xref: /openbsd-src/sys/netinet6/ip6_mroute.c (revision b2ea75c1b17e1a9a339660e7ed45cd24946b230e)
1 /*	$OpenBSD: ip6_mroute.c,v 1.16 2001/06/09 06:43:38 angelos Exp $	*/
2 /*	$KAME: ip6_mroute.c,v 1.45 2001/03/25 08:38:51 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 #ifndef _KERNEL
49 # ifdef KERNEL
50 #  define _KERNEL
51 # endif
52 #endif
53 
54 #include <sys/param.h>
55 #include <sys/systm.h>
56 #include <sys/timeout.h>
57 #include <sys/mbuf.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/sockio.h>
61 #include <sys/protosw.h>
62 #include <sys/errno.h>
63 #include <sys/time.h>
64 #include <sys/kernel.h>
65 #include <sys/ioctl.h>
66 #include <sys/syslog.h>
67 
68 #include <net/if.h>
69 #include <net/route.h>
70 #include <net/raw_cb.h>
71 
72 #include <netinet/in.h>
73 #include <netinet/in_var.h>
74 
75 #include <netinet/ip6.h>
76 #include <netinet6/ip6_var.h>
77 #include <netinet6/ip6_mroute.h>
78 #include <netinet6/pim6.h>
79 #include <netinet6/pim6_var.h>
80 
81 #define M_HASCL(m) ((m)->m_flags & M_EXT)
82 
83 static int ip6_mdq __P((struct mbuf *, struct ifnet *, struct mf6c *));
84 static void phyint_send __P((struct ip6_hdr *, struct mif6 *, struct mbuf *));
85 
86 static int set_pim6 __P((int *));
87 static int get_pim6 __P((struct mbuf *));
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 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 timeout expire_upcalls_ch;
225 
226 /*
227  * Handle MRT setsockopt commands to modify the multicast routing tables.
228  */
229 int
230 ip6_mrouter_set(cmd, so, m)
231 	int cmd;
232 	struct socket *so;
233 	struct mbuf *m;
234 {
235 	if (cmd != MRT6_INIT && so != ip6_mrouter)
236 		return EACCES;
237 
238 	switch (cmd) {
239 #ifdef MRT6_OINIT
240 	case MRT6_OINIT:	return ip6_mrouter_init(so, m, cmd);
241 #endif
242 	case MRT6_INIT:		return ip6_mrouter_init(so, m, cmd);
243 	case MRT6_DONE:		return ip6_mrouter_done();
244 	case MRT6_ADD_MIF:	return add_m6if(mtod(m, struct mif6ctl *));
245 	case MRT6_DEL_MIF:	return del_m6if(mtod(m, mifi_t *));
246 	case MRT6_ADD_MFC:	return add_m6fc(mtod(m, struct mf6cctl *));
247 	case MRT6_DEL_MFC:	return del_m6fc(mtod(m, struct mf6cctl *));
248 	case MRT6_PIM:		return set_pim6(mtod(m, int *));
249 	default:		return EOPNOTSUPP;
250 	}
251 }
252 
253 /*
254  * Handle MRT getsockopt commands
255  */
256 int
257 ip6_mrouter_get(cmd, so, m)
258 	int cmd;
259 	struct socket *so;
260 	struct mbuf **m;
261 {
262 	struct mbuf *mb;
263 
264 	if (so != ip6_mrouter) return EACCES;
265 
266 	*m = mb = m_get(M_WAIT, MT_SOOPTS);
267 
268 	switch (cmd) {
269 	case MRT6_PIM:
270 		return get_pim6(mb);
271 	default:
272 		m_free(mb);
273 		return EOPNOTSUPP;
274 	}
275 }
276 
277 /*
278  * Handle ioctl commands to obtain information from the cache
279  */
280 int
281 mrt6_ioctl(cmd, data)
282 	int cmd;
283 	caddr_t data;
284 {
285 	int error = 0;
286 
287 	switch (cmd) {
288 	case SIOCGETSGCNT_IN6:
289 		return(get_sg_cnt((struct sioc_sg_req6 *)data));
290 		break;		/* for safety */
291 	case SIOCGETMIFCNT_IN6:
292 		return(get_mif6_cnt((struct sioc_mif_req6 *)data));
293 		break;		/* for safety */
294 	default:
295 		return (EINVAL);
296 		break;
297 	}
298 	return error;
299 }
300 
301 /*
302  * returns the packet, byte, rpf-failure count for the source group provided
303  */
304 static int
305 get_sg_cnt(req)
306 	struct sioc_sg_req6 *req;
307 {
308 	struct mf6c *rt;
309 	int s;
310 
311 	s = splnet();
312 
313 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
314 	splx(s);
315 	if (rt != NULL) {
316 		req->pktcnt = rt->mf6c_pkt_cnt;
317 		req->bytecnt = rt->mf6c_byte_cnt;
318 		req->wrong_if = rt->mf6c_wrong_if;
319 	} else
320 		return(ESRCH);
321 #if 0
322 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
323 #endif
324 
325 	return 0;
326 }
327 
328 /*
329  * returns the input and output packet and byte counts on the mif provided
330  */
331 static int
332 get_mif6_cnt(req)
333 	struct sioc_mif_req6 *req;
334 {
335 	mifi_t mifi = req->mifi;
336 
337 	if (mifi >= nummifs)
338 		return EINVAL;
339 
340 	req->icount = mif6table[mifi].m6_pkt_in;
341 	req->ocount = mif6table[mifi].m6_pkt_out;
342 	req->ibytes = mif6table[mifi].m6_bytes_in;
343 	req->obytes = mif6table[mifi].m6_bytes_out;
344 
345 	return 0;
346 }
347 
348 /*
349  * Get PIM processiong global
350  */
351 static int
352 get_pim6(m)
353 	struct mbuf *m;
354 {
355 	int *i;
356 
357 	i = mtod(m, int *);
358 
359 	*i = pim6;
360 
361 	return 0;
362 }
363 
364 static int
365 set_pim6(i)
366 	int *i;
367 {
368 	if ((*i != 1) && (*i != 0))
369 		return EINVAL;
370 
371 	pim6 = *i;
372 
373 	return 0;
374 }
375 
376 /*
377  * Enable multicast routing
378  */
379 static int
380 ip6_mrouter_init(so, m, cmd)
381 	struct socket *so;
382 	struct mbuf *m;
383 	int cmd;
384 {
385 	int *v;
386 
387 #ifdef MRT6DEBUG
388 	if (mrt6debug)
389 		log(LOG_DEBUG,
390 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
391 		    so->so_type, so->so_proto->pr_protocol);
392 #endif
393 
394 	if (so->so_type != SOCK_RAW ||
395 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
396 		return EOPNOTSUPP;
397 
398 	if (!m || (m->m_len != sizeof(int *)))
399 		return ENOPROTOOPT;
400 
401 	v = mtod(m, int *);
402 	if (*v != 1)
403 		return ENOPROTOOPT;
404 
405 	if (ip6_mrouter != NULL) return EADDRINUSE;
406 
407 	ip6_mrouter = so;
408 	ip6_mrouter_ver = cmd;
409 
410 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
411 	bzero((caddr_t)nexpire, sizeof(nexpire));
412 
413 	pim6 = 0;/* used for stubbing out/in pim stuff */
414 
415 	timeout_set(&expire_upcalls_ch, expire_upcalls, NULL);
416 	timeout_add(&expire_upcalls_ch, EXPIRE_TIMEOUT);
417 
418 #ifdef MRT6DEBUG
419 	if (mrt6debug)
420 		log(LOG_DEBUG, "ip6_mrouter_init\n");
421 #endif
422 
423 	return 0;
424 }
425 
426 /*
427  * Disable multicast routing
428  */
429 int
430 ip6_mrouter_done()
431 {
432 	mifi_t mifi;
433 	int i;
434 	struct ifnet *ifp;
435 	struct in6_ifreq ifr;
436 	struct mf6c *rt;
437 	struct rtdetq *rte;
438 	int s;
439 
440 	s = splnet();
441 
442 	/*
443 	 * For each phyint in use, disable promiscuous reception of all IPv6
444 	 * multicasts.
