xref: /freebsd/sys/netinet/ip_mroute.c (revision 4cf49a43559ed9fdad601bdcccd2c55963008675)
1 /*
2  * IP multicast forwarding procedures
3  *
4  * Written by David Waitzman, BBN Labs, August 1988.
5  * Modified by Steve Deering, Stanford, February 1989.
6  * Modified by Mark J. Steiglitz, Stanford, May, 1991
7  * Modified by Van Jacobson, LBL, January 1993
8  * Modified by Ajit Thyagarajan, PARC, August 1993
9  * Modified by Bill Fenner, PARC, April 1995
10  *
11  * MROUTING Revision: 3.5
12  * $FreeBSD$
13  */
14 
15 #include "opt_mrouting.h"
16 
17 #include <sys/param.h>
18 #include <sys/systm.h>
19 #include <sys/malloc.h>
20 #include <sys/mbuf.h>
21 #include <sys/socket.h>
22 #include <sys/socketvar.h>
23 #include <sys/protosw.h>
24 #include <sys/time.h>
25 #include <sys/kernel.h>
26 #include <sys/sockio.h>
27 #include <sys/syslog.h>
28 #include <net/if.h>
29 #include <net/route.h>
30 #include <netinet/in.h>
31 #include <netinet/in_systm.h>
32 #include <netinet/ip.h>
33 #include <netinet/ip_var.h>
34 #include <netinet/in_var.h>
35 #include <netinet/igmp.h>
36 #include <netinet/ip_mroute.h>
37 #include <netinet/udp.h>
38 
39 #ifndef NTOHL
40 #if BYTE_ORDER != BIG_ENDIAN
41 #define NTOHL(d) ((d) = ntohl((d)))
42 #define NTOHS(d) ((d) = ntohs((u_short)(d)))
43 #define HTONL(d) ((d) = htonl((d)))
44 #define HTONS(d) ((d) = htons((u_short)(d)))
45 #else
46 #define NTOHL(d)
47 #define NTOHS(d)
48 #define HTONL(d)
49 #define HTONS(d)
50 #endif
51 #endif
52 
53 #ifndef MROUTING
54 extern u_long	_ip_mcast_src __P((int vifi));
55 extern int	_ip_mforward __P((struct ip *ip, struct ifnet *ifp,
56 				  struct mbuf *m, struct ip_moptions *imo));
57 extern int	_ip_mrouter_done __P((void));
58 extern int	_ip_mrouter_get __P((struct socket *so, struct sockopt *sopt));
59 extern int	_ip_mrouter_set __P((struct socket *so, struct sockopt *sopt));
60 extern int	_mrt_ioctl __P((int req, caddr_t data, struct proc *p));
61 
62 /*
63  * Dummy routines and globals used when multicast routing is not compiled in.
64  */
65 
66 struct socket  *ip_mrouter  = NULL;
67 u_int		rsvpdebug = 0;
68 
69 int
70 _ip_mrouter_set(so, sopt)
71 	struct socket *so;
72 	struct sockopt *sopt;
73 {
74 	return(EOPNOTSUPP);
75 }
76 
77 int (*ip_mrouter_set)(struct socket *, struct sockopt *) = _ip_mrouter_set;
78 
79 
80 int
81 _ip_mrouter_get(so, sopt)
82 	struct socket *so;
83 	struct sockopt *sopt;
84 {
85 	return(EOPNOTSUPP);
86 }
87 
88 int (*ip_mrouter_get)(struct socket *, struct sockopt *) = _ip_mrouter_get;
89 
90 int
91 _ip_mrouter_done()
92 {
93 	return(0);
94 }
95 
96 int (*ip_mrouter_done)(void) = _ip_mrouter_done;
97 
98 int
99 _ip_mforward(ip, ifp, m, imo)
100 	struct ip *ip;
101 	struct ifnet *ifp;
102 	struct mbuf *m;
103 	struct ip_moptions *imo;
104 {
105 	return(0);
106 }
107 
108 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
109 		   struct ip_moptions *) = _ip_mforward;
110 
111 int
112 _mrt_ioctl(int req, caddr_t data, struct proc *p)
113 {
114 	return EOPNOTSUPP;
115 }
116 
117 int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl;
118 
119 void
120 rsvp_input(m, iphlen)		/* XXX must fixup manually */
121 	struct mbuf *m;
122 	int iphlen;
123 {
124     /* Can still get packets with rsvp_on = 0 if there is a local member
125      * of the group to which the RSVP packet is addressed.  But in this
126      * case we want to throw the packet away.
127      */
128     if (!rsvp_on) {
129 	m_freem(m);
130 	return;
131     }
132 
133     if (ip_rsvpd != NULL) {
134 	if (rsvpdebug)
135 	    printf("rsvp_input: Sending packet up old-style socket\n");
136 	rip_input(m, iphlen);
137 	return;
138     }
139     /* Drop the packet */
140     m_freem(m);
141 }
142 
143 void ipip_input(struct mbuf *m, int iphlen) { /* XXX must fixup manually */
144 	rip_input(m, iphlen);
145 }
146 
147 int (*legal_vif_num)(int) = 0;
148 
149 /*
150  * This should never be called, since IP_MULTICAST_VIF should fail, but
151  * just in case it does get called, the code a little lower in ip_output
152  * will assign the packet a local address.
153  */
154 u_long
155 _ip_mcast_src(int vifi) { return INADDR_ANY; }
156 u_long (*ip_mcast_src)(int) = _ip_mcast_src;
157 
158 int
159 ip_rsvp_vif_init(so, sopt)
160     struct socket *so;
161     struct sockopt *sopt;
162 {
163     return(EINVAL);
164 }
165 
166 int
167 ip_rsvp_vif_done(so, sopt)
168     struct socket *so;
169     struct sockopt *sopt;
170 {
171     return(EINVAL);
172 }
173 
174 void
175 ip_rsvp_force_done(so)
176     struct socket *so;
177 {
178     return;
179 }
180 
181 #else /* MROUTING */
182 
183 #define M_HASCL(m)	((m)->m_flags & M_EXT)
184 
185 #define INSIZ		sizeof(struct in_addr)
186 #define	same(a1, a2) \
187 	(bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0)
188 
189 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables");
190 
191 /*
192  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
193  * except for netstat or debugging purposes.
194  */
195 #ifndef MROUTE_LKM
196 struct socket  *ip_mrouter  = NULL;
197 static struct mrtstat	mrtstat;
198 #else /* MROUTE_LKM */
199 extern void	X_ipip_input __P((struct mbuf *m, int iphlen));
200 extern struct mrtstat mrtstat;
201 static int ip_mrtproto;
202 #endif
203 
204 #define NO_RTE_FOUND 	0x1
205 #define RTE_FOUND	0x2
206 
207 static struct mfc	*mfctable[MFCTBLSIZ];
208 static u_char		nexpire[MFCTBLSIZ];
209 static struct vif	viftable[MAXVIFS];
210 static u_int	mrtdebug = 0;	  /* debug level 	*/
211 #define		DEBUG_MFC	0x02
212 #define		DEBUG_FORWARD	0x04
213 #define		DEBUG_EXPIRE	0x08
214 #define		DEBUG_XMIT	0x10
215 static u_int  	tbfdebug = 0;     /* tbf debug level 	*/
216 static u_int	rsvpdebug = 0;	  /* rsvp debug level   */
217 
218 static struct callout_handle expire_upcalls_ch;
219 
220 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
221 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
222 
223 /*
224  * Define the token bucket filter structures
225  * tbftable -> each vif has one of these for storing info
226  */
227 
228 static struct tbf tbftable[MAXVIFS];
229 #define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
230 
231 /*
232  * 'Interfaces' associated with decapsulator (so we can tell
233  * packets that went through it from ones that get reflected
234  * by a broken gateway).  These interfaces are never linked into
235  * the system ifnet list & no routes point to them.  I.e., packets
236  * can't be sent this way.  They only exist as a placeholder for
237  * multicast source verification.
238  */
239 static struct ifnet multicast_decap_if[MAXVIFS];
240 
241 #define ENCAP_TTL 64
242 #define ENCAP_PROTO IPPROTO_IPIP	/* 4 */
243 
244 /* prototype IP hdr for encapsulated packets */
245 static struct ip multicast_encap_iphdr = {
246 #if BYTE_ORDER == LITTLE_ENDIAN
247 	sizeof(struct ip) >> 2, IPVERSION,
248 #else
249 	IPVERSION, sizeof(struct ip) >> 2,
250 #endif
251 	0,				/* tos */
252 	sizeof(struct ip),		/* total length */
253 	0,				/* id */
254 	0,				/* frag offset */
255 	ENCAP_TTL, ENCAP_PROTO,
256 	0,				/* checksum */
257 };
258 
259 /*
260  * Private variables.
261  */
262 static vifi_t	   numvifs = 0;
263 static int have_encap_tunnel = 0;
264 
265 /*
266  * one-back cache used by ipip_input to locate a tunnel's vif
267  * given a datagram's src ip address.
