xref: /freebsd/sys/netinet/ip_mroute.c (revision 4ed925457ab06e83238a5db33e89ccc94b99a713)
1 /*-
2  * Copyright (c) 1989 Stephen Deering
3  * Copyright (c) 1992, 1993
4  *      The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Stephen Deering of Stanford University.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
34  */
35 
36 /*
37  * IP multicast forwarding procedures
38  *
39  * Written by David Waitzman, BBN Labs, August 1988.
40  * Modified by Steve Deering, Stanford, February 1989.
41  * Modified by Mark J. Steiglitz, Stanford, May, 1991
42  * Modified by Van Jacobson, LBL, January 1993
43  * Modified by Ajit Thyagarajan, PARC, August 1993
44  * Modified by Bill Fenner, PARC, April 1995
45  * Modified by Ahmed Helmy, SGI, June 1996
46  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
47  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
48  * Modified by Hitoshi Asaeda, WIDE, August 2000
49  * Modified by Pavlin Radoslavov, ICSI, October 2002
50  *
51  * MROUTING Revision: 3.5
52  * and PIM-SMv2 and PIM-DM support, advanced API support,
53  * bandwidth metering and signaling
54  */
55 
56 /*
57  * TODO: Prefix functions with ipmf_.
58  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
59  * domain attachment (if_afdata) so we can track consumers of that service.
60  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
61  * move it to socket options.
62  * TODO: Cleanup LSRR removal further.
63  * TODO: Push RSVP stubs into raw_ip.c.
64  * TODO: Use bitstring.h for vif set.
65  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
66  * TODO: Sync ip6_mroute.c with this file.
67  */
68 
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
71 
72 #include "opt_inet.h"
73 #include "opt_mrouting.h"
74 
75 #define _PIM_VT 1
76 
77 #include <sys/param.h>
78 #include <sys/kernel.h>
79 #include <sys/stddef.h>
80 #include <sys/lock.h>
81 #include <sys/ktr.h>
82 #include <sys/malloc.h>
83 #include <sys/mbuf.h>
84 #include <sys/module.h>
85 #include <sys/priv.h>
86 #include <sys/protosw.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/socketvar.h>
90 #include <sys/sockio.h>
91 #include <sys/sx.h>
92 #include <sys/sysctl.h>
93 #include <sys/syslog.h>
94 #include <sys/systm.h>
95 #include <sys/time.h>
96 
97 #include <net/if.h>
98 #include <net/netisr.h>
99 #include <net/route.h>
100 #include <net/vnet.h>
101 
102 #include <netinet/in.h>
103 #include <netinet/igmp.h>
104 #include <netinet/in_systm.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip.h>
107 #include <netinet/ip_encap.h>
108 #include <netinet/ip_mroute.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/ip_options.h>
111 #include <netinet/pim.h>
112 #include <netinet/pim_var.h>
113 #include <netinet/udp.h>
114 
115 #include <machine/in_cksum.h>
116 
117 #ifndef KTR_IPMF
118 #define KTR_IPMF KTR_INET
119 #endif
120 
121 #define		VIFI_INVALID	((vifi_t) -1)
122 #define		M_HASCL(m)	((m)->m_flags & M_EXT)
123 
124 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
125 
126 /*
127  * Locking.  We use two locks: one for the virtual interface table and
128  * one for the forwarding table.  These locks may be nested in which case
129  * the VIF lock must always be taken first.  Note that each lock is used
130  * to cover not only the specific data structure but also related data
131  * structures.
132  */
133 
134 static struct mtx mrouter_mtx;
135 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
136 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
137 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
138 #define	MROUTER_LOCK_INIT()						\
139 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
140 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
141 
142 static struct mrtstat	mrtstat;
143 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
144     &mrtstat, mrtstat,
145     "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
146     "netinet/ip_mroute.h)");
147 
148 static u_long			 mfchash;
149 #define MFCHASH(a, g)							\
150 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
151 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
152 #define MFCHASHSIZE	256
153 
154 static u_char			*nexpire;	/* 0..mfchashsize-1 */
155 static u_long			 mfchashsize;	/* Hash size */
156 LIST_HEAD(mfchashhdr, mfc)	*mfchashtbl;
157 
158 static struct mtx mfc_mtx;
159 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
160 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
161 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
162 #define	MFC_LOCK_INIT()							\
163 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
164 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
165 
166 static vifi_t		numvifs;
167 static struct vif	viftable[MAXVIFS];
168 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
169     &viftable, sizeof(viftable), "S,vif[MAXVIFS]",
170     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
171 
172 static struct mtx vif_mtx;
173 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
174 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
175 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
176 #define	VIF_LOCK_INIT()							\
177 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
178 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
179 
180 static eventhandler_tag if_detach_event_tag = NULL;
181 
182 static struct callout expire_upcalls_ch;
183 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
184 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
185 
186 /*
187  * Bandwidth meter variables and constants
188  */
189 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
190 /*
191  * Pending timeouts are stored in a hash table, the key being the
192  * expiration time. Periodically, the entries are analysed and processed.
193  */
194 #define BW_METER_BUCKETS	1024
195 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
196 static struct callout bw_meter_ch;
197 #define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
198 
199 /*
200  * Pending upcalls are stored in a vector which is flushed when
201  * full, or periodically
202  */
203 static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
204 static u_int	bw_upcalls_n; /* # of pending upcalls */
205 static struct callout bw_upcalls_ch;
206 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
207 
208 static struct pimstat pimstat;
209 
210 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
211 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
212     &pimstat, pimstat,
213     "PIM Statistics (struct pimstat, netinet/pim_var.h)");
214 
215 static u_long	pim_squelch_wholepkt = 0;
216 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
217     &pim_squelch_wholepkt, 0,
218     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
219 
220 extern  struct domain inetdomain;
221 static const struct protosw in_pim_protosw = {
222 	.pr_type =		SOCK_RAW,
223 	.pr_domain =		&inetdomain,
224 	.pr_protocol =		IPPROTO_PIM,
225 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
226 	.pr_input =		pim_input,
227 	.pr_output =		(pr_output_t*)rip_output,
228 	.pr_ctloutput =		rip_ctloutput,
229 	.pr_usrreqs =		&rip_usrreqs
230 };
231 static const struct encaptab *pim_encap_cookie;
232 
233 static int pim_encapcheck(const struct mbuf *, int, int, void *);
234 
235 /*
236  * Note: the PIM Register encapsulation adds the following in front of a
237  * data packet:
238  *
239  * struct pim_encap_hdr {
240  *    struct ip ip;
241  *    struct pim_encap_pimhdr  pim;
242  * }
243  *
244  */
245 
246 struct pim_encap_pimhdr {
247 	struct pim pim;
248 	uint32_t   flags;
249 };
250 #define		PIM_ENCAP_TTL	64
251 
252 static struct ip pim_encap_iphdr = {
253 #if BYTE_ORDER == LITTLE_ENDIAN
254 	sizeof(struct ip) >> 2,
255 	IPVERSION,
256 #else
257 	IPVERSION,
258 	sizeof(struct ip) >> 2,
259 #endif
260 	0,			/* tos */
261 	sizeof(struct ip),	/* total length */
262 	0,			/* id */
263 	0,			/* frag offset */
264 	PIM_ENCAP_TTL,
265 	IPPROTO_PIM,
266 	0,			/* checksum */
267 };
268 
269 static struct pim_encap_pimhdr pim_encap_pimhdr = {
270     {
271 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
272 	0,			/* reserved */
273 	0,			/* checksum */
274     },
275     0				/* flags */
276 };
277 
278 static struct ifnet multicast_register_if;
279 static vifi_t reg_vif_num = VIFI_INVALID;
280 
281 /*
282  * Private variables.
283  */
284 
285 static u_long	X_ip_mcast_src(int);
286 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
287 		    struct ip_moptions *);
288 static int	X_ip_mrouter_done(void);
289 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
290 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
291 static int	X_legal_vif_num(int);
292 static int	X_mrt_ioctl(u_long, caddr_t, int);
293 
294 static int	add_bw_upcall(struct bw_upcall *);
295 static int	add_mfc(struct mfcctl2 *);
296 static int	add_vif(struct vifctl *);
297 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
298 static void	bw_meter_process(void);
299 static void	bw_meter_receive_packet(struct bw_meter *, int,
300 		    struct timeval *);
301 static void	bw_upcalls_send(void);
302 static int	del_bw_upcall(struct bw_upcall *);
303 static int	del_mfc(struct mfcctl2 *);
304 static int	del_vif(vifi_t);
305 static int	del_vif_locked(vifi_t);
306 static void	expire_bw_meter_process(void *);
307 static void	expire_bw_upcalls_send(void *);
308 static void	expire_mfc(struct mfc *);
309 static void	expire_upcalls(void *);
310 static void	free_bw_list(struct bw_meter *);
311 static int	get_sg_cnt(struct sioc_sg_req *);
312 static int	get_vif_cnt(struct sioc_vif_req *);
313 static void	if_detached_event(void *, struct ifnet *);
314 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
315 static int	ip_mrouter_init(struct socket *, int);
316 static __inline struct mfc *
317 		mfc_find(struct in_addr *, struct in_addr *);
318 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
319 static struct mbuf *
320 		pim_register_prepare(struct ip *, struct mbuf *);
321 static int	pim_register_send(struct ip *, struct vif *,
322 		    struct mbuf *, struct mfc *);
323 static int	pim_register_send_rp(struct ip *, struct vif *,
324 		    struct mbuf *, struct mfc *);
325 static int	pim_register_send_upcall(struct ip *, struct vif *,
326 		    struct mbuf *, struct mfc *);
327 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
328 static void	send_packet(struct vif *, struct mbuf *);
329 static int	set_api_config(uint32_t *);
330 static int	set_assert(int);
331 static int	socket_send(struct socket *, struct mbuf *,
332 		    struct sockaddr_in *);
333 static void	unschedule_bw_meter(struct bw_meter *);
334 
335 /*
336  * Kernel multicast forwarding API capabilities and setup.
