xref: /freebsd/sys/netinet/ip_mroute.c (revision a98ff317388a00b992f1bf8404dee596f9383f5e)
1 /*-
2  * Copyright (c) 1989 Stephen Deering
3  * Copyright (c) 1992, 1993
4  *      The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Stephen Deering of Stanford University.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
34  */
35 
36 /*
37  * IP multicast forwarding procedures
38  *
39  * Written by David Waitzman, BBN Labs, August 1988.
40  * Modified by Steve Deering, Stanford, February 1989.
41  * Modified by Mark J. Steiglitz, Stanford, May, 1991
42  * Modified by Van Jacobson, LBL, January 1993
43  * Modified by Ajit Thyagarajan, PARC, August 1993
44  * Modified by Bill Fenner, PARC, April 1995
45  * Modified by Ahmed Helmy, SGI, June 1996
46  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
47  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
48  * Modified by Hitoshi Asaeda, WIDE, August 2000
49  * Modified by Pavlin Radoslavov, ICSI, October 2002
50  *
51  * MROUTING Revision: 3.5
52  * and PIM-SMv2 and PIM-DM support, advanced API support,
53  * bandwidth metering and signaling
54  */
55 
56 /*
57  * TODO: Prefix functions with ipmf_.
58  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
59  * domain attachment (if_afdata) so we can track consumers of that service.
60  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
61  * move it to socket options.
62  * TODO: Cleanup LSRR removal further.
63  * TODO: Push RSVP stubs into raw_ip.c.
64  * TODO: Use bitstring.h for vif set.
65  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
66  * TODO: Sync ip6_mroute.c with this file.
67  */
68 
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
71 
72 #include "opt_inet.h"
73 #include "opt_mrouting.h"
74 
75 #define _PIM_VT 1
76 
77 #include <sys/param.h>
78 #include <sys/kernel.h>
79 #include <sys/stddef.h>
80 #include <sys/lock.h>
81 #include <sys/ktr.h>
82 #include <sys/malloc.h>
83 #include <sys/mbuf.h>
84 #include <sys/module.h>
85 #include <sys/priv.h>
86 #include <sys/protosw.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/socketvar.h>
90 #include <sys/sockio.h>
91 #include <sys/sx.h>
92 #include <sys/sysctl.h>
93 #include <sys/syslog.h>
94 #include <sys/systm.h>
95 #include <sys/time.h>
96 #include <sys/counter.h>
97 
98 #include <net/if.h>
99 #include <net/netisr.h>
100 #include <net/route.h>
101 #include <net/vnet.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 
116 #include <machine/in_cksum.h>
117 
118 #ifndef KTR_IPMF
119 #define KTR_IPMF KTR_INET
120 #endif
121 
122 #define		VIFI_INVALID	((vifi_t) -1)
123 #define		M_HASCL(m)	((m)->m_flags & M_EXT)
124 
125 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
126 #define	V_last_tv_sec	VNET(last_tv_sec)
127 
128 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
129 
130 /*
131  * Locking.  We use two locks: one for the virtual interface table and
132  * one for the forwarding table.  These locks may be nested in which case
133  * the VIF lock must always be taken first.  Note that each lock is used
134  * to cover not only the specific data structure but also related data
135  * structures.
136  */
137 
138 static struct mtx mrouter_mtx;
139 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
140 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
141 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
142 #define	MROUTER_LOCK_INIT()						\
143 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
144 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
145 
146 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
147 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
148 
149 static VNET_PCPUSTAT_DEFINE(struct mrtstat, mrtstat);
150 VNET_PCPUSTAT_SYSINIT(mrtstat);
151 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
152 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
153     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
154     "netinet/ip_mroute.h)");
155 
156 static VNET_DEFINE(u_long, mfchash);
157 #define	V_mfchash		VNET(mfchash)
158 #define	MFCHASH(a, g)							\
159 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
160 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
161 #define	MFCHASHSIZE	256
162 
163 static u_long mfchashsize;			/* Hash size */
164 static VNET_DEFINE(u_char *, nexpire);		/* 0..mfchashsize-1 */
165 #define	V_nexpire		VNET(nexpire)
166 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
167 #define	V_mfchashtbl		VNET(mfchashtbl)
168 
169 static struct mtx mfc_mtx;
170 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
171 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
172 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
173 #define	MFC_LOCK_INIT()							\
174 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
175 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
176 
177 static VNET_DEFINE(vifi_t, numvifs);
178 #define	V_numvifs		VNET(numvifs)
179 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
180 #define	V_viftable		VNET(viftable)
181 SYSCTL_VNET_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
182     &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
183     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
184 
185 static struct mtx vif_mtx;
186 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
187 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
188 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
189 #define	VIF_LOCK_INIT()							\
190 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
191 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
192 
193 static eventhandler_tag if_detach_event_tag = NULL;
194 
195 static VNET_DEFINE(struct callout, expire_upcalls_ch);
196 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
197 
198 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
199 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
200 
201 /*
202  * Bandwidth meter variables and constants
203  */
204 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
205 /*
206  * Pending timeouts are stored in a hash table, the key being the
207  * expiration time. Periodically, the entries are analysed and processed.
208  */
209 #define	BW_METER_BUCKETS	1024
210 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
211 #define	V_bw_meter_timers	VNET(bw_meter_timers)
212 static VNET_DEFINE(struct callout, bw_meter_ch);
213 #define	V_bw_meter_ch		VNET(bw_meter_ch)
214 #define	BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
215 
216 /*
217  * Pending upcalls are stored in a vector which is flushed when
218  * full, or periodically
219  */
220 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
221 #define	V_bw_upcalls		VNET(bw_upcalls)
222 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
223 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
224 static VNET_DEFINE(struct callout, bw_upcalls_ch);
225 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
226 
227 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
228 
229 static VNET_PCPUSTAT_DEFINE(struct pimstat, pimstat);
230 VNET_PCPUSTAT_SYSINIT(pimstat);
231 VNET_PCPUSTAT_SYSUNINIT(pimstat);
232 
233 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
234 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
235     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
236 
237 static u_long	pim_squelch_wholepkt = 0;
238 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
239     &pim_squelch_wholepkt, 0,
240     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
241 
242 extern  struct domain inetdomain;
243 static const struct protosw in_pim_protosw = {
244 	.pr_type =		SOCK_RAW,
245 	.pr_domain =		&inetdomain,
246 	.pr_protocol =		IPPROTO_PIM,
247 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
248 	.pr_input =		pim_input,
249 	.pr_output =		(pr_output_t*)rip_output,
250 	.pr_ctloutput =		rip_ctloutput,
251 	.pr_usrreqs =		&rip_usrreqs
252 };
253 static const struct encaptab *pim_encap_cookie;
254 
255 static int pim_encapcheck(const struct mbuf *, int, int, void *);
256 
257 /*
258  * Note: the PIM Register encapsulation adds the following in front of a
259  * data packet:
260  *
261  * struct pim_encap_hdr {
262  *    struct ip ip;
263  *    struct pim_encap_pimhdr  pim;
264  * }
265  *
266  */
267 
268 struct pim_encap_pimhdr {
269 	struct pim pim;
270 	uint32_t   flags;
271 };
272 #define		PIM_ENCAP_TTL	64
273 
274 static struct ip pim_encap_iphdr = {
275 #if BYTE_ORDER == LITTLE_ENDIAN
276 	sizeof(struct ip) >> 2,
277 	IPVERSION,
278 #else
279 	IPVERSION,
280 	sizeof(struct ip) >> 2,
281 #endif
282 	0,			/* tos */
283 	sizeof(struct ip),	/* total length */
284 	0,			/* id */
285 	0,			/* frag offset */
286 	PIM_ENCAP_TTL,
287 	IPPROTO_PIM,
288 	0,			/* checksum */
289 };
290 
291 static struct pim_encap_pimhdr pim_encap_pimhdr = {
292     {
293 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
294 	0,			/* reserved */
295 	0,			/* checksum */
296     },
297     0				/* flags */
298 };
299 
300 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
301 #define	V_reg_vif_num		VNET(reg_vif_num)
302 static VNET_DEFINE(struct ifnet, multicast_register_if);
303 #define	V_multicast_register_if	VNET(multicast_register_if)
304 
305 /*
306  * Private variables.
