xref: /freebsd/sys/netinet6/ip6_mroute.c (revision b601c69bdbe8755d26570261d7fd4c02ee4eff74)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_mroute.c,v 1.24 2000/05/19 07:37:05 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*	BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp	*/
34 
35 /*
36  * IP multicast forwarding procedures
37  *
38  * Written by David Waitzman, BBN Labs, August 1988.
39  * Modified by Steve Deering, Stanford, February 1989.
40  * Modified by Mark J. Steiglitz, Stanford, May, 1991
41  * Modified by Van Jacobson, LBL, January 1993
42  * Modified by Ajit Thyagarajan, PARC, August 1993
43  * Modified by Bill Fenenr, PARC, April 1994
44  *
45  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
46  */
47 
48 #include "opt_inet.h"
49 #include "opt_inet6.h"
50 
51 #include <sys/param.h>
52 #include <sys/systm.h>
53 #include <sys/malloc.h>
54 #include <sys/mbuf.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sockio.h>
58 #include <sys/protosw.h>
59 #include <sys/errno.h>
60 #include <sys/time.h>
61 #include <sys/kernel.h>
62 #include <sys/syslog.h>
63 
64 #include <net/if.h>
65 #include <net/route.h>
66 #include <net/raw_cb.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 
71 #include <netinet/ip6.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/ip6_mroute.h>
74 #include <netinet6/pim6.h>
75 #include <netinet6/pim6_var.h>
76 
77 static MALLOC_DEFINE(M_MRTABLE, "mf6c", "multicast forwarding cache entry");
78 
79 #define M_HASCL(m) ((m)->m_flags & M_EXT)
80 
81 static int ip6_mdq __P((struct mbuf *, struct ifnet *, struct mf6c *));
82 static void phyint_send __P((struct ip6_hdr *, struct mif6 *, struct mbuf *));
83 
84 static int set_pim6 __P((int *));
85 static int socket_send __P((struct socket *, struct mbuf *,
86 			    struct sockaddr_in6 *));
87 static int register_send __P((struct ip6_hdr *, struct mif6 *,
88 			      struct mbuf *));
89 
90 /*
91  * Globals.  All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static,
92  * except for netstat or debugging purposes.
93  */
94 struct socket  *ip6_mrouter  = NULL;
95 int		ip6_mrouter_ver = 0;
96 int		ip6_mrtproto = IPPROTO_PIM;    /* for netstat only */
97 struct mrt6stat	mrt6stat;
98 
99 #define NO_RTE_FOUND 	0x1
100 #define RTE_FOUND	0x2
101 
102 struct mf6c	*mf6ctable[MF6CTBLSIZ];
103 u_char		nexpire[MF6CTBLSIZ];
104 static struct mif6 mif6table[MAXMIFS];
105 #ifdef MRT6DEBUG
106 u_int		mrt6debug = 0;	  /* debug level 	*/
107 #define		DEBUG_MFC	0x02
108 #define		DEBUG_FORWARD	0x04
109 #define		DEBUG_EXPIRE	0x08
110 #define		DEBUG_XMIT	0x10
111 #define         DEBUG_REG       0x20
112 #define         DEBUG_PIM       0x40
113 #endif
114 
115 static void	expire_upcalls __P((void *));
116 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
117 #define		UPCALL_EXPIRE	6		/* number of timeouts */
118 
119 #ifdef INET
120 #ifdef MROUTING
121 extern struct socket *ip_mrouter;
122 #endif
123 #endif
124 
125 /*
126  * 'Interfaces' associated with decapsulator (so we can tell
127  * packets that went through it from ones that get reflected
128  * by a broken gateway).  These interfaces are never linked into
129  * the system ifnet list & no routes point to them.  I.e., packets
130  * can't be sent this way.  They only exist as a placeholder for
131  * multicast source verification.
132  */
133 struct ifnet multicast_register_if;
134 
135 #define ENCAP_HOPS 64
136 
137 /*
138  * Private variables.
139  */
140 static mifi_t nummifs = 0;
141 static mifi_t reg_mif_num = (mifi_t)-1;
142 
143 static struct pim6stat pim6stat;
144 static struct callout_handle expire_upcalls_ch;
145 
146 /*
147  * one-back cache used by ipip_input to locate a tunnel's mif
148  * given a datagram's src ip address.
149  */
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 	register 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 	    register 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 /*
223  * Handle MRT setsockopt commands to modify the multicast routing tables.
224  */
225 int
226 ip6_mrouter_set(so, sopt)
227 	struct socket *so;
228 	struct sockopt *sopt;
229 {
230 	int	error = 0;
231 	struct mbuf *m;
232 
233 	if (so != ip6_mrouter && sopt->sopt_name != MRT6_INIT)
234 		return (EACCES);
235 
236 	if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
237 		return (error);
238 	if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
239 		return (error);
240 
241 	switch (sopt->sopt_name) {
242 	 case MRT6_INIT:
243 #ifdef MRT6_OINIT
244 	 case MRT6_OINIT:
245 #endif
246 		 error = ip6_mrouter_init(so, m, sopt->sopt_name);
247 		 break;
248 	 case MRT6_DONE:
249 		 error = ip6_mrouter_done();
250 		 break;
251 	 case MRT6_ADD_MIF:
252 		 error = add_m6if(mtod(m, struct mif6ctl *));
253 		 break;
254 	 case MRT6_DEL_MIF:
255 		 error = del_m6if(mtod(m, mifi_t *));
256 		 break;
257 	 case MRT6_ADD_MFC:
258 		 error = add_m6fc(mtod(m, struct mf6cctl *));
259 		 break;
260 	 case MRT6_DEL_MFC:
261 		 error = del_m6fc(mtod(m, struct mf6cctl *));
