xref: /netbsd-src/sys/netinet/ip_mroute.c (revision 76dfffe33547c37f8bdd446e3e4ab0f3c16cea4b)
1 /*	$NetBSD: ip_mroute.c,v 1.34 1996/10/13 02:03:06 christos Exp $	*/
2 
3 /*
4  * IP multicast forwarding procedures
5  *
6  * Written by David Waitzman, BBN Labs, August 1988.
7  * Modified by Steve Deering, Stanford, February 1989.
8  * Modified by Mark J. Steiglitz, Stanford, May, 1991
9  * Modified by Van Jacobson, LBL, January 1993
10  * Modified by Ajit Thyagarajan, PARC, August 1993
11  * Modified by Bill Fenner, PARC, April 1994
12  * Modified by Charles M. Hannum, NetBSD, May 1995.
13  *
14  * MROUTING Revision: 1.2
15  */
16 
17 #include <sys/param.h>
18 #include <sys/systm.h>
19 #include <sys/mbuf.h>
20 #include <sys/socket.h>
21 #include <sys/socketvar.h>
22 #include <sys/protosw.h>
23 #include <sys/errno.h>
24 #include <sys/time.h>
25 #include <sys/kernel.h>
26 #include <sys/ioctl.h>
27 #include <sys/syslog.h>
28 #include <net/if.h>
29 #include <net/route.h>
30 #include <net/raw_cb.h>
31 #include <netinet/in.h>
32 #include <netinet/in_var.h>
33 #include <netinet/in_systm.h>
34 #include <netinet/ip.h>
35 #include <netinet/ip_var.h>
36 #include <netinet/in_pcb.h>
37 #include <netinet/udp.h>
38 #include <netinet/igmp.h>
39 #include <netinet/igmp_var.h>
40 #include <netinet/ip_mroute.h>
41 
42 #include <machine/stdarg.h>
43 
44 #define IP_MULTICASTOPTS 0
45 #define	M_PULLUP(m, len) \
46 	do { \
47 		if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
48 			(m) = m_pullup((m), (len)); \
49 	} while (0)
50 
51 /*
52  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
53  * except for netstat or debugging purposes.
54  */
55 struct socket  *ip_mrouter  = 0;
56 int		ip_mrtproto = IGMP_DVMRP;    /* for netstat only */
57 
58 #define NO_RTE_FOUND 	0x1
59 #define RTE_FOUND	0x2
60 
61 #define	MFCHASH(a, g) \
62 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
63 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
64 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
65 u_long	mfchash;
66 
67 u_char		nexpire[MFCTBLSIZ];
68 struct vif	viftable[MAXVIFS];
69 struct mrtstat	mrtstat;
70 u_int		mrtdebug = 0;	  /* debug level 	*/
71 #define		DEBUG_MFC	0x02
72 #define		DEBUG_FORWARD	0x04
73 #define		DEBUG_EXPIRE	0x08
74 #define		DEBUG_XMIT	0x10
75 u_int       	tbfdebug = 0;     /* tbf debug level 	*/
76 #ifdef RSVP_ISI
77 u_int		rsvpdebug = 0;	  /* rsvp debug level   */
78 extern struct socket *ip_rsvpd;
79 extern int rsvp_on;
80 #endif /* RSVP_ISI */
81 
82 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
83 #define		UPCALL_EXPIRE	6		/* number of timeouts */
84 
85 /*
86  * Define the token bucket filter structures
87  */
88 
89 #define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
90 
91 static int get_sg_cnt __P((struct sioc_sg_req *));
92 static int get_vif_cnt __P((struct sioc_vif_req *));
93 static int ip_mrouter_init __P((struct socket *, struct mbuf *));
94 static int get_version __P((struct mbuf *));
95 static int set_assert __P((struct mbuf *));
96 static int get_assert __P((struct mbuf *));
97 static int add_vif __P((struct mbuf *));
98 static int del_vif __P((struct mbuf *));
99 static void update_mfc __P((struct mfcctl *, struct mfc *));
100 static void expire_mfc __P((struct mfc *));
101 static int add_mfc __P((struct mbuf *));
102 #ifdef UPCALL_TIMING
103 static void collate __P((struct timeval *));
104 #endif
105 static int del_mfc __P((struct mbuf *));
106 static int socket_send __P((struct socket *, struct mbuf *,
107 			    struct sockaddr_in *));
108 static void expire_upcalls __P((void *));
109 #ifdef RSVP_ISI
110 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *, vifi_t));
111 #else
112 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *));
113 #endif
114 static void phyint_send __P((struct ip *, struct vif *, struct mbuf *));
115 static void encap_send __P((struct ip *, struct vif *, struct mbuf *));
116 static void tbf_control __P((struct vif *, struct mbuf *, struct ip *,
117 			     u_int32_t));
118 static void tbf_queue __P((struct vif *, struct mbuf *));
119 static void tbf_process_q __P((struct vif *));
120 static void tbf_reprocess_q __P((void *));
121 static int tbf_dq_sel __P((struct vif *, struct ip *));
122 static void tbf_send_packet __P((struct vif *, struct mbuf *));
123 static void tbf_update_tokens __P((struct vif *));
124 static int priority __P((struct vif *, struct ip *));
125 
126 /*
127  * 'Interfaces' associated with decapsulator (so we can tell
128  * packets that went through it from ones that get reflected
129  * by a broken gateway).  These interfaces are never linked into
130  * the system ifnet list & no routes point to them.  I.e., packets
131  * can't be sent this way.  They only exist as a placeholder for
132  * multicast source verification.
133  */
134 #if 0
135 struct ifnet multicast_decap_if[MAXVIFS];
136 #endif
137 
138 #define	ENCAP_TTL	64
139 #define	ENCAP_PROTO	IPPROTO_IPIP	/* 4 */
140 
141 /* prototype IP hdr for encapsulated packets */
142 struct ip multicast_encap_iphdr = {
143 #if BYTE_ORDER == LITTLE_ENDIAN
144 	sizeof(struct ip) >> 2, IPVERSION,
145 #else
146 	IPVERSION, sizeof(struct ip) >> 2,
147 #endif
148 	0,				/* tos */
149 	sizeof(struct ip),		/* total length */
150 	0,				/* id */
151 	0,				/* frag offset */
152 	ENCAP_TTL, ENCAP_PROTO,
153 	0,				/* checksum */
154 };
155 
156 /*
157  * Private variables.
158  */
159 static vifi_t	   numvifs = 0;
160 static int have_encap_tunnel = 0;
161 
162 /*
163  * one-back cache used by ipip_input to locate a tunnel's vif
164  * given a datagram's src ip address.
165  */
166 static struct in_addr last_encap_src;
167 static struct vif *last_encap_vif;
168 
169 /*
170  * whether or not special PIM assert processing is enabled.
171  */
172 static int pim_assert;
173 /*
174  * Rate limit for assert notification messages, in usec
175  */
176 #define ASSERT_MSG_TIME		3000000
177 
178 /*
179  * Find a route for a given origin IP address and Multicast group address
180  * Type of service parameter to be added in the future!!!
181  */
182 
183 #define MFCFIND(o, g, rt) { \
184 	register struct mfc *_rt; \
185 	(rt) = 0; \
186 	++mrtstat.mrts_mfc_lookups; \
187 	for (_rt = mfchashtbl[MFCHASH(o, g)].lh_first; \
188 	     _rt; _rt = _rt->mfc_hash.le_next) { \
189 		if (in_hosteq(_rt->mfc_origin, (o)) && \
190 		    in_hosteq(_rt->mfc_mcastgrp, (g)) && \
191 		    _rt->mfc_stall == 0) { \
192 			(rt) = _rt; \
193 			break; \
194 		} \
195 	} \
196 	if ((rt) == 0) \
197 		++mrtstat.mrts_mfc_misses; \
198 }
199 
200 /*
201  * Macros to compute elapsed time efficiently
202  * Borrowed from Van Jacobson's scheduling code
203  */
204 #define TV_DELTA(a, b, delta) { \
205 	register int xxs; \
206 	delta = (a).tv_usec - (b).tv_usec; \
207 	xxs = (a).tv_sec - (b).tv_sec; \
208 	switch (xxs) { \
209 	case 2: \
210 		delta += 1000000; \
211 		/* fall through */ \
212 	case 1: \
213 		delta += 1000000; \
214 		/* fall through */ \
215 	case 0: \
216 		break; \
217 	default: \
218 		delta += (1000000 * xxs); \
219 		break; \
220 	} \
221 }
222 
223 #ifdef UPCALL_TIMING
224 u_int32_t upcall_data[51];
225 #endif /* UPCALL_TIMING */
226 
227 /*
228  * Handle MRT setsockopt commands to modify the multicast routing tables.
