xref: /netbsd-src/sys/netinet/ip_mroute.c (revision d909946ca08dceb44d7d0f22ec9488679695d976)
1 /*	$NetBSD: ip_mroute.c,v 1.144 2016/08/01 03:15:30 ozaki-r 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 1.144 2016/08/01 03:15:30 ozaki-r Exp $");
97 
98 #ifdef _KERNEL_OPT
99 #include "opt_inet.h"
100 #include "opt_ipsec.h"
101 #include "opt_pim.h"
102 #endif
103 
104 #ifdef PIM
105 #define _PIM_VT 1
106 #endif
107 
108 #include <sys/param.h>
109 #include <sys/systm.h>
110 #include <sys/callout.h>
111 #include <sys/mbuf.h>
112 #include <sys/socket.h>
113 #include <sys/socketvar.h>
114 #include <sys/protosw.h>
115 #include <sys/errno.h>
116 #include <sys/time.h>
117 #include <sys/kernel.h>
118 #include <sys/kmem.h>
119 #include <sys/ioctl.h>
120 #include <sys/syslog.h>
121 
122 #include <net/if.h>
123 #include <net/raw_cb.h>
124 
125 #include <netinet/in.h>
126 #include <netinet/in_var.h>
127 #include <netinet/in_systm.h>
128 #include <netinet/ip.h>
129 #include <netinet/ip_var.h>
130 #include <netinet/in_pcb.h>
131 #include <netinet/udp.h>
132 #include <netinet/igmp.h>
133 #include <netinet/igmp_var.h>
134 #include <netinet/ip_mroute.h>
135 #ifdef PIM
136 #include <netinet/pim.h>
137 #include <netinet/pim_var.h>
138 #endif
139 #include <netinet/ip_encap.h>
140 
141 #ifdef IPSEC
142 #include <netipsec/ipsec.h>
143 #include <netipsec/key.h>
144 #endif
145 
146 #define IP_MULTICASTOPTS 0
147 #define	M_PULLUP(m, len)						 \
148 	do {								 \
149 		if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
150 			(m) = m_pullup((m), (len));			 \
151 	} while (/*CONSTCOND*/ 0)
152 
153 /*
154  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
155  * except for netstat or debugging purposes.
156  */
157 struct socket  *ip_mrouter  = NULL;
158 int		ip_mrtproto = IGMP_DVMRP;    /* for netstat only */
159 
160 #define NO_RTE_FOUND 	0x1
161 #define RTE_FOUND	0x2
162 
163 #define	MFCHASH(a, g)							\
164 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^	\
165 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
166 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
167 u_long	mfchash;
168 
169 u_char		nexpire[MFCTBLSIZ];
170 struct vif	viftable[MAXVIFS];
171 struct mrtstat	mrtstat;
172 u_int		mrtdebug = 0;	  /* debug level 	*/
173 #define		DEBUG_MFC	0x02
174 #define		DEBUG_FORWARD	0x04
175 #define		DEBUG_EXPIRE	0x08
176 #define		DEBUG_XMIT	0x10
177 #define		DEBUG_PIM	0x20
178 
179 #define		VIFI_INVALID	((vifi_t) -1)
180 
181 u_int       	tbfdebug = 0;     /* tbf debug level 	*/
182 #ifdef RSVP_ISI
183 u_int		rsvpdebug = 0;	  /* rsvp debug level   */
184 #define	RSVP_DPRINTF(a)	do if (rsvpdebug) printf a; while (/*CONSTCOND*/0)
185 extern struct socket *ip_rsvpd;
186 extern int rsvp_on;
187 #else
188 #define	RSVP_DPRINTF(a)	do {} while (/*CONSTCOND*/0)
189 #endif /* RSVP_ISI */
190 
191 /* vif attachment using sys/netinet/ip_encap.c */
192 static void vif_input(struct mbuf *, int, int);
193 static int vif_encapcheck(struct mbuf *, int, int, void *);
194 
195 static const struct encapsw vif_encapsw = {
196 	.encapsw4 = {
197 		.pr_input	= vif_input,
198 		.pr_ctlinput	= NULL,
199 	}
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 ifnet *ifp;
789 	int error, s;
790 	struct sockaddr_in sin;
791 
792 	if (vifcp->vifc_vifi >= MAXVIFS)
793 		return (EINVAL);
794 	if (in_nullhost(vifcp->vifc_lcl_addr))
795 		return (EADDRNOTAVAIL);
796 
797 	vifp = &viftable[vifcp->vifc_vifi];
798 	if (!in_nullhost(vifp->v_lcl_addr))
799 		return (EADDRINUSE);
800 
801 	/* Find the interface with an address in AF_INET family. */
802 #ifdef PIM
803 	if (vifcp->vifc_flags & VIFF_REGISTER) {
804 		/*
805 		 * XXX: Because VIFF_REGISTER does not really need a valid
806 		 * local interface (e.g. it could be 127.0.0.2), we don't
807 		 * check its address.
808 		 */
809 	    ifp = NULL;
810 	} else
811 #endif
812 	{
813 		struct ifaddr *ifa;
814 
815 		sockaddr_in_init(&sin, &vifcp->vifc_lcl_addr, 0);
816 		s = pserialize_read_enter();
817 		ifa = ifa_ifwithaddr(sintosa(&sin));
818 		if (ifa == NULL) {
819 			pserialize_read_exit(s);
820 			return EADDRNOTAVAIL;
821 		}
822 		ifp = ifa->ifa_ifp;
823 		/* FIXME NOMPSAFE */
824 		pserialize_read_exit(s);
825 	}
826 
827 	if (vifcp->vifc_flags & VIFF_TUNNEL) {
828 		if (vifcp->vifc_flags & VIFF_SRCRT) {
829 			log(LOG_ERR, "source routed tunnels not supported\n");
830 			return (EOPNOTSUPP);
831 		}
832 
833 		/* attach this vif to decapsulator dispatch table */
834 		/*
835 		 * XXX Use addresses in registration so that matching
836 		 * can be done with radix tree in decapsulator.  But,
837 		 * we need to check inner header for multicast, so
838 		 * this requires both radix tree lookup and then a
839 		 * function to check, and this is not supported yet.
840 		 */
841 		error = encap_lock_enter();
842 		if (error)
843 			return error;
844 		vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
845 		    vif_encapcheck, &vif_encapsw, vifp);
846 		encap_lock_exit();
847 		if (!vifp->v_encap_cookie)
848 			return (EINVAL);
849 
850 		/* Create a fake encapsulation interface. */
851 		ifp = malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK|M_ZERO);
852 		snprintf(ifp->if_xname, sizeof(ifp->if_xname),
853 			 "mdecap%d", vifcp->vifc_vifi);
854 
855 		/* Prepare cached route entry. */
856 		memset(&vifp->v_route, 0, sizeof(vifp->v_route));
857 #ifdef PIM
858 	} else if (vifcp->vifc_flags & VIFF_REGISTER) {
859 		ifp = &multicast_register_if;
860 		if (mrtdebug)
861 			log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
862 			    (void *)ifp);
863 		if (reg_vif_num == VIFI_INVALID) {
864 			memset(ifp, 0, sizeof(*ifp));
865 			snprintf(ifp->if_xname, sizeof(ifp->if_xname),
866 				 "register_vif");
867 			ifp->if_flags = IFF_LOOPBACK;
868 			memset(&vifp->v_route, 0, sizeof(vifp->v_route));
869 			reg_vif_num = vifcp->vifc_vifi;
870 		}
871 #endif
872 	} else {
873 		/* Make sure the interface supports multicast. */
874 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
875 			return (EOPNOTSUPP);
876 
877 		/* Enable promiscuous reception of all IP multicasts. */
878 		sockaddr_in_init(&sin, &zeroin_addr, 0);
879 		error = if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin));
880 		if (error)
881 			return (error);
882 	}
883 
884 	s = splsoftnet();
885 
886 	/* Define parameters for the tbf structure. */
887 	vifp->tbf_q = NULL;
888 	vifp->tbf_t = &vifp->tbf_q;
889 	microtime(&vifp->tbf_last_pkt_t);
890 	vifp->tbf_n_tok = 0;
891 	vifp->tbf_q_len = 0;
892 	vifp->tbf_max_q_len = MAXQSIZE;
893 
894 	vifp->v_flags = vifcp->vifc_flags;
895 	vifp->v_threshold = vifcp->vifc_threshold;
896 	/* scaling up here allows division by 1024 in critical code */
897 	vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
898 	vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
899 	vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
900 	vifp->v_ifp = ifp;
901 	/* Initialize per vif pkt counters. */
902 	vifp->v_pkt_in = 0;
903 	vifp->v_pkt_out = 0;
904 	vifp->v_bytes_in = 0;
905 	vifp->v_bytes_out = 0;
906 
907 	callout_init(&vifp->v_repq_ch, 0);
908 
909 #ifdef RSVP_ISI
910 	vifp->v_rsvp_on = 0;
911 	vifp->v_rsvpd = NULL;
912 #endif /* RSVP_ISI */
913 
914 	splx(s);
915 
916 	/* Adjust numvifs up if the vifi is higher than numvifs. */
917 	if (numvifs <= vifcp->vifc_vifi)
918 		numvifs = vifcp->vifc_vifi + 1;
919 
920 	if (mrtdebug)
921 		log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
922 		    vifcp->vifc_vifi,
923 		    ntohl(vifcp->vifc_lcl_addr.s_addr),
924 		    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
925 		    ntohl(vifcp->vifc_rmt_addr.s_addr),
926 		    vifcp->vifc_threshold,
927 		    vifcp->vifc_rate_limit);
928 
929 	return (0);
930 }
931 
932 void
933 reset_vif(struct vif *vifp)
934 {
935 	struct mbuf *m, *n;
936 	struct ifnet *ifp;
937 	struct sockaddr_in sin;
938 
939 	callout_stop(&vifp->v_repq_ch);
940 
941 	/* detach this vif from decapsulator dispatch table */
942 	encap_lock_enter();
943 	encap_detach(vifp->v_encap_cookie);
944 	encap_lock_exit();
945 	vifp->v_encap_cookie = NULL;
946 
947 	/*
948 	 * Free packets queued at the interface
949 	 */
950 	for (m = vifp->tbf_q; m != NULL; m = n) {
951 		n = m->m_nextpkt;
952 		m_freem(m);
953 	}
954 
955 	if (vifp->v_flags & VIFF_TUNNEL)
956 		free(vifp->v_ifp, M_MRTABLE);
957 	else if (vifp->v_flags & VIFF_REGISTER) {
958 #ifdef PIM
959 		reg_vif_num = VIFI_INVALID;
960 #endif
961 	} else {
962 		sockaddr_in_init(&sin, &zeroin_addr, 0);
963 		ifp = vifp->v_ifp;
964 		if_mcast_op(ifp, SIOCDELMULTI, sintosa(&sin));
965 	}
966 	memset((void *)vifp, 0, sizeof(*vifp));
967 }
968 
969 /*
970  * Delete a vif from the vif table
971  */
972 static int
973 del_vif(vifi_t *vifip)
974 {
975 	struct vif *vifp;
976 	vifi_t vifi;
977 	int s;
978 
979 	if (*vifip >= numvifs)
980 		return (EINVAL);
981 
982 	vifp = &viftable[*vifip];
983 	if (in_nullhost(vifp->v_lcl_addr))
984 		return (EADDRNOTAVAIL);
985 
986 	s = splsoftnet();
987 
988 	reset_vif(vifp);
989 
990 	/* Adjust numvifs down */
991 	for (vifi = numvifs; vifi > 0; vifi--)
992 		if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
993 			break;
994 	numvifs = vifi;
995 
996 	splx(s);
997 
998 	if (mrtdebug)
999 		log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
1000 
1001 	return (0);
1002 }
1003 
1004 /*
1005  * update an mfc entry without resetting counters and S,G addresses.
