xref: /freebsd/sys/netinet/ip_mroute.c (revision 74d9553e43cfafc29448d0bb836916aa21dea0de)
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/eventhandler.h>
81 #include <sys/lock.h>
82 #include <sys/ktr.h>
83 #include <sys/malloc.h>
84 #include <sys/mbuf.h>
85 #include <sys/module.h>
86 #include <sys/priv.h>
87 #include <sys/protosw.h>
88 #include <sys/signalvar.h>
89 #include <sys/socket.h>
90 #include <sys/socketvar.h>
91 #include <sys/sockio.h>
92 #include <sys/sx.h>
93 #include <sys/sysctl.h>
94 #include <sys/syslog.h>
95 #include <sys/systm.h>
96 #include <sys/time.h>
97 #include <sys/counter.h>
98 
99 #include <net/if.h>
100 #include <net/if_var.h>
101 #include <net/netisr.h>
102 #include <net/route.h>
103 #include <net/vnet.h>
104 
105 #include <netinet/in.h>
106 #include <netinet/igmp.h>
107 #include <netinet/in_systm.h>
108 #include <netinet/in_var.h>
109 #include <netinet/ip.h>
110 #include <netinet/ip_encap.h>
111 #include <netinet/ip_mroute.h>
112 #include <netinet/ip_var.h>
113 #include <netinet/ip_options.h>
114 #include <netinet/pim.h>
115 #include <netinet/pim_var.h>
116 #include <netinet/udp.h>
117 
118 #include <machine/in_cksum.h>
119 
120 #ifndef KTR_IPMF
121 #define KTR_IPMF KTR_INET
122 #endif
123 
124 #define		VIFI_INVALID	((vifi_t) -1)
125 
126 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
127 #define	V_last_tv_sec	VNET(last_tv_sec)
128 
129 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
130 
131 /*
132  * Locking.  We use two locks: one for the virtual interface table and
133  * one for the forwarding table.  These locks may be nested in which case
134  * the VIF lock must always be taken first.  Note that each lock is used
135  * to cover not only the specific data structure but also related data
136  * structures.
137  */
138 
139 static struct mtx mrouter_mtx;
140 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
141 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
142 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
143 #define	MROUTER_LOCK_INIT()						\
144 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
145 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
146 
147 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
148 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
149 
150 static VNET_PCPUSTAT_DEFINE(struct mrtstat, mrtstat);
151 VNET_PCPUSTAT_SYSINIT(mrtstat);
152 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
153 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
154     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
155     "netinet/ip_mroute.h)");
156 
157 static VNET_DEFINE(u_long, mfchash);
158 #define	V_mfchash		VNET(mfchash)
159 #define	MFCHASH(a, g)							\
160 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
161 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
162 #define	MFCHASHSIZE	256
163 
164 static u_long mfchashsize;			/* Hash size */
165 static VNET_DEFINE(u_char *, nexpire);		/* 0..mfchashsize-1 */
166 #define	V_nexpire		VNET(nexpire)
167 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
168 #define	V_mfchashtbl		VNET(mfchashtbl)
169 
170 static struct mtx mfc_mtx;
171 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
172 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
173 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
174 #define	MFC_LOCK_INIT()							\
175 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
176 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
177 
178 static VNET_DEFINE(vifi_t, numvifs);
179 #define	V_numvifs		VNET(numvifs)
180 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
181 #define	V_viftable		VNET(viftable)
182 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_VNET | CTLFLAG_RD,
183     &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
184     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
185 
186 static struct mtx vif_mtx;
187 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
188 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
189 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
190 #define	VIF_LOCK_INIT()							\
191 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
192 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
193 
194 static eventhandler_tag if_detach_event_tag = NULL;
195 
196 static VNET_DEFINE(struct callout, expire_upcalls_ch);
197 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
198 
199 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
200 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
201 
202 /*
203  * Bandwidth meter variables and constants
204  */
205 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
206 /*
207  * Pending timeouts are stored in a hash table, the key being the
208  * expiration time. Periodically, the entries are analysed and processed.
209  */
210 #define	BW_METER_BUCKETS	1024
211 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
212 #define	V_bw_meter_timers	VNET(bw_meter_timers)
213 static VNET_DEFINE(struct callout, bw_meter_ch);
214 #define	V_bw_meter_ch		VNET(bw_meter_ch)
215 #define	BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
216 
217 /*
218  * Pending upcalls are stored in a vector which is flushed when
219  * full, or periodically
220  */
221 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
222 #define	V_bw_upcalls		VNET(bw_upcalls)
223 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
224 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
225 static VNET_DEFINE(struct callout, bw_upcalls_ch);
226 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
227 
228 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
229 
230 static VNET_PCPUSTAT_DEFINE(struct pimstat, pimstat);
231 VNET_PCPUSTAT_SYSINIT(pimstat);
232 VNET_PCPUSTAT_SYSUNINIT(pimstat);
233 
234 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
235 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
236     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
237 
238 static u_long	pim_squelch_wholepkt = 0;
239 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
240     &pim_squelch_wholepkt, 0,
241     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
242 
243 extern  struct domain inetdomain;
244 static const struct protosw in_pim_protosw = {
245 	.pr_type =		SOCK_RAW,
246 	.pr_domain =		&inetdomain,
247 	.pr_protocol =		IPPROTO_PIM,
248 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
249 	.pr_input =		pim_input,
250 	.pr_output =		rip_output,
251 	.pr_ctloutput =		rip_ctloutput,
252 	.pr_usrreqs =		&rip_usrreqs
253 };
254 static const struct encaptab *pim_encap_cookie;
255 
256 static int pim_encapcheck(const struct mbuf *, int, int, void *);
257 
258 /*
259  * Note: the PIM Register encapsulation adds the following in front of a
260  * data packet:
261  *
262  * struct pim_encap_hdr {
263  *    struct ip ip;
264  *    struct pim_encap_pimhdr  pim;
265  * }
266  *
267  */
268 
269 struct pim_encap_pimhdr {
270 	struct pim pim;
271 	uint32_t   flags;
272 };
273 #define		PIM_ENCAP_TTL	64
274 
275 static struct ip pim_encap_iphdr = {
276 #if BYTE_ORDER == LITTLE_ENDIAN
277 	sizeof(struct ip) >> 2,
278 	IPVERSION,
279 #else
280 	IPVERSION,
281 	sizeof(struct ip) >> 2,
282 #endif
283 	0,			/* tos */
284 	sizeof(struct ip),	/* total length */
285 	0,			/* id */
286 	0,			/* frag offset */
287 	PIM_ENCAP_TTL,
288 	IPPROTO_PIM,
289 	0,			/* checksum */
290 };
291 
292 static struct pim_encap_pimhdr pim_encap_pimhdr = {
293     {
294 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
295 	0,			/* reserved */
296 	0,			/* checksum */
297     },
298     0				/* flags */
299 };
300 
301 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
302 #define	V_reg_vif_num		VNET(reg_vif_num)
303 static VNET_DEFINE(struct ifnet, multicast_register_if);
304 #define	V_multicast_register_if	VNET(multicast_register_if)
305 
306 /*
307  * Private variables.
308  */
309 
310 static u_long	X_ip_mcast_src(int);
311 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
312 		    struct ip_moptions *);
313 static int	X_ip_mrouter_done(void);
314 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
315 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
316 static int	X_legal_vif_num(int);
317 static int	X_mrt_ioctl(u_long, caddr_t, int);
318 
319 static int	add_bw_upcall(struct bw_upcall *);
320 static int	add_mfc(struct mfcctl2 *);
321 static int	add_vif(struct vifctl *);
322 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
323 static void	bw_meter_process(void);
324 static void	bw_meter_receive_packet(struct bw_meter *, int,
325 		    struct timeval *);
326 static void	bw_upcalls_send(void);
327 static int	del_bw_upcall(struct bw_upcall *);
328 static int	del_mfc(struct mfcctl2 *);
329 static int	del_vif(vifi_t);
330 static int	del_vif_locked(vifi_t);
331 static void	expire_bw_meter_process(void *);
332 static void	expire_bw_upcalls_send(void *);
333 static void	expire_mfc(struct mfc *);
334 static void	expire_upcalls(void *);
335 static void	free_bw_list(struct bw_meter *);
336 static int	get_sg_cnt(struct sioc_sg_req *);
337 static int	get_vif_cnt(struct sioc_vif_req *);
338 static void	if_detached_event(void *, struct ifnet *);
339 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
340 static int	ip_mrouter_init(struct socket *, int);
341 static __inline struct mfc *
342 		mfc_find(struct in_addr *, struct in_addr *);
343 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
344 static struct mbuf *
345 		pim_register_prepare(struct ip *, struct mbuf *);
346 static int	pim_register_send(struct ip *, struct vif *,
347 		    struct mbuf *, struct mfc *);
348 static int	pim_register_send_rp(struct ip *, struct vif *,
349 		    struct mbuf *, struct mfc *);
350 static int	pim_register_send_upcall(struct ip *, struct vif *,
351 		    struct mbuf *, struct mfc *);
352 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
353 static void	send_packet(struct vif *, struct mbuf *);
354 static int	set_api_config(uint32_t *);
355 static int	set_assert(int);
356 static int	socket_send(struct socket *, struct mbuf *,
357 		    struct sockaddr_in *);
358 static void	unschedule_bw_meter(struct bw_meter *);
359 
360 /*
361  * Kernel multicast forwarding API capabilities and setup.
362  * If more API capabilities are added to the kernel, they should be
363  * recorded in `mrt_api_support'.
364  */
365 #define MRT_API_VERSION		0x0305
366 
367 static const int mrt_api_version = MRT_API_VERSION;
368 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
369 					 MRT_MFC_FLAGS_BORDER_VIF |
370 					 MRT_MFC_RP |
371 					 MRT_MFC_BW_UPCALL);
372 static VNET_DEFINE(uint32_t, mrt_api_config);
373 #define	V_mrt_api_config	VNET(mrt_api_config)
374 static VNET_DEFINE(int, pim_assert_enabled);
375 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
376 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
377 
378 /*
379  * Find a route for a given origin IP address and multicast group address.
380  * Statistics must be updated by the caller.
