xref: /freebsd/sys/netinet6/ip6_mroute.c (revision c7046f76c2c027b00c0e6ba57cfd28f1a78f5e23)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1989 Stephen Deering
36  * Copyright (c) 1992, 1993
37  *      The Regents of the University of California.  All rights reserved.
38  *
39  * This code is derived from software contributed to Berkeley by
40  * Stephen Deering of Stanford University.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 3. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)ip_mroute.c	8.2 (Berkeley) 11/15/93
67  *	BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp
68  */
69 
70 /*
71  * IP multicast forwarding procedures
72  *
73  * Written by David Waitzman, BBN Labs, August 1988.
74  * Modified by Steve Deering, Stanford, February 1989.
75  * Modified by Mark J. Steiglitz, Stanford, May, 1991
76  * Modified by Van Jacobson, LBL, January 1993
77  * Modified by Ajit Thyagarajan, PARC, August 1993
78  * Modified by Bill Fenner, PARC, April 1994
79  *
80  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
81  */
82 
83 #include <sys/cdefs.h>
84 __FBSDID("$FreeBSD$");
85 
86 #include "opt_inet6.h"
87 
88 #include <sys/param.h>
89 #include <sys/callout.h>
90 #include <sys/errno.h>
91 #include <sys/kernel.h>
92 #include <sys/lock.h>
93 #include <sys/malloc.h>
94 #include <sys/mbuf.h>
95 #include <sys/module.h>
96 #include <sys/domain.h>
97 #include <sys/protosw.h>
98 #include <sys/sdt.h>
99 #include <sys/signalvar.h>
100 #include <sys/socket.h>
101 #include <sys/socketvar.h>
102 #include <sys/sockio.h>
103 #include <sys/sx.h>
104 #include <sys/sysctl.h>
105 #include <sys/syslog.h>
106 #include <sys/systm.h>
107 #include <sys/time.h>
108 
109 #include <net/if.h>
110 #include <net/if_var.h>
111 #include <net/if_types.h>
112 #include <net/vnet.h>
113 
114 #include <netinet/in.h>
115 #include <netinet/in_var.h>
116 #include <netinet/icmp6.h>
117 #include <netinet/ip_encap.h>
118 
119 #include <netinet/ip6.h>
120 #include <netinet/in_kdtrace.h>
121 #include <netinet6/ip6_var.h>
122 #include <netinet6/scope6_var.h>
123 #include <netinet6/nd6.h>
124 #include <netinet6/ip6_mroute.h>
125 #include <netinet6/pim6.h>
126 #include <netinet6/pim6_var.h>
127 
128 static MALLOC_DEFINE(M_MRTABLE6, "mf6c", "multicast forwarding cache entry");
129 
130 static int	ip6_mdq(struct mbuf *, struct ifnet *, struct mf6c *);
131 static void	phyint_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
132 static int	register_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
133 static int	set_pim6(int *);
134 static int	socket_send(struct socket *, struct mbuf *,
135 		    struct sockaddr_in6 *);
136 
137 extern int in6_mcast_loop;
138 extern struct domain inet6domain;
139 
140 static const struct encaptab *pim6_encap_cookie;
141 static int pim6_encapcheck(const struct mbuf *, int, int, void *);
142 static int pim6_input(struct mbuf *, int, int, void *);
143 
144 static const struct encap_config ipv6_encap_cfg = {
145 	.proto = IPPROTO_PIM,
146 	.min_length = sizeof(struct ip6_hdr) + PIM_MINLEN,
147 	.exact_match = 8,
148 	.check = pim6_encapcheck,
149 	.input = pim6_input
150 };
151 
152 VNET_DEFINE_STATIC(int, ip6_mrouter_ver) = 0;
153 #define	V_ip6_mrouter_ver	VNET(ip6_mrouter_ver)
154 
155 SYSCTL_DECL(_net_inet6);
156 SYSCTL_DECL(_net_inet6_ip6);
157 static SYSCTL_NODE(_net_inet6, IPPROTO_PIM, pim,
158     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
159     "PIM");
160 
161 static struct mrt6stat mrt6stat;
162 SYSCTL_STRUCT(_net_inet6_ip6, OID_AUTO, mrt6stat, CTLFLAG_RW,
163     &mrt6stat, mrt6stat,
164     "Multicast Routing Statistics (struct mrt6stat, netinet6/ip6_mroute.h)");
165 
166 #define	MRT6STAT_INC(name)	mrt6stat.name += 1
167 #define NO_RTE_FOUND	0x1
168 #define RTE_FOUND	0x2
169 
170 static struct sx mrouter6_mtx;
171 #define	MROUTER6_LOCKPTR()	(&mrouter6_mtx)
172 #define	MROUTER6_LOCK()		sx_xlock(MROUTER6_LOCKPTR())
173 #define	MROUTER6_UNLOCK()	sx_xunlock(MROUTER6_LOCKPTR())
174 #define	MROUTER6_LOCK_ASSERT()	sx_assert(MROUTER6_LOCKPTR(), SA_XLOCKED
175 #define	MROUTER6_LOCK_INIT()	sx_init(MROUTER6_LOCKPTR(), "mrouter6")
176 #define	MROUTER6_LOCK_DESTROY()	sx_destroy(MROUTER6_LOCKPTR())
177 
178 static struct mf6c *mf6ctable[MF6CTBLSIZ];
179 SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mf6ctable, CTLFLAG_RD,
180     &mf6ctable, sizeof(mf6ctable), "S,*mf6ctable[MF6CTBLSIZ]",
181     "IPv6 Multicast Forwarding Table (struct *mf6ctable[MF6CTBLSIZ], "
182     "netinet6/ip6_mroute.h)");
183 
184 static struct mtx mfc6_mtx;
185 #define	MFC6_LOCKPTR()		(&mfc6_mtx)
186 #define	MFC6_LOCK()		mtx_lock(MFC6_LOCKPTR())
187 #define	MFC6_UNLOCK()		mtx_unlock(MFC6_LOCKPTR())
188 #define	MFC6_LOCK_ASSERT()	mtx_assert(MFC6_LOCKPTR(), MA_OWNED)
189 #define	MFC6_LOCK_INIT()	mtx_init(MFC6_LOCKPTR(),		\
190 				    "IPv6 multicast forwarding cache",	\
191 				    NULL, MTX_DEF)
192 #define	MFC6_LOCK_DESTROY()	mtx_destroy(MFC6_LOCKPTR())
193 
194 static u_char n6expire[MF6CTBLSIZ];
195 
196 static struct mif6 mif6table[MAXMIFS];
197 static int
198 sysctl_mif6table(SYSCTL_HANDLER_ARGS)
199 {
200 	struct mif6_sctl *out;
201 	int error;
202 
203 	out = malloc(sizeof(struct mif6_sctl) * MAXMIFS, M_TEMP,
204 	    M_WAITOK | M_ZERO);
205 	for (int i = 0; i < MAXMIFS; i++) {
206 		out[i].m6_flags		= mif6table[i].m6_flags;
207 		out[i].m6_rate_limit	= mif6table[i].m6_rate_limit;
208 		out[i].m6_lcl_addr	= mif6table[i].m6_lcl_addr;
209 		if (mif6table[i].m6_ifp != NULL)
210 			out[i].m6_ifp	= mif6table[i].m6_ifp->if_index;
211 		else
212 			out[i].m6_ifp	= 0;
213 		out[i].m6_pkt_in	= mif6table[i].m6_pkt_in;
214 		out[i].m6_pkt_out	= mif6table[i].m6_pkt_out;
215 		out[i].m6_bytes_in	= mif6table[i].m6_bytes_in;
216 		out[i].m6_bytes_out	= mif6table[i].m6_bytes_out;
217 	}
218 	error = SYSCTL_OUT(req, out, sizeof(struct mif6_sctl) * MAXMIFS);
219 	free(out, M_TEMP);
220 	return (error);
221 }
222 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, mif6table,
223     CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
224     NULL, 0, sysctl_mif6table, "S,mif6_sctl[MAXMIFS]",
225     "IPv6 Multicast Interfaces (struct mif6_sctl[MAXMIFS], "
226     "netinet6/ip6_mroute.h)");
227 
228 static struct mtx mif6_mtx;
229 #define	MIF6_LOCKPTR()		(&mif6_mtx)
230 #define	MIF6_LOCK()		mtx_lock(MIF6_LOCKPTR())
231 #define	MIF6_UNLOCK()		mtx_unlock(MIF6_LOCKPTR())
232 #define	MIF6_LOCK_ASSERT()	mtx_assert(MIF6_LOCKPTR(), MA_OWNED)
233 #define	MIF6_LOCK_INIT()	\
234 	mtx_init(MIF6_LOCKPTR(), "IPv6 multicast interfaces", NULL, MTX_DEF)
235 #define	MIF6_LOCK_DESTROY()	mtx_destroy(MIF6_LOCKPTR())
236 
237 #ifdef MRT6DEBUG
238 VNET_DEFINE_STATIC(u_int, mrt6debug) = 0;	/* debug level */
239 #define	V_mrt6debug		VNET(mrt6debug)
240 #define DEBUG_MFC	0x02
241 #define DEBUG_FORWARD	0x04
242 #define DEBUG_EXPIRE	0x08
243 #define DEBUG_XMIT	0x10
244 #define DEBUG_REG	0x20
245 #define DEBUG_PIM	0x40
246 #define	DEBUG_ERR	0x80
247 #define	DEBUG_ANY	0x7f
248 #define	MRT6_DLOG(m, fmt, ...)	\
249 	if (V_mrt6debug & (m))	\
250 		log(((m) & DEBUG_ERR) ? LOG_ERR: LOG_DEBUG, \
251 		    "%s: " fmt "\n", __func__, ##__VA_ARGS__)
252 #else
253 #define	MRT6_DLOG(m, fmt, ...)