445 	 */
446 #ifdef INET
447 #ifdef MROUTING
448 	/*
449 	 * If there is still IPv4 multicast routing daemon,
450 	 * we remain interfaces to receive all muliticasted packets.
451 	 * XXX: there may be an interface in which the IPv4 multicast
452 	 * daemon is not interested...
453 	 */
454 	if (!ip_mrouter)
455 #endif
456 #endif
457 	{
458 		for (mifi = 0; mifi < nummifs; mifi++) {
459 			if (mif6table[mifi].m6_ifp &&
460 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
461 				ifr.ifr_addr.sin6_family = AF_INET6;
462 				ifr.ifr_addr.sin6_addr= in6addr_any;
463 				ifp = mif6table[mifi].m6_ifp;
464 				(*ifp->if_ioctl)(ifp, SIOCDELMULTI,
465 						 (caddr_t)&ifr);
466 			}
467 		}
468 	}
469 #ifdef notyet
470 	bzero((caddr_t)qtable, sizeof(qtable));
471 	bzero((caddr_t)tbftable, sizeof(tbftable));
472 #endif
473 	bzero((caddr_t)mif6table, sizeof(mif6table));
474 	nummifs = 0;
475 
476 	pim6 = 0; /* used to stub out/in pim specific code */
477 
478 	timeout_del(&expire_upcalls_ch);
479 
480 	/*
481 	 * Free all multicast forwarding cache entries.
482 	 */
483 	for (i = 0; i < MF6CTBLSIZ; i++) {
484 		rt = mf6ctable[i];
485 		while (rt) {
486 			struct mf6c *frt;
487 
488 			for (rte = rt->mf6c_stall; rte != NULL; ) {
489 				struct rtdetq *n = rte->next;
490 
491 				m_free(rte->m);
492 				free(rte, M_MRTABLE);
493 				rte = n;
494 			}
495 			frt = rt;
496 			rt = rt->mf6c_next;
497 			free(frt, M_MRTABLE);
498 		}
499 	}
500 
501 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
502 
503 	/*
504 	 * Reset de-encapsulation cache
505 	 */
506 	reg_mif_num = -1;
507 
508 	ip6_mrouter = NULL;
509 	ip6_mrouter_ver = 0;
510 
511 	splx(s);
512 
513 #ifdef MRT6DEBUG
514 	if (mrt6debug)
515 		log(LOG_DEBUG, "ip6_mrouter_done\n");
516 #endif
517 
518 	return 0;
519 }
520 
521 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
522 
523 /*
524  * Add a mif to the mif table
525  */
526 static int
527 add_m6if(mifcp)
528 	struct mif6ctl *mifcp;
529 {
530 	struct mif6 *mifp;
531 	struct ifnet *ifp;
532 	struct in6_ifreq ifr;
533 	int error, s;
534 #ifdef notyet
535 	struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
536 #endif
537 
538 	if (mifcp->mif6c_mifi >= MAXMIFS)
539 		return EINVAL;
540 	mifp = mif6table + mifcp->mif6c_mifi;
541 	if (mifp->m6_ifp)
542 		return EADDRINUSE; /* XXX: is it appropriate? */
543 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
544 		return ENXIO;
545 	ifp = ifindex2ifnet[mifcp->mif6c_pifi];
546 
547 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
548 		if (reg_mif_num == (mifi_t)-1) {
549 			strcpy(multicast_register_if.if_xname,
550 			       "register_mif"); /* XXX */
551 			multicast_register_if.if_flags |= IFF_LOOPBACK;
552 			multicast_register_if.if_index = mifcp->mif6c_mifi;
553 			reg_mif_num = mifcp->mif6c_mifi;
554 		}
555 
556 		ifp = &multicast_register_if;
557 
558 	} /* if REGISTER */
559 	else {
560 		/* Make sure the interface supports multicast */
561 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
562 			return EOPNOTSUPP;
563 
564 		s = splnet();
565 
566 		/*
567 		 * Enable promiscuous reception of all IPv6 multicasts
568 		 * from the interface.
569 		 */
570 		ifr.ifr_addr.sin6_family = AF_INET6;
571 		ifr.ifr_addr.sin6_addr = in6addr_any;
572 		error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
573 
574 		splx(s);
575 		if (error)
576 			return error;
577 	}
578 
579 	s = splnet();
580 
581 	mifp->m6_flags     = mifcp->mif6c_flags;
582 	mifp->m6_ifp       = ifp;
583 #ifdef notyet
584 	/* scaling up here allows division by 1024 in critical code */
585 	mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
586 #endif
587 	/* initialize per mif pkt counters */
588 	mifp->m6_pkt_in    = 0;
589 	mifp->m6_pkt_out   = 0;
590 	mifp->m6_bytes_in  = 0;
591 	mifp->m6_bytes_out = 0;
592 	splx(s);
593 
594 	/* Adjust nummifs up if the mifi is higher than nummifs */
595 	if (nummifs <= mifcp->mif6c_mifi)
596 		nummifs = mifcp->mif6c_mifi + 1;
597 
598 #ifdef MRT6DEBUG
599 	if (mrt6debug)
600 		log(LOG_DEBUG,
601 		    "add_mif #%d, phyint %s%d\n",
602 		    mifcp->mif6c_mifi,
603 		    ifp->if_name, ifp->if_unit);
604 #endif
605 
606 	return 0;
607 }
608 
609 /*
610  * Delete a mif from the mif table
611  */
612 static int
613 del_m6if(mifip)
614 	mifi_t *mifip;
615 {
616 	struct mif6 *mifp = mif6table + *mifip;
617 	mifi_t mifi;
618 	struct ifnet *ifp;
619 	struct in6_ifreq ifr;
620 	int s;
621 
622 	if (*mifip >= nummifs)
623 		return EINVAL;
624 	if (mifp->m6_ifp == NULL)
625 		return EINVAL;
626 
627 	s = splnet();
628 
629 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
630 		/*
631 		 * XXX: what if there is yet IPv4 multicast daemon
632 		 *      using the interface?
