xref: /freebsd/sys/netinet6/ip6_mroute.c (revision 3332f1b444d4a73238e9f59cca27bfc95fe936bd)
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 mif6 *mifp;
681 	struct ifnet *ifp;
682 	int error;
683 
684 	MIF6_LOCK();
685 
686 	if (mifcp->mif6c_mifi >= MAXMIFS) {
687 		MIF6_UNLOCK();
688 		return (EINVAL);
689 	}
690 	mifp = mif6table + mifcp->mif6c_mifi;
691 	if (mifp->m6_ifp != NULL) {
692 		MIF6_UNLOCK();
693 		return (EADDRINUSE); /* XXX: is it appropriate? */
694 	}
695 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > V_if_index) {
696 		MIF6_UNLOCK();
697 		return (ENXIO);
698 	}
699 
700 	ifp = ifnet_byindex(mifcp->mif6c_pifi);
701 
702 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
703 		if (reg_mif_num == (mifi_t)-1) {
704 			ifp = if_alloc(IFT_OTHER);
705 
706 			if_initname(ifp, "register_mif", 0);
707 			ifp->if_flags |= IFF_LOOPBACK;
708 			if_attach(ifp);
709 			multicast_register_if6 = ifp;
710 			reg_mif_num = mifcp->mif6c_mifi;
711 			/*
712 			 * it is impossible to guess the ifindex of the
713 			 * register interface.  So mif6c_pifi is automatically
714 			 * calculated.
715 			 */
716 			mifcp->mif6c_pifi = ifp->if_index;
717 		} else {
718 			ifp = multicast_register_if6;
719 		}
720 	} else {
721 		/* Make sure the interface supports multicast */
722 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
723 			MIF6_UNLOCK();
724 			return (EOPNOTSUPP);
725 		}
726 
727 		error = if_allmulti(ifp, 1);
728 		if (error) {
729 			MIF6_UNLOCK();
730 			return (error);
731 		}
732 	}
733 
734 	mifp->m6_flags     = mifcp->mif6c_flags;
735 	mifp->m6_ifp       = ifp;
736 
737 	/* initialize per mif pkt counters */
738 	mifp->m6_pkt_in    = 0;
739 	mifp->m6_pkt_out   = 0;
740 	mifp->m6_bytes_in  = 0;
741 	mifp->m6_bytes_out = 0;
742 
743 	/* Adjust nummifs up if the mifi is higher than nummifs */
744 	if (nummifs <= mifcp->mif6c_mifi)
745 		nummifs = mifcp->mif6c_mifi + 1;
746 
747 	MIF6_UNLOCK();
748 	MRT6_DLOG(DEBUG_ANY, "mif #%d, phyint %s", mifcp->mif6c_mifi,
749 	    if_name(ifp));
750 
751 	return (0);
752 }
753 
754 /*
755  * Delete a mif from the mif table
756  */
757 static int
758 del_m6if_locked(mifi_t *mifip)
759 {
760 	struct mif6 *mifp = mif6table + *mifip;
761 	mifi_t mifi;
762 	struct ifnet *ifp;
763 
764 	MIF6_LOCK_ASSERT();
765 
766 	if (*mifip >= nummifs)
767 		return (EINVAL);
768 	if (mifp->m6_ifp == NULL)
769 		return (EINVAL);
770 
771 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
772 		/* XXX: TODO: Maintain an ALLMULTI refcount in struct ifnet. */
773 		ifp = mifp->m6_ifp;
774 		if_allmulti(ifp, 0);
775 	} else {
776 		if (reg_mif_num != (mifi_t)-1 &&
777 		    multicast_register_if6 != NULL) {
778 			if_detach(multicast_register_if6);
779 			if_free(multicast_register_if6);
780 			reg_mif_num = (mifi_t)-1;
781 			multicast_register_if6 = NULL;
782 		}
783 	}
784 
785 	bzero((caddr_t)mifp, sizeof(*mifp));
786 
787 	/* Adjust nummifs down */
788 	for (mifi = nummifs; mifi > 0; mifi--)
789 		if (mif6table[mifi - 1].m6_ifp)
790 			break;
791 	nummifs = mifi;
792 	MRT6_DLOG(DEBUG_ANY, "mif %d, nummifs %d", *mifip, nummifs);
793 
794 	return (0);
795 }
796 
797 static int
798 del_m6if(mifi_t *mifip)
799 {
800 	int cc;
801 
802 	MIF6_LOCK();
803 	cc = del_m6if_locked(mifip);
804 	MIF6_UNLOCK();
805 
806 	return (cc);
807 }
808 
809 /*
810  * Add an mfc entry
811  */
812 static int
813 add_m6fc(struct mf6cctl *mfccp)
814 {
815 	struct mf6c *rt;
816 	u_long hash;
817 	struct rtdetq *rte;
818 	u_short nstl;
819 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
820 
821 	MFC6_LOCK();
822 
823 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
824 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
825 
826 	/* If an entry already exists, just update the fields */
827 	if (rt) {
828 		MRT6_DLOG(DEBUG_MFC, "no upcall o %s g %s p %x",
829 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
830 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
831 		    mfccp->mf6cc_parent);
832 
833 		rt->mf6c_parent = mfccp->mf6cc_parent;
834 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
835 
836 		MFC6_UNLOCK();
837 		return (0);
838 	}
839 
840 	/*
841 	 * Find the entry for which the upcall was made and update
842 	 */
843 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
844 			mfccp->mf6cc_mcastgrp.sin6_addr);
845 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
846 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
847 				       &mfccp->mf6cc_origin.sin6_addr) &&
848 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
849 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
850 		    (rt->mf6c_stall != NULL)) {
851 			if (nstl++)
852 				log(LOG_ERR,
853 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
854 				    "multiple kernel entries",
855 				    ip6_sprintf(ip6bufo,
856 					    &mfccp->mf6cc_origin.sin6_addr),
857 				    ip6_sprintf(ip6bufg,
858 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
859 				    mfccp->mf6cc_parent, rt->mf6c_stall);
860 
861 			MRT6_DLOG(DEBUG_MFC, "o %s g %s p %x dbg %p",
862 			    ip6_sprintf(ip6bufo,
863 			    &mfccp->mf6cc_origin.sin6_addr),
864 			    ip6_sprintf(ip6bufg,
865 				&mfccp->mf6cc_mcastgrp.sin6_addr),
