xref: /netbsd-src/sys/netinet/ip_mroute.c (revision ca453df649ce9db45b64d73678ba06cbccf9aa11)
1 /*	$NetBSD: ip_mroute.c,v 1.119 2011/07/17 20:54:53 joerg 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 1.119 2011/07/17 20:54:53 joerg Exp $");
97 
98 #include "opt_inet.h"
99 #include "opt_ipsec.h"
100 #include "opt_pim.h"
101 
102 #ifdef PIM
103 #define _PIM_VT 1
104 #endif
105 
106 #include <sys/param.h>
107 #include <sys/systm.h>
108 #include <sys/callout.h>
109 #include <sys/mbuf.h>
110 #include <sys/socket.h>
111 #include <sys/socketvar.h>
112 #include <sys/protosw.h>
113 #include <sys/errno.h>
114 #include <sys/time.h>
115 #include <sys/kernel.h>
116 #include <sys/ioctl.h>
117 #include <sys/syslog.h>
118 
119 #include <net/if.h>
120 #include <net/route.h>
121 #include <net/raw_cb.h>
122 
123 #include <netinet/in.h>
124 #include <netinet/in_var.h>
125 #include <netinet/in_systm.h>
126 #include <netinet/ip.h>
127 #include <netinet/ip_var.h>
128 #include <netinet/in_pcb.h>
129 #include <netinet/udp.h>
130 #include <netinet/igmp.h>
131 #include <netinet/igmp_var.h>
132 #include <netinet/ip_mroute.h>
133 #ifdef PIM
134 #include <netinet/pim.h>
135 #include <netinet/pim_var.h>
136 #endif
137 #include <netinet/ip_encap.h>
138 
139 #ifdef IPSEC
140 #include <netinet6/ipsec.h>
141 #include <netkey/key.h>
142 #endif
143 
144 #ifdef FAST_IPSEC
145 #include <netipsec/ipsec.h>
146 #include <netipsec/key.h>
147 #endif
148 
149 #define IP_MULTICASTOPTS 0
150 #define	M_PULLUP(m, len)						 \
151 	do {								 \
152 		if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
153 			(m) = m_pullup((m), (len));			 \
154 	} while (/*CONSTCOND*/ 0)
155 
156 /*
157  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
158  * except for netstat or debugging purposes.
159  */
160 struct socket  *ip_mrouter  = NULL;
161 int		ip_mrtproto = IGMP_DVMRP;    /* for netstat only */
162 
163 #define NO_RTE_FOUND 	0x1
164 #define RTE_FOUND	0x2
165 
166 #define	MFCHASH(a, g)							\
167 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^	\
168 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
169 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
170 u_long	mfchash;
171 
172 u_char		nexpire[MFCTBLSIZ];
173 struct vif	viftable[MAXVIFS];
174 struct mrtstat	mrtstat;
175 u_int		mrtdebug = 0;	  /* debug level 	*/
176 #define		DEBUG_MFC	0x02
177 #define		DEBUG_FORWARD	0x04
178 #define		DEBUG_EXPIRE	0x08
179 #define		DEBUG_XMIT	0x10
180 #define		DEBUG_PIM	0x20
181 
182 #define		VIFI_INVALID	((vifi_t) -1)
183 
184 u_int       	tbfdebug = 0;     /* tbf debug level 	*/
185 #ifdef RSVP_ISI
186 u_int		rsvpdebug = 0;	  /* rsvp debug level   */
187 extern struct socket *ip_rsvpd;
188 extern int rsvp_on;
189 #endif /* RSVP_ISI */
190 
191 /* vif attachment using sys/netinet/ip_encap.c */
192 static void vif_input(struct mbuf *, ...);
193 static int vif_encapcheck(struct mbuf *, int, int, void *);
194 
195 static const struct protosw vif_protosw =
196 { SOCK_RAW,	&inetdomain,	IPPROTO_IPV4,	PR_ATOMIC|PR_ADDR,
197   vif_input,	rip_output,	0,		rip_ctloutput,
198   rip_usrreq,
199   0,            0,              0,              0,
200 };
201 
202 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
203 #define		UPCALL_EXPIRE	6		/* number of timeouts */
204 
205 /*
206  * Define the token bucket filter structures
207  */
208 
209 #define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
210 
211 static int get_sg_cnt(struct sioc_sg_req *);
212 static int get_vif_cnt(struct sioc_vif_req *);
213 static int ip_mrouter_init(struct socket *, int);
214 static int set_assert(int);
215 static int add_vif(struct vifctl *);
216 static int del_vif(vifi_t *);
217 static void update_mfc_params(struct mfc *, struct mfcctl2 *);
218 static void init_mfc_params(struct mfc *, struct mfcctl2 *);
219 static void expire_mfc(struct mfc *);
220 static int add_mfc(struct sockopt *);
221 #ifdef UPCALL_TIMING
222 static void collate(struct timeval *);
223 #endif
224 static int del_mfc(struct sockopt *);
225 static int set_api_config(struct sockopt *); /* chose API capabilities */
226 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
227 static void expire_upcalls(void *);
228 #ifdef RSVP_ISI
229 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
230 #else
231 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *);
232 #endif
233 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
234 static void encap_send(struct ip *, struct vif *, struct mbuf *);
235 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_int32_t);
236 static void tbf_queue(struct vif *, struct mbuf *);
237 static void tbf_process_q(struct vif *);
238 static void tbf_reprocess_q(void *);
239 static int tbf_dq_sel(struct vif *, struct ip *);
240 static void tbf_send_packet(struct vif *, struct mbuf *);
241 static void tbf_update_tokens(struct vif *);
242 static int priority(struct vif *, struct ip *);
243 
244 /*
245  * Bandwidth monitoring
246  */
247 static void free_bw_list(struct bw_meter *);
248 static int add_bw_upcall(struct bw_upcall *);
249 static int del_bw_upcall(struct bw_upcall *);
250 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *);
251 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
252 static void bw_upcalls_send(void);
253 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
254 static void unschedule_bw_meter(struct bw_meter *);
255 static void bw_meter_process(void);
256 static void expire_bw_upcalls_send(void *);
257 static void expire_bw_meter_process(void *);
258 
259 #ifdef PIM
260 static int pim_register_send(struct ip *, struct vif *,
261 		struct mbuf *, struct mfc *);
262 static int pim_register_send_rp(struct ip *, struct vif *,
263 		struct mbuf *, struct mfc *);
264 static int pim_register_send_upcall(struct ip *, struct vif *,
265 		struct mbuf *, struct mfc *);
266 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
267 #endif
268 
269 /*
270  * 'Interfaces' associated with decapsulator (so we can tell
271  * packets that went through it from ones that get reflected
272  * by a broken gateway).  These interfaces are never linked into
273  * the system ifnet list & no routes point to them.  I.e., packets
274  * can't be sent this way.  They only exist as a placeholder for
275  * multicast source verification.
276  */
277 #if 0
278 struct ifnet multicast_decap_if[MAXVIFS];
279 #endif
280 
281 #define	ENCAP_TTL	64
282 #define	ENCAP_PROTO	IPPROTO_IPIP	/* 4 */
283 
284 /* prototype IP hdr for encapsulated packets */
285 struct ip multicast_encap_iphdr = {
286 	.ip_hl = sizeof(struct ip) >> 2,
287 	.ip_v = IPVERSION,
288 	.ip_len = sizeof(struct ip),
289 	.ip_ttl = ENCAP_TTL,
290 	.ip_p = ENCAP_PROTO,
291 };
292 
293 /*
294  * Bandwidth meter variables and constants
295  */
296 
297 /*
298  * Pending timeouts are stored in a hash table, the key being the
299  * expiration time. Periodically, the entries are analysed and processed.
300  */
301 #define BW_METER_BUCKETS	1024
302 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
303 struct callout bw_meter_ch;
304 #define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
305 
306 /*
307  * Pending upcalls are stored in a vector which is flushed when
308  * full, or periodically
309  */
310 static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
311 static u_int	bw_upcalls_n; /* # of pending upcalls */
312 struct callout	bw_upcalls_ch;
313 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
314 
315 #ifdef PIM
316 struct pimstat pimstat;
317 
318 /*
319  * Note: the PIM Register encapsulation adds the following in front of a
320  * data packet:
321  *
322  * struct pim_encap_hdr {
323  *    struct ip ip;
324  *    struct pim_encap_pimhdr  pim;
325  * }
326  *
327  */
328 
329 struct pim_encap_pimhdr {
330 	struct pim pim;
331 	uint32_t   flags;
332 };
333 
334 static struct ip pim_encap_iphdr = {
335 	.ip_v = IPVERSION,
336 	.ip_hl = sizeof(struct ip) >> 2,
337 	.ip_len = sizeof(struct ip),
338 	.ip_ttl = ENCAP_TTL,
339 	.ip_p = IPPROTO_PIM,
340 };
341 
342 static struct pim_encap_pimhdr pim_encap_pimhdr = {
343     {
344 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
345 	0,			/* reserved */
346 	0,			/* checksum */
347     },
348     0				/* flags */
349 };
350 
351 static struct ifnet multicast_register_if;
352 static vifi_t reg_vif_num = VIFI_INVALID;
353 #endif /* PIM */
354 
355 
356 /*
357  * Private variables.
358  */
359 static vifi_t	   numvifs = 0;
360 
361 static struct callout expire_upcalls_ch;
362 
363 /*
364  * whether or not special PIM assert processing is enabled.
365  */
366 static int pim_assert;
367 /*
368  * Rate limit for assert notification messages, in usec
369  */
370 #define ASSERT_MSG_TIME		3000000
371 
372 /*
373  * Kernel multicast routing API capabilities and setup.
374  * If more API capabilities are added to the kernel, they should be
375  * recorded in `mrt_api_support'.
376  */
377 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
378 					  MRT_MFC_FLAGS_BORDER_VIF |
379 					  MRT_MFC_RP |
380 					  MRT_MFC_BW_UPCALL);
381 static u_int32_t mrt_api_config = 0;
382 
383 /*
384  * Find a route for a given origin IP address and Multicast group address
385  * Type of service parameter to be added in the future!!!
386  * Statistics are updated by the caller if needed
387  * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
388  */
389 static struct mfc *
390 mfc_find(struct in_addr *o, struct in_addr *g)
391 {
392 	struct mfc *rt;
393 
394 	LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
395 		if (in_hosteq(rt->mfc_origin, *o) &&
396 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
397 		    (rt->mfc_stall == NULL))
398 			break;
399 	}
400 
401 	return (rt);
402 }
403 
404 /*
405  * Macros to compute elapsed time efficiently
406  * Borrowed from Van Jacobson's scheduling code
407  */
408 #define TV_DELTA(a, b, delta) do {					\
409 	int xxs;							\
410 	delta = (a).tv_usec - (b).tv_usec;				\
411 	xxs = (a).tv_sec - (b).tv_sec;					\
412 	switch (xxs) {							\
413 	case 2:								\
414 		delta += 1000000;					\
415 		/* fall through */					\
416 	case 1:								\
417 		delta += 1000000;					\
418 		/* fall through */					\
419 	case 0:								\
420 		break;							\
421 	default:							\
422 		delta += (1000000 * xxs);				\
423 		break;							\
424 	}								\
425 } while (/*CONSTCOND*/ 0)
426 
427 #ifdef UPCALL_TIMING
428 u_int32_t upcall_data[51];
429 #endif /* UPCALL_TIMING */
430 
431 /*
432  * Handle MRT setsockopt commands to modify the multicast routing tables.
