xref: /freebsd/sys/netinet6/ip6_mroute.c (revision 66df505066f51e6d8411b966765d828817f88971)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1989 Stephen Deering
36  * Copyright (c) 1992, 1993
37  *      The Regents of the University of California.  All rights reserved.
38  *
39  * This code is derived from software contributed to Berkeley by
40  * Stephen Deering of Stanford University.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 3. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)ip_mroute.c	8.2 (Berkeley) 11/15/93
67  *	BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp
68  */
69 
70 /*
71  * IP multicast forwarding procedures
72  *
73  * Written by David Waitzman, BBN Labs, August 1988.
74  * Modified by Steve Deering, Stanford, February 1989.
75  * Modified by Mark J. Steiglitz, Stanford, May, 1991
76  * Modified by Van Jacobson, LBL, January 1993
77  * Modified by Ajit Thyagarajan, PARC, August 1993
78  * Modified by Bill Fenner, PARC, April 1994
79  *
80  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
81  */
82 
83 #include <sys/cdefs.h>
84 __FBSDID("$FreeBSD$");
85 
86 #include "opt_inet6.h"
87 
88 #include <sys/param.h>
89 #include <sys/callout.h>
90 #include <sys/errno.h>
91 #include <sys/kernel.h>
92 #include <sys/lock.h>
93 #include <sys/malloc.h>
94 #include <sys/mbuf.h>
95 #include <sys/module.h>
96 #include <sys/domain.h>
97 #include <sys/protosw.h>
98 #include <sys/sdt.h>
99 #include <sys/signalvar.h>
100 #include <sys/socket.h>
101 #include <sys/socketvar.h>
102 #include <sys/sockio.h>
103 #include <sys/sx.h>
104 #include <sys/sysctl.h>
105 #include <sys/syslog.h>
106 #include <sys/systm.h>
107 #include <sys/time.h>
108 
109 #include <net/if.h>
110 #include <net/if_var.h>
111 #include <net/if_types.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/pim6.h>
126 #include <netinet6/pim6_var.h>
127 
128 static MALLOC_DEFINE(M_MRTABLE6, "mf6c", "multicast forwarding cache entry");
129 
130 static int	ip6_mdq(struct mbuf *, struct ifnet *, struct mf6c *);
131 static void	phyint_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
132 static int	register_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
133 static int	set_pim6(int *);
134 static int	socket_send(struct socket *, struct mbuf *,
135 		    struct sockaddr_in6 *);
136 
137 extern int in6_mcast_loop;
138 extern struct domain inet6domain;
139 
140 static const struct encaptab *pim6_encap_cookie;
141 static int pim6_encapcheck(const struct mbuf *, int, int, void *);
142 static int pim6_input(struct mbuf *, int, int, void *);
143 
144 static const struct encap_config ipv6_encap_cfg = {
145 	.proto = IPPROTO_PIM,
146 	.min_length = sizeof(struct ip6_hdr) + PIM_MINLEN,
147 	.exact_match = 8,
148 	.check = pim6_encapcheck,
149 	.input = pim6_input
150 };
151 
152 VNET_DEFINE_STATIC(int, ip6_mrouter_ver) = 0;
153 #define	V_ip6_mrouter_ver	VNET(ip6_mrouter_ver)
154 
155 SYSCTL_DECL(_net_inet6);
156 SYSCTL_DECL(_net_inet6_ip6);
157 static SYSCTL_NODE(_net_inet6, IPPROTO_PIM, pim,
158     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
159     "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	MRT6STAT_INC(name)	mrt6stat.name += 1
167 #define NO_RTE_FOUND	0x1
168 #define RTE_FOUND	0x2
169 
170 static struct sx mrouter6_mtx;
171 #define	MROUTER6_LOCKPTR()	(&mrouter6_mtx)
172 #define	MROUTER6_LOCK()		sx_xlock(MROUTER6_LOCKPTR())
173 #define	MROUTER6_UNLOCK()	sx_xunlock(MROUTER6_LOCKPTR())
174 #define	MROUTER6_LOCK_ASSERT()	sx_assert(MROUTER6_LOCKPTR(), SA_XLOCKED
175 #define	MROUTER6_LOCK_INIT()	sx_init(MROUTER6_LOCKPTR(), "mrouter6")
176 #define	MROUTER6_LOCK_DESTROY()	sx_destroy(MROUTER6_LOCKPTR())
177 
178 static struct mf6c *mf6ctable[MF6CTBLSIZ];
179 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mf6ctable, CTLFLAG_RD,
180     &mf6ctable, sizeof(mf6ctable), "S,*mf6ctable[MF6CTBLSIZ]",
181     "IPv6 Multicast Forwarding Table (struct *mf6ctable[MF6CTBLSIZ], "
182     "netinet6/ip6_mroute.h)");
183 
184 static struct mtx mfc6_mtx;
185 #define	MFC6_LOCKPTR()		(&mfc6_mtx)
186 #define	MFC6_LOCK()		mtx_lock(MFC6_LOCKPTR())
187 #define	MFC6_UNLOCK()		mtx_unlock(MFC6_LOCKPTR())
188 #define	MFC6_LOCK_ASSERT()	mtx_assert(MFC6_LOCKPTR(), MA_OWNED)
189 #define	MFC6_LOCK_INIT()	mtx_init(MFC6_LOCKPTR(),		\
190 				    "IPv6 multicast forwarding cache",	\
191 				    NULL, MTX_DEF)
192 #define	MFC6_LOCK_DESTROY()	mtx_destroy(MFC6_LOCKPTR())
193 
194 static u_char n6expire[MF6CTBLSIZ];
195 
196 static struct mif6 mif6table[MAXMIFS];
197 static int
198 sysctl_mif6table(SYSCTL_HANDLER_ARGS)
199 {
200 	struct mif6_sctl *out;
201 	int error;
202 
203 	out = malloc(sizeof(struct mif6_sctl) * MAXMIFS, M_TEMP,
204 	    M_WAITOK | M_ZERO);
205 	for (int i = 0; i < MAXMIFS; i++) {
206 		out[i].m6_flags		= mif6table[i].m6_flags;
207 		out[i].m6_rate_limit	= mif6table[i].m6_rate_limit;
208 		out[i].m6_lcl_addr	= mif6table[i].m6_lcl_addr;
209 		if (mif6table[i].m6_ifp != NULL)
210 			out[i].m6_ifp	= mif6table[i].m6_ifp->if_index;
211 		else
212 			out[i].m6_ifp	= 0;
213 		out[i].m6_pkt_in	= mif6table[i].m6_pkt_in;
214 		out[i].m6_pkt_out	= mif6table[i].m6_pkt_out;
215 		out[i].m6_bytes_in	= mif6table[i].m6_bytes_in;
216 		out[i].m6_bytes_out	= mif6table[i].m6_bytes_out;
217 	}
218 	error = SYSCTL_OUT(req, out, sizeof(struct mif6_sctl) * MAXMIFS);
219 	free(out, M_TEMP);
220 	return (error);
221 }
222 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, mif6table,
223     CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
224     NULL, 0, sysctl_mif6table, "S,mif6_sctl[MAXMIFS]",
225     "IPv6 Multicast Interfaces (struct mif6_sctl[MAXMIFS], "
226     "netinet6/ip6_mroute.h)");
227 
228 static struct mtx mif6_mtx;
229 #define	MIF6_LOCKPTR()		(&mif6_mtx)
230 #define	MIF6_LOCK()		mtx_lock(MIF6_LOCKPTR())
231 #define	MIF6_UNLOCK()		mtx_unlock(MIF6_LOCKPTR())
232 #define	MIF6_LOCK_ASSERT()	mtx_assert(MIF6_LOCKPTR(), MA_OWNED)
233 #define	MIF6_LOCK_INIT()	\
234 	mtx_init(MIF6_LOCKPTR(), "IPv6 multicast interfaces", NULL, MTX_DEF)
235 #define	MIF6_LOCK_DESTROY()	mtx_destroy(MIF6_LOCKPTR())
236 
237 #ifdef MRT6DEBUG
238 VNET_DEFINE_STATIC(u_int, mrt6debug) = 0;	/* debug level */
239 #define	V_mrt6debug		VNET(mrt6debug)
240 #define DEBUG_MFC	0x02
241 #define DEBUG_FORWARD	0x04
242 #define DEBUG_EXPIRE	0x08
243 #define DEBUG_XMIT	0x10
244 #define DEBUG_REG	0x20
245 #define DEBUG_PIM	0x40
246 #define	DEBUG_ERR	0x80
247 #define	DEBUG_ANY	0x7f
248 #define	MRT6_DLOG(m, fmt, ...)	\
249 	if (V_mrt6debug & (m))	\
250 		log(((m) & DEBUG_ERR) ? LOG_ERR: LOG_DEBUG, \
251 		    "%s: " fmt "\n", __func__, ##__VA_ARGS__)
252 #else
253 #define	MRT6_DLOG(m, fmt, ...)
