xref: /freebsd/sys/netinet/ip_mroute.c (revision a9148abd9da5db2f1c682fb17bed791845fc41c9)
1 /*-
2  * Copyright (c) 1989 Stephen Deering
3  * Copyright (c) 1992, 1993
4  *      The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Stephen Deering of Stanford University.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
34  */
35 
36 /*
37  * IP multicast forwarding procedures
38  *
39  * Written by David Waitzman, BBN Labs, August 1988.
40  * Modified by Steve Deering, Stanford, February 1989.
41  * Modified by Mark J. Steiglitz, Stanford, May, 1991
42  * Modified by Van Jacobson, LBL, January 1993
43  * Modified by Ajit Thyagarajan, PARC, August 1993
44  * Modified by Bill Fenner, PARC, April 1995
45  * Modified by Ahmed Helmy, SGI, June 1996
46  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
47  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
48  * Modified by Hitoshi Asaeda, WIDE, August 2000
49  * Modified by Pavlin Radoslavov, ICSI, October 2002
50  *
51  * MROUTING Revision: 3.5
52  * and PIM-SMv2 and PIM-DM support, advanced API support,
53  * bandwidth metering and signaling
54  */
55 
56 #include <sys/cdefs.h>
57 __FBSDID("$FreeBSD$");
58 
59 #include "opt_inet.h"
60 #include "opt_inet6.h"
61 #include "opt_mac.h"
62 #include "opt_mrouting.h"
63 
64 #define _PIM_VT 1
65 
66 #include <sys/param.h>
67 #include <sys/kernel.h>
68 #include <sys/lock.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/module.h>
72 #include <sys/priv.h>
73 #include <sys/protosw.h>
74 #include <sys/signalvar.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/sockio.h>
78 #include <sys/sx.h>
79 #include <sys/sysctl.h>
80 #include <sys/syslog.h>
81 #include <sys/systm.h>
82 #include <sys/time.h>
83 #include <sys/vimage.h>
84 #include <net/if.h>
85 #include <net/netisr.h>
86 #include <net/route.h>
87 #include <netinet/in.h>
88 #include <netinet/igmp.h>
89 #include <netinet/in_systm.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip.h>
92 #include <netinet/ip_encap.h>
93 #include <netinet/ip_mroute.h>
94 #include <netinet/ip_var.h>
95 #include <netinet/ip_options.h>
96 #include <netinet/pim.h>
97 #include <netinet/pim_var.h>
98 #include <netinet/udp.h>
99 #ifdef INET6
100 #include <netinet/ip6.h>
101 #include <netinet6/in6_var.h>
102 #include <netinet6/ip6_mroute.h>
103 #include <netinet6/ip6_var.h>
104 #endif
105 #include <machine/in_cksum.h>
106 
107 #include <security/mac/mac_framework.h>
108 
109 /*
110  * Control debugging code for rsvp and multicast routing code.
111  * Can only set them with the debugger.
112  */
113 static u_int    rsvpdebug;		/* non-zero enables debugging	*/
114 
115 static u_int	mrtdebug;		/* any set of the flags below	*/
116 #define		DEBUG_MFC	0x02
117 #define		DEBUG_FORWARD	0x04
118 #define		DEBUG_EXPIRE	0x08
119 #define		DEBUG_XMIT	0x10
120 #define		DEBUG_PIM	0x20
121 
122 #define		VIFI_INVALID	((vifi_t) -1)
123 
124 #define M_HASCL(m)	((m)->m_flags & M_EXT)
125 
126 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables");
127 
128 /*
129  * Locking.  We use two locks: one for the virtual interface table and
130  * one for the forwarding table.  These locks may be nested in which case
131  * the VIF lock must always be taken first.  Note that each lock is used
132  * to cover not only the specific data structure but also related data
133  * structures.  It may be better to add more fine-grained locking later;
134  * it's not clear how performance-critical this code is.
135  *
136  * XXX: This module could particularly benefit from being cleaned
137  *      up to use the <sys/queue.h> macros.
138  *
139  */
140 
141 static struct mrtstat	mrtstat;
142 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
143     &mrtstat, mrtstat,
144     "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)");
145 
146 static struct mfc	*mfctable[MFCTBLSIZ];
147 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
148     &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]",
149     "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)");
150 
151 static struct mtx mrouter_mtx;
152 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
153 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
154 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
155 #define	MROUTER_LOCK_INIT()	\
156 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
157 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
158 
159 static struct mtx mfc_mtx;
160 #define	MFC_LOCK()	mtx_lock(&mfc_mtx)
161 #define	MFC_UNLOCK()	mtx_unlock(&mfc_mtx)
162 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
163 #define	MFC_LOCK_INIT()	mtx_init(&mfc_mtx, "mroute mfc table", NULL, MTX_DEF)
164 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
165 
166 static struct vif	viftable[MAXVIFS];
167 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
168     &viftable, sizeof(viftable), "S,vif[MAXVIFS]",
169     "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
170 
171 static struct mtx vif_mtx;
172 #define	VIF_LOCK()	mtx_lock(&vif_mtx)
173 #define	VIF_UNLOCK()	mtx_unlock(&vif_mtx)
174 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
175 #define	VIF_LOCK_INIT()	mtx_init(&vif_mtx, "mroute vif table", NULL, MTX_DEF)
176 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
177 
178 static u_char		nexpire[MFCTBLSIZ];
179 
180 static eventhandler_tag if_detach_event_tag = NULL;
181 
182 static struct callout expire_upcalls_ch;
183 
184 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
185 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
186 
187 #define ENCAP_TTL 64
188 
189 /*
190  * Bandwidth meter variables and constants
191  */
192 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
193 /*
194  * Pending timeouts are stored in a hash table, the key being the
195  * expiration time. Periodically, the entries are analysed and processed.
196  */
197 #define BW_METER_BUCKETS	1024
198 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
199 static struct callout bw_meter_ch;
200 #define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
201 
202 /*
203  * Pending upcalls are stored in a vector which is flushed when
204  * full, or periodically
205  */
206 static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
207 static u_int	bw_upcalls_n; /* # of pending upcalls */
208 static struct callout bw_upcalls_ch;
209 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
210 
211 static struct pimstat pimstat;
212 
213 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
214 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
215     &pimstat, pimstat,
216     "PIM Statistics (struct pimstat, netinet/pim_var.h)");
217 
218 static u_long	pim_squelch_wholepkt = 0;
219 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
220     &pim_squelch_wholepkt, 0,
221     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
222 
223 extern  struct domain inetdomain;
224 struct protosw in_pim_protosw = {
225 	.pr_type =		SOCK_RAW,
226 	.pr_domain =		&inetdomain,
227 	.pr_protocol =		IPPROTO_PIM,
228 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
229 	.pr_input =		pim_input,
230 	.pr_output =		(pr_output_t*)rip_output,
231 	.pr_ctloutput =		rip_ctloutput,
232 	.pr_usrreqs =		&rip_usrreqs
233 };
234 static const struct encaptab *pim_encap_cookie;
235 
236 #ifdef INET6
237 /* ip6_mroute.c glue */
238 extern struct in6_protosw in6_pim_protosw;
239 static const struct encaptab *pim6_encap_cookie;
240 
241 extern int X_ip6_mrouter_set(struct socket *, struct sockopt *);
242 extern int X_ip6_mrouter_get(struct socket *, struct sockopt *);
243 extern int X_ip6_mrouter_done(void);
244 extern int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *);
245 extern int X_mrt6_ioctl(int, caddr_t);
246 #endif
247 
248 static int pim_encapcheck(const struct mbuf *, int, int, void *);
249 
250 /*
251  * Note: the PIM Register encapsulation adds the following in front of a
252  * data packet:
253  *
254  * struct pim_encap_hdr {
255  *    struct ip ip;
256  *    struct pim_encap_pimhdr  pim;
257  * }
258  *
259  */
260 
261 struct pim_encap_pimhdr {
262 	struct pim pim;
263 	uint32_t   flags;
264 };
265 
266 static struct ip pim_encap_iphdr = {
267 #if BYTE_ORDER == LITTLE_ENDIAN
268 	sizeof(struct ip) >> 2,
269 	IPVERSION,
270 #else
271 	IPVERSION,
272 	sizeof(struct ip) >> 2,
273 #endif
274 	0,			/* tos */
275 	sizeof(struct ip),	/* total length */
276 	0,			/* id */
277 	0,			/* frag offset */
278 	ENCAP_TTL,
279 	IPPROTO_PIM,
280 	0,			/* checksum */
281 };
282 
283 static struct pim_encap_pimhdr pim_encap_pimhdr = {
284     {
285 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
286 	0,			/* reserved */
287 	0,			/* checksum */
288     },
289     0				/* flags */
290 };
291 
292 static struct ifnet multicast_register_if;
293 static vifi_t reg_vif_num = VIFI_INVALID;
294 
295 /*
296  * Private variables.
297  */
298 static vifi_t	   numvifs;
299 
300 static u_long	X_ip_mcast_src(int vifi);
301 static int	X_ip_mforward(struct ip *ip, struct ifnet *ifp,
302 			struct mbuf *m, struct ip_moptions *imo);
303 static int	X_ip_mrouter_done(void);
304 static int	X_ip_mrouter_get(struct socket *so, struct sockopt *m);
305 static int	X_ip_mrouter_set(struct socket *so, struct sockopt *m);
306 static int	X_legal_vif_num(int vif);
307 static int	X_mrt_ioctl(int cmd, caddr_t data, int fibnum);
308 
309 static int get_sg_cnt(struct sioc_sg_req *);
310 static int get_vif_cnt(struct sioc_vif_req *);
311 static void if_detached_event(void *arg __unused, struct ifnet *);
312 static int ip_mrouter_init(struct socket *, int);
313 static int add_vif(struct vifctl *);
314 static int del_vif_locked(vifi_t);
315 static int del_vif(vifi_t);
316 static int add_mfc(struct mfcctl2 *);
317 static int del_mfc(struct mfcctl2 *);
318 static int set_api_config(uint32_t *); /* chose API capabilities */
319 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
320 static int set_assert(int);
321 static void expire_upcalls(void *);
322 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
323 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
324 static void send_packet(struct vif *, struct mbuf *);
325 
326 /*
327  * Bandwidth monitoring
328  */
329 static void free_bw_list(struct bw_meter *list);
330 static int add_bw_upcall(struct bw_upcall *);
331 static int del_bw_upcall(struct bw_upcall *);
332 static void bw_meter_receive_packet(struct bw_meter *x, int plen,
333 		struct timeval *nowp);
334 static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp);
335 static void bw_upcalls_send(void);
336 static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp);
337 static void unschedule_bw_meter(struct bw_meter *x);
338 static void bw_meter_process(void);
339 static void expire_bw_upcalls_send(void *);
340 static void expire_bw_meter_process(void *);
341 
342 static int pim_register_send(struct ip *, struct vif *,
343 		struct mbuf *, struct mfc *);
344 static int pim_register_send_rp(struct ip *, struct vif *,
345 		struct mbuf *, struct mfc *);
346 static int pim_register_send_upcall(struct ip *, struct vif *,
347 		struct mbuf *, struct mfc *);
348 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
349 
350 /*
351  * whether or not special PIM assert processing is enabled.
352  */
353 static int pim_assert;
354 /*
355  * Rate limit for assert notification messages, in usec
356  */
357 #define ASSERT_MSG_TIME		3000000
358 
359 /*
360  * Kernel multicast routing API capabilities and setup.
361  * If more API capabilities are added to the kernel, they should be
362  * recorded in `mrt_api_support'.
363  */
364 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
365 					 MRT_MFC_FLAGS_BORDER_VIF |
366 					 MRT_MFC_RP |
367 					 MRT_MFC_BW_UPCALL);
368 static uint32_t mrt_api_config = 0;
369 
370 /*
371  * Hash function for a source, group entry
372  */
373 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
374 			((g) >> 20) ^ ((g) >> 10) ^ (g))
375 
376 /*
377  * Find a route for a given origin IP address and Multicast group address
378  * Statistics are updated by the caller if needed
379  * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
380  */
381 static struct mfc *
382 mfc_find(in_addr_t o, in_addr_t g)
383 {
384     struct mfc *rt;
385 
386     MFC_LOCK_ASSERT();
387 
388     for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next)
389 	if ((rt->mfc_origin.s_addr == o) &&
390 		(rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL))
391 	    break;
392     return rt;
393 }
394 
395 /*
396  * Macros to compute elapsed time efficiently
397  * Borrowed from Van Jacobson's scheduling code
398  */
399 #define TV_DELTA(a, b, delta) {					\
400 	int xxs;						\
401 	delta = (a).tv_usec - (b).tv_usec;			\
402 	if ((xxs = (a).tv_sec - (b).tv_sec)) {			\
403 		switch (xxs) {					\
404 		case 2:						\
405 		      delta += 1000000;				\
406 		      /* FALLTHROUGH */				\
407 		case 1:						\
408 		      delta += 1000000;				\
409 		      break;					\
410 		default:					\
411 		      delta += (1000000 * xxs);			\
412 		}						\
413 	}							\
414 }
415 
416 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
417 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
418 
419 /*
420  * Handle MRT setsockopt commands to modify the multicast routing tables.