433  */
434 int
435 ip_mrouter_set(struct socket *so, struct sockopt *sopt)
436 {
437 	int error;
438 	int optval;
439 	struct vifctl vifc;
440 	vifi_t vifi;
441 	struct bw_upcall bwuc;
442 
443 	if (sopt->sopt_name != MRT_INIT && so != ip_mrouter)
444 		error = ENOPROTOOPT;
445 	else {
446 		switch (sopt->sopt_name) {
447 		case MRT_INIT:
448 			error = sockopt_getint(sopt, &optval);
449 			if (error)
450 				break;
451 
452 			error = ip_mrouter_init(so, optval);
453 			break;
454 		case MRT_DONE:
455 			error = ip_mrouter_done();
456 			break;
457 		case MRT_ADD_VIF:
458 			error = sockopt_get(sopt, &vifc, sizeof(vifc));
459 			if (error)
460 				break;
461 			error = add_vif(&vifc);
462 			break;
463 		case MRT_DEL_VIF:
464 			error = sockopt_get(sopt, &vifi, sizeof(vifi));
465 			if (error)
466 				break;
467 			error = del_vif(&vifi);
468 			break;
469 		case MRT_ADD_MFC:
470 			error = add_mfc(sopt);
471 			break;
472 		case MRT_DEL_MFC:
473 			error = del_mfc(sopt);
474 			break;
475 		case MRT_ASSERT:
476 			error = sockopt_getint(sopt, &optval);
477 			if (error)
478 				break;
479 			error = set_assert(optval);
480 			break;
481 		case MRT_API_CONFIG:
482 			error = set_api_config(sopt);
483 			break;
484 		case MRT_ADD_BW_UPCALL:
485 			error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
486 			if (error)
487 				break;
488 			error = add_bw_upcall(&bwuc);
489 			break;
490 		case MRT_DEL_BW_UPCALL:
491 			error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
492 			if (error)
493 				break;
494 			error = del_bw_upcall(&bwuc);
495 			break;
496 		default:
497 			error = ENOPROTOOPT;
498 			break;
499 		}
500 	}
501 	return (error);
502 }
503 
504 /*
505  * Handle MRT getsockopt commands
506  */
507 int
508 ip_mrouter_get(struct socket *so, struct sockopt *sopt)
509 {
510 	int error;
511 
512 	if (so != ip_mrouter)
513 		error = ENOPROTOOPT;
514 	else {
515 		switch (sopt->sopt_name) {
516 		case MRT_VERSION:
517 			error = sockopt_setint(sopt, 0x0305); /* XXX !!!! */
518 			break;
519 		case MRT_ASSERT:
520 			error = sockopt_setint(sopt, pim_assert);
521 			break;
522 		case MRT_API_SUPPORT:
523 			error = sockopt_set(sopt, &mrt_api_support,
524 			    sizeof(mrt_api_support));
525 			break;
526 		case MRT_API_CONFIG:
527 			error = sockopt_set(sopt, &mrt_api_config,
528 			    sizeof(mrt_api_config));
529 			break;
530 		default:
531 			error = ENOPROTOOPT;
532 			break;
533 		}
534 	}
535 	return (error);
536 }
537 
538 /*
539  * Handle ioctl commands to obtain information from the cache
540  */
541 int
542 mrt_ioctl(struct socket *so, u_long cmd, void *data)
543 {
544 	int error;
545 
546 	if (so != ip_mrouter)
547 		error = EINVAL;
548 	else
549 		switch (cmd) {
550 		case SIOCGETVIFCNT:
551 			error = get_vif_cnt((struct sioc_vif_req *)data);
552 			break;
553 		case SIOCGETSGCNT:
554 			error = get_sg_cnt((struct sioc_sg_req *)data);
555 			break;
556 		default:
557 			error = EINVAL;
558 			break;
559 		}
560 
561 	return (error);
562 }
563 
564 /*
565  * returns the packet, byte, rpf-failure count for the source group provided
566  */
567 static int
568 get_sg_cnt(struct sioc_sg_req *req)
569 {
570 	int s;
571 	struct mfc *rt;
572 
573 	s = splsoftnet();
574 	rt = mfc_find(&req->src, &req->grp);
575 	if (rt == NULL) {
576 		splx(s);
577 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
578 		return (EADDRNOTAVAIL);
579 	}
580 	req->pktcnt = rt->mfc_pkt_cnt;
581 	req->bytecnt = rt->mfc_byte_cnt;
582 	req->wrong_if = rt->mfc_wrong_if;
583 	splx(s);
584 
585 	return (0);
586 }
587 
588 /*
589  * returns the input and output packet and byte counts on the vif provided
590  */
591 static int
592 get_vif_cnt(struct sioc_vif_req *req)
593 {
594 	vifi_t vifi = req->vifi;
595 
596 	if (vifi >= numvifs)
597 		return (EINVAL);
598 
599 	req->icount = viftable[vifi].v_pkt_in;
600 	req->ocount = viftable[vifi].v_pkt_out;
601 	req->ibytes = viftable[vifi].v_bytes_in;
602 	req->obytes = viftable[vifi].v_bytes_out;
603 
604 	return (0);
605 }
606 
607 /*
608  * Enable multicast routing
609  */
610 static int
611 ip_mrouter_init(struct socket *so, int v)
612 {
613 	if (mrtdebug)
614 		log(LOG_DEBUG,
615 		    "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
616 		    so->so_type, so->so_proto->pr_protocol);
617 
618 	if (so->so_type != SOCK_RAW ||
619 	    so->so_proto->pr_protocol != IPPROTO_IGMP)
620 		return (EOPNOTSUPP);
621 
622 	if (v != 1)
623 		return (EINVAL);
624 
625 	if (ip_mrouter != NULL)
626 		return (EADDRINUSE);
627 
628 	ip_mrouter = so;
629 
630 	mfchashtbl = hashinit(MFCTBLSIZ, HASH_LIST, true, &mfchash);
631 	memset((void *)nexpire, 0, sizeof(nexpire));
632 
633 	pim_assert = 0;
634 
635 	callout_init(&expire_upcalls_ch, 0);
636 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
637 		      expire_upcalls, NULL);
638 
639 	callout_init(&bw_upcalls_ch, 0);
640 	callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
641 		      expire_bw_upcalls_send, NULL);
642 
643 	callout_init(&bw_meter_ch, 0);
644 	callout_reset(&bw_meter_ch, BW_METER_PERIOD,
645 		      expire_bw_meter_process, NULL);
646 
647 	if (mrtdebug)
648 		log(LOG_DEBUG, "ip_mrouter_init\n");
649 
650 	return (0);
651 }
652 
653 /*
654  * Disable multicast routing
655  */
656 int
657 ip_mrouter_done(void)
658 {
659 	vifi_t vifi;
660 	struct vif *vifp;
661 	int i;
662 	int s;
663 
664 	s = splsoftnet();
665 
666 	/* Clear out all the vifs currently in use. */
667 	for (vifi = 0; vifi < numvifs; vifi++) {
668 		vifp = &viftable[vifi];
669 		if (!in_nullhost(vifp->v_lcl_addr))
670 			reset_vif(vifp);
671 	}
672 
673 	numvifs = 0;
674 	pim_assert = 0;
675 	mrt_api_config = 0;
676 
677 	callout_stop(&expire_upcalls_ch);
678 	callout_stop(&bw_upcalls_ch);
679 	callout_stop(&bw_meter_ch);
680 
681 	/*
682 	 * Free all multicast forwarding cache entries.
683 	 */
684 	for (i = 0; i < MFCTBLSIZ; i++) {
685 		struct mfc *rt, *nrt;
686 
687 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
688 			nrt = LIST_NEXT(rt, mfc_hash);
689 
690 			expire_mfc(rt);
691 		}
692 	}
693 
694 	memset((void *)nexpire, 0, sizeof(nexpire));
695 	hashdone(mfchashtbl, HASH_LIST, mfchash);
696 	mfchashtbl = NULL;
697 
698 	bw_upcalls_n = 0;
699 	memset(bw_meter_timers, 0, sizeof(bw_meter_timers));
700 
701 	/* Reset de-encapsulation cache. */
702 
703 	ip_mrouter = NULL;
704 
705 	splx(s);
706 
707 	if (mrtdebug)
708 		log(LOG_DEBUG, "ip_mrouter_done\n");
709 
710 	return (0);
711 }
712 
713 void
714 ip_mrouter_detach(struct ifnet *ifp)
715 {
716 	int vifi, i;
717 	struct vif *vifp;
718 	struct mfc *rt;
719 	struct rtdetq *rte;
720 
721 	/* XXX not sure about side effect to userland routing daemon */
722 	for (vifi = 0; vifi < numvifs; vifi++) {
723 		vifp = &viftable[vifi];
724 		if (vifp->v_ifp == ifp)
725 			reset_vif(vifp);
726 	}
727 	for (i = 0; i < MFCTBLSIZ; i++) {
728 		if (nexpire[i] == 0)
729 			continue;
730 		LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
731 			for (rte = rt->mfc_stall; rte; rte = rte->next) {
732 				if (rte->ifp == ifp)
733 					rte->ifp = NULL;
734 			}
735 		}
736 	}
737 }
738 
739 /*
740  * Set PIM assert processing global
741  */
742 static int
743 set_assert(int i)
744 {
745 	pim_assert = !!i;
746 	return (0);
747 }
748 
749 /*
750  * Configure API capabilities
751  */
752 static int
753 set_api_config(struct sockopt *sopt)
754 {
755 	u_int32_t apival;
756 	int i, error;
757 
758 	/*
759 	 * We can set the API capabilities only if it is the first operation
760 	 * after MRT_INIT. I.e.:
761 	 *  - there are no vifs installed
762 	 *  - pim_assert is not enabled
763 	 *  - the MFC table is empty
764 	 */
765 	error = sockopt_get(sopt, &apival, sizeof(apival));
766 	if (error)
767 		return (error);
768 	if (numvifs > 0)
769 		return (EPERM);
770 	if (pim_assert)
771 		return (EPERM);
772 	for (i = 0; i < MFCTBLSIZ; i++) {
773 		if (LIST_FIRST(&mfchashtbl[i]) != NULL)
774 			return (EPERM);
775 	}
776 
777 	mrt_api_config = apival & mrt_api_support;
778 	return (0);
779 }
780 
781 /*
782  * Add a vif to the vif table
783  */
784 static int
785 add_vif(struct vifctl *vifcp)
786 {
787 	struct vif *vifp;
788 	struct ifaddr *ifa;
789 	struct ifnet *ifp;
790 	struct ifreq ifr;
791 	int error, s;
792 	struct sockaddr_in sin;
793 
794 	if (vifcp->vifc_vifi >= MAXVIFS)
795 		return (EINVAL);
796 	if (in_nullhost(vifcp->vifc_lcl_addr))
797 		return (EADDRNOTAVAIL);
798 
799 	vifp = &viftable[vifcp->vifc_vifi];
800 	if (!in_nullhost(vifp->v_lcl_addr))
801 		return (EADDRINUSE);
802 
803 	/* Find the interface with an address in AF_INET family. */
804 #ifdef PIM
805 	if (vifcp->vifc_flags & VIFF_REGISTER) {
806 		/*
807 		 * XXX: Because VIFF_REGISTER does not really need a valid
808 		 * local interface (e.g. it could be 127.0.0.2), we don't
809 		 * check its address.
810 		 */
811 	    ifp = NULL;
812 	} else
813 #endif
814 	{
815 		sockaddr_in_init(&sin, &vifcp->vifc_lcl_addr, 0);
816 		ifa = ifa_ifwithaddr(sintosa(&sin));
817 		if (ifa == NULL)
818 			return (EADDRNOTAVAIL);
819 		ifp = ifa->ifa_ifp;
820 	}
821 
822 	if (vifcp->vifc_flags & VIFF_TUNNEL) {
823 		if (vifcp->vifc_flags & VIFF_SRCRT) {
824 			log(LOG_ERR, "source routed tunnels not supported\n");
825 			return (EOPNOTSUPP);
826 		}
827 
828 		/* attach this vif to decapsulator dispatch table */
829 		/*
830 		 * XXX Use addresses in registration so that matching
831 		 * can be done with radix tree in decapsulator.  But,
832 		 * we need to check inner header for multicast, so
833 		 * this requires both radix tree lookup and then a
834 		 * function to check, and this is not supported yet.
835 		 */
836 		vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
837 		    vif_encapcheck, &vif_protosw, vifp);
838 		if (!vifp->v_encap_cookie)
839 			return (EINVAL);
840 
841 		/* Create a fake encapsulation interface. */
842 		ifp = malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK|M_ZERO);
843 		snprintf(ifp->if_xname, sizeof(ifp->if_xname),
844 			 "mdecap%d", vifcp->vifc_vifi);
845 
846 		/* Prepare cached route entry. */
847 		memset(&vifp->v_route, 0, sizeof(vifp->v_route));
848 #ifdef PIM
849 	} else if (vifcp->vifc_flags & VIFF_REGISTER) {
850 		ifp = &multicast_register_if;
851 		if (mrtdebug)
852 			log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
853 			    (void *)ifp);
854 		if (reg_vif_num == VIFI_INVALID) {
855 			memset(ifp, 0, sizeof(*ifp));
856 			snprintf(ifp->if_xname, sizeof(ifp->if_xname),
857 				 "register_vif");
858 			ifp->if_flags = IFF_LOOPBACK;
859 			memset(&vifp->v_route, 0, sizeof(vifp->v_route));
860 			reg_vif_num = vifcp->vifc_vifi;
861 		}
862 #endif
863 	} else {
864 		/* Make sure the interface supports multicast. */
865 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
866 			return (EOPNOTSUPP);
867 
868 		/* Enable promiscuous reception of all IP multicasts. */
869 		sockaddr_in_init(&sin, &zeroin_addr, 0);
870 		ifreq_setaddr(SIOCADDMULTI, &ifr, sintosa(&sin));
871 		error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, &ifr);
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 ifreq ifr;
930 	struct sockaddr_in sin;
931 
932 	callout_stop(&vifp->v_repq_ch);
933 
934 	/* detach this vif from decapsulator dispatch table */
935 	encap_detach(vifp->v_encap_cookie);
936 	vifp->v_encap_cookie = NULL;
937 
938 	/*
939 	 * Free packets queued at the interface
940 	 */
941 	for (m = vifp->tbf_q; m != NULL; m = n) {
942 		n = m->m_nextpkt;
943 		m_freem(m);
944 	}
945 
946 	if (vifp->v_flags & VIFF_TUNNEL)
947 		free(vifp->v_ifp, M_MRTABLE);
948 	else if (vifp->v_flags & VIFF_REGISTER) {
949 #ifdef PIM
950 		reg_vif_num = VIFI_INVALID;
951 #endif
952 	} else {
953 		sockaddr_in_init(&sin, &zeroin_addr, 0);
954 		ifreq_setaddr(SIOCDELMULTI, &ifr, sintosa(&sin));
955 		ifp = vifp->v_ifp;
956 		(*ifp->if_ioctl)(ifp, SIOCDELMULTI, &ifr);
957 	}
958 	memset((void *)vifp, 0, sizeof(*vifp));
959 }
960 
961 /*
962  * Delete a vif from the vif table
963  */
964 static int
965 del_vif(vifi_t *vifip)
966 {
967 	struct vif *vifp;
968 	vifi_t vifi;
969 	int s;
970 
971 	if (*vifip >= numvifs)
972 		return (EINVAL);
973 
974 	vifp = &viftable[*vifip];
975 	if (in_nullhost(vifp->v_lcl_addr))
976 		return (EADDRNOTAVAIL);
977 
978 	s = splsoftnet();
979 
980 	reset_vif(vifp);
981 
982 	/* Adjust numvifs down */
983 	for (vifi = numvifs; vifi > 0; vifi--)
984 		if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
985 			break;
986 	numvifs = vifi;
987 
988 	splx(s);
989 
990 	if (mrtdebug)
991 		log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
992 
993 	return (0);
994 }
995 
996 /*
997  * update an mfc entry without resetting counters and S,G addresses.
