xref: /freebsd/sys/netinet6/ip6_mroute.c (revision 77a0943ded95b9e6438f7db70c4a28e4d93946d4)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_mroute.c,v 1.33 2000/10/19 02:23:43 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*	BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp	*/
34 
35 /*
36  * IP multicast forwarding procedures
37  *
38  * Written by David Waitzman, BBN Labs, August 1988.
39  * Modified by Steve Deering, Stanford, February 1989.
40  * Modified by Mark J. Steiglitz, Stanford, May, 1991
41  * Modified by Van Jacobson, LBL, January 1993
42  * Modified by Ajit Thyagarajan, PARC, August 1993
43  * Modified by Bill Fenenr, PARC, April 1994
44  *
45  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
46  */
47 
48 #include "opt_inet.h"
49 #include "opt_inet6.h"
50 
51 #include <sys/param.h>
52 #include <sys/systm.h>
53 #include <sys/malloc.h>
54 #include <sys/mbuf.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sockio.h>
58 #include <sys/protosw.h>
59 #include <sys/errno.h>
60 #include <sys/time.h>
61 #include <sys/kernel.h>
62 #include <sys/syslog.h>
63 
64 #include <net/if.h>
65 #include <net/route.h>
66 #include <net/raw_cb.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 
71 #include <netinet/ip6.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/ip6_mroute.h>
74 #include <netinet6/pim6.h>
75 #include <netinet6/pim6_var.h>
76 
77 #include <net/net_osdep.h>
78 
79 static MALLOC_DEFINE(M_MRTABLE, "mf6c", "multicast forwarding cache entry");
80 
81 #define M_HASCL(m) ((m)->m_flags & M_EXT)
82 
83 static int ip6_mdq __P((struct mbuf *, struct ifnet *, struct mf6c *));
84 static void phyint_send __P((struct ip6_hdr *, struct mif6 *, struct mbuf *));
85 
86 static int set_pim6 __P((int *));
87 static int socket_send __P((struct socket *, struct mbuf *,
88 			    struct sockaddr_in6 *));
89 static int register_send __P((struct ip6_hdr *, struct mif6 *,
90 			      struct mbuf *));
91 
92 /*
93  * Globals.  All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static,
94  * except for netstat or debugging purposes.
95  */
96 struct socket  *ip6_mrouter  = NULL;
97 int		ip6_mrouter_ver = 0;
98 int		ip6_mrtproto = IPPROTO_PIM;    /* for netstat only */
99 struct mrt6stat	mrt6stat;
100 
101 #define NO_RTE_FOUND 	0x1
102 #define RTE_FOUND	0x2
103 
104 struct mf6c	*mf6ctable[MF6CTBLSIZ];
105 u_char		nexpire[MF6CTBLSIZ];
106 static struct mif6 mif6table[MAXMIFS];
107 #ifdef MRT6DEBUG
108 u_int		mrt6debug = 0;	  /* debug level 	*/
109 #define		DEBUG_MFC	0x02
110 #define		DEBUG_FORWARD	0x04
111 #define		DEBUG_EXPIRE	0x08
112 #define		DEBUG_XMIT	0x10
113 #define         DEBUG_REG       0x20
114 #define         DEBUG_PIM       0x40
115 #endif
116 
117 static void	expire_upcalls __P((void *));
118 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
119 #define		UPCALL_EXPIRE	6		/* number of timeouts */
120 
121 #ifdef INET
122 #ifdef MROUTING
123 extern struct socket *ip_mrouter;
124 #endif
125 #endif
126 
127 /*
128  * 'Interfaces' associated with decapsulator (so we can tell
129  * packets that went through it from ones that get reflected
130  * by a broken gateway).  These interfaces are never linked into
131  * the system ifnet list & no routes point to them.  I.e., packets
132  * can't be sent this way.  They only exist as a placeholder for
133  * multicast source verification.
134  */
135 struct ifnet multicast_register_if;
136 
137 #define ENCAP_HOPS 64
138 
139 /*
140  * Private variables.
141  */
142 static mifi_t nummifs = 0;
143 static mifi_t reg_mif_num = (mifi_t)-1;
144 
145 static struct pim6stat pim6stat;
146 static struct callout_handle expire_upcalls_ch;
147 
148 /*
149  * one-back cache used by ipip_input to locate a tunnel's mif
150  * given a datagram's src ip address.
151  */
152 static int pim6;
153 
154 /*
155  * Hash function for a source, group entry
156  */
157 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
158 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
159 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
160 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
161 
162 /*
163  * Find a route for a given origin IPv6 address and Multicast group address.
164  * Quality of service parameter to be added in the future!!!
165  */
166 
167 #define MF6CFIND(o, g, rt) do { \
168 	register struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
169 	rt = NULL; \
170 	mrt6stat.mrt6s_mfc_lookups++; \
171 	while (_rt) { \
172 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
173 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
174 		    (_rt->mf6c_stall == NULL)) { \
175 			rt = _rt; \
176 			break; \
177 		} \
178 		_rt = _rt->mf6c_next; \
179 	} \
180 	if (rt == NULL) { \
181 		mrt6stat.mrt6s_mfc_misses++; \
182 	} \
183 } while (0)
184 
185 /*
186  * Macros to compute elapsed time efficiently
187  * Borrowed from Van Jacobson's scheduling code
188  */
189 #define TV_DELTA(a, b, delta) do { \
190 	    register int xxs; \
191 		\
192 	    delta = (a).tv_usec - (b).tv_usec; \
193 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
194 	       switch (xxs) { \
195 		      case 2: \
196 			  delta += 1000000; \
197 			      /* fall through */ \
198 		      case 1: \
199 			  delta += 1000000; \
200 			  break; \
201 		      default: \
202 			  delta += (1000000 * xxs); \
203 	       } \
204 	    } \
205 } while (0)
206 
207 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
208 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
209 
210 #ifdef UPCALL_TIMING
211 #define UPCALL_MAX	50
212 u_long upcall_data[UPCALL_MAX + 1];
213 static void collate();
214 #endif /* UPCALL_TIMING */
215 
216 static int get_sg_cnt __P((struct sioc_sg_req6 *));
217 static int get_mif6_cnt __P((struct sioc_mif_req6 *));
218 static int ip6_mrouter_init __P((struct socket *, struct mbuf *, int));
219 static int add_m6if __P((struct mif6ctl *));
220 static int del_m6if __P((mifi_t *));
221 static int add_m6fc __P((struct mf6cctl *));
222 static int del_m6fc __P((struct mf6cctl *));
223 
224 /*
225  * Handle MRT setsockopt commands to modify the multicast routing tables.