381  */
382 static __inline struct mfc *
383 mfc_find(struct in_addr *o, struct in_addr *g)
384 {
385 	struct mfc *rt;
386 
387 	MFC_LOCK_ASSERT();
388 
389 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
390 		if (in_hosteq(rt->mfc_origin, *o) &&
391 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
392 		    TAILQ_EMPTY(&rt->mfc_stall))
393 			break;
394 	}
395 
396 	return (rt);
397 }
398 
399 /*
400  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
401  */
402 static int
403 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
404 {
405     int	error, optval;
406     vifi_t	vifi;
407     struct	vifctl vifc;
408     struct	mfcctl2 mfc;
409     struct	bw_upcall bw_upcall;
410     uint32_t	i;
411 
412     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
413 	return EPERM;
414 
415     error = 0;
416     switch (sopt->sopt_name) {
417     case MRT_INIT:
418 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
419 	if (error)
420 	    break;
421 	error = ip_mrouter_init(so, optval);
422 	break;
423 
424     case MRT_DONE:
425 	error = ip_mrouter_done();
426 	break;
427 
428     case MRT_ADD_VIF:
429 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
430 	if (error)
431 	    break;
432 	error = add_vif(&vifc);
433 	break;
434 
435     case MRT_DEL_VIF:
436 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
437 	if (error)
438 	    break;
439 	error = del_vif(vifi);
440 	break;
441 
442     case MRT_ADD_MFC:
443     case MRT_DEL_MFC:
444 	/*
445 	 * select data size depending on API version.
446 	 */
447 	if (sopt->sopt_name == MRT_ADD_MFC &&
448 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
449 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
450 				sizeof(struct mfcctl2));
451 	} else {
452 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
453 				sizeof(struct mfcctl));
454 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
455 			sizeof(mfc) - sizeof(struct mfcctl));
456 	}
457 	if (error)
458 	    break;
459 	if (sopt->sopt_name == MRT_ADD_MFC)
460 	    error = add_mfc(&mfc);
461 	else
462 	    error = del_mfc(&mfc);
463 	break;
464 
465     case MRT_ASSERT:
466 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
467 	if (error)
468 	    break;
469 	set_assert(optval);
470 	break;
471 
472     case MRT_API_CONFIG:
473 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
474 	if (!error)
475 	    error = set_api_config(&i);
476 	if (!error)
477 	    error = sooptcopyout(sopt, &i, sizeof i);
478 	break;
479 
480     case MRT_ADD_BW_UPCALL:
481     case MRT_DEL_BW_UPCALL:
482 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
483 				sizeof bw_upcall);
484 	if (error)
485 	    break;
486 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
487 	    error = add_bw_upcall(&bw_upcall);
488 	else
489 	    error = del_bw_upcall(&bw_upcall);
490 	break;
491 
492     default:
493 	error = EOPNOTSUPP;
494 	break;
495     }
496     return error;
497 }
498 
499 /*
500  * Handle MRT getsockopt commands
501  */
502 static int
503 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
504 {
505     int error;
506 
507     switch (sopt->sopt_name) {
508     case MRT_VERSION:
509 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
510 	break;
511 
512     case MRT_ASSERT:
513 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
514 	    sizeof V_pim_assert_enabled);
515 	break;
516 
517     case MRT_API_SUPPORT:
518 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
519 	break;
520 
521     case MRT_API_CONFIG:
522 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
523 	break;
524 
525     default:
526 	error = EOPNOTSUPP;
527 	break;
528     }
529     return error;
530 }
531 
532 /*
533  * Handle ioctl commands to obtain information from the cache
534  */
535 static int
536 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
537 {
538     int error = 0;
539 
540     /*
541      * Currently the only function calling this ioctl routine is rtioctl_fib().
542      * Typically, only root can create the raw socket in order to execute
543      * this ioctl method, however the request might be coming from a prison
544      */
545     error = priv_check(curthread, PRIV_NETINET_MROUTE);
546     if (error)
547 	return (error);
548     switch (cmd) {
549     case (SIOCGETVIFCNT):
550 	error = get_vif_cnt((struct sioc_vif_req *)data);
551 	break;
552 
553     case (SIOCGETSGCNT):
554 	error = get_sg_cnt((struct sioc_sg_req *)data);
555 	break;
556 
557     default:
558 	error = EINVAL;
559 	break;
560     }
561     return error;
562 }
563 
564 /*
565  * returns the packet, byte, rpf-failure count for the source group provided
566  */
567 static int
568 get_sg_cnt(struct sioc_sg_req *req)
569 {
570     struct mfc *rt;
571 
572     MFC_LOCK();
573     rt = mfc_find(&req->src, &req->grp);
574     if (rt == NULL) {
575 	MFC_UNLOCK();
576 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
577 	return EADDRNOTAVAIL;
578     }
579     req->pktcnt = rt->mfc_pkt_cnt;
580     req->bytecnt = rt->mfc_byte_cnt;
581     req->wrong_if = rt->mfc_wrong_if;
582     MFC_UNLOCK();
583     return 0;
584 }
585 
586 /*
587  * returns the input and output packet and byte counts on the vif provided
588  */
589 static int
590 get_vif_cnt(struct sioc_vif_req *req)
591 {
592     vifi_t vifi = req->vifi;
593 
594     VIF_LOCK();
595     if (vifi >= V_numvifs) {
596 	VIF_UNLOCK();
597 	return EINVAL;
598     }
599 
600     req->icount = V_viftable[vifi].v_pkt_in;
601     req->ocount = V_viftable[vifi].v_pkt_out;
602     req->ibytes = V_viftable[vifi].v_bytes_in;
603     req->obytes = V_viftable[vifi].v_bytes_out;
604     VIF_UNLOCK();
605 
606     return 0;
607 }
608 
609 static void
610 if_detached_event(void *arg __unused, struct ifnet *ifp)
611 {
612     vifi_t vifi;
613     u_long i;
614 
615     MROUTER_LOCK();
616 
617     if (V_ip_mrouter == NULL) {
618 	MROUTER_UNLOCK();
619 	return;
620     }
621 
622     VIF_LOCK();
623     MFC_LOCK();
624 
625     /*
626      * Tear down multicast forwarder state associated with this ifnet.
627      * 1. Walk the vif list, matching vifs against this ifnet.
628      * 2. Walk the multicast forwarding cache (mfc) looking for
629      *    inner matches with this vif's index.
630      * 3. Expire any matching multicast forwarding cache entries.
631      * 4. Free vif state. This should disable ALLMULTI on the interface.
632      */
633     for (vifi = 0; vifi < V_numvifs; vifi++) {
634 	if (V_viftable[vifi].v_ifp != ifp)
635 		continue;
636 	for (i = 0; i < mfchashsize; i++) {
637 		struct mfc *rt, *nrt;
638 
639 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
640 			if (rt->mfc_parent == vifi) {
641 				expire_mfc(rt);
642 			}
643 		}
644 	}
645 	del_vif_locked(vifi);
646     }
647 
648     MFC_UNLOCK();
649     VIF_UNLOCK();
650 
651     MROUTER_UNLOCK();
652 }
653 
654 /*
655  * Enable multicast forwarding.
656  */
657 static int
658 ip_mrouter_init(struct socket *so, int version)
659 {
660 
661     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
662         so->so_type, so->so_proto->pr_protocol);
663 
664     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
665 	return EOPNOTSUPP;
666 
667     if (version != 1)
668 	return ENOPROTOOPT;
669 
670     MROUTER_LOCK();
671 
672     if (ip_mrouter_unloading) {
673 	MROUTER_UNLOCK();
674 	return ENOPROTOOPT;
675     }
676 
677     if (V_ip_mrouter != NULL) {
678 	MROUTER_UNLOCK();
679 	return EADDRINUSE;
680     }
681 
682     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
683 	HASH_NOWAIT);
684 
685     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
686 	curvnet);
687     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
688 	curvnet);
689     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
690 	curvnet);
691 
692     V_ip_mrouter = so;
693     ip_mrouter_cnt++;
694 
695     MROUTER_UNLOCK();
696 
697     CTR1(KTR_IPMF, "%s: done", __func__);
698 
699     return 0;
700 }
701 
702 /*
703  * Disable multicast forwarding.
704  */
705 static int
706 X_ip_mrouter_done(void)
707 {
708     struct ifnet *ifp;
709     u_long i;
710     vifi_t vifi;
711 
712     MROUTER_LOCK();
713 
714     if (V_ip_mrouter == NULL) {
715 	MROUTER_UNLOCK();
716 	return EINVAL;
717     }
718 
719     /*
720      * Detach/disable hooks to the reset of the system.
721      */
722     V_ip_mrouter = NULL;
723     ip_mrouter_cnt--;
724     V_mrt_api_config = 0;
725 
726     VIF_LOCK();
727 
728     /*
729      * For each phyint in use, disable promiscuous reception of all IP
730      * multicasts.
731      */
732     for (vifi = 0; vifi < V_numvifs; vifi++) {
733 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
734 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
735 	    ifp = V_viftable[vifi].v_ifp;
736 	    if_allmulti(ifp, 0);
737 	}
738     }
739     bzero((caddr_t)V_viftable, sizeof(V_viftable));
740     V_numvifs = 0;
741     V_pim_assert_enabled = 0;
742 
743     VIF_UNLOCK();
744 
745     callout_stop(&V_expire_upcalls_ch);
746     callout_stop(&V_bw_upcalls_ch);
747     callout_stop(&V_bw_meter_ch);
748 
749     MFC_LOCK();
750 
751     /*
752      * Free all multicast forwarding cache entries.
753      * Do not use hashdestroy(), as we must perform other cleanup.
754      */
755     for (i = 0; i < mfchashsize; i++) {
756 	struct mfc *rt, *nrt;
757 
758 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
759 		expire_mfc(rt);
760 	}
761     }
762     free(V_mfchashtbl, M_MRTABLE);
763     V_mfchashtbl = NULL;
764 
765     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
766 
767     V_bw_upcalls_n = 0;
768     bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
769 
770     MFC_UNLOCK();
771 
772     V_reg_vif_num = VIFI_INVALID;
773 
774     MROUTER_UNLOCK();
775 
776     CTR1(KTR_IPMF, "%s: done", __func__);
777 
778     return 0;
779 }
780 
781 /*
782  * Set PIM assert processing global
783  */
784 static int
785 set_assert(int i)
786 {
787     if ((i != 1) && (i != 0))
788 	return EINVAL;
789 
790     V_pim_assert_enabled = i;
791 
792     return 0;
793 }
794 
795 /*
796  * Configure API capabilities
797  */
798 int
799 set_api_config(uint32_t *apival)
800 {
801     u_long i;
802 
803     /*
804      * We can set the API capabilities only if it is the first operation
805      * after MRT_INIT. I.e.:
806      *  - there are no vifs installed
807      *  - pim_assert is not enabled
808      *  - the MFC table is empty
809      */
810     if (V_numvifs > 0) {
811 	*apival = 0;
812 	return EPERM;
813     }
814     if (V_pim_assert_enabled) {
815 	*apival = 0;
816 	return EPERM;
817     }
818 
819     MFC_LOCK();
820 
821     for (i = 0; i < mfchashsize; i++) {
822 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
823 	    MFC_UNLOCK();
824 	    *apival = 0;
825 	    return EPERM;
826 	}
827     }
828 
829     MFC_UNLOCK();
830 
831     V_mrt_api_config = *apival & mrt_api_support;
832     *apival = V_mrt_api_config;
833 
834     return 0;
835 }
836 
837 /*
838  * Add a vif to the vif table
839  */
840 static int
841 add_vif(struct vifctl *vifcp)
842 {
843     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
844     struct sockaddr_in sin = {sizeof sin, AF_INET};
845     struct ifaddr *ifa;
846     struct ifnet *ifp;
847     int error;
848 #ifdef KTR
849     char addrbuf[INET_ADDRSTRLEN];
850 #endif
851 
852     VIF_LOCK();
853     if (vifcp->vifc_vifi >= MAXVIFS) {
854 	VIF_UNLOCK();
855 	return EINVAL;
856     }
857     /* rate limiting is no longer supported by this code */
858     if (vifcp->vifc_rate_limit != 0) {
859 	log(LOG_ERR, "rate limiting is no longer supported\n");
860 	VIF_UNLOCK();
861 	return EINVAL;
862     }
863     if (!in_nullhost(vifp->v_lcl_addr)) {
864 	VIF_UNLOCK();
865 	return EADDRINUSE;
866     }
867     if (in_nullhost(vifcp->vifc_lcl_addr)) {
868 	VIF_UNLOCK();
869 	return EADDRNOTAVAIL;
870     }
871 
872     /* Find the interface with an address in AF_INET family */
873     if (vifcp->vifc_flags & VIFF_REGISTER) {
874 	/*
875 	 * XXX: Because VIFF_REGISTER does not really need a valid
876 	 * local interface (e.g. it could be 127.0.0.2), we don't
877 	 * check its address.
