xref: /freebsd/sys/netinet6/ip6_mroute.c (revision 57718be8fa0bd5edc11ab9a72e68cc71982939a6)
1 /*-
2  * Copyright (C) 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1989 Stephen Deering
34  * Copyright (c) 1992, 1993
35  *      The Regents of the University of California.  All rights reserved.
36  *
37  * This code is derived from software contributed to Berkeley by
38  * Stephen Deering of Stanford University.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 4. Neither the name of the University nor the names of its contributors
49  *    may be used to endorse or promote products derived from this software
50  *    without specific prior written permission.
51  *
52  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62  * SUCH DAMAGE.
63  *
64  *	@(#)ip_mroute.c	8.2 (Berkeley) 11/15/93
65  *	BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp
66  */
67 
68 /*
69  * IP multicast forwarding procedures
70  *
71  * Written by David Waitzman, BBN Labs, August 1988.
72  * Modified by Steve Deering, Stanford, February 1989.
73  * Modified by Mark J. Steiglitz, Stanford, May, 1991
74  * Modified by Van Jacobson, LBL, January 1993
75  * Modified by Ajit Thyagarajan, PARC, August 1993
76  * Modified by Bill Fenner, PARC, April 1994
77  *
78  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
79  */
80 
81 #include <sys/cdefs.h>
82 __FBSDID("$FreeBSD$");
83 
84 #include "opt_inet6.h"
85 
86 #include <sys/param.h>
87 #include <sys/callout.h>
88 #include <sys/errno.h>
89 #include <sys/kernel.h>
90 #include <sys/lock.h>
91 #include <sys/malloc.h>
92 #include <sys/mbuf.h>
93 #include <sys/module.h>
94 #include <sys/domain.h>
95 #include <sys/protosw.h>
96 #include <sys/sdt.h>
97 #include <sys/signalvar.h>
98 #include <sys/socket.h>
99 #include <sys/socketvar.h>
100 #include <sys/sockio.h>
101 #include <sys/sx.h>
102 #include <sys/sysctl.h>
103 #include <sys/syslog.h>
104 #include <sys/systm.h>
105 #include <sys/time.h>
106 
107 #include <net/if.h>
108 #include <net/if_var.h>
109 #include <net/if_types.h>
110 #include <net/raw_cb.h>
111 #include <net/vnet.h>
112 
113 #include <netinet/in.h>
114 #include <netinet/in_var.h>
115 #include <netinet/icmp6.h>
116 #include <netinet/ip_encap.h>
117 
118 #include <netinet/ip6.h>
119 #include <netinet/in_kdtrace.h>
120 #include <netinet6/ip6_var.h>
121 #include <netinet6/scope6_var.h>
122 #include <netinet6/nd6.h>
123 #include <netinet6/ip6_mroute.h>
124 #include <netinet6/pim6.h>
125 #include <netinet6/pim6_var.h>
126 
127 static MALLOC_DEFINE(M_MRTABLE6, "mf6c", "multicast forwarding cache entry");
128 
129 /* XXX: this is a very common idiom; move to <sys/mbuf.h> ? */
130 #define M_HASCL(m) ((m)->m_flags & M_EXT)
131 
132 static int	ip6_mdq(struct mbuf *, struct ifnet *, struct mf6c *);
133 static void	phyint_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
134 static int	register_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
135 static int	set_pim6(int *);
136 static int	socket_send(struct socket *, struct mbuf *,
137 		    struct sockaddr_in6 *);
138 
139 extern int in6_mcast_loop;
140 extern struct domain inet6domain;
141 
142 static const struct encaptab *pim6_encap_cookie;
143 static const struct protosw in6_pim_protosw = {
144 	.pr_type =		SOCK_RAW,
145 	.pr_domain =		&inet6domain,
146 	.pr_protocol =		IPPROTO_PIM,
147 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
148 	.pr_input =		pim6_input,
149 	.pr_output =		rip6_output,
150 	.pr_ctloutput =		rip6_ctloutput,
151 	.pr_usrreqs =		&rip6_usrreqs
152 };
153 static int pim6_encapcheck(const struct mbuf *, int, int, void *);
154 
155 static VNET_DEFINE(int, ip6_mrouter_ver) = 0;
156 #define	V_ip6_mrouter_ver	VNET(ip6_mrouter_ver)
157 
158 SYSCTL_DECL(_net_inet6);
159 SYSCTL_DECL(_net_inet6_ip6);
160 static SYSCTL_NODE(_net_inet6, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
161 
162 static struct mrt6stat mrt6stat;
163 SYSCTL_STRUCT(_net_inet6_ip6, OID_AUTO, mrt6stat, CTLFLAG_RW,
164     &mrt6stat, mrt6stat,
165     "Multicast Routing Statistics (struct mrt6stat, netinet6/ip6_mroute.h)");
166 
167 #define	MRT6STAT_INC(name)	mrt6stat.name += 1
168 #define NO_RTE_FOUND	0x1
169 #define RTE_FOUND	0x2
170 
171 static struct mtx mrouter6_mtx;
172 #define	MROUTER6_LOCK()		mtx_lock(&mrouter6_mtx)
173 #define	MROUTER6_UNLOCK()	mtx_unlock(&mrouter6_mtx)
174 #define	MROUTER6_LOCK_ASSERT()	do {					\
175 	mtx_assert(&mrouter6_mtx, MA_OWNED);				\
176 	NET_ASSERT_GIANT();						\
177 } while (0)
178 #define	MROUTER6_LOCK_INIT()	\
179 	mtx_init(&mrouter6_mtx, "IPv6 multicast forwarding", NULL, MTX_DEF)
180 #define	MROUTER6_LOCK_DESTROY()	mtx_destroy(&mrouter6_mtx)
181 
182 static struct mf6c *mf6ctable[MF6CTBLSIZ];
183 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mf6ctable, CTLFLAG_RD,
184     &mf6ctable, sizeof(mf6ctable), "S,*mf6ctable[MF6CTBLSIZ]",
185     "IPv6 Multicast Forwarding Table (struct *mf6ctable[MF6CTBLSIZ], "
186     "netinet6/ip6_mroute.h)");
187 
188 static struct mtx mfc6_mtx;
189 #define	MFC6_LOCK()		mtx_lock(&mfc6_mtx)
190 #define	MFC6_UNLOCK()		mtx_unlock(&mfc6_mtx)
191 #define	MFC6_LOCK_ASSERT()	do {					\
192 	mtx_assert(&mfc6_mtx, MA_OWNED);				\
193 	NET_ASSERT_GIANT();						\
194 } while (0)
195 #define	MFC6_LOCK_INIT()		\
196 	mtx_init(&mfc6_mtx, "IPv6 multicast forwarding cache", NULL, MTX_DEF)
197 #define	MFC6_LOCK_DESTROY()	mtx_destroy(&mfc6_mtx)
198 
199 static u_char n6expire[MF6CTBLSIZ];
200 
201 static struct mif6 mif6table[MAXMIFS];
202 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mif6table, CTLFLAG_RD,
203     &mif6table, sizeof(mif6table), "S,mif6[MAXMIFS]",
204     "IPv6 Multicast Interfaces (struct mif6[MAXMIFS], netinet6/ip6_mroute.h)");
205 
206 static struct mtx mif6_mtx;
207 #define	MIF6_LOCK()		mtx_lock(&mif6_mtx)
208 #define	MIF6_UNLOCK()		mtx_unlock(&mif6_mtx)
209 #define	MIF6_LOCK_ASSERT()	mtx_assert(&mif6_mtx, MA_OWNED)
210 #define	MIF6_LOCK_INIT()	\
211 	mtx_init(&mif6_mtx, "IPv6 multicast interfaces", NULL, MTX_DEF)
212 #define	MIF6_LOCK_DESTROY()	mtx_destroy(&mif6_mtx)
213 
214 #ifdef MRT6DEBUG
215 static VNET_DEFINE(u_int, mrt6debug) = 0;	/* debug level */
216 #define	V_mrt6debug		VNET(mrt6debug)
217 #define DEBUG_MFC	0x02
218 #define DEBUG_FORWARD	0x04
219 #define DEBUG_EXPIRE	0x08
220 #define DEBUG_XMIT	0x10
221 #define DEBUG_REG	0x20
222 #define DEBUG_PIM	0x40
223 #define	DEBUG_ERR	0x80
224 #define	DEBUG_ANY	0x7f
225 #define	MRT6_DLOG(m, fmt, ...)	\
226 	if (V_mrt6debug & (m))	\
227 		log(((m) & DEBUG_ERR) ? LOG_ERR: LOG_DEBUG, \
228 		    "%s: " fmt "\n", __func__, ##__VA_ARGS__)
229 #else
230 #define	MRT6_DLOG(m, fmt, ...)
