);
548 
549 	V_ip6_mrouter_ver = 0;
550 
551 #ifdef MRT6DEBUG
552 	V_mrt6debug = 0;
553 
554 	if (V_mrt6debug)
555 		log(LOG_DEBUG,
556 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
557 		    so->so_type, so->so_proto->pr_protocol);
558 #endif
559 
560 	if (so->so_type != SOCK_RAW ||
561 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
562 		return (EOPNOTSUPP);
563 
564 	if (v != 1)
565 		return (ENOPROTOOPT);
566 
567 	MROUTER6_LOCK();
568 
569 	if (V_ip6_mrouter != NULL) {
570 		MROUTER6_UNLOCK();
571 		return (EADDRINUSE);
572 	}
573 
574 	V_ip6_mrouter = so;
575 	V_ip6_mrouter_ver = cmd;
576 
577 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
578 	bzero((caddr_t)n6expire, sizeof(n6expire));
579 
580 	V_pim6 = 0;/* used for stubbing out/in pim stuff */
581 
582 	callout_init(&expire_upcalls_ch, 0);
583 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
584 	    expire_upcalls, NULL);
585 
586 	MROUTER6_UNLOCK();
587 
588 #ifdef MRT6DEBUG
589 	if (V_mrt6debug)
590 		log(LOG_DEBUG, "ip6_mrouter_init\n");
591 #endif
592 
593 	return (0);
594 }
595 
596 /*
597  * Disable IPv6 multicast forwarding.
598  */
599 int
600 X_ip6_mrouter_done(void)
601 {
602 	INIT_VNET_INET6(curvnet);
603 	mifi_t mifi;
604 	int i;
605 	struct mf6c *rt;
606 	struct rtdetq *rte;
607 
608 	MROUTER6_LOCK();
609 
610 	if (V_ip6_mrouter == NULL) {
611 		MROUTER6_UNLOCK();
612 		return (EINVAL);
613 	}
614 
615 	/*
616 	 * For each phyint in use, disable promiscuous reception of all IPv6
617 	 * multicasts.
618 	 */
619 	for (mifi = 0; mifi < nummifs; mifi++) {
620 		if (mif6table[mifi].m6_ifp &&
621 		    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
622 			if_allmulti(mif6table[mifi].m6_ifp, 0);
623 		}
624 	}
625 	bzero((caddr_t)mif6table, sizeof(mif6table));
626 	nummifs = 0;
627 
628 	V_pim6 = 0; /* used to stub out/in pim specific code */
629 
630 	callout_stop(&expire_upcalls_ch);
631 
632 	/*
633 	 * Free all multicast forwarding cache entries.
634 	 */
635 	MFC6_LOCK();
636 	for (i = 0; i < MF6CTBLSIZ; i++) {
637 		rt = mf6ctable[i];
638 		while (rt) {
639 			struct mf6c *frt;
640 
641 			for (rte = rt->mf6c_stall; rte != NULL; ) {
642 				struct rtdetq *n = rte->next;
643 
644 				m_free(rte->m);
645 				free(rte, M_MRTABLE6);
646 				rte = n;
647 			}
648 			frt = rt;
649 			rt = rt->mf6c_next;
650 			free(frt, M_MRTABLE6);
651 		}
652 	}
653 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
654 	MFC6_UNLOCK();
655 
656 	/*
657 	 * Reset register interface
658 	 */
659 	if (reg_mif_num != (mifi_t)-1 && multicast_register_if6 != NULL) {
660 		if_detach(multicast_register_if6);
661 		if_free(multicast_register_if6);
662 		reg_mif_num = (mifi_t)-1;
663 		multicast_register_if6 = NULL;
664 	}
665 
666 	V_ip6_mrouter = NULL;
667 	V_ip6_mrouter_ver = 0;
668 
669 	MROUTER6_UNLOCK();
670 
671 #ifdef MRT6DEBUG
672 	if (V_mrt6debug)
673 		log(LOG_DEBUG, "ip6_mrouter_done\n");
674 #endif
675 
676 	return (0);
677 }
678 
679 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
680 
681 /*
682  * Add a mif to the mif table
683  */
684 static int
685 add_m6if(struct mif6ctl *mifcp)
686 {
687 	INIT_VNET_NET(curvnet);
688 	struct mif6 *mifp;
689 	struct ifnet *ifp;
690 	int error;
691 
692 	MIF6_LOCK();
693 
694 	if (mifcp->mif6c_mifi >= MAXMIFS) {
695 		MIF6_UNLOCK();
696 		return (EINVAL);
697 	}
698 	mifp = mif6table + mifcp->mif6c_mifi;
699 	if (mifp->m6_ifp != NULL) {
700 		MIF6_UNLOCK();
701 		return (EADDRINUSE); /* XXX: is it appropriate? */
702 	}
703 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > V_if_index) {
704 		MIF6_UNLOCK();
705 		return (ENXIO);
706 	}
707 
708 	ifp = ifnet_byindex(mifcp->mif6c_pifi);
709 
710 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
711 		if (reg_mif_num == (mifi_t)-1) {
712 			ifp = if_alloc(IFT_OTHER);
713 
714 			if_initname(ifp, "register_mif", 0);
715 			ifp->if_flags |= IFF_LOOPBACK;
716 			if_attach(ifp);
717 			multicast_register_if6 = ifp;
718 			reg_mif_num = mifcp->mif6c_mifi;
719 			/*
720 			 * it is impossible to guess the ifindex of the
721 			 * register interface.  So mif6c_pifi is automatically
722 			 * calculated.