254 #endif
255 
256 static void	expire_upcalls(void *);
257 #define	EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
258 #define	UPCALL_EXPIRE	6		/* number of timeouts */
259 
260 /*
261  * XXX TODO: maintain a count to if_allmulti() calls in struct ifnet.
262  */
263 
264 /*
265  * 'Interfaces' associated with decapsulator (so we can tell
266  * packets that went through it from ones that get reflected
267  * by a broken gateway).  Different from IPv4 register_if,
268  * these interfaces are linked into the system ifnet list,
269  * because per-interface IPv6 statistics are maintained in
270  * ifp->if_afdata.  But it does not have any routes point
271  * to them.  I.e., packets can't be sent this way.  They
272  * only exist as a placeholder for multicast source
273  * verification.
274  */
275 static struct ifnet *multicast_register_if6;
276 
277 #define ENCAP_HOPS 64
278 
279 /*
280  * Private variables.
281  */
282 static mifi_t nummifs = 0;
283 static mifi_t reg_mif_num = (mifi_t)-1;
284 
285 static struct pim6stat pim6stat;
286 SYSCTL_STRUCT(_net_inet6_pim, PIM6CTL_STATS, stats, CTLFLAG_RW,
287     &pim6stat, pim6stat,
288     "PIM Statistics (struct pim6stat, netinet6/pim6_var.h)");
289 
290 #define	PIM6STAT_INC(name)	pim6stat.name += 1
291 VNET_DEFINE_STATIC(int, pim6);
292 #define	V_pim6		VNET(pim6)
293 
294 /*
295  * Hash function for a source, group entry
296  */
297 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
298 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
299 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
300 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
301 
302 /*
303  * Find a route for a given origin IPv6 address and Multicast group address.
304  */
305 #define MF6CFIND(o, g, rt) do { \
306 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
307 	rt = NULL; \
308 	while (_rt) { \
309 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
310 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
311 		    (_rt->mf6c_stall == NULL)) { \
312 			rt = _rt; \
313 			break; \
314 		} \
315 		_rt = _rt->mf6c_next; \
316 	} \
317 	if (rt == NULL) { \
318 		MRT6STAT_INC(mrt6s_mfc_misses); \
319 	} \
320 } while (/*CONSTCOND*/ 0)
321 
322 /*
323  * Macros to compute elapsed time efficiently
324  * Borrowed from Van Jacobson's scheduling code
325  * XXX: replace with timersub() ?
326  */
327 #define TV_DELTA(a, b, delta) do { \
328 	    int xxs; \
329 		\
330 	    delta = (a).tv_usec - (b).tv_usec; \
331 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
332 	       switch (xxs) { \
333 		      case 2: \
334 			  delta += 1000000; \
335 			      /* FALLTHROUGH */ \
336 		      case 1: \
337 			  delta += 1000000; \
338 			  break; \
339 		      default: \
340 			  delta += (1000000 * xxs); \
341 	       } \
342 	    } \
343 } while (/*CONSTCOND*/ 0)
344 
345 /* XXX: replace with timercmp(a, b, <) ? */
346 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
347 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
348 
349 #ifdef UPCALL_TIMING
350 #define UPCALL_MAX	50
351 static u_long upcall_data[UPCALL_MAX + 1];
352 static void collate();
353 #endif /* UPCALL_TIMING */
354 
355 static int ip6_mrouter_init(struct socket *, int, int);
356 static int add_m6fc(struct mf6cctl *);
357 static int add_m6if(struct mif6ctl *);
358 static int del_m6fc(struct mf6cctl *);
359 static int del_m6if(mifi_t *);
360 static int del_m6if_locked(mifi_t *);
361 static int get_mif6_cnt(struct sioc_mif_req6 *);
362 static int get_sg_cnt(struct sioc_sg_req6 *);
363 
364 static struct callout expire_upcalls_ch;
365 
366 int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *);
367 int X_ip6_mrouter_done(void);
368 int X_ip6_mrouter_set(struct socket *, struct sockopt *);
369 int X_ip6_mrouter_get(struct socket *, struct sockopt *);
370 int X_mrt6_ioctl(u_long, caddr_t);
371 
372 /*
373  * Handle MRT setsockopt commands to modify the multicast routing tables.
374  */
375 int
376 X_ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
377 {
378 	int error = 0;
379 	int optval;
380 	struct mif6ctl mifc;
381 	struct mf6cctl mfcc;
382 	mifi_t mifi;
383 
384 	if (so != V_ip6_mrouter && sopt->sopt_name != MRT6_INIT)
385 		return (EPERM);
386 
387 	switch (sopt->sopt_name) {
388 	case MRT6_INIT:
389 #ifdef MRT6_OINIT
390 	case MRT6_OINIT:
391 #endif
392 		error = sooptcopyin(sopt, &optval, sizeof(optval),
393 		    sizeof(optval));
394 		if (error)
395 			break;
396 		error = ip6_mrouter_init(so, optval, sopt->sopt_name);
397 		break;
398 	case MRT6_DONE:
399 		error = X_ip6_mrouter_done();
400 		break;
401 	case MRT6_ADD_MIF:
402 		error = sooptcopyin(sopt, &mifc, sizeof(mifc), sizeof(mifc));
403 		if (error)
404 			break;
405 		error = add_m6if(&mifc);
406 		break;
407 	case MRT6_ADD_MFC:
408 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
409 		if (error)
410 			break;
411 		error = add_m6fc(&mfcc);
412 		break;
413 	case MRT6_DEL_MFC:
414 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
415 		if (error)
416 			break;
417 		error = del_m6fc(&mfcc);
418 		break;
419 	case MRT6_DEL_MIF:
420 		error = sooptcopyin(sopt, &mifi, sizeof(mifi), sizeof(mifi));
421 		if (error)
422 			break;
423 		error = del_m6if(&mifi);
424 		break;
425 	case MRT6_PIM:
426 		error = sooptcopyin(sopt, &optval, sizeof(optval),
427 		    sizeof(optval));
428 		if (error)
429 			break;
430 		error = set_pim6(&optval);
431 		break;
432 	default:
433 		error = EOPNOTSUPP;
434 		break;
435 	}
436 
437 	return (error);
438 }
439 
440 /*
441  * Handle MRT getsockopt commands
442  */
443 int
444 X_ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
445 {
446 	int error = 0;
447 
448 	if (so != V_ip6_mrouter)
449 		return (EACCES);
450 
451 	switch (sopt->sopt_name) {
452 		case MRT6_PIM:
453 			error = sooptcopyout(sopt, &V_pim6, sizeof(V_pim6));
