xref: /freebsd/sys/netinet6/ip6_mroute.c (revision 884a2a699669ec61e2366e3e358342dbc94be24a)
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 
87 #include <sys/param.h>
88 #include <sys/callout.h>
89 #include <sys/errno.h>
90 #include <sys/kernel.h>
91 #include <sys/lock.h>
92 #include <sys/malloc.h>
93 #include <sys/mbuf.h>
94 #include <sys/module.h>
95 #include <sys/domain.h>
96 #include <sys/protosw.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_types.h>
109 #include <net/raw_cb.h>
110 #include <net/vnet.h>
111 
112 #include <netinet/in.h>
113 #include <netinet/in_var.h>
114 #include <netinet/icmp6.h>
115 #include <netinet/ip_encap.h>
116 
117 #include <netinet/ip6.h>
118 #include <netinet6/ip6_var.h>
119 #include <netinet6/scope6_var.h>
120 #include <netinet6/nd6.h>
121 #include <netinet6/ip6_mroute.h>
122 #include <netinet6/ip6protosw.h>
123 #include <netinet6/pim6.h>
124 #include <netinet6/pim6_var.h>
125 
126 static MALLOC_DEFINE(M_MRTABLE6, "mf6c", "multicast forwarding cache entry");
127 
128 /* XXX: this is a very common idiom; move to <sys/mbuf.h> ? */
129 #define M_HASCL(m) ((m)->m_flags & M_EXT)
130 
131 static int	ip6_mdq(struct mbuf *, struct ifnet *, struct mf6c *);
132 static void	phyint_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
133 static int	register_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
134 static int	set_pim6(int *);
135 static int	socket_send(struct socket *, struct mbuf *,
136 		    struct sockaddr_in6 *);
137 
138 extern int in6_mcast_loop;
139 extern struct domain inet6domain;
140 
141 static const struct encaptab *pim6_encap_cookie;
142 static const struct ip6protosw in6_pim_protosw = {
143 	.pr_type =		SOCK_RAW,
144 	.pr_domain =		&inet6domain,
145 	.pr_protocol =		IPPROTO_PIM,
146 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
147 	.pr_input =		pim6_input,
148 	.pr_output =		rip6_output,
149 	.pr_ctloutput =		rip6_ctloutput,
150 	.pr_usrreqs =		&rip6_usrreqs
151 };
152 static int pim6_encapcheck(const struct mbuf *, int, int, void *);
153 
154 static VNET_DEFINE(int, ip6_mrouter_ver) = 0;
155 #define	V_ip6_mrouter_ver	VNET(ip6_mrouter_ver)
156 
157 SYSCTL_DECL(_net_inet6);
158 SYSCTL_DECL(_net_inet6_ip6);
159 SYSCTL_NODE(_net_inet6, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
160 
161 static struct mrt6stat mrt6stat;
162 SYSCTL_STRUCT(_net_inet6_ip6, OID_AUTO, mrt6stat, CTLFLAG_RW,
163     &mrt6stat, mrt6stat,
164     "Multicast Routing Statistics (struct mrt6stat, netinet6/ip6_mroute.h)");
165 
166 #define NO_RTE_FOUND	0x1
167 #define RTE_FOUND	0x2
168 
169 static struct mtx mrouter6_mtx;
170 #define	MROUTER6_LOCK()		mtx_lock(&mrouter6_mtx)
171 #define	MROUTER6_UNLOCK()	mtx_unlock(&mrouter6_mtx)
172 #define	MROUTER6_LOCK_ASSERT()	do {					\
173 	mtx_assert(&mrouter6_mtx, MA_OWNED);				\
174 	NET_ASSERT_GIANT();						\
175 } while (0)
176 #define	MROUTER6_LOCK_INIT()	\
177 	mtx_init(&mrouter6_mtx, "IPv6 multicast forwarding", NULL, MTX_DEF)
178 #define	MROUTER6_LOCK_DESTROY()	mtx_destroy(&mrouter6_mtx)
179 
180 static struct mf6c *mf6ctable[MF6CTBLSIZ];
181 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mf6ctable, CTLFLAG_RD,
182     &mf6ctable, sizeof(mf6ctable), "S,*mf6ctable[MF6CTBLSIZ]",
183     "IPv6 Multicast Forwarding Table (struct *mf6ctable[MF6CTBLSIZ], "
184     "netinet6/ip6_mroute.h)");
185 
186 static struct mtx mfc6_mtx;
187 #define	MFC6_LOCK()		mtx_lock(&mfc6_mtx)
188 #define	MFC6_UNLOCK()		mtx_unlock(&mfc6_mtx)
189 #define	MFC6_LOCK_ASSERT()	do {					\
190 	mtx_assert(&mfc6_mtx, MA_OWNED);				\
191 	NET_ASSERT_GIANT();						\
192 } while (0)
193 #define	MFC6_LOCK_INIT()		\
194 	mtx_init(&mfc6_mtx, "IPv6 multicast forwarding cache", NULL, MTX_DEF)
195 #define	MFC6_LOCK_DESTROY()	mtx_destroy(&mfc6_mtx)
196 
197 static u_char n6expire[MF6CTBLSIZ];
198 
199 static struct mif6 mif6table[MAXMIFS];
200 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mif6table, CTLFLAG_RD,
201     &mif6table, sizeof(mif6table), "S,mif6[MAXMIFS]",
202     "IPv6 Multicast Interfaces (struct mif6[MAXMIFS], netinet6/ip6_mroute.h)");
203 
204 static struct mtx mif6_mtx;
205 #define	MIF6_LOCK()		mtx_lock(&mif6_mtx)
206 #define	MIF6_UNLOCK()		mtx_unlock(&mif6_mtx)
207 #define	MIF6_LOCK_ASSERT()	mtx_assert(&mif6_mtx, MA_OWNED)
208 #define	MIF6_LOCK_INIT()	\
209 	mtx_init(&mif6_mtx, "IPv6 multicast interfaces", NULL, MTX_DEF)
210 #define	MIF6_LOCK_DESTROY()	mtx_destroy(&mif6_mtx)
211 
212 #ifdef MRT6DEBUG
213 static VNET_DEFINE(u_int, mrt6debug) = 0;	/* debug level */
214 #define	V_mrt6debug		VNET(mrt6debug)
215 #define DEBUG_MFC	0x02
216 #define DEBUG_FORWARD	0x04
217 #define DEBUG_EXPIRE	0x08
218 #define DEBUG_XMIT	0x10
219 #define DEBUG_REG	0x20
220 #define DEBUG_PIM	0x40
221 #endif
222 
223 static void	expire_upcalls(void *);
224 #define	EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
225 #define	UPCALL_EXPIRE	6		/* number of timeouts */
226 
227 /*
228  * XXX TODO: maintain a count to if_allmulti() calls in struct ifnet.
229  */
230 
231 /*
232  * 'Interfaces' associated with decapsulator (so we can tell
233  * packets that went through it from ones that get reflected
234  * by a broken gateway).  Different from IPv4 register_if,
235  * these interfaces are linked into the system ifnet list,
236  * because per-interface IPv6 statistics are maintained in
237  * ifp->if_afdata.  But it does not have any routes point
238  * to them.  I.e., packets can't be sent this way.  They
239  * only exist as a placeholder for multicast source
240  * verification.
241  */
242 static struct ifnet *multicast_register_if6;
243 
244 #define ENCAP_HOPS 64
245 
246 /*
247  * Private variables.
248  */
249 static mifi_t nummifs = 0;
250 static mifi_t reg_mif_num = (mifi_t)-1;
251 
252 static struct pim6stat pim6stat;
253 SYSCTL_STRUCT(_net_inet6_pim, PIM6CTL_STATS, stats, CTLFLAG_RD,
254     &pim6stat, pim6stat,
255     "PIM Statistics (struct pim6stat, netinet6/pim_var.h)");
256 
257 static VNET_DEFINE(int, pim6);
258 #define	V_pim6		VNET(pim6)
259 
260 /*
261  * Hash function for a source, group entry
262  */
263 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
264 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
265 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
266 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
267 
268 /*
269  * Find a route for a given origin IPv6 address and Multicast group address.