262 		 break;
263 	 case MRT6_PIM:
264 		 error = set_pim6(mtod(m, int *));
265 		 break;
266 	 default:
267 		 error = EOPNOTSUPP;
268 		 break;
269 	}
270 
271 	(void)m_freem(m);
272 	return(error);
273 }
274 
275 /*
276  * Handle MRT getsockopt commands
277  */
278 int
279 ip6_mrouter_get(so, sopt)
280 	struct socket *so;
281 	struct sockopt *sopt;
282 {
283 	int error = 0;
284 
285 	if (so != ip6_mrouter) return EACCES;
286 
287 	switch (sopt->sopt_name) {
288 		case MRT6_PIM:
289 			error = sooptcopyout(sopt, &pim6, sizeof(pim6));
290 			break;
291 	}
292 	return (error);
293 }
294 
295 /*
296  * Handle ioctl commands to obtain information from the cache
297  */
298 int
299 mrt6_ioctl(cmd, data)
300 	int cmd;
301 	caddr_t data;
302 {
303     int error = 0;
304 
305     switch (cmd) {
306      case SIOCGETSGCNT_IN6:
307 	     return(get_sg_cnt((struct sioc_sg_req6 *)data));
308 	     break;		/* for safety */
309      case SIOCGETMIFCNT_IN6:
310 	     return(get_mif6_cnt((struct sioc_mif_req6 *)data));
311 	     break;		/* for safety */
312      default:
313 	     return (EINVAL);
314 	     break;
315     }
316     return error;
317 }
318 
319 /*
320  * returns the packet, byte, rpf-failure count for the source group provided
321  */
322 static int
323 get_sg_cnt(req)
324 	register struct sioc_sg_req6 *req;
325 {
326 	register struct mf6c *rt;
327 	int s;
328 
329 	s = splnet();
330 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
331 	splx(s);
332 	if (rt != NULL) {
333 		req->pktcnt = rt->mf6c_pkt_cnt;
334 		req->bytecnt = rt->mf6c_byte_cnt;
335 		req->wrong_if = rt->mf6c_wrong_if;
336 	} else
337 		return(ESRCH);
338 #if 0
339 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
340 #endif
341 
342 	return 0;
343 }
344 
345 /*
346  * returns the input and output packet and byte counts on the mif provided
347  */
348 static int
349 get_mif6_cnt(req)
350 	register struct sioc_mif_req6 *req;
351 {
352 	register mifi_t mifi = req->mifi;
353 
354 	if (mifi >= nummifs)
355 		return EINVAL;
356 
357 	req->icount = mif6table[mifi].m6_pkt_in;
358 	req->ocount = mif6table[mifi].m6_pkt_out;
359 	req->ibytes = mif6table[mifi].m6_bytes_in;
360 	req->obytes = mif6table[mifi].m6_bytes_out;
361 
362 	return 0;
363 }
364 
365 static int
366 set_pim6(i)
367 	int *i;
368 {
369 	if ((*i != 1) && (*i != 0))
370 		return EINVAL;
371 
372 	pim6 = *i;
373 
374 	return 0;
375 }
376 
377 /*
378  * Enable multicast routing
379  */
380 static int
381 ip6_mrouter_init(so, m, cmd)
382 	struct socket *so;
383 	struct mbuf *m;
384 	int cmd;
385 {
386 	int *v;
387 
388 #ifdef MRT6DEBUG
389 	if (mrt6debug)
390 		log(LOG_DEBUG,
391 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
392 		    so->so_type, so->so_proto->pr_protocol);
393 #endif
394 
395 	if (so->so_type != SOCK_RAW ||
396 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
397 		return EOPNOTSUPP;
398 
399 	if (!m || (m->m_len != sizeof(int *)))
400 		return ENOPROTOOPT;
401 
402 	v = mtod(m, int *);
403 	if (*v != 1)
404 		return ENOPROTOOPT;
405 
406 	if (ip6_mrouter != NULL) return EADDRINUSE;
407 
408 	ip6_mrouter = so;
409 	ip6_mrouter_ver = cmd;
410 
411 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
412 	bzero((caddr_t)nexpire, sizeof(nexpire));
413 
414 	pim6 = 0;/* used for stubbing out/in pim stuff */
415 
416 	expire_upcalls_ch =
417 		timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
418 
419 #ifdef MRT6DEBUG
420 	if (mrt6debug)
421 		log(LOG_DEBUG, "ip6_mrouter_init\n");
422 #endif
423 
424 	return 0;
425 }
426 
427 /*
428  * Disable multicast routing
429  */
430 int
431 ip6_mrouter_done()
432 {
433 	mifi_t mifi;
434 	int i;
435 	struct ifnet *ifp;
436 	struct in6_ifreq ifr;
437 	struct mf6c *rt;
438 	struct rtdetq *rte;
439 	int s;
440 
441 	s = splnet();
442 
443 	/*
444 	 * For each phyint in use, disable promiscuous reception of all IPv6
445 	 * multicasts.
446 	 */
447 #ifdef INET
448 #ifdef MROUTING
449 	/*
450 	 * If there is still IPv4 multicast routing daemon,
451 	 * we remain interfaces to receive all muliticasted packets.
452 	 * XXX: there may be an interface in which the IPv4 multicast
453 	 * daemon is not interested...
454 	 */
455 	if (!ip_mrouter)
456 #endif
457 #endif
458 	{
459 		for (mifi = 0; mifi < nummifs; mifi++) {
460 			if (mif6table[mifi].m6_ifp &&
461 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
462 				ifr.ifr_addr.sin6_family = AF_INET6;
463 				ifr.ifr_addr.sin6_addr= in6addr_any;
464 				ifp = mif6table[mifi].m6_ifp;
465 				(*ifp->if_ioctl)(ifp, SIOCDELMULTI,
466 						 (caddr_t)&ifr);
467 			}
468 		}
469 	}
470 #ifdef notyet
471 	bzero((caddr_t)qtable, sizeof(qtable));
472 	bzero((caddr_t)tbftable, sizeof(tbftable));
473 #endif
474 	bzero((caddr_t)mif6table, sizeof(mif6table));
475 	nummifs = 0;
476 
477 	pim6 = 0; /* used to stub out/in pim specific code */
478 
479 	untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch);
480 
481 	/*
482 	 * Free all multicast forwarding cache entries.