1006  */
1007 static void
1008 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1009 {
1010 	int i;
1011 
1012 	rt->mfc_parent = mfccp->mfcc_parent;
1013 	for (i = 0; i < numvifs; i++) {
1014 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1015 		rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1016 			MRT_MFC_FLAGS_ALL;
1017 	}
1018 	/* set the RP address */
1019 	if (mrt_api_config & MRT_MFC_RP)
1020 		rt->mfc_rp = mfccp->mfcc_rp;
1021 	else
1022 		rt->mfc_rp = zeroin_addr;
1023 }
1024 
1025 /*
1026  * fully initialize an mfc entry from the parameter.
1027  */
1028 static void
1029 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1030 {
1031 	rt->mfc_origin     = mfccp->mfcc_origin;
1032 	rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1033 
1034 	update_mfc_params(rt, mfccp);
1035 
1036 	/* initialize pkt counters per src-grp */
1037 	rt->mfc_pkt_cnt    = 0;
1038 	rt->mfc_byte_cnt   = 0;
1039 	rt->mfc_wrong_if   = 0;
1040 	timerclear(&rt->mfc_last_assert);
1041 }
1042 
1043 static void
1044 expire_mfc(struct mfc *rt)
1045 {
1046 	struct rtdetq *rte, *nrte;
1047 
1048 	free_bw_list(rt->mfc_bw_meter);
1049 
1050 	for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1051 		nrte = rte->next;
1052 		m_freem(rte->m);
1053 		free(rte, M_MRTABLE);
1054 	}
1055 
1056 	LIST_REMOVE(rt, mfc_hash);
1057 	free(rt, M_MRTABLE);
1058 }
1059 
1060 /*
1061  * Add an mfc entry
1062  */
1063 static int
1064 add_mfc(struct sockopt *sopt)
1065 {
1066 	struct mfcctl2 mfcctl2;
1067 	struct mfcctl2 *mfccp;
1068 	struct mfc *rt;
1069 	u_int32_t hash = 0;
1070 	struct rtdetq *rte, *nrte;
1071 	u_short nstl;
1072 	int s;
1073 	int error;
1074 
1075 	/*
1076 	 * select data size depending on API version.
1077 	 */
1078 	mfccp = &mfcctl2;
1079 	memset(&mfcctl2, 0, sizeof(mfcctl2));
1080 
1081 	if (mrt_api_config & MRT_API_FLAGS_ALL)
1082 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1083 	else
1084 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1085 
1086 	if (error)
1087 		return (error);
1088 
1089 	s = splsoftnet();
1090 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1091 
1092 	/* If an entry already exists, just update the fields */
1093 	if (rt) {
1094 		if (mrtdebug & DEBUG_MFC)
1095 			log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1096 			    ntohl(mfccp->mfcc_origin.s_addr),
1097 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1098 			    mfccp->mfcc_parent);
1099 
1100 		update_mfc_params(rt, mfccp);
1101 
1102 		splx(s);
1103 		return (0);
1104 	}
1105 
1106 	/*
1107 	 * Find the entry for which the upcall was made and update
1108 	 */
1109 	nstl = 0;
1110 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1111 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1112 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1113 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1114 		    rt->mfc_stall != NULL) {
1115 			if (nstl++)
1116 				log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
1117 				    "multiple kernel entries",
1118 				    ntohl(mfccp->mfcc_origin.s_addr),
1119 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1120 				    mfccp->mfcc_parent, rt->mfc_stall);
1121 
1122 			if (mrtdebug & DEBUG_MFC)
1123 				log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n",
1124 				    ntohl(mfccp->mfcc_origin.s_addr),
1125 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1126 				    mfccp->mfcc_parent, rt->mfc_stall);
1127 
1128 			rte = rt->mfc_stall;
1129 			init_mfc_params(rt, mfccp);
1130 			rt->mfc_stall = NULL;
1131 
1132 			rt->mfc_expire = 0; /* Don't clean this guy up */
1133 			nexpire[hash]--;
1134 
1135 			/* free packets Qed at the end of this entry */
1136 			for (; rte != NULL; rte = nrte) {
1137 				nrte = rte->next;
1138 				if (rte->ifp) {
1139 #ifdef RSVP_ISI
1140 					ip_mdq(rte->m, rte->ifp, rt, -1);
1141 #else
1142 					ip_mdq(rte->m, rte->ifp, rt);
1143 #endif /* RSVP_ISI */
1144 				}
1145 				m_freem(rte->m);
1146 #ifdef UPCALL_TIMING
1147 				collate(&rte->t);
1148 #endif /* UPCALL_TIMING */
1149 				free(rte, M_MRTABLE);
1150 			}
1151 		}
1152 	}
1153 
1154 	/*
1155 	 * It is possible that an entry is being inserted without an upcall
1156 	 */
1157 	if (nstl == 0) {
1158 		/*
1159 		 * No mfc; make a new one
1160 		 */
1161 		if (mrtdebug & DEBUG_MFC)
1162 			log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1163 			    ntohl(mfccp->mfcc_origin.s_addr),
1164 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1165 			    mfccp->mfcc_parent);
1166 
1167 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1168 			if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1169 			    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1170 				init_mfc_params(rt, mfccp);
1171 				if (rt->mfc_expire)
1172 					nexpire[hash]--;
1173 				rt->mfc_expire = 0;
1174 				break; /* XXX */
1175 			}
1176 		}
1177 		if (rt == NULL) {	/* no upcall, so make a new entry */
1178 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1179 						  M_NOWAIT);
1180 			if (rt == NULL) {
1181 				splx(s);
1182 				return (ENOBUFS);
1183 			}
1184 
1185 			init_mfc_params(rt, mfccp);
1186 			rt->mfc_expire	= 0;
1187 			rt->mfc_stall	= NULL;
1188 			rt->mfc_bw_meter = NULL;
1189 
1190 			/* insert new entry at head of hash chain */
1191 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1192 		}
1193 	}
1194 
1195 	splx(s);
1196 	return (0);
1197 }
1198 
1199 #ifdef UPCALL_TIMING
1200 /*
1201  * collect delay statistics on the upcalls
1202  */
1203 static void
1204 collate(struct timeval *t)
1205 {
1206 	u_int32_t d;
1207 	struct timeval tp;
1208 	u_int32_t delta;
1209 
1210 	microtime(&tp);
1211 
1212 	if (timercmp(t, &tp, <)) {
1213 		TV_DELTA(tp, *t, delta);
1214 
1215 		d = delta >> 10;
1216 		if (d > 50)
1217 			d = 50;
1218 
1219 		++upcall_data[d];
1220 	}
1221 }
1222 #endif /* UPCALL_TIMING */
1223 
1224 /*
1225  * Delete an mfc entry
1226  */
1227 static int
1228 del_mfc(struct sockopt *sopt)
1229 {
1230 	struct mfcctl2 mfcctl2;
1231 	struct mfcctl2 *mfccp;
1232 	struct mfc *rt;
1233 	int s;
1234 	int error;
1235 
1236 	/*
1237 	 * XXX: for deleting MFC entries the information in entries
1238 	 * of size "struct mfcctl" is sufficient.
1239 	 */
1240 
1241 	mfccp = &mfcctl2;
1242 	memset(&mfcctl2, 0, sizeof(mfcctl2));
1243 
1244 	error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1245 	if (error) {
1246 		/* Try with the size of mfcctl2. */
1247 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1248 		if (error)
1249 			return (error);
1250 	}
1251 
1252 	if (mrtdebug & DEBUG_MFC)
1253 		log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1254 		    ntohl(mfccp->mfcc_origin.s_addr),
1255 		    ntohl(mfccp->mfcc_mcastgrp.s_addr));
1256 
1257 	s = splsoftnet();
1258 
1259 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1260 	if (rt == NULL) {
1261 		splx(s);
1262 		return (EADDRNOTAVAIL);
1263 	}
1264 
1265 	/*
1266 	 * free the bw_meter entries
1267 	 */
1268 	free_bw_list(rt->mfc_bw_meter);
1269 	rt->mfc_bw_meter = NULL;
1270 
1271 	LIST_REMOVE(rt, mfc_hash);
1272 	free(rt, M_MRTABLE);
1273 
1274 	splx(s);
1275 	return (0);
1276 }
1277 
1278 static int
1279 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1280 {
1281 	if (s) {
1282 		if (sbappendaddr(&s->so_rcv, sintosa(src), mm, NULL) != 0) {
1283 			sorwakeup(s);
1284 			return (0);
1285 		}
1286 	}
1287 	m_freem(mm);
1288 	return (-1);
1289 }
1290 
1291 /*
1292  * IP multicast forwarding function. This function assumes that the packet
1293  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1294  * pointed to by "ifp", and the packet is to be relayed to other networks
1295  * that have members of the packet's destination IP multicast group.
1296  *
1297  * The packet is returned unscathed to the caller, unless it is
1298  * erroneous, in which case a non-zero return value tells the caller to
1299  * discard it.
1300  */
1301 
1302 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1303 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1304 
1305 int
1306 #ifdef RSVP_ISI
1307 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo)
1308 #else
1309 ip_mforward(struct mbuf *m, struct ifnet *ifp)
1310 #endif /* RSVP_ISI */
1311 {
1312 	struct ip *ip = mtod(m, struct ip *);
1313 	struct mfc *rt;
1314 	static int srctun = 0;
1315 	struct mbuf *mm;
1316 	struct sockaddr_in sin;
1317 	int s;
1318 	vifi_t vifi;
1319 
1320 	if (mrtdebug & DEBUG_FORWARD)
1321 		log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1322 		    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1323 
1324 	if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1325 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1326 		/*
1327 		 * Packet arrived via a physical interface or
1328 		 * an encapsulated tunnel or a register_vif.
1329 		 */
1330 	} else {
1331 		/*
1332 		 * Packet arrived through a source-route tunnel.
1333 		 * Source-route tunnels are no longer supported.
1334 		 */
1335 		if ((srctun++ % 1000) == 0)
1336 			log(LOG_ERR,
1337 			    "ip_mforward: received source-routed packet from %x\n",
1338 			    ntohl(ip->ip_src.s_addr));
1339 
1340 		return (1);
1341 	}
1342 
1343 	/*
1344 	 * Clear any in-bound checksum flags for this packet.
1345 	 */
1346 	m->m_pkthdr.csum_flags = 0;
1347 
1348 #ifdef RSVP_ISI
1349 	if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1350 		if (ip->ip_ttl < MAXTTL)
1351 			ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1352 		if (ip->ip_p == IPPROTO_RSVP) {
1353 			struct vif *vifp = viftable + vifi;
1354 			RSVP_DPRINTF(("%s: Sending IPPROTO_RSVP from %x to %x"
1355 			    " on vif %d (%s%s)\n", __func__,
1356 			    ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1357 			    (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1358 			    vifp->v_ifp->if_xname));
1359 		}
1360 		return (ip_mdq(m, ifp, NULL, vifi));
1361 	}
1362 	if (ip->ip_p == IPPROTO_RSVP) {
1363 		RSVP_DPRINTF(("%s: Warning: IPPROTO_RSVP from %x to %x"
1364 		    " without vif option\n", __func__,
1365 		    ntohl(ip->ip_src), ntohl(ip->ip_dst));
1366 	}
1367 #endif /* RSVP_ISI */
1368 
1369 	/*
1370 	 * Don't forward a packet with time-to-live of zero or one,
1371 	 * or a packet destined to a local-only group.