337  * If more API capabilities are added to the kernel, they should be
338  * recorded in `mrt_api_support'.
339  */
340 #define MRT_API_VERSION		0x0305
341 
342 static const int mrt_api_version = MRT_API_VERSION;
343 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
344 					 MRT_MFC_FLAGS_BORDER_VIF |
345 					 MRT_MFC_RP |
346 					 MRT_MFC_BW_UPCALL);
347 static uint32_t mrt_api_config = 0;
348 
349 static int pim_assert_enabled;
350 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
351 
352 /*
353  * Find a route for a given origin IP address and multicast group address.
354  * Statistics must be updated by the caller.
355  */
356 static __inline struct mfc *
357 mfc_find(struct in_addr *o, struct in_addr *g)
358 {
359 	struct mfc *rt;
360 
361 	MFC_LOCK_ASSERT();
362 
363 	LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
364 		if (in_hosteq(rt->mfc_origin, *o) &&
365 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
366 		    TAILQ_EMPTY(&rt->mfc_stall))
367 			break;
368 	}
369 
370 	return (rt);
371 }
372 
373 /*
374  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
375  */
376 static int
377 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
378 {
379     int	error, optval;
380     vifi_t	vifi;
381     struct	vifctl vifc;
382     struct	mfcctl2 mfc;
383     struct	bw_upcall bw_upcall;
384     uint32_t	i;
385 
386     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
387 	return EPERM;
388 
389     error = 0;
390     switch (sopt->sopt_name) {
391     case MRT_INIT:
392 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
393 	if (error)
394 	    break;
395 	error = ip_mrouter_init(so, optval);
396 	break;
397 
398     case MRT_DONE:
399 	error = ip_mrouter_done();
400 	break;
401 
402     case MRT_ADD_VIF:
403 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
404 	if (error)
405 	    break;
406 	error = add_vif(&vifc);
407 	break;
408 
409     case MRT_DEL_VIF:
410 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
411 	if (error)
412 	    break;
413 	error = del_vif(vifi);
414 	break;
415 
416     case MRT_ADD_MFC:
417     case MRT_DEL_MFC:
418 	/*
419 	 * select data size depending on API version.
420 	 */
421 	if (sopt->sopt_name == MRT_ADD_MFC &&
422 		mrt_api_config & MRT_API_FLAGS_ALL) {
423 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
424 				sizeof(struct mfcctl2));
425 	} else {
426 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
427 				sizeof(struct mfcctl));
428 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
429 			sizeof(mfc) - sizeof(struct mfcctl));
430 	}
431 	if (error)
432 	    break;
433 	if (sopt->sopt_name == MRT_ADD_MFC)
434 	    error = add_mfc(&mfc);
435 	else
436 	    error = del_mfc(&mfc);
437 	break;
438 
439     case MRT_ASSERT:
440 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
441 	if (error)
442 	    break;
443 	set_assert(optval);
444 	break;
445 
446     case MRT_API_CONFIG:
447 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
448 	if (!error)
449 	    error = set_api_config(&i);
450 	if (!error)
451 	    error = sooptcopyout(sopt, &i, sizeof i);
452 	break;
453 
454     case MRT_ADD_BW_UPCALL:
455     case MRT_DEL_BW_UPCALL:
456 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
457 				sizeof bw_upcall);
458 	if (error)
459 	    break;
460 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
461 	    error = add_bw_upcall(&bw_upcall);
462 	else
463 	    error = del_bw_upcall(&bw_upcall);
464 	break;
465 
466     default:
467 	error = EOPNOTSUPP;
468 	break;
469     }
470     return error;
471 }
472 
473 /*
474  * Handle MRT getsockopt commands
475  */
476 static int
477 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
478 {
479     int error;
480 
481     switch (sopt->sopt_name) {
482     case MRT_VERSION:
483 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
484 	break;
485 
486     case MRT_ASSERT:
487 	error = sooptcopyout(sopt, &pim_assert_enabled,
488 	    sizeof pim_assert_enabled);
489 	break;
490 
491     case MRT_API_SUPPORT:
492 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
493 	break;
494 
495     case MRT_API_CONFIG:
496 	error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config);
497 	break;
498 
499     default:
500 	error = EOPNOTSUPP;
501 	break;
502     }
503     return error;
504 }
505 
506 /*
507  * Handle ioctl commands to obtain information from the cache
508  */
509 static int
510 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
511 {
512     int error = 0;
513 
514     /*
515      * Currently the only function calling this ioctl routine is rtioctl().
516      * Typically, only root can create the raw socket in order to execute
517      * this ioctl method, however the request might be coming from a prison
518      */
519     error = priv_check(curthread, PRIV_NETINET_MROUTE);
520     if (error)
521 	return (error);
522     switch (cmd) {
523     case (SIOCGETVIFCNT):
524 	error = get_vif_cnt((struct sioc_vif_req *)data);
525 	break;
526 
527     case (SIOCGETSGCNT):
528 	error = get_sg_cnt((struct sioc_sg_req *)data);
529 	break;
530 
531     default:
532 	error = EINVAL;
533 	break;
534     }
535     return error;
536 }
537 
538 /*
539  * returns the packet, byte, rpf-failure count for the source group provided
540  */
541 static int
542 get_sg_cnt(struct sioc_sg_req *req)
543 {
544     struct mfc *rt;
545 
546     MFC_LOCK();
547     rt = mfc_find(&req->src, &req->grp);
548     if (rt == NULL) {
549 	MFC_UNLOCK();
550 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
551 	return EADDRNOTAVAIL;
552     }
553     req->pktcnt = rt->mfc_pkt_cnt;
554     req->bytecnt = rt->mfc_byte_cnt;
555     req->wrong_if = rt->mfc_wrong_if;
556     MFC_UNLOCK();
557     return 0;
558 }
559 
560 /*
561  * returns the input and output packet and byte counts on the vif provided
562  */
563 static int
564 get_vif_cnt(struct sioc_vif_req *req)
565 {
566     vifi_t vifi = req->vifi;
567 
568     VIF_LOCK();
569     if (vifi >= numvifs) {
570 	VIF_UNLOCK();
571 	return EINVAL;
572     }
573 
574     req->icount = viftable[vifi].v_pkt_in;
575     req->ocount = viftable[vifi].v_pkt_out;
576     req->ibytes = viftable[vifi].v_bytes_in;
577     req->obytes = viftable[vifi].v_bytes_out;
578     VIF_UNLOCK();
579 
580     return 0;
581 }
582 
583 static void
584 ip_mrouter_reset(void)
585 {
586 
587     pim_assert_enabled = 0;
588     mrt_api_config = 0;
589 
590     callout_init(&expire_upcalls_ch, CALLOUT_MPSAFE);
591 
592     bw_upcalls_n = 0;
593     bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers));
594     callout_init(&bw_upcalls_ch, CALLOUT_MPSAFE);
595     callout_init(&bw_meter_ch, CALLOUT_MPSAFE);
596 }
597 
598 static void
599 if_detached_event(void *arg __unused, struct ifnet *ifp)
600 {
601     vifi_t vifi;
602     int i;
603 
604     MROUTER_LOCK();
605 
606     if (V_ip_mrouter == NULL) {
607 	MROUTER_UNLOCK();
608 	return;
609     }
610 
611     VIF_LOCK();
612     MFC_LOCK();
613 
614     /*
615      * Tear down multicast forwarder state associated with this ifnet.
616      * 1. Walk the vif list, matching vifs against this ifnet.
617      * 2. Walk the multicast forwarding cache (mfc) looking for
618      *    inner matches with this vif's index.
619      * 3. Expire any matching multicast forwarding cache entries.
620      * 4. Free vif state. This should disable ALLMULTI on the interface.
621      */
622     for (vifi = 0; vifi < numvifs; vifi++) {
623 	if (viftable[vifi].v_ifp != ifp)
624 		continue;
625 	for (i = 0; i < mfchashsize; i++) {
626 		struct mfc *rt, *nrt;
627 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
628 			nrt = LIST_NEXT(rt, mfc_hash);
629 			if (rt->mfc_parent == vifi) {
630 				expire_mfc(rt);
631 			}
632 		}
633 	}
634 	del_vif_locked(vifi);
635     }
636 
637     MFC_UNLOCK();
638     VIF_UNLOCK();
639 
640     MROUTER_UNLOCK();
641 }
642 
643 /*
644  * Enable multicast forwarding.