231 #endif
232 
233 static void	expire_upcalls(void *);
234 #define	EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
235 #define	UPCALL_EXPIRE	6		/* number of timeouts */
236 
237 /*
238  * XXX TODO: maintain a count to if_allmulti() calls in struct ifnet.
239  */
240 
241 /*
242  * 'Interfaces' associated with decapsulator (so we can tell
243  * packets that went through it from ones that get reflected
244  * by a broken gateway).  Different from IPv4 register_if,
245  * these interfaces are linked into the system ifnet list,
246  * because per-interface IPv6 statistics are maintained in
247  * ifp->if_afdata.  But it does not have any routes point
248  * to them.  I.e., packets can't be sent this way.  They
249  * only exist as a placeholder for multicast source
250  * verification.
251  */
252 static struct ifnet *multicast_register_if6;
253 
254 #define ENCAP_HOPS 64
255 
256 /*
257  * Private variables.
258  */
259 static mifi_t nummifs = 0;
260 static mifi_t reg_mif_num = (mifi_t)-1;
261 
262 static struct pim6stat pim6stat;
263 SYSCTL_STRUCT(_net_inet6_pim, PIM6CTL_STATS, stats, CTLFLAG_RW,
264     &pim6stat, pim6stat,
265     "PIM Statistics (struct pim6stat, netinet6/pim6_var.h)");
266 
267 #define	PIM6STAT_INC(name)	pim6stat.name += 1
268 static VNET_DEFINE(int, pim6);
269 #define	V_pim6		VNET(pim6)
270 
271 /*
272  * Hash function for a source, group entry
273  */
274 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
275 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
276 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
277 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
278 
279 /*
280  * Find a route for a given origin IPv6 address and Multicast group address.
281  */
282 #define MF6CFIND(o, g, rt) do { \
283 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
284 	rt = NULL; \
285 	while (_rt) { \
286 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
287 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
288 		    (_rt->mf6c_stall == NULL)) { \
289 			rt = _rt; \
290 			break; \
291 		} \
292 		_rt = _rt->mf6c_next; \
293 	} \
294 	if (rt == NULL) { \
295 		MRT6STAT_INC(mrt6s_mfc_misses); \
296 	} \
297 } while (/*CONSTCOND*/ 0)
298 
299 /*
300  * Macros to compute elapsed time efficiently
301  * Borrowed from Van Jacobson's scheduling code
302  * XXX: replace with timersub() ?
303  */
304 #define TV_DELTA(a, b, delta) do { \
305 	    int xxs; \
306 		\
307 	    delta = (a).tv_usec - (b).tv_usec; \
308 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
309 	       switch (xxs) { \
310 		      case 2: \
311 			  delta += 1000000; \
312 			      /* FALLTHROUGH */ \
313 		      case 1: \
314 			  delta += 1000000; \
315 			  break; \
316 		      default: \
317 			  delta += (1000000 * xxs); \
318 	       } \
319 	    } \
320 } while (/*CONSTCOND*/ 0)
321 
322 /* XXX: replace with timercmp(a, b, <) ? */
323 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
324 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
325 
326 #ifdef UPCALL_TIMING
327 #define UPCALL_MAX	50
328 static u_long upcall_data[UPCALL_MAX + 1];
329 static void collate();
330 #endif /* UPCALL_TIMING */
331 
332 static int ip6_mrouter_init(struct socket *, int, int);
333 static int add_m6fc(struct mf6cctl *);
334 static int add_m6if(struct mif6ctl *);
335 static int del_m6fc(struct mf6cctl *);
336 static int del_m6if(mifi_t *);
337 static int del_m6if_locked(mifi_t *);
338 static int get_mif6_cnt(struct sioc_mif_req6 *);
339 static int get_sg_cnt(struct sioc_sg_req6 *);
340 
341 static struct callout expire_upcalls_ch;
342 
343 int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *);
344 int X_ip6_mrouter_done(void);
345 int X_ip6_mrouter_set(struct socket *, struct sockopt *);
346 int X_ip6_mrouter_get(struct socket *, struct sockopt *);
347 int X_mrt6_ioctl(u_long, caddr_t);
348 
349 /*
350  * Handle MRT setsockopt commands to modify the multicast routing tables.
351  */
352 int
353 X_ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
354 {
355 	int error = 0;
356 	int optval;
357 	struct mif6ctl mifc;
358 	struct mf6cctl mfcc;
359 	mifi_t mifi;
360 
361 	if (so != V_ip6_mrouter && sopt->sopt_name != MRT6_INIT)
362 		return (EPERM);
363 
364 	switch (sopt->sopt_name) {
365 	case MRT6_INIT:
366 #ifdef MRT6_OINIT
367 	case MRT6_OINIT:
368 #endif
369 		error = sooptcopyin(sopt, &optval, sizeof(optval),
370 		    sizeof(optval));
371 		if (error)
372 			break;
373 		error = ip6_mrouter_init(so, optval, sopt->sopt_name);
374 		break;
375 	case MRT6_DONE:
376 		error = X_ip6_mrouter_done();
377 		break;
378 	case MRT6_ADD_MIF:
379 		error = sooptcopyin(sopt, &mifc, sizeof(mifc), sizeof(mifc));
380 		if (error)
381 			break;
382 		error = add_m6if(&mifc);
383 		break;
384 	case MRT6_ADD_MFC:
385 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
386 		if (error)
387 			break;
388 		error = add_m6fc(&mfcc);
389 		break;
390 	case MRT6_DEL_MFC:
391 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
392 		if (error)
393 			break;
394 		error = del_m6fc(&mfcc);
395 		break;
396 	case MRT6_DEL_MIF:
397 		error = sooptcopyin(sopt, &mifi, sizeof(mifi), sizeof(mifi));
398 		if (error)
399 			break;
400 		error = del_m6if(&mifi);
401 		break;
402 	case MRT6_PIM:
403 		error = sooptcopyin(sopt, &optval, sizeof(optval),
404 		    sizeof(optval));
405 		if (error)
406 			break;
407 		error = set_pim6(&optval);
408 		break;
409 	default:
410 		error = EOPNOTSUPP;
411 		break;
412 	}
413 
414 	return (error);
415 }
416 
417 /*
418  * Handle MRT getsockopt commands
419  */
420 int
421 X_ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
422 {
423 	int error = 0;
424 
425 	if (so != V_ip6_mrouter)
426 		return (EACCES);
427 
428 	switch (sopt->sopt_name) {
429 		case MRT6_PIM:
