xref: /freebsd/sys/netinet6/ip6_mroute.c (revision 6ef6ba9950260f42b47499d17874d00ca9290955)
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 #include "opt_kdtrace.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/sdt.h>
98 #include <sys/signalvar.h>
99 #include <sys/socket.h>
100 #include <sys/socketvar.h>
101 #include <sys/sockio.h>
102 #include <sys/sx.h>
103 #include <sys/sysctl.h>
104 #include <sys/syslog.h>
105 #include <sys/systm.h>
106 #include <sys/time.h>
107 
108 #include <net/if.h>
109 #include <net/if_var.h>
110 #include <net/if_types.h>
111 #include <net/raw_cb.h>
112 #include <net/vnet.h>
113 
114 #include <netinet/in.h>
115 #include <netinet/in_var.h>
116 #include <netinet/icmp6.h>
117 #include <netinet/ip_encap.h>
118 
119 #include <netinet/ip6.h>
120 #include <netinet/in_kdtrace.h>
121 #include <netinet6/ip6_var.h>
122 #include <netinet6/scope6_var.h>
123 #include <netinet6/nd6.h>
124 #include <netinet6/ip6_mroute.h>
125 #include <netinet6/ip6protosw.h>
126 #include <netinet6/pim6.h>
127 #include <netinet6/pim6_var.h>
128 
129 static MALLOC_DEFINE(M_MRTABLE6, "mf6c", "multicast forwarding cache entry");
130 
131 /* XXX: this is a very common idiom; move to <sys/mbuf.h> ? */
132 #define M_HASCL(m) ((m)->m_flags & M_EXT)
133 
134 static int	ip6_mdq(struct mbuf *, struct ifnet *, struct mf6c *);
135 static void	phyint_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
136 static int	register_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
137 static int	set_pim6(int *);
138 static int	socket_send(struct socket *, struct mbuf *,
139 		    struct sockaddr_in6 *);
140 
141 extern int in6_mcast_loop;
142 extern struct domain inet6domain;
143 
144 static const struct encaptab *pim6_encap_cookie;
145 static const struct ip6protosw in6_pim_protosw = {
146 	.pr_type =		SOCK_RAW,
147 	.pr_domain =		&inet6domain,
148 	.pr_protocol =		IPPROTO_PIM,
149 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
150 	.pr_input =		pim6_input,
151 	.pr_output =		rip6_output,
152 	.pr_ctloutput =		rip6_ctloutput,
153 	.pr_usrreqs =		&rip6_usrreqs
154 };
155 static int pim6_encapcheck(const struct mbuf *, int, int, void *);
156 
157 static VNET_DEFINE(int, ip6_mrouter_ver) = 0;
158 #define	V_ip6_mrouter_ver	VNET(ip6_mrouter_ver)
159 
160 SYSCTL_DECL(_net_inet6);
161 SYSCTL_DECL(_net_inet6_ip6);
162 static SYSCTL_NODE(_net_inet6, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
163 
164 static struct mrt6stat mrt6stat;
165 SYSCTL_STRUCT(_net_inet6_ip6, OID_AUTO, mrt6stat, CTLFLAG_RW,
166     &mrt6stat, mrt6stat,
167     "Multicast Routing Statistics (struct mrt6stat, netinet6/ip6_mroute.h)");
168 
169 #define	MRT6STAT_INC(name)	mrt6stat.name += 1
170 #define NO_RTE_FOUND	0x1
171 #define RTE_FOUND	0x2
172 
173 static struct mtx mrouter6_mtx;
174 #define	MROUTER6_LOCK()		mtx_lock(&mrouter6_mtx)
175 #define	MROUTER6_UNLOCK()	mtx_unlock(&mrouter6_mtx)
176 #define	MROUTER6_LOCK_ASSERT()	do {					\
177 	mtx_assert(&mrouter6_mtx, MA_OWNED);				\
178 	NET_ASSERT_GIANT();						\
179 } while (0)
180 #define	MROUTER6_LOCK_INIT()	\
181 	mtx_init(&mrouter6_mtx, "IPv6 multicast forwarding", NULL, MTX_DEF)
182 #define	MROUTER6_LOCK_DESTROY()	mtx_destroy(&mrouter6_mtx)
183 
184 static struct mf6c *mf6ctable[MF6CTBLSIZ];
185 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mf6ctable, CTLFLAG_RD,
186     &mf6ctable, sizeof(mf6ctable), "S,*mf6ctable[MF6CTBLSIZ]",
187     "IPv6 Multicast Forwarding Table (struct *mf6ctable[MF6CTBLSIZ], "
188     "netinet6/ip6_mroute.h)");
189 
190 static struct mtx mfc6_mtx;
191 #define	MFC6_LOCK()		mtx_lock(&mfc6_mtx)
192 #define	MFC6_UNLOCK()		mtx_unlock(&mfc6_mtx)
193 #define	MFC6_LOCK_ASSERT()	do {					\
194 	mtx_assert(&mfc6_mtx, MA_OWNED);				\
195 	NET_ASSERT_GIANT();						\
196 } while (0)
197 #define	MFC6_LOCK_INIT()		\
198 	mtx_init(&mfc6_mtx, "IPv6 multicast forwarding cache", NULL, MTX_DEF)
199 #define	MFC6_LOCK_DESTROY()	mtx_destroy(&mfc6_mtx)
200 
201 static u_char n6expire[MF6CTBLSIZ];
202 
203 static struct mif6 mif6table[MAXMIFS];
204 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mif6table, CTLFLAG_RD,
205     &mif6table, sizeof(mif6table), "S,mif6[MAXMIFS]",
206     "IPv6 Multicast Interfaces (struct mif6[MAXMIFS], netinet6/ip6_mroute.h)");
207 
208 static struct mtx mif6_mtx;
209 #define	MIF6_LOCK()		mtx_lock(&mif6_mtx)
210 #define	MIF6_UNLOCK()		mtx_unlock(&mif6_mtx)
211 #define	MIF6_LOCK_ASSERT()	mtx_assert(&mif6_mtx, MA_OWNED)
212 #define	MIF6_LOCK_INIT()	\
213 	mtx_init(&mif6_mtx, "IPv6 multicast interfaces", NULL, MTX_DEF)
214 #define	MIF6_LOCK_DESTROY()	mtx_destroy(&mif6_mtx)
215 
216 #ifdef MRT6DEBUG
217 static VNET_DEFINE(u_int, mrt6debug) = 0;	/* debug level */
218 #define	V_mrt6debug		VNET(mrt6debug)
219 #define DEBUG_MFC	0x02
220 #define DEBUG_FORWARD	0x04
221 #define DEBUG_EXPIRE	0x08
222 #define DEBUG_XMIT	0x10
223 #define DEBUG_REG	0x20
224 #define DEBUG_PIM	0x40
225 #endif
226 
227 static void	expire_upcalls(void *);
228 #define	EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
229 #define	UPCALL_EXPIRE	6		/* number of timeouts */
230 
231 /*
232  * XXX TODO: maintain a count to if_allmulti() calls in struct ifnet.
233  */
234 
235 /*
236  * 'Interfaces' associated with decapsulator (so we can tell
237  * packets that went through it from ones that get reflected
238  * by a broken gateway).  Different from IPv4 register_if,
239  * these interfaces are linked into the system ifnet list,
240  * because per-interface IPv6 statistics are maintained in
241  * ifp->if_afdata.  But it does not have any routes point
242  * to them.  I.e., packets can't be sent this way.  They
243  * only exist as a placeholder for multicast source
244  * verification.
245  */
246 static struct ifnet *multicast_register_if6;
247 
248 #define ENCAP_HOPS 64
249 
250 /*
251  * Private variables.
252  */
253 static mifi_t nummifs = 0;
254 static mifi_t reg_mif_num = (mifi_t)-1;
255 
256 static struct pim6stat pim6stat;
257 SYSCTL_STRUCT(_net_inet6_pim, PIM6CTL_STATS, stats, CTLFLAG_RW,
258     &pim6stat, pim6stat,
259     "PIM Statistics (struct pim6stat, netinet6/pim6_var.h)");
260 
261 #define	PIM6STAT_INC(name)	pim6stat.name += 1
262 static VNET_DEFINE(int, pim6);
263 #define	V_pim6		VNET(pim6)
264 
265 /*
266  * Hash function for a source, group entry
267  */
268 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
269 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
270 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
271 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
272 
273 /*
274  * Find a route for a given origin IPv6 address and Multicast group address.