645  */
646 static int
647 ip_mrouter_init(struct socket *so, int version)
648 {
649 
650     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
651         so->so_type, so->so_proto->pr_protocol);
652 
653     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
654 	return EOPNOTSUPP;
655 
656     if (version != 1)
657 	return ENOPROTOOPT;
658 
659     MROUTER_LOCK();
660 
661     if (V_ip_mrouter != NULL) {
662 	MROUTER_UNLOCK();
663 	return EADDRINUSE;
664     }
665 
666     if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
667         if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
668     if (if_detach_event_tag == NULL) {
669 	MROUTER_UNLOCK();
670 	return (ENOMEM);
671     }
672 
673     mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &mfchash, HASH_NOWAIT);
674 
675     callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
676 
677     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
678 	expire_bw_upcalls_send, NULL);
679     callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
680 
681     V_ip_mrouter = so;
682 
683     MROUTER_UNLOCK();
684 
685     CTR1(KTR_IPMF, "%s: done", __func__);
686 
687     return 0;
688 }
689 
690 /*
691  * Disable multicast forwarding.
692  */
693 static int
694 X_ip_mrouter_done(void)
695 {
696     vifi_t vifi;
697     int i;
698     struct ifnet *ifp;
699     struct ifreq ifr;
700 
701     MROUTER_LOCK();
702 
703     if (V_ip_mrouter == NULL) {
704 	MROUTER_UNLOCK();
705 	return EINVAL;
706     }
707 
708     /*
709      * Detach/disable hooks to the reset of the system.
710      */
711     V_ip_mrouter = NULL;
712     mrt_api_config = 0;
713 
714     VIF_LOCK();
715 
716     /*
717      * For each phyint in use, disable promiscuous reception of all IP
718      * multicasts.
719      */
720     for (vifi = 0; vifi < numvifs; vifi++) {
721 	if (!in_nullhost(viftable[vifi].v_lcl_addr) &&
722 		!(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
723 	    struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
724 
725 	    so->sin_len = sizeof(struct sockaddr_in);
726 	    so->sin_family = AF_INET;
727 	    so->sin_addr.s_addr = INADDR_ANY;
728 	    ifp = viftable[vifi].v_ifp;
729 	    if_allmulti(ifp, 0);
730 	}
731     }
732     bzero((caddr_t)viftable, sizeof(viftable));
733     numvifs = 0;
734     pim_assert_enabled = 0;
735 
736     VIF_UNLOCK();
737 
738     EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
739 
740     callout_stop(&expire_upcalls_ch);
741     callout_stop(&bw_upcalls_ch);
742     callout_stop(&bw_meter_ch);
743 
744     MFC_LOCK();
745 
746     /*
747      * Free all multicast forwarding cache entries.
748      * Do not use hashdestroy(), as we must perform other cleanup.
749      */
750     for (i = 0; i < mfchashsize; i++) {
751 	struct mfc *rt, *nrt;
752 	for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
753 		nrt = LIST_NEXT(rt, mfc_hash);
754 		expire_mfc(rt);
755 	}
756     }
757     free(mfchashtbl, M_MRTABLE);
758     mfchashtbl = NULL;
759 
760     bzero(nexpire, sizeof(nexpire[0]) * mfchashsize);
761 
762     bw_upcalls_n = 0;
763     bzero(bw_meter_timers, sizeof(bw_meter_timers));
764 
765     MFC_UNLOCK();
766 
767     reg_vif_num = VIFI_INVALID;
768 
769     MROUTER_UNLOCK();
770 
771     CTR1(KTR_IPMF, "%s: done", __func__);
772 
773     return 0;
774 }
775 
776 /*
777  * Set PIM assert processing global
778  */
779 static int
780 set_assert(int i)
781 {
782     if ((i != 1) && (i != 0))
783 	return EINVAL;
784 
785     pim_assert_enabled = i;
786 
787     return 0;
788 }
789 
790 /*
791  * Configure API capabilities
792  */
793 int
794 set_api_config(uint32_t *apival)
795 {
796     int i;
797 
798     /*
799      * We can set the API capabilities only if it is the first operation
800      * after MRT_INIT. I.e.:
801      *  - there are no vifs installed
802      *  - pim_assert is not enabled
803      *  - the MFC table is empty
804      */
805     if (numvifs > 0) {
806 	*apival = 0;
807 	return EPERM;
808     }
809     if (pim_assert_enabled) {
810 	*apival = 0;
811 	return EPERM;
812     }
813 
814     MFC_LOCK();
815 
816     for (i = 0; i < mfchashsize; i++) {
817 	if (LIST_FIRST(&mfchashtbl[i]) != NULL) {
818 	    *apival = 0;
819 	    return EPERM;
820 	}
821     }
822 
823     MFC_UNLOCK();
824 
825     mrt_api_config = *apival & mrt_api_support;
826     *apival = mrt_api_config;
827 
828     return 0;
829 }
830 
831 /*
832  * Add a vif to the vif table
833  */
834 static int
835 add_vif(struct vifctl *vifcp)
836 {
837     struct vif *vifp = viftable + vifcp->vifc_vifi;
838     struct sockaddr_in sin = {sizeof sin, AF_INET};
839     struct ifaddr *ifa;
840     struct ifnet *ifp;
841     int error;
842 
843     VIF_LOCK();
844     if (vifcp->vifc_vifi >= MAXVIFS) {
845 	VIF_UNLOCK();
846 	return EINVAL;
847     }
848     /* rate limiting is no longer supported by this code */
849     if (vifcp->vifc_rate_limit != 0) {
850 	log(LOG_ERR, "rate limiting is no longer supported\n");
851 	VIF_UNLOCK();
852 	return EINVAL;
853     }
854     if (!in_nullhost(vifp->v_lcl_addr)) {
855 	VIF_UNLOCK();
856 	return EADDRINUSE;
857     }
858     if (in_nullhost(vifcp->vifc_lcl_addr)) {
859 	VIF_UNLOCK();
860 	return EADDRNOTAVAIL;
861     }
862 
863     /* Find the interface with an address in AF_INET family */
864     if (vifcp->vifc_flags & VIFF_REGISTER) {
865 	/*
866 	 * XXX: Because VIFF_REGISTER does not really need a valid
867 	 * local interface (e.g. it could be 127.0.0.2), we don't
868 	 * check its address.
869 	 */
870 	ifp = NULL;
871     } else {
872 	sin.sin_addr = vifcp->vifc_lcl_addr;
873 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
874 	if (ifa == NULL) {
875 	    VIF_UNLOCK();
876 	    return EADDRNOTAVAIL;
877 	}
878 	ifp = ifa->ifa_ifp;
879 	ifa_free(ifa);
880     }
881 
882     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
883 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
884 	VIF_UNLOCK();
885 	return EOPNOTSUPP;
886     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
887 	ifp = &multicast_register_if;
888 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
889 	if (reg_vif_num == VIFI_INVALID) {
890 	    if_initname(&multicast_register_if, "register_vif", 0);
891 	    multicast_register_if.if_flags = IFF_LOOPBACK;
892 	    reg_vif_num = vifcp->vifc_vifi;
893 	}
894     } else {		/* Make sure the interface supports multicast */
895 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
896 	    VIF_UNLOCK();
897 	    return EOPNOTSUPP;
898 	}
899 
900 	/* Enable promiscuous reception of all IP multicasts from the if */
901 	error = if_allmulti(ifp, 1);
902 	if (error) {
903 	    VIF_UNLOCK();
904 	    return error;
905 	}
906     }
907 
908     vifp->v_flags     = vifcp->vifc_flags;
909     vifp->v_threshold = vifcp->vifc_threshold;
910     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
911     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
912     vifp->v_ifp       = ifp;
913     /* initialize per vif pkt counters */
914     vifp->v_pkt_in    = 0;
915     vifp->v_pkt_out   = 0;
916     vifp->v_bytes_in  = 0;
917     vifp->v_bytes_out = 0;
918     bzero(&vifp->v_route, sizeof(vifp->v_route));
919 
920     /* Adjust numvifs up if the vifi is higher than numvifs */
921     if (numvifs <= vifcp->vifc_vifi)
922 	numvifs = vifcp->vifc_vifi + 1;
923 
924     VIF_UNLOCK();
925 
926     CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
927 	(int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr),
928 	(int)vifcp->vifc_threshold);
929 
930     return 0;
931 }
932 
933 /*
934  * Delete a vif from the vif table
935  */
936 static int
937 del_vif_locked(vifi_t vifi)
938 {
939     struct vif *vifp;
940 
941     VIF_LOCK_ASSERT();
942 
943     if (vifi >= numvifs) {
944 	return EINVAL;
945     }
946     vifp = &viftable[vifi];
947     if (in_nullhost(vifp->v_lcl_addr)) {
948 	return EADDRNOTAVAIL;
949     }
950 
951     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
952 	if_allmulti(vifp->v_ifp, 0);
953 
954     if (vifp->v_flags & VIFF_REGISTER)
955 	reg_vif_num = VIFI_INVALID;
956 
957     bzero((caddr_t)vifp, sizeof (*vifp));
958 
959     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
960 
961     /* Adjust numvifs down */
962     for (vifi = numvifs; vifi > 0; vifi--)
963 	if (!in_nullhost(viftable[vifi-1].v_lcl_addr))
964 	    break;
965     numvifs = vifi;
966 
967     return 0;
968 }
969 
970 static int
971 del_vif(vifi_t vifi)
972 {
973     int cc;
974 
975     VIF_LOCK();
976     cc = del_vif_locked(vifi);
977     VIF_UNLOCK();
978 
979     return cc;
980 }
981 
982 /*
983  * update an mfc entry without resetting counters and S,G addresses.
