xref: /netbsd-src/sys/netinet/ip_mroute.c (revision 3b01aba77a7a698587faaae455bbfe740923c1f5)
1 /*	$NetBSD: ip_mroute.c,v 1.56 2001/07/22 13:34:11 wiz 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 "opt_ipsec.h"
18 
19 #include <sys/param.h>
20 #include <sys/systm.h>
21 #include <sys/callout.h>
22 #include <sys/mbuf.h>
23 #include <sys/socket.h>
24 #include <sys/socketvar.h>
25 #include <sys/protosw.h>
26 #include <sys/errno.h>
27 #include <sys/time.h>
28 #include <sys/kernel.h>
29 #include <sys/ioctl.h>
30 #include <sys/syslog.h>
31 #include <net/if.h>
32 #include <net/route.h>
33 #include <net/raw_cb.h>
34 #include <netinet/in.h>
35 #include <netinet/in_var.h>
36 #include <netinet/in_systm.h>
37 #include <netinet/ip.h>
38 #include <netinet/ip_var.h>
39 #include <netinet/in_pcb.h>
40 #include <netinet/udp.h>
41 #include <netinet/igmp.h>
42 #include <netinet/igmp_var.h>
43 #include <netinet/ip_mroute.h>
44 #include <netinet/ip_encap.h>
45 
46 #include <machine/stdarg.h>
47 
48 #define IP_MULTICASTOPTS 0
49 #define	M_PULLUP(m, len) \
50 	do { \
51 		if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
52 			(m) = m_pullup((m), (len)); \
53 	} while (0)
54 
55 /*
56  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
57  * except for netstat or debugging purposes.
58  */
59 struct socket  *ip_mrouter  = 0;
60 int		ip_mrtproto = IGMP_DVMRP;    /* for netstat only */
61 
62 #define NO_RTE_FOUND 	0x1
63 #define RTE_FOUND	0x2
64 
65 #define	MFCHASH(a, g) \
66 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
67 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
68 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
69 u_long	mfchash;
70 
71 u_char		nexpire[MFCTBLSIZ];
72 struct vif	viftable[MAXVIFS];
73 struct mrtstat	mrtstat;
74 u_int		mrtdebug = 0;	  /* debug level 	*/
75 #define		DEBUG_MFC	0x02
76 #define		DEBUG_FORWARD	0x04
77 #define		DEBUG_EXPIRE	0x08
78 #define		DEBUG_XMIT	0x10
79 u_int       	tbfdebug = 0;     /* tbf debug level 	*/
80 #ifdef RSVP_ISI
81 u_int		rsvpdebug = 0;	  /* rsvp debug level   */
82 extern struct socket *ip_rsvpd;
83 extern int rsvp_on;
84 #endif /* RSVP_ISI */
85 
86 /* vif attachment using sys/netinet/ip_encap.c */
87 extern struct domain inetdomain;
88 static void vif_input __P((struct mbuf *, ...));
89 static int vif_encapcheck __P((const struct mbuf *, int, int, void *));
90 static struct protosw vif_protosw =
91 { SOCK_RAW,	&inetdomain,	IPPROTO_IPV4,	PR_ATOMIC|PR_ADDR,
92   vif_input,	rip_output,	0,		rip_ctloutput,
93   rip_usrreq,
94   0,            0,              0,              0,
95 };
96 
97 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
98 #define		UPCALL_EXPIRE	6		/* number of timeouts */
99 
100 /*
101  * Define the token bucket filter structures
102  */
103 
104 #define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
105 
106 static int get_sg_cnt __P((struct sioc_sg_req *));
107 static int get_vif_cnt __P((struct sioc_vif_req *));
108 static int ip_mrouter_init __P((struct socket *, struct mbuf *));
109 static int get_version __P((struct mbuf *));
110 static int set_assert __P((struct mbuf *));
111 static int get_assert __P((struct mbuf *));
112 static int add_vif __P((struct mbuf *));
113 static int del_vif __P((struct mbuf *));
114 static void update_mfc __P((struct mfcctl *, struct mfc *));
115 static void expire_mfc __P((struct mfc *));
116 static int add_mfc __P((struct mbuf *));
117 #ifdef UPCALL_TIMING
118 static void collate __P((struct timeval *));
119 #endif
120 static int del_mfc __P((struct mbuf *));
121 static int socket_send __P((struct socket *, struct mbuf *,
122 			    struct sockaddr_in *));
123 static void expire_upcalls __P((void *));
124 #ifdef RSVP_ISI
125 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *, vifi_t));
126 #else
127 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *));
128 #endif
129 static void phyint_send __P((struct ip *, struct vif *, struct mbuf *));
130 static void encap_send __P((struct ip *, struct vif *, struct mbuf *));
131 static void tbf_control __P((struct vif *, struct mbuf *, struct ip *,
132 			     u_int32_t));
133 static void tbf_queue __P((struct vif *, struct mbuf *));
134 static void tbf_process_q __P((struct vif *));
135 static void tbf_reprocess_q __P((void *));
136 static int tbf_dq_sel __P((struct vif *, struct ip *));
137 static void tbf_send_packet __P((struct vif *, struct mbuf *));
138 static void tbf_update_tokens __P((struct vif *));
139 static int priority __P((struct vif *, struct ip *));
140 
141 /*
142  * 'Interfaces' associated with decapsulator (so we can tell
143  * packets that went through it from ones that get reflected
144  * by a broken gateway).  These interfaces are never linked into
145  * the system ifnet list & no routes point to them.  I.e., packets
146  * can't be sent this way.  They only exist as a placeholder for
147  * multicast source verification.
148  */
149 #if 0
150 struct ifnet multicast_decap_if[MAXVIFS];
151 #endif
152 
153 #define	ENCAP_TTL	64
154 #define	ENCAP_PROTO	IPPROTO_IPIP	/* 4 */
155 
156 /* prototype IP hdr for encapsulated packets */
157 struct ip multicast_encap_iphdr = {
158 #if BYTE_ORDER == LITTLE_ENDIAN
159 	sizeof(struct ip) >> 2, IPVERSION,
160 #else
161 	IPVERSION, sizeof(struct ip) >> 2,
162 #endif
163 	0,				/* tos */
164 	sizeof(struct ip),		/* total length */
165 	0,				/* id */
166 	0,				/* frag offset */
167 	ENCAP_TTL, ENCAP_PROTO,
168 	0,				/* checksum */
169 };
170 
171 /*
172  * Private variables.
173  */
174 static vifi_t	   numvifs = 0;
175 static int have_encap_tunnel = 0;
176 
177 static struct callout expire_upcalls_ch;
178 
179 /*
180  * one-back cache used by mrt_ipip_input to locate a tunnel's vif
181  * given a datagram's src ip address.
182  */
183 static struct in_addr last_encap_src;
184 static struct vif *last_encap_vif;
185 
186 /*
187  * whether or not special PIM assert processing is enabled.
188  */
189 static int pim_assert;
190 /*
191  * Rate limit for assert notification messages, in usec
192  */
193 #define ASSERT_MSG_TIME		3000000
194 
195 /*
196  * Find a route for a given origin IP address and Multicast group address
197  * Type of service parameter to be added in the future!!!
198  */
199 
200 #define MFCFIND(o, g, rt) { \
201 	struct mfc *_rt; \
202 	(rt) = 0; \
203 	++mrtstat.mrts_mfc_lookups; \
204 	for (_rt = mfchashtbl[MFCHASH(o, g)].lh_first; \
205 	     _rt; _rt = _rt->mfc_hash.le_next) { \
206 		if (in_hosteq(_rt->mfc_origin, (o)) && \
207 		    in_hosteq(_rt->mfc_mcastgrp, (g)) && \
208 		    _rt->mfc_stall == 0) { \
209 			(rt) = _rt; \
210 			break; \
211 		} \
212 	} \
213 	if ((rt) == 0) \
214 		++mrtstat.mrts_mfc_misses; \
215 }
216 
217 /*
218  * Macros to compute elapsed time efficiently
219  * Borrowed from Van Jacobson's scheduling code
220  */
221 #define TV_DELTA(a, b, delta) { \
222 	int xxs; \
223 	delta = (a).tv_usec - (b).tv_usec; \
224 	xxs = (a).tv_sec - (b).tv_sec; \
225 	switch (xxs) { \
226 	case 2: \
227 		delta += 1000000; \
228 		/* fall through */ \
229 	case 1: \
230 		delta += 1000000; \
231 		/* fall through */ \
232 	case 0: \
233 		break; \
234 	default: \
235 		delta += (1000000 * xxs); \
236 		break; \
237 	} \
238 }
239 
240 #ifdef UPCALL_TIMING
241 u_int32_t upcall_data[51];
242 #endif /* UPCALL_TIMING */
243 
244 /*
245  * Handle MRT setsockopt commands to modify the multicast routing tables.
