xref: /freebsd/sys/netinet/ip_mroute.c (revision 0ea3482342b4d7d6e71f3007ce4dafe445c639fd)
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  * $Id: ip_mroute.c,v 1.23 1995/10/06 19:30:43 wollman Exp $
13  */
14 
15 
16 #include <sys/param.h>
17 #include <sys/systm.h>
18 #include <sys/mbuf.h>
19 #include <sys/socket.h>
20 #include <sys/socketvar.h>
21 #include <sys/protosw.h>
22 #include <sys/errno.h>
23 #include <sys/time.h>
24 #include <sys/kernel.h>
25 #include <sys/ioctl.h>
26 #include <sys/syslog.h>
27 #include <sys/queue.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_pcb.h>
35 #include <netinet/in_var.h>
36 #include <netinet/igmp.h>
37 #include <netinet/igmp_var.h>
38 #include <netinet/ip_mroute.h>
39 #include <netinet/udp.h>
40 
41 #ifndef NTOHL
42 #if BYTE_ORDER != BIG_ENDIAN
43 #define NTOHL(d) ((d) = ntohl((d)))
44 #define NTOHS(d) ((d) = ntohs((u_short)(d)))
45 #define HTONL(d) ((d) = htonl((d)))
46 #define HTONS(d) ((d) = htons((u_short)(d)))
47 #else
48 #define NTOHL(d)
49 #define NTOHS(d)
50 #define HTONL(d)
51 #define HTONS(d)
52 #endif
53 #endif
54 
55 #ifndef MROUTING
56 /*
57  * Dummy routines and globals used when multicast routing is not compiled in.
58  */
59 
60 struct socket  *ip_mrouter  = NULL;
61 u_int		ip_mrtproto = 0;
62 struct mrtstat	mrtstat;
63 u_int		rsvpdebug = 0;
64 
65 int
66 _ip_mrouter_set(cmd, so, m)
67 	int cmd;
68 	struct socket *so;
69 	struct mbuf *m;
70 {
71 	return(EOPNOTSUPP);
72 }
73 
74 int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = _ip_mrouter_set;
75 
76 
77 int
78 _ip_mrouter_get(cmd, so, m)
79 	int cmd;
80 	struct socket *so;
81 	struct mbuf **m;
82 {
83 	return(EOPNOTSUPP);
84 }
85 
86 int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = _ip_mrouter_get;
87 
88 int
89 _ip_mrouter_done()
90 {
91 	return(0);
92 }
93 
94 int (*ip_mrouter_done)(void) = _ip_mrouter_done;
95 
96 int
97 _ip_mforward(ip, ifp, m, imo)
98 	struct ip *ip;
99 	struct ifnet *ifp;
100 	struct mbuf *m;
101 	struct ip_moptions *imo;
102 {
103 	return(0);
104 }
105 
106 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
107 		   struct ip_moptions *) = _ip_mforward;
108 
109 int
110 _mrt_ioctl(int req, caddr_t data, struct proc *p)
111 {
112 	return EOPNOTSUPP;
113 }
114 
115 int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl;
116 
117 void
118 rsvp_input(m, iphlen)		/* XXX must fixup manually */
119 	struct mbuf *m;
120 	int iphlen;
121 {
122     /* Can still get packets with rsvp_on = 0 if there is a local member
123      * of the group to which the RSVP packet is addressed.  But in this
124      * case we want to throw the packet away.
125      */
126     if (!rsvp_on) {
127 	m_freem(m);
128 	return;
129     }
130 
131     if (ip_rsvpd != NULL) {
132 	if (rsvpdebug)
133 	    printf("rsvp_input: Sending packet up old-style socket\n");
134 	rip_input(m);
135 	return;
136     }
137     /* Drop the packet */
138     m_freem(m);
139 }
140 
141 void ipip_input(struct mbuf *m) { /* XXX must fixup manually */
142 	rip_input(m);
143 }
144 
145 int (*legal_vif_num)(int) = 0;
146 
147 /*
148  * This should never be called, since IP_MULTICAST_VIF should fail, but
149  * just in case it does get called, the code a little lower in ip_output
150  * will assign the packet a local address.
151  */
152 u_long
153 _ip_mcast_src(int vifi) { return INADDR_ANY; }
154 u_long (*ip_mcast_src)(int) = _ip_mcast_src;
155 
156 int
157 ip_rsvp_vif_init(so, m)
158     struct socket *so;
159     struct mbuf *m;
160 {
161     return(EINVAL);
162 }
163 
164 int
165 ip_rsvp_vif_done(so, m)
166     struct socket *so;
167     struct mbuf *m;
168 {
169     return(EINVAL);
170 }
171 
172 void
173 ip_rsvp_force_done(so)
174     struct socket *so;
175 {
176     return;
177 }
178 
179 #else /* MROUTING */
180 
181 #define M_HASCL(m)	((m)->m_flags & M_EXT)
182 
183 #define INSIZ		sizeof(struct in_addr)
184 #define	same(a1, a2) \
185 	(bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0)
186 
187 #define MT_MRTABLE MT_RTABLE	/* since nothing else uses it */
188 
189 /*
190  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
191  * except for netstat or debugging purposes.
192  */
193 #ifndef MROUTE_LKM
194 struct socket  *ip_mrouter  = NULL;
195 struct mrtstat	mrtstat;
196 
197 int		ip_mrtproto = IGMP_DVMRP;    /* for netstat only */
198 #else /* MROUTE_LKM */
199 extern struct mrtstat mrtstat;
200 extern int ip_mrtproto;
201 #endif
202 
203 #define NO_RTE_FOUND 	0x1
204 #define RTE_FOUND	0x2
205 
206 struct mbuf    *mfctable[MFCTBLSIZ];
207 u_char		nexpire[MFCTBLSIZ];
208 struct vif	viftable[MAXVIFS];
209 u_int		mrtdebug = 0;	  /* debug level 	*/
210 #define		DEBUG_MFC	0x02
211 #define		DEBUG_FORWARD	0x04
212 #define		DEBUG_EXPIRE	0x08
213 #define		DEBUG_XMIT	0x10
214 u_int       	tbfdebug = 0;     /* tbf debug level 	*/
215 u_int		rsvpdebug = 0;	  /* rsvp debug level   */
216 
217 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
218 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
219 
220 /*
221  * Define the token bucket filter structures
222  * tbftable -> each vif has one of these for storing info
223  */
224 
225 struct tbf tbftable[MAXVIFS];
226 #define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
227 
228 /*
229  * 'Interfaces' associated with decapsulator (so we can tell
230  * packets that went through it from ones that get reflected
231  * by a broken gateway).  These interfaces are never linked into
232  * the system ifnet list & no routes point to them.  I.e., packets
233  * can't be sent this way.  They only exist as a placeholder for
234  * multicast source verification.
235  */
236 struct ifnet multicast_decap_if[MAXVIFS];
237 
238 #define ENCAP_TTL 64
239 #define ENCAP_PROTO IPPROTO_IPIP	/* 4 */
240 
241 /* prototype IP hdr for encapsulated packets */
242 struct ip multicast_encap_iphdr = {
243 #if BYTE_ORDER == LITTLE_ENDIAN
244 	sizeof(struct ip) >> 2, IPVERSION,
245 #else
246 	IPVERSION, sizeof(struct ip) >> 2,
247 #endif
248 	0,				/* tos */
249 	sizeof(struct ip),		/* total length */
250 	0,				/* id */
251 	0,				/* frag offset */
252 	ENCAP_TTL, ENCAP_PROTO,
253 	0,				/* checksum */
254 };
255 
256 /*
257  * Private variables.
258  */
259 static vifi_t	   numvifs = 0;
260 static void (*encap_oldrawip)() = 0;
261 static int have_encap_tunnel = 0;
262 
263 /*
264  * one-back cache used by ipip_input to locate a tunnel's vif
265  * given a datagram's src ip address.
