xref: /dragonfly/sys/netinet6/ip6_mroute.c (revision 2cd2d2b5)
1 /*	$FreeBSD: src/sys/netinet6/ip6_mroute.c,v 1.2.2.9 2003/01/23 21:06:47 sam Exp $	*/
2 /*	$DragonFly: src/sys/netinet6/ip6_mroute.c,v 1.8 2004/06/02 14:43:01 eirikn Exp $	*/
3 /*	$KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $	*/
4 
5 /*
6  * Copyright (C) 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*	BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp	*/
35 
36 /*
37  * IP multicast forwarding procedures
38  *
39  * Written by David Waitzman, BBN Labs, August 1988.
40  * Modified by Steve Deering, Stanford, February 1989.
41  * Modified by Mark J. Steiglitz, Stanford, May, 1991
42  * Modified by Van Jacobson, LBL, January 1993
43  * Modified by Ajit Thyagarajan, PARC, August 1993
44  * Modified by Bill Fenenr, PARC, April 1994
45  *
46  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
47  */
48 
49 #include "opt_inet.h"
50 #include "opt_inet6.h"
51 
52 #include <sys/param.h>
53 #include <sys/systm.h>
54 #include <sys/callout.h>
55 #include <sys/malloc.h>
56 #include <sys/mbuf.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sockio.h>
60 #include <sys/protosw.h>
61 #include <sys/errno.h>
62 #include <sys/time.h>
63 #include <sys/kernel.h>
64 #include <sys/syslog.h>
65 
66 #include <net/if.h>
67 #include <net/route.h>
68 #include <net/raw_cb.h>
69 
70 #include <netinet/in.h>
71 #include <netinet/in_var.h>
72 
73 #include <netinet/ip6.h>
74 #include <netinet6/ip6_var.h>
75 #include <netinet6/ip6_mroute.h>
76 #include <netinet6/pim6.h>
77 #include <netinet6/pim6_var.h>
78 
79 #include <net/net_osdep.h>
80 
81 static MALLOC_DEFINE(M_MRTABLE, "mf6c", "multicast forwarding cache entry");
82 
83 #define M_HASCL(m) ((m)->m_flags & M_EXT)
84 
85 static int ip6_mdq (struct mbuf *, struct ifnet *, struct mf6c *);
86 static void phyint_send (struct ip6_hdr *, struct mif6 *, struct mbuf *);
87 
88 static int set_pim6 (int *);
89 static int socket_send (struct socket *, struct mbuf *,
90 			    struct sockaddr_in6 *);
91 static int register_send (struct ip6_hdr *, struct mif6 *,
92 			      struct mbuf *);
93 
94 /*
95  * Globals.  All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static,
96  * except for netstat or debugging purposes.
97  */
98 struct socket  *ip6_mrouter = NULL;
99 int		ip6_mrouter_ver = 0;
100 int		ip6_mrtproto = IPPROTO_PIM;    /* for netstat only */
101 struct mrt6stat	mrt6stat;
102 
103 #define NO_RTE_FOUND 	0x1
104 #define RTE_FOUND	0x2
105 
106 struct mf6c	*mf6ctable[MF6CTBLSIZ];
107 u_char		n6expire[MF6CTBLSIZ];
108 static struct mif6 mif6table[MAXMIFS];
109 #ifdef MRT6DEBUG
110 u_int		mrt6debug = 0;	  /* debug level 	*/
111 #define		DEBUG_MFC	0x02
112 #define		DEBUG_FORWARD	0x04
113 #define		DEBUG_EXPIRE	0x08
114 #define		DEBUG_XMIT	0x10
115 #define         DEBUG_REG       0x20
116 #define         DEBUG_PIM       0x40
117 #endif
118 
119 static void	expire_upcalls (void *);
120 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
121 #define		UPCALL_EXPIRE	6		/* number of timeouts */
122 
123 #ifdef INET
124 #ifdef MROUTING
125 extern struct socket *ip_mrouter;
126 #endif
127 #endif
128 
129 /*
130  * 'Interfaces' associated with decapsulator (so we can tell
131  * packets that went through it from ones that get reflected
132  * by a broken gateway).  These interfaces are never linked into
133  * the system ifnet list & no routes point to them.  I.e., packets
134  * can't be sent this way.  They only exist as a placeholder for
135  * multicast source verification.
136  */
137 struct ifnet multicast_register_if;
138 
139 #define ENCAP_HOPS 64
140 
141 /*
142  * Private variables.
143  */
144 static mifi_t nummifs = 0;
145 static mifi_t reg_mif_num = (mifi_t)-1;
146 
147 static struct pim6stat pim6stat;
148 static int pim6;
149 
150 /*
151  * Hash function for a source, group entry
152  */
153 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
154 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
155 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
156 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
157 
158 /*
159  * Find a route for a given origin IPv6 address and Multicast group address.
160  * Quality of service parameter to be added in the future!!!
161  */
162 
163 #define MF6CFIND(o, g, rt) do { \
164 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
165 	rt = NULL; \
166 	mrt6stat.mrt6s_mfc_lookups++; \
167 	while (_rt) { \
168 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
169 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
170 		    (_rt->mf6c_stall == NULL)) { \
171 			rt = _rt; \
172 			break; \
173 		} \
174 		_rt = _rt->mf6c_next; \
175 	} \
176 	if (rt == NULL) { \
177 		mrt6stat.mrt6s_mfc_misses++; \
178 	} \
179 } while (0)
180 
181 /*
182  * Macros to compute elapsed time efficiently
183  * Borrowed from Van Jacobson's scheduling code
184  */
185 #define TV_DELTA(a, b, delta) do { \
186 	    int xxs; \
187 		\
188 	    delta = (a).tv_usec - (b).tv_usec; \
189 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
190 	       switch (xxs) { \
191 		      case 2: \
192 			  delta += 1000000; \
193 			      /* fall through */ \
194 		      case 1: \
195 			  delta += 1000000; \
196 			  break; \
197 		      default: \
198 			  delta += (1000000 * xxs); \
199 	       } \
200 	    } \
201 } while (0)
202 
203 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
204 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
205 
206 #ifdef UPCALL_TIMING
207 #define UPCALL_MAX	50
208 u_long upcall_data[UPCALL_MAX + 1];
209 static void collate();
210 #endif /* UPCALL_TIMING */
211 
212 static int get_sg_cnt (struct sioc_sg_req6 *);
213 static int get_mif6_cnt (struct sioc_mif_req6 *);
214 static int ip6_mrouter_init (struct socket *, struct mbuf *, int);
215 static int add_m6if (struct mif6ctl *);
216 static int del_m6if (mifi_t *);
217 static int add_m6fc (struct mf6cctl *);
218 static int del_m6fc (struct mf6cctl *);
219 
220 static struct callout expire_upcalls_ch;
221 
222 /*
223  * Handle MRT setsockopt commands to modify the multicast routing tables.
