xref: /freebsd/sys/netinet/ip_mroute.c (revision 38d120bc13ac1de5b739b67b87016b9122149374)
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 /*
57  * TODO: Prefix functions with ipmf_.
58  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
59  * domain attachment (if_afdata) so we can track consumers of that service.
60  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
61  * move it to socket options.
62  * TODO: Cleanup LSRR removal further.
63  * TODO: Push RSVP stubs into raw_ip.c.
64  * TODO: Use bitstring.h for vif set.
65  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
66  * TODO: Sync ip6_mroute.c with this file.
67  */
68 
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
71 
72 #include "opt_inet.h"
73 #include "opt_mrouting.h"
74 
75 #define _PIM_VT 1
76 
77 #include <sys/param.h>
78 #include <sys/kernel.h>
79 #include <sys/stddef.h>
80 #include <sys/lock.h>
81 #include <sys/ktr.h>
82 #include <sys/malloc.h>
83 #include <sys/mbuf.h>
84 #include <sys/module.h>
85 #include <sys/priv.h>
86 #include <sys/protosw.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/socketvar.h>
90 #include <sys/sockio.h>
91 #include <sys/sx.h>
92 #include <sys/sysctl.h>
93 #include <sys/syslog.h>
94 #include <sys/systm.h>
95 #include <sys/time.h>
96 #include <sys/counter.h>
97 
98 #include <net/if.h>
99 #include <net/if_var.h>
100 #include <net/netisr.h>
101 #include <net/route.h>
102 #include <net/vnet.h>
103 
104 #include <netinet/in.h>
105 #include <netinet/igmp.h>
106 #include <netinet/in_systm.h>
107 #include <netinet/in_var.h>
108 #include <netinet/ip.h>
109 #include <netinet/ip_encap.h>
110 #include <netinet/ip_mroute.h>
111 #include <netinet/ip_var.h>
112 #include <netinet/ip_options.h>
113 #include <netinet/pim.h>
114 #include <netinet/pim_var.h>
115 #include <netinet/udp.h>
116 
117 #include <machine/in_cksum.h>
118 
119 #ifndef KTR_IPMF
120 #define KTR_IPMF KTR_INET
121 #endif
122 
123 #define		VIFI_INVALID	((vifi_t) -1)
124 
125 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
126 #define	V_last_tv_sec	VNET(last_tv_sec)
127 
128 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
129 
130 /*
131  * Locking.  We use two locks: one for the virtual interface table and
132  * one for the forwarding table.  These locks may be nested in which case
133  * the VIF lock must always be taken first.  Note that each lock is used
134  * to cover not only the specific data structure but also related data
135  * structures.
136  */
137 
138 static struct mtx mrouter_mtx;
139 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
140 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
141 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
142 #define	MROUTER_LOCK_INIT()						\
143 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
144 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
145 
146 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
147 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
148 
149 static VNET_PCPUSTAT_DEFINE(struct mrtstat, mrtstat);
150 VNET_PCPUSTAT_SYSINIT(mrtstat);
151 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
152 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
153     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
154     "netinet/ip_mroute.h)");
155 
156 static VNET_DEFINE(u_long, mfchash);
157 #define	V_mfchash		VNET(mfchash)
158 #define	MFCHASH(a, g)							\
159 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
160 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
161 #define	MFCHASHSIZE	256
162 
163 static u_long mfchashsize;			/* Hash size */
164 static VNET_DEFINE(u_char *, nexpire);		/* 0..mfchashsize-1 */
165 #define	V_nexpire		VNET(nexpire)
166 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
167 #define	V_mfchashtbl		VNET(mfchashtbl)
168 
169 static struct mtx mfc_mtx;
170 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
171 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
172 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
173 #define	MFC_LOCK_INIT()							\
174 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
175 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
176 
177 static VNET_DEFINE(vifi_t, numvifs);
178 #define	V_numvifs		VNET(numvifs)
179 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
180 #define	V_viftable		VNET(viftable)
181 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_VNET | CTLFLAG_RD,
182     &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
183     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
184 
185 static struct mtx vif_mtx;
186 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
187 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
188 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
189 #define	VIF_LOCK_INIT()							\
190 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
191 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
192 
193 static eventhandler_tag if_detach_event_tag = NULL;
194 
195 static VNET_DEFINE(struct callout, expire_upcalls_ch);
196 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
197 
198 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
199 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
200 
201 /*
202  * Bandwidth meter variables and constants
203  */
204 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
205 /*
206  * Pending timeouts are stored in a hash table, the key being the
207  * expiration time. Periodically, the entries are analysed and processed.
208  */
209 #define	BW_METER_BUCKETS	1024
210 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
211 #define	V_bw_meter_timers	VNET(bw_meter_timers)
212 static VNET_DEFINE(struct callout, bw_meter_ch);
213 #define	V_bw_meter_ch		VNET(bw_meter_ch)
214 #define	BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
215 
216 /*
217  * Pending upcalls are stored in a vector which is flushed when
218  * full, or periodically
219  */
220 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
221 #define	V_bw_upcalls		VNET(bw_upcalls)
222 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
223 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
224 static VNET_DEFINE(struct callout, bw_upcalls_ch);
225 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
226 
227 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
228 
229 static VNET_PCPUSTAT_DEFINE(struct pimstat, pimstat);
230 VNET_PCPUSTAT_SYSINIT(pimstat);
231 VNET_PCPUSTAT_SYSUNINIT(pimstat);
232 
233 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
234 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
235     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
236 
237 static u_long	pim_squelch_wholepkt = 0;
238 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
239     &pim_squelch_wholepkt, 0,
240     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
241 
242 extern  struct domain inetdomain;
243 static const struct protosw in_pim_protosw = {
244 	.pr_type =		SOCK_RAW,
245 	.pr_domain =		&inetdomain,
246 	.pr_protocol =		IPPROTO_PIM,
247 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
248 	.pr_input =		pim_input,
249 	.pr_output =		rip_output,
250 	.pr_ctloutput =		rip_ctloutput,
251 	.pr_usrreqs =		&rip_usrreqs
252 };
253 static const struct encaptab *pim_encap_cookie;
254 
255 static int pim_encapcheck(const struct mbuf *, int, int, void *);
256 
257 /*
258  * Note: the PIM Register encapsulation adds the following in front of a
259  * data packet:
260  *
261  * struct pim_encap_hdr {
262  *    struct ip ip;
263  *    struct pim_encap_pimhdr  pim;
264  * }
265  *
266  */
267 
268 struct pim_encap_pimhdr {
269 	struct pim pim;
270 	uint32_t   flags;
271 };
272 #define		PIM_ENCAP_TTL	64
273 
274 static struct ip pim_encap_iphdr = {
275 #if BYTE_ORDER == LITTLE_ENDIAN
276 	sizeof(struct ip) >> 2,
277 	IPVERSION,
278 #else
279 	IPVERSION,
280 	sizeof(struct ip) >> 2,
281 #endif
282 	0,			/* tos */
283 	sizeof(struct ip),	/* total length */
284 	0,			/* id */
285 	0,			/* frag offset */
286 	PIM_ENCAP_TTL,
287 	IPPROTO_PIM,
288 	0,			/* checksum */
289 };
290 
291 static struct pim_encap_pimhdr pim_encap_pimhdr = {
292     {
293 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
294 	0,			/* reserved */
295 	0,			/* checksum */
296     },
297     0				/* flags */
298 };
299 
300 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
301 #define	V_reg_vif_num		VNET(reg_vif_num)
302 static VNET_DEFINE(struct ifnet, multicast_register_if);
303 #define	V_multicast_register_if	VNET(multicast_register_if)
304 
305 /*
306  * Private variables.
307  */
308 
309 static u_long	X_ip_mcast_src(int);
310 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
311 		    struct ip_moptions *);
312 static int	X_ip_mrouter_done(void);
313 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
314 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
315 static int	X_legal_vif_num(int);
316 static int	X_mrt_ioctl(u_long, caddr_t, int);
317 
318 static int	add_bw_upcall(struct bw_upcall *);
319 static int	add_mfc(struct mfcctl2 *);
320 static int	add_vif(struct vifctl *);
321 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
322 static void	bw_meter_process(void);
323 static void	bw_meter_receive_packet(struct bw_meter *, int,
324 		    struct timeval *);
325 static void	bw_upcalls_send(void);
326 static int	del_bw_upcall(struct bw_upcall *);
327 static int	del_mfc(struct mfcctl2 *);
328 static int	del_vif(vifi_t);
329 static int	del_vif_locked(vifi_t);
330 static void	expire_bw_meter_process(void *);
331 static void	expire_bw_upcalls_send(void *);
332 static void	expire_mfc(struct mfc *);
333 static void	expire_upcalls(void *);
334 static void	free_bw_list(struct bw_meter *);
335 static int	get_sg_cnt(struct sioc_sg_req *);
336 static int	get_vif_cnt(struct sioc_vif_req *);
337 static void	if_detached_event(void *, struct ifnet *);
338 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
339 static int	ip_mrouter_init(struct socket *, int);
340 static __inline struct mfc *
341 		mfc_find(struct in_addr *, struct in_addr *);
342 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
343 static struct mbuf *
344 		pim_register_prepare(struct ip *, struct mbuf *);
345 static int	pim_register_send(struct ip *, struct vif *,
346 		    struct mbuf *, struct mfc *);
347 static int	pim_register_send_rp(struct ip *, struct vif *,
348 		    struct mbuf *, struct mfc *);
349 static int	pim_register_send_upcall(struct ip *, struct vif *,
350 		    struct mbuf *, struct mfc *);
351 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
352 static void	send_packet(struct vif *, struct mbuf *);
353 static int	set_api_config(uint32_t *);
354 static int	set_assert(int);
355 static int	socket_send(struct socket *, struct mbuf *,
356 		    struct sockaddr_in *);
357 static void	unschedule_bw_meter(struct bw_meter *);
358 
359 /*
360  * Kernel multicast forwarding 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 #define MRT_API_VERSION		0x0305
365 
366 static const int mrt_api_version = MRT_API_VERSION;
367 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
368 					 MRT_MFC_FLAGS_BORDER_VIF |
369 					 MRT_MFC_RP |
370 					 MRT_MFC_BW_UPCALL);
371 static VNET_DEFINE(uint32_t, mrt_api_config);
372 #define	V_mrt_api_config	VNET(mrt_api_config)
373 static VNET_DEFINE(int, pim_assert_enabled);
374 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
375 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
376 
377 /*
378  * Find a route for a given origin IP address and multicast group address.
379  * Statistics must be updated by the caller.
