xref: /freebsd/sys/netinet6/ip6_mroute.c (revision 1b6c76a2fe091c74f08427e6c870851025a9cf67)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_mroute.c,v 1.46 2001/04/04 05:17:30 itojun 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  * IP multicast forwarding procedures
37  *
38  * Written by David Waitzman, BBN Labs, August 1988.
39  * Modified by Steve Deering, Stanford, February 1989.
40  * Modified by Mark J. Steiglitz, Stanford, May, 1991
41  * Modified by Van Jacobson, LBL, January 1993
42  * Modified by Ajit Thyagarajan, PARC, August 1993
43  * Modified by Bill Fenenr, PARC, April 1994
44  *
45  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
46  */
47 
48 #include "opt_inet.h"
49 #include "opt_inet6.h"
50 
51 #include <sys/param.h>
52 #include <sys/systm.h>
53 #include <sys/callout.h>
54 #include <sys/malloc.h>
55 #include <sys/mbuf.h>
56 #include <sys/socket.h>
57 #include <sys/socketvar.h>
58 #include <sys/sockio.h>
59 #include <sys/protosw.h>
60 #include <sys/errno.h>
61 #include <sys/time.h>
62 #include <sys/kernel.h>
63 #include <sys/syslog.h>
64 
65 #include <net/if.h>
66 #include <net/route.h>
67 #include <net/raw_cb.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_var.h>
71 
72 #include <netinet/ip6.h>
73 #include <netinet6/ip6_var.h>
74 #include <netinet6/ip6_mroute.h>
75 #include <netinet6/pim6.h>
76 #include <netinet6/pim6_var.h>
77 
78 #include <net/net_osdep.h>
79 
80 static MALLOC_DEFINE(M_MRTABLE, "mf6c", "multicast forwarding cache entry");
81 
82 #define M_HASCL(m) ((m)->m_flags & M_EXT)
83 
84 static int ip6_mdq __P((struct mbuf *, struct ifnet *, struct mf6c *));
85 static void phyint_send __P((struct ip6_hdr *, struct mif6 *, struct mbuf *));
86 
87 static int set_pim6 __P((int *));
88 static int socket_send __P((struct socket *, struct mbuf *,
89 			    struct sockaddr_in6 *));
90 static int register_send __P((struct ip6_hdr *, struct mif6 *,
91 			      struct mbuf *));
92 
93 /*
94  * Globals.  All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static,
95  * except for netstat or debugging purposes.
96  */
97 struct socket  *ip6_mrouter  = NULL;
98 int		ip6_mrouter_ver = 0;
99 int		ip6_mrtproto = IPPROTO_PIM;    /* for netstat only */
100 struct mrt6stat	mrt6stat;
101 
102 #define NO_RTE_FOUND 	0x1
103 #define RTE_FOUND	0x2
104 
105 struct mf6c	*mf6ctable[MF6CTBLSIZ];
106 u_char		nexpire[MF6CTBLSIZ];
107 static struct mif6 mif6table[MAXMIFS];
108 #ifdef MRT6DEBUG
109 u_int		mrt6debug = 0;	  /* debug level 	*/
110 #define		DEBUG_MFC	0x02
111 #define		DEBUG_FORWARD	0x04
112 #define		DEBUG_EXPIRE	0x08
113 #define		DEBUG_XMIT	0x10
114 #define         DEBUG_REG       0x20
115 #define         DEBUG_PIM       0x40
116 #endif
117 
118 static void	expire_upcalls __P((void *));
119 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
120 #define		UPCALL_EXPIRE	6		/* number of timeouts */
121 
122 #ifdef INET
123 #ifdef MROUTING
124 extern struct socket *ip_mrouter;
125 #endif
126 #endif
127 
128 /*
129  * 'Interfaces' associated with decapsulator (so we can tell
130  * packets that went through it from ones that get reflected
131  * by a broken gateway).  These interfaces are never linked into
132  * the system ifnet list & no routes point to them.  I.e., packets
133  * can't be sent this way.  They only exist as a placeholder for
134  * multicast source verification.
135  */
136 struct ifnet multicast_register_if;
137 
138 #define ENCAP_HOPS 64
139 
140 /*
141  * Private variables.
142  */
143 static mifi_t nummifs = 0;
144 static mifi_t reg_mif_num = (mifi_t)-1;
145 
146 static struct pim6stat pim6stat;
147 
148 /*
149  * one-back cache used by ipip_input to locate a tunnel's mif
150  * given a datagram's src ip address.
151  */
152 static int pim6;
153 
154 /*
155  * Hash function for a source, group entry
156  */
157 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
158 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
159 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
160 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
161 
162 /*
163  * Find a route for a given origin IPv6 address and Multicast group address.
164  * Quality of service parameter to be added in the future!!!
165  */
166 
167 #define MF6CFIND(o, g, rt) do { \
168 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
169 	rt = NULL; \
170 	mrt6stat.mrt6s_mfc_lookups++; \
171 	while (_rt) { \
172 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
173 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
174 		    (_rt->mf6c_stall == NULL)) { \
175 			rt = _rt; \
176 			break; \
177 		} \
178 		_rt = _rt->mf6c_next; \
179 	} \
180 	if (rt == NULL) { \
181 		mrt6stat.mrt6s_mfc_misses++; \
182 	} \
183 } while (0)
184 
185 /*
186  * Macros to compute elapsed time efficiently
187  * Borrowed from Van Jacobson's scheduling code
188  */
189 #define TV_DELTA(a, b, delta) do { \
190 	    int xxs; \
191 		\
192 	    delta = (a).tv_usec - (b).tv_usec; \
193 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
194 	       switch (xxs) { \
195 		      case 2: \
196 			  delta += 1000000; \
197 			      /* fall through */ \
198 		      case 1: \
199 			  delta += 1000000; \
200 			  break; \
201 		      default: \
202 			  delta += (1000000 * xxs); \
203 	       } \
204 	    } \
205 } while (0)
206 
207 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
208 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
209 
210 #ifdef UPCALL_TIMING
211 #define UPCALL_MAX	50
212 u_long upcall_data[UPCALL_MAX + 1];
213 static void collate();
214 #endif /* UPCALL_TIMING */
215 
216 static int get_sg_cnt __P((struct sioc_sg_req6 *));
217 static int get_mif6_cnt __P((struct sioc_mif_req6 *));
218 static int ip6_mrouter_init __P((struct socket *, struct mbuf *, int));
219 static int add_m6if __P((struct mif6ctl *));
220 static int del_m6if __P((mifi_t *));
221 static int add_m6fc __P((struct mf6cctl *));
222 static int del_m6fc __P((struct mf6cctl *));
223 
224 static struct callout expire_upcalls_ch;
225 
226 /*
227  * Handle MRT setsockopt commands to modify the multicast routing tables.
