ip_mroute.c revision 22967
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 * $Id$
13 */
14
15#include "opt_mrouting.h"
16
17#include <sys/param.h>
18#include <sys/queue.h>
19#include <sys/systm.h>
20#include <sys/mbuf.h>
21#include <sys/socket.h>
22#include <sys/socketvar.h>
23#include <sys/protosw.h>
24#include <sys/errno.h>
25#include <sys/time.h>
26#include <sys/kernel.h>
27#include <sys/ioctl.h>
28#include <sys/syslog.h>
29#include <net/if.h>
30#include <net/route.h>
31#include <netinet/in.h>
32#include <netinet/in_systm.h>
33#include <netinet/ip.h>
34#include <netinet/ip_var.h>
35#include <netinet/in_pcb.h>
36#include <netinet/in_var.h>
37#include <netinet/igmp.h>
38#include <netinet/igmp_var.h>
39#include <netinet/ip_mroute.h>
40#include <netinet/udp.h>
41
42#ifndef NTOHL
43#if BYTE_ORDER != BIG_ENDIAN
44#define NTOHL(d) ((d) = ntohl((d)))
45#define NTOHS(d) ((d) = ntohs((u_short)(d)))
46#define HTONL(d) ((d) = htonl((d)))
47#define HTONS(d) ((d) = htons((u_short)(d)))
48#else
49#define NTOHL(d)
50#define NTOHS(d)
51#define HTONL(d)
52#define HTONS(d)
53#endif
54#endif
55
56#ifndef MROUTING
57extern u_long	_ip_mcast_src __P((int vifi));
58extern int	_ip_mforward __P((struct ip *ip, struct ifnet *ifp,
59				  struct mbuf *m, struct ip_moptions *imo));
60extern int	_ip_mrouter_done __P((void));
61extern int	_ip_mrouter_get __P((int cmd, struct socket *so,
62				     struct mbuf **m));
63extern int	_ip_mrouter_set __P((int cmd, struct socket *so,
64				     struct mbuf *m));
65extern int	_mrt_ioctl __P((int req, caddr_t data, struct proc *p));
66
67/*
68 * Dummy routines and globals used when multicast routing is not compiled in.
69 */
70
71struct socket  *ip_mrouter  = NULL;
72static u_int		ip_mrtproto = 0;
73static struct mrtstat	mrtstat;
74u_int		rsvpdebug = 0;
75
76int
77_ip_mrouter_set(cmd, so, m)
78	int cmd;
79	struct socket *so;
80	struct mbuf *m;
81{
82	return(EOPNOTSUPP);
83}
84
85int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = _ip_mrouter_set;
86
87
88int
89_ip_mrouter_get(cmd, so, m)
90	int cmd;
91	struct socket *so;
92	struct mbuf **m;
93{
94	return(EOPNOTSUPP);
95}
96
97int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = _ip_mrouter_get;
98
99int
100_ip_mrouter_done()
101{
102	return(0);
103}
104
105int (*ip_mrouter_done)(void) = _ip_mrouter_done;
106
107int
108_ip_mforward(ip, ifp, m, imo)
109	struct ip *ip;
110	struct ifnet *ifp;
111	struct mbuf *m;
112	struct ip_moptions *imo;
113{
114	return(0);
115}
116
117int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
118		   struct ip_moptions *) = _ip_mforward;
119
120int
121_mrt_ioctl(int req, caddr_t data, struct proc *p)
122{
123	return EOPNOTSUPP;
124}
125
126int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl;
127
128void
129rsvp_input(m, iphlen)		/* XXX must fixup manually */
130	struct mbuf *m;
131	int iphlen;
132{
133    /* Can still get packets with rsvp_on = 0 if there is a local member
134     * of the group to which the RSVP packet is addressed.  But in this
135     * case we want to throw the packet away.
136     */
137    if (!rsvp_on) {
138	m_freem(m);
139	return;
140    }
141
142    if (ip_rsvpd != NULL) {
143	if (rsvpdebug)
144	    printf("rsvp_input: Sending packet up old-style socket\n");
145	rip_input(m, iphlen);
146	return;
147    }
148    /* Drop the packet */
149    m_freem(m);
150}
151
152void ipip_input(struct mbuf *m, int iphlen) { /* XXX must fixup manually */
153	rip_input(m, iphlen);
154}
155
156int (*legal_vif_num)(int) = 0;
157
158/*
159 * This should never be called, since IP_MULTICAST_VIF should fail, but
160 * just in case it does get called, the code a little lower in ip_output
161 * will assign the packet a local address.
162 */
163u_long
164_ip_mcast_src(int vifi) { return INADDR_ANY; }
165u_long (*ip_mcast_src)(int) = _ip_mcast_src;
166
167int
168ip_rsvp_vif_init(so, m)
169    struct socket *so;
170    struct mbuf *m;
171{
172    return(EINVAL);
173}
174
175int
176ip_rsvp_vif_done(so, m)
177    struct socket *so;
178    struct mbuf *m;
179{
180    return(EINVAL);
181}
182
183void
184ip_rsvp_force_done(so)
185    struct socket *so;
186{
187    return;
188}
189
190#else /* MROUTING */
191
192#define M_HASCL(m)	((m)->m_flags & M_EXT)
193
194#define INSIZ		sizeof(struct in_addr)
195#define	same(a1, a2) \
196	(bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0)
197
198#define MT_MRTABLE MT_RTABLE	/* since nothing else uses it */
199
200/*
201 * Globals.  All but ip_mrouter and ip_mrtproto could be static,
202 * except for netstat or debugging purposes.
203 */
204#ifndef MROUTE_LKM
205struct socket  *ip_mrouter  = NULL;
206struct mrtstat	mrtstat;
207
208int		ip_mrtproto = IGMP_DVMRP;    /* for netstat only */
209#else /* MROUTE_LKM */
210extern void	X_ipip_input __P((struct mbuf *m, int iphlen));
211extern struct mrtstat mrtstat;
212static int ip_mrtproto;
213#endif
214
215#define NO_RTE_FOUND 	0x1
216#define RTE_FOUND	0x2
217
218static struct mbuf    *mfctable[MFCTBLSIZ];
219static u_char		nexpire[MFCTBLSIZ];
220static struct vif	viftable[MAXVIFS];
221static u_int	mrtdebug = 0;	  /* debug level 	*/
222#define		DEBUG_MFC	0x02
223#define		DEBUG_FORWARD	0x04
224#define		DEBUG_EXPIRE	0x08
225#define		DEBUG_XMIT	0x10
226static u_int  	tbfdebug = 0;     /* tbf debug level 	*/
227static u_int	rsvpdebug = 0;	  /* rsvp debug level   */
228
229#define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
230#define		UPCALL_EXPIRE	6		/* number of timeouts	*/
231
232/*
233 * Define the token bucket filter structures
234 * tbftable -> each vif has one of these for storing info
235 */
236
237static struct tbf tbftable[MAXVIFS];
238#define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
239
240/*
241 * 'Interfaces' associated with decapsulator (so we can tell
242 * packets that went through it from ones that get reflected
243 * by a broken gateway).  These interfaces are never linked into
244 * the system ifnet list & no routes point to them.  I.e., packets
245 * can't be sent this way.  They only exist as a placeholder for
246 * multicast source verification.
247 */
248static struct ifnet multicast_decap_if[MAXVIFS];
249
250#define ENCAP_TTL 64
251#define ENCAP_PROTO IPPROTO_IPIP	/* 4 */
252
253/* prototype IP hdr for encapsulated packets */
254static struct ip multicast_encap_iphdr = {
255#if BYTE_ORDER == LITTLE_ENDIAN
256	sizeof(struct ip) >> 2, IPVERSION,
257#else
258	IPVERSION, sizeof(struct ip) >> 2,
259#endif
260	0,				/* tos */
261	sizeof(struct ip),		/* total length */
262	0,				/* id */
263	0,				/* frag offset */
264	ENCAP_TTL, ENCAP_PROTO,
265	0,				/* checksum */
266};
267
268/*
269 * Private variables.
270 */
271static vifi_t	   numvifs = 0;
272static int have_encap_tunnel = 0;
273
274/*
275 * one-back cache used by ipip_input to locate a tunnel's vif
276 * given a datagram's src ip address.