229  */
230 int
231 ip_mrouter_set(so, optname, m)
232 	struct socket *so;
233 	int optname;
234 	struct mbuf **m;
235 {
236 	int error;
237 
238 	if (optname != MRT_INIT && so != ip_mrouter)
239 		error = ENOPROTOOPT;
240 	else
241 		switch (optname) {
242 		case MRT_INIT:
243 			error = ip_mrouter_init(so, *m);
244 			break;
245 		case MRT_DONE:
246 			error = ip_mrouter_done();
247 			break;
248 		case MRT_ADD_VIF:
249 			error = add_vif(*m);
250 			break;
251 		case MRT_DEL_VIF:
252 			error = del_vif(*m);
253 			break;
254 		case MRT_ADD_MFC:
255 			error = add_mfc(*m);
256 			break;
257 		case MRT_DEL_MFC:
258 			error = del_mfc(*m);
259 			break;
260 		case MRT_ASSERT:
261 			error = set_assert(*m);
262 			break;
263 		default:
264 			error = ENOPROTOOPT;
265 			break;
266 		}
267 
268 	if (*m)
269 		m_free(*m);
270 	return (error);
271 }
272 
273 /*
274  * Handle MRT getsockopt commands
275  */
276 int
277 ip_mrouter_get(so, optname, m)
278 	struct socket *so;
279 	int optname;
280 	struct mbuf **m;
281 {
282 	int error;
283 
284 	if (so != ip_mrouter)
285 		error = ENOPROTOOPT;
286 	else {
287 		*m = m_get(M_WAIT, MT_SOOPTS);
288 
289 		switch (optname) {
290 		case MRT_VERSION:
291 			error = get_version(*m);
292 			break;
293 		case MRT_ASSERT:
294 			error = get_assert(*m);
295 			break;
296 		default:
297 			error = ENOPROTOOPT;
298 			break;
299 		}
300 
301 		if (error)
302 			m_free(*m);
303 	}
304 
305 	return (error);
306 }
307 
308 /*
309  * Handle ioctl commands to obtain information from the cache
310  */
311 int
312 mrt_ioctl(so, cmd, data)
313 	struct socket *so;
314 	u_long cmd;
315 	caddr_t data;
316 {
317 	int error;
318 
319 	if (so != ip_mrouter)
320 		error = EINVAL;
321 	else
322 		switch (cmd) {
323 		case SIOCGETVIFCNT:
324 			error = get_vif_cnt((struct sioc_vif_req *)data);
325 			break;
326 		case SIOCGETSGCNT:
327 			error = get_sg_cnt((struct sioc_sg_req *)data);
328 			break;
329 		default:
330 			error = EINVAL;
331 			break;
332 		}
333 
334 	return (error);
335 }
336 
337 /*
338  * returns the packet, byte, rpf-failure count for the source group provided
339  */
340 static int
341 get_sg_cnt(req)
342 	register struct sioc_sg_req *req;
343 {
344 	register struct mfc *rt;
345 	int s;
346 
347 	s = splsoftnet();
348 	MFCFIND(req->src, req->grp, rt);
349 	splx(s);
350 	if (rt != 0) {
351 		req->pktcnt = rt->mfc_pkt_cnt;
352 		req->bytecnt = rt->mfc_byte_cnt;
353 		req->wrong_if = rt->mfc_wrong_if;
354 	} else
355 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
356 
357 	return (0);
358 }
359 
360 /*
361  * returns the input and output packet and byte counts on the vif provided
362  */
363 static int
364 get_vif_cnt(req)
365 	register struct sioc_vif_req *req;
366 {
367 	register vifi_t vifi = req->vifi;
368 
369 	if (vifi >= numvifs)
370 		return (EINVAL);
371 
372 	req->icount = viftable[vifi].v_pkt_in;
373 	req->ocount = viftable[vifi].v_pkt_out;
374 	req->ibytes = viftable[vifi].v_bytes_in;
375 	req->obytes = viftable[vifi].v_bytes_out;
376 
377 	return (0);
378 }
379 
380 /*
381  * Enable multicast routing
382  */
383 static int
384 ip_mrouter_init(so, m)
385 	struct socket *so;
386 	struct mbuf *m;
387 {
388 	int *v;
389 
390 	if (mrtdebug)
391 		log(LOG_DEBUG,
392 		    "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
393 		    so->so_type, so->so_proto->pr_protocol);
394 
395 	if (so->so_type != SOCK_RAW ||
396 	    so->so_proto->pr_protocol != IPPROTO_IGMP)
397 		return (EOPNOTSUPP);
398 
399 	if (m == 0 || m->m_len < sizeof(int))
400 		return (EINVAL);
401 
402 	v = mtod(m, int *);
403 	if (*v != 1)
404 		return (EINVAL);
405 
406 	if (ip_mrouter != 0)
407 		return (EADDRINUSE);
408 
409 	ip_mrouter = so;
410 
411 	mfchashtbl = hashinit(MFCTBLSIZ, M_MRTABLE, &mfchash);
412 	bzero((caddr_t)nexpire, sizeof(nexpire));
413 
414 	pim_assert = 0;
415 
416 	timeout(expire_upcalls, (caddr_t)0, EXPIRE_TIMEOUT);
417 
418 	if (mrtdebug)
419 		log(LOG_DEBUG, "ip_mrouter_init\n");
420 
421 	return (0);
422 }
423 
424 /*
425  * Disable multicast routing
426  */
427 int
428 ip_mrouter_done()
429 {
430 	vifi_t vifi;
431 	register struct vif *vifp;
432 	int i;
433 	int s;
434 
435 	s = splsoftnet();
436 
437 	/* Clear out all the vifs currently in use. */
438 	for (vifi = 0; vifi < numvifs; vifi++) {
439 		vifp = &viftable[vifi];
440 		if (!in_nullhost(vifp->v_lcl_addr))
441 			reset_vif(vifp);
442 	}
443 
444 	numvifs = 0;
445 	pim_assert = 0;
446 
447 	untimeout(expire_upcalls, (caddr_t)0);
448 
449 	/*
450 	 * Free all multicast forwarding cache entries.