228  */
229 int
230 ip6_mrouter_set(so, sopt)
231 	struct socket *so;
232 	struct sockopt *sopt;
233 {
234 	int	error = 0;
235 	struct mbuf *m;
236 
237 	if (so != ip6_mrouter && sopt->sopt_name != MRT6_INIT)
238 		return (EACCES);
239 
240 	if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
241 		return (error);
242 	if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
243 		return (error);
244 
245 	switch (sopt->sopt_name) {
246 	case MRT6_INIT:
247 #ifdef MRT6_OINIT
248 	case MRT6_OINIT:
249 #endif
250 		error = ip6_mrouter_init(so, m, sopt->sopt_name);
251 		break;
252 	case MRT6_DONE:
253 		error = ip6_mrouter_done();
254 		break;
255 	case MRT6_ADD_MIF:
256 		error = add_m6if(mtod(m, struct mif6ctl *));
257 		break;
258 	case MRT6_DEL_MIF:
259 		error = del_m6if(mtod(m, mifi_t *));
260 		break;
261 	case MRT6_ADD_MFC:
262 		error = add_m6fc(mtod(m, struct mf6cctl *));
263 		break;
264 	case MRT6_DEL_MFC:
265 		error = del_m6fc(mtod(m, struct mf6cctl *));
266 		break;
267 	case MRT6_PIM:
268 		error = set_pim6(mtod(m, int *));
269 		break;
270 	default:
271 		error = EOPNOTSUPP;
272 		break;
273 	}
274 
275 	(void)m_freem(m);
276 	return(error);
277 }
278 
279 /*
280  * Handle MRT getsockopt commands
281  */
282 int
283 ip6_mrouter_get(so, sopt)
284 	struct socket *so;
285 	struct sockopt *sopt;
286 {
287 	int error = 0;
288 
289 	if (so != ip6_mrouter) return EACCES;
290 
291 	switch (sopt->sopt_name) {
292 		case MRT6_PIM:
293 			error = sooptcopyout(sopt, &pim6, sizeof(pim6));
294 			break;
295 	}
296 	return (error);
297 }
298 
299 /*
300  * Handle ioctl commands to obtain information from the cache
301  */
302 int
303 mrt6_ioctl(cmd, data)
304 	int cmd;
305 	caddr_t data;
306 {
307 	int error = 0;
308 
309 	switch (cmd) {
310 	case SIOCGETSGCNT_IN6:
311 		return(get_sg_cnt((struct sioc_sg_req6 *)data));
312 		break;		/* for safety */
313 	case SIOCGETMIFCNT_IN6:
314 		return(get_mif6_cnt((struct sioc_mif_req6 *)data));
315 		break;		/* for safety */
316 	default:
317 		return (EINVAL);
318 		break;
319 	}
320 	return error;
321 }
322 
323 /*
324  * returns the packet, byte, rpf-failure count for the source group provided
325  */
326 static int
327 get_sg_cnt(req)
328 	struct sioc_sg_req6 *req;
329 {
330 	struct mf6c *rt;
331 	int s;
332 
333 	s = splnet();
334 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
335 	splx(s);
336 	if (rt != NULL) {
337 		req->pktcnt = rt->mf6c_pkt_cnt;
338 		req->bytecnt = rt->mf6c_byte_cnt;
339 		req->wrong_if = rt->mf6c_wrong_if;
340 	} else
341 		return(ESRCH);
342 #if 0
343 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
344 #endif
345 
346 	return 0;
347 }
348 
349 /*
350  * returns the input and output packet and byte counts on the mif provided
351  */
352 static int
353 get_mif6_cnt(req)
354 	struct sioc_mif_req6 *req;
355 {
356 	mifi_t mifi = req->mifi;
357 
358 	if (mifi >= nummifs)
359 		return EINVAL;
360 
361 	req->icount = mif6table[mifi].m6_pkt_in;
362 	req->ocount = mif6table[mifi].m6_pkt_out;
363 	req->ibytes = mif6table[mifi].m6_bytes_in;
364 	req->obytes = mif6table[mifi].m6_bytes_out;
365 
366 	return 0;
367 }
368 
369 static int
370 set_pim6(i)
371 	int *i;
372 {
373 	if ((*i != 1) && (*i != 0))
374 		return EINVAL;
375 
376 	pim6 = *i;
377 
378 	return 0;
379 }
380 
381 /*
382  * Enable multicast routing
383  */
384 static int
385 ip6_mrouter_init(so, m, cmd)
386 	struct socket *so;
387 	struct mbuf *m;
388 	int cmd;
389 {
390 	int *v;
391 
392 #ifdef MRT6DEBUG
393 	if (mrt6debug)
394 		log(LOG_DEBUG,
395 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
396 		    so->so_type, so->so_proto->pr_protocol);
397 #endif
398 
399 	if (so->so_type != SOCK_RAW ||
400 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
401 		return EOPNOTSUPP;
402 
403 	if (!m || (m->m_len != sizeof(int *)))
404 		return ENOPROTOOPT;
405 
406 	v = mtod(m, int *);
407 	if (*v != 1)
408 		return ENOPROTOOPT;
409 
410 	if (ip6_mrouter != NULL) return EADDRINUSE;
411 
412 	ip6_mrouter = so;
413 	ip6_mrouter_ver = cmd;
414 
415 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
416 	bzero((caddr_t)nexpire, sizeof(nexpire));
417 
418 	pim6 = 0;/* used for stubbing out/in pim stuff */
419 
420 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
421 	    expire_upcalls, NULL);
422 
423 #ifdef MRT6DEBUG
424 	if (mrt6debug)
425 		log(LOG_DEBUG, "ip6_mrouter_init\n");
426 #endif
427 
428 	return 0;
429 }
430 
431 /*
432  * Disable multicast routing
433  */
434 int
435 ip6_mrouter_done()
436 {
437 	mifi_t mifi;
438 	int i;
439 	struct ifnet *ifp;
440 	struct in6_ifreq ifr;
441 	struct mf6c *rt;
442 	struct rtdetq *rte;
443 	int s;
444 
445 	s = splnet();
446 
447 	/*
448 	 * For each phyint in use, disable promiscuous reception of all IPv6
449 	 * multicasts.
450 	 */
451 #ifdef INET
452 #ifdef MROUTING
453 	/*
454 	 * If there is still IPv4 multicast routing daemon,
455 	 * we remain interfaces to receive all muliticasted packets.
456 	 * XXX: there may be an interface in which the IPv4 multicast
457 	 * daemon is not interested...
458 	 */
459 	if (!ip_mrouter)
460 #endif
461 #endif
462 	{
463 		for (mifi = 0; mifi < nummifs; mifi++) {
464 			if (mif6table[mifi].m6_ifp &&
465 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
466 				ifr.ifr_addr.sin6_family = AF_INET6;
467 				ifr.ifr_addr.sin6_addr= in6addr_any;
468 				ifp = mif6table[mifi].m6_ifp;
469 				(*ifp->if_ioctl)(ifp, SIOCDELMULTI,
470 						 (caddr_t)&ifr);
471 			}
472 		}
473 	}
474 #ifdef notyet
475 	bzero((caddr_t)qtable, sizeof(qtable));
476 	bzero((caddr_t)tbftable, sizeof(tbftable));
477 #endif
478 	bzero((caddr_t)mif6table, sizeof(mif6table));
479 	nummifs = 0;
480 
481 	pim6 = 0; /* used to stub out/in pim specific code */
482 
483 	callout_stop(&expire_upcalls_ch);
484 
485 	/*
486 	 * Free all multicast forwarding cache entries.