421  */
422 static int
423 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
424 {
425     INIT_VNET_INET(curvnet);
426     int	error, optval;
427     vifi_t	vifi;
428     struct	vifctl vifc;
429     struct	mfcctl2 mfc;
430     struct	bw_upcall bw_upcall;
431     uint32_t	i;
432 
433     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
434 	return EPERM;
435 
436     error = 0;
437     switch (sopt->sopt_name) {
438     case MRT_INIT:
439 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
440 	if (error)
441 	    break;
442 	error = ip_mrouter_init(so, optval);
443 	break;
444 
445     case MRT_DONE:
446 	error = ip_mrouter_done();
447 	break;
448 
449     case MRT_ADD_VIF:
450 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
451 	if (error)
452 	    break;
453 	error = add_vif(&vifc);
454 	break;
455 
456     case MRT_DEL_VIF:
457 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
458 	if (error)
459 	    break;
460 	error = del_vif(vifi);
461 	break;
462 
463     case MRT_ADD_MFC:
464     case MRT_DEL_MFC:
465 	/*
466 	 * select data size depending on API version.
467 	 */
468 	if (sopt->sopt_name == MRT_ADD_MFC &&
469 		mrt_api_config & MRT_API_FLAGS_ALL) {
470 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
471 				sizeof(struct mfcctl2));
472 	} else {
473 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
474 				sizeof(struct mfcctl));
475 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
476 			sizeof(mfc) - sizeof(struct mfcctl));
477 	}
478 	if (error)
479 	    break;
480 	if (sopt->sopt_name == MRT_ADD_MFC)
481 	    error = add_mfc(&mfc);
482 	else
483 	    error = del_mfc(&mfc);
484 	break;
485 
486     case MRT_ASSERT:
487 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
488 	if (error)
489 	    break;
490 	set_assert(optval);
491 	break;
492 
493     case MRT_API_CONFIG:
494 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
495 	if (!error)
496 	    error = set_api_config(&i);
497 	if (!error)
498 	    error = sooptcopyout(sopt, &i, sizeof i);
499 	break;
500 
501     case MRT_ADD_BW_UPCALL:
502     case MRT_DEL_BW_UPCALL:
503 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
504 				sizeof bw_upcall);
505 	if (error)
506 	    break;
507 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
508 	    error = add_bw_upcall(&bw_upcall);
509 	else
510 	    error = del_bw_upcall(&bw_upcall);
511 	break;
512 
513     default:
514 	error = EOPNOTSUPP;
515 	break;
516     }
517     return error;
518 }
519 
520 /*
521  * Handle MRT getsockopt commands
522  */
523 static int
524 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
525 {
526     int error;
527     static int version = 0x0305; /* !!! why is this here? XXX */
528 
529     switch (sopt->sopt_name) {
530     case MRT_VERSION:
531 	error = sooptcopyout(sopt, &version, sizeof version);
532 	break;
533 
534     case MRT_ASSERT:
535 	error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert);
536 	break;
537 
538     case MRT_API_SUPPORT:
539 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
540 	break;
541 
542     case MRT_API_CONFIG:
543 	error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config);
544 	break;
545 
546     default:
547 	error = EOPNOTSUPP;
548 	break;
549     }
550     return error;
551 }
552 
553 /*
554  * Handle ioctl commands to obtain information from the cache
555  */
556 static int
557 X_mrt_ioctl(int cmd, caddr_t data, int fibnum)
558 {
559     int error = 0;
560 
561     /*
562      * Currently the only function calling this ioctl routine is rtioctl().
563      * Typically, only root can create the raw socket in order to execute
564      * this ioctl method, however the request might be coming from a prison
565      */
566     error = priv_check(curthread, PRIV_NETINET_MROUTE);
567     if (error)
568 	return (error);
569     switch (cmd) {
570     case (SIOCGETVIFCNT):
571 	error = get_vif_cnt((struct sioc_vif_req *)data);
572 	break;
573 
574     case (SIOCGETSGCNT):
575 	error = get_sg_cnt((struct sioc_sg_req *)data);
576 	break;
577 
578     default:
579 	error = EINVAL;
580 	break;
581     }
582     return error;
583 }
584 
585 /*
586  * returns the packet, byte, rpf-failure count for the source group provided
587  */
588 static int
589 get_sg_cnt(struct sioc_sg_req *req)
590 {
591     struct mfc *rt;
592 
593     MFC_LOCK();
594     rt = mfc_find(req->src.s_addr, req->grp.s_addr);
595     if (rt == NULL) {
596 	MFC_UNLOCK();
597 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
598 	return EADDRNOTAVAIL;
599     }
600     req->pktcnt = rt->mfc_pkt_cnt;
601     req->bytecnt = rt->mfc_byte_cnt;
602     req->wrong_if = rt->mfc_wrong_if;
603     MFC_UNLOCK();
604     return 0;
605 }
606 
607 /*
608  * returns the input and output packet and byte counts on the vif provided
609  */
610 static int
611 get_vif_cnt(struct sioc_vif_req *req)
612 {
613     vifi_t vifi = req->vifi;
614 
615     VIF_LOCK();
616     if (vifi >= numvifs) {
617 	VIF_UNLOCK();
618 	return EINVAL;
619     }
620 
621     req->icount = viftable[vifi].v_pkt_in;
622     req->ocount = viftable[vifi].v_pkt_out;
623     req->ibytes = viftable[vifi].v_bytes_in;
624     req->obytes = viftable[vifi].v_bytes_out;
625     VIF_UNLOCK();
626 
627     return 0;
628 }
629 
630 static void
631 ip_mrouter_reset(void)
632 {
633     bzero((caddr_t)mfctable, sizeof(mfctable));
634     bzero((caddr_t)nexpire, sizeof(nexpire));
635 
636     pim_assert = 0;
637     mrt_api_config = 0;
638 
639     callout_init(&expire_upcalls_ch, CALLOUT_MPSAFE);
640 
641     bw_upcalls_n = 0;
642     bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers));
643     callout_init(&bw_upcalls_ch, CALLOUT_MPSAFE);
644     callout_init(&bw_meter_ch, CALLOUT_MPSAFE);
645 }
646 
647 static void
648 if_detached_event(void *arg __unused, struct ifnet *ifp)
649 {
650     INIT_VNET_INET(curvnet);
651     vifi_t vifi;
652     int i;
653     struct mfc *mfc;
654     struct mfc *nmfc;
655     struct mfc **ppmfc;	/* Pointer to previous node's next-pointer */
656     struct rtdetq *pq;
657     struct rtdetq *npq;
658 
659     MROUTER_LOCK();
660     if (V_ip_mrouter == NULL) {
661 	MROUTER_UNLOCK();
662     }
663 
664     /*
665      * Tear down multicast forwarder state associated with this ifnet.
666      * 1. Walk the vif list, matching vifs against this ifnet.
667      * 2. Walk the multicast forwarding cache (mfc) looking for
668      *    inner matches with this vif's index.
669      * 3. Free any pending mbufs for this mfc.
670      * 4. Free the associated mfc entry and state associated with this vif.
671      *    Be very careful about unlinking from a singly-linked list whose
672      *    "head node" is a pointer in a simple array.
673      * 5. Free vif state. This should disable ALLMULTI on the interface.
674      */
675     VIF_LOCK();
676     MFC_LOCK();
677     for (vifi = 0; vifi < numvifs; vifi++) {
678 	if (viftable[vifi].v_ifp != ifp)
679 		continue;
680 	for (i = 0; i < MFCTBLSIZ; i++) {
681 	    ppmfc = &mfctable[i];
682 	    for (mfc = mfctable[i]; mfc != NULL; ) {
683 		nmfc = mfc->mfc_next;
684 		if (mfc->mfc_parent == vifi) {
685 		    for (pq = mfc->mfc_stall; pq != NULL; ) {
686 			npq = pq->next;
687 			m_freem(pq->m);
688 			free(pq, M_MRTABLE);
689 			pq = npq;
690 		    }
691 		    free_bw_list(mfc->mfc_bw_meter);
692 		    free(mfc, M_MRTABLE);
693 		    *ppmfc = nmfc;
694 		} else {
695 		    ppmfc = &mfc->mfc_next;
696 		}
697 		mfc = nmfc;
698 	    }
699 	}
700 	del_vif_locked(vifi);
701     }
702     MFC_UNLOCK();
703     VIF_UNLOCK();
704 
705     MROUTER_UNLOCK();
706 }
707 
708 /*
709  * Enable multicast routing
710  */
711 static int
712 ip_mrouter_init(struct socket *so, int version)
713 {
714     INIT_VNET_INET(curvnet);
715 
716     if (mrtdebug)
717 	log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
718 	    so->so_type, so->so_proto->pr_protocol);
719 
720     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
721 	return EOPNOTSUPP;
722 
723     if (version != 1)
724 	return ENOPROTOOPT;
725 
726     MROUTER_LOCK();
727 
728     if (V_ip_mrouter != NULL) {
729 	MROUTER_UNLOCK();
730 	return EADDRINUSE;
731     }
732 
733     if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
734         if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
735     if (if_detach_event_tag == NULL) {
736 	MROUTER_UNLOCK();
737 	return (ENOMEM);
738     }
739 
740     callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
741 
742     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
743 	expire_bw_upcalls_send, NULL);
744     callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
745 
746     V_ip_mrouter = so;
747 
748     MROUTER_UNLOCK();
749 
750     if (mrtdebug)
751 	log(LOG_DEBUG, "ip_mrouter_init\n");
752 
753     return 0;
754 }
755 
756 /*
757  * Disable multicast routing
758  */
759 static int
760 X_ip_mrouter_done(void)
761 {
762     INIT_VNET_INET(curvnet);
763     vifi_t vifi;
764     int i;
765     struct ifnet *ifp;
766     struct ifreq ifr;
767     struct mfc *rt;
768     struct rtdetq *rte;
769 
770     MROUTER_LOCK();
771 
772     if (V_ip_mrouter == NULL) {
773 	MROUTER_UNLOCK();
774 	return EINVAL;
775     }
776 
777     /*
778      * Detach/disable hooks to the reset of the system.
779      */
780     V_ip_mrouter = NULL;
781     mrt_api_config = 0;
782 
783     VIF_LOCK();
784     /*
785      * For each phyint in use, disable promiscuous reception of all IP
786      * multicasts.
787      */
788     for (vifi = 0; vifi < numvifs; vifi++) {
789 	if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
790 		!(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
791 	    struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
792 
793 	    so->sin_len = sizeof(struct sockaddr_in);
794 	    so->sin_family = AF_INET;
795 	    so->sin_addr.s_addr = INADDR_ANY;
796 	    ifp = viftable[vifi].v_ifp;
797 	    if_allmulti(ifp, 0);
798 	}
799     }
800     bzero((caddr_t)viftable, sizeof(viftable));
801     numvifs = 0;
802     pim_assert = 0;
803     VIF_UNLOCK();
804     EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
805 
806     /*
807      * Free all multicast forwarding cache entries.
