xref: /freebsd/sys/netinet/ip_mroute.c (revision 23f282aa31e9b6fceacd449020e936e98d6f2298)
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, off, proto)		/* XXX must fixup manually */
121 	struct mbuf *m;
122 	int off;
123 	int proto;
124 {
125     /* Can still get packets with rsvp_on = 0 if there is a local member
126      * of the group to which the RSVP packet is addressed.  But in this
127      * case we want to throw the packet away.
128      */
129     if (!rsvp_on) {
130 	m_freem(m);
131 	return;
132     }
133 
134     if (ip_rsvpd != NULL) {
135 	if (rsvpdebug)
136 	    printf("rsvp_input: Sending packet up old-style socket\n");
137 	rip_input(m, off, proto);
138 	return;
139     }
140     /* Drop the packet */
141     m_freem(m);
142 }
143 
144 void ipip_input(struct mbuf *m, int off, int proto) { /* XXX must fixup manually */
145 	rip_input(m, off, proto);
146 }
147 
148 int (*legal_vif_num)(int) = 0;
149 
150 /*
151  * This should never be called, since IP_MULTICAST_VIF should fail, but
152  * just in case it does get called, the code a little lower in ip_output
153  * will assign the packet a local address.
154  */
155 u_long
156 _ip_mcast_src(int vifi) { return INADDR_ANY; }
157 u_long (*ip_mcast_src)(int) = _ip_mcast_src;
158 
159 int
160 ip_rsvp_vif_init(so, sopt)
161     struct socket *so;
162     struct sockopt *sopt;
163 {
164     return(EINVAL);
165 }
166 
167 int
168 ip_rsvp_vif_done(so, sopt)
169     struct socket *so;
170     struct sockopt *sopt;
171 {
172     return(EINVAL);
173 }
174 
175 void
176 ip_rsvp_force_done(so)
177     struct socket *so;
178 {
179     return;
180 }
181 
182 #else /* MROUTING */
183 
184 #define M_HASCL(m)	((m)->m_flags & M_EXT)
185 
186 #define INSIZ		sizeof(struct in_addr)
187 #define	same(a1, a2) \
188 	(bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0)
189 
190 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables");
191 
192 /*
193  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
194  * except for netstat or debugging purposes.
195  */
196 #ifndef MROUTE_LKM
197 struct socket  *ip_mrouter  = NULL;
198 static struct mrtstat	mrtstat;
199 #else /* MROUTE_LKM */
200 extern void	X_ipip_input __P((struct mbuf *m, int iphlen));
201 extern struct mrtstat mrtstat;
202 static int ip_mrtproto;
203 #endif
204 
205 #define NO_RTE_FOUND 	0x1
206 #define RTE_FOUND	0x2
207 
208 static struct mfc	*mfctable[MFCTBLSIZ];
209 static u_char		nexpire[MFCTBLSIZ];
210 static struct vif	viftable[MAXVIFS];
211 static u_int	mrtdebug = 0;	  /* debug level 	*/
212 #define		DEBUG_MFC	0x02
213 #define		DEBUG_FORWARD	0x04
214 #define		DEBUG_EXPIRE	0x08
215 #define		DEBUG_XMIT	0x10
216 static u_int  	tbfdebug = 0;     /* tbf debug level 	*/
217 static u_int	rsvpdebug = 0;	  /* rsvp debug level   */
218 
219 static struct callout_handle expire_upcalls_ch;
220 
221 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
222 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
223 
224 /*
225  * Define the token bucket filter structures
226  * tbftable -> each vif has one of these for storing info
227  */
228 
229 static struct tbf tbftable[MAXVIFS];
230 #define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
231 
232 /*
233  * 'Interfaces' associated with decapsulator (so we can tell
234  * packets that went through it from ones that get reflected
235  * by a broken gateway).  These interfaces are never linked into
236  * the system ifnet list & no routes point to them.  I.e., packets
237  * can't be sent this way.  They only exist as a placeholder for
238  * multicast source verification.
239  */
240 static struct ifnet multicast_decap_if[MAXVIFS];
241 
242 #define ENCAP_TTL 64
243 #define ENCAP_PROTO IPPROTO_IPIP	/* 4 */
244 
245 /* prototype IP hdr for encapsulated packets */
246 static struct ip multicast_encap_iphdr = {
247 #if BYTE_ORDER == LITTLE_ENDIAN
248 	sizeof(struct ip) >> 2, IPVERSION,
249 #else
250 	IPVERSION, sizeof(struct ip) >> 2,
251 #endif
252 	0,				/* tos */
253 	sizeof(struct ip),		/* total length */
254 	0,				/* id */
255 	0,				/* frag offset */
256 	ENCAP_TTL, ENCAP_PROTO,
257 	0,				/* checksum */
258 };
259 
260 /*
261  * Private variables.
262  */
263 static vifi_t	   numvifs = 0;
264 static int have_encap_tunnel = 0;
265 
266 /*
267  * one-back cache used by ipip_input to locate a tunnel's vif
268  * given a datagram's src ip address.
269  */
270 static u_long last_encap_src;
271 static struct vif *last_encap_vif;
272 
273 static u_long	X_ip_mcast_src __P((int vifi));
274 static int	X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo));
275 static int	X_ip_mrouter_done __P((void));
276 static int	X_ip_mrouter_get __P((struct socket *so, struct sockopt *m));
277 static int	X_ip_mrouter_set __P((struct socket *so, struct sockopt *m));
278 static int	X_legal_vif_num __P((int vif));
279 static int	X_mrt_ioctl __P((int cmd, caddr_t data));
280 
281 static int get_sg_cnt(struct sioc_sg_req *);
282 static int get_vif_cnt(struct sioc_vif_req *);
283 static int ip_mrouter_init(struct socket *, int);
284 static int add_vif(struct vifctl *);
285 static int del_vif(vifi_t);
286 static int add_mfc(struct mfcctl *);
287 static int del_mfc(struct mfcctl *);
288 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
289 static int set_assert(int);
290 static void expire_upcalls(void *);
291 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *,
292 		  vifi_t);
293 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
294 static void encap_send(struct ip *, struct vif *, struct mbuf *);
295 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
296 static void tbf_queue(struct vif *, struct mbuf *);
297 static void tbf_process_q(struct vif *);
298 static void tbf_reprocess_q(void *);
299 static int tbf_dq_sel(struct vif *, struct ip *);
300 static void tbf_send_packet(struct vif *, struct mbuf *);
301 static void tbf_update_tokens(struct vif *);
302 static int priority(struct vif *, struct ip *);
303 void multiencap_decap(struct mbuf *);
304 
305 /*
306  * whether or not special PIM assert processing is enabled.
307  */
308 static int pim_assert;
309 /*
310  * Rate limit for assert notification messages, in usec
311  */
312 #define ASSERT_MSG_TIME		3000000
313 
314 /*
315  * Hash function for a source, group entry
316  */
317 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
318 			((g) >> 20) ^ ((g) >> 10) ^ (g))
319 
320 /*
321  * Find a route for a given origin IP address and Multicast group address
322  * Type of service parameter to be added in the future!!!
323  */
324 
325 #define MFCFIND(o, g, rt) { \
326 	register struct mfc *_rt = mfctable[MFCHASH(o,g)]; \
327 	rt = NULL; \
328 	++mrtstat.mrts_mfc_lookups; \
329 	while (_rt) { \
330 		if ((_rt->mfc_origin.s_addr == o) && \
331 		    (_rt->mfc_mcastgrp.s_addr == g) && \
332 		    (_rt->mfc_stall == NULL)) { \
333 			rt = _rt; \
334 			break; \
335 		} \
336 		_rt = _rt->mfc_next; \
337 	} \
338 	if (rt == NULL) { \
339 		++mrtstat.mrts_mfc_misses; \
340 	} \
341 }
342 
343 
344 /*
345  * Macros to compute elapsed time efficiently
346  * Borrowed from Van Jacobson's scheduling code
347  */
348 #define TV_DELTA(a, b, delta) { \
349 	    register int xxs; \
350 		\
351 	    delta = (a).tv_usec - (b).tv_usec; \
352 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
353 	       switch (xxs) { \
354 		      case 2: \
355 			  delta += 1000000; \
356 			      /* fall through */ \
357 		      case 1: \
358 			  delta += 1000000; \
359 			  break; \
360 		      default: \
361 			  delta += (1000000 * xxs); \
362 	       } \
363 	    } \
364 }
365 
366 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
367 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
368 
369 #ifdef UPCALL_TIMING
370 u_long upcall_data[51];
371 static void collate(struct timeval *);
372 #endif /* UPCALL_TIMING */
373 
374 
375 /*
376  * Handle MRT setsockopt commands to modify the multicast routing tables.
