xref: /netbsd-src/sys/netinet/ip_mroute.c (revision e89934bbf778a6d6d6894877c4da59d0c7835b0f)
1 /*	$NetBSD: ip_mroute.c,v 1.146 2017/01/24 07:09:24 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 1.146 2017/01/24 07:09:24 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/errno.h>
115 #include <sys/time.h>
116 #include <sys/kernel.h>
117 #include <sys/kmem.h>
118 #include <sys/ioctl.h>
119 #include <sys/syslog.h>
120 
121 #include <net/if.h>
122 #include <net/raw_cb.h>
123 
124 #include <netinet/in.h>
125 #include <netinet/in_var.h>
126 #include <netinet/in_systm.h>
127 #include <netinet/ip.h>
128 #include <netinet/ip_var.h>
129 #include <netinet/in_pcb.h>
130 #include <netinet/udp.h>
131 #include <netinet/igmp.h>
132 #include <netinet/igmp_var.h>
133 #include <netinet/ip_mroute.h>
134 #ifdef PIM
135 #include <netinet/pim.h>
136 #include <netinet/pim_var.h>
137 #endif
138 #include <netinet/ip_encap.h>
139 
140 #ifdef IPSEC
141 #include <netipsec/ipsec.h>
142 #include <netipsec/key.h>
143 #endif
144 
145 #define IP_MULTICASTOPTS 0
146 #define	M_PULLUP(m, len)						 \
147 	do {								 \
148 		if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
149 			(m) = m_pullup((m), (len));			 \
150 	} while (/*CONSTCOND*/ 0)
151 
152 /*
153  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
154  * except for netstat or debugging purposes.
155  */
156 struct socket  *ip_mrouter  = NULL;
157 int		ip_mrtproto = IGMP_DVMRP;    /* for netstat only */
158 
159 #define NO_RTE_FOUND 	0x1
160 #define RTE_FOUND	0x2
161 
162 #define	MFCHASH(a, g)							\
163 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^	\
164 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
165 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
166 u_long	mfchash;
167 
168 u_char		nexpire[MFCTBLSIZ];
169 struct vif	viftable[MAXVIFS];
170 struct mrtstat	mrtstat;
171 u_int		mrtdebug = 0;	  /* debug level 	*/
172 #define		DEBUG_MFC	0x02
173 #define		DEBUG_FORWARD	0x04
174 #define		DEBUG_EXPIRE	0x08
175 #define		DEBUG_XMIT	0x10
176 #define		DEBUG_PIM	0x20
177 
178 #define		VIFI_INVALID	((vifi_t) -1)
179 
180 u_int       	tbfdebug = 0;     /* tbf debug level 	*/
181 #ifdef RSVP_ISI
182 u_int		rsvpdebug = 0;	  /* rsvp debug level   */
183 #define	RSVP_DPRINTF(a)	do if (rsvpdebug) printf a; while (/*CONSTCOND*/0)
184 extern struct socket *ip_rsvpd;
185 extern int rsvp_on;
186 #else
187 #define	RSVP_DPRINTF(a)	do {} while (/*CONSTCOND*/0)
188 #endif /* RSVP_ISI */
189 
190 /* vif attachment using sys/netinet/ip_encap.c */
191 static void vif_input(struct mbuf *, int, int);
192 static int vif_encapcheck(struct mbuf *, int, int, void *);
193 
194 static const struct encapsw vif_encapsw = {
195 	.encapsw4 = {
196 		.pr_input	= vif_input,
197 		.pr_ctlinput	= NULL,
198 	}
199 };
200 
201 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
202 #define		UPCALL_EXPIRE	6		/* number of timeouts */
203 
204 /*
205  * Define the token bucket filter structures
206  */
207 
208 #define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
209 
210 static int get_sg_cnt(struct sioc_sg_req *);
211 static int get_vif_cnt(struct sioc_vif_req *);
212 static int ip_mrouter_init(struct socket *, int);
213 static int set_assert(int);
214 static int add_vif(struct vifctl *);
215 static int del_vif(vifi_t *);
216 static void update_mfc_params(struct mfc *, struct mfcctl2 *);
217 static void init_mfc_params(struct mfc *, struct mfcctl2 *);
218 static void expire_mfc(struct mfc *);
219 static int add_mfc(struct sockopt *);
220 #ifdef UPCALL_TIMING
221 static void collate(struct timeval *);
222 #endif
223 static int del_mfc(struct sockopt *);
224 static int set_api_config(struct sockopt *); /* chose API capabilities */
225 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
226 static void expire_upcalls(void *);
227 #ifdef RSVP_ISI
228 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
229 #else
230 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *);
231 #endif
232 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
233 static void encap_send(struct ip *, struct vif *, struct mbuf *);
234 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_int32_t);
235 static void tbf_queue(struct vif *, struct mbuf *);
236 static void tbf_process_q(struct vif *);
237 static void tbf_reprocess_q(void *);
238 static int tbf_dq_sel(struct vif *, struct ip *);
239 static void tbf_send_packet(struct vif *, struct mbuf *);
240 static void tbf_update_tokens(struct vif *);
241 static int priority(struct vif *, struct ip *);
242 
243 /*
244  * Bandwidth monitoring
245  */
246 static void free_bw_list(struct bw_meter *);
247 static int add_bw_upcall(struct bw_upcall *);
248 static int del_bw_upcall(struct bw_upcall *);
249 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *);
250 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
251 static void bw_upcalls_send(void);
252 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
253 static void unschedule_bw_meter(struct bw_meter *);
254 static void bw_meter_process(void);
255 static void expire_bw_upcalls_send(void *);
256 static void expire_bw_meter_process(void *);
257 
258 #ifdef PIM
259 static int pim_register_send(struct ip *, struct vif *,
260 		struct mbuf *, struct mfc *);
261 static int pim_register_send_rp(struct ip *, struct vif *,
262 		struct mbuf *, struct mfc *);
263 static int pim_register_send_upcall(struct ip *, struct vif *,
264 		struct mbuf *, struct mfc *);
265 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
266 #endif
267 
268 /*
269  * 'Interfaces' associated with decapsulator (so we can tell
270  * packets that went through it from ones that get reflected
271  * by a broken gateway).  These interfaces are never linked into
272  * the system ifnet list & no routes point to them.  I.e., packets
273  * can't be sent this way.  They only exist as a placeholder for
274  * multicast source verification.
275  */
276 #if 0
277 struct ifnet multicast_decap_if[MAXVIFS];
278 #endif
279 
280 #define	ENCAP_TTL	64
281 #define	ENCAP_PROTO	IPPROTO_IPIP	/* 4 */
282 
283 /* prototype IP hdr for encapsulated packets */
284 struct ip multicast_encap_iphdr = {
285 	.ip_hl = sizeof(struct ip) >> 2,
286 	.ip_v = IPVERSION,
287 	.ip_len = sizeof(struct ip),
288 	.ip_ttl = ENCAP_TTL,
289 	.ip_p = ENCAP_PROTO,
290 };
291 
292 /*
293  * Bandwidth meter variables and constants
294  */
295 
296 /*
297  * Pending timeouts are stored in a hash table, the key being the
298  * expiration time. Periodically, the entries are analysed and processed.
299  */
300 #define BW_METER_BUCKETS	1024
301 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
302 struct callout bw_meter_ch;
303 #define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
304 
305 /*
306  * Pending upcalls are stored in a vector which is flushed when
307  * full, or periodically
308  */
309 static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
310 static u_int	bw_upcalls_n; /* # of pending upcalls */
311 struct callout	bw_upcalls_ch;
312 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
313 
314 #ifdef PIM
315 struct pimstat pimstat;
316 
317 /*
318  * Note: the PIM Register encapsulation adds the following in front of a
319  * data packet:
320  *
321  * struct pim_encap_hdr {
322  *    struct ip ip;
323  *    struct pim_encap_pimhdr  pim;
324  * }
325  *
326  */
327 
328 struct pim_encap_pimhdr {
329 	struct pim pim;
330 	uint32_t   flags;
331 };
332 
333 static struct ip pim_encap_iphdr = {
334 	.ip_v = IPVERSION,
335 	.ip_hl = sizeof(struct ip) >> 2,
336 	.ip_len = sizeof(struct ip),
337 	.ip_ttl = ENCAP_TTL,
338 	.ip_p = IPPROTO_PIM,
339 };
340 
341 static struct pim_encap_pimhdr pim_encap_pimhdr = {
342     {
343 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
344 	0,			/* reserved */
345 	0,			/* checksum */
346     },
347     0				/* flags */
348 };
349 
350 static struct ifnet multicast_register_if;
351 static vifi_t reg_vif_num = VIFI_INVALID;
352 #endif /* PIM */
353 
354 
355 /*
356  * Private variables.
357  */
358 static vifi_t	   numvifs = 0;
359 
360 static struct callout expire_upcalls_ch;
361 
362 /*
363  * whether or not special PIM assert processing is enabled.
364  */
365 static int pim_assert;
366 /*
367  * Rate limit for assert notification messages, in usec
368  */
369 #define ASSERT_MSG_TIME		3000000
370 
371 /*
372  * Kernel multicast routing API capabilities and setup.
373  * If more API capabilities are added to the kernel, they should be
374  * recorded in `mrt_api_support'.
375  */
376 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
377 					  MRT_MFC_FLAGS_BORDER_VIF |
378 					  MRT_MFC_RP |
379 					  MRT_MFC_BW_UPCALL);
380 static u_int32_t mrt_api_config = 0;
381 
382 /*
383  * Find a route for a given origin IP address and Multicast group address
384  * Type of service parameter to be added in the future!!!
385  * Statistics are updated by the caller if needed
386  * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
387  */
388 static struct mfc *
389 mfc_find(struct in_addr *o, struct in_addr *g)
390 {
391 	struct mfc *rt;
392 
393 	LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
394 		if (in_hosteq(rt->mfc_origin, *o) &&
395 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
396 		    (rt->mfc_stall == NULL))
397 			break;
398 	}
399 
400 	return (rt);
401 }
402 
403 /*
404  * Macros to compute elapsed time efficiently
405  * Borrowed from Van Jacobson's scheduling code
406  */
407 #define TV_DELTA(a, b, delta) do {					\
408 	int xxs;							\
409 	delta = (a).tv_usec - (b).tv_usec;				\
410 	xxs = (a).tv_sec - (b).tv_sec;					\
411 	switch (xxs) {							\
412 	case 2:								\
413 		delta += 1000000;					\
414 		/* fall through */					\
415 	case 1:								\
416 		delta += 1000000;					\
417 		/* fall through */					\
418 	case 0:								\
419 		break;							\
420 	default:							\
421 		delta += (1000000 * xxs);				\
422 		break;							\
423 	}								\
424 } while (/*CONSTCOND*/ 0)
425 
426 #ifdef UPCALL_TIMING
427 u_int32_t upcall_data[51];
428 #endif /* UPCALL_TIMING */
429 
430 /*
431  * Handle MRT setsockopt commands to modify the multicast routing tables.