454 			break;
455 	}
456 	return (error);
457 }
458 
459 /*
460  * Handle ioctl commands to obtain information from the cache
461  */
462 int
463 X_mrt6_ioctl(u_long cmd, caddr_t data)
464 {
465 	int ret;
466 
467 	ret = EINVAL;
468 
469 	switch (cmd) {
470 	case SIOCGETSGCNT_IN6:
471 		ret = get_sg_cnt((struct sioc_sg_req6 *)data);
472 		break;
473 
474 	case SIOCGETMIFCNT_IN6:
475 		ret = get_mif6_cnt((struct sioc_mif_req6 *)data);
476 		break;
477 
478 	default:
479 		break;
480 	}
481 
482 	return (ret);
483 }
484 
485 /*
486  * returns the packet, byte, rpf-failure count for the source group provided
487  */
488 static int
489 get_sg_cnt(struct sioc_sg_req6 *req)
490 {
491 	struct mf6c *rt;
492 	int ret;
493 
494 	ret = 0;
495 
496 	MFC6_LOCK();
497 
498 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
499 	if (rt == NULL) {
500 		ret = ESRCH;
501 	} else {
502 		req->pktcnt = rt->mf6c_pkt_cnt;
503 		req->bytecnt = rt->mf6c_byte_cnt;
504 		req->wrong_if = rt->mf6c_wrong_if;
505 	}
506 
507 	MFC6_UNLOCK();
508 
509 	return (ret);
510 }
511 
512 /*
513  * returns the input and output packet and byte counts on the mif provided
514  */
515 static int
516 get_mif6_cnt(struct sioc_mif_req6 *req)
517 {
518 	mifi_t mifi;
519 	int ret;
520 
521 	ret = 0;
522 	mifi = req->mifi;
523 
524 	MIF6_LOCK();
525 
526 	if (mifi >= nummifs) {
527 		ret = EINVAL;
528 	} else {
529 		req->icount = mif6table[mifi].m6_pkt_in;
530 		req->ocount = mif6table[mifi].m6_pkt_out;
531 		req->ibytes = mif6table[mifi].m6_bytes_in;
532 		req->obytes = mif6table[mifi].m6_bytes_out;
533 	}
534 
535 	MIF6_UNLOCK();
536 
537 	return (ret);
538 }
539 
540 static int
541 set_pim6(int *i)
542 {
543 	if ((*i != 1) && (*i != 0))
544 		return (EINVAL);
545 
546 	V_pim6 = *i;
547 
548 	return (0);
549 }
550 
551 /*
552  * Enable multicast routing
553  */
554 static int
555 ip6_mrouter_init(struct socket *so, int v, int cmd)
556 {
557 
558 	MRT6_DLOG(DEBUG_ANY, "so_type = %d, pr_protocol = %d",
559 	    so->so_type, so->so_proto->pr_protocol);
560 
561 	if (so->so_type != SOCK_RAW ||
562 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
563 		return (EOPNOTSUPP);
564 
565 	if (v != 1)
566 		return (ENOPROTOOPT);
567 
568 	MROUTER6_LOCK();
569 
570 	if (V_ip6_mrouter != NULL) {
571 		MROUTER6_UNLOCK();
572 		return (EADDRINUSE);
573 	}
574 
575 	V_ip6_mrouter = so;
576 	V_ip6_mrouter_ver = cmd;
577 
578 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
579 	bzero((caddr_t)n6expire, sizeof(n6expire));
580 
581 	V_pim6 = 0;/* used for stubbing out/in pim stuff */
582 
583 	callout_init_mtx(&expire_upcalls_ch, MFC6_LOCKPTR(), 0);
584 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
585 	    expire_upcalls, NULL);
586 
587 	MROUTER6_UNLOCK();
588 
589 	MRT6_DLOG(DEBUG_ANY, "finished");
590 
591 	return (0);
592 }
593 
594 /*
595  * Disable IPv6 multicast forwarding.
596  */
597 int
598 X_ip6_mrouter_done(void)
599 {
600 	mifi_t mifi;
601 	u_long i;
602 	struct mf6c *rt;
603 	struct rtdetq *rte;
604 
605 	MROUTER6_LOCK();
606 
607 	if (V_ip6_mrouter == NULL) {
608 		MROUTER6_UNLOCK();
609 		return (EINVAL);
610 	}
611 
612 	/*
613 	 * For each phyint in use, disable promiscuous reception of all IPv6
614 	 * multicasts.
615 	 */
616 	for (mifi = 0; mifi < nummifs; mifi++) {
617 		if (mif6table[mifi].m6_ifp &&
618 		    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
619 			if_allmulti(mif6table[mifi].m6_ifp, 0);
620 		}
621 	}
622 	bzero((caddr_t)mif6table, sizeof(mif6table));
623 	nummifs = 0;
624 
625 	V_pim6 = 0; /* used to stub out/in pim specific code */
626 
627 	/*
628 	 * Free all multicast forwarding cache entries.
629 	 */
630 	MFC6_LOCK();
631 	for (i = 0; i < MF6CTBLSIZ; i++) {
632 		rt = mf6ctable[i];
633 		while (rt) {
634 			struct mf6c *frt;
635 
636 			for (rte = rt->mf6c_stall; rte != NULL; ) {
637 				struct rtdetq *n = rte->next;
638 
639 				m_freem(rte->m);
640 				free(rte, M_MRTABLE6);
641 				rte = n;
642 			}
643 			frt = rt;
644 			rt = rt->mf6c_next;
645 			free(frt, M_MRTABLE6);
646 		}
647 	}
648 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
649 	MFC6_UNLOCK();
650 
651 	callout_drain(&expire_upcalls_ch);
652 
653 	/*
654 	 * Reset register interface
655 	 */
656 	if (reg_mif_num != (mifi_t)-1 && multicast_register_if6 != NULL) {
657 		if_detach(multicast_register_if6);
658 		if_free(multicast_register_if6);
659 		reg_mif_num = (mifi_t)-1;
660 		multicast_register_if6 = NULL;
661 	}
662 
663 	V_ip6_mrouter = NULL;
664 	V_ip6_mrouter_ver = 0;
665 
666 	MROUTER6_UNLOCK();
667 	MRT6_DLOG(DEBUG_ANY, "finished");
668 
669 	return (0);
670 }
671 
672 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
673 
674 /*
675  * Add a mif to the mif table
676  */
677 static int
678 add_m6if(struct mif6ctl *mifcp)
679 {
680 	struct epoch_tracker et;
681 	struct mif6 *mifp;
682 	struct ifnet *ifp;
683 	int error;
684 
685 	MIF6_LOCK();
686 
687 	if (mifcp->mif6c_mifi >= MAXMIFS) {
688 		MIF6_UNLOCK();
689 		return (EINVAL);
690 	}
691 	mifp = mif6table + mifcp->mif6c_mifi;
692 	if (mifp->m6_ifp != NULL) {
693 		MIF6_UNLOCK();
694 		return (EADDRINUSE); /* XXX: is it appropriate? */
695 	}
696 
697 	NET_EPOCH_ENTER(et);
698 	if ((ifp = ifnet_byindex(mifcp->mif6c_pifi)) == NULL) {
699 		NET_EPOCH_EXIT(et);
700 		MIF6_UNLOCK();
701 		return (ENXIO);
702 	}
703 	NET_EPOCH_EXIT(et);	/* XXXGL: unsafe ifp */
704 
705 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
706 		if (reg_mif_num == (mifi_t)-1) {
707 			ifp = if_alloc(IFT_OTHER);
708 
709 			if_initname(ifp, "register_mif", 0);
710 			ifp->if_flags |= IFF_LOOPBACK;
711 			if_attach(ifp);
712 			multicast_register_if6 = ifp;
713 			reg_mif_num = mifcp->mif6c_mifi;
714 			/*
715 			 * it is impossible to guess the ifindex of the
716 			 * register interface.  So mif6c_pifi is automatically
717 			 * calculated.