633 		 */
634 		ifp = mifp->m6_ifp;
635 
636 		ifr.ifr_addr.sin6_family = AF_INET6;
637 		ifr.ifr_addr.sin6_addr = in6addr_any;
638 		(*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
639 	}
640 
641 #ifdef notyet
642 	bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
643 	bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
644 #endif
645 	bzero((caddr_t)mifp, sizeof (*mifp));
646 
647 	/* Adjust nummifs down */
648 	for (mifi = nummifs; mifi > 0; mifi--)
649 		if (mif6table[mifi - 1].m6_ifp)
650 			break;
651 	nummifs = mifi;
652 
653 	splx(s);
654 
655 #ifdef MRT6DEBUG
656 	if (mrt6debug)
657 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
658 #endif
659 
660 	return 0;
661 }
662 
663 /*
664  * Add an mfc entry
665  */
666 static int
667 add_m6fc(mfccp)
668 	struct mf6cctl *mfccp;
669 {
670 	struct mf6c *rt;
671 	u_long hash;
672 	struct rtdetq *rte;
673 	u_short nstl;
674 	int s;
675 
676 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
677 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
678 
679 	/* If an entry already exists, just update the fields */
680 	if (rt) {
681 #ifdef MRT6DEBUG
682 		if (mrt6debug & DEBUG_MFC)
683 			log(LOG_DEBUG,"add_m6fc update o %s g %s p %x\n",
684 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
685 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
686 			    mfccp->mf6cc_parent);
687 #endif
688 
689 		s = splnet();
690 
691 		rt->mf6c_parent = mfccp->mf6cc_parent;
692 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
693 		splx(s);
694 		return 0;
695 	}
696 
697 	/*
698 	 * Find the entry for which the upcall was made and update
699 	 */
700 	s = splnet();
701 
702 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
703 			mfccp->mf6cc_mcastgrp.sin6_addr);
704 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
705 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
706 				       &mfccp->mf6cc_origin.sin6_addr) &&
707 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
708 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
709 		    (rt->mf6c_stall != NULL)) {
710 
711 			if (nstl++)
712 				log(LOG_ERR,
713 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
714 				    "multiple kernel entries",
715 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
716 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
717 				    mfccp->mf6cc_parent, rt->mf6c_stall);
718 
719 #ifdef MRT6DEBUG
720 			if (mrt6debug & DEBUG_MFC)
721 				log(LOG_DEBUG,
722 				    "add_m6fc o %s g %s p %x dbg %x\n",
723 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
724 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
725 				    mfccp->mf6cc_parent, rt->mf6c_stall);
726 #endif
727 
728 			rt->mf6c_origin     = mfccp->mf6cc_origin;
729 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
730 			rt->mf6c_parent     = mfccp->mf6cc_parent;
731 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
732 			/* initialize pkt counters per src-grp */
733 			rt->mf6c_pkt_cnt    = 0;
734 			rt->mf6c_byte_cnt   = 0;
735 			rt->mf6c_wrong_if   = 0;
736 
737 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
738 			nexpire[hash]--;
739 
740 			/* free packets Qed at the end of this entry */
741 			for (rte = rt->mf6c_stall; rte != NULL; ) {
742 				struct rtdetq *n = rte->next;
743 				ip6_mdq(rte->m, rte->ifp, rt);
744 				m_freem(rte->m);
745 #ifdef UPCALL_TIMING
746 				collate(&(rte->t));
747 #endif /* UPCALL_TIMING */
748 				free(rte, M_MRTABLE);
749 				rte = n;
750 			}
751 			rt->mf6c_stall = NULL;
752 		}
753 	}
754 
755 	/*
756 	 * It is possible that an entry is being inserted without an upcall
757 	 */
758 	if (nstl == 0) {
759 #ifdef MRT6DEBUG
760 		if (mrt6debug & DEBUG_MFC)
761 			log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n",
762 			    hash,
763 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
764 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
765 			    mfccp->mf6cc_parent);
766 #endif
767 
768 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
769 
770 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
771 					       &mfccp->mf6cc_origin.sin6_addr)&&
772 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
773 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
774 
775 				rt->mf6c_origin     = mfccp->mf6cc_origin;
776 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
777 				rt->mf6c_parent     = mfccp->mf6cc_parent;
778 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
779 				/* initialize pkt counters per src-grp */
780 				rt->mf6c_pkt_cnt    = 0;
781 				rt->mf6c_byte_cnt   = 0;
782 				rt->mf6c_wrong_if   = 0;
783 
784 				if (rt->mf6c_expire)
785 					nexpire[hash]--;
786 				rt->mf6c_expire	   = 0;
787 			}
788 		}
789 		if (rt == NULL) {
790 			/* no upcall, so make a new entry */
791 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
792 						  M_NOWAIT);
793 			if (rt == NULL) {
794 				splx(s);
795 				return ENOBUFS;
796 			}
797 
798 			/* insert new entry at head of hash chain */
799 			rt->mf6c_origin     = mfccp->mf6cc_origin;
800 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
801 			rt->mf6c_parent     = mfccp->mf6cc_parent;
802 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
803 			/* initialize pkt counters per src-grp */
804 			rt->mf6c_pkt_cnt    = 0;
805 			rt->mf6c_byte_cnt   = 0;
806 			rt->mf6c_wrong_if   = 0;
807 			rt->mf6c_expire     = 0;
808 			rt->mf6c_stall = NULL;
809 
810 			/* link into table */
811 			rt->mf6c_next  = mf6ctable[hash];
