xref: /netbsd-src/sys/netinet/ip_mroute.c (revision a5847cc334d9a7029f6352b847e9e8d71a0f9e0c)
1 /*	$NetBSD: ip_mroute.c,v 1.121 2011/10/19 01:52:22 dyoung 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 1.121 2011/10/19 01:52:22 dyoung 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 	int error, s;
791 	struct sockaddr_in sin;
792 
793 	if (vifcp->vifc_vifi >= MAXVIFS)
794 		return (EINVAL);
795 	if (in_nullhost(vifcp->vifc_lcl_addr))
796 		return (EADDRNOTAVAIL);
797 
798 	vifp = &viftable[vifcp->vifc_vifi];
799 	if (!in_nullhost(vifp->v_lcl_addr))
800 		return (EADDRINUSE);
801 
802 	/* Find the interface with an address in AF_INET family. */
803 #ifdef PIM
804 	if (vifcp->vifc_flags & VIFF_REGISTER) {
805 		/*
806 		 * XXX: Because VIFF_REGISTER does not really need a valid
807 		 * local interface (e.g. it could be 127.0.0.2), we don't
808 		 * check its address.
809 		 */
810 	    ifp = NULL;
811 	} else
812 #endif
813 	{
814 		sockaddr_in_init(&sin, &vifcp->vifc_lcl_addr, 0);
815 		ifa = ifa_ifwithaddr(sintosa(&sin));
816 		if (ifa == NULL)
817 			return (EADDRNOTAVAIL);
818 		ifp = ifa->ifa_ifp;
819 	}
820 
821 	if (vifcp->vifc_flags & VIFF_TUNNEL) {
822 		if (vifcp->vifc_flags & VIFF_SRCRT) {
823 			log(LOG_ERR, "source routed tunnels not supported\n");
824 			return (EOPNOTSUPP);
825 		}
826 
827 		/* attach this vif to decapsulator dispatch table */
828 		/*
829 		 * XXX Use addresses in registration so that matching
830 		 * can be done with radix tree in decapsulator.  But,
831 		 * we need to check inner header for multicast, so
832 		 * this requires both radix tree lookup and then a
833 		 * function to check, and this is not supported yet.
834 		 */
835 		vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
836 		    vif_encapcheck, &vif_protosw, vifp);
837 		if (!vifp->v_encap_cookie)
838 			return (EINVAL);
839 
840 		/* Create a fake encapsulation interface. */
841 		ifp = malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK|M_ZERO);
842 		snprintf(ifp->if_xname, sizeof(ifp->if_xname),
843 			 "mdecap%d", vifcp->vifc_vifi);
844 
845 		/* Prepare cached route entry. */
846 		memset(&vifp->v_route, 0, sizeof(vifp->v_route));
847 #ifdef PIM
848 	} else if (vifcp->vifc_flags & VIFF_REGISTER) {
849 		ifp = &multicast_register_if;
850 		if (mrtdebug)
851 			log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
852 			    (void *)ifp);
853 		if (reg_vif_num == VIFI_INVALID) {
854 			memset(ifp, 0, sizeof(*ifp));
855 			snprintf(ifp->if_xname, sizeof(ifp->if_xname),
856 				 "register_vif");
857 			ifp->if_flags = IFF_LOOPBACK;
858 			memset(&vifp->v_route, 0, sizeof(vifp->v_route));
859 			reg_vif_num = vifcp->vifc_vifi;
860 		}
861 #endif
862 	} else {
863 		/* Make sure the interface supports multicast. */
864 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
865 			return (EOPNOTSUPP);
866 
867 		/* Enable promiscuous reception of all IP multicasts. */
868 		sockaddr_in_init(&sin, &zeroin_addr, 0);
869 		error = if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin));
870 		if (error)
871 			return (error);
872 	}
873 
874 	s = splsoftnet();
875 
876 	/* Define parameters for the tbf structure. */
877 	vifp->tbf_q = NULL;
878 	vifp->tbf_t = &vifp->tbf_q;
879 	microtime(&vifp->tbf_last_pkt_t);
880 	vifp->tbf_n_tok = 0;
881 	vifp->tbf_q_len = 0;
882 	vifp->tbf_max_q_len = MAXQSIZE;
883 
884 	vifp->v_flags = vifcp->vifc_flags;
885 	vifp->v_threshold = vifcp->vifc_threshold;
886 	/* scaling up here allows division by 1024 in critical code */
887 	vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
888 	vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
889 	vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
890 	vifp->v_ifp = ifp;
891 	/* Initialize per vif pkt counters. */
892 	vifp->v_pkt_in = 0;
893 	vifp->v_pkt_out = 0;
894 	vifp->v_bytes_in = 0;
895 	vifp->v_bytes_out = 0;
896 
897 	callout_init(&vifp->v_repq_ch, 0);
898 
899 #ifdef RSVP_ISI
900 	vifp->v_rsvp_on = 0;
901 	vifp->v_rsvpd = NULL;
902 #endif /* RSVP_ISI */
903 
904 	splx(s);
905 
906 	/* Adjust numvifs up if the vifi is higher than numvifs. */
907 	if (numvifs <= vifcp->vifc_vifi)
908 		numvifs = vifcp->vifc_vifi + 1;
909 
910 	if (mrtdebug)
911 		log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
912 		    vifcp->vifc_vifi,
913 		    ntohl(vifcp->vifc_lcl_addr.s_addr),
914 		    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
915 		    ntohl(vifcp->vifc_rmt_addr.s_addr),
916 		    vifcp->vifc_threshold,
917 		    vifcp->vifc_rate_limit);
918 
919 	return (0);
920 }
921 
922 void
923 reset_vif(struct vif *vifp)
924 {
925 	struct mbuf *m, *n;
926 	struct ifnet *ifp;
927 	struct sockaddr_in sin;
928 
929 	callout_stop(&vifp->v_repq_ch);
930 
931 	/* detach this vif from decapsulator dispatch table */
932 	encap_detach(vifp->v_encap_cookie);
933 	vifp->v_encap_cookie = NULL;
934 
935 	/*
936 	 * Free packets queued at the interface
937 	 */
938 	for (m = vifp->tbf_q; m != NULL; m = n) {
939 		n = m->m_nextpkt;
940 		m_freem(m);
941 	}
942 
943 	if (vifp->v_flags & VIFF_TUNNEL)
944 		free(vifp->v_ifp, M_MRTABLE);
945 	else if (vifp->v_flags & VIFF_REGISTER) {
946 #ifdef PIM
947 		reg_vif_num = VIFI_INVALID;
948 #endif
949 	} else {
950 		sockaddr_in_init(&sin, &zeroin_addr, 0);
951 		ifp = vifp->v_ifp;
952 		if_mcast_op(ifp, SIOCDELMULTI, sintosa(&sin));
953 	}
954 	memset((void *)vifp, 0, sizeof(*vifp));
955 }
956 
957 /*
958  * Delete a vif from the vif table
959  */
960 static int
961 del_vif(vifi_t *vifip)
962 {
963 	struct vif *vifp;
964 	vifi_t vifi;
965 	int s;
966 
967 	if (*vifip >= numvifs)
968 		return (EINVAL);
969 
970 	vifp = &viftable[*vifip];
971 	if (in_nullhost(vifp->v_lcl_addr))
972 		return (EADDRNOTAVAIL);
973 
974 	s = splsoftnet();
975 
976 	reset_vif(vifp);
977 
978 	/* Adjust numvifs down */
979 	for (vifi = numvifs; vifi > 0; vifi--)
980 		if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
981 			break;
982 	numvifs = vifi;
983 
984 	splx(s);
985 
986 	if (mrtdebug)
987 		log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
988 
989 	return (0);
990 }
991 
992 /*
993  * update an mfc entry without resetting counters and S,G addresses.
994  */
995 static void
996 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
997 {
998 	int i;
999 
1000 	rt->mfc_parent = mfccp->mfcc_parent;
1001 	for (i = 0; i < numvifs; i++) {
1002 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1003 		rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1004 			MRT_MFC_FLAGS_ALL;
1005 	}
1006 	/* set the RP address */
1007 	if (mrt_api_config & MRT_MFC_RP)
1008 		rt->mfc_rp = mfccp->mfcc_rp;
1009 	else
1010 		rt->mfc_rp = zeroin_addr;
1011 }
1012 
1013 /*
1014  * fully initialize an mfc entry from the parameter.
1015  */
1016 static void
1017 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1018 {
1019 	rt->mfc_origin     = mfccp->mfcc_origin;
1020 	rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1021 
1022 	update_mfc_params(rt, mfccp);
1023 
1024 	/* initialize pkt counters per src-grp */
1025 	rt->mfc_pkt_cnt    = 0;
1026 	rt->mfc_byte_cnt   = 0;
1027 	rt->mfc_wrong_if   = 0;
1028 	timerclear(&rt->mfc_last_assert);
1029 }
1030 
1031 static void
1032 expire_mfc(struct mfc *rt)
1033 {
1034 	struct rtdetq *rte, *nrte;
1035 
1036 	free_bw_list(rt->mfc_bw_meter);
1037 
1038 	for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1039 		nrte = rte->next;
1040 		m_freem(rte->m);
1041 		free(rte, M_MRTABLE);
1042 	}
1043 
1044 	LIST_REMOVE(rt, mfc_hash);
1045 	free(rt, M_MRTABLE);
1046 }
1047 
1048 /*
1049  * Add an mfc entry
1050  */
1051 static int
1052 add_mfc(struct sockopt *sopt)
1053 {
1054 	struct mfcctl2 mfcctl2;
1055 	struct mfcctl2 *mfccp;
1056 	struct mfc *rt;
1057 	u_int32_t hash = 0;
1058 	struct rtdetq *rte, *nrte;
1059 	u_short nstl;
1060 	int s;
1061 	int mfcctl_size = sizeof(struct mfcctl);
1062 	int error;
1063 
1064 	if (mrt_api_config & MRT_API_FLAGS_ALL)
1065 		mfcctl_size = sizeof(struct mfcctl2);
1066 
1067 	/*
1068 	 * select data size depending on API version.