718 			 */
719 			mifcp->mif6c_pifi = ifp->if_index;
720 		} else {
721 			ifp = multicast_register_if6;
722 		}
723 	} else {
724 		/* Make sure the interface supports multicast */
725 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
726 			MIF6_UNLOCK();
727 			return (EOPNOTSUPP);
728 		}
729 
730 		error = if_allmulti(ifp, 1);
731 		if (error) {
732 			MIF6_UNLOCK();
733 			return (error);
734 		}
735 	}
736 
737 	mifp->m6_flags     = mifcp->mif6c_flags;
738 	mifp->m6_ifp       = ifp;
739 
740 	/* initialize per mif pkt counters */
741 	mifp->m6_pkt_in    = 0;
742 	mifp->m6_pkt_out   = 0;
743 	mifp->m6_bytes_in  = 0;
744 	mifp->m6_bytes_out = 0;
745 
746 	/* Adjust nummifs up if the mifi is higher than nummifs */
747 	if (nummifs <= mifcp->mif6c_mifi)
748 		nummifs = mifcp->mif6c_mifi + 1;
749 
750 	MIF6_UNLOCK();
751 	MRT6_DLOG(DEBUG_ANY, "mif #%d, phyint %s", mifcp->mif6c_mifi,
752 	    if_name(ifp));
753 
754 	return (0);
755 }
756 
757 /*
758  * Delete a mif from the mif table
759  */
760 static int
761 del_m6if_locked(mifi_t *mifip)
762 {
763 	struct mif6 *mifp = mif6table + *mifip;
764 	mifi_t mifi;
765 	struct ifnet *ifp;
766 
767 	MIF6_LOCK_ASSERT();
768 
769 	if (*mifip >= nummifs)
770 		return (EINVAL);
771 	if (mifp->m6_ifp == NULL)
772 		return (EINVAL);
773 
774 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
775 		/* XXX: TODO: Maintain an ALLMULTI refcount in struct ifnet. */
776 		ifp = mifp->m6_ifp;
777 		if_allmulti(ifp, 0);
778 	} else {
779 		if (reg_mif_num != (mifi_t)-1 &&
780 		    multicast_register_if6 != NULL) {
781 			if_detach(multicast_register_if6);
782 			if_free(multicast_register_if6);
783 			reg_mif_num = (mifi_t)-1;
784 			multicast_register_if6 = NULL;
785 		}
786 	}
787 
788 	bzero((caddr_t)mifp, sizeof(*mifp));
789 
790 	/* Adjust nummifs down */
791 	for (mifi = nummifs; mifi > 0; mifi--)
792 		if (mif6table[mifi - 1].m6_ifp)
793 			break;
794 	nummifs = mifi;
795 	MRT6_DLOG(DEBUG_ANY, "mif %d, nummifs %d", *mifip, nummifs);
796 
797 	return (0);
798 }
799 
800 static int
801 del_m6if(mifi_t *mifip)
802 {
803 	int cc;
804 
805 	MIF6_LOCK();
806 	cc = del_m6if_locked(mifip);
807 	MIF6_UNLOCK();
808 
809 	return (cc);
810 }
811 
812 /*
813  * Add an mfc entry
814  */
815 static int
816 add_m6fc(struct mf6cctl *mfccp)
817 {
818 	struct mf6c *rt;
819 	u_long hash;
820 	struct rtdetq *rte;
821 	u_short nstl;
822 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
823 
824 	MFC6_LOCK();
825 
826 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
827 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
828 
829 	/* If an entry already exists, just update the fields */
830 	if (rt) {
831 		MRT6_DLOG(DEBUG_MFC, "no upcall o %s g %s p %x",
832 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
833 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
834 		    mfccp->mf6cc_parent);
835 
836 		rt->mf6c_parent = mfccp->mf6cc_parent;
837 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
838 
839 		MFC6_UNLOCK();
840 		return (0);
841 	}
842 
843 	/*
844 	 * Find the entry for which the upcall was made and update
845 	 */
846 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
847 			mfccp->mf6cc_mcastgrp.sin6_addr);
848 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
849 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
850 				       &mfccp->mf6cc_origin.sin6_addr) &&
851 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
852 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
853 		    (rt->mf6c_stall != NULL)) {
854 			if (nstl++)
855 				log(LOG_ERR,
856 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
857 				    "multiple kernel entries",
858 				    ip6_sprintf(ip6bufo,
859 					    &mfccp->mf6cc_origin.sin6_addr),
860 				    ip6_sprintf(ip6bufg,
861 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
862 				    mfccp->mf6cc_parent, rt->mf6c_stall);
863 
864 			MRT6_DLOG(DEBUG_MFC, "o %s g %s p %x dbg %p",
865 			    ip6_sprintf(ip6bufo,
866 			    &mfccp->mf6cc_origin.sin6_addr),
867 			    ip6_sprintf(ip6bufg,
868 				&mfccp->mf6cc_mcastgrp.sin6_addr),
869 			    mfccp->mf6cc_parent, rt->mf6c_stall);
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 		MRT6_DLOG(DEBUG_MFC, "no upcall h %lu o %s g %s p %x", hash,
903 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
904 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
905 		    mfccp->mf6cc_parent);
906 
907 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
908 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
909 					       &mfccp->mf6cc_origin.sin6_addr)&&
910 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
911 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
912 				rt->mf6c_origin     = mfccp->mf6cc_origin;
913 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
914 				rt->mf6c_parent     = mfccp->mf6cc_parent;
915 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
916 				/* initialize pkt counters per src-grp */
917 				rt->mf6c_pkt_cnt    = 0;
918 				rt->mf6c_byte_cnt   = 0;
919 				rt->mf6c_wrong_if   = 0;
920 
921 				if (rt->mf6c_expire)
922 					n6expire[hash]--;
923 				rt->mf6c_expire	   = 0;
924 			}
925 		}
926 		if (rt == NULL) {
927 			/* no upcall, so make a new entry */
928 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
929 						  M_NOWAIT);
930 			if (rt == NULL) {
931 				MFC6_UNLOCK();
932 				return (ENOBUFS);
933 			}
934 
935 			/* insert new entry at head of hash chain */
936 			rt->mf6c_origin     = mfccp->mf6cc_origin;
937 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
938 			rt->mf6c_parent     = mfccp->mf6cc_parent;
939 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
940 			/* initialize pkt counters per src-grp */
941 			rt->mf6c_pkt_cnt    = 0;
942 			rt->mf6c_byte_cnt   = 0;
943 			rt->mf6c_wrong_if   = 0;
944 			rt->mf6c_expire     = 0;
945 			rt->mf6c_stall = NULL;
946 
947 			/* link into table */
948 			rt->mf6c_next  = mf6ctable[hash];
949 			mf6ctable[hash] = rt;
950 		}
951 	}
952 
953 	MFC6_UNLOCK();
954 	return (0);
955 }
956 
957 #ifdef UPCALL_TIMING
958 /*
959  * collect delay statistics on the upcalls
960  */
961 static void
962 collate(struct timeval *t)
963 {
964 	u_long d;
965 	struct timeval tp;
966 	u_long delta;
967 
968 	GET_TIME(tp);
969 
970 	if (TV_LT(*t, tp))
971 	{
972 		TV_DELTA(tp, *t, delta);
973 
974 		d = delta >> 10;
975 		if (d > UPCALL_MAX)
976 			d = UPCALL_MAX;
977 
978 		++upcall_data[d];
979 	}
980 }
981 #endif /* UPCALL_TIMING */
982 
983 /*
984  * Delete an mfc entry
985  */
986 static int
987 del_m6fc(struct mf6cctl *mfccp)
988 {
989 #ifdef MRT6DEBUG
990 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
991 #endif
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 	MRT6_DLOG(DEBUG_MFC, "orig %s mcastgrp %s",
1003 	    ip6_sprintf(ip6bufo, &origin.sin6_addr),
1004 	    ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr));
1005 
1006 	MFC6_LOCK();
1007 
1008 	nptr = &mf6ctable[hash];
1009 	while ((rt = *nptr) != NULL) {
1010 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
1011 				       &rt->mf6c_origin.sin6_addr) &&
1012 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
1013 				       &rt->mf6c_mcastgrp.sin6_addr) &&
1014 		    rt->mf6c_stall == NULL)
1015 			break;
1016 
1017 		nptr = &rt->mf6c_next;
1018 	}
1019 	if (rt == NULL) {
1020 		MFC6_UNLOCK();
1021 		return (EADDRNOTAVAIL);
1022 	}
1023 
1024 	*nptr = rt->mf6c_next;
1025 	free(rt, M_MRTABLE6);
1026 
1027 	MFC6_UNLOCK();
1028 
1029 	return (0);
1030 }
1031 
1032 static int
1033 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
1034 {
1035 
1036 	if (s) {
1037 		if (sbappendaddr(&s->so_rcv,
1038 				 (struct sockaddr *)src,
1039 				 mm, (struct mbuf *)0) != 0) {
1040 			sorwakeup(s);
1041 			return (0);
1042 		} else
1043 			soroverflow(s);
1044 	}
1045 	m_freem(mm);
1046 	return (-1);
1047 }
1048 
1049 /*
1050  * IPv6 multicast forwarding function. This function assumes that the packet
1051  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
1052  * pointed to by "ifp", and the packet is to be relayed to other networks
1053  * that have members of the packet's destination IPv6 multicast group.