377  */
378 static int
379 X_ip_mrouter_set(so, sopt)
380 	struct socket *so;
381 	struct sockopt *sopt;
382 {
383 	int	error, optval;
384 	vifi_t	vifi;
385 	struct	vifctl vifc;
386 	struct	mfcctl mfc;
387 
388 	if (so != ip_mrouter && sopt->sopt_name != MRT_INIT)
389 		return (EPERM);
390 
391 	error = 0;
392 	switch (sopt->sopt_name) {
393 	case MRT_INIT:
394 		error = sooptcopyin(sopt, &optval, sizeof optval,
395 				    sizeof optval);
396 		if (error)
397 			break;
398 		error = ip_mrouter_init(so, optval);
399 		break;
400 
401 	case MRT_DONE:
402 		error = ip_mrouter_done();
403 		break;
404 
405 	case MRT_ADD_VIF:
406 		error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
407 		if (error)
408 			break;
409 		error = add_vif(&vifc);
410 		break;
411 
412 	case MRT_DEL_VIF:
413 		error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
414 		if (error)
415 			break;
416 		error = del_vif(vifi);
417 		break;
418 
419 	case MRT_ADD_MFC:
420 	case MRT_DEL_MFC:
421 		error = sooptcopyin(sopt, &mfc, sizeof mfc, sizeof mfc);
422 		if (error)
423 			break;
424 		if (sopt->sopt_name == MRT_ADD_MFC)
425 			error = add_mfc(&mfc);
426 		else
427 			error = del_mfc(&mfc);
428 		break;
429 
430 	case MRT_ASSERT:
431 		error = sooptcopyin(sopt, &optval, sizeof optval,
432 				    sizeof optval);
433 		if (error)
434 			break;
435 		set_assert(optval);
436 		break;
437 
438 	default:
439 		error = EOPNOTSUPP;
440 		break;
441 	}
442 	return (error);
443 }
444 
445 #ifndef MROUTE_LKM
446 int (*ip_mrouter_set)(struct socket *, struct sockopt *) = X_ip_mrouter_set;
447 #endif
448 
449 /*
450  * Handle MRT getsockopt commands
451  */
452 static int
453 X_ip_mrouter_get(so, sopt)
454 	struct socket *so;
455 	struct sockopt *sopt;
456 {
457 	int error;
458 	static int version = 0x0305; /* !!! why is this here? XXX */
459 
460 	switch (sopt->sopt_name) {
461 	case MRT_VERSION:
462 		error = sooptcopyout(sopt, &version, sizeof version);
463 		break;
464 
465 	case MRT_ASSERT:
466 		error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert);
467 		break;
468 	default:
469 		error = EOPNOTSUPP;
470 		break;
471 	}
472 	return (error);
473 }
474 
475 #ifndef MROUTE_LKM
476 int (*ip_mrouter_get)(struct socket *, struct sockopt *) = X_ip_mrouter_get;
477 #endif
478 
479 /*
480  * Handle ioctl commands to obtain information from the cache
481  */
482 static int
483 X_mrt_ioctl(cmd, data)
484     int cmd;
485     caddr_t data;
486 {
487     int error = 0;
488 
489     switch (cmd) {
490 	case (SIOCGETVIFCNT):
491 	    return (get_vif_cnt((struct sioc_vif_req *)data));
492 	    break;
493 	case (SIOCGETSGCNT):
494 	    return (get_sg_cnt((struct sioc_sg_req *)data));
495 	    break;
496 	default:
497 	    return (EINVAL);
498 	    break;
499     }
500     return error;
501 }
502 
503 #ifndef MROUTE_LKM
504 int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl;
505 #endif
506 
507 /*
508  * returns the packet, byte, rpf-failure count for the source group provided
509  */
510 static int
511 get_sg_cnt(req)
512     register struct sioc_sg_req *req;
513 {
514     register struct mfc *rt;
515     int s;
516 
517     s = splnet();
518     MFCFIND(req->src.s_addr, req->grp.s_addr, rt);
519     splx(s);
520     if (rt != NULL) {
521 	req->pktcnt = rt->mfc_pkt_cnt;
522 	req->bytecnt = rt->mfc_byte_cnt;
523 	req->wrong_if = rt->mfc_wrong_if;
524     } else
525 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
526 
527     return 0;
528 }
529 
530 /*
531  * returns the input and output packet and byte counts on the vif provided
532  */
533 static int
534 get_vif_cnt(req)
535     register struct sioc_vif_req *req;
536 {
537     register vifi_t vifi = req->vifi;
538 
539     if (vifi >= numvifs) return EINVAL;
540 
541     req->icount = viftable[vifi].v_pkt_in;
542     req->ocount = viftable[vifi].v_pkt_out;
543     req->ibytes = viftable[vifi].v_bytes_in;
544     req->obytes = viftable[vifi].v_bytes_out;
545 
546     return 0;
547 }
548 
549 /*
550  * Enable multicast routing
551  */
552 static int
553 ip_mrouter_init(so, version)
554 	struct socket *so;
555 	int version;
556 {
557     if (mrtdebug)
558 	log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
559 		so->so_type, so->so_proto->pr_protocol);
560 
561     if (so->so_type != SOCK_RAW ||
562 	so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP;
563 
564     if (version != 1)
565 	return ENOPROTOOPT;
566 
567     if (ip_mrouter != NULL) return EADDRINUSE;
568 
569     ip_mrouter = so;
570 
571     bzero((caddr_t)mfctable, sizeof(mfctable));
572     bzero((caddr_t)nexpire, sizeof(nexpire));
573 
574     pim_assert = 0;
575 
576     expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
577 
578     if (mrtdebug)
579 	log(LOG_DEBUG, "ip_mrouter_init\n");
580 
581     return 0;
582 }
583 
584 /*
585  * Disable multicast routing
586  */
587 static int
588 X_ip_mrouter_done()
589 {
590     vifi_t vifi;
591     int i;
592     struct ifnet *ifp;
593     struct ifreq ifr;
594     struct mfc *rt;
595     struct rtdetq *rte;
596     int s;
597 
598     s = splnet();
599 
600     /*
601      * For each phyint in use, disable promiscuous reception of all IP
602      * multicasts.
603      */
604     for (vifi = 0; vifi < numvifs; vifi++) {
605 	if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
606 	    !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
607 	    ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
608 	    ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr
609 								= INADDR_ANY;
610 	    ifp = viftable[vifi].v_ifp;
611 	    if_allmulti(ifp, 0);
612 	}
613     }
614     bzero((caddr_t)tbftable, sizeof(tbftable));
615     bzero((caddr_t)viftable, sizeof(viftable));
616     numvifs = 0;
617     pim_assert = 0;
618 
619     untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch);
620 
621     /*
622      * Free all multicast forwarding cache entries.
623      */
624     for (i = 0; i < MFCTBLSIZ; i++) {
625 	for (rt = mfctable[i]; rt != NULL; ) {
626 	    struct mfc *nr = rt->mfc_next;
627 
628 	    for (rte = rt->mfc_stall; rte != NULL; ) {
629 		struct rtdetq *n = rte->next;
630 
631 		m_freem(rte->m);
632 		free(rte, M_MRTABLE);
633 		rte = n;
634 	    }
635 	    free(rt, M_MRTABLE);
636 	    rt = nr;
637 	}
638     }
639 
640     bzero((caddr_t)mfctable, sizeof(mfctable));
641 
642     /*
643      * Reset de-encapsulation cache
644      */
645     last_encap_src = 0;
646     last_encap_vif = NULL;
647     have_encap_tunnel = 0;
648 
649     ip_mrouter = NULL;
650 
651     splx(s);
652 
653     if (mrtdebug)
654 	log(LOG_DEBUG, "ip_mrouter_done\n");
655 
656     return 0;
657 }
658 
659 #ifndef MROUTE_LKM
660 int (*ip_mrouter_done)(void) = X_ip_mrouter_done;
661 #endif
662 
663 /*
664  * Set PIM assert processing global
665  */
666 static int
667 set_assert(i)
668 	int i;
669 {
670     if ((i != 1) && (i != 0))
671 	return EINVAL;
672 
673     pim_assert = i;
674 
675     return 0;
676 }
677 
678 /*
679  * Add a vif to the vif table
680  */
681 static int
682 add_vif(vifcp)
683     register struct vifctl *vifcp;
684 {
685     register struct vif *vifp = viftable + vifcp->vifc_vifi;
686     static struct sockaddr_in sin = {sizeof sin, AF_INET};
687     struct ifaddr *ifa;
688     struct ifnet *ifp;
689     int error, s;
690     struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
691 
692     if (vifcp->vifc_vifi >= MAXVIFS)  return EINVAL;
693     if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE;
694 
695     /* Find the interface with an address in AF_INET family */
696     sin.sin_addr = vifcp->vifc_lcl_addr;
697     ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
698     if (ifa == 0) return EADDRNOTAVAIL;
699     ifp = ifa->ifa_ifp;
700 
701     if (vifcp->vifc_flags & VIFF_TUNNEL) {
702 	if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
703 		/*
704 		 * An encapsulating tunnel is wanted.  Tell ipip_input() to
705 		 * start paying attention to encapsulated packets.