268  */
269 static u_long last_encap_src;
270 static struct vif *last_encap_vif;
271 
272 static u_long	X_ip_mcast_src __P((int vifi));
273 static int	X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo));
274 static int	X_ip_mrouter_done __P((void));
275 static int	X_ip_mrouter_get __P((struct socket *so, struct sockopt *m));
276 static int	X_ip_mrouter_set __P((struct socket *so, struct sockopt *m));
277 static int	X_legal_vif_num __P((int vif));
278 static int	X_mrt_ioctl __P((int cmd, caddr_t data));
279 
280 static int get_sg_cnt(struct sioc_sg_req *);
281 static int get_vif_cnt(struct sioc_vif_req *);
282 static int ip_mrouter_init(struct socket *, int);
283 static int add_vif(struct vifctl *);
284 static int del_vif(vifi_t);
285 static int add_mfc(struct mfcctl *);
286 static int del_mfc(struct mfcctl *);
287 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
288 static int set_assert(int);
289 static void expire_upcalls(void *);
290 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *,
291 		  vifi_t);
292 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
293 static void encap_send(struct ip *, struct vif *, struct mbuf *);
294 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
295 static void tbf_queue(struct vif *, struct mbuf *);
296 static void tbf_process_q(struct vif *);
297 static void tbf_reprocess_q(void *);
298 static int tbf_dq_sel(struct vif *, struct ip *);
299 static void tbf_send_packet(struct vif *, struct mbuf *);
300 static void tbf_update_tokens(struct vif *);
301 static int priority(struct vif *, struct ip *);
302 void multiencap_decap(struct mbuf *);
303 
304 /*
305  * whether or not special PIM assert processing is enabled.
306  */
307 static int pim_assert;
308 /*
309  * Rate limit for assert notification messages, in usec
310  */
311 #define ASSERT_MSG_TIME		3000000
312 
313 /*
314  * Hash function for a source, group entry
315  */
316 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
317 			((g) >> 20) ^ ((g) >> 10) ^ (g))
318 
319 /*
320  * Find a route for a given origin IP address and Multicast group address
321  * Type of service parameter to be added in the future!!!
322  */
323 
324 #define MFCFIND(o, g, rt) { \
325 	register struct mfc *_rt = mfctable[MFCHASH(o,g)]; \
326 	rt = NULL; \
327 	++mrtstat.mrts_mfc_lookups; \
328 	while (_rt) { \
329 		if ((_rt->mfc_origin.s_addr == o) && \
330 		    (_rt->mfc_mcastgrp.s_addr == g) && \
331 		    (_rt->mfc_stall == NULL)) { \
332 			rt = _rt; \
333 			break; \
334 		} \
335 		_rt = _rt->mfc_next; \
336 	} \
337 	if (rt == NULL) { \
338 		++mrtstat.mrts_mfc_misses; \
339 	} \
340 }
341 
342 
343 /*
344  * Macros to compute elapsed time efficiently
345  * Borrowed from Van Jacobson's scheduling code
346  */
347 #define TV_DELTA(a, b, delta) { \
348 	    register int xxs; \
349 		\
350 	    delta = (a).tv_usec - (b).tv_usec; \
351 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
352 	       switch (xxs) { \
353 		      case 2: \
354 			  delta += 1000000; \
355 			      /* fall through */ \
356 		      case 1: \
357 			  delta += 1000000; \
358 			  break; \
359 		      default: \
360 			  delta += (1000000 * xxs); \
361 	       } \
362 	    } \
363 }
364 
365 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
366 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
367 
368 #ifdef UPCALL_TIMING
369 u_long upcall_data[51];
370 static void collate(struct timeval *);
371 #endif /* UPCALL_TIMING */
372 
373 
374 /*
375  * Handle MRT setsockopt commands to modify the multicast routing tables.
376  */
377 static int
378 X_ip_mrouter_set(so, sopt)
379 	struct socket *so;
380 	struct sockopt *sopt;
381 {
382 	int	error, optval;
383 	vifi_t	vifi;
384 	struct	vifctl vifc;
385 	struct	mfcctl mfc;
386 
387 	if (so != ip_mrouter && sopt->sopt_name != MRT_INIT)
388 		return (EPERM);
389 
390 	error = 0;
391 	switch (sopt->sopt_name) {
392 	case MRT_INIT:
393 		error = sooptcopyin(sopt, &optval, sizeof optval,
394 				    sizeof optval);
395 		if (error)
396 			break;
397 		error = ip_mrouter_init(so, optval);
398 		break;
399 
400 	case MRT_DONE:
401 		error = ip_mrouter_done();
402 		break;
403 
404 	case MRT_ADD_VIF:
405 		error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
406 		if (error)
407 			break;
408 		error = add_vif(&vifc);
409 		break;
410 
411 	case MRT_DEL_VIF:
412 		error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
413 		if (error)
414 			break;
415 		error = del_vif(vifi);
416 		break;
417 
418 	case MRT_ADD_MFC:
419 	case MRT_DEL_MFC:
420 		error = sooptcopyin(sopt, &mfc, sizeof mfc, sizeof mfc);
421 		if (error)
422 			break;
423 		if (sopt->sopt_name == MRT_ADD_MFC)
424 			error = add_mfc(&mfc);
425 		else
426 			error = del_mfc(&mfc);
427 		break;
428 
429 	case MRT_ASSERT:
430 		error = sooptcopyin(sopt, &optval, sizeof optval,
431 				    sizeof optval);
432 		if (error)
433 			break;
434 		set_assert(optval);
435 		break;
436 
437 	default:
438 		error = EOPNOTSUPP;
439 		break;
440 	}
441 	return (error);
442 }
443 
444 #ifndef MROUTE_LKM
445 int (*ip_mrouter_set)(struct socket *, struct sockopt *) = X_ip_mrouter_set;
446 #endif
447 
448 /*
449  * Handle MRT getsockopt commands
450  */
451 static int
452 X_ip_mrouter_get(so, sopt)
453 	struct socket *so;
454 	struct sockopt *sopt;
455 {
456 	int error;
457 	static int version = 0x0305; /* !!! why is this here? XXX */
458 
459 	switch (sopt->sopt_name) {
460 	case MRT_VERSION:
461 		error = sooptcopyout(sopt, &version, sizeof version);
462 		break;
463 
464 	case MRT_ASSERT:
465 		error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert);
466 		break;
467 	default:
468 		error = EOPNOTSUPP;
469 		break;
470 	}
471 	return (error);
472 }
473 
474 #ifndef MROUTE_LKM
475 int (*ip_mrouter_get)(struct socket *, struct sockopt *) = X_ip_mrouter_get;
476 #endif
477 
478 /*
479  * Handle ioctl commands to obtain information from the cache
480  */
481 static int
482 X_mrt_ioctl(cmd, data)
483     int cmd;
484     caddr_t data;
485 {
486     int error = 0;
487 
488     switch (cmd) {
489 	case (SIOCGETVIFCNT):
490 	    return (get_vif_cnt((struct sioc_vif_req *)data));
491 	    break;
492 	case (SIOCGETSGCNT):
493 	    return (get_sg_cnt((struct sioc_sg_req *)data));
494 	    break;
495 	default:
496 	    return (EINVAL);
497 	    break;
498     }
499     return error;
500 }
501 
502 #ifndef MROUTE_LKM
503 int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl;
504 #endif
505 
506 /*
507  * returns the packet, byte, rpf-failure count for the source group provided
508  */
509 static int
510 get_sg_cnt(req)
511     register struct sioc_sg_req *req;
512 {
513     register struct mfc *rt;
514     int s;
515 
516     s = splnet();
517     MFCFIND(req->src.s_addr, req->grp.s_addr, rt);
518     splx(s);
519     if (rt != NULL) {
520 	req->pktcnt = rt->mfc_pkt_cnt;
521 	req->bytecnt = rt->mfc_byte_cnt;
522 	req->wrong_if = rt->mfc_wrong_if;
523     } else
524 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
525 
526     return 0;
527 }
528 
529 /*
530  * returns the input and output packet and byte counts on the vif provided
531  */
532 static int
533 get_vif_cnt(req)
534     register struct sioc_vif_req *req;
535 {
536     register vifi_t vifi = req->vifi;
537 
538     if (vifi >= numvifs) return EINVAL;
539 
540     req->icount = viftable[vifi].v_pkt_in;
541     req->ocount = viftable[vifi].v_pkt_out;
542     req->ibytes = viftable[vifi].v_bytes_in;
543     req->obytes = viftable[vifi].v_bytes_out;
544 
545     return 0;
546 }
547 
548 /*
549  * Enable multicast routing
550  */
551 static int
552 ip_mrouter_init(so, version)
553 	struct socket *so;
554 	int version;
555 {
556     if (mrtdebug)
557 	log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
558 		so->so_type, so->so_proto->pr_protocol);
559 
560     if (so->so_type != SOCK_RAW ||
561 	so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP;
562 
563     if (version != 1)
564 	return ENOPROTOOPT;
565 
566     if (ip_mrouter != NULL) return EADDRINUSE;
567 
568     ip_mrouter = so;
569 
570     bzero((caddr_t)mfctable, sizeof(mfctable));
571     bzero((caddr_t)nexpire, sizeof(nexpire));
572 
573     pim_assert = 0;
574 
575     expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
576 
577     if (mrtdebug)
578 	log(LOG_DEBUG, "ip_mrouter_init\n");
579 
580     return 0;
581 }
582 
583 /*
584  * Disable multicast routing
585  */
586 static int
587 X_ip_mrouter_done()
588 {
589     vifi_t vifi;
590     int i;
591     struct ifnet *ifp;
592     struct ifreq ifr;
593     struct mfc *rt;
594     struct rtdetq *rte;
595     int s;
596 
597     s = splnet();
598 
599     /*
600      * For each phyint in use, disable promiscuous reception of all IP
601      * multicasts.