432  */
433 int
434 ip_mrouter_set(struct socket *so, struct sockopt *sopt)
435 {
436 	int error;
437 	int optval;
438 	struct vifctl vifc;
439 	vifi_t vifi;
440 	struct bw_upcall bwuc;
441 
442 	if (sopt->sopt_name != MRT_INIT && so != ip_mrouter)
443 		error = ENOPROTOOPT;
444 	else {
445 		switch (sopt->sopt_name) {
446 		case MRT_INIT:
447 			error = sockopt_getint(sopt, &optval);
448 			if (error)
449 				break;
450 
451 			error = ip_mrouter_init(so, optval);
452 			break;
453 		case MRT_DONE:
454 			error = ip_mrouter_done();
455 			break;
456 		case MRT_ADD_VIF:
457 			error = sockopt_get(sopt, &vifc, sizeof(vifc));
458 			if (error)
459 				break;
460 			error = add_vif(&vifc);
461 			break;
462 		case MRT_DEL_VIF:
463 			error = sockopt_get(sopt, &vifi, sizeof(vifi));
464 			if (error)
465 				break;
466 			error = del_vif(&vifi);
467 			break;
468 		case MRT_ADD_MFC:
469 			error = add_mfc(sopt);
470 			break;
471 		case MRT_DEL_MFC:
472 			error = del_mfc(sopt);
473 			break;
474 		case MRT_ASSERT:
475 			error = sockopt_getint(sopt, &optval);
476 			if (error)
477 				break;
478 			error = set_assert(optval);
479 			break;
480 		case MRT_API_CONFIG:
481 			error = set_api_config(sopt);
482 			break;
483 		case MRT_ADD_BW_UPCALL:
484 			error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
485 			if (error)
486 				break;
487 			error = add_bw_upcall(&bwuc);
488 			break;
489 		case MRT_DEL_BW_UPCALL:
490 			error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
491 			if (error)
492 				break;
493 			error = del_bw_upcall(&bwuc);
494 			break;
495 		default:
496 			error = ENOPROTOOPT;
497 			break;
498 		}
499 	}
500 	return (error);
501 }
502 
503 /*
504  * Handle MRT getsockopt commands
505  */
506 int
507 ip_mrouter_get(struct socket *so, struct sockopt *sopt)
508 {
509 	int error;
510 
511 	if (so != ip_mrouter)
512 		error = ENOPROTOOPT;
513 	else {
514 		switch (sopt->sopt_name) {
515 		case MRT_VERSION:
516 			error = sockopt_setint(sopt, 0x0305); /* XXX !!!! */
517 			break;
518 		case MRT_ASSERT:
519 			error = sockopt_setint(sopt, pim_assert);
520 			break;
521 		case MRT_API_SUPPORT:
522 			error = sockopt_set(sopt, &mrt_api_support,
523 			    sizeof(mrt_api_support));
524 			break;
525 		case MRT_API_CONFIG:
526 			error = sockopt_set(sopt, &mrt_api_config,
527 			    sizeof(mrt_api_config));
528 			break;
529 		default:
530 			error = ENOPROTOOPT;
531 			break;
532 		}
533 	}
534 	return (error);
535 }
536 
537 /*
538  * Handle ioctl commands to obtain information from the cache
539  */
540 int
541 mrt_ioctl(struct socket *so, u_long cmd, void *data)
542 {
543 	int error;
544 
545 	if (so != ip_mrouter)
546 		error = EINVAL;
547 	else
548 		switch (cmd) {
549 		case SIOCGETVIFCNT:
550 			error = get_vif_cnt((struct sioc_vif_req *)data);
551 			break;
552 		case SIOCGETSGCNT:
553 			error = get_sg_cnt((struct sioc_sg_req *)data);
554 			break;
555 		default:
556 			error = EINVAL;
557 			break;
558 		}
559 
560 	return (error);
561 }
562 
563 /*
564  * returns the packet, byte, rpf-failure count for the source group provided
565  */
566 static int
567 get_sg_cnt(struct sioc_sg_req *req)
568 {
569 	int s;
570 	struct mfc *rt;
571 
572 	s = splsoftnet();
573 	rt = mfc_find(&req->src, &req->grp);
574 	if (rt == NULL) {
575 		splx(s);
576 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
577 		return (EADDRNOTAVAIL);
578 	}
579 	req->pktcnt = rt->mfc_pkt_cnt;
580 	req->bytecnt = rt->mfc_byte_cnt;
581 	req->wrong_if = rt->mfc_wrong_if;
582 	splx(s);
583 
584 	return (0);
585 }
586 
587 /*
588  * returns the input and output packet and byte counts on the vif provided
589  */
590 static int
591 get_vif_cnt(struct sioc_vif_req *req)
592 {
593 	vifi_t vifi = req->vifi;
594 
595 	if (vifi >= numvifs)
596 		return (EINVAL);
597 
598 	req->icount = viftable[vifi].v_pkt_in;
599 	req->ocount = viftable[vifi].v_pkt_out;
600 	req->ibytes = viftable[vifi].v_bytes_in;
601 	req->obytes = viftable[vifi].v_bytes_out;
602 
603 	return (0);
604 }
605 
606 /*
607  * Enable multicast routing
608  */
609 static int
610 ip_mrouter_init(struct socket *so, int v)
611 {
612 	if (mrtdebug)
613 		log(LOG_DEBUG,
614 		    "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
615 		    so->so_type, so->so_proto->pr_protocol);
616 
617 	if (so->so_type != SOCK_RAW ||
618 	    so->so_proto->pr_protocol != IPPROTO_IGMP)
619 		return (EOPNOTSUPP);
620 
621 	if (v != 1)
622 		return (EINVAL);
623 
624 	if (ip_mrouter != NULL)
625 		return (EADDRINUSE);
626 
627 	ip_mrouter = so;
628 
629 	mfchashtbl = hashinit(MFCTBLSIZ, HASH_LIST, true, &mfchash);
630 	memset((void *)nexpire, 0, sizeof(nexpire));
631 
632 	pim_assert = 0;
633 
634 	callout_init(&expire_upcalls_ch, 0);
635 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
636 		      expire_upcalls, NULL);
637 
638 	callout_init(&bw_upcalls_ch, 0);
639 	callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
640 		      expire_bw_upcalls_send, NULL);
641 
642 	callout_init(&bw_meter_ch, 0);
643 	callout_reset(&bw_meter_ch, BW_METER_PERIOD,
644 		      expire_bw_meter_process, NULL);
645 
646 	if (mrtdebug)
647 		log(LOG_DEBUG, "ip_mrouter_init\n");
648 
649 	return (0);
650 }
651 
652 /*
653  * Disable multicast routing
654  */
655 int
656 ip_mrouter_done(void)
657 {
658 	vifi_t vifi;
659 	struct vif *vifp;
660 	int i;
661 	int s;
662 
663 	s = splsoftnet();
664 
665 	/* Clear out all the vifs currently in use. */
666 	for (vifi = 0; vifi < numvifs; vifi++) {
667 		vifp = &viftable[vifi];
668 		if (!in_nullhost(vifp->v_lcl_addr))
669 			reset_vif(vifp);
670 	}
671 
672 	numvifs = 0;
673 	pim_assert = 0;
674 	mrt_api_config = 0;
675 
676 	callout_stop(&expire_upcalls_ch);
677 	callout_stop(&bw_upcalls_ch);
678 	callout_stop(&bw_meter_ch);
679 
680 	/*
681 	 * Free all multicast forwarding cache entries.
682 	 */
683 	for (i = 0; i < MFCTBLSIZ; i++) {
684 		struct mfc *rt, *nrt;
685 
686 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
687 			nrt = LIST_NEXT(rt, mfc_hash);
688 
689 			expire_mfc(rt);
690 		}
691 	}
692 
693 	memset((void *)nexpire, 0, sizeof(nexpire));
694 	hashdone(mfchashtbl, HASH_LIST, mfchash);
695 	mfchashtbl = NULL;
696 
697 	bw_upcalls_n = 0;
698 	memset(bw_meter_timers, 0, sizeof(bw_meter_timers));
699 
700 	/* Reset de-encapsulation cache. */
701 
702 	ip_mrouter = NULL;
703 
704 	splx(s);
705 
706 	if (mrtdebug)
707 		log(LOG_DEBUG, "ip_mrouter_done\n");
708 
709 	return (0);
710 }
711 
712 void
713 ip_mrouter_detach(struct ifnet *ifp)
714 {
715 	int vifi, i;
716 	struct vif *vifp;
717 	struct mfc *rt;
718 	struct rtdetq *rte;
719 
720 	/* XXX not sure about side effect to userland routing daemon */
721 	for (vifi = 0; vifi < numvifs; vifi++) {
722 		vifp = &viftable[vifi];
723 		if (vifp->v_ifp == ifp)
724 			reset_vif(vifp);
725 	}
726 	for (i = 0; i < MFCTBLSIZ; i++) {
727 		if (nexpire[i] == 0)
728 			continue;
729 		LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
730 			for (rte = rt->mfc_stall; rte; rte = rte->next) {
731 				if (rte->ifp == ifp)
732 					rte->ifp = NULL;
733 			}
734 		}
735 	}
736 }
737 
738 /*
739  * Set PIM assert processing global
740  */
741 static int
742 set_assert(int i)
743 {
744 	pim_assert = !!i;
745 	return (0);
746 }
747 
748 /*
749  * Configure API capabilities
750  */
751 static int
752 set_api_config(struct sockopt *sopt)
753 {
754 	u_int32_t apival;
755 	int i, error;
756 
757 	/*
758 	 * We can set the API capabilities only if it is the first operation
759 	 * after MRT_INIT. I.e.:
760 	 *  - there are no vifs installed
761 	 *  - pim_assert is not enabled
762 	 *  - the MFC table is empty
763 	 */
764 	error = sockopt_get(sopt, &apival, sizeof(apival));
765 	if (error)
766 		return (error);
767 	if (numvifs > 0)
768 		return (EPERM);
769 	if (pim_assert)
770 		return (EPERM);
771 	for (i = 0; i < MFCTBLSIZ; i++) {
772 		if (LIST_FIRST(&mfchashtbl[i]) != NULL)
773 			return (EPERM);
774 	}
775 
776 	mrt_api_config = apival & mrt_api_support;
777 	return (0);
778 }
779 
780 /*
781  * Add a vif to the vif table
782  */
783 static int
784 add_vif(struct vifctl *vifcp)
785 {
786 	struct vif *vifp;
787 	struct ifnet *ifp;
788 	int error, s;
789 	struct sockaddr_in sin;
790 
791 	if (vifcp->vifc_vifi >= MAXVIFS)
792 		return (EINVAL);
793 	if (in_nullhost(vifcp->vifc_lcl_addr))
794 		return (EADDRNOTAVAIL);
795 
796 	vifp = &viftable[vifcp->vifc_vifi];
797 	if (!in_nullhost(vifp->v_lcl_addr))
798 		return (EADDRINUSE);
799 
800 	/* Find the interface with an address in AF_INET family. */
801 #ifdef PIM
802 	if (vifcp->vifc_flags & VIFF_REGISTER) {
803 		/*
804 		 * XXX: Because VIFF_REGISTER does not really need a valid
805 		 * local interface (e.g. it could be 127.0.0.2), we don't
806 		 * check its address.
807 		 */
808 	    ifp = NULL;
809 	} else
810 #endif
811 	{
812 		struct ifaddr *ifa;
813 
814 		sockaddr_in_init(&sin, &vifcp->vifc_lcl_addr, 0);
815 		s = pserialize_read_enter();
816 		ifa = ifa_ifwithaddr(sintosa(&sin));
817 		if (ifa == NULL) {
818 			pserialize_read_exit(s);
819 			return EADDRNOTAVAIL;
820 		}
821 		ifp = ifa->ifa_ifp;
822 		/* FIXME NOMPSAFE */
823 		pserialize_read_exit(s);
824 	}
825 
826 	if (vifcp->vifc_flags & VIFF_TUNNEL) {
827 		if (vifcp->vifc_flags & VIFF_SRCRT) {
828 			log(LOG_ERR, "source routed tunnels not supported\n");
829 			return (EOPNOTSUPP);
830 		}
831 
832 		/* attach this vif to decapsulator dispatch table */
833 		/*
834 		 * XXX Use addresses in registration so that matching
835 		 * can be done with radix tree in decapsulator.  But,
836 		 * we need to check inner header for multicast, so
837 		 * this requires both radix tree lookup and then a
838 		 * function to check, and this is not supported yet.
839 		 */
840 		error = encap_lock_enter();
841 		if (error)
842 			return error;
843 		vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
844 		    vif_encapcheck, &vif_encapsw, vifp);
845 		encap_lock_exit();
846 		if (!vifp->v_encap_cookie)
847 			return (EINVAL);
848 
849 		/* Create a fake encapsulation interface. */
850 		ifp = malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK|M_ZERO);
851 		snprintf(ifp->if_xname, sizeof(ifp->if_xname),
852 			 "mdecap%d", vifcp->vifc_vifi);
853 
854 		/* Prepare cached route entry. */
855 		memset(&vifp->v_route, 0, sizeof(vifp->v_route));
856 #ifdef PIM
857 	} else if (vifcp->vifc_flags & VIFF_REGISTER) {
858 		ifp = &multicast_register_if;
859 		if (mrtdebug)
860 			log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
861 			    (void *)ifp);
862 		if (reg_vif_num == VIFI_INVALID) {
863 			memset(ifp, 0, sizeof(*ifp));
864 			snprintf(ifp->if_xname, sizeof(ifp->if_xname),
865 				 "register_vif");
866 			ifp->if_flags = IFF_LOOPBACK;
867 			memset(&vifp->v_route, 0, sizeof(vifp->v_route));
868 			reg_vif_num = vifcp->vifc_vifi;
869 		}
870 #endif
871 	} else {
872 		/* Make sure the interface supports multicast. */
873 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
874 			return (EOPNOTSUPP);
875 
876 		/* Enable promiscuous reception of all IP multicasts. */
877 		sockaddr_in_init(&sin, &zeroin_addr, 0);
878 		error = if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin));
879 		if (error)
880 			return (error);
881 	}
882 
883 	s = splsoftnet();
884 
885 	/* Define parameters for the tbf structure. */
886 	vifp->tbf_q = NULL;
887 	vifp->tbf_t = &vifp->tbf_q;
888 	microtime(&vifp->tbf_last_pkt_t);
889 	vifp->tbf_n_tok = 0;
890 	vifp->tbf_q_len = 0;
891 	vifp->tbf_max_q_len = MAXQSIZE;
892 
893 	vifp->v_flags = vifcp->vifc_flags;
894 	vifp->v_threshold = vifcp->vifc_threshold;
895 	/* scaling up here allows division by 1024 in critical code */
896 	vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
897 	vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
898 	vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
899 	vifp->v_ifp = ifp;
900 	/* Initialize per vif pkt counters. */
901 	vifp->v_pkt_in = 0;
902 	vifp->v_pkt_out = 0;
903 	vifp->v_bytes_in = 0;
904 	vifp->v_bytes_out = 0;
905 
906 	callout_init(&vifp->v_repq_ch, 0);
907 
908 #ifdef RSVP_ISI
909 	vifp->v_rsvp_on = 0;
910 	vifp->v_rsvpd = NULL;
911 #endif /* RSVP_ISI */
912 
913 	splx(s);
914 
915 	/* Adjust numvifs up if the vifi is higher than numvifs. */
916 	if (numvifs <= vifcp->vifc_vifi)
917 		numvifs = vifcp->vifc_vifi + 1;
918 
919 	if (mrtdebug)
920 		log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
921 		    vifcp->vifc_vifi,
922 		    ntohl(vifcp->vifc_lcl_addr.s_addr),
923 		    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
924 		    ntohl(vifcp->vifc_rmt_addr.s_addr),
925 		    vifcp->vifc_threshold,
926 		    vifcp->vifc_rate_limit);
927 
928 	return (0);
929 }
930 
931 void
932 reset_vif(struct vif *vifp)
933 {
934 	struct mbuf *m, *n;
935 	struct ifnet *ifp;
936 	struct sockaddr_in sin;
937 
938 	callout_stop(&vifp->v_repq_ch);
939 
940 	/* detach this vif from decapsulator dispatch table */
941 	encap_lock_enter();
942 	encap_detach(vifp->v_encap_cookie);
943 	encap_lock_exit();
944 	vifp->v_encap_cookie = NULL;
945 
946 	/*
947 	 * Free packets queued at the interface
948 	 */
949 	for (m = vifp->tbf_q; m != NULL; m = n) {
950 		n = m->m_nextpkt;
951 		m_freem(m);
952 	}
953 
954 	if (vifp->v_flags & VIFF_TUNNEL)
955 		free(vifp->v_ifp, M_MRTABLE);
956 	else if (vifp->v_flags & VIFF_REGISTER) {
957 #ifdef PIM
958 		reg_vif_num = VIFI_INVALID;
959 #endif
960 	} else {
961 		sockaddr_in_init(&sin, &zeroin_addr, 0);
962 		ifp = vifp->v_ifp;
963 		if_mcast_op(ifp, SIOCDELMULTI, sintosa(&sin));
964 	}
965 	memset((void *)vifp, 0, sizeof(*vifp));
966 }
967 
968 /*
969  * Delete a vif from the vif table
970  */
971 static int
972 del_vif(vifi_t *vifip)
973 {
974 	struct vif *vifp;
975 	vifi_t vifi;
976 	int s;
977 
978 	if (*vifip >= numvifs)
979 		return (EINVAL);
980 
981 	vifp = &viftable[*vifip];
982 	if (in_nullhost(vifp->v_lcl_addr))
983 		return (EADDRNOTAVAIL);
984 
985 	s = splsoftnet();
986 
987 	reset_vif(vifp);
988 
989 	/* Adjust numvifs down */
990 	for (vifi = numvifs; vifi > 0; vifi--)
991 		if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
992 			break;
993 	numvifs = vifi;
994 
995 	splx(s);
996 
997 	if (mrtdebug)
998 		log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
999 
1000 	return (0);
1001 }
1002 
1003 /*
1004  * update an mfc entry without resetting counters and S,G addresses.