808      */
809     callout_stop(&expire_upcalls_ch);
810     callout_stop(&bw_upcalls_ch);
811     callout_stop(&bw_meter_ch);
812 
813     MFC_LOCK();
814     for (i = 0; i < MFCTBLSIZ; i++) {
815 	for (rt = mfctable[i]; rt != NULL; ) {
816 	    struct mfc *nr = rt->mfc_next;
817 
818 	    for (rte = rt->mfc_stall; rte != NULL; ) {
819 		struct rtdetq *n = rte->next;
820 
821 		m_freem(rte->m);
822 		free(rte, M_MRTABLE);
823 		rte = n;
824 	    }
825 	    free_bw_list(rt->mfc_bw_meter);
826 	    free(rt, M_MRTABLE);
827 	    rt = nr;
828 	}
829     }
830     bzero((caddr_t)mfctable, sizeof(mfctable));
831     bzero((caddr_t)nexpire, sizeof(nexpire));
832     bw_upcalls_n = 0;
833     bzero(bw_meter_timers, sizeof(bw_meter_timers));
834     MFC_UNLOCK();
835 
836     reg_vif_num = VIFI_INVALID;
837 
838     MROUTER_UNLOCK();
839 
840     if (mrtdebug)
841 	log(LOG_DEBUG, "ip_mrouter_done\n");
842 
843     return 0;
844 }
845 
846 /*
847  * Set PIM assert processing global
848  */
849 static int
850 set_assert(int i)
851 {
852     if ((i != 1) && (i != 0))
853 	return EINVAL;
854 
855     pim_assert = i;
856 
857     return 0;
858 }
859 
860 /*
861  * Configure API capabilities
862  */
863 int
864 set_api_config(uint32_t *apival)
865 {
866     int i;
867 
868     /*
869      * We can set the API capabilities only if it is the first operation
870      * after MRT_INIT. I.e.:
871      *  - there are no vifs installed
872      *  - pim_assert is not enabled
873      *  - the MFC table is empty
874      */
875     if (numvifs > 0) {
876 	*apival = 0;
877 	return EPERM;
878     }
879     if (pim_assert) {
880 	*apival = 0;
881 	return EPERM;
882     }
883     for (i = 0; i < MFCTBLSIZ; i++) {
884 	if (mfctable[i] != NULL) {
885 	    *apival = 0;
886 	    return EPERM;
887 	}
888     }
889 
890     mrt_api_config = *apival & mrt_api_support;
891     *apival = mrt_api_config;
892 
893     return 0;
894 }
895 
896 /*
897  * Add a vif to the vif table
898  */
899 static int
900 add_vif(struct vifctl *vifcp)
901 {
902     struct vif *vifp = viftable + vifcp->vifc_vifi;
903     struct sockaddr_in sin = {sizeof sin, AF_INET};
904     struct ifaddr *ifa;
905     struct ifnet *ifp;
906     int error;
907 
908     VIF_LOCK();
909     if (vifcp->vifc_vifi >= MAXVIFS) {
910 	VIF_UNLOCK();
911 	return EINVAL;
912     }
913     /* rate limiting is no longer supported by this code */
914     if (vifcp->vifc_rate_limit != 0) {
915 	log(LOG_ERR, "rate limiting is no longer supported\n");
916 	VIF_UNLOCK();
917 	return EINVAL;
918     }
919     if (vifp->v_lcl_addr.s_addr != INADDR_ANY) {
920 	VIF_UNLOCK();
921 	return EADDRINUSE;
922     }
923     if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) {
924 	VIF_UNLOCK();
925 	return EADDRNOTAVAIL;
926     }
927 
928     /* Find the interface with an address in AF_INET family */
929     if (vifcp->vifc_flags & VIFF_REGISTER) {
930 	/*
931 	 * XXX: Because VIFF_REGISTER does not really need a valid
932 	 * local interface (e.g. it could be 127.0.0.2), we don't
933 	 * check its address.
934 	 */
935 	ifp = NULL;
936     } else {
937 	sin.sin_addr = vifcp->vifc_lcl_addr;
938 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
939 	if (ifa == NULL) {
940 	    VIF_UNLOCK();
941 	    return EADDRNOTAVAIL;
942 	}
943 	ifp = ifa->ifa_ifp;
944     }
945 
946     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
947 	log(LOG_ERR, "tunnels are no longer supported\n");
948 	VIF_UNLOCK();
949 	return EOPNOTSUPP;
950     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
951 	ifp = &multicast_register_if;
952 	if (mrtdebug)
953 	    log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
954 		    (void *)&multicast_register_if);
955 	if (reg_vif_num == VIFI_INVALID) {
956 	    if_initname(&multicast_register_if, "register_vif", 0);
957 	    multicast_register_if.if_flags = IFF_LOOPBACK;
958 	    reg_vif_num = vifcp->vifc_vifi;
959 	}
960     } else {		/* Make sure the interface supports multicast */
961 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
962 	    VIF_UNLOCK();
963 	    return EOPNOTSUPP;
964 	}
965 
966 	/* Enable promiscuous reception of all IP multicasts from the if */
967 	error = if_allmulti(ifp, 1);
968 	if (error) {
969 	    VIF_UNLOCK();
970 	    return error;
971 	}
972     }
973 
974     vifp->v_flags     = vifcp->vifc_flags;
975     vifp->v_threshold = vifcp->vifc_threshold;
976     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
977     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
978     vifp->v_ifp       = ifp;
979     vifp->v_rsvp_on   = 0;
980     vifp->v_rsvpd     = NULL;
981     /* initialize per vif pkt counters */
982     vifp->v_pkt_in    = 0;
983     vifp->v_pkt_out   = 0;
984     vifp->v_bytes_in  = 0;
985     vifp->v_bytes_out = 0;
986     bzero(&vifp->v_route, sizeof(vifp->v_route));
987 
988     /* Adjust numvifs up if the vifi is higher than numvifs */
989     if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
990 
991     VIF_UNLOCK();
992 
993     if (mrtdebug)
994 	log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x\n",
995 	    vifcp->vifc_vifi,
996 	    (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr),
997 	    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
998 	    (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr),
999 	    vifcp->vifc_threshold);
1000 
1001     return 0;
1002 }
1003 
1004 /*
1005  * Delete a vif from the vif table
1006  */
1007 static int
1008 del_vif_locked(vifi_t vifi)
1009 {
1010     struct vif *vifp;
1011 
1012     VIF_LOCK_ASSERT();
1013 
1014     if (vifi >= numvifs) {
1015 	return EINVAL;
1016     }
1017     vifp = &viftable[vifi];
1018     if (vifp->v_lcl_addr.s_addr == INADDR_ANY) {
1019 	return EADDRNOTAVAIL;
1020     }
1021 
1022     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1023 	if_allmulti(vifp->v_ifp, 0);
1024 
1025     if (vifp->v_flags & VIFF_REGISTER)
1026 	reg_vif_num = VIFI_INVALID;
1027 
1028     bzero((caddr_t)vifp, sizeof (*vifp));
1029 
1030     if (mrtdebug)
1031 	log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs);
1032 
1033     /* Adjust numvifs down */
1034     for (vifi = numvifs; vifi > 0; vifi--)
1035 	if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY)
1036 	    break;
1037     numvifs = vifi;
1038 
1039     return 0;
1040 }
1041 
1042 static int
1043 del_vif(vifi_t vifi)
1044 {
1045     int cc;
1046 
1047     VIF_LOCK();
1048     cc = del_vif_locked(vifi);
1049     VIF_UNLOCK();
1050 
1051     return cc;
1052 }
1053 
1054 /*
1055  * update an mfc entry without resetting counters and S,G addresses.
1056  */
1057 static void
1058 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1059 {
1060     int i;
1061 
1062     rt->mfc_parent = mfccp->mfcc_parent;
1063     for (i = 0; i < numvifs; i++) {
1064 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1065 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1066 	    MRT_MFC_FLAGS_ALL;
1067     }
1068     /* set the RP address */
1069     if (mrt_api_config & MRT_MFC_RP)
1070 	rt->mfc_rp = mfccp->mfcc_rp;
1071     else
1072 	rt->mfc_rp.s_addr = INADDR_ANY;
1073 }
1074 
1075 /*
1076  * fully initialize an mfc entry from the parameter.
1077  */
1078 static void
1079 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1080 {
1081     rt->mfc_origin     = mfccp->mfcc_origin;
1082     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1083 
1084     update_mfc_params(rt, mfccp);
1085 
1086     /* initialize pkt counters per src-grp */
1087     rt->mfc_pkt_cnt    = 0;
1088     rt->mfc_byte_cnt   = 0;
1089     rt->mfc_wrong_if   = 0;
1090     rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
1091 }
1092 
1093 
1094 /*
1095  * Add an mfc entry
1096  */
1097 static int
1098 add_mfc(struct mfcctl2 *mfccp)
1099 {
1100     struct mfc *rt;
1101     u_long hash;
1102     struct rtdetq *rte;
1103     u_short nstl;
1104 
1105     VIF_LOCK();
1106     MFC_LOCK();
1107 
1108     rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
1109 
1110     /* If an entry already exists, just update the fields */
1111     if (rt) {
1112 	if (mrtdebug & DEBUG_MFC)
1113 	    log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n",
1114 		(u_long)ntohl(mfccp->mfcc_origin.s_addr),
1115 		(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1116 		mfccp->mfcc_parent);
1117 
1118 	update_mfc_params(rt, mfccp);
1119 	MFC_UNLOCK();
1120 	VIF_UNLOCK();
1121 	return 0;
1122     }
1123 
1124     /*
1125      * Find the entry for which the upcall was made and update
1126      */
1127     hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
1128     for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) {
1129 
1130 	if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1131 		(rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
1132 		(rt->mfc_stall != NULL)) {
1133 
1134 	    if (nstl++)
1135 		log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n",
1136 		    "multiple kernel entries",
1137 		    (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1138 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1139 		    mfccp->mfcc_parent, (void *)rt->mfc_stall);
1140 
1141 	    if (mrtdebug & DEBUG_MFC)
1142 		log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n",
1143 		    (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1144 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1145 		    mfccp->mfcc_parent, (void *)rt->mfc_stall);
1146 
1147 	    init_mfc_params(rt, mfccp);
1148 
1149 	    rt->mfc_expire = 0;	/* Don't clean this guy up */
1150 	    nexpire[hash]--;
1151 
1152 	    /* free packets Qed at the end of this entry */
1153 	    for (rte = rt->mfc_stall; rte != NULL; ) {
1154 		struct rtdetq *n = rte->next;
1155 
1156 		ip_mdq(rte->m, rte->ifp, rt, -1);
1157 		m_freem(rte->m);
1158 		free(rte, M_MRTABLE);
1159 		rte = n;
1160 	    }
1161 	    rt->mfc_stall = NULL;
1162 	}
1163     }
1164 
1165     /*
1166      * It is possible that an entry is being inserted without an upcall
1167      */
1168     if (nstl == 0) {
1169 	if (mrtdebug & DEBUG_MFC)
1170 	    log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n",
1171 		hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1172 		(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1173 		mfccp->mfcc_parent);
1174 
1175 	for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) {
1176 	    if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1177 		    (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
1178 		init_mfc_params(rt, mfccp);
1179 		if (rt->mfc_expire)
1180 		    nexpire[hash]--;
1181 		rt->mfc_expire = 0;
1182 		break; /* XXX */
1183 	    }
1184 	}
1185 	if (rt == NULL) {		/* no upcall, so make a new entry */
1186 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1187 	    if (rt == NULL) {
1188 		MFC_UNLOCK();
1189 		VIF_UNLOCK();
1190 		return ENOBUFS;
1191 	    }
1192 
1193 	    init_mfc_params(rt, mfccp);
1194 	    rt->mfc_expire     = 0;
1195 	    rt->mfc_stall      = NULL;
1196 
1197 	    rt->mfc_bw_meter = NULL;
1198 	    /* insert new entry at head of hash chain */
1199 	    rt->mfc_next = mfctable[hash];
1200 	    mfctable[hash] = rt;
1201 	}
1202     }
1203     MFC_UNLOCK();
1204     VIF_UNLOCK();
1205     return 0;
1206 }
1207 
1208 /*
1209  * Delete an mfc entry
1210  */
1211 static int
1212 del_mfc(struct mfcctl2 *mfccp)
1213 {
1214     struct in_addr	origin;
1215     struct in_addr	mcastgrp;
1216     struct mfc		*rt;
1217     struct mfc		**nptr;
1218     u_long		hash;
1219     struct bw_meter	*list;
1220 
1221     origin = mfccp->mfcc_origin;
1222     mcastgrp = mfccp->mfcc_mcastgrp;
1223 
1224     if (mrtdebug & DEBUG_MFC)
1225 	log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n",
1226 	    (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1227 
1228     MFC_LOCK();
1229 
1230     hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1231     for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next)
1232 	if (origin.s_addr == rt->mfc_origin.s_addr &&
1233 		mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1234 		rt->mfc_stall == NULL)
1235 	    break;
1236     if (rt == NULL) {
1237 	MFC_UNLOCK();
1238 	return EADDRNOTAVAIL;
1239     }
1240 
1241     *nptr = rt->mfc_next;
1242 
1243     /*
1244      * free the bw_meter entries
1245      */
1246     list = rt->mfc_bw_meter;
1247     rt->mfc_bw_meter = NULL;
1248 
1249     free(rt, M_MRTABLE);
1250 
1251     free_bw_list(list);
1252 
1253     MFC_UNLOCK();
1254 
1255     return 0;
1256 }
1257 
1258 /*
1259  * Send a message to the routing daemon on the multicast routing socket
1260  */
1261 static int
1262 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1263 {
1264     if (s) {
1265 	SOCKBUF_LOCK(&s->so_rcv);
1266 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1267 	    NULL) != 0) {
1268 	    sorwakeup_locked(s);
1269 	    return 0;
1270 	}
1271 	SOCKBUF_UNLOCK(&s->so_rcv);
1272     }
1273     m_freem(mm);
1274     return -1;
1275 }
1276 
1277 /*
1278  * IP multicast forwarding function. This function assumes that the packet
1279  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1280  * pointed to by "ifp", and the packet is to be relayed to other networks
1281  * that have members of the packet's destination IP multicast group.