723 			 */
724 			mifcp->mif6c_pifi = ifp->if_index;
725 		} else {
726 			ifp = multicast_register_if6;
727 		}
728 	} else {
729 		/* Make sure the interface supports multicast */
730 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
731 			MIF6_UNLOCK();
732 			return (EOPNOTSUPP);
733 		}
734 
735 		error = if_allmulti(ifp, 1);
736 		if (error) {
737 			MIF6_UNLOCK();
738 			return (error);
739 		}
740 	}
741 
742 	mifp->m6_flags     = mifcp->mif6c_flags;
743 	mifp->m6_ifp       = ifp;
744 
745 	/* initialize per mif pkt counters */
746 	mifp->m6_pkt_in    = 0;
747 	mifp->m6_pkt_out   = 0;
748 	mifp->m6_bytes_in  = 0;
749 	mifp->m6_bytes_out = 0;
750 	bzero(&mifp->m6_route, sizeof(mifp->m6_route));
751 
752 	/* Adjust nummifs up if the mifi is higher than nummifs */
753 	if (nummifs <= mifcp->mif6c_mifi)
754 		nummifs = mifcp->mif6c_mifi + 1;
755 
756 	MIF6_UNLOCK();
757 
758 #ifdef MRT6DEBUG
759 	if (V_mrt6debug)
760 		log(LOG_DEBUG,
761 		    "add_mif #%d, phyint %s\n",
762 		    mifcp->mif6c_mifi,
763 		    ifp->if_xname);
764 #endif
765 
766 	return (0);
767 }
768 
769 /*
770  * Delete a mif from the mif table
771  */
772 static int
773 del_m6if_locked(mifi_t *mifip)
774 {
775 	struct mif6 *mifp = mif6table + *mifip;
776 	mifi_t mifi;
777 	struct ifnet *ifp;
778 
779 	MIF6_LOCK_ASSERT();
780 
781 	if (*mifip >= nummifs)
782 		return (EINVAL);
783 	if (mifp->m6_ifp == NULL)
784 		return (EINVAL);
785 
786 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
787 		/* XXX: TODO: Maintain an ALLMULTI refcount in struct ifnet. */
788 		ifp = mifp->m6_ifp;
789 		if_allmulti(ifp, 0);
790 	} else {
791 		if (reg_mif_num != (mifi_t)-1 &&
792 		    multicast_register_if6 != NULL) {
793 			if_detach(multicast_register_if6);
794 			if_free(multicast_register_if6);
795 			reg_mif_num = (mifi_t)-1;
796 			multicast_register_if6 = NULL;
797 		}
798 	}
799 
800 	bzero((caddr_t)mifp, sizeof(*mifp));
801 
802 	/* Adjust nummifs down */
803 	for (mifi = nummifs; mifi > 0; mifi--)
804 		if (mif6table[mifi - 1].m6_ifp)
805 			break;
806 	nummifs = mifi;
807 
808 #ifdef MRT6DEBUG
809 	if (V_mrt6debug)
810 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
811 #endif
812 
813 	return (0);
814 }
815 
816 static int
817 del_m6if(mifi_t *mifip)
818 {
819 	int cc;
820 
821 	MIF6_LOCK();
822 	cc = del_m6if_locked(mifip);
823 	MIF6_UNLOCK();
824 
825 	return (cc);
826 }
827 
828 /*
829  * Add an mfc entry
830  */
831 static int
832 add_m6fc(struct mf6cctl *mfccp)
833 {
834 	struct mf6c *rt;
835 	u_long hash;
836 	struct rtdetq *rte;
837 	u_short nstl;
838 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
839 
840 	MFC6_LOCK();
841 
842 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
843 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
844 
845 	/* If an entry already exists, just update the fields */
846 	if (rt) {
847 #ifdef MRT6DEBUG
848 		if (V_mrt6debug & DEBUG_MFC) {
849 		    log(LOG_DEBUG,
850 			"add_m6fc no upcall h %d o %s g %s p %x\n",
851 			ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
852 			ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
853 			mfccp->mf6cc_parent);
854 		}
855 #endif
856 
857 		rt->mf6c_parent = mfccp->mf6cc_parent;
858 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
859 
860 		MFC6_UNLOCK();
861 		return (0);
862 	}
863 
864 	/*
865 	 * Find the entry for which the upcall was made and update
866 	 */
867 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
868 			mfccp->mf6cc_mcastgrp.sin6_addr);
869 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
870 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
871 				       &mfccp->mf6cc_origin.sin6_addr) &&
872 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
873 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
874 		    (rt->mf6c_stall != NULL)) {
875 
876 			if (nstl++)
877 				log(LOG_ERR,
878 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
879 				    "multiple kernel entries",
880 				    ip6_sprintf(ip6bufo,
881 					    &mfccp->mf6cc_origin.sin6_addr),
882 				    ip6_sprintf(ip6bufg,
883 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
884 				    mfccp->mf6cc_parent, rt->mf6c_stall);
885 
886 #ifdef MRT6DEBUG
887 			if (V_mrt6debug & DEBUG_MFC)
888 				log(LOG_DEBUG,
889 				    "add_m6fc o %s g %s p %x dbg %x\n",
890 				    ip6_sprintf(ip6bufo,
891 					    &mfccp->mf6cc_origin.sin6_addr),
892 				    ip6_sprintf(ip6bufg,
893 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
894 				    mfccp->mf6cc_parent, rt->mf6c_stall);
895 #endif
896 
897 			rt->mf6c_origin     = mfccp->mf6cc_origin;
898 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
899 			rt->mf6c_parent     = mfccp->mf6cc_parent;
900 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
901 			/* initialize pkt counters per src-grp */
902 			rt->mf6c_pkt_cnt    = 0;
903 			rt->mf6c_byte_cnt   = 0;
904 			rt->mf6c_wrong_if   = 0;
905 
906 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
907 			n6expire[hash]--;
908 
909 			/* free packets Qed at the end of this entry */
910 			for (rte = rt->mf6c_stall; rte != NULL; ) {
911 				struct rtdetq *n = rte->next;
912 				ip6_mdq(rte->m, rte->ifp, rt);
913 				m_freem(rte->m);
914 #ifdef UPCALL_TIMING
915 				collate(&(rte->t));
916 #endif /* UPCALL_TIMING */
917 				free(rte, M_MRTABLE6);
918 				rte = n;
919 			}
920 			rt->mf6c_stall = NULL;
921 		}
922 	}
923 
924 	/*
925 	 * It is possible that an entry is being inserted without an upcall
926 	 */
927 	if (nstl == 0) {
928 #ifdef MRT6DEBUG
929 		if (V_mrt6debug & DEBUG_MFC)
930 		    log(LOG_DEBUG,
931 			"add_mfc no upcall h %d o %s g %s p %x\n",
932 			hash,
933 			ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
934 			ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
935 			mfccp->mf6cc_parent);
936 #endif
937 
938 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
939 
940 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
941 					       &mfccp->mf6cc_origin.sin6_addr)&&
942 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
943 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
944 
945 				rt->mf6c_origin     = mfccp->mf6cc_origin;
946 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
947 				rt->mf6c_parent     = mfccp->mf6cc_parent;
948 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
949 				/* initialize pkt counters per src-grp */
950 				rt->mf6c_pkt_cnt    = 0;
951 				rt->mf6c_byte_cnt   = 0;
952 				rt->mf6c_wrong_if   = 0;
953 
954 				if (rt->mf6c_expire)
955 					n6expire[hash]--;
956 				rt->mf6c_expire	   = 0;
957 			}
958 		}
959 		if (rt == NULL) {
960 			/* no upcall, so make a new entry */
961 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
