xref: /freebsd/sys/netinet/ip_mroute.c (revision 195ebc7e9e4b129de810833791a19dfb4349d6a9)
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_mac.h"
74 #include "opt_mrouting.h"
75 
76 #define _PIM_VT 1
77 
78 #include <sys/param.h>
79 #include <sys/kernel.h>
80 #include <sys/stddef.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/vimage.h>
98 
99 #include <net/if.h>
100 #include <net/netisr.h>
101 #include <net/route.h>
102 
103 #include <netinet/in.h>
104 #include <netinet/igmp.h>
105 #include <netinet/in_systm.h>
106 #include <netinet/in_var.h>
107 #include <netinet/ip.h>
108 #include <netinet/ip_encap.h>
109 #include <netinet/ip_mroute.h>
110 #include <netinet/ip_var.h>
111 #include <netinet/ip_options.h>
112 #include <netinet/pim.h>
113 #include <netinet/pim_var.h>
114 #include <netinet/udp.h>
115 #include <netinet/vinet.h>
116 
117 #include <machine/in_cksum.h>
118 
119 #include <security/mac/mac_framework.h>
120 
121 #ifndef KTR_IPMF
122 #define KTR_IPMF KTR_INET
123 #endif
124 
125 #define		VIFI_INVALID	((vifi_t) -1)
126 #define		M_HASCL(m)	((m)->m_flags & M_EXT)
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 struct mrtstat	mrtstat;
147 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
148     &mrtstat, mrtstat,
149     "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
150     "netinet/ip_mroute.h)");
151 
152 static u_long			 mfchash;
153 #define MFCHASH(a, g)							\
154 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
155 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
156 #define MFCHASHSIZE	256
157 
158 static u_char			*nexpire;	/* 0..mfchashsize-1 */
159 static u_long			 mfchashsize;	/* Hash size */
160 LIST_HEAD(mfchashhdr, mfc)	*mfchashtbl;
161 
162 static struct mtx mfc_mtx;
163 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
164 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
165 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
166 #define	MFC_LOCK_INIT()							\
167 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
168 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
169 
170 static vifi_t		numvifs;
171 static struct vif	viftable[MAXVIFS];
172 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
173     &viftable, sizeof(viftable), "S,vif[MAXVIFS]",
174     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
175 
176 static struct mtx vif_mtx;
177 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
178 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
179 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
180 #define	VIF_LOCK_INIT()							\
181 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
182 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
183 
184 static eventhandler_tag if_detach_event_tag = NULL;
185 
186 static struct callout expire_upcalls_ch;
187 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
188 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
189 
190 /*
191  * Bandwidth meter variables and constants
192  */
193 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
194 /*
195  * Pending timeouts are stored in a hash table, the key being the
196  * expiration time. Periodically, the entries are analysed and processed.
197  */
198 #define BW_METER_BUCKETS	1024
199 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
200 static struct callout bw_meter_ch;
201 #define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
202 
203 /*
204  * Pending upcalls are stored in a vector which is flushed when
205  * full, or periodically
206  */
207 static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
208 static u_int	bw_upcalls_n; /* # of pending upcalls */
209 static struct callout bw_upcalls_ch;
210 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
211 
212 static struct pimstat pimstat;
213 
214 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
215 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
216     &pimstat, pimstat,
217     "PIM Statistics (struct pimstat, netinet/pim_var.h)");
218 
219 static u_long	pim_squelch_wholepkt = 0;
220 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
221     &pim_squelch_wholepkt, 0,
222     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
223 
224 extern  struct domain inetdomain;
225 static const struct protosw in_pim_protosw = {
226 	.pr_type =		SOCK_RAW,
227 	.pr_domain =		&inetdomain,
228 	.pr_protocol =		IPPROTO_PIM,
229 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
230 	.pr_input =		pim_input,
231 	.pr_output =		(pr_output_t*)rip_output,
232 	.pr_ctloutput =		rip_ctloutput,
233 	.pr_usrreqs =		&rip_usrreqs
234 };
235 static const struct encaptab *pim_encap_cookie;
236 
237 static int pim_encapcheck(const struct mbuf *, int, int, void *);
238 
239 /*
240  * Note: the PIM Register encapsulation adds the following in front of a
241  * data packet:
242  *
243  * struct pim_encap_hdr {
244  *    struct ip ip;
245  *    struct pim_encap_pimhdr  pim;
246  * }
247  *
248  */
249 
250 struct pim_encap_pimhdr {
251 	struct pim pim;
252 	uint32_t   flags;
253 };
254 #define		PIM_ENCAP_TTL	64
255 
256 static struct ip pim_encap_iphdr = {
257 #if BYTE_ORDER == LITTLE_ENDIAN
258 	sizeof(struct ip) >> 2,
259 	IPVERSION,
260 #else
261 	IPVERSION,
262 	sizeof(struct ip) >> 2,
263 #endif
264 	0,			/* tos */
265 	sizeof(struct ip),	/* total length */
266 	0,			/* id */
267 	0,			/* frag offset */
268 	PIM_ENCAP_TTL,
269 	IPPROTO_PIM,
270 	0,			/* checksum */
271 };
272 
273 static struct pim_encap_pimhdr pim_encap_pimhdr = {
274     {
275 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
276 	0,			/* reserved */
277 	0,			/* checksum */
278     },
279     0				/* flags */
280 };
281 
282 static struct ifnet multicast_register_if;
283 static vifi_t reg_vif_num = VIFI_INVALID;
284 
285 /*
286  * Private variables.
287  */
288 
289 static u_long	X_ip_mcast_src(int);
290 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
291 		    struct ip_moptions *);
292 static int	X_ip_mrouter_done(void);
293 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
294 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
295 static int	X_legal_vif_num(int);
296 static int	X_mrt_ioctl(int, caddr_t, int);
297 
298 static int	add_bw_upcall(struct bw_upcall *);
299 static int	add_mfc(struct mfcctl2 *);
300 static int	add_vif(struct vifctl *);
301 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
302 static void	bw_meter_process(void);
303 static void	bw_meter_receive_packet(struct bw_meter *, int,
304 		    struct timeval *);
305 static void	bw_upcalls_send(void);
306 static int	del_bw_upcall(struct bw_upcall *);
307 static int	del_mfc(struct mfcctl2 *);
308 static int	del_vif(vifi_t);
309 static int	del_vif_locked(vifi_t);
310 static void	expire_bw_meter_process(void *);
311 static void	expire_bw_upcalls_send(void *);
312 static void	expire_mfc(struct mfc *);
313 static void	expire_upcalls(void *);
314 static void	free_bw_list(struct bw_meter *);
315 static int	get_sg_cnt(struct sioc_sg_req *);
316 static int	get_vif_cnt(struct sioc_vif_req *);
317 static void	if_detached_event(void *, struct ifnet *);
318 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
319 static int	ip_mrouter_init(struct socket *, int);
320 static __inline struct mfc *
321 		mfc_find(struct in_addr *, struct in_addr *);
322 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
323 static struct mbuf *
324 		pim_register_prepare(struct ip *, struct mbuf *);
325 static int	pim_register_send(struct ip *, struct vif *,
326 		    struct mbuf *, struct mfc *);
327 static int	pim_register_send_rp(struct ip *, struct vif *,
328 		    struct mbuf *, struct mfc *);
329 static int	pim_register_send_upcall(struct ip *, struct vif *,
330 		    struct mbuf *, struct mfc *);
331 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
332 static void	send_packet(struct vif *, struct mbuf *);
333 static int	set_api_config(uint32_t *);
334 static int	set_assert(int);
335 static int	socket_send(struct socket *, struct mbuf *,
336 		    struct sockaddr_in *);
337 static void	unschedule_bw_meter(struct bw_meter *);
338 
339 /*
340  * Kernel multicast forwarding API capabilities and setup.
341  * If more API capabilities are added to the kernel, they should be
342  * recorded in `mrt_api_support'.
343  */
344 #define MRT_API_VERSION		0x0305
345 
346 static const int mrt_api_version = MRT_API_VERSION;
347 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
348 					 MRT_MFC_FLAGS_BORDER_VIF |
349 					 MRT_MFC_RP |
350 					 MRT_MFC_BW_UPCALL);
351 static uint32_t mrt_api_config = 0;
352 
353 static int pim_assert_enabled;
354 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
355 
356 /*
357  * Find a route for a given origin IP address and multicast group address.
358  * Statistics must be updated by the caller.
359  */
360 static __inline struct mfc *
361 mfc_find(struct in_addr *o, struct in_addr *g)
362 {
363 	struct mfc *rt;
364 
365 	MFC_LOCK_ASSERT();
366 
367 	LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
368 		if (in_hosteq(rt->mfc_origin, *o) &&
369 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
370 		    TAILQ_EMPTY(&rt->mfc_stall))
371 			break;
372 	}
373 
374 	return (rt);
375 }
376 
377 /*
378  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
379  */
380 static int
381 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
382 {
383     INIT_VNET_INET(curvnet);
384     int	error, optval;
385     vifi_t	vifi;
386     struct	vifctl vifc;
387     struct	mfcctl2 mfc;
388     struct	bw_upcall bw_upcall;
389     uint32_t	i;
390 
391     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
392 	return EPERM;
393 
394     error = 0;
395     switch (sopt->sopt_name) {
396     case MRT_INIT:
397 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
398 	if (error)
399 	    break;
400 	error = ip_mrouter_init(so, optval);
401 	break;
402 
403     case MRT_DONE:
404 	error = ip_mrouter_done();
405 	break;
406 
407     case MRT_ADD_VIF:
408 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
409 	if (error)
410 	    break;
411 	error = add_vif(&vifc);
412 	break;
413 
414     case MRT_DEL_VIF:
415 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
416 	if (error)
417 	    break;
418 	error = del_vif(vifi);
419 	break;
420 
421     case MRT_ADD_MFC:
422     case MRT_DEL_MFC:
423 	/*
424 	 * select data size depending on API version.