706 		 */
707 		if (have_encap_tunnel == 0) {
708 			have_encap_tunnel = 1;
709 			for (s = 0; s < MAXVIFS; ++s) {
710 				multicast_decap_if[s].if_name = "mdecap";
711 				multicast_decap_if[s].if_unit = s;
712 			}
713 		}
714 		/*
715 		 * Set interface to fake encapsulator interface
716 		 */
717 		ifp = &multicast_decap_if[vifcp->vifc_vifi];
718 		/*
719 		 * Prepare cached route entry
720 		 */
721 		bzero(&vifp->v_route, sizeof(vifp->v_route));
722 	} else {
723 	    log(LOG_ERR, "source routed tunnels not supported\n");
724 	    return EOPNOTSUPP;
725 	}
726     } else {
727 	/* Make sure the interface supports multicast */
728 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
729 	    return EOPNOTSUPP;
730 
731 	/* Enable promiscuous reception of all IP multicasts from the if */
732 	s = splnet();
733 	error = if_allmulti(ifp, 1);
734 	splx(s);
735 	if (error)
736 	    return error;
737     }
738 
739     s = splnet();
740     /* define parameters for the tbf structure */
741     vifp->v_tbf = v_tbf;
742     GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
743     vifp->v_tbf->tbf_n_tok = 0;
744     vifp->v_tbf->tbf_q_len = 0;
745     vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
746     vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
747 
748     vifp->v_flags     = vifcp->vifc_flags;
749     vifp->v_threshold = vifcp->vifc_threshold;
750     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
751     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
752     vifp->v_ifp       = ifp;
753     /* scaling up here allows division by 1024 in critical code */
754     vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
755     vifp->v_rsvp_on   = 0;
756     vifp->v_rsvpd     = NULL;
757     /* initialize per vif pkt counters */
758     vifp->v_pkt_in    = 0;
759     vifp->v_pkt_out   = 0;
760     vifp->v_bytes_in  = 0;
761     vifp->v_bytes_out = 0;
762     splx(s);
763 
764     /* Adjust numvifs up if the vifi is higher than numvifs */
765     if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
766 
767     if (mrtdebug)
768 	log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n",
769 	    vifcp->vifc_vifi,
770 	    (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr),
771 	    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
772 	    (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr),
773 	    vifcp->vifc_threshold,
774 	    vifcp->vifc_rate_limit);
775 
776     return 0;
777 }
778 
779 /*
780  * Delete a vif from the vif table
781  */
782 static int
783 del_vif(vifi)
784 	vifi_t vifi;
785 {
786     register struct vif *vifp = &viftable[vifi];
787     register struct mbuf *m;
788     struct ifnet *ifp;
789     struct ifreq ifr;
790     int s;
791 
792     if (vifi >= numvifs) return EINVAL;
793     if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL;
794 
795     s = splnet();
796 
797     if (!(vifp->v_flags & VIFF_TUNNEL)) {
798 	((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
799 	((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
800 	ifp = vifp->v_ifp;
801 	if_allmulti(ifp, 0);
802     }
803 
804     if (vifp == last_encap_vif) {
805 	last_encap_vif = 0;
806 	last_encap_src = 0;
807     }
808 
809     /*
810      * Free packets queued at the interface
811      */
812     while (vifp->v_tbf->tbf_q) {
813 	m = vifp->v_tbf->tbf_q;
814 	vifp->v_tbf->tbf_q = m->m_act;
815 	m_freem(m);
816     }
817 
818     bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
819     bzero((caddr_t)vifp, sizeof (*vifp));
820 
821     if (mrtdebug)
822       log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs);
823 
824     /* Adjust numvifs down */
825     for (vifi = numvifs; vifi > 0; vifi--)
826 	if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break;
827     numvifs = vifi;
828 
829     splx(s);
830 
831     return 0;
832 }
833 
834 /*
835  * Add an mfc entry
836  */
837 static int
838 add_mfc(mfccp)
839     struct mfcctl *mfccp;
840 {
841     struct mfc *rt;
842     u_long hash;
843     struct rtdetq *rte;
844     register u_short nstl;
845     int s;
846     int i;
847 
848     MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt);
849 
850     /* If an entry already exists, just update the fields */
851     if (rt) {
852 	if (mrtdebug & DEBUG_MFC)
853 	    log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n",
854 		(u_long)ntohl(mfccp->mfcc_origin.s_addr),
855 		(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
856 		mfccp->mfcc_parent);
857 
858 	s = splnet();
859 	rt->mfc_parent = mfccp->mfcc_parent;
860 	for (i = 0; i < numvifs; i++)
861 	    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
862 	splx(s);
863 	return 0;
864     }
865 
866     /*
867      * Find the entry for which the upcall was made and update
868      */
869     s = splnet();
870     hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
871     for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) {
872 
873 	if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
874 	    (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
875 	    (rt->mfc_stall != NULL)) {
876 
877 	    if (nstl++)
878 		log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n",
879 		    "multiple kernel entries",
880 		    (u_long)ntohl(mfccp->mfcc_origin.s_addr),
881 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
882 		    mfccp->mfcc_parent, (void *)rt->mfc_stall);
883 
884 	    if (mrtdebug & DEBUG_MFC)
885 		log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n",
886 		    (u_long)ntohl(mfccp->mfcc_origin.s_addr),
887 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
888 		    mfccp->mfcc_parent, (void *)rt->mfc_stall);
889 
890 	    rt->mfc_origin     = mfccp->mfcc_origin;
891 	    rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
892 	    rt->mfc_parent     = mfccp->mfcc_parent;
893 	    for (i = 0; i < numvifs; i++)
894 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
895 	    /* initialize pkt counters per src-grp */
896 	    rt->mfc_pkt_cnt    = 0;
897 	    rt->mfc_byte_cnt   = 0;
898 	    rt->mfc_wrong_if   = 0;
899 	    rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
900 
901 	    rt->mfc_expire = 0;	/* Don't clean this guy up */
902 	    nexpire[hash]--;
903 
904 	    /* free packets Qed at the end of this entry */
905 	    for (rte = rt->mfc_stall; rte != NULL; ) {
906 		struct rtdetq *n = rte->next;
907 
908 		ip_mdq(rte->m, rte->ifp, rt, -1);
909 		m_freem(rte->m);
910 #ifdef UPCALL_TIMING
911 		collate(&(rte->t));
912 #endif /* UPCALL_TIMING */
913 		free(rte, M_MRTABLE);
914 		rte = n;
915 	    }
916 	    rt->mfc_stall = NULL;
917 	}
918     }
919 
920     /*
921      * It is possible that an entry is being inserted without an upcall
922      */
923     if (nstl == 0) {
924 	if (mrtdebug & DEBUG_MFC)
925 	    log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n",
926 		hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr),
927 		(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
928 		mfccp->mfcc_parent);
929 
930 	for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) {
931 
932 	    if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
933 		(rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
934 
935 		rt->mfc_origin     = mfccp->mfcc_origin;
936 		rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
937 		rt->mfc_parent     = mfccp->mfcc_parent;
938 		for (i = 0; i < numvifs; i++)
939 		    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