878 	 */
879 	ifp = NULL;
880     } else {
881 	sin.sin_addr = vifcp->vifc_lcl_addr;
882 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
883 	if (ifa == NULL) {
884 	    VIF_UNLOCK();
885 	    return EADDRNOTAVAIL;
886 	}
887 	ifp = ifa->ifa_ifp;
888 	ifa_free(ifa);
889     }
890 
891     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
892 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
893 	VIF_UNLOCK();
894 	return EOPNOTSUPP;
895     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
896 	ifp = &V_multicast_register_if;
897 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
898 	if (V_reg_vif_num == VIFI_INVALID) {
899 	    if_initname(&V_multicast_register_if, "register_vif", 0);
900 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
901 	    V_reg_vif_num = vifcp->vifc_vifi;
902 	}
903     } else {		/* Make sure the interface supports multicast */
904 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
905 	    VIF_UNLOCK();
906 	    return EOPNOTSUPP;
907 	}
908 
909 	/* Enable promiscuous reception of all IP multicasts from the if */
910 	error = if_allmulti(ifp, 1);
911 	if (error) {
912 	    VIF_UNLOCK();
913 	    return error;
914 	}
915     }
916 
917     vifp->v_flags     = vifcp->vifc_flags;
918     vifp->v_threshold = vifcp->vifc_threshold;
919     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
920     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
921     vifp->v_ifp       = ifp;
922     /* initialize per vif pkt counters */
923     vifp->v_pkt_in    = 0;
924     vifp->v_pkt_out   = 0;
925     vifp->v_bytes_in  = 0;
926     vifp->v_bytes_out = 0;
927 
928     /* Adjust numvifs up if the vifi is higher than numvifs */
929     if (V_numvifs <= vifcp->vifc_vifi)
930 	V_numvifs = vifcp->vifc_vifi + 1;
931 
932     VIF_UNLOCK();
933 
934     CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
935 	(int)vifcp->vifc_vifi, inet_ntoa_r(vifcp->vifc_lcl_addr, addrbuf),
936 	(int)vifcp->vifc_threshold);
937 
938     return 0;
939 }
940 
941 /*
942  * Delete a vif from the vif table
943  */
944 static int
945 del_vif_locked(vifi_t vifi)
946 {
947     struct vif *vifp;
948 
949     VIF_LOCK_ASSERT();
950 
951     if (vifi >= V_numvifs) {
952 	return EINVAL;
953     }
954     vifp = &V_viftable[vifi];
955     if (in_nullhost(vifp->v_lcl_addr)) {
956 	return EADDRNOTAVAIL;
957     }
958 
959     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
960 	if_allmulti(vifp->v_ifp, 0);
961 
962     if (vifp->v_flags & VIFF_REGISTER)
963 	V_reg_vif_num = VIFI_INVALID;
964 
965     bzero((caddr_t)vifp, sizeof (*vifp));
966 
967     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
968 
969     /* Adjust numvifs down */
970     for (vifi = V_numvifs; vifi > 0; vifi--)
971 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
972 	    break;
973     V_numvifs = vifi;
974 
975     return 0;
976 }
977 
978 static int
979 del_vif(vifi_t vifi)
980 {
981     int cc;
982 
983     VIF_LOCK();
984     cc = del_vif_locked(vifi);
985     VIF_UNLOCK();
986 
987     return cc;
988 }
989 
990 /*
991  * update an mfc entry without resetting counters and S,G addresses.
992  */
993 static void
994 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
995 {
996     int i;
997 
998     rt->mfc_parent = mfccp->mfcc_parent;
999     for (i = 0; i < V_numvifs; i++) {
1000 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1001 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1002 	    MRT_MFC_FLAGS_ALL;
1003     }
1004     /* set the RP address */
1005     if (V_mrt_api_config & MRT_MFC_RP)
1006 	rt->mfc_rp = mfccp->mfcc_rp;
1007     else
1008 	rt->mfc_rp.s_addr = INADDR_ANY;
1009 }
1010 
1011 /*
1012  * fully initialize an mfc entry from the parameter.
1013  */
1014 static void
1015 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1016 {
1017     rt->mfc_origin     = mfccp->mfcc_origin;
1018     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1019 
1020     update_mfc_params(rt, mfccp);
1021 
1022     /* initialize pkt counters per src-grp */
1023     rt->mfc_pkt_cnt    = 0;
1024     rt->mfc_byte_cnt   = 0;
1025     rt->mfc_wrong_if   = 0;
1026     timevalclear(&rt->mfc_last_assert);
1027 }
1028 
1029 static void
1030 expire_mfc(struct mfc *rt)
1031 {
1032 	struct rtdetq *rte, *nrte;
1033 
1034 	MFC_LOCK_ASSERT();
1035 
1036 	free_bw_list(rt->mfc_bw_meter);
1037 
1038 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1039 		m_freem(rte->m);
1040 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1041 		free(rte, M_MRTABLE);
1042 	}
1043 
1044 	LIST_REMOVE(rt, mfc_hash);
1045 	free(rt, M_MRTABLE);
1046 }
1047 
1048 /*
1049  * Add an mfc entry
1050  */
1051 static int
1052 add_mfc(struct mfcctl2 *mfccp)
1053 {
1054     struct mfc *rt;
1055     struct rtdetq *rte, *nrte;
1056     u_long hash = 0;
1057     u_short nstl;
1058 #ifdef KTR
1059     char addrbuf[INET_ADDRSTRLEN];
1060 #endif
1061 
1062     VIF_LOCK();
1063     MFC_LOCK();
1064 
1065     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1066 
1067     /* If an entry already exists, just update the fields */
1068     if (rt) {
1069 	CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x",
1070 	    __func__, inet_ntoa_r(mfccp->mfcc_origin, addrbuf),
1071 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1072 	    mfccp->mfcc_parent);
1073 	update_mfc_params(rt, mfccp);
1074 	MFC_UNLOCK();
1075 	VIF_UNLOCK();
1076 	return (0);
1077     }
1078 
1079     /*
1080      * Find the entry for which the upcall was made and update
1081      */
1082     nstl = 0;
1083     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1084     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1085 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1086 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1087 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1088 		CTR5(KTR_IPMF,
1089 		    "%s: add mfc orig %s group %lx parent %x qh %p",
1090 		    __func__, inet_ntoa_r(mfccp->mfcc_origin, addrbuf),
1091 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1092 		    mfccp->mfcc_parent,
1093 		    TAILQ_FIRST(&rt->mfc_stall));
1094 		if (nstl++)
1095 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1096 
1097 		init_mfc_params(rt, mfccp);
1098 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1099 		V_nexpire[hash]--;
1100 
1101 		/* Free queued packets, but attempt to forward them first. */
1102 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1103 			if (rte->ifp != NULL)
1104 				ip_mdq(rte->m, rte->ifp, rt, -1);
1105 			m_freem(rte->m);
1106 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1107 			rt->mfc_nstall--;
1108 			free(rte, M_MRTABLE);
1109 		}
1110 	}
1111     }
1112 
1113     /*
1114      * It is possible that an entry is being inserted without an upcall
1115      */
1116     if (nstl == 0) {
1117 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1118 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1119 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1120 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1121 			init_mfc_params(rt, mfccp);
1122 			if (rt->mfc_expire)
1123 			    V_nexpire[hash]--;
1124 			rt->mfc_expire = 0;
1125 			break; /* XXX */
1126 		}
1127 	}
1128 
1129 	if (rt == NULL) {		/* no upcall, so make a new entry */
1130 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1131 	    if (rt == NULL) {
1132 		MFC_UNLOCK();
1133 		VIF_UNLOCK();
1134 		return (ENOBUFS);
1135 	    }
1136 
1137 	    init_mfc_params(rt, mfccp);
1138 	    TAILQ_INIT(&rt->mfc_stall);
1139 	    rt->mfc_nstall = 0;
1140 
1141 	    rt->mfc_expire     = 0;
1142 	    rt->mfc_bw_meter = NULL;
1143 
1144 	    /* insert new entry at head of hash chain */
1145 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1146 	}
1147     }
1148 
1149     MFC_UNLOCK();
1150     VIF_UNLOCK();
1151 
1152     return (0);
1153 }
1154 
1155 /*
1156  * Delete an mfc entry
1157  */
1158 static int
1159 del_mfc(struct mfcctl2 *mfccp)
1160 {
1161     struct in_addr	origin;
1162     struct in_addr	mcastgrp;
1163     struct mfc		*rt;
1164 #ifdef KTR
1165     char		addrbuf[INET_ADDRSTRLEN];
1166 #endif
1167 
1168     origin = mfccp->mfcc_origin;
1169     mcastgrp = mfccp->mfcc_mcastgrp;
1170 
1171     CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__,
1172 	inet_ntoa_r(origin, addrbuf), (u_long)ntohl(mcastgrp.s_addr));
1173 
1174     MFC_LOCK();
1175 
1176     rt = mfc_find(&origin, &mcastgrp);
1177     if (rt == NULL) {
1178 	MFC_UNLOCK();
1179 	return EADDRNOTAVAIL;
1180     }
1181 
1182     /*
1183      * free the bw_meter entries
1184      */
1185     free_bw_list(rt->mfc_bw_meter);
1186     rt->mfc_bw_meter = NULL;
1187 
1188     LIST_REMOVE(rt, mfc_hash);
1189     free(rt, M_MRTABLE);
1190 
1191     MFC_UNLOCK();
1192 
1193     return (0);
1194 }
1195 
1196 /*
1197  * Send a message to the routing daemon on the multicast routing socket.