275  */
276 #define MF6CFIND(o, g, rt) do { \
277 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
278 	rt = NULL; \
279 	MRT6STAT_INC(mrt6s_mfc_lookups); \
280 	while (_rt) { \
281 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
282 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
283 		    (_rt->mf6c_stall == NULL)) { \
284 			rt = _rt; \
285 			break; \
286 		} \
287 		_rt = _rt->mf6c_next; \
288 	} \
289 	if (rt == NULL) { \
290 		MRT6STAT_INC(mrt6s_mfc_misses); \
291 	} \
292 } while (/*CONSTCOND*/ 0)
293 
294 /*
295  * Macros to compute elapsed time efficiently
296  * Borrowed from Van Jacobson's scheduling code
297  * XXX: replace with timersub() ?
298  */
299 #define TV_DELTA(a, b, delta) do { \
300 	    int xxs; \
301 		\
302 	    delta = (a).tv_usec - (b).tv_usec; \
303 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
304 	       switch (xxs) { \
305 		      case 2: \
306 			  delta += 1000000; \
307 			      /* FALLTHROUGH */ \
308 		      case 1: \
309 			  delta += 1000000; \
310 			  break; \
311 		      default: \
312 			  delta += (1000000 * xxs); \
313 	       } \
314 	    } \
315 } while (/*CONSTCOND*/ 0)
316 
317 /* XXX: replace with timercmp(a, b, <) ? */
318 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
319 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
320 
321 #ifdef UPCALL_TIMING
322 #define UPCALL_MAX	50
323 static u_long upcall_data[UPCALL_MAX + 1];
324 static void collate();
325 #endif /* UPCALL_TIMING */
326 
327 static int ip6_mrouter_init(struct socket *, int, int);
328 static int add_m6fc(struct mf6cctl *);
329 static int add_m6if(struct mif6ctl *);
330 static int del_m6fc(struct mf6cctl *);
331 static int del_m6if(mifi_t *);
332 static int del_m6if_locked(mifi_t *);
333 static int get_mif6_cnt(struct sioc_mif_req6 *);
334 static int get_sg_cnt(struct sioc_sg_req6 *);
335 
336 static struct callout expire_upcalls_ch;
337 
338 int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *);
339 int X_ip6_mrouter_done(void);
340 int X_ip6_mrouter_set(struct socket *, struct sockopt *);
341 int X_ip6_mrouter_get(struct socket *, struct sockopt *);
342 int X_mrt6_ioctl(u_long, caddr_t);
343 
344 /*
345  * Handle MRT setsockopt commands to modify the multicast routing tables.
346  */
347 int
348 X_ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
349 {
350 	int error = 0;
351 	int optval;
352 	struct mif6ctl mifc;
353 	struct mf6cctl mfcc;
354 	mifi_t mifi;
355 
356 	if (so != V_ip6_mrouter && sopt->sopt_name != MRT6_INIT)
357 		return (EACCES);
358 
359 	switch (sopt->sopt_name) {
360 	case MRT6_INIT:
361 #ifdef MRT6_OINIT
362 	case MRT6_OINIT:
363 #endif
364 		error = sooptcopyin(sopt, &optval, sizeof(optval),
365 		    sizeof(optval));
366 		if (error)
367 			break;
368 		error = ip6_mrouter_init(so, optval, sopt->sopt_name);
369 		break;
370 	case MRT6_DONE:
371 		error = X_ip6_mrouter_done();
372 		break;
373 	case MRT6_ADD_MIF:
374 		error = sooptcopyin(sopt, &mifc, sizeof(mifc), sizeof(mifc));
375 		if (error)
376 			break;
377 		error = add_m6if(&mifc);
378 		break;
379 	case MRT6_ADD_MFC:
380 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
381 		if (error)
382 			break;
383 		error = add_m6fc(&mfcc);
384 		break;
385 	case MRT6_DEL_MFC:
386 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
387 		if (error)
388 			break;
389 		error = del_m6fc(&mfcc);
390 		break;
391 	case MRT6_DEL_MIF:
392 		error = sooptcopyin(sopt, &mifi, sizeof(mifi), sizeof(mifi));
393 		if (error)
394 			break;
395 		error = del_m6if(&mifi);
396 		break;
397 	case MRT6_PIM:
398 		error = sooptcopyin(sopt, &optval, sizeof(optval),
399 		    sizeof(optval));
400 		if (error)
401 			break;
402 		error = set_pim6(&optval);
403 		break;
404 	default:
405 		error = EOPNOTSUPP;
406 		break;
407 	}
408 
409 	return (error);
410 }
411 
412 /*
413  * Handle MRT getsockopt commands
414  */
415 int
416 X_ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
417 {
418 	int error = 0;
419 
420 	if (so != V_ip6_mrouter)
421 		return (EACCES);
422 
423 	switch (sopt->sopt_name) {
424 		case MRT6_PIM:
425 			error = sooptcopyout(sopt, &V_pim6, sizeof(V_pim6));
426 			break;
427 	}
428 	return (error);
429 }
430 
431 /*
432  * Handle ioctl commands to obtain information from the cache
433  */
434 int
435 X_mrt6_ioctl(u_long cmd, caddr_t data)
436 {
437 	int ret;
438 
439 	ret = EINVAL;
440 
441 	switch (cmd) {
442 	case SIOCGETSGCNT_IN6:
443 		ret = get_sg_cnt((struct sioc_sg_req6 *)data);
444 		break;
445 
446 	case SIOCGETMIFCNT_IN6:
447 		ret = get_mif6_cnt((struct sioc_mif_req6 *)data);
448 		break;
449 
450 	default:
451 		break;
452 	}
453 
454 	return (ret);
455 }
456 
457 /*
458  * returns the packet, byte, rpf-failure count for the source group provided
459  */
460 static int
461 get_sg_cnt(struct sioc_sg_req6 *req)
462 {
463 	struct mf6c *rt;
464 	int ret;
465 
466 	ret = 0;
467 
468 	MFC6_LOCK();
469 
470 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
471 	if (rt == NULL) {
472 		ret = ESRCH;
473 	} else {
474 		req->pktcnt = rt->mf6c_pkt_cnt;
475 		req->bytecnt = rt->mf6c_byte_cnt;
476 		req->wrong_if = rt->mf6c_wrong_if;
477 	}
478 
479 	MFC6_UNLOCK();
480 
481 	return (ret);
482 }
483 
484 /*
485  * returns the input and output packet and byte counts on the mif provided
486  */
487 static int
488 get_mif6_cnt(struct sioc_mif_req6 *req)
489 {
490 	mifi_t mifi;
491 	int ret;
492 
493 	ret = 0;
494 	mifi = req->mifi;
495 
496 	MIF6_LOCK();
497 
498 	if (mifi >= nummifs) {
499 		ret = EINVAL;
500 	} else {
501 		req->icount = mif6table[mifi].m6_pkt_in;
502 		req->ocount = mif6table[mifi].m6_pkt_out;
503 		req->ibytes = mif6table[mifi].m6_bytes_in;
504 		req->obytes = mif6table[mifi].m6_bytes_out;
505 	}
506 
507 	MIF6_UNLOCK();
508 
509 	return (ret);
510 }
511 
512 static int
513 set_pim6(int *i)
514 {
515 	if ((*i != 1) && (*i != 0))
516 		return (EINVAL);
517 
518 	V_pim6 = *i;
519 
520 	return (0);
521 }
522 
523 /*
524  * Enable multicast routing
525  */
526 static int
527 ip6_mrouter_init(struct socket *so, int v, int cmd)
528 {
529 
530 #ifdef MRT6DEBUG
531 	if (V_mrt6debug)
532 		log(LOG_DEBUG,
533 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
534 		    so->so_type, so->so_proto->pr_protocol);
535 #endif
536 
537 	if (so->so_type != SOCK_RAW ||
538 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
539 		return (EOPNOTSUPP);
540 
541 	if (v != 1)
542 		return (ENOPROTOOPT);
543 
544 	MROUTER6_LOCK();
545 
546 	if (V_ip6_mrouter != NULL) {
547 		MROUTER6_UNLOCK();
548 		return (EADDRINUSE);
549 	}
550 
551 	V_ip6_mrouter = so;
552 	V_ip6_mrouter_ver = cmd;
553 
554 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
555 	bzero((caddr_t)n6expire, sizeof(n6expire));
556 
557 	V_pim6 = 0;/* used for stubbing out/in pim stuff */
558 
559 	callout_init(&expire_upcalls_ch, 0);
560 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
561 	    expire_upcalls, NULL);
562 
563 	MROUTER6_UNLOCK();
564 
565 #ifdef MRT6DEBUG
566 	if (V_mrt6debug)
567 		log(LOG_DEBUG, "ip6_mrouter_init\n");
568 #endif
569 
570 	return (0);
571 }
572 
573 /*
574  * Disable IPv6 multicast forwarding.