254 #endif
255 
256 static void	expire_upcalls(void *);
257 #define	EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
258 #define	UPCALL_EXPIRE	6		/* number of timeouts */
259 
260 /*
261  * XXX TODO: maintain a count to if_allmulti() calls in struct ifnet.
262  */
263 
264 /*
265  * 'Interfaces' associated with decapsulator (so we can tell
266  * packets that went through it from ones that get reflected
267  * by a broken gateway).  Different from IPv4 register_if,
268  * these interfaces are linked into the system ifnet list,
269  * because per-interface IPv6 statistics are maintained in
270  * ifp->if_afdata.  But it does not have any routes point
271  * to them.  I.e., packets can't be sent this way.  They
272  * only exist as a placeholder for multicast source
273  * verification.
274  */
275 static struct ifnet *multicast_register_if6;
276 
277 #define ENCAP_HOPS 64
278 
279 /*
280  * Private variables.
281  */
282 static mifi_t nummifs = 0;
283 static mifi_t reg_mif_num = (mifi_t)-1;
284 
285 static struct pim6stat pim6stat;
286 SYSCTL_STRUCT(_net_inet6_pim, PIM6CTL_STATS, stats, CTLFLAG_RW,
287     &pim6stat, pim6stat,
288     "PIM Statistics (struct pim6stat, netinet6/pim6_var.h)");
289 
290 #define	PIM6STAT_INC(name)	pim6stat.name += 1
291 VNET_DEFINE_STATIC(int, pim6);
292 #define	V_pim6		VNET(pim6)
293 
294 /*
295  * Hash function for a source, group entry
296  */
297 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
298 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
299 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
300 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
301 
302 /*
303  * Find a route for a given origin IPv6 address and Multicast group address.
304  */
305 #define MF6CFIND(o, g, rt) do { \
306 	struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
307 	rt = NULL; \
308 	while (_rt) { \
309 		if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
310 		    IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
311 		    (_rt->mf6c_stall == NULL)) { \
312 			rt = _rt; \
313 			break; \
314 		} \
315 		_rt = _rt->mf6c_next; \
316 	} \
317 	if (rt == NULL) { \
318 		MRT6STAT_INC(mrt6s_mfc_misses); \
319 	} \
320 } while (/*CONSTCOND*/ 0)
321 
322 /*
323  * Macros to compute elapsed time efficiently
324  * Borrowed from Van Jacobson's scheduling code
325  * XXX: replace with timersub() ?
326  */
327 #define TV_DELTA(a, b, delta) do { \
328 	    int xxs; \
329 		\
330 	    delta = (a).tv_usec - (b).tv_usec; \
331 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
332 	       switch (xxs) { \
333 		      case 2: \
334 			  delta += 1000000; \
335 			      /* FALLTHROUGH */ \
336 		      case 1: \
337 			  delta += 1000000; \
338 			  break; \
339 		      default: \
340 			  delta += (1000000 * xxs); \
341 	       } \
342 	    } \
343 } while (/*CONSTCOND*/ 0)
344 
345 /* XXX: replace with timercmp(a, b, <) ? */
346 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
347 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
348 
349 #ifdef UPCALL_TIMING
350 #define UPCALL_MAX	50
351 static u_long upcall_data[UPCALL_MAX + 1];
352 static void collate();
353 #endif /* UPCALL_TIMING */
354 
355 static int ip6_mrouter_init(struct socket *, int, int);
356 static int add_m6fc(struct mf6cctl *);
357 static int add_m6if(struct mif6ctl *);
358 static int del_m6fc(struct mf6cctl *);
359 static int del_m6if(mifi_t *);
360 static int del_m6if_locked(mifi_t *);
361 static int get_mif6_cnt(struct sioc_mif_req6 *);
362 static int get_sg_cnt(struct sioc_sg_req6 *);
363 
364 static struct callout expire_upcalls_ch;
365 
366 int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *);
367 int X_ip6_mrouter_done(void);
368 int X_ip6_mrouter_set(struct socket *, struct sockopt *);
369 int X_ip6_mrouter_get(struct socket *, struct sockopt *);
370 int X_mrt6_ioctl(u_long, caddr_t);
371 
372 /*
373  * Handle MRT setsockopt commands to modify the multicast routing tables.
374  */
375 int
376 X_ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
377 {
378 	int error = 0;
379 	int optval;
380 	struct mif6ctl mifc;
381 	struct mf6cctl mfcc;
382 	mifi_t mifi;
383 
384 	if (so != V_ip6_mrouter && sopt->sopt_name != MRT6_INIT)
385 		return (EPERM);
386 
387 	switch (sopt->sopt_name) {
388 	case MRT6_INIT:
389 #ifdef MRT6_OINIT
390 	case MRT6_OINIT:
391 #endif
392 		error = sooptcopyin(sopt, &optval, sizeof(optval),
393 		    sizeof(optval));
394 		if (error)
395 			break;
396 		error = ip6_mrouter_init(so, optval, sopt->sopt_name);
397 		break;
398 	case MRT6_DONE:
399 		error = X_ip6_mrouter_done();
400 		break;
401 	case MRT6_ADD_MIF:
402 		error = sooptcopyin(sopt, &mifc, sizeof(mifc), sizeof(mifc));
403 		if (error)
404 			break;
405 		error = add_m6if(&mifc);
406 		break;
407 	case MRT6_ADD_MFC:
408 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
409 		if (error)
410 			break;
411 		error = add_m6fc(&mfcc);
412 		break;
413 	case MRT6_DEL_MFC:
414 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
415 		if (error)
416 			break;
417 		error = del_m6fc(&mfcc);
418 		break;
419 	case MRT6_DEL_MIF:
420 		error = sooptcopyin(sopt, &mifi, sizeof(mifi), sizeof(mifi));
421 		if (error)
422 			break;
423 		error = del_m6if(&mifi);
424 		break;
425 	case MRT6_PIM:
426 		error = sooptcopyin(sopt, &optval, sizeof(optval),
427 		    sizeof(optval));
428 		if (error)
429 			break;
430 		error = set_pim6(&optval);
431 		break;
432 	default:
433 		error = EOPNOTSUPP;
434 		break;
435 	}
436 
437 	return (error);
438 }
439 
440 /*
441  * Handle MRT getsockopt commands
442  */
443 int
444 X_ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
445 {
446 	int error = 0;
447 
448 	if (so != V_ip6_mrouter)
449 		return (EACCES);
450 
451 	switch (sopt->sopt_name) {
