xref: /netbsd-src/sys/netinet6/ip6_mroute.c (revision 3b01aba77a7a698587faaae455bbfe740923c1f5)
1 /*	$NetBSD: ip6_mroute.c,v 1.22 2001/07/25 09:23:46 itojun Exp $	*/
2 /*	$KAME: ip6_mroute.c,v 1.49 2001/07/25 09:21:18 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*	BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp	*/
34 
35 /*
36  * 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 
50 #ifndef _KERNEL
51 # ifdef KERNEL
52 #  define _KERNEL
53 # endif
54 #endif
55 
56 #include <sys/param.h>
57 #include <sys/systm.h>
58 #include <sys/callout.h>
59 #include <sys/mbuf.h>
60 #include <sys/socket.h>
61 #include <sys/socketvar.h>
62 #include <sys/sockio.h>
63 #include <sys/protosw.h>
64 #include <sys/errno.h>
65 #include <sys/time.h>
66 #include <sys/kernel.h>
67 #include <sys/ioctl.h>
68 #include <sys/syslog.h>
69 
70 #include <net/if.h>
71 #include <net/route.h>
72 #include <net/raw_cb.h>
73 
74 #include <netinet/in.h>
75 #include <netinet/in_var.h>
76 
77 #include <netinet/ip6.h>
78 #include <netinet6/ip6_var.h>
79 #include <netinet6/ip6_mroute.h>
80 #include <netinet6/pim6.h>
81 #include <netinet6/pim6_var.h>
82 
83 #include <net/net_osdep.h>
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 get_pim6 __P((struct mbuf *));
92 static int socket_send __P((struct socket *, struct mbuf *,
93 			    struct sockaddr_in6 *));
94 static int register_send __P((struct ip6_hdr *, struct mif6 *,
95 			      struct mbuf *));
96 
97 /*
98  * Globals.  All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static,
99  * except for netstat or debugging purposes.
100  */
101 struct socket  *ip6_mrouter  = NULL;
102 int		ip6_mrouter_ver = 0;
103 int		ip6_mrtproto = IPPROTO_PIM;    /* for netstat only */
104 struct mrt6stat	mrt6stat;
105 
106 #define NO_RTE_FOUND 	0x1
107 #define RTE_FOUND	0x2
108 
109 struct mf6c	*mf6ctable[MF6CTBLSIZ];
110 u_char		nexpire[MF6CTBLSIZ];
111 static struct mif6 mif6table[MAXMIFS];
112 #ifdef MRT6DEBUG
113 u_int		mrt6debug = 0;	  /* debug level 	*/
114 #define		DEBUG_MFC	0x02
115 #define		DEBUG_FORWARD	0x04
116 #define		DEBUG_EXPIRE	0x08
117 #define		DEBUG_XMIT	0x10
118 #define         DEBUG_REG       0x20
119 #define         DEBUG_PIM       0x40
120 #endif
121 
122 static void	expire_upcalls __P((void *));
123 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
124 #define		UPCALL_EXPIRE	6		/* number of timeouts */
125 
126 #ifdef INET
127 #ifdef MROUTING
128 extern struct socket *ip_mrouter;
129 #endif
130 #endif
131 
132 /*
133  * 'Interfaces' associated with decapsulator (so we can tell
134  * packets that went through it from ones that get reflected
135  * by a broken gateway).  These interfaces are never linked into
136  * the system ifnet list & no routes point to them.  I.e., packets
137  * can't be sent this way.  They only exist as a placeholder for
138  * multicast source verification.
139  */
140 struct ifnet multicast_register_if;
141 
142 #define ENCAP_HOPS 64
143 
144 /*
145  * Private variables.
146  */
147 static mifi_t nummifs = 0;
148 static mifi_t reg_mif_num = (mifi_t)-1;
149 
150 static struct pim6stat pim6stat;
151 
152 /*
153  * one-back cache used by ipip_input to locate a tunnel's mif
154  * given a datagram's src ip address.
155  */
156 static int pim6;
157 
158 /*
159  * Hash function for a source, group entry
160  */
161 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
162 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
163 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
164 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
165 
166 /*
167  * Find a route for a given origin IPv6 address and Multicast group address.
168  * Quality of service parameter to be added in the future!!!
169  */
170 
171 #define MF6CFIND(o, g, rt) do { \
172 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
173 	rt = NULL; \
174 	mrt6stat.mrt6s_mfc_lookups++; \
175 	while (_rt) { \
176 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
177 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
178 		    (_rt->mf6c_stall == NULL)) { \
179 			rt = _rt; \
180 			break; \
181 		} \
182 		_rt = _rt->mf6c_next; \
183 	} \
184 	if (rt == NULL) { \
185 		mrt6stat.mrt6s_mfc_misses++; \
186 	} \
187 } while (0)
188 
189 /*
190  * Macros to compute elapsed time efficiently
191  * Borrowed from Van Jacobson's scheduling code
192  */
193 #define TV_DELTA(a, b, delta) do { \
194 	    int xxs; \
195 		\
196 	    delta = (a).tv_usec - (b).tv_usec; \
197 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
198 	       switch (xxs) { \
199 		      case 2: \
200 			  delta += 1000000; \
201 			      /* fall through */ \
202 		      case 1: \
203 			  delta += 1000000; \
204 			  break; \
205 		      default: \
206 			  delta += (1000000 * xxs); \
207 	       } \
208 	    } \
209 } while (0)
210 
211 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
212 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
213 
214 #ifdef UPCALL_TIMING
215 #define UPCALL_MAX	50
216 u_long upcall_data[UPCALL_MAX + 1];
217 static void collate();
218 #endif /* UPCALL_TIMING */
219 
220 static int get_sg_cnt __P((struct sioc_sg_req6 *));
221 static int get_mif6_cnt __P((struct sioc_mif_req6 *));
222 static int ip6_mrouter_init __P((struct socket *, struct mbuf *, int));
223 static int add_m6if __P((struct mif6ctl *));
224 static int del_m6if __P((mifi_t *));
225 static int add_m6fc __P((struct mf6cctl *));
226 static int del_m6fc __P((struct mf6cctl *));
227 
228 static struct callout expire_upcalls_ch = CALLOUT_INITIALIZER;
229 
230 /*
231  * Handle MRT setsockopt commands to modify the multicast routing tables.