575  */
576 int
577 X_ip6_mrouter_done(void)
578 {
579 	mifi_t mifi;
580 	u_long i;
581 	struct mf6c *rt;
582 	struct rtdetq *rte;
583 
584 	MROUTER6_LOCK();
585 
586 	if (V_ip6_mrouter == NULL) {
587 		MROUTER6_UNLOCK();
588 		return (EINVAL);
589 	}
590 
591 	/*
592 	 * For each phyint in use, disable promiscuous reception of all IPv6
593 	 * multicasts.
594 	 */
595 	for (mifi = 0; mifi < nummifs; mifi++) {
596 		if (mif6table[mifi].m6_ifp &&
597 		    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
598 			if_allmulti(mif6table[mifi].m6_ifp, 0);
599 		}
600 	}
601 	bzero((caddr_t)mif6table, sizeof(mif6table));
602 	nummifs = 0;
603 
604 	V_pim6 = 0; /* used to stub out/in pim specific code */
605 
606 	callout_stop(&expire_upcalls_ch);
607 
608 	/*
609 	 * Free all multicast forwarding cache entries.
610 	 */
611 	MFC6_LOCK();
612 	for (i = 0; i < MF6CTBLSIZ; i++) {
613 		rt = mf6ctable[i];
614 		while (rt) {
615 			struct mf6c *frt;
616 
617 			for (rte = rt->mf6c_stall; rte != NULL; ) {
618 				struct rtdetq *n = rte->next;
619 
620 				m_freem(rte->m);
621 				free(rte, M_MRTABLE6);
622 				rte = n;
623 			}
624 			frt = rt;
625 			rt = rt->mf6c_next;
626 			free(frt, M_MRTABLE6);
627 		}
628 	}
629 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
630 	MFC6_UNLOCK();
631 
632 	/*
633 	 * Reset register interface
634 	 */
635 	if (reg_mif_num != (mifi_t)-1 && multicast_register_if6 != NULL) {
636 		if_detach(multicast_register_if6);
637 		if_free(multicast_register_if6);
638 		reg_mif_num = (mifi_t)-1;
639 		multicast_register_if6 = NULL;
640 	}
641 
642 	V_ip6_mrouter = NULL;
643 	V_ip6_mrouter_ver = 0;
644 
645 	MROUTER6_UNLOCK();
646 
647 #ifdef MRT6DEBUG
648 	if (V_mrt6debug)
649 		log(LOG_DEBUG, "ip6_mrouter_done\n");
650 #endif
651 
652 	return (0);
653 }
654 
655 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
656 
657 /*
658  * Add a mif to the mif table
659  */
660 static int
661 add_m6if(struct mif6ctl *mifcp)
662 {
663 	struct mif6 *mifp;
664 	struct ifnet *ifp;
665 	int error;
666 
667 	MIF6_LOCK();
668 
669 	if (mifcp->mif6c_mifi >= MAXMIFS) {
670 		MIF6_UNLOCK();
671 		return (EINVAL);
672 	}
673 	mifp = mif6table + mifcp->mif6c_mifi;
674 	if (mifp->m6_ifp != NULL) {
675 		MIF6_UNLOCK();
676 		return (EADDRINUSE); /* XXX: is it appropriate? */
677 	}
678 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > V_if_index) {
679 		MIF6_UNLOCK();
680 		return (ENXIO);
681 	}
682 
683 	ifp = ifnet_byindex(mifcp->mif6c_pifi);
684 
685 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
686 		if (reg_mif_num == (mifi_t)-1) {
687 			ifp = if_alloc(IFT_OTHER);
688 
689 			if_initname(ifp, "register_mif", 0);
690 			ifp->if_flags |= IFF_LOOPBACK;
691 			if_attach(ifp);
692 			multicast_register_if6 = ifp;
693 			reg_mif_num = mifcp->mif6c_mifi;
694 			/*
695 			 * it is impossible to guess the ifindex of the
696 			 * register interface.  So mif6c_pifi is automatically
697 			 * calculated.
698 			 */
699 			mifcp->mif6c_pifi = ifp->if_index;
700 		} else {
701 			ifp = multicast_register_if6;
702 		}
703 	} else {
704 		/* Make sure the interface supports multicast */
705 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
706 			MIF6_UNLOCK();
707 			return (EOPNOTSUPP);
708 		}
709 
710 		error = if_allmulti(ifp, 1);
711 		if (error) {
712 			MIF6_UNLOCK();
713 			return (error);
714 		}
715 	}
716 
717 	mifp->m6_flags     = mifcp->mif6c_flags;
718 	mifp->m6_ifp       = ifp;
719 
720 	/* initialize per mif pkt counters */
721 	mifp->m6_pkt_in    = 0;
722 	mifp->m6_pkt_out   = 0;
723 	mifp->m6_bytes_in  = 0;
724 	mifp->m6_bytes_out = 0;
725 
726 	/* Adjust nummifs up if the mifi is higher than nummifs */
727 	if (nummifs <= mifcp->mif6c_mifi)
728 		nummifs = mifcp->mif6c_mifi + 1;
729 
730 	MIF6_UNLOCK();
731 
732 #ifdef MRT6DEBUG
733 	if (V_mrt6debug)
734 		log(LOG_DEBUG,
735 		    "add_mif #%d, phyint %s\n",
736 		    mifcp->mif6c_mifi,
737 		    ifp->if_xname);
738 #endif
739 
740 	return (0);
741 }
742 
743 /*
744  * Delete a mif from the mif table
745  */
746 static int
747 del_m6if_locked(mifi_t *mifip)
748 {
749 	struct mif6 *mifp = mif6table + *mifip;
750 	mifi_t mifi;
751 	struct ifnet *ifp;
752 
753 	MIF6_LOCK_ASSERT();
754 
755 	if (*mifip >= nummifs)
756 		return (EINVAL);
757 	if (mifp->m6_ifp == NULL)
758 		return (EINVAL);
759 
760 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
761 		/* XXX: TODO: Maintain an ALLMULTI refcount in struct ifnet. */
762 		ifp = mifp->m6_ifp;
763 		if_allmulti(ifp, 0);
764 	} else {
765 		if (reg_mif_num != (mifi_t)-1 &&
766 		    multicast_register_if6 != NULL) {
767 			if_detach(multicast_register_if6);
768 			if_free(multicast_register_if6);
769 			reg_mif_num = (mifi_t)-1;
770 			multicast_register_if6 = NULL;
771 		}
772 	}
773 
774 	bzero((caddr_t)mifp, sizeof(*mifp));
775 
776 	/* Adjust nummifs down */
777 	for (mifi = nummifs; mifi > 0; mifi--)
778 		if (mif6table[mifi - 1].m6_ifp)
779 			break;
780 	nummifs = mifi;
781 
782 #ifdef MRT6DEBUG
783 	if (V_mrt6debug)
784 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
785 #endif
786 
787 	return (0);
788 }
789 
790 static int
791 del_m6if(mifi_t *mifip)
792 {
793 	int cc;
794 
795 	MIF6_LOCK();
796 	cc = del_m6if_locked(mifip);
797 	MIF6_UNLOCK();
798 
799 	return (cc);
800 }
801 
802 /*
803  * Add an mfc entry
804  */
805 static int
806 add_m6fc(struct mf6cctl *mfccp)
807 {
808 	struct mf6c *rt;
809 	u_long hash;
810 	struct rtdetq *rte;
811 	u_short nstl;
812 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
813 
814 	MFC6_LOCK();
815 
816 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
817 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
818 
819 	/* If an entry already exists, just update the fields */
820 	if (rt) {
821 #ifdef MRT6DEBUG
822 		if (V_mrt6debug & DEBUG_MFC) {
823 		    log(LOG_DEBUG,
824 			"add_m6fc no upcall h %d o %s g %s p %x\n",
825 			ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
826 			ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
827 			mfccp->mf6cc_parent);
828 		}
829 #endif
830 
831 		rt->mf6c_parent = mfccp->mf6cc_parent;
832 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
833 
834 		MFC6_UNLOCK();
835 		return (0);
836 	}
837 
838 	/*
839 	 * Find the entry for which the upcall was made and update
840 	 */
841 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
842 			mfccp->mf6cc_mcastgrp.sin6_addr);
843 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
844 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
845 				       &mfccp->mf6cc_origin.sin6_addr) &&
846 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
847 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
848 		    (rt->mf6c_stall != NULL)) {
849 
850 			if (nstl++)
851 				log(LOG_ERR,
852 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
853 				    "multiple kernel entries",
854 				    ip6_sprintf(ip6bufo,
855 					    &mfccp->mf6cc_origin.sin6_addr),
856 				    ip6_sprintf(ip6bufg,
857 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