307  */
308 
309 static u_long	X_ip_mcast_src(int);
310 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
311 		    struct ip_moptions *);
312 static int	X_ip_mrouter_done(void);
313 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
314 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
315 static int	X_legal_vif_num(int);
316 static int	X_mrt_ioctl(u_long, caddr_t, int);
317 
318 static int	add_bw_upcall(struct bw_upcall *);
319 static int	add_mfc(struct mfcctl2 *);
320 static int	add_vif(struct vifctl *);
321 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
322 static void	bw_meter_process(void);
323 static void	bw_meter_receive_packet(struct bw_meter *, int,
324 		    struct timeval *);
325 static void	bw_upcalls_send(void);
326 static int	del_bw_upcall(struct bw_upcall *);
327 static int	del_mfc(struct mfcctl2 *);
328 static int	del_vif(vifi_t);
329 static int	del_vif_locked(vifi_t);
330 static void	expire_bw_meter_process(void *);
331 static void	expire_bw_upcalls_send(void *);
332 static void	expire_mfc(struct mfc *);
333 static void	expire_upcalls(void *);
334 static void	free_bw_list(struct bw_meter *);
335 static int	get_sg_cnt(struct sioc_sg_req *);
336 static int	get_vif_cnt(struct sioc_vif_req *);
337 static void	if_detached_event(void *, struct ifnet *);
338 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
339 static int	ip_mrouter_init(struct socket *, int);
340 static __inline struct mfc *
341 		mfc_find(struct in_addr *, struct in_addr *);
342 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
343 static struct mbuf *
344 		pim_register_prepare(struct ip *, struct mbuf *);
345 static int	pim_register_send(struct ip *, struct vif *,
346 		    struct mbuf *, struct mfc *);
347 static int	pim_register_send_rp(struct ip *, struct vif *,
348 		    struct mbuf *, struct mfc *);
349 static int	pim_register_send_upcall(struct ip *, struct vif *,
350 		    struct mbuf *, struct mfc *);
351 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
352 static void	send_packet(struct vif *, struct mbuf *);
353 static int	set_api_config(uint32_t *);
354 static int	set_assert(int);
355 static int	socket_send(struct socket *, struct mbuf *,
356 		    struct sockaddr_in *);
357 static void	unschedule_bw_meter(struct bw_meter *);
358 
359 /*
360  * Kernel multicast forwarding API capabilities and setup.
361  * If more API capabilities are added to the kernel, they should be
362  * recorded in `mrt_api_support'.
363  */
364 #define MRT_API_VERSION		0x0305
365 
366 static const int mrt_api_version = MRT_API_VERSION;
367 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
368 					 MRT_MFC_FLAGS_BORDER_VIF |
369 					 MRT_MFC_RP |
370 					 MRT_MFC_BW_UPCALL);
371 static VNET_DEFINE(uint32_t, mrt_api_config);
372 #define	V_mrt_api_config	VNET(mrt_api_config)
373 static VNET_DEFINE(int, pim_assert_enabled);
374 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
375 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
376 
377 /*
378  * Find a route for a given origin IP address and multicast group address.
379  * Statistics must be updated by the caller.
380  */
381 static __inline struct mfc *
382 mfc_find(struct in_addr *o, struct in_addr *g)
383 {
384 	struct mfc *rt;
385 
386 	MFC_LOCK_ASSERT();
387 
388 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
389 		if (in_hosteq(rt->mfc_origin, *o) &&
390 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
391 		    TAILQ_EMPTY(&rt->mfc_stall))
392 			break;
393 	}
394 
395 	return (rt);
396 }
397 
398 /*
399  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
400  */
401 static int
402 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
403 {
404     int	error, optval;
405     vifi_t	vifi;
406     struct	vifctl vifc;
407     struct	mfcctl2 mfc;
408     struct	bw_upcall bw_upcall;
409     uint32_t	i;
410 
411     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
412 	return EPERM;
413 
414     error = 0;
415     switch (sopt->sopt_name) {
416     case MRT_INIT:
417 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
418 	if (error)
419 	    break;
420 	error = ip_mrouter_init(so, optval);
421 	break;
422 
423     case MRT_DONE:
424 	error = ip_mrouter_done();
425 	break;
426 
427     case MRT_ADD_VIF:
428 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
429 	if (error)
430 	    break;
431 	error = add_vif(&vifc);
432 	break;
433 
434     case MRT_DEL_VIF:
435 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
436 	if (error)
437 	    break;
438 	error = del_vif(vifi);
439 	break;
440 
441     case MRT_ADD_MFC:
442     case MRT_DEL_MFC:
443 	/*
444 	 * select data size depending on API version.
445 	 */
446 	if (sopt->sopt_name == MRT_ADD_MFC &&
447 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
448 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
449 				sizeof(struct mfcctl2));
450 	} else {
451 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
452 				sizeof(struct mfcctl));
453 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
454 			sizeof(mfc) - sizeof(struct mfcctl));
455 	}
456 	if (error)
457 	    break;
458 	if (sopt->sopt_name == MRT_ADD_MFC)
459 	    error = add_mfc(&mfc);
460 	else
461 	    error = del_mfc(&mfc);
462 	break;
463 
464     case MRT_ASSERT:
465 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
466 	if (error)
467 	    break;
468 	set_assert(optval);
469 	break;
470 
471     case MRT_API_CONFIG:
472 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
473 	if (!error)
474 	    error = set_api_config(&i);
475 	if (!error)
476 	    error = sooptcopyout(sopt, &i, sizeof i);
477 	break;
478 
479     case MRT_ADD_BW_UPCALL:
480     case MRT_DEL_BW_UPCALL:
481 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
482 				sizeof bw_upcall);
483 	if (error)
484 	    break;
485 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
486 	    error = add_bw_upcall(&bw_upcall);
487 	else
488 	    error = del_bw_upcall(&bw_upcall);
489 	break;
490 
491     default:
492 	error = EOPNOTSUPP;
493 	break;
494     }
495     return error;
496 }
497 
498 /*
499  * Handle MRT getsockopt commands
500  */
501 static int
502 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
503 {
504     int error;
505 
506     switch (sopt->sopt_name) {
507     case MRT_VERSION:
508 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
509 	break;
510 
511     case MRT_ASSERT:
512 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
513 	    sizeof V_pim_assert_enabled);
514 	break;
515 
516     case MRT_API_SUPPORT:
517 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
518 	break;
519 
520     case MRT_API_CONFIG:
521 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
522 	break;
523 
524     default:
525 	error = EOPNOTSUPP;
526 	break;
527     }
528     return error;
529 }
530 
531 /*
532  * Handle ioctl commands to obtain information from the cache
533  */
534 static int
535 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
536 {
537     int error = 0;
538 
539     /*
540      * Currently the only function calling this ioctl routine is rtioctl().
541      * Typically, only root can create the raw socket in order to execute
542      * this ioctl method, however the request might be coming from a prison
543      */
544     error = priv_check(curthread, PRIV_NETINET_MROUTE);
545     if (error)
546 	return (error);
547     switch (cmd) {
548     case (SIOCGETVIFCNT):
549 	error = get_vif_cnt((struct sioc_vif_req *)data);
550 	break;
551 
552     case (SIOCGETSGCNT):
553 	error = get_sg_cnt((struct sioc_sg_req *)data);
554 	break;
555 
556     default:
557 	error = EINVAL;
558 	break;
559     }
560     return error;
561 }
562 
563 /*
564  * returns the packet, byte, rpf-failure count for the source group provided
565  */
566 static int
567 get_sg_cnt(struct sioc_sg_req *req)
568 {
569     struct mfc *rt;
570 
571     MFC_LOCK();
572     rt = mfc_find(&req->src, &req->grp);
573     if (rt == NULL) {
574 	MFC_UNLOCK();
575 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
576 	return EADDRNOTAVAIL;
577     }
578     req->pktcnt = rt->mfc_pkt_cnt;
579     req->bytecnt = rt->mfc_byte_cnt;
580     req->wrong_if = rt->mfc_wrong_if;
581     MFC_UNLOCK();
582     return 0;
583 }
584 
585 /*
586  * returns the input and output packet and byte counts on the vif provided
587  */
588 static int
589 get_vif_cnt(struct sioc_vif_req *req)
590 {
591     vifi_t vifi = req->vifi;
592 
593     VIF_LOCK();
594     if (vifi >= V_numvifs) {
595 	VIF_UNLOCK();
596 	return EINVAL;
597     }
598 
599     req->icount = V_viftable[vifi].v_pkt_in;
600     req->ocount = V_viftable[vifi].v_pkt_out;
601     req->ibytes = V_viftable[vifi].v_bytes_in;
602     req->obytes = V_viftable[vifi].v_bytes_out;
603     VIF_UNLOCK();
604 
605     return 0;
606 }
607 
608 static void
609 if_detached_event(void *arg __unused, struct ifnet *ifp)
610 {
611     vifi_t vifi;
612     int i;
613 
614     MROUTER_LOCK();
615 
616     if (V_ip_mrouter == NULL) {
617 	MROUTER_UNLOCK();
618 	return;
619     }
620 
621     VIF_LOCK();
622     MFC_LOCK();
623 
624     /*
625      * Tear down multicast forwarder state associated with this ifnet.
626      * 1. Walk the vif list, matching vifs against this ifnet.
627      * 2. Walk the multicast forwarding cache (mfc) looking for
628      *    inner matches with this vif's index.