1005  */
1006 static void
1007 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1008 {
1009 	int i;
1010 
1011 	rt->mfc_parent = mfccp->mfcc_parent;
1012 	for (i = 0; i < numvifs; i++) {
1013 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1014 		rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1015 			MRT_MFC_FLAGS_ALL;
1016 	}
1017 	/* set the RP address */
1018 	if (mrt_api_config & MRT_MFC_RP)
1019 		rt->mfc_rp = mfccp->mfcc_rp;
1020 	else
1021 		rt->mfc_rp = zeroin_addr;
1022 }
1023 
1024 /*
1025  * fully initialize an mfc entry from the parameter.
1026  */
1027 static void
1028 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1029 {
1030 	rt->mfc_origin     = mfccp->mfcc_origin;
1031 	rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1032 
1033 	update_mfc_params(rt, mfccp);
1034 
1035 	/* initialize pkt counters per src-grp */
1036 	rt->mfc_pkt_cnt    = 0;
1037 	rt->mfc_byte_cnt   = 0;
1038 	rt->mfc_wrong_if   = 0;
1039 	timerclear(&rt->mfc_last_assert);
1040 }
1041 
1042 static void
1043 expire_mfc(struct mfc *rt)
1044 {
1045 	struct rtdetq *rte, *nrte;
1046 
1047 	free_bw_list(rt->mfc_bw_meter);
1048 
1049 	for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1050 		nrte = rte->next;
1051 		m_freem(rte->m);
1052 		free(rte, M_MRTABLE);
1053 	}
1054 
1055 	LIST_REMOVE(rt, mfc_hash);
1056 	free(rt, M_MRTABLE);
1057 }
1058 
1059 /*
1060  * Add an mfc entry
1061  */
1062 static int
1063 add_mfc(struct sockopt *sopt)
1064 {
1065 	struct mfcctl2 mfcctl2;
1066 	struct mfcctl2 *mfccp;
1067 	struct mfc *rt;
1068 	u_int32_t hash = 0;
1069 	struct rtdetq *rte, *nrte;
1070 	u_short nstl;
1071 	int s;
1072 	int error;
1073 
1074 	/*
1075 	 * select data size depending on API version.
1076 	 */
1077 	mfccp = &mfcctl2;
1078 	memset(&mfcctl2, 0, sizeof(mfcctl2));
1079 
1080 	if (mrt_api_config & MRT_API_FLAGS_ALL)
1081 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1082 	else
1083 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1084 
1085 	if (error)
1086 		return (error);
1087 
1088 	s = splsoftnet();
1089 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1090 
1091 	/* If an entry already exists, just update the fields */
1092 	if (rt) {
1093 		if (mrtdebug & DEBUG_MFC)
1094 			log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1095 			    ntohl(mfccp->mfcc_origin.s_addr),
1096 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1097 			    mfccp->mfcc_parent);
1098 
1099 		update_mfc_params(rt, mfccp);
1100 
1101 		splx(s);
1102 		return (0);
1103 	}
1104 
1105 	/*
1106 	 * Find the entry for which the upcall was made and update
1107 	 */
1108 	nstl = 0;
1109 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1110 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1111 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1112 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1113 		    rt->mfc_stall != NULL) {
1114 			if (nstl++)
1115 				log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
1116 				    "multiple kernel entries",
1117 				    ntohl(mfccp->mfcc_origin.s_addr),
1118 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1119 				    mfccp->mfcc_parent, rt->mfc_stall);
1120 
1121 			if (mrtdebug & DEBUG_MFC)
1122 				log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n",
1123 				    ntohl(mfccp->mfcc_origin.s_addr),
1124 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1125 				    mfccp->mfcc_parent, rt->mfc_stall);
1126 
1127 			rte = rt->mfc_stall;
1128 			init_mfc_params(rt, mfccp);
1129 			rt->mfc_stall = NULL;
1130 
1131 			rt->mfc_expire = 0; /* Don't clean this guy up */
1132 			nexpire[hash]--;
1133 
1134 			/* free packets Qed at the end of this entry */
1135 			for (; rte != NULL; rte = nrte) {
1136 				nrte = rte->next;
1137 				if (rte->ifp) {
1138 #ifdef RSVP_ISI
1139 					ip_mdq(rte->m, rte->ifp, rt, -1);
1140 #else
1141 					ip_mdq(rte->m, rte->ifp, rt);
1142 #endif /* RSVP_ISI */
1143 				}
1144 				m_freem(rte->m);
1145 #ifdef UPCALL_TIMING
1146 				collate(&rte->t);
1147 #endif /* UPCALL_TIMING */
1148 				free(rte, M_MRTABLE);
1149 			}
1150 		}
1151 	}
1152 
1153 	/*
1154 	 * It is possible that an entry is being inserted without an upcall
1155 	 */
1156 	if (nstl == 0) {
1157 		/*
1158 		 * No mfc; make a new one
1159 		 */
1160 		if (mrtdebug & DEBUG_MFC)
1161 			log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1162 			    ntohl(mfccp->mfcc_origin.s_addr),
1163 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1164 			    mfccp->mfcc_parent);
1165 
1166 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1167 			if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1168 			    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1169 				init_mfc_params(rt, mfccp);
1170 				if (rt->mfc_expire)
1171 					nexpire[hash]--;
1172 				rt->mfc_expire = 0;
1173 				break; /* XXX */
1174 			}
1175 		}
1176 		if (rt == NULL) {	/* no upcall, so make a new entry */
1177 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1178 						  M_NOWAIT);
1179 			if (rt == NULL) {
1180 				splx(s);
1181 				return (ENOBUFS);
1182 			}
1183 
1184 			init_mfc_params(rt, mfccp);
1185 			rt->mfc_expire	= 0;
1186 			rt->mfc_stall	= NULL;
1187 			rt->mfc_bw_meter = NULL;
1188 
1189 			/* insert new entry at head of hash chain */
1190 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1191 		}
1192 	}
1193 
1194 	splx(s);
1195 	return (0);
1196 }
1197 
1198 #ifdef UPCALL_TIMING
1199 /*
1200  * collect delay statistics on the upcalls
1201  */
1202 static void
1203 collate(struct timeval *t)
1204 {
1205 	u_int32_t d;
1206 	struct timeval tp;
1207 	u_int32_t delta;
1208 
1209 	microtime(&tp);
1210 
1211 	if (timercmp(t, &tp, <)) {
1212 		TV_DELTA(tp, *t, delta);
1213 
1214 		d = delta >> 10;
1215 		if (d > 50)
1216 			d = 50;
1217 
1218 		++upcall_data[d];
1219 	}
1220 }
1221 #endif /* UPCALL_TIMING */
1222 
1223 /*
1224  * Delete an mfc entry
1225  */
1226 static int
1227 del_mfc(struct sockopt *sopt)
1228 {
1229 	struct mfcctl2 mfcctl2;
1230 	struct mfcctl2 *mfccp;
1231 	struct mfc *rt;
1232 	int s;
1233 	int error;
1234 
1235 	/*
1236 	 * XXX: for deleting MFC entries the information in entries
1237 	 * of size "struct mfcctl" is sufficient.
1238 	 */
1239 
1240 	mfccp = &mfcctl2;
1241 	memset(&mfcctl2, 0, sizeof(mfcctl2));
1242 
1243 	error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1244 	if (error) {
1245 		/* Try with the size of mfcctl2. */
1246 		error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1247 		if (error)
1248 			return (error);
1249 	}
1250 
1251 	if (mrtdebug & DEBUG_MFC)
1252 		log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1253 		    ntohl(mfccp->mfcc_origin.s_addr),
1254 		    ntohl(mfccp->mfcc_mcastgrp.s_addr));
1255 
1256 	s = splsoftnet();
1257 
1258 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1259 	if (rt == NULL) {
1260 		splx(s);
1261 		return (EADDRNOTAVAIL);
1262 	}
1263 
1264 	/*
1265 	 * free the bw_meter entries
1266 	 */
1267 	free_bw_list(rt->mfc_bw_meter);
1268 	rt->mfc_bw_meter = NULL;
1269 
1270 	LIST_REMOVE(rt, mfc_hash);
1271 	free(rt, M_MRTABLE);
1272 
1273 	splx(s);
1274 	return (0);
1275 }
1276 
1277 static int
1278 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1279 {
1280 	if (s) {
1281 		if (sbappendaddr(&s->so_rcv, sintosa(src), mm, NULL) != 0) {
1282 			sorwakeup(s);
1283 			return (0);
1284 		}
1285 	}
1286 	m_freem(mm);
1287 	return (-1);
1288 }
1289 
1290 /*
1291  * IP multicast forwarding function. This function assumes that the packet
1292  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1293  * pointed to by "ifp", and the packet is to be relayed to other networks
1294  * that have members of the packet's destination IP multicast group.
1295  *
1296  * The packet is returned unscathed to the caller, unless it is
1297  * erroneous, in which case a non-zero return value tells the caller to
1298  * discard it.
1299  */
1300 
1301 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1302 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1303 
1304 int
1305 #ifdef RSVP_ISI
1306 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo)
1307 #else
1308 ip_mforward(struct mbuf *m, struct ifnet *ifp)
1309 #endif /* RSVP_ISI */
1310 {
1311 	struct ip *ip = mtod(m, struct ip *);
1312 	struct mfc *rt;
1313 	static int srctun = 0;
1314 	struct mbuf *mm;
1315 	struct sockaddr_in sin;
1316 	int s;
1317 	vifi_t vifi;
1318 
1319 	if (mrtdebug & DEBUG_FORWARD)
1320 		log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1321 		    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1322 
1323 	if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1324 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1325 		/*
1326 		 * Packet arrived via a physical interface or
1327 		 * an encapsulated tunnel or a register_vif.
1328 		 */
1329 	} else {
1330 		/*
1331 		 * Packet arrived through a source-route tunnel.
1332 		 * Source-route tunnels are no longer supported.
1333 		 */
1334 		if ((srctun++ % 1000) == 0)
1335 			log(LOG_ERR,
1336 			    "ip_mforward: received source-routed packet from %x\n",
1337 			    ntohl(ip->ip_src.s_addr));
1338 
1339 		return (1);
1340 	}
1341 
1342 	/*
1343 	 * Clear any in-bound checksum flags for this packet.
1344 	 */
1345 	m->m_pkthdr.csum_flags = 0;
1346 
1347 #ifdef RSVP_ISI
1348 	if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1349 		if (ip->ip_ttl < MAXTTL)
1350 			ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1351 		if (ip->ip_p == IPPROTO_RSVP) {
1352 			struct vif *vifp = viftable + vifi;
1353 			RSVP_DPRINTF(("%s: Sending IPPROTO_RSVP from %x to %x"
1354 			    " on vif %d (%s%s)\n", __func__,
1355 			    ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1356 			    (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1357 			    vifp->v_ifp->if_xname));
1358 		}
1359 		return (ip_mdq(m, ifp, NULL, vifi));
1360 	}
1361 	if (ip->ip_p == IPPROTO_RSVP) {
1362 		RSVP_DPRINTF(("%s: Warning: IPPROTO_RSVP from %x to %x"
1363 		    " without vif option\n", __func__,
1364 		    ntohl(ip->ip_src), ntohl(ip->ip_dst));
1365 	}
1366 #endif /* RSVP_ISI */
1367 
1368 	/*
1369 	 * Don't forward a packet with time-to-live of zero or one,
1370 	 * or a packet destined to a local-only group.