226  */
227 int
228 ip6_mrouter_set(so, sopt)
229 	struct socket *so;
230 	struct sockopt *sopt;
231 {
232 	int	error = 0;
233 	struct mbuf *m;
234 
235 	if (so != ip6_mrouter && sopt->sopt_name != MRT6_INIT)
236 		return (EACCES);
237 
238 	if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
239 		return (error);
240 	if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
241 		return (error);
242 
243 	switch (sopt->sopt_name) {
244 	 case MRT6_INIT:
245 #ifdef MRT6_OINIT
246 	 case MRT6_OINIT:
247 #endif
248 		 error = ip6_mrouter_init(so, m, sopt->sopt_name);
249 		 break;
250 	 case MRT6_DONE:
251 		 error = ip6_mrouter_done();
252 		 break;
253 	 case MRT6_ADD_MIF:
254 		 error = add_m6if(mtod(m, struct mif6ctl *));
255 		 break;
256 	 case MRT6_DEL_MIF:
257 		 error = del_m6if(mtod(m, mifi_t *));
258 		 break;
259 	 case MRT6_ADD_MFC:
260 		 error = add_m6fc(mtod(m, struct mf6cctl *));
261 		 break;
262 	 case MRT6_DEL_MFC:
263 		 error = del_m6fc(mtod(m, struct mf6cctl *));
264 		 break;
265 	 case MRT6_PIM:
266 		 error = set_pim6(mtod(m, int *));
267 		 break;
268 	 default:
269 		 error = EOPNOTSUPP;
270 		 break;
271 	}
272 
273 	(void)m_freem(m);
274 	return(error);
275 }
276 
277 /*
278  * Handle MRT getsockopt commands
279  */
280 int
281 ip6_mrouter_get(so, sopt)
282 	struct socket *so;
283 	struct sockopt *sopt;
284 {
285 	int error = 0;
286 
287 	if (so != ip6_mrouter) return EACCES;
288 
289 	switch (sopt->sopt_name) {
290 		case MRT6_PIM:
291 			error = sooptcopyout(sopt, &pim6, sizeof(pim6));
292 			break;
293 	}
294 	return (error);
295 }
296 
297 /*
298  * Handle ioctl commands to obtain information from the cache
299  */
300 int
301 mrt6_ioctl(cmd, data)
302 	int cmd;
303 	caddr_t data;
304 {
305     int error = 0;
306 
307     switch (cmd) {
308      case SIOCGETSGCNT_IN6:
309 	     return(get_sg_cnt((struct sioc_sg_req6 *)data));
310 	     break;		/* for safety */
311      case SIOCGETMIFCNT_IN6:
312 	     return(get_mif6_cnt((struct sioc_mif_req6 *)data));
313 	     break;		/* for safety */
314      default:
315 	     return (EINVAL);
316 	     break;
317     }
318     return error;
319 }
320 
321 /*
322  * returns the packet, byte, rpf-failure count for the source group provided
323  */
324 static int
325 get_sg_cnt(req)
326 	register struct sioc_sg_req6 *req;
327 {
328 	register struct mf6c *rt;
329 	int s;
330 
331 	s = splnet();
332 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
333 	splx(s);
334 	if (rt != NULL) {
335 		req->pktcnt = rt->mf6c_pkt_cnt;
336 		req->bytecnt = rt->mf6c_byte_cnt;
337 		req->wrong_if = rt->mf6c_wrong_if;
338 	} else
339 		return(ESRCH);
340 #if 0
341 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
342 #endif
343 
344 	return 0;
345 }
346 
347 /*
348  * returns the input and output packet and byte counts on the mif provided
349  */
350 static int
351 get_mif6_cnt(req)
352 	register struct sioc_mif_req6 *req;
353 {
354 	register mifi_t mifi = req->mifi;
355 
356 	if (mifi >= nummifs)
357 		return EINVAL;
358 
359 	req->icount = mif6table[mifi].m6_pkt_in;
360 	req->ocount = mif6table[mifi].m6_pkt_out;
361 	req->ibytes = mif6table[mifi].m6_bytes_in;
362 	req->obytes = mif6table[mifi].m6_bytes_out;
363 
364 	return 0;
365 }
366 
367 static int
368 set_pim6(i)
369 	int *i;
370 {
371 	if ((*i != 1) && (*i != 0))
372 		return EINVAL;
373 
374 	pim6 = *i;
375 
376 	return 0;
377 }
378 
379 /*
380  * Enable multicast routing
381  */
382 static int
383 ip6_mrouter_init(so, m, cmd)
384 	struct socket *so;
385 	struct mbuf *m;
386 	int cmd;
387 {
388 	int *v;
389 
390 #ifdef MRT6DEBUG
391 	if (mrt6debug)
392 		log(LOG_DEBUG,
393 		    "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
394 		    so->so_type, so->so_proto->pr_protocol);
395 #endif
396 
397 	if (so->so_type != SOCK_RAW ||
398 	    so->so_proto->pr_protocol != IPPROTO_ICMPV6)
399 		return EOPNOTSUPP;
400 
401 	if (!m || (m->m_len != sizeof(int *)))
402 		return ENOPROTOOPT;
403 
404 	v = mtod(m, int *);
405 	if (*v != 1)
406 		return ENOPROTOOPT;
407 
408 	if (ip6_mrouter != NULL) return EADDRINUSE;
409 
410 	ip6_mrouter = so;
411 	ip6_mrouter_ver = cmd;
412 
413 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
414 	bzero((caddr_t)nexpire, sizeof(nexpire));
415 
416 	pim6 = 0;/* used for stubbing out/in pim stuff */
417 
418 	expire_upcalls_ch =
419 		timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
420 
421 #ifdef MRT6DEBUG
422 	if (mrt6debug)
423 		log(LOG_DEBUG, "ip6_mrouter_init\n");
424 #endif
425 
426 	return 0;
427 }
428 
429 /*
430  * Disable multicast routing
431  */
432 int
433 ip6_mrouter_done()
434 {
435 	mifi_t mifi;
436 	int i;
437 	struct ifnet *ifp;
438 	struct in6_ifreq ifr;
439 	struct mf6c *rt;
440 	struct rtdetq *rte;
441 	int s;
442 
443 	s = splnet();
444 
445 	/*
446 	 * For each phyint in use, disable promiscuous reception of all IPv6
447 	 * multicasts.
448 	 */
449 #ifdef INET
450 #ifdef MROUTING
451 	/*
452 	 * If there is still IPv4 multicast routing daemon,
453 	 * we remain interfaces to receive all muliticasted packets.
454 	 * XXX: there may be an interface in which the IPv4 multicast
455 	 * daemon is not interested...
456 	 */
457 	if (!ip_mrouter)
458 #endif
459 #endif
460 	{
461 		for (mifi = 0; mifi < nummifs; mifi++) {
462 			if (mif6table[mifi].m6_ifp &&
463 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
464 				ifr.ifr_addr.sin6_family = AF_INET6;
465 				ifr.ifr_addr.sin6_addr= in6addr_any;
466 				ifp = mif6table[mifi].m6_ifp;
467 				(*ifp->if_ioctl)(ifp, SIOCDELMULTI,
468 						 (caddr_t)&ifr);
469 			}
470 		}
471 	}
472 #ifdef notyet
473 	bzero((caddr_t)qtable, sizeof(qtable));
474 	bzero((caddr_t)tbftable, sizeof(tbftable));
475 #endif
476 	bzero((caddr_t)mif6table, sizeof(mif6table));
477 	nummifs = 0;
478 
479 	pim6 = 0; /* used to stub out/in pim specific code */
480 
481 	untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch);
482 
483 	/*
484 	 * Free all multicast forwarding cache entries.