451 	 */
452 	for (i = 0; i < MFCTBLSIZ; i++) {
453 		register struct mfc *rt, *nrt;
454 
455 		for (rt = mfchashtbl[i].lh_first; rt; rt = nrt) {
456 			nrt = rt->mfc_hash.le_next;
457 
458 			expire_mfc(rt);
459 		}
460 	}
461 	free(mfchashtbl, M_MRTABLE);
462 
463 	/* Reset de-encapsulation cache. */
464 	have_encap_tunnel = 0;
465 
466 	ip_mrouter = 0;
467 
468 	splx(s);
469 
470 	if (mrtdebug)
471 		log(LOG_DEBUG, "ip_mrouter_done\n");
472 
473 	return (0);
474 }
475 
476 static int
477 get_version(m)
478 	struct mbuf *m;
479 {
480 	int *v = mtod(m, int *);
481 
482 	*v = 0x0305;	/* XXX !!!! */
483 	m->m_len = sizeof(int);
484 	return (0);
485 }
486 
487 /*
488  * Set PIM assert processing global
489  */
490 static int
491 set_assert(m)
492 	struct mbuf *m;
493 {
494 	int *i;
495 
496 	if (m == 0 || m->m_len < sizeof(int))
497 		return (EINVAL);
498 
499 	i = mtod(m, int *);
500 	pim_assert = !!*i;
501 	return (0);
502 }
503 
504 /*
505  * Get PIM assert processing global
506  */
507 static int
508 get_assert(m)
509 	struct mbuf *m;
510 {
511 	int *i = mtod(m, int *);
512 
513 	*i = pim_assert;
514 	m->m_len = sizeof(int);
515 	return (0);
516 }
517 
518 static struct sockaddr_in sin = { sizeof(sin), AF_INET };
519 
520 /*
521  * Add a vif to the vif table
522  */
523 static int
524 add_vif(m)
525 	struct mbuf *m;
526 {
527 	register struct vifctl *vifcp;
528 	register struct vif *vifp;
529 	struct ifaddr *ifa;
530 	struct ifnet *ifp;
531 	struct ifreq ifr;
532 	int error, s;
533 
534 	if (m == 0 || m->m_len < sizeof(struct vifctl))
535 		return (EINVAL);
536 
537 	vifcp = mtod(m, struct vifctl *);
538 	if (vifcp->vifc_vifi >= MAXVIFS)
539 		return (EINVAL);
540 
541 	vifp = &viftable[vifcp->vifc_vifi];
542 	if (!in_nullhost(vifp->v_lcl_addr))
543 		return (EADDRINUSE);
544 
545 	/* Find the interface with an address in AF_INET family. */
546 	sin.sin_addr = vifcp->vifc_lcl_addr;
547 	ifa = ifa_ifwithaddr(sintosa(&sin));
548 	if (ifa == 0)
549 		return (EADDRNOTAVAIL);
550 
551 	if (vifcp->vifc_flags & VIFF_TUNNEL) {
552 		if (vifcp->vifc_flags & VIFF_SRCRT) {
553 			log(LOG_ERR, "Source routed tunnels not supported\n");
554 			return (EOPNOTSUPP);
555 		}
556 
557 		/* Create a fake encapsulation interface. */
558 		ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK);
559 		bzero(ifp, sizeof(*ifp));
560 		sprintf(ifp->if_xname, "mdecap%d", vifcp->vifc_vifi);
561 
562 		/* Prepare cached route entry. */
563 		bzero(&vifp->v_route, sizeof(vifp->v_route));
564 
565 		/* Tell ipip_input() to start looking at encapsulated packets. */
566 		have_encap_tunnel = 1;
567 	} else {
568 		/* Use the physical interface associated with the address. */
569 		ifp = ifa->ifa_ifp;
570 
571 		/* Make sure the interface supports multicast. */
572 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
573 			return (EOPNOTSUPP);
574 
575 		/* Enable promiscuous reception of all IP multicasts. */
576 		satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
577 		satosin(&ifr.ifr_addr)->sin_family = AF_INET;
578 		satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
579 		error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
580 		if (error)
581 			return (error);
582 	}
583 
584 	s = splsoftnet();
585 
586 	/* Define parameters for the tbf structure. */
587 	vifp->tbf_q = 0;
588 	vifp->tbf_t = &vifp->tbf_q;
589 	microtime(&vifp->tbf_last_pkt_t);
590 	vifp->tbf_n_tok = 0;
591 	vifp->tbf_q_len = 0;
592 	vifp->tbf_max_q_len = MAXQSIZE;
593 
594 	vifp->v_flags = vifcp->vifc_flags;
595 	vifp->v_threshold = vifcp->vifc_threshold;
596 	/* scaling up here allows division by 1024 in critical code */
597 	vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
598 	vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
599 	vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
600 	vifp->v_ifp = ifp;
601 	/* Initialize per vif pkt counters. */
602 	vifp->v_pkt_in = 0;
603 	vifp->v_pkt_out = 0;
604 	vifp->v_bytes_in = 0;
605 	vifp->v_bytes_out = 0;
606 #ifdef RSVP_ISI
607 	vifp->v_rsvp_on = 0;
608 	vifp->v_rsvpd = 0;
609 #endif /* RSVP_ISI */
610 
611 	splx(s);
612 
613 	/* Adjust numvifs up if the vifi is higher than numvifs. */
614 	if (numvifs <= vifcp->vifc_vifi)
615 		numvifs = vifcp->vifc_vifi + 1;
616 
617 	if (mrtdebug)
618 		log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
619 		    vifcp->vifc_vifi,
620 		    ntohl(vifcp->vifc_lcl_addr.s_addr),
621 		    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
622 		    ntohl(vifcp->vifc_rmt_addr.s_addr),
623 		    vifcp->vifc_threshold,
624 		    vifcp->vifc_rate_limit);
625 
626 	return (0);
627 }
628 
629 void
630 reset_vif(vifp)
631 	register struct vif *vifp;
632 {
633 	register struct mbuf *m, *n;
634 	struct ifnet *ifp;
635 	struct ifreq ifr;
636 
637 	for (m = vifp->tbf_q; m != 0; m = n) {
638 		n = m->m_nextpkt;
639 		m_freem(m);
640 	}
641 
642 	if (vifp->v_flags & VIFF_TUNNEL) {
643 		free(vifp->v_ifp, M_MRTABLE);
644 		if (vifp == last_encap_vif) {
645 			last_encap_vif = 0;
646 			last_encap_src = zeroin_addr;
647 		}
648 	} else {
649 		satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
650 		satosin(&ifr.ifr_addr)->sin_family = AF_INET;
651 		satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
652 		ifp = vifp->v_ifp;
653 		(*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
654 	}
655 	bzero((caddr_t)vifp, sizeof(*vifp));
656 }
657 
658 /*
659  * Delete a vif from the vif table
660  */
661 static int
662 del_vif(m)
663 	struct mbuf *m;
664 {
665 	vifi_t *vifip;
666 	register struct vif *vifp;
667 	register vifi_t vifi;
668 	int s;
669 
670 	if (m == 0 || m->m_len < sizeof(vifi_t))
671 		return (EINVAL);
672 
673 	vifip = mtod(m, vifi_t *);
674 	if (*vifip >= numvifs)
675 		return (EINVAL);
676 
677 	vifp = &viftable[*vifip];
678 	if (in_nullhost(vifp->v_lcl_addr))
679 		return (EADDRNOTAVAIL);
680 
681 	s = splsoftnet();
682 
683 	reset_vif(vifp);
684 
685 	/* Adjust numvifs down */
686 	for (vifi = numvifs; vifi > 0; vifi--)
687 		if (!in_nullhost(viftable[vifi-1].v_lcl_addr))
688 			break;
689 	numvifs = vifi;
690 
691 	splx(s);
692 
693 	if (mrtdebug)
694 		log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
695 
696 	return (0);
697 }
698 
699 static void
700 update_mfc(mfccp, rt)
701 	struct mfcctl *mfccp;
702 	struct mfc *rt;
703 {
704 	vifi_t vifi;
705 
706 	rt->mfc_parent = mfccp->mfcc_parent;
707 	for (vifi = 0; vifi < numvifs; vifi++)
708 		rt->mfc_ttls[vifi] = mfccp->mfcc_ttls[vifi];
709 	rt->mfc_expire = 0;
710 	rt->mfc_stall = 0;
711 }
712 
713 static void
714 expire_mfc(rt)
715 	struct mfc *rt;
716 {
717 	struct rtdetq *rte, *nrte;
718 
719 	for (rte = rt->mfc_stall; rte != 0; rte = nrte) {
720 		nrte = rte->next;
721 		m_freem(rte->m);
722 		free(rte, M_MRTABLE);
723 	}
724 
725 	LIST_REMOVE(rt, mfc_hash);
726 	free(rt, M_MRTABLE);
727 }
728 
729 /*
730  * Add an mfc entry
731  */
732 static int
733 add_mfc(m)
734 	struct mbuf *m;
735 {
736 	struct mfcctl *mfccp;
737 	struct mfc *rt;
738 	u_int32_t hash = 0;
739 	struct rtdetq *rte, *nrte;
740 	register u_short nstl;
741 	int s;
742 
743 	if (m == 0 || m->m_len < sizeof(struct mfcctl))
744 		return (EINVAL);
745 
746 	mfccp = mtod(m, struct mfcctl *);
747 
748 	s = splsoftnet();
749 	MFCFIND(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp, rt);
750 
751 	/* If an entry already exists, just update the fields */
752 	if (rt) {
753 		if (mrtdebug & DEBUG_MFC)
754 			log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n",
755 			    ntohl(mfccp->mfcc_origin.s_addr),
756 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
757 			    mfccp->mfcc_parent);
758 
759 		if (rt->mfc_expire)
760 			nexpire[hash]--;
761 
762 		update_mfc(mfccp, rt);
763 
764 		splx(s);
765 		return (0);
766 	}
767 
768 	/*
769 	 * Find the entry for which the upcall was made and update
770 	 */
771 	nstl = 0;
772 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