998  */
999 static void
1000 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1001 {
1002 	int i;
1003 
1004 	rt->mfc_parent = mfccp->mfcc_parent;
1005 	for (i = 0; i < numvifs; i++) {
1006 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1007 		rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1008 			MRT_MFC_FLAGS_ALL;
1009 	}
1010 	/* set the RP address */
1011 	if (mrt_api_config & MRT_MFC_RP)
1012 		rt->mfc_rp = mfccp->mfcc_rp;
1013 	else
1014 		rt->mfc_rp = zeroin_addr;
1015 }
1016 
1017 /*
1018  * fully initialize an mfc entry from the parameter.
1019  */
1020 static void
1021 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1022 {
1023 	rt->mfc_origin     = mfccp->mfcc_origin;
1024 	rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1025 
1026 	update_mfc_params(rt, mfccp);
1027 
1028 	/* initialize pkt counters per src-grp */
1029 	rt->mfc_pkt_cnt    = 0;
1030 	rt->mfc_byte_cnt   = 0;
1031 	rt->mfc_wrong_if   = 0;
1032 	timerclear(&rt->mfc_last_assert);
1033 }
1034 
1035 static void
1036 expire_mfc(struct mfc *rt)
1037 {
1038 	struct rtdetq *rte, *nrte;
1039 
1040 	free_bw_list(rt->mfc_bw_meter);
1041 
1042 	for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1043 		nrte = rte->next;
1044 		m_freem(rte->m);
1045 		free(rte, M_MRTABLE);
1046 	}
1047 
1048 	LIST_REMOVE(rt, mfc_hash);
1049 	free(rt, M_MRTABLE);
1050 }
1051 
1052 /*
1053  * Add an mfc entry
1054  */
1055 static int
1056 add_mfc(struct sockopt *sopt)
1057 {
1058 	struct mfcctl2 mfcctl2;
1059 	struct mfcctl2 *mfccp;
1060 	struct mfc *rt;
1061 	u_int32_t hash = 0;
1062 	struct rtdetq *rte, *nrte;
1063 	u_short nstl;
1064 	int s;
1065 	int mfcctl_size = sizeof(struct mfcctl);
1066 	int error;
1067 
1068 	if (mrt_api_config & MRT_API_FLAGS_ALL)
1069 		mfcctl_size = sizeof(struct mfcctl2);
1070 
1071 	/*
1072 	 * select data size depending on API version.
1073 	 */
1074 	mfccp = &mfcctl2;
1075 	memset(&mfcctl2, 0, sizeof(mfcctl2));
1076 
1077 	if (mrt_api_config & MRT_API_FLAGS_ALL)
1078 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1079 	else
1080 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1081 
1082 	if (error)
1083 		return (error);
1084 
1085 	s = splsoftnet();
1086 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1087 
1088 	/* If an entry already exists, just update the fields */
1089 	if (rt) {
1090 		if (mrtdebug & DEBUG_MFC)
1091 			log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1092 			    ntohl(mfccp->mfcc_origin.s_addr),
1093 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1094 			    mfccp->mfcc_parent);
1095 
1096 		update_mfc_params(rt, mfccp);
1097 
1098 		splx(s);
1099 		return (0);
1100 	}
1101 
1102 	/*
1103 	 * Find the entry for which the upcall was made and update
1104 	 */
1105 	nstl = 0;
1106 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1107 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1108 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1109 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1110 		    rt->mfc_stall != NULL) {
1111 			if (nstl++)
1112 				log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
1113 				    "multiple kernel entries",
1114 				    ntohl(mfccp->mfcc_origin.s_addr),
1115 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1116 				    mfccp->mfcc_parent, rt->mfc_stall);
1117 
1118 			if (mrtdebug & DEBUG_MFC)
1119 				log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n",
1120 				    ntohl(mfccp->mfcc_origin.s_addr),
1121 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1122 				    mfccp->mfcc_parent, rt->mfc_stall);
1123 
1124 			rte = rt->mfc_stall;
1125 			init_mfc_params(rt, mfccp);
1126 			rt->mfc_stall = NULL;
1127 
1128 			rt->mfc_expire = 0; /* Don't clean this guy up */
1129 			nexpire[hash]--;
1130 
1131 			/* free packets Qed at the end of this entry */
1132 			for (; rte != NULL; rte = nrte) {
1133 				nrte = rte->next;
1134 				if (rte->ifp) {
1135 #ifdef RSVP_ISI
1136 					ip_mdq(rte->m, rte->ifp, rt, -1);
1137 #else
1138 					ip_mdq(rte->m, rte->ifp, rt);
1139 #endif /* RSVP_ISI */
1140 				}
1141 				m_freem(rte->m);
1142 #ifdef UPCALL_TIMING
1143 				collate(&rte->t);
1144 #endif /* UPCALL_TIMING */
1145 				free(rte, M_MRTABLE);
1146 			}
1147 		}
1148 	}
1149 
1150 	/*
1151 	 * It is possible that an entry is being inserted without an upcall
1152 	 */
1153 	if (nstl == 0) {
1154 		/*
1155 		 * No mfc; make a new one
1156 		 */
1157 		if (mrtdebug & DEBUG_MFC)
1158 			log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1159 			    ntohl(mfccp->mfcc_origin.s_addr),
1160 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1161 			    mfccp->mfcc_parent);
1162 
1163 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1164 			if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1165 			    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1166 				init_mfc_params(rt, mfccp);
1167 				if (rt->mfc_expire)
1168 					nexpire[hash]--;
1169 				rt->mfc_expire = 0;
1170 				break; /* XXX */
1171 			}
1172 		}
1173 		if (rt == NULL) {	/* no upcall, so make a new entry */
1174 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1175 						  M_NOWAIT);
1176 			if (rt == NULL) {
1177 				splx(s);
1178 				return (ENOBUFS);
1179 			}
1180 
1181 			init_mfc_params(rt, mfccp);
1182 			rt->mfc_expire	= 0;
1183 			rt->mfc_stall	= NULL;
1184 			rt->mfc_bw_meter = NULL;
1185 
1186 			/* insert new entry at head of hash chain */
1187 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1188 		}
1189 	}
1190 
1191 	splx(s);
1192 	return (0);
1193 }
1194 
1195 #ifdef UPCALL_TIMING
1196 /*
1197  * collect delay statistics on the upcalls
1198  */
1199 static void
1200 collate(struct timeval *t)
1201 {
1202 	u_int32_t d;
1203 	struct timeval tp;
1204 	u_int32_t delta;
1205 
1206 	microtime(&tp);
1207 
1208 	if (timercmp(t, &tp, <)) {
1209 		TV_DELTA(tp, *t, delta);
1210 
1211 		d = delta >> 10;
1212 		if (d > 50)
1213 			d = 50;
1214 
1215 		++upcall_data[d];
1216 	}
1217 }
1218 #endif /* UPCALL_TIMING */
1219 
1220 /*
1221  * Delete an mfc entry
1222  */
1223 static int
1224 del_mfc(struct sockopt *sopt)
1225 {
1226 	struct mfcctl2 mfcctl2;
1227 	struct mfcctl2 *mfccp;
1228 	struct mfc *rt;
1229 	int s;
1230 	int error;
1231 
1232 	/*
1233 	 * XXX: for deleting MFC entries the information in entries
1234 	 * of size "struct mfcctl" is sufficient.
1235 	 */
1236 
1237 	mfccp = &mfcctl2;
1238 	memset(&mfcctl2, 0, sizeof(mfcctl2));
1239 
1240 	error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1241 	if (error) {
1242 		/* Try with the size of mfcctl2. */
1243 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1244 		if (error)
1245 			return (error);
1246 	}
1247 
1248 	if (mrtdebug & DEBUG_MFC)
1249 		log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1250 		    ntohl(mfccp->mfcc_origin.s_addr),
1251 		    ntohl(mfccp->mfcc_mcastgrp.s_addr));
1252 
1253 	s = splsoftnet();
1254 
1255 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1256 	if (rt == NULL) {
1257 		splx(s);
1258 		return (EADDRNOTAVAIL);
1259 	}
1260 
1261 	/*
1262 	 * free the bw_meter entries
1263 	 */
1264 	free_bw_list(rt->mfc_bw_meter);
1265 	rt->mfc_bw_meter = NULL;
1266 
1267 	LIST_REMOVE(rt, mfc_hash);
1268 	free(rt, M_MRTABLE);
1269 
1270 	splx(s);
1271 	return (0);
1272 }
1273 
1274 static int
1275 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1276 {
1277 	if (s) {
1278 		if (sbappendaddr(&s->so_rcv, sintosa(src), mm,
1279 		    (struct mbuf *)NULL) != 0) {
1280 			sorwakeup(s);
1281 			return (0);
1282 		}
1283 	}
1284 	m_freem(mm);
1285 	return (-1);
1286 }
1287 
1288 /*
1289  * IP multicast forwarding function. This function assumes that the packet
1290  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1291  * pointed to by "ifp", and the packet is to be relayed to other networks
1292  * that have members of the packet's destination IP multicast group.
1293  *
1294  * The packet is returned unscathed to the caller, unless it is
1295  * erroneous, in which case a non-zero return value tells the caller to
1296  * discard it.
1297  */
1298 
1299 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1300 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1301 
1302 int
1303 #ifdef RSVP_ISI
1304 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo)
1305 #else
1306 ip_mforward(struct mbuf *m, struct ifnet *ifp)
1307 #endif /* RSVP_ISI */
1308 {
1309 	struct ip *ip = mtod(m, struct ip *);
1310 	struct mfc *rt;
1311 	static int srctun = 0;
1312 	struct mbuf *mm;
1313 	struct sockaddr_in sin;
1314 	int s;
1315 	vifi_t vifi;
1316 
1317 	if (mrtdebug & DEBUG_FORWARD)
1318 		log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1319 		    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1320 
1321 	if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1322 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1323 		/*
1324 		 * Packet arrived via a physical interface or
1325 		 * an encapsulated tunnel or a register_vif.
1326 		 */
1327 	} else {
1328 		/*
1329 		 * Packet arrived through a source-route tunnel.
1330 		 * Source-route tunnels are no longer supported.
1331 		 */
1332 		if ((srctun++ % 1000) == 0)
1333 			log(LOG_ERR,
1334 			    "ip_mforward: received source-routed packet from %x\n",
1335 			    ntohl(ip->ip_src.s_addr));
1336 
1337 		return (1);
1338 	}
1339 
1340 	/*
1341 	 * Clear any in-bound checksum flags for this packet.
1342 	 */
1343 	m->m_pkthdr.csum_flags = 0;
1344 
1345 #ifdef RSVP_ISI
1346 	if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1347 		if (ip->ip_ttl < MAXTTL)
1348 			ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1349 		if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1350 			struct vif *vifp = viftable + vifi;
1351 			printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n",
1352 			    ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1353 			    (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1354 			    vifp->v_ifp->if_xname);
1355 		}
1356 		return (ip_mdq(m, ifp, (struct mfc *)NULL, vifi));
1357 	}
1358 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1359 		printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
1360 		    ntohl(ip->ip_src), ntohl(ip->ip_dst));
1361 	}
1362 #endif /* RSVP_ISI */
1363 
1364 	/*
1365 	 * Don't forward a packet with time-to-live of zero or one,
1366 	 * or a packet destined to a local-only group.
1367 	 */
1368 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1369 		return (0);
1370 
1371 	/*
1372 	 * Determine forwarding vifs from the forwarding cache table
1373 	 */
1374 	s = splsoftnet();
1375 	++mrtstat.mrts_mfc_lookups;
1376 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1377 
1378 	/* Entry exists, so forward if necessary */
1379 	if (rt != NULL) {
1380 		splx(s);
1381 #ifdef RSVP_ISI
1382 		return (ip_mdq(m, ifp, rt, -1));
1383 #else
1384 		return (ip_mdq(m, ifp, rt));
1385 #endif /* RSVP_ISI */
1386 	} else {
1387 		/*
1388 		 * If we don't have a route for packet's origin,
1389 		 * Make a copy of the packet & send message to routing daemon
1390 		 */
1391 
1392 		struct mbuf *mb0;
1393 		struct rtdetq *rte;
1394 		u_int32_t hash;
1395 		int hlen = ip->ip_hl << 2;
1396 #ifdef UPCALL_TIMING
1397 		struct timeval tp;
1398 
1399 		microtime(&tp);
1400 #endif /* UPCALL_TIMING */
1401 
1402 		++mrtstat.mrts_mfc_misses;
1403 
1404 		mrtstat.mrts_no_route++;
1405 		if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1406 			log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1407 			    ntohl(ip->ip_src.s_addr),
1408 			    ntohl(ip->ip_dst.s_addr));
1409 
1410 		/*
1411 		 * Allocate mbufs early so that we don't do extra work if we are
1412 		 * just going to fail anyway.  Make sure to pullup the header so
1413 		 * that other people can't step on it.