1371 	 */
1372 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1373 		return (0);
1374 
1375 	/*
1376 	 * Determine forwarding vifs from the forwarding cache table
1377 	 */
1378 	s = splsoftnet();
1379 	++mrtstat.mrts_mfc_lookups;
1380 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1381 
1382 	/* Entry exists, so forward if necessary */
1383 	if (rt != NULL) {
1384 		splx(s);
1385 #ifdef RSVP_ISI
1386 		return (ip_mdq(m, ifp, rt, -1));
1387 #else
1388 		return (ip_mdq(m, ifp, rt));
1389 #endif /* RSVP_ISI */
1390 	} else {
1391 		/*
1392 		 * If we don't have a route for packet's origin,
1393 		 * Make a copy of the packet & send message to routing daemon
1394 		 */
1395 
1396 		struct mbuf *mb0;
1397 		struct rtdetq *rte;
1398 		u_int32_t hash;
1399 		int hlen = ip->ip_hl << 2;
1400 #ifdef UPCALL_TIMING
1401 		struct timeval tp;
1402 
1403 		microtime(&tp);
1404 #endif /* UPCALL_TIMING */
1405 
1406 		++mrtstat.mrts_mfc_misses;
1407 
1408 		mrtstat.mrts_no_route++;
1409 		if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1410 			log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1411 			    ntohl(ip->ip_src.s_addr),
1412 			    ntohl(ip->ip_dst.s_addr));
1413 
1414 		/*
1415 		 * Allocate mbufs early so that we don't do extra work if we are
1416 		 * just going to fail anyway.  Make sure to pullup the header so
1417 		 * that other people can't step on it.
1418 		 */
1419 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1420 					      M_NOWAIT);
1421 		if (rte == NULL) {
1422 			splx(s);
1423 			return (ENOBUFS);
1424 		}
1425 		mb0 = m_copypacket(m, M_DONTWAIT);
1426 		M_PULLUP(mb0, hlen);
1427 		if (mb0 == NULL) {
1428 			free(rte, M_MRTABLE);
1429 			splx(s);
1430 			return (ENOBUFS);
1431 		}
1432 
1433 		/* is there an upcall waiting for this flow? */
1434 		hash = MFCHASH(ip->ip_src, ip->ip_dst);
1435 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1436 			if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1437 			    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1438 			    rt->mfc_stall != NULL)
1439 				break;
1440 		}
1441 
1442 		if (rt == NULL) {
1443 			int i;
1444 			struct igmpmsg *im;
1445 
1446 			/*
1447 			 * Locate the vifi for the incoming interface for
1448 			 * this packet.
1449 			 * If none found, drop packet.
1450 			 */
1451 			for (vifi = 0; vifi < numvifs &&
1452 				 viftable[vifi].v_ifp != ifp; vifi++)
1453 				;
1454 			if (vifi >= numvifs) /* vif not found, drop packet */
1455 				goto non_fatal;
1456 
1457 			/* no upcall, so make a new entry */
1458 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1459 						  M_NOWAIT);
1460 			if (rt == NULL)
1461 				goto fail;
1462 
1463 			/*
1464 			 * Make a copy of the header to send to the user level
1465 			 * process
1466 			 */
1467 			mm = m_copym(m, 0, hlen, M_DONTWAIT);
1468 			M_PULLUP(mm, hlen);
1469 			if (mm == NULL)
1470 				goto fail1;
1471 
1472 			/*
1473 			 * Send message to routing daemon to install
1474 			 * a route into the kernel table
1475 			 */
1476 
1477 			im = mtod(mm, struct igmpmsg *);
1478 			im->im_msgtype = IGMPMSG_NOCACHE;
1479 			im->im_mbz = 0;
1480 			im->im_vif = vifi;
1481 
1482 			mrtstat.mrts_upcalls++;
1483 
1484 			sockaddr_in_init(&sin, &ip->ip_src, 0);
1485 			if (socket_send(ip_mrouter, mm, &sin) < 0) {
1486 				log(LOG_WARNING,
1487 				    "ip_mforward: ip_mrouter socket queue full\n");
1488 				++mrtstat.mrts_upq_sockfull;
1489 			fail1:
1490 				free(rt, M_MRTABLE);
1491 			fail:
1492 				free(rte, M_MRTABLE);
1493 				m_freem(mb0);
1494 				splx(s);
1495 				return (ENOBUFS);
1496 			}
1497 
1498 			/* insert new entry at head of hash chain */
1499 			rt->mfc_origin = ip->ip_src;
1500 			rt->mfc_mcastgrp = ip->ip_dst;
1501 			rt->mfc_pkt_cnt = 0;
1502 			rt->mfc_byte_cnt = 0;
1503 			rt->mfc_wrong_if = 0;
1504 			rt->mfc_expire = UPCALL_EXPIRE;
1505 			nexpire[hash]++;
1506 			for (i = 0; i < numvifs; i++) {
1507 				rt->mfc_ttls[i] = 0;
1508 				rt->mfc_flags[i] = 0;
1509 			}
1510 			rt->mfc_parent = -1;
1511 
1512 			/* clear the RP address */
1513 			rt->mfc_rp = zeroin_addr;
1514 
1515 			rt->mfc_bw_meter = NULL;
1516 
1517 			/* link into table */
1518 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1519 			/* Add this entry to the end of the queue */
1520 			rt->mfc_stall = rte;
1521 		} else {
1522 			/* determine if q has overflowed */
1523 			struct rtdetq **p;
1524 			int npkts = 0;
1525 
1526 			/*
1527 			 * XXX ouch! we need to append to the list, but we
1528 			 * only have a pointer to the front, so we have to
1529 			 * scan the entire list every time.
1530 			 */
1531 			for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1532 				if (++npkts > MAX_UPQ) {
1533 					mrtstat.mrts_upq_ovflw++;
1534 				non_fatal:
1535 					free(rte, M_MRTABLE);
1536 					m_freem(mb0);
1537 					splx(s);
1538 					return (0);
1539 				}
1540 
1541 			/* Add this entry to the end of the queue */
1542 			*p = rte;
1543 		}
1544 
1545 		rte->next = NULL;
1546 		rte->m = mb0;
1547 		rte->ifp = ifp;
1548 #ifdef UPCALL_TIMING
1549 		rte->t = tp;
1550 #endif /* UPCALL_TIMING */
1551 
1552 		splx(s);
1553 
1554 		return (0);
1555 	}
1556 }
1557 
1558 
1559 /*ARGSUSED*/
1560 static void
1561 expire_upcalls(void *v)
1562 {
1563 	int i;
1564 
1565 	/* XXX NOMPSAFE still need softnet_lock */
1566 	mutex_enter(softnet_lock);
1567 	KERNEL_LOCK(1, NULL);
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 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1604 	    expire_upcalls, NULL);
1605 
1606 	KERNEL_UNLOCK_ONE(NULL);
1607 	mutex_exit(softnet_lock);
1608 }
1609 
1610 /*
1611  * Packet forwarding routine once entry in the cache is made
1612  */
1613 static int
1614 #ifdef RSVP_ISI
1615 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1616 #else
1617 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1618 #endif /* RSVP_ISI */
1619 {
1620 	struct ip  *ip = mtod(m, struct ip *);
1621 	vifi_t vifi;
1622 	struct vif *vifp;
1623 	struct sockaddr_in sin;
1624 	int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1625 
1626 /*
1627  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1628  * input, they shouldn't get counted on output, so statistics keeping is
1629  * separate.
1630  */
1631 #define MC_SEND(ip, vifp, m) do {					\
1632 	if ((vifp)->v_flags & VIFF_TUNNEL)				\
1633 		encap_send((ip), (vifp), (m));				\
1634 	else								\
1635 		phyint_send((ip), (vifp), (m));				\
1636 } while (/*CONSTCOND*/ 0)
1637 
1638 #ifdef RSVP_ISI
1639 	/*
1640 	 * If xmt_vif is not -1, send on only the requested vif.
1641 	 *
1642 	 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1643 	 */
1644 	if (xmt_vif < numvifs) {
1645 #ifdef PIM
1646 		if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1647 			pim_register_send(ip, viftable + xmt_vif, m, rt);
1648 		else
1649 #endif
1650 		MC_SEND(ip, viftable + xmt_vif, m);
1651 		return (1);
1652 	}
1653 #endif /* RSVP_ISI */
1654 
1655 	/*
1656 	 * Don't forward if it didn't arrive from the parent vif for its origin.
1657 	 */
1658 	vifi = rt->mfc_parent;
1659 	if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1660 		/* came in the wrong interface */
1661 		if (mrtdebug & DEBUG_FORWARD)
1662 			log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1663 			    ifp, vifi,
1664 			    vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1665 		++mrtstat.mrts_wrong_if;
1666 		++rt->mfc_wrong_if;
1667 		/*
1668 		 * If we are doing PIM assert processing, send a message
1669 		 * to the routing daemon.
1670 		 *
1671 		 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1672 		 * can complete the SPT switch, regardless of the type
1673 		 * of the iif (broadcast media, GRE tunnel, etc).
1674 		 */
1675 		if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1676 			struct timeval now;
1677 			u_int32_t delta;
1678 
1679 #ifdef PIM
1680 			if (ifp == &multicast_register_if)
1681 				pimstat.pims_rcv_registers_wrongiif++;
1682 #endif
1683 
1684 			/* Get vifi for the incoming packet */
1685 			for (vifi = 0;
1686 			     vifi < numvifs && viftable[vifi].v_ifp != ifp;
1687 			     vifi++)
1688 			    ;
1689 			if (vifi >= numvifs) {
1690 				/* The iif is not found: ignore the packet. */
1691 				return (0);
1692 			}
1693 
1694 			if (rt->mfc_flags[vifi] &
1695 			    MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1696 				/* WRONGVIF disabled: ignore the packet */
1697 				return (0);
1698 			}
1699 
1700 			microtime(&now);
1701 
1702 			TV_DELTA(rt->mfc_last_assert, now, delta);
1703 
1704 			if (delta > ASSERT_MSG_TIME) {
1705 				struct igmpmsg *im;
1706 				int hlen = ip->ip_hl << 2;
1707 				struct mbuf *mm =
1708 				    m_copym(m, 0, hlen, M_DONTWAIT);
1709 
1710 				M_PULLUP(mm, hlen);
1711 				if (mm == NULL)
1712 					return (ENOBUFS);
1713 
1714 				rt->mfc_last_assert = now;
1715 
1716 				im = mtod(mm, struct igmpmsg *);
1717 				im->im_msgtype	= IGMPMSG_WRONGVIF;
1718 				im->im_mbz	= 0;
1719 				im->im_vif	= vifi;
1720 
1721 				mrtstat.mrts_upcalls++;
1722 
1723 				sockaddr_in_init(&sin, &im->im_src, 0);
1724 				if (socket_send(ip_mrouter, mm, &sin) < 0) {
1725 					log(LOG_WARNING,
1726 					    "ip_mforward: ip_mrouter socket queue full\n");
1727 					++mrtstat.mrts_upq_sockfull;
1728 					return (ENOBUFS);
1729 				}
1730 			}
1731 		}
1732 		return (0);
1733 	}
1734 
1735 	/* If I sourced this packet, it counts as output, else it was input. */
1736 	if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1737 		viftable[vifi].v_pkt_out++;
1738 		viftable[vifi].v_bytes_out += plen;
1739 	} else {
1740 		viftable[vifi].v_pkt_in++;
1741 		viftable[vifi].v_bytes_in += plen;
1742 	}
1743 	rt->mfc_pkt_cnt++;
1744 	rt->mfc_byte_cnt += plen;
1745 
1746 	/*
1747 	 * For each vif, decide if a copy of the packet should be forwarded.
1748 	 * Forward if:
1749 	 *		- the ttl exceeds the vif's threshold
1750 	 *		- there are group members downstream on interface
1751 	 */
1752 	for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1753 		if ((rt->mfc_ttls[vifi] > 0) &&
1754 			(ip->ip_ttl > rt->mfc_ttls[vifi])) {
1755 			vifp->v_pkt_out++;
1756 			vifp->v_bytes_out += plen;
1757 #ifdef PIM
1758 			if (vifp->v_flags & VIFF_REGISTER)
1759 				pim_register_send(ip, vifp, m, rt);
1760 			else
1761 #endif
1762 			MC_SEND(ip, vifp, m);
1763 		}
1764 
1765 	/*
1766 	 * Perform upcall-related bw measuring.
1767 	 */
1768 	if (rt->mfc_bw_meter != NULL) {
1769 		struct bw_meter *x;
1770 		struct timeval now;
1771 
1772 		microtime(&now);
1773 		for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1774 			bw_meter_receive_packet(x, plen, &now);
1775 	}
1776 
1777 	return (0);
1778 }
1779 
1780 #ifdef RSVP_ISI
1781 /*
1782  * check if a vif number is legal/ok. This is used by ip_output.
1783  */
1784 int
1785 legal_vif_num(int vif)
1786 {
1787 	if (vif >= 0 && vif < numvifs)
1788 		return (1);
1789 	else
1790 		return (0);
1791 }
1792 #endif /* RSVP_ISI */
1793 
1794 static void
1795 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1796 {
1797 	struct mbuf *mb_copy;
1798 	int hlen = ip->ip_hl << 2;
1799 
1800 	/*
1801 	 * Make a new reference to the packet; make sure that
1802 	 * the IP header is actually copied, not just referenced,
1803 	 * so that ip_output() only scribbles on the copy.