452 		case MRT6_PIM:
453 			error = sooptcopyout(sopt, &V_pim6, sizeof(V_pim6));
454 			break;
455 	}
456 	return (error);
457 }
458 
459 /*
460  * Handle ioctl commands to obtain information from the cache
461  */
462 int
463 X_mrt6_ioctl(u_long cmd, caddr_t data)
464 {
465 	int ret;
466 
467 	ret = EINVAL;
468 
469 	switch (cmd) {
470 	case SIOCGETSGCNT_IN6:
471 		ret = get_sg_cnt((struct sioc_sg_req6 *)data);
472 		break;
473 
474 	case SIOCGETMIFCNT_IN6:
475 		ret = get_mif6_cnt((struct sioc_mif_req6 *)data);
476 		break;
477 
478 	default:
479 		break;
480 	}
481 
482 	return (ret);
483 }
484 
485 /*
486  * returns the packet, byte, rpf-failure count for the source group provided
487  */
488 static int
489 get_sg_cnt(struct sioc_sg_req6 *req)
490 {
491 	struct mf6c *rt;
492 	int ret;
493 
494 	ret = 0;
495 
496 	MFC6_LOCK();
497 
498 	MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
499 	if (rt == NULL) {
500 		ret = ESRCH;
501 	} else {
502 		req->pktcnt = rt->mf6c_pkt_cnt;
503 		req->bytecnt = rt->mf6c_byte_cnt;
504 		req->wrong_if = rt->mf6c_wrong_if;
505 	}
506 
507 	MFC6_UNLOCK();
508 
509 	return (ret);
510 }
511 
512 /*
513  * returns the input and output packet and byte counts on the mif provided
514  */
515 static int
516 get_mif6_cnt(struct sioc_mif_req6 *req)
517 {
518 	mifi_t mifi;
519 	int ret;
520 
521 	ret = 0;
522 	mifi = req->mifi;
523 
524 	MIF6_LOCK();
525 
526 	if (mifi >= nummifs) {
527 		ret = EINVAL;
528 	} else {
529 		req->icount = mif6table[mifi].m6_pkt_in;
530 		req->ocount = mif6table[mifi].m6_pkt_out;
531 		req->ibytes = mif6table[mifi].m6_bytes_in;
532 		req->obytes = mif6table[mifi].m6_bytes_out;
533 	}
534 
535 	MIF6_UNLOCK();
536 
537 	return (ret);
538 }
539 
540 static int
541 set_pim6(int *i)
542 {
543 	if ((*i != 1) && (*i != 0))
544 		return (EINVAL);
545 
546 	V_pim6 = *i;
547 
548 	return (0);
549 }
550 
551 /*
552  * Enable multicast routing
553  */
554 static int
555 ip6_mrouter_init(struct socket *so, int v, int cmd)
556 {
557 
558 	MRT6_DLOG(DEBUG_ANY, "%s: socket %p", __func__, so);
559 
560 	if (v != 1)
561 		return (ENOPROTOOPT);
562 
563 	MROUTER6_LOCK();
564 
565 	if (V_ip6_mrouter != NULL) {
566 		MROUTER6_UNLOCK();
567 		return (EADDRINUSE);
568 	}
569 
570 	V_ip6_mrouter = so;
571 	V_ip6_mrouter_ver = cmd;
572 
573 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
574 	bzero((caddr_t)n6expire, sizeof(n6expire));
575 
576 	V_pim6 = 0;/* used for stubbing out/in pim stuff */
577 
578 	callout_init_mtx(&expire_upcalls_ch, MFC6_LOCKPTR(), 0);
579 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
580 	    expire_upcalls, NULL);
581 
582 	MROUTER6_UNLOCK();
583 
584 	MRT6_DLOG(DEBUG_ANY, "finished");
585 
586 	return (0);
587 }
588 
589 /*
590  * Disable IPv6 multicast forwarding.
591  */
592 int
593 X_ip6_mrouter_done(void)
594 {
595 	mifi_t mifi;
596 	u_long i;
597 	struct mf6c *rt;
598 	struct rtdetq *rte;
599 
600 	MROUTER6_LOCK();
601 
602 	if (V_ip6_mrouter == NULL) {
603 		MROUTER6_UNLOCK();
604 		return (EINVAL);
605 	}
606 
607 	/*
608 	 * For each phyint in use, disable promiscuous reception of all IPv6
609 	 * multicasts.
610 	 */
611 	for (mifi = 0; mifi < nummifs; mifi++) {
612 		if (mif6table[mifi].m6_ifp &&
613 		    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
614 			if_allmulti(mif6table[mifi].m6_ifp, 0);
615 		}
616 	}
617 	bzero((caddr_t)mif6table, sizeof(mif6table));
618 	nummifs = 0;
619 
620 	V_pim6 = 0; /* used to stub out/in pim specific code */
621 
622 	/*
623 	 * Free all multicast forwarding cache entries.
624 	 */
625 	MFC6_LOCK();
626 	for (i = 0; i < MF6CTBLSIZ; i++) {
627 		rt = mf6ctable[i];
628 		while (rt) {
629 			struct mf6c *frt;
630 
631 			for (rte = rt->mf6c_stall; rte != NULL; ) {
632 				struct rtdetq *n = rte->next;
633 
634 				m_freem(rte->m);
635 				free(rte, M_MRTABLE6);
636 				rte = n;
637 			}
638 			frt = rt;
639 			rt = rt->mf6c_next;
640 			free(frt, M_MRTABLE6);
641 		}
642 	}
643 	bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
644 	MFC6_UNLOCK();
645 
646 	callout_drain(&expire_upcalls_ch);
647 
648 	/*
649 	 * Reset register interface
650 	 */
651 	if (reg_mif_num != (mifi_t)-1 && multicast_register_if6 != NULL) {
652 		if_detach(multicast_register_if6);
653 		if_free(multicast_register_if6);
654 		reg_mif_num = (mifi_t)-1;
655 		multicast_register_if6 = NULL;
656 	}
657 
658 	V_ip6_mrouter = NULL;
659 	V_ip6_mrouter_ver = 0;
660 
661 	MROUTER6_UNLOCK();
662 	MRT6_DLOG(DEBUG_ANY, "finished");
663 
664 	return (0);
665 }
666 
667 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
668 
669 /*
670  * Add a mif to the mif table
671  */
672 static int
673 add_m6if(struct mif6ctl *mifcp)
674 {
675 	struct epoch_tracker et;
676 	struct mif6 *mifp;
677 	struct ifnet *ifp;
678 	int error;
679 
680 	MIF6_LOCK();
681 
682 	if (mifcp->mif6c_mifi >= MAXMIFS) {
683 		MIF6_UNLOCK();
684 		return (EINVAL);
685 	}
686 	mifp = mif6table + mifcp->mif6c_mifi;
687 	if (mifp->m6_ifp != NULL) {
688 		MIF6_UNLOCK();
689 		return (EADDRINUSE); /* XXX: is it appropriate? */
690 	}
691 
692 	NET_EPOCH_ENTER(et);
693 	if ((ifp = ifnet_byindex(mifcp->mif6c_pifi)) == NULL) {
694 		NET_EPOCH_EXIT(et);
695 		MIF6_UNLOCK();
696 		return (ENXIO);
697 	}
698 	NET_EPOCH_EXIT(et);	/* XXXGL: unsafe ifp */
699 
700 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
701 		if (reg_mif_num == (mifi_t)-1) {
702 			ifp = if_alloc(IFT_OTHER);
703 
704 			if_initname(ifp, "register_mif", 0);
705 			ifp->if_flags |= IFF_LOOPBACK;
706 			if_attach(ifp);
707 			multicast_register_if6 = ifp;
708 			reg_mif_num = mifcp->mif6c_mifi;
709 			/*
710 			 * it is impossible to guess the ifindex of the
711 			 * register interface.  So mif6c_pifi is automatically
712 			 * calculated.