858 				    mfccp->mf6cc_parent, rt->mf6c_stall);
859 
860 #ifdef MRT6DEBUG
861 			if (V_mrt6debug & DEBUG_MFC)
862 				log(LOG_DEBUG,
863 				    "add_m6fc o %s g %s p %x dbg %x\n",
864 				    ip6_sprintf(ip6bufo,
865 					    &mfccp->mf6cc_origin.sin6_addr),
866 				    ip6_sprintf(ip6bufg,
867 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
868 				    mfccp->mf6cc_parent, rt->mf6c_stall);
869 #endif
870 
871 			rt->mf6c_origin     = mfccp->mf6cc_origin;
872 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
873 			rt->mf6c_parent     = mfccp->mf6cc_parent;
874 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
875 			/* initialize pkt counters per src-grp */
876 			rt->mf6c_pkt_cnt    = 0;
877 			rt->mf6c_byte_cnt   = 0;
878 			rt->mf6c_wrong_if   = 0;
879 
880 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
881 			n6expire[hash]--;
882 
883 			/* free packets Qed at the end of this entry */
884 			for (rte = rt->mf6c_stall; rte != NULL; ) {
885 				struct rtdetq *n = rte->next;
886 				ip6_mdq(rte->m, rte->ifp, rt);
887 				m_freem(rte->m);
888 #ifdef UPCALL_TIMING
889 				collate(&(rte->t));
890 #endif /* UPCALL_TIMING */
891 				free(rte, M_MRTABLE6);
892 				rte = n;
893 			}
894 			rt->mf6c_stall = NULL;
895 		}
896 	}
897 
898 	/*
899 	 * It is possible that an entry is being inserted without an upcall
900 	 */
901 	if (nstl == 0) {
902 #ifdef MRT6DEBUG
903 		if (V_mrt6debug & DEBUG_MFC)
904 		    log(LOG_DEBUG,
905 			"add_mfc no upcall h %d o %s g %s p %x\n",
906 			hash,
907 			ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
908 			ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
909 			mfccp->mf6cc_parent);
910 #endif
911 
912 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
913 
914 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
915 					       &mfccp->mf6cc_origin.sin6_addr)&&
916 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
917 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
918 
919 				rt->mf6c_origin     = mfccp->mf6cc_origin;
920 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
921 				rt->mf6c_parent     = mfccp->mf6cc_parent;
922 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
923 				/* initialize pkt counters per src-grp */
924 				rt->mf6c_pkt_cnt    = 0;
925 				rt->mf6c_byte_cnt   = 0;
926 				rt->mf6c_wrong_if   = 0;
927 
928 				if (rt->mf6c_expire)
929 					n6expire[hash]--;
930 				rt->mf6c_expire	   = 0;
931 			}
932 		}
933 		if (rt == NULL) {
934 			/* no upcall, so make a new entry */
935 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
936 						  M_NOWAIT);
937 			if (rt == NULL) {
938 				MFC6_UNLOCK();
939 				return (ENOBUFS);
940 			}
941 
942 			/* insert new entry at head of hash chain */
943 			rt->mf6c_origin     = mfccp->mf6cc_origin;
944 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
945 			rt->mf6c_parent     = mfccp->mf6cc_parent;
946 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
947 			/* initialize pkt counters per src-grp */
948 			rt->mf6c_pkt_cnt    = 0;
949 			rt->mf6c_byte_cnt   = 0;
950 			rt->mf6c_wrong_if   = 0;
951 			rt->mf6c_expire     = 0;
952 			rt->mf6c_stall = NULL;
953 
954 			/* link into table */
955 			rt->mf6c_next  = mf6ctable[hash];
956 			mf6ctable[hash] = rt;
957 		}
958 	}
959 
960 	MFC6_UNLOCK();
961 	return (0);
962 }
963 
964 #ifdef UPCALL_TIMING
965 /*
966  * collect delay statistics on the upcalls
967  */
968 static void
969 collate(struct timeval *t)
970 {
971 	u_long d;
972 	struct timeval tp;
973 	u_long delta;
974 
975 	GET_TIME(tp);
976 
977 	if (TV_LT(*t, tp))
978 	{
979 		TV_DELTA(tp, *t, delta);
980 
981 		d = delta >> 10;
982 		if (d > UPCALL_MAX)
983 			d = UPCALL_MAX;
984 
985 		++upcall_data[d];
986 	}
987 }
988 #endif /* UPCALL_TIMING */
989 
990 /*
991  * Delete an mfc entry
992  */
993 static int
994 del_m6fc(struct mf6cctl *mfccp)
995 {
996 	struct sockaddr_in6	origin;
997 	struct sockaddr_in6	mcastgrp;
998 	struct mf6c		*rt;
999 	struct mf6c		**nptr;
1000 	u_long		hash;
1001 
1002 	origin = mfccp->mf6cc_origin;
1003 	mcastgrp = mfccp->mf6cc_mcastgrp;
1004 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
1005 
1006 #ifdef MRT6DEBUG
1007 	if (V_mrt6debug & DEBUG_MFC) {
1008 		char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1009 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
1010 		    ip6_sprintf(ip6bufo, &origin.sin6_addr),
1011 		    ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr));
1012 	}
1013 #endif
1014 
1015 	MFC6_LOCK();
1016 
1017 	nptr = &mf6ctable[hash];
1018 	while ((rt = *nptr) != NULL) {
1019 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
1020 				       &rt->mf6c_origin.sin6_addr) &&
1021 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
1022 				       &rt->mf6c_mcastgrp.sin6_addr) &&
1023 		    rt->mf6c_stall == NULL)
1024 			break;
1025 
1026 		nptr = &rt->mf6c_next;
1027 	}
1028 	if (rt == NULL) {
1029 		MFC6_UNLOCK();
1030 		return (EADDRNOTAVAIL);
1031 	}
1032 
1033 	*nptr = rt->mf6c_next;
1034 	free(rt, M_MRTABLE6);
1035 
1036 	MFC6_UNLOCK();
1037 
1038 	return (0);
1039 }
1040 
1041 static int
1042 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
1043 {
1044 
1045 	if (s) {
1046 		if (sbappendaddr(&s->so_rcv,
1047 				 (struct sockaddr *)src,
1048 				 mm, (struct mbuf *)0) != 0) {
1049 			sorwakeup(s);
1050 			return (0);
1051 		}
1052 	}
1053 	m_freem(mm);
1054 	return (-1);
1055 }
1056 
1057 /*
1058  * IPv6 multicast forwarding function. This function assumes that the packet
1059  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
1060  * pointed to by "ifp", and the packet is to be relayed to other networks
1061  * that have members of the packet's destination IPv6 multicast group.
1062  *
1063  * The packet is returned unscathed to the caller, unless it is
1064  * erroneous, in which case a non-zero return value tells the caller to
1065  * discard it.
1066  *
1067  * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff
1068  * this function is called in the originating context (i.e., not when
1069  * forwarding a packet from other node).  ip6_output(), which is currently the
1070  * only function that calls this function is called in the originating context,
1071  * explicitly ensures this condition.  It is caller's responsibility to ensure
1072  * that if this function is called from somewhere else in the originating
1073  * context in the future.
1074  */
1075 int
1076 X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
1077 {
1078 	struct mf6c *rt;
1079 	struct mif6 *mifp;
1080 	struct mbuf *mm;
1081 	mifi_t mifi;
1082 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1083 
1084 #ifdef MRT6DEBUG
1085 	if (V_mrt6debug & DEBUG_FORWARD)
1086 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
1087 		    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1088 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1089 		    ifp->if_index);
1090 #endif
1091 
1092 	/*
1093 	 * Don't forward a packet with Hop limit of zero or one,
1094 	 * or a packet destined to a local-only group.