1804 	 */
1805 	mb_copy = m_copypacket(m, M_DONTWAIT);
1806 	M_PULLUP(mb_copy, hlen);
1807 	if (mb_copy == NULL)
1808 		return;
1809 
1810 	if (vifp->v_rate_limit <= 0)
1811 		tbf_send_packet(vifp, mb_copy);
1812 	else
1813 		tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1814 		    ntohs(ip->ip_len));
1815 }
1816 
1817 static void
1818 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1819 {
1820 	struct mbuf *mb_copy;
1821 	struct ip *ip_copy;
1822 	int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1823 
1824 	/* Take care of delayed checksums */
1825 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1826 		in_delayed_cksum(m);
1827 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1828 	}
1829 
1830 	/*
1831 	 * copy the old packet & pullup its IP header into the
1832 	 * new mbuf so we can modify it.  Try to fill the new
1833 	 * mbuf since if we don't the ethernet driver will.
1834 	 */
1835 	MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1836 	if (mb_copy == NULL)
1837 		return;
1838 	mb_copy->m_data += max_linkhdr;
1839 	mb_copy->m_pkthdr.len = len;
1840 	mb_copy->m_len = sizeof(multicast_encap_iphdr);
1841 
1842 	if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
1843 		m_freem(mb_copy);
1844 		return;
1845 	}
1846 	i = MHLEN - max_linkhdr;
1847 	if (i > len)
1848 		i = len;
1849 	mb_copy = m_pullup(mb_copy, i);
1850 	if (mb_copy == NULL)
1851 		return;
1852 
1853 	/*
1854 	 * fill in the encapsulating IP header.
1855 	 */
1856 	ip_copy = mtod(mb_copy, struct ip *);
1857 	*ip_copy = multicast_encap_iphdr;
1858 	if (len < IP_MINFRAGSIZE)
1859 		ip_copy->ip_id = 0;
1860 	else
1861 		ip_copy->ip_id = ip_newid(NULL);
1862 	ip_copy->ip_len = htons(len);
1863 	ip_copy->ip_src = vifp->v_lcl_addr;
1864 	ip_copy->ip_dst = vifp->v_rmt_addr;
1865 
1866 	/*
1867 	 * turn the encapsulated IP header back into a valid one.
1868 	 */
1869 	ip = (struct ip *)((char *)ip_copy + sizeof(multicast_encap_iphdr));
1870 	--ip->ip_ttl;
1871 	ip->ip_sum = 0;
1872 	mb_copy->m_data += sizeof(multicast_encap_iphdr);
1873 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1874 	mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1875 
1876 	if (vifp->v_rate_limit <= 0)
1877 		tbf_send_packet(vifp, mb_copy);
1878 	else
1879 		tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1880 }
1881 
1882 /*
1883  * De-encapsulate a packet and feed it back through ip input.
1884  */
1885 static void
1886 vif_input(struct mbuf *m, int off, int proto)
1887 {
1888 	struct vif *vifp;
1889 
1890 	vifp = (struct vif *)encap_getarg(m);
1891 	if (!vifp || proto != ENCAP_PROTO) {
1892 		m_freem(m);
1893 		mrtstat.mrts_bad_tunnel++;
1894 		return;
1895 	}
1896 
1897 	m_adj(m, off);
1898 	m_set_rcvif(m, vifp->v_ifp);
1899 
1900 	if (__predict_false(!pktq_enqueue(ip_pktq, m, 0))) {
1901 		m_freem(m);
1902 	}
1903 }
1904 
1905 /*
1906  * Check if the packet should be received on the vif denoted by arg.
1907  * (The encap selection code will call this once per vif since each is
1908  * registered separately.)
1909  */
1910 static int
1911 vif_encapcheck(struct mbuf *m, int off, int proto, void *arg)
1912 {
1913 	struct vif *vifp;
1914 	struct ip ip;
1915 
1916 #ifdef DIAGNOSTIC
1917 	if (!arg || proto != IPPROTO_IPV4)
1918 		panic("unexpected arg in vif_encapcheck");
1919 #endif
1920 
1921 	/*
1922 	 * Accept the packet only if the inner heaader is multicast
1923 	 * and the outer header matches a tunnel-mode vif.  Order
1924 	 * checks in the hope that common non-matching packets will be
1925 	 * rejected quickly.  Assume that unicast IPv4 traffic in a
1926 	 * parallel tunnel (e.g. gif(4)) is unlikely.
1927 	 */
1928 
1929 	/* Obtain the outer IP header and the vif pointer. */
1930 	m_copydata((struct mbuf *)m, 0, sizeof(ip), (void *)&ip);
1931 	vifp = (struct vif *)arg;
1932 
1933 	/*
1934 	 * The outer source must match the vif's remote peer address.
1935 	 * For a multicast router with several tunnels, this is the
1936 	 * only check that will fail on packets in other tunnels,
1937 	 * assuming the local address is the same.
1938 	 */
1939 	if (!in_hosteq(vifp->v_rmt_addr, ip.ip_src))
1940 		return 0;
1941 
1942 	/* The outer destination must match the vif's local address. */
1943 	if (!in_hosteq(vifp->v_lcl_addr, ip.ip_dst))
1944 		return 0;
1945 
1946 	/* The vif must be of tunnel type. */
1947 	if ((vifp->v_flags & VIFF_TUNNEL) == 0)
1948 		return 0;
1949 
1950 	/* Check that the inner destination is multicast. */
1951 	m_copydata((struct mbuf *)m, off, sizeof(ip), (void *)&ip);
1952 	if (!IN_MULTICAST(ip.ip_dst.s_addr))
1953 		return 0;
1954 
1955 	/*
1956 	 * We have checked that both the outer src and dst addresses
1957 	 * match the vif, and that the inner destination is multicast
1958 	 * (224/5).  By claiming more than 64, we intend to
1959 	 * preferentially take packets that also match a parallel
1960 	 * gif(4).
1961 	 */
1962 	return 32 + 32 + 5;
1963 }
1964 
1965 /*
1966  * Token bucket filter module
1967  */
1968 static void
1969 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
1970 {
1971 
1972 	if (len > MAX_BKT_SIZE) {
1973 		/* drop if packet is too large */
1974 		mrtstat.mrts_pkt2large++;
1975 		m_freem(m);
1976 		return;
1977 	}
1978 
1979 	tbf_update_tokens(vifp);
1980 
1981 	/*
1982 	 * If there are enough tokens, and the queue is empty, send this packet
1983 	 * out immediately.  Otherwise, try to insert it on this vif's queue.
1984 	 */
1985 	if (vifp->tbf_q_len == 0) {
1986 		if (len <= vifp->tbf_n_tok) {
1987 			vifp->tbf_n_tok -= len;
1988 			tbf_send_packet(vifp, m);
1989 		} else {
1990 			/* queue packet and timeout till later */
1991 			tbf_queue(vifp, m);
1992 			callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
1993 			    tbf_reprocess_q, vifp);
1994 		}
1995 	} else {
1996 		if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
1997 		    !tbf_dq_sel(vifp, ip)) {
1998 			/* queue full, and couldn't make room */
1999 			mrtstat.mrts_q_overflow++;
2000 			m_freem(m);
2001 		} else {
2002 			/* queue length low enough, or made room */
2003 			tbf_queue(vifp, m);
2004 			tbf_process_q(vifp);
2005 		}
2006 	}
2007 }
2008 
2009 /*
2010  * adds a packet to the queue at the interface
2011  */
2012 static void
2013 tbf_queue(struct vif *vifp, struct mbuf *m)
2014 {
2015 	int s = splsoftnet();
2016 
2017 	/* insert at tail */
2018 	*vifp->tbf_t = m;
2019 	vifp->tbf_t = &m->m_nextpkt;
2020 	vifp->tbf_q_len++;
2021 
2022 	splx(s);
2023 }
2024 
2025 
2026 /*
2027  * processes the queue at the interface
2028  */
2029 static void
2030 tbf_process_q(struct vif *vifp)
2031 {
2032 	struct mbuf *m;
2033 	int len;
2034 	int s = splsoftnet();
2035 
2036 	/*
2037 	 * Loop through the queue at the interface and send as many packets
2038 	 * as possible.
2039 	 */
2040 	for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
2041 		len = ntohs(mtod(m, struct ip *)->ip_len);
2042 
2043 		/* determine if the packet can be sent */
2044 		if (len <= vifp->tbf_n_tok) {
2045 			/* if so,
2046 			 * reduce no of tokens, dequeue the packet,
2047 			 * send the packet.
2048 			 */
2049 			if ((vifp->tbf_q = m->m_nextpkt) == NULL)
2050 				vifp->tbf_t = &vifp->tbf_q;
2051 			--vifp->tbf_q_len;
2052 
2053 			m->m_nextpkt = NULL;
2054 			vifp->tbf_n_tok -= len;
2055 			tbf_send_packet(vifp, m);
2056 		} else
2057 			break;
2058 	}
2059 	splx(s);
2060 }
2061 
2062 static void
2063 tbf_reprocess_q(void *arg)
2064 {
2065 	struct vif *vifp = arg;
2066 
2067 	if (ip_mrouter == NULL)
2068 		return;
2069 
2070 	tbf_update_tokens(vifp);
2071 	tbf_process_q(vifp);
2072 
2073 	if (vifp->tbf_q_len != 0)
2074 		callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2075 		    tbf_reprocess_q, vifp);
2076 }
2077 
2078 /* function that will selectively discard a member of the queue
2079  * based on the precedence value and the priority
2080  */
2081 static int
2082 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2083 {
2084 	u_int p;
2085 	struct mbuf **mp, *m;
2086 	int s = splsoftnet();
2087 
2088 	p = priority(vifp, ip);
2089 
2090 	for (mp = &vifp->tbf_q, m = *mp;
2091 	    m != NULL;
2092 	    mp = &m->m_nextpkt, m = *mp) {
2093 		if (p > priority(vifp, mtod(m, struct ip *))) {
2094 			if ((*mp = m->m_nextpkt) == NULL)
2095 				vifp->tbf_t = mp;
2096 			--vifp->tbf_q_len;
2097 
2098 			m_freem(m);
2099 			mrtstat.mrts_drop_sel++;
2100 			splx(s);
2101 			return (1);
2102 		}
2103 	}
2104 	splx(s);
2105 	return (0);
2106 }
2107 
2108 static void
2109 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2110 {
2111 	int error;
2112 	int s = splsoftnet();
2113 
2114 	if (vifp->v_flags & VIFF_TUNNEL) {
2115 		/* If tunnel options */
2116 		ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL);
2117 	} else {
2118 		/* if physical interface option, extract the options and then send */
2119 		struct ip_moptions imo;
2120 
2121 		imo.imo_multicast_if_index = if_get_index(vifp->v_ifp);
2122 		imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2123 		imo.imo_multicast_loop = 1;
2124 #ifdef RSVP_ISI
2125 		imo.imo_multicast_vif = -1;
2126 #endif
2127 
2128 		error = ip_output(m, NULL, NULL, IP_FORWARDING|IP_MULTICASTOPTS,
2129 		    &imo, NULL);
2130 
2131 		if (mrtdebug & DEBUG_XMIT)
2132 			log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2133 			    (long)(vifp - viftable), error);
2134 	}
2135 	splx(s);
2136 }
2137 
2138 /* determine the current time and then
2139  * the elapsed time (between the last time and time now)
2140  * in milliseconds & update the no. of tokens in the bucket
2141  */
2142 static void
2143 tbf_update_tokens(struct vif *vifp)
2144 {
2145 	struct timeval tp;
2146 	u_int32_t tm;
2147 	int s = splsoftnet();
2148 
2149 	microtime(&tp);
2150 
2151 	TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2152 
2153 	/*
2154 	 * This formula is actually
2155 	 * "time in seconds" * "bytes/second".
2156 	 *
2157 	 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2158 	 *
2159 	 * The (1000/1024) was introduced in add_vif to optimize
2160 	 * this divide into a shift.
2161 	 */
2162 	vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2163 	vifp->tbf_last_pkt_t = tp;
2164 
2165 	if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2166 		vifp->tbf_n_tok = MAX_BKT_SIZE;
2167 
2168 	splx(s);
2169 }
2170 
2171 static int
2172 priority(struct vif *vifp, struct ip *ip)
2173 {
2174 	int prio = 50;	/* the lowest priority -- default case */
2175 
2176 	/* temporary hack; may add general packet classifier some day */
2177 
2178 	/*
2179 	 * The UDP port space is divided up into four priority ranges:
2180 	 * [0, 16384)     : unclassified - lowest priority
2181 	 * [16384, 32768) : audio - highest priority
2182 	 * [32768, 49152) : whiteboard - medium priority
2183 	 * [49152, 65536) : video - low priority
2184 	 */
2185 	if (ip->ip_p == IPPROTO_UDP) {
2186 		struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2187 
2188 		switch (ntohs(udp->uh_dport) & 0xc000) {
2189 		case 0x4000:
2190 			prio = 70;
2191 			break;
2192 		case 0x8000:
2193 			prio = 60;
2194 			break;
2195 		case 0xc000:
2196 			prio = 55;
2197 			break;
2198 		}
2199 
2200 		if (tbfdebug > 1)
2201 			log(LOG_DEBUG, "port %x prio %d\n",
2202 			    ntohs(udp->uh_dport), prio);
2203 	}
2204 
2205 	return (prio);
2206 }
2207 
2208 /*
2209  * End of token bucket filter modifications
2210  */
2211 #ifdef RSVP_ISI
2212 int
2213 ip_rsvp_vif_init(struct socket *so, struct mbuf *m)
2214 {
2215 	int vifi, s;
2216 
2217 	RSVP_DPRINTF(("%s: so_type = %d, pr_protocol = %d\n", __func__
2218 	    so->so_type, so->so_proto->pr_protocol));
2219 
2220 	if (so->so_type != SOCK_RAW ||
2221 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2222 		return (EOPNOTSUPP);
2223 
2224 	/* Check mbuf. */
2225 	if (m == NULL || m->m_len != sizeof(int)) {
2226 		return (EINVAL);
2227 	}
2228 	vifi = *(mtod(m, int *));
2229 
2230 	RSVP_DPRINTF(("%s: vif = %d rsvp_on = %d\n", __func__, vifi, rsvp_on));
2231 
2232 	s = splsoftnet();
2233 
2234 	/* Check vif. */
2235 	if (!legal_vif_num(vifi)) {
2236 		splx(s);
2237 		return (EADDRNOTAVAIL);
2238 	}
2239 
2240 	/* Check if socket is available. */
2241 	if (viftable[vifi].v_rsvpd != NULL) {
2242 		splx(s);
2243 		return (EADDRINUSE);
2244 	}
2245 
2246 	viftable[vifi].v_rsvpd = so;
2247 	/*
2248 	 * This may seem silly, but we need to be sure we don't over-increment
2249 	 * the RSVP counter, in case something slips up.