629      * 3. Expire any matching multicast forwarding cache entries.
630      * 4. Free vif state. This should disable ALLMULTI on the interface.
631      */
632     for (vifi = 0; vifi < V_numvifs; vifi++) {
633 	if (V_viftable[vifi].v_ifp != ifp)
634 		continue;
635 	for (i = 0; i < mfchashsize; i++) {
636 		struct mfc *rt, *nrt;
637 		for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
638 			nrt = LIST_NEXT(rt, mfc_hash);
639 			if (rt->mfc_parent == vifi) {
640 				expire_mfc(rt);
641 			}
642 		}
643 	}
644 	del_vif_locked(vifi);
645     }
646 
647     MFC_UNLOCK();
648     VIF_UNLOCK();
649 
650     MROUTER_UNLOCK();
651 }
652 
653 /*
654  * Enable multicast forwarding.
655  */
656 static int
657 ip_mrouter_init(struct socket *so, int version)
658 {
659 
660     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
661         so->so_type, so->so_proto->pr_protocol);
662 
663     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
664 	return EOPNOTSUPP;
665 
666     if (version != 1)
667 	return ENOPROTOOPT;
668 
669     MROUTER_LOCK();
670 
671     if (ip_mrouter_unloading) {
672 	MROUTER_UNLOCK();
673 	return ENOPROTOOPT;
674     }
675 
676     if (V_ip_mrouter != NULL) {
677 	MROUTER_UNLOCK();
678 	return EADDRINUSE;
679     }
680 
681     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
682 	HASH_NOWAIT);
683 
684     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
685 	curvnet);
686     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
687 	curvnet);
688     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
689 	curvnet);
690 
691     V_ip_mrouter = so;
692     ip_mrouter_cnt++;
693 
694     MROUTER_UNLOCK();
695 
696     CTR1(KTR_IPMF, "%s: done", __func__);
697 
698     return 0;
699 }
700 
701 /*
702  * Disable multicast forwarding.
703  */
704 static int
705 X_ip_mrouter_done(void)
706 {
707     vifi_t vifi;
708     int i;
709     struct ifnet *ifp;
710     struct ifreq ifr;
711 
712     MROUTER_LOCK();
713 
714     if (V_ip_mrouter == NULL) {
715 	MROUTER_UNLOCK();
716 	return EINVAL;
717     }
718 
719     /*
720      * Detach/disable hooks to the reset of the system.
721      */
722     V_ip_mrouter = NULL;
723     ip_mrouter_cnt--;
724     V_mrt_api_config = 0;
725 
726     VIF_LOCK();
727 
728     /*
729      * For each phyint in use, disable promiscuous reception of all IP
730      * multicasts.
731      */
732     for (vifi = 0; vifi < V_numvifs; vifi++) {
733 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
734 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
735 	    struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
736 
737 	    so->sin_len = sizeof(struct sockaddr_in);
738 	    so->sin_family = AF_INET;
739 	    so->sin_addr.s_addr = INADDR_ANY;
740 	    ifp = V_viftable[vifi].v_ifp;
741 	    if_allmulti(ifp, 0);
742 	}
743     }
744     bzero((caddr_t)V_viftable, sizeof(V_viftable));
745     V_numvifs = 0;
746     V_pim_assert_enabled = 0;
747 
748     VIF_UNLOCK();
749 
750     callout_stop(&V_expire_upcalls_ch);
751     callout_stop(&V_bw_upcalls_ch);
752     callout_stop(&V_bw_meter_ch);
753 
754     MFC_LOCK();
755 
756     /*
757      * Free all multicast forwarding cache entries.
758      * Do not use hashdestroy(), as we must perform other cleanup.
759      */
760     for (i = 0; i < mfchashsize; i++) {
761 	struct mfc *rt, *nrt;
762 	for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
763 		nrt = LIST_NEXT(rt, mfc_hash);
764 		expire_mfc(rt);
765 	}
766     }
767     free(V_mfchashtbl, M_MRTABLE);
768     V_mfchashtbl = NULL;
769 
770     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
771 
772     V_bw_upcalls_n = 0;
773     bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
774 
775     MFC_UNLOCK();
776 
777     V_reg_vif_num = VIFI_INVALID;
778 
779     MROUTER_UNLOCK();
780 
781     CTR1(KTR_IPMF, "%s: done", __func__);
782 
783     return 0;
784 }
785 
786 /*
787  * Set PIM assert processing global
788  */
789 static int
790 set_assert(int i)
791 {
792     if ((i != 1) && (i != 0))
793 	return EINVAL;
794 
795     V_pim_assert_enabled = i;
796 
797     return 0;
798 }
799 
800 /*
801  * Configure API capabilities
802  */
803 int
804 set_api_config(uint32_t *apival)
805 {
806     int i;
807 
808     /*
809      * We can set the API capabilities only if it is the first operation
810      * after MRT_INIT. I.e.:
811      *  - there are no vifs installed
812      *  - pim_assert is not enabled
813      *  - the MFC table is empty
814      */
815     if (V_numvifs > 0) {
816 	*apival = 0;
817 	return EPERM;
818     }
819     if (V_pim_assert_enabled) {
820 	*apival = 0;
821 	return EPERM;
822     }
823 
824     MFC_LOCK();
825 
826     for (i = 0; i < mfchashsize; i++) {
827 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
828 	    MFC_UNLOCK();
829 	    *apival = 0;
830 	    return EPERM;
831 	}
832     }
833 
834     MFC_UNLOCK();
835 
836     V_mrt_api_config = *apival & mrt_api_support;
837     *apival = V_mrt_api_config;
838 
839     return 0;
840 }
841 
842 /*
843  * Add a vif to the vif table
844  */
845 static int
846 add_vif(struct vifctl *vifcp)
847 {
848     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
849     struct sockaddr_in sin = {sizeof sin, AF_INET};
850     struct ifaddr *ifa;
851     struct ifnet *ifp;
852     int error;
853 
854     VIF_LOCK();
855     if (vifcp->vifc_vifi >= MAXVIFS) {
856 	VIF_UNLOCK();
857 	return EINVAL;
858     }
859     /* rate limiting is no longer supported by this code */
860     if (vifcp->vifc_rate_limit != 0) {
861 	log(LOG_ERR, "rate limiting is no longer supported\n");
862 	VIF_UNLOCK();
863 	return EINVAL;
864     }
865     if (!in_nullhost(vifp->v_lcl_addr)) {
866 	VIF_UNLOCK();
867 	return EADDRINUSE;
868     }
869     if (in_nullhost(vifcp->vifc_lcl_addr)) {
870 	VIF_UNLOCK();
871 	return EADDRNOTAVAIL;
872     }
873 
874     /* Find the interface with an address in AF_INET family */
875     if (vifcp->vifc_flags & VIFF_REGISTER) {
876 	/*
877 	 * XXX: Because VIFF_REGISTER does not really need a valid
878 	 * local interface (e.g. it could be 127.0.0.2), we don't
879 	 * check its address.
880 	 */
881 	ifp = NULL;
882     } else {
883 	sin.sin_addr = vifcp->vifc_lcl_addr;
884 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
885 	if (ifa == NULL) {
886 	    VIF_UNLOCK();
887 	    return EADDRNOTAVAIL;
888 	}
889 	ifp = ifa->ifa_ifp;
890 	ifa_free(ifa);
891     }
892 
893     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
894 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
895 	VIF_UNLOCK();
896 	return EOPNOTSUPP;
897     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
898 	ifp = &V_multicast_register_if;
899 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
900 	if (V_reg_vif_num == VIFI_INVALID) {
901 	    if_initname(&V_multicast_register_if, "register_vif", 0);
902 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
903 	    V_reg_vif_num = vifcp->vifc_vifi;
904 	}
905     } else {		/* Make sure the interface supports multicast */
906 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
907 	    VIF_UNLOCK();
908 	    return EOPNOTSUPP;
909 	}
910 
911 	/* Enable promiscuous reception of all IP multicasts from the if */
912 	error = if_allmulti(ifp, 1);
913 	if (error) {
914 	    VIF_UNLOCK();
915 	    return error;
916 	}
917     }
918 
919     vifp->v_flags     = vifcp->vifc_flags;
920     vifp->v_threshold = vifcp->vifc_threshold;
921     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
922     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
923     vifp->v_ifp       = ifp;
924     /* initialize per vif pkt counters */
925     vifp->v_pkt_in    = 0;
926     vifp->v_pkt_out   = 0;
927     vifp->v_bytes_in  = 0;
928     vifp->v_bytes_out = 0;
929 
930     /* Adjust numvifs up if the vifi is higher than numvifs */
931     if (V_numvifs <= vifcp->vifc_vifi)
932 	V_numvifs = vifcp->vifc_vifi + 1;
933 
934     VIF_UNLOCK();
935 
936     CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
937 	(int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr),
938 	(int)vifcp->vifc_threshold);
939 
940     return 0;
941 }
942 
943 /*
944  * Delete a vif from the vif table
945  */
946 static int
947 del_vif_locked(vifi_t vifi)
948 {
949     struct vif *vifp;
950 
951     VIF_LOCK_ASSERT();
952 
953     if (vifi >= V_numvifs) {
954 	return EINVAL;
955     }
956     vifp = &V_viftable[vifi];
957     if (in_nullhost(vifp->v_lcl_addr)) {
958 	return EADDRNOTAVAIL;
959     }
960 
961     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
962 	if_allmulti(vifp->v_ifp, 0);
963 
964     if (vifp->v_flags & VIFF_REGISTER)
965 	V_reg_vif_num = VIFI_INVALID;
966 
967     bzero((caddr_t)vifp, sizeof (*vifp));
968 
969     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
970 
971     /* Adjust numvifs down */
972     for (vifi = V_numvifs; vifi > 0; vifi--)
973 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
974 	    break;
975     V_numvifs = vifi;
976 
977     return 0;
978 }
979 
980 static int
981 del_vif(vifi_t vifi)
982 {
983     int cc;
984 
985     VIF_LOCK();
986     cc = del_vif_locked(vifi);
987     VIF_UNLOCK();
988 
989     return cc;
990 }
991 
992 /*
993  * update an mfc entry without resetting counters and S,G addresses.