1198  */
1199 static int
1200 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1201 {
1202     if (s) {
1203 	SOCKBUF_LOCK(&s->so_rcv);
1204 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1205 	    NULL) != 0) {
1206 	    sorwakeup_locked(s);
1207 	    return 0;
1208 	}
1209 	SOCKBUF_UNLOCK(&s->so_rcv);
1210     }
1211     m_freem(mm);
1212     return -1;
1213 }
1214 
1215 /*
1216  * IP multicast forwarding function. This function assumes that the packet
1217  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1218  * pointed to by "ifp", and the packet is to be relayed to other networks
1219  * that have members of the packet's destination IP multicast group.
1220  *
1221  * The packet is returned unscathed to the caller, unless it is
1222  * erroneous, in which case a non-zero return value tells the caller to
1223  * discard it.
1224  */
1225 
1226 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1227 
1228 static int
1229 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1230     struct ip_moptions *imo)
1231 {
1232     struct mfc *rt;
1233     int error;
1234     vifi_t vifi;
1235 #ifdef KTR
1236     char addrbuf[INET_ADDRSTRLEN];
1237 #endif
1238 
1239     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p",
1240 	inet_ntoa_r(ip->ip_src, addrbuf), (u_long)ntohl(ip->ip_dst.s_addr),
1241 	ifp);
1242 
1243     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1244 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1245 	/*
1246 	 * Packet arrived via a physical interface or
1247 	 * an encapsulated tunnel or a register_vif.
1248 	 */
1249     } else {
1250 	/*
1251 	 * Packet arrived through a source-route tunnel.
1252 	 * Source-route tunnels are no longer supported.
1253 	 */
1254 	return (1);
1255     }
1256 
1257     VIF_LOCK();
1258     MFC_LOCK();
1259     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1260 	if (ip->ip_ttl < MAXTTL)
1261 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1262 	error = ip_mdq(m, ifp, NULL, vifi);
1263 	MFC_UNLOCK();
1264 	VIF_UNLOCK();
1265 	return error;
1266     }
1267 
1268     /*
1269      * Don't forward a packet with time-to-live of zero or one,
1270      * or a packet destined to a local-only group.
1271      */
1272     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1273 	MFC_UNLOCK();
1274 	VIF_UNLOCK();
1275 	return 0;
1276     }
1277 
1278     /*
1279      * Determine forwarding vifs from the forwarding cache table
1280      */
1281     MRTSTAT_INC(mrts_mfc_lookups);
1282     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1283 
1284     /* Entry exists, so forward if necessary */
1285     if (rt != NULL) {
1286 	error = ip_mdq(m, ifp, rt, -1);
1287 	MFC_UNLOCK();
1288 	VIF_UNLOCK();
1289 	return error;
1290     } else {
1291 	/*
1292 	 * If we don't have a route for packet's origin,
1293 	 * Make a copy of the packet & send message to routing daemon
1294 	 */
1295 
1296 	struct mbuf *mb0;
1297 	struct rtdetq *rte;
1298 	u_long hash;
1299 	int hlen = ip->ip_hl << 2;
1300 
1301 	MRTSTAT_INC(mrts_mfc_misses);
1302 	MRTSTAT_INC(mrts_no_route);
1303 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)",
1304 	    inet_ntoa_r(ip->ip_src, addrbuf), (u_long)ntohl(ip->ip_dst.s_addr));
1305 
1306 	/*
1307 	 * Allocate mbufs early so that we don't do extra work if we are
1308 	 * just going to fail anyway.  Make sure to pullup the header so
1309 	 * that other people can't step on it.
1310 	 */
1311 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1312 	    M_NOWAIT|M_ZERO);
1313 	if (rte == NULL) {
1314 	    MFC_UNLOCK();
1315 	    VIF_UNLOCK();
1316 	    return ENOBUFS;
1317 	}
1318 
1319 	mb0 = m_copypacket(m, M_NOWAIT);
1320 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1321 	    mb0 = m_pullup(mb0, hlen);
1322 	if (mb0 == NULL) {
1323 	    free(rte, M_MRTABLE);
1324 	    MFC_UNLOCK();
1325 	    VIF_UNLOCK();
1326 	    return ENOBUFS;
1327 	}
1328 
1329 	/* is there an upcall waiting for this flow ? */
1330 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1331 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1332 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1333 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1334 		    !TAILQ_EMPTY(&rt->mfc_stall))
1335 			break;
1336 	}
1337 
1338 	if (rt == NULL) {
1339 	    int i;
1340 	    struct igmpmsg *im;
1341 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1342 	    struct mbuf *mm;
1343 
1344 	    /*
1345 	     * Locate the vifi for the incoming interface for this packet.
1346 	     * If none found, drop packet.
1347 	     */
1348 	    for (vifi = 0; vifi < V_numvifs &&
1349 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1350 		;
1351 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1352 		goto non_fatal;
1353 
1354 	    /* no upcall, so make a new entry */
1355 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1356 	    if (rt == NULL)
1357 		goto fail;
1358 
1359 	    /* Make a copy of the header to send to the user level process */
1360 	    mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1361 	    if (mm == NULL)
1362 		goto fail1;
1363 
1364 	    /*
1365 	     * Send message to routing daemon to install
1366 	     * a route into the kernel table
1367 	     */
1368 
1369 	    im = mtod(mm, struct igmpmsg *);
1370 	    im->im_msgtype = IGMPMSG_NOCACHE;
1371 	    im->im_mbz = 0;
1372 	    im->im_vif = vifi;
1373 
1374 	    MRTSTAT_INC(mrts_upcalls);
1375 
1376 	    k_igmpsrc.sin_addr = ip->ip_src;
1377 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1378 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1379 		MRTSTAT_INC(mrts_upq_sockfull);
1380 fail1:
1381 		free(rt, M_MRTABLE);
1382 fail:
1383 		free(rte, M_MRTABLE);
1384 		m_freem(mb0);
1385 		MFC_UNLOCK();
1386 		VIF_UNLOCK();
1387 		return ENOBUFS;
1388 	    }
1389 
1390 	    /* insert new entry at head of hash chain */
1391 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1392 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1393 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1394 	    V_nexpire[hash]++;
1395 	    for (i = 0; i < V_numvifs; i++) {
1396 		rt->mfc_ttls[i] = 0;
1397 		rt->mfc_flags[i] = 0;
1398 	    }
1399 	    rt->mfc_parent = -1;
1400 
1401 	    /* clear the RP address */
1402 	    rt->mfc_rp.s_addr = INADDR_ANY;
1403 	    rt->mfc_bw_meter = NULL;
1404 
1405 	    /* initialize pkt counters per src-grp */
1406 	    rt->mfc_pkt_cnt = 0;
1407 	    rt->mfc_byte_cnt = 0;
1408 	    rt->mfc_wrong_if = 0;
1409 	    timevalclear(&rt->mfc_last_assert);
1410 
1411 	    TAILQ_INIT(&rt->mfc_stall);
1412 	    rt->mfc_nstall = 0;
1413 
1414 	    /* link into table */
1415 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1416 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1417 	    rt->mfc_nstall++;
1418 
1419 	} else {
1420 	    /* determine if queue has overflowed */
1421 	    if (rt->mfc_nstall > MAX_UPQ) {
1422 		MRTSTAT_INC(mrts_upq_ovflw);
1423 non_fatal:
1424 		free(rte, M_MRTABLE);
1425 		m_freem(mb0);
1426 		MFC_UNLOCK();
1427 		VIF_UNLOCK();
1428 		return (0);
1429 	    }
1430 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1431 	    rt->mfc_nstall++;
1432 	}
1433 
1434 	rte->m			= mb0;
1435 	rte->ifp		= ifp;
1436 
1437 	MFC_UNLOCK();
1438 	VIF_UNLOCK();
1439 
1440 	return 0;
1441     }
1442 }
1443 
1444 /*
1445  * Clean up the cache entry if upcall is not serviced
1446  */
1447 static void
1448 expire_upcalls(void *arg)
1449 {
1450     u_long i;
1451 
1452     CURVNET_SET((struct vnet *) arg);
1453 
1454     MFC_LOCK();
1455 
1456     for (i = 0; i < mfchashsize; i++) {
1457 	struct mfc *rt, *nrt;
1458 
1459 	if (V_nexpire[i] == 0)
1460 	    continue;
1461 
1462 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1463 		if (TAILQ_EMPTY(&rt->mfc_stall))
1464 			continue;
1465 
1466 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1467 			continue;
1468 
1469 		/*
1470 		 * free the bw_meter entries
1471 		 */
1472 		while (rt->mfc_bw_meter != NULL) {
1473 		    struct bw_meter *x = rt->mfc_bw_meter;
1474 
1475 		    rt->mfc_bw_meter = x->bm_mfc_next;
1476 		    free(x, M_BWMETER);
1477 		}
1478 
1479 		MRTSTAT_INC(mrts_cache_cleanups);
1480 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1481 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1482 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1483 
1484 		expire_mfc(rt);
1485 	    }
1486     }
1487 
1488     MFC_UNLOCK();
1489 
1490     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1491 	curvnet);
1492 
1493     CURVNET_RESTORE();
1494 }
1495 
1496 /*
1497  * Packet forwarding routine once entry in the cache is made
1498  */
1499 static int
1500 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1501 {
1502     struct ip  *ip = mtod(m, struct ip *);
1503     vifi_t vifi;
1504     int plen = ntohs(ip->ip_len);
1505 
1506     VIF_LOCK_ASSERT();
1507 
1508     /*
1509      * If xmt_vif is not -1, send on only the requested vif.
1510      *
1511      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1512      */
1513     if (xmt_vif < V_numvifs) {
1514 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1515 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1516 	else
1517 		phyint_send(ip, V_viftable + xmt_vif, m);
1518 	return 1;
1519     }
1520 
1521     /*
1522      * Don't forward if it didn't arrive from the parent vif for its origin.
1523      */
1524     vifi = rt->mfc_parent;
1525     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1526 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1527 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1528 	MRTSTAT_INC(mrts_wrong_if);
1529 	++rt->mfc_wrong_if;
1530 	/*
1531 	 * If we are doing PIM assert processing, send a message
1532 	 * to the routing daemon.
1533 	 *
1534 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1535 	 * can complete the SPT switch, regardless of the type
1536 	 * of the iif (broadcast media, GRE tunnel, etc).
