xref: /netbsd-src/sys/netinet6/ip6_mroute.c (revision 37b34d511dea595d3ba03a661cf3b775038ea5f8)
1 /*	$NetBSD: ip6_mroute.c,v 1.38 2002/09/23 05:51:15 simonb 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 Fenner, PARC, April 1994
44  *
45  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
46  */
47 
48 #include <sys/cdefs.h>
49 __KERNEL_RCSID(0, "$NetBSD: ip6_mroute.c,v 1.38 2002/09/23 05:51:15 simonb Exp $");
50 
51 #include "opt_inet.h"
52 
53 #ifndef _KERNEL
54 # ifdef KERNEL
55 #  define _KERNEL
56 # endif
57 #endif
58 
59 #include <sys/param.h>
60 #include <sys/systm.h>
61 #include <sys/callout.h>
62 #include <sys/mbuf.h>
63 #include <sys/socket.h>
64 #include <sys/socketvar.h>
65 #include <sys/sockio.h>
66 #include <sys/protosw.h>
67 #include <sys/errno.h>
68 #include <sys/time.h>
69 #include <sys/kernel.h>
70 #include <sys/ioctl.h>
71 #include <sys/syslog.h>
72 
73 #include <net/if.h>
74 #include <net/route.h>
75 #include <net/raw_cb.h>
76 
77 #include <netinet/in.h>
78 #include <netinet/in_var.h>
79 
80 #include <netinet/ip6.h>
81 #include <netinet6/ip6_var.h>
82 #include <netinet6/ip6_mroute.h>
83 #include <netinet6/pim6.h>
84 #include <netinet6/pim6_var.h>
85 #include <netinet6/nd6.h>
86 
87 #include <net/net_osdep.h>
88 
89 static int ip6_mdq __P((struct mbuf *, struct ifnet *, struct mf6c *));
90 static void phyint_send __P((struct ip6_hdr *, struct mif6 *, struct mbuf *));
91 
92 static int set_pim6 __P((int *));
93 static int get_pim6 __P((struct mbuf *));
94 static int socket_send __P((struct socket *, struct mbuf *,
95 			    struct sockaddr_in6 *));
96 static int register_send __P((struct ip6_hdr *, struct mif6 *,
97 			      struct mbuf *));
98 
99 /*
100  * Globals.  All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static,
101  * except for netstat or debugging purposes.
102  */
103 struct socket  *ip6_mrouter = NULL;
104 int		ip6_mrouter_ver = 0;
105 int		ip6_mrtproto = IPPROTO_PIM;    /* for netstat only */
106 struct mrt6stat	mrt6stat;
107 
108 #define NO_RTE_FOUND 	0x1
109 #define RTE_FOUND	0x2
110 
111 struct mf6c	*mf6ctable[MF6CTBLSIZ];
112 u_char		n6expire[MF6CTBLSIZ];
113 static struct mif6 mif6table[MAXMIFS];
114 #ifdef MRT6DEBUG
115 u_int		mrt6debug = 0;	  /* debug level 	*/
116 #define		DEBUG_MFC	0x02
117 #define		DEBUG_FORWARD	0x04
118 #define		DEBUG_EXPIRE	0x08
119 #define		DEBUG_XMIT	0x10
120 #define         DEBUG_REG       0x20
121 #define         DEBUG_PIM       0x40
122 #endif
123 
124 static void	expire_upcalls __P((void *));
125 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
126 #define		UPCALL_EXPIRE	6		/* number of timeouts */
127 
128 #ifdef INET
129 #ifdef MROUTING
130 extern struct socket *ip_mrouter;
131 #endif
132 #endif
133 
134 /*
135  * 'Interfaces' associated with decapsulator (so we can tell
136  * packets that went through it from ones that get reflected
137  * by a broken gateway).  These interfaces are never linked into
138  * the system ifnet list & no routes point to them.  I.e., packets
139  * can't be sent this way.  They only exist as a placeholder for
140  * multicast source verification.
141  */
142 struct ifnet multicast_register_if;
143 
144 #define ENCAP_HOPS 64
145 
146 /*
147  * Private variables.
148  */
149 static mifi_t nummifs = 0;
150 static mifi_t reg_mif_num = (mifi_t)-1;
151 
152 static struct pim6stat pim6stat;
153 static int pim6;
154 
155 /*
156  * Hash function for a source, group entry
157  */
158 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
159 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
160 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
161 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
162 
163 /*
164  * Find a route for a given origin IPv6 address and Multicast group address.
165  * Quality of service parameter to be added in the future!!!
166  */
167 
168 #define MF6CFIND(o, g, rt) do { \
169 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
170 	rt = NULL; \
171 	mrt6stat.mrt6s_mfc_lookups++; \
172 	while (_rt) { \
173 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
174 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
175 		    (_rt->mf6c_stall == NULL)) { \
176 			rt = _rt; \
177 			break; \
178 		} \
179 		_rt = _rt->mf6c_next; \
180 	} \
181 	if (rt == NULL) { \
182 		mrt6stat.mrt6s_mfc_misses++; \
183 	} \
184 } while (0)
185 
186 /*
187  * Macros to compute elapsed time efficiently
188  * Borrowed from Van Jacobson's scheduling code
189  */
190 #define TV_DELTA(a, b, delta) do { \
191 	    int xxs; \
192 		\
193 	    delta = (a).tv_usec - (b).tv_usec; \
194 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
195 	       switch (xxs) { \
196 		      case 2: \
197 			  delta += 1000000; \
198 			      /* FALLTHROUGH */ \
199 		      case 1: \
200 			  delta += 1000000; \
201 			  break; \
202 		      default: \
203 			  delta += (1000000 * xxs); \
204 	       } \
205 	    } \
206 } while (0)
207 
208 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
209 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
210 
211 #ifdef UPCALL_TIMING
212 #define UPCALL_MAX	50
213 u_long upcall_data[UPCALL_MAX + 1];
214 static void collate();
215 #endif /* UPCALL_TIMING */
216 
217 static int get_sg_cnt __P((struct sioc_sg_req6 *));
218 static int get_mif6_cnt __P((struct sioc_mif_req6 *));
219 static int ip6_mrouter_init __P((struct socket *, struct mbuf *, int));
220 static int add_m6if __P((struct mif6ctl *));
221 static int del_m6if __P((mifi_t *));
222 static int add_m6fc __P((struct mf6cctl *));
223 static int del_m6fc __P((struct mf6cctl *));
224 
225 static struct callout expire_upcalls_ch = CALLOUT_INITIALIZER;
226 
227 /*
228  * Handle MRT setsockopt commands to modify the multicast routing tables.