713 			 */
714 			mifcp->mif6c_pifi = ifp->if_index;
715 		} else {
716 			ifp = multicast_register_if6;
717 		}
718 	} else {
719 		/* Make sure the interface supports multicast */
720 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
721 			MIF6_UNLOCK();
722 			return (EOPNOTSUPP);
723 		}
724 
725 		error = if_allmulti(ifp, 1);
726 		if (error) {
727 			MIF6_UNLOCK();
728 			return (error);
729 		}
730 	}
731 
732 	mifp->m6_flags     = mifcp->mif6c_flags;
733 	mifp->m6_ifp       = ifp;
734 
735 	/* initialize per mif pkt counters */
736 	mifp->m6_pkt_in    = 0;
737 	mifp->m6_pkt_out   = 0;
738 	mifp->m6_bytes_in  = 0;
739 	mifp->m6_bytes_out = 0;
740 
741 	/* Adjust nummifs up if the mifi is higher than nummifs */
742 	if (nummifs <= mifcp->mif6c_mifi)
743 		nummifs = mifcp->mif6c_mifi + 1;
744 
745 	MIF6_UNLOCK();
746 	MRT6_DLOG(DEBUG_ANY, "mif #%d, phyint %s", mifcp->mif6c_mifi,
747 	    if_name(ifp));
748 
749 	return (0);
750 }
751 
752 /*
753  * Delete a mif from the mif table
754  */
755 static int
756 del_m6if_locked(mifi_t *mifip)
757 {
758 	struct mif6 *mifp = mif6table + *mifip;
759 	mifi_t mifi;
760 	struct ifnet *ifp;
761 
762 	MIF6_LOCK_ASSERT();
763 
764 	if (*mifip >= nummifs)
765 		return (EINVAL);
766 	if (mifp->m6_ifp == NULL)
767 		return (EINVAL);
768 
769 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
770 		/* XXX: TODO: Maintain an ALLMULTI refcount in struct ifnet. */
771 		ifp = mifp->m6_ifp;
772 		if_allmulti(ifp, 0);
773 	} else {
774 		if (reg_mif_num != (mifi_t)-1 &&
775 		    multicast_register_if6 != NULL) {
776 			if_detach(multicast_register_if6);
777 			if_free(multicast_register_if6);
778 			reg_mif_num = (mifi_t)-1;
779 			multicast_register_if6 = NULL;
780 		}
781 	}
782 
783 	bzero((caddr_t)mifp, sizeof(*mifp));
784 
785 	/* Adjust nummifs down */
786 	for (mifi = nummifs; mifi > 0; mifi--)
787 		if (mif6table[mifi - 1].m6_ifp)
788 			break;
789 	nummifs = mifi;
790 	MRT6_DLOG(DEBUG_ANY, "mif %d, nummifs %d", *mifip, nummifs);
791 
792 	return (0);
793 }
794 
795 static int
796 del_m6if(mifi_t *mifip)
797 {
798 	int cc;
799 
800 	MIF6_LOCK();
801 	cc = del_m6if_locked(mifip);
802 	MIF6_UNLOCK();
803 
804 	return (cc);
805 }
806 
807 /*
808  * Add an mfc entry
809  */
810 static int
811 add_m6fc(struct mf6cctl *mfccp)
812 {
813 	struct mf6c *rt;
814 	u_long hash;
815 	struct rtdetq *rte;
816 	u_short nstl;
817 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
818 
819 	MFC6_LOCK();
820 
821 	MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
822 		 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
823 
824 	/* If an entry already exists, just update the fields */
825 	if (rt) {
826 		MRT6_DLOG(DEBUG_MFC, "no upcall o %s g %s p %x",
827 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
828 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
829 		    mfccp->mf6cc_parent);
830 
831 		rt->mf6c_parent = mfccp->mf6cc_parent;
832 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
833 
834 		MFC6_UNLOCK();
835 		return (0);
836 	}
837 
838 	/*
839 	 * Find the entry for which the upcall was made and update
840 	 */
841 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
842 			mfccp->mf6cc_mcastgrp.sin6_addr);
843 	for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
844 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
845 				       &mfccp->mf6cc_origin.sin6_addr) &&
846 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
847 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
848 		    (rt->mf6c_stall != NULL)) {
849 			if (nstl++)
850 				log(LOG_ERR,
851 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
852 				    "multiple kernel entries",
853 				    ip6_sprintf(ip6bufo,
854 					    &mfccp->mf6cc_origin.sin6_addr),
855 				    ip6_sprintf(ip6bufg,
856 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
857 				    mfccp->mf6cc_parent, rt->mf6c_stall);
858 
859 			MRT6_DLOG(DEBUG_MFC, "o %s g %s p %x dbg %p",
860 			    ip6_sprintf(ip6bufo,
861 			    &mfccp->mf6cc_origin.sin6_addr),
862 			    ip6_sprintf(ip6bufg,
863 				&mfccp->mf6cc_mcastgrp.sin6_addr),
864 			    mfccp->mf6cc_parent, rt->mf6c_stall);
865 
866 			rt->mf6c_origin     = mfccp->mf6cc_origin;
867 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
868 			rt->mf6c_parent     = mfccp->mf6cc_parent;
869 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
870 			/* initialize pkt counters per src-grp */
871 			rt->mf6c_pkt_cnt    = 0;
872 			rt->mf6c_byte_cnt   = 0;
873 			rt->mf6c_wrong_if   = 0;
874 
875 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
876 			n6expire[hash]--;
877 
878 			/* free packets Qed at the end of this entry */
879 			for (rte = rt->mf6c_stall; rte != NULL; ) {
880 				struct rtdetq *n = rte->next;
881 				ip6_mdq(rte->m, rte->ifp, rt);
882 				m_freem(rte->m);
883 #ifdef UPCALL_TIMING
884 				collate(&(rte->t));
885 #endif /* UPCALL_TIMING */
886 				free(rte, M_MRTABLE6);
887 				rte = n;
888 			}
889 			rt->mf6c_stall = NULL;
890 		}
891 	}
892 
893 	/*
894 	 * It is possible that an entry is being inserted without an upcall
895 	 */
896 	if (nstl == 0) {
897 		MRT6_DLOG(DEBUG_MFC, "no upcall h %lu o %s g %s p %x", hash,
898 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
899 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
900 		    mfccp->mf6cc_parent);
901 
902 		for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
903 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
904 					       &mfccp->mf6cc_origin.sin6_addr)&&
905 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
906 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
907 				rt->mf6c_origin     = mfccp->mf6cc_origin;
908 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
909 				rt->mf6c_parent     = mfccp->mf6cc_parent;
910 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
911 				/* initialize pkt counters per src-grp */
912 				rt->mf6c_pkt_cnt    = 0;
913 				rt->mf6c_byte_cnt   = 0;
914 				rt->mf6c_wrong_if   = 0;
915 
916 				if (rt->mf6c_expire)
917 					n6expire[hash]--;
918 				rt->mf6c_expire	   = 0;
919 			}
920 		}
921 		if (rt == NULL) {
922 			/* no upcall, so make a new entry */
923 			rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6,
924 						  M_NOWAIT);
925 			if (rt == NULL) {
926 				MFC6_UNLOCK();
927 				return (ENOBUFS);
928 			}
929 
930 			/* insert new entry at head of hash chain */
931 			rt->mf6c_origin     = mfccp->mf6cc_origin;
932 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
933 			rt->mf6c_parent     = mfccp->mf6cc_parent;
934 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
935 			/* initialize pkt counters per src-grp */
936 			rt->mf6c_pkt_cnt    = 0;
937 			rt->mf6c_byte_cnt   = 0;
938 			rt->mf6c_wrong_if   = 0;
939 			rt->mf6c_expire     = 0;
940 			rt->mf6c_stall = NULL;
941 
942 			/* link into table */
943 			rt->mf6c_next  = mf6ctable[hash];
944 			mf6ctable[hash] = rt;
945 		}
946 	}
947 
948 	MFC6_UNLOCK();
949 	return (0);
950 }
951 
952 #ifdef UPCALL_TIMING
953 /*
954  * collect delay statistics on the upcalls
955  */
956 static void
957 collate(struct timeval *t)
958 {
959 	u_long d;
960 	struct timeval tp;
961 	u_long delta;
962 
963 	GET_TIME(tp);
964 
965 	if (TV_LT(*t, tp))
966 	{
967 		TV_DELTA(tp, *t, delta);
968 
969 		d = delta >> 10;
970 		if (d > UPCALL_MAX)
971 			d = UPCALL_MAX;
972 
973 		++upcall_data[d];
974 	}
975 }
976 #endif /* UPCALL_TIMING */
977 
978 /*
979  * Delete an mfc entry
980  */
981 static int
982 del_m6fc(struct mf6cctl *mfccp)
983 {
984 #ifdef MRT6DEBUG
985 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
986 #endif
987 	struct sockaddr_in6	origin;
988 	struct sockaddr_in6	mcastgrp;
989 	struct mf6c		*rt;
990 	struct mf6c		**nptr;
991 	u_long		hash;
992 
993 	origin = mfccp->mf6cc_origin;
994 	mcastgrp = mfccp->mf6cc_mcastgrp;
995 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
996 
997 	MRT6_DLOG(DEBUG_MFC, "orig %s mcastgrp %s",
998 	    ip6_sprintf(ip6bufo, &origin.sin6_addr),
999 	    ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr));
1000 
1001 	MFC6_LOCK();
1002 
1003 	nptr = &mf6ctable[hash];
1004 	while ((rt = *nptr) != NULL) {
1005 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
1006 				       &rt->mf6c_origin.sin6_addr) &&
1007 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
1008 				       &rt->mf6c_mcastgrp.sin6_addr) &&
1009 		    rt->mf6c_stall == NULL)
1010 			break;
1011 
1012 		nptr = &rt->mf6c_next;
1013 	}
1014 	if (rt == NULL) {
1015 		MFC6_UNLOCK();
1016 		return (EADDRNOTAVAIL);
1017 	}
1018 
1019 	*nptr = rt->mf6c_next;
1020 	free(rt, M_MRTABLE6);
1021 
1022 	MFC6_UNLOCK();
1023 
1024 	return (0);
1025 }
1026 
1027 static int
1028 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
1029 {
1030 
1031 	if (s) {
1032 		if (sbappendaddr(&s->so_rcv,
1033 				 (struct sockaddr *)src,
1034 				 mm, (struct mbuf *)0) != 0) {
1035 			sorwakeup(s);
1036 			return (0);
1037 		} else
1038 			soroverflow(s);
1039 	}
1040 	m_freem(mm);
1041 	return (-1);
1042 }
1043 
1044 /*
1045  * IPv6 multicast forwarding function. This function assumes that the packet
1046  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
1047  * pointed to by "ifp", and the packet is to be relayed to other networks
1048  * that have members of the packet's destination IPv6 multicast group.