1537 	 */
1538 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1539 	    V_viftable[vifi].v_ifp) {
1540 
1541 	    if (ifp == &V_multicast_register_if)
1542 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1543 
1544 	    /* Get vifi for the incoming packet */
1545 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1546 		vifi++)
1547 		;
1548 	    if (vifi >= V_numvifs)
1549 		return 0;	/* The iif is not found: ignore the packet. */
1550 
1551 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1552 		return 0;	/* WRONGVIF disabled: ignore the packet */
1553 
1554 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1555 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1556 		struct igmpmsg *im;
1557 		int hlen = ip->ip_hl << 2;
1558 		struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1559 
1560 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1561 		    mm = m_pullup(mm, hlen);
1562 		if (mm == NULL)
1563 		    return ENOBUFS;
1564 
1565 		im = mtod(mm, struct igmpmsg *);
1566 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1567 		im->im_mbz		= 0;
1568 		im->im_vif		= vifi;
1569 
1570 		MRTSTAT_INC(mrts_upcalls);
1571 
1572 		k_igmpsrc.sin_addr = im->im_src;
1573 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1574 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1575 		    MRTSTAT_INC(mrts_upq_sockfull);
1576 		    return ENOBUFS;
1577 		}
1578 	    }
1579 	}
1580 	return 0;
1581     }
1582 
1583 
1584     /* If I sourced this packet, it counts as output, else it was input. */
1585     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1586 	V_viftable[vifi].v_pkt_out++;
1587 	V_viftable[vifi].v_bytes_out += plen;
1588     } else {
1589 	V_viftable[vifi].v_pkt_in++;
1590 	V_viftable[vifi].v_bytes_in += plen;
1591     }
1592     rt->mfc_pkt_cnt++;
1593     rt->mfc_byte_cnt += plen;
1594 
1595     /*
1596      * For each vif, decide if a copy of the packet should be forwarded.
1597      * Forward if:
1598      *		- the ttl exceeds the vif's threshold
1599      *		- there are group members downstream on interface
1600      */
1601     for (vifi = 0; vifi < V_numvifs; vifi++)
1602 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1603 	    V_viftable[vifi].v_pkt_out++;
1604 	    V_viftable[vifi].v_bytes_out += plen;
1605 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1606 		pim_register_send(ip, V_viftable + vifi, m, rt);
1607 	    else
1608 		phyint_send(ip, V_viftable + vifi, m);
1609 	}
1610 
1611     /*
1612      * Perform upcall-related bw measuring.
1613      */
1614     if (rt->mfc_bw_meter != NULL) {
1615 	struct bw_meter *x;
1616 	struct timeval now;
1617 
1618 	microtime(&now);
1619 	MFC_LOCK_ASSERT();
1620 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1621 	    bw_meter_receive_packet(x, plen, &now);
1622     }
1623 
1624     return 0;
1625 }
1626 
1627 /*
1628  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1629  */
1630 static int
1631 X_legal_vif_num(int vif)
1632 {
1633 	int ret;
1634 
1635 	ret = 0;
1636 	if (vif < 0)
1637 		return (ret);
1638 
1639 	VIF_LOCK();
1640 	if (vif < V_numvifs)
1641 		ret = 1;
1642 	VIF_UNLOCK();
1643 
1644 	return (ret);
1645 }
1646 
1647 /*
1648  * Return the local address used by this vif
1649  */
1650 static u_long
1651 X_ip_mcast_src(int vifi)
1652 {
1653 	in_addr_t addr;
1654 
1655 	addr = INADDR_ANY;
1656 	if (vifi < 0)
1657 		return (addr);
1658 
1659 	VIF_LOCK();
1660 	if (vifi < V_numvifs)
1661 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1662 	VIF_UNLOCK();
1663 
1664 	return (addr);
1665 }
1666 
1667 static void
1668 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1669 {
1670     struct mbuf *mb_copy;
1671     int hlen = ip->ip_hl << 2;
1672 
1673     VIF_LOCK_ASSERT();
1674 
1675     /*
1676      * Make a new reference to the packet; make sure that
1677      * the IP header is actually copied, not just referenced,
1678      * so that ip_output() only scribbles on the copy.
1679      */
1680     mb_copy = m_copypacket(m, M_NOWAIT);
1681     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1682 	mb_copy = m_pullup(mb_copy, hlen);
1683     if (mb_copy == NULL)
1684 	return;
1685 
1686     send_packet(vifp, mb_copy);
1687 }
1688 
1689 static void
1690 send_packet(struct vif *vifp, struct mbuf *m)
1691 {
1692 	struct ip_moptions imo;
1693 	struct in_multi *imm[2];
1694 	int error;
1695 
1696 	VIF_LOCK_ASSERT();
1697 
1698 	imo.imo_multicast_ifp  = vifp->v_ifp;
1699 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1700 	imo.imo_multicast_loop = 1;
1701 	imo.imo_multicast_vif  = -1;
1702 	imo.imo_num_memberships = 0;
1703 	imo.imo_max_memberships = 2;
1704 	imo.imo_membership  = &imm[0];
1705 
1706 	/*
1707 	 * Re-entrancy should not be a problem here, because
1708 	 * the packets that we send out and are looped back at us
1709 	 * should get rejected because they appear to come from
1710 	 * the loopback interface, thus preventing looping.
1711 	 */
1712 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1713 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1714 	    (ptrdiff_t)(vifp - V_viftable), error);
1715 }
1716 
1717 /*
1718  * Stubs for old RSVP socket shim implementation.
1719  */
1720 
1721 static int
1722 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1723 {
1724 
1725 	return (EOPNOTSUPP);
1726 }
1727 
1728 static void
1729 X_ip_rsvp_force_done(struct socket *so __unused)
1730 {
1731 
1732 }
1733 
1734 static int
1735 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1736 {
1737 	struct mbuf *m;
1738 
1739 	m = *mp;
1740 	*mp = NULL;
1741 	if (!V_rsvp_on)
1742 		m_freem(m);
1743 	return (IPPROTO_DONE);
1744 }
1745 
1746 /*
1747  * Code for bandwidth monitors
1748  */
1749 
1750 /*
1751  * Define common interface for timeval-related methods
1752  */
1753 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1754 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1755 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1756 
1757 static uint32_t
1758 compute_bw_meter_flags(struct bw_upcall *req)
1759 {
1760     uint32_t flags = 0;
1761 
1762     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1763 	flags |= BW_METER_UNIT_PACKETS;
1764     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1765 	flags |= BW_METER_UNIT_BYTES;
1766     if (req->bu_flags & BW_UPCALL_GEQ)
1767 	flags |= BW_METER_GEQ;
1768     if (req->bu_flags & BW_UPCALL_LEQ)
1769 	flags |= BW_METER_LEQ;
1770 
1771     return flags;
1772 }
1773 
1774 /*
1775  * Add a bw_meter entry
1776  */
1777 static int
1778 add_bw_upcall(struct bw_upcall *req)
1779 {
1780     struct mfc *mfc;
1781     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1782 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1783     struct timeval now;
1784     struct bw_meter *x;
1785     uint32_t flags;
1786 
1787     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1788 	return EOPNOTSUPP;
1789 
1790     /* Test if the flags are valid */
1791     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1792 	return EINVAL;
1793     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1794 	return EINVAL;
1795     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1796 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1797 	return EINVAL;
1798 
1799     /* Test if the threshold time interval is valid */
1800     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1801 	return EINVAL;
1802 
1803     flags = compute_bw_meter_flags(req);
1804 
1805     /*
1806      * Find if we have already same bw_meter entry
1807      */
1808     MFC_LOCK();
1809     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1810     if (mfc == NULL) {
1811 	MFC_UNLOCK();
1812 	return EADDRNOTAVAIL;
1813     }
1814     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1815 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1816 			   &req->bu_threshold.b_time, ==)) &&
1817 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1818 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1819 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1820 	    MFC_UNLOCK();
1821 	    return 0;		/* XXX Already installed */
1822 	}
1823     }
1824 
1825     /* Allocate the new bw_meter entry */
1826     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1827     if (x == NULL) {
1828 	MFC_UNLOCK();
1829 	return ENOBUFS;
1830     }
1831 
1832     /* Set the new bw_meter entry */
1833     x->bm_threshold.b_time = req->bu_threshold.b_time;
1834     microtime(&now);
1835     x->bm_start_time = now;
1836     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1837     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1838     x->bm_measured.b_packets = 0;
1839     x->bm_measured.b_bytes = 0;
1840     x->bm_flags = flags;
1841     x->bm_time_next = NULL;
1842     x->bm_time_hash = BW_METER_BUCKETS;
1843 
1844     /* Add the new bw_meter entry to the front of entries for this MFC */
1845     x->bm_mfc = mfc;
1846     x->bm_mfc_next = mfc->mfc_bw_meter;
1847     mfc->mfc_bw_meter = x;
1848     schedule_bw_meter(x, &now);
1849     MFC_UNLOCK();
1850 
1851     return 0;
1852 }
1853 
1854 static void
1855 free_bw_list(struct bw_meter *list)
1856 {
1857     while (list != NULL) {
1858 	struct bw_meter *x = list;
1859 
1860 	list = list->bm_mfc_next;
1861 	unschedule_bw_meter(x);
1862 	free(x, M_BWMETER);
1863     }
1864 }
1865 
1866 /*
1867  * Delete one or multiple bw_meter entries
1868  */
1869 static int
1870 del_bw_upcall(struct bw_upcall *req)
1871 {
1872     struct mfc *mfc;
1873     struct bw_meter *x;
1874 
1875     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1876 	return EOPNOTSUPP;
1877 
1878     MFC_LOCK();
1879 
1880     /* Find the corresponding MFC entry */
1881     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1882     if (mfc == NULL) {
1883 	MFC_UNLOCK();
1884 	return EADDRNOTAVAIL;
1885     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1886 	/*
1887 	 * Delete all bw_meter entries for this mfc
1888 	 */
1889 	struct bw_meter *list;
1890 
1891 	list = mfc->mfc_bw_meter;
1892 	mfc->mfc_bw_meter = NULL;
1893 	free_bw_list(list);
1894 	MFC_UNLOCK();
1895 	return 0;
1896     } else {			/* Delete a single bw_meter entry */
1897 	struct bw_meter *prev;
1898 	uint32_t flags = 0;
1899 
1900 	flags = compute_bw_meter_flags(req);
1901 
1902 	/* Find the bw_meter entry to delete */
1903 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1904 	     prev = x, x = x->bm_mfc_next) {
1905 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1906 			       &req->bu_threshold.b_time, ==)) &&
1907 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1908 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1909 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1910 		break;
1911 	}
1912 	if (x != NULL) { /* Delete entry from the list for this MFC */
1913 	    if (prev != NULL)
1914 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1915 	    else
1916 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1917 
1918 	    unschedule_bw_meter(x);
1919 	    MFC_UNLOCK();
1920 	    /* Free the bw_meter entry */
1921 	    free(x, M_BWMETER);
1922 	    return 0;
1923 	} else {
1924 	    MFC_UNLOCK();
1925 	    return EINVAL;
1926 	}
1927     }
1928     /* NOTREACHED */
1929 }
1930 
1931 /*
1932  * Perform bandwidth measurement processing that may result in an upcall
1933  */
1934 static void
1935 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1936 {
1937     struct timeval delta;
1938 
1939     MFC_LOCK_ASSERT();
1940 
1941     delta = *nowp;
1942     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1943 
1944     if (x->bm_flags & BW_METER_GEQ) {
1945 	/*
1946 	 * Processing for ">=" type of bw_meter entry
1947 	 */
1948 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1949 	    /* Reset the bw_meter entry */
1950 	    x->bm_start_time = *nowp;
1951 	    x->bm_measured.b_packets = 0;
1952 	    x->bm_measured.b_bytes = 0;
1953 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1954 	}
1955 
1956 	/* Record that a packet is received */
1957 	x->bm_measured.b_packets++;
1958 	x->bm_measured.b_bytes += plen;
1959 
1960 	/*
1961 	 * Test if we should deliver an upcall
1962 	 */
1963 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1964 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1965 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1966 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1967 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1968 		/* Prepare an upcall for delivery */
1969 		bw_meter_prepare_upcall(x, nowp);
1970 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1971 	    }
1972 	}
1973     } else if (x->bm_flags & BW_METER_LEQ) {
1974 	/*
1975 	 * Processing for "<=" type of bw_meter entry
1976 	 */
1977 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1978 	    /*
1979 	     * We are behind time with the multicast forwarding table
1980 	     * scanning for "<=" type of bw_meter entries, so test now
1981 	     * if we should deliver an upcall.
