xref: /freebsd/sys/netinet6/ip6_mroute.c (revision f9218d3d4fd34f082473b3a021c6d4d109fb47cf)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 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/callout.h>
53 #include <sys/errno.h>
54 #include <sys/kernel.h>
55 #include <sys/lock.h>
56 #include <sys/malloc.h>
57 #include <sys/mbuf.h>
58 #include <sys/protosw.h>
59 #include <sys/signalvar.h>
60 #include <sys/socket.h>
61 #include <sys/socketvar.h>
62 #include <sys/sockio.h>
63 #include <sys/sx.h>
64 #include <sys/syslog.h>
65 #include <sys/systm.h>
66 #include <sys/time.h>
67 
68 #include <net/if.h>
69 #include <net/raw_cb.h>
70 #include <net/route.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 #include <net/net_osdep.h>
82 
83 static MALLOC_DEFINE(M_MRTABLE, "mf6c", "multicast forwarding cache entry");
84 
85 #define M_HASCL(m) ((m)->m_flags & M_EXT)
86 
87 static int ip6_mdq __P((struct mbuf *, struct ifnet *, struct mf6c *));
88 static void phyint_send __P((struct ip6_hdr *, struct mif6 *, struct mbuf *));
89 
90 static int set_pim6 __P((int *));
91 static int socket_send __P((struct socket *, struct mbuf *,
92 			    struct sockaddr_in6 *));
93 static int register_send __P((struct ip6_hdr *, struct mif6 *,
94 			      struct mbuf *));
95 
96 /*
97  * Globals.  All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static,
98  * except for netstat or debugging purposes.
99  */
100 struct socket  *ip6_mrouter = NULL;
101 int		ip6_mrouter_ver = 0;
102 int		ip6_mrtproto = IPPROTO_PIM;    /* for netstat only */
103 struct mrt6stat	mrt6stat;
104 
105 #define NO_RTE_FOUND 	0x1
106 #define RTE_FOUND	0x2
107 
108 struct mf6c	*mf6ctable[MF6CTBLSIZ];
109 u_char		n6expire[MF6CTBLSIZ];
110 static struct mif6 mif6table[MAXMIFS];
111 #ifdef MRT6DEBUG
112 u_int		mrt6debug = 0;	  /* debug level 	*/
113 #define		DEBUG_MFC	0x02
114 #define		DEBUG_FORWARD	0x04
115 #define		DEBUG_EXPIRE	0x08
116 #define		DEBUG_XMIT	0x10
117 #define         DEBUG_REG       0x20
118 #define         DEBUG_PIM       0x40
119 #endif
120 
121 static void	expire_upcalls __P((void *));
122 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
123 #define		UPCALL_EXPIRE	6		/* number of timeouts */
124 
125 #ifdef INET
126 #ifdef MROUTING
127 extern struct socket *ip_mrouter;
128 #endif
129 #endif
130 
131 /*
132  * 'Interfaces' associated with decapsulator (so we can tell
133  * packets that went through it from ones that get reflected
134  * by a broken gateway).  These interfaces are never linked into
135  * the system ifnet list & no routes point to them.  I.e., packets
136  * can't be sent this way.  They only exist as a placeholder for
137  * multicast source verification.
138  */
139 struct ifnet multicast_register_if;
140 
141 #define ENCAP_HOPS 64
142 
143 /*
144  * Private variables.
145  */
146 static mifi_t nummifs = 0;
147 static mifi_t reg_mif_num = (mifi_t)-1;
148 
149 static struct pim6stat pim6stat;
150 static int pim6;
151 
152 /*
153  * Hash function for a source, group entry
154  */
155 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
156 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
157 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
158 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
159 
160 /*
161  * Find a route for a given origin IPv6 address and Multicast group address.
162  * Quality of service parameter to be added in the future!!!
163  */
164 
165 #define MF6CFIND(o, g, rt) do { \
166 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
167 	rt = NULL; \
168 	mrt6stat.mrt6s_mfc_lookups++; \
169 	while (_rt) { \
170 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
171 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
172 		    (_rt->mf6c_stall == NULL)) { \
173 			rt = _rt; \
174 			break; \
175 		} \
176 		_rt = _rt->mf6c_next; \
177 	} \
178 	if (rt == NULL) { \
179 		mrt6stat.mrt6s_mfc_misses++; \
180 	} \
181 } while (0)
182 
183 /*
184  * Macros to compute elapsed time efficiently
185  * Borrowed from Van Jacobson's scheduling code
186  */
187 #define TV_DELTA(a, b, delta) do { \
188 	    int xxs; \
189 		\
190 	    delta = (a).tv_usec - (b).tv_usec; \
191 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
192 	       switch (xxs) { \
193 		      case 2: \
194 			  delta += 1000000; \
195 			      /* fall through */ \
196 		      case 1: \
197 			  delta += 1000000; \
198 			  break; \
199 		      default: \
200 			  delta += (1000000 * xxs); \
201 	       } \
202 	    } \
203 } while (0)
204 
205 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
206 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
207 
208 #ifdef UPCALL_TIMING
209 #define UPCALL_MAX	50
210 u_long upcall_data[UPCALL_MAX + 1];
211 static void collate();
212 #endif /* UPCALL_TIMING */
213 
214 static int get_sg_cnt __P((struct sioc_sg_req6 *));
215 static int get_mif6_cnt __P((struct sioc_mif_req6 *));
216 static int ip6_mrouter_init __P((struct socket *, struct mbuf *, int));
217 static int add_m6if __P((struct mif6ctl *));
218 static int del_m6if __P((mifi_t *));
219 static int add_m6fc __P((struct mf6cctl *));
220 static int del_m6fc __P((struct mf6cctl *));
221 
222 static struct callout expire_upcalls_ch;
223 
224 /*
225  * Handle MRT setsockopt commands to modify the multicast routing tables.