1069 	 */
1070 	mfccp = &mfcctl2;
1071 	memset(&mfcctl2, 0, sizeof(mfcctl2));
1072 
1073 	if (mrt_api_config & MRT_API_FLAGS_ALL)
1074 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1075 	else
1076 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1077 
1078 	if (error)
1079 		return (error);
1080 
1081 	s = splsoftnet();
1082 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1083 
1084 	/* If an entry already exists, just update the fields */
1085 	if (rt) {
1086 		if (mrtdebug & DEBUG_MFC)
1087 			log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1088 			    ntohl(mfccp->mfcc_origin.s_addr),
1089 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1090 			    mfccp->mfcc_parent);
1091 
1092 		update_mfc_params(rt, mfccp);
1093 
1094 		splx(s);
1095 		return (0);
1096 	}
1097 
1098 	/*
1099 	 * Find the entry for which the upcall was made and update
1100 	 */
1101 	nstl = 0;
1102 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1103 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1104 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1105 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1106 		    rt->mfc_stall != NULL) {
1107 			if (nstl++)
1108 				log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
1109 				    "multiple kernel entries",
1110 				    ntohl(mfccp->mfcc_origin.s_addr),
1111 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1112 				    mfccp->mfcc_parent, rt->mfc_stall);
1113 
1114 			if (mrtdebug & DEBUG_MFC)
1115 				log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n",
1116 				    ntohl(mfccp->mfcc_origin.s_addr),
1117 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1118 				    mfccp->mfcc_parent, rt->mfc_stall);
1119 
1120 			rte = rt->mfc_stall;
1121 			init_mfc_params(rt, mfccp);
1122 			rt->mfc_stall = NULL;
1123 
1124 			rt->mfc_expire = 0; /* Don't clean this guy up */
1125 			nexpire[hash]--;
1126 
1127 			/* free packets Qed at the end of this entry */
1128 			for (; rte != NULL; rte = nrte) {
1129 				nrte = rte->next;
1130 				if (rte->ifp) {
1131 #ifdef RSVP_ISI
1132 					ip_mdq(rte->m, rte->ifp, rt, -1);
1133 #else
1134 					ip_mdq(rte->m, rte->ifp, rt);
1135 #endif /* RSVP_ISI */
1136 				}
1137 				m_freem(rte->m);
1138 #ifdef UPCALL_TIMING
1139 				collate(&rte->t);
1140 #endif /* UPCALL_TIMING */
1141 				free(rte, M_MRTABLE);
1142 			}
1143 		}
1144 	}
1145 
1146 	/*
1147 	 * It is possible that an entry is being inserted without an upcall
1148 	 */
1149 	if (nstl == 0) {
1150 		/*
1151 		 * No mfc; make a new one
1152 		 */
1153 		if (mrtdebug & DEBUG_MFC)
1154 			log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1155 			    ntohl(mfccp->mfcc_origin.s_addr),
1156 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1157 			    mfccp->mfcc_parent);
1158 
1159 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1160 			if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1161 			    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1162 				init_mfc_params(rt, mfccp);
1163 				if (rt->mfc_expire)
1164 					nexpire[hash]--;
1165 				rt->mfc_expire = 0;
1166 				break; /* XXX */
1167 			}
1168 		}
1169 		if (rt == NULL) {	/* no upcall, so make a new entry */
1170 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1171 						  M_NOWAIT);
1172 			if (rt == NULL) {
1173 				splx(s);
1174 				return (ENOBUFS);
1175 			}
1176 
1177 			init_mfc_params(rt, mfccp);
1178 			rt->mfc_expire	= 0;
1179 			rt->mfc_stall	= NULL;
1180 			rt->mfc_bw_meter = NULL;
1181 
1182 			/* insert new entry at head of hash chain */
1183 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1184 		}
1185 	}
1186 
1187 	splx(s);
1188 	return (0);
1189 }
1190 
1191 #ifdef UPCALL_TIMING
1192 /*
1193  * collect delay statistics on the upcalls
1194  */
1195 static void
1196 collate(struct timeval *t)
1197 {
1198 	u_int32_t d;
1199 	struct timeval tp;
1200 	u_int32_t delta;
1201 
1202 	microtime(&tp);
1203 
1204 	if (timercmp(t, &tp, <)) {
1205 		TV_DELTA(tp, *t, delta);
1206 
1207 		d = delta >> 10;
1208 		if (d > 50)
1209 			d = 50;
1210 
1211 		++upcall_data[d];
1212 	}
1213 }
1214 #endif /* UPCALL_TIMING */
1215 
1216 /*
1217  * Delete an mfc entry
1218  */
1219 static int
1220 del_mfc(struct sockopt *sopt)
1221 {
1222 	struct mfcctl2 mfcctl2;
1223 	struct mfcctl2 *mfccp;
1224 	struct mfc *rt;
1225 	int s;
1226 	int error;
1227 
1228 	/*
1229 	 * XXX: for deleting MFC entries the information in entries
1230 	 * of size "struct mfcctl" is sufficient.
1231 	 */
1232 
1233 	mfccp = &mfcctl2;
1234 	memset(&mfcctl2, 0, sizeof(mfcctl2));
1235 
1236 	error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1237 	if (error) {
1238 		/* Try with the size of mfcctl2. */
1239 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1240 		if (error)
1241 			return (error);
1242 	}
1243 
1244 	if (mrtdebug & DEBUG_MFC)
1245 		log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1246 		    ntohl(mfccp->mfcc_origin.s_addr),
1247 		    ntohl(mfccp->mfcc_mcastgrp.s_addr));
1248 
1249 	s = splsoftnet();
1250 
1251 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1252 	if (rt == NULL) {
1253 		splx(s);
1254 		return (EADDRNOTAVAIL);
1255 	}
1256 
1257 	/*
1258 	 * free the bw_meter entries
1259 	 */
1260 	free_bw_list(rt->mfc_bw_meter);
1261 	rt->mfc_bw_meter = NULL;
1262 
1263 	LIST_REMOVE(rt, mfc_hash);
1264 	free(rt, M_MRTABLE);
1265 
1266 	splx(s);
1267 	return (0);
1268 }
1269 
1270 static int
1271 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1272 {
1273 	if (s) {
1274 		if (sbappendaddr(&s->so_rcv, sintosa(src), mm, NULL) != 0) {
1275 			sorwakeup(s);
1276 			return (0);
1277 		}
1278 	}
1279 	m_freem(mm);
1280 	return (-1);
1281 }
1282 
1283 /*
1284  * IP multicast forwarding function. This function assumes that the packet
1285  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1286  * pointed to by "ifp", and the packet is to be relayed to other networks
1287  * that have members of the packet's destination IP multicast group.
1288  *
1289  * The packet is returned unscathed to the caller, unless it is
1290  * erroneous, in which case a non-zero return value tells the caller to
1291  * discard it.
1292  */
1293 
1294 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1295 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1296 
1297 int
1298 #ifdef RSVP_ISI
1299 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo)
1300 #else
1301 ip_mforward(struct mbuf *m, struct ifnet *ifp)
1302 #endif /* RSVP_ISI */
1303 {
1304 	struct ip *ip = mtod(m, struct ip *);
1305 	struct mfc *rt;
1306 	static int srctun = 0;
1307 	struct mbuf *mm;
1308 	struct sockaddr_in sin;
1309 	int s;
1310 	vifi_t vifi;
1311 
1312 	if (mrtdebug & DEBUG_FORWARD)
1313 		log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1314 		    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1315 
1316 	if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1317 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1318 		/*
1319 		 * Packet arrived via a physical interface or
1320 		 * an encapsulated tunnel or a register_vif.
1321 		 */
1322 	} else {
1323 		/*
1324 		 * Packet arrived through a source-route tunnel.
1325 		 * Source-route tunnels are no longer supported.
1326 		 */
1327 		if ((srctun++ % 1000) == 0)
1328 			log(LOG_ERR,
1329 			    "ip_mforward: received source-routed packet from %x\n",
1330 			    ntohl(ip->ip_src.s_addr));
1331 
1332 		return (1);
1333 	}
1334 
1335 	/*
1336 	 * Clear any in-bound checksum flags for this packet.
1337 	 */
1338 	m->m_pkthdr.csum_flags = 0;
1339 
1340 #ifdef RSVP_ISI
1341 	if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1342 		if (ip->ip_ttl < MAXTTL)
1343 			ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1344 		if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1345 			struct vif *vifp = viftable + vifi;
1346 			printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n",
1347 			    ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1348 			    (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1349 			    vifp->v_ifp->if_xname);
1350 		}
1351 		return (ip_mdq(m, ifp, NULL, vifi));
1352 	}
1353 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1354 		printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
1355 		    ntohl(ip->ip_src), ntohl(ip->ip_dst));
1356 	}
1357 #endif /* RSVP_ISI */
1358 
1359 	/*
1360 	 * Don't forward a packet with time-to-live of zero or one,
1361 	 * or a packet destined to a local-only group.
1362 	 */
1363 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1364 		return (0);
1365 
1366 	/*
1367 	 * Determine forwarding vifs from the forwarding cache table
1368 	 */
1369 	s = splsoftnet();
1370 	++mrtstat.mrts_mfc_lookups;
1371 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1372 
1373 	/* Entry exists, so forward if necessary */
1374 	if (rt != NULL) {
1375 		splx(s);
1376 #ifdef RSVP_ISI
1377 		return (ip_mdq(m, ifp, rt, -1));
1378 #else
1379 		return (ip_mdq(m, ifp, rt));
1380 #endif /* RSVP_ISI */
1381 	} else {
1382 		/*
1383 		 * If we don't have a route for packet's origin,
1384 		 * Make a copy of the packet & send message to routing daemon
1385 		 */
1386 
1387 		struct mbuf *mb0;
1388 		struct rtdetq *rte;
1389 		u_int32_t hash;
1390 		int hlen = ip->ip_hl << 2;
1391 #ifdef UPCALL_TIMING
1392 		struct timeval tp;
1393 
1394 		microtime(&tp);
1395 #endif /* UPCALL_TIMING */
1396 
1397 		++mrtstat.mrts_mfc_misses;
1398 
1399 		mrtstat.mrts_no_route++;
1400 		if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1401 			log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1402 			    ntohl(ip->ip_src.s_addr),
1403 			    ntohl(ip->ip_dst.s_addr));
1404 
1405 		/*
1406 		 * Allocate mbufs early so that we don't do extra work if we are
1407 		 * just going to fail anyway.  Make sure to pullup the header so
1408 		 * that other people can't step on it.
1409 		 */
1410 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1411 					      M_NOWAIT);
1412 		if (rte == NULL) {
1413 			splx(s);
1414 			return (ENOBUFS);
1415 		}
1416 		mb0 = m_copypacket(m, M_DONTWAIT);
1417 		M_PULLUP(mb0, hlen);
1418 		if (mb0 == NULL) {
1419 			free(rte, M_MRTABLE);
1420 			splx(s);
1421 			return (ENOBUFS);
1422 		}
1423 
1424 		/* is there an upcall waiting for this flow? */
1425 		hash = MFCHASH(ip->ip_src, ip->ip_dst);
1426 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1427 			if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1428 			    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1429 			    rt->mfc_stall != NULL)
1430 				break;
1431 		}
1432 
1433 		if (rt == NULL) {
1434 			int i;
1435 			struct igmpmsg *im;
1436 
1437 			/*
1438 			 * Locate the vifi for the incoming interface for
1439 			 * this packet.
1440 			 * If none found, drop packet.