430 			error = sooptcopyout(sopt, &V_pim6, sizeof(V_pim6));
431 			break;
432 	}
433 	return (error);
434 }
435 
436 /*
437  * Handle ioctl commands to obtain information from the cache
438  */
439 int
440 X_mrt6_ioctl(u_long cmd, caddr_t data)
441 {
442 	int ret;
443 
444 	ret = EINVAL;
445 
446 	switch (cmd) {
447 	case SIOCGETSGCNT_IN6:
448 		ret = get_sg_cnt((struct sioc_sg_req6 *)data);
449 		break;
450 
451 	case SIOCGETMIFCNT_IN6:
452 		ret = get_mif6_cnt((struct sioc_mif_req6 *)data);
453 		break;
454 
455 	default:
456 		break;
457 	}
458 
459 	return (ret);
460 }
461 
462 /*
463  * returns the packet, byte, rpf-failure count for the source group provided
464  */
465 static int
466 get_sg_cnt(struct sioc_sg_req6 *req)
467 {
468 	struct mf6c *rt;
469 	int ret;
470 
471 	ret = 0;
472 
473 	MFC6_LOCK();
474 
475 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
476 	if (rt == NULL) {
477 		ret = ESRCH;
478 	} else {
479 		req->pktcnt = rt->mf6c_pkt_cnt;
480 		req->bytecnt = rt->mf6c_byte_cnt;
481 		req->wrong_if = rt->mf6c_wrong_if;
482 	}
483 
484 	MFC6_UNLOCK();
485 
486 	return (ret);
487 }
488 
489 /*
490  * returns the input and output packet and byte counts on the mif provided
491  */
492 static int
493 get_mif6_cnt(struct sioc_mif_req6 *req)
494 {
495 	mifi_t mifi;
496 	int ret;
497 
498 	ret = 0;
499 	mifi = req->mifi;
500 
501 	MIF6_LOCK();
502 
503 	if (mifi >= nummifs) {
504 		ret = EINVAL;
505 	} else {
506 		req->icount = mif6table[mifi].m6_pkt_in;
507 		req->ocount = mif6table[mifi].m6_pkt_out;
508 		req->ibytes = mif6table[mifi].m6_bytes_in;
509 		req->obytes = mif6table[mifi].m6_bytes_out;
510 	}
511 
512 	MIF6_UNLOCK();
513 
514 	return (ret);
515 }
516 
517 static int
518 set_pim6(int *i)
519 {
520 	if ((*i != 1) && (*i != 0))
521 		return (EINVAL);
522 
523 	V_pim6 = *i;
524 
525 	return (0);
526 }
527 
528 /*
529  * Enable multicast routing
530  */
531 static int
532 ip6_mrouter_init(struct socket *so, int v, int cmd)
533 {
534 
535 	MRT6_DLOG(DEBUG_ANY, "so_type = %d, pr_protocol = %d",
536 	    so->so_type, so->so_proto->pr_protocol);
537 
538 	if (so->so_type != SOCK_RAW ||
539 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
540 		return (EOPNOTSUPP);
541 
542 	if (v != 1)
543 		return (ENOPROTOOPT);
544 
545 	MROUTER6_LOCK();
546 
547 	if (V_ip6_mrouter != NULL) {
548 		MROUTER6_UNLOCK();
549 		return (EADDRINUSE);
550 	}
551 
552 	V_ip6_mrouter = so;
553 	V_ip6_mrouter_ver = cmd;
554 
555 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
556 	bzero((caddr_t)n6expire, sizeof(n6expire));
557 
558 	V_pim6 = 0;/* used for stubbing out/in pim stuff */
559 
560 	callout_init(&expire_upcalls_ch, 0);
561 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
562 	    expire_upcalls, NULL);
563 
564 	MROUTER6_UNLOCK();
565 	MRT6_DLOG(DEBUG_ANY, "finished");
566 
567 	return (0);
568 }
569 
570 /*
571  * Disable IPv6 multicast forwarding.
572  */
573 int
574 X_ip6_mrouter_done(void)
575 {
576 	mifi_t mifi;
577 	u_long i;
578 	struct mf6c *rt;
579 	struct rtdetq *rte;
580 
581 	MROUTER6_LOCK();
582 
583 	if (V_ip6_mrouter == NULL) {
584 		MROUTER6_UNLOCK();
585 		return (EINVAL);
586 	}
587 
588 	/*
589 	 * For each phyint in use, disable promiscuous reception of all IPv6
590 	 * multicasts.
591 	 */
592 	for (mifi = 0; mifi < nummifs; mifi++) {
593 		if (mif6table[mifi].m6_ifp &&
594 		    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
595 			if_allmulti(mif6table[mifi].m6_ifp, 0);
596 		}
597 	}
598 	bzero((caddr_t)mif6table, sizeof(mif6table));
599 	nummifs = 0;
600 
601 	V_pim6 = 0; /* used to stub out/in pim specific code */
602 
603 	callout_stop(&expire_upcalls_ch);
604 
605 	/*
606 	 * Free all multicast forwarding cache entries.
607 	 */
608 	MFC6_LOCK();
609 	for (i = 0; i < MF6CTBLSIZ; i++) {
610 		rt = mf6ctable[i];
611 		while (rt) {
612 			struct mf6c *frt;
613 
614 			for (rte = rt->mf6c_stall; rte != NULL; ) {
615 				struct rtdetq *n = rte->next;
616 
617 				m_freem(rte->m);
618 				free(rte, M_MRTABLE6);
619 				rte = n;
620 			}
621 			frt = rt;
622 			rt = rt->mf6c_next;
623 			free(frt, M_MRTABLE6);
624 		}
625 	}
626 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
627 	MFC6_UNLOCK();
628 
629 	/*
630 	 * Reset register interface
631 	 */
632 	if (reg_mif_num != (mifi_t)-1 && multicast_register_if6 != NULL) {
633 		if_detach(multicast_register_if6);
634 		if_free(multicast_register_if6);
635 		reg_mif_num = (mifi_t)-1;
636 		multicast_register_if6 = NULL;
637 	}
638 
639 	V_ip6_mrouter = NULL;
640 	V_ip6_mrouter_ver = 0;
641 
642 	MROUTER6_UNLOCK();
643 	MRT6_DLOG(DEBUG_ANY, "finished");
644 
645 	return (0);
646 }
647 
648 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
649 
650 /*
651  * Add a mif to the mif table
652  */
653 static int
654 add_m6if(struct mif6ctl *mifcp)
655 {
656 	struct mif6 *mifp;
657 	struct ifnet *ifp;
658 	int error;
659 
660 	MIF6_LOCK();
661 
662 	if (mifcp->mif6c_mifi >= MAXMIFS) {
663 		MIF6_UNLOCK();
664 		return (EINVAL);
665 	}
666 	mifp = mif6table + mifcp->mif6c_mifi;
667 	if (mifp->m6_ifp != NULL) {
668 		MIF6_UNLOCK();
669 		return (EADDRINUSE); /* XXX: is it appropriate? */
670 	}
671 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > V_if_index) {
672 		MIF6_UNLOCK();
673 		return (ENXIO);
674 	}
675 
676 	ifp = ifnet_byindex(mifcp->mif6c_pifi);
677 
678 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
679 		if (reg_mif_num == (mifi_t)-1) {
680 			ifp = if_alloc(IFT_OTHER);
681 
682 			if_initname(ifp, "register_mif", 0);
683 			ifp->if_flags |= IFF_LOOPBACK;
684 			if_attach(ifp);
685 			multicast_register_if6 = ifp;
686 			reg_mif_num = mifcp->mif6c_mifi;
687 			/*
688 			 * it is impossible to guess the ifindex of the
689 			 * register interface.  So mif6c_pifi is automatically
690 			 * calculated.