226  */
227 int
228 ip6_mrouter_set(so, sopt)
229 	struct socket *so;
230 	struct sockopt *sopt;
231 {
232 	int	error = 0;
233 	struct mbuf *m;
234 
235 	if (so != ip6_mrouter && sopt->sopt_name != MRT6_INIT)
236 		return (EACCES);
237 
238 	if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
239 		return (error);
240 	if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
241 		return (error);
242 
243 	switch (sopt->sopt_name) {
244 	case MRT6_INIT:
245 #ifdef MRT6_OINIT
246 	case MRT6_OINIT:
247 #endif
248 		error = ip6_mrouter_init(so, m, sopt->sopt_name);
249 		break;
250 	case MRT6_DONE:
251 		error = ip6_mrouter_done();
252 		break;
253 	case MRT6_ADD_MIF:
254 		error = add_m6if(mtod(m, struct mif6ctl *));
255 		break;
256 	case MRT6_DEL_MIF:
257 		error = del_m6if(mtod(m, mifi_t *));
258 		break;
259 	case MRT6_ADD_MFC:
260 		error = add_m6fc(mtod(m, struct mf6cctl *));
261 		break;
262 	case MRT6_DEL_MFC:
263 		error = del_m6fc(mtod(m, struct mf6cctl *));
264 		break;
265 	case MRT6_PIM:
266 		error = set_pim6(mtod(m, int *));
267 		break;
268 	default:
269 		error = EOPNOTSUPP;
270 		break;
271 	}
272 
273 	(void)m_freem(m);
274 	return(error);
275 }
276 
277 /*
278  * Handle MRT getsockopt commands
279  */
280 int
281 ip6_mrouter_get(so, sopt)
282 	struct socket *so;
283 	struct sockopt *sopt;
284 {
285 	int error = 0;
286 
287 	if (so != ip6_mrouter) return EACCES;
288 
289 	switch (sopt->sopt_name) {
290 		case MRT6_PIM:
291 			error = sooptcopyout(sopt, &pim6, sizeof(pim6));
292 			break;
293 	}
294 	return (error);
295 }
296 
297 /*
298  * Handle ioctl commands to obtain information from the cache
299  */
300 int
301 mrt6_ioctl(cmd, data)
302 	int cmd;
303 	caddr_t data;
304 {
305 	int error = 0;
306 
307 	switch (cmd) {
308 	case SIOCGETSGCNT_IN6:
309 		return(get_sg_cnt((struct sioc_sg_req6 *)data));
310 		break;		/* for safety */
311 	case SIOCGETMIFCNT_IN6:
312 		return(get_mif6_cnt((struct sioc_mif_req6 *)data));
313 		break;		/* for safety */
314 	default:
315 		return (EINVAL);
316 		break;
317 	}
318 	return error;
319 }
320 
321 /*
322  * returns the packet, byte, rpf-failure count for the source group provided
323  */
324 static int
325 get_sg_cnt(req)
326 	struct sioc_sg_req6 *req;
327 {
328 	struct mf6c *rt;
329 	int s;
330 
331 	s = splnet();
332 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
333 	splx(s);
334 	if (rt != NULL) {
335 		req->pktcnt = rt->mf6c_pkt_cnt;
336 		req->bytecnt = rt->mf6c_byte_cnt;
337 		req->wrong_if = rt->mf6c_wrong_if;
338 	} else
339 		return(ESRCH);
340 #if 0
341 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
342 #endif
343 
344 	return 0;
345 }
346 
347 /*
348  * returns the input and output packet and byte counts on the mif provided
349  */
350 static int
351 get_mif6_cnt(req)
352 	struct sioc_mif_req6 *req;
353 {
354 	mifi_t mifi = req->mifi;
355 
356 	if (mifi >= nummifs)
357 		return EINVAL;
358 
359 	req->icount = mif6table[mifi].m6_pkt_in;
360 	req->ocount = mif6table[mifi].m6_pkt_out;
361 	req->ibytes = mif6table[mifi].m6_bytes_in;
362 	req->obytes = mif6table[mifi].m6_bytes_out;
363 
364 	return 0;
365 }
366 
367 static int
368 set_pim6(i)
369 	int *i;
370 {
371 	if ((*i != 1) && (*i != 0))
372 		return EINVAL;
373 
374 	pim6 = *i;
375 
376 	return 0;
377 }
378 
379 /*
380  * Enable multicast routing
381  */
382 static int
383 ip6_mrouter_init(so, m, cmd)
384 	struct socket *so;
385 	struct mbuf *m;
386 	int cmd;
387 {
388 	int *v;
389 
390 #ifdef MRT6DEBUG
391 	if (mrt6debug)
392 		log(LOG_DEBUG,
393 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
394 		    so->so_type, so->so_proto->pr_protocol);
395 #endif
396 
397 	if (so->so_type != SOCK_RAW ||
398 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
399 		return EOPNOTSUPP;
400 
401 	if (!m || (m->m_len != sizeof(int *)))
402 		return ENOPROTOOPT;
403 
404 	v = mtod(m, int *);
405 	if (*v != 1)
406 		return ENOPROTOOPT;
407 
408 	if (ip6_mrouter != NULL) return EADDRINUSE;
409 
410 	ip6_mrouter = so;
411 	ip6_mrouter_ver = cmd;
412 
413 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
414 	bzero((caddr_t)n6expire, sizeof(n6expire));
415 
416 	pim6 = 0;/* used for stubbing out/in pim stuff */
417 
418 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
419 	    expire_upcalls, NULL);
420 
421 #ifdef MRT6DEBUG
422 	if (mrt6debug)
423 		log(LOG_DEBUG, "ip6_mrouter_init\n");
424 #endif
425 
426 	return 0;
427 }
428 
429 /*
430  * Disable multicast routing
431  */
432 int
433 ip6_mrouter_done()
434 {
435 	mifi_t mifi;
436 	int i;
437 	struct ifnet *ifp;
438 	struct in6_ifreq ifr;
439 	struct mf6c *rt;
440 	struct rtdetq *rte;
441 	int s;
442 
443 	s = splnet();
444 
445 	/*
446 	 * For each phyint in use, disable promiscuous reception of all IPv6
447 	 * multicasts.
448 	 */
449 #ifdef INET
450 #ifdef MROUTING
451 	/*
452 	 * If there is still IPv4 multicast routing daemon,
453 	 * we remain interfaces to receive all muliticasted packets.
454 	 * XXX: there may be an interface in which the IPv4 multicast
455 	 * daemon is not interested...
456 	 */
457 	if (!ip_mrouter)
458 #endif
459 #endif
460 	{
461 		for (mifi = 0; mifi < nummifs; mifi++) {
462 			if (mif6table[mifi].m6_ifp &&
463 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
464 				ifr.ifr_addr.sin6_family = AF_INET6;
465 				ifr.ifr_addr.sin6_addr= in6addr_any;
466 				ifp = mif6table[mifi].m6_ifp;
467 				(*ifp->if_ioctl)(ifp, SIOCDELMULTI,
468 						 (caddr_t)&ifr);
469 			}
470 		}
471 	}
472 #ifdef notyet
473 	bzero((caddr_t)qtable, sizeof(qtable));
474 	bzero((caddr_t)tbftable, sizeof(tbftable));
475 #endif
476 	bzero((caddr_t)mif6table, sizeof(mif6table));
477 	nummifs = 0;
478 
479 	pim6 = 0; /* used to stub out/in pim specific code */
480 
481 	callout_stop(&expire_upcalls_ch);
482 
483 	/*
484 	 * Free all multicast forwarding cache entries.