483 	 */
484 	for (i = 0; i < MF6CTBLSIZ; i++) {
485 		rt = mf6ctable[i];
486 		while (rt) {
487 			struct mf6c *frt;
488 
489 			for (rte = rt->mf6c_stall; rte != NULL; ) {
490 				struct rtdetq *n = rte->next;
491 
492 				m_free(rte->m);
493 				free(rte, M_MRTABLE);
494 				rte = n;
495 			}
496 			frt = rt;
497 			rt = rt->mf6c_next;
498 			free(frt, M_MRTABLE);
499 		}
500 	}
501 
502 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
503 
504 	/*
505 	 * Reset de-encapsulation cache
506 	 */
507 	reg_mif_num = -1;
508 
509 	ip6_mrouter = NULL;
510 	ip6_mrouter_ver = 0;
511 
512 	splx(s);
513 
514 #ifdef MRT6DEBUG
515 	if (mrt6debug)
516 		log(LOG_DEBUG, "ip6_mrouter_done\n");
517 #endif
518 
519 	return 0;
520 }
521 
522 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
523 
524 /*
525  * Add a mif to the mif table
526  */
527 static int
528 add_m6if(mifcp)
529 	register struct mif6ctl *mifcp;
530 {
531 	register struct mif6 *mifp;
532 	struct ifnet *ifp;
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 			multicast_register_if.if_name = "register_mif";
550 			multicast_register_if.if_flags |= IFF_LOOPBACK;
551 			multicast_register_if.if_index = mifcp->mif6c_mifi;
552 			reg_mif_num = mifcp->mif6c_mifi;
553 		}
554 
555 		ifp = &multicast_register_if;
556 
557 	} /* if REGISTER */
558 	else {
559 		/* Make sure the interface supports multicast */
560 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
561 			return EOPNOTSUPP;
562 
563 		s = splnet();
564 		error = if_allmulti(ifp, 1);
565 		splx(s);
566 		if (error)
567 			return error;
568 	}
569 
570 	s = splnet();
571 	mifp->m6_flags     = mifcp->mif6c_flags;
572 	mifp->m6_ifp       = ifp;
573 #ifdef notyet
574 	/* scaling up here allows division by 1024 in critical code */
575 	mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
576 #endif
577 	/* initialize per mif pkt counters */
578 	mifp->m6_pkt_in    = 0;
579 	mifp->m6_pkt_out   = 0;
580 	mifp->m6_bytes_in  = 0;
581 	mifp->m6_bytes_out = 0;
582 	splx(s);
583 
584 	/* Adjust nummifs up if the mifi is higher than nummifs */
585 	if (nummifs <= mifcp->mif6c_mifi)
586 		nummifs = mifcp->mif6c_mifi + 1;
587 
588 #ifdef MRT6DEBUG
589 	if (mrt6debug)
590 		log(LOG_DEBUG,
591 		    "add_mif #%d, phyint %s%d\n",
592 		    mifcp->mif6c_mifi,
593 		    ifp->if_name, ifp->if_unit);
594 #endif
595 
596 	return 0;
597 }
598 
599 /*
600  * Delete a mif from the mif table
601  */
602 static int
603 del_m6if(mifip)
604 	mifi_t *mifip;
605 {
606 	register struct mif6 *mifp = mif6table + *mifip;
607 	register mifi_t mifi;
608 	struct ifnet *ifp;
609 	int s;
610 
611 	if (*mifip >= nummifs)
612 		return EINVAL;
613 	if (mifp->m6_ifp == NULL)
614 		return EINVAL;
615 
616 	s = splnet();
617 
618 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
619 		/*
620 		 * XXX: what if there is yet IPv4 multicast daemon
621 		 *      using the interface?
622 		 */
623 		ifp = mifp->m6_ifp;
624 
625 		if_allmulti(ifp, 0);
626 	}
627 
628 #ifdef notyet
629 	bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
630 	bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
631 #endif
632 	bzero((caddr_t)mifp, sizeof (*mifp));
633 
634 	/* Adjust nummifs down */
635 	for (mifi = nummifs; mifi > 0; mifi--)
636 		if (mif6table[mifi - 1].m6_ifp)
637 			break;
638 	nummifs = mifi;
639 
640 	splx(s);
641 
642 #ifdef MRT6DEBUG
643 	if (mrt6debug)
644 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
645 #endif
646 
647 	return 0;
648 }
649 
650 /*
651  * Add an mfc entry
652  */
653 static int
654 add_m6fc(mfccp)
655 	struct mf6cctl *mfccp;
656 {
657 	struct mf6c *rt;
658 	u_long hash;
659 	struct rtdetq *rte;
660 	register u_short nstl;
661 	int s;
662 
663 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
664 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
665 
666 	/* If an entry already exists, just update the fields */
667 	if (rt) {
668 #ifdef MRT6DEBUG
669 		if (mrt6debug & DEBUG_MFC)
670 			log(LOG_DEBUG,"add_m6fc update o %s g %s p %x\n",
671 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
672 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
673 			    mfccp->mf6cc_parent);
674 #endif
675 
676 		s = splnet();
677 		rt->mf6c_parent = mfccp->mf6cc_parent;
678 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
679 		splx(s);
680 		return 0;
681 	}
682 
683 	/*
684 	 * Find the entry for which the upcall was made and update
685 	 */
686 	s = splnet();
687 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
688 			mfccp->mf6cc_mcastgrp.sin6_addr);
689 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
690 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
691 				       &mfccp->mf6cc_origin.sin6_addr) &&
692 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
693 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
694 		    (rt->mf6c_stall != NULL)) {
695 
696 			if (nstl++)
697 				log(LOG_ERR,
698 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
699 				    "multiple kernel entries",
700 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
701 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
702 				    mfccp->mf6cc_parent, rt->mf6c_stall);
703 
704 #ifdef MRT6DEBUG
705 			if (mrt6debug & DEBUG_MFC)
706 				log(LOG_DEBUG,
707 				    "add_m6fc o %s g %s p %x dbg %x\n",
708 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
709 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
710 				    mfccp->mf6cc_parent, rt->mf6c_stall);
711 #endif
712 
713 			rt->mf6c_origin     = mfccp->mf6cc_origin;
714 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
715 			rt->mf6c_parent     = mfccp->mf6cc_parent;
716 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
717 			/* initialize pkt counters per src-grp */
718 			rt->mf6c_pkt_cnt    = 0;
719 			rt->mf6c_byte_cnt   = 0;
720 			rt->mf6c_wrong_if   = 0;
721 
722 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
723 			nexpire[hash]--;
724 
725 			/* free packets Qed at the end of this entry */
726 			for (rte = rt->mf6c_stall; rte != NULL; ) {
727 				struct rtdetq *n = rte->next;
728 				ip6_mdq(rte->m, rte->ifp, rt);
729 				m_freem(rte->m);
730 #ifdef UPCALL_TIMING
731 				collate(&(rte->t));
732 #endif /* UPCALL_TIMING */
733 				free(rte, M_MRTABLE);
734 				rte = n;
735 			}
736 			rt->mf6c_stall = NULL;
737 		}
738 	}
739 
740 	/*
741 	 * It is possible that an entry is being inserted without an upcall
742 	 */
743 	if (nstl == 0) {
744 #ifdef MRT6DEBUG