277 */
278static u_long last_encap_src;
279static struct vif *last_encap_vif;
280
281static u_long	X_ip_mcast_src __P((int vifi));
282static int	X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo));
283static int	X_ip_mrouter_done __P((void));
284static int	X_ip_mrouter_get __P((int cmd, struct socket *so, struct mbuf **m));
285static int	X_ip_mrouter_set __P((int cmd, struct socket *so, struct mbuf *m));
286static int	X_legal_vif_num __P((int vif));
287static int	X_mrt_ioctl __P((int cmd, caddr_t data));
288
289static int get_sg_cnt(struct sioc_sg_req *);
290static int get_vif_cnt(struct sioc_vif_req *);
291static int ip_mrouter_init(struct socket *, struct mbuf *);
292static int add_vif(struct vifctl *);
293static int del_vif(vifi_t *);
294static int add_mfc(struct mfcctl *);
295static int del_mfc(struct mfcctl *);
296static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
297static int get_version(struct mbuf *);
298static int get_assert(struct mbuf *);
299static int set_assert(int *);
300static void expire_upcalls(void *);
301static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *,
302		  vifi_t);
303static void phyint_send(struct ip *, struct vif *, struct mbuf *);
304static void encap_send(struct ip *, struct vif *, struct mbuf *);
305static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
306static void tbf_queue(struct vif *, struct mbuf *);
307static void tbf_process_q(struct vif *);
308static void tbf_reprocess_q(void *);
309static int tbf_dq_sel(struct vif *, struct ip *);
310static void tbf_send_packet(struct vif *, struct mbuf *);
311static void tbf_update_tokens(struct vif *);
312static int priority(struct vif *, struct ip *);
313void multiencap_decap(struct mbuf *);
314
315/*
316 * whether or not special PIM assert processing is enabled.
317 */
318static int pim_assert;
319/*
320 * Rate limit for assert notification messages, in usec
321 */
322#define ASSERT_MSG_TIME		3000000
323
324/*
325 * Hash function for a source, group entry
326 */
327#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
328			((g) >> 20) ^ ((g) >> 10) ^ (g))
329
330/*
331 * Find a route for a given origin IP address and Multicast group address
332 * Type of service parameter to be added in the future!!!
333 */
334
335#define MFCFIND(o, g, rt) { \
336	register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \
337	register struct mfc *_rt = NULL; \
338	rt = NULL; \
339	++mrtstat.mrts_mfc_lookups; \
340	while (_mb_rt) { \
341		_rt = mtod(_mb_rt, struct mfc *); \
342		if ((_rt->mfc_origin.s_addr == o) && \
343		    (_rt->mfc_mcastgrp.s_addr == g) && \
344		    (_mb_rt->m_act == NULL)) { \
345			rt = _rt; \
346			break; \
347		} \
348		_mb_rt = _mb_rt->m_next; \
349	} \
350	if (rt == NULL) { \
351		++mrtstat.mrts_mfc_misses; \
352	} \
353}
354
355
356/*
357 * Macros to compute elapsed time efficiently
358 * Borrowed from Van Jacobson's scheduling code
359 */
360#define TV_DELTA(a, b, delta) { \
361	    register int xxs; \
362		\
363	    delta = (a).tv_usec - (b).tv_usec; \
364	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
365	       switch (xxs) { \
366		      case 2: \
367			  delta += 1000000; \
368			      /* fall through */ \
369		      case 1: \
370			  delta += 1000000; \
371			  break; \
372		      default: \
373			  delta += (1000000 * xxs); \
374	       } \
375	    } \
376}
377
378#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
379	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
380
381#ifdef UPCALL_TIMING
382u_long upcall_data[51];
383static void collate(struct timeval *);
384#endif /* UPCALL_TIMING */
385
386
387/*
388 * Handle MRT setsockopt commands to modify the multicast routing tables.
389 */
390static int
391X_ip_mrouter_set(cmd, so, m)
392    int cmd;
393    struct socket *so;
394    struct mbuf *m;
395{
396   if (cmd != MRT_INIT && so != ip_mrouter) return EACCES;
397
398    switch (cmd) {
399	case MRT_INIT:     return ip_mrouter_init(so, m);
400	case MRT_DONE:     return ip_mrouter_done();
401	case MRT_ADD_VIF:  return add_vif (mtod(m, struct vifctl *));
402	case MRT_DEL_VIF:  return del_vif (mtod(m, vifi_t *));
403	case MRT_ADD_MFC:  return add_mfc (mtod(m, struct mfcctl *));
404	case MRT_DEL_MFC:  return del_mfc (mtod(m, struct mfcctl *));
405	case MRT_ASSERT:   return set_assert(mtod(m, int *));
406	default:             return EOPNOTSUPP;
407    }
408}
409
410#ifndef MROUTE_LKM
411int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = X_ip_mrouter_set;
412#endif
413
414/*
415 * Handle MRT getsockopt commands
416 */
417static int
418X_ip_mrouter_get(cmd, so, m)
419    int cmd;
420    struct socket *so;
421    struct mbuf **m;
422{
423    struct mbuf *mb;
424
425    if (so != ip_mrouter) return EACCES;
426
427    *m = mb = m_get(M_WAIT, MT_SOOPTS);
428
429    switch (cmd) {
430	case MRT_VERSION:   return get_version(mb);
431	case MRT_ASSERT:    return get_assert(mb);
432	default:            return EOPNOTSUPP;
433    }
434}
435
436#ifndef MROUTE_LKM
437int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = X_ip_mrouter_get;
438#endif
439
440/*
441 * Handle ioctl commands to obtain information from the cache
442 */
443static int
444X_mrt_ioctl(cmd, data)
445    int cmd;
446    caddr_t data;
447{
448    int error = 0;
449
450    switch (cmd) {
451	case (SIOCGETVIFCNT):
452	    return (get_vif_cnt((struct sioc_vif_req *)data));
453	    break;
454	case (SIOCGETSGCNT):
455	    return (get_sg_cnt((struct sioc_sg_req *)data));
456	    break;
457	default:
458	    return (EINVAL);
459	    break;
460    }
461    return error;
462}
463
464#ifndef MROUTE_LKM
465int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl;
466#endif
467
468/*
469 * returns the packet, byte, rpf-failure count for the source group provided
470 */
471static int
472get_sg_cnt(req)
473    register struct sioc_sg_req *req;
474{
475    register struct mfc *rt;
476    int s;
477
478    s = splnet();
479    MFCFIND(req->src.s_addr, req->grp.s_addr, rt);
480    splx(s);
481    if (rt != NULL) {
482	req->pktcnt = rt->mfc_pkt_cnt;
483	req->bytecnt = rt->mfc_byte_cnt;
484	req->wrong_if = rt->mfc_wrong_if;
485    } else
486	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
487
488    return 0;
489}
490
491/*
492 * returns the input and output packet and byte counts on the vif provided
493 */
494static int
495get_vif_cnt(req)
496    register struct sioc_vif_req *req;
497{
498    register vifi_t vifi = req->vifi;
499
500    if (vifi >= numvifs) return EINVAL;
501
502    req->icount = viftable[vifi].v_pkt_in;
503    req->ocount = viftable[vifi].v_pkt_out;
504    req->ibytes = viftable[vifi].v_bytes_in;
505    req->obytes = viftable[vifi].v_bytes_out;
506
507    return 0;
508}
509
510/*
511 * Enable multicast routing
512 */
513static int
514ip_mrouter_init(so, m)
515	struct socket *so;
516	struct mbuf *m;
517{
518    int *v;
519
520    if (mrtdebug)
521	log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
522		so->so_type, so->so_proto->pr_protocol);
523
524    if (so->so_type != SOCK_RAW ||
525	so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP;
526
527    if (!m || (m->m_len != sizeof(int *)))
528	return ENOPROTOOPT;
529
530    v = mtod(m, int *);
531    if (*v != 1)
532	return ENOPROTOOPT;
533
534    if (ip_mrouter != NULL) return EADDRINUSE;
535
536    ip_mrouter = so;
537
538    bzero((caddr_t)mfctable, sizeof(mfctable));
539    bzero((caddr_t)nexpire, sizeof(nexpire));
540
541    pim_assert = 0;
542
543    timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
544
545    if (mrtdebug)
546	log(LOG_DEBUG, "ip_mrouter_init\n");
547
548    return 0;
549}
550
551/*
552 * Disable multicast routing
553 */
554static int
555X_ip_mrouter_done()
556{
557    vifi_t vifi;
558    int i;
559    struct ifnet *ifp;
560    struct ifreq ifr;
561    struct mbuf *mb_rt;
562    struct mbuf *m;
563    struct rtdetq *rte;
564    int s;
565
566    s = splnet();
567
568    /*
569     * For each phyint in use, disable promiscuous reception of all IP
570     * multicasts.
571     */
572    for (vifi = 0; vifi < numvifs; vifi++) {
573	if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
574	    !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
575	    ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
576	    ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr
577								= INADDR_ANY;
578	    ifp = viftable[vifi].v_ifp;
579	    if_allmulti(ifp, 0);
580	}
581    }
582    bzero((caddr_t)tbftable, sizeof(tbftable));
583    bzero((caddr_t)viftable, sizeof(viftable));
584    numvifs = 0;
585    pim_assert = 0;
586
587    untimeout(expire_upcalls, (caddr_t)NULL);
588
589    /*
590     * Free all multicast forwarding cache entries.