1372 	 */
1373 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1374 		return (0);
1375 
1376 	/*
1377 	 * Determine forwarding vifs from the forwarding cache table
1378 	 */
1379 	s = splsoftnet();
1380 	++mrtstat.mrts_mfc_lookups;
1381 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1382 
1383 	/* Entry exists, so forward if necessary */
1384 	if (rt != NULL) {
1385 		splx(s);
1386 #ifdef RSVP_ISI
1387 		return (ip_mdq(m, ifp, rt, -1));
1388 #else
1389 		return (ip_mdq(m, ifp, rt));
1390 #endif /* RSVP_ISI */
1391 	} else {
1392 		/*
1393 		 * If we don't have a route for packet's origin,
1394 		 * Make a copy of the packet & send message to routing daemon
1395 		 */
1396 
1397 		struct mbuf *mb0;
1398 		struct rtdetq *rte;
1399 		u_int32_t hash;
1400 		int hlen = ip->ip_hl << 2;
1401 #ifdef UPCALL_TIMING
1402 		struct timeval tp;
1403 
1404 		microtime(&tp);
1405 #endif /* UPCALL_TIMING */
1406 
1407 		++mrtstat.mrts_mfc_misses;
1408 
1409 		mrtstat.mrts_no_route++;
1410 		if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1411 			log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1412 			    ntohl(ip->ip_src.s_addr),
1413 			    ntohl(ip->ip_dst.s_addr));
1414 
1415 		/*
1416 		 * Allocate mbufs early so that we don't do extra work if we are
1417 		 * just going to fail anyway.  Make sure to pullup the header so
1418 		 * that other people can't step on it.
1419 		 */
1420 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1421 					      M_NOWAIT);
1422 		if (rte == NULL) {
1423 			splx(s);
1424 			return (ENOBUFS);
1425 		}
1426 		mb0 = m_copypacket(m, M_DONTWAIT);
1427 		M_PULLUP(mb0, hlen);
1428 		if (mb0 == NULL) {
1429 			free(rte, M_MRTABLE);
1430 			splx(s);
1431 			return (ENOBUFS);
1432 		}
1433 
1434 		/* is there an upcall waiting for this flow? */
1435 		hash = MFCHASH(ip->ip_src, ip->ip_dst);
1436 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1437 			if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1438 			    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1439 			    rt->mfc_stall != NULL)
1440 				break;
1441 		}
1442 
1443 		if (rt == NULL) {
1444 			int i;
1445 			struct igmpmsg *im;
1446 
1447 			/*
1448 			 * Locate the vifi for the incoming interface for
1449 			 * this packet.
1450 			 * If none found, drop packet.
1451 			 */
1452 			for (vifi = 0; vifi < numvifs &&
1453 				 viftable[vifi].v_ifp != ifp; vifi++)
1454 				;
1455 			if (vifi >= numvifs) /* vif not found, drop packet */
1456 				goto non_fatal;
1457 
1458 			/* no upcall, so make a new entry */
1459 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1460 						  M_NOWAIT);
1461 			if (rt == NULL)
1462 				goto fail;
1463 
1464 			/*
1465 			 * Make a copy of the header to send to the user level
1466 			 * process
1467 			 */
1468 			mm = m_copym(m, 0, hlen, M_DONTWAIT);
1469 			M_PULLUP(mm, hlen);
1470 			if (mm == NULL)
1471 				goto fail1;
1472 
1473 			/*
1474 			 * Send message to routing daemon to install
1475 			 * a route into the kernel table
1476 			 */
1477 
1478 			im = mtod(mm, struct igmpmsg *);
1479 			im->im_msgtype = IGMPMSG_NOCACHE;
1480 			im->im_mbz = 0;
1481 			im->im_vif = vifi;
1482 
1483 			mrtstat.mrts_upcalls++;
1484 
1485 			sockaddr_in_init(&sin, &ip->ip_src, 0);
1486 			if (socket_send(ip_mrouter, mm, &sin) < 0) {
1487 				log(LOG_WARNING,
1488 				    "ip_mforward: ip_mrouter socket queue full\n");
1489 				++mrtstat.mrts_upq_sockfull;
1490 			fail1:
1491 				free(rt, M_MRTABLE);
1492 			fail:
1493 				free(rte, M_MRTABLE);
1494 				m_freem(mb0);
1495 				splx(s);
1496 				return (ENOBUFS);
1497 			}
1498 
1499 			/* insert new entry at head of hash chain */
1500 			rt->mfc_origin = ip->ip_src;
1501 			rt->mfc_mcastgrp = ip->ip_dst;
1502 			rt->mfc_pkt_cnt = 0;
1503 			rt->mfc_byte_cnt = 0;
1504 			rt->mfc_wrong_if = 0;
1505 			rt->mfc_expire = UPCALL_EXPIRE;
1506 			nexpire[hash]++;
1507 			for (i = 0; i < numvifs; i++) {
1508 				rt->mfc_ttls[i] = 0;
1509 				rt->mfc_flags[i] = 0;
1510 			}
1511 			rt->mfc_parent = -1;
1512 
1513 			/* clear the RP address */
1514 			rt->mfc_rp = zeroin_addr;
1515 
1516 			rt->mfc_bw_meter = NULL;
1517 
1518 			/* link into table */
1519 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1520 			/* Add this entry to the end of the queue */
1521 			rt->mfc_stall = rte;
1522 		} else {
1523 			/* determine if q has overflowed */
1524 			struct rtdetq **p;
1525 			int npkts = 0;
1526 
1527 			/*
1528 			 * XXX ouch! we need to append to the list, but we
1529 			 * only have a pointer to the front, so we have to
1530 			 * scan the entire list every time.
1531 			 */
1532 			for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1533 				if (++npkts > MAX_UPQ) {
1534 					mrtstat.mrts_upq_ovflw++;
1535 				non_fatal:
1536 					free(rte, M_MRTABLE);
1537 					m_freem(mb0);
1538 					splx(s);
1539 					return (0);
1540 				}
1541 
1542 			/* Add this entry to the end of the queue */
1543 			*p = rte;
1544 		}
1545 
1546 		rte->next = NULL;
1547 		rte->m = mb0;
1548 		rte->ifp = ifp;
1549 #ifdef UPCALL_TIMING
1550 		rte->t = tp;
1551 #endif /* UPCALL_TIMING */
1552 
1553 		splx(s);
1554 
1555 		return (0);
1556 	}
1557 }
1558 
1559 
1560 /*ARGSUSED*/
1561 static void
1562 expire_upcalls(void *v)
1563 {
1564 	int i;
1565 	int s;
1566 
1567 	s = splsoftnet();
1568 
1569 	for (i = 0; i < MFCTBLSIZ; i++) {
1570 		struct mfc *rt, *nrt;
1571 
1572 		if (nexpire[i] == 0)
1573 			continue;
1574 
1575 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1576 			nrt = LIST_NEXT(rt, mfc_hash);
1577 
1578 			if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1579 				continue;
1580 			nexpire[i]--;
1581 
1582 			/*
1583 			 * free the bw_meter entries
1584 			 */
1585 			while (rt->mfc_bw_meter != NULL) {
1586 				struct bw_meter *x = rt->mfc_bw_meter;
1587 
1588 				rt->mfc_bw_meter = x->bm_mfc_next;
1589 				kmem_free(x, sizeof(*x));
1590 			}
1591 
1592 			++mrtstat.mrts_cache_cleanups;
1593 			if (mrtdebug & DEBUG_EXPIRE)
1594 				log(LOG_DEBUG,
1595 				    "expire_upcalls: expiring (%x %x)\n",
1596 				    ntohl(rt->mfc_origin.s_addr),
1597 				    ntohl(rt->mfc_mcastgrp.s_addr));
1598 
1599 			expire_mfc(rt);
1600 		}
1601 	}
1602 
1603 	splx(s);
1604 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1605 	    expire_upcalls, NULL);
1606 }
1607 
1608 /*
1609  * Packet forwarding routine once entry in the cache is made
1610  */
1611 static int
1612 #ifdef RSVP_ISI
1613 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1614 #else
1615 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1616 #endif /* RSVP_ISI */
1617 {
1618 	struct ip  *ip = mtod(m, struct ip *);
1619 	vifi_t vifi;
1620 	struct vif *vifp;
1621 	struct sockaddr_in sin;
1622 	int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1623 
1624 /*
1625  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1626  * input, they shouldn't get counted on output, so statistics keeping is
1627  * separate.
1628  */
1629 #define MC_SEND(ip, vifp, m) do {					\
1630 	if ((vifp)->v_flags & VIFF_TUNNEL)				\
1631 		encap_send((ip), (vifp), (m));				\
1632 	else								\
1633 		phyint_send((ip), (vifp), (m));				\
1634 } while (/*CONSTCOND*/ 0)
1635 
1636 #ifdef RSVP_ISI
1637 	/*
1638 	 * If xmt_vif is not -1, send on only the requested vif.
1639 	 *
1640 	 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1641 	 */
1642 	if (xmt_vif < numvifs) {
1643 #ifdef PIM
1644 		if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1645 			pim_register_send(ip, viftable + xmt_vif, m, rt);
1646 		else
1647 #endif
1648 		MC_SEND(ip, viftable + xmt_vif, m);
1649 		return (1);
1650 	}
1651 #endif /* RSVP_ISI */
1652 
1653 	/*
1654 	 * Don't forward if it didn't arrive from the parent vif for its origin.
1655 	 */
1656 	vifi = rt->mfc_parent;
1657 	if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1658 		/* came in the wrong interface */
1659 		if (mrtdebug & DEBUG_FORWARD)
1660 			log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1661 			    ifp, vifi,
1662 			    vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1663 		++mrtstat.mrts_wrong_if;
1664 		++rt->mfc_wrong_if;
1665 		/*
1666 		 * If we are doing PIM assert processing, send a message
1667 		 * to the routing daemon.
1668 		 *
1669 		 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1670 		 * can complete the SPT switch, regardless of the type
1671 		 * of the iif (broadcast media, GRE tunnel, etc).
1672 		 */
1673 		if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1674 			struct timeval now;
1675 			u_int32_t delta;
1676 
1677 #ifdef PIM
1678 			if (ifp == &multicast_register_if)
1679 				pimstat.pims_rcv_registers_wrongiif++;
1680 #endif
1681 
1682 			/* Get vifi for the incoming packet */
1683 			for (vifi = 0;
1684 			     vifi < numvifs && viftable[vifi].v_ifp != ifp;
1685 			     vifi++)
1686 			    ;
1687 			if (vifi >= numvifs) {
1688 				/* The iif is not found: ignore the packet. */
1689 				return (0);
1690 			}
1691 
1692 			if (rt->mfc_flags[vifi] &
1693 			    MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1694 				/* WRONGVIF disabled: ignore the packet */
1695 				return (0);
1696 			}
1697 
1698 			microtime(&now);
1699 
1700 			TV_DELTA(rt->mfc_last_assert, now, delta);
1701 
1702 			if (delta > ASSERT_MSG_TIME) {
1703 				struct igmpmsg *im;
1704 				int hlen = ip->ip_hl << 2;
1705 				struct mbuf *mm =
1706 				    m_copym(m, 0, hlen, M_DONTWAIT);
1707 
1708 				M_PULLUP(mm, hlen);
1709 				if (mm == NULL)
1710 					return (ENOBUFS);
1711 
1712 				rt->mfc_last_assert = now;
1713 
1714 				im = mtod(mm, struct igmpmsg *);
1715 				im->im_msgtype	= IGMPMSG_WRONGVIF;
1716 				im->im_mbz	= 0;
1717 				im->im_vif	= vifi;
1718 
1719 				mrtstat.mrts_upcalls++;
1720 
1721 				sockaddr_in_init(&sin, &im->im_src, 0);
1722 				if (socket_send(ip_mrouter, mm, &sin) < 0) {
1723 					log(LOG_WARNING,
1724 					    "ip_mforward: ip_mrouter socket queue full\n");
1725 					++mrtstat.mrts_upq_sockfull;
1726 					return (ENOBUFS);
1727 				}
1728 			}
1729 		}
1730 		return (0);
1731 	}
1732 
1733 	/* If I sourced this packet, it counts as output, else it was input. */
1734 	if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1735 		viftable[vifi].v_pkt_out++;
1736 		viftable[vifi].v_bytes_out += plen;
1737 	} else {
1738 		viftable[vifi].v_pkt_in++;
1739 		viftable[vifi].v_bytes_in += plen;
1740 	}
1741 	rt->mfc_pkt_cnt++;
1742 	rt->mfc_byte_cnt += plen;
1743 
1744 	/*
1745 	 * For each vif, decide if a copy of the packet should be forwarded.
