xref: /freebsd/sys/netinet6/ip6_mroute.c (revision 7fdf597e96a02165cfe22ff357b857d5fa15ed8a)
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  *	BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp
66  */
67 
68 /*
69  * IP multicast forwarding procedures
70  *
71  * Written by David Waitzman, BBN Labs, August 1988.
72  * Modified by Steve Deering, Stanford, February 1989.
73  * Modified by Mark J. Steiglitz, Stanford, May, 1991
74  * Modified by Van Jacobson, LBL, January 1993
75  * Modified by Ajit Thyagarajan, PARC, August 1993
76  * Modified by Bill Fenner, PARC, April 1994
77  *
78  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
79  */
80 
81 #include <sys/cdefs.h>
82 #include "opt_inet6.h"
83 
84 #include <sys/param.h>
85 #include <sys/callout.h>
86 #include <sys/errno.h>
87 #include <sys/kernel.h>
88 #include <sys/lock.h>
89 #include <sys/malloc.h>
90 #include <sys/mbuf.h>
91 #include <sys/module.h>
92 #include <sys/domain.h>
93 #include <sys/protosw.h>
94 #include <sys/sdt.h>
95 #include <sys/signalvar.h>
96 #include <sys/socket.h>
97 #include <sys/socketvar.h>
98 #include <sys/sockio.h>
99 #include <sys/sx.h>
100 #include <sys/sysctl.h>
101 #include <sys/syslog.h>
102 #include <sys/systm.h>
103 #include <sys/time.h>
104 
105 #include <net/if.h>
106 #include <net/if_var.h>
107 #include <net/if_private.h>
108 #include <net/if_types.h>
109 #include <net/vnet.h>
110 
111 #include <netinet/in.h>
112 #include <netinet/in_var.h>
113 #include <netinet/icmp6.h>
114 #include <netinet/ip_encap.h>
115 
116 #include <netinet/ip6.h>
117 #include <netinet/in_kdtrace.h>
118 #include <netinet6/ip6_var.h>
119 #include <netinet6/scope6_var.h>
120 #include <netinet6/nd6.h>
121 #include <netinet6/ip6_mroute.h>
122 #include <netinet6/pim6.h>
123 #include <netinet6/pim6_var.h>
124 
125 static MALLOC_DEFINE(M_MRTABLE6, "mf6c", "multicast forwarding cache entry");
126 
127 static int	ip6_mdq(struct mbuf *, struct ifnet *, struct mf6c *);
128 static void	phyint_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
129 static int	register_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
130 static int	set_pim6(int *);
131 static int	socket_send(struct socket *, struct mbuf *,
132 		    struct sockaddr_in6 *);
133 
134 extern int in6_mcast_loop;
135 extern struct domain inet6domain;
136 
137 static const struct encaptab *pim6_encap_cookie;
138 static int pim6_encapcheck(const struct mbuf *, int, int, void *);
139 static int pim6_input(struct mbuf *, int, int, void *);
140 
141 static const struct encap_config ipv6_encap_cfg = {
142 	.proto = IPPROTO_PIM,
143 	.min_length = sizeof(struct ip6_hdr) + PIM_MINLEN,
144 	.exact_match = 8,
145 	.check = pim6_encapcheck,
146 	.input = pim6_input
147 };
148 
149 VNET_DEFINE_STATIC(int, ip6_mrouter_ver) = 0;
150 #define	V_ip6_mrouter_ver	VNET(ip6_mrouter_ver)
151 
152 SYSCTL_DECL(_net_inet6);
153 SYSCTL_DECL(_net_inet6_ip6);
154 static SYSCTL_NODE(_net_inet6, IPPROTO_PIM, pim,
155     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
156     "PIM");
157 
158 static struct mrt6stat mrt6stat;
159 SYSCTL_STRUCT(_net_inet6_ip6, OID_AUTO, mrt6stat, CTLFLAG_RW,
160     &mrt6stat, mrt6stat,
161     "Multicast Routing Statistics (struct mrt6stat, netinet6/ip6_mroute.h)");
162 
163 #define	MRT6STAT_INC(name)	mrt6stat.name += 1
164 #define NO_RTE_FOUND	0x1
165 #define RTE_FOUND	0x2
166 
167 static struct sx mrouter6_mtx;
168 #define	MROUTER6_LOCKPTR()	(&mrouter6_mtx)
169 #define	MROUTER6_LOCK()		sx_xlock(MROUTER6_LOCKPTR())
170 #define	MROUTER6_UNLOCK()	sx_xunlock(MROUTER6_LOCKPTR())
171 #define	MROUTER6_LOCK_ASSERT()	sx_assert(MROUTER6_LOCKPTR(), SA_XLOCKED
172 #define	MROUTER6_LOCK_INIT()	sx_init(MROUTER6_LOCKPTR(), "mrouter6")
173 #define	MROUTER6_LOCK_DESTROY()	sx_destroy(MROUTER6_LOCKPTR())
174 
175 static struct mf6c *mf6ctable[MF6CTBLSIZ];
176 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mf6ctable, CTLFLAG_RD,
177     &mf6ctable, sizeof(mf6ctable), "S,*mf6ctable[MF6CTBLSIZ]",
178     "IPv6 Multicast Forwarding Table (struct *mf6ctable[MF6CTBLSIZ], "
179     "netinet6/ip6_mroute.h)");
180 
181 static struct mtx mfc6_mtx;
182 #define	MFC6_LOCKPTR()		(&mfc6_mtx)
183 #define	MFC6_LOCK()		mtx_lock(MFC6_LOCKPTR())
184 #define	MFC6_UNLOCK()		mtx_unlock(MFC6_LOCKPTR())
185 #define	MFC6_LOCK_ASSERT()	mtx_assert(MFC6_LOCKPTR(), MA_OWNED)
186 #define	MFC6_LOCK_INIT()	mtx_init(MFC6_LOCKPTR(),		\
187 				    "IPv6 multicast forwarding cache",	\
188 				    NULL, MTX_DEF)
189 #define	MFC6_LOCK_DESTROY()	mtx_destroy(MFC6_LOCKPTR())
190 
191 static u_char n6expire[MF6CTBLSIZ];
192 
193 static struct mif6 mif6table[MAXMIFS];
194 static int
195 sysctl_mif6table(SYSCTL_HANDLER_ARGS)
196 {
197 	struct mif6_sctl *out;
198 	int error;
199 
200 	out = malloc(sizeof(struct mif6_sctl) * MAXMIFS, M_TEMP,
201 	    M_WAITOK | M_ZERO);
202 	for (int i = 0; i < MAXMIFS; i++) {
203 		out[i].m6_flags		= mif6table[i].m6_flags;
204 		out[i].m6_rate_limit	= mif6table[i].m6_rate_limit;
205 		out[i].m6_lcl_addr	= mif6table[i].m6_lcl_addr;
206 		if (mif6table[i].m6_ifp != NULL)
207 			out[i].m6_ifp	= mif6table[i].m6_ifp->if_index;
208 		else
209 			out[i].m6_ifp	= 0;
210 		out[i].m6_pkt_in	= mif6table[i].m6_pkt_in;
211 		out[i].m6_pkt_out	= mif6table[i].m6_pkt_out;
212 		out[i].m6_bytes_in	= mif6table[i].m6_bytes_in;
213 		out[i].m6_bytes_out	= mif6table[i].m6_bytes_out;
214 	}
215 	error = SYSCTL_OUT(req, out, sizeof(struct mif6_sctl) * MAXMIFS);
216 	free(out, M_TEMP);
217 	return (error);
218 }
219 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, mif6table,
220     CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
221     NULL, 0, sysctl_mif6table, "S,mif6_sctl[MAXMIFS]",
222     "IPv6 Multicast Interfaces (struct mif6_sctl[MAXMIFS], "
223     "netinet6/ip6_mroute.h)");
224 
225 static struct mtx mif6_mtx;
226 #define	MIF6_LOCKPTR()		(&mif6_mtx)
227 #define	MIF6_LOCK()		mtx_lock(MIF6_LOCKPTR())
228 #define	MIF6_UNLOCK()		mtx_unlock(MIF6_LOCKPTR())
229 #define	MIF6_LOCK_ASSERT()	mtx_assert(MIF6_LOCKPTR(), MA_OWNED)
230 #define	MIF6_LOCK_INIT()	\
231 	mtx_init(MIF6_LOCKPTR(), "IPv6 multicast interfaces", NULL, MTX_DEF)
232 #define	MIF6_LOCK_DESTROY()	mtx_destroy(MIF6_LOCKPTR())
233 
234 #ifdef MRT6DEBUG
235 VNET_DEFINE_STATIC(u_int, mrt6debug) = 0;	/* debug level */
236 #define	V_mrt6debug		VNET(mrt6debug)
237 #define DEBUG_MFC	0x02
238 #define DEBUG_FORWARD	0x04
239 #define DEBUG_EXPIRE	0x08
240 #define DEBUG_XMIT	0x10
241 #define DEBUG_REG	0x20
242 #define DEBUG_PIM	0x40
243 #define	DEBUG_ERR	0x80
244 #define	DEBUG_ANY	0x7f
245 #define	MRT6_DLOG(m, fmt, ...)	\
246 	if (V_mrt6debug & (m))	\
247 		log(((m) & DEBUG_ERR) ? LOG_ERR: LOG_DEBUG, \
248 		    "%s: " fmt "\n", __func__, ##__VA_ARGS__)
249 #else
250 #define	MRT6_DLOG(m, fmt, ...)