380  */
381 static __inline struct mfc *
382 mfc_find(struct in_addr *o, struct in_addr *g)
383 {
384 	struct mfc *rt;
385 
386 	MFC_LOCK_ASSERT();
387 
388 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
389 		if (in_hosteq(rt->mfc_origin, *o) &&
390 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
391 		    TAILQ_EMPTY(&rt->mfc_stall))
392 			break;
393 	}
394 
395 	return (rt);
396 }
397 
398 /*
399  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
400  */
401 static int
402 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
403 {
404     int	error, optval;
405     vifi_t	vifi;
406     struct	vifctl vifc;
407     struct	mfcctl2 mfc;
408     struct	bw_upcall bw_upcall;
409     uint32_t	i;
410 
411     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
412 	return EPERM;
413 
414     error = 0;
415     switch (sopt->sopt_name) {
416     case MRT_INIT:
417 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
418 	if (error)
419 	    break;
420 	error = ip_mrouter_init(so, optval);
421 	break;
422 
423     case MRT_DONE:
424 	error = ip_mrouter_done();
425 	break;
426 
427     case MRT_ADD_VIF:
428 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
429 	if (error)
430 	    break;
431 	error = add_vif(&vifc);
432 	break;
433 
434     case MRT_DEL_VIF:
435 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
436 	if (error)
437 	    break;
438 	error = del_vif(vifi);
439 	break;
440 
441     case MRT_ADD_MFC:
442     case MRT_DEL_MFC:
443 	/*
444 	 * select data size depending on API version.
445 	 */
446 	if (sopt->sopt_name == MRT_ADD_MFC &&
447 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
448 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
449 				sizeof(struct mfcctl2));
450 	} else {
451 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
452 				sizeof(struct mfcctl));
453 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
454 			sizeof(mfc) - sizeof(struct mfcctl));
455 	}
456 	if (error)
457 	    break;
458 	if (sopt->sopt_name == MRT_ADD_MFC)
459 	    error = add_mfc(&mfc);
460 	else
461 	    error = del_mfc(&mfc);
462 	break;
463 
464     case MRT_ASSERT:
465 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
466 	if (error)
467 	    break;
468 	set_assert(optval);
469 	break;
470 
471     case MRT_API_CONFIG:
472 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
473 	if (!error)
474 	    error = set_api_config(&i);
475 	if (!error)
476 	    error = sooptcopyout(sopt, &i, sizeof i);
477 	break;
478 
479     case MRT_ADD_BW_UPCALL:
480     case MRT_DEL_BW_UPCALL:
481 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
482 				sizeof bw_upcall);
483 	if (error)
484 	    break;
485 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
486 	    error = add_bw_upcall(&bw_upcall);
487 	else
488 	    error = del_bw_upcall(&bw_upcall);
489 	break;
490 
491     default:
492 	error = EOPNOTSUPP;
493 	break;
494     }
495     return error;
496 }
497 
498 /*
499  * Handle MRT getsockopt commands
500  */
501 static int
502 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
503 {
504     int error;
505 
506     switch (sopt->sopt_name) {
507     case MRT_VERSION:
508 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
509 	break;
510 
511     case MRT_ASSERT:
512 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
513 	    sizeof V_pim_assert_enabled);
514 	break;
515 
516     case MRT_API_SUPPORT:
517 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
518 	break;
519 
520     case MRT_API_CONFIG:
521 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
522 	break;
523 
524     default:
525 	error = EOPNOTSUPP;
526 	break;
527     }
528     return error;
529 }
530 
531 /*
532  * Handle ioctl commands to obtain information from the cache
533  */
534 static int
535 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
536 {
537     int error = 0;
538 
539     /*
540      * Currently the only function calling this ioctl routine is rtioctl().
541      * Typically, only root can create the raw socket in order to execute
542      * this ioctl method, however the request might be coming from a prison
543      */
544     error = priv_check(curthread, PRIV_NETINET_MROUTE);
545     if (error)
546 	return (error);
547     switch (cmd) {
548     case (SIOCGETVIFCNT):
549 	error = get_vif_cnt((struct sioc_vif_req *)data);
550 	break;
551 
552     case (SIOCGETSGCNT):
553 	error = get_sg_cnt((struct sioc_sg_req *)data);
554 	break;
555 
556     default:
557 	error = EINVAL;
558 	break;
559     }
560     return error;
561 }
562 
563 /*
564  * returns the packet, byte, rpf-failure count for the source group provided
565  */
566 static int
567 get_sg_cnt(struct sioc_sg_req *req)
568 {
569     struct mfc *rt;
570 
571     MFC_LOCK();
572     rt = mfc_find(&req->src, &req->grp);
573     if (rt == NULL) {
574 	MFC_UNLOCK();
575 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
576 	return EADDRNOTAVAIL;
577     }
578     req->pktcnt = rt->mfc_pkt_cnt;
579     req->bytecnt = rt->mfc_byte_cnt;
580     req->wrong_if = rt->mfc_wrong_if;
581     MFC_UNLOCK();
582     return 0;
583 }
584 
585 /*
586  * returns the input and output packet and byte counts on the vif provided
587  */
588 static int
589 get_vif_cnt(struct sioc_vif_req *req)
590 {
591     vifi_t vifi = req->vifi;
592 
593     VIF_LOCK();
594     if (vifi >= V_numvifs) {
595 	VIF_UNLOCK();
596 	return EINVAL;
597     }
598 
599     req->icount = V_viftable[vifi].v_pkt_in;
600     req->ocount = V_viftable[vifi].v_pkt_out;
601     req->ibytes = V_viftable[vifi].v_bytes_in;
602     req->obytes = V_viftable[vifi].v_bytes_out;
603     VIF_UNLOCK();
604 
605     return 0;
606 }
607 
608 static void
609 if_detached_event(void *arg __unused, struct ifnet *ifp)
610 {
611     vifi_t vifi;
612     u_long i;
613 
614     MROUTER_LOCK();
615 
616     if (V_ip_mrouter == NULL) {
617 	MROUTER_UNLOCK();
618 	return;
619     }
620 
621     VIF_LOCK();
622     MFC_LOCK();
623 
624     /*
625      * Tear down multicast forwarder state associated with this ifnet.
626      * 1. Walk the vif list, matching vifs against this ifnet.
627      * 2. Walk the multicast forwarding cache (mfc) looking for
628      *    inner matches with this vif's index.
629      * 3. Expire any matching multicast forwarding cache entries.
630      * 4. Free vif state. This should disable ALLMULTI on the interface.
631      */
632     for (vifi = 0; vifi < V_numvifs; vifi++) {
633 	if (V_viftable[vifi].v_ifp != ifp)
634 		continue;
635 	for (i = 0; i < mfchashsize; i++) {
636 		struct mfc *rt, *nrt;
637 
638 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
639 			if (rt->mfc_parent == vifi) {
640 				expire_mfc(rt);
641 			}
642 		}
643 	}
644 	del_vif_locked(vifi);
645     }
646 
647     MFC_UNLOCK();
648     VIF_UNLOCK();
649 
650     MROUTER_UNLOCK();
651 }
652 
653 /*
654  * Enable multicast forwarding.
655  */
656 static int
657 ip_mrouter_init(struct socket *so, int version)
658 {
659 
660     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
661         so->so_type, so->so_proto->pr_protocol);
662 
663     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
664 	return EOPNOTSUPP;
665 
666     if (version != 1)
667 	return ENOPROTOOPT;
668 
669     MROUTER_LOCK();
670 
671     if (ip_mrouter_unloading) {
672 	MROUTER_UNLOCK();
673 	return ENOPROTOOPT;
674     }
675 
676     if (V_ip_mrouter != NULL) {
677 	MROUTER_UNLOCK();
678 	return EADDRINUSE;
679     }
680 
681     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
682 	HASH_NOWAIT);
683 
684     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
685 	curvnet);
686     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
687 	curvnet);
688     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
689 	curvnet);
690 
691     V_ip_mrouter = so;
692     ip_mrouter_cnt++;
693 
694     MROUTER_UNLOCK();
695 
696     CTR1(KTR_IPMF, "%s: done", __func__);
697 
698     return 0;
699 }
700 
701 /*
702  * Disable multicast forwarding.
703  */
704 static int
705 X_ip_mrouter_done(void)
706 {
707     struct ifnet *ifp;
708     u_long i;
709     vifi_t vifi;
710 
711     MROUTER_LOCK();
712 
713     if (V_ip_mrouter == NULL) {
714 	MROUTER_UNLOCK();
715 	return EINVAL;
716     }
717 
718     /*
719      * Detach/disable hooks to the reset of the system.
720      */
721     V_ip_mrouter = NULL;
722     ip_mrouter_cnt--;
723     V_mrt_api_config = 0;
724 
725     VIF_LOCK();
726 
727     /*
728      * For each phyint in use, disable promiscuous reception of all IP
729      * multicasts.
730      */
731     for (vifi = 0; vifi < V_numvifs; vifi++) {
732 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
733 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
734 	    ifp = V_viftable[vifi].v_ifp;
735 	    if_allmulti(ifp, 0);
736 	}
737     }
738     bzero((caddr_t)V_viftable, sizeof(V_viftable));
739     V_numvifs = 0;
740     V_pim_assert_enabled = 0;
741 
742     VIF_UNLOCK();
743 
744     callout_stop(&V_expire_upcalls_ch);
745     callout_stop(&V_bw_upcalls_ch);
746     callout_stop(&V_bw_meter_ch);
747 
748     MFC_LOCK();
749 
750     /*
751      * Free all multicast forwarding cache entries.
752      * Do not use hashdestroy(), as we must perform other cleanup.