591     */
592    for (i = 0; i < MFCTBLSIZ; i++) {
593	mb_rt = mfctable[i];
594	while (mb_rt) {
595	    if (mb_rt->m_act != NULL) {
596		while (mb_rt->m_act) {
597		    m = mb_rt->m_act;
598		    mb_rt->m_act = m->m_act;
599		    rte = mtod(m, struct rtdetq *);
600		    m_freem(rte->m);
601		    m_free(m);
602		}
603	    }
604	    mb_rt = m_free(mb_rt);
605	}
606    }
607
608    bzero((caddr_t)mfctable, sizeof(mfctable));
609
610    /*
611     * Reset de-encapsulation cache
612     */
613    last_encap_src = 0;
614    last_encap_vif = NULL;
615    have_encap_tunnel = 0;
616
617    ip_mrouter = NULL;
618
619    splx(s);
620
621    if (mrtdebug)
622	log(LOG_DEBUG, "ip_mrouter_done\n");
623
624    return 0;
625}
626
627#ifndef MROUTE_LKM
628int (*ip_mrouter_done)(void) = X_ip_mrouter_done;
629#endif
630
631static int
632get_version(mb)
633    struct mbuf *mb;
634{
635    int *v;
636
637    v = mtod(mb, int *);
638
639    *v = 0x0305;	/* XXX !!!! */
640    mb->m_len = sizeof(int);
641
642    return 0;
643}
644
645/*
646 * Set PIM assert processing global
647 */
648static int
649set_assert(i)
650    int *i;
651{
652    if ((*i != 1) && (*i != 0))
653	return EINVAL;
654
655    pim_assert = *i;
656
657    return 0;
658}
659
660/*
661 * Get PIM assert processing global
662 */
663static int
664get_assert(m)
665    struct mbuf *m;
666{
667    int *i;
668
669    i = mtod(m, int *);
670
671    *i = pim_assert;
672
673    return 0;
674}
675
676/*
677 * Add a vif to the vif table
678 */
679static int
680add_vif(vifcp)
681    register struct vifctl *vifcp;
682{
683    register struct vif *vifp = viftable + vifcp->vifc_vifi;
684    static struct sockaddr_in sin = {sizeof sin, AF_INET};
685    struct ifaddr *ifa;
686    struct ifnet *ifp;
687    struct ifreq ifr;
688    int error, s;
689    struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
690
691    if (vifcp->vifc_vifi >= MAXVIFS)  return EINVAL;
692    if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE;
693
694    /* Find the interface with an address in AF_INET family */
695    sin.sin_addr = vifcp->vifc_lcl_addr;
696    ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
697    if (ifa == 0) return EADDRNOTAVAIL;
698    ifp = ifa->ifa_ifp;
699
700    if (vifcp->vifc_flags & VIFF_TUNNEL) {
701	if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
702		/*
703		 * An encapsulating tunnel is wanted.  Tell ipip_input() to
704		 * start paying attention to encapsulated packets.
705		 */
706		if (have_encap_tunnel == 0) {
707			have_encap_tunnel = 1;
708			for (s = 0; s < MAXVIFS; ++s) {
709				multicast_decap_if[s].if_name = "mdecap";
710				multicast_decap_if[s].if_unit = s;
711			}
712		}
713		/*
714		 * Set interface to fake encapsulator interface
715		 */
716		ifp = &multicast_decap_if[vifcp->vifc_vifi];
717		/*
718		 * Prepare cached route entry
719		 */
720		bzero(&vifp->v_route, sizeof(vifp->v_route));
721	} else {
722	    log(LOG_ERR, "source routed tunnels not supported\n");
723	    return EOPNOTSUPP;
724	}
725    } else {
726	/* Make sure the interface supports multicast */
727	if ((ifp->if_flags & IFF_MULTICAST) == 0)
728	    return EOPNOTSUPP;
729
730	/* Enable promiscuous reception of all IP multicasts from the if */
731	s = splnet();
732	error = if_allmulti(ifp, 1);
733	splx(s);
734	if (error)
735	    return error;
736    }
737
738    s = splnet();
739    /* define parameters for the tbf structure */
740    vifp->v_tbf = v_tbf;
741    GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
742    vifp->v_tbf->tbf_n_tok = 0;
743    vifp->v_tbf->tbf_q_len = 0;
744    vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
745    vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
746
747    vifp->v_flags     = vifcp->vifc_flags;
748    vifp->v_threshold = vifcp->vifc_threshold;
749    vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
750    vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
751    vifp->v_ifp       = ifp;
752    /* scaling up here allows division by 1024 in critical code */
753    vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
754    vifp->v_rsvp_on   = 0;
755    vifp->v_rsvpd     = NULL;
756    /* initialize per vif pkt counters */
757    vifp->v_pkt_in    = 0;
758    vifp->v_pkt_out   = 0;
759    vifp->v_bytes_in  = 0;
760    vifp->v_bytes_out = 0;
761    splx(s);
762
763    /* Adjust numvifs up if the vifi is higher than numvifs */
764    if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
765
766    if (mrtdebug)
767	log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
768	    vifcp->vifc_vifi,
769	    ntohl(vifcp->vifc_lcl_addr.s_addr),
770	    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
771	    ntohl(vifcp->vifc_rmt_addr.s_addr),
772	    vifcp->vifc_threshold,
773	    vifcp->vifc_rate_limit);
774
775    return 0;
776}
777
778/*
779 * Delete a vif from the vif table
780 */
781static int
782del_vif(vifip)
783    vifi_t *vifip;
784{
785    register struct vif *vifp = viftable + *vifip;
786    register vifi_t vifi;
787    register struct mbuf *m;
788    struct ifnet *ifp;
789    struct ifreq ifr;
790    int s;
791
792    if (*vifip >= 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    /* Adjust numvifs down */
822    for (vifi = numvifs; vifi > 0; vifi--)
823	if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break;
824    numvifs = vifi;
825
826    splx(s);
827
828    if (mrtdebug)
829      log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
830
831    return 0;
832}
833
834/*
835 * Add an mfc entry
836 */
837static int
838add_mfc(mfccp)
839    struct mfcctl *mfccp;
840{
841    struct mfc *rt;
842    register struct mbuf *mb_rt;
843    u_long hash;
844    struct mbuf *mb_ntry;
845    struct rtdetq *rte;
846    register u_short nstl;
847    int s;
848    int i;
849
850    MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt);
851
852    /* If an entry already exists, just update the fields */
853    if (rt) {
854	if (mrtdebug & DEBUG_MFC)
855	    log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n",
856		ntohl(mfccp->mfcc_origin.s_addr),
857		ntohl(mfccp->mfcc_mcastgrp.s_addr),
858		mfccp->mfcc_parent);
859
860	s = splnet();
861	rt->mfc_parent = mfccp->mfcc_parent;
862	for (i = 0; i < numvifs; i++)
863	    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
864	splx(s);
865	return 0;
866    }
867
868    /*
869     * Find the entry for which the upcall was made and update
870     */
871    s = splnet();
872    hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
873    for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) {
874
875	rt = mtod(mb_rt, struct mfc *);
876	if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
877	    (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
878	    (mb_rt->m_act != NULL)) {
879
880	    if (nstl++)
881		log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %x\n",
882		    "multiple kernel entries",
883		    ntohl(mfccp->mfcc_origin.s_addr),
884		    ntohl(mfccp->mfcc_mcastgrp.s_addr),
885		    mfccp->mfcc_parent, mb_rt->m_act);
886
887	    if (mrtdebug & DEBUG_MFC)
888		log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %x\n",
889		    ntohl(mfccp->mfcc_origin.s_addr),
890		    ntohl(mfccp->mfcc_mcastgrp.s_addr),
891		    mfccp->mfcc_parent, mb_rt->m_act);
892
893	    rt->mfc_origin     = mfccp->mfcc_origin;
894	    rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
895	    rt->mfc_parent     = mfccp->mfcc_parent;
896	    for (i = 0; i < numvifs; i++)
897		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
898	    /* initialize pkt counters per src-grp */
899	    rt->mfc_pkt_cnt    = 0;
900	    rt->mfc_byte_cnt   = 0;
901	    rt->mfc_wrong_if   = 0;
902	    rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
903
904	    rt->mfc_expire = 0;	/* Don't clean this guy up */
905	    nexpire[hash]--;
906
907	    /* free packets Qed at the end of this entry */
908	    while (mb_rt->m_act) {
909		mb_ntry = mb_rt->m_act;
910		rte = mtod(mb_ntry, struct rtdetq *);
911/* #ifdef RSVP_ISI */
912		ip_mdq(rte->m, rte->ifp, rt, -1);
913/* #endif */
914		mb_rt->m_act = mb_ntry->m_act;
915		m_freem(rte->m);
916#ifdef UPCALL_TIMING
917		collate(&(rte->t));
918#endif /* UPCALL_TIMING */
919		m_free(mb_ntry);