812 			mf6ctable[hash] = rt;
813 		}
814 	}
815 	splx(s);
816 	return 0;
817 }
818 
819 #ifdef UPCALL_TIMING
820 /*
821  * collect delay statistics on the upcalls
822  */
823 static void
824 collate(t)
825 	struct timeval *t;
826 {
827 	u_long d;
828 	struct timeval tp;
829 	u_long delta;
830 
831 	GET_TIME(tp);
832 
833 	if (TV_LT(*t, tp))
834 	{
835 		TV_DELTA(tp, *t, delta);
836 
837 		d = delta >> 10;
838 		if (d > UPCALL_MAX)
839 			d = UPCALL_MAX;
840 
841 		++upcall_data[d];
842 	}
843 }
844 #endif /* UPCALL_TIMING */
845 
846 /*
847  * Delete an mfc entry
848  */
849 static int
850 del_m6fc(mfccp)
851 	struct mf6cctl *mfccp;
852 {
853 	struct sockaddr_in6 	origin;
854 	struct sockaddr_in6 	mcastgrp;
855 	struct mf6c 		*rt;
856 	struct mf6c	 	**nptr;
857 	u_long 		hash;
858 	int s;
859 
860 	origin = mfccp->mf6cc_origin;
861 	mcastgrp = mfccp->mf6cc_mcastgrp;
862 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
863 
864 #ifdef MRT6DEBUG
865 	if (mrt6debug & DEBUG_MFC)
866 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
867 		    ip6_sprintf(&origin.sin6_addr),
868 		    ip6_sprintf(&mcastgrp.sin6_addr));
869 #endif
870 
871 	s = splnet();
872 
873 	nptr = &mf6ctable[hash];
874 	while ((rt = *nptr) != NULL) {
875 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
876 				       &rt->mf6c_origin.sin6_addr) &&
877 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
878 				       &rt->mf6c_mcastgrp.sin6_addr) &&
879 		    rt->mf6c_stall == NULL)
880 			break;
881 
882 		nptr = &rt->mf6c_next;
883 	}
884 	if (rt == NULL) {
885 		splx(s);
886 		return EADDRNOTAVAIL;
887 	}
888 
889 	*nptr = rt->mf6c_next;
890 	free(rt, M_MRTABLE);
891 
892 	splx(s);
893 
894 	return 0;
895 }
896 
897 static int
898 socket_send(s, mm, src)
899 	struct socket *s;
900 	struct mbuf *mm;
901 	struct sockaddr_in6 *src;
902 {
903 	if (s) {
904 		if (sbappendaddr(&s->so_rcv,
905 				 (struct sockaddr *)src,
906 				 mm, (struct mbuf *)0) != 0) {
907 			sorwakeup(s);
908 			return 0;
909 		}
910 	}
911 	m_freem(mm);
912 	return -1;
913 }
914 
915 /*
916  * IPv6 multicast forwarding function. This function assumes that the packet
917  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
918  * pointed to by "ifp", and the packet is to be relayed to other networks
919  * that have members of the packet's destination IPv6 multicast group.
920  *
921  * The packet is returned unscathed to the caller, unless it is
922  * erroneous, in which case a non-zero return value tells the caller to
923  * discard it.
924  */
925 
926 int
927 ip6_mforward(ip6, ifp, m)
928 	struct ip6_hdr *ip6;
929 	struct ifnet *ifp;
930 	struct mbuf *m;
931 {
932 	struct mf6c *rt;
933 	struct mif6 *mifp;
934 	struct mbuf *mm;
935 	int s;
936 	mifi_t mifi;
937 	long time_second = time.tv_sec;
938 
939 #ifdef MRT6DEBUG
940 	if (mrt6debug & DEBUG_FORWARD)
941 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
942 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
943 		    ifp->if_index);
944 #endif
945 
946 	/*
947 	 * Don't forward a packet with Hop limit of zero or one,
948 	 * or a packet destined to a local-only group.
949 	 */
950 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
951 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
952 		return 0;
953 	ip6->ip6_hlim--;
954 
955 	/*
956 	 * Source address check: do not forward packets with unspecified
957 	 * source. It was discussed in July 2000, on ipngwg mailing list.
958 	 * This is rather more serious than unicast cases, because some
959 	 * MLD packets can be sent with the unspecified source address
960 	 * (although such packets must normally set 1 to the hop limit field).
961 	 */
962 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
963 		ip6stat.ip6s_cantforward++;
964 		if (ip6_log_time + ip6_log_interval < time_second) {
965 			ip6_log_time = time_second;
966 			log(LOG_DEBUG,
967 			    "cannot forward "
968 			    "from %s to %s nxt %d received on %s\n",
969 			    ip6_sprintf(&ip6->ip6_src),
970 			    ip6_sprintf(&ip6->ip6_dst),
971 			    ip6->ip6_nxt,
972 			    m->m_pkthdr.rcvif->if_xname);
973 		}
974 		return 0;
975 	}
976 
977 	/*
978 	 * Determine forwarding mifs from the forwarding cache table
979 	 */
980 	s = splnet();
981 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
982 
983 	/* Entry exists, so forward if necessary */
984 	if (rt) {
985 		splx(s);
986 		return (ip6_mdq(m, ifp, rt));
987 	} else {
988 		/*
989 		 * If we don't have a route for packet's origin,
990 		 * Make a copy of the packet &
991 		 * send message to routing daemon
992 		 */
993 
994 		struct mbuf *mb0;
995 		struct rtdetq *rte;
996 		u_long hash;
997 /*		int i, npkts;*/
998 #ifdef UPCALL_TIMING
999 		struct timeval tp;
1000 
1001 		GET_TIME(tp);
1002 #endif /* UPCALL_TIMING */
1003 
1004 		mrt6stat.mrt6s_no_route++;
1005 #ifdef MRT6DEBUG
1006 		if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
1007 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
1008 			    ip6_sprintf(&ip6->ip6_src),
1009 			    ip6_sprintf(&ip6->ip6_dst));
1010 #endif
1011 
1012 		/*
1013 		 * Allocate mbufs early so that we don't do extra work if we
1014 		 * are just going to fail anyway.
1015 		 */
1016 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1017 					      M_NOWAIT);
1018 		if (rte == NULL) {
1019 			splx(s);
1020 			return ENOBUFS;
1021 		}
1022 		mb0 = m_copy(m, 0, M_COPYALL);
1023 		/*
1024 		 * Pullup packet header if needed before storing it,
1025 		 * as other references may modify it in the meantime.