270  */
271 #define MF6CFIND(o, g, rt) do { \
272 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
273 	rt = NULL; \
274 	mrt6stat.mrt6s_mfc_lookups++; \
275 	while (_rt) { \
276 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
277 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
278 		    (_rt->mf6c_stall == NULL)) { \
279 			rt = _rt; \
280 			break; \
281 		} \
282 		_rt = _rt->mf6c_next; \
283 	} \
284 	if (rt == NULL) { \
285 		mrt6stat.mrt6s_mfc_misses++; \
286 	} \
287 } while (/*CONSTCOND*/ 0)
288 
289 /*
290  * Macros to compute elapsed time efficiently
291  * Borrowed from Van Jacobson's scheduling code
292  * XXX: replace with timersub() ?
293  */
294 #define TV_DELTA(a, b, delta) do { \
295 	    int xxs; \
296 		\
297 	    delta = (a).tv_usec - (b).tv_usec; \
298 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
299 	       switch (xxs) { \
300 		      case 2: \
301 			  delta += 1000000; \
302 			      /* FALLTHROUGH */ \
303 		      case 1: \
304 			  delta += 1000000; \
305 			  break; \
306 		      default: \
307 			  delta += (1000000 * xxs); \
308 	       } \
309 	    } \
310 } while (/*CONSTCOND*/ 0)
311 
312 /* XXX: replace with timercmp(a, b, <) ? */
313 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
314 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
315 
316 #ifdef UPCALL_TIMING
317 #define UPCALL_MAX	50
318 static u_long upcall_data[UPCALL_MAX + 1];
319 static void collate();
320 #endif /* UPCALL_TIMING */
321 
322 static int ip6_mrouter_init(struct socket *, int, int);
323 static int add_m6fc(struct mf6cctl *);
324 static int add_m6if(struct mif6ctl *);
325 static int del_m6fc(struct mf6cctl *);
326 static int del_m6if(mifi_t *);
327 static int del_m6if_locked(mifi_t *);
328 static int get_mif6_cnt(struct sioc_mif_req6 *);
329 static int get_sg_cnt(struct sioc_sg_req6 *);
330 
331 static struct callout expire_upcalls_ch;
332 
333 int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *);
334 int X_ip6_mrouter_done(void);
335 int X_ip6_mrouter_set(struct socket *, struct sockopt *);
336 int X_ip6_mrouter_get(struct socket *, struct sockopt *);
337 int X_mrt6_ioctl(u_long, caddr_t);
338 
339 /*
340  * Handle MRT setsockopt commands to modify the multicast routing tables.
341  */
342 int
343 X_ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
344 {
345 	int error = 0;
346 	int optval;
347 	struct mif6ctl mifc;
348 	struct mf6cctl mfcc;
349 	mifi_t mifi;
350 
351 	if (so != V_ip6_mrouter && sopt->sopt_name != MRT6_INIT)
352 		return (EACCES);
353 
354 	switch (sopt->sopt_name) {
355 	case MRT6_INIT:
356 #ifdef MRT6_OINIT
357 	case MRT6_OINIT:
358 #endif
359 		error = sooptcopyin(sopt, &optval, sizeof(optval),
360 		    sizeof(optval));
361 		if (error)
362 			break;
363 		error = ip6_mrouter_init(so, optval, sopt->sopt_name);
364 		break;
365 	case MRT6_DONE:
366 		error = X_ip6_mrouter_done();
367 		break;
368 	case MRT6_ADD_MIF:
369 		error = sooptcopyin(sopt, &mifc, sizeof(mifc), sizeof(mifc));
370 		if (error)
371 			break;
372 		error = add_m6if(&mifc);
373 		break;
374 	case MRT6_ADD_MFC:
375 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
376 		if (error)
377 			break;
378 		error = add_m6fc(&mfcc);
379 		break;
380 	case MRT6_DEL_MFC:
381 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
382 		if (error)
383 			break;
384 		error = del_m6fc(&mfcc);
385 		break;
386 	case MRT6_DEL_MIF:
387 		error = sooptcopyin(sopt, &mifi, sizeof(mifi), sizeof(mifi));
388 		if (error)
389 			break;
390 		error = del_m6if(&mifi);
391 		break;
392 	case MRT6_PIM:
393 		error = sooptcopyin(sopt, &optval, sizeof(optval),
394 		    sizeof(optval));
395 		if (error)
396 			break;
397 		error = set_pim6(&optval);
398 		break;
399 	default:
400 		error = EOPNOTSUPP;
401 		break;
402 	}
403 
404 	return (error);
405 }
406 
407 /*
408  * Handle MRT getsockopt commands
409  */
410 int
411 X_ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
412 {
413 	int error = 0;
414 
415 	if (so != V_ip6_mrouter)
416 		return (EACCES);
417 
418 	switch (sopt->sopt_name) {
419 		case MRT6_PIM:
420 			error = sooptcopyout(sopt, &V_pim6, sizeof(V_pim6));
421 			break;
422 	}
423 	return (error);
424 }
425 
426 /*
427  * Handle ioctl commands to obtain information from the cache
428  */
429 int
430 X_mrt6_ioctl(u_long cmd, caddr_t data)
431 {
432 	int ret;
433 
434 	ret = EINVAL;
435 
436 	switch (cmd) {
437 	case SIOCGETSGCNT_IN6:
438 		ret = get_sg_cnt((struct sioc_sg_req6 *)data);
439 		break;
440 
441 	case SIOCGETMIFCNT_IN6:
442 		ret = get_mif6_cnt((struct sioc_mif_req6 *)data);
443 		break;
444 
445 	default:
446 		break;
447 	}
448 
449 	return (ret);
450 }
451 
452 /*
453  * returns the packet, byte, rpf-failure count for the source group provided
454  */
455 static int
456 get_sg_cnt(struct sioc_sg_req6 *req)
457 {
458 	struct mf6c *rt;
459 	int ret;
460 
461 	ret = 0;
462 
463 	MFC6_LOCK();
464 
465 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
466 	if (rt == NULL) {
467 		ret = ESRCH;
468 	} else {
469 		req->pktcnt = rt->mf6c_pkt_cnt;
470 		req->bytecnt = rt->mf6c_byte_cnt;
471 		req->wrong_if = rt->mf6c_wrong_if;
472 	}
473 
474 	MFC6_UNLOCK();
475 
476 	return (ret);
477 }
478 
479 /*
480  * returns the input and output packet and byte counts on the mif provided
481  */
482 static int
483 get_mif6_cnt(struct sioc_mif_req6 *req)
484 {
485 	mifi_t mifi;
486 	int ret;
487 
488 	ret = 0;
489 	mifi = req->mifi;
490 
491 	MIF6_LOCK();
492 
493 	if (mifi >= nummifs) {
494 		ret = EINVAL;
495 	} else {
496 		req->icount = mif6table[mifi].m6_pkt_in;
497 		req->ocount = mif6table[mifi].m6_pkt_out;
498 		req->ibytes = mif6table[mifi].m6_bytes_in;
499 		req->obytes = mif6table[mifi].m6_bytes_out;
500 	}
501 
502 	MIF6_UNLOCK();
503 
504 	return (ret);
505 }
506 
507 static int
508 set_pim6(int *i)
509 {
510 	if ((*i != 1) && (*i != 0))
511 		return (EINVAL);
512 
513 	V_pim6 = *i;
514 
515 	return (0);
516 }
517 
518 /*
519  * Enable multicast routing
520  */
521 static int
522 ip6_mrouter_init(struct socket *so, int v, int cmd)
523 {
524 
525 #ifdef MRT6DEBUG
526 	if (V_mrt6debug)
527 		log(LOG_DEBUG,
528 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
529 		    so->so_type, so->so_proto->pr_protocol);
530 #endif
531 
532 	if (so->so_type != SOCK_RAW ||
533 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
534 		return (EOPNOTSUPP);
535 
536 	if (v != 1)
537 		return (ENOPROTOOPT);
538 
539 	MROUTER6_LOCK();
540 
541 	if (V_ip6_mrouter != NULL) {
542 		MROUTER6_UNLOCK();
543 		return (EADDRINUSE);
544 	}
545 
546 	V_ip6_mrouter = so;
547 	V_ip6_mrouter_ver = cmd;
548 
549 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
550 	bzero((caddr_t)n6expire, sizeof(n6expire));
551 
552 	V_pim6 = 0;/* used for stubbing out/in pim stuff */
553 
554 	callout_init(&expire_upcalls_ch, 0);
555 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
556 	    expire_upcalls, NULL);
557 
558 	MROUTER6_UNLOCK();
559 
560 #ifdef MRT6DEBUG
561 	if (V_mrt6debug)
562 		log(LOG_DEBUG, "ip6_mrouter_init\n");
563 #endif
564 
565 	return (0);
566 }
567 
568 /*
569  * Disable IPv6 multicast forwarding.