866 			    mfccp->mf6cc_parent, rt->mf6c_stall);
867 
868 			rt->mf6c_origin     = mfccp->mf6cc_origin;
869 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
870 			rt->mf6c_parent     = mfccp->mf6cc_parent;
871 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
872 			/* initialize pkt counters per src-grp */
873 			rt->mf6c_pkt_cnt    = 0;
874 			rt->mf6c_byte_cnt   = 0;
875 			rt->mf6c_wrong_if   = 0;
876 
877 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
878 			n6expire[hash]--;
879 
880 			/* free packets Qed at the end of this entry */
881 			for (rte = rt->mf6c_stall; rte != NULL; ) {
882 				struct rtdetq *n = rte->next;
883 				ip6_mdq(rte->m, rte->ifp, rt);
884 				m_freem(rte->m);
885 #ifdef UPCALL_TIMING
886 				collate(&(rte->t));
887 #endif /* UPCALL_TIMING */
888 				free(rte, M_MRTABLE6);
889 				rte = n;
890 			}
891 			rt->mf6c_stall = NULL;
892 		}
893 	}
894 
895 	/*
896 	 * It is possible that an entry is being inserted without an upcall
897 	 */
898 	if (nstl == 0) {
899 		MRT6_DLOG(DEBUG_MFC, "no upcall h %lu o %s g %s p %x", hash,
900 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
901 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
902 		    mfccp->mf6cc_parent);
903 
904 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
905 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
906 					       &mfccp->mf6cc_origin.sin6_addr)&&
907 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
908 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
909 				rt->mf6c_origin     = mfccp->mf6cc_origin;
910 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
911 				rt->mf6c_parent     = mfccp->mf6cc_parent;
912 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
913 				/* initialize pkt counters per src-grp */
914 				rt->mf6c_pkt_cnt    = 0;
915 				rt->mf6c_byte_cnt   = 0;
916 				rt->mf6c_wrong_if   = 0;
917 
918 				if (rt->mf6c_expire)
919 					n6expire[hash]--;
920 				rt->mf6c_expire	   = 0;
921 			}
922 		}
923 		if (rt == NULL) {
924 			/* no upcall, so make a new entry */
925 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
926 						  M_NOWAIT);
927 			if (rt == NULL) {
928 				MFC6_UNLOCK();
929 				return (ENOBUFS);
930 			}
931 
932 			/* insert new entry at head of hash chain */
933 			rt->mf6c_origin     = mfccp->mf6cc_origin;
934 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
935 			rt->mf6c_parent     = mfccp->mf6cc_parent;
936 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
937 			/* initialize pkt counters per src-grp */
938 			rt->mf6c_pkt_cnt    = 0;
939 			rt->mf6c_byte_cnt   = 0;
940 			rt->mf6c_wrong_if   = 0;
941 			rt->mf6c_expire     = 0;
942 			rt->mf6c_stall = NULL;
943 
944 			/* link into table */
945 			rt->mf6c_next  = mf6ctable[hash];
946 			mf6ctable[hash] = rt;
947 		}
948 	}
949 
950 	MFC6_UNLOCK();
951 	return (0);
952 }
953 
954 #ifdef UPCALL_TIMING
955 /*
956  * collect delay statistics on the upcalls
957  */
958 static void
959 collate(struct timeval *t)
960 {
961 	u_long d;
962 	struct timeval tp;
963 	u_long delta;
964 
965 	GET_TIME(tp);
966 
967 	if (TV_LT(*t, tp))
968 	{
969 		TV_DELTA(tp, *t, delta);
970 
971 		d = delta >> 10;
972 		if (d > UPCALL_MAX)
973 			d = UPCALL_MAX;
974 
975 		++upcall_data[d];
976 	}
977 }
978 #endif /* UPCALL_TIMING */
979 
980 /*
981  * Delete an mfc entry
982  */
983 static int
984 del_m6fc(struct mf6cctl *mfccp)
985 {
986 #ifdef MRT6DEBUG
987 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
988 #endif
989 	struct sockaddr_in6	origin;
990 	struct sockaddr_in6	mcastgrp;
991 	struct mf6c		*rt;
992 	struct mf6c		**nptr;
993 	u_long		hash;
994 
995 	origin = mfccp->mf6cc_origin;
996 	mcastgrp = mfccp->mf6cc_mcastgrp;
997 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
998 
999 	MRT6_DLOG(DEBUG_MFC, "orig %s mcastgrp %s",
1000 	    ip6_sprintf(ip6bufo, &origin.sin6_addr),
1001 	    ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr));
1002 
1003 	MFC6_LOCK();
1004 
1005 	nptr = &mf6ctable[hash];
1006 	while ((rt = *nptr) != NULL) {
1007 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
1008 				       &rt->mf6c_origin.sin6_addr) &&
1009 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
1010 				       &rt->mf6c_mcastgrp.sin6_addr) &&
1011 		    rt->mf6c_stall == NULL)
1012 			break;
1013 
1014 		nptr = &rt->mf6c_next;
1015 	}
1016 	if (rt == NULL) {
1017 		MFC6_UNLOCK();
1018 		return (EADDRNOTAVAIL);
1019 	}
1020 
1021 	*nptr = rt->mf6c_next;
1022 	free(rt, M_MRTABLE6);
1023 
1024 	MFC6_UNLOCK();
1025 
1026 	return (0);
1027 }
1028 
1029 static int
1030 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
1031 {
1032 
1033 	if (s) {
1034 		if (sbappendaddr(&s->so_rcv,
1035 				 (struct sockaddr *)src,
1036 				 mm, (struct mbuf *)0) != 0) {
1037 			sorwakeup(s);
1038 			return (0);
1039 		} else
1040 			soroverflow(s);
1041 	}
1042 	m_freem(mm);
1043 	return (-1);
1044 }
1045 
1046 /*
1047  * IPv6 multicast forwarding function. This function assumes that the packet
1048  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
1049  * pointed to by "ifp", and the packet is to be relayed to other networks
1050  * that have members of the packet's destination IPv6 multicast group.
