xref: /netbsd-src/sys/netinet6/ip6_mroute.c (revision d710132b4b8ce7f7cccaaf660cb16aa16b4077a0)
1 /*	$NetBSD: ip6_mroute.c,v 1.49 2003/06/24 07:39:26 itojun Exp $	*/
2 /*	$KAME: ip6_mroute.c,v 1.49 2001/07/25 09:21:18 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*	BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp	*/
34 
35 /*
36  * Copyright (c) 1989 Stephen Deering
37  * Copyright (c) 1992, 1993
38  *      The Regents of the University of California.  All rights reserved.
39  *
40  * This code is derived from software contributed to Berkeley by
41  * Stephen Deering of Stanford University.
42  *
43  * Redistribution and use in source and binary forms, with or without
44  * modification, are permitted provided that the following conditions
45  * are met:
46  * 1. Redistributions of source code must retain the above copyright
47  *    notice, this list of conditions and the following disclaimer.
48  * 2. Redistributions in binary form must reproduce the above copyright
49  *    notice, this list of conditions and the following disclaimer in the
50  *    documentation and/or other materials provided with the distribution.
51  * 3. All advertising materials mentioning features or use of this software
52  *    must display the following acknowledgement:
53  *      This product includes software developed by the University of
54  *      California, Berkeley and its contributors.
55  * 4. Neither the name of the University nor the names of its contributors
56  *    may be used to endorse or promote products derived from this software
57  *    without specific prior written permission.
58  *
59  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
60  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
61  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
62  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
63  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
64  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
65  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
66  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
67  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
68  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
69  * SUCH DAMAGE.
70  *
71  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
72  */
73 
74 /*
75  * IP multicast forwarding procedures
76  *
77  * Written by David Waitzman, BBN Labs, August 1988.
78  * Modified by Steve Deering, Stanford, February 1989.
79  * Modified by Mark J. Steiglitz, Stanford, May, 1991
80  * Modified by Van Jacobson, LBL, January 1993
81  * Modified by Ajit Thyagarajan, PARC, August 1993
82  * Modified by Bill Fenner, PARC, April 1994
83  *
84  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
85  */
86 
87 #include <sys/cdefs.h>
88 __KERNEL_RCSID(0, "$NetBSD: ip6_mroute.c,v 1.49 2003/06/24 07:39:26 itojun Exp $");
89 
90 #include "opt_inet.h"
91 #include "opt_mrouting.h"
92 
93 #ifndef _KERNEL
94 # ifdef KERNEL
95 #  define _KERNEL
96 # endif
97 #endif
98 
99 #include <sys/param.h>
100 #include <sys/systm.h>
101 #include <sys/callout.h>
102 #include <sys/mbuf.h>
103 #include <sys/socket.h>
104 #include <sys/socketvar.h>
105 #include <sys/sockio.h>
106 #include <sys/protosw.h>
107 #include <sys/errno.h>
108 #include <sys/time.h>
109 #include <sys/kernel.h>
110 #include <sys/ioctl.h>
111 #include <sys/syslog.h>
112 
113 #include <net/if.h>
114 #include <net/route.h>
115 #include <net/raw_cb.h>
116 
117 #include <netinet/in.h>
118 #include <netinet/in_var.h>
119 #ifdef MULTICAST_PMTUD
120 #include <netinet/icmp6.h>
121 #endif
122 
123 #include <netinet/ip6.h>
124 #include <netinet6/ip6_var.h>
125 #include <netinet6/ip6_mroute.h>
126 #include <netinet6/pim6.h>
127 #include <netinet6/pim6_var.h>
128 #include <netinet6/nd6.h>
129 
130 #include <net/net_osdep.h>
131 
132 static int ip6_mdq __P((struct mbuf *, struct ifnet *, struct mf6c *));
133 static void phyint_send __P((struct ip6_hdr *, struct mif6 *, struct mbuf *));
134 
135 static int set_pim6 __P((int *));
136 static int get_pim6 __P((struct mbuf *));
137 static int socket_send __P((struct socket *, struct mbuf *,
138 			    struct sockaddr_in6 *));
139 static int register_send __P((struct ip6_hdr *, struct mif6 *,
140 			      struct mbuf *));
141 
142 /*
143  * Globals.  All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static,
144  * except for netstat or debugging purposes.
145  */
146 struct socket  *ip6_mrouter = NULL;
147 int		ip6_mrouter_ver = 0;
148 int		ip6_mrtproto = IPPROTO_PIM;    /* for netstat only */
149 struct mrt6stat	mrt6stat;
150 
151 #define NO_RTE_FOUND 	0x1
152 #define RTE_FOUND	0x2
153 
154 struct mf6c	*mf6ctable[MF6CTBLSIZ];
155 u_char		n6expire[MF6CTBLSIZ];
156 static struct mif6 mif6table[MAXMIFS];
157 #ifdef MRT6DEBUG
158 u_int		mrt6debug = 0;	  /* debug level 	*/
159 #define		DEBUG_MFC	0x02
160 #define		DEBUG_FORWARD	0x04
161 #define		DEBUG_EXPIRE	0x08
162 #define		DEBUG_XMIT	0x10
163 #define         DEBUG_REG       0x20
164 #define         DEBUG_PIM       0x40
165 #endif
166 
167 static void	expire_upcalls __P((void *));
168 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
169 #define		UPCALL_EXPIRE	6		/* number of timeouts */
170 
171 #ifdef INET
172 #ifdef MROUTING
173 extern struct socket *ip_mrouter;
174 #endif
175 #endif
176 
177 /*
178  * 'Interfaces' associated with decapsulator (so we can tell
179  * packets that went through it from ones that get reflected
180  * by a broken gateway).  These interfaces are never linked into
181  * the system ifnet list & no routes point to them.  I.e., packets
182  * can't be sent this way.  They only exist as a placeholder for
183  * multicast source verification.
184  */
185 struct ifnet multicast_register_if;
186 
187 #define ENCAP_HOPS 64
188 
189 /*
190  * Private variables.
191  */
192 static mifi_t nummifs = 0;
193 static mifi_t reg_mif_num = (mifi_t)-1;
194 
195 static struct pim6stat pim6stat;
196 static int pim6;
197 
198 /*
199  * Hash function for a source, group entry
200  */
201 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
202 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
203 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
204 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
205 
206 /*
207  * Find a route for a given origin IPv6 address and Multicast group address.
208  * Quality of service parameter to be added in the future!!!