1282  *
1283  * The packet is returned unscathed to the caller, unless it is
1284  * erroneous, in which case a non-zero return value tells the caller to
1285  * discard it.
1286  */
1287 
1288 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1289 
1290 static int
1291 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1292     struct ip_moptions *imo)
1293 {
1294     INIT_VNET_INET(curvnet);
1295     struct mfc *rt;
1296     int error;
1297     vifi_t vifi;
1298 
1299     if (mrtdebug & DEBUG_FORWARD)
1300 	log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n",
1301 	    (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr),
1302 	    (void *)ifp);
1303 
1304     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1305 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1306 	/*
1307 	 * Packet arrived via a physical interface or
1308 	 * an encapsulated tunnel or a register_vif.
1309 	 */
1310     } else {
1311 	/*
1312 	 * Packet arrived through a source-route tunnel.
1313 	 * Source-route tunnels are no longer supported.
1314 	 */
1315 	static int last_log;
1316 	if (last_log != time_uptime) {
1317 	    last_log = time_uptime;
1318 	    log(LOG_ERR,
1319 		"ip_mforward: received source-routed packet from %lx\n",
1320 		(u_long)ntohl(ip->ip_src.s_addr));
1321 	}
1322 	return 1;
1323     }
1324 
1325     VIF_LOCK();
1326     MFC_LOCK();
1327     if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1328 	if (ip->ip_ttl < MAXTTL)
1329 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1330 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1331 	    struct vif *vifp = viftable + vifi;
1332 
1333 	    printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n",
1334 		(long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr),
1335 		vifi,
1336 		(vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1337 		vifp->v_ifp->if_xname);
1338 	}
1339 	error = ip_mdq(m, ifp, NULL, vifi);
1340 	MFC_UNLOCK();
1341 	VIF_UNLOCK();
1342 	return error;
1343     }
1344     if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1345 	printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1346 	    (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr));
1347 	if (!imo)
1348 	    printf("In fact, no options were specified at all\n");
1349     }
1350 
1351     /*
1352      * Don't forward a packet with time-to-live of zero or one,
1353      * or a packet destined to a local-only group.
1354      */
1355     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1356 	MFC_UNLOCK();
1357 	VIF_UNLOCK();
1358 	return 0;
1359     }
1360 
1361     /*
1362      * Determine forwarding vifs from the forwarding cache table
1363      */
1364     ++mrtstat.mrts_mfc_lookups;
1365     rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1366 
1367     /* Entry exists, so forward if necessary */
1368     if (rt != NULL) {
1369 	error = ip_mdq(m, ifp, rt, -1);
1370 	MFC_UNLOCK();
1371 	VIF_UNLOCK();
1372 	return error;
1373     } else {
1374 	/*
1375 	 * If we don't have a route for packet's origin,
1376 	 * Make a copy of the packet & send message to routing daemon
1377 	 */
1378 
1379 	struct mbuf *mb0;
1380 	struct rtdetq *rte;
1381 	u_long hash;
1382 	int hlen = ip->ip_hl << 2;
1383 
1384 	++mrtstat.mrts_mfc_misses;
1385 
1386 	mrtstat.mrts_no_route++;
1387 	if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1388 	    log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n",
1389 		(u_long)ntohl(ip->ip_src.s_addr),
1390 		(u_long)ntohl(ip->ip_dst.s_addr));
1391 
1392 	/*
1393 	 * Allocate mbufs early so that we don't do extra work if we are
1394 	 * just going to fail anyway.  Make sure to pullup the header so
1395 	 * that other people can't step on it.
1396 	 */
1397 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT);
1398 	if (rte == NULL) {
1399 	    MFC_UNLOCK();
1400 	    VIF_UNLOCK();
1401 	    return ENOBUFS;
1402 	}
1403 	mb0 = m_copypacket(m, M_DONTWAIT);
1404 	if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1405 	    mb0 = m_pullup(mb0, hlen);
1406 	if (mb0 == NULL) {
1407 	    free(rte, M_MRTABLE);
1408 	    MFC_UNLOCK();
1409 	    VIF_UNLOCK();
1410 	    return ENOBUFS;
1411 	}
1412 
1413 	/* is there an upcall waiting for this flow ? */
1414 	hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1415 	for (rt = mfctable[hash]; rt; rt = rt->mfc_next) {
1416 	    if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1417 		    (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1418 		    (rt->mfc_stall != NULL))
1419 		break;
1420 	}
1421 
1422 	if (rt == NULL) {
1423 	    int i;
1424 	    struct igmpmsg *im;
1425 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1426 	    struct mbuf *mm;
1427 
1428 	    /*
1429 	     * Locate the vifi for the incoming interface for this packet.
1430 	     * If none found, drop packet.
1431 	     */
1432 	    for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1433 		;
1434 	    if (vifi >= numvifs)	/* vif not found, drop packet */
1435 		goto non_fatal;
1436 
1437 	    /* no upcall, so make a new entry */
1438 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1439 	    if (rt == NULL)
1440 		goto fail;
1441 	    /* Make a copy of the header to send to the user level process */
1442 	    mm = m_copy(mb0, 0, hlen);
1443 	    if (mm == NULL)
1444 		goto fail1;
1445 
1446 	    /*
1447 	     * Send message to routing daemon to install
1448 	     * a route into the kernel table
1449 	     */
1450 
1451 	    im = mtod(mm, struct igmpmsg *);
1452 	    im->im_msgtype = IGMPMSG_NOCACHE;
1453 	    im->im_mbz = 0;
1454 	    im->im_vif = vifi;
1455 
1456 	    mrtstat.mrts_upcalls++;
1457 
1458 	    k_igmpsrc.sin_addr = ip->ip_src;
1459 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1460 		log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1461 		++mrtstat.mrts_upq_sockfull;
1462 fail1:
1463 		free(rt, M_MRTABLE);
1464 fail:
1465 		free(rte, M_MRTABLE);
1466 		m_freem(mb0);
1467 		MFC_UNLOCK();
1468 		VIF_UNLOCK();
1469 		return ENOBUFS;
1470 	    }
1471 
1472 	    /* insert new entry at head of hash chain */
1473 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1474 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1475 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1476 	    nexpire[hash]++;
1477 	    for (i = 0; i < numvifs; i++) {
1478 		rt->mfc_ttls[i] = 0;
1479 		rt->mfc_flags[i] = 0;
1480 	    }
1481 	    rt->mfc_parent = -1;
1482 
1483 	    rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */
1484 
1485 	    rt->mfc_bw_meter = NULL;
1486 
1487 	    /* link into table */
1488 	    rt->mfc_next   = mfctable[hash];
1489 	    mfctable[hash] = rt;
1490 	    rt->mfc_stall = rte;
1491 
1492 	} else {
1493 	    /* determine if q has overflowed */
1494 	    int npkts = 0;
1495 	    struct rtdetq **p;
1496 
1497 	    /*
1498 	     * XXX ouch! we need to append to the list, but we
1499 	     * only have a pointer to the front, so we have to
1500 	     * scan the entire list every time.
1501 	     */
1502 	    for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1503 		npkts++;
1504 
1505 	    if (npkts > MAX_UPQ) {
1506 		mrtstat.mrts_upq_ovflw++;
1507 non_fatal:
1508 		free(rte, M_MRTABLE);
1509 		m_freem(mb0);
1510 		MFC_UNLOCK();
1511 		VIF_UNLOCK();
1512 		return 0;
1513 	    }
1514 
1515 	    /* Add this entry to the end of the queue */
1516 	    *p = rte;
1517 	}
1518 
1519 	rte->m			= mb0;
1520 	rte->ifp		= ifp;
1521 	rte->next		= NULL;
1522 
1523 	MFC_UNLOCK();
1524 	VIF_UNLOCK();
1525 
1526 	return 0;
1527     }
1528 }
1529 
1530 /*
1531  * Clean up the cache entry if upcall is not serviced
1532  */
1533 static void
1534 expire_upcalls(void *unused)
1535 {
1536     struct rtdetq *rte;
1537     struct mfc *mfc, **nptr;
1538     int i;
1539 
1540     MFC_LOCK();
1541     for (i = 0; i < MFCTBLSIZ; i++) {
1542 	if (nexpire[i] == 0)
1543 	    continue;
1544 	nptr = &mfctable[i];
1545 	for (mfc = *nptr; mfc != NULL; mfc = *nptr) {
1546 	    /*
1547 	     * Skip real cache entries
1548 	     * Make sure it wasn't marked to not expire (shouldn't happen)
1549 	     * If it expires now
1550 	     */
1551 	    if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 &&
1552 		    --mfc->mfc_expire == 0) {
1553 		if (mrtdebug & DEBUG_EXPIRE)
1554 		    log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n",
1555 			(u_long)ntohl(mfc->mfc_origin.s_addr),
1556 			(u_long)ntohl(mfc->mfc_mcastgrp.s_addr));
1557 		/*
1558 		 * drop all the packets
1559 		 * free the mbuf with the pkt, if, timing info
1560 		 */
1561 		for (rte = mfc->mfc_stall; rte; ) {
1562 		    struct rtdetq *n = rte->next;
1563 
1564 		    m_freem(rte->m);
1565 		    free(rte, M_MRTABLE);
1566 		    rte = n;
1567 		}
1568 		++mrtstat.mrts_cache_cleanups;
1569 		nexpire[i]--;
1570 
1571 		/*
1572 		 * free the bw_meter entries
1573 		 */
1574 		while (mfc->mfc_bw_meter != NULL) {
1575 		    struct bw_meter *x = mfc->mfc_bw_meter;
1576 
1577 		    mfc->mfc_bw_meter = x->bm_mfc_next;
1578 		    free(x, M_BWMETER);
1579 		}
1580 
1581 		*nptr = mfc->mfc_next;
1582 		free(mfc, M_MRTABLE);
1583 	    } else {
1584 		nptr = &mfc->mfc_next;
1585 	    }
1586 	}
1587     }
1588     MFC_UNLOCK();
1589 
1590     callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
1591 }
1592 
1593 /*
1594  * Packet forwarding routine once entry in the cache is made
1595  */
1596 static int
1597 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1598 {
1599     INIT_VNET_INET(curvnet);
1600     struct ip  *ip = mtod(m, struct ip *);
1601     vifi_t vifi;
1602     int plen = ip->ip_len;
1603 
1604     VIF_LOCK_ASSERT();
1605 
1606     /*
1607      * If xmt_vif is not -1, send on only the requested vif.
1608      *
1609      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1610      */
1611     if (xmt_vif < numvifs) {
1612 	if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1613 		pim_register_send(ip, viftable + xmt_vif, m, rt);
1614 	else
1615 		phyint_send(ip, viftable + xmt_vif, m);
1616 	return 1;
1617     }
1618 
1619     /*
1620      * Don't forward if it didn't arrive from the parent vif for its origin.
1621      */
1622     vifi = rt->mfc_parent;
1623     if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1624 	/* came in the wrong interface */
1625 	if (mrtdebug & DEBUG_FORWARD)
1626 	    log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1627 		(void *)ifp, vifi, (void *)viftable[vifi].v_ifp);
1628 	++mrtstat.mrts_wrong_if;
1629 	++rt->mfc_wrong_if;
1630 	/*
1631 	 * If we are doing PIM assert processing, send a message
1632 	 * to the routing daemon.
1633 	 *
1634 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1635 	 * can complete the SPT switch, regardless of the type
1636 	 * of the iif (broadcast media, GRE tunnel, etc).