425 	 */
426 	if (sopt->sopt_name == MRT_ADD_MFC &&
427 		mrt_api_config & MRT_API_FLAGS_ALL) {
428 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
429 				sizeof(struct mfcctl2));
430 	} else {
431 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
432 				sizeof(struct mfcctl));
433 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
434 			sizeof(mfc) - sizeof(struct mfcctl));
435 	}
436 	if (error)
437 	    break;
438 	if (sopt->sopt_name == MRT_ADD_MFC)
439 	    error = add_mfc(&mfc);
440 	else
441 	    error = del_mfc(&mfc);
442 	break;
443 
444     case MRT_ASSERT:
445 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
446 	if (error)
447 	    break;
448 	set_assert(optval);
449 	break;
450 
451     case MRT_API_CONFIG:
452 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
453 	if (!error)
454 	    error = set_api_config(&i);
455 	if (!error)
456 	    error = sooptcopyout(sopt, &i, sizeof i);
457 	break;
458 
459     case MRT_ADD_BW_UPCALL:
460     case MRT_DEL_BW_UPCALL:
461 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
462 				sizeof bw_upcall);
463 	if (error)
464 	    break;
465 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
466 	    error = add_bw_upcall(&bw_upcall);
467 	else
468 	    error = del_bw_upcall(&bw_upcall);
469 	break;
470 
471     default:
472 	error = EOPNOTSUPP;
473 	break;
474     }
475     return error;
476 }
477 
478 /*
479  * Handle MRT getsockopt commands
480  */
481 static int
482 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
483 {
484     int error;
485 
486     switch (sopt->sopt_name) {
487     case MRT_VERSION:
488 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
489 	break;
490 
491     case MRT_ASSERT:
492 	error = sooptcopyout(sopt, &pim_assert_enabled,
493 	    sizeof pim_assert_enabled);
494 	break;
495 
496     case MRT_API_SUPPORT:
497 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
498 	break;
499 
500     case MRT_API_CONFIG:
501 	error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config);
502 	break;
503 
504     default:
505 	error = EOPNOTSUPP;
506 	break;
507     }
508     return error;
509 }
510 
511 /*
512  * Handle ioctl commands to obtain information from the cache
513  */
514 static int
515 X_mrt_ioctl(int cmd, caddr_t data, int fibnum __unused)
516 {
517     int error = 0;
518 
519     /*
520      * Currently the only function calling this ioctl routine is rtioctl().
521      * Typically, only root can create the raw socket in order to execute
522      * this ioctl method, however the request might be coming from a prison
523      */
524     error = priv_check(curthread, PRIV_NETINET_MROUTE);
525     if (error)
526 	return (error);
527     switch (cmd) {
528     case (SIOCGETVIFCNT):
529 	error = get_vif_cnt((struct sioc_vif_req *)data);
530 	break;
531 
532     case (SIOCGETSGCNT):
533 	error = get_sg_cnt((struct sioc_sg_req *)data);
534 	break;
535 
536     default:
537 	error = EINVAL;
538 	break;
539     }
540     return error;
541 }
542 
543 /*
544  * returns the packet, byte, rpf-failure count for the source group provided
545  */
546 static int
547 get_sg_cnt(struct sioc_sg_req *req)
548 {
549     struct mfc *rt;
550 
551     MFC_LOCK();
552     rt = mfc_find(&req->src, &req->grp);
553     if (rt == NULL) {
554 	MFC_UNLOCK();
555 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
556 	return EADDRNOTAVAIL;
557     }
558     req->pktcnt = rt->mfc_pkt_cnt;
559     req->bytecnt = rt->mfc_byte_cnt;
560     req->wrong_if = rt->mfc_wrong_if;
561     MFC_UNLOCK();
562     return 0;
563 }
564 
565 /*
566  * returns the input and output packet and byte counts on the vif provided
567  */
568 static int
569 get_vif_cnt(struct sioc_vif_req *req)
570 {
571     vifi_t vifi = req->vifi;
572 
573     VIF_LOCK();
574     if (vifi >= numvifs) {
575 	VIF_UNLOCK();
576 	return EINVAL;
577     }
578 
579     req->icount = viftable[vifi].v_pkt_in;
580     req->ocount = viftable[vifi].v_pkt_out;
581     req->ibytes = viftable[vifi].v_bytes_in;
582     req->obytes = viftable[vifi].v_bytes_out;
583     VIF_UNLOCK();
584 
585     return 0;
586 }
587 
588 static void
589 ip_mrouter_reset(void)
590 {
591 
592     pim_assert_enabled = 0;
593     mrt_api_config = 0;
594 
595     callout_init(&expire_upcalls_ch, CALLOUT_MPSAFE);
596 
597     bw_upcalls_n = 0;
598     bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers));
599     callout_init(&bw_upcalls_ch, CALLOUT_MPSAFE);
600     callout_init(&bw_meter_ch, CALLOUT_MPSAFE);
601 }
602 
603 static void
604 if_detached_event(void *arg __unused, struct ifnet *ifp)
605 {
606     INIT_VNET_INET(curvnet);
607     vifi_t vifi;
608     int i;
609 
610     MROUTER_LOCK();
611 
612     if (V_ip_mrouter == NULL) {
613 	MROUTER_UNLOCK();
614 	return;
615     }
616 
617     VIF_LOCK();
618     MFC_LOCK();
619 
620     /*
621      * Tear down multicast forwarder state associated with this ifnet.
622      * 1. Walk the vif list, matching vifs against this ifnet.
623      * 2. Walk the multicast forwarding cache (mfc) looking for
624      *    inner matches with this vif's index.
625      * 3. Expire any matching multicast forwarding cache entries.
626      * 4. Free vif state. This should disable ALLMULTI on the interface.
627      */
628     for (vifi = 0; vifi < numvifs; vifi++) {
629 	if (viftable[vifi].v_ifp != ifp)
630 		continue;
631 	for (i = 0; i < mfchashsize; i++) {
632 		struct mfc *rt, *nrt;
633 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
634 			nrt = LIST_NEXT(rt, mfc_hash);
635 			if (rt->mfc_parent == vifi) {
636 				expire_mfc(rt);
637 			}
638 		}
639 	}
640 	del_vif_locked(vifi);
641     }
642 
643     MFC_UNLOCK();
644     VIF_UNLOCK();
645 
646     MROUTER_UNLOCK();
647 }
648 
649 /*
650  * Enable multicast forwarding.
651  */
652 static int
653 ip_mrouter_init(struct socket *so, int version)
654 {
655     INIT_VNET_INET(curvnet);
656 
657     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
658         so->so_type, so->so_proto->pr_protocol);
659 
660     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
661 	return EOPNOTSUPP;
662 
663     if (version != 1)
664 	return ENOPROTOOPT;
665 
666     MROUTER_LOCK();
667 
668     if (V_ip_mrouter != NULL) {
669 	MROUTER_UNLOCK();
670 	return EADDRINUSE;
671     }
672 
673     if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
674         if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
675     if (if_detach_event_tag == NULL) {
676 	MROUTER_UNLOCK();
677 	return (ENOMEM);
678     }
679 
680     mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &mfchash, HASH_NOWAIT);
681 
682     callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
683 
684     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
685 	expire_bw_upcalls_send, NULL);
686     callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
687 
688     V_ip_mrouter = so;
689 
690     MROUTER_UNLOCK();
691 
692     CTR1(KTR_IPMF, "%s: done", __func__);
693 
694     return 0;
695 }
696 
697 /*
698  * Disable multicast forwarding.
699  */
700 static int
701 X_ip_mrouter_done(void)
702 {
703     INIT_VNET_INET(curvnet);
704     vifi_t vifi;
705     int i;
706     struct ifnet *ifp;
707     struct ifreq ifr;
708 
709     MROUTER_LOCK();
710 
711     if (V_ip_mrouter == NULL) {
712 	MROUTER_UNLOCK();
713 	return EINVAL;
714     }
715 
716     /*
717      * Detach/disable hooks to the reset of the system.
718      */
719     V_ip_mrouter = NULL;
720     mrt_api_config = 0;
721 
722     VIF_LOCK();
723 
724     /*
725      * For each phyint in use, disable promiscuous reception of all IP
726      * multicasts.
727      */
728     for (vifi = 0; vifi < numvifs; vifi++) {
729 	if (!in_nullhost(viftable[vifi].v_lcl_addr) &&
730 		!(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
731 	    struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
732 
733 	    so->sin_len = sizeof(struct sockaddr_in);
734 	    so->sin_family = AF_INET;
735 	    so->sin_addr.s_addr = INADDR_ANY;
736 	    ifp = viftable[vifi].v_ifp;
737 	    if_allmulti(ifp, 0);
738 	}
739     }
740     bzero((caddr_t)viftable, sizeof(viftable));
741     numvifs = 0;
742     pim_assert_enabled = 0;
743 
744     VIF_UNLOCK();
745 
746     EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
747 
748     callout_stop(&expire_upcalls_ch);
749     callout_stop(&bw_upcalls_ch);
750     callout_stop(&bw_meter_ch);
751 
752     MFC_LOCK();
753 
754     /*
755      * Free all multicast forwarding cache entries.
756      * Do not use hashdestroy(), as we must perform other cleanup.
757      */
758     for (i = 0; i < mfchashsize; i++) {
759 	struct mfc *rt, *nrt;
760 	for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
761 		nrt = LIST_NEXT(rt, mfc_hash);
762 		expire_mfc(rt);
763 	}
764     }
765     free(mfchashtbl, M_MRTABLE);
766     mfchashtbl = NULL;
767 
768     bzero(nexpire, sizeof(nexpire[0]) * mfchashsize);
769 
770     bw_upcalls_n = 0;
771     bzero(bw_meter_timers, sizeof(bw_meter_timers));
772 
773     MFC_UNLOCK();
774 
775     reg_vif_num = VIFI_INVALID;
776 
777     MROUTER_UNLOCK();
778 
779     CTR1(KTR_IPMF, "%s: done", __func__);
780 
781     return 0;
782 }
783 
784 /*
785  * Set PIM assert processing global
786  */
787 static int
788 set_assert(int i)
789 {
790     if ((i != 1) && (i != 0))
791 	return EINVAL;
792 
793     pim_assert_enabled = i;
794 
795     return 0;
796 }
797 
798 /*
799  * Configure API capabilities
800  */
801 int
802 set_api_config(uint32_t *apival)
803 {
804     int i;
805 
806     /*
807      * We can set the API capabilities only if it is the first operation
808      * after MRT_INIT. I.e.:
809      *  - there are no vifs installed
810      *  - pim_assert is not enabled
811      *  - the MFC table is empty
812      */
813     if (numvifs > 0) {
814 	*apival = 0;
815 	return EPERM;
816     }
817     if (pim_assert_enabled) {
818 	*apival = 0;
819 	return EPERM;
820     }
821 
822     MFC_LOCK();
823 
824     for (i = 0; i < mfchashsize; i++) {
825 	if (LIST_FIRST(&mfchashtbl[i]) != NULL) {
826 	    *apival = 0;
827 	    return EPERM;
828 	}
829     }
830 
831     MFC_UNLOCK();
832 
833     mrt_api_config = *apival & mrt_api_support;
834     *apival = mrt_api_config;
835 
836     return 0;
837 }
838 
839 /*
840  * Add a vif to the vif table
841  */
842 static int
843 add_vif(struct vifctl *vifcp)
844 {
845     struct vif *vifp = viftable + vifcp->vifc_vifi;
846     struct sockaddr_in sin = {sizeof sin, AF_INET};
847     struct ifaddr *ifa;
848     struct ifnet *ifp;
849     int error;
850 
851     VIF_LOCK();
852     if (vifcp->vifc_vifi >= MAXVIFS) {
853 	VIF_UNLOCK();
854 	return EINVAL;
855     }
856     /* rate limiting is no longer supported by this code */
857     if (vifcp->vifc_rate_limit != 0) {
858 	log(LOG_ERR, "rate limiting is no longer supported\n");
859 	VIF_UNLOCK();
860 	return EINVAL;
861     }
862     if (!in_nullhost(vifp->v_lcl_addr)) {
863 	VIF_UNLOCK();
864 	return EADDRINUSE;
865     }
866     if (in_nullhost(vifcp->vifc_lcl_addr)) {
867 	VIF_UNLOCK();
868 	return EADDRNOTAVAIL;
869     }
870 
871     /* Find the interface with an address in AF_INET family */
872     if (vifcp->vifc_flags & VIFF_REGISTER) {
873 	/*
874 	 * XXX: Because VIFF_REGISTER does not really need a valid
875 	 * local interface (e.g. it could be 127.0.0.2), we don't
876 	 * check its address.