266  */
267 static u_long last_encap_src;
268 static struct vif *last_encap_vif;
269 
270 static int get_sg_cnt(struct sioc_sg_req *);
271 static int get_vif_cnt(struct sioc_vif_req *);
272 int ip_mrouter_init(struct socket *, struct mbuf *);
273 static int add_vif(struct vifctl *);
274 static int del_vif(vifi_t *);
275 static int add_mfc(struct mfcctl *);
276 static int del_mfc(struct mfcctl *);
277 static int get_version(struct mbuf *);
278 static int get_assert(struct mbuf *);
279 static int set_assert(int *);
280 static void expire_upcalls(void *);
281 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *,
282 		  vifi_t);
283 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
284 static void encap_send(struct ip *, struct vif *, struct mbuf *);
285 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
286 static void tbf_queue(struct vif *, struct mbuf *);
287 static void tbf_process_q(struct vif *);
288 static void tbf_reprocess_q(void *);
289 static int tbf_dq_sel(struct vif *, struct ip *);
290 static void tbf_send_packet(struct vif *, struct mbuf *);
291 static void tbf_update_tokens(struct vif *);
292 static int priority(struct vif *, struct ip *);
293 void multiencap_decap(struct mbuf *);
294 
295 /*
296  * whether or not special PIM assert processing is enabled.
297  */
298 static int pim_assert;
299 /*
300  * Rate limit for assert notification messages, in usec
301  */
302 #define ASSERT_MSG_TIME		3000000
303 
304 /*
305  * Hash function for a source, group entry
306  */
307 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
308 			((g) >> 20) ^ ((g) >> 10) ^ (g))
309 
310 /*
311  * Find a route for a given origin IP address and Multicast group address
312  * Type of service parameter to be added in the future!!!
313  */
314 
315 #define MFCFIND(o, g, rt) { \
316 	register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \
317 	register struct mfc *_rt = NULL; \
318 	rt = NULL; \
319 	++mrtstat.mrts_mfc_lookups; \
320 	while (_mb_rt) { \
321 		_rt = mtod(_mb_rt, struct mfc *); \
322 		if ((_rt->mfc_origin.s_addr == o) && \
323 		    (_rt->mfc_mcastgrp.s_addr == g) && \
324 		    (_mb_rt->m_act == NULL)) { \
325 			rt = _rt; \
326 			break; \
327 		} \
328 		_mb_rt = _mb_rt->m_next; \
329 	} \
330 	if (rt == NULL) { \
331 		++mrtstat.mrts_mfc_misses; \
332 	} \
333 }
334 
335 
336 /*
337  * Macros to compute elapsed time efficiently
338  * Borrowed from Van Jacobson's scheduling code
339  */
340 #define TV_DELTA(a, b, delta) { \
341 	    register int xxs; \
342 		\
343 	    delta = (a).tv_usec - (b).tv_usec; \
344 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
345 	       switch (xxs) { \
346 		      case 2: \
347 			  delta += 1000000; \
348 			      /* fall through */ \
349 		      case 1: \
350 			  delta += 1000000; \
351 			  break; \
352 		      default: \
353 			  delta += (1000000 * xxs); \
354 	       } \
355 	    } \
356 }
357 
358 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
359 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
360 
361 #ifdef UPCALL_TIMING
362 u_long upcall_data[51];
363 static void collate(struct timeval *);
364 #endif /* UPCALL_TIMING */
365 
366 
367 /*
368  * Handle MRT setsockopt commands to modify the multicast routing tables.
369  */
370 int
371 X_ip_mrouter_set(cmd, so, m)
372     int cmd;
373     struct socket *so;
374     struct mbuf *m;
375 {
376    if (cmd != MRT_INIT && so != ip_mrouter) return EACCES;
377 
378     switch (cmd) {
379 	case MRT_INIT:     return ip_mrouter_init(so, m);
380 	case MRT_DONE:     return ip_mrouter_done();
381 	case MRT_ADD_VIF:  return add_vif (mtod(m, struct vifctl *));
382 	case MRT_DEL_VIF:  return del_vif (mtod(m, vifi_t *));
383 	case MRT_ADD_MFC:  return add_mfc (mtod(m, struct mfcctl *));
384 	case MRT_DEL_MFC:  return del_mfc (mtod(m, struct mfcctl *));
385 	case MRT_ASSERT:   return set_assert(mtod(m, int *));
386 	default:             return EOPNOTSUPP;
387     }
388 }
389 
390 #ifndef MROUTE_LKM
391 int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = X_ip_mrouter_set;
392 #endif
393 
394 /*
395  * Handle MRT getsockopt commands
396  */
397 int
398 X_ip_mrouter_get(cmd, so, m)
399     int cmd;
400     struct socket *so;
401     struct mbuf **m;
402 {
403     struct mbuf *mb;
404 
405     if (so != ip_mrouter) return EACCES;
406 
407     *m = mb = m_get(M_WAIT, MT_SOOPTS);
408 
409     switch (cmd) {
410 	case MRT_VERSION:   return get_version(mb);
411 	case MRT_ASSERT:    return get_assert(mb);
412 	default:            return EOPNOTSUPP;
413     }
414 }
415 
416 #ifndef MROUTE_LKM
417 int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = X_ip_mrouter_get;
418 #endif
419 
420 /*
421  * Handle ioctl commands to obtain information from the cache
422  */
423 int
424 X_mrt_ioctl(cmd, data)
425     int cmd;
426     caddr_t data;
427 {
428     int error = 0;
429 
430     switch (cmd) {
431 	case (SIOCGETVIFCNT):
432 	    return (get_vif_cnt((struct sioc_vif_req *)data));
433 	    break;
434 	case (SIOCGETSGCNT):
435 	    return (get_sg_cnt((struct sioc_sg_req *)data));
436 	    break;
437 	default:
438 	    return (EINVAL);
439 	    break;
440     }
441     return error;
442 }
443 
444 #ifndef MROUTE_LKM
445 int (*mrt_ioctl)(int, caddr_t, struct proc *) = X_mrt_ioctl;
446 #endif
447 
448 /*
449  * returns the packet, byte, rpf-failure count for the source group provided
450  */
451 static int
452 get_sg_cnt(req)
453     register struct sioc_sg_req *req;
454 {
455     register struct mfc *rt;
456     int s;
457 
458     s = splnet();
459     MFCFIND(req->src.s_addr, req->grp.s_addr, rt);
460     splx(s);
461     if (rt != NULL) {
462 	req->pktcnt = rt->mfc_pkt_cnt;
463 	req->bytecnt = rt->mfc_byte_cnt;
464 	req->wrong_if = rt->mfc_wrong_if;
465     } else
466 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
467 
468     return 0;
469 }
470 
471 /*
472  * returns the input and output packet and byte counts on the vif provided
473  */
474 static int
475 get_vif_cnt(req)
476     register struct sioc_vif_req *req;
477 {
478     register vifi_t vifi = req->vifi;
479 
480     if (vifi >= numvifs) return EINVAL;
481 
482     req->icount = viftable[vifi].v_pkt_in;
483     req->ocount = viftable[vifi].v_pkt_out;
484     req->ibytes = viftable[vifi].v_bytes_in;
485     req->obytes = viftable[vifi].v_bytes_out;
486 
487     return 0;
488 }
489 
490 /*
491  * Enable multicast routing
492  */
493 int
494 ip_mrouter_init(so, m)
495 	struct socket *so;
496 	struct mbuf *m;
497 {
498     int *v;
499 
500     if (mrtdebug)
501 	log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
502 		so->so_type, so->so_proto->pr_protocol);
503 
504     if (so->so_type != SOCK_RAW ||
505 	so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP;
506 
507     if (!m || (m->m_len != sizeof(int *)))
508 	return ENOPROTOOPT;
509 
510     v = mtod(m, int *);
511     if (*v != 1)
512 	return ENOPROTOOPT;
513 
514     if (ip_mrouter != NULL) return EADDRINUSE;
515 
516     ip_mrouter = so;
517 
518     bzero((caddr_t)mfctable, sizeof(mfctable));
519     bzero((caddr_t)nexpire, sizeof(nexpire));
520 
521     pim_assert = 0;
522 
523     timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
524 
525     if (mrtdebug)
526 	log(LOG_DEBUG, "ip_mrouter_init\n");
527 
528     return 0;
529 }
530 
531 /*
532  * Disable multicast routing
533  */
534 int
535 X_ip_mrouter_done()
536 {
537     vifi_t vifi;
538     int i;
539     struct ifnet *ifp;
540     struct ifreq ifr;
541     struct mbuf *mb_rt;
542     struct mbuf *m;
543     struct rtdetq *rte;
544     int s;
545 
546     s = splnet();
547 
548     /*
549      * For each phyint in use, disable promiscuous reception of all IP
550      * multicasts.
