xref: /netbsd-src/sys/netinet/ip_mroute.c (revision bf1e9b32e27832f0c493206710fb8b58a980838a)
1 /*	$NetBSD: ip_mroute.c,v 1.94 2005/06/06 06:06:50 martin Exp $	*/
2 
3 /*
4  * Copyright (c) 1992, 1993
5  *      The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Stephen Deering of Stanford University.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
35  */
36 
37 /*
38  * Copyright (c) 1989 Stephen Deering
39  *
40  * This code is derived from software contributed to Berkeley by
41  * Stephen Deering of Stanford University.
42  *
43  * Redistribution and use in source and binary forms, with or without
44  * modification, are permitted provided that the following conditions
45  * are met:
46  * 1. Redistributions of source code must retain the above copyright
47  *    notice, this list of conditions and the following disclaimer.
48  * 2. Redistributions in binary form must reproduce the above copyright
49  *    notice, this list of conditions and the following disclaimer in the
50  *    documentation and/or other materials provided with the distribution.
51  * 3. All advertising materials mentioning features or use of this software
52  *    must display the following acknowledgement:
53  *      This product includes software developed by the University of
54  *      California, Berkeley and its contributors.
55  * 4. Neither the name of the University nor the names of its contributors
56  *    may be used to endorse or promote products derived from this software
57  *    without specific prior written permission.
58  *
59  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
60  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
61  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
62  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
63  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
64  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
65  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
66  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
67  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
68  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
69  * SUCH DAMAGE.
70  *
71  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
72  */
73 
74 /*
75  * IP multicast forwarding procedures
76  *
77  * Written by David Waitzman, BBN Labs, August 1988.
78  * Modified by Steve Deering, Stanford, February 1989.
79  * Modified by Mark J. Steiglitz, Stanford, May, 1991
80  * Modified by Van Jacobson, LBL, January 1993
81  * Modified by Ajit Thyagarajan, PARC, August 1993
82  * Modified by Bill Fenner, PARC, April 1994
83  * Modified by Charles M. Hannum, NetBSD, May 1995.
84  * Modified by Ahmed Helmy, SGI, June 1996
85  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
86  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
87  * Modified by Hitoshi Asaeda, WIDE, August 2000
88  * Modified by Pavlin Radoslavov, ICSI, October 2002
89  *
90  * MROUTING Revision: 1.2
91  * and PIM-SMv2 and PIM-DM support, advanced API support,
92  * bandwidth metering and signaling
93  */
94 
95 #include <sys/cdefs.h>
96 __KERNEL_RCSID(0, "$NetBSD: ip_mroute.c,v 1.94 2005/06/06 06:06:50 martin Exp $");
97 
98 #include "opt_inet.h"
99 #include "opt_ipsec.h"
100 #include "opt_pim.h"
101 
102 #ifdef PIM
103 #define _PIM_VT 1
104 #endif
105 
106 #include <sys/param.h>
107 #include <sys/systm.h>
108 #include <sys/callout.h>
109 #include <sys/mbuf.h>
110 #include <sys/socket.h>
111 #include <sys/socketvar.h>
112 #include <sys/protosw.h>
113 #include <sys/errno.h>
114 #include <sys/time.h>
115 #include <sys/kernel.h>
116 #include <sys/ioctl.h>
117 #include <sys/syslog.h>
118 
119 #include <net/if.h>
120 #include <net/route.h>
121 #include <net/raw_cb.h>
122 
123 #include <netinet/in.h>
124 #include <netinet/in_var.h>
125 #include <netinet/in_systm.h>
126 #include <netinet/ip.h>
127 #include <netinet/ip_var.h>
128 #include <netinet/in_pcb.h>
129 #include <netinet/udp.h>
130 #include <netinet/igmp.h>
131 #include <netinet/igmp_var.h>
132 #include <netinet/ip_mroute.h>
133 #ifdef PIM
134 #include <netinet/pim.h>
135 #include <netinet/pim_var.h>
136 #endif
137 #include <netinet/ip_encap.h>
138 
139 #ifdef IPSEC
140 #include <netinet6/ipsec.h>
141 #include <netkey/key.h>
142 #endif
143 
144 #ifdef FAST_IPSEC
145 #include <netipsec/ipsec.h>
146 #include <netipsec/key.h>
147 #endif
148 
149 #include <machine/stdarg.h>
150 
151 #define IP_MULTICASTOPTS 0
152 #define	M_PULLUP(m, len)						 \
153 	do {								 \
154 		if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
155 			(m) = m_pullup((m), (len));			 \
156 	} while (/*CONSTCOND*/ 0)
157 
158 /*
159  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
160  * except for netstat or debugging purposes.
161  */
162 struct socket  *ip_mrouter  = NULL;
163 int		ip_mrtproto = IGMP_DVMRP;    /* for netstat only */
164 
165 #define NO_RTE_FOUND 	0x1
166 #define RTE_FOUND	0x2
167 
168 #define	MFCHASH(a, g)							\
169 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^	\
170 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
171 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
172 u_long	mfchash;
173 
174 u_char		nexpire[MFCTBLSIZ];
175 struct vif	viftable[MAXVIFS];
176 struct mrtstat	mrtstat;
177 u_int		mrtdebug = 0;	  /* debug level 	*/
178 #define		DEBUG_MFC	0x02
179 #define		DEBUG_FORWARD	0x04
180 #define		DEBUG_EXPIRE	0x08
181 #define		DEBUG_XMIT	0x10
182 #define		DEBUG_PIM	0x20
183 
184 #define		VIFI_INVALID	((vifi_t) -1)
185 
186 u_int       	tbfdebug = 0;     /* tbf debug level 	*/
187 #ifdef RSVP_ISI
188 u_int		rsvpdebug = 0;	  /* rsvp debug level   */
189 extern struct socket *ip_rsvpd;
190 extern int rsvp_on;
191 #endif /* RSVP_ISI */
192 
193 /* vif attachment using sys/netinet/ip_encap.c */
194 static void vif_input(struct mbuf *, ...);
195 static int vif_encapcheck(struct mbuf *, int, int, void *);
196 
197 static const struct protosw vif_protosw =
198 { SOCK_RAW,	&inetdomain,	IPPROTO_IPV4,	PR_ATOMIC|PR_ADDR,
199   vif_input,	rip_output,	0,		rip_ctloutput,
200   rip_usrreq,
201   0,            0,              0,              0,
202 };
203 
204 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
205 #define		UPCALL_EXPIRE	6		/* number of timeouts */
206 
207 /*
208  * Define the token bucket filter structures
209  */
210 
211 #define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
212 
213 static int get_sg_cnt(struct sioc_sg_req *);
214 static int get_vif_cnt(struct sioc_vif_req *);
215 static int ip_mrouter_init(struct socket *, struct mbuf *);
216 static int get_version(struct mbuf *);
217 static int set_assert(struct mbuf *);
218 static int get_assert(struct mbuf *);
219 static int add_vif(struct mbuf *);
220 static int del_vif(struct mbuf *);
221 static void update_mfc_params(struct mfc *, struct mfcctl2 *);
222 static void init_mfc_params(struct mfc *, struct mfcctl2 *);
223 static void expire_mfc(struct mfc *);
224 static int add_mfc(struct mbuf *);
225 #ifdef UPCALL_TIMING
226 static void collate(struct timeval *);
227 #endif
228 static int del_mfc(struct mbuf *);
229 static int set_api_config(struct mbuf *); /* chose API capabilities */
230 static int get_api_support(struct mbuf *);
231 static int get_api_config(struct mbuf *);
232 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
233 static void expire_upcalls(void *);
234 #ifdef RSVP_ISI
235 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
236 #else
237 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *);
238 #endif
239 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
240 static void encap_send(struct ip *, struct vif *, struct mbuf *);
241 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_int32_t);
242 static void tbf_queue(struct vif *, struct mbuf *);
243 static void tbf_process_q(struct vif *);
244 static void tbf_reprocess_q(void *);
245 static int tbf_dq_sel(struct vif *, struct ip *);
246 static void tbf_send_packet(struct vif *, struct mbuf *);
247 static void tbf_update_tokens(struct vif *);
248 static int priority(struct vif *, struct ip *);
249 
250 /*
251  * Bandwidth monitoring
252  */
253 static void free_bw_list(struct bw_meter *);
254 static int add_bw_upcall(struct mbuf *);
255 static int del_bw_upcall(struct mbuf *);
256 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *);
257 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
258 static void bw_upcalls_send(void);
259 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
260 static void unschedule_bw_meter(struct bw_meter *);
261 static void bw_meter_process(void);
262 static void expire_bw_upcalls_send(void *);
263 static void expire_bw_meter_process(void *);
264 
265 #ifdef PIM
266 static int pim_register_send(struct ip *, struct vif *,
267 		struct mbuf *, struct mfc *);
268 static int pim_register_send_rp(struct ip *, struct vif *,
269 		struct mbuf *, struct mfc *);
270 static int pim_register_send_upcall(struct ip *, struct vif *,
271 		struct mbuf *, struct mfc *);
272 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
273 #endif
274 
275 /*
276  * 'Interfaces' associated with decapsulator (so we can tell
277  * packets that went through it from ones that get reflected
278  * by a broken gateway).  These interfaces are never linked into
279  * the system ifnet list & no routes point to them.  I.e., packets
280  * can't be sent this way.  They only exist as a placeholder for
281  * multicast source verification.
282  */
283 #if 0
284 struct ifnet multicast_decap_if[MAXVIFS];
285 #endif
286 
287 #define	ENCAP_TTL	64
288 #define	ENCAP_PROTO	IPPROTO_IPIP	/* 4 */
289 
290 /* prototype IP hdr for encapsulated packets */
291 struct ip multicast_encap_iphdr = {
292 #if BYTE_ORDER == LITTLE_ENDIAN
293 	sizeof(struct ip) >> 2, IPVERSION,
294 #else
295 	IPVERSION, sizeof(struct ip) >> 2,
296 #endif
297 	0,				/* tos */
298 	sizeof(struct ip),		/* total length */
299 	0,				/* id */
300 	0,				/* frag offset */
301 	ENCAP_TTL, ENCAP_PROTO,
302 	0,				/* checksum */
303 };
304 
305 /*
306  * Bandwidth meter variables and constants
307  */
308 
309 /*
310  * Pending timeouts are stored in a hash table, the key being the
311  * expiration time. Periodically, the entries are analysed and processed.
312  */
313 #define BW_METER_BUCKETS	1024
314 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
315 struct callout bw_meter_ch;
316 #define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
317 
318 /*
319  * Pending upcalls are stored in a vector which is flushed when
320  * full, or periodically
321  */
322 static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
323 static u_int	bw_upcalls_n; /* # of pending upcalls */
324 struct callout	bw_upcalls_ch;
325 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
326 
327 #ifdef PIM
328 struct pimstat pimstat;
329 
330 /*
331  * Note: the PIM Register encapsulation adds the following in front of a
332  * data packet:
333  *
334  * struct pim_encap_hdr {
335  *    struct ip ip;
336  *    struct pim_encap_pimhdr  pim;
337  * }
338  *
339  */
340 
341 struct pim_encap_pimhdr {
342 	struct pim pim;
343 	uint32_t   flags;
344 };
345 
346 static struct ip pim_encap_iphdr = {
347 #if BYTE_ORDER == LITTLE_ENDIAN
348 	sizeof(struct ip) >> 2,
349 	IPVERSION,
350 #else
351 	IPVERSION,
352 	sizeof(struct ip) >> 2,
353 #endif
354 	0,			/* tos */
355 	sizeof(struct ip),	/* total length */
356 	0,			/* id */
357 	0,			/* frag offset */
358 	ENCAP_TTL,
359 	IPPROTO_PIM,
360 	0,			/* checksum */
361 };
362 
363 static struct pim_encap_pimhdr pim_encap_pimhdr = {
364     {
365 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
366 	0,			/* reserved */
367 	0,			/* checksum */
368     },
369     0				/* flags */
370 };
371 
372 static struct ifnet multicast_register_if;
373 static vifi_t reg_vif_num = VIFI_INVALID;
374 #endif /* PIM */
375 
376 
377 /*
378  * Private variables.
379  */
380 static vifi_t	   numvifs = 0;
381 
382 static struct callout expire_upcalls_ch;
383 
384 /*
385  * one-back cache used by vif_encapcheck to locate a tunnel's vif
386  * given a datagram's src ip address.
387  */
388 static struct in_addr last_encap_src;
389 static struct vif *last_encap_vif;
390 
391 /*
392  * whether or not special PIM assert processing is enabled.
393  */
394 static int pim_assert;
395 /*
396  * Rate limit for assert notification messages, in usec
397  */
398 #define ASSERT_MSG_TIME		3000000
399 
400 /*
401  * Kernel multicast routing API capabilities and setup.
402  * If more API capabilities are added to the kernel, they should be
403  * recorded in `mrt_api_support'.
404  */
405 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
406 					  MRT_MFC_FLAGS_BORDER_VIF |
407 					  MRT_MFC_RP |
408 					  MRT_MFC_BW_UPCALL);
409 static u_int32_t mrt_api_config = 0;
410 
411 /*
412  * Find a route for a given origin IP address and Multicast group address
413  * Type of service parameter to be added in the future!!!
414  * Statistics are updated by the caller if needed
415  * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
416  */
417 static struct mfc *
418 mfc_find(struct in_addr *o, struct in_addr *g)
419 {
420 	struct mfc *rt;
421 
422 	LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
423 		if (in_hosteq(rt->mfc_origin, *o) &&
424 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
425 		    (rt->mfc_stall == NULL))
426 			break;
427 	}
428 
429 	return (rt);
430 }
431 
432 /*
433  * Macros to compute elapsed time efficiently
434  * Borrowed from Van Jacobson's scheduling code
435  */
436 #define TV_DELTA(a, b, delta) do {					\
437 	int xxs;							\
438 	delta = (a).tv_usec - (b).tv_usec;				\
439 	xxs = (a).tv_sec - (b).tv_sec;					\
440 	switch (xxs) {							\
441 	case 2:								\
442 		delta += 1000000;					\
443 		/* fall through */					\
444 	case 1:								\
445 		delta += 1000000;					\
446 		/* fall through */					\
447 	case 0:								\
448 		break;							\
449 	default:							\
450 		delta += (1000000 * xxs);				\
451 		break;							\
452 	}								\
453 } while (/*CONSTCOND*/ 0)
454 
455 #ifdef UPCALL_TIMING
456 u_int32_t upcall_data[51];
457 #endif /* UPCALL_TIMING */
458 
459 /*
460  * Handle MRT setsockopt commands to modify the multicast routing tables.