1051  *
1052  * The packet is returned unscathed to the caller, unless it is
1053  * erroneous, in which case a non-zero return value tells the caller to
1054  * discard it.
1055  *
1056  * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff
1057  * this function is called in the originating context (i.e., not when
1058  * forwarding a packet from other node).  ip6_output(), which is currently the
1059  * only function that calls this function is called in the originating context,
1060  * explicitly ensures this condition.  It is caller's responsibility to ensure
1061  * that if this function is called from somewhere else in the originating
1062  * context in the future.
1063  */
1064 int
1065 X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
1066 {
1067 	struct rtdetq *rte;
1068 	struct mbuf *mb0;
1069 	struct mf6c *rt;
1070 	struct mif6 *mifp;
1071 	struct mbuf *mm;
1072 	u_long hash;
1073 	mifi_t mifi;
1074 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1075 #ifdef UPCALL_TIMING
1076 	struct timeval tp;
1077 
1078 	GET_TIME(tp);
1079 #endif /* UPCALL_TIMING */
1080 
1081 	MRT6_DLOG(DEBUG_FORWARD, "src %s, dst %s, ifindex %d",
1082 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1083 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst), ifp->if_index);
1084 
1085 	/*
1086 	 * Don't forward a packet with Hop limit of zero or one,
1087 	 * or a packet destined to a local-only group.
1088 	 */
1089 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
1090 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1091 		return (0);
1092 	ip6->ip6_hlim--;
1093 
1094 	/*
1095 	 * Source address check: do not forward packets with unspecified
1096 	 * source. It was discussed in July 2000, on ipngwg mailing list.
1097 	 * This is rather more serious than unicast cases, because some
1098 	 * MLD packets can be sent with the unspecified source address
1099 	 * (although such packets must normally set 1 to the hop limit field).
1100 	 */
1101 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1102 		IP6STAT_INC(ip6s_cantforward);
1103 		if (V_ip6_log_time + V_ip6_log_interval < time_uptime) {
1104 			V_ip6_log_time = time_uptime;
1105 			log(LOG_DEBUG,
1106 			    "cannot forward "
1107 			    "from %s to %s nxt %d received on %s\n",
1108 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1109 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1110 			    ip6->ip6_nxt,
1111 			    if_name(m->m_pkthdr.rcvif));
1112 		}
1113 		return (0);
1114 	}
1115 
1116 	MFC6_LOCK();
1117 
1118 	/*
1119 	 * Determine forwarding mifs from the forwarding cache table
1120 	 */
1121 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
1122 	MRT6STAT_INC(mrt6s_mfc_lookups);
1123 
1124 	/* Entry exists, so forward if necessary */
1125 	if (rt) {
1126 		MFC6_UNLOCK();
1127 		return (ip6_mdq(m, ifp, rt));
1128 	}
1129 
1130 	/*
1131 	 * If we don't have a route for packet's origin,
1132 	 * Make a copy of the packet & send message to routing daemon.
1133 	 */
1134 	MRT6STAT_INC(mrt6s_no_route);
1135 	MRT6_DLOG(DEBUG_FORWARD | DEBUG_MFC, "no rte s %s g %s",
1136 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1137 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1138 
1139 	/*
1140 	 * Allocate mbufs early so that we don't do extra work if we
1141 	 * are just going to fail anyway.
1142 	 */
1143 	rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE6, M_NOWAIT);
1144 	if (rte == NULL) {
1145 		MFC6_UNLOCK();
1146 		return (ENOBUFS);
1147 	}
1148 	mb0 = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1149 	/*
1150 	 * Pullup packet header if needed before storing it,
1151 	 * as other references may modify it in the meantime.
