xref: /freebsd/sys/netinet6/ip6_mroute.c (revision b1f9167f94059fd55c630891d359bcff987bd7eb)
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/ip6protosw.h>
125 #include <netinet6/pim6.h>
126 #include <netinet6/pim6_var.h>
127 
128 static MALLOC_DEFINE(M_MRTABLE6, "mf6c", "multicast forwarding cache entry");
129 
130 /* XXX: this is a very common idiom; move to <sys/mbuf.h> ? */
131 #define M_HASCL(m) ((m)->m_flags & M_EXT)
132 
133 static int	ip6_mdq(struct mbuf *, struct ifnet *, struct mf6c *);
134 static void	phyint_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
135 static int	register_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
136 static int	set_pim6(int *);
137 static int	socket_send(struct socket *, struct mbuf *,
138 		    struct sockaddr_in6 *);
139 
140 extern int in6_mcast_loop;
141 extern struct domain inet6domain;
142 
143 static const struct encaptab *pim6_encap_cookie;
144 static const struct ip6protosw in6_pim_protosw = {
145 	.pr_type =		SOCK_RAW,
146 	.pr_domain =		&inet6domain,
147 	.pr_protocol =		IPPROTO_PIM,
148 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
149 	.pr_input =		pim6_input,
150 	.pr_output =		rip6_output,
151 	.pr_ctloutput =		rip6_ctloutput,
152 	.pr_usrreqs =		&rip6_usrreqs
153 };
154 static int pim6_encapcheck(const struct mbuf *, int, int, void *);
155 
156 static VNET_DEFINE(int, ip6_mrouter_ver) = 0;
157 #define	V_ip6_mrouter_ver	VNET(ip6_mrouter_ver)
158 
159 SYSCTL_DECL(_net_inet6);
160 SYSCTL_DECL(_net_inet6_ip6);
161 static SYSCTL_NODE(_net_inet6, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
162 
163 static struct mrt6stat mrt6stat;
164 SYSCTL_STRUCT(_net_inet6_ip6, OID_AUTO, mrt6stat, CTLFLAG_RW,
165     &mrt6stat, mrt6stat,
166     "Multicast Routing Statistics (struct mrt6stat, netinet6/ip6_mroute.h)");
167 
168 #define	MRT6STAT_INC(name)	mrt6stat.name += 1
169 #define NO_RTE_FOUND	0x1
170 #define RTE_FOUND	0x2
171 
172 static struct mtx mrouter6_mtx;
173 #define	MROUTER6_LOCK()		mtx_lock(&mrouter6_mtx)
174 #define	MROUTER6_UNLOCK()	mtx_unlock(&mrouter6_mtx)
175 #define	MROUTER6_LOCK_ASSERT()	do {					\
176 	mtx_assert(&mrouter6_mtx, MA_OWNED);				\
177 	NET_ASSERT_GIANT();						\
178 } while (0)
179 #define	MROUTER6_LOCK_INIT()	\
180 	mtx_init(&mrouter6_mtx, "IPv6 multicast forwarding", NULL, MTX_DEF)
181 #define	MROUTER6_LOCK_DESTROY()	mtx_destroy(&mrouter6_mtx)
182 
183 static struct mf6c *mf6ctable[MF6CTBLSIZ];
184 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mf6ctable, CTLFLAG_RD,
185     &mf6ctable, sizeof(mf6ctable), "S,*mf6ctable[MF6CTBLSIZ]",
186     "IPv6 Multicast Forwarding Table (struct *mf6ctable[MF6CTBLSIZ], "
187     "netinet6/ip6_mroute.h)");
188 
189 static struct mtx mfc6_mtx;
190 #define	MFC6_LOCK()		mtx_lock(&mfc6_mtx)
191 #define	MFC6_UNLOCK()		mtx_unlock(&mfc6_mtx)
192 #define	MFC6_LOCK_ASSERT()	do {					\
193 	mtx_assert(&mfc6_mtx, MA_OWNED);				\
194 	NET_ASSERT_GIANT();						\
195 } while (0)
196 #define	MFC6_LOCK_INIT()		\
197 	mtx_init(&mfc6_mtx, "IPv6 multicast forwarding cache", NULL, MTX_DEF)
198 #define	MFC6_LOCK_DESTROY()	mtx_destroy(&mfc6_mtx)
199 
200 static u_char n6expire[MF6CTBLSIZ];
201 
202 static struct mif6 mif6table[MAXMIFS];
203 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mif6table, CTLFLAG_RD,
204     &mif6table, sizeof(mif6table), "S,mif6[MAXMIFS]",
205     "IPv6 Multicast Interfaces (struct mif6[MAXMIFS], netinet6/ip6_mroute.h)");
206 
207 static struct mtx mif6_mtx;
208 #define	MIF6_LOCK()		mtx_lock(&mif6_mtx)
209 #define	MIF6_UNLOCK()		mtx_unlock(&mif6_mtx)
210 #define	MIF6_LOCK_ASSERT()	mtx_assert(&mif6_mtx, MA_OWNED)
211 #define	MIF6_LOCK_INIT()	\
212 	mtx_init(&mif6_mtx, "IPv6 multicast interfaces", NULL, MTX_DEF)
213 #define	MIF6_LOCK_DESTROY()	mtx_destroy(&mif6_mtx)
214 
215 #ifdef MRT6DEBUG
216 static VNET_DEFINE(u_int, mrt6debug) = 0;	/* debug level */
217 #define	V_mrt6debug		VNET(mrt6debug)
218 #define DEBUG_MFC	0x02
219 #define DEBUG_FORWARD	0x04
220 #define DEBUG_EXPIRE	0x08
221 #define DEBUG_XMIT	0x10
222 #define DEBUG_REG	0x20
223 #define DEBUG_PIM	0x40
224 #define	DEBUG_ERR	0x80
225 #define	DEBUG_ANY	0x7f
226 #define	MRT6_DLOG(m, fmt, ...)	\
227 	if (V_mrt6debug & (m))	\
228 		log(((m) & DEBUG_ERR) ? LOG_ERR: LOG_DEBUG, \
229 		    "%s: " fmt "\n", __func__, ##__VA_ARGS__)
230 #else
231 #define	MRT6_DLOG(m, fmt, ...)
232 #endif
233 
234 static void	expire_upcalls(void *);
235 #define	EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
236 #define	UPCALL_EXPIRE	6		/* number of timeouts */
237 
238 /*
239  * XXX TODO: maintain a count to if_allmulti() calls in struct ifnet.
240  */
241 
242 /*
243  * 'Interfaces' associated with decapsulator (so we can tell
244  * packets that went through it from ones that get reflected
245  * by a broken gateway).  Different from IPv4 register_if,
246  * these interfaces are linked into the system ifnet list,
247  * because per-interface IPv6 statistics are maintained in
248  * ifp->if_afdata.  But it does not have any routes point
249  * to them.  I.e., packets can't be sent this way.  They
250  * only exist as a placeholder for multicast source
251  * verification.
252  */
253 static struct ifnet *multicast_register_if6;
254 
255 #define ENCAP_HOPS 64
256 
257 /*
258  * Private variables.
259  */
260 static mifi_t nummifs = 0;
261 static mifi_t reg_mif_num = (mifi_t)-1;
262 
263 static struct pim6stat pim6stat;
264 SYSCTL_STRUCT(_net_inet6_pim, PIM6CTL_STATS, stats, CTLFLAG_RW,
265     &pim6stat, pim6stat,
266     "PIM Statistics (struct pim6stat, netinet6/pim6_var.h)");
267 
268 #define	PIM6STAT_INC(name)	pim6stat.name += 1
269 static VNET_DEFINE(int, pim6);
270 #define	V_pim6		VNET(pim6)
271 
272 /*
273  * Hash function for a source, group entry
274  */
275 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
276 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
277 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
278 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
279 
280 /*
281  * Find a route for a given origin IPv6 address and Multicast group address.