487 	 */
488 	for (i = 0; i < MF6CTBLSIZ; i++) {
489 		rt = mf6ctable[i];
490 		while (rt) {
491 			struct mf6c *frt;
492 
493 			for (rte = rt->mf6c_stall; rte != NULL; ) {
494 				struct rtdetq *n = rte->next;
495 
496 				m_free(rte->m);
497 				free(rte, M_MRTABLE);
498 				rte = n;
499 			}
500 			frt = rt;
501 			rt = rt->mf6c_next;
502 			free(frt, M_MRTABLE);
503 		}
504 	}
505 
506 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
507 
508 	/*
509 	 * Reset de-encapsulation cache
510 	 */
511 	reg_mif_num = -1;
512 
513 	ip6_mrouter = NULL;
514 	ip6_mrouter_ver = 0;
515 
516 	splx(s);
517 
518 #ifdef MRT6DEBUG
519 	if (mrt6debug)
520 		log(LOG_DEBUG, "ip6_mrouter_done\n");
521 #endif
522 
523 	return 0;
524 }
525 
526 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
527 
528 /*
529  * Add a mif to the mif table
530  */
531 static int
532 add_m6if(mifcp)
533 	struct mif6ctl *mifcp;
534 {
535 	struct mif6 *mifp;
536 	struct ifnet *ifp;
537 	int error, s;
538 #ifdef notyet
539 	struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
540 #endif
541 
542 	if (mifcp->mif6c_mifi >= MAXMIFS)
543 		return EINVAL;
544 	mifp = mif6table + mifcp->mif6c_mifi;
545 	if (mifp->m6_ifp)
546 		return EADDRINUSE; /* XXX: is it appropriate? */
547 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
548 		return ENXIO;
549 	ifp = ifindex2ifnet[mifcp->mif6c_pifi];
550 
551 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
552 		if (reg_mif_num == (mifi_t)-1) {
553 			multicast_register_if.if_name = "register_mif";
554 			multicast_register_if.if_flags |= IFF_LOOPBACK;
555 			multicast_register_if.if_index = mifcp->mif6c_mifi;
556 			reg_mif_num = mifcp->mif6c_mifi;
557 		}
558 
559 		ifp = &multicast_register_if;
560 
561 	} /* if REGISTER */
562 	else {
563 		/* Make sure the interface supports multicast */
564 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
565 			return EOPNOTSUPP;
566 
567 		s = splnet();
568 		error = if_allmulti(ifp, 1);
569 		splx(s);
570 		if (error)
571 			return error;
572 	}
573 
574 	s = splnet();
575 	mifp->m6_flags     = mifcp->mif6c_flags;
576 	mifp->m6_ifp       = ifp;
577 #ifdef notyet
578 	/* scaling up here allows division by 1024 in critical code */
579 	mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
580 #endif
581 	/* initialize per mif pkt counters */
582 	mifp->m6_pkt_in    = 0;
583 	mifp->m6_pkt_out   = 0;
584 	mifp->m6_bytes_in  = 0;
585 	mifp->m6_bytes_out = 0;
586 	splx(s);
587 
588 	/* Adjust nummifs up if the mifi is higher than nummifs */
589 	if (nummifs <= mifcp->mif6c_mifi)
590 		nummifs = mifcp->mif6c_mifi + 1;
591 
592 #ifdef MRT6DEBUG
593 	if (mrt6debug)
594 		log(LOG_DEBUG,
595 		    "add_mif #%d, phyint %s%d\n",
596 		    mifcp->mif6c_mifi,
597 		    ifp->if_name, ifp->if_unit);
598 #endif
599 
600 	return 0;
601 }
602 
603 /*
604  * Delete a mif from the mif table
605  */
606 static int
607 del_m6if(mifip)
608 	mifi_t *mifip;
609 {
610 	struct mif6 *mifp = mif6table + *mifip;
611 	mifi_t mifi;
612 	struct ifnet *ifp;
613 	int s;
614 
615 	if (*mifip >= nummifs)
616 		return EINVAL;
617 	if (mifp->m6_ifp == NULL)
618 		return EINVAL;
619 
620 	s = splnet();
621 
622 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
623 		/*
624 		 * XXX: what if there is yet IPv4 multicast daemon
625 		 *      using the interface?
626 		 */
627 		ifp = mifp->m6_ifp;
628 
629 		if_allmulti(ifp, 0);
630 	}
631 
632 #ifdef notyet
633 	bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
634 	bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
635 #endif
636 	bzero((caddr_t)mifp, sizeof (*mifp));
637 
638 	/* Adjust nummifs down */
639 	for (mifi = nummifs; mifi > 0; mifi--)
640 		if (mif6table[mifi - 1].m6_ifp)
641 			break;
642 	nummifs = mifi;
643 
644 	splx(s);
645 
646 #ifdef MRT6DEBUG
647 	if (mrt6debug)
648 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
649 #endif
650 
651 	return 0;
652 }
653 
654 /*
655  * Add an mfc entry
656  */
657 static int
658 add_m6fc(mfccp)
659 	struct mf6cctl *mfccp;
660 {
661 	struct mf6c *rt;
662 	u_long hash;
663 	struct rtdetq *rte;
664 	u_short nstl;
665 	int s;
666 
667 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
668 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
669 
670 	/* If an entry already exists, just update the fields */
671 	if (rt) {
672 #ifdef MRT6DEBUG
673 		if (mrt6debug & DEBUG_MFC)
674 			log(LOG_DEBUG,"add_m6fc update o %s g %s p %x\n",
675 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
676 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
677 			    mfccp->mf6cc_parent);
678 #endif
679 
680 		s = splnet();
681 		rt->mf6c_parent = mfccp->mf6cc_parent;
682 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
683 		splx(s);
684 		return 0;
685 	}
686 
687 	/*
688 	 * Find the entry for which the upcall was made and update
689 	 */
690 	s = splnet();
691 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
692 			mfccp->mf6cc_mcastgrp.sin6_addr);
693 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
694 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
695 				       &mfccp->mf6cc_origin.sin6_addr) &&
696 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
697 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
698 		    (rt->mf6c_stall != NULL)) {
699 
700 			if (nstl++)
701 				log(LOG_ERR,
702 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
703 				    "multiple kernel entries",
704 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
705 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
706 				    mfccp->mf6cc_parent, rt->mf6c_stall);
707 
708 #ifdef MRT6DEBUG
709 			if (mrt6debug & DEBUG_MFC)
710 				log(LOG_DEBUG,
711 				    "add_m6fc o %s g %s p %x dbg %x\n",
712 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
713 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
714 				    mfccp->mf6cc_parent, rt->mf6c_stall);
715 #endif
716 
717 			rt->mf6c_origin     = mfccp->mf6cc_origin;
718 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
719 			rt->mf6c_parent     = mfccp->mf6cc_parent;
720 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
721 			/* initialize pkt counters per src-grp */