920	    }
921	}
922    }
923
924    /*
925     * It is possible that an entry is being inserted without an upcall
926     */
927    if (nstl == 0) {
928	if (mrtdebug & DEBUG_MFC)
929	    log(LOG_DEBUG,"add_mfc no upcall h %d o %x g %x p %x\n",
930		hash, ntohl(mfccp->mfcc_origin.s_addr),
931		ntohl(mfccp->mfcc_mcastgrp.s_addr),
932		mfccp->mfcc_parent);
933
934	for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
935
936	    rt = mtod(mb_rt, struct mfc *);
937	    if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
938		(rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
939
940		rt->mfc_origin     = mfccp->mfcc_origin;
941		rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
942		rt->mfc_parent     = mfccp->mfcc_parent;
943		for (i = 0; i < numvifs; i++)
944		    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
945		/* initialize pkt counters per src-grp */
946		rt->mfc_pkt_cnt    = 0;
947		rt->mfc_byte_cnt   = 0;
948		rt->mfc_wrong_if   = 0;
949		rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
950		if (rt->mfc_expire)
951		    nexpire[hash]--;
952		rt->mfc_expire	   = 0;
953	    }
954	}
955	if (mb_rt == NULL) {
956	    /* no upcall, so make a new entry */
957	    MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
958	    if (mb_rt == NULL) {
959		splx(s);
960		return ENOBUFS;
961	    }
962
963	    rt = mtod(mb_rt, struct mfc *);
964
965	    /* insert new entry at head of hash chain */
966	    rt->mfc_origin     = mfccp->mfcc_origin;
967	    rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
968	    rt->mfc_parent     = mfccp->mfcc_parent;
969	    for (i = 0; i < numvifs; i++)
970		    rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
971	    /* initialize pkt counters per src-grp */
972	    rt->mfc_pkt_cnt    = 0;
973	    rt->mfc_byte_cnt   = 0;
974	    rt->mfc_wrong_if   = 0;
975	    rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
976	    rt->mfc_expire     = 0;
977
978	    /* link into table */
979	    mb_rt->m_next  = mfctable[hash];
980	    mfctable[hash] = mb_rt;
981	    mb_rt->m_act = NULL;
982	}
983    }
984    splx(s);
985    return 0;
986}
987
988#ifdef UPCALL_TIMING
989/*
990 * collect delay statistics on the upcalls
991 */
992static void collate(t)
993register struct timeval *t;
994{
995    register u_long d;
996    register struct timeval tp;
997    register u_long delta;
998
999    GET_TIME(tp);
1000
1001    if (TV_LT(*t, tp))
1002    {
1003	TV_DELTA(tp, *t, delta);
1004
1005	d = delta >> 10;
1006	if (d > 50)
1007	    d = 50;
1008
1009	++upcall_data[d];
1010    }
1011}
1012#endif /* UPCALL_TIMING */
1013
1014/*
1015 * Delete an mfc entry
1016 */
1017static int
1018del_mfc(mfccp)
1019    struct mfcctl *mfccp;
1020{
1021    struct in_addr 	origin;
1022    struct in_addr 	mcastgrp;
1023    struct mfc 		*rt;
1024    struct mbuf 	*mb_rt;
1025    struct mbuf 	**nptr;
1026    u_long 		hash;
1027    int s;
1028
1029    origin = mfccp->mfcc_origin;
1030    mcastgrp = mfccp->mfcc_mcastgrp;
1031    hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1032
1033    if (mrtdebug & DEBUG_MFC)
1034	log(LOG_DEBUG,"del_mfc orig %x mcastgrp %x\n",
1035	    ntohl(origin.s_addr), ntohl(mcastgrp.s_addr));
1036
1037    s = splnet();
1038
1039    nptr = &mfctable[hash];
1040    while ((mb_rt = *nptr) != NULL) {
1041        rt = mtod(mb_rt, struct mfc *);
1042	if (origin.s_addr == rt->mfc_origin.s_addr &&
1043	    mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1044	    mb_rt->m_act == NULL)
1045	    break;
1046
1047	nptr = &mb_rt->m_next;
1048    }
1049    if (mb_rt == NULL) {
1050	splx(s);
1051	return EADDRNOTAVAIL;
1052    }
1053
1054    MFREE(mb_rt, *nptr);
1055
1056    splx(s);
1057
1058    return 0;
1059}
1060
1061/*
1062 * Send a message to mrouted on the multicast routing socket
1063 */
1064static int
1065socket_send(s, mm, src)
1066	struct socket *s;
1067	struct mbuf *mm;
1068	struct sockaddr_in *src;
1069{
1070	if (s) {
1071		if (sbappendaddr(&s->so_rcv,
1072				 (struct sockaddr *)src,
1073				 mm, (struct mbuf *)0) != 0) {
1074			sorwakeup(s);
1075			return 0;
1076		}
1077	}
1078	m_freem(mm);
1079	return -1;
1080}
1081
1082/*
1083 * IP multicast forwarding function. This function assumes that the packet
1084 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1085 * pointed to by "ifp", and the packet is to be relayed to other networks
1086 * that have members of the packet's destination IP multicast group.
1087 *
1088 * The packet is returned unscathed to the caller, unless it is
1089 * erroneous, in which case a non-zero return value tells the caller to
1090 * discard it.
1091 */
1092
1093#define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1094#define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1095
1096static int
1097X_ip_mforward(ip, ifp, m, imo)
1098    register struct ip *ip;
1099    struct ifnet *ifp;
1100    struct mbuf *m;
1101    struct ip_moptions *imo;
1102{
1103    register struct mfc *rt;
1104    register u_char *ipoptions;
1105    static struct sockaddr_in 	k_igmpsrc	= { sizeof k_igmpsrc, AF_INET };
1106    static int srctun = 0;
1107    register struct mbuf *mm;
1108    int s;
1109    vifi_t vifi;
1110    struct vif *vifp;
1111
1112    if (mrtdebug & DEBUG_FORWARD)
1113	log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %x\n",
1114	    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1115
1116    if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1117	(ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1118	/*
1119	 * Packet arrived via a physical interface or
1120	 * an encapsulated tunnel.
1121	 */
1122    } else {
1123	/*
1124	 * Packet arrived through a source-route tunnel.
1125	 * Source-route tunnels are no longer supported.
1126	 */
1127	if ((srctun++ % 1000) == 0)
1128	    log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n",
1129		ntohl(ip->ip_src.s_addr));
1130
1131	return 1;
1132    }
1133
1134    if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1135	if (ip->ip_ttl < 255)
1136		ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1137	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1138	    vifp = viftable + vifi;
1139	    printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n",
1140		ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi,
1141		(vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1142		vifp->v_ifp->if_name, vifp->v_ifp->if_unit);
1143	}
1144	return (ip_mdq(m, ifp, NULL, vifi));
1145    }
1146    if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1147	printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1148	    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr));
1149	if(!imo)
1150		printf("In fact, no options were specified at all\n");
1151    }
1152
1153    /*
1154     * Don't forward a packet with time-to-live of zero or one,
1155     * or a packet destined to a local-only group.
1156     */
1157    if (ip->ip_ttl <= 1 ||
1158	ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
1159	return 0;
1160
1161    /*
1162     * Determine forwarding vifs from the forwarding cache table
1163     */
1164    s = splnet();
1165    MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt);
1166
1167    /* Entry exists, so forward if necessary */
1168    if (rt != NULL) {
1169	splx(s);
1170	return (ip_mdq(m, ifp, rt, -1));
1171    } else {
1172	/*
1173	 * If we don't have a route for packet's origin,
1174	 * Make a copy of the packet &
1175	 * send message to routing daemon
1176	 */
1177
1178	register struct mbuf *mb_rt;
1179	register struct mbuf *mb_ntry;
1180	register struct mbuf *mb0;
1181	register struct rtdetq *rte;
1182	register struct mbuf *rte_m;
1183	register u_long hash;
1184	register int npkts;
1185	int hlen = ip->ip_hl << 2;
1186#ifdef UPCALL_TIMING
1187	struct timeval tp;
1188
1189	GET_TIME(tp);
1190#endif
1191
1192	mrtstat.mrts_no_route++;
1193	if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1194	    log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1195		ntohl(ip->ip_src.s_addr),
1196		ntohl(ip->ip_dst.s_addr));
1197
1198	/*
1199	 * Allocate mbufs early so that we don't do extra work if we are
1200	 * just going to fail anyway.  Make sure to pullup the header so
1201	 * that other people can't step on it.