1637 	 */
1638 	if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1639 	    struct timeval now;
1640 	    u_long delta;
1641 
1642 	    if (ifp == &multicast_register_if)
1643 		pimstat.pims_rcv_registers_wrongiif++;
1644 
1645 	    /* Get vifi for the incoming packet */
1646 	    for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1647 		;
1648 	    if (vifi >= numvifs)
1649 		return 0;	/* The iif is not found: ignore the packet. */
1650 
1651 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1652 		return 0;	/* WRONGVIF disabled: ignore the packet */
1653 
1654 	    GET_TIME(now);
1655 
1656 	    TV_DELTA(now, rt->mfc_last_assert, delta);
1657 
1658 	    if (delta > ASSERT_MSG_TIME) {
1659 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1660 		struct igmpmsg *im;
1661 		int hlen = ip->ip_hl << 2;
1662 		struct mbuf *mm = m_copy(m, 0, hlen);
1663 
1664 		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1665 		    mm = m_pullup(mm, hlen);
1666 		if (mm == NULL)
1667 		    return ENOBUFS;
1668 
1669 		rt->mfc_last_assert = now;
1670 
1671 		im = mtod(mm, struct igmpmsg *);
1672 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1673 		im->im_mbz		= 0;
1674 		im->im_vif		= vifi;
1675 
1676 		mrtstat.mrts_upcalls++;
1677 
1678 		k_igmpsrc.sin_addr = im->im_src;
1679 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1680 		    log(LOG_WARNING,
1681 			"ip_mforward: ip_mrouter socket queue full\n");
1682 		    ++mrtstat.mrts_upq_sockfull;
1683 		    return ENOBUFS;
1684 		}
1685 	    }
1686 	}
1687 	return 0;
1688     }
1689 
1690     /* If I sourced this packet, it counts as output, else it was input. */
1691     if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1692 	viftable[vifi].v_pkt_out++;
1693 	viftable[vifi].v_bytes_out += plen;
1694     } else {
1695 	viftable[vifi].v_pkt_in++;
1696 	viftable[vifi].v_bytes_in += plen;
1697     }
1698     rt->mfc_pkt_cnt++;
1699     rt->mfc_byte_cnt += plen;
1700 
1701     /*
1702      * For each vif, decide if a copy of the packet should be forwarded.
1703      * Forward if:
1704      *		- the ttl exceeds the vif's threshold
1705      *		- there are group members downstream on interface
1706      */
1707     for (vifi = 0; vifi < numvifs; vifi++)
1708 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1709 	    viftable[vifi].v_pkt_out++;
1710 	    viftable[vifi].v_bytes_out += plen;
1711 	    if (viftable[vifi].v_flags & VIFF_REGISTER)
1712 		pim_register_send(ip, viftable + vifi, m, rt);
1713 	    else
1714 		phyint_send(ip, viftable + vifi, m);
1715 	}
1716 
1717     /*
1718      * Perform upcall-related bw measuring.
1719      */
1720     if (rt->mfc_bw_meter != NULL) {
1721 	struct bw_meter *x;
1722 	struct timeval now;
1723 
1724 	GET_TIME(now);
1725 	MFC_LOCK_ASSERT();
1726 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1727 	    bw_meter_receive_packet(x, plen, &now);
1728     }
1729 
1730     return 0;
1731 }
1732 
1733 /*
1734  * check if a vif number is legal/ok. This is used by ip_output.
1735  */
1736 static int
1737 X_legal_vif_num(int vif)
1738 {
1739     /* XXX unlocked, matter? */
1740     return (vif >= 0 && vif < numvifs);
1741 }
1742 
1743 /*
1744  * Return the local address used by this vif
1745  */
1746 static u_long
1747 X_ip_mcast_src(int vifi)
1748 {
1749     /* XXX unlocked, matter? */
1750     if (vifi >= 0 && vifi < numvifs)
1751 	return viftable[vifi].v_lcl_addr.s_addr;
1752     else
1753 	return INADDR_ANY;
1754 }
1755 
1756 static void
1757 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1758 {
1759     struct mbuf *mb_copy;
1760     int hlen = ip->ip_hl << 2;
1761 
1762     VIF_LOCK_ASSERT();
1763 
1764     /*
1765      * Make a new reference to the packet; make sure that
1766      * the IP header is actually copied, not just referenced,
1767      * so that ip_output() only scribbles on the copy.
1768      */
1769     mb_copy = m_copypacket(m, M_DONTWAIT);
1770     if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1771 	mb_copy = m_pullup(mb_copy, hlen);
1772     if (mb_copy == NULL)
1773 	return;
1774 
1775     send_packet(vifp, mb_copy);
1776 }
1777 
1778 static void
1779 send_packet(struct vif *vifp, struct mbuf *m)
1780 {
1781 	struct ip_moptions imo;
1782 	struct in_multi *imm[2];
1783 	int error;
1784 
1785 	VIF_LOCK_ASSERT();
1786 
1787 	imo.imo_multicast_ifp  = vifp->v_ifp;
1788 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1789 	imo.imo_multicast_loop = 1;
1790 	imo.imo_multicast_vif  = -1;
1791 	imo.imo_num_memberships = 0;
1792 	imo.imo_max_memberships = 2;
1793 	imo.imo_membership  = &imm[0];
1794 
1795 	/*
1796 	 * Re-entrancy should not be a problem here, because
1797 	 * the packets that we send out and are looped back at us
1798 	 * should get rejected because they appear to come from
1799 	 * the loopback interface, thus preventing looping.
1800 	 */
1801 	error = ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, &imo, NULL);
1802 	if (mrtdebug & DEBUG_XMIT) {
1803 	    log(LOG_DEBUG, "phyint_send on vif %td err %d\n",
1804 		vifp - viftable, error);
1805 	}
1806 }
1807 
1808 static int
1809 X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt)
1810 {
1811     INIT_VNET_INET(curvnet);
1812     int error, vifi;
1813 
1814     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
1815 	return EOPNOTSUPP;
1816 
1817     error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
1818     if (error)
1819 	return error;
1820 
1821     VIF_LOCK();
1822 
1823     if (vifi < 0 || vifi >= numvifs) {	/* Error if vif is invalid */
1824 	VIF_UNLOCK();
1825 	return EADDRNOTAVAIL;
1826     }
1827 
1828     if (sopt->sopt_name == IP_RSVP_VIF_ON) {
1829 	/* Check if socket is available. */
1830 	if (viftable[vifi].v_rsvpd != NULL) {
1831 	    VIF_UNLOCK();
1832 	    return EADDRINUSE;
1833 	}
1834 
1835 	viftable[vifi].v_rsvpd = so;
1836 	/* This may seem silly, but we need to be sure we don't over-increment
1837 	 * the RSVP counter, in case something slips up.
1838 	 */
1839 	if (!viftable[vifi].v_rsvp_on) {
1840 	    viftable[vifi].v_rsvp_on = 1;
1841 	    V_rsvp_on++;
1842 	}
1843     } else { /* must be VIF_OFF */
1844 	/*
1845 	 * XXX as an additional consistency check, one could make sure
1846 	 * that viftable[vifi].v_rsvpd == so, otherwise passing so as
1847 	 * first parameter is pretty useless.
1848 	 */
1849 	viftable[vifi].v_rsvpd = NULL;
1850 	/*
1851 	 * This may seem silly, but we need to be sure we don't over-decrement
1852 	 * the RSVP counter, in case something slips up.
1853 	 */
1854 	if (viftable[vifi].v_rsvp_on) {
1855 	    viftable[vifi].v_rsvp_on = 0;
1856 	    V_rsvp_on--;
1857 	}
1858     }
1859     VIF_UNLOCK();
1860     return 0;
1861 }
1862 
1863 static void
1864 X_ip_rsvp_force_done(struct socket *so)
1865 {
1866     INIT_VNET_INET(curvnet);
1867     int vifi;
1868 
1869     /* Don't bother if it is not the right type of socket. */
1870     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
1871 	return;
1872 
1873     VIF_LOCK();
1874 
1875     /* The socket may be attached to more than one vif...this
1876      * is perfectly legal.
1877      */
1878     for (vifi = 0; vifi < numvifs; vifi++) {
1879 	if (viftable[vifi].v_rsvpd == so) {
1880 	    viftable[vifi].v_rsvpd = NULL;
1881 	    /* This may seem silly, but we need to be sure we don't
1882 	     * over-decrement the RSVP counter, in case something slips up.
1883 	     */
1884 	    if (viftable[vifi].v_rsvp_on) {
1885 		viftable[vifi].v_rsvp_on = 0;
1886 		V_rsvp_on--;
1887 	    }
1888 	}
1889     }
1890 
1891     VIF_UNLOCK();
1892 }
1893 
1894 static void
1895 X_rsvp_input(struct mbuf *m, int off)
1896 {
1897     INIT_VNET_INET(curvnet);
1898     int vifi;
1899     struct ip *ip = mtod(m, struct ip *);
1900     struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
1901     struct ifnet *ifp;
1902 
1903     if (rsvpdebug)
1904 	printf("rsvp_input: rsvp_on %d\n", V_rsvp_on);
1905 
1906     /* Can still get packets with rsvp_on = 0 if there is a local member
1907      * of the group to which the RSVP packet is addressed.  But in this
1908      * case we want to throw the packet away.
1909      */
1910     if (!V_rsvp_on) {
1911 	m_freem(m);
1912 	return;
1913     }
1914 
1915     if (rsvpdebug)
1916 	printf("rsvp_input: check vifs\n");
1917 
1918 #ifdef DIAGNOSTIC
1919     M_ASSERTPKTHDR(m);
1920 #endif
1921 
1922     ifp = m->m_pkthdr.rcvif;
1923 
1924     VIF_LOCK();
1925     /* Find which vif the packet arrived on. */
1926     for (vifi = 0; vifi < numvifs; vifi++)
1927 	if (viftable[vifi].v_ifp == ifp)
1928 	    break;
1929 
1930     if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) {
1931 	/*
1932 	 * Drop the lock here to avoid holding it across rip_input.
1933 	 * This could make rsvpdebug printfs wrong.  If you care,
1934 	 * record the state of stuff before dropping the lock.
1935 	 */
1936 	VIF_UNLOCK();
1937 	/*
1938 	 * If the old-style non-vif-associated socket is set,
1939 	 * then use it.  Otherwise, drop packet since there
1940 	 * is no specific socket for this vif.