209  */
210 
211 #define MF6CFIND(o, g, rt) do { \
212 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
213 	rt = NULL; \
214 	mrt6stat.mrt6s_mfc_lookups++; \
215 	while (_rt) { \
216 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
217 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
218 		    (_rt->mf6c_stall == NULL)) { \
219 			rt = _rt; \
220 			break; \
221 		} \
222 		_rt = _rt->mf6c_next; \
223 	} \
224 	if (rt == NULL) { \
225 		mrt6stat.mrt6s_mfc_misses++; \
226 	} \
227 } while (/*CONSTCOND*/ 0)
228 
229 /*
230  * Macros to compute elapsed time efficiently
231  * Borrowed from Van Jacobson's scheduling code
232  */
233 #define TV_DELTA(a, b, delta) do { \
234 	    int xxs; \
235 		\
236 	    delta = (a).tv_usec - (b).tv_usec; \
237 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
238 	       switch (xxs) { \
239 		      case 2: \
240 			  delta += 1000000; \
241 			      /* FALLTHROUGH */ \
242 		      case 1: \
243 			  delta += 1000000; \
244 			  break; \
245 		      default: \
246 			  delta += (1000000 * xxs); \
247 	       } \
248 	    } \
249 } while (/*CONSTCOND*/ 0)
250 
251 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
252 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
253 
254 #ifdef UPCALL_TIMING
255 #define UPCALL_MAX	50
256 u_long upcall_data[UPCALL_MAX + 1];
257 static void collate();
258 #endif /* UPCALL_TIMING */
259 
260 static int get_sg_cnt __P((struct sioc_sg_req6 *));
261 static int get_mif6_cnt __P((struct sioc_mif_req6 *));
262 static int ip6_mrouter_init __P((struct socket *, struct mbuf *, int));
263 static int add_m6if __P((struct mif6ctl *));
264 static int del_m6if __P((mifi_t *));
265 static int add_m6fc __P((struct mf6cctl *));
266 static int del_m6fc __P((struct mf6cctl *));
267 
268 static struct callout expire_upcalls_ch = CALLOUT_INITIALIZER;
269 
270 /*
271  * Handle MRT setsockopt commands to modify the multicast routing tables.
272  */
273 int
274 ip6_mrouter_set(cmd, so, m)
275 	int cmd;
276 	struct socket *so;
277 	struct mbuf *m;
278 {
279 	if (cmd != MRT6_INIT && so != ip6_mrouter)
280 		return EACCES;
281 
282 	switch (cmd) {
283 	case MRT6_OINIT:	return ip6_mrouter_init(so, m, cmd);
284 	case MRT6_INIT:		return ip6_mrouter_init(so, m, cmd);
285 	case MRT6_DONE:		return ip6_mrouter_done();
286 	case MRT6_ADD_MIF:	return add_m6if(mtod(m, struct mif6ctl *));
287 	case MRT6_DEL_MIF:	return del_m6if(mtod(m, mifi_t *));
288 	case MRT6_ADD_MFC:	return add_m6fc(mtod(m, struct mf6cctl *));
289 	case MRT6_DEL_MFC:	return del_m6fc(mtod(m, struct mf6cctl *));
290 	case MRT6_PIM:		return set_pim6(mtod(m, int *));
291 	default:		return EOPNOTSUPP;
292 	}
293 }
294 
295 /*
296  * Handle MRT getsockopt commands
297  */
298 int
299 ip6_mrouter_get(cmd, so, m)
300 	int cmd;
301 	struct socket *so;
302 	struct mbuf **m;
303 {
304 	struct mbuf *mb;
305 
306 	if (so != ip6_mrouter) return EACCES;
307 
308 	*m = mb = m_get(M_WAIT, MT_SOOPTS);
309 
310 	switch (cmd) {
311 	case MRT6_PIM:
312 		return get_pim6(mb);
313 	default:
314 		m_free(mb);
315 		return EOPNOTSUPP;
316 	}
317 }
318 
319 /*
320  * Handle ioctl commands to obtain information from the cache
321  */
322 int
323 mrt6_ioctl(cmd, data)
324 	int cmd;
325 	caddr_t data;
326 {
327 
328 	switch (cmd) {
329 	case SIOCGETSGCNT_IN6:
330 		return (get_sg_cnt((struct sioc_sg_req6 *)data));
331 	case SIOCGETMIFCNT_IN6:
332 		return (get_mif6_cnt((struct sioc_mif_req6 *)data));
333 	default:
334 		return (EINVAL);
335 	}
336 }
337 
338 /*
339  * returns the packet, byte, rpf-failure count for the source group provided
340  */
341 static int
342 get_sg_cnt(req)
343 	struct sioc_sg_req6 *req;
344 {
345 	struct mf6c *rt;
346 	int s;
347 
348 	s = splsoftnet();
349 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
350 	splx(s);
351 	if (rt != NULL) {
352 		req->pktcnt = rt->mf6c_pkt_cnt;
353 		req->bytecnt = rt->mf6c_byte_cnt;
354 		req->wrong_if = rt->mf6c_wrong_if;
355 	} else
356 		return (ESRCH);
357 #if 0
358 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
359 #endif
360 
361 	return 0;
362 }
363 
364 /*
365  * returns the input and output packet and byte counts on the mif provided
366  */
367 static int
368 get_mif6_cnt(req)
369 	struct sioc_mif_req6 *req;
370 {
371 	mifi_t mifi = req->mifi;
372 
373 	if (mifi >= nummifs)
374 		return EINVAL;
375 
376 	req->icount = mif6table[mifi].m6_pkt_in;
377 	req->ocount = mif6table[mifi].m6_pkt_out;
378 	req->ibytes = mif6table[mifi].m6_bytes_in;
379 	req->obytes = mif6table[mifi].m6_bytes_out;
380 
381 	return 0;
382 }
383 
384 /*
385  * Get PIM processiong global
386  */
387 static int
388 get_pim6(m)
389 	struct mbuf *m;
390 {
391 	int *i;
392 
393 	i = mtod(m, int *);
394 
395 	*i = pim6;
396 
397 	return 0;
398 }
399 
400 static int
401 set_pim6(i)
402 	int *i;
403 {
404 	if ((*i != 1) && (*i != 0))
405 		return EINVAL;
406 
407 	pim6 = *i;
408 
409 	return 0;
410 }
411 
412 /*
413  * Enable multicast routing
414  */
415 static int
416 ip6_mrouter_init(so, m, cmd)
417 	struct socket *so;
418 	struct mbuf *m;
419 	int cmd;
420 {
421 	int *v;
422 
423 #ifdef MRT6DEBUG
424 	if (mrt6debug)
425 		log(LOG_DEBUG,
426 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
427 		    so->so_type, so->so_proto->pr_protocol);
428 #endif
429 
430 	if (so->so_type != SOCK_RAW ||
431 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
432 		return EOPNOTSUPP;
433 
434 	if (!m || (m->m_len != sizeof(int *)))
435 		return ENOPROTOOPT;
436 
437 	v = mtod(m, int *);
438 	if (*v != 1)
439 		return ENOPROTOOPT;
440 
441 	if (ip6_mrouter != NULL) return EADDRINUSE;
442 
443 	ip6_mrouter = so;
444 	ip6_mrouter_ver = cmd;
445 
446 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
447 	bzero((caddr_t)n6expire, sizeof(n6expire));
448 
449 	pim6 = 0;/* used for stubbing out/in pim stuff */
450 
451 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
452 	    expire_upcalls, NULL);
453 
454 #ifdef MRT6DEBUG
455 	if (mrt6debug)
456 		log(LOG_DEBUG, "ip6_mrouter_init\n");
457 #endif
458 
459 	return 0;
460 }
461 
462 /*
463  * Disable multicast routing
464  */
465 int
466 ip6_mrouter_done()
467 {
468 	mifi_t mifi;
469 	int i;
470 	struct ifnet *ifp;
471 	struct in6_ifreq ifr;
472 	struct mf6c *rt;
473 	struct rtdetq *rte;
474 	int s;
475 
476 	s = splsoftnet();
477 
478 	/*
479 	 * For each phyint in use, disable promiscuous reception of all IPv6
480 	 * multicasts.
481 	 */
482 #ifdef INET
483 #ifdef MROUTING
484 	/*
485 	 * If there is still IPv4 multicast routing daemon,
486 	 * we remain interfaces to receive all muliticasted packets.
487 	 * XXX: there may be an interface in which the IPv4 multicast
488 	 * daemon is not interested...