994  */
995 static void
996 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
997 {
998     int i;
999 
1000     rt->mfc_parent = mfccp->mfcc_parent;
1001     for (i = 0; i < V_numvifs; i++) {
1002 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1003 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1004 	    MRT_MFC_FLAGS_ALL;
1005     }
1006     /* set the RP address */
1007     if (V_mrt_api_config & MRT_MFC_RP)
1008 	rt->mfc_rp = mfccp->mfcc_rp;
1009     else
1010 	rt->mfc_rp.s_addr = INADDR_ANY;
1011 }
1012 
1013 /*
1014  * fully initialize an mfc entry from the parameter.
1015  */
1016 static void
1017 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1018 {
1019     rt->mfc_origin     = mfccp->mfcc_origin;
1020     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1021 
1022     update_mfc_params(rt, mfccp);
1023 
1024     /* initialize pkt counters per src-grp */
1025     rt->mfc_pkt_cnt    = 0;
1026     rt->mfc_byte_cnt   = 0;
1027     rt->mfc_wrong_if   = 0;
1028     timevalclear(&rt->mfc_last_assert);
1029 }
1030 
1031 static void
1032 expire_mfc(struct mfc *rt)
1033 {
1034 	struct rtdetq *rte, *nrte;
1035 
1036 	MFC_LOCK_ASSERT();
1037 
1038 	free_bw_list(rt->mfc_bw_meter);
1039 
1040 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1041 		m_freem(rte->m);
1042 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1043 		free(rte, M_MRTABLE);
1044 	}
1045 
1046 	LIST_REMOVE(rt, mfc_hash);
1047 	free(rt, M_MRTABLE);
1048 }
1049 
1050 /*
1051  * Add an mfc entry
1052  */
1053 static int
1054 add_mfc(struct mfcctl2 *mfccp)
1055 {
1056     struct mfc *rt;
1057     struct rtdetq *rte, *nrte;
1058     u_long hash = 0;
1059     u_short nstl;
1060 
1061     VIF_LOCK();
1062     MFC_LOCK();
1063 
1064     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1065 
1066     /* If an entry already exists, just update the fields */
1067     if (rt) {
1068 	CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x",
1069 	    __func__, inet_ntoa(mfccp->mfcc_origin),
1070 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1071 	    mfccp->mfcc_parent);
1072 	update_mfc_params(rt, mfccp);
1073 	MFC_UNLOCK();
1074 	VIF_UNLOCK();
1075 	return (0);
1076     }
1077 
1078     /*
1079      * Find the entry for which the upcall was made and update
1080      */
1081     nstl = 0;
1082     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1083     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1084 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1085 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1086 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1087 		CTR5(KTR_IPMF,
1088 		    "%s: add mfc orig %s group %lx parent %x qh %p",
1089 		    __func__, inet_ntoa(mfccp->mfcc_origin),
1090 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1091 		    mfccp->mfcc_parent,
1092 		    TAILQ_FIRST(&rt->mfc_stall));
1093 		if (nstl++)
1094 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1095 
1096 		init_mfc_params(rt, mfccp);
1097 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1098 		V_nexpire[hash]--;
1099 
1100 		/* Free queued packets, but attempt to forward them first. */
1101 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1102 			if (rte->ifp != NULL)
1103 				ip_mdq(rte->m, rte->ifp, rt, -1);
1104 			m_freem(rte->m);
1105 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1106 			rt->mfc_nstall--;
1107 			free(rte, M_MRTABLE);
1108 		}
1109 	}
1110     }
1111 
1112     /*
1113      * It is possible that an entry is being inserted without an upcall
1114      */
1115     if (nstl == 0) {
1116 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1117 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1118 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1119 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1120 			init_mfc_params(rt, mfccp);
1121 			if (rt->mfc_expire)
1122 			    V_nexpire[hash]--;
1123 			rt->mfc_expire = 0;
1124 			break; /* XXX */
1125 		}
1126 	}
1127 
1128 	if (rt == NULL) {		/* no upcall, so make a new entry */
1129 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1130 	    if (rt == NULL) {
1131 		MFC_UNLOCK();
1132 		VIF_UNLOCK();
1133 		return (ENOBUFS);
1134 	    }
1135 
1136 	    init_mfc_params(rt, mfccp);
1137 	    TAILQ_INIT(&rt->mfc_stall);
1138 	    rt->mfc_nstall = 0;
1139 
1140 	    rt->mfc_expire     = 0;
1141 	    rt->mfc_bw_meter = NULL;
1142 
1143 	    /* insert new entry at head of hash chain */
1144 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1145 	}
1146     }
1147 
1148     MFC_UNLOCK();
1149     VIF_UNLOCK();
1150 
1151     return (0);
1152 }
1153 
1154 /*
1155  * Delete an mfc entry
1156  */
1157 static int
1158 del_mfc(struct mfcctl2 *mfccp)
1159 {
1160     struct in_addr	origin;
1161     struct in_addr	mcastgrp;
1162     struct mfc		*rt;
1163 
1164     origin = mfccp->mfcc_origin;
1165     mcastgrp = mfccp->mfcc_mcastgrp;
1166 
1167     CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__,
1168 	inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr));
1169 
1170     MFC_LOCK();
1171 
1172     rt = mfc_find(&origin, &mcastgrp);
1173     if (rt == NULL) {
1174 	MFC_UNLOCK();
1175 	return EADDRNOTAVAIL;
1176     }
1177 
1178     /*
1179      * free the bw_meter entries
1180      */
1181     free_bw_list(rt->mfc_bw_meter);
1182     rt->mfc_bw_meter = NULL;
1183 
1184     LIST_REMOVE(rt, mfc_hash);
1185     free(rt, M_MRTABLE);
1186 
1187     MFC_UNLOCK();
1188 
1189     return (0);
1190 }
1191 
1192 /*
1193  * Send a message to the routing daemon on the multicast routing socket.
1194  */
1195 static int
1196 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1197 {
1198     if (s) {
1199 	SOCKBUF_LOCK(&s->so_rcv);
1200 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1201 	    NULL) != 0) {
1202 	    sorwakeup_locked(s);
1203 	    return 0;
1204 	}
1205 	SOCKBUF_UNLOCK(&s->so_rcv);
1206     }
1207     m_freem(mm);
1208     return -1;
1209 }
1210 
1211 /*
1212  * IP multicast forwarding function. This function assumes that the packet
1213  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1214  * pointed to by "ifp", and the packet is to be relayed to other networks
1215  * that have members of the packet's destination IP multicast group.
1216  *
1217  * The packet is returned unscathed to the caller, unless it is
1218  * erroneous, in which case a non-zero return value tells the caller to
1219  * discard it.
1220  */
1221 
1222 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1223 
1224 static int
1225 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1226     struct ip_moptions *imo)
1227 {
1228     struct mfc *rt;
1229     int error;
1230     vifi_t vifi;
1231 
1232     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p",
1233 	inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1234 
1235     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1236 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1237 	/*
1238 	 * Packet arrived via a physical interface or
1239 	 * an encapsulated tunnel or a register_vif.
1240 	 */
1241     } else {
1242 	/*
1243 	 * Packet arrived through a source-route tunnel.
1244 	 * Source-route tunnels are no longer supported.