1054  *
1055  * The packet is returned unscathed to the caller, unless it is
1056  * erroneous, in which case a non-zero return value tells the caller to
1057  * discard it.
1058  *
1059  * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff
1060  * this function is called in the originating context (i.e., not when
1061  * forwarding a packet from other node).  ip6_output(), which is currently the
1062  * only function that calls this function is called in the originating context,
1063  * explicitly ensures this condition.  It is caller's responsibility to ensure
1064  * that if this function is called from somewhere else in the originating
1065  * context in the future.
1066  */
1067 int
1068 X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
1069 {
1070 	struct rtdetq *rte;
1071 	struct mbuf *mb0;
1072 	struct mf6c *rt;
1073 	struct mif6 *mifp;
1074 	struct mbuf *mm;
1075 	u_long hash;
1076 	mifi_t mifi;
1077 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1078 #ifdef UPCALL_TIMING
1079 	struct timeval tp;
1080 
1081 	GET_TIME(tp);
1082 #endif /* UPCALL_TIMING */
1083 
1084 	MRT6_DLOG(DEBUG_FORWARD, "src %s, dst %s, ifindex %d",
1085 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1086 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst), ifp->if_index);
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 		IP6STAT_INC(ip6s_cantforward);
1106 		if (V_ip6_log_time + V_ip6_log_interval < time_uptime) {
1107 			V_ip6_log_time = time_uptime;
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 	MRT6STAT_INC(mrt6s_mfc_lookups);
1126 
1127 	/* Entry exists, so forward if necessary */
1128 	if (rt) {
1129 		MFC6_UNLOCK();
1130 		return (ip6_mdq(m, ifp, rt));
1131 	}
1132 
1133 	/*
1134 	 * If we don't have a route for packet's origin,
1135 	 * Make a copy of the packet & send message to routing daemon.
1136 	 */
1137 	MRT6STAT_INC(mrt6s_no_route);
1138 	MRT6_DLOG(DEBUG_FORWARD | DEBUG_MFC, "no rte s %s g %s",
1139 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1140 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1141 
1142 	/*
1143 	 * Allocate mbufs early so that we don't do extra work if we
1144 	 * are just going to fail anyway.
1145 	 */
1146 	rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE6, M_NOWAIT);
1147 	if (rte == NULL) {
1148 		MFC6_UNLOCK();
1149 		return (ENOBUFS);
1150 	}
1151 	mb0 = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1152 	/*
1153 	 * Pullup packet header if needed before storing it,
1154 	 * as other references may modify it in the meantime.
1155 	 */
1156 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1157 		mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1158 	if (mb0 == NULL) {
1159 		free(rte, M_MRTABLE6);
1160 		MFC6_UNLOCK();
1161 		return (ENOBUFS);
1162 	}
1163 
1164 	/* is there an upcall waiting for this packet? */
1165 	hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1166 	for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1167 		if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1168 		    &rt->mf6c_origin.sin6_addr) &&
1169 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1170 		    &rt->mf6c_mcastgrp.sin6_addr) && (rt->mf6c_stall != NULL))
1171 			break;
1172 	}
1173 
1174 	if (rt == NULL) {
1175 		struct mrt6msg *im;
1176 #ifdef MRT6_OINIT
1177 		struct omrt6msg *oim;
1178 #endif
1179 		/* no upcall, so make a new entry */
1180 		rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6, M_NOWAIT);
1181 		if (rt == NULL) {
1182 			free(rte, M_MRTABLE6);
1183 			m_freem(mb0);
1184 			MFC6_UNLOCK();
1185 			return (ENOBUFS);
1186 		}
1187 		/*
1188 		 * Make a copy of the header to send to the user
1189 		 * level process
1190 		 */
1191 		mm = m_copym(mb0, 0, sizeof(struct ip6_hdr), M_NOWAIT);
1192 		if (mm == NULL) {
1193 			free(rte, M_MRTABLE6);
1194 			m_freem(mb0);
1195 			free(rt, M_MRTABLE6);
1196 			MFC6_UNLOCK();
1197 			return (ENOBUFS);
1198 		}
1199 
1200 		/*
1201 		 * Send message to routing daemon
1202 		 */
1203 		sin6.sin6_addr = ip6->ip6_src;
1204 		im = NULL;
1205 #ifdef MRT6_OINIT
1206 		oim = NULL;
1207 #endif
1208 		switch (V_ip6_mrouter_ver) {
1209 #ifdef MRT6_OINIT
1210 		case MRT6_OINIT:
1211 			oim = mtod(mm, struct omrt6msg *);
1212 			oim->im6_msgtype = MRT6MSG_NOCACHE;
1213 			oim->im6_mbz = 0;
1214 			break;
1215 #endif
1216 		case MRT6_INIT:
1217 			im = mtod(mm, struct mrt6msg *);
1218 			im->im6_msgtype = MRT6MSG_NOCACHE;
1219 			im->im6_mbz = 0;
1220 			break;
1221 		default:
1222 			free(rte, M_MRTABLE6);
1223 			m_freem(mb0);
1224 			free(rt, M_MRTABLE6);
1225 			MFC6_UNLOCK();
1226 			return (EINVAL);
1227 		}
1228 
1229 		MRT6_DLOG(DEBUG_FORWARD, "getting the iif info in the kernel");
1230 		for (mifp = mif6table, mifi = 0;
1231 		    mifi < nummifs && mifp->m6_ifp != ifp; mifp++, mifi++)
1232 				;
1233 
1234 		switch (V_ip6_mrouter_ver) {
1235 #ifdef MRT6_OINIT
1236 		case MRT6_OINIT:
1237 			oim->im6_mif = mifi;
1238 			break;
1239 #endif
1240 		case MRT6_INIT:
1241 			im->im6_mif = mifi;
1242 			break;
1243 		}
1244 
1245 		if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1246 			log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1247 			    "socket queue full\n");
1248 			MRT6STAT_INC(mrt6s_upq_sockfull);
1249 			free(rte, M_MRTABLE6);
1250 			m_freem(mb0);
1251 			free(rt, M_MRTABLE6);
1252 			MFC6_UNLOCK();
1253 			return (ENOBUFS);
1254 		}
1255 
1256 		MRT6STAT_INC(mrt6s_upcalls);
1257 
1258 		/* insert new entry at head of hash chain */
1259 		bzero(rt, sizeof(*rt));
1260 		rt->mf6c_origin.sin6_family = AF_INET6;
1261 		rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1262 		rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1263 		rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1264 		rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1265 		rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1266 		rt->mf6c_expire = UPCALL_EXPIRE;
1267 		n6expire[hash]++;
1268 		rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1269 
1270 		/* link into table */
1271 		rt->mf6c_next  = mf6ctable[hash];
1272 		mf6ctable[hash] = rt;
1273 		/* Add this entry to the end of the queue */
1274 		rt->mf6c_stall = rte;
1275 	} else {
1276 		/* determine if q has overflowed */
1277 		struct rtdetq **p;
1278 		int npkts = 0;
1279 
1280 		for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1281 			if (++npkts > MAX_UPQ6) {
1282 				MRT6STAT_INC(mrt6s_upq_ovflw);
1283 				free(rte, M_MRTABLE6);
1284 				m_freem(mb0);
1285 				MFC6_UNLOCK();
1286 				return (0);
1287 			}
1288 
1289 		/* Add this entry to the end of the queue */
1290 		*p = rte;
1291 	}
1292 
1293 	rte->next = NULL;
1294 	rte->m = mb0;
1295 	rte->ifp = ifp;
1296 #ifdef UPCALL_TIMING
1297 	rte->t = tp;
1298 #endif /* UPCALL_TIMING */
1299 
1300 	MFC6_UNLOCK();
1301 
1302 	return (0);
1303 }
1304 
1305 /*
1306  * Clean up cache entries if upcalls are not serviced
1307  * Call from the Slow Timeout mechanism, every half second.