246  */
247 int
248 ip_mrouter_set(so, optname, m)
249 	struct socket *so;
250 	int optname;
251 	struct mbuf **m;
252 {
253 	int error;
254 
255 	if (optname != MRT_INIT && so != ip_mrouter)
256 		error = ENOPROTOOPT;
257 	else
258 		switch (optname) {
259 		case MRT_INIT:
260 			error = ip_mrouter_init(so, *m);
261 			break;
262 		case MRT_DONE:
263 			error = ip_mrouter_done();
264 			break;
265 		case MRT_ADD_VIF:
266 			error = add_vif(*m);
267 			break;
268 		case MRT_DEL_VIF:
269 			error = del_vif(*m);
270 			break;
271 		case MRT_ADD_MFC:
272 			error = add_mfc(*m);
273 			break;
274 		case MRT_DEL_MFC:
275 			error = del_mfc(*m);
276 			break;
277 		case MRT_ASSERT:
278 			error = set_assert(*m);
279 			break;
280 		default:
281 			error = ENOPROTOOPT;
282 			break;
283 		}
284 
285 	if (*m)
286 		m_free(*m);
287 	return (error);
288 }
289 
290 /*
291  * Handle MRT getsockopt commands
292  */
293 int
294 ip_mrouter_get(so, optname, m)
295 	struct socket *so;
296 	int optname;
297 	struct mbuf **m;
298 {
299 	int error;
300 
301 	if (so != ip_mrouter)
302 		error = ENOPROTOOPT;
303 	else {
304 		*m = m_get(M_WAIT, MT_SOOPTS);
305 
306 		switch (optname) {
307 		case MRT_VERSION:
308 			error = get_version(*m);
309 			break;
310 		case MRT_ASSERT:
311 			error = get_assert(*m);
312 			break;
313 		default:
314 			error = ENOPROTOOPT;
315 			break;
316 		}
317 
318 		if (error)
319 			m_free(*m);
320 	}
321 
322 	return (error);
323 }
324 
325 /*
326  * Handle ioctl commands to obtain information from the cache
327  */
328 int
329 mrt_ioctl(so, cmd, data)
330 	struct socket *so;
331 	u_long cmd;
332 	caddr_t data;
333 {
334 	int error;
335 
336 	if (so != ip_mrouter)
337 		error = EINVAL;
338 	else
339 		switch (cmd) {
340 		case SIOCGETVIFCNT:
341 			error = get_vif_cnt((struct sioc_vif_req *)data);
342 			break;
343 		case SIOCGETSGCNT:
344 			error = get_sg_cnt((struct sioc_sg_req *)data);
345 			break;
346 		default:
347 			error = EINVAL;
348 			break;
349 		}
350 
351 	return (error);
352 }
353 
354 /*
355  * returns the packet, byte, rpf-failure count for the source group provided
356  */
357 static int
358 get_sg_cnt(req)
359 	struct sioc_sg_req *req;
360 {
361 	struct mfc *rt;
362 	int s;
363 
364 	s = splsoftnet();
365 	MFCFIND(req->src, req->grp, rt);
366 	splx(s);
367 	if (rt != 0) {
368 		req->pktcnt = rt->mfc_pkt_cnt;
369 		req->bytecnt = rt->mfc_byte_cnt;
370 		req->wrong_if = rt->mfc_wrong_if;
371 	} else
372 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
373 
374 	return (0);
375 }
376 
377 /*
378  * returns the input and output packet and byte counts on the vif provided
379  */
380 static int
381 get_vif_cnt(req)
382 	struct sioc_vif_req *req;
383 {
384 	vifi_t vifi = req->vifi;
385 
386 	if (vifi >= numvifs)
387 		return (EINVAL);
388 
389 	req->icount = viftable[vifi].v_pkt_in;
390 	req->ocount = viftable[vifi].v_pkt_out;
391 	req->ibytes = viftable[vifi].v_bytes_in;
392 	req->obytes = viftable[vifi].v_bytes_out;
393 
394 	return (0);
395 }
396 
397 /*
398  * Enable multicast routing
399  */
400 static int
401 ip_mrouter_init(so, m)
402 	struct socket *so;
403 	struct mbuf *m;
404 {
405 	int *v;
406 
407 	if (mrtdebug)
408 		log(LOG_DEBUG,
409 		    "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
410 		    so->so_type, so->so_proto->pr_protocol);
411 
412 	if (so->so_type != SOCK_RAW ||
413 	    so->so_proto->pr_protocol != IPPROTO_IGMP)
414 		return (EOPNOTSUPP);
415 
416 	if (m == 0 || m->m_len < sizeof(int))
417 		return (EINVAL);
418 
419 	v = mtod(m, int *);
420 	if (*v != 1)
421 		return (EINVAL);
422 
423 	if (ip_mrouter != 0)
424 		return (EADDRINUSE);
425 
426 	ip_mrouter = so;
427 
428 	mfchashtbl =
429 	    hashinit(MFCTBLSIZ, HASH_LIST, M_MRTABLE, M_WAITOK, &mfchash);
430 	bzero((caddr_t)nexpire, sizeof(nexpire));
431 
432 	pim_assert = 0;
433 
434 	callout_init(&expire_upcalls_ch);
435 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
436 	    expire_upcalls, NULL);
437 
438 	if (mrtdebug)
439 		log(LOG_DEBUG, "ip_mrouter_init\n");
440 
441 	return (0);
442 }
443 
444 /*
445  * Disable multicast routing
446  */
447 int
448 ip_mrouter_done()
449 {
450 	vifi_t vifi;
451 	struct vif *vifp;
452 	int i;
453 	int s;
454 
455 	s = splsoftnet();
456 
457 	/* Clear out all the vifs currently in use. */
458 	for (vifi = 0; vifi < numvifs; vifi++) {
459 		vifp = &viftable[vifi];
460 		if (!in_nullhost(vifp->v_lcl_addr))
461 			reset_vif(vifp);
462 	}
463 
464 	numvifs = 0;
465 	pim_assert = 0;
466 
467 	callout_stop(&expire_upcalls_ch);
468 
469 	/*
470 	 * Free all multicast forwarding cache entries.