224  */
225 int
226 ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
227 {
228 	int	error = 0;
229 	struct mbuf *m;
230 
231 	if (so != ip6_mrouter && sopt->sopt_name != MRT6_INIT)
232 		return (EACCES);
233 
234 	if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
235 		return (error);
236 	if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
237 		return (error);
238 
239 	switch (sopt->sopt_name) {
240 	case MRT6_INIT:
241 #ifdef MRT6_OINIT
242 	case MRT6_OINIT:
243 #endif
244 		error = ip6_mrouter_init(so, m, sopt->sopt_name);
245 		break;
246 	case MRT6_DONE:
247 		error = ip6_mrouter_done();
248 		break;
249 	case MRT6_ADD_MIF:
250 		error = add_m6if(mtod(m, struct mif6ctl *));
251 		break;
252 	case MRT6_DEL_MIF:
253 		error = del_m6if(mtod(m, mifi_t *));
254 		break;
255 	case MRT6_ADD_MFC:
256 		error = add_m6fc(mtod(m, struct mf6cctl *));
257 		break;
258 	case MRT6_DEL_MFC:
259 		error = del_m6fc(mtod(m, struct mf6cctl *));
260 		break;
261 	case MRT6_PIM:
262 		error = set_pim6(mtod(m, int *));
263 		break;
264 	default:
265 		error = EOPNOTSUPP;
266 		break;
267 	}
268 
269 	(void)m_freem(m);
270 	return(error);
271 }
272 
273 /*
274  * Handle MRT getsockopt commands
275  */
276 int
277 ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
278 {
279 	int error = 0;
280 
281 	if (so != ip6_mrouter) return EACCES;
282 
283 	switch (sopt->sopt_name) {
284 		case MRT6_PIM:
285 			error = sooptcopyout(sopt, &pim6, sizeof(pim6));
286 			break;
287 	}
288 	return (error);
289 }
290 
291 /*
292  * Handle ioctl commands to obtain information from the cache
293  */
294 int
295 mrt6_ioctl(int cmd, caddr_t data)
296 {
297 	int error = 0;
298 
299 	switch (cmd) {
300 	case SIOCGETSGCNT_IN6:
301 		return(get_sg_cnt((struct sioc_sg_req6 *)data));
302 		break;		/* for safety */
303 	case SIOCGETMIFCNT_IN6:
304 		return(get_mif6_cnt((struct sioc_mif_req6 *)data));
305 		break;		/* for safety */
306 	default:
307 		return (EINVAL);
308 		break;
309 	}
310 	return error;
311 }
312 
313 /*
314  * returns the packet, byte, rpf-failure count for the source group provided
315  */
316 static int
317 get_sg_cnt(struct sioc_sg_req6 *req)
318 {
319 	struct mf6c *rt;
320 	int s;
321 
322 	s = splnet();
323 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
324 	splx(s);
325 	if (rt != NULL) {
326 		req->pktcnt = rt->mf6c_pkt_cnt;
327 		req->bytecnt = rt->mf6c_byte_cnt;
328 		req->wrong_if = rt->mf6c_wrong_if;
329 	} else
330 		return(ESRCH);
331 #if 0
332 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
333 #endif
334 
335 	return 0;
336 }
337 
338 /*
339  * returns the input and output packet and byte counts on the mif provided
340  */
341 static int
342 get_mif6_cnt(struct sioc_mif_req6 *req)
343 {
344 	mifi_t mifi = req->mifi;
345 
346 	if (mifi >= nummifs)
347 		return EINVAL;
348 
349 	req->icount = mif6table[mifi].m6_pkt_in;
350 	req->ocount = mif6table[mifi].m6_pkt_out;
351 	req->ibytes = mif6table[mifi].m6_bytes_in;
352 	req->obytes = mif6table[mifi].m6_bytes_out;
353 
354 	return 0;
355 }
356 
357 static int
358 set_pim6(int *i)
359 {
360 	if ((*i != 1) && (*i != 0))
361 		return EINVAL;
362 
363 	pim6 = *i;
364 
365 	return 0;
366 }
367 
368 /*
369  * Enable multicast routing
370  */
371 static int
372 ip6_mrouter_init(struct socket *so, struct mbuf *m, int cmd)
373 {
374 	int *v;
375 
376 #ifdef MRT6DEBUG
377 	if (mrt6debug)
378 		log(LOG_DEBUG,
379 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
380 		    so->so_type, so->so_proto->pr_protocol);
381 #endif
382 
383 	if (so->so_type != SOCK_RAW ||
384 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
385 		return EOPNOTSUPP;
386 
387 	if (!m || (m->m_len != sizeof(int *)))
388 		return ENOPROTOOPT;
389 
390 	v = mtod(m, int *);
391 	if (*v != 1)
392 		return ENOPROTOOPT;
393 
394 	if (ip6_mrouter != NULL) return EADDRINUSE;
395 
396 	ip6_mrouter = so;
397 	ip6_mrouter_ver = cmd;
398 
399 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
400 	bzero((caddr_t)n6expire, sizeof(n6expire));
401 
402 	pim6 = 0;/* used for stubbing out/in pim stuff */
403 
404 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
405 	    expire_upcalls, NULL);
406 
407 #ifdef MRT6DEBUG
408 	if (mrt6debug)
409 		log(LOG_DEBUG, "ip6_mrouter_init\n");
410 #endif
411 
412 	return 0;
413 }
414 
415 /*
416  * Disable multicast routing
417  */
418 int
419 ip6_mrouter_done(void)
420 {
421 	mifi_t mifi;
422 	int i;
423 	struct ifnet *ifp;
424 	struct in6_ifreq ifr;
425 	struct mf6c *rt;
426 	struct rtdetq *rte;
427 	int s;
428 
429 	s = splnet();
430 
431 	/*
432 	 * For each phyint in use, disable promiscuous reception of all IPv6
433 	 * multicasts.
434 	 */
435 #ifdef INET
436 #ifdef MROUTING
437 	/*
438 	 * If there is still IPv4 multicast routing daemon,
439 	 * we remain interfaces to receive all muliticasted packets.
440 	 * XXX: there may be an interface in which the IPv4 multicast
441 	 * daemon is not interested...
442 	 */
443 	if (!ip_mrouter)
444 #endif
445 #endif
446 	{
447 		for (mifi = 0; mifi < nummifs; mifi++) {
448 			if (mif6table[mifi].m6_ifp &&
449 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
450 				ifr.ifr_addr.sin6_family = AF_INET6;
451 				ifr.ifr_addr.sin6_addr= in6addr_any;
452 				ifp = mif6table[mifi].m6_ifp;
453 				(*ifp->if_ioctl)(ifp, SIOCDELMULTI,
454 						 (caddr_t)&ifr,
455 						 (struct ucred *)NULL);
456 			}
457 		}
458 	}
459 #ifdef notyet
460 	bzero((caddr_t)qtable, sizeof(qtable));
461 	bzero((caddr_t)tbftable, sizeof(tbftable));
462 #endif
463 	bzero((caddr_t)mif6table, sizeof(mif6table));
464 	nummifs = 0;
465 
466 	pim6 = 0; /* used to stub out/in pim specific code */
467 
468 	callout_stop(&expire_upcalls_ch);
469 
470 	/*
471 	 * Free all multicast forwarding cache entries.