745 		if (mrt6debug & DEBUG_MFC)
746 			log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n",
747 			    hash,
748 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
749 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
750 			    mfccp->mf6cc_parent);
751 #endif
752 
753 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
754 
755 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
756 					       &mfccp->mf6cc_origin.sin6_addr)&&
757 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
758 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
759 
760 				rt->mf6c_origin     = mfccp->mf6cc_origin;
761 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
762 				rt->mf6c_parent     = mfccp->mf6cc_parent;
763 				/* initialize pkt counters per src-grp */
764 				rt->mf6c_pkt_cnt    = 0;
765 				rt->mf6c_byte_cnt   = 0;
766 				rt->mf6c_wrong_if   = 0;
767 
768 				if (rt->mf6c_expire)
769 					nexpire[hash]--;
770 				rt->mf6c_expire	   = 0;
771 			}
772 		}
773 		if (rt == NULL) {
774 			/* no upcall, so make a new entry */
775 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
776 						  M_NOWAIT);
777 			if (rt == NULL) {
778 				splx(s);
779 				return ENOBUFS;
780 			}
781 
782 			/* insert new entry at head of hash chain */
783 			rt->mf6c_origin     = mfccp->mf6cc_origin;
784 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
785 			rt->mf6c_parent     = mfccp->mf6cc_parent;
786 			/* initialize pkt counters per src-grp */
787 			rt->mf6c_pkt_cnt    = 0;
788 			rt->mf6c_byte_cnt   = 0;
789 			rt->mf6c_wrong_if   = 0;
790 			rt->mf6c_expire     = 0;
791 			rt->mf6c_stall = NULL;
792 
793 			/* link into table */
794 			rt->mf6c_next  = mf6ctable[hash];
795 			mf6ctable[hash] = rt;
796 		}
797 	}
798 	splx(s);
799 	return 0;
800 }
801 
802 #ifdef UPCALL_TIMING
803 /*
804  * collect delay statistics on the upcalls
805  */
806 static void
807 collate(t)
808 	register struct timeval *t;
809 {
810 	register u_long d;
811 	register struct timeval tp;
812 	register u_long delta;
813 
814 	GET_TIME(tp);
815 
816 	if (TV_LT(*t, tp))
817 	{
818 		TV_DELTA(tp, *t, delta);
819 
820 		d = delta >> 10;
821 		if (d > UPCALL_MAX)
822 			d = UPCALL_MAX;
823 
824 		++upcall_data[d];
825 	}
826 }
827 #endif /* UPCALL_TIMING */
828 
829 /*
830  * Delete an mfc entry
831  */
832 static int
833 del_m6fc(mfccp)
834 	struct mf6cctl *mfccp;
835 {
836 	struct sockaddr_in6 	origin;
837 	struct sockaddr_in6 	mcastgrp;
838 	struct mf6c 		*rt;
839 	struct mf6c	 	**nptr;
840 	u_long 		hash;
841 	int s;
842 
843 	origin = mfccp->mf6cc_origin;
844 	mcastgrp = mfccp->mf6cc_mcastgrp;
845 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
846 
847 #ifdef MRT6DEBUG
848 	if (mrt6debug & DEBUG_MFC)
849 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
850 		    ip6_sprintf(&origin.sin6_addr),
851 		    ip6_sprintf(&mcastgrp.sin6_addr));
852 #endif
853 
854 	s = splnet();
855 
856 	nptr = &mf6ctable[hash];
857 	while ((rt = *nptr) != NULL) {
858 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
859 				       &rt->mf6c_origin.sin6_addr) &&
860 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
861 				       &rt->mf6c_mcastgrp.sin6_addr) &&
862 		    rt->mf6c_stall == NULL)
863 			break;
864 
865 		nptr = &rt->mf6c_next;
866 	}
867 	if (rt == NULL) {
868 		splx(s);
869 		return EADDRNOTAVAIL;
870 	}
871 
872 	*nptr = rt->mf6c_next;
873 	free(rt, M_MRTABLE);
874 
875 	splx(s);
876 
877 	return 0;
878 }
879 
880 static int
881 socket_send(s, mm, src)
882 	struct socket *s;
883 	struct mbuf *mm;
884 	struct sockaddr_in6 *src;
885 {
886 	if (s) {
887 		if (sbappendaddr(&s->so_rcv,
888 				 (struct sockaddr *)src,
889 				 mm, (struct mbuf *)0) != 0) {
890 			sorwakeup(s);
891 			return 0;
892 		}
893 	}
894 	m_freem(mm);
895 	return -1;
896 }
897 
898 /*
899  * IPv6 multicast forwarding function. This function assumes that the packet
900  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
901  * pointed to by "ifp", and the packet is to be relayed to other networks
902  * that have members of the packet's destination IPv6 multicast group.
903  *
904  * The packet is returned unscathed to the caller, unless it is
905  * erroneous, in which case a non-zero return value tells the caller to
906  * discard it.
907  */
908 
909 int
910 ip6_mforward(ip6, ifp, m)
911 	register struct ip6_hdr *ip6;
912 	struct ifnet *ifp;
913 	struct mbuf *m;
914 {
915 	register struct mf6c *rt;
916 	register struct mif6 *mifp;
917 	register struct mbuf *mm;
918 	int s;
919 	mifi_t mifi;
920 
921 #ifdef MRT6DEBUG
922 	if (mrt6debug & DEBUG_FORWARD)
923 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
924 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
925 		    ifp->if_index);
926 #endif
927 
928 	/*
929 	 * Don't forward a packet with Hop limit of zero or one,
930 	 * or a packet destined to a local-only group.
931 	 */
932 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
933 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
934 		return 0;
935 	ip6->ip6_hlim--;
936 
937 	/*
938 	 * Determine forwarding mifs from the forwarding cache table
939 	 */
940 	s = splnet();
941 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
942 
943 	/* Entry exists, so forward if necessary */
944 	if (rt) {
945 		splx(s);
946 		return (ip6_mdq(m, ifp, rt));
947 	} else {
948 		/*
949 		 * If we don't have a route for packet's origin,
950 		 * Make a copy of the packet &
951 		 * send message to routing daemon
952 		 */
953 
954 		register struct mbuf *mb0;
955 		register struct rtdetq *rte;
956 		register u_long hash;
957 /*	register int i, npkts;*/
958 #ifdef UPCALL_TIMING
959 		struct timeval tp;
960 
961 		GET_TIME(tp);
962 #endif /* UPCALL_TIMING */
963 
964 		mrt6stat.mrt6s_no_route++;
965 #ifdef MRT6DEBUG
966 		if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
967 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
968 			    ip6_sprintf(&ip6->ip6_src),
969 			    ip6_sprintf(&ip6->ip6_dst));
970 #endif
971 
972 		/*
973 		 * Allocate mbufs early so that we don't do extra work if we
974 		 * are just going to fail anyway.
975 		 */
976 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
977 					      M_NOWAIT);
978 		if (rte == NULL) {
979 			splx(s);
980 			return ENOBUFS;
981 		}
982 		mb0 = m_copy(m, 0, M_COPYALL);
983 		/*
984 		 * Pullup packet header if needed before storing it,
985 		 * as other references may modify it in the meantime.