602      */
603     for (vifi = 0; vifi < numvifs; vifi++) {
604 	if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
605 	    !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
606 	    ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
607 	    ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr
608 								= INADDR_ANY;
609 	    ifp = viftable[vifi].v_ifp;
610 	    if_allmulti(ifp, 0);
611 	}
612     }
613     bzero((caddr_t)tbftable, sizeof(tbftable));
614     bzero((caddr_t)viftable, sizeof(viftable));
615     numvifs = 0;
616     pim_assert = 0;
617 
618     untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch);
619 
620     /*
621      * Free all multicast forwarding cache entries.
622      */
623     for (i = 0; i < MFCTBLSIZ; i++) {
624 	for (rt = mfctable[i]; rt != NULL; ) {
625 	    struct mfc *nr = rt->mfc_next;
626 
627 	    for (rte = rt->mfc_stall; rte != NULL; ) {
628 		struct rtdetq *n = rte->next;
629 
630 		m_freem(rte->m);
631 		free(rte, M_MRTABLE);
632 		rte = n;
633 	    }
634 	    free(rt, M_MRTABLE);
635 	    rt = nr;
636 	}
637     }
638 
639     bzero((caddr_t)mfctable, sizeof(mfctable));
640 
641     /*
642      * Reset de-encapsulation cache
643      */
644     last_encap_src = 0;
645     last_encap_vif = NULL;
646     have_encap_tunnel = 0;
647 
648     ip_mrouter = NULL;
649 
650     splx(s);
651 
652     if (mrtdebug)
653 	log(LOG_DEBUG, "ip_mrouter_done\n");
654 
655     return 0;
656 }
657 
658 #ifndef MROUTE_LKM
659 int (*ip_mrouter_done)(void) = X_ip_mrouter_done;
660 #endif
661 
662 /*
663  * Set PIM assert processing global
664  */
665 static int
666 set_assert(i)
667 	int i;
668 {
669     if ((i != 1) && (i != 0))
670 	return EINVAL;
671 
672     pim_assert = i;
673 
674     return 0;
675 }
676 
677 /*
678  * Add a vif to the vif table
679  */
680 static int
681 add_vif(vifcp)
682     register struct vifctl *vifcp;
683 {
684     register struct vif *vifp = viftable + vifcp->vifc_vifi;
685     static struct sockaddr_in sin = {sizeof sin, AF_INET};
686     struct ifaddr *ifa;
687     struct ifnet *ifp;
688     int error, s;
689     struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
690 
691     if (vifcp->vifc_vifi >= MAXVIFS)  return EINVAL;
692     if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE;
693 
694     /* Find the interface with an address in AF_INET family */
695     sin.sin_addr = vifcp->vifc_lcl_addr;
696     ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
697     if (ifa == 0) return EADDRNOTAVAIL;
698     ifp = ifa->ifa_ifp;
699 
700     if (vifcp->vifc_flags & VIFF_TUNNEL) {
701 	if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
702 		/*
703 		 * An encapsulating tunnel is wanted.  Tell ipip_input() to
704 		 * start paying attention to encapsulated packets.
705 		 */
706 		if (have_encap_tunnel == 0) {
707 			have_encap_tunnel = 1;
708 			for (s = 0; s < MAXVIFS; ++s) {
709 				multicast_decap_if[s].if_name = "mdecap";
710 				multicast_decap_if[s].if_unit = s;
711 			}
712 		}
713 		/*
714 		 * Set interface to fake encapsulator interface
715 		 */
716 		ifp = &multicast_decap_if[vifcp->vifc_vifi];
717 		/*
718 		 * Prepare cached route entry
719 		 */
720 		bzero(&vifp->v_route, sizeof(vifp->v_route));
721 	} else {
722 	    log(LOG_ERR, "source routed tunnels not supported\n");
723 	    return EOPNOTSUPP;
724 	}
725     } else {
726 	/* Make sure the interface supports multicast */
727 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
728 	    return EOPNOTSUPP;
729 
730 	/* Enable promiscuous reception of all IP multicasts from the if */
731 	s = splnet();
732 	error = if_allmulti(ifp, 1);
733 	splx(s);
734 	if (error)
735 	    return error;
736     }
737 
738     s = splnet();
739     /* define parameters for the tbf structure */
740     vifp->v_tbf = v_tbf;
741     GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
742     vifp->v_tbf->tbf_n_tok = 0;
743     vifp->v_tbf->tbf_q_len = 0;
744     vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
745     vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
746 
747     vifp->v_flags     = vifcp->vifc_flags;
748     vifp->v_threshold = vifcp->vifc_threshold;
749     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
750     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
751     vifp->v_ifp       = ifp;
752     /* scaling up here allows division by 1024 in critical code */
753     vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
754     vifp->v_rsvp_on   = 0;
755     vifp->v_rsvpd     = NULL;
756     /* initialize per vif pkt counters */
757     vifp->v_pkt_in    = 0;
758     vifp->v_pkt_out   = 0;
759     vifp->v_bytes_in  = 0;
760     vifp->v_bytes_out = 0;
761     splx(s);
762 
763     /* Adjust numvifs up if the vifi is higher than numvifs */
764     if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
765 
766     if (mrtdebug)
767 	log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n",
768 	    vifcp->vifc_vifi,
769 	    (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr),
770 	    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
771 	    (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr),
772 	    vifcp->vifc_threshold,
773 	    vifcp->vifc_rate_limit);
774 
775     return 0;
776 }
777 
778 /*
779  * Delete a vif from the vif table
780  */
781 static int
782 del_vif(vifi)
783 	vifi_t vifi;
784 {
785     register struct vif *vifp = &viftable[vifi];
786     register struct mbuf *m;
787     struct ifnet *ifp;
788     struct ifreq ifr;
789     int s;
790 
791     if (vifi >= numvifs) return EINVAL;
792     if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL;
793 
794     s = splnet();
795 
796     if (!(vifp->v_flags & VIFF_TUNNEL)) {
797 	((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
798 	((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
799 	ifp = vifp->v_ifp;
800 	if_allmulti(ifp, 0);
801     }
802 
803     if (vifp == last_encap_vif) {
804 	last_encap_vif = 0;
805 	last_encap_src = 0;
806     }
807 
808     /*
809      * Free packets queued at the interface
810      */
811     while (vifp->v_tbf->tbf_q) {
812 	m = vifp->v_tbf->tbf_q;
813 	vifp->v_tbf->tbf_q = m->m_act;
814 	m_freem(m);
815     }
816 
817     bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
818     bzero((caddr_t)vifp, sizeof (*vifp));
819 
820     if (mrtdebug)
821       log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs);
822 
823     /* Adjust numvifs down */
824     for (vifi = numvifs; vifi > 0; vifi--)
825 	if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break;
826     numvifs = vifi;
827 
828     splx(s);
829 
830     return 0;
831 }
832 
833 /*
834  * Add an mfc entry
835  */
836 static int
837 add_mfc(mfccp)
838     struct mfcctl *mfccp;
839 {
840     struct mfc *rt;
841     u_long hash;
842     struct rtdetq *rte;
843     register u_short nstl;
844     int s;
845     int i;
846 
847     MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt);
848 
849     /* If an entry already exists, just update the fields */
850     if (rt) {
851 	if (mrtdebug & DEBUG_MFC)
852 	    log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n",
853 		(u_long)ntohl(mfccp->mfcc_origin.s_addr),
854 		(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
855 		mfccp->mfcc_parent);
856 
857 	s = splnet();
858 	rt->mfc_parent = mfccp->mfcc_parent;
859 	for (i = 0; i < numvifs; i++)
860 	    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
861 	splx(s);
862 	return 0;
863     }
864 
865     /*
866      * Find the entry for which the upcall was made and update
867      */
868     s = splnet();
869     hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
870     for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) {
871 
872 	if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
873 	    (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
874 	    (rt->mfc_stall != NULL)) {
875 
876 	    if (nstl++)
877 		log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n",
878 		    "multiple kernel entries",
879 		    (u_long)ntohl(mfccp->mfcc_origin.s_addr),
880 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
881 		    mfccp->mfcc_parent, (void *)rt->mfc_stall);
882 
883 	    if (mrtdebug & DEBUG_MFC)
884 		log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n",
885 		    (u_long)ntohl(mfccp->mfcc_origin.s_addr),
886 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
887 		    mfccp->mfcc_parent, (void *)rt->mfc_stall);
888 
889 	    rt->mfc_origin     = mfccp->mfcc_origin;
890 	    rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
891 	    rt->mfc_parent     = mfccp->mfcc_parent;
892 	    for (i = 0; i < numvifs; i++)
893 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
894 	    /* initialize pkt counters per src-grp */
895 	    rt->mfc_pkt_cnt    = 0;
896 	    rt->mfc_byte_cnt   = 0;