1746 	 * Forward if:
1747 	 *		- the ttl exceeds the vif's threshold
1748 	 *		- there are group members downstream on interface
1749 	 */
1750 	for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1751 		if ((rt->mfc_ttls[vifi] > 0) &&
1752 			(ip->ip_ttl > rt->mfc_ttls[vifi])) {
1753 			vifp->v_pkt_out++;
1754 			vifp->v_bytes_out += plen;
1755 #ifdef PIM
1756 			if (vifp->v_flags & VIFF_REGISTER)
1757 				pim_register_send(ip, vifp, m, rt);
1758 			else
1759 #endif
1760 			MC_SEND(ip, vifp, m);
1761 		}
1762 
1763 	/*
1764 	 * Perform upcall-related bw measuring.
1765 	 */
1766 	if (rt->mfc_bw_meter != NULL) {
1767 		struct bw_meter *x;
1768 		struct timeval now;
1769 
1770 		microtime(&now);
1771 		for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1772 			bw_meter_receive_packet(x, plen, &now);
1773 	}
1774 
1775 	return (0);
1776 }
1777 
1778 #ifdef RSVP_ISI
1779 /*
1780  * check if a vif number is legal/ok. This is used by ip_output.
1781  */
1782 int
1783 legal_vif_num(int vif)
1784 {
1785 	if (vif >= 0 && vif < numvifs)
1786 		return (1);
1787 	else
1788 		return (0);
1789 }
1790 #endif /* RSVP_ISI */
1791 
1792 static void
1793 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1794 {
1795 	struct mbuf *mb_copy;
1796 	int hlen = ip->ip_hl << 2;
1797 
1798 	/*
1799 	 * Make a new reference to the packet; make sure that
1800 	 * the IP header is actually copied, not just referenced,
1801 	 * so that ip_output() only scribbles on the copy.
1802 	 */
1803 	mb_copy = m_copypacket(m, M_DONTWAIT);
1804 	M_PULLUP(mb_copy, hlen);
1805 	if (mb_copy == NULL)
1806 		return;
1807 
1808 	if (vifp->v_rate_limit <= 0)
1809 		tbf_send_packet(vifp, mb_copy);
1810 	else
1811 		tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1812 		    ntohs(ip->ip_len));
1813 }
1814 
1815 static void
1816 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1817 {
1818 	struct mbuf *mb_copy;
1819 	struct ip *ip_copy;
1820 	int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1821 
1822 	/* Take care of delayed checksums */
1823 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1824 		in_delayed_cksum(m);
1825 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1826 	}
1827 
1828 	/*
1829 	 * copy the old packet & pullup its IP header into the
1830 	 * new mbuf so we can modify it.  Try to fill the new
1831 	 * mbuf since if we don't the ethernet driver will.
1832 	 */
1833 	MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1834 	if (mb_copy == NULL)
1835 		return;
1836 	mb_copy->m_data += max_linkhdr;
1837 	mb_copy->m_pkthdr.len = len;
1838 	mb_copy->m_len = sizeof(multicast_encap_iphdr);
1839 
1840 	if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
1841 		m_freem(mb_copy);
1842 		return;
1843 	}
1844 	i = MHLEN - max_linkhdr;
1845 	if (i > len)
1846 		i = len;
1847 	mb_copy = m_pullup(mb_copy, i);
1848 	if (mb_copy == NULL)
1849 		return;
1850 
1851 	/*
1852 	 * fill in the encapsulating IP header.
1853 	 */
1854 	ip_copy = mtod(mb_copy, struct ip *);
1855 	*ip_copy = multicast_encap_iphdr;
1856 	if (len < IP_MINFRAGSIZE)
1857 		ip_copy->ip_id = 0;
1858 	else
1859 		ip_copy->ip_id = ip_newid(NULL);
1860 	ip_copy->ip_len = htons(len);
1861 	ip_copy->ip_src = vifp->v_lcl_addr;
1862 	ip_copy->ip_dst = vifp->v_rmt_addr;
1863 
1864 	/*
1865 	 * turn the encapsulated IP header back into a valid one.
1866 	 */
1867 	ip = (struct ip *)((char *)ip_copy + sizeof(multicast_encap_iphdr));
1868 	--ip->ip_ttl;
1869 	ip->ip_sum = 0;
1870 	mb_copy->m_data += sizeof(multicast_encap_iphdr);
1871 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1872 	mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1873 
1874 	if (vifp->v_rate_limit <= 0)
1875 		tbf_send_packet(vifp, mb_copy);
1876 	else
1877 		tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1878 }
1879 
1880 /*
1881  * De-encapsulate a packet and feed it back through ip input.
1882  */
1883 static void
1884 vif_input(struct mbuf *m, int off, int proto)
1885 {
1886 	struct vif *vifp;
1887 
1888 	vifp = (struct vif *)encap_getarg(m);
1889 	if (!vifp || proto != ENCAP_PROTO) {
1890 		m_freem(m);
1891 		mrtstat.mrts_bad_tunnel++;
1892 		return;
1893 	}
1894 
1895 	m_adj(m, off);
1896 	m_set_rcvif(m, vifp->v_ifp);
1897 
1898 	if (__predict_false(!pktq_enqueue(ip_pktq, m, 0))) {
1899 		m_freem(m);
1900 	}
1901 }
1902 
1903 /*
1904  * Check if the packet should be received on the vif denoted by arg.
1905  * (The encap selection code will call this once per vif since each is
1906  * registered separately.)
1907  */
1908 static int
1909 vif_encapcheck(struct mbuf *m, int off, int proto, void *arg)
1910 {
1911 	struct vif *vifp;
1912 	struct ip ip;
1913 
1914 #ifdef DIAGNOSTIC
1915 	if (!arg || proto != IPPROTO_IPV4)
1916 		panic("unexpected arg in vif_encapcheck");
1917 #endif
1918 
1919 	/*
1920 	 * Accept the packet only if the inner heaader is multicast
1921 	 * and the outer header matches a tunnel-mode vif.  Order
1922 	 * checks in the hope that common non-matching packets will be
1923 	 * rejected quickly.  Assume that unicast IPv4 traffic in a
1924 	 * parallel tunnel (e.g. gif(4)) is unlikely.
1925 	 */
1926 
1927 	/* Obtain the outer IP header and the vif pointer. */
1928 	m_copydata((struct mbuf *)m, 0, sizeof(ip), (void *)&ip);
1929 	vifp = (struct vif *)arg;
1930 
1931 	/*
1932 	 * The outer source must match the vif's remote peer address.
1933 	 * For a multicast router with several tunnels, this is the
1934 	 * only check that will fail on packets in other tunnels,
1935 	 * assuming the local address is the same.
1936 	 */
1937 	if (!in_hosteq(vifp->v_rmt_addr, ip.ip_src))
1938 		return 0;
1939 
1940 	/* The outer destination must match the vif's local address. */
1941 	if (!in_hosteq(vifp->v_lcl_addr, ip.ip_dst))
1942 		return 0;
1943 
1944 	/* The vif must be of tunnel type. */
1945 	if ((vifp->v_flags & VIFF_TUNNEL) == 0)
1946 		return 0;
1947 
1948 	/* Check that the inner destination is multicast. */
1949 	m_copydata((struct mbuf *)m, off, sizeof(ip), (void *)&ip);
1950 	if (!IN_MULTICAST(ip.ip_dst.s_addr))
1951 		return 0;
1952 
1953 	/*
1954 	 * We have checked that both the outer src and dst addresses
1955 	 * match the vif, and that the inner destination is multicast
1956 	 * (224/5).  By claiming more than 64, we intend to
1957 	 * preferentially take packets that also match a parallel
1958 	 * gif(4).
1959 	 */
1960 	return 32 + 32 + 5;
1961 }
1962 
1963 /*
1964  * Token bucket filter module
1965  */
1966 static void
1967 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
1968 {
1969 
1970 	if (len > MAX_BKT_SIZE) {
1971 		/* drop if packet is too large */
1972 		mrtstat.mrts_pkt2large++;
1973 		m_freem(m);
1974 		return;
1975 	}
1976 
1977 	tbf_update_tokens(vifp);
1978 
1979 	/*
1980 	 * If there are enough tokens, and the queue is empty, send this packet
1981 	 * out immediately.  Otherwise, try to insert it on this vif's queue.
1982 	 */
1983 	if (vifp->tbf_q_len == 0) {
1984 		if (len <= vifp->tbf_n_tok) {
1985 			vifp->tbf_n_tok -= len;
1986 			tbf_send_packet(vifp, m);
1987 		} else {
1988 			/* queue packet and timeout till later */
1989 			tbf_queue(vifp, m);
1990 			callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
1991 			    tbf_reprocess_q, vifp);
1992 		}
1993 	} else {
1994 		if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
1995 		    !tbf_dq_sel(vifp, ip)) {
1996 			/* queue full, and couldn't make room */
1997 			mrtstat.mrts_q_overflow++;
1998 			m_freem(m);
1999 		} else {
2000 			/* queue length low enough, or made room */
2001 			tbf_queue(vifp, m);
2002 			tbf_process_q(vifp);
2003 		}
2004 	}
2005 }
2006 
2007 /*
2008  * adds a packet to the queue at the interface
2009  */
2010 static void
2011 tbf_queue(struct vif *vifp, struct mbuf *m)
2012 {
2013 	int s = splsoftnet();
2014 
2015 	/* insert at tail */
2016 	*vifp->tbf_t = m;
2017 	vifp->tbf_t = &m->m_nextpkt;
2018 	vifp->tbf_q_len++;
2019 
2020 	splx(s);
2021 }
2022 
2023 
2024 /*
2025  * processes the queue at the interface
2026  */
2027 static void
2028 tbf_process_q(struct vif *vifp)
2029 {
2030 	struct mbuf *m;
2031 	int len;
2032 	int s = splsoftnet();
2033 
2034 	/*
2035 	 * Loop through the queue at the interface and send as many packets
2036 	 * as possible.
2037 	 */
2038 	for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
2039 		len = ntohs(mtod(m, struct ip *)->ip_len);
2040 
2041 		/* determine if the packet can be sent */
2042 		if (len <= vifp->tbf_n_tok) {
2043 			/* if so,
2044 			 * reduce no of tokens, dequeue the packet,
2045 			 * send the packet.