229  */
230 int
231 ip6_mrouter_set(cmd, so, m)
232 	int cmd;
233 	struct socket *so;
234 	struct mbuf *m;
235 {
236 	if (cmd != MRT6_INIT && so != ip6_mrouter)
237 		return EACCES;
238 
239 	switch (cmd) {
240 	case MRT6_OINIT:	return ip6_mrouter_init(so, m, cmd);
241 	case MRT6_INIT:		return ip6_mrouter_init(so, m, cmd);
242 	case MRT6_DONE:		return ip6_mrouter_done();
243 	case MRT6_ADD_MIF:	return add_m6if(mtod(m, struct mif6ctl *));
244 	case MRT6_DEL_MIF:	return del_m6if(mtod(m, mifi_t *));
245 	case MRT6_ADD_MFC:	return add_m6fc(mtod(m, struct mf6cctl *));
246 	case MRT6_DEL_MFC:	return del_m6fc(mtod(m, struct mf6cctl *));
247 	case MRT6_PIM:		return set_pim6(mtod(m, int *));
248 	default:		return EOPNOTSUPP;
249 	}
250 }
251 
252 /*
253  * Handle MRT getsockopt commands
254  */
255 int
256 ip6_mrouter_get(cmd, so, m)
257 	int cmd;
258 	struct socket *so;
259 	struct mbuf **m;
260 {
261 	struct mbuf *mb;
262 
263 	if (so != ip6_mrouter) return EACCES;
264 
265 	*m = mb = m_get(M_WAIT, MT_SOOPTS);
266 
267 	switch (cmd) {
268 	case MRT6_PIM:
269 		return get_pim6(mb);
270 	default:
271 		m_free(mb);
272 		return EOPNOTSUPP;
273 	}
274 }
275 
276 /*
277  * Handle ioctl commands to obtain information from the cache
278  */
279 int
280 mrt6_ioctl(cmd, data)
281 	int cmd;
282 	caddr_t data;
283 {
284 
285 	switch (cmd) {
286 	case SIOCGETSGCNT_IN6:
287 		return (get_sg_cnt((struct sioc_sg_req6 *)data));
288 	case SIOCGETMIFCNT_IN6:
289 		return (get_mif6_cnt((struct sioc_mif_req6 *)data));
290 	default:
291 		return (EINVAL);
292 	}
293 }
294 
295 /*
296  * returns the packet, byte, rpf-failure count for the source group provided
297  */
298 static int
299 get_sg_cnt(req)
300 	struct sioc_sg_req6 *req;
301 {
302 	struct mf6c *rt;
303 	int s;
304 
305 	s = splsoftnet();
306 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
307 	splx(s);
308 	if (rt != NULL) {
309 		req->pktcnt = rt->mf6c_pkt_cnt;
310 		req->bytecnt = rt->mf6c_byte_cnt;
311 		req->wrong_if = rt->mf6c_wrong_if;
312 	} else
313 		return (ESRCH);
314 #if 0
315 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
316 #endif
317 
318 	return 0;
319 }
320 
321 /*
322  * returns the input and output packet and byte counts on the mif provided
323  */
324 static int
325 get_mif6_cnt(req)
326 	struct sioc_mif_req6 *req;
327 {
328 	mifi_t mifi = req->mifi;
329 
330 	if (mifi >= nummifs)
331 		return EINVAL;
332 
333 	req->icount = mif6table[mifi].m6_pkt_in;
334 	req->ocount = mif6table[mifi].m6_pkt_out;
335 	req->ibytes = mif6table[mifi].m6_bytes_in;
336 	req->obytes = mif6table[mifi].m6_bytes_out;
337 
338 	return 0;
339 }
340 
341 /*
342  * Get PIM processiong global
343  */
344 static int
345 get_pim6(m)
346 	struct mbuf *m;
347 {
348 	int *i;
349 
350 	i = mtod(m, int *);
351 
352 	*i = pim6;
353 
354 	return 0;
355 }
356 
357 static int
358 set_pim6(i)
359 	int *i;
360 {
361 	if ((*i != 1) && (*i != 0))
362 		return EINVAL;
363 
364 	pim6 = *i;
365 
366 	return 0;
367 }
368 
369 /*
370  * Enable multicast routing
371  */
372 static int
373 ip6_mrouter_init(so, m, cmd)
374 	struct socket *so;
375 	struct mbuf *m;
376 	int cmd;
377 {
378 	int *v;
379 
380 #ifdef MRT6DEBUG
381 	if (mrt6debug)
382 		log(LOG_DEBUG,
383 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
384 		    so->so_type, so->so_proto->pr_protocol);
385 #endif
386 
387 	if (so->so_type != SOCK_RAW ||
388 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
389 		return EOPNOTSUPP;
390 
391 	if (!m || (m->m_len != sizeof(int *)))
392 		return ENOPROTOOPT;
393 
394 	v = mtod(m, int *);
395 	if (*v != 1)
396 		return ENOPROTOOPT;
397 
398 	if (ip6_mrouter != NULL) return EADDRINUSE;
399 
400 	ip6_mrouter = so;
401 	ip6_mrouter_ver = cmd;
402 
403 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
404 	bzero((caddr_t)n6expire, sizeof(n6expire));
405 
406 	pim6 = 0;/* used for stubbing out/in pim stuff */
407 
408 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
409 	    expire_upcalls, NULL);
410 
411 #ifdef MRT6DEBUG
412 	if (mrt6debug)
413 		log(LOG_DEBUG, "ip6_mrouter_init\n");
414 #endif
415 
416 	return 0;
417 }
418 
419 /*
420  * Disable multicast routing
421  */
422 int
423 ip6_mrouter_done()
424 {
425 	mifi_t mifi;
426 	int i;
427 	struct ifnet *ifp;
428 	struct in6_ifreq ifr;
429 	struct mf6c *rt;
430 	struct rtdetq *rte;
431 	int s;
432 
433 	s = splsoftnet();
434 
435 	/*
436 	 * For each phyint in use, disable promiscuous reception of all IPv6
437 	 * multicasts.
438 	 */
439 #ifdef INET
440 #ifdef MROUTING
441 	/*
442 	 * If there is still IPv4 multicast routing daemon,
443 	 * we remain interfaces to receive all muliticasted packets.
444 	 * XXX: there may be an interface in which the IPv4 multicast
445 	 * daemon is not interested...
446 	 */
447 	if (!ip_mrouter)
448 #endif
449 #endif
450 	{
451 		for (mifi = 0; mifi < nummifs; mifi++) {
452 			if (mif6table[mifi].m6_ifp &&
453 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
454 				ifr.ifr_addr.sin6_family = AF_INET6;
455 				ifr.ifr_addr.sin6_addr= in6addr_any;
456 				ifp = mif6table[mifi].m6_ifp;
457 				(*ifp->if_ioctl)(ifp, SIOCDELMULTI,
458 						 (caddr_t)&ifr);
459 			}
460 		}
461 	}
462 #ifdef notyet
463 	bzero((caddr_t)qtable, sizeof(qtable));
464 	bzero((caddr_t)tbftable, sizeof(tbftable));
465 #endif
466 	bzero((caddr_t)mif6table, sizeof(mif6table));
467 	nummifs = 0;
468 
469 	pim6 = 0; /* used to stub out/in pim specific code */
470 
471 	callout_stop(&expire_upcalls_ch);
472 
473 	/*
474 	 * Free all multicast forwarding cache entries.