485 	 */
486 	for (i = 0; i < MF6CTBLSIZ; i++) {
487 		rt = mf6ctable[i];
488 		while (rt) {
489 			struct mf6c *frt;
490 
491 			for (rte = rt->mf6c_stall; rte != NULL; ) {
492 				struct rtdetq *n = rte->next;
493 
494 				m_free(rte->m);
495 				free(rte, M_MRTABLE);
496 				rte = n;
497 			}
498 			frt = rt;
499 			rt = rt->mf6c_next;
500 			free(frt, M_MRTABLE);
501 		}
502 	}
503 
504 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
505 
506 	/*
507 	 * Reset de-encapsulation cache
508 	 */
509 	reg_mif_num = -1;
510 
511 	ip6_mrouter = NULL;
512 	ip6_mrouter_ver = 0;
513 
514 	splx(s);
515 
516 #ifdef MRT6DEBUG
517 	if (mrt6debug)
518 		log(LOG_DEBUG, "ip6_mrouter_done\n");
519 #endif
520 
521 	return 0;
522 }
523 
524 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
525 
526 /*
527  * Add a mif to the mif table
528  */
529 static int
530 add_m6if(mifcp)
531 	struct mif6ctl *mifcp;
532 {
533 	struct mif6 *mifp;
534 	struct ifnet *ifp;
535 	int error, s;
536 #ifdef notyet
537 	struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
538 #endif
539 
540 	if (mifcp->mif6c_mifi >= MAXMIFS)
541 		return EINVAL;
542 	mifp = mif6table + mifcp->mif6c_mifi;
543 	if (mifp->m6_ifp)
544 		return EADDRINUSE; /* XXX: is it appropriate? */
545 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
546 		return ENXIO;
547 	ifp = ifnet_byindex(mifcp->mif6c_pifi);
548 
549 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
550 		if (reg_mif_num == (mifi_t)-1) {
551 			multicast_register_if.if_name = "register_mif";
552 			multicast_register_if.if_flags |= IFF_LOOPBACK;
553 			multicast_register_if.if_index = mifcp->mif6c_mifi;
554 			reg_mif_num = mifcp->mif6c_mifi;
555 		}
556 
557 		ifp = &multicast_register_if;
558 
559 	} /* if REGISTER */
560 	else {
561 		/* Make sure the interface supports multicast */
562 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
563 			return EOPNOTSUPP;
564 
565 		s = splnet();
566 		error = if_allmulti(ifp, 1);
567 		splx(s);
568 		if (error)
569 			return error;
570 	}
571 
572 	s = splnet();
573 	mifp->m6_flags     = mifcp->mif6c_flags;
574 	mifp->m6_ifp       = ifp;
575 #ifdef notyet
576 	/* scaling up here allows division by 1024 in critical code */
577 	mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
578 #endif
579 	/* initialize per mif pkt counters */
580 	mifp->m6_pkt_in    = 0;
581 	mifp->m6_pkt_out   = 0;
582 	mifp->m6_bytes_in  = 0;
583 	mifp->m6_bytes_out = 0;
584 	splx(s);
585 
586 	/* Adjust nummifs up if the mifi is higher than nummifs */
587 	if (nummifs <= mifcp->mif6c_mifi)
588 		nummifs = mifcp->mif6c_mifi + 1;
589 
590 #ifdef MRT6DEBUG
591 	if (mrt6debug)
592 		log(LOG_DEBUG,
593 		    "add_mif #%d, phyint %s%d\n",
594 		    mifcp->mif6c_mifi,
595 		    ifp->if_name, ifp->if_unit);
596 #endif
597 
598 	return 0;
599 }
600 
601 /*
602  * Delete a mif from the mif table
603  */
604 static int
605 del_m6if(mifip)
606 	mifi_t *mifip;
607 {
608 	struct mif6 *mifp = mif6table + *mifip;
609 	mifi_t mifi;
610 	struct ifnet *ifp;
611 	int s;
612 
613 	if (*mifip >= nummifs)
614 		return EINVAL;
615 	if (mifp->m6_ifp == NULL)
616 		return EINVAL;
617 
618 	s = splnet();
619 
620 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
621 		/*
622 		 * XXX: what if there is yet IPv4 multicast daemon
623 		 *      using the interface?
624 		 */
625 		ifp = mifp->m6_ifp;
626 
627 		if_allmulti(ifp, 0);
628 	}
629 
630 #ifdef notyet
631 	bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
632 	bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
633 #endif
634 	bzero((caddr_t)mifp, sizeof (*mifp));
635 
636 	/* Adjust nummifs down */
637 	for (mifi = nummifs; mifi > 0; mifi--)
638 		if (mif6table[mifi - 1].m6_ifp)
639 			break;
640 	nummifs = mifi;
641 
642 	splx(s);
643 
644 #ifdef MRT6DEBUG
645 	if (mrt6debug)
646 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
647 #endif
648 
649 	return 0;
650 }
651 
652 /*
653  * Add an mfc entry
654  */
655 static int
656 add_m6fc(mfccp)
657 	struct mf6cctl *mfccp;
658 {
659 	struct mf6c *rt;
660 	u_long hash;
661 	struct rtdetq *rte;
662 	u_short nstl;
663 	int s;
664 
665 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
666 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
667 
668 	/* If an entry already exists, just update the fields */
669 	if (rt) {
670 #ifdef MRT6DEBUG
671 		if (mrt6debug & DEBUG_MFC)
672 			log(LOG_DEBUG,
673 			    "add_m6fc no upcall h %d o %s g %s p %x\n",
674 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
675 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
676 			    mfccp->mf6cc_parent);
677 #endif
678 
679 		s = splnet();
680 		rt->mf6c_parent = mfccp->mf6cc_parent;
681 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
682 		splx(s);
683 		return 0;
684 	}
685 
686 	/*
687 	 * Find the entry for which the upcall was made and update
688 	 */
689 	s = splnet();
690 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
691 			mfccp->mf6cc_mcastgrp.sin6_addr);
692 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
693 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
694 				       &mfccp->mf6cc_origin.sin6_addr) &&
695 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
696 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
697 		    (rt->mf6c_stall != NULL)) {
698 
699 			if (nstl++)
700 				log(LOG_ERR,
701 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
702 				    "multiple kernel entries",
703 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
704 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
705 				    mfccp->mf6cc_parent, rt->mf6c_stall);
706 
707 #ifdef MRT6DEBUG
708 			if (mrt6debug & DEBUG_MFC)
709 				log(LOG_DEBUG,
710 				    "add_m6fc o %s g %s p %x dbg %x\n",
711 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
712 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
713 				    mfccp->mf6cc_parent, rt->mf6c_stall);
714 #endif
715 
716 			rt->mf6c_origin     = mfccp->mf6cc_origin;
717 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
718 			rt->mf6c_parent     = mfccp->mf6cc_parent;
719 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
720 			/* initialize pkt counters per src-grp */
721 			rt->mf6c_pkt_cnt    = 0;
722 			rt->mf6c_byte_cnt   = 0;
723 			rt->mf6c_wrong_if   = 0;
724 
725 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
726 			n6expire[hash]--;
727 
728 			/* free packets Qed at the end of this entry */
729 			for (rte = rt->mf6c_stall; rte != NULL; ) {
730 				struct rtdetq *n = rte->next;
731 				ip6_mdq(rte->m, rte->ifp, rt);
732 				m_freem(rte->m);
733 #ifdef UPCALL_TIMING
734 				collate(&(rte->t));
735 #endif /* UPCALL_TIMING */
736 				free(rte, M_MRTABLE);
737 				rte = n;
738 			}
739 			rt->mf6c_stall = NULL;
740 		}
741 	}
742 
743 	/*
744 	 * It is possible that an entry is being inserted without an upcall
745 	 */
746 	if (nstl == 0) {
747 #ifdef MRT6DEBUG
748 		if (mrt6debug & DEBUG_MFC)