551      */
552     for (vifi = 0; vifi < numvifs; vifi++) {
553 	if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
554 	    !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
555 	    ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
556 	    ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr
557 								= INADDR_ANY;
558 	    ifp = viftable[vifi].v_ifp;
559 	    (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
560 	}
561     }
562     bzero((caddr_t)tbftable, sizeof(tbftable));
563     bzero((caddr_t)viftable, sizeof(viftable));
564     numvifs = 0;
565     pim_assert = 0;
566 
567     untimeout(expire_upcalls, (caddr_t)NULL);
568 
569     /*
570      * Free all multicast forwarding cache entries.
571      */
572     for (i = 0; i < MFCTBLSIZ; i++) {
573 	mb_rt = mfctable[i];
574 	while (mb_rt) {
575 	    if (mb_rt->m_act != NULL) {
576 		while (mb_rt->m_act) {
577 		    m = mb_rt->m_act;
578 		    mb_rt->m_act = m->m_act;
579 		    rte = mtod(m, struct rtdetq *);
580 		    m_freem(rte->m);
581 		    m_free(m);
582 		}
583 	    }
584 	    mb_rt = m_free(mb_rt);
585 	}
586     }
587 
588     bzero((caddr_t)mfctable, sizeof(mfctable));
589 
590     /*
591      * Reset de-encapsulation cache
592      */
593     last_encap_src = NULL;
594     last_encap_vif = NULL;
595     have_encap_tunnel = 0;
596 
597     ip_mrouter = NULL;
598 
599     splx(s);
600 
601     if (mrtdebug)
602 	log(LOG_DEBUG, "ip_mrouter_done\n");
603 
604     return 0;
605 }
606 
607 #ifndef MROUTE_LKM
608 int (*ip_mrouter_done)(void) = X_ip_mrouter_done;
609 #endif
610 
611 static int
612 get_version(mb)
613     struct mbuf *mb;
614 {
615     int *v;
616 
617     v = mtod(mb, int *);
618 
619     *v = 0x0305;	/* XXX !!!! */
620     mb->m_len = sizeof(int);
621 
622     return 0;
623 }
624 
625 /*
626  * Set PIM assert processing global
627  */
628 static int
629 set_assert(i)
630     int *i;
631 {
632     if ((*i != 1) && (*i != 0))
633 	return EINVAL;
634 
635     pim_assert = *i;
636 
637     return 0;
638 }
639 
640 /*
641  * Get PIM assert processing global
642  */
643 static int
644 get_assert(m)
645     struct mbuf *m;
646 {
647     int *i;
648 
649     i = mtod(m, int *);
650 
651     *i = pim_assert;
652 
653     return 0;
654 }
655 
656 /*
657  * Add a vif to the vif table
658  */
659 static int
660 add_vif(vifcp)
661     register struct vifctl *vifcp;
662 {
663     register struct vif *vifp = viftable + vifcp->vifc_vifi;
664     static struct sockaddr_in sin = {sizeof sin, AF_INET};
665     struct ifaddr *ifa;
666     struct ifnet *ifp;
667     struct ifreq ifr;
668     int error, s;
669     struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
670 
671     if (vifcp->vifc_vifi >= MAXVIFS)  return EINVAL;
672     if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE;
673 
674     /* Find the interface with an address in AF_INET family */
675     sin.sin_addr = vifcp->vifc_lcl_addr;
676     ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
677     if (ifa == 0) return EADDRNOTAVAIL;
678     ifp = ifa->ifa_ifp;
679 
680     if (vifcp->vifc_flags & VIFF_TUNNEL) {
681 	if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
682 		/*
683 		 * An encapsulating tunnel is wanted.  Tell ipip_input() to
684 		 * start paying attention to encapsulated packets.
685 		 */
686 		if (have_encap_tunnel == 0) {
687 			have_encap_tunnel = 1;
688 			for (s = 0; s < MAXVIFS; ++s) {
689 				multicast_decap_if[s].if_name = "mdecap";
690 				multicast_decap_if[s].if_unit = s;
691 			}
692 		}
693 		/*
694 		 * Set interface to fake encapsulator interface
695 		 */
696 		ifp = &multicast_decap_if[vifcp->vifc_vifi];
697 		/*
698 		 * Prepare cached route entry
699 		 */
700 		bzero(&vifp->v_route, sizeof(vifp->v_route));
701 	} else {
702 	    log(LOG_ERR, "source routed tunnels not supported\n");
703 	    return EOPNOTSUPP;
704 	}
705     } else {
706 	/* Make sure the interface supports multicast */
707 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
708 	    return EOPNOTSUPP;
709 
710 	/* Enable promiscuous reception of all IP multicasts from the if */
711 	((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
712 	((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
713 	s = splnet();
714 	error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
715 	splx(s);
716 	if (error)
717 	    return error;
718     }
719 
720     s = splnet();
721     /* define parameters for the tbf structure */
722     vifp->v_tbf = v_tbf;
723     GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
724     vifp->v_tbf->tbf_n_tok = 0;
725     vifp->v_tbf->tbf_q_len = 0;
726     vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
727     vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
728 
729     vifp->v_flags     = vifcp->vifc_flags;
730     vifp->v_threshold = vifcp->vifc_threshold;
731     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
732     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
733     vifp->v_ifp       = ifp;
734     /* scaling up here allows division by 1024 in critical code */
735     vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
736     vifp->v_rsvp_on   = 0;
737     vifp->v_rsvpd     = NULL;
738     /* initialize per vif pkt counters */
739     vifp->v_pkt_in    = 0;
740     vifp->v_pkt_out   = 0;
741     vifp->v_bytes_in  = 0;
742     vifp->v_bytes_out = 0;
743     splx(s);
744 
745     /* Adjust numvifs up if the vifi is higher than numvifs */
746     if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
747 
748     if (mrtdebug)
749 	log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
750 	    vifcp->vifc_vifi,
751 	    ntohl(vifcp->vifc_lcl_addr.s_addr),
752 	    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
753 	    ntohl(vifcp->vifc_rmt_addr.s_addr),
754 	    vifcp->vifc_threshold,
755 	    vifcp->vifc_rate_limit);
756 
757     return 0;
758 }
759 
760 /*
761  * Delete a vif from the vif table
762  */
763 static int
764 del_vif(vifip)
765     vifi_t *vifip;
766 {
767     register struct vif *vifp = viftable + *vifip;
768     register vifi_t vifi;
769     register struct mbuf *m;
770     struct ifnet *ifp;
771     struct ifreq ifr;
772     int s;
773 
774     if (*vifip >= numvifs) return EINVAL;
775     if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL;
776 
777     s = splnet();
778 
779     if (!(vifp->v_flags & VIFF_TUNNEL)) {
780 	((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
781 	((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
782 	ifp = vifp->v_ifp;
783 	(*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
784     }
785 
786     if (vifp == last_encap_vif) {
787 	last_encap_vif = 0;
788 	last_encap_src = 0;
789     }
790 
791     /*
792      * Free packets queued at the interface
793      */
794     while (vifp->v_tbf->tbf_q) {
795 	m = vifp->v_tbf->tbf_q;
796 	vifp->v_tbf->tbf_q = m->m_act;
797 	m_freem(m);
798     }
799 
800     bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
801     bzero((caddr_t)vifp, sizeof (*vifp));
802 
803     /* Adjust numvifs down */
804     for (vifi = numvifs; vifi > 0; vifi--)
805 	if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break;
806     numvifs = vifi;
807 
808     splx(s);
809 
810     if (mrtdebug)
811       log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
812 
813     return 0;
814 }
815 
816 /*
817  * Add an mfc entry