1095 	 */
1096 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
1097 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1098 		return (0);
1099 	ip6->ip6_hlim--;
1100 
1101 	/*
1102 	 * Source address check: do not forward packets with unspecified
1103 	 * source. It was discussed in July 2000, on ipngwg mailing list.
1104 	 * This is rather more serious than unicast cases, because some
1105 	 * MLD packets can be sent with the unspecified source address
1106 	 * (although such packets must normally set 1 to the hop limit field).
1107 	 */
1108 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1109 		IP6STAT_INC(ip6s_cantforward);
1110 		if (V_ip6_log_time + V_ip6_log_interval < time_uptime) {
1111 			V_ip6_log_time = time_uptime;
1112 			log(LOG_DEBUG,
1113 			    "cannot forward "
1114 			    "from %s to %s nxt %d received on %s\n",
1115 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1116 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1117 			    ip6->ip6_nxt,
1118 			    if_name(m->m_pkthdr.rcvif));
1119 		}
1120 		return (0);
1121 	}
1122 
1123 	MFC6_LOCK();
1124 
1125 	/*
1126 	 * Determine forwarding mifs from the forwarding cache table
1127 	 */
1128 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
1129 
1130 	/* Entry exists, so forward if necessary */
1131 	if (rt) {
1132 		MFC6_UNLOCK();
1133 		return (ip6_mdq(m, ifp, rt));
1134 	} else {
1135 		/*
1136 		 * If we don't have a route for packet's origin,
1137 		 * Make a copy of the packet &
1138 		 * send message to routing daemon
1139 		 */
1140 
1141 		struct mbuf *mb0;
1142 		struct rtdetq *rte;
1143 		u_long hash;
1144 /*		int i, npkts;*/
1145 #ifdef UPCALL_TIMING
1146 		struct timeval tp;
1147 
1148 		GET_TIME(tp);
1149 #endif /* UPCALL_TIMING */
1150 
1151 		MRT6STAT_INC(mrt6s_no_route);
1152 #ifdef MRT6DEBUG
1153 		if (V_mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
1154 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
1155 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1156 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1157 #endif
1158 
1159 		/*
1160 		 * Allocate mbufs early so that we don't do extra work if we
1161 		 * are just going to fail anyway.
1162 		 */
1163 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE6,
1164 					      M_NOWAIT);
1165 		if (rte == NULL) {
1166 			MFC6_UNLOCK();
1167 			return (ENOBUFS);
1168 		}
1169 		mb0 = m_copy(m, 0, M_COPYALL);
1170 		/*
1171 		 * Pullup packet header if needed before storing it,
1172 		 * as other references may modify it in the meantime.
1173 		 */
1174 		if (mb0 &&
1175 		    (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1176 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1177 		if (mb0 == NULL) {
1178 			free(rte, M_MRTABLE6);
1179 			MFC6_UNLOCK();
1180 			return (ENOBUFS);
1181 		}
1182 
1183 		/* is there an upcall waiting for this packet? */
1184 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1185 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1186 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1187 					       &rt->mf6c_origin.sin6_addr) &&
1188 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1189 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1190 			    (rt->mf6c_stall != NULL))
1191 				break;
1192 		}
1193 
1194 		if (rt == NULL) {
1195 			struct mrt6msg *im;
1196 #ifdef MRT6_OINIT
1197 			struct omrt6msg *oim;
1198 #endif
1199 
1200 			/* no upcall, so make a new entry */
1201 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
1202 						  M_NOWAIT);
1203 			if (rt == NULL) {
1204 				free(rte, M_MRTABLE6);
1205 				m_freem(mb0);
1206 				MFC6_UNLOCK();
1207 				return (ENOBUFS);
1208 			}
1209 			/*
1210 			 * Make a copy of the header to send to the user
1211 			 * level process
1212 			 */
1213 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1214 
1215 			if (mm == NULL) {
1216 				free(rte, M_MRTABLE6);
1217 				m_freem(mb0);
1218 				free(rt, M_MRTABLE6);
1219 				MFC6_UNLOCK();
1220 				return (ENOBUFS);
1221 			}
1222 
1223 			/*
1224 			 * Send message to routing daemon
1225 			 */
1226 			sin6.sin6_addr = ip6->ip6_src;
1227 
1228 			im = NULL;
1229 #ifdef MRT6_OINIT
1230 			oim = NULL;
1231 #endif
1232 			switch (V_ip6_mrouter_ver) {
1233 #ifdef MRT6_OINIT
1234 			case MRT6_OINIT:
1235 				oim = mtod(mm, struct omrt6msg *);
1236 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1237 				oim->im6_mbz = 0;
1238 				break;
1239 #endif
1240 			case MRT6_INIT:
1241 				im = mtod(mm, struct mrt6msg *);
1242 				im->im6_msgtype = MRT6MSG_NOCACHE;
1243 				im->im6_mbz = 0;
1244 				break;
1245 			default:
1246 				free(rte, M_MRTABLE6);
1247 				m_freem(mb0);
1248 				free(rt, M_MRTABLE6);
1249 				MFC6_UNLOCK();
1250 				return (EINVAL);
1251 			}
1252 
1253 #ifdef MRT6DEBUG
1254 			if (V_mrt6debug & DEBUG_FORWARD)
1255 				log(LOG_DEBUG,
1256 				    "getting the iif info in the kernel\n");
1257 #endif
1258 
1259 			for (mifp = mif6table, mifi = 0;
1260 			     mifi < nummifs && mifp->m6_ifp != ifp;
1261 			     mifp++, mifi++)
1262 				;
1263 
1264 			switch (V_ip6_mrouter_ver) {
1265 #ifdef MRT6_OINIT
1266 			case MRT6_OINIT:
1267 				oim->im6_mif = mifi;
1268 				break;
1269 #endif
1270 			case MRT6_INIT:
1271 				im->im6_mif = mifi;
1272 				break;
1273 			}
1274 
1275 			if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1276 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1277 				    "socket queue full\n");
1278 				MRT6STAT_INC(mrt6s_upq_sockfull);
1279 				free(rte, M_MRTABLE6);
1280 				m_freem(mb0);
1281 				free(rt, M_MRTABLE6);
1282 				MFC6_UNLOCK();
1283 				return (ENOBUFS);
1284 			}
1285 
1286 			MRT6STAT_INC(mrt6s_upcalls);
1287 
1288 			/* insert new entry at head of hash chain */
1289 			bzero(rt, sizeof(*rt));
1290 			rt->mf6c_origin.sin6_family = AF_INET6;
1291 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1292 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1293 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1294 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1295 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1296 			rt->mf6c_expire = UPCALL_EXPIRE;
1297 			n6expire[hash]++;
1298 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1299 
1300 			/* link into table */
1301 			rt->mf6c_next  = mf6ctable[hash];
1302 			mf6ctable[hash] = rt;
1303 			/* Add this entry to the end of the queue */
1304 			rt->mf6c_stall = rte;
1305 		} else {
1306 			/* determine if q has overflowed */
1307 			struct rtdetq **p;
1308 			int npkts = 0;
1309 
1310 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1311 				if (++npkts > MAX_UPQ6) {
1312 					MRT6STAT_INC(mrt6s_upq_ovflw);
1313 					free(rte, M_MRTABLE6);
1314 					m_freem(mb0);
1315 					MFC6_UNLOCK();
1316 					return (0);
1317 				}
1318 
1319 			/* Add this entry to the end of the queue */
1320 			*p = rte;
1321 		}
1322 
1323 		rte->next = NULL;
1324 		rte->m = mb0;
1325 		rte->ifp = ifp;
1326 #ifdef UPCALL_TIMING
1327 		rte->t = tp;
1328 #endif /* UPCALL_TIMING */
1329 
1330 		MFC6_UNLOCK();
1331 
1332 		return (0);
1333 	}
1334 }
1335 
1336 /*
1337  * Clean up cache entries if upcalls are not serviced
1338  * Call from the Slow Timeout mechanism, every half second.