282  */
283 #define MF6CFIND(o, g, rt) do { \
284 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
285 	rt = NULL; \
286 	while (_rt) { \
287 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
288 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
289 		    (_rt->mf6c_stall == NULL)) { \
290 			rt = _rt; \
291 			break; \
292 		} \
293 		_rt = _rt->mf6c_next; \
294 	} \
295 	if (rt == NULL) { \
296 		MRT6STAT_INC(mrt6s_mfc_misses); \
297 	} \
298 } while (/*CONSTCOND*/ 0)
299 
300 /*
301  * Macros to compute elapsed time efficiently
302  * Borrowed from Van Jacobson's scheduling code
303  * XXX: replace with timersub() ?
304  */
305 #define TV_DELTA(a, b, delta) do { \
306 	    int xxs; \
307 		\
308 	    delta = (a).tv_usec - (b).tv_usec; \
309 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
310 	       switch (xxs) { \
311 		      case 2: \
312 			  delta += 1000000; \
313 			      /* FALLTHROUGH */ \
314 		      case 1: \
315 			  delta += 1000000; \
316 			  break; \
317 		      default: \
318 			  delta += (1000000 * xxs); \
319 	       } \
320 	    } \
321 } while (/*CONSTCOND*/ 0)
322 
323 /* XXX: replace with timercmp(a, b, <) ? */
324 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
325 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
326 
327 #ifdef UPCALL_TIMING
328 #define UPCALL_MAX	50
329 static u_long upcall_data[UPCALL_MAX + 1];
330 static void collate();
331 #endif /* UPCALL_TIMING */
332 
333 static int ip6_mrouter_init(struct socket *, int, int);
334 static int add_m6fc(struct mf6cctl *);
335 static int add_m6if(struct mif6ctl *);
336 static int del_m6fc(struct mf6cctl *);
337 static int del_m6if(mifi_t *);
338 static int del_m6if_locked(mifi_t *);
339 static int get_mif6_cnt(struct sioc_mif_req6 *);
340 static int get_sg_cnt(struct sioc_sg_req6 *);
341 
342 static struct callout expire_upcalls_ch;
343 
344 int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *);
345 int X_ip6_mrouter_done(void);
346 int X_ip6_mrouter_set(struct socket *, struct sockopt *);
347 int X_ip6_mrouter_get(struct socket *, struct sockopt *);
348 int X_mrt6_ioctl(u_long, caddr_t);
349 
350 /*
351  * Handle MRT setsockopt commands to modify the multicast routing tables.
352  */
353 int
354 X_ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
355 {
356 	int error = 0;
357 	int optval;
358 	struct mif6ctl mifc;
359 	struct mf6cctl mfcc;
360 	mifi_t mifi;
361 
362 	if (so != V_ip6_mrouter && sopt->sopt_name != MRT6_INIT)
363 		return (EPERM);
364 
365 	switch (sopt->sopt_name) {
366 	case MRT6_INIT:
367 #ifdef MRT6_OINIT
368 	case MRT6_OINIT:
369 #endif
370 		error = sooptcopyin(sopt, &optval, sizeof(optval),
371 		    sizeof(optval));
372 		if (error)
373 			break;
374 		error = ip6_mrouter_init(so, optval, sopt->sopt_name);
375 		break;
376 	case MRT6_DONE:
377 		error = X_ip6_mrouter_done();
378 		break;
379 	case MRT6_ADD_MIF:
380 		error = sooptcopyin(sopt, &mifc, sizeof(mifc), sizeof(mifc));
381 		if (error)
382 			break;
383 		error = add_m6if(&mifc);
384 		break;
385 	case MRT6_ADD_MFC:
386 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
387 		if (error)
388 			break;
389 		error = add_m6fc(&mfcc);
390 		break;
391 	case MRT6_DEL_MFC:
392 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
393 		if (error)
394 			break;
395 		error = del_m6fc(&mfcc);
396 		break;
397 	case MRT6_DEL_MIF:
398 		error = sooptcopyin(sopt, &mifi, sizeof(mifi), sizeof(mifi));
399 		if (error)
400 			break;
401 		error = del_m6if(&mifi);
402 		break;
403 	case MRT6_PIM:
404 		error = sooptcopyin(sopt, &optval, sizeof(optval),
405 		    sizeof(optval));
406 		if (error)
407 			break;
408 		error = set_pim6(&optval);
409 		break;
410 	default:
411 		error = EOPNOTSUPP;
412 		break;
413 	}
414 
415 	return (error);
416 }
417 
418 /*
419  * Handle MRT getsockopt commands
420  */
421 int
422 X_ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
423 {
424 	int error = 0;
425 
426 	if (so != V_ip6_mrouter)
427 		return (EACCES);
428 
429 	switch (sopt->sopt_name) {
430 		case MRT6_PIM:
431 			error = sooptcopyout(sopt, &V_pim6, sizeof(V_pim6));
432 			break;
433 	}
434 	return (error);
435 }
436 
437 /*
438  * Handle ioctl commands to obtain information from the cache
439  */
440 int
441 X_mrt6_ioctl(u_long cmd, caddr_t data)
442 {
443 	int ret;
444 
445 	ret = EINVAL;
446 
447 	switch (cmd) {
448 	case SIOCGETSGCNT_IN6:
449 		ret = get_sg_cnt((struct sioc_sg_req6 *)data);
450 		break;
451 
452 	case SIOCGETMIFCNT_IN6:
453 		ret = get_mif6_cnt((struct sioc_mif_req6 *)data);
454 		break;
455 
456 	default:
457 		break;
458 	}
459 
460 	return (ret);
461 }
462 
463 /*
464  * returns the packet, byte, rpf-failure count for the source group provided
465  */
466 static int
467 get_sg_cnt(struct sioc_sg_req6 *req)
468 {
469 	struct mf6c *rt;
470 	int ret;
471 
472 	ret = 0;
473 
474 	MFC6_LOCK();
475 
476 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
477 	if (rt == NULL) {
478 		ret = ESRCH;
479 	} else {
480 		req->pktcnt = rt->mf6c_pkt_cnt;
481 		req->bytecnt = rt->mf6c_byte_cnt;
482 		req->wrong_if = rt->mf6c_wrong_if;
483 	}
484 
485 	MFC6_UNLOCK();
486 
487 	return (ret);
488 }
489 
490 /*
491  * returns the input and output packet and byte counts on the mif provided
492  */
493 static int
494 get_mif6_cnt(struct sioc_mif_req6 *req)
495 {
496 	mifi_t mifi;
497 	int ret;
498 
499 	ret = 0;
500 	mifi = req->mifi;
501 
502 	MIF6_LOCK();
503 
504 	if (mifi >= nummifs) {
505 		ret = EINVAL;
506 	} else {
507 		req->icount = mif6table[mifi].m6_pkt_in;
508 		req->ocount = mif6table[mifi].m6_pkt_out;
509 		req->ibytes = mif6table[mifi].m6_bytes_in;
510 		req->obytes = mif6table[mifi].m6_bytes_out;
511 	}
512 
513 	MIF6_UNLOCK();
514 
515 	return (ret);
516 }
517 
518 static int
519 set_pim6(int *i)
520 {
521 	if ((*i != 1) && (*i != 0))
522 		return (EINVAL);
523 
524 	V_pim6 = *i;
525 
526 	return (0);
527 }
528 
529 /*
530  * Enable multicast routing
531  */
532 static int
533 ip6_mrouter_init(struct socket *so, int v, int cmd)
534 {
535 
536 	MRT6_DLOG(DEBUG_ANY, "so_type = %d, pr_protocol = %d",
537 	    so->so_type, so->so_proto->pr_protocol);
538 
539 	if (so->so_type != SOCK_RAW ||
540 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
541 		return (EOPNOTSUPP);
542 
543 	if (v != 1)
544 		return (ENOPROTOOPT);
545 
546 	MROUTER6_LOCK();
547 
548 	if (V_ip6_mrouter != NULL) {
549 		MROUTER6_UNLOCK();
550 		return (EADDRINUSE);
551 	}
552 
553 	V_ip6_mrouter = so;
554 	V_ip6_mrouter_ver = cmd;
555 
556 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
557 	bzero((caddr_t)n6expire, sizeof(n6expire));
558 
559 	V_pim6 = 0;/* used for stubbing out/in pim stuff */
560 
561 	callout_init(&expire_upcalls_ch, 0);
562 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
563 	    expire_upcalls, NULL);
564 
565 	MROUTER6_UNLOCK();
566 	MRT6_DLOG(DEBUG_ANY, "finished");
567 
568 	return (0);
569 }
570 
571 /*
572  * Disable IPv6 multicast forwarding.