984  */
985 static void
986 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
987 {
988     int i;
989 
990     rt->mfc_parent = mfccp->mfcc_parent;
991     for (i = 0; i < numvifs; i++) {
992 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
993 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
994 	    MRT_MFC_FLAGS_ALL;
995     }
996     /* set the RP address */
997     if (mrt_api_config & MRT_MFC_RP)
998 	rt->mfc_rp = mfccp->mfcc_rp;
999     else
1000 	rt->mfc_rp.s_addr = INADDR_ANY;
1001 }
1002 
1003 /*
1004  * fully initialize an mfc entry from the parameter.
1005  */
1006 static void
1007 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1008 {
1009     rt->mfc_origin     = mfccp->mfcc_origin;
1010     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1011 
1012     update_mfc_params(rt, mfccp);
1013 
1014     /* initialize pkt counters per src-grp */
1015     rt->mfc_pkt_cnt    = 0;
1016     rt->mfc_byte_cnt   = 0;
1017     rt->mfc_wrong_if   = 0;
1018     timevalclear(&rt->mfc_last_assert);
1019 }
1020 
1021 static void
1022 expire_mfc(struct mfc *rt)
1023 {
1024 	struct rtdetq *rte, *nrte;
1025 
1026 	MFC_LOCK_ASSERT();
1027 
1028 	free_bw_list(rt->mfc_bw_meter);
1029 
1030 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1031 		m_freem(rte->m);
1032 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1033 		free(rte, M_MRTABLE);
1034 	}
1035 
1036 	LIST_REMOVE(rt, mfc_hash);
1037 	free(rt, M_MRTABLE);
1038 }
1039 
1040 /*
1041  * Add an mfc entry
1042  */
1043 static int
1044 add_mfc(struct mfcctl2 *mfccp)
1045 {
1046     struct mfc *rt;
1047     struct rtdetq *rte, *nrte;
1048     u_long hash = 0;
1049     u_short nstl;
1050 
1051     VIF_LOCK();
1052     MFC_LOCK();
1053 
1054     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1055 
1056     /* If an entry already exists, just update the fields */
1057     if (rt) {
1058 	CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x",
1059 	    __func__, inet_ntoa(mfccp->mfcc_origin),
1060 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1061 	    mfccp->mfcc_parent);
1062 	update_mfc_params(rt, mfccp);
1063 	MFC_UNLOCK();
1064 	VIF_UNLOCK();
1065 	return (0);
1066     }
1067 
1068     /*
1069      * Find the entry for which the upcall was made and update
1070      */
1071     nstl = 0;
1072     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1073     LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1074 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1075 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1076 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1077 		CTR5(KTR_IPMF,
1078 		    "%s: add mfc orig %s group %lx parent %x qh %p",
1079 		    __func__, inet_ntoa(mfccp->mfcc_origin),
1080 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1081 		    mfccp->mfcc_parent,
1082 		    TAILQ_FIRST(&rt->mfc_stall));
1083 		if (nstl++)
1084 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1085 
1086 		init_mfc_params(rt, mfccp);
1087 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1088 		nexpire[hash]--;
1089 
1090 		/* Free queued packets, but attempt to forward them first. */
1091 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1092 			if (rte->ifp != NULL)
1093 				ip_mdq(rte->m, rte->ifp, rt, -1);
1094 			m_freem(rte->m);
1095 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1096 			rt->mfc_nstall--;
1097 			free(rte, M_MRTABLE);
1098 		}
1099 	}
1100     }
1101 
1102     /*
1103      * It is possible that an entry is being inserted without an upcall
1104      */
1105     if (nstl == 0) {
1106 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1107 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1108 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1109 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1110 			init_mfc_params(rt, mfccp);
1111 			if (rt->mfc_expire)
1112 			    nexpire[hash]--;
1113 			rt->mfc_expire = 0;
1114 			break; /* XXX */
1115 		}
1116 	}
1117 
1118 	if (rt == NULL) {		/* no upcall, so make a new entry */
1119 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1120 	    if (rt == NULL) {
1121 		MFC_UNLOCK();
1122 		VIF_UNLOCK();
1123 		return (ENOBUFS);
1124 	    }
1125 
1126 	    init_mfc_params(rt, mfccp);
1127 	    TAILQ_INIT(&rt->mfc_stall);
1128 	    rt->mfc_nstall = 0;
1129 
1130 	    rt->mfc_expire     = 0;
1131 	    rt->mfc_bw_meter = NULL;
1132 
1133 	    /* insert new entry at head of hash chain */
1134 	    LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1135 	}
1136     }
1137 
1138     MFC_UNLOCK();
1139     VIF_UNLOCK();
1140 
1141     return (0);
1142 }
1143 
1144 /*
1145  * Delete an mfc entry
1146  */
1147 static int
1148 del_mfc(struct mfcctl2 *mfccp)
1149 {
1150     struct in_addr	origin;
1151     struct in_addr	mcastgrp;
1152     struct mfc		*rt;
1153 
1154     origin = mfccp->mfcc_origin;
1155     mcastgrp = mfccp->mfcc_mcastgrp;
1156 
1157     CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__,
1158 	inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr));
1159 
1160     MFC_LOCK();
1161 
1162     rt = mfc_find(&origin, &mcastgrp);
1163     if (rt == NULL) {
1164 	MFC_UNLOCK();
1165 	return EADDRNOTAVAIL;
1166     }
1167 
1168     /*
1169      * free the bw_meter entries
1170      */
1171     free_bw_list(rt->mfc_bw_meter);
1172     rt->mfc_bw_meter = NULL;
1173 
1174     LIST_REMOVE(rt, mfc_hash);
1175     free(rt, M_MRTABLE);
1176 
1177     MFC_UNLOCK();
1178 
1179     return (0);
1180 }
1181 
1182 /*
1183  * Send a message to the routing daemon on the multicast routing socket.
1184  */
1185 static int
1186 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1187 {
1188     if (s) {
1189 	SOCKBUF_LOCK(&s->so_rcv);
1190 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1191 	    NULL) != 0) {
1192 	    sorwakeup_locked(s);
1193 	    return 0;
1194 	}
1195 	SOCKBUF_UNLOCK(&s->so_rcv);
1196     }
1197     m_freem(mm);
1198     return -1;
1199 }
1200 
1201 /*
1202  * IP multicast forwarding function. This function assumes that the packet
1203  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1204  * pointed to by "ifp", and the packet is to be relayed to other networks
1205  * that have members of the packet's destination IP multicast group.
1206  *
1207  * The packet is returned unscathed to the caller, unless it is
1208  * erroneous, in which case a non-zero return value tells the caller to
1209  * discard it.
1210  */
1211 
1212 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1213 
1214 static int
1215 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1216     struct ip_moptions *imo)
1217 {
1218     struct mfc *rt;
1219     int error;
1220     vifi_t vifi;
1221 
1222     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p",
1223 	inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1224 
1225     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1226 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1227 	/*
1228 	 * Packet arrived via a physical interface or
1229 	 * an encapsulated tunnel or a register_vif.
1230 	 */
1231     } else {
1232 	/*
1233 	 * Packet arrived through a source-route tunnel.
1234 	 * Source-route tunnels are no longer supported.
1235 	 */
1236 	return (1);
1237     }
1238 
1239     VIF_LOCK();
1240     MFC_LOCK();
1241     if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1242 	if (ip->ip_ttl < MAXTTL)
1243 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1244 	error = ip_mdq(m, ifp, NULL, vifi);
1245 	MFC_UNLOCK();
1246 	VIF_UNLOCK();
1247 	return error;
1248     }
1249 
1250     /*
1251      * Don't forward a packet with time-to-live of zero or one,
1252      * or a packet destined to a local-only group.
1253      */
1254     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1255 	MFC_UNLOCK();
1256 	VIF_UNLOCK();
1257 	return 0;
1258     }
1259 
1260     /*
1261      * Determine forwarding vifs from the forwarding cache table
1262      */
1263     MRTSTAT_INC(mrts_mfc_lookups);
1264     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1265 
1266     /* Entry exists, so forward if necessary */
1267     if (rt != NULL) {
1268 	error = ip_mdq(m, ifp, rt, -1);
1269 	MFC_UNLOCK();
1270 	VIF_UNLOCK();
1271 	return error;
1272     } else {
1273 	/*
1274 	 * If we don't have a route for packet's origin,
1275 	 * Make a copy of the packet & send message to routing daemon
1276 	 */
1277 
1278 	struct mbuf *mb0;
1279 	struct rtdetq *rte;
1280 	u_long hash;
1281 	int hlen = ip->ip_hl << 2;
1282 
1283 	MRTSTAT_INC(mrts_mfc_misses);
1284 	MRTSTAT_INC(mrts_no_route);
1285 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)",
1286 	    inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr));
1287 
1288 	/*
1289 	 * Allocate mbufs early so that we don't do extra work if we are
1290 	 * just going to fail anyway.  Make sure to pullup the header so
1291 	 * that other people can't step on it.
1292 	 */
1293 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1294 	    M_NOWAIT|M_ZERO);
1295 	if (rte == NULL) {
1296 	    MFC_UNLOCK();
1297 	    VIF_UNLOCK();
1298 	    return ENOBUFS;
1299 	}
1300 
1301 	mb0 = m_copypacket(m, M_DONTWAIT);
1302 	if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1303 	    mb0 = m_pullup(mb0, hlen);
1304 	if (mb0 == NULL) {
1305 	    free(rte, M_MRTABLE);
1306 	    MFC_UNLOCK();
1307 	    VIF_UNLOCK();
1308 	    return ENOBUFS;
1309 	}
1310 
1311 	/* is there an upcall waiting for this flow ? */
1312 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1313 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1314 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1315 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1316 		    !TAILQ_EMPTY(&rt->mfc_stall))
1317 			break;
1318 	}
1319 
1320 	if (rt == NULL) {
1321 	    int i;
1322 	    struct igmpmsg *im;
1323 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1324 	    struct mbuf *mm;
1325 
1326 	    /*
1327 	     * Locate the vifi for the incoming interface for this packet.