475 	 */
476 	for (i = 0; i < MF6CTBLSIZ; i++) {
477 		rt = mf6ctable[i];
478 		while (rt) {
479 			struct mf6c *frt;
480 
481 			for (rte = rt->mf6c_stall; rte != NULL; ) {
482 				struct rtdetq *n = rte->next;
483 
484 				m_free(rte->m);
485 				free(rte, M_MRTABLE);
486 				rte = n;
487 			}
488 			frt = rt;
489 			rt = rt->mf6c_next;
490 			free(frt, M_MRTABLE);
491 		}
492 	}
493 
494 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
495 
496 	/*
497 	 * Reset de-encapsulation cache
498 	 */
499 	reg_mif_num = -1;
500 
501 	ip6_mrouter = NULL;
502 	ip6_mrouter_ver = 0;
503 
504 	splx(s);
505 
506 #ifdef MRT6DEBUG
507 	if (mrt6debug)
508 		log(LOG_DEBUG, "ip6_mrouter_done\n");
509 #endif
510 
511 	return 0;
512 }
513 
514 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
515 
516 /*
517  * Add a mif to the mif table
518  */
519 static int
520 add_m6if(mifcp)
521 	struct mif6ctl *mifcp;
522 {
523 	struct mif6 *mifp;
524 	struct ifnet *ifp;
525 	struct in6_ifreq ifr;
526 	int error, s;
527 #ifdef notyet
528 	struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
529 #endif
530 
531 	if (mifcp->mif6c_mifi >= MAXMIFS)
532 		return EINVAL;
533 	mifp = mif6table + mifcp->mif6c_mifi;
534 	if (mifp->m6_ifp)
535 		return EADDRINUSE; /* XXX: is it appropriate? */
536 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
537 		return ENXIO;
538 	/*
539 	 * XXX: some OSes can remove ifp and clear ifindex2ifnet[id]
540 	 * even for id between 0 and if_index.
541 	 */
542 	if ((ifp = ifindex2ifnet[mifcp->mif6c_pifi]) == NULL)
543 		return ENXIO;
544 
545 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
546 		if (reg_mif_num == (mifi_t)-1) {
547 			strcpy(multicast_register_if.if_xname,
548 			       "register_mif"); /* XXX */
549 			multicast_register_if.if_flags |= IFF_LOOPBACK;
550 			multicast_register_if.if_index = mifcp->mif6c_mifi;
551 			reg_mif_num = mifcp->mif6c_mifi;
552 		}
553 
554 		ifp = &multicast_register_if;
555 
556 	} /* if REGISTER */
557 	else {
558 		/* Make sure the interface supports multicast */
559 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
560 			return EOPNOTSUPP;
561 
562 		s = splsoftnet();
563 		/*
564 		 * Enable promiscuous reception of all IPv6 multicasts
565 		 * from the interface.
566 		 */
567 		ifr.ifr_addr.sin6_family = AF_INET6;
568 		ifr.ifr_addr.sin6_addr = in6addr_any;
569 		error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
570 		splx(s);
571 		if (error)
572 			return error;
573 	}
574 
575 	s = splsoftnet();
576 	mifp->m6_flags     = mifcp->mif6c_flags;
577 	mifp->m6_ifp       = ifp;
578 #ifdef notyet
579 	/* scaling up here allows division by 1024 in critical code */
580 	mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
581 #endif
582 	/* initialize per mif pkt counters */
583 	mifp->m6_pkt_in    = 0;
584 	mifp->m6_pkt_out   = 0;
585 	mifp->m6_bytes_in  = 0;
586 	mifp->m6_bytes_out = 0;
587 	splx(s);
588 
589 	/* Adjust nummifs up if the mifi is higher than nummifs */
590 	if (nummifs <= mifcp->mif6c_mifi)
591 		nummifs = mifcp->mif6c_mifi + 1;
592 
593 #ifdef MRT6DEBUG
594 	if (mrt6debug)
595 		log(LOG_DEBUG,
596 		    "add_mif #%d, phyint %s%d\n",
597 		    mifcp->mif6c_mifi,
598 		    ifp->if_name, ifp->if_unit);
599 #endif
600 
601 	return 0;
602 }
603 
604 /*
605  * Delete a mif from the mif table
606  */
607 static int
608 del_m6if(mifip)
609 	mifi_t *mifip;
610 {
611 	struct mif6 *mifp = mif6table + *mifip;
612 	mifi_t mifi;
613 	struct ifnet *ifp;
614 	struct in6_ifreq ifr;
615 	int s;
616 
617 	if (*mifip >= nummifs)
618 		return EINVAL;
619 	if (mifp->m6_ifp == NULL)
620 		return EINVAL;
621 
622 	s = splsoftnet();
623 
624 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
625 		/*
626 		 * XXX: what if there is yet IPv4 multicast daemon
627 		 *      using the interface?
628 		 */
629 		ifp = mifp->m6_ifp;
630 
631 		ifr.ifr_addr.sin6_family = AF_INET6;
632 		ifr.ifr_addr.sin6_addr = in6addr_any;
633 		(*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
634 	}
635 
636 #ifdef notyet
637 	bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
638 	bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
639 #endif
640 	bzero((caddr_t)mifp, sizeof (*mifp));
641 
642 	/* Adjust nummifs down */
643 	for (mifi = nummifs; mifi > 0; mifi--)
644 		if (mif6table[mifi - 1].m6_ifp)
645 			break;
646 	nummifs = mifi;
647 
648 	splx(s);
649 
650 #ifdef MRT6DEBUG
651 	if (mrt6debug)
652 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
653 #endif
654 
655 	return 0;
656 }
657 
658 /*
659  * Add an mfc entry
660  */
661 static int
662 add_m6fc(mfccp)
663 	struct mf6cctl *mfccp;
664 {
665 	struct mf6c *rt;
666 	u_long hash;
667 	struct rtdetq *rte;
668 	u_short nstl;
669 	int s;
670 
671 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
672 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
673 
674 	/* If an entry already exists, just update the fields */
675 	if (rt) {
676 #ifdef MRT6DEBUG
677 		if (mrt6debug & DEBUG_MFC)
678 			log(LOG_DEBUG,"add_m6fc update o %s g %s p %x\n",
679 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
680 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
681 			    mfccp->mf6cc_parent);
682 #endif
683 
684 		s = splsoftnet();
685 		rt->mf6c_parent = mfccp->mf6cc_parent;
686 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
687 		splx(s);
688 		return 0;
689 	}
690 
691 	/*
692 	 * Find the entry for which the upcall was made and update
693 	 */
694 	s = splsoftnet();
695 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
696 			mfccp->mf6cc_mcastgrp.sin6_addr);
697 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
698 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
699 				       &mfccp->mf6cc_origin.sin6_addr) &&
700 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
701 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
702 		    (rt->mf6c_stall != NULL)) {
703 
704 			if (nstl++)
705 				log(LOG_ERR,
706 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
707 				    "multiple kernel entries",
708 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
709 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
710 				    mfccp->mf6cc_parent, rt->mf6c_stall);
711 
712 #ifdef MRT6DEBUG
713 			if (mrt6debug & DEBUG_MFC)
714 				log(LOG_DEBUG,
715 				    "add_m6fc o %s g %s p %x dbg %x\n",
716 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
717 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
718 				    mfccp->mf6cc_parent, rt->mf6c_stall);
719 #endif
720 
721 			rt->mf6c_origin     = mfccp->mf6cc_origin;
722 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
723 			rt->mf6c_parent     = mfccp->mf6cc_parent;
724 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