485 	 */
486 	for (i = 0; i < MF6CTBLSIZ; i++) {
487 		rt = mf6ctable[i];
488 		while (rt) {
489 			struct mf6c *frt;
490 
491 			for (rte = rt->mf6c_stall; rte != NULL; ) {
492 				struct rtdetq *n = rte->next;
493 
494 				m_free(rte->m);
495 				free(rte, M_MRTABLE);
496 				rte = n;
497 			}
498 			frt = rt;
499 			rt = rt->mf6c_next;
500 			free(frt, M_MRTABLE);
501 		}
502 	}
503 
504 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
505 
506 	/*
507 	 * Reset de-encapsulation cache
508 	 */
509 	reg_mif_num = -1;
510 
511 	ip6_mrouter = NULL;
512 	ip6_mrouter_ver = 0;
513 
514 	splx(s);
515 
516 #ifdef MRT6DEBUG
517 	if (mrt6debug)
518 		log(LOG_DEBUG, "ip6_mrouter_done\n");
519 #endif
520 
521 	return 0;
522 }
523 
524 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
525 
526 /*
527  * Add a mif to the mif table
528  */
529 static int
530 add_m6if(mifcp)
531 	register struct mif6ctl *mifcp;
532 {
533 	register struct mif6 *mifp;
534 	struct ifnet *ifp;
535 	int error, s;
536 #ifdef notyet
537 	struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
538 #endif
539 
540 	if (mifcp->mif6c_mifi >= MAXMIFS)
541 		return EINVAL;
542 	mifp = mif6table + mifcp->mif6c_mifi;
543 	if (mifp->m6_ifp)
544 		return EADDRINUSE; /* XXX: is it appropriate? */
545 	if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
546 		return ENXIO;
547 	ifp = ifindex2ifnet[mifcp->mif6c_pifi];
548 
549 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
550 		if (reg_mif_num == (mifi_t)-1) {
551 			multicast_register_if.if_name = "register_mif";
552 			multicast_register_if.if_flags |= IFF_LOOPBACK;
553 			multicast_register_if.if_index = mifcp->mif6c_mifi;
554 			reg_mif_num = mifcp->mif6c_mifi;
555 		}
556 
557 		ifp = &multicast_register_if;
558 
559 	} /* if REGISTER */
560 	else {
561 		/* Make sure the interface supports multicast */
562 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
563 			return EOPNOTSUPP;
564 
565 		s = splnet();
566 		error = if_allmulti(ifp, 1);
567 		splx(s);
568 		if (error)
569 			return error;
570 	}
571 
572 	s = splnet();
573 	mifp->m6_flags     = mifcp->mif6c_flags;
574 	mifp->m6_ifp       = ifp;
575 #ifdef notyet
576 	/* scaling up here allows division by 1024 in critical code */
577 	mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
578 #endif
579 	/* initialize per mif pkt counters */
580 	mifp->m6_pkt_in    = 0;
581 	mifp->m6_pkt_out   = 0;
582 	mifp->m6_bytes_in  = 0;
583 	mifp->m6_bytes_out = 0;
584 	splx(s);
585 
586 	/* Adjust nummifs up if the mifi is higher than nummifs */
587 	if (nummifs <= mifcp->mif6c_mifi)
588 		nummifs = mifcp->mif6c_mifi + 1;
589 
590 #ifdef MRT6DEBUG
591 	if (mrt6debug)
592 		log(LOG_DEBUG,
593 		    "add_mif #%d, phyint %s%d\n",
594 		    mifcp->mif6c_mifi,
595 		    ifp->if_name, ifp->if_unit);
596 #endif
597 
598 	return 0;
599 }
600 
601 /*
602  * Delete a mif from the mif table
603  */
604 static int
605 del_m6if(mifip)
606 	mifi_t *mifip;
607 {
608 	register struct mif6 *mifp = mif6table + *mifip;
609 	register mifi_t mifi;
610 	struct ifnet *ifp;
611 	int s;
612 
613 	if (*mifip >= nummifs)
614 		return EINVAL;
615 	if (mifp->m6_ifp == NULL)
616 		return EINVAL;
617 
618 	s = splnet();
619 
620 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
621 		/*
622 		 * XXX: what if there is yet IPv4 multicast daemon
623 		 *      using the interface?
624 		 */
625 		ifp = mifp->m6_ifp;
626 
627 		if_allmulti(ifp, 0);
628 	}
629 
630 #ifdef notyet
631 	bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
632 	bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
633 #endif
634 	bzero((caddr_t)mifp, sizeof (*mifp));
635 
636 	/* Adjust nummifs down */
637 	for (mifi = nummifs; mifi > 0; mifi--)
638 		if (mif6table[mifi - 1].m6_ifp)
639 			break;
640 	nummifs = mifi;
641 
642 	splx(s);
643 
644 #ifdef MRT6DEBUG
645 	if (mrt6debug)
646 		log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
647 #endif
648 
649 	return 0;
650 }
651 
652 /*
653  * Add an mfc entry
654  */
655 static int
656 add_m6fc(mfccp)
657 	struct mf6cctl *mfccp;
658 {
659 	struct mf6c *rt;
660 	u_long hash;
661 	struct rtdetq *rte;
662 	register u_short nstl;
663 	int s;
664 
665 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
666 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
667 
668 	/* If an entry already exists, just update the fields */
669 	if (rt) {
670 #ifdef MRT6DEBUG
671 		if (mrt6debug & DEBUG_MFC)
672 			log(LOG_DEBUG,"add_m6fc update o %s g %s p %x\n",
673 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
674 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
675 			    mfccp->mf6cc_parent);
676 #endif
677 
678 		s = splnet();
679 		rt->mf6c_parent = mfccp->mf6cc_parent;
680 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
681 		splx(s);
682 		return 0;
683 	}
684 
685 	/*
686 	 * Find the entry for which the upcall was made and update
687 	 */
688 	s = splnet();
689 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
690 			mfccp->mf6cc_mcastgrp.sin6_addr);
691 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
692 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
693 				       &mfccp->mf6cc_origin.sin6_addr) &&
694 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
695 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
696 		    (rt->mf6c_stall != NULL)) {
697 
698 			if (nstl++)
699 				log(LOG_ERR,
700 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
701 				    "multiple kernel entries",
702 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
703 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
704 				    mfccp->mf6cc_parent, rt->mf6c_stall);
705 
706 #ifdef MRT6DEBUG
707 			if (mrt6debug & DEBUG_MFC)
708 				log(LOG_DEBUG,
709 				    "add_m6fc o %s g %s p %x dbg %x\n",
710 				    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
711 				    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
712 				    mfccp->mf6cc_parent, rt->mf6c_stall);
713 #endif
714 