461  */
462 int
463 ip_mrouter_set(struct socket *so, int optname, struct mbuf **m)
464 {
465 	int error;
466 
467 	if (optname != MRT_INIT && so != ip_mrouter)
468 		error = ENOPROTOOPT;
469 	else
470 		switch (optname) {
471 		case MRT_INIT:
472 			error = ip_mrouter_init(so, *m);
473 			break;
474 		case MRT_DONE:
475 			error = ip_mrouter_done();
476 			break;
477 		case MRT_ADD_VIF:
478 			error = add_vif(*m);
479 			break;
480 		case MRT_DEL_VIF:
481 			error = del_vif(*m);
482 			break;
483 		case MRT_ADD_MFC:
484 			error = add_mfc(*m);
485 			break;
486 		case MRT_DEL_MFC:
487 			error = del_mfc(*m);
488 			break;
489 		case MRT_ASSERT:
490 			error = set_assert(*m);
491 			break;
492 		case MRT_API_CONFIG:
493 			error = set_api_config(*m);
494 			break;
495 		case MRT_ADD_BW_UPCALL:
496 			error = add_bw_upcall(*m);
497 			break;
498 		case MRT_DEL_BW_UPCALL:
499 			error = del_bw_upcall(*m);
500 			break;
501 		default:
502 			error = ENOPROTOOPT;
503 			break;
504 		}
505 
506 	if (*m)
507 		m_free(*m);
508 	return (error);
509 }
510 
511 /*
512  * Handle MRT getsockopt commands
513  */
514 int
515 ip_mrouter_get(struct socket *so, int optname, struct mbuf **m)
516 {
517 	int error;
518 
519 	if (so != ip_mrouter)
520 		error = ENOPROTOOPT;
521 	else {
522 		*m = m_get(M_WAIT, MT_SOOPTS);
523 		MCLAIM(*m, so->so_mowner);
524 
525 		switch (optname) {
526 		case MRT_VERSION:
527 			error = get_version(*m);
528 			break;
529 		case MRT_ASSERT:
530 			error = get_assert(*m);
531 			break;
532 		case MRT_API_SUPPORT:
533 			error = get_api_support(*m);
534 			break;
535 		case MRT_API_CONFIG:
536 			error = get_api_config(*m);
537 			break;
538 		default:
539 			error = ENOPROTOOPT;
540 			break;
541 		}
542 
543 		if (error)
544 			m_free(*m);
545 	}
546 
547 	return (error);
548 }
549 
550 /*
551  * Handle ioctl commands to obtain information from the cache
552  */
553 int
554 mrt_ioctl(struct socket *so, u_long cmd, caddr_t data)
555 {
556 	int error;
557 
558 	if (so != ip_mrouter)
559 		error = EINVAL;
560 	else
561 		switch (cmd) {
562 		case SIOCGETVIFCNT:
563 			error = get_vif_cnt((struct sioc_vif_req *)data);
564 			break;
565 		case SIOCGETSGCNT:
566 			error = get_sg_cnt((struct sioc_sg_req *)data);
567 			break;
568 		default:
569 			error = EINVAL;
570 			break;
571 		}
572 
573 	return (error);
574 }
575 
576 /*
577  * returns the packet, byte, rpf-failure count for the source group provided
578  */
579 static int
580 get_sg_cnt(struct sioc_sg_req *req)
581 {
582 	int s;
583 	struct mfc *rt;
584 
585 	s = splsoftnet();
586 	rt = mfc_find(&req->src, &req->grp);
587 	if (rt == NULL) {
588 		splx(s);
589 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
590 		return (EADDRNOTAVAIL);
591 	}
592 	req->pktcnt = rt->mfc_pkt_cnt;
593 	req->bytecnt = rt->mfc_byte_cnt;
594 	req->wrong_if = rt->mfc_wrong_if;
595 	splx(s);
596 
597 	return (0);
598 }
599 
600 /*
601  * returns the input and output packet and byte counts on the vif provided
602  */
603 static int
604 get_vif_cnt(struct sioc_vif_req *req)
605 {
606 	vifi_t vifi = req->vifi;
607 
608 	if (vifi >= numvifs)
609 		return (EINVAL);
610 
611 	req->icount = viftable[vifi].v_pkt_in;
612 	req->ocount = viftable[vifi].v_pkt_out;
613 	req->ibytes = viftable[vifi].v_bytes_in;
614 	req->obytes = viftable[vifi].v_bytes_out;
615 
616 	return (0);
617 }
618 
619 /*
620  * Enable multicast routing
621  */
622 static int
623 ip_mrouter_init(struct socket *so, struct mbuf *m)
624 {
625 	int *v;
626 
627 	if (mrtdebug)
628 		log(LOG_DEBUG,
629 		    "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
630 		    so->so_type, so->so_proto->pr_protocol);
631 
632 	if (so->so_type != SOCK_RAW ||
633 	    so->so_proto->pr_protocol != IPPROTO_IGMP)
634 		return (EOPNOTSUPP);
635 
636 	if (m == NULL || m->m_len < sizeof(int))
637 		return (EINVAL);
638 
639 	v = mtod(m, int *);
640 	if (*v != 1)
641 		return (EINVAL);
642 
643 	if (ip_mrouter != NULL)
644 		return (EADDRINUSE);
645 
646 	ip_mrouter = so;
647 
648 	mfchashtbl =
649 	    hashinit(MFCTBLSIZ, HASH_LIST, M_MRTABLE, M_WAITOK, &mfchash);
650 	bzero((caddr_t)nexpire, sizeof(nexpire));
651 
652 	pim_assert = 0;
653 
654 	callout_init(&expire_upcalls_ch);
655 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
656 		      expire_upcalls, NULL);
657 
658 	callout_init(&bw_upcalls_ch);
659 	callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
660 		      expire_bw_upcalls_send, NULL);
661 
662 	callout_init(&bw_meter_ch);
663 	callout_reset(&bw_meter_ch, BW_METER_PERIOD,
664 		      expire_bw_meter_process, NULL);
665 
666 	if (mrtdebug)
667 		log(LOG_DEBUG, "ip_mrouter_init\n");
668 
669 	return (0);
670 }
671 
672 /*
673  * Disable multicast routing
674  */
675 int
676 ip_mrouter_done(void)
677 {
678 	vifi_t vifi;
679 	struct vif *vifp;
680 	int i;
681 	int s;
682 
683 	s = splsoftnet();
684 
685 	/* Clear out all the vifs currently in use. */
686 	for (vifi = 0; vifi < numvifs; vifi++) {
687 		vifp = &viftable[vifi];
688 		if (!in_nullhost(vifp->v_lcl_addr))
689 			reset_vif(vifp);
690 	}
691 
692 	numvifs = 0;
693 	pim_assert = 0;
694 	mrt_api_config = 0;
695 
696 	callout_stop(&expire_upcalls_ch);
697 	callout_stop(&bw_upcalls_ch);
698 	callout_stop(&bw_meter_ch);
699 
700 	/*
701 	 * Free all multicast forwarding cache entries.
702 	 */
703 	for (i = 0; i < MFCTBLSIZ; i++) {
704 		struct mfc *rt, *nrt;
705 
706 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
707 			nrt = LIST_NEXT(rt, mfc_hash);
708 
709 			expire_mfc(rt);
710 		}
711 	}
712 
713 	bzero((caddr_t)nexpire, sizeof(nexpire));
714 	free(mfchashtbl, M_MRTABLE);
715 	mfchashtbl = NULL;
716 
717 	bw_upcalls_n = 0;
718 	bzero(bw_meter_timers, sizeof(bw_meter_timers));
719 
720 	/* Reset de-encapsulation cache. */
721 
722 	ip_mrouter = NULL;
723 
724 	splx(s);
725 
726 	if (mrtdebug)
727 		log(LOG_DEBUG, "ip_mrouter_done\n");
728 
729 	return (0);
730 }
731 
732 void
733 ip_mrouter_detach(struct ifnet *ifp)
734 {
735 	int vifi, i;
736 	struct vif *vifp;
737 	struct mfc *rt;
738 	struct rtdetq *rte;
739 
740 	/* XXX not sure about side effect to userland routing daemon */
741 	for (vifi = 0; vifi < numvifs; vifi++) {
742 		vifp = &viftable[vifi];
743 		if (vifp->v_ifp == ifp)
744 			reset_vif(vifp);
745 	}
746 	for (i = 0; i < MFCTBLSIZ; i++) {
747 		if (nexpire[i] == 0)
748 			continue;
749 		LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
750 			for (rte = rt->mfc_stall; rte; rte = rte->next) {
751 				if (rte->ifp == ifp)
752 					rte->ifp = NULL;
753 			}
754 		}
755 	}
756 }
757 
758 static int
759 get_version(struct mbuf *m)
760 {
761 	int *v = mtod(m, int *);
762 
763 	*v = 0x0305;	/* XXX !!!! */
764 	m->m_len = sizeof(int);
765 	return (0);
766 }
767 
768 /*
769  * Set PIM assert processing global
770  */
771 static int
772 set_assert(struct mbuf *m)
773 {
774 	int *i;
775 
776 	if (m == NULL || m->m_len < sizeof(int))
777 		return (EINVAL);
778 
779 	i = mtod(m, int *);
780 	pim_assert = !!*i;
781 	return (0);
782 }
783 
784 /*
785  * Get PIM assert processing global
786  */
787 static int
788 get_assert(struct mbuf *m)
789 {
790 	int *i = mtod(m, int *);
791 
792 	*i = pim_assert;
793 	m->m_len = sizeof(int);
794 	return (0);
795 }
796 
797 /*
798  * Configure API capabilities
799  */
800 static int
801 set_api_config(struct mbuf *m)
802 {
803 	int i;
804 	u_int32_t *apival;
805 
806 	if (m == NULL || m->m_len < sizeof(u_int32_t))
807 		return (EINVAL);
808 
809 	apival = mtod(m, u_int32_t *);
810 
811 	/*
812 	 * We can set the API capabilities only if it is the first operation
813 	 * after MRT_INIT. I.e.:
814 	 *  - there are no vifs installed
815 	 *  - pim_assert is not enabled
816 	 *  - the MFC table is empty
817 	 */
818 	if (numvifs > 0) {
819 		*apival = 0;
820 		return (EPERM);
821 	}
822 	if (pim_assert) {
823 		*apival = 0;
824 		return (EPERM);
825 	}
826 	for (i = 0; i < MFCTBLSIZ; i++) {
827 		if (LIST_FIRST(&mfchashtbl[i]) != NULL) {
828 			*apival = 0;
829 			return (EPERM);
830 		}
831 	}
832 
833 	mrt_api_config = *apival & mrt_api_support;
834 	*apival = mrt_api_config;
835 
836 	return (0);
837 }
838 
839 /*
840  * Get API capabilities
841  */
842 static int
843 get_api_support(struct mbuf *m)
844 {
845 	u_int32_t *apival;
846 
847 	if (m == NULL || m->m_len < sizeof(u_int32_t))
848 		return (EINVAL);
849 
850 	apival = mtod(m, u_int32_t *);
851 
852 	*apival = mrt_api_support;
853 
854 	return (0);
855 }
856 
857 /*
858  * Get API configured capabilities
859  */
860 static int
861 get_api_config(struct mbuf *m)
862 {
863 	u_int32_t *apival;
864 
865 	if (m == NULL || m->m_len < sizeof(u_int32_t))
866 		return (EINVAL);
867 
868 	apival = mtod(m, u_int32_t *);
869 
870 	*apival = mrt_api_config;
871 
872 	return (0);
873 }
874 
875 static struct sockaddr_in sin = { sizeof(sin), AF_INET };
876 
877 /*
878  * Add a vif to the vif table
879  */
880 static int
881 add_vif(struct mbuf *m)
882 {
883 	struct vifctl *vifcp;
884 	struct vif *vifp;
885 	struct ifaddr *ifa;
886 	struct ifnet *ifp;
887 	struct ifreq ifr;
888 	int error, s;
889 
890 	if (m == NULL || m->m_len < sizeof(struct vifctl))
891 		return (EINVAL);
892 
893 	vifcp = mtod(m, struct vifctl *);
894 	if (vifcp->vifc_vifi >= MAXVIFS)
895 		return (EINVAL);
896 	if (in_nullhost(vifcp->vifc_lcl_addr))
897 		return (EADDRNOTAVAIL);
898 
899 	vifp = &viftable[vifcp->vifc_vifi];
900 	if (!in_nullhost(vifp->v_lcl_addr))
901 		return (EADDRINUSE);
902 
903 	/* Find the interface with an address in AF_INET family. */
904 #ifdef PIM
905 	if (vifcp->vifc_flags & VIFF_REGISTER) {
906 		/*
907 		 * XXX: Because VIFF_REGISTER does not really need a valid
908 		 * local interface (e.g. it could be 127.0.0.2), we don't
909 		 * check its address.