570  */
571 int
572 X_ip6_mrouter_done(void)
573 {
574 	mifi_t mifi;
575 	int i;
576 	struct mf6c *rt;
577 	struct rtdetq *rte;
578 
579 	MROUTER6_LOCK();
580 
581 	if (V_ip6_mrouter == NULL) {
582 		MROUTER6_UNLOCK();
583 		return (EINVAL);
584 	}
585 
586 	/*
587 	 * For each phyint in use, disable promiscuous reception of all IPv6
588 	 * multicasts.
589 	 */
590 	for (mifi = 0; mifi < nummifs; mifi++) {
591 		if (mif6table[mifi].m6_ifp &&
592 		    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
593 			if_allmulti(mif6table[mifi].m6_ifp, 0);
594 		}
595 	}
596 	bzero((caddr_t)mif6table, sizeof(mif6table));
597 	nummifs = 0;
598 
599 	V_pim6 = 0; /* used to stub out/in pim specific code */
600 
601 	callout_stop(&expire_upcalls_ch);
602 
603 	/*
604 	 * Free all multicast forwarding cache entries.
605 	 */
606 	MFC6_LOCK();
607 	for (i = 0; i < MF6CTBLSIZ; i++) {
608 		rt = mf6ctable[i];
609 		while (rt) {
610 			struct mf6c *frt;
611 
612 			for (rte = rt->mf6c_stall; rte != NULL; ) {
613 				struct rtdetq *n = rte->next;
614 
615 				m_free(rte->m);
616 				free(rte, M_MRTABLE6);
617 				rte = n;
618 			}
619 			frt = rt;
620 			rt = rt->mf6c_next;
621 			free(frt, M_MRTABLE6);
622 		}
623 	}
624 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
625 	MFC6_UNLOCK();
626 
627 	/*
628 	 * Reset register interface
629 	 */
630 	if (reg_mif_num != (mifi_t)-1 && multicast_register_if6 != NULL) {
631 		if_detach(multicast_register_if6);
632 		if_free(multicast_register_if6);
633 		reg_mif_num = (mifi_t)-1;
634 		multicast_register_if6 = NULL;
635 	}
636 
637 	V_ip6_mrouter = NULL;
638 	V_ip6_mrouter_ver = 0;
639 
640 	MROUTER6_UNLOCK();
641 
642 #ifdef MRT6DEBUG
643 	if (V_mrt6debug)
644 		log(LOG_DEBUG, "ip6_mrouter_done\n");
645 #endif
646 
647 	return (0);
648 }
649 
650 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
651 
652 /*
653  * Add a mif to the mif table
654  */
655 static int
656 add_m6if(struct mif6ctl *mifcp)
657 {
658 	struct mif6 *mifp;
659 	struct ifnet *ifp;
660 	int error;
661 
662 	MIF6_LOCK();
663 
664 	if (mifcp->mif6c_mifi >= MAXMIFS) {
665 		MIF6_UNLOCK();
666 		return (EINVAL);
667 	}
668 	mifp = mif6table + mifcp->mif6c_mifi;
669 	if (mifp->m6_ifp != NULL) {
670 		MIF6_UNLOCK();
671 		return (EADDRINUSE); /* XXX: is it appropriate? */
672 	}
673 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > V_if_index) {
674 		MIF6_UNLOCK();
675 		return (ENXIO);
676 	}
677 
678 	ifp = ifnet_byindex(mifcp->mif6c_pifi);
679 
680 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
681 		if (reg_mif_num == (mifi_t)-1) {
682 			ifp = if_alloc(IFT_OTHER);
683 
684 			if_initname(ifp, "register_mif", 0);
685 			ifp->if_flags |= IFF_LOOPBACK;
686 			if_attach(ifp);
687 			multicast_register_if6 = ifp;
688 			reg_mif_num = mifcp->mif6c_mifi;
689 			/*
690 			 * it is impossible to guess the ifindex of the
691 			 * register interface.  So mif6c_pifi is automatically
692 			 * calculated.
693 			 */
694 			mifcp->mif6c_pifi = ifp->if_index;
695 		} else {
696 			ifp = multicast_register_if6;
697 		}
698 	} else {
699 		/* Make sure the interface supports multicast */
700 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
701 			MIF6_UNLOCK();
702 			return (EOPNOTSUPP);
703 		}
704 
705 		error = if_allmulti(ifp, 1);
706 		if (error) {
707 			MIF6_UNLOCK();
708 			return (error);
709 		}
710 	}
711 
712 	mifp->m6_flags     = mifcp->mif6c_flags;
713 	mifp->m6_ifp       = ifp;
714 
715 	/* initialize per mif pkt counters */
716 	mifp->m6_pkt_in    = 0;
717 	mifp->m6_pkt_out   = 0;
718 	mifp->m6_bytes_in  = 0;
719 	mifp->m6_bytes_out = 0;
720 	bzero(&mifp->m6_route, sizeof(mifp->m6_route));
721 
722 	/* Adjust nummifs up if the mifi is higher than nummifs */
723 	if (nummifs <= mifcp->mif6c_mifi)
724 		nummifs = mifcp->mif6c_mifi + 1;
725 
726 	MIF6_UNLOCK();
727 
728 #ifdef MRT6DEBUG
729 	if (V_mrt6debug)
730 		log(LOG_DEBUG,
731 		    "add_mif #%d, phyint %s\n",
732 		    mifcp->mif6c_mifi,
733 		    ifp->if_xname);
734 #endif
735 
736 	return (0);
737 }
738 
739 /*
740  * Delete a mif from the mif table
741  */
742 static int
743 del_m6if_locked(mifi_t *mifip)
744 {
745 	struct mif6 *mifp = mif6table + *mifip;
746 	mifi_t mifi;
747 	struct ifnet *ifp;
748 
749 	MIF6_LOCK_ASSERT();
750 
751 	if (*mifip >= nummifs)
752 		return (EINVAL);
753 	if (mifp->m6_ifp == NULL)
754 		return (EINVAL);
755 
756 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
757 		/* XXX: TODO: Maintain an ALLMULTI refcount in struct ifnet. */
758 		ifp = mifp->m6_ifp;
759 		if_allmulti(ifp, 0);
760 	} else {
761 		if (reg_mif_num != (mifi_t)-1 &&
762 		    multicast_register_if6 != NULL) {
763 			if_detach(multicast_register_if6);
764 			if_free(multicast_register_if6);
765 			reg_mif_num = (mifi_t)-1;
766 			multicast_register_if6 = NULL;
767 		}
768 	}
769 
770 	bzero((caddr_t)mifp, sizeof(*mifp));
771 
772 	/* Adjust nummifs down */
773 	for (mifi = nummifs; mifi > 0; mifi--)
774 		if (mif6table[mifi - 1].m6_ifp)
775 			break;
776 	nummifs = mifi;
777 
778 #ifdef MRT6DEBUG
779 	if (V_mrt6debug)
780 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
781 #endif
782 
783 	return (0);
784 }
785 
786 static int
787 del_m6if(mifi_t *mifip)
788 {
789 	int cc;
790 
791 	MIF6_LOCK();
792 	cc = del_m6if_locked(mifip);
793 	MIF6_UNLOCK();
794 
795 	return (cc);
796 }
797 
798 /*
799  * Add an mfc entry
800  */
801 static int
802 add_m6fc(struct mf6cctl *mfccp)
803 {
804 	struct mf6c *rt;
805 	u_long hash;
806 	struct rtdetq *rte;
807 	u_short nstl;
808 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
809 
810 	MFC6_LOCK();
811 
812 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
813 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
814 
815 	/* If an entry already exists, just update the fields */
816 	if (rt) {
817 #ifdef MRT6DEBUG
818 		if (V_mrt6debug & DEBUG_MFC) {
819 		    log(LOG_DEBUG,
820 			"add_m6fc no upcall h %d o %s g %s p %x\n",
821 			ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
822 			ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
823 			mfccp->mf6cc_parent);
824 		}
825 #endif
826 
827 		rt->mf6c_parent = mfccp->mf6cc_parent;
828 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
829 
830 		MFC6_UNLOCK();
831 		return (0);
832 	}
833 
834 	/*
835 	 * Find the entry for which the upcall was made and update
836 	 */
837 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
838 			mfccp->mf6cc_mcastgrp.sin6_addr);
839 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
840 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
841 				       &mfccp->mf6cc_origin.sin6_addr) &&
842 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
843 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
844 		    (rt->mf6c_stall != NULL)) {
845 
846 			if (nstl++)
847 				log(LOG_ERR,
848 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
849 				    "multiple kernel entries",
850 				    ip6_sprintf(ip6bufo,
851 					    &mfccp->mf6cc_origin.sin6_addr),
852 				    ip6_sprintf(ip6bufg,
853 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