1339  */
1340 static void
1341 expire_upcalls(void *unused)
1342 {
1343 	struct rtdetq *rte;
1344 	struct mf6c *mfc, **nptr;
1345 	u_long i;
1346 
1347 	MFC6_LOCK();
1348 	for (i = 0; i < MF6CTBLSIZ; i++) {
1349 		if (n6expire[i] == 0)
1350 			continue;
1351 		nptr = &mf6ctable[i];
1352 		while ((mfc = *nptr) != NULL) {
1353 			rte = mfc->mf6c_stall;
1354 			/*
1355 			 * Skip real cache entries
1356 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1357 			 * If it expires now
1358 			 */
1359 			if (rte != NULL &&
1360 			    mfc->mf6c_expire != 0 &&
1361 			    --mfc->mf6c_expire == 0) {
1362 #ifdef MRT6DEBUG
1363 				if (V_mrt6debug & DEBUG_EXPIRE) {
1364 					char ip6bufo[INET6_ADDRSTRLEN];
1365 					char ip6bufg[INET6_ADDRSTRLEN];
1366 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1367 					    ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr),
1368 					    ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr));
1369 				}
1370 #endif
1371 				/*
1372 				 * drop all the packets
1373 				 * free the mbuf with the pkt, if, timing info
1374 				 */
1375 				do {
1376 					struct rtdetq *n = rte->next;
1377 					m_freem(rte->m);
1378 					free(rte, M_MRTABLE6);
1379 					rte = n;
1380 				} while (rte != NULL);
1381 				MRT6STAT_INC(mrt6s_cache_cleanups);
1382 				n6expire[i]--;
1383 
1384 				*nptr = mfc->mf6c_next;
1385 				free(mfc, M_MRTABLE6);
1386 			} else {
1387 				nptr = &mfc->mf6c_next;
1388 			}
1389 		}
1390 	}
1391 	MFC6_UNLOCK();
1392 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1393 	    expire_upcalls, NULL);
1394 }
1395 
1396 /*
1397  * Packet forwarding routine once entry in the cache is made
1398  */
1399 static int
1400 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1401 {
1402 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1403 	mifi_t mifi, iif;
1404 	struct mif6 *mifp;
1405 	int plen = m->m_pkthdr.len;
1406 	struct in6_addr src0, dst0; /* copies for local work */
1407 	u_int32_t iszone, idzone, oszone, odzone;
1408 	int error = 0;
1409 
1410 /*
1411  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1412  * input, they shouldn't get counted on output, so statistics keeping is
1413  * separate.
1414  */
1415 
1416 #define MC6_SEND(ip6, mifp, m) do {				\
1417 	if ((mifp)->m6_flags & MIFF_REGISTER)			\
1418 		register_send((ip6), (mifp), (m));		\
1419 	else							\
1420 		phyint_send((ip6), (mifp), (m));		\
1421 } while (/*CONSTCOND*/ 0)
1422 
1423 	/*
1424 	 * Don't forward if it didn't arrive from the parent mif
1425 	 * for its origin.
1426 	 */
1427 	mifi = rt->mf6c_parent;
1428 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1429 		/* came in the wrong interface */
1430 #ifdef MRT6DEBUG
1431 		if (V_mrt6debug & DEBUG_FORWARD)
1432 			log(LOG_DEBUG,
1433 			    "wrong if: ifid %d mifi %d mififid %x\n",
1434 			    ifp->if_index, mifi,
1435 			    mif6table[mifi].m6_ifp->if_index);
1436 #endif
1437 		MRT6STAT_INC(mrt6s_wrong_if);
1438 		rt->mf6c_wrong_if++;
1439 		/*
1440 		 * If we are doing PIM processing, and we are forwarding
1441 		 * packets on this interface, send a message to the
1442 		 * routing daemon.
1443 		 */
1444 		/* have to make sure this is a valid mif */
1445 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1446 			if (V_pim6 && (m->m_flags & M_LOOP) == 0) {
1447 				/*
1448 				 * Check the M_LOOP flag to avoid an
1449 				 * unnecessary PIM assert.
1450 				 * XXX: M_LOOP is an ad-hoc hack...
1451 				 */
1452 				static struct sockaddr_in6 sin6 =
1453 				{ sizeof(sin6), AF_INET6 };
1454 
1455 				struct mbuf *mm;
1456 				struct mrt6msg *im;
1457 #ifdef MRT6_OINIT
1458 				struct omrt6msg *oim;
1459 #endif
1460 
1461 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1462 				if (mm &&
1463 				    (M_HASCL(mm) ||
1464 				     mm->m_len < sizeof(struct ip6_hdr)))
1465 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1466 				if (mm == NULL)
1467 					return (ENOBUFS);
1468 
1469 #ifdef MRT6_OINIT
1470 				oim = NULL;
1471 #endif
1472 				im = NULL;
1473 				switch (V_ip6_mrouter_ver) {
1474 #ifdef MRT6_OINIT
1475 				case MRT6_OINIT:
1476 					oim = mtod(mm, struct omrt6msg *);
1477 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1478 					oim->im6_mbz = 0;
1479 					break;
1480 #endif
1481 				case MRT6_INIT:
1482 					im = mtod(mm, struct mrt6msg *);
1483 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1484 					im->im6_mbz = 0;
1485 					break;
1486 				default:
1487 					m_freem(mm);
1488 					return (EINVAL);
1489 				}
1490 
1491 				for (mifp = mif6table, iif = 0;
1492 				     iif < nummifs && mifp &&
1493 					     mifp->m6_ifp != ifp;
1494 				     mifp++, iif++)
1495 					;
1496 
1497 				switch (V_ip6_mrouter_ver) {
1498 #ifdef MRT6_OINIT
1499 				case MRT6_OINIT:
1500 					oim->im6_mif = iif;
1501 					sin6.sin6_addr = oim->im6_src;
1502 					break;
1503 #endif
1504 				case MRT6_INIT:
1505 					im->im6_mif = iif;
1506 					sin6.sin6_addr = im->im6_src;
1507 					break;
1508 				}
1509 
1510 				MRT6STAT_INC(mrt6s_upcalls);
1511 
1512 				if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1513 #ifdef MRT6DEBUG
1514 					if (V_mrt6debug)
1515 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1516 #endif
1517 					MRT6STAT_INC(mrt6s_upq_sockfull);
1518 					return (ENOBUFS);
1519 				}	/* if socket Q full */
1520 			}		/* if PIM */
1521 		return (0);
1522 	}			/* if wrong iif */
1523 
1524 	/* If I sourced this packet, it counts as output, else it was input. */
1525 	if (m->m_pkthdr.rcvif == NULL) {
1526 		/* XXX: is rcvif really NULL when output?? */
1527 		mif6table[mifi].m6_pkt_out++;
1528 		mif6table[mifi].m6_bytes_out += plen;
1529 	} else {
1530 		mif6table[mifi].m6_pkt_in++;
1531 		mif6table[mifi].m6_bytes_in += plen;
1532 	}
1533 	rt->mf6c_pkt_cnt++;
1534 	rt->mf6c_byte_cnt += plen;
1535 
1536 	/*
1537 	 * For each mif, forward a copy of the packet if there are group
1538 	 * members downstream on the interface.
1539 	 */
1540 	src0 = ip6->ip6_src;
1541 	dst0 = ip6->ip6_dst;
1542 	if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 ||
1543 	    (error = in6_setscope(&dst0, ifp, &idzone)) != 0) {
1544 		IP6STAT_INC(ip6s_badscope);
1545 		return (error);
1546 	}
1547 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) {
1548 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1549 			/*
1550 			 * check if the outgoing packet is going to break
1551 			 * a scope boundary.
1552 			 * XXX For packets through PIM register tunnel
1553 			 * interface, we believe a routing daemon.
1554 			 */
1555 			if (!(mif6table[rt->mf6c_parent].m6_flags &
1556 			      MIFF_REGISTER) &&
1557 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
1558 				if (in6_setscope(&src0, mif6table[mifi].m6_ifp,
1559 				    &oszone) ||
1560 				    in6_setscope(&dst0, mif6table[mifi].m6_ifp,
1561 				    &odzone) ||
1562 				    iszone != oszone ||
1563 				    idzone != odzone) {
1564 					IP6STAT_INC(ip6s_badscope);
1565 					continue;
1566 				}
1567 			}
1568 
1569 			mifp->m6_pkt_out++;
1570 			mifp->m6_bytes_out += plen;
1571 			MC6_SEND(ip6, mifp, m);
1572 		}
1573 	}
1574 	return (0);
1575 }
1576 
1577 static void
1578 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1579 {
1580 	struct mbuf *mb_copy;
1581 	struct ifnet *ifp = mifp->m6_ifp;
1582 	int error = 0;
1583 	u_long linkmtu;
1584 
1585 	/*
1586 	 * Make a new reference to the packet; make sure that
1587 	 * the IPv6 header is actually copied, not just referenced,
1588 	 * so that ip6_output() only scribbles on the copy.