472 	 */
473 	for (i = 0; i < MF6CTBLSIZ; i++) {
474 		rt = mf6ctable[i];
475 		while (rt) {
476 			struct mf6c *frt;
477 
478 			for (rte = rt->mf6c_stall; rte != NULL; ) {
479 				struct rtdetq *n = rte->next;
480 
481 				m_free(rte->m);
482 				free(rte, M_MRTABLE);
483 				rte = n;
484 			}
485 			frt = rt;
486 			rt = rt->mf6c_next;
487 			free(frt, M_MRTABLE);
488 		}
489 	}
490 
491 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
492 
493 	/*
494 	 * Reset de-encapsulation cache
495 	 */
496 	reg_mif_num = -1;
497 
498 	ip6_mrouter = NULL;
499 	ip6_mrouter_ver = 0;
500 
501 	splx(s);
502 
503 #ifdef MRT6DEBUG
504 	if (mrt6debug)
505 		log(LOG_DEBUG, "ip6_mrouter_done\n");
506 #endif
507 
508 	return 0;
509 }
510 
511 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
512 
513 /*
514  * Add a mif to the mif table
515  */
516 static int
517 add_m6if(struct mif6ctl *mifcp)
518 {
519 	struct mif6 *mifp;
520 	struct ifnet *ifp;
521 	int error, s;
522 #ifdef notyet
523 	struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
524 #endif
525 
526 	if (mifcp->mif6c_mifi >= MAXMIFS)
527 		return EINVAL;
528 	mifp = mif6table + mifcp->mif6c_mifi;
529 	if (mifp->m6_ifp)
530 		return EADDRINUSE; /* XXX: is it appropriate? */
531 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
532 		return ENXIO;
533 	ifp = ifindex2ifnet[mifcp->mif6c_pifi];
534 
535 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
536 		if (reg_mif_num == (mifi_t)-1) {
537 			strlcpy(multicast_register_if.if_xname, "register_mif",
538 			    IFNAMSIZ);
539 			multicast_register_if.if_flags |= IFF_LOOPBACK;
540 			multicast_register_if.if_index = mifcp->mif6c_mifi;
541 			reg_mif_num = mifcp->mif6c_mifi;
542 		}
543 
544 		ifp = &multicast_register_if;
545 
546 	} /* if REGISTER */
547 	else {
548 		/* Make sure the interface supports multicast */
549 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
550 			return EOPNOTSUPP;
551 
552 		s = splnet();
553 		error = if_allmulti(ifp, 1);
554 		splx(s);
555 		if (error)
556 			return error;
557 	}
558 
559 	s = splnet();
560 	mifp->m6_flags     = mifcp->mif6c_flags;
561 	mifp->m6_ifp       = ifp;
562 #ifdef notyet
563 	/* scaling up here allows division by 1024 in critical code */
564 	mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
565 #endif
566 	/* initialize per mif pkt counters */
567 	mifp->m6_pkt_in    = 0;
568 	mifp->m6_pkt_out   = 0;
569 	mifp->m6_bytes_in  = 0;
570 	mifp->m6_bytes_out = 0;
571 	splx(s);
572 
573 	/* Adjust nummifs up if the mifi is higher than nummifs */
574 	if (nummifs <= mifcp->mif6c_mifi)
575 		nummifs = mifcp->mif6c_mifi + 1;
576 
577 #ifdef MRT6DEBUG
578 	if (mrt6debug)
579 		log(LOG_DEBUG,
580 		    "add_mif #%d, phyint %s\n",
581 		    mifcp->mif6c_mifi,
582 		    ifp->if_xname);
583 #endif
584 
585 	return 0;
586 }
587 
588 /*
589  * Delete a mif from the mif table
590  */
591 static int
592 del_m6if(mifi_t *mifip)
593 {
594 	struct mif6 *mifp = mif6table + *mifip;
595 	mifi_t mifi;
596 	struct ifnet *ifp;
597 	int s;
598 
599 	if (*mifip >= nummifs)
600 		return EINVAL;
601 	if (mifp->m6_ifp == NULL)
602 		return EINVAL;
603 
604 	s = splnet();
605 
606 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
607 		/*
608 		 * XXX: what if there is yet IPv4 multicast daemon
609 		 *      using the interface?
610 		 */
611 		ifp = mifp->m6_ifp;
612 
613 		if_allmulti(ifp, 0);
614 	}
615 
616 #ifdef notyet
617 	bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
618 	bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
619 #endif
620 	bzero((caddr_t)mifp, sizeof (*mifp));
621 
622 	/* Adjust nummifs down */
623 	for (mifi = nummifs; mifi > 0; mifi--)
624 		if (mif6table[mifi - 1].m6_ifp)
625 			break;
626 	nummifs = mifi;
627 
628 	splx(s);
629 
630 #ifdef MRT6DEBUG
631 	if (mrt6debug)
632 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
633 #endif
634 
635 	return 0;
636 }
637 
638 /*
639  * Add an mfc entry
640  */
641 static int
642 add_m6fc(struct mf6cctl *mfccp)
643 {
644 	struct mf6c *rt;
645 	u_long hash;
646 	struct rtdetq *rte;
647 	u_short nstl;
648 	int s;
649 
650 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
651 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
652 
653 	/* If an entry already exists, just update the fields */
654 	if (rt) {
655 #ifdef MRT6DEBUG
656 		if (mrt6debug & DEBUG_MFC)
657 			log(LOG_DEBUG,
658 			    "add_m6fc no upcall h %d o %s g %s p %x\n",
659 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
660 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
661 			    mfccp->mf6cc_parent);
662 #endif
663 
664 		s = splnet();
665 		rt->mf6c_parent = mfccp->mf6cc_parent;
666 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
667 		splx(s);
668 		return 0;
669 	}
670 
671 	/*
672 	 * Find the entry for which the upcall was made and update
673 	 */
674 	s = splnet();
675 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
676 			mfccp->mf6cc_mcastgrp.sin6_addr);
677 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
678 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
679 				       &mfccp->mf6cc_origin.sin6_addr) &&
680 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
681 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
682 		    (rt->mf6c_stall != NULL)) {
683 
684 			if (nstl++)
685 				log(LOG_ERR,
686 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
687 				    "multiple kernel entries",
688 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
689 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
690 				    mfccp->mf6cc_parent, rt->mf6c_stall);
691 
692 #ifdef MRT6DEBUG
693 			if (mrt6debug & DEBUG_MFC)
694 				log(LOG_DEBUG,
695 				    "add_m6fc o %s g %s p %x dbg %x\n",
696 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
697 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
698 				    mfccp->mf6cc_parent, rt->mf6c_stall);
699 #endif
700 
701 			rt->mf6c_origin     = mfccp->mf6cc_origin;
702 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
703 			rt->mf6c_parent     = mfccp->mf6cc_parent;
704 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