897 	    rt->mfc_wrong_if   = 0;
898 	    rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
899 
900 	    rt->mfc_expire = 0;	/* Don't clean this guy up */
901 	    nexpire[hash]--;
902 
903 	    /* free packets Qed at the end of this entry */
904 	    for (rte = rt->mfc_stall; rte != NULL; ) {
905 		struct rtdetq *n = rte->next;
906 
907 		ip_mdq(rte->m, rte->ifp, rt, -1);
908 		m_freem(rte->m);
909 #ifdef UPCALL_TIMING
910 		collate(&(rte->t));
911 #endif /* UPCALL_TIMING */
912 		free(rte, M_MRTABLE);
913 		rte = n;
914 	    }
915 	    rt->mfc_stall = NULL;
916 	}
917     }
918 
919     /*
920      * It is possible that an entry is being inserted without an upcall
921      */
922     if (nstl == 0) {
923 	if (mrtdebug & DEBUG_MFC)
924 	    log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n",
925 		hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr),
926 		(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
927 		mfccp->mfcc_parent);
928 
929 	for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) {
930 
931 	    if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
932 		(rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
933 
934 		rt->mfc_origin     = mfccp->mfcc_origin;
935 		rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
936 		rt->mfc_parent     = mfccp->mfcc_parent;
937 		for (i = 0; i < numvifs; i++)
938 		    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
939 		/* initialize pkt counters per src-grp */
940 		rt->mfc_pkt_cnt    = 0;
941 		rt->mfc_byte_cnt   = 0;
942 		rt->mfc_wrong_if   = 0;
943 		rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
944 		if (rt->mfc_expire)
945 		    nexpire[hash]--;
946 		rt->mfc_expire	   = 0;
947 	    }
948 	}
949 	if (rt == NULL) {
950 	    /* no upcall, so make a new entry */
951 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
952 	    if (rt == NULL) {
953 		splx(s);
954 		return ENOBUFS;
955 	    }
956 
957 	    /* insert new entry at head of hash chain */
958 	    rt->mfc_origin     = mfccp->mfcc_origin;
959 	    rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
960 	    rt->mfc_parent     = mfccp->mfcc_parent;
961 	    for (i = 0; i < numvifs; i++)
962 		    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
963 	    /* initialize pkt counters per src-grp */
964 	    rt->mfc_pkt_cnt    = 0;
965 	    rt->mfc_byte_cnt   = 0;
966 	    rt->mfc_wrong_if   = 0;
967 	    rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
968 	    rt->mfc_expire     = 0;
969 	    rt->mfc_stall      = NULL;
970 
971 	    /* link into table */
972 	    rt->mfc_next = mfctable[hash];
973 	    mfctable[hash] = rt;
974 	}
975     }
976     splx(s);
977     return 0;
978 }
979 
980 #ifdef UPCALL_TIMING
981 /*
982  * collect delay statistics on the upcalls
983  */
984 static void collate(t)
985 register struct timeval *t;
986 {
987     register u_long d;
988     register struct timeval tp;
989     register u_long delta;
990 
991     GET_TIME(tp);
992 
993     if (TV_LT(*t, tp))
994     {
995 	TV_DELTA(tp, *t, delta);
996 
997 	d = delta >> 10;
998 	if (d > 50)
999 	    d = 50;
1000 
1001 	++upcall_data[d];
1002     }
1003 }
1004 #endif /* UPCALL_TIMING */
1005 
1006 /*
1007  * Delete an mfc entry
1008  */
1009 static int
1010 del_mfc(mfccp)
1011     struct mfcctl *mfccp;
1012 {
1013     struct in_addr 	origin;
1014     struct in_addr 	mcastgrp;
1015     struct mfc 		*rt;
1016     struct mfc	 	**nptr;
1017     u_long 		hash;
1018     int s;
1019 
1020     origin = mfccp->mfcc_origin;
1021     mcastgrp = mfccp->mfcc_mcastgrp;
1022     hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1023 
1024     if (mrtdebug & DEBUG_MFC)
1025 	log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n",
1026 	    (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1027 
1028     s = splnet();
1029 
1030     nptr = &mfctable[hash];
1031     while ((rt = *nptr) != NULL) {
1032 	if (origin.s_addr == rt->mfc_origin.s_addr &&
1033 	    mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1034 	    rt->mfc_stall == NULL)
1035 	    break;
1036 
1037 	nptr = &rt->mfc_next;
1038     }
1039     if (rt == NULL) {
1040 	splx(s);
1041 	return EADDRNOTAVAIL;
1042     }
1043 
1044     *nptr = rt->mfc_next;
1045     free(rt, M_MRTABLE);
1046 
1047     splx(s);
1048 
1049     return 0;
1050 }
1051 
1052 /*
1053  * Send a message to mrouted on the multicast routing socket
1054  */
1055 static int
1056 socket_send(s, mm, src)
1057 	struct socket *s;
1058 	struct mbuf *mm;
1059 	struct sockaddr_in *src;
1060 {
1061 	if (s) {
1062 		if (sbappendaddr(&s->so_rcv,
1063 				 (struct sockaddr *)src,
1064 				 mm, (struct mbuf *)0) != 0) {
1065 			sorwakeup(s);
1066 			return 0;
1067 		}
1068 	}
1069 	m_freem(mm);
1070 	return -1;
1071 }
1072 
1073 /*
1074  * IP multicast forwarding function. This function assumes that the packet
1075  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1076  * pointed to by "ifp", and the packet is to be relayed to other networks
1077  * that have members of the packet's destination IP multicast group.
1078  *
1079  * The packet is returned unscathed to the caller, unless it is
1080  * erroneous, in which case a non-zero return value tells the caller to
1081  * discard it.
1082  */
1083 
1084 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1085 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1086 
1087 static int
1088 X_ip_mforward(ip, ifp, m, imo)
1089     register struct ip *ip;
1090     struct ifnet *ifp;
1091     struct mbuf *m;
1092     struct ip_moptions *imo;
1093 {
1094     register struct mfc *rt;
1095     register u_char *ipoptions;
1096     static struct sockaddr_in 	k_igmpsrc	= { sizeof k_igmpsrc, AF_INET };
1097     static int srctun = 0;
1098     register struct mbuf *mm;
1099     int s;
1100     vifi_t vifi;
1101     struct vif *vifp;
1102 
1103     if (mrtdebug & DEBUG_FORWARD)
1104 	log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n",
1105 	    (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr),
1106 	    (void *)ifp);
1107 
1108     if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1109 	(ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1110 	/*
1111 	 * Packet arrived via a physical interface or
1112 	 * an encapsulated tunnel.
1113 	 */
1114     } else {
1115 	/*
1116 	 * Packet arrived through a source-route tunnel.
1117 	 * Source-route tunnels are no longer supported.
1118 	 */
1119 	if ((srctun++ % 1000) == 0)
1120 	    log(LOG_ERR,
1121 		"ip_mforward: received source-routed packet from %lx\n",
1122 		(u_long)ntohl(ip->ip_src.s_addr));
1123 
1124 	return 1;
1125     }
1126 
1127     if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1128 	if (ip->ip_ttl < 255)
1129 		ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1130 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1131 	    vifp = viftable + vifi;
1132 	    printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n",
1133 		ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi,
1134 		(vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1135 		vifp->v_ifp->if_name, vifp->v_ifp->if_unit);
1136 	}
1137 	return (ip_mdq(m, ifp, NULL, vifi));
1138     }
1139     if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1140 	printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1141 	    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr));
1142 	if(!imo)
1143 		printf("In fact, no options were specified at all\n");
1144     }
1145 
1146     /*
1147      * Don't forward a packet with time-to-live of zero or one,
1148      * or a packet destined to a local-only group.
1149      */
1150     if (ip->ip_ttl <= 1 ||
1151 	ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
1152 	return 0;
1153 
1154     /*
1155      * Determine forwarding vifs from the forwarding cache table
1156      */
1157     s = splnet();
1158     MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt);
1159 
1160     /* Entry exists, so forward if necessary */
1161     if (rt != NULL) {
1162 	splx(s);
1163 	return (ip_mdq(m, ifp, rt, -1));
1164     } else {
1165 	/*
1166 	 * If we don't have a route for packet's origin,
1167 	 * Make a copy of the packet &
1168 	 * send message to routing daemon
1169 	 */
1170 
1171 	register struct mbuf *mb0;
1172 	register struct rtdetq *rte;
1173 	register u_long hash;
1174 	int hlen = ip->ip_hl << 2;
1175 #ifdef UPCALL_TIMING
1176 	struct timeval tp;
1177 
1178 	GET_TIME(tp);
1179 #endif
1180 
1181 	mrtstat.mrts_no_route++;
1182 	if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1183 	    log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n",
1184 		(u_long)ntohl(ip->ip_src.s_addr),
1185 		(u_long)ntohl(ip->ip_dst.s_addr));
1186 
1187 	/*
1188 	 * Allocate mbufs early so that we don't do extra work if we are
1189 	 * just going to fail anyway.  Make sure to pullup the header so
1190 	 * that other people can't step on it.