1589 	 */
1590 	mb_copy = m_copy(m, 0, M_COPYALL);
1591 	if (mb_copy &&
1592 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1593 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1594 	if (mb_copy == NULL) {
1595 		return;
1596 	}
1597 	/* set MCAST flag to the outgoing packet */
1598 	mb_copy->m_flags |= M_MCAST;
1599 
1600 	/*
1601 	 * If we sourced the packet, call ip6_output since we may devide
1602 	 * the packet into fragments when the packet is too big for the
1603 	 * outgoing interface.
1604 	 * Otherwise, we can simply send the packet to the interface
1605 	 * sending queue.
1606 	 */
1607 	if (m->m_pkthdr.rcvif == NULL) {
1608 		struct ip6_moptions im6o;
1609 
1610 		im6o.im6o_multicast_ifp = ifp;
1611 		/* XXX: ip6_output will override ip6->ip6_hlim */
1612 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1613 		im6o.im6o_multicast_loop = 1;
1614 		error = ip6_output(mb_copy, NULL, NULL, IPV6_FORWARDING, &im6o,
1615 		    NULL, NULL);
1616 
1617 #ifdef MRT6DEBUG
1618 		if (V_mrt6debug & DEBUG_XMIT)
1619 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1620 			    mifp - mif6table, error);
1621 #endif
1622 		return;
1623 	}
1624 
1625 	/*
1626 	 * If configured to loop back multicasts by default,
1627 	 * loop back a copy now.
1628 	 */
1629 	if (in6_mcast_loop) {
1630 		struct sockaddr_in6 dst6;
1631 
1632 		bzero(&dst6, sizeof(dst6));
1633 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1634 		dst6.sin6_family = AF_INET6;
1635 		dst6.sin6_addr = ip6->ip6_dst;
1636 		ip6_mloopback(ifp, m, &dst6);
1637 	}
1638 
1639 	/*
1640 	 * Put the packet into the sending queue of the outgoing interface
1641 	 * if it would fit in the MTU of the interface.
1642 	 */
1643 	linkmtu = IN6_LINKMTU(ifp);
1644 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1645 		struct sockaddr_in6 dst6;
1646 
1647 		bzero(&dst6, sizeof(dst6));
1648 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1649 		dst6.sin6_family = AF_INET6;
1650 		dst6.sin6_addr = ip6->ip6_dst;
1651 
1652 		IP_PROBE(send, NULL, NULL, ip6, ifp, NULL, ip6);
1653 		/*
1654 		 * We just call if_output instead of nd6_output here, since
1655 		 * we need no ND for a multicast forwarded packet...right?
1656 		 */
1657 		m_clrprotoflags(m);	/* Avoid confusing lower layers. */
1658 		error = (*ifp->if_output)(ifp, mb_copy,
1659 		    (struct sockaddr *)&dst6, NULL);
1660 #ifdef MRT6DEBUG
1661 		if (V_mrt6debug & DEBUG_XMIT)
1662 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1663 			    mifp - mif6table, error);
1664 #endif
1665 	} else {
1666 		/*
1667 		 * pMTU discovery is intentionally disabled by default, since
1668 		 * various router may notify pMTU in multicast, which can be
1669 		 * a DDoS to a router
1670 		 */
1671 		if (V_ip6_mcast_pmtu)
1672 			icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1673 		else {
1674 #ifdef MRT6DEBUG
1675 			if (V_mrt6debug & DEBUG_XMIT) {
1676 				char ip6bufs[INET6_ADDRSTRLEN];
1677 				char ip6bufd[INET6_ADDRSTRLEN];
1678 				log(LOG_DEBUG,
1679 				    "phyint_send: packet too big on %s o %s "
1680 				    "g %s size %d(discarded)\n",
1681 				    if_name(ifp),
1682 				    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1683 				    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1684 				    mb_copy->m_pkthdr.len);
1685 			}
1686 #endif /* MRT6DEBUG */
1687 			m_freem(mb_copy); /* simply discard the packet */
1688 		}
1689 	}
1690 }
1691 
1692 static int
1693 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1694 {
1695 	struct mbuf *mm;
1696 	int i, len = m->m_pkthdr.len;
1697 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1698 	struct mrt6msg *im6;
1699 
1700 #ifdef MRT6DEBUG
1701 	if (V_mrt6debug) {
1702 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1703 		log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1704 		    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1705 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1706 	}
1707 #endif
1708 	PIM6STAT_INC(pim6s_snd_registers);
1709 
1710 	/* Make a copy of the packet to send to the user level process. */
1711 	mm = m_gethdr(M_NOWAIT, MT_DATA);
1712 	if (mm == NULL)
1713 		return (ENOBUFS);
1714 	mm->m_data += max_linkhdr;
1715 	mm->m_len = sizeof(struct ip6_hdr);
1716 
1717 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1718 		m_freem(mm);
1719 		return (ENOBUFS);
1720 	}
1721 	i = MHLEN - M_LEADINGSPACE(mm);
1722 	if (i > len)
1723 		i = len;
1724 	mm = m_pullup(mm, i);
1725 	if (mm == NULL)
1726 		return (ENOBUFS);
1727 /* TODO: check it! */
1728 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1729 
1730 	/*
1731 	 * Send message to routing daemon
1732 	 */
1733 	sin6.sin6_addr = ip6->ip6_src;
1734 
1735 	im6 = mtod(mm, struct mrt6msg *);
1736 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1737 	im6->im6_mbz          = 0;
1738 
1739 	im6->im6_mif = mif - mif6table;
1740 
1741 	/* iif info is not given for reg. encap.n */
1742 	MRT6STAT_INC(mrt6s_upcalls);
1743 
1744 	if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1745 #ifdef MRT6DEBUG
1746 		if (V_mrt6debug)
1747 			log(LOG_WARNING,
1748 			    "register_send: ip6_mrouter socket queue full\n");
1749 #endif
1750 		MRT6STAT_INC(mrt6s_upq_sockfull);
1751 		return (ENOBUFS);
1752 	}
1753 	return (0);
1754 }
1755 
1756 /*
1757  * pim6_encapcheck() is called by the encap6_input() path at runtime to
1758  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
1759  * into the kernel.
1760  */
1761 static int
1762 pim6_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
1763 {
1764 
1765 #ifdef DIAGNOSTIC
1766     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
1767 #endif
1768     if (proto != IPPROTO_PIM)
1769 	return 0;	/* not for us; reject the datagram. */
1770 
1771     return 64;		/* claim the datagram. */
1772 }
1773 
1774 /*
1775  * PIM sparse mode hook
1776  * Receives the pim control messages, and passes them up to the listening
1777  * socket, using rip6_input.
1778  * The only message processed is the REGISTER pim message; the pim header
1779  * is stripped off, and the inner packet is passed to register_mforward.
1780  */
1781 int
1782 pim6_input(struct mbuf **mp, int *offp, int proto)
1783 {
1784 	struct pim *pim; /* pointer to a pim struct */
1785 	struct ip6_hdr *ip6;
1786 	int pimlen;
1787 	struct mbuf *m = *mp;
1788 	int minlen;
1789 	int off = *offp;
1790 
1791 	PIM6STAT_INC(pim6s_rcv_total);
1792 
1793 	ip6 = mtod(m, struct ip6_hdr *);
1794 	pimlen = m->m_pkthdr.len - *offp;
1795 
1796 	/*
1797 	 * Validate lengths
1798 	 */
1799 	if (pimlen < PIM_MINLEN) {
1800 		PIM6STAT_INC(pim6s_rcv_tooshort);
1801 #ifdef MRT6DEBUG
1802 		if (V_mrt6debug & DEBUG_PIM)
1803 			log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1804 #endif
1805 		m_freem(m);
1806 		return (IPPROTO_DONE);
1807 	}
1808 
1809 	/*
1810 	 * if the packet is at least as big as a REGISTER, go ahead
1811 	 * and grab the PIM REGISTER header size, to avoid another
1812 	 * possible m_pullup() later.
1813 	 *
1814 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1815 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1816 	 */
1817 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1818 
1819 	/*
1820 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1821 	 * possibly the PIM REGISTER header
1822 	 */
1823 #ifndef PULLDOWN_TEST
1824 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1825 	/* adjust pointer */
1826 	ip6 = mtod(m, struct ip6_hdr *);
1827 
1828 	/* adjust mbuf to point to the PIM header */
1829 	pim = (struct pim *)((caddr_t)ip6 + off);
1830 #else
1831 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1832 	if (pim == NULL) {
1833 		PIM6STAT_INC(pim6s_rcv_tooshort);
1834 		return (IPPROTO_DONE);
1835 	}
1836 #endif
1837 
1838 #define PIM6_CHECKSUM
1839 #ifdef PIM6_CHECKSUM
1840 	{
1841 		int cksumlen;
1842 
1843 		/*
1844 		 * Validate checksum.