471 	 */
472 	for (i = 0; i < MFCTBLSIZ; i++) {
473 		struct mfc *rt, *nrt;
474 
475 		for (rt = mfchashtbl[i].lh_first; rt; rt = nrt) {
476 			nrt = rt->mfc_hash.le_next;
477 
478 			expire_mfc(rt);
479 		}
480 	}
481 
482 	free(mfchashtbl, M_MRTABLE);
483 	mfchashtbl = 0;
484 
485 	/* Reset de-encapsulation cache. */
486 	have_encap_tunnel = 0;
487 
488 	ip_mrouter = 0;
489 
490 	splx(s);
491 
492 	if (mrtdebug)
493 		log(LOG_DEBUG, "ip_mrouter_done\n");
494 
495 	return (0);
496 }
497 
498 static int
499 get_version(m)
500 	struct mbuf *m;
501 {
502 	int *v = mtod(m, int *);
503 
504 	*v = 0x0305;	/* XXX !!!! */
505 	m->m_len = sizeof(int);
506 	return (0);
507 }
508 
509 /*
510  * Set PIM assert processing global
511  */
512 static int
513 set_assert(m)
514 	struct mbuf *m;
515 {
516 	int *i;
517 
518 	if (m == 0 || m->m_len < sizeof(int))
519 		return (EINVAL);
520 
521 	i = mtod(m, int *);
522 	pim_assert = !!*i;
523 	return (0);
524 }
525 
526 /*
527  * Get PIM assert processing global
528  */
529 static int
530 get_assert(m)
531 	struct mbuf *m;
532 {
533 	int *i = mtod(m, int *);
534 
535 	*i = pim_assert;
536 	m->m_len = sizeof(int);
537 	return (0);
538 }
539 
540 static struct sockaddr_in sin = { sizeof(sin), AF_INET };
541 
542 /*
543  * Add a vif to the vif table
544  */
545 static int
546 add_vif(m)
547 	struct mbuf *m;
548 {
549 	struct vifctl *vifcp;
550 	struct vif *vifp;
551 	struct ifaddr *ifa;
552 	struct ifnet *ifp;
553 	struct ifreq ifr;
554 	int error, s;
555 
556 	if (m == 0 || m->m_len < sizeof(struct vifctl))
557 		return (EINVAL);
558 
559 	vifcp = mtod(m, struct vifctl *);
560 	if (vifcp->vifc_vifi >= MAXVIFS)
561 		return (EINVAL);
562 
563 	vifp = &viftable[vifcp->vifc_vifi];
564 	if (!in_nullhost(vifp->v_lcl_addr))
565 		return (EADDRINUSE);
566 
567 	/* Find the interface with an address in AF_INET family. */
568 	sin.sin_addr = vifcp->vifc_lcl_addr;
569 	ifa = ifa_ifwithaddr(sintosa(&sin));
570 	if (ifa == 0)
571 		return (EADDRNOTAVAIL);
572 
573 	if (vifcp->vifc_flags & VIFF_TUNNEL) {
574 		if (vifcp->vifc_flags & VIFF_SRCRT) {
575 			log(LOG_ERR, "Source routed tunnels not supported\n");
576 			return (EOPNOTSUPP);
577 		}
578 
579 		/* attach this vif to decapsulator dispatch table */
580 		vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
581 		    vif_encapcheck, &vif_protosw, vifp);
582 		if (!vifp->v_encap_cookie)
583 			return (EINVAL);
584 
585 		/* Create a fake encapsulation interface. */
586 		ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK);
587 		bzero(ifp, sizeof(*ifp));
588 		sprintf(ifp->if_xname, "mdecap%d", vifcp->vifc_vifi);
589 
590 		/* Prepare cached route entry. */
591 		bzero(&vifp->v_route, sizeof(vifp->v_route));
592 
593 		/*
594 		 * Tell mrt_ipip_input() to start looking at encapsulated
595 		 * packets.
596 		 */
597 		have_encap_tunnel = 1;
598 	} else {
599 		/* Use the physical interface associated with the address. */
600 		ifp = ifa->ifa_ifp;
601 
602 		/* Make sure the interface supports multicast. */
603 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
604 			return (EOPNOTSUPP);
605 
606 		/* Enable promiscuous reception of all IP multicasts. */
607 		satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
608 		satosin(&ifr.ifr_addr)->sin_family = AF_INET;
609 		satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
610 		error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
611 		if (error)
612 			return (error);
613 	}
614 
615 	s = splsoftnet();
616 
617 	/* Define parameters for the tbf structure. */
618 	vifp->tbf_q = 0;
619 	vifp->tbf_t = &vifp->tbf_q;
620 	microtime(&vifp->tbf_last_pkt_t);
621 	vifp->tbf_n_tok = 0;
622 	vifp->tbf_q_len = 0;
623 	vifp->tbf_max_q_len = MAXQSIZE;
624 
625 	vifp->v_flags = vifcp->vifc_flags;
626 	vifp->v_threshold = vifcp->vifc_threshold;
627 	/* scaling up here allows division by 1024 in critical code */
628 	vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
629 	vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
630 	vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
631 	vifp->v_ifp = ifp;
632 	/* Initialize per vif pkt counters. */
633 	vifp->v_pkt_in = 0;
634 	vifp->v_pkt_out = 0;
635 	vifp->v_bytes_in = 0;
636 	vifp->v_bytes_out = 0;
637 
638 	callout_init(&vifp->v_repq_ch);
639 
640 #ifdef RSVP_ISI
641 	vifp->v_rsvp_on = 0;
642 	vifp->v_rsvpd = 0;
643 #endif /* RSVP_ISI */
644 
645 	splx(s);
646 
647 	/* Adjust numvifs up if the vifi is higher than numvifs. */
648 	if (numvifs <= vifcp->vifc_vifi)
649 		numvifs = vifcp->vifc_vifi + 1;
650 
651 	if (mrtdebug)
652 		log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
653 		    vifcp->vifc_vifi,
654 		    ntohl(vifcp->vifc_lcl_addr.s_addr),
655 		    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
656 		    ntohl(vifcp->vifc_rmt_addr.s_addr),
657 		    vifcp->vifc_threshold,
658 		    vifcp->vifc_rate_limit);
659 
660 	return (0);
661 }
662 
663 void
664 reset_vif(vifp)
665 	struct vif *vifp;
666 {
667 	struct mbuf *m, *n;
668 	struct ifnet *ifp;
669 	struct ifreq ifr;
670 
671 	callout_stop(&vifp->v_repq_ch);
672 
673 	/* detach this vif from decapsulator dispatch table */
674 	encap_detach(vifp->v_encap_cookie);
675 	vifp->v_encap_cookie = NULL;
676 
677 	for (m = vifp->tbf_q; m != 0; m = n) {
678 		n = m->m_nextpkt;
679 		m_freem(m);
680 	}
681 
682 	if (vifp->v_flags & VIFF_TUNNEL) {
683 		free(vifp->v_ifp, M_MRTABLE);
684 		if (vifp == last_encap_vif) {
685 			last_encap_vif = 0;
686 			last_encap_src = zeroin_addr;
687 		}
688 	} else {
689 		satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
690 		satosin(&ifr.ifr_addr)->sin_family = AF_INET;
691 		satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
692 		ifp = vifp->v_ifp;
693 		(*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
694 	}
695 	bzero((caddr_t)vifp, sizeof(*vifp));
696 }
697 
698 /*
699  * Delete a vif from the vif table
700  */
701 static int
702 del_vif(m)
703 	struct mbuf *m;
704 {
705 	vifi_t *vifip;
706 	struct vif *vifp;
707 	vifi_t vifi;
708 	int s;
709 
710 	if (m == 0 || m->m_len < sizeof(vifi_t))
711 		return (EINVAL);
712 
713 	vifip = mtod(m, vifi_t *);
714 	if (*vifip >= numvifs)
715 		return (EINVAL);
716 
717 	vifp = &viftable[*vifip];
718 	if (in_nullhost(vifp->v_lcl_addr))
719 		return (EADDRNOTAVAIL);
720 
721 	s = splsoftnet();
722 
723 	reset_vif(vifp);
724 
725 	/* Adjust numvifs down */
726 	for (vifi = numvifs; vifi > 0; vifi--)
727 		if (!in_nullhost(viftable[vifi-1].v_lcl_addr))
728 			break;
729 	numvifs = vifi;
730 
731 	splx(s);
732 
733 	if (mrtdebug)
734 		log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
735 
736 	return (0);
737 }
738 
739 static void
740 update_mfc(mfccp, rt)
741 	struct mfcctl *mfccp;
742 	struct mfc *rt;
743 {
744 	vifi_t vifi;
745 
746 	rt->mfc_parent = mfccp->mfcc_parent;
747 	for (vifi = 0; vifi < numvifs; vifi++)
748 		rt->mfc_ttls[vifi] = mfccp->mfcc_ttls[vifi];
749 	rt->mfc_expire = 0;
750 	rt->mfc_stall = 0;
751 }
752 
753 static void
754 expire_mfc(rt)
755 	struct mfc *rt;
756 {
757 	struct rtdetq *rte, *nrte;
758 
759 	for (rte = rt->mfc_stall; rte != 0; rte = nrte) {
760 		nrte = rte->next;
761 		m_freem(rte->m);
762 		free(rte, M_MRTABLE);
763 	}
764 
765 	LIST_REMOVE(rt, mfc_hash);
766 	free(rt, M_MRTABLE);
767 }
768 
769 /*
770  * Add an mfc entry
771  */
772 static int
773 add_mfc(m)
774 	struct mbuf *m;
775 {
776 	struct mfcctl *mfccp;
777 	struct mfc *rt;
778 	u_int32_t hash = 0;
779 	struct rtdetq *rte, *nrte;