725 			/* initialize pkt counters per src-grp */
726 			rt->mf6c_pkt_cnt    = 0;
727 			rt->mf6c_byte_cnt   = 0;
728 			rt->mf6c_wrong_if   = 0;
729 
730 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
731 			n6expire[hash]--;
732 
733 			/* free packets Qed at the end of this entry */
734 			for (rte = rt->mf6c_stall; rte != NULL; ) {
735 				struct rtdetq *n = rte->next;
736 				ip6_mdq(rte->m, rte->ifp, rt);
737 				m_freem(rte->m);
738 #ifdef UPCALL_TIMING
739 				collate(&(rte->t));
740 #endif /* UPCALL_TIMING */
741 				free(rte, M_MRTABLE);
742 				rte = n;
743 			}
744 			rt->mf6c_stall = NULL;
745 		}
746 	}
747 
748 	/*
749 	 * It is possible that an entry is being inserted without an upcall
750 	 */
751 	if (nstl == 0) {
752 #ifdef MRT6DEBUG
753 		if (mrt6debug & DEBUG_MFC)
754 			log(LOG_DEBUG,
755 			    "add_mfc no upcall h %d o %s g %s p %x\n",
756 			    hash,
757 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
758 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
759 			    mfccp->mf6cc_parent);
760 #endif
761 
762 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
763 
764 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
765 					       &mfccp->mf6cc_origin.sin6_addr)&&
766 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
767 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
768 
769 				rt->mf6c_origin     = mfccp->mf6cc_origin;
770 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
771 				rt->mf6c_parent     = mfccp->mf6cc_parent;
772 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
773 				/* initialize pkt counters per src-grp */
774 				rt->mf6c_pkt_cnt    = 0;
775 				rt->mf6c_byte_cnt   = 0;
776 				rt->mf6c_wrong_if   = 0;
777 
778 				if (rt->mf6c_expire)
779 					n6expire[hash]--;
780 				rt->mf6c_expire	   = 0;
781 			}
782 		}
783 		if (rt == NULL) {
784 			/* no upcall, so make a new entry */
785 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
786 						  M_NOWAIT);
787 			if (rt == NULL) {
788 				splx(s);
789 				return ENOBUFS;
790 			}
791 
792 			/* insert new entry at head of hash chain */
793 			rt->mf6c_origin     = mfccp->mf6cc_origin;
794 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
795 			rt->mf6c_parent     = mfccp->mf6cc_parent;
796 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
797 			/* initialize pkt counters per src-grp */
798 			rt->mf6c_pkt_cnt    = 0;
799 			rt->mf6c_byte_cnt   = 0;
800 			rt->mf6c_wrong_if   = 0;
801 			rt->mf6c_expire     = 0;
802 			rt->mf6c_stall = NULL;
803 
804 			/* link into table */
805 			rt->mf6c_next  = mf6ctable[hash];
806 			mf6ctable[hash] = rt;
807 		}
808 	}
809 	splx(s);
810 	return 0;
811 }
812 
813 #ifdef UPCALL_TIMING
814 /*
815  * collect delay statistics on the upcalls
816  */
817 static void
818 collate(t)
819 	struct timeval *t;
820 {
821 	u_long d;
822 	struct timeval tp;
823 	u_long delta;
824 
825 	GET_TIME(tp);
826 
827 	if (TV_LT(*t, tp))
828 	{
829 		TV_DELTA(tp, *t, delta);
830 
831 		d = delta >> 10;
832 		if (d > UPCALL_MAX)
833 			d = UPCALL_MAX;
834 
835 		++upcall_data[d];
836 	}
837 }
838 #endif /* UPCALL_TIMING */
839 
840 /*
841  * Delete an mfc entry
842  */
843 static int
844 del_m6fc(mfccp)
845 	struct mf6cctl *mfccp;
846 {
847 	struct sockaddr_in6 	origin;
848 	struct sockaddr_in6 	mcastgrp;
849 	struct mf6c 		*rt;
850 	struct mf6c	 	**nptr;
851 	u_long 		hash;
852 	int s;
853 
854 	origin = mfccp->mf6cc_origin;
855 	mcastgrp = mfccp->mf6cc_mcastgrp;
856 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
857 
858 #ifdef MRT6DEBUG
859 	if (mrt6debug & DEBUG_MFC)
860 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
861 		    ip6_sprintf(&origin.sin6_addr),
862 		    ip6_sprintf(&mcastgrp.sin6_addr));
863 #endif
864 
865 	s = splsoftnet();
866 
867 	nptr = &mf6ctable[hash];
868 	while ((rt = *nptr) != NULL) {
869 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
870 				       &rt->mf6c_origin.sin6_addr) &&
871 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
872 				       &rt->mf6c_mcastgrp.sin6_addr) &&
873 		    rt->mf6c_stall == NULL)
874 			break;
875 
876 		nptr = &rt->mf6c_next;
877 	}
878 	if (rt == NULL) {
879 		splx(s);
880 		return EADDRNOTAVAIL;
881 	}
882 
883 	*nptr = rt->mf6c_next;
884 	free(rt, M_MRTABLE);
885 
886 	splx(s);
887 
888 	return 0;
889 }
890 
891 static int
892 socket_send(s, mm, src)
893 	struct socket *s;
894 	struct mbuf *mm;
895 	struct sockaddr_in6 *src;
896 {
897 	if (s) {
898 		if (sbappendaddr(&s->so_rcv,
899 				 (struct sockaddr *)src,
900 				 mm, (struct mbuf *)0) != 0) {
901 			sorwakeup(s);
902 			return 0;
903 		}
904 	}
905 	m_freem(mm);
906 	return -1;
907 }
908 
909 /*
910  * IPv6 multicast forwarding function. This function assumes that the packet
911  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
912  * pointed to by "ifp", and the packet is to be relayed to other networks
913  * that have members of the packet's destination IPv6 multicast group.
914  *
915  * The packet is returned unscathed to the caller, unless it is
916  * erroneous, in which case a non-zero return value tells the caller to
917  * discard it.
918  */
919 
920 int
921 ip6_mforward(ip6, ifp, m)
922 	struct ip6_hdr *ip6;
923 	struct ifnet *ifp;
924 	struct mbuf *m;
925 {
926 	struct mf6c *rt;
927 	struct mif6 *mifp;
928 	struct mbuf *mm;
929 	int s;
930 	mifi_t mifi;
931 	long time_second = time.tv_sec;
932 
933 #ifdef MRT6DEBUG
934 	if (mrt6debug & DEBUG_FORWARD)
935 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
936 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
937 		    ifp->if_index);
938 #endif
939 
940 	/*
941 	 * Don't forward a packet with Hop limit of zero or one,
942 	 * or a packet destined to a local-only group.
943 	 */
944 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
945 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
946 		return 0;
947 	ip6->ip6_hlim--;
948 
949 	/*
950 	 * Source address check: do not forward packets with unspecified
951 	 * source. It was discussed in July 2000, on ipngwg mailing list.
952 	 * This is rather more serious than unicast cases, because some
953 	 * MLD packets can be sent with the unspecified source address
954 	 * (although such packets must normally set 1 to the hop limit field).
955 	 */
956 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
957 		ip6stat.ip6s_cantforward++;
958 		if (ip6_log_time + ip6_log_interval < time_second) {
959 			ip6_log_time = time_second;
960 			log(LOG_DEBUG,
961 			    "cannot forward "
962 			    "from %s to %s nxt %d received on %s\n",
963 			    ip6_sprintf(&ip6->ip6_src),
964 			    ip6_sprintf(&ip6->ip6_dst),
965 			    ip6->ip6_nxt,
966 			    m->m_pkthdr.rcvif ?