489 	 */
490 	if (!ip_mrouter)
491 #endif
492 #endif
493 	{
494 		for (mifi = 0; mifi < nummifs; mifi++) {
495 			if (mif6table[mifi].m6_ifp &&
496 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
497 				ifr.ifr_addr.sin6_family = AF_INET6;
498 				ifr.ifr_addr.sin6_addr= in6addr_any;
499 				ifp = mif6table[mifi].m6_ifp;
500 				(*ifp->if_ioctl)(ifp, SIOCDELMULTI,
501 						 (caddr_t)&ifr);
502 			}
503 		}
504 	}
505 #ifdef notyet
506 	bzero((caddr_t)qtable, sizeof(qtable));
507 	bzero((caddr_t)tbftable, sizeof(tbftable));
508 #endif
509 	bzero((caddr_t)mif6table, sizeof(mif6table));
510 	nummifs = 0;
511 
512 	pim6 = 0; /* used to stub out/in pim specific code */
513 
514 	callout_stop(&expire_upcalls_ch);
515 
516 	/*
517 	 * Free all multicast forwarding cache entries.
518 	 */
519 	for (i = 0; i < MF6CTBLSIZ; i++) {
520 		rt = mf6ctable[i];
521 		while (rt) {
522 			struct mf6c *frt;
523 
524 			for (rte = rt->mf6c_stall; rte != NULL; ) {
525 				struct rtdetq *n = rte->next;
526 
527 				m_free(rte->m);
528 				free(rte, M_MRTABLE);
529 				rte = n;
530 			}
531 			frt = rt;
532 			rt = rt->mf6c_next;
533 			free(frt, M_MRTABLE);
534 		}
535 	}
536 
537 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
538 
539 	/*
540 	 * Reset de-encapsulation cache
541 	 */
542 	reg_mif_num = -1;
543 
544 	ip6_mrouter = NULL;
545 	ip6_mrouter_ver = 0;
546 
547 	splx(s);
548 
549 #ifdef MRT6DEBUG
550 	if (mrt6debug)
551 		log(LOG_DEBUG, "ip6_mrouter_done\n");
552 #endif
553 
554 	return 0;
555 }
556 
557 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
558 
559 /*
560  * Add a mif to the mif table
561  */
562 static int
563 add_m6if(mifcp)
564 	struct mif6ctl *mifcp;
565 {
566 	struct mif6 *mifp;
567 	struct ifnet *ifp;
568 	struct in6_ifreq ifr;
569 	int error, s;
570 #ifdef notyet
571 	struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
572 #endif
573 
574 	if (mifcp->mif6c_mifi >= MAXMIFS)
575 		return EINVAL;
576 	mifp = mif6table + mifcp->mif6c_mifi;
577 	if (mifp->m6_ifp)
578 		return EADDRINUSE; /* XXX: is it appropriate? */
579 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
580 		return ENXIO;
581 	/*
582 	 * XXX: some OSes can remove ifp and clear ifindex2ifnet[id]
583 	 * even for id between 0 and if_index.
584 	 */
585 	if ((ifp = ifindex2ifnet[mifcp->mif6c_pifi]) == NULL)
586 		return ENXIO;
587 
588 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
589 		if (reg_mif_num == (mifi_t)-1) {
590 			strlcpy(multicast_register_if.if_xname,
591 			    "register_mif",
592 			    sizeof(multicast_register_if.if_xname));
593 			multicast_register_if.if_flags |= IFF_LOOPBACK;
594 			multicast_register_if.if_index = mifcp->mif6c_mifi;
595 			reg_mif_num = mifcp->mif6c_mifi;
596 		}
597 
598 		ifp = &multicast_register_if;
599 
600 	} /* if REGISTER */
601 	else {
602 		/* Make sure the interface supports multicast */
603 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
604 			return EOPNOTSUPP;
605 
606 		s = splsoftnet();
607 		/*
608 		 * Enable promiscuous reception of all IPv6 multicasts
609 		 * from the interface.
610 		 */
611 		ifr.ifr_addr.sin6_family = AF_INET6;
612 		ifr.ifr_addr.sin6_addr = in6addr_any;
613 		error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
614 		splx(s);
615 		if (error)
616 			return error;
617 	}
618 
619 	s = splsoftnet();
620 	mifp->m6_flags     = mifcp->mif6c_flags;
621 	mifp->m6_ifp       = ifp;
622 #ifdef notyet
623 	/* scaling up here allows division by 1024 in critical code */
624 	mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
625 #endif
626 	/* initialize per mif pkt counters */
627 	mifp->m6_pkt_in    = 0;
628 	mifp->m6_pkt_out   = 0;
629 	mifp->m6_bytes_in  = 0;
630 	mifp->m6_bytes_out = 0;
631 	splx(s);
632 
633 	/* Adjust nummifs up if the mifi is higher than nummifs */
634 	if (nummifs <= mifcp->mif6c_mifi)
635 		nummifs = mifcp->mif6c_mifi + 1;
636 
637 #ifdef MRT6DEBUG
638 	if (mrt6debug)
639 		log(LOG_DEBUG,
640 		    "add_mif #%d, phyint %s%d\n",
641 		    mifcp->mif6c_mifi,
642 		    ifp->if_name, ifp->if_unit);
643 #endif
644 
645 	return 0;
646 }
647 
648 /*
649  * Delete a mif from the mif table
650  */
651 static int
652 del_m6if(mifip)
653 	mifi_t *mifip;
654 {
655 	struct mif6 *mifp = mif6table + *mifip;
656 	mifi_t mifi;
657 	struct ifnet *ifp;
658 	struct in6_ifreq ifr;
659 	int s;
660 
661 	if (*mifip >= nummifs)
662 		return EINVAL;
663 	if (mifp->m6_ifp == NULL)
664 		return EINVAL;
665 
666 	s = splsoftnet();
667 
668 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
669 		/*
670 		 * XXX: what if there is yet IPv4 multicast daemon
671 		 *      using the interface?