2250 	 */
2251 	if (!viftable[vifi].v_rsvp_on) {
2252 		viftable[vifi].v_rsvp_on = 1;
2253 		rsvp_on++;
2254 	}
2255 
2256 	splx(s);
2257 	return (0);
2258 }
2259 
2260 int
2261 ip_rsvp_vif_done(struct socket *so, struct mbuf *m)
2262 {
2263 	int vifi, s;
2264 
2265 	RSVP_DPRINTF(("%s: so_type = %d, pr_protocol = %d\n", __func__,
2266 	    so->so_type, so->so_proto->pr_protocol));
2267 
2268 	if (so->so_type != SOCK_RAW ||
2269 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2270 		return (EOPNOTSUPP);
2271 
2272 	/* Check mbuf. */
2273 	if (m == NULL || m->m_len != sizeof(int)) {
2274 		return (EINVAL);
2275 	}
2276 	vifi = *(mtod(m, int *));
2277 
2278 	s = splsoftnet();
2279 
2280 	/* Check vif. */
2281 	if (!legal_vif_num(vifi)) {
2282 		splx(s);
2283 		return (EADDRNOTAVAIL);
2284 	}
2285 
2286 	RSVP_DPRINTF(("%s: v_rsvpd = %x so = %x\n", __func__,
2287 	    viftable[vifi].v_rsvpd, so));
2288 
2289 	viftable[vifi].v_rsvpd = NULL;
2290 	/*
2291 	 * This may seem silly, but we need to be sure we don't over-decrement
2292 	 * the RSVP counter, in case something slips up.
2293 	 */
2294 	if (viftable[vifi].v_rsvp_on) {
2295 		viftable[vifi].v_rsvp_on = 0;
2296 		rsvp_on--;
2297 	}
2298 
2299 	splx(s);
2300 	return (0);
2301 }
2302 
2303 void
2304 ip_rsvp_force_done(struct socket *so)
2305 {
2306 	int vifi, s;
2307 
2308 	/* Don't bother if it is not the right type of socket. */
2309 	if (so->so_type != SOCK_RAW ||
2310 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2311 		return;
2312 
2313 	s = splsoftnet();
2314 
2315 	/*
2316 	 * The socket may be attached to more than one vif...this
2317 	 * is perfectly legal.
2318 	 */
2319 	for (vifi = 0; vifi < numvifs; vifi++) {
2320 		if (viftable[vifi].v_rsvpd == so) {
2321 			viftable[vifi].v_rsvpd = NULL;
2322 			/*
2323 			 * This may seem silly, but we need to be sure we don't
2324 			 * over-decrement the RSVP counter, in case something
2325 			 * slips up.
2326 			 */
2327 			if (viftable[vifi].v_rsvp_on) {
2328 				viftable[vifi].v_rsvp_on = 0;
2329 				rsvp_on--;
2330 			}
2331 		}
2332 	}
2333 
2334 	splx(s);
2335 	return;
2336 }
2337 
2338 void
2339 rsvp_input(struct mbuf *m, struct ifnet *ifp)
2340 {
2341 	int vifi, s;
2342 	struct ip *ip = mtod(m, struct ip *);
2343 	struct sockaddr_in rsvp_src;
2344 
2345 	RSVP_DPRINTF(("%s: rsvp_on %d\n", __func__, rsvp_on));
2346 
2347 	/*
2348 	 * Can still get packets with rsvp_on = 0 if there is a local member
2349 	 * of the group to which the RSVP packet is addressed.  But in this
2350 	 * case we want to throw the packet away.
2351 	 */
2352 	if (!rsvp_on) {
2353 		m_freem(m);
2354 		return;
2355 	}
2356 
2357 	/*
2358 	 * If the old-style non-vif-associated socket is set, then use
2359 	 * it and ignore the new ones.
2360 	 */
2361 	if (ip_rsvpd != NULL) {
2362 		RSVP_DPRINTF(("%s: Sending packet up old-style socket\n",
2363 		    __func__));
2364 		rip_input(m);	/*XXX*/
2365 		return;
2366 	}
2367 
2368 	s = splsoftnet();
2369 
2370 	RSVP_DPRINTF(("%s: check vifs\n", __func__));
2371 
2372 	/* Find which vif the packet arrived on. */
2373 	for (vifi = 0; vifi < numvifs; vifi++) {
2374 		if (viftable[vifi].v_ifp == ifp)
2375 			break;
2376 	}
2377 
2378 	if (vifi == numvifs) {
2379 		/* Can't find vif packet arrived on. Drop packet. */
2380 		RSVP_DPRINTF("%s: Can't find vif for packet...dropping it.\n",
2381 		    __func__));
2382 		m_freem(m);
2383 		splx(s);
2384 		return;
2385 	}
2386 
2387 	RSVP_DPRINTF(("%s: check socket\n", __func__));
2388 
2389 	if (viftable[vifi].v_rsvpd == NULL) {
2390 		/*
2391 		 * drop packet, since there is no specific socket for this
2392 		 * interface
2393 		 */
2394 		RSVP_DPRINTF(("%s: No socket defined for vif %d\n", __func__,
2395 		    vifi));
2396 		m_freem(m);
2397 		splx(s);
2398 		return;
2399 	}
2400 
2401 	sockaddr_in_init(&rsvp_src, &ip->ip_src, 0);
2402 
2403 	if (m)
2404 		RSVP_DPRINTF(("%s: m->m_len = %d, sbspace() = %d\n", __func__,
2405 		    m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv)));
2406 
2407 	if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2408 		RSVP_DPRINTF(("%s: Failed to append to socket\n", __func__));
2409 	else
2410 		RSVP_DPRINTF(("%s: send packet up\n", __func__));
2411 
2412 	splx(s);
2413 }
2414 #endif /* RSVP_ISI */
2415 
2416 /*
2417  * Code for bandwidth monitors
2418  */
2419 
2420 /*
2421  * Define common interface for timeval-related methods
2422  */
2423 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2424 #define	BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2425 #define	BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2426 
2427 static uint32_t
2428 compute_bw_meter_flags(struct bw_upcall *req)
2429 {
2430     uint32_t flags = 0;
2431 
2432     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2433 	flags |= BW_METER_UNIT_PACKETS;
2434     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2435 	flags |= BW_METER_UNIT_BYTES;
2436     if (req->bu_flags & BW_UPCALL_GEQ)
2437 	flags |= BW_METER_GEQ;
2438     if (req->bu_flags & BW_UPCALL_LEQ)
2439 	flags |= BW_METER_LEQ;
2440 
2441     return flags;
2442 }
2443 
2444 /*
2445  * Add a bw_meter entry
2446  */
2447 static int
2448 add_bw_upcall(struct bw_upcall *req)
2449 {
2450     int s;
2451     struct mfc *mfc;
2452     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2453 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2454     struct timeval now;
2455     struct bw_meter *x;
2456     uint32_t flags;
2457 
2458     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2459 	return EOPNOTSUPP;
2460 
2461     /* Test if the flags are valid */
2462     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2463 	return EINVAL;
2464     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2465 	return EINVAL;
2466     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2467 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2468 	return EINVAL;
2469 
2470     /* Test if the threshold time interval is valid */
2471     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2472 	return EINVAL;
2473 
2474     flags = compute_bw_meter_flags(req);
2475 
2476     /*
2477      * Find if we have already same bw_meter entry
2478      */
2479     s = splsoftnet();
2480     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2481     if (mfc == NULL) {
2482 	splx(s);
2483 	return EADDRNOTAVAIL;
2484     }
2485     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2486 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2487 			   &req->bu_threshold.b_time, ==)) &&
2488 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2489 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2490 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
2491 	    splx(s);
2492 	    return 0;		/* XXX Already installed */
2493 	}
2494     }
2495 
2496     /* Allocate the new bw_meter entry */
2497     x = kmem_intr_alloc(sizeof(*x), KM_NOSLEEP);
2498     if (x == NULL) {
2499 	splx(s);
2500 	return ENOBUFS;
2501     }
2502 
2503     /* Set the new bw_meter entry */
2504     x->bm_threshold.b_time = req->bu_threshold.b_time;
2505     microtime(&now);
2506     x->bm_start_time = now;
2507     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2508     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2509     x->bm_measured.b_packets = 0;
2510     x->bm_measured.b_bytes = 0;
2511     x->bm_flags = flags;
2512     x->bm_time_next = NULL;
2513     x->bm_time_hash = BW_METER_BUCKETS;
2514 
2515     /* Add the new bw_meter entry to the front of entries for this MFC */
2516     x->bm_mfc = mfc;
2517     x->bm_mfc_next = mfc->mfc_bw_meter;
2518     mfc->mfc_bw_meter = x;
2519     schedule_bw_meter(x, &now);
2520     splx(s);
2521 
2522     return 0;
2523 }
2524 
2525 static void
2526 free_bw_list(struct bw_meter *list)
2527 {
2528     while (list != NULL) {
2529 	struct bw_meter *x = list;
2530 
2531 	list = list->bm_mfc_next;
2532 	unschedule_bw_meter(x);
2533 	kmem_free(x, sizeof(*x));
2534     }
2535 }
2536 
2537 /*
2538  * Delete one or multiple bw_meter entries
2539  */
2540 static int
2541 del_bw_upcall(struct bw_upcall *req)
2542 {
2543     int s;
2544     struct mfc *mfc;
2545     struct bw_meter *x;
2546 
2547     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2548 	return EOPNOTSUPP;
2549 
2550     s = splsoftnet();
2551     /* Find the corresponding MFC entry */
2552     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2553     if (mfc == NULL) {
2554 	splx(s);
2555 	return EADDRNOTAVAIL;
2556     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2557 	/*
2558 	 * Delete all bw_meter entries for this mfc
2559 	 */
2560 	struct bw_meter *list;
2561 
2562 	list = mfc->mfc_bw_meter;
2563 	mfc->mfc_bw_meter = NULL;
2564 	free_bw_list(list);
2565 	splx(s);
2566 	return 0;
2567     } else {			/* Delete a single bw_meter entry */
2568 	struct bw_meter *prev;
2569 	uint32_t flags = 0;
2570 
2571 	flags = compute_bw_meter_flags(req);
2572 
2573 	/* Find the bw_meter entry to delete */
2574 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2575 	     prev = x, x = x->bm_mfc_next) {
2576 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2577 			       &req->bu_threshold.b_time, ==)) &&
2578 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2579 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2580 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2581 		break;
2582 	}
2583 	if (x != NULL) { /* Delete entry from the list for this MFC */
2584 	    if (prev != NULL)
2585 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2586 	    else
2587 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2588 
2589 	    unschedule_bw_meter(x);
2590 	    splx(s);
2591 	    /* Free the bw_meter entry */
2592 	    kmem_free(x, sizeof(*x));
2593 	    return 0;
2594 	} else {
2595 	    splx(s);
2596 	    return EINVAL;
2597 	}
2598     }
2599     /* NOTREACHED */
2600 }
2601 
2602 /*
2603  * Perform bandwidth measurement processing that may result in an upcall
2604  */
2605 static void
2606 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2607 {
2608     struct timeval delta;
2609 
2610     delta = *nowp;
2611     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2612 
2613     if (x->bm_flags & BW_METER_GEQ) {
2614 	/*
2615 	 * Processing for ">=" type of bw_meter entry
2616 	 */
2617 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2618 	    /* Reset the bw_meter entry */
2619 	    x->bm_start_time = *nowp;
2620 	    x->bm_measured.b_packets = 0;
2621 	    x->bm_measured.b_bytes = 0;
2622 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2623 	}
2624 
2625 	/* Record that a packet is received */
2626 	x->bm_measured.b_packets++;
2627 	x->bm_measured.b_bytes += plen;
2628 
2629 	/*
2630 	 * Test if we should deliver an upcall
2631 	 */
2632 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2633 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2634 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2635 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2636 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2637 		/* Prepare an upcall for delivery */
2638 		bw_meter_prepare_upcall(x, nowp);
2639 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2640 	    }
2641 	}
2642     } else if (x->bm_flags & BW_METER_LEQ) {
2643 	/*
2644 	 * Processing for "<=" type of bw_meter entry
2645 	 */
2646 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2647 	    /*
2648 	     * We are behind time with the multicast forwarding table
2649 	     * scanning for "<=" type of bw_meter entries, so test now
2650 	     * if we should deliver an upcall.