705 			/* initialize pkt counters per src-grp */
706 			rt->mf6c_pkt_cnt    = 0;
707 			rt->mf6c_byte_cnt   = 0;
708 			rt->mf6c_wrong_if   = 0;
709 
710 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
711 			n6expire[hash]--;
712 
713 			/* free packets Qed at the end of this entry */
714 			for (rte = rt->mf6c_stall; rte != NULL; ) {
715 				struct rtdetq *n = rte->next;
716 				ip6_mdq(rte->m, rte->ifp, rt);
717 				m_freem(rte->m);
718 #ifdef UPCALL_TIMING
719 				collate(&(rte->t));
720 #endif /* UPCALL_TIMING */
721 				free(rte, M_MRTABLE);
722 				rte = n;
723 			}
724 			rt->mf6c_stall = NULL;
725 		}
726 	}
727 
728 	/*
729 	 * It is possible that an entry is being inserted without an upcall
730 	 */
731 	if (nstl == 0) {
732 #ifdef MRT6DEBUG
733 		if (mrt6debug & DEBUG_MFC)
734 			log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n",
735 			    hash,
736 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
737 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
738 			    mfccp->mf6cc_parent);
739 #endif
740 
741 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
742 
743 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
744 					       &mfccp->mf6cc_origin.sin6_addr)&&
745 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
746 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
747 
748 				rt->mf6c_origin     = mfccp->mf6cc_origin;
749 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
750 				rt->mf6c_parent     = mfccp->mf6cc_parent;
751 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
752 				/* initialize pkt counters per src-grp */
753 				rt->mf6c_pkt_cnt    = 0;
754 				rt->mf6c_byte_cnt   = 0;
755 				rt->mf6c_wrong_if   = 0;
756 
757 				if (rt->mf6c_expire)
758 					n6expire[hash]--;
759 				rt->mf6c_expire	   = 0;
760 			}
761 		}
762 		if (rt == NULL) {
763 			/* no upcall, so make a new entry */
764 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
765 						  M_NOWAIT);
766 			if (rt == NULL) {
767 				splx(s);
768 				return ENOBUFS;
769 			}
770 
771 			/* insert new entry at head of hash chain */
772 			rt->mf6c_origin     = mfccp->mf6cc_origin;
773 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
774 			rt->mf6c_parent     = mfccp->mf6cc_parent;
775 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
776 			/* initialize pkt counters per src-grp */
777 			rt->mf6c_pkt_cnt    = 0;
778 			rt->mf6c_byte_cnt   = 0;
779 			rt->mf6c_wrong_if   = 0;
780 			rt->mf6c_expire     = 0;
781 			rt->mf6c_stall = NULL;
782 
783 			/* link into table */
784 			rt->mf6c_next  = mf6ctable[hash];
785 			mf6ctable[hash] = rt;
786 		}
787 	}
788 	splx(s);
789 	return 0;
790 }
791 
792 #ifdef UPCALL_TIMING
793 /*
794  * collect delay statistics on the upcalls
795  */
796 static void
797 collate(struct timeval *t)
798 {
799 	u_long d;
800 	struct timeval tp;
801 	u_long delta;
802 
803 	GET_TIME(tp);
804 
805 	if (TV_LT(*t, tp))
806 	{
807 		TV_DELTA(tp, *t, delta);
808 
809 		d = delta >> 10;
810 		if (d > UPCALL_MAX)
811 			d = UPCALL_MAX;
812 
813 		++upcall_data[d];
814 	}
815 }
816 #endif /* UPCALL_TIMING */
817 
818 /*
819  * Delete an mfc entry
820  */
821 static int
822 del_m6fc(struct mf6cctl *mfccp)
823 {
824 	struct sockaddr_in6 	origin;
825 	struct sockaddr_in6 	mcastgrp;
826 	struct mf6c 		*rt;
827 	struct mf6c	 	**nptr;
828 	u_long 		hash;
829 	int s;
830 
831 	origin = mfccp->mf6cc_origin;
832 	mcastgrp = mfccp->mf6cc_mcastgrp;
833 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
834 
835 #ifdef MRT6DEBUG
836 	if (mrt6debug & DEBUG_MFC)
837 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
838 		    ip6_sprintf(&origin.sin6_addr),
839 		    ip6_sprintf(&mcastgrp.sin6_addr));
840 #endif
841 
842 	s = splnet();
843 
844 	nptr = &mf6ctable[hash];
845 	while ((rt = *nptr) != NULL) {
846 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
847 				       &rt->mf6c_origin.sin6_addr) &&
848 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
849 				       &rt->mf6c_mcastgrp.sin6_addr) &&
850 		    rt->mf6c_stall == NULL)
851 			break;
852 
853 		nptr = &rt->mf6c_next;
854 	}
855 	if (rt == NULL) {
856 		splx(s);
857 		return EADDRNOTAVAIL;
858 	}
859 
860 	*nptr = rt->mf6c_next;
861 	free(rt, M_MRTABLE);
862 
863 	splx(s);
864 
865 	return 0;
866 }
867 
868 static int
869 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
870 {
871 	if (s) {
872 		if (sbappendaddr(&s->so_rcv,
873 				 (struct sockaddr *)src,
874 				 mm, (struct mbuf *)0) != 0) {
875 			sorwakeup(s);
876 			return 0;
877 		}
878 	}
879 	m_freem(mm);
880 	return -1;
881 }
882 
883 /*
884  * IPv6 multicast forwarding function. This function assumes that the packet
885  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
886  * pointed to by "ifp", and the packet is to be relayed to other networks
887  * that have members of the packet's destination IPv6 multicast group.
888  *
889  * The packet is returned unscathed to the caller, unless it is
890  * erroneous, in which case a non-zero return value tells the caller to
891  * discard it.
892  */
893 
894 int
895 ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
896 {
897 	struct mf6c *rt;
898 	struct mif6 *mifp;
899 	struct mbuf *mm;
900 	int s;
901 	mifi_t mifi;
902 
903 #ifdef MRT6DEBUG
904 	if (mrt6debug & DEBUG_FORWARD)
905 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
906 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
907 		    ifp->if_index);
908 #endif
909 
910 	/*
911 	 * Don't forward a packet with Hop limit of zero or one,
912 	 * or a packet destined to a local-only group.
913 	 */
914 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
915 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
916 		return 0;
917 	ip6->ip6_hlim--;
918 
919 	/*
920 	 * Source address check: do not forward packets with unspecified
921 	 * source. It was discussed in July 2000, on ipngwg mailing list.
922 	 * This is rather more serious than unicast cases, because some
923 	 * MLD packets can be sent with the unspecified source address
924 	 * (although such packets must normally set 1 to the hop limit field).