1202	 */
1203	MGET(mb_ntry, M_DONTWAIT, MT_DATA);
1204	if (mb_ntry == NULL) {
1205	    splx(s);
1206	    return ENOBUFS;
1207	}
1208	mb0 = m_copy(m, 0, M_COPYALL);
1209	if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1210	    mb0 = m_pullup(mb0, hlen);
1211	if (mb0 == NULL) {
1212	    m_free(mb_ntry);
1213	    splx(s);
1214	    return ENOBUFS;
1215	}
1216
1217	/* is there an upcall waiting for this packet? */
1218	hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1219	for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
1220	    rt = mtod(mb_rt, struct mfc *);
1221	    if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1222		(ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1223		(mb_rt->m_act != NULL))
1224		break;
1225	}
1226
1227	if (mb_rt == NULL) {
1228	    int i;
1229	    struct igmpmsg *im;
1230
1231	    /* no upcall, so make a new entry */
1232	    MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
1233	    if (mb_rt == NULL) {
1234		m_free(mb_ntry);
1235		m_freem(mb0);
1236		splx(s);
1237		return ENOBUFS;
1238	    }
1239	    /* Make a copy of the header to send to the user level process */
1240	    mm = m_copy(mb0, 0, hlen);
1241	    if (mm == NULL) {
1242		m_free(mb_ntry);
1243		m_freem(mb0);
1244		m_free(mb_rt);
1245		splx(s);
1246		return ENOBUFS;
1247	    }
1248
1249	    /*
1250	     * Send message to routing daemon to install
1251	     * a route into the kernel table
1252	     */
1253	    k_igmpsrc.sin_addr = ip->ip_src;
1254
1255	    im = mtod(mm, struct igmpmsg *);
1256	    im->im_msgtype	= IGMPMSG_NOCACHE;
1257	    im->im_mbz		= 0;
1258
1259	    mrtstat.mrts_upcalls++;
1260
1261	    if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1262		log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1263		++mrtstat.mrts_upq_sockfull;
1264		m_free(mb_ntry);
1265		m_freem(mb0);
1266		m_free(mb_rt);
1267		splx(s);
1268		return ENOBUFS;
1269	    }
1270
1271	    rt = mtod(mb_rt, struct mfc *);
1272
1273	    /* insert new entry at head of hash chain */
1274	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1275	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1276	    rt->mfc_expire	      = UPCALL_EXPIRE;
1277	    nexpire[hash]++;
1278	    for (i = 0; i < numvifs; i++)
1279		rt->mfc_ttls[i] = 0;
1280	    rt->mfc_parent = -1;
1281
1282	    /* link into table */
1283	    mb_rt->m_next  = mfctable[hash];
1284	    mfctable[hash] = mb_rt;
1285	    mb_rt->m_act = NULL;
1286
1287	    rte_m = mb_rt;
1288	} else {
1289	    /* determine if q has overflowed */
1290	    for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act)
1291		npkts++;
1292
1293	    if (npkts > MAX_UPQ) {
1294		mrtstat.mrts_upq_ovflw++;
1295		m_free(mb_ntry);
1296		m_freem(mb0);
1297		splx(s);
1298		return 0;
1299	    }
1300	}
1301
1302	mb_ntry->m_act = NULL;
1303	rte = mtod(mb_ntry, struct rtdetq *);
1304
1305	rte->m 			= mb0;
1306	rte->ifp 		= ifp;
1307#ifdef UPCALL_TIMING
1308	rte->t			= tp;
1309#endif
1310
1311	/* Add this entry to the end of the queue */
1312	rte_m->m_act		= mb_ntry;
1313
1314	splx(s);
1315
1316	return 0;
1317    }
1318}
1319
1320#ifndef MROUTE_LKM
1321int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
1322		   struct ip_moptions *) = X_ip_mforward;
1323#endif
1324
1325/*
1326 * Clean up the cache entry if upcall is not serviced
1327 */
1328static void
1329expire_upcalls(void *unused)
1330{
1331    struct mbuf *mb_rt, *m, **nptr;
1332    struct rtdetq *rte;
1333    struct mfc *mfc;
1334    int i;
1335    int s;
1336
1337    s = splnet();
1338    for (i = 0; i < MFCTBLSIZ; i++) {
1339	if (nexpire[i] == 0)
1340	    continue;
1341	nptr = &mfctable[i];
1342	for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) {
1343	    mfc = mtod(mb_rt, struct mfc *);
1344
1345	    /*
1346	     * Skip real cache entries
1347	     * Make sure it wasn't marked to not expire (shouldn't happen)
1348	     * If it expires now
1349	     */
1350	    if (mb_rt->m_act != NULL &&
1351	        mfc->mfc_expire != 0 &&
1352		--mfc->mfc_expire == 0) {
1353		if (mrtdebug & DEBUG_EXPIRE)
1354		    log(LOG_DEBUG, "expire_upcalls: expiring (%x %x)\n",
1355			ntohl(mfc->mfc_origin.s_addr),
1356			ntohl(mfc->mfc_mcastgrp.s_addr));
1357		/*
1358		 * drop all the packets
1359		 * free the mbuf with the pkt, if, timing info
1360		 */
1361		while (mb_rt->m_act) {
1362		    m = mb_rt->m_act;
1363		    mb_rt->m_act = m->m_act;
1364
1365		    rte = mtod(m, struct rtdetq *);
1366		    m_freem(rte->m);
1367		    m_free(m);
1368		}
1369		++mrtstat.mrts_cache_cleanups;
1370		nexpire[i]--;
1371
1372		MFREE(mb_rt, *nptr);
1373	    } else {
1374		nptr = &mb_rt->m_next;
1375	    }
1376	}
1377    }
1378    splx(s);
1379    timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
1380}
1381
1382/*
1383 * Packet forwarding routine once entry in the cache is made
1384 */
1385static int
1386ip_mdq(m, ifp, rt, xmt_vif)
1387    register struct mbuf *m;
1388    register struct ifnet *ifp;
1389    register struct mfc *rt;
1390    register vifi_t xmt_vif;
1391{
1392    register struct ip  *ip = mtod(m, struct ip *);
1393    register vifi_t vifi;
1394    register struct vif *vifp;
1395    register int plen = ntohs(ip->ip_len);
1396
1397/*
1398 * Macro to send packet on vif.  Since RSVP packets don't get counted on
1399 * input, they shouldn't get counted on output, so statistics keeping is
1400 * seperate.
1401 */
1402#define MC_SEND(ip,vifp,m) {                             \
1403                if ((vifp)->v_flags & VIFF_TUNNEL)  	 \
1404                    encap_send((ip), (vifp), (m));       \
1405                else                                     \
1406                    phyint_send((ip), (vifp), (m));      \
1407}
1408
1409    /*
1410     * If xmt_vif is not -1, send on only the requested vif.
1411     *
1412     * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1413     */
1414    if (xmt_vif < numvifs) {
1415	MC_SEND(ip, viftable + xmt_vif, m);
1416	return 1;
1417    }
1418
1419    /*
1420     * Don't forward if it didn't arrive from the parent vif for its origin.
1421     */
1422    vifi = rt->mfc_parent;
1423    if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1424	/* came in the wrong interface */
1425	if (mrtdebug & DEBUG_FORWARD)
1426	    log(LOG_DEBUG, "wrong if: ifp %x vifi %d vififp %x\n",
1427		ifp, vifi, viftable[vifi].v_ifp);
1428	++mrtstat.mrts_wrong_if;
1429	++rt->mfc_wrong_if;
1430	/*
1431	 * If we are doing PIM assert processing, and we are forwarding
1432	 * packets on this interface, and it is a broadcast medium
1433	 * interface (and not a tunnel), send a message to the routing daemon.
1434	 */
1435	if (pim_assert && rt->mfc_ttls[vifi] &&
1436		(ifp->if_flags & IFF_BROADCAST) &&
1437		!(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1438	    struct sockaddr_in k_igmpsrc;
1439	    struct mbuf *mm;
1440	    struct igmpmsg *im;
1441	    int hlen = ip->ip_hl << 2;
1442	    struct timeval now;
1443	    register u_long delta;
1444
1445	    GET_TIME(now);
1446
1447	    TV_DELTA(rt->mfc_last_assert, now, delta);
1448
1449	    if (delta > ASSERT_MSG_TIME) {
1450		mm = m_copy(m, 0, hlen);
1451		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1452		    mm = m_pullup(mm, hlen);
1453		if (mm == NULL) {
1454		    return ENOBUFS;
1455		}
1456
1457		rt->mfc_last_assert = now;
1458
1459		im = mtod(mm, struct igmpmsg *);
1460		im->im_msgtype	= IGMPMSG_WRONGVIF;
1461		im->im_mbz		= 0;
1462		im->im_vif		= vifi;
1463
1464		k_igmpsrc.sin_addr = im->im_src;
1465
1466		socket_send(ip_mrouter, mm, &k_igmpsrc);
1467	    }
1468	}
1469	return 0;
1470    }
1471
1472    /* If I sourced this packet, it counts as output, else it was input. */
1473    if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1474	viftable[vifi].v_pkt_out++;
1475	viftable[vifi].v_bytes_out += plen;
1476    } else {
1477	viftable[vifi].v_pkt_in++;
1478	viftable[vifi].v_bytes_in += plen;
1479    }
1480    rt->mfc_pkt_cnt++;
1481    rt->mfc_byte_cnt += plen;
1482
1483    /*
1484     * For each vif, decide if a copy of the packet should be forwarded.