722 			rt->mf6c_pkt_cnt    = 0;
723 			rt->mf6c_byte_cnt   = 0;
724 			rt->mf6c_wrong_if   = 0;
725 
726 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
727 			nexpire[hash]--;
728 
729 			/* free packets Qed at the end of this entry */
730 			for (rte = rt->mf6c_stall; rte != NULL; ) {
731 				struct rtdetq *n = rte->next;
732 				ip6_mdq(rte->m, rte->ifp, rt);
733 				m_freem(rte->m);
734 #ifdef UPCALL_TIMING
735 				collate(&(rte->t));
736 #endif /* UPCALL_TIMING */
737 				free(rte, M_MRTABLE);
738 				rte = n;
739 			}
740 			rt->mf6c_stall = NULL;
741 		}
742 	}
743 
744 	/*
745 	 * It is possible that an entry is being inserted without an upcall
746 	 */
747 	if (nstl == 0) {
748 #ifdef MRT6DEBUG
749 		if (mrt6debug & DEBUG_MFC)
750 			log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n",
751 			    hash,
752 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
753 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
754 			    mfccp->mf6cc_parent);
755 #endif
756 
757 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
758 
759 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
760 					       &mfccp->mf6cc_origin.sin6_addr)&&
761 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
762 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
763 
764 				rt->mf6c_origin     = mfccp->mf6cc_origin;
765 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
766 				rt->mf6c_parent     = mfccp->mf6cc_parent;
767 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
768 				/* initialize pkt counters per src-grp */
769 				rt->mf6c_pkt_cnt    = 0;
770 				rt->mf6c_byte_cnt   = 0;
771 				rt->mf6c_wrong_if   = 0;
772 
773 				if (rt->mf6c_expire)
774 					nexpire[hash]--;
775 				rt->mf6c_expire	   = 0;
776 			}
777 		}
778 		if (rt == NULL) {
779 			/* no upcall, so make a new entry */
780 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
781 						  M_NOWAIT);
782 			if (rt == NULL) {
783 				splx(s);
784 				return ENOBUFS;
785 			}
786 
787 			/* insert new entry at head of hash chain */
788 			rt->mf6c_origin     = mfccp->mf6cc_origin;
789 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
790 			rt->mf6c_parent     = mfccp->mf6cc_parent;
791 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
792 			/* initialize pkt counters per src-grp */
793 			rt->mf6c_pkt_cnt    = 0;
794 			rt->mf6c_byte_cnt   = 0;
795 			rt->mf6c_wrong_if   = 0;
796 			rt->mf6c_expire     = 0;
797 			rt->mf6c_stall = NULL;
798 
799 			/* link into table */
800 			rt->mf6c_next  = mf6ctable[hash];
801 			mf6ctable[hash] = rt;
802 		}
803 	}
804 	splx(s);
805 	return 0;
806 }
807 
808 #ifdef UPCALL_TIMING
809 /*
810  * collect delay statistics on the upcalls
811  */
812 static void
813 collate(t)
814 	struct timeval *t;
815 {
816 	u_long d;
817 	struct timeval tp;
818 	u_long delta;
819 
820 	GET_TIME(tp);
821 
822 	if (TV_LT(*t, tp))
823 	{
824 		TV_DELTA(tp, *t, delta);
825 
826 		d = delta >> 10;
827 		if (d > UPCALL_MAX)
828 			d = UPCALL_MAX;
829 
830 		++upcall_data[d];
831 	}
832 }
833 #endif /* UPCALL_TIMING */
834 
835 /*
836  * Delete an mfc entry
837  */
838 static int
839 del_m6fc(mfccp)
840 	struct mf6cctl *mfccp;
841 {
842 	struct sockaddr_in6 	origin;
843 	struct sockaddr_in6 	mcastgrp;
844 	struct mf6c 		*rt;
845 	struct mf6c	 	**nptr;
846 	u_long 		hash;
847 	int s;
848 
849 	origin = mfccp->mf6cc_origin;
850 	mcastgrp = mfccp->mf6cc_mcastgrp;
851 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
852 
853 #ifdef MRT6DEBUG
854 	if (mrt6debug & DEBUG_MFC)
855 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
856 		    ip6_sprintf(&origin.sin6_addr),
857 		    ip6_sprintf(&mcastgrp.sin6_addr));
858 #endif
859 
860 	s = splnet();
861 
862 	nptr = &mf6ctable[hash];
863 	while ((rt = *nptr) != NULL) {
864 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
865 				       &rt->mf6c_origin.sin6_addr) &&
866 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
867 				       &rt->mf6c_mcastgrp.sin6_addr) &&
868 		    rt->mf6c_stall == NULL)
869 			break;
870 
871 		nptr = &rt->mf6c_next;
872 	}
873 	if (rt == NULL) {
874 		splx(s);
875 		return EADDRNOTAVAIL;
876 	}
877 
878 	*nptr = rt->mf6c_next;
879 	free(rt, M_MRTABLE);
880 
881 	splx(s);
882 
883 	return 0;
884 }
885 
886 static int
887 socket_send(s, mm, src)
888 	struct socket *s;
889 	struct mbuf *mm;
890 	struct sockaddr_in6 *src;
891 {
892 	if (s) {
893 		if (sbappendaddr(&s->so_rcv,
894 				 (struct sockaddr *)src,
895 				 mm, (struct mbuf *)0) != 0) {
896 			sorwakeup(s);
897 			return 0;
898 		}
899 	}
900 	m_freem(mm);
901 	return -1;
902 }
903 
904 /*
905  * IPv6 multicast forwarding function. This function assumes that the packet
906  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
907  * pointed to by "ifp", and the packet is to be relayed to other networks
908  * that have members of the packet's destination IPv6 multicast group.
909  *
910  * The packet is returned unscathed to the caller, unless it is
911  * erroneous, in which case a non-zero return value tells the caller to
912  * discard it.
913  */
914 
915 int
916 ip6_mforward(ip6, ifp, m)
917 	struct ip6_hdr *ip6;
918 	struct ifnet *ifp;
919 	struct mbuf *m;
920 {
921 	struct mf6c *rt;
922 	struct mif6 *mifp;
923 	struct mbuf *mm;
924 	int s;
925 	mifi_t mifi;
926 
927 #ifdef MRT6DEBUG
928 	if (mrt6debug & DEBUG_FORWARD)
929 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
930 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
931 		    ifp->if_index);
932 #endif
933 
934 	/*
935 	 * Don't forward a packet with Hop limit of zero or one,
936 	 * or a packet destined to a local-only group.
937 	 */
938 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
939 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
940 		return 0;
941 	ip6->ip6_hlim--;
942 
943 	/*
944 	 * Source address check: do not forward packets with unspecified
945 	 * source. It was discussed in July 2000, on ipngwg mailing list.
946 	 * This is rather more serious than unicast cases, because some
947 	 * MLD packets can be sent with the unspecified source address
948 	 * (although such packets must normally set 1 to the hop limit field).