251 #endif
252 
253 static void	expire_upcalls(void *);
254 #define	EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
255 #define	UPCALL_EXPIRE	6		/* number of timeouts */
256 
257 /*
258  * XXX TODO: maintain a count to if_allmulti() calls in struct ifnet.
259  */
260 
261 /*
262  * 'Interfaces' associated with decapsulator (so we can tell
263  * packets that went through it from ones that get reflected
264  * by a broken gateway).  Different from IPv4 register_if,
265  * these interfaces are linked into the system ifnet list,
266  * because per-interface IPv6 statistics are maintained in
267  * ifp->if_afdata.  But it does not have any routes point
268  * to them.  I.e., packets can't be sent this way.  They
269  * only exist as a placeholder for multicast source
270  * verification.
271  */
272 static struct ifnet *multicast_register_if6;
273 
274 #define ENCAP_HOPS 64
275 
276 /*
277  * Private variables.
278  */
279 static mifi_t nummifs = 0;
280 static mifi_t reg_mif_num = (mifi_t)-1;
281 
282 static struct pim6stat pim6stat;
283 SYSCTL_STRUCT(_net_inet6_pim, PIM6CTL_STATS, stats, CTLFLAG_RW,
284     &pim6stat, pim6stat,
285     "PIM Statistics (struct pim6stat, netinet6/pim6_var.h)");
286 
287 #define	PIM6STAT_INC(name)	pim6stat.name += 1
288 VNET_DEFINE_STATIC(int, pim6);
289 #define	V_pim6		VNET(pim6)
290 
291 /*
292  * Hash function for a source, group entry
293  */
294 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
295 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
296 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
297 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
298 
299 /*
300  * Find a route for a given origin IPv6 address and Multicast group address.
301  */
302 #define MF6CFIND(o, g, rt) do { \
303 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
304 	rt = NULL; \
305 	while (_rt) { \
306 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
307 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
308 		    (_rt->mf6c_stall == NULL)) { \
309 			rt = _rt; \
310 			break; \
311 		} \
312 		_rt = _rt->mf6c_next; \
313 	} \
314 	if (rt == NULL) { \
315 		MRT6STAT_INC(mrt6s_mfc_misses); \
316 	} \
317 } while (/*CONSTCOND*/ 0)
318 
319 /*
320  * Macros to compute elapsed time efficiently
321  * Borrowed from Van Jacobson's scheduling code
322  * XXX: replace with timersub() ?
323  */
324 #define TV_DELTA(a, b, delta) do { \
325 	    int xxs; \
326 		\
327 	    delta = (a).tv_usec - (b).tv_usec; \
328 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
329 	       switch (xxs) { \
330 		      case 2: \
331 			  delta += 1000000; \
332 			      /* FALLTHROUGH */ \
333 		      case 1: \
334 			  delta += 1000000; \
335 			  break; \
336 		      default: \
337 			  delta += (1000000 * xxs); \
338 	       } \
339 	    } \
340 } while (/*CONSTCOND*/ 0)
341 
342 /* XXX: replace with timercmp(a, b, <) ? */
343 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
344 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
345 
346 #ifdef UPCALL_TIMING
347 #define UPCALL_MAX	50
348 static u_long upcall_data[UPCALL_MAX + 1];
349 static void collate();
350 #endif /* UPCALL_TIMING */
351 
352 static int ip6_mrouter_init(struct socket *, int, int);
353 static int add_m6fc(struct mf6cctl *);
354 static int add_m6if(struct mif6ctl *);
355 static int del_m6fc(struct mf6cctl *);
356 static int del_m6if(mifi_t *);
357 static int del_m6if_locked(mifi_t *);
358 static int get_mif6_cnt(struct sioc_mif_req6 *);
359 static int get_sg_cnt(struct sioc_sg_req6 *);
360 
361 static struct callout expire_upcalls_ch;
362 
363 int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *);
364 int X_ip6_mrouter_done(void);
365 int X_ip6_mrouter_set(struct socket *, struct sockopt *);
366 int X_ip6_mrouter_get(struct socket *, struct sockopt *);
367 int X_mrt6_ioctl(u_long, caddr_t);
368 
369 /*
370  * Handle MRT setsockopt commands to modify the multicast routing tables.
371  */
372 int
373 X_ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
374 {
375 	int error = 0;
376 	int optval;
377 	struct mif6ctl mifc;
378 	struct mf6cctl mfcc;
379 	mifi_t mifi;
380 
381 	if (so != V_ip6_mrouter && sopt->sopt_name != MRT6_INIT)
382 		return (EPERM);
383 
384 	switch (sopt->sopt_name) {
385 	case MRT6_INIT:
386 #ifdef MRT6_OINIT
387 	case MRT6_OINIT:
388 #endif
389 		error = sooptcopyin(sopt, &optval, sizeof(optval),
390 		    sizeof(optval));
391 		if (error)
392 			break;
393 		error = ip6_mrouter_init(so, optval, sopt->sopt_name);
394 		break;
395 	case MRT6_DONE:
396 		error = X_ip6_mrouter_done();
397 		break;
398 	case MRT6_ADD_MIF:
399 		error = sooptcopyin(sopt, &mifc, sizeof(mifc), sizeof(mifc));
400 		if (error)
401 			break;
402 		error = add_m6if(&mifc);
403 		break;
404 	case MRT6_ADD_MFC:
405 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
406 		if (error)
407 			break;
408 		error = add_m6fc(&mfcc);
409 		break;
410 	case MRT6_DEL_MFC:
411 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
412 		if (error)
413 			break;
414 		error = del_m6fc(&mfcc);
415 		break;
416 	case MRT6_DEL_MIF:
417 		error = sooptcopyin(sopt, &mifi, sizeof(mifi), sizeof(mifi));
418 		if (error)
419 			break;
420 		error = del_m6if(&mifi);
421 		break;
422 	case MRT6_PIM:
423 		error = sooptcopyin(sopt, &optval, sizeof(optval),
424 		    sizeof(optval));
425 		if (error)
426 			break;
427 		error = set_pim6(&optval);
428 		break;
429 	default:
430 		error = EOPNOTSUPP;
431 		break;
432 	}
433 
434 	return (error);
435 }
436 
437 /*
438  * Handle MRT getsockopt commands
439  */
440 int
441 X_ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
442 {
443 	int error = 0;
444 
445 	if (so != V_ip6_mrouter)
446 		return (EACCES);
447 
448 	switch (sopt->sopt_name) {