1245 	 */
1246 	return (1);
1247     }
1248 
1249     VIF_LOCK();
1250     MFC_LOCK();
1251     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1252 	if (ip->ip_ttl < MAXTTL)
1253 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1254 	error = ip_mdq(m, ifp, NULL, vifi);
1255 	MFC_UNLOCK();
1256 	VIF_UNLOCK();
1257 	return error;
1258     }
1259 
1260     /*
1261      * Don't forward a packet with time-to-live of zero or one,
1262      * or a packet destined to a local-only group.
1263      */
1264     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1265 	MFC_UNLOCK();
1266 	VIF_UNLOCK();
1267 	return 0;
1268     }
1269 
1270     /*
1271      * Determine forwarding vifs from the forwarding cache table
1272      */
1273     MRTSTAT_INC(mrts_mfc_lookups);
1274     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1275 
1276     /* Entry exists, so forward if necessary */
1277     if (rt != NULL) {
1278 	error = ip_mdq(m, ifp, rt, -1);
1279 	MFC_UNLOCK();
1280 	VIF_UNLOCK();
1281 	return error;
1282     } else {
1283 	/*
1284 	 * If we don't have a route for packet's origin,
1285 	 * Make a copy of the packet & send message to routing daemon
1286 	 */
1287 
1288 	struct mbuf *mb0;
1289 	struct rtdetq *rte;
1290 	u_long hash;
1291 	int hlen = ip->ip_hl << 2;
1292 
1293 	MRTSTAT_INC(mrts_mfc_misses);
1294 	MRTSTAT_INC(mrts_no_route);
1295 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)",
1296 	    inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr));
1297 
1298 	/*
1299 	 * Allocate mbufs early so that we don't do extra work if we are
1300 	 * just going to fail anyway.  Make sure to pullup the header so
1301 	 * that other people can't step on it.
1302 	 */
1303 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1304 	    M_NOWAIT|M_ZERO);
1305 	if (rte == NULL) {
1306 	    MFC_UNLOCK();
1307 	    VIF_UNLOCK();
1308 	    return ENOBUFS;
1309 	}
1310 
1311 	mb0 = m_copypacket(m, M_NOWAIT);
1312 	if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1313 	    mb0 = m_pullup(mb0, hlen);
1314 	if (mb0 == NULL) {
1315 	    free(rte, M_MRTABLE);
1316 	    MFC_UNLOCK();
1317 	    VIF_UNLOCK();
1318 	    return ENOBUFS;
1319 	}
1320 
1321 	/* is there an upcall waiting for this flow ? */
1322 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1323 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1324 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1325 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1326 		    !TAILQ_EMPTY(&rt->mfc_stall))
1327 			break;
1328 	}
1329 
1330 	if (rt == NULL) {
1331 	    int i;
1332 	    struct igmpmsg *im;
1333 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1334 	    struct mbuf *mm;
1335 
1336 	    /*
1337 	     * Locate the vifi for the incoming interface for this packet.
1338 	     * If none found, drop packet.
1339 	     */
1340 	    for (vifi = 0; vifi < V_numvifs &&
1341 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1342 		;
1343 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1344 		goto non_fatal;
1345 
1346 	    /* no upcall, so make a new entry */
1347 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1348 	    if (rt == NULL)
1349 		goto fail;
1350 
1351 	    /* Make a copy of the header to send to the user level process */
1352 	    mm = m_copy(mb0, 0, hlen);
1353 	    if (mm == NULL)
1354 		goto fail1;
1355 
1356 	    /*
1357 	     * Send message to routing daemon to install
1358 	     * a route into the kernel table
1359 	     */
1360 
1361 	    im = mtod(mm, struct igmpmsg *);
1362 	    im->im_msgtype = IGMPMSG_NOCACHE;
1363 	    im->im_mbz = 0;
1364 	    im->im_vif = vifi;
1365 
1366 	    MRTSTAT_INC(mrts_upcalls);
1367 
1368 	    k_igmpsrc.sin_addr = ip->ip_src;
1369 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1370 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1371 		MRTSTAT_INC(mrts_upq_sockfull);
1372 fail1:
1373 		free(rt, M_MRTABLE);
1374 fail:
1375 		free(rte, M_MRTABLE);
1376 		m_freem(mb0);
1377 		MFC_UNLOCK();
1378 		VIF_UNLOCK();
1379 		return ENOBUFS;
1380 	    }
1381 
1382 	    /* insert new entry at head of hash chain */
1383 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1384 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1385 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1386 	    V_nexpire[hash]++;
1387 	    for (i = 0; i < V_numvifs; i++) {
1388 		rt->mfc_ttls[i] = 0;
1389 		rt->mfc_flags[i] = 0;
1390 	    }
1391 	    rt->mfc_parent = -1;
1392 
1393 	    /* clear the RP address */
1394 	    rt->mfc_rp.s_addr = INADDR_ANY;
1395 	    rt->mfc_bw_meter = NULL;
1396 
1397 	    /* initialize pkt counters per src-grp */
1398 	    rt->mfc_pkt_cnt = 0;
1399 	    rt->mfc_byte_cnt = 0;
1400 	    rt->mfc_wrong_if = 0;
1401 	    timevalclear(&rt->mfc_last_assert);
1402 
1403 	    TAILQ_INIT(&rt->mfc_stall);
1404 	    rt->mfc_nstall = 0;
1405 
1406 	    /* link into table */
1407 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1408 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1409 	    rt->mfc_nstall++;
1410 
1411 	} else {
1412 	    /* determine if queue has overflowed */
1413 	    if (rt->mfc_nstall > MAX_UPQ) {
1414 		MRTSTAT_INC(mrts_upq_ovflw);
1415 non_fatal:
1416 		free(rte, M_MRTABLE);
1417 		m_freem(mb0);
1418 		MFC_UNLOCK();
1419 		VIF_UNLOCK();
1420 		return (0);
1421 	    }
1422 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1423 	    rt->mfc_nstall++;
1424 	}
1425 
1426 	rte->m			= mb0;
1427 	rte->ifp		= ifp;
1428 
1429 	MFC_UNLOCK();
1430 	VIF_UNLOCK();
1431 
1432 	return 0;
1433     }
1434 }
1435 
1436 /*
1437  * Clean up the cache entry if upcall is not serviced
1438  */
1439 static void
1440 expire_upcalls(void *arg)
1441 {
1442     int i;
1443 
1444     CURVNET_SET((struct vnet *) arg);
1445 
1446     MFC_LOCK();
1447 
1448     for (i = 0; i < mfchashsize; i++) {
1449 	struct mfc *rt, *nrt;
1450 
1451 	if (V_nexpire[i] == 0)
1452 	    continue;
1453 
1454 	for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
1455 		nrt = LIST_NEXT(rt, mfc_hash);
1456 
1457 		if (TAILQ_EMPTY(&rt->mfc_stall))
1458 			continue;
1459 
1460 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1461 			continue;
1462 
1463 		/*
1464 		 * free the bw_meter entries
1465 		 */
1466 		while (rt->mfc_bw_meter != NULL) {
1467 		    struct bw_meter *x = rt->mfc_bw_meter;
1468 
1469 		    rt->mfc_bw_meter = x->bm_mfc_next;
1470 		    free(x, M_BWMETER);
1471 		}
1472 
1473 		MRTSTAT_INC(mrts_cache_cleanups);
1474 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1475 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1476 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1477 
1478 		expire_mfc(rt);
1479 	    }
1480     }
1481 
1482     MFC_UNLOCK();
1483 
1484     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1485 	curvnet);
1486 
1487     CURVNET_RESTORE();
1488 }
1489 
1490 /*
1491  * Packet forwarding routine once entry in the cache is made
1492  */
1493 static int
1494 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1495 {
1496     struct ip  *ip = mtod(m, struct ip *);
1497     vifi_t vifi;
1498     int plen = ntohs(ip->ip_len);
1499 
1500     VIF_LOCK_ASSERT();
1501 
1502     /*
1503      * If xmt_vif is not -1, send on only the requested vif.
1504      *
1505      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1506      */
1507     if (xmt_vif < V_numvifs) {
1508 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1509 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1510 	else
1511 		phyint_send(ip, V_viftable + xmt_vif, m);
1512 	return 1;
1513     }
1514 
1515     /*
1516      * Don't forward if it didn't arrive from the parent vif for its origin.
1517      */
1518     vifi = rt->mfc_parent;
1519     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1520 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1521 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1522 	MRTSTAT_INC(mrts_wrong_if);
1523 	++rt->mfc_wrong_if;
1524 	/*
1525 	 * If we are doing PIM assert processing, send a message
1526 	 * to the routing daemon.
1527 	 *
1528 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1529 	 * can complete the SPT switch, regardless of the type
1530 	 * of the iif (broadcast media, GRE tunnel, etc).