1941 	 */
1942 	if (V_ip_rsvpd != NULL) {
1943 	    if (rsvpdebug)
1944 		printf("rsvp_input: Sending packet up old-style socket\n");
1945 	    rip_input(m, off);  /* xxx */
1946 	} else {
1947 	    if (rsvpdebug && vifi == numvifs)
1948 		printf("rsvp_input: Can't find vif for packet.\n");
1949 	    else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL)
1950 		printf("rsvp_input: No socket defined for vif %d\n",vifi);
1951 	    m_freem(m);
1952 	}
1953 	return;
1954     }
1955     rsvp_src.sin_addr = ip->ip_src;
1956 
1957     if (rsvpdebug && m)
1958 	printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n",
1959 	       m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
1960 
1961     if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) {
1962 	if (rsvpdebug)
1963 	    printf("rsvp_input: Failed to append to socket\n");
1964     } else {
1965 	if (rsvpdebug)
1966 	    printf("rsvp_input: send packet up\n");
1967     }
1968     VIF_UNLOCK();
1969 }
1970 
1971 /*
1972  * Code for bandwidth monitors
1973  */
1974 
1975 /*
1976  * Define common interface for timeval-related methods
1977  */
1978 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1979 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1980 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1981 
1982 static uint32_t
1983 compute_bw_meter_flags(struct bw_upcall *req)
1984 {
1985     uint32_t flags = 0;
1986 
1987     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1988 	flags |= BW_METER_UNIT_PACKETS;
1989     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1990 	flags |= BW_METER_UNIT_BYTES;
1991     if (req->bu_flags & BW_UPCALL_GEQ)
1992 	flags |= BW_METER_GEQ;
1993     if (req->bu_flags & BW_UPCALL_LEQ)
1994 	flags |= BW_METER_LEQ;
1995 
1996     return flags;
1997 }
1998 
1999 /*
2000  * Add a bw_meter entry
2001  */
2002 static int
2003 add_bw_upcall(struct bw_upcall *req)
2004 {
2005     struct mfc *mfc;
2006     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2007 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2008     struct timeval now;
2009     struct bw_meter *x;
2010     uint32_t flags;
2011 
2012     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2013 	return EOPNOTSUPP;
2014 
2015     /* Test if the flags are valid */
2016     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2017 	return EINVAL;
2018     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2019 	return EINVAL;
2020     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2021 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2022 	return EINVAL;
2023 
2024     /* Test if the threshold time interval is valid */
2025     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2026 	return EINVAL;
2027 
2028     flags = compute_bw_meter_flags(req);
2029 
2030     /*
2031      * Find if we have already same bw_meter entry
2032      */
2033     MFC_LOCK();
2034     mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
2035     if (mfc == NULL) {
2036 	MFC_UNLOCK();
2037 	return EADDRNOTAVAIL;
2038     }
2039     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2040 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2041 			   &req->bu_threshold.b_time, ==)) &&
2042 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2043 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2044 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
2045 	    MFC_UNLOCK();
2046 	    return 0;		/* XXX Already installed */
2047 	}
2048     }
2049 
2050     /* Allocate the new bw_meter entry */
2051     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2052     if (x == NULL) {
2053 	MFC_UNLOCK();
2054 	return ENOBUFS;
2055     }
2056 
2057     /* Set the new bw_meter entry */
2058     x->bm_threshold.b_time = req->bu_threshold.b_time;
2059     GET_TIME(now);
2060     x->bm_start_time = now;
2061     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2062     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2063     x->bm_measured.b_packets = 0;
2064     x->bm_measured.b_bytes = 0;
2065     x->bm_flags = flags;
2066     x->bm_time_next = NULL;
2067     x->bm_time_hash = BW_METER_BUCKETS;
2068 
2069     /* Add the new bw_meter entry to the front of entries for this MFC */
2070     x->bm_mfc = mfc;
2071     x->bm_mfc_next = mfc->mfc_bw_meter;
2072     mfc->mfc_bw_meter = x;
2073     schedule_bw_meter(x, &now);
2074     MFC_UNLOCK();
2075 
2076     return 0;
2077 }
2078 
2079 static void
2080 free_bw_list(struct bw_meter *list)
2081 {
2082     while (list != NULL) {
2083 	struct bw_meter *x = list;
2084 
2085 	list = list->bm_mfc_next;
2086 	unschedule_bw_meter(x);
2087 	free(x, M_BWMETER);
2088     }
2089 }
2090 
2091 /*
2092  * Delete one or multiple bw_meter entries
2093  */
2094 static int
2095 del_bw_upcall(struct bw_upcall *req)
2096 {
2097     struct mfc *mfc;
2098     struct bw_meter *x;
2099 
2100     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2101 	return EOPNOTSUPP;
2102 
2103     MFC_LOCK();
2104     /* Find the corresponding MFC entry */
2105     mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
2106     if (mfc == NULL) {
2107 	MFC_UNLOCK();
2108 	return EADDRNOTAVAIL;
2109     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2110 	/*
2111 	 * Delete all bw_meter entries for this mfc
2112 	 */
2113 	struct bw_meter *list;
2114 
2115 	list = mfc->mfc_bw_meter;
2116 	mfc->mfc_bw_meter = NULL;
2117 	free_bw_list(list);
2118 	MFC_UNLOCK();
2119 	return 0;
2120     } else {			/* Delete a single bw_meter entry */
2121 	struct bw_meter *prev;
2122 	uint32_t flags = 0;
2123 
2124 	flags = compute_bw_meter_flags(req);
2125 
2126 	/* Find the bw_meter entry to delete */
2127 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2128 	     prev = x, x = x->bm_mfc_next) {
2129 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2130 			       &req->bu_threshold.b_time, ==)) &&
2131 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2132 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2133 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2134 		break;
2135 	}
2136 	if (x != NULL) { /* Delete entry from the list for this MFC */
2137 	    if (prev != NULL)
2138 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2139 	    else
2140 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2141 
2142 	    unschedule_bw_meter(x);
2143 	    MFC_UNLOCK();
2144 	    /* Free the bw_meter entry */
2145 	    free(x, M_BWMETER);
2146 	    return 0;
2147 	} else {
2148 	    MFC_UNLOCK();
2149 	    return EINVAL;
2150 	}
2151     }
2152     /* NOTREACHED */
2153 }
2154 
2155 /*
2156  * Perform bandwidth measurement processing that may result in an upcall
2157  */
2158 static void
2159 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2160 {
2161     struct timeval delta;
2162 
2163     MFC_LOCK_ASSERT();
2164 
2165     delta = *nowp;
2166     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2167 
2168     if (x->bm_flags & BW_METER_GEQ) {
2169 	/*
2170 	 * Processing for ">=" type of bw_meter entry
2171 	 */
2172 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2173 	    /* Reset the bw_meter entry */
2174 	    x->bm_start_time = *nowp;
2175 	    x->bm_measured.b_packets = 0;
2176 	    x->bm_measured.b_bytes = 0;
2177 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2178 	}
2179 
2180 	/* Record that a packet is received */
2181 	x->bm_measured.b_packets++;
2182 	x->bm_measured.b_bytes += plen;
2183 
2184 	/*
2185 	 * Test if we should deliver an upcall
2186 	 */
2187 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2188 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2189 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2190 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2191 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2192 		/* Prepare an upcall for delivery */
2193 		bw_meter_prepare_upcall(x, nowp);
2194 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2195 	    }
2196 	}
2197     } else if (x->bm_flags & BW_METER_LEQ) {
2198 	/*
2199 	 * Processing for "<=" type of bw_meter entry
2200 	 */
2201 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2202 	    /*
2203 	     * We are behind time with the multicast forwarding table
2204 	     * scanning for "<=" type of bw_meter entries, so test now
2205 	     * if we should deliver an upcall.
2206 	     */
2207 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2208 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2209 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2210 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2211 		/* Prepare an upcall for delivery */
2212 		bw_meter_prepare_upcall(x, nowp);
2213 	    }
2214 	    /* Reschedule the bw_meter entry */
2215 	    unschedule_bw_meter(x);
2216 	    schedule_bw_meter(x, nowp);
2217 	}
2218 
2219 	/* Record that a packet is received */
2220 	x->bm_measured.b_packets++;
2221 	x->bm_measured.b_bytes += plen;
2222 
2223 	/*
2224 	 * Test if we should restart the measuring interval
2225 	 */
2226 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2227 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2228 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
2229 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2230 	    /* Don't restart the measuring interval */
2231 	} else {
2232 	    /* Do restart the measuring interval */
2233 	    /*
2234 	     * XXX: note that we don't unschedule and schedule, because this
2235 	     * might be too much overhead per packet. Instead, when we process
2236 	     * all entries for a given timer hash bin, we check whether it is
2237 	     * really a timeout. If not, we reschedule at that time.
2238 	     */
2239 	    x->bm_start_time = *nowp;
2240 	    x->bm_measured.b_packets = 0;
2241 	    x->bm_measured.b_bytes = 0;
2242 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2243 	}
2244     }
2245 }
2246 
2247 /*
2248  * Prepare a bandwidth-related upcall
2249  */
2250 static void
2251 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2252 {
2253     struct timeval delta;
2254     struct bw_upcall *u;
2255 
2256     MFC_LOCK_ASSERT();
2257 
2258     /*
2259      * Compute the measured time interval
2260      */
2261     delta = *nowp;
2262     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2263 
2264     /*
2265      * If there are too many pending upcalls, deliver them now
2266      */
2267     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2268 	bw_upcalls_send();
2269 
2270     /*
2271      * Set the bw_upcall entry
2272      */
2273     u = &bw_upcalls[bw_upcalls_n++];
2274     u->bu_src = x->bm_mfc->mfc_origin;
2275     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2276     u->bu_threshold.b_time = x->bm_threshold.b_time;
2277     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2278     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2279     u->bu_measured.b_time = delta;
2280     u->bu_measured.b_packets = x->bm_measured.b_packets;
2281     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2282     u->bu_flags = 0;
2283     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2284 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2285     if (x->bm_flags & BW_METER_UNIT_BYTES)
2286 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2287     if (x->bm_flags & BW_METER_GEQ)
2288 	u->bu_flags |= BW_UPCALL_GEQ;
2289     if (x->bm_flags & BW_METER_LEQ)
2290 	u->bu_flags |= BW_UPCALL_LEQ;
2291 }
2292 
2293 /*
2294  * Send the pending bandwidth-related upcalls
2295  */
2296 static void
2297 bw_upcalls_send(void)
2298 {
2299     INIT_VNET_INET(curvnet);
2300     struct mbuf *m;
2301     int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2302     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2303     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2304 				      0,		/* unused2 */
2305 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2306 				      0,		/* im_mbz  */
2307 				      0,		/* im_vif  */
2308 				      0,		/* unused3 */
2309 				      { 0 },		/* im_src  */
2310 				      { 0 } };		/* im_dst  */
2311 
2312     MFC_LOCK_ASSERT();
2313 
2314     if (bw_upcalls_n == 0)
2315 	return;			/* No pending upcalls */
2316 
2317     bw_upcalls_n = 0;
2318 
2319     /*
2320      * Allocate a new mbuf, initialize it with the header and
2321      * the payload for the pending calls.
2322      */
2323     MGETHDR(m, M_DONTWAIT, MT_DATA);
2324     if (m == NULL) {
2325 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2326 	return;
2327     }
2328 
2329     m->m_len = m->m_pkthdr.len = 0;
2330     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2331     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2332 
2333     /*
2334      * Send the upcalls
2335      * XXX do we need to set the address in k_igmpsrc ?
2336      */
2337     mrtstat.mrts_upcalls++;
2338     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2339 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2340 	++mrtstat.mrts_upq_sockfull;
2341     }
2342 }
2343 
2344 /*
2345  * Compute the timeout hash value for the bw_meter entries
2346  */
2347 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2348     do {								\
2349 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2350 									\
2351 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2352 	(hash) = next_timeval.tv_sec;					\
2353 	if (next_timeval.tv_usec)					\
2354 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2355 	(hash) %= BW_METER_BUCKETS;					\
2356     } while (0)
2357 
2358 /*
2359  * Schedule a timer to process periodically bw_meter entry of type "<="
2360  * by linking the entry in the proper hash bucket.
2361  */
2362 static void
2363 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2364 {
2365     int time_hash;
2366 
2367     MFC_LOCK_ASSERT();
2368 
2369     if (!(x->bm_flags & BW_METER_LEQ))
2370 	return;		/* XXX: we schedule timers only for "<=" entries */
2371 
2372     /*
2373      * Reset the bw_meter entry
2374      */
2375     x->bm_start_time = *nowp;
2376     x->bm_measured.b_packets = 0;
2377     x->bm_measured.b_bytes = 0;
2378     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2379 
2380     /*
2381      * Compute the timeout hash value and insert the entry
2382      */
2383     BW_METER_TIMEHASH(x, time_hash);
2384     x->bm_time_next = bw_meter_timers[time_hash];
2385     bw_meter_timers[time_hash] = x;
2386     x->bm_time_hash = time_hash;
2387 }
2388 
2389 /*
2390  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2391  * by removing the entry from the proper hash bucket.
2392  */
2393 static void
2394 unschedule_bw_meter(struct bw_meter *x)
2395 {
2396     int time_hash;
2397     struct bw_meter *prev, *tmp;
2398 
2399     MFC_LOCK_ASSERT();
2400 
2401     if (!(x->bm_flags & BW_METER_LEQ))
2402 	return;		/* XXX: we schedule timers only for "<=" entries */
2403 
2404     /*
2405      * Compute the timeout hash value and delete the entry
2406      */
2407     time_hash = x->bm_time_hash;
2408     if (time_hash >= BW_METER_BUCKETS)
2409 	return;		/* Entry was not scheduled */
2410 
2411     for (prev = NULL, tmp = bw_meter_timers[time_hash];
2412 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2413 	if (tmp == x)
2414 	    break;
2415 
2416     if (tmp == NULL)
2417 	panic("unschedule_bw_meter: bw_meter entry not found");
2418 
2419     if (prev != NULL)
2420 	prev->bm_time_next = x->bm_time_next;
2421     else
2422 	bw_meter_timers[time_hash] = x->bm_time_next;
2423 
2424     x->bm_time_next = NULL;
2425     x->bm_time_hash = BW_METER_BUCKETS;
2426 }
2427 
2428 
2429 /*
2430  * Process all "<=" type of bw_meter that should be processed now,
2431  * and for each entry prepare an upcall if necessary. Each processed
2432  * entry is rescheduled again for the (periodic) processing.
2433  *
2434  * This is run periodically (once per second normally). On each round,
2435  * all the potentially matching entries are in the hash slot that we are
2436  * looking at.