753      */
754     for (i = 0; i < mfchashsize; i++) {
755 	struct mfc *rt, *nrt;
756 
757 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
758 		expire_mfc(rt);
759 	}
760     }
761     free(V_mfchashtbl, M_MRTABLE);
762     V_mfchashtbl = NULL;
763 
764     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
765 
766     V_bw_upcalls_n = 0;
767     bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
768 
769     MFC_UNLOCK();
770 
771     V_reg_vif_num = VIFI_INVALID;
772 
773     MROUTER_UNLOCK();
774 
775     CTR1(KTR_IPMF, "%s: done", __func__);
776 
777     return 0;
778 }
779 
780 /*
781  * Set PIM assert processing global
782  */
783 static int
784 set_assert(int i)
785 {
786     if ((i != 1) && (i != 0))
787 	return EINVAL;
788 
789     V_pim_assert_enabled = i;
790 
791     return 0;
792 }
793 
794 /*
795  * Configure API capabilities
796  */
797 int
798 set_api_config(uint32_t *apival)
799 {
800     u_long i;
801 
802     /*
803      * We can set the API capabilities only if it is the first operation
804      * after MRT_INIT. I.e.:
805      *  - there are no vifs installed
806      *  - pim_assert is not enabled
807      *  - the MFC table is empty
808      */
809     if (V_numvifs > 0) {
810 	*apival = 0;
811 	return EPERM;
812     }
813     if (V_pim_assert_enabled) {
814 	*apival = 0;
815 	return EPERM;
816     }
817 
818     MFC_LOCK();
819 
820     for (i = 0; i < mfchashsize; i++) {
821 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
822 	    MFC_UNLOCK();
823 	    *apival = 0;
824 	    return EPERM;
825 	}
826     }
827 
828     MFC_UNLOCK();
829 
830     V_mrt_api_config = *apival & mrt_api_support;
831     *apival = V_mrt_api_config;
832 
833     return 0;
834 }
835 
836 /*
837  * Add a vif to the vif table
838  */
839 static int
840 add_vif(struct vifctl *vifcp)
841 {
842     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
843     struct sockaddr_in sin = {sizeof sin, AF_INET};
844     struct ifaddr *ifa;
845     struct ifnet *ifp;
846     int error;
847 
848     VIF_LOCK();
849     if (vifcp->vifc_vifi >= MAXVIFS) {
850 	VIF_UNLOCK();
851 	return EINVAL;
852     }
853     /* rate limiting is no longer supported by this code */
854     if (vifcp->vifc_rate_limit != 0) {
855 	log(LOG_ERR, "rate limiting is no longer supported\n");
856 	VIF_UNLOCK();
857 	return EINVAL;
858     }
859     if (!in_nullhost(vifp->v_lcl_addr)) {
860 	VIF_UNLOCK();
861 	return EADDRINUSE;
862     }
863     if (in_nullhost(vifcp->vifc_lcl_addr)) {
864 	VIF_UNLOCK();
865 	return EADDRNOTAVAIL;
866     }
867 
868     /* Find the interface with an address in AF_INET family */
869     if (vifcp->vifc_flags & VIFF_REGISTER) {
870 	/*
871 	 * XXX: Because VIFF_REGISTER does not really need a valid
872 	 * local interface (e.g. it could be 127.0.0.2), we don't
873 	 * check its address.
874 	 */
875 	ifp = NULL;
876     } else {
877 	sin.sin_addr = vifcp->vifc_lcl_addr;
878 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
879 	if (ifa == NULL) {
880 	    VIF_UNLOCK();
881 	    return EADDRNOTAVAIL;
882 	}
883 	ifp = ifa->ifa_ifp;
884 	ifa_free(ifa);
885     }
886 
887     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
888 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
889 	VIF_UNLOCK();
890 	return EOPNOTSUPP;
891     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
892 	ifp = &V_multicast_register_if;
893 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
894 	if (V_reg_vif_num == VIFI_INVALID) {
895 	    if_initname(&V_multicast_register_if, "register_vif", 0);
896 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
897 	    V_reg_vif_num = vifcp->vifc_vifi;
898 	}
899     } else {		/* Make sure the interface supports multicast */
900 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
901 	    VIF_UNLOCK();
902 	    return EOPNOTSUPP;
903 	}
904 
905 	/* Enable promiscuous reception of all IP multicasts from the if */
906 	error = if_allmulti(ifp, 1);
907 	if (error) {
908 	    VIF_UNLOCK();
909 	    return error;
910 	}
911     }
912 
913     vifp->v_flags     = vifcp->vifc_flags;
914     vifp->v_threshold = vifcp->vifc_threshold;
915     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
916     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
917     vifp->v_ifp       = ifp;
918     /* initialize per vif pkt counters */
919     vifp->v_pkt_in    = 0;
920     vifp->v_pkt_out   = 0;
921     vifp->v_bytes_in  = 0;
922     vifp->v_bytes_out = 0;
923 
924     /* Adjust numvifs up if the vifi is higher than numvifs */
925     if (V_numvifs <= vifcp->vifc_vifi)
926 	V_numvifs = vifcp->vifc_vifi + 1;
927 
928     VIF_UNLOCK();
929 
930     CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
931 	(int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr),
932 	(int)vifcp->vifc_threshold);
933 
934     return 0;
935 }
936 
937 /*
938  * Delete a vif from the vif table
939  */
940 static int
941 del_vif_locked(vifi_t vifi)
942 {
943     struct vif *vifp;
944 
945     VIF_LOCK_ASSERT();
946 
947     if (vifi >= V_numvifs) {
948 	return EINVAL;
949     }
950     vifp = &V_viftable[vifi];
951     if (in_nullhost(vifp->v_lcl_addr)) {
952 	return EADDRNOTAVAIL;
953     }
954 
955     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
956 	if_allmulti(vifp->v_ifp, 0);
957 
958     if (vifp->v_flags & VIFF_REGISTER)
959 	V_reg_vif_num = VIFI_INVALID;
960 
961     bzero((caddr_t)vifp, sizeof (*vifp));
962 
963     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
964 
965     /* Adjust numvifs down */
966     for (vifi = V_numvifs; vifi > 0; vifi--)
967 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
968 	    break;
969     V_numvifs = vifi;
970 
971     return 0;
972 }
973 
974 static int
975 del_vif(vifi_t vifi)
976 {
977     int cc;
978 
979     VIF_LOCK();
980     cc = del_vif_locked(vifi);
981     VIF_UNLOCK();
982 
983     return cc;
984 }
985 
986 /*
987  * update an mfc entry without resetting counters and S,G addresses.
988  */
989 static void
990 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
991 {
992     int i;
993 
994     rt->mfc_parent = mfccp->mfcc_parent;
995     for (i = 0; i < V_numvifs; i++) {
996 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
997 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
998 	    MRT_MFC_FLAGS_ALL;
999     }
1000     /* set the RP address */
1001     if (V_mrt_api_config & MRT_MFC_RP)
1002 	rt->mfc_rp = mfccp->mfcc_rp;
1003     else
1004 	rt->mfc_rp.s_addr = INADDR_ANY;
1005 }
1006 
1007 /*
1008  * fully initialize an mfc entry from the parameter.
1009  */
1010 static void
1011 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1012 {
1013     rt->mfc_origin     = mfccp->mfcc_origin;
1014     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1015 
1016     update_mfc_params(rt, mfccp);
1017 
1018     /* initialize pkt counters per src-grp */
1019     rt->mfc_pkt_cnt    = 0;
1020     rt->mfc_byte_cnt   = 0;
1021     rt->mfc_wrong_if   = 0;
1022     timevalclear(&rt->mfc_last_assert);
1023 }
1024 
1025 static void
1026 expire_mfc(struct mfc *rt)
1027 {
1028 	struct rtdetq *rte, *nrte;
1029 
1030 	MFC_LOCK_ASSERT();
1031 
1032 	free_bw_list(rt->mfc_bw_meter);
1033 
1034 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1035 		m_freem(rte->m);
1036 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1037 		free(rte, M_MRTABLE);
1038 	}
1039 
1040 	LIST_REMOVE(rt, mfc_hash);
1041 	free(rt, M_MRTABLE);
1042 }
1043 
1044 /*
1045  * Add an mfc entry
1046  */
1047 static int
1048 add_mfc(struct mfcctl2 *mfccp)
1049 {
1050     struct mfc *rt;
1051     struct rtdetq *rte, *nrte;
1052     u_long hash = 0;
1053     u_short nstl;
1054 
1055     VIF_LOCK();
1056     MFC_LOCK();
1057 
1058     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1059 
1060     /* If an entry already exists, just update the fields */
1061     if (rt) {
1062 	CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x",
1063 	    __func__, inet_ntoa(mfccp->mfcc_origin),
1064 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1065 	    mfccp->mfcc_parent);
1066 	update_mfc_params(rt, mfccp);
1067 	MFC_UNLOCK();
1068 	VIF_UNLOCK();
1069 	return (0);
1070     }
1071 
1072     /*
1073      * Find the entry for which the upcall was made and update
1074      */
1075     nstl = 0;
1076     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1077     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1078 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1079 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1080 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1081 		CTR5(KTR_IPMF,
1082 		    "%s: add mfc orig %s group %lx parent %x qh %p",
1083 		    __func__, inet_ntoa(mfccp->mfcc_origin),
1084 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1085 		    mfccp->mfcc_parent,
1086 		    TAILQ_FIRST(&rt->mfc_stall));
1087 		if (nstl++)
1088 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1089 
1090 		init_mfc_params(rt, mfccp);
1091 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1092 		V_nexpire[hash]--;
1093 
1094 		/* Free queued packets, but attempt to forward them first. */
1095 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1096 			if (rte->ifp != NULL)
1097 				ip_mdq(rte->m, rte->ifp, rt, -1);
1098 			m_freem(rte->m);
1099 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1100 			rt->mfc_nstall--;
1101 			free(rte, M_MRTABLE);
1102 		}
1103 	}
1104     }
1105 
1106     /*
1107      * It is possible that an entry is being inserted without an upcall
1108      */
1109     if (nstl == 0) {
1110 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1111 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1112 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1113 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1114 			init_mfc_params(rt, mfccp);
1115 			if (rt->mfc_expire)
1116 			    V_nexpire[hash]--;
1117 			rt->mfc_expire = 0;
1118 			break; /* XXX */
1119 		}
1120 	}
1121 
1122 	if (rt == NULL) {		/* no upcall, so make a new entry */
1123 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1124 	    if (rt == NULL) {
1125 		MFC_UNLOCK();
1126 		VIF_UNLOCK();
1127 		return (ENOBUFS);
1128 	    }
1129 
1130 	    init_mfc_params(rt, mfccp);
1131 	    TAILQ_INIT(&rt->mfc_stall);
1132 	    rt->mfc_nstall = 0;
1133 
1134 	    rt->mfc_expire     = 0;
1135 	    rt->mfc_bw_meter = NULL;
1136 
1137 	    /* insert new entry at head of hash chain */
1138 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1139 	}
1140     }
1141 
1142     MFC_UNLOCK();
1143     VIF_UNLOCK();
1144 
1145     return (0);
1146 }
1147 
1148 /*
1149  * Delete an mfc entry
1150  */
1151 static int
1152 del_mfc(struct mfcctl2 *mfccp)
1153 {
1154     struct in_addr	origin;
1155     struct in_addr	mcastgrp;
1156     struct mfc		*rt;
1157 
1158     origin = mfccp->mfcc_origin;
1159     mcastgrp = mfccp->mfcc_mcastgrp;
1160 
1161     CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__,
1162 	inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr));
1163 
1164     MFC_LOCK();
1165 
1166     rt = mfc_find(&origin, &mcastgrp);
1167     if (rt == NULL) {
1168 	MFC_UNLOCK();
1169 	return EADDRNOTAVAIL;
1170     }
1171 
1172     /*
1173      * free the bw_meter entries
1174      */
1175     free_bw_list(rt->mfc_bw_meter);
1176     rt->mfc_bw_meter = NULL;
1177 
1178     LIST_REMOVE(rt, mfc_hash);
1179     free(rt, M_MRTABLE);
1180 
1181     MFC_UNLOCK();
1182 
1183     return (0);
1184 }
1185 
1186 /*
1187  * Send a message to the routing daemon on the multicast routing socket.