1191 	 */
1192 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT);
1193 	if (rte == NULL) {
1194 	    splx(s);
1195 	    return ENOBUFS;
1196 	}
1197 	mb0 = m_copy(m, 0, M_COPYALL);
1198 	if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1199 	    mb0 = m_pullup(mb0, hlen);
1200 	if (mb0 == NULL) {
1201 	    free(rte, M_MRTABLE);
1202 	    splx(s);
1203 	    return ENOBUFS;
1204 	}
1205 
1206 	/* is there an upcall waiting for this packet? */
1207 	hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1208 	for (rt = mfctable[hash]; rt; rt = rt->mfc_next) {
1209 	    if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1210 		(ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1211 		(rt->mfc_stall != NULL))
1212 		break;
1213 	}
1214 
1215 	if (rt == NULL) {
1216 	    int i;
1217 	    struct igmpmsg *im;
1218 
1219 	    /* no upcall, so make a new entry */
1220 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1221 	    if (rt == NULL) {
1222 		free(rte, M_MRTABLE);
1223 		m_freem(mb0);
1224 		splx(s);
1225 		return ENOBUFS;
1226 	    }
1227 	    /* Make a copy of the header to send to the user level process */
1228 	    mm = m_copy(mb0, 0, hlen);
1229 	    if (mm == NULL) {
1230 		free(rte, M_MRTABLE);
1231 		m_freem(mb0);
1232 		free(rt, M_MRTABLE);
1233 		splx(s);
1234 		return ENOBUFS;
1235 	    }
1236 
1237 	    /*
1238 	     * Send message to routing daemon to install
1239 	     * a route into the kernel table
1240 	     */
1241 	    k_igmpsrc.sin_addr = ip->ip_src;
1242 
1243 	    im = mtod(mm, struct igmpmsg *);
1244 	    im->im_msgtype	= IGMPMSG_NOCACHE;
1245 	    im->im_mbz		= 0;
1246 
1247 	    mrtstat.mrts_upcalls++;
1248 
1249 	    if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1250 		log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1251 		++mrtstat.mrts_upq_sockfull;
1252 		free(rte, M_MRTABLE);
1253 		m_freem(mb0);
1254 		free(rt, M_MRTABLE);
1255 		splx(s);
1256 		return ENOBUFS;
1257 	    }
1258 
1259 	    /* insert new entry at head of hash chain */
1260 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1261 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1262 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1263 	    nexpire[hash]++;
1264 	    for (i = 0; i < numvifs; i++)
1265 		rt->mfc_ttls[i] = 0;
1266 	    rt->mfc_parent = -1;
1267 
1268 	    /* link into table */
1269 	    rt->mfc_next   = mfctable[hash];
1270 	    mfctable[hash] = rt;
1271 	    rt->mfc_stall = rte;
1272 
1273 	} else {
1274 	    /* determine if q has overflowed */
1275 	    int npkts = 0;
1276 	    struct rtdetq **p;
1277 
1278 	    for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1279 		npkts++;
1280 
1281 	    if (npkts > MAX_UPQ) {
1282 		mrtstat.mrts_upq_ovflw++;
1283 		free(rte, M_MRTABLE);
1284 		m_freem(mb0);
1285 		splx(s);
1286 		return 0;
1287 	    }
1288 
1289 	    /* Add this entry to the end of the queue */
1290 	    *p = rte;
1291 	}
1292 
1293 	rte->m 			= mb0;
1294 	rte->ifp 		= ifp;
1295 #ifdef UPCALL_TIMING
1296 	rte->t			= tp;
1297 #endif
1298 	rte->next		= NULL;
1299 
1300 	splx(s);
1301 
1302 	return 0;
1303     }
1304 }
1305 
1306 #ifndef MROUTE_LKM
1307 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
1308 		   struct ip_moptions *) = X_ip_mforward;
1309 #endif
1310 
1311 /*
1312  * Clean up the cache entry if upcall is not serviced
1313  */
1314 static void
1315 expire_upcalls(void *unused)
1316 {
1317     struct rtdetq *rte;
1318     struct mfc *mfc, **nptr;
1319     int i;
1320     int s;
1321 
1322     s = splnet();
1323     for (i = 0; i < MFCTBLSIZ; i++) {
1324 	if (nexpire[i] == 0)
1325 	    continue;
1326 	nptr = &mfctable[i];
1327 	for (mfc = *nptr; mfc != NULL; mfc = *nptr) {
1328 	    /*
1329 	     * Skip real cache entries
1330 	     * Make sure it wasn't marked to not expire (shouldn't happen)
1331 	     * If it expires now
1332 	     */
1333 	    if (mfc->mfc_stall != NULL &&
1334 	        mfc->mfc_expire != 0 &&
1335 		--mfc->mfc_expire == 0) {
1336 		if (mrtdebug & DEBUG_EXPIRE)
1337 		    log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n",
1338 			(u_long)ntohl(mfc->mfc_origin.s_addr),
1339 			(u_long)ntohl(mfc->mfc_mcastgrp.s_addr));
1340 		/*
1341 		 * drop all the packets
1342 		 * free the mbuf with the pkt, if, timing info
1343 		 */
1344 		for (rte = mfc->mfc_stall; rte; ) {
1345 		    struct rtdetq *n = rte->next;
1346 
1347 		    m_freem(rte->m);
1348 		    free(rte, M_MRTABLE);
1349 		    rte = n;
1350 		}
1351 		++mrtstat.mrts_cache_cleanups;
1352 		nexpire[i]--;
1353 
1354 		*nptr = mfc->mfc_next;
1355 		free(mfc, M_MRTABLE);
1356 	    } else {
1357 		nptr = &mfc->mfc_next;
1358 	    }
1359 	}
1360     }
1361     splx(s);
1362     expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
1363 }
1364 
1365 /*
1366  * Packet forwarding routine once entry in the cache is made
1367  */
1368 static int
1369 ip_mdq(m, ifp, rt, xmt_vif)
1370     register struct mbuf *m;
1371     register struct ifnet *ifp;
1372     register struct mfc *rt;
1373     register vifi_t xmt_vif;
1374 {
1375     register struct ip  *ip = mtod(m, struct ip *);
1376     register vifi_t vifi;
1377     register struct vif *vifp;
1378     register int plen = ip->ip_len;
1379 
1380 /*
1381  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1382  * input, they shouldn't get counted on output, so statistics keeping is
1383  * seperate.
1384  */
1385 #define MC_SEND(ip,vifp,m) {                             \
1386                 if ((vifp)->v_flags & VIFF_TUNNEL)  	 \
1387                     encap_send((ip), (vifp), (m));       \
1388                 else                                     \
1389                     phyint_send((ip), (vifp), (m));      \
1390 }
1391 
1392     /*
1393      * If xmt_vif is not -1, send on only the requested vif.
1394      *
1395      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1396      */
1397     if (xmt_vif < numvifs) {
1398 	MC_SEND(ip, viftable + xmt_vif, m);
1399 	return 1;
1400     }
1401 
1402     /*
1403      * Don't forward if it didn't arrive from the parent vif for its origin.
1404      */
1405     vifi = rt->mfc_parent;
1406     if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1407 	/* came in the wrong interface */
1408 	if (mrtdebug & DEBUG_FORWARD)
1409 	    log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1410 		(void *)ifp, vifi, (void *)viftable[vifi].v_ifp);
1411 	++mrtstat.mrts_wrong_if;
1412 	++rt->mfc_wrong_if;
1413 	/*
1414 	 * If we are doing PIM assert processing, and we are forwarding
1415 	 * packets on this interface, and it is a broadcast medium
1416 	 * interface (and not a tunnel), send a message to the routing daemon.
1417 	 */
1418 	if (pim_assert && rt->mfc_ttls[vifi] &&
1419 		(ifp->if_flags & IFF_BROADCAST) &&
1420 		!(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1421 	    struct sockaddr_in k_igmpsrc;
1422 	    struct mbuf *mm;
1423 	    struct igmpmsg *im;
1424 	    int hlen = ip->ip_hl << 2;
1425 	    struct timeval now;
1426 	    register u_long delta;
1427 
1428 	    GET_TIME(now);
1429 
1430 	    TV_DELTA(rt->mfc_last_assert, now, delta);
1431 
1432 	    if (delta > ASSERT_MSG_TIME) {
1433 		mm = m_copy(m, 0, hlen);
1434 		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1435 		    mm = m_pullup(mm, hlen);
1436 		if (mm == NULL) {
1437 		    return ENOBUFS;
1438 		}
1439 
1440 		rt->mfc_last_assert = now;
1441 
1442 		im = mtod(mm, struct igmpmsg *);
1443 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1444 		im->im_mbz		= 0;
1445 		im->im_vif		= vifi;
1446 
1447 		k_igmpsrc.sin_addr = im->im_src;
1448 
1449 		socket_send(ip_mrouter, mm, &k_igmpsrc);
1450 	    }
1451 	}
1452 	return 0;
1453     }
1454 
1455     /* If I sourced this packet, it counts as output, else it was input. */
1456     if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1457 	viftable[vifi].v_pkt_out++;
1458 	viftable[vifi].v_bytes_out += plen;
1459     } else {
1460 	viftable[vifi].v_pkt_in++;
1461 	viftable[vifi].v_bytes_in += plen;
1462     }
1463     rt->mfc_pkt_cnt++;
1464     rt->mfc_byte_cnt += plen;
1465 
1466     /*
1467      * For each vif, decide if a copy of the packet should be forwarded.