910 		 */
911 	    ifp = NULL;
912 	} else
913 #endif
914 	{
915 		sin.sin_addr = vifcp->vifc_lcl_addr;
916 		ifa = ifa_ifwithaddr(sintosa(&sin));
917 		if (ifa == NULL)
918 			return (EADDRNOTAVAIL);
919 		ifp = ifa->ifa_ifp;
920 	}
921 
922 	if (vifcp->vifc_flags & VIFF_TUNNEL) {
923 		if (vifcp->vifc_flags & VIFF_SRCRT) {
924 			log(LOG_ERR, "source routed tunnels not supported\n");
925 			return (EOPNOTSUPP);
926 		}
927 
928 		/* attach this vif to decapsulator dispatch table */
929 		vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
930 		    vif_encapcheck, &vif_protosw, vifp);
931 		if (!vifp->v_encap_cookie)
932 			return (EINVAL);
933 
934 		/* Create a fake encapsulation interface. */
935 		ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK);
936 		bzero(ifp, sizeof(*ifp));
937 		snprintf(ifp->if_xname, sizeof(ifp->if_xname),
938 			 "mdecap%d", vifcp->vifc_vifi);
939 
940 		/* Prepare cached route entry. */
941 		bzero(&vifp->v_route, sizeof(vifp->v_route));
942 #ifdef PIM
943 	} else if (vifcp->vifc_flags & VIFF_REGISTER) {
944 		ifp = &multicast_register_if;
945 		if (mrtdebug)
946 			log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
947 			    (void *)ifp);
948 		if (reg_vif_num == VIFI_INVALID) {
949 			bzero(ifp, sizeof(*ifp));
950 			snprintf(ifp->if_xname, sizeof(ifp->if_xname),
951 				 "register_vif");
952 			ifp->if_flags = IFF_LOOPBACK;
953 			bzero(&vifp->v_route, sizeof(vifp->v_route));
954 			reg_vif_num = vifcp->vifc_vifi;
955 		}
956 #endif
957 	} else {
958 		/* Make sure the interface supports multicast. */
959 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
960 			return (EOPNOTSUPP);
961 
962 		/* Enable promiscuous reception of all IP multicasts. */
963 		satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
964 		satosin(&ifr.ifr_addr)->sin_family = AF_INET;
965 		satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
966 		error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
967 		if (error)
968 			return (error);
969 	}
970 
971 	s = splsoftnet();
972 
973 	/* Define parameters for the tbf structure. */
974 	vifp->tbf_q = NULL;
975 	vifp->tbf_t = &vifp->tbf_q;
976 	microtime(&vifp->tbf_last_pkt_t);
977 	vifp->tbf_n_tok = 0;
978 	vifp->tbf_q_len = 0;
979 	vifp->tbf_max_q_len = MAXQSIZE;
980 
981 	vifp->v_flags = vifcp->vifc_flags;
982 	vifp->v_threshold = vifcp->vifc_threshold;
983 	/* scaling up here allows division by 1024 in critical code */
984 	vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
985 	vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
986 	vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
987 	vifp->v_ifp = ifp;
988 	/* Initialize per vif pkt counters. */
989 	vifp->v_pkt_in = 0;
990 	vifp->v_pkt_out = 0;
991 	vifp->v_bytes_in = 0;
992 	vifp->v_bytes_out = 0;
993 
994 	callout_init(&vifp->v_repq_ch);
995 
996 #ifdef RSVP_ISI
997 	vifp->v_rsvp_on = 0;
998 	vifp->v_rsvpd = NULL;
999 #endif /* RSVP_ISI */
1000 
1001 	splx(s);
1002 
1003 	/* Adjust numvifs up if the vifi is higher than numvifs. */
1004 	if (numvifs <= vifcp->vifc_vifi)
1005 		numvifs = vifcp->vifc_vifi + 1;
1006 
1007 	if (mrtdebug)
1008 		log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
1009 		    vifcp->vifc_vifi,
1010 		    ntohl(vifcp->vifc_lcl_addr.s_addr),
1011 		    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
1012 		    ntohl(vifcp->vifc_rmt_addr.s_addr),
1013 		    vifcp->vifc_threshold,
1014 		    vifcp->vifc_rate_limit);
1015 
1016 	return (0);
1017 }
1018 
1019 void
1020 reset_vif(struct vif *vifp)
1021 {
1022 	struct mbuf *m, *n;
1023 	struct ifnet *ifp;
1024 	struct ifreq ifr;
1025 
1026 	callout_stop(&vifp->v_repq_ch);
1027 
1028 	/* detach this vif from decapsulator dispatch table */
1029 	encap_detach(vifp->v_encap_cookie);
1030 	vifp->v_encap_cookie = NULL;
1031 
1032 	/*
1033 	 * Free packets queued at the interface
1034 	 */
1035 	for (m = vifp->tbf_q; m != NULL; m = n) {
1036 		n = m->m_nextpkt;
1037 		m_freem(m);
1038 	}
1039 
1040 	if (vifp->v_flags & VIFF_TUNNEL) {
1041 		free(vifp->v_ifp, M_MRTABLE);
1042 		if (vifp == last_encap_vif) {
1043 			last_encap_vif = NULL;
1044 			last_encap_src = zeroin_addr;
1045 		}
1046 	} else if (vifp->v_flags & VIFF_REGISTER) {
1047 #ifdef PIM
1048 		reg_vif_num = VIFI_INVALID;
1049 #endif
1050 	} else {
1051 		satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
1052 		satosin(&ifr.ifr_addr)->sin_family = AF_INET;
1053 		satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
1054 		ifp = vifp->v_ifp;
1055 		(*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
1056 	}
1057 	bzero((caddr_t)vifp, sizeof(*vifp));
1058 }
1059 
1060 /*
1061  * Delete a vif from the vif table
1062  */
1063 static int
1064 del_vif(struct mbuf *m)
1065 {
1066 	vifi_t *vifip;
1067 	struct vif *vifp;
1068 	vifi_t vifi;
1069 	int s;
1070 
1071 	if (m == NULL || m->m_len < sizeof(vifi_t))
1072 		return (EINVAL);
1073 
1074 	vifip = mtod(m, vifi_t *);
1075 	if (*vifip >= numvifs)
1076 		return (EINVAL);
1077 
1078 	vifp = &viftable[*vifip];
1079 	if (in_nullhost(vifp->v_lcl_addr))
1080 		return (EADDRNOTAVAIL);
1081 
1082 	s = splsoftnet();
1083 
1084 	reset_vif(vifp);
1085 
1086 	/* Adjust numvifs down */
1087 	for (vifi = numvifs; vifi > 0; vifi--)
1088 		if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
1089 			break;
1090 	numvifs = vifi;
1091 
1092 	splx(s);
1093 
1094 	if (mrtdebug)
1095 		log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
1096 
1097 	return (0);
1098 }
1099 
1100 /*
1101  * update an mfc entry without resetting counters and S,G addresses.
1102  */
1103 static void
1104 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1105 {
1106 	int i;
1107 
1108 	rt->mfc_parent = mfccp->mfcc_parent;
1109 	for (i = 0; i < numvifs; i++) {
1110 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1111 		rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1112 			MRT_MFC_FLAGS_ALL;
1113 	}
1114 	/* set the RP address */
1115 	if (mrt_api_config & MRT_MFC_RP)
1116 		rt->mfc_rp = mfccp->mfcc_rp;
1117 	else
1118 		rt->mfc_rp = zeroin_addr;
1119 }
1120 
1121 /*
1122  * fully initialize an mfc entry from the parameter.
1123  */
1124 static void
1125 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1126 {
1127 	rt->mfc_origin     = mfccp->mfcc_origin;
1128 	rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1129 
1130 	update_mfc_params(rt, mfccp);
1131 
1132 	/* initialize pkt counters per src-grp */
1133 	rt->mfc_pkt_cnt    = 0;
1134 	rt->mfc_byte_cnt   = 0;
1135 	rt->mfc_wrong_if   = 0;
1136 	timerclear(&rt->mfc_last_assert);
1137 }
1138 
1139 static void
1140 expire_mfc(struct mfc *rt)
1141 {
1142 	struct rtdetq *rte, *nrte;
1143 
1144 	free_bw_list(rt->mfc_bw_meter);
1145 
1146 	for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1147 		nrte = rte->next;
1148 		m_freem(rte->m);
1149 		free(rte, M_MRTABLE);
1150 	}
1151 
1152 	LIST_REMOVE(rt, mfc_hash);
1153 	free(rt, M_MRTABLE);
1154 }
1155 
1156 /*
1157  * Add an mfc entry
1158  */
1159 static int
1160 add_mfc(struct mbuf *m)
1161 {
1162 	struct mfcctl2 mfcctl2;
1163 	struct mfcctl2 *mfccp;
1164 	struct mfc *rt;
1165 	u_int32_t hash = 0;
1166 	struct rtdetq *rte, *nrte;
1167 	u_short nstl;
1168 	int s;
1169 	int mfcctl_size = sizeof(struct mfcctl);
1170 
1171 	if (mrt_api_config & MRT_API_FLAGS_ALL)
1172 		mfcctl_size = sizeof(struct mfcctl2);
1173 
1174 	if (m == NULL || m->m_len < mfcctl_size)
1175 		return (EINVAL);
1176 
1177 	/*
1178 	 * select data size depending on API version.
1179 	 */
1180 	if (mrt_api_config & MRT_API_FLAGS_ALL) {
1181 		struct mfcctl2 *mp2 = mtod(m, struct mfcctl2 *);
1182 		bcopy(mp2, (caddr_t)&mfcctl2, sizeof(*mp2));
1183 	} else {
1184 		struct mfcctl *mp = mtod(m, struct mfcctl *);
1185 		bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp));
1186 		bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl),
1187 		      sizeof(mfcctl2) - sizeof(struct mfcctl));
1188 	}
1189 	mfccp = &mfcctl2;
1190 
1191 	s = splsoftnet();
1192 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1193 
1194 	/* If an entry already exists, just update the fields */
1195 	if (rt) {
1196 		if (mrtdebug & DEBUG_MFC)
1197 			log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1198 			    ntohl(mfccp->mfcc_origin.s_addr),
1199 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1200 			    mfccp->mfcc_parent);
1201 
1202 		update_mfc_params(rt, mfccp);
1203 
1204 		splx(s);
1205 		return (0);
1206 	}
1207 
1208 	/*
1209 	 * Find the entry for which the upcall was made and update
1210 	 */
1211 	nstl = 0;
1212 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1213 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1214 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1215 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1216 		    rt->mfc_stall != NULL) {
1217 			if (nstl++)
1218 				log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
1219 				    "multiple kernel entries",
1220 				    ntohl(mfccp->mfcc_origin.s_addr),
1221 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1222 				    mfccp->mfcc_parent, rt->mfc_stall);
1223 
1224 			if (mrtdebug & DEBUG_MFC)
1225 				log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n",
1226 				    ntohl(mfccp->mfcc_origin.s_addr),
1227 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1228 				    mfccp->mfcc_parent, rt->mfc_stall);
1229 
1230 			rte = rt->mfc_stall;
1231 			init_mfc_params(rt, mfccp);
1232 			rt->mfc_stall = NULL;
1233 
1234 			rt->mfc_expire = 0; /* Don't clean this guy up */
1235 			nexpire[hash]--;
1236 
1237 			/* free packets Qed at the end of this entry */
1238 			for (; rte != NULL; rte = nrte) {
1239 				nrte = rte->next;
1240 				if (rte->ifp) {
1241 #ifdef RSVP_ISI
1242 					ip_mdq(rte->m, rte->ifp, rt, -1);
1243 #else
1244 					ip_mdq(rte->m, rte->ifp, rt);
1245 #endif /* RSVP_ISI */
1246 				}
1247 				m_freem(rte->m);
1248 #ifdef UPCALL_TIMING
1249 				collate(&rte->t);
1250 #endif /* UPCALL_TIMING */
1251 				free(rte, M_MRTABLE);
1252 			}
1253 		}
1254 	}
1255 
1256 	/*
1257 	 * It is possible that an entry is being inserted without an upcall
1258 	 */
1259 	if (nstl == 0) {
1260 		/*
1261 		 * No mfc; make a new one
1262 		 */
1263 		if (mrtdebug & DEBUG_MFC)
1264 			log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1265 			    ntohl(mfccp->mfcc_origin.s_addr),
1266 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1267 			    mfccp->mfcc_parent);
1268 
1269 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1270 			if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1271 			    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1272 				init_mfc_params(rt, mfccp);
1273 				if (rt->mfc_expire)
1274 					nexpire[hash]--;
1275 				rt->mfc_expire = 0;
1276 				break; /* XXX */
1277 			}
1278 		}
1279 		if (rt == NULL) {	/* no upcall, so make a new entry */
1280 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1281 						  M_NOWAIT);
1282 			if (rt == NULL) {
1283 				splx(s);
1284 				return (ENOBUFS);
1285 			}
1286 
1287 			init_mfc_params(rt, mfccp);
1288 			rt->mfc_expire	= 0;
1289 			rt->mfc_stall	= NULL;
1290 			rt->mfc_bw_meter = NULL;
1291 
1292 			/* insert new entry at head of hash chain */
1293 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1294 		}
1295 	}
1296 
1297 	splx(s);
1298 	return (0);
1299 }
1300 
1301 #ifdef UPCALL_TIMING
1302 /*
1303  * collect delay statistics on the upcalls
1304  */
1305 static void
1306 collate(struct timeval *t)
1307 {
1308 	u_int32_t d;
1309 	struct timeval tp;
1310 	u_int32_t delta;
1311 
1312 	microtime(&tp);
1313 
1314 	if (timercmp(t, &tp, <)) {
1315 		TV_DELTA(tp, *t, delta);
1316 
1317 		d = delta >> 10;
1318 		if (d > 50)
1319 			d = 50;
1320 
1321 		++upcall_data[d];
1322 	}
1323 }
1324 #endif /* UPCALL_TIMING */
1325 
1326 /*
1327  * Delete an mfc entry
1328  */
1329 static int
1330 del_mfc(struct mbuf *m)
1331 {
1332 	struct mfcctl2 mfcctl2;
1333 	struct mfcctl2 *mfccp;
1334 	struct mfc *rt;
1335 	int s;
1336 	int mfcctl_size = sizeof(struct mfcctl);
1337 	struct mfcctl *mp = mtod(m, struct mfcctl *);
1338 
1339 	/*
1340 	 * XXX: for deleting MFC entries the information in entries
1341 	 * of size "struct mfcctl" is sufficient.