1441 			 */
1442 			for (vifi = 0; vifi < numvifs &&
1443 				 viftable[vifi].v_ifp != ifp; vifi++)
1444 				;
1445 			if (vifi >= numvifs) /* vif not found, drop packet */
1446 				goto non_fatal;
1447 
1448 			/* no upcall, so make a new entry */
1449 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1450 						  M_NOWAIT);
1451 			if (rt == NULL)
1452 				goto fail;
1453 
1454 			/*
1455 			 * Make a copy of the header to send to the user level
1456 			 * process
1457 			 */
1458 			mm = m_copym(m, 0, hlen, M_DONTWAIT);
1459 			M_PULLUP(mm, hlen);
1460 			if (mm == NULL)
1461 				goto fail1;
1462 
1463 			/*
1464 			 * Send message to routing daemon to install
1465 			 * a route into the kernel table
1466 			 */
1467 
1468 			im = mtod(mm, struct igmpmsg *);
1469 			im->im_msgtype = IGMPMSG_NOCACHE;
1470 			im->im_mbz = 0;
1471 			im->im_vif = vifi;
1472 
1473 			mrtstat.mrts_upcalls++;
1474 
1475 			sockaddr_in_init(&sin, &ip->ip_src, 0);
1476 			if (socket_send(ip_mrouter, mm, &sin) < 0) {
1477 				log(LOG_WARNING,
1478 				    "ip_mforward: ip_mrouter socket queue full\n");
1479 				++mrtstat.mrts_upq_sockfull;
1480 			fail1:
1481 				free(rt, M_MRTABLE);
1482 			fail:
1483 				free(rte, M_MRTABLE);
1484 				m_freem(mb0);
1485 				splx(s);
1486 				return (ENOBUFS);
1487 			}
1488 
1489 			/* insert new entry at head of hash chain */
1490 			rt->mfc_origin = ip->ip_src;
1491 			rt->mfc_mcastgrp = ip->ip_dst;
1492 			rt->mfc_pkt_cnt = 0;
1493 			rt->mfc_byte_cnt = 0;
1494 			rt->mfc_wrong_if = 0;
1495 			rt->mfc_expire = UPCALL_EXPIRE;
1496 			nexpire[hash]++;
1497 			for (i = 0; i < numvifs; i++) {
1498 				rt->mfc_ttls[i] = 0;
1499 				rt->mfc_flags[i] = 0;
1500 			}
1501 			rt->mfc_parent = -1;
1502 
1503 			/* clear the RP address */
1504 			rt->mfc_rp = zeroin_addr;
1505 
1506 			rt->mfc_bw_meter = NULL;
1507 
1508 			/* link into table */
1509 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1510 			/* Add this entry to the end of the queue */
1511 			rt->mfc_stall = rte;
1512 		} else {
1513 			/* determine if q has overflowed */
1514 			struct rtdetq **p;
1515 			int npkts = 0;
1516 
1517 			/*
1518 			 * XXX ouch! we need to append to the list, but we
1519 			 * only have a pointer to the front, so we have to
1520 			 * scan the entire list every time.
1521 			 */
1522 			for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1523 				if (++npkts > MAX_UPQ) {
1524 					mrtstat.mrts_upq_ovflw++;
1525 				non_fatal:
1526 					free(rte, M_MRTABLE);
1527 					m_freem(mb0);
1528 					splx(s);
1529 					return (0);
1530 				}
1531 
1532 			/* Add this entry to the end of the queue */
1533 			*p = rte;
1534 		}
1535 
1536 		rte->next = NULL;
1537 		rte->m = mb0;
1538 		rte->ifp = ifp;
1539 #ifdef UPCALL_TIMING
1540 		rte->t = tp;
1541 #endif /* UPCALL_TIMING */
1542 
1543 		splx(s);
1544 
1545 		return (0);
1546 	}
1547 }
1548 
1549 
1550 /*ARGSUSED*/
1551 static void
1552 expire_upcalls(void *v)
1553 {
1554 	int i;
1555 	int s;
1556 
1557 	s = splsoftnet();
1558 
1559 	for (i = 0; i < MFCTBLSIZ; i++) {
1560 		struct mfc *rt, *nrt;
1561 
1562 		if (nexpire[i] == 0)
1563 			continue;
1564 
1565 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1566 			nrt = LIST_NEXT(rt, mfc_hash);
1567 
1568 			if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1569 				continue;
1570 			nexpire[i]--;
1571 
1572 			/*
1573 			 * free the bw_meter entries
1574 			 */
1575 			while (rt->mfc_bw_meter != NULL) {
1576 				struct bw_meter *x = rt->mfc_bw_meter;
1577 
1578 				rt->mfc_bw_meter = x->bm_mfc_next;
1579 				free(x, M_BWMETER);
1580 			}
1581 
1582 			++mrtstat.mrts_cache_cleanups;
1583 			if (mrtdebug & DEBUG_EXPIRE)
1584 				log(LOG_DEBUG,
1585 				    "expire_upcalls: expiring (%x %x)\n",
1586 				    ntohl(rt->mfc_origin.s_addr),
1587 				    ntohl(rt->mfc_mcastgrp.s_addr));
1588 
1589 			expire_mfc(rt);
1590 		}
1591 	}
1592 
1593 	splx(s);
1594 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1595 	    expire_upcalls, NULL);
1596 }
1597 
1598 /*
1599  * Packet forwarding routine once entry in the cache is made
1600  */
1601 static int
1602 #ifdef RSVP_ISI
1603 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1604 #else
1605 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1606 #endif /* RSVP_ISI */
1607 {
1608 	struct ip  *ip = mtod(m, struct ip *);
1609 	vifi_t vifi;
1610 	struct vif *vifp;
1611 	struct sockaddr_in sin;
1612 	int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1613 
1614 /*
1615  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1616  * input, they shouldn't get counted on output, so statistics keeping is
1617  * separate.
1618  */
1619 #define MC_SEND(ip, vifp, m) do {					\
1620 	if ((vifp)->v_flags & VIFF_TUNNEL)				\
1621 		encap_send((ip), (vifp), (m));				\
1622 	else								\
1623 		phyint_send((ip), (vifp), (m));				\
1624 } while (/*CONSTCOND*/ 0)
1625 
1626 #ifdef RSVP_ISI
1627 	/*
1628 	 * If xmt_vif is not -1, send on only the requested vif.
1629 	 *
1630 	 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1631 	 */
1632 	if (xmt_vif < numvifs) {
1633 #ifdef PIM
1634 		if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1635 			pim_register_send(ip, viftable + xmt_vif, m, rt);
1636 		else
1637 #endif
1638 		MC_SEND(ip, viftable + xmt_vif, m);
1639 		return (1);
1640 	}
1641 #endif /* RSVP_ISI */
1642 
1643 	/*
1644 	 * Don't forward if it didn't arrive from the parent vif for its origin.
1645 	 */
1646 	vifi = rt->mfc_parent;
1647 	if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1648 		/* came in the wrong interface */
1649 		if (mrtdebug & DEBUG_FORWARD)
1650 			log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1651 			    ifp, vifi,
1652 			    vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1653 		++mrtstat.mrts_wrong_if;
1654 		++rt->mfc_wrong_if;
1655 		/*
1656 		 * If we are doing PIM assert processing, send a message
1657 		 * to the routing daemon.
1658 		 *
1659 		 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1660 		 * can complete the SPT switch, regardless of the type
1661 		 * of the iif (broadcast media, GRE tunnel, etc).
1662 		 */
1663 		if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1664 			struct timeval now;
1665 			u_int32_t delta;
1666 
1667 #ifdef PIM
1668 			if (ifp == &multicast_register_if)
1669 				pimstat.pims_rcv_registers_wrongiif++;
1670 #endif
1671 
1672 			/* Get vifi for the incoming packet */
1673 			for (vifi = 0;
1674 			     vifi < numvifs && viftable[vifi].v_ifp != ifp;
1675 			     vifi++)
1676 			    ;
1677 			if (vifi >= numvifs) {
1678 				/* The iif is not found: ignore the packet. */
1679 				return (0);
1680 			}
1681 
1682 			if (rt->mfc_flags[vifi] &
1683 			    MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1684 				/* WRONGVIF disabled: ignore the packet */
1685 				return (0);
1686 			}
1687 
1688 			microtime(&now);
1689 
1690 			TV_DELTA(rt->mfc_last_assert, now, delta);
1691 
1692 			if (delta > ASSERT_MSG_TIME) {
1693 				struct igmpmsg *im;
1694 				int hlen = ip->ip_hl << 2;
1695 				struct mbuf *mm =
1696 				    m_copym(m, 0, hlen, M_DONTWAIT);
1697 
1698 				M_PULLUP(mm, hlen);
1699 				if (mm == NULL)
1700 					return (ENOBUFS);
1701 
1702 				rt->mfc_last_assert = now;
1703 
1704 				im = mtod(mm, struct igmpmsg *);
1705 				im->im_msgtype	= IGMPMSG_WRONGVIF;
1706 				im->im_mbz	= 0;
1707 				im->im_vif	= vifi;
1708 
1709 				mrtstat.mrts_upcalls++;
1710 
1711 				sockaddr_in_init(&sin, &im->im_src, 0);
1712 				if (socket_send(ip_mrouter, mm, &sin) < 0) {
1713 					log(LOG_WARNING,
1714 					    "ip_mforward: ip_mrouter socket queue full\n");
1715 					++mrtstat.mrts_upq_sockfull;
1716 					return (ENOBUFS);
1717 				}
1718 			}
1719 		}
1720 		return (0);
1721 	}
1722 
1723 	/* If I sourced this packet, it counts as output, else it was input. */
1724 	if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1725 		viftable[vifi].v_pkt_out++;
1726 		viftable[vifi].v_bytes_out += plen;
1727 	} else {
1728 		viftable[vifi].v_pkt_in++;
1729 		viftable[vifi].v_bytes_in += plen;
1730 	}
1731 	rt->mfc_pkt_cnt++;
1732 	rt->mfc_byte_cnt += plen;
1733 
1734 	/*
1735 	 * For each vif, decide if a copy of the packet should be forwarded.
1736 	 * Forward if:
1737 	 *		- the ttl exceeds the vif's threshold
1738 	 *		- there are group members downstream on interface
1739 	 */
1740 	for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1741 		if ((rt->mfc_ttls[vifi] > 0) &&
1742 			(ip->ip_ttl > rt->mfc_ttls[vifi])) {
1743 			vifp->v_pkt_out++;
1744 			vifp->v_bytes_out += plen;
1745 #ifdef PIM
1746 			if (vifp->v_flags & VIFF_REGISTER)
1747 				pim_register_send(ip, vifp, m, rt);
1748 			else
1749 #endif
1750 			MC_SEND(ip, vifp, m);
1751 		}
1752 
1753 	/*
1754 	 * Perform upcall-related bw measuring.
1755 	 */
1756 	if (rt->mfc_bw_meter != NULL) {
1757 		struct bw_meter *x;
1758 		struct timeval now;
1759 
1760 		microtime(&now);
1761 		for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1762 			bw_meter_receive_packet(x, plen, &now);
1763 	}
1764 
1765 	return (0);
1766 }
1767 
1768 #ifdef RSVP_ISI
1769 /*
1770  * check if a vif number is legal/ok. This is used by ip_output.
1771  */
1772 int
1773 legal_vif_num(int vif)
1774 {
1775 	if (vif >= 0 && vif < numvifs)
1776 		return (1);
1777 	else
1778 		return (0);
1779 }
1780 #endif /* RSVP_ISI */
1781 
1782 static void
1783 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1784 {
1785 	struct mbuf *mb_copy;
1786 	int hlen = ip->ip_hl << 2;
1787 
1788 	/*
1789 	 * Make a new reference to the packet; make sure that
1790 	 * the IP header is actually copied, not just referenced,
1791 	 * so that ip_output() only scribbles on the copy.