749 			log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n",
750 			    hash,
751 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
752 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
753 			    mfccp->mf6cc_parent);
754 #endif
755 
756 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
757 
758 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
759 					       &mfccp->mf6cc_origin.sin6_addr)&&
760 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
761 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
762 
763 				rt->mf6c_origin     = mfccp->mf6cc_origin;
764 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
765 				rt->mf6c_parent     = mfccp->mf6cc_parent;
766 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
767 				/* initialize pkt counters per src-grp */
768 				rt->mf6c_pkt_cnt    = 0;
769 				rt->mf6c_byte_cnt   = 0;
770 				rt->mf6c_wrong_if   = 0;
771 
772 				if (rt->mf6c_expire)
773 					n6expire[hash]--;
774 				rt->mf6c_expire	   = 0;
775 			}
776 		}
777 		if (rt == NULL) {
778 			/* no upcall, so make a new entry */
779 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
780 						  M_NOWAIT);
781 			if (rt == NULL) {
782 				splx(s);
783 				return ENOBUFS;
784 			}
785 
786 			/* insert new entry at head of hash chain */
787 			rt->mf6c_origin     = mfccp->mf6cc_origin;
788 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
789 			rt->mf6c_parent     = mfccp->mf6cc_parent;
790 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
791 			/* initialize pkt counters per src-grp */
792 			rt->mf6c_pkt_cnt    = 0;
793 			rt->mf6c_byte_cnt   = 0;
794 			rt->mf6c_wrong_if   = 0;
795 			rt->mf6c_expire     = 0;
796 			rt->mf6c_stall = NULL;
797 
798 			/* link into table */
799 			rt->mf6c_next  = mf6ctable[hash];
800 			mf6ctable[hash] = rt;
801 		}
802 	}
803 	splx(s);
804 	return 0;
805 }
806 
807 #ifdef UPCALL_TIMING
808 /*
809  * collect delay statistics on the upcalls
810  */
811 static void
812 collate(t)
813 	struct timeval *t;
814 {
815 	u_long d;
816 	struct timeval tp;
817 	u_long delta;
818 
819 	GET_TIME(tp);
820 
821 	if (TV_LT(*t, tp))
822 	{
823 		TV_DELTA(tp, *t, delta);
824 
825 		d = delta >> 10;
826 		if (d > UPCALL_MAX)
827 			d = UPCALL_MAX;
828 
829 		++upcall_data[d];
830 	}
831 }
832 #endif /* UPCALL_TIMING */
833 
834 /*
835  * Delete an mfc entry
836  */
837 static int
838 del_m6fc(mfccp)
839 	struct mf6cctl *mfccp;
840 {
841 	struct sockaddr_in6 	origin;
842 	struct sockaddr_in6 	mcastgrp;
843 	struct mf6c 		*rt;
844 	struct mf6c	 	**nptr;
845 	u_long 		hash;
846 	int s;
847 
848 	origin = mfccp->mf6cc_origin;
849 	mcastgrp = mfccp->mf6cc_mcastgrp;
850 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
851 
852 #ifdef MRT6DEBUG
853 	if (mrt6debug & DEBUG_MFC)
854 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
855 		    ip6_sprintf(&origin.sin6_addr),
856 		    ip6_sprintf(&mcastgrp.sin6_addr));
857 #endif
858 
859 	s = splnet();
860 
861 	nptr = &mf6ctable[hash];
862 	while ((rt = *nptr) != NULL) {
863 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
864 				       &rt->mf6c_origin.sin6_addr) &&
865 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
866 				       &rt->mf6c_mcastgrp.sin6_addr) &&
867 		    rt->mf6c_stall == NULL)
868 			break;
869 
870 		nptr = &rt->mf6c_next;
871 	}
872 	if (rt == NULL) {
873 		splx(s);
874 		return EADDRNOTAVAIL;
875 	}
876 
877 	*nptr = rt->mf6c_next;
878 	free(rt, M_MRTABLE);
879 
880 	splx(s);
881 
882 	return 0;
883 }
884 
885 static int
886 socket_send(s, mm, src)
887 	struct socket *s;
888 	struct mbuf *mm;
889 	struct sockaddr_in6 *src;
890 {
891 	if (s) {
892 		if (sbappendaddr(&s->so_rcv,
893 				 (struct sockaddr *)src,
894 				 mm, (struct mbuf *)0) != 0) {
895 			sorwakeup(s);
896 			return 0;
897 		}
898 	}
899 	m_freem(mm);
900 	return -1;
901 }
902 
903 /*
904  * IPv6 multicast forwarding function. This function assumes that the packet
905  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
906  * pointed to by "ifp", and the packet is to be relayed to other networks
907  * that have members of the packet's destination IPv6 multicast group.
908  *
909  * The packet is returned unscathed to the caller, unless it is
910  * erroneous, in which case a non-zero return value tells the caller to
911  * discard it.
912  */
913 
914 int
915 ip6_mforward(ip6, ifp, m)
916 	struct ip6_hdr *ip6;
917 	struct ifnet *ifp;
918 	struct mbuf *m;
919 {
920 	struct mf6c *rt;
921 	struct mif6 *mifp;
922 	struct mbuf *mm;
923 	int s;
924 	mifi_t mifi;
925 
926 #ifdef MRT6DEBUG
927 	if (mrt6debug & DEBUG_FORWARD)
928 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
929 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
930 		    ifp->if_index);
931 #endif
932 
933 	/*
934 	 * Don't forward a packet with Hop limit of zero or one,
935 	 * or a packet destined to a local-only group.
936 	 */
937 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
938 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
939 		return 0;
940 	ip6->ip6_hlim--;
941 
942 	/*
943 	 * Source address check: do not forward packets with unspecified
944 	 * source. It was discussed in July 2000, on ipngwg mailing list.
945 	 * This is rather more serious than unicast cases, because some
946 	 * MLD packets can be sent with the unspecified source address
947 	 * (although such packets must normally set 1 to the hop limit field).
948 	 */
949 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
950 		ip6stat.ip6s_cantforward++;
951 		if (ip6_log_time + ip6_log_interval < time_second) {
952 			ip6_log_time = time_second;
953 			log(LOG_DEBUG,
954 			    "cannot forward "
955 			    "from %s to %s nxt %d received on %s\n",
956 			    ip6_sprintf(&ip6->ip6_src),
957 			    ip6_sprintf(&ip6->ip6_dst),
958 			    ip6->ip6_nxt,
959 			    if_name(m->m_pkthdr.rcvif));
960 		}
961 		return 0;
962 	}
963 
964 	/*
965 	 * Determine forwarding mifs from the forwarding cache table
966 	 */
967 	s = splnet();
968 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
969 
970 	/* Entry exists, so forward if necessary */
971 	if (rt) {
972 		splx(s);
973 		return (ip6_mdq(m, ifp, rt));
974 	} else {
975 		/*
976 		 * If we don't have a route for packet's origin,
977 		 * Make a copy of the packet &
978 		 * send message to routing daemon
979 		 */
980 
981 		struct mbuf *mb0;
982 		struct rtdetq *rte;
983 		u_long hash;
984 /*		int i, npkts;*/
985 #ifdef UPCALL_TIMING
986 		struct timeval tp;
987 
988 		GET_TIME(tp);
989 #endif /* UPCALL_TIMING */
990 
991 		mrt6stat.mrt6s_no_route++;
992 #ifdef MRT6DEBUG
993 		if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
994 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
995 			    ip6_sprintf(&ip6->ip6_src),
996 			    ip6_sprintf(&ip6->ip6_dst));
997 #endif
998 
999 		/*
1000 		 * Allocate mbufs early so that we don't do extra work if we
1001 		 * are just going to fail anyway.