232  */
233 int
234 ip6_mrouter_set(cmd, so, m)
235 	int cmd;
236 	struct socket *so;
237 	struct mbuf *m;
238 {
239 	if (cmd != MRT6_INIT && so != ip6_mrouter)
240 		return EACCES;
241 
242 	switch (cmd) {
243 	 case MRT6_OINIT:     return ip6_mrouter_init(so, m, cmd);
244 	 case MRT6_INIT:      return ip6_mrouter_init(so, m, cmd);
245 	 case MRT6_DONE:      return ip6_mrouter_done();
246 	 case MRT6_ADD_MIF:   return add_m6if(mtod(m, struct mif6ctl *));
247 	 case MRT6_DEL_MIF:   return del_m6if(mtod(m, mifi_t *));
248 	 case MRT6_ADD_MFC:   return add_m6fc(mtod(m, struct mf6cctl *));
249 	 case MRT6_DEL_MFC:   return del_m6fc(mtod(m, struct mf6cctl *));
250 	 case MRT6_PIM:       return set_pim6(mtod(m, int *));
251 	 default:            return EOPNOTSUPP;
252 	}
253 }
254 
255 /*
256  * Handle MRT getsockopt commands
257  */
258 int
259 ip6_mrouter_get(cmd, so, m)
260 	int cmd;
261 	struct socket *so;
262 	struct mbuf **m;
263 {
264 	struct mbuf *mb;
265 
266 	if (so != ip6_mrouter) return EACCES;
267 
268 	*m = mb = m_get(M_WAIT, MT_SOOPTS);
269 
270 	switch (cmd) {
271 	 case MRT6_PIM:       return get_pim6(mb);
272 	 default:
273 		 m_free(mb);
274 		 return EOPNOTSUPP;
275 	}
276 }
277 
278 /*
279  * Handle ioctl commands to obtain information from the cache
280  */
281 int
282 mrt6_ioctl(cmd, data)
283 	int cmd;
284 	caddr_t data;
285 {
286     int error = 0;
287 
288     switch (cmd) {
289      case SIOCGETSGCNT_IN6:
290 	     return(get_sg_cnt((struct sioc_sg_req6 *)data));
291 	     break;		/* for safety */
292      case SIOCGETMIFCNT_IN6:
293 	     return(get_mif6_cnt((struct sioc_mif_req6 *)data));
294 	     break;		/* for safety */
295      default:
296 	     return (EINVAL);
297 	     break;
298     }
299     return error;
300 }
301 
302 /*
303  * returns the packet, byte, rpf-failure count for the source group provided
304  */
305 static int
306 get_sg_cnt(req)
307 	struct sioc_sg_req6 *req;
308 {
309 	struct mf6c *rt;
310 	int s;
311 
312 	s = splsoftnet();
313 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
314 	splx(s);
315 	if (rt != NULL) {
316 		req->pktcnt = rt->mf6c_pkt_cnt;
317 		req->bytecnt = rt->mf6c_byte_cnt;
318 		req->wrong_if = rt->mf6c_wrong_if;
319 	} else
320 		return(ESRCH);
321 #if 0
322 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
323 #endif
324 
325 	return 0;
326 }
327 
328 /*
329  * returns the input and output packet and byte counts on the mif provided
330  */
331 static int
332 get_mif6_cnt(req)
333 	struct sioc_mif_req6 *req;
334 {
335 	mifi_t mifi = req->mifi;
336 
337 	if (mifi >= nummifs)
338 		return EINVAL;
339 
340 	req->icount = mif6table[mifi].m6_pkt_in;
341 	req->ocount = mif6table[mifi].m6_pkt_out;
342 	req->ibytes = mif6table[mifi].m6_bytes_in;
343 	req->obytes = mif6table[mifi].m6_bytes_out;
344 
345 	return 0;
346 }
347 
348 /*
349  * Get PIM processiong global
350  */
351 static int
352 get_pim6(m)
353 	struct mbuf *m;
354 {
355 	int *i;
356 
357 	i = mtod(m, int *);
358 
359 	*i = pim6;
360 
361 	return 0;
362 }
363 
364 static int
365 set_pim6(i)
366 	int *i;
367 {
368 	if ((*i != 1) && (*i != 0))
369 		return EINVAL;
370 
371 	pim6 = *i;
372 
373 	return 0;
374 }
375 
376 /*
377  * Enable multicast routing
378  */
379 static int
380 ip6_mrouter_init(so, m, cmd)
381 	struct socket *so;
382 	struct mbuf *m;
383 	int cmd;
384 {
385 	int *v;
386 
387 #ifdef MRT6DEBUG
388 	if (mrt6debug)
389 		log(LOG_DEBUG,
390 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
391 		    so->so_type, so->so_proto->pr_protocol);
392 #endif
393 
394 	if (so->so_type != SOCK_RAW ||
395 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
396 		return EOPNOTSUPP;
397 
398 	if (!m || (m->m_len != sizeof(int *)))
399 		return ENOPROTOOPT;
400 
401 	v = mtod(m, int *);
402 	if (*v != 1)
403 		return ENOPROTOOPT;
404 
405 	if (ip6_mrouter != NULL) return EADDRINUSE;
406 
407 	ip6_mrouter = so;
408 	ip6_mrouter_ver = cmd;
409 
410 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
411 	bzero((caddr_t)nexpire, sizeof(nexpire));
412 
413 	pim6 = 0;/* used for stubbing out/in pim stuff */
414 
415 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
416 	    expire_upcalls, NULL);
417 
418 #ifdef MRT6DEBUG
419 	if (mrt6debug)
420 		log(LOG_DEBUG, "ip6_mrouter_init\n");
421 #endif
422 
423 	return 0;
424 }
425 
426 /*
427  * Disable multicast routing
428  */
429 int
430 ip6_mrouter_done()
431 {
432 	mifi_t mifi;
433 	int i;
434 	struct ifnet *ifp;
435 	struct in6_ifreq ifr;
436 	struct mf6c *rt;
437 	struct rtdetq *rte;
438 	int s;
439 
440 	s = splsoftnet();
441 
442 	/*
443 	 * For each phyint in use, disable promiscuous reception of all IPv6
444 	 * multicasts.