1792 	 */
1793 	mb_copy = m_copypacket(m, M_DONTWAIT);
1794 	M_PULLUP(mb_copy, hlen);
1795 	if (mb_copy == NULL)
1796 		return;
1797 
1798 	if (vifp->v_rate_limit <= 0)
1799 		tbf_send_packet(vifp, mb_copy);
1800 	else
1801 		tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1802 		    ntohs(ip->ip_len));
1803 }
1804 
1805 static void
1806 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1807 {
1808 	struct mbuf *mb_copy;
1809 	struct ip *ip_copy;
1810 	int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1811 
1812 	/* Take care of delayed checksums */
1813 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1814 		in_delayed_cksum(m);
1815 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1816 	}
1817 
1818 	/*
1819 	 * copy the old packet & pullup it's IP header into the
1820 	 * new mbuf so we can modify it.  Try to fill the new
1821 	 * mbuf since if we don't the ethernet driver will.
1822 	 */
1823 	MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1824 	if (mb_copy == NULL)
1825 		return;
1826 	mb_copy->m_data += max_linkhdr;
1827 	mb_copy->m_pkthdr.len = len;
1828 	mb_copy->m_len = sizeof(multicast_encap_iphdr);
1829 
1830 	if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
1831 		m_freem(mb_copy);
1832 		return;
1833 	}
1834 	i = MHLEN - max_linkhdr;
1835 	if (i > len)
1836 		i = len;
1837 	mb_copy = m_pullup(mb_copy, i);
1838 	if (mb_copy == NULL)
1839 		return;
1840 
1841 	/*
1842 	 * fill in the encapsulating IP header.
1843 	 */
1844 	ip_copy = mtod(mb_copy, struct ip *);
1845 	*ip_copy = multicast_encap_iphdr;
1846 	if (len < IP_MINFRAGSIZE)
1847 		ip_copy->ip_id = 0;
1848 	else
1849 		ip_copy->ip_id = ip_newid(NULL);
1850 	ip_copy->ip_len = htons(len);
1851 	ip_copy->ip_src = vifp->v_lcl_addr;
1852 	ip_copy->ip_dst = vifp->v_rmt_addr;
1853 
1854 	/*
1855 	 * turn the encapsulated IP header back into a valid one.
1856 	 */
1857 	ip = (struct ip *)((char *)ip_copy + sizeof(multicast_encap_iphdr));
1858 	--ip->ip_ttl;
1859 	ip->ip_sum = 0;
1860 	mb_copy->m_data += sizeof(multicast_encap_iphdr);
1861 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1862 	mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1863 
1864 	if (vifp->v_rate_limit <= 0)
1865 		tbf_send_packet(vifp, mb_copy);
1866 	else
1867 		tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1868 }
1869 
1870 /*
1871  * De-encapsulate a packet and feed it back through ip input.
1872  */
1873 static void
1874 vif_input(struct mbuf *m, ...)
1875 {
1876 	int off, proto;
1877 	va_list ap;
1878 	struct vif *vifp;
1879 	int s;
1880 	struct ifqueue *ifq;
1881 
1882 	va_start(ap, m);
1883 	off = va_arg(ap, int);
1884 	proto = va_arg(ap, int);
1885 	va_end(ap);
1886 
1887 	vifp = (struct vif *)encap_getarg(m);
1888 	if (!vifp || proto != ENCAP_PROTO) {
1889 		m_freem(m);
1890 		mrtstat.mrts_bad_tunnel++;
1891 		return;
1892 	}
1893 
1894 	m_adj(m, off);
1895 	m->m_pkthdr.rcvif = vifp->v_ifp;
1896 	ifq = &ipintrq;
1897 	s = splnet();
1898 	if (IF_QFULL(ifq)) {
1899 		IF_DROP(ifq);
1900 		m_freem(m);
1901 	} else {
1902 		IF_ENQUEUE(ifq, m);
1903 		/*
1904 		 * normally we would need a "schednetisr(NETISR_IP)"
1905 		 * here but we were called by ip_input and it is going
1906 		 * to loop back & try to dequeue the packet we just
1907 		 * queued as soon as we return so we avoid the
1908 		 * unnecessary software interrrupt.
1909 		 */
1910 	}
1911 	splx(s);
1912 }
1913 
1914 /*
1915  * Check if the packet should be received on the vif denoted by arg.
1916  * (The encap selection code will call this once per vif since each is
1917  * registered separately.)
1918  */
1919 static int
1920 vif_encapcheck(struct mbuf *m, int off, int proto, void *arg)
1921 {
1922 	struct vif *vifp;
1923 	struct ip ip;
1924 
1925 #ifdef DIAGNOSTIC
1926 	if (!arg || proto != IPPROTO_IPV4)
1927 		panic("unexpected arg in vif_encapcheck");
1928 #endif
1929 
1930 	/*
1931 	 * Accept the packet only if the inner heaader is multicast
1932 	 * and the outer header matches a tunnel-mode vif.  Order
1933 	 * checks in the hope that common non-matching packets will be
1934 	 * rejected quickly.  Assume that unicast IPv4 traffic in a
1935 	 * parallel tunnel (e.g. gif(4)) is unlikely.
1936 	 */
1937 
1938 	/* Obtain the outer IP header and the vif pointer. */
1939 	m_copydata((struct mbuf *)m, 0, sizeof(ip), (void *)&ip);
1940 	vifp = (struct vif *)arg;
1941 
1942 	/*
1943 	 * The outer source must match the vif's remote peer address.
1944 	 * For a multicast router with several tunnels, this is the
1945 	 * only check that will fail on packets in other tunnels,
1946 	 * assuming the local address is the same.
1947 	 */
1948 	if (!in_hosteq(vifp->v_rmt_addr, ip.ip_src))
1949 		return 0;
1950 
1951 	/* The outer destination must match the vif's local address. */
1952 	if (!in_hosteq(vifp->v_lcl_addr, ip.ip_dst))
1953 		return 0;
1954 
1955 	/* The vif must be of tunnel type. */
1956 	if ((vifp->v_flags & VIFF_TUNNEL) == 0)
1957 		return 0;
1958 
1959 	/* Check that the inner destination is multicast. */
1960 	m_copydata((struct mbuf *)m, off, sizeof(ip), (void *)&ip);
1961 	if (!IN_MULTICAST(ip.ip_dst.s_addr))
1962 		return 0;
1963 
1964 	/*
1965 	 * We have checked that both the outer src and dst addresses
1966 	 * match the vif, and that the inner destination is multicast
1967 	 * (224/5).  By claiming more than 64, we intend to
1968 	 * preferentially take packets that also match a parallel
1969 	 * gif(4).
1970 	 */
1971 	return 32 + 32 + 5;
1972 }
1973 
1974 /*
1975  * Token bucket filter module
1976  */
1977 static void
1978 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
1979 {
1980 
1981 	if (len > MAX_BKT_SIZE) {
1982 		/* drop if packet is too large */
1983 		mrtstat.mrts_pkt2large++;
1984 		m_freem(m);
1985 		return;
1986 	}
1987 
1988 	tbf_update_tokens(vifp);
1989 
1990 	/*
1991 	 * If there are enough tokens, and the queue is empty, send this packet
1992 	 * out immediately.  Otherwise, try to insert it on this vif's queue.
1993 	 */
1994 	if (vifp->tbf_q_len == 0) {
1995 		if (len <= vifp->tbf_n_tok) {
1996 			vifp->tbf_n_tok -= len;
1997 			tbf_send_packet(vifp, m);
1998 		} else {
1999 			/* queue packet and timeout till later */
2000 			tbf_queue(vifp, m);
2001 			callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2002 			    tbf_reprocess_q, vifp);
2003 		}
2004 	} else {
2005 		if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
2006 		    !tbf_dq_sel(vifp, ip)) {
2007 			/* queue full, and couldn't make room */
2008 			mrtstat.mrts_q_overflow++;
2009 			m_freem(m);
2010 		} else {
2011 			/* queue length low enough, or made room */
2012 			tbf_queue(vifp, m);
2013 			tbf_process_q(vifp);
2014 		}
2015 	}
2016 }
2017 
2018 /*
2019  * adds a packet to the queue at the interface
2020  */
2021 static void
2022 tbf_queue(struct vif *vifp, struct mbuf *m)
2023 {
2024 	int s = splsoftnet();
2025 
2026 	/* insert at tail */
2027 	*vifp->tbf_t = m;
2028 	vifp->tbf_t = &m->m_nextpkt;
2029 	vifp->tbf_q_len++;
2030 
2031 	splx(s);
2032 }
2033 
2034 
2035 /*
2036  * processes the queue at the interface
2037  */
2038 static void
2039 tbf_process_q(struct vif *vifp)
2040 {
2041 	struct mbuf *m;
2042 	int len;
2043 	int s = splsoftnet();
2044 
2045 	/*
2046 	 * Loop through the queue at the interface and send as many packets
2047 	 * as possible.
2048 	 */
2049 	for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
2050 		len = ntohs(mtod(m, struct ip *)->ip_len);
2051 
2052 		/* determine if the packet can be sent */
2053 		if (len <= vifp->tbf_n_tok) {
2054 			/* if so,
2055 			 * reduce no of tokens, dequeue the packet,
2056 			 * send the packet.
2057 			 */
2058 			if ((vifp->tbf_q = m->m_nextpkt) == NULL)
2059 				vifp->tbf_t = &vifp->tbf_q;
2060 			--vifp->tbf_q_len;
2061 
2062 			m->m_nextpkt = NULL;
2063 			vifp->tbf_n_tok -= len;
2064 			tbf_send_packet(vifp, m);
2065 		} else
2066 			break;
2067 	}
2068 	splx(s);
2069 }
2070 
2071 static void
2072 tbf_reprocess_q(void *arg)
2073 {
2074 	struct vif *vifp = arg;
2075 
2076 	if (ip_mrouter == NULL)
2077 		return;
2078 
2079 	tbf_update_tokens(vifp);
2080 	tbf_process_q(vifp);
2081 
2082 	if (vifp->tbf_q_len != 0)
2083 		callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2084 		    tbf_reprocess_q, vifp);
2085 }
2086 
2087 /* function that will selectively discard a member of the queue
2088  * based on the precedence value and the priority
2089  */
2090 static int
2091 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2092 {
2093 	u_int p;
2094 	struct mbuf **mp, *m;
2095 	int s = splsoftnet();
2096 
2097 	p = priority(vifp, ip);
2098 
2099 	for (mp = &vifp->tbf_q, m = *mp;
2100 	    m != NULL;
2101 	    mp = &m->m_nextpkt, m = *mp) {
2102 		if (p > priority(vifp, mtod(m, struct ip *))) {
2103 			if ((*mp = m->m_nextpkt) == NULL)
2104 				vifp->tbf_t = mp;
2105 			--vifp->tbf_q_len;
2106 
2107 			m_freem(m);
2108 			mrtstat.mrts_drop_sel++;
2109 			splx(s);
2110 			return (1);
2111 		}
2112 	}
2113 	splx(s);
2114 	return (0);
2115 }
2116 
2117 static void
2118 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2119 {
2120 	int error;
2121 	int s = splsoftnet();
2122 
2123 	if (vifp->v_flags & VIFF_TUNNEL) {
2124 		/* If tunnel options */
2125 		ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL);
2126 	} else {
2127 		/* if physical interface option, extract the options and then send */
2128 		struct ip_moptions imo;
2129 
2130 		imo.imo_multicast_ifp = vifp->v_ifp;
2131 		imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2132 		imo.imo_multicast_loop = 1;
2133 #ifdef RSVP_ISI
2134 		imo.imo_multicast_vif = -1;
2135 #endif
2136 
2137 		error = ip_output(m, NULL, NULL, IP_FORWARDING|IP_MULTICASTOPTS,
2138 		    &imo, NULL);
2139 
2140 		if (mrtdebug & DEBUG_XMIT)
2141 			log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2142 			    (long)(vifp - viftable), error);
2143 	}
2144 	splx(s);
2145 }
2146 
2147 /* determine the current time and then
2148  * the elapsed time (between the last time and time now)
2149  * in milliseconds & update the no. of tokens in the bucket
2150  */
2151 static void
2152 tbf_update_tokens(struct vif *vifp)
2153 {
2154 	struct timeval tp;
2155 	u_int32_t tm;
2156 	int s = splsoftnet();
2157 
2158 	microtime(&tp);
2159 
2160 	TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2161 
2162 	/*
2163 	 * This formula is actually
2164 	 * "time in seconds" * "bytes/second".