1049  *
1050  * The packet is returned unscathed to the caller, unless it is
1051  * erroneous, in which case a non-zero return value tells the caller to
1052  * discard it.
1053  *
1054  * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff
1055  * this function is called in the originating context (i.e., not when
1056  * forwarding a packet from other node).  ip6_output(), which is currently the
1057  * only function that calls this function is called in the originating context,
1058  * explicitly ensures this condition.  It is caller's responsibility to ensure
1059  * that if this function is called from somewhere else in the originating
1060  * context in the future.
1061  */
1062 int
1063 X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
1064 {
1065 	struct rtdetq *rte;
1066 	struct mbuf *mb0;
1067 	struct mf6c *rt;
1068 	struct mif6 *mifp;
1069 	struct mbuf *mm;
1070 	u_long hash;
1071 	mifi_t mifi;
1072 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1073 #ifdef UPCALL_TIMING
1074 	struct timeval tp;
1075 
1076 	GET_TIME(tp);
1077 #endif /* UPCALL_TIMING */
1078 
1079 	MRT6_DLOG(DEBUG_FORWARD, "src %s, dst %s, ifindex %d",
1080 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1081 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst), ifp->if_index);
1082 
1083 	/*
1084 	 * Don't forward a packet with Hop limit of zero or one,
1085 	 * or a packet destined to a local-only group.
1086 	 */
1087 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
1088 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1089 		return (0);
1090 	ip6->ip6_hlim--;
1091 
1092 	/*
1093 	 * Source address check: do not forward packets with unspecified
1094 	 * source. It was discussed in July 2000, on ipngwg mailing list.
1095 	 * This is rather more serious than unicast cases, because some
1096 	 * MLD packets can be sent with the unspecified source address
1097 	 * (although such packets must normally set 1 to the hop limit field).
1098 	 */
1099 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1100 		IP6STAT_INC(ip6s_cantforward);
1101 		if (V_ip6_log_time + V_ip6_log_interval < time_uptime) {
1102 			V_ip6_log_time = time_uptime;
1103 			log(LOG_DEBUG,
1104 			    "cannot forward "
1105 			    "from %s to %s nxt %d received on %s\n",
1106 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1107 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1108 			    ip6->ip6_nxt,
1109 			    if_name(m->m_pkthdr.rcvif));
1110 		}
1111 		return (0);
1112 	}
1113 
1114 	MFC6_LOCK();
1115 
1116 	/*
1117 	 * Determine forwarding mifs from the forwarding cache table
1118 	 */
1119 	MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
1120 	MRT6STAT_INC(mrt6s_mfc_lookups);
1121 
1122 	/* Entry exists, so forward if necessary */
1123 	if (rt) {
1124 		MFC6_UNLOCK();
1125 		return (ip6_mdq(m, ifp, rt));
1126 	}
1127 
1128 	/*
1129 	 * If we don't have a route for packet's origin,
1130 	 * Make a copy of the packet & send message to routing daemon.
1131 	 */
1132 	MRT6STAT_INC(mrt6s_no_route);
1133 	MRT6_DLOG(DEBUG_FORWARD | DEBUG_MFC, "no rte s %s g %s",
1134 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1135 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1136 
1137 	/*
1138 	 * Allocate mbufs early so that we don't do extra work if we
1139 	 * are just going to fail anyway.
1140 	 */
1141 	rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE6, M_NOWAIT);
1142 	if (rte == NULL) {
1143 		MFC6_UNLOCK();
1144 		return (ENOBUFS);
1145 	}
1146 	mb0 = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1147 	/*
1148 	 * Pullup packet header if needed before storing it,
1149 	 * as other references may modify it in the meantime.
1150 	 */
1151 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1152 		mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1153 	if (mb0 == NULL) {
1154 		free(rte, M_MRTABLE6);
1155 		MFC6_UNLOCK();
1156 		return (ENOBUFS);
1157 	}
1158 
1159 	/* is there an upcall waiting for this packet? */
1160 	hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1161 	for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1162 		if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1163 		    &rt->mf6c_origin.sin6_addr) &&
1164 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1165 		    &rt->mf6c_mcastgrp.sin6_addr) && (rt->mf6c_stall != NULL))
1166 			break;
1167 	}
1168 
1169 	if (rt == NULL) {
1170 		struct mrt6msg *im;
1171 #ifdef MRT6_OINIT
1172 		struct omrt6msg *oim;
1173 #endif
1174 		/* no upcall, so make a new entry */
1175 		rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6, M_NOWAIT);
1176 		if (rt == NULL) {
1177 			free(rte, M_MRTABLE6);
1178 			m_freem(mb0);
1179 			MFC6_UNLOCK();
1180 			return (ENOBUFS);
1181 		}
1182 		/*
1183 		 * Make a copy of the header to send to the user
1184 		 * level process
1185 		 */
1186 		mm = m_copym(mb0, 0, sizeof(struct ip6_hdr), M_NOWAIT);
1187 		if (mm == NULL) {
1188 			free(rte, M_MRTABLE6);
1189 			m_freem(mb0);
1190 			free(rt, M_MRTABLE6);
1191 			MFC6_UNLOCK();
1192 			return (ENOBUFS);
1193 		}
1194 
1195 		/*
1196 		 * Send message to routing daemon
1197 		 */
1198 		sin6.sin6_addr = ip6->ip6_src;
1199 		im = NULL;
1200 #ifdef MRT6_OINIT
1201 		oim = NULL;
1202 #endif
1203 		switch (V_ip6_mrouter_ver) {
1204 #ifdef MRT6_OINIT
1205 		case MRT6_OINIT:
1206 			oim = mtod(mm, struct omrt6msg *);
1207 			oim->im6_msgtype = MRT6MSG_NOCACHE;
1208 			oim->im6_mbz = 0;
1209 			break;
1210 #endif
1211 		case MRT6_INIT:
1212 			im = mtod(mm, struct mrt6msg *);
1213 			im->im6_msgtype = MRT6MSG_NOCACHE;
1214 			im->im6_mbz = 0;
1215 			break;
1216 		default:
1217 			free(rte, M_MRTABLE6);
1218 			m_freem(mb0);
1219 			free(rt, M_MRTABLE6);
1220 			MFC6_UNLOCK();
1221 			return (EINVAL);
1222 		}
1223 
1224 		MRT6_DLOG(DEBUG_FORWARD, "getting the iif info in the kernel");
1225 		for (mifp = mif6table, mifi = 0;
1226 		    mifi < nummifs && mifp->m6_ifp != ifp; mifp++, mifi++)
1227 				;
1228 
1229 		switch (V_ip6_mrouter_ver) {
1230 #ifdef MRT6_OINIT
1231 		case MRT6_OINIT:
1232 			oim->im6_mif = mifi;
1233 			break;
1234 #endif
1235 		case MRT6_INIT:
1236 			im->im6_mif = mifi;
1237 			break;
1238 		}
1239 
1240 		if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1241 			log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1242 			    "socket queue full\n");
1243 			MRT6STAT_INC(mrt6s_upq_sockfull);
1244 			free(rte, M_MRTABLE6);
1245 			m_freem(mb0);
1246 			free(rt, M_MRTABLE6);
1247 			MFC6_UNLOCK();
1248 			return (ENOBUFS);
1249 		}
1250 
1251 		MRT6STAT_INC(mrt6s_upcalls);
1252 
1253 		/* insert new entry at head of hash chain */
1254 		bzero(rt, sizeof(*rt));
1255 		rt->mf6c_origin.sin6_family = AF_INET6;
1256 		rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1257 		rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1258 		rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1259 		rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1260 		rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1261 		rt->mf6c_expire = UPCALL_EXPIRE;
1262 		n6expire[hash]++;
1263 		rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1264 
1265 		/* link into table */
1266 		rt->mf6c_next  = mf6ctable[hash];
1267 		mf6ctable[hash] = rt;
1268 		/* Add this entry to the end of the queue */
1269 		rt->mf6c_stall = rte;
1270 	} else {
1271 		/* determine if q has overflowed */
1272 		struct rtdetq **p;
1273 		int npkts = 0;
1274 
1275 		for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1276 			if (++npkts > MAX_UPQ6) {
1277 				MRT6STAT_INC(mrt6s_upq_ovflw);
1278 				free(rte, M_MRTABLE6);
1279 				m_freem(mb0);
1280 				MFC6_UNLOCK();
1281 				return (0);
1282 			}
1283 
1284 		/* Add this entry to the end of the queue */
1285 		*p = rte;
1286 	}
1287 
1288 	rte->next = NULL;
1289 	rte->m = mb0;
1290 	rte->ifp = ifp;
1291 #ifdef UPCALL_TIMING
1292 	rte->t = tp;
1293 #endif /* UPCALL_TIMING */
1294 
1295 	MFC6_UNLOCK();
1296 
1297 	return (0);
1298 }
1299 
1300 /*
1301  * Clean up cache entries if upcalls are not serviced
1302  * Call from the Slow Timeout mechanism, every half second.