1308  */
1309 static void
1310 expire_upcalls(void *unused)
1311 {
1312 #ifdef MRT6DEBUG
1313 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1314 #endif
1315 	struct rtdetq *rte;
1316 	struct mf6c *mfc, **nptr;
1317 	u_long i;
1318 
1319 	MFC6_LOCK_ASSERT();
1320 
1321 	for (i = 0; i < MF6CTBLSIZ; i++) {
1322 		if (n6expire[i] == 0)
1323 			continue;
1324 		nptr = &mf6ctable[i];
1325 		while ((mfc = *nptr) != NULL) {
1326 			rte = mfc->mf6c_stall;
1327 			/*
1328 			 * Skip real cache entries
1329 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1330 			 * If it expires now
1331 			 */
1332 			if (rte != NULL &&
1333 			    mfc->mf6c_expire != 0 &&
1334 			    --mfc->mf6c_expire == 0) {
1335 				MRT6_DLOG(DEBUG_EXPIRE, "expiring (%s %s)",
1336 				    ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr),
1337 				    ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr));
1338 				/*
1339 				 * drop all the packets
1340 				 * free the mbuf with the pkt, if, timing info
1341 				 */
1342 				do {
1343 					struct rtdetq *n = rte->next;
1344 					m_freem(rte->m);
1345 					free(rte, M_MRTABLE6);
1346 					rte = n;
1347 				} while (rte != NULL);
1348 				MRT6STAT_INC(mrt6s_cache_cleanups);
1349 				n6expire[i]--;
1350 
1351 				*nptr = mfc->mf6c_next;
1352 				free(mfc, M_MRTABLE6);
1353 			} else {
1354 				nptr = &mfc->mf6c_next;
1355 			}
1356 		}
1357 	}
1358 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1359 	    expire_upcalls, NULL);
1360 }
1361 
1362 /*
1363  * Packet forwarding routine once entry in the cache is made
1364  */
1365 static int
1366 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1367 {
1368 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1369 	mifi_t mifi, iif;
1370 	struct mif6 *mifp;
1371 	int plen = m->m_pkthdr.len;
1372 	struct in6_addr src0, dst0; /* copies for local work */
1373 	u_int32_t iszone, idzone, oszone, odzone;
1374 	int error = 0;
1375 
1376 	/*
1377 	 * Don't forward if it didn't arrive from the parent mif
1378 	 * for its origin.
1379 	 */
1380 	mifi = rt->mf6c_parent;
1381 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1382 		/* came in the wrong interface */
1383 		MRT6_DLOG(DEBUG_FORWARD,
1384 		    "wrong if: ifid %d mifi %d mififid %x", ifp->if_index,
1385 		    mifi, mif6table[mifi].m6_ifp->if_index);
1386 		MRT6STAT_INC(mrt6s_wrong_if);
1387 		rt->mf6c_wrong_if++;
1388 		/*
1389 		 * If we are doing PIM processing, and we are forwarding
1390 		 * packets on this interface, send a message to the
1391 		 * routing daemon.
1392 		 */
1393 		/* have to make sure this is a valid mif */
1394 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1395 			if (V_pim6 && (m->m_flags & M_LOOP) == 0) {
1396 				/*
1397 				 * Check the M_LOOP flag to avoid an
1398 				 * unnecessary PIM assert.
1399 				 * XXX: M_LOOP is an ad-hoc hack...
1400 				 */
1401 				static struct sockaddr_in6 sin6 =
1402 				{ sizeof(sin6), AF_INET6 };
1403 
1404 				struct mbuf *mm;
1405 				struct mrt6msg *im;
1406 #ifdef MRT6_OINIT
1407 				struct omrt6msg *oim;
1408 #endif
1409 
1410 				mm = m_copym(m, 0, sizeof(struct ip6_hdr),
1411 				    M_NOWAIT);
1412 				if (mm &&
1413 				    (!M_WRITABLE(mm) ||
1414 				     mm->m_len < sizeof(struct ip6_hdr)))
1415 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1416 				if (mm == NULL)
1417 					return (ENOBUFS);
1418 
1419 #ifdef MRT6_OINIT
1420 				oim = NULL;
1421 #endif
1422 				im = NULL;
1423 				switch (V_ip6_mrouter_ver) {
1424 #ifdef MRT6_OINIT
1425 				case MRT6_OINIT:
1426 					oim = mtod(mm, struct omrt6msg *);
1427 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1428 					oim->im6_mbz = 0;
1429 					break;
1430 #endif
1431 				case MRT6_INIT:
1432 					im = mtod(mm, struct mrt6msg *);
1433 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1434 					im->im6_mbz = 0;
1435 					break;
1436 				default:
1437 					m_freem(mm);
1438 					return (EINVAL);
1439 				}
1440 
1441 				for (mifp = mif6table, iif = 0;
1442 				     iif < nummifs && mifp &&
1443 					     mifp->m6_ifp != ifp;
1444 				     mifp++, iif++)
1445 					;
1446 
1447 				switch (V_ip6_mrouter_ver) {
1448 #ifdef MRT6_OINIT
1449 				case MRT6_OINIT:
1450 					oim->im6_mif = iif;
1451 					sin6.sin6_addr = oim->im6_src;
1452 					break;
1453 #endif
1454 				case MRT6_INIT:
1455 					im->im6_mif = iif;
1456 					sin6.sin6_addr = im->im6_src;
1457 					break;
1458 				}
1459 
1460 				MRT6STAT_INC(mrt6s_upcalls);
1461 
1462 				if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1463 					MRT6_DLOG(DEBUG_ANY,
1464 					    "ip6_mrouter socket queue full");
1465 					MRT6STAT_INC(mrt6s_upq_sockfull);
1466 					return (ENOBUFS);
1467 				}	/* if socket Q full */
1468 			}		/* if PIM */
1469 		return (0);
1470 	}			/* if wrong iif */
1471 
1472 	/* If I sourced this packet, it counts as output, else it was input. */
1473 	if (m->m_pkthdr.rcvif == NULL) {
1474 		/* XXX: is rcvif really NULL when output?? */
1475 		mif6table[mifi].m6_pkt_out++;
1476 		mif6table[mifi].m6_bytes_out += plen;
1477 	} else {
1478 		mif6table[mifi].m6_pkt_in++;
1479 		mif6table[mifi].m6_bytes_in += plen;
1480 	}
1481 	rt->mf6c_pkt_cnt++;
1482 	rt->mf6c_byte_cnt += plen;
1483 
1484 	/*
1485 	 * For each mif, forward a copy of the packet if there are group
1486 	 * members downstream on the interface.
1487 	 */
1488 	src0 = ip6->ip6_src;
1489 	dst0 = ip6->ip6_dst;
1490 	if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 ||
1491 	    (error = in6_setscope(&dst0, ifp, &idzone)) != 0) {
1492 		IP6STAT_INC(ip6s_badscope);
1493 		return (error);
1494 	}
1495 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) {
1496 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1497 			/*
1498 			 * check if the outgoing packet is going to break
1499 			 * a scope boundary.
1500 			 * XXX For packets through PIM register tunnel
1501 			 * interface, we believe a routing daemon.
