xref: /netbsd-src/sys/netinet/ip_mroute.c (revision 63aea4bd5b445e491ff0389fe27ec78b3099dba3)
1 /*	$NetBSD: ip_mroute.c,v 1.132 2015/08/24 22:21:26 pooka 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 1.132 2015/08/24 22:21:26 pooka Exp $");
97 
98 #ifdef _KERNEL_OPT
99 #include "opt_inet.h"
100 #include "opt_ipsec.h"
101 #include "opt_pim.h"
102 #endif
103 
104 #ifdef PIM
105 #define _PIM_VT 1
106 #endif
107 
108 #include <sys/param.h>
109 #include <sys/systm.h>
110 #include <sys/callout.h>
111 #include <sys/mbuf.h>
112 #include <sys/socket.h>
113 #include <sys/socketvar.h>
114 #include <sys/protosw.h>
115 #include <sys/errno.h>
116 #include <sys/time.h>
117 #include <sys/kernel.h>
118 #include <sys/kmem.h>
119 #include <sys/ioctl.h>
120 #include <sys/syslog.h>
121 
122 #include <net/if.h>
123 #include <net/route.h>
124 #include <net/raw_cb.h>
125 
126 #include <netinet/in.h>
127 #include <netinet/in_var.h>
128 #include <netinet/in_systm.h>
129 #include <netinet/ip.h>
130 #include <netinet/ip_var.h>
131 #include <netinet/in_pcb.h>
132 #include <netinet/udp.h>
133 #include <netinet/igmp.h>
134 #include <netinet/igmp_var.h>
135 #include <netinet/ip_mroute.h>
136 #ifdef PIM
137 #include <netinet/pim.h>
138 #include <netinet/pim_var.h>
139 #endif
140 #include <netinet/ip_encap.h>
141 
142 #ifdef IPSEC
143 #include <netipsec/ipsec.h>
144 #include <netipsec/key.h>
145 #endif
146 
147 #define IP_MULTICASTOPTS 0
148 #define	M_PULLUP(m, len)						 \
149 	do {								 \
150 		if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
151 			(m) = m_pullup((m), (len));			 \
152 	} while (/*CONSTCOND*/ 0)
153 
154 /*
155  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
156  * except for netstat or debugging purposes.
157  */
158 struct socket  *ip_mrouter  = NULL;
159 int		ip_mrtproto = IGMP_DVMRP;    /* for netstat only */
160 
161 #define NO_RTE_FOUND 	0x1
162 #define RTE_FOUND	0x2
163 
164 #define	MFCHASH(a, g)							\
165 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^	\
166 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
167 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
168 u_long	mfchash;
169 
170 u_char		nexpire[MFCTBLSIZ];
171 struct vif	viftable[MAXVIFS];
172 struct mrtstat	mrtstat;
173 u_int		mrtdebug = 0;	  /* debug level 	*/
174 #define		DEBUG_MFC	0x02
175 #define		DEBUG_FORWARD	0x04
176 #define		DEBUG_EXPIRE	0x08
177 #define		DEBUG_XMIT	0x10
178 #define		DEBUG_PIM	0x20
179 
180 #define		VIFI_INVALID	((vifi_t) -1)
181 
182 u_int       	tbfdebug = 0;     /* tbf debug level 	*/
183 #ifdef RSVP_ISI
184 u_int		rsvpdebug = 0;	  /* rsvp debug level   */
185 extern struct socket *ip_rsvpd;
186 extern int rsvp_on;
187 #endif /* RSVP_ISI */
188 
189 /* vif attachment using sys/netinet/ip_encap.c */
190 static void vif_input(struct mbuf *, ...);
191 static int vif_encapcheck(struct mbuf *, int, int, void *);
192 
193 static const struct protosw vif_protosw = {
194 	.pr_type	= SOCK_RAW,
195 	.pr_domain	= &inetdomain,
196 	.pr_protocol	= IPPROTO_IPV4,
197 	.pr_flags	= PR_ATOMIC|PR_ADDR,
198 	.pr_input	= vif_input,
199 	.pr_output	= rip_output,
200 	.pr_ctloutput	= rip_ctloutput,
201 	.pr_usrreqs	= &rip_usrreqs,
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 *, int);
216 static int set_assert(int);
217 static int add_vif(struct vifctl *);
218 static int del_vif(vifi_t *);
219 static void update_mfc_params(struct mfc *, struct mfcctl2 *);
220 static void init_mfc_params(struct mfc *, struct mfcctl2 *);
221 static void expire_mfc(struct mfc *);
222 static int add_mfc(struct sockopt *);
223 #ifdef UPCALL_TIMING
224 static void collate(struct timeval *);
225 #endif
226 static int del_mfc(struct sockopt *);
227 static int set_api_config(struct sockopt *); /* chose API capabilities */
228 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
229 static void expire_upcalls(void *);
230 #ifdef RSVP_ISI
231 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
232 #else
233 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *);
234 #endif
235 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
236 static void encap_send(struct ip *, struct vif *, struct mbuf *);
237 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_int32_t);
238 static void tbf_queue(struct vif *, struct mbuf *);
239 static void tbf_process_q(struct vif *);
240 static void tbf_reprocess_q(void *);
241 static int tbf_dq_sel(struct vif *, struct ip *);
242 static void tbf_send_packet(struct vif *, struct mbuf *);
243 static void tbf_update_tokens(struct vif *);
244 static int priority(struct vif *, struct ip *);
245 
246 /*
247  * Bandwidth monitoring
248  */
249 static void free_bw_list(struct bw_meter *);
250 static int add_bw_upcall(struct bw_upcall *);
251 static int del_bw_upcall(struct bw_upcall *);
252 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *);
253 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
254 static void bw_upcalls_send(void);
255 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
256 static void unschedule_bw_meter(struct bw_meter *);
257 static void bw_meter_process(void);
258 static void expire_bw_upcalls_send(void *);
259 static void expire_bw_meter_process(void *);
260 
261 #ifdef PIM
262 static int pim_register_send(struct ip *, struct vif *,
263 		struct mbuf *, struct mfc *);
264 static int pim_register_send_rp(struct ip *, struct vif *,
265 		struct mbuf *, struct mfc *);
266 static int pim_register_send_upcall(struct ip *, struct vif *,
267 		struct mbuf *, struct mfc *);
268 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
269 #endif
270 
271 /*
272  * 'Interfaces' associated with decapsulator (so we can tell
273  * packets that went through it from ones that get reflected
274  * by a broken gateway).  These interfaces are never linked into
275  * the system ifnet list & no routes point to them.  I.e., packets
276  * can't be sent this way.  They only exist as a placeholder for
277  * multicast source verification.
278  */
279 #if 0
280 struct ifnet multicast_decap_if[MAXVIFS];
281 #endif
282 
283 #define	ENCAP_TTL	64
284 #define	ENCAP_PROTO	IPPROTO_IPIP	/* 4 */
285 
286 /* prototype IP hdr for encapsulated packets */
287 struct ip multicast_encap_iphdr = {
288 	.ip_hl = sizeof(struct ip) >> 2,
289 	.ip_v = IPVERSION,
290 	.ip_len = sizeof(struct ip),
291 	.ip_ttl = ENCAP_TTL,
292 	.ip_p = ENCAP_PROTO,
293 };
294 
295 /*
296  * Bandwidth meter variables and constants
297  */
298 
299 /*
300  * Pending timeouts are stored in a hash table, the key being the
301  * expiration time. Periodically, the entries are analysed and processed.
302  */
303 #define BW_METER_BUCKETS	1024
304 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
305 struct callout bw_meter_ch;
306 #define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
307 
308 /*
309  * Pending upcalls are stored in a vector which is flushed when
310  * full, or periodically
311  */
312 static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
313 static u_int	bw_upcalls_n; /* # of pending upcalls */
314 struct callout	bw_upcalls_ch;
315 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
316 
317 #ifdef PIM
318 struct pimstat pimstat;
319 
320 /*
321  * Note: the PIM Register encapsulation adds the following in front of a
322  * data packet:
323  *
324  * struct pim_encap_hdr {
325  *    struct ip ip;
326  *    struct pim_encap_pimhdr  pim;
327  * }
328  *
329  */
330 
331 struct pim_encap_pimhdr {
332 	struct pim pim;
333 	uint32_t   flags;
334 };
335 
336 static struct ip pim_encap_iphdr = {
337 	.ip_v = IPVERSION,
338 	.ip_hl = sizeof(struct ip) >> 2,
339 	.ip_len = sizeof(struct ip),
340 	.ip_ttl = ENCAP_TTL,
341 	.ip_p = IPPROTO_PIM,
342 };
343 
344 static struct pim_encap_pimhdr pim_encap_pimhdr = {
345     {
346 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
347 	0,			/* reserved */
348 	0,			/* checksum */
349     },
350     0				/* flags */
351 };
352 
353 static struct ifnet multicast_register_if;
354 static vifi_t reg_vif_num = VIFI_INVALID;
355 #endif /* PIM */
356 
357 
358 /*
359  * Private variables.
360  */
361 static vifi_t	   numvifs = 0;
362 
363 static struct callout expire_upcalls_ch;
364 
365 /*
366  * whether or not special PIM assert processing is enabled.
367  */
368 static int pim_assert;
369 /*
370  * Rate limit for assert notification messages, in usec
371  */
372 #define ASSERT_MSG_TIME		3000000
373 
374 /*
375  * Kernel multicast routing API capabilities and setup.
376  * If more API capabilities are added to the kernel, they should be
377  * recorded in `mrt_api_support'.
378  */
379 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
380 					  MRT_MFC_FLAGS_BORDER_VIF |
381 					  MRT_MFC_RP |
382 					  MRT_MFC_BW_UPCALL);
383 static u_int32_t mrt_api_config = 0;
384 
385 /*
386  * Find a route for a given origin IP address and Multicast group address
387  * Type of service parameter to be added in the future!!!
388  * Statistics are updated by the caller if needed
389  * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
390  */
391 static struct mfc *
392 mfc_find(struct in_addr *o, struct in_addr *g)
393 {
394 	struct mfc *rt;
395 
396 	LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
397 		if (in_hosteq(rt->mfc_origin, *o) &&
398 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
399 		    (rt->mfc_stall == NULL))
400 			break;
401 	}
402 
403 	return (rt);
404 }
405 
406 /*
407  * Macros to compute elapsed time efficiently
408  * Borrowed from Van Jacobson's scheduling code
409  */
410 #define TV_DELTA(a, b, delta) do {					\
411 	int xxs;							\
412 	delta = (a).tv_usec - (b).tv_usec;				\
413 	xxs = (a).tv_sec - (b).tv_sec;					\
414 	switch (xxs) {							\
415 	case 2:								\
416 		delta += 1000000;					\
417 		/* fall through */					\
418 	case 1:								\
419 		delta += 1000000;					\
420 		/* fall through */					\
421 	case 0:								\
422 		break;							\
423 	default:							\
424 		delta += (1000000 * xxs);				\
425 		break;							\
426 	}								\
427 } while (/*CONSTCOND*/ 0)
428 
429 #ifdef UPCALL_TIMING
430 u_int32_t upcall_data[51];
431 #endif /* UPCALL_TIMING */
432 
433 /*
434  * Handle MRT setsockopt commands to modify the multicast routing tables.