672 		 */
673 		ifp = mifp->m6_ifp;
674 
675 		ifr.ifr_addr.sin6_family = AF_INET6;
676 		ifr.ifr_addr.sin6_addr = in6addr_any;
677 		(*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
678 	}
679 
680 #ifdef notyet
681 	bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
682 	bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
683 #endif
684 	bzero((caddr_t)mifp, sizeof (*mifp));
685 
686 	/* Adjust nummifs down */
687 	for (mifi = nummifs; mifi > 0; mifi--)
688 		if (mif6table[mifi - 1].m6_ifp)
689 			break;
690 	nummifs = mifi;
691 
692 	splx(s);
693 
694 #ifdef MRT6DEBUG
695 	if (mrt6debug)
696 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
697 #endif
698 
699 	return 0;
700 }
701 
702 /*
703  * Add an mfc entry
704  */
705 static int
706 add_m6fc(mfccp)
707 	struct mf6cctl *mfccp;
708 {
709 	struct mf6c *rt;
710 	u_long hash;
711 	struct rtdetq *rte;
712 	u_short nstl;
713 	int s;
714 
715 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
716 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
717 
718 	/* If an entry already exists, just update the fields */
719 	if (rt) {
720 #ifdef MRT6DEBUG
721 		if (mrt6debug & DEBUG_MFC)
722 			log(LOG_DEBUG,"add_m6fc update o %s g %s p %x\n",
723 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
724 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
725 			    mfccp->mf6cc_parent);
726 #endif
727 
728 		s = splsoftnet();
729 		rt->mf6c_parent = mfccp->mf6cc_parent;
730 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
731 		splx(s);
732 		return 0;
733 	}
734 
735 	/*
736 	 * Find the entry for which the upcall was made and update
737 	 */
738 	s = splsoftnet();
739 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
740 			mfccp->mf6cc_mcastgrp.sin6_addr);
741 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
742 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
743 				       &mfccp->mf6cc_origin.sin6_addr) &&
744 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
745 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
746 		    (rt->mf6c_stall != NULL)) {
747 
748 			if (nstl++)
749 				log(LOG_ERR,
750 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
751 				    "multiple kernel entries",
752 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
753 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
754 				    mfccp->mf6cc_parent, rt->mf6c_stall);
755 
756 #ifdef MRT6DEBUG
757 			if (mrt6debug & DEBUG_MFC)
758 				log(LOG_DEBUG,
759 				    "add_m6fc o %s g %s p %x dbg %x\n",
760 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
761 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
762 				    mfccp->mf6cc_parent, rt->mf6c_stall);
763 #endif
764 
765 			rt->mf6c_origin     = mfccp->mf6cc_origin;
766 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
767 			rt->mf6c_parent     = mfccp->mf6cc_parent;
768 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
769 			/* initialize pkt counters per src-grp */
770 			rt->mf6c_pkt_cnt    = 0;
771 			rt->mf6c_byte_cnt   = 0;
772 			rt->mf6c_wrong_if   = 0;
773 
774 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
775 			n6expire[hash]--;
776 
777 			/* free packets Qed at the end of this entry */
778 			for (rte = rt->mf6c_stall; rte != NULL; ) {
779 				struct rtdetq *n = rte->next;
780 				ip6_mdq(rte->m, rte->ifp, rt);
781 				m_freem(rte->m);
782 #ifdef UPCALL_TIMING
783 				collate(&(rte->t));
784 #endif /* UPCALL_TIMING */
785 				free(rte, M_MRTABLE);
786 				rte = n;
787 			}
788 			rt->mf6c_stall = NULL;
789 		}
790 	}
791 
792 	/*
793 	 * It is possible that an entry is being inserted without an upcall
794 	 */
795 	if (nstl == 0) {
796 #ifdef MRT6DEBUG
797 		if (mrt6debug & DEBUG_MFC)
798 			log(LOG_DEBUG,
799 			    "add_mfc no upcall h %d o %s g %s p %x\n",
800 			    hash,
801 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
802 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
803 			    mfccp->mf6cc_parent);
804 #endif
805 
806 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
807 
808 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
809 					       &mfccp->mf6cc_origin.sin6_addr)&&
810 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
811 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
812 
813 				rt->mf6c_origin     = mfccp->mf6cc_origin;
814 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
815 				rt->mf6c_parent     = mfccp->mf6cc_parent;
816 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
817 				/* initialize pkt counters per src-grp */
818 				rt->mf6c_pkt_cnt    = 0;
819 				rt->mf6c_byte_cnt   = 0;
820 				rt->mf6c_wrong_if   = 0;
821 
822 				if (rt->mf6c_expire)
823 					n6expire[hash]--;
824 				rt->mf6c_expire	   = 0;
825 			}
826 		}
827 		if (rt == NULL) {
828 			/* no upcall, so make a new entry */
829 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
830 						  M_NOWAIT);
831 			if (rt == NULL) {
832 				splx(s);
833 				return ENOBUFS;
834 			}
835 
836 			/* insert new entry at head of hash chain */
837 			rt->mf6c_origin     = mfccp->mf6cc_origin;
838 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
839 			rt->mf6c_parent     = mfccp->mf6cc_parent;
840 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
841 			/* initialize pkt counters per src-grp */
842 			rt->mf6c_pkt_cnt    = 0;
843 			rt->mf6c_byte_cnt   = 0;
844 			rt->mf6c_wrong_if   = 0;
845 			rt->mf6c_expire     = 0;
846 			rt->mf6c_stall = NULL;
847 
848 			/* link into table */
849 			rt->mf6c_next  = mf6ctable[hash];
850 			mf6ctable[hash] = rt;
851 		}
852 	}
853 	splx(s);
854 	return 0;
855 }
856 
857 #ifdef UPCALL_TIMING
858 /*
859  * collect delay statistics on the upcalls
860  */
861 static void
862 collate(t)
863 	struct timeval *t;
864 {
865 	u_long d;
866 	struct timeval tp;
867 	u_long delta;
868 
869 	GET_TIME(tp);
870 
871 	if (TV_LT(*t, tp))
872 	{
873 		TV_DELTA(tp, *t, delta);
874 
875 		d = delta >> 10;
876 		if (d > UPCALL_MAX)
877 			d = UPCALL_MAX;
878 
879 		++upcall_data[d];
880 	}
881 }
882 #endif /* UPCALL_TIMING */
883 
884 /*
885  * Delete an mfc entry
886  */
887 static int
888 del_m6fc(mfccp)
889 	struct mf6cctl *mfccp;
890 {
891 	struct sockaddr_in6 	origin;
892 	struct sockaddr_in6 	mcastgrp;
893 	struct mf6c 		*rt;
894 	struct mf6c	 	**nptr;
895 	u_long 		hash;
896 	int s;
897 
898 	origin = mfccp->mf6cc_origin;
899 	mcastgrp = mfccp->mf6cc_mcastgrp;
900 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
901 
902 #ifdef MRT6DEBUG
903 	if (mrt6debug & DEBUG_MFC)
904 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
905 		    ip6_sprintf(&origin.sin6_addr),
906 		    ip6_sprintf(&mcastgrp.sin6_addr));
907 #endif
908 
909 	s = splsoftnet();
910 
911 	nptr = &mf6ctable[hash];
912 	while ((rt = *nptr) != NULL) {
913 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
914 				       &rt->mf6c_origin.sin6_addr) &&
915 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
916 				       &rt->mf6c_mcastgrp.sin6_addr) &&
917 		    rt->mf6c_stall == NULL)
918 			break;
919 
920 		nptr = &rt->mf6c_next;
921 	}
922 	if (rt == NULL) {
923 		splx(s);
924 		return EADDRNOTAVAIL;
925 	}
926 
927 	*nptr = rt->mf6c_next;
928 	free(rt, M_MRTABLE);
929 
930 	splx(s);
931 
932 	return 0;
933 }
934 
935 static int
936 socket_send(s, mm, src)
937 	struct socket *s;
938 	struct mbuf *mm;
939 	struct sockaddr_in6 *src;
940 {
941 	if (s) {
942 		if (sbappendaddr(&s->so_rcv,
943 				 (struct sockaddr *)src,
944 				 mm, (struct mbuf *)0) != 0) {
945 			sorwakeup(s);
946 			return 0;
947 		}
948 	}
949 	m_freem(mm);
950 	return -1;
951 }
952 
953 /*
954  * IPv6 multicast forwarding function. This function assumes that the packet
955  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
956  * pointed to by "ifp", and the packet is to be relayed to other networks
957  * that have members of the packet's destination IPv6 multicast group.
958  *
959  * The packet is returned unscathed to the caller, unless it is
960  * erroneous, in which case a non-zero return value tells the caller to
961  * discard it.
962  */
963 
964 int
965 ip6_mforward(ip6, ifp, m)
966 	struct ip6_hdr *ip6;
967 	struct ifnet *ifp;
968 	struct mbuf *m;
969 {
970 	struct mf6c *rt;
971 	struct mif6 *mifp;
972 	struct mbuf *mm;
973 	int s;
974 	mifi_t mifi;
975 
976 #ifdef MRT6DEBUG
977 	if (mrt6debug & DEBUG_FORWARD)
978 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
979 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
980 		    ifp->if_index);
981 #endif
982 
983 	/*
984 	 * Don't forward a packet with Hop limit of zero or one,
985 	 * or a packet destined to a local-only group.