445 	 */
446 #ifdef INET
447 #ifdef MROUTING
448 	/*
449 	 * If there is still IPv4 multicast routing daemon,
450 	 * we remain interfaces to receive all muliticasted packets.
451 	 * XXX: there may be an interface in which the IPv4 multicast
452 	 * daemon is not interested...
453 	 */
454 	if (!ip_mrouter)
455 #endif
456 #endif
457 	{
458 		for (mifi = 0; mifi < nummifs; mifi++) {
459 			if (mif6table[mifi].m6_ifp &&
460 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
461 				ifr.ifr_addr.sin6_family = AF_INET6;
462 				ifr.ifr_addr.sin6_addr= in6addr_any;
463 				ifp = mif6table[mifi].m6_ifp;
464 				(*ifp->if_ioctl)(ifp, SIOCDELMULTI,
465 						 (caddr_t)&ifr);
466 			}
467 		}
468 	}
469 #ifdef notyet
470 	bzero((caddr_t)qtable, sizeof(qtable));
471 	bzero((caddr_t)tbftable, sizeof(tbftable));
472 #endif
473 	bzero((caddr_t)mif6table, sizeof(mif6table));
474 	nummifs = 0;
475 
476 	pim6 = 0; /* used to stub out/in pim specific code */
477 
478 	callout_stop(&expire_upcalls_ch);
479 
480 	/*
481 	 * Free all multicast forwarding cache entries.
482 	 */
483 	for (i = 0; i < MF6CTBLSIZ; i++) {
484 		rt = mf6ctable[i];
485 		while (rt) {
486 			struct mf6c *frt;
487 
488 			for (rte = rt->mf6c_stall; rte != NULL; ) {
489 				struct rtdetq *n = rte->next;
490 
491 				m_free(rte->m);
492 				free(rte, M_MRTABLE);
493 				rte = n;
494 			}
495 			frt = rt;
496 			rt = rt->mf6c_next;
497 			free(frt, M_MRTABLE);
498 		}
499 	}
500 
501 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
502 
503 	/*
504 	 * Reset de-encapsulation cache
505 	 */
506 	reg_mif_num = -1;
507 
508 	ip6_mrouter = NULL;
509 	ip6_mrouter_ver = 0;
510 
511 	splx(s);
512 
513 #ifdef MRT6DEBUG
514 	if (mrt6debug)
515 		log(LOG_DEBUG, "ip6_mrouter_done\n");
516 #endif
517 
518 	return 0;
519 }
520 
521 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
522 
523 /*
524  * Add a mif to the mif table
525  */
526 static int
527 add_m6if(mifcp)
528 	struct mif6ctl *mifcp;
529 {
530 	struct mif6 *mifp;
531 	struct ifnet *ifp;
532 	struct in6_ifreq ifr;
533 	int error, s;
534 #ifdef notyet
535 	struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
536 #endif
537 
538 	if (mifcp->mif6c_mifi >= MAXMIFS)
539 		return EINVAL;
540 	mifp = mif6table + mifcp->mif6c_mifi;
541 	if (mifp->m6_ifp)
542 		return EADDRINUSE; /* XXX: is it appropriate? */
543 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
544 		return ENXIO;
545 	/*
546 	 * XXX: some OSes can remove ifp and clear ifindex2ifnet[id]
547 	 * even for id between 0 and if_index.
548 	 */
549 	if ((ifp = ifindex2ifnet[mifcp->mif6c_pifi]) == NULL)
550 		return ENXIO;
551 
552 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
553 		if (reg_mif_num == (mifi_t)-1) {
554 			strcpy(multicast_register_if.if_xname,
555 			       "register_mif"); /* XXX */
556 			multicast_register_if.if_flags |= IFF_LOOPBACK;
557 			multicast_register_if.if_index = mifcp->mif6c_mifi;
558 			reg_mif_num = mifcp->mif6c_mifi;
559 		}
560 
561 		ifp = &multicast_register_if;
562 
563 	} /* if REGISTER */
564 	else {
565 		/* Make sure the interface supports multicast */
566 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
567 			return EOPNOTSUPP;
568 
569 		s = splsoftnet();
570 		/*
571 		 * Enable promiscuous reception of all IPv6 multicasts
572 		 * from the interface.
573 		 */
574 		ifr.ifr_addr.sin6_family = AF_INET6;
575 		ifr.ifr_addr.sin6_addr = in6addr_any;
576 		error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
577 		splx(s);
578 		if (error)
579 			return error;
580 	}
581 
582 	s = splsoftnet();
583 	mifp->m6_flags     = mifcp->mif6c_flags;
584 	mifp->m6_ifp       = ifp;
585 #ifdef notyet
586 	/* scaling up here allows division by 1024 in critical code */
587 	mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
588 #endif
589 	/* initialize per mif pkt counters */
590 	mifp->m6_pkt_in    = 0;
591 	mifp->m6_pkt_out   = 0;
592 	mifp->m6_bytes_in  = 0;
593 	mifp->m6_bytes_out = 0;
594 	splx(s);
595 
596 	/* Adjust nummifs up if the mifi is higher than nummifs */
597 	if (nummifs <= mifcp->mif6c_mifi)
598 		nummifs = mifcp->mif6c_mifi + 1;
599 
600 #ifdef MRT6DEBUG
601 	if (mrt6debug)
602 		log(LOG_DEBUG,
603 		    "add_mif #%d, phyint %s%d\n",
604 		    mifcp->mif6c_mifi,
605 		    ifp->if_name, ifp->if_unit);
606 #endif
607 
608 	return 0;
609 }
610 
611 /*
612  * Delete a mif from the mif table
613  */
614 static int
615 del_m6if(mifip)
616 	mifi_t *mifip;
617 {
618 	struct mif6 *mifp = mif6table + *mifip;
619 	mifi_t mifi;
620 	struct ifnet *ifp;
621 	struct in6_ifreq ifr;
622 	int s;
623 
624 	if (*mifip >= nummifs)
625 		return EINVAL;
626 	if (mifp->m6_ifp == NULL)
627 		return EINVAL;
628 
629 	s = splsoftnet();
630 
631 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
632 		/*
633 		 * XXX: what if there is yet IPv4 multicast daemon
634 		 *      using the interface?