435  */
436 int
437 ip_mrouter_set(struct socket *so, struct sockopt *sopt)
438 {
439 	int error;
440 	int optval;
441 	struct vifctl vifc;
442 	vifi_t vifi;
443 	struct bw_upcall bwuc;
444 
445 	if (sopt->sopt_name != MRT_INIT && so != ip_mrouter)
446 		error = ENOPROTOOPT;
447 	else {
448 		switch (sopt->sopt_name) {
449 		case MRT_INIT:
450 			error = sockopt_getint(sopt, &optval);
451 			if (error)
452 				break;
453 
454 			error = ip_mrouter_init(so, optval);
455 			break;
456 		case MRT_DONE:
457 			error = ip_mrouter_done();
458 			break;
459 		case MRT_ADD_VIF:
460 			error = sockopt_get(sopt, &vifc, sizeof(vifc));
461 			if (error)
462 				break;
463 			error = add_vif(&vifc);
464 			break;
465 		case MRT_DEL_VIF:
466 			error = sockopt_get(sopt, &vifi, sizeof(vifi));
467 			if (error)
468 				break;
469 			error = del_vif(&vifi);
470 			break;
471 		case MRT_ADD_MFC:
472 			error = add_mfc(sopt);
473 			break;
474 		case MRT_DEL_MFC:
475 			error = del_mfc(sopt);
476 			break;
477 		case MRT_ASSERT:
478 			error = sockopt_getint(sopt, &optval);
479 			if (error)
480 				break;
481 			error = set_assert(optval);
482 			break;
483 		case MRT_API_CONFIG:
484 			error = set_api_config(sopt);
485 			break;
486 		case MRT_ADD_BW_UPCALL:
487 			error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
488 			if (error)
489 				break;
490 			error = add_bw_upcall(&bwuc);
491 			break;
492 		case MRT_DEL_BW_UPCALL:
493 			error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
494 			if (error)
495 				break;
496 			error = del_bw_upcall(&bwuc);
497 			break;
498 		default:
499 			error = ENOPROTOOPT;
500 			break;
501 		}
502 	}
503 	return (error);
504 }
505 
506 /*
507  * Handle MRT getsockopt commands
508  */
509 int
510 ip_mrouter_get(struct socket *so, struct sockopt *sopt)
511 {
512 	int error;
513 
514 	if (so != ip_mrouter)
515 		error = ENOPROTOOPT;
516 	else {
517 		switch (sopt->sopt_name) {
518 		case MRT_VERSION:
519 			error = sockopt_setint(sopt, 0x0305); /* XXX !!!! */
520 			break;
521 		case MRT_ASSERT:
522 			error = sockopt_setint(sopt, pim_assert);
523 			break;
524 		case MRT_API_SUPPORT:
525 			error = sockopt_set(sopt, &mrt_api_support,
526 			    sizeof(mrt_api_support));
527 			break;
528 		case MRT_API_CONFIG:
529 			error = sockopt_set(sopt, &mrt_api_config,
530 			    sizeof(mrt_api_config));
531 			break;
532 		default:
533 			error = ENOPROTOOPT;
534 			break;
535 		}
536 	}
537 	return (error);
538 }
539 
540 /*
541  * Handle ioctl commands to obtain information from the cache
542  */
543 int
544 mrt_ioctl(struct socket *so, u_long cmd, void *data)
545 {
546 	int error;
547 
548 	if (so != ip_mrouter)
549 		error = EINVAL;
550 	else
551 		switch (cmd) {
552 		case SIOCGETVIFCNT:
553 			error = get_vif_cnt((struct sioc_vif_req *)data);
554 			break;
555 		case SIOCGETSGCNT:
556 			error = get_sg_cnt((struct sioc_sg_req *)data);
557 			break;
558 		default:
559 			error = EINVAL;
560 			break;
561 		}
562 
563 	return (error);
564 }
565 
566 /*
567  * returns the packet, byte, rpf-failure count for the source group provided
568  */
569 static int
570 get_sg_cnt(struct sioc_sg_req *req)
571 {
572 	int s;
573 	struct mfc *rt;
574 
575 	s = splsoftnet();
576 	rt = mfc_find(&req->src, &req->grp);
577 	if (rt == NULL) {
578 		splx(s);
579 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
580 		return (EADDRNOTAVAIL);
581 	}
582 	req->pktcnt = rt->mfc_pkt_cnt;
583 	req->bytecnt = rt->mfc_byte_cnt;
584 	req->wrong_if = rt->mfc_wrong_if;
585 	splx(s);
586 
587 	return (0);
588 }
589 
590 /*
591  * returns the input and output packet and byte counts on the vif provided
592  */
593 static int
594 get_vif_cnt(struct sioc_vif_req *req)
595 {
596 	vifi_t vifi = req->vifi;
597 
598 	if (vifi >= numvifs)
599 		return (EINVAL);
600 
601 	req->icount = viftable[vifi].v_pkt_in;
602 	req->ocount = viftable[vifi].v_pkt_out;
603 	req->ibytes = viftable[vifi].v_bytes_in;
604 	req->obytes = viftable[vifi].v_bytes_out;
605 
606 	return (0);
607 }
608 
609 /*
610  * Enable multicast routing
611  */
612 static int
613 ip_mrouter_init(struct socket *so, int v)
614 {
615 	if (mrtdebug)
616 		log(LOG_DEBUG,
617 		    "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
618 		    so->so_type, so->so_proto->pr_protocol);
619 
620 	if (so->so_type != SOCK_RAW ||
621 	    so->so_proto->pr_protocol != IPPROTO_IGMP)
622 		return (EOPNOTSUPP);
623 
624 	if (v != 1)
625 		return (EINVAL);
626 
627 	if (ip_mrouter != NULL)
628 		return (EADDRINUSE);
629 
630 	ip_mrouter = so;
631 
632 	mfchashtbl = hashinit(MFCTBLSIZ, HASH_LIST, true, &mfchash);
633 	memset((void *)nexpire, 0, sizeof(nexpire));
634 
635 	pim_assert = 0;
636 
637 	callout_init(&expire_upcalls_ch, 0);
638 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
639 		      expire_upcalls, NULL);
640 
641 	callout_init(&bw_upcalls_ch, 0);
642 	callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
643 		      expire_bw_upcalls_send, NULL);
644 
645 	callout_init(&bw_meter_ch, 0);
646 	callout_reset(&bw_meter_ch, BW_METER_PERIOD,
647 		      expire_bw_meter_process, NULL);
648 
649 	if (mrtdebug)
650 		log(LOG_DEBUG, "ip_mrouter_init\n");
651 
652 	return (0);
653 }
654 
655 /*
656  * Disable multicast routing
657  */
658 int
659 ip_mrouter_done(void)
660 {
661 	vifi_t vifi;
662 	struct vif *vifp;
663 	int i;
664 	int s;
665 
666 	s = splsoftnet();
667 
668 	/* Clear out all the vifs currently in use. */
669 	for (vifi = 0; vifi < numvifs; vifi++) {
670 		vifp = &viftable[vifi];
671 		if (!in_nullhost(vifp->v_lcl_addr))
672 			reset_vif(vifp);
673 	}
674 
675 	numvifs = 0;
676 	pim_assert = 0;
677 	mrt_api_config = 0;
678 
679 	callout_stop(&expire_upcalls_ch);
680 	callout_stop(&bw_upcalls_ch);
681 	callout_stop(&bw_meter_ch);
682 
683 	/*
684 	 * Free all multicast forwarding cache entries.
685 	 */
686 	for (i = 0; i < MFCTBLSIZ; i++) {
687 		struct mfc *rt, *nrt;
688 
689 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
690 			nrt = LIST_NEXT(rt, mfc_hash);
691 
692 			expire_mfc(rt);
693 		}
694 	}
695 
696 	memset((void *)nexpire, 0, sizeof(nexpire));
697 	hashdone(mfchashtbl, HASH_LIST, mfchash);
698 	mfchashtbl = NULL;
699 
700 	bw_upcalls_n = 0;
701 	memset(bw_meter_timers, 0, sizeof(bw_meter_timers));
702 
703 	/* Reset de-encapsulation cache. */
704 
705 	ip_mrouter = NULL;
706 
707 	splx(s);
708 
709 	if (mrtdebug)
710 		log(LOG_DEBUG, "ip_mrouter_done\n");
711 
712 	return (0);
713 }
714 
715 void
716 ip_mrouter_detach(struct ifnet *ifp)
717 {
718 	int vifi, i;
719 	struct vif *vifp;
720 	struct mfc *rt;
721 	struct rtdetq *rte;
722 
723 	/* XXX not sure about side effect to userland routing daemon */
724 	for (vifi = 0; vifi < numvifs; vifi++) {
725 		vifp = &viftable[vifi];
726 		if (vifp->v_ifp == ifp)
727 			reset_vif(vifp);
728 	}
729 	for (i = 0; i < MFCTBLSIZ; i++) {
730 		if (nexpire[i] == 0)
731 			continue;
732 		LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
733 			for (rte = rt->mfc_stall; rte; rte = rte->next) {
734 				if (rte->ifp == ifp)
735 					rte->ifp = NULL;
736 			}
737 		}
738 	}
739 }
740 
741 /*
742  * Set PIM assert processing global
743  */
744 static int
745 set_assert(int i)
746 {
747 	pim_assert = !!i;
748 	return (0);
749 }
750 
751 /*
752  * Configure API capabilities
753  */
754 static int
755 set_api_config(struct sockopt *sopt)
756 {
757 	u_int32_t apival;
758 	int i, error;
759 
760 	/*
761 	 * We can set the API capabilities only if it is the first operation
762 	 * after MRT_INIT. I.e.:
763 	 *  - there are no vifs installed
764 	 *  - pim_assert is not enabled
765 	 *  - the MFC table is empty
766 	 */
767 	error = sockopt_get(sopt, &apival, sizeof(apival));
768 	if (error)
769 		return (error);
770 	if (numvifs > 0)
771 		return (EPERM);
772 	if (pim_assert)
773 		return (EPERM);
774 	for (i = 0; i < MFCTBLSIZ; i++) {
775 		if (LIST_FIRST(&mfchashtbl[i]) != NULL)
776 			return (EPERM);
777 	}
778 
779 	mrt_api_config = apival & mrt_api_support;
780 	return (0);
781 }
782 
783 /*
784  * Add a vif to the vif table
785  */
786 static int
787 add_vif(struct vifctl *vifcp)
788 {
789 	struct vif *vifp;
790 	struct ifaddr *ifa;
791 	struct ifnet *ifp;
792 	int error, s;
793 	struct sockaddr_in sin;
794 
795 	if (vifcp->vifc_vifi >= MAXVIFS)
796 		return (EINVAL);
797 	if (in_nullhost(vifcp->vifc_lcl_addr))
798 		return (EADDRNOTAVAIL);
799 
800 	vifp = &viftable[vifcp->vifc_vifi];
801 	if (!in_nullhost(vifp->v_lcl_addr))
802 		return (EADDRINUSE);
803 
804 	/* Find the interface with an address in AF_INET family. */
805 #ifdef PIM
806 	if (vifcp->vifc_flags & VIFF_REGISTER) {
807 		/*
808 		 * XXX: Because VIFF_REGISTER does not really need a valid
809 		 * local interface (e.g. it could be 127.0.0.2), we don't
810 		 * check its address.
811 		 */
812 	    ifp = NULL;
813 	} else
814 #endif
815 	{
816 		sockaddr_in_init(&sin, &vifcp->vifc_lcl_addr, 0);
817 		ifa = ifa_ifwithaddr(sintosa(&sin));
818 		if (ifa == NULL)
819 			return (EADDRNOTAVAIL);
820 		ifp = ifa->ifa_ifp;
821 	}
822 
823 	if (vifcp->vifc_flags & VIFF_TUNNEL) {
824 		if (vifcp->vifc_flags & VIFF_SRCRT) {
825 			log(LOG_ERR, "source routed tunnels not supported\n");
826 			return (EOPNOTSUPP);
827 		}
828 
829 		/* attach this vif to decapsulator dispatch table */
830 		/*
831 		 * XXX Use addresses in registration so that matching
832 		 * can be done with radix tree in decapsulator.  But,
833 		 * we need to check inner header for multicast, so
834 		 * this requires both radix tree lookup and then a
835 		 * function to check, and this is not supported yet.
836 		 */
837 		vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
838 		    vif_encapcheck, &vif_protosw, vifp);
839 		if (!vifp->v_encap_cookie)
840 			return (EINVAL);
841 
842 		/* Create a fake encapsulation interface. */
843 		ifp = malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK|M_ZERO);
844 		snprintf(ifp->if_xname, sizeof(ifp->if_xname),
845 			 "mdecap%d", vifcp->vifc_vifi);
846 
847 		/* Prepare cached route entry. */
848 		memset(&vifp->v_route, 0, sizeof(vifp->v_route));
849 #ifdef PIM
850 	} else if (vifcp->vifc_flags & VIFF_REGISTER) {
851 		ifp = &multicast_register_if;
852 		if (mrtdebug)
853 			log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
854 			    (void *)ifp);
855 		if (reg_vif_num == VIFI_INVALID) {
856 			memset(ifp, 0, sizeof(*ifp));
857 			snprintf(ifp->if_xname, sizeof(ifp->if_xname),
858 				 "register_vif");
859 			ifp->if_flags = IFF_LOOPBACK;
860 			memset(&vifp->v_route, 0, sizeof(vifp->v_route));
861 			reg_vif_num = vifcp->vifc_vifi;
862 		}
863 #endif
864 	} else {
865 		/* Make sure the interface supports multicast. */
866 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
867 			return (EOPNOTSUPP);
868 
869 		/* Enable promiscuous reception of all IP multicasts. */
870 		sockaddr_in_init(&sin, &zeroin_addr, 0);
871 		error = if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin));
872 		if (error)
873 			return (error);
874 	}
875 
876 	s = splsoftnet();
877 
878 	/* Define parameters for the tbf structure. */
879 	vifp->tbf_q = NULL;
880 	vifp->tbf_t = &vifp->tbf_q;
881 	microtime(&vifp->tbf_last_pkt_t);
882 	vifp->tbf_n_tok = 0;
883 	vifp->tbf_q_len = 0;
884 	vifp->tbf_max_q_len = MAXQSIZE;
885 
886 	vifp->v_flags = vifcp->vifc_flags;
887 	vifp->v_threshold = vifcp->vifc_threshold;
888 	/* scaling up here allows division by 1024 in critical code */
889 	vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
890 	vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
891 	vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
892 	vifp->v_ifp = ifp;
893 	/* Initialize per vif pkt counters. */
894 	vifp->v_pkt_in = 0;
895 	vifp->v_pkt_out = 0;
896 	vifp->v_bytes_in = 0;
897 	vifp->v_bytes_out = 0;
898 
899 	callout_init(&vifp->v_repq_ch, 0);
900 
901 #ifdef RSVP_ISI
902 	vifp->v_rsvp_on = 0;
903 	vifp->v_rsvpd = NULL;
904 #endif /* RSVP_ISI */
905 
906 	splx(s);
907 
908 	/* Adjust numvifs up if the vifi is higher than numvifs. */
909 	if (numvifs <= vifcp->vifc_vifi)
910 		numvifs = vifcp->vifc_vifi + 1;
911 
912 	if (mrtdebug)
913 		log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
914 		    vifcp->vifc_vifi,
915 		    ntohl(vifcp->vifc_lcl_addr.s_addr),
916 		    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
917 		    ntohl(vifcp->vifc_rmt_addr.s_addr),
918 		    vifcp->vifc_threshold,
919 		    vifcp->vifc_rate_limit);
920 
921 	return (0);
922 }
923 
924 void
925 reset_vif(struct vif *vifp)
926 {
927 	struct mbuf *m, *n;
928 	struct ifnet *ifp;
929 	struct sockaddr_in sin;
930 
931 	callout_stop(&vifp->v_repq_ch);
932 
933 	/* detach this vif from decapsulator dispatch table */
934 	encap_detach(vifp->v_encap_cookie);
935 	vifp->v_encap_cookie = NULL;
936 
937 	/*
938 	 * Free packets queued at the interface
939 	 */
940 	for (m = vifp->tbf_q; m != NULL; m = n) {
941 		n = m->m_nextpkt;
942 		m_freem(m);
943 	}
944 
945 	if (vifp->v_flags & VIFF_TUNNEL)
946 		free(vifp->v_ifp, M_MRTABLE);
947 	else if (vifp->v_flags & VIFF_REGISTER) {
948 #ifdef PIM
949 		reg_vif_num = VIFI_INVALID;
950 #endif
951 	} else {
952 		sockaddr_in_init(&sin, &zeroin_addr, 0);
953 		ifp = vifp->v_ifp;
954 		if_mcast_op(ifp, SIOCDELMULTI, sintosa(&sin));
955 	}
956 	memset((void *)vifp, 0, sizeof(*vifp));
957 }
958 
959 /*
960  * Delete a vif from the vif table
961  */
962 static int
963 del_vif(vifi_t *vifip)
964 {
965 	struct vif *vifp;
966 	vifi_t vifi;
967 	int s;
968 
969 	if (*vifip >= numvifs)
970 		return (EINVAL);
971 
972 	vifp = &viftable[*vifip];
973 	if (in_nullhost(vifp->v_lcl_addr))
974 		return (EADDRNOTAVAIL);
975 
976 	s = splsoftnet();
977 
978 	reset_vif(vifp);
979 
980 	/* Adjust numvifs down */
981 	for (vifi = numvifs; vifi > 0; vifi--)
982 		if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
983 			break;
984 	numvifs = vifi;
985 
986 	splx(s);
987 
988 	if (mrtdebug)
989 		log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
990 
991 	return (0);
992 }
993 
994 /*
995  * update an mfc entry without resetting counters and S,G addresses.