949 	 */
950 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
951 		ip6stat.ip6s_cantforward++;
952 		if (ip6_log_time + ip6_log_interval < time_second) {
953 			ip6_log_time = time_second;
954 			log(LOG_DEBUG,
955 			    "cannot forward "
956 			    "from %s to %s nxt %d received on %s\n",
957 			    ip6_sprintf(&ip6->ip6_src),
958 			    ip6_sprintf(&ip6->ip6_dst),
959 			    ip6->ip6_nxt,
960 			    if_name(m->m_pkthdr.rcvif));
961 		}
962 		return 0;
963 	}
964 
965 	/*
966 	 * Determine forwarding mifs from the forwarding cache table
967 	 */
968 	s = splnet();
969 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
970 
971 	/* Entry exists, so forward if necessary */
972 	if (rt) {
973 		splx(s);
974 		return (ip6_mdq(m, ifp, rt));
975 	} else {
976 		/*
977 		 * If we don't have a route for packet's origin,
978 		 * Make a copy of the packet &
979 		 * send message to routing daemon
980 		 */
981 
982 		struct mbuf *mb0;
983 		struct rtdetq *rte;
984 		u_long hash;
985 /*		int i, npkts;*/
986 #ifdef UPCALL_TIMING
987 		struct timeval tp;
988 
989 		GET_TIME(tp);
990 #endif /* UPCALL_TIMING */
991 
992 		mrt6stat.mrt6s_no_route++;
993 #ifdef MRT6DEBUG
994 		if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
995 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
996 			    ip6_sprintf(&ip6->ip6_src),
997 			    ip6_sprintf(&ip6->ip6_dst));
998 #endif
999 
1000 		/*
1001 		 * Allocate mbufs early so that we don't do extra work if we
1002 		 * are just going to fail anyway.
1003 		 */
1004 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1005 					      M_NOWAIT);
1006 		if (rte == NULL) {
1007 			splx(s);
1008 			return ENOBUFS;
1009 		}
1010 		mb0 = m_copy(m, 0, M_COPYALL);
1011 		/*
1012 		 * Pullup packet header if needed before storing it,
1013 		 * as other references may modify it in the meantime.
1014 		 */
1015 		if (mb0 &&
1016 		    (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1017 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1018 		if (mb0 == NULL) {
1019 			free(rte, M_MRTABLE);
1020 			splx(s);
1021 			return ENOBUFS;
1022 		}
1023 
1024 		/* is there an upcall waiting for this packet? */
1025 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1026 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1027 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1028 					       &rt->mf6c_origin.sin6_addr) &&
1029 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1030 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1031 			    (rt->mf6c_stall != NULL))
1032 				break;
1033 		}
1034 
1035 		if (rt == NULL) {
1036 			struct mrt6msg *im;
1037 #ifdef MRT6_OINIT
1038 			struct omrt6msg *oim;
1039 #endif
1040 
1041 			/* no upcall, so make a new entry */
1042 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1043 						  M_NOWAIT);
1044 			if (rt == NULL) {
1045 				free(rte, M_MRTABLE);
1046 				m_freem(mb0);
1047 				splx(s);
1048 				return ENOBUFS;
1049 			}
1050 			/*
1051 			 * Make a copy of the header to send to the user
1052 			 * level process
1053 			 */
1054 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1055 
1056 			if (mm == NULL) {
1057 				free(rte, M_MRTABLE);
1058 				m_freem(mb0);
1059 				free(rt, M_MRTABLE);
1060 				splx(s);
1061 				return ENOBUFS;
1062 			}
1063 
1064 			/*
1065 			 * Send message to routing daemon
1066 			 */
1067 			sin6.sin6_addr = ip6->ip6_src;
1068 
1069 			im = NULL;
1070 #ifdef MRT6_OINIT
1071 			oim = NULL;
1072 #endif
1073 			switch (ip6_mrouter_ver) {
1074 #ifdef MRT6_OINIT
1075 			case MRT6_OINIT:
1076 				oim = mtod(mm, struct omrt6msg *);
1077 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1078 				oim->im6_mbz = 0;
1079 				break;
1080 #endif
1081 			case MRT6_INIT:
1082 				im = mtod(mm, struct mrt6msg *);
1083 				im->im6_msgtype = MRT6MSG_NOCACHE;
1084 				im->im6_mbz = 0;
1085 				break;
1086 			default:
1087 				free(rte, M_MRTABLE);
1088 				m_freem(mb0);
1089 				free(rt, M_MRTABLE);
1090 				splx(s);
1091 				return EINVAL;
1092 			}
1093 
1094 #ifdef MRT6DEBUG
1095 			if (mrt6debug & DEBUG_FORWARD)
1096 				log(LOG_DEBUG,
1097 				    "getting the iif info in the kernel\n");
1098 #endif
1099 
1100 			for (mifp = mif6table, mifi = 0;
1101 			     mifi < nummifs && mifp->m6_ifp != ifp;
1102 			     mifp++, mifi++)
1103 				;
1104 
1105 			switch (ip6_mrouter_ver) {
1106 #ifdef MRT6_OINIT
1107 			case MRT6_OINIT:
1108 				oim->im6_mif = mifi;
1109 				break;
1110 #endif
1111 			case MRT6_INIT:
1112 				im->im6_mif = mifi;
1113 				break;
1114 			}
1115 
1116 			if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1117 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1118 				    "socket queue full\n");
1119 				mrt6stat.mrt6s_upq_sockfull++;
1120 				free(rte, M_MRTABLE);
1121 				m_freem(mb0);
1122 				free(rt, M_MRTABLE);
1123 				splx(s);
1124 				return ENOBUFS;
1125 			}
1126 
1127 			mrt6stat.mrt6s_upcalls++;
1128 
1129 			/* insert new entry at head of hash chain */
1130 			bzero(rt, sizeof(*rt));
1131 			rt->mf6c_origin.sin6_family = AF_INET6;
1132 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1133 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1134 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1135 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1136 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1137 			rt->mf6c_expire = UPCALL_EXPIRE;
1138 			nexpire[hash]++;
1139 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1140 
1141 			/* link into table */
1142 			rt->mf6c_next  = mf6ctable[hash];
1143 			mf6ctable[hash] = rt;
1144 			/* Add this entry to the end of the queue */
1145 			rt->mf6c_stall = rte;
1146 		} else {
1147 			/* determine if q has overflowed */
1148 			struct rtdetq **p;
1149 			int npkts = 0;
1150 
1151 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1152 				if (++npkts > MAX_UPQ6) {
1153 					mrt6stat.mrt6s_upq_ovflw++;
1154 					free(rte, M_MRTABLE);
1155 					m_freem(mb0);
1156 					splx(s);
1157 					return 0;
1158 				}
1159 
1160 			/* Add this entry to the end of the queue */
1161 			*p = rte;
1162 		}
1163 
1164 		rte->next = NULL;
1165 		rte->m = mb0;
1166 		rte->ifp = ifp;
1167 #ifdef UPCALL_TIMING
1168 		rte->t = tp;
1169 #endif /* UPCALL_TIMING */
1170 
1171 		splx(s);
1172 
1173 		return 0;
1174 	}
1175 }
1176 
1177 /*
1178  * Clean up cache entries if upcalls are not serviced
1179  * Call from the Slow Timeout mechanism, every half second.