986 		 */
987 		if (mb0 &&
988 		    (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
989 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
990 		if (mb0 == NULL) {
991 			free(rte, M_MRTABLE);
992 			splx(s);
993 			return ENOBUFS;
994 		}
995 
996 		/* is there an upcall waiting for this packet? */
997 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
998 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
999 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1000 					       &rt->mf6c_origin.sin6_addr) &&
1001 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1002 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1003 			    (rt->mf6c_stall != NULL))
1004 				break;
1005 		}
1006 
1007 		if (rt == NULL) {
1008 			struct mrt6msg *im;
1009 #ifdef MRT6_OINIT
1010 			struct omrt6msg *oim;
1011 #endif
1012 
1013 			/* no upcall, so make a new entry */
1014 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1015 						  M_NOWAIT);
1016 			if (rt == NULL) {
1017 				free(rte, M_MRTABLE);
1018 				m_freem(mb0);
1019 				splx(s);
1020 				return ENOBUFS;
1021 			}
1022 			/*
1023 			 * Make a copy of the header to send to the user
1024 			 * level process
1025 			 */
1026 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1027 
1028 			if (mm == NULL) {
1029 				free(rte, M_MRTABLE);
1030 				m_freem(mb0);
1031 				free(rt, M_MRTABLE);
1032 				splx(s);
1033 				return ENOBUFS;
1034 			}
1035 
1036 			/*
1037 			 * Send message to routing daemon
1038 			 */
1039 			sin6.sin6_addr = ip6->ip6_src;
1040 
1041 			im = NULL;
1042 #ifdef MRT6_OINIT
1043 			oim = NULL;
1044 #endif
1045 			switch (ip6_mrouter_ver) {
1046 #ifdef MRT6_OINIT
1047 			case MRT6_OINIT:
1048 				oim = mtod(mm, struct omrt6msg *);
1049 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1050 				oim->im6_mbz = 0;
1051 				break;
1052 #endif
1053 			case MRT6_INIT:
1054 				im = mtod(mm, struct mrt6msg *);
1055 				im->im6_msgtype = MRT6MSG_NOCACHE;
1056 				im->im6_mbz = 0;
1057 				break;
1058 			default:
1059 				free(rte, M_MRTABLE);
1060 				m_freem(mb0);
1061 				free(rt, M_MRTABLE);
1062 				splx(s);
1063 				return EINVAL;
1064 			}
1065 
1066 #ifdef MRT6DEBUG
1067 			if (mrt6debug & DEBUG_FORWARD)
1068 				log(LOG_DEBUG,
1069 				    "getting the iif info in the kernel\n");
1070 #endif
1071 
1072 			for (mifp = mif6table, mifi = 0;
1073 			     mifi < nummifs && mifp->m6_ifp != ifp;
1074 			     mifp++, mifi++)
1075 				;
1076 
1077 			switch (ip6_mrouter_ver) {
1078 #ifdef MRT6_OINIT
1079 			case MRT6_OINIT:
1080 				oim->im6_mif = mifi;
1081 				break;
1082 #endif
1083 			case MRT6_INIT:
1084 				im->im6_mif = mifi;
1085 				break;
1086 			}
1087 
1088 			if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1089 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1090 				    "socket queue full\n");
1091 				mrt6stat.mrt6s_upq_sockfull++;
1092 				free(rte, M_MRTABLE);
1093 				m_freem(mb0);
1094 				free(rt, M_MRTABLE);
1095 				splx(s);
1096 				return ENOBUFS;
1097 			}
1098 
1099 			mrt6stat.mrt6s_upcalls++;
1100 
1101 			/* insert new entry at head of hash chain */
1102 			bzero(rt, sizeof(*rt));
1103 			rt->mf6c_origin.sin6_family = AF_INET6;
1104 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1105 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1106 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1107 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1108 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1109 			rt->mf6c_expire = UPCALL_EXPIRE;
1110 			nexpire[hash]++;
1111 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1112 
1113 			/* link into table */
1114 			rt->mf6c_next  = mf6ctable[hash];
1115 			mf6ctable[hash] = rt;
1116 			/* Add this entry to the end of the queue */
1117 			rt->mf6c_stall = rte;
1118 		} else {
1119 			/* determine if q has overflowed */
1120 			struct rtdetq **p;
1121 			register int npkts = 0;
1122 
1123 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1124 				if (++npkts > MAX_UPQ6) {
1125 					mrt6stat.mrt6s_upq_ovflw++;
1126 					free(rte, M_MRTABLE);
1127 					m_freem(mb0);
1128 					splx(s);
1129 					return 0;
1130 				}
1131 
1132 			/* Add this entry to the end of the queue */
1133 			*p = rte;
1134 		}
1135 
1136 		rte->next = NULL;
1137 		rte->m = mb0;
1138 		rte->ifp = ifp;
1139 #ifdef UPCALL_TIMING
1140 		rte->t = tp;
1141 #endif /* UPCALL_TIMING */
1142 
1143 		splx(s);
1144 
1145 		return 0;
1146 	}
1147 }
1148 
1149 /*
1150  * Clean up cache entries if upcalls are not serviced
1151  * Call from the Slow Timeout mechanism, every half second.
1152  */
1153 static void
1154 expire_upcalls(unused)
1155 	void *unused;
1156 {
1157 	struct rtdetq *rte;
1158 	struct mf6c *mfc, **nptr;
1159 	int i;
1160 	int s;
1161 
1162 	s = splnet();
1163 	for (i = 0; i < MF6CTBLSIZ; i++) {
1164 		if (nexpire[i] == 0)
1165 			continue;
1166 		nptr = &mf6ctable[i];
1167 		while ((mfc = *nptr) != NULL) {
1168 			rte = mfc->mf6c_stall;
1169 			/*
1170 			 * Skip real cache entries
1171 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1172 			 * If it expires now
1173 			 */
1174 			if (rte != NULL &&
1175 			    mfc->mf6c_expire != 0 &&
1176 			    --mfc->mf6c_expire == 0) {
1177 #ifdef MRT6DEBUG
1178 				if (mrt6debug & DEBUG_EXPIRE)
1179 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1180 					    ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1181 					    ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1182 #endif
1183 				/*
1184 				 * drop all the packets
1185 				 * free the mbuf with the pkt, if, timing info
1186 				 */
1187 				do {
1188 					struct rtdetq *n = rte->next;
1189 					m_freem(rte->m);
1190 					free(rte, M_MRTABLE);
1191 					rte = n;
1192 				} while (rte != NULL);
1193 				mrt6stat.mrt6s_cache_cleanups++;
1194 				nexpire[i]--;
1195 
1196 				*nptr = mfc->mf6c_next;
1197 				free(mfc, M_MRTABLE);
1198 			} else {
1199 				nptr = &mfc->mf6c_next;
1200 			}
1201 		}
1202 	}
1203 	splx(s);
1204 	expire_upcalls_ch =
1205 	timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
1206 }
1207 
1208 /*
1209  * Packet forwarding routine once entry in the cache is made
1210  */
1211 static int
1212 ip6_mdq(m, ifp, rt)
1213 	register struct mbuf *m;
1214 	register struct ifnet *ifp;
1215 	register struct mf6c *rt;
1216 {
1217 	register struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1218 	register mifi_t mifi, iif;
1219 	register struct mif6 *mifp;
1220 	register int plen = m->m_pkthdr.len;
1221 
1222 /*
1223  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1224  * input, they shouldn't get counted on output, so statistics keeping is
1225  * seperate.