2046 			 */
2047 			if ((vifp->tbf_q = m->m_nextpkt) == NULL)
2048 				vifp->tbf_t = &vifp->tbf_q;
2049 			--vifp->tbf_q_len;
2050 
2051 			m->m_nextpkt = NULL;
2052 			vifp->tbf_n_tok -= len;
2053 			tbf_send_packet(vifp, m);
2054 		} else
2055 			break;
2056 	}
2057 	splx(s);
2058 }
2059 
2060 static void
2061 tbf_reprocess_q(void *arg)
2062 {
2063 	struct vif *vifp = arg;
2064 
2065 	if (ip_mrouter == NULL)
2066 		return;
2067 
2068 	tbf_update_tokens(vifp);
2069 	tbf_process_q(vifp);
2070 
2071 	if (vifp->tbf_q_len != 0)
2072 		callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2073 		    tbf_reprocess_q, vifp);
2074 }
2075 
2076 /* function that will selectively discard a member of the queue
2077  * based on the precedence value and the priority
2078  */
2079 static int
2080 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2081 {
2082 	u_int p;
2083 	struct mbuf **mp, *m;
2084 	int s = splsoftnet();
2085 
2086 	p = priority(vifp, ip);
2087 
2088 	for (mp = &vifp->tbf_q, m = *mp;
2089 	    m != NULL;
2090 	    mp = &m->m_nextpkt, m = *mp) {
2091 		if (p > priority(vifp, mtod(m, struct ip *))) {
2092 			if ((*mp = m->m_nextpkt) == NULL)
2093 				vifp->tbf_t = mp;
2094 			--vifp->tbf_q_len;
2095 
2096 			m_freem(m);
2097 			mrtstat.mrts_drop_sel++;
2098 			splx(s);
2099 			return (1);
2100 		}
2101 	}
2102 	splx(s);
2103 	return (0);
2104 }
2105 
2106 static void
2107 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2108 {
2109 	int error;
2110 	int s = splsoftnet();
2111 
2112 	if (vifp->v_flags & VIFF_TUNNEL) {
2113 		/* If tunnel options */
2114 		ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL);
2115 	} else {
2116 		/* if physical interface option, extract the options and then send */
2117 		struct ip_moptions imo;
2118 
2119 		imo.imo_multicast_if_index = if_get_index(vifp->v_ifp);
2120 		imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2121 		imo.imo_multicast_loop = 1;
2122 #ifdef RSVP_ISI
2123 		imo.imo_multicast_vif = -1;
2124 #endif
2125 
2126 		error = ip_output(m, NULL, NULL, IP_FORWARDING|IP_MULTICASTOPTS,
2127 		    &imo, NULL);
2128 
2129 		if (mrtdebug & DEBUG_XMIT)
2130 			log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2131 			    (long)(vifp - viftable), error);
2132 	}
2133 	splx(s);
2134 }
2135 
2136 /* determine the current time and then
2137  * the elapsed time (between the last time and time now)
2138  * in milliseconds & update the no. of tokens in the bucket
2139  */
2140 static void
2141 tbf_update_tokens(struct vif *vifp)
2142 {
2143 	struct timeval tp;
2144 	u_int32_t tm;
2145 	int s = splsoftnet();
2146 
2147 	microtime(&tp);
2148 
2149 	TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2150 
2151 	/*
2152 	 * This formula is actually
2153 	 * "time in seconds" * "bytes/second".
2154 	 *
2155 	 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2156 	 *
2157 	 * The (1000/1024) was introduced in add_vif to optimize
2158 	 * this divide into a shift.
2159 	 */
2160 	vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2161 	vifp->tbf_last_pkt_t = tp;
2162 
2163 	if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2164 		vifp->tbf_n_tok = MAX_BKT_SIZE;
2165 
2166 	splx(s);
2167 }
2168 
2169 static int
2170 priority(struct vif *vifp, struct ip *ip)
2171 {
2172 	int prio = 50;	/* the lowest priority -- default case */
2173 
2174 	/* temporary hack; may add general packet classifier some day */
2175 
2176 	/*
2177 	 * The UDP port space is divided up into four priority ranges:
2178 	 * [0, 16384)     : unclassified - lowest priority
2179 	 * [16384, 32768) : audio - highest priority
2180 	 * [32768, 49152) : whiteboard - medium priority
2181 	 * [49152, 65536) : video - low priority
2182 	 */
2183 	if (ip->ip_p == IPPROTO_UDP) {
2184 		struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2185 
2186 		switch (ntohs(udp->uh_dport) & 0xc000) {
2187 		case 0x4000:
2188 			prio = 70;
2189 			break;
2190 		case 0x8000:
2191 			prio = 60;
2192 			break;
2193 		case 0xc000:
2194 			prio = 55;
2195 			break;
2196 		}
2197 
2198 		if (tbfdebug > 1)
2199 			log(LOG_DEBUG, "port %x prio %d\n",
2200 			    ntohs(udp->uh_dport), prio);
2201 	}
2202 
2203 	return (prio);
2204 }
2205 
2206 /*
2207  * End of token bucket filter modifications
2208  */
2209 #ifdef RSVP_ISI
2210 int
2211 ip_rsvp_vif_init(struct socket *so, struct mbuf *m)
2212 {
2213 	int vifi, s;
2214 
2215 	RSVP_DPRINTF(("%s: so_type = %d, pr_protocol = %d\n", __func__
2216 	    so->so_type, so->so_proto->pr_protocol));
2217 
2218 	if (so->so_type != SOCK_RAW ||
2219 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2220 		return (EOPNOTSUPP);
2221 
2222 	/* Check mbuf. */
2223 	if (m == NULL || m->m_len != sizeof(int)) {
2224 		return (EINVAL);
2225 	}
2226 	vifi = *(mtod(m, int *));
2227 
2228 	RSVP_DPRINTF(("%s: vif = %d rsvp_on = %d\n", __func__, vifi, rsvp_on));
2229 
2230 	s = splsoftnet();
2231 
2232 	/* Check vif. */
2233 	if (!legal_vif_num(vifi)) {
2234 		splx(s);
2235 		return (EADDRNOTAVAIL);
2236 	}
2237 
2238 	/* Check if socket is available. */
2239 	if (viftable[vifi].v_rsvpd != NULL) {
2240 		splx(s);
2241 		return (EADDRINUSE);
2242 	}
2243 
2244 	viftable[vifi].v_rsvpd = so;
2245 	/*
2246 	 * This may seem silly, but we need to be sure we don't over-increment
2247 	 * the RSVP counter, in case something slips up.
2248 	 */
2249 	if (!viftable[vifi].v_rsvp_on) {
2250 		viftable[vifi].v_rsvp_on = 1;
2251 		rsvp_on++;
2252 	}
2253 
2254 	splx(s);
2255 	return (0);
2256 }
2257 
2258 int
2259 ip_rsvp_vif_done(struct socket *so, struct mbuf *m)
2260 {
2261 	int vifi, s;
2262 
2263 	RSVP_DPRINTF(("%s: so_type = %d, pr_protocol = %d\n", __func__,
2264 	    so->so_type, so->so_proto->pr_protocol));
2265 
2266 	if (so->so_type != SOCK_RAW ||
2267 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2268 		return (EOPNOTSUPP);
2269 
2270 	/* Check mbuf. */
2271 	if (m == NULL || m->m_len != sizeof(int)) {
2272 		return (EINVAL);
2273 	}
2274 	vifi = *(mtod(m, int *));
2275 
2276 	s = splsoftnet();
2277 
2278 	/* Check vif. */
2279 	if (!legal_vif_num(vifi)) {
2280 		splx(s);
2281 		return (EADDRNOTAVAIL);
2282 	}
2283 
2284 	RSVP_DPRINTF(("%s: v_rsvpd = %x so = %x\n", __func__,
2285 	    viftable[vifi].v_rsvpd, so));
2286 
2287 	viftable[vifi].v_rsvpd = NULL;
2288 	/*
2289 	 * This may seem silly, but we need to be sure we don't over-decrement
2290 	 * the RSVP counter, in case something slips up.
2291 	 */
2292 	if (viftable[vifi].v_rsvp_on) {
2293 		viftable[vifi].v_rsvp_on = 0;
2294 		rsvp_on--;
2295 	}
2296 
2297 	splx(s);
2298 	return (0);
2299 }
2300 
2301 void
2302 ip_rsvp_force_done(struct socket *so)
2303 {
2304 	int vifi, s;
2305 
2306 	/* Don't bother if it is not the right type of socket. */
2307 	if (so->so_type != SOCK_RAW ||
2308 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2309 		return;
2310 
2311 	s = splsoftnet();
2312 
2313 	/*
2314 	 * The socket may be attached to more than one vif...this
2315 	 * is perfectly legal.
2316 	 */
2317 	for (vifi = 0; vifi < numvifs; vifi++) {
2318 		if (viftable[vifi].v_rsvpd == so) {
2319 			viftable[vifi].v_rsvpd = NULL;
2320 			/*
2321 			 * This may seem silly, but we need to be sure we don't
2322 			 * over-decrement the RSVP counter, in case something
2323 			 * slips up.
2324 			 */
2325 			if (viftable[vifi].v_rsvp_on) {
2326 				viftable[vifi].v_rsvp_on = 0;
2327 				rsvp_on--;
2328 			}
2329 		}
2330 	}
2331 
2332 	splx(s);
2333 	return;
2334 }
2335 
2336 void
2337 rsvp_input(struct mbuf *m, struct ifnet *ifp)
2338 {
2339 	int vifi, s;
2340 	struct ip *ip = mtod(m, struct ip *);
2341 	struct sockaddr_in rsvp_src;
2342 
2343 	RSVP_DPRINTF(("%s: rsvp_on %d\n", __func__, rsvp_on));
2344 
2345 	/*
2346 	 * Can still get packets with rsvp_on = 0 if there is a local member
2347 	 * of the group to which the RSVP packet is addressed.  But in this
2348 	 * case we want to throw the packet away.
2349 	 */
2350 	if (!rsvp_on) {
2351 		m_freem(m);
2352 		return;
2353 	}
2354 
2355 	/*
2356 	 * If the old-style non-vif-associated socket is set, then use
2357 	 * it and ignore the new ones.