1342 	 */
1343 
1344 	if (m == NULL || m->m_len < mfcctl_size)
1345 		return (EINVAL);
1346 
1347 	bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp));
1348 	bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl),
1349 	      sizeof(mfcctl2) - sizeof(struct mfcctl));
1350 
1351 	mfccp = &mfcctl2;
1352 
1353 	if (mrtdebug & DEBUG_MFC)
1354 		log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1355 		    ntohl(mfccp->mfcc_origin.s_addr),
1356 		    ntohl(mfccp->mfcc_mcastgrp.s_addr));
1357 
1358 	s = splsoftnet();
1359 
1360 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1361 	if (rt == NULL) {
1362 		splx(s);
1363 		return (EADDRNOTAVAIL);
1364 	}
1365 
1366 	/*
1367 	 * free the bw_meter entries
1368 	 */
1369 	free_bw_list(rt->mfc_bw_meter);
1370 	rt->mfc_bw_meter = NULL;
1371 
1372 	LIST_REMOVE(rt, mfc_hash);
1373 	free(rt, M_MRTABLE);
1374 
1375 	splx(s);
1376 	return (0);
1377 }
1378 
1379 static int
1380 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1381 {
1382 	if (s) {
1383 		if (sbappendaddr(&s->so_rcv, sintosa(src), mm,
1384 		    (struct mbuf *)NULL) != 0) {
1385 			sorwakeup(s);
1386 			return (0);
1387 		}
1388 	}
1389 	m_freem(mm);
1390 	return (-1);
1391 }
1392 
1393 /*
1394  * IP multicast forwarding function. This function assumes that the packet
1395  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1396  * pointed to by "ifp", and the packet is to be relayed to other networks
1397  * that have members of the packet's destination IP multicast group.
1398  *
1399  * The packet is returned unscathed to the caller, unless it is
1400  * erroneous, in which case a non-zero return value tells the caller to
1401  * discard it.
1402  */
1403 
1404 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1405 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1406 
1407 int
1408 #ifdef RSVP_ISI
1409 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo)
1410 #else
1411 ip_mforward(struct mbuf *m, struct ifnet *ifp)
1412 #endif /* RSVP_ISI */
1413 {
1414 	struct ip *ip = mtod(m, struct ip *);
1415 	struct mfc *rt;
1416 	static int srctun = 0;
1417 	struct mbuf *mm;
1418 	int s;
1419 	vifi_t vifi;
1420 
1421 	if (mrtdebug & DEBUG_FORWARD)
1422 		log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1423 		    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1424 
1425 	if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1426 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1427 		/*
1428 		 * Packet arrived via a physical interface or
1429 		 * an encapsulated tunnel or a register_vif.
1430 		 */
1431 	} else {
1432 		/*
1433 		 * Packet arrived through a source-route tunnel.
1434 		 * Source-route tunnels are no longer supported.
1435 		 */
1436 		if ((srctun++ % 1000) == 0)
1437 			log(LOG_ERR,
1438 			    "ip_mforward: received source-routed packet from %x\n",
1439 			    ntohl(ip->ip_src.s_addr));
1440 
1441 		return (1);
1442 	}
1443 
1444 #ifdef RSVP_ISI
1445 	if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1446 		if (ip->ip_ttl < 255)
1447 			ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1448 		if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1449 			struct vif *vifp = viftable + vifi;
1450 			printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n",
1451 			    ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1452 			    (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1453 			    vifp->v_ifp->if_xname);
1454 		}
1455 		return (ip_mdq(m, ifp, (struct mfc *)NULL, vifi));
1456 	}
1457 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1458 		printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
1459 		    ntohl(ip->ip_src), ntohl(ip->ip_dst));
1460 	}
1461 #endif /* RSVP_ISI */
1462 
1463 	/*
1464 	 * Don't forward a packet with time-to-live of zero or one,
1465 	 * or a packet destined to a local-only group.
1466 	 */
1467 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1468 		return (0);
1469 
1470 	/*
1471 	 * Determine forwarding vifs from the forwarding cache table
1472 	 */
1473 	s = splsoftnet();
1474 	++mrtstat.mrts_mfc_lookups;
1475 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1476 
1477 	/* Entry exists, so forward if necessary */
1478 	if (rt != NULL) {
1479 		splx(s);
1480 #ifdef RSVP_ISI
1481 		return (ip_mdq(m, ifp, rt, -1));
1482 #else
1483 		return (ip_mdq(m, ifp, rt));
1484 #endif /* RSVP_ISI */
1485 	} else {
1486 		/*
1487 		 * If we don't have a route for packet's origin,
1488 		 * Make a copy of the packet & send message to routing daemon
1489 		 */
1490 
1491 		struct mbuf *mb0;
1492 		struct rtdetq *rte;
1493 		u_int32_t hash;
1494 		int hlen = ip->ip_hl << 2;
1495 #ifdef UPCALL_TIMING
1496 		struct timeval tp;
1497 
1498 		microtime(&tp);
1499 #endif /* UPCALL_TIMING */
1500 
1501 		++mrtstat.mrts_mfc_misses;
1502 
1503 		mrtstat.mrts_no_route++;
1504 		if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1505 			log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1506 			    ntohl(ip->ip_src.s_addr),
1507 			    ntohl(ip->ip_dst.s_addr));
1508 
1509 		/*
1510 		 * Allocate mbufs early so that we don't do extra work if we are
1511 		 * just going to fail anyway.  Make sure to pullup the header so
1512 		 * that other people can't step on it.
1513 		 */
1514 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1515 					      M_NOWAIT);
1516 		if (rte == NULL) {
1517 			splx(s);
1518 			return (ENOBUFS);
1519 		}
1520 		mb0 = m_copy(m, 0, M_COPYALL);
1521 		M_PULLUP(mb0, hlen);
1522 		if (mb0 == NULL) {
1523 			free(rte, M_MRTABLE);
1524 			splx(s);
1525 			return (ENOBUFS);
1526 		}
1527 
1528 		/* is there an upcall waiting for this flow? */
1529 		hash = MFCHASH(ip->ip_src, ip->ip_dst);
1530 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1531 			if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1532 			    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1533 			    rt->mfc_stall != NULL)
1534 				break;
1535 		}
1536 
1537 		if (rt == NULL) {
1538 			int i;
1539 			struct igmpmsg *im;
1540 
1541 			/*
1542 			 * Locate the vifi for the incoming interface for
1543 			 * this packet.
1544 			 * If none found, drop packet.
1545 			 */
1546 			for (vifi = 0; vifi < numvifs &&
1547 				 viftable[vifi].v_ifp != ifp; vifi++)
1548 				;
1549 			if (vifi >= numvifs) /* vif not found, drop packet */
1550 				goto non_fatal;
1551 
1552 			/* no upcall, so make a new entry */
1553 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1554 						  M_NOWAIT);
1555 			if (rt == NULL)
1556 				goto fail;
1557 
1558 			/*
1559 			 * Make a copy of the header to send to the user level
1560 			 * process
1561 			 */
1562 			mm = m_copy(m, 0, hlen);
1563 			M_PULLUP(mm, hlen);
1564 			if (mm == NULL)
1565 				goto fail1;
1566 
1567 			/*
1568 			 * Send message to routing daemon to install
1569 			 * a route into the kernel table
1570 			 */
1571 
1572 			im = mtod(mm, struct igmpmsg *);
1573 			im->im_msgtype = IGMPMSG_NOCACHE;
1574 			im->im_mbz = 0;
1575 			im->im_vif = vifi;
1576 
1577 			mrtstat.mrts_upcalls++;
1578 
1579 			sin.sin_addr = ip->ip_src;
1580 			if (socket_send(ip_mrouter, mm, &sin) < 0) {
1581 				log(LOG_WARNING,
1582 				    "ip_mforward: ip_mrouter socket queue full\n");
1583 				++mrtstat.mrts_upq_sockfull;
1584 			fail1:
1585 				free(rt, M_MRTABLE);
1586 			fail:
1587 				free(rte, M_MRTABLE);
1588 				m_freem(mb0);
1589 				splx(s);
1590 				return (ENOBUFS);
1591 			}
1592 
1593 			/* insert new entry at head of hash chain */
1594 			rt->mfc_origin = ip->ip_src;
1595 			rt->mfc_mcastgrp = ip->ip_dst;
1596 			rt->mfc_pkt_cnt = 0;
1597 			rt->mfc_byte_cnt = 0;
1598 			rt->mfc_wrong_if = 0;
1599 			rt->mfc_expire = UPCALL_EXPIRE;
1600 			nexpire[hash]++;
1601 			for (i = 0; i < numvifs; i++) {
1602 				rt->mfc_ttls[i] = 0;
1603 				rt->mfc_flags[i] = 0;
1604 			}
1605 			rt->mfc_parent = -1;
1606 
1607 			/* clear the RP address */
1608 			rt->mfc_rp = zeroin_addr;
1609 
1610 			rt->mfc_bw_meter = NULL;
1611 
1612 			/* link into table */
1613 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1614 			/* Add this entry to the end of the queue */
1615 			rt->mfc_stall = rte;
1616 		} else {
1617 			/* determine if q has overflowed */
1618 			struct rtdetq **p;
1619 			int npkts = 0;
1620 
1621 			/*
1622 			 * XXX ouch! we need to append to the list, but we
1623 			 * only have a pointer to the front, so we have to
1624 			 * scan the entire list every time.
1625 			 */
1626 			for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1627 				if (++npkts > MAX_UPQ) {
1628 					mrtstat.mrts_upq_ovflw++;
1629 				non_fatal:
1630 					free(rte, M_MRTABLE);
1631 					m_freem(mb0);
1632 					splx(s);
1633 					return (0);
1634 				}
1635 
1636 			/* Add this entry to the end of the queue */
1637 			*p = rte;
1638 		}
1639 
1640 		rte->next = NULL;
1641 		rte->m = mb0;
1642 		rte->ifp = ifp;
1643 #ifdef UPCALL_TIMING
1644 		rte->t = tp;
1645 #endif /* UPCALL_TIMING */
1646 
1647 		splx(s);
1648 
1649 		return (0);
1650 	}
1651 }
1652 
1653 
1654 /*ARGSUSED*/
1655 static void
1656 expire_upcalls(void *v)
1657 {
1658 	int i;
1659 	int s;
1660 
1661 	s = splsoftnet();
1662 
1663 	for (i = 0; i < MFCTBLSIZ; i++) {
1664 		struct mfc *rt, *nrt;
1665 
1666 		if (nexpire[i] == 0)
1667 			continue;
1668 
1669 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1670 			nrt = LIST_NEXT(rt, mfc_hash);
1671 
1672 			if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1673 				continue;
1674 			nexpire[i]--;
1675 
1676 			/*
1677 			 * free the bw_meter entries
1678 			 */
1679 			while (rt->mfc_bw_meter != NULL) {
1680 				struct bw_meter *x = rt->mfc_bw_meter;
1681 
1682 				rt->mfc_bw_meter = x->bm_mfc_next;
1683 				free(x, M_BWMETER);
1684 			}
1685 
1686 			++mrtstat.mrts_cache_cleanups;
1687 			if (mrtdebug & DEBUG_EXPIRE)
1688 				log(LOG_DEBUG,
1689 				    "expire_upcalls: expiring (%x %x)\n",
1690 				    ntohl(rt->mfc_origin.s_addr),
1691 				    ntohl(rt->mfc_mcastgrp.s_addr));
1692 
1693 			expire_mfc(rt);
1694 		}
1695 	}
1696 
1697 	splx(s);
1698 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1699 	    expire_upcalls, NULL);
1700 }
1701 
1702 /*
1703  * Packet forwarding routine once entry in the cache is made
1704  */
1705 static int
1706 #ifdef RSVP_ISI
1707 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1708 #else
1709 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1710 #endif /* RSVP_ISI */
1711 {
1712 	struct ip  *ip = mtod(m, struct ip *);
1713 	vifi_t vifi;
1714 	struct vif *vifp;
1715 	int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1716 
1717 /*
1718  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1719  * input, they shouldn't get counted on output, so statistics keeping is
1720  * separate.
1721  */
1722 #define MC_SEND(ip, vifp, m) do {					\
1723 	if ((vifp)->v_flags & VIFF_TUNNEL)				\
1724 		encap_send((ip), (vifp), (m));				\
1725 	else								\
1726 		phyint_send((ip), (vifp), (m));				\
1727 } while (/*CONSTCOND*/ 0)
1728 
1729 #ifdef RSVP_ISI
1730 	/*
1731 	 * If xmt_vif is not -1, send on only the requested vif.
1732 	 *
1733 	 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1734 	 */
1735 	if (xmt_vif < numvifs) {
1736 #ifdef PIM
1737 		if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1738 			pim_register_send(ip, viftable + xmt_vif, m, rt);
1739 		else
1740 #endif
1741 		MC_SEND(ip, viftable + xmt_vif, m);
1742 		return (1);
1743 	}
1744 #endif /* RSVP_ISI */
1745 
1746 	/*
1747 	 * Don't forward if it didn't arrive from the parent vif for its origin.
1748 	 */
1749 	vifi = rt->mfc_parent;
1750 	if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1751 		/* came in the wrong interface */
1752 		if (mrtdebug & DEBUG_FORWARD)
1753 			log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1754 			    ifp, vifi,
1755 			    vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1756 		++mrtstat.mrts_wrong_if;
1757 		++rt->mfc_wrong_if;
1758 		/*
1759 		 * If we are doing PIM assert processing, send a message
1760 		 * to the routing daemon.
1761 		 *
1762 		 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1763 		 * can complete the SPT switch, regardless of the type
1764 		 * of the iif (broadcast media, GRE tunnel, etc).