940 		/* initialize pkt counters per src-grp */
941 		rt->mfc_pkt_cnt    = 0;
942 		rt->mfc_byte_cnt   = 0;
943 		rt->mfc_wrong_if   = 0;
944 		rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
945 		if (rt->mfc_expire)
946 		    nexpire[hash]--;
947 		rt->mfc_expire	   = 0;
948 	    }
949 	}
950 	if (rt == NULL) {
951 	    /* no upcall, so make a new entry */
952 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
953 	    if (rt == NULL) {
954 		splx(s);
955 		return ENOBUFS;
956 	    }
957 
958 	    /* insert new entry at head of hash chain */
959 	    rt->mfc_origin     = mfccp->mfcc_origin;
960 	    rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
961 	    rt->mfc_parent     = mfccp->mfcc_parent;
962 	    for (i = 0; i < numvifs; i++)
963 		    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
964 	    /* initialize pkt counters per src-grp */
965 	    rt->mfc_pkt_cnt    = 0;
966 	    rt->mfc_byte_cnt   = 0;
967 	    rt->mfc_wrong_if   = 0;
968 	    rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
969 	    rt->mfc_expire     = 0;
970 	    rt->mfc_stall      = NULL;
971 
972 	    /* link into table */
973 	    rt->mfc_next = mfctable[hash];
974 	    mfctable[hash] = rt;
975 	}
976     }
977     splx(s);
978     return 0;
979 }
980 
981 #ifdef UPCALL_TIMING
982 /*
983  * collect delay statistics on the upcalls
984  */
985 static void collate(t)
986 register struct timeval *t;
987 {
988     register u_long d;
989     register struct timeval tp;
990     register u_long delta;
991 
992     GET_TIME(tp);
993 
994     if (TV_LT(*t, tp))
995     {
996 	TV_DELTA(tp, *t, delta);
997 
998 	d = delta >> 10;
999 	if (d > 50)
1000 	    d = 50;
1001 
1002 	++upcall_data[d];
1003     }
1004 }
1005 #endif /* UPCALL_TIMING */
1006 
1007 /*
1008  * Delete an mfc entry
1009  */
1010 static int
1011 del_mfc(mfccp)
1012     struct mfcctl *mfccp;
1013 {
1014     struct in_addr 	origin;
1015     struct in_addr 	mcastgrp;
1016     struct mfc 		*rt;
1017     struct mfc	 	**nptr;
1018     u_long 		hash;
1019     int s;
1020 
1021     origin = mfccp->mfcc_origin;
1022     mcastgrp = mfccp->mfcc_mcastgrp;
1023     hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1024 
1025     if (mrtdebug & DEBUG_MFC)
1026 	log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n",
1027 	    (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1028 
1029     s = splnet();
1030 
1031     nptr = &mfctable[hash];
1032     while ((rt = *nptr) != NULL) {
1033 	if (origin.s_addr == rt->mfc_origin.s_addr &&
1034 	    mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1035 	    rt->mfc_stall == NULL)
1036 	    break;
1037 
1038 	nptr = &rt->mfc_next;
1039     }
1040     if (rt == NULL) {
1041 	splx(s);
1042 	return EADDRNOTAVAIL;
1043     }
1044 
1045     *nptr = rt->mfc_next;
1046     free(rt, M_MRTABLE);
1047 
1048     splx(s);
1049 
1050     return 0;
1051 }
1052 
1053 /*
1054  * Send a message to mrouted on the multicast routing socket
1055  */
1056 static int
1057 socket_send(s, mm, src)
1058 	struct socket *s;
1059 	struct mbuf *mm;
1060 	struct sockaddr_in *src;
1061 {
1062 	if (s) {
1063 		if (sbappendaddr(&s->so_rcv,
1064 				 (struct sockaddr *)src,
1065 				 mm, (struct mbuf *)0) != 0) {
1066 			sorwakeup(s);
1067 			return 0;
1068 		}
1069 	}
1070 	m_freem(mm);
1071 	return -1;
1072 }
1073 
1074 /*
1075  * IP multicast forwarding function. This function assumes that the packet
1076  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1077  * pointed to by "ifp", and the packet is to be relayed to other networks
1078  * that have members of the packet's destination IP multicast group.
1079  *
1080  * The packet is returned unscathed to the caller, unless it is
1081  * erroneous, in which case a non-zero return value tells the caller to
1082  * discard it.
1083  */
1084 
1085 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1086 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1087 
1088 static int
1089 X_ip_mforward(ip, ifp, m, imo)
1090     register struct ip *ip;
1091     struct ifnet *ifp;
1092     struct mbuf *m;
1093     struct ip_moptions *imo;
1094 {
1095     register struct mfc *rt;
1096     register u_char *ipoptions;
1097     static struct sockaddr_in 	k_igmpsrc	= { sizeof k_igmpsrc, AF_INET };
1098     static int srctun = 0;
1099     register struct mbuf *mm;
1100     int s;
1101     vifi_t vifi;
1102     struct vif *vifp;
1103 
1104     if (mrtdebug & DEBUG_FORWARD)
1105 	log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n",
1106 	    (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr),
1107 	    (void *)ifp);
1108 
1109     if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1110 	(ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1111 	/*
1112 	 * Packet arrived via a physical interface or
1113 	 * an encapsulated tunnel.
1114 	 */
1115     } else {
1116 	/*
1117 	 * Packet arrived through a source-route tunnel.
1118 	 * Source-route tunnels are no longer supported.
1119 	 */
1120 	if ((srctun++ % 1000) == 0)
1121 	    log(LOG_ERR,
1122 		"ip_mforward: received source-routed packet from %lx\n",
1123 		(u_long)ntohl(ip->ip_src.s_addr));
1124 
1125 	return 1;
1126     }
1127 
1128     if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1129 	if (ip->ip_ttl < 255)
1130 		ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1131 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1132 	    vifp = viftable + vifi;
1133 	    printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n",
1134 		ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi,
1135 		(vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1136 		vifp->v_ifp->if_name, vifp->v_ifp->if_unit);
1137 	}
1138 	return (ip_mdq(m, ifp, NULL, vifi));
1139     }
1140     if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1141 	printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1142 	    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr));
1143 	if(!imo)
1144 		printf("In fact, no options were specified at all\n");
1145     }
1146 
1147     /*
1148      * Don't forward a packet with time-to-live of zero or one,
1149      * or a packet destined to a local-only group.
1150      */
1151     if (ip->ip_ttl <= 1 ||
1152 	ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
1153 	return 0;
1154 
1155     /*
1156      * Determine forwarding vifs from the forwarding cache table
1157      */
1158     s = splnet();
1159     MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt);
1160 
1161     /* Entry exists, so forward if necessary */
1162     if (rt != NULL) {
1163 	splx(s);
1164 	return (ip_mdq(m, ifp, rt, -1));
1165     } else {
1166 	/*
1167 	 * If we don't have a route for packet's origin,
1168 	 * Make a copy of the packet &
1169 	 * send message to routing daemon
1170 	 */
1171 
1172 	register struct mbuf *mb0;
1173 	register struct rtdetq *rte;
1174 	register u_long hash;
1175 	int hlen = ip->ip_hl << 2;
1176 #ifdef UPCALL_TIMING
1177 	struct timeval tp;
1178 
1179 	GET_TIME(tp);
1180 #endif
1181 
1182 	mrtstat.mrts_no_route++;
1183 	if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1184 	    log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n",
1185 		(u_long)ntohl(ip->ip_src.s_addr),
1186 		(u_long)ntohl(ip->ip_dst.s_addr));
1187 
1188 	/*
1189 	 * Allocate mbufs early so that we don't do extra work if we are
1190 	 * just going to fail anyway.  Make sure to pullup the header so
1191 	 * that other people can't step on it.