1414 		 */
1415 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1416 					      M_NOWAIT);
1417 		if (rte == NULL) {
1418 			splx(s);
1419 			return (ENOBUFS);
1420 		}
1421 		mb0 = m_copypacket(m, M_DONTWAIT);
1422 		M_PULLUP(mb0, hlen);
1423 		if (mb0 == NULL) {
1424 			free(rte, M_MRTABLE);
1425 			splx(s);
1426 			return (ENOBUFS);
1427 		}
1428 
1429 		/* is there an upcall waiting for this flow? */
1430 		hash = MFCHASH(ip->ip_src, ip->ip_dst);
1431 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1432 			if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1433 			    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1434 			    rt->mfc_stall != NULL)
1435 				break;
1436 		}
1437 
1438 		if (rt == NULL) {
1439 			int i;
1440 			struct igmpmsg *im;
1441 
1442 			/*
1443 			 * Locate the vifi for the incoming interface for
1444 			 * this packet.
1445 			 * If none found, drop packet.
1446 			 */
1447 			for (vifi = 0; vifi < numvifs &&
1448 				 viftable[vifi].v_ifp != ifp; vifi++)
1449 				;
1450 			if (vifi >= numvifs) /* vif not found, drop packet */
1451 				goto non_fatal;
1452 
1453 			/* no upcall, so make a new entry */
1454 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1455 						  M_NOWAIT);
1456 			if (rt == NULL)
1457 				goto fail;
1458 
1459 			/*
1460 			 * Make a copy of the header to send to the user level
1461 			 * process
1462 			 */
1463 			mm = m_copym(m, 0, hlen, M_DONTWAIT);
1464 			M_PULLUP(mm, hlen);
1465 			if (mm == NULL)
1466 				goto fail1;
1467 
1468 			/*
1469 			 * Send message to routing daemon to install
1470 			 * a route into the kernel table
1471 			 */
1472 
1473 			im = mtod(mm, struct igmpmsg *);
1474 			im->im_msgtype = IGMPMSG_NOCACHE;
1475 			im->im_mbz = 0;
1476 			im->im_vif = vifi;
1477 
1478 			mrtstat.mrts_upcalls++;
1479 
1480 			sockaddr_in_init(&sin, &ip->ip_src, 0);
1481 			if (socket_send(ip_mrouter, mm, &sin) < 0) {
1482 				log(LOG_WARNING,
1483 				    "ip_mforward: ip_mrouter socket queue full\n");
1484 				++mrtstat.mrts_upq_sockfull;
1485 			fail1:
1486 				free(rt, M_MRTABLE);
1487 			fail:
1488 				free(rte, M_MRTABLE);
1489 				m_freem(mb0);
1490 				splx(s);
1491 				return (ENOBUFS);
1492 			}
1493 
1494 			/* insert new entry at head of hash chain */
1495 			rt->mfc_origin = ip->ip_src;
1496 			rt->mfc_mcastgrp = ip->ip_dst;
1497 			rt->mfc_pkt_cnt = 0;
1498 			rt->mfc_byte_cnt = 0;
1499 			rt->mfc_wrong_if = 0;
1500 			rt->mfc_expire = UPCALL_EXPIRE;
1501 			nexpire[hash]++;
1502 			for (i = 0; i < numvifs; i++) {
1503 				rt->mfc_ttls[i] = 0;
1504 				rt->mfc_flags[i] = 0;
1505 			}
1506 			rt->mfc_parent = -1;
1507 
1508 			/* clear the RP address */
1509 			rt->mfc_rp = zeroin_addr;
1510 
1511 			rt->mfc_bw_meter = NULL;
1512 
1513 			/* link into table */
1514 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1515 			/* Add this entry to the end of the queue */
1516 			rt->mfc_stall = rte;
1517 		} else {
1518 			/* determine if q has overflowed */
1519 			struct rtdetq **p;
1520 			int npkts = 0;
1521 
1522 			/*
1523 			 * XXX ouch! we need to append to the list, but we
1524 			 * only have a pointer to the front, so we have to
1525 			 * scan the entire list every time.
1526 			 */
1527 			for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1528 				if (++npkts > MAX_UPQ) {
1529 					mrtstat.mrts_upq_ovflw++;
1530 				non_fatal:
1531 					free(rte, M_MRTABLE);
1532 					m_freem(mb0);
1533 					splx(s);
1534 					return (0);
1535 				}
1536 
1537 			/* Add this entry to the end of the queue */
1538 			*p = rte;
1539 		}
1540 
1541 		rte->next = NULL;
1542 		rte->m = mb0;
1543 		rte->ifp = ifp;
1544 #ifdef UPCALL_TIMING
1545 		rte->t = tp;
1546 #endif /* UPCALL_TIMING */
1547 
1548 		splx(s);
1549 
1550 		return (0);
1551 	}
1552 }
1553 
1554 
1555 /*ARGSUSED*/
1556 static void
1557 expire_upcalls(void *v)
1558 {
1559 	int i;
1560 	int s;
1561 
1562 	s = splsoftnet();
1563 
1564 	for (i = 0; i < MFCTBLSIZ; i++) {
1565 		struct mfc *rt, *nrt;
1566 
1567 		if (nexpire[i] == 0)
1568 			continue;
1569 
1570 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1571 			nrt = LIST_NEXT(rt, mfc_hash);
1572 
1573 			if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1574 				continue;
1575 			nexpire[i]--;
1576 
1577 			/*
1578 			 * free the bw_meter entries
1579 			 */
1580 			while (rt->mfc_bw_meter != NULL) {
1581 				struct bw_meter *x = rt->mfc_bw_meter;
1582 
1583 				rt->mfc_bw_meter = x->bm_mfc_next;
1584 				free(x, M_BWMETER);
1585 			}
1586 
1587 			++mrtstat.mrts_cache_cleanups;
1588 			if (mrtdebug & DEBUG_EXPIRE)
1589 				log(LOG_DEBUG,
1590 				    "expire_upcalls: expiring (%x %x)\n",
1591 				    ntohl(rt->mfc_origin.s_addr),
1592 				    ntohl(rt->mfc_mcastgrp.s_addr));
1593 
1594 			expire_mfc(rt);
1595 		}
1596 	}
1597 
1598 	splx(s);
1599 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1600 	    expire_upcalls, NULL);
1601 }
1602 
1603 /*
1604  * Packet forwarding routine once entry in the cache is made
1605  */
1606 static int
1607 #ifdef RSVP_ISI
1608 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1609 #else
1610 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1611 #endif /* RSVP_ISI */
1612 {
1613 	struct ip  *ip = mtod(m, struct ip *);
1614 	vifi_t vifi;
1615 	struct vif *vifp;
1616 	struct sockaddr_in sin;
1617 	int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1618 
1619 /*
1620  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1621  * input, they shouldn't get counted on output, so statistics keeping is
1622  * separate.
1623  */
1624 #define MC_SEND(ip, vifp, m) do {					\
1625 	if ((vifp)->v_flags & VIFF_TUNNEL)				\
1626 		encap_send((ip), (vifp), (m));				\
1627 	else								\
1628 		phyint_send((ip), (vifp), (m));				\
1629 } while (/*CONSTCOND*/ 0)
1630 
1631 #ifdef RSVP_ISI
1632 	/*
1633 	 * If xmt_vif is not -1, send on only the requested vif.
1634 	 *
1635 	 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1636 	 */
1637 	if (xmt_vif < numvifs) {
1638 #ifdef PIM
1639 		if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1640 			pim_register_send(ip, viftable + xmt_vif, m, rt);
1641 		else
1642 #endif
1643 		MC_SEND(ip, viftable + xmt_vif, m);
1644 		return (1);
1645 	}
1646 #endif /* RSVP_ISI */
1647 
1648 	/*
1649 	 * Don't forward if it didn't arrive from the parent vif for its origin.
1650 	 */
1651 	vifi = rt->mfc_parent;
1652 	if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1653 		/* came in the wrong interface */
1654 		if (mrtdebug & DEBUG_FORWARD)
1655 			log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1656 			    ifp, vifi,
1657 			    vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1658 		++mrtstat.mrts_wrong_if;
1659 		++rt->mfc_wrong_if;
1660 		/*
1661 		 * If we are doing PIM assert processing, send a message
1662 		 * to the routing daemon.
1663 		 *
1664 		 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1665 		 * can complete the SPT switch, regardless of the type
1666 		 * of the iif (broadcast media, GRE tunnel, etc).
1667 		 */
1668 		if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1669 			struct timeval now;
1670 			u_int32_t delta;
1671 
1672 #ifdef PIM
1673 			if (ifp == &multicast_register_if)
1674 				pimstat.pims_rcv_registers_wrongiif++;
1675 #endif
1676 
1677 			/* Get vifi for the incoming packet */
1678 			for (vifi = 0;
1679 			     vifi < numvifs && viftable[vifi].v_ifp != ifp;
1680 			     vifi++)
1681 			    ;
1682 			if (vifi >= numvifs) {
1683 				/* The iif is not found: ignore the packet. */
1684 				return (0);
1685 			}
1686 
1687 			if (rt->mfc_flags[vifi] &
1688 			    MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1689 				/* WRONGVIF disabled: ignore the packet */
1690 				return (0);
1691 			}
1692 
1693 			microtime(&now);
1694 
1695 			TV_DELTA(rt->mfc_last_assert, now, delta);
1696 
1697 			if (delta > ASSERT_MSG_TIME) {
1698 				struct igmpmsg *im;
1699 				int hlen = ip->ip_hl << 2;
1700 				struct mbuf *mm =
1701 				    m_copym(m, 0, hlen, M_DONTWAIT);
1702 
1703 				M_PULLUP(mm, hlen);
1704 				if (mm == NULL)
1705 					return (ENOBUFS);
1706 
1707 				rt->mfc_last_assert = now;
1708 
1709 				im = mtod(mm, struct igmpmsg *);
1710 				im->im_msgtype	= IGMPMSG_WRONGVIF;
1711 				im->im_mbz	= 0;
1712 				im->im_vif	= vifi;
1713 
1714 				mrtstat.mrts_upcalls++;
1715 
1716 				sockaddr_in_init(&sin, &im->im_src, 0);
1717 				if (socket_send(ip_mrouter, mm, &sin) < 0) {
1718 					log(LOG_WARNING,
1719 					    "ip_mforward: ip_mrouter socket queue full\n");
1720 					++mrtstat.mrts_upq_sockfull;
1721 					return (ENOBUFS);
1722 				}
1723 			}
1724 		}
1725 		return (0);
1726 	}
1727 
1728 	/* If I sourced this packet, it counts as output, else it was input. */
1729 	if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1730 		viftable[vifi].v_pkt_out++;
1731 		viftable[vifi].v_bytes_out += plen;
1732 	} else {
1733 		viftable[vifi].v_pkt_in++;
1734 		viftable[vifi].v_bytes_in += plen;
1735 	}
1736 	rt->mfc_pkt_cnt++;
1737 	rt->mfc_byte_cnt += plen;
1738 
1739 	/*
1740 	 * For each vif, decide if a copy of the packet should be forwarded.
1741 	 * Forward if:
1742 	 *		- the ttl exceeds the vif's threshold
1743 	 *		- there are group members downstream on interface
1744 	 */
1745 	for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1746 		if ((rt->mfc_ttls[vifi] > 0) &&
1747 			(ip->ip_ttl > rt->mfc_ttls[vifi])) {
1748 			vifp->v_pkt_out++;
1749 			vifp->v_bytes_out += plen;
1750 #ifdef PIM
1751 			if (vifp->v_flags & VIFF_REGISTER)
1752 				pim_register_send(ip, vifp, m, rt);
1753 			else
1754 #endif
1755 			MC_SEND(ip, vifp, m);
1756 		}
1757 
1758 	/*
1759 	 * Perform upcall-related bw measuring.
1760 	 */
1761 	if (rt->mfc_bw_meter != NULL) {
1762 		struct bw_meter *x;
1763 		struct timeval now;
1764 
1765 		microtime(&now);
1766 		for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1767 			bw_meter_receive_packet(x, plen, &now);
1768 	}
1769 
1770 	return (0);
1771 }
1772 
1773 #ifdef RSVP_ISI
1774 /*
1775  * check if a vif number is legal/ok. This is used by ip_output.
1776  */
1777 int
1778 legal_vif_num(int vif)
1779 {
1780 	if (vif >= 0 && vif < numvifs)
1781 		return (1);
1782 	else
1783 		return (0);
1784 }
1785 #endif /* RSVP_ISI */
1786 
1787 static void
1788 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1789 {
1790 	struct mbuf *mb_copy;
1791 	int hlen = ip->ip_hl << 2;
1792 
1793 	/*
1794 	 * Make a new reference to the packet; make sure that
1795 	 * the IP header is actually copied, not just referenced,
1796 	 * so that ip_output() only scribbles on the copy.
1797 	 */
1798 	mb_copy = m_copypacket(m, M_DONTWAIT);
1799 	M_PULLUP(mb_copy, hlen);
1800 	if (mb_copy == NULL)
1801 		return;
1802 
1803 	if (vifp->v_rate_limit <= 0)
1804 		tbf_send_packet(vifp, mb_copy);
1805 	else
1806 		tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1807 		    ntohs(ip->ip_len));
1808 }
1809 
1810 static void
1811 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1812 {
1813 	struct mbuf *mb_copy;
1814 	struct ip *ip_copy;
1815 	int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1816 
1817 	/* Take care of delayed checksums */
1818 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1819 		in_delayed_cksum(m);
1820 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1821 	}
1822 
1823 	/*
1824 	 * copy the old packet & pullup it's IP header into the
1825 	 * new mbuf so we can modify it.  Try to fill the new
1826 	 * mbuf since if we don't the ethernet driver will.