854 				    mfccp->mf6cc_parent, rt->mf6c_stall);
855 
856 #ifdef MRT6DEBUG
857 			if (V_mrt6debug & DEBUG_MFC)
858 				log(LOG_DEBUG,
859 				    "add_m6fc o %s g %s p %x dbg %x\n",
860 				    ip6_sprintf(ip6bufo,
861 					    &mfccp->mf6cc_origin.sin6_addr),
862 				    ip6_sprintf(ip6bufg,
863 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
864 				    mfccp->mf6cc_parent, rt->mf6c_stall);
865 #endif
866 
867 			rt->mf6c_origin     = mfccp->mf6cc_origin;
868 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
869 			rt->mf6c_parent     = mfccp->mf6cc_parent;
870 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
871 			/* initialize pkt counters per src-grp */
872 			rt->mf6c_pkt_cnt    = 0;
873 			rt->mf6c_byte_cnt   = 0;
874 			rt->mf6c_wrong_if   = 0;
875 
876 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
877 			n6expire[hash]--;
878 
879 			/* free packets Qed at the end of this entry */
880 			for (rte = rt->mf6c_stall; rte != NULL; ) {
881 				struct rtdetq *n = rte->next;
882 				ip6_mdq(rte->m, rte->ifp, rt);
883 				m_freem(rte->m);
884 #ifdef UPCALL_TIMING
885 				collate(&(rte->t));
886 #endif /* UPCALL_TIMING */
887 				free(rte, M_MRTABLE6);
888 				rte = n;
889 			}
890 			rt->mf6c_stall = NULL;
891 		}
892 	}
893 
894 	/*
895 	 * It is possible that an entry is being inserted without an upcall
896 	 */
897 	if (nstl == 0) {
898 #ifdef MRT6DEBUG
899 		if (V_mrt6debug & DEBUG_MFC)
900 		    log(LOG_DEBUG,
901 			"add_mfc no upcall h %d o %s g %s p %x\n",
902 			hash,
903 			ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
904 			ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
905 			mfccp->mf6cc_parent);
906 #endif
907 
908 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
909 
910 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
911 					       &mfccp->mf6cc_origin.sin6_addr)&&
912 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
913 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
914 
915 				rt->mf6c_origin     = mfccp->mf6cc_origin;
916 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
917 				rt->mf6c_parent     = mfccp->mf6cc_parent;
918 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
919 				/* initialize pkt counters per src-grp */
920 				rt->mf6c_pkt_cnt    = 0;
921 				rt->mf6c_byte_cnt   = 0;
922 				rt->mf6c_wrong_if   = 0;
923 
924 				if (rt->mf6c_expire)
925 					n6expire[hash]--;
926 				rt->mf6c_expire	   = 0;
927 			}
928 		}
929 		if (rt == NULL) {
930 			/* no upcall, so make a new entry */
931 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
932 						  M_NOWAIT);
933 			if (rt == NULL) {
934 				MFC6_UNLOCK();
935 				return (ENOBUFS);
936 			}
937 
938 			/* insert new entry at head of hash chain */
939 			rt->mf6c_origin     = mfccp->mf6cc_origin;
940 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
941 			rt->mf6c_parent     = mfccp->mf6cc_parent;
942 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
943 			/* initialize pkt counters per src-grp */
944 			rt->mf6c_pkt_cnt    = 0;
945 			rt->mf6c_byte_cnt   = 0;
946 			rt->mf6c_wrong_if   = 0;
947 			rt->mf6c_expire     = 0;
948 			rt->mf6c_stall = NULL;
949 
950 			/* link into table */
951 			rt->mf6c_next  = mf6ctable[hash];
952 			mf6ctable[hash] = rt;
953 		}
954 	}
955 
956 	MFC6_UNLOCK();
957 	return (0);
958 }
959 
960 #ifdef UPCALL_TIMING
961 /*
962  * collect delay statistics on the upcalls
963  */
964 static void
965 collate(struct timeval *t)
966 {
967 	u_long d;
968 	struct timeval tp;
969 	u_long delta;
970 
971 	GET_TIME(tp);
972 
973 	if (TV_LT(*t, tp))
974 	{
975 		TV_DELTA(tp, *t, delta);
976 
977 		d = delta >> 10;
978 		if (d > UPCALL_MAX)
979 			d = UPCALL_MAX;
980 
981 		++upcall_data[d];
982 	}
983 }
984 #endif /* UPCALL_TIMING */
985 
986 /*
987  * Delete an mfc entry
988  */
989 static int
990 del_m6fc(struct mf6cctl *mfccp)
991 {
992 	struct sockaddr_in6	origin;
993 	struct sockaddr_in6	mcastgrp;
994 	struct mf6c		*rt;
995 	struct mf6c		**nptr;
996 	u_long		hash;
997 
998 	origin = mfccp->mf6cc_origin;
999 	mcastgrp = mfccp->mf6cc_mcastgrp;
1000 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
1001 
1002 #ifdef MRT6DEBUG
1003 	if (V_mrt6debug & DEBUG_MFC) {
1004 		char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1005 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
1006 		    ip6_sprintf(ip6bufo, &origin.sin6_addr),
1007 		    ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr));
1008 	}
1009 #endif
1010 
1011 	MFC6_LOCK();
1012 
1013 	nptr = &mf6ctable[hash];
1014 	while ((rt = *nptr) != NULL) {
1015 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
1016 				       &rt->mf6c_origin.sin6_addr) &&
1017 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
1018 				       &rt->mf6c_mcastgrp.sin6_addr) &&
1019 		    rt->mf6c_stall == NULL)
1020 			break;
1021 
1022 		nptr = &rt->mf6c_next;
1023 	}
1024 	if (rt == NULL) {
1025 		MFC6_UNLOCK();
1026 		return (EADDRNOTAVAIL);
1027 	}
1028 
1029 	*nptr = rt->mf6c_next;
1030 	free(rt, M_MRTABLE6);
1031 
1032 	MFC6_UNLOCK();
1033 
1034 	return (0);
1035 }
1036 
1037 static int
1038 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
1039 {
1040 
1041 	if (s) {
1042 		if (sbappendaddr(&s->so_rcv,
1043 				 (struct sockaddr *)src,
1044 				 mm, (struct mbuf *)0) != 0) {
1045 			sorwakeup(s);
1046 			return (0);
1047 		}
1048 	}
1049 	m_freem(mm);
1050 	return (-1);
1051 }
1052 
1053 /*
1054  * IPv6 multicast forwarding function. This function assumes that the packet
1055  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
1056  * pointed to by "ifp", and the packet is to be relayed to other networks
1057  * that have members of the packet's destination IPv6 multicast group.
1058  *
1059  * The packet is returned unscathed to the caller, unless it is
1060  * erroneous, in which case a non-zero return value tells the caller to
1061  * discard it.
1062  *
1063  * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff
1064  * this function is called in the originating context (i.e., not when
1065  * forwarding a packet from other node).  ip6_output(), which is currently the
1066  * only function that calls this function is called in the originating context,
1067  * explicitly ensures this condition.  It is caller's responsibility to ensure
1068  * that if this function is called from somewhere else in the originating
1069  * context in the future.
1070  */
1071 int
1072 X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
1073 {
1074 	struct mf6c *rt;
1075 	struct mif6 *mifp;
1076 	struct mbuf *mm;
1077 	mifi_t mifi;
1078 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1079 
1080 #ifdef MRT6DEBUG
1081 	if (V_mrt6debug & DEBUG_FORWARD)
1082 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
1083 		    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1084 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1085 		    ifp->if_index);
1086 #endif
1087 
1088 	/*
1089 	 * Don't forward a packet with Hop limit of zero or one,
1090 	 * or a packet destined to a local-only group.