2358 	 */
2359 	if (ip_rsvpd != NULL) {
2360 		RSVP_DPRINTF(("%s: Sending packet up old-style socket\n",
2361 		    __func__));
2362 		rip_input(m);	/*XXX*/
2363 		return;
2364 	}
2365 
2366 	s = splsoftnet();
2367 
2368 	RSVP_DPRINTF(("%s: check vifs\n", __func__));
2369 
2370 	/* Find which vif the packet arrived on. */
2371 	for (vifi = 0; vifi < numvifs; vifi++) {
2372 		if (viftable[vifi].v_ifp == ifp)
2373 			break;
2374 	}
2375 
2376 	if (vifi == numvifs) {
2377 		/* Can't find vif packet arrived on. Drop packet. */
2378 		RSVP_DPRINTF("%s: Can't find vif for packet...dropping it.\n",
2379 		    __func__));
2380 		m_freem(m);
2381 		splx(s);
2382 		return;
2383 	}
2384 
2385 	RSVP_DPRINTF(("%s: check socket\n", __func__));
2386 
2387 	if (viftable[vifi].v_rsvpd == NULL) {
2388 		/*
2389 		 * drop packet, since there is no specific socket for this
2390 		 * interface
2391 		 */
2392 		RSVP_DPRINTF(("%s: No socket defined for vif %d\n", __func__,
2393 		    vifi));
2394 		m_freem(m);
2395 		splx(s);
2396 		return;
2397 	}
2398 
2399 	sockaddr_in_init(&rsvp_src, &ip->ip_src, 0);
2400 
2401 	if (m)
2402 		RSVP_DPRINTF(("%s: m->m_len = %d, sbspace() = %d\n", __func__,
2403 		    m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv)));
2404 
2405 	if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2406 		RSVP_DPRINTF(("%s: Failed to append to socket\n", __func__));
2407 	else
2408 		RSVP_DPRINTF(("%s: send packet up\n", __func__));
2409 
2410 	splx(s);
2411 }
2412 #endif /* RSVP_ISI */
2413 
2414 /*
2415  * Code for bandwidth monitors
2416  */
2417 
2418 /*
2419  * Define common interface for timeval-related methods
2420  */
2421 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2422 #define	BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2423 #define	BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2424 
2425 static uint32_t
2426 compute_bw_meter_flags(struct bw_upcall *req)
2427 {
2428     uint32_t flags = 0;
2429 
2430     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2431 	flags |= BW_METER_UNIT_PACKETS;
2432     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2433 	flags |= BW_METER_UNIT_BYTES;
2434     if (req->bu_flags & BW_UPCALL_GEQ)
2435 	flags |= BW_METER_GEQ;
2436     if (req->bu_flags & BW_UPCALL_LEQ)
2437 	flags |= BW_METER_LEQ;
2438 
2439     return flags;
2440 }
2441 
2442 /*
2443  * Add a bw_meter entry
2444  */
2445 static int
2446 add_bw_upcall(struct bw_upcall *req)
2447 {
2448     int s;
2449     struct mfc *mfc;
2450     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2451 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2452     struct timeval now;
2453     struct bw_meter *x;
2454     uint32_t flags;
2455 
2456     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2457 	return EOPNOTSUPP;
2458 
2459     /* Test if the flags are valid */
2460     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2461 	return EINVAL;
2462     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2463 	return EINVAL;
2464     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2465 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2466 	return EINVAL;
2467 
2468     /* Test if the threshold time interval is valid */
2469     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2470 	return EINVAL;
2471 
2472     flags = compute_bw_meter_flags(req);
2473 
2474     /*
2475      * Find if we have already same bw_meter entry
2476      */
2477     s = splsoftnet();
2478     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2479     if (mfc == NULL) {
2480 	splx(s);
2481 	return EADDRNOTAVAIL;
2482     }
2483     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2484 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2485 			   &req->bu_threshold.b_time, ==)) &&
2486 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2487 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2488 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
2489 	    splx(s);
2490 	    return 0;		/* XXX Already installed */
2491 	}
2492     }
2493 
2494     /* Allocate the new bw_meter entry */
2495     x = kmem_intr_alloc(sizeof(*x), KM_NOSLEEP);
2496     if (x == NULL) {
2497 	splx(s);
2498 	return ENOBUFS;
2499     }
2500 
2501     /* Set the new bw_meter entry */
2502     x->bm_threshold.b_time = req->bu_threshold.b_time;
2503     microtime(&now);
2504     x->bm_start_time = now;
2505     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2506     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2507     x->bm_measured.b_packets = 0;
2508     x->bm_measured.b_bytes = 0;
2509     x->bm_flags = flags;
2510     x->bm_time_next = NULL;
2511     x->bm_time_hash = BW_METER_BUCKETS;
2512 
2513     /* Add the new bw_meter entry to the front of entries for this MFC */
2514     x->bm_mfc = mfc;
2515     x->bm_mfc_next = mfc->mfc_bw_meter;
2516     mfc->mfc_bw_meter = x;
2517     schedule_bw_meter(x, &now);
2518     splx(s);
2519 
2520     return 0;
2521 }
2522 
2523 static void
2524 free_bw_list(struct bw_meter *list)
2525 {
2526     while (list != NULL) {
2527 	struct bw_meter *x = list;
2528 
2529 	list = list->bm_mfc_next;
2530 	unschedule_bw_meter(x);
2531 	kmem_free(x, sizeof(*x));
2532     }
2533 }
2534 
2535 /*
2536  * Delete one or multiple bw_meter entries
2537  */
2538 static int
2539 del_bw_upcall(struct bw_upcall *req)
2540 {
2541     int s;
2542     struct mfc *mfc;
2543     struct bw_meter *x;
2544 
2545     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2546 	return EOPNOTSUPP;
2547 
2548     s = splsoftnet();
2549     /* Find the corresponding MFC entry */
2550     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2551     if (mfc == NULL) {
2552 	splx(s);
2553 	return EADDRNOTAVAIL;
2554     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2555 	/*
2556 	 * Delete all bw_meter entries for this mfc
2557 	 */
2558 	struct bw_meter *list;
2559 
2560 	list = mfc->mfc_bw_meter;
2561 	mfc->mfc_bw_meter = NULL;
2562 	free_bw_list(list);
2563 	splx(s);
2564 	return 0;
2565     } else {			/* Delete a single bw_meter entry */
2566 	struct bw_meter *prev;
2567 	uint32_t flags = 0;
2568 
2569 	flags = compute_bw_meter_flags(req);
2570 
2571 	/* Find the bw_meter entry to delete */
2572 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2573 	     prev = x, x = x->bm_mfc_next) {
2574 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2575 			       &req->bu_threshold.b_time, ==)) &&
2576 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2577 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2578 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2579 		break;
2580 	}
2581 	if (x != NULL) { /* Delete entry from the list for this MFC */
2582 	    if (prev != NULL)
2583 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2584 	    else
2585 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2586 
2587 	    unschedule_bw_meter(x);
2588 	    splx(s);
2589 	    /* Free the bw_meter entry */
2590 	    kmem_free(x, sizeof(*x));
2591 	    return 0;
2592 	} else {
2593 	    splx(s);
2594 	    return EINVAL;
2595 	}
2596     }
2597     /* NOTREACHED */
2598 }
2599 
2600 /*
2601  * Perform bandwidth measurement processing that may result in an upcall
2602  */
2603 static void
2604 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2605 {
2606     struct timeval delta;
2607 
2608     delta = *nowp;
2609     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2610 
2611     if (x->bm_flags & BW_METER_GEQ) {
2612 	/*
2613 	 * Processing for ">=" type of bw_meter entry
2614 	 */
2615 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2616 	    /* Reset the bw_meter entry */
2617 	    x->bm_start_time = *nowp;
2618 	    x->bm_measured.b_packets = 0;
2619 	    x->bm_measured.b_bytes = 0;
2620 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2621 	}
2622 
2623 	/* Record that a packet is received */
2624 	x->bm_measured.b_packets++;
2625 	x->bm_measured.b_bytes += plen;
2626 
2627 	/*
2628 	 * Test if we should deliver an upcall
2629 	 */
2630 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2631 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2632 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2633 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2634 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2635 		/* Prepare an upcall for delivery */
2636 		bw_meter_prepare_upcall(x, nowp);
2637 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2638 	    }
2639 	}
2640     } else if (x->bm_flags & BW_METER_LEQ) {
2641 	/*
2642 	 * Processing for "<=" type of bw_meter entry
2643 	 */
2644 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2645 	    /*
2646 	     * We are behind time with the multicast forwarding table
2647 	     * scanning for "<=" type of bw_meter entries, so test now
2648 	     * if we should deliver an upcall.
2649 	     */
2650 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2651 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2652 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2653 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2654 		/* Prepare an upcall for delivery */
2655 		bw_meter_prepare_upcall(x, nowp);
2656 	    }
2657 	    /* Reschedule the bw_meter entry */
2658 	    unschedule_bw_meter(x);
2659 	    schedule_bw_meter(x, nowp);
2660 	}
2661 
2662 	/* Record that a packet is received */
2663 	x->bm_measured.b_packets++;
2664 	x->bm_measured.b_bytes += plen;
2665 
2666 	/*
2667 	 * Test if we should restart the measuring interval
2668 	 */
2669 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2670 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2671 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
2672 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2673 	    /* Don't restart the measuring interval */
2674 	} else {
2675 	    /* Do restart the measuring interval */
2676 	    /*
2677 	     * XXX: note that we don't unschedule and schedule, because this
2678 	     * might be too much overhead per packet. Instead, when we process
2679 	     * all entries for a given timer hash bin, we check whether it is
2680 	     * really a timeout. If not, we reschedule at that time.
2681 	     */
2682 	    x->bm_start_time = *nowp;
2683 	    x->bm_measured.b_packets = 0;
2684 	    x->bm_measured.b_bytes = 0;
2685 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2686 	}
2687     }
2688 }
2689 
2690 /*
2691  * Prepare a bandwidth-related upcall
2692  */
2693 static void
2694 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2695 {
2696     struct timeval delta;
2697     struct bw_upcall *u;
2698 
2699     /*
2700      * Compute the measured time interval
2701      */
2702     delta = *nowp;
2703     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2704 
2705     /*
2706      * If there are too many pending upcalls, deliver them now
2707      */
2708     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2709 	bw_upcalls_send();
2710 
2711     /*
2712      * Set the bw_upcall entry
2713      */
2714     u = &bw_upcalls[bw_upcalls_n++];
2715     u->bu_src = x->bm_mfc->mfc_origin;
2716     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2717     u->bu_threshold.b_time = x->bm_threshold.b_time;
2718     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2719     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2720     u->bu_measured.b_time = delta;
2721     u->bu_measured.b_packets = x->bm_measured.b_packets;
2722     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2723     u->bu_flags = 0;
2724     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2725 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2726     if (x->bm_flags & BW_METER_UNIT_BYTES)
2727 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2728     if (x->bm_flags & BW_METER_GEQ)
2729 	u->bu_flags |= BW_UPCALL_GEQ;
2730     if (x->bm_flags & BW_METER_LEQ)
2731 	u->bu_flags |= BW_UPCALL_LEQ;
2732 }
2733 
2734 /*
2735  * Send the pending bandwidth-related upcalls
2736  */
2737 static void
2738 bw_upcalls_send(void)
2739 {
2740     struct mbuf *m;
2741     int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2742     struct sockaddr_in k_igmpsrc = {
2743 	    .sin_len = sizeof(k_igmpsrc),
2744 	    .sin_family = AF_INET,
2745     };
2746     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2747 				      0,		/* unused2 */
2748 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2749 				      0,		/* im_mbz  */
2750 				      0,		/* im_vif  */
2751 				      0,		/* unused3 */
2752 				      { 0 },		/* im_src  */
2753 				      { 0 } };		/* im_dst  */
2754 
2755     if (bw_upcalls_n == 0)
2756 	return;			/* No pending upcalls */
2757 
2758     bw_upcalls_n = 0;
2759 
2760     /*
2761      * Allocate a new mbuf, initialize it with the header and
2762      * the payload for the pending calls.
2763      */
2764     MGETHDR(m, M_DONTWAIT, MT_HEADER);
2765     if (m == NULL) {
2766 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2767 	return;
2768     }
2769 
2770     m->m_len = m->m_pkthdr.len = 0;
2771     m_copyback(m, 0, sizeof(struct igmpmsg), (void *)&igmpmsg);
2772     m_copyback(m, sizeof(struct igmpmsg), len, (void *)&bw_upcalls[0]);
2773 
2774     /*
2775      * Send the upcalls
2776      * XXX do we need to set the address in k_igmpsrc ?