818  */
819 static int
820 add_mfc(mfccp)
821     struct mfcctl *mfccp;
822 {
823     struct mfc *rt;
824     register struct mbuf *mb_rt;
825     u_long hash;
826     struct mbuf *mb_ntry;
827     struct rtdetq *rte;
828     register u_short nstl;
829     int s;
830     int i;
831 
832     MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt);
833 
834     /* If an entry already exists, just update the fields */
835     if (rt) {
836 	if (mrtdebug & DEBUG_MFC)
837 	    log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n",
838 		ntohl(mfccp->mfcc_origin.s_addr),
839 		ntohl(mfccp->mfcc_mcastgrp.s_addr),
840 		mfccp->mfcc_parent);
841 
842 	s = splnet();
843 	rt->mfc_parent = mfccp->mfcc_parent;
844 	for (i = 0; i < numvifs; i++)
845 	    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
846 	splx(s);
847 	return 0;
848     }
849 
850     /*
851      * Find the entry for which the upcall was made and update
852      */
853     s = splnet();
854     hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
855     for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) {
856 
857 	rt = mtod(mb_rt, struct mfc *);
858 	if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
859 	    (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
860 	    (mb_rt->m_act != NULL)) {
861 
862 	    if (nstl++)
863 		log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %x\n",
864 		    "multiple kernel entries",
865 		    ntohl(mfccp->mfcc_origin.s_addr),
866 		    ntohl(mfccp->mfcc_mcastgrp.s_addr),
867 		    mfccp->mfcc_parent, mb_rt->m_act);
868 
869 	    if (mrtdebug & DEBUG_MFC)
870 		log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %x\n",
871 		    ntohl(mfccp->mfcc_origin.s_addr),
872 		    ntohl(mfccp->mfcc_mcastgrp.s_addr),
873 		    mfccp->mfcc_parent, mb_rt->m_act);
874 
875 	    rt->mfc_origin     = mfccp->mfcc_origin;
876 	    rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
877 	    rt->mfc_parent     = mfccp->mfcc_parent;
878 	    for (i = 0; i < numvifs; i++)
879 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
880 	    /* initialize pkt counters per src-grp */
881 	    rt->mfc_pkt_cnt    = 0;
882 	    rt->mfc_byte_cnt   = 0;
883 	    rt->mfc_wrong_if   = 0;
884 	    rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
885 
886 	    rt->mfc_expire = 0;	/* Don't clean this guy up */
887 	    nexpire[hash]--;
888 
889 	    /* free packets Qed at the end of this entry */
890 	    while (mb_rt->m_act) {
891 		mb_ntry = mb_rt->m_act;
892 		rte = mtod(mb_ntry, struct rtdetq *);
893 /* #ifdef RSVP_ISI */
894 		ip_mdq(rte->m, rte->ifp, rt, -1);
895 /* #endif */
896 		mb_rt->m_act = mb_ntry->m_act;
897 		m_freem(rte->m);
898 #ifdef UPCALL_TIMING
899 		collate(&(rte->t));
900 #endif /* UPCALL_TIMING */
901 		m_free(mb_ntry);
902 	    }
903 	}
904     }
905 
906     /*
907      * It is possible that an entry is being inserted without an upcall
908      */
909     if (nstl == 0) {
910 	if (mrtdebug & DEBUG_MFC)
911 	    log(LOG_DEBUG,"add_mfc no upcall h %d o %x g %x p %x\n",
912 		hash, ntohl(mfccp->mfcc_origin.s_addr),
913 		ntohl(mfccp->mfcc_mcastgrp.s_addr),
914 		mfccp->mfcc_parent);
915 
916 	for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
917 
918 	    rt = mtod(mb_rt, struct mfc *);
919 	    if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
920 		(rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
921 
922 		rt->mfc_origin     = mfccp->mfcc_origin;
923 		rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
924 		rt->mfc_parent     = mfccp->mfcc_parent;
925 		for (i = 0; i < numvifs; i++)
926 		    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
927 		/* initialize pkt counters per src-grp */
928 		rt->mfc_pkt_cnt    = 0;
929 		rt->mfc_byte_cnt   = 0;
930 		rt->mfc_wrong_if   = 0;
931 		rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
932 		if (rt->mfc_expire)
933 		    nexpire[hash]--;
934 		rt->mfc_expire	   = 0;
935 	    }
936 	}
937 	if (mb_rt == NULL) {
938 	    /* no upcall, so make a new entry */
939 	    MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
940 	    if (mb_rt == NULL) {
941 		splx(s);
942 		return ENOBUFS;
943 	    }
944 
945 	    rt = mtod(mb_rt, struct mfc *);
946 
947 	    /* insert new entry at head of hash chain */
948 	    rt->mfc_origin     = mfccp->mfcc_origin;
949 	    rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
950 	    rt->mfc_parent     = mfccp->mfcc_parent;
951 	    for (i = 0; i < numvifs; i++)
952 		    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
953 	    /* initialize pkt counters per src-grp */
954 	    rt->mfc_pkt_cnt    = 0;
955 	    rt->mfc_byte_cnt   = 0;
956 	    rt->mfc_wrong_if   = 0;
957 	    rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
958 	    rt->mfc_expire     = 0;
959 
960 	    /* link into table */
961 	    mb_rt->m_next  = mfctable[hash];
962 	    mfctable[hash] = mb_rt;
963 	    mb_rt->m_act = NULL;
964 	}
965     }
966     splx(s);
967     return 0;
968 }
969 
970 #ifdef UPCALL_TIMING
971 /*
972  * collect delay statistics on the upcalls
973  */
974 static void collate(t)
975 register struct timeval *t;
976 {
977     register u_long d;
978     register struct timeval tp;
979     register u_long delta;
980 
981     GET_TIME(tp);
982 
983     if (TV_LT(*t, tp))
984     {
985 	TV_DELTA(tp, *t, delta);
986 
987 	d = delta >> 10;
988 	if (d > 50)
989 	    d = 50;
990 
991 	++upcall_data[d];
992     }
993 }
994 #endif /* UPCALL_TIMING */
995 
996 /*
997  * Delete an mfc entry
998  */
999 static int
1000 del_mfc(mfccp)
1001     struct mfcctl *mfccp;
1002 {
1003     struct in_addr 	origin;
1004     struct in_addr 	mcastgrp;
1005     struct mfc 		*rt;
1006     struct mbuf 	*mb_rt;
1007     struct mbuf 	**nptr;
1008     u_long 		hash;
1009     int s;
1010 
1011     origin = mfccp->mfcc_origin;
1012     mcastgrp = mfccp->mfcc_mcastgrp;
1013     hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1014 
1015     if (mrtdebug & DEBUG_MFC)
1016 	log(LOG_DEBUG,"del_mfc orig %x mcastgrp %x\n",
1017 	    ntohl(origin.s_addr), ntohl(mcastgrp.s_addr));
1018 
1019     s = splnet();
1020 
1021     nptr = &mfctable[hash];
1022     while ((mb_rt = *nptr) != NULL) {
1023         rt = mtod(mb_rt, struct mfc *);
1024 	if (origin.s_addr == rt->mfc_origin.s_addr &&
1025 	    mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1026 	    mb_rt->m_act == NULL)
1027 	    break;
1028 
1029 	nptr = &mb_rt->m_next;
1030     }
1031     if (mb_rt == NULL) {
1032 	splx(s);
1033 	return EADDRNOTAVAIL;
1034     }
1035 
1036     MFREE(mb_rt, *nptr);
1037 
1038     splx(s);
1039 
1040     return 0;
1041 }
1042 
1043 /*
1044  * Send a message to mrouted on the multicast routing socket
1045  */
1046 static int
1047 socket_send(s, mm, src)
1048 	struct socket *s;
1049 	struct mbuf *mm;
1050 	struct sockaddr_in *src;
1051 {
1052 	if (s) {
1053 		if (sbappendaddr(&s->so_rcv,
1054 				 (struct sockaddr *)src,
1055 				 mm, (struct mbuf *)0) != 0) {
1056 			sorwakeup(s);
1057 			return 0;
1058 		}
1059 	}
1060 	m_freem(mm);
1061 	return -1;
1062 }
1063 
1064 /*
1065  * IP multicast forwarding function. This function assumes that the packet
1066  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1067  * pointed to by "ifp", and the packet is to be relayed to other networks
1068  * that have members of the packet's destination IP multicast group.