1328 	     * If none found, drop packet.
1329 	     */
1330 	    for (vifi = 0; vifi < numvifs &&
1331 		    viftable[vifi].v_ifp != ifp; vifi++)
1332 		;
1333 	    if (vifi >= numvifs)	/* vif not found, drop packet */
1334 		goto non_fatal;
1335 
1336 	    /* no upcall, so make a new entry */
1337 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1338 	    if (rt == NULL)
1339 		goto fail;
1340 
1341 	    /* Make a copy of the header to send to the user level process */
1342 	    mm = m_copy(mb0, 0, hlen);
1343 	    if (mm == NULL)
1344 		goto fail1;
1345 
1346 	    /*
1347 	     * Send message to routing daemon to install
1348 	     * a route into the kernel table
1349 	     */
1350 
1351 	    im = mtod(mm, struct igmpmsg *);
1352 	    im->im_msgtype = IGMPMSG_NOCACHE;
1353 	    im->im_mbz = 0;
1354 	    im->im_vif = vifi;
1355 
1356 	    MRTSTAT_INC(mrts_upcalls);
1357 
1358 	    k_igmpsrc.sin_addr = ip->ip_src;
1359 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1360 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1361 		MRTSTAT_INC(mrts_upq_sockfull);
1362 fail1:
1363 		free(rt, M_MRTABLE);
1364 fail:
1365 		free(rte, M_MRTABLE);
1366 		m_freem(mb0);
1367 		MFC_UNLOCK();
1368 		VIF_UNLOCK();
1369 		return ENOBUFS;
1370 	    }
1371 
1372 	    /* insert new entry at head of hash chain */
1373 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1374 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1375 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1376 	    nexpire[hash]++;
1377 	    for (i = 0; i < numvifs; i++) {
1378 		rt->mfc_ttls[i] = 0;
1379 		rt->mfc_flags[i] = 0;
1380 	    }
1381 	    rt->mfc_parent = -1;
1382 
1383 	    /* clear the RP address */
1384 	    rt->mfc_rp.s_addr = INADDR_ANY;
1385 	    rt->mfc_bw_meter = NULL;
1386 
1387 	    /* initialize pkt counters per src-grp */
1388 	    rt->mfc_pkt_cnt = 0;
1389 	    rt->mfc_byte_cnt = 0;
1390 	    rt->mfc_wrong_if = 0;
1391 	    timevalclear(&rt->mfc_last_assert);
1392 
1393 	    TAILQ_INIT(&rt->mfc_stall);
1394 	    rt->mfc_nstall = 0;
1395 
1396 	    /* link into table */
1397 	    LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1398 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1399 	    rt->mfc_nstall++;
1400 
1401 	} else {
1402 	    /* determine if queue has overflowed */
1403 	    if (rt->mfc_nstall > MAX_UPQ) {
1404 		MRTSTAT_INC(mrts_upq_ovflw);
1405 non_fatal:
1406 		free(rte, M_MRTABLE);
1407 		m_freem(mb0);
1408 		MFC_UNLOCK();
1409 		VIF_UNLOCK();
1410 		return (0);
1411 	    }
1412 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1413 	    rt->mfc_nstall++;
1414 	}
1415 
1416 	rte->m			= mb0;
1417 	rte->ifp		= ifp;
1418 
1419 	MFC_UNLOCK();
1420 	VIF_UNLOCK();
1421 
1422 	return 0;
1423     }
1424 }
1425 
1426 /*
1427  * Clean up the cache entry if upcall is not serviced
1428  */
1429 static void
1430 expire_upcalls(void *unused)
1431 {
1432     int i;
1433 
1434     MFC_LOCK();
1435 
1436     for (i = 0; i < mfchashsize; i++) {
1437 	struct mfc *rt, *nrt;
1438 
1439 	if (nexpire[i] == 0)
1440 	    continue;
1441 
1442 	for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1443 		nrt = LIST_NEXT(rt, mfc_hash);
1444 
1445 		if (TAILQ_EMPTY(&rt->mfc_stall))
1446 			continue;
1447 
1448 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1449 			continue;
1450 
1451 		/*
1452 		 * free the bw_meter entries
1453 		 */
1454 		while (rt->mfc_bw_meter != NULL) {
1455 		    struct bw_meter *x = rt->mfc_bw_meter;
1456 
1457 		    rt->mfc_bw_meter = x->bm_mfc_next;
1458 		    free(x, M_BWMETER);
1459 		}
1460 
1461 		MRTSTAT_INC(mrts_cache_cleanups);
1462 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1463 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1464 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1465 
1466 		expire_mfc(rt);
1467 	    }
1468     }
1469 
1470     MFC_UNLOCK();
1471 
1472     callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
1473 }
1474 
1475 /*
1476  * Packet forwarding routine once entry in the cache is made
1477  */
1478 static int
1479 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1480 {
1481     struct ip  *ip = mtod(m, struct ip *);
1482     vifi_t vifi;
1483     int plen = ip->ip_len;
1484 
1485     VIF_LOCK_ASSERT();
1486 
1487     /*
1488      * If xmt_vif is not -1, send on only the requested vif.
1489      *
1490      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1491      */
1492     if (xmt_vif < numvifs) {
1493 	if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1494 		pim_register_send(ip, viftable + xmt_vif, m, rt);
1495 	else
1496 		phyint_send(ip, viftable + xmt_vif, m);
1497 	return 1;
1498     }
1499 
1500     /*
1501      * Don't forward if it didn't arrive from the parent vif for its origin.
1502      */
1503     vifi = rt->mfc_parent;
1504     if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1505 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1506 	    __func__, ifp, (int)vifi, viftable[vifi].v_ifp);
1507 	MRTSTAT_INC(mrts_wrong_if);
1508 	++rt->mfc_wrong_if;
1509 	/*
1510 	 * If we are doing PIM assert processing, send a message
1511 	 * to the routing daemon.
1512 	 *
1513 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1514 	 * can complete the SPT switch, regardless of the type
1515 	 * of the iif (broadcast media, GRE tunnel, etc).
1516 	 */
1517 	if (pim_assert_enabled && (vifi < numvifs) && viftable[vifi].v_ifp) {
1518 
1519 	    if (ifp == &multicast_register_if)
1520 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1521 
1522 	    /* Get vifi for the incoming packet */
1523 	    for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1524 		;
1525 	    if (vifi >= numvifs)
1526 		return 0;	/* The iif is not found: ignore the packet. */
1527 
1528 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1529 		return 0;	/* WRONGVIF disabled: ignore the packet */
1530 
1531 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1532 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1533 		struct igmpmsg *im;
1534 		int hlen = ip->ip_hl << 2;
1535 		struct mbuf *mm = m_copy(m, 0, hlen);
1536 
1537 		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1538 		    mm = m_pullup(mm, hlen);
1539 		if (mm == NULL)
1540 		    return ENOBUFS;
1541 
1542 		im = mtod(mm, struct igmpmsg *);
1543 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1544 		im->im_mbz		= 0;
1545 		im->im_vif		= vifi;
1546 
1547 		MRTSTAT_INC(mrts_upcalls);
1548 
1549 		k_igmpsrc.sin_addr = im->im_src;
1550 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1551 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1552 		    MRTSTAT_INC(mrts_upq_sockfull);
1553 		    return ENOBUFS;
1554 		}
1555 	    }
1556 	}
1557 	return 0;
1558     }
1559 
1560 
1561     /* If I sourced this packet, it counts as output, else it was input. */
1562     if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1563 	viftable[vifi].v_pkt_out++;
1564 	viftable[vifi].v_bytes_out += plen;
1565     } else {
1566 	viftable[vifi].v_pkt_in++;
1567 	viftable[vifi].v_bytes_in += plen;
1568     }
1569     rt->mfc_pkt_cnt++;
1570     rt->mfc_byte_cnt += plen;
1571 
1572     /*
1573      * For each vif, decide if a copy of the packet should be forwarded.
1574      * Forward if:
1575      *		- the ttl exceeds the vif's threshold
1576      *		- there are group members downstream on interface
1577      */
1578     for (vifi = 0; vifi < numvifs; vifi++)
1579 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1580 	    viftable[vifi].v_pkt_out++;
1581 	    viftable[vifi].v_bytes_out += plen;
1582 	    if (viftable[vifi].v_flags & VIFF_REGISTER)
1583 		pim_register_send(ip, viftable + vifi, m, rt);
1584 	    else
1585 		phyint_send(ip, viftable + vifi, m);
1586 	}
1587 
1588     /*
1589      * Perform upcall-related bw measuring.