1982 	     */
1983 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1984 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1985 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1986 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1987 		/* Prepare an upcall for delivery */
1988 		bw_meter_prepare_upcall(x, nowp);
1989 	    }
1990 	    /* Reschedule the bw_meter entry */
1991 	    unschedule_bw_meter(x);
1992 	    schedule_bw_meter(x, nowp);
1993 	}
1994 
1995 	/* Record that a packet is received */
1996 	x->bm_measured.b_packets++;
1997 	x->bm_measured.b_bytes += plen;
1998 
1999 	/*
2000 	 * Test if we should restart the measuring interval
2001 	 */
2002 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2003 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2004 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
2005 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2006 	    /* Don't restart the measuring interval */
2007 	} else {
2008 	    /* Do restart the measuring interval */
2009 	    /*
2010 	     * XXX: note that we don't unschedule and schedule, because this
2011 	     * might be too much overhead per packet. Instead, when we process
2012 	     * all entries for a given timer hash bin, we check whether it is
2013 	     * really a timeout. If not, we reschedule at that time.
2014 	     */
2015 	    x->bm_start_time = *nowp;
2016 	    x->bm_measured.b_packets = 0;
2017 	    x->bm_measured.b_bytes = 0;
2018 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2019 	}
2020     }
2021 }
2022 
2023 /*
2024  * Prepare a bandwidth-related upcall
2025  */
2026 static void
2027 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2028 {
2029     struct timeval delta;
2030     struct bw_upcall *u;
2031 
2032     MFC_LOCK_ASSERT();
2033 
2034     /*
2035      * Compute the measured time interval
2036      */
2037     delta = *nowp;
2038     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2039 
2040     /*
2041      * If there are too many pending upcalls, deliver them now
2042      */
2043     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2044 	bw_upcalls_send();
2045 
2046     /*
2047      * Set the bw_upcall entry
2048      */
2049     u = &V_bw_upcalls[V_bw_upcalls_n++];
2050     u->bu_src = x->bm_mfc->mfc_origin;
2051     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2052     u->bu_threshold.b_time = x->bm_threshold.b_time;
2053     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2054     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2055     u->bu_measured.b_time = delta;
2056     u->bu_measured.b_packets = x->bm_measured.b_packets;
2057     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2058     u->bu_flags = 0;
2059     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2060 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2061     if (x->bm_flags & BW_METER_UNIT_BYTES)
2062 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2063     if (x->bm_flags & BW_METER_GEQ)
2064 	u->bu_flags |= BW_UPCALL_GEQ;
2065     if (x->bm_flags & BW_METER_LEQ)
2066 	u->bu_flags |= BW_UPCALL_LEQ;
2067 }
2068 
2069 /*
2070  * Send the pending bandwidth-related upcalls
2071  */
2072 static void
2073 bw_upcalls_send(void)
2074 {
2075     struct mbuf *m;
2076     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2077     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2078     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2079 				      0,		/* unused2 */
2080 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2081 				      0,		/* im_mbz  */
2082 				      0,		/* im_vif  */
2083 				      0,		/* unused3 */
2084 				      { 0 },		/* im_src  */
2085 				      { 0 } };		/* im_dst  */
2086 
2087     MFC_LOCK_ASSERT();
2088 
2089     if (V_bw_upcalls_n == 0)
2090 	return;			/* No pending upcalls */
2091 
2092     V_bw_upcalls_n = 0;
2093 
2094     /*
2095      * Allocate a new mbuf, initialize it with the header and
2096      * the payload for the pending calls.
2097      */
2098     m = m_gethdr(M_NOWAIT, MT_DATA);
2099     if (m == NULL) {
2100 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2101 	return;
2102     }
2103 
2104     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2105     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2106 
2107     /*
2108      * Send the upcalls
2109      * XXX do we need to set the address in k_igmpsrc ?
2110      */
2111     MRTSTAT_INC(mrts_upcalls);
2112     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2113 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2114 	MRTSTAT_INC(mrts_upq_sockfull);
2115     }
2116 }
2117 
2118 /*
2119  * Compute the timeout hash value for the bw_meter entries
2120  */
2121 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2122     do {								\
2123 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2124 									\
2125 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2126 	(hash) = next_timeval.tv_sec;					\
2127 	if (next_timeval.tv_usec)					\
2128 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2129 	(hash) %= BW_METER_BUCKETS;					\
2130     } while (0)
2131 
2132 /*
2133  * Schedule a timer to process periodically bw_meter entry of type "<="
2134  * by linking the entry in the proper hash bucket.
2135  */
2136 static void
2137 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2138 {
2139     int time_hash;
2140 
2141     MFC_LOCK_ASSERT();
2142 
2143     if (!(x->bm_flags & BW_METER_LEQ))
2144 	return;		/* XXX: we schedule timers only for "<=" entries */
2145 
2146     /*
2147      * Reset the bw_meter entry
2148      */
2149     x->bm_start_time = *nowp;
2150     x->bm_measured.b_packets = 0;
2151     x->bm_measured.b_bytes = 0;
2152     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2153 
2154     /*
2155      * Compute the timeout hash value and insert the entry
2156      */
2157     BW_METER_TIMEHASH(x, time_hash);
2158     x->bm_time_next = V_bw_meter_timers[time_hash];
2159     V_bw_meter_timers[time_hash] = x;
2160     x->bm_time_hash = time_hash;
2161 }
2162 
2163 /*
2164  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2165  * by removing the entry from the proper hash bucket.
2166  */
2167 static void
2168 unschedule_bw_meter(struct bw_meter *x)
2169 {
2170     int time_hash;
2171     struct bw_meter *prev, *tmp;
2172 
2173     MFC_LOCK_ASSERT();
2174 
2175     if (!(x->bm_flags & BW_METER_LEQ))
2176 	return;		/* XXX: we schedule timers only for "<=" entries */
2177 
2178     /*
2179      * Compute the timeout hash value and delete the entry
2180      */
2181     time_hash = x->bm_time_hash;
2182     if (time_hash >= BW_METER_BUCKETS)
2183 	return;		/* Entry was not scheduled */
2184 
2185     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2186 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2187 	if (tmp == x)
2188 	    break;
2189 
2190     if (tmp == NULL)
2191 	panic("unschedule_bw_meter: bw_meter entry not found");
2192 
2193     if (prev != NULL)
2194 	prev->bm_time_next = x->bm_time_next;
2195     else
2196 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2197 
2198     x->bm_time_next = NULL;
2199     x->bm_time_hash = BW_METER_BUCKETS;
2200 }
2201 
2202 
2203 /*
2204  * Process all "<=" type of bw_meter that should be processed now,
2205  * and for each entry prepare an upcall if necessary. Each processed
2206  * entry is rescheduled again for the (periodic) processing.
2207  *
2208  * This is run periodically (once per second normally). On each round,
2209  * all the potentially matching entries are in the hash slot that we are
2210  * looking at.
2211  */
2212 static void
2213 bw_meter_process()
2214 {
2215     uint32_t loops;
2216     int i;
2217     struct timeval now, process_endtime;
2218 
2219     microtime(&now);
2220     if (V_last_tv_sec == now.tv_sec)
2221 	return;		/* nothing to do */
2222 
2223     loops = now.tv_sec - V_last_tv_sec;
2224     V_last_tv_sec = now.tv_sec;
2225     if (loops > BW_METER_BUCKETS)
2226 	loops = BW_METER_BUCKETS;
2227 
2228     MFC_LOCK();
2229     /*
2230      * Process all bins of bw_meter entries from the one after the last
2231      * processed to the current one. On entry, i points to the last bucket
2232      * visited, so we need to increment i at the beginning of the loop.
2233      */
2234     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2235 	struct bw_meter *x, *tmp_list;
2236 
2237 	if (++i >= BW_METER_BUCKETS)
2238 	    i = 0;
2239 
2240 	/* Disconnect the list of bw_meter entries from the bin */
2241 	tmp_list = V_bw_meter_timers[i];
2242 	V_bw_meter_timers[i] = NULL;
2243 
2244 	/* Process the list of bw_meter entries */
2245 	while (tmp_list != NULL) {
2246 	    x = tmp_list;
2247 	    tmp_list = tmp_list->bm_time_next;
2248 
2249 	    /* Test if the time interval is over */
2250 	    process_endtime = x->bm_start_time;
2251 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2252 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2253 		/* Not yet: reschedule, but don't reset */
2254 		int time_hash;
2255 
2256 		BW_METER_TIMEHASH(x, time_hash);
2257 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2258 		    /*
2259 		     * XXX: somehow the bin processing is a bit ahead of time.
2260 		     * Put the entry in the next bin.