962 						  M_NOWAIT);
963 			if (rt == NULL) {
964 				MFC6_UNLOCK();
965 				return (ENOBUFS);
966 			}
967 
968 			/* insert new entry at head of hash chain */
969 			rt->mf6c_origin     = mfccp->mf6cc_origin;
970 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
971 			rt->mf6c_parent     = mfccp->mf6cc_parent;
972 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
973 			/* initialize pkt counters per src-grp */
974 			rt->mf6c_pkt_cnt    = 0;
975 			rt->mf6c_byte_cnt   = 0;
976 			rt->mf6c_wrong_if   = 0;
977 			rt->mf6c_expire     = 0;
978 			rt->mf6c_stall = NULL;
979 
980 			/* link into table */
981 			rt->mf6c_next  = mf6ctable[hash];
982 			mf6ctable[hash] = rt;
983 		}
984 	}
985 
986 	MFC6_UNLOCK();
987 	return (0);
988 }
989 
990 #ifdef UPCALL_TIMING
991 /*
992  * collect delay statistics on the upcalls
993  */
994 static void
995 collate(struct timeval *t)
996 {
997 	u_long d;
998 	struct timeval tp;
999 	u_long delta;
1000 
1001 	GET_TIME(tp);
1002 
1003 	if (TV_LT(*t, tp))
1004 	{
1005 		TV_DELTA(tp, *t, delta);
1006 
1007 		d = delta >> 10;
1008 		if (d > UPCALL_MAX)
1009 			d = UPCALL_MAX;
1010 
1011 		++upcall_data[d];
1012 	}
1013 }
1014 #endif /* UPCALL_TIMING */
1015 
1016 /*
1017  * Delete an mfc entry
1018  */
1019 static int
1020 del_m6fc(struct mf6cctl *mfccp)
1021 {
1022 	struct sockaddr_in6	origin;
1023 	struct sockaddr_in6	mcastgrp;
1024 	struct mf6c		*rt;
1025 	struct mf6c		**nptr;
1026 	u_long		hash;
1027 
1028 	origin = mfccp->mf6cc_origin;
1029 	mcastgrp = mfccp->mf6cc_mcastgrp;
1030 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
1031 
1032 #ifdef MRT6DEBUG
1033 	if (V_mrt6debug & DEBUG_MFC) {
1034 		char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1035 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
1036 		    ip6_sprintf(ip6bufo, &origin.sin6_addr),
1037 		    ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr));
1038 	}
1039 #endif
1040 
1041 	MFC6_LOCK();
1042 
1043 	nptr = &mf6ctable[hash];
1044 	while ((rt = *nptr) != NULL) {
1045 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
1046 				       &rt->mf6c_origin.sin6_addr) &&
1047 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
1048 				       &rt->mf6c_mcastgrp.sin6_addr) &&
1049 		    rt->mf6c_stall == NULL)
1050 			break;
1051 
1052 		nptr = &rt->mf6c_next;
1053 	}
1054 	if (rt == NULL) {
1055 		MFC6_UNLOCK();
1056 		return (EADDRNOTAVAIL);
1057 	}
1058 
1059 	*nptr = rt->mf6c_next;
1060 	free(rt, M_MRTABLE6);
1061 
1062 	MFC6_UNLOCK();
1063 
1064 	return (0);
1065 }
1066 
1067 static int
1068 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
1069 {
1070 
1071 	if (s) {
1072 		if (sbappendaddr(&s->so_rcv,
1073 				 (struct sockaddr *)src,
1074 				 mm, (struct mbuf *)0) != 0) {
1075 			sorwakeup(s);
1076 			return (0);
1077 		}
1078 	}
1079 	m_freem(mm);
1080 	return (-1);
1081 }
1082 
1083 /*
1084  * IPv6 multicast forwarding function. This function assumes that the packet
1085  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
1086  * pointed to by "ifp", and the packet is to be relayed to other networks
1087  * that have members of the packet's destination IPv6 multicast group.
1088  *
1089  * The packet is returned unscathed to the caller, unless it is
1090  * erroneous, in which case a non-zero return value tells the caller to
1091  * discard it.
1092  *
1093  * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff
1094  * this function is called in the originating context (i.e., not when
1095  * forwarding a packet from other node).  ip6_output(), which is currently the
1096  * only function that calls this function is called in the originating context,
1097  * explicitly ensures this condition.  It is caller's responsibility to ensure
1098  * that if this function is called from somewhere else in the originating
1099  * context in the future.
1100  */
1101 int
1102 X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
1103 {
1104 	INIT_VNET_INET6(curvnet);
1105 	struct mf6c *rt;
1106 	struct mif6 *mifp;
1107 	struct mbuf *mm;
1108 	mifi_t mifi;
1109 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1110 
1111 #ifdef MRT6DEBUG
1112 	if (V_mrt6debug & DEBUG_FORWARD)
1113 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
1114 		    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1115 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1116 		    ifp->if_index);
1117 #endif
1118 
1119 	/*
1120 	 * Don't forward a packet with Hop limit of zero or one,
1121 	 * or a packet destined to a local-only group.
1122 	 */
1123 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
1124 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1125 		return (0);
1126 	ip6->ip6_hlim--;
1127 
1128 	/*
1129 	 * Source address check: do not forward packets with unspecified
1130 	 * source. It was discussed in July 2000, on ipngwg mailing list.
1131 	 * This is rather more serious than unicast cases, because some
1132 	 * MLD packets can be sent with the unspecified source address
1133 	 * (although such packets must normally set 1 to the hop limit field).
1134 	 */
1135 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1136 		V_ip6stat.ip6s_cantforward++;
1137 		if (V_ip6_log_time + V_ip6_log_interval < time_second) {
1138 			V_ip6_log_time = time_second;
1139 			log(LOG_DEBUG,
1140 			    "cannot forward "
1141 			    "from %s to %s nxt %d received on %s\n",
1142 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1143 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1144 			    ip6->ip6_nxt,
1145 			    if_name(m->m_pkthdr.rcvif));
1146 		}
1147 		return (0);
1148 	}
1149 
1150 	MFC6_LOCK();
1151 
1152 	/*
1153 	 * Determine forwarding mifs from the forwarding cache table
1154 	 */
1155 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
1156 
1157 	/* Entry exists, so forward if necessary */
1158 	if (rt) {
1159 		MFC6_UNLOCK();
1160 		return (ip6_mdq(m, ifp, rt));
1161 	} else {
1162 		/*
1163 		 * If we don't have a route for packet's origin,
1164 		 * Make a copy of the packet &
1165 		 * send message to routing daemon
1166 		 */
1167 
1168 		struct mbuf *mb0;
1169 		struct rtdetq *rte;
1170 		u_long hash;
1171 /*		int i, npkts;*/
1172 #ifdef UPCALL_TIMING
1173 		struct timeval tp;
1174 
1175 		GET_TIME(tp);
1176 #endif /* UPCALL_TIMING */
1177 
1178 		mrt6stat.mrt6s_no_route++;
1179 #ifdef MRT6DEBUG
1180 		if (V_mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
1181 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
1182 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1183 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1184 #endif
1185 
1186 		/*
1187 		 * Allocate mbufs early so that we don't do extra work if we
1188 		 * are just going to fail anyway.