573  */
574 int
575 X_ip6_mrouter_done(void)
576 {
577 	mifi_t mifi;
578 	u_long i;
579 	struct mf6c *rt;
580 	struct rtdetq *rte;
581 
582 	MROUTER6_LOCK();
583 
584 	if (V_ip6_mrouter == NULL) {
585 		MROUTER6_UNLOCK();
586 		return (EINVAL);
587 	}
588 
589 	/*
590 	 * For each phyint in use, disable promiscuous reception of all IPv6
591 	 * multicasts.
592 	 */
593 	for (mifi = 0; mifi < nummifs; mifi++) {
594 		if (mif6table[mifi].m6_ifp &&
595 		    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
596 			if_allmulti(mif6table[mifi].m6_ifp, 0);
597 		}
598 	}
599 	bzero((caddr_t)mif6table, sizeof(mif6table));
600 	nummifs = 0;
601 
602 	V_pim6 = 0; /* used to stub out/in pim specific code */
603 
604 	callout_stop(&expire_upcalls_ch);
605 
606 	/*
607 	 * Free all multicast forwarding cache entries.
608 	 */
609 	MFC6_LOCK();
610 	for (i = 0; i < MF6CTBLSIZ; i++) {
611 		rt = mf6ctable[i];
612 		while (rt) {
613 			struct mf6c *frt;
614 
615 			for (rte = rt->mf6c_stall; rte != NULL; ) {
616 				struct rtdetq *n = rte->next;
617 
618 				m_freem(rte->m);
619 				free(rte, M_MRTABLE6);
620 				rte = n;
621 			}
622 			frt = rt;
623 			rt = rt->mf6c_next;
624 			free(frt, M_MRTABLE6);
625 		}
626 	}
627 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
628 	MFC6_UNLOCK();
629 
630 	/*
631 	 * Reset register interface
632 	 */
633 	if (reg_mif_num != (mifi_t)-1 && multicast_register_if6 != NULL) {
634 		if_detach(multicast_register_if6);
635 		if_free(multicast_register_if6);
636 		reg_mif_num = (mifi_t)-1;
637 		multicast_register_if6 = NULL;
638 	}
639 
640 	V_ip6_mrouter = NULL;
641 	V_ip6_mrouter_ver = 0;
642 
643 	MROUTER6_UNLOCK();
644 	MRT6_DLOG(DEBUG_ANY, "finished");
645 
646 	return (0);
647 }
648 
649 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
650 
651 /*
652  * Add a mif to the mif table
653  */
654 static int
655 add_m6if(struct mif6ctl *mifcp)
656 {
657 	struct mif6 *mifp;
658 	struct ifnet *ifp;
659 	int error;
660 
661 	MIF6_LOCK();
662 
663 	if (mifcp->mif6c_mifi >= MAXMIFS) {
664 		MIF6_UNLOCK();
665 		return (EINVAL);
666 	}
667 	mifp = mif6table + mifcp->mif6c_mifi;
668 	if (mifp->m6_ifp != NULL) {
669 		MIF6_UNLOCK();
670 		return (EADDRINUSE); /* XXX: is it appropriate? */
671 	}
672 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > V_if_index) {
673 		MIF6_UNLOCK();
674 		return (ENXIO);
675 	}
676 
677 	ifp = ifnet_byindex(mifcp->mif6c_pifi);
678 
679 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
680 		if (reg_mif_num == (mifi_t)-1) {
681 			ifp = if_alloc(IFT_OTHER);
682 
683 			if_initname(ifp, "register_mif", 0);
684 			ifp->if_flags |= IFF_LOOPBACK;
685 			if_attach(ifp);
686 			multicast_register_if6 = ifp;
687 			reg_mif_num = mifcp->mif6c_mifi;
688 			/*
689 			 * it is impossible to guess the ifindex of the
690 			 * register interface.  So mif6c_pifi is automatically
691 			 * calculated.
692 			 */
693 			mifcp->mif6c_pifi = ifp->if_index;
694 		} else {
695 			ifp = multicast_register_if6;
696 		}
697 	} else {
698 		/* Make sure the interface supports multicast */
699 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
700 			MIF6_UNLOCK();
701 			return (EOPNOTSUPP);
702 		}
703 
704 		error = if_allmulti(ifp, 1);
705 		if (error) {
706 			MIF6_UNLOCK();
707 			return (error);
708 		}
709 	}
710 
711 	mifp->m6_flags     = mifcp->mif6c_flags;
712 	mifp->m6_ifp       = ifp;
713 
714 	/* initialize per mif pkt counters */
715 	mifp->m6_pkt_in    = 0;
716 	mifp->m6_pkt_out   = 0;
717 	mifp->m6_bytes_in  = 0;
718 	mifp->m6_bytes_out = 0;
719 
720 	/* Adjust nummifs up if the mifi is higher than nummifs */
721 	if (nummifs <= mifcp->mif6c_mifi)
722 		nummifs = mifcp->mif6c_mifi + 1;
723 
724 	MIF6_UNLOCK();
725 	MRT6_DLOG(DEBUG_ANY, "mif #%d, phyint %s", mifcp->mif6c_mifi,
726 	    if_name(ifp));
727 
728 	return (0);
729 }
730 
731 /*
732  * Delete a mif from the mif table
733  */
734 static int
735 del_m6if_locked(mifi_t *mifip)
736 {
737 	struct mif6 *mifp = mif6table + *mifip;
738 	mifi_t mifi;
739 	struct ifnet *ifp;
740 
741 	MIF6_LOCK_ASSERT();
742 
743 	if (*mifip >= nummifs)
744 		return (EINVAL);
745 	if (mifp->m6_ifp == NULL)
746 		return (EINVAL);
747 
748 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
749 		/* XXX: TODO: Maintain an ALLMULTI refcount in struct ifnet. */
750 		ifp = mifp->m6_ifp;
751 		if_allmulti(ifp, 0);
752 	} else {
753 		if (reg_mif_num != (mifi_t)-1 &&
754 		    multicast_register_if6 != NULL) {
755 			if_detach(multicast_register_if6);
756 			if_free(multicast_register_if6);
757 			reg_mif_num = (mifi_t)-1;
758 			multicast_register_if6 = NULL;
759 		}
760 	}
761 
762 	bzero((caddr_t)mifp, sizeof(*mifp));
763 
764 	/* Adjust nummifs down */
765 	for (mifi = nummifs; mifi > 0; mifi--)
766 		if (mif6table[mifi - 1].m6_ifp)
767 			break;
768 	nummifs = mifi;
769 	MRT6_DLOG(DEBUG_ANY, "mif %d, nummifs %d", *mifip, nummifs);
770 
771 	return (0);
772 }
773 
774 static int
775 del_m6if(mifi_t *mifip)
776 {
777 	int cc;
778 
779 	MIF6_LOCK();
780 	cc = del_m6if_locked(mifip);
781 	MIF6_UNLOCK();
782 
783 	return (cc);
784 }
785 
786 /*
787  * Add an mfc entry
788  */
789 static int
790 add_m6fc(struct mf6cctl *mfccp)
791 {
792 	struct mf6c *rt;
793 	u_long hash;
794 	struct rtdetq *rte;
795 	u_short nstl;
796 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
797 
798 	MFC6_LOCK();
799 
800 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
801 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
802 
803 	/* If an entry already exists, just update the fields */
804 	if (rt) {
805 		MRT6_DLOG(DEBUG_MFC, "no upcall o %s g %s p %x",
806 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
807 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
808 		    mfccp->mf6cc_parent);
809 
810 		rt->mf6c_parent = mfccp->mf6cc_parent;
811 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
812 
813 		MFC6_UNLOCK();
814 		return (0);
815 	}
816 
817 	/*
818 	 * Find the entry for which the upcall was made and update
819 	 */
820 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
821 			mfccp->mf6cc_mcastgrp.sin6_addr);
822 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
823 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
824 				       &mfccp->mf6cc_origin.sin6_addr) &&
825 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
826 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
827 		    (rt->mf6c_stall != NULL)) {