1180  */
1181 static void
1182 expire_upcalls(unused)
1183 	void *unused;
1184 {
1185 	struct rtdetq *rte;
1186 	struct mf6c *mfc, **nptr;
1187 	int i;
1188 	int s;
1189 
1190 	s = splnet();
1191 	for (i = 0; i < MF6CTBLSIZ; i++) {
1192 		if (nexpire[i] == 0)
1193 			continue;
1194 		nptr = &mf6ctable[i];
1195 		while ((mfc = *nptr) != NULL) {
1196 			rte = mfc->mf6c_stall;
1197 			/*
1198 			 * Skip real cache entries
1199 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1200 			 * If it expires now
1201 			 */
1202 			if (rte != NULL &&
1203 			    mfc->mf6c_expire != 0 &&
1204 			    --mfc->mf6c_expire == 0) {
1205 #ifdef MRT6DEBUG
1206 				if (mrt6debug & DEBUG_EXPIRE)
1207 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1208 					    ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1209 					    ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1210 #endif
1211 				/*
1212 				 * drop all the packets
1213 				 * free the mbuf with the pkt, if, timing info
1214 				 */
1215 				do {
1216 					struct rtdetq *n = rte->next;
1217 					m_freem(rte->m);
1218 					free(rte, M_MRTABLE);
1219 					rte = n;
1220 				} while (rte != NULL);
1221 				mrt6stat.mrt6s_cache_cleanups++;
1222 				nexpire[i]--;
1223 
1224 				*nptr = mfc->mf6c_next;
1225 				free(mfc, M_MRTABLE);
1226 			} else {
1227 				nptr = &mfc->mf6c_next;
1228 			}
1229 		}
1230 	}
1231 	splx(s);
1232 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1233 	    expire_upcalls, NULL);
1234 }
1235 
1236 /*
1237  * Packet forwarding routine once entry in the cache is made
1238  */
1239 static int
1240 ip6_mdq(m, ifp, rt)
1241 	struct mbuf *m;
1242 	struct ifnet *ifp;
1243 	struct mf6c *rt;
1244 {
1245 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1246 	mifi_t mifi, iif;
1247 	struct mif6 *mifp;
1248 	int plen = m->m_pkthdr.len;
1249 
1250 /*
1251  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1252  * input, they shouldn't get counted on output, so statistics keeping is
1253  * separate.
1254  */
1255 
1256 #define MC6_SEND(ip6, mifp, m) do {				\
1257 		if ((mifp)->m6_flags & MIFF_REGISTER)		\
1258 		    register_send((ip6), (mifp), (m));		\
1259 		else						\
1260 		    phyint_send((ip6), (mifp), (m));		\
1261 } while (0)
1262 
1263 	/*
1264 	 * Don't forward if it didn't arrive from the parent mif
1265 	 * for its origin.
1266 	 */
1267 	mifi = rt->mf6c_parent;
1268 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1269 		/* came in the wrong interface */
1270 #ifdef MRT6DEBUG
1271 		if (mrt6debug & DEBUG_FORWARD)
1272 			log(LOG_DEBUG,
1273 			    "wrong if: ifid %d mifi %d mififid %x\n",
1274 			    ifp->if_index, mifi,
1275 			    mif6table[mifi].m6_ifp->if_index);
1276 #endif
1277 		mrt6stat.mrt6s_wrong_if++;
1278 		rt->mf6c_wrong_if++;
1279 		/*
1280 		 * If we are doing PIM processing, and we are forwarding
1281 		 * packets on this interface, send a message to the
1282 		 * routing daemon.
1283 		 */
1284 		/* have to make sure this is a valid mif */
1285 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1286 			if (pim6 && (m->m_flags & M_LOOP) == 0) {
1287 				/*
1288 				 * Check the M_LOOP flag to avoid an
1289 				 * unnecessary PIM assert.
1290 				 * XXX: M_LOOP is an ad-hoc hack...
1291 				 */
1292 				static struct sockaddr_in6 sin6 =
1293 				{ sizeof(sin6), AF_INET6 };
1294 
1295 				struct mbuf *mm;
1296 				struct mrt6msg *im;
1297 #ifdef MRT6_OINIT
1298 				struct omrt6msg *oim;
1299 #endif
1300 
1301 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1302 				if (mm &&
1303 				    (M_HASCL(mm) ||
1304 				     mm->m_len < sizeof(struct ip6_hdr)))
1305 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1306 				if (mm == NULL)
1307 					return ENOBUFS;
1308 
1309 #ifdef MRT6_OINIT
1310 				oim = NULL;
1311 #endif
1312 				im = NULL;
1313 				switch (ip6_mrouter_ver) {
1314 #ifdef MRT6_OINIT
1315 				case MRT6_OINIT:
1316 					oim = mtod(mm, struct omrt6msg *);
1317 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1318 					oim->im6_mbz = 0;
1319 					break;
1320 #endif
1321 				case MRT6_INIT:
1322 					im = mtod(mm, struct mrt6msg *);
1323 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1324 					im->im6_mbz = 0;
1325 					break;
1326 				default:
1327 					m_freem(mm);
1328 					return EINVAL;
1329 				}
1330 
1331 				for (mifp = mif6table, iif = 0;
1332 				     iif < nummifs && mifp &&
1333 					     mifp->m6_ifp != ifp;
1334 				     mifp++, iif++)
1335 					;
1336 
1337 				switch (ip6_mrouter_ver) {
1338 #ifdef MRT6_OINIT
1339 				case MRT6_OINIT:
1340 					oim->im6_mif = iif;
1341 					sin6.sin6_addr = oim->im6_src;
1342 					break;
1343 #endif
1344 				case MRT6_INIT:
1345 					im->im6_mif = iif;
1346 					sin6.sin6_addr = im->im6_src;
1347 					break;
1348 				}
1349 
1350 				mrt6stat.mrt6s_upcalls++;
1351 
1352 				if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1353 #ifdef MRT6DEBUG
1354 					if (mrt6debug)
1355 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1356 #endif
1357 					++mrt6stat.mrt6s_upq_sockfull;
1358 					return ENOBUFS;
1359 				}	/* if socket Q full */
1360 			}		/* if PIM */
1361 		return 0;
1362 	}			/* if wrong iif */
1363 
1364 	/* If I sourced this packet, it counts as output, else it was input. */
1365 	if (m->m_pkthdr.rcvif == NULL) {
1366 		/* XXX: is rcvif really NULL when output?? */
1367 		mif6table[mifi].m6_pkt_out++;
1368 		mif6table[mifi].m6_bytes_out += plen;
1369 	} else {
1370 		mif6table[mifi].m6_pkt_in++;
1371 		mif6table[mifi].m6_bytes_in += plen;
1372 	}
1373 	rt->mf6c_pkt_cnt++;
1374 	rt->mf6c_byte_cnt += plen;
1375 
1376 	/*
1377 	 * For each mif, forward a copy of the packet if there are group
1378 	 * members downstream on the interface.
1379 	 */
1380 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1381 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1382 			/*
1383 			 * check if the outgoing packet is going to break
1384 			 * a scope boundary.
1385 			 * XXX For packets through PIM register tunnel
1386 			 * interface, we believe a routing daemon.
1387 			 */
1388 			if ((mif6table[rt->mf6c_parent].m6_flags &
1389 			     MIFF_REGISTER) == 0 &&
1390 			    (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1391 			    (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1392 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1393 					      &ip6->ip6_dst) ||
1394 			     in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1395 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1396 					      &ip6->ip6_src))) {
1397 				ip6stat.ip6s_badscope++;
1398 				continue;
1399 			}
1400 
1401 			mifp->m6_pkt_out++;
1402 			mifp->m6_bytes_out += plen;
1403 			MC6_SEND(ip6, mifp, m);
1404 		}
1405 	return 0;
1406 }
1407 
1408 static void
1409 phyint_send(ip6, mifp, m)
1410     struct ip6_hdr *ip6;
1411     struct mif6 *mifp;
1412     struct mbuf *m;
1413 {
1414 	struct mbuf *mb_copy;
1415 	struct ifnet *ifp = mifp->m6_ifp;
1416 	int error = 0;
1417 	int s = splnet();	/* needs to protect static "ro" below. */
1418 	static struct route_in6 ro;
1419 	struct	in6_multi *in6m;
1420 	struct sockaddr_in6 *dst6;
1421 
1422 	/*
1423 	 * Make a new reference to the packet; make sure that
1424 	 * the IPv6 header is actually copied, not just referenced,
1425 	 * so that ip6_output() only scribbles on the copy.