1485     * Forward if:
1486     *		- the ttl exceeds the vif's threshold
1487     *		- there are group members downstream on interface
1488     */
1489    for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1490	if ((rt->mfc_ttls[vifi] > 0) &&
1491	    (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1492	    vifp->v_pkt_out++;
1493	    vifp->v_bytes_out += plen;
1494	    MC_SEND(ip, vifp, m);
1495	}
1496
1497    return 0;
1498}
1499
1500/*
1501 * check if a vif number is legal/ok. This is used by ip_output, to export
1502 * numvifs there,
1503 */
1504static int
1505X_legal_vif_num(vif)
1506    int vif;
1507{
1508    if (vif >= 0 && vif < numvifs)
1509       return(1);
1510    else
1511       return(0);
1512}
1513
1514#ifndef MROUTE_LKM
1515int (*legal_vif_num)(int) = X_legal_vif_num;
1516#endif
1517
1518/*
1519 * Return the local address used by this vif
1520 */
1521static u_long
1522X_ip_mcast_src(vifi)
1523    int vifi;
1524{
1525    if (vifi >= 0 && vifi < numvifs)
1526	return viftable[vifi].v_lcl_addr.s_addr;
1527    else
1528	return INADDR_ANY;
1529}
1530
1531#ifndef MROUTE_LKM
1532u_long (*ip_mcast_src)(int) = X_ip_mcast_src;
1533#endif
1534
1535static void
1536phyint_send(ip, vifp, m)
1537    struct ip *ip;
1538    struct vif *vifp;
1539    struct mbuf *m;
1540{
1541    register struct mbuf *mb_copy;
1542    register int hlen = ip->ip_hl << 2;
1543
1544    /*
1545     * Make a new reference to the packet; make sure that
1546     * the IP header is actually copied, not just referenced,
1547     * so that ip_output() only scribbles on the copy.
1548     */
1549    mb_copy = m_copy(m, 0, M_COPYALL);
1550    if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1551	mb_copy = m_pullup(mb_copy, hlen);
1552    if (mb_copy == NULL)
1553	return;
1554
1555    if (vifp->v_rate_limit <= 0)
1556	tbf_send_packet(vifp, mb_copy);
1557    else
1558	tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1559}
1560
1561static void
1562encap_send(ip, vifp, m)
1563    register struct ip *ip;
1564    register struct vif *vifp;
1565    register struct mbuf *m;
1566{
1567    register struct mbuf *mb_copy;
1568    register struct ip *ip_copy;
1569    register int i, len = ip->ip_len;
1570
1571    /*
1572     * copy the old packet & pullup it's IP header into the
1573     * new mbuf so we can modify it.  Try to fill the new
1574     * mbuf since if we don't the ethernet driver will.
1575     */
1576    MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER);
1577    if (mb_copy == NULL)
1578	return;
1579    mb_copy->m_data += max_linkhdr;
1580    mb_copy->m_len = sizeof(multicast_encap_iphdr);
1581
1582    if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1583	m_freem(mb_copy);
1584	return;
1585    }
1586    i = MHLEN - M_LEADINGSPACE(mb_copy);
1587    if (i > len)
1588	i = len;
1589    mb_copy = m_pullup(mb_copy, i);
1590    if (mb_copy == NULL)
1591	return;
1592    mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1593
1594    /*
1595     * fill in the encapsulating IP header.
1596     */
1597    ip_copy = mtod(mb_copy, struct ip *);
1598    *ip_copy = multicast_encap_iphdr;
1599    ip_copy->ip_id = htons(ip_id++);
1600    ip_copy->ip_len += len;
1601    ip_copy->ip_src = vifp->v_lcl_addr;
1602    ip_copy->ip_dst = vifp->v_rmt_addr;
1603
1604    /*
1605     * turn the encapsulated IP header back into a valid one.
1606     */
1607    ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1608    --ip->ip_ttl;
1609    HTONS(ip->ip_len);
1610    HTONS(ip->ip_off);
1611    ip->ip_sum = 0;
1612#if defined(LBL) && !defined(ultrix)
1613    ip->ip_sum = ~oc_cksum((caddr_t)ip, ip->ip_hl << 2, 0);
1614#else
1615    mb_copy->m_data += sizeof(multicast_encap_iphdr);
1616    ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1617    mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1618#endif
1619
1620    if (vifp->v_rate_limit <= 0)
1621	tbf_send_packet(vifp, mb_copy);
1622    else
1623	tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1624}
1625
1626/*
1627 * De-encapsulate a packet and feed it back through ip input (this
1628 * routine is called whenever IP gets a packet with proto type
1629 * ENCAP_PROTO and a local destination address).
1630 */
1631void
1632#ifdef MROUTE_LKM
1633X_ipip_input(m, iphlen)
1634#else
1635ipip_input(m, iphlen)
1636#endif
1637	register struct mbuf *m;
1638	int iphlen;
1639{
1640    struct ifnet *ifp = m->m_pkthdr.rcvif;
1641    register struct ip *ip = mtod(m, struct ip *);
1642    register int hlen = ip->ip_hl << 2;
1643    register int s;
1644    register struct ifqueue *ifq;
1645    register struct vif *vifp;
1646
1647    if (!have_encap_tunnel) {
1648	    rip_input(m, iphlen);
1649	    return;
1650    }
1651    /*
1652     * dump the packet if it's not to a multicast destination or if
1653     * we don't have an encapsulating tunnel with the source.
1654     * Note:  This code assumes that the remote site IP address
1655     * uniquely identifies the tunnel (i.e., that this site has
1656     * at most one tunnel with the remote site).
1657     */
1658    if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) {
1659	++mrtstat.mrts_bad_tunnel;
1660	m_freem(m);
1661	return;
1662    }
1663    if (ip->ip_src.s_addr != last_encap_src) {
1664	register struct vif *vife;
1665
1666	vifp = viftable;
1667	vife = vifp + numvifs;
1668	last_encap_src = ip->ip_src.s_addr;
1669	last_encap_vif = 0;
1670	for ( ; vifp < vife; ++vifp)
1671	    if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
1672		if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT))
1673		    == VIFF_TUNNEL)
1674		    last_encap_vif = vifp;
1675		break;
1676	    }
1677    }
1678    if ((vifp = last_encap_vif) == 0) {
1679	last_encap_src = 0;
1680	mrtstat.mrts_cant_tunnel++; /*XXX*/
1681	m_freem(m);
1682	if (mrtdebug)
1683          log(LOG_DEBUG, "ip_mforward: no tunnel with %x\n",
1684		ntohl(ip->ip_src.s_addr));
1685	return;
1686    }
1687    ifp = vifp->v_ifp;
1688
1689    if (hlen > IP_HDR_LEN)
1690      ip_stripoptions(m, (struct mbuf *) 0);
1691    m->m_data += IP_HDR_LEN;
1692    m->m_len -= IP_HDR_LEN;
1693    m->m_pkthdr.len -= IP_HDR_LEN;
1694    m->m_pkthdr.rcvif = ifp;
1695
1696    ifq = &ipintrq;
1697    s = splimp();
1698    if (IF_QFULL(ifq)) {
1699	IF_DROP(ifq);
1700	m_freem(m);
1701    } else {
1702	IF_ENQUEUE(ifq, m);
1703	/*
1704	 * normally we would need a "schednetisr(NETISR_IP)"
1705	 * here but we were called by ip_input and it is going
1706	 * to loop back & try to dequeue the packet we just
1707	 * queued as soon as we return so we avoid the
1708	 * unnecessary software interrrupt.