828 
829 			if (nstl++)
830 				log(LOG_ERR,
831 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
832 				    "multiple kernel entries",
833 				    ip6_sprintf(ip6bufo,
834 					    &mfccp->mf6cc_origin.sin6_addr),
835 				    ip6_sprintf(ip6bufg,
836 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
837 				    mfccp->mf6cc_parent, rt->mf6c_stall);
838 
839 			MRT6_DLOG(DEBUG_MFC, "o %s g %s p %x dbg %p",
840 			    ip6_sprintf(ip6bufo,
841 			    &mfccp->mf6cc_origin.sin6_addr),
842 			    ip6_sprintf(ip6bufg,
843 				&mfccp->mf6cc_mcastgrp.sin6_addr),
844 			    mfccp->mf6cc_parent, rt->mf6c_stall);
845 
846 			rt->mf6c_origin     = mfccp->mf6cc_origin;
847 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
848 			rt->mf6c_parent     = mfccp->mf6cc_parent;
849 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
850 			/* initialize pkt counters per src-grp */
851 			rt->mf6c_pkt_cnt    = 0;
852 			rt->mf6c_byte_cnt   = 0;
853 			rt->mf6c_wrong_if   = 0;
854 
855 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
856 			n6expire[hash]--;
857 
858 			/* free packets Qed at the end of this entry */
859 			for (rte = rt->mf6c_stall; rte != NULL; ) {
860 				struct rtdetq *n = rte->next;
861 				ip6_mdq(rte->m, rte->ifp, rt);
862 				m_freem(rte->m);
863 #ifdef UPCALL_TIMING
864 				collate(&(rte->t));
865 #endif /* UPCALL_TIMING */
866 				free(rte, M_MRTABLE6);
867 				rte = n;
868 			}
869 			rt->mf6c_stall = NULL;
870 		}
871 	}
872 
873 	/*
874 	 * It is possible that an entry is being inserted without an upcall
875 	 */
876 	if (nstl == 0) {
877 		MRT6_DLOG(DEBUG_MFC, "no upcall h %lu o %s g %s p %x", hash,
878 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
879 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
880 		    mfccp->mf6cc_parent);
881 
882 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
883 
884 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
885 					       &mfccp->mf6cc_origin.sin6_addr)&&
886 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
887 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
888 
889 				rt->mf6c_origin     = mfccp->mf6cc_origin;
890 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
891 				rt->mf6c_parent     = mfccp->mf6cc_parent;
892 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
893 				/* initialize pkt counters per src-grp */
894 				rt->mf6c_pkt_cnt    = 0;
895 				rt->mf6c_byte_cnt   = 0;
896 				rt->mf6c_wrong_if   = 0;
897 
898 				if (rt->mf6c_expire)
899 					n6expire[hash]--;
900 				rt->mf6c_expire	   = 0;
901 			}
902 		}
903 		if (rt == NULL) {
904 			/* no upcall, so make a new entry */
905 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
906 						  M_NOWAIT);
907 			if (rt == NULL) {
908 				MFC6_UNLOCK();
909 				return (ENOBUFS);
910 			}
911 
912 			/* insert new entry at head of hash chain */
913 			rt->mf6c_origin     = mfccp->mf6cc_origin;
914 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
915 			rt->mf6c_parent     = mfccp->mf6cc_parent;
916 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
917 			/* initialize pkt counters per src-grp */
918 			rt->mf6c_pkt_cnt    = 0;
919 			rt->mf6c_byte_cnt   = 0;
920 			rt->mf6c_wrong_if   = 0;
921 			rt->mf6c_expire     = 0;
922 			rt->mf6c_stall = NULL;
923 
924 			/* link into table */
925 			rt->mf6c_next  = mf6ctable[hash];
926 			mf6ctable[hash] = rt;
927 		}
928 	}
929 
930 	MFC6_UNLOCK();
931 	return (0);
932 }
933 
934 #ifdef UPCALL_TIMING
935 /*
936  * collect delay statistics on the upcalls
937  */
938 static void
939 collate(struct timeval *t)
940 {
941 	u_long d;
942 	struct timeval tp;
943 	u_long delta;
944 
945 	GET_TIME(tp);
946 
947 	if (TV_LT(*t, tp))
948 	{
949 		TV_DELTA(tp, *t, delta);
950 
951 		d = delta >> 10;
952 		if (d > UPCALL_MAX)
953 			d = UPCALL_MAX;
954 
955 		++upcall_data[d];
956 	}
957 }
958 #endif /* UPCALL_TIMING */
959 
960 /*
961  * Delete an mfc entry
962  */
963 static int
964 del_m6fc(struct mf6cctl *mfccp)
965 {
966 #ifdef MRT6DEBUG
967 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
968 #endif
969 	struct sockaddr_in6	origin;
970 	struct sockaddr_in6	mcastgrp;
971 	struct mf6c		*rt;
972 	struct mf6c		**nptr;
973 	u_long		hash;
974 
975 	origin = mfccp->mf6cc_origin;
976 	mcastgrp = mfccp->mf6cc_mcastgrp;
977 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
978 
979 	MRT6_DLOG(DEBUG_MFC, "orig %s mcastgrp %s",
980 	    ip6_sprintf(ip6bufo, &origin.sin6_addr),
981 	    ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr));
982 
983 	MFC6_LOCK();
984 
985 	nptr = &mf6ctable[hash];
986 	while ((rt = *nptr) != NULL) {
987 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
988 				       &rt->mf6c_origin.sin6_addr) &&
989 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
990 				       &rt->mf6c_mcastgrp.sin6_addr) &&
991 		    rt->mf6c_stall == NULL)
992 			break;
993 
994 		nptr = &rt->mf6c_next;
995 	}
996 	if (rt == NULL) {
997 		MFC6_UNLOCK();
998 		return (EADDRNOTAVAIL);
999 	}
1000 
1001 	*nptr = rt->mf6c_next;
1002 	free(rt, M_MRTABLE6);
1003 
1004 	MFC6_UNLOCK();
1005 
1006 	return (0);
1007 }
1008 
1009 static int
1010 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
1011 {
1012 
1013 	if (s) {
1014 		if (sbappendaddr(&s->so_rcv,
1015 				 (struct sockaddr *)src,
1016 				 mm, (struct mbuf *)0) != 0) {
1017 			sorwakeup(s);
1018 			return (0);
1019 		}
1020 	}
1021 	m_freem(mm);
1022 	return (-1);
1023 }
1024 
1025 /*
1026  * IPv6 multicast forwarding function. This function assumes that the packet
1027  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
1028  * pointed to by "ifp", and the packet is to be relayed to other networks
1029  * that have members of the packet's destination IPv6 multicast group.
1030  *
1031  * The packet is returned unscathed to the caller, unless it is
1032  * erroneous, in which case a non-zero return value tells the caller to
1033  * discard it.
1034  *
1035  * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff
1036  * this function is called in the originating context (i.e., not when
1037  * forwarding a packet from other node).  ip6_output(), which is currently the
1038  * only function that calls this function is called in the originating context,
1039  * explicitly ensures this condition.  It is caller's responsibility to ensure
1040  * that if this function is called from somewhere else in the originating
1041  * context in the future.