1189 		 */
1190 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE6,
1191 					      M_NOWAIT);
1192 		if (rte == NULL) {
1193 			MFC6_UNLOCK();
1194 			return (ENOBUFS);
1195 		}
1196 		mb0 = m_copy(m, 0, M_COPYALL);
1197 		/*
1198 		 * Pullup packet header if needed before storing it,
1199 		 * as other references may modify it in the meantime.
1200 		 */
1201 		if (mb0 &&
1202 		    (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1203 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1204 		if (mb0 == NULL) {
1205 			free(rte, M_MRTABLE6);
1206 			MFC6_UNLOCK();
1207 			return (ENOBUFS);
1208 		}
1209 
1210 		/* is there an upcall waiting for this packet? */
1211 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1212 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1213 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1214 					       &rt->mf6c_origin.sin6_addr) &&
1215 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1216 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1217 			    (rt->mf6c_stall != NULL))
1218 				break;
1219 		}
1220 
1221 		if (rt == NULL) {
1222 			struct mrt6msg *im;
1223 #ifdef MRT6_OINIT
1224 			struct omrt6msg *oim;
1225 #endif
1226 
1227 			/* no upcall, so make a new entry */
1228 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
1229 						  M_NOWAIT);
1230 			if (rt == NULL) {
1231 				free(rte, M_MRTABLE6);
1232 				m_freem(mb0);
1233 				MFC6_UNLOCK();
1234 				return (ENOBUFS);
1235 			}
1236 			/*
1237 			 * Make a copy of the header to send to the user
1238 			 * level process
1239 			 */
1240 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1241 
1242 			if (mm == NULL) {
1243 				free(rte, M_MRTABLE6);
1244 				m_freem(mb0);
1245 				free(rt, M_MRTABLE6);
1246 				MFC6_UNLOCK();
1247 				return (ENOBUFS);
1248 			}
1249 
1250 			/*
1251 			 * Send message to routing daemon
1252 			 */
1253 			sin6.sin6_addr = ip6->ip6_src;
1254 
1255 			im = NULL;
1256 #ifdef MRT6_OINIT
1257 			oim = NULL;
1258 #endif
1259 			switch (V_ip6_mrouter_ver) {
1260 #ifdef MRT6_OINIT
1261 			case MRT6_OINIT:
1262 				oim = mtod(mm, struct omrt6msg *);
1263 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1264 				oim->im6_mbz = 0;
1265 				break;
1266 #endif
1267 			case MRT6_INIT:
1268 				im = mtod(mm, struct mrt6msg *);
1269 				im->im6_msgtype = MRT6MSG_NOCACHE;
1270 				im->im6_mbz = 0;
1271 				break;
1272 			default:
1273 				free(rte, M_MRTABLE6);
1274 				m_freem(mb0);
1275 				free(rt, M_MRTABLE6);
1276 				MFC6_UNLOCK();
1277 				return (EINVAL);
1278 			}
1279 
1280 #ifdef MRT6DEBUG
1281 			if (V_mrt6debug & DEBUG_FORWARD)
1282 				log(LOG_DEBUG,
1283 				    "getting the iif info in the kernel\n");
1284 #endif
1285 
1286 			for (mifp = mif6table, mifi = 0;
1287 			     mifi < nummifs && mifp->m6_ifp != ifp;
1288 			     mifp++, mifi++)
1289 				;
1290 
1291 			switch (V_ip6_mrouter_ver) {
1292 #ifdef MRT6_OINIT
1293 			case MRT6_OINIT:
1294 				oim->im6_mif = mifi;
1295 				break;
1296 #endif
1297 			case MRT6_INIT:
1298 				im->im6_mif = mifi;
1299 				break;
1300 			}
1301 
1302 			if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1303 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1304 				    "socket queue full\n");
1305 				mrt6stat.mrt6s_upq_sockfull++;
1306 				free(rte, M_MRTABLE6);
1307 				m_freem(mb0);
1308 				free(rt, M_MRTABLE6);
1309 				MFC6_UNLOCK();
1310 				return (ENOBUFS);
1311 			}
1312 
1313 			mrt6stat.mrt6s_upcalls++;
1314 
1315 			/* insert new entry at head of hash chain */
1316 			bzero(rt, sizeof(*rt));
1317 			rt->mf6c_origin.sin6_family = AF_INET6;
1318 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1319 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1320 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1321 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1322 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1323 			rt->mf6c_expire = UPCALL_EXPIRE;
1324 			n6expire[hash]++;
1325 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1326 
1327 			/* link into table */
1328 			rt->mf6c_next  = mf6ctable[hash];
1329 			mf6ctable[hash] = rt;
1330 			/* Add this entry to the end of the queue */
1331 			rt->mf6c_stall = rte;
1332 		} else {
1333 			/* determine if q has overflowed */
1334 			struct rtdetq **p;
1335 			int npkts = 0;
1336 
1337 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1338 				if (++npkts > MAX_UPQ6) {
1339 					mrt6stat.mrt6s_upq_ovflw++;
1340 					free(rte, M_MRTABLE6);
1341 					m_freem(mb0);
1342 					MFC6_UNLOCK();
1343 					return (0);
1344 				}
1345 
1346 			/* Add this entry to the end of the queue */
1347 			*p = rte;
1348 		}
1349 
1350 		rte->next = NULL;
1351 		rte->m = mb0;
1352 		rte->ifp = ifp;
1353 #ifdef UPCALL_TIMING
1354 		rte->t = tp;
1355 #endif /* UPCALL_TIMING */
1356 
1357 		MFC6_UNLOCK();
1358 
1359 		return (0);
1360 	}
1361 }
1362 
1363 /*
1364  * Clean up cache entries if upcalls are not serviced
1365  * Call from the Slow Timeout mechanism, every half second.