877 	 */
878 	ifp = NULL;
879     } else {
880 	sin.sin_addr = vifcp->vifc_lcl_addr;
881 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
882 	if (ifa == NULL) {
883 	    VIF_UNLOCK();
884 	    return EADDRNOTAVAIL;
885 	}
886 	ifp = ifa->ifa_ifp;
887     }
888 
889     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
890 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
891 	VIF_UNLOCK();
892 	return EOPNOTSUPP;
893     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
894 	ifp = &multicast_register_if;
895 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
896 	if (reg_vif_num == VIFI_INVALID) {
897 	    if_initname(&multicast_register_if, "register_vif", 0);
898 	    multicast_register_if.if_flags = IFF_LOOPBACK;
899 	    reg_vif_num = vifcp->vifc_vifi;
900 	}
901     } else {		/* Make sure the interface supports multicast */
902 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
903 	    VIF_UNLOCK();
904 	    return EOPNOTSUPP;
905 	}
906 
907 	/* Enable promiscuous reception of all IP multicasts from the if */
908 	error = if_allmulti(ifp, 1);
909 	if (error) {
910 	    VIF_UNLOCK();
911 	    return error;
912 	}
913     }
914 
915     vifp->v_flags     = vifcp->vifc_flags;
916     vifp->v_threshold = vifcp->vifc_threshold;
917     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
918     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
919     vifp->v_ifp       = ifp;
920     /* initialize per vif pkt counters */
921     vifp->v_pkt_in    = 0;
922     vifp->v_pkt_out   = 0;
923     vifp->v_bytes_in  = 0;
924     vifp->v_bytes_out = 0;
925     bzero(&vifp->v_route, sizeof(vifp->v_route));
926 
927     /* Adjust numvifs up if the vifi is higher than numvifs */
928     if (numvifs <= vifcp->vifc_vifi)
929 	numvifs = vifcp->vifc_vifi + 1;
930 
931     VIF_UNLOCK();
932 
933     CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
934 	(int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr),
935 	(int)vifcp->vifc_threshold);
936 
937     return 0;
938 }
939 
940 /*
941  * Delete a vif from the vif table
942  */
943 static int
944 del_vif_locked(vifi_t vifi)
945 {
946     struct vif *vifp;
947 
948     VIF_LOCK_ASSERT();
949 
950     if (vifi >= numvifs) {
951 	return EINVAL;
952     }
953     vifp = &viftable[vifi];
954     if (in_nullhost(vifp->v_lcl_addr)) {
955 	return EADDRNOTAVAIL;
956     }
957 
958     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
959 	if_allmulti(vifp->v_ifp, 0);
960 
961     if (vifp->v_flags & VIFF_REGISTER)
962 	reg_vif_num = VIFI_INVALID;
963 
964     bzero((caddr_t)vifp, sizeof (*vifp));
965 
966     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
967 
968     /* Adjust numvifs down */
969     for (vifi = numvifs; vifi > 0; vifi--)
970 	if (!in_nullhost(viftable[vifi-1].v_lcl_addr))
971 	    break;
972     numvifs = vifi;
973 
974     return 0;
975 }
976 
977 static int
978 del_vif(vifi_t vifi)
979 {
980     int cc;
981 
982     VIF_LOCK();
983     cc = del_vif_locked(vifi);
984     VIF_UNLOCK();
985 
986     return cc;
987 }
988 
989 /*
990  * update an mfc entry without resetting counters and S,G addresses.
991  */
992 static void
993 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
994 {
995     int i;
996 
997     rt->mfc_parent = mfccp->mfcc_parent;
998     for (i = 0; i < numvifs; i++) {
999 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1000 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1001 	    MRT_MFC_FLAGS_ALL;
1002     }
1003     /* set the RP address */
1004     if (mrt_api_config & MRT_MFC_RP)
1005 	rt->mfc_rp = mfccp->mfcc_rp;
1006     else
1007 	rt->mfc_rp.s_addr = INADDR_ANY;
1008 }
1009 
1010 /*
1011  * fully initialize an mfc entry from the parameter.
1012  */
1013 static void
1014 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1015 {
1016     rt->mfc_origin     = mfccp->mfcc_origin;
1017     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1018 
1019     update_mfc_params(rt, mfccp);
1020 
1021     /* initialize pkt counters per src-grp */
1022     rt->mfc_pkt_cnt    = 0;
1023     rt->mfc_byte_cnt   = 0;
1024     rt->mfc_wrong_if   = 0;
1025     timevalclear(&rt->mfc_last_assert);
1026 }
1027 
1028 static void
1029 expire_mfc(struct mfc *rt)
1030 {
1031 	struct rtdetq *rte, *nrte;
1032 
1033 	free_bw_list(rt->mfc_bw_meter);
1034 
1035 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1036 		m_freem(rte->m);
1037 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1038 		free(rte, M_MRTABLE);
1039 	}
1040 
1041 	LIST_REMOVE(rt, mfc_hash);
1042 	free(rt, M_MRTABLE);
1043 }
1044 
1045 /*
1046  * Add an mfc entry
1047  */
1048 static int
1049 add_mfc(struct mfcctl2 *mfccp)
1050 {
1051     struct mfc *rt;
1052     struct rtdetq *rte, *nrte;
1053     u_long hash = 0;
1054     u_short nstl;
1055 
1056     VIF_LOCK();
1057     MFC_LOCK();
1058 
1059     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1060 
1061     /* If an entry already exists, just update the fields */
1062     if (rt) {
1063 	CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x",
1064 	    __func__, inet_ntoa(mfccp->mfcc_origin),
1065 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1066 	    mfccp->mfcc_parent);
1067 	update_mfc_params(rt, mfccp);
1068 	MFC_UNLOCK();
1069 	VIF_UNLOCK();
1070 	return (0);
1071     }
1072 
1073     /*
1074      * Find the entry for which the upcall was made and update
1075      */
1076     nstl = 0;
1077     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1078     LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1079 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1080 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1081 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1082 		CTR5(KTR_IPMF,
1083 		    "%s: add mfc orig %s group %lx parent %x qh %p",
1084 		    __func__, inet_ntoa(mfccp->mfcc_origin),
1085 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1086 		    mfccp->mfcc_parent,
1087 		    TAILQ_FIRST(&rt->mfc_stall));
1088 		if (nstl++)
1089 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1090 
1091 		init_mfc_params(rt, mfccp);
1092 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1093 		nexpire[hash]--;
1094 
1095 		/* Free queued packets, but attempt to forward them first. */
1096 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1097 			if (rte->ifp != NULL)
1098 				ip_mdq(rte->m, rte->ifp, rt, -1);
1099 			m_freem(rte->m);
1100 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1101 			rt->mfc_nstall--;
1102 			free(rte, M_MRTABLE);
1103 		}
1104 	}
1105     }
1106 
1107     /*
1108      * It is possible that an entry is being inserted without an upcall
1109      */
1110     if (nstl == 0) {
1111 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1112 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1113 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1114 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1115 			init_mfc_params(rt, mfccp);
1116 			if (rt->mfc_expire)
1117 			    nexpire[hash]--;
1118 			rt->mfc_expire = 0;
1119 			break; /* XXX */
1120 		}
1121 	}
1122 
1123 	if (rt == NULL) {		/* no upcall, so make a new entry */
1124 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1125 	    if (rt == NULL) {
1126 		MFC_UNLOCK();
1127 		VIF_UNLOCK();
1128 		return (ENOBUFS);
1129 	    }
1130 
1131 	    init_mfc_params(rt, mfccp);
1132 	    TAILQ_INIT(&rt->mfc_stall);
1133 	    rt->mfc_nstall = 0;
1134 
1135 	    rt->mfc_expire     = 0;
1136 	    rt->mfc_bw_meter = NULL;
1137 
1138 	    /* insert new entry at head of hash chain */
1139 	    LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1140 	}
1141     }
1142 
1143     MFC_UNLOCK();
1144     VIF_UNLOCK();
1145 
1146     return (0);
1147 }
1148 
1149 /*
1150  * Delete an mfc entry
1151  */
1152 static int
1153 del_mfc(struct mfcctl2 *mfccp)
1154 {
1155     struct in_addr	origin;
1156     struct in_addr	mcastgrp;
1157     struct mfc		*rt;
1158 
1159     origin = mfccp->mfcc_origin;
1160     mcastgrp = mfccp->mfcc_mcastgrp;
1161 
1162     CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__,
1163 	inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr));
1164 
1165     MFC_LOCK();
1166 
1167     rt = mfc_find(&origin, &mcastgrp);
1168     if (rt == NULL) {
1169 	MFC_UNLOCK();
1170 	return EADDRNOTAVAIL;
1171     }
1172 
1173     /*
1174      * free the bw_meter entries
1175      */
1176     free_bw_list(rt->mfc_bw_meter);
1177     rt->mfc_bw_meter = NULL;
1178 
1179     LIST_REMOVE(rt, mfc_hash);
1180     free(rt, M_MRTABLE);
1181 
1182     MFC_UNLOCK();
1183 
1184     return (0);
1185 }
1186 
1187 /*
1188  * Send a message to the routing daemon on the multicast routing socket.