2777      */
2778     mrtstat.mrts_upcalls++;
2779     if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2780 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2781 	++mrtstat.mrts_upq_sockfull;
2782     }
2783 }
2784 
2785 /*
2786  * Compute the timeout hash value for the bw_meter entries
2787  */
2788 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2789     do {								\
2790 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2791 									\
2792 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2793 	(hash) = next_timeval.tv_sec;					\
2794 	if (next_timeval.tv_usec)					\
2795 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2796 	(hash) %= BW_METER_BUCKETS;					\
2797     } while (/*CONSTCOND*/ 0)
2798 
2799 /*
2800  * Schedule a timer to process periodically bw_meter entry of type "<="
2801  * by linking the entry in the proper hash bucket.
2802  */
2803 static void
2804 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2805 {
2806     int time_hash;
2807 
2808     if (!(x->bm_flags & BW_METER_LEQ))
2809 	return;		/* XXX: we schedule timers only for "<=" entries */
2810 
2811     /*
2812      * Reset the bw_meter entry
2813      */
2814     x->bm_start_time = *nowp;
2815     x->bm_measured.b_packets = 0;
2816     x->bm_measured.b_bytes = 0;
2817     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2818 
2819     /*
2820      * Compute the timeout hash value and insert the entry
2821      */
2822     BW_METER_TIMEHASH(x, time_hash);
2823     x->bm_time_next = bw_meter_timers[time_hash];
2824     bw_meter_timers[time_hash] = x;
2825     x->bm_time_hash = time_hash;
2826 }
2827 
2828 /*
2829  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2830  * by removing the entry from the proper hash bucket.
2831  */
2832 static void
2833 unschedule_bw_meter(struct bw_meter *x)
2834 {
2835     int time_hash;
2836     struct bw_meter *prev, *tmp;
2837 
2838     if (!(x->bm_flags & BW_METER_LEQ))
2839 	return;		/* XXX: we schedule timers only for "<=" entries */
2840 
2841     /*
2842      * Compute the timeout hash value and delete the entry
2843      */
2844     time_hash = x->bm_time_hash;
2845     if (time_hash >= BW_METER_BUCKETS)
2846 	return;		/* Entry was not scheduled */
2847 
2848     for (prev = NULL, tmp = bw_meter_timers[time_hash];
2849 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2850 	if (tmp == x)
2851 	    break;
2852 
2853     if (tmp == NULL)
2854 	panic("unschedule_bw_meter: bw_meter entry not found");
2855 
2856     if (prev != NULL)
2857 	prev->bm_time_next = x->bm_time_next;
2858     else
2859 	bw_meter_timers[time_hash] = x->bm_time_next;
2860 
2861     x->bm_time_next = NULL;
2862     x->bm_time_hash = BW_METER_BUCKETS;
2863 }
2864 
2865 /*
2866  * Process all "<=" type of bw_meter that should be processed now,
2867  * and for each entry prepare an upcall if necessary. Each processed
2868  * entry is rescheduled again for the (periodic) processing.
2869  *
2870  * This is run periodically (once per second normally). On each round,
2871  * all the potentially matching entries are in the hash slot that we are
2872  * looking at.
2873  */
2874 static void
2875 bw_meter_process(void)
2876 {
2877     int s;
2878     static uint32_t last_tv_sec;	/* last time we processed this */
2879 
2880     uint32_t loops;
2881     int i;
2882     struct timeval now, process_endtime;
2883 
2884     microtime(&now);
2885     if (last_tv_sec == now.tv_sec)
2886 	return;		/* nothing to do */
2887 
2888     loops = now.tv_sec - last_tv_sec;
2889     last_tv_sec = now.tv_sec;
2890     if (loops > BW_METER_BUCKETS)
2891 	loops = BW_METER_BUCKETS;
2892 
2893     s = splsoftnet();
2894     /*
2895      * Process all bins of bw_meter entries from the one after the last
2896      * processed to the current one. On entry, i points to the last bucket
2897      * visited, so we need to increment i at the beginning of the loop.
2898      */
2899     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2900 	struct bw_meter *x, *tmp_list;
2901 
2902 	if (++i >= BW_METER_BUCKETS)
2903 	    i = 0;
2904 
2905 	/* Disconnect the list of bw_meter entries from the bin */
2906 	tmp_list = bw_meter_timers[i];
2907 	bw_meter_timers[i] = NULL;
2908 
2909 	/* Process the list of bw_meter entries */
2910 	while (tmp_list != NULL) {
2911 	    x = tmp_list;
2912 	    tmp_list = tmp_list->bm_time_next;
2913 
2914 	    /* Test if the time interval is over */
2915 	    process_endtime = x->bm_start_time;
2916 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2917 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2918 		/* Not yet: reschedule, but don't reset */
2919 		int time_hash;
2920 
2921 		BW_METER_TIMEHASH(x, time_hash);
2922 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2923 		    /*
2924 		     * XXX: somehow the bin processing is a bit ahead of time.
2925 		     * Put the entry in the next bin.
2926 		     */
2927 		    if (++time_hash >= BW_METER_BUCKETS)
2928 			time_hash = 0;
2929 		}
2930 		x->bm_time_next = bw_meter_timers[time_hash];
2931 		bw_meter_timers[time_hash] = x;
2932 		x->bm_time_hash = time_hash;
2933 
2934 		continue;
2935 	    }
2936 
2937 	    /*
2938 	     * Test if we should deliver an upcall
2939 	     */
2940 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2941 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2942 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2943 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2944 		/* Prepare an upcall for delivery */
2945 		bw_meter_prepare_upcall(x, &now);
2946 	    }
2947 
2948 	    /*
2949 	     * Reschedule for next processing
2950 	     */
2951 	    schedule_bw_meter(x, &now);
2952 	}
2953     }
2954 
2955     /* Send all upcalls that are pending delivery */
2956     bw_upcalls_send();
2957 
2958     splx(s);
2959 }
2960 
2961 /*
2962  * A periodic function for sending all upcalls that are pending delivery
2963  */
2964 static void
2965 expire_bw_upcalls_send(void *unused)
2966 {
2967     int s;
2968 
2969     s = splsoftnet();
2970     bw_upcalls_send();
2971     splx(s);
2972 
2973     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2974 		  expire_bw_upcalls_send, NULL);
2975 }
2976 
2977 /*
2978  * A periodic function for periodic scanning of the multicast forwarding
2979  * table for processing all "<=" bw_meter entries.
2980  */
2981 static void
2982 expire_bw_meter_process(void *unused)
2983 {
2984     if (mrt_api_config & MRT_MFC_BW_UPCALL)
2985 	bw_meter_process();
2986 
2987     callout_reset(&bw_meter_ch, BW_METER_PERIOD,
2988 		  expire_bw_meter_process, NULL);
2989 }
2990 
2991 /*
2992  * End of bandwidth monitoring code
2993  */
2994 
2995 #ifdef PIM
2996 /*
2997  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2998  */
2999 static int
3000 pim_register_send(struct ip *ip, struct vif *vifp,
3001 	struct mbuf *m, struct mfc *rt)
3002 {
3003     struct mbuf *mb_copy, *mm;
3004 
3005     if (mrtdebug & DEBUG_PIM)
3006         log(LOG_DEBUG, "pim_register_send: \n");
3007 
3008     mb_copy = pim_register_prepare(ip, m);
3009     if (mb_copy == NULL)
3010 	return ENOBUFS;
3011 
3012     /*
3013      * Send all the fragments. Note that the mbuf for each fragment
3014      * is freed by the sending machinery.
3015      */
3016     for (mm = mb_copy; mm; mm = mb_copy) {
3017 	mb_copy = mm->m_nextpkt;
3018 	mm->m_nextpkt = NULL;
3019 	mm = m_pullup(mm, sizeof(struct ip));
3020 	if (mm != NULL) {
3021 	    ip = mtod(mm, struct ip *);
3022 	    if ((mrt_api_config & MRT_MFC_RP) &&
3023 		!in_nullhost(rt->mfc_rp)) {
3024 		pim_register_send_rp(ip, vifp, mm, rt);
3025 	    } else {
3026 		pim_register_send_upcall(ip, vifp, mm, rt);
3027 	    }
3028 	}
3029     }
3030 
3031     return 0;
3032 }
3033 
3034 /*
3035  * Return a copy of the data packet that is ready for PIM Register
3036  * encapsulation.
3037  * XXX: Note that in the returned copy the IP header is a valid one.
3038  */
3039 static struct mbuf *
3040 pim_register_prepare(struct ip *ip, struct mbuf *m)
3041 {
3042     struct mbuf *mb_copy = NULL;
3043     int mtu;
3044 
3045     /* Take care of delayed checksums */
3046     if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3047 	in_delayed_cksum(m);
3048 	m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3049     }
3050 
3051     /*
3052      * Copy the old packet & pullup its IP header into the
3053      * new mbuf so we can modify it.
3054      */
3055     mb_copy = m_copypacket(m, M_DONTWAIT);
3056     if (mb_copy == NULL)
3057 	return NULL;
3058     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3059     if (mb_copy == NULL)
3060 	return NULL;
3061 
3062     /* take care of the TTL */
3063     ip = mtod(mb_copy, struct ip *);
3064     --ip->ip_ttl;
3065 
3066     /* Compute the MTU after the PIM Register encapsulation */
3067     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3068 
3069     if (ntohs(ip->ip_len) <= mtu) {
3070 	/* Turn the IP header into a valid one */
3071 	ip->ip_sum = 0;
3072 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3073     } else {
3074 	/* Fragment the packet */
3075 	if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3076 	    /* XXX: mb_copy was freed by ip_fragment() */
3077 	    return NULL;
3078 	}
3079     }
3080     return mb_copy;
3081 }
3082 
3083 /*
3084  * Send an upcall with the data packet to the user-level process.
3085  */
3086 static int
3087 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3088     struct mbuf *mb_copy, struct mfc *rt)
3089 {
3090     struct mbuf *mb_first;
3091     int len = ntohs(ip->ip_len);
3092     struct igmpmsg *im;
3093     struct sockaddr_in k_igmpsrc = {
3094 	    .sin_len = sizeof(k_igmpsrc),
3095 	    .sin_family = AF_INET,
3096     };
3097 
3098     /*
3099      * Add a new mbuf with an upcall header
3100      */
3101     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3102     if (mb_first == NULL) {
3103 	m_freem(mb_copy);
3104 	return ENOBUFS;
3105     }
3106     mb_first->m_data += max_linkhdr;
3107     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3108     mb_first->m_len = sizeof(struct igmpmsg);
3109     mb_first->m_next = mb_copy;
3110 
3111     /* Send message to routing daemon */
3112     im = mtod(mb_first, struct igmpmsg *);
3113     im->im_msgtype	= IGMPMSG_WHOLEPKT;
3114     im->im_mbz		= 0;
3115     im->im_vif		= vifp - viftable;
3116     im->im_src		= ip->ip_src;
3117     im->im_dst		= ip->ip_dst;
3118 
3119     k_igmpsrc.sin_addr	= ip->ip_src;
3120 
3121     mrtstat.mrts_upcalls++;
3122 
3123     if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3124 	if (mrtdebug & DEBUG_PIM)
3125 	    log(LOG_WARNING,
3126 		"mcast: pim_register_send_upcall: ip_mrouter socket queue full\n");
3127 	++mrtstat.mrts_upq_sockfull;
3128 	return ENOBUFS;
3129     }
3130 
3131     /* Keep statistics */
3132     pimstat.pims_snd_registers_msgs++;
3133     pimstat.pims_snd_registers_bytes += len;
3134 
3135     return 0;
3136 }
3137 
3138 /*
3139  * Encapsulate the data packet in PIM Register message and send it to the RP.