1002 		 */
1003 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1004 					      M_NOWAIT);
1005 		if (rte == NULL) {
1006 			splx(s);
1007 			return ENOBUFS;
1008 		}
1009 		mb0 = m_copy(m, 0, M_COPYALL);
1010 		/*
1011 		 * Pullup packet header if needed before storing it,
1012 		 * as other references may modify it in the meantime.
1013 		 */
1014 		if (mb0 &&
1015 		    (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1016 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1017 		if (mb0 == NULL) {
1018 			free(rte, M_MRTABLE);
1019 			splx(s);
1020 			return ENOBUFS;
1021 		}
1022 
1023 		/* is there an upcall waiting for this packet? */
1024 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1025 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1026 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1027 					       &rt->mf6c_origin.sin6_addr) &&
1028 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1029 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1030 			    (rt->mf6c_stall != NULL))
1031 				break;
1032 		}
1033 
1034 		if (rt == NULL) {
1035 			struct mrt6msg *im;
1036 #ifdef MRT6_OINIT
1037 			struct omrt6msg *oim;
1038 #endif
1039 
1040 			/* no upcall, so make a new entry */
1041 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1042 						  M_NOWAIT);
1043 			if (rt == NULL) {
1044 				free(rte, M_MRTABLE);
1045 				m_freem(mb0);
1046 				splx(s);
1047 				return ENOBUFS;
1048 			}
1049 			/*
1050 			 * Make a copy of the header to send to the user
1051 			 * level process
1052 			 */
1053 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1054 
1055 			if (mm == NULL) {
1056 				free(rte, M_MRTABLE);
1057 				m_freem(mb0);
1058 				free(rt, M_MRTABLE);
1059 				splx(s);
1060 				return ENOBUFS;
1061 			}
1062 
1063 			/*
1064 			 * Send message to routing daemon
1065 			 */
1066 			sin6.sin6_addr = ip6->ip6_src;
1067 
1068 			im = NULL;
1069 #ifdef MRT6_OINIT
1070 			oim = NULL;
1071 #endif
1072 			switch (ip6_mrouter_ver) {
1073 #ifdef MRT6_OINIT
1074 			case MRT6_OINIT:
1075 				oim = mtod(mm, struct omrt6msg *);
1076 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1077 				oim->im6_mbz = 0;
1078 				break;
1079 #endif
1080 			case MRT6_INIT:
1081 				im = mtod(mm, struct mrt6msg *);
1082 				im->im6_msgtype = MRT6MSG_NOCACHE;
1083 				im->im6_mbz = 0;
1084 				break;
1085 			default:
1086 				free(rte, M_MRTABLE);
1087 				m_freem(mb0);
1088 				free(rt, M_MRTABLE);
1089 				splx(s);
1090 				return EINVAL;
1091 			}
1092 
1093 #ifdef MRT6DEBUG
1094 			if (mrt6debug & DEBUG_FORWARD)
1095 				log(LOG_DEBUG,
1096 				    "getting the iif info in the kernel\n");
1097 #endif
1098 
1099 			for (mifp = mif6table, mifi = 0;
1100 			     mifi < nummifs && mifp->m6_ifp != ifp;
1101 			     mifp++, mifi++)
1102 				;
1103 
1104 			switch (ip6_mrouter_ver) {
1105 #ifdef MRT6_OINIT
1106 			case MRT6_OINIT:
1107 				oim->im6_mif = mifi;
1108 				break;
1109 #endif
1110 			case MRT6_INIT:
1111 				im->im6_mif = mifi;
1112 				break;
1113 			}
1114 
1115 			if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1116 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1117 				    "socket queue full\n");
1118 				mrt6stat.mrt6s_upq_sockfull++;
1119 				free(rte, M_MRTABLE);
1120 				m_freem(mb0);
1121 				free(rt, M_MRTABLE);
1122 				splx(s);
1123 				return ENOBUFS;
1124 			}
1125 
1126 			mrt6stat.mrt6s_upcalls++;
1127 
1128 			/* insert new entry at head of hash chain */
1129 			bzero(rt, sizeof(*rt));
1130 			rt->mf6c_origin.sin6_family = AF_INET6;
1131 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1132 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1133 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1134 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1135 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1136 			rt->mf6c_expire = UPCALL_EXPIRE;
1137 			n6expire[hash]++;
1138 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1139 
1140 			/* link into table */
1141 			rt->mf6c_next  = mf6ctable[hash];
1142 			mf6ctable[hash] = rt;
1143 			/* Add this entry to the end of the queue */
1144 			rt->mf6c_stall = rte;
1145 		} else {
1146 			/* determine if q has overflowed */
1147 			struct rtdetq **p;
1148 			int npkts = 0;
1149 
1150 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1151 				if (++npkts > MAX_UPQ6) {
1152 					mrt6stat.mrt6s_upq_ovflw++;
1153 					free(rte, M_MRTABLE);
1154 					m_freem(mb0);
1155 					splx(s);
1156 					return 0;
1157 				}
1158 
1159 			/* Add this entry to the end of the queue */
1160 			*p = rte;
1161 		}
1162 
1163 		rte->next = NULL;
1164 		rte->m = mb0;
1165 		rte->ifp = ifp;
1166 #ifdef UPCALL_TIMING
1167 		rte->t = tp;
1168 #endif /* UPCALL_TIMING */
1169 
1170 		splx(s);
1171 
1172 		return 0;
1173 	}
1174 }
1175 
1176 /*
1177  * Clean up cache entries if upcalls are not serviced
1178  * Call from the Slow Timeout mechanism, every half second.