2165 	 *
2166 	 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2167 	 *
2168 	 * The (1000/1024) was introduced in add_vif to optimize
2169 	 * this divide into a shift.
2170 	 */
2171 	vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2172 	vifp->tbf_last_pkt_t = tp;
2173 
2174 	if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2175 		vifp->tbf_n_tok = MAX_BKT_SIZE;
2176 
2177 	splx(s);
2178 }
2179 
2180 static int
2181 priority(struct vif *vifp, struct ip *ip)
2182 {
2183 	int prio = 50;	/* the lowest priority -- default case */
2184 
2185 	/* temporary hack; may add general packet classifier some day */
2186 
2187 	/*
2188 	 * The UDP port space is divided up into four priority ranges:
2189 	 * [0, 16384)     : unclassified - lowest priority
2190 	 * [16384, 32768) : audio - highest priority
2191 	 * [32768, 49152) : whiteboard - medium priority
2192 	 * [49152, 65536) : video - low priority
2193 	 */
2194 	if (ip->ip_p == IPPROTO_UDP) {
2195 		struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2196 
2197 		switch (ntohs(udp->uh_dport) & 0xc000) {
2198 		case 0x4000:
2199 			prio = 70;
2200 			break;
2201 		case 0x8000:
2202 			prio = 60;
2203 			break;
2204 		case 0xc000:
2205 			prio = 55;
2206 			break;
2207 		}
2208 
2209 		if (tbfdebug > 1)
2210 			log(LOG_DEBUG, "port %x prio %d\n",
2211 			    ntohs(udp->uh_dport), prio);
2212 	}
2213 
2214 	return (prio);
2215 }
2216 
2217 /*
2218  * End of token bucket filter modifications
2219  */
2220 #ifdef RSVP_ISI
2221 int
2222 ip_rsvp_vif_init(struct socket *so, struct mbuf *m)
2223 {
2224 	int vifi, s;
2225 
2226 	if (rsvpdebug)
2227 		printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2228 		    so->so_type, so->so_proto->pr_protocol);
2229 
2230 	if (so->so_type != SOCK_RAW ||
2231 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2232 		return (EOPNOTSUPP);
2233 
2234 	/* Check mbuf. */
2235 	if (m == NULL || m->m_len != sizeof(int)) {
2236 		return (EINVAL);
2237 	}
2238 	vifi = *(mtod(m, int *));
2239 
2240 	if (rsvpdebug)
2241 		printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",
2242 		       vifi, rsvp_on);
2243 
2244 	s = splsoftnet();
2245 
2246 	/* Check vif. */
2247 	if (!legal_vif_num(vifi)) {
2248 		splx(s);
2249 		return (EADDRNOTAVAIL);
2250 	}
2251 
2252 	/* Check if socket is available. */
2253 	if (viftable[vifi].v_rsvpd != NULL) {
2254 		splx(s);
2255 		return (EADDRINUSE);
2256 	}
2257 
2258 	viftable[vifi].v_rsvpd = so;
2259 	/*
2260 	 * This may seem silly, but we need to be sure we don't over-increment
2261 	 * the RSVP counter, in case something slips up.
2262 	 */
2263 	if (!viftable[vifi].v_rsvp_on) {
2264 		viftable[vifi].v_rsvp_on = 1;
2265 		rsvp_on++;
2266 	}
2267 
2268 	splx(s);
2269 	return (0);
2270 }
2271 
2272 int
2273 ip_rsvp_vif_done(struct socket *so, struct mbuf *m)
2274 {
2275 	int vifi, s;
2276 
2277 	if (rsvpdebug)
2278 		printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2279 		    so->so_type, so->so_proto->pr_protocol);
2280 
2281 	if (so->so_type != SOCK_RAW ||
2282 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2283 		return (EOPNOTSUPP);
2284 
2285 	/* Check mbuf. */
2286 	if (m == NULL || m->m_len != sizeof(int)) {
2287 		return (EINVAL);
2288 	}
2289 	vifi = *(mtod(m, int *));
2290 
2291 	s = splsoftnet();
2292 
2293 	/* Check vif. */
2294 	if (!legal_vif_num(vifi)) {
2295 		splx(s);
2296 		return (EADDRNOTAVAIL);
2297 	}
2298 
2299 	if (rsvpdebug)
2300 		printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2301 		    viftable[vifi].v_rsvpd, so);
2302 
2303 	viftable[vifi].v_rsvpd = NULL;
2304 	/*
2305 	 * This may seem silly, but we need to be sure we don't over-decrement
2306 	 * the RSVP counter, in case something slips up.
2307 	 */
2308 	if (viftable[vifi].v_rsvp_on) {
2309 		viftable[vifi].v_rsvp_on = 0;
2310 		rsvp_on--;
2311 	}
2312 
2313 	splx(s);
2314 	return (0);
2315 }
2316 
2317 void
2318 ip_rsvp_force_done(struct socket *so)
2319 {
2320 	int vifi, s;
2321 
2322 	/* Don't bother if it is not the right type of socket. */
2323 	if (so->so_type != SOCK_RAW ||
2324 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2325 		return;
2326 
2327 	s = splsoftnet();
2328 
2329 	/*
2330 	 * The socket may be attached to more than one vif...this
2331 	 * is perfectly legal.
2332 	 */
2333 	for (vifi = 0; vifi < numvifs; vifi++) {
2334 		if (viftable[vifi].v_rsvpd == so) {
2335 			viftable[vifi].v_rsvpd = NULL;
2336 			/*
2337 			 * This may seem silly, but we need to be sure we don't
2338 			 * over-decrement the RSVP counter, in case something
2339 			 * slips up.
2340 			 */
2341 			if (viftable[vifi].v_rsvp_on) {
2342 				viftable[vifi].v_rsvp_on = 0;
2343 				rsvp_on--;
2344 			}
2345 		}
2346 	}
2347 
2348 	splx(s);
2349 	return;
2350 }
2351 
2352 void
2353 rsvp_input(struct mbuf *m, struct ifnet *ifp)
2354 {
2355 	int vifi, s;
2356 	struct ip *ip = mtod(m, struct ip *);
2357 	struct sockaddr_in rsvp_src;
2358 
2359 	if (rsvpdebug)
2360 		printf("rsvp_input: rsvp_on %d\n", rsvp_on);
2361 
2362 	/*
2363 	 * Can still get packets with rsvp_on = 0 if there is a local member
2364 	 * of the group to which the RSVP packet is addressed.  But in this
2365 	 * case we want to throw the packet away.
2366 	 */
2367 	if (!rsvp_on) {
2368 		m_freem(m);
2369 		return;
2370 	}
2371 
2372 	/*
2373 	 * If the old-style non-vif-associated socket is set, then use
2374 	 * it and ignore the new ones.