1188  */
1189 static int
1190 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1191 {
1192     if (s) {
1193 	SOCKBUF_LOCK(&s->so_rcv);
1194 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1195 	    NULL) != 0) {
1196 	    sorwakeup_locked(s);
1197 	    return 0;
1198 	}
1199 	SOCKBUF_UNLOCK(&s->so_rcv);
1200     }
1201     m_freem(mm);
1202     return -1;
1203 }
1204 
1205 /*
1206  * IP multicast forwarding function. This function assumes that the packet
1207  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1208  * pointed to by "ifp", and the packet is to be relayed to other networks
1209  * that have members of the packet's destination IP multicast group.
1210  *
1211  * The packet is returned unscathed to the caller, unless it is
1212  * erroneous, in which case a non-zero return value tells the caller to
1213  * discard it.
1214  */
1215 
1216 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1217 
1218 static int
1219 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1220     struct ip_moptions *imo)
1221 {
1222     struct mfc *rt;
1223     int error;
1224     vifi_t vifi;
1225 
1226     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p",
1227 	inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1228 
1229     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1230 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1231 	/*
1232 	 * Packet arrived via a physical interface or
1233 	 * an encapsulated tunnel or a register_vif.
1234 	 */
1235     } else {
1236 	/*
1237 	 * Packet arrived through a source-route tunnel.
1238 	 * Source-route tunnels are no longer supported.
1239 	 */
1240 	return (1);
1241     }
1242 
1243     VIF_LOCK();
1244     MFC_LOCK();
1245     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1246 	if (ip->ip_ttl < MAXTTL)
1247 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1248 	error = ip_mdq(m, ifp, NULL, vifi);
1249 	MFC_UNLOCK();
1250 	VIF_UNLOCK();
1251 	return error;
1252     }
1253 
1254     /*
1255      * Don't forward a packet with time-to-live of zero or one,
1256      * or a packet destined to a local-only group.
1257      */
1258     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1259 	MFC_UNLOCK();
1260 	VIF_UNLOCK();
1261 	return 0;
1262     }
1263 
1264     /*
1265      * Determine forwarding vifs from the forwarding cache table
1266      */
1267     MRTSTAT_INC(mrts_mfc_lookups);
1268     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1269 
1270     /* Entry exists, so forward if necessary */
1271     if (rt != NULL) {
1272 	error = ip_mdq(m, ifp, rt, -1);
1273 	MFC_UNLOCK();
1274 	VIF_UNLOCK();
1275 	return error;
1276     } else {
1277 	/*
1278 	 * If we don't have a route for packet's origin,
1279 	 * Make a copy of the packet & send message to routing daemon
1280 	 */
1281 
1282 	struct mbuf *mb0;
1283 	struct rtdetq *rte;
1284 	u_long hash;
1285 	int hlen = ip->ip_hl << 2;
1286 
1287 	MRTSTAT_INC(mrts_mfc_misses);
1288 	MRTSTAT_INC(mrts_no_route);
1289 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)",
1290 	    inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr));
1291 
1292 	/*
1293 	 * Allocate mbufs early so that we don't do extra work if we are
1294 	 * just going to fail anyway.  Make sure to pullup the header so
1295 	 * that other people can't step on it.
1296 	 */
1297 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1298 	    M_NOWAIT|M_ZERO);
1299 	if (rte == NULL) {
1300 	    MFC_UNLOCK();
1301 	    VIF_UNLOCK();
1302 	    return ENOBUFS;
1303 	}
1304 
1305 	mb0 = m_copypacket(m, M_NOWAIT);
1306 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1307 	    mb0 = m_pullup(mb0, hlen);
1308 	if (mb0 == NULL) {
1309 	    free(rte, M_MRTABLE);
1310 	    MFC_UNLOCK();
1311 	    VIF_UNLOCK();
1312 	    return ENOBUFS;
1313 	}
1314 
1315 	/* is there an upcall waiting for this flow ? */
1316 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1317 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1318 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1319 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1320 		    !TAILQ_EMPTY(&rt->mfc_stall))
1321 			break;
1322 	}
1323 
1324 	if (rt == NULL) {
1325 	    int i;
1326 	    struct igmpmsg *im;
1327 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1328 	    struct mbuf *mm;
1329 
1330 	    /*
1331 	     * Locate the vifi for the incoming interface for this packet.
1332 	     * If none found, drop packet.
1333 	     */
1334 	    for (vifi = 0; vifi < V_numvifs &&
1335 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1336 		;
1337 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1338 		goto non_fatal;
1339 
1340 	    /* no upcall, so make a new entry */
1341 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1342 	    if (rt == NULL)
1343 		goto fail;
1344 
1345 	    /* Make a copy of the header to send to the user level process */
1346 	    mm = m_copy(mb0, 0, hlen);
1347 	    if (mm == NULL)
1348 		goto fail1;
1349 
1350 	    /*
1351 	     * Send message to routing daemon to install
1352 	     * a route into the kernel table
1353 	     */
1354 
1355 	    im = mtod(mm, struct igmpmsg *);
1356 	    im->im_msgtype = IGMPMSG_NOCACHE;
1357 	    im->im_mbz = 0;
1358 	    im->im_vif = vifi;
1359 
1360 	    MRTSTAT_INC(mrts_upcalls);
1361 
1362 	    k_igmpsrc.sin_addr = ip->ip_src;
1363 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1364 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1365 		MRTSTAT_INC(mrts_upq_sockfull);
1366 fail1:
1367 		free(rt, M_MRTABLE);
1368 fail:
1369 		free(rte, M_MRTABLE);
1370 		m_freem(mb0);
1371 		MFC_UNLOCK();
1372 		VIF_UNLOCK();
1373 		return ENOBUFS;
1374 	    }
1375 
1376 	    /* insert new entry at head of hash chain */
1377 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1378 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1379 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1380 	    V_nexpire[hash]++;
1381 	    for (i = 0; i < V_numvifs; i++) {
1382 		rt->mfc_ttls[i] = 0;
1383 		rt->mfc_flags[i] = 0;
1384 	    }
1385 	    rt->mfc_parent = -1;
1386 
1387 	    /* clear the RP address */
1388 	    rt->mfc_rp.s_addr = INADDR_ANY;
1389 	    rt->mfc_bw_meter = NULL;
1390 
1391 	    /* initialize pkt counters per src-grp */
1392 	    rt->mfc_pkt_cnt = 0;
1393 	    rt->mfc_byte_cnt = 0;
1394 	    rt->mfc_wrong_if = 0;
1395 	    timevalclear(&rt->mfc_last_assert);
1396 
1397 	    TAILQ_INIT(&rt->mfc_stall);
1398 	    rt->mfc_nstall = 0;
1399 
1400 	    /* link into table */
1401 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1402 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1403 	    rt->mfc_nstall++;
1404 
1405 	} else {
1406 	    /* determine if queue has overflowed */
1407 	    if (rt->mfc_nstall > MAX_UPQ) {
1408 		MRTSTAT_INC(mrts_upq_ovflw);
1409 non_fatal:
1410 		free(rte, M_MRTABLE);
1411 		m_freem(mb0);
1412 		MFC_UNLOCK();
1413 		VIF_UNLOCK();
1414 		return (0);
1415 	    }
1416 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1417 	    rt->mfc_nstall++;
1418 	}
1419 
1420 	rte->m			= mb0;
1421 	rte->ifp		= ifp;
1422 
1423 	MFC_UNLOCK();
1424 	VIF_UNLOCK();
1425 
1426 	return 0;
1427     }
1428 }
1429 
1430 /*
1431  * Clean up the cache entry if upcall is not serviced
1432  */
1433 static void
1434 expire_upcalls(void *arg)
1435 {
1436     u_long i;
1437 
1438     CURVNET_SET((struct vnet *) arg);
1439 
1440     MFC_LOCK();
1441 
1442     for (i = 0; i < mfchashsize; i++) {
1443 	struct mfc *rt, *nrt;
1444 
1445 	if (V_nexpire[i] == 0)
1446 	    continue;
1447 
1448 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1449 		if (TAILQ_EMPTY(&rt->mfc_stall))
1450 			continue;
1451 
1452 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1453 			continue;
1454 
1455 		/*
1456 		 * free the bw_meter entries
1457 		 */
1458 		while (rt->mfc_bw_meter != NULL) {
1459 		    struct bw_meter *x = rt->mfc_bw_meter;
1460 
1461 		    rt->mfc_bw_meter = x->bm_mfc_next;
1462 		    free(x, M_BWMETER);
1463 		}
1464 
1465 		MRTSTAT_INC(mrts_cache_cleanups);
1466 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1467 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1468 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1469 
1470 		expire_mfc(rt);
1471 	    }
1472     }
1473 
1474     MFC_UNLOCK();
1475 
1476     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1477 	curvnet);
1478 
1479     CURVNET_RESTORE();
1480 }
1481 
1482 /*
1483  * Packet forwarding routine once entry in the cache is made
1484  */
1485 static int
1486 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1487 {
1488     struct ip  *ip = mtod(m, struct ip *);
1489     vifi_t vifi;
1490     int plen = ntohs(ip->ip_len);
1491 
1492     VIF_LOCK_ASSERT();
1493 
1494     /*
1495      * If xmt_vif is not -1, send on only the requested vif.
1496      *
1497      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1498      */
1499     if (xmt_vif < V_numvifs) {
1500 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1501 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1502 	else
1503 		phyint_send(ip, V_viftable + xmt_vif, m);
1504 	return 1;
1505     }
1506 
1507     /*
1508      * Don't forward if it didn't arrive from the parent vif for its origin.
1509      */
1510     vifi = rt->mfc_parent;
1511     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1512 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1513 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1514 	MRTSTAT_INC(mrts_wrong_if);
1515 	++rt->mfc_wrong_if;
1516 	/*
1517 	 * If we are doing PIM assert processing, send a message
1518 	 * to the routing daemon.
1519 	 *
1520 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1521 	 * can complete the SPT switch, regardless of the type
1522 	 * of the iif (broadcast media, GRE tunnel, etc).