773 	for (rt = mfchashtbl[hash].lh_first; rt; rt = rt->mfc_hash.le_next) {
774 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
775 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
776 		    rt->mfc_stall != 0) {
777 			if (nstl++)
778 				log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
779 				    "multiple kernel entries",
780 				    ntohl(mfccp->mfcc_origin.s_addr),
781 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
782 				    mfccp->mfcc_parent, rt->mfc_stall);
783 
784 			if (mrtdebug & DEBUG_MFC)
785 				log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %p\n",
786 				    ntohl(mfccp->mfcc_origin.s_addr),
787 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
788 				    mfccp->mfcc_parent, rt->mfc_stall);
789 
790 			if (rt->mfc_expire)
791 				nexpire[hash]--;
792 
793 			/* free packets Qed at the end of this entry */
794 			for (rte = rt->mfc_stall; rte != 0; rte = nrte) {
795 				nrte = rte->next;
796 #ifdef RSVP_ISI
797 				ip_mdq(rte->m, rte->ifp, rt, -1);
798 #else
799 				ip_mdq(rte->m, rte->ifp, rt);
800 #endif /* RSVP_ISI */
801 				m_freem(rte->m);
802 #ifdef UPCALL_TIMING
803 				collate(&rte->t);
804 #endif /* UPCALL_TIMING */
805 				free(rte, M_MRTABLE);
806 			}
807 
808 			update_mfc(mfccp, rt);
809 		}
810 	}
811 
812 	if (nstl == 0) {
813 		/*
814 		 * No mfc; make a new one
815 		 */
816 		if (mrtdebug & DEBUG_MFC)
817 			log(LOG_DEBUG,"add_mfc no upcall o %x g %x p %x\n",
818 			    ntohl(mfccp->mfcc_origin.s_addr),
819 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
820 			    mfccp->mfcc_parent);
821 
822 		rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
823 		if (rt == 0) {
824 			splx(s);
825 			return (ENOBUFS);
826 		}
827 
828 		rt->mfc_origin = mfccp->mfcc_origin;
829 		rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
830 		/* initialize pkt counters per src-grp */
831 		rt->mfc_pkt_cnt = 0;
832 		rt->mfc_byte_cnt = 0;
833 		rt->mfc_wrong_if = 0;
834 		timerclear(&rt->mfc_last_assert);
835 		update_mfc(mfccp, rt);
836 
837 		/* insert new entry at head of hash chain */
838 		LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
839 	}
840 
841 	splx(s);
842 	return (0);
843 }
844 
845 #ifdef UPCALL_TIMING
846 /*
847  * collect delay statistics on the upcalls
848  */
849 static void collate(t)
850 register struct timeval *t;
851 {
852     register u_int32_t d;
853     register struct timeval tp;
854     register u_int32_t delta;
855 
856     microtime(&tp);
857 
858     if (timercmp(t, &tp, <)) {
859 	TV_DELTA(tp, *t, delta);
860 
861 	d = delta >> 10;
862 	if (d > 50)
863 	    d = 50;
864 
865 	++upcall_data[d];
866     }
867 }
868 #endif /* UPCALL_TIMING */
869 
870 /*
871  * Delete an mfc entry
872  */
873 static int
874 del_mfc(m)
875 	struct mbuf *m;
876 {
877 	struct mfcctl *mfccp;
878 	struct mfc *rt;
879 	int s;
880 
881 	if (m == 0 || m->m_len < sizeof(struct mfcctl))
882 		return (EINVAL);
883 
884 	mfccp = mtod(m, struct mfcctl *);
885 
886 	if (mrtdebug & DEBUG_MFC)
887 		log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
888 		    ntohl(mfccp->mfcc_origin.s_addr),
889 		    ntohl(mfccp->mfcc_mcastgrp.s_addr));
890 
891 	s = splsoftnet();
892 
893 	MFCFIND(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp, rt);
894 	if (rt == 0) {
895 		splx(s);
896 		return (EADDRNOTAVAIL);
897 	}
898 
899 	LIST_REMOVE(rt, mfc_hash);
900 	free(rt, M_MRTABLE);
901 
902 	splx(s);
903 	return (0);
904 }
905 
906 static int
907 socket_send(s, mm, src)
908     struct socket *s;
909     struct mbuf *mm;
910     struct sockaddr_in *src;
911 {
912     if (s) {
913 	if (sbappendaddr(&s->so_rcv, sintosa(src), mm, (struct mbuf *)0) != 0) {
914 	    sorwakeup(s);
915 	    return (0);
916 	}
917     }
918     m_freem(mm);
919     return (-1);
920 }
921 
922 /*
923  * IP multicast forwarding function. This function assumes that the packet
924  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
925  * pointed to by "ifp", and the packet is to be relayed to other networks
926  * that have members of the packet's destination IP multicast group.
927  *
928  * The packet is returned unscathed to the caller, unless it is
929  * erroneous, in which case a non-zero return value tells the caller to
930  * discard it.
931  */
932 
933 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
934 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
935 
936 int
937 #ifdef RSVP_ISI
938 ip_mforward(m, ifp, imo)
939 #else
940 ip_mforward(m, ifp)
941 #endif /* RSVP_ISI */
942     struct mbuf *m;
943     struct ifnet *ifp;
944 #ifdef RSVP_ISI
945     struct ip_moptions *imo;
946 #endif /* RSVP_ISI */
947 {
948     register struct ip *ip = mtod(m, struct ip *);
949     register struct mfc *rt;
950     register u_char *ipoptions;
951     static int srctun = 0;
952     register struct mbuf *mm;
953     int s;
954 #ifdef RSVP_ISI
955     register struct vif *vifp;
956     vifi_t vifi;
957 #endif /* RSVP_ISI */
958 
959     if (mrtdebug & DEBUG_FORWARD)
960 	log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
961 	    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
962 
963     if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
964 	(ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR) {
965 	/*
966 	 * Packet arrived via a physical interface or
967 	 * an encapuslated tunnel.
968 	 */
969     } else {
970 	/*
971 	 * Packet arrived through a source-route tunnel.
972 	 * Source-route tunnels are no longer supported.
973 	 */
974 	if ((srctun++ % 1000) == 0)
975 	    log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n",
976 		ntohl(ip->ip_src.s_addr));
977 
978 	return (1);
979     }
980 
981 #ifdef RSVP_ISI
982     if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
983 	if (ip->ip_ttl < 255)
984 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
985 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
986 	    vifp = viftable + vifi;
987 	    printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n",
988 		ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
989 		(vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
990 		vifp->v_ifp->if_xname);
991 	}
992 	return (ip_mdq(m, ifp, (struct mfc *)0, vifi));
993     }
994     if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
995 	printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
996 	    ntohl(ip->ip_src), ntohl(ip->ip_dst));
997     }
998 #endif /* RSVP_ISI */
999 
1000     /*
1001      * Don't forward a packet with time-to-live of zero or one,
1002      * or a packet destined to a local-only group.
1003      */
1004     if (ip->ip_ttl <= 1 ||
1005 	IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1006 	return (0);
1007 
1008     /*
1009      * Determine forwarding vifs from the forwarding cache table
1010      */
1011     s = splsoftnet();
1012     MFCFIND(ip->ip_src, ip->ip_dst, rt);
1013 
1014     /* Entry exists, so forward if necessary */
1015     if (rt != 0) {
1016 	splx(s);
1017 #ifdef RSVP_ISI
1018 	return (ip_mdq(m, ifp, rt, -1));
1019 #else
1020 	return (ip_mdq(m, ifp, rt));
1021 #endif /* RSVP_ISI */
1022     } else {
1023 	/*
1024 	 * If we don't have a route for packet's origin,
1025 	 * Make a copy of the packet &
1026 	 * send message to routing daemon
1027 	 */
1028 
1029 	register struct mbuf *mb0;
1030 	register struct rtdetq *rte;
1031 	register u_int32_t hash;
1032 	int hlen = ip->ip_hl << 2;
1033 #ifdef UPCALL_TIMING
1034 	struct timeval tp;
1035 
1036 	microtime(&tp);
1037 #endif /* UPCALL_TIMING */
1038 
1039 	mrtstat.mrts_no_route++;
1040 	if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1041 	    log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1042 		ntohl(ip->ip_src.s_addr),
1043 		ntohl(ip->ip_dst.s_addr));
1044 
1045 	/*
1046 	 * Allocate mbufs early so that we don't do extra work if we are
1047 	 * just going to fail anyway.  Make sure to pullup the header so
1048 	 * that other people can't step on it.