715 			rt->mf6c_origin     = mfccp->mf6cc_origin;
716 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
717 			rt->mf6c_parent     = mfccp->mf6cc_parent;
718 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
719 			/* initialize pkt counters per src-grp */
720 			rt->mf6c_pkt_cnt    = 0;
721 			rt->mf6c_byte_cnt   = 0;
722 			rt->mf6c_wrong_if   = 0;
723 
724 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
725 			nexpire[hash]--;
726 
727 			/* free packets Qed at the end of this entry */
728 			for (rte = rt->mf6c_stall; rte != NULL; ) {
729 				struct rtdetq *n = rte->next;
730 				ip6_mdq(rte->m, rte->ifp, rt);
731 				m_freem(rte->m);
732 #ifdef UPCALL_TIMING
733 				collate(&(rte->t));
734 #endif /* UPCALL_TIMING */
735 				free(rte, M_MRTABLE);
736 				rte = n;
737 			}
738 			rt->mf6c_stall = NULL;
739 		}
740 	}
741 
742 	/*
743 	 * It is possible that an entry is being inserted without an upcall
744 	 */
745 	if (nstl == 0) {
746 #ifdef MRT6DEBUG
747 		if (mrt6debug & DEBUG_MFC)
748 			log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n",
749 			    hash,
750 			    ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
751 			    ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
752 			    mfccp->mf6cc_parent);
753 #endif
754 
755 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
756 
757 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
758 					       &mfccp->mf6cc_origin.sin6_addr)&&
759 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
760 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
761 
762 				rt->mf6c_origin     = mfccp->mf6cc_origin;
763 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
764 				rt->mf6c_parent     = mfccp->mf6cc_parent;
765 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
766 				/* initialize pkt counters per src-grp */
767 				rt->mf6c_pkt_cnt    = 0;
768 				rt->mf6c_byte_cnt   = 0;
769 				rt->mf6c_wrong_if   = 0;
770 
771 				if (rt->mf6c_expire)
772 					nexpire[hash]--;
773 				rt->mf6c_expire	   = 0;
774 			}
775 		}
776 		if (rt == NULL) {
777 			/* no upcall, so make a new entry */
778 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
779 						  M_NOWAIT);
780 			if (rt == NULL) {
781 				splx(s);
782 				return ENOBUFS;
783 			}
784 
785 			/* insert new entry at head of hash chain */
786 			rt->mf6c_origin     = mfccp->mf6cc_origin;
787 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
788 			rt->mf6c_parent     = mfccp->mf6cc_parent;
789 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
790 			/* initialize pkt counters per src-grp */
791 			rt->mf6c_pkt_cnt    = 0;
792 			rt->mf6c_byte_cnt   = 0;
793 			rt->mf6c_wrong_if   = 0;
794 			rt->mf6c_expire     = 0;
795 			rt->mf6c_stall = NULL;
796 
797 			/* link into table */
798 			rt->mf6c_next  = mf6ctable[hash];
799 			mf6ctable[hash] = rt;
800 		}
801 	}
802 	splx(s);
803 	return 0;
804 }
805 
806 #ifdef UPCALL_TIMING
807 /*
808  * collect delay statistics on the upcalls
809  */
810 static void
811 collate(t)
812 	register struct timeval *t;
813 {
814 	register u_long d;
815 	register struct timeval tp;
816 	register u_long delta;
817 
818 	GET_TIME(tp);
819 
820 	if (TV_LT(*t, tp))
821 	{
822 		TV_DELTA(tp, *t, delta);
823 
824 		d = delta >> 10;
825 		if (d > UPCALL_MAX)
826 			d = UPCALL_MAX;
827 
828 		++upcall_data[d];
829 	}
830 }
831 #endif /* UPCALL_TIMING */
832 
833 /*
834  * Delete an mfc entry
835  */
836 static int
837 del_m6fc(mfccp)
838 	struct mf6cctl *mfccp;
839 {
840 	struct sockaddr_in6 	origin;
841 	struct sockaddr_in6 	mcastgrp;
842 	struct mf6c 		*rt;
843 	struct mf6c	 	**nptr;
844 	u_long 		hash;
845 	int s;
846 
847 	origin = mfccp->mf6cc_origin;
848 	mcastgrp = mfccp->mf6cc_mcastgrp;
849 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
850 
851 #ifdef MRT6DEBUG
852 	if (mrt6debug & DEBUG_MFC)
853 		log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
854 		    ip6_sprintf(&origin.sin6_addr),
855 		    ip6_sprintf(&mcastgrp.sin6_addr));
856 #endif
857 
858 	s = splnet();
859 
860 	nptr = &mf6ctable[hash];
861 	while ((rt = *nptr) != NULL) {
862 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
863 				       &rt->mf6c_origin.sin6_addr) &&
864 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
865 				       &rt->mf6c_mcastgrp.sin6_addr) &&
866 		    rt->mf6c_stall == NULL)
867 			break;
868 
869 		nptr = &rt->mf6c_next;
870 	}
871 	if (rt == NULL) {
872 		splx(s);
873 		return EADDRNOTAVAIL;
874 	}
875 
876 	*nptr = rt->mf6c_next;
877 	free(rt, M_MRTABLE);
878 
879 	splx(s);
880 
881 	return 0;
882 }
883 
884 static int
885 socket_send(s, mm, src)
886 	struct socket *s;
887 	struct mbuf *mm;
888 	struct sockaddr_in6 *src;
889 {
890 	if (s) {
891 		if (sbappendaddr(&s->so_rcv,
892 				 (struct sockaddr *)src,
893 				 mm, (struct mbuf *)0) != 0) {
894 			sorwakeup(s);
895 			return 0;
896 		}
897 	}
898 	m_freem(mm);
899 	return -1;
900 }
901 
902 /*
903  * IPv6 multicast forwarding function. This function assumes that the packet
904  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
905  * pointed to by "ifp", and the packet is to be relayed to other networks
906  * that have members of the packet's destination IPv6 multicast group.
907  *
908  * The packet is returned unscathed to the caller, unless it is
909  * erroneous, in which case a non-zero return value tells the caller to
910  * discard it.
911  */
912 
913 int
914 ip6_mforward(ip6, ifp, m)
915 	register struct ip6_hdr *ip6;
916 	struct ifnet *ifp;
917 	struct mbuf *m;
918 {
919 	register struct mf6c *rt;
920 	register struct mif6 *mifp;
921 	register struct mbuf *mm;
922 	int s;
923 	mifi_t mifi;
924 
925 #ifdef MRT6DEBUG
926 	if (mrt6debug & DEBUG_FORWARD)
927 		log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
928 		    ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
929 		    ifp->if_index);
930 #endif
931 
932 	/*
933 	 * Don't forward a packet with Hop limit of zero or one,
934 	 * or a packet destined to a local-only group.