3140  */
3141 static int
3142 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3143 	struct mbuf *mb_copy, struct mfc *rt)
3144 {
3145     struct mbuf *mb_first;
3146     struct ip *ip_outer;
3147     struct pim_encap_pimhdr *pimhdr;
3148     int len = ntohs(ip->ip_len);
3149     vifi_t vifi = rt->mfc_parent;
3150 
3151     if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3152 	m_freem(mb_copy);
3153 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
3154     }
3155 
3156     /*
3157      * Add a new mbuf with the encapsulating header
3158      */
3159     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3160     if (mb_first == NULL) {
3161 	m_freem(mb_copy);
3162 	return ENOBUFS;
3163     }
3164     mb_first->m_data += max_linkhdr;
3165     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3166     mb_first->m_next = mb_copy;
3167 
3168     mb_first->m_pkthdr.len = len + mb_first->m_len;
3169 
3170     /*
3171      * Fill in the encapsulating IP and PIM header
3172      */
3173     ip_outer = mtod(mb_first, struct ip *);
3174     *ip_outer = pim_encap_iphdr;
3175      if (mb_first->m_pkthdr.len < IP_MINFRAGSIZE)
3176 	ip_outer->ip_id = 0;
3177     else
3178 	ip_outer->ip_id = ip_newid(NULL);
3179     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3180 			     sizeof(pim_encap_pimhdr));
3181     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3182     ip_outer->ip_dst = rt->mfc_rp;
3183     /*
3184      * Copy the inner header TOS to the outer header, and take care of the
3185      * IP_DF bit.
3186      */
3187     ip_outer->ip_tos = ip->ip_tos;
3188     if (ntohs(ip->ip_off) & IP_DF)
3189 	ip_outer->ip_off |= htons(IP_DF);
3190     pimhdr = (struct pim_encap_pimhdr *)((char *)ip_outer
3191 					 + sizeof(pim_encap_iphdr));
3192     *pimhdr = pim_encap_pimhdr;
3193     /* If the iif crosses a border, set the Border-bit */
3194     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3195 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3196 
3197     mb_first->m_data += sizeof(pim_encap_iphdr);
3198     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3199     mb_first->m_data -= sizeof(pim_encap_iphdr);
3200 
3201     if (vifp->v_rate_limit == 0)
3202 	tbf_send_packet(vifp, mb_first);
3203     else
3204 	tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3205 
3206     /* Keep statistics */
3207     pimstat.pims_snd_registers_msgs++;
3208     pimstat.pims_snd_registers_bytes += len;
3209 
3210     return 0;
3211 }
3212 
3213 /*
3214  * PIM-SMv2 and PIM-DM messages processing.
3215  * Receives and verifies the PIM control messages, and passes them
3216  * up to the listening socket, using rip_input().
3217  * The only message with special processing is the PIM_REGISTER message
3218  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3219  * is passed to if_simloop().
3220  */
3221 void
3222 pim_input(struct mbuf *m, ...)
3223 {
3224     struct ip *ip = mtod(m, struct ip *);
3225     struct pim *pim;
3226     int minlen;
3227     int datalen;
3228     int ip_tos;
3229     int proto;
3230     int iphlen;
3231     va_list ap;
3232 
3233     va_start(ap, m);
3234     iphlen = va_arg(ap, int);
3235     proto = va_arg(ap, int);
3236     va_end(ap);
3237 
3238     datalen = ntohs(ip->ip_len) - iphlen;
3239 
3240     /* Keep statistics */
3241     pimstat.pims_rcv_total_msgs++;
3242     pimstat.pims_rcv_total_bytes += datalen;
3243 
3244     /*
3245      * Validate lengths
3246      */
3247     if (datalen < PIM_MINLEN) {
3248 	pimstat.pims_rcv_tooshort++;
3249 	log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3250 	    datalen, (u_long)ip->ip_src.s_addr);
3251 	m_freem(m);
3252 	return;
3253     }
3254 
3255     /*
3256      * If the packet is at least as big as a REGISTER, go agead
3257      * and grab the PIM REGISTER header size, to avoid another
3258      * possible m_pullup() later.
3259      *
3260      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
3261      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3262      */
3263     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3264     /*
3265      * Get the IP and PIM headers in contiguous memory, and
3266      * possibly the PIM REGISTER header.
3267      */
3268     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3269 	(m = m_pullup(m, minlen)) == NULL) {
3270 	log(LOG_ERR, "pim_input: m_pullup failure\n");
3271 	return;
3272     }
3273     /* m_pullup() may have given us a new mbuf so reset ip. */
3274     ip = mtod(m, struct ip *);
3275     ip_tos = ip->ip_tos;
3276 
3277     /* adjust mbuf to point to the PIM header */
3278     m->m_data += iphlen;
3279     m->m_len  -= iphlen;
3280     pim = mtod(m, struct pim *);
3281 
3282     /*
3283      * Validate checksum. If PIM REGISTER, exclude the data packet.
3284      *
3285      * XXX: some older PIMv2 implementations don't make this distinction,
3286      * so for compatibility reason perform the checksum over part of the
3287      * message, and if error, then over the whole message.
3288      */
3289     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3290 	/* do nothing, checksum okay */
3291     } else if (in_cksum(m, datalen)) {
3292 	pimstat.pims_rcv_badsum++;
3293 	if (mrtdebug & DEBUG_PIM)
3294 	    log(LOG_DEBUG, "pim_input: invalid checksum\n");
3295 	m_freem(m);
3296 	return;
3297     }
3298 
3299     /* PIM version check */
3300     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3301 	pimstat.pims_rcv_badversion++;
3302 	log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3303 	    PIM_VT_V(pim->pim_vt), PIM_VERSION);
3304 	m_freem(m);
3305 	return;
3306     }
3307 
3308     /* restore mbuf back to the outer IP */
3309     m->m_data -= iphlen;
3310     m->m_len  += iphlen;
3311 
3312     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3313 	/*
3314 	 * Since this is a REGISTER, we'll make a copy of the register
3315 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3316 	 * routing daemon.
3317 	 */
3318 	int s;
3319 	struct sockaddr_in dst = {
3320 		.sin_len = sizeof(dst),
3321 		.sin_family = AF_INET,
3322 	};
3323 	struct mbuf *mcp;
3324 	struct ip *encap_ip;
3325 	u_int32_t *reghdr;
3326 	struct ifnet *vifp;
3327 
3328 	s = splsoftnet();
3329 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3330 	    splx(s);
3331 	    if (mrtdebug & DEBUG_PIM)
3332 		log(LOG_DEBUG,
3333 		    "pim_input: register vif not set: %d\n", reg_vif_num);
3334 	    m_freem(m);
3335 	    return;
3336 	}
3337 	/* XXX need refcnt? */
3338 	vifp = viftable[reg_vif_num].v_ifp;
3339 	splx(s);
3340 
3341 	/*
3342 	 * Validate length
3343 	 */
3344 	if (datalen < PIM_REG_MINLEN) {
3345 	    pimstat.pims_rcv_tooshort++;
3346 	    pimstat.pims_rcv_badregisters++;
3347 	    log(LOG_ERR,
3348 		"pim_input: register packet size too small %d from %lx\n",
3349 		datalen, (u_long)ip->ip_src.s_addr);
3350 	    m_freem(m);
3351 	    return;
3352 	}
3353 
3354 	reghdr = (u_int32_t *)(pim + 1);
3355 	encap_ip = (struct ip *)(reghdr + 1);
3356 
3357 	if (mrtdebug & DEBUG_PIM) {
3358 	    log(LOG_DEBUG,
3359 		"pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3360 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3361 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3362 		ntohs(encap_ip->ip_len));
3363 	}
3364 
3365 	/* verify the version number of the inner packet */
3366 	if (encap_ip->ip_v != IPVERSION) {
3367 	    pimstat.pims_rcv_badregisters++;
3368 	    if (mrtdebug & DEBUG_PIM) {
3369 		log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3370 		    "of the inner packet\n", encap_ip->ip_v);
3371 	    }
3372 	    m_freem(m);
3373 	    return;
3374 	}
3375 
3376 	/* verify the inner packet is destined to a mcast group */
3377 	if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3378 	    pimstat.pims_rcv_badregisters++;
3379 	    if (mrtdebug & DEBUG_PIM)
3380 		log(LOG_DEBUG,
3381 		    "pim_input: inner packet of register is not "
3382 		    "multicast %lx\n",
3383 		    (u_long)ntohl(encap_ip->ip_dst.s_addr));
3384 	    m_freem(m);
3385 	    return;
3386 	}
3387 
3388 	/* If a NULL_REGISTER, pass it to the daemon */
3389 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3390 	    goto pim_input_to_daemon;
3391 
3392 	/*
3393 	 * Copy the TOS from the outer IP header to the inner IP header.
3394 	 */
3395 	if (encap_ip->ip_tos != ip_tos) {
3396 	    /* Outer TOS -> inner TOS */
3397 	    encap_ip->ip_tos = ip_tos;
3398 	    /* Recompute the inner header checksum. Sigh... */
3399 
3400 	    /* adjust mbuf to point to the inner IP header */
3401 	    m->m_data += (iphlen + PIM_MINLEN);
3402 	    m->m_len  -= (iphlen + PIM_MINLEN);
3403 
3404 	    encap_ip->ip_sum = 0;
3405 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3406 
3407 	    /* restore mbuf to point back to the outer IP header */
3408 	    m->m_data -= (iphlen + PIM_MINLEN);
3409 	    m->m_len  += (iphlen + PIM_MINLEN);
3410 	}
3411 
3412 	/*
3413 	 * Decapsulate the inner IP packet and loopback to forward it
3414 	 * as a normal multicast packet. Also, make a copy of the
3415 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
3416 	 * to pass to the daemon later, so it can take the appropriate
3417 	 * actions (e.g., send back PIM_REGISTER_STOP).
3418 	 * XXX: here m->m_data points to the outer IP header.
3419 	 */
3420 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_DONTWAIT);
3421 	if (mcp == NULL) {
3422 	    log(LOG_ERR,
3423 		"pim_input: pim register: could not copy register head\n");
3424 	    m_freem(m);
3425 	    return;
3426 	}
3427 
3428 	/* Keep statistics */
3429 	/* XXX: registers_bytes include only the encap. mcast pkt */
3430 	pimstat.pims_rcv_registers_msgs++;
3431 	pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3432 
3433 	/*
3434 	 * forward the inner ip packet; point m_data at the inner ip.
3435 	 */
3436 	m_adj(m, iphlen + PIM_MINLEN);
3437 
3438 	if (mrtdebug & DEBUG_PIM) {
3439 	    log(LOG_DEBUG,
3440 		"pim_input: forwarding decapsulated register: "
3441 		"src %lx, dst %lx, vif %d\n",
3442 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3443 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3444 		reg_vif_num);
3445 	}
3446 	/* NB: vifp was collected above; can it change on us? */
3447 	looutput(vifp, m, (struct sockaddr *)&dst, NULL);
3448 
3449 	/* prepare the register head to send to the mrouting daemon */
3450 	m = mcp;
3451     }
3452 
3453 pim_input_to_daemon:
3454     /*
3455      * Pass the PIM message up to the daemon; if it is a Register message,
3456      * pass the 'head' only up to the daemon. This includes the
3457      * outer IP header, PIM header, PIM-Register header and the
3458      * inner IP header.
3459      * XXX: the outer IP header pkt size of a Register is not adjust to
3460      * reflect the fact that the inner multicast data is truncated.
3461      */
3462     rip_input(m, iphlen, proto);
3463 
3464     return;
3465 }
3466 #endif /* PIM */
3467