1049 	 */
1050 	rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, M_NOWAIT);
1051 	if (rte == 0) {
1052 	    splx(s);
1053 	    return (ENOBUFS);
1054 	}
1055 	mb0 = m_copy(m, 0, M_COPYALL);
1056 	M_PULLUP(mb0, hlen);
1057 	if (mb0 == 0) {
1058 	    free(rte, M_MRTABLE);
1059 	    splx(s);
1060 	    return (ENOBUFS);
1061 	}
1062 
1063 	/* is there an upcall waiting for this packet? */
1064 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1065 	for (rt = mfchashtbl[hash].lh_first; rt; rt = rt->mfc_hash.le_next) {
1066 	    if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1067 		in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1068 		rt->mfc_stall != 0)
1069 		break;
1070 	}
1071 
1072 	if (rt == 0) {
1073 	    int i;
1074 	    struct igmpmsg *im;
1075 
1076 	    /* no upcall, so make a new entry */
1077 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1078 	    if (rt == 0) {
1079 		free(rte, M_MRTABLE);
1080 		m_freem(mb0);
1081 		splx(s);
1082 		return (ENOBUFS);
1083 	    }
1084 	    /* Make a copy of the header to send to the user level process */
1085 	    mm = m_copy(m, 0, hlen);
1086 	    M_PULLUP(mm, hlen);
1087 	    if (mm == 0) {
1088 		free(rte, M_MRTABLE);
1089 		m_freem(mb0);
1090 		free(rt, M_MRTABLE);
1091 		splx(s);
1092 		return (ENOBUFS);
1093 	    }
1094 
1095 	    /*
1096 	     * Send message to routing daemon to install
1097 	     * a route into the kernel table
1098 	     */
1099 	    sin.sin_addr = ip->ip_src;
1100 
1101 	    im = mtod(mm, struct igmpmsg *);
1102 	    im->im_msgtype	= IGMPMSG_NOCACHE;
1103 	    im->im_mbz		= 0;
1104 
1105 	    mrtstat.mrts_upcalls++;
1106 
1107 	    if (socket_send(ip_mrouter, mm, &sin) < 0) {
1108 		log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1109 		++mrtstat.mrts_upq_sockfull;
1110 		free(rte, M_MRTABLE);
1111 		m_freem(mb0);
1112 		free(rt, M_MRTABLE);
1113 		splx(s);
1114 		return (ENOBUFS);
1115 	    }
1116 
1117 	    /* insert new entry at head of hash chain */
1118 	    rt->mfc_origin = ip->ip_src;
1119 	    rt->mfc_mcastgrp = ip->ip_dst;
1120 	    rt->mfc_pkt_cnt = 0;
1121 	    rt->mfc_byte_cnt = 0;
1122 	    rt->mfc_wrong_if = 0;
1123 	    rt->mfc_expire = UPCALL_EXPIRE;
1124 	    nexpire[hash]++;
1125 	    for (i = 0; i < numvifs; i++)
1126 		rt->mfc_ttls[i] = 0;
1127 	    rt->mfc_parent = -1;
1128 
1129 	    /* link into table */
1130 	    LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1131 	    /* Add this entry to the end of the queue */
1132 	    rt->mfc_stall = rte;
1133 	} else {
1134 	    /* determine if q has overflowed */
1135 	    struct rtdetq **p;
1136 	    register int npkts = 0;
1137 
1138 	    for (p = &rt->mfc_stall; *p != 0; p = &(*p)->next)
1139 		if (++npkts > MAX_UPQ) {
1140 		    mrtstat.mrts_upq_ovflw++;
1141 		    free(rte, M_MRTABLE);
1142 		    m_freem(mb0);
1143 		    splx(s);
1144 		    return (0);
1145 	        }
1146 
1147 	    /* Add this entry to the end of the queue */
1148 	    *p = rte;
1149 	}
1150 
1151 	rte->next		= 0;
1152 	rte->m 			= mb0;
1153 	rte->ifp 		= ifp;
1154 #ifdef UPCALL_TIMING
1155 	rte->t			= tp;
1156 #endif /* UPCALL_TIMING */
1157 
1158 
1159 	splx(s);
1160 
1161 	return (0);
1162     }
1163 }
1164 
1165 
1166 /*ARGSUSED*/
1167 static void
1168 expire_upcalls(v)
1169 	void *v;
1170 {
1171 	int i;
1172 	int s;
1173 
1174 	s = splsoftnet();
1175 
1176 	for (i = 0; i < MFCTBLSIZ; i++) {
1177 		register struct mfc *rt, *nrt;
1178 
1179 		if (nexpire[i] == 0)
1180 			continue;
1181 
1182 		for (rt = mfchashtbl[i].lh_first; rt; rt = nrt) {
1183 			nrt = rt->mfc_hash.le_next;
1184 
1185 			if (rt->mfc_expire == 0 ||
1186 			    --rt->mfc_expire > 0)
1187 				continue;
1188 			nexpire[i]--;
1189 
1190 			++mrtstat.mrts_cache_cleanups;
1191 			if (mrtdebug & DEBUG_EXPIRE)
1192 				log(LOG_DEBUG,
1193 				    "expire_upcalls: expiring (%x %x)\n",
1194 				    ntohl(rt->mfc_origin.s_addr),
1195 				    ntohl(rt->mfc_mcastgrp.s_addr));
1196 
1197 			expire_mfc(rt);
1198 		}
1199 	}
1200 
1201 	splx(s);
1202 	timeout(expire_upcalls, (caddr_t)0, EXPIRE_TIMEOUT);
1203 }
1204 
1205 /*
1206  * Packet forwarding routine once entry in the cache is made
1207  */
1208 static int
1209 #ifdef RSVP_ISI
1210 ip_mdq(m, ifp, rt, xmt_vif)
1211 #else
1212 ip_mdq(m, ifp, rt)
1213 #endif /* RSVP_ISI */
1214     register struct mbuf *m;
1215     register struct ifnet *ifp;
1216     register struct mfc *rt;
1217 #ifdef RSVP_ISI
1218     register vifi_t xmt_vif;
1219 #endif /* RSVP_ISI */
1220 {
1221     register struct ip  *ip = mtod(m, struct ip *);
1222     register vifi_t vifi;
1223     register struct vif *vifp;
1224     register int plen = ntohs(ip->ip_len);
1225 
1226 /*
1227  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1228  * input, they shouldn't get counted on output, so statistics keeping is
1229  * seperate.
1230  */
1231 #define MC_SEND(ip,vifp,m) {                             \
1232                 if ((vifp)->v_flags & VIFF_TUNNEL)	 \
1233                     encap_send((ip), (vifp), (m));       \
1234                 else                                     \
1235                     phyint_send((ip), (vifp), (m));      \
1236 }
1237 
1238 #ifdef RSVP_ISI
1239     /*
1240      * If xmt_vif is not -1, send on only the requested vif.
1241      *
1242      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1243      */
1244     if (xmt_vif < numvifs) {
1245         MC_SEND(ip, viftable + xmt_vif, m);
1246 	return (1);
1247     }
1248 #endif /* RSVP_ISI */
1249 
1250     /*
1251      * Don't forward if it didn't arrive from the parent vif for its origin.
1252      */
1253     vifi = rt->mfc_parent;
1254     if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1255 	/* came in the wrong interface */
1256 	if (mrtdebug & DEBUG_FORWARD)
1257 	    log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1258 		ifp, vifi, viftable[vifi].v_ifp);
1259 	++mrtstat.mrts_wrong_if;
1260 	++rt->mfc_wrong_if;
1261 	/*
1262 	 * If we are doing PIM assert processing, and we are forwarding
1263 	 * packets on this interface, and it is a broadcast medium
1264 	 * interface (and not a tunnel), send a message to the routing daemon.