1590      */
1591     if (rt->mfc_bw_meter != NULL) {
1592 	struct bw_meter *x;
1593 	struct timeval now;
1594 
1595 	microtime(&now);
1596 	MFC_LOCK_ASSERT();
1597 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1598 	    bw_meter_receive_packet(x, plen, &now);
1599     }
1600 
1601     return 0;
1602 }
1603 
1604 /*
1605  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1606  */
1607 static int
1608 X_legal_vif_num(int vif)
1609 {
1610 	int ret;
1611 
1612 	ret = 0;
1613 	if (vif < 0)
1614 		return (ret);
1615 
1616 	VIF_LOCK();
1617 	if (vif < numvifs)
1618 		ret = 1;
1619 	VIF_UNLOCK();
1620 
1621 	return (ret);
1622 }
1623 
1624 /*
1625  * Return the local address used by this vif
1626  */
1627 static u_long
1628 X_ip_mcast_src(int vifi)
1629 {
1630 	in_addr_t addr;
1631 
1632 	addr = INADDR_ANY;
1633 	if (vifi < 0)
1634 		return (addr);
1635 
1636 	VIF_LOCK();
1637 	if (vifi < numvifs)
1638 		addr = viftable[vifi].v_lcl_addr.s_addr;
1639 	VIF_UNLOCK();
1640 
1641 	return (addr);
1642 }
1643 
1644 static void
1645 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1646 {
1647     struct mbuf *mb_copy;
1648     int hlen = ip->ip_hl << 2;
1649 
1650     VIF_LOCK_ASSERT();
1651 
1652     /*
1653      * Make a new reference to the packet; make sure that
1654      * the IP header is actually copied, not just referenced,
1655      * so that ip_output() only scribbles on the copy.
1656      */
1657     mb_copy = m_copypacket(m, M_DONTWAIT);
1658     if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1659 	mb_copy = m_pullup(mb_copy, hlen);
1660     if (mb_copy == NULL)
1661 	return;
1662 
1663     send_packet(vifp, mb_copy);
1664 }
1665 
1666 static void
1667 send_packet(struct vif *vifp, struct mbuf *m)
1668 {
1669 	struct ip_moptions imo;
1670 	struct in_multi *imm[2];
1671 	int error;
1672 
1673 	VIF_LOCK_ASSERT();
1674 
1675 	imo.imo_multicast_ifp  = vifp->v_ifp;
1676 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1677 	imo.imo_multicast_loop = 1;
1678 	imo.imo_multicast_vif  = -1;
1679 	imo.imo_num_memberships = 0;
1680 	imo.imo_max_memberships = 2;
1681 	imo.imo_membership  = &imm[0];
1682 
1683 	/*
1684 	 * Re-entrancy should not be a problem here, because
1685 	 * the packets that we send out and are looped back at us
1686 	 * should get rejected because they appear to come from
1687 	 * the loopback interface, thus preventing looping.
1688 	 */
1689 	error = ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, &imo, NULL);
1690 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1691 	    (ptrdiff_t)(vifp - viftable), error);
1692 }
1693 
1694 /*
1695  * Stubs for old RSVP socket shim implementation.
1696  */
1697 
1698 static int
1699 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1700 {
1701 
1702 	return (EOPNOTSUPP);
1703 }
1704 
1705 static void
1706 X_ip_rsvp_force_done(struct socket *so __unused)
1707 {
1708 
1709 }
1710 
1711 static void
1712 X_rsvp_input(struct mbuf *m, int off __unused)
1713 {
1714 
1715 	if (!V_rsvp_on)
1716 		m_freem(m);
1717 }
1718 
1719 /*
1720  * Code for bandwidth monitors
1721  */
1722 
1723 /*
1724  * Define common interface for timeval-related methods
1725  */
1726 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1727 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1728 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1729 
1730 static uint32_t
1731 compute_bw_meter_flags(struct bw_upcall *req)
1732 {
1733     uint32_t flags = 0;
1734 
1735     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1736 	flags |= BW_METER_UNIT_PACKETS;
1737     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1738 	flags |= BW_METER_UNIT_BYTES;
1739     if (req->bu_flags & BW_UPCALL_GEQ)
1740 	flags |= BW_METER_GEQ;
1741     if (req->bu_flags & BW_UPCALL_LEQ)
1742 	flags |= BW_METER_LEQ;
1743 
1744     return flags;
1745 }
1746 
1747 /*
1748  * Add a bw_meter entry
1749  */
1750 static int
1751 add_bw_upcall(struct bw_upcall *req)
1752 {
1753     struct mfc *mfc;
1754     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1755 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1756     struct timeval now;
1757     struct bw_meter *x;
1758     uint32_t flags;
1759 
1760     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
1761 	return EOPNOTSUPP;
1762 
1763     /* Test if the flags are valid */
1764     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1765 	return EINVAL;
1766     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1767 	return EINVAL;
1768     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1769 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1770 	return EINVAL;
1771 
1772     /* Test if the threshold time interval is valid */
1773     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1774 	return EINVAL;
1775 
1776     flags = compute_bw_meter_flags(req);
1777 
1778     /*
1779      * Find if we have already same bw_meter entry
1780      */
1781     MFC_LOCK();
1782     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1783     if (mfc == NULL) {
1784 	MFC_UNLOCK();
1785 	return EADDRNOTAVAIL;
1786     }
1787     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1788 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1789 			   &req->bu_threshold.b_time, ==)) &&
1790 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1791 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1792 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1793 	    MFC_UNLOCK();
1794 	    return 0;		/* XXX Already installed */
1795 	}
1796     }
1797 
1798     /* Allocate the new bw_meter entry */
1799     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1800     if (x == NULL) {
1801 	MFC_UNLOCK();
1802 	return ENOBUFS;
1803     }
1804 
1805     /* Set the new bw_meter entry */
1806     x->bm_threshold.b_time = req->bu_threshold.b_time;
1807     microtime(&now);
1808     x->bm_start_time = now;
1809     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1810     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1811     x->bm_measured.b_packets = 0;
1812     x->bm_measured.b_bytes = 0;
1813     x->bm_flags = flags;
1814     x->bm_time_next = NULL;
1815     x->bm_time_hash = BW_METER_BUCKETS;
1816 
1817     /* Add the new bw_meter entry to the front of entries for this MFC */
1818     x->bm_mfc = mfc;
1819     x->bm_mfc_next = mfc->mfc_bw_meter;
1820     mfc->mfc_bw_meter = x;
1821     schedule_bw_meter(x, &now);
1822     MFC_UNLOCK();
1823 
1824     return 0;
1825 }
1826 
1827 static void
1828 free_bw_list(struct bw_meter *list)
1829 {
1830     while (list != NULL) {
1831 	struct bw_meter *x = list;
1832 
1833 	list = list->bm_mfc_next;
1834 	unschedule_bw_meter(x);
1835 	free(x, M_BWMETER);
1836     }
1837 }
1838 
1839 /*
1840  * Delete one or multiple bw_meter entries
1841  */
1842 static int
1843 del_bw_upcall(struct bw_upcall *req)
1844 {
1845     struct mfc *mfc;
1846     struct bw_meter *x;
1847 
1848     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
1849 	return EOPNOTSUPP;
1850 
1851     MFC_LOCK();
1852 
1853     /* Find the corresponding MFC entry */
1854     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1855     if (mfc == NULL) {
1856 	MFC_UNLOCK();
1857 	return EADDRNOTAVAIL;
1858     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1859 	/*
1860 	 * Delete all bw_meter entries for this mfc
1861 	 */
1862 	struct bw_meter *list;
1863 
1864 	list = mfc->mfc_bw_meter;
1865 	mfc->mfc_bw_meter = NULL;
1866 	free_bw_list(list);
1867 	MFC_UNLOCK();
1868 	return 0;
1869     } else {			/* Delete a single bw_meter entry */
1870 	struct bw_meter *prev;
1871 	uint32_t flags = 0;
1872 
1873 	flags = compute_bw_meter_flags(req);
1874 
1875 	/* Find the bw_meter entry to delete */
1876 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1877 	     prev = x, x = x->bm_mfc_next) {
1878 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1879 			       &req->bu_threshold.b_time, ==)) &&
1880 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1881 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1882 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1883 		break;
1884 	}
1885 	if (x != NULL) { /* Delete entry from the list for this MFC */
1886 	    if (prev != NULL)
1887 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1888 	    else
1889 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1890 
1891 	    unschedule_bw_meter(x);
1892 	    MFC_UNLOCK();
1893 	    /* Free the bw_meter entry */
1894 	    free(x, M_BWMETER);
1895 	    return 0;
1896 	} else {
1897 	    MFC_UNLOCK();
1898 	    return EINVAL;
1899 	}
1900     }
1901     /* NOTREACHED */
1902 }
1903 
1904 /*
1905  * Perform bandwidth measurement processing that may result in an upcall
1906  */
1907 static void
1908 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1909 {
1910     struct timeval delta;
1911 
1912     MFC_LOCK_ASSERT();
1913 
1914     delta = *nowp;
1915     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1916 
1917     if (x->bm_flags & BW_METER_GEQ) {
1918 	/*
1919 	 * Processing for ">=" type of bw_meter entry
1920 	 */
1921 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1922 	    /* Reset the bw_meter entry */
1923 	    x->bm_start_time = *nowp;
1924 	    x->bm_measured.b_packets = 0;
1925 	    x->bm_measured.b_bytes = 0;
1926 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1927 	}
1928 
1929 	/* Record that a packet is received */
1930 	x->bm_measured.b_packets++;
1931 	x->bm_measured.b_bytes += plen;
1932 
1933 	/*
1934 	 * Test if we should deliver an upcall
1935 	 */
1936 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1937 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1938 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1939 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1940 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1941 		/* Prepare an upcall for delivery */
1942 		bw_meter_prepare_upcall(x, nowp);
1943 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1944 	    }
1945 	}
1946     } else if (x->bm_flags & BW_METER_LEQ) {
1947 	/*
1948 	 * Processing for "<=" type of bw_meter entry
1949 	 */
1950 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1951 	    /*
1952 	     * We are behind time with the multicast forwarding table
1953 	     * scanning for "<=" type of bw_meter entries, so test now
1954 	     * if we should deliver an upcall.