1531 	 */
1532 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1533 	    V_viftable[vifi].v_ifp) {
1534 
1535 	    if (ifp == &V_multicast_register_if)
1536 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1537 
1538 	    /* Get vifi for the incoming packet */
1539 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1540 		vifi++)
1541 		;
1542 	    if (vifi >= V_numvifs)
1543 		return 0;	/* The iif is not found: ignore the packet. */
1544 
1545 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1546 		return 0;	/* WRONGVIF disabled: ignore the packet */
1547 
1548 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1549 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1550 		struct igmpmsg *im;
1551 		int hlen = ip->ip_hl << 2;
1552 		struct mbuf *mm = m_copy(m, 0, hlen);
1553 
1554 		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1555 		    mm = m_pullup(mm, hlen);
1556 		if (mm == NULL)
1557 		    return ENOBUFS;
1558 
1559 		im = mtod(mm, struct igmpmsg *);
1560 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1561 		im->im_mbz		= 0;
1562 		im->im_vif		= vifi;
1563 
1564 		MRTSTAT_INC(mrts_upcalls);
1565 
1566 		k_igmpsrc.sin_addr = im->im_src;
1567 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1568 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1569 		    MRTSTAT_INC(mrts_upq_sockfull);
1570 		    return ENOBUFS;
1571 		}
1572 	    }
1573 	}
1574 	return 0;
1575     }
1576 
1577 
1578     /* If I sourced this packet, it counts as output, else it was input. */
1579     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1580 	V_viftable[vifi].v_pkt_out++;
1581 	V_viftable[vifi].v_bytes_out += plen;
1582     } else {
1583 	V_viftable[vifi].v_pkt_in++;
1584 	V_viftable[vifi].v_bytes_in += plen;
1585     }
1586     rt->mfc_pkt_cnt++;
1587     rt->mfc_byte_cnt += plen;
1588 
1589     /*
1590      * For each vif, decide if a copy of the packet should be forwarded.
1591      * Forward if:
1592      *		- the ttl exceeds the vif's threshold
1593      *		- there are group members downstream on interface
1594      */
1595     for (vifi = 0; vifi < V_numvifs; vifi++)
1596 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1597 	    V_viftable[vifi].v_pkt_out++;
1598 	    V_viftable[vifi].v_bytes_out += plen;
1599 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1600 		pim_register_send(ip, V_viftable + vifi, m, rt);
1601 	    else
1602 		phyint_send(ip, V_viftable + vifi, m);
1603 	}
1604 
1605     /*
1606      * Perform upcall-related bw measuring.
1607      */
1608     if (rt->mfc_bw_meter != NULL) {
1609 	struct bw_meter *x;
1610 	struct timeval now;
1611 
1612 	microtime(&now);
1613 	MFC_LOCK_ASSERT();
1614 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1615 	    bw_meter_receive_packet(x, plen, &now);
1616     }
1617 
1618     return 0;
1619 }
1620 
1621 /*
1622  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1623  */
1624 static int
1625 X_legal_vif_num(int vif)
1626 {
1627 	int ret;
1628 
1629 	ret = 0;
1630 	if (vif < 0)
1631 		return (ret);
1632 
1633 	VIF_LOCK();
1634 	if (vif < V_numvifs)
1635 		ret = 1;
1636 	VIF_UNLOCK();
1637 
1638 	return (ret);
1639 }
1640 
1641 /*
1642  * Return the local address used by this vif
1643  */
1644 static u_long
1645 X_ip_mcast_src(int vifi)
1646 {
1647 	in_addr_t addr;
1648 
1649 	addr = INADDR_ANY;
1650 	if (vifi < 0)
1651 		return (addr);
1652 
1653 	VIF_LOCK();
1654 	if (vifi < V_numvifs)
1655 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1656 	VIF_UNLOCK();
1657 
1658 	return (addr);
1659 }
1660 
1661 static void
1662 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1663 {
1664     struct mbuf *mb_copy;
1665     int hlen = ip->ip_hl << 2;
1666 
1667     VIF_LOCK_ASSERT();
1668 
1669     /*
1670      * Make a new reference to the packet; make sure that
1671      * the IP header is actually copied, not just referenced,
1672      * so that ip_output() only scribbles on the copy.
1673      */
1674     mb_copy = m_copypacket(m, M_NOWAIT);
1675     if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1676 	mb_copy = m_pullup(mb_copy, hlen);
1677     if (mb_copy == NULL)
1678 	return;
1679 
1680     send_packet(vifp, mb_copy);
1681 }
1682 
1683 static void
1684 send_packet(struct vif *vifp, struct mbuf *m)
1685 {
1686 	struct ip_moptions imo;
1687 	struct in_multi *imm[2];
1688 	int error;
1689 
1690 	VIF_LOCK_ASSERT();
1691 
1692 	imo.imo_multicast_ifp  = vifp->v_ifp;
1693 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1694 	imo.imo_multicast_loop = 1;
1695 	imo.imo_multicast_vif  = -1;
1696 	imo.imo_num_memberships = 0;
1697 	imo.imo_max_memberships = 2;
1698 	imo.imo_membership  = &imm[0];
1699 
1700 	/*
1701 	 * Re-entrancy should not be a problem here, because
1702 	 * the packets that we send out and are looped back at us
1703 	 * should get rejected because they appear to come from
1704 	 * the loopback interface, thus preventing looping.
1705 	 */
1706 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1707 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1708 	    (ptrdiff_t)(vifp - V_viftable), error);
1709 }
1710 
1711 /*
1712  * Stubs for old RSVP socket shim implementation.
1713  */
1714 
1715 static int
1716 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1717 {
1718 
1719 	return (EOPNOTSUPP);
1720 }
1721 
1722 static void
1723 X_ip_rsvp_force_done(struct socket *so __unused)
1724 {
1725 
1726 }
1727 
1728 static void
1729 X_rsvp_input(struct mbuf *m, int off __unused)
1730 {
1731 
1732 	if (!V_rsvp_on)
1733 		m_freem(m);
1734 }
1735 
1736 /*
1737  * Code for bandwidth monitors
1738  */
1739 
1740 /*
1741  * Define common interface for timeval-related methods
1742  */
1743 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1744 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1745 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1746 
1747 static uint32_t
1748 compute_bw_meter_flags(struct bw_upcall *req)
1749 {
1750     uint32_t flags = 0;
1751 
1752     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1753 	flags |= BW_METER_UNIT_PACKETS;
1754     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1755 	flags |= BW_METER_UNIT_BYTES;
1756     if (req->bu_flags & BW_UPCALL_GEQ)
1757 	flags |= BW_METER_GEQ;
1758     if (req->bu_flags & BW_UPCALL_LEQ)
1759 	flags |= BW_METER_LEQ;
1760 
1761     return flags;
1762 }
1763 
1764 /*
1765  * Add a bw_meter entry
1766  */
1767 static int
1768 add_bw_upcall(struct bw_upcall *req)
1769 {
1770     struct mfc *mfc;
1771     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1772 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1773     struct timeval now;
1774     struct bw_meter *x;
1775     uint32_t flags;
1776 
1777     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1778 	return EOPNOTSUPP;
1779 
1780     /* Test if the flags are valid */
1781     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1782 	return EINVAL;
1783     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1784 	return EINVAL;
1785     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1786 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1787 	return EINVAL;
1788 
1789     /* Test if the threshold time interval is valid */
1790     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1791 	return EINVAL;
1792 
1793     flags = compute_bw_meter_flags(req);
1794 
1795     /*
1796      * Find if we have already same bw_meter entry
1797      */
1798     MFC_LOCK();
1799     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1800     if (mfc == NULL) {
1801 	MFC_UNLOCK();
1802 	return EADDRNOTAVAIL;
1803     }
1804     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1805 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1806 			   &req->bu_threshold.b_time, ==)) &&
1807 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1808 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1809 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1810 	    MFC_UNLOCK();
1811 	    return 0;		/* XXX Already installed */
1812 	}
1813     }
1814 
1815     /* Allocate the new bw_meter entry */
1816     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1817     if (x == NULL) {
1818 	MFC_UNLOCK();
1819 	return ENOBUFS;
1820     }
1821 
1822     /* Set the new bw_meter entry */
1823     x->bm_threshold.b_time = req->bu_threshold.b_time;
1824     microtime(&now);
1825     x->bm_start_time = now;
1826     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1827     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1828     x->bm_measured.b_packets = 0;
1829     x->bm_measured.b_bytes = 0;
1830     x->bm_flags = flags;
1831     x->bm_time_next = NULL;
1832     x->bm_time_hash = BW_METER_BUCKETS;
1833 
1834     /* Add the new bw_meter entry to the front of entries for this MFC */
1835     x->bm_mfc = mfc;
1836     x->bm_mfc_next = mfc->mfc_bw_meter;
1837     mfc->mfc_bw_meter = x;
1838     schedule_bw_meter(x, &now);
1839     MFC_UNLOCK();
1840 
1841     return 0;
1842 }
1843 
1844 static void
1845 free_bw_list(struct bw_meter *list)
1846 {
1847     while (list != NULL) {
1848 	struct bw_meter *x = list;
1849 
1850 	list = list->bm_mfc_next;
1851 	unschedule_bw_meter(x);
1852 	free(x, M_BWMETER);
1853     }
1854 }
1855 
1856 /*
1857  * Delete one or multiple bw_meter entries
1858  */
1859 static int
1860 del_bw_upcall(struct bw_upcall *req)
1861 {
1862     struct mfc *mfc;
1863     struct bw_meter *x;
1864 
1865     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1866 	return EOPNOTSUPP;
1867 