1192 	 */
1193 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT);
1194 	if (rte == NULL) {
1195 	    splx(s);
1196 	    return ENOBUFS;
1197 	}
1198 	mb0 = m_copy(m, 0, M_COPYALL);
1199 	if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1200 	    mb0 = m_pullup(mb0, hlen);
1201 	if (mb0 == NULL) {
1202 	    free(rte, M_MRTABLE);
1203 	    splx(s);
1204 	    return ENOBUFS;
1205 	}
1206 
1207 	/* is there an upcall waiting for this packet? */
1208 	hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1209 	for (rt = mfctable[hash]; rt; rt = rt->mfc_next) {
1210 	    if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1211 		(ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1212 		(rt->mfc_stall != NULL))
1213 		break;
1214 	}
1215 
1216 	if (rt == NULL) {
1217 	    int i;
1218 	    struct igmpmsg *im;
1219 
1220 	    /* no upcall, so make a new entry */
1221 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1222 	    if (rt == NULL) {
1223 		free(rte, M_MRTABLE);
1224 		m_freem(mb0);
1225 		splx(s);
1226 		return ENOBUFS;
1227 	    }
1228 	    /* Make a copy of the header to send to the user level process */
1229 	    mm = m_copy(mb0, 0, hlen);
1230 	    if (mm == NULL) {
1231 		free(rte, M_MRTABLE);
1232 		m_freem(mb0);
1233 		free(rt, M_MRTABLE);
1234 		splx(s);
1235 		return ENOBUFS;
1236 	    }
1237 
1238 	    /*
1239 	     * Send message to routing daemon to install
1240 	     * a route into the kernel table
1241 	     */
1242 	    k_igmpsrc.sin_addr = ip->ip_src;
1243 
1244 	    im = mtod(mm, struct igmpmsg *);
1245 	    im->im_msgtype	= IGMPMSG_NOCACHE;
1246 	    im->im_mbz		= 0;
1247 
1248 	    mrtstat.mrts_upcalls++;
1249 
1250 	    if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1251 		log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1252 		++mrtstat.mrts_upq_sockfull;
1253 		free(rte, M_MRTABLE);
1254 		m_freem(mb0);
1255 		free(rt, M_MRTABLE);
1256 		splx(s);
1257 		return ENOBUFS;
1258 	    }
1259 
1260 	    /* insert new entry at head of hash chain */
1261 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1262 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1263 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1264 	    nexpire[hash]++;
1265 	    for (i = 0; i < numvifs; i++)
1266 		rt->mfc_ttls[i] = 0;
1267 	    rt->mfc_parent = -1;
1268 
1269 	    /* link into table */
1270 	    rt->mfc_next   = mfctable[hash];
1271 	    mfctable[hash] = rt;
1272 	    rt->mfc_stall = rte;
1273 
1274 	} else {
1275 	    /* determine if q has overflowed */
1276 	    int npkts = 0;
1277 	    struct rtdetq **p;
1278 
1279 	    for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1280 		npkts++;
1281 
1282 	    if (npkts > MAX_UPQ) {
1283 		mrtstat.mrts_upq_ovflw++;
1284 		free(rte, M_MRTABLE);
1285 		m_freem(mb0);
1286 		splx(s);
1287 		return 0;
1288 	    }
1289 
1290 	    /* Add this entry to the end of the queue */
1291 	    *p = rte;
1292 	}
1293 
1294 	rte->m 			= mb0;
1295 	rte->ifp 		= ifp;
1296 #ifdef UPCALL_TIMING
1297 	rte->t			= tp;
1298 #endif
1299 	rte->next		= NULL;
1300 
1301 	splx(s);
1302 
1303 	return 0;
1304     }
1305 }
1306 
1307 #ifndef MROUTE_LKM
1308 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
1309 		   struct ip_moptions *) = X_ip_mforward;
1310 #endif
1311 
1312 /*
1313  * Clean up the cache entry if upcall is not serviced
1314  */
1315 static void
1316 expire_upcalls(void *unused)
1317 {
1318     struct rtdetq *rte;
1319     struct mfc *mfc, **nptr;
1320     int i;
1321     int s;
1322 
1323     s = splnet();
1324     for (i = 0; i < MFCTBLSIZ; i++) {
1325 	if (nexpire[i] == 0)
1326 	    continue;
1327 	nptr = &mfctable[i];
1328 	for (mfc = *nptr; mfc != NULL; mfc = *nptr) {
1329 	    /*
1330 	     * Skip real cache entries
1331 	     * Make sure it wasn't marked to not expire (shouldn't happen)
1332 	     * If it expires now
1333 	     */
1334 	    if (mfc->mfc_stall != NULL &&
1335 	        mfc->mfc_expire != 0 &&
1336 		--mfc->mfc_expire == 0) {
1337 		if (mrtdebug & DEBUG_EXPIRE)
1338 		    log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n",
1339 			(u_long)ntohl(mfc->mfc_origin.s_addr),
1340 			(u_long)ntohl(mfc->mfc_mcastgrp.s_addr));
1341 		/*
1342 		 * drop all the packets
1343 		 * free the mbuf with the pkt, if, timing info
1344 		 */
1345 		for (rte = mfc->mfc_stall; rte; ) {
1346 		    struct rtdetq *n = rte->next;
1347 
1348 		    m_freem(rte->m);
1349 		    free(rte, M_MRTABLE);
1350 		    rte = n;
1351 		}
1352 		++mrtstat.mrts_cache_cleanups;
1353 		nexpire[i]--;
1354 
1355 		*nptr = mfc->mfc_next;
1356 		free(mfc, M_MRTABLE);
1357 	    } else {
1358 		nptr = &mfc->mfc_next;
1359 	    }
1360 	}
1361     }
1362     splx(s);
1363     expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
1364 }
1365 
1366 /*
1367  * Packet forwarding routine once entry in the cache is made
1368  */
1369 static int
1370 ip_mdq(m, ifp, rt, xmt_vif)
1371     register struct mbuf *m;
1372     register struct ifnet *ifp;
1373     register struct mfc *rt;
1374     register vifi_t xmt_vif;
1375 {
1376     register struct ip  *ip = mtod(m, struct ip *);
1377     register vifi_t vifi;
1378     register struct vif *vifp;
1379     register int plen = ip->ip_len;
1380 
1381 /*
1382  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1383  * input, they shouldn't get counted on output, so statistics keeping is
1384  * seperate.
1385  */
1386 #define MC_SEND(ip,vifp,m) {                             \
1387                 if ((vifp)->v_flags & VIFF_TUNNEL)  	 \
1388                     encap_send((ip), (vifp), (m));       \
1389                 else                                     \
1390                     phyint_send((ip), (vifp), (m));      \
1391 }
1392 
1393     /*
1394      * If xmt_vif is not -1, send on only the requested vif.
1395      *
1396      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1397      */
1398     if (xmt_vif < numvifs) {
1399 	MC_SEND(ip, viftable + xmt_vif, m);
1400 	return 1;
1401     }
1402 
1403     /*
1404      * Don't forward if it didn't arrive from the parent vif for its origin.
1405      */
1406     vifi = rt->mfc_parent;
1407     if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1408 	/* came in the wrong interface */
1409 	if (mrtdebug & DEBUG_FORWARD)
1410 	    log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1411 		(void *)ifp, vifi, (void *)viftable[vifi].v_ifp);
1412 	++mrtstat.mrts_wrong_if;
1413 	++rt->mfc_wrong_if;
1414 	/*
1415 	 * If we are doing PIM assert processing, and we are forwarding
1416 	 * packets on this interface, and it is a broadcast medium
1417 	 * interface (and not a tunnel), send a message to the routing daemon.
1418 	 */
1419 	if (pim_assert && rt->mfc_ttls[vifi] &&
1420 		(ifp->if_flags & IFF_BROADCAST) &&
1421 		!(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1422 	    struct sockaddr_in k_igmpsrc;
1423 	    struct mbuf *mm;
1424 	    struct igmpmsg *im;
1425 	    int hlen = ip->ip_hl << 2;
1426 	    struct timeval now;
1427 	    register u_long delta;
1428 
1429 	    GET_TIME(now);
1430 
1431 	    TV_DELTA(rt->mfc_last_assert, now, delta);
1432 
1433 	    if (delta > ASSERT_MSG_TIME) {
1434 		mm = m_copy(m, 0, hlen);
1435 		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1436 		    mm = m_pullup(mm, hlen);
1437 		if (mm == NULL) {
1438 		    return ENOBUFS;
1439 		}
1440 
1441 		rt->mfc_last_assert = now;
1442 
1443 		im = mtod(mm, struct igmpmsg *);
1444 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1445 		im->im_mbz		= 0;
1446 		im->im_vif		= vifi;
1447 
1448 		k_igmpsrc.sin_addr = im->im_src;
1449 
1450 		socket_send(ip_mrouter, mm, &k_igmpsrc);
1451 	    }
1452 	}
1453 	return 0;
1454     }
1455 
1456     /* If I sourced this packet, it counts as output, else it was input. */
1457     if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1458 	viftable[vifi].v_pkt_out++;
1459 	viftable[vifi].v_bytes_out += plen;
1460     } else {
1461 	viftable[vifi].v_pkt_in++;
1462 	viftable[vifi].v_bytes_in += plen;
1463     }
1464     rt->mfc_pkt_cnt++;
1465     rt->mfc_byte_cnt += plen;
1466 
1467     /*
1468      * For each vif, decide if a copy of the packet should be forwarded.