780 	u_short nstl;
781 	int s;
782 
783 	if (m == 0 || m->m_len < sizeof(struct mfcctl))
784 		return (EINVAL);
785 
786 	mfccp = mtod(m, struct mfcctl *);
787 
788 	s = splsoftnet();
789 	MFCFIND(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp, rt);
790 
791 	/* If an entry already exists, just update the fields */
792 	if (rt) {
793 		if (mrtdebug & DEBUG_MFC)
794 			log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n",
795 			    ntohl(mfccp->mfcc_origin.s_addr),
796 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
797 			    mfccp->mfcc_parent);
798 
799 		if (rt->mfc_expire)
800 			nexpire[hash]--;
801 
802 		update_mfc(mfccp, rt);
803 
804 		splx(s);
805 		return (0);
806 	}
807 
808 	/*
809 	 * Find the entry for which the upcall was made and update
810 	 */
811 	nstl = 0;
812 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
813 	for (rt = mfchashtbl[hash].lh_first; rt; rt = rt->mfc_hash.le_next) {
814 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
815 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
816 		    rt->mfc_stall != 0) {
817 			if (nstl++)
818 				log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
819 				    "multiple kernel entries",
820 				    ntohl(mfccp->mfcc_origin.s_addr),
821 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
822 				    mfccp->mfcc_parent, rt->mfc_stall);
823 
824 			if (mrtdebug & DEBUG_MFC)
825 				log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %p\n",
826 				    ntohl(mfccp->mfcc_origin.s_addr),
827 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
828 				    mfccp->mfcc_parent, rt->mfc_stall);
829 
830 			if (rt->mfc_expire)
831 				nexpire[hash]--;
832 
833 			rte = rt->mfc_stall;
834 			update_mfc(mfccp, rt);
835 
836 			/* free packets Qed at the end of this entry */
837 			for (; rte != 0; rte = nrte) {
838 				nrte = rte->next;
839 #ifdef RSVP_ISI
840 				ip_mdq(rte->m, rte->ifp, rt, -1);
841 #else
842 				ip_mdq(rte->m, rte->ifp, rt);
843 #endif /* RSVP_ISI */
844 				m_freem(rte->m);
845 #ifdef UPCALL_TIMING
846 				collate(&rte->t);
847 #endif /* UPCALL_TIMING */
848 				free(rte, M_MRTABLE);
849 			}
850 		}
851 	}
852 
853 	if (nstl == 0) {
854 		/*
855 		 * No mfc; make a new one
856 		 */
857 		if (mrtdebug & DEBUG_MFC)
858 			log(LOG_DEBUG,"add_mfc no upcall o %x g %x p %x\n",
859 			    ntohl(mfccp->mfcc_origin.s_addr),
860 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
861 			    mfccp->mfcc_parent);
862 
863 		rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
864 		if (rt == 0) {
865 			splx(s);
866 			return (ENOBUFS);
867 		}
868 
869 		rt->mfc_origin = mfccp->mfcc_origin;
870 		rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
871 		/* initialize pkt counters per src-grp */
872 		rt->mfc_pkt_cnt = 0;
873 		rt->mfc_byte_cnt = 0;
874 		rt->mfc_wrong_if = 0;
875 		timerclear(&rt->mfc_last_assert);
876 		update_mfc(mfccp, rt);
877 
878 		/* insert new entry at head of hash chain */
879 		LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
880 	}
881 
882 	splx(s);
883 	return (0);
884 }
885 
886 #ifdef UPCALL_TIMING
887 /*
888  * collect delay statistics on the upcalls
889  */
890 static void collate(t)
891 struct timeval *t;
892 {
893     u_int32_t d;
894     struct timeval tp;
895     u_int32_t delta;
896 
897     microtime(&tp);
898 
899     if (timercmp(t, &tp, <)) {
900 	TV_DELTA(tp, *t, delta);
901 
902 	d = delta >> 10;
903 	if (d > 50)
904 	    d = 50;
905 
906 	++upcall_data[d];
907     }
908 }
909 #endif /* UPCALL_TIMING */
910 
911 /*
912  * Delete an mfc entry
913  */
914 static int
915 del_mfc(m)
916 	struct mbuf *m;
917 {
918 	struct mfcctl *mfccp;
919 	struct mfc *rt;
920 	int s;
921 
922 	if (m == 0 || m->m_len < sizeof(struct mfcctl))
923 		return (EINVAL);
924 
925 	mfccp = mtod(m, struct mfcctl *);
926 
927 	if (mrtdebug & DEBUG_MFC)
928 		log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
929 		    ntohl(mfccp->mfcc_origin.s_addr),
930 		    ntohl(mfccp->mfcc_mcastgrp.s_addr));
931 
932 	s = splsoftnet();
933 
934 	MFCFIND(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp, rt);
935 	if (rt == 0) {
936 		splx(s);
937 		return (EADDRNOTAVAIL);
938 	}
939 
940 	LIST_REMOVE(rt, mfc_hash);
941 	free(rt, M_MRTABLE);
942 
943 	splx(s);
944 	return (0);
945 }
946 
947 static int
948 socket_send(s, mm, src)
949     struct socket *s;
950     struct mbuf *mm;
951     struct sockaddr_in *src;
952 {
953     if (s) {
954 	if (sbappendaddr(&s->so_rcv, sintosa(src), mm, (struct mbuf *)0) != 0) {
955 	    sorwakeup(s);
956 	    return (0);
957 	}
958     }
959     m_freem(mm);
960     return (-1);
961 }
962 
963 /*
964  * IP multicast forwarding function. This function assumes that the packet
965  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
966  * pointed to by "ifp", and the packet is to be relayed to other networks
967  * that have members of the packet's destination IP multicast group.
968  *
969  * The packet is returned unscathed to the caller, unless it is
970  * erroneous, in which case a non-zero return value tells the caller to
971  * discard it.
972  */
973 
974 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
975 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
976 
977 int
978 #ifdef RSVP_ISI
979 ip_mforward(m, ifp, imo)
980 #else
981 ip_mforward(m, ifp)
982 #endif /* RSVP_ISI */
983     struct mbuf *m;
984     struct ifnet *ifp;
985 #ifdef RSVP_ISI
986     struct ip_moptions *imo;
987 #endif /* RSVP_ISI */
988 {
989     struct ip *ip = mtod(m, struct ip *);
990     struct mfc *rt;
991     u_char *ipoptions;
992     static int srctun = 0;
993     struct mbuf *mm;
994     int s;
995 #ifdef RSVP_ISI
996     struct vif *vifp;
997     vifi_t vifi;
998 #endif /* RSVP_ISI */
999 
1000     /*
1001      * Clear any in-bound checksum flags for this packet.
1002      */
1003     m->m_pkthdr.csum_flags = 0;
1004 
1005     if (mrtdebug & DEBUG_FORWARD)
1006 	log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1007 	    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1008 
1009     if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1010 	(ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1011 	/*
1012 	 * Packet arrived via a physical interface or
1013 	 * an encapuslated tunnel.
1014 	 */
1015     } else {
1016 	/*
1017 	 * Packet arrived through a source-route tunnel.
1018 	 * Source-route tunnels are no longer supported.
1019 	 */
1020 	if ((srctun++ % 1000) == 0)
1021 	    log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n",
1022 		ntohl(ip->ip_src.s_addr));
1023 
1024 	return (1);
1025     }
1026 
1027 #ifdef RSVP_ISI
1028     if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1029 	if (ip->ip_ttl < 255)
1030 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1031 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1032 	    vifp = viftable + vifi;
1033 	    printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n",
1034 		ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1035 		(vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1036 		vifp->v_ifp->if_xname);
1037 	}
1038 	return (ip_mdq(m, ifp, (struct mfc *)0, vifi));
1039     }
1040     if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1041 	printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
1042 	    ntohl(ip->ip_src), ntohl(ip->ip_dst));
1043     }
1044 #endif /* RSVP_ISI */
1045 
1046     /*
1047      * Don't forward a packet with time-to-live of zero or one,
1048      * or a packet destined to a local-only group.