1827 	 */
1828 	MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1829 	if (mb_copy == NULL)
1830 		return;
1831 	mb_copy->m_data += max_linkhdr;
1832 	mb_copy->m_pkthdr.len = len;
1833 	mb_copy->m_len = sizeof(multicast_encap_iphdr);
1834 
1835 	if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
1836 		m_freem(mb_copy);
1837 		return;
1838 	}
1839 	i = MHLEN - max_linkhdr;
1840 	if (i > len)
1841 		i = len;
1842 	mb_copy = m_pullup(mb_copy, i);
1843 	if (mb_copy == NULL)
1844 		return;
1845 
1846 	/*
1847 	 * fill in the encapsulating IP header.
1848 	 */
1849 	ip_copy = mtod(mb_copy, struct ip *);
1850 	*ip_copy = multicast_encap_iphdr;
1851 	if (len < IP_MINFRAGSIZE)
1852 		ip_copy->ip_id = 0;
1853 	else
1854 		ip_copy->ip_id = ip_newid(NULL);
1855 	ip_copy->ip_len = htons(len);
1856 	ip_copy->ip_src = vifp->v_lcl_addr;
1857 	ip_copy->ip_dst = vifp->v_rmt_addr;
1858 
1859 	/*
1860 	 * turn the encapsulated IP header back into a valid one.
1861 	 */
1862 	ip = (struct ip *)((char *)ip_copy + sizeof(multicast_encap_iphdr));
1863 	--ip->ip_ttl;
1864 	ip->ip_sum = 0;
1865 	mb_copy->m_data += sizeof(multicast_encap_iphdr);
1866 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1867 	mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1868 
1869 	if (vifp->v_rate_limit <= 0)
1870 		tbf_send_packet(vifp, mb_copy);
1871 	else
1872 		tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1873 }
1874 
1875 /*
1876  * De-encapsulate a packet and feed it back through ip input.
1877  */
1878 static void
1879 vif_input(struct mbuf *m, ...)
1880 {
1881 	int off, proto;
1882 	va_list ap;
1883 	struct vif *vifp;
1884 	int s;
1885 	struct ifqueue *ifq;
1886 
1887 	va_start(ap, m);
1888 	off = va_arg(ap, int);
1889 	proto = va_arg(ap, int);
1890 	va_end(ap);
1891 
1892 	vifp = (struct vif *)encap_getarg(m);
1893 	if (!vifp || proto != ENCAP_PROTO) {
1894 		m_freem(m);
1895 		mrtstat.mrts_bad_tunnel++;
1896 		return;
1897 	}
1898 
1899 	m_adj(m, off);
1900 	m->m_pkthdr.rcvif = vifp->v_ifp;
1901 	ifq = &ipintrq;
1902 	s = splnet();
1903 	if (IF_QFULL(ifq)) {
1904 		IF_DROP(ifq);
1905 		m_freem(m);
1906 	} else {
1907 		IF_ENQUEUE(ifq, m);
1908 		/*
1909 		 * normally we would need a "schednetisr(NETISR_IP)"
1910 		 * here but we were called by ip_input and it is going
1911 		 * to loop back & try to dequeue the packet we just
1912 		 * queued as soon as we return so we avoid the
1913 		 * unnecessary software interrrupt.
1914 		 */
1915 	}
1916 	splx(s);
1917 }
1918 
1919 /*
1920  * Check if the packet should be received on the vif denoted by arg.
1921  * (The encap selection code will call this once per vif since each is
1922  * registered separately.)
1923  */
1924 static int
1925 vif_encapcheck(struct mbuf *m, int off, int proto, void *arg)
1926 {
1927 	struct vif *vifp;
1928 	struct ip ip;
1929 
1930 #ifdef DIAGNOSTIC
1931 	if (!arg || proto != IPPROTO_IPV4)
1932 		panic("unexpected arg in vif_encapcheck");
1933 #endif
1934 
1935 	/*
1936 	 * Accept the packet only if the inner heaader is multicast
1937 	 * and the outer header matches a tunnel-mode vif.  Order
1938 	 * checks in the hope that common non-matching packets will be
1939 	 * rejected quickly.  Assume that unicast IPv4 traffic in a
1940 	 * parallel tunnel (e.g. gif(4)) is unlikely.
1941 	 */
1942 
1943 	/* Obtain the outer IP header and the vif pointer. */
1944 	m_copydata((struct mbuf *)m, 0, sizeof(ip), (void *)&ip);
1945 	vifp = (struct vif *)arg;
1946 
1947 	/*
1948 	 * The outer source must match the vif's remote peer address.
1949 	 * For a multicast router with several tunnels, this is the
1950 	 * only check that will fail on packets in other tunnels,
1951 	 * assuming the local address is the same.
1952 	 */
1953 	if (!in_hosteq(vifp->v_rmt_addr, ip.ip_src))
1954 		return 0;
1955 
1956 	/* The outer destination must match the vif's local address. */
1957 	if (!in_hosteq(vifp->v_lcl_addr, ip.ip_dst))
1958 		return 0;
1959 
1960 	/* The vif must be of tunnel type. */
1961 	if ((vifp->v_flags & VIFF_TUNNEL) == 0)
1962 		return 0;
1963 
1964 	/* Check that the inner destination is multicast. */
1965 	m_copydata((struct mbuf *)m, off, sizeof(ip), (void *)&ip);
1966 	if (!IN_MULTICAST(ip.ip_dst.s_addr))
1967 		return 0;
1968 
1969 	/*
1970 	 * We have checked that both the outer src and dst addresses
1971 	 * match the vif, and that the inner destination is multicast
1972 	 * (224/5).  By claiming more than 64, we intend to
1973 	 * preferentially take packets that also match a parallel
1974 	 * gif(4).
1975 	 */
1976 	return 32 + 32 + 5;
1977 }
1978 
1979 /*
1980  * Token bucket filter module
1981  */
1982 static void
1983 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
1984 {
1985 
1986 	if (len > MAX_BKT_SIZE) {
1987 		/* drop if packet is too large */
1988 		mrtstat.mrts_pkt2large++;
1989 		m_freem(m);
1990 		return;
1991 	}
1992 
1993 	tbf_update_tokens(vifp);
1994 
1995 	/*
1996 	 * If there are enough tokens, and the queue is empty, send this packet
1997 	 * out immediately.  Otherwise, try to insert it on this vif's queue.
1998 	 */
1999 	if (vifp->tbf_q_len == 0) {
2000 		if (len <= vifp->tbf_n_tok) {
2001 			vifp->tbf_n_tok -= len;
2002 			tbf_send_packet(vifp, m);
2003 		} else {
2004 			/* queue packet and timeout till later */
2005 			tbf_queue(vifp, m);
2006 			callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2007 			    tbf_reprocess_q, vifp);
2008 		}
2009 	} else {
2010 		if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
2011 		    !tbf_dq_sel(vifp, ip)) {
2012 			/* queue full, and couldn't make room */
2013 			mrtstat.mrts_q_overflow++;
2014 			m_freem(m);
2015 		} else {
2016 			/* queue length low enough, or made room */
2017 			tbf_queue(vifp, m);
2018 			tbf_process_q(vifp);
2019 		}
2020 	}
2021 }
2022 
2023 /*
2024  * adds a packet to the queue at the interface
2025  */
2026 static void
2027 tbf_queue(struct vif *vifp, struct mbuf *m)
2028 {
2029 	int s = splsoftnet();
2030 
2031 	/* insert at tail */
2032 	*vifp->tbf_t = m;
2033 	vifp->tbf_t = &m->m_nextpkt;
2034 	vifp->tbf_q_len++;
2035 
2036 	splx(s);
2037 }
2038 
2039 
2040 /*
2041  * processes the queue at the interface
2042  */
2043 static void
2044 tbf_process_q(struct vif *vifp)
2045 {
2046 	struct mbuf *m;
2047 	int len;
2048 	int s = splsoftnet();
2049 
2050 	/*
2051 	 * Loop through the queue at the interface and send as many packets
2052 	 * as possible.
2053 	 */
2054 	for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
2055 		len = ntohs(mtod(m, struct ip *)->ip_len);
2056 
2057 		/* determine if the packet can be sent */
2058 		if (len <= vifp->tbf_n_tok) {
2059 			/* if so,
2060 			 * reduce no of tokens, dequeue the packet,
2061 			 * send the packet.
2062 			 */
2063 			if ((vifp->tbf_q = m->m_nextpkt) == NULL)
2064 				vifp->tbf_t = &vifp->tbf_q;
2065 			--vifp->tbf_q_len;
2066 
2067 			m->m_nextpkt = NULL;
2068 			vifp->tbf_n_tok -= len;
2069 			tbf_send_packet(vifp, m);
2070 		} else
2071 			break;
2072 	}
2073 	splx(s);
2074 }
2075 
2076 static void
2077 tbf_reprocess_q(void *arg)
2078 {
2079 	struct vif *vifp = arg;
2080 
2081 	if (ip_mrouter == NULL)
2082 		return;
2083 
2084 	tbf_update_tokens(vifp);
2085 	tbf_process_q(vifp);
2086 
2087 	if (vifp->tbf_q_len != 0)
2088 		callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2089 		    tbf_reprocess_q, vifp);
2090 }
2091 
2092 /* function that will selectively discard a member of the queue
2093  * based on the precedence value and the priority
2094  */
2095 static int
2096 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2097 {
2098 	u_int p;
2099 	struct mbuf **mp, *m;
2100 	int s = splsoftnet();
2101 
2102 	p = priority(vifp, ip);
2103 
2104 	for (mp = &vifp->tbf_q, m = *mp;
2105 	    m != NULL;
2106 	    mp = &m->m_nextpkt, m = *mp) {
2107 		if (p > priority(vifp, mtod(m, struct ip *))) {
2108 			if ((*mp = m->m_nextpkt) == NULL)
2109 				vifp->tbf_t = mp;
2110 			--vifp->tbf_q_len;
2111 
2112 			m_freem(m);
2113 			mrtstat.mrts_drop_sel++;
2114 			splx(s);
2115 			return (1);
2116 		}
2117 	}
2118 	splx(s);
2119 	return (0);
2120 }
2121 
2122 static void
2123 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2124 {
2125 	int error;
2126 	int s = splsoftnet();
2127 
2128 	if (vifp->v_flags & VIFF_TUNNEL) {
2129 		/* If tunnel options */
2130 		ip_output(m, (struct mbuf *)NULL, &vifp->v_route,
2131 		    IP_FORWARDING, (struct ip_moptions *)NULL,
2132 		    (struct socket *)NULL);
2133 	} else {
2134 		/* if physical interface option, extract the options and then send */
2135 		struct ip_moptions imo;
2136 
2137 		imo.imo_multicast_ifp = vifp->v_ifp;
2138 		imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2139 		imo.imo_multicast_loop = 1;
2140 #ifdef RSVP_ISI
2141 		imo.imo_multicast_vif = -1;
2142 #endif
2143 
2144 		error = ip_output(m, NULL, NULL, IP_FORWARDING|IP_MULTICASTOPTS,
2145 		    &imo, NULL);
2146 
2147 		if (mrtdebug & DEBUG_XMIT)
2148 			log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2149 			    (long)(vifp - viftable), error);
2150 	}
2151 	splx(s);
2152 }
2153 
2154 /* determine the current time and then
2155  * the elapsed time (between the last time and time now)
2156  * in milliseconds & update the no. of tokens in the bucket
2157  */
2158 static void
2159 tbf_update_tokens(struct vif *vifp)
2160 {
2161 	struct timeval tp;
2162 	u_int32_t tm;
2163 	int s = splsoftnet();
2164 
2165 	microtime(&tp);
2166 
2167 	TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2168 
2169 	/*
2170 	 * This formula is actually
2171 	 * "time in seconds" * "bytes/second".
2172 	 *
2173 	 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2174 	 *
2175 	 * The (1000/1024) was introduced in add_vif to optimize
2176 	 * this divide into a shift.
2177 	 */
2178 	vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2179 	vifp->tbf_last_pkt_t = tp;
2180 
2181 	if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2182 		vifp->tbf_n_tok = MAX_BKT_SIZE;
2183 
2184 	splx(s);
2185 }
2186 
2187 static int
2188 priority(struct vif *vifp, struct ip *ip)
2189 {
2190 	int prio = 50;	/* the lowest priority -- default case */
2191 
2192 	/* temporary hack; may add general packet classifier some day */
2193 
2194 	/*
2195 	 * The UDP port space is divided up into four priority ranges:
2196 	 * [0, 16384)     : unclassified - lowest priority
2197 	 * [16384, 32768) : audio - highest priority
2198 	 * [32768, 49152) : whiteboard - medium priority
2199 	 * [49152, 65536) : video - low priority
2200 	 */
2201 	if (ip->ip_p == IPPROTO_UDP) {
2202 		struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2203 
2204 		switch (ntohs(udp->uh_dport) & 0xc000) {
2205 		case 0x4000:
2206 			prio = 70;
2207 			break;
2208 		case 0x8000:
2209 			prio = 60;
2210 			break;
2211 		case 0xc000:
2212 			prio = 55;
2213 			break;
2214 		}
2215 
2216 		if (tbfdebug > 1)
2217 			log(LOG_DEBUG, "port %x prio %d\n",
2218 			    ntohs(udp->uh_dport), prio);
2219 	}
2220 
2221 	return (prio);
2222 }
2223 
2224 /*
2225  * End of token bucket filter modifications
2226  */
2227 #ifdef RSVP_ISI
2228 int
2229 ip_rsvp_vif_init(struct socket *so, struct mbuf *m)
2230 {
2231 	int vifi, s;
2232 
2233 	if (rsvpdebug)
2234 		printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2235 		    so->so_type, so->so_proto->pr_protocol);
2236 
2237 	if (so->so_type != SOCK_RAW ||
2238 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2239 		return (EOPNOTSUPP);
2240 
2241 	/* Check mbuf. */
2242 	if (m == NULL || m->m_len != sizeof(int)) {
2243 		return (EINVAL);
2244 	}
2245 	vifi = *(mtod(m, int *));
2246 
2247 	if (rsvpdebug)
2248 		printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",
2249 		       vifi, rsvp_on);
2250 
2251 	s = splsoftnet();
2252 
2253 	/* Check vif. */
2254 	if (!legal_vif_num(vifi)) {
2255 		splx(s);
2256 		return (EADDRNOTAVAIL);
2257 	}
2258 
2259 	/* Check if socket is available. */
2260 	if (viftable[vifi].v_rsvpd != NULL) {
2261 		splx(s);
2262 		return (EADDRINUSE);
2263 	}
2264 
2265 	viftable[vifi].v_rsvpd = so;
2266 	/*
2267 	 * This may seem silly, but we need to be sure we don't over-increment
2268 	 * the RSVP counter, in case something slips up.