967 			    if_name(m->m_pkthdr.rcvif) : "?");
968 		}
969 		return 0;
970 	}
971 
972 	/*
973 	 * Determine forwarding mifs from the forwarding cache table
974 	 */
975 	s = splsoftnet();
976 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
977 
978 	/* Entry exists, so forward if necessary */
979 	if (rt) {
980 		splx(s);
981 		return (ip6_mdq(m, ifp, rt));
982 	} else {
983 		/*
984 		 * If we don't have a route for packet's origin,
985 		 * Make a copy of the packet &
986 		 * send message to routing daemon
987 		 */
988 
989 		struct mbuf *mb0;
990 		struct rtdetq *rte;
991 		u_long hash;
992 /*		int i, npkts;*/
993 #ifdef UPCALL_TIMING
994 		struct timeval tp;
995 
996 		GET_TIME(tp);
997 #endif /* UPCALL_TIMING */
998 
999 		mrt6stat.mrt6s_no_route++;
1000 #ifdef MRT6DEBUG
1001 		if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
1002 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
1003 			    ip6_sprintf(&ip6->ip6_src),
1004 			    ip6_sprintf(&ip6->ip6_dst));
1005 #endif
1006 
1007 		/*
1008 		 * Allocate mbufs early so that we don't do extra work if we
1009 		 * are just going to fail anyway.
1010 		 */
1011 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1012 					      M_NOWAIT);
1013 		if (rte == NULL) {
1014 			splx(s);
1015 			return ENOBUFS;
1016 		}
1017 		mb0 = m_copy(m, 0, M_COPYALL);
1018 		/*
1019 		 * Pullup packet header if needed before storing it,
1020 		 * as other references may modify it in the meantime.
1021 		 */
1022 		if (mb0 &&
1023 		    (M_READONLY(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1024 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1025 		if (mb0 == NULL) {
1026 			free(rte, M_MRTABLE);
1027 			splx(s);
1028 			return ENOBUFS;
1029 		}
1030 
1031 		/* is there an upcall waiting for this packet? */
1032 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1033 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1034 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1035 					       &rt->mf6c_origin.sin6_addr) &&
1036 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1037 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1038 			    (rt->mf6c_stall != NULL))
1039 				break;
1040 		}
1041 
1042 		if (rt == NULL) {
1043 			struct mrt6msg *im;
1044 			struct omrt6msg *oim;
1045 
1046 			/* no upcall, so make a new entry */
1047 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1048 						  M_NOWAIT);
1049 			if (rt == NULL) {
1050 				free(rte, M_MRTABLE);
1051 				m_freem(mb0);
1052 				splx(s);
1053 				return ENOBUFS;
1054 			}
1055 			/*
1056 			 * Make a copy of the header to send to the user
1057 			 * level process
1058 			 */
1059 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1060 
1061 			if (mm == NULL) {
1062 				free(rte, M_MRTABLE);
1063 				m_freem(mb0);
1064 				free(rt, M_MRTABLE);
1065 				splx(s);
1066 				return ENOBUFS;
1067 			}
1068 
1069 			/*
1070 			 * Send message to routing daemon
1071 			 */
1072 			sin6.sin6_addr = ip6->ip6_src;
1073 
1074 			im = NULL;
1075 			oim = NULL;
1076 			switch (ip6_mrouter_ver) {
1077 			case MRT6_OINIT:
1078 				oim = mtod(mm, struct omrt6msg *);
1079 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1080 				oim->im6_mbz = 0;
1081 				break;
1082 			case MRT6_INIT:
1083 				im = mtod(mm, struct mrt6msg *);
1084 				im->im6_msgtype = MRT6MSG_NOCACHE;
1085 				im->im6_mbz = 0;
1086 				break;
1087 			default:
1088 				free(rte, M_MRTABLE);
1089 				m_freem(mb0);
1090 				free(rt, M_MRTABLE);
1091 				splx(s);
1092 				return EINVAL;
1093 			}
1094 
1095 #ifdef MRT6DEBUG
1096 			if (mrt6debug & DEBUG_FORWARD)
1097 				log(LOG_DEBUG,
1098 				    "getting the iif info in the kernel\n");
1099 #endif
1100 
1101 			for (mifp = mif6table, mifi = 0;
1102 			     mifi < nummifs && mifp->m6_ifp != ifp;
1103 			     mifp++, mifi++)
1104 				;
1105 
1106 			switch (ip6_mrouter_ver) {
1107 			case MRT6_OINIT:
1108 				oim->im6_mif = mifi;
1109 				break;
1110 			case MRT6_INIT:
1111 				im->im6_mif = mifi;
1112 				break;
1113 			}
1114 
1115 			if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1116 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1117 				    "socket queue full\n");
1118 				mrt6stat.mrt6s_upq_sockfull++;
1119 				free(rte, M_MRTABLE);
1120 				m_freem(mb0);
1121 				free(rt, M_MRTABLE);
1122 				splx(s);
1123 				return ENOBUFS;
1124 			}
1125 
1126 			mrt6stat.mrt6s_upcalls++;
1127 
1128 			/* insert new entry at head of hash chain */
1129 			bzero(rt, sizeof(*rt));
1130 			rt->mf6c_origin.sin6_family = AF_INET6;
1131 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1132 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1133 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1134 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1135 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1136 			rt->mf6c_expire = UPCALL_EXPIRE;
1137 			n6expire[hash]++;
1138 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1139 
1140 			/* link into table */
1141 			rt->mf6c_next  = mf6ctable[hash];
1142 			mf6ctable[hash] = rt;
1143 			/* Add this entry to the end of the queue */
1144 			rt->mf6c_stall = rte;
1145 		} else {
1146 			/* determine if q has overflowed */
1147 			struct rtdetq **p;
1148 			int npkts = 0;
1149 
1150 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1151 				if (++npkts > MAX_UPQ6) {
1152 					mrt6stat.mrt6s_upq_ovflw++;
1153 					free(rte, M_MRTABLE);
1154 					m_freem(mb0);
1155 					splx(s);
1156 					return 0;
1157 				}
1158 
1159 			/* Add this entry to the end of the queue */
1160 			*p = rte;
1161 		}
1162 
1163 		rte->next = NULL;
1164 		rte->m = mb0;
1165 		rte->ifp = ifp;
1166 #ifdef UPCALL_TIMING
1167 		rte->t = tp;
1168 #endif /* UPCALL_TIMING */
1169 
1170 		splx(s);
1171 
1172 		return 0;
1173 	}
1174 }
1175 
1176 /*
1177  * Clean up cache entries if upcalls are not serviced
1178  * Call from the Slow Timeout mechanism, every half second.