1189  */
1190 static int
1191 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1192 {
1193     if (s) {
1194 	SOCKBUF_LOCK(&s->so_rcv);
1195 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1196 	    NULL) != 0) {
1197 	    sorwakeup_locked(s);
1198 	    return 0;
1199 	}
1200 	SOCKBUF_UNLOCK(&s->so_rcv);
1201     }
1202     m_freem(mm);
1203     return -1;
1204 }
1205 
1206 /*
1207  * IP multicast forwarding function. This function assumes that the packet
1208  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1209  * pointed to by "ifp", and the packet is to be relayed to other networks
1210  * that have members of the packet's destination IP multicast group.
1211  *
1212  * The packet is returned unscathed to the caller, unless it is
1213  * erroneous, in which case a non-zero return value tells the caller to
1214  * discard it.
1215  */
1216 
1217 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1218 
1219 static int
1220 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1221     struct ip_moptions *imo)
1222 {
1223     INIT_VNET_INET(curvnet);
1224     struct mfc *rt;
1225     int error;
1226     vifi_t vifi;
1227 
1228     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p",
1229 	inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1230 
1231     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1232 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1233 	/*
1234 	 * Packet arrived via a physical interface or
1235 	 * an encapsulated tunnel or a register_vif.
1236 	 */
1237     } else {
1238 	/*
1239 	 * Packet arrived through a source-route tunnel.
1240 	 * Source-route tunnels are no longer supported.
1241 	 */
1242 	return (1);
1243     }
1244 
1245     VIF_LOCK();
1246     MFC_LOCK();
1247     if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1248 	if (ip->ip_ttl < MAXTTL)
1249 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1250 	error = ip_mdq(m, ifp, NULL, vifi);
1251 	MFC_UNLOCK();
1252 	VIF_UNLOCK();
1253 	return error;
1254     }
1255 
1256     /*
1257      * Don't forward a packet with time-to-live of zero or one,
1258      * or a packet destined to a local-only group.
1259      */
1260     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1261 	MFC_UNLOCK();
1262 	VIF_UNLOCK();
1263 	return 0;
1264     }
1265 
1266     /*
1267      * Determine forwarding vifs from the forwarding cache table
1268      */
1269     MRTSTAT_INC(mrts_mfc_lookups);
1270     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1271 
1272     /* Entry exists, so forward if necessary */
1273     if (rt != NULL) {
1274 	error = ip_mdq(m, ifp, rt, -1);
1275 	MFC_UNLOCK();
1276 	VIF_UNLOCK();
1277 	return error;
1278     } else {
1279 	/*
1280 	 * If we don't have a route for packet's origin,
1281 	 * Make a copy of the packet & send message to routing daemon
1282 	 */
1283 
1284 	struct mbuf *mb0;
1285 	struct rtdetq *rte;
1286 	u_long hash;
1287 	int hlen = ip->ip_hl << 2;
1288 
1289 	MRTSTAT_INC(mrts_mfc_misses);
1290 	MRTSTAT_INC(mrts_no_route);
1291 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)",
1292 	    inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr));
1293 
1294 	/*
1295 	 * Allocate mbufs early so that we don't do extra work if we are
1296 	 * just going to fail anyway.  Make sure to pullup the header so
1297 	 * that other people can't step on it.
1298 	 */
1299 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1300 	    M_NOWAIT|M_ZERO);
1301 	if (rte == NULL) {
1302 	    MFC_UNLOCK();
1303 	    VIF_UNLOCK();
1304 	    return ENOBUFS;
1305 	}
1306 
1307 	mb0 = m_copypacket(m, M_DONTWAIT);
1308 	if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1309 	    mb0 = m_pullup(mb0, hlen);
1310 	if (mb0 == NULL) {
1311 	    free(rte, M_MRTABLE);
1312 	    MFC_UNLOCK();
1313 	    VIF_UNLOCK();
1314 	    return ENOBUFS;
1315 	}
1316 
1317 	/* is there an upcall waiting for this flow ? */
1318 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1319 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1320 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1321 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1322 		    !TAILQ_EMPTY(&rt->mfc_stall))
1323 			break;
1324 	}
1325 
1326 	if (rt == NULL) {
1327 	    int i;
1328 	    struct igmpmsg *im;
1329 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1330 	    struct mbuf *mm;
1331 
1332 	    /*
1333 	     * Locate the vifi for the incoming interface for this packet.
1334 	     * If none found, drop packet.
1335 	     */
1336 	    for (vifi = 0; vifi < numvifs &&
1337 		    viftable[vifi].v_ifp != ifp; vifi++)
1338 		;
1339 	    if (vifi >= numvifs)	/* vif not found, drop packet */
1340 		goto non_fatal;
1341 
1342 	    /* no upcall, so make a new entry */
1343 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1344 	    if (rt == NULL)
1345 		goto fail;
1346 
1347 	    /* Make a copy of the header to send to the user level process */
1348 	    mm = m_copy(mb0, 0, hlen);
1349 	    if (mm == NULL)
1350 		goto fail1;
1351 
1352 	    /*
1353 	     * Send message to routing daemon to install
1354 	     * a route into the kernel table
1355 	     */
1356 
1357 	    im = mtod(mm, struct igmpmsg *);
1358 	    im->im_msgtype = IGMPMSG_NOCACHE;
1359 	    im->im_mbz = 0;
1360 	    im->im_vif = vifi;
1361 
1362 	    MRTSTAT_INC(mrts_upcalls);
1363 
1364 	    k_igmpsrc.sin_addr = ip->ip_src;
1365 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1366 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1367 		MRTSTAT_INC(mrts_upq_sockfull);
1368 fail1:
1369 		free(rt, M_MRTABLE);
1370 fail:
1371 		free(rte, M_MRTABLE);
1372 		m_freem(mb0);
1373 		MFC_UNLOCK();
1374 		VIF_UNLOCK();
1375 		return ENOBUFS;
1376 	    }
1377 
1378 	    /* insert new entry at head of hash chain */
1379 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1380 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1381 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1382 	    nexpire[hash]++;
1383 	    for (i = 0; i < numvifs; i++) {
1384 		rt->mfc_ttls[i] = 0;
1385 		rt->mfc_flags[i] = 0;
1386 	    }
1387 	    rt->mfc_parent = -1;
1388 
1389 	    /* clear the RP address */
1390 	    rt->mfc_rp.s_addr = INADDR_ANY;
1391 	    rt->mfc_bw_meter = NULL;
1392 
1393 	    /* link into table */
1394 	    LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1395 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1396 	    rt->mfc_nstall++;
1397 
1398 	} else {
1399 	    /* determine if queue has overflowed */
1400 	    if (rt->mfc_nstall > MAX_UPQ) {
1401 		MRTSTAT_INC(mrts_upq_ovflw);
1402 non_fatal:
1403 		free(rte, M_MRTABLE);
1404 		m_freem(mb0);
1405 		MFC_UNLOCK();
1406 		VIF_UNLOCK();
1407 		return (0);
1408 	    }
1409 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1410 	    rt->mfc_nstall++;
1411 	}
1412 
1413 	rte->m			= mb0;
1414 	rte->ifp		= ifp;
1415 
1416 	MFC_UNLOCK();
1417 	VIF_UNLOCK();
1418 
1419 	return 0;
1420     }
1421 }
1422 
1423 /*
1424  * Clean up the cache entry if upcall is not serviced
1425  */
1426 static void
1427 expire_upcalls(void *unused)
1428 {
1429     int i;
1430 
1431     MFC_LOCK();
1432 
1433     for (i = 0; i < mfchashsize; i++) {
1434 	struct mfc *rt, *nrt;
1435 
1436 	if (nexpire[i] == 0)
1437 	    continue;
1438 
1439 	for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1440 		nrt = LIST_NEXT(rt, mfc_hash);
1441 
1442 		if (TAILQ_EMPTY(&rt->mfc_stall))
1443 			continue;
1444 
1445 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1446 			continue;
1447 
1448 		/*
1449 		 * free the bw_meter entries
1450 		 */
1451 		while (rt->mfc_bw_meter != NULL) {
1452 		    struct bw_meter *x = rt->mfc_bw_meter;
1453 
1454 		    rt->mfc_bw_meter = x->bm_mfc_next;
1455 		    free(x, M_BWMETER);
1456 		}
1457 
1458 		MRTSTAT_INC(mrts_cache_cleanups);
1459 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1460 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1461 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1462 
1463 		expire_mfc(rt);
1464 	    }
1465     }
1466 
1467     MFC_UNLOCK();
1468 
1469     callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
1470 }
1471 
1472 /*
1473  * Packet forwarding routine once entry in the cache is made
1474  */
1475 static int
1476 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1477 {
1478     INIT_VNET_INET(curvnet);
1479     struct ip  *ip = mtod(m, struct ip *);
1480     vifi_t vifi;
1481     int plen = ip->ip_len;
1482 
1483     VIF_LOCK_ASSERT();
1484 
1485     /*
1486      * If xmt_vif is not -1, send on only the requested vif.
1487      *
1488      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1489      */
1490     if (xmt_vif < numvifs) {
1491 	if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1492 		pim_register_send(ip, viftable + xmt_vif, m, rt);
1493 	else
1494 		phyint_send(ip, viftable + xmt_vif, m);
1495 	return 1;
1496     }
1497 
1498     /*
1499      * Don't forward if it didn't arrive from the parent vif for its origin.
1500      */
1501     vifi = rt->mfc_parent;
1502     if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1503 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1504 	    __func__, ifp, (int)vifi, viftable[vifi].v_ifp);
1505 	MRTSTAT_INC(mrts_wrong_if);
1506 	++rt->mfc_wrong_if;
1507 	/*
1508 	 * If we are doing PIM assert processing, send a message
1509 	 * to the routing daemon.
1510 	 *
1511 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1512 	 * can complete the SPT switch, regardless of the type
1513 	 * of the iif (broadcast media, GRE tunnel, etc).