2269 	 */
2270 	if (!viftable[vifi].v_rsvp_on) {
2271 		viftable[vifi].v_rsvp_on = 1;
2272 		rsvp_on++;
2273 	}
2274 
2275 	splx(s);
2276 	return (0);
2277 }
2278 
2279 int
2280 ip_rsvp_vif_done(struct socket *so, struct mbuf *m)
2281 {
2282 	int vifi, s;
2283 
2284 	if (rsvpdebug)
2285 		printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2286 		    so->so_type, so->so_proto->pr_protocol);
2287 
2288 	if (so->so_type != SOCK_RAW ||
2289 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2290 		return (EOPNOTSUPP);
2291 
2292 	/* Check mbuf. */
2293 	if (m == NULL || m->m_len != sizeof(int)) {
2294 		return (EINVAL);
2295 	}
2296 	vifi = *(mtod(m, int *));
2297 
2298 	s = splsoftnet();
2299 
2300 	/* Check vif. */
2301 	if (!legal_vif_num(vifi)) {
2302 		splx(s);
2303 		return (EADDRNOTAVAIL);
2304 	}
2305 
2306 	if (rsvpdebug)
2307 		printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2308 		    viftable[vifi].v_rsvpd, so);
2309 
2310 	viftable[vifi].v_rsvpd = NULL;
2311 	/*
2312 	 * This may seem silly, but we need to be sure we don't over-decrement
2313 	 * the RSVP counter, in case something slips up.
2314 	 */
2315 	if (viftable[vifi].v_rsvp_on) {
2316 		viftable[vifi].v_rsvp_on = 0;
2317 		rsvp_on--;
2318 	}
2319 
2320 	splx(s);
2321 	return (0);
2322 }
2323 
2324 void
2325 ip_rsvp_force_done(struct socket *so)
2326 {
2327 	int vifi, s;
2328 
2329 	/* Don't bother if it is not the right type of socket. */
2330 	if (so->so_type != SOCK_RAW ||
2331 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2332 		return;
2333 
2334 	s = splsoftnet();
2335 
2336 	/*
2337 	 * The socket may be attached to more than one vif...this
2338 	 * is perfectly legal.
2339 	 */
2340 	for (vifi = 0; vifi < numvifs; vifi++) {
2341 		if (viftable[vifi].v_rsvpd == so) {
2342 			viftable[vifi].v_rsvpd = NULL;
2343 			/*
2344 			 * This may seem silly, but we need to be sure we don't
2345 			 * over-decrement the RSVP counter, in case something
2346 			 * slips up.
2347 			 */
2348 			if (viftable[vifi].v_rsvp_on) {
2349 				viftable[vifi].v_rsvp_on = 0;
2350 				rsvp_on--;
2351 			}
2352 		}
2353 	}
2354 
2355 	splx(s);
2356 	return;
2357 }
2358 
2359 void
2360 rsvp_input(struct mbuf *m, struct ifnet *ifp)
2361 {
2362 	int vifi, s;
2363 	struct ip *ip = mtod(m, struct ip *);
2364 	struct sockaddr_in rsvp_src;
2365 
2366 	if (rsvpdebug)
2367 		printf("rsvp_input: rsvp_on %d\n", rsvp_on);
2368 
2369 	/*
2370 	 * Can still get packets with rsvp_on = 0 if there is a local member
2371 	 * of the group to which the RSVP packet is addressed.  But in this
2372 	 * case we want to throw the packet away.
2373 	 */
2374 	if (!rsvp_on) {
2375 		m_freem(m);
2376 		return;
2377 	}
2378 
2379 	/*
2380 	 * If the old-style non-vif-associated socket is set, then use
2381 	 * it and ignore the new ones.
2382 	 */
2383 	if (ip_rsvpd != NULL) {
2384 		if (rsvpdebug)
2385 			printf("rsvp_input: "
2386 			    "Sending packet up old-style socket\n");
2387 		rip_input(m);	/*XXX*/
2388 		return;
2389 	}
2390 
2391 	s = splsoftnet();
2392 
2393 	if (rsvpdebug)
2394 		printf("rsvp_input: check vifs\n");
2395 
2396 	/* Find which vif the packet arrived on. */
2397 	for (vifi = 0; vifi < numvifs; vifi++) {
2398 		if (viftable[vifi].v_ifp == ifp)
2399 			break;
2400 	}
2401 
2402 	if (vifi == numvifs) {
2403 		/* Can't find vif packet arrived on. Drop packet. */
2404 		if (rsvpdebug)
2405 			printf("rsvp_input: "
2406 			    "Can't find vif for packet...dropping it.\n");
2407 		m_freem(m);
2408 		splx(s);
2409 		return;
2410 	}
2411 
2412 	if (rsvpdebug)
2413 		printf("rsvp_input: check socket\n");
2414 
2415 	if (viftable[vifi].v_rsvpd == NULL) {
2416 		/*
2417 		 * drop packet, since there is no specific socket for this
2418 		 * interface
2419 		 */
2420 		if (rsvpdebug)
2421 			printf("rsvp_input: No socket defined for vif %d\n",
2422 			    vifi);
2423 		m_freem(m);
2424 		splx(s);
2425 		return;
2426 	}
2427 
2428 	sockaddr_in_init(&rsvp_src, &ip->ip_src, 0);
2429 
2430 	if (rsvpdebug && m)
2431 		printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
2432 		    m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv));
2433 
2434 	if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2435 		if (rsvpdebug)
2436 			printf("rsvp_input: Failed to append to socket\n");
2437 	else
2438 		if (rsvpdebug)
2439 			printf("rsvp_input: send packet up\n");
2440 
2441 	splx(s);
2442 }
2443 #endif /* RSVP_ISI */
2444 
2445 /*
2446  * Code for bandwidth monitors
2447  */
2448 
2449 /*
2450  * Define common interface for timeval-related methods
2451  */
2452 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2453 #define	BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2454 #define	BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2455 
2456 static uint32_t
2457 compute_bw_meter_flags(struct bw_upcall *req)
2458 {
2459     uint32_t flags = 0;
2460 
2461     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2462 	flags |= BW_METER_UNIT_PACKETS;
2463     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2464 	flags |= BW_METER_UNIT_BYTES;
2465     if (req->bu_flags & BW_UPCALL_GEQ)
2466 	flags |= BW_METER_GEQ;
2467     if (req->bu_flags & BW_UPCALL_LEQ)
2468 	flags |= BW_METER_LEQ;
2469 
2470     return flags;
2471 }
2472 
2473 /*
2474  * Add a bw_meter entry
2475  */
2476 static int
2477 add_bw_upcall(struct bw_upcall *req)
2478 {
2479     int s;
2480     struct mfc *mfc;
2481     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2482 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2483     struct timeval now;
2484     struct bw_meter *x;
2485     uint32_t flags;
2486 
2487     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2488 	return EOPNOTSUPP;
2489 
2490     /* Test if the flags are valid */
2491     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2492 	return EINVAL;
2493     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2494 	return EINVAL;
2495     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2496 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2497 	return EINVAL;
2498 
2499     /* Test if the threshold time interval is valid */
2500     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2501 	return EINVAL;
2502 
2503     flags = compute_bw_meter_flags(req);
2504 
2505     /*
2506      * Find if we have already same bw_meter entry
2507      */
2508     s = splsoftnet();
2509     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2510     if (mfc == NULL) {
2511 	splx(s);
2512 	return EADDRNOTAVAIL;
2513     }
2514     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2515 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2516 			   &req->bu_threshold.b_time, ==)) &&
2517 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2518 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2519 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
2520 	    splx(s);
2521 	    return 0;		/* XXX Already installed */
2522 	}
2523     }
2524 
2525     /* Allocate the new bw_meter entry */
2526     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2527     if (x == NULL) {
2528 	splx(s);
2529 	return ENOBUFS;
2530     }
2531 
2532     /* Set the new bw_meter entry */
2533     x->bm_threshold.b_time = req->bu_threshold.b_time;
2534     microtime(&now);
2535     x->bm_start_time = now;
2536     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2537     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2538     x->bm_measured.b_packets = 0;
2539     x->bm_measured.b_bytes = 0;
2540     x->bm_flags = flags;
2541     x->bm_time_next = NULL;
2542     x->bm_time_hash = BW_METER_BUCKETS;
2543 
2544     /* Add the new bw_meter entry to the front of entries for this MFC */
2545     x->bm_mfc = mfc;
2546     x->bm_mfc_next = mfc->mfc_bw_meter;
2547     mfc->mfc_bw_meter = x;
2548     schedule_bw_meter(x, &now);
2549     splx(s);
2550 
2551     return 0;
2552 }
2553 
2554 static void
2555 free_bw_list(struct bw_meter *list)
2556 {
2557     while (list != NULL) {
2558 	struct bw_meter *x = list;
2559 
2560 	list = list->bm_mfc_next;
2561 	unschedule_bw_meter(x);
2562 	free(x, M_BWMETER);
2563     }
2564 }
2565 
2566 /*
2567  * Delete one or multiple bw_meter entries
2568  */
2569 static int
2570 del_bw_upcall(struct bw_upcall *req)
2571 {
2572     int s;
2573     struct mfc *mfc;
2574     struct bw_meter *x;
2575 
2576     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2577 	return EOPNOTSUPP;
2578 
2579     s = splsoftnet();
2580     /* Find the corresponding MFC entry */
2581     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2582     if (mfc == NULL) {
2583 	splx(s);
2584 	return EADDRNOTAVAIL;
2585     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2586 	/*
2587 	 * Delete all bw_meter entries for this mfc
2588 	 */
2589 	struct bw_meter *list;
2590 
2591 	list = mfc->mfc_bw_meter;
2592 	mfc->mfc_bw_meter = NULL;
2593 	free_bw_list(list);
2594 	splx(s);
2595 	return 0;
2596     } else {			/* Delete a single bw_meter entry */
2597 	struct bw_meter *prev;
2598 	uint32_t flags = 0;
2599 
2600 	flags = compute_bw_meter_flags(req);
2601 
2602 	/* Find the bw_meter entry to delete */
2603 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2604 	     prev = x, x = x->bm_mfc_next) {
2605 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2606 			       &req->bu_threshold.b_time, ==)) &&
2607 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2608 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2609 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2610 		break;
2611 	}
2612 	if (x != NULL) { /* Delete entry from the list for this MFC */
2613 	    if (prev != NULL)
2614 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2615 	    else
2616 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2617 
2618 	    unschedule_bw_meter(x);
2619 	    splx(s);
2620 	    /* Free the bw_meter entry */
2621 	    free(x, M_BWMETER);
2622 	    return 0;
2623 	} else {
2624 	    splx(s);
2625 	    return EINVAL;
2626 	}
2627     }
2628     /* NOTREACHED */
2629 }
2630 
2631 /*
2632  * Perform bandwidth measurement processing that may result in an upcall
2633  */
2634 static void
2635 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2636 {
2637     struct timeval delta;
2638 
2639     delta = *nowp;
2640     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2641 
2642     if (x->bm_flags & BW_METER_GEQ) {
2643 	/*
2644 	 * Processing for ">=" type of bw_meter entry
2645 	 */
2646 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2647 	    /* Reset the bw_meter entry */
2648 	    x->bm_start_time = *nowp;
2649 	    x->bm_measured.b_packets = 0;
2650 	    x->bm_measured.b_bytes = 0;
2651 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2652 	}
2653 
2654 	/* Record that a packet is received */
2655 	x->bm_measured.b_packets++;
2656 	x->bm_measured.b_bytes += plen;
2657 
2658 	/*
2659 	 * Test if we should deliver an upcall
2660 	 */
2661 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2662 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2663 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2664 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2665 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2666 		/* Prepare an upcall for delivery */
2667 		bw_meter_prepare_upcall(x, nowp);
2668 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2669 	    }
2670 	}
2671     } else if (x->bm_flags & BW_METER_LEQ) {
2672 	/*
2673 	 * Processing for "<=" type of bw_meter entry
2674 	 */
2675 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2676 	    /*
2677 	     * We are behind time with the multicast forwarding table
2678 	     * scanning for "<=" type of bw_meter entries, so test now
2679 	     * if we should deliver an upcall.