1179  */
1180 static void
1181 expire_upcalls(unused)
1182 	void *unused;
1183 {
1184 	struct rtdetq *rte;
1185 	struct mf6c *mfc, **nptr;
1186 	int i;
1187 	int s;
1188 
1189 	s = splsoftnet();
1190 	for (i = 0; i < MF6CTBLSIZ; i++) {
1191 		if (n6expire[i] == 0)
1192 			continue;
1193 		nptr = &mf6ctable[i];
1194 		while ((mfc = *nptr) != NULL) {
1195 			rte = mfc->mf6c_stall;
1196 			/*
1197 			 * Skip real cache entries
1198 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1199 			 * If it expires now
1200 			 */
1201 			if (rte != NULL &&
1202 			    mfc->mf6c_expire != 0 &&
1203 			    --mfc->mf6c_expire == 0) {
1204 #ifdef MRT6DEBUG
1205 				if (mrt6debug & DEBUG_EXPIRE)
1206 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1207 					    ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1208 					    ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1209 #endif
1210 				/*
1211 				 * drop all the packets
1212 				 * free the mbuf with the pkt, if, timing info
1213 				 */
1214 				do {
1215 					struct rtdetq *n = rte->next;
1216 					m_freem(rte->m);
1217 					free(rte, M_MRTABLE);
1218 					rte = n;
1219 				} while (rte != NULL);
1220 				mrt6stat.mrt6s_cache_cleanups++;
1221 				n6expire[i]--;
1222 
1223 				*nptr = mfc->mf6c_next;
1224 				free(mfc, M_MRTABLE);
1225 			} else {
1226 				nptr = &mfc->mf6c_next;
1227 			}
1228 		}
1229 	}
1230 	splx(s);
1231 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1232 	    expire_upcalls, NULL);
1233 }
1234 
1235 /*
1236  * Packet forwarding routine once entry in the cache is made
1237  */
1238 static int
1239 ip6_mdq(m, ifp, rt)
1240 	struct mbuf *m;
1241 	struct ifnet *ifp;
1242 	struct mf6c *rt;
1243 {
1244 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1245 	mifi_t mifi, iif;
1246 	struct mif6 *mifp;
1247 	int plen = m->m_pkthdr.len;
1248 
1249 /*
1250  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1251  * input, they shouldn't get counted on output, so statistics keeping is
1252  * separate.
1253  */
1254 
1255 #define MC6_SEND(ip6, mifp, m) do {				\
1256 		if ((mifp)->m6_flags & MIFF_REGISTER)		\
1257 		    register_send((ip6), (mifp), (m));		\
1258 		else						\
1259 		    phyint_send((ip6), (mifp), (m));		\
1260 } while (0)
1261 
1262 	/*
1263 	 * Don't forward if it didn't arrive from the parent mif
1264 	 * for its origin.
1265 	 */
1266 	mifi = rt->mf6c_parent;
1267 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1268 		/* came in the wrong interface */
1269 #ifdef MRT6DEBUG
1270 		if (mrt6debug & DEBUG_FORWARD)
1271 			log(LOG_DEBUG,
1272 			    "wrong if: ifid %d mifi %d mififid %x\n",
1273 			    ifp->if_index, mifi,
1274 			    mif6table[mifi].m6_ifp->if_index);
1275 #endif
1276 		mrt6stat.mrt6s_wrong_if++;
1277 		rt->mf6c_wrong_if++;
1278 		/*
1279 		 * If we are doing PIM processing, and we are forwarding
1280 		 * packets on this interface, send a message to the
1281 		 * routing daemon.
1282 		 */
1283 		/* have to make sure this is a valid mif */
1284 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1285 			if (pim6 && (m->m_flags & M_LOOP) == 0) {
1286 				/*
1287 				 * Check the M_LOOP flag to avoid an
1288 				 * unnecessary PIM assert.
1289 				 * XXX: M_LOOP is an ad-hoc hack...
1290 				 */
1291 				static struct sockaddr_in6 sin6 =
1292 				{ sizeof(sin6), AF_INET6 };
1293 
1294 				struct mbuf *mm;
1295 				struct mrt6msg *im;
1296 				struct omrt6msg *oim;
1297 
1298 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1299 				if (mm &&
1300 				    (M_READONLY(mm) ||
1301 				     mm->m_len < sizeof(struct ip6_hdr)))
1302 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1303 				if (mm == NULL)
1304 					return ENOBUFS;
1305 
1306 				oim = NULL;
1307 				im = NULL;
1308 				switch (ip6_mrouter_ver) {
1309 				case MRT6_OINIT:
1310 					oim = mtod(mm, struct omrt6msg *);
1311 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1312 					oim->im6_mbz = 0;
1313 					break;
1314 				case MRT6_INIT:
1315 					im = mtod(mm, struct mrt6msg *);
1316 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1317 					im->im6_mbz = 0;
1318 					break;
1319 				default:
1320 					m_freem(mm);
1321 					return EINVAL;
1322 				}
1323 
1324 				for (mifp = mif6table, iif = 0;
1325 				     iif < nummifs && mifp &&
1326 					     mifp->m6_ifp != ifp;
1327 				     mifp++, iif++)
1328 					;
1329 
1330 				switch (ip6_mrouter_ver) {
1331 				case MRT6_OINIT:
1332 					oim->im6_mif = iif;
1333 					sin6.sin6_addr = oim->im6_src;
1334 					break;
1335 				case MRT6_INIT:
1336 					im->im6_mif = iif;
1337 					sin6.sin6_addr = im->im6_src;
1338 					break;
1339 				}
1340 
1341 				mrt6stat.mrt6s_upcalls++;
1342 
1343 				if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1344 #ifdef MRT6DEBUG
1345 					if (mrt6debug)
1346 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1347 #endif
1348 					++mrt6stat.mrt6s_upq_sockfull;
1349 					return ENOBUFS;
1350 				}	/* if socket Q full */
1351 			}		/* if PIM */
1352 		return 0;
1353 	}			/* if wrong iif */
1354 
1355 	/* If I sourced this packet, it counts as output, else it was input. */
1356 	if (m->m_pkthdr.rcvif == NULL) {
1357 		/* XXX: is rcvif really NULL when output?? */
1358 		mif6table[mifi].m6_pkt_out++;
1359 		mif6table[mifi].m6_bytes_out += plen;
1360 	} else {
1361 		mif6table[mifi].m6_pkt_in++;
1362 		mif6table[mifi].m6_bytes_in += plen;
1363 	}
1364 	rt->mf6c_pkt_cnt++;
1365 	rt->mf6c_byte_cnt += plen;
1366 
1367 	/*
1368 	 * For each mif, forward a copy of the packet if there are group
1369 	 * members downstream on the interface.
1370 	 */
1371 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1372 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1373 #ifdef notyet
1374 			/*
1375 			 * check if the outgoing packet is going to break
1376 			 * a scope boundary.
1377 			 * XXX For packets through PIM register tunnel
1378 			 * interface, we believe a routing daemon.
1379 			 */
1380 			if ((mif6table[rt->mf6c_parent].m6_flags &
1381 			     MIFF_REGISTER) == 0 &&
1382 			    (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1383 			    (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1384 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1385 					      &ip6->ip6_dst) ||
1386 			     in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1387 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1388 					      &ip6->ip6_src))) {
1389 				ip6stat.ip6s_badscope++;
1390 				continue;
1391 			}
1392 #endif
1393 
1394 			mifp->m6_pkt_out++;
1395 			mifp->m6_bytes_out += plen;
1396 			MC6_SEND(ip6, mifp, m);
1397 		}
1398 	return 0;
1399 }
1400 
1401 static void
1402 phyint_send(ip6, mifp, m)
1403     struct ip6_hdr *ip6;
1404     struct mif6 *mifp;
1405     struct mbuf *m;
1406 {
1407 	struct mbuf *mb_copy;
1408 	struct ifnet *ifp = mifp->m6_ifp;
1409 	int error = 0;
1410 	int s = splsoftnet();
1411 	static struct route_in6 ro;
1412 	struct	in6_multi *in6m;
1413 	struct sockaddr_in6 *dst6;
1414 	u_long linkmtu;
1415 
1416 	/*
1417 	 * Make a new reference to the packet; make sure that
1418 	 * the IPv6 header is actually copied, not just referenced,
1419 	 * so that ip6_output() only scribbles on the copy.