996  */
997 static void
998 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
999 {
1000 	int i;
1001 
1002 	rt->mfc_parent = mfccp->mfcc_parent;
1003 	for (i = 0; i < numvifs; i++) {
1004 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1005 		rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1006 			MRT_MFC_FLAGS_ALL;
1007 	}
1008 	/* set the RP address */
1009 	if (mrt_api_config & MRT_MFC_RP)
1010 		rt->mfc_rp = mfccp->mfcc_rp;
1011 	else
1012 		rt->mfc_rp = zeroin_addr;
1013 }
1014 
1015 /*
1016  * fully initialize an mfc entry from the parameter.
1017  */
1018 static void
1019 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1020 {
1021 	rt->mfc_origin     = mfccp->mfcc_origin;
1022 	rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1023 
1024 	update_mfc_params(rt, mfccp);
1025 
1026 	/* initialize pkt counters per src-grp */
1027 	rt->mfc_pkt_cnt    = 0;
1028 	rt->mfc_byte_cnt   = 0;
1029 	rt->mfc_wrong_if   = 0;
1030 	timerclear(&rt->mfc_last_assert);
1031 }
1032 
1033 static void
1034 expire_mfc(struct mfc *rt)
1035 {
1036 	struct rtdetq *rte, *nrte;
1037 
1038 	free_bw_list(rt->mfc_bw_meter);
1039 
1040 	for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1041 		nrte = rte->next;
1042 		m_freem(rte->m);
1043 		free(rte, M_MRTABLE);
1044 	}
1045 
1046 	LIST_REMOVE(rt, mfc_hash);
1047 	free(rt, M_MRTABLE);
1048 }
1049 
1050 /*
1051  * Add an mfc entry
1052  */
1053 static int
1054 add_mfc(struct sockopt *sopt)
1055 {
1056 	struct mfcctl2 mfcctl2;
1057 	struct mfcctl2 *mfccp;
1058 	struct mfc *rt;
1059 	u_int32_t hash = 0;
1060 	struct rtdetq *rte, *nrte;
1061 	u_short nstl;
1062 	int s;
1063 	int error;
1064 
1065 	/*
1066 	 * select data size depending on API version.
1067 	 */
1068 	mfccp = &mfcctl2;
1069 	memset(&mfcctl2, 0, sizeof(mfcctl2));
1070 
1071 	if (mrt_api_config & MRT_API_FLAGS_ALL)
1072 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1073 	else
1074 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1075 
1076 	if (error)
1077 		return (error);
1078 
1079 	s = splsoftnet();
1080 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1081 
1082 	/* If an entry already exists, just update the fields */
1083 	if (rt) {
1084 		if (mrtdebug & DEBUG_MFC)
1085 			log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1086 			    ntohl(mfccp->mfcc_origin.s_addr),
1087 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1088 			    mfccp->mfcc_parent);
1089 
1090 		update_mfc_params(rt, mfccp);
1091 
1092 		splx(s);
1093 		return (0);
1094 	}
1095 
1096 	/*
1097 	 * Find the entry for which the upcall was made and update
1098 	 */
1099 	nstl = 0;
1100 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1101 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1102 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1103 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1104 		    rt->mfc_stall != NULL) {
1105 			if (nstl++)
1106 				log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
1107 				    "multiple kernel entries",
1108 				    ntohl(mfccp->mfcc_origin.s_addr),
1109 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1110 				    mfccp->mfcc_parent, rt->mfc_stall);
1111 
1112 			if (mrtdebug & DEBUG_MFC)
1113 				log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n",
1114 				    ntohl(mfccp->mfcc_origin.s_addr),
1115 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1116 				    mfccp->mfcc_parent, rt->mfc_stall);
1117 
1118 			rte = rt->mfc_stall;
1119 			init_mfc_params(rt, mfccp);
1120 			rt->mfc_stall = NULL;
1121 
1122 			rt->mfc_expire = 0; /* Don't clean this guy up */
1123 			nexpire[hash]--;
1124 
1125 			/* free packets Qed at the end of this entry */
1126 			for (; rte != NULL; rte = nrte) {
1127 				nrte = rte->next;
1128 				if (rte->ifp) {
1129 #ifdef RSVP_ISI
1130 					ip_mdq(rte->m, rte->ifp, rt, -1);
1131 #else
1132 					ip_mdq(rte->m, rte->ifp, rt);
1133 #endif /* RSVP_ISI */
1134 				}
1135 				m_freem(rte->m);
1136 #ifdef UPCALL_TIMING
1137 				collate(&rte->t);
1138 #endif /* UPCALL_TIMING */
1139 				free(rte, M_MRTABLE);
1140 			}
1141 		}
1142 	}
1143 
1144 	/*
1145 	 * It is possible that an entry is being inserted without an upcall
1146 	 */
1147 	if (nstl == 0) {
1148 		/*
1149 		 * No mfc; make a new one
1150 		 */
1151 		if (mrtdebug & DEBUG_MFC)
1152 			log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1153 			    ntohl(mfccp->mfcc_origin.s_addr),
1154 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1155 			    mfccp->mfcc_parent);
1156 
1157 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1158 			if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1159 			    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1160 				init_mfc_params(rt, mfccp);
1161 				if (rt->mfc_expire)
1162 					nexpire[hash]--;
1163 				rt->mfc_expire = 0;
1164 				break; /* XXX */
1165 			}
1166 		}
1167 		if (rt == NULL) {	/* no upcall, so make a new entry */
1168 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1169 						  M_NOWAIT);
1170 			if (rt == NULL) {
1171 				splx(s);
1172 				return (ENOBUFS);
1173 			}
1174 
1175 			init_mfc_params(rt, mfccp);
1176 			rt->mfc_expire	= 0;
1177 			rt->mfc_stall	= NULL;
1178 			rt->mfc_bw_meter = NULL;
1179 
1180 			/* insert new entry at head of hash chain */
1181 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1182 		}
1183 	}
1184 
1185 	splx(s);
1186 	return (0);
1187 }
1188 
1189 #ifdef UPCALL_TIMING
1190 /*
1191  * collect delay statistics on the upcalls
1192  */
1193 static void
1194 collate(struct timeval *t)
1195 {
1196 	u_int32_t d;
1197 	struct timeval tp;
1198 	u_int32_t delta;
1199 
1200 	microtime(&tp);
1201 
1202 	if (timercmp(t, &tp, <)) {
1203 		TV_DELTA(tp, *t, delta);
1204 
1205 		d = delta >> 10;
1206 		if (d > 50)
1207 			d = 50;
1208 
1209 		++upcall_data[d];
1210 	}
1211 }
1212 #endif /* UPCALL_TIMING */
1213 
1214 /*
1215  * Delete an mfc entry
1216  */
1217 static int
1218 del_mfc(struct sockopt *sopt)
1219 {
1220 	struct mfcctl2 mfcctl2;
1221 	struct mfcctl2 *mfccp;
1222 	struct mfc *rt;
1223 	int s;
1224 	int error;
1225 
1226 	/*
1227 	 * XXX: for deleting MFC entries the information in entries
1228 	 * of size "struct mfcctl" is sufficient.
1229 	 */
1230 
1231 	mfccp = &mfcctl2;
1232 	memset(&mfcctl2, 0, sizeof(mfcctl2));
1233 
1234 	error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1235 	if (error) {
1236 		/* Try with the size of mfcctl2. */
1237 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1238 		if (error)
1239 			return (error);
1240 	}
1241 
1242 	if (mrtdebug & DEBUG_MFC)
1243 		log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1244 		    ntohl(mfccp->mfcc_origin.s_addr),
1245 		    ntohl(mfccp->mfcc_mcastgrp.s_addr));
1246 
1247 	s = splsoftnet();
1248 
1249 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1250 	if (rt == NULL) {
1251 		splx(s);
1252 		return (EADDRNOTAVAIL);
1253 	}
1254 
1255 	/*
1256 	 * free the bw_meter entries
1257 	 */
1258 	free_bw_list(rt->mfc_bw_meter);
1259 	rt->mfc_bw_meter = NULL;
1260 
1261 	LIST_REMOVE(rt, mfc_hash);
1262 	free(rt, M_MRTABLE);
1263 
1264 	splx(s);
1265 	return (0);
1266 }
1267 
1268 static int
1269 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1270 {
1271 	if (s) {
1272 		if (sbappendaddr(&s->so_rcv, sintosa(src), mm, NULL) != 0) {
1273 			sorwakeup(s);
1274 			return (0);
1275 		}
1276 	}
1277 	m_freem(mm);
1278 	return (-1);
1279 }
1280 
1281 /*
1282  * IP multicast forwarding function. This function assumes that the packet
1283  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1284  * pointed to by "ifp", and the packet is to be relayed to other networks
1285  * that have members of the packet's destination IP multicast group.
1286  *
1287  * The packet is returned unscathed to the caller, unless it is
1288  * erroneous, in which case a non-zero return value tells the caller to
1289  * discard it.
1290  */
1291 
1292 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1293 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1294 
1295 int
1296 #ifdef RSVP_ISI
1297 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo)
1298 #else
1299 ip_mforward(struct mbuf *m, struct ifnet *ifp)
1300 #endif /* RSVP_ISI */
1301 {
1302 	struct ip *ip = mtod(m, struct ip *);
1303 	struct mfc *rt;
1304 	static int srctun = 0;
1305 	struct mbuf *mm;
1306 	struct sockaddr_in sin;
1307 	int s;
1308 	vifi_t vifi;
1309 
1310 	if (mrtdebug & DEBUG_FORWARD)
1311 		log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1312 		    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1313 
1314 	if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1315 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1316 		/*
1317 		 * Packet arrived via a physical interface or
1318 		 * an encapsulated tunnel or a register_vif.
1319 		 */
1320 	} else {
1321 		/*
1322 		 * Packet arrived through a source-route tunnel.
1323 		 * Source-route tunnels are no longer supported.
1324 		 */
1325 		if ((srctun++ % 1000) == 0)
1326 			log(LOG_ERR,
1327 			    "ip_mforward: received source-routed packet from %x\n",
1328 			    ntohl(ip->ip_src.s_addr));
1329 
1330 		return (1);
1331 	}
1332 
1333 	/*
1334 	 * Clear any in-bound checksum flags for this packet.
1335 	 */
1336 	m->m_pkthdr.csum_flags = 0;
1337 
1338 #ifdef RSVP_ISI
1339 	if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1340 		if (ip->ip_ttl < MAXTTL)
1341 			ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1342 		if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1343 			struct vif *vifp = viftable + vifi;
1344 			printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n",
1345 			    ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1346 			    (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1347 			    vifp->v_ifp->if_xname);
1348 		}
1349 		return (ip_mdq(m, ifp, NULL, vifi));
1350 	}
1351 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1352 		printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
1353 		    ntohl(ip->ip_src), ntohl(ip->ip_dst));
1354 	}
1355 #endif /* RSVP_ISI */
1356 
1357 	/*
1358 	 * Don't forward a packet with time-to-live of zero or one,
1359 	 * or a packet destined to a local-only group.
1360 	 */
1361 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1362 		return (0);
1363 
1364 	/*
1365 	 * Determine forwarding vifs from the forwarding cache table
1366 	 */
1367 	s = splsoftnet();
1368 	++mrtstat.mrts_mfc_lookups;
1369 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1370 
1371 	/* Entry exists, so forward if necessary */
1372 	if (rt != NULL) {
1373 		splx(s);
1374 #ifdef RSVP_ISI
1375 		return (ip_mdq(m, ifp, rt, -1));
1376 #else
1377 		return (ip_mdq(m, ifp, rt));
1378 #endif /* RSVP_ISI */
1379 	} else {
1380 		/*
1381 		 * If we don't have a route for packet's origin,
1382 		 * Make a copy of the packet & send message to routing daemon
1383 		 */
1384 
1385 		struct mbuf *mb0;
1386 		struct rtdetq *rte;
1387 		u_int32_t hash;
1388 		int hlen = ip->ip_hl << 2;
1389 #ifdef UPCALL_TIMING
1390 		struct timeval tp;
1391 
1392 		microtime(&tp);
1393 #endif /* UPCALL_TIMING */
1394 
1395 		++mrtstat.mrts_mfc_misses;
1396 
1397 		mrtstat.mrts_no_route++;
1398 		if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1399 			log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1400 			    ntohl(ip->ip_src.s_addr),
1401 			    ntohl(ip->ip_dst.s_addr));
1402 
1403 		/*
1404 		 * Allocate mbufs early so that we don't do extra work if we are
1405 		 * just going to fail anyway.  Make sure to pullup the header so
1406 		 * that other people can't step on it.