691 			 */
692 			mifcp->mif6c_pifi = ifp->if_index;
693 		} else {
694 			ifp = multicast_register_if6;
695 		}
696 	} else {
697 		/* Make sure the interface supports multicast */
698 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
699 			MIF6_UNLOCK();
700 			return (EOPNOTSUPP);
701 		}
702 
703 		error = if_allmulti(ifp, 1);
704 		if (error) {
705 			MIF6_UNLOCK();
706 			return (error);
707 		}
708 	}
709 
710 	mifp->m6_flags     = mifcp->mif6c_flags;
711 	mifp->m6_ifp       = ifp;
712 
713 	/* initialize per mif pkt counters */
714 	mifp->m6_pkt_in    = 0;
715 	mifp->m6_pkt_out   = 0;
716 	mifp->m6_bytes_in  = 0;
717 	mifp->m6_bytes_out = 0;
718 
719 	/* Adjust nummifs up if the mifi is higher than nummifs */
720 	if (nummifs <= mifcp->mif6c_mifi)
721 		nummifs = mifcp->mif6c_mifi + 1;
722 
723 	MIF6_UNLOCK();
724 	MRT6_DLOG(DEBUG_ANY, "mif #%d, phyint %s", mifcp->mif6c_mifi,
725 	    if_name(ifp));
726 
727 	return (0);
728 }
729 
730 /*
731  * Delete a mif from the mif table
732  */
733 static int
734 del_m6if_locked(mifi_t *mifip)
735 {
736 	struct mif6 *mifp = mif6table + *mifip;
737 	mifi_t mifi;
738 	struct ifnet *ifp;
739 
740 	MIF6_LOCK_ASSERT();
741 
742 	if (*mifip >= nummifs)
743 		return (EINVAL);
744 	if (mifp->m6_ifp == NULL)
745 		return (EINVAL);
746 
747 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
748 		/* XXX: TODO: Maintain an ALLMULTI refcount in struct ifnet. */
749 		ifp = mifp->m6_ifp;
750 		if_allmulti(ifp, 0);
751 	} else {
752 		if (reg_mif_num != (mifi_t)-1 &&
753 		    multicast_register_if6 != NULL) {
754 			if_detach(multicast_register_if6);
755 			if_free(multicast_register_if6);
756 			reg_mif_num = (mifi_t)-1;
757 			multicast_register_if6 = NULL;
758 		}
759 	}
760 
761 	bzero((caddr_t)mifp, sizeof(*mifp));
762 
763 	/* Adjust nummifs down */
764 	for (mifi = nummifs; mifi > 0; mifi--)
765 		if (mif6table[mifi - 1].m6_ifp)
766 			break;
767 	nummifs = mifi;
768 	MRT6_DLOG(DEBUG_ANY, "mif %d, nummifs %d", *mifip, nummifs);
769 
770 	return (0);
771 }
772 
773 static int
774 del_m6if(mifi_t *mifip)
775 {
776 	int cc;
777 
778 	MIF6_LOCK();
779 	cc = del_m6if_locked(mifip);
780 	MIF6_UNLOCK();
781 
782 	return (cc);
783 }
784 
785 /*
786  * Add an mfc entry
787  */
788 static int
789 add_m6fc(struct mf6cctl *mfccp)
790 {
791 	struct mf6c *rt;
792 	u_long hash;
793 	struct rtdetq *rte;
794 	u_short nstl;
795 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
796 
797 	MFC6_LOCK();
798 
799 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
800 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
801 
802 	/* If an entry already exists, just update the fields */
803 	if (rt) {
804 		MRT6_DLOG(DEBUG_MFC, "no upcall o %s g %s p %x",
805 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
806 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
807 		    mfccp->mf6cc_parent);
808 
809 		rt->mf6c_parent = mfccp->mf6cc_parent;
810 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
811 
812 		MFC6_UNLOCK();
813 		return (0);
814 	}
815 
816 	/*
817 	 * Find the entry for which the upcall was made and update
818 	 */
819 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
820 			mfccp->mf6cc_mcastgrp.sin6_addr);
821 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
822 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
823 				       &mfccp->mf6cc_origin.sin6_addr) &&
824 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
825 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
826 		    (rt->mf6c_stall != NULL)) {
827 
828 			if (nstl++)
829 				log(LOG_ERR,
830 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
831 				    "multiple kernel entries",
832 				    ip6_sprintf(ip6bufo,
833 					    &mfccp->mf6cc_origin.sin6_addr),
834 				    ip6_sprintf(ip6bufg,
835 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
836 				    mfccp->mf6cc_parent, rt->mf6c_stall);
837 
838 			MRT6_DLOG(DEBUG_MFC, "o %s g %s p %x dbg %p",
839 			    ip6_sprintf(ip6bufo,
840 			    &mfccp->mf6cc_origin.sin6_addr),
841 			    ip6_sprintf(ip6bufg,
842 				&mfccp->mf6cc_mcastgrp.sin6_addr),
843 			    mfccp->mf6cc_parent, rt->mf6c_stall);
844 
845 			rt->mf6c_origin     = mfccp->mf6cc_origin;
846 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
847 			rt->mf6c_parent     = mfccp->mf6cc_parent;
848 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
849 			/* initialize pkt counters per src-grp */
850 			rt->mf6c_pkt_cnt    = 0;
851 			rt->mf6c_byte_cnt   = 0;
852 			rt->mf6c_wrong_if   = 0;
853 
854 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
855 			n6expire[hash]--;
856 
857 			/* free packets Qed at the end of this entry */
858 			for (rte = rt->mf6c_stall; rte != NULL; ) {
859 				struct rtdetq *n = rte->next;
860 				ip6_mdq(rte->m, rte->ifp, rt);
861 				m_freem(rte->m);
862 #ifdef UPCALL_TIMING
863 				collate(&(rte->t));
864 #endif /* UPCALL_TIMING */
865 				free(rte, M_MRTABLE6);
866 				rte = n;
867 			}
868 			rt->mf6c_stall = NULL;
869 		}
870 	}
871 
872 	/*
873 	 * It is possible that an entry is being inserted without an upcall
874 	 */
875 	if (nstl == 0) {
876 		MRT6_DLOG(DEBUG_MFC, "no upcall h %lu o %s g %s p %x", hash,
877 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
878 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
879 		    mfccp->mf6cc_parent);
880 
881 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
882 
883 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
884 					       &mfccp->mf6cc_origin.sin6_addr)&&
885 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
886 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
887 
888 				rt->mf6c_origin     = mfccp->mf6cc_origin;
889 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
890 				rt->mf6c_parent     = mfccp->mf6cc_parent;
891 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
892 				/* initialize pkt counters per src-grp */
893 				rt->mf6c_pkt_cnt    = 0;
894 				rt->mf6c_byte_cnt   = 0;
895 				rt->mf6c_wrong_if   = 0;
896 
897 				if (rt->mf6c_expire)
898 					n6expire[hash]--;
899 				rt->mf6c_expire	   = 0;
900 			}
901 		}
902 		if (rt == NULL) {
903 			/* no upcall, so make a new entry */
904 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
905 						  M_NOWAIT);
906 			if (rt == NULL) {
907 				MFC6_UNLOCK();
908 				return (ENOBUFS);
909 			}
910 
911 			/* insert new entry at head of hash chain */
912 			rt->mf6c_origin     = mfccp->mf6cc_origin;
913 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
914 			rt->mf6c_parent     = mfccp->mf6cc_parent;
915 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
916 			/* initialize pkt counters per src-grp */
917 			rt->mf6c_pkt_cnt    = 0;
918 			rt->mf6c_byte_cnt   = 0;
919 			rt->mf6c_wrong_if   = 0;
920 			rt->mf6c_expire     = 0;
921 			rt->mf6c_stall = NULL;
922 
923 			/* link into table */
924 			rt->mf6c_next  = mf6ctable[hash];
925 			mf6ctable[hash] = rt;
926 		}
927 	}
928 
929 	MFC6_UNLOCK();
930 	return (0);
931 }
932 
933 #ifdef UPCALL_TIMING
934 /*
935  * collect delay statistics on the upcalls
936  */
937 static void
938 collate(struct timeval *t)
939 {
940 	u_long d;
941 	struct timeval tp;
942 	u_long delta;
943 
944 	GET_TIME(tp);
945 
946 	if (TV_LT(*t, tp))
947 	{
948 		TV_DELTA(tp, *t, delta);
949 
950 		d = delta >> 10;
951 		if (d > UPCALL_MAX)
952 			d = UPCALL_MAX;
953 
954 		++upcall_data[d];
955 	}
956 }
957 #endif /* UPCALL_TIMING */
958 
959 /*
960  * Delete an mfc entry
961  */
962 static int
963 del_m6fc(struct mf6cctl *mfccp)
964 {
965 #ifdef MRT6DEBUG
966 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
967 #endif
968 	struct sockaddr_in6	origin;
969 	struct sockaddr_in6	mcastgrp;
970 	struct mf6c		*rt;
971 	struct mf6c		**nptr;
972 	u_long		hash;
973 
974 	origin = mfccp->mf6cc_origin;
975 	mcastgrp = mfccp->mf6cc_mcastgrp;
976 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
977 
978 	MRT6_DLOG(DEBUG_MFC, "orig %s mcastgrp %s",
979 	    ip6_sprintf(ip6bufo, &origin.sin6_addr),
980 	    ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr));
981 
982 	MFC6_LOCK();
983 
984 	nptr = &mf6ctable[hash];
985 	while ((rt = *nptr) != NULL) {
986 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
987 				       &rt->mf6c_origin.sin6_addr) &&
988 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
989 				       &rt->mf6c_mcastgrp.sin6_addr) &&
990 		    rt->mf6c_stall == NULL)
991 			break;
992 
993 		nptr = &rt->mf6c_next;
994 	}
995 	if (rt == NULL) {
996 		MFC6_UNLOCK();
997 		return (EADDRNOTAVAIL);
998 	}
999 
1000 	*nptr = rt->mf6c_next;
1001 	free(rt, M_MRTABLE6);
1002 
1003 	MFC6_UNLOCK();
1004 
1005 	return (0);
1006 }
1007 
1008 static int
1009 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
1010 {
1011 
1012 	if (s) {
1013 		if (sbappendaddr(&s->so_rcv,
1014 				 (struct sockaddr *)src,
1015 				 mm, (struct mbuf *)0) != 0) {
1016 			sorwakeup(s);
1017 			return (0);
1018 		}
1019 	}
1020 	m_freem(mm);
1021 	return (-1);
1022 }
1023 
1024 /*
1025  * IPv6 multicast forwarding function. This function assumes that the packet
1026  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
1027  * pointed to by "ifp", and the packet is to be relayed to other networks
1028  * that have members of the packet's destination IPv6 multicast group.