1091 	 */
1092 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
1093 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1094 		return (0);
1095 	ip6->ip6_hlim--;
1096 
1097 	/*
1098 	 * Source address check: do not forward packets with unspecified
1099 	 * source. It was discussed in July 2000, on ipngwg mailing list.
1100 	 * This is rather more serious than unicast cases, because some
1101 	 * MLD packets can be sent with the unspecified source address
1102 	 * (although such packets must normally set 1 to the hop limit field).
1103 	 */
1104 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1105 		V_ip6stat.ip6s_cantforward++;
1106 		if (V_ip6_log_time + V_ip6_log_interval < time_second) {
1107 			V_ip6_log_time = time_second;
1108 			log(LOG_DEBUG,
1109 			    "cannot forward "
1110 			    "from %s to %s nxt %d received on %s\n",
1111 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1112 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1113 			    ip6->ip6_nxt,
1114 			    if_name(m->m_pkthdr.rcvif));
1115 		}
1116 		return (0);
1117 	}
1118 
1119 	MFC6_LOCK();
1120 
1121 	/*
1122 	 * Determine forwarding mifs from the forwarding cache table
1123 	 */
1124 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
1125 
1126 	/* Entry exists, so forward if necessary */
1127 	if (rt) {
1128 		MFC6_UNLOCK();
1129 		return (ip6_mdq(m, ifp, rt));
1130 	} else {
1131 		/*
1132 		 * If we don't have a route for packet's origin,
1133 		 * Make a copy of the packet &
1134 		 * send message to routing daemon
1135 		 */
1136 
1137 		struct mbuf *mb0;
1138 		struct rtdetq *rte;
1139 		u_long hash;
1140 /*		int i, npkts;*/
1141 #ifdef UPCALL_TIMING
1142 		struct timeval tp;
1143 
1144 		GET_TIME(tp);
1145 #endif /* UPCALL_TIMING */
1146 
1147 		mrt6stat.mrt6s_no_route++;
1148 #ifdef MRT6DEBUG
1149 		if (V_mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
1150 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
1151 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1152 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1153 #endif
1154 
1155 		/*
1156 		 * Allocate mbufs early so that we don't do extra work if we
1157 		 * are just going to fail anyway.
1158 		 */
1159 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE6,
1160 					      M_NOWAIT);
1161 		if (rte == NULL) {
1162 			MFC6_UNLOCK();
1163 			return (ENOBUFS);
1164 		}
1165 		mb0 = m_copy(m, 0, M_COPYALL);
1166 		/*
1167 		 * Pullup packet header if needed before storing it,
1168 		 * as other references may modify it in the meantime.
1169 		 */
1170 		if (mb0 &&
1171 		    (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1172 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1173 		if (mb0 == NULL) {
1174 			free(rte, M_MRTABLE6);
1175 			MFC6_UNLOCK();
1176 			return (ENOBUFS);
1177 		}
1178 
1179 		/* is there an upcall waiting for this packet? */
1180 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1181 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1182 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1183 					       &rt->mf6c_origin.sin6_addr) &&
1184 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1185 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1186 			    (rt->mf6c_stall != NULL))
1187 				break;
1188 		}
1189 
1190 		if (rt == NULL) {
1191 			struct mrt6msg *im;
1192 #ifdef MRT6_OINIT
1193 			struct omrt6msg *oim;
1194 #endif
1195 
1196 			/* no upcall, so make a new entry */
1197 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
1198 						  M_NOWAIT);
1199 			if (rt == NULL) {
1200 				free(rte, M_MRTABLE6);
1201 				m_freem(mb0);
1202 				MFC6_UNLOCK();
1203 				return (ENOBUFS);
1204 			}
1205 			/*
1206 			 * Make a copy of the header to send to the user
1207 			 * level process
1208 			 */
1209 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1210 
1211 			if (mm == NULL) {
1212 				free(rte, M_MRTABLE6);
1213 				m_freem(mb0);
1214 				free(rt, M_MRTABLE6);
1215 				MFC6_UNLOCK();
1216 				return (ENOBUFS);
1217 			}
1218 
1219 			/*
1220 			 * Send message to routing daemon
1221 			 */
1222 			sin6.sin6_addr = ip6->ip6_src;
1223 
1224 			im = NULL;
1225 #ifdef MRT6_OINIT
1226 			oim = NULL;
1227 #endif
1228 			switch (V_ip6_mrouter_ver) {
1229 #ifdef MRT6_OINIT
1230 			case MRT6_OINIT:
1231 				oim = mtod(mm, struct omrt6msg *);
1232 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1233 				oim->im6_mbz = 0;
1234 				break;
1235 #endif
1236 			case MRT6_INIT:
1237 				im = mtod(mm, struct mrt6msg *);
1238 				im->im6_msgtype = MRT6MSG_NOCACHE;
1239 				im->im6_mbz = 0;
1240 				break;
1241 			default:
1242 				free(rte, M_MRTABLE6);
1243 				m_freem(mb0);
1244 				free(rt, M_MRTABLE6);
1245 				MFC6_UNLOCK();
1246 				return (EINVAL);
1247 			}
1248 
1249 #ifdef MRT6DEBUG
1250 			if (V_mrt6debug & DEBUG_FORWARD)
1251 				log(LOG_DEBUG,
1252 				    "getting the iif info in the kernel\n");
1253 #endif
1254 
1255 			for (mifp = mif6table, mifi = 0;
1256 			     mifi < nummifs && mifp->m6_ifp != ifp;
1257 			     mifp++, mifi++)
1258 				;
1259 
1260 			switch (V_ip6_mrouter_ver) {
1261 #ifdef MRT6_OINIT
1262 			case MRT6_OINIT:
1263 				oim->im6_mif = mifi;
1264 				break;
1265 #endif
1266 			case MRT6_INIT:
1267 				im->im6_mif = mifi;
1268 				break;
1269 			}
1270 
1271 			if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1272 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1273 				    "socket queue full\n");
1274 				mrt6stat.mrt6s_upq_sockfull++;
1275 				free(rte, M_MRTABLE6);
1276 				m_freem(mb0);
1277 				free(rt, M_MRTABLE6);
1278 				MFC6_UNLOCK();
1279 				return (ENOBUFS);
1280 			}
1281 
1282 			mrt6stat.mrt6s_upcalls++;
1283 
1284 			/* insert new entry at head of hash chain */
1285 			bzero(rt, sizeof(*rt));
1286 			rt->mf6c_origin.sin6_family = AF_INET6;
1287 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1288 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1289 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1290 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1291 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1292 			rt->mf6c_expire = UPCALL_EXPIRE;
1293 			n6expire[hash]++;
1294 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1295 
1296 			/* link into table */
1297 			rt->mf6c_next  = mf6ctable[hash];
1298 			mf6ctable[hash] = rt;
1299 			/* Add this entry to the end of the queue */
1300 			rt->mf6c_stall = rte;
1301 		} else {
1302 			/* determine if q has overflowed */
1303 			struct rtdetq **p;
1304 			int npkts = 0;
1305 
1306 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1307 				if (++npkts > MAX_UPQ6) {
1308 					mrt6stat.mrt6s_upq_ovflw++;
1309 					free(rte, M_MRTABLE6);
1310 					m_freem(mb0);
1311 					MFC6_UNLOCK();
1312 					return (0);
1313 				}
1314 
1315 			/* Add this entry to the end of the queue */
1316 			*p = rte;
1317 		}
1318 
1319 		rte->next = NULL;
1320 		rte->m = mb0;
1321 		rte->ifp = ifp;
1322 #ifdef UPCALL_TIMING
1323 		rte->t = tp;
1324 #endif /* UPCALL_TIMING */
1325 
1326 		MFC6_UNLOCK();
1327 
1328 		return (0);
1329 	}
1330 }
1331 
1332 /*
1333  * Clean up cache entries if upcalls are not serviced
1334  * Call from the Slow Timeout mechanism, every half second.