1069  *
1070  * The packet is returned unscathed to the caller, unless it is
1071  * erroneous, in which case a non-zero return value tells the caller to
1072  * discard it.
1073  */
1074 
1075 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1076 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1077 
1078 int
1079 X_ip_mforward(ip, ifp, m, imo)
1080     register struct ip *ip;
1081     struct ifnet *ifp;
1082     struct mbuf *m;
1083     struct ip_moptions *imo;
1084 {
1085     register struct mfc *rt = 0; /* XXX uninit warning */
1086     register u_char *ipoptions;
1087     static struct sockaddr_in 	k_igmpsrc	= { sizeof k_igmpsrc, AF_INET };
1088     static int srctun = 0;
1089     register struct mbuf *mm;
1090     int s;
1091     vifi_t vifi;
1092     struct vif *vifp;
1093 
1094     if (mrtdebug & DEBUG_FORWARD)
1095 	log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %x\n",
1096 	    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1097 
1098     if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1099 	(ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1100 	/*
1101 	 * Packet arrived via a physical interface or
1102 	 * an encapsulated tunnel.
1103 	 */
1104     } else {
1105 	/*
1106 	 * Packet arrived through a source-route tunnel.
1107 	 * Source-route tunnels are no longer supported.
1108 	 */
1109 	if ((srctun++ % 1000) == 0)
1110 	    log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n",
1111 		ntohl(ip->ip_src.s_addr));
1112 
1113 	return 1;
1114     }
1115 
1116     if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1117 	if (ip->ip_ttl < 255)
1118 		ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1119 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1120 	    vifp = viftable + vifi;
1121 	    printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s%d)\n",
1122 		ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi,
1123 		(vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1124 		vifp->v_ifp->if_name, vifp->v_ifp->if_unit);
1125 	}
1126 	return (ip_mdq(m, ifp, rt, vifi));
1127     }
1128     if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1129 	printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
1130 	    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr));
1131 	if(!imo)
1132 		printf("In fact, no options were specified at all\n");
1133     }
1134 
1135     /*
1136      * Don't forward a packet with time-to-live of zero or one,
1137      * or a packet destined to a local-only group.
1138      */
1139     if (ip->ip_ttl <= 1 ||
1140 	ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
1141 	return 0;
1142 
1143     /*
1144      * Determine forwarding vifs from the forwarding cache table
1145      */
1146     s = splnet();
1147     MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt);
1148 
1149     /* Entry exists, so forward if necessary */
1150     if (rt != NULL) {
1151 	splx(s);
1152 	return (ip_mdq(m, ifp, rt, -1));
1153     } else {
1154 	/*
1155 	 * If we don't have a route for packet's origin,
1156 	 * Make a copy of the packet &
1157 	 * send message to routing daemon
1158 	 */
1159 
1160 	register struct mbuf *mb_rt;
1161 	register struct mbuf *mb_ntry;
1162 	register struct mbuf *mb0;
1163 	register struct rtdetq *rte;
1164 	register struct mbuf *rte_m;
1165 	register u_long hash;
1166 	register int npkts;
1167 #ifdef UPCALL_TIMING
1168 	struct timeval tp;
1169 
1170 	GET_TIME(tp);
1171 #endif
1172 
1173 	mrtstat.mrts_no_route++;
1174 	if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1175 	    log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1176 		ntohl(ip->ip_src.s_addr),
1177 		ntohl(ip->ip_dst.s_addr));
1178 
1179 	/*
1180 	 * Allocate mbufs early so that we don't do extra work if we are
1181 	 * just going to fail anyway.
1182 	 */
1183 	MGET(mb_ntry, M_DONTWAIT, MT_DATA);
1184 	if (mb_ntry == NULL) {
1185 	    splx(s);
1186 	    return ENOBUFS;
1187 	}
1188 	mb0 = m_copy(m, 0, M_COPYALL);
1189 	if (mb0 == NULL) {
1190 	    m_free(mb_ntry);
1191 	    splx(s);
1192 	    return ENOBUFS;
1193 	}
1194 
1195 	/* is there an upcall waiting for this packet? */
1196 	hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1197 	for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
1198 	    rt = mtod(mb_rt, struct mfc *);
1199 	    if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1200 		(ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1201 		(mb_rt->m_act != NULL))
1202 		break;
1203 	}
1204 
1205 	if (mb_rt == NULL) {
1206 	    int hlen = ip->ip_hl << 2;
1207 	    int i;
1208 	    struct igmpmsg *im;
1209 
1210 	    /* no upcall, so make a new entry */
1211 	    MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
1212 	    if (mb_rt == NULL) {
1213 		m_free(mb_ntry);
1214 		m_freem(mb0);
1215 		splx(s);
1216 		return ENOBUFS;
1217 	    }
1218 	    /* Make a copy of the header to send to the user level process */
1219 	    mm = m_copy(m, 0, hlen);
1220 	    if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1221 		mm = m_pullup(mm, hlen);
1222 	    if (mm == NULL) {
1223 		m_free(mb_ntry);
1224 		m_freem(mb0);
1225 		m_free(mb_rt);
1226 		splx(s);
1227 		return ENOBUFS;
1228 	    }
1229 
1230 	    /*
1231 	     * Send message to routing daemon to install
1232 	     * a route into the kernel table
1233 	     */
1234 	    k_igmpsrc.sin_addr = ip->ip_src;
1235 
1236 	    im = mtod(mm, struct igmpmsg *);
1237 	    im->im_msgtype	= IGMPMSG_NOCACHE;
1238 	    im->im_mbz		= 0;
1239 
1240 	    mrtstat.mrts_upcalls++;
1241 
1242 	    if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1243 		log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1244 		++mrtstat.mrts_upq_sockfull;
1245 		m_free(mb_ntry);
1246 		m_freem(mb0);
1247 		m_free(mb_rt);
1248 		splx(s);
1249 		return ENOBUFS;
1250 	    }
1251 
1252 	    rt = mtod(mb_rt, struct mfc *);
1253 
1254 	    /* insert new entry at head of hash chain */
1255 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1256 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1257 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1258 	    nexpire[hash]++;
1259 	    for (i = 0; i < numvifs; i++)
1260 		rt->mfc_ttls[i] = 0;
1261 	    rt->mfc_parent = -1;
1262 
1263 	    /* link into table */
1264 	    mb_rt->m_next  = mfctable[hash];
1265 	    mfctable[hash] = mb_rt;
1266 	    mb_rt->m_act = NULL;
1267 
1268 	    rte_m = mb_rt;
1269 	} else {
1270 	    /* determine if q has overflowed */
1271 	    for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act)
1272 		npkts++;
1273 
1274 	    if (npkts > MAX_UPQ) {
1275 		mrtstat.mrts_upq_ovflw++;
1276 		m_free(mb_ntry);
1277 		m_freem(mb0);
1278 		splx(s);
1279 		return 0;
1280 	    }
1281 	}
1282 
1283 	mb_ntry->m_act = NULL;
1284 	rte = mtod(mb_ntry, struct rtdetq *);
1285 
1286 	rte->m 			= mb0;
1287 	rte->ifp 		= ifp;
1288 #ifdef UPCALL_TIMING
1289 	rte->t			= tp;