925 	 */
926 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
927 		ip6stat.ip6s_cantforward++;
928 		if (ip6_log_time + ip6_log_interval < time_second) {
929 			ip6_log_time = time_second;
930 			log(LOG_DEBUG,
931 			    "cannot forward "
932 			    "from %s to %s nxt %d received on %s\n",
933 			    ip6_sprintf(&ip6->ip6_src),
934 			    ip6_sprintf(&ip6->ip6_dst),
935 			    ip6->ip6_nxt,
936 			    if_name(m->m_pkthdr.rcvif));
937 		}
938 		return 0;
939 	}
940 
941 	/*
942 	 * Determine forwarding mifs from the forwarding cache table
943 	 */
944 	s = splnet();
945 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
946 
947 	/* Entry exists, so forward if necessary */
948 	if (rt) {
949 		splx(s);
950 		return (ip6_mdq(m, ifp, rt));
951 	} else {
952 		/*
953 		 * If we don't have a route for packet's origin,
954 		 * Make a copy of the packet &
955 		 * send message to routing daemon
956 		 */
957 
958 		struct mbuf *mb0;
959 		struct rtdetq *rte;
960 		u_long hash;
961 /*		int i, npkts;*/
962 #ifdef UPCALL_TIMING
963 		struct timeval tp;
964 
965 		GET_TIME(tp);
966 #endif /* UPCALL_TIMING */
967 
968 		mrt6stat.mrt6s_no_route++;
969 #ifdef MRT6DEBUG
970 		if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
971 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
972 			    ip6_sprintf(&ip6->ip6_src),
973 			    ip6_sprintf(&ip6->ip6_dst));
974 #endif
975 
976 		/*
977 		 * Allocate mbufs early so that we don't do extra work if we
978 		 * are just going to fail anyway.
979 		 */
980 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
981 					      M_NOWAIT);
982 		if (rte == NULL) {
983 			splx(s);
984 			return ENOBUFS;
985 		}
986 		mb0 = m_copy(m, 0, M_COPYALL);
987 		/*
988 		 * Pullup packet header if needed before storing it,
989 		 * as other references may modify it in the meantime.
990 		 */
991 		if (mb0 &&
992 		    (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
993 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
994 		if (mb0 == NULL) {
995 			free(rte, M_MRTABLE);
996 			splx(s);
997 			return ENOBUFS;
998 		}
999 
1000 		/* is there an upcall waiting for this packet? */
1001 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1002 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1003 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1004 					       &rt->mf6c_origin.sin6_addr) &&
1005 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1006 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1007 			    (rt->mf6c_stall != NULL))
1008 				break;
1009 		}
1010 
1011 		if (rt == NULL) {
1012 			struct mrt6msg *im;
1013 #ifdef MRT6_OINIT
1014 			struct omrt6msg *oim;
1015 #endif
1016 
1017 			/* no upcall, so make a new entry */
1018 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1019 						  M_NOWAIT);
1020 			if (rt == NULL) {
1021 				free(rte, M_MRTABLE);
1022 				m_freem(mb0);
1023 				splx(s);
1024 				return ENOBUFS;
1025 			}
1026 			/*
1027 			 * Make a copy of the header to send to the user
1028 			 * level process
1029 			 */
1030 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1031 
1032 			if (mm == NULL) {
1033 				free(rte, M_MRTABLE);
1034 				m_freem(mb0);
1035 				free(rt, M_MRTABLE);
1036 				splx(s);
1037 				return ENOBUFS;
1038 			}
1039 
1040 			/*
1041 			 * Send message to routing daemon
1042 			 */
1043 			sin6.sin6_addr = ip6->ip6_src;
1044 
1045 			im = NULL;
1046 #ifdef MRT6_OINIT
1047 			oim = NULL;
1048 #endif
1049 			switch (ip6_mrouter_ver) {
1050 #ifdef MRT6_OINIT
1051 			case MRT6_OINIT:
1052 				oim = mtod(mm, struct omrt6msg *);
1053 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1054 				oim->im6_mbz = 0;
1055 				break;
1056 #endif
1057 			case MRT6_INIT:
1058 				im = mtod(mm, struct mrt6msg *);
1059 				im->im6_msgtype = MRT6MSG_NOCACHE;
1060 				im->im6_mbz = 0;
1061 				break;
1062 			default:
1063 				free(rte, M_MRTABLE);
1064 				m_freem(mb0);
1065 				free(rt, M_MRTABLE);
1066 				splx(s);
1067 				return EINVAL;
1068 			}
1069 
1070 #ifdef MRT6DEBUG
1071 			if (mrt6debug & DEBUG_FORWARD)
1072 				log(LOG_DEBUG,
1073 				    "getting the iif info in the kernel\n");
1074 #endif
1075 
1076 			for (mifp = mif6table, mifi = 0;
1077 			     mifi < nummifs && mifp->m6_ifp != ifp;
1078 			     mifp++, mifi++)
1079 				;
1080 
1081 			switch (ip6_mrouter_ver) {
1082 #ifdef MRT6_OINIT
1083 			case MRT6_OINIT:
1084 				oim->im6_mif = mifi;
1085 				break;
1086 #endif
1087 			case MRT6_INIT:
1088 				im->im6_mif = mifi;
1089 				break;
1090 			}
1091 
1092 			if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1093 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1094 				    "socket queue full\n");
1095 				mrt6stat.mrt6s_upq_sockfull++;
1096 				free(rte, M_MRTABLE);
1097 				m_freem(mb0);
1098 				free(rt, M_MRTABLE);
1099 				splx(s);
1100 				return ENOBUFS;
1101 			}
1102 
1103 			mrt6stat.mrt6s_upcalls++;
1104 
1105 			/* insert new entry at head of hash chain */
1106 			bzero(rt, sizeof(*rt));
1107 			rt->mf6c_origin.sin6_family = AF_INET6;
1108 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1109 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1110 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1111 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1112 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1113 			rt->mf6c_expire = UPCALL_EXPIRE;
1114 			n6expire[hash]++;
1115 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1116 
1117 			/* link into table */
1118 			rt->mf6c_next  = mf6ctable[hash];
1119 			mf6ctable[hash] = rt;
1120 			/* Add this entry to the end of the queue */
1121 			rt->mf6c_stall = rte;
1122 		} else {
1123 			/* determine if q has overflowed */
1124 			struct rtdetq **p;
1125 			int npkts = 0;
1126 
1127 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1128 				if (++npkts > MAX_UPQ6) {
1129 					mrt6stat.mrt6s_upq_ovflw++;
1130 					free(rte, M_MRTABLE);
1131 					m_freem(mb0);
1132 					splx(s);
1133 					return 0;
1134 				}
1135 
1136 			/* Add this entry to the end of the queue */
1137 			*p = rte;
1138 		}
1139 
1140 		rte->next = NULL;
1141 		rte->m = mb0;
1142 		rte->ifp = ifp;
1143 #ifdef UPCALL_TIMING
1144 		rte->t = tp;
1145 #endif /* UPCALL_TIMING */
1146 
1147 		splx(s);
1148 
1149 		return 0;
1150 	}
1151 }
1152 
1153 /*
1154  * Clean up cache entries if upcalls are not serviced
1155  * Call from the Slow Timeout mechanism, every half second.