1502 			 */
1503 			if (!(mif6table[rt->mf6c_parent].m6_flags &
1504 			      MIFF_REGISTER) &&
1505 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
1506 				if (in6_setscope(&src0, mif6table[mifi].m6_ifp,
1507 				    &oszone) ||
1508 				    in6_setscope(&dst0, mif6table[mifi].m6_ifp,
1509 				    &odzone) ||
1510 				    iszone != oszone ||
1511 				    idzone != odzone) {
1512 					IP6STAT_INC(ip6s_badscope);
1513 					continue;
1514 				}
1515 			}
1516 
1517 			mifp->m6_pkt_out++;
1518 			mifp->m6_bytes_out += plen;
1519 			if (mifp->m6_flags & MIFF_REGISTER)
1520 				register_send(ip6, mifp, m);
1521 			else
1522 				phyint_send(ip6, mifp, m);
1523 		}
1524 	}
1525 	return (0);
1526 }
1527 
1528 static void
1529 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1530 {
1531 #ifdef MRT6DEBUG
1532 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1533 #endif
1534 	struct mbuf *mb_copy;
1535 	struct ifnet *ifp = mifp->m6_ifp;
1536 	int error = 0;
1537 	u_long linkmtu;
1538 
1539 	/*
1540 	 * Make a new reference to the packet; make sure that
1541 	 * the IPv6 header is actually copied, not just referenced,
1542 	 * so that ip6_output() only scribbles on the copy.
1543 	 */
1544 	mb_copy = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1545 	if (mb_copy &&
1546 	    (!M_WRITABLE(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1547 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1548 	if (mb_copy == NULL) {
1549 		return;
1550 	}
1551 	/* set MCAST flag to the outgoing packet */
1552 	mb_copy->m_flags |= M_MCAST;
1553 
1554 	/*
1555 	 * If we sourced the packet, call ip6_output since we may devide
1556 	 * the packet into fragments when the packet is too big for the
1557 	 * outgoing interface.
1558 	 * Otherwise, we can simply send the packet to the interface
1559 	 * sending queue.
1560 	 */
1561 	if (m->m_pkthdr.rcvif == NULL) {
1562 		struct ip6_moptions im6o;
1563 		struct epoch_tracker et;
1564 
1565 		im6o.im6o_multicast_ifp = ifp;
1566 		/* XXX: ip6_output will override ip6->ip6_hlim */
1567 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1568 		im6o.im6o_multicast_loop = 1;
1569 		NET_EPOCH_ENTER(et);
1570 		error = ip6_output(mb_copy, NULL, NULL, IPV6_FORWARDING, &im6o,
1571 		    NULL, NULL);
1572 		NET_EPOCH_EXIT(et);
1573 
1574 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1575 		    (uint16_t)(mifp - mif6table), error);
1576 		return;
1577 	}
1578 
1579 	/*
1580 	 * If configured to loop back multicasts by default,
1581 	 * loop back a copy now.
1582 	 */
1583 	if (in6_mcast_loop)
1584 		ip6_mloopback(ifp, m);
1585 
1586 	/*
1587 	 * Put the packet into the sending queue of the outgoing interface
1588 	 * if it would fit in the MTU of the interface.
1589 	 */
1590 	linkmtu = IN6_LINKMTU(ifp);
1591 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1592 		struct sockaddr_in6 dst6;
1593 
1594 		bzero(&dst6, sizeof(dst6));
1595 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1596 		dst6.sin6_family = AF_INET6;
1597 		dst6.sin6_addr = ip6->ip6_dst;
1598 
1599 		IP_PROBE(send, NULL, NULL, ip6, ifp, NULL, ip6);
1600 		/*
1601 		 * We just call if_output instead of nd6_output here, since
1602 		 * we need no ND for a multicast forwarded packet...right?
1603 		 */
1604 		m_clrprotoflags(m);	/* Avoid confusing lower layers. */
1605 		error = (*ifp->if_output)(ifp, mb_copy,
1606 		    (struct sockaddr *)&dst6, NULL);
1607 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1608 		    (uint16_t)(mifp - mif6table), error);
1609 	} else {
1610 		/*
1611 		 * pMTU discovery is intentionally disabled by default, since
1612 		 * various router may notify pMTU in multicast, which can be
1613 		 * a DDoS to a router
1614 		 */
1615 		if (V_ip6_mcast_pmtu)
1616 			icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1617 		else {
1618 			MRT6_DLOG(DEBUG_XMIT, " packet too big on %s o %s "
1619 			    "g %s size %d (discarded)", if_name(ifp),
1620 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1621 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1622 			    mb_copy->m_pkthdr.len);
1623 			m_freem(mb_copy); /* simply discard the packet */
1624 		}
1625 	}
1626 }
1627 
1628 static int
1629 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1630 {
1631 #ifdef MRT6DEBUG
1632 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1633 #endif
1634 	struct mbuf *mm;
1635 	int i, len = m->m_pkthdr.len;
1636 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1637 	struct mrt6msg *im6;
1638 
1639 	MRT6_DLOG(DEBUG_ANY, "src %s dst %s",
1640 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1641 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1642 	PIM6STAT_INC(pim6s_snd_registers);
1643 
1644 	/* Make a copy of the packet to send to the user level process. */
1645 	mm = m_gethdr(M_NOWAIT, MT_DATA);
1646 	if (mm == NULL)
1647 		return (ENOBUFS);
1648 	mm->m_data += max_linkhdr;
1649 	mm->m_len = sizeof(struct ip6_hdr);
1650 
1651 	if ((mm->m_next = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
1652 		m_freem(mm);
1653 		return (ENOBUFS);
1654 	}
1655 	i = MHLEN - M_LEADINGSPACE(mm);
1656 	if (i > len)
1657 		i = len;
1658 	mm = m_pullup(mm, i);
1659 	if (mm == NULL)
1660 		return (ENOBUFS);
1661 /* TODO: check it! */
1662 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1663 
1664 	/*
1665 	 * Send message to routing daemon
1666 	 */
1667 	sin6.sin6_addr = ip6->ip6_src;
1668 
1669 	im6 = mtod(mm, struct mrt6msg *);
1670 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1671 	im6->im6_mbz          = 0;
1672 
1673 	im6->im6_mif = mif - mif6table;
1674 
1675 	/* iif info is not given for reg. encap.n */
1676 	MRT6STAT_INC(mrt6s_upcalls);
1677 
1678 	if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1679 		MRT6_DLOG(DEBUG_ANY, "ip6_mrouter socket queue full");
1680 		MRT6STAT_INC(mrt6s_upq_sockfull);
1681 		return (ENOBUFS);
1682 	}
1683 	return (0);
1684 }
1685 
1686 /*
1687  * pim6_encapcheck() is called by the encap6_input() path at runtime to
1688  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
1689  * into the kernel.
1690  */
1691 static int
1692 pim6_encapcheck(const struct mbuf *m __unused, int off __unused,
1693     int proto __unused, void *arg __unused)
1694 {
1695 
1696     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
1697     return (8);		/* claim the datagram. */
1698 }
1699 
1700 /*
1701  * PIM sparse mode hook
1702  * Receives the pim control messages, and passes them up to the listening
1703  * socket, using rip6_input.
1704  * The only message processed is the REGISTER pim message; the pim header
1705  * is stripped off, and the inner packet is passed to register_mforward.
1706  */
1707 static int
1708 pim6_input(struct mbuf *m, int off, int proto, void *arg __unused)
1709 {
1710 	struct pim *pim; /* pointer to a pim struct */
1711 	struct ip6_hdr *ip6;
1712 	int pimlen;
1713 	int minlen;
1714 
1715 	PIM6STAT_INC(pim6s_rcv_total);
1716 
1717 	/*
1718 	 * Validate lengths
1719 	 */
1720 	pimlen = m->m_pkthdr.len - off;
1721 	if (pimlen < PIM_MINLEN) {
1722 		PIM6STAT_INC(pim6s_rcv_tooshort);
1723 		MRT6_DLOG(DEBUG_PIM, "PIM packet too short");
1724 		m_freem(m);
1725 		return (IPPROTO_DONE);
1726 	}
1727 
1728 	/*
1729 	 * if the packet is at least as big as a REGISTER, go ahead
1730 	 * and grab the PIM REGISTER header size, to avoid another
1731 	 * possible m_pullup() later.