1426 	 */
1427 	mb_copy = m_copy(m, 0, M_COPYALL);
1428 	if (mb_copy &&
1429 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1430 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1431 	if (mb_copy == NULL) {
1432 		splx(s);
1433 		return;
1434 	}
1435 	/* set MCAST flag to the outgoing packet */
1436 	mb_copy->m_flags |= M_MCAST;
1437 
1438 	/*
1439 	 * If we sourced the packet, call ip6_output since we may devide
1440 	 * the packet into fragments when the packet is too big for the
1441 	 * outgoing interface.
1442 	 * Otherwise, we can simply send the packet to the interface
1443 	 * sending queue.
1444 	 */
1445 	if (m->m_pkthdr.rcvif == NULL) {
1446 		struct ip6_moptions im6o;
1447 
1448 		im6o.im6o_multicast_ifp = ifp;
1449 		/* XXX: ip6_output will override ip6->ip6_hlim */
1450 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1451 		im6o.im6o_multicast_loop = 1;
1452 		error = ip6_output(mb_copy, NULL, &ro,
1453 				   IPV6_FORWARDING, &im6o, NULL);
1454 
1455 #ifdef MRT6DEBUG
1456 		if (mrt6debug & DEBUG_XMIT)
1457 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1458 			    mifp - mif6table, error);
1459 #endif
1460 		splx(s);
1461 		return;
1462 	}
1463 
1464 	/*
1465 	 * If we belong to the destination multicast group
1466 	 * on the outgoing interface, loop back a copy.
1467 	 */
1468 	dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
1469 	IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1470 	if (in6m != NULL) {
1471 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1472 		dst6->sin6_family = AF_INET6;
1473 		dst6->sin6_addr = ip6->ip6_dst;
1474 		ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst);
1475 	}
1476 	/*
1477 	 * Put the packet into the sending queue of the outgoing interface
1478 	 * if it would fit in the MTU of the interface.
1479 	 */
1480 	if (mb_copy->m_pkthdr.len < ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
1481 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1482 		dst6->sin6_family = AF_INET6;
1483 		dst6->sin6_addr = ip6->ip6_dst;
1484 		/*
1485 		 * We just call if_output instead of nd6_output here, since
1486 		 * we need no ND for a multicast forwarded packet...right?
1487 		 */
1488 		error = (*ifp->if_output)(ifp, mb_copy,
1489 		    (struct sockaddr *)&ro.ro_dst, NULL);
1490 #ifdef MRT6DEBUG
1491 		if (mrt6debug & DEBUG_XMIT)
1492 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1493 			    mifp - mif6table, error);
1494 #endif
1495 	} else {
1496 #ifdef MULTICAST_PMTUD
1497 		icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
1498 #else
1499 #ifdef MRT6DEBUG
1500 		if (mrt6debug & DEBUG_XMIT)
1501 			log(LOG_DEBUG,
1502 			    "phyint_send: packet too big on %s o %s g %s"
1503 			    " size %d(discarded)\n",
1504 			    if_name(ifp),
1505 			    ip6_sprintf(&ip6->ip6_src),
1506 			    ip6_sprintf(&ip6->ip6_dst),
1507 			    mb_copy->m_pkthdr.len);
1508 #endif /* MRT6DEBUG */
1509 		m_freem(mb_copy); /* simply discard the packet */
1510 #endif
1511 	}
1512 
1513 	splx(s);
1514 }
1515 
1516 static int
1517 register_send(ip6, mif, m)
1518 	struct ip6_hdr *ip6;
1519 	struct mif6 *mif;
1520 	struct mbuf *m;
1521 {
1522 	struct mbuf *mm;
1523 	int i, len = m->m_pkthdr.len;
1524 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1525 	struct mrt6msg *im6;
1526 
1527 #ifdef MRT6DEBUG
1528 	if (mrt6debug)
1529 		log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1530 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1531 #endif
1532 	++pim6stat.pim6s_snd_registers;
1533 
1534 	/* Make a copy of the packet to send to the user level process */
1535 	MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1536 	if (mm == NULL)
1537 		return ENOBUFS;
1538 	mm->m_pkthdr.rcvif = NULL;
1539 	mm->m_data += max_linkhdr;
1540 	mm->m_len = sizeof(struct ip6_hdr);
1541 
1542 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1543 		m_freem(mm);
1544 		return ENOBUFS;
1545 	}
1546 	i = MHLEN - M_LEADINGSPACE(mm);
1547 	if (i > len)
1548 		i = len;
1549 	mm = m_pullup(mm, i);
1550 	if (mm == NULL){
1551 		m_freem(mm);
1552 		return ENOBUFS;
1553 	}
1554 /* TODO: check it! */
1555 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1556 
1557 	/*
1558 	 * Send message to routing daemon
1559 	 */
1560 	sin6.sin6_addr = ip6->ip6_src;
1561 
1562 	im6 = mtod(mm, struct mrt6msg *);
1563 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1564 	im6->im6_mbz          = 0;
1565 
1566 	im6->im6_mif = mif - mif6table;
1567 
1568 	/* iif info is not given for reg. encap.n */
1569 	mrt6stat.mrt6s_upcalls++;
1570 
1571 	if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1572 #ifdef MRT6DEBUG
1573 		if (mrt6debug)
1574 			log(LOG_WARNING,
1575 			    "register_send: ip_mrouter socket queue full\n");
1576 #endif
1577 		++mrt6stat.mrt6s_upq_sockfull;
1578 		return ENOBUFS;
1579 	}
1580 	return 0;
1581 }
1582 
1583 /*
1584  * PIM sparse mode hook
1585  * Receives the pim control messages, and passes them up to the listening
1586  * socket, using rip6_input.
1587  * The only message processed is the REGISTER pim message; the pim header
1588  * is stripped off, and the inner packet is passed to register_mforward.
1589  */
1590 int
1591 pim6_input(mp, offp, proto)
1592 	struct mbuf **mp;
1593 	int *offp, proto;
1594 {
1595 	struct pim *pim; /* pointer to a pim struct */
1596 	struct ip6_hdr *ip6;
1597 	int pimlen;
1598 	struct mbuf *m = *mp;
1599 	int minlen;
1600 	int off = *offp;
1601 
1602 	++pim6stat.pim6s_rcv_total;
1603 
1604 	ip6 = mtod(m, struct ip6_hdr *);
1605 	pimlen = m->m_pkthdr.len - *offp;
1606 
1607 	/*
1608 	 * Validate lengths
1609 	 */
1610 	if (pimlen < PIM_MINLEN) {
1611 		++pim6stat.pim6s_rcv_tooshort;
1612 #ifdef MRT6DEBUG
1613 		if (mrt6debug & DEBUG_PIM)
1614 			log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1615 #endif
1616 		m_freem(m);
1617 		return(IPPROTO_DONE);
1618 	}
1619 
1620 	/*
1621 	 * if the packet is at least as big as a REGISTER, go ahead
1622 	 * and grab the PIM REGISTER header size, to avoid another
1623 	 * possible m_pullup() later.