1049      */
1050     if (ip->ip_ttl <= 1 ||
1051 	IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1052 	return (0);
1053 
1054     /*
1055      * Determine forwarding vifs from the forwarding cache table
1056      */
1057     s = splsoftnet();
1058     MFCFIND(ip->ip_src, ip->ip_dst, rt);
1059 
1060     /* Entry exists, so forward if necessary */
1061     if (rt != 0) {
1062 	splx(s);
1063 #ifdef RSVP_ISI
1064 	return (ip_mdq(m, ifp, rt, -1));
1065 #else
1066 	return (ip_mdq(m, ifp, rt));
1067 #endif /* RSVP_ISI */
1068     } else {
1069 	/*
1070 	 * If we don't have a route for packet's origin,
1071 	 * Make a copy of the packet &
1072 	 * send message to routing daemon
1073 	 */
1074 
1075 	struct mbuf *mb0;
1076 	struct rtdetq *rte;
1077 	u_int32_t hash;
1078 	int hlen = ip->ip_hl << 2;
1079 #ifdef UPCALL_TIMING
1080 	struct timeval tp;
1081 
1082 	microtime(&tp);
1083 #endif /* UPCALL_TIMING */
1084 
1085 	mrtstat.mrts_no_route++;
1086 	if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1087 	    log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1088 		ntohl(ip->ip_src.s_addr),
1089 		ntohl(ip->ip_dst.s_addr));
1090 
1091 	/*
1092 	 * Allocate mbufs early so that we don't do extra work if we are
1093 	 * just going to fail anyway.  Make sure to pullup the header so
1094 	 * that other people can't step on it.
1095 	 */
1096 	rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, M_NOWAIT);
1097 	if (rte == 0) {
1098 	    splx(s);
1099 	    return (ENOBUFS);
1100 	}
1101 	mb0 = m_copy(m, 0, M_COPYALL);
1102 	M_PULLUP(mb0, hlen);
1103 	if (mb0 == 0) {
1104 	    free(rte, M_MRTABLE);
1105 	    splx(s);
1106 	    return (ENOBUFS);
1107 	}
1108 
1109 	/* is there an upcall waiting for this packet? */
1110 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1111 	for (rt = mfchashtbl[hash].lh_first; rt; rt = rt->mfc_hash.le_next) {
1112 	    if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1113 		in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1114 		rt->mfc_stall != 0)
1115 		break;
1116 	}
1117 
1118 	if (rt == 0) {
1119 	    int i;
1120 	    struct igmpmsg *im;
1121 
1122 	    /* no upcall, so make a new entry */
1123 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1124 	    if (rt == 0) {
1125 		free(rte, M_MRTABLE);
1126 		m_freem(mb0);
1127 		splx(s);
1128 		return (ENOBUFS);
1129 	    }
1130 	    /* Make a copy of the header to send to the user level process */
1131 	    mm = m_copy(m, 0, hlen);
1132 	    M_PULLUP(mm, hlen);
1133 	    if (mm == 0) {
1134 		free(rte, M_MRTABLE);
1135 		m_freem(mb0);
1136 		free(rt, M_MRTABLE);
1137 		splx(s);
1138 		return (ENOBUFS);
1139 	    }
1140 
1141 	    /*
1142 	     * Send message to routing daemon to install
1143 	     * a route into the kernel table
1144 	     */
1145 	    sin.sin_addr = ip->ip_src;
1146 
1147 	    im = mtod(mm, struct igmpmsg *);
1148 	    im->im_msgtype	= IGMPMSG_NOCACHE;
1149 	    im->im_mbz		= 0;
1150 
1151 	    mrtstat.mrts_upcalls++;
1152 
1153 	    if (socket_send(ip_mrouter, mm, &sin) < 0) {
1154 		log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1155 		++mrtstat.mrts_upq_sockfull;
1156 		free(rte, M_MRTABLE);
1157 		m_freem(mb0);
1158 		free(rt, M_MRTABLE);
1159 		splx(s);
1160 		return (ENOBUFS);
1161 	    }
1162 
1163 	    /* insert new entry at head of hash chain */
1164 	    rt->mfc_origin = ip->ip_src;
1165 	    rt->mfc_mcastgrp = ip->ip_dst;
1166 	    rt->mfc_pkt_cnt = 0;
1167 	    rt->mfc_byte_cnt = 0;
1168 	    rt->mfc_wrong_if = 0;
1169 	    rt->mfc_expire = UPCALL_EXPIRE;
1170 	    nexpire[hash]++;
1171 	    for (i = 0; i < numvifs; i++)
1172 		rt->mfc_ttls[i] = 0;
1173 	    rt->mfc_parent = -1;
1174 
1175 	    /* link into table */
1176 	    LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1177 	    /* Add this entry to the end of the queue */
1178 	    rt->mfc_stall = rte;
1179 	} else {
1180 	    /* determine if q has overflowed */
1181 	    struct rtdetq **p;
1182 	    int npkts = 0;
1183 
1184 	    for (p = &rt->mfc_stall; *p != 0; p = &(*p)->next)
1185 		if (++npkts > MAX_UPQ) {
1186 		    mrtstat.mrts_upq_ovflw++;
1187 		    free(rte, M_MRTABLE);
1188 		    m_freem(mb0);
1189 		    splx(s);
1190 		    return (0);
1191 	        }
1192 
1193 	    /* Add this entry to the end of the queue */
1194 	    *p = rte;
1195 	}
1196 
1197 	rte->next		= 0;
1198 	rte->m 			= mb0;
1199 	rte->ifp 		= ifp;
1200 #ifdef UPCALL_TIMING
1201 	rte->t			= tp;
1202 #endif /* UPCALL_TIMING */
1203 
1204 
1205 	splx(s);
1206 
1207 	return (0);
1208     }
1209 }
1210 
1211 
1212 /*ARGSUSED*/
1213 static void
1214 expire_upcalls(v)
1215 	void *v;
1216 {
1217 	int i;
1218 	int s;
1219 
1220 	s = splsoftnet();
1221 
1222 	for (i = 0; i < MFCTBLSIZ; i++) {
1223 		struct mfc *rt, *nrt;
1224 
1225 		if (nexpire[i] == 0)
1226 			continue;
1227 
1228 		for (rt = mfchashtbl[i].lh_first; rt; rt = nrt) {
1229 			nrt = rt->mfc_hash.le_next;
1230 
1231 			if (rt->mfc_expire == 0 ||
1232 			    --rt->mfc_expire > 0)
1233 				continue;
1234 			nexpire[i]--;
1235 
1236 			++mrtstat.mrts_cache_cleanups;
1237 			if (mrtdebug & DEBUG_EXPIRE)
1238 				log(LOG_DEBUG,
1239 				    "expire_upcalls: expiring (%x %x)\n",
1240 				    ntohl(rt->mfc_origin.s_addr),
1241 				    ntohl(rt->mfc_mcastgrp.s_addr));
1242 
1243 			expire_mfc(rt);
1244 		}
1245 	}
1246 
1247 	splx(s);
1248 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1249 	    expire_upcalls, NULL);
1250 }
1251 
1252 /*
1253  * Packet forwarding routine once entry in the cache is made
1254  */
1255 static int
1256 #ifdef RSVP_ISI
1257 ip_mdq(m, ifp, rt, xmt_vif)
1258 #else
1259 ip_mdq(m, ifp, rt)
1260 #endif /* RSVP_ISI */
1261     struct mbuf *m;
1262     struct ifnet *ifp;
1263     struct mfc *rt;
1264 #ifdef RSVP_ISI
1265     vifi_t xmt_vif;
1266 #endif /* RSVP_ISI */
1267 {
1268     struct ip  *ip = mtod(m, struct ip *);
1269     vifi_t vifi;
1270     struct vif *vifp;
1271     int plen = ntohs(ip->ip_len);
1272 
1273 /*
1274  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1275  * input, they shouldn't get counted on output, so statistics keeping is
1276  * separate.
1277  */
1278 #define MC_SEND(ip,vifp,m) {                             \
1279                 if ((vifp)->v_flags & VIFF_TUNNEL)	 \
1280                     encap_send((ip), (vifp), (m));       \
1281                 else                                     \
1282                     phyint_send((ip), (vifp), (m));      \
1283 }
1284 
1285 #ifdef RSVP_ISI
1286     /*
1287      * If xmt_vif is not -1, send on only the requested vif.
1288      *
1289      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1290      */
1291     if (xmt_vif < numvifs) {
1292         MC_SEND(ip, viftable + xmt_vif, m);
1293 	return (1);
1294     }
1295 #endif /* RSVP_ISI */
1296 
1297     /*
1298      * Don't forward if it didn't arrive from the parent vif for its origin.