1514 	 */
1515 	if (pim_assert_enabled && (vifi < numvifs) && viftable[vifi].v_ifp) {
1516 
1517 	    if (ifp == &multicast_register_if)
1518 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1519 
1520 	    /* Get vifi for the incoming packet */
1521 	    for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1522 		;
1523 	    if (vifi >= numvifs)
1524 		return 0;	/* The iif is not found: ignore the packet. */
1525 
1526 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1527 		return 0;	/* WRONGVIF disabled: ignore the packet */
1528 
1529 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1530 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1531 		struct igmpmsg *im;
1532 		int hlen = ip->ip_hl << 2;
1533 		struct mbuf *mm = m_copy(m, 0, hlen);
1534 
1535 		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1536 		    mm = m_pullup(mm, hlen);
1537 		if (mm == NULL)
1538 		    return ENOBUFS;
1539 
1540 		im = mtod(mm, struct igmpmsg *);
1541 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1542 		im->im_mbz		= 0;
1543 		im->im_vif		= vifi;
1544 
1545 		MRTSTAT_INC(mrts_upcalls);
1546 
1547 		k_igmpsrc.sin_addr = im->im_src;
1548 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1549 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1550 		    MRTSTAT_INC(mrts_upq_sockfull);
1551 		    return ENOBUFS;
1552 		}
1553 	    }
1554 	}
1555 	return 0;
1556     }
1557 
1558 
1559     /* If I sourced this packet, it counts as output, else it was input. */
1560     if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1561 	viftable[vifi].v_pkt_out++;
1562 	viftable[vifi].v_bytes_out += plen;
1563     } else {
1564 	viftable[vifi].v_pkt_in++;
1565 	viftable[vifi].v_bytes_in += plen;
1566     }
1567     rt->mfc_pkt_cnt++;
1568     rt->mfc_byte_cnt += plen;
1569 
1570     /*
1571      * For each vif, decide if a copy of the packet should be forwarded.
1572      * Forward if:
1573      *		- the ttl exceeds the vif's threshold
1574      *		- there are group members downstream on interface
1575      */
1576     for (vifi = 0; vifi < numvifs; vifi++)
1577 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1578 	    viftable[vifi].v_pkt_out++;
1579 	    viftable[vifi].v_bytes_out += plen;
1580 	    if (viftable[vifi].v_flags & VIFF_REGISTER)
1581 		pim_register_send(ip, viftable + vifi, m, rt);
1582 	    else
1583 		phyint_send(ip, viftable + vifi, m);
1584 	}
1585 
1586     /*
1587      * Perform upcall-related bw measuring.
1588      */
1589     if (rt->mfc_bw_meter != NULL) {
1590 	struct bw_meter *x;
1591 	struct timeval now;
1592 
1593 	microtime(&now);
1594 	MFC_LOCK_ASSERT();
1595 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1596 	    bw_meter_receive_packet(x, plen, &now);
1597     }
1598 
1599     return 0;
1600 }
1601 
1602 /*
1603  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1604  */
1605 static int
1606 X_legal_vif_num(int vif)
1607 {
1608 	int ret;
1609 
1610 	ret = 0;
1611 	if (vif < 0)
1612 		return (ret);
1613 
1614 	VIF_LOCK();
1615 	if (vif < numvifs)
1616 		ret = 1;
1617 	VIF_UNLOCK();
1618 
1619 	return (ret);
1620 }
1621 
1622 /*
1623  * Return the local address used by this vif
1624  */
1625 static u_long
1626 X_ip_mcast_src(int vifi)
1627 {
1628 	in_addr_t addr;
1629 
1630 	addr = INADDR_ANY;
1631 	if (vifi < 0)
1632 		return (addr);
1633 
1634 	VIF_LOCK();
1635 	if (vifi < numvifs)
1636 		addr = viftable[vifi].v_lcl_addr.s_addr;
1637 	VIF_UNLOCK();
1638 
1639 	return (addr);
1640 }
1641 
1642 static void
1643 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1644 {
1645     struct mbuf *mb_copy;
1646     int hlen = ip->ip_hl << 2;
1647 
1648     VIF_LOCK_ASSERT();
1649 
1650     /*
1651      * Make a new reference to the packet; make sure that
1652      * the IP header is actually copied, not just referenced,
1653      * so that ip_output() only scribbles on the copy.
1654      */
1655     mb_copy = m_copypacket(m, M_DONTWAIT);
1656     if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1657 	mb_copy = m_pullup(mb_copy, hlen);
1658     if (mb_copy == NULL)
1659 	return;
1660 
1661     send_packet(vifp, mb_copy);
1662 }
1663 
1664 static void
1665 send_packet(struct vif *vifp, struct mbuf *m)
1666 {
1667 	struct ip_moptions imo;
1668 	struct in_multi *imm[2];
1669 	int error;
1670 
1671 	VIF_LOCK_ASSERT();
1672 
1673 	imo.imo_multicast_ifp  = vifp->v_ifp;
1674 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1675 	imo.imo_multicast_loop = 1;
1676 	imo.imo_multicast_vif  = -1;
1677 	imo.imo_num_memberships = 0;
1678 	imo.imo_max_memberships = 2;
1679 	imo.imo_membership  = &imm[0];
1680 
1681 	/*
1682 	 * Re-entrancy should not be a problem here, because
1683 	 * the packets that we send out and are looped back at us
1684 	 * should get rejected because they appear to come from
1685 	 * the loopback interface, thus preventing looping.
1686 	 */
1687 	error = ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, &imo, NULL);
1688 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1689 	    (ptrdiff_t)(vifp - viftable), error);
1690 }
1691 
1692 /*
1693  * Stubs for old RSVP socket shim implementation.
1694  */
1695 
1696 static int
1697 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1698 {
1699 
1700 	return (EOPNOTSUPP);
1701 }
1702 
1703 static void
1704 X_ip_rsvp_force_done(struct socket *so __unused)
1705 {
1706 
1707 }
1708 
1709 static void
1710 X_rsvp_input(struct mbuf *m, int off __unused)
1711 {
1712 	INIT_VNET_INET(curvnet);
1713 
1714 	if (!V_rsvp_on)
1715 		m_freem(m);
1716 }
1717 
1718 /*
1719  * Code for bandwidth monitors
1720  */
1721 
1722 /*
1723  * Define common interface for timeval-related methods
1724  */
1725 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1726 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1727 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1728 
1729 static uint32_t
1730 compute_bw_meter_flags(struct bw_upcall *req)
1731 {
1732     uint32_t flags = 0;
1733 
1734     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1735 	flags |= BW_METER_UNIT_PACKETS;
1736     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1737 	flags |= BW_METER_UNIT_BYTES;
1738     if (req->bu_flags & BW_UPCALL_GEQ)
1739 	flags |= BW_METER_GEQ;
1740     if (req->bu_flags & BW_UPCALL_LEQ)
1741 	flags |= BW_METER_LEQ;
1742 
1743     return flags;
1744 }
1745 
1746 /*
1747  * Add a bw_meter entry
1748  */
1749 static int
1750 add_bw_upcall(struct bw_upcall *req)
1751 {
1752     struct mfc *mfc;
1753     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1754 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1755     struct timeval now;
1756     struct bw_meter *x;
1757     uint32_t flags;
1758 
1759     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
1760 	return EOPNOTSUPP;
1761 
1762     /* Test if the flags are valid */
1763     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1764 	return EINVAL;
1765     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1766 	return EINVAL;
1767     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1768 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1769 	return EINVAL;
1770 
1771     /* Test if the threshold time interval is valid */
1772     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1773 	return EINVAL;
1774 
1775     flags = compute_bw_meter_flags(req);
1776 
1777     /*
1778      * Find if we have already same bw_meter entry
1779      */
1780     MFC_LOCK();
1781     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1782     if (mfc == NULL) {
1783 	MFC_UNLOCK();
1784 	return EADDRNOTAVAIL;
1785     }
1786     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1787 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1788 			   &req->bu_threshold.b_time, ==)) &&
1789 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1790 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1791 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1792 	    MFC_UNLOCK();
1793 	    return 0;		/* XXX Already installed */
1794 	}
1795     }
1796 
1797     /* Allocate the new bw_meter entry */
1798     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1799     if (x == NULL) {
1800 	MFC_UNLOCK();
1801 	return ENOBUFS;
1802     }
1803 
1804     /* Set the new bw_meter entry */
1805     x->bm_threshold.b_time = req->bu_threshold.b_time;
1806     microtime(&now);
1807     x->bm_start_time = now;
1808     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1809     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1810     x->bm_measured.b_packets = 0;
1811     x->bm_measured.b_bytes = 0;
1812     x->bm_flags = flags;
1813     x->bm_time_next = NULL;
1814     x->bm_time_hash = BW_METER_BUCKETS;
1815 
1816     /* Add the new bw_meter entry to the front of entries for this MFC */
1817     x->bm_mfc = mfc;
1818     x->bm_mfc_next = mfc->mfc_bw_meter;
1819     mfc->mfc_bw_meter = x;
1820     schedule_bw_meter(x, &now);
1821     MFC_UNLOCK();
1822 
1823     return 0;
1824 }
1825 
1826 static void
1827 free_bw_list(struct bw_meter *list)
1828 {
1829     while (list != NULL) {
1830 	struct bw_meter *x = list;
1831 
1832 	list = list->bm_mfc_next;
1833 	unschedule_bw_meter(x);
1834 	free(x, M_BWMETER);
1835     }
1836 }
1837 
1838 /*
1839  * Delete one or multiple bw_meter entries
1840  */
1841 static int
1842 del_bw_upcall(struct bw_upcall *req)
1843 {
1844     struct mfc *mfc;
1845     struct bw_meter *x;
1846 
1847     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
1848 	return EOPNOTSUPP;
1849 
1850     MFC_LOCK();
1851 
1852     /* Find the corresponding MFC entry */
1853     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1854     if (mfc == NULL) {
1855 	MFC_UNLOCK();
1856 	return EADDRNOTAVAIL;
1857     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1858 	/*
1859 	 * Delete all bw_meter entries for this mfc
1860 	 */
1861 	struct bw_meter *list;
1862 
1863 	list = mfc->mfc_bw_meter;
1864 	mfc->mfc_bw_meter = NULL;
1865 	free_bw_list(list);
1866 	MFC_UNLOCK();
1867 	return 0;
1868     } else {			/* Delete a single bw_meter entry */
1869 	struct bw_meter *prev;
1870 	uint32_t flags = 0;
1871 
1872 	flags = compute_bw_meter_flags(req);
1873 
1874 	/* Find the bw_meter entry to delete */
1875 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1876 	     prev = x, x = x->bm_mfc_next) {
1877 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1878 			       &req->bu_threshold.b_time, ==)) &&
1879 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1880 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1881 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1882 		break;
1883 	}
1884 	if (x != NULL) { /* Delete entry from the list for this MFC */
1885 	    if (prev != NULL)
1886 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1887 	    else
1888 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1889 
1890 	    unschedule_bw_meter(x);
1891 	    MFC_UNLOCK();
1892 	    /* Free the bw_meter entry */
1893 	    free(x, M_BWMETER);
1894 	    return 0;
1895 	} else {
1896 	    MFC_UNLOCK();
1897 	    return EINVAL;
1898 	}
1899     }
1900     /* NOTREACHED */
1901 }
1902 
1903 /*
1904  * Perform bandwidth measurement processing that may result in an upcall
1905  */
1906 static void
1907 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1908 {
1909     struct timeval delta;
1910 
1911     MFC_LOCK_ASSERT();
1912 
1913     delta = *nowp;
1914     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1915 
1916     if (x->bm_flags & BW_METER_GEQ) {
1917 	/*
1918 	 * Processing for ">=" type of bw_meter entry
1919 	 */
1920 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1921 	    /* Reset the bw_meter entry */
1922 	    x->bm_start_time = *nowp;
1923 	    x->bm_measured.b_packets = 0;
1924 	    x->bm_measured.b_bytes = 0;
1925 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1926 	}
1927 
1928 	/* Record that a packet is received */
1929 	x->bm_measured.b_packets++;
1930 	x->bm_measured.b_bytes += plen;
1931 
1932 	/*
1933 	 * Test if we should deliver an upcall
1934 	 */
1935 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1936 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1937 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1938 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1939 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1940 		/* Prepare an upcall for delivery */
1941 		bw_meter_prepare_upcall(x, nowp);
1942 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1943 	    }
1944 	}
1945     } else if (x->bm_flags & BW_METER_LEQ) {
1946 	/*
1947 	 * Processing for "<=" type of bw_meter entry
1948 	 */
1949 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1950 	    /*
1951 	     * We are behind time with the multicast forwarding table
1952 	     * scanning for "<=" type of bw_meter entries, so test now
1953 	     * if we should deliver an upcall.