1265 	 */
1266 	if (pim_assert && rt->mfc_ttls[vifi] &&
1267 		(ifp->if_flags & IFF_BROADCAST) &&
1268 		!(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1269 	    struct mbuf *mm;
1270 	    struct igmpmsg *im;
1271 	    int hlen = ip->ip_hl << 2;
1272 	    struct timeval now;
1273 	    register u_int32_t delta;
1274 
1275 	    microtime(&now);
1276 
1277 	    TV_DELTA(rt->mfc_last_assert, now, delta);
1278 
1279 	    if (delta > ASSERT_MSG_TIME) {
1280 		mm = m_copy(m, 0, hlen);
1281 		M_PULLUP(mm, hlen);
1282 		if (mm == 0) {
1283 		    return (ENOBUFS);
1284 		}
1285 
1286 		rt->mfc_last_assert = now;
1287 
1288 		im = mtod(mm, struct igmpmsg *);
1289 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1290 		im->im_mbz	= 0;
1291 		im->im_vif	= vifi;
1292 
1293 		sin.sin_addr = im->im_src;
1294 
1295 		socket_send(ip_mrouter, mm, &sin);
1296 	    }
1297 	}
1298 	return (0);
1299     }
1300 
1301     /* If I sourced this packet, it counts as output, else it was input. */
1302     if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1303 	viftable[vifi].v_pkt_out++;
1304 	viftable[vifi].v_bytes_out += plen;
1305     } else {
1306 	viftable[vifi].v_pkt_in++;
1307 	viftable[vifi].v_bytes_in += plen;
1308     }
1309     rt->mfc_pkt_cnt++;
1310     rt->mfc_byte_cnt += plen;
1311 
1312     /*
1313      * For each vif, decide if a copy of the packet should be forwarded.
1314      * Forward if:
1315      *		- the ttl exceeds the vif's threshold
1316      *		- there are group members downstream on interface
1317      */
1318     for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1319 	if ((rt->mfc_ttls[vifi] > 0) &&
1320 	    (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1321 	    vifp->v_pkt_out++;
1322 	    vifp->v_bytes_out += plen;
1323 	    MC_SEND(ip, vifp, m);
1324 	}
1325 
1326     return (0);
1327 }
1328 
1329 #ifdef RSVP_ISI
1330 /*
1331  * check if a vif number is legal/ok. This is used by ip_output, to export
1332  * numvifs there,
1333  */
1334 int
1335 legal_vif_num(vif)
1336     int vif;
1337 {
1338     if (vif >= 0 && vif < numvifs)
1339        return (1);
1340     else
1341        return (0);
1342 }
1343 #endif /* RSVP_ISI */
1344 
1345 static void
1346 phyint_send(ip, vifp, m)
1347 	struct ip *ip;
1348 	struct vif *vifp;
1349 	struct mbuf *m;
1350 {
1351 	register struct mbuf *mb_copy;
1352 	register int hlen = ip->ip_hl << 2;
1353 
1354 	/*
1355 	 * Make a new reference to the packet; make sure that
1356 	 * the IP header is actually copied, not just referenced,
1357 	 * so that ip_output() only scribbles on the copy.
1358 	 */
1359 	mb_copy = m_copy(m, 0, M_COPYALL);
1360 	M_PULLUP(mb_copy, hlen);
1361 	if (mb_copy == 0)
1362 		return;
1363 
1364 	if (vifp->v_rate_limit <= 0)
1365 		tbf_send_packet(vifp, mb_copy);
1366 	else
1367 		tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1368 }
1369 
1370 static void
1371 encap_send(ip, vifp, m)
1372 	register struct ip *ip;
1373 	register struct vif *vifp;
1374 	register struct mbuf *m;
1375 {
1376 	register struct mbuf *mb_copy;
1377 	register struct ip *ip_copy;
1378 	register int i, len = ip->ip_len + sizeof(multicast_encap_iphdr);
1379 
1380 	/*
1381 	 * copy the old packet & pullup it's IP header into the
1382 	 * new mbuf so we can modify it.  Try to fill the new
1383 	 * mbuf since if we don't the ethernet driver will.
1384 	 */
1385 	MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1386 	if (mb_copy == 0)
1387 		return;
1388 	mb_copy->m_data += max_linkhdr;
1389 	mb_copy->m_pkthdr.len = len;
1390 	mb_copy->m_len = sizeof(multicast_encap_iphdr);
1391 
1392 	if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == 0) {
1393 		m_freem(mb_copy);
1394 		return;
1395 	}
1396 	i = MHLEN - max_linkhdr;
1397 	if (i > len)
1398 		i = len;
1399 	mb_copy = m_pullup(mb_copy, i);
1400 	if (mb_copy == 0)
1401 		return;
1402 
1403 	/*
1404 	 * fill in the encapsulating IP header.
1405 	 */
1406 	ip_copy = mtod(mb_copy, struct ip *);
1407 	*ip_copy = multicast_encap_iphdr;
1408 	ip_copy->ip_id = htons(ip_id++);
1409 	ip_copy->ip_len = len;
1410 	ip_copy->ip_src = vifp->v_lcl_addr;
1411 	ip_copy->ip_dst = vifp->v_rmt_addr;
1412 
1413 	/*
1414 	 * turn the encapsulated IP header back into a valid one.
1415 	 */
1416 	ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1417 	--ip->ip_ttl;
1418 	HTONS(ip->ip_len);
1419 	HTONS(ip->ip_off);
1420 	ip->ip_sum = 0;
1421 #if defined(LBL) && !defined(ultrix) && !defined(i386)
1422 	ip->ip_sum = ~oc_cksum((caddr_t)ip, ip->ip_hl << 2, 0);
1423 #else
1424 	mb_copy->m_data += sizeof(multicast_encap_iphdr);
1425 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1426 	mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1427 #endif
1428 
1429 	if (vifp->v_rate_limit <= 0)
1430 		tbf_send_packet(vifp, mb_copy);
1431 	else
1432 		tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1433 }
1434 
1435 /*
1436  * De-encapsulate a packet and feed it back through ip input (this
1437  * routine is called whenever IP gets a packet with proto type
1438  * ENCAP_PROTO and a local destination address).
1439  */
1440 void
1441 #if __STDC__
1442 ipip_input(struct mbuf *m, ...)
1443 #else
1444 ipip_input(m, va_alist)
1445 	struct mbuf *m;
1446 	va_dcl
1447 #endif
1448 {
1449 	register int hlen;
1450 	register struct ip *ip = mtod(m, struct ip *);
1451 	register int s;
1452 	register struct ifqueue *ifq;
1453 	register struct vif *vifp;
1454 	va_list ap;
1455 
1456 	va_start(ap, m);
1457 	hlen = va_arg(ap, int);
1458 	va_end(ap);
1459 
1460 	if (!have_encap_tunnel) {
1461 		rip_input(m);
1462 		return;
1463 	}
1464 
1465 	/*
1466 	 * dump the packet if it's not to a multicast destination or if
1467 	 * we don't have an encapsulating tunnel with the source.
1468 	 * Note:  This code assumes that the remote site IP address
1469 	 * uniquely identifies the tunnel (i.e., that this site has
1470 	 * at most one tunnel with the remote site).