1765 		 */
1766 		if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1767 			struct timeval now;
1768 			u_int32_t delta;
1769 
1770 #ifdef PIM
1771 			if (ifp == &multicast_register_if)
1772 				pimstat.pims_rcv_registers_wrongiif++;
1773 #endif
1774 
1775 			/* Get vifi for the incoming packet */
1776 			for (vifi = 0;
1777 			     vifi < numvifs && viftable[vifi].v_ifp != ifp;
1778 			     vifi++)
1779 			    ;
1780 			if (vifi >= numvifs) {
1781 				/* The iif is not found: ignore the packet. */
1782 				return (0);
1783 			}
1784 
1785 			if (rt->mfc_flags[vifi] &
1786 			    MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1787 				/* WRONGVIF disabled: ignore the packet */
1788 				return (0);
1789 			}
1790 
1791 			microtime(&now);
1792 
1793 			TV_DELTA(rt->mfc_last_assert, now, delta);
1794 
1795 			if (delta > ASSERT_MSG_TIME) {
1796 				struct igmpmsg *im;
1797 				int hlen = ip->ip_hl << 2;
1798 				struct mbuf *mm = m_copy(m, 0, hlen);
1799 
1800 				M_PULLUP(mm, hlen);
1801 				if (mm == NULL)
1802 					return (ENOBUFS);
1803 
1804 				rt->mfc_last_assert = now;
1805 
1806 				im = mtod(mm, struct igmpmsg *);
1807 				im->im_msgtype	= IGMPMSG_WRONGVIF;
1808 				im->im_mbz	= 0;
1809 				im->im_vif	= vifi;
1810 
1811 				mrtstat.mrts_upcalls++;
1812 
1813 				sin.sin_addr = im->im_src;
1814 				if (socket_send(ip_mrouter, mm, &sin) < 0) {
1815 					log(LOG_WARNING,
1816 					    "ip_mforward: ip_mrouter socket queue full\n");
1817 					++mrtstat.mrts_upq_sockfull;
1818 					return (ENOBUFS);
1819 				}
1820 			}
1821 		}
1822 		return (0);
1823 	}
1824 
1825 	/* If I sourced this packet, it counts as output, else it was input. */
1826 	if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1827 		viftable[vifi].v_pkt_out++;
1828 		viftable[vifi].v_bytes_out += plen;
1829 	} else {
1830 		viftable[vifi].v_pkt_in++;
1831 		viftable[vifi].v_bytes_in += plen;
1832 	}
1833 	rt->mfc_pkt_cnt++;
1834 	rt->mfc_byte_cnt += plen;
1835 
1836 	/*
1837 	 * For each vif, decide if a copy of the packet should be forwarded.
1838 	 * Forward if:
1839 	 *		- the ttl exceeds the vif's threshold
1840 	 *		- there are group members downstream on interface
1841 	 */
1842 	for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1843 		if ((rt->mfc_ttls[vifi] > 0) &&
1844 			(ip->ip_ttl > rt->mfc_ttls[vifi])) {
1845 			vifp->v_pkt_out++;
1846 			vifp->v_bytes_out += plen;
1847 #ifdef PIM
1848 			if (vifp->v_flags & VIFF_REGISTER)
1849 				pim_register_send(ip, vifp, m, rt);
1850 			else
1851 #endif
1852 			MC_SEND(ip, vifp, m);
1853 		}
1854 
1855 	/*
1856 	 * Perform upcall-related bw measuring.
1857 	 */
1858 	if (rt->mfc_bw_meter != NULL) {
1859 		struct bw_meter *x;
1860 		struct timeval now;
1861 
1862 		microtime(&now);
1863 		for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1864 			bw_meter_receive_packet(x, plen, &now);
1865 	}
1866 
1867 	return (0);
1868 }
1869 
1870 #ifdef RSVP_ISI
1871 /*
1872  * check if a vif number is legal/ok. This is used by ip_output.
1873  */
1874 int
1875 legal_vif_num(int vif)
1876 {
1877 	if (vif >= 0 && vif < numvifs)
1878 		return (1);
1879 	else
1880 		return (0);
1881 }
1882 #endif /* RSVP_ISI */
1883 
1884 static void
1885 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1886 {
1887 	struct mbuf *mb_copy;
1888 	int hlen = ip->ip_hl << 2;
1889 
1890 	/*
1891 	 * Make a new reference to the packet; make sure that
1892 	 * the IP header is actually copied, not just referenced,
1893 	 * so that ip_output() only scribbles on the copy.
1894 	 */
1895 	mb_copy = m_copy(m, 0, M_COPYALL);
1896 	M_PULLUP(mb_copy, hlen);
1897 	if (mb_copy == NULL)
1898 		return;
1899 
1900 	if (vifp->v_rate_limit <= 0)
1901 		tbf_send_packet(vifp, mb_copy);
1902 	else
1903 		tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1904 		    ntohs(ip->ip_len));
1905 }
1906 
1907 static void
1908 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1909 {
1910 	struct mbuf *mb_copy;
1911 	struct ip *ip_copy;
1912 	int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1913 
1914 	/* Take care of delayed checksums */
1915 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1916 		in_delayed_cksum(m);
1917 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1918 	}
1919 
1920 	/*
1921 	 * copy the old packet & pullup it's IP header into the
1922 	 * new mbuf so we can modify it.  Try to fill the new
1923 	 * mbuf since if we don't the ethernet driver will.
1924 	 */
1925 	MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1926 	if (mb_copy == NULL)
1927 		return;
1928 	mb_copy->m_data += max_linkhdr;
1929 	mb_copy->m_pkthdr.len = len;
1930 	mb_copy->m_len = sizeof(multicast_encap_iphdr);
1931 
1932 	if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1933 		m_freem(mb_copy);
1934 		return;
1935 	}
1936 	i = MHLEN - max_linkhdr;
1937 	if (i > len)
1938 		i = len;
1939 	mb_copy = m_pullup(mb_copy, i);
1940 	if (mb_copy == NULL)
1941 		return;
1942 
1943 	/*
1944 	 * fill in the encapsulating IP header.
1945 	 */
1946 	ip_copy = mtod(mb_copy, struct ip *);
1947 	*ip_copy = multicast_encap_iphdr;
1948 	ip_copy->ip_id = ip_newid();
1949 	ip_copy->ip_len = htons(len);
1950 	ip_copy->ip_src = vifp->v_lcl_addr;
1951 	ip_copy->ip_dst = vifp->v_rmt_addr;
1952 
1953 	/*
1954 	 * turn the encapsulated IP header back into a valid one.
1955 	 */
1956 	ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1957 	--ip->ip_ttl;
1958 	ip->ip_sum = 0;
1959 	mb_copy->m_data += sizeof(multicast_encap_iphdr);
1960 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1961 	mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1962 
1963 	if (vifp->v_rate_limit <= 0)
1964 		tbf_send_packet(vifp, mb_copy);
1965 	else
1966 		tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1967 }
1968 
1969 /*
1970  * De-encapsulate a packet and feed it back through ip input.
1971  */
1972 static void
1973 vif_input(struct mbuf *m, ...)
1974 {
1975 	int off, proto;
1976 	va_list ap;
1977 	struct vif *vifp;
1978 	int s;
1979 	struct ifqueue *ifq;
1980 
1981 	va_start(ap, m);
1982 	off = va_arg(ap, int);
1983 	proto = va_arg(ap, int);
1984 	va_end(ap);
1985 
1986 	vifp = (struct vif *)encap_getarg(m);
1987 	if (!vifp || proto != AF_INET) {
1988 		m_freem(m);
1989 		mrtstat.mrts_bad_tunnel++;
1990 		return;
1991 	}
1992 
1993 	m_adj(m, off);
1994 	m->m_pkthdr.rcvif = vifp->v_ifp;
1995 	ifq = &ipintrq;
1996 	s = splnet();
1997 	if (IF_QFULL(ifq)) {
1998 		IF_DROP(ifq);
1999 		m_freem(m);
2000 	} else {
2001 		IF_ENQUEUE(ifq, m);
2002 		/*
2003 		 * normally we would need a "schednetisr(NETISR_IP)"
2004 		 * here but we were called by ip_input and it is going
2005 		 * to loop back & try to dequeue the packet we just
2006 		 * queued as soon as we return so we avoid the
2007 		 * unnecessary software interrrupt.
2008 		 */
2009 	}
2010 	splx(s);
2011 }
2012 
2013 /*
2014  * Check if the packet should be grabbed by us.
2015  */
2016 static int
2017 vif_encapcheck(struct mbuf *m, int off, int proto, void *arg)
2018 {
2019 	struct vif *vifp;
2020 	struct ip ip;
2021 
2022 #ifdef DIAGNOSTIC
2023 	if (!arg || proto != IPPROTO_IPV4)
2024 		panic("unexpected arg in vif_encapcheck");
2025 #endif
2026 
2027 	/*
2028 	 * do not grab the packet if it's not to a multicast destination or if
2029 	 * we don't have an encapsulating tunnel with the source.
2030 	 * Note:  This code assumes that the remote site IP address
2031 	 * uniquely identifies the tunnel (i.e., that this site has
2032 	 * at most one tunnel with the remote site).
2033 	 */
2034 
2035 	m_copydata(m, off, sizeof(ip), (caddr_t)&ip);
2036 	if (!IN_MULTICAST(ip.ip_dst.s_addr))
2037 		return 0;
2038 
2039 	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
2040 	if (!in_hosteq(ip.ip_src, last_encap_src)) {
2041 		vifp = (struct vif *)arg;
2042 		if (vifp->v_flags & VIFF_TUNNEL &&
2043 		    in_hosteq(vifp->v_rmt_addr, ip.ip_src))
2044 			;
2045 		else
2046 			return 0;
2047 		last_encap_vif = vifp;
2048 		last_encap_src = ip.ip_src;
2049 	} else
2050 		vifp = last_encap_vif;
2051 
2052 	/* 32bit match, since we have checked ip_src only */
2053 	return 32;
2054 }
2055 
2056 /*
2057  * Token bucket filter module
2058  */
2059 static void
2060 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
2061 {
2062 
2063 	if (len > MAX_BKT_SIZE) {
2064 		/* drop if packet is too large */
2065 		mrtstat.mrts_pkt2large++;
2066 		m_freem(m);
2067 		return;
2068 	}
2069 
2070 	tbf_update_tokens(vifp);
2071 
2072 	/*
2073 	 * If there are enough tokens, and the queue is empty, send this packet
2074 	 * out immediately.  Otherwise, try to insert it on this vif's queue.
2075 	 */
2076 	if (vifp->tbf_q_len == 0) {
2077 		if (len <= vifp->tbf_n_tok) {
2078 			vifp->tbf_n_tok -= len;
2079 			tbf_send_packet(vifp, m);
2080 		} else {
2081 			/* queue packet and timeout till later */
2082 			tbf_queue(vifp, m);
2083 			callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2084 			    tbf_reprocess_q, vifp);
2085 		}
2086 	} else {
2087 		if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
2088 		    !tbf_dq_sel(vifp, ip)) {
2089 			/* queue full, and couldn't make room */
2090 			mrtstat.mrts_q_overflow++;
2091 			m_freem(m);
2092 		} else {
2093 			/* queue length low enough, or made room */
2094 			tbf_queue(vifp, m);
2095 			tbf_process_q(vifp);
2096 		}
2097 	}
2098 }
2099 
2100 /*
2101  * adds a packet to the queue at the interface
2102  */
2103 static void
2104 tbf_queue(struct vif *vifp, struct mbuf *m)
2105 {
2106 	int s = splsoftnet();
2107 
2108 	/* insert at tail */
2109 	*vifp->tbf_t = m;
2110 	vifp->tbf_t = &m->m_nextpkt;
2111 	vifp->tbf_q_len++;
2112 
2113 	splx(s);
2114 }
2115 
2116 
2117 /*
2118  * processes the queue at the interface
2119  */
2120 static void
2121 tbf_process_q(struct vif *vifp)
2122 {
2123 	struct mbuf *m;
2124 	int len;
2125 	int s = splsoftnet();
2126 
2127 	/*
2128 	 * Loop through the queue at the interface and send as many packets
2129 	 * as possible.
2130 	 */
2131 	for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
2132 		len = ntohs(mtod(m, struct ip *)->ip_len);
2133 
2134 		/* determine if the packet can be sent */
2135 		if (len <= vifp->tbf_n_tok) {
2136 			/* if so,
2137 			 * reduce no of tokens, dequeue the packet,
2138 			 * send the packet.
2139 			 */
2140 			if ((vifp->tbf_q = m->m_nextpkt) == NULL)
2141 				vifp->tbf_t = &vifp->tbf_q;
2142 			--vifp->tbf_q_len;
2143 
2144 			m->m_nextpkt = NULL;
2145 			vifp->tbf_n_tok -= len;
2146 			tbf_send_packet(vifp, m);
2147 		} else
2148 			break;
2149 	}
2150 	splx(s);
2151 }
2152 
2153 static void
2154 tbf_reprocess_q(void *arg)
2155 {
2156 	struct vif *vifp = arg;
2157 
2158 	if (ip_mrouter == NULL)
2159 		return;
2160 
2161 	tbf_update_tokens(vifp);
2162 	tbf_process_q(vifp);
2163 
2164 	if (vifp->tbf_q_len != 0)
2165 		callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2166 		    tbf_reprocess_q, vifp);
2167 }
2168 
2169 /* function that will selectively discard a member of the queue
2170  * based on the precedence value and the priority
2171  */
2172 static int
2173 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2174 {
2175 	u_int p;
2176 	struct mbuf **mp, *m;
2177 	int s = splsoftnet();
2178 
2179 	p = priority(vifp, ip);
2180 
2181 	for (mp = &vifp->tbf_q, m = *mp;
2182 	    m != NULL;
2183 	    mp = &m->m_nextpkt, m = *mp) {
2184 		if (p > priority(vifp, mtod(m, struct ip *))) {
2185 			if ((*mp = m->m_nextpkt) == NULL)
2186 				vifp->tbf_t = mp;
2187 			--vifp->tbf_q_len;
2188 
2189 			m_freem(m);
2190 			mrtstat.mrts_drop_sel++;
2191 			splx(s);
2192 			return (1);
2193 		}
2194 	}
2195 	splx(s);
2196 	return (0);
2197 }
2198 
2199 static void
2200 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2201 {
2202 	int error;
2203 	int s = splsoftnet();
2204 
2205 	if (vifp->v_flags & VIFF_TUNNEL) {
2206 		/* If tunnel options */
2207 		ip_output(m, (struct mbuf *)NULL, &vifp->v_route,
2208 		    IP_FORWARDING, (struct ip_moptions *)NULL,
2209 		    (struct socket *)NULL);
2210 	} else {
2211 		/* if physical interface option, extract the options and then send */
2212 		struct ip_moptions imo;
2213 
2214 		imo.imo_multicast_ifp = vifp->v_ifp;
2215 		imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2216 		imo.imo_multicast_loop = 1;
2217 #ifdef RSVP_ISI
2218 		imo.imo_multicast_vif = -1;
2219 #endif
2220 
2221 		error = ip_output(m, (struct mbuf *)NULL, (struct route *)NULL,
2222 		    IP_FORWARDING|IP_MULTICASTOPTS, &imo,
2223 		    (struct socket *)NULL);
2224 
2225 		if (mrtdebug & DEBUG_XMIT)
2226 			log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2227 			    (long)(vifp - viftable), error);
2228 	}
2229 	splx(s);
2230 }
2231 
2232 /* determine the current time and then
2233  * the elapsed time (between the last time and time now)
2234  * in milliseconds & update the no. of tokens in the bucket
2235  */
2236 static void
2237 tbf_update_tokens(struct vif *vifp)
2238 {
2239 	struct timeval tp;
2240 	u_int32_t tm;
2241 	int s = splsoftnet();
2242 
2243 	microtime(&tp);
2244 
2245 	TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2246 
2247 	/*
2248 	 * This formula is actually
2249 	 * "time in seconds" * "bytes/second".