1868     MFC_LOCK();
1869 
1870     /* Find the corresponding MFC entry */
1871     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1872     if (mfc == NULL) {
1873 	MFC_UNLOCK();
1874 	return EADDRNOTAVAIL;
1875     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1876 	/*
1877 	 * Delete all bw_meter entries for this mfc
1878 	 */
1879 	struct bw_meter *list;
1880 
1881 	list = mfc->mfc_bw_meter;
1882 	mfc->mfc_bw_meter = NULL;
1883 	free_bw_list(list);
1884 	MFC_UNLOCK();
1885 	return 0;
1886     } else {			/* Delete a single bw_meter entry */
1887 	struct bw_meter *prev;
1888 	uint32_t flags = 0;
1889 
1890 	flags = compute_bw_meter_flags(req);
1891 
1892 	/* Find the bw_meter entry to delete */
1893 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1894 	     prev = x, x = x->bm_mfc_next) {
1895 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1896 			       &req->bu_threshold.b_time, ==)) &&
1897 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1898 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1899 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1900 		break;
1901 	}
1902 	if (x != NULL) { /* Delete entry from the list for this MFC */
1903 	    if (prev != NULL)
1904 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1905 	    else
1906 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1907 
1908 	    unschedule_bw_meter(x);
1909 	    MFC_UNLOCK();
1910 	    /* Free the bw_meter entry */
1911 	    free(x, M_BWMETER);
1912 	    return 0;
1913 	} else {
1914 	    MFC_UNLOCK();
1915 	    return EINVAL;
1916 	}
1917     }
1918     /* NOTREACHED */
1919 }
1920 
1921 /*
1922  * Perform bandwidth measurement processing that may result in an upcall
1923  */
1924 static void
1925 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1926 {
1927     struct timeval delta;
1928 
1929     MFC_LOCK_ASSERT();
1930 
1931     delta = *nowp;
1932     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1933 
1934     if (x->bm_flags & BW_METER_GEQ) {
1935 	/*
1936 	 * Processing for ">=" type of bw_meter entry
1937 	 */
1938 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1939 	    /* Reset the bw_meter entry */
1940 	    x->bm_start_time = *nowp;
1941 	    x->bm_measured.b_packets = 0;
1942 	    x->bm_measured.b_bytes = 0;
1943 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1944 	}
1945 
1946 	/* Record that a packet is received */
1947 	x->bm_measured.b_packets++;
1948 	x->bm_measured.b_bytes += plen;
1949 
1950 	/*
1951 	 * Test if we should deliver an upcall
1952 	 */
1953 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1954 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1955 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1956 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1957 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1958 		/* Prepare an upcall for delivery */
1959 		bw_meter_prepare_upcall(x, nowp);
1960 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1961 	    }
1962 	}
1963     } else if (x->bm_flags & BW_METER_LEQ) {
1964 	/*
1965 	 * Processing for "<=" type of bw_meter entry
1966 	 */
1967 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1968 	    /*
1969 	     * We are behind time with the multicast forwarding table
1970 	     * scanning for "<=" type of bw_meter entries, so test now
1971 	     * if we should deliver an upcall.
1972 	     */
1973 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1974 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1975 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1976 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1977 		/* Prepare an upcall for delivery */
1978 		bw_meter_prepare_upcall(x, nowp);
1979 	    }
1980 	    /* Reschedule the bw_meter entry */
1981 	    unschedule_bw_meter(x);
1982 	    schedule_bw_meter(x, nowp);
1983 	}
1984 
1985 	/* Record that a packet is received */
1986 	x->bm_measured.b_packets++;
1987 	x->bm_measured.b_bytes += plen;
1988 
1989 	/*
1990 	 * Test if we should restart the measuring interval
1991 	 */
1992 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1993 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1994 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1995 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1996 	    /* Don't restart the measuring interval */
1997 	} else {
1998 	    /* Do restart the measuring interval */
1999 	    /*
2000 	     * XXX: note that we don't unschedule and schedule, because this
2001 	     * might be too much overhead per packet. Instead, when we process
2002 	     * all entries for a given timer hash bin, we check whether it is
2003 	     * really a timeout. If not, we reschedule at that time.
2004 	     */
2005 	    x->bm_start_time = *nowp;
2006 	    x->bm_measured.b_packets = 0;
2007 	    x->bm_measured.b_bytes = 0;
2008 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2009 	}
2010     }
2011 }
2012 
2013 /*
2014  * Prepare a bandwidth-related upcall
2015  */
2016 static void
2017 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2018 {
2019     struct timeval delta;
2020     struct bw_upcall *u;
2021 
2022     MFC_LOCK_ASSERT();
2023 
2024     /*
2025      * Compute the measured time interval
2026      */
2027     delta = *nowp;
2028     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2029 
2030     /*
2031      * If there are too many pending upcalls, deliver them now
2032      */
2033     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2034 	bw_upcalls_send();
2035 
2036     /*
2037      * Set the bw_upcall entry
2038      */
2039     u = &V_bw_upcalls[V_bw_upcalls_n++];
2040     u->bu_src = x->bm_mfc->mfc_origin;
2041     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2042     u->bu_threshold.b_time = x->bm_threshold.b_time;
2043     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2044     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2045     u->bu_measured.b_time = delta;
2046     u->bu_measured.b_packets = x->bm_measured.b_packets;
2047     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2048     u->bu_flags = 0;
2049     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2050 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2051     if (x->bm_flags & BW_METER_UNIT_BYTES)
2052 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2053     if (x->bm_flags & BW_METER_GEQ)
2054 	u->bu_flags |= BW_UPCALL_GEQ;
2055     if (x->bm_flags & BW_METER_LEQ)
2056 	u->bu_flags |= BW_UPCALL_LEQ;
2057 }
2058 
2059 /*
2060  * Send the pending bandwidth-related upcalls
2061  */
2062 static void
2063 bw_upcalls_send(void)
2064 {
2065     struct mbuf *m;
2066     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2067     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2068     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2069 				      0,		/* unused2 */
2070 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2071 				      0,		/* im_mbz  */
2072 				      0,		/* im_vif  */
2073 				      0,		/* unused3 */
2074 				      { 0 },		/* im_src  */
2075 				      { 0 } };		/* im_dst  */
2076 
2077     MFC_LOCK_ASSERT();
2078 
2079     if (V_bw_upcalls_n == 0)
2080 	return;			/* No pending upcalls */
2081 
2082     V_bw_upcalls_n = 0;
2083 
2084     /*
2085      * Allocate a new mbuf, initialize it with the header and
2086      * the payload for the pending calls.
2087      */
2088     m = m_gethdr(M_NOWAIT, MT_DATA);
2089     if (m == NULL) {
2090 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2091 	return;
2092     }
2093 
2094     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2095     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2096 
2097     /*
2098      * Send the upcalls
2099      * XXX do we need to set the address in k_igmpsrc ?
2100      */
2101     MRTSTAT_INC(mrts_upcalls);
2102     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2103 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2104 	MRTSTAT_INC(mrts_upq_sockfull);
2105     }
2106 }
2107 
2108 /*
2109  * Compute the timeout hash value for the bw_meter entries
2110  */
2111 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2112     do {								\
2113 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2114 									\
2115 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2116 	(hash) = next_timeval.tv_sec;					\
2117 	if (next_timeval.tv_usec)					\
2118 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2119 	(hash) %= BW_METER_BUCKETS;					\
2120     } while (0)
2121 
2122 /*
2123  * Schedule a timer to process periodically bw_meter entry of type "<="
2124  * by linking the entry in the proper hash bucket.
2125  */
2126 static void
2127 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2128 {
2129     int time_hash;
2130 
2131     MFC_LOCK_ASSERT();
2132 
2133     if (!(x->bm_flags & BW_METER_LEQ))
2134 	return;		/* XXX: we schedule timers only for "<=" entries */
2135 
2136     /*
2137      * Reset the bw_meter entry
2138      */
2139     x->bm_start_time = *nowp;
2140     x->bm_measured.b_packets = 0;
2141     x->bm_measured.b_bytes = 0;
2142     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2143 
2144     /*
2145      * Compute the timeout hash value and insert the entry
2146      */
2147     BW_METER_TIMEHASH(x, time_hash);
2148     x->bm_time_next = V_bw_meter_timers[time_hash];
2149     V_bw_meter_timers[time_hash] = x;
2150     x->bm_time_hash = time_hash;
2151 }
2152 
2153 /*
2154  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2155  * by removing the entry from the proper hash bucket.