1471 	 */
1472 	if (!IN_MULTICAST(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr)) {
1473 		++mrtstat.mrts_bad_tunnel;
1474 		m_freem(m);
1475 		return;
1476 	}
1477 
1478 	if (!in_hosteq(ip->ip_src, last_encap_src)) {
1479 		register struct vif *vife;
1480 
1481 		vifp = viftable;
1482 		vife = vifp + numvifs;
1483 		for (; vifp < vife; vifp++)
1484 			if (vifp->v_flags & VIFF_TUNNEL &&
1485 			    in_hosteq(vifp->v_rmt_addr, ip->ip_src))
1486 				break;
1487 		if (vifp == vife) {
1488 			mrtstat.mrts_cant_tunnel++; /*XXX*/
1489 			m_freem(m);
1490 			if (mrtdebug)
1491 				log(LOG_DEBUG, "ip_mforward: no tunnel with %x\n",
1492 				    ntohl(ip->ip_src.s_addr));
1493 			return;
1494 		}
1495 		last_encap_vif = vifp;
1496 		last_encap_src = ip->ip_src;
1497 	} else
1498 		vifp = last_encap_vif;
1499 
1500 	m->m_data += hlen;
1501 	m->m_len -= hlen;
1502 	m->m_pkthdr.len -= hlen;
1503 	m->m_pkthdr.rcvif = vifp->v_ifp;
1504 	ifq = &ipintrq;
1505 	s = splimp();
1506 	if (IF_QFULL(ifq)) {
1507 		IF_DROP(ifq);
1508 		m_freem(m);
1509 	} else {
1510 		IF_ENQUEUE(ifq, m);
1511 		/*
1512 		 * normally we would need a "schednetisr(NETISR_IP)"
1513 		 * here but we were called by ip_input and it is going
1514 		 * to loop back & try to dequeue the packet we just
1515 		 * queued as soon as we return so we avoid the
1516 		 * unnecessary software interrrupt.
1517 		 */
1518 	}
1519 	splx(s);
1520 }
1521 
1522 /*
1523  * Token bucket filter module
1524  */
1525 static void
1526 tbf_control(vifp, m, ip, len)
1527 	register struct vif *vifp;
1528 	register struct mbuf *m;
1529 	register struct ip *ip;
1530 	register u_int32_t len;
1531 {
1532 
1533 	if (len > MAX_BKT_SIZE) {
1534 		/* drop if packet is too large */
1535 		mrtstat.mrts_pkt2large++;
1536 		m_freem(m);
1537 		return;
1538 	}
1539 
1540 	tbf_update_tokens(vifp);
1541 
1542 	/*
1543 	 * If there are enough tokens, and the queue is empty, send this packet
1544 	 * out immediately.  Otherwise, try to insert it on this vif's queue.
1545 	 */
1546 	if (vifp->tbf_q_len == 0) {
1547 		if (len <= vifp->tbf_n_tok) {
1548 			vifp->tbf_n_tok -= len;
1549 			tbf_send_packet(vifp, m);
1550 		} else {
1551 			/* queue packet and timeout till later */
1552 			tbf_queue(vifp, m);
1553 			timeout(tbf_reprocess_q, vifp, TBF_REPROCESS);
1554 		}
1555 	} else {
1556 		if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
1557 		    !tbf_dq_sel(vifp, ip)) {
1558 			/* queue length too much, and couldn't make room */
1559 			mrtstat.mrts_q_overflow++;
1560 			m_freem(m);
1561 		} else {
1562 			/* queue length low enough, or made room */
1563 			tbf_queue(vifp, m);
1564 			tbf_process_q(vifp);
1565 		}
1566 	}
1567 }
1568 
1569 /*
1570  * adds a packet to the queue at the interface
1571  */
1572 static void
1573 tbf_queue(vifp, m)
1574 	register struct vif *vifp;
1575 	register struct mbuf *m;
1576 {
1577 	register int s = splsoftnet();
1578 
1579 	/* insert at tail */
1580 	*vifp->tbf_t = m;
1581 	vifp->tbf_t = &m->m_nextpkt;
1582 	vifp->tbf_q_len++;
1583 
1584 	splx(s);
1585 }
1586 
1587 
1588 /*
1589  * processes the queue at the interface
1590  */
1591 static void
1592 tbf_process_q(vifp)
1593 	register struct vif *vifp;
1594 {
1595 	register struct mbuf *m;
1596 	register int len;
1597 	register int s = splsoftnet();
1598 
1599 	/*
1600 	 * Loop through the queue at the interface and send as many packets
1601 	 * as possible.
1602 	 */
1603 	for (m = vifp->tbf_q;
1604 	    m != 0;
1605 	    m = vifp->tbf_q) {
1606 		len = mtod(m, struct ip *)->ip_len;
1607 
1608 		/* determine if the packet can be sent */
1609 		if (len <= vifp->tbf_n_tok) {
1610 			/* if so,
1611 			 * reduce no of tokens, dequeue the packet,
1612 			 * send the packet.
1613 			 */
1614 			if ((vifp->tbf_q = m->m_nextpkt) == 0)
1615 				vifp->tbf_t = &vifp->tbf_q;
1616 			--vifp->tbf_q_len;
1617 
1618 			m->m_nextpkt = 0;
1619 			vifp->tbf_n_tok -= len;
1620 			tbf_send_packet(vifp, m);
1621 		} else
1622 			break;
1623 	}
1624 	splx(s);
1625 }
1626 
1627 static void
1628 tbf_reprocess_q(arg)
1629 	void *arg;
1630 {
1631 	register struct vif *vifp = arg;
1632 
1633 	if (ip_mrouter == 0)
1634 		return;
1635 
1636 	tbf_update_tokens(vifp);
1637 	tbf_process_q(vifp);
1638 
1639 	if (vifp->tbf_q_len != 0)
1640 		timeout(tbf_reprocess_q, vifp, TBF_REPROCESS);
1641 }
1642 
1643 /* function that will selectively discard a member of the queue
1644  * based on the precedence value and the priority
1645  */
1646 static int
1647 tbf_dq_sel(vifp, ip)
1648 	register struct vif *vifp;
1649 	register struct ip *ip;
1650 {
1651 	register u_int p;
1652 	register struct mbuf **mp, *m;
1653 	register int s = splsoftnet();
1654 
1655 	p = priority(vifp, ip);
1656 
1657 	for (mp = &vifp->tbf_q, m = *mp;
1658 	    m != 0;
1659 	    mp = &m->m_nextpkt, m = *mp) {
1660 		if (p > priority(vifp, mtod(m, struct ip *))) {
1661 			if ((*mp = m->m_nextpkt) == 0)
1662 				vifp->tbf_t = mp;
1663 			--vifp->tbf_q_len;
1664 
1665 			m_freem(m);
1666 			mrtstat.mrts_drop_sel++;
1667 			splx(s);
1668 			return (1);
1669 		}
1670 	}
1671 	splx(s);
1672 	return (0);
1673 }
1674 
1675 static void
1676 tbf_send_packet(vifp, m)
1677 	register struct vif *vifp;
1678 	register struct mbuf *m;
1679 {
1680 	int error;
1681 	int s = splsoftnet();
1682 
1683 	if (vifp->v_flags & VIFF_TUNNEL) {
1684 		/* If tunnel options */
1685 		ip_output(m, (struct mbuf *)0, &vifp->v_route,
1686 			  IP_FORWARDING, (struct ip_moptions *)0);
1687 	} else {
1688 		/* if physical interface option, extract the options and then send */
1689 		struct ip_moptions imo;
1690 
1691 		imo.imo_multicast_ifp = vifp->v_ifp;
1692 		imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1693 		imo.imo_multicast_loop = 1;
1694 #ifdef RSVP_ISI
1695 		imo.imo_multicast_vif = -1;
1696 #endif
1697 
1698 		error = ip_output(m, (struct mbuf *)0, (struct route *)0,
1699 				  IP_FORWARDING|IP_MULTICASTOPTS, &imo);
1700 
1701 		if (mrtdebug & DEBUG_XMIT)
1702 			log(LOG_DEBUG, "phyint_send on vif %d err %d\n",
1703 			    vifp-viftable, error);
1704 	}
1705 	splx(s);
1706 }
1707 
1708 /* determine the current time and then
1709  * the elapsed time (between the last time and time now)
1710  * in milliseconds & update the no. of tokens in the bucket
1711  */
1712 static void
1713 tbf_update_tokens(vifp)
1714 	register struct vif *vifp;
1715 {
1716 	struct timeval tp;
1717 	register u_int32_t tm;
1718 	register int s = splsoftnet();
1719 
1720 	microtime(&tp);
1721 
1722 	TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
1723 
1724 	/*
1725 	 * This formula is actually
1726 	 * "time in seconds" * "bytes/second".
1727 	 *
1728 	 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
1729 	 *
1730 	 * The (1000/1024) was introduced in add_vif to optimize
1731 	 * this divide into a shift.