1290 #endif
1291 
1292 	/* Add this entry to the end of the queue */
1293 	rte_m->m_act		= mb_ntry;
1294 
1295 	splx(s);
1296 
1297 	return 0;
1298     }
1299 }
1300 
1301 #ifndef MROUTE_LKM
1302 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
1303 		   struct ip_moptions *) = X_ip_mforward;
1304 #endif
1305 
1306 /*
1307  * Clean up the cache entry if upcall is not serviced
1308  */
1309 static void
1310 expire_upcalls(void *unused)
1311 {
1312     struct mbuf *mb_rt, *m, **nptr;
1313     struct rtdetq *rte;
1314     struct mfc *mfc;
1315     int i;
1316     int s;
1317 
1318     s = splnet();
1319     for (i = 0; i < MFCTBLSIZ; i++) {
1320 	if (nexpire[i] == 0)
1321 	    continue;
1322 	nptr = &mfctable[i];
1323 	for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) {
1324 	    mfc = mtod(mb_rt, struct mfc *);
1325 
1326 	    /*
1327 	     * Skip real cache entries
1328 	     * Make sure it wasn't marked to not expire (shouldn't happen)
1329 	     * If it expires now
1330 	     */
1331 	    if (mb_rt->m_act != NULL &&
1332 	        mfc->mfc_expire != 0 &&
1333 		--mfc->mfc_expire == 0) {
1334 		if (mrtdebug & DEBUG_EXPIRE)
1335 		    log(LOG_DEBUG, "expire_upcalls: expiring (%x %x)\n",
1336 			ntohl(mfc->mfc_origin.s_addr),
1337 			ntohl(mfc->mfc_mcastgrp.s_addr));
1338 		/*
1339 		 * drop all the packets
1340 		 * free the mbuf with the pkt, if, timing info
1341 		 */
1342 		while (mb_rt->m_act) {
1343 		    m = mb_rt->m_act;
1344 		    mb_rt->m_act = m->m_act;
1345 
1346 		    rte = mtod(m, struct rtdetq *);
1347 		    m_freem(rte->m);
1348 		    m_free(m);
1349 		}
1350 		++mrtstat.mrts_cache_cleanups;
1351 		nexpire[i]--;
1352 
1353 		MFREE(mb_rt, *nptr);
1354 	    } else {
1355 		nptr = &mb_rt->m_next;
1356 	    }
1357 	}
1358     }
1359     splx(s);
1360     timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
1361 }
1362 
1363 /*
1364  * Packet forwarding routine once entry in the cache is made
1365  */
1366 static int
1367 ip_mdq(m, ifp, rt, xmt_vif)
1368     register struct mbuf *m;
1369     register struct ifnet *ifp;
1370     register struct mfc *rt;
1371     register vifi_t xmt_vif;
1372 {
1373     register struct ip  *ip = mtod(m, struct ip *);
1374     register vifi_t vifi;
1375     register struct vif *vifp;
1376     register int plen = ntohs(ip->ip_len);
1377 
1378 /*
1379  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1380  * input, they shouldn't get counted on output, so statistics keeping is
1381  * seperate.
1382  */
1383 #define MC_SEND(ip,vifp,m) {                             \
1384                 if ((vifp)->v_flags & VIFF_TUNNEL)  	 \
1385                     encap_send((ip), (vifp), (m));       \
1386                 else                                     \
1387                     phyint_send((ip), (vifp), (m));      \
1388 }
1389 
1390     /*
1391      * If xmt_vif is not -1, send on only the requested vif.
1392      *
1393      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1394      */
1395     if (xmt_vif < numvifs) {
1396 	MC_SEND(ip, viftable + xmt_vif, m);
1397 	return 1;
1398     }
1399 
1400     /*
1401      * Don't forward if it didn't arrive from the parent vif for its origin.
1402      */
1403     vifi = rt->mfc_parent;
1404     if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1405 	/* came in the wrong interface */
1406 	if (mrtdebug & DEBUG_FORWARD)
1407 	    log(LOG_DEBUG, "wrong if: ifp %x vifi %d vififp %x\n",
1408 		ifp, vifi, viftable[vifi].v_ifp);
1409 	++mrtstat.mrts_wrong_if;
1410 	++rt->mfc_wrong_if;
1411 	/*
1412 	 * If we are doing PIM assert processing, and we are forwarding
1413 	 * packets on this interface, and it is a broadcast medium
1414 	 * interface (and not a tunnel), send a message to the routing daemon.
1415 	 */
1416 	if (pim_assert && rt->mfc_ttls[vifi] &&
1417 		(ifp->if_flags & IFF_BROADCAST) &&
1418 		!(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1419 	    struct sockaddr_in k_igmpsrc;
1420 	    struct mbuf *mm;
1421 	    struct igmpmsg *im;
1422 	    int hlen = ip->ip_hl << 2;
1423 	    struct timeval now;
1424 	    register u_long delta;
1425 
1426 	    GET_TIME(now);
1427 
1428 	    TV_DELTA(rt->mfc_last_assert, now, delta);
1429 
1430 	    if (delta > ASSERT_MSG_TIME) {
1431 		mm = m_copy(m, 0, hlen);
1432 		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1433 		    mm = m_pullup(mm, hlen);
1434 		if (mm == NULL) {
1435 		    return ENOBUFS;
1436 		}
1437 
1438 		rt->mfc_last_assert = now;
1439 
1440 		im = mtod(mm, struct igmpmsg *);
1441 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1442 		im->im_mbz		= 0;
1443 		im->im_vif		= vifi;
1444 
1445 		k_igmpsrc.sin_addr = im->im_src;
1446 
1447 		socket_send(ip_mrouter, mm, &k_igmpsrc);
1448 	    }
1449 	}
1450 	return 0;
1451     }
1452 
1453     /* If I sourced this packet, it counts as output, else it was input. */
1454     if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1455 	viftable[vifi].v_pkt_out++;
1456 	viftable[vifi].v_bytes_out += plen;
1457     } else {
1458 	viftable[vifi].v_pkt_in++;
1459 	viftable[vifi].v_bytes_in += plen;
1460     }
1461     rt->mfc_pkt_cnt++;
1462     rt->mfc_byte_cnt += plen;
1463 
1464     /*
1465      * For each vif, decide if a copy of the packet should be forwarded.
1466      * Forward if:
1467      *		- the ttl exceeds the vif's threshold
1468      *		- there are group members downstream on interface
1469      */
1470     for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1471 	if ((rt->mfc_ttls[vifi] > 0) &&
1472 	    (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1473 	    vifp->v_pkt_out++;
1474 	    vifp->v_bytes_out += plen;
1475 	    MC_SEND(ip, vifp, m);
1476 	}
1477 
1478     return 0;
1479 }
1480 
1481 /*
1482  * check if a vif number is legal/ok. This is used by ip_output, to export
1483  * numvifs there,
1484  */
1485 int
1486 X_legal_vif_num(vif)
1487     int vif;
1488 {
1489     if (vif >= 0 && vif < numvifs)
1490        return(1);
1491     else
1492        return(0);
1493 }
1494 
1495 #ifndef MROUTE_LKM
1496 int (*legal_vif_num)(int) = X_legal_vif_num;
1497 #endif
1498 
1499 /*
1500  * Return the local address used by this vif
1501  */
1502 u_long
1503 X_ip_mcast_src(vifi)
1504     int vifi;
1505 {
1506     if (vifi >= 0 && vifi < numvifs)
1507 	return viftable[vifi].v_lcl_addr.s_addr;
1508     else
1509 	return INADDR_ANY;
1510 }
1511 
1512 #ifndef MROUTE_LKM
1513 u_long (*ip_mcast_src)(int) = X_ip_mcast_src;
1514 #endif
1515 
1516 static void
1517 phyint_send(ip, vifp, m)
1518     struct ip *ip;
1519     struct vif *vifp;
1520     struct mbuf *m;
1521 {
1522     register struct mbuf *mb_copy;
1523     register int hlen = ip->ip_hl << 2;
1524 
1525     /*
1526      * Make a new reference to the packet; make sure that
1527      * the IP header is actually copied, not just referenced,
1528      * so that ip_output() only scribbles on the copy.