1732 	 *
1733 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1734 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1735 	 */
1736 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1737 
1738 	/*
1739 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1740 	 * possibly the PIM REGISTER header
1741 	 */
1742 	if (m->m_len < off + minlen) {
1743 		m = m_pullup(m, off + minlen);
1744 		if (m == NULL) {
1745 			IP6STAT_INC(ip6s_exthdrtoolong);
1746 			return (IPPROTO_DONE);
1747 		}
1748 	}
1749 	ip6 = mtod(m, struct ip6_hdr *);
1750 	pim = (struct pim *)((caddr_t)ip6 + off);
1751 
1752 #define PIM6_CHECKSUM
1753 #ifdef PIM6_CHECKSUM
1754 	{
1755 		int cksumlen;
1756 
1757 		/*
1758 		 * Validate checksum.
1759 		 * If PIM REGISTER, exclude the data packet
1760 		 */
1761 		if (pim->pim_type == PIM_REGISTER)
1762 			cksumlen = PIM_MINLEN;
1763 		else
1764 			cksumlen = pimlen;
1765 
1766 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1767 			PIM6STAT_INC(pim6s_rcv_badsum);
1768 			MRT6_DLOG(DEBUG_PIM, "invalid checksum");
1769 			m_freem(m);
1770 			return (IPPROTO_DONE);
1771 		}
1772 	}
1773 #endif /* PIM_CHECKSUM */
1774 
1775 	/* PIM version check */
1776 	if (pim->pim_ver != PIM_VERSION) {
1777 		PIM6STAT_INC(pim6s_rcv_badversion);
1778 		MRT6_DLOG(DEBUG_ANY | DEBUG_ERR,
1779 		    "incorrect version %d, expecting %d",
1780 		    pim->pim_ver, PIM_VERSION);
1781 		m_freem(m);
1782 		return (IPPROTO_DONE);
1783 	}
1784 
1785 	if (pim->pim_type == PIM_REGISTER) {
1786 		/*
1787 		 * since this is a REGISTER, we'll make a copy of the register
1788 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1789 		 * routing daemon.
1790 		 */
1791 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1792 
1793 		struct mbuf *mcp;
1794 		struct ip6_hdr *eip6;
1795 		u_int32_t *reghdr;
1796 		int rc;
1797 #ifdef MRT6DEBUG
1798 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1799 #endif
1800 
1801 		PIM6STAT_INC(pim6s_rcv_registers);
1802 
1803 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1804 			MRT6_DLOG(DEBUG_PIM, "register mif not set: %d",
1805 			    reg_mif_num);
1806 			m_freem(m);
1807 			return (IPPROTO_DONE);
1808 		}
1809 
1810 		reghdr = (u_int32_t *)(pim + 1);
1811 
1812 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1813 			goto pim6_input_to_daemon;
1814 
1815 		/*
1816 		 * Validate length
1817 		 */
1818 		if (pimlen < PIM6_REG_MINLEN) {
1819 			PIM6STAT_INC(pim6s_rcv_tooshort);
1820 			PIM6STAT_INC(pim6s_rcv_badregisters);
1821 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "register packet "
1822 			    "size too small %d from %s",
1823 			    pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src));
1824 			m_freem(m);
1825 			return (IPPROTO_DONE);
1826 		}
1827 
1828 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1829 		MRT6_DLOG(DEBUG_PIM, "eip6: %s -> %s, eip6 plen %d",
1830 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1831 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1832 		    ntohs(eip6->ip6_plen));
1833 
1834 		/* verify the version number of the inner packet */
1835 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1836 			PIM6STAT_INC(pim6s_rcv_badregisters);
1837 			MRT6_DLOG(DEBUG_ANY, "invalid IP version (%d) "
1838 			    "of the inner packet",
1839 			    (eip6->ip6_vfc & IPV6_VERSION));
1840 			m_freem(m);
1841 			return (IPPROTO_DONE);
1842 		}
1843 
1844 		/* verify the inner packet is destined to a mcast group */
1845 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1846 			PIM6STAT_INC(pim6s_rcv_badregisters);
1847 			MRT6_DLOG(DEBUG_PIM, "inner packet of register "
1848 			    "is not multicast %s",
1849 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst));
1850 			m_freem(m);
1851 			return (IPPROTO_DONE);
1852 		}
1853 
1854 		/*
1855 		 * make a copy of the whole header to pass to the daemon later.
1856 		 */
1857 		mcp = m_copym(m, 0, off + PIM6_REG_MINLEN, M_NOWAIT);
1858 		if (mcp == NULL) {
1859 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "pim register: "
1860 			    "could not copy register head");
1861 			m_freem(m);
1862 			return (IPPROTO_DONE);
1863 		}
1864 
1865 		/*
1866 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1867 		 */
1868 		m_adj(m, off + PIM_MINLEN);
1869 		MRT6_DLOG(DEBUG_PIM, "forwarding decapsulated register: "
1870 		    "src %s, dst %s, mif %d",
1871 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1872 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst), reg_mif_num);
1873 
1874 		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1875 				dst.sin6_family, 0);
1876 
1877 		/* prepare the register head to send to the mrouting daemon */
1878 		m = mcp;
1879 	}
1880 
1881 	/*
1882 	 * Pass the PIM message up to the daemon; if it is a register message
1883 	 * pass the 'head' only up to the daemon. This includes the
1884 	 * encapsulator ip6 header, pim header, register header and the
1885 	 * encapsulated ip6 header.
1886 	 */
1887   pim6_input_to_daemon:
1888 	return (rip6_input(&m, &off, proto));
1889 }
1890 
1891 static int
1892 ip6_mroute_modevent(module_t mod, int type, void *unused)
1893 {
1894 
1895 	switch (type) {
1896 	case MOD_LOAD:
1897 		MROUTER6_LOCK_INIT();
1898 		MFC6_LOCK_INIT();
1899 		MIF6_LOCK_INIT();
1900 
1901 		pim6_encap_cookie = ip6_encap_attach(&ipv6_encap_cfg,
1902 		    NULL, M_WAITOK);
1903 		if (pim6_encap_cookie == NULL) {
1904 			printf("ip6_mroute: unable to attach pim6 encap\n");
1905 			MIF6_LOCK_DESTROY();
1906 			MFC6_LOCK_DESTROY();
1907 			MROUTER6_LOCK_DESTROY();
1908 			return (EINVAL);
1909 		}
1910 
1911 		ip6_mforward = X_ip6_mforward;
1912 		ip6_mrouter_done = X_ip6_mrouter_done;
1913 		ip6_mrouter_get = X_ip6_mrouter_get;
1914 		ip6_mrouter_set = X_ip6_mrouter_set;
1915 		mrt6_ioctl = X_mrt6_ioctl;
1916 		break;
1917 
1918 	case MOD_UNLOAD:
1919 		if (V_ip6_mrouter != NULL)
1920 			return EINVAL;
1921 
1922 		if (pim6_encap_cookie) {
1923 			ip6_encap_detach(pim6_encap_cookie);
1924 			pim6_encap_cookie = NULL;
1925 		}
1926 		X_ip6_mrouter_done();
1927 		ip6_mforward = NULL;
1928 		ip6_mrouter_done = NULL;
1929 		ip6_mrouter_get = NULL;
1930 		ip6_mrouter_set = NULL;
1931 		mrt6_ioctl = NULL;
1932 
1933 		MIF6_LOCK_DESTROY();
1934 		MFC6_LOCK_DESTROY();
1935 		MROUTER6_LOCK_DESTROY();
1936 		break;
1937 
1938 	default:
1939 		return (EOPNOTSUPP);
1940 	}
1941 
1942 	return (0);
1943 }
1944 
1945 static moduledata_t ip6_mroutemod = {
1946 	"ip6_mroute",
1947 	ip6_mroute_modevent,
1948 	0
1949 };
1950 
1951 DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_ANY);
1952