1226  */
1227 
1228 #define MC6_SEND(ip6, mifp, m) do {				\
1229 		if ((mifp)->m6_flags & MIFF_REGISTER)		\
1230 		    register_send((ip6), (mifp), (m));		\
1231 		else						\
1232 		    phyint_send((ip6), (mifp), (m));		\
1233 } while (0)
1234 
1235 	/*
1236 	 * Don't forward if it didn't arrive from the parent mif
1237 	 * for its origin.
1238 	 */
1239 	mifi = rt->mf6c_parent;
1240 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1241 		/* came in the wrong interface */
1242 #ifdef MRT6DEBUG
1243 		if (mrt6debug & DEBUG_FORWARD)
1244 			log(LOG_DEBUG,
1245 			    "wrong if: ifid %d mifi %d mififid %x\n",
1246 			    ifp->if_index, mifi,
1247 			    mif6table[mifi].m6_ifp->if_index);
1248 #endif
1249 		mrt6stat.mrt6s_wrong_if++;
1250 		rt->mf6c_wrong_if++;
1251 		/*
1252 		 * If we are doing PIM processing, and we are forwarding
1253 		 * packets on this interface, send a message to the
1254 		 * routing daemon.
1255 		 */
1256 		/* have to make sure this is a valid mif */
1257 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1258 			if (pim6 && (m->m_flags & M_LOOP) == 0) {
1259 				/*
1260 				 * Check the M_LOOP flag to avoid an
1261 				 * unnecessary PIM assert.
1262 				 * XXX: M_LOOP is an ad-hoc hack...
1263 				 */
1264 				static struct sockaddr_in6 sin6 =
1265 				{ sizeof(sin6), AF_INET6 };
1266 
1267 				register struct mbuf *mm;
1268 				struct mrt6msg *im;
1269 #ifdef MRT6_OINIT
1270 				struct omrt6msg *oim;
1271 #endif
1272 
1273 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1274 				if (mm &&
1275 				    (M_HASCL(mm) ||
1276 				     mm->m_len < sizeof(struct ip6_hdr)))
1277 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1278 				if (mm == NULL)
1279 					return ENOBUFS;
1280 
1281 #ifdef MRT6_OINIT
1282 				oim = NULL;
1283 #endif
1284 				im = NULL;
1285 				switch (ip6_mrouter_ver) {
1286 #ifdef MRT6_OINIT
1287 				case MRT6_OINIT:
1288 					oim = mtod(mm, struct omrt6msg *);
1289 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1290 					oim->im6_mbz = 0;
1291 					break;
1292 #endif
1293 				case MRT6_INIT:
1294 					im = mtod(mm, struct mrt6msg *);
1295 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1296 					break;
1297 				default:
1298 					m_freem(mm);
1299 					return EINVAL;
1300 				}
1301 
1302 				for (mifp = mif6table, iif = 0;
1303 				     iif < nummifs && mifp &&
1304 					     mifp->m6_ifp != ifp;
1305 				     mifp++, iif++)
1306 					;
1307 
1308 				switch (ip6_mrouter_ver) {
1309 #ifdef MRT6_OINIT
1310 				case MRT6_OINIT:
1311 					oim->im6_mif = iif;
1312 					sin6.sin6_addr = oim->im6_src;
1313 					break;
1314 #endif
1315 				case MRT6_INIT:
1316 					im->im6_mif = iif;
1317 					sin6.sin6_addr = im->im6_src;
1318 					break;
1319 				}
1320 
1321 				mrt6stat.mrt6s_upcalls++;
1322 
1323 				if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1324 #ifdef MRT6DEBUG
1325 					if (mrt6debug)
1326 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1327 #endif
1328 					++mrt6stat.mrt6s_upq_sockfull;
1329 					return ENOBUFS;
1330 				}	/* if socket Q full */
1331 			}		/* if PIM */
1332 		return 0;
1333 	}			/* if wrong iif */
1334 
1335 	/* If I sourced this packet, it counts as output, else it was input. */
1336 	if (m->m_pkthdr.rcvif == NULL) {
1337 		/* XXX: is rcvif really NULL when output?? */
1338 		mif6table[mifi].m6_pkt_out++;
1339 		mif6table[mifi].m6_bytes_out += plen;
1340 	} else {
1341 		mif6table[mifi].m6_pkt_in++;
1342 		mif6table[mifi].m6_bytes_in += plen;
1343 	}
1344 	rt->mf6c_pkt_cnt++;
1345 	rt->mf6c_byte_cnt += plen;
1346 
1347 	/*
1348 	 * For each mif, forward a copy of the packet if there are group
1349 	 * members downstream on the interface.
1350 	 */
1351 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1352 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1353 			/*
1354 			 * check if the outgoing packet is going to break
1355 			 * a scope boundary.
1356 			 * XXX For packets through PIM register tunnel
1357 			 * interface, we believe a routing daemon.
1358 			 */
1359 			if ((mif6table[rt->mf6c_parent].m6_flags &
1360 			     MIFF_REGISTER) == 0 &&
1361 			    (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1362 			    (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1363 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1364 					      &ip6->ip6_dst) ||
1365 			     in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1366 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1367 					      &ip6->ip6_src))) {
1368 				ip6stat.ip6s_badscope++;
1369 				continue;
1370 			}
1371 
1372 			mifp->m6_pkt_out++;
1373 			mifp->m6_bytes_out += plen;
1374 			MC6_SEND(ip6, mifp, m);
1375 		}
1376 	return 0;
1377 }
1378 
1379 static void
1380 phyint_send(ip6, mifp, m)
1381     struct ip6_hdr *ip6;
1382     struct mif6 *mifp;
1383     struct mbuf *m;
1384 {
1385 	register struct mbuf *mb_copy;
1386 	struct ifnet *ifp = mifp->m6_ifp;
1387 	int error = 0;
1388 	int s = splnet();
1389 	static struct route_in6 ro6;
1390 	struct	in6_multi *in6m;
1391 
1392 	/*
1393 	 * Make a new reference to the packet; make sure that
1394 	 * the IPv6 header is actually copied, not just referenced,
1395 	 * so that ip6_output() only scribbles on the copy.