1529      */
1530     mb_copy = m_copy(m, 0, M_COPYALL);
1531     if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1532 	mb_copy = m_pullup(mb_copy, hlen);
1533     if (mb_copy == NULL)
1534 	return;
1535 
1536     if (vifp->v_rate_limit <= 0)
1537 	tbf_send_packet(vifp, mb_copy);
1538     else
1539 	tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1540 }
1541 
1542 static void
1543 encap_send(ip, vifp, m)
1544     register struct ip *ip;
1545     register struct vif *vifp;
1546     register struct mbuf *m;
1547 {
1548     register struct mbuf *mb_copy;
1549     register struct ip *ip_copy;
1550     register int i, len = ip->ip_len;
1551 
1552     /*
1553      * copy the old packet & pullup it's IP header into the
1554      * new mbuf so we can modify it.  Try to fill the new
1555      * mbuf since if we don't the ethernet driver will.
1556      */
1557     MGET(mb_copy, M_DONTWAIT, MT_DATA);
1558     if (mb_copy == NULL)
1559 	return;
1560     mb_copy->m_data += 16;
1561     mb_copy->m_len = sizeof(multicast_encap_iphdr);
1562 
1563     if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1564 	m_freem(mb_copy);
1565 	return;
1566     }
1567     i = MHLEN - M_LEADINGSPACE(mb_copy);
1568     if (i > len)
1569 	i = len;
1570     mb_copy = m_pullup(mb_copy, i);
1571     if (mb_copy == NULL)
1572 	return;
1573     mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1574 
1575     /*
1576      * fill in the encapsulating IP header.
1577      */
1578     ip_copy = mtod(mb_copy, struct ip *);
1579     *ip_copy = multicast_encap_iphdr;
1580     ip_copy->ip_id = htons(ip_id++);
1581     ip_copy->ip_len += len;
1582     ip_copy->ip_src = vifp->v_lcl_addr;
1583     ip_copy->ip_dst = vifp->v_rmt_addr;
1584 
1585     /*
1586      * turn the encapsulated IP header back into a valid one.
1587      */
1588     ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1589     --ip->ip_ttl;
1590     HTONS(ip->ip_len);
1591     HTONS(ip->ip_off);
1592     ip->ip_sum = 0;
1593 #if defined(LBL) && !defined(ultrix)
1594     ip->ip_sum = ~oc_cksum((caddr_t)ip, ip->ip_hl << 2, 0);
1595 #else
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 #endif
1600 
1601     if (vifp->v_rate_limit <= 0)
1602 	tbf_send_packet(vifp, mb_copy);
1603     else
1604 	tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1605 }
1606 
1607 /*
1608  * De-encapsulate a packet and feed it back through ip input (this
1609  * routine is called whenever IP gets a packet with proto type
1610  * ENCAP_PROTO and a local destination address).
1611  */
1612 void
1613 #ifdef MROUTE_LKM
1614 X_ipip_input(m)
1615 #else
1616 ipip_input(m, iphlen)
1617 #endif
1618 	register struct mbuf *m;
1619 	int iphlen;
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);
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 %x\n",
1665 		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     int s = splnet();
1889 
1890     if (vifp->v_flags & VIFF_TUNNEL) {
1891 	/* If tunnel options */
1892 	ip_output(m, (struct mbuf *)0, (struct route *)0,
1893 		  IP_FORWARDING, (struct ip_moptions *)0);
1894     } else {
1895 	imo.imo_multicast_ifp  = vifp->v_ifp;
1896 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1897 	imo.imo_multicast_loop = 1;
1898 	imo.imo_multicast_vif  = -1;
1899 
1900 	error = ip_output(m, (struct mbuf *)0, (struct route *)0,
1901 			  IP_FORWARDING, &imo);
1902 
1903 	if (mrtdebug & DEBUG_XMIT)
1904 	    log(LOG_DEBUG, "phyint_send on vif %d err %d\n",
1905 		vifp - viftable, error);
1906     }
1907     splx(s);
1908 }
1909 
1910 /* determine the current time and then
1911  * the elapsed time (between the last time and time now)
1912  * in milliseconds & update the no. of tokens in the bucket
1913  */
1914 static void
1915 tbf_update_tokens(vifp)
1916     register struct vif *vifp;
1917 {
1918     struct timeval tp;
1919     register u_long tm;
1920     register int s = splnet();
1921     register struct tbf *t = vifp->v_tbf;
1922 
1923     GET_TIME(tp);
1924 
1925     TV_DELTA(tp, t->tbf_last_pkt_t, tm);
1926 
1927     /*
1928      * This formula is actually
1929      * "time in seconds" * "bytes/second".
1930      *
1931      * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
1932      *
1933      * The (1000/1024) was introduced in add_vif to optimize
1934      * this divide into a shift.
1935      */
1936     t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8;
1937     t->tbf_last_pkt_t = tp;
1938 
1939     if (t->tbf_n_tok > MAX_BKT_SIZE)
1940 	t->tbf_n_tok = MAX_BKT_SIZE;
1941 
1942     splx(s);
1943 }
1944 
1945 static int
1946 priority(vifp, ip)
1947     register struct vif *vifp;
1948     register struct ip *ip;
1949 {
1950     register int prio;
1951 
1952     /* temporary hack; may add general packet classifier some day */
1953 
1954     /*
1955      * The UDP port space is divided up into four priority ranges:
1956      * [0, 16384)     : unclassified - lowest priority
1957      * [16384, 32768) : audio - highest priority
1958      * [32768, 49152) : whiteboard - medium priority
1959      * [49152, 65536) : video - low priority
1960      */
1961     if (ip->ip_p == IPPROTO_UDP) {
1962 	struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
1963 	switch (ntohs(udp->uh_dport) & 0xc000) {
1964 	    case 0x4000:
1965 		prio = 70;
1966 		break;
1967 	    case 0x8000:
1968 		prio = 60;
1969 		break;
1970 	    case 0xc000:
1971 		prio = 55;
1972 		break;
1973 	    default:
1974 		prio = 50;
1975 		break;
1976 	}
1977 	if (tbfdebug > 1)
1978 		log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio);
1979     } else {
1980 	    prio = 50;
1981     }
1982     return prio;
1983 }
1984 
1985 /*
1986  * End of token bucket filter modifications
1987  */
1988 
1989 int
1990 ip_rsvp_vif_init(so, m)
1991     struct socket *so;
1992     struct mbuf *m;
1993 {
1994     int i;
1995     register int s;
1996 
1997     if (rsvpdebug)
1998 	printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
1999 	       so->so_type, so->so_proto->pr_protocol);
2000 
2001     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2002 	return EOPNOTSUPP;
2003 
2004     /* Check mbuf. */
2005     if (m == NULL || m->m_len != sizeof(int)) {
2006 	return EINVAL;
2007     }
2008     i = *(mtod(m, int *));
2009 
2010     if (rsvpdebug)
2011 	printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on);
2012 
2013     s = splnet();
2014 
2015     /* Check vif. */
2016     if (!legal_vif_num(i)) {
2017 	splx(s);
2018 	return EADDRNOTAVAIL;
2019     }
2020 
2021     /* Check if socket is available. */
2022     if (viftable[i].v_rsvpd != NULL) {
2023 	splx(s);
2024 	return EADDRINUSE;
2025     }
2026 
2027     viftable[i].v_rsvpd = so;
2028     /* This may seem silly, but we need to be sure we don't over-increment
2029      * the RSVP counter, in case something slips up.