1420 	 */
1421 	mb_copy = m_copy(m, 0, M_COPYALL);
1422 	if (mb_copy &&
1423 	    (M_READONLY(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1424 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1425 	if (mb_copy == NULL) {
1426 		splx(s);
1427 		return;
1428 	}
1429 	/* set MCAST flag to the outgoing packet */
1430 	mb_copy->m_flags |= M_MCAST;
1431 
1432 	/*
1433 	 * If we sourced the packet, call ip6_output since we may devide
1434 	 * the packet into fragments when the packet is too big for the
1435 	 * outgoing interface.
1436 	 * Otherwise, we can simply send the packet to the interface
1437 	 * sending queue.
1438 	 */
1439 	if (m->m_pkthdr.rcvif == NULL) {
1440 		struct ip6_moptions im6o;
1441 
1442 		im6o.im6o_multicast_ifp = ifp;
1443 		/* XXX: ip6_output will override ip6->ip6_hlim */
1444 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1445 		im6o.im6o_multicast_loop = 1;
1446 		error = ip6_output(mb_copy, NULL, &ro,
1447 				   IPV6_FORWARDING, &im6o, NULL);
1448 
1449 #ifdef MRT6DEBUG
1450 		if (mrt6debug & DEBUG_XMIT)
1451 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1452 			    mifp - mif6table, error);
1453 #endif
1454 		splx(s);
1455 		return;
1456 	}
1457 
1458 	/*
1459 	 * If we belong to the destination multicast group
1460 	 * on the outgoing interface, loop back a copy.
1461 	 */
1462 	dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
1463 	IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1464 	if (in6m != NULL) {
1465 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1466 		dst6->sin6_family = AF_INET6;
1467 		dst6->sin6_addr = ip6->ip6_dst;
1468 		ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst);
1469 	}
1470 	/*
1471 	 * Put the packet into the sending queue of the outgoing interface
1472 	 * if it would fit in the MTU of the interface.
1473 	 */
1474 	linkmtu = IN6_LINKMTU(ifp);
1475 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1476 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1477 		dst6->sin6_family = AF_INET6;
1478 		dst6->sin6_addr = ip6->ip6_dst;
1479 		/*
1480 		 * We just call if_output instead of nd6_output here, since
1481 		 * we need no ND for a multicast forwarded packet...right?
1482 		 */
1483 		error = (*ifp->if_output)(ifp, mb_copy,
1484 		    (struct sockaddr *)&ro.ro_dst, NULL);
1485 #ifdef MRT6DEBUG
1486 		if (mrt6debug & DEBUG_XMIT)
1487 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1488 			    mifp - mif6table, error);
1489 #endif
1490 	} else {
1491 #ifdef MULTICAST_PMTUD
1492 		icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1493 #else
1494 #ifdef MRT6DEBUG
1495 		if (mrt6debug & DEBUG_XMIT)
1496 			log(LOG_DEBUG,
1497 			    "phyint_send: packet too big on %s o %s g %s"
1498 			    " size %d(discarded)\n",
1499 			    if_name(ifp),
1500 			    ip6_sprintf(&ip6->ip6_src),
1501 			    ip6_sprintf(&ip6->ip6_dst),
1502 			    mb_copy->m_pkthdr.len);
1503 #endif /* MRT6DEBUG */
1504 		m_freem(mb_copy); /* simply discard the packet */
1505 #endif
1506 	}
1507 
1508 	splx(s);
1509 }
1510 
1511 static int
1512 register_send(ip6, mif, m)
1513 	struct ip6_hdr *ip6;
1514 	struct mif6 *mif;
1515 	struct mbuf *m;
1516 {
1517 	struct mbuf *mm;
1518 	int i, len = m->m_pkthdr.len;
1519 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1520 	struct mrt6msg *im6;
1521 
1522 #ifdef MRT6DEBUG
1523 	if (mrt6debug)
1524 		log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1525 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1526 #endif
1527 	++pim6stat.pim6s_snd_registers;
1528 
1529 	/* Make a copy of the packet to send to the user level process */
1530 	MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1531 	if (mm == NULL)
1532 		return ENOBUFS;
1533 	mm->m_data += max_linkhdr;
1534 	mm->m_len = sizeof(struct ip6_hdr);
1535 
1536 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1537 		m_freem(mm);
1538 		return ENOBUFS;
1539 	}
1540 	i = MHLEN - M_LEADINGSPACE(mm);
1541 	if (i > len)
1542 		i = len;
1543 	mm = m_pullup(mm, i);
1544 	if (mm == NULL)
1545 		return ENOBUFS;
1546 /* TODO: check it! */
1547 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1548 
1549 	/*
1550 	 * Send message to routing daemon
1551 	 */
1552 	sin6.sin6_addr = ip6->ip6_src;
1553 
1554 	im6 = mtod(mm, struct mrt6msg *);
1555 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1556 	im6->im6_mbz          = 0;
1557 
1558 	im6->im6_mif = mif - mif6table;
1559 
1560 	/* iif info is not given for reg. encap.n */
1561 	mrt6stat.mrt6s_upcalls++;
1562 
1563 	if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1564 #ifdef MRT6DEBUG
1565 		if (mrt6debug)
1566 			log(LOG_WARNING,
1567 			    "register_send: ip6_mrouter socket queue full\n");
1568 #endif
1569 		++mrt6stat.mrt6s_upq_sockfull;
1570 		return ENOBUFS;
1571 	}
1572 	return 0;
1573 }
1574 
1575 /*
1576  * PIM sparse mode hook
1577  * Receives the pim control messages, and passes them up to the listening
1578  * socket, using rip6_input.
1579  * The only message processed is the REGISTER pim message; the pim header
1580  * is stripped off, and the inner packet is passed to register_mforward.
1581  */
1582 int
1583 pim6_input(mp, offp, proto)
1584 	struct mbuf **mp;
1585 	int *offp, proto;
1586 {
1587 	struct pim *pim; /* pointer to a pim struct */
1588 	struct ip6_hdr *ip6;
1589 	int pimlen;
1590 	struct mbuf *m = *mp;
1591 	int minlen;
1592 	int off = *offp;
1593 
1594 	++pim6stat.pim6s_rcv_total;
1595 
1596 	ip6 = mtod(m, struct ip6_hdr *);
1597 	pimlen = m->m_pkthdr.len - *offp;
1598 
1599 	/*
1600 	 * Validate lengths
1601 	 */
1602 	if (pimlen < PIM_MINLEN) {
1603 		++pim6stat.pim6s_rcv_tooshort;
1604 #ifdef MRT6DEBUG
1605 		if (mrt6debug & DEBUG_PIM)
1606 			log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1607 #endif
1608 		m_freem(m);
1609 		return (IPPROTO_DONE);
1610 	}
1611 
1612 	/*
1613 	 * if the packet is at least as big as a REGISTER, go ahead
1614 	 * and grab the PIM REGISTER header size, to avoid another
1615 	 * possible m_pullup() later.