1026 		 */
1027 		if (mb0 &&
1028 		    (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1029 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1030 		if (mb0 == NULL) {
1031 			free(rte, M_MRTABLE);
1032 			splx(s);
1033 			return ENOBUFS;
1034 		}
1035 
1036 		/* is there an upcall waiting for this packet? */
1037 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1038 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1039 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1040 					       &rt->mf6c_origin.sin6_addr) &&
1041 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1042 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1043 			    (rt->mf6c_stall != NULL))
1044 				break;
1045 		}
1046 
1047 		if (rt == NULL) {
1048 			struct mrt6msg *im;
1049 #ifdef MRT6_OINIT
1050 			struct omrt6msg *oim;
1051 #endif
1052 
1053 			/* no upcall, so make a new entry */
1054 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1055 						  M_NOWAIT);
1056 			if (rt == NULL) {
1057 				free(rte, M_MRTABLE);
1058 				m_freem(mb0);
1059 				splx(s);
1060 				return ENOBUFS;
1061 			}
1062 			/*
1063 			 * Make a copy of the header to send to the user
1064 			 * level process
1065 			 */
1066 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1067 
1068 			if (mm == NULL) {
1069 				free(rte, M_MRTABLE);
1070 				m_freem(mb0);
1071 				free(rt, M_MRTABLE);
1072 				splx(s);
1073 				return ENOBUFS;
1074 			}
1075 
1076 			/*
1077 			 * Send message to routing daemon
1078 			 */
1079 			sin6.sin6_addr = ip6->ip6_src;
1080 
1081 			im = NULL;
1082 #ifdef MRT6_OINIT
1083 			oim = NULL;
1084 #endif
1085 			switch (ip6_mrouter_ver) {
1086 #ifdef MRT6_OINIT
1087 			case MRT6_OINIT:
1088 				oim = mtod(mm, struct omrt6msg *);
1089 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1090 				oim->im6_mbz = 0;
1091 				break;
1092 #endif
1093 			case MRT6_INIT:
1094 				im = mtod(mm, struct mrt6msg *);
1095 				im->im6_msgtype = MRT6MSG_NOCACHE;
1096 				im->im6_mbz = 0;
1097 				break;
1098 			default:
1099 				free(rte, M_MRTABLE);
1100 				m_freem(mb0);
1101 				free(rt, M_MRTABLE);
1102 				splx(s);
1103 				return EINVAL;
1104 			}
1105 
1106 #ifdef MRT6DEBUG
1107 			if (mrt6debug & DEBUG_FORWARD)
1108 				log(LOG_DEBUG,
1109 				    "getting the iif info in the kernel\n");
1110 #endif
1111 
1112 			for (mifp = mif6table, mifi = 0;
1113 			     mifi < nummifs && mifp->m6_ifp != ifp;
1114 			     mifp++, mifi++)
1115 				;
1116 
1117 			switch (ip6_mrouter_ver) {
1118 #ifdef MRT6_OINIT
1119 			case MRT6_OINIT:
1120 				oim->im6_mif = mifi;
1121 				break;
1122 #endif
1123 			case MRT6_INIT:
1124 				im->im6_mif = mifi;
1125 				break;
1126 			}
1127 
1128 			if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1129 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1130 				    "socket queue full\n");
1131 				mrt6stat.mrt6s_upq_sockfull++;
1132 				free(rte, M_MRTABLE);
1133 				m_freem(mb0);
1134 				free(rt, M_MRTABLE);
1135 				splx(s);
1136 				return ENOBUFS;
1137 			}
1138 
1139 			mrt6stat.mrt6s_upcalls++;
1140 
1141 			/* insert new entry at head of hash chain */
1142 			bzero(rt, sizeof(*rt));
1143 			rt->mf6c_origin.sin6_family = AF_INET6;
1144 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1145 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1146 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1147 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1148 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1149 			rt->mf6c_expire = UPCALL_EXPIRE;
1150 			nexpire[hash]++;
1151 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1152 
1153 			/* link into table */
1154 			rt->mf6c_next  = mf6ctable[hash];
1155 			mf6ctable[hash] = rt;
1156 			/* Add this entry to the end of the queue */
1157 			rt->mf6c_stall = rte;
1158 		} else {
1159 			/* determine if q has overflowed */
1160 			struct rtdetq **p;
1161 			int npkts = 0;
1162 
1163 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1164 				if (++npkts > MAX_UPQ6) {
1165 					mrt6stat.mrt6s_upq_ovflw++;
1166 					free(rte, M_MRTABLE);
1167 					m_freem(mb0);
1168 					splx(s);
1169 					return 0;
1170 				}
1171 
1172 			/* Add this entry to the end of the queue */
1173 			*p = rte;
1174 		}
1175 
1176 		rte->next = NULL;
1177 		rte->m = mb0;
1178 		rte->ifp = ifp;
1179 #ifdef UPCALL_TIMING
1180 		rte->t = tp;
1181 #endif /* UPCALL_TIMING */
1182 
1183 		splx(s);
1184 
1185 		return 0;
1186 	}
1187 }
1188 
1189 /*
1190  * Clean up cache entries if upcalls are not serviced
1191  * Call from the Slow Timeout mechanism, every half second.
1192  */
1193 static void
1194 expire_upcalls(unused)
1195 	void *unused;
1196 {
1197 	struct rtdetq *rte;
1198 	struct mf6c *mfc, **nptr;
1199 	int i;
1200 	int s;
1201 
1202 	s = splnet();
1203 
1204 	for (i = 0; i < MF6CTBLSIZ; i++) {
1205 		if (nexpire[i] == 0)
1206 			continue;
1207 		nptr = &mf6ctable[i];
1208 		while ((mfc = *nptr) != NULL) {
1209 			rte = mfc->mf6c_stall;
1210 			/*
1211 			 * Skip real cache entries
1212 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1213 			 * If it expires now
1214 			 */
1215 			if (rte != NULL &&
1216 			    mfc->mf6c_expire != 0 &&
1217 			    --mfc->mf6c_expire == 0) {
1218 #ifdef MRT6DEBUG
1219 				if (mrt6debug & DEBUG_EXPIRE)
1220 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1221 					    ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1222 					    ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1223 #endif
1224 				/*
1225 				 * drop all the packets
1226 				 * free the mbuf with the pkt, if, timing info
1227 				 */
1228 				do {
1229 					struct rtdetq *n = rte->next;
1230 					m_freem(rte->m);
1231 					free(rte, M_MRTABLE);
1232 					rte = n;
1233 				} while (rte != NULL);
1234 				mrt6stat.mrt6s_cache_cleanups++;
1235 				nexpire[i]--;
1236 
1237 				*nptr = mfc->mf6c_next;
1238 				free(mfc, M_MRTABLE);
1239 			} else {
1240 				nptr = &mfc->mf6c_next;
1241 			}
1242 		}
1243 	}
1244 	splx(s);
1245 	timeout_set(&expire_upcalls_ch, expire_upcalls, NULL);
1246 	timeout_add(&expire_upcalls_ch, EXPIRE_TIMEOUT);
1247 }
1248 
1249 /*
1250  * Packet forwarding routine once entry in the cache is made
1251  */
1252 static int
1253 ip6_mdq(m, ifp, rt)
1254 	struct mbuf *m;
1255 	struct ifnet *ifp;
1256 	struct mf6c *rt;
1257 {
1258 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1259 	mifi_t mifi, iif;
1260 	struct mif6 *mifp;
1261 	int plen = m->m_pkthdr.len;
1262 
1263 /*
1264  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1265  * input, they shouldn't get counted on output, so statistics keeping is
1266  * seperate.