1335  */
1336 static void
1337 expire_upcalls(void *unused)
1338 {
1339 	struct rtdetq *rte;
1340 	struct mf6c *mfc, **nptr;
1341 	int i;
1342 
1343 	MFC6_LOCK();
1344 	for (i = 0; i < MF6CTBLSIZ; i++) {
1345 		if (n6expire[i] == 0)
1346 			continue;
1347 		nptr = &mf6ctable[i];
1348 		while ((mfc = *nptr) != NULL) {
1349 			rte = mfc->mf6c_stall;
1350 			/*
1351 			 * Skip real cache entries
1352 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1353 			 * If it expires now
1354 			 */
1355 			if (rte != NULL &&
1356 			    mfc->mf6c_expire != 0 &&
1357 			    --mfc->mf6c_expire == 0) {
1358 #ifdef MRT6DEBUG
1359 				if (V_mrt6debug & DEBUG_EXPIRE) {
1360 					char ip6bufo[INET6_ADDRSTRLEN];
1361 					char ip6bufg[INET6_ADDRSTRLEN];
1362 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1363 					    ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr),
1364 					    ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr));
1365 				}
1366 #endif
1367 				/*
1368 				 * drop all the packets
1369 				 * free the mbuf with the pkt, if, timing info
1370 				 */
1371 				do {
1372 					struct rtdetq *n = rte->next;
1373 					m_freem(rte->m);
1374 					free(rte, M_MRTABLE6);
1375 					rte = n;
1376 				} while (rte != NULL);
1377 				mrt6stat.mrt6s_cache_cleanups++;
1378 				n6expire[i]--;
1379 
1380 				*nptr = mfc->mf6c_next;
1381 				free(mfc, M_MRTABLE6);
1382 			} else {
1383 				nptr = &mfc->mf6c_next;
1384 			}
1385 		}
1386 	}
1387 	MFC6_UNLOCK();
1388 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1389 	    expire_upcalls, NULL);
1390 }
1391 
1392 /*
1393  * Packet forwarding routine once entry in the cache is made
1394  */
1395 static int
1396 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1397 {
1398 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1399 	mifi_t mifi, iif;
1400 	struct mif6 *mifp;
1401 	int plen = m->m_pkthdr.len;
1402 	struct in6_addr src0, dst0; /* copies for local work */
1403 	u_int32_t iszone, idzone, oszone, odzone;
1404 	int error = 0;
1405 
1406 /*
1407  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1408  * input, they shouldn't get counted on output, so statistics keeping is
1409  * separate.
1410  */
1411 
1412 #define MC6_SEND(ip6, mifp, m) do {				\
1413 	if ((mifp)->m6_flags & MIFF_REGISTER)			\
1414 		register_send((ip6), (mifp), (m));		\
1415 	else							\
1416 		phyint_send((ip6), (mifp), (m));		\
1417 } while (/*CONSTCOND*/ 0)
1418 
1419 	/*
1420 	 * Don't forward if it didn't arrive from the parent mif
1421 	 * for its origin.
1422 	 */
1423 	mifi = rt->mf6c_parent;
1424 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1425 		/* came in the wrong interface */
1426 #ifdef MRT6DEBUG
1427 		if (V_mrt6debug & DEBUG_FORWARD)
1428 			log(LOG_DEBUG,
1429 			    "wrong if: ifid %d mifi %d mififid %x\n",
1430 			    ifp->if_index, mifi,
1431 			    mif6table[mifi].m6_ifp->if_index);
1432 #endif
1433 		mrt6stat.mrt6s_wrong_if++;
1434 		rt->mf6c_wrong_if++;
1435 		/*
1436 		 * If we are doing PIM processing, and we are forwarding
1437 		 * packets on this interface, send a message to the
1438 		 * routing daemon.
1439 		 */
1440 		/* have to make sure this is a valid mif */
1441 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1442 			if (V_pim6 && (m->m_flags & M_LOOP) == 0) {
1443 				/*
1444 				 * Check the M_LOOP flag to avoid an
1445 				 * unnecessary PIM assert.
1446 				 * XXX: M_LOOP is an ad-hoc hack...
1447 				 */
1448 				static struct sockaddr_in6 sin6 =
1449 				{ sizeof(sin6), AF_INET6 };
1450 
1451 				struct mbuf *mm;
1452 				struct mrt6msg *im;
1453 #ifdef MRT6_OINIT
1454 				struct omrt6msg *oim;
1455 #endif
1456 
1457 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1458 				if (mm &&
1459 				    (M_HASCL(mm) ||
1460 				     mm->m_len < sizeof(struct ip6_hdr)))
1461 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1462 				if (mm == NULL)
1463 					return (ENOBUFS);
1464 
1465 #ifdef MRT6_OINIT
1466 				oim = NULL;
1467 #endif
1468 				im = NULL;
1469 				switch (V_ip6_mrouter_ver) {
1470 #ifdef MRT6_OINIT
1471 				case MRT6_OINIT:
1472 					oim = mtod(mm, struct omrt6msg *);
1473 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1474 					oim->im6_mbz = 0;
1475 					break;
1476 #endif
1477 				case MRT6_INIT:
1478 					im = mtod(mm, struct mrt6msg *);
1479 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1480 					im->im6_mbz = 0;
1481 					break;
1482 				default:
1483 					m_freem(mm);
1484 					return (EINVAL);
1485 				}
1486 
1487 				for (mifp = mif6table, iif = 0;
1488 				     iif < nummifs && mifp &&
1489 					     mifp->m6_ifp != ifp;
1490 				     mifp++, iif++)
1491 					;
1492 
1493 				switch (V_ip6_mrouter_ver) {
1494 #ifdef MRT6_OINIT
1495 				case MRT6_OINIT:
1496 					oim->im6_mif = iif;
1497 					sin6.sin6_addr = oim->im6_src;
1498 					break;
1499 #endif
1500 				case MRT6_INIT:
1501 					im->im6_mif = iif;
1502 					sin6.sin6_addr = im->im6_src;
1503 					break;
1504 				}
1505 
1506 				mrt6stat.mrt6s_upcalls++;
1507 
1508 				if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1509 #ifdef MRT6DEBUG
1510 					if (V_mrt6debug)
1511 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1512 #endif
1513 					++mrt6stat.mrt6s_upq_sockfull;
1514 					return (ENOBUFS);
1515 				}	/* if socket Q full */
1516 			}		/* if PIM */
1517 		return (0);
1518 	}			/* if wrong iif */
1519 
1520 	/* If I sourced this packet, it counts as output, else it was input. */
1521 	if (m->m_pkthdr.rcvif == NULL) {
1522 		/* XXX: is rcvif really NULL when output?? */
1523 		mif6table[mifi].m6_pkt_out++;
1524 		mif6table[mifi].m6_bytes_out += plen;
1525 	} else {
1526 		mif6table[mifi].m6_pkt_in++;
1527 		mif6table[mifi].m6_bytes_in += plen;
1528 	}
1529 	rt->mf6c_pkt_cnt++;
1530 	rt->mf6c_byte_cnt += plen;
1531 
1532 	/*
1533 	 * For each mif, forward a copy of the packet if there are group
1534 	 * members downstream on the interface.
1535 	 */
1536 	src0 = ip6->ip6_src;
1537 	dst0 = ip6->ip6_dst;
1538 	if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 ||
1539 	    (error = in6_setscope(&dst0, ifp, &idzone)) != 0) {
1540 		V_ip6stat.ip6s_badscope++;
1541 		return (error);
1542 	}
1543 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) {
1544 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1545 			/*
1546 			 * check if the outgoing packet is going to break
1547 			 * a scope boundary.
1548 			 * XXX For packets through PIM register tunnel
1549 			 * interface, we believe a routing daemon.
1550 			 */
1551 			if (!(mif6table[rt->mf6c_parent].m6_flags &
1552 			      MIFF_REGISTER) &&
1553 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
1554 				if (in6_setscope(&src0, mif6table[mifi].m6_ifp,
1555 				    &oszone) ||
1556 				    in6_setscope(&dst0, mif6table[mifi].m6_ifp,
1557 				    &odzone) ||
1558 				    iszone != oszone ||
1559 				    idzone != odzone) {
1560 					V_ip6stat.ip6s_badscope++;
1561 					continue;
1562 				}
1563 			}
1564 
1565 			mifp->m6_pkt_out++;
1566 			mifp->m6_bytes_out += plen;
1567 			MC6_SEND(ip6, mifp, m);
1568 		}
1569 	}
1570 	return (0);
1571 }
1572 
1573 static void
1574 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1575 {
1576 	struct mbuf *mb_copy;
1577 	struct ifnet *ifp = mifp->m6_ifp;
1578 	int error = 0;
1579 	struct sockaddr_in6 *dst6;
1580 	u_long linkmtu;
1581 
1582 	dst6 = &mifp->m6_route.ro_dst;
1583 
1584 	/*
1585 	 * Make a new reference to the packet; make sure that
1586 	 * the IPv6 header is actually copied, not just referenced,
1587 	 * so that ip6_output() only scribbles on the copy.