1299      */
1300     vifi = rt->mfc_parent;
1301     if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1302 	/* came in the wrong interface */
1303 	if (mrtdebug & DEBUG_FORWARD)
1304 	    log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1305 		ifp, vifi, viftable[vifi].v_ifp);
1306 	++mrtstat.mrts_wrong_if;
1307 	++rt->mfc_wrong_if;
1308 	/*
1309 	 * If we are doing PIM assert processing, and we are forwarding
1310 	 * packets on this interface, and it is a broadcast medium
1311 	 * interface (and not a tunnel), send a message to the routing daemon.
1312 	 */
1313 	if (pim_assert && rt->mfc_ttls[vifi] &&
1314 		(ifp->if_flags & IFF_BROADCAST) &&
1315 		!(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1316 	    struct mbuf *mm;
1317 	    struct igmpmsg *im;
1318 	    int hlen = ip->ip_hl << 2;
1319 	    struct timeval now;
1320 	    u_int32_t delta;
1321 
1322 	    microtime(&now);
1323 
1324 	    TV_DELTA(rt->mfc_last_assert, now, delta);
1325 
1326 	    if (delta > ASSERT_MSG_TIME) {
1327 		mm = m_copy(m, 0, hlen);
1328 		M_PULLUP(mm, hlen);
1329 		if (mm == 0) {
1330 		    return (ENOBUFS);
1331 		}
1332 
1333 		rt->mfc_last_assert = now;
1334 
1335 		im = mtod(mm, struct igmpmsg *);
1336 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1337 		im->im_mbz	= 0;
1338 		im->im_vif	= vifi;
1339 
1340 		sin.sin_addr = im->im_src;
1341 
1342 		socket_send(ip_mrouter, mm, &sin);
1343 	    }
1344 	}
1345 	return (0);
1346     }
1347 
1348     /* If I sourced this packet, it counts as output, else it was input. */
1349     if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1350 	viftable[vifi].v_pkt_out++;
1351 	viftable[vifi].v_bytes_out += plen;
1352     } else {
1353 	viftable[vifi].v_pkt_in++;
1354 	viftable[vifi].v_bytes_in += plen;
1355     }
1356     rt->mfc_pkt_cnt++;
1357     rt->mfc_byte_cnt += plen;
1358 
1359     /*
1360      * For each vif, decide if a copy of the packet should be forwarded.
1361      * Forward if:
1362      *		- the ttl exceeds the vif's threshold
1363      *		- there are group members downstream on interface
1364      */
1365     for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1366 	if ((rt->mfc_ttls[vifi] > 0) &&
1367 	    (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1368 	    vifp->v_pkt_out++;
1369 	    vifp->v_bytes_out += plen;
1370 	    MC_SEND(ip, vifp, m);
1371 	}
1372 
1373     return (0);
1374 }
1375 
1376 #ifdef RSVP_ISI
1377 /*
1378  * check if a vif number is legal/ok. This is used by ip_output, to export
1379  * numvifs there,
1380  */
1381 int
1382 legal_vif_num(vif)
1383     int vif;
1384 {
1385     if (vif >= 0 && vif < numvifs)
1386        return (1);
1387     else
1388        return (0);
1389 }
1390 #endif /* RSVP_ISI */
1391 
1392 static void
1393 phyint_send(ip, vifp, m)
1394 	struct ip *ip;
1395 	struct vif *vifp;
1396 	struct mbuf *m;
1397 {
1398 	struct mbuf *mb_copy;
1399 	int hlen = ip->ip_hl << 2;
1400 
1401 	/*
1402 	 * Make a new reference to the packet; make sure that
1403 	 * the IP header is actually copied, not just referenced,
1404 	 * so that ip_output() only scribbles on the copy.
1405 	 */
1406 	mb_copy = m_copy(m, 0, M_COPYALL);
1407 	M_PULLUP(mb_copy, hlen);
1408 	if (mb_copy == 0)
1409 		return;
1410 
1411 	if (vifp->v_rate_limit <= 0)
1412 		tbf_send_packet(vifp, mb_copy);
1413 	else
1414 		tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1415 }
1416 
1417 static void
1418 encap_send(ip, vifp, m)
1419 	struct ip *ip;
1420 	struct vif *vifp;
1421 	struct mbuf *m;
1422 {
1423 	struct mbuf *mb_copy;
1424 	struct ip *ip_copy;
1425 	int i, len = ip->ip_len + sizeof(multicast_encap_iphdr);
1426 
1427 	/*
1428 	 * copy the old packet & pullup it's IP header into the
1429 	 * new mbuf so we can modify it.  Try to fill the new
1430 	 * mbuf since if we don't the ethernet driver will.
1431 	 */
1432 	MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1433 	if (mb_copy == 0)
1434 		return;
1435 	mb_copy->m_data += max_linkhdr;
1436 	mb_copy->m_pkthdr.len = len;
1437 	mb_copy->m_len = sizeof(multicast_encap_iphdr);
1438 
1439 	if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == 0) {
1440 		m_freem(mb_copy);
1441 		return;
1442 	}
1443 	i = MHLEN - max_linkhdr;
1444 	if (i > len)
1445 		i = len;
1446 	mb_copy = m_pullup(mb_copy, i);
1447 	if (mb_copy == 0)
1448 		return;
1449 
1450 	/*
1451 	 * fill in the encapsulating IP header.
1452 	 */
1453 	ip_copy = mtod(mb_copy, struct ip *);
1454 	*ip_copy = multicast_encap_iphdr;
1455 	ip_copy->ip_id = htons(ip_id++);
1456 	ip_copy->ip_len = len;
1457 	ip_copy->ip_src = vifp->v_lcl_addr;
1458 	ip_copy->ip_dst = vifp->v_rmt_addr;
1459 
1460 	/*
1461 	 * turn the encapsulated IP header back into a valid one.
1462 	 */
1463 	ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1464 	--ip->ip_ttl;
1465 	HTONS(ip->ip_len);
1466 	HTONS(ip->ip_off);
1467 	ip->ip_sum = 0;
1468 	mb_copy->m_data += sizeof(multicast_encap_iphdr);
1469 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1470 	mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1471 
1472 	if (vifp->v_rate_limit <= 0)
1473 		tbf_send_packet(vifp, mb_copy);
1474 	else
1475 		tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1476 }
1477 
1478 /*
1479  * De-encapsulate a packet and feed it back through ip input.
1480  */
1481 static void
1482 #if __STDC__
1483 vif_input(struct mbuf *m, ...)
1484 #else
1485 vif_input(m, va_alist)
1486 	struct mbuf *m;
1487 	va_dcl
1488 #endif
1489 {
1490 	int off, proto;
1491 	va_list ap;
1492 	struct ip *ip;
1493 	struct vif *vifp;
1494 	int s;
1495 	struct ifqueue *ifq;
1496 
1497 	va_start(ap, m);
1498 	off = va_arg(ap, int);
1499 	proto = va_arg(ap, int);
1500 	va_end(ap);
1501 
1502 	vifp = (struct vif *)encap_getarg(m);
1503 	if (!vifp || proto != AF_INET) {
1504 		m_freem(m);
1505 		mrtstat.mrts_bad_tunnel++;
1506 		return;
1507 	}
1508 
1509 	ip = mtod(m, struct ip *);
1510 
1511 	m_adj(m, off);
1512 	m->m_pkthdr.rcvif = vifp->v_ifp;
1513 	ifq = &ipintrq;
1514 	s = splnet();
1515 	if (IF_QFULL(ifq)) {
1516 		IF_DROP(ifq);
1517 		m_freem(m);
1518 	} else {
1519 		IF_ENQUEUE(ifq, m);
1520 		/*
1521 		 * normally we would need a "schednetisr(NETISR_IP)"
1522 		 * here but we were called by ip_input and it is going
1523 		 * to loop back & try to dequeue the packet we just
1524 		 * queued as soon as we return so we avoid the
1525 		 * unnecessary software interrrupt.
1526 		 */
1527 	}
1528 	splx(s);
1529 }
1530 
1531 /*
1532  * Check if the packet should be grabbed by us.
1533  */
1534 static int
1535 vif_encapcheck(m, off, proto, arg)
1536 	const struct mbuf *m;
1537 	int off;
1538 	int proto;
1539 	void *arg;
1540 {
1541 	struct vif *vifp;
1542 	struct ip ip;
1543 
1544 #ifdef DIAGNOSTIC
1545 	if (!arg || proto != IPPROTO_IPV4)
1546 		panic("unexpected arg in vif_encapcheck");
1547 #endif
1548 
1549 	/*
1550 	 * do not grab the packet if it's not to a multicast destination or if
1551 	 * we don't have an encapsulating tunnel with the source.