1616 	 *
1617 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1618 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1619 	 */
1620 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1621 
1622 	/*
1623 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1624 	 * possibly the PIM REGISTER header
1625 	 */
1626 #ifndef PULLDOWN_TEST
1627 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1628 	/* adjust pointer */
1629 	ip6 = mtod(m, struct ip6_hdr *);
1630 
1631 	/* adjust mbuf to point to the PIM header */
1632 	pim = (struct pim *)((caddr_t)ip6 + off);
1633 #else
1634 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1635 	if (pim == NULL) {
1636 		pim6stat.pim6s_rcv_tooshort++;
1637 		return IPPROTO_DONE;
1638 	}
1639 #endif
1640 
1641 #define PIM6_CHECKSUM
1642 #ifdef PIM6_CHECKSUM
1643 	{
1644 		int cksumlen;
1645 
1646 		/*
1647 		 * Validate checksum.
1648 		 * If PIM REGISTER, exclude the data packet
1649 		 */
1650 		if (pim->pim_type == PIM_REGISTER)
1651 			cksumlen = PIM_MINLEN;
1652 		else
1653 			cksumlen = pimlen;
1654 
1655 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1656 			++pim6stat.pim6s_rcv_badsum;
1657 #ifdef MRT6DEBUG
1658 			if (mrt6debug & DEBUG_PIM)
1659 				log(LOG_DEBUG,
1660 				    "pim6_input: invalid checksum\n");
1661 #endif
1662 			m_freem(m);
1663 			return (IPPROTO_DONE);
1664 		}
1665 	}
1666 #endif /* PIM_CHECKSUM */
1667 
1668 	/* PIM version check */
1669 	if (pim->pim_ver != PIM_VERSION) {
1670 		++pim6stat.pim6s_rcv_badversion;
1671 #ifdef MRT6DEBUG
1672 		log(LOG_ERR,
1673 		    "pim6_input: incorrect version %d, expecting %d\n",
1674 		    pim->pim_ver, PIM_VERSION);
1675 #endif
1676 		m_freem(m);
1677 		return (IPPROTO_DONE);
1678 	}
1679 
1680 	if (pim->pim_type == PIM_REGISTER) {
1681 		/*
1682 		 * since this is a REGISTER, we'll make a copy of the register
1683 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1684 		 * routing daemon.
1685 		 */
1686 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1687 
1688 		struct mbuf *mcp;
1689 		struct ip6_hdr *eip6;
1690 		u_int32_t *reghdr;
1691 		int rc;
1692 
1693 		++pim6stat.pim6s_rcv_registers;
1694 
1695 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1696 #ifdef MRT6DEBUG
1697 			if (mrt6debug & DEBUG_PIM)
1698 				log(LOG_DEBUG,
1699 				    "pim6_input: register mif not set: %d\n",
1700 				    reg_mif_num);
1701 #endif
1702 			m_freem(m);
1703 			return (IPPROTO_DONE);
1704 		}
1705 
1706 		reghdr = (u_int32_t *)(pim + 1);
1707 
1708 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1709 			goto pim6_input_to_daemon;
1710 
1711 		/*
1712 		 * Validate length
1713 		 */
1714 		if (pimlen < PIM6_REG_MINLEN) {
1715 			++pim6stat.pim6s_rcv_tooshort;
1716 			++pim6stat.pim6s_rcv_badregisters;
1717 #ifdef MRT6DEBUG
1718 			log(LOG_ERR,
1719 			    "pim6_input: register packet size too "
1720 			    "small %d from %s\n",
1721 			    pimlen, ip6_sprintf(&ip6->ip6_src));
1722 #endif
1723 			m_freem(m);
1724 			return (IPPROTO_DONE);
1725 		}
1726 
1727 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1728 #ifdef MRT6DEBUG
1729 		if (mrt6debug & DEBUG_PIM)
1730 			log(LOG_DEBUG,
1731 			    "pim6_input[register], eip6: %s -> %s, "
1732 			    "eip6 plen %d\n",
1733 			    ip6_sprintf(&eip6->ip6_src),
1734 			    ip6_sprintf(&eip6->ip6_dst),
1735 			    ntohs(eip6->ip6_plen));
1736 #endif
1737 
1738 		/* verify the version number of the inner packet */
1739 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1740 			++pim6stat.pim6s_rcv_badregisters;
1741 #ifdef MRT6DEBUG
1742 			log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1743 			    "of the inner packet\n",
1744 			    (eip6->ip6_vfc & IPV6_VERSION));
1745 #endif
1746 			m_freem(m);
1747 			return (IPPROTO_NONE);
1748 		}
1749 
1750 		/* verify the inner packet is destined to a mcast group */
1751 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1752 			++pim6stat.pim6s_rcv_badregisters;
1753 #ifdef MRT6DEBUG
1754 			if (mrt6debug & DEBUG_PIM)
1755 				log(LOG_DEBUG,
1756 				    "pim6_input: inner packet of register "
1757 				    "is not multicast %s\n",
1758 				    ip6_sprintf(&eip6->ip6_dst));
1759 #endif
1760 			m_freem(m);
1761 			return (IPPROTO_DONE);
1762 		}
1763 
1764 		/*
1765 		 * make a copy of the whole header to pass to the daemon later.
1766 		 */
1767 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1768 		if (mcp == NULL) {
1769 #ifdef MRT6DEBUG
1770 			log(LOG_ERR,
1771 			    "pim6_input: pim register: "
1772 			    "could not copy register head\n");
1773 #endif
1774 			m_freem(m);
1775 			return (IPPROTO_DONE);
1776 		}
1777 
1778 		/*
1779 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1780 		 */
1781 		m_adj(m, off + PIM_MINLEN);
1782 #ifdef MRT6DEBUG
1783 		if (mrt6debug & DEBUG_PIM) {
1784 			log(LOG_DEBUG,
1785 			    "pim6_input: forwarding decapsulated register: "
1786 			    "src %s, dst %s, mif %d\n",
1787 			    ip6_sprintf(&eip6->ip6_src),
1788 			    ip6_sprintf(&eip6->ip6_dst),
1789 			    reg_mif_num);
1790 		}
1791 #endif
1792 
1793 		rc = looutput(mif6table[reg_mif_num].m6_ifp, m,
1794 			      (struct sockaddr *) &dst,
1795 			      (struct rtentry *) NULL);
1796 
1797 		/* prepare the register head to send to the mrouting daemon */
1798 		m = mcp;
1799 	}
1800 
1801 	/*
1802 	 * Pass the PIM message up to the daemon; if it is a register message
1803 	 * pass the 'head' only up to the daemon. This includes the
1804 	 * encapsulator ip6 header, pim header, register header and the
1805 	 * encapsulated ip6 header.
1806 	 */
1807   pim6_input_to_daemon:
1808 	rip6_input(&m, offp, proto);
1809 	return (IPPROTO_DONE);
1810 }
1811