1407 		 */
1408 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1409 					      M_NOWAIT);
1410 		if (rte == NULL) {
1411 			splx(s);
1412 			return (ENOBUFS);
1413 		}
1414 		mb0 = m_copypacket(m, M_DONTWAIT);
1415 		M_PULLUP(mb0, hlen);
1416 		if (mb0 == NULL) {
1417 			free(rte, M_MRTABLE);
1418 			splx(s);
1419 			return (ENOBUFS);
1420 		}
1421 
1422 		/* is there an upcall waiting for this flow? */
1423 		hash = MFCHASH(ip->ip_src, ip->ip_dst);
1424 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1425 			if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1426 			    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1427 			    rt->mfc_stall != NULL)
1428 				break;
1429 		}
1430 
1431 		if (rt == NULL) {
1432 			int i;
1433 			struct igmpmsg *im;
1434 
1435 			/*
1436 			 * Locate the vifi for the incoming interface for
1437 			 * this packet.
1438 			 * If none found, drop packet.
1439 			 */
1440 			for (vifi = 0; vifi < numvifs &&
1441 				 viftable[vifi].v_ifp != ifp; vifi++)
1442 				;
1443 			if (vifi >= numvifs) /* vif not found, drop packet */
1444 				goto non_fatal;
1445 
1446 			/* no upcall, so make a new entry */
1447 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1448 						  M_NOWAIT);
1449 			if (rt == NULL)
1450 				goto fail;
1451 
1452 			/*
1453 			 * Make a copy of the header to send to the user level
1454 			 * process
1455 			 */
1456 			mm = m_copym(m, 0, hlen, M_DONTWAIT);
1457 			M_PULLUP(mm, hlen);
1458 			if (mm == NULL)
1459 				goto fail1;
1460 
1461 			/*
1462 			 * Send message to routing daemon to install
1463 			 * a route into the kernel table
1464 			 */
1465 
1466 			im = mtod(mm, struct igmpmsg *);
1467 			im->im_msgtype = IGMPMSG_NOCACHE;
1468 			im->im_mbz = 0;
1469 			im->im_vif = vifi;
1470 
1471 			mrtstat.mrts_upcalls++;
1472 
1473 			sockaddr_in_init(&sin, &ip->ip_src, 0);
1474 			if (socket_send(ip_mrouter, mm, &sin) < 0) {
1475 				log(LOG_WARNING,
1476 				    "ip_mforward: ip_mrouter socket queue full\n");
1477 				++mrtstat.mrts_upq_sockfull;
1478 			fail1:
1479 				free(rt, M_MRTABLE);
1480 			fail:
1481 				free(rte, M_MRTABLE);
1482 				m_freem(mb0);
1483 				splx(s);
1484 				return (ENOBUFS);
1485 			}
1486 
1487 			/* insert new entry at head of hash chain */
1488 			rt->mfc_origin = ip->ip_src;
1489 			rt->mfc_mcastgrp = ip->ip_dst;
1490 			rt->mfc_pkt_cnt = 0;
1491 			rt->mfc_byte_cnt = 0;
1492 			rt->mfc_wrong_if = 0;
1493 			rt->mfc_expire = UPCALL_EXPIRE;
1494 			nexpire[hash]++;
1495 			for (i = 0; i < numvifs; i++) {
1496 				rt->mfc_ttls[i] = 0;
1497 				rt->mfc_flags[i] = 0;
1498 			}
1499 			rt->mfc_parent = -1;
1500 
1501 			/* clear the RP address */
1502 			rt->mfc_rp = zeroin_addr;
1503 
1504 			rt->mfc_bw_meter = NULL;
1505 
1506 			/* link into table */
1507 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1508 			/* Add this entry to the end of the queue */
1509 			rt->mfc_stall = rte;
1510 		} else {
1511 			/* determine if q has overflowed */
1512 			struct rtdetq **p;
1513 			int npkts = 0;
1514 
1515 			/*
1516 			 * XXX ouch! we need to append to the list, but we
1517 			 * only have a pointer to the front, so we have to
1518 			 * scan the entire list every time.
1519 			 */
1520 			for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1521 				if (++npkts > MAX_UPQ) {
1522 					mrtstat.mrts_upq_ovflw++;
1523 				non_fatal:
1524 					free(rte, M_MRTABLE);
1525 					m_freem(mb0);
1526 					splx(s);
1527 					return (0);
1528 				}
1529 
1530 			/* Add this entry to the end of the queue */
1531 			*p = rte;
1532 		}
1533 
1534 		rte->next = NULL;
1535 		rte->m = mb0;
1536 		rte->ifp = ifp;
1537 #ifdef UPCALL_TIMING
1538 		rte->t = tp;
1539 #endif /* UPCALL_TIMING */
1540 
1541 		splx(s);
1542 
1543 		return (0);
1544 	}
1545 }
1546 
1547 
1548 /*ARGSUSED*/
1549 static void
1550 expire_upcalls(void *v)
1551 {
1552 	int i;
1553 	int s;
1554 
1555 	s = splsoftnet();
1556 
1557 	for (i = 0; i < MFCTBLSIZ; i++) {
1558 		struct mfc *rt, *nrt;
1559 
1560 		if (nexpire[i] == 0)
1561 			continue;
1562 
1563 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1564 			nrt = LIST_NEXT(rt, mfc_hash);
1565 
1566 			if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1567 				continue;
1568 			nexpire[i]--;
1569 
1570 			/*
1571 			 * free the bw_meter entries
1572 			 */
1573 			while (rt->mfc_bw_meter != NULL) {
1574 				struct bw_meter *x = rt->mfc_bw_meter;
1575 
1576 				rt->mfc_bw_meter = x->bm_mfc_next;
1577 				kmem_free(x, sizeof(*x));
1578 			}
1579 
1580 			++mrtstat.mrts_cache_cleanups;
1581 			if (mrtdebug & DEBUG_EXPIRE)
1582 				log(LOG_DEBUG,
1583 				    "expire_upcalls: expiring (%x %x)\n",
1584 				    ntohl(rt->mfc_origin.s_addr),
1585 				    ntohl(rt->mfc_mcastgrp.s_addr));
1586 
1587 			expire_mfc(rt);
1588 		}
1589 	}
1590 
1591 	splx(s);
1592 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1593 	    expire_upcalls, NULL);
1594 }
1595 
1596 /*
1597  * Packet forwarding routine once entry in the cache is made
1598  */
1599 static int
1600 #ifdef RSVP_ISI
1601 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1602 #else
1603 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1604 #endif /* RSVP_ISI */
1605 {
1606 	struct ip  *ip = mtod(m, struct ip *);
1607 	vifi_t vifi;
1608 	struct vif *vifp;
1609 	struct sockaddr_in sin;
1610 	int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1611 
1612 /*
1613  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1614  * input, they shouldn't get counted on output, so statistics keeping is
1615  * separate.
1616  */
1617 #define MC_SEND(ip, vifp, m) do {					\
1618 	if ((vifp)->v_flags & VIFF_TUNNEL)				\
1619 		encap_send((ip), (vifp), (m));				\
1620 	else								\
1621 		phyint_send((ip), (vifp), (m));				\
1622 } while (/*CONSTCOND*/ 0)
1623 
1624 #ifdef RSVP_ISI
1625 	/*
1626 	 * If xmt_vif is not -1, send on only the requested vif.
1627 	 *
1628 	 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1629 	 */
1630 	if (xmt_vif < numvifs) {
1631 #ifdef PIM
1632 		if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1633 			pim_register_send(ip, viftable + xmt_vif, m, rt);
1634 		else
1635 #endif
1636 		MC_SEND(ip, viftable + xmt_vif, m);
1637 		return (1);
1638 	}
1639 #endif /* RSVP_ISI */
1640 
1641 	/*
1642 	 * Don't forward if it didn't arrive from the parent vif for its origin.
1643 	 */
1644 	vifi = rt->mfc_parent;
1645 	if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1646 		/* came in the wrong interface */
1647 		if (mrtdebug & DEBUG_FORWARD)
1648 			log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1649 			    ifp, vifi,
1650 			    vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1651 		++mrtstat.mrts_wrong_if;
1652 		++rt->mfc_wrong_if;
1653 		/*
1654 		 * If we are doing PIM assert processing, send a message
1655 		 * to the routing daemon.
1656 		 *
1657 		 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1658 		 * can complete the SPT switch, regardless of the type
1659 		 * of the iif (broadcast media, GRE tunnel, etc).
1660 		 */
1661 		if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1662 			struct timeval now;
1663 			u_int32_t delta;
1664 
1665 #ifdef PIM
1666 			if (ifp == &multicast_register_if)
1667 				pimstat.pims_rcv_registers_wrongiif++;
1668 #endif
1669 
1670 			/* Get vifi for the incoming packet */
1671 			for (vifi = 0;
1672 			     vifi < numvifs && viftable[vifi].v_ifp != ifp;
1673 			     vifi++)
1674 			    ;
1675 			if (vifi >= numvifs) {
1676 				/* The iif is not found: ignore the packet. */
1677 				return (0);
1678 			}
1679 
1680 			if (rt->mfc_flags[vifi] &
1681 			    MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1682 				/* WRONGVIF disabled: ignore the packet */
1683 				return (0);
1684 			}
1685 
1686 			microtime(&now);
1687 
1688 			TV_DELTA(rt->mfc_last_assert, now, delta);
1689 
1690 			if (delta > ASSERT_MSG_TIME) {
1691 				struct igmpmsg *im;
1692 				int hlen = ip->ip_hl << 2;
1693 				struct mbuf *mm =
1694 				    m_copym(m, 0, hlen, M_DONTWAIT);
1695 
1696 				M_PULLUP(mm, hlen);
1697 				if (mm == NULL)
1698 					return (ENOBUFS);
1699 
1700 				rt->mfc_last_assert = now;
1701 
1702 				im = mtod(mm, struct igmpmsg *);
1703 				im->im_msgtype	= IGMPMSG_WRONGVIF;
1704 				im->im_mbz	= 0;
1705 				im->im_vif	= vifi;
1706 
1707 				mrtstat.mrts_upcalls++;
1708 
1709 				sockaddr_in_init(&sin, &im->im_src, 0);
1710 				if (socket_send(ip_mrouter, mm, &sin) < 0) {
1711 					log(LOG_WARNING,
1712 					    "ip_mforward: ip_mrouter socket queue full\n");
1713 					++mrtstat.mrts_upq_sockfull;
1714 					return (ENOBUFS);
1715 				}
1716 			}
1717 		}
1718 		return (0);
1719 	}
1720 
1721 	/* If I sourced this packet, it counts as output, else it was input. */
1722 	if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1723 		viftable[vifi].v_pkt_out++;
1724 		viftable[vifi].v_bytes_out += plen;
1725 	} else {
1726 		viftable[vifi].v_pkt_in++;
1727 		viftable[vifi].v_bytes_in += plen;
1728 	}
1729 	rt->mfc_pkt_cnt++;
1730 	rt->mfc_byte_cnt += plen;
1731 
1732 	/*
1733 	 * For each vif, decide if a copy of the packet should be forwarded.
1734 	 * Forward if:
1735 	 *		- the ttl exceeds the vif's threshold
1736 	 *		- there are group members downstream on interface
1737 	 */
1738 	for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1739 		if ((rt->mfc_ttls[vifi] > 0) &&
1740 			(ip->ip_ttl > rt->mfc_ttls[vifi])) {
1741 			vifp->v_pkt_out++;
1742 			vifp->v_bytes_out += plen;
1743 #ifdef PIM
1744 			if (vifp->v_flags & VIFF_REGISTER)
1745 				pim_register_send(ip, vifp, m, rt);
1746 			else
1747 #endif
1748 			MC_SEND(ip, vifp, m);
1749 		}
1750 
1751 	/*
1752 	 * Perform upcall-related bw measuring.
1753 	 */
1754 	if (rt->mfc_bw_meter != NULL) {
1755 		struct bw_meter *x;
1756 		struct timeval now;
1757 
1758 		microtime(&now);
1759 		for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1760 			bw_meter_receive_packet(x, plen, &now);
1761 	}
1762 
1763 	return (0);
1764 }
1765 
1766 #ifdef RSVP_ISI
1767 /*
1768  * check if a vif number is legal/ok. This is used by ip_output.
1769  */
1770 int
1771 legal_vif_num(int vif)
1772 {
1773 	if (vif >= 0 && vif < numvifs)
1774 		return (1);
1775 	else
1776 		return (0);
1777 }
1778 #endif /* RSVP_ISI */
1779 
1780 static void
1781 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1782 {
1783 	struct mbuf *mb_copy;
1784 	int hlen = ip->ip_hl << 2;
1785 
1786 	/*
1787 	 * Make a new reference to the packet; make sure that
1788 	 * the IP header is actually copied, not just referenced,
1789 	 * so that ip_output() only scribbles on the copy.