935 	 */
936 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
937 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
938 		return 0;
939 	ip6->ip6_hlim--;
940 
941 	/*
942 	 * Source address check: do not forward packets with unspecified
943 	 * source. It was discussed in July 2000, on ipngwg mailing list.
944 	 * This is rather more serious than unicast cases, because some
945 	 * MLD packets can be sent with the unspecified source address
946 	 * (although such packets must normally set 1 to the hop limit field).
947 	 */
948 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
949 		ip6stat.ip6s_cantforward++;
950 		if (ip6_log_time + ip6_log_interval < time_second) {
951 			ip6_log_time = time_second;
952 			log(LOG_DEBUG,
953 			    "cannot forward "
954 			    "from %s to %s nxt %d received on %s\n",
955 			    ip6_sprintf(&ip6->ip6_src),
956 			    ip6_sprintf(&ip6->ip6_dst),
957 			    ip6->ip6_nxt,
958 			    if_name(m->m_pkthdr.rcvif));
959 		}
960 		return 0;
961 	}
962 
963 	/*
964 	 * Determine forwarding mifs from the forwarding cache table
965 	 */
966 	s = splnet();
967 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
968 
969 	/* Entry exists, so forward if necessary */
970 	if (rt) {
971 		splx(s);
972 		return (ip6_mdq(m, ifp, rt));
973 	} else {
974 		/*
975 		 * If we don't have a route for packet's origin,
976 		 * Make a copy of the packet &
977 		 * send message to routing daemon
978 		 */
979 
980 		register struct mbuf *mb0;
981 		register struct rtdetq *rte;
982 		register u_long hash;
983 /*	register int i, npkts;*/
984 #ifdef UPCALL_TIMING
985 		struct timeval tp;
986 
987 		GET_TIME(tp);
988 #endif /* UPCALL_TIMING */
989 
990 		mrt6stat.mrt6s_no_route++;
991 #ifdef MRT6DEBUG
992 		if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
993 			log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
994 			    ip6_sprintf(&ip6->ip6_src),
995 			    ip6_sprintf(&ip6->ip6_dst));
996 #endif
997 
998 		/*
999 		 * Allocate mbufs early so that we don't do extra work if we
1000 		 * are just going to fail anyway.
1001 		 */
1002 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1003 					      M_NOWAIT);
1004 		if (rte == NULL) {
1005 			splx(s);
1006 			return ENOBUFS;
1007 		}
1008 		mb0 = m_copy(m, 0, M_COPYALL);
1009 		/*
1010 		 * Pullup packet header if needed before storing it,
1011 		 * as other references may modify it in the meantime.
1012 		 */
1013 		if (mb0 &&
1014 		    (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1015 			mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1016 		if (mb0 == NULL) {
1017 			free(rte, M_MRTABLE);
1018 			splx(s);
1019 			return ENOBUFS;
1020 		}
1021 
1022 		/* is there an upcall waiting for this packet? */
1023 		hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1024 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1025 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1026 					       &rt->mf6c_origin.sin6_addr) &&
1027 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1028 					       &rt->mf6c_mcastgrp.sin6_addr) &&
1029 			    (rt->mf6c_stall != NULL))
1030 				break;
1031 		}
1032 
1033 		if (rt == NULL) {
1034 			struct mrt6msg *im;
1035 #ifdef MRT6_OINIT
1036 			struct omrt6msg *oim;
1037 #endif
1038 
1039 			/* no upcall, so make a new entry */
1040 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1041 						  M_NOWAIT);
1042 			if (rt == NULL) {
1043 				free(rte, M_MRTABLE);
1044 				m_freem(mb0);
1045 				splx(s);
1046 				return ENOBUFS;
1047 			}
1048 			/*
1049 			 * Make a copy of the header to send to the user
1050 			 * level process
1051 			 */
1052 			mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1053 
1054 			if (mm == NULL) {
1055 				free(rte, M_MRTABLE);
1056 				m_freem(mb0);
1057 				free(rt, M_MRTABLE);
1058 				splx(s);
1059 				return ENOBUFS;
1060 			}
1061 
1062 			/*
1063 			 * Send message to routing daemon
1064 			 */
1065 			sin6.sin6_addr = ip6->ip6_src;
1066 
1067 			im = NULL;
1068 #ifdef MRT6_OINIT
1069 			oim = NULL;
1070 #endif
1071 			switch (ip6_mrouter_ver) {
1072 #ifdef MRT6_OINIT
1073 			case MRT6_OINIT:
1074 				oim = mtod(mm, struct omrt6msg *);
1075 				oim->im6_msgtype = MRT6MSG_NOCACHE;
1076 				oim->im6_mbz = 0;
1077 				break;
1078 #endif
1079 			case MRT6_INIT:
1080 				im = mtod(mm, struct mrt6msg *);
1081 				im->im6_msgtype = MRT6MSG_NOCACHE;
1082 				im->im6_mbz = 0;
1083 				break;
1084 			default:
1085 				free(rte, M_MRTABLE);
1086 				m_freem(mb0);
1087 				free(rt, M_MRTABLE);
1088 				splx(s);
1089 				return EINVAL;
1090 			}
1091 
1092 #ifdef MRT6DEBUG
1093 			if (mrt6debug & DEBUG_FORWARD)
1094 				log(LOG_DEBUG,
1095 				    "getting the iif info in the kernel\n");
1096 #endif
1097 
1098 			for (mifp = mif6table, mifi = 0;
1099 			     mifi < nummifs && mifp->m6_ifp != ifp;
1100 			     mifp++, mifi++)
1101 				;
1102 
1103 			switch (ip6_mrouter_ver) {
1104 #ifdef MRT6_OINIT
1105 			case MRT6_OINIT:
1106 				oim->im6_mif = mifi;
1107 				break;
1108 #endif
1109 			case MRT6_INIT:
1110 				im->im6_mif = mifi;
1111 				break;
1112 			}
1113 
1114 			if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1115 				log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1116 				    "socket queue full\n");
1117 				mrt6stat.mrt6s_upq_sockfull++;
1118 				free(rte, M_MRTABLE);
1119 				m_freem(mb0);
1120 				free(rt, M_MRTABLE);
1121 				splx(s);
1122 				return ENOBUFS;
1123 			}
1124 
1125 			mrt6stat.mrt6s_upcalls++;
1126 
1127 			/* insert new entry at head of hash chain */
1128 			bzero(rt, sizeof(*rt));
1129 			rt->mf6c_origin.sin6_family = AF_INET6;
1130 			rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1131 			rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1132 			rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1133 			rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1134 			rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1135 			rt->mf6c_expire = UPCALL_EXPIRE;
1136 			nexpire[hash]++;
1137 			rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1138 
1139 			/* link into table */
1140 			rt->mf6c_next  = mf6ctable[hash];
1141 			mf6ctable[hash] = rt;
1142 			/* Add this entry to the end of the queue */
1143 			rt->mf6c_stall = rte;
1144 		} else {
1145 			/* determine if q has overflowed */
1146 			struct rtdetq **p;
1147 			register int npkts = 0;
1148 
1149 			for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1150 				if (++npkts > MAX_UPQ6) {
1151 					mrt6stat.mrt6s_upq_ovflw++;
1152 					free(rte, M_MRTABLE);
1153 					m_freem(mb0);
1154 					splx(s);
1155 					return 0;
1156 				}
1157 
1158 			/* Add this entry to the end of the queue */
1159 			*p = rte;
1160 		}
1161 
1162 		rte->next = NULL;
1163 		rte->m = mb0;
1164 		rte->ifp = ifp;
1165 #ifdef UPCALL_TIMING
1166 		rte->t = tp;
1167 #endif /* UPCALL_TIMING */
1168 
1169 		splx(s);
1170 
1171 		return 0;
1172 	}
1173 }
1174 
1175 /*
1176  * Clean up cache entries if upcalls are not serviced
1177  * Call from the Slow Timeout mechanism, every half second.