1523 	 */
1524 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1525 	    V_viftable[vifi].v_ifp) {
1526 
1527 	    if (ifp == &V_multicast_register_if)
1528 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1529 
1530 	    /* Get vifi for the incoming packet */
1531 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1532 		vifi++)
1533 		;
1534 	    if (vifi >= V_numvifs)
1535 		return 0;	/* The iif is not found: ignore the packet. */
1536 
1537 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1538 		return 0;	/* WRONGVIF disabled: ignore the packet */
1539 
1540 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1541 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1542 		struct igmpmsg *im;
1543 		int hlen = ip->ip_hl << 2;
1544 		struct mbuf *mm = m_copy(m, 0, hlen);
1545 
1546 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1547 		    mm = m_pullup(mm, hlen);
1548 		if (mm == NULL)
1549 		    return ENOBUFS;
1550 
1551 		im = mtod(mm, struct igmpmsg *);
1552 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1553 		im->im_mbz		= 0;
1554 		im->im_vif		= vifi;
1555 
1556 		MRTSTAT_INC(mrts_upcalls);
1557 
1558 		k_igmpsrc.sin_addr = im->im_src;
1559 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1560 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1561 		    MRTSTAT_INC(mrts_upq_sockfull);
1562 		    return ENOBUFS;
1563 		}
1564 	    }
1565 	}
1566 	return 0;
1567     }
1568 
1569 
1570     /* If I sourced this packet, it counts as output, else it was input. */
1571     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1572 	V_viftable[vifi].v_pkt_out++;
1573 	V_viftable[vifi].v_bytes_out += plen;
1574     } else {
1575 	V_viftable[vifi].v_pkt_in++;
1576 	V_viftable[vifi].v_bytes_in += plen;
1577     }
1578     rt->mfc_pkt_cnt++;
1579     rt->mfc_byte_cnt += plen;
1580 
1581     /*
1582      * For each vif, decide if a copy of the packet should be forwarded.
1583      * Forward if:
1584      *		- the ttl exceeds the vif's threshold
1585      *		- there are group members downstream on interface
1586      */
1587     for (vifi = 0; vifi < V_numvifs; vifi++)
1588 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1589 	    V_viftable[vifi].v_pkt_out++;
1590 	    V_viftable[vifi].v_bytes_out += plen;
1591 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1592 		pim_register_send(ip, V_viftable + vifi, m, rt);
1593 	    else
1594 		phyint_send(ip, V_viftable + vifi, m);
1595 	}
1596 
1597     /*
1598      * Perform upcall-related bw measuring.
1599      */
1600     if (rt->mfc_bw_meter != NULL) {
1601 	struct bw_meter *x;
1602 	struct timeval now;
1603 
1604 	microtime(&now);
1605 	MFC_LOCK_ASSERT();
1606 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1607 	    bw_meter_receive_packet(x, plen, &now);
1608     }
1609 
1610     return 0;
1611 }
1612 
1613 /*
1614  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1615  */
1616 static int
1617 X_legal_vif_num(int vif)
1618 {
1619 	int ret;
1620 
1621 	ret = 0;
1622 	if (vif < 0)
1623 		return (ret);
1624 
1625 	VIF_LOCK();
1626 	if (vif < V_numvifs)
1627 		ret = 1;
1628 	VIF_UNLOCK();
1629 
1630 	return (ret);
1631 }
1632 
1633 /*
1634  * Return the local address used by this vif
1635  */
1636 static u_long
1637 X_ip_mcast_src(int vifi)
1638 {
1639 	in_addr_t addr;
1640 
1641 	addr = INADDR_ANY;
1642 	if (vifi < 0)
1643 		return (addr);
1644 
1645 	VIF_LOCK();
1646 	if (vifi < V_numvifs)
1647 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1648 	VIF_UNLOCK();
1649 
1650 	return (addr);
1651 }
1652 
1653 static void
1654 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1655 {
1656     struct mbuf *mb_copy;
1657     int hlen = ip->ip_hl << 2;
1658 
1659     VIF_LOCK_ASSERT();
1660 
1661     /*
1662      * Make a new reference to the packet; make sure that
1663      * the IP header is actually copied, not just referenced,
1664      * so that ip_output() only scribbles on the copy.
1665      */
1666     mb_copy = m_copypacket(m, M_NOWAIT);
1667     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1668 	mb_copy = m_pullup(mb_copy, hlen);
1669     if (mb_copy == NULL)
1670 	return;
1671 
1672     send_packet(vifp, mb_copy);
1673 }
1674 
1675 static void
1676 send_packet(struct vif *vifp, struct mbuf *m)
1677 {
1678 	struct ip_moptions imo;
1679 	struct in_multi *imm[2];
1680 	int error;
1681 
1682 	VIF_LOCK_ASSERT();
1683 
1684 	imo.imo_multicast_ifp  = vifp->v_ifp;
1685 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1686 	imo.imo_multicast_loop = 1;
1687 	imo.imo_multicast_vif  = -1;
1688 	imo.imo_num_memberships = 0;
1689 	imo.imo_max_memberships = 2;
1690 	imo.imo_membership  = &imm[0];
1691 
1692 	/*
1693 	 * Re-entrancy should not be a problem here, because
1694 	 * the packets that we send out and are looped back at us
1695 	 * should get rejected because they appear to come from
1696 	 * the loopback interface, thus preventing looping.
1697 	 */
1698 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1699 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1700 	    (ptrdiff_t)(vifp - V_viftable), error);
1701 }
1702 
1703 /*
1704  * Stubs for old RSVP socket shim implementation.
1705  */
1706 
1707 static int
1708 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1709 {
1710 
1711 	return (EOPNOTSUPP);
1712 }
1713 
1714 static void
1715 X_ip_rsvp_force_done(struct socket *so __unused)
1716 {
1717 
1718 }
1719 
1720 static int
1721 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1722 {
1723 	struct mbuf *m;
1724 
1725 	m = *mp;
1726 	*mp = NULL;
1727 	if (!V_rsvp_on)
1728 		m_freem(m);
1729 	return (IPPROTO_DONE);
1730 }
1731 
1732 /*
1733  * Code for bandwidth monitors
1734  */
1735 
1736 /*
1737  * Define common interface for timeval-related methods
1738  */
1739 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1740 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1741 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1742 
1743 static uint32_t
1744 compute_bw_meter_flags(struct bw_upcall *req)
1745 {
1746     uint32_t flags = 0;
1747 
1748     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1749 	flags |= BW_METER_UNIT_PACKETS;
1750     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1751 	flags |= BW_METER_UNIT_BYTES;
1752     if (req->bu_flags & BW_UPCALL_GEQ)
1753 	flags |= BW_METER_GEQ;
1754     if (req->bu_flags & BW_UPCALL_LEQ)
1755 	flags |= BW_METER_LEQ;
1756 
1757     return flags;
1758 }
1759 
1760 /*
1761  * Add a bw_meter entry
1762  */
1763 static int
1764 add_bw_upcall(struct bw_upcall *req)
1765 {
1766     struct mfc *mfc;
1767     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1768 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1769     struct timeval now;
1770     struct bw_meter *x;
1771     uint32_t flags;
1772 
1773     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1774 	return EOPNOTSUPP;
1775 
1776     /* Test if the flags are valid */
1777     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1778 	return EINVAL;
1779     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1780 	return EINVAL;
1781     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1782 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1783 	return EINVAL;
1784 
1785     /* Test if the threshold time interval is valid */
1786     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1787 	return EINVAL;
1788 
1789     flags = compute_bw_meter_flags(req);
1790 
1791     /*
1792      * Find if we have already same bw_meter entry
1793      */
1794     MFC_LOCK();
1795     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1796     if (mfc == NULL) {
1797 	MFC_UNLOCK();
1798 	return EADDRNOTAVAIL;
1799     }
1800     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1801 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1802 			   &req->bu_threshold.b_time, ==)) &&
1803 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1804 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1805 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1806 	    MFC_UNLOCK();
1807 	    return 0;		/* XXX Already installed */
1808 	}
1809     }
1810 
1811     /* Allocate the new bw_meter entry */
1812     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1813     if (x == NULL) {
1814 	MFC_UNLOCK();
1815 	return ENOBUFS;
1816     }
1817 
1818     /* Set the new bw_meter entry */
1819     x->bm_threshold.b_time = req->bu_threshold.b_time;
1820     microtime(&now);
1821     x->bm_start_time = now;
1822     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1823     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1824     x->bm_measured.b_packets = 0;
1825     x->bm_measured.b_bytes = 0;
1826     x->bm_flags = flags;
1827     x->bm_time_next = NULL;
1828     x->bm_time_hash = BW_METER_BUCKETS;
1829 
1830     /* Add the new bw_meter entry to the front of entries for this MFC */
1831     x->bm_mfc = mfc;
1832     x->bm_mfc_next = mfc->mfc_bw_meter;
1833     mfc->mfc_bw_meter = x;
1834     schedule_bw_meter(x, &now);
1835     MFC_UNLOCK();
1836 
1837     return 0;
1838 }
1839 
1840 static void
1841 free_bw_list(struct bw_meter *list)
1842 {
1843     while (list != NULL) {
1844 	struct bw_meter *x = list;
1845 
1846 	list = list->bm_mfc_next;
1847 	unschedule_bw_meter(x);
1848 	free(x, M_BWMETER);
1849     }
1850 }
1851 
1852 /*
1853  * Delete one or multiple bw_meter entries
1854  */
1855 static int
1856 del_bw_upcall(struct bw_upcall *req)
1857 {
1858     struct mfc *mfc;
1859     struct bw_meter *x;
1860 
1861     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1862 	return EOPNOTSUPP;
1863 
1864     MFC_LOCK();
1865 
1866     /* Find the corresponding MFC entry */
1867     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1868     if (mfc == NULL) {
1869 	MFC_UNLOCK();
1870 	return EADDRNOTAVAIL;
1871     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1872 	/*
1873 	 * Delete all bw_meter entries for this mfc
1874 	 */
1875 	struct bw_meter *list;
1876 
1877 	list = mfc->mfc_bw_meter;
1878 	mfc->mfc_bw_meter = NULL;
1879 	free_bw_list(list);
1880 	MFC_UNLOCK();
1881 	return 0;
1882     } else {			/* Delete a single bw_meter entry */
1883 	struct bw_meter *prev;
1884 	uint32_t flags = 0;
1885 
1886 	flags = compute_bw_meter_flags(req);
1887 
1888 	/* Find the bw_meter entry to delete */
1889 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1890 	     prev = x, x = x->bm_mfc_next) {
1891 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1892 			       &req->bu_threshold.b_time, ==)) &&
1893 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1894 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1895 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1896 		break;
1897 	}
1898 	if (x != NULL) { /* Delete entry from the list for this MFC */
1899 	    if (prev != NULL)
1900 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1901 	    else
1902 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1903 
1904 	    unschedule_bw_meter(x);
1905 	    MFC_UNLOCK();
1906 	    /* Free the bw_meter entry */
1907 	    free(x, M_BWMETER);
1908 	    return 0;
1909 	} else {
1910 	    MFC_UNLOCK();
1911 	    return EINVAL;
1912 	}
1913     }
1914     /* NOTREACHED */
1915 }
1916 
1917 /*
1918  * Perform bandwidth measurement processing that may result in an upcall
1919  */
1920 static void
1921 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1922 {
1923     struct timeval delta;
1924 
1925     MFC_LOCK_ASSERT();
1926 
1927     delta = *nowp;
1928     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1929 
1930     if (x->bm_flags & BW_METER_GEQ) {
1931 	/*
1932 	 * Processing for ">=" type of bw_meter entry
1933 	 */
1934 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1935 	    /* Reset the bw_meter entry */
1936 	    x->bm_start_time = *nowp;
1937 	    x->bm_measured.b_packets = 0;
1938 	    x->bm_measured.b_bytes = 0;
1939 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1940 	}
1941 
1942 	/* Record that a packet is received */
1943 	x->bm_measured.b_packets++;
1944 	x->bm_measured.b_bytes += plen;
1945 
1946 	/*
1947 	 * Test if we should deliver an upcall
1948 	 */
1949 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1950 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1951 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1952 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1953 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1954 		/* Prepare an upcall for delivery */
1955 		bw_meter_prepare_upcall(x, nowp);
1956 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1957 	    }
1958 	}
1959     } else if (x->bm_flags & BW_METER_LEQ) {
1960 	/*
1961 	 * Processing for "<=" type of bw_meter entry
1962 	 */
1963 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1964 	    /*
1965 	     * We are behind time with the multicast forwarding table
1966 	     * scanning for "<=" type of bw_meter entries, so test now
1967 	     * if we should deliver an upcall.