1790 	 */
1791 	mb_copy = m_copypacket(m, M_DONTWAIT);
1792 	M_PULLUP(mb_copy, hlen);
1793 	if (mb_copy == NULL)
1794 		return;
1795 
1796 	if (vifp->v_rate_limit <= 0)
1797 		tbf_send_packet(vifp, mb_copy);
1798 	else
1799 		tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1800 		    ntohs(ip->ip_len));
1801 }
1802 
1803 static void
1804 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1805 {
1806 	struct mbuf *mb_copy;
1807 	struct ip *ip_copy;
1808 	int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1809 
1810 	/* Take care of delayed checksums */
1811 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1812 		in_delayed_cksum(m);
1813 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1814 	}
1815 
1816 	/*
1817 	 * copy the old packet & pullup its IP header into the
1818 	 * new mbuf so we can modify it.  Try to fill the new
1819 	 * mbuf since if we don't the ethernet driver will.
1820 	 */
1821 	MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1822 	if (mb_copy == NULL)
1823 		return;
1824 	mb_copy->m_data += max_linkhdr;
1825 	mb_copy->m_pkthdr.len = len;
1826 	mb_copy->m_len = sizeof(multicast_encap_iphdr);
1827 
1828 	if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
1829 		m_freem(mb_copy);
1830 		return;
1831 	}
1832 	i = MHLEN - max_linkhdr;
1833 	if (i > len)
1834 		i = len;
1835 	mb_copy = m_pullup(mb_copy, i);
1836 	if (mb_copy == NULL)
1837 		return;
1838 
1839 	/*
1840 	 * fill in the encapsulating IP header.
1841 	 */
1842 	ip_copy = mtod(mb_copy, struct ip *);
1843 	*ip_copy = multicast_encap_iphdr;
1844 	if (len < IP_MINFRAGSIZE)
1845 		ip_copy->ip_id = 0;
1846 	else
1847 		ip_copy->ip_id = ip_newid(NULL);
1848 	ip_copy->ip_len = htons(len);
1849 	ip_copy->ip_src = vifp->v_lcl_addr;
1850 	ip_copy->ip_dst = vifp->v_rmt_addr;
1851 
1852 	/*
1853 	 * turn the encapsulated IP header back into a valid one.
1854 	 */
1855 	ip = (struct ip *)((char *)ip_copy + sizeof(multicast_encap_iphdr));
1856 	--ip->ip_ttl;
1857 	ip->ip_sum = 0;
1858 	mb_copy->m_data += sizeof(multicast_encap_iphdr);
1859 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1860 	mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1861 
1862 	if (vifp->v_rate_limit <= 0)
1863 		tbf_send_packet(vifp, mb_copy);
1864 	else
1865 		tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1866 }
1867 
1868 /*
1869  * De-encapsulate a packet and feed it back through ip input.
1870  */
1871 static void
1872 vif_input(struct mbuf *m, ...)
1873 {
1874 	int off, proto;
1875 	va_list ap;
1876 	struct vif *vifp;
1877 
1878 	va_start(ap, m);
1879 	off = va_arg(ap, int);
1880 	proto = va_arg(ap, int);
1881 	va_end(ap);
1882 
1883 	vifp = (struct vif *)encap_getarg(m);
1884 	if (!vifp || proto != ENCAP_PROTO) {
1885 		m_freem(m);
1886 		mrtstat.mrts_bad_tunnel++;
1887 		return;
1888 	}
1889 
1890 	m_adj(m, off);
1891 	m->m_pkthdr.rcvif = vifp->v_ifp;
1892 
1893 	if (__predict_false(!pktq_enqueue(ip_pktq, m, 0))) {
1894 		m_freem(m);
1895 	}
1896 }
1897 
1898 /*
1899  * Check if the packet should be received on the vif denoted by arg.
1900  * (The encap selection code will call this once per vif since each is
1901  * registered separately.)
1902  */
1903 static int
1904 vif_encapcheck(struct mbuf *m, int off, int proto, void *arg)
1905 {
1906 	struct vif *vifp;
1907 	struct ip ip;
1908 
1909 #ifdef DIAGNOSTIC
1910 	if (!arg || proto != IPPROTO_IPV4)
1911 		panic("unexpected arg in vif_encapcheck");
1912 #endif
1913 
1914 	/*
1915 	 * Accept the packet only if the inner heaader is multicast
1916 	 * and the outer header matches a tunnel-mode vif.  Order
1917 	 * checks in the hope that common non-matching packets will be
1918 	 * rejected quickly.  Assume that unicast IPv4 traffic in a
1919 	 * parallel tunnel (e.g. gif(4)) is unlikely.
1920 	 */
1921 
1922 	/* Obtain the outer IP header and the vif pointer. */
1923 	m_copydata((struct mbuf *)m, 0, sizeof(ip), (void *)&ip);
1924 	vifp = (struct vif *)arg;
1925 
1926 	/*
1927 	 * The outer source must match the vif's remote peer address.
1928 	 * For a multicast router with several tunnels, this is the
1929 	 * only check that will fail on packets in other tunnels,
1930 	 * assuming the local address is the same.
1931 	 */
1932 	if (!in_hosteq(vifp->v_rmt_addr, ip.ip_src))
1933 		return 0;
1934 
1935 	/* The outer destination must match the vif's local address. */
1936 	if (!in_hosteq(vifp->v_lcl_addr, ip.ip_dst))
1937 		return 0;
1938 
1939 	/* The vif must be of tunnel type. */
1940 	if ((vifp->v_flags & VIFF_TUNNEL) == 0)
1941 		return 0;
1942 
1943 	/* Check that the inner destination is multicast. */
1944 	m_copydata((struct mbuf *)m, off, sizeof(ip), (void *)&ip);
1945 	if (!IN_MULTICAST(ip.ip_dst.s_addr))
1946 		return 0;
1947 
1948 	/*
1949 	 * We have checked that both the outer src and dst addresses
1950 	 * match the vif, and that the inner destination is multicast
1951 	 * (224/5).  By claiming more than 64, we intend to
1952 	 * preferentially take packets that also match a parallel
1953 	 * gif(4).
1954 	 */
1955 	return 32 + 32 + 5;
1956 }
1957 
1958 /*
1959  * Token bucket filter module
1960  */
1961 static void
1962 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
1963 {
1964 
1965 	if (len > MAX_BKT_SIZE) {
1966 		/* drop if packet is too large */
1967 		mrtstat.mrts_pkt2large++;
1968 		m_freem(m);
1969 		return;
1970 	}
1971 
1972 	tbf_update_tokens(vifp);
1973 
1974 	/*
1975 	 * If there are enough tokens, and the queue is empty, send this packet
1976 	 * out immediately.  Otherwise, try to insert it on this vif's queue.
1977 	 */
1978 	if (vifp->tbf_q_len == 0) {
1979 		if (len <= vifp->tbf_n_tok) {
1980 			vifp->tbf_n_tok -= len;
1981 			tbf_send_packet(vifp, m);
1982 		} else {
1983 			/* queue packet and timeout till later */
1984 			tbf_queue(vifp, m);
1985 			callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
1986 			    tbf_reprocess_q, vifp);
1987 		}
1988 	} else {
1989 		if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
1990 		    !tbf_dq_sel(vifp, ip)) {
1991 			/* queue full, and couldn't make room */
1992 			mrtstat.mrts_q_overflow++;
1993 			m_freem(m);
1994 		} else {
1995 			/* queue length low enough, or made room */
1996 			tbf_queue(vifp, m);
1997 			tbf_process_q(vifp);
1998 		}
1999 	}
2000 }
2001 
2002 /*
2003  * adds a packet to the queue at the interface
2004  */
2005 static void
2006 tbf_queue(struct vif *vifp, struct mbuf *m)
2007 {
2008 	int s = splsoftnet();
2009 
2010 	/* insert at tail */
2011 	*vifp->tbf_t = m;
2012 	vifp->tbf_t = &m->m_nextpkt;
2013 	vifp->tbf_q_len++;
2014 
2015 	splx(s);
2016 }
2017 
2018 
2019 /*
2020  * processes the queue at the interface
2021  */
2022 static void
2023 tbf_process_q(struct vif *vifp)
2024 {
2025 	struct mbuf *m;
2026 	int len;
2027 	int s = splsoftnet();
2028 
2029 	/*
2030 	 * Loop through the queue at the interface and send as many packets
2031 	 * as possible.
2032 	 */
2033 	for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
2034 		len = ntohs(mtod(m, struct ip *)->ip_len);
2035 
2036 		/* determine if the packet can be sent */
2037 		if (len <= vifp->tbf_n_tok) {
2038 			/* if so,
2039 			 * reduce no of tokens, dequeue the packet,
2040 			 * send the packet.
2041 			 */
2042 			if ((vifp->tbf_q = m->m_nextpkt) == NULL)
2043 				vifp->tbf_t = &vifp->tbf_q;
2044 			--vifp->tbf_q_len;
2045 
2046 			m->m_nextpkt = NULL;
2047 			vifp->tbf_n_tok -= len;
2048 			tbf_send_packet(vifp, m);
2049 		} else
2050 			break;
2051 	}
2052 	splx(s);
2053 }
2054 
2055 static void
2056 tbf_reprocess_q(void *arg)
2057 {
2058 	struct vif *vifp = arg;
2059 
2060 	if (ip_mrouter == NULL)
2061 		return;
2062 
2063 	tbf_update_tokens(vifp);
2064 	tbf_process_q(vifp);
2065 
2066 	if (vifp->tbf_q_len != 0)
2067 		callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2068 		    tbf_reprocess_q, vifp);
2069 }
2070 
2071 /* function that will selectively discard a member of the queue
2072  * based on the precedence value and the priority
2073  */
2074 static int
2075 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2076 {
2077 	u_int p;
2078 	struct mbuf **mp, *m;
2079 	int s = splsoftnet();
2080 
2081 	p = priority(vifp, ip);
2082 
2083 	for (mp = &vifp->tbf_q, m = *mp;
2084 	    m != NULL;
2085 	    mp = &m->m_nextpkt, m = *mp) {
2086 		if (p > priority(vifp, mtod(m, struct ip *))) {
2087 			if ((*mp = m->m_nextpkt) == NULL)
2088 				vifp->tbf_t = mp;
2089 			--vifp->tbf_q_len;
2090 
2091 			m_freem(m);
2092 			mrtstat.mrts_drop_sel++;
2093 			splx(s);
2094 			return (1);
2095 		}
2096 	}
2097 	splx(s);
2098 	return (0);
2099 }
2100 
2101 static void
2102 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2103 {
2104 	int error;
2105 	int s = splsoftnet();
2106 
2107 	if (vifp->v_flags & VIFF_TUNNEL) {
2108 		/* If tunnel options */
2109 		ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL);
2110 	} else {
2111 		/* if physical interface option, extract the options and then send */
2112 		struct ip_moptions imo;
2113 
2114 		imo.imo_multicast_ifp = vifp->v_ifp;
2115 		imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2116 		imo.imo_multicast_loop = 1;
2117 #ifdef RSVP_ISI
2118 		imo.imo_multicast_vif = -1;
2119 #endif
2120 
2121 		error = ip_output(m, NULL, NULL, IP_FORWARDING|IP_MULTICASTOPTS,
2122 		    &imo, NULL);
2123 
2124 		if (mrtdebug & DEBUG_XMIT)
2125 			log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2126 			    (long)(vifp - viftable), error);
2127 	}
2128 	splx(s);
2129 }
2130 
2131 /* determine the current time and then
2132  * the elapsed time (between the last time and time now)
2133  * in milliseconds & update the no. of tokens in the bucket
2134  */
2135 static void
2136 tbf_update_tokens(struct vif *vifp)
2137 {
2138 	struct timeval tp;
2139 	u_int32_t tm;
2140 	int s = splsoftnet();
2141 
2142 	microtime(&tp);
2143 
2144 	TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2145 
2146 	/*
2147 	 * This formula is actually
2148 	 * "time in seconds" * "bytes/second".
2149 	 *
2150 	 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2151 	 *
2152 	 * The (1000/1024) was introduced in add_vif to optimize
2153 	 * this divide into a shift.
2154 	 */
2155 	vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2156 	vifp->tbf_last_pkt_t = tp;
2157 
2158 	if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2159 		vifp->tbf_n_tok = MAX_BKT_SIZE;
2160 
2161 	splx(s);
2162 }
2163 
2164 static int
2165 priority(struct vif *vifp, struct ip *ip)
2166 {
2167 	int prio = 50;	/* the lowest priority -- default case */
2168 
2169 	/* temporary hack; may add general packet classifier some day */
2170 
2171 	/*
2172 	 * The UDP port space is divided up into four priority ranges:
2173 	 * [0, 16384)     : unclassified - lowest priority
2174 	 * [16384, 32768) : audio - highest priority
2175 	 * [32768, 49152) : whiteboard - medium priority
2176 	 * [49152, 65536) : video - low priority
2177 	 */
2178 	if (ip->ip_p == IPPROTO_UDP) {
2179 		struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2180 
2181 		switch (ntohs(udp->uh_dport) & 0xc000) {
2182 		case 0x4000:
2183 			prio = 70;
2184 			break;
2185 		case 0x8000:
2186 			prio = 60;
2187 			break;
2188 		case 0xc000:
2189 			prio = 55;
2190 			break;
2191 		}
2192 
2193 		if (tbfdebug > 1)
2194 			log(LOG_DEBUG, "port %x prio %d\n",
2195 			    ntohs(udp->uh_dport), prio);
2196 	}
2197 
2198 	return (prio);
2199 }
2200 
2201 /*
2202  * End of token bucket filter modifications
2203  */
2204 #ifdef RSVP_ISI
2205 int
2206 ip_rsvp_vif_init(struct socket *so, struct mbuf *m)
2207 {
2208 	int vifi, s;
2209 
2210 	if (rsvpdebug)
2211 		printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2212 		    so->so_type, so->so_proto->pr_protocol);
2213 
2214 	if (so->so_type != SOCK_RAW ||
2215 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2216 		return (EOPNOTSUPP);
2217 
2218 	/* Check mbuf. */
2219 	if (m == NULL || m->m_len != sizeof(int)) {
2220 		return (EINVAL);
2221 	}
2222 	vifi = *(mtod(m, int *));
2223 
2224 	if (rsvpdebug)
2225 		printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",
2226 		       vifi, rsvp_on);
2227 
2228 	s = splsoftnet();
2229 
2230 	/* Check vif. */
2231 	if (!legal_vif_num(vifi)) {
2232 		splx(s);
2233 		return (EADDRNOTAVAIL);
2234 	}
2235 
2236 	/* Check if socket is available. */
2237 	if (viftable[vifi].v_rsvpd != NULL) {
2238 		splx(s);
2239 		return (EADDRINUSE);
2240 	}
2241 
2242 	viftable[vifi].v_rsvpd = so;
2243 	/*
2244 	 * This may seem silly, but we need to be sure we don't over-increment
2245 	 * the RSVP counter, in case something slips up.