1152 	 */
1153 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1154 		mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1155 	if (mb0 == NULL) {
1156 		free(rte, M_MRTABLE6);
1157 		MFC6_UNLOCK();
1158 		return (ENOBUFS);
1159 	}
1160 
1161 	/* is there an upcall waiting for this packet? */
1162 	hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1163 	for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1164 		if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1165 		    &rt->mf6c_origin.sin6_addr) &&
1166 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1167 		    &rt->mf6c_mcastgrp.sin6_addr) && (rt->mf6c_stall != NULL))
1168 			break;
1169 	}
1170 
1171 	if (rt == NULL) {
1172 		struct mrt6msg *im;
1173 #ifdef MRT6_OINIT
1174 		struct omrt6msg *oim;
1175 #endif
1176 		/* no upcall, so make a new entry */
1177 		rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6, M_NOWAIT);
1178 		if (rt == NULL) {
1179 			free(rte, M_MRTABLE6);
1180 			m_freem(mb0);
1181 			MFC6_UNLOCK();
1182 			return (ENOBUFS);
1183 		}
1184 		/*
1185 		 * Make a copy of the header to send to the user
1186 		 * level process
1187 		 */
1188 		mm = m_copym(mb0, 0, sizeof(struct ip6_hdr), M_NOWAIT);
1189 		if (mm == NULL) {
1190 			free(rte, M_MRTABLE6);
1191 			m_freem(mb0);
1192 			free(rt, M_MRTABLE6);
1193 			MFC6_UNLOCK();
1194 			return (ENOBUFS);
1195 		}
1196 
1197 		/*
1198 		 * Send message to routing daemon
1199 		 */
1200 		sin6.sin6_addr = ip6->ip6_src;
1201 		im = NULL;
1202 #ifdef MRT6_OINIT
1203 		oim = NULL;
1204 #endif
1205 		switch (V_ip6_mrouter_ver) {
1206 #ifdef MRT6_OINIT
1207 		case MRT6_OINIT:
1208 			oim = mtod(mm, struct omrt6msg *);
1209 			oim->im6_msgtype = MRT6MSG_NOCACHE;
1210 			oim->im6_mbz = 0;
1211 			break;
1212 #endif
1213 		case MRT6_INIT:
1214 			im = mtod(mm, struct mrt6msg *);
1215 			im->im6_msgtype = MRT6MSG_NOCACHE;
1216 			im->im6_mbz = 0;
1217 			break;
1218 		default:
1219 			free(rte, M_MRTABLE6);
1220 			m_freem(mb0);
1221 			free(rt, M_MRTABLE6);
1222 			MFC6_UNLOCK();
1223 			return (EINVAL);
1224 		}
1225 
1226 		MRT6_DLOG(DEBUG_FORWARD, "getting the iif info in the kernel");
1227 		for (mifp = mif6table, mifi = 0;
1228 		    mifi < nummifs && mifp->m6_ifp != ifp; mifp++, mifi++)
1229 				;
1230 
1231 		switch (V_ip6_mrouter_ver) {
1232 #ifdef MRT6_OINIT
1233 		case MRT6_OINIT:
1234 			oim->im6_mif = mifi;
1235 			break;
1236 #endif
1237 		case MRT6_INIT:
1238 			im->im6_mif = mifi;
1239 			break;
1240 		}
1241 
1242 		if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1243 			log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1244 			    "socket queue full\n");
1245 			MRT6STAT_INC(mrt6s_upq_sockfull);
1246 			free(rte, M_MRTABLE6);
1247 			m_freem(mb0);
1248 			free(rt, M_MRTABLE6);
1249 			MFC6_UNLOCK();
1250 			return (ENOBUFS);
1251 		}
1252 
1253 		MRT6STAT_INC(mrt6s_upcalls);
1254 
1255 		/* insert new entry at head of hash chain */
1256 		bzero(rt, sizeof(*rt));
1257 		rt->mf6c_origin.sin6_family = AF_INET6;
1258 		rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1259 		rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1260 		rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1261 		rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1262 		rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1263 		rt->mf6c_expire = UPCALL_EXPIRE;
1264 		n6expire[hash]++;
1265 		rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1266 
1267 		/* link into table */
1268 		rt->mf6c_next  = mf6ctable[hash];
1269 		mf6ctable[hash] = rt;
1270 		/* Add this entry to the end of the queue */
1271 		rt->mf6c_stall = rte;
1272 	} else {
1273 		/* determine if q has overflowed */
1274 		struct rtdetq **p;
1275 		int npkts = 0;
1276 
1277 		for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1278 			if (++npkts > MAX_UPQ6) {
1279 				MRT6STAT_INC(mrt6s_upq_ovflw);
1280 				free(rte, M_MRTABLE6);
1281 				m_freem(mb0);
1282 				MFC6_UNLOCK();
1283 				return (0);
1284 			}
1285 
1286 		/* Add this entry to the end of the queue */
1287 		*p = rte;
1288 	}
1289 
1290 	rte->next = NULL;
1291 	rte->m = mb0;
1292 	rte->ifp = ifp;
1293 #ifdef UPCALL_TIMING
1294 	rte->t = tp;
1295 #endif /* UPCALL_TIMING */
1296 
1297 	MFC6_UNLOCK();
1298 
1299 	return (0);
1300 }
1301 
1302 /*
1303  * Clean up cache entries if upcalls are not serviced
1304  * Call from the Slow Timeout mechanism, every half second.