1029  *
1030  * The packet is returned unscathed to the caller, unless it is
1031  * erroneous, in which case a non-zero return value tells the caller to
1032  * discard it.
1033  *
1034  * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff
1035  * this function is called in the originating context (i.e., not when
1036  * forwarding a packet from other node).  ip6_output(), which is currently the
1037  * only function that calls this function is called in the originating context,
1038  * explicitly ensures this condition.  It is caller's responsibility to ensure
1039  * that if this function is called from somewhere else in the originating
1040  * context in the future.
1041  */
1042 int
1043 X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
1044 {
1045 	struct rtdetq *rte;
1046 	struct mbuf *mb0;
1047 	struct mf6c *rt;
1048 	struct mif6 *mifp;
1049 	struct mbuf *mm;
1050 	u_long hash;
1051 	mifi_t mifi;
1052 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1053 #ifdef UPCALL_TIMING
1054 	struct timeval tp;
1055 
1056 	GET_TIME(tp);
1057 #endif /* UPCALL_TIMING */
1058 
1059 	MRT6_DLOG(DEBUG_FORWARD, "src %s, dst %s, ifindex %d",
1060 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1061 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst), ifp->if_index);
1062 
1063 	/*
1064 	 * Don't forward a packet with Hop limit of zero or one,
1065 	 * or a packet destined to a local-only group.
1066 	 */
1067 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
1068 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1069 		return (0);
1070 	ip6->ip6_hlim--;
1071 
1072 	/*
1073 	 * Source address check: do not forward packets with unspecified
1074 	 * source. It was discussed in July 2000, on ipngwg mailing list.
1075 	 * This is rather more serious than unicast cases, because some
1076 	 * MLD packets can be sent with the unspecified source address
1077 	 * (although such packets must normally set 1 to the hop limit field).
1078 	 */
1079 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1080 		IP6STAT_INC(ip6s_cantforward);
1081 		if (V_ip6_log_time + V_ip6_log_interval < time_uptime) {
1082 			V_ip6_log_time = time_uptime;
1083 			log(LOG_DEBUG,
1084 			    "cannot forward "
1085 			    "from %s to %s nxt %d received on %s\n",
1086 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1087 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1088 			    ip6->ip6_nxt,
1089 			    if_name(m->m_pkthdr.rcvif));
1090 		}
1091 		return (0);
1092 	}
1093 
1094 	MFC6_LOCK();
1095 
1096 	/*
1097 	 * Determine forwarding mifs from the forwarding cache table
1098 	 */
1099 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
1100 	MRT6STAT_INC(mrt6s_mfc_lookups);
1101 
1102 	/* Entry exists, so forward if necessary */
1103 	if (rt) {
1104 		MFC6_UNLOCK();
1105 		return (ip6_mdq(m, ifp, rt));
1106 	}
1107 
1108 	/*
1109 	 * If we don't have a route for packet's origin,
1110 	 * Make a copy of the packet & send message to routing daemon.
1111 	 */
1112 	MRT6STAT_INC(mrt6s_no_route);
1113 	MRT6_DLOG(DEBUG_FORWARD | DEBUG_MFC, "no rte s %s g %s",
1114 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1115 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1116 
1117 	/*
1118 	 * Allocate mbufs early so that we don't do extra work if we
1119 	 * are just going to fail anyway.
1120 	 */
1121 	rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE6, M_NOWAIT);
1122 	if (rte == NULL) {
1123 		MFC6_UNLOCK();
1124 		return (ENOBUFS);
1125 	}
1126 	mb0 = m_copy(m, 0, M_COPYALL);
1127 	/*
1128 	 * Pullup packet header if needed before storing it,
1129 	 * as other references may modify it in the meantime.
1130 	 */
1131 	if (mb0 && (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1132 		mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1133 	if (mb0 == NULL) {
1134 		free(rte, M_MRTABLE6);
1135 		MFC6_UNLOCK();
1136 		return (ENOBUFS);
1137 	}
1138 
1139 	/* is there an upcall waiting for this packet? */
1140 	hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1141 	for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1142 		if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1143 		    &rt->mf6c_origin.sin6_addr) &&
1144 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1145 		    &rt->mf6c_mcastgrp.sin6_addr) && (rt->mf6c_stall != NULL))
1146 			break;
1147 	}
1148 
1149 	if (rt == NULL) {
1150 		struct mrt6msg *im;
1151 #ifdef MRT6_OINIT
1152 		struct omrt6msg *oim;
1153 #endif
1154 		/* no upcall, so make a new entry */
1155 		rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6, M_NOWAIT);
1156 		if (rt == NULL) {
1157 			free(rte, M_MRTABLE6);
1158 			m_freem(mb0);
1159 			MFC6_UNLOCK();
1160 			return (ENOBUFS);
1161 		}
1162 		/*
1163 		 * Make a copy of the header to send to the user
1164 		 * level process
1165 		 */
1166 		mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1167 		if (mm == NULL) {
1168 			free(rte, M_MRTABLE6);
1169 			m_freem(mb0);
1170 			free(rt, M_MRTABLE6);
1171 			MFC6_UNLOCK();
1172 			return (ENOBUFS);
1173 		}
1174 
1175 		/*
1176 		 * Send message to routing daemon
1177 		 */
1178 		sin6.sin6_addr = ip6->ip6_src;
1179 		im = NULL;
1180 #ifdef MRT6_OINIT
1181 		oim = NULL;
1182 #endif
1183 		switch (V_ip6_mrouter_ver) {
1184 #ifdef MRT6_OINIT
1185 		case MRT6_OINIT:
1186 			oim = mtod(mm, struct omrt6msg *);
1187 			oim->im6_msgtype = MRT6MSG_NOCACHE;
1188 			oim->im6_mbz = 0;
1189 			break;
1190 #endif
1191 		case MRT6_INIT:
1192 			im = mtod(mm, struct mrt6msg *);
1193 			im->im6_msgtype = MRT6MSG_NOCACHE;
1194 			im->im6_mbz = 0;
1195 			break;
1196 		default:
1197 			free(rte, M_MRTABLE6);
1198 			m_freem(mb0);
1199 			free(rt, M_MRTABLE6);
1200 			MFC6_UNLOCK();
1201 			return (EINVAL);
1202 		}
1203 
1204 		MRT6_DLOG(DEBUG_FORWARD, "getting the iif info in the kernel");
1205 		for (mifp = mif6table, mifi = 0;
1206 		    mifi < nummifs && mifp->m6_ifp != ifp; mifp++, mifi++)
1207 				;
1208 
1209 		switch (V_ip6_mrouter_ver) {
1210 #ifdef MRT6_OINIT
1211 		case MRT6_OINIT:
1212 			oim->im6_mif = mifi;
1213 			break;
1214 #endif
1215 		case MRT6_INIT:
1216 			im->im6_mif = mifi;
1217 			break;
1218 		}
1219 
1220 		if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1221 			log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1222 			    "socket queue full\n");
1223 			MRT6STAT_INC(mrt6s_upq_sockfull);
1224 			free(rte, M_MRTABLE6);
1225 			m_freem(mb0);
1226 			free(rt, M_MRTABLE6);
1227 			MFC6_UNLOCK();
1228 			return (ENOBUFS);
1229 		}
1230 
1231 		MRT6STAT_INC(mrt6s_upcalls);
1232 
1233 		/* insert new entry at head of hash chain */
1234 		bzero(rt, sizeof(*rt));
1235 		rt->mf6c_origin.sin6_family = AF_INET6;
1236 		rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1237 		rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1238 		rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1239 		rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1240 		rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1241 		rt->mf6c_expire = UPCALL_EXPIRE;
1242 		n6expire[hash]++;
1243 		rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1244 
1245 		/* link into table */
1246 		rt->mf6c_next  = mf6ctable[hash];
1247 		mf6ctable[hash] = rt;
1248 		/* Add this entry to the end of the queue */
1249 		rt->mf6c_stall = rte;
1250 	} else {
1251 		/* determine if q has overflowed */
1252 		struct rtdetq **p;
1253 		int npkts = 0;
1254 
1255 		for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1256 			if (++npkts > MAX_UPQ6) {
1257 				MRT6STAT_INC(mrt6s_upq_ovflw);
1258 				free(rte, M_MRTABLE6);
1259 				m_freem(mb0);
1260 				MFC6_UNLOCK();
1261 				return (0);
1262 			}
1263 
1264 		/* Add this entry to the end of the queue */
1265 		*p = rte;
1266 	}
1267 
1268 	rte->next = NULL;
1269 	rte->m = mb0;
1270 	rte->ifp = ifp;
1271 #ifdef UPCALL_TIMING
1272 	rte->t = tp;
1273 #endif /* UPCALL_TIMING */
1274 
1275 	MFC6_UNLOCK();
1276 
1277 	return (0);
1278 }
1279 
1280 /*
1281  * Clean up cache entries if upcalls are not serviced
1282  * Call from the Slow Timeout mechanism, every half second.