1396 	 */
1397 	mb_copy = m_copy(m, 0, M_COPYALL);
1398 	if (mb_copy &&
1399 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1400 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1401 	if (mb_copy == NULL)
1402 		return;
1403 	/* set MCAST flag to the outgoing packet */
1404 	mb_copy->m_flags |= M_MCAST;
1405 
1406 	/*
1407 	 * If we sourced the packet, call ip6_output since we may devide
1408 	 * the packet into fragments when the packet is too big for the
1409 	 * outgoing interface.
1410 	 * Otherwise, we can simply send the packet to the interface
1411 	 * sending queue.
1412 	 */
1413 	if (m->m_pkthdr.rcvif == NULL) {
1414 		struct ip6_moptions im6o;
1415 
1416 		im6o.im6o_multicast_ifp = ifp;
1417 		/* XXX: ip6_output will override ip6->ip6_hlim */
1418 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1419 		im6o.im6o_multicast_loop = 1;
1420 		error = ip6_output(mb_copy, NULL, &ro6,
1421 				   IPV6_FORWARDING, &im6o, NULL);
1422 
1423 #ifdef MRT6DEBUG
1424 		if (mrt6debug & DEBUG_XMIT)
1425 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1426 			    mifp - mif6table, error);
1427 #endif
1428 		splx(s);
1429 		return;
1430 	}
1431 
1432 	/*
1433 	 * If we belong to the destination multicast group
1434 	 * on the outgoing interface, loop back a copy.
1435 	 */
1436 	IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1437 	if (in6m != NULL) {
1438 		ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
1439 		ro6.ro_dst.sin6_family = AF_INET6;
1440 		ro6.ro_dst.sin6_addr = ip6->ip6_dst;
1441 		ip6_mloopback(ifp, m, &ro6.ro_dst);
1442 	}
1443 	/*
1444 	 * Put the packet into the sending queue of the outgoing interface
1445 	 * if it would fit in the MTU of the interface.
1446 	 */
1447 	if (mb_copy->m_pkthdr.len < ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
1448 		ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
1449 		ro6.ro_dst.sin6_family = AF_INET6;
1450 		ro6.ro_dst.sin6_addr = ip6->ip6_dst;
1451 		/*
1452 		 * We just call if_output instead of nd6_output here, since
1453 		 * we need no ND for a multicast forwarded packet...right?
1454 		 */
1455 		error = (*ifp->if_output)(ifp, mb_copy,
1456 					  (struct sockaddr *)&ro6.ro_dst,
1457 					  NULL);
1458 #ifdef MRT6DEBUG
1459 		if (mrt6debug & DEBUG_XMIT)
1460 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1461 			    mifp - mif6table, error);
1462 #endif
1463 	}
1464 	else {
1465 #ifdef MULTICAST_PMTUD
1466 		icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
1467 		return;
1468 #else
1469 #ifdef MRT6DEBUG
1470 		if (mrt6debug & DEBUG_XMIT)
1471 			log(LOG_DEBUG,
1472 			    "phyint_send: packet too big on %s%u o %s g %s"
1473 			    " size %d(discarded)\n",
1474 			    ifp->if_name, ifp->if_unit,
1475 			    ip6_sprintf(&ip6->ip6_src),
1476 			    ip6_sprintf(&ip6->ip6_dst),
1477 			    mb_copy->m_pkthdr.len);
1478 #endif /* MRT6DEBUG */
1479 		m_freem(mb_copy); /* simply discard the packet */
1480 		return;
1481 #endif
1482 	}
1483 }
1484 
1485 static int
1486 register_send(ip6, mif, m)
1487 	register struct ip6_hdr *ip6;
1488 	struct mif6 *mif;
1489 	register struct mbuf *m;
1490 {
1491 	register struct mbuf *mm;
1492 	register int i, len = m->m_pkthdr.len;
1493 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1494 	struct mrt6msg *im6;
1495 
1496 #ifdef MRT6DEBUG
1497 	if (mrt6debug)
1498 		log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1499 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1500 #endif
1501 	++pim6stat.pim6s_snd_registers;
1502 
1503 	/* Make a copy of the packet to send to the user level process */
1504 	MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1505 	if (mm == NULL)
1506 		return ENOBUFS;
1507 	mm->m_data += max_linkhdr;
1508 	mm->m_len = sizeof(struct ip6_hdr);
1509 
1510 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1511 		m_freem(mm);
1512 		return ENOBUFS;
1513 	}
1514 	i = MHLEN - M_LEADINGSPACE(mm);
1515 	if (i > len)
1516 		i = len;
1517 	mm = m_pullup(mm, i);
1518 	if (mm == NULL){
1519 		m_freem(mm);
1520 		return ENOBUFS;
1521 	}
1522 /* TODO: check it! */
1523 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1524 
1525 	/*
1526 	 * Send message to routing daemon
1527 	 */
1528 	sin6.sin6_addr = ip6->ip6_src;
1529 
1530 	im6 = mtod(mm, struct mrt6msg *);
1531 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1532 	im6->im6_mbz          = 0;
1533 
1534 	im6->im6_mif = mif - mif6table;
1535 
1536 	/* iif info is not given for reg. encap.n */
1537 	mrt6stat.mrt6s_upcalls++;
1538 
1539 	if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1540 #ifdef MRT6DEBUG
1541 		if (mrt6debug)
1542 			log(LOG_WARNING,
1543 			    "register_send: ip_mrouter socket queue full\n");
1544 #endif
1545                 ++mrt6stat.mrt6s_upq_sockfull;
1546                 return ENOBUFS;
1547 	}
1548 	return 0;
1549 }
1550 
1551 /*
1552  * PIM sparse mode hook
1553  * Receives the pim control messages, and passes them up to the listening
1554  * socket, using rip6_input.
1555  * The only message processed is the REGISTER pim message; the pim header
1556  * is stripped off, and the inner packet is passed to register_mforward.
1557  */
1558 int
1559 pim6_input(mp, offp, proto)
1560 	struct mbuf **mp;
1561 	int *offp, proto;
1562 {
1563         register struct pim *pim; /* pointer to a pim struct */
1564         register struct ip6_hdr *ip6;
1565         register int pimlen;
1566 	struct mbuf *m = *mp;
1567         int minlen;
1568 	int off = *offp;
1569 
1570 	++pim6stat.pim6s_rcv_total;
1571 
1572         ip6 = mtod(m, struct ip6_hdr *);
1573         pimlen = m->m_pkthdr.len - *offp;
1574 
1575         /*
1576          * Validate lengths
1577          */
1578 	if (pimlen < PIM_MINLEN) {
1579 		++pim6stat.pim6s_rcv_tooshort;
1580 #ifdef MRT6DEBUG
1581 		if (mrt6debug & DEBUG_PIM)
1582 			log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1583 #endif
1584 		m_freem(m);
1585 		return(IPPROTO_DONE);
1586 	}
1587 
1588 	/*
1589 	 * if the packet is at least as big as a REGISTER, go ahead
1590 	 * and grab the PIM REGISTER header size, to avoid another
1591 	 * possible m_pullup() later.