1305  */
1306 static void
1307 expire_upcalls(void *unused)
1308 {
1309 #ifdef MRT6DEBUG
1310 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1311 #endif
1312 	struct rtdetq *rte;
1313 	struct mf6c *mfc, **nptr;
1314 	u_long i;
1315 
1316 	MFC6_LOCK_ASSERT();
1317 
1318 	for (i = 0; i < MF6CTBLSIZ; i++) {
1319 		if (n6expire[i] == 0)
1320 			continue;
1321 		nptr = &mf6ctable[i];
1322 		while ((mfc = *nptr) != NULL) {
1323 			rte = mfc->mf6c_stall;
1324 			/*
1325 			 * Skip real cache entries
1326 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1327 			 * If it expires now
1328 			 */
1329 			if (rte != NULL &&
1330 			    mfc->mf6c_expire != 0 &&
1331 			    --mfc->mf6c_expire == 0) {
1332 				MRT6_DLOG(DEBUG_EXPIRE, "expiring (%s %s)",
1333 				    ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr),
1334 				    ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr));
1335 				/*
1336 				 * drop all the packets
1337 				 * free the mbuf with the pkt, if, timing info
1338 				 */
1339 				do {
1340 					struct rtdetq *n = rte->next;
1341 					m_freem(rte->m);
1342 					free(rte, M_MRTABLE6);
1343 					rte = n;
1344 				} while (rte != NULL);
1345 				MRT6STAT_INC(mrt6s_cache_cleanups);
1346 				n6expire[i]--;
1347 
1348 				*nptr = mfc->mf6c_next;
1349 				free(mfc, M_MRTABLE6);
1350 			} else {
1351 				nptr = &mfc->mf6c_next;
1352 			}
1353 		}
1354 	}
1355 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1356 	    expire_upcalls, NULL);
1357 }
1358 
1359 /*
1360  * Packet forwarding routine once entry in the cache is made
1361  */
1362 static int
1363 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1364 {
1365 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1366 	mifi_t mifi, iif;
1367 	struct mif6 *mifp;
1368 	int plen = m->m_pkthdr.len;
1369 	struct in6_addr src0, dst0; /* copies for local work */
1370 	u_int32_t iszone, idzone, oszone, odzone;
1371 	int error = 0;
1372 
1373 	/*
1374 	 * Don't forward if it didn't arrive from the parent mif
1375 	 * for its origin.
1376 	 */
1377 	mifi = rt->mf6c_parent;
1378 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1379 		/* came in the wrong interface */
1380 		MRT6_DLOG(DEBUG_FORWARD,
1381 		    "wrong if: ifid %d mifi %d mififid %x", ifp->if_index,
1382 		    mifi, mif6table[mifi].m6_ifp->if_index);
1383 		MRT6STAT_INC(mrt6s_wrong_if);
1384 		rt->mf6c_wrong_if++;
1385 		/*
1386 		 * If we are doing PIM processing, and we are forwarding
1387 		 * packets on this interface, send a message to the
1388 		 * routing daemon.
1389 		 */
1390 		/* have to make sure this is a valid mif */
1391 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1392 			if (V_pim6 && (m->m_flags & M_LOOP) == 0) {
1393 				/*
1394 				 * Check the M_LOOP flag to avoid an
1395 				 * unnecessary PIM assert.
1396 				 * XXX: M_LOOP is an ad-hoc hack...
1397 				 */
1398 				static struct sockaddr_in6 sin6 =
1399 				{ sizeof(sin6), AF_INET6 };
1400 
1401 				struct mbuf *mm;
1402 				struct mrt6msg *im;
1403 #ifdef MRT6_OINIT
1404 				struct omrt6msg *oim;
1405 #endif
1406 
1407 				mm = m_copym(m, 0, sizeof(struct ip6_hdr),
1408 				    M_NOWAIT);
1409 				if (mm &&
1410 				    (!M_WRITABLE(mm) ||
1411 				     mm->m_len < sizeof(struct ip6_hdr)))
1412 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1413 				if (mm == NULL)
1414 					return (ENOBUFS);
1415 
1416 #ifdef MRT6_OINIT
1417 				oim = NULL;
1418 #endif
1419 				im = NULL;
1420 				switch (V_ip6_mrouter_ver) {
1421 #ifdef MRT6_OINIT
1422 				case MRT6_OINIT:
1423 					oim = mtod(mm, struct omrt6msg *);
1424 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1425 					oim->im6_mbz = 0;
1426 					break;
1427 #endif
1428 				case MRT6_INIT:
1429 					im = mtod(mm, struct mrt6msg *);
1430 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1431 					im->im6_mbz = 0;
1432 					break;
1433 				default:
1434 					m_freem(mm);
1435 					return (EINVAL);
1436 				}
1437 
1438 				for (mifp = mif6table, iif = 0;
1439 				     iif < nummifs && mifp &&
1440 					     mifp->m6_ifp != ifp;
1441 				     mifp++, iif++)
1442 					;
1443 
1444 				switch (V_ip6_mrouter_ver) {
1445 #ifdef MRT6_OINIT
1446 				case MRT6_OINIT:
1447 					oim->im6_mif = iif;
1448 					sin6.sin6_addr = oim->im6_src;
1449 					break;
1450 #endif
1451 				case MRT6_INIT:
1452 					im->im6_mif = iif;
1453 					sin6.sin6_addr = im->im6_src;
1454 					break;
1455 				}
1456 
1457 				MRT6STAT_INC(mrt6s_upcalls);
1458 
1459 				if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1460 					MRT6_DLOG(DEBUG_ANY,
1461 					    "ip6_mrouter socket queue full");
1462 					MRT6STAT_INC(mrt6s_upq_sockfull);
1463 					return (ENOBUFS);
1464 				}	/* if socket Q full */
1465 			}		/* if PIM */
1466 		return (0);
1467 	}			/* if wrong iif */
1468 
1469 	/* If I sourced this packet, it counts as output, else it was input. */
1470 	if (m->m_pkthdr.rcvif == NULL) {
1471 		/* XXX: is rcvif really NULL when output?? */
1472 		mif6table[mifi].m6_pkt_out++;
1473 		mif6table[mifi].m6_bytes_out += plen;
1474 	} else {
1475 		mif6table[mifi].m6_pkt_in++;
1476 		mif6table[mifi].m6_bytes_in += plen;
1477 	}
1478 	rt->mf6c_pkt_cnt++;
1479 	rt->mf6c_byte_cnt += plen;
1480 
1481 	/*
1482 	 * For each mif, forward a copy of the packet if there are group
1483 	 * members downstream on the interface.