449 		case MRT6_PIM:
450 			error = sooptcopyout(sopt, &V_pim6, sizeof(V_pim6));
451 			break;
452 	}
453 	return (error);
454 }
455 
456 /*
457  * Handle ioctl commands to obtain information from the cache
458  */
459 int
460 X_mrt6_ioctl(u_long cmd, caddr_t data)
461 {
462 	int ret;
463 
464 	ret = EINVAL;
465 
466 	switch (cmd) {
467 	case SIOCGETSGCNT_IN6:
468 		ret = get_sg_cnt((struct sioc_sg_req6 *)data);
469 		break;
470 
471 	case SIOCGETMIFCNT_IN6:
472 		ret = get_mif6_cnt((struct sioc_mif_req6 *)data);
473 		break;
474 
475 	default:
476 		break;
477 	}
478 
479 	return (ret);
480 }
481 
482 /*
483  * returns the packet, byte, rpf-failure count for the source group provided
484  */
485 static int
486 get_sg_cnt(struct sioc_sg_req6 *req)
487 {
488 	struct mf6c *rt;
489 	int ret;
490 
491 	ret = 0;
492 
493 	MFC6_LOCK();
494 
495 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
496 	if (rt == NULL) {
497 		ret = ESRCH;
498 	} else {
499 		req->pktcnt = rt->mf6c_pkt_cnt;
500 		req->bytecnt = rt->mf6c_byte_cnt;
501 		req->wrong_if = rt->mf6c_wrong_if;
502 	}
503 
504 	MFC6_UNLOCK();
505 
506 	return (ret);
507 }
508 
509 /*
510  * returns the input and output packet and byte counts on the mif provided
511  */
512 static int
513 get_mif6_cnt(struct sioc_mif_req6 *req)
514 {
515 	mifi_t mifi;
516 	int ret;
517 
518 	ret = 0;
519 	mifi = req->mifi;
520 
521 	MIF6_LOCK();
522 
523 	if (mifi >= nummifs) {
524 		ret = EINVAL;
525 	} else {
526 		req->icount = mif6table[mifi].m6_pkt_in;
527 		req->ocount = mif6table[mifi].m6_pkt_out;
528 		req->ibytes = mif6table[mifi].m6_bytes_in;
529 		req->obytes = mif6table[mifi].m6_bytes_out;
530 	}
531 
532 	MIF6_UNLOCK();
533 
534 	return (ret);
535 }
536 
537 static int
538 set_pim6(int *i)
539 {
540 	if ((*i != 1) && (*i != 0))
541 		return (EINVAL);
542 
543 	V_pim6 = *i;
544 
545 	return (0);
546 }
547 
548 /*
549  * Enable multicast routing
550  */
551 static int
552 ip6_mrouter_init(struct socket *so, int v, int cmd)
553 {
554 
555 	MRT6_DLOG(DEBUG_ANY, "%s: socket %p", __func__, so);
556 
557 	if (v != 1)
558 		return (ENOPROTOOPT);
559 
560 	MROUTER6_LOCK();
561 
562 	if (V_ip6_mrouter != NULL) {
563 		MROUTER6_UNLOCK();
564 		return (EADDRINUSE);
565 	}
566 
567 	V_ip6_mrouter = so;
568 	V_ip6_mrouter_ver = cmd;
569 
570 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
571 	bzero((caddr_t)n6expire, sizeof(n6expire));
572 
573 	V_pim6 = 0;/* used for stubbing out/in pim stuff */
574 
575 	callout_init_mtx(&expire_upcalls_ch, MFC6_LOCKPTR(), 0);
576 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
577 	    expire_upcalls, NULL);
578 
579 	MROUTER6_UNLOCK();
580 
581 	MRT6_DLOG(DEBUG_ANY, "finished");
582 
583 	return (0);
584 }
585 
586 /*
587  * Disable IPv6 multicast forwarding.
588  */
589 int
590 X_ip6_mrouter_done(void)
591 {
592 	mifi_t mifi;
593 	u_long i;
594 	struct mf6c *rt;
595 	struct rtdetq *rte;
596 
597 	MROUTER6_LOCK();
598 
599 	if (V_ip6_mrouter == NULL) {
600 		MROUTER6_UNLOCK();
601 		return (EINVAL);
602 	}
603 
604 	/*
605 	 * For each phyint in use, disable promiscuous reception of all IPv6
606 	 * multicasts.
607 	 */
608 	for (mifi = 0; mifi < nummifs; mifi++) {
609 		if (mif6table[mifi].m6_ifp &&
610 		    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
611 			if_allmulti(mif6table[mifi].m6_ifp, 0);
612 		}
613 	}
614 	bzero((caddr_t)mif6table, sizeof(mif6table));
615 	nummifs = 0;
616 
617 	V_pim6 = 0; /* used to stub out/in pim specific code */
618 
619 	/*
620 	 * Free all multicast forwarding cache entries.
621 	 */
622 	MFC6_LOCK();
623 	for (i = 0; i < MF6CTBLSIZ; i++) {
624 		rt = mf6ctable[i];
625 		while (rt) {
626 			struct mf6c *frt;
627 
628 			for (rte = rt->mf6c_stall; rte != NULL; ) {
629 				struct rtdetq *n = rte->next;
630 
631 				m_freem(rte->m);
632 				free(rte, M_MRTABLE6);
633 				rte = n;
634 			}
635 			frt = rt;
636 			rt = rt->mf6c_next;
637 			free(frt, M_MRTABLE6);
638 		}
639 	}
640 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
641 	MFC6_UNLOCK();
642 
643 	callout_drain(&expire_upcalls_ch);
644 
645 	/*
646 	 * Reset register interface
647 	 */
648 	if (reg_mif_num != (mifi_t)-1 && multicast_register_if6 != NULL) {
649 		if_detach(multicast_register_if6);
650 		if_free(multicast_register_if6);
651 		reg_mif_num = (mifi_t)-1;
652 		multicast_register_if6 = NULL;
653 	}
654 
655 	V_ip6_mrouter = NULL;
656 	V_ip6_mrouter_ver = 0;
657 
658 	MROUTER6_UNLOCK();
659 	MRT6_DLOG(DEBUG_ANY, "finished");
660 
661 	return (0);
662 }
663 
664 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
665 
666 /*
667  * Add a mif to the mif table
668  */
669 static int
670 add_m6if(struct mif6ctl *mifcp)
671 {
672 	struct epoch_tracker et;
673 	struct mif6 *mifp;
674 	struct ifnet *ifp;
675 	int error;
676 
677 	MIF6_LOCK();
678 
679 	if (mifcp->mif6c_mifi >= MAXMIFS) {
680 		MIF6_UNLOCK();
681 		return (EINVAL);
682 	}
683 	mifp = mif6table + mifcp->mif6c_mifi;
684 	if (mifp->m6_ifp != NULL) {
685 		MIF6_UNLOCK();
686 		return (EADDRINUSE); /* XXX: is it appropriate? */
687 	}
688 
689 	NET_EPOCH_ENTER(et);
690 	if ((ifp = ifnet_byindex(mifcp->mif6c_pifi)) == NULL) {
691 		NET_EPOCH_EXIT(et);
692 		MIF6_UNLOCK();
693 		return (ENXIO);
694 	}
695 	NET_EPOCH_EXIT(et);	/* XXXGL: unsafe ifp */
696 
697 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
698 		if (reg_mif_num == (mifi_t)-1) {
699 			ifp = if_alloc(IFT_OTHER);
700 
701 			if_initname(ifp, "register_mif", 0);
702 			ifp->if_flags |= IFF_LOOPBACK;
703 			if_attach(ifp);
704 			multicast_register_if6 = ifp;
705 			reg_mif_num = mifcp->mif6c_mifi;
706 			/*
707 			 * it is impossible to guess the ifindex of the
708 			 * register interface.  So mif6c_pifi is automatically
709 			 * calculated.