1709	 */
1710    }
1711    splx(s);
1712}
1713
1714/*
1715 * Token bucket filter module
1716 */
1717
1718static void
1719tbf_control(vifp, m, ip, p_len)
1720	register struct vif *vifp;
1721	register struct mbuf *m;
1722	register struct ip *ip;
1723	register u_long p_len;
1724{
1725    register struct tbf *t = vifp->v_tbf;
1726
1727    if (p_len > MAX_BKT_SIZE) {
1728	/* drop if packet is too large */
1729	mrtstat.mrts_pkt2large++;
1730	m_freem(m);
1731	return;
1732    }
1733
1734    tbf_update_tokens(vifp);
1735
1736    /* if there are enough tokens,
1737     * and the queue is empty,
1738     * send this packet out
1739     */
1740
1741    if (t->tbf_q_len == 0) {
1742	/* queue empty, send packet if enough tokens */
1743	if (p_len <= t->tbf_n_tok) {
1744	    t->tbf_n_tok -= p_len;
1745	    tbf_send_packet(vifp, m);
1746	} else {
1747	    /* queue packet and timeout till later */
1748	    tbf_queue(vifp, m);
1749	    timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1750	}
1751    } else if (t->tbf_q_len < t->tbf_max_q_len) {
1752	/* finite queue length, so queue pkts and process queue */
1753	tbf_queue(vifp, m);
1754	tbf_process_q(vifp);
1755    } else {
1756	/* queue length too much, try to dq and queue and process */
1757	if (!tbf_dq_sel(vifp, ip)) {
1758	    mrtstat.mrts_q_overflow++;
1759	    m_freem(m);
1760	    return;
1761	} else {
1762	    tbf_queue(vifp, m);
1763	    tbf_process_q(vifp);
1764	}
1765    }
1766    return;
1767}
1768
1769/*
1770 * adds a packet to the queue at the interface
1771 */
1772static void
1773tbf_queue(vifp, m)
1774	register struct vif *vifp;
1775	register struct mbuf *m;
1776{
1777    register int s = splnet();
1778    register struct tbf *t = vifp->v_tbf;
1779
1780    if (t->tbf_t == NULL) {
1781	/* Queue was empty */
1782	t->tbf_q = m;
1783    } else {
1784	/* Insert at tail */
1785	t->tbf_t->m_act = m;
1786    }
1787
1788    /* Set new tail pointer */
1789    t->tbf_t = m;
1790
1791#ifdef DIAGNOSTIC
1792    /* Make sure we didn't get fed a bogus mbuf */
1793    if (m->m_act)
1794	panic("tbf_queue: m_act");
1795#endif
1796    m->m_act = NULL;
1797
1798    t->tbf_q_len++;
1799
1800    splx(s);
1801}
1802
1803
1804/*
1805 * processes the queue at the interface
1806 */
1807static void
1808tbf_process_q(vifp)
1809    register struct vif *vifp;
1810{
1811    register struct mbuf *m;
1812    register int len;
1813    register int s = splnet();
1814    register struct tbf *t = vifp->v_tbf;
1815
1816    /* loop through the queue at the interface and send as many packets
1817     * as possible
1818     */
1819    while (t->tbf_q_len > 0) {
1820	m = t->tbf_q;
1821
1822	len = mtod(m, struct ip *)->ip_len;
1823
1824	/* determine if the packet can be sent */
1825	if (len <= t->tbf_n_tok) {
1826	    /* if so,
1827	     * reduce no of tokens, dequeue the packet,
1828	     * send the packet.
1829	     */
1830	    t->tbf_n_tok -= len;
1831
1832	    t->tbf_q = m->m_act;
1833	    if (--t->tbf_q_len == 0)
1834		t->tbf_t = NULL;
1835
1836	    m->m_act = NULL;
1837	    tbf_send_packet(vifp, m);
1838
1839	} else break;
1840    }
1841    splx(s);
1842}
1843
1844static void
1845tbf_reprocess_q(xvifp)
1846	void *xvifp;
1847{
1848    register struct vif *vifp = xvifp;
1849    if (ip_mrouter == NULL)
1850	return;
1851
1852    tbf_update_tokens(vifp);
1853
1854    tbf_process_q(vifp);
1855
1856    if (vifp->v_tbf->tbf_q_len)
1857	timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1858}
1859
1860/* function that will selectively discard a member of the queue
1861 * based on the precedence value and the priority
1862 */
1863static int
1864tbf_dq_sel(vifp, ip)
1865    register struct vif *vifp;
1866    register struct ip *ip;
1867{
1868    register int s = splnet();
1869    register u_int p;
1870    register struct mbuf *m, *last;
1871    register struct mbuf **np;
1872    register struct tbf *t = vifp->v_tbf;
1873
1874    p = priority(vifp, ip);
1875
1876    np = &t->tbf_q;
1877    last = NULL;
1878    while ((m = *np) != NULL) {
1879	if (p > priority(vifp, mtod(m, struct ip *))) {
1880	    *np = m->m_act;
1881	    /* If we're removing the last packet, fix the tail pointer */
1882	    if (m == t->tbf_t)
1883		t->tbf_t = last;
1884	    m_freem(m);
1885	    /* it's impossible for the queue to be empty, but
1886	     * we check anyway. */
1887	    if (--t->tbf_q_len == 0)
1888		t->tbf_t = NULL;
1889	    splx(s);
1890	    mrtstat.mrts_drop_sel++;
1891	    return(1);
1892	}
1893	np = &m->m_act;
1894	last = m;
1895    }
1896    splx(s);
1897    return(0);
1898}
1899
1900static void
1901tbf_send_packet(vifp, m)
1902    register struct vif *vifp;
1903    register struct mbuf *m;
1904{
1905    struct ip_moptions imo;
1906    int error;
1907    static struct route ro;
1908    int s = splnet();
1909
1910    if (vifp->v_flags & VIFF_TUNNEL) {
1911	/* If tunnel options */
1912	ip_output(m, (struct mbuf *)0, &vifp->v_route,
1913		  IP_FORWARDING, (struct ip_moptions *)0);
1914    } else {
1915	imo.imo_multicast_ifp  = vifp->v_ifp;
1916	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1917	imo.imo_multicast_loop = 1;
1918	imo.imo_multicast_vif  = -1;
1919
1920	/*
1921	 * Re-entrancy should not be a problem here, because
1922	 * the packets that we send out and are looped back at us
1923	 * should get rejected because they appear to come from
1924	 * the loopback interface, thus preventing looping.
1925	 */
1926	error = ip_output(m, (struct mbuf *)0, &ro,
1927			  IP_FORWARDING, &imo);
1928
1929	if (mrtdebug & DEBUG_XMIT)
1930	    log(LOG_DEBUG, "phyint_send on vif %d err %d\n",
1931		vifp - viftable, error);
1932    }
1933    splx(s);
1934}
1935
1936/* determine the current time and then
1937 * the elapsed time (between the last time and time now)
1938 * in milliseconds & update the no. of tokens in the bucket
1939 */
1940static void
1941tbf_update_tokens(vifp)
1942    register struct vif *vifp;
1943{
1944    struct timeval tp;
1945    register u_long tm;
1946    register int s = splnet();
1947    register struct tbf *t = vifp->v_tbf;
1948
1949    GET_TIME(tp);
1950
1951    TV_DELTA(tp, t->tbf_last_pkt_t, tm);
1952
1953    /*
1954     * This formula is actually
1955     * "time in seconds" * "bytes/second".
1956     *
1957     * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
1958     *
1959     * The (1000/1024) was introduced in add_vif to optimize
1960     * this divide into a shift.
1961     */
1962    t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8;
1963    t->tbf_last_pkt_t = tp;
1964
1965    if (t->tbf_n_tok > MAX_BKT_SIZE)
1966	t->tbf_n_tok = MAX_BKT_SIZE;
1967
1968    splx(s);
1969}
1970
1971static int
1972priority(vifp, ip)
1973    register struct vif *vifp;
1974    register struct ip *ip;
1975{
1976    register int prio;
1977
1978    /* temporary hack; may add general packet classifier some day */
1979
1980    /*
1981     * The UDP port space is divided up into four priority ranges:
1982     * [0, 16384)     : unclassified - lowest priority
1983     * [16384, 32768) : audio - highest priority
1984     * [32768, 49152) : whiteboard - medium priority
1985     * [49152, 65536) : video - low priority
1986     */
1987    if (ip->ip_p == IPPROTO_UDP) {
1988	struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
1989	switch (ntohs(udp->uh_dport) & 0xc000) {
1990	    case 0x4000:
1991		prio = 70;
1992		break;
1993	    case 0x8000:
1994		prio = 60;
1995		break;
1996	    case 0xc000:
1997		prio = 55;
1998		break;
1999	    default:
2000		prio = 50;
2001		break;
2002	}
2003	if (tbfdebug > 1)
2004		log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio);
2005    } else {
2006	    prio = 50;
2007    }
2008    return prio;
2009}
2010
2011/*
2012 * End of token bucket filter modifications
2013 */
2014
2015int
2016ip_rsvp_vif_init(so, m)
2017    struct socket *so;
2018    struct mbuf *m;
2019{
2020    int i;
2021    register int s;
2022
2023    if (rsvpdebug)
2024	printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2025	       so->so_type, so->so_proto->pr_protocol);
2026
2027    if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2028	return EOPNOTSUPP;
2029
2030    /* Check mbuf. */
2031    if (m == NULL || m->m_len != sizeof(int)) {
2032	return EINVAL;
2033    }
2034    i = *(mtod(m, int *));
2035
2036    if (rsvpdebug)
2037	printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on);
2038
2039    s = splnet();
2040
2041    /* Check vif. */
2042    if (!legal_vif_num(i)) {
2043	splx(s);
2044	return EADDRNOTAVAIL;
2045    }
2046
2047    /* Check if socket is available. */
2048    if (viftable[i].v_rsvpd != NULL) {
2049	splx(s);
2050	return EADDRINUSE;
2051    }
2052
2053    viftable[i].v_rsvpd = so;
2054    /* This may seem silly, but we need to be sure we don't over-increment
2055     * the RSVP counter, in case something slips up.