986 	 */
987 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
988 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
989 		return 0;
990 	ip6->ip6_hlim--;
991 
992 	/*
993 	 * Source address check: do not forward packets with unspecified
994 	 * source. It was discussed in July 2000, on ipngwg mailing list.
995 	 * This is rather more serious than unicast cases, because some
996 	 * MLD packets can be sent with the unspecified source address
997 	 * (although such packets must normally set 1 to the hop limit field).
998 	 */
999 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1000 		ip6stat.ip6s_cantforward++;
1001 		if (ip6_log_time + ip6_log_interval < time.tv_sec) {
1002 			ip6_log_time = time.tv_sec;
1003 			log(LOG_DEBUG,
1004 			    "cannot forward "
1005 			    "from %s to %s nxt %d received on %s\n",
1006 			    ip6_sprintf(&ip6->ip6_src),
1007 			    ip6_sprintf(&ip6->ip6_dst),
1008 			    ip6->ip6_nxt,
1009 			    m->m_pkthdr.rcvif ?
1010 			    if_name(m->m_pkthdr.rcvif) : "?");
1011 		}
1012 		return 0;
1013 	}
1014 
1015 	/*
1016 	 * Determine forwarding mifs from the forwarding cache table
1017 	 */
1018 	s = splsoftnet();
1019 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
1020 
1021 	/* Entry exists, so forward if necessary */
1022 	if (rt) {
1023 		splx(s);
1024 		return (ip6_mdq(m, ifp, rt));
1025 	} else {
1026 		/*
1027 		 * If we don't have a route for packet's origin,
1028 		 * Make a copy of the packet &
1029 		 * send message to routing daemon
1030 		 */
1031 
1032 		struct mbuf *mb0;
1033 		struct rtdetq *rte;
1034 		u_long hash;
1035 /*		int i, npkts;*/
1036 #ifdef UPCALL_TIMING
1037 		struct timeval tp;
1038 
1039 		GET_TIME(tp);
1040 #endif /* UPCALL_TIMING */
1041 
1042 		mrt6stat.mrt6s_no_route++;
1043 #ifdef MRT6DEBUG
1044 		if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
1045 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
1046 			    ip6_sprintf(&ip6->ip6_src),
1047 			    ip6_sprintf(&ip6->ip6_dst));
1048 #endif
1049 
1050 		/*
1051 		 * Allocate mbufs early so that we don't do extra work if we
1052 		 * are just going to fail anyway.
1053 		 */
1054 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1055 					      M_NOWAIT);
1056 		if (rte == NULL) {
1057 			splx(s);
1058 			return ENOBUFS;
1059 		}
1060 		mb0 = m_copy(m, 0, M_COPYALL);
1061 		/*
1062 		 * Pullup packet header if needed before storing it,
1063 		 * as other references may modify it in the meantime.
1064 		 */
1065 		if (mb0 &&
1066 		    (M_READONLY(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1067 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1068 		if (mb0 == NULL) {
1069 			free(rte, M_MRTABLE);
1070 			splx(s);
1071 			return ENOBUFS;
1072 		}
1073 
1074 		/* is there an upcall waiting for this packet? */
1075 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1076 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1077 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1078 					       &rt->mf6c_origin.sin6_addr) &&
1079 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1080 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1081 			    (rt->mf6c_stall != NULL))
1082 				break;
1083 		}
1084 
1085 		if (rt == NULL) {
1086 			struct mrt6msg *im;
1087 			struct omrt6msg *oim;
1088 
1089 			/* no upcall, so make a new entry */
1090 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1091 						  M_NOWAIT);
1092 			if (rt == NULL) {
1093 				free(rte, M_MRTABLE);
1094 				m_freem(mb0);
1095 				splx(s);
1096 				return ENOBUFS;
1097 			}
1098 			/*
1099 			 * Make a copy of the header to send to the user
1100 			 * level process
1101 			 */
1102 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1103 
1104 			if (mm == NULL) {
1105 				free(rte, M_MRTABLE);
1106 				m_freem(mb0);
1107 				free(rt, M_MRTABLE);
1108 				splx(s);
1109 				return ENOBUFS;
1110 			}
1111 
1112 			/*
1113 			 * Send message to routing daemon
1114 			 */
1115 			sin6.sin6_addr = ip6->ip6_src;
1116 
1117 			im = NULL;
1118 			oim = NULL;
1119 			switch (ip6_mrouter_ver) {
1120 			case MRT6_OINIT:
1121 				oim = mtod(mm, struct omrt6msg *);
1122 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1123 				oim->im6_mbz = 0;
1124 				break;
1125 			case MRT6_INIT:
1126 				im = mtod(mm, struct mrt6msg *);
1127 				im->im6_msgtype = MRT6MSG_NOCACHE;
1128 				im->im6_mbz = 0;
1129 				break;
1130 			default:
1131 				free(rte, M_MRTABLE);
1132 				m_freem(mb0);
1133 				free(rt, M_MRTABLE);
1134 				splx(s);
1135 				return EINVAL;
1136 			}
1137 
1138 #ifdef MRT6DEBUG
1139 			if (mrt6debug & DEBUG_FORWARD)
1140 				log(LOG_DEBUG,
1141 				    "getting the iif info in the kernel\n");
1142 #endif
1143 
1144 			for (mifp = mif6table, mifi = 0;
1145 			     mifi < nummifs && mifp->m6_ifp != ifp;
1146 			     mifp++, mifi++)
1147 				;
1148 
1149 			switch (ip6_mrouter_ver) {
1150 			case MRT6_OINIT:
1151 				oim->im6_mif = mifi;
1152 				break;
1153 			case MRT6_INIT:
1154 				im->im6_mif = mifi;
1155 				break;
1156 			}
1157 
1158 			if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1159 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1160 				    "socket queue full\n");
1161 				mrt6stat.mrt6s_upq_sockfull++;
1162 				free(rte, M_MRTABLE);
1163 				m_freem(mb0);
1164 				free(rt, M_MRTABLE);
1165 				splx(s);
1166 				return ENOBUFS;
1167 			}
1168 
1169 			mrt6stat.mrt6s_upcalls++;
1170 
1171 			/* insert new entry at head of hash chain */
1172 			bzero(rt, sizeof(*rt));
1173 			rt->mf6c_origin.sin6_family = AF_INET6;
1174 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1175 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1176 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1177 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1178 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1179 			rt->mf6c_expire = UPCALL_EXPIRE;
1180 			n6expire[hash]++;
1181 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1182 
1183 			/* link into table */
1184 			rt->mf6c_next  = mf6ctable[hash];
1185 			mf6ctable[hash] = rt;
1186 			/* Add this entry to the end of the queue */
1187 			rt->mf6c_stall = rte;
1188 		} else {
1189 			/* determine if q has overflowed */
1190 			struct rtdetq **p;
1191 			int npkts = 0;
1192 
1193 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1194 				if (++npkts > MAX_UPQ6) {
1195 					mrt6stat.mrt6s_upq_ovflw++;
1196 					free(rte, M_MRTABLE);
1197 					m_freem(mb0);
1198 					splx(s);
1199 					return 0;
1200 				}
1201 
1202 			/* Add this entry to the end of the queue */
1203 			*p = rte;
1204 		}
1205 
1206 		rte->next = NULL;
1207 		rte->m = mb0;
1208 		rte->ifp = ifp;
1209 #ifdef UPCALL_TIMING
1210 		rte->t = tp;
1211 #endif /* UPCALL_TIMING */
1212 
1213 		splx(s);
1214 
1215 		return 0;
1216 	}
1217 }
1218 
1219 /*
1220  * Clean up cache entries if upcalls are not serviced
1221  * Call from the Slow Timeout mechanism, every half second.