2680 	     */
2681 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2682 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2683 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2684 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2685 		/* Prepare an upcall for delivery */
2686 		bw_meter_prepare_upcall(x, nowp);
2687 	    }
2688 	    /* Reschedule the bw_meter entry */
2689 	    unschedule_bw_meter(x);
2690 	    schedule_bw_meter(x, nowp);
2691 	}
2692 
2693 	/* Record that a packet is received */
2694 	x->bm_measured.b_packets++;
2695 	x->bm_measured.b_bytes += plen;
2696 
2697 	/*
2698 	 * Test if we should restart the measuring interval
2699 	 */
2700 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2701 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2702 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
2703 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2704 	    /* Don't restart the measuring interval */
2705 	} else {
2706 	    /* Do restart the measuring interval */
2707 	    /*
2708 	     * XXX: note that we don't unschedule and schedule, because this
2709 	     * might be too much overhead per packet. Instead, when we process
2710 	     * all entries for a given timer hash bin, we check whether it is
2711 	     * really a timeout. If not, we reschedule at that time.
2712 	     */
2713 	    x->bm_start_time = *nowp;
2714 	    x->bm_measured.b_packets = 0;
2715 	    x->bm_measured.b_bytes = 0;
2716 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2717 	}
2718     }
2719 }
2720 
2721 /*
2722  * Prepare a bandwidth-related upcall
2723  */
2724 static void
2725 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2726 {
2727     struct timeval delta;
2728     struct bw_upcall *u;
2729 
2730     /*
2731      * Compute the measured time interval
2732      */
2733     delta = *nowp;
2734     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2735 
2736     /*
2737      * If there are too many pending upcalls, deliver them now
2738      */
2739     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2740 	bw_upcalls_send();
2741 
2742     /*
2743      * Set the bw_upcall entry
2744      */
2745     u = &bw_upcalls[bw_upcalls_n++];
2746     u->bu_src = x->bm_mfc->mfc_origin;
2747     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2748     u->bu_threshold.b_time = x->bm_threshold.b_time;
2749     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2750     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2751     u->bu_measured.b_time = delta;
2752     u->bu_measured.b_packets = x->bm_measured.b_packets;
2753     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2754     u->bu_flags = 0;
2755     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2756 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2757     if (x->bm_flags & BW_METER_UNIT_BYTES)
2758 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2759     if (x->bm_flags & BW_METER_GEQ)
2760 	u->bu_flags |= BW_UPCALL_GEQ;
2761     if (x->bm_flags & BW_METER_LEQ)
2762 	u->bu_flags |= BW_UPCALL_LEQ;
2763 }
2764 
2765 /*
2766  * Send the pending bandwidth-related upcalls
2767  */
2768 static void
2769 bw_upcalls_send(void)
2770 {
2771     struct mbuf *m;
2772     int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2773     struct sockaddr_in k_igmpsrc = {
2774 	    .sin_len = sizeof(k_igmpsrc),
2775 	    .sin_family = AF_INET,
2776     };
2777     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2778 				      0,		/* unused2 */
2779 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2780 				      0,		/* im_mbz  */
2781 				      0,		/* im_vif  */
2782 				      0,		/* unused3 */
2783 				      { 0 },		/* im_src  */
2784 				      { 0 } };		/* im_dst  */
2785 
2786     if (bw_upcalls_n == 0)
2787 	return;			/* No pending upcalls */
2788 
2789     bw_upcalls_n = 0;
2790 
2791     /*
2792      * Allocate a new mbuf, initialize it with the header and
2793      * the payload for the pending calls.
2794      */
2795     MGETHDR(m, M_DONTWAIT, MT_HEADER);
2796     if (m == NULL) {
2797 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2798 	return;
2799     }
2800 
2801     m->m_len = m->m_pkthdr.len = 0;
2802     m_copyback(m, 0, sizeof(struct igmpmsg), (void *)&igmpmsg);
2803     m_copyback(m, sizeof(struct igmpmsg), len, (void *)&bw_upcalls[0]);
2804 
2805     /*
2806      * Send the upcalls
2807      * XXX do we need to set the address in k_igmpsrc ?
2808      */
2809     mrtstat.mrts_upcalls++;
2810     if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2811 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2812 	++mrtstat.mrts_upq_sockfull;
2813     }
2814 }
2815 
2816 /*
2817  * Compute the timeout hash value for the bw_meter entries
2818  */
2819 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2820     do {								\
2821 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2822 									\
2823 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2824 	(hash) = next_timeval.tv_sec;					\
2825 	if (next_timeval.tv_usec)					\
2826 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2827 	(hash) %= BW_METER_BUCKETS;					\
2828     } while (/*CONSTCOND*/ 0)
2829 
2830 /*
2831  * Schedule a timer to process periodically bw_meter entry of type "<="
2832  * by linking the entry in the proper hash bucket.
2833  */
2834 static void
2835 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2836 {
2837     int time_hash;
2838 
2839     if (!(x->bm_flags & BW_METER_LEQ))
2840 	return;		/* XXX: we schedule timers only for "<=" entries */
2841 
2842     /*
2843      * Reset the bw_meter entry
2844      */
2845     x->bm_start_time = *nowp;
2846     x->bm_measured.b_packets = 0;
2847     x->bm_measured.b_bytes = 0;
2848     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2849 
2850     /*
2851      * Compute the timeout hash value and insert the entry
2852      */
2853     BW_METER_TIMEHASH(x, time_hash);
2854     x->bm_time_next = bw_meter_timers[time_hash];
2855     bw_meter_timers[time_hash] = x;
2856     x->bm_time_hash = time_hash;
2857 }
2858 
2859 /*
2860  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2861  * by removing the entry from the proper hash bucket.
2862  */
2863 static void
2864 unschedule_bw_meter(struct bw_meter *x)
2865 {
2866     int time_hash;
2867     struct bw_meter *prev, *tmp;
2868 
2869     if (!(x->bm_flags & BW_METER_LEQ))
2870 	return;		/* XXX: we schedule timers only for "<=" entries */
2871 
2872     /*
2873      * Compute the timeout hash value and delete the entry
2874      */
2875     time_hash = x->bm_time_hash;
2876     if (time_hash >= BW_METER_BUCKETS)
2877 	return;		/* Entry was not scheduled */
2878 
2879     for (prev = NULL, tmp = bw_meter_timers[time_hash];
2880 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2881 	if (tmp == x)
2882 	    break;
2883 
2884     if (tmp == NULL)
2885 	panic("unschedule_bw_meter: bw_meter entry not found");
2886 
2887     if (prev != NULL)
2888 	prev->bm_time_next = x->bm_time_next;
2889     else
2890 	bw_meter_timers[time_hash] = x->bm_time_next;
2891 
2892     x->bm_time_next = NULL;
2893     x->bm_time_hash = BW_METER_BUCKETS;
2894 }
2895 
2896 /*
2897  * Process all "<=" type of bw_meter that should be processed now,
2898  * and for each entry prepare an upcall if necessary. Each processed
2899  * entry is rescheduled again for the (periodic) processing.
2900  *
2901  * This is run periodically (once per second normally). On each round,
2902  * all the potentially matching entries are in the hash slot that we are
2903  * looking at.
2904  */
2905 static void
2906 bw_meter_process(void)
2907 {
2908     int s;
2909     static uint32_t last_tv_sec;	/* last time we processed this */
2910 
2911     uint32_t loops;
2912     int i;
2913     struct timeval now, process_endtime;
2914 
2915     microtime(&now);
2916     if (last_tv_sec == now.tv_sec)
2917 	return;		/* nothing to do */
2918 
2919     loops = now.tv_sec - last_tv_sec;
2920     last_tv_sec = now.tv_sec;
2921     if (loops > BW_METER_BUCKETS)
2922 	loops = BW_METER_BUCKETS;
2923 
2924     s = splsoftnet();
2925     /*
2926      * Process all bins of bw_meter entries from the one after the last
2927      * processed to the current one. On entry, i points to the last bucket
2928      * visited, so we need to increment i at the beginning of the loop.
2929      */
2930     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2931 	struct bw_meter *x, *tmp_list;
2932 
2933 	if (++i >= BW_METER_BUCKETS)
2934 	    i = 0;
2935 
2936 	/* Disconnect the list of bw_meter entries from the bin */
2937 	tmp_list = bw_meter_timers[i];
2938 	bw_meter_timers[i] = NULL;
2939 
2940 	/* Process the list of bw_meter entries */
2941 	while (tmp_list != NULL) {
2942 	    x = tmp_list;
2943 	    tmp_list = tmp_list->bm_time_next;
2944 
2945 	    /* Test if the time interval is over */
2946 	    process_endtime = x->bm_start_time;
2947 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2948 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2949 		/* Not yet: reschedule, but don't reset */
2950 		int time_hash;
2951 
2952 		BW_METER_TIMEHASH(x, time_hash);
2953 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2954 		    /*
2955 		     * XXX: somehow the bin processing is a bit ahead of time.
2956 		     * Put the entry in the next bin.
2957 		     */
2958 		    if (++time_hash >= BW_METER_BUCKETS)
2959 			time_hash = 0;
2960 		}
2961 		x->bm_time_next = bw_meter_timers[time_hash];
2962 		bw_meter_timers[time_hash] = x;
2963 		x->bm_time_hash = time_hash;
2964 
2965 		continue;
2966 	    }
2967 
2968 	    /*
2969 	     * Test if we should deliver an upcall
2970 	     */
2971 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2972 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2973 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2974 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2975 		/* Prepare an upcall for delivery */
2976 		bw_meter_prepare_upcall(x, &now);
2977 	    }
2978 
2979 	    /*
2980 	     * Reschedule for next processing
2981 	     */
2982 	    schedule_bw_meter(x, &now);
2983 	}
2984     }
2985 
2986     /* Send all upcalls that are pending delivery */
2987     bw_upcalls_send();
2988 
2989     splx(s);
2990 }
2991 
2992 /*
2993  * A periodic function for sending all upcalls that are pending delivery
2994  */
2995 static void
2996 expire_bw_upcalls_send(void *unused)
2997 {
2998     int s;
2999 
3000     s = splsoftnet();
3001     bw_upcalls_send();
3002     splx(s);
3003 
3004     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
3005 		  expire_bw_upcalls_send, NULL);
3006 }
3007 
3008 /*
3009  * A periodic function for periodic scanning of the multicast forwarding
3010  * table for processing all "<=" bw_meter entries.
3011  */
3012 static void
3013 expire_bw_meter_process(void *unused)
3014 {
3015     if (mrt_api_config & MRT_MFC_BW_UPCALL)
3016 	bw_meter_process();
3017 
3018     callout_reset(&bw_meter_ch, BW_METER_PERIOD,
3019 		  expire_bw_meter_process, NULL);
3020 }
3021 
3022 /*
3023  * End of bandwidth monitoring code
3024  */
3025 
3026 #ifdef PIM
3027 /*
3028  * Send the packet up to the user daemon, or eventually do kernel encapsulation
3029  */
3030 static int
3031 pim_register_send(struct ip *ip, struct vif *vifp,
3032 	struct mbuf *m, struct mfc *rt)
3033 {
3034     struct mbuf *mb_copy, *mm;
3035 
3036     if (mrtdebug & DEBUG_PIM)
3037         log(LOG_DEBUG, "pim_register_send: ");
3038 
3039     mb_copy = pim_register_prepare(ip, m);
3040     if (mb_copy == NULL)
3041 	return ENOBUFS;
3042 
3043     /*
3044      * Send all the fragments. Note that the mbuf for each fragment
3045      * is freed by the sending machinery.
3046      */
3047     for (mm = mb_copy; mm; mm = mb_copy) {
3048 	mb_copy = mm->m_nextpkt;
3049 	mm->m_nextpkt = NULL;
3050 	mm = m_pullup(mm, sizeof(struct ip));
3051 	if (mm != NULL) {
3052 	    ip = mtod(mm, struct ip *);
3053 	    if ((mrt_api_config & MRT_MFC_RP) &&
3054 		!in_nullhost(rt->mfc_rp)) {
3055 		pim_register_send_rp(ip, vifp, mm, rt);
3056 	    } else {
3057 		pim_register_send_upcall(ip, vifp, mm, rt);
3058 	    }
3059 	}
3060     }
3061 
3062     return 0;
3063 }
3064 
3065 /*
3066  * Return a copy of the data packet that is ready for PIM Register
3067  * encapsulation.
3068  * XXX: Note that in the returned copy the IP header is a valid one.
3069  */
3070 static struct mbuf *
3071 pim_register_prepare(struct ip *ip, struct mbuf *m)
3072 {
3073     struct mbuf *mb_copy = NULL;
3074     int mtu;
3075 
3076     /* Take care of delayed checksums */
3077     if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3078 	in_delayed_cksum(m);
3079 	m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3080     }
3081 
3082     /*
3083      * Copy the old packet & pullup its IP header into the
3084      * new mbuf so we can modify it.
3085      */
3086     mb_copy = m_copypacket(m, M_DONTWAIT);
3087     if (mb_copy == NULL)
3088 	return NULL;
3089     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3090     if (mb_copy == NULL)
3091 	return NULL;
3092 
3093     /* take care of the TTL */
3094     ip = mtod(mb_copy, struct ip *);
3095     --ip->ip_ttl;
3096 
3097     /* Compute the MTU after the PIM Register encapsulation */
3098     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3099 
3100     if (ntohs(ip->ip_len) <= mtu) {
3101 	/* Turn the IP header into a valid one */
3102 	ip->ip_sum = 0;
3103 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3104     } else {
3105 	/* Fragment the packet */
3106 	if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3107 	    /* XXX: mb_copy was freed by ip_fragment() */
3108 	    return NULL;
3109 	}
3110     }
3111     return mb_copy;
3112 }
3113 
3114 /*
3115  * Send an upcall with the data packet to the user-level process.