710 			 */
711 			mifcp->mif6c_pifi = ifp->if_index;
712 		} else {
713 			ifp = multicast_register_if6;
714 		}
715 	} else {
716 		/* Make sure the interface supports multicast */
717 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
718 			MIF6_UNLOCK();
719 			return (EOPNOTSUPP);
720 		}
721 
722 		error = if_allmulti(ifp, 1);
723 		if (error) {
724 			MIF6_UNLOCK();
725 			return (error);
726 		}
727 	}
728 
729 	mifp->m6_flags     = mifcp->mif6c_flags;
730 	mifp->m6_ifp       = ifp;
731 
732 	/* initialize per mif pkt counters */
733 	mifp->m6_pkt_in    = 0;
734 	mifp->m6_pkt_out   = 0;
735 	mifp->m6_bytes_in  = 0;
736 	mifp->m6_bytes_out = 0;
737 
738 	/* Adjust nummifs up if the mifi is higher than nummifs */
739 	if (nummifs <= mifcp->mif6c_mifi)
740 		nummifs = mifcp->mif6c_mifi + 1;
741 
742 	MIF6_UNLOCK();
743 	MRT6_DLOG(DEBUG_ANY, "mif #%d, phyint %s", mifcp->mif6c_mifi,
744 	    if_name(ifp));
745 
746 	return (0);
747 }
748 
749 /*
750  * Delete a mif from the mif table
751  */
752 static int
753 del_m6if_locked(mifi_t *mifip)
754 {
755 	struct mif6 *mifp = mif6table + *mifip;
756 	mifi_t mifi;
757 	struct ifnet *ifp;
758 
759 	MIF6_LOCK_ASSERT();
760 
761 	if (*mifip >= nummifs)
762 		return (EINVAL);
763 	if (mifp->m6_ifp == NULL)
764 		return (EINVAL);
765 
766 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
767 		/* XXX: TODO: Maintain an ALLMULTI refcount in struct ifnet. */
768 		ifp = mifp->m6_ifp;
769 		if_allmulti(ifp, 0);
770 	} else {
771 		if (reg_mif_num != (mifi_t)-1 &&
772 		    multicast_register_if6 != NULL) {
773 			if_detach(multicast_register_if6);
774 			if_free(multicast_register_if6);
775 			reg_mif_num = (mifi_t)-1;
776 			multicast_register_if6 = NULL;
777 		}
778 	}
779 
780 	bzero((caddr_t)mifp, sizeof(*mifp));
781 
782 	/* Adjust nummifs down */
783 	for (mifi = nummifs; mifi > 0; mifi--)
784 		if (mif6table[mifi - 1].m6_ifp)
785 			break;
786 	nummifs = mifi;
787 	MRT6_DLOG(DEBUG_ANY, "mif %d, nummifs %d", *mifip, nummifs);
788 
789 	return (0);
790 }
791 
792 static int
793 del_m6if(mifi_t *mifip)
794 {
795 	int cc;
796 
797 	MIF6_LOCK();
798 	cc = del_m6if_locked(mifip);
799 	MIF6_UNLOCK();
800 
801 	return (cc);
802 }
803 
804 /*
805  * Add an mfc entry
806  */
807 static int
808 add_m6fc(struct mf6cctl *mfccp)
809 {
810 	struct mf6c *rt;
811 	u_long hash;
812 	struct rtdetq *rte;
813 	u_short nstl;
814 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
815 
816 	MFC6_LOCK();
817 
818 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
819 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
820 
821 	/* If an entry already exists, just update the fields */
822 	if (rt) {
823 		MRT6_DLOG(DEBUG_MFC, "no upcall o %s g %s p %x",
824 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
825 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
826 		    mfccp->mf6cc_parent);
827 
828 		rt->mf6c_parent = mfccp->mf6cc_parent;
829 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
830 
831 		MFC6_UNLOCK();
832 		return (0);
833 	}
834 
835 	/*
836 	 * Find the entry for which the upcall was made and update
837 	 */
838 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
839 			mfccp->mf6cc_mcastgrp.sin6_addr);
840 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
841 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
842 				       &mfccp->mf6cc_origin.sin6_addr) &&
843 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
844 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
845 		    (rt->mf6c_stall != NULL)) {
846 			if (nstl++)
847 				log(LOG_ERR,
848 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
849 				    "multiple kernel entries",
850 				    ip6_sprintf(ip6bufo,
851 					    &mfccp->mf6cc_origin.sin6_addr),
852 				    ip6_sprintf(ip6bufg,
853 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
854 				    mfccp->mf6cc_parent, rt->mf6c_stall);
855 
856 			MRT6_DLOG(DEBUG_MFC, "o %s g %s p %x dbg %p",
857 			    ip6_sprintf(ip6bufo,
858 			    &mfccp->mf6cc_origin.sin6_addr),
859 			    ip6_sprintf(ip6bufg,
860 				&mfccp->mf6cc_mcastgrp.sin6_addr),
861 			    mfccp->mf6cc_parent, rt->mf6c_stall);
862 
863 			rt->mf6c_origin     = mfccp->mf6cc_origin;
864 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
865 			rt->mf6c_parent     = mfccp->mf6cc_parent;
866 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
867 			/* initialize pkt counters per src-grp */
868 			rt->mf6c_pkt_cnt    = 0;
869 			rt->mf6c_byte_cnt   = 0;
870 			rt->mf6c_wrong_if   = 0;
871 
872 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
873 			n6expire[hash]--;
874 
875 			/* free packets Qed at the end of this entry */
876 			for (rte = rt->mf6c_stall; rte != NULL; ) {
877 				struct rtdetq *n = rte->next;
878 				ip6_mdq(rte->m, rte->ifp, rt);
879 				m_freem(rte->m);
880 #ifdef UPCALL_TIMING
881 				collate(&(rte->t));
882 #endif /* UPCALL_TIMING */
883 				free(rte, M_MRTABLE6);
884 				rte = n;
885 			}
886 			rt->mf6c_stall = NULL;
887 		}
888 	}
889 
890 	/*
891 	 * It is possible that an entry is being inserted without an upcall
892 	 */
893 	if (nstl == 0) {
894 		MRT6_DLOG(DEBUG_MFC, "no upcall h %lu o %s g %s p %x", hash,
895 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
896 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
897 		    mfccp->mf6cc_parent);
898 
899 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
900 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
901 					       &mfccp->mf6cc_origin.sin6_addr)&&
902 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
903 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
904 				rt->mf6c_origin     = mfccp->mf6cc_origin;
905 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
906 				rt->mf6c_parent     = mfccp->mf6cc_parent;
907 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
908 				/* initialize pkt counters per src-grp */
909 				rt->mf6c_pkt_cnt    = 0;
910 				rt->mf6c_byte_cnt   = 0;
911 				rt->mf6c_wrong_if   = 0;
912 
913 				if (rt->mf6c_expire)
914 					n6expire[hash]--;
915 				rt->mf6c_expire	   = 0;
916 			}
917 		}
918 		if (rt == NULL) {
919 			/* no upcall, so make a new entry */
920 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
921 						  M_NOWAIT);
922 			if (rt == NULL) {
923 				MFC6_UNLOCK();
924 				return (ENOBUFS);
925 			}
926 
927 			/* insert new entry at head of hash chain */
928 			rt->mf6c_origin     = mfccp->mf6cc_origin;
929 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
930 			rt->mf6c_parent     = mfccp->mf6cc_parent;
931 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
932 			/* initialize pkt counters per src-grp */
933 			rt->mf6c_pkt_cnt    = 0;
934 			rt->mf6c_byte_cnt   = 0;
935 			rt->mf6c_wrong_if   = 0;
936 			rt->mf6c_expire     = 0;
937 			rt->mf6c_stall = NULL;
938 
939 			/* link into table */
940 			rt->mf6c_next  = mf6ctable[hash];
941 			mf6ctable[hash] = rt;
942 		}
943 	}
944 
945 	MFC6_UNLOCK();
946 	return (0);
947 }
948 
949 #ifdef UPCALL_TIMING
950 /*
951  * collect delay statistics on the upcalls
952  */
953 static void
954 collate(struct timeval *t)
955 {
956 	u_long d;
957 	struct timeval tp;
958 	u_long delta;
959 
960 	GET_TIME(tp);
961 
962 	if (TV_LT(*t, tp))
963 	{
964 		TV_DELTA(tp, *t, delta);
965 
966 		d = delta >> 10;
967 		if (d > UPCALL_MAX)
968 			d = UPCALL_MAX;
969 
970 		++upcall_data[d];
971 	}
972 }
973 #endif /* UPCALL_TIMING */
974 
975 /*
976  * Delete an mfc entry
977  */
978 static int
979 del_m6fc(struct mf6cctl *mfccp)
980 {
981 #ifdef MRT6DEBUG
982 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
983 #endif
984 	struct sockaddr_in6	origin;
985 	struct sockaddr_in6	mcastgrp;
986 	struct mf6c		*rt;
987 	struct mf6c		**nptr;
988 	u_long		hash;
989 
990 	origin = mfccp->mf6cc_origin;
991 	mcastgrp = mfccp->mf6cc_mcastgrp;
992 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
993 
994 	MRT6_DLOG(DEBUG_MFC, "orig %s mcastgrp %s",
995 	    ip6_sprintf(ip6bufo, &origin.sin6_addr),
996 	    ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr));
997 
998 	MFC6_LOCK();
999 
1000 	nptr = &mf6ctable[hash];
1001 	while ((rt = *nptr) != NULL) {
1002 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
1003 				       &rt->mf6c_origin.sin6_addr) &&
1004 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
1005 				       &rt->mf6c_mcastgrp.sin6_addr) &&
1006 		    rt->mf6c_stall == NULL)
1007 			break;
1008 
1009 		nptr = &rt->mf6c_next;
1010 	}
1011 	if (rt == NULL) {
1012 		MFC6_UNLOCK();
1013 		return (EADDRNOTAVAIL);
1014 	}
1015 
1016 	*nptr = rt->mf6c_next;
1017 	free(rt, M_MRTABLE6);
1018 
1019 	MFC6_UNLOCK();
1020 
1021 	return (0);
1022 }
1023 
1024 static int
1025 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
1026 {
1027 
1028 	if (s) {
1029 		if (sbappendaddr(&s->so_rcv,
1030 				 (struct sockaddr *)src,
1031 				 mm, (struct mbuf *)0) != 0) {
1032 			sorwakeup(s);
1033 			return (0);
1034 		} else
1035 			soroverflow(s);
1036 	}
1037 	m_freem(mm);
1038 	return (-1);
1039 }
1040 
1041 /*
1042  * IPv6 multicast forwarding function. This function assumes that the packet
1043  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
1044  * pointed to by "ifp", and the packet is to be relayed to other networks
1045  * that have members of the packet's destination IPv6 multicast group.