1468      * Forward if:
1469      *		- the ttl exceeds the vif's threshold
1470      *		- there are group members downstream on interface
1471      */
1472     for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1473 	if ((rt->mfc_ttls[vifi] > 0) &&
1474 	    (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1475 	    vifp->v_pkt_out++;
1476 	    vifp->v_bytes_out += plen;
1477 	    MC_SEND(ip, vifp, m);
1478 	}
1479 
1480     return 0;
1481 }
1482 
1483 /*
1484  * check if a vif number is legal/ok. This is used by ip_output, to export
1485  * numvifs there,
1486  */
1487 static int
1488 X_legal_vif_num(vif)
1489     int vif;
1490 {
1491     if (vif >= 0 && vif < numvifs)
1492        return(1);
1493     else
1494        return(0);
1495 }
1496 
1497 #ifndef MROUTE_LKM
1498 int (*legal_vif_num)(int) = X_legal_vif_num;
1499 #endif
1500 
1501 /*
1502  * Return the local address used by this vif
1503  */
1504 static u_long
1505 X_ip_mcast_src(vifi)
1506     int vifi;
1507 {
1508     if (vifi >= 0 && vifi < numvifs)
1509 	return viftable[vifi].v_lcl_addr.s_addr;
1510     else
1511 	return INADDR_ANY;
1512 }
1513 
1514 #ifndef MROUTE_LKM
1515 u_long (*ip_mcast_src)(int) = X_ip_mcast_src;
1516 #endif
1517 
1518 static void
1519 phyint_send(ip, vifp, m)
1520     struct ip *ip;
1521     struct vif *vifp;
1522     struct mbuf *m;
1523 {
1524     register struct mbuf *mb_copy;
1525     register int hlen = ip->ip_hl << 2;
1526 
1527     /*
1528      * Make a new reference to the packet; make sure that
1529      * the IP header is actually copied, not just referenced,
1530      * so that ip_output() only scribbles on the copy.
1531      */
1532     mb_copy = m_copy(m, 0, M_COPYALL);
1533     if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1534 	mb_copy = m_pullup(mb_copy, hlen);
1535     if (mb_copy == NULL)
1536 	return;
1537 
1538     if (vifp->v_rate_limit == 0)
1539 	tbf_send_packet(vifp, mb_copy);
1540     else
1541 	tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1542 }
1543 
1544 static void
1545 encap_send(ip, vifp, m)
1546     register struct ip *ip;
1547     register struct vif *vifp;
1548     register struct mbuf *m;
1549 {
1550     register struct mbuf *mb_copy;
1551     register struct ip *ip_copy;
1552     register int i, len = ip->ip_len;
1553 
1554     /*
1555      * copy the old packet & pullup its IP header into the
1556      * new mbuf so we can modify it.  Try to fill the new
1557      * mbuf since if we don't the ethernet driver will.
1558      */
1559     MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER);
1560     if (mb_copy == NULL)
1561 	return;
1562     mb_copy->m_data += max_linkhdr;
1563     mb_copy->m_len = sizeof(multicast_encap_iphdr);
1564 
1565     if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1566 	m_freem(mb_copy);
1567 	return;
1568     }
1569     i = MHLEN - M_LEADINGSPACE(mb_copy);
1570     if (i > len)
1571 	i = len;
1572     mb_copy = m_pullup(mb_copy, i);
1573     if (mb_copy == NULL)
1574 	return;
1575     mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1576 
1577     /*
1578      * fill in the encapsulating IP header.
1579      */
1580     ip_copy = mtod(mb_copy, struct ip *);
1581     *ip_copy = multicast_encap_iphdr;
1582     ip_copy->ip_id = htons(ip_id++);
1583     ip_copy->ip_len += len;
1584     ip_copy->ip_src = vifp->v_lcl_addr;
1585     ip_copy->ip_dst = vifp->v_rmt_addr;
1586 
1587     /*
1588      * turn the encapsulated IP header back into a valid one.
1589      */
1590     ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1591     --ip->ip_ttl;
1592     HTONS(ip->ip_len);
1593     HTONS(ip->ip_off);
1594     ip->ip_sum = 0;
1595     mb_copy->m_data += sizeof(multicast_encap_iphdr);
1596     ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1597     mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1598 
1599     if (vifp->v_rate_limit == 0)
1600 	tbf_send_packet(vifp, mb_copy);
1601     else
1602 	tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1603 }
1604 
1605 /*
1606  * De-encapsulate a packet and feed it back through ip input (this
1607  * routine is called whenever IP gets a packet with proto type
1608  * ENCAP_PROTO and a local destination address).
1609  */
1610 void
1611 #ifdef MROUTE_LKM
1612 X_ipip_input(m, iphlen)
1613 #else
1614 ipip_input(m, iphlen)
1615 #endif
1616 	register struct mbuf *m;
1617 	int iphlen;
1618 {
1619     struct ifnet *ifp = m->m_pkthdr.rcvif;
1620     register struct ip *ip = mtod(m, struct ip *);
1621     register int hlen = ip->ip_hl << 2;
1622     register int s;
1623     register struct ifqueue *ifq;
1624     register struct vif *vifp;
1625 
1626     if (!have_encap_tunnel) {
1627 	    rip_input(m, iphlen);
1628 	    return;
1629     }
1630     /*
1631      * dump the packet if it's not to a multicast destination or if
1632      * we don't have an encapsulating tunnel with the source.
1633      * Note:  This code assumes that the remote site IP address
1634      * uniquely identifies the tunnel (i.e., that this site has
1635      * at most one tunnel with the remote site).
1636      */
1637     if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) {
1638 	++mrtstat.mrts_bad_tunnel;
1639 	m_freem(m);
1640 	return;
1641     }
1642     if (ip->ip_src.s_addr != last_encap_src) {
1643 	register struct vif *vife;
1644 
1645 	vifp = viftable;
1646 	vife = vifp + numvifs;
1647 	last_encap_src = ip->ip_src.s_addr;
1648 	last_encap_vif = 0;
1649 	for ( ; vifp < vife; ++vifp)
1650 	    if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
1651 		if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT))
1652 		    == VIFF_TUNNEL)
1653 		    last_encap_vif = vifp;
1654 		break;
1655 	    }
1656     }
1657     if ((vifp = last_encap_vif) == 0) {
1658 	last_encap_src = 0;
1659 	mrtstat.mrts_cant_tunnel++; /*XXX*/
1660 	m_freem(m);
1661 	if (mrtdebug)
1662 	  log(LOG_DEBUG, "ip_mforward: no tunnel with %lx\n",
1663 		(u_long)ntohl(ip->ip_src.s_addr));
1664 	return;
1665     }
1666     ifp = vifp->v_ifp;
1667 
1668     if (hlen > IP_HDR_LEN)
1669       ip_stripoptions(m, (struct mbuf *) 0);
1670     m->m_data += IP_HDR_LEN;
1671     m->m_len -= IP_HDR_LEN;
1672     m->m_pkthdr.len -= IP_HDR_LEN;
1673     m->m_pkthdr.rcvif = ifp;
1674 
1675     ifq = &ipintrq;
1676     s = splimp();
1677     if (IF_QFULL(ifq)) {
1678 	IF_DROP(ifq);
1679 	m_freem(m);
1680     } else {
1681 	IF_ENQUEUE(ifq, m);
1682 	/*
1683 	 * normally we would need a "schednetisr(NETISR_IP)"
1684 	 * here but we were called by ip_input and it is going
1685 	 * to loop back & try to dequeue the packet we just
1686 	 * queued as soon as we return so we avoid the
1687 	 * unnecessary software interrrupt.