1624 	 *
1625 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1626 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1627 	 */
1628 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1629 
1630 	/*
1631 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1632 	 * possibly the PIM REGISTER header
1633 	 */
1634 #ifndef PULLDOWN_TEST
1635 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1636 	/* adjust pointer */
1637 	ip6 = mtod(m, struct ip6_hdr *);
1638 
1639 	/* adjust mbuf to point to the PIM header */
1640 	pim = (struct pim *)((caddr_t)ip6 + off);
1641 #else
1642 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1643 	if (pim == NULL) {
1644 		pim6stat.pim6s_rcv_tooshort++;
1645 		return IPPROTO_DONE;
1646 	}
1647 #endif
1648 
1649 #define PIM6_CHECKSUM
1650 #ifdef PIM6_CHECKSUM
1651 	{
1652 		int cksumlen;
1653 
1654 		/*
1655 		 * Validate checksum.
1656 		 * If PIM REGISTER, exclude the data packet
1657 		 */
1658 		if (pim->pim_type == PIM_REGISTER)
1659 			cksumlen = PIM_MINLEN;
1660 		else
1661 			cksumlen = pimlen;
1662 
1663 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1664 			++pim6stat.pim6s_rcv_badsum;
1665 #ifdef MRT6DEBUG
1666 			if (mrt6debug & DEBUG_PIM)
1667 				log(LOG_DEBUG,
1668 				    "pim6_input: invalid checksum\n");
1669 #endif
1670 			m_freem(m);
1671 			return(IPPROTO_DONE);
1672 		}
1673 	}
1674 #endif /* PIM_CHECKSUM */
1675 
1676 	/* PIM version check */
1677 	if (pim->pim_ver != PIM_VERSION) {
1678 		++pim6stat.pim6s_rcv_badversion;
1679 #ifdef MRT6DEBUG
1680 		log(LOG_ERR,
1681 		    "pim6_input: incorrect version %d, expecting %d\n",
1682 		    pim->pim_ver, PIM_VERSION);
1683 #endif
1684 		m_freem(m);
1685 		return(IPPROTO_DONE);
1686 	}
1687 
1688 	if (pim->pim_type == PIM_REGISTER) {
1689 		/*
1690 		 * since this is a REGISTER, we'll make a copy of the register
1691 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1692 		 * routing daemon.
1693 		 */
1694 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1695 
1696 		struct mbuf *mcp;
1697 		struct ip6_hdr *eip6;
1698 		u_int32_t *reghdr;
1699 		int rc;
1700 
1701 		++pim6stat.pim6s_rcv_registers;
1702 
1703 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1704 #ifdef MRT6DEBUG
1705 			if (mrt6debug & DEBUG_PIM)
1706 				log(LOG_DEBUG,
1707 				    "pim6_input: register mif not set: %d\n",
1708 				    reg_mif_num);
1709 #endif
1710 			m_freem(m);
1711 			return(IPPROTO_DONE);
1712 		}
1713 
1714 		reghdr = (u_int32_t *)(pim + 1);
1715 
1716 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1717 			goto pim6_input_to_daemon;
1718 
1719 		/*
1720 		 * Validate length
1721 		 */
1722 		if (pimlen < PIM6_REG_MINLEN) {
1723 			++pim6stat.pim6s_rcv_tooshort;
1724 			++pim6stat.pim6s_rcv_badregisters;
1725 #ifdef MRT6DEBUG
1726 			log(LOG_ERR,
1727 			    "pim6_input: register packet size too "
1728 			    "small %d from %s\n",
1729 			    pimlen, ip6_sprintf(&ip6->ip6_src));
1730 #endif
1731 			m_freem(m);
1732 			return(IPPROTO_DONE);
1733 		}
1734 
1735 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1736 #ifdef MRT6DEBUG
1737 		if (mrt6debug & DEBUG_PIM)
1738 			log(LOG_DEBUG,
1739 			    "pim6_input[register], eip6: %s -> %s, "
1740 			    "eip6 plen %d\n",
1741 			    ip6_sprintf(&eip6->ip6_src),
1742 			    ip6_sprintf(&eip6->ip6_dst),
1743 			    ntohs(eip6->ip6_plen));
1744 #endif
1745 
1746 		/* verify the version number of the inner packet */
1747 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1748 			++pim6stat.pim6s_rcv_badregisters;
1749 #ifdef MRT6DEBUG
1750 			log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1751 			    "of the inner packet\n",
1752 			    (eip6->ip6_vfc & IPV6_VERSION));
1753 #endif
1754 			m_freem(m);
1755 			return(IPPROTO_NONE);
1756 		}
1757 
1758 		/* verify the inner packet is destined to a mcast group */
1759 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1760 			++pim6stat.pim6s_rcv_badregisters;
1761 #ifdef MRT6DEBUG
1762 			if (mrt6debug & DEBUG_PIM)
1763 				log(LOG_DEBUG,
1764 				    "pim6_input: inner packet of register "
1765 				    "is not multicast %s\n",
1766 				    ip6_sprintf(&eip6->ip6_dst));
1767 #endif
1768 			m_freem(m);
1769 			return(IPPROTO_DONE);
1770 		}
1771 
1772 		/*
1773 		 * make a copy of the whole header to pass to the daemon later.
1774 		 */
1775 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1776 		if (mcp == NULL) {
1777 #ifdef MRT6DEBUG
1778 			log(LOG_ERR,
1779 			    "pim6_input: pim register: "
1780 			    "could not copy register head\n");
1781 #endif
1782 			m_freem(m);
1783 			return(IPPROTO_DONE);
1784 		}
1785 
1786 		/*
1787 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1788 		 */
1789 		m_adj(m, off + PIM_MINLEN);
1790 #ifdef MRT6DEBUG
1791 		if (mrt6debug & DEBUG_PIM) {
1792 			log(LOG_DEBUG,
1793 			    "pim6_input: forwarding decapsulated register: "
1794 			    "src %s, dst %s, mif %d\n",
1795 			    ip6_sprintf(&eip6->ip6_src),
1796 			    ip6_sprintf(&eip6->ip6_dst),
1797 			    reg_mif_num);
1798 		}
1799 #endif
1800 
1801  		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1802 				dst.sin6_family, NULL);
1803 
1804 		/* prepare the register head to send to the mrouting daemon */
1805 		m = mcp;
1806 	}
1807 
1808 	/*
1809 	 * Pass the PIM message up to the daemon; if it is a register message
1810 	 * pass the 'head' only up to the daemon. This includes the
1811 	 * encapsulator ip6 header, pim header, register header and the
1812 	 * encapsulated ip6 header.
1813 	 */
1814   pim6_input_to_daemon:
1815 	rip6_input(&m, offp, proto);
1816 	return(IPPROTO_DONE);
1817 }
1818