1222  */
1223 static void
1224 expire_upcalls(unused)
1225 	void *unused;
1226 {
1227 	struct rtdetq *rte;
1228 	struct mf6c *mfc, **nptr;
1229 	int i;
1230 	int s;
1231 
1232 	s = splsoftnet();
1233 	for (i = 0; i < MF6CTBLSIZ; i++) {
1234 		if (n6expire[i] == 0)
1235 			continue;
1236 		nptr = &mf6ctable[i];
1237 		while ((mfc = *nptr) != NULL) {
1238 			rte = mfc->mf6c_stall;
1239 			/*
1240 			 * Skip real cache entries
1241 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1242 			 * If it expires now
1243 			 */
1244 			if (rte != NULL &&
1245 			    mfc->mf6c_expire != 0 &&
1246 			    --mfc->mf6c_expire == 0) {
1247 #ifdef MRT6DEBUG
1248 				if (mrt6debug & DEBUG_EXPIRE)
1249 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1250 					    ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1251 					    ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1252 #endif
1253 				/*
1254 				 * drop all the packets
1255 				 * free the mbuf with the pkt, if, timing info
1256 				 */
1257 				do {
1258 					struct rtdetq *n = rte->next;
1259 					m_freem(rte->m);
1260 					free(rte, M_MRTABLE);
1261 					rte = n;
1262 				} while (rte != NULL);
1263 				mrt6stat.mrt6s_cache_cleanups++;
1264 				n6expire[i]--;
1265 
1266 				*nptr = mfc->mf6c_next;
1267 				free(mfc, M_MRTABLE);
1268 			} else {
1269 				nptr = &mfc->mf6c_next;
1270 			}
1271 		}
1272 	}
1273 	splx(s);
1274 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1275 	    expire_upcalls, NULL);
1276 }
1277 
1278 /*
1279  * Packet forwarding routine once entry in the cache is made
1280  */
1281 static int
1282 ip6_mdq(m, ifp, rt)
1283 	struct mbuf *m;
1284 	struct ifnet *ifp;
1285 	struct mf6c *rt;
1286 {
1287 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1288 	mifi_t mifi, iif;
1289 	struct mif6 *mifp;
1290 	int plen = m->m_pkthdr.len;
1291 
1292 /*
1293  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1294  * input, they shouldn't get counted on output, so statistics keeping is
1295  * separate.
1296  */
1297 
1298 #define MC6_SEND(ip6, mifp, m) do {				\
1299 	if ((mifp)->m6_flags & MIFF_REGISTER)			\
1300 		register_send((ip6), (mifp), (m));		\
1301 	else							\
1302 		phyint_send((ip6), (mifp), (m));		\
1303 } while (/*CONSTCOND*/ 0)
1304 
1305 	/*
1306 	 * Don't forward if it didn't arrive from the parent mif
1307 	 * for its origin.
1308 	 */
1309 	mifi = rt->mf6c_parent;
1310 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1311 		/* came in the wrong interface */
1312 #ifdef MRT6DEBUG
1313 		if (mrt6debug & DEBUG_FORWARD)
1314 			log(LOG_DEBUG,
1315 			    "wrong if: ifid %d mifi %d mififid %x\n",
1316 			    ifp->if_index, mifi,
1317 			    mif6table[mifi].m6_ifp->if_index);
1318 #endif
1319 		mrt6stat.mrt6s_wrong_if++;
1320 		rt->mf6c_wrong_if++;
1321 		/*
1322 		 * If we are doing PIM processing, and we are forwarding
1323 		 * packets on this interface, send a message to the
1324 		 * routing daemon.
1325 		 */
1326 		/* have to make sure this is a valid mif */
1327 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1328 			if (pim6 && (m->m_flags & M_LOOP) == 0) {
1329 				/*
1330 				 * Check the M_LOOP flag to avoid an
1331 				 * unnecessary PIM assert.
1332 				 * XXX: M_LOOP is an ad-hoc hack...
1333 				 */
1334 				static struct sockaddr_in6 sin6 =
1335 				{ sizeof(sin6), AF_INET6 };
1336 
1337 				struct mbuf *mm;
1338 				struct mrt6msg *im;
1339 				struct omrt6msg *oim;
1340 
1341 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1342 				if (mm &&
1343 				    (M_READONLY(mm) ||
1344 				     mm->m_len < sizeof(struct ip6_hdr)))
1345 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1346 				if (mm == NULL)
1347 					return ENOBUFS;
1348 
1349 				oim = NULL;
1350 				im = NULL;
1351 				switch (ip6_mrouter_ver) {
1352 				case MRT6_OINIT:
1353 					oim = mtod(mm, struct omrt6msg *);
1354 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1355 					oim->im6_mbz = 0;
1356 					break;
1357 				case MRT6_INIT:
1358 					im = mtod(mm, struct mrt6msg *);
1359 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1360 					im->im6_mbz = 0;
1361 					break;
1362 				default:
1363 					m_freem(mm);
1364 					return EINVAL;
1365 				}
1366 
1367 				for (mifp = mif6table, iif = 0;
1368 				     iif < nummifs && mifp &&
1369 					     mifp->m6_ifp != ifp;
1370 				     mifp++, iif++)
1371 					;
1372 
1373 				switch (ip6_mrouter_ver) {
1374 				case MRT6_OINIT:
1375 					oim->im6_mif = iif;
1376 					sin6.sin6_addr = oim->im6_src;
1377 					break;
1378 				case MRT6_INIT:
1379 					im->im6_mif = iif;
1380 					sin6.sin6_addr = im->im6_src;
1381 					break;
1382 				}
1383 
1384 				mrt6stat.mrt6s_upcalls++;
1385 
1386 				if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1387 #ifdef MRT6DEBUG
1388 					if (mrt6debug)
1389 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1390 #endif
1391 					++mrt6stat.mrt6s_upq_sockfull;
1392 					return ENOBUFS;
1393 				}	/* if socket Q full */
1394 			}		/* if PIM */
1395 		return 0;
1396 	}			/* if wrong iif */
1397 
1398 	/* If I sourced this packet, it counts as output, else it was input. */
1399 	if (m->m_pkthdr.rcvif == NULL) {
1400 		/* XXX: is rcvif really NULL when output?? */
1401 		mif6table[mifi].m6_pkt_out++;
1402 		mif6table[mifi].m6_bytes_out += plen;
1403 	} else {
1404 		mif6table[mifi].m6_pkt_in++;
1405 		mif6table[mifi].m6_bytes_in += plen;
1406 	}
1407 	rt->mf6c_pkt_cnt++;
1408 	rt->mf6c_byte_cnt += plen;
1409 
1410 	/*
1411 	 * For each mif, forward a copy of the packet if there are group
1412 	 * members downstream on the interface.
1413 	 */
1414 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1415 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1416 			if (mif6table[mifi].m6_ifp == NULL)
1417 				continue;
1418 #ifdef notyet
1419 			/*
1420 			 * check if the outgoing packet is going to break
1421 			 * a scope boundary.