1046  *
1047  * The packet is returned unscathed to the caller, unless it is
1048  * erroneous, in which case a non-zero return value tells the caller to
1049  * discard it.
1050  *
1051  * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff
1052  * this function is called in the originating context (i.e., not when
1053  * forwarding a packet from other node).  ip6_output(), which is currently the
1054  * only function that calls this function is called in the originating context,
1055  * explicitly ensures this condition.  It is caller's responsibility to ensure
1056  * that if this function is called from somewhere else in the originating
1057  * context in the future.
1058  */
1059 int
1060 X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
1061 {
1062 	struct rtdetq *rte;
1063 	struct mbuf *mb0;
1064 	struct mf6c *rt;
1065 	struct mif6 *mifp;
1066 	struct mbuf *mm;
1067 	u_long hash;
1068 	mifi_t mifi;
1069 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1070 #ifdef UPCALL_TIMING
1071 	struct timeval tp;
1072 
1073 	GET_TIME(tp);
1074 #endif /* UPCALL_TIMING */
1075 
1076 	M_ASSERTMAPPED(m);
1077 	MRT6_DLOG(DEBUG_FORWARD, "src %s, dst %s, ifindex %d",
1078 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1079 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst), ifp->if_index);
1080 
1081 	/*
1082 	 * Don't forward a packet with Hop limit of zero or one,
1083 	 * or a packet destined to a local-only group.
1084 	 */
1085 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
1086 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1087 		return (0);
1088 	ip6->ip6_hlim--;
1089 
1090 	/*
1091 	 * Source address check: do not forward packets with unspecified
1092 	 * source. It was discussed in July 2000, on ipngwg mailing list.
1093 	 * This is rather more serious than unicast cases, because some
1094 	 * MLD packets can be sent with the unspecified source address
1095 	 * (although such packets must normally set 1 to the hop limit field).
1096 	 */
1097 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1098 		IP6STAT_INC(ip6s_cantforward);
1099 		if (V_ip6_log_cannot_forward && ip6_log_ratelimit()) {
1100 			log(LOG_DEBUG,
1101 			    "cannot forward "
1102 			    "from %s to %s nxt %d received on %s\n",
1103 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1104 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1105 			    ip6->ip6_nxt,
1106 			    if_name(m->m_pkthdr.rcvif));
1107 		}
1108 		return (0);
1109 	}
1110 
1111 	MFC6_LOCK();
1112 
1113 	/*
1114 	 * Determine forwarding mifs from the forwarding cache table
1115 	 */
1116 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
1117 	MRT6STAT_INC(mrt6s_mfc_lookups);
1118 
1119 	/* Entry exists, so forward if necessary */
1120 	if (rt) {
1121 		MFC6_UNLOCK();
1122 		return (ip6_mdq(m, ifp, rt));
1123 	}
1124 
1125 	/*
1126 	 * If we don't have a route for packet's origin,
1127 	 * Make a copy of the packet & send message to routing daemon.
1128 	 */
1129 	MRT6STAT_INC(mrt6s_no_route);
1130 	MRT6_DLOG(DEBUG_FORWARD | DEBUG_MFC, "no rte s %s g %s",
1131 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1132 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1133 
1134 	/*
1135 	 * Allocate mbufs early so that we don't do extra work if we
1136 	 * are just going to fail anyway.
1137 	 */
1138 	rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE6, M_NOWAIT);
1139 	if (rte == NULL) {
1140 		MFC6_UNLOCK();
1141 		return (ENOBUFS);
1142 	}
1143 	mb0 = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1144 	/*
1145 	 * Pullup packet header if needed before storing it,
1146 	 * as other references may modify it in the meantime.
1147 	 */
1148 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1149 		mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1150 	if (mb0 == NULL) {
1151 		free(rte, M_MRTABLE6);
1152 		MFC6_UNLOCK();
1153 		return (ENOBUFS);
1154 	}
1155 
1156 	/* is there an upcall waiting for this packet? */
1157 	hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1158 	for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1159 		if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1160 		    &rt->mf6c_origin.sin6_addr) &&
1161 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1162 		    &rt->mf6c_mcastgrp.sin6_addr) && (rt->mf6c_stall != NULL))
1163 			break;
1164 	}
1165 
1166 	if (rt == NULL) {
1167 		struct mrt6msg *im;
1168 #ifdef MRT6_OINIT
1169 		struct omrt6msg *oim;
1170 #endif
1171 		/* no upcall, so make a new entry */
1172 		rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6, M_NOWAIT);
1173 		if (rt == NULL) {
1174 			free(rte, M_MRTABLE6);
1175 			m_freem(mb0);
1176 			MFC6_UNLOCK();
1177 			return (ENOBUFS);
1178 		}
1179 		/*
1180 		 * Make a copy of the header to send to the user
1181 		 * level process
1182 		 */
1183 		mm = m_copym(mb0, 0, sizeof(struct ip6_hdr), M_NOWAIT);
1184 		if (mm == NULL) {
1185 			free(rte, M_MRTABLE6);
1186 			m_freem(mb0);
1187 			free(rt, M_MRTABLE6);
1188 			MFC6_UNLOCK();
1189 			return (ENOBUFS);
1190 		}
1191 
1192 		/*
1193 		 * Send message to routing daemon
1194 		 */
1195 		sin6.sin6_addr = ip6->ip6_src;
1196 		im = NULL;
1197 #ifdef MRT6_OINIT
1198 		oim = NULL;
1199 #endif
1200 		switch (V_ip6_mrouter_ver) {
1201 #ifdef MRT6_OINIT
1202 		case MRT6_OINIT:
1203 			oim = mtod(mm, struct omrt6msg *);
1204 			oim->im6_msgtype = MRT6MSG_NOCACHE;
1205 			oim->im6_mbz = 0;
1206 			break;
1207 #endif
1208 		case MRT6_INIT:
1209 			im = mtod(mm, struct mrt6msg *);
1210 			im->im6_msgtype = MRT6MSG_NOCACHE;
1211 			im->im6_mbz = 0;
1212 			break;
1213 		default:
1214 			free(rte, M_MRTABLE6);
1215 			m_freem(mb0);
1216 			free(rt, M_MRTABLE6);
1217 			MFC6_UNLOCK();
1218 			return (EINVAL);
1219 		}
1220 
1221 		MRT6_DLOG(DEBUG_FORWARD, "getting the iif info in the kernel");
1222 		for (mifp = mif6table, mifi = 0;
1223 		    mifi < nummifs && mifp->m6_ifp != ifp; mifp++, mifi++)
1224 				;
1225 
1226 		switch (V_ip6_mrouter_ver) {
1227 #ifdef MRT6_OINIT
1228 		case MRT6_OINIT:
1229 			oim->im6_mif = mifi;
1230 			break;
1231 #endif
1232 		case MRT6_INIT:
1233 			im->im6_mif = mifi;
1234 			break;
1235 		}
1236 
1237 		if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1238 			log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1239 			    "socket queue full\n");
1240 			MRT6STAT_INC(mrt6s_upq_sockfull);
1241 			free(rte, M_MRTABLE6);
1242 			m_freem(mb0);
1243 			free(rt, M_MRTABLE6);
1244 			MFC6_UNLOCK();
1245 			return (ENOBUFS);
1246 		}
1247 
1248 		MRT6STAT_INC(mrt6s_upcalls);
1249 
1250 		/* insert new entry at head of hash chain */
1251 		bzero(rt, sizeof(*rt));
1252 		rt->mf6c_origin.sin6_family = AF_INET6;
1253 		rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1254 		rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1255 		rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1256 		rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1257 		rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1258 		rt->mf6c_expire = UPCALL_EXPIRE;
1259 		n6expire[hash]++;
1260 		rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1261 
1262 		/* link into table */
1263 		rt->mf6c_next  = mf6ctable[hash];
1264 		mf6ctable[hash] = rt;
1265 		/* Add this entry to the end of the queue */
1266 		rt->mf6c_stall = rte;
1267 	} else {
1268 		/* determine if q has overflowed */
1269 		struct rtdetq **p;
1270 		int npkts = 0;
1271 
1272 		for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1273 			if (++npkts > MAX_UPQ6) {
1274 				MRT6STAT_INC(mrt6s_upq_ovflw);
1275 				free(rte, M_MRTABLE6);
1276 				m_freem(mb0);
1277 				MFC6_UNLOCK();
1278 				return (0);
1279 			}
1280 
1281 		/* Add this entry to the end of the queue */
1282 		*p = rte;
1283 	}
1284 
1285 	rte->next = NULL;
1286 	rte->m = mb0;
1287 	rte->ifp = ifp;
1288 #ifdef UPCALL_TIMING
1289 	rte->t = tp;
1290 #endif /* UPCALL_TIMING */
1291 
1292 	MFC6_UNLOCK();
1293 
1294 	return (0);
1295 }
1296 
1297 /*
1298  * Clean up cache entries if upcalls are not serviced
1299  * Call from the Slow Timeout mechanism, every half second.