1592 	 *
1593 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1594 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1595 	 */
1596 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1597 
1598 	/*
1599 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1600 	 * possibly the PIM REGISTER header
1601 	 */
1602 #ifndef PULLDOWN_TEST
1603 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1604 	/* adjust pointer */
1605 	ip6 = mtod(m, struct ip6_hdr *);
1606 
1607 	/* adjust mbuf to point to the PIM header */
1608 	pim = (struct pim *)((caddr_t)ip6 + off);
1609 #else
1610 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1611 	if (pim == NULL) {
1612 		pim6stat.pim6s_rcv_tooshort++;
1613 		return IPPROTO_DONE;
1614 	}
1615 #endif
1616 
1617 #define PIM6_CHECKSUM
1618 #ifdef PIM6_CHECKSUM
1619 	{
1620 		int cksumlen;
1621 
1622 		/*
1623 		 * Validate checksum.
1624 		 * If PIM REGISTER, exclude the data packet
1625 		 */
1626 		if (pim->pim_type == PIM_REGISTER)
1627 			cksumlen = PIM_MINLEN;
1628 		else
1629 			cksumlen = pimlen;
1630 
1631 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1632 			++pim6stat.pim6s_rcv_badsum;
1633 #ifdef MRT6DEBUG
1634 			if (mrt6debug & DEBUG_PIM)
1635 				log(LOG_DEBUG,
1636 				    "pim6_input: invalid checksum\n");
1637 #endif
1638 			m_freem(m);
1639 			return(IPPROTO_DONE);
1640 		}
1641 	}
1642 #endif /* PIM_CHECKSUM */
1643 
1644 	/* PIM version check */
1645 	if (pim->pim_ver != PIM_VERSION) {
1646 		++pim6stat.pim6s_rcv_badversion;
1647 #ifdef MRT6DEBUG
1648 		log(LOG_ERR,
1649 		    "pim6_input: incorrect version %d, expecting %d\n",
1650 		    pim->pim_ver, PIM_VERSION);
1651 #endif
1652 		m_freem(m);
1653 		return(IPPROTO_DONE);
1654 	}
1655 
1656 	if (pim->pim_type == PIM_REGISTER) {
1657 		/*
1658 		 * since this is a REGISTER, we'll make a copy of the register
1659 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1660 		 * routing daemon.
1661 		 */
1662 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1663 
1664 		struct mbuf *mcp;
1665 		struct ip6_hdr *eip6;
1666 		u_int32_t *reghdr;
1667 		int rc;
1668 
1669 		++pim6stat.pim6s_rcv_registers;
1670 
1671 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1672 #ifdef MRT6DEBUG
1673 			if (mrt6debug & DEBUG_PIM)
1674 				log(LOG_DEBUG,
1675 				    "pim6_input: register mif not set: %d\n",
1676 				    reg_mif_num);
1677 #endif
1678 			m_freem(m);
1679 			return(IPPROTO_DONE);
1680 		}
1681 
1682 		reghdr = (u_int32_t *)(pim + 1);
1683 
1684 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1685 			goto pim6_input_to_daemon;
1686 
1687 		/*
1688 		 * Validate length
1689 		 */
1690 		if (pimlen < PIM6_REG_MINLEN) {
1691 			++pim6stat.pim6s_rcv_tooshort;
1692 			++pim6stat.pim6s_rcv_badregisters;
1693 #ifdef MRT6DEBUG
1694 			log(LOG_ERR,
1695 			    "pim6_input: register packet size too "
1696 			    "small %d from %s\n",
1697 			    pimlen, ip6_sprintf(&ip6->ip6_src));
1698 #endif
1699 			m_freem(m);
1700 			return(IPPROTO_DONE);
1701 		}
1702 
1703 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1704 #ifdef MRT6DEBUG
1705 		if (mrt6debug & DEBUG_PIM)
1706 			log(LOG_DEBUG,
1707 			    "pim6_input[register], eip6: %s -> %s, "
1708 			    "eip6 plen %d\n",
1709 			    ip6_sprintf(&eip6->ip6_src),
1710 			    ip6_sprintf(&eip6->ip6_dst),
1711 			    ntohs(eip6->ip6_plen));
1712 #endif
1713 
1714 		/* verify the inner packet is destined to a mcast group */
1715 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1716 			++pim6stat.pim6s_rcv_badregisters;
1717 #ifdef MRT6DEBUG
1718 			if (mrt6debug & DEBUG_PIM)
1719 				log(LOG_DEBUG,
1720 				    "pim6_input: inner packet of register "
1721 				    "is not multicast %s\n",
1722 				    ip6_sprintf(&eip6->ip6_dst));
1723 #endif
1724 			m_freem(m);
1725 			return(IPPROTO_DONE);
1726 		}
1727 
1728 		/*
1729 		 * make a copy of the whole header to pass to the daemon later.
1730 		 */
1731 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1732 		if (mcp == NULL) {
1733 #ifdef MRT6DEBUG
1734 			log(LOG_ERR,
1735 			    "pim6_input: pim register: "
1736 			    "could not copy register head\n");
1737 #endif
1738 			m_freem(m);
1739 			return(IPPROTO_DONE);
1740 		}
1741 
1742 		/*
1743 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1744 		 */
1745 		m_adj(m, off + PIM_MINLEN);
1746 #ifdef MRT6DEBUG
1747 		if (mrt6debug & DEBUG_PIM) {
1748 			log(LOG_DEBUG,
1749 			    "pim6_input: forwarding decapsulated register: "
1750 			    "src %s, dst %s, mif %d\n",
1751 			    ip6_sprintf(&eip6->ip6_src),
1752 			    ip6_sprintf(&eip6->ip6_dst),
1753 			    reg_mif_num);
1754 		}
1755 #endif
1756 
1757  		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1758 			       dst.sin6_family, NULL);
1759 
1760 		/* prepare the register head to send to the mrouting daemon */
1761 		m = mcp;
1762 	}
1763 
1764 	/*
1765 	 * Pass the PIM message up to the daemon; if it is a register message
1766 	 * pass the 'head' only up to the daemon. This includes the
1767 	 * encapsulator ip6 header, pim header, register header and the
1768 	 * encapsulated ip6 header.
1769 	 */
1770   pim6_input_to_daemon:
1771 	rip6_input(&m, offp, proto);
1772 	return(IPPROTO_DONE);
1773 }
1774