2246 	 */
2247 	if (!viftable[vifi].v_rsvp_on) {
2248 		viftable[vifi].v_rsvp_on = 1;
2249 		rsvp_on++;
2250 	}
2251 
2252 	splx(s);
2253 	return (0);
2254 }
2255 
2256 int
2257 ip_rsvp_vif_done(struct socket *so, struct mbuf *m)
2258 {
2259 	int vifi, s;
2260 
2261 	if (rsvpdebug)
2262 		printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2263 		    so->so_type, so->so_proto->pr_protocol);
2264 
2265 	if (so->so_type != SOCK_RAW ||
2266 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2267 		return (EOPNOTSUPP);
2268 
2269 	/* Check mbuf. */
2270 	if (m == NULL || m->m_len != sizeof(int)) {
2271 		return (EINVAL);
2272 	}
2273 	vifi = *(mtod(m, int *));
2274 
2275 	s = splsoftnet();
2276 
2277 	/* Check vif. */
2278 	if (!legal_vif_num(vifi)) {
2279 		splx(s);
2280 		return (EADDRNOTAVAIL);
2281 	}
2282 
2283 	if (rsvpdebug)
2284 		printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2285 		    viftable[vifi].v_rsvpd, so);
2286 
2287 	viftable[vifi].v_rsvpd = NULL;
2288 	/*
2289 	 * This may seem silly, but we need to be sure we don't over-decrement
2290 	 * the RSVP counter, in case something slips up.
2291 	 */
2292 	if (viftable[vifi].v_rsvp_on) {
2293 		viftable[vifi].v_rsvp_on = 0;
2294 		rsvp_on--;
2295 	}
2296 
2297 	splx(s);
2298 	return (0);
2299 }
2300 
2301 void
2302 ip_rsvp_force_done(struct socket *so)
2303 {
2304 	int vifi, s;
2305 
2306 	/* Don't bother if it is not the right type of socket. */
2307 	if (so->so_type != SOCK_RAW ||
2308 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2309 		return;
2310 
2311 	s = splsoftnet();
2312 
2313 	/*
2314 	 * The socket may be attached to more than one vif...this
2315 	 * is perfectly legal.
2316 	 */
2317 	for (vifi = 0; vifi < numvifs; vifi++) {
2318 		if (viftable[vifi].v_rsvpd == so) {
2319 			viftable[vifi].v_rsvpd = NULL;
2320 			/*
2321 			 * This may seem silly, but we need to be sure we don't
2322 			 * over-decrement the RSVP counter, in case something
2323 			 * slips up.
2324 			 */
2325 			if (viftable[vifi].v_rsvp_on) {
2326 				viftable[vifi].v_rsvp_on = 0;
2327 				rsvp_on--;
2328 			}
2329 		}
2330 	}
2331 
2332 	splx(s);
2333 	return;
2334 }
2335 
2336 void
2337 rsvp_input(struct mbuf *m, struct ifnet *ifp)
2338 {
2339 	int vifi, s;
2340 	struct ip *ip = mtod(m, struct ip *);
2341 	struct sockaddr_in rsvp_src;
2342 
2343 	if (rsvpdebug)
2344 		printf("rsvp_input: rsvp_on %d\n", rsvp_on);
2345 
2346 	/*
2347 	 * Can still get packets with rsvp_on = 0 if there is a local member
2348 	 * of the group to which the RSVP packet is addressed.  But in this
2349 	 * case we want to throw the packet away.
2350 	 */
2351 	if (!rsvp_on) {
2352 		m_freem(m);
2353 		return;
2354 	}
2355 
2356 	/*
2357 	 * If the old-style non-vif-associated socket is set, then use
2358 	 * it and ignore the new ones.
2359 	 */
2360 	if (ip_rsvpd != NULL) {
2361 		if (rsvpdebug)
2362 			printf("rsvp_input: "
2363 			    "Sending packet up old-style socket\n");
2364 		rip_input(m);	/*XXX*/
2365 		return;
2366 	}
2367 
2368 	s = splsoftnet();
2369 
2370 	if (rsvpdebug)
2371 		printf("rsvp_input: check vifs\n");
2372 
2373 	/* Find which vif the packet arrived on. */
2374 	for (vifi = 0; vifi < numvifs; vifi++) {
2375 		if (viftable[vifi].v_ifp == ifp)
2376 			break;
2377 	}
2378 
2379 	if (vifi == numvifs) {
2380 		/* Can't find vif packet arrived on. Drop packet. */
2381 		if (rsvpdebug)
2382 			printf("rsvp_input: "
2383 			    "Can't find vif for packet...dropping it.\n");
2384 		m_freem(m);
2385 		splx(s);
2386 		return;
2387 	}
2388 
2389 	if (rsvpdebug)
2390 		printf("rsvp_input: check socket\n");
2391 
2392 	if (viftable[vifi].v_rsvpd == NULL) {
2393 		/*
2394 		 * drop packet, since there is no specific socket for this
2395 		 * interface
2396 		 */
2397 		if (rsvpdebug)
2398 			printf("rsvp_input: No socket defined for vif %d\n",
2399 			    vifi);
2400 		m_freem(m);
2401 		splx(s);
2402 		return;
2403 	}
2404 
2405 	sockaddr_in_init(&rsvp_src, &ip->ip_src, 0);
2406 
2407 	if (rsvpdebug && m)
2408 		printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
2409 		    m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv));
2410 
2411 	if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2412 		if (rsvpdebug)
2413 			printf("rsvp_input: Failed to append to socket\n");
2414 	else
2415 		if (rsvpdebug)
2416 			printf("rsvp_input: send packet up\n");
2417 
2418 	splx(s);
2419 }
2420 #endif /* RSVP_ISI */
2421 
2422 /*
2423  * Code for bandwidth monitors
2424  */
2425 
2426 /*
2427  * Define common interface for timeval-related methods
2428  */
2429 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2430 #define	BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2431 #define	BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2432 
2433 static uint32_t
2434 compute_bw_meter_flags(struct bw_upcall *req)
2435 {
2436     uint32_t flags = 0;
2437 
2438     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2439 	flags |= BW_METER_UNIT_PACKETS;
2440     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2441 	flags |= BW_METER_UNIT_BYTES;
2442     if (req->bu_flags & BW_UPCALL_GEQ)
2443 	flags |= BW_METER_GEQ;
2444     if (req->bu_flags & BW_UPCALL_LEQ)
2445 	flags |= BW_METER_LEQ;
2446 
2447     return flags;
2448 }
2449 
2450 /*
2451  * Add a bw_meter entry
2452  */
2453 static int
2454 add_bw_upcall(struct bw_upcall *req)
2455 {
2456     int s;
2457     struct mfc *mfc;
2458     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2459 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2460     struct timeval now;
2461     struct bw_meter *x;
2462     uint32_t flags;
2463 
2464     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2465 	return EOPNOTSUPP;
2466 
2467     /* Test if the flags are valid */
2468     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2469 	return EINVAL;
2470     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2471 	return EINVAL;
2472     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2473 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2474 	return EINVAL;
2475 
2476     /* Test if the threshold time interval is valid */
2477     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2478 	return EINVAL;
2479 
2480     flags = compute_bw_meter_flags(req);
2481 
2482     /*
2483      * Find if we have already same bw_meter entry
2484      */
2485     s = splsoftnet();
2486     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2487     if (mfc == NULL) {
2488 	splx(s);
2489 	return EADDRNOTAVAIL;
2490     }
2491     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2492 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2493 			   &req->bu_threshold.b_time, ==)) &&
2494 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2495 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2496 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
2497 	    splx(s);
2498 	    return 0;		/* XXX Already installed */
2499 	}
2500     }
2501 
2502     /* Allocate the new bw_meter entry */
2503     x = kmem_intr_alloc(sizeof(*x), KM_NOSLEEP);
2504     if (x == NULL) {
2505 	splx(s);
2506 	return ENOBUFS;
2507     }
2508 
2509     /* Set the new bw_meter entry */
2510     x->bm_threshold.b_time = req->bu_threshold.b_time;
2511     microtime(&now);
2512     x->bm_start_time = now;
2513     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2514     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2515     x->bm_measured.b_packets = 0;
2516     x->bm_measured.b_bytes = 0;
2517     x->bm_flags = flags;
2518     x->bm_time_next = NULL;
2519     x->bm_time_hash = BW_METER_BUCKETS;
2520 
2521     /* Add the new bw_meter entry to the front of entries for this MFC */
2522     x->bm_mfc = mfc;
2523     x->bm_mfc_next = mfc->mfc_bw_meter;
2524     mfc->mfc_bw_meter = x;
2525     schedule_bw_meter(x, &now);
2526     splx(s);
2527 
2528     return 0;
2529 }
2530 
2531 static void
2532 free_bw_list(struct bw_meter *list)
2533 {
2534     while (list != NULL) {
2535 	struct bw_meter *x = list;
2536 
2537 	list = list->bm_mfc_next;
2538 	unschedule_bw_meter(x);
2539 	kmem_free(x, sizeof(*x));
2540     }
2541 }
2542 
2543 /*
2544  * Delete one or multiple bw_meter entries
2545  */
2546 static int
2547 del_bw_upcall(struct bw_upcall *req)
2548 {
2549     int s;
2550     struct mfc *mfc;
2551     struct bw_meter *x;
2552 
2553     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2554 	return EOPNOTSUPP;
2555 
2556     s = splsoftnet();
2557     /* Find the corresponding MFC entry */
2558     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2559     if (mfc == NULL) {
2560 	splx(s);
2561 	return EADDRNOTAVAIL;
2562     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2563 	/*
2564 	 * Delete all bw_meter entries for this mfc
2565 	 */
2566 	struct bw_meter *list;
2567 
2568 	list = mfc->mfc_bw_meter;
2569 	mfc->mfc_bw_meter = NULL;
2570 	free_bw_list(list);
2571 	splx(s);
2572 	return 0;
2573     } else {			/* Delete a single bw_meter entry */
2574 	struct bw_meter *prev;
2575 	uint32_t flags = 0;
2576 
2577 	flags = compute_bw_meter_flags(req);
2578 
2579 	/* Find the bw_meter entry to delete */
2580 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2581 	     prev = x, x = x->bm_mfc_next) {
2582 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2583 			       &req->bu_threshold.b_time, ==)) &&
2584 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2585 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2586 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2587 		break;
2588 	}
2589 	if (x != NULL) { /* Delete entry from the list for this MFC */
2590 	    if (prev != NULL)
2591 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2592 	    else
2593 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2594 
2595 	    unschedule_bw_meter(x);
2596 	    splx(s);
2597 	    /* Free the bw_meter entry */
2598 	    kmem_free(x, sizeof(*x));
2599 	    return 0;
2600 	} else {
2601 	    splx(s);
2602 	    return EINVAL;
2603 	}
2604     }
2605     /* NOTREACHED */
2606 }
2607 
2608 /*
2609  * Perform bandwidth measurement processing that may result in an upcall
2610  */
2611 static void
2612 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2613 {
2614     struct timeval delta;
2615 
2616     delta = *nowp;
2617     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2618 
2619     if (x->bm_flags & BW_METER_GEQ) {
2620 	/*
2621 	 * Processing for ">=" type of bw_meter entry
2622 	 */
2623 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2624 	    /* Reset the bw_meter entry */
2625 	    x->bm_start_time = *nowp;
2626 	    x->bm_measured.b_packets = 0;
2627 	    x->bm_measured.b_bytes = 0;
2628 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2629 	}
2630 
2631 	/* Record that a packet is received */
2632 	x->bm_measured.b_packets++;
2633 	x->bm_measured.b_bytes += plen;
2634 
2635 	/*
2636 	 * Test if we should deliver an upcall
2637 	 */
2638 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2639 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2640 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2641 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2642 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2643 		/* Prepare an upcall for delivery */
2644 		bw_meter_prepare_upcall(x, nowp);
2645 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2646 	    }
2647 	}
2648     } else if (x->bm_flags & BW_METER_LEQ) {
2649 	/*
2650 	 * Processing for "<=" type of bw_meter entry
2651 	 */
2652 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2653 	    /*
2654 	     * We are behind time with the multicast forwarding table
2655 	     * scanning for "<=" type of bw_meter entries, so test now
2656 	     * if we should deliver an upcall.