2651 	     */
2652 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2653 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2654 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2655 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2656 		/* Prepare an upcall for delivery */
2657 		bw_meter_prepare_upcall(x, nowp);
2658 	    }
2659 	    /* Reschedule the bw_meter entry */
2660 	    unschedule_bw_meter(x);
2661 	    schedule_bw_meter(x, nowp);
2662 	}
2663 
2664 	/* Record that a packet is received */
2665 	x->bm_measured.b_packets++;
2666 	x->bm_measured.b_bytes += plen;
2667 
2668 	/*
2669 	 * Test if we should restart the measuring interval
2670 	 */
2671 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2672 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2673 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
2674 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2675 	    /* Don't restart the measuring interval */
2676 	} else {
2677 	    /* Do restart the measuring interval */
2678 	    /*
2679 	     * XXX: note that we don't unschedule and schedule, because this
2680 	     * might be too much overhead per packet. Instead, when we process
2681 	     * all entries for a given timer hash bin, we check whether it is
2682 	     * really a timeout. If not, we reschedule at that time.
2683 	     */
2684 	    x->bm_start_time = *nowp;
2685 	    x->bm_measured.b_packets = 0;
2686 	    x->bm_measured.b_bytes = 0;
2687 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2688 	}
2689     }
2690 }
2691 
2692 /*
2693  * Prepare a bandwidth-related upcall
2694  */
2695 static void
2696 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2697 {
2698     struct timeval delta;
2699     struct bw_upcall *u;
2700 
2701     /*
2702      * Compute the measured time interval
2703      */
2704     delta = *nowp;
2705     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2706 
2707     /*
2708      * If there are too many pending upcalls, deliver them now
2709      */
2710     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2711 	bw_upcalls_send();
2712 
2713     /*
2714      * Set the bw_upcall entry
2715      */
2716     u = &bw_upcalls[bw_upcalls_n++];
2717     u->bu_src = x->bm_mfc->mfc_origin;
2718     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2719     u->bu_threshold.b_time = x->bm_threshold.b_time;
2720     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2721     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2722     u->bu_measured.b_time = delta;
2723     u->bu_measured.b_packets = x->bm_measured.b_packets;
2724     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2725     u->bu_flags = 0;
2726     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2727 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2728     if (x->bm_flags & BW_METER_UNIT_BYTES)
2729 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2730     if (x->bm_flags & BW_METER_GEQ)
2731 	u->bu_flags |= BW_UPCALL_GEQ;
2732     if (x->bm_flags & BW_METER_LEQ)
2733 	u->bu_flags |= BW_UPCALL_LEQ;
2734 }
2735 
2736 /*
2737  * Send the pending bandwidth-related upcalls
2738  */
2739 static void
2740 bw_upcalls_send(void)
2741 {
2742     struct mbuf *m;
2743     int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2744     struct sockaddr_in k_igmpsrc = {
2745 	    .sin_len = sizeof(k_igmpsrc),
2746 	    .sin_family = AF_INET,
2747     };
2748     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2749 				      0,		/* unused2 */
2750 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2751 				      0,		/* im_mbz  */
2752 				      0,		/* im_vif  */
2753 				      0,		/* unused3 */
2754 				      { 0 },		/* im_src  */
2755 				      { 0 } };		/* im_dst  */
2756 
2757     if (bw_upcalls_n == 0)
2758 	return;			/* No pending upcalls */
2759 
2760     bw_upcalls_n = 0;
2761 
2762     /*
2763      * Allocate a new mbuf, initialize it with the header and
2764      * the payload for the pending calls.
2765      */
2766     MGETHDR(m, M_DONTWAIT, MT_HEADER);
2767     if (m == NULL) {
2768 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2769 	return;
2770     }
2771 
2772     m->m_len = m->m_pkthdr.len = 0;
2773     m_copyback(m, 0, sizeof(struct igmpmsg), (void *)&igmpmsg);
2774     m_copyback(m, sizeof(struct igmpmsg), len, (void *)&bw_upcalls[0]);
2775 
2776     /*
2777      * Send the upcalls
2778      * XXX do we need to set the address in k_igmpsrc ?
2779      */
2780     mrtstat.mrts_upcalls++;
2781     if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2782 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2783 	++mrtstat.mrts_upq_sockfull;
2784     }
2785 }
2786 
2787 /*
2788  * Compute the timeout hash value for the bw_meter entries
2789  */
2790 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2791     do {								\
2792 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2793 									\
2794 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2795 	(hash) = next_timeval.tv_sec;					\
2796 	if (next_timeval.tv_usec)					\
2797 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2798 	(hash) %= BW_METER_BUCKETS;					\
2799     } while (/*CONSTCOND*/ 0)
2800 
2801 /*
2802  * Schedule a timer to process periodically bw_meter entry of type "<="
2803  * by linking the entry in the proper hash bucket.
2804  */
2805 static void
2806 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2807 {
2808     int time_hash;
2809 
2810     if (!(x->bm_flags & BW_METER_LEQ))
2811 	return;		/* XXX: we schedule timers only for "<=" entries */
2812 
2813     /*
2814      * Reset the bw_meter entry
2815      */
2816     x->bm_start_time = *nowp;
2817     x->bm_measured.b_packets = 0;
2818     x->bm_measured.b_bytes = 0;
2819     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2820 
2821     /*
2822      * Compute the timeout hash value and insert the entry
2823      */
2824     BW_METER_TIMEHASH(x, time_hash);
2825     x->bm_time_next = bw_meter_timers[time_hash];
2826     bw_meter_timers[time_hash] = x;
2827     x->bm_time_hash = time_hash;
2828 }
2829 
2830 /*
2831  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2832  * by removing the entry from the proper hash bucket.
2833  */
2834 static void
2835 unschedule_bw_meter(struct bw_meter *x)
2836 {
2837     int time_hash;
2838     struct bw_meter *prev, *tmp;
2839 
2840     if (!(x->bm_flags & BW_METER_LEQ))
2841 	return;		/* XXX: we schedule timers only for "<=" entries */
2842 
2843     /*
2844      * Compute the timeout hash value and delete the entry
2845      */
2846     time_hash = x->bm_time_hash;
2847     if (time_hash >= BW_METER_BUCKETS)
2848 	return;		/* Entry was not scheduled */
2849 
2850     for (prev = NULL, tmp = bw_meter_timers[time_hash];
2851 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2852 	if (tmp == x)
2853 	    break;
2854 
2855     if (tmp == NULL)
2856 	panic("unschedule_bw_meter: bw_meter entry not found");
2857 
2858     if (prev != NULL)
2859 	prev->bm_time_next = x->bm_time_next;
2860     else
2861 	bw_meter_timers[time_hash] = x->bm_time_next;
2862 
2863     x->bm_time_next = NULL;
2864     x->bm_time_hash = BW_METER_BUCKETS;
2865 }
2866 
2867 /*
2868  * Process all "<=" type of bw_meter that should be processed now,
2869  * and for each entry prepare an upcall if necessary. Each processed
2870  * entry is rescheduled again for the (periodic) processing.
2871  *
2872  * This is run periodically (once per second normally). On each round,
2873  * all the potentially matching entries are in the hash slot that we are
2874  * looking at.
2875  */
2876 static void
2877 bw_meter_process(void)
2878 {
2879     int s;
2880     static uint32_t last_tv_sec;	/* last time we processed this */
2881 
2882     uint32_t loops;
2883     int i;
2884     struct timeval now, process_endtime;
2885 
2886     microtime(&now);
2887     if (last_tv_sec == now.tv_sec)
2888 	return;		/* nothing to do */
2889 
2890     loops = now.tv_sec - last_tv_sec;
2891     last_tv_sec = now.tv_sec;
2892     if (loops > BW_METER_BUCKETS)
2893 	loops = BW_METER_BUCKETS;
2894 
2895     s = splsoftnet();
2896     /*
2897      * Process all bins of bw_meter entries from the one after the last
2898      * processed to the current one. On entry, i points to the last bucket
2899      * visited, so we need to increment i at the beginning of the loop.
2900      */
2901     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2902 	struct bw_meter *x, *tmp_list;
2903 
2904 	if (++i >= BW_METER_BUCKETS)
2905 	    i = 0;
2906 
2907 	/* Disconnect the list of bw_meter entries from the bin */
2908 	tmp_list = bw_meter_timers[i];
2909 	bw_meter_timers[i] = NULL;
2910 
2911 	/* Process the list of bw_meter entries */
2912 	while (tmp_list != NULL) {
2913 	    x = tmp_list;
2914 	    tmp_list = tmp_list->bm_time_next;
2915 
2916 	    /* Test if the time interval is over */
2917 	    process_endtime = x->bm_start_time;
2918 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2919 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2920 		/* Not yet: reschedule, but don't reset */
2921 		int time_hash;
2922 
2923 		BW_METER_TIMEHASH(x, time_hash);
2924 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2925 		    /*
2926 		     * XXX: somehow the bin processing is a bit ahead of time.
2927 		     * Put the entry in the next bin.
2928 		     */
2929 		    if (++time_hash >= BW_METER_BUCKETS)
2930 			time_hash = 0;
2931 		}
2932 		x->bm_time_next = bw_meter_timers[time_hash];
2933 		bw_meter_timers[time_hash] = x;
2934 		x->bm_time_hash = time_hash;
2935 
2936 		continue;
2937 	    }
2938 
2939 	    /*
2940 	     * Test if we should deliver an upcall
2941 	     */
2942 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2943 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2944 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2945 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2946 		/* Prepare an upcall for delivery */
2947 		bw_meter_prepare_upcall(x, &now);
2948 	    }
2949 
2950 	    /*
2951 	     * Reschedule for next processing
2952 	     */
2953 	    schedule_bw_meter(x, &now);
2954 	}
2955     }
2956 
2957     /* Send all upcalls that are pending delivery */
2958     bw_upcalls_send();
2959 
2960     splx(s);
2961 }
2962 
2963 /*
2964  * A periodic function for sending all upcalls that are pending delivery
2965  */
2966 static void
2967 expire_bw_upcalls_send(void *unused)
2968 {
2969     int s;
2970 
2971     s = splsoftnet();
2972     bw_upcalls_send();
2973     splx(s);
2974 
2975     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2976 		  expire_bw_upcalls_send, NULL);
2977 }
2978 
2979 /*
2980  * A periodic function for periodic scanning of the multicast forwarding
2981  * table for processing all "<=" bw_meter entries.
2982  */
2983 static void
2984 expire_bw_meter_process(void *unused)
2985 {
2986     if (mrt_api_config & MRT_MFC_BW_UPCALL)
2987 	bw_meter_process();
2988 
2989     callout_reset(&bw_meter_ch, BW_METER_PERIOD,
2990 		  expire_bw_meter_process, NULL);
2991 }
2992 
2993 /*
2994  * End of bandwidth monitoring code
2995  */
2996 
2997 #ifdef PIM
2998 /*
2999  * Send the packet up to the user daemon, or eventually do kernel encapsulation
3000  */
3001 static int
3002 pim_register_send(struct ip *ip, struct vif *vifp,
3003 	struct mbuf *m, struct mfc *rt)
3004 {
3005     struct mbuf *mb_copy, *mm;
3006 
3007     if (mrtdebug & DEBUG_PIM)
3008         log(LOG_DEBUG, "pim_register_send: \n");
3009 
3010     mb_copy = pim_register_prepare(ip, m);
3011     if (mb_copy == NULL)
3012 	return ENOBUFS;
3013 
3014     /*
3015      * Send all the fragments. Note that the mbuf for each fragment
3016      * is freed by the sending machinery.
3017      */
3018     for (mm = mb_copy; mm; mm = mb_copy) {
3019 	mb_copy = mm->m_nextpkt;
3020 	mm->m_nextpkt = NULL;
3021 	mm = m_pullup(mm, sizeof(struct ip));
3022 	if (mm != NULL) {
3023 	    ip = mtod(mm, struct ip *);
3024 	    if ((mrt_api_config & MRT_MFC_RP) &&
3025 		!in_nullhost(rt->mfc_rp)) {
3026 		pim_register_send_rp(ip, vifp, mm, rt);
3027 	    } else {
3028 		pim_register_send_upcall(ip, vifp, mm, rt);
3029 	    }
3030 	}
3031     }
3032 
3033     return 0;
3034 }
3035 
3036 /*
3037  * Return a copy of the data packet that is ready for PIM Register
3038  * encapsulation.
3039  * XXX: Note that in the returned copy the IP header is a valid one.
3040  */
3041 static struct mbuf *
3042 pim_register_prepare(struct ip *ip, struct mbuf *m)
3043 {
3044     struct mbuf *mb_copy = NULL;
3045     int mtu;
3046 
3047     /* Take care of delayed checksums */
3048     if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3049 	in_delayed_cksum(m);
3050 	m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3051     }
3052 
3053     /*
3054      * Copy the old packet & pullup its IP header into the
3055      * new mbuf so we can modify it.
3056      */
3057     mb_copy = m_copypacket(m, M_DONTWAIT);
3058     if (mb_copy == NULL)
3059 	return NULL;
3060     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3061     if (mb_copy == NULL)
3062 	return NULL;
3063 
3064     /* take care of the TTL */
3065     ip = mtod(mb_copy, struct ip *);
3066     --ip->ip_ttl;
3067 
3068     /* Compute the MTU after the PIM Register encapsulation */
3069     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3070 
3071     if (ntohs(ip->ip_len) <= mtu) {
3072 	/* Turn the IP header into a valid one */
3073 	ip->ip_sum = 0;
3074 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3075     } else {
3076 	/* Fragment the packet */
3077 	if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3078 	    /* XXX: mb_copy was freed by ip_fragment() */
3079 	    return NULL;
3080 	}
3081     }
3082     return mb_copy;
3083 }
3084 
3085 /*
3086  * Send an upcall with the data packet to the user-level process.