1267  */
1268 
1269 #define MC6_SEND(ip6, mifp, m) do {				\
1270 		if ((mifp)->m6_flags & MIFF_REGISTER)		\
1271 		    register_send((ip6), (mifp), (m));		\
1272 		else						\
1273 		    phyint_send((ip6), (mifp), (m));		\
1274 } while (0)
1275 
1276 	/*
1277 	 * Don't forward if it didn't arrive from the parent mif
1278 	 * for its origin.
1279 	 */
1280 	mifi = rt->mf6c_parent;
1281 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1282 		/* came in the wrong interface */
1283 #ifdef MRT6DEBUG
1284 		if (mrt6debug & DEBUG_FORWARD)
1285 			log(LOG_DEBUG,
1286 			    "wrong if: ifid %d mifi %d mififid %x\n",
1287 			    ifp->if_index, mifi,
1288 			    mif6table[mifi].m6_ifp->if_index);
1289 #endif
1290 		mrt6stat.mrt6s_wrong_if++;
1291 		rt->mf6c_wrong_if++;
1292 		/*
1293 		 * If we are doing PIM processing, and we are forwarding
1294 		 * packets on this interface, send a message to the
1295 		 * routing daemon.
1296 		 */
1297 		/* have to make sure this is a valid mif */
1298 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1299 			if (pim6 && (m->m_flags & M_LOOP) == 0) {
1300 				/*
1301 				 * Check the M_LOOP flag to avoid an
1302 				 * unnecessary PIM assert.
1303 				 * XXX: M_LOOP is an ad-hoc hack...
1304 				 */
1305 				static struct sockaddr_in6 sin6 =
1306 				{ sizeof(sin6), AF_INET6 };
1307 
1308 				struct mbuf *mm;
1309 				struct mrt6msg *im;
1310 #ifdef MRT6_OINIT
1311 				struct omrt6msg *oim;
1312 #endif
1313 
1314 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1315 				if (mm &&
1316 				    (M_HASCL(mm) ||
1317 				     mm->m_len < sizeof(struct ip6_hdr)))
1318 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1319 				if (mm == NULL)
1320 					return ENOBUFS;
1321 
1322 #ifdef MRT6_OINIT
1323 				oim = NULL;
1324 #endif
1325 				im = NULL;
1326 				switch (ip6_mrouter_ver) {
1327 #ifdef MRT6_OINIT
1328 				case MRT6_OINIT:
1329 					oim = mtod(mm, struct omrt6msg *);
1330 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1331 					oim->im6_mbz = 0;
1332 					break;
1333 #endif
1334 				case MRT6_INIT:
1335 					im = mtod(mm, struct mrt6msg *);
1336 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1337 					im->im6_mbz = 0;
1338 					break;
1339 				default:
1340 					m_freem(mm);
1341 					return EINVAL;
1342 				}
1343 
1344 				for (mifp = mif6table, iif = 0;
1345 				     iif < nummifs && mifp &&
1346 					     mifp->m6_ifp != ifp;
1347 				     mifp++, iif++)
1348 					;
1349 
1350 				switch (ip6_mrouter_ver) {
1351 #ifdef MRT6_OINIT
1352 				case MRT6_OINIT:
1353 					oim->im6_mif = iif;
1354 					sin6.sin6_addr = oim->im6_src;
1355 					break;
1356 #endif
1357 				case MRT6_INIT:
1358 					im->im6_mif = iif;
1359 					sin6.sin6_addr = im->im6_src;
1360 					break;
1361 				}
1362 
1363 				mrt6stat.mrt6s_upcalls++;
1364 
1365 				if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1366 #ifdef MRT6DEBUG
1367 					if (mrt6debug)
1368 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1369 #endif
1370 					++mrt6stat.mrt6s_upq_sockfull;
1371 					return ENOBUFS;
1372 				}	/* if socket Q full */
1373 			}		/* if PIM */
1374 		return 0;
1375 	}			/* if wrong iif */
1376 
1377 	/* If I sourced this packet, it counts as output, else it was input. */
1378 	if (m->m_pkthdr.rcvif == NULL) {
1379 		/* XXX: is rcvif really NULL when output?? */
1380 		mif6table[mifi].m6_pkt_out++;
1381 		mif6table[mifi].m6_bytes_out += plen;
1382 	} else {
1383 		mif6table[mifi].m6_pkt_in++;
1384 		mif6table[mifi].m6_bytes_in += plen;
1385 	}
1386 	rt->mf6c_pkt_cnt++;
1387 	rt->mf6c_byte_cnt += plen;
1388 
1389 	/*
1390 	 * For each mif, forward a copy of the packet if there are group
1391 	 * members downstream on the interface.
1392 	 */
1393 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1394 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1395 			/*
1396 			 * check if the outgoing packet is going to break
1397 			 * a scope boundary.
1398 			 * XXX For packets through PIM register tunnel
1399 			 * interface, we believe a routing daemon.
1400 			 */
1401 			if ((mif6table[rt->mf6c_parent].m6_flags &
1402 			     MIFF_REGISTER) == 0 &&
1403 			    (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1404 			    (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1405 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1406 					      &ip6->ip6_dst) ||
1407 			     in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1408 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1409 					      &ip6->ip6_src))) {
1410 				ip6stat.ip6s_badscope++;
1411 				continue;
1412 			}
1413 
1414 			mifp->m6_pkt_out++;
1415 			mifp->m6_bytes_out += plen;
1416 			MC6_SEND(ip6, mifp, m);
1417 		}
1418 	return 0;
1419 }
1420 
1421 static void
1422 phyint_send(ip6, mifp, m)
1423     struct ip6_hdr *ip6;
1424     struct mif6 *mifp;
1425     struct mbuf *m;
1426 {
1427 	struct mbuf *mb_copy;
1428 	struct ifnet *ifp = mifp->m6_ifp;
1429 	int error = 0;
1430 	int s = splnet();
1431 	static struct route_in6 ro;
1432 	struct	in6_multi *in6m;
1433 	struct sockaddr_in6 *dst6;
1434 
1435 	/*
1436 	 * Make a new reference to the packet; make sure that
1437 	 * the IPv6 header is actually copied, not just referenced,
1438 	 * so that ip6_output() only scribbles on the copy.
1439 	 */
1440 	mb_copy = m_copy(m, 0, M_COPYALL);
1441 	if (mb_copy &&
1442 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1443 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1444 	if (mb_copy == NULL) {
1445 		splx(s);
1446 		return;
1447 	}
1448 	/* set MCAST flag to the outgoing packet */
1449 	mb_copy->m_flags |= M_MCAST;
1450 
1451 	/*
1452 	 * If we sourced the packet, call ip6_output since we may devide
1453 	 * the packet into fragments when the packet is too big for the
1454 	 * outgoing interface.