2056     */
2057    if (!viftable[i].v_rsvp_on) {
2058	viftable[i].v_rsvp_on = 1;
2059	rsvp_on++;
2060    }
2061
2062    splx(s);
2063    return 0;
2064}
2065
2066int
2067ip_rsvp_vif_done(so, m)
2068    struct socket *so;
2069    struct mbuf *m;
2070{
2071	int i;
2072	register int s;
2073
2074    if (rsvpdebug)
2075	printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2076	       so->so_type, so->so_proto->pr_protocol);
2077
2078    if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2079	return EOPNOTSUPP;
2080
2081    /* Check mbuf. */
2082    if (m == NULL || m->m_len != sizeof(int)) {
2083	    return EINVAL;
2084    }
2085    i = *(mtod(m, int *));
2086
2087    s = splnet();
2088
2089    /* Check vif. */
2090    if (!legal_vif_num(i)) {
2091	splx(s);
2092        return EADDRNOTAVAIL;
2093    }
2094
2095    if (rsvpdebug)
2096	printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n",
2097	       viftable[i].v_rsvpd, so);
2098
2099    viftable[i].v_rsvpd = NULL;
2100    /* This may seem silly, but we need to be sure we don't over-decrement
2101     * the RSVP counter, in case something slips up.
2102     */
2103    if (viftable[i].v_rsvp_on) {
2104	viftable[i].v_rsvp_on = 0;
2105	rsvp_on--;
2106    }
2107
2108    splx(s);
2109    return 0;
2110}
2111
2112void
2113ip_rsvp_force_done(so)
2114    struct socket *so;
2115{
2116    int vifi;
2117    register int s;
2118
2119    /* Don't bother if it is not the right type of socket. */
2120    if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2121	return;
2122
2123    s = splnet();
2124
2125    /* The socket may be attached to more than one vif...this
2126     * is perfectly legal.
2127     */
2128    for (vifi = 0; vifi < numvifs; vifi++) {
2129	if (viftable[vifi].v_rsvpd == so) {
2130	    viftable[vifi].v_rsvpd = NULL;
2131	    /* This may seem silly, but we need to be sure we don't
2132	     * over-decrement the RSVP counter, in case something slips up.
2133	     */
2134	    if (viftable[vifi].v_rsvp_on) {
2135		viftable[vifi].v_rsvp_on = 0;
2136		rsvp_on--;
2137	    }
2138	}
2139    }
2140
2141    splx(s);
2142    return;
2143}
2144
2145void
2146rsvp_input(m, iphlen)
2147	struct mbuf *m;
2148	int iphlen;
2149{
2150    int vifi;
2151    register struct ip *ip = mtod(m, struct ip *);
2152    static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
2153    register int s;
2154    struct ifnet *ifp;
2155
2156    if (rsvpdebug)
2157	printf("rsvp_input: rsvp_on %d\n",rsvp_on);
2158
2159    /* Can still get packets with rsvp_on = 0 if there is a local member
2160     * of the group to which the RSVP packet is addressed.  But in this
2161     * case we want to throw the packet away.
2162     */
2163    if (!rsvp_on) {
2164	m_freem(m);
2165	return;
2166    }
2167
2168    /* If the old-style non-vif-associated socket is set, then use
2169     * it and ignore the new ones.
2170     */
2171    if (ip_rsvpd != NULL) {
2172	if (rsvpdebug)
2173	    printf("rsvp_input: Sending packet up old-style socket\n");
2174	rip_input(m, iphlen);
2175	return;
2176    }
2177
2178    s = splnet();
2179
2180    if (rsvpdebug)
2181	printf("rsvp_input: check vifs\n");
2182
2183#ifdef DIAGNOSTIC
2184    if (!(m->m_flags & M_PKTHDR))
2185	    panic("rsvp_input no hdr");
2186#endif
2187
2188    ifp = m->m_pkthdr.rcvif;
2189    /* Find which vif the packet arrived on. */
2190    for (vifi = 0; vifi < numvifs; vifi++) {
2191	if (viftable[vifi].v_ifp == ifp)
2192 		break;
2193 	}
2194
2195    if (vifi == numvifs) {
2196	/* Can't find vif packet arrived on. Drop packet. */
2197	if (rsvpdebug)
2198	    printf("rsvp_input: Can't find vif for packet...dropping it.\n");
2199	m_freem(m);
2200	splx(s);
2201	return;
2202    }
2203
2204    if (rsvpdebug)
2205	printf("rsvp_input: check socket\n");
2206
2207    if (viftable[vifi].v_rsvpd == NULL) {
2208	/* drop packet, since there is no specific socket for this
2209	 * interface */
2210	    if (rsvpdebug)
2211		    printf("rsvp_input: No socket defined for vif %d\n",vifi);
2212	    m_freem(m);
2213	    splx(s);
2214	    return;
2215    }
2216    rsvp_src.sin_addr = ip->ip_src;
2217
2218    if (rsvpdebug && m)
2219	printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n",
2220	       m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
2221
2222    if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2223	if (rsvpdebug)
2224	    printf("rsvp_input: Failed to append to socket\n");
2225    else
2226	if (rsvpdebug)
2227	    printf("rsvp_input: send packet up\n");
2228
2229    splx(s);
2230}
2231
2232#ifdef MROUTE_LKM
2233#include <sys/conf.h>
2234#include <sys/exec.h>
2235#include <sys/sysent.h>
2236#include <sys/lkm.h>
2237
2238MOD_MISC("ip_mroute_mod")
2239
2240static int
2241ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd)
2242{
2243	int i;
2244	struct lkm_misc	*args = lkmtp->private.lkm_misc;
2245	int err = 0;
2246
2247	switch(cmd) {
2248		static int (*old_ip_mrouter_cmd)();
2249		static int (*old_ip_mrouter_done)();
2250		static int (*old_ip_mforward)();
2251		static int (*old_mrt_ioctl)();
2252		static void (*old_proto4_input)();
2253		static int (*old_legal_vif_num)();
2254		extern struct protosw inetsw[];
2255
2256	case LKM_E_LOAD:
2257		if(lkmexists(lkmtp) || ip_mrtproto)
2258		  return(EEXIST);
2259		old_ip_mrouter_cmd = ip_mrouter_cmd;
2260		ip_mrouter_cmd = X_ip_mrouter_cmd;
2261		old_ip_mrouter_done = ip_mrouter_done;
2262		ip_mrouter_done = X_ip_mrouter_done;
2263		old_ip_mforward = ip_mforward;
2264		ip_mforward = X_ip_mforward;
2265		old_mrt_ioctl = mrt_ioctl;
2266		mrt_ioctl = X_mrt_ioctl;
2267              old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input;
2268              inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input;
2269		old_legal_vif_num = legal_vif_num;
2270		legal_vif_num = X_legal_vif_num;
2271		ip_mrtproto = IGMP_DVMRP;
2272
2273		printf("\nIP multicast routing loaded\n");
2274		break;
2275
2276	case LKM_E_UNLOAD:
2277		if (ip_mrouter)
2278		  return EINVAL;
2279
2280		ip_mrouter_cmd = old_ip_mrouter_cmd;
2281		ip_mrouter_done = old_ip_mrouter_done;
2282		ip_mforward = old_ip_mforward;
2283		mrt_ioctl = old_mrt_ioctl;
2284              inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input;
2285		legal_vif_num = old_legal_vif_num;
2286		ip_mrtproto = 0;
2287		break;
2288
2289	default:
2290		err = EINVAL;
2291		break;
2292	}
2293
2294	return(err);
2295}
2296
2297int
2298ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) {
2299	DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle,
2300		 nosys);
2301}
2302
2303#endif /* MROUTE_LKM */
2304#endif /* MROUTING */
2305