635 		 */
636 		ifp = mifp->m6_ifp;
637 
638 		ifr.ifr_addr.sin6_family = AF_INET6;
639 		ifr.ifr_addr.sin6_addr = in6addr_any;
640 		(*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
641 	}
642 
643 #ifdef notyet
644 	bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
645 	bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
646 #endif
647 	bzero((caddr_t)mifp, sizeof (*mifp));
648 
649 	/* Adjust nummifs down */
650 	for (mifi = nummifs; mifi > 0; mifi--)
651 		if (mif6table[mifi - 1].m6_ifp)
652 			break;
653 	nummifs = mifi;
654 
655 	splx(s);
656 
657 #ifdef MRT6DEBUG
658 	if (mrt6debug)
659 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
660 #endif
661 
662 	return 0;
663 }
664 
665 /*
666  * Add an mfc entry
667  */
668 static int
669 add_m6fc(mfccp)
670 	struct mf6cctl *mfccp;
671 {
672 	struct mf6c *rt;
673 	u_long hash;
674 	struct rtdetq *rte;
675 	u_short nstl;
676 	int s;
677 
678 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
679 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
680 
681 	/* If an entry already exists, just update the fields */
682 	if (rt) {
683 #ifdef MRT6DEBUG
684 		if (mrt6debug & DEBUG_MFC)
685 			log(LOG_DEBUG,"add_m6fc update o %s g %s p %x\n",
686 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
687 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
688 			    mfccp->mf6cc_parent);
689 #endif
690 
691 		s = splsoftnet();
692 		rt->mf6c_parent = mfccp->mf6cc_parent;
693 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
694 		splx(s);
695 		return 0;
696 	}
697 
698 	/*
699 	 * Find the entry for which the upcall was made and update
700 	 */
701 	s = splsoftnet();
702 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
703 			mfccp->mf6cc_mcastgrp.sin6_addr);
704 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
705 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
706 				       &mfccp->mf6cc_origin.sin6_addr) &&
707 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
708 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
709 		    (rt->mf6c_stall != NULL)) {
710 
711 			if (nstl++)
712 				log(LOG_ERR,
713 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
714 				    "multiple kernel entries",
715 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
716 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
717 				    mfccp->mf6cc_parent, rt->mf6c_stall);
718 
719 #ifdef MRT6DEBUG
720 			if (mrt6debug & DEBUG_MFC)
721 				log(LOG_DEBUG,
722 				    "add_m6fc o %s g %s p %x dbg %x\n",
723 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
724 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
725 				    mfccp->mf6cc_parent, rt->mf6c_stall);
726 #endif
727 
728 			rt->mf6c_origin     = mfccp->mf6cc_origin;
729 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
730 			rt->mf6c_parent     = mfccp->mf6cc_parent;
731 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
732 			/* initialize pkt counters per src-grp */
733 			rt->mf6c_pkt_cnt    = 0;
734 			rt->mf6c_byte_cnt   = 0;
735 			rt->mf6c_wrong_if   = 0;
736 
737 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
738 			nexpire[hash]--;
739 
740 			/* free packets Qed at the end of this entry */
741 			for (rte = rt->mf6c_stall; rte != NULL; ) {
742 				struct rtdetq *n = rte->next;
743 				ip6_mdq(rte->m, rte->ifp, rt);
744 				m_freem(rte->m);
745 #ifdef UPCALL_TIMING
746 				collate(&(rte->t));
747 #endif /* UPCALL_TIMING */
748 				free(rte, M_MRTABLE);
749 				rte = n;
750 			}
751 			rt->mf6c_stall = NULL;
752 		}
753 	}
754 
755 	/*
756 	 * It is possible that an entry is being inserted without an upcall
757 	 */
758 	if (nstl == 0) {
759 #ifdef MRT6DEBUG
760 		if (mrt6debug & DEBUG_MFC)
761 			log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n",
762 			    hash,
763 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
764 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
765 			    mfccp->mf6cc_parent);
766 #endif
767 
768 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
769 
770 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
771 					       &mfccp->mf6cc_origin.sin6_addr)&&
772 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
773 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
774 
775 				rt->mf6c_origin     = mfccp->mf6cc_origin;
776 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
777 				rt->mf6c_parent     = mfccp->mf6cc_parent;
778 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
779 				/* initialize pkt counters per src-grp */
780 				rt->mf6c_pkt_cnt    = 0;
781 				rt->mf6c_byte_cnt   = 0;
782 				rt->mf6c_wrong_if   = 0;
783 
784 				if (rt->mf6c_expire)
785 					nexpire[hash]--;
786 				rt->mf6c_expire	   = 0;
787 			}
788 		}
789 		if (rt == NULL) {
790 			/* no upcall, so make a new entry */
791 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
792 						  M_NOWAIT);
793 			if (rt == NULL) {
794 				splx(s);
795 				return ENOBUFS;
796 			}
797 
798 			/* insert new entry at head of hash chain */
799 			rt->mf6c_origin     = mfccp->mf6cc_origin;
800 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
801 			rt->mf6c_parent     = mfccp->mf6cc_parent;
802 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
803 			/* initialize pkt counters per src-grp */
804 			rt->mf6c_pkt_cnt    = 0;
805 			rt->mf6c_byte_cnt   = 0;
806 			rt->mf6c_wrong_if   = 0;
807 			rt->mf6c_expire     = 0;
808 			rt->mf6c_stall = NULL;
809 
810 			/* link into table */
811 			rt->mf6c_next  = mf6ctable[hash];
812 			mf6ctable[hash] = rt;