2437  */
2438 static void
2439 bw_meter_process()
2440 {
2441     static uint32_t last_tv_sec;	/* last time we processed this */
2442 
2443     uint32_t loops;
2444     int i;
2445     struct timeval now, process_endtime;
2446 
2447     GET_TIME(now);
2448     if (last_tv_sec == now.tv_sec)
2449 	return;		/* nothing to do */
2450 
2451     loops = now.tv_sec - last_tv_sec;
2452     last_tv_sec = now.tv_sec;
2453     if (loops > BW_METER_BUCKETS)
2454 	loops = BW_METER_BUCKETS;
2455 
2456     MFC_LOCK();
2457     /*
2458      * Process all bins of bw_meter entries from the one after the last
2459      * processed to the current one. On entry, i points to the last bucket
2460      * visited, so we need to increment i at the beginning of the loop.
2461      */
2462     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2463 	struct bw_meter *x, *tmp_list;
2464 
2465 	if (++i >= BW_METER_BUCKETS)
2466 	    i = 0;
2467 
2468 	/* Disconnect the list of bw_meter entries from the bin */
2469 	tmp_list = bw_meter_timers[i];
2470 	bw_meter_timers[i] = NULL;
2471 
2472 	/* Process the list of bw_meter entries */
2473 	while (tmp_list != NULL) {
2474 	    x = tmp_list;
2475 	    tmp_list = tmp_list->bm_time_next;
2476 
2477 	    /* Test if the time interval is over */
2478 	    process_endtime = x->bm_start_time;
2479 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2480 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2481 		/* Not yet: reschedule, but don't reset */
2482 		int time_hash;
2483 
2484 		BW_METER_TIMEHASH(x, time_hash);
2485 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2486 		    /*
2487 		     * XXX: somehow the bin processing is a bit ahead of time.
2488 		     * Put the entry in the next bin.
2489 		     */
2490 		    if (++time_hash >= BW_METER_BUCKETS)
2491 			time_hash = 0;
2492 		}
2493 		x->bm_time_next = bw_meter_timers[time_hash];
2494 		bw_meter_timers[time_hash] = x;
2495 		x->bm_time_hash = time_hash;
2496 
2497 		continue;
2498 	    }
2499 
2500 	    /*
2501 	     * Test if we should deliver an upcall
2502 	     */
2503 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2504 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2505 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2506 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2507 		/* Prepare an upcall for delivery */
2508 		bw_meter_prepare_upcall(x, &now);
2509 	    }
2510 
2511 	    /*
2512 	     * Reschedule for next processing
2513 	     */
2514 	    schedule_bw_meter(x, &now);
2515 	}
2516     }
2517 
2518     /* Send all upcalls that are pending delivery */
2519     bw_upcalls_send();
2520 
2521     MFC_UNLOCK();
2522 }
2523 
2524 /*
2525  * A periodic function for sending all upcalls that are pending delivery
2526  */
2527 static void
2528 expire_bw_upcalls_send(void *unused)
2529 {
2530     MFC_LOCK();
2531     bw_upcalls_send();
2532     MFC_UNLOCK();
2533 
2534     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2535 	expire_bw_upcalls_send, NULL);
2536 }
2537 
2538 /*
2539  * A periodic function for periodic scanning of the multicast forwarding
2540  * table for processing all "<=" bw_meter entries.
2541  */
2542 static void
2543 expire_bw_meter_process(void *unused)
2544 {
2545     if (mrt_api_config & MRT_MFC_BW_UPCALL)
2546 	bw_meter_process();
2547 
2548     callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
2549 }
2550 
2551 /*
2552  * End of bandwidth monitoring code
2553  */
2554 
2555 /*
2556  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2557  *
2558  */
2559 static int
2560 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2561     struct mfc *rt)
2562 {
2563     struct mbuf *mb_copy, *mm;
2564 
2565     if (mrtdebug & DEBUG_PIM)
2566 	log(LOG_DEBUG, "pim_register_send: ");
2567 
2568     /*
2569      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2570      * rendezvous point was unspecified, and we were told not to.
2571      */
2572     if (pim_squelch_wholepkt != 0 && (mrt_api_config & MRT_MFC_RP) &&
2573 	(rt->mfc_rp.s_addr == INADDR_ANY))
2574 	return 0;
2575 
2576     mb_copy = pim_register_prepare(ip, m);
2577     if (mb_copy == NULL)
2578 	return ENOBUFS;
2579 
2580     /*
2581      * Send all the fragments. Note that the mbuf for each fragment
2582      * is freed by the sending machinery.
2583      */
2584     for (mm = mb_copy; mm; mm = mb_copy) {
2585 	mb_copy = mm->m_nextpkt;
2586 	mm->m_nextpkt = 0;
2587 	mm = m_pullup(mm, sizeof(struct ip));
2588 	if (mm != NULL) {
2589 	    ip = mtod(mm, struct ip *);
2590 	    if ((mrt_api_config & MRT_MFC_RP) &&
2591 		(rt->mfc_rp.s_addr != INADDR_ANY)) {
2592 		pim_register_send_rp(ip, vifp, mm, rt);
2593 	    } else {
2594 		pim_register_send_upcall(ip, vifp, mm, rt);
2595 	    }
2596 	}
2597     }
2598 
2599     return 0;
2600 }
2601 
2602 /*
2603  * Return a copy of the data packet that is ready for PIM Register
2604  * encapsulation.
2605  * XXX: Note that in the returned copy the IP header is a valid one.
2606  */
2607 static struct mbuf *
2608 pim_register_prepare(struct ip *ip, struct mbuf *m)
2609 {
2610     struct mbuf *mb_copy = NULL;
2611     int mtu;
2612 
2613     /* Take care of delayed checksums */
2614     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2615 	in_delayed_cksum(m);
2616 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2617     }
2618 
2619     /*
2620      * Copy the old packet & pullup its IP header into the
2621      * new mbuf so we can modify it.
2622      */
2623     mb_copy = m_copypacket(m, M_DONTWAIT);
2624     if (mb_copy == NULL)
2625 	return NULL;
2626     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2627     if (mb_copy == NULL)
2628 	return NULL;
2629 
2630     /* take care of the TTL */
2631     ip = mtod(mb_copy, struct ip *);
2632     --ip->ip_ttl;
2633 
2634     /* Compute the MTU after the PIM Register encapsulation */
2635     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2636 
2637     if (ip->ip_len <= mtu) {
2638 	/* Turn the IP header into a valid one */
2639 	ip->ip_len = htons(ip->ip_len);
2640 	ip->ip_off = htons(ip->ip_off);
2641 	ip->ip_sum = 0;
2642 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2643     } else {
2644 	/* Fragment the packet */
2645 	if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
2646 	    m_freem(mb_copy);
2647 	    return NULL;
2648 	}
2649     }
2650     return mb_copy;
2651 }
2652 
2653 /*
2654  * Send an upcall with the data packet to the user-level process.
2655  */
2656 static int
2657 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2658     struct mbuf *mb_copy, struct mfc *rt)
2659 {
2660     INIT_VNET_INET(curvnet);
2661     struct mbuf *mb_first;
2662     int len = ntohs(ip->ip_len);
2663     struct igmpmsg *im;
2664     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2665 
2666     VIF_LOCK_ASSERT();
2667 
2668     /*
2669      * Add a new mbuf with an upcall header
2670      */
2671     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2672     if (mb_first == NULL) {
2673 	m_freem(mb_copy);
2674 	return ENOBUFS;
2675     }
2676     mb_first->m_data += max_linkhdr;
2677     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2678     mb_first->m_len = sizeof(struct igmpmsg);
2679     mb_first->m_next = mb_copy;
2680 
2681     /* Send message to routing daemon */
2682     im = mtod(mb_first, struct igmpmsg *);
2683     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2684     im->im_mbz		= 0;
2685     im->im_vif		= vifp - viftable;
2686     im->im_src		= ip->ip_src;
2687     im->im_dst		= ip->ip_dst;
2688 
2689     k_igmpsrc.sin_addr	= ip->ip_src;
2690 
2691     mrtstat.mrts_upcalls++;
2692 
2693     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2694 	if (mrtdebug & DEBUG_PIM)
2695 	    log(LOG_WARNING,
2696 		"mcast: pim_register_send_upcall: ip_mrouter socket queue full");
2697 	++mrtstat.mrts_upq_sockfull;
2698 	return ENOBUFS;
2699     }
2700 
2701     /* Keep statistics */
2702     pimstat.pims_snd_registers_msgs++;
2703     pimstat.pims_snd_registers_bytes += len;
2704 
2705     return 0;
2706 }
2707 
2708 /*
2709  * Encapsulate the data packet in PIM Register message and send it to the RP.
2710  */
2711 static int
2712 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2713     struct mfc *rt)
2714 {
2715     INIT_VNET_INET(curvnet);
2716     struct mbuf *mb_first;
2717     struct ip *ip_outer;
2718     struct pim_encap_pimhdr *pimhdr;
2719     int len = ntohs(ip->ip_len);
2720     vifi_t vifi = rt->mfc_parent;
2721 
2722     VIF_LOCK_ASSERT();
2723 
2724     if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) {
2725 	m_freem(mb_copy);
2726 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2727     }
2728 
2729     /*
2730      * Add a new mbuf with the encapsulating header
2731      */
2732     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2733     if (mb_first == NULL) {
2734 	m_freem(mb_copy);
2735 	return ENOBUFS;
2736     }
2737     mb_first->m_data += max_linkhdr;
2738     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2739     mb_first->m_next = mb_copy;
2740 
2741     mb_first->m_pkthdr.len = len + mb_first->m_len;
2742 
2743     /*
2744      * Fill in the encapsulating IP and PIM header
2745      */
2746     ip_outer = mtod(mb_first, struct ip *);
2747     *ip_outer = pim_encap_iphdr;
2748     ip_outer->ip_id = ip_newid();
2749     ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2750     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
2751     ip_outer->ip_dst = rt->mfc_rp;
2752     /*
2753      * Copy the inner header TOS to the outer header, and take care of the
2754      * IP_DF bit.
2755      */
2756     ip_outer->ip_tos = ip->ip_tos;
2757     if (ntohs(ip->ip_off) & IP_DF)
2758 	ip_outer->ip_off |= IP_DF;
2759     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2760 					 + sizeof(pim_encap_iphdr));
2761     *pimhdr = pim_encap_pimhdr;
2762     /* If the iif crosses a border, set the Border-bit */
2763     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
2764 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2765 
2766     mb_first->m_data += sizeof(pim_encap_iphdr);
2767     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2768     mb_first->m_data -= sizeof(pim_encap_iphdr);
2769 
2770     send_packet(vifp, mb_first);
2771 
2772     /* Keep statistics */
2773     pimstat.pims_snd_registers_msgs++;
2774     pimstat.pims_snd_registers_bytes += len;
2775 
2776     return 0;
2777 }
2778 
2779 /*
2780  * pim_encapcheck() is called by the encap[46]_input() path at runtime to
2781  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2782  * into the kernel.
2783  */
2784 static int
2785 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2786 {
2787 
2788 #ifdef DIAGNOSTIC
2789     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2790 #endif
2791     if (proto != IPPROTO_PIM)
2792 	return 0;	/* not for us; reject the datagram. */
2793 
2794     return 64;		/* claim the datagram. */
2795 }
2796 
2797 /*
2798  * PIM-SMv2 and PIM-DM messages processing.
2799  * Receives and verifies the PIM control messages, and passes them
2800  * up to the listening socket, using rip_input().
2801  * The only message with special processing is the PIM_REGISTER message
2802  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2803  * is passed to if_simloop().
2804  */
2805 void
2806 pim_input(struct mbuf *m, int off)
2807 {
2808     struct ip *ip = mtod(m, struct ip *);
2809     struct pim *pim;
2810     int minlen;
2811     int datalen = ip->ip_len;
2812     int ip_tos;
2813     int iphlen = off;
2814 
2815     /* Keep statistics */
2816     pimstat.pims_rcv_total_msgs++;
2817     pimstat.pims_rcv_total_bytes += datalen;
2818 
2819     /*
2820      * Validate lengths
2821      */
2822     if (datalen < PIM_MINLEN) {
2823 	pimstat.pims_rcv_tooshort++;
2824 	log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
2825 	    datalen, (u_long)ip->ip_src.s_addr);
2826 	m_freem(m);
2827 	return;
2828     }
2829 
2830     /*
2831      * If the packet is at least as big as a REGISTER, go agead
2832      * and grab the PIM REGISTER header size, to avoid another
2833      * possible m_pullup() later.