1178  */
1179 static void
1180 expire_upcalls(unused)
1181 	void *unused;
1182 {
1183 	struct rtdetq *rte;
1184 	struct mf6c *mfc, **nptr;
1185 	int i;
1186 	int s;
1187 
1188 	s = splnet();
1189 	for (i = 0; i < MF6CTBLSIZ; i++) {
1190 		if (nexpire[i] == 0)
1191 			continue;
1192 		nptr = &mf6ctable[i];
1193 		while ((mfc = *nptr) != NULL) {
1194 			rte = mfc->mf6c_stall;
1195 			/*
1196 			 * Skip real cache entries
1197 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1198 			 * If it expires now
1199 			 */
1200 			if (rte != NULL &&
1201 			    mfc->mf6c_expire != 0 &&
1202 			    --mfc->mf6c_expire == 0) {
1203 #ifdef MRT6DEBUG
1204 				if (mrt6debug & DEBUG_EXPIRE)
1205 					log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1206 					    ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1207 					    ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1208 #endif
1209 				/*
1210 				 * drop all the packets
1211 				 * free the mbuf with the pkt, if, timing info
1212 				 */
1213 				do {
1214 					struct rtdetq *n = rte->next;
1215 					m_freem(rte->m);
1216 					free(rte, M_MRTABLE);
1217 					rte = n;
1218 				} while (rte != NULL);
1219 				mrt6stat.mrt6s_cache_cleanups++;
1220 				nexpire[i]--;
1221 
1222 				*nptr = mfc->mf6c_next;
1223 				free(mfc, M_MRTABLE);
1224 			} else {
1225 				nptr = &mfc->mf6c_next;
1226 			}
1227 		}
1228 	}
1229 	splx(s);
1230 	expire_upcalls_ch =
1231 	timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
1232 }
1233 
1234 /*
1235  * Packet forwarding routine once entry in the cache is made
1236  */
1237 static int
1238 ip6_mdq(m, ifp, rt)
1239 	register struct mbuf *m;
1240 	register struct ifnet *ifp;
1241 	register struct mf6c *rt;
1242 {
1243 	register struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1244 	register mifi_t mifi, iif;
1245 	register struct mif6 *mifp;
1246 	register int plen = m->m_pkthdr.len;
1247 
1248 /*
1249  * Macro to send packet on mif.  Since RSVP packets don't get counted on
1250  * input, they shouldn't get counted on output, so statistics keeping is
1251  * seperate.
1252  */
1253 
1254 #define MC6_SEND(ip6, mifp, m) do {				\
1255 		if ((mifp)->m6_flags & MIFF_REGISTER)		\
1256 		    register_send((ip6), (mifp), (m));		\
1257 		else						\
1258 		    phyint_send((ip6), (mifp), (m));		\
1259 } while (0)
1260 
1261 	/*
1262 	 * Don't forward if it didn't arrive from the parent mif
1263 	 * for its origin.
1264 	 */
1265 	mifi = rt->mf6c_parent;
1266 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1267 		/* came in the wrong interface */
1268 #ifdef MRT6DEBUG
1269 		if (mrt6debug & DEBUG_FORWARD)
1270 			log(LOG_DEBUG,
1271 			    "wrong if: ifid %d mifi %d mififid %x\n",
1272 			    ifp->if_index, mifi,
1273 			    mif6table[mifi].m6_ifp->if_index);
1274 #endif
1275 		mrt6stat.mrt6s_wrong_if++;
1276 		rt->mf6c_wrong_if++;
1277 		/*
1278 		 * If we are doing PIM processing, and we are forwarding
1279 		 * packets on this interface, send a message to the
1280 		 * routing daemon.
1281 		 */
1282 		/* have to make sure this is a valid mif */
1283 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1284 			if (pim6 && (m->m_flags & M_LOOP) == 0) {
1285 				/*
1286 				 * Check the M_LOOP flag to avoid an
1287 				 * unnecessary PIM assert.
1288 				 * XXX: M_LOOP is an ad-hoc hack...
1289 				 */
1290 				static struct sockaddr_in6 sin6 =
1291 				{ sizeof(sin6), AF_INET6 };
1292 
1293 				register struct mbuf *mm;
1294 				struct mrt6msg *im;
1295 #ifdef MRT6_OINIT
1296 				struct omrt6msg *oim;
1297 #endif
1298 
1299 				mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1300 				if (mm &&
1301 				    (M_HASCL(mm) ||
1302 				     mm->m_len < sizeof(struct ip6_hdr)))
1303 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1304 				if (mm == NULL)
1305 					return ENOBUFS;
1306 
1307 #ifdef MRT6_OINIT
1308 				oim = NULL;
1309 #endif
1310 				im = NULL;
1311 				switch (ip6_mrouter_ver) {
1312 #ifdef MRT6_OINIT
1313 				case MRT6_OINIT:
1314 					oim = mtod(mm, struct omrt6msg *);
1315 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1316 					oim->im6_mbz = 0;
1317 					break;
1318 #endif
1319 				case MRT6_INIT:
1320 					im = mtod(mm, struct mrt6msg *);
1321 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1322 					break;
1323 				default:
1324 					m_freem(mm);
1325 					return EINVAL;
1326 				}
1327 
1328 				for (mifp = mif6table, iif = 0;
1329 				     iif < nummifs && mifp &&
1330 					     mifp->m6_ifp != ifp;
1331 				     mifp++, iif++)
1332 					;
1333 
1334 				switch (ip6_mrouter_ver) {
1335 #ifdef MRT6_OINIT
1336 				case MRT6_OINIT:
1337 					oim->im6_mif = iif;
1338 					sin6.sin6_addr = oim->im6_src;
1339 					break;
1340 #endif
1341 				case MRT6_INIT:
1342 					im->im6_mif = iif;
1343 					sin6.sin6_addr = im->im6_src;
1344 					break;
1345 				}
1346 
1347 				mrt6stat.mrt6s_upcalls++;
1348 
1349 				if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1350 #ifdef MRT6DEBUG
1351 					if (mrt6debug)
1352 						log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1353 #endif
1354 					++mrt6stat.mrt6s_upq_sockfull;
1355 					return ENOBUFS;
1356 				}	/* if socket Q full */
1357 			}		/* if PIM */
1358 		return 0;
1359 	}			/* if wrong iif */
1360 
1361 	/* If I sourced this packet, it counts as output, else it was input. */
1362 	if (m->m_pkthdr.rcvif == NULL) {
1363 		/* XXX: is rcvif really NULL when output?? */
1364 		mif6table[mifi].m6_pkt_out++;
1365 		mif6table[mifi].m6_bytes_out += plen;
1366 	} else {
1367 		mif6table[mifi].m6_pkt_in++;
1368 		mif6table[mifi].m6_bytes_in += plen;
1369 	}
1370 	rt->mf6c_pkt_cnt++;
1371 	rt->mf6c_byte_cnt += plen;
1372 
1373 	/*
1374 	 * For each mif, forward a copy of the packet if there are group
1375 	 * members downstream on the interface.