1484 	 */
1485 	src0 = ip6->ip6_src;
1486 	dst0 = ip6->ip6_dst;
1487 	if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 ||
1488 	    (error = in6_setscope(&dst0, ifp, &idzone)) != 0) {
1489 		IP6STAT_INC(ip6s_badscope);
1490 		return (error);
1491 	}
1492 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) {
1493 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1494 			/*
1495 			 * check if the outgoing packet is going to break
1496 			 * a scope boundary.
1497 			 * XXX For packets through PIM register tunnel
1498 			 * interface, we believe a routing daemon.
1499 			 */
1500 			if (!(mif6table[rt->mf6c_parent].m6_flags &
1501 			      MIFF_REGISTER) &&
1502 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
1503 				if (in6_setscope(&src0, mif6table[mifi].m6_ifp,
1504 				    &oszone) ||
1505 				    in6_setscope(&dst0, mif6table[mifi].m6_ifp,
1506 				    &odzone) ||
1507 				    iszone != oszone ||
1508 				    idzone != odzone) {
1509 					IP6STAT_INC(ip6s_badscope);
1510 					continue;
1511 				}
1512 			}
1513 
1514 			mifp->m6_pkt_out++;
1515 			mifp->m6_bytes_out += plen;
1516 			if (mifp->m6_flags & MIFF_REGISTER)
1517 				register_send(ip6, mifp, m);
1518 			else
1519 				phyint_send(ip6, mifp, m);
1520 		}
1521 	}
1522 	return (0);
1523 }
1524 
1525 static void
1526 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1527 {
1528 #ifdef MRT6DEBUG
1529 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1530 #endif
1531 	struct mbuf *mb_copy;
1532 	struct ifnet *ifp = mifp->m6_ifp;
1533 	int error = 0;
1534 	u_long linkmtu;
1535 
1536 	/*
1537 	 * Make a new reference to the packet; make sure that
1538 	 * the IPv6 header is actually copied, not just referenced,
1539 	 * so that ip6_output() only scribbles on the copy.
1540 	 */
1541 	mb_copy = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1542 	if (mb_copy &&
1543 	    (!M_WRITABLE(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1544 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1545 	if (mb_copy == NULL) {
1546 		return;
1547 	}
1548 	/* set MCAST flag to the outgoing packet */
1549 	mb_copy->m_flags |= M_MCAST;
1550 
1551 	/*
1552 	 * If we sourced the packet, call ip6_output since we may devide
1553 	 * the packet into fragments when the packet is too big for the
1554 	 * outgoing interface.
1555 	 * Otherwise, we can simply send the packet to the interface
1556 	 * sending queue.
1557 	 */
1558 	if (m->m_pkthdr.rcvif == NULL) {
1559 		struct ip6_moptions im6o;
1560 		struct epoch_tracker et;
1561 
1562 		im6o.im6o_multicast_ifp = ifp;
1563 		/* XXX: ip6_output will override ip6->ip6_hlim */
1564 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1565 		im6o.im6o_multicast_loop = 1;
1566 		NET_EPOCH_ENTER(et);
1567 		error = ip6_output(mb_copy, NULL, NULL, IPV6_FORWARDING, &im6o,
1568 		    NULL, NULL);
1569 		NET_EPOCH_EXIT(et);
1570 
1571 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1572 		    (uint16_t)(mifp - mif6table), error);
1573 		return;
1574 	}
1575 
1576 	/*
1577 	 * If configured to loop back multicasts by default,
1578 	 * loop back a copy now.
1579 	 */
1580 	if (in6_mcast_loop)
1581 		ip6_mloopback(ifp, m);
1582 
1583 	/*
1584 	 * Put the packet into the sending queue of the outgoing interface
1585 	 * if it would fit in the MTU of the interface.
1586 	 */
1587 	linkmtu = IN6_LINKMTU(ifp);
1588 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1589 		struct sockaddr_in6 dst6;
1590 
1591 		bzero(&dst6, sizeof(dst6));
1592 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1593 		dst6.sin6_family = AF_INET6;
1594 		dst6.sin6_addr = ip6->ip6_dst;
1595 
1596 		IP_PROBE(send, NULL, NULL, ip6, ifp, NULL, ip6);
1597 		/*
1598 		 * We just call if_output instead of nd6_output here, since
1599 		 * we need no ND for a multicast forwarded packet...right?