3116  */
3117 static int
3118 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3119     struct mbuf *mb_copy, struct mfc *rt)
3120 {
3121     struct mbuf *mb_first;
3122     int len = ntohs(ip->ip_len);
3123     struct igmpmsg *im;
3124     struct sockaddr_in k_igmpsrc = {
3125 	    .sin_len = sizeof(k_igmpsrc),
3126 	    .sin_family = AF_INET,
3127     };
3128 
3129     /*
3130      * Add a new mbuf with an upcall header
3131      */
3132     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3133     if (mb_first == NULL) {
3134 	m_freem(mb_copy);
3135 	return ENOBUFS;
3136     }
3137     mb_first->m_data += max_linkhdr;
3138     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3139     mb_first->m_len = sizeof(struct igmpmsg);
3140     mb_first->m_next = mb_copy;
3141 
3142     /* Send message to routing daemon */
3143     im = mtod(mb_first, struct igmpmsg *);
3144     im->im_msgtype	= IGMPMSG_WHOLEPKT;
3145     im->im_mbz		= 0;
3146     im->im_vif		= vifp - viftable;
3147     im->im_src		= ip->ip_src;
3148     im->im_dst		= ip->ip_dst;
3149 
3150     k_igmpsrc.sin_addr	= ip->ip_src;
3151 
3152     mrtstat.mrts_upcalls++;
3153 
3154     if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3155 	if (mrtdebug & DEBUG_PIM)
3156 	    log(LOG_WARNING,
3157 		"mcast: pim_register_send_upcall: ip_mrouter socket queue full");
3158 	++mrtstat.mrts_upq_sockfull;
3159 	return ENOBUFS;
3160     }
3161 
3162     /* Keep statistics */
3163     pimstat.pims_snd_registers_msgs++;
3164     pimstat.pims_snd_registers_bytes += len;
3165 
3166     return 0;
3167 }
3168 
3169 /*
3170  * Encapsulate the data packet in PIM Register message and send it to the RP.
3171  */
3172 static int
3173 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3174 	struct mbuf *mb_copy, struct mfc *rt)
3175 {
3176     struct mbuf *mb_first;
3177     struct ip *ip_outer;
3178     struct pim_encap_pimhdr *pimhdr;
3179     int len = ntohs(ip->ip_len);
3180     vifi_t vifi = rt->mfc_parent;
3181 
3182     if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3183 	m_freem(mb_copy);
3184 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
3185     }
3186 
3187     /*
3188      * Add a new mbuf with the encapsulating header
3189      */
3190     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3191     if (mb_first == NULL) {
3192 	m_freem(mb_copy);
3193 	return ENOBUFS;
3194     }
3195     mb_first->m_data += max_linkhdr;
3196     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3197     mb_first->m_next = mb_copy;
3198 
3199     mb_first->m_pkthdr.len = len + mb_first->m_len;
3200 
3201     /*
3202      * Fill in the encapsulating IP and PIM header
3203      */
3204     ip_outer = mtod(mb_first, struct ip *);
3205     *ip_outer = pim_encap_iphdr;
3206      if (mb_first->m_pkthdr.len < IP_MINFRAGSIZE)
3207 	ip_outer->ip_id = 0;
3208     else
3209 	ip_outer->ip_id = ip_newid(NULL);
3210     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3211 			     sizeof(pim_encap_pimhdr));
3212     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3213     ip_outer->ip_dst = rt->mfc_rp;
3214     /*
3215      * Copy the inner header TOS to the outer header, and take care of the
3216      * IP_DF bit.
3217      */
3218     ip_outer->ip_tos = ip->ip_tos;
3219     if (ntohs(ip->ip_off) & IP_DF)
3220 	ip_outer->ip_off |= htons(IP_DF);
3221     pimhdr = (struct pim_encap_pimhdr *)((char *)ip_outer
3222 					 + sizeof(pim_encap_iphdr));
3223     *pimhdr = pim_encap_pimhdr;
3224     /* If the iif crosses a border, set the Border-bit */
3225     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3226 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3227 
3228     mb_first->m_data += sizeof(pim_encap_iphdr);
3229     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3230     mb_first->m_data -= sizeof(pim_encap_iphdr);
3231 
3232     if (vifp->v_rate_limit == 0)
3233 	tbf_send_packet(vifp, mb_first);
3234     else
3235 	tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3236 
3237     /* Keep statistics */
3238     pimstat.pims_snd_registers_msgs++;
3239     pimstat.pims_snd_registers_bytes += len;
3240 
3241     return 0;
3242 }
3243 
3244 /*
3245  * PIM-SMv2 and PIM-DM messages processing.
3246  * Receives and verifies the PIM control messages, and passes them
3247  * up to the listening socket, using rip_input().
3248  * The only message with special processing is the PIM_REGISTER message
3249  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3250  * is passed to if_simloop().
3251  */
3252 void
3253 pim_input(struct mbuf *m, ...)
3254 {
3255     struct ip *ip = mtod(m, struct ip *);
3256     struct pim *pim;
3257     int minlen;
3258     int datalen;
3259     int ip_tos;
3260     int proto;
3261     int iphlen;
3262     va_list ap;
3263 
3264     va_start(ap, m);
3265     iphlen = va_arg(ap, int);
3266     proto = va_arg(ap, int);
3267     va_end(ap);
3268 
3269     datalen = ntohs(ip->ip_len) - iphlen;
3270 
3271     /* Keep statistics */
3272     pimstat.pims_rcv_total_msgs++;
3273     pimstat.pims_rcv_total_bytes += datalen;
3274 
3275     /*
3276      * Validate lengths
3277      */
3278     if (datalen < PIM_MINLEN) {
3279 	pimstat.pims_rcv_tooshort++;
3280 	log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3281 	    datalen, (u_long)ip->ip_src.s_addr);
3282 	m_freem(m);
3283 	return;
3284     }
3285 
3286     /*
3287      * If the packet is at least as big as a REGISTER, go agead
3288      * and grab the PIM REGISTER header size, to avoid another
3289      * possible m_pullup() later.
3290      *
3291      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
3292      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3293      */
3294     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3295     /*
3296      * Get the IP and PIM headers in contiguous memory, and
3297      * possibly the PIM REGISTER header.
3298      */
3299     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3300 	(m = m_pullup(m, minlen)) == NULL) {
3301 	log(LOG_ERR, "pim_input: m_pullup failure\n");
3302 	return;
3303     }
3304     /* m_pullup() may have given us a new mbuf so reset ip. */
3305     ip = mtod(m, struct ip *);
3306     ip_tos = ip->ip_tos;
3307 
3308     /* adjust mbuf to point to the PIM header */
3309     m->m_data += iphlen;
3310     m->m_len  -= iphlen;
3311     pim = mtod(m, struct pim *);
3312 
3313     /*
3314      * Validate checksum. If PIM REGISTER, exclude the data packet.
3315      *
3316      * XXX: some older PIMv2 implementations don't make this distinction,
3317      * so for compatibility reason perform the checksum over part of the
3318      * message, and if error, then over the whole message.
3319      */
3320     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3321 	/* do nothing, checksum okay */
3322     } else if (in_cksum(m, datalen)) {
3323 	pimstat.pims_rcv_badsum++;
3324 	if (mrtdebug & DEBUG_PIM)
3325 	    log(LOG_DEBUG, "pim_input: invalid checksum");
3326 	m_freem(m);
3327 	return;
3328     }
3329 
3330     /* PIM version check */
3331     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3332 	pimstat.pims_rcv_badversion++;
3333 	log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3334 	    PIM_VT_V(pim->pim_vt), PIM_VERSION);
3335 	m_freem(m);
3336 	return;
3337     }
3338 
3339     /* restore mbuf back to the outer IP */
3340     m->m_data -= iphlen;
3341     m->m_len  += iphlen;
3342 
3343     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3344 	/*
3345 	 * Since this is a REGISTER, we'll make a copy of the register
3346 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3347 	 * routing daemon.
3348 	 */
3349 	int s;
3350 	struct sockaddr_in dst = {
3351 		.sin_len = sizeof(dst),
3352 		.sin_family = AF_INET,
3353 	};
3354 	struct mbuf *mcp;
3355 	struct ip *encap_ip;
3356 	u_int32_t *reghdr;
3357 	struct ifnet *vifp;
3358 
3359 	s = splsoftnet();
3360 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3361 	    splx(s);
3362 	    if (mrtdebug & DEBUG_PIM)
3363 		log(LOG_DEBUG,
3364 		    "pim_input: register vif not set: %d\n", reg_vif_num);
3365 	    m_freem(m);
3366 	    return;
3367 	}
3368 	/* XXX need refcnt? */
3369 	vifp = viftable[reg_vif_num].v_ifp;
3370 	splx(s);
3371 
3372 	/*
3373 	 * Validate length
3374 	 */
3375 	if (datalen < PIM_REG_MINLEN) {
3376 	    pimstat.pims_rcv_tooshort++;
3377 	    pimstat.pims_rcv_badregisters++;
3378 	    log(LOG_ERR,
3379 		"pim_input: register packet size too small %d from %lx\n",
3380 		datalen, (u_long)ip->ip_src.s_addr);
3381 	    m_freem(m);
3382 	    return;
3383 	}
3384 
3385 	reghdr = (u_int32_t *)(pim + 1);
3386 	encap_ip = (struct ip *)(reghdr + 1);
3387 
3388 	if (mrtdebug & DEBUG_PIM) {
3389 	    log(LOG_DEBUG,
3390 		"pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3391 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3392 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3393 		ntohs(encap_ip->ip_len));
3394 	}
3395 
3396 	/* verify the version number of the inner packet */
3397 	if (encap_ip->ip_v != IPVERSION) {
3398 	    pimstat.pims_rcv_badregisters++;
3399 	    if (mrtdebug & DEBUG_PIM) {
3400 		log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3401 		    "of the inner packet\n", encap_ip->ip_v);
3402 	    }
3403 	    m_freem(m);
3404 	    return;
3405 	}
3406 
3407 	/* verify the inner packet is destined to a mcast group */
3408 	if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3409 	    pimstat.pims_rcv_badregisters++;
3410 	    if (mrtdebug & DEBUG_PIM)
3411 		log(LOG_DEBUG,
3412 		    "pim_input: inner packet of register is not "
3413 		    "multicast %lx\n",
3414 		    (u_long)ntohl(encap_ip->ip_dst.s_addr));
3415 	    m_freem(m);
3416 	    return;
3417 	}
3418 
3419 	/* If a NULL_REGISTER, pass it to the daemon */
3420 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3421 	    goto pim_input_to_daemon;
3422 
3423 	/*
3424 	 * Copy the TOS from the outer IP header to the inner IP header.
3425 	 */
3426 	if (encap_ip->ip_tos != ip_tos) {
3427 	    /* Outer TOS -> inner TOS */
3428 	    encap_ip->ip_tos = ip_tos;
3429 	    /* Recompute the inner header checksum. Sigh... */
3430 
3431 	    /* adjust mbuf to point to the inner IP header */
3432 	    m->m_data += (iphlen + PIM_MINLEN);
3433 	    m->m_len  -= (iphlen + PIM_MINLEN);
3434 
3435 	    encap_ip->ip_sum = 0;
3436 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3437 
3438 	    /* restore mbuf to point back to the outer IP header */
3439 	    m->m_data -= (iphlen + PIM_MINLEN);
3440 	    m->m_len  += (iphlen + PIM_MINLEN);
3441 	}
3442 
3443 	/*
3444 	 * Decapsulate the inner IP packet and loopback to forward it
3445 	 * as a normal multicast packet. Also, make a copy of the
3446 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
3447 	 * to pass to the daemon later, so it can take the appropriate
3448 	 * actions (e.g., send back PIM_REGISTER_STOP).
3449 	 * XXX: here m->m_data points to the outer IP header.
3450 	 */
3451 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_DONTWAIT);
3452 	if (mcp == NULL) {
3453 	    log(LOG_ERR,
3454 		"pim_input: pim register: could not copy register head\n");
3455 	    m_freem(m);
3456 	    return;
3457 	}
3458 
3459 	/* Keep statistics */
3460 	/* XXX: registers_bytes include only the encap. mcast pkt */
3461 	pimstat.pims_rcv_registers_msgs++;
3462 	pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3463 
3464 	/*
3465 	 * forward the inner ip packet; point m_data at the inner ip.
3466 	 */
3467 	m_adj(m, iphlen + PIM_MINLEN);
3468 
3469 	if (mrtdebug & DEBUG_PIM) {
3470 	    log(LOG_DEBUG,
3471 		"pim_input: forwarding decapsulated register: "
3472 		"src %lx, dst %lx, vif %d\n",
3473 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3474 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3475 		reg_vif_num);
3476 	}
3477 	/* NB: vifp was collected above; can it change on us? */
3478 	looutput(vifp, m, (struct sockaddr *)&dst, (struct rtentry *)NULL);
3479 
3480 	/* prepare the register head to send to the mrouting daemon */
3481 	m = mcp;
3482     }
3483 
3484 pim_input_to_daemon:
3485     /*
3486      * Pass the PIM message up to the daemon; if it is a Register message,
3487      * pass the 'head' only up to the daemon. This includes the
3488      * outer IP header, PIM header, PIM-Register header and the
3489      * inner IP header.
3490      * XXX: the outer IP header pkt size of a Register is not adjust to
3491      * reflect the fact that the inner multicast data is truncated.
3492      */
3493     rip_input(m, iphlen, proto);
3494 
3495     return;
3496 }
3497 #endif /* PIM */
3498