1469      * Forward if:
1470      *		- the ttl exceeds the vif's threshold
1471      *		- there are group members downstream on interface
1472      */
1473     for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1474 	if ((rt->mfc_ttls[vifi] > 0) &&
1475 	    (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1476 	    vifp->v_pkt_out++;
1477 	    vifp->v_bytes_out += plen;
1478 	    MC_SEND(ip, vifp, m);
1479 	}
1480 
1481     return 0;
1482 }
1483 
1484 /*
1485  * check if a vif number is legal/ok. This is used by ip_output, to export
1486  * numvifs there,
1487  */
1488 static int
1489 X_legal_vif_num(vif)
1490     int vif;
1491 {
1492     if (vif >= 0 && vif < numvifs)
1493        return(1);
1494     else
1495        return(0);
1496 }
1497 
1498 #ifndef MROUTE_LKM
1499 int (*legal_vif_num)(int) = X_legal_vif_num;
1500 #endif
1501 
1502 /*
1503  * Return the local address used by this vif
1504  */
1505 static u_long
1506 X_ip_mcast_src(vifi)
1507     int vifi;
1508 {
1509     if (vifi >= 0 && vifi < numvifs)
1510 	return viftable[vifi].v_lcl_addr.s_addr;
1511     else
1512 	return INADDR_ANY;
1513 }
1514 
1515 #ifndef MROUTE_LKM
1516 u_long (*ip_mcast_src)(int) = X_ip_mcast_src;
1517 #endif
1518 
1519 static void
1520 phyint_send(ip, vifp, m)
1521     struct ip *ip;
1522     struct vif *vifp;
1523     struct mbuf *m;
1524 {
1525     register struct mbuf *mb_copy;
1526     register int hlen = ip->ip_hl << 2;
1527 
1528     /*
1529      * Make a new reference to the packet; make sure that
1530      * the IP header is actually copied, not just referenced,
1531      * so that ip_output() only scribbles on the copy.
1532      */
1533     mb_copy = m_copy(m, 0, M_COPYALL);
1534     if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1535 	mb_copy = m_pullup(mb_copy, hlen);
1536     if (mb_copy == NULL)
1537 	return;
1538 
1539     if (vifp->v_rate_limit == 0)
1540 	tbf_send_packet(vifp, mb_copy);
1541     else
1542 	tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1543 }
1544 
1545 static void
1546 encap_send(ip, vifp, m)
1547     register struct ip *ip;
1548     register struct vif *vifp;
1549     register struct mbuf *m;
1550 {
1551     register struct mbuf *mb_copy;
1552     register struct ip *ip_copy;
1553     register int i, len = ip->ip_len;
1554 
1555     /*
1556      * copy the old packet & pullup its IP header into the
1557      * new mbuf so we can modify it.  Try to fill the new
1558      * mbuf since if we don't the ethernet driver will.
1559      */
1560     MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER);
1561     if (mb_copy == NULL)
1562 	return;
1563     mb_copy->m_data += max_linkhdr;
1564     mb_copy->m_len = sizeof(multicast_encap_iphdr);
1565 
1566     if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1567 	m_freem(mb_copy);
1568 	return;
1569     }
1570     i = MHLEN - M_LEADINGSPACE(mb_copy);
1571     if (i > len)
1572 	i = len;
1573     mb_copy = m_pullup(mb_copy, i);
1574     if (mb_copy == NULL)
1575 	return;
1576     mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1577 
1578     /*
1579      * fill in the encapsulating IP header.
1580      */
1581     ip_copy = mtod(mb_copy, struct ip *);
1582     *ip_copy = multicast_encap_iphdr;
1583     ip_copy->ip_id = htons(ip_id++);
1584     ip_copy->ip_len += len;
1585     ip_copy->ip_src = vifp->v_lcl_addr;
1586     ip_copy->ip_dst = vifp->v_rmt_addr;
1587 
1588     /*
1589      * turn the encapsulated IP header back into a valid one.
1590      */
1591     ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1592     --ip->ip_ttl;
1593     HTONS(ip->ip_len);
1594     HTONS(ip->ip_off);
1595     ip->ip_sum = 0;
1596     mb_copy->m_data += sizeof(multicast_encap_iphdr);
1597     ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1598     mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1599 
1600     if (vifp->v_rate_limit == 0)
1601 	tbf_send_packet(vifp, mb_copy);
1602     else
1603 	tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1604 }
1605 
1606 /*
1607  * De-encapsulate a packet and feed it back through ip input (this
1608  * routine is called whenever IP gets a packet with proto type
1609  * ENCAP_PROTO and a local destination address).
1610  */
1611 void
1612 #ifdef MROUTE_LKM
1613 X_ipip_input(m, off, proto)
1614 #else
1615 ipip_input(m, off, proto)
1616 #endif
1617 	register struct mbuf *m;
1618 	int off;
1619 	int proto;
1620 {
1621     struct ifnet *ifp = m->m_pkthdr.rcvif;
1622     register struct ip *ip = mtod(m, struct ip *);
1623     register int hlen = ip->ip_hl << 2;
1624     register int s;
1625     register struct ifqueue *ifq;
1626     register struct vif *vifp;
1627 
1628     if (!have_encap_tunnel) {
1629 	    rip_input(m, off, proto);
1630 	    return;
1631     }
1632     /*
1633      * dump the packet if it's not to a multicast destination or if
1634      * we don't have an encapsulating tunnel with the source.
1635      * Note:  This code assumes that the remote site IP address
1636      * uniquely identifies the tunnel (i.e., that this site has
1637      * at most one tunnel with the remote site).
1638      */
1639     if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) {
1640 	++mrtstat.mrts_bad_tunnel;
1641 	m_freem(m);
1642 	return;
1643     }
1644     if (ip->ip_src.s_addr != last_encap_src) {
1645 	register struct vif *vife;
1646 
1647 	vifp = viftable;
1648 	vife = vifp + numvifs;
1649 	last_encap_src = ip->ip_src.s_addr;
1650 	last_encap_vif = 0;
1651 	for ( ; vifp < vife; ++vifp)
1652 	    if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
1653 		if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT))
1654 		    == VIFF_TUNNEL)
1655 		    last_encap_vif = vifp;
1656 		break;
1657 	    }
1658     }
1659     if ((vifp = last_encap_vif) == 0) {
1660 	last_encap_src = 0;
1661 	mrtstat.mrts_cant_tunnel++; /*XXX*/
1662 	m_freem(m);
1663 	if (mrtdebug)
1664 	  log(LOG_DEBUG, "ip_mforward: no tunnel with %lx\n",
1665 		(u_long)ntohl(ip->ip_src.s_addr));
1666 	return;
1667     }
1668     ifp = vifp->v_ifp;
1669 
1670     if (hlen > IP_HDR_LEN)
1671       ip_stripoptions(m, (struct mbuf *) 0);
1672     m->m_data += IP_HDR_LEN;
1673     m->m_len -= IP_HDR_LEN;
1674     m->m_pkthdr.len -= IP_HDR_LEN;
1675     m->m_pkthdr.rcvif = ifp;
1676 
1677     ifq = &ipintrq;
1678     s = splimp();
1679     if (IF_QFULL(ifq)) {
1680 	IF_DROP(ifq);
1681 	m_freem(m);
1682     } else {
1683 	IF_ENQUEUE(ifq, m);
1684 	/*
1685 	 * normally we would need a "schednetisr(NETISR_IP)"
1686 	 * here but we were called by ip_input and it is going
1687 	 * to loop back & try to dequeue the packet we just
1688 	 * queued as soon as we return so we avoid the
1689 	 * unnecessary software interrrupt.