1366  */
1367 static void
1368 expire_upcalls(void *unused)
1369 {
1370 	struct rtdetq *rte;
1371 	struct mf6c *mfc, **nptr;
1372 	int i;
1373 
1374 	MFC6_LOCK();
1375 	for (i = 0; i < MF6CTBLSIZ; i++) {
1376 		if (n6expire[i] == 0)
1377 			continue;
1378 		nptr = &mf6ctable[i];
1379 		while ((mfc = *nptr) != NULL) {
1380 			rte = mfc->mf6c_stall;
1381 			/*
1382 			 * Skip real cache entries
1383 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1384 			 * If it expires now
1385 			 */
1386 			if (rte != NULL &&
1387 			    mfc->mf6c_expire != 0 &&
1388 			    --mfc->mf6c_expire == 0) {
1389 #ifdef MRT6DEBUG
1390 				if (V_mrt6debug & DEBUG_EXPIRE) {
1391 					char ip6bufo[INET6_ADDRSTRLEN];
1392 					char ip6bufg[INET6_ADDRSTRLEN];
1393 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1394 					    ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr),
1395 					    ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr));
1396 				}
1397 #endif
1398 				/*
1399 				 * drop all the packets
1400 				 * free the mbuf with the pkt, if, timing info
1401 				 */
1402 				do {
1403 					struct rtdetq *n = rte->next;
1404 					m_freem(rte->m);
1405 					free(rte, M_MRTABLE6);
1406 					rte = n;
1407 				} while (rte != NULL);
1408 				mrt6stat.mrt6s_cache_cleanups++;
1409 				n6expire[i]--;
1410 
1411 				*nptr = mfc->mf6c_next;
1412 				free(mfc, M_MRTABLE6);
1413 			} else {
1414 				nptr = &mfc->mf6c_next;
1415 			}
1416 		}
1417 	}
1418 	MFC6_UNLOCK();
1419 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1420 	    expire_upcalls, NULL);
1421 }
1422 
1423 /*
1424  * Packet forwarding routine once entry in the cache is made
1425  */
1426 static int
1427 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1428 {
1429 	INIT_VNET_INET6(curvnet);
1430 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1431 	mifi_t mifi, iif;
1432 	struct mif6 *mifp;
1433 	int plen = m->m_pkthdr.len;
1434 	struct in6_addr src0, dst0; /* copies for local work */
1435 	u_int32_t iszone, idzone, oszone, odzone;
1436 	int error = 0;
1437 
1438 /*
1439  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1440  * input, they shouldn't get counted on output, so statistics keeping is
1441  * separate.
1442  */
1443 
1444 #define MC6_SEND(ip6, mifp, m) do {				\
1445 	if ((mifp)->m6_flags & MIFF_REGISTER)			\
1446 		register_send((ip6), (mifp), (m));		\
1447 	else							\
1448 		phyint_send((ip6), (mifp), (m));		\
1449 } while (/*CONSTCOND*/ 0)
1450 
1451 	/*
1452 	 * Don't forward if it didn't arrive from the parent mif
1453 	 * for its origin.
1454 	 */
1455 	mifi = rt->mf6c_parent;
1456 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1457 		/* came in the wrong interface */
1458 #ifdef MRT6DEBUG
1459 		if (V_mrt6debug & DEBUG_FORWARD)
1460 			log(LOG_DEBUG,
1461 			    "wrong if: ifid %d mifi %d mififid %x\n",
1462 			    ifp->if_index, mifi,
1463 			    mif6table[mifi].m6_ifp->if_index);
1464 #endif
1465 		mrt6stat.mrt6s_wrong_if++;
1466 		rt->mf6c_wrong_if++;
1467 		/*
1468 		 * If we are doing PIM processing, and we are forwarding
1469 		 * packets on this interface, send a message to the
1470 		 * routing daemon.
1471 		 */
1472 		/* have to make sure this is a valid mif */
1473 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1474 			if (V_pim6 && (m->m_flags & M_LOOP) == 0) {
1475 				/*
1476 				 * Check the M_LOOP flag to avoid an
1477 				 * unnecessary PIM assert.
1478 				 * XXX: M_LOOP is an ad-hoc hack...
1479 				 */
1480 				static struct sockaddr_in6 sin6 =
1481 				{ sizeof(sin6), AF_INET6 };
1482 
1483 				struct mbuf *mm;
1484 				struct mrt6msg *im;
1485 #ifdef MRT6_OINIT
1486 				struct omrt6msg *oim;
1487 #endif
1488 
1489 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1490 				if (mm &&
1491 				    (M_HASCL(mm) ||
1492 				     mm->m_len < sizeof(struct ip6_hdr)))
1493 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1494 				if (mm == NULL)
1495 					return (ENOBUFS);
1496 
1497 #ifdef MRT6_OINIT
1498 				oim = NULL;
1499 #endif
1500 				im = NULL;
1501 				switch (V_ip6_mrouter_ver) {
1502 #ifdef MRT6_OINIT
1503 				case MRT6_OINIT:
1504 					oim = mtod(mm, struct omrt6msg *);
1505 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1506 					oim->im6_mbz = 0;
1507 					break;
1508 #endif
1509 				case MRT6_INIT:
1510 					im = mtod(mm, struct mrt6msg *);
1511 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1512 					im->im6_mbz = 0;
1513 					break;
1514 				default:
1515 					m_freem(mm);
1516 					return (EINVAL);
1517 				}
1518 
1519 				for (mifp = mif6table, iif = 0;
1520 				     iif < nummifs && mifp &&
1521 					     mifp->m6_ifp != ifp;
1522 				     mifp++, iif++)
1523 					;
1524 
1525 				switch (V_ip6_mrouter_ver) {
1526 #ifdef MRT6_OINIT
1527 				case MRT6_OINIT:
1528 					oim->im6_mif = iif;
1529 					sin6.sin6_addr = oim->im6_src;
1530 					break;
1531 #endif
1532 				case MRT6_INIT:
1533 					im->im6_mif = iif;
1534 					sin6.sin6_addr = im->im6_src;
1535 					break;
1536 				}
1537 
1538 				mrt6stat.mrt6s_upcalls++;
1539 
1540 				if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1541 #ifdef MRT6DEBUG
1542 					if (V_mrt6debug)
1543 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1544 #endif
1545 					++mrt6stat.mrt6s_upq_sockfull;
1546 					return (ENOBUFS);
1547 				}	/* if socket Q full */
1548 			}		/* if PIM */
1549 		return (0);
1550 	}			/* if wrong iif */
1551 
1552 	/* If I sourced this packet, it counts as output, else it was input. */
1553 	if (m->m_pkthdr.rcvif == NULL) {
1554 		/* XXX: is rcvif really NULL when output?? */
1555 		mif6table[mifi].m6_pkt_out++;
1556 		mif6table[mifi].m6_bytes_out += plen;
1557 	} else {
1558 		mif6table[mifi].m6_pkt_in++;
1559 		mif6table[mifi].m6_bytes_in += plen;
1560 	}
1561 	rt->mf6c_pkt_cnt++;
1562 	rt->mf6c_byte_cnt += plen;
1563 
1564 	/*
1565 	 * For each mif, forward a copy of the packet if there are group
1566 	 * members downstream on the interface.