1042  */
1043 int
1044 X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
1045 {
1046 	struct rtdetq *rte;
1047 	struct mbuf *mb0;
1048 	struct mf6c *rt;
1049 	struct mif6 *mifp;
1050 	struct mbuf *mm;
1051 	u_long hash;
1052 	mifi_t mifi;
1053 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1054 #ifdef UPCALL_TIMING
1055 	struct timeval tp;
1056 
1057 	GET_TIME(tp);
1058 #endif /* UPCALL_TIMING */
1059 
1060 	MRT6_DLOG(DEBUG_FORWARD, "src %s, dst %s, ifindex %d",
1061 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1062 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst), ifp->if_index);
1063 
1064 	/*
1065 	 * Don't forward a packet with Hop limit of zero or one,
1066 	 * or a packet destined to a local-only group.
1067 	 */
1068 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
1069 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1070 		return (0);
1071 	ip6->ip6_hlim--;
1072 
1073 	/*
1074 	 * Source address check: do not forward packets with unspecified
1075 	 * source. It was discussed in July 2000, on ipngwg mailing list.
1076 	 * This is rather more serious than unicast cases, because some
1077 	 * MLD packets can be sent with the unspecified source address
1078 	 * (although such packets must normally set 1 to the hop limit field).
1079 	 */
1080 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1081 		IP6STAT_INC(ip6s_cantforward);
1082 		if (V_ip6_log_time + V_ip6_log_interval < time_uptime) {
1083 			V_ip6_log_time = time_uptime;
1084 			log(LOG_DEBUG,
1085 			    "cannot forward "
1086 			    "from %s to %s nxt %d received on %s\n",
1087 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1088 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1089 			    ip6->ip6_nxt,
1090 			    if_name(m->m_pkthdr.rcvif));
1091 		}
1092 		return (0);
1093 	}
1094 
1095 	MFC6_LOCK();
1096 
1097 	/*
1098 	 * Determine forwarding mifs from the forwarding cache table
1099 	 */
1100 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
1101 	MRT6STAT_INC(mrt6s_mfc_lookups);
1102 
1103 	/* Entry exists, so forward if necessary */
1104 	if (rt) {
1105 		MFC6_UNLOCK();
1106 		return (ip6_mdq(m, ifp, rt));
1107 	}
1108 
1109 	/*
1110 	 * If we don't have a route for packet's origin,
1111 	 * Make a copy of the packet & send message to routing daemon.
1112 	 */
1113 	MRT6STAT_INC(mrt6s_no_route);
1114 	MRT6_DLOG(DEBUG_FORWARD | DEBUG_MFC, "no rte s %s g %s",
1115 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1116 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1117 
1118 	/*
1119 	 * Allocate mbufs early so that we don't do extra work if we
1120 	 * are just going to fail anyway.
1121 	 */
1122 	rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE6, M_NOWAIT);
1123 	if (rte == NULL) {
1124 		MFC6_UNLOCK();
1125 		return (ENOBUFS);
1126 	}
1127 	mb0 = m_copy(m, 0, M_COPYALL);
1128 	/*
1129 	 * Pullup packet header if needed before storing it,
1130 	 * as other references may modify it in the meantime.
1131 	 */
1132 	if (mb0 && (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1133 		mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1134 	if (mb0 == NULL) {
1135 		free(rte, M_MRTABLE6);
1136 		MFC6_UNLOCK();
1137 		return (ENOBUFS);
1138 	}
1139 
1140 	/* is there an upcall waiting for this packet? */
1141 	hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1142 	for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1143 		if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1144 		    &rt->mf6c_origin.sin6_addr) &&
1145 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1146 		    &rt->mf6c_mcastgrp.sin6_addr) && (rt->mf6c_stall != NULL))
1147 			break;
1148 	}
1149 
1150 	if (rt == NULL) {
1151 		struct mrt6msg *im;
1152 #ifdef MRT6_OINIT
1153 		struct omrt6msg *oim;
1154 #endif
1155 		/* no upcall, so make a new entry */
1156 		rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6, M_NOWAIT);
1157 		if (rt == NULL) {
1158 			free(rte, M_MRTABLE6);
1159 			m_freem(mb0);
1160 			MFC6_UNLOCK();
1161 			return (ENOBUFS);
1162 		}
1163 		/*
1164 		 * Make a copy of the header to send to the user
1165 		 * level process
1166 		 */
1167 		mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1168 		if (mm == NULL) {
1169 			free(rte, M_MRTABLE6);
1170 			m_freem(mb0);
1171 			free(rt, M_MRTABLE6);
1172 			MFC6_UNLOCK();
1173 			return (ENOBUFS);
1174 		}
1175 
1176 		/*
1177 		 * Send message to routing daemon
1178 		 */
1179 		sin6.sin6_addr = ip6->ip6_src;
1180 		im = NULL;
1181 #ifdef MRT6_OINIT
1182 		oim = NULL;
1183 #endif
1184 		switch (V_ip6_mrouter_ver) {
1185 #ifdef MRT6_OINIT
1186 		case MRT6_OINIT:
1187 			oim = mtod(mm, struct omrt6msg *);
1188 			oim->im6_msgtype = MRT6MSG_NOCACHE;
1189 			oim->im6_mbz = 0;
1190 			break;
1191 #endif
1192 		case MRT6_INIT:
1193 			im = mtod(mm, struct mrt6msg *);
1194 			im->im6_msgtype = MRT6MSG_NOCACHE;
1195 			im->im6_mbz = 0;
1196 			break;
1197 		default:
1198 			free(rte, M_MRTABLE6);
1199 			m_freem(mb0);
1200 			free(rt, M_MRTABLE6);
1201 			MFC6_UNLOCK();
1202 			return (EINVAL);
1203 		}
1204 
1205 		MRT6_DLOG(DEBUG_FORWARD, "getting the iif info in the kernel");
1206 		for (mifp = mif6table, mifi = 0;
1207 		    mifi < nummifs && mifp->m6_ifp != ifp; mifp++, mifi++)
1208 				;
1209 
1210 		switch (V_ip6_mrouter_ver) {
1211 #ifdef MRT6_OINIT
1212 		case MRT6_OINIT:
1213 			oim->im6_mif = mifi;
1214 			break;
1215 #endif
1216 		case MRT6_INIT:
1217 			im->im6_mif = mifi;
1218 			break;
1219 		}
1220 
1221 		if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1222 			log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1223 			    "socket queue full\n");
1224 			MRT6STAT_INC(mrt6s_upq_sockfull);
1225 			free(rte, M_MRTABLE6);
1226 			m_freem(mb0);
1227 			free(rt, M_MRTABLE6);
1228 			MFC6_UNLOCK();
1229 			return (ENOBUFS);
1230 		}
1231 
1232 		MRT6STAT_INC(mrt6s_upcalls);
1233 
1234 		/* insert new entry at head of hash chain */
1235 		bzero(rt, sizeof(*rt));
1236 		rt->mf6c_origin.sin6_family = AF_INET6;
1237 		rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1238 		rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1239 		rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1240 		rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1241 		rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1242 		rt->mf6c_expire = UPCALL_EXPIRE;
1243 		n6expire[hash]++;
1244 		rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1245 
1246 		/* link into table */
1247 		rt->mf6c_next  = mf6ctable[hash];
1248 		mf6ctable[hash] = rt;
1249 		/* Add this entry to the end of the queue */
1250 		rt->mf6c_stall = rte;
1251 	} else {
1252 		/* determine if q has overflowed */
1253 		struct rtdetq **p;
1254 		int npkts = 0;
1255 
1256 		for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1257 			if (++npkts > MAX_UPQ6) {
1258 				MRT6STAT_INC(mrt6s_upq_ovflw);
1259 				free(rte, M_MRTABLE6);
1260 				m_freem(mb0);
1261 				MFC6_UNLOCK();
1262 				return (0);
1263 			}
1264 
1265 		/* Add this entry to the end of the queue */
1266 		*p = rte;
1267 	}
1268 
1269 	rte->next = NULL;
1270 	rte->m = mb0;
1271 	rte->ifp = ifp;
1272 #ifdef UPCALL_TIMING
1273 	rte->t = tp;
1274 #endif /* UPCALL_TIMING */
1275 
1276 	MFC6_UNLOCK();
1277 
1278 	return (0);
1279 }
1280 
1281 /*
1282  * Clean up cache entries if upcalls are not serviced
1283  * Call from the Slow Timeout mechanism, every half second.