1455 	 * Otherwise, we can simply send the packet to the interface
1456 	 * sending queue.
1457 	 */
1458 	if (m->m_pkthdr.rcvif == NULL) {
1459 		struct ip6_moptions im6o;
1460 
1461 		im6o.im6o_multicast_ifp = ifp;
1462 		/* XXX: ip6_output will override ip6->ip6_hlim */
1463 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1464 		im6o.im6o_multicast_loop = 1;
1465 		error = ip6_output(mb_copy, NULL, &ro,
1466 				   IPV6_FORWARDING, &im6o, NULL);
1467 
1468 #ifdef MRT6DEBUG
1469 		if (mrt6debug & DEBUG_XMIT)
1470 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1471 			    mifp - mif6table, error);
1472 #endif
1473 		splx(s);
1474 		return;
1475 	}
1476 
1477 	/*
1478 	 * If we belong to the destination multicast group
1479 	 * on the outgoing interface, loop back a copy.
1480 	 */
1481 	dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
1482 	IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1483 	if (in6m != NULL) {
1484 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1485 		dst6->sin6_family = AF_INET6;
1486 		dst6->sin6_addr = ip6->ip6_dst;
1487 		ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst);
1488 	}
1489 	/*
1490 	 * Put the packet into the sending queue of the outgoing interface
1491 	 * if it would fit in the MTU of the interface.
1492 	 */
1493 	if (mb_copy->m_pkthdr.len < ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
1494 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1495 		dst6->sin6_family = AF_INET6;
1496 		dst6->sin6_addr = ip6->ip6_dst;
1497 		/*
1498 		 * We just call if_output instead of nd6_output here, since
1499 		 * we need no ND for a multicast forwarded packet...right?
1500 		 */
1501 		error = (*ifp->if_output)(ifp, mb_copy,
1502 		    (struct sockaddr *)&ro.ro_dst, NULL);
1503 #ifdef MRT6DEBUG
1504 		if (mrt6debug & DEBUG_XMIT)
1505 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1506 			    mifp - mif6table, error);
1507 #endif
1508 	} else {
1509 #ifdef MULTICAST_PMTUD
1510 		icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
1511 #else
1512 #ifdef MRT6DEBUG
1513 		if (mrt6debug & DEBUG_XMIT)
1514 			log(LOG_DEBUG,
1515 			    "phyint_send: packet too big on %s o %s g %s"
1516 			    " size %d(discarded)\n",
1517 			    ifp->if_xname,
1518 			    ip6_sprintf(&ip6->ip6_src),
1519 			    ip6_sprintf(&ip6->ip6_dst),
1520 			    mb_copy->m_pkthdr.len);
1521 #endif /* MRT6DEBUG */
1522 		m_freem(mb_copy); /* simply discard the packet */
1523 #endif
1524 	}
1525 
1526 	splx(s);
1527 }
1528 
1529 static int
1530 register_send(ip6, mif, m)
1531 	struct ip6_hdr *ip6;
1532 	struct mif6 *mif;
1533 	struct mbuf *m;
1534 {
1535 	struct mbuf *mm;
1536 	int i, len = m->m_pkthdr.len;
1537 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1538 	struct mrt6msg *im6;
1539 
1540 #ifdef MRT6DEBUG
1541 	if (mrt6debug)
1542 		log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1543 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1544 #endif
1545 	++pim6stat.pim6s_snd_registers;
1546 
1547 	/* Make a copy of the packet to send to the user level process */
1548 	MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1549 	if (mm == NULL)
1550 		return ENOBUFS;
1551 	mm->m_data += max_linkhdr;
1552 	mm->m_len = sizeof(struct ip6_hdr);
1553 
1554 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1555 		m_freem(mm);
1556 		return ENOBUFS;
1557 	}
1558 	i = MHLEN - M_LEADINGSPACE(mm);
1559 	if (i > len)
1560 		i = len;
1561 	mm = m_pullup(mm, i);
1562 	if (mm == NULL){
1563 		m_freem(mm);
1564 		return ENOBUFS;
1565 	}
1566 /* TODO: check it! */
1567 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1568 
1569 	/*
1570 	 * Send message to routing daemon
1571 	 */
1572 	sin6.sin6_addr = ip6->ip6_src;
1573 
1574 	im6 = mtod(mm, struct mrt6msg *);
1575 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1576 	im6->im6_mbz          = 0;
1577 
1578 	im6->im6_mif = mif - mif6table;
1579 
1580 	/* iif info is not given for reg. encap.n */
1581 	mrt6stat.mrt6s_upcalls++;
1582 
1583 	if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1584 #ifdef MRT6DEBUG
1585 		if (mrt6debug)
1586 			log(LOG_WARNING,
1587 			    "register_send: ip_mrouter socket queue full\n");
1588 #endif
1589 		++mrt6stat.mrt6s_upq_sockfull;
1590 		return ENOBUFS;
1591 	}
1592 	return 0;
1593 }
1594 
1595 /*
1596  * PIM sparse mode hook
1597  * Receives the pim control messages, and passes them up to the listening
1598  * socket, using rip6_input.
1599  * The only message processed is the REGISTER pim message; the pim header
1600  * is stripped off, and the inner packet is passed to register_mforward.
1601  */
1602 int
1603 pim6_input(mp, offp, proto)
1604 	struct mbuf **mp;
1605 	int *offp, proto;
1606 {
1607 	struct pim *pim; /* pointer to a pim struct */
1608 	struct ip6_hdr *ip6;
1609 	int pimlen;
1610 	struct mbuf *m = *mp;
1611 	int minlen;
1612 	int off = *offp;
1613 
1614 	++pim6stat.pim6s_rcv_total;
1615 
1616 	ip6 = mtod(m, struct ip6_hdr *);
1617 	pimlen = m->m_pkthdr.len - *offp;
1618 
1619 	/*
1620 	 * Validate lengths
1621 	 */
1622 	if (pimlen < PIM_MINLEN) {
1623 		++pim6stat.pim6s_rcv_tooshort;
1624 #ifdef MRT6DEBUG
1625 		if (mrt6debug & DEBUG_PIM)
1626 			log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1627 #endif
1628 		m_freem(m);
1629 		return(IPPROTO_DONE);
1630 	}
1631 
1632 	/*
1633 	 * if the packet is at least as big as a REGISTER, go ahead
1634 	 * and grab the PIM REGISTER header size, to avoid another
1635 	 * possible m_pullup() later.