1567 	 */
1568 	src0 = ip6->ip6_src;
1569 	dst0 = ip6->ip6_dst;
1570 	if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 ||
1571 	    (error = in6_setscope(&dst0, ifp, &idzone)) != 0) {
1572 		V_ip6stat.ip6s_badscope++;
1573 		return (error);
1574 	}
1575 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) {
1576 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1577 			/*
1578 			 * check if the outgoing packet is going to break
1579 			 * a scope boundary.
1580 			 * XXX For packets through PIM register tunnel
1581 			 * interface, we believe a routing daemon.
1582 			 */
1583 			if (!(mif6table[rt->mf6c_parent].m6_flags &
1584 			      MIFF_REGISTER) &&
1585 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
1586 				if (in6_setscope(&src0, mif6table[mifi].m6_ifp,
1587 				    &oszone) ||
1588 				    in6_setscope(&dst0, mif6table[mifi].m6_ifp,
1589 				    &odzone) ||
1590 				    iszone != oszone ||
1591 				    idzone != odzone) {
1592 					V_ip6stat.ip6s_badscope++;
1593 					continue;
1594 				}
1595 			}
1596 
1597 			mifp->m6_pkt_out++;
1598 			mifp->m6_bytes_out += plen;
1599 			MC6_SEND(ip6, mifp, m);
1600 		}
1601 	}
1602 	return (0);
1603 }
1604 
1605 static void
1606 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1607 {
1608 	INIT_VNET_INET6(curvnet);
1609 	struct mbuf *mb_copy;
1610 	struct ifnet *ifp = mifp->m6_ifp;
1611 	int error = 0;
1612 	struct sockaddr_in6 *dst6;
1613 	u_long linkmtu;
1614 
1615 	dst6 = &mifp->m6_route.ro_dst;
1616 
1617 	/*
1618 	 * Make a new reference to the packet; make sure that
1619 	 * the IPv6 header is actually copied, not just referenced,
1620 	 * so that ip6_output() only scribbles on the copy.
1621 	 */
1622 	mb_copy = m_copy(m, 0, M_COPYALL);
1623 	if (mb_copy &&
1624 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1625 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1626 	if (mb_copy == NULL) {
1627 		return;
1628 	}
1629 	/* set MCAST flag to the outgoing packet */
1630 	mb_copy->m_flags |= M_MCAST;
1631 
1632 	/*
1633 	 * If we sourced the packet, call ip6_output since we may devide
1634 	 * the packet into fragments when the packet is too big for the
1635 	 * outgoing interface.
1636 	 * Otherwise, we can simply send the packet to the interface
1637 	 * sending queue.
1638 	 */
1639 	if (m->m_pkthdr.rcvif == NULL) {
1640 		struct ip6_moptions im6o;
1641 
1642 		im6o.im6o_multicast_ifp = ifp;
1643 		/* XXX: ip6_output will override ip6->ip6_hlim */
1644 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1645 		im6o.im6o_multicast_loop = 1;
1646 		error = ip6_output(mb_copy, NULL, &mifp->m6_route,
1647 				   IPV6_FORWARDING, &im6o, NULL, NULL);
1648 
1649 #ifdef MRT6DEBUG
1650 		if (V_mrt6debug & DEBUG_XMIT)
1651 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1652 			    mifp - mif6table, error);
1653 #endif
1654 		return;
1655 	}
1656 
1657 	/*
1658 	 * If configured to loop back multicasts by default,
1659 	 * loop back a copy now.
1660 	 */
1661 	if (in6_mcast_loop) {
1662 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1663 		dst6->sin6_family = AF_INET6;
1664 		dst6->sin6_addr = ip6->ip6_dst;
1665 		ip6_mloopback(ifp, m, &mifp->m6_route.ro_dst);
1666 	}
1667 
1668 	/*
1669 	 * Put the packet into the sending queue of the outgoing interface
1670 	 * if it would fit in the MTU of the interface.
1671 	 */
1672 	linkmtu = IN6_LINKMTU(ifp);
1673 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1674 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1675 		dst6->sin6_family = AF_INET6;
1676 		dst6->sin6_addr = ip6->ip6_dst;
1677 		/*
1678 		 * We just call if_output instead of nd6_output here, since
1679 		 * we need no ND for a multicast forwarded packet...right?