2375 	 */
2376 	if (ip_rsvpd != NULL) {
2377 		if (rsvpdebug)
2378 			printf("rsvp_input: "
2379 			    "Sending packet up old-style socket\n");
2380 		rip_input(m);	/*XXX*/
2381 		return;
2382 	}
2383 
2384 	s = splsoftnet();
2385 
2386 	if (rsvpdebug)
2387 		printf("rsvp_input: check vifs\n");
2388 
2389 	/* Find which vif the packet arrived on. */
2390 	for (vifi = 0; vifi < numvifs; vifi++) {
2391 		if (viftable[vifi].v_ifp == ifp)
2392 			break;
2393 	}
2394 
2395 	if (vifi == numvifs) {
2396 		/* Can't find vif packet arrived on. Drop packet. */
2397 		if (rsvpdebug)
2398 			printf("rsvp_input: "
2399 			    "Can't find vif for packet...dropping it.\n");
2400 		m_freem(m);
2401 		splx(s);
2402 		return;
2403 	}
2404 
2405 	if (rsvpdebug)
2406 		printf("rsvp_input: check socket\n");
2407 
2408 	if (viftable[vifi].v_rsvpd == NULL) {
2409 		/*
2410 		 * drop packet, since there is no specific socket for this
2411 		 * interface
2412 		 */
2413 		if (rsvpdebug)
2414 			printf("rsvp_input: No socket defined for vif %d\n",
2415 			    vifi);
2416 		m_freem(m);
2417 		splx(s);
2418 		return;
2419 	}
2420 
2421 	sockaddr_in_init(&rsvp_src, &ip->ip_src, 0);
2422 
2423 	if (rsvpdebug && m)
2424 		printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
2425 		    m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv));
2426 
2427 	if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2428 		if (rsvpdebug)
2429 			printf("rsvp_input: Failed to append to socket\n");
2430 	else
2431 		if (rsvpdebug)
2432 			printf("rsvp_input: send packet up\n");
2433 
2434 	splx(s);
2435 }
2436 #endif /* RSVP_ISI */
2437 
2438 /*
2439  * Code for bandwidth monitors
2440  */
2441 
2442 /*
2443  * Define common interface for timeval-related methods
2444  */
2445 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2446 #define	BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2447 #define	BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2448 
2449 static uint32_t
2450 compute_bw_meter_flags(struct bw_upcall *req)
2451 {
2452     uint32_t flags = 0;
2453 
2454     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2455 	flags |= BW_METER_UNIT_PACKETS;
2456     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2457 	flags |= BW_METER_UNIT_BYTES;
2458     if (req->bu_flags & BW_UPCALL_GEQ)
2459 	flags |= BW_METER_GEQ;
2460     if (req->bu_flags & BW_UPCALL_LEQ)
2461 	flags |= BW_METER_LEQ;
2462 
2463     return flags;
2464 }
2465 
2466 /*
2467  * Add a bw_meter entry
2468  */
2469 static int
2470 add_bw_upcall(struct bw_upcall *req)
2471 {
2472     int s;
2473     struct mfc *mfc;
2474     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2475 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2476     struct timeval now;
2477     struct bw_meter *x;
2478     uint32_t flags;
2479 
2480     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2481 	return EOPNOTSUPP;
2482 
2483     /* Test if the flags are valid */
2484     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2485 	return EINVAL;
2486     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2487 	return EINVAL;
2488     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2489 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2490 	return EINVAL;
2491 
2492     /* Test if the threshold time interval is valid */
2493     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2494 	return EINVAL;
2495 
2496     flags = compute_bw_meter_flags(req);
2497 
2498     /*
2499      * Find if we have already same bw_meter entry
2500      */
2501     s = splsoftnet();
2502     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2503     if (mfc == NULL) {
2504 	splx(s);
2505 	return EADDRNOTAVAIL;
2506     }
2507     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2508 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2509 			   &req->bu_threshold.b_time, ==)) &&
2510 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2511 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2512 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
2513 	    splx(s);
2514 	    return 0;		/* XXX Already installed */
2515 	}
2516     }
2517 
2518     /* Allocate the new bw_meter entry */
2519     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2520     if (x == NULL) {
2521 	splx(s);
2522 	return ENOBUFS;
2523     }
2524 
2525     /* Set the new bw_meter entry */
2526     x->bm_threshold.b_time = req->bu_threshold.b_time;
2527     microtime(&now);
2528     x->bm_start_time = now;
2529     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2530     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2531     x->bm_measured.b_packets = 0;
2532     x->bm_measured.b_bytes = 0;
2533     x->bm_flags = flags;
2534     x->bm_time_next = NULL;
2535     x->bm_time_hash = BW_METER_BUCKETS;
2536 
2537     /* Add the new bw_meter entry to the front of entries for this MFC */
2538     x->bm_mfc = mfc;
2539     x->bm_mfc_next = mfc->mfc_bw_meter;
2540     mfc->mfc_bw_meter = x;
2541     schedule_bw_meter(x, &now);
2542     splx(s);
2543 
2544     return 0;
2545 }
2546 
2547 static void
2548 free_bw_list(struct bw_meter *list)
2549 {
2550     while (list != NULL) {
2551 	struct bw_meter *x = list;
2552 
2553 	list = list->bm_mfc_next;
2554 	unschedule_bw_meter(x);
2555 	free(x, M_BWMETER);
2556     }
2557 }
2558 
2559 /*
2560  * Delete one or multiple bw_meter entries
2561  */
2562 static int
2563 del_bw_upcall(struct bw_upcall *req)
2564 {
2565     int s;
2566     struct mfc *mfc;
2567     struct bw_meter *x;
2568 
2569     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2570 	return EOPNOTSUPP;
2571 
2572     s = splsoftnet();
2573     /* Find the corresponding MFC entry */
2574     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2575     if (mfc == NULL) {
2576 	splx(s);
2577 	return EADDRNOTAVAIL;
2578     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2579 	/*
2580 	 * Delete all bw_meter entries for this mfc
2581 	 */
2582 	struct bw_meter *list;
2583 
2584 	list = mfc->mfc_bw_meter;
2585 	mfc->mfc_bw_meter = NULL;
2586 	free_bw_list(list);
2587 	splx(s);
2588 	return 0;
2589     } else {			/* Delete a single bw_meter entry */
2590 	struct bw_meter *prev;
2591 	uint32_t flags = 0;
2592 
2593 	flags = compute_bw_meter_flags(req);
2594 
2595 	/* Find the bw_meter entry to delete */
2596 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2597 	     prev = x, x = x->bm_mfc_next) {
2598 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2599 			       &req->bu_threshold.b_time, ==)) &&
2600 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2601 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2602 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2603 		break;
2604 	}
2605 	if (x != NULL) { /* Delete entry from the list for this MFC */
2606 	    if (prev != NULL)
2607 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2608 	    else
2609 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2610 
2611 	    unschedule_bw_meter(x);
2612 	    splx(s);
2613 	    /* Free the bw_meter entry */
2614 	    free(x, M_BWMETER);
2615 	    return 0;
2616 	} else {
2617 	    splx(s);
2618 	    return EINVAL;
2619 	}
2620     }
2621     /* NOTREACHED */
2622 }
2623 
2624 /*
2625  * Perform bandwidth measurement processing that may result in an upcall
2626  */
2627 static void
2628 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2629 {
2630     struct timeval delta;
2631 
2632     delta = *nowp;
2633     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2634 
2635     if (x->bm_flags & BW_METER_GEQ) {
2636 	/*
2637 	 * Processing for ">=" type of bw_meter entry
2638 	 */
2639 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2640 	    /* Reset the bw_meter entry */
2641 	    x->bm_start_time = *nowp;
2642 	    x->bm_measured.b_packets = 0;
2643 	    x->bm_measured.b_bytes = 0;
2644 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2645 	}
2646 
2647 	/* Record that a packet is received */
2648 	x->bm_measured.b_packets++;
2649 	x->bm_measured.b_bytes += plen;
2650 
2651 	/*
2652 	 * Test if we should deliver an upcall
2653 	 */
2654 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2655 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2656 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2657 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2658 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2659 		/* Prepare an upcall for delivery */
2660 		bw_meter_prepare_upcall(x, nowp);
2661 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2662 	    }
2663 	}
2664     } else if (x->bm_flags & BW_METER_LEQ) {
2665 	/*
2666 	 * Processing for "<=" type of bw_meter entry
2667 	 */
2668 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2669 	    /*
2670 	     * We are behind time with the multicast forwarding table
2671 	     * scanning for "<=" type of bw_meter entries, so test now
2672 	     * if we should deliver an upcall.
2673 	     */
2674 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2675 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2676 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2677 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2678 		/* Prepare an upcall for delivery */
2679 		bw_meter_prepare_upcall(x, nowp);
2680 	    }
2681 	    /* Reschedule the bw_meter entry */
2682 	    unschedule_bw_meter(x);
2683 	    schedule_bw_meter(x, nowp);
2684 	}
2685 
2686 	/* Record that a packet is received */
2687 	x->bm_measured.b_packets++;
2688 	x->bm_measured.b_bytes += plen;
2689 
2690 	/*
2691 	 * Test if we should restart the measuring interval
2692 	 */
2693 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2694 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2695 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
2696 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2697 	    /* Don't restart the measuring interval */
2698 	} else {
2699 	    /* Do restart the measuring interval */
2700 	    /*
2701 	     * XXX: note that we don't unschedule and schedule, because this
2702 	     * might be too much overhead per packet. Instead, when we process
2703 	     * all entries for a given timer hash bin, we check whether it is
2704 	     * really a timeout. If not, we reschedule at that time.
2705 	     */
2706 	    x->bm_start_time = *nowp;
2707 	    x->bm_measured.b_packets = 0;
2708 	    x->bm_measured.b_bytes = 0;
2709 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2710 	}
2711     }
2712 }
2713 
2714 /*
2715  * Prepare a bandwidth-related upcall
2716  */
2717 static void
2718 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2719 {
2720     struct timeval delta;
2721     struct bw_upcall *u;
2722 
2723     /*
2724      * Compute the measured time interval
2725      */
2726     delta = *nowp;
2727     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2728 
2729     /*
2730      * If there are too many pending upcalls, deliver them now
2731      */
2732     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2733 	bw_upcalls_send();
2734 
2735     /*
2736      * Set the bw_upcall entry
2737      */
2738     u = &bw_upcalls[bw_upcalls_n++];
2739     u->bu_src = x->bm_mfc->mfc_origin;
2740     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2741     u->bu_threshold.b_time = x->bm_threshold.b_time;
2742     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2743     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2744     u->bu_measured.b_time = delta;
2745     u->bu_measured.b_packets = x->bm_measured.b_packets;
2746     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2747     u->bu_flags = 0;
2748     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2749 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2750     if (x->bm_flags & BW_METER_UNIT_BYTES)
2751 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2752     if (x->bm_flags & BW_METER_GEQ)
2753 	u->bu_flags |= BW_UPCALL_GEQ;
2754     if (x->bm_flags & BW_METER_LEQ)
2755 	u->bu_flags |= BW_UPCALL_LEQ;
2756 }
2757 
2758 /*
2759  * Send the pending bandwidth-related upcalls
2760  */
2761 static void
2762 bw_upcalls_send(void)
2763 {
2764     struct mbuf *m;
2765     int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2766     struct sockaddr_in k_igmpsrc = {
2767 	    .sin_len = sizeof(k_igmpsrc),
2768 	    .sin_family = AF_INET,
2769     };
2770     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2771 				      0,		/* unused2 */
2772 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2773 				      0,		/* im_mbz  */
2774 				      0,		/* im_vif  */
2775 				      0,		/* unused3 */
2776 				      { 0 },		/* im_src  */
2777 				      { 0 } };		/* im_dst  */
2778 
2779     if (bw_upcalls_n == 0)
2780 	return;			/* No pending upcalls */
2781 
2782     bw_upcalls_n = 0;
2783 
2784     /*
2785      * Allocate a new mbuf, initialize it with the header and
2786      * the payload for the pending calls.
2787      */
2788     MGETHDR(m, M_DONTWAIT, MT_HEADER);
2789     if (m == NULL) {
2790 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2791 	return;
2792     }
2793 
2794     m->m_len = m->m_pkthdr.len = 0;
2795     m_copyback(m, 0, sizeof(struct igmpmsg), (void *)&igmpmsg);
2796     m_copyback(m, sizeof(struct igmpmsg), len, (void *)&bw_upcalls[0]);
2797 
2798     /*
2799      * Send the upcalls
2800      * XXX do we need to set the address in k_igmpsrc ?
2801      */
2802     mrtstat.mrts_upcalls++;
2803     if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2804 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2805 	++mrtstat.mrts_upq_sockfull;
2806     }
2807 }
2808 
2809 /*
2810  * Compute the timeout hash value for the bw_meter entries
2811  */
2812 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2813     do {								\
2814 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2815 									\
2816 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2817 	(hash) = next_timeval.tv_sec;					\
2818 	if (next_timeval.tv_usec)					\
2819 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2820 	(hash) %= BW_METER_BUCKETS;					\
2821     } while (/*CONSTCOND*/ 0)
2822 
2823 /*
2824  * Schedule a timer to process periodically bw_meter entry of type "<="
2825  * by linking the entry in the proper hash bucket.
2826  */
2827 static void
2828 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2829 {
2830     int time_hash;
2831 
2832     if (!(x->bm_flags & BW_METER_LEQ))
2833 	return;		/* XXX: we schedule timers only for "<=" entries */
2834 
2835     /*
2836      * Reset the bw_meter entry
2837      */
2838     x->bm_start_time = *nowp;
2839     x->bm_measured.b_packets = 0;
2840     x->bm_measured.b_bytes = 0;
2841     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2842 
2843     /*
2844      * Compute the timeout hash value and insert the entry
2845      */
2846     BW_METER_TIMEHASH(x, time_hash);
2847     x->bm_time_next = bw_meter_timers[time_hash];
2848     bw_meter_timers[time_hash] = x;
2849     x->bm_time_hash = time_hash;
2850 }
2851 
2852 /*
2853  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2854  * by removing the entry from the proper hash bucket.