1845 		 * If PIM REGISTER, exclude the data packet
1846 		 */
1847 		if (pim->pim_type == PIM_REGISTER)
1848 			cksumlen = PIM_MINLEN;
1849 		else
1850 			cksumlen = pimlen;
1851 
1852 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1853 			PIM6STAT_INC(pim6s_rcv_badsum);
1854 #ifdef MRT6DEBUG
1855 			if (V_mrt6debug & DEBUG_PIM)
1856 				log(LOG_DEBUG,
1857 				    "pim6_input: invalid checksum\n");
1858 #endif
1859 			m_freem(m);
1860 			return (IPPROTO_DONE);
1861 		}
1862 	}
1863 #endif /* PIM_CHECKSUM */
1864 
1865 	/* PIM version check */
1866 	if (pim->pim_ver != PIM_VERSION) {
1867 		PIM6STAT_INC(pim6s_rcv_badversion);
1868 #ifdef MRT6DEBUG
1869 		log(LOG_ERR,
1870 		    "pim6_input: incorrect version %d, expecting %d\n",
1871 		    pim->pim_ver, PIM_VERSION);
1872 #endif
1873 		m_freem(m);
1874 		return (IPPROTO_DONE);
1875 	}
1876 
1877 	if (pim->pim_type == PIM_REGISTER) {
1878 		/*
1879 		 * since this is a REGISTER, we'll make a copy of the register
1880 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1881 		 * routing daemon.
1882 		 */
1883 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1884 
1885 		struct mbuf *mcp;
1886 		struct ip6_hdr *eip6;
1887 		u_int32_t *reghdr;
1888 		int rc;
1889 #ifdef MRT6DEBUG
1890 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1891 #endif
1892 
1893 		PIM6STAT_INC(pim6s_rcv_registers);
1894 
1895 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1896 #ifdef MRT6DEBUG
1897 			if (V_mrt6debug & DEBUG_PIM)
1898 				log(LOG_DEBUG,
1899 				    "pim6_input: register mif not set: %d\n",
1900 				    reg_mif_num);
1901 #endif
1902 			m_freem(m);
1903 			return (IPPROTO_DONE);
1904 		}
1905 
1906 		reghdr = (u_int32_t *)(pim + 1);
1907 
1908 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1909 			goto pim6_input_to_daemon;
1910 
1911 		/*
1912 		 * Validate length
1913 		 */
1914 		if (pimlen < PIM6_REG_MINLEN) {
1915 			PIM6STAT_INC(pim6s_rcv_tooshort);
1916 			PIM6STAT_INC(pim6s_rcv_badregisters);
1917 #ifdef MRT6DEBUG
1918 			log(LOG_ERR,
1919 			    "pim6_input: register packet size too "
1920 			    "small %d from %s\n",
1921 			    pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src));
1922 #endif
1923 			m_freem(m);
1924 			return (IPPROTO_DONE);
1925 		}
1926 
1927 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1928 #ifdef MRT6DEBUG
1929 		if (V_mrt6debug & DEBUG_PIM)
1930 			log(LOG_DEBUG,
1931 			    "pim6_input[register], eip6: %s -> %s, "
1932 			    "eip6 plen %d\n",
1933 			    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1934 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1935 			    ntohs(eip6->ip6_plen));
1936 #endif
1937 
1938 		/* verify the version number of the inner packet */
1939 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1940 			PIM6STAT_INC(pim6s_rcv_badregisters);
1941 #ifdef MRT6DEBUG
1942 			log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1943 			    "of the inner packet\n",
1944 			    (eip6->ip6_vfc & IPV6_VERSION));
1945 #endif
1946 			m_freem(m);
1947 			return (IPPROTO_NONE);
1948 		}
1949 
1950 		/* verify the inner packet is destined to a mcast group */
1951 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1952 			PIM6STAT_INC(pim6s_rcv_badregisters);
1953 #ifdef MRT6DEBUG
1954 			if (V_mrt6debug & DEBUG_PIM)
1955 				log(LOG_DEBUG,
1956 				    "pim6_input: inner packet of register "
1957 				    "is not multicast %s\n",
1958 				    ip6_sprintf(ip6bufd, &eip6->ip6_dst));
1959 #endif
1960 			m_freem(m);
1961 			return (IPPROTO_DONE);
1962 		}
1963 
1964 		/*
1965 		 * make a copy of the whole header to pass to the daemon later.
1966 		 */
1967 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1968 		if (mcp == NULL) {
1969 #ifdef MRT6DEBUG
1970 			log(LOG_ERR,
1971 			    "pim6_input: pim register: "
1972 			    "could not copy register head\n");
1973 #endif
1974 			m_freem(m);
1975 			return (IPPROTO_DONE);
1976 		}
1977 
1978 		/*
1979 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1980 		 */
1981 		m_adj(m, off + PIM_MINLEN);
1982 #ifdef MRT6DEBUG
1983 		if (V_mrt6debug & DEBUG_PIM) {
1984 			log(LOG_DEBUG,
1985 			    "pim6_input: forwarding decapsulated register: "
1986 			    "src %s, dst %s, mif %d\n",
1987 			    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1988 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1989 			    reg_mif_num);
1990 		}
1991 #endif
1992 
1993 		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1994 				dst.sin6_family, 0);
1995 
1996 		/* prepare the register head to send to the mrouting daemon */
1997 		m = mcp;
1998 	}
1999 
2000 	/*
2001 	 * Pass the PIM message up to the daemon; if it is a register message
2002 	 * pass the 'head' only up to the daemon. This includes the
2003 	 * encapsulator ip6 header, pim header, register header and the
2004 	 * encapsulated ip6 header.
2005 	 */
2006   pim6_input_to_daemon:
2007 	rip6_input(&m, offp, proto);
2008 	return (IPPROTO_DONE);
2009 }
2010 
2011 static int
2012 ip6_mroute_modevent(module_t mod, int type, void *unused)
2013 {
2014 
2015 	switch (type) {
2016 	case MOD_LOAD:
2017 		MROUTER6_LOCK_INIT();
2018 		MFC6_LOCK_INIT();
2019 		MIF6_LOCK_INIT();
2020 
2021 		pim6_encap_cookie = encap_attach_func(AF_INET6, IPPROTO_PIM,
2022 			pim6_encapcheck,
2023 			(const struct protosw *)&in6_pim_protosw, NULL);
2024 		if (pim6_encap_cookie == NULL) {
2025 			printf("ip6_mroute: unable to attach pim6 encap\n");
2026 			MIF6_LOCK_DESTROY();
2027 			MFC6_LOCK_DESTROY();
2028 			MROUTER6_LOCK_DESTROY();
2029 			return (EINVAL);
2030 		}
2031 
2032 		ip6_mforward = X_ip6_mforward;
2033 		ip6_mrouter_done = X_ip6_mrouter_done;
2034 		ip6_mrouter_get = X_ip6_mrouter_get;
2035 		ip6_mrouter_set = X_ip6_mrouter_set;
2036 		mrt6_ioctl = X_mrt6_ioctl;
2037 		break;
2038 
2039 	case MOD_UNLOAD:
2040 		if (V_ip6_mrouter != NULL)
2041 			return EINVAL;
2042 
2043 		if (pim6_encap_cookie) {
2044 			encap_detach(pim6_encap_cookie);
2045 			pim6_encap_cookie = NULL;
2046 		}
2047 		X_ip6_mrouter_done();
2048 		ip6_mforward = NULL;
2049 		ip6_mrouter_done = NULL;
2050 		ip6_mrouter_get = NULL;
2051 		ip6_mrouter_set = NULL;
2052 		mrt6_ioctl = NULL;
2053 
2054 		MIF6_LOCK_DESTROY();
2055 		MFC6_LOCK_DESTROY();
2056 		MROUTER6_LOCK_DESTROY();
2057 		break;
2058 
2059 	default:
2060 		return (EOPNOTSUPP);
2061 	}
2062 
2063 	return (0);
2064 }
2065 
2066 static moduledata_t ip6_mroutemod = {
2067 	"ip6_mroute",
2068 	ip6_mroute_modevent,
2069 	0
2070 };
2071 
2072 DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
2073