1680 		 */
1681 		error = (*ifp->if_output)(ifp, mb_copy,
1682 		    (struct sockaddr *)&mifp->m6_route.ro_dst, NULL);
1683 #ifdef MRT6DEBUG
1684 		if (V_mrt6debug & DEBUG_XMIT)
1685 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1686 			    mifp - mif6table, error);
1687 #endif
1688 	} else {
1689 		/*
1690 		 * pMTU discovery is intentionally disabled by default, since
1691 		 * various router may notify pMTU in multicast, which can be
1692 		 * a DDoS to a router
1693 		 */
1694 		if (V_ip6_mcast_pmtu)
1695 			icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1696 		else {
1697 #ifdef MRT6DEBUG
1698 			if (V_mrt6debug & DEBUG_XMIT) {
1699 				char ip6bufs[INET6_ADDRSTRLEN];
1700 				char ip6bufd[INET6_ADDRSTRLEN];
1701 				log(LOG_DEBUG,
1702 				    "phyint_send: packet too big on %s o %s "
1703 				    "g %s size %d(discarded)\n",
1704 				    if_name(ifp),
1705 				    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1706 				    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1707 				    mb_copy->m_pkthdr.len);
1708 			}
1709 #endif /* MRT6DEBUG */
1710 			m_freem(mb_copy); /* simply discard the packet */
1711 		}
1712 	}
1713 }
1714 
1715 static int
1716 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1717 {
1718 	struct mbuf *mm;
1719 	int i, len = m->m_pkthdr.len;
1720 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1721 	struct mrt6msg *im6;
1722 
1723 #ifdef MRT6DEBUG
1724 	if (V_mrt6debug) {
1725 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1726 		log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1727 		    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1728 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1729 	}
1730 #endif
1731 	++pim6stat.pim6s_snd_registers;
1732 
1733 	/* Make a copy of the packet to send to the user level process */
1734 	MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1735 	if (mm == NULL)
1736 		return (ENOBUFS);
1737 	mm->m_pkthdr.rcvif = NULL;
1738 	mm->m_data += max_linkhdr;
1739 	mm->m_len = sizeof(struct ip6_hdr);
1740 
1741 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1742 		m_freem(mm);
1743 		return (ENOBUFS);
1744 	}
1745 	i = MHLEN - M_LEADINGSPACE(mm);
1746 	if (i > len)
1747 		i = len;
1748 	mm = m_pullup(mm, i);
1749 	if (mm == NULL)
1750 		return (ENOBUFS);
1751 /* TODO: check it! */
1752 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1753 
1754 	/*
1755 	 * Send message to routing daemon
1756 	 */
1757 	sin6.sin6_addr = ip6->ip6_src;
1758 
1759 	im6 = mtod(mm, struct mrt6msg *);
1760 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1761 	im6->im6_mbz          = 0;
1762 
1763 	im6->im6_mif = mif - mif6table;
1764 
1765 	/* iif info is not given for reg. encap.n */
1766 	mrt6stat.mrt6s_upcalls++;
1767 
1768 	if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1769 #ifdef MRT6DEBUG
1770 		if (V_mrt6debug)
1771 			log(LOG_WARNING,
1772 			    "register_send: ip6_mrouter socket queue full\n");
1773 #endif
1774 		++mrt6stat.mrt6s_upq_sockfull;
1775 		return (ENOBUFS);
1776 	}
1777 	return (0);
1778 }
1779 
1780 /*
1781  * pim6_encapcheck() is called by the encap6_input() path at runtime to
1782  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
1783  * into the kernel.
1784  */
1785 static int
1786 pim6_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
1787 {
1788 
1789 #ifdef DIAGNOSTIC
1790     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
1791 #endif
1792     if (proto != IPPROTO_PIM)
1793 	return 0;	/* not for us; reject the datagram. */
1794 
1795     return 64;		/* claim the datagram. */
1796 }
1797 
1798 /*
1799  * PIM sparse mode hook
1800  * Receives the pim control messages, and passes them up to the listening
1801  * socket, using rip6_input.
1802  * The only message processed is the REGISTER pim message; the pim header
1803  * is stripped off, and the inner packet is passed to register_mforward.
1804  */
1805 int
1806 pim6_input(struct mbuf **mp, int *offp, int proto)
1807 {
1808 	INIT_VNET_INET6(curvnet);
1809 	struct pim *pim; /* pointer to a pim struct */
1810 	struct ip6_hdr *ip6;
1811 	int pimlen;
1812 	struct mbuf *m = *mp;
1813 	int minlen;
1814 	int off = *offp;
1815 
1816 	++pim6stat.pim6s_rcv_total;
1817 
1818 	ip6 = mtod(m, struct ip6_hdr *);
1819 	pimlen = m->m_pkthdr.len - *offp;
1820 
1821 	/*
1822 	 * Validate lengths
1823 	 */
1824 	if (pimlen < PIM_MINLEN) {
1825 		++pim6stat.pim6s_rcv_tooshort;
1826 #ifdef MRT6DEBUG
1827 		if (V_mrt6debug & DEBUG_PIM)
1828 			log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1829 #endif
1830 		m_freem(m);
1831 		return (IPPROTO_DONE);
1832 	}
1833 
1834 	/*
1835 	 * if the packet is at least as big as a REGISTER, go ahead
1836 	 * and grab the PIM REGISTER header size, to avoid another
1837 	 * possible m_pullup() later.
1838 	 *
1839 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1840 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1841 	 */
1842 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1843 
1844 	/*
1845 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1846 	 * possibly the PIM REGISTER header
1847 	 */
1848 #ifndef PULLDOWN_TEST
1849 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1850 	/* adjust pointer */
1851 	ip6 = mtod(m, struct ip6_hdr *);
1852 
1853 	/* adjust mbuf to point to the PIM header */
1854 	pim = (struct pim *)((caddr_t)ip6 + off);
1855 #else
1856 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1857 	if (pim == NULL) {
1858 		pim6stat.pim6s_rcv_tooshort++;
1859 		return (IPPROTO_DONE);
1860 	}
1861 #endif
1862 
1863 #define PIM6_CHECKSUM
1864 #ifdef PIM6_CHECKSUM
1865 	{
1866 		int cksumlen;
1867 
1868 		/*
1869 		 * Validate checksum.
1870 		 * If PIM REGISTER, exclude the data packet
1871 		 */
1872 		if (pim->pim_type == PIM_REGISTER)
1873 			cksumlen = PIM_MINLEN;
1874 		else
1875 			cksumlen = pimlen;
1876 
1877 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1878 			++pim6stat.pim6s_rcv_badsum;
1879 #ifdef MRT6DEBUG
1880 			if (V_mrt6debug & DEBUG_PIM)
1881 				log(LOG_DEBUG,
1882 				    "pim6_input: invalid checksum\n");
1883 #endif
1884 			m_freem(m);
1885 			return (IPPROTO_DONE);
1886 		}
1887 	}
1888 #endif /* PIM_CHECKSUM */
1889 
1890 	/* PIM version check */
1891 	if (pim->pim_ver != PIM_VERSION) {
1892 		++pim6stat.pim6s_rcv_badversion;
1893 #ifdef MRT6DEBUG
1894 		log(LOG_ERR,
1895 		    "pim6_input: incorrect version %d, expecting %d\n",
1896 		    pim->pim_ver, PIM_VERSION);
1897 #endif
1898 		m_freem(m);
1899 		return (IPPROTO_DONE);
1900 	}
1901 
1902 	if (pim->pim_type == PIM_REGISTER) {
1903 		/*
1904 		 * since this is a REGISTER, we'll make a copy of the register
1905 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1906 		 * routing daemon.