2250 	 *
2251 	 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2252 	 *
2253 	 * The (1000/1024) was introduced in add_vif to optimize
2254 	 * this divide into a shift.
2255 	 */
2256 	vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2257 	vifp->tbf_last_pkt_t = tp;
2258 
2259 	if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2260 		vifp->tbf_n_tok = MAX_BKT_SIZE;
2261 
2262 	splx(s);
2263 }
2264 
2265 static int
2266 priority(struct vif *vifp, struct ip *ip)
2267 {
2268 	int prio = 50;	/* the lowest priority -- default case */
2269 
2270 	/* temporary hack; may add general packet classifier some day */
2271 
2272 	/*
2273 	 * The UDP port space is divided up into four priority ranges:
2274 	 * [0, 16384)     : unclassified - lowest priority
2275 	 * [16384, 32768) : audio - highest priority
2276 	 * [32768, 49152) : whiteboard - medium priority
2277 	 * [49152, 65536) : video - low priority
2278 	 */
2279 	if (ip->ip_p == IPPROTO_UDP) {
2280 		struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2281 
2282 		switch (ntohs(udp->uh_dport) & 0xc000) {
2283 		case 0x4000:
2284 			prio = 70;
2285 			break;
2286 		case 0x8000:
2287 			prio = 60;
2288 			break;
2289 		case 0xc000:
2290 			prio = 55;
2291 			break;
2292 		}
2293 
2294 		if (tbfdebug > 1)
2295 			log(LOG_DEBUG, "port %x prio %d\n",
2296 			    ntohs(udp->uh_dport), prio);
2297 	}
2298 
2299 	return (prio);
2300 }
2301 
2302 /*
2303  * End of token bucket filter modifications
2304  */
2305 #ifdef RSVP_ISI
2306 int
2307 ip_rsvp_vif_init(struct socket *so, struct mbuf *m)
2308 {
2309 	int vifi, s;
2310 
2311 	if (rsvpdebug)
2312 		printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2313 		    so->so_type, so->so_proto->pr_protocol);
2314 
2315 	if (so->so_type != SOCK_RAW ||
2316 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2317 		return (EOPNOTSUPP);
2318 
2319 	/* Check mbuf. */
2320 	if (m == NULL || m->m_len != sizeof(int)) {
2321 		return (EINVAL);
2322 	}
2323 	vifi = *(mtod(m, int *));
2324 
2325 	if (rsvpdebug)
2326 		printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",
2327 		       vifi, rsvp_on);
2328 
2329 	s = splsoftnet();
2330 
2331 	/* Check vif. */
2332 	if (!legal_vif_num(vifi)) {
2333 		splx(s);
2334 		return (EADDRNOTAVAIL);
2335 	}
2336 
2337 	/* Check if socket is available. */
2338 	if (viftable[vifi].v_rsvpd != NULL) {
2339 		splx(s);
2340 		return (EADDRINUSE);
2341 	}
2342 
2343 	viftable[vifi].v_rsvpd = so;
2344 	/*
2345 	 * This may seem silly, but we need to be sure we don't over-increment
2346 	 * the RSVP counter, in case something slips up.
2347 	 */
2348 	if (!viftable[vifi].v_rsvp_on) {
2349 		viftable[vifi].v_rsvp_on = 1;
2350 		rsvp_on++;
2351 	}
2352 
2353 	splx(s);
2354 	return (0);
2355 }
2356 
2357 int
2358 ip_rsvp_vif_done(struct socket *so, struct mbuf *m)
2359 {
2360 	int vifi, s;
2361 
2362 	if (rsvpdebug)
2363 		printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2364 		    so->so_type, so->so_proto->pr_protocol);
2365 
2366 	if (so->so_type != SOCK_RAW ||
2367 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2368 		return (EOPNOTSUPP);
2369 
2370 	/* Check mbuf. */
2371 	if (m == NULL || m->m_len != sizeof(int)) {
2372 		return (EINVAL);
2373 	}
2374 	vifi = *(mtod(m, int *));
2375 
2376 	s = splsoftnet();
2377 
2378 	/* Check vif. */
2379 	if (!legal_vif_num(vifi)) {
2380 		splx(s);
2381 		return (EADDRNOTAVAIL);
2382 	}
2383 
2384 	if (rsvpdebug)
2385 		printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2386 		    viftable[vifi].v_rsvpd, so);
2387 
2388 	viftable[vifi].v_rsvpd = NULL;
2389 	/*
2390 	 * This may seem silly, but we need to be sure we don't over-decrement
2391 	 * the RSVP counter, in case something slips up.
2392 	 */
2393 	if (viftable[vifi].v_rsvp_on) {
2394 		viftable[vifi].v_rsvp_on = 0;
2395 		rsvp_on--;
2396 	}
2397 
2398 	splx(s);
2399 	return (0);
2400 }
2401 
2402 void
2403 ip_rsvp_force_done(struct socket *so)
2404 {
2405 	int vifi, s;
2406 
2407 	/* Don't bother if it is not the right type of socket. */
2408 	if (so->so_type != SOCK_RAW ||
2409 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2410 		return;
2411 
2412 	s = splsoftnet();
2413 
2414 	/*
2415 	 * The socket may be attached to more than one vif...this
2416 	 * is perfectly legal.
2417 	 */
2418 	for (vifi = 0; vifi < numvifs; vifi++) {
2419 		if (viftable[vifi].v_rsvpd == so) {
2420 			viftable[vifi].v_rsvpd = NULL;
2421 			/*
2422 			 * This may seem silly, but we need to be sure we don't
2423 			 * over-decrement the RSVP counter, in case something
2424 			 * slips up.
2425 			 */
2426 			if (viftable[vifi].v_rsvp_on) {
2427 				viftable[vifi].v_rsvp_on = 0;
2428 				rsvp_on--;
2429 			}
2430 		}
2431 	}
2432 
2433 	splx(s);
2434 	return;
2435 }
2436 
2437 void
2438 rsvp_input(struct mbuf *m, struct ifnet *ifp)
2439 {
2440 	int vifi, s;
2441 	struct ip *ip = mtod(m, struct ip *);
2442 	static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET };
2443 
2444 	if (rsvpdebug)
2445 		printf("rsvp_input: rsvp_on %d\n", rsvp_on);
2446 
2447 	/*
2448 	 * Can still get packets with rsvp_on = 0 if there is a local member
2449 	 * of the group to which the RSVP packet is addressed.  But in this
2450 	 * case we want to throw the packet away.
2451 	 */
2452 	if (!rsvp_on) {
2453 		m_freem(m);
2454 		return;
2455 	}
2456 
2457 	/*
2458 	 * If the old-style non-vif-associated socket is set, then use
2459 	 * it and ignore the new ones.
2460 	 */
2461 	if (ip_rsvpd != NULL) {
2462 		if (rsvpdebug)
2463 			printf("rsvp_input: "
2464 			    "Sending packet up old-style socket\n");
2465 		rip_input(m);	/*XXX*/
2466 		return;
2467 	}
2468 
2469 	s = splsoftnet();
2470 
2471 	if (rsvpdebug)
2472 		printf("rsvp_input: check vifs\n");
2473 
2474 	/* Find which vif the packet arrived on. */
2475 	for (vifi = 0; vifi < numvifs; vifi++) {
2476 		if (viftable[vifi].v_ifp == ifp)
2477 			break;
2478 	}
2479 
2480 	if (vifi == numvifs) {
2481 		/* Can't find vif packet arrived on. Drop packet. */
2482 		if (rsvpdebug)
2483 			printf("rsvp_input: "
2484 			    "Can't find vif for packet...dropping it.\n");
2485 		m_freem(m);
2486 		splx(s);
2487 		return;
2488 	}
2489 
2490 	if (rsvpdebug)
2491 		printf("rsvp_input: check socket\n");
2492 
2493 	if (viftable[vifi].v_rsvpd == NULL) {
2494 		/*
2495 		 * drop packet, since there is no specific socket for this
2496 		 * interface
2497 		 */
2498 		if (rsvpdebug)
2499 			printf("rsvp_input: No socket defined for vif %d\n",
2500 			    vifi);
2501 		m_freem(m);
2502 		splx(s);
2503 		return;
2504 	}
2505 
2506 	rsvp_src.sin_addr = ip->ip_src;
2507 
2508 	if (rsvpdebug && m)
2509 		printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
2510 		    m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv));
2511 
2512 	if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2513 		if (rsvpdebug)
2514 			printf("rsvp_input: Failed to append to socket\n");
2515 	else
2516 		if (rsvpdebug)
2517 			printf("rsvp_input: send packet up\n");
2518 
2519 	splx(s);
2520 }
2521 #endif /* RSVP_ISI */
2522 
2523 /*
2524  * Code for bandwidth monitors
2525  */
2526 
2527 /*
2528  * Define common interface for timeval-related methods
2529  */
2530 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2531 #define	BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2532 #define	BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2533 
2534 static uint32_t
2535 compute_bw_meter_flags(struct bw_upcall *req)
2536 {
2537     uint32_t flags = 0;
2538 
2539     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2540 	flags |= BW_METER_UNIT_PACKETS;
2541     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2542 	flags |= BW_METER_UNIT_BYTES;
2543     if (req->bu_flags & BW_UPCALL_GEQ)
2544 	flags |= BW_METER_GEQ;
2545     if (req->bu_flags & BW_UPCALL_LEQ)
2546 	flags |= BW_METER_LEQ;
2547 
2548     return flags;
2549 }
2550 
2551 /*
2552  * Add a bw_meter entry
2553  */
2554 static int
2555 add_bw_upcall(struct mbuf *m)
2556 {
2557     int s;
2558     struct mfc *mfc;
2559     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2560 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2561     struct timeval now;
2562     struct bw_meter *x;
2563     uint32_t flags;
2564     struct bw_upcall *req;
2565 
2566     if (m == NULL || m->m_len < sizeof(struct bw_upcall))
2567 	return EINVAL;
2568 
2569     req = mtod(m, struct bw_upcall *);
2570 
2571     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2572 	return EOPNOTSUPP;
2573 
2574     /* Test if the flags are valid */
2575     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2576 	return EINVAL;
2577     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2578 	return EINVAL;
2579     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2580 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2581 	return EINVAL;
2582 
2583     /* Test if the threshold time interval is valid */
2584     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2585 	return EINVAL;
2586 
2587     flags = compute_bw_meter_flags(req);
2588 
2589     /*
2590      * Find if we have already same bw_meter entry
2591      */
2592     s = splsoftnet();
2593     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2594     if (mfc == NULL) {
2595 	splx(s);
2596 	return EADDRNOTAVAIL;
2597     }
2598     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2599 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2600 			   &req->bu_threshold.b_time, ==)) &&
2601 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2602 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2603 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
2604 	    splx(s);
2605 	    return 0;		/* XXX Already installed */
2606 	}
2607     }
2608 
2609     /* Allocate the new bw_meter entry */
2610     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2611     if (x == NULL) {
2612 	splx(s);
2613 	return ENOBUFS;
2614     }
2615 
2616     /* Set the new bw_meter entry */
2617     x->bm_threshold.b_time = req->bu_threshold.b_time;
2618     microtime(&now);
2619     x->bm_start_time = now;
2620     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2621     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2622     x->bm_measured.b_packets = 0;
2623     x->bm_measured.b_bytes = 0;
2624     x->bm_flags = flags;
2625     x->bm_time_next = NULL;
2626     x->bm_time_hash = BW_METER_BUCKETS;
2627 
2628     /* Add the new bw_meter entry to the front of entries for this MFC */
2629     x->bm_mfc = mfc;
2630     x->bm_mfc_next = mfc->mfc_bw_meter;
2631     mfc->mfc_bw_meter = x;
2632     schedule_bw_meter(x, &now);
2633     splx(s);
2634 
2635     return 0;
2636 }
2637 
2638 static void
2639 free_bw_list(struct bw_meter *list)
2640 {
2641     while (list != NULL) {
2642 	struct bw_meter *x = list;
2643 
2644 	list = list->bm_mfc_next;
2645 	unschedule_bw_meter(x);
2646 	free(x, M_BWMETER);
2647     }
2648 }
2649 
2650 /*
2651  * Delete one or multiple bw_meter entries
2652  */
2653 static int
2654 del_bw_upcall(struct mbuf *m)
2655 {
2656     int s;
2657     struct mfc *mfc;
2658     struct bw_meter *x;
2659     struct bw_upcall *req;
2660 
2661     if (m == NULL || m->m_len < sizeof(struct bw_upcall))
2662 	return EINVAL;
2663 
2664     req = mtod(m, struct bw_upcall *);
2665 
2666     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2667 	return EOPNOTSUPP;
2668 
2669     s = splsoftnet();
2670     /* Find the corresponding MFC entry */
2671     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2672     if (mfc == NULL) {
2673 	splx(s);
2674 	return EADDRNOTAVAIL;
2675     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2676 	/*
2677 	 * Delete all bw_meter entries for this mfc
2678 	 */
2679 	struct bw_meter *list;
2680 
2681 	list = mfc->mfc_bw_meter;
2682 	mfc->mfc_bw_meter = NULL;
2683 	free_bw_list(list);
2684 	splx(s);
2685 	return 0;
2686     } else {			/* Delete a single bw_meter entry */
2687 	struct bw_meter *prev;
2688 	uint32_t flags = 0;
2689 
2690 	flags = compute_bw_meter_flags(req);
2691 
2692 	/* Find the bw_meter entry to delete */
2693 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2694 	     prev = x, x = x->bm_mfc_next) {
2695 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2696 			       &req->bu_threshold.b_time, ==)) &&
2697 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2698 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2699 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2700 		break;
2701 	}
2702 	if (x != NULL) { /* Delete entry from the list for this MFC */
2703 	    if (prev != NULL)
2704 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2705 	    else
2706 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2707 
2708 	    unschedule_bw_meter(x);
2709 	    splx(s);
2710 	    /* Free the bw_meter entry */
2711 	    free(x, M_BWMETER);
2712 	    return 0;
2713 	} else {
2714 	    splx(s);
2715 	    return EINVAL;
2716 	}
2717     }
2718     /* NOTREACHED */
2719 }
2720 
2721 /*
2722  * Perform bandwidth measurement processing that may result in an upcall
2723  */
2724 static void
2725 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2726 {
2727     struct timeval delta;
2728 
2729     delta = *nowp;
2730     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2731 
2732     if (x->bm_flags & BW_METER_GEQ) {
2733 	/*
2734 	 * Processing for ">=" type of bw_meter entry
2735 	 */
2736 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2737 	    /* Reset the bw_meter entry */
2738 	    x->bm_start_time = *nowp;
2739 	    x->bm_measured.b_packets = 0;
2740 	    x->bm_measured.b_bytes = 0;
2741 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2742 	}
2743 
2744 	/* Record that a packet is received */
2745 	x->bm_measured.b_packets++;
2746 	x->bm_measured.b_bytes += plen;
2747 
2748 	/*
2749 	 * Test if we should deliver an upcall
2750 	 */
2751 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2752 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2753 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2754 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2755 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2756 		/* Prepare an upcall for delivery */
2757 		bw_meter_prepare_upcall(x, nowp);
2758 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2759 	    }
2760 	}
2761     } else if (x->bm_flags & BW_METER_LEQ) {
2762 	/*
2763 	 * Processing for "<=" type of bw_meter entry
2764 	 */
2765 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2766 	    /*
2767 	     * We are behind time with the multicast forwarding table
2768 	     * scanning for "<=" type of bw_meter entries, so test now
2769 	     * if we should deliver an upcall.