2657 	     */
2658 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2659 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2660 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2661 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2662 		/* Prepare an upcall for delivery */
2663 		bw_meter_prepare_upcall(x, nowp);
2664 	    }
2665 	    /* Reschedule the bw_meter entry */
2666 	    unschedule_bw_meter(x);
2667 	    schedule_bw_meter(x, nowp);
2668 	}
2669 
2670 	/* Record that a packet is received */
2671 	x->bm_measured.b_packets++;
2672 	x->bm_measured.b_bytes += plen;
2673 
2674 	/*
2675 	 * Test if we should restart the measuring interval
2676 	 */
2677 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2678 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2679 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
2680 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2681 	    /* Don't restart the measuring interval */
2682 	} else {
2683 	    /* Do restart the measuring interval */
2684 	    /*
2685 	     * XXX: note that we don't unschedule and schedule, because this
2686 	     * might be too much overhead per packet. Instead, when we process
2687 	     * all entries for a given timer hash bin, we check whether it is
2688 	     * really a timeout. If not, we reschedule at that time.
2689 	     */
2690 	    x->bm_start_time = *nowp;
2691 	    x->bm_measured.b_packets = 0;
2692 	    x->bm_measured.b_bytes = 0;
2693 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2694 	}
2695     }
2696 }
2697 
2698 /*
2699  * Prepare a bandwidth-related upcall
2700  */
2701 static void
2702 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2703 {
2704     struct timeval delta;
2705     struct bw_upcall *u;
2706 
2707     /*
2708      * Compute the measured time interval
2709      */
2710     delta = *nowp;
2711     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2712 
2713     /*
2714      * If there are too many pending upcalls, deliver them now
2715      */
2716     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2717 	bw_upcalls_send();
2718 
2719     /*
2720      * Set the bw_upcall entry
2721      */
2722     u = &bw_upcalls[bw_upcalls_n++];
2723     u->bu_src = x->bm_mfc->mfc_origin;
2724     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2725     u->bu_threshold.b_time = x->bm_threshold.b_time;
2726     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2727     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2728     u->bu_measured.b_time = delta;
2729     u->bu_measured.b_packets = x->bm_measured.b_packets;
2730     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2731     u->bu_flags = 0;
2732     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2733 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2734     if (x->bm_flags & BW_METER_UNIT_BYTES)
2735 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2736     if (x->bm_flags & BW_METER_GEQ)
2737 	u->bu_flags |= BW_UPCALL_GEQ;
2738     if (x->bm_flags & BW_METER_LEQ)
2739 	u->bu_flags |= BW_UPCALL_LEQ;
2740 }
2741 
2742 /*
2743  * Send the pending bandwidth-related upcalls
2744  */
2745 static void
2746 bw_upcalls_send(void)
2747 {
2748     struct mbuf *m;
2749     int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2750     struct sockaddr_in k_igmpsrc = {
2751 	    .sin_len = sizeof(k_igmpsrc),
2752 	    .sin_family = AF_INET,
2753     };
2754     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2755 				      0,		/* unused2 */
2756 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2757 				      0,		/* im_mbz  */
2758 				      0,		/* im_vif  */
2759 				      0,		/* unused3 */
2760 				      { 0 },		/* im_src  */
2761 				      { 0 } };		/* im_dst  */
2762 
2763     if (bw_upcalls_n == 0)
2764 	return;			/* No pending upcalls */
2765 
2766     bw_upcalls_n = 0;
2767 
2768     /*
2769      * Allocate a new mbuf, initialize it with the header and
2770      * the payload for the pending calls.
2771      */
2772     MGETHDR(m, M_DONTWAIT, MT_HEADER);
2773     if (m == NULL) {
2774 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2775 	return;
2776     }
2777 
2778     m->m_len = m->m_pkthdr.len = 0;
2779     m_copyback(m, 0, sizeof(struct igmpmsg), (void *)&igmpmsg);
2780     m_copyback(m, sizeof(struct igmpmsg), len, (void *)&bw_upcalls[0]);
2781 
2782     /*
2783      * Send the upcalls
2784      * XXX do we need to set the address in k_igmpsrc ?
2785      */
2786     mrtstat.mrts_upcalls++;
2787     if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2788 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2789 	++mrtstat.mrts_upq_sockfull;
2790     }
2791 }
2792 
2793 /*
2794  * Compute the timeout hash value for the bw_meter entries
2795  */
2796 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2797     do {								\
2798 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2799 									\
2800 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2801 	(hash) = next_timeval.tv_sec;					\
2802 	if (next_timeval.tv_usec)					\
2803 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2804 	(hash) %= BW_METER_BUCKETS;					\
2805     } while (/*CONSTCOND*/ 0)
2806 
2807 /*
2808  * Schedule a timer to process periodically bw_meter entry of type "<="
2809  * by linking the entry in the proper hash bucket.
2810  */
2811 static void
2812 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2813 {
2814     int time_hash;
2815 
2816     if (!(x->bm_flags & BW_METER_LEQ))
2817 	return;		/* XXX: we schedule timers only for "<=" entries */
2818 
2819     /*
2820      * Reset the bw_meter entry
2821      */
2822     x->bm_start_time = *nowp;
2823     x->bm_measured.b_packets = 0;
2824     x->bm_measured.b_bytes = 0;
2825     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2826 
2827     /*
2828      * Compute the timeout hash value and insert the entry
2829      */
2830     BW_METER_TIMEHASH(x, time_hash);
2831     x->bm_time_next = bw_meter_timers[time_hash];
2832     bw_meter_timers[time_hash] = x;
2833     x->bm_time_hash = time_hash;
2834 }
2835 
2836 /*
2837  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2838  * by removing the entry from the proper hash bucket.
2839  */
2840 static void
2841 unschedule_bw_meter(struct bw_meter *x)
2842 {
2843     int time_hash;
2844     struct bw_meter *prev, *tmp;
2845 
2846     if (!(x->bm_flags & BW_METER_LEQ))
2847 	return;		/* XXX: we schedule timers only for "<=" entries */
2848 
2849     /*
2850      * Compute the timeout hash value and delete the entry
2851      */
2852     time_hash = x->bm_time_hash;
2853     if (time_hash >= BW_METER_BUCKETS)
2854 	return;		/* Entry was not scheduled */
2855 
2856     for (prev = NULL, tmp = bw_meter_timers[time_hash];
2857 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2858 	if (tmp == x)
2859 	    break;
2860 
2861     if (tmp == NULL)
2862 	panic("unschedule_bw_meter: bw_meter entry not found");
2863 
2864     if (prev != NULL)
2865 	prev->bm_time_next = x->bm_time_next;
2866     else
2867 	bw_meter_timers[time_hash] = x->bm_time_next;
2868 
2869     x->bm_time_next = NULL;
2870     x->bm_time_hash = BW_METER_BUCKETS;
2871 }
2872 
2873 /*
2874  * Process all "<=" type of bw_meter that should be processed now,
2875  * and for each entry prepare an upcall if necessary. Each processed
2876  * entry is rescheduled again for the (periodic) processing.
2877  *
2878  * This is run periodically (once per second normally). On each round,
2879  * all the potentially matching entries are in the hash slot that we are
2880  * looking at.
2881  */
2882 static void
2883 bw_meter_process(void)
2884 {
2885     int s;
2886     static uint32_t last_tv_sec;	/* last time we processed this */
2887 
2888     uint32_t loops;
2889     int i;
2890     struct timeval now, process_endtime;
2891 
2892     microtime(&now);
2893     if (last_tv_sec == now.tv_sec)
2894 	return;		/* nothing to do */
2895 
2896     loops = now.tv_sec - last_tv_sec;
2897     last_tv_sec = now.tv_sec;
2898     if (loops > BW_METER_BUCKETS)
2899 	loops = BW_METER_BUCKETS;
2900 
2901     s = splsoftnet();
2902     /*
2903      * Process all bins of bw_meter entries from the one after the last
2904      * processed to the current one. On entry, i points to the last bucket
2905      * visited, so we need to increment i at the beginning of the loop.
2906      */
2907     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2908 	struct bw_meter *x, *tmp_list;
2909 
2910 	if (++i >= BW_METER_BUCKETS)
2911 	    i = 0;
2912 
2913 	/* Disconnect the list of bw_meter entries from the bin */
2914 	tmp_list = bw_meter_timers[i];
2915 	bw_meter_timers[i] = NULL;
2916 
2917 	/* Process the list of bw_meter entries */
2918 	while (tmp_list != NULL) {
2919 	    x = tmp_list;
2920 	    tmp_list = tmp_list->bm_time_next;
2921 
2922 	    /* Test if the time interval is over */
2923 	    process_endtime = x->bm_start_time;
2924 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2925 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2926 		/* Not yet: reschedule, but don't reset */
2927 		int time_hash;
2928 
2929 		BW_METER_TIMEHASH(x, time_hash);
2930 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2931 		    /*
2932 		     * XXX: somehow the bin processing is a bit ahead of time.
2933 		     * Put the entry in the next bin.
2934 		     */
2935 		    if (++time_hash >= BW_METER_BUCKETS)
2936 			time_hash = 0;
2937 		}
2938 		x->bm_time_next = bw_meter_timers[time_hash];
2939 		bw_meter_timers[time_hash] = x;
2940 		x->bm_time_hash = time_hash;
2941 
2942 		continue;
2943 	    }
2944 
2945 	    /*
2946 	     * Test if we should deliver an upcall
2947 	     */
2948 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2949 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2950 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2951 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2952 		/* Prepare an upcall for delivery */
2953 		bw_meter_prepare_upcall(x, &now);
2954 	    }
2955 
2956 	    /*
2957 	     * Reschedule for next processing
2958 	     */
2959 	    schedule_bw_meter(x, &now);
2960 	}
2961     }
2962 
2963     /* Send all upcalls that are pending delivery */
2964     bw_upcalls_send();
2965 
2966     splx(s);
2967 }
2968 
2969 /*
2970  * A periodic function for sending all upcalls that are pending delivery
2971  */
2972 static void
2973 expire_bw_upcalls_send(void *unused)
2974 {
2975     int s;
2976 
2977     s = splsoftnet();
2978     bw_upcalls_send();
2979     splx(s);
2980 
2981     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2982 		  expire_bw_upcalls_send, NULL);
2983 }
2984 
2985 /*
2986  * A periodic function for periodic scanning of the multicast forwarding
2987  * table for processing all "<=" bw_meter entries.
2988  */
2989 static void
2990 expire_bw_meter_process(void *unused)
2991 {
2992     if (mrt_api_config & MRT_MFC_BW_UPCALL)
2993 	bw_meter_process();
2994 
2995     callout_reset(&bw_meter_ch, BW_METER_PERIOD,
2996 		  expire_bw_meter_process, NULL);
2997 }
2998 
2999 /*
3000  * End of bandwidth monitoring code
3001  */
3002 
3003 #ifdef PIM
3004 /*
3005  * Send the packet up to the user daemon, or eventually do kernel encapsulation
3006  */
3007 static int
3008 pim_register_send(struct ip *ip, struct vif *vifp,
3009 	struct mbuf *m, struct mfc *rt)
3010 {
3011     struct mbuf *mb_copy, *mm;
3012 
3013     if (mrtdebug & DEBUG_PIM)
3014         log(LOG_DEBUG, "pim_register_send: \n");
3015 
3016     mb_copy = pim_register_prepare(ip, m);
3017     if (mb_copy == NULL)
3018 	return ENOBUFS;
3019 
3020     /*
3021      * Send all the fragments. Note that the mbuf for each fragment
3022      * is freed by the sending machinery.
3023      */
3024     for (mm = mb_copy; mm; mm = mb_copy) {
3025 	mb_copy = mm->m_nextpkt;
3026 	mm->m_nextpkt = NULL;
3027 	mm = m_pullup(mm, sizeof(struct ip));
3028 	if (mm != NULL) {
3029 	    ip = mtod(mm, struct ip *);
3030 	    if ((mrt_api_config & MRT_MFC_RP) &&
3031 		!in_nullhost(rt->mfc_rp)) {
3032 		pim_register_send_rp(ip, vifp, mm, rt);
3033 	    } else {
3034 		pim_register_send_upcall(ip, vifp, mm, rt);
3035 	    }
3036 	}
3037     }
3038 
3039     return 0;
3040 }
3041 
3042 /*
3043  * Return a copy of the data packet that is ready for PIM Register
3044  * encapsulation.
3045  * XXX: Note that in the returned copy the IP header is a valid one.
3046  */
3047 static struct mbuf *
3048 pim_register_prepare(struct ip *ip, struct mbuf *m)
3049 {
3050     struct mbuf *mb_copy = NULL;
3051     int mtu;
3052 
3053     /* Take care of delayed checksums */
3054     if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3055 	in_delayed_cksum(m);
3056 	m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3057     }
3058 
3059     /*
3060      * Copy the old packet & pullup its IP header into the
3061      * new mbuf so we can modify it.
3062      */
3063     mb_copy = m_copypacket(m, M_DONTWAIT);
3064     if (mb_copy == NULL)
3065 	return NULL;
3066     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3067     if (mb_copy == NULL)
3068 	return NULL;
3069 
3070     /* take care of the TTL */
3071     ip = mtod(mb_copy, struct ip *);
3072     --ip->ip_ttl;
3073 
3074     /* Compute the MTU after the PIM Register encapsulation */
3075     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3076 
3077     if (ntohs(ip->ip_len) <= mtu) {
3078 	/* Turn the IP header into a valid one */
3079 	ip->ip_sum = 0;
3080 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3081     } else {
3082 	/* Fragment the packet */
3083 	if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3084 	    /* XXX: mb_copy was freed by ip_fragment() */
3085 	    return NULL;
3086 	}
3087     }
3088     return mb_copy;
3089 }
3090 
3091 /*
3092  * Send an upcall with the data packet to the user-level process.