1552 	 * Note:  This code assumes that the remote site IP address
1553 	 * uniquely identifies the tunnel (i.e., that this site has
1554 	 * at most one tunnel with the remote site).
1555 	 */
1556 
1557 	/* LINTED const cast */
1558 	m_copydata((struct mbuf *)m, off, sizeof(ip), (caddr_t)&ip);
1559 	if (!IN_MULTICAST(ip.ip_dst.s_addr))
1560 		return 0;
1561 
1562 	/* LINTED const cast */
1563 	m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip);
1564 	if (!in_hosteq(ip.ip_src, last_encap_src)) {
1565 		vifp = (struct vif *)arg;
1566 		if (vifp->v_flags & VIFF_TUNNEL &&
1567 		    in_hosteq(vifp->v_rmt_addr, ip.ip_src))
1568 			;
1569 		else
1570 			return 0;
1571 		last_encap_vif = vifp;
1572 		last_encap_src = ip.ip_src;
1573 	} else
1574 		vifp = last_encap_vif;
1575 
1576 	/* 32bit match, since we have checked ip_src only */
1577 	return 32;
1578 }
1579 
1580 /*
1581  * Token bucket filter module
1582  */
1583 static void
1584 tbf_control(vifp, m, ip, len)
1585 	struct vif *vifp;
1586 	struct mbuf *m;
1587 	struct ip *ip;
1588 	u_int32_t len;
1589 {
1590 
1591 	if (len > MAX_BKT_SIZE) {
1592 		/* drop if packet is too large */
1593 		mrtstat.mrts_pkt2large++;
1594 		m_freem(m);
1595 		return;
1596 	}
1597 
1598 	tbf_update_tokens(vifp);
1599 
1600 	/*
1601 	 * If there are enough tokens, and the queue is empty, send this packet
1602 	 * out immediately.  Otherwise, try to insert it on this vif's queue.
1603 	 */
1604 	if (vifp->tbf_q_len == 0) {
1605 		if (len <= vifp->tbf_n_tok) {
1606 			vifp->tbf_n_tok -= len;
1607 			tbf_send_packet(vifp, m);
1608 		} else {
1609 			/* queue packet and timeout till later */
1610 			tbf_queue(vifp, m);
1611 			callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
1612 			    tbf_reprocess_q, vifp);
1613 		}
1614 	} else {
1615 		if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
1616 		    !tbf_dq_sel(vifp, ip)) {
1617 			/* queue length too much, and couldn't make room */
1618 			mrtstat.mrts_q_overflow++;
1619 			m_freem(m);
1620 		} else {
1621 			/* queue length low enough, or made room */
1622 			tbf_queue(vifp, m);
1623 			tbf_process_q(vifp);
1624 		}
1625 	}
1626 }
1627 
1628 /*
1629  * adds a packet to the queue at the interface
1630  */
1631 static void
1632 tbf_queue(vifp, m)
1633 	struct vif *vifp;
1634 	struct mbuf *m;
1635 {
1636 	int s = splsoftnet();
1637 
1638 	/* insert at tail */
1639 	*vifp->tbf_t = m;
1640 	vifp->tbf_t = &m->m_nextpkt;
1641 	vifp->tbf_q_len++;
1642 
1643 	splx(s);
1644 }
1645 
1646 
1647 /*
1648  * processes the queue at the interface
1649  */
1650 static void
1651 tbf_process_q(vifp)
1652 	struct vif *vifp;
1653 {
1654 	struct mbuf *m;
1655 	int len;
1656 	int s = splsoftnet();
1657 
1658 	/*
1659 	 * Loop through the queue at the interface and send as many packets
1660 	 * as possible.
1661 	 */
1662 	for (m = vifp->tbf_q;
1663 	    m != 0;
1664 	    m = vifp->tbf_q) {
1665 		len = mtod(m, struct ip *)->ip_len;
1666 
1667 		/* determine if the packet can be sent */
1668 		if (len <= vifp->tbf_n_tok) {
1669 			/* if so,
1670 			 * reduce no of tokens, dequeue the packet,
1671 			 * send the packet.
1672 			 */
1673 			if ((vifp->tbf_q = m->m_nextpkt) == 0)
1674 				vifp->tbf_t = &vifp->tbf_q;
1675 			--vifp->tbf_q_len;
1676 
1677 			m->m_nextpkt = 0;
1678 			vifp->tbf_n_tok -= len;
1679 			tbf_send_packet(vifp, m);
1680 		} else
1681 			break;
1682 	}
1683 	splx(s);
1684 }
1685 
1686 static void
1687 tbf_reprocess_q(arg)
1688 	void *arg;
1689 {
1690 	struct vif *vifp = arg;
1691 
1692 	if (ip_mrouter == 0)
1693 		return;
1694 
1695 	tbf_update_tokens(vifp);
1696 	tbf_process_q(vifp);
1697 
1698 	if (vifp->tbf_q_len != 0)
1699 		callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
1700 		    tbf_reprocess_q, vifp);
1701 }
1702 
1703 /* function that will selectively discard a member of the queue
1704  * based on the precedence value and the priority
1705  */
1706 static int
1707 tbf_dq_sel(vifp, ip)
1708 	struct vif *vifp;
1709 	struct ip *ip;
1710 {
1711 	u_int p;
1712 	struct mbuf **mp, *m;
1713 	int s = splsoftnet();
1714 
1715 	p = priority(vifp, ip);
1716 
1717 	for (mp = &vifp->tbf_q, m = *mp;
1718 	    m != 0;
1719 	    mp = &m->m_nextpkt, m = *mp) {
1720 		if (p > priority(vifp, mtod(m, struct ip *))) {
1721 			if ((*mp = m->m_nextpkt) == 0)
1722 				vifp->tbf_t = mp;
1723 			--vifp->tbf_q_len;
1724 
1725 			m_freem(m);
1726 			mrtstat.mrts_drop_sel++;
1727 			splx(s);
1728 			return (1);
1729 		}
1730 	}
1731 	splx(s);
1732 	return (0);
1733 }
1734 
1735 static void
1736 tbf_send_packet(vifp, m)
1737 	struct vif *vifp;
1738 	struct mbuf *m;
1739 {
1740 	int error;
1741 	int s = splsoftnet();
1742 
1743 	if (vifp->v_flags & VIFF_TUNNEL) {
1744 		/* If tunnel options */
1745 #ifdef IPSEC
1746 		/* Don't lookup socket in forwading case */
1747 		(void)ipsec_setsocket(m, NULL);
1748 #endif
1749 		ip_output(m, (struct mbuf *)0, &vifp->v_route,
1750 			  IP_FORWARDING, (struct ip_moptions *)0);
1751 	} else {
1752 		/* if physical interface option, extract the options and then send */
1753 		struct ip_moptions imo;
1754 
1755 		imo.imo_multicast_ifp = vifp->v_ifp;
1756 		imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1757 		imo.imo_multicast_loop = 1;
1758 #ifdef RSVP_ISI
1759 		imo.imo_multicast_vif = -1;
1760 #endif
1761 
1762 #ifdef IPSEC
1763 		/* Don't lookup socket in forwading case */
1764 		(void)ipsec_setsocket(m, NULL);
1765 #endif
1766 		error = ip_output(m, (struct mbuf *)0, (struct route *)0,
1767 				  IP_FORWARDING|IP_MULTICASTOPTS, &imo);
1768 
1769 		if (mrtdebug & DEBUG_XMIT)
1770 			log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
1771 			    (long)(vifp-viftable), error);
1772 	}
1773 	splx(s);
1774 }
1775 
1776 /* determine the current time and then
1777  * the elapsed time (between the last time and time now)
1778  * in milliseconds & update the no. of tokens in the bucket
1779  */
1780 static void
1781 tbf_update_tokens(vifp)
1782 	struct vif *vifp;
1783 {
1784 	struct timeval tp;
1785 	u_int32_t tm;
1786 	int s = splsoftnet();
1787 
1788 	microtime(&tp);
1789 
1790 	TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
1791 
1792 	/*
1793 	 * This formula is actually
1794 	 * "time in seconds" * "bytes/second".