2770 	     */
2771 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2772 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2773 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2774 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2775 		/* Prepare an upcall for delivery */
2776 		bw_meter_prepare_upcall(x, nowp);
2777 	    }
2778 	    /* Reschedule the bw_meter entry */
2779 	    unschedule_bw_meter(x);
2780 	    schedule_bw_meter(x, nowp);
2781 	}
2782 
2783 	/* Record that a packet is received */
2784 	x->bm_measured.b_packets++;
2785 	x->bm_measured.b_bytes += plen;
2786 
2787 	/*
2788 	 * Test if we should restart the measuring interval
2789 	 */
2790 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2791 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2792 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
2793 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2794 	    /* Don't restart the measuring interval */
2795 	} else {
2796 	    /* Do restart the measuring interval */
2797 	    /*
2798 	     * XXX: note that we don't unschedule and schedule, because this
2799 	     * might be too much overhead per packet. Instead, when we process
2800 	     * all entries for a given timer hash bin, we check whether it is
2801 	     * really a timeout. If not, we reschedule at that time.
2802 	     */
2803 	    x->bm_start_time = *nowp;
2804 	    x->bm_measured.b_packets = 0;
2805 	    x->bm_measured.b_bytes = 0;
2806 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2807 	}
2808     }
2809 }
2810 
2811 /*
2812  * Prepare a bandwidth-related upcall
2813  */
2814 static void
2815 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2816 {
2817     struct timeval delta;
2818     struct bw_upcall *u;
2819 
2820     /*
2821      * Compute the measured time interval
2822      */
2823     delta = *nowp;
2824     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2825 
2826     /*
2827      * If there are too many pending upcalls, deliver them now
2828      */
2829     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2830 	bw_upcalls_send();
2831 
2832     /*
2833      * Set the bw_upcall entry
2834      */
2835     u = &bw_upcalls[bw_upcalls_n++];
2836     u->bu_src = x->bm_mfc->mfc_origin;
2837     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2838     u->bu_threshold.b_time = x->bm_threshold.b_time;
2839     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2840     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2841     u->bu_measured.b_time = delta;
2842     u->bu_measured.b_packets = x->bm_measured.b_packets;
2843     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2844     u->bu_flags = 0;
2845     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2846 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2847     if (x->bm_flags & BW_METER_UNIT_BYTES)
2848 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2849     if (x->bm_flags & BW_METER_GEQ)
2850 	u->bu_flags |= BW_UPCALL_GEQ;
2851     if (x->bm_flags & BW_METER_LEQ)
2852 	u->bu_flags |= BW_UPCALL_LEQ;
2853 }
2854 
2855 /*
2856  * Send the pending bandwidth-related upcalls
2857  */
2858 static void
2859 bw_upcalls_send(void)
2860 {
2861     struct mbuf *m;
2862     int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2863     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2864     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2865 				      0,		/* unused2 */
2866 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2867 				      0,		/* im_mbz  */
2868 				      0,		/* im_vif  */
2869 				      0,		/* unused3 */
2870 				      { 0 },		/* im_src  */
2871 				      { 0 } };		/* im_dst  */
2872 
2873     if (bw_upcalls_n == 0)
2874 	return;			/* No pending upcalls */
2875 
2876     bw_upcalls_n = 0;
2877 
2878     /*
2879      * Allocate a new mbuf, initialize it with the header and
2880      * the payload for the pending calls.
2881      */
2882     MGETHDR(m, M_DONTWAIT, MT_HEADER);
2883     if (m == NULL) {
2884 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2885 	return;
2886     }
2887 
2888     m->m_len = m->m_pkthdr.len = 0;
2889     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2890     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2891 
2892     /*
2893      * Send the upcalls
2894      * XXX do we need to set the address in k_igmpsrc ?
2895      */
2896     mrtstat.mrts_upcalls++;
2897     if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2898 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2899 	++mrtstat.mrts_upq_sockfull;
2900     }
2901 }
2902 
2903 /*
2904  * Compute the timeout hash value for the bw_meter entries
2905  */
2906 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2907     do {								\
2908 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2909 									\
2910 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2911 	(hash) = next_timeval.tv_sec;					\
2912 	if (next_timeval.tv_usec)					\
2913 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2914 	(hash) %= BW_METER_BUCKETS;					\
2915     } while (/*CONSTCOND*/ 0)
2916 
2917 /*
2918  * Schedule a timer to process periodically bw_meter entry of type "<="
2919  * by linking the entry in the proper hash bucket.
2920  */
2921 static void
2922 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2923 {
2924     int time_hash;
2925 
2926     if (!(x->bm_flags & BW_METER_LEQ))
2927 	return;		/* XXX: we schedule timers only for "<=" entries */
2928 
2929     /*
2930      * Reset the bw_meter entry
2931      */
2932     x->bm_start_time = *nowp;
2933     x->bm_measured.b_packets = 0;
2934     x->bm_measured.b_bytes = 0;
2935     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2936 
2937     /*
2938      * Compute the timeout hash value and insert the entry
2939      */
2940     BW_METER_TIMEHASH(x, time_hash);
2941     x->bm_time_next = bw_meter_timers[time_hash];
2942     bw_meter_timers[time_hash] = x;
2943     x->bm_time_hash = time_hash;
2944 }
2945 
2946 /*
2947  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2948  * by removing the entry from the proper hash bucket.
2949  */
2950 static void
2951 unschedule_bw_meter(struct bw_meter *x)
2952 {
2953     int time_hash;
2954     struct bw_meter *prev, *tmp;
2955 
2956     if (!(x->bm_flags & BW_METER_LEQ))
2957 	return;		/* XXX: we schedule timers only for "<=" entries */
2958 
2959     /*
2960      * Compute the timeout hash value and delete the entry
2961      */
2962     time_hash = x->bm_time_hash;
2963     if (time_hash >= BW_METER_BUCKETS)
2964 	return;		/* Entry was not scheduled */
2965 
2966     for (prev = NULL, tmp = bw_meter_timers[time_hash];
2967 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2968 	if (tmp == x)
2969 	    break;
2970 
2971     if (tmp == NULL)
2972 	panic("unschedule_bw_meter: bw_meter entry not found");
2973 
2974     if (prev != NULL)
2975 	prev->bm_time_next = x->bm_time_next;
2976     else
2977 	bw_meter_timers[time_hash] = x->bm_time_next;
2978 
2979     x->bm_time_next = NULL;
2980     x->bm_time_hash = BW_METER_BUCKETS;
2981 }
2982 
2983 /*
2984  * Process all "<=" type of bw_meter that should be processed now,
2985  * and for each entry prepare an upcall if necessary. Each processed
2986  * entry is rescheduled again for the (periodic) processing.
2987  *
2988  * This is run periodically (once per second normally). On each round,
2989  * all the potentially matching entries are in the hash slot that we are
2990  * looking at.
2991  */
2992 static void
2993 bw_meter_process(void)
2994 {
2995     int s;
2996     static uint32_t last_tv_sec;	/* last time we processed this */
2997 
2998     uint32_t loops;
2999     int i;
3000     struct timeval now, process_endtime;
3001 
3002     microtime(&now);
3003     if (last_tv_sec == now.tv_sec)
3004 	return;		/* nothing to do */
3005 
3006     loops = now.tv_sec - last_tv_sec;
3007     last_tv_sec = now.tv_sec;
3008     if (loops > BW_METER_BUCKETS)
3009 	loops = BW_METER_BUCKETS;
3010 
3011     s = splsoftnet();
3012     /*
3013      * Process all bins of bw_meter entries from the one after the last
3014      * processed to the current one. On entry, i points to the last bucket
3015      * visited, so we need to increment i at the beginning of the loop.
3016      */
3017     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
3018 	struct bw_meter *x, *tmp_list;
3019 
3020 	if (++i >= BW_METER_BUCKETS)
3021 	    i = 0;
3022 
3023 	/* Disconnect the list of bw_meter entries from the bin */
3024 	tmp_list = bw_meter_timers[i];
3025 	bw_meter_timers[i] = NULL;
3026 
3027 	/* Process the list of bw_meter entries */
3028 	while (tmp_list != NULL) {
3029 	    x = tmp_list;
3030 	    tmp_list = tmp_list->bm_time_next;
3031 
3032 	    /* Test if the time interval is over */
3033 	    process_endtime = x->bm_start_time;
3034 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
3035 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
3036 		/* Not yet: reschedule, but don't reset */
3037 		int time_hash;
3038 
3039 		BW_METER_TIMEHASH(x, time_hash);
3040 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
3041 		    /*
3042 		     * XXX: somehow the bin processing is a bit ahead of time.
3043 		     * Put the entry in the next bin.
3044 		     */
3045 		    if (++time_hash >= BW_METER_BUCKETS)
3046 			time_hash = 0;
3047 		}
3048 		x->bm_time_next = bw_meter_timers[time_hash];
3049 		bw_meter_timers[time_hash] = x;
3050 		x->bm_time_hash = time_hash;
3051 
3052 		continue;
3053 	    }
3054 
3055 	    /*
3056 	     * Test if we should deliver an upcall
3057 	     */
3058 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
3059 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
3060 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
3061 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
3062 		/* Prepare an upcall for delivery */
3063 		bw_meter_prepare_upcall(x, &now);
3064 	    }
3065 
3066 	    /*
3067 	     * Reschedule for next processing
3068 	     */
3069 	    schedule_bw_meter(x, &now);
3070 	}
3071     }
3072 
3073     /* Send all upcalls that are pending delivery */
3074     bw_upcalls_send();
3075 
3076     splx(s);
3077 }
3078 
3079 /*
3080  * A periodic function for sending all upcalls that are pending delivery
3081  */
3082 static void
3083 expire_bw_upcalls_send(void *unused)
3084 {
3085     int s;
3086 
3087     s = splsoftnet();
3088     bw_upcalls_send();
3089     splx(s);
3090 
3091     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
3092 		  expire_bw_upcalls_send, NULL);
3093 }
3094 
3095 /*
3096  * A periodic function for periodic scanning of the multicast forwarding
3097  * table for processing all "<=" bw_meter entries.
3098  */
3099 static void
3100 expire_bw_meter_process(void *unused)
3101 {
3102     if (mrt_api_config & MRT_MFC_BW_UPCALL)
3103 	bw_meter_process();
3104 
3105     callout_reset(&bw_meter_ch, BW_METER_PERIOD,
3106 		  expire_bw_meter_process, NULL);
3107 }
3108 
3109 /*
3110  * End of bandwidth monitoring code
3111  */
3112 
3113 #ifdef PIM
3114 /*
3115  * Send the packet up to the user daemon, or eventually do kernel encapsulation
3116  */
3117 static int
3118 pim_register_send(struct ip *ip, struct vif *vifp,
3119 	struct mbuf *m, struct mfc *rt)
3120 {
3121     struct mbuf *mb_copy, *mm;
3122 
3123     if (mrtdebug & DEBUG_PIM)
3124         log(LOG_DEBUG, "pim_register_send: ");
3125 
3126     mb_copy = pim_register_prepare(ip, m);
3127     if (mb_copy == NULL)
3128 	return ENOBUFS;
3129 
3130     /*
3131      * Send all the fragments. Note that the mbuf for each fragment
3132      * is freed by the sending machinery.
3133      */
3134     for (mm = mb_copy; mm; mm = mb_copy) {
3135 	mb_copy = mm->m_nextpkt;
3136 	mm->m_nextpkt = NULL;
3137 	mm = m_pullup(mm, sizeof(struct ip));
3138 	if (mm != NULL) {
3139 	    ip = mtod(mm, struct ip *);
3140 	    if ((mrt_api_config & MRT_MFC_RP) &&
3141 		!in_nullhost(rt->mfc_rp)) {
3142 		pim_register_send_rp(ip, vifp, mm, rt);
3143 	    } else {
3144 		pim_register_send_upcall(ip, vifp, mm, rt);
3145 	    }
3146 	}
3147     }
3148 
3149     return 0;
3150 }
3151 
3152 /*
3153  * Return a copy of the data packet that is ready for PIM Register
3154  * encapsulation.
3155  * XXX: Note that in the returned copy the IP header is a valid one.