1303  */
1304 static void
1305 expire_upcalls(void *unused)
1306 {
1307 #ifdef MRT6DEBUG
1308 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1309 #endif
1310 	struct rtdetq *rte;
1311 	struct mf6c *mfc, **nptr;
1312 	u_long i;
1313 
1314 	MFC6_LOCK_ASSERT();
1315 
1316 	for (i = 0; i < MF6CTBLSIZ; i++) {
1317 		if (n6expire[i] == 0)
1318 			continue;
1319 		nptr = &mf6ctable[i];
1320 		while ((mfc = *nptr) != NULL) {
1321 			rte = mfc->mf6c_stall;
1322 			/*
1323 			 * Skip real cache entries
1324 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1325 			 * If it expires now
1326 			 */
1327 			if (rte != NULL &&
1328 			    mfc->mf6c_expire != 0 &&
1329 			    --mfc->mf6c_expire == 0) {
1330 				MRT6_DLOG(DEBUG_EXPIRE, "expiring (%s %s)",
1331 				    ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr),
1332 				    ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr));
1333 				/*
1334 				 * drop all the packets
1335 				 * free the mbuf with the pkt, if, timing info
1336 				 */
1337 				do {
1338 					struct rtdetq *n = rte->next;
1339 					m_freem(rte->m);
1340 					free(rte, M_MRTABLE6);
1341 					rte = n;
1342 				} while (rte != NULL);
1343 				MRT6STAT_INC(mrt6s_cache_cleanups);
1344 				n6expire[i]--;
1345 
1346 				*nptr = mfc->mf6c_next;
1347 				free(mfc, M_MRTABLE6);
1348 			} else {
1349 				nptr = &mfc->mf6c_next;
1350 			}
1351 		}
1352 	}
1353 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1354 	    expire_upcalls, NULL);
1355 }
1356 
1357 /*
1358  * Packet forwarding routine once entry in the cache is made
1359  */
1360 static int
1361 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1362 {
1363 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1364 	mifi_t mifi, iif;
1365 	struct mif6 *mifp;
1366 	int plen = m->m_pkthdr.len;
1367 	struct in6_addr src0, dst0; /* copies for local work */
1368 	u_int32_t iszone, idzone, oszone, odzone;
1369 	int error = 0;
1370 
1371 	/*
1372 	 * Don't forward if it didn't arrive from the parent mif
1373 	 * for its origin.
1374 	 */
1375 	mifi = rt->mf6c_parent;
1376 	if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1377 		/* came in the wrong interface */
1378 		MRT6_DLOG(DEBUG_FORWARD,
1379 		    "wrong if: ifid %d mifi %d mififid %x", ifp->if_index,
1380 		    mifi, mif6table[mifi].m6_ifp->if_index);
1381 		MRT6STAT_INC(mrt6s_wrong_if);
1382 		rt->mf6c_wrong_if++;
1383 		/*
1384 		 * If we are doing PIM processing, and we are forwarding
1385 		 * packets on this interface, send a message to the
1386 		 * routing daemon.
1387 		 */
1388 		/* have to make sure this is a valid mif */
1389 		if (mifi < nummifs && mif6table[mifi].m6_ifp)
1390 			if (V_pim6 && (m->m_flags & M_LOOP) == 0) {
1391 				/*
1392 				 * Check the M_LOOP flag to avoid an
1393 				 * unnecessary PIM assert.
1394 				 * XXX: M_LOOP is an ad-hoc hack...
1395 				 */
1396 				static struct sockaddr_in6 sin6 =
1397 				{ sizeof(sin6), AF_INET6 };
1398 
1399 				struct mbuf *mm;
1400 				struct mrt6msg *im;
1401 #ifdef MRT6_OINIT
1402 				struct omrt6msg *oim;
1403 #endif
1404 
1405 				mm = m_copym(m, 0, sizeof(struct ip6_hdr),
1406 				    M_NOWAIT);
1407 				if (mm &&
1408 				    (!M_WRITABLE(mm) ||
1409 				     mm->m_len < sizeof(struct ip6_hdr)))
1410 					mm = m_pullup(mm, sizeof(struct ip6_hdr));
1411 				if (mm == NULL)
1412 					return (ENOBUFS);
1413 
1414 #ifdef MRT6_OINIT
1415 				oim = NULL;
1416 #endif
1417 				im = NULL;
1418 				switch (V_ip6_mrouter_ver) {
1419 #ifdef MRT6_OINIT
1420 				case MRT6_OINIT:
1421 					oim = mtod(mm, struct omrt6msg *);
1422 					oim->im6_msgtype = MRT6MSG_WRONGMIF;
1423 					oim->im6_mbz = 0;
1424 					break;
1425 #endif
1426 				case MRT6_INIT:
1427 					im = mtod(mm, struct mrt6msg *);
1428 					im->im6_msgtype = MRT6MSG_WRONGMIF;
1429 					im->im6_mbz = 0;
1430 					break;
1431 				default:
1432 					m_freem(mm);
1433 					return (EINVAL);
1434 				}
1435 
1436 				for (mifp = mif6table, iif = 0;
1437 				     iif < nummifs && mifp &&
1438 					     mifp->m6_ifp != ifp;
1439 				     mifp++, iif++)
1440 					;
1441 
1442 				switch (V_ip6_mrouter_ver) {
1443 #ifdef MRT6_OINIT
1444 				case MRT6_OINIT:
1445 					oim->im6_mif = iif;
1446 					sin6.sin6_addr = oim->im6_src;
1447 					break;
1448 #endif
1449 				case MRT6_INIT:
1450 					im->im6_mif = iif;
1451 					sin6.sin6_addr = im->im6_src;
1452 					break;
1453 				}
1454 
1455 				MRT6STAT_INC(mrt6s_upcalls);
1456 
1457 				if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1458 					MRT6_DLOG(DEBUG_ANY,
1459 					    "ip6_mrouter socket queue full");
1460 					MRT6STAT_INC(mrt6s_upq_sockfull);
1461 					return (ENOBUFS);
1462 				}	/* if socket Q full */
1463 			}		/* if PIM */
1464 		return (0);
1465 	}			/* if wrong iif */
1466 
1467 	/* If I sourced this packet, it counts as output, else it was input. */
1468 	if (m->m_pkthdr.rcvif == NULL) {
1469 		/* XXX: is rcvif really NULL when output?? */
1470 		mif6table[mifi].m6_pkt_out++;
1471 		mif6table[mifi].m6_bytes_out += plen;
1472 	} else {
1473 		mif6table[mifi].m6_pkt_in++;
1474 		mif6table[mifi].m6_bytes_in += plen;
1475 	}
1476 	rt->mf6c_pkt_cnt++;
1477 	rt->mf6c_byte_cnt += plen;
1478 
1479 	/*
1480 	 * For each mif, forward a copy of the packet if there are group
1481 	 * members downstream on the interface.