1968 	     */
1969 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1970 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1971 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1972 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1973 		/* Prepare an upcall for delivery */
1974 		bw_meter_prepare_upcall(x, nowp);
1975 	    }
1976 	    /* Reschedule the bw_meter entry */
1977 	    unschedule_bw_meter(x);
1978 	    schedule_bw_meter(x, nowp);
1979 	}
1980 
1981 	/* Record that a packet is received */
1982 	x->bm_measured.b_packets++;
1983 	x->bm_measured.b_bytes += plen;
1984 
1985 	/*
1986 	 * Test if we should restart the measuring interval
1987 	 */
1988 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1989 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1990 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1991 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1992 	    /* Don't restart the measuring interval */
1993 	} else {
1994 	    /* Do restart the measuring interval */
1995 	    /*
1996 	     * XXX: note that we don't unschedule and schedule, because this
1997 	     * might be too much overhead per packet. Instead, when we process
1998 	     * all entries for a given timer hash bin, we check whether it is
1999 	     * really a timeout. If not, we reschedule at that time.
2000 	     */
2001 	    x->bm_start_time = *nowp;
2002 	    x->bm_measured.b_packets = 0;
2003 	    x->bm_measured.b_bytes = 0;
2004 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2005 	}
2006     }
2007 }
2008 
2009 /*
2010  * Prepare a bandwidth-related upcall
2011  */
2012 static void
2013 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2014 {
2015     struct timeval delta;
2016     struct bw_upcall *u;
2017 
2018     MFC_LOCK_ASSERT();
2019 
2020     /*
2021      * Compute the measured time interval
2022      */
2023     delta = *nowp;
2024     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2025 
2026     /*
2027      * If there are too many pending upcalls, deliver them now
2028      */
2029     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2030 	bw_upcalls_send();
2031 
2032     /*
2033      * Set the bw_upcall entry
2034      */
2035     u = &V_bw_upcalls[V_bw_upcalls_n++];
2036     u->bu_src = x->bm_mfc->mfc_origin;
2037     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2038     u->bu_threshold.b_time = x->bm_threshold.b_time;
2039     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2040     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2041     u->bu_measured.b_time = delta;
2042     u->bu_measured.b_packets = x->bm_measured.b_packets;
2043     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2044     u->bu_flags = 0;
2045     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2046 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2047     if (x->bm_flags & BW_METER_UNIT_BYTES)
2048 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2049     if (x->bm_flags & BW_METER_GEQ)
2050 	u->bu_flags |= BW_UPCALL_GEQ;
2051     if (x->bm_flags & BW_METER_LEQ)
2052 	u->bu_flags |= BW_UPCALL_LEQ;
2053 }
2054 
2055 /*
2056  * Send the pending bandwidth-related upcalls
2057  */
2058 static void
2059 bw_upcalls_send(void)
2060 {
2061     struct mbuf *m;
2062     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2063     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2064     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2065 				      0,		/* unused2 */
2066 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2067 				      0,		/* im_mbz  */
2068 				      0,		/* im_vif  */
2069 				      0,		/* unused3 */
2070 				      { 0 },		/* im_src  */
2071 				      { 0 } };		/* im_dst  */
2072 
2073     MFC_LOCK_ASSERT();
2074 
2075     if (V_bw_upcalls_n == 0)
2076 	return;			/* No pending upcalls */
2077 
2078     V_bw_upcalls_n = 0;
2079 
2080     /*
2081      * Allocate a new mbuf, initialize it with the header and
2082      * the payload for the pending calls.
2083      */
2084     m = m_gethdr(M_NOWAIT, MT_DATA);
2085     if (m == NULL) {
2086 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2087 	return;
2088     }
2089 
2090     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2091     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2092 
2093     /*
2094      * Send the upcalls
2095      * XXX do we need to set the address in k_igmpsrc ?
2096      */
2097     MRTSTAT_INC(mrts_upcalls);
2098     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2099 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2100 	MRTSTAT_INC(mrts_upq_sockfull);
2101     }
2102 }
2103 
2104 /*
2105  * Compute the timeout hash value for the bw_meter entries
2106  */
2107 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2108     do {								\
2109 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2110 									\
2111 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2112 	(hash) = next_timeval.tv_sec;					\
2113 	if (next_timeval.tv_usec)					\
2114 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2115 	(hash) %= BW_METER_BUCKETS;					\
2116     } while (0)
2117 
2118 /*
2119  * Schedule a timer to process periodically bw_meter entry of type "<="
2120  * by linking the entry in the proper hash bucket.
2121  */
2122 static void
2123 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2124 {
2125     int time_hash;
2126 
2127     MFC_LOCK_ASSERT();
2128 
2129     if (!(x->bm_flags & BW_METER_LEQ))
2130 	return;		/* XXX: we schedule timers only for "<=" entries */
2131 
2132     /*
2133      * Reset the bw_meter entry
2134      */
2135     x->bm_start_time = *nowp;
2136     x->bm_measured.b_packets = 0;
2137     x->bm_measured.b_bytes = 0;
2138     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2139 
2140     /*
2141      * Compute the timeout hash value and insert the entry
2142      */
2143     BW_METER_TIMEHASH(x, time_hash);
2144     x->bm_time_next = V_bw_meter_timers[time_hash];
2145     V_bw_meter_timers[time_hash] = x;
2146     x->bm_time_hash = time_hash;
2147 }
2148 
2149 /*
2150  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2151  * by removing the entry from the proper hash bucket.
2152  */
2153 static void
2154 unschedule_bw_meter(struct bw_meter *x)
2155 {
2156     int time_hash;
2157     struct bw_meter *prev, *tmp;
2158 
2159     MFC_LOCK_ASSERT();
2160 
2161     if (!(x->bm_flags & BW_METER_LEQ))
2162 	return;		/* XXX: we schedule timers only for "<=" entries */
2163 
2164     /*
2165      * Compute the timeout hash value and delete the entry
2166      */
2167     time_hash = x->bm_time_hash;
2168     if (time_hash >= BW_METER_BUCKETS)
2169 	return;		/* Entry was not scheduled */
2170 
2171     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2172 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2173 	if (tmp == x)
2174 	    break;
2175 
2176     if (tmp == NULL)
2177 	panic("unschedule_bw_meter: bw_meter entry not found");
2178 
2179     if (prev != NULL)
2180 	prev->bm_time_next = x->bm_time_next;
2181     else
2182 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2183 
2184     x->bm_time_next = NULL;
2185     x->bm_time_hash = BW_METER_BUCKETS;
2186 }
2187 
2188 
2189 /*
2190  * Process all "<=" type of bw_meter that should be processed now,
2191  * and for each entry prepare an upcall if necessary. Each processed
2192  * entry is rescheduled again for the (periodic) processing.
2193  *
2194  * This is run periodically (once per second normally). On each round,
2195  * all the potentially matching entries are in the hash slot that we are
2196  * looking at.
2197  */
2198 static void
2199 bw_meter_process()
2200 {
2201     uint32_t loops;
2202     int i;
2203     struct timeval now, process_endtime;
2204 
2205     microtime(&now);
2206     if (V_last_tv_sec == now.tv_sec)
2207 	return;		/* nothing to do */
2208 
2209     loops = now.tv_sec - V_last_tv_sec;
2210     V_last_tv_sec = now.tv_sec;
2211     if (loops > BW_METER_BUCKETS)
2212 	loops = BW_METER_BUCKETS;
2213 
2214     MFC_LOCK();
2215     /*
2216      * Process all bins of bw_meter entries from the one after the last
2217      * processed to the current one. On entry, i points to the last bucket
2218      * visited, so we need to increment i at the beginning of the loop.
2219      */
2220     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2221 	struct bw_meter *x, *tmp_list;
2222 
2223 	if (++i >= BW_METER_BUCKETS)
2224 	    i = 0;
2225 
2226 	/* Disconnect the list of bw_meter entries from the bin */
2227 	tmp_list = V_bw_meter_timers[i];
2228 	V_bw_meter_timers[i] = NULL;
2229 
2230 	/* Process the list of bw_meter entries */
2231 	while (tmp_list != NULL) {
2232 	    x = tmp_list;
2233 	    tmp_list = tmp_list->bm_time_next;
2234 
2235 	    /* Test if the time interval is over */
2236 	    process_endtime = x->bm_start_time;
2237 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2238 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2239 		/* Not yet: reschedule, but don't reset */
2240 		int time_hash;
2241 
2242 		BW_METER_TIMEHASH(x, time_hash);
2243 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2244 		    /*
2245 		     * XXX: somehow the bin processing is a bit ahead of time.
2246 		     * Put the entry in the next bin.