1955 	     */
1956 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1957 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1958 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1959 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1960 		/* Prepare an upcall for delivery */
1961 		bw_meter_prepare_upcall(x, nowp);
1962 	    }
1963 	    /* Reschedule the bw_meter entry */
1964 	    unschedule_bw_meter(x);
1965 	    schedule_bw_meter(x, nowp);
1966 	}
1967 
1968 	/* Record that a packet is received */
1969 	x->bm_measured.b_packets++;
1970 	x->bm_measured.b_bytes += plen;
1971 
1972 	/*
1973 	 * Test if we should restart the measuring interval
1974 	 */
1975 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1976 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1977 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1978 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1979 	    /* Don't restart the measuring interval */
1980 	} else {
1981 	    /* Do restart the measuring interval */
1982 	    /*
1983 	     * XXX: note that we don't unschedule and schedule, because this
1984 	     * might be too much overhead per packet. Instead, when we process
1985 	     * all entries for a given timer hash bin, we check whether it is
1986 	     * really a timeout. If not, we reschedule at that time.
1987 	     */
1988 	    x->bm_start_time = *nowp;
1989 	    x->bm_measured.b_packets = 0;
1990 	    x->bm_measured.b_bytes = 0;
1991 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1992 	}
1993     }
1994 }
1995 
1996 /*
1997  * Prepare a bandwidth-related upcall
1998  */
1999 static void
2000 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2001 {
2002     struct timeval delta;
2003     struct bw_upcall *u;
2004 
2005     MFC_LOCK_ASSERT();
2006 
2007     /*
2008      * Compute the measured time interval
2009      */
2010     delta = *nowp;
2011     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2012 
2013     /*
2014      * If there are too many pending upcalls, deliver them now
2015      */
2016     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2017 	bw_upcalls_send();
2018 
2019     /*
2020      * Set the bw_upcall entry
2021      */
2022     u = &bw_upcalls[bw_upcalls_n++];
2023     u->bu_src = x->bm_mfc->mfc_origin;
2024     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2025     u->bu_threshold.b_time = x->bm_threshold.b_time;
2026     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2027     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2028     u->bu_measured.b_time = delta;
2029     u->bu_measured.b_packets = x->bm_measured.b_packets;
2030     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2031     u->bu_flags = 0;
2032     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2033 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2034     if (x->bm_flags & BW_METER_UNIT_BYTES)
2035 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2036     if (x->bm_flags & BW_METER_GEQ)
2037 	u->bu_flags |= BW_UPCALL_GEQ;
2038     if (x->bm_flags & BW_METER_LEQ)
2039 	u->bu_flags |= BW_UPCALL_LEQ;
2040 }
2041 
2042 /*
2043  * Send the pending bandwidth-related upcalls
2044  */
2045 static void
2046 bw_upcalls_send(void)
2047 {
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     struct mbuf *mb_first;
2405     int len = ntohs(ip->ip_len);
2406     struct igmpmsg *im;
2407     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2408 
2409     VIF_LOCK_ASSERT();
2410 
2411     /*
2412      * Add a new mbuf with an upcall header
2413      */
2414     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2415     if (mb_first == NULL) {
2416 	m_freem(mb_copy);
2417 	return ENOBUFS;
2418     }
2419     mb_first->m_data += max_linkhdr;
2420     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2421     mb_first->m_len = sizeof(struct igmpmsg);
2422     mb_first->m_next = mb_copy;
2423 
2424     /* Send message to routing daemon */
2425     im = mtod(mb_first, struct igmpmsg *);
2426     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2427     im->im_mbz		= 0;
2428     im->im_vif		= vifp - viftable;
2429     im->im_src		= ip->ip_src;
2430     im->im_dst		= ip->ip_dst;
2431 
2432     k_igmpsrc.sin_addr	= ip->ip_src;
2433 
2434     MRTSTAT_INC(mrts_upcalls);
2435 
2436     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2437 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2438 	MRTSTAT_INC(mrts_upq_sockfull);
2439 	return ENOBUFS;
2440     }
2441 
2442     /* Keep statistics */
2443     PIMSTAT_INC(pims_snd_registers_msgs);
2444     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2445 
2446     return 0;
2447 }
2448 
2449 /*
2450  * Encapsulate the data packet in PIM Register message and send it to the RP.
2451  */
2452 static int
2453 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2454     struct mfc *rt)
2455 {
2456     struct mbuf *mb_first;
2457     struct ip *ip_outer;
2458     struct pim_encap_pimhdr *pimhdr;
2459     int len = ntohs(ip->ip_len);
2460     vifi_t vifi = rt->mfc_parent;
2461 
2462     VIF_LOCK_ASSERT();
2463 
2464     if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
2465 	m_freem(mb_copy);
2466 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2467     }
2468 
2469     /*
2470      * Add a new mbuf with the encapsulating header
2471      */
2472     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2473     if (mb_first == NULL) {
2474 	m_freem(mb_copy);
2475 	return ENOBUFS;
2476     }
2477     mb_first->m_data += max_linkhdr;
2478     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2479     mb_first->m_next = mb_copy;
2480 
2481     mb_first->m_pkthdr.len = len + mb_first->m_len;
2482 
2483     /*
2484      * Fill in the encapsulating IP and PIM header
2485      */
2486     ip_outer = mtod(mb_first, struct ip *);
2487     *ip_outer = pim_encap_iphdr;
2488     ip_outer->ip_id = ip_newid();
2489     ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2490     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
2491     ip_outer->ip_dst = rt->mfc_rp;
2492     /*
2493      * Copy the inner header TOS to the outer header, and take care of the
2494      * IP_DF bit.
2495      */
2496     ip_outer->ip_tos = ip->ip_tos;
2497     if (ntohs(ip->ip_off) & IP_DF)
2498 	ip_outer->ip_off |= IP_DF;
2499     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2500 					 + sizeof(pim_encap_iphdr));
2501     *pimhdr = pim_encap_pimhdr;
2502     /* If the iif crosses a border, set the Border-bit */
2503     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
2504 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2505 
2506     mb_first->m_data += sizeof(pim_encap_iphdr);
2507     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2508     mb_first->m_data -= sizeof(pim_encap_iphdr);
2509 
2510     send_packet(vifp, mb_first);
2511 
2512     /* Keep statistics */
2513     PIMSTAT_INC(pims_snd_registers_msgs);
2514     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2515 
2516     return 0;
2517 }
2518 
2519 /*
2520  * pim_encapcheck() is called by the encap4_input() path at runtime to
2521  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2522  * into the kernel.
2523  */
2524 static int
2525 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2526 {
2527 
2528 #ifdef DIAGNOSTIC
2529     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2530 #endif
2531     if (proto != IPPROTO_PIM)
2532 	return 0;	/* not for us; reject the datagram. */
2533 
2534     return 64;		/* claim the datagram. */
2535 }
2536 
2537 /*
2538  * PIM-SMv2 and PIM-DM messages processing.
2539  * Receives and verifies the PIM control messages, and passes them
2540  * up to the listening socket, using rip_input().
2541  * The only message with special processing is the PIM_REGISTER message
2542  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2543  * is passed to if_simloop().
2544  */
2545 void
2546 pim_input(struct mbuf *m, int off)
2547 {
2548     struct ip *ip = mtod(m, struct ip *);
2549     struct pim *pim;
2550     int minlen;
2551     int datalen = ip->ip_len;
2552     int ip_tos;
2553     int iphlen = off;
2554 
2555     /* Keep statistics */
2556     PIMSTAT_INC(pims_rcv_total_msgs);
2557     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2558 
2559     /*
2560      * Validate lengths
2561      */
2562     if (datalen < PIM_MINLEN) {
2563 	PIMSTAT_INC(pims_rcv_tooshort);
2564 	CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2565 	    __func__, datalen, inet_ntoa(ip->ip_src));
2566 	m_freem(m);
2567 	return;
2568     }
2569 
2570     /*
2571      * If the packet is at least as big as a REGISTER, go agead
2572      * and grab the PIM REGISTER header size, to avoid another
2573      * possible m_pullup() later.
2574      *
2575      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2576      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2577      */
2578     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2579     /*
2580      * Get the IP and PIM headers in contiguous memory, and
2581      * possibly the PIM REGISTER header.
2582      */
2583     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
2584 	(m = m_pullup(m, minlen)) == 0) {
2585 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2586 	return;
2587     }
2588 
2589     /* m_pullup() may have given us a new mbuf so reset ip. */
2590     ip = mtod(m, struct ip *);
2591     ip_tos = ip->ip_tos;
2592 
2593     /* adjust mbuf to point to the PIM header */
2594     m->m_data += iphlen;
2595     m->m_len  -= iphlen;
2596     pim = mtod(m, struct pim *);
2597 
2598     /*
2599      * Validate checksum. If PIM REGISTER, exclude the data packet.
2600      *
2601      * XXX: some older PIMv2 implementations don't make this distinction,
2602      * so for compatibility reason perform the checksum over part of the
2603      * message, and if error, then over the whole message.