1482 	 */
1483 	src0 = ip6->ip6_src;
1484 	dst0 = ip6->ip6_dst;
1485 	if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 ||
1486 	    (error = in6_setscope(&dst0, ifp, &idzone)) != 0) {
1487 		IP6STAT_INC(ip6s_badscope);
1488 		return (error);
1489 	}
1490 	for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) {
1491 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1492 			/*
1493 			 * check if the outgoing packet is going to break
1494 			 * a scope boundary.
1495 			 * XXX For packets through PIM register tunnel
1496 			 * interface, we believe a routing daemon.
1497 			 */
1498 			if (!(mif6table[rt->mf6c_parent].m6_flags &
1499 			      MIFF_REGISTER) &&
1500 			    !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
1501 				if (in6_setscope(&src0, mif6table[mifi].m6_ifp,
1502 				    &oszone) ||
1503 				    in6_setscope(&dst0, mif6table[mifi].m6_ifp,
1504 				    &odzone) ||
1505 				    iszone != oszone ||
1506 				    idzone != odzone) {
1507 					IP6STAT_INC(ip6s_badscope);
1508 					continue;
1509 				}
1510 			}
1511 
1512 			mifp->m6_pkt_out++;
1513 			mifp->m6_bytes_out += plen;
1514 			if (mifp->m6_flags & MIFF_REGISTER)
1515 				register_send(ip6, mifp, m);
1516 			else
1517 				phyint_send(ip6, mifp, m);
1518 		}
1519 	}
1520 	return (0);
1521 }
1522 
1523 static void
1524 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1525 {
1526 #ifdef MRT6DEBUG
1527 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1528 #endif
1529 	struct mbuf *mb_copy;
1530 	struct ifnet *ifp = mifp->m6_ifp;
1531 	int error __unused = 0;
1532 	u_long linkmtu;
1533 
1534 	/*
1535 	 * Make a new reference to the packet; make sure that
1536 	 * the IPv6 header is actually copied, not just referenced,
1537 	 * so that ip6_output() only scribbles on the copy.
1538 	 */
1539 	mb_copy = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1540 	if (mb_copy &&
1541 	    (!M_WRITABLE(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1542 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1543 	if (mb_copy == NULL) {
1544 		return;
1545 	}
1546 	/* set MCAST flag to the outgoing packet */
1547 	mb_copy->m_flags |= M_MCAST;
1548 
1549 	/*
1550 	 * If we sourced the packet, call ip6_output since we may devide
1551 	 * the packet into fragments when the packet is too big for the
1552 	 * outgoing interface.
1553 	 * Otherwise, we can simply send the packet to the interface
1554 	 * sending queue.
1555 	 */
1556 	if (m->m_pkthdr.rcvif == NULL) {
1557 		struct ip6_moptions im6o;
1558 		struct epoch_tracker et;
1559 
1560 		im6o.im6o_multicast_ifp = ifp;
1561 		/* XXX: ip6_output will override ip6->ip6_hlim */
1562 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1563 		im6o.im6o_multicast_loop = 1;
1564 		NET_EPOCH_ENTER(et);
1565 		error = ip6_output(mb_copy, NULL, NULL, IPV6_FORWARDING, &im6o,
1566 		    NULL, NULL);
1567 		NET_EPOCH_EXIT(et);
1568 
1569 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1570 		    (uint16_t)(mifp - mif6table), error);
1571 		return;
1572 	}
1573 
1574 	/*
1575 	 * If configured to loop back multicasts by default,
1576 	 * loop back a copy now.
1577 	 */
1578 	if (in6_mcast_loop)
1579 		ip6_mloopback(ifp, m);
1580 
1581 	/*
1582 	 * Put the packet into the sending queue of the outgoing interface
1583 	 * if it would fit in the MTU of the interface.
1584 	 */
1585 	linkmtu = IN6_LINKMTU(ifp);
1586 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1587 		struct sockaddr_in6 dst6;
1588 
1589 		bzero(&dst6, sizeof(dst6));
1590 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1591 		dst6.sin6_family = AF_INET6;
1592 		dst6.sin6_addr = ip6->ip6_dst;
1593 
1594 		IP_PROBE(send, NULL, NULL, ip6, ifp, NULL, ip6);
1595 		/*
1596 		 * We just call if_output instead of nd6_output here, since
1597 		 * we need no ND for a multicast forwarded packet...right?
1598 		 */
1599 		m_clrprotoflags(m);	/* Avoid confusing lower layers. */
1600 		error = (*ifp->if_output)(ifp, mb_copy,
1601 		    (struct sockaddr *)&dst6, NULL);
1602 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1603 		    (uint16_t)(mifp - mif6table), error);
1604 	} else {
1605 		/*
1606 		 * pMTU discovery is intentionally disabled by default, since
1607 		 * various router may notify pMTU in multicast, which can be
1608 		 * a DDoS to a router
1609 		 */
1610 		if (V_ip6_mcast_pmtu)
1611 			icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1612 		else {
1613 			MRT6_DLOG(DEBUG_XMIT, " packet too big on %s o %s "
1614 			    "g %s size %d (discarded)", if_name(ifp),
1615 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1616 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1617 			    mb_copy->m_pkthdr.len);
1618 			m_freem(mb_copy); /* simply discard the packet */
1619 		}
1620 	}
1621 }
1622 
1623 static int
1624 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1625 {
1626 #ifdef MRT6DEBUG
1627 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1628 #endif
1629 	struct mbuf *mm;
1630 	int i, len = m->m_pkthdr.len;
1631 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1632 	struct mrt6msg *im6;
1633 
1634 	MRT6_DLOG(DEBUG_ANY, "src %s dst %s",
1635 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1636 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1637 	PIM6STAT_INC(pim6s_snd_registers);
1638 
1639 	/* Make a copy of the packet to send to the user level process. */
1640 	mm = m_gethdr(M_NOWAIT, MT_DATA);
1641 	if (mm == NULL)
1642 		return (ENOBUFS);
1643 	mm->m_data += max_linkhdr;
1644 	mm->m_len = sizeof(struct ip6_hdr);
1645 
1646 	if ((mm->m_next = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
1647 		m_freem(mm);
1648 		return (ENOBUFS);
1649 	}
1650 	i = MHLEN - M_LEADINGSPACE(mm);
1651 	if (i > len)
1652 		i = len;
1653 	mm = m_pullup(mm, i);
1654 	if (mm == NULL)
1655 		return (ENOBUFS);
1656 /* TODO: check it! */
1657 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1658 
1659 	/*
1660 	 * Send message to routing daemon
1661 	 */
1662 	sin6.sin6_addr = ip6->ip6_src;
1663 
1664 	im6 = mtod(mm, struct mrt6msg *);
1665 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1666 	im6->im6_mbz          = 0;
1667 
1668 	im6->im6_mif = mif - mif6table;
1669 
1670 	/* iif info is not given for reg. encap.n */
1671 	MRT6STAT_INC(mrt6s_upcalls);
1672 
1673 	if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) {
1674 		MRT6_DLOG(DEBUG_ANY, "ip6_mrouter socket queue full");
1675 		MRT6STAT_INC(mrt6s_upq_sockfull);
1676 		return (ENOBUFS);
1677 	}
1678 	return (0);
1679 }
1680 
1681 /*
1682  * pim6_encapcheck() is called by the encap6_input() path at runtime to
1683  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
1684  * into the kernel.
1685  */
1686 static int
1687 pim6_encapcheck(const struct mbuf *m __unused, int off __unused,
1688     int proto __unused, void *arg __unused)
1689 {
1690 
1691     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
1692     return (8);		/* claim the datagram. */
1693 }
1694 
1695 /*
1696  * PIM sparse mode hook
1697  * Receives the pim control messages, and passes them up to the listening
1698  * socket, using rip6_input.
1699  * The only message processed is the REGISTER pim message; the pim header
1700  * is stripped off, and the inner packet is passed to register_mforward.