813 		}
814 	}
815 	splx(s);
816 	return 0;
817 }
818 
819 #ifdef UPCALL_TIMING
820 /*
821  * collect delay statistics on the upcalls
822  */
823 static void
824 collate(t)
825 	struct timeval *t;
826 {
827 	u_long d;
828 	struct timeval tp;
829 	u_long delta;
830 
831 	GET_TIME(tp);
832 
833 	if (TV_LT(*t, tp))
834 	{
835 		TV_DELTA(tp, *t, delta);
836 
837 		d = delta >> 10;
838 		if (d > UPCALL_MAX)
839 			d = UPCALL_MAX;
840 
841 		++upcall_data[d];
842 	}
843 }
844 #endif /* UPCALL_TIMING */
845 
846 /*
847  * Delete an mfc entry
848  */
849 static int
850 del_m6fc(mfccp)
851 	struct mf6cctl *mfccp;
852 {
853 	struct sockaddr_in6 	origin;
854 	struct sockaddr_in6 	mcastgrp;
855 	struct mf6c 		*rt;
856 	struct mf6c	 	**nptr;
857 	u_long 		hash;
858 	int s;
859 
860 	origin = mfccp->mf6cc_origin;
861 	mcastgrp = mfccp->mf6cc_mcastgrp;
862 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
863 
864 #ifdef MRT6DEBUG
865 	if (mrt6debug & DEBUG_MFC)
866 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
867 		    ip6_sprintf(&origin.sin6_addr),
868 		    ip6_sprintf(&mcastgrp.sin6_addr));
869 #endif
870 
871 	s = splsoftnet();
872 
873 	nptr = &mf6ctable[hash];
874 	while ((rt = *nptr) != NULL) {
875 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
876 				       &rt->mf6c_origin.sin6_addr) &&
877 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
878 				       &rt->mf6c_mcastgrp.sin6_addr) &&
879 		    rt->mf6c_stall == NULL)
880 			break;
881 
882 		nptr = &rt->mf6c_next;
883 	}
884 	if (rt == NULL) {
885 		splx(s);
886 		return EADDRNOTAVAIL;
887 	}
888 
889 	*nptr = rt->mf6c_next;
890 	free(rt, M_MRTABLE);
891 
892 	splx(s);
893 
894 	return 0;
895 }
896 
897 static int
898 socket_send(s, mm, src)
899 	struct socket *s;
900 	struct mbuf *mm;
901 	struct sockaddr_in6 *src;
902 {
903 	if (s) {
904 		if (sbappendaddr(&s->so_rcv,
905 				 (struct sockaddr *)src,
906 				 mm, (struct mbuf *)0) != 0) {
907 			sorwakeup(s);
908 			return 0;
909 		}
910 	}
911 	m_freem(mm);
912 	return -1;
913 }
914 
915 /*
916  * IPv6 multicast forwarding function. This function assumes that the packet
917  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
918  * pointed to by "ifp", and the packet is to be relayed to other networks
919  * that have members of the packet's destination IPv6 multicast group.
920  *
921  * The packet is returned unscathed to the caller, unless it is
922  * erroneous, in which case a non-zero return value tells the caller to
923  * discard it.
924  */
925 
926 int
927 ip6_mforward(ip6, ifp, m)
928 	struct ip6_hdr *ip6;
929 	struct ifnet *ifp;
930 	struct mbuf *m;
931 {
932 	struct mf6c *rt;
933 	struct mif6 *mifp;
934 	struct mbuf *mm;
935 	int s;
936 	mifi_t mifi;
937 	long time_second = time.tv_sec;
938 
939 #ifdef MRT6DEBUG
940 	if (mrt6debug & DEBUG_FORWARD)
941 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
942 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
943 		    ifp->if_index);
944 #endif
945 
946 	/*
947 	 * Don't forward a packet with Hop limit of zero or one,
948 	 * or a packet destined to a local-only group.
949 	 */
950 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
951 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
952 		return 0;
953 	ip6->ip6_hlim--;
954 
955 	/*
956 	 * Source address check: do not forward packets with unspecified
957 	 * source. It was discussed in July 2000, on ipngwg mailing list.
958 	 * This is rather more serious than unicast cases, because some
959 	 * MLD packets can be sent with the unspecified source address
960 	 * (although such packets must normally set 1 to the hop limit field).
961 	 */
962 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
963 		ip6stat.ip6s_cantforward++;
964 		if (ip6_log_time + ip6_log_interval < time_second) {
965 			ip6_log_time = time_second;
966 			log(LOG_DEBUG,
967 			    "cannot forward "
968 			    "from %s to %s nxt %d received on %s\n",
969 			    ip6_sprintf(&ip6->ip6_src),
970 			    ip6_sprintf(&ip6->ip6_dst),
971 			    ip6->ip6_nxt,
972 			    if_name(m->m_pkthdr.rcvif));
973 		}
974 		return 0;
975 	}
976 
977 	/*
978 	 * Determine forwarding mifs from the forwarding cache table
979 	 */
980 	s = splsoftnet();
981 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
982 
983 	/* Entry exists, so forward if necessary */
984 	if (rt) {
985 		splx(s);
986 		return (ip6_mdq(m, ifp, rt));
987 	} else {
988 		/*
989 		 * If we don't have a route for packet's origin,
990 		 * Make a copy of the packet &
991 		 * send message to routing daemon
992 		 */
993 
994 		struct mbuf *mb0;
995 		struct rtdetq *rte;
996 		u_long hash;
997 /*		int i, npkts;*/
998 #ifdef UPCALL_TIMING
999 		struct timeval tp;
1000 
1001 		GET_TIME(tp);
1002 #endif /* UPCALL_TIMING */
1003 
1004 		mrt6stat.mrt6s_no_route++;
1005 #ifdef MRT6DEBUG
1006 		if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
1007 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
1008 			    ip6_sprintf(&ip6->ip6_src),
1009 			    ip6_sprintf(&ip6->ip6_dst));
1010 #endif
1011 
1012 		/*
1013 		 * Allocate mbufs early so that we don't do extra work if we
1014 		 * are just going to fail anyway.
1015 		 */
1016 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1017 					      M_NOWAIT);
1018 		if (rte == NULL) {
1019 			splx(s);
1020 			return ENOBUFS;
1021 		}
1022 		mb0 = m_copy(m, 0, M_COPYALL);
1023 		/*
1024 		 * Pullup packet header if needed before storing it,
1025 		 * as other references may modify it in the meantime.