1954 	     */
1955 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1956 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1957 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1958 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1959 		/* Prepare an upcall for delivery */
1960 		bw_meter_prepare_upcall(x, nowp);
1961 	    }
1962 	    /* Reschedule the bw_meter entry */
1963 	    unschedule_bw_meter(x);
1964 	    schedule_bw_meter(x, nowp);
1965 	}
1966 
1967 	/* Record that a packet is received */
1968 	x->bm_measured.b_packets++;
1969 	x->bm_measured.b_bytes += plen;
1970 
1971 	/*
1972 	 * Test if we should restart the measuring interval
1973 	 */
1974 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1975 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1976 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1977 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1978 	    /* Don't restart the measuring interval */
1979 	} else {
1980 	    /* Do restart the measuring interval */
1981 	    /*
1982 	     * XXX: note that we don't unschedule and schedule, because this
1983 	     * might be too much overhead per packet. Instead, when we process
1984 	     * all entries for a given timer hash bin, we check whether it is
1985 	     * really a timeout. If not, we reschedule at that time.
1986 	     */
1987 	    x->bm_start_time = *nowp;
1988 	    x->bm_measured.b_packets = 0;
1989 	    x->bm_measured.b_bytes = 0;
1990 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1991 	}
1992     }
1993 }
1994 
1995 /*
1996  * Prepare a bandwidth-related upcall
1997  */
1998 static void
1999 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2000 {
2001     struct timeval delta;
2002     struct bw_upcall *u;
2003 
2004     MFC_LOCK_ASSERT();
2005 
2006     /*
2007      * Compute the measured time interval
2008      */
2009     delta = *nowp;
2010     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2011 
2012     /*
2013      * If there are too many pending upcalls, deliver them now
2014      */
2015     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2016 	bw_upcalls_send();
2017 
2018     /*
2019      * Set the bw_upcall entry
2020      */
2021     u = &bw_upcalls[bw_upcalls_n++];
2022     u->bu_src = x->bm_mfc->mfc_origin;
2023     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2024     u->bu_threshold.b_time = x->bm_threshold.b_time;
2025     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2026     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2027     u->bu_measured.b_time = delta;
2028     u->bu_measured.b_packets = x->bm_measured.b_packets;
2029     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2030     u->bu_flags = 0;
2031     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2032 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2033     if (x->bm_flags & BW_METER_UNIT_BYTES)
2034 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2035     if (x->bm_flags & BW_METER_GEQ)
2036 	u->bu_flags |= BW_UPCALL_GEQ;
2037     if (x->bm_flags & BW_METER_LEQ)
2038 	u->bu_flags |= BW_UPCALL_LEQ;
2039 }
2040 
2041 /*
2042  * Send the pending bandwidth-related upcalls
2043  */
2044 static void
2045 bw_upcalls_send(void)
2046 {
2047     INIT_VNET_INET(curvnet);
2048     struct mbuf *m;
2049     int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2050     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2051     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2052 				      0,		/* unused2 */
2053 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2054 				      0,		/* im_mbz  */
2055 				      0,		/* im_vif  */
2056 				      0,		/* unused3 */
2057 				      { 0 },		/* im_src  */
2058 				      { 0 } };		/* im_dst  */
2059 
2060     MFC_LOCK_ASSERT();
2061 
2062     if (bw_upcalls_n == 0)
2063 	return;			/* No pending upcalls */
2064 
2065     bw_upcalls_n = 0;
2066 
2067     /*
2068      * Allocate a new mbuf, initialize it with the header and
2069      * the payload for the pending calls.
2070      */
2071     MGETHDR(m, M_DONTWAIT, MT_DATA);
2072     if (m == NULL) {
2073 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2074 	return;
2075     }
2076 
2077     m->m_len = m->m_pkthdr.len = 0;
2078     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2079     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2080 
2081     /*
2082      * Send the upcalls
2083      * XXX do we need to set the address in k_igmpsrc ?
2084      */
2085     MRTSTAT_INC(mrts_upcalls);
2086     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2087 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2088 	MRTSTAT_INC(mrts_upq_sockfull);
2089     }
2090 }
2091 
2092 /*
2093  * Compute the timeout hash value for the bw_meter entries
2094  */
2095 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2096     do {								\
2097 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2098 									\
2099 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2100 	(hash) = next_timeval.tv_sec;					\
2101 	if (next_timeval.tv_usec)					\
2102 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2103 	(hash) %= BW_METER_BUCKETS;					\
2104     } while (0)
2105 
2106 /*
2107  * Schedule a timer to process periodically bw_meter entry of type "<="
2108  * by linking the entry in the proper hash bucket.
2109  */
2110 static void
2111 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2112 {
2113     int time_hash;
2114 
2115     MFC_LOCK_ASSERT();
2116 
2117     if (!(x->bm_flags & BW_METER_LEQ))
2118 	return;		/* XXX: we schedule timers only for "<=" entries */
2119 
2120     /*
2121      * Reset the bw_meter entry
2122      */
2123     x->bm_start_time = *nowp;
2124     x->bm_measured.b_packets = 0;
2125     x->bm_measured.b_bytes = 0;
2126     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2127 
2128     /*
2129      * Compute the timeout hash value and insert the entry
2130      */
2131     BW_METER_TIMEHASH(x, time_hash);
2132     x->bm_time_next = bw_meter_timers[time_hash];
2133     bw_meter_timers[time_hash] = x;
2134     x->bm_time_hash = time_hash;
2135 }
2136 
2137 /*
2138  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2139  * by removing the entry from the proper hash bucket.
2140  */
2141 static void
2142 unschedule_bw_meter(struct bw_meter *x)
2143 {
2144     int time_hash;
2145     struct bw_meter *prev, *tmp;
2146 
2147     MFC_LOCK_ASSERT();
2148 
2149     if (!(x->bm_flags & BW_METER_LEQ))
2150 	return;		/* XXX: we schedule timers only for "<=" entries */
2151 
2152     /*
2153      * Compute the timeout hash value and delete the entry
2154      */
2155     time_hash = x->bm_time_hash;
2156     if (time_hash >= BW_METER_BUCKETS)
2157 	return;		/* Entry was not scheduled */
2158 
2159     for (prev = NULL, tmp = bw_meter_timers[time_hash];
2160 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2161 	if (tmp == x)
2162 	    break;
2163 
2164     if (tmp == NULL)
2165 	panic("unschedule_bw_meter: bw_meter entry not found");
2166 
2167     if (prev != NULL)
2168 	prev->bm_time_next = x->bm_time_next;
2169     else
2170 	bw_meter_timers[time_hash] = x->bm_time_next;
2171 
2172     x->bm_time_next = NULL;
2173     x->bm_time_hash = BW_METER_BUCKETS;
2174 }
2175 
2176 
2177 /*
2178  * Process all "<=" type of bw_meter that should be processed now,
2179  * and for each entry prepare an upcall if necessary. Each processed
2180  * entry is rescheduled again for the (periodic) processing.
2181  *
2182  * This is run periodically (once per second normally). On each round,
2183  * all the potentially matching entries are in the hash slot that we are
2184  * looking at.
2185  */
2186 static void
2187 bw_meter_process()
2188 {
2189     static uint32_t last_tv_sec;	/* last time we processed this */
2190 
2191     uint32_t loops;
2192     int i;
2193     struct timeval now, process_endtime;
2194 
2195     microtime(&now);
2196     if (last_tv_sec == now.tv_sec)
2197 	return;		/* nothing to do */
2198 
2199     loops = now.tv_sec - last_tv_sec;
2200     last_tv_sec = now.tv_sec;
2201     if (loops > BW_METER_BUCKETS)
2202 	loops = BW_METER_BUCKETS;
2203 
2204     MFC_LOCK();
2205     /*
2206      * Process all bins of bw_meter entries from the one after the last
2207      * processed to the current one. On entry, i points to the last bucket
2208      * visited, so we need to increment i at the beginning of the loop.
2209      */
2210     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2211 	struct bw_meter *x, *tmp_list;
2212 
2213 	if (++i >= BW_METER_BUCKETS)
2214 	    i = 0;
2215 
2216 	/* Disconnect the list of bw_meter entries from the bin */
2217 	tmp_list = bw_meter_timers[i];
2218 	bw_meter_timers[i] = NULL;
2219 
2220 	/* Process the list of bw_meter entries */
2221 	while (tmp_list != NULL) {
2222 	    x = tmp_list;
2223 	    tmp_list = tmp_list->bm_time_next;
2224 
2225 	    /* Test if the time interval is over */
2226 	    process_endtime = x->bm_start_time;
2227 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2228 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2229 		/* Not yet: reschedule, but don't reset */
2230 		int time_hash;
2231 
2232 		BW_METER_TIMEHASH(x, time_hash);
2233 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2234 		    /*
2235 		     * XXX: somehow the bin processing is a bit ahead of time.