2030      */
2031     if (!viftable[i].v_rsvp_on) {
2032 	viftable[i].v_rsvp_on = 1;
2033 	rsvp_on++;
2034     }
2035 
2036     splx(s);
2037     return 0;
2038 }
2039 
2040 int
2041 ip_rsvp_vif_done(so, m)
2042     struct socket *so;
2043     struct mbuf *m;
2044 {
2045 	int i;
2046 	register int s;
2047 
2048     if (rsvpdebug)
2049 	printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2050 	       so->so_type, so->so_proto->pr_protocol);
2051 
2052     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2053 	return EOPNOTSUPP;
2054 
2055     /* Check mbuf. */
2056     if (m == NULL || m->m_len != sizeof(int)) {
2057 	    return EINVAL;
2058     }
2059     i = *(mtod(m, int *));
2060 
2061     s = splnet();
2062 
2063     /* Check vif. */
2064     if (!legal_vif_num(i)) {
2065 	splx(s);
2066         return EADDRNOTAVAIL;
2067     }
2068 
2069     if (rsvpdebug)
2070 	printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2071 	       viftable[i].v_rsvpd, so);
2072 
2073     viftable[i].v_rsvpd = NULL;
2074     /* This may seem silly, but we need to be sure we don't over-decrement
2075      * the RSVP counter, in case something slips up.
2076      */
2077     if (viftable[i].v_rsvp_on) {
2078 	viftable[i].v_rsvp_on = 0;
2079 	rsvp_on--;
2080     }
2081 
2082     splx(s);
2083     return 0;
2084 }
2085 
2086 void
2087 ip_rsvp_force_done(so)
2088     struct socket *so;
2089 {
2090     int vifi;
2091     register int s;
2092 
2093     /* Don't bother if it is not the right type of socket. */
2094     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2095 	return;
2096 
2097     s = splnet();
2098 
2099     /* The socket may be attached to more than one vif...this
2100      * is perfectly legal.
2101      */
2102     for (vifi = 0; vifi < numvifs; vifi++) {
2103 	if (viftable[vifi].v_rsvpd == so) {
2104 	    viftable[vifi].v_rsvpd = NULL;
2105 	    /* This may seem silly, but we need to be sure we don't
2106 	     * over-decrement the RSVP counter, in case something slips up.
2107 	     */
2108 	    if (viftable[vifi].v_rsvp_on) {
2109 		viftable[vifi].v_rsvp_on = 0;
2110 		rsvp_on--;
2111 	    }
2112 	}
2113     }
2114 
2115     splx(s);
2116     return;
2117 }
2118 
2119 void
2120 rsvp_input(m, iphlen)
2121 	struct mbuf *m;
2122 	int iphlen;
2123 {
2124     int vifi;
2125     register struct ip *ip = mtod(m, struct ip *);
2126     static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
2127     register int s;
2128     struct ifnet *ifp;
2129 
2130     if (rsvpdebug)
2131 	printf("rsvp_input: rsvp_on %d\n",rsvp_on);
2132 
2133     /* Can still get packets with rsvp_on = 0 if there is a local member
2134      * of the group to which the RSVP packet is addressed.  But in this
2135      * case we want to throw the packet away.
2136      */
2137     if (!rsvp_on) {
2138 	m_freem(m);
2139 	return;
2140     }
2141 
2142     /* If the old-style non-vif-associated socket is set, then use
2143      * it and ignore the new ones.
2144      */
2145     if (ip_rsvpd != NULL) {
2146 	if (rsvpdebug)
2147 	    printf("rsvp_input: Sending packet up old-style socket\n");
2148 	rip_input(m);
2149 	return;
2150     }
2151 
2152     s = splnet();
2153 
2154     if (rsvpdebug)
2155 	printf("rsvp_input: check vifs\n");
2156 
2157 #ifdef DIAGNOSTIC
2158     if (!(m->m_flags & M_PKTHDR))
2159 	    panic("rsvp_input no hdr");
2160 #endif
2161 
2162     ifp = m->m_pkthdr.rcvif;
2163     /* Find which vif the packet arrived on. */
2164     for (vifi = 0; vifi < numvifs; vifi++) {
2165 	if (viftable[vifi].v_ifp == ifp)
2166  		break;
2167  	}
2168 
2169     if (vifi == numvifs) {
2170 	/* Can't find vif packet arrived on. Drop packet. */
2171 	if (rsvpdebug)
2172 	    printf("rsvp_input: Can't find vif for packet...dropping it.\n");
2173 	m_freem(m);
2174 	splx(s);
2175 	return;
2176     }
2177 
2178     if (rsvpdebug)
2179 	printf("rsvp_input: check socket\n");
2180 
2181     if (viftable[vifi].v_rsvpd == NULL) {
2182 	/* drop packet, since there is no specific socket for this
2183 	 * interface */
2184 	    if (rsvpdebug)
2185 		    printf("rsvp_input: No socket defined for vif %d\n",vifi);
2186 	    m_freem(m);
2187 	    splx(s);
2188 	    return;
2189     }
2190     rsvp_src.sin_addr = ip->ip_src;
2191 
2192     if (rsvpdebug && m)
2193 	printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
2194 	       m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
2195 
2196     if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2197 	if (rsvpdebug)
2198 	    printf("rsvp_input: Failed to append to socket\n");
2199     else
2200 	if (rsvpdebug)
2201 	    printf("rsvp_input: send packet up\n");
2202 
2203     splx(s);
2204 }
2205 
2206 #ifdef MROUTE_LKM
2207 #include <sys/conf.h>
2208 #include <sys/exec.h>
2209 #include <sys/sysent.h>
2210 #include <sys/lkm.h>
2211 
2212 MOD_MISC("ip_mroute_mod")
2213 
2214 static int
2215 ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd)
2216 {
2217 	int i;
2218 	struct lkm_misc	*args = lkmtp->private.lkm_misc;
2219 	int err = 0;
2220 
2221 	switch(cmd) {
2222 		static int (*old_ip_mrouter_cmd)();
2223 		static int (*old_ip_mrouter_done)();
2224 		static int (*old_ip_mforward)();
2225 		static int (*old_mrt_ioctl)();
2226 		static void (*old_proto4_input)();
2227 		static int (*old_legal_vif_num)();
2228 		extern struct protosw inetsw[];
2229 
2230 	case LKM_E_LOAD:
2231 		if(lkmexists(lkmtp) || ip_mrtproto)
2232 		  return(EEXIST);
2233 		old_ip_mrouter_cmd = ip_mrouter_cmd;
2234 		ip_mrouter_cmd = X_ip_mrouter_cmd;
2235 		old_ip_mrouter_done = ip_mrouter_done;
2236 		ip_mrouter_done = X_ip_mrouter_done;
2237 		old_ip_mforward = ip_mforward;
2238 		ip_mforward = X_ip_mforward;
2239 		old_mrt_ioctl = mrt_ioctl;
2240 		mrt_ioctl = X_mrt_ioctl;
2241               old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input;
2242               inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input;
2243 		old_legal_vif_num = legal_vif_num;
2244 		legal_vif_num = X_legal_vif_num;
2245 		ip_mrtproto = IGMP_DVMRP;
2246 
2247 		printf("\nIP multicast routing loaded\n");
2248 		break;
2249 
2250 	case LKM_E_UNLOAD:
2251 		if (ip_mrouter)
2252 		  return EINVAL;
2253 
2254 		ip_mrouter_cmd = old_ip_mrouter_cmd;
2255 		ip_mrouter_done = old_ip_mrouter_done;
2256 		ip_mforward = old_ip_mforward;
2257 		mrt_ioctl = old_mrt_ioctl;
2258               inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input;
2259 		legal_vif_num = old_legal_vif_num;
2260 		ip_mrtproto = 0;
2261 		break;
2262 
2263 	default:
2264 		err = EINVAL;
2265 		break;
2266 	}
2267 
2268 	return(err);
2269 }
2270 
2271 int
2272 ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) {
2273 	DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle,
2274 		 nosys);
2275 }
2276 
2277 #endif /* MROUTE_LKM */
2278 #endif /* MROUTING */
2279