1732 	 */
1733 	vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
1734 	vifp->tbf_last_pkt_t = tp;
1735 
1736 	if (vifp->tbf_n_tok > MAX_BKT_SIZE)
1737 		vifp->tbf_n_tok = MAX_BKT_SIZE;
1738 
1739 	splx(s);
1740 }
1741 
1742 static int
1743 priority(vifp, ip)
1744     register struct vif *vifp;
1745     register struct ip *ip;
1746 {
1747     register int prio;
1748 
1749     /* temporary hack; may add general packet classifier some day */
1750 
1751     /*
1752      * The UDP port space is divided up into four priority ranges:
1753      * [0, 16384)     : unclassified - lowest priority
1754      * [16384, 32768) : audio - highest priority
1755      * [32768, 49152) : whiteboard - medium priority
1756      * [49152, 65536) : video - low priority
1757      */
1758     if (ip->ip_p == IPPROTO_UDP) {
1759 	struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
1760 
1761 	switch (ntohs(udp->uh_dport) & 0xc000) {
1762 	    case 0x4000:
1763 		prio = 70;
1764 		break;
1765 	    case 0x8000:
1766 		prio = 60;
1767 		break;
1768 	    case 0xc000:
1769 		prio = 55;
1770 		break;
1771 	    default:
1772 		prio = 50;
1773 		break;
1774 	}
1775 
1776 	if (tbfdebug > 1)
1777 	    log(LOG_DEBUG, "port %x prio %d\n", ntohs(udp->uh_dport), prio);
1778     } else
1779 	prio = 50;
1780 
1781 
1782     return (prio);
1783 }
1784 
1785 /*
1786  * End of token bucket filter modifications
1787  */
1788 
1789 #ifdef RSVP_ISI
1790 
1791 int
1792 ip_rsvp_vif_init(so, m)
1793     struct socket *so;
1794     struct mbuf *m;
1795 {
1796     int i;
1797     register int s;
1798 
1799     if (rsvpdebug)
1800 	printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
1801 	    so->so_type, so->so_proto->pr_protocol);
1802 
1803     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
1804 	return (EOPNOTSUPP);
1805 
1806     /* Check mbuf. */
1807     if (m == 0 || m->m_len != sizeof(int)) {
1808 	return (EINVAL);
1809     }
1810     i = *(mtod(m, int *));
1811 
1812     if (rsvpdebug)
1813 	printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on);
1814 
1815     s = splsoftnet();
1816 
1817     /* Check vif. */
1818     if (!legal_vif_num(i)) {
1819 	splx(s);
1820 	return (EADDRNOTAVAIL);
1821     }
1822 
1823     /* Check if socket is available. */
1824     if (viftable[i].v_rsvpd != 0) {
1825 	splx(s);
1826 	return (EADDRINUSE);
1827     }
1828 
1829     viftable[i].v_rsvpd = so;
1830     /* This may seem silly, but we need to be sure we don't over-increment
1831      * the RSVP counter, in case something slips up.
1832      */
1833     if (!viftable[i].v_rsvp_on) {
1834 	viftable[i].v_rsvp_on = 1;
1835 	rsvp_on++;
1836     }
1837 
1838     splx(s);
1839     return (0);
1840 }
1841 
1842 int
1843 ip_rsvp_vif_done(so, m)
1844     struct socket *so;
1845     struct mbuf *m;
1846 {
1847     int i;
1848     register int s;
1849 
1850     if (rsvpdebug)
1851 	printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
1852 	       so->so_type, so->so_proto->pr_protocol);
1853 
1854     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
1855 	return (EOPNOTSUPP);
1856 
1857     /* Check mbuf. */
1858     if (m == 0 || m->m_len != sizeof(int)) {
1859 	return (EINVAL);
1860     }
1861     i = *(mtod(m, int *));
1862 
1863     s = splsoftnet();
1864 
1865     /* Check vif. */
1866     if (!legal_vif_num(i)) {
1867 	splx(s);
1868         return (EADDRNOTAVAIL);
1869     }
1870 
1871     if (rsvpdebug)
1872 	printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
1873 	    viftable[i].v_rsvpd, so);
1874 
1875     viftable[i].v_rsvpd = 0;
1876     /* This may seem silly, but we need to be sure we don't over-decrement
1877      * the RSVP counter, in case something slips up.
1878      */
1879     if (viftable[i].v_rsvp_on) {
1880 	viftable[i].v_rsvp_on = 0;
1881 	rsvp_on--;
1882     }
1883 
1884     splx(s);
1885     return (0);
1886 }
1887 
1888 void
1889 ip_rsvp_force_done(so)
1890     struct socket *so;
1891 {
1892     int vifi;
1893     register int s;
1894 
1895     /* Don't bother if it is not the right type of socket. */
1896     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
1897 	return;
1898 
1899     s = splsoftnet();
1900 
1901     /* The socket may be attached to more than one vif...this
1902      * is perfectly legal.
1903      */
1904     for (vifi = 0; vifi < numvifs; vifi++) {
1905 	if (viftable[vifi].v_rsvpd == so) {
1906 	    viftable[vifi].v_rsvpd = 0;
1907 	    /* This may seem silly, but we need to be sure we don't
1908 	     * over-decrement the RSVP counter, in case something slips up.
1909 	     */
1910 	    if (viftable[vifi].v_rsvp_on) {
1911 		viftable[vifi].v_rsvp_on = 0;
1912 		rsvp_on--;
1913 	    }
1914 	}
1915     }
1916 
1917     splx(s);
1918     return;
1919 }
1920 
1921 void
1922 rsvp_input(m, ifp)
1923     struct mbuf *m;
1924     struct ifnet *ifp;
1925 {
1926     int vifi;
1927     register struct ip *ip = mtod(m, struct ip *);
1928     static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET };
1929     register int s;
1930 
1931     if (rsvpdebug)
1932 	printf("rsvp_input: rsvp_on %d\n",rsvp_on);
1933 
1934     /* Can still get packets with rsvp_on = 0 if there is a local member
1935      * of the group to which the RSVP packet is addressed.  But in this
1936      * case we want to throw the packet away.
1937      */
1938     if (!rsvp_on) {
1939 	m_freem(m);
1940 	return;
1941     }
1942 
1943     /* If the old-style non-vif-associated socket is set, then use
1944      * it and ignore the new ones.
1945      */
1946     if (ip_rsvpd != 0) {
1947 	if (rsvpdebug)
1948 	    printf("rsvp_input: Sending packet up old-style socket\n");
1949 	rip_input(m);
1950 	return;
1951     }
1952 
1953     s = splsoftnet();
1954 
1955     if (rsvpdebug)
1956 	printf("rsvp_input: check vifs\n");
1957 
1958     /* Find which vif the packet arrived on. */
1959     for (vifi = 0; vifi < numvifs; vifi++) {
1960 	if (viftable[vifi].v_ifp == ifp)
1961 	    break;
1962     }
1963 
1964     if (vifi == numvifs) {
1965 	/* Can't find vif packet arrived on. Drop packet. */
1966 	if (rsvpdebug)
1967 	    printf("rsvp_input: Can't find vif for packet...dropping it.\n");
1968 	m_freem(m);
1969 	splx(s);
1970 	return;
1971     }
1972 
1973     if (rsvpdebug)
1974 	printf("rsvp_input: check socket\n");
1975 
1976     if (viftable[vifi].v_rsvpd == 0) {
1977 	/* drop packet, since there is no specific socket for this
1978 	 * interface */
1979 	if (rsvpdebug)
1980 	    printf("rsvp_input: No socket defined for vif %d\n",vifi);
1981 	m_freem(m);
1982 	splx(s);
1983 	return;
1984     }
1985 
1986     rsvp_src.sin_addr = ip->ip_src;
1987 
1988     if (rsvpdebug && m)
1989 	printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
1990 	       m->m_len,sbspace(&viftable[vifi].v_rsvpd->so_rcv));
1991 
1992     if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
1993 	if (rsvpdebug)
1994 	    printf("rsvp_input: Failed to append to socket\n");
1995     else
1996 	if (rsvpdebug)
1997 	    printf("rsvp_input: send packet up\n");
1998 
1999     splx(s);
2000 }
2001 #endif /* RSVP_ISI */
2002