1795 	 *
1796 	 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
1797 	 *
1798 	 * The (1000/1024) was introduced in add_vif to optimize
1799 	 * this divide into a shift.
1800 	 */
1801 	vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
1802 	vifp->tbf_last_pkt_t = tp;
1803 
1804 	if (vifp->tbf_n_tok > MAX_BKT_SIZE)
1805 		vifp->tbf_n_tok = MAX_BKT_SIZE;
1806 
1807 	splx(s);
1808 }
1809 
1810 static int
1811 priority(vifp, ip)
1812     struct vif *vifp;
1813     struct ip *ip;
1814 {
1815     int prio;
1816 
1817     /* temporary hack; may add general packet classifier some day */
1818 
1819     /*
1820      * The UDP port space is divided up into four priority ranges:
1821      * [0, 16384)     : unclassified - lowest priority
1822      * [16384, 32768) : audio - highest priority
1823      * [32768, 49152) : whiteboard - medium priority
1824      * [49152, 65536) : video - low priority
1825      */
1826     if (ip->ip_p == IPPROTO_UDP) {
1827 	struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
1828 
1829 	switch (ntohs(udp->uh_dport) & 0xc000) {
1830 	    case 0x4000:
1831 		prio = 70;
1832 		break;
1833 	    case 0x8000:
1834 		prio = 60;
1835 		break;
1836 	    case 0xc000:
1837 		prio = 55;
1838 		break;
1839 	    default:
1840 		prio = 50;
1841 		break;
1842 	}
1843 
1844 	if (tbfdebug > 1)
1845 	    log(LOG_DEBUG, "port %x prio %d\n", ntohs(udp->uh_dport), prio);
1846     } else
1847 	prio = 50;
1848 
1849 
1850     return (prio);
1851 }
1852 
1853 /*
1854  * End of token bucket filter modifications
1855  */
1856 
1857 #ifdef RSVP_ISI
1858 
1859 int
1860 ip_rsvp_vif_init(so, m)
1861     struct socket *so;
1862     struct mbuf *m;
1863 {
1864     int i;
1865     int s;
1866 
1867     if (rsvpdebug)
1868 	printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
1869 	    so->so_type, so->so_proto->pr_protocol);
1870 
1871     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
1872 	return (EOPNOTSUPP);
1873 
1874     /* Check mbuf. */
1875     if (m == 0 || m->m_len != sizeof(int)) {
1876 	return (EINVAL);
1877     }
1878     i = *(mtod(m, int *));
1879 
1880     if (rsvpdebug)
1881 	printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on);
1882 
1883     s = splsoftnet();
1884 
1885     /* Check vif. */
1886     if (!legal_vif_num(i)) {
1887 	splx(s);
1888 	return (EADDRNOTAVAIL);
1889     }
1890 
1891     /* Check if socket is available. */
1892     if (viftable[i].v_rsvpd != 0) {
1893 	splx(s);
1894 	return (EADDRINUSE);
1895     }
1896 
1897     viftable[i].v_rsvpd = so;
1898     /* This may seem silly, but we need to be sure we don't over-increment
1899      * the RSVP counter, in case something slips up.
1900      */
1901     if (!viftable[i].v_rsvp_on) {
1902 	viftable[i].v_rsvp_on = 1;
1903 	rsvp_on++;
1904     }
1905 
1906     splx(s);
1907     return (0);
1908 }
1909 
1910 int
1911 ip_rsvp_vif_done(so, m)
1912     struct socket *so;
1913     struct mbuf *m;
1914 {
1915     int i;
1916     int s;
1917 
1918     if (rsvpdebug)
1919 	printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
1920 	       so->so_type, so->so_proto->pr_protocol);
1921 
1922     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
1923 	return (EOPNOTSUPP);
1924 
1925     /* Check mbuf. */
1926     if (m == 0 || m->m_len != sizeof(int)) {
1927 	return (EINVAL);
1928     }
1929     i = *(mtod(m, int *));
1930 
1931     s = splsoftnet();
1932 
1933     /* Check vif. */
1934     if (!legal_vif_num(i)) {
1935 	splx(s);
1936         return (EADDRNOTAVAIL);
1937     }
1938 
1939     if (rsvpdebug)
1940 	printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
1941 	    viftable[i].v_rsvpd, so);
1942 
1943     viftable[i].v_rsvpd = 0;
1944     /* This may seem silly, but we need to be sure we don't over-decrement
1945      * the RSVP counter, in case something slips up.
1946      */
1947     if (viftable[i].v_rsvp_on) {
1948 	viftable[i].v_rsvp_on = 0;
1949 	rsvp_on--;
1950     }
1951 
1952     splx(s);
1953     return (0);
1954 }
1955 
1956 void
1957 ip_rsvp_force_done(so)
1958     struct socket *so;
1959 {
1960     int vifi;
1961     int s;
1962 
1963     /* Don't bother if it is not the right type of socket. */
1964     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
1965 	return;
1966 
1967     s = splsoftnet();
1968 
1969     /* The socket may be attached to more than one vif...this
1970      * is perfectly legal.
1971      */
1972     for (vifi = 0; vifi < numvifs; vifi++) {
1973 	if (viftable[vifi].v_rsvpd == so) {
1974 	    viftable[vifi].v_rsvpd = 0;
1975 	    /* This may seem silly, but we need to be sure we don't
1976 	     * over-decrement the RSVP counter, in case something slips up.
1977 	     */
1978 	    if (viftable[vifi].v_rsvp_on) {
1979 		viftable[vifi].v_rsvp_on = 0;
1980 		rsvp_on--;
1981 	    }
1982 	}
1983     }
1984 
1985     splx(s);
1986     return;
1987 }
1988 
1989 void
1990 rsvp_input(m, ifp)
1991     struct mbuf *m;
1992     struct ifnet *ifp;
1993 {
1994     int vifi;
1995     struct ip *ip = mtod(m, struct ip *);
1996     static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET };
1997     int s;
1998 
1999     if (rsvpdebug)
2000 	printf("rsvp_input: rsvp_on %d\n",rsvp_on);
2001 
2002     /* Can still get packets with rsvp_on = 0 if there is a local member
2003      * of the group to which the RSVP packet is addressed.  But in this
2004      * case we want to throw the packet away.
2005      */
2006     if (!rsvp_on) {
2007 	m_freem(m);
2008 	return;
2009     }
2010 
2011     /* If the old-style non-vif-associated socket is set, then use
2012      * it and ignore the new ones.
2013      */
2014     if (ip_rsvpd != 0) {
2015 	if (rsvpdebug)
2016 	    printf("rsvp_input: Sending packet up old-style socket\n");
2017 	rip_input(m);	/*XXX*/
2018 	return;
2019     }
2020 
2021     s = splsoftnet();
2022 
2023     if (rsvpdebug)
2024 	printf("rsvp_input: check vifs\n");
2025 
2026     /* Find which vif the packet arrived on. */
2027     for (vifi = 0; vifi < numvifs; vifi++) {
2028 	if (viftable[vifi].v_ifp == ifp)
2029 	    break;
2030     }
2031 
2032     if (vifi == numvifs) {
2033 	/* Can't find vif packet arrived on. Drop packet. */
2034 	if (rsvpdebug)
2035 	    printf("rsvp_input: Can't find vif for packet...dropping it.\n");
2036 	m_freem(m);
2037 	splx(s);
2038 	return;
2039     }
2040 
2041     if (rsvpdebug)
2042 	printf("rsvp_input: check socket\n");
2043 
2044     if (viftable[vifi].v_rsvpd == 0) {
2045 	/* drop packet, since there is no specific socket for this
2046 	 * interface */
2047 	if (rsvpdebug)
2048 	    printf("rsvp_input: No socket defined for vif %d\n",vifi);
2049 	m_freem(m);
2050 	splx(s);
2051 	return;
2052     }
2053 
2054     rsvp_src.sin_addr = ip->ip_src;
2055 
2056     if (rsvpdebug && m)
2057 	printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
2058 	       m->m_len,sbspace(&viftable[vifi].v_rsvpd->so_rcv));
2059 
2060     if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2061 	if (rsvpdebug)
2062 	    printf("rsvp_input: Failed to append to socket\n");
2063     else
2064 	if (rsvpdebug)
2065 	    printf("rsvp_input: send packet up\n");
2066 
2067     splx(s);
2068 }
2069 #endif /* RSVP_ISI */
2070