1600 		 */
1601 		m_clrprotoflags(m);	/* Avoid confusing lower layers. */
1602 		error = (*ifp->if_output)(ifp, mb_copy,
1603 		    (struct sockaddr *)&dst6, NULL);
1604 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1605 		    (uint16_t)(mifp - mif6table), error);
1606 	} else {
1607 		/*
1608 		 * pMTU discovery is intentionally disabled by default, since
1609 		 * various router may notify pMTU in multicast, which can be
1610 		 * a DDoS to a router
1611 		 */
1612 		if (V_ip6_mcast_pmtu)
1613 			icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1614 		else {
1615 			MRT6_DLOG(DEBUG_XMIT, " packet too big on %s o %s "
1616 			    "g %s size %d (discarded)", if_name(ifp),
1617 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1618 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1619 			    mb_copy->m_pkthdr.len);
1620 			m_freem(mb_copy); /* simply discard the packet */
1621 		}
1622 	}
1623 }
1624 
1625 static int
1626 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1627 {
1628 #ifdef MRT6DEBUG
1629 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1630 #endif
1631 	struct mbuf *mm;
1632 	int i, len = m->m_pkthdr.len;
1633 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1634 	struct mrt6msg *im6;
1635 
1636 	MRT6_DLOG(DEBUG_ANY, "src %s dst %s",
1637 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1638 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1639 	PIM6STAT_INC(pim6s_snd_registers);
1640 
1641 	/* Make a copy of the packet to send to the user level process. */
1642 	mm = m_gethdr(M_NOWAIT, MT_DATA);
1643 	if (mm == NULL)
1644 		return (ENOBUFS);
1645 	mm->m_data += max_linkhdr;
1646 	mm->m_len = sizeof(struct ip6_hdr);
1647 
1648 	if ((mm->m_next = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
1649 		m_freem(mm);
1650 		return (ENOBUFS);
1651 	}
1652 	i = MHLEN - M_LEADINGSPACE(mm);
1653 	if (i > len)
1654 		i = len;
1655 	mm = m_pullup(mm, i);
1656 	if (mm == NULL)
1657 		return (ENOBUFS);
1658 /* TODO: check it! */
1659 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1660 
1661 	/*
1662 	 * Send message to routing daemon
1663 	 */
1664 	sin6.sin6_addr = ip6->ip6_src;
1665 
1666 	im6 = mtod(mm, struct mrt6msg *);
1667 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1668 	im6->im6_mbz          = 0;
1669 
1670 	im6->im6_mif = mif - mif6table;
1671 
1672 	/* iif info is not given for reg. encap.n */
1673 	MRT6STAT_INC(mrt6s_upcalls);
1674 
1675 	if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1676 		MRT6_DLOG(DEBUG_ANY, "ip6_mrouter socket queue full");
1677 		MRT6STAT_INC(mrt6s_upq_sockfull);
1678 		return (ENOBUFS);
1679 	}
1680 	return (0);
1681 }
1682 
1683 /*
1684  * pim6_encapcheck() is called by the encap6_input() path at runtime to
1685  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
1686  * into the kernel.
1687  */
1688 static int
1689 pim6_encapcheck(const struct mbuf *m __unused, int off __unused,
1690     int proto __unused, void *arg __unused)
1691 {
1692 
1693     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
1694     return (8);		/* claim the datagram. */
1695 }
1696 
1697 /*
1698  * PIM sparse mode hook
1699  * Receives the pim control messages, and passes them up to the listening
1700  * socket, using rip6_input.
1701  * The only message processed is the REGISTER pim message; the pim header
1702  * is stripped off, and the inner packet is passed to register_mforward.
1703  */
1704 static int
1705 pim6_input(struct mbuf *m, int off, int proto, void *arg __unused)
1706 {
1707 	struct pim *pim; /* pointer to a pim struct */
1708 	struct ip6_hdr *ip6;
1709 	int pimlen;
1710 	int minlen;
1711 
1712 	PIM6STAT_INC(pim6s_rcv_total);
1713 
1714 	/*
1715 	 * Validate lengths
1716 	 */
1717 	pimlen = m->m_pkthdr.len - off;
1718 	if (pimlen < PIM_MINLEN) {
1719 		PIM6STAT_INC(pim6s_rcv_tooshort);
1720 		MRT6_DLOG(DEBUG_PIM, "PIM packet too short");
1721 		m_freem(m);
1722 		return (IPPROTO_DONE);
1723 	}
1724 
1725 	/*
1726 	 * if the packet is at least as big as a REGISTER, go ahead
1727 	 * and grab the PIM REGISTER header size, to avoid another
1728 	 * possible m_pullup() later.
1729 	 *
1730 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1731 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1732 	 */
1733 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1734 
1735 	/*
1736 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1737 	 * possibly the PIM REGISTER header
1738 	 */
1739 	if (m->m_len < off + minlen) {
1740 		m = m_pullup(m, off + minlen);
1741 		if (m == NULL) {
1742 			IP6STAT_INC(ip6s_exthdrtoolong);
1743 			return (IPPROTO_DONE);
1744 		}
1745 	}
1746 	ip6 = mtod(m, struct ip6_hdr *);
1747 	pim = (struct pim *)((caddr_t)ip6 + off);
1748 
1749 #define PIM6_CHECKSUM
1750 #ifdef PIM6_CHECKSUM
1751 	{
1752 		int cksumlen;
1753 
1754 		/*
1755 		 * Validate checksum.
1756 		 * If PIM REGISTER, exclude the data packet
1757 		 */
1758 		if (pim->pim_type == PIM_REGISTER)
1759 			cksumlen = PIM_MINLEN;
1760 		else
1761 			cksumlen = pimlen;
1762 
1763 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1764 			PIM6STAT_INC(pim6s_rcv_badsum);
1765 			MRT6_DLOG(DEBUG_PIM, "invalid checksum");
1766 			m_freem(m);
1767 			return (IPPROTO_DONE);
1768 		}
1769 	}
1770 #endif /* PIM_CHECKSUM */
1771 
1772 	/* PIM version check */
1773 	if (pim->pim_ver != PIM_VERSION) {
1774 		PIM6STAT_INC(pim6s_rcv_badversion);
1775 		MRT6_DLOG(DEBUG_ANY | DEBUG_ERR,
1776 		    "incorrect version %d, expecting %d",
1777 		    pim->pim_ver, PIM_VERSION);
1778 		m_freem(m);
1779 		return (IPPROTO_DONE);
1780 	}
1781 
1782 	if (pim->pim_type == PIM_REGISTER) {
1783 		/*
1784 		 * since this is a REGISTER, we'll make a copy of the register
1785 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1786 		 * routing daemon.