1588 	 */
1589 	mb_copy = m_copy(m, 0, M_COPYALL);
1590 	if (mb_copy &&
1591 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1592 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1593 	if (mb_copy == NULL) {
1594 		return;
1595 	}
1596 	/* set MCAST flag to the outgoing packet */
1597 	mb_copy->m_flags |= M_MCAST;
1598 
1599 	/*
1600 	 * If we sourced the packet, call ip6_output since we may devide
1601 	 * the packet into fragments when the packet is too big for the
1602 	 * outgoing interface.
1603 	 * Otherwise, we can simply send the packet to the interface
1604 	 * sending queue.
1605 	 */
1606 	if (m->m_pkthdr.rcvif == NULL) {
1607 		struct ip6_moptions im6o;
1608 
1609 		im6o.im6o_multicast_ifp = ifp;
1610 		/* XXX: ip6_output will override ip6->ip6_hlim */
1611 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1612 		im6o.im6o_multicast_loop = 1;
1613 		error = ip6_output(mb_copy, NULL, &mifp->m6_route,
1614 				   IPV6_FORWARDING, &im6o, NULL, NULL);
1615 
1616 #ifdef MRT6DEBUG
1617 		if (V_mrt6debug & DEBUG_XMIT)
1618 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1619 			    mifp - mif6table, error);
1620 #endif
1621 		return;
1622 	}
1623 
1624 	/*
1625 	 * If configured to loop back multicasts by default,
1626 	 * loop back a copy now.
1627 	 */
1628 	if (in6_mcast_loop) {
1629 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1630 		dst6->sin6_family = AF_INET6;
1631 		dst6->sin6_addr = ip6->ip6_dst;
1632 		ip6_mloopback(ifp, m, &mifp->m6_route.ro_dst);
1633 	}
1634 
1635 	/*
1636 	 * Put the packet into the sending queue of the outgoing interface
1637 	 * if it would fit in the MTU of the interface.
1638 	 */
1639 	linkmtu = IN6_LINKMTU(ifp);
1640 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1641 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1642 		dst6->sin6_family = AF_INET6;
1643 		dst6->sin6_addr = ip6->ip6_dst;
1644 		/*
1645 		 * We just call if_output instead of nd6_output here, since
1646 		 * we need no ND for a multicast forwarded packet...right?
1647 		 */
1648 		error = (*ifp->if_output)(ifp, mb_copy,
1649 		    (struct sockaddr *)&mifp->m6_route.ro_dst, NULL);
1650 #ifdef MRT6DEBUG
1651 		if (V_mrt6debug & DEBUG_XMIT)
1652 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1653 			    mifp - mif6table, error);
1654 #endif
1655 	} else {
1656 		/*
1657 		 * pMTU discovery is intentionally disabled by default, since
1658 		 * various router may notify pMTU in multicast, which can be
1659 		 * a DDoS to a router
1660 		 */
1661 		if (V_ip6_mcast_pmtu)
1662 			icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1663 		else {
1664 #ifdef MRT6DEBUG
1665 			if (V_mrt6debug & DEBUG_XMIT) {
1666 				char ip6bufs[INET6_ADDRSTRLEN];
1667 				char ip6bufd[INET6_ADDRSTRLEN];
1668 				log(LOG_DEBUG,
1669 				    "phyint_send: packet too big on %s o %s "
1670 				    "g %s size %d(discarded)\n",
1671 				    if_name(ifp),
1672 				    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1673 				    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1674 				    mb_copy->m_pkthdr.len);
1675 			}
1676 #endif /* MRT6DEBUG */
1677 			m_freem(mb_copy); /* simply discard the packet */
1678 		}
1679 	}
1680 }
1681 
1682 static int
1683 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1684 {
1685 	struct mbuf *mm;
1686 	int i, len = m->m_pkthdr.len;
1687 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1688 	struct mrt6msg *im6;
1689 
1690 #ifdef MRT6DEBUG
1691 	if (V_mrt6debug) {
1692 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1693 		log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1694 		    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1695 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1696 	}
1697 #endif
1698 	++pim6stat.pim6s_snd_registers;
1699 
1700 	/* Make a copy of the packet to send to the user level process */
1701 	MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1702 	if (mm == NULL)
1703 		return (ENOBUFS);
1704 	mm->m_pkthdr.rcvif = NULL;
1705 	mm->m_data += max_linkhdr;
1706 	mm->m_len = sizeof(struct ip6_hdr);
1707 
1708 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1709 		m_freem(mm);
1710 		return (ENOBUFS);
1711 	}
1712 	i = MHLEN - M_LEADINGSPACE(mm);
1713 	if (i > len)
1714 		i = len;
1715 	mm = m_pullup(mm, i);
1716 	if (mm == NULL)
1717 		return (ENOBUFS);
1718 /* TODO: check it! */
1719 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1720 
1721 	/*
1722 	 * Send message to routing daemon
1723 	 */
1724 	sin6.sin6_addr = ip6->ip6_src;
1725 
1726 	im6 = mtod(mm, struct mrt6msg *);
1727 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1728 	im6->im6_mbz          = 0;
1729 
1730 	im6->im6_mif = mif - mif6table;
1731 
1732 	/* iif info is not given for reg. encap.n */
1733 	mrt6stat.mrt6s_upcalls++;
1734 
1735 	if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1736 #ifdef MRT6DEBUG
1737 		if (V_mrt6debug)
1738 			log(LOG_WARNING,
1739 			    "register_send: ip6_mrouter socket queue full\n");
1740 #endif
1741 		++mrt6stat.mrt6s_upq_sockfull;
1742 		return (ENOBUFS);
1743 	}
1744 	return (0);
1745 }
1746 
1747 /*
1748  * pim6_encapcheck() is called by the encap6_input() path at runtime to
1749  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
1750  * into the kernel.
1751  */
1752 static int
1753 pim6_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
1754 {
1755 
1756 #ifdef DIAGNOSTIC
1757     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
1758 #endif
1759     if (proto != IPPROTO_PIM)
1760 	return 0;	/* not for us; reject the datagram. */
1761 
1762     return 64;		/* claim the datagram. */
1763 }
1764 
1765 /*
1766  * PIM sparse mode hook
1767  * Receives the pim control messages, and passes them up to the listening
1768  * socket, using rip6_input.
1769  * The only message processed is the REGISTER pim message; the pim header
1770  * is stripped off, and the inner packet is passed to register_mforward.
1771  */
1772 int
1773 pim6_input(struct mbuf **mp, int *offp, int proto)
1774 {
1775 	struct pim *pim; /* pointer to a pim struct */
1776 	struct ip6_hdr *ip6;
1777 	int pimlen;
1778 	struct mbuf *m = *mp;
1779 	int minlen;
1780 	int off = *offp;
1781 
1782 	++pim6stat.pim6s_rcv_total;
1783 
1784 	ip6 = mtod(m, struct ip6_hdr *);
1785 	pimlen = m->m_pkthdr.len - *offp;
1786 
1787 	/*
1788 	 * Validate lengths
1789 	 */
1790 	if (pimlen < PIM_MINLEN) {
1791 		++pim6stat.pim6s_rcv_tooshort;
1792 #ifdef MRT6DEBUG
1793 		if (V_mrt6debug & DEBUG_PIM)
1794 			log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1795 #endif
1796 		m_freem(m);
1797 		return (IPPROTO_DONE);
1798 	}
1799 
1800 	/*
1801 	 * if the packet is at least as big as a REGISTER, go ahead
1802 	 * and grab the PIM REGISTER header size, to avoid another
1803 	 * possible m_pullup() later.
1804 	 *
1805 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1806 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1807 	 */
1808 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1809 
1810 	/*
1811 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1812 	 * possibly the PIM REGISTER header
1813 	 */
1814 #ifndef PULLDOWN_TEST
1815 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1816 	/* adjust pointer */
1817 	ip6 = mtod(m, struct ip6_hdr *);
1818 
1819 	/* adjust mbuf to point to the PIM header */
1820 	pim = (struct pim *)((caddr_t)ip6 + off);
1821 #else
1822 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1823 	if (pim == NULL) {
1824 		pim6stat.pim6s_rcv_tooshort++;
1825 		return (IPPROTO_DONE);
1826 	}
1827 #endif
1828 
1829 #define PIM6_CHECKSUM
1830 #ifdef PIM6_CHECKSUM
1831 	{
1832 		int cksumlen;
1833 
1834 		/*
1835 		 * Validate checksum.