1300  */
1301 static void
1302 expire_upcalls(void *unused)
1303 {
1304 #ifdef MRT6DEBUG
1305 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1306 #endif
1307 	struct rtdetq *rte;
1308 	struct mf6c *mfc, **nptr;
1309 	u_long i;
1310 
1311 	MFC6_LOCK_ASSERT();
1312 
1313 	for (i = 0; i < MF6CTBLSIZ; i++) {
1314 		if (n6expire[i] == 0)
1315 			continue;
1316 		nptr = &mf6ctable[i];
1317 		while ((mfc = *nptr) != NULL) {
1318 			rte = mfc->mf6c_stall;
1319 			/*
1320 			 * Skip real cache entries
1321 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1322 			 * If it expires now
1323 			 */
1324 			if (rte != NULL &&
1325 			    mfc->mf6c_expire != 0 &&
1326 			    --mfc->mf6c_expire == 0) {
1327 				MRT6_DLOG(DEBUG_EXPIRE, "expiring (%s %s)",
1328 				    ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr),
1329 				    ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr));
1330 				/*
1331 				 * drop all the packets
1332 				 * free the mbuf with the pkt, if, timing info
1333 				 */
1334 				do {
1335 					struct rtdetq *n = rte->next;
1336 					m_freem(rte->m);
1337 					free(rte, M_MRTABLE6);
1338 					rte = n;
1339 				} while (rte != NULL);
1340 				MRT6STAT_INC(mrt6s_cache_cleanups);
1341 				n6expire[i]--;
1342 
1343 				*nptr = mfc->mf6c_next;
1344 				free(mfc, M_MRTABLE6);
1345 			} else {
1346 				nptr = &mfc->mf6c_next;
1347 			}
1348 		}
1349 	}
1350 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1351 	    expire_upcalls, NULL);
1352 }
1353 
1354 /*
1355  * Packet forwarding routine once entry in the cache is made
1356  */
1357 static int
1358 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1359 {
1360 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1361 	mifi_t mifi, iif;
1362 	struct mif6 *mifp;
1363 	int plen = m->m_pkthdr.len;
1364 	struct in6_addr src0, dst0; /* copies for local work */
1365 	u_int32_t iszone, idzone, oszone, odzone;
1366 	int error = 0;
1367 
1368 	M_ASSERTMAPPED(m);
1369 
1370 	/*
1371 	 * Don't forward if it didn't arrive from the parent mif
1372 	 * for its origin.
1373 	 */
1374 	mifi = rt->mf6c_parent;
1375 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1376 		/* came in the wrong interface */
1377 		MRT6_DLOG(DEBUG_FORWARD,
1378 		    "wrong if: ifid %d mifi %d mififid %x", ifp->if_index,
1379 		    mifi, mif6table[mifi].m6_ifp->if_index);
1380 		MRT6STAT_INC(mrt6s_wrong_if);
1381 		rt->mf6c_wrong_if++;
1382 		/*
1383 		 * If we are doing PIM processing, and we are forwarding
1384 		 * packets on this interface, send a message to the
1385 		 * routing daemon.
1386 		 */
1387 		/* have to make sure this is a valid mif */
1388 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1389 			if (V_pim6 && (m->m_flags & M_LOOP) == 0) {
1390 				/*
1391 				 * Check the M_LOOP flag to avoid an
1392 				 * unnecessary PIM assert.
1393 				 * XXX: M_LOOP is an ad-hoc hack...
1394 				 */
1395 				static struct sockaddr_in6 sin6 =
1396 				{ sizeof(sin6), AF_INET6 };
1397 
1398 				struct mbuf *mm;
1399 				struct mrt6msg *im;
1400 #ifdef MRT6_OINIT
1401 				struct omrt6msg *oim;
1402 #endif
1403 
1404 				mm = m_copym(m, 0, sizeof(struct ip6_hdr),
1405 				    M_NOWAIT);
1406 				if (mm &&
1407 				    (!M_WRITABLE(mm) ||
1408 				     mm->m_len < sizeof(struct ip6_hdr)))
1409 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1410 				if (mm == NULL)
1411 					return (ENOBUFS);
1412 
1413 #ifdef MRT6_OINIT
1414 				oim = NULL;
1415 #endif
1416 				im = NULL;
1417 				switch (V_ip6_mrouter_ver) {
1418 #ifdef MRT6_OINIT
1419 				case MRT6_OINIT:
1420 					oim = mtod(mm, struct omrt6msg *);
1421 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1422 					oim->im6_mbz = 0;
1423 					break;
1424 #endif
1425 				case MRT6_INIT:
1426 					im = mtod(mm, struct mrt6msg *);
1427 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1428 					im->im6_mbz = 0;
1429 					break;
1430 				default:
1431 					m_freem(mm);
1432 					return (EINVAL);
1433 				}
1434 
1435 				for (mifp = mif6table, iif = 0;
1436 				     iif < nummifs && mifp &&
1437 					     mifp->m6_ifp != ifp;
1438 				     mifp++, iif++)
1439 					;
1440 
1441 				switch (V_ip6_mrouter_ver) {
1442 #ifdef MRT6_OINIT
1443 				case MRT6_OINIT:
1444 					oim->im6_mif = iif;
1445 					sin6.sin6_addr = oim->im6_src;
1446 					break;
1447 #endif
1448 				case MRT6_INIT:
1449 					im->im6_mif = iif;
1450 					sin6.sin6_addr = im->im6_src;
1451 					break;
1452 				}
1453 
1454 				MRT6STAT_INC(mrt6s_upcalls);
1455 
1456 				if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1457 					MRT6_DLOG(DEBUG_ANY,
1458 					    "ip6_mrouter socket queue full");
1459 					MRT6STAT_INC(mrt6s_upq_sockfull);
1460 					return (ENOBUFS);
1461 				}	/* if socket Q full */
1462 			}		/* if PIM */
1463 		return (0);
1464 	}			/* if wrong iif */
1465 
1466 	/* If I sourced this packet, it counts as output, else it was input. */
1467 	if (m->m_pkthdr.rcvif == NULL) {
1468 		/* XXX: is rcvif really NULL when output?? */
1469 		mif6table[mifi].m6_pkt_out++;
1470 		mif6table[mifi].m6_bytes_out += plen;
1471 	} else {
1472 		mif6table[mifi].m6_pkt_in++;
1473 		mif6table[mifi].m6_bytes_in += plen;
1474 	}
1475 	rt->mf6c_pkt_cnt++;
1476 	rt->mf6c_byte_cnt += plen;
1477 
1478 	/*
1479 	 * For each mif, forward a copy of the packet if there are group
1480 	 * members downstream on the interface.