1179  */
1180 static void
1181 expire_upcalls(unused)
1182 	void *unused;
1183 {
1184 	struct rtdetq *rte;
1185 	struct mf6c *mfc, **nptr;
1186 	int i;
1187 	int s;
1188 
1189 	s = splnet();
1190 	for (i = 0; i < MF6CTBLSIZ; i++) {
1191 		if (n6expire[i] == 0)
1192 			continue;
1193 		nptr = &mf6ctable[i];
1194 		while ((mfc = *nptr) != NULL) {
1195 			rte = mfc->mf6c_stall;
1196 			/*
1197 			 * Skip real cache entries
1198 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1199 			 * If it expires now
1200 			 */
1201 			if (rte != NULL &&
1202 			    mfc->mf6c_expire != 0 &&
1203 			    --mfc->mf6c_expire == 0) {
1204 #ifdef MRT6DEBUG
1205 				if (mrt6debug & DEBUG_EXPIRE)
1206 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1207 					    ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1208 					    ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1209 #endif
1210 				/*
1211 				 * drop all the packets
1212 				 * free the mbuf with the pkt, if, timing info
1213 				 */
1214 				do {
1215 					struct rtdetq *n = rte->next;
1216 					m_freem(rte->m);
1217 					free(rte, M_MRTABLE);
1218 					rte = n;
1219 				} while (rte != NULL);
1220 				mrt6stat.mrt6s_cache_cleanups++;
1221 				n6expire[i]--;
1222 
1223 				*nptr = mfc->mf6c_next;
1224 				free(mfc, M_MRTABLE);
1225 			} else {
1226 				nptr = &mfc->mf6c_next;
1227 			}
1228 		}
1229 	}
1230 	splx(s);
1231 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1232 	    expire_upcalls, NULL);
1233 }
1234 
1235 /*
1236  * Packet forwarding routine once entry in the cache is made
1237  */
1238 static int
1239 ip6_mdq(m, ifp, rt)
1240 	struct mbuf *m;
1241 	struct ifnet *ifp;
1242 	struct mf6c *rt;
1243 {
1244 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1245 	mifi_t mifi, iif;
1246 	struct mif6 *mifp;
1247 	int plen = m->m_pkthdr.len;
1248 
1249 /*
1250  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1251  * input, they shouldn't get counted on output, so statistics keeping is
1252  * separate.
1253  */
1254 
1255 #define MC6_SEND(ip6, mifp, m) do {				\
1256 		if ((mifp)->m6_flags & MIFF_REGISTER)		\
1257 		    register_send((ip6), (mifp), (m));		\
1258 		else						\
1259 		    phyint_send((ip6), (mifp), (m));		\
1260 } while (0)
1261 
1262 	/*
1263 	 * Don't forward if it didn't arrive from the parent mif
1264 	 * for its origin.
1265 	 */
1266 	mifi = rt->mf6c_parent;
1267 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1268 		/* came in the wrong interface */
1269 #ifdef MRT6DEBUG
1270 		if (mrt6debug & DEBUG_FORWARD)
1271 			log(LOG_DEBUG,
1272 			    "wrong if: ifid %d mifi %d mififid %x\n",
1273 			    ifp->if_index, mifi,
1274 			    mif6table[mifi].m6_ifp->if_index);
1275 #endif
1276 		mrt6stat.mrt6s_wrong_if++;
1277 		rt->mf6c_wrong_if++;
1278 		/*
1279 		 * If we are doing PIM processing, and we are forwarding
1280 		 * packets on this interface, send a message to the
1281 		 * routing daemon.
1282 		 */
1283 		/* have to make sure this is a valid mif */
1284 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1285 			if (pim6 && (m->m_flags & M_LOOP) == 0) {
1286 				/*
1287 				 * Check the M_LOOP flag to avoid an
1288 				 * unnecessary PIM assert.
1289 				 * XXX: M_LOOP is an ad-hoc hack...
1290 				 */
1291 				static struct sockaddr_in6 sin6 =
1292 				{ sizeof(sin6), AF_INET6 };
1293 
1294 				struct mbuf *mm;
1295 				struct mrt6msg *im;
1296 #ifdef MRT6_OINIT
1297 				struct omrt6msg *oim;
1298 #endif
1299 
1300 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1301 				if (mm &&
1302 				    (M_HASCL(mm) ||
1303 				     mm->m_len < sizeof(struct ip6_hdr)))
1304 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1305 				if (mm == NULL)
1306 					return ENOBUFS;
1307 
1308 #ifdef MRT6_OINIT
1309 				oim = NULL;
1310 #endif
1311 				im = NULL;
1312 				switch (ip6_mrouter_ver) {
1313 #ifdef MRT6_OINIT
1314 				case MRT6_OINIT:
1315 					oim = mtod(mm, struct omrt6msg *);
1316 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1317 					oim->im6_mbz = 0;
1318 					break;
1319 #endif
1320 				case MRT6_INIT:
1321 					im = mtod(mm, struct mrt6msg *);
1322 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1323 					im->im6_mbz = 0;
1324 					break;
1325 				default:
1326 					m_freem(mm);
1327 					return EINVAL;
1328 				}
1329 
1330 				for (mifp = mif6table, iif = 0;
1331 				     iif < nummifs && mifp &&
1332 					     mifp->m6_ifp != ifp;
1333 				     mifp++, iif++)
1334 					;
1335 
1336 				switch (ip6_mrouter_ver) {
1337 #ifdef MRT6_OINIT
1338 				case MRT6_OINIT:
1339 					oim->im6_mif = iif;
1340 					sin6.sin6_addr = oim->im6_src;
1341 					break;
1342 #endif
1343 				case MRT6_INIT:
1344 					im->im6_mif = iif;
1345 					sin6.sin6_addr = im->im6_src;
1346 					break;
1347 				}
1348 
1349 				mrt6stat.mrt6s_upcalls++;
1350 
1351 				if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1352 #ifdef MRT6DEBUG
1353 					if (mrt6debug)
1354 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1355 #endif
1356 					++mrt6stat.mrt6s_upq_sockfull;
1357 					return ENOBUFS;
1358 				}	/* if socket Q full */
1359 			}		/* if PIM */
1360 		return 0;
1361 	}			/* if wrong iif */
1362 
1363 	/* If I sourced this packet, it counts as output, else it was input. */
1364 	if (m->m_pkthdr.rcvif == NULL) {
1365 		/* XXX: is rcvif really NULL when output?? */
1366 		mif6table[mifi].m6_pkt_out++;
1367 		mif6table[mifi].m6_bytes_out += plen;
1368 	} else {
1369 		mif6table[mifi].m6_pkt_in++;
1370 		mif6table[mifi].m6_bytes_in += plen;
1371 	}
1372 	rt->mf6c_pkt_cnt++;
1373 	rt->mf6c_byte_cnt += plen;
1374 
1375 	/*
1376 	 * For each mif, forward a copy of the packet if there are group
1377 	 * members downstream on the interface.
1378 	 */
1379 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1380 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1381 			/*
1382 			 * check if the outgoing packet is going to break
1383 			 * a scope boundary.
1384 			 * XXX For packets through PIM register tunnel
1385 			 * interface, we believe a routing daemon.
1386 			 */
1387 			if ((mif6table[rt->mf6c_parent].m6_flags &
1388 			     MIFF_REGISTER) == 0 &&
1389 			    (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1390 			    (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1391 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1392 					      &ip6->ip6_dst) ||
1393 			     in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1394 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1395 					      &ip6->ip6_src))) {
1396 				ip6stat.ip6s_badscope++;
1397 				continue;
1398 			}
1399 
1400 			mifp->m6_pkt_out++;
1401 			mifp->m6_bytes_out += plen;
1402 			MC6_SEND(ip6, mifp, m);
1403 		}
1404 	return 0;
1405 }
1406 
1407 static void
1408 phyint_send(ip6, mifp, m)
1409     struct ip6_hdr *ip6;
1410     struct mif6 *mifp;
1411     struct mbuf *m;
1412 {
1413 	struct mbuf *mb_copy;
1414 	struct ifnet *ifp = mifp->m6_ifp;
1415 	int error = 0;
1416 	int s = splnet();	/* needs to protect static "ro" below. */
1417 	static struct route_in6 ro;
1418 	struct	in6_multi *in6m;
1419 	struct sockaddr_in6 *dst6;
1420 
1421 	/*
1422 	 * Make a new reference to the packet; make sure that
1423 	 * the IPv6 header is actually copied, not just referenced,
1424 	 * so that ip6_output() only scribbles on the copy.