2604      */
2605     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2606 	/* do nothing, checksum okay */
2607     } else if (in_cksum(m, datalen)) {
2608 	PIMSTAT_INC(pims_rcv_badsum);
2609 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2610 	m_freem(m);
2611 	return;
2612     }
2613 
2614     /* PIM version check */
2615     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2616 	PIMSTAT_INC(pims_rcv_badversion);
2617 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2618 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2619 	m_freem(m);
2620 	return;
2621     }
2622 
2623     /* restore mbuf back to the outer IP */
2624     m->m_data -= iphlen;
2625     m->m_len  += iphlen;
2626 
2627     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2628 	/*
2629 	 * Since this is a REGISTER, we'll make a copy of the register
2630 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2631 	 * routing daemon.
2632 	 */
2633 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2634 	struct mbuf *mcp;
2635 	struct ip *encap_ip;
2636 	u_int32_t *reghdr;
2637 	struct ifnet *vifp;
2638 
2639 	VIF_LOCK();
2640 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
2641 	    VIF_UNLOCK();
2642 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2643 		(int)reg_vif_num);
2644 	    m_freem(m);
2645 	    return;
2646 	}
2647 	/* XXX need refcnt? */
2648 	vifp = viftable[reg_vif_num].v_ifp;
2649 	VIF_UNLOCK();
2650 
2651 	/*
2652 	 * Validate length
2653 	 */
2654 	if (datalen < PIM_REG_MINLEN) {
2655 	    PIMSTAT_INC(pims_rcv_tooshort);
2656 	    PIMSTAT_INC(pims_rcv_badregisters);
2657 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2658 	    m_freem(m);
2659 	    return;
2660 	}
2661 
2662 	reghdr = (u_int32_t *)(pim + 1);
2663 	encap_ip = (struct ip *)(reghdr + 1);
2664 
2665 	CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2666 	    __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2667 
2668 	/* verify the version number of the inner packet */
2669 	if (encap_ip->ip_v != IPVERSION) {
2670 	    PIMSTAT_INC(pims_rcv_badregisters);
2671 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2672 	    m_freem(m);
2673 	    return;
2674 	}
2675 
2676 	/* verify the inner packet is destined to a mcast group */
2677 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2678 	    PIMSTAT_INC(pims_rcv_badregisters);
2679 	    CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2680 		inet_ntoa(encap_ip->ip_dst));
2681 	    m_freem(m);
2682 	    return;
2683 	}
2684 
2685 	/* If a NULL_REGISTER, pass it to the daemon */
2686 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2687 	    goto pim_input_to_daemon;
2688 
2689 	/*
2690 	 * Copy the TOS from the outer IP header to the inner IP header.
2691 	 */
2692 	if (encap_ip->ip_tos != ip_tos) {
2693 	    /* Outer TOS -> inner TOS */
2694 	    encap_ip->ip_tos = ip_tos;
2695 	    /* Recompute the inner header checksum. Sigh... */
2696 
2697 	    /* adjust mbuf to point to the inner IP header */
2698 	    m->m_data += (iphlen + PIM_MINLEN);
2699 	    m->m_len  -= (iphlen + PIM_MINLEN);
2700 
2701 	    encap_ip->ip_sum = 0;
2702 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2703 
2704 	    /* restore mbuf to point back to the outer IP header */
2705 	    m->m_data -= (iphlen + PIM_MINLEN);
2706 	    m->m_len  += (iphlen + PIM_MINLEN);
2707 	}
2708 
2709 	/*
2710 	 * Decapsulate the inner IP packet and loopback to forward it
2711 	 * as a normal multicast packet. Also, make a copy of the
2712 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2713 	 * to pass to the daemon later, so it can take the appropriate
2714 	 * actions (e.g., send back PIM_REGISTER_STOP).
2715 	 * XXX: here m->m_data points to the outer IP header.
2716 	 */
2717 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2718 	if (mcp == NULL) {
2719 	    CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2720 	    m_freem(m);
2721 	    return;
2722 	}
2723 
2724 	/* Keep statistics */
2725 	/* XXX: registers_bytes include only the encap. mcast pkt */
2726 	PIMSTAT_INC(pims_rcv_registers_msgs);
2727 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2728 
2729 	/*
2730 	 * forward the inner ip packet; point m_data at the inner ip.
2731 	 */
2732 	m_adj(m, iphlen + PIM_MINLEN);
2733 
2734 	CTR4(KTR_IPMF,
2735 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2736 	    __func__,
2737 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2738 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2739 	    (int)reg_vif_num);
2740 
2741 	/* NB: vifp was collected above; can it change on us? */
2742 	if_simloop(vifp, m, dst.sin_family, 0);
2743 
2744 	/* prepare the register head to send to the mrouting daemon */
2745 	m = mcp;
2746     }
2747 
2748 pim_input_to_daemon:
2749     /*
2750      * Pass the PIM message up to the daemon; if it is a Register message,
2751      * pass the 'head' only up to the daemon. This includes the
2752      * outer IP header, PIM header, PIM-Register header and the
2753      * inner IP header.
2754      * XXX: the outer IP header pkt size of a Register is not adjust to
2755      * reflect the fact that the inner multicast data is truncated.
2756      */
2757     rip_input(m, iphlen);
2758 
2759     return;
2760 }
2761 
2762 static int
2763 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2764 {
2765 	struct mfc	*rt;
2766 	int		 error, i;
2767 
2768 	if (req->newptr)
2769 		return (EPERM);
2770 	if (mfchashtbl == NULL)	/* XXX unlocked */
2771 		return (0);
2772 	error = sysctl_wire_old_buffer(req, 0);
2773 	if (error)
2774 		return (error);
2775 
2776 	MFC_LOCK();
2777 	for (i = 0; i < mfchashsize; i++) {
2778 		LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
2779 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2780 			if (error)
2781 				goto out_locked;
2782 		}
2783 	}
2784 out_locked:
2785 	MFC_UNLOCK();
2786 	return (error);
2787 }
2788 
2789 SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, sysctl_mfctable,
2790     "IPv4 Multicast Forwarding Table (struct *mfc[mfchashsize], "
2791     "netinet/ip_mroute.h)");
2792 
2793 static int
2794 ip_mroute_modevent(module_t mod, int type, void *unused)
2795 {
2796 
2797     switch (type) {
2798     case MOD_LOAD:
2799 	MROUTER_LOCK_INIT();
2800 	MFC_LOCK_INIT();
2801 	VIF_LOCK_INIT();
2802 
2803 	mfchashsize = MFCHASHSIZE;
2804 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2805 	    !powerof2(mfchashsize)) {
2806 		printf("WARNING: %s not a power of 2; using default\n",
2807 		    "net.inet.ip.mfchashsize");
2808 		mfchashsize = MFCHASHSIZE;
2809 	}
2810 	MALLOC(nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2811 
2812 	pim_squelch_wholepkt = 0;
2813 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2814 	    &pim_squelch_wholepkt);
2815 	ip_mrouter_reset();
2816 
2817 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2818 	    pim_encapcheck, &in_pim_protosw, NULL);
2819 	if (pim_encap_cookie == NULL) {
2820 		printf("ip_mroute: unable to attach pim encap\n");
2821 		VIF_LOCK_DESTROY();
2822 		MFC_LOCK_DESTROY();
2823 		MROUTER_LOCK_DESTROY();
2824 		return (EINVAL);
2825 	}
2826 
2827 	ip_mcast_src = X_ip_mcast_src;
2828 	ip_mforward = X_ip_mforward;
2829 	ip_mrouter_done = X_ip_mrouter_done;
2830 	ip_mrouter_get = X_ip_mrouter_get;
2831 	ip_mrouter_set = X_ip_mrouter_set;
2832 
2833 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2834 	ip_rsvp_vif = X_ip_rsvp_vif;
2835 
2836 	legal_vif_num = X_legal_vif_num;
2837 	mrt_ioctl = X_mrt_ioctl;
2838 	rsvp_input_p = X_rsvp_input;
2839 	break;
2840 
2841     case MOD_UNLOAD:
2842 	/*
2843 	 * Typically module unload happens after the user-level
2844 	 * process has shutdown the kernel services (the check
2845 	 * below insures someone can't just yank the module out
2846 	 * from under a running process).  But if the module is
2847 	 * just loaded and then unloaded w/o starting up a user
2848 	 * process we still need to cleanup.
2849 	 */
2850 	if (V_ip_mrouter != NULL)
2851 	    return (EINVAL);
2852 
2853 	if (pim_encap_cookie) {
2854 	    encap_detach(pim_encap_cookie);
2855 	    pim_encap_cookie = NULL;
2856 	}
2857 	X_ip_mrouter_done();
2858 
2859 	FREE(nexpire, M_MRTABLE);
2860 	nexpire = NULL;
2861 
2862 	ip_mcast_src = NULL;
2863 	ip_mforward = NULL;
2864 	ip_mrouter_done = NULL;
2865 	ip_mrouter_get = NULL;
2866 	ip_mrouter_set = NULL;
2867 
2868 	ip_rsvp_force_done = NULL;
2869 	ip_rsvp_vif = NULL;
2870 
2871 	legal_vif_num = NULL;
2872 	mrt_ioctl = NULL;
2873 	rsvp_input_p = NULL;
2874 
2875 	VIF_LOCK_DESTROY();
2876 	MFC_LOCK_DESTROY();
2877 	MROUTER_LOCK_DESTROY();
2878 	break;
2879 
2880     default:
2881 	return EOPNOTSUPP;
2882     }
2883     return 0;
2884 }
2885 
2886 static moduledata_t ip_mroutemod = {
2887     "ip_mroute",
2888     ip_mroute_modevent,
2889     0
2890 };
2891 
2892 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
2893