1701  */
1702 static int
1703 pim6_input(struct mbuf *m, int off, int proto, void *arg __unused)
1704 {
1705 	struct pim *pim; /* pointer to a pim struct */
1706 	struct ip6_hdr *ip6;
1707 	int pimlen;
1708 	int minlen;
1709 
1710 	PIM6STAT_INC(pim6s_rcv_total);
1711 
1712 	/*
1713 	 * Validate lengths
1714 	 */
1715 	pimlen = m->m_pkthdr.len - off;
1716 	if (pimlen < PIM_MINLEN) {
1717 		PIM6STAT_INC(pim6s_rcv_tooshort);
1718 		MRT6_DLOG(DEBUG_PIM, "PIM packet too short");
1719 		m_freem(m);
1720 		return (IPPROTO_DONE);
1721 	}
1722 
1723 	/*
1724 	 * if the packet is at least as big as a REGISTER, go ahead
1725 	 * and grab the PIM REGISTER header size, to avoid another
1726 	 * possible m_pullup() later.
1727 	 *
1728 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1729 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1730 	 */
1731 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1732 
1733 	/*
1734 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1735 	 * possibly the PIM REGISTER header
1736 	 */
1737 	if (m->m_len < off + minlen) {
1738 		m = m_pullup(m, off + minlen);
1739 		if (m == NULL) {
1740 			IP6STAT_INC(ip6s_exthdrtoolong);
1741 			return (IPPROTO_DONE);
1742 		}
1743 	}
1744 	ip6 = mtod(m, struct ip6_hdr *);
1745 	pim = (struct pim *)((caddr_t)ip6 + off);
1746 
1747 #define PIM6_CHECKSUM
1748 #ifdef PIM6_CHECKSUM
1749 	{
1750 		int cksumlen;
1751 
1752 		/*
1753 		 * Validate checksum.
1754 		 * If PIM REGISTER, exclude the data packet
1755 		 */
1756 		if (pim->pim_type == PIM_REGISTER)
1757 			cksumlen = PIM_MINLEN;
1758 		else
1759 			cksumlen = pimlen;
1760 
1761 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1762 			PIM6STAT_INC(pim6s_rcv_badsum);
1763 			MRT6_DLOG(DEBUG_PIM, "invalid checksum");
1764 			m_freem(m);
1765 			return (IPPROTO_DONE);
1766 		}
1767 	}
1768 #endif /* PIM_CHECKSUM */
1769 
1770 	/* PIM version check */
1771 	if (pim->pim_ver != PIM_VERSION) {
1772 		PIM6STAT_INC(pim6s_rcv_badversion);
1773 		MRT6_DLOG(DEBUG_ANY | DEBUG_ERR,
1774 		    "incorrect version %d, expecting %d",
1775 		    pim->pim_ver, PIM_VERSION);
1776 		m_freem(m);
1777 		return (IPPROTO_DONE);
1778 	}
1779 
1780 	if (pim->pim_type == PIM_REGISTER) {
1781 		/*
1782 		 * since this is a REGISTER, we'll make a copy of the register
1783 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1784 		 * routing daemon.
1785 		 */
1786 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1787 
1788 		struct mbuf *mcp;
1789 		struct ip6_hdr *eip6;
1790 		u_int32_t *reghdr;
1791 #ifdef MRT6DEBUG
1792 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1793 #endif
1794 
1795 		PIM6STAT_INC(pim6s_rcv_registers);
1796 
1797 		if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1798 			MRT6_DLOG(DEBUG_PIM, "register mif not set: %d",
1799 			    reg_mif_num);
1800 			m_freem(m);
1801 			return (IPPROTO_DONE);
1802 		}
1803 
1804 		reghdr = (u_int32_t *)(pim + 1);
1805 
1806 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1807 			goto pim6_input_to_daemon;
1808 
1809 		/*
1810 		 * Validate length
1811 		 */
1812 		if (pimlen < PIM6_REG_MINLEN) {
1813 			PIM6STAT_INC(pim6s_rcv_tooshort);
1814 			PIM6STAT_INC(pim6s_rcv_badregisters);
1815 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "register packet "
1816 			    "size too small %d from %s",
1817 			    pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src));
1818 			m_freem(m);
1819 			return (IPPROTO_DONE);
1820 		}
1821 
1822 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1823 		MRT6_DLOG(DEBUG_PIM, "eip6: %s -> %s, eip6 plen %d",
1824 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1825 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1826 		    ntohs(eip6->ip6_plen));
1827 
1828 		/* verify the version number of the inner packet */
1829 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1830 			PIM6STAT_INC(pim6s_rcv_badregisters);
1831 			MRT6_DLOG(DEBUG_ANY, "invalid IP version (%d) "
1832 			    "of the inner packet",
1833 			    (eip6->ip6_vfc & IPV6_VERSION));
1834 			m_freem(m);
1835 			return (IPPROTO_DONE);
1836 		}
1837 
1838 		/* verify the inner packet is destined to a mcast group */
1839 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1840 			PIM6STAT_INC(pim6s_rcv_badregisters);
1841 			MRT6_DLOG(DEBUG_PIM, "inner packet of register "
1842 			    "is not multicast %s",
1843 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst));
1844 			m_freem(m);
1845 			return (IPPROTO_DONE);
1846 		}
1847 
1848 		/*
1849 		 * make a copy of the whole header to pass to the daemon later.
1850 		 */
1851 		mcp = m_copym(m, 0, off + PIM6_REG_MINLEN, M_NOWAIT);
1852 		if (mcp == NULL) {
1853 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "pim register: "
1854 			    "could not copy register head");
1855 			m_freem(m);
1856 			return (IPPROTO_DONE);
1857 		}
1858 
1859 		/*
1860 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1861 		 */
1862 		m_adj(m, off + PIM_MINLEN);
1863 		MRT6_DLOG(DEBUG_PIM, "forwarding decapsulated register: "
1864 		    "src %s, dst %s, mif %d",
1865 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1866 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst), reg_mif_num);
1867 
1868 		if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1869 				dst.sin6_family, 0);
1870 
1871 		/* prepare the register head to send to the mrouting daemon */
1872 		m = mcp;
1873 	}
1874 
1875 	/*
1876 	 * Pass the PIM message up to the daemon; if it is a register message
1877 	 * pass the 'head' only up to the daemon. This includes the
1878 	 * encapsulator ip6 header, pim header, register header and the
1879 	 * encapsulated ip6 header.
1880 	 */
1881   pim6_input_to_daemon:
1882 	return (rip6_input(&m, &off, proto));
1883 }
1884 
1885 static int
1886 ip6_mroute_modevent(module_t mod, int type, void *unused)
1887 {
1888 
1889 	switch (type) {
1890 	case MOD_LOAD:
1891 		MROUTER6_LOCK_INIT();
1892 		MFC6_LOCK_INIT();
1893 		MIF6_LOCK_INIT();
1894 
1895 		pim6_encap_cookie = ip6_encap_attach(&ipv6_encap_cfg,
1896 		    NULL, M_WAITOK);
1897 		if (pim6_encap_cookie == NULL) {
1898 			printf("ip6_mroute: unable to attach pim6 encap\n");
1899 			MIF6_LOCK_DESTROY();
1900 			MFC6_LOCK_DESTROY();
1901 			MROUTER6_LOCK_DESTROY();
1902 			return (EINVAL);
1903 		}
1904 
1905 		ip6_mforward = X_ip6_mforward;
1906 		ip6_mrouter_done = X_ip6_mrouter_done;
1907 		ip6_mrouter_get = X_ip6_mrouter_get;
1908 		ip6_mrouter_set = X_ip6_mrouter_set;
1909 		mrt6_ioctl = X_mrt6_ioctl;
1910 		break;
1911 
1912 	case MOD_UNLOAD:
1913 		if (V_ip6_mrouter != NULL)
1914 			return EINVAL;
1915 
1916 		if (pim6_encap_cookie) {
1917 			ip6_encap_detach(pim6_encap_cookie);
1918 			pim6_encap_cookie = NULL;
1919 		}
1920 		X_ip6_mrouter_done();
1921 		ip6_mforward = NULL;
1922 		ip6_mrouter_done = NULL;
1923 		ip6_mrouter_get = NULL;
1924 		ip6_mrouter_set = NULL;
1925 		mrt6_ioctl = NULL;
1926 
1927 		MIF6_LOCK_DESTROY();
1928 		MFC6_LOCK_DESTROY();
1929 		MROUTER6_LOCK_DESTROY();
1930 		break;
1931 
1932 	default:
1933 		return (EOPNOTSUPP);
1934 	}
1935 
1936 	return (0);
1937 }
1938 
1939 static moduledata_t ip6_mroutemod = {
1940 	"ip6_mroute",
1941 	ip6_mroute_modevent,
1942 	0
1943 };
1944 
1945 DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_ANY);
1946