2247 		     */
2248 		    if (++time_hash >= BW_METER_BUCKETS)
2249 			time_hash = 0;
2250 		}
2251 		x->bm_time_next = V_bw_meter_timers[time_hash];
2252 		V_bw_meter_timers[time_hash] = x;
2253 		x->bm_time_hash = time_hash;
2254 
2255 		continue;
2256 	    }
2257 
2258 	    /*
2259 	     * Test if we should deliver an upcall
2260 	     */
2261 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2262 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2263 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2264 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2265 		/* Prepare an upcall for delivery */
2266 		bw_meter_prepare_upcall(x, &now);
2267 	    }
2268 
2269 	    /*
2270 	     * Reschedule for next processing
2271 	     */
2272 	    schedule_bw_meter(x, &now);
2273 	}
2274     }
2275 
2276     /* Send all upcalls that are pending delivery */
2277     bw_upcalls_send();
2278 
2279     MFC_UNLOCK();
2280 }
2281 
2282 /*
2283  * A periodic function for sending all upcalls that are pending delivery
2284  */
2285 static void
2286 expire_bw_upcalls_send(void *arg)
2287 {
2288     CURVNET_SET((struct vnet *) arg);
2289 
2290     MFC_LOCK();
2291     bw_upcalls_send();
2292     MFC_UNLOCK();
2293 
2294     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2295 	curvnet);
2296     CURVNET_RESTORE();
2297 }
2298 
2299 /*
2300  * A periodic function for periodic scanning of the multicast forwarding
2301  * table for processing all "<=" bw_meter entries.
2302  */
2303 static void
2304 expire_bw_meter_process(void *arg)
2305 {
2306     CURVNET_SET((struct vnet *) arg);
2307 
2308     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2309 	bw_meter_process();
2310 
2311     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2312 	curvnet);
2313     CURVNET_RESTORE();
2314 }
2315 
2316 /*
2317  * End of bandwidth monitoring code
2318  */
2319 
2320 /*
2321  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2322  *
2323  */
2324 static int
2325 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2326     struct mfc *rt)
2327 {
2328     struct mbuf *mb_copy, *mm;
2329 
2330     /*
2331      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2332      * rendezvous point was unspecified, and we were told not to.
2333      */
2334     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2335 	in_nullhost(rt->mfc_rp))
2336 	return 0;
2337 
2338     mb_copy = pim_register_prepare(ip, m);
2339     if (mb_copy == NULL)
2340 	return ENOBUFS;
2341 
2342     /*
2343      * Send all the fragments. Note that the mbuf for each fragment
2344      * is freed by the sending machinery.
2345      */
2346     for (mm = mb_copy; mm; mm = mb_copy) {
2347 	mb_copy = mm->m_nextpkt;
2348 	mm->m_nextpkt = 0;
2349 	mm = m_pullup(mm, sizeof(struct ip));
2350 	if (mm != NULL) {
2351 	    ip = mtod(mm, struct ip *);
2352 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2353 		pim_register_send_rp(ip, vifp, mm, rt);
2354 	    } else {
2355 		pim_register_send_upcall(ip, vifp, mm, rt);
2356 	    }
2357 	}
2358     }
2359 
2360     return 0;
2361 }
2362 
2363 /*
2364  * Return a copy of the data packet that is ready for PIM Register
2365  * encapsulation.
2366  * XXX: Note that in the returned copy the IP header is a valid one.
2367  */
2368 static struct mbuf *
2369 pim_register_prepare(struct ip *ip, struct mbuf *m)
2370 {
2371     struct mbuf *mb_copy = NULL;
2372     int mtu;
2373 
2374     /* Take care of delayed checksums */
2375     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2376 	in_delayed_cksum(m);
2377 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2378     }
2379 
2380     /*
2381      * Copy the old packet & pullup its IP header into the
2382      * new mbuf so we can modify it.
2383      */
2384     mb_copy = m_copypacket(m, M_NOWAIT);
2385     if (mb_copy == NULL)
2386 	return NULL;
2387     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2388     if (mb_copy == NULL)
2389 	return NULL;
2390 
2391     /* take care of the TTL */
2392     ip = mtod(mb_copy, struct ip *);
2393     --ip->ip_ttl;
2394 
2395     /* Compute the MTU after the PIM Register encapsulation */
2396     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2397 
2398     if (ntohs(ip->ip_len) <= mtu) {
2399 	/* Turn the IP header into a valid one */
2400 	ip->ip_sum = 0;
2401 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2402     } else {
2403 	/* Fragment the packet */
2404 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2405 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2406 	    m_freem(mb_copy);
2407 	    return NULL;
2408 	}
2409     }
2410     return mb_copy;
2411 }
2412 
2413 /*
2414  * Send an upcall with the data packet to the user-level process.
2415  */
2416 static int
2417 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2418     struct mbuf *mb_copy, struct mfc *rt)
2419 {
2420     struct mbuf *mb_first;
2421     int len = ntohs(ip->ip_len);
2422     struct igmpmsg *im;
2423     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2424 
2425     VIF_LOCK_ASSERT();
2426 
2427     /*
2428      * Add a new mbuf with an upcall header
2429      */
2430     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2431     if (mb_first == NULL) {
2432 	m_freem(mb_copy);
2433 	return ENOBUFS;
2434     }
2435     mb_first->m_data += max_linkhdr;
2436     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2437     mb_first->m_len = sizeof(struct igmpmsg);
2438     mb_first->m_next = mb_copy;
2439 
2440     /* Send message to routing daemon */
2441     im = mtod(mb_first, struct igmpmsg *);
2442     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2443     im->im_mbz		= 0;
2444     im->im_vif		= vifp - V_viftable;
2445     im->im_src		= ip->ip_src;
2446     im->im_dst		= ip->ip_dst;
2447 
2448     k_igmpsrc.sin_addr	= ip->ip_src;
2449 
2450     MRTSTAT_INC(mrts_upcalls);
2451 
2452     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2453 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2454 	MRTSTAT_INC(mrts_upq_sockfull);
2455 	return ENOBUFS;
2456     }
2457 
2458     /* Keep statistics */
2459     PIMSTAT_INC(pims_snd_registers_msgs);
2460     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2461 
2462     return 0;
2463 }
2464 
2465 /*
2466  * Encapsulate the data packet in PIM Register message and send it to the RP.
2467  */
2468 static int
2469 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2470     struct mfc *rt)
2471 {
2472     struct mbuf *mb_first;
2473     struct ip *ip_outer;
2474     struct pim_encap_pimhdr *pimhdr;
2475     int len = ntohs(ip->ip_len);
2476     vifi_t vifi = rt->mfc_parent;
2477 
2478     VIF_LOCK_ASSERT();
2479 
2480     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2481 	m_freem(mb_copy);
2482 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2483     }
2484 
2485     /*
2486      * Add a new mbuf with the encapsulating header
2487      */
2488     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2489     if (mb_first == NULL) {
2490 	m_freem(mb_copy);
2491 	return ENOBUFS;
2492     }
2493     mb_first->m_data += max_linkhdr;
2494     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2495     mb_first->m_next = mb_copy;
2496 
2497     mb_first->m_pkthdr.len = len + mb_first->m_len;
2498 
2499     /*
2500      * Fill in the encapsulating IP and PIM header
2501      */
2502     ip_outer = mtod(mb_first, struct ip *);
2503     *ip_outer = pim_encap_iphdr;
2504     ip_outer->ip_id = ip_newid();
2505     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2506 	sizeof(pim_encap_pimhdr));
2507     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2508     ip_outer->ip_dst = rt->mfc_rp;
2509     /*
2510      * Copy the inner header TOS to the outer header, and take care of the
2511      * IP_DF bit.
2512      */
2513     ip_outer->ip_tos = ip->ip_tos;
2514     if (ip->ip_off & htons(IP_DF))
2515 	ip_outer->ip_off |= htons(IP_DF);
2516     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2517 					 + sizeof(pim_encap_iphdr));
2518     *pimhdr = pim_encap_pimhdr;
2519     /* If the iif crosses a border, set the Border-bit */
2520     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2521 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2522 
2523     mb_first->m_data += sizeof(pim_encap_iphdr);
2524     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2525     mb_first->m_data -= sizeof(pim_encap_iphdr);
2526 
2527     send_packet(vifp, mb_first);
2528 
2529     /* Keep statistics */
2530     PIMSTAT_INC(pims_snd_registers_msgs);
2531     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2532 
2533     return 0;
2534 }
2535 
2536 /*
2537  * pim_encapcheck() is called by the encap4_input() path at runtime to
2538  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2539  * into the kernel.
2540  */
2541 static int
2542 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2543 {
2544 
2545 #ifdef DIAGNOSTIC
2546     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2547 #endif
2548     if (proto != IPPROTO_PIM)
2549 	return 0;	/* not for us; reject the datagram. */
2550 
2551     return 64;		/* claim the datagram. */
2552 }
2553 
2554 /*
2555  * PIM-SMv2 and PIM-DM messages processing.
2556  * Receives and verifies the PIM control messages, and passes them
2557  * up to the listening socket, using rip_input().
2558  * The only message with special processing is the PIM_REGISTER message
2559  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2560  * is passed to if_simloop().
2561  */
2562 int
2563 pim_input(struct mbuf **mp, int *offp, int proto)
2564 {
2565     struct mbuf *m = *mp;
2566     struct ip *ip = mtod(m, struct ip *);
2567     struct pim *pim;
2568     int iphlen = *offp;
2569     int minlen;
2570     int datalen = ntohs(ip->ip_len) - iphlen;
2571     int ip_tos;
2572 
2573     *mp = NULL;
2574 
2575     /* Keep statistics */
2576     PIMSTAT_INC(pims_rcv_total_msgs);
2577     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2578 
2579     /*
2580      * Validate lengths
2581      */
2582     if (datalen < PIM_MINLEN) {
2583 	PIMSTAT_INC(pims_rcv_tooshort);
2584 	CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2585 	    __func__, datalen, inet_ntoa(ip->ip_src));
2586 	m_freem(m);
2587 	return (IPPROTO_DONE);
2588     }
2589 
2590     /*
2591      * If the packet is at least as big as a REGISTER, go agead
2592      * and grab the PIM REGISTER header size, to avoid another
2593      * possible m_pullup() later.
2594      *
2595      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2596      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2597      */
2598     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2599     /*
2600      * Get the IP and PIM headers in contiguous memory, and
2601      * possibly the PIM REGISTER header.
2602      */
2603     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == 0) {
2604 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2605 	return (IPPROTO_DONE);
2606     }
2607 
2608     /* m_pullup() may have given us a new mbuf so reset ip. */
2609     ip = mtod(m, struct ip *);
2610     ip_tos = ip->ip_tos;
2611 
2612     /* adjust mbuf to point to the PIM header */
2613     m->m_data += iphlen;
2614     m->m_len  -= iphlen;
2615     pim = mtod(m, struct pim *);
2616 
2617     /*
2618      * Validate checksum. If PIM REGISTER, exclude the data packet.
2619      *
2620      * XXX: some older PIMv2 implementations don't make this distinction,
2621      * so for compatibility reason perform the checksum over part of the
2622      * message, and if error, then over the whole message.