1156  */
1157 static void
1158 expire_upcalls(void *unused)
1159 {
1160 	struct rtdetq *rte;
1161 	struct mf6c *mfc, **nptr;
1162 	int i;
1163 	int s;
1164 
1165 	s = splnet();
1166 	for (i = 0; i < MF6CTBLSIZ; i++) {
1167 		if (n6expire[i] == 0)
1168 			continue;
1169 		nptr = &mf6ctable[i];
1170 		while ((mfc = *nptr) != NULL) {
1171 			rte = mfc->mf6c_stall;
1172 			/*
1173 			 * Skip real cache entries
1174 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1175 			 * If it expires now
1176 			 */
1177 			if (rte != NULL &&
1178 			    mfc->mf6c_expire != 0 &&
1179 			    --mfc->mf6c_expire == 0) {
1180 #ifdef MRT6DEBUG
1181 				if (mrt6debug & DEBUG_EXPIRE)
1182 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1183 					    ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1184 					    ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1185 #endif
1186 				/*
1187 				 * drop all the packets
1188 				 * free the mbuf with the pkt, if, timing info
1189 				 */
1190 				do {
1191 					struct rtdetq *n = rte->next;
1192 					m_freem(rte->m);
1193 					free(rte, M_MRTABLE);
1194 					rte = n;
1195 				} while (rte != NULL);
1196 				mrt6stat.mrt6s_cache_cleanups++;
1197 				n6expire[i]--;
1198 
1199 				*nptr = mfc->mf6c_next;
1200 				free(mfc, M_MRTABLE);
1201 			} else {
1202 				nptr = &mfc->mf6c_next;
1203 			}
1204 		}
1205 	}
1206 	splx(s);
1207 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1208 	    expire_upcalls, NULL);
1209 }
1210 
1211 /*
1212  * Packet forwarding routine once entry in the cache is made
1213  */
1214 static int
1215 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1216 {
1217 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1218 	mifi_t mifi, iif;
1219 	struct mif6 *mifp;
1220 	int plen = m->m_pkthdr.len;
1221 
1222 /*
1223  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1224  * input, they shouldn't get counted on output, so statistics keeping is
1225  * separate.
1226  */
1227 
1228 #define MC6_SEND(ip6, mifp, m) do {				\
1229 		if ((mifp)->m6_flags & MIFF_REGISTER)		\
1230 		    register_send((ip6), (mifp), (m));		\
1231 		else						\
1232 		    phyint_send((ip6), (mifp), (m));		\
1233 } while (0)
1234 
1235 	/*
1236 	 * Don't forward if it didn't arrive from the parent mif
1237 	 * for its origin.
1238 	 */
1239 	mifi = rt->mf6c_parent;
1240 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1241 		/* came in the wrong interface */
1242 #ifdef MRT6DEBUG
1243 		if (mrt6debug & DEBUG_FORWARD)
1244 			log(LOG_DEBUG,
1245 			    "wrong if: ifid %d mifi %d mififid %x\n",
1246 			    ifp->if_index, mifi,
1247 			    mif6table[mifi].m6_ifp->if_index);
1248 #endif
1249 		mrt6stat.mrt6s_wrong_if++;
1250 		rt->mf6c_wrong_if++;
1251 		/*
1252 		 * If we are doing PIM processing, and we are forwarding
1253 		 * packets on this interface, send a message to the
1254 		 * routing daemon.
1255 		 */
1256 		/* have to make sure this is a valid mif */
1257 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1258 			if (pim6 && (m->m_flags & M_LOOP) == 0) {
1259 				/*
1260 				 * Check the M_LOOP flag to avoid an
1261 				 * unnecessary PIM assert.
1262 				 * XXX: M_LOOP is an ad-hoc hack...
1263 				 */
1264 				static struct sockaddr_in6 sin6 =
1265 				{ sizeof(sin6), AF_INET6 };
1266 
1267 				struct mbuf *mm;
1268 				struct mrt6msg *im;
1269 #ifdef MRT6_OINIT
1270 				struct omrt6msg *oim;
1271 #endif
1272 
1273 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1274 				if (mm &&
1275 				    (M_HASCL(mm) ||
1276 				     mm->m_len < sizeof(struct ip6_hdr)))
1277 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1278 				if (mm == NULL)
1279 					return ENOBUFS;
1280 
1281 #ifdef MRT6_OINIT
1282 				oim = NULL;
1283 #endif
1284 				im = NULL;
1285 				switch (ip6_mrouter_ver) {
1286 #ifdef MRT6_OINIT
1287 				case MRT6_OINIT:
1288 					oim = mtod(mm, struct omrt6msg *);
1289 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1290 					oim->im6_mbz = 0;
1291 					break;
1292 #endif
1293 				case MRT6_INIT:
1294 					im = mtod(mm, struct mrt6msg *);
1295 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1296 					im->im6_mbz = 0;
1297 					break;
1298 				default:
1299 					m_freem(mm);
1300 					return EINVAL;
1301 				}
1302 
1303 				for (mifp = mif6table, iif = 0;
1304 				     iif < nummifs && mifp &&
1305 					     mifp->m6_ifp != ifp;
1306 				     mifp++, iif++)
1307 					;
1308 
1309 				switch (ip6_mrouter_ver) {
1310 #ifdef MRT6_OINIT
1311 				case MRT6_OINIT:
1312 					oim->im6_mif = iif;
1313 					sin6.sin6_addr = oim->im6_src;
1314 					break;
1315 #endif
1316 				case MRT6_INIT:
1317 					im->im6_mif = iif;
1318 					sin6.sin6_addr = im->im6_src;
1319 					break;
1320 				}
1321 
1322 				mrt6stat.mrt6s_upcalls++;
1323 
1324 				if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1325 #ifdef MRT6DEBUG
1326 					if (mrt6debug)
1327 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1328 #endif
1329 					++mrt6stat.mrt6s_upq_sockfull;
1330 					return ENOBUFS;
1331 				}	/* if socket Q full */
1332 			}		/* if PIM */
1333 		return 0;
1334 	}			/* if wrong iif */
1335 
1336 	/* If I sourced this packet, it counts as output, else it was input. */
1337 	if (m->m_pkthdr.rcvif == NULL) {
1338 		/* XXX: is rcvif really NULL when output?? */
1339 		mif6table[mifi].m6_pkt_out++;
1340 		mif6table[mifi].m6_bytes_out += plen;
1341 	} else {
1342 		mif6table[mifi].m6_pkt_in++;
1343 		mif6table[mifi].m6_bytes_in += plen;
1344 	}
1345 	rt->mf6c_pkt_cnt++;
1346 	rt->mf6c_byte_cnt += plen;
1347 
1348 	/*
1349 	 * For each mif, forward a copy of the packet if there are group
1350 	 * members downstream on the interface.
1351 	 */
1352 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1353 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1354 			/*
1355 			 * check if the outgoing packet is going to break
1356 			 * a scope boundary.
1357 			 * XXX For packets through PIM register tunnel
1358 			 * interface, we believe a routing daemon.
1359 			 */
1360 			if ((mif6table[rt->mf6c_parent].m6_flags &
1361 			     MIFF_REGISTER) == 0 &&
1362 			    (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1363 			    (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1364 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1365 					      &ip6->ip6_dst) ||
1366 			     in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1367 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1368 					      &ip6->ip6_src))) {
1369 				ip6stat.ip6s_badscope++;
1370 				continue;
1371 			}
1372 
1373 			mifp->m6_pkt_out++;
1374 			mifp->m6_bytes_out += plen;
1375 			MC6_SEND(ip6, mifp, m);
1376 		}
1377 	return 0;
1378 }
1379 
1380 static void
1381 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1382 {
1383 	struct mbuf *mb_copy;
1384 	struct ifnet *ifp = mifp->m6_ifp;
1385 	int error = 0;
1386 	int s = splnet();	/* needs to protect static "ro" below. */
1387 	static struct route_in6 ro;
1388 	struct	in6_multi *in6m;
1389 	struct sockaddr_in6 *dst6;
1390 
1391 	/*
1392 	 * Make a new reference to the packet; make sure that
1393 	 * the IPv6 header is actually copied, not just referenced,
1394 	 * so that ip6_output() only scribbles on the copy.