1284  */
1285 static void
1286 expire_upcalls(void *unused)
1287 {
1288 #ifdef MRT6DEBUG
1289 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1290 #endif
1291 	struct rtdetq *rte;
1292 	struct mf6c *mfc, **nptr;
1293 	u_long i;
1294 
1295 	MFC6_LOCK();
1296 	for (i = 0; i < MF6CTBLSIZ; i++) {
1297 		if (n6expire[i] == 0)
1298 			continue;
1299 		nptr = &mf6ctable[i];
1300 		while ((mfc = *nptr) != NULL) {
1301 			rte = mfc->mf6c_stall;
1302 			/*
1303 			 * Skip real cache entries
1304 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1305 			 * If it expires now
1306 			 */
1307 			if (rte != NULL &&
1308 			    mfc->mf6c_expire != 0 &&
1309 			    --mfc->mf6c_expire == 0) {
1310 				MRT6_DLOG(DEBUG_EXPIRE, "expiring (%s %s)",
1311 				    ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr),
1312 				    ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr));
1313 				/*
1314 				 * drop all the packets
1315 				 * free the mbuf with the pkt, if, timing info
1316 				 */
1317 				do {
1318 					struct rtdetq *n = rte->next;
1319 					m_freem(rte->m);
1320 					free(rte, M_MRTABLE6);
1321 					rte = n;
1322 				} while (rte != NULL);
1323 				MRT6STAT_INC(mrt6s_cache_cleanups);
1324 				n6expire[i]--;
1325 
1326 				*nptr = mfc->mf6c_next;
1327 				free(mfc, M_MRTABLE6);
1328 			} else {
1329 				nptr = &mfc->mf6c_next;
1330 			}
1331 		}
1332 	}
1333 	MFC6_UNLOCK();
1334 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1335 	    expire_upcalls, NULL);
1336 }
1337 
1338 /*
1339  * Packet forwarding routine once entry in the cache is made
1340  */
1341 static int
1342 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1343 {
1344 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1345 	mifi_t mifi, iif;
1346 	struct mif6 *mifp;
1347 	int plen = m->m_pkthdr.len;
1348 	struct in6_addr src0, dst0; /* copies for local work */
1349 	u_int32_t iszone, idzone, oszone, odzone;
1350 	int error = 0;
1351 
1352 /*
1353  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1354  * input, they shouldn't get counted on output, so statistics keeping is
1355  * separate.
1356  */
1357 
1358 #define MC6_SEND(ip6, mifp, m) do {				\
1359 	if ((mifp)->m6_flags & MIFF_REGISTER)			\
1360 		register_send((ip6), (mifp), (m));		\
1361 	else							\
1362 		phyint_send((ip6), (mifp), (m));		\
1363 } while (/*CONSTCOND*/ 0)
1364 
1365 	/*
1366 	 * Don't forward if it didn't arrive from the parent mif
1367 	 * for its origin.
1368 	 */
1369 	mifi = rt->mf6c_parent;
1370 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1371 		/* came in the wrong interface */
1372 		MRT6_DLOG(DEBUG_FORWARD,
1373 		    "wrong if: ifid %d mifi %d mififid %x", ifp->if_index,
1374 		    mifi, mif6table[mifi].m6_ifp->if_index);
1375 		MRT6STAT_INC(mrt6s_wrong_if);
1376 		rt->mf6c_wrong_if++;
1377 		/*
1378 		 * If we are doing PIM processing, and we are forwarding
1379 		 * packets on this interface, send a message to the
1380 		 * routing daemon.
1381 		 */
1382 		/* have to make sure this is a valid mif */
1383 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1384 			if (V_pim6 && (m->m_flags & M_LOOP) == 0) {
1385 				/*
1386 				 * Check the M_LOOP flag to avoid an
1387 				 * unnecessary PIM assert.
1388 				 * XXX: M_LOOP is an ad-hoc hack...
1389 				 */
1390 				static struct sockaddr_in6 sin6 =
1391 				{ sizeof(sin6), AF_INET6 };
1392 
1393 				struct mbuf *mm;
1394 				struct mrt6msg *im;
1395 #ifdef MRT6_OINIT
1396 				struct omrt6msg *oim;
1397 #endif
1398 
1399 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1400 				if (mm &&
1401 				    (M_HASCL(mm) ||
1402 				     mm->m_len < sizeof(struct ip6_hdr)))
1403 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1404 				if (mm == NULL)
1405 					return (ENOBUFS);
1406 
1407 #ifdef MRT6_OINIT
1408 				oim = NULL;
1409 #endif
1410 				im = NULL;
1411 				switch (V_ip6_mrouter_ver) {
1412 #ifdef MRT6_OINIT
1413 				case MRT6_OINIT:
1414 					oim = mtod(mm, struct omrt6msg *);
1415 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1416 					oim->im6_mbz = 0;
1417 					break;
1418 #endif
1419 				case MRT6_INIT:
1420 					im = mtod(mm, struct mrt6msg *);
1421 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1422 					im->im6_mbz = 0;
1423 					break;
1424 				default:
1425 					m_freem(mm);
1426 					return (EINVAL);
1427 				}
1428 
1429 				for (mifp = mif6table, iif = 0;
1430 				     iif < nummifs && mifp &&
1431 					     mifp->m6_ifp != ifp;
1432 				     mifp++, iif++)
1433 					;
1434 
1435 				switch (V_ip6_mrouter_ver) {
1436 #ifdef MRT6_OINIT
1437 				case MRT6_OINIT:
1438 					oim->im6_mif = iif;
1439 					sin6.sin6_addr = oim->im6_src;
1440 					break;
1441 #endif
1442 				case MRT6_INIT:
1443 					im->im6_mif = iif;
1444 					sin6.sin6_addr = im->im6_src;
1445 					break;
1446 				}
1447 
1448 				MRT6STAT_INC(mrt6s_upcalls);
1449 
1450 				if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1451 					MRT6_DLOG(DEBUG_ANY,
1452 					    "ip6_mrouter socket queue full");
1453 					MRT6STAT_INC(mrt6s_upq_sockfull);
1454 					return (ENOBUFS);
1455 				}	/* if socket Q full */
1456 			}		/* if PIM */
1457 		return (0);
1458 	}			/* if wrong iif */
1459 
1460 	/* If I sourced this packet, it counts as output, else it was input. */
1461 	if (m->m_pkthdr.rcvif == NULL) {
1462 		/* XXX: is rcvif really NULL when output?? */
1463 		mif6table[mifi].m6_pkt_out++;
1464 		mif6table[mifi].m6_bytes_out += plen;
1465 	} else {
1466 		mif6table[mifi].m6_pkt_in++;
1467 		mif6table[mifi].m6_bytes_in += plen;
1468 	}
1469 	rt->mf6c_pkt_cnt++;
1470 	rt->mf6c_byte_cnt += plen;
1471 
1472 	/*
1473 	 * For each mif, forward a copy of the packet if there are group
1474 	 * members downstream on the interface.
1475 	 */
1476 	src0 = ip6->ip6_src;
1477 	dst0 = ip6->ip6_dst;
1478 	if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 ||
1479 	    (error = in6_setscope(&dst0, ifp, &idzone)) != 0) {
1480 		IP6STAT_INC(ip6s_badscope);
1481 		return (error);
1482 	}
1483 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) {
1484 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1485 			/*
1486 			 * check if the outgoing packet is going to break
1487 			 * a scope boundary.
1488 			 * XXX For packets through PIM register tunnel
1489 			 * interface, we believe a routing daemon.