1283  */
1284 static void
1285 expire_upcalls(void *unused)
1286 {
1287 #ifdef MRT6DEBUG
1288 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1289 #endif
1290 	struct rtdetq *rte;
1291 	struct mf6c *mfc, **nptr;
1292 	u_long i;
1293 
1294 	MFC6_LOCK();
1295 	for (i = 0; i < MF6CTBLSIZ; i++) {
1296 		if (n6expire[i] == 0)
1297 			continue;
1298 		nptr = &mf6ctable[i];
1299 		while ((mfc = *nptr) != NULL) {
1300 			rte = mfc->mf6c_stall;
1301 			/*
1302 			 * Skip real cache entries
1303 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1304 			 * If it expires now
1305 			 */
1306 			if (rte != NULL &&
1307 			    mfc->mf6c_expire != 0 &&
1308 			    --mfc->mf6c_expire == 0) {
1309 				MRT6_DLOG(DEBUG_EXPIRE, "expiring (%s %s)",
1310 				    ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr),
1311 				    ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr));
1312 				/*
1313 				 * drop all the packets
1314 				 * free the mbuf with the pkt, if, timing info
1315 				 */
1316 				do {
1317 					struct rtdetq *n = rte->next;
1318 					m_freem(rte->m);
1319 					free(rte, M_MRTABLE6);
1320 					rte = n;
1321 				} while (rte != NULL);
1322 				MRT6STAT_INC(mrt6s_cache_cleanups);
1323 				n6expire[i]--;
1324 
1325 				*nptr = mfc->mf6c_next;
1326 				free(mfc, M_MRTABLE6);
1327 			} else {
1328 				nptr = &mfc->mf6c_next;
1329 			}
1330 		}
1331 	}
1332 	MFC6_UNLOCK();
1333 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1334 	    expire_upcalls, NULL);
1335 }
1336 
1337 /*
1338  * Packet forwarding routine once entry in the cache is made
1339  */
1340 static int
1341 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1342 {
1343 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1344 	mifi_t mifi, iif;
1345 	struct mif6 *mifp;
1346 	int plen = m->m_pkthdr.len;
1347 	struct in6_addr src0, dst0; /* copies for local work */
1348 	u_int32_t iszone, idzone, oszone, odzone;
1349 	int error = 0;
1350 
1351 /*
1352  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1353  * input, they shouldn't get counted on output, so statistics keeping is
1354  * separate.
1355  */
1356 
1357 #define MC6_SEND(ip6, mifp, m) do {				\
1358 	if ((mifp)->m6_flags & MIFF_REGISTER)			\
1359 		register_send((ip6), (mifp), (m));		\
1360 	else							\
1361 		phyint_send((ip6), (mifp), (m));		\
1362 } while (/*CONSTCOND*/ 0)
1363 
1364 	/*
1365 	 * Don't forward if it didn't arrive from the parent mif
1366 	 * for its origin.
1367 	 */
1368 	mifi = rt->mf6c_parent;
1369 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1370 		/* came in the wrong interface */
1371 		MRT6_DLOG(DEBUG_FORWARD,
1372 		    "wrong if: ifid %d mifi %d mififid %x", ifp->if_index,
1373 		    mifi, mif6table[mifi].m6_ifp->if_index);
1374 		MRT6STAT_INC(mrt6s_wrong_if);
1375 		rt->mf6c_wrong_if++;
1376 		/*
1377 		 * If we are doing PIM processing, and we are forwarding
1378 		 * packets on this interface, send a message to the
1379 		 * routing daemon.
1380 		 */
1381 		/* have to make sure this is a valid mif */
1382 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1383 			if (V_pim6 && (m->m_flags & M_LOOP) == 0) {
1384 				/*
1385 				 * Check the M_LOOP flag to avoid an
1386 				 * unnecessary PIM assert.
1387 				 * XXX: M_LOOP is an ad-hoc hack...
1388 				 */
1389 				static struct sockaddr_in6 sin6 =
1390 				{ sizeof(sin6), AF_INET6 };
1391 
1392 				struct mbuf *mm;
1393 				struct mrt6msg *im;
1394 #ifdef MRT6_OINIT
1395 				struct omrt6msg *oim;
1396 #endif
1397 
1398 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1399 				if (mm &&
1400 				    (M_HASCL(mm) ||
1401 				     mm->m_len < sizeof(struct ip6_hdr)))
1402 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1403 				if (mm == NULL)
1404 					return (ENOBUFS);
1405 
1406 #ifdef MRT6_OINIT
1407 				oim = NULL;
1408 #endif
1409 				im = NULL;
1410 				switch (V_ip6_mrouter_ver) {
1411 #ifdef MRT6_OINIT
1412 				case MRT6_OINIT:
1413 					oim = mtod(mm, struct omrt6msg *);
1414 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1415 					oim->im6_mbz = 0;
1416 					break;
1417 #endif
1418 				case MRT6_INIT:
1419 					im = mtod(mm, struct mrt6msg *);
1420 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1421 					im->im6_mbz = 0;
1422 					break;
1423 				default:
1424 					m_freem(mm);
1425 					return (EINVAL);
1426 				}
1427 
1428 				for (mifp = mif6table, iif = 0;
1429 				     iif < nummifs && mifp &&
1430 					     mifp->m6_ifp != ifp;
1431 				     mifp++, iif++)
1432 					;
1433 
1434 				switch (V_ip6_mrouter_ver) {
1435 #ifdef MRT6_OINIT
1436 				case MRT6_OINIT:
1437 					oim->im6_mif = iif;
1438 					sin6.sin6_addr = oim->im6_src;
1439 					break;
1440 #endif
1441 				case MRT6_INIT:
1442 					im->im6_mif = iif;
1443 					sin6.sin6_addr = im->im6_src;
1444 					break;
1445 				}
1446 
1447 				MRT6STAT_INC(mrt6s_upcalls);
1448 
1449 				if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1450 					MRT6_DLOG(DEBUG_ANY,
1451 					    "ip6_mrouter socket queue full");
1452 					MRT6STAT_INC(mrt6s_upq_sockfull);
1453 					return (ENOBUFS);
1454 				}	/* if socket Q full */
1455 			}		/* if PIM */
1456 		return (0);
1457 	}			/* if wrong iif */
1458 
1459 	/* If I sourced this packet, it counts as output, else it was input. */
1460 	if (m->m_pkthdr.rcvif == NULL) {
1461 		/* XXX: is rcvif really NULL when output?? */
1462 		mif6table[mifi].m6_pkt_out++;
1463 		mif6table[mifi].m6_bytes_out += plen;
1464 	} else {
1465 		mif6table[mifi].m6_pkt_in++;
1466 		mif6table[mifi].m6_bytes_in += plen;
1467 	}
1468 	rt->mf6c_pkt_cnt++;
1469 	rt->mf6c_byte_cnt += plen;
1470 
1471 	/*
1472 	 * For each mif, forward a copy of the packet if there are group
1473 	 * members downstream on the interface.