1636 	 *
1637 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1638 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1639 	 */
1640 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1641 
1642 	/*
1643 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1644 	 * possibly the PIM REGISTER header
1645 	 */
1646 #ifndef PULLDOWN_TEST
1647 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1648 	/* adjust pointer */
1649 	ip6 = mtod(m, struct ip6_hdr *);
1650 
1651 	/* adjust mbuf to point to the PIM header */
1652 	pim = (struct pim *)((caddr_t)ip6 + off);
1653 #else
1654 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1655 	if (pim == NULL) {
1656 		pim6stat.pim6s_rcv_tooshort++;
1657 		return IPPROTO_DONE;
1658 	}
1659 #endif
1660 
1661 #define PIM6_CHECKSUM
1662 #ifdef PIM6_CHECKSUM
1663 	{
1664 		int cksumlen;
1665 
1666 		/*
1667 		 * Validate checksum.
1668 		 * If PIM REGISTER, exclude the data packet
1669 		 */
1670 		if (pim->pim_type == PIM_REGISTER)
1671 			cksumlen = PIM_MINLEN;
1672 		else
1673 			cksumlen = pimlen;
1674 
1675 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1676 			++pim6stat.pim6s_rcv_badsum;
1677 #ifdef MRT6DEBUG
1678 			if (mrt6debug & DEBUG_PIM)
1679 				log(LOG_DEBUG,
1680 				    "pim6_input: invalid checksum\n");
1681 #endif
1682 			m_freem(m);
1683 			return(IPPROTO_DONE);
1684 		}
1685 	}
1686 #endif /* PIM_CHECKSUM */
1687 
1688 	/* PIM version check */
1689 	if (pim->pim_ver != PIM_VERSION) {
1690 		++pim6stat.pim6s_rcv_badversion;
1691 #ifdef MRT6DEBUG
1692 		log(LOG_ERR,
1693 		    "pim6_input: incorrect version %d, expecting %d\n",
1694 		    pim->pim_ver, PIM_VERSION);
1695 #endif
1696 		m_freem(m);
1697 		return(IPPROTO_DONE);
1698 	}
1699 
1700 	if (pim->pim_type == PIM_REGISTER) {
1701 		/*
1702 		 * since this is a REGISTER, we'll make a copy of the register
1703 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1704 		 * routing daemon.
1705 		 */
1706 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1707 
1708 		struct mbuf *mcp;
1709 		struct ip6_hdr *eip6;
1710 		u_int32_t *reghdr;
1711 		int rc;
1712 
1713 		++pim6stat.pim6s_rcv_registers;
1714 
1715 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1716 #ifdef MRT6DEBUG
1717 			if (mrt6debug & DEBUG_PIM)
1718 				log(LOG_DEBUG,
1719 				    "pim6_input: register mif not set: %d\n",
1720 				    reg_mif_num);
1721 #endif
1722 			m_freem(m);
1723 			return(IPPROTO_DONE);
1724 		}
1725 
1726 		reghdr = (u_int32_t *)(pim + 1);
1727 
1728 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1729 			goto pim6_input_to_daemon;
1730 
1731 		/*
1732 		 * Validate length
1733 		 */
1734 		if (pimlen < PIM6_REG_MINLEN) {
1735 			++pim6stat.pim6s_rcv_tooshort;
1736 			++pim6stat.pim6s_rcv_badregisters;
1737 #ifdef MRT6DEBUG
1738 			log(LOG_ERR,
1739 			    "pim6_input: register packet size too "
1740 			    "small %d from %s\n",
1741 			    pimlen, ip6_sprintf(&ip6->ip6_src));
1742 #endif
1743 			m_freem(m);
1744 			return(IPPROTO_DONE);
1745 		}
1746 
1747 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1748 #ifdef MRT6DEBUG
1749 		if (mrt6debug & DEBUG_PIM)
1750 			log(LOG_DEBUG,
1751 			    "pim6_input[register], eip6: %s -> %s, "
1752 			    "eip6 plen %d\n",
1753 			    ip6_sprintf(&eip6->ip6_src),
1754 			    ip6_sprintf(&eip6->ip6_dst),
1755 			    ntohs(eip6->ip6_plen));
1756 #endif
1757 
1758 		/* verify the version number of the inner packet */
1759 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1760 			++pim6stat.pim6s_rcv_badregisters;
1761 #ifdef MRT6DEBUG
1762 			log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1763 			    "of the inner packet\n",
1764 			    (eip6->ip6_vfc & IPV6_VERSION));
1765 #endif
1766 			m_freem(m);
1767 			return(IPPROTO_NONE);
1768 		}
1769 
1770 		/* verify the inner packet is destined to a mcast group */
1771 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1772 			++pim6stat.pim6s_rcv_badregisters;
1773 #ifdef MRT6DEBUG
1774 			if (mrt6debug & DEBUG_PIM)
1775 				log(LOG_DEBUG,
1776 				    "pim6_input: inner packet of register "
1777 				    "is not multicast %s\n",
1778 				    ip6_sprintf(&eip6->ip6_dst));
1779 #endif
1780 			m_freem(m);
1781 			return(IPPROTO_DONE);
1782 		}
1783 
1784 		/*
1785 		 * make a copy of the whole header to pass to the daemon later.
1786 		 */
1787 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1788 		if (mcp == NULL) {
1789 #ifdef MRT6DEBUG
1790 			log(LOG_ERR,
1791 			    "pim6_input: pim register: "
1792 			    "could not copy register head\n");
1793 #endif
1794 			m_freem(m);
1795 			return(IPPROTO_DONE);
1796 		}
1797 
1798 		/*
1799 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1800 		 */
1801 		m_adj(m, off + PIM_MINLEN);
1802 #ifdef MRT6DEBUG
1803 		if (mrt6debug & DEBUG_PIM) {
1804 			log(LOG_DEBUG,
1805 			    "pim6_input: forwarding decapsulated register: "
1806 			    "src %s, dst %s, mif %d\n",
1807 			    ip6_sprintf(&eip6->ip6_src),
1808 			    ip6_sprintf(&eip6->ip6_dst),
1809 			    reg_mif_num);
1810 		}
1811 #endif
1812 
1813  		rc = looutput(mif6table[reg_mif_num].m6_ifp, m,
1814 			      (struct sockaddr *) &dst,
1815 			      (struct rtentry *) NULL);
1816 
1817 		/* prepare the register head to send to the mrouting daemon */
1818 		m = mcp;
1819 	}
1820 
1821 	/*
1822 	 * Pass the PIM message up to the daemon; if it is a register message
1823 	 * pass the 'head' only up to the daemon. This includes the
1824 	 * encapsulator ip6 header, pim header, register header and the
1825 	 * encapsulated ip6 header.
1826 	 */
1827   pim6_input_to_daemon:
1828 	rip6_input(&m, offp, proto);
1829 	return(IPPROTO_DONE);
1830 }
1831