1395 	 */
1396 	mb_copy = m_copy(m, 0, M_COPYALL);
1397 	if (mb_copy &&
1398 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1399 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1400 	if (mb_copy == NULL) {
1401 		splx(s);
1402 		return;
1403 	}
1404 	/* set MCAST flag to the outgoing packet */
1405 	mb_copy->m_flags |= M_MCAST;
1406 
1407 	/*
1408 	 * If we sourced the packet, call ip6_output since we may devide
1409 	 * the packet into fragments when the packet is too big for the
1410 	 * outgoing interface.
1411 	 * Otherwise, we can simply send the packet to the interface
1412 	 * sending queue.
1413 	 */
1414 	if (m->m_pkthdr.rcvif == NULL) {
1415 		struct ip6_moptions im6o;
1416 
1417 		im6o.im6o_multicast_ifp = ifp;
1418 		/* XXX: ip6_output will override ip6->ip6_hlim */
1419 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1420 		im6o.im6o_multicast_loop = 1;
1421 		error = ip6_output(mb_copy, NULL, &ro,
1422 				   IPV6_FORWARDING, &im6o, NULL, NULL);
1423 
1424 #ifdef MRT6DEBUG
1425 		if (mrt6debug & DEBUG_XMIT)
1426 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1427 			    mifp - mif6table, error);
1428 #endif
1429 		splx(s);
1430 		return;
1431 	}
1432 
1433 	/*
1434 	 * If we belong to the destination multicast group
1435 	 * on the outgoing interface, loop back a copy.
1436 	 */
1437 	dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
1438 	IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1439 	if (in6m != NULL) {
1440 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1441 		dst6->sin6_family = AF_INET6;
1442 		dst6->sin6_addr = ip6->ip6_dst;
1443 		ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst);
1444 	}
1445 	/*
1446 	 * Put the packet into the sending queue of the outgoing interface
1447 	 * if it would fit in the MTU of the interface.
1448 	 */
1449 	if (mb_copy->m_pkthdr.len <= ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
1450 		dst6->sin6_len = sizeof(struct sockaddr_in6);
1451 		dst6->sin6_family = AF_INET6;
1452 		dst6->sin6_addr = ip6->ip6_dst;
1453 		/*
1454 		 * We just call if_output instead of nd6_output here, since
1455 		 * we need no ND for a multicast forwarded packet...right?
1456 		 */
1457 		error = (*ifp->if_output)(ifp, mb_copy,
1458 		    (struct sockaddr *)&ro.ro_dst, NULL);
1459 #ifdef MRT6DEBUG
1460 		if (mrt6debug & DEBUG_XMIT)
1461 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1462 			    mifp - mif6table, error);
1463 #endif
1464 	} else {
1465 #ifdef MULTICAST_PMTUD
1466 		icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
1467 #else
1468 #ifdef MRT6DEBUG
1469 		if (mrt6debug & DEBUG_XMIT)
1470 			log(LOG_DEBUG,
1471 			    "phyint_send: packet too big on %s o %s g %s"
1472 			    " size %d(discarded)\n",
1473 			    if_name(ifp),
1474 			    ip6_sprintf(&ip6->ip6_src),
1475 			    ip6_sprintf(&ip6->ip6_dst),
1476 			    mb_copy->m_pkthdr.len);
1477 #endif /* MRT6DEBUG */
1478 		m_freem(mb_copy); /* simply discard the packet */
1479 #endif
1480 	}
1481 
1482 	splx(s);
1483 }
1484 
1485 static int
1486 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1487 {
1488 	struct mbuf *mm;
1489 	int i, len = m->m_pkthdr.len;
1490 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1491 	struct mrt6msg *im6;
1492 
1493 #ifdef MRT6DEBUG
1494 	if (mrt6debug)
1495 		log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1496 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1497 #endif
1498 	++pim6stat.pim6s_snd_registers;
1499 
1500 	/* Make a copy of the packet to send to the user level process */
1501 	MGETHDR(mm, MB_DONTWAIT, MT_HEADER);
1502 	if (mm == NULL)
1503 		return ENOBUFS;
1504 	mm->m_pkthdr.rcvif = NULL;
1505 	mm->m_data += max_linkhdr;
1506 	mm->m_len = sizeof(struct ip6_hdr);
1507 
1508 	if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1509 		m_freem(mm);
1510 		return ENOBUFS;
1511 	}
1512 	i = MHLEN - M_LEADINGSPACE(mm);
1513 	if (i > len)
1514 		i = len;
1515 	mm = m_pullup(mm, i);
1516 	if (mm == NULL)
1517 		return ENOBUFS;
1518 /* TODO: check it! */
1519 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1520 
1521 	/*
1522 	 * Send message to routing daemon
1523 	 */
1524 	sin6.sin6_addr = ip6->ip6_src;
1525 
1526 	im6 = mtod(mm, struct mrt6msg *);
1527 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1528 	im6->im6_mbz          = 0;
1529 
1530 	im6->im6_mif = mif - mif6table;
1531 
1532 	/* iif info is not given for reg. encap.n */
1533 	mrt6stat.mrt6s_upcalls++;
1534 
1535 	if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1536 #ifdef MRT6DEBUG
1537 		if (mrt6debug)
1538 			log(LOG_WARNING,
1539 			    "register_send: ip6_mrouter socket queue full\n");
1540 #endif
1541 		++mrt6stat.mrt6s_upq_sockfull;
1542 		return ENOBUFS;
1543 	}
1544 	return 0;
1545 }
1546 
1547 /*
1548  * PIM sparse mode hook
1549  * Receives the pim control messages, and passes them up to the listening
1550  * socket, using rip6_input.
1551  * The only message processed is the REGISTER pim message; the pim header
1552  * is stripped off, and the inner packet is passed to register_mforward.
1553  */
1554 int
1555 pim6_input(struct mbuf **mp, int *offp, int proto)
1556 {
1557 	struct pim *pim; /* pointer to a pim struct */
1558 	struct ip6_hdr *ip6;
1559 	int pimlen;
1560 	struct mbuf *m = *mp;
1561 	int minlen;
1562 	int off = *offp;
1563 
1564 	++pim6stat.pim6s_rcv_total;
1565 
1566 	ip6 = mtod(m, struct ip6_hdr *);
1567 	pimlen = m->m_pkthdr.len - *offp;
1568 
1569 	/*
1570 	 * Validate lengths
1571 	 */
1572 	if (pimlen < PIM_MINLEN) {
1573 		++pim6stat.pim6s_rcv_tooshort;
1574 #ifdef MRT6DEBUG
1575 		if (mrt6debug & DEBUG_PIM)
1576 			log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1577 #endif
1578 		m_freem(m);
1579 		return(IPPROTO_DONE);
1580 	}
1581 
1582 	/*
1583 	 * if the packet is at least as big as a REGISTER, go ahead
1584 	 * and grab the PIM REGISTER header size, to avoid another
1585 	 * possible m_pullup() later.