1474 	 */
1475 	src0 = ip6->ip6_src;
1476 	dst0 = ip6->ip6_dst;
1477 	if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 ||
1478 	    (error = in6_setscope(&dst0, ifp, &idzone)) != 0) {
1479 		IP6STAT_INC(ip6s_badscope);
1480 		return (error);
1481 	}
1482 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) {
1483 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1484 			/*
1485 			 * check if the outgoing packet is going to break
1486 			 * a scope boundary.
1487 			 * XXX For packets through PIM register tunnel
1488 			 * interface, we believe a routing daemon.
1489 			 */
1490 			if (!(mif6table[rt->mf6c_parent].m6_flags &
1491 			      MIFF_REGISTER) &&
1492 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
1493 				if (in6_setscope(&src0, mif6table[mifi].m6_ifp,
1494 				    &oszone) ||
1495 				    in6_setscope(&dst0, mif6table[mifi].m6_ifp,
1496 				    &odzone) ||
1497 				    iszone != oszone ||
1498 				    idzone != odzone) {
1499 					IP6STAT_INC(ip6s_badscope);
1500 					continue;
1501 				}
1502 			}
1503 
1504 			mifp->m6_pkt_out++;
1505 			mifp->m6_bytes_out += plen;
1506 			MC6_SEND(ip6, mifp, m);
1507 		}
1508 	}
1509 	return (0);
1510 }
1511 
1512 static void
1513 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1514 {
1515 #ifdef MRT6DEBUG
1516 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1517 #endif
1518 	struct mbuf *mb_copy;
1519 	struct ifnet *ifp = mifp->m6_ifp;
1520 	int error = 0;
1521 	u_long linkmtu;
1522 
1523 	/*
1524 	 * Make a new reference to the packet; make sure that
1525 	 * the IPv6 header is actually copied, not just referenced,
1526 	 * so that ip6_output() only scribbles on the copy.
1527 	 */
1528 	mb_copy = m_copy(m, 0, M_COPYALL);
1529 	if (mb_copy &&
1530 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1531 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1532 	if (mb_copy == NULL) {
1533 		return;
1534 	}
1535 	/* set MCAST flag to the outgoing packet */
1536 	mb_copy->m_flags |= M_MCAST;
1537 
1538 	/*
1539 	 * If we sourced the packet, call ip6_output since we may devide
1540 	 * the packet into fragments when the packet is too big for the
1541 	 * outgoing interface.
1542 	 * Otherwise, we can simply send the packet to the interface
1543 	 * sending queue.
1544 	 */
1545 	if (m->m_pkthdr.rcvif == NULL) {
1546 		struct ip6_moptions im6o;
1547 
1548 		im6o.im6o_multicast_ifp = ifp;
1549 		/* XXX: ip6_output will override ip6->ip6_hlim */
1550 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1551 		im6o.im6o_multicast_loop = 1;
1552 		error = ip6_output(mb_copy, NULL, NULL, IPV6_FORWARDING, &im6o,
1553 		    NULL, NULL);
1554 
1555 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1556 		    (uint16_t)(mifp - mif6table), error);
1557 		return;
1558 	}
1559 
1560 	/*
1561 	 * If configured to loop back multicasts by default,
1562 	 * loop back a copy now.
1563 	 */
1564 	if (in6_mcast_loop) {
1565 		struct sockaddr_in6 dst6;
1566 
1567 		bzero(&dst6, sizeof(dst6));
1568 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1569 		dst6.sin6_family = AF_INET6;
1570 		dst6.sin6_addr = ip6->ip6_dst;
1571 		ip6_mloopback(ifp, m, &dst6);
1572 	}
1573 
1574 	/*
1575 	 * Put the packet into the sending queue of the outgoing interface
1576 	 * if it would fit in the MTU of the interface.
1577 	 */
1578 	linkmtu = IN6_LINKMTU(ifp);
1579 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1580 		struct sockaddr_in6 dst6;
1581 
1582 		bzero(&dst6, sizeof(dst6));
1583 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1584 		dst6.sin6_family = AF_INET6;
1585 		dst6.sin6_addr = ip6->ip6_dst;
1586 
1587 		IP_PROBE(send, NULL, NULL, ip6, ifp, NULL, ip6);
1588 		/*
1589 		 * We just call if_output instead of nd6_output here, since
1590 		 * we need no ND for a multicast forwarded packet...right?
1591 		 */
1592 		m_clrprotoflags(m);	/* Avoid confusing lower layers. */
1593 		error = (*ifp->if_output)(ifp, mb_copy,
1594 		    (struct sockaddr *)&dst6, NULL);
1595 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1596 		    (uint16_t)(mifp - mif6table), error);
1597 	} else {
1598 		/*
1599 		 * pMTU discovery is intentionally disabled by default, since
1600 		 * various router may notify pMTU in multicast, which can be
1601 		 * a DDoS to a router
1602 		 */
1603 		if (V_ip6_mcast_pmtu)
1604 			icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1605 		else {
1606 			MRT6_DLOG(DEBUG_XMIT, " packet too big on %s o %s "
1607 			    "g %s size %d (discarded)", if_name(ifp),
1608 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1609 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1610 			    mb_copy->m_pkthdr.len);
1611 			m_freem(mb_copy); /* simply discard the packet */
1612 		}
1613 	}
1614 }
1615 
1616 static int
1617 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1618 {
1619 #ifdef MRT6DEBUG
1620 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1621 #endif
1622 	struct mbuf *mm;
1623 	int i, len = m->m_pkthdr.len;
1624 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1625 	struct mrt6msg *im6;
1626 
1627 	MRT6_DLOG(DEBUG_ANY, "src %s dst %s",
1628 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1629 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1630 	PIM6STAT_INC(pim6s_snd_registers);
1631 
1632 	/* Make a copy of the packet to send to the user level process. */
1633 	mm = m_gethdr(M_NOWAIT, MT_DATA);
1634 	if (mm == NULL)
1635 		return (ENOBUFS);
1636 	mm->m_data += max_linkhdr;
1637 	mm->m_len = sizeof(struct ip6_hdr);
1638 
1639 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1640 		m_freem(mm);
1641 		return (ENOBUFS);
1642 	}
1643 	i = MHLEN - M_LEADINGSPACE(mm);
1644 	if (i > len)
1645 		i = len;
1646 	mm = m_pullup(mm, i);
1647 	if (mm == NULL)
1648 		return (ENOBUFS);
1649 /* TODO: check it! */
1650 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1651 
1652 	/*
1653 	 * Send message to routing daemon
1654 	 */
1655 	sin6.sin6_addr = ip6->ip6_src;
1656 
1657 	im6 = mtod(mm, struct mrt6msg *);
1658 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1659 	im6->im6_mbz          = 0;
1660 
1661 	im6->im6_mif = mif - mif6table;
1662 
1663 	/* iif info is not given for reg. encap.n */
1664 	MRT6STAT_INC(mrt6s_upcalls);
1665 
1666 	if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1667 		MRT6_DLOG(DEBUG_ANY, "ip6_mrouter socket queue full");
1668 		MRT6STAT_INC(mrt6s_upq_sockfull);
1669 		return (ENOBUFS);
1670 	}
1671 	return (0);
1672 }
1673 
1674 /*
1675  * pim6_encapcheck() is called by the encap6_input() path at runtime to
1676  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
1677  * into the kernel.
1678  */
1679 static int
1680 pim6_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
1681 {
1682 
1683 #ifdef DIAGNOSTIC
1684     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
1685 #endif
1686     if (proto != IPPROTO_PIM)
1687 	return 0;	/* not for us; reject the datagram. */
1688 
1689     return 64;		/* claim the datagram. */
1690 }
1691 
1692 /*
1693  * PIM sparse mode hook
1694  * Receives the pim control messages, and passes them up to the listening
1695  * socket, using rip6_input.
1696  * The only message processed is the REGISTER pim message; the pim header
1697  * is stripped off, and the inner packet is passed to register_mforward.