2855  */
2856 static void
2857 unschedule_bw_meter(struct bw_meter *x)
2858 {
2859     int time_hash;
2860     struct bw_meter *prev, *tmp;
2861 
2862     if (!(x->bm_flags & BW_METER_LEQ))
2863 	return;		/* XXX: we schedule timers only for "<=" entries */
2864 
2865     /*
2866      * Compute the timeout hash value and delete the entry
2867      */
2868     time_hash = x->bm_time_hash;
2869     if (time_hash >= BW_METER_BUCKETS)
2870 	return;		/* Entry was not scheduled */
2871 
2872     for (prev = NULL, tmp = bw_meter_timers[time_hash];
2873 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2874 	if (tmp == x)
2875 	    break;
2876 
2877     if (tmp == NULL)
2878 	panic("unschedule_bw_meter: bw_meter entry not found");
2879 
2880     if (prev != NULL)
2881 	prev->bm_time_next = x->bm_time_next;
2882     else
2883 	bw_meter_timers[time_hash] = x->bm_time_next;
2884 
2885     x->bm_time_next = NULL;
2886     x->bm_time_hash = BW_METER_BUCKETS;
2887 }
2888 
2889 /*
2890  * Process all "<=" type of bw_meter that should be processed now,
2891  * and for each entry prepare an upcall if necessary. Each processed
2892  * entry is rescheduled again for the (periodic) processing.
2893  *
2894  * This is run periodically (once per second normally). On each round,
2895  * all the potentially matching entries are in the hash slot that we are
2896  * looking at.
2897  */
2898 static void
2899 bw_meter_process(void)
2900 {
2901     int s;
2902     static uint32_t last_tv_sec;	/* last time we processed this */
2903 
2904     uint32_t loops;
2905     int i;
2906     struct timeval now, process_endtime;
2907 
2908     microtime(&now);
2909     if (last_tv_sec == now.tv_sec)
2910 	return;		/* nothing to do */
2911 
2912     loops = now.tv_sec - last_tv_sec;
2913     last_tv_sec = now.tv_sec;
2914     if (loops > BW_METER_BUCKETS)
2915 	loops = BW_METER_BUCKETS;
2916 
2917     s = splsoftnet();
2918     /*
2919      * Process all bins of bw_meter entries from the one after the last
2920      * processed to the current one. On entry, i points to the last bucket
2921      * visited, so we need to increment i at the beginning of the loop.
2922      */
2923     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2924 	struct bw_meter *x, *tmp_list;
2925 
2926 	if (++i >= BW_METER_BUCKETS)
2927 	    i = 0;
2928 
2929 	/* Disconnect the list of bw_meter entries from the bin */
2930 	tmp_list = bw_meter_timers[i];
2931 	bw_meter_timers[i] = NULL;
2932 
2933 	/* Process the list of bw_meter entries */
2934 	while (tmp_list != NULL) {
2935 	    x = tmp_list;
2936 	    tmp_list = tmp_list->bm_time_next;
2937 
2938 	    /* Test if the time interval is over */
2939 	    process_endtime = x->bm_start_time;
2940 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2941 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2942 		/* Not yet: reschedule, but don't reset */
2943 		int time_hash;
2944 
2945 		BW_METER_TIMEHASH(x, time_hash);
2946 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2947 		    /*
2948 		     * XXX: somehow the bin processing is a bit ahead of time.
2949 		     * Put the entry in the next bin.
2950 		     */
2951 		    if (++time_hash >= BW_METER_BUCKETS)
2952 			time_hash = 0;
2953 		}
2954 		x->bm_time_next = bw_meter_timers[time_hash];
2955 		bw_meter_timers[time_hash] = x;
2956 		x->bm_time_hash = time_hash;
2957 
2958 		continue;
2959 	    }
2960 
2961 	    /*
2962 	     * Test if we should deliver an upcall
2963 	     */
2964 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2965 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2966 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2967 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2968 		/* Prepare an upcall for delivery */
2969 		bw_meter_prepare_upcall(x, &now);
2970 	    }
2971 
2972 	    /*
2973 	     * Reschedule for next processing
2974 	     */
2975 	    schedule_bw_meter(x, &now);
2976 	}
2977     }
2978 
2979     /* Send all upcalls that are pending delivery */
2980     bw_upcalls_send();
2981 
2982     splx(s);
2983 }
2984 
2985 /*
2986  * A periodic function for sending all upcalls that are pending delivery
2987  */
2988 static void
2989 expire_bw_upcalls_send(void *unused)
2990 {
2991     int s;
2992 
2993     s = splsoftnet();
2994     bw_upcalls_send();
2995     splx(s);
2996 
2997     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2998 		  expire_bw_upcalls_send, NULL);
2999 }
3000 
3001 /*
3002  * A periodic function for periodic scanning of the multicast forwarding
3003  * table for processing all "<=" bw_meter entries.
3004  */
3005 static void
3006 expire_bw_meter_process(void *unused)
3007 {
3008     if (mrt_api_config & MRT_MFC_BW_UPCALL)
3009 	bw_meter_process();
3010 
3011     callout_reset(&bw_meter_ch, BW_METER_PERIOD,
3012 		  expire_bw_meter_process, NULL);
3013 }
3014 
3015 /*
3016  * End of bandwidth monitoring code
3017  */
3018 
3019 #ifdef PIM
3020 /*
3021  * Send the packet up to the user daemon, or eventually do kernel encapsulation
3022  */
3023 static int
3024 pim_register_send(struct ip *ip, struct vif *vifp,
3025 	struct mbuf *m, struct mfc *rt)
3026 {
3027     struct mbuf *mb_copy, *mm;
3028 
3029     if (mrtdebug & DEBUG_PIM)
3030         log(LOG_DEBUG, "pim_register_send: ");
3031 
3032     mb_copy = pim_register_prepare(ip, m);
3033     if (mb_copy == NULL)
3034 	return ENOBUFS;
3035 
3036     /*
3037      * Send all the fragments. Note that the mbuf for each fragment
3038      * is freed by the sending machinery.
3039      */
3040     for (mm = mb_copy; mm; mm = mb_copy) {
3041 	mb_copy = mm->m_nextpkt;
3042 	mm->m_nextpkt = NULL;
3043 	mm = m_pullup(mm, sizeof(struct ip));
3044 	if (mm != NULL) {
3045 	    ip = mtod(mm, struct ip *);
3046 	    if ((mrt_api_config & MRT_MFC_RP) &&
3047 		!in_nullhost(rt->mfc_rp)) {
3048 		pim_register_send_rp(ip, vifp, mm, rt);
3049 	    } else {
3050 		pim_register_send_upcall(ip, vifp, mm, rt);
3051 	    }
3052 	}
3053     }
3054 
3055     return 0;
3056 }
3057 
3058 /*
3059  * Return a copy of the data packet that is ready for PIM Register
3060  * encapsulation.
3061  * XXX: Note that in the returned copy the IP header is a valid one.
3062  */
3063 static struct mbuf *
3064 pim_register_prepare(struct ip *ip, struct mbuf *m)
3065 {
3066     struct mbuf *mb_copy = NULL;
3067     int mtu;
3068 
3069     /* Take care of delayed checksums */
3070     if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3071 	in_delayed_cksum(m);
3072 	m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3073     }
3074 
3075     /*
3076      * Copy the old packet & pullup its IP header into the
3077      * new mbuf so we can modify it.
3078      */
3079     mb_copy = m_copypacket(m, M_DONTWAIT);
3080     if (mb_copy == NULL)
3081 	return NULL;
3082     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3083     if (mb_copy == NULL)
3084 	return NULL;
3085 
3086     /* take care of the TTL */
3087     ip = mtod(mb_copy, struct ip *);
3088     --ip->ip_ttl;
3089 
3090     /* Compute the MTU after the PIM Register encapsulation */
3091     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3092 
3093     if (ntohs(ip->ip_len) <= mtu) {
3094 	/* Turn the IP header into a valid one */
3095 	ip->ip_sum = 0;
3096 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3097     } else {
3098 	/* Fragment the packet */
3099 	if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3100 	    /* XXX: mb_copy was freed by ip_fragment() */
3101 	    return NULL;
3102 	}
3103     }
3104     return mb_copy;
3105 }
3106 
3107 /*
3108  * Send an upcall with the data packet to the user-level process.
3109  */
3110 static int
3111 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3112     struct mbuf *mb_copy, struct mfc *rt)
3113 {
3114     struct mbuf *mb_first;
3115     int len = ntohs(ip->ip_len);
3116     struct igmpmsg *im;
3117     struct sockaddr_in k_igmpsrc = {
3118 	    .sin_len = sizeof(k_igmpsrc),
3119 	    .sin_family = AF_INET,
3120     };
3121 
3122     /*
3123      * Add a new mbuf with an upcall header
3124      */
3125     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3126     if (mb_first == NULL) {
3127 	m_freem(mb_copy);
3128 	return ENOBUFS;
3129     }
3130     mb_first->m_data += max_linkhdr;
3131     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3132     mb_first->m_len = sizeof(struct igmpmsg);
3133     mb_first->m_next = mb_copy;
3134 
3135     /* Send message to routing daemon */
3136     im = mtod(mb_first, struct igmpmsg *);
3137     im->im_msgtype	= IGMPMSG_WHOLEPKT;
3138     im->im_mbz		= 0;
3139     im->im_vif		= vifp - viftable;
3140     im->im_src		= ip->ip_src;
3141     im->im_dst		= ip->ip_dst;
3142 
3143     k_igmpsrc.sin_addr	= ip->ip_src;
3144 
3145     mrtstat.mrts_upcalls++;
3146 
3147     if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3148 	if (mrtdebug & DEBUG_PIM)
3149 	    log(LOG_WARNING,
3150 		"mcast: pim_register_send_upcall: ip_mrouter socket queue full");
3151 	++mrtstat.mrts_upq_sockfull;
3152 	return ENOBUFS;
3153     }
3154 
3155     /* Keep statistics */
3156     pimstat.pims_snd_registers_msgs++;
3157     pimstat.pims_snd_registers_bytes += len;
3158 
3159     return 0;
3160 }
3161 
3162 /*
3163  * Encapsulate the data packet in PIM Register message and send it to the RP.