3093  */
3094 static int
3095 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3096     struct mbuf *mb_copy, struct mfc *rt)
3097 {
3098     struct mbuf *mb_first;
3099     int len = ntohs(ip->ip_len);
3100     struct igmpmsg *im;
3101     struct sockaddr_in k_igmpsrc = {
3102 	    .sin_len = sizeof(k_igmpsrc),
3103 	    .sin_family = AF_INET,
3104     };
3105 
3106     /*
3107      * Add a new mbuf with an upcall header
3108      */
3109     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3110     if (mb_first == NULL) {
3111 	m_freem(mb_copy);
3112 	return ENOBUFS;
3113     }
3114     mb_first->m_data += max_linkhdr;
3115     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3116     mb_first->m_len = sizeof(struct igmpmsg);
3117     mb_first->m_next = mb_copy;
3118 
3119     /* Send message to routing daemon */
3120     im = mtod(mb_first, struct igmpmsg *);
3121     im->im_msgtype	= IGMPMSG_WHOLEPKT;
3122     im->im_mbz		= 0;
3123     im->im_vif		= vifp - viftable;
3124     im->im_src		= ip->ip_src;
3125     im->im_dst		= ip->ip_dst;
3126 
3127     k_igmpsrc.sin_addr	= ip->ip_src;
3128 
3129     mrtstat.mrts_upcalls++;
3130 
3131     if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3132 	if (mrtdebug & DEBUG_PIM)
3133 	    log(LOG_WARNING,
3134 		"mcast: pim_register_send_upcall: ip_mrouter socket queue full\n");
3135 	++mrtstat.mrts_upq_sockfull;
3136 	return ENOBUFS;
3137     }
3138 
3139     /* Keep statistics */
3140     pimstat.pims_snd_registers_msgs++;
3141     pimstat.pims_snd_registers_bytes += len;
3142 
3143     return 0;
3144 }
3145 
3146 /*
3147  * Encapsulate the data packet in PIM Register message and send it to the RP.
3148  */
3149 static int
3150 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3151 	struct mbuf *mb_copy, struct mfc *rt)
3152 {
3153     struct mbuf *mb_first;
3154     struct ip *ip_outer;
3155     struct pim_encap_pimhdr *pimhdr;
3156     int len = ntohs(ip->ip_len);
3157     vifi_t vifi = rt->mfc_parent;
3158 
3159     if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3160 	m_freem(mb_copy);
3161 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
3162     }
3163 
3164     /*
3165      * Add a new mbuf with the encapsulating header
3166      */
3167     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3168     if (mb_first == NULL) {
3169 	m_freem(mb_copy);
3170 	return ENOBUFS;
3171     }
3172     mb_first->m_data += max_linkhdr;
3173     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3174     mb_first->m_next = mb_copy;
3175 
3176     mb_first->m_pkthdr.len = len + mb_first->m_len;
3177 
3178     /*
3179      * Fill in the encapsulating IP and PIM header
3180      */
3181     ip_outer = mtod(mb_first, struct ip *);
3182     *ip_outer = pim_encap_iphdr;
3183      if (mb_first->m_pkthdr.len < IP_MINFRAGSIZE)
3184 	ip_outer->ip_id = 0;
3185     else
3186 	ip_outer->ip_id = ip_newid(NULL);
3187     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3188 			     sizeof(pim_encap_pimhdr));
3189     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3190     ip_outer->ip_dst = rt->mfc_rp;
3191     /*
3192      * Copy the inner header TOS to the outer header, and take care of the
3193      * IP_DF bit.
3194      */
3195     ip_outer->ip_tos = ip->ip_tos;
3196     if (ntohs(ip->ip_off) & IP_DF)
3197 	ip_outer->ip_off |= htons(IP_DF);
3198     pimhdr = (struct pim_encap_pimhdr *)((char *)ip_outer
3199 					 + sizeof(pim_encap_iphdr));
3200     *pimhdr = pim_encap_pimhdr;
3201     /* If the iif crosses a border, set the Border-bit */
3202     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3203 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3204 
3205     mb_first->m_data += sizeof(pim_encap_iphdr);
3206     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3207     mb_first->m_data -= sizeof(pim_encap_iphdr);
3208 
3209     if (vifp->v_rate_limit == 0)
3210 	tbf_send_packet(vifp, mb_first);
3211     else
3212 	tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3213 
3214     /* Keep statistics */
3215     pimstat.pims_snd_registers_msgs++;
3216     pimstat.pims_snd_registers_bytes += len;
3217 
3218     return 0;
3219 }
3220 
3221 /*
3222  * PIM-SMv2 and PIM-DM messages processing.
3223  * Receives and verifies the PIM control messages, and passes them
3224  * up to the listening socket, using rip_input().
3225  * The only message with special processing is the PIM_REGISTER message
3226  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3227  * is passed to if_simloop().
3228  */
3229 void
3230 pim_input(struct mbuf *m, ...)
3231 {
3232     struct ip *ip = mtod(m, struct ip *);
3233     struct pim *pim;
3234     int minlen;
3235     int datalen;
3236     int ip_tos;
3237     int proto;
3238     int iphlen;
3239     va_list ap;
3240 
3241     va_start(ap, m);
3242     iphlen = va_arg(ap, int);
3243     proto = va_arg(ap, int);
3244     va_end(ap);
3245 
3246     datalen = ntohs(ip->ip_len) - iphlen;
3247 
3248     /* Keep statistics */
3249     pimstat.pims_rcv_total_msgs++;
3250     pimstat.pims_rcv_total_bytes += datalen;
3251 
3252     /*
3253      * Validate lengths
3254      */
3255     if (datalen < PIM_MINLEN) {
3256 	pimstat.pims_rcv_tooshort++;
3257 	log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3258 	    datalen, (u_long)ip->ip_src.s_addr);
3259 	m_freem(m);
3260 	return;
3261     }
3262 
3263     /*
3264      * If the packet is at least as big as a REGISTER, go agead
3265      * and grab the PIM REGISTER header size, to avoid another
3266      * possible m_pullup() later.
3267      *
3268      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
3269      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3270      */
3271     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3272     /*
3273      * Get the IP and PIM headers in contiguous memory, and
3274      * possibly the PIM REGISTER header.
3275      */
3276     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3277 	(m = m_pullup(m, minlen)) == NULL) {
3278 	log(LOG_ERR, "pim_input: m_pullup failure\n");
3279 	return;
3280     }
3281     /* m_pullup() may have given us a new mbuf so reset ip. */
3282     ip = mtod(m, struct ip *);
3283     ip_tos = ip->ip_tos;
3284 
3285     /* adjust mbuf to point to the PIM header */
3286     m->m_data += iphlen;
3287     m->m_len  -= iphlen;
3288     pim = mtod(m, struct pim *);
3289 
3290     /*
3291      * Validate checksum. If PIM REGISTER, exclude the data packet.
3292      *
3293      * XXX: some older PIMv2 implementations don't make this distinction,
3294      * so for compatibility reason perform the checksum over part of the
3295      * message, and if error, then over the whole message.
3296      */
3297     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3298 	/* do nothing, checksum okay */
3299     } else if (in_cksum(m, datalen)) {
3300 	pimstat.pims_rcv_badsum++;
3301 	if (mrtdebug & DEBUG_PIM)
3302 	    log(LOG_DEBUG, "pim_input: invalid checksum\n");
3303 	m_freem(m);
3304 	return;
3305     }
3306 
3307     /* PIM version check */
3308     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3309 	pimstat.pims_rcv_badversion++;
3310 	log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3311 	    PIM_VT_V(pim->pim_vt), PIM_VERSION);
3312 	m_freem(m);
3313 	return;
3314     }
3315 
3316     /* restore mbuf back to the outer IP */
3317     m->m_data -= iphlen;
3318     m->m_len  += iphlen;
3319 
3320     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3321 	/*
3322 	 * Since this is a REGISTER, we'll make a copy of the register
3323 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3324 	 * routing daemon.
3325 	 */
3326 	int s;
3327 	struct sockaddr_in dst = {
3328 		.sin_len = sizeof(dst),
3329 		.sin_family = AF_INET,
3330 	};
3331 	struct mbuf *mcp;
3332 	struct ip *encap_ip;
3333 	u_int32_t *reghdr;
3334 	struct ifnet *vifp;
3335 
3336 	s = splsoftnet();
3337 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3338 	    splx(s);
3339 	    if (mrtdebug & DEBUG_PIM)
3340 		log(LOG_DEBUG,
3341 		    "pim_input: register vif not set: %d\n", reg_vif_num);
3342 	    m_freem(m);
3343 	    return;
3344 	}
3345 	/* XXX need refcnt? */
3346 	vifp = viftable[reg_vif_num].v_ifp;
3347 	splx(s);
3348 
3349 	/*
3350 	 * Validate length
3351 	 */
3352 	if (datalen < PIM_REG_MINLEN) {
3353 	    pimstat.pims_rcv_tooshort++;
3354 	    pimstat.pims_rcv_badregisters++;
3355 	    log(LOG_ERR,
3356 		"pim_input: register packet size too small %d from %lx\n",
3357 		datalen, (u_long)ip->ip_src.s_addr);
3358 	    m_freem(m);
3359 	    return;
3360 	}
3361 
3362 	reghdr = (u_int32_t *)(pim + 1);
3363 	encap_ip = (struct ip *)(reghdr + 1);
3364 
3365 	if (mrtdebug & DEBUG_PIM) {
3366 	    log(LOG_DEBUG,
3367 		"pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3368 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3369 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3370 		ntohs(encap_ip->ip_len));
3371 	}
3372 
3373 	/* verify the version number of the inner packet */
3374 	if (encap_ip->ip_v != IPVERSION) {
3375 	    pimstat.pims_rcv_badregisters++;
3376 	    if (mrtdebug & DEBUG_PIM) {
3377 		log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3378 		    "of the inner packet\n", encap_ip->ip_v);
3379 	    }
3380 	    m_freem(m);
3381 	    return;
3382 	}
3383 
3384 	/* verify the inner packet is destined to a mcast group */
3385 	if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3386 	    pimstat.pims_rcv_badregisters++;
3387 	    if (mrtdebug & DEBUG_PIM)
3388 		log(LOG_DEBUG,
3389 		    "pim_input: inner packet of register is not "
3390 		    "multicast %lx\n",
3391 		    (u_long)ntohl(encap_ip->ip_dst.s_addr));
3392 	    m_freem(m);
3393 	    return;
3394 	}
3395 
3396 	/* If a NULL_REGISTER, pass it to the daemon */
3397 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3398 	    goto pim_input_to_daemon;
3399 
3400 	/*
3401 	 * Copy the TOS from the outer IP header to the inner IP header.
3402 	 */
3403 	if (encap_ip->ip_tos != ip_tos) {
3404 	    /* Outer TOS -> inner TOS */
3405 	    encap_ip->ip_tos = ip_tos;
3406 	    /* Recompute the inner header checksum. Sigh... */
3407 
3408 	    /* adjust mbuf to point to the inner IP header */
3409 	    m->m_data += (iphlen + PIM_MINLEN);
3410 	    m->m_len  -= (iphlen + PIM_MINLEN);
3411 
3412 	    encap_ip->ip_sum = 0;
3413 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3414 
3415 	    /* restore mbuf to point back to the outer IP header */
3416 	    m->m_data -= (iphlen + PIM_MINLEN);
3417 	    m->m_len  += (iphlen + PIM_MINLEN);
3418 	}
3419 
3420 	/*
3421 	 * Decapsulate the inner IP packet and loopback to forward it
3422 	 * as a normal multicast packet. Also, make a copy of the
3423 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
3424 	 * to pass to the daemon later, so it can take the appropriate
3425 	 * actions (e.g., send back PIM_REGISTER_STOP).
3426 	 * XXX: here m->m_data points to the outer IP header.
3427 	 */
3428 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_DONTWAIT);
3429 	if (mcp == NULL) {
3430 	    log(LOG_ERR,
3431 		"pim_input: pim register: could not copy register head\n");
3432 	    m_freem(m);
3433 	    return;
3434 	}
3435 
3436 	/* Keep statistics */
3437 	/* XXX: registers_bytes include only the encap. mcast pkt */
3438 	pimstat.pims_rcv_registers_msgs++;
3439 	pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3440 
3441 	/*
3442 	 * forward the inner ip packet; point m_data at the inner ip.
3443 	 */
3444 	m_adj(m, iphlen + PIM_MINLEN);
3445 
3446 	if (mrtdebug & DEBUG_PIM) {
3447 	    log(LOG_DEBUG,
3448 		"pim_input: forwarding decapsulated register: "
3449 		"src %lx, dst %lx, vif %d\n",
3450 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3451 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3452 		reg_vif_num);
3453 	}
3454 	/* NB: vifp was collected above; can it change on us? */
3455 	looutput(vifp, m, (struct sockaddr *)&dst, NULL);
3456 
3457 	/* prepare the register head to send to the mrouting daemon */
3458 	m = mcp;
3459     }
3460 
3461 pim_input_to_daemon:
3462     /*
3463      * Pass the PIM message up to the daemon; if it is a Register message,
3464      * pass the 'head' only up to the daemon. This includes the
3465      * outer IP header, PIM header, PIM-Register header and the
3466      * inner IP header.
3467      * XXX: the outer IP header pkt size of a Register is not adjust to
3468      * reflect the fact that the inner multicast data is truncated.
3469      */
3470     rip_input(m, iphlen, proto);
3471 
3472     return;
3473 }
3474 #endif /* PIM */
3475