3087  */
3088 static int
3089 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3090     struct mbuf *mb_copy, struct mfc *rt)
3091 {
3092     struct mbuf *mb_first;
3093     int len = ntohs(ip->ip_len);
3094     struct igmpmsg *im;
3095     struct sockaddr_in k_igmpsrc = {
3096 	    .sin_len = sizeof(k_igmpsrc),
3097 	    .sin_family = AF_INET,
3098     };
3099 
3100     /*
3101      * Add a new mbuf with an upcall header
3102      */
3103     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3104     if (mb_first == NULL) {
3105 	m_freem(mb_copy);
3106 	return ENOBUFS;
3107     }
3108     mb_first->m_data += max_linkhdr;
3109     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3110     mb_first->m_len = sizeof(struct igmpmsg);
3111     mb_first->m_next = mb_copy;
3112 
3113     /* Send message to routing daemon */
3114     im = mtod(mb_first, struct igmpmsg *);
3115     im->im_msgtype	= IGMPMSG_WHOLEPKT;
3116     im->im_mbz		= 0;
3117     im->im_vif		= vifp - viftable;
3118     im->im_src		= ip->ip_src;
3119     im->im_dst		= ip->ip_dst;
3120 
3121     k_igmpsrc.sin_addr	= ip->ip_src;
3122 
3123     mrtstat.mrts_upcalls++;
3124 
3125     if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3126 	if (mrtdebug & DEBUG_PIM)
3127 	    log(LOG_WARNING,
3128 		"mcast: pim_register_send_upcall: ip_mrouter socket queue full\n");
3129 	++mrtstat.mrts_upq_sockfull;
3130 	return ENOBUFS;
3131     }
3132 
3133     /* Keep statistics */
3134     pimstat.pims_snd_registers_msgs++;
3135     pimstat.pims_snd_registers_bytes += len;
3136 
3137     return 0;
3138 }
3139 
3140 /*
3141  * Encapsulate the data packet in PIM Register message and send it to the RP.
3142  */
3143 static int
3144 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3145 	struct mbuf *mb_copy, struct mfc *rt)
3146 {
3147     struct mbuf *mb_first;
3148     struct ip *ip_outer;
3149     struct pim_encap_pimhdr *pimhdr;
3150     int len = ntohs(ip->ip_len);
3151     vifi_t vifi = rt->mfc_parent;
3152 
3153     if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3154 	m_freem(mb_copy);
3155 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
3156     }
3157 
3158     /*
3159      * Add a new mbuf with the encapsulating header
3160      */
3161     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3162     if (mb_first == NULL) {
3163 	m_freem(mb_copy);
3164 	return ENOBUFS;
3165     }
3166     mb_first->m_data += max_linkhdr;
3167     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3168     mb_first->m_next = mb_copy;
3169 
3170     mb_first->m_pkthdr.len = len + mb_first->m_len;
3171 
3172     /*
3173      * Fill in the encapsulating IP and PIM header
3174      */
3175     ip_outer = mtod(mb_first, struct ip *);
3176     *ip_outer = pim_encap_iphdr;
3177      if (mb_first->m_pkthdr.len < IP_MINFRAGSIZE)
3178 	ip_outer->ip_id = 0;
3179     else
3180 	ip_outer->ip_id = ip_newid(NULL);
3181     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3182 			     sizeof(pim_encap_pimhdr));
3183     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3184     ip_outer->ip_dst = rt->mfc_rp;
3185     /*
3186      * Copy the inner header TOS to the outer header, and take care of the
3187      * IP_DF bit.
3188      */
3189     ip_outer->ip_tos = ip->ip_tos;
3190     if (ntohs(ip->ip_off) & IP_DF)
3191 	ip_outer->ip_off |= htons(IP_DF);
3192     pimhdr = (struct pim_encap_pimhdr *)((char *)ip_outer
3193 					 + sizeof(pim_encap_iphdr));
3194     *pimhdr = pim_encap_pimhdr;
3195     /* If the iif crosses a border, set the Border-bit */
3196     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3197 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3198 
3199     mb_first->m_data += sizeof(pim_encap_iphdr);
3200     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3201     mb_first->m_data -= sizeof(pim_encap_iphdr);
3202 
3203     if (vifp->v_rate_limit == 0)
3204 	tbf_send_packet(vifp, mb_first);
3205     else
3206 	tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3207 
3208     /* Keep statistics */
3209     pimstat.pims_snd_registers_msgs++;
3210     pimstat.pims_snd_registers_bytes += len;
3211 
3212     return 0;
3213 }
3214 
3215 /*
3216  * PIM-SMv2 and PIM-DM messages processing.
3217  * Receives and verifies the PIM control messages, and passes them
3218  * up to the listening socket, using rip_input().
3219  * The only message with special processing is the PIM_REGISTER message
3220  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3221  * is passed to if_simloop().
3222  */
3223 void
3224 pim_input(struct mbuf *m, ...)
3225 {
3226     struct ip *ip = mtod(m, struct ip *);
3227     struct pim *pim;
3228     int minlen;
3229     int datalen;
3230     int ip_tos;
3231     int proto;
3232     int iphlen;
3233     va_list ap;
3234 
3235     va_start(ap, m);
3236     iphlen = va_arg(ap, int);
3237     proto = va_arg(ap, int);
3238     va_end(ap);
3239 
3240     datalen = ntohs(ip->ip_len) - iphlen;
3241 
3242     /* Keep statistics */
3243     pimstat.pims_rcv_total_msgs++;
3244     pimstat.pims_rcv_total_bytes += datalen;
3245 
3246     /*
3247      * Validate lengths
3248      */
3249     if (datalen < PIM_MINLEN) {
3250 	pimstat.pims_rcv_tooshort++;
3251 	log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3252 	    datalen, (u_long)ip->ip_src.s_addr);
3253 	m_freem(m);
3254 	return;
3255     }
3256 
3257     /*
3258      * If the packet is at least as big as a REGISTER, go agead
3259      * and grab the PIM REGISTER header size, to avoid another
3260      * possible m_pullup() later.
3261      *
3262      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
3263      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3264      */
3265     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3266     /*
3267      * Get the IP and PIM headers in contiguous memory, and
3268      * possibly the PIM REGISTER header.
3269      */
3270     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3271 	(m = m_pullup(m, minlen)) == NULL) {
3272 	log(LOG_ERR, "pim_input: m_pullup failure\n");
3273 	return;
3274     }
3275     /* m_pullup() may have given us a new mbuf so reset ip. */
3276     ip = mtod(m, struct ip *);
3277     ip_tos = ip->ip_tos;
3278 
3279     /* adjust mbuf to point to the PIM header */
3280     m->m_data += iphlen;
3281     m->m_len  -= iphlen;
3282     pim = mtod(m, struct pim *);
3283 
3284     /*
3285      * Validate checksum. If PIM REGISTER, exclude the data packet.
3286      *
3287      * XXX: some older PIMv2 implementations don't make this distinction,
3288      * so for compatibility reason perform the checksum over part of the
3289      * message, and if error, then over the whole message.
3290      */
3291     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3292 	/* do nothing, checksum okay */
3293     } else if (in_cksum(m, datalen)) {
3294 	pimstat.pims_rcv_badsum++;
3295 	if (mrtdebug & DEBUG_PIM)
3296 	    log(LOG_DEBUG, "pim_input: invalid checksum\n");
3297 	m_freem(m);
3298 	return;
3299     }
3300 
3301     /* PIM version check */
3302     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3303 	pimstat.pims_rcv_badversion++;
3304 	log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3305 	    PIM_VT_V(pim->pim_vt), PIM_VERSION);
3306 	m_freem(m);
3307 	return;
3308     }
3309 
3310     /* restore mbuf back to the outer IP */
3311     m->m_data -= iphlen;
3312     m->m_len  += iphlen;
3313 
3314     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3315 	/*
3316 	 * Since this is a REGISTER, we'll make a copy of the register
3317 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3318 	 * routing daemon.
3319 	 */
3320 	int s;
3321 	struct sockaddr_in dst = {
3322 		.sin_len = sizeof(dst),
3323 		.sin_family = AF_INET,
3324 	};
3325 	struct mbuf *mcp;
3326 	struct ip *encap_ip;
3327 	u_int32_t *reghdr;
3328 	struct ifnet *vifp;
3329 
3330 	s = splsoftnet();
3331 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3332 	    splx(s);
3333 	    if (mrtdebug & DEBUG_PIM)
3334 		log(LOG_DEBUG,
3335 		    "pim_input: register vif not set: %d\n", reg_vif_num);
3336 	    m_freem(m);
3337 	    return;
3338 	}
3339 	/* XXX need refcnt? */
3340 	vifp = viftable[reg_vif_num].v_ifp;
3341 	splx(s);
3342 
3343 	/*
3344 	 * Validate length
3345 	 */
3346 	if (datalen < PIM_REG_MINLEN) {
3347 	    pimstat.pims_rcv_tooshort++;
3348 	    pimstat.pims_rcv_badregisters++;
3349 	    log(LOG_ERR,
3350 		"pim_input: register packet size too small %d from %lx\n",
3351 		datalen, (u_long)ip->ip_src.s_addr);
3352 	    m_freem(m);
3353 	    return;
3354 	}
3355 
3356 	reghdr = (u_int32_t *)(pim + 1);
3357 	encap_ip = (struct ip *)(reghdr + 1);
3358 
3359 	if (mrtdebug & DEBUG_PIM) {
3360 	    log(LOG_DEBUG,
3361 		"pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3362 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3363 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3364 		ntohs(encap_ip->ip_len));
3365 	}
3366 
3367 	/* verify the version number of the inner packet */
3368 	if (encap_ip->ip_v != IPVERSION) {
3369 	    pimstat.pims_rcv_badregisters++;
3370 	    if (mrtdebug & DEBUG_PIM) {
3371 		log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3372 		    "of the inner packet\n", encap_ip->ip_v);
3373 	    }
3374 	    m_freem(m);
3375 	    return;
3376 	}
3377 
3378 	/* verify the inner packet is destined to a mcast group */
3379 	if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3380 	    pimstat.pims_rcv_badregisters++;
3381 	    if (mrtdebug & DEBUG_PIM)
3382 		log(LOG_DEBUG,
3383 		    "pim_input: inner packet of register is not "
3384 		    "multicast %lx\n",
3385 		    (u_long)ntohl(encap_ip->ip_dst.s_addr));
3386 	    m_freem(m);
3387 	    return;
3388 	}
3389 
3390 	/* If a NULL_REGISTER, pass it to the daemon */
3391 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3392 	    goto pim_input_to_daemon;
3393 
3394 	/*
3395 	 * Copy the TOS from the outer IP header to the inner IP header.
3396 	 */
3397 	if (encap_ip->ip_tos != ip_tos) {
3398 	    /* Outer TOS -> inner TOS */
3399 	    encap_ip->ip_tos = ip_tos;
3400 	    /* Recompute the inner header checksum. Sigh... */
3401 
3402 	    /* adjust mbuf to point to the inner IP header */
3403 	    m->m_data += (iphlen + PIM_MINLEN);
3404 	    m->m_len  -= (iphlen + PIM_MINLEN);
3405 
3406 	    encap_ip->ip_sum = 0;
3407 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3408 
3409 	    /* restore mbuf to point back to the outer IP header */
3410 	    m->m_data -= (iphlen + PIM_MINLEN);
3411 	    m->m_len  += (iphlen + PIM_MINLEN);
3412 	}
3413 
3414 	/*
3415 	 * Decapsulate the inner IP packet and loopback to forward it
3416 	 * as a normal multicast packet. Also, make a copy of the
3417 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
3418 	 * to pass to the daemon later, so it can take the appropriate
3419 	 * actions (e.g., send back PIM_REGISTER_STOP).
3420 	 * XXX: here m->m_data points to the outer IP header.
3421 	 */
3422 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_DONTWAIT);
3423 	if (mcp == NULL) {
3424 	    log(LOG_ERR,
3425 		"pim_input: pim register: could not copy register head\n");
3426 	    m_freem(m);
3427 	    return;
3428 	}
3429 
3430 	/* Keep statistics */
3431 	/* XXX: registers_bytes include only the encap. mcast pkt */
3432 	pimstat.pims_rcv_registers_msgs++;
3433 	pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3434 
3435 	/*
3436 	 * forward the inner ip packet; point m_data at the inner ip.
3437 	 */
3438 	m_adj(m, iphlen + PIM_MINLEN);
3439 
3440 	if (mrtdebug & DEBUG_PIM) {
3441 	    log(LOG_DEBUG,
3442 		"pim_input: forwarding decapsulated register: "
3443 		"src %lx, dst %lx, vif %d\n",
3444 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3445 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3446 		reg_vif_num);
3447 	}
3448 	/* NB: vifp was collected above; can it change on us? */
3449 	looutput(vifp, m, (struct sockaddr *)&dst, NULL);
3450 
3451 	/* prepare the register head to send to the mrouting daemon */
3452 	m = mcp;
3453     }
3454 
3455 pim_input_to_daemon:
3456     /*
3457      * Pass the PIM message up to the daemon; if it is a Register message,
3458      * pass the 'head' only up to the daemon. This includes the
3459      * outer IP header, PIM header, PIM-Register header and the
3460      * inner IP header.
3461      * XXX: the outer IP header pkt size of a Register is not adjust to
3462      * reflect the fact that the inner multicast data is truncated.
3463      */
3464     rip_input(m, iphlen, proto);
3465 
3466     return;
3467 }
3468 #endif /* PIM */
3469