1376 	 */
1377 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1378 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1379 			/*
1380 			 * check if the outgoing packet is going to break
1381 			 * a scope boundary.
1382 			 * XXX For packets through PIM register tunnel
1383 			 * interface, we believe a routing daemon.
1384 			 */
1385 			if ((mif6table[rt->mf6c_parent].m6_flags &
1386 			     MIFF_REGISTER) == 0 &&
1387 			    (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1388 			    (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1389 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1390 					      &ip6->ip6_dst) ||
1391 			     in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1392 			     in6_addr2scopeid(mif6table[mifi].m6_ifp,
1393 					      &ip6->ip6_src))) {
1394 				ip6stat.ip6s_badscope++;
1395 				continue;
1396 			}
1397 
1398 			mifp->m6_pkt_out++;
1399 			mifp->m6_bytes_out += plen;
1400 			MC6_SEND(ip6, mifp, m);
1401 		}
1402 	return 0;
1403 }
1404 
1405 static void
1406 phyint_send(ip6, mifp, m)
1407     struct ip6_hdr *ip6;
1408     struct mif6 *mifp;
1409     struct mbuf *m;
1410 {
1411 	register struct mbuf *mb_copy;
1412 	struct ifnet *ifp = mifp->m6_ifp;
1413 	int error = 0;
1414 	int s = splnet();
1415 	static struct route_in6 ro6;
1416 	struct	in6_multi *in6m;
1417 
1418 	/*
1419 	 * Make a new reference to the packet; make sure that
1420 	 * the IPv6 header is actually copied, not just referenced,
1421 	 * so that ip6_output() only scribbles on the copy.
1422 	 */
1423 	mb_copy = m_copy(m, 0, M_COPYALL);
1424 	if (mb_copy &&
1425 	    (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1426 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1427 	if (mb_copy == NULL)
1428 		return;
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, &ro6,
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 	IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1463 	if (in6m != NULL) {
1464 		ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
1465 		ro6.ro_dst.sin6_family = AF_INET6;
1466 		ro6.ro_dst.sin6_addr = ip6->ip6_dst;
1467 		ip6_mloopback(ifp, m, &ro6.ro_dst);
1468 	}
1469 	/*
1470 	 * Put the packet into the sending queue of the outgoing interface
1471 	 * if it would fit in the MTU of the interface.
1472 	 */
1473 	if (mb_copy->m_pkthdr.len < ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
1474 		ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
1475 		ro6.ro_dst.sin6_family = AF_INET6;
1476 		ro6.ro_dst.sin6_addr = ip6->ip6_dst;
1477 		/*
1478 		 * We just call if_output instead of nd6_output here, since
1479 		 * we need no ND for a multicast forwarded packet...right?
1480 		 */
1481 		error = (*ifp->if_output)(ifp, mb_copy,
1482 					  (struct sockaddr *)&ro6.ro_dst,
1483 					  NULL);
1484 #ifdef MRT6DEBUG
1485 		if (mrt6debug & DEBUG_XMIT)
1486 			log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1487 			    mifp - mif6table, error);
1488 #endif
1489 	}
1490 	else {
1491 #ifdef MULTICAST_PMTUD
1492 		icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
1493 		return;
1494 #else
1495 #ifdef MRT6DEBUG
1496 		if (mrt6debug & DEBUG_XMIT)
1497 			log(LOG_DEBUG,
1498 			    "phyint_send: packet too big on %s%u o %s g %s"
1499 			    " size %d(discarded)\n",
1500 			    ifp->if_name, ifp->if_unit,
1501 			    ip6_sprintf(&ip6->ip6_src),
1502 			    ip6_sprintf(&ip6->ip6_dst),
1503 			    mb_copy->m_pkthdr.len);
1504 #endif /* MRT6DEBUG */
1505 		m_freem(mb_copy); /* simply discard the packet */
1506 		return;
1507 #endif
1508 	}
1509 }
1510 
1511 static int
1512 register_send(ip6, mif, m)
1513 	register struct ip6_hdr *ip6;
1514 	struct mif6 *mif;
1515 	register struct mbuf *m;
1516 {
1517 	register struct mbuf *mm;
1518 	register 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 		m_freem(mm);
1546 		return ENOBUFS;
1547 	}
1548 /* TODO: check it! */
1549 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1550 
1551 	/*
1552 	 * Send message to routing daemon
1553 	 */
1554 	sin6.sin6_addr = ip6->ip6_src;
1555 
1556 	im6 = mtod(mm, struct mrt6msg *);
1557 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1558 	im6->im6_mbz          = 0;
1559 
1560 	im6->im6_mif = mif - mif6table;
1561 
1562 	/* iif info is not given for reg. encap.n */
1563 	mrt6stat.mrt6s_upcalls++;
1564 
1565 	if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1566 #ifdef MRT6DEBUG
1567 		if (mrt6debug)
1568 			log(LOG_WARNING,
1569 			    "register_send: ip_mrouter socket queue full\n");
1570 #endif
1571                 ++mrt6stat.mrt6s_upq_sockfull;
1572                 return ENOBUFS;
1573 	}
1574 	return 0;
1575 }
1576 
1577 /*
1578  * PIM sparse mode hook
1579  * Receives the pim control messages, and passes them up to the listening
1580  * socket, using rip6_input.
1581  * The only message processed is the REGISTER pim message; the pim header
1582  * is stripped off, and the inner packet is passed to register_mforward.
1583  */
1584 int
1585 pim6_input(mp, offp, proto)
1586 	struct mbuf **mp;
1587 	int *offp, proto;
1588 {
1589         register struct pim *pim; /* pointer to a pim struct */
1590         register struct ip6_hdr *ip6;
1591         register int pimlen;
1592 	struct mbuf *m = *mp;
1593         int minlen;
1594 	int off = *offp;
1595 
1596 	++pim6stat.pim6s_rcv_total;
1597 
1598         ip6 = mtod(m, struct ip6_hdr *);
1599         pimlen = m->m_pkthdr.len - *offp;
1600 
1601         /*
1602          * Validate lengths
1603          */
1604 	if (pimlen < PIM_MINLEN) {
1605 		++pim6stat.pim6s_rcv_tooshort;
1606 #ifdef MRT6DEBUG
1607 		if (mrt6debug & DEBUG_PIM)
1608 			log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1609 #endif
1610 		m_freem(m);
1611 		return(IPPROTO_DONE);
1612 	}
1613 
1614 	/*
1615 	 * if the packet is at least as big as a REGISTER, go ahead
1616 	 * and grab the PIM REGISTER header size, to avoid another
1617 	 * possible m_pullup() later.