3164  */
3165 static int
3166 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3167 	struct mbuf *mb_copy, struct mfc *rt)
3168 {
3169     struct mbuf *mb_first;
3170     struct ip *ip_outer;
3171     struct pim_encap_pimhdr *pimhdr;
3172     int len = ntohs(ip->ip_len);
3173     vifi_t vifi = rt->mfc_parent;
3174 
3175     if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3176 	m_freem(mb_copy);
3177 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
3178     }
3179 
3180     /*
3181      * Add a new mbuf with the encapsulating header
3182      */
3183     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3184     if (mb_first == NULL) {
3185 	m_freem(mb_copy);
3186 	return ENOBUFS;
3187     }
3188     mb_first->m_data += max_linkhdr;
3189     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3190     mb_first->m_next = mb_copy;
3191 
3192     mb_first->m_pkthdr.len = len + mb_first->m_len;
3193 
3194     /*
3195      * Fill in the encapsulating IP and PIM header
3196      */
3197     ip_outer = mtod(mb_first, struct ip *);
3198     *ip_outer = pim_encap_iphdr;
3199      if (mb_first->m_pkthdr.len < IP_MINFRAGSIZE)
3200 	ip_outer->ip_id = 0;
3201     else
3202 	ip_outer->ip_id = ip_newid(NULL);
3203     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3204 			     sizeof(pim_encap_pimhdr));
3205     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3206     ip_outer->ip_dst = rt->mfc_rp;
3207     /*
3208      * Copy the inner header TOS to the outer header, and take care of the
3209      * IP_DF bit.
3210      */
3211     ip_outer->ip_tos = ip->ip_tos;
3212     if (ntohs(ip->ip_off) & IP_DF)
3213 	ip_outer->ip_off |= htons(IP_DF);
3214     pimhdr = (struct pim_encap_pimhdr *)((char *)ip_outer
3215 					 + sizeof(pim_encap_iphdr));
3216     *pimhdr = pim_encap_pimhdr;
3217     /* If the iif crosses a border, set the Border-bit */
3218     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3219 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3220 
3221     mb_first->m_data += sizeof(pim_encap_iphdr);
3222     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3223     mb_first->m_data -= sizeof(pim_encap_iphdr);
3224 
3225     if (vifp->v_rate_limit == 0)
3226 	tbf_send_packet(vifp, mb_first);
3227     else
3228 	tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3229 
3230     /* Keep statistics */
3231     pimstat.pims_snd_registers_msgs++;
3232     pimstat.pims_snd_registers_bytes += len;
3233 
3234     return 0;
3235 }
3236 
3237 /*
3238  * PIM-SMv2 and PIM-DM messages processing.
3239  * Receives and verifies the PIM control messages, and passes them
3240  * up to the listening socket, using rip_input().
3241  * The only message with special processing is the PIM_REGISTER message
3242  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3243  * is passed to if_simloop().
3244  */
3245 void
3246 pim_input(struct mbuf *m, ...)
3247 {
3248     struct ip *ip = mtod(m, struct ip *);
3249     struct pim *pim;
3250     int minlen;
3251     int datalen;
3252     int ip_tos;
3253     int proto;
3254     int iphlen;
3255     va_list ap;
3256 
3257     va_start(ap, m);
3258     iphlen = va_arg(ap, int);
3259     proto = va_arg(ap, int);
3260     va_end(ap);
3261 
3262     datalen = ntohs(ip->ip_len) - iphlen;
3263 
3264     /* Keep statistics */
3265     pimstat.pims_rcv_total_msgs++;
3266     pimstat.pims_rcv_total_bytes += datalen;
3267 
3268     /*
3269      * Validate lengths
3270      */
3271     if (datalen < PIM_MINLEN) {
3272 	pimstat.pims_rcv_tooshort++;
3273 	log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3274 	    datalen, (u_long)ip->ip_src.s_addr);
3275 	m_freem(m);
3276 	return;
3277     }
3278 
3279     /*
3280      * If the packet is at least as big as a REGISTER, go agead
3281      * and grab the PIM REGISTER header size, to avoid another
3282      * possible m_pullup() later.
3283      *
3284      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
3285      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3286      */
3287     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3288     /*
3289      * Get the IP and PIM headers in contiguous memory, and
3290      * possibly the PIM REGISTER header.
3291      */
3292     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3293 	(m = m_pullup(m, minlen)) == NULL) {
3294 	log(LOG_ERR, "pim_input: m_pullup failure\n");
3295 	return;
3296     }
3297     /* m_pullup() may have given us a new mbuf so reset ip. */
3298     ip = mtod(m, struct ip *);
3299     ip_tos = ip->ip_tos;
3300 
3301     /* adjust mbuf to point to the PIM header */
3302     m->m_data += iphlen;
3303     m->m_len  -= iphlen;
3304     pim = mtod(m, struct pim *);
3305 
3306     /*
3307      * Validate checksum. If PIM REGISTER, exclude the data packet.
3308      *
3309      * XXX: some older PIMv2 implementations don't make this distinction,
3310      * so for compatibility reason perform the checksum over part of the
3311      * message, and if error, then over the whole message.
3312      */
3313     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3314 	/* do nothing, checksum okay */
3315     } else if (in_cksum(m, datalen)) {
3316 	pimstat.pims_rcv_badsum++;
3317 	if (mrtdebug & DEBUG_PIM)
3318 	    log(LOG_DEBUG, "pim_input: invalid checksum");
3319 	m_freem(m);
3320 	return;
3321     }
3322 
3323     /* PIM version check */
3324     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3325 	pimstat.pims_rcv_badversion++;
3326 	log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3327 	    PIM_VT_V(pim->pim_vt), PIM_VERSION);
3328 	m_freem(m);
3329 	return;
3330     }
3331 
3332     /* restore mbuf back to the outer IP */
3333     m->m_data -= iphlen;
3334     m->m_len  += iphlen;
3335 
3336     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3337 	/*
3338 	 * Since this is a REGISTER, we'll make a copy of the register
3339 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3340 	 * routing daemon.
3341 	 */
3342 	int s;
3343 	struct sockaddr_in dst = {
3344 		.sin_len = sizeof(dst),
3345 		.sin_family = AF_INET,
3346 	};
3347 	struct mbuf *mcp;
3348 	struct ip *encap_ip;
3349 	u_int32_t *reghdr;
3350 	struct ifnet *vifp;
3351 
3352 	s = splsoftnet();
3353 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3354 	    splx(s);
3355 	    if (mrtdebug & DEBUG_PIM)
3356 		log(LOG_DEBUG,
3357 		    "pim_input: register vif not set: %d\n", reg_vif_num);
3358 	    m_freem(m);
3359 	    return;
3360 	}
3361 	/* XXX need refcnt? */
3362 	vifp = viftable[reg_vif_num].v_ifp;
3363 	splx(s);
3364 
3365 	/*
3366 	 * Validate length
3367 	 */
3368 	if (datalen < PIM_REG_MINLEN) {
3369 	    pimstat.pims_rcv_tooshort++;
3370 	    pimstat.pims_rcv_badregisters++;
3371 	    log(LOG_ERR,
3372 		"pim_input: register packet size too small %d from %lx\n",
3373 		datalen, (u_long)ip->ip_src.s_addr);
3374 	    m_freem(m);
3375 	    return;
3376 	}
3377 
3378 	reghdr = (u_int32_t *)(pim + 1);
3379 	encap_ip = (struct ip *)(reghdr + 1);
3380 
3381 	if (mrtdebug & DEBUG_PIM) {
3382 	    log(LOG_DEBUG,
3383 		"pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3384 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3385 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3386 		ntohs(encap_ip->ip_len));
3387 	}
3388 
3389 	/* verify the version number of the inner packet */
3390 	if (encap_ip->ip_v != IPVERSION) {
3391 	    pimstat.pims_rcv_badregisters++;
3392 	    if (mrtdebug & DEBUG_PIM) {
3393 		log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3394 		    "of the inner packet\n", encap_ip->ip_v);
3395 	    }
3396 	    m_freem(m);
3397 	    return;
3398 	}
3399 
3400 	/* verify the inner packet is destined to a mcast group */
3401 	if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3402 	    pimstat.pims_rcv_badregisters++;
3403 	    if (mrtdebug & DEBUG_PIM)
3404 		log(LOG_DEBUG,
3405 		    "pim_input: inner packet of register is not "
3406 		    "multicast %lx\n",
3407 		    (u_long)ntohl(encap_ip->ip_dst.s_addr));
3408 	    m_freem(m);
3409 	    return;
3410 	}
3411 
3412 	/* If a NULL_REGISTER, pass it to the daemon */
3413 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3414 	    goto pim_input_to_daemon;
3415 
3416 	/*
3417 	 * Copy the TOS from the outer IP header to the inner IP header.
3418 	 */
3419 	if (encap_ip->ip_tos != ip_tos) {
3420 	    /* Outer TOS -> inner TOS */
3421 	    encap_ip->ip_tos = ip_tos;
3422 	    /* Recompute the inner header checksum. Sigh... */
3423 
3424 	    /* adjust mbuf to point to the inner IP header */
3425 	    m->m_data += (iphlen + PIM_MINLEN);
3426 	    m->m_len  -= (iphlen + PIM_MINLEN);
3427 
3428 	    encap_ip->ip_sum = 0;
3429 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3430 
3431 	    /* restore mbuf to point back to the outer IP header */
3432 	    m->m_data -= (iphlen + PIM_MINLEN);
3433 	    m->m_len  += (iphlen + PIM_MINLEN);
3434 	}
3435 
3436 	/*
3437 	 * Decapsulate the inner IP packet and loopback to forward it
3438 	 * as a normal multicast packet. Also, make a copy of the
3439 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
3440 	 * to pass to the daemon later, so it can take the appropriate
3441 	 * actions (e.g., send back PIM_REGISTER_STOP).
3442 	 * XXX: here m->m_data points to the outer IP header.
3443 	 */
3444 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_DONTWAIT);
3445 	if (mcp == NULL) {
3446 	    log(LOG_ERR,
3447 		"pim_input: pim register: could not copy register head\n");
3448 	    m_freem(m);
3449 	    return;
3450 	}
3451 
3452 	/* Keep statistics */
3453 	/* XXX: registers_bytes include only the encap. mcast pkt */
3454 	pimstat.pims_rcv_registers_msgs++;
3455 	pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3456 
3457 	/*
3458 	 * forward the inner ip packet; point m_data at the inner ip.
3459 	 */
3460 	m_adj(m, iphlen + PIM_MINLEN);
3461 
3462 	if (mrtdebug & DEBUG_PIM) {
3463 	    log(LOG_DEBUG,
3464 		"pim_input: forwarding decapsulated register: "
3465 		"src %lx, dst %lx, vif %d\n",
3466 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3467 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3468 		reg_vif_num);
3469 	}
3470 	/* NB: vifp was collected above; can it change on us? */
3471 	looutput(vifp, m, (struct sockaddr *)&dst, NULL);
3472 
3473 	/* prepare the register head to send to the mrouting daemon */
3474 	m = mcp;
3475     }
3476 
3477 pim_input_to_daemon:
3478     /*
3479      * Pass the PIM message up to the daemon; if it is a Register message,
3480      * pass the 'head' only up to the daemon. This includes the
3481      * outer IP header, PIM header, PIM-Register header and the
3482      * inner IP header.
3483      * XXX: the outer IP header pkt size of a Register is not adjust to
3484      * reflect the fact that the inner multicast data is truncated.
3485      */
3486     rip_input(m, iphlen, proto);
3487 
3488     return;
3489 }
3490 #endif /* PIM */
3491