1425 	 */
1426 	mb_copy = m_copy(m, 0, M_COPYALL);
1427 	if (mb_copy &&
1428 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1429 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1430 	if (mb_copy == NULL) {
1431 		splx(s);
1432 		return;
1433 	}
1434 	/* set MCAST flag to the outgoing packet */
1435 	mb_copy->m_flags |= M_MCAST;
1436 
1437 	/*
1438 	 * If we sourced the packet, call ip6_output since we may devide
1439 	 * the packet into fragments when the packet is too big for the
1440 	 * outgoing interface.
1441 	 * Otherwise, we can simply send the packet to the interface
1442 	 * sending queue.
1443 	 */
1444 	if (m->m_pkthdr.rcvif == NULL) {
1445 		struct ip6_moptions im6o;
1446 
1447 		im6o.im6o_multicast_ifp = ifp;
1448 		/* XXX: ip6_output will override ip6->ip6_hlim */
1449 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1450 		im6o.im6o_multicast_loop = 1;
1451 		error = ip6_output(mb_copy, NULL, &ro,
1452 				   IPV6_FORWARDING, &im6o, NULL, NULL);
1453 
1454 #ifdef MRT6DEBUG
1455 		if (mrt6debug & DEBUG_XMIT)
1456 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1457 			    mifp - mif6table, error);
1458 #endif
1459 		splx(s);
1460 		return;
1461 	}
1462 
1463 	/*
1464 	 * If we belong to the destination multicast group
1465 	 * on the outgoing interface, loop back a copy.
1466 	 */
1467 	dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
1468 	IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1469 	if (in6m != NULL) {
1470 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1471 		dst6->sin6_family = AF_INET6;
1472 		dst6->sin6_addr = ip6->ip6_dst;
1473 		ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst);
1474 	}
1475 	/*
1476 	 * Put the packet into the sending queue of the outgoing interface
1477 	 * if it would fit in the MTU of the interface.
1478 	 */
1479 	if (mb_copy->m_pkthdr.len <= ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
1480 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1481 		dst6->sin6_family = AF_INET6;
1482 		dst6->sin6_addr = ip6->ip6_dst;
1483 		/*
1484 		 * We just call if_output instead of nd6_output here, since
1485 		 * we need no ND for a multicast forwarded packet...right?
1486 		 */
1487 		error = (*ifp->if_output)(ifp, mb_copy,
1488 		    (struct sockaddr *)&ro.ro_dst, NULL);
1489 #ifdef MRT6DEBUG
1490 		if (mrt6debug & DEBUG_XMIT)
1491 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1492 			    mifp - mif6table, error);
1493 #endif
1494 	} else {
1495 #ifdef MULTICAST_PMTUD
1496 		icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
1497 #else
1498 #ifdef MRT6DEBUG
1499 		if (mrt6debug & DEBUG_XMIT)
1500 			log(LOG_DEBUG,
1501 			    "phyint_send: packet too big on %s o %s g %s"
1502 			    " size %d(discarded)\n",
1503 			    if_name(ifp),
1504 			    ip6_sprintf(&ip6->ip6_src),
1505 			    ip6_sprintf(&ip6->ip6_dst),
1506 			    mb_copy->m_pkthdr.len);
1507 #endif /* MRT6DEBUG */
1508 		m_freem(mb_copy); /* simply discard the packet */
1509 #endif
1510 	}
1511 
1512 	splx(s);
1513 }
1514 
1515 static int
1516 register_send(ip6, mif, m)
1517 	struct ip6_hdr *ip6;
1518 	struct mif6 *mif;
1519 	struct mbuf *m;
1520 {
1521 	struct mbuf *mm;
1522 	int i, len = m->m_pkthdr.len;
1523 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1524 	struct mrt6msg *im6;
1525 
1526 #ifdef MRT6DEBUG
1527 	if (mrt6debug)
1528 		log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1529 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1530 #endif
1531 	++pim6stat.pim6s_snd_registers;
1532 
1533 	/* Make a copy of the packet to send to the user level process */
1534 	MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1535 	if (mm == NULL)
1536 		return ENOBUFS;
1537 	mm->m_pkthdr.rcvif = NULL;
1538 	mm->m_data += max_linkhdr;
1539 	mm->m_len = sizeof(struct ip6_hdr);
1540 
1541 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1542 		m_freem(mm);
1543 		return ENOBUFS;
1544 	}
1545 	i = MHLEN - M_LEADINGSPACE(mm);
1546 	if (i > len)
1547 		i = len;
1548 	mm = m_pullup(mm, i);
1549 	if (mm == NULL)
1550 		return ENOBUFS;
1551 /* TODO: check it! */
1552 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1553 
1554 	/*
1555 	 * Send message to routing daemon
1556 	 */
1557 	sin6.sin6_addr = ip6->ip6_src;
1558 
1559 	im6 = mtod(mm, struct mrt6msg *);
1560 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1561 	im6->im6_mbz          = 0;
1562 
1563 	im6->im6_mif = mif - mif6table;
1564 
1565 	/* iif info is not given for reg. encap.n */
1566 	mrt6stat.mrt6s_upcalls++;
1567 
1568 	if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1569 #ifdef MRT6DEBUG
1570 		if (mrt6debug)
1571 			log(LOG_WARNING,
1572 			    "register_send: ip6_mrouter socket queue full\n");
1573 #endif
1574 		++mrt6stat.mrt6s_upq_sockfull;
1575 		return ENOBUFS;
1576 	}
1577 	return 0;
1578 }
1579 
1580 /*
1581  * PIM sparse mode hook
1582  * Receives the pim control messages, and passes them up to the listening
1583  * socket, using rip6_input.
1584  * The only message processed is the REGISTER pim message; the pim header
1585  * is stripped off, and the inner packet is passed to register_mforward.
1586  */
1587 int
1588 pim6_input(mp, offp, proto)
1589 	struct mbuf **mp;
1590 	int *offp, proto;
1591 {
1592 	struct pim *pim; /* pointer to a pim struct */
1593 	struct ip6_hdr *ip6;
1594 	int pimlen;
1595 	struct mbuf *m = *mp;
1596 	int minlen;
1597 	int off = *offp;
1598 
1599 	++pim6stat.pim6s_rcv_total;
1600 
1601 	ip6 = mtod(m, struct ip6_hdr *);
1602 	pimlen = m->m_pkthdr.len - *offp;
1603 
1604 	/*
1605 	 * Validate lengths
1606 	 */
1607 	if (pimlen < PIM_MINLEN) {
1608 		++pim6stat.pim6s_rcv_tooshort;
1609 #ifdef MRT6DEBUG
1610 		if (mrt6debug & DEBUG_PIM)
1611 			log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1612 #endif
1613 		m_freem(m);
1614 		return(IPPROTO_DONE);
1615 	}
1616 
1617 	/*
1618 	 * if the packet is at least as big as a REGISTER, go ahead
1619 	 * and grab the PIM REGISTER header size, to avoid another
1620 	 * possible m_pullup() later.