1688 	 */
1689     }
1690     splx(s);
1691 }
1692 
1693 /*
1694  * Token bucket filter module
1695  */
1696 
1697 static void
1698 tbf_control(vifp, m, ip, p_len)
1699 	register struct vif *vifp;
1700 	register struct mbuf *m;
1701 	register struct ip *ip;
1702 	register u_long p_len;
1703 {
1704     register struct tbf *t = vifp->v_tbf;
1705 
1706     if (p_len > MAX_BKT_SIZE) {
1707 	/* drop if packet is too large */
1708 	mrtstat.mrts_pkt2large++;
1709 	m_freem(m);
1710 	return;
1711     }
1712 
1713     tbf_update_tokens(vifp);
1714 
1715     /* if there are enough tokens,
1716      * and the queue is empty,
1717      * send this packet out
1718      */
1719 
1720     if (t->tbf_q_len == 0) {
1721 	/* queue empty, send packet if enough tokens */
1722 	if (p_len <= t->tbf_n_tok) {
1723 	    t->tbf_n_tok -= p_len;
1724 	    tbf_send_packet(vifp, m);
1725 	} else {
1726 	    /* queue packet and timeout till later */
1727 	    tbf_queue(vifp, m);
1728 	    timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1729 	}
1730     } else if (t->tbf_q_len < t->tbf_max_q_len) {
1731 	/* finite queue length, so queue pkts and process queue */
1732 	tbf_queue(vifp, m);
1733 	tbf_process_q(vifp);
1734     } else {
1735 	/* queue length too much, try to dq and queue and process */
1736 	if (!tbf_dq_sel(vifp, ip)) {
1737 	    mrtstat.mrts_q_overflow++;
1738 	    m_freem(m);
1739 	    return;
1740 	} else {
1741 	    tbf_queue(vifp, m);
1742 	    tbf_process_q(vifp);
1743 	}
1744     }
1745     return;
1746 }
1747 
1748 /*
1749  * adds a packet to the queue at the interface
1750  */
1751 static void
1752 tbf_queue(vifp, m)
1753 	register struct vif *vifp;
1754 	register struct mbuf *m;
1755 {
1756     register int s = splnet();
1757     register struct tbf *t = vifp->v_tbf;
1758 
1759     if (t->tbf_t == NULL) {
1760 	/* Queue was empty */
1761 	t->tbf_q = m;
1762     } else {
1763 	/* Insert at tail */
1764 	t->tbf_t->m_act = m;
1765     }
1766 
1767     /* Set new tail pointer */
1768     t->tbf_t = m;
1769 
1770 #ifdef DIAGNOSTIC
1771     /* Make sure we didn't get fed a bogus mbuf */
1772     if (m->m_act)
1773 	panic("tbf_queue: m_act");
1774 #endif
1775     m->m_act = NULL;
1776 
1777     t->tbf_q_len++;
1778 
1779     splx(s);
1780 }
1781 
1782 
1783 /*
1784  * processes the queue at the interface
1785  */
1786 static void
1787 tbf_process_q(vifp)
1788     register struct vif *vifp;
1789 {
1790     register struct mbuf *m;
1791     register int len;
1792     register int s = splnet();
1793     register struct tbf *t = vifp->v_tbf;
1794 
1795     /* loop through the queue at the interface and send as many packets
1796      * as possible
1797      */
1798     while (t->tbf_q_len > 0) {
1799 	m = t->tbf_q;
1800 
1801 	len = mtod(m, struct ip *)->ip_len;
1802 
1803 	/* determine if the packet can be sent */
1804 	if (len <= t->tbf_n_tok) {
1805 	    /* if so,
1806 	     * reduce no of tokens, dequeue the packet,
1807 	     * send the packet.
1808 	     */
1809 	    t->tbf_n_tok -= len;
1810 
1811 	    t->tbf_q = m->m_act;
1812 	    if (--t->tbf_q_len == 0)
1813 		t->tbf_t = NULL;
1814 
1815 	    m->m_act = NULL;
1816 	    tbf_send_packet(vifp, m);
1817 
1818 	} else break;
1819     }
1820     splx(s);
1821 }
1822 
1823 static void
1824 tbf_reprocess_q(xvifp)
1825 	void *xvifp;
1826 {
1827     register struct vif *vifp = xvifp;
1828     if (ip_mrouter == NULL)
1829 	return;
1830 
1831     tbf_update_tokens(vifp);
1832 
1833     tbf_process_q(vifp);
1834 
1835     if (vifp->v_tbf->tbf_q_len)
1836 	timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1837 }
1838 
1839 /* function that will selectively discard a member of the queue
1840  * based on the precedence value and the priority
1841  */
1842 static int
1843 tbf_dq_sel(vifp, ip)
1844     register struct vif *vifp;
1845     register struct ip *ip;
1846 {
1847     register int s = splnet();
1848     register u_int p;
1849     register struct mbuf *m, *last;
1850     register struct mbuf **np;
1851     register struct tbf *t = vifp->v_tbf;
1852 
1853     p = priority(vifp, ip);
1854 
1855     np = &t->tbf_q;
1856     last = NULL;
1857     while ((m = *np) != NULL) {
1858 	if (p > priority(vifp, mtod(m, struct ip *))) {
1859 	    *np = m->m_act;
1860 	    /* If we're removing the last packet, fix the tail pointer */
1861 	    if (m == t->tbf_t)
1862 		t->tbf_t = last;
1863 	    m_freem(m);
1864 	    /* it's impossible for the queue to be empty, but
1865 	     * we check anyway. */
1866 	    if (--t->tbf_q_len == 0)
1867 		t->tbf_t = NULL;
1868 	    splx(s);
1869 	    mrtstat.mrts_drop_sel++;
1870 	    return(1);
1871 	}
1872 	np = &m->m_act;
1873 	last = m;
1874     }
1875     splx(s);
1876     return(0);
1877 }
1878 
1879 static void
1880 tbf_send_packet(vifp, m)
1881     register struct vif *vifp;
1882     register struct mbuf *m;
1883 {
1884     struct ip_moptions imo;
1885     int error;
1886     static struct route ro;
1887     int s = splnet();
1888 
1889     if (vifp->v_flags & VIFF_TUNNEL) {
1890 	/* If tunnel options */
1891 	ip_output(m, (struct mbuf *)0, &vifp->v_route,
1892 		  IP_FORWARDING, (struct ip_moptions *)0);
1893     } else {
1894 	imo.imo_multicast_ifp  = vifp->v_ifp;
1895 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1896 	imo.imo_multicast_loop = 1;
1897 	imo.imo_multicast_vif  = -1;
1898 
1899 	/*
1900 	 * Re-entrancy should not be a problem here, because
1901 	 * the packets that we send out and are looped back at us
1902 	 * should get rejected because they appear to come from
1903 	 * the loopback interface, thus preventing looping.
1904 	 */
1905 	error = ip_output(m, (struct mbuf *)0, &ro,
1906 			  IP_FORWARDING, &imo);
1907 
1908 	if (mrtdebug & DEBUG_XMIT)
1909 	    log(LOG_DEBUG, "phyint_send on vif %d err %d\n",
1910 		vifp - viftable, error);
1911     }
1912     splx(s);
1913 }
1914 
1915 /* determine the current time and then
1916  * the elapsed time (between the last time and time now)
1917  * in milliseconds & update the no. of tokens in the bucket
1918  */
1919 static void
1920 tbf_update_tokens(vifp)
1921     register struct vif *vifp;
1922 {
1923     struct timeval tp;
1924     register u_long tm;
1925     register int s = splnet();
1926     register struct tbf *t = vifp->v_tbf;
1927 
1928     GET_TIME(tp);
1929 
1930     TV_DELTA(tp, t->tbf_last_pkt_t, tm);
1931 
1932     /*
1933      * This formula is actually
1934      * "time in seconds" * "bytes/second".
1935      *
1936      * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
1937      *
1938      * The (1000/1024) was introduced in add_vif to optimize
1939      * this divide into a shift.
1940      */
1941     t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8;
1942     t->tbf_last_pkt_t = tp;
1943 
1944     if (t->tbf_n_tok > MAX_BKT_SIZE)
1945 	t->tbf_n_tok = MAX_BKT_SIZE;
1946 
1947     splx(s);
1948 }
1949 
1950 static int
1951 priority(vifp, ip)
1952     register struct vif *vifp;
1953     register struct ip *ip;
1954 {
1955     register int prio;
1956 
1957     /* temporary hack; may add general packet classifier some day */
1958 
1959     /*
1960      * The UDP port space is divided up into four priority ranges:
1961      * [0, 16384)     : unclassified - lowest priority
1962      * [16384, 32768) : audio - highest priority
1963      * [32768, 49152) : whiteboard - medium priority
1964      * [49152, 65536) : video - low priority
1965      */
1966     if (ip->ip_p == IPPROTO_UDP) {
1967 	struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
1968 	switch (ntohs(udp->uh_dport) & 0xc000) {
1969 	    case 0x4000:
1970 		prio = 70;
1971 		break;
1972 	    case 0x8000:
1973 		prio = 60;
1974 		break;
1975 	    case 0xc000:
1976 		prio = 55;
1977 		break;
1978 	    default:
1979 		prio = 50;
1980 		break;
1981 	}
1982 	if (tbfdebug > 1)
1983 		log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio);
1984     } else {
1985 	    prio = 50;
1986     }
1987     return prio;
1988 }
1989 
1990 /*
1991  * End of token bucket filter modifications
1992  */
1993 
1994 int
1995 ip_rsvp_vif_init(so, sopt)
1996 	struct socket *so;
1997 	struct sockopt *sopt;
1998 {
1999     int error, i, s;
2000 
2001     if (rsvpdebug)
2002 	printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2003 	       so->so_type, so->so_proto->pr_protocol);
2004 
2005     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2006 	return EOPNOTSUPP;
2007 
2008     /* Check mbuf. */
2009     error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
2010     if (error)
2011 	    return (error);
2012 
2013     if (rsvpdebug)
2014 	printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", i, rsvp_on);
2015 
2016     s = splnet();
2017 
2018     /* Check vif. */
2019     if (!legal_vif_num(i)) {
2020 	splx(s);
2021 	return EADDRNOTAVAIL;
2022     }
2023 
2024     /* Check if socket is available. */
2025     if (viftable[i].v_rsvpd != NULL) {
2026 	splx(s);
2027 	return EADDRINUSE;
2028     }
2029 
2030     viftable[i].v_rsvpd = so;
2031     /* This may seem silly, but we need to be sure we don't over-increment
2032      * the RSVP counter, in case something slips up.