1787 		 */
1788 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1789 
1790 		struct mbuf *mcp;
1791 		struct ip6_hdr *eip6;
1792 		u_int32_t *reghdr;
1793 		int rc;
1794 #ifdef MRT6DEBUG
1795 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1796 #endif
1797 
1798 		PIM6STAT_INC(pim6s_rcv_registers);
1799 
1800 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1801 			MRT6_DLOG(DEBUG_PIM, "register mif not set: %d",
1802 			    reg_mif_num);
1803 			m_freem(m);
1804 			return (IPPROTO_DONE);
1805 		}
1806 
1807 		reghdr = (u_int32_t *)(pim + 1);
1808 
1809 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1810 			goto pim6_input_to_daemon;
1811 
1812 		/*
1813 		 * Validate length
1814 		 */
1815 		if (pimlen < PIM6_REG_MINLEN) {
1816 			PIM6STAT_INC(pim6s_rcv_tooshort);
1817 			PIM6STAT_INC(pim6s_rcv_badregisters);
1818 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "register packet "
1819 			    "size too small %d from %s",
1820 			    pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src));
1821 			m_freem(m);
1822 			return (IPPROTO_DONE);
1823 		}
1824 
1825 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1826 		MRT6_DLOG(DEBUG_PIM, "eip6: %s -> %s, eip6 plen %d",
1827 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1828 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1829 		    ntohs(eip6->ip6_plen));
1830 
1831 		/* verify the version number of the inner packet */
1832 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1833 			PIM6STAT_INC(pim6s_rcv_badregisters);
1834 			MRT6_DLOG(DEBUG_ANY, "invalid IP version (%d) "
1835 			    "of the inner packet",
1836 			    (eip6->ip6_vfc & IPV6_VERSION));
1837 			m_freem(m);
1838 			return (IPPROTO_DONE);
1839 		}
1840 
1841 		/* verify the inner packet is destined to a mcast group */
1842 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1843 			PIM6STAT_INC(pim6s_rcv_badregisters);
1844 			MRT6_DLOG(DEBUG_PIM, "inner packet of register "
1845 			    "is not multicast %s",
1846 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst));
1847 			m_freem(m);
1848 			return (IPPROTO_DONE);
1849 		}
1850 
1851 		/*
1852 		 * make a copy of the whole header to pass to the daemon later.
1853 		 */
1854 		mcp = m_copym(m, 0, off + PIM6_REG_MINLEN, M_NOWAIT);
1855 		if (mcp == NULL) {
1856 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "pim register: "
1857 			    "could not copy register head");
1858 			m_freem(m);
1859 			return (IPPROTO_DONE);
1860 		}
1861 
1862 		/*
1863 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1864 		 */
1865 		m_adj(m, off + PIM_MINLEN);
1866 		MRT6_DLOG(DEBUG_PIM, "forwarding decapsulated register: "
1867 		    "src %s, dst %s, mif %d",
1868 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1869 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst), reg_mif_num);
1870 
1871 		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1872 				dst.sin6_family, 0);
1873 
1874 		/* prepare the register head to send to the mrouting daemon */
1875 		m = mcp;
1876 	}
1877 
1878 	/*
1879 	 * Pass the PIM message up to the daemon; if it is a register message
1880 	 * pass the 'head' only up to the daemon. This includes the
1881 	 * encapsulator ip6 header, pim header, register header and the
1882 	 * encapsulated ip6 header.
1883 	 */
1884   pim6_input_to_daemon:
1885 	return (rip6_input(&m, &off, proto));
1886 }
1887 
1888 static int
1889 ip6_mroute_modevent(module_t mod, int type, void *unused)
1890 {
1891 
1892 	switch (type) {
1893 	case MOD_LOAD:
1894 		MROUTER6_LOCK_INIT();
1895 		MFC6_LOCK_INIT();
1896 		MIF6_LOCK_INIT();
1897 
1898 		pim6_encap_cookie = ip6_encap_attach(&ipv6_encap_cfg,
1899 		    NULL, M_WAITOK);
1900 		if (pim6_encap_cookie == NULL) {
1901 			printf("ip6_mroute: unable to attach pim6 encap\n");
1902 			MIF6_LOCK_DESTROY();
1903 			MFC6_LOCK_DESTROY();
1904 			MROUTER6_LOCK_DESTROY();
1905 			return (EINVAL);
1906 		}
1907 
1908 		ip6_mforward = X_ip6_mforward;
1909 		ip6_mrouter_done = X_ip6_mrouter_done;
1910 		ip6_mrouter_get = X_ip6_mrouter_get;
1911 		ip6_mrouter_set = X_ip6_mrouter_set;
1912 		mrt6_ioctl = X_mrt6_ioctl;
1913 		break;
1914 
1915 	case MOD_UNLOAD:
1916 		if (V_ip6_mrouter != NULL)
1917 			return EINVAL;
1918 
1919 		if (pim6_encap_cookie) {
1920 			ip6_encap_detach(pim6_encap_cookie);
1921 			pim6_encap_cookie = NULL;
1922 		}
1923 		X_ip6_mrouter_done();
1924 		ip6_mforward = NULL;
1925 		ip6_mrouter_done = NULL;
1926 		ip6_mrouter_get = NULL;
1927 		ip6_mrouter_set = NULL;
1928 		mrt6_ioctl = NULL;
1929 
1930 		MIF6_LOCK_DESTROY();
1931 		MFC6_LOCK_DESTROY();
1932 		MROUTER6_LOCK_DESTROY();
1933 		break;
1934 
1935 	default:
1936 		return (EOPNOTSUPP);
1937 	}
1938 
1939 	return (0);
1940 }
1941 
1942 static moduledata_t ip6_mroutemod = {
1943 	"ip6_mroute",
1944 	ip6_mroute_modevent,
1945 	0
1946 };
1947 
1948 DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_ANY);
1949