1698  */
1699 int
1700 pim6_input(struct mbuf **mp, int *offp, int proto)
1701 {
1702 	struct pim *pim; /* pointer to a pim struct */
1703 	struct ip6_hdr *ip6;
1704 	int pimlen;
1705 	struct mbuf *m = *mp;
1706 	int minlen;
1707 	int off = *offp;
1708 
1709 	PIM6STAT_INC(pim6s_rcv_total);
1710 
1711 	ip6 = mtod(m, struct ip6_hdr *);
1712 	pimlen = m->m_pkthdr.len - *offp;
1713 
1714 	/*
1715 	 * Validate lengths
1716 	 */
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 #ifndef PULLDOWN_TEST
1739 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1740 	/* adjust pointer */
1741 	ip6 = mtod(m, struct ip6_hdr *);
1742 
1743 	/* adjust mbuf to point to the PIM header */
1744 	pim = (struct pim *)((caddr_t)ip6 + off);
1745 #else
1746 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1747 	if (pim == NULL) {
1748 		PIM6STAT_INC(pim6s_rcv_tooshort);
1749 		return (IPPROTO_DONE);
1750 	}
1751 #endif
1752 
1753 #define PIM6_CHECKSUM
1754 #ifdef PIM6_CHECKSUM
1755 	{
1756 		int cksumlen;
1757 
1758 		/*
1759 		 * Validate checksum.
1760 		 * If PIM REGISTER, exclude the data packet
1761 		 */
1762 		if (pim->pim_type == PIM_REGISTER)
1763 			cksumlen = PIM_MINLEN;
1764 		else
1765 			cksumlen = pimlen;
1766 
1767 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1768 			PIM6STAT_INC(pim6s_rcv_badsum);
1769 			MRT6_DLOG(DEBUG_PIM, "invalid checksum");
1770 			m_freem(m);
1771 			return (IPPROTO_DONE);
1772 		}
1773 	}
1774 #endif /* PIM_CHECKSUM */
1775 
1776 	/* PIM version check */
1777 	if (pim->pim_ver != PIM_VERSION) {
1778 		PIM6STAT_INC(pim6s_rcv_badversion);
1779 		MRT6_DLOG(DEBUG_ANY | DEBUG_ERR,
1780 		    "incorrect version %d, expecting %d",
1781 		    pim->pim_ver, PIM_VERSION);
1782 		m_freem(m);
1783 		return (IPPROTO_DONE);
1784 	}
1785 
1786 	if (pim->pim_type == PIM_REGISTER) {
1787 		/*
1788 		 * since this is a REGISTER, we'll make a copy of the register
1789 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1790 		 * routing daemon.
1791 		 */
1792 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1793 
1794 		struct mbuf *mcp;
1795 		struct ip6_hdr *eip6;
1796 		u_int32_t *reghdr;
1797 		int rc;
1798 #ifdef MRT6DEBUG
1799 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1800 #endif
1801 
1802 		PIM6STAT_INC(pim6s_rcv_registers);
1803 
1804 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1805 			MRT6_DLOG(DEBUG_PIM, "register mif not set: %d",
1806 			    reg_mif_num);
1807 			m_freem(m);
1808 			return (IPPROTO_DONE);
1809 		}
1810 
1811 		reghdr = (u_int32_t *)(pim + 1);
1812 
1813 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1814 			goto pim6_input_to_daemon;
1815 
1816 		/*
1817 		 * Validate length
1818 		 */
1819 		if (pimlen < PIM6_REG_MINLEN) {
1820 			PIM6STAT_INC(pim6s_rcv_tooshort);
1821 			PIM6STAT_INC(pim6s_rcv_badregisters);
1822 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "register packet "
1823 			    "size too small %d from %s",
1824 			    pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src));
1825 			m_freem(m);
1826 			return (IPPROTO_DONE);
1827 		}
1828 
1829 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1830 		MRT6_DLOG(DEBUG_PIM, "eip6: %s -> %s, eip6 plen %d",
1831 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1832 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1833 		    ntohs(eip6->ip6_plen));
1834 
1835 		/* verify the version number of the inner packet */
1836 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1837 			PIM6STAT_INC(pim6s_rcv_badregisters);
1838 			MRT6_DLOG(DEBUG_ANY, "invalid IP version (%d) "
1839 			    "of the inner packet",
1840 			    (eip6->ip6_vfc & IPV6_VERSION));
1841 			m_freem(m);
1842 			return (IPPROTO_NONE);
1843 		}
1844 
1845 		/* verify the inner packet is destined to a mcast group */
1846 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1847 			PIM6STAT_INC(pim6s_rcv_badregisters);
1848 			MRT6_DLOG(DEBUG_PIM, "inner packet of register "
1849 			    "is not multicast %s",
1850 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst));
1851 			m_freem(m);
1852 			return (IPPROTO_DONE);
1853 		}
1854 
1855 		/*
1856 		 * make a copy of the whole header to pass to the daemon later.
1857 		 */
1858 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1859 		if (mcp == NULL) {
1860 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "pim register: "
1861 			    "could not copy register head");
1862 			m_freem(m);
1863 			return (IPPROTO_DONE);
1864 		}
1865 
1866 		/*
1867 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1868 		 */
1869 		m_adj(m, off + PIM_MINLEN);
1870 		MRT6_DLOG(DEBUG_PIM, "forwarding decapsulated register: "
1871 		    "src %s, dst %s, mif %d",
1872 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1873 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst), reg_mif_num);
1874 
1875 		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1876 				dst.sin6_family, 0);
1877 
1878 		/* prepare the register head to send to the mrouting daemon */
1879 		m = mcp;
1880 	}
1881 
1882 	/*
1883 	 * Pass the PIM message up to the daemon; if it is a register message
1884 	 * pass the 'head' only up to the daemon. This includes the
1885 	 * encapsulator ip6 header, pim header, register header and the
1886 	 * encapsulated ip6 header.
1887 	 */
1888   pim6_input_to_daemon:
1889 	rip6_input(&m, offp, proto);
1890 	return (IPPROTO_DONE);
1891 }
1892 
1893 static int
1894 ip6_mroute_modevent(module_t mod, int type, void *unused)
1895 {
1896 
1897 	switch (type) {
1898 	case MOD_LOAD:
1899 		MROUTER6_LOCK_INIT();
1900 		MFC6_LOCK_INIT();
1901 		MIF6_LOCK_INIT();
1902 
1903 		pim6_encap_cookie = encap_attach_func(AF_INET6, IPPROTO_PIM,
1904 			pim6_encapcheck,
1905 			(const struct protosw *)&in6_pim_protosw, NULL);
1906 		if (pim6_encap_cookie == NULL) {
1907 			printf("ip6_mroute: unable to attach pim6 encap\n");
1908 			MIF6_LOCK_DESTROY();
1909 			MFC6_LOCK_DESTROY();
1910 			MROUTER6_LOCK_DESTROY();
1911 			return (EINVAL);
1912 		}
1913 
1914 		ip6_mforward = X_ip6_mforward;
1915 		ip6_mrouter_done = X_ip6_mrouter_done;
1916 		ip6_mrouter_get = X_ip6_mrouter_get;
1917 		ip6_mrouter_set = X_ip6_mrouter_set;
1918 		mrt6_ioctl = X_mrt6_ioctl;
1919 		break;
1920 
1921 	case MOD_UNLOAD:
1922 		if (V_ip6_mrouter != NULL)
1923 			return EINVAL;
1924 
1925 		if (pim6_encap_cookie) {
1926 			encap_detach(pim6_encap_cookie);
1927 			pim6_encap_cookie = NULL;
1928 		}
1929 		X_ip6_mrouter_done();
1930 		ip6_mforward = NULL;
1931 		ip6_mrouter_done = NULL;
1932 		ip6_mrouter_get = NULL;
1933 		ip6_mrouter_set = NULL;
1934 		mrt6_ioctl = NULL;
1935 
1936 		MIF6_LOCK_DESTROY();
1937 		MFC6_LOCK_DESTROY();
1938 		MROUTER6_LOCK_DESTROY();
1939 		break;
1940 
1941 	default:
1942 		return (EOPNOTSUPP);
1943 	}
1944 
1945 	return (0);
1946 }
1947 
1948 static moduledata_t ip6_mroutemod = {
1949 	"ip6_mroute",
1950 	ip6_mroute_modevent,
1951 	0
1952 };
1953 
1954 DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
1955