1836 		 * If PIM REGISTER, exclude the data packet
1837 		 */
1838 		if (pim->pim_type == PIM_REGISTER)
1839 			cksumlen = PIM_MINLEN;
1840 		else
1841 			cksumlen = pimlen;
1842 
1843 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1844 			++pim6stat.pim6s_rcv_badsum;
1845 #ifdef MRT6DEBUG
1846 			if (V_mrt6debug & DEBUG_PIM)
1847 				log(LOG_DEBUG,
1848 				    "pim6_input: invalid checksum\n");
1849 #endif
1850 			m_freem(m);
1851 			return (IPPROTO_DONE);
1852 		}
1853 	}
1854 #endif /* PIM_CHECKSUM */
1855 
1856 	/* PIM version check */
1857 	if (pim->pim_ver != PIM_VERSION) {
1858 		++pim6stat.pim6s_rcv_badversion;
1859 #ifdef MRT6DEBUG
1860 		log(LOG_ERR,
1861 		    "pim6_input: incorrect version %d, expecting %d\n",
1862 		    pim->pim_ver, PIM_VERSION);
1863 #endif
1864 		m_freem(m);
1865 		return (IPPROTO_DONE);
1866 	}
1867 
1868 	if (pim->pim_type == PIM_REGISTER) {
1869 		/*
1870 		 * since this is a REGISTER, we'll make a copy of the register
1871 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1872 		 * routing daemon.
1873 		 */
1874 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1875 
1876 		struct mbuf *mcp;
1877 		struct ip6_hdr *eip6;
1878 		u_int32_t *reghdr;
1879 		int rc;
1880 #ifdef MRT6DEBUG
1881 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1882 #endif
1883 
1884 		++pim6stat.pim6s_rcv_registers;
1885 
1886 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1887 #ifdef MRT6DEBUG
1888 			if (V_mrt6debug & DEBUG_PIM)
1889 				log(LOG_DEBUG,
1890 				    "pim6_input: register mif not set: %d\n",
1891 				    reg_mif_num);
1892 #endif
1893 			m_freem(m);
1894 			return (IPPROTO_DONE);
1895 		}
1896 
1897 		reghdr = (u_int32_t *)(pim + 1);
1898 
1899 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1900 			goto pim6_input_to_daemon;
1901 
1902 		/*
1903 		 * Validate length
1904 		 */
1905 		if (pimlen < PIM6_REG_MINLEN) {
1906 			++pim6stat.pim6s_rcv_tooshort;
1907 			++pim6stat.pim6s_rcv_badregisters;
1908 #ifdef MRT6DEBUG
1909 			log(LOG_ERR,
1910 			    "pim6_input: register packet size too "
1911 			    "small %d from %s\n",
1912 			    pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src));
1913 #endif
1914 			m_freem(m);
1915 			return (IPPROTO_DONE);
1916 		}
1917 
1918 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1919 #ifdef MRT6DEBUG
1920 		if (V_mrt6debug & DEBUG_PIM)
1921 			log(LOG_DEBUG,
1922 			    "pim6_input[register], eip6: %s -> %s, "
1923 			    "eip6 plen %d\n",
1924 			    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1925 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1926 			    ntohs(eip6->ip6_plen));
1927 #endif
1928 
1929 		/* verify the version number of the inner packet */
1930 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1931 			++pim6stat.pim6s_rcv_badregisters;
1932 #ifdef MRT6DEBUG
1933 			log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1934 			    "of the inner packet\n",
1935 			    (eip6->ip6_vfc & IPV6_VERSION));
1936 #endif
1937 			m_freem(m);
1938 			return (IPPROTO_NONE);
1939 		}
1940 
1941 		/* verify the inner packet is destined to a mcast group */
1942 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1943 			++pim6stat.pim6s_rcv_badregisters;
1944 #ifdef MRT6DEBUG
1945 			if (V_mrt6debug & DEBUG_PIM)
1946 				log(LOG_DEBUG,
1947 				    "pim6_input: inner packet of register "
1948 				    "is not multicast %s\n",
1949 				    ip6_sprintf(ip6bufd, &eip6->ip6_dst));
1950 #endif
1951 			m_freem(m);
1952 			return (IPPROTO_DONE);
1953 		}
1954 
1955 		/*
1956 		 * make a copy of the whole header to pass to the daemon later.
1957 		 */
1958 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1959 		if (mcp == NULL) {
1960 #ifdef MRT6DEBUG
1961 			log(LOG_ERR,
1962 			    "pim6_input: pim register: "
1963 			    "could not copy register head\n");
1964 #endif
1965 			m_freem(m);
1966 			return (IPPROTO_DONE);
1967 		}
1968 
1969 		/*
1970 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1971 		 */
1972 		m_adj(m, off + PIM_MINLEN);
1973 #ifdef MRT6DEBUG
1974 		if (V_mrt6debug & DEBUG_PIM) {
1975 			log(LOG_DEBUG,
1976 			    "pim6_input: forwarding decapsulated register: "
1977 			    "src %s, dst %s, mif %d\n",
1978 			    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1979 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1980 			    reg_mif_num);
1981 		}
1982 #endif
1983 
1984 		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1985 				dst.sin6_family, 0);
1986 
1987 		/* prepare the register head to send to the mrouting daemon */
1988 		m = mcp;
1989 	}
1990 
1991 	/*
1992 	 * Pass the PIM message up to the daemon; if it is a register message
1993 	 * pass the 'head' only up to the daemon. This includes the
1994 	 * encapsulator ip6 header, pim header, register header and the
1995 	 * encapsulated ip6 header.
1996 	 */
1997   pim6_input_to_daemon:
1998 	rip6_input(&m, offp, proto);
1999 	return (IPPROTO_DONE);
2000 }
2001 
2002 static int
2003 ip6_mroute_modevent(module_t mod, int type, void *unused)
2004 {
2005 
2006 	switch (type) {
2007 	case MOD_LOAD:
2008 		MROUTER6_LOCK_INIT();
2009 		MFC6_LOCK_INIT();
2010 		MIF6_LOCK_INIT();
2011 
2012 		pim6_encap_cookie = encap_attach_func(AF_INET6, IPPROTO_PIM,
2013 			pim6_encapcheck,
2014 			(const struct protosw *)&in6_pim_protosw, NULL);
2015 		if (pim6_encap_cookie == NULL) {
2016 			printf("ip6_mroute: unable to attach pim6 encap\n");
2017 			MIF6_LOCK_DESTROY();
2018 			MFC6_LOCK_DESTROY();
2019 			MROUTER6_LOCK_DESTROY();
2020 			return (EINVAL);
2021 		}
2022 
2023 		ip6_mforward = X_ip6_mforward;
2024 		ip6_mrouter_done = X_ip6_mrouter_done;
2025 		ip6_mrouter_get = X_ip6_mrouter_get;
2026 		ip6_mrouter_set = X_ip6_mrouter_set;
2027 		mrt6_ioctl = X_mrt6_ioctl;
2028 		break;
2029 
2030 	case MOD_UNLOAD:
2031 		if (V_ip6_mrouter != NULL)
2032 			return EINVAL;
2033 
2034 		if (pim6_encap_cookie) {
2035 			encap_detach(pim6_encap_cookie);
2036 			pim6_encap_cookie = NULL;
2037 		}
2038 		X_ip6_mrouter_done();
2039 		ip6_mforward = NULL;
2040 		ip6_mrouter_done = NULL;
2041 		ip6_mrouter_get = NULL;
2042 		ip6_mrouter_set = NULL;
2043 		mrt6_ioctl = NULL;
2044 
2045 		MIF6_LOCK_DESTROY();
2046 		MFC6_LOCK_DESTROY();
2047 		MROUTER6_LOCK_DESTROY();
2048 		break;
2049 
2050 	default:
2051 		return (EOPNOTSUPP);
2052 	}
2053 
2054 	return (0);
2055 }
2056 
2057 static moduledata_t ip6_mroutemod = {
2058 	"ip6_mroute",
2059 	ip6_mroute_modevent,
2060 	0
2061 };
2062 
2063 DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
2064