1586 	 *
1587 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1588 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1589 	 */
1590 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1591 
1592 	/*
1593 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1594 	 * possibly the PIM REGISTER header
1595 	 */
1596 #ifndef PULLDOWN_TEST
1597 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1598 	/* adjust pointer */
1599 	ip6 = mtod(m, struct ip6_hdr *);
1600 
1601 	/* adjust mbuf to point to the PIM header */
1602 	pim = (struct pim *)((caddr_t)ip6 + off);
1603 #else
1604 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1605 	if (pim == NULL) {
1606 		pim6stat.pim6s_rcv_tooshort++;
1607 		return IPPROTO_DONE;
1608 	}
1609 #endif
1610 
1611 #define PIM6_CHECKSUM
1612 #ifdef PIM6_CHECKSUM
1613 	{
1614 		int cksumlen;
1615 
1616 		/*
1617 		 * Validate checksum.
1618 		 * If PIM REGISTER, exclude the data packet
1619 		 */
1620 		if (pim->pim_type == PIM_REGISTER)
1621 			cksumlen = PIM_MINLEN;
1622 		else
1623 			cksumlen = pimlen;
1624 
1625 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1626 			++pim6stat.pim6s_rcv_badsum;
1627 #ifdef MRT6DEBUG
1628 			if (mrt6debug & DEBUG_PIM)
1629 				log(LOG_DEBUG,
1630 				    "pim6_input: invalid checksum\n");
1631 #endif
1632 			m_freem(m);
1633 			return(IPPROTO_DONE);
1634 		}
1635 	}
1636 #endif /* PIM_CHECKSUM */
1637 
1638 	/* PIM version check */
1639 	if (pim->pim_ver != PIM_VERSION) {
1640 		++pim6stat.pim6s_rcv_badversion;
1641 #ifdef MRT6DEBUG
1642 		log(LOG_ERR,
1643 		    "pim6_input: incorrect version %d, expecting %d\n",
1644 		    pim->pim_ver, PIM_VERSION);
1645 #endif
1646 		m_freem(m);
1647 		return(IPPROTO_DONE);
1648 	}
1649 
1650 	if (pim->pim_type == PIM_REGISTER) {
1651 		/*
1652 		 * since this is a REGISTER, we'll make a copy of the register
1653 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1654 		 * routing daemon.
1655 		 */
1656 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1657 
1658 		struct mbuf *mcp;
1659 		struct ip6_hdr *eip6;
1660 		u_int32_t *reghdr;
1661 		int rc;
1662 
1663 		++pim6stat.pim6s_rcv_registers;
1664 
1665 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1666 #ifdef MRT6DEBUG
1667 			if (mrt6debug & DEBUG_PIM)
1668 				log(LOG_DEBUG,
1669 				    "pim6_input: register mif not set: %d\n",
1670 				    reg_mif_num);
1671 #endif
1672 			m_freem(m);
1673 			return(IPPROTO_DONE);
1674 		}
1675 
1676 		reghdr = (u_int32_t *)(pim + 1);
1677 
1678 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1679 			goto pim6_input_to_daemon;
1680 
1681 		/*
1682 		 * Validate length
1683 		 */
1684 		if (pimlen < PIM6_REG_MINLEN) {
1685 			++pim6stat.pim6s_rcv_tooshort;
1686 			++pim6stat.pim6s_rcv_badregisters;
1687 #ifdef MRT6DEBUG
1688 			log(LOG_ERR,
1689 			    "pim6_input: register packet size too "
1690 			    "small %d from %s\n",
1691 			    pimlen, ip6_sprintf(&ip6->ip6_src));
1692 #endif
1693 			m_freem(m);
1694 			return(IPPROTO_DONE);
1695 		}
1696 
1697 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1698 #ifdef MRT6DEBUG
1699 		if (mrt6debug & DEBUG_PIM)
1700 			log(LOG_DEBUG,
1701 			    "pim6_input[register], eip6: %s -> %s, "
1702 			    "eip6 plen %d\n",
1703 			    ip6_sprintf(&eip6->ip6_src),
1704 			    ip6_sprintf(&eip6->ip6_dst),
1705 			    ntohs(eip6->ip6_plen));
1706 #endif
1707 
1708 		/* verify the version number of the inner packet */
1709 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1710 			++pim6stat.pim6s_rcv_badregisters;
1711 #ifdef MRT6DEBUG
1712 			log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1713 			    "of the inner packet\n",
1714 			    (eip6->ip6_vfc & IPV6_VERSION));
1715 #endif
1716 			m_freem(m);
1717 			return(IPPROTO_NONE);
1718 		}
1719 
1720 		/* verify the inner packet is destined to a mcast group */
1721 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1722 			++pim6stat.pim6s_rcv_badregisters;
1723 #ifdef MRT6DEBUG
1724 			if (mrt6debug & DEBUG_PIM)
1725 				log(LOG_DEBUG,
1726 				    "pim6_input: inner packet of register "
1727 				    "is not multicast %s\n",
1728 				    ip6_sprintf(&eip6->ip6_dst));
1729 #endif
1730 			m_freem(m);
1731 			return(IPPROTO_DONE);
1732 		}
1733 
1734 		/*
1735 		 * make a copy of the whole header to pass to the daemon later.
1736 		 */
1737 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1738 		if (mcp == NULL) {
1739 #ifdef MRT6DEBUG
1740 			log(LOG_ERR,
1741 			    "pim6_input: pim register: "
1742 			    "could not copy register head\n");
1743 #endif
1744 			m_freem(m);
1745 			return(IPPROTO_DONE);
1746 		}
1747 
1748 		/*
1749 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1750 		 */
1751 		m_adj(m, off + PIM_MINLEN);
1752 #ifdef MRT6DEBUG
1753 		if (mrt6debug & DEBUG_PIM) {
1754 			log(LOG_DEBUG,
1755 			    "pim6_input: forwarding decapsulated register: "
1756 			    "src %s, dst %s, mif %d\n",
1757 			    ip6_sprintf(&eip6->ip6_src),
1758 			    ip6_sprintf(&eip6->ip6_dst),
1759 			    reg_mif_num);
1760 		}
1761 #endif
1762 
1763  		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1764 				dst.sin6_family, NULL);
1765 
1766 		/* prepare the register head to send to the mrouting daemon */
1767 		m = mcp;
1768 	}
1769 
1770 	/*
1771 	 * Pass the PIM message up to the daemon; if it is a register message
1772 	 * pass the 'head' only up to the daemon. This includes the
1773 	 * encapsulator ip6 header, pim header, register header and the
1774 	 * encapsulated ip6 header.
1775 	 */
1776   pim6_input_to_daemon:
1777 	rip6_input(&m, offp, proto);
1778 	return(IPPROTO_DONE);
1779 }
1780