1690 	 */
1691     }
1692     splx(s);
1693 }
1694 
1695 /*
1696  * Token bucket filter module
1697  */
1698 
1699 static void
1700 tbf_control(vifp, m, ip, p_len)
1701 	register struct vif *vifp;
1702 	register struct mbuf *m;
1703 	register struct ip *ip;
1704 	register u_long p_len;
1705 {
1706     register struct tbf *t = vifp->v_tbf;
1707 
1708     if (p_len > MAX_BKT_SIZE) {
1709 	/* drop if packet is too large */
1710 	mrtstat.mrts_pkt2large++;
1711 	m_freem(m);
1712 	return;
1713     }
1714 
1715     tbf_update_tokens(vifp);
1716 
1717     /* if there are enough tokens,
1718      * and the queue is empty,
1719      * send this packet out
1720      */
1721 
1722     if (t->tbf_q_len == 0) {
1723 	/* queue empty, send packet if enough tokens */
1724 	if (p_len <= t->tbf_n_tok) {
1725 	    t->tbf_n_tok -= p_len;
1726 	    tbf_send_packet(vifp, m);
1727 	} else {
1728 	    /* queue packet and timeout till later */
1729 	    tbf_queue(vifp, m);
1730 	    timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1731 	}
1732     } else if (t->tbf_q_len < t->tbf_max_q_len) {
1733 	/* finite queue length, so queue pkts and process queue */
1734 	tbf_queue(vifp, m);
1735 	tbf_process_q(vifp);
1736     } else {
1737 	/* queue length too much, try to dq and queue and process */
1738 	if (!tbf_dq_sel(vifp, ip)) {
1739 	    mrtstat.mrts_q_overflow++;
1740 	    m_freem(m);
1741 	    return;
1742 	} else {
1743 	    tbf_queue(vifp, m);
1744 	    tbf_process_q(vifp);
1745 	}
1746     }
1747     return;
1748 }
1749 
1750 /*
1751  * adds a packet to the queue at the interface
1752  */
1753 static void
1754 tbf_queue(vifp, m)
1755 	register struct vif *vifp;
1756 	register struct mbuf *m;
1757 {
1758     register int s = splnet();
1759     register struct tbf *t = vifp->v_tbf;
1760 
1761     if (t->tbf_t == NULL) {
1762 	/* Queue was empty */
1763 	t->tbf_q = m;
1764     } else {
1765 	/* Insert at tail */
1766 	t->tbf_t->m_act = m;
1767     }
1768 
1769     /* Set new tail pointer */
1770     t->tbf_t = m;
1771 
1772 #ifdef DIAGNOSTIC
1773     /* Make sure we didn't get fed a bogus mbuf */
1774     if (m->m_act)
1775 	panic("tbf_queue: m_act");
1776 #endif
1777     m->m_act = NULL;
1778 
1779     t->tbf_q_len++;
1780 
1781     splx(s);
1782 }
1783 
1784 
1785 /*
1786  * processes the queue at the interface
1787  */
1788 static void
1789 tbf_process_q(vifp)
1790     register struct vif *vifp;
1791 {
1792     register struct mbuf *m;
1793     register int len;
1794     register int s = splnet();
1795     register struct tbf *t = vifp->v_tbf;
1796 
1797     /* loop through the queue at the interface and send as many packets
1798      * as possible
1799      */
1800     while (t->tbf_q_len > 0) {
1801 	m = t->tbf_q;
1802 
1803 	len = mtod(m, struct ip *)->ip_len;
1804 
1805 	/* determine if the packet can be sent */
1806 	if (len <= t->tbf_n_tok) {
1807 	    /* if so,
1808 	     * reduce no of tokens, dequeue the packet,
1809 	     * send the packet.
1810 	     */
1811 	    t->tbf_n_tok -= len;
1812 
1813 	    t->tbf_q = m->m_act;
1814 	    if (--t->tbf_q_len == 0)
1815 		t->tbf_t = NULL;
1816 
1817 	    m->m_act = NULL;
1818 	    tbf_send_packet(vifp, m);
1819 
1820 	} else break;
1821     }
1822     splx(s);
1823 }
1824 
1825 static void
1826 tbf_reprocess_q(xvifp)
1827 	void *xvifp;
1828 {
1829     register struct vif *vifp = xvifp;
1830     if (ip_mrouter == NULL)
1831 	return;
1832 
1833     tbf_update_tokens(vifp);
1834 
1835     tbf_process_q(vifp);
1836 
1837     if (vifp->v_tbf->tbf_q_len)
1838 	timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1839 }
1840 
1841 /* function that will selectively discard a member of the queue
1842  * based on the precedence value and the priority
1843  */
1844 static int
1845 tbf_dq_sel(vifp, ip)
1846     register struct vif *vifp;
1847     register struct ip *ip;
1848 {
1849     register int s = splnet();
1850     register u_int p;
1851     register struct mbuf *m, *last;
1852     register struct mbuf **np;
1853     register struct tbf *t = vifp->v_tbf;
1854 
1855     p = priority(vifp, ip);
1856 
1857     np = &t->tbf_q;
1858     last = NULL;
1859     while ((m = *np) != NULL) {
1860 	if (p > priority(vifp, mtod(m, struct ip *))) {
1861 	    *np = m->m_act;
1862 	    /* If we're removing the last packet, fix the tail pointer */
1863 	    if (m == t->tbf_t)
1864 		t->tbf_t = last;
1865 	    m_freem(m);
1866 	    /* it's impossible for the queue to be empty, but
1867 	     * we check anyway. */
1868 	    if (--t->tbf_q_len == 0)
1869 		t->tbf_t = NULL;
1870 	    splx(s);
1871 	    mrtstat.mrts_drop_sel++;
1872 	    return(1);
1873 	}
1874 	np = &m->m_act;
1875 	last = m;
1876     }
1877     splx(s);
1878     return(0);
1879 }
1880 
1881 static void
1882 tbf_send_packet(vifp, m)
1883     register struct vif *vifp;
1884     register struct mbuf *m;
1885 {
1886     struct ip_moptions imo;
1887     int error;
1888     static struct route ro;
1889     int s = splnet();
1890 
1891     if (vifp->v_flags & VIFF_TUNNEL) {
1892 	/* If tunnel options */
1893 	ip_output(m, (struct mbuf *)0, &vifp->v_route,
1894 		  IP_FORWARDING, (struct ip_moptions *)0);
1895     } else {
1896 	imo.imo_multicast_ifp  = vifp->v_ifp;
1897 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1898 	imo.imo_multicast_loop = 1;
1899 	imo.imo_multicast_vif  = -1;
1900 
1901 	/*
1902 	 * Re-entrancy should not be a problem here, because
1903 	 * the packets that we send out and are looped back at us
1904 	 * should get rejected because they appear to come from
1905 	 * the loopback interface, thus preventing looping.
1906 	 */
1907 	error = ip_output(m, (struct mbuf *)0, &ro,
1908 			  IP_FORWARDING, &imo);
1909 
1910 	if (mrtdebug & DEBUG_XMIT)
1911 	    log(LOG_DEBUG, "phyint_send on vif %d err %d\n",
1912 		vifp - viftable, error);
1913     }
1914     splx(s);
1915 }
1916 
1917 /* determine the current time and then
1918  * the elapsed time (between the last time and time now)
1919  * in milliseconds & update the no. of tokens in the bucket
1920  */
1921 static void
1922 tbf_update_tokens(vifp)
1923     register struct vif *vifp;
1924 {
1925     struct timeval tp;
1926     register u_long tm;
1927     register int s = splnet();
1928     register struct tbf *t = vifp->v_tbf;
1929 
1930     GET_TIME(tp);
1931 
1932     TV_DELTA(tp, t->tbf_last_pkt_t, tm);
1933 
1934     /*
1935      * This formula is actually
1936      * "time in seconds" * "bytes/second".
1937      *
1938      * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
1939      *
1940      * The (1000/1024) was introduced in add_vif to optimize
1941      * this divide into a shift.
1942      */
1943     t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8;
1944     t->tbf_last_pkt_t = tp;
1945 
1946     if (t->tbf_n_tok > MAX_BKT_SIZE)
1947 	t->tbf_n_tok = MAX_BKT_SIZE;
1948 
1949     splx(s);
1950 }
1951 
1952 static int
1953 priority(vifp, ip)
1954     register struct vif *vifp;
1955     register struct ip *ip;
1956 {
1957     register int prio;
1958 
1959     /* temporary hack; may add general packet classifier some day */
1960 
1961     /*
1962      * The UDP port space is divided up into four priority ranges:
1963      * [0, 16384)     : unclassified - lowest priority
1964      * [16384, 32768) : audio - highest priority
1965      * [32768, 49152) : whiteboard - medium priority
1966      * [49152, 65536) : video - low priority
1967      */
1968     if (ip->ip_p == IPPROTO_UDP) {
1969 	struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
1970 	switch (ntohs(udp->uh_dport) & 0xc000) {
1971 	    case 0x4000:
1972 		prio = 70;
1973 		break;
1974 	    case 0x8000:
1975 		prio = 60;
1976 		break;
1977 	    case 0xc000:
1978 		prio = 55;
1979 		break;
1980 	    default:
1981 		prio = 50;
1982 		break;
1983 	}
1984 	if (tbfdebug > 1)
1985 		log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio);
1986     } else {
1987 	    prio = 50;
1988     }
1989     return prio;
1990 }
1991 
1992 /*
1993  * End of token bucket filter modifications
1994  */
1995 
1996 int
1997 ip_rsvp_vif_init(so, sopt)
1998 	struct socket *so;
1999 	struct sockopt *sopt;
2000 {
2001     int error, i, s;
2002 
2003     if (rsvpdebug)
2004 	printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2005 	       so->so_type, so->so_proto->pr_protocol);
2006 
2007     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2008 	return EOPNOTSUPP;
2009 
2010     /* Check mbuf. */
2011     error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
2012     if (error)
2013 	    return (error);
2014 
2015     if (rsvpdebug)
2016 	printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", i, rsvp_on);
2017 
2018     s = splnet();
2019 
2020     /* Check vif. */
2021     if (!legal_vif_num(i)) {
2022 	splx(s);
2023 	return EADDRNOTAVAIL;
2024     }
2025 
2026     /* Check if socket is available. */
2027     if (viftable[i].v_rsvpd != NULL) {
2028 	splx(s);
2029 	return EADDRINUSE;
2030     }
2031 
2032     viftable[i].v_rsvpd = so;
2033     /* This may seem silly, but we need to be sure we don't over-increment
2034      * the RSVP counter, in case something slips up.