2623      */
2624     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2625 	/* do nothing, checksum okay */
2626     } else if (in_cksum(m, datalen)) {
2627 	PIMSTAT_INC(pims_rcv_badsum);
2628 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2629 	m_freem(m);
2630 	return (IPPROTO_DONE);
2631     }
2632 
2633     /* PIM version check */
2634     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2635 	PIMSTAT_INC(pims_rcv_badversion);
2636 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2637 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2638 	m_freem(m);
2639 	return (IPPROTO_DONE);
2640     }
2641 
2642     /* restore mbuf back to the outer IP */
2643     m->m_data -= iphlen;
2644     m->m_len  += iphlen;
2645 
2646     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2647 	/*
2648 	 * Since this is a REGISTER, we'll make a copy of the register
2649 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2650 	 * routing daemon.
2651 	 */
2652 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2653 	struct mbuf *mcp;
2654 	struct ip *encap_ip;
2655 	u_int32_t *reghdr;
2656 	struct ifnet *vifp;
2657 
2658 	VIF_LOCK();
2659 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2660 	    VIF_UNLOCK();
2661 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2662 		(int)V_reg_vif_num);
2663 	    m_freem(m);
2664 	    return (IPPROTO_DONE);
2665 	}
2666 	/* XXX need refcnt? */
2667 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2668 	VIF_UNLOCK();
2669 
2670 	/*
2671 	 * Validate length
2672 	 */
2673 	if (datalen < PIM_REG_MINLEN) {
2674 	    PIMSTAT_INC(pims_rcv_tooshort);
2675 	    PIMSTAT_INC(pims_rcv_badregisters);
2676 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2677 	    m_freem(m);
2678 	    return (IPPROTO_DONE);
2679 	}
2680 
2681 	reghdr = (u_int32_t *)(pim + 1);
2682 	encap_ip = (struct ip *)(reghdr + 1);
2683 
2684 	CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2685 	    __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2686 
2687 	/* verify the version number of the inner packet */
2688 	if (encap_ip->ip_v != IPVERSION) {
2689 	    PIMSTAT_INC(pims_rcv_badregisters);
2690 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2691 	    m_freem(m);
2692 	    return (IPPROTO_DONE);
2693 	}
2694 
2695 	/* verify the inner packet is destined to a mcast group */
2696 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2697 	    PIMSTAT_INC(pims_rcv_badregisters);
2698 	    CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2699 		inet_ntoa(encap_ip->ip_dst));
2700 	    m_freem(m);
2701 	    return (IPPROTO_DONE);
2702 	}
2703 
2704 	/* If a NULL_REGISTER, pass it to the daemon */
2705 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2706 	    goto pim_input_to_daemon;
2707 
2708 	/*
2709 	 * Copy the TOS from the outer IP header to the inner IP header.
2710 	 */
2711 	if (encap_ip->ip_tos != ip_tos) {
2712 	    /* Outer TOS -> inner TOS */
2713 	    encap_ip->ip_tos = ip_tos;
2714 	    /* Recompute the inner header checksum. Sigh... */
2715 
2716 	    /* adjust mbuf to point to the inner IP header */
2717 	    m->m_data += (iphlen + PIM_MINLEN);
2718 	    m->m_len  -= (iphlen + PIM_MINLEN);
2719 
2720 	    encap_ip->ip_sum = 0;
2721 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2722 
2723 	    /* restore mbuf to point back to the outer IP header */
2724 	    m->m_data -= (iphlen + PIM_MINLEN);
2725 	    m->m_len  += (iphlen + PIM_MINLEN);
2726 	}
2727 
2728 	/*
2729 	 * Decapsulate the inner IP packet and loopback to forward it
2730 	 * as a normal multicast packet. Also, make a copy of the
2731 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2732 	 * to pass to the daemon later, so it can take the appropriate
2733 	 * actions (e.g., send back PIM_REGISTER_STOP).
2734 	 * XXX: here m->m_data points to the outer IP header.
2735 	 */
2736 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2737 	if (mcp == NULL) {
2738 	    CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2739 	    m_freem(m);
2740 	    return (IPPROTO_DONE);
2741 	}
2742 
2743 	/* Keep statistics */
2744 	/* XXX: registers_bytes include only the encap. mcast pkt */
2745 	PIMSTAT_INC(pims_rcv_registers_msgs);
2746 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2747 
2748 	/*
2749 	 * forward the inner ip packet; point m_data at the inner ip.
2750 	 */
2751 	m_adj(m, iphlen + PIM_MINLEN);
2752 
2753 	CTR4(KTR_IPMF,
2754 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2755 	    __func__,
2756 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2757 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2758 	    (int)V_reg_vif_num);
2759 
2760 	/* NB: vifp was collected above; can it change on us? */
2761 	if_simloop(vifp, m, dst.sin_family, 0);
2762 
2763 	/* prepare the register head to send to the mrouting daemon */
2764 	m = mcp;
2765     }
2766 
2767 pim_input_to_daemon:
2768     /*
2769      * Pass the PIM message up to the daemon; if it is a Register message,
2770      * pass the 'head' only up to the daemon. This includes the
2771      * outer IP header, PIM header, PIM-Register header and the
2772      * inner IP header.
2773      * XXX: the outer IP header pkt size of a Register is not adjust to
2774      * reflect the fact that the inner multicast data is truncated.
2775      */
2776     *mp = m;
2777     rip_input(mp, offp, proto);
2778 
2779     return (IPPROTO_DONE);
2780 }
2781 
2782 static int
2783 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2784 {
2785 	struct mfc	*rt;
2786 	int		 error, i;
2787 
2788 	if (req->newptr)
2789 		return (EPERM);
2790 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2791 		return (0);
2792 	error = sysctl_wire_old_buffer(req, 0);
2793 	if (error)
2794 		return (error);
2795 
2796 	MFC_LOCK();
2797 	for (i = 0; i < mfchashsize; i++) {
2798 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2799 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2800 			if (error)
2801 				goto out_locked;
2802 		}
2803 	}
2804 out_locked:
2805 	MFC_UNLOCK();
2806 	return (error);
2807 }
2808 
2809 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2810     sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2811     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2812 
2813 static void
2814 vnet_mroute_init(const void *unused __unused)
2815 {
2816 
2817 	MALLOC(V_nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2818 	bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2819 	callout_init(&V_expire_upcalls_ch, CALLOUT_MPSAFE);
2820 	callout_init(&V_bw_upcalls_ch, CALLOUT_MPSAFE);
2821 	callout_init(&V_bw_meter_ch, CALLOUT_MPSAFE);
2822 }
2823 
2824 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mroute_init,
2825 	NULL);
2826 
2827 static void
2828 vnet_mroute_uninit(const void *unused __unused)
2829 {
2830 
2831 	FREE(V_nexpire, M_MRTABLE);
2832 	V_nexpire = NULL;
2833 }
2834 
2835 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE,
2836 	vnet_mroute_uninit, NULL);
2837 
2838 static int
2839 ip_mroute_modevent(module_t mod, int type, void *unused)
2840 {
2841 
2842     switch (type) {
2843     case MOD_LOAD:
2844 	MROUTER_LOCK_INIT();
2845 
2846 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2847 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2848 	if (if_detach_event_tag == NULL) {
2849 		printf("ip_mroute: unable to register "
2850 		    "ifnet_departure_event handler\n");
2851 		MROUTER_LOCK_DESTROY();
2852 		return (EINVAL);
2853 	}
2854 
2855 	MFC_LOCK_INIT();
2856 	VIF_LOCK_INIT();
2857 
2858 	mfchashsize = MFCHASHSIZE;
2859 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2860 	    !powerof2(mfchashsize)) {
2861 		printf("WARNING: %s not a power of 2; using default\n",
2862 		    "net.inet.ip.mfchashsize");
2863 		mfchashsize = MFCHASHSIZE;
2864 	}
2865 
2866 	pim_squelch_wholepkt = 0;
2867 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2868 	    &pim_squelch_wholepkt);
2869 
2870 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2871 	    pim_encapcheck, &in_pim_protosw, NULL);
2872 	if (pim_encap_cookie == NULL) {
2873 		printf("ip_mroute: unable to attach pim encap\n");
2874 		VIF_LOCK_DESTROY();
2875 		MFC_LOCK_DESTROY();
2876 		MROUTER_LOCK_DESTROY();
2877 		return (EINVAL);
2878 	}
2879 
2880 	ip_mcast_src = X_ip_mcast_src;
2881 	ip_mforward = X_ip_mforward;
2882 	ip_mrouter_done = X_ip_mrouter_done;
2883 	ip_mrouter_get = X_ip_mrouter_get;
2884 	ip_mrouter_set = X_ip_mrouter_set;
2885 
2886 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2887 	ip_rsvp_vif = X_ip_rsvp_vif;
2888 
2889 	legal_vif_num = X_legal_vif_num;
2890 	mrt_ioctl = X_mrt_ioctl;
2891 	rsvp_input_p = X_rsvp_input;
2892 	break;
2893 
2894     case MOD_UNLOAD:
2895 	/*
2896 	 * Typically module unload happens after the user-level
2897 	 * process has shutdown the kernel services (the check
2898 	 * below insures someone can't just yank the module out
2899 	 * from under a running process).  But if the module is
2900 	 * just loaded and then unloaded w/o starting up a user
2901 	 * process we still need to cleanup.
2902 	 */
2903 	MROUTER_LOCK();
2904 	if (ip_mrouter_cnt != 0) {
2905 	    MROUTER_UNLOCK();
2906 	    return (EINVAL);
2907 	}
2908 	ip_mrouter_unloading = 1;
2909 	MROUTER_UNLOCK();
2910 
2911 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2912 
2913 	if (pim_encap_cookie) {
2914 	    encap_detach(pim_encap_cookie);
2915 	    pim_encap_cookie = NULL;
2916 	}
2917 
2918 	ip_mcast_src = NULL;
2919 	ip_mforward = NULL;
2920 	ip_mrouter_done = NULL;
2921 	ip_mrouter_get = NULL;
2922 	ip_mrouter_set = NULL;
2923 
2924 	ip_rsvp_force_done = NULL;
2925 	ip_rsvp_vif = NULL;
2926 
2927 	legal_vif_num = NULL;
2928 	mrt_ioctl = NULL;
2929 	rsvp_input_p = NULL;
2930 
2931 	VIF_LOCK_DESTROY();
2932 	MFC_LOCK_DESTROY();
2933 	MROUTER_LOCK_DESTROY();
2934 	break;
2935 
2936     default:
2937 	return EOPNOTSUPP;
2938     }
2939     return 0;
2940 }
2941 
2942 static moduledata_t ip_mroutemod = {
2943     "ip_mroute",
2944     ip_mroute_modevent,
2945     0
2946 };
2947 
2948 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_MIDDLE);
2949