2261 		     */
2262 		    if (++time_hash >= BW_METER_BUCKETS)
2263 			time_hash = 0;
2264 		}
2265 		x->bm_time_next = V_bw_meter_timers[time_hash];
2266 		V_bw_meter_timers[time_hash] = x;
2267 		x->bm_time_hash = time_hash;
2268 
2269 		continue;
2270 	    }
2271 
2272 	    /*
2273 	     * Test if we should deliver an upcall
2274 	     */
2275 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2276 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2277 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2278 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2279 		/* Prepare an upcall for delivery */
2280 		bw_meter_prepare_upcall(x, &now);
2281 	    }
2282 
2283 	    /*
2284 	     * Reschedule for next processing
2285 	     */
2286 	    schedule_bw_meter(x, &now);
2287 	}
2288     }
2289 
2290     /* Send all upcalls that are pending delivery */
2291     bw_upcalls_send();
2292 
2293     MFC_UNLOCK();
2294 }
2295 
2296 /*
2297  * A periodic function for sending all upcalls that are pending delivery
2298  */
2299 static void
2300 expire_bw_upcalls_send(void *arg)
2301 {
2302     CURVNET_SET((struct vnet *) arg);
2303 
2304     MFC_LOCK();
2305     bw_upcalls_send();
2306     MFC_UNLOCK();
2307 
2308     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2309 	curvnet);
2310     CURVNET_RESTORE();
2311 }
2312 
2313 /*
2314  * A periodic function for periodic scanning of the multicast forwarding
2315  * table for processing all "<=" bw_meter entries.
2316  */
2317 static void
2318 expire_bw_meter_process(void *arg)
2319 {
2320     CURVNET_SET((struct vnet *) arg);
2321 
2322     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2323 	bw_meter_process();
2324 
2325     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2326 	curvnet);
2327     CURVNET_RESTORE();
2328 }
2329 
2330 /*
2331  * End of bandwidth monitoring code
2332  */
2333 
2334 /*
2335  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2336  *
2337  */
2338 static int
2339 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2340     struct mfc *rt)
2341 {
2342     struct mbuf *mb_copy, *mm;
2343 
2344     /*
2345      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2346      * rendezvous point was unspecified, and we were told not to.
2347      */
2348     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2349 	in_nullhost(rt->mfc_rp))
2350 	return 0;
2351 
2352     mb_copy = pim_register_prepare(ip, m);
2353     if (mb_copy == NULL)
2354 	return ENOBUFS;
2355 
2356     /*
2357      * Send all the fragments. Note that the mbuf for each fragment
2358      * is freed by the sending machinery.
2359      */
2360     for (mm = mb_copy; mm; mm = mb_copy) {
2361 	mb_copy = mm->m_nextpkt;
2362 	mm->m_nextpkt = 0;
2363 	mm = m_pullup(mm, sizeof(struct ip));
2364 	if (mm != NULL) {
2365 	    ip = mtod(mm, struct ip *);
2366 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2367 		pim_register_send_rp(ip, vifp, mm, rt);
2368 	    } else {
2369 		pim_register_send_upcall(ip, vifp, mm, rt);
2370 	    }
2371 	}
2372     }
2373 
2374     return 0;
2375 }
2376 
2377 /*
2378  * Return a copy of the data packet that is ready for PIM Register
2379  * encapsulation.
2380  * XXX: Note that in the returned copy the IP header is a valid one.
2381  */
2382 static struct mbuf *
2383 pim_register_prepare(struct ip *ip, struct mbuf *m)
2384 {
2385     struct mbuf *mb_copy = NULL;
2386     int mtu;
2387 
2388     /* Take care of delayed checksums */
2389     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2390 	in_delayed_cksum(m);
2391 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2392     }
2393 
2394     /*
2395      * Copy the old packet & pullup its IP header into the
2396      * new mbuf so we can modify it.
2397      */
2398     mb_copy = m_copypacket(m, M_NOWAIT);
2399     if (mb_copy == NULL)
2400 	return NULL;
2401     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2402     if (mb_copy == NULL)
2403 	return NULL;
2404 
2405     /* take care of the TTL */
2406     ip = mtod(mb_copy, struct ip *);
2407     --ip->ip_ttl;
2408 
2409     /* Compute the MTU after the PIM Register encapsulation */
2410     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2411 
2412     if (ntohs(ip->ip_len) <= mtu) {
2413 	/* Turn the IP header into a valid one */
2414 	ip->ip_sum = 0;
2415 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2416     } else {
2417 	/* Fragment the packet */
2418 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2419 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2420 	    m_freem(mb_copy);
2421 	    return NULL;
2422 	}
2423     }
2424     return mb_copy;
2425 }
2426 
2427 /*
2428  * Send an upcall with the data packet to the user-level process.
2429  */
2430 static int
2431 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2432     struct mbuf *mb_copy, struct mfc *rt)
2433 {
2434     struct mbuf *mb_first;
2435     int len = ntohs(ip->ip_len);
2436     struct igmpmsg *im;
2437     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2438 
2439     VIF_LOCK_ASSERT();
2440 
2441     /*
2442      * Add a new mbuf with an upcall header
2443      */
2444     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2445     if (mb_first == NULL) {
2446 	m_freem(mb_copy);
2447 	return ENOBUFS;
2448     }
2449     mb_first->m_data += max_linkhdr;
2450     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2451     mb_first->m_len = sizeof(struct igmpmsg);
2452     mb_first->m_next = mb_copy;
2453 
2454     /* Send message to routing daemon */
2455     im = mtod(mb_first, struct igmpmsg *);
2456     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2457     im->im_mbz		= 0;
2458     im->im_vif		= vifp - V_viftable;
2459     im->im_src		= ip->ip_src;
2460     im->im_dst		= ip->ip_dst;
2461 
2462     k_igmpsrc.sin_addr	= ip->ip_src;
2463 
2464     MRTSTAT_INC(mrts_upcalls);
2465 
2466     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2467 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2468 	MRTSTAT_INC(mrts_upq_sockfull);
2469 	return ENOBUFS;
2470     }
2471 
2472     /* Keep statistics */
2473     PIMSTAT_INC(pims_snd_registers_msgs);
2474     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2475 
2476     return 0;
2477 }
2478 
2479 /*
2480  * Encapsulate the data packet in PIM Register message and send it to the RP.
2481  */
2482 static int
2483 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2484     struct mfc *rt)
2485 {
2486     struct mbuf *mb_first;
2487     struct ip *ip_outer;
2488     struct pim_encap_pimhdr *pimhdr;
2489     int len = ntohs(ip->ip_len);
2490     vifi_t vifi = rt->mfc_parent;
2491 
2492     VIF_LOCK_ASSERT();
2493 
2494     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2495 	m_freem(mb_copy);
2496 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2497     }
2498 
2499     /*
2500      * Add a new mbuf with the encapsulating header
2501      */
2502     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2503     if (mb_first == NULL) {
2504 	m_freem(mb_copy);
2505 	return ENOBUFS;
2506     }
2507     mb_first->m_data += max_linkhdr;
2508     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2509     mb_first->m_next = mb_copy;
2510 
2511     mb_first->m_pkthdr.len = len + mb_first->m_len;
2512 
2513     /*
2514      * Fill in the encapsulating IP and PIM header
2515      */
2516     ip_outer = mtod(mb_first, struct ip *);
2517     *ip_outer = pim_encap_iphdr;
2518     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2519 	sizeof(pim_encap_pimhdr));
2520     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2521     ip_outer->ip_dst = rt->mfc_rp;
2522     /*
2523      * Copy the inner header TOS to the outer header, and take care of the
2524      * IP_DF bit.
2525      */
2526     ip_outer->ip_tos = ip->ip_tos;
2527     if (ip->ip_off & htons(IP_DF))
2528 	ip_outer->ip_off |= htons(IP_DF);
2529     ip_fillid(ip_outer);
2530     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2531 					 + sizeof(pim_encap_iphdr));
2532     *pimhdr = pim_encap_pimhdr;
2533     /* If the iif crosses a border, set the Border-bit */
2534     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2535 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2536 
2537     mb_first->m_data += sizeof(pim_encap_iphdr);
2538     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2539     mb_first->m_data -= sizeof(pim_encap_iphdr);
2540 
2541     send_packet(vifp, mb_first);
2542 
2543     /* Keep statistics */
2544     PIMSTAT_INC(pims_snd_registers_msgs);
2545     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2546 
2547     return 0;
2548 }
2549 
2550 /*
2551  * pim_encapcheck() is called by the encap4_input() path at runtime to
2552  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2553  * into the kernel.
2554  */
2555 static int
2556 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2557 {
2558 
2559 #ifdef DIAGNOSTIC
2560     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2561 #endif
2562     if (proto != IPPROTO_PIM)
2563 	return 0;	/* not for us; reject the datagram. */
2564 
2565     return 64;		/* claim the datagram. */
2566 }
2567 
2568 /*
2569  * PIM-SMv2 and PIM-DM messages processing.
2570  * Receives and verifies the PIM control messages, and passes them
2571  * up to the listening socket, using rip_input().
2572  * The only message with special processing is the PIM_REGISTER message
2573  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2574  * is passed to if_simloop().
2575  */
2576 int
2577 pim_input(struct mbuf **mp, int *offp, int proto)
2578 {
2579     struct mbuf *m = *mp;
2580     struct ip *ip = mtod(m, struct ip *);
2581     struct pim *pim;
2582     int iphlen = *offp;
2583     int minlen;
2584     int datalen = ntohs(ip->ip_len) - iphlen;
2585     int ip_tos;
2586 #ifdef KTR
2587     char addrbuf[INET_ADDRSTRLEN];
2588 #endif
2589 
2590     *mp = NULL;
2591 
2592     /* Keep statistics */
2593     PIMSTAT_INC(pims_rcv_total_msgs);
2594     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2595 
2596     /*
2597      * Validate lengths
2598      */
2599     if (datalen < PIM_MINLEN) {
2600 	PIMSTAT_INC(pims_rcv_tooshort);
2601 	CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2602 	    __func__, datalen, inet_ntoa_r(ip->ip_src, addrbuf));
2603 	m_freem(m);
2604 	return (IPPROTO_DONE);
2605     }
2606 
2607     /*
2608      * If the packet is at least as big as a REGISTER, go agead
2609      * and grab the PIM REGISTER header size, to avoid another
2610      * possible m_pullup() later.
2611      *
2612      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2613      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2614      */
2615     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2616     /*
2617      * Get the IP and PIM headers in contiguous memory, and
2618      * possibly the PIM REGISTER header.
2619      */
2620     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2621 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2622 	return (IPPROTO_DONE);
2623     }
2624 
2625     /* m_pullup() may have given us a new mbuf so reset ip. */
2626     ip = mtod(m, struct ip *);
2627     ip_tos = ip->ip_tos;
2628 
2629     /* adjust mbuf to point to the PIM header */
2630     m->m_data += iphlen;
2631     m->m_len  -= iphlen;
2632     pim = mtod(m, struct pim *);
2633 
2634     /*
2635      * Validate checksum. If PIM REGISTER, exclude the data packet.
2636      *
2637      * XXX: some older PIMv2 implementations don't make this distinction,
2638      * so for compatibility reason perform the checksum over part of the
2639      * message, and if error, then over the whole message.