2035      */
2036     if (!viftable[i].v_rsvp_on) {
2037 	viftable[i].v_rsvp_on = 1;
2038 	rsvp_on++;
2039     }
2040 
2041     splx(s);
2042     return 0;
2043 }
2044 
2045 int
2046 ip_rsvp_vif_done(so, sopt)
2047 	struct socket *so;
2048 	struct sockopt *sopt;
2049 {
2050 	int error, i, s;
2051 
2052 	if (rsvpdebug)
2053 		printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2054 		       so->so_type, so->so_proto->pr_protocol);
2055 
2056 	if (so->so_type != SOCK_RAW ||
2057 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2058 		return EOPNOTSUPP;
2059 
2060 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
2061 	if (error)
2062 		return (error);
2063 
2064 	s = splnet();
2065 
2066 	/* Check vif. */
2067 	if (!legal_vif_num(i)) {
2068 		splx(s);
2069 		return EADDRNOTAVAIL;
2070 	}
2071 
2072 	if (rsvpdebug)
2073 		printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n",
2074 		       viftable[i].v_rsvpd, so);
2075 
2076 	viftable[i].v_rsvpd = NULL;
2077 	/*
2078 	 * This may seem silly, but we need to be sure we don't over-decrement
2079 	 * the RSVP counter, in case something slips up.
2080 	 */
2081 	if (viftable[i].v_rsvp_on) {
2082 		viftable[i].v_rsvp_on = 0;
2083 		rsvp_on--;
2084 	}
2085 
2086 	splx(s);
2087 	return 0;
2088 }
2089 
2090 void
2091 ip_rsvp_force_done(so)
2092     struct socket *so;
2093 {
2094     int vifi;
2095     register int s;
2096 
2097     /* Don't bother if it is not the right type of socket. */
2098     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2099 	return;
2100 
2101     s = splnet();
2102 
2103     /* The socket may be attached to more than one vif...this
2104      * is perfectly legal.
2105      */
2106     for (vifi = 0; vifi < numvifs; vifi++) {
2107 	if (viftable[vifi].v_rsvpd == so) {
2108 	    viftable[vifi].v_rsvpd = NULL;
2109 	    /* This may seem silly, but we need to be sure we don't
2110 	     * over-decrement the RSVP counter, in case something slips up.
2111 	     */
2112 	    if (viftable[vifi].v_rsvp_on) {
2113 		viftable[vifi].v_rsvp_on = 0;
2114 		rsvp_on--;
2115 	    }
2116 	}
2117     }
2118 
2119     splx(s);
2120     return;
2121 }
2122 
2123 void
2124 rsvp_input(m, off, proto)
2125 	struct mbuf *m;
2126 	int off;
2127 	int proto;
2128 {
2129     int vifi;
2130     register struct ip *ip = mtod(m, struct ip *);
2131     static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
2132     register int s;
2133     struct ifnet *ifp;
2134 
2135     if (rsvpdebug)
2136 	printf("rsvp_input: rsvp_on %d\n",rsvp_on);
2137 
2138     /* Can still get packets with rsvp_on = 0 if there is a local member
2139      * of the group to which the RSVP packet is addressed.  But in this
2140      * case we want to throw the packet away.
2141      */
2142     if (!rsvp_on) {
2143 	m_freem(m);
2144 	return;
2145     }
2146 
2147     /* If the old-style non-vif-associated socket is set, then use
2148      * it and ignore the new ones.
2149      */
2150     if (ip_rsvpd != NULL) {
2151 	if (rsvpdebug)
2152 	    printf("rsvp_input: Sending packet up old-style socket\n");
2153 	rip_input(m, off, proto);  /* xxx */
2154 	return;
2155     }
2156 
2157     s = splnet();
2158 
2159     if (rsvpdebug)
2160 	printf("rsvp_input: check vifs\n");
2161 
2162 #ifdef DIAGNOSTIC
2163     if (!(m->m_flags & M_PKTHDR))
2164 	    panic("rsvp_input no hdr");
2165 #endif
2166 
2167     ifp = m->m_pkthdr.rcvif;
2168     /* Find which vif the packet arrived on. */
2169     for (vifi = 0; vifi < numvifs; vifi++) {
2170 	if (viftable[vifi].v_ifp == ifp)
2171  		break;
2172  	}
2173 
2174     if (vifi == numvifs) {
2175 	/* Can't find vif packet arrived on. Drop packet. */
2176 	if (rsvpdebug)
2177 	    printf("rsvp_input: Can't find vif for packet...dropping it.\n");
2178 	m_freem(m);
2179 	splx(s);
2180 	return;
2181     }
2182 
2183     if (rsvpdebug)
2184 	printf("rsvp_input: check socket\n");
2185 
2186     if (viftable[vifi].v_rsvpd == NULL) {
2187 	/* drop packet, since there is no specific socket for this
2188 	 * interface */
2189 	    if (rsvpdebug)
2190 		    printf("rsvp_input: No socket defined for vif %d\n",vifi);
2191 	    m_freem(m);
2192 	    splx(s);
2193 	    return;
2194     }
2195     rsvp_src.sin_addr = ip->ip_src;
2196 
2197     if (rsvpdebug && m)
2198 	printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n",
2199 	       m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
2200 
2201     if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) {
2202 	if (rsvpdebug)
2203 	    printf("rsvp_input: Failed to append to socket\n");
2204     } else {
2205 	if (rsvpdebug)
2206 	    printf("rsvp_input: send packet up\n");
2207     }
2208 
2209     splx(s);
2210 }
2211 
2212 #ifdef MROUTE_LKM
2213 #include <sys/conf.h>
2214 #include <sys/exec.h>
2215 #include <sys/sysent.h>
2216 #include <sys/lkm.h>
2217 
2218 MOD_MISC("ip_mroute_mod")
2219 
2220 static int
2221 ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd)
2222 {
2223 	int i;
2224 	struct lkm_misc	*args = lkmtp->private.lkm_misc;
2225 	int err = 0;
2226 
2227 	switch(cmd) {
2228 		static int (*old_ip_mrouter_cmd)();
2229 		static int (*old_ip_mrouter_done)();
2230 		static int (*old_ip_mforward)();
2231 		static int (*old_mrt_ioctl)();
2232 		static void (*old_proto4_input)();
2233 		static int (*old_legal_vif_num)();
2234 		extern struct protosw inetsw[];
2235 
2236 	case LKM_E_LOAD:
2237 		if(lkmexists(lkmtp) || ip_mrtproto)
2238 		  return(EEXIST);
2239 		old_ip_mrouter_cmd = ip_mrouter_cmd;
2240 		ip_mrouter_cmd = X_ip_mrouter_cmd;
2241 		old_ip_mrouter_done = ip_mrouter_done;
2242 		ip_mrouter_done = X_ip_mrouter_done;
2243 		old_ip_mforward = ip_mforward;
2244 		ip_mforward = X_ip_mforward;
2245 		old_mrt_ioctl = mrt_ioctl;
2246 		mrt_ioctl = X_mrt_ioctl;
2247               old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input;
2248               inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input;
2249 		old_legal_vif_num = legal_vif_num;
2250 		legal_vif_num = X_legal_vif_num;
2251 		ip_mrtproto = IGMP_DVMRP;
2252 
2253 		printf("\nIP multicast routing loaded\n");
2254 		break;
2255 
2256 	case LKM_E_UNLOAD:
2257 		if (ip_mrouter)
2258 		  return EINVAL;
2259 
2260 		ip_mrouter_cmd = old_ip_mrouter_cmd;
2261 		ip_mrouter_done = old_ip_mrouter_done;
2262 		ip_mforward = old_ip_mforward;
2263 		mrt_ioctl = old_mrt_ioctl;
2264               inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input;
2265 		legal_vif_num = old_legal_vif_num;
2266 		ip_mrtproto = 0;
2267 		break;
2268 
2269 	default:
2270 		err = EINVAL;
2271 		break;
2272 	}
2273 
2274 	return(err);
2275 }
2276 
2277 int
2278 ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) {
2279 	DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle,
2280 		 nosys);
2281 }
2282 
2283 #endif /* MROUTE_LKM */
2284 #endif /* MROUTING */
2285