1490 			 */
1491 			if (!(mif6table[rt->mf6c_parent].m6_flags &
1492 			      MIFF_REGISTER) &&
1493 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
1494 				if (in6_setscope(&src0, mif6table[mifi].m6_ifp,
1495 				    &oszone) ||
1496 				    in6_setscope(&dst0, mif6table[mifi].m6_ifp,
1497 				    &odzone) ||
1498 				    iszone != oszone ||
1499 				    idzone != odzone) {
1500 					IP6STAT_INC(ip6s_badscope);
1501 					continue;
1502 				}
1503 			}
1504 
1505 			mifp->m6_pkt_out++;
1506 			mifp->m6_bytes_out += plen;
1507 			MC6_SEND(ip6, mifp, m);
1508 		}
1509 	}
1510 	return (0);
1511 }
1512 
1513 static void
1514 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1515 {
1516 #ifdef MRT6DEBUG
1517 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1518 #endif
1519 	struct mbuf *mb_copy;
1520 	struct ifnet *ifp = mifp->m6_ifp;
1521 	int error = 0;
1522 	u_long linkmtu;
1523 
1524 	/*
1525 	 * Make a new reference to the packet; make sure that
1526 	 * the IPv6 header is actually copied, not just referenced,
1527 	 * so that ip6_output() only scribbles on the copy.
1528 	 */
1529 	mb_copy = m_copy(m, 0, M_COPYALL);
1530 	if (mb_copy &&
1531 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1532 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1533 	if (mb_copy == NULL) {
1534 		return;
1535 	}
1536 	/* set MCAST flag to the outgoing packet */
1537 	mb_copy->m_flags |= M_MCAST;
1538 
1539 	/*
1540 	 * If we sourced the packet, call ip6_output since we may devide
1541 	 * the packet into fragments when the packet is too big for the
1542 	 * outgoing interface.
1543 	 * Otherwise, we can simply send the packet to the interface
1544 	 * sending queue.
1545 	 */
1546 	if (m->m_pkthdr.rcvif == NULL) {
1547 		struct ip6_moptions im6o;
1548 
1549 		im6o.im6o_multicast_ifp = ifp;
1550 		/* XXX: ip6_output will override ip6->ip6_hlim */
1551 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1552 		im6o.im6o_multicast_loop = 1;
1553 		error = ip6_output(mb_copy, NULL, NULL, IPV6_FORWARDING, &im6o,
1554 		    NULL, NULL);
1555 
1556 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1557 		    (uint16_t)(mifp - mif6table), error);
1558 		return;
1559 	}
1560 
1561 	/*
1562 	 * If configured to loop back multicasts by default,
1563 	 * loop back a copy now.
1564 	 */
1565 	if (in6_mcast_loop) {
1566 		struct sockaddr_in6 dst6;
1567 
1568 		bzero(&dst6, sizeof(dst6));
1569 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1570 		dst6.sin6_family = AF_INET6;
1571 		dst6.sin6_addr = ip6->ip6_dst;
1572 		ip6_mloopback(ifp, m, &dst6);
1573 	}
1574 
1575 	/*
1576 	 * Put the packet into the sending queue of the outgoing interface
1577 	 * if it would fit in the MTU of the interface.
1578 	 */
1579 	linkmtu = IN6_LINKMTU(ifp);
1580 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1581 		struct sockaddr_in6 dst6;
1582 
1583 		bzero(&dst6, sizeof(dst6));
1584 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1585 		dst6.sin6_family = AF_INET6;
1586 		dst6.sin6_addr = ip6->ip6_dst;
1587 
1588 		IP_PROBE(send, NULL, NULL, ip6, ifp, NULL, ip6);
1589 		/*
1590 		 * We just call if_output instead of nd6_output here, since
1591 		 * we need no ND for a multicast forwarded packet...right?
1592 		 */
1593 		m_clrprotoflags(m);	/* Avoid confusing lower layers. */
1594 		error = (*ifp->if_output)(ifp, mb_copy,
1595 		    (struct sockaddr *)&dst6, NULL);
1596 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1597 		    (uint16_t)(mifp - mif6table), error);
1598 	} else {
1599 		/*
1600 		 * pMTU discovery is intentionally disabled by default, since
1601 		 * various router may notify pMTU in multicast, which can be
1602 		 * a DDoS to a router
1603 		 */
1604 		if (V_ip6_mcast_pmtu)
1605 			icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1606 		else {
1607 			MRT6_DLOG(DEBUG_XMIT, " packet too big on %s o %s "
1608 			    "g %s size %d (discarded)", if_name(ifp),
1609 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1610 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1611 			    mb_copy->m_pkthdr.len);
1612 			m_freem(mb_copy); /* simply discard the packet */
1613 		}
1614 	}
1615 }
1616 
1617 static int
1618 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1619 {
1620 #ifdef MRT6DEBUG
1621 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1622 #endif
1623 	struct mbuf *mm;
1624 	int i, len = m->m_pkthdr.len;
1625 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1626 	struct mrt6msg *im6;
1627 
1628 	MRT6_DLOG(DEBUG_ANY, "src %s dst %s",
1629 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1630 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1631 	PIM6STAT_INC(pim6s_snd_registers);
1632 
1633 	/* Make a copy of the packet to send to the user level process. */
1634 	mm = m_gethdr(M_NOWAIT, MT_DATA);
1635 	if (mm == NULL)
1636 		return (ENOBUFS);
1637 	mm->m_data += max_linkhdr;
1638 	mm->m_len = sizeof(struct ip6_hdr);
1639 
1640 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1641 		m_freem(mm);
1642 		return (ENOBUFS);
1643 	}
1644 	i = MHLEN - M_LEADINGSPACE(mm);
1645 	if (i > len)
1646 		i = len;
1647 	mm = m_pullup(mm, i);
1648 	if (mm == NULL)
1649 		return (ENOBUFS);
1650 /* TODO: check it! */
1651 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1652 
1653 	/*
1654 	 * Send message to routing daemon
1655 	 */
1656 	sin6.sin6_addr = ip6->ip6_src;
1657 
1658 	im6 = mtod(mm, struct mrt6msg *);
1659 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1660 	im6->im6_mbz          = 0;
1661 
1662 	im6->im6_mif = mif - mif6table;
1663 
1664 	/* iif info is not given for reg. encap.n */
1665 	MRT6STAT_INC(mrt6s_upcalls);
1666 
1667 	if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1668 		MRT6_DLOG(DEBUG_ANY, "ip6_mrouter socket queue full");
1669 		MRT6STAT_INC(mrt6s_upq_sockfull);
1670 		return (ENOBUFS);
1671 	}
1672 	return (0);
1673 }
1674 
1675 /*
1676  * pim6_encapcheck() is called by the encap6_input() path at runtime to
1677  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
1678  * into the kernel.
1679  */
1680 static int
1681 pim6_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
1682 {
1683 
1684 #ifdef DIAGNOSTIC
1685     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
1686 #endif
1687     if (proto != IPPROTO_PIM)
1688 	return 0;	/* not for us; reject the datagram. */
1689 
1690     return 64;		/* claim the datagram. */
1691 }
1692 
1693 /*
1694  * PIM sparse mode hook
1695  * Receives the pim control messages, and passes them up to the listening
1696  * socket, using rip6_input.
1697  * The only message processed is the REGISTER pim message; the pim header
1698  * is stripped off, and the inner packet is passed to register_mforward.
1699  */
1700 int
1701 pim6_input(struct mbuf **mp, int *offp, int proto)
1702 {
1703 	struct pim *pim; /* pointer to a pim struct */
1704 	struct ip6_hdr *ip6;
1705 	int pimlen;
1706 	struct mbuf *m = *mp;
1707 	int minlen;
1708 	int off = *offp;
1709 
1710 	PIM6STAT_INC(pim6s_rcv_total);
1711 
1712 	ip6 = mtod(m, struct ip6_hdr *);
1713 	pimlen = m->m_pkthdr.len - *offp;
1714 
1715 	/*
1716 	 * Validate lengths
1717 	 */
1718 	if (pimlen < PIM_MINLEN) {
1719 		PIM6STAT_INC(pim6s_rcv_tooshort);
1720 		MRT6_DLOG(DEBUG_PIM, "PIM packet too short");
1721 		m_freem(m);
1722 		return (IPPROTO_DONE);
1723 	}
1724 
1725 	/*
1726 	 * if the packet is at least as big as a REGISTER, go ahead
1727 	 * and grab the PIM REGISTER header size, to avoid another
1728 	 * possible m_pullup() later.