1907 		 */
1908 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1909 
1910 		struct mbuf *mcp;
1911 		struct ip6_hdr *eip6;
1912 		u_int32_t *reghdr;
1913 		int rc;
1914 #ifdef MRT6DEBUG
1915 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1916 #endif
1917 
1918 		++pim6stat.pim6s_rcv_registers;
1919 
1920 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1921 #ifdef MRT6DEBUG
1922 			if (V_mrt6debug & DEBUG_PIM)
1923 				log(LOG_DEBUG,
1924 				    "pim6_input: register mif not set: %d\n",
1925 				    reg_mif_num);
1926 #endif
1927 			m_freem(m);
1928 			return (IPPROTO_DONE);
1929 		}
1930 
1931 		reghdr = (u_int32_t *)(pim + 1);
1932 
1933 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1934 			goto pim6_input_to_daemon;
1935 
1936 		/*
1937 		 * Validate length
1938 		 */
1939 		if (pimlen < PIM6_REG_MINLEN) {
1940 			++pim6stat.pim6s_rcv_tooshort;
1941 			++pim6stat.pim6s_rcv_badregisters;
1942 #ifdef MRT6DEBUG
1943 			log(LOG_ERR,
1944 			    "pim6_input: register packet size too "
1945 			    "small %d from %s\n",
1946 			    pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src));
1947 #endif
1948 			m_freem(m);
1949 			return (IPPROTO_DONE);
1950 		}
1951 
1952 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1953 #ifdef MRT6DEBUG
1954 		if (V_mrt6debug & DEBUG_PIM)
1955 			log(LOG_DEBUG,
1956 			    "pim6_input[register], eip6: %s -> %s, "
1957 			    "eip6 plen %d\n",
1958 			    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1959 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1960 			    ntohs(eip6->ip6_plen));
1961 #endif
1962 
1963 		/* verify the version number of the inner packet */
1964 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1965 			++pim6stat.pim6s_rcv_badregisters;
1966 #ifdef MRT6DEBUG
1967 			log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1968 			    "of the inner packet\n",
1969 			    (eip6->ip6_vfc & IPV6_VERSION));
1970 #endif
1971 			m_freem(m);
1972 			return (IPPROTO_NONE);
1973 		}
1974 
1975 		/* verify the inner packet is destined to a mcast group */
1976 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1977 			++pim6stat.pim6s_rcv_badregisters;
1978 #ifdef MRT6DEBUG
1979 			if (V_mrt6debug & DEBUG_PIM)
1980 				log(LOG_DEBUG,
1981 				    "pim6_input: inner packet of register "
1982 				    "is not multicast %s\n",
1983 				    ip6_sprintf(ip6bufd, &eip6->ip6_dst));
1984 #endif
1985 			m_freem(m);
1986 			return (IPPROTO_DONE);
1987 		}
1988 
1989 		/*
1990 		 * make a copy of the whole header to pass to the daemon later.
1991 		 */
1992 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1993 		if (mcp == NULL) {
1994 #ifdef MRT6DEBUG
1995 			log(LOG_ERR,
1996 			    "pim6_input: pim register: "
1997 			    "could not copy register head\n");
1998 #endif
1999 			m_freem(m);
2000 			return (IPPROTO_DONE);
2001 		}
2002 
2003 		/*
2004 		 * forward the inner ip6 packet; point m_data at the inner ip6.
2005 		 */
2006 		m_adj(m, off + PIM_MINLEN);
2007 #ifdef MRT6DEBUG
2008 		if (V_mrt6debug & DEBUG_PIM) {
2009 			log(LOG_DEBUG,
2010 			    "pim6_input: forwarding decapsulated register: "
2011 			    "src %s, dst %s, mif %d\n",
2012 			    ip6_sprintf(ip6bufs, &eip6->ip6_src),
2013 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
2014 			    reg_mif_num);
2015 		}
2016 #endif
2017 
2018 		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
2019 				dst.sin6_family, 0);
2020 
2021 		/* prepare the register head to send to the mrouting daemon */
2022 		m = mcp;
2023 	}
2024 
2025 	/*
2026 	 * Pass the PIM message up to the daemon; if it is a register message
2027 	 * pass the 'head' only up to the daemon. This includes the
2028 	 * encapsulator ip6 header, pim header, register header and the
2029 	 * encapsulated ip6 header.
2030 	 */
2031   pim6_input_to_daemon:
2032 	rip6_input(&m, offp, proto);
2033 	return (IPPROTO_DONE);
2034 }
2035 
2036 static int
2037 ip6_mroute_modevent(module_t mod, int type, void *unused)
2038 {
2039 
2040 	switch (type) {
2041 	case MOD_LOAD:
2042 		MROUTER6_LOCK_INIT();
2043 		MFC6_LOCK_INIT();
2044 		MIF6_LOCK_INIT();
2045 
2046 		pim6_encap_cookie = encap_attach_func(AF_INET6, IPPROTO_PIM,
2047 			pim6_encapcheck,
2048 			(const struct protosw *)&in6_pim_protosw, NULL);
2049 		if (pim6_encap_cookie == NULL) {
2050 			printf("ip6_mroute: unable to attach pim6 encap\n");
2051 			MIF6_LOCK_DESTROY();
2052 			MFC6_LOCK_DESTROY();
2053 			MROUTER6_LOCK_DESTROY();
2054 			return (EINVAL);
2055 		}
2056 
2057 		ip6_mforward = X_ip6_mforward;
2058 		ip6_mrouter_done = X_ip6_mrouter_done;
2059 		ip6_mrouter_get = X_ip6_mrouter_get;
2060 		ip6_mrouter_set = X_ip6_mrouter_set;
2061 		mrt6_ioctl = X_mrt6_ioctl;
2062 		break;
2063 
2064 	case MOD_UNLOAD:
2065 		if (V_ip6_mrouter != NULL)
2066 			return EINVAL;
2067 
2068 		if (pim6_encap_cookie) {
2069 			encap_detach(pim6_encap_cookie);
2070 			pim6_encap_cookie = NULL;
2071 		}
2072 		X_ip6_mrouter_done();
2073 		ip6_mforward = NULL;
2074 		ip6_mrouter_done = NULL;
2075 		ip6_mrouter_get = NULL;
2076 		ip6_mrouter_set = NULL;
2077 		mrt6_ioctl = NULL;
2078 
2079 		MIF6_LOCK_DESTROY();
2080 		MFC6_LOCK_DESTROY();
2081 		MROUTER6_LOCK_DESTROY();
2082 		break;
2083 
2084 	default:
2085 		return (EOPNOTSUPP);
2086 	}
2087 
2088 	return (0);
2089 }
2090 
2091 static moduledata_t ip6_mroutemod = {
2092 	"ip6_mroute",
2093 	ip6_mroute_modevent,
2094 	0
2095 };
2096 
2097 DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
2098