2640      */
2641     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2642 	/* do nothing, checksum okay */
2643     } else if (in_cksum(m, datalen)) {
2644 	PIMSTAT_INC(pims_rcv_badsum);
2645 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2646 	m_freem(m);
2647 	return (IPPROTO_DONE);
2648     }
2649 
2650     /* PIM version check */
2651     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2652 	PIMSTAT_INC(pims_rcv_badversion);
2653 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2654 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2655 	m_freem(m);
2656 	return (IPPROTO_DONE);
2657     }
2658 
2659     /* restore mbuf back to the outer IP */
2660     m->m_data -= iphlen;
2661     m->m_len  += iphlen;
2662 
2663     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2664 	/*
2665 	 * Since this is a REGISTER, we'll make a copy of the register
2666 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2667 	 * routing daemon.
2668 	 */
2669 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2670 	struct mbuf *mcp;
2671 	struct ip *encap_ip;
2672 	u_int32_t *reghdr;
2673 	struct ifnet *vifp;
2674 
2675 	VIF_LOCK();
2676 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2677 	    VIF_UNLOCK();
2678 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2679 		(int)V_reg_vif_num);
2680 	    m_freem(m);
2681 	    return (IPPROTO_DONE);
2682 	}
2683 	/* XXX need refcnt? */
2684 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2685 	VIF_UNLOCK();
2686 
2687 	/*
2688 	 * Validate length
2689 	 */
2690 	if (datalen < PIM_REG_MINLEN) {
2691 	    PIMSTAT_INC(pims_rcv_tooshort);
2692 	    PIMSTAT_INC(pims_rcv_badregisters);
2693 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2694 	    m_freem(m);
2695 	    return (IPPROTO_DONE);
2696 	}
2697 
2698 	reghdr = (u_int32_t *)(pim + 1);
2699 	encap_ip = (struct ip *)(reghdr + 1);
2700 
2701 	CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2702 	    __func__, inet_ntoa_r(encap_ip->ip_src, addrbuf),
2703 	    ntohs(encap_ip->ip_len));
2704 
2705 	/* verify the version number of the inner packet */
2706 	if (encap_ip->ip_v != IPVERSION) {
2707 	    PIMSTAT_INC(pims_rcv_badregisters);
2708 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2709 	    m_freem(m);
2710 	    return (IPPROTO_DONE);
2711 	}
2712 
2713 	/* verify the inner packet is destined to a mcast group */
2714 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2715 	    PIMSTAT_INC(pims_rcv_badregisters);
2716 	    CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2717 		inet_ntoa_r(encap_ip->ip_dst, addrbuf));
2718 	    m_freem(m);
2719 	    return (IPPROTO_DONE);
2720 	}
2721 
2722 	/* If a NULL_REGISTER, pass it to the daemon */
2723 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2724 	    goto pim_input_to_daemon;
2725 
2726 	/*
2727 	 * Copy the TOS from the outer IP header to the inner IP header.
2728 	 */
2729 	if (encap_ip->ip_tos != ip_tos) {
2730 	    /* Outer TOS -> inner TOS */
2731 	    encap_ip->ip_tos = ip_tos;
2732 	    /* Recompute the inner header checksum. Sigh... */
2733 
2734 	    /* adjust mbuf to point to the inner IP header */
2735 	    m->m_data += (iphlen + PIM_MINLEN);
2736 	    m->m_len  -= (iphlen + PIM_MINLEN);
2737 
2738 	    encap_ip->ip_sum = 0;
2739 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2740 
2741 	    /* restore mbuf to point back to the outer IP header */
2742 	    m->m_data -= (iphlen + PIM_MINLEN);
2743 	    m->m_len  += (iphlen + PIM_MINLEN);
2744 	}
2745 
2746 	/*
2747 	 * Decapsulate the inner IP packet and loopback to forward it
2748 	 * as a normal multicast packet. Also, make a copy of the
2749 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2750 	 * to pass to the daemon later, so it can take the appropriate
2751 	 * actions (e.g., send back PIM_REGISTER_STOP).
2752 	 * XXX: here m->m_data points to the outer IP header.
2753 	 */
2754 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2755 	if (mcp == NULL) {
2756 	    CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2757 	    m_freem(m);
2758 	    return (IPPROTO_DONE);
2759 	}
2760 
2761 	/* Keep statistics */
2762 	/* XXX: registers_bytes include only the encap. mcast pkt */
2763 	PIMSTAT_INC(pims_rcv_registers_msgs);
2764 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2765 
2766 	/*
2767 	 * forward the inner ip packet; point m_data at the inner ip.
2768 	 */
2769 	m_adj(m, iphlen + PIM_MINLEN);
2770 
2771 	CTR4(KTR_IPMF,
2772 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2773 	    __func__,
2774 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2775 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2776 	    (int)V_reg_vif_num);
2777 
2778 	/* NB: vifp was collected above; can it change on us? */
2779 	if_simloop(vifp, m, dst.sin_family, 0);
2780 
2781 	/* prepare the register head to send to the mrouting daemon */
2782 	m = mcp;
2783     }
2784 
2785 pim_input_to_daemon:
2786     /*
2787      * Pass the PIM message up to the daemon; if it is a Register message,
2788      * pass the 'head' only up to the daemon. This includes the
2789      * outer IP header, PIM header, PIM-Register header and the
2790      * inner IP header.
2791      * XXX: the outer IP header pkt size of a Register is not adjust to
2792      * reflect the fact that the inner multicast data is truncated.
2793      */
2794     *mp = m;
2795     rip_input(mp, offp, proto);
2796 
2797     return (IPPROTO_DONE);
2798 }
2799 
2800 static int
2801 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2802 {
2803 	struct mfc	*rt;
2804 	int		 error, i;
2805 
2806 	if (req->newptr)
2807 		return (EPERM);
2808 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2809 		return (0);
2810 	error = sysctl_wire_old_buffer(req, 0);
2811 	if (error)
2812 		return (error);
2813 
2814 	MFC_LOCK();
2815 	for (i = 0; i < mfchashsize; i++) {
2816 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2817 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2818 			if (error)
2819 				goto out_locked;
2820 		}
2821 	}
2822 out_locked:
2823 	MFC_UNLOCK();
2824 	return (error);
2825 }
2826 
2827 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2828     sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2829     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2830 
2831 static void
2832 vnet_mroute_init(const void *unused __unused)
2833 {
2834 
2835 	MALLOC(V_nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2836 	bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2837 	callout_init(&V_expire_upcalls_ch, 1);
2838 	callout_init(&V_bw_upcalls_ch, 1);
2839 	callout_init(&V_bw_meter_ch, 1);
2840 }
2841 
2842 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2843 	NULL);
2844 
2845 static void
2846 vnet_mroute_uninit(const void *unused __unused)
2847 {
2848 
2849 	FREE(V_nexpire, M_MRTABLE);
2850 	V_nexpire = NULL;
2851 }
2852 
2853 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2854 	vnet_mroute_uninit, NULL);
2855 
2856 static int
2857 ip_mroute_modevent(module_t mod, int type, void *unused)
2858 {
2859 
2860     switch (type) {
2861     case MOD_LOAD:
2862 	MROUTER_LOCK_INIT();
2863 
2864 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2865 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2866 	if (if_detach_event_tag == NULL) {
2867 		printf("ip_mroute: unable to register "
2868 		    "ifnet_departure_event handler\n");
2869 		MROUTER_LOCK_DESTROY();
2870 		return (EINVAL);
2871 	}
2872 
2873 	MFC_LOCK_INIT();
2874 	VIF_LOCK_INIT();
2875 
2876 	mfchashsize = MFCHASHSIZE;
2877 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2878 	    !powerof2(mfchashsize)) {
2879 		printf("WARNING: %s not a power of 2; using default\n",
2880 		    "net.inet.ip.mfchashsize");
2881 		mfchashsize = MFCHASHSIZE;
2882 	}
2883 
2884 	pim_squelch_wholepkt = 0;
2885 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2886 	    &pim_squelch_wholepkt);
2887 
2888 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2889 	    pim_encapcheck, &in_pim_protosw, NULL);
2890 	if (pim_encap_cookie == NULL) {
2891 		printf("ip_mroute: unable to attach pim encap\n");
2892 		VIF_LOCK_DESTROY();
2893 		MFC_LOCK_DESTROY();
2894 		MROUTER_LOCK_DESTROY();
2895 		return (EINVAL);
2896 	}
2897 
2898 	ip_mcast_src = X_ip_mcast_src;
2899 	ip_mforward = X_ip_mforward;
2900 	ip_mrouter_done = X_ip_mrouter_done;
2901 	ip_mrouter_get = X_ip_mrouter_get;
2902 	ip_mrouter_set = X_ip_mrouter_set;
2903 
2904 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2905 	ip_rsvp_vif = X_ip_rsvp_vif;
2906 
2907 	legal_vif_num = X_legal_vif_num;
2908 	mrt_ioctl = X_mrt_ioctl;
2909 	rsvp_input_p = X_rsvp_input;
2910 	break;
2911 
2912     case MOD_UNLOAD:
2913 	/*
2914 	 * Typically module unload happens after the user-level
2915 	 * process has shutdown the kernel services (the check
2916 	 * below insures someone can't just yank the module out
2917 	 * from under a running process).  But if the module is
2918 	 * just loaded and then unloaded w/o starting up a user
2919 	 * process we still need to cleanup.
2920 	 */
2921 	MROUTER_LOCK();
2922 	if (ip_mrouter_cnt != 0) {
2923 	    MROUTER_UNLOCK();
2924 	    return (EINVAL);
2925 	}
2926 	ip_mrouter_unloading = 1;
2927 	MROUTER_UNLOCK();
2928 
2929 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2930 
2931 	if (pim_encap_cookie) {
2932 	    encap_detach(pim_encap_cookie);
2933 	    pim_encap_cookie = NULL;
2934 	}
2935 
2936 	ip_mcast_src = NULL;
2937 	ip_mforward = NULL;
2938 	ip_mrouter_done = NULL;
2939 	ip_mrouter_get = NULL;
2940 	ip_mrouter_set = NULL;
2941 
2942 	ip_rsvp_force_done = NULL;
2943 	ip_rsvp_vif = NULL;
2944 
2945 	legal_vif_num = NULL;
2946 	mrt_ioctl = NULL;
2947 	rsvp_input_p = NULL;
2948 
2949 	VIF_LOCK_DESTROY();
2950 	MFC_LOCK_DESTROY();
2951 	MROUTER_LOCK_DESTROY();
2952 	break;
2953 
2954     default:
2955 	return EOPNOTSUPP;
2956     }
2957     return 0;
2958 }
2959 
2960 static moduledata_t ip_mroutemod = {
2961     "ip_mroute",
2962     ip_mroute_modevent,
2963     0
2964 };
2965 
2966 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2967