1729 	 *
1730 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1731 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1732 	 */
1733 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1734 
1735 	/*
1736 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1737 	 * possibly the PIM REGISTER header
1738 	 */
1739 #ifndef PULLDOWN_TEST
1740 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1741 	/* adjust pointer */
1742 	ip6 = mtod(m, struct ip6_hdr *);
1743 
1744 	/* adjust mbuf to point to the PIM header */
1745 	pim = (struct pim *)((caddr_t)ip6 + off);
1746 #else
1747 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1748 	if (pim == NULL) {
1749 		PIM6STAT_INC(pim6s_rcv_tooshort);
1750 		return (IPPROTO_DONE);
1751 	}
1752 #endif
1753 
1754 #define PIM6_CHECKSUM
1755 #ifdef PIM6_CHECKSUM
1756 	{
1757 		int cksumlen;
1758 
1759 		/*
1760 		 * Validate checksum.
1761 		 * If PIM REGISTER, exclude the data packet
1762 		 */
1763 		if (pim->pim_type == PIM_REGISTER)
1764 			cksumlen = PIM_MINLEN;
1765 		else
1766 			cksumlen = pimlen;
1767 
1768 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1769 			PIM6STAT_INC(pim6s_rcv_badsum);
1770 			MRT6_DLOG(DEBUG_PIM, "invalid checksum");
1771 			m_freem(m);
1772 			return (IPPROTO_DONE);
1773 		}
1774 	}
1775 #endif /* PIM_CHECKSUM */
1776 
1777 	/* PIM version check */
1778 	if (pim->pim_ver != PIM_VERSION) {
1779 		PIM6STAT_INC(pim6s_rcv_badversion);
1780 		MRT6_DLOG(DEBUG_ANY | DEBUG_ERR,
1781 		    "incorrect version %d, expecting %d",
1782 		    pim->pim_ver, PIM_VERSION);
1783 		m_freem(m);
1784 		return (IPPROTO_DONE);
1785 	}
1786 
1787 	if (pim->pim_type == PIM_REGISTER) {
1788 		/*
1789 		 * since this is a REGISTER, we'll make a copy of the register
1790 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1791 		 * routing daemon.
1792 		 */
1793 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1794 
1795 		struct mbuf *mcp;
1796 		struct ip6_hdr *eip6;
1797 		u_int32_t *reghdr;
1798 		int rc;
1799 #ifdef MRT6DEBUG
1800 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1801 #endif
1802 
1803 		PIM6STAT_INC(pim6s_rcv_registers);
1804 
1805 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1806 			MRT6_DLOG(DEBUG_PIM, "register mif not set: %d",
1807 			    reg_mif_num);
1808 			m_freem(m);
1809 			return (IPPROTO_DONE);
1810 		}
1811 
1812 		reghdr = (u_int32_t *)(pim + 1);
1813 
1814 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1815 			goto pim6_input_to_daemon;
1816 
1817 		/*
1818 		 * Validate length
1819 		 */
1820 		if (pimlen < PIM6_REG_MINLEN) {
1821 			PIM6STAT_INC(pim6s_rcv_tooshort);
1822 			PIM6STAT_INC(pim6s_rcv_badregisters);
1823 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "register packet "
1824 			    "size too small %d from %s",
1825 			    pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src));
1826 			m_freem(m);
1827 			return (IPPROTO_DONE);
1828 		}
1829 
1830 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1831 		MRT6_DLOG(DEBUG_PIM, "eip6: %s -> %s, eip6 plen %d",
1832 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1833 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1834 		    ntohs(eip6->ip6_plen));
1835 
1836 		/* verify the version number of the inner packet */
1837 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1838 			PIM6STAT_INC(pim6s_rcv_badregisters);
1839 			MRT6_DLOG(DEBUG_ANY, "invalid IP version (%d) "
1840 			    "of the inner packet",
1841 			    (eip6->ip6_vfc & IPV6_VERSION));
1842 			m_freem(m);
1843 			return (IPPROTO_NONE);
1844 		}
1845 
1846 		/* verify the inner packet is destined to a mcast group */
1847 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1848 			PIM6STAT_INC(pim6s_rcv_badregisters);
1849 			MRT6_DLOG(DEBUG_PIM, "inner packet of register "
1850 			    "is not multicast %s",
1851 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst));
1852 			m_freem(m);
1853 			return (IPPROTO_DONE);
1854 		}
1855 
1856 		/*
1857 		 * make a copy of the whole header to pass to the daemon later.
1858 		 */
1859 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1860 		if (mcp == NULL) {
1861 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "pim register: "
1862 			    "could not copy register head");
1863 			m_freem(m);
1864 			return (IPPROTO_DONE);
1865 		}
1866 
1867 		/*
1868 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1869 		 */
1870 		m_adj(m, off + PIM_MINLEN);
1871 		MRT6_DLOG(DEBUG_PIM, "forwarding decapsulated register: "
1872 		    "src %s, dst %s, mif %d",
1873 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1874 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst), reg_mif_num);
1875 
1876 		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1877 				dst.sin6_family, 0);
1878 
1879 		/* prepare the register head to send to the mrouting daemon */
1880 		m = mcp;
1881 	}
1882 
1883 	/*
1884 	 * Pass the PIM message up to the daemon; if it is a register message
1885 	 * pass the 'head' only up to the daemon. This includes the
1886 	 * encapsulator ip6 header, pim header, register header and the
1887 	 * encapsulated ip6 header.
1888 	 */
1889   pim6_input_to_daemon:
1890 	rip6_input(&m, offp, proto);
1891 	return (IPPROTO_DONE);
1892 }
1893 
1894 static int
1895 ip6_mroute_modevent(module_t mod, int type, void *unused)
1896 {
1897 
1898 	switch (type) {
1899 	case MOD_LOAD:
1900 		MROUTER6_LOCK_INIT();
1901 		MFC6_LOCK_INIT();
1902 		MIF6_LOCK_INIT();
1903 
1904 		pim6_encap_cookie = encap_attach_func(AF_INET6, IPPROTO_PIM,
1905 			pim6_encapcheck,
1906 			(const struct protosw *)&in6_pim_protosw, NULL);
1907 		if (pim6_encap_cookie == NULL) {
1908 			printf("ip6_mroute: unable to attach pim6 encap\n");
1909 			MIF6_LOCK_DESTROY();
1910 			MFC6_LOCK_DESTROY();
1911 			MROUTER6_LOCK_DESTROY();
1912 			return (EINVAL);
1913 		}
1914 
1915 		ip6_mforward = X_ip6_mforward;
1916 		ip6_mrouter_done = X_ip6_mrouter_done;
1917 		ip6_mrouter_get = X_ip6_mrouter_get;
1918 		ip6_mrouter_set = X_ip6_mrouter_set;
1919 		mrt6_ioctl = X_mrt6_ioctl;
1920 		break;
1921 
1922 	case MOD_UNLOAD:
1923 		if (V_ip6_mrouter != NULL)
1924 			return EINVAL;
1925 
1926 		if (pim6_encap_cookie) {
1927 			encap_detach(pim6_encap_cookie);
1928 			pim6_encap_cookie = NULL;
1929 		}
1930 		X_ip6_mrouter_done();
1931 		ip6_mforward = NULL;
1932 		ip6_mrouter_done = NULL;
1933 		ip6_mrouter_get = NULL;
1934 		ip6_mrouter_set = NULL;
1935 		mrt6_ioctl = NULL;
1936 
1937 		MIF6_LOCK_DESTROY();
1938 		MFC6_LOCK_DESTROY();
1939 		MROUTER6_LOCK_DESTROY();
1940 		break;
1941 
1942 	default:
1943 		return (EOPNOTSUPP);
1944 	}
1945 
1946 	return (0);
1947 }
1948 
1949 static moduledata_t ip6_mroutemod = {
1950 	"ip6_mroute",
1951 	ip6_mroute_modevent,
1952 	0
1953 };
1954 
1955 DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
1956