1618 	 *
1619 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1620 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1621 	 */
1622 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1623 
1624 	/*
1625 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1626 	 * possibly the PIM REGISTER header
1627 	 */
1628 #ifndef PULLDOWN_TEST
1629 	IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1630 	/* adjust pointer */
1631 	ip6 = mtod(m, struct ip6_hdr *);
1632 
1633 	/* adjust mbuf to point to the PIM header */
1634 	pim = (struct pim *)((caddr_t)ip6 + off);
1635 #else
1636 	IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1637 	if (pim == NULL) {
1638 		pim6stat.pim6s_rcv_tooshort++;
1639 		return IPPROTO_DONE;
1640 	}
1641 #endif
1642 
1643 #define PIM6_CHECKSUM
1644 #ifdef PIM6_CHECKSUM
1645 	{
1646 		int cksumlen;
1647 
1648 		/*
1649 		 * Validate checksum.
1650 		 * If PIM REGISTER, exclude the data packet
1651 		 */
1652 		if (pim->pim_type == PIM_REGISTER)
1653 			cksumlen = PIM_MINLEN;
1654 		else
1655 			cksumlen = pimlen;
1656 
1657 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1658 			++pim6stat.pim6s_rcv_badsum;
1659 #ifdef MRT6DEBUG
1660 			if (mrt6debug & DEBUG_PIM)
1661 				log(LOG_DEBUG,
1662 				    "pim6_input: invalid checksum\n");
1663 #endif
1664 			m_freem(m);
1665 			return(IPPROTO_DONE);
1666 		}
1667 	}
1668 #endif /* PIM_CHECKSUM */
1669 
1670 	/* PIM version check */
1671 	if (pim->pim_ver != PIM_VERSION) {
1672 		++pim6stat.pim6s_rcv_badversion;
1673 #ifdef MRT6DEBUG
1674 		log(LOG_ERR,
1675 		    "pim6_input: incorrect version %d, expecting %d\n",
1676 		    pim->pim_ver, PIM_VERSION);
1677 #endif
1678 		m_freem(m);
1679 		return(IPPROTO_DONE);
1680 	}
1681 
1682 	if (pim->pim_type == PIM_REGISTER) {
1683 		/*
1684 		 * since this is a REGISTER, we'll make a copy of the register
1685 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1686 		 * routing daemon.
1687 		 */
1688 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1689 
1690 		struct mbuf *mcp;
1691 		struct ip6_hdr *eip6;
1692 		u_int32_t *reghdr;
1693 		int rc;
1694 
1695 		++pim6stat.pim6s_rcv_registers;
1696 
1697 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1698 #ifdef MRT6DEBUG
1699 			if (mrt6debug & DEBUG_PIM)
1700 				log(LOG_DEBUG,
1701 				    "pim6_input: register mif not set: %d\n",
1702 				    reg_mif_num);
1703 #endif
1704 			m_freem(m);
1705 			return(IPPROTO_DONE);
1706 		}
1707 
1708 		reghdr = (u_int32_t *)(pim + 1);
1709 
1710 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1711 			goto pim6_input_to_daemon;
1712 
1713 		/*
1714 		 * Validate length
1715 		 */
1716 		if (pimlen < PIM6_REG_MINLEN) {
1717 			++pim6stat.pim6s_rcv_tooshort;
1718 			++pim6stat.pim6s_rcv_badregisters;
1719 #ifdef MRT6DEBUG
1720 			log(LOG_ERR,
1721 			    "pim6_input: register packet size too "
1722 			    "small %d from %s\n",
1723 			    pimlen, ip6_sprintf(&ip6->ip6_src));
1724 #endif
1725 			m_freem(m);
1726 			return(IPPROTO_DONE);
1727 		}
1728 
1729 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1730 #ifdef MRT6DEBUG
1731 		if (mrt6debug & DEBUG_PIM)
1732 			log(LOG_DEBUG,
1733 			    "pim6_input[register], eip6: %s -> %s, "
1734 			    "eip6 plen %d\n",
1735 			    ip6_sprintf(&eip6->ip6_src),
1736 			    ip6_sprintf(&eip6->ip6_dst),
1737 			    ntohs(eip6->ip6_plen));
1738 #endif
1739 
1740 		/* verify the inner packet is destined to a mcast group */
1741 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1742 			++pim6stat.pim6s_rcv_badregisters;
1743 #ifdef MRT6DEBUG
1744 			if (mrt6debug & DEBUG_PIM)
1745 				log(LOG_DEBUG,
1746 				    "pim6_input: inner packet of register "
1747 				    "is not multicast %s\n",
1748 				    ip6_sprintf(&eip6->ip6_dst));
1749 #endif
1750 			m_freem(m);
1751 			return(IPPROTO_DONE);
1752 		}
1753 
1754 		/*
1755 		 * make a copy of the whole header to pass to the daemon later.
1756 		 */
1757 		mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1758 		if (mcp == NULL) {
1759 #ifdef MRT6DEBUG
1760 			log(LOG_ERR,
1761 			    "pim6_input: pim register: "
1762 			    "could not copy register head\n");
1763 #endif
1764 			m_freem(m);
1765 			return(IPPROTO_DONE);
1766 		}
1767 
1768 		/*
1769 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1770 		 */
1771 		m_adj(m, off + PIM_MINLEN);
1772 #ifdef MRT6DEBUG
1773 		if (mrt6debug & DEBUG_PIM) {
1774 			log(LOG_DEBUG,
1775 			    "pim6_input: forwarding decapsulated register: "
1776 			    "src %s, dst %s, mif %d\n",
1777 			    ip6_sprintf(&eip6->ip6_src),
1778 			    ip6_sprintf(&eip6->ip6_dst),
1779 			    reg_mif_num);
1780 		}
1781 #endif
1782 
1783  		rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1784 			       dst.sin6_family, NULL);
1785 
1786 		/* prepare the register head to send to the mrouting daemon */
1787 		m = mcp;
1788 	}
1789 
1790 	/*
1791 	 * Pass the PIM message up to the daemon; if it is a register message
1792 	 * pass the 'head' only up to the daemon. This includes the
1793 	 * encapsulator ip6 header, pim header, register header and the
1794 	 * encapsulated ip6 header.
1795 	 */
1796   pim6_input_to_daemon:
1797 	rip6_input(&m, offp, proto);
1798 	return(IPPROTO_DONE);
1799 }
1800