1481 	 */
1482 	src0 = ip6->ip6_src;
1483 	dst0 = ip6->ip6_dst;
1484 	if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 ||
1485 	    (error = in6_setscope(&dst0, ifp, &idzone)) != 0) {
1486 		IP6STAT_INC(ip6s_badscope);
1487 		return (error);
1488 	}
1489 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) {
1490 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1491 			/*
1492 			 * check if the outgoing packet is going to break
1493 			 * a scope boundary.
1494 			 * XXX For packets through PIM register tunnel
1495 			 * interface, we believe a routing daemon.
1496 			 */
1497 			if (!(mif6table[rt->mf6c_parent].m6_flags &
1498 			      MIFF_REGISTER) &&
1499 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
1500 				if (in6_setscope(&src0, mif6table[mifi].m6_ifp,
1501 				    &oszone) ||
1502 				    in6_setscope(&dst0, mif6table[mifi].m6_ifp,
1503 				    &odzone) ||
1504 				    iszone != oszone ||
1505 				    idzone != odzone) {
1506 					IP6STAT_INC(ip6s_badscope);
1507 					continue;
1508 				}
1509 			}
1510 
1511 			mifp->m6_pkt_out++;
1512 			mifp->m6_bytes_out += plen;
1513 			if (mifp->m6_flags & MIFF_REGISTER)
1514 				register_send(ip6, mifp, m);
1515 			else
1516 				phyint_send(ip6, mifp, m);
1517 		}
1518 	}
1519 	return (0);
1520 }
1521 
1522 static void
1523 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1524 {
1525 #ifdef MRT6DEBUG
1526 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1527 #endif
1528 	struct mbuf *mb_copy;
1529 	struct ifnet *ifp = mifp->m6_ifp;
1530 	int error __unused = 0;
1531 	u_long linkmtu;
1532 
1533 	M_ASSERTMAPPED(m);
1534 
1535 	/*
1536 	 * Make a new reference to the packet; make sure that
1537 	 * the IPv6 header is actually copied, not just referenced,
1538 	 * so that ip6_output() only scribbles on the copy.
1539 	 */
1540 	mb_copy = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1541 	if (mb_copy &&
1542 	    (!M_WRITABLE(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1543 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1544 	if (mb_copy == NULL) {
1545 		return;
1546 	}
1547 	/* set MCAST flag to the outgoing packet */
1548 	mb_copy->m_flags |= M_MCAST;
1549 
1550 	/*
1551 	 * If we sourced the packet, call ip6_output since we may devide
1552 	 * the packet into fragments when the packet is too big for the
1553 	 * outgoing interface.
1554 	 * Otherwise, we can simply send the packet to the interface
1555 	 * sending queue.
1556 	 */
1557 	if (m->m_pkthdr.rcvif == NULL) {
1558 		struct ip6_moptions im6o;
1559 		struct epoch_tracker et;
1560 
1561 		im6o.im6o_multicast_ifp = ifp;
1562 		/* XXX: ip6_output will override ip6->ip6_hlim */
1563 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1564 		im6o.im6o_multicast_loop = 1;
1565 		NET_EPOCH_ENTER(et);
1566 		error = ip6_output(mb_copy, NULL, NULL, IPV6_FORWARDING, &im6o,
1567 		    NULL, NULL);
1568 		NET_EPOCH_EXIT(et);
1569 
1570 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1571 		    (uint16_t)(mifp - mif6table), error);
1572 		return;
1573 	}
1574 
1575 	/*
1576 	 * If configured to loop back multicasts by default,
1577 	 * loop back a copy now.
1578 	 */
1579 	if (in6_mcast_loop)
1580 		ip6_mloopback(ifp, m);
1581 
1582 	/*
1583 	 * Put the packet into the sending queue of the outgoing interface
1584 	 * if it would fit in the MTU of the interface.
1585 	 */
1586 	linkmtu = IN6_LINKMTU(ifp);
1587 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1588 		struct sockaddr_in6 dst6;
1589 
1590 		bzero(&dst6, sizeof(dst6));
1591 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1592 		dst6.sin6_family = AF_INET6;
1593 		dst6.sin6_addr = ip6->ip6_dst;
1594 
1595 		IP_PROBE(send, NULL, NULL, ip6, ifp, NULL, ip6);
1596 		/*
1597 		 * We just call if_output instead of nd6_output here, since
1598 		 * we need no ND for a multicast forwarded packet...right?
1599 		 */
1600 		m_clrprotoflags(m);	/* Avoid confusing lower layers. */
1601 		error = (*ifp->if_output)(ifp, mb_copy,
1602 		    (struct sockaddr *)&dst6, NULL);
1603 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1604 		    (uint16_t)(mifp - mif6table), error);
1605 	} else {
1606 		/*
1607 		 * pMTU discovery is intentionally disabled by default, since
1608 		 * various router may notify pMTU in multicast, which can be
1609 		 * a DDoS to a router
1610 		 */
1611 		if (V_ip6_mcast_pmtu)
1612 			icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1613 		else {
1614 			MRT6_DLOG(DEBUG_XMIT, " packet too big on %s o %s "
1615 			    "g %s size %d (discarded)", if_name(ifp),
1616 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1617 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1618 			    mb_copy->m_pkthdr.len);
1619 			m_freem(mb_copy); /* simply discard the packet */
1620 		}
1621 	}
1622 }
1623 
1624 static int
1625 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1626 {
1627 #ifdef MRT6DEBUG
1628 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1629 #endif
1630 	struct mbuf *mm;
1631 	int i, len = m->m_pkthdr.len;
1632 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1633 	struct mrt6msg *im6;
1634 
1635 	MRT6_DLOG(DEBUG_ANY, "src %s dst %s",
1636 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1637 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1638 	PIM6STAT_INC(pim6s_snd_registers);
1639 
1640 	/* Make a copy of the packet to send to the user level process. */
1641 	mm = m_gethdr(M_NOWAIT, MT_DATA);
1642 	if (mm == NULL)
1643 		return (ENOBUFS);
1644 	mm->m_data += max_linkhdr;
1645 	mm->m_len = sizeof(struct ip6_hdr);
1646 
1647 	if ((mm->m_next = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
1648 		m_freem(mm);
1649 		return (ENOBUFS);
1650 	}
1651 	i = MHLEN - M_LEADINGSPACE(mm);
1652 	if (i > len)
1653 		i = len;
1654 	mm = m_pullup(mm, i);
1655 	if (mm == NULL)
1656 		return (ENOBUFS);
1657 /* TODO: check it! */
1658 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1659 
1660 	/*
1661 	 * Send message to routing daemon
1662 	 */
1663 	sin6.sin6_addr = ip6->ip6_src;
1664 
1665 	im6 = mtod(mm, struct mrt6msg *);
1666 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1667 	im6->im6_mbz          = 0;
1668 
1669 	im6->im6_mif = mif - mif6table;
1670 
1671 	/* iif info is not given for reg. encap.n */
1672 	MRT6STAT_INC(mrt6s_upcalls);
1673 
1674 	if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1675 		MRT6_DLOG(DEBUG_ANY, "ip6_mrouter socket queue full");
1676 		MRT6STAT_INC(mrt6s_upq_sockfull);
1677 		return (ENOBUFS);
1678 	}
1679 	return (0);
1680 }
1681 
1682 /*
1683  * pim6_encapcheck() is called by the encap6_input() path at runtime to
1684  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
1685  * into the kernel.
1686  */
1687 static int
1688 pim6_encapcheck(const struct mbuf *m __unused, int off __unused,
1689     int proto __unused, void *arg __unused)
1690 {
1691 
1692     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
1693     return (8);		/* claim the datagram. */
1694 }
1695 
1696 /*
1697  * PIM sparse mode hook
1698  * Receives the pim control messages, and passes them up to the listening
1699  * socket, using rip6_input.
1700  * The only message processed is the REGISTER pim message; the pim header
1701  * is stripped off, and the inner packet is passed to register_mforward.