1422 			 * XXX For packets through PIM register tunnel
1423 			 * interface, we believe a routing daemon.
1424 			 */
1425 			if ((mif6table[rt->mf6c_parent].m6_flags &
1426 			     MIFF_REGISTER) == 0 &&
1427 			    (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1428 			    (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1429 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1430 					      &ip6->ip6_dst) ||
1431 			     in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1432 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1433 					      &ip6->ip6_src))) {
1434 				ip6stat.ip6s_badscope++;
1435 				continue;
1436 			}
1437 #endif
1438 
1439 			mifp->m6_pkt_out++;
1440 			mifp->m6_bytes_out += plen;
1441 			MC6_SEND(ip6, mifp, m);
1442 		}
1443 	return 0;
1444 }
1445 
1446 static void
1447 phyint_send(ip6, mifp, m)
1448 	struct ip6_hdr *ip6;
1449 	struct mif6 *mifp;
1450 	struct mbuf *m;
1451 {
1452 	struct mbuf *mb_copy;
1453 	struct ifnet *ifp = mifp->m6_ifp;
1454 	int error = 0;
1455 	int s = splsoftnet();
1456 	static struct route_in6 ro;
1457 	struct	in6_multi *in6m;
1458 	struct sockaddr_in6 *dst6;
1459 	u_long linkmtu;
1460 
1461 	/*
1462 	 * Make a new reference to the packet; make sure that
1463 	 * the IPv6 header is actually copied, not just referenced,
1464 	 * so that ip6_output() only scribbles on the copy.
1465 	 */
1466 	mb_copy = m_copy(m, 0, M_COPYALL);
1467 	if (mb_copy &&
1468 	    (M_READONLY(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1469 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1470 	if (mb_copy == NULL) {
1471 		splx(s);
1472 		return;
1473 	}
1474 	/* set MCAST flag to the outgoing packet */
1475 	mb_copy->m_flags |= M_MCAST;
1476 
1477 	/*
1478 	 * If we sourced the packet, call ip6_output since we may devide
1479 	 * the packet into fragments when the packet is too big for the
1480 	 * outgoing interface.
1481 	 * Otherwise, we can simply send the packet to the interface
1482 	 * sending queue.
1483 	 */
1484 	if (m->m_pkthdr.rcvif == NULL) {
1485 		struct ip6_moptions im6o;
1486 
1487 		im6o.im6o_multicast_ifp = ifp;
1488 		/* XXX: ip6_output will override ip6->ip6_hlim */
1489 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1490 		im6o.im6o_multicast_loop = 1;
1491 		error = ip6_output(mb_copy, NULL, &ro,
1492 				   IPV6_FORWARDING, &im6o, NULL);
1493 
1494 #ifdef MRT6DEBUG
1495 		if (mrt6debug & DEBUG_XMIT)
1496 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1497 			    mifp - mif6table, error);
1498 #endif
1499 		splx(s);
1500 		return;
1501 	}
1502 
1503 	/*
1504 	 * If we belong to the destination multicast group
1505 	 * on the outgoing interface, loop back a copy.
1506 	 */
1507 	dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
1508 	IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1509 	if (in6m != NULL) {
1510 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1511 		dst6->sin6_family = AF_INET6;
1512 		dst6->sin6_addr = ip6->ip6_dst;
1513 		ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst);
1514 	}
1515 	/*
1516 	 * Put the packet into the sending queue of the outgoing interface
1517 	 * if it would fit in the MTU of the interface.
1518 	 */
1519 	linkmtu = IN6_LINKMTU(ifp);
1520 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1521 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1522 		dst6->sin6_family = AF_INET6;
1523 		dst6->sin6_addr = ip6->ip6_dst;
1524 		/*
1525 		 * We just call if_output instead of nd6_output here, since
1526 		 * we need no ND for a multicast forwarded packet...right?
1527 		 */
1528 		error = (*ifp->if_output)(ifp, mb_copy,
1529 		    (struct sockaddr *)&ro.ro_dst, NULL);
1530 #ifdef MRT6DEBUG
1531 		if (mrt6debug & DEBUG_XMIT)
1532 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1533 			    mifp - mif6table, error);
1534 #endif
1535 	} else {
1536 #ifdef MULTICAST_PMTUD
1537 		icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1538 #else
1539 #ifdef MRT6DEBUG
1540 		if (mrt6debug & DEBUG_XMIT)
1541 			log(LOG_DEBUG,
1542 			    "phyint_send: packet too big on %s o %s g %s"
1543 			    " size %d(discarded)\n",
1544 			    if_name(ifp),
1545 			    ip6_sprintf(&ip6->ip6_src),
1546 			    ip6_sprintf(&ip6->ip6_dst),
1547 			    mb_copy->m_pkthdr.len);
1548 #endif /* MRT6DEBUG */
1549 		m_freem(mb_copy); /* simply discard the packet */
1550 #endif
1551 	}
1552 
1553 	splx(s);
1554 }
1555 
1556 static int
1557 register_send(ip6, mif, m)
1558 	struct ip6_hdr *ip6;
1559 	struct mif6 *mif;
1560 	struct mbuf *m;
1561 {
1562 	struct mbuf *mm;
1563 	int i, len = m->m_pkthdr.len;
1564 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1565 	struct mrt6msg *im6;
1566 
1567 #ifdef MRT6DEBUG
1568 	if (mrt6debug)
1569 		log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1570 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1571 #endif
1572 	++pim6stat.pim6s_snd_registers;
1573 
1574 	/* Make a copy of the packet to send to the user level process */
1575 	MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1576 	if (mm == NULL)
1577 		return ENOBUFS;
1578 	mm->m_data += max_linkhdr;
1579 	mm->m_len = sizeof(struct ip6_hdr);
1580 
1581 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1582 		m_freem(mm);
1583 		return ENOBUFS;
1584 	}
1585 	i = MHLEN - M_LEADINGSPACE(mm);
1586 	if (i > len)
1587 		i = len;
1588 	mm = m_pullup(mm, i);
1589 	if (mm == NULL)
1590 		return ENOBUFS;
1591 /* TODO: check it! */
1592 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1593 
1594 	/*
1595 	 * Send message to routing daemon
1596 	 */
1597 	sin6.sin6_addr = ip6->ip6_src;
1598 
1599 	im6 = mtod(mm, struct mrt6msg *);
1600 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1601 	im6->im6_mbz          = 0;
1602 
1603 	im6->im6_mif = mif - mif6table;
1604 
1605 	/* iif info is not given for reg. encap.n */
1606 	mrt6stat.mrt6s_upcalls++;
1607 
1608 	if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1609 #ifdef MRT6DEBUG
1610 		if (mrt6debug)
1611 			log(LOG_WARNING,
1612 			    "register_send: ip6_mrouter socket queue full\n");
1613 #endif
1614 		++mrt6stat.mrt6s_upq_sockfull;
1615 		return ENOBUFS;
1616 	}
1617 	return 0;
1618 }
1619 
1620 /*
1621  * PIM sparse mode hook
1622  * Receives the pim control messages, and passes them up to the listening
1623  * socket, using rip6_input.
1624  * The only message processed is the REGISTER pim message; the pim header
1625  * is stripped off, and the inner packet is passed to register_mforward.