3156  */
3157 static struct mbuf *
3158 pim_register_prepare(struct ip *ip, struct mbuf *m)
3159 {
3160     struct mbuf *mb_copy = NULL;
3161     int mtu;
3162 
3163     /* Take care of delayed checksums */
3164     if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3165 	in_delayed_cksum(m);
3166 	m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3167     }
3168 
3169     /*
3170      * Copy the old packet & pullup its IP header into the
3171      * new mbuf so we can modify it.
3172      */
3173     mb_copy = m_copy(m, 0, M_COPYALL);
3174     if (mb_copy == NULL)
3175 	return NULL;
3176     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3177     if (mb_copy == NULL)
3178 	return NULL;
3179 
3180     /* take care of the TTL */
3181     ip = mtod(mb_copy, struct ip *);
3182     --ip->ip_ttl;
3183 
3184     /* Compute the MTU after the PIM Register encapsulation */
3185     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3186 
3187     if (ntohs(ip->ip_len) <= mtu) {
3188 	/* Turn the IP header into a valid one */
3189 	ip->ip_sum = 0;
3190 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3191     } else {
3192 	/* Fragment the packet */
3193 	if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3194 	    /* XXX: mb_copy was freed by ip_fragment() */
3195 	    return NULL;
3196 	}
3197     }
3198     return mb_copy;
3199 }
3200 
3201 /*
3202  * Send an upcall with the data packet to the user-level process.
3203  */
3204 static int
3205 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3206 	struct mbuf *mb_copy, struct mfc *rt)
3207 {
3208     struct mbuf *mb_first;
3209     int len = ntohs(ip->ip_len);
3210     struct igmpmsg *im;
3211     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
3212 
3213     /*
3214      * Add a new mbuf with an upcall header
3215      */
3216     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3217     if (mb_first == NULL) {
3218 	m_freem(mb_copy);
3219 	return ENOBUFS;
3220     }
3221     mb_first->m_data += max_linkhdr;
3222     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3223     mb_first->m_len = sizeof(struct igmpmsg);
3224     mb_first->m_next = mb_copy;
3225 
3226     /* Send message to routing daemon */
3227     im = mtod(mb_first, struct igmpmsg *);
3228     im->im_msgtype	= IGMPMSG_WHOLEPKT;
3229     im->im_mbz		= 0;
3230     im->im_vif		= vifp - viftable;
3231     im->im_src		= ip->ip_src;
3232     im->im_dst		= ip->ip_dst;
3233 
3234     k_igmpsrc.sin_addr	= ip->ip_src;
3235 
3236     mrtstat.mrts_upcalls++;
3237 
3238     if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3239 	if (mrtdebug & DEBUG_PIM)
3240 	    log(LOG_WARNING,
3241 		"mcast: pim_register_send_upcall: ip_mrouter socket queue full");
3242 	++mrtstat.mrts_upq_sockfull;
3243 	return ENOBUFS;
3244     }
3245 
3246     /* Keep statistics */
3247     pimstat.pims_snd_registers_msgs++;
3248     pimstat.pims_snd_registers_bytes += len;
3249 
3250     return 0;
3251 }
3252 
3253 /*
3254  * Encapsulate the data packet in PIM Register message and send it to the RP.
3255  */
3256 static int
3257 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3258 	struct mbuf *mb_copy, struct mfc *rt)
3259 {
3260     struct mbuf *mb_first;
3261     struct ip *ip_outer;
3262     struct pim_encap_pimhdr *pimhdr;
3263     int len = ntohs(ip->ip_len);
3264     vifi_t vifi = rt->mfc_parent;
3265 
3266     if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3267 	m_freem(mb_copy);
3268 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
3269     }
3270 
3271     /*
3272      * Add a new mbuf with the encapsulating header
3273      */
3274     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3275     if (mb_first == NULL) {
3276 	m_freem(mb_copy);
3277 	return ENOBUFS;
3278     }
3279     mb_first->m_data += max_linkhdr;
3280     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3281     mb_first->m_next = mb_copy;
3282 
3283     mb_first->m_pkthdr.len = len + mb_first->m_len;
3284 
3285     /*
3286      * Fill in the encapsulating IP and PIM header
3287      */
3288     ip_outer = mtod(mb_first, struct ip *);
3289     *ip_outer = pim_encap_iphdr;
3290     ip_outer->ip_id = ip_newid();
3291     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3292 			     sizeof(pim_encap_pimhdr));
3293     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3294     ip_outer->ip_dst = rt->mfc_rp;
3295     /*
3296      * Copy the inner header TOS to the outer header, and take care of the
3297      * IP_DF bit.
3298      */
3299     ip_outer->ip_tos = ip->ip_tos;
3300     if (ntohs(ip->ip_off) & IP_DF)
3301 	ip_outer->ip_off |= IP_DF;
3302     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
3303 					 + sizeof(pim_encap_iphdr));
3304     *pimhdr = pim_encap_pimhdr;
3305     /* If the iif crosses a border, set the Border-bit */
3306     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3307 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3308 
3309     mb_first->m_data += sizeof(pim_encap_iphdr);
3310     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3311     mb_first->m_data -= sizeof(pim_encap_iphdr);
3312 
3313     if (vifp->v_rate_limit == 0)
3314 	tbf_send_packet(vifp, mb_first);
3315     else
3316 	tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3317 
3318     /* Keep statistics */
3319     pimstat.pims_snd_registers_msgs++;
3320     pimstat.pims_snd_registers_bytes += len;
3321 
3322     return 0;
3323 }
3324 
3325 /*
3326  * PIM-SMv2 and PIM-DM messages processing.
3327  * Receives and verifies the PIM control messages, and passes them
3328  * up to the listening socket, using rip_input().
3329  * The only message with special processing is the PIM_REGISTER message
3330  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3331  * is passed to if_simloop().
3332  */
3333 void
3334 pim_input(struct mbuf *m, ...)
3335 {
3336     struct ip *ip = mtod(m, struct ip *);
3337     struct pim *pim;
3338     int minlen;
3339     int datalen;
3340     int ip_tos;
3341     int proto;
3342     int iphlen;
3343     va_list ap;
3344 
3345     va_start(ap, m);
3346     iphlen = va_arg(ap, int);
3347     proto = va_arg(ap, int);
3348     va_end(ap);
3349 
3350     datalen = ntohs(ip->ip_len) - iphlen;
3351 
3352     /* Keep statistics */
3353     pimstat.pims_rcv_total_msgs++;
3354     pimstat.pims_rcv_total_bytes += datalen;
3355 
3356     /*
3357      * Validate lengths
3358      */
3359     if (datalen < PIM_MINLEN) {
3360 	pimstat.pims_rcv_tooshort++;
3361 	log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3362 	    datalen, (u_long)ip->ip_src.s_addr);
3363 	m_freem(m);
3364 	return;
3365     }
3366 
3367     /*
3368      * If the packet is at least as big as a REGISTER, go agead
3369      * and grab the PIM REGISTER header size, to avoid another
3370      * possible m_pullup() later.
3371      *
3372      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
3373      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3374      */
3375     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3376     /*
3377      * Get the IP and PIM headers in contiguous memory, and
3378      * possibly the PIM REGISTER header.
3379      */
3380     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3381 	(m = m_pullup(m, minlen)) == NULL) {
3382 	log(LOG_ERR, "pim_input: m_pullup failure\n");
3383 	return;
3384     }
3385     /* m_pullup() may have given us a new mbuf so reset ip. */
3386     ip = mtod(m, struct ip *);
3387     ip_tos = ip->ip_tos;
3388 
3389     /* adjust mbuf to point to the PIM header */
3390     m->m_data += iphlen;
3391     m->m_len  -= iphlen;
3392     pim = mtod(m, struct pim *);
3393 
3394     /*
3395      * Validate checksum. If PIM REGISTER, exclude the data packet.
3396      *
3397      * XXX: some older PIMv2 implementations don't make this distinction,
3398      * so for compatibility reason perform the checksum over part of the
3399      * message, and if error, then over the whole message.
3400      */
3401     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3402 	/* do nothing, checksum okay */
3403     } else if (in_cksum(m, datalen)) {
3404 	pimstat.pims_rcv_badsum++;
3405 	if (mrtdebug & DEBUG_PIM)
3406 	    log(LOG_DEBUG, "pim_input: invalid checksum");
3407 	m_freem(m);
3408 	return;
3409     }
3410 
3411     /* PIM version check */
3412     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3413 	pimstat.pims_rcv_badversion++;
3414 	log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3415 	    PIM_VT_V(pim->pim_vt), PIM_VERSION);
3416 	m_freem(m);
3417 	return;
3418     }
3419 
3420     /* restore mbuf back to the outer IP */
3421     m->m_data -= iphlen;
3422     m->m_len  += iphlen;
3423 
3424     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3425 	/*
3426 	 * Since this is a REGISTER, we'll make a copy of the register
3427 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3428 	 * routing daemon.
3429 	 */
3430 	int s;
3431 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
3432 	struct mbuf *mcp;
3433 	struct ip *encap_ip;
3434 	u_int32_t *reghdr;
3435 	struct ifnet *vifp;
3436 
3437 	s = splsoftnet();
3438 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3439 	    splx(s);
3440 	    if (mrtdebug & DEBUG_PIM)
3441 		log(LOG_DEBUG,
3442 		    "pim_input: register vif not set: %d\n", reg_vif_num);
3443 	    m_freem(m);
3444 	    return;
3445 	}
3446 	/* XXX need refcnt? */
3447 	vifp = viftable[reg_vif_num].v_ifp;
3448 	splx(s);
3449 
3450 	/*
3451 	 * Validate length
3452 	 */
3453 	if (datalen < PIM_REG_MINLEN) {
3454 	    pimstat.pims_rcv_tooshort++;
3455 	    pimstat.pims_rcv_badregisters++;
3456 	    log(LOG_ERR,
3457 		"pim_input: register packet size too small %d from %lx\n",
3458 		datalen, (u_long)ip->ip_src.s_addr);
3459 	    m_freem(m);
3460 	    return;
3461 	}
3462 
3463 	reghdr = (u_int32_t *)(pim + 1);
3464 	encap_ip = (struct ip *)(reghdr + 1);
3465 
3466 	if (mrtdebug & DEBUG_PIM) {
3467 	    log(LOG_DEBUG,
3468 		"pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3469 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3470 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3471 		ntohs(encap_ip->ip_len));
3472 	}
3473 
3474 	/* verify the version number of the inner packet */
3475 	if (encap_ip->ip_v != IPVERSION) {
3476 	    pimstat.pims_rcv_badregisters++;
3477 	    if (mrtdebug & DEBUG_PIM) {
3478 		log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3479 		    "of the inner packet\n", encap_ip->ip_v);
3480 	    }
3481 	    m_freem(m);
3482 	    return;
3483 	}
3484 
3485 	/* verify the inner packet is destined to a mcast group */
3486 	if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3487 	    pimstat.pims_rcv_badregisters++;
3488 	    if (mrtdebug & DEBUG_PIM)
3489 		log(LOG_DEBUG,
3490 		    "pim_input: inner packet of register is not "
3491 		    "multicast %lx\n",
3492 		    (u_long)ntohl(encap_ip->ip_dst.s_addr));
3493 	    m_freem(m);
3494 	    return;
3495 	}
3496 
3497 	/* If a NULL_REGISTER, pass it to the daemon */
3498 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3499 	    goto pim_input_to_daemon;
3500 
3501 	/*
3502 	 * Copy the TOS from the outer IP header to the inner IP header.
3503 	 */
3504 	if (encap_ip->ip_tos != ip_tos) {
3505 	    /* Outer TOS -> inner TOS */
3506 	    encap_ip->ip_tos = ip_tos;
3507 	    /* Recompute the inner header checksum. Sigh... */
3508 
3509 	    /* adjust mbuf to point to the inner IP header */
3510 	    m->m_data += (iphlen + PIM_MINLEN);
3511 	    m->m_len  -= (iphlen + PIM_MINLEN);
3512 
3513 	    encap_ip->ip_sum = 0;
3514 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3515 
3516 	    /* restore mbuf to point back to the outer IP header */
3517 	    m->m_data -= (iphlen + PIM_MINLEN);
3518 	    m->m_len  += (iphlen + PIM_MINLEN);
3519 	}
3520 
3521 	/*
3522 	 * Decapsulate the inner IP packet and loopback to forward it
3523 	 * as a normal multicast packet. Also, make a copy of the
3524 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
3525 	 * to pass to the daemon later, so it can take the appropriate
3526 	 * actions (e.g., send back PIM_REGISTER_STOP).
3527 	 * XXX: here m->m_data points to the outer IP header.
3528 	 */
3529 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
3530 	if (mcp == NULL) {
3531 	    log(LOG_ERR,
3532 		"pim_input: pim register: could not copy register head\n");
3533 	    m_freem(m);
3534 	    return;
3535 	}
3536 
3537 	/* Keep statistics */
3538 	/* XXX: registers_bytes include only the encap. mcast pkt */
3539 	pimstat.pims_rcv_registers_msgs++;
3540 	pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3541 
3542 	/*
3543 	 * forward the inner ip packet; point m_data at the inner ip.
3544 	 */
3545 	m_adj(m, iphlen + PIM_MINLEN);
3546 
3547 	if (mrtdebug & DEBUG_PIM) {
3548 	    log(LOG_DEBUG,
3549 		"pim_input: forwarding decapsulated register: "
3550 		"src %lx, dst %lx, vif %d\n",
3551 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3552 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3553 		reg_vif_num);
3554 	}
3555 	/* NB: vifp was collected above; can it change on us? */
3556 	looutput(vifp, m, (struct sockaddr *)&dst, (struct rtentry *)NULL);
3557 
3558 	/* prepare the register head to send to the mrouting daemon */
3559 	m = mcp;
3560     }
3561 
3562 pim_input_to_daemon:
3563     /*
3564      * Pass the PIM message up to the daemon; if it is a Register message,
3565      * pass the 'head' only up to the daemon. This includes the
3566      * outer IP header, PIM header, PIM-Register header and the
3567      * inner IP header.
3568      * XXX: the outer IP header pkt size of a Register is not adjust to
3569      * reflect the fact that the inner multicast data is truncated.
3570      */
3571     rip_input(m, iphlen, proto);
3572 
3573     return;
3574 }
3575 #endif /* PIM */
3576