1702  */
1703 static int
1704 pim6_input(struct mbuf *m, int off, int proto, void *arg __unused)
1705 {
1706 	struct pim *pim; /* pointer to a pim struct */
1707 	struct ip6_hdr *ip6;
1708 	int pimlen;
1709 	int minlen;
1710 
1711 	PIM6STAT_INC(pim6s_rcv_total);
1712 
1713 	/*
1714 	 * Validate lengths
1715 	 */
1716 	pimlen = m->m_pkthdr.len - off;
1717 	if (pimlen < PIM_MINLEN) {
1718 		PIM6STAT_INC(pim6s_rcv_tooshort);
1719 		MRT6_DLOG(DEBUG_PIM, "PIM packet too short");
1720 		m_freem(m);
1721 		return (IPPROTO_DONE);
1722 	}
1723 
1724 	/*
1725 	 * if the packet is at least as big as a REGISTER, go ahead
1726 	 * and grab the PIM REGISTER header size, to avoid another
1727 	 * possible m_pullup() later.
1728 	 *
1729 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1730 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1731 	 */
1732 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1733 
1734 	/*
1735 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1736 	 * possibly the PIM REGISTER header
1737 	 */
1738 	if (m->m_len < off + minlen) {
1739 		m = m_pullup(m, off + minlen);
1740 		if (m == NULL) {
1741 			IP6STAT_INC(ip6s_exthdrtoolong);
1742 			return (IPPROTO_DONE);
1743 		}
1744 	}
1745 	ip6 = mtod(m, struct ip6_hdr *);
1746 	pim = (struct pim *)((caddr_t)ip6 + off);
1747 
1748 #define PIM6_CHECKSUM
1749 #ifdef PIM6_CHECKSUM
1750 	{
1751 		int cksumlen;
1752 
1753 		/*
1754 		 * Validate checksum.
1755 		 * If PIM REGISTER, exclude the data packet
1756 		 */
1757 		if (pim->pim_type == PIM_REGISTER)
1758 			cksumlen = PIM_MINLEN;
1759 		else
1760 			cksumlen = pimlen;
1761 
1762 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1763 			PIM6STAT_INC(pim6s_rcv_badsum);
1764 			MRT6_DLOG(DEBUG_PIM, "invalid checksum");
1765 			m_freem(m);
1766 			return (IPPROTO_DONE);
1767 		}
1768 	}
1769 #endif /* PIM_CHECKSUM */
1770 
1771 	/* PIM version check */
1772 	if (pim->pim_ver != PIM_VERSION) {
1773 		PIM6STAT_INC(pim6s_rcv_badversion);
1774 		MRT6_DLOG(DEBUG_ANY | DEBUG_ERR,
1775 		    "incorrect version %d, expecting %d",
1776 		    pim->pim_ver, PIM_VERSION);
1777 		m_freem(m);
1778 		return (IPPROTO_DONE);
1779 	}
1780 
1781 	if (pim->pim_type == PIM_REGISTER) {
1782 		/*
1783 		 * since this is a REGISTER, we'll make a copy of the register
1784 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1785 		 * routing daemon.
1786 		 */
1787 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1788 
1789 		struct mbuf *mcp;
1790 		struct ip6_hdr *eip6;
1791 		u_int32_t *reghdr;
1792 #ifdef MRT6DEBUG
1793 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1794 #endif
1795 
1796 		PIM6STAT_INC(pim6s_rcv_registers);
1797 
1798 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1799 			MRT6_DLOG(DEBUG_PIM, "register mif not set: %d",
1800 			    reg_mif_num);
1801 			m_freem(m);
1802 			return (IPPROTO_DONE);
1803 		}
1804 
1805 		reghdr = (u_int32_t *)(pim + 1);
1806 
1807 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1808 			goto pim6_input_to_daemon;
1809 
1810 		/*
1811 		 * Validate length
1812 		 */
1813 		if (pimlen < PIM6_REG_MINLEN) {
1814 			PIM6STAT_INC(pim6s_rcv_tooshort);
1815 			PIM6STAT_INC(pim6s_rcv_badregisters);
1816 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "register packet "
1817 			    "size too small %d from %s",
1818 			    pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src));
1819 			m_freem(m);
1820 			return (IPPROTO_DONE);
1821 		}
1822 
1823 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1824 		MRT6_DLOG(DEBUG_PIM, "eip6: %s -> %s, eip6 plen %d",
1825 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1826 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1827 		    ntohs(eip6->ip6_plen));
1828 
1829 		/* verify the version number of the inner packet */
1830 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1831 			PIM6STAT_INC(pim6s_rcv_badregisters);
1832 			MRT6_DLOG(DEBUG_ANY, "invalid IP version (%d) "
1833 			    "of the inner packet",
1834 			    (eip6->ip6_vfc & IPV6_VERSION));
1835 			m_freem(m);
1836 			return (IPPROTO_DONE);
1837 		}
1838 
1839 		/* verify the inner packet is destined to a mcast group */
1840 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1841 			PIM6STAT_INC(pim6s_rcv_badregisters);
1842 			MRT6_DLOG(DEBUG_PIM, "inner packet of register "
1843 			    "is not multicast %s",
1844 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst));
1845 			m_freem(m);
1846 			return (IPPROTO_DONE);
1847 		}
1848 
1849 		/*
1850 		 * make a copy of the whole header to pass to the daemon later.
1851 		 */
1852 		mcp = m_copym(m, 0, off + PIM6_REG_MINLEN, M_NOWAIT);
1853 		if (mcp == NULL) {
1854 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "pim register: "
1855 			    "could not copy register head");
1856 			m_freem(m);
1857 			return (IPPROTO_DONE);
1858 		}
1859 
1860 		/*
1861 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1862 		 */
1863 		m_adj(m, off + PIM_MINLEN);
1864 		MRT6_DLOG(DEBUG_PIM, "forwarding decapsulated register: "
1865 		    "src %s, dst %s, mif %d",
1866 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1867 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst), reg_mif_num);
1868 
1869 		if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1870 				dst.sin6_family, 0);
1871 
1872 		/* prepare the register head to send to the mrouting daemon */
1873 		m = mcp;
1874 	}
1875 
1876 	/*
1877 	 * Pass the PIM message up to the daemon; if it is a register message
1878 	 * pass the 'head' only up to the daemon. This includes the
1879 	 * encapsulator ip6 header, pim header, register header and the
1880 	 * encapsulated ip6 header.
1881 	 */
1882   pim6_input_to_daemon:
1883 	return (rip6_input(&m, &off, proto));
1884 }
1885 
1886 static int
1887 ip6_mroute_modevent(module_t mod, int type, void *unused)
1888 {
1889 
1890 	switch (type) {
1891 	case MOD_LOAD:
1892 		MROUTER6_LOCK_INIT();
1893 		MFC6_LOCK_INIT();
1894 		MIF6_LOCK_INIT();
1895 
1896 		pim6_encap_cookie = ip6_encap_attach(&ipv6_encap_cfg,
1897 		    NULL, M_WAITOK);
1898 		if (pim6_encap_cookie == NULL) {
1899 			printf("ip6_mroute: unable to attach pim6 encap\n");
1900 			MIF6_LOCK_DESTROY();
1901 			MFC6_LOCK_DESTROY();
1902 			MROUTER6_LOCK_DESTROY();
1903 			return (EINVAL);
1904 		}
1905 
1906 		ip6_mforward = X_ip6_mforward;
1907 		ip6_mrouter_done = X_ip6_mrouter_done;
1908 		ip6_mrouter_get = X_ip6_mrouter_get;
1909 		ip6_mrouter_set = X_ip6_mrouter_set;
1910 		mrt6_ioctl = X_mrt6_ioctl;
1911 		break;
1912 
1913 	case MOD_UNLOAD:
1914 		if (V_ip6_mrouter != NULL)
1915 			return EINVAL;
1916 
1917 		if (pim6_encap_cookie) {
1918 			ip6_encap_detach(pim6_encap_cookie);
1919 			pim6_encap_cookie = NULL;
1920 		}
1921 		X_ip6_mrouter_done();
1922 		ip6_mforward = NULL;
1923 		ip6_mrouter_done = NULL;
1924 		ip6_mrouter_get = NULL;
1925 		ip6_mrouter_set = NULL;
1926 		mrt6_ioctl = NULL;
1927 
1928 		MIF6_LOCK_DESTROY();
1929 		MFC6_LOCK_DESTROY();
1930 		MROUTER6_LOCK_DESTROY();
1931 		break;
1932 
1933 	default:
1934 		return (EOPNOTSUPP);
1935 	}
1936 
1937 	return (0);
1938 }
1939 
1940 static moduledata_t ip6_mroutemod = {
1941 	"ip6_mroute",
1942 	ip6_mroute_modevent,
1943 	0
1944 };
1945 
1946 DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_ANY);
1947