2236 		     * Put the entry in the next bin.
2237 		     */
2238 		    if (++time_hash >= BW_METER_BUCKETS)
2239 			time_hash = 0;
2240 		}
2241 		x->bm_time_next = bw_meter_timers[time_hash];
2242 		bw_meter_timers[time_hash] = x;
2243 		x->bm_time_hash = time_hash;
2244 
2245 		continue;
2246 	    }
2247 
2248 	    /*
2249 	     * Test if we should deliver an upcall
2250 	     */
2251 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2252 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2253 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2254 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2255 		/* Prepare an upcall for delivery */
2256 		bw_meter_prepare_upcall(x, &now);
2257 	    }
2258 
2259 	    /*
2260 	     * Reschedule for next processing
2261 	     */
2262 	    schedule_bw_meter(x, &now);
2263 	}
2264     }
2265 
2266     /* Send all upcalls that are pending delivery */
2267     bw_upcalls_send();
2268 
2269     MFC_UNLOCK();
2270 }
2271 
2272 /*
2273  * A periodic function for sending all upcalls that are pending delivery
2274  */
2275 static void
2276 expire_bw_upcalls_send(void *unused)
2277 {
2278     MFC_LOCK();
2279     bw_upcalls_send();
2280     MFC_UNLOCK();
2281 
2282     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2283 	expire_bw_upcalls_send, NULL);
2284 }
2285 
2286 /*
2287  * A periodic function for periodic scanning of the multicast forwarding
2288  * table for processing all "<=" bw_meter entries.
2289  */
2290 static void
2291 expire_bw_meter_process(void *unused)
2292 {
2293     if (mrt_api_config & MRT_MFC_BW_UPCALL)
2294 	bw_meter_process();
2295 
2296     callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
2297 }
2298 
2299 /*
2300  * End of bandwidth monitoring code
2301  */
2302 
2303 /*
2304  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2305  *
2306  */
2307 static int
2308 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2309     struct mfc *rt)
2310 {
2311     struct mbuf *mb_copy, *mm;
2312 
2313     /*
2314      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2315      * rendezvous point was unspecified, and we were told not to.
2316      */
2317     if (pim_squelch_wholepkt != 0 && (mrt_api_config & MRT_MFC_RP) &&
2318 	in_nullhost(rt->mfc_rp))
2319 	return 0;
2320 
2321     mb_copy = pim_register_prepare(ip, m);
2322     if (mb_copy == NULL)
2323 	return ENOBUFS;
2324 
2325     /*
2326      * Send all the fragments. Note that the mbuf for each fragment
2327      * is freed by the sending machinery.
2328      */
2329     for (mm = mb_copy; mm; mm = mb_copy) {
2330 	mb_copy = mm->m_nextpkt;
2331 	mm->m_nextpkt = 0;
2332 	mm = m_pullup(mm, sizeof(struct ip));
2333 	if (mm != NULL) {
2334 	    ip = mtod(mm, struct ip *);
2335 	    if ((mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2336 		pim_register_send_rp(ip, vifp, mm, rt);
2337 	    } else {
2338 		pim_register_send_upcall(ip, vifp, mm, rt);
2339 	    }
2340 	}
2341     }
2342 
2343     return 0;
2344 }
2345 
2346 /*
2347  * Return a copy of the data packet that is ready for PIM Register
2348  * encapsulation.
2349  * XXX: Note that in the returned copy the IP header is a valid one.
2350  */
2351 static struct mbuf *
2352 pim_register_prepare(struct ip *ip, struct mbuf *m)
2353 {
2354     struct mbuf *mb_copy = NULL;
2355     int mtu;
2356 
2357     /* Take care of delayed checksums */
2358     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2359 	in_delayed_cksum(m);
2360 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2361     }
2362 
2363     /*
2364      * Copy the old packet & pullup its IP header into the
2365      * new mbuf so we can modify it.
2366      */
2367     mb_copy = m_copypacket(m, M_DONTWAIT);
2368     if (mb_copy == NULL)
2369 	return NULL;
2370     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2371     if (mb_copy == NULL)
2372 	return NULL;
2373 
2374     /* take care of the TTL */
2375     ip = mtod(mb_copy, struct ip *);
2376     --ip->ip_ttl;
2377 
2378     /* Compute the MTU after the PIM Register encapsulation */
2379     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2380 
2381     if (ip->ip_len <= mtu) {
2382 	/* Turn the IP header into a valid one */
2383 	ip->ip_len = htons(ip->ip_len);
2384 	ip->ip_off = htons(ip->ip_off);
2385 	ip->ip_sum = 0;
2386 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2387     } else {
2388 	/* Fragment the packet */
2389 	if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
2390 	    m_freem(mb_copy);
2391 	    return NULL;
2392 	}
2393     }
2394     return mb_copy;
2395 }
2396 
2397 /*
2398  * Send an upcall with the data packet to the user-level process.
2399  */
2400 static int
2401 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2402     struct mbuf *mb_copy, struct mfc *rt)
2403 {
2404     INIT_VNET_INET(curvnet);
2405     struct mbuf *mb_first;
2406     int len = ntohs(ip->ip_len);
2407     struct igmpmsg *im;
2408     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2409 
2410     VIF_LOCK_ASSERT();
2411 
2412     /*
2413      * Add a new mbuf with an upcall header
2414      */
2415     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2416     if (mb_first == NULL) {
2417 	m_freem(mb_copy);
2418 	return ENOBUFS;
2419     }
2420     mb_first->m_data += max_linkhdr;
2421     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2422     mb_first->m_len = sizeof(struct igmpmsg);
2423     mb_first->m_next = mb_copy;
2424 
2425     /* Send message to routing daemon */
2426     im = mtod(mb_first, struct igmpmsg *);
2427     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2428     im->im_mbz		= 0;
2429     im->im_vif		= vifp - viftable;
2430     im->im_src		= ip->ip_src;
2431     im->im_dst		= ip->ip_dst;
2432 
2433     k_igmpsrc.sin_addr	= ip->ip_src;
2434 
2435     MRTSTAT_INC(mrts_upcalls);
2436 
2437     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2438 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2439 	MRTSTAT_INC(mrts_upq_sockfull);
2440 	return ENOBUFS;
2441     }
2442 
2443     /* Keep statistics */
2444     PIMSTAT_INC(pims_snd_registers_msgs);
2445     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2446 
2447     return 0;
2448 }
2449 
2450 /*
2451  * Encapsulate the data packet in PIM Register message and send it to the RP.
2452  */
2453 static int
2454 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2455     struct mfc *rt)
2456 {
2457     INIT_VNET_INET(curvnet);
2458     struct mbuf *mb_first;
2459     struct ip *ip_outer;
2460     struct pim_encap_pimhdr *pimhdr;
2461     int len = ntohs(ip->ip_len);
2462     vifi_t vifi = rt->mfc_parent;
2463 
2464     VIF_LOCK_ASSERT();
2465 
2466     if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
2467 	m_freem(mb_copy);
2468 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2469     }
2470 
2471     /*
2472      * Add a new mbuf with the encapsulating header
2473      */
2474     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2475     if (mb_first == NULL) {
2476 	m_freem(mb_copy);
2477 	return ENOBUFS;
2478     }
2479     mb_first->m_data += max_linkhdr;
2480     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2481     mb_first->m_next = mb_copy;
2482 
2483     mb_first->m_pkthdr.len = len + mb_first->m_len;
2484 
2485     /*
2486      * Fill in the encapsulating IP and PIM header
2487      */
2488     ip_outer = mtod(mb_first, struct ip *);
2489     *ip_outer = pim_encap_iphdr;
2490     ip_outer->ip_id = ip_newid();
2491     ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2492     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
2493     ip_outer->ip_dst = rt->mfc_rp;
2494     /*
2495      * Copy the inner header TOS to the outer header, and take care of the
2496      * IP_DF bit.
2497      */
2498     ip_outer->ip_tos = ip->ip_tos;
2499     if (ntohs(ip->ip_off) & IP_DF)
2500 	ip_outer->ip_off |= IP_DF;
2501     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2502 					 + sizeof(pim_encap_iphdr));
2503     *pimhdr = pim_encap_pimhdr;
2504     /* If the iif crosses a border, set the Border-bit */
2505     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
2506 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2507 
2508     mb_first->m_data += sizeof(pim_encap_iphdr);
2509     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2510     mb_first->m_data -= sizeof(pim_encap_iphdr);
2511 
2512     send_packet(vifp, mb_first);
2513 
2514     /* Keep statistics */
2515     PIMSTAT_INC(pims_snd_registers_msgs);
2516     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2517 
2518     return 0;
2519 }
2520 
2521 /*
2522  * pim_encapcheck() is called by the encap4_input() path at runtime to
2523  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2524  * into the kernel.
2525  */
2526 static int
2527 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2528 {
2529 
2530 #ifdef DIAGNOSTIC
2531     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2532 #endif
2533     if (proto != IPPROTO_PIM)
2534 	return 0;	/* not for us; reject the datagram. */
2535 
2536     return 64;		/* claim the datagram. */
2537 }
2538 
2539 /*
2540  * PIM-SMv2 and PIM-DM messages processing.
2541  * Receives and verifies the PIM control messages, and passes them
2542  * up to the listening socket, using rip_input().
2543  * The only message with special processing is the PIM_REGISTER message
2544  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2545  * is passed to if_simloop().
2546  */
2547 void
2548 pim_input(struct mbuf *m, int off)
2549 {
2550     struct ip *ip = mtod(m, struct ip *);
2551     struct pim *pim;
2552     int minlen;
2553     int datalen = ip->ip_len;
2554     int ip_tos;
2555     int iphlen = off;
2556 
2557     /* Keep statistics */
2558     PIMSTAT_INC(pims_rcv_total_msgs);
2559     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2560 
2561     /*
2562      * Validate lengths
2563      */
2564     if (datalen < PIM_MINLEN) {
2565 	PIMSTAT_INC(pims_rcv_tooshort);
2566 	CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2567 	    __func__, datalen, inet_ntoa(ip->ip_src));
2568 	m_freem(m);
2569 	return;
2570     }
2571 
2572     /*
2573      * If the packet is at least as big as a REGISTER, go agead
2574      * and grab the PIM REGISTER header size, to avoid another
2575      * possible m_pullup() later.
2576      *
2577      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2578      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2579      */
2580     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2581     /*
2582      * Get the IP and PIM headers in contiguous memory, and
2583      * possibly the PIM REGISTER header.