1621 	 *
1622 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1623 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1624 	 */
1625 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1626 
1627 	/*
1628 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1629 	 * possibly the PIM REGISTER header
1630 	 */
1631 #ifndef PULLDOWN_TEST
1632 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1633 	/* adjust pointer */
1634 	ip6 = mtod(m, struct ip6_hdr *);
1635 
1636 	/* adjust mbuf to point to the PIM header */
1637 	pim = (struct pim *)((caddr_t)ip6 + off);
1638 #else
1639 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1640 	if (pim == NULL) {
1641 		pim6stat.pim6s_rcv_tooshort++;
1642 		return IPPROTO_DONE;
1643 	}
1644 #endif
1645 
1646 #define PIM6_CHECKSUM
1647 #ifdef PIM6_CHECKSUM
1648 	{
1649 		int cksumlen;
1650 
1651 		/*
1652 		 * Validate checksum.
1653 		 * If PIM REGISTER, exclude the data packet
1654 		 */
1655 		if (pim->pim_type == PIM_REGISTER)
1656 			cksumlen = PIM_MINLEN;
1657 		else
1658 			cksumlen = pimlen;
1659 
1660 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1661 			++pim6stat.pim6s_rcv_badsum;
1662 #ifdef MRT6DEBUG
1663 			if (mrt6debug & DEBUG_PIM)
1664 				log(LOG_DEBUG,
1665 				    "pim6_input: invalid checksum\n");
1666 #endif
1667 			m_freem(m);
1668 			return(IPPROTO_DONE);
1669 		}
1670 	}
1671 #endif /* PIM_CHECKSUM */
1672 
1673 	/* PIM version check */
1674 	if (pim->pim_ver != PIM_VERSION) {
1675 		++pim6stat.pim6s_rcv_badversion;
1676 #ifdef MRT6DEBUG
1677 		log(LOG_ERR,
1678 		    "pim6_input: incorrect version %d, expecting %d\n",
1679 		    pim->pim_ver, PIM_VERSION);
1680 #endif
1681 		m_freem(m);
1682 		return(IPPROTO_DONE);
1683 	}
1684 
1685 	if (pim->pim_type == PIM_REGISTER) {
1686 		/*
1687 		 * since this is a REGISTER, we'll make a copy of the register
1688 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1689 		 * routing daemon.
1690 		 */
1691 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1692 
1693 		struct mbuf *mcp;
1694 		struct ip6_hdr *eip6;
1695 		u_int32_t *reghdr;
1696 		int rc;
1697 
1698 		++pim6stat.pim6s_rcv_registers;
1699 
1700 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1701 #ifdef MRT6DEBUG
1702 			if (mrt6debug & DEBUG_PIM)
1703 				log(LOG_DEBUG,
1704 				    "pim6_input: register mif not set: %d\n",
1705 				    reg_mif_num);
1706 #endif
1707 			m_freem(m);
1708 			return(IPPROTO_DONE);
1709 		}
1710 
1711 		reghdr = (u_int32_t *)(pim + 1);
1712 
1713 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1714 			goto pim6_input_to_daemon;
1715 
1716 		/*
1717 		 * Validate length
1718 		 */
1719 		if (pimlen < PIM6_REG_MINLEN) {
1720 			++pim6stat.pim6s_rcv_tooshort;
1721 			++pim6stat.pim6s_rcv_badregisters;
1722 #ifdef MRT6DEBUG
1723 			log(LOG_ERR,
1724 			    "pim6_input: register packet size too "
1725 			    "small %d from %s\n",
1726 			    pimlen, ip6_sprintf(&ip6->ip6_src));
1727 #endif
1728 			m_freem(m);
1729 			return(IPPROTO_DONE);
1730 		}
1731 
1732 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1733 #ifdef MRT6DEBUG
1734 		if (mrt6debug & DEBUG_PIM)
1735 			log(LOG_DEBUG,
1736 			    "pim6_input[register], eip6: %s -> %s, "
1737 			    "eip6 plen %d\n",
1738 			    ip6_sprintf(&eip6->ip6_src),
1739 			    ip6_sprintf(&eip6->ip6_dst),
1740 			    ntohs(eip6->ip6_plen));
1741 #endif
1742 
1743 		/* verify the version number of the inner packet */
1744 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1745 			++pim6stat.pim6s_rcv_badregisters;
1746 #ifdef MRT6DEBUG
1747 			log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1748 			    "of the inner packet\n",
1749 			    (eip6->ip6_vfc & IPV6_VERSION));
1750 #endif
1751 			m_freem(m);
1752 			return(IPPROTO_NONE);
1753 		}
1754 
1755 		/* verify the inner packet is destined to a mcast group */
1756 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1757 			++pim6stat.pim6s_rcv_badregisters;
1758 #ifdef MRT6DEBUG
1759 			if (mrt6debug & DEBUG_PIM)
1760 				log(LOG_DEBUG,
1761 				    "pim6_input: inner packet of register "
1762 				    "is not multicast %s\n",
1763 				    ip6_sprintf(&eip6->ip6_dst));
1764 #endif
1765 			m_freem(m);
1766 			return(IPPROTO_DONE);
1767 		}
1768 
1769 		/*
1770 		 * make a copy of the whole header to pass to the daemon later.
1771 		 */
1772 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1773 		if (mcp == NULL) {
1774 #ifdef MRT6DEBUG
1775 			log(LOG_ERR,
1776 			    "pim6_input: pim register: "
1777 			    "could not copy register head\n");
1778 #endif
1779 			m_freem(m);
1780 			return(IPPROTO_DONE);
1781 		}
1782 
1783 		/*
1784 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1785 		 */
1786 		m_adj(m, off + PIM_MINLEN);
1787 #ifdef MRT6DEBUG
1788 		if (mrt6debug & DEBUG_PIM) {
1789 			log(LOG_DEBUG,
1790 			    "pim6_input: forwarding decapsulated register: "
1791 			    "src %s, dst %s, mif %d\n",
1792 			    ip6_sprintf(&eip6->ip6_src),
1793 			    ip6_sprintf(&eip6->ip6_dst),
1794 			    reg_mif_num);
1795 		}
1796 #endif
1797 
1798  		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1799 				dst.sin6_family, 0);
1800 
1801 		/* prepare the register head to send to the mrouting daemon */
1802 		m = mcp;
1803 	}
1804 
1805 	/*
1806 	 * Pass the PIM message up to the daemon; if it is a register message
1807 	 * pass the 'head' only up to the daemon. This includes the
1808 	 * encapsulator ip6 header, pim header, register header and the
1809 	 * encapsulated ip6 header.
1810 	 */
1811   pim6_input_to_daemon:
1812 	rip6_input(&m, offp, proto);
1813 	return(IPPROTO_DONE);
1814 }
1815