1026 		 */
1027 		if (mb0 &&
1028 		    (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1029 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1030 		if (mb0 == NULL) {
1031 			free(rte, M_MRTABLE);
1032 			splx(s);
1033 			return ENOBUFS;
1034 		}
1035 
1036 		/* is there an upcall waiting for this packet? */
1037 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1038 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1039 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1040 					       &rt->mf6c_origin.sin6_addr) &&
1041 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1042 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1043 			    (rt->mf6c_stall != NULL))
1044 				break;
1045 		}
1046 
1047 		if (rt == NULL) {
1048 			struct mrt6msg *im;
1049 			struct omrt6msg *oim;
1050 
1051 			/* no upcall, so make a new entry */
1052 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1053 						  M_NOWAIT);
1054 			if (rt == NULL) {
1055 				free(rte, M_MRTABLE);
1056 				m_freem(mb0);
1057 				splx(s);
1058 				return ENOBUFS;
1059 			}
1060 			/*
1061 			 * Make a copy of the header to send to the user
1062 			 * level process
1063 			 */
1064 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1065 
1066 			if (mm == NULL) {
1067 				free(rte, M_MRTABLE);
1068 				m_freem(mb0);
1069 				free(rt, M_MRTABLE);
1070 				splx(s);
1071 				return ENOBUFS;
1072 			}
1073 
1074 			/*
1075 			 * Send message to routing daemon
1076 			 */
1077 			sin6.sin6_addr = ip6->ip6_src;
1078 
1079 			im = NULL;
1080 			oim = NULL;
1081 			switch (ip6_mrouter_ver) {
1082 			case MRT6_OINIT:
1083 				oim = mtod(mm, struct omrt6msg *);
1084 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1085 				oim->im6_mbz = 0;
1086 				break;
1087 			case MRT6_INIT:
1088 				im = mtod(mm, struct mrt6msg *);
1089 				im->im6_msgtype = MRT6MSG_NOCACHE;
1090 				im->im6_mbz = 0;
1091 				break;
1092 			default:
1093 				free(rte, M_MRTABLE);
1094 				m_freem(mb0);
1095 				free(rt, M_MRTABLE);
1096 				splx(s);
1097 				return EINVAL;
1098 			}
1099 
1100 #ifdef MRT6DEBUG
1101 			if (mrt6debug & DEBUG_FORWARD)
1102 				log(LOG_DEBUG,
1103 				    "getting the iif info in the kernel\n");
1104 #endif
1105 
1106 			for (mifp = mif6table, mifi = 0;
1107 			     mifi < nummifs && mifp->m6_ifp != ifp;
1108 			     mifp++, mifi++)
1109 				;
1110 
1111 			switch (ip6_mrouter_ver) {
1112 			case MRT6_OINIT:
1113 				oim->im6_mif = mifi;
1114 				break;
1115 			case MRT6_INIT:
1116 				im->im6_mif = mifi;
1117 				break;
1118 			}
1119 
1120 			if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1121 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1122 				    "socket queue full\n");
1123 				mrt6stat.mrt6s_upq_sockfull++;
1124 				free(rte, M_MRTABLE);
1125 				m_freem(mb0);
1126 				free(rt, M_MRTABLE);
1127 				splx(s);
1128 				return ENOBUFS;
1129 			}
1130 
1131 			mrt6stat.mrt6s_upcalls++;
1132 
1133 			/* insert new entry at head of hash chain */
1134 			bzero(rt, sizeof(*rt));
1135 			rt->mf6c_origin.sin6_family = AF_INET6;
1136 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1137 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1138 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1139 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1140 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1141 			rt->mf6c_expire = UPCALL_EXPIRE;
1142 			nexpire[hash]++;
1143 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1144 
1145 			/* link into table */
1146 			rt->mf6c_next  = mf6ctable[hash];
1147 			mf6ctable[hash] = rt;
1148 			/* Add this entry to the end of the queue */
1149 			rt->mf6c_stall = rte;
1150 		} else {
1151 			/* determine if q has overflowed */
1152 			struct rtdetq **p;
1153 			int npkts = 0;
1154 
1155 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1156 				if (++npkts > MAX_UPQ6) {
1157 					mrt6stat.mrt6s_upq_ovflw++;
1158 					free(rte, M_MRTABLE);
1159 					m_freem(mb0);
1160 					splx(s);
1161 					return 0;
1162 				}
1163 
1164 			/* Add this entry to the end of the queue */
1165 			*p = rte;
1166 		}
1167 
1168 		rte->next = NULL;
1169 		rte->m = mb0;
1170 		rte->ifp = ifp;
1171 #ifdef UPCALL_TIMING
1172 		rte->t = tp;
1173 #endif /* UPCALL_TIMING */
1174 
1175 		splx(s);
1176 
1177 		return 0;
1178 	}
1179 }
1180 
1181 /*
1182  * Clean up cache entries if upcalls are not serviced
1183  * Call from the Slow Timeout mechanism, every half second.
1184  */
1185 static void
1186 expire_upcalls(unused)
1187 	void *unused;
1188 {
1189 	struct rtdetq *rte;
1190 	struct mf6c *mfc, **nptr;
1191 	int i;
1192 	int s;
1193 
1194 	s = splsoftnet();
1195 	for (i = 0; i < MF6CTBLSIZ; i++) {
1196 		if (nexpire[i] == 0)
1197 			continue;
1198 		nptr = &mf6ctable[i];
1199 		while ((mfc = *nptr) != NULL) {
1200 			rte = mfc->mf6c_stall;
1201 			/*
1202 			 * Skip real cache entries
1203 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1204 			 * If it expires now
1205 			 */
1206 			if (rte != NULL &&
1207 			    mfc->mf6c_expire != 0 &&
1208 			    --mfc->mf6c_expire == 0) {
1209 #ifdef MRT6DEBUG
1210 				if (mrt6debug & DEBUG_EXPIRE)
1211 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1212 					    ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1213 					    ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1214 #endif
1215 				/*
1216 				 * drop all the packets
1217 				 * free the mbuf with the pkt, if, timing info
1218 				 */
1219 				do {
1220 					struct rtdetq *n = rte->next;
1221 					m_freem(rte->m);
1222 					free(rte, M_MRTABLE);
1223 					rte = n;
1224 				} while (rte != NULL);
1225 				mrt6stat.mrt6s_cache_cleanups++;
1226 				nexpire[i]--;
1227 
1228 				*nptr = mfc->mf6c_next;
1229 				free(mfc, M_MRTABLE);
1230 			} else {
1231 				nptr = &mfc->mf6c_next;
1232 			}
1233 		}
1234 	}
1235 	splx(s);
1236 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1237 	    expire_upcalls, NULL);
1238 }
1239 
1240 /*
1241  * Packet forwarding routine once entry in the cache is made
1242  */
1243 static int
1244 ip6_mdq(m, ifp, rt)
1245 	struct mbuf *m;
1246 	struct ifnet *ifp;
1247 	struct mf6c *rt;
1248 {
1249 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1250 	mifi_t mifi, iif;
1251 	struct mif6 *mifp;
1252 	int plen = m->m_pkthdr.len;
1253 
1254 /*
1255  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1256  * input, they shouldn't get counted on output, so statistics keeping is
1257  * separate.