1626  */
1627 int
1628 pim6_input(mp, offp, proto)
1629 	struct mbuf **mp;
1630 	int *offp, proto;
1631 {
1632 	struct pim *pim; /* pointer to a pim struct */
1633 	struct ip6_hdr *ip6;
1634 	int pimlen;
1635 	struct mbuf *m = *mp;
1636 	int minlen;
1637 	int off = *offp;
1638 
1639 	++pim6stat.pim6s_rcv_total;
1640 
1641 	ip6 = mtod(m, struct ip6_hdr *);
1642 	pimlen = m->m_pkthdr.len - *offp;
1643 
1644 	/*
1645 	 * Validate lengths
1646 	 */
1647 	if (pimlen < PIM_MINLEN) {
1648 		++pim6stat.pim6s_rcv_tooshort;
1649 #ifdef MRT6DEBUG
1650 		if (mrt6debug & DEBUG_PIM)
1651 			log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1652 #endif
1653 		m_freem(m);
1654 		return (IPPROTO_DONE);
1655 	}
1656 
1657 	/*
1658 	 * if the packet is at least as big as a REGISTER, go ahead
1659 	 * and grab the PIM REGISTER header size, to avoid another
1660 	 * possible m_pullup() later.
1661 	 *
1662 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1663 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1664 	 */
1665 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1666 
1667 	/*
1668 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1669 	 * possibly the PIM REGISTER header
1670 	 */
1671 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1672 	if (pim == NULL) {
1673 		pim6stat.pim6s_rcv_tooshort++;
1674 		return IPPROTO_DONE;
1675 	}
1676 
1677 	/* PIM version check */
1678 	if (pim->pim_ver != PIM_VERSION) {
1679 		++pim6stat.pim6s_rcv_badversion;
1680 #ifdef MRT6DEBUG
1681 		log(LOG_ERR,
1682 		    "pim6_input: incorrect version %d, expecting %d\n",
1683 		    pim->pim_ver, PIM_VERSION);
1684 #endif
1685 		m_freem(m);
1686 		return (IPPROTO_DONE);
1687 	}
1688 
1689 #define PIM6_CHECKSUM
1690 #ifdef PIM6_CHECKSUM
1691 	{
1692 		int cksumlen;
1693 
1694 		/*
1695 		 * Validate checksum.
1696 		 * If PIM REGISTER, exclude the data packet
1697 		 */
1698 		if (pim->pim_type == PIM_REGISTER)
1699 			cksumlen = PIM_MINLEN;
1700 		else
1701 			cksumlen = pimlen;
1702 
1703 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1704 			++pim6stat.pim6s_rcv_badsum;
1705 #ifdef MRT6DEBUG
1706 			if (mrt6debug & DEBUG_PIM)
1707 				log(LOG_DEBUG,
1708 				    "pim6_input: invalid checksum\n");
1709 #endif
1710 			m_freem(m);
1711 			return (IPPROTO_DONE);
1712 		}
1713 	}
1714 #endif /* PIM_CHECKSUM */
1715 
1716 	if (pim->pim_type == PIM_REGISTER) {
1717 		/*
1718 		 * since this is a REGISTER, we'll make a copy of the register
1719 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1720 		 * routing daemon.
1721 		 */
1722 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1723 
1724 		struct mbuf *mcp;
1725 		struct ip6_hdr *eip6;
1726 		u_int32_t *reghdr;
1727 		int rc;
1728 
1729 		++pim6stat.pim6s_rcv_registers;
1730 
1731 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1732 #ifdef MRT6DEBUG
1733 			if (mrt6debug & DEBUG_PIM)
1734 				log(LOG_DEBUG,
1735 				    "pim6_input: register mif not set: %d\n",
1736 				    reg_mif_num);
1737 #endif
1738 			m_freem(m);
1739 			return (IPPROTO_DONE);
1740 		}
1741 
1742 		reghdr = (u_int32_t *)(pim + 1);
1743 
1744 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1745 			goto pim6_input_to_daemon;
1746 
1747 		/*
1748 		 * Validate length
1749 		 */
1750 		if (pimlen < PIM6_REG_MINLEN) {
1751 			++pim6stat.pim6s_rcv_tooshort;
1752 			++pim6stat.pim6s_rcv_badregisters;
1753 #ifdef MRT6DEBUG
1754 			log(LOG_ERR,
1755 			    "pim6_input: register packet size too "
1756 			    "small %d from %s\n",
1757 			    pimlen, ip6_sprintf(&ip6->ip6_src));
1758 #endif
1759 			m_freem(m);
1760 			return (IPPROTO_DONE);
1761 		}
1762 
1763 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1764 #ifdef MRT6DEBUG
1765 		if (mrt6debug & DEBUG_PIM)
1766 			log(LOG_DEBUG,
1767 			    "pim6_input[register], eip6: %s -> %s, "
1768 			    "eip6 plen %d\n",
1769 			    ip6_sprintf(&eip6->ip6_src),
1770 			    ip6_sprintf(&eip6->ip6_dst),
1771 			    ntohs(eip6->ip6_plen));
1772 #endif
1773 
1774 		/* verify the version number of the inner packet */
1775 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1776 			++pim6stat.pim6s_rcv_badregisters;
1777 #ifdef MRT6DEBUG
1778 			log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1779 			    "of the inner packet\n",
1780 			    (eip6->ip6_vfc & IPV6_VERSION));
1781 #endif
1782 			m_freem(m);
1783 			return (IPPROTO_NONE);
1784 		}
1785 
1786 		/* verify the inner packet is destined to a mcast group */
1787 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1788 			++pim6stat.pim6s_rcv_badregisters;
1789 #ifdef MRT6DEBUG
1790 			if (mrt6debug & DEBUG_PIM)
1791 				log(LOG_DEBUG,
1792 				    "pim6_input: inner packet of register "
1793 				    "is not multicast %s\n",
1794 				    ip6_sprintf(&eip6->ip6_dst));
1795 #endif
1796 			m_freem(m);
1797 			return (IPPROTO_DONE);
1798 		}
1799 
1800 		/*
1801 		 * make a copy of the whole header to pass to the daemon later.
1802 		 */
1803 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1804 		if (mcp == NULL) {
1805 #ifdef MRT6DEBUG
1806 			log(LOG_ERR,
1807 			    "pim6_input: pim register: "
1808 			    "could not copy register head\n");
1809 #endif
1810 			m_freem(m);
1811 			return (IPPROTO_DONE);
1812 		}
1813 
1814 		/*
1815 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1816 		 */
1817 		m_adj(m, off + PIM_MINLEN);
1818 #ifdef MRT6DEBUG
1819 		if (mrt6debug & DEBUG_PIM) {
1820 			log(LOG_DEBUG,
1821 			    "pim6_input: forwarding decapsulated register: "
1822 			    "src %s, dst %s, mif %d\n",
1823 			    ip6_sprintf(&eip6->ip6_src),
1824 			    ip6_sprintf(&eip6->ip6_dst),
1825 			    reg_mif_num);
1826 		}
1827 #endif
1828 
1829 		rc = looutput(mif6table[reg_mif_num].m6_ifp, m,
1830 			      (struct sockaddr *) &dst,
1831 			      (struct rtentry *) NULL);
1832 
1833 		/* prepare the register head to send to the mrouting daemon */
1834 		m = mcp;
1835 	}
1836 
1837 	/*
1838 	 * Pass the PIM message up to the daemon; if it is a register message
1839 	 * pass the 'head' only up to the daemon. This includes the
1840 	 * encapsulator ip6 header, pim header, register header and the
1841 	 * encapsulated ip6 header.
1842 	 */
1843   pim6_input_to_daemon:
1844 	rip6_input(&m, offp, proto);
1845 	return (IPPROTO_DONE);
1846 }
1847