2584      */
2585     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
2586 	(m = m_pullup(m, minlen)) == 0) {
2587 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2588 	return;
2589     }
2590 
2591     /* m_pullup() may have given us a new mbuf so reset ip. */
2592     ip = mtod(m, struct ip *);
2593     ip_tos = ip->ip_tos;
2594 
2595     /* adjust mbuf to point to the PIM header */
2596     m->m_data += iphlen;
2597     m->m_len  -= iphlen;
2598     pim = mtod(m, struct pim *);
2599 
2600     /*
2601      * Validate checksum. If PIM REGISTER, exclude the data packet.
2602      *
2603      * XXX: some older PIMv2 implementations don't make this distinction,
2604      * so for compatibility reason perform the checksum over part of the
2605      * message, and if error, then over the whole message.
2606      */
2607     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2608 	/* do nothing, checksum okay */
2609     } else if (in_cksum(m, datalen)) {
2610 	PIMSTAT_INC(pims_rcv_badsum);
2611 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2612 	m_freem(m);
2613 	return;
2614     }
2615 
2616     /* PIM version check */
2617     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2618 	PIMSTAT_INC(pims_rcv_badversion);
2619 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2620 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2621 	m_freem(m);
2622 	return;
2623     }
2624 
2625     /* restore mbuf back to the outer IP */
2626     m->m_data -= iphlen;
2627     m->m_len  += iphlen;
2628 
2629     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2630 	/*
2631 	 * Since this is a REGISTER, we'll make a copy of the register
2632 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2633 	 * routing daemon.
2634 	 */
2635 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2636 	struct mbuf *mcp;
2637 	struct ip *encap_ip;
2638 	u_int32_t *reghdr;
2639 	struct ifnet *vifp;
2640 
2641 	VIF_LOCK();
2642 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
2643 	    VIF_UNLOCK();
2644 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2645 		(int)reg_vif_num);
2646 	    m_freem(m);
2647 	    return;
2648 	}
2649 	/* XXX need refcnt? */
2650 	vifp = viftable[reg_vif_num].v_ifp;
2651 	VIF_UNLOCK();
2652 
2653 	/*
2654 	 * Validate length
2655 	 */
2656 	if (datalen < PIM_REG_MINLEN) {
2657 	    PIMSTAT_INC(pims_rcv_tooshort);
2658 	    PIMSTAT_INC(pims_rcv_badregisters);
2659 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2660 	    m_freem(m);
2661 	    return;
2662 	}
2663 
2664 	reghdr = (u_int32_t *)(pim + 1);
2665 	encap_ip = (struct ip *)(reghdr + 1);
2666 
2667 	CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2668 	    __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2669 
2670 	/* verify the version number of the inner packet */
2671 	if (encap_ip->ip_v != IPVERSION) {
2672 	    PIMSTAT_INC(pims_rcv_badregisters);
2673 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2674 	    m_freem(m);
2675 	    return;
2676 	}
2677 
2678 	/* verify the inner packet is destined to a mcast group */
2679 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2680 	    PIMSTAT_INC(pims_rcv_badregisters);
2681 	    CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2682 		inet_ntoa(encap_ip->ip_dst));
2683 	    m_freem(m);
2684 	    return;
2685 	}
2686 
2687 	/* If a NULL_REGISTER, pass it to the daemon */
2688 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2689 	    goto pim_input_to_daemon;
2690 
2691 	/*
2692 	 * Copy the TOS from the outer IP header to the inner IP header.
2693 	 */
2694 	if (encap_ip->ip_tos != ip_tos) {
2695 	    /* Outer TOS -> inner TOS */
2696 	    encap_ip->ip_tos = ip_tos;
2697 	    /* Recompute the inner header checksum. Sigh... */
2698 
2699 	    /* adjust mbuf to point to the inner IP header */
2700 	    m->m_data += (iphlen + PIM_MINLEN);
2701 	    m->m_len  -= (iphlen + PIM_MINLEN);
2702 
2703 	    encap_ip->ip_sum = 0;
2704 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2705 
2706 	    /* restore mbuf to point back to the outer IP header */
2707 	    m->m_data -= (iphlen + PIM_MINLEN);
2708 	    m->m_len  += (iphlen + PIM_MINLEN);
2709 	}
2710 
2711 	/*
2712 	 * Decapsulate the inner IP packet and loopback to forward it
2713 	 * as a normal multicast packet. Also, make a copy of the
2714 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2715 	 * to pass to the daemon later, so it can take the appropriate
2716 	 * actions (e.g., send back PIM_REGISTER_STOP).
2717 	 * XXX: here m->m_data points to the outer IP header.
2718 	 */
2719 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2720 	if (mcp == NULL) {
2721 	    CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2722 	    m_freem(m);
2723 	    return;
2724 	}
2725 
2726 	/* Keep statistics */
2727 	/* XXX: registers_bytes include only the encap. mcast pkt */
2728 	PIMSTAT_INC(pims_rcv_registers_msgs);
2729 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2730 
2731 	/*
2732 	 * forward the inner ip packet; point m_data at the inner ip.
2733 	 */
2734 	m_adj(m, iphlen + PIM_MINLEN);
2735 
2736 	CTR4(KTR_IPMF,
2737 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2738 	    __func__,
2739 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2740 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2741 	    (int)reg_vif_num);
2742 
2743 	/* NB: vifp was collected above; can it change on us? */
2744 	if_simloop(vifp, m, dst.sin_family, 0);
2745 
2746 	/* prepare the register head to send to the mrouting daemon */
2747 	m = mcp;
2748     }
2749 
2750 pim_input_to_daemon:
2751     /*
2752      * Pass the PIM message up to the daemon; if it is a Register message,
2753      * pass the 'head' only up to the daemon. This includes the
2754      * outer IP header, PIM header, PIM-Register header and the
2755      * inner IP header.
2756      * XXX: the outer IP header pkt size of a Register is not adjust to
2757      * reflect the fact that the inner multicast data is truncated.
2758      */
2759     rip_input(m, iphlen);
2760 
2761     return;
2762 }
2763 
2764 static int
2765 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2766 {
2767 	struct mfc	*rt;
2768 	int		 error, i;
2769 
2770 	if (req->newptr)
2771 		return (EPERM);
2772 	if (mfchashtbl == NULL)	/* XXX unlocked */
2773 		return (0);
2774 	error = sysctl_wire_old_buffer(req, 0);
2775 	if (error)
2776 		return (error);
2777 
2778 	MFC_LOCK();
2779 	for (i = 0; i < mfchashsize; i++) {
2780 		LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
2781 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2782 			if (error)
2783 				goto out_locked;
2784 		}
2785 	}
2786 out_locked:
2787 	MFC_UNLOCK();
2788 	return (error);
2789 }
2790 
2791 SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, sysctl_mfctable,
2792     "IPv4 Multicast Forwarding Table (struct *mfc[mfchashsize], "
2793     "netinet/ip_mroute.h)");
2794 
2795 static int
2796 ip_mroute_modevent(module_t mod, int type, void *unused)
2797 {
2798     INIT_VNET_INET(curvnet);
2799 
2800     switch (type) {
2801     case MOD_LOAD:
2802 	MROUTER_LOCK_INIT();
2803 	MFC_LOCK_INIT();
2804 	VIF_LOCK_INIT();
2805 
2806 	mfchashsize = MFCHASHSIZE;
2807 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2808 	    !powerof2(mfchashsize)) {
2809 		printf("WARNING: %s not a power of 2; using default\n",
2810 		    "net.inet.ip.mfchashsize");
2811 		mfchashsize = MFCHASHSIZE;
2812 	}
2813 	MALLOC(nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2814 
2815 	pim_squelch_wholepkt = 0;
2816 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2817 	    &pim_squelch_wholepkt);
2818 	ip_mrouter_reset();
2819 
2820 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2821 	    pim_encapcheck, &in_pim_protosw, NULL);
2822 	if (pim_encap_cookie == NULL) {
2823 		printf("ip_mroute: unable to attach pim encap\n");
2824 		VIF_LOCK_DESTROY();
2825 		MFC_LOCK_DESTROY();
2826 		MROUTER_LOCK_DESTROY();
2827 		return (EINVAL);
2828 	}
2829 
2830 	ip_mcast_src = X_ip_mcast_src;
2831 	ip_mforward = X_ip_mforward;
2832 	ip_mrouter_done = X_ip_mrouter_done;
2833 	ip_mrouter_get = X_ip_mrouter_get;
2834 	ip_mrouter_set = X_ip_mrouter_set;
2835 
2836 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2837 	ip_rsvp_vif = X_ip_rsvp_vif;
2838 
2839 	legal_vif_num = X_legal_vif_num;
2840 	mrt_ioctl = X_mrt_ioctl;
2841 	rsvp_input_p = X_rsvp_input;
2842 	break;
2843 
2844     case MOD_UNLOAD:
2845 	/*
2846 	 * Typically module unload happens after the user-level
2847 	 * process has shutdown the kernel services (the check
2848 	 * below insures someone can't just yank the module out
2849 	 * from under a running process).  But if the module is
2850 	 * just loaded and then unloaded w/o starting up a user
2851 	 * process we still need to cleanup.
2852 	 */
2853 	if (V_ip_mrouter != NULL)
2854 	    return (EINVAL);
2855 
2856 	if (pim_encap_cookie) {
2857 	    encap_detach(pim_encap_cookie);
2858 	    pim_encap_cookie = NULL;
2859 	}
2860 	X_ip_mrouter_done();
2861 
2862 	FREE(nexpire, M_MRTABLE);
2863 	nexpire = NULL;
2864 
2865 	ip_mcast_src = NULL;
2866 	ip_mforward = NULL;
2867 	ip_mrouter_done = NULL;
2868 	ip_mrouter_get = NULL;
2869 	ip_mrouter_set = NULL;
2870 
2871 	ip_rsvp_force_done = NULL;
2872 	ip_rsvp_vif = NULL;
2873 
2874 	legal_vif_num = NULL;
2875 	mrt_ioctl = NULL;
2876 	rsvp_input_p = NULL;
2877 
2878 	VIF_LOCK_DESTROY();
2879 	MFC_LOCK_DESTROY();
2880 	MROUTER_LOCK_DESTROY();
2881 	break;
2882 
2883     default:
2884 	return EOPNOTSUPP;
2885     }
2886     return 0;
2887 }
2888 
2889 static moduledata_t ip_mroutemod = {
2890     "ip_mroute",
2891     ip_mroute_modevent,
2892     0
2893 };
2894 
2895 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
2896