1258  */
1259 
1260 #define MC6_SEND(ip6, mifp, m) do {				\
1261 		if ((mifp)->m6_flags & MIFF_REGISTER)		\
1262 		    register_send((ip6), (mifp), (m));		\
1263 		else						\
1264 		    phyint_send((ip6), (mifp), (m));		\
1265 } while (0)
1266 
1267 	/*
1268 	 * Don't forward if it didn't arrive from the parent mif
1269 	 * for its origin.
1270 	 */
1271 	mifi = rt->mf6c_parent;
1272 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1273 		/* came in the wrong interface */
1274 #ifdef MRT6DEBUG
1275 		if (mrt6debug & DEBUG_FORWARD)
1276 			log(LOG_DEBUG,
1277 			    "wrong if: ifid %d mifi %d mififid %x\n",
1278 			    ifp->if_index, mifi,
1279 			    mif6table[mifi].m6_ifp->if_index);
1280 #endif
1281 		mrt6stat.mrt6s_wrong_if++;
1282 		rt->mf6c_wrong_if++;
1283 		/*
1284 		 * If we are doing PIM processing, and we are forwarding
1285 		 * packets on this interface, send a message to the
1286 		 * routing daemon.
1287 		 */
1288 		/* have to make sure this is a valid mif */
1289 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1290 			if (pim6 && (m->m_flags & M_LOOP) == 0) {
1291 				/*
1292 				 * Check the M_LOOP flag to avoid an
1293 				 * unnecessary PIM assert.
1294 				 * XXX: M_LOOP is an ad-hoc hack...
1295 				 */
1296 				static struct sockaddr_in6 sin6 =
1297 				{ sizeof(sin6), AF_INET6 };
1298 
1299 				struct mbuf *mm;
1300 				struct mrt6msg *im;
1301 				struct omrt6msg *oim;
1302 
1303 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1304 				if (mm &&
1305 				    (M_HASCL(mm) ||
1306 				     mm->m_len < sizeof(struct ip6_hdr)))
1307 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1308 				if (mm == NULL)
1309 					return ENOBUFS;
1310 
1311 				oim = NULL;
1312 				im = NULL;
1313 				switch (ip6_mrouter_ver) {
1314 				case MRT6_OINIT:
1315 					oim = mtod(mm, struct omrt6msg *);
1316 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1317 					oim->im6_mbz = 0;
1318 					break;
1319 				case MRT6_INIT:
1320 					im = mtod(mm, struct mrt6msg *);
1321 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1322 					im->im6_mbz = 0;
1323 					break;
1324 				default:
1325 					m_freem(mm);
1326 					return EINVAL;
1327 				}
1328 
1329 				for (mifp = mif6table, iif = 0;
1330 				     iif < nummifs && mifp &&
1331 					     mifp->m6_ifp != ifp;
1332 				     mifp++, iif++)
1333 					;
1334 
1335 				switch (ip6_mrouter_ver) {
1336 				case MRT6_OINIT:
1337 					oim->im6_mif = iif;
1338 					sin6.sin6_addr = oim->im6_src;
1339 					break;
1340 				case MRT6_INIT:
1341 					im->im6_mif = iif;
1342 					sin6.sin6_addr = im->im6_src;
1343 					break;
1344 				}
1345 
1346 				mrt6stat.mrt6s_upcalls++;
1347 
1348 				if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1349 #ifdef MRT6DEBUG
1350 					if (mrt6debug)
1351 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1352 #endif
1353 					++mrt6stat.mrt6s_upq_sockfull;
1354 					return ENOBUFS;
1355 				}	/* if socket Q full */
1356 			}		/* if PIM */
1357 		return 0;
1358 	}			/* if wrong iif */
1359 
1360 	/* If I sourced this packet, it counts as output, else it was input. */
1361 	if (m->m_pkthdr.rcvif == NULL) {
1362 		/* XXX: is rcvif really NULL when output?? */
1363 		mif6table[mifi].m6_pkt_out++;
1364 		mif6table[mifi].m6_bytes_out += plen;
1365 	} else {
1366 		mif6table[mifi].m6_pkt_in++;
1367 		mif6table[mifi].m6_bytes_in += plen;
1368 	}
1369 	rt->mf6c_pkt_cnt++;
1370 	rt->mf6c_byte_cnt += plen;
1371 
1372 	/*
1373 	 * For each mif, forward a copy of the packet if there are group
1374 	 * members downstream on the interface.
1375 	 */
1376 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1377 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1378 #ifdef notyet
1379 			/*
1380 			 * check if the outgoing packet is going to break
1381 			 * a scope boundary.
1382 			 * XXX For packets through PIM register tunnel
1383 			 * interface, we believe a routing daemon.