2834      *
2835      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2836      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2837      */
2838     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2839     /*
2840      * Get the IP and PIM headers in contiguous memory, and
2841      * possibly the PIM REGISTER header.
2842      */
2843     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
2844 	(m = m_pullup(m, minlen)) == 0) {
2845 	log(LOG_ERR, "pim_input: m_pullup failure\n");
2846 	return;
2847     }
2848     /* m_pullup() may have given us a new mbuf so reset ip. */
2849     ip = mtod(m, struct ip *);
2850     ip_tos = ip->ip_tos;
2851 
2852     /* adjust mbuf to point to the PIM header */
2853     m->m_data += iphlen;
2854     m->m_len  -= iphlen;
2855     pim = mtod(m, struct pim *);
2856 
2857     /*
2858      * Validate checksum. If PIM REGISTER, exclude the data packet.
2859      *
2860      * XXX: some older PIMv2 implementations don't make this distinction,
2861      * so for compatibility reason perform the checksum over part of the
2862      * message, and if error, then over the whole message.
2863      */
2864     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2865 	/* do nothing, checksum okay */
2866     } else if (in_cksum(m, datalen)) {
2867 	pimstat.pims_rcv_badsum++;
2868 	if (mrtdebug & DEBUG_PIM)
2869 	    log(LOG_DEBUG, "pim_input: invalid checksum");
2870 	m_freem(m);
2871 	return;
2872     }
2873 
2874     /* PIM version check */
2875     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2876 	pimstat.pims_rcv_badversion++;
2877 	log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
2878 	    PIM_VT_V(pim->pim_vt), PIM_VERSION);
2879 	m_freem(m);
2880 	return;
2881     }
2882 
2883     /* restore mbuf back to the outer IP */
2884     m->m_data -= iphlen;
2885     m->m_len  += iphlen;
2886 
2887     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2888 	/*
2889 	 * Since this is a REGISTER, we'll make a copy of the register
2890 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2891 	 * routing daemon.
2892 	 */
2893 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2894 	struct mbuf *mcp;
2895 	struct ip *encap_ip;
2896 	u_int32_t *reghdr;
2897 	struct ifnet *vifp;
2898 
2899 	VIF_LOCK();
2900 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
2901 	    VIF_UNLOCK();
2902 	    if (mrtdebug & DEBUG_PIM)
2903 		log(LOG_DEBUG,
2904 		    "pim_input: register vif not set: %d\n", reg_vif_num);
2905 	    m_freem(m);
2906 	    return;
2907 	}
2908 	/* XXX need refcnt? */
2909 	vifp = viftable[reg_vif_num].v_ifp;
2910 	VIF_UNLOCK();
2911 
2912 	/*
2913 	 * Validate length
2914 	 */
2915 	if (datalen < PIM_REG_MINLEN) {
2916 	    pimstat.pims_rcv_tooshort++;
2917 	    pimstat.pims_rcv_badregisters++;
2918 	    log(LOG_ERR,
2919 		"pim_input: register packet size too small %d from %lx\n",
2920 		datalen, (u_long)ip->ip_src.s_addr);
2921 	    m_freem(m);
2922 	    return;
2923 	}
2924 
2925 	reghdr = (u_int32_t *)(pim + 1);
2926 	encap_ip = (struct ip *)(reghdr + 1);
2927 
2928 	if (mrtdebug & DEBUG_PIM) {
2929 	    log(LOG_DEBUG,
2930 		"pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
2931 		(u_long)ntohl(encap_ip->ip_src.s_addr),
2932 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
2933 		ntohs(encap_ip->ip_len));
2934 	}
2935 
2936 	/* verify the version number of the inner packet */
2937 	if (encap_ip->ip_v != IPVERSION) {
2938 	    pimstat.pims_rcv_badregisters++;
2939 	    if (mrtdebug & DEBUG_PIM) {
2940 		log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
2941 		    "of the inner packet\n", encap_ip->ip_v);
2942 	    }
2943 	    m_freem(m);
2944 	    return;
2945 	}
2946 
2947 	/* verify the inner packet is destined to a mcast group */
2948 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2949 	    pimstat.pims_rcv_badregisters++;
2950 	    if (mrtdebug & DEBUG_PIM)
2951 		log(LOG_DEBUG,
2952 		    "pim_input: inner packet of register is not "
2953 		    "multicast %lx\n",
2954 		    (u_long)ntohl(encap_ip->ip_dst.s_addr));
2955 	    m_freem(m);
2956 	    return;
2957 	}
2958 
2959 	/* If a NULL_REGISTER, pass it to the daemon */
2960 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2961 	    goto pim_input_to_daemon;
2962 
2963 	/*
2964 	 * Copy the TOS from the outer IP header to the inner IP header.
2965 	 */
2966 	if (encap_ip->ip_tos != ip_tos) {
2967 	    /* Outer TOS -> inner TOS */
2968 	    encap_ip->ip_tos = ip_tos;
2969 	    /* Recompute the inner header checksum. Sigh... */
2970 
2971 	    /* adjust mbuf to point to the inner IP header */
2972 	    m->m_data += (iphlen + PIM_MINLEN);
2973 	    m->m_len  -= (iphlen + PIM_MINLEN);
2974 
2975 	    encap_ip->ip_sum = 0;
2976 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2977 
2978 	    /* restore mbuf to point back to the outer IP header */
2979 	    m->m_data -= (iphlen + PIM_MINLEN);
2980 	    m->m_len  += (iphlen + PIM_MINLEN);
2981 	}
2982 
2983 	/*
2984 	 * Decapsulate the inner IP packet and loopback to forward it
2985 	 * as a normal multicast packet. Also, make a copy of the
2986 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2987 	 * to pass to the daemon later, so it can take the appropriate
2988 	 * actions (e.g., send back PIM_REGISTER_STOP).
2989 	 * XXX: here m->m_data points to the outer IP header.
2990 	 */
2991 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2992 	if (mcp == NULL) {
2993 	    log(LOG_ERR,
2994 		"pim_input: pim register: could not copy register head\n");
2995 	    m_freem(m);
2996 	    return;
2997 	}
2998 
2999 	/* Keep statistics */
3000 	/* XXX: registers_bytes include only the encap. mcast pkt */
3001 	pimstat.pims_rcv_registers_msgs++;
3002 	pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3003 
3004 	/*
3005 	 * forward the inner ip packet; point m_data at the inner ip.
3006 	 */
3007 	m_adj(m, iphlen + PIM_MINLEN);
3008 
3009 	if (mrtdebug & DEBUG_PIM) {
3010 	    log(LOG_DEBUG,
3011 		"pim_input: forwarding decapsulated register: "
3012 		"src %lx, dst %lx, vif %d\n",
3013 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3014 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3015 		reg_vif_num);
3016 	}
3017 	/* NB: vifp was collected above; can it change on us? */
3018 	if_simloop(vifp, m, dst.sin_family, 0);
3019 
3020 	/* prepare the register head to send to the mrouting daemon */
3021 	m = mcp;
3022     }
3023 
3024 pim_input_to_daemon:
3025     /*
3026      * Pass the PIM message up to the daemon; if it is a Register message,
3027      * pass the 'head' only up to the daemon. This includes the
3028      * outer IP header, PIM header, PIM-Register header and the
3029      * inner IP header.
3030      * XXX: the outer IP header pkt size of a Register is not adjust to
3031      * reflect the fact that the inner multicast data is truncated.
3032      */
3033     rip_input(m, iphlen);
3034 
3035     return;
3036 }
3037 
3038 /*
3039  * XXX: This is common code for dealing with initialization for both
3040  * the IPv4 and IPv6 multicast forwarding paths. It could do with cleanup.
3041  */
3042 static int
3043 ip_mroute_modevent(module_t mod, int type, void *unused)
3044 {
3045     INIT_VNET_INET(curvnet);
3046 
3047     switch (type) {
3048     case MOD_LOAD:
3049 	MROUTER_LOCK_INIT();
3050 	MFC_LOCK_INIT();
3051 	VIF_LOCK_INIT();
3052 	ip_mrouter_reset();
3053 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
3054 	    &pim_squelch_wholepkt);
3055 
3056 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
3057 	    pim_encapcheck, &in_pim_protosw, NULL);
3058 	if (pim_encap_cookie == NULL) {
3059 		printf("ip_mroute: unable to attach pim encap\n");
3060 		VIF_LOCK_DESTROY();
3061 		MFC_LOCK_DESTROY();
3062 		MROUTER_LOCK_DESTROY();
3063 		return (EINVAL);
3064 	}
3065 
3066 #ifdef INET6
3067 	pim6_encap_cookie = encap_attach_func(AF_INET6, IPPROTO_PIM,
3068 	    pim_encapcheck, (struct protosw *)&in6_pim_protosw, NULL);
3069 	if (pim6_encap_cookie == NULL) {
3070 		printf("ip_mroute: unable to attach pim6 encap\n");
3071 		if (pim_encap_cookie) {
3072 		    encap_detach(pim_encap_cookie);
3073 		    pim_encap_cookie = NULL;
3074 		}
3075 		VIF_LOCK_DESTROY();
3076 		MFC_LOCK_DESTROY();
3077 		MROUTER_LOCK_DESTROY();
3078 		return (EINVAL);
3079 	}
3080 #endif
3081 
3082 	ip_mcast_src = X_ip_mcast_src;
3083 	ip_mforward = X_ip_mforward;
3084 	ip_mrouter_done = X_ip_mrouter_done;
3085 	ip_mrouter_get = X_ip_mrouter_get;
3086 	ip_mrouter_set = X_ip_mrouter_set;
3087 
3088 #ifdef INET6
3089 	ip6_mforward = X_ip6_mforward;
3090 	ip6_mrouter_done = X_ip6_mrouter_done;
3091 	ip6_mrouter_get = X_ip6_mrouter_get;
3092 	ip6_mrouter_set = X_ip6_mrouter_set;
3093 	mrt6_ioctl = X_mrt6_ioctl;
3094 #endif
3095 
3096 	ip_rsvp_force_done = X_ip_rsvp_force_done;
3097 	ip_rsvp_vif = X_ip_rsvp_vif;
3098 
3099 	legal_vif_num = X_legal_vif_num;
3100 	mrt_ioctl = X_mrt_ioctl;
3101 	rsvp_input_p = X_rsvp_input;
3102 	break;
3103 
3104     case MOD_UNLOAD:
3105 	/*
3106 	 * Typically module unload happens after the user-level
3107 	 * process has shutdown the kernel services (the check
3108 	 * below insures someone can't just yank the module out
3109 	 * from under a running process).  But if the module is
3110 	 * just loaded and then unloaded w/o starting up a user
3111 	 * process we still need to cleanup.
3112 	 */
3113 	if (V_ip_mrouter
3114 #ifdef INET6
3115 	    || ip6_mrouter
3116 #endif
3117 	)
3118 	    return EINVAL;
3119 
3120 #ifdef INET6
3121 	if (pim6_encap_cookie) {
3122 	    encap_detach(pim6_encap_cookie);
3123 	    pim6_encap_cookie = NULL;
3124 	}
3125 	X_ip6_mrouter_done();
3126 	ip6_mforward = NULL;
3127 	ip6_mrouter_done = NULL;
3128 	ip6_mrouter_get = NULL;
3129 	ip6_mrouter_set = NULL;
3130 	mrt6_ioctl = NULL;
3131 #endif
3132 
3133 	if (pim_encap_cookie) {
3134 	    encap_detach(pim_encap_cookie);
3135 	    pim_encap_cookie = NULL;
3136 	}
3137 	X_ip_mrouter_done();
3138 	ip_mcast_src = NULL;
3139 	ip_mforward = NULL;
3140 	ip_mrouter_done = NULL;
3141 	ip_mrouter_get = NULL;
3142 	ip_mrouter_set = NULL;
3143 
3144 	ip_rsvp_force_done = NULL;
3145 	ip_rsvp_vif = NULL;
3146 
3147 	legal_vif_num = NULL;
3148 	mrt_ioctl = NULL;
3149 	rsvp_input_p = NULL;
3150 
3151 	VIF_LOCK_DESTROY();
3152 	MFC_LOCK_DESTROY();
3153 	MROUTER_LOCK_DESTROY();
3154 	break;
3155 
3156     default:
3157 	return EOPNOTSUPP;
3158     }
3159     return 0;
3160 }
3161 
3162 static moduledata_t ip_mroutemod = {
3163     "ip_mroute",
3164     ip_mroute_modevent,
3165     0
3166 };
3167 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
3168