2033      */
2034     if (!viftable[i].v_rsvp_on) {
2035 	viftable[i].v_rsvp_on = 1;
2036 	rsvp_on++;
2037     }
2038 
2039     splx(s);
2040     return 0;
2041 }
2042 
2043 int
2044 ip_rsvp_vif_done(so, sopt)
2045 	struct socket *so;
2046 	struct sockopt *sopt;
2047 {
2048 	int error, i, s;
2049 
2050 	if (rsvpdebug)
2051 		printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2052 		       so->so_type, so->so_proto->pr_protocol);
2053 
2054 	if (so->so_type != SOCK_RAW ||
2055 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2056 		return EOPNOTSUPP;
2057 
2058 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
2059 	if (error)
2060 		return (error);
2061 
2062 	s = splnet();
2063 
2064 	/* Check vif. */
2065 	if (!legal_vif_num(i)) {
2066 		splx(s);
2067 		return EADDRNOTAVAIL;
2068 	}
2069 
2070 	if (rsvpdebug)
2071 		printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n",
2072 		       viftable[i].v_rsvpd, so);
2073 
2074 	viftable[i].v_rsvpd = NULL;
2075 	/*
2076 	 * This may seem silly, but we need to be sure we don't over-decrement
2077 	 * the RSVP counter, in case something slips up.
2078 	 */
2079 	if (viftable[i].v_rsvp_on) {
2080 		viftable[i].v_rsvp_on = 0;
2081 		rsvp_on--;
2082 	}
2083 
2084 	splx(s);
2085 	return 0;
2086 }
2087 
2088 void
2089 ip_rsvp_force_done(so)
2090     struct socket *so;
2091 {
2092     int vifi;
2093     register int s;
2094 
2095     /* Don't bother if it is not the right type of socket. */
2096     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2097 	return;
2098 
2099     s = splnet();
2100 
2101     /* The socket may be attached to more than one vif...this
2102      * is perfectly legal.
2103      */
2104     for (vifi = 0; vifi < numvifs; vifi++) {
2105 	if (viftable[vifi].v_rsvpd == so) {
2106 	    viftable[vifi].v_rsvpd = NULL;
2107 	    /* This may seem silly, but we need to be sure we don't
2108 	     * over-decrement the RSVP counter, in case something slips up.
2109 	     */
2110 	    if (viftable[vifi].v_rsvp_on) {
2111 		viftable[vifi].v_rsvp_on = 0;
2112 		rsvp_on--;
2113 	    }
2114 	}
2115     }
2116 
2117     splx(s);
2118     return;
2119 }
2120 
2121 void
2122 rsvp_input(m, iphlen)
2123 	struct mbuf *m;
2124 	int iphlen;
2125 {
2126     int vifi;
2127     register struct ip *ip = mtod(m, struct ip *);
2128     static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
2129     register int s;
2130     struct ifnet *ifp;
2131 
2132     if (rsvpdebug)
2133 	printf("rsvp_input: rsvp_on %d\n",rsvp_on);
2134 
2135     /* Can still get packets with rsvp_on = 0 if there is a local member
2136      * of the group to which the RSVP packet is addressed.  But in this
2137      * case we want to throw the packet away.
2138      */
2139     if (!rsvp_on) {
2140 	m_freem(m);
2141 	return;
2142     }
2143 
2144     /* If the old-style non-vif-associated socket is set, then use
2145      * it and ignore the new ones.
2146      */
2147     if (ip_rsvpd != NULL) {
2148 	if (rsvpdebug)
2149 	    printf("rsvp_input: Sending packet up old-style socket\n");
2150 	rip_input(m, iphlen);
2151 	return;
2152     }
2153 
2154     s = splnet();
2155 
2156     if (rsvpdebug)
2157 	printf("rsvp_input: check vifs\n");
2158 
2159 #ifdef DIAGNOSTIC
2160     if (!(m->m_flags & M_PKTHDR))
2161 	    panic("rsvp_input no hdr");
2162 #endif
2163 
2164     ifp = m->m_pkthdr.rcvif;
2165     /* Find which vif the packet arrived on. */
2166     for (vifi = 0; vifi < numvifs; vifi++) {
2167 	if (viftable[vifi].v_ifp == ifp)
2168  		break;
2169  	}
2170 
2171     if (vifi == numvifs) {
2172 	/* Can't find vif packet arrived on. Drop packet. */
2173 	if (rsvpdebug)
2174 	    printf("rsvp_input: Can't find vif for packet...dropping it.\n");
2175 	m_freem(m);
2176 	splx(s);
2177 	return;
2178     }
2179 
2180     if (rsvpdebug)
2181 	printf("rsvp_input: check socket\n");
2182 
2183     if (viftable[vifi].v_rsvpd == NULL) {
2184 	/* drop packet, since there is no specific socket for this
2185 	 * interface */
2186 	    if (rsvpdebug)
2187 		    printf("rsvp_input: No socket defined for vif %d\n",vifi);
2188 	    m_freem(m);
2189 	    splx(s);
2190 	    return;
2191     }
2192     rsvp_src.sin_addr = ip->ip_src;
2193 
2194     if (rsvpdebug && m)
2195 	printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n",
2196 	       m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
2197 
2198     if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) {
2199 	if (rsvpdebug)
2200 	    printf("rsvp_input: Failed to append to socket\n");
2201     } else {
2202 	if (rsvpdebug)
2203 	    printf("rsvp_input: send packet up\n");
2204     }
2205 
2206     splx(s);
2207 }
2208 
2209 #ifdef MROUTE_LKM
2210 #include <sys/conf.h>
2211 #include <sys/exec.h>
2212 #include <sys/sysent.h>
2213 #include <sys/lkm.h>
2214 
2215 MOD_MISC("ip_mroute_mod")
2216 
2217 static int
2218 ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd)
2219 {
2220 	int i;
2221 	struct lkm_misc	*args = lkmtp->private.lkm_misc;
2222 	int err = 0;
2223 
2224 	switch(cmd) {
2225 		static int (*old_ip_mrouter_cmd)();
2226 		static int (*old_ip_mrouter_done)();
2227 		static int (*old_ip_mforward)();
2228 		static int (*old_mrt_ioctl)();
2229 		static void (*old_proto4_input)();
2230 		static int (*old_legal_vif_num)();
2231 		extern struct protosw inetsw[];
2232 
2233 	case LKM_E_LOAD:
2234 		if(lkmexists(lkmtp) || ip_mrtproto)
2235 		  return(EEXIST);
2236 		old_ip_mrouter_cmd = ip_mrouter_cmd;
2237 		ip_mrouter_cmd = X_ip_mrouter_cmd;
2238 		old_ip_mrouter_done = ip_mrouter_done;
2239 		ip_mrouter_done = X_ip_mrouter_done;
2240 		old_ip_mforward = ip_mforward;
2241 		ip_mforward = X_ip_mforward;
2242 		old_mrt_ioctl = mrt_ioctl;
2243 		mrt_ioctl = X_mrt_ioctl;
2244               old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input;
2245               inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input;
2246 		old_legal_vif_num = legal_vif_num;
2247 		legal_vif_num = X_legal_vif_num;
2248 		ip_mrtproto = IGMP_DVMRP;
2249 
2250 		printf("\nIP multicast routing loaded\n");
2251 		break;
2252 
2253 	case LKM_E_UNLOAD:
2254 		if (ip_mrouter)
2255 		  return EINVAL;
2256 
2257 		ip_mrouter_cmd = old_ip_mrouter_cmd;
2258 		ip_mrouter_done = old_ip_mrouter_done;
2259 		ip_mforward = old_ip_mforward;
2260 		mrt_ioctl = old_mrt_ioctl;
2261               inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input;
2262 		legal_vif_num = old_legal_vif_num;
2263 		ip_mrtproto = 0;
2264 		break;
2265 
2266 	default:
2267 		err = EINVAL;
2268 		break;
2269 	}
2270 
2271 	return(err);
2272 }
2273 
2274 int
2275 ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) {
2276 	DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle,
2277 		 nosys);
2278 }
2279 
2280 #endif /* MROUTE_LKM */
2281 #endif /* MROUTING */
2282