1384 			 */
1385 			if ((mif6table[rt->mf6c_parent].m6_flags &
1386 			     MIFF_REGISTER) == 0 &&
1387 			    (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1388 			    (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1389 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1390 					      &ip6->ip6_dst) ||
1391 			     in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1392 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1393 					      &ip6->ip6_src))) {
1394 				ip6stat.ip6s_badscope++;
1395 				continue;
1396 			}
1397 #endif
1398 
1399 			mifp->m6_pkt_out++;
1400 			mifp->m6_bytes_out += plen;
1401 			MC6_SEND(ip6, mifp, m);
1402 		}
1403 	return 0;
1404 }
1405 
1406 static void
1407 phyint_send(ip6, mifp, m)
1408     struct ip6_hdr *ip6;
1409     struct mif6 *mifp;
1410     struct mbuf *m;
1411 {
1412 	struct mbuf *mb_copy;
1413 	struct ifnet *ifp = mifp->m6_ifp;
1414 	int error = 0;
1415 	int s = splsoftnet();
1416 	static struct route_in6 ro;
1417 	struct	in6_multi *in6m;
1418 	struct sockaddr_in6 *dst6;
1419 
1420 	/*
1421 	 * Make a new reference to the packet; make sure that
1422 	 * the IPv6 header is actually copied, not just referenced,
1423 	 * so that ip6_output() only scribbles on the copy.
1424 	 */
1425 	mb_copy = m_copy(m, 0, M_COPYALL);
1426 	if (mb_copy &&
1427 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1428 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1429 	if (mb_copy == NULL) {
1430 		splx(s);
1431 		return;
1432 	}
1433 	/* set MCAST flag to the outgoing packet */
1434 	mb_copy->m_flags |= M_MCAST;
1435 
1436 	/*
1437 	 * If we sourced the packet, call ip6_output since we may devide
1438 	 * the packet into fragments when the packet is too big for the
1439 	 * outgoing interface.
1440 	 * Otherwise, we can simply send the packet to the interface
1441 	 * sending queue.
1442 	 */
1443 	if (m->m_pkthdr.rcvif == NULL) {
1444 		struct ip6_moptions im6o;
1445 
1446 		im6o.im6o_multicast_ifp = ifp;
1447 		/* XXX: ip6_output will override ip6->ip6_hlim */
1448 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1449 		im6o.im6o_multicast_loop = 1;
1450 		error = ip6_output(mb_copy, NULL, &ro,
1451 				   IPV6_FORWARDING, &im6o, NULL);
1452 
1453 #ifdef MRT6DEBUG
1454 		if (mrt6debug & DEBUG_XMIT)
1455 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1456 			    mifp - mif6table, error);
1457 #endif
1458 		splx(s);
1459 		return;
1460 	}
1461 
1462 	/*
1463 	 * If we belong to the destination multicast group
1464 	 * on the outgoing interface, loop back a copy.
1465 	 */
1466 	dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
1467 	IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1468 	if (in6m != NULL) {
1469 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1470 		dst6->sin6_family = AF_INET6;
1471 		dst6->sin6_addr = ip6->ip6_dst;
1472 		ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst);
1473 	}
1474 	/*
1475 	 * Put the packet into the sending queue of the outgoing interface
1476 	 * if it would fit in the MTU of the interface.
1477 	 */
1478 	if (mb_copy->m_pkthdr.len < ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
1479 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1480 		dst6->sin6_family = AF_INET6;
1481 		dst6->sin6_addr = ip6->ip6_dst;
1482 		/*
1483 		 * We just call if_output instead of nd6_output here, since
1484 		 * we need no ND for a multicast forwarded packet...right?
1485 		 */
1486 		error = (*ifp->if_output)(ifp, mb_copy,
1487 		    (struct sockaddr *)&ro.ro_dst, NULL);
1488 #ifdef MRT6DEBUG
1489 		if (mrt6debug & DEBUG_XMIT)
1490 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1491 			    mifp - mif6table, error);
1492 #endif
1493 	} else {
1494 #ifdef MULTICAST_PMTUD
1495 		icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
1496 #else
1497 #ifdef MRT6DEBUG
1498 		if (mrt6debug & DEBUG_XMIT)
1499 			log(LOG_DEBUG,
1500 			    "phyint_send: packet too big on %s o %s g %s"
1501 			    " size %d(discarded)\n",
1502 			    if_name(ifp),
1503 			    ip6_sprintf(&ip6->ip6_src),
1504 			    ip6_sprintf(&ip6->ip6_dst),
1505 			    mb_copy->m_pkthdr.len);
1506 #endif /* MRT6DEBUG */
1507 		m_freem(mb_copy); /* simply discard the packet */
1508 #endif
1509 	}
1510 
1511 	splx(s);
1512 }
1513 
1514 static int
1515 register_send(ip6, mif, m)
1516 	struct ip6_hdr *ip6;
1517 	struct mif6 *mif;
1518 	struct mbuf *m;
1519 {
1520 	struct mbuf *mm;
1521 	int i, len = m->m_pkthdr.len;
1522 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1523 	struct mrt6msg *im6;
1524 
1525 #ifdef MRT6DEBUG
1526 	if (mrt6debug)
1527 		log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1528 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1529 #endif
1530 	++pim6stat.pim6s_snd_registers;
1531 
1532 	/* Make a copy of the packet to send to the user level process */
1533 	MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1534 	if (mm == NULL)
1535 		return ENOBUFS;
1536 	mm->m_data += max_linkhdr;
1537 	mm->m_len = sizeof(struct ip6_hdr);
1538 
1539 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1540 		m_freem(mm);
1541 		return ENOBUFS;
1542 	}
1543 	i = MHLEN - M_LEADINGSPACE(mm);
1544 	if (i > len)
1545 		i = len;
1546 	mm = m_pullup(mm, i);
1547 	if (mm == NULL){
1548 		m_freem(mm);
1549 		return ENOBUFS;
1550 	}
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: ip_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 = looutput(mif6table[reg_mif_num].m6_ifp, m,
1799 			      (struct sockaddr *) &dst,
1800 			      (struct rtentry *) NULL);
1801 
1802 		/* prepare the register head to send to the mrouting daemon */
1803 		m = mcp;
1804 	}
1805 
1806 	/*
1807 	 * Pass the PIM message up to the daemon; if it is a register message
1808 	 * pass the 'head' only up to the daemon. This includes the
1809 	 * encapsulator ip6 header, pim header, register header and the
1810 	 * encapsulated ip6 header.
1811 	 */
1812   pim6_input_to_daemon:
1813 	rip6_input(&m, offp, proto);
1814 	return(IPPROTO_DONE);
1815 }
1816