2156  */
2157 static void
2158 unschedule_bw_meter(struct bw_meter *x)
2159 {
2160     int time_hash;
2161     struct bw_meter *prev, *tmp;
2162 
2163     MFC_LOCK_ASSERT();
2164 
2165     if (!(x->bm_flags & BW_METER_LEQ))
2166 	return;		/* XXX: we schedule timers only for "<=" entries */
2167 
2168     /*
2169      * Compute the timeout hash value and delete the entry
2170      */
2171     time_hash = x->bm_time_hash;
2172     if (time_hash >= BW_METER_BUCKETS)
2173 	return;		/* Entry was not scheduled */
2174 
2175     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2176 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2177 	if (tmp == x)
2178 	    break;
2179 
2180     if (tmp == NULL)
2181 	panic("unschedule_bw_meter: bw_meter entry not found");
2182 
2183     if (prev != NULL)
2184 	prev->bm_time_next = x->bm_time_next;
2185     else
2186 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2187 
2188     x->bm_time_next = NULL;
2189     x->bm_time_hash = BW_METER_BUCKETS;
2190 }
2191 
2192 
2193 /*
2194  * Process all "<=" type of bw_meter that should be processed now,
2195  * and for each entry prepare an upcall if necessary. Each processed
2196  * entry is rescheduled again for the (periodic) processing.
2197  *
2198  * This is run periodically (once per second normally). On each round,
2199  * all the potentially matching entries are in the hash slot that we are
2200  * looking at.
2201  */
2202 static void
2203 bw_meter_process()
2204 {
2205     uint32_t loops;
2206     int i;
2207     struct timeval now, process_endtime;
2208 
2209     microtime(&now);
2210     if (V_last_tv_sec == now.tv_sec)
2211 	return;		/* nothing to do */
2212 
2213     loops = now.tv_sec - V_last_tv_sec;
2214     V_last_tv_sec = now.tv_sec;
2215     if (loops > BW_METER_BUCKETS)
2216 	loops = BW_METER_BUCKETS;
2217 
2218     MFC_LOCK();
2219     /*
2220      * Process all bins of bw_meter entries from the one after the last
2221      * processed to the current one. On entry, i points to the last bucket
2222      * visited, so we need to increment i at the beginning of the loop.
2223      */
2224     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2225 	struct bw_meter *x, *tmp_list;
2226 
2227 	if (++i >= BW_METER_BUCKETS)
2228 	    i = 0;
2229 
2230 	/* Disconnect the list of bw_meter entries from the bin */
2231 	tmp_list = V_bw_meter_timers[i];
2232 	V_bw_meter_timers[i] = NULL;
2233 
2234 	/* Process the list of bw_meter entries */
2235 	while (tmp_list != NULL) {
2236 	    x = tmp_list;
2237 	    tmp_list = tmp_list->bm_time_next;
2238 
2239 	    /* Test if the time interval is over */
2240 	    process_endtime = x->bm_start_time;
2241 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2242 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2243 		/* Not yet: reschedule, but don't reset */
2244 		int time_hash;
2245 
2246 		BW_METER_TIMEHASH(x, time_hash);
2247 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2248 		    /*
2249 		     * XXX: somehow the bin processing is a bit ahead of time.
2250 		     * Put the entry in the next bin.
2251 		     */
2252 		    if (++time_hash >= BW_METER_BUCKETS)
2253 			time_hash = 0;
2254 		}
2255 		x->bm_time_next = V_bw_meter_timers[time_hash];
2256 		V_bw_meter_timers[time_hash] = x;
2257 		x->bm_time_hash = time_hash;
2258 
2259 		continue;
2260 	    }
2261 
2262 	    /*
2263 	     * Test if we should deliver an upcall
2264 	     */
2265 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2266 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2267 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2268 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2269 		/* Prepare an upcall for delivery */
2270 		bw_meter_prepare_upcall(x, &now);
2271 	    }
2272 
2273 	    /*
2274 	     * Reschedule for next processing
2275 	     */
2276 	    schedule_bw_meter(x, &now);
2277 	}
2278     }
2279 
2280     /* Send all upcalls that are pending delivery */
2281     bw_upcalls_send();
2282 
2283     MFC_UNLOCK();
2284 }
2285 
2286 /*
2287  * A periodic function for sending all upcalls that are pending delivery
2288  */
2289 static void
2290 expire_bw_upcalls_send(void *arg)
2291 {
2292     CURVNET_SET((struct vnet *) arg);
2293 
2294     MFC_LOCK();
2295     bw_upcalls_send();
2296     MFC_UNLOCK();
2297 
2298     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2299 	curvnet);
2300     CURVNET_RESTORE();
2301 }
2302 
2303 /*
2304  * A periodic function for periodic scanning of the multicast forwarding
2305  * table for processing all "<=" bw_meter entries.
2306  */
2307 static void
2308 expire_bw_meter_process(void *arg)
2309 {
2310     CURVNET_SET((struct vnet *) arg);
2311 
2312     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2313 	bw_meter_process();
2314 
2315     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2316 	curvnet);
2317     CURVNET_RESTORE();
2318 }
2319 
2320 /*
2321  * End of bandwidth monitoring code
2322  */
2323 
2324 /*
2325  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2326  *
2327  */
2328 static int
2329 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2330     struct mfc *rt)
2331 {
2332     struct mbuf *mb_copy, *mm;
2333 
2334     /*
2335      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2336      * rendezvous point was unspecified, and we were told not to.
2337      */
2338     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2339 	in_nullhost(rt->mfc_rp))
2340 	return 0;
2341 
2342     mb_copy = pim_register_prepare(ip, m);
2343     if (mb_copy == NULL)
2344 	return ENOBUFS;
2345 
2346     /*
2347      * Send all the fragments. Note that the mbuf for each fragment
2348      * is freed by the sending machinery.
2349      */
2350     for (mm = mb_copy; mm; mm = mb_copy) {
2351 	mb_copy = mm->m_nextpkt;
2352 	mm->m_nextpkt = 0;
2353 	mm = m_pullup(mm, sizeof(struct ip));
2354 	if (mm != NULL) {
2355 	    ip = mtod(mm, struct ip *);
2356 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2357 		pim_register_send_rp(ip, vifp, mm, rt);
2358 	    } else {
2359 		pim_register_send_upcall(ip, vifp, mm, rt);
2360 	    }
2361 	}
2362     }
2363 
2364     return 0;
2365 }
2366 
2367 /*
2368  * Return a copy of the data packet that is ready for PIM Register
2369  * encapsulation.
2370  * XXX: Note that in the returned copy the IP header is a valid one.
2371  */
2372 static struct mbuf *
2373 pim_register_prepare(struct ip *ip, struct mbuf *m)
2374 {
2375     struct mbuf *mb_copy = NULL;
2376     int mtu;
2377 
2378     /* Take care of delayed checksums */
2379     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2380 	in_delayed_cksum(m);
2381 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2382     }
2383 
2384     /*
2385      * Copy the old packet & pullup its IP header into the
2386      * new mbuf so we can modify it.
2387      */
2388     mb_copy = m_copypacket(m, M_NOWAIT);
2389     if (mb_copy == NULL)
2390 	return NULL;
2391     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2392     if (mb_copy == NULL)
2393 	return NULL;
2394 
2395     /* take care of the TTL */
2396     ip = mtod(mb_copy, struct ip *);
2397     --ip->ip_ttl;
2398 
2399     /* Compute the MTU after the PIM Register encapsulation */
2400     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2401 
2402     if (ntohs(ip->ip_len) <= mtu) {
2403 	/* Turn the IP header into a valid one */
2404 	ip->ip_sum = 0;
2405 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2406     } else {
2407 	/* Fragment the packet */
2408 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2409 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2410 	    m_freem(mb_copy);
2411 	    return NULL;
2412 	}
2413     }
2414     return mb_copy;
2415 }
2416 
2417 /*
2418  * Send an upcall with the data packet to the user-level process.
2419  */
2420 static int
2421 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2422     struct mbuf *mb_copy, struct mfc *rt)
2423 {
2424     struct mbuf *mb_first;
2425     int len = ntohs(ip->ip_len);
2426     struct igmpmsg *im;
2427     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2428 
2429     VIF_LOCK_ASSERT();
2430 
2431     /*
2432      * Add a new mbuf with an upcall header
2433      */
2434     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2435     if (mb_first == NULL) {
2436 	m_freem(mb_copy);
2437 	return ENOBUFS;
2438     }
2439     mb_first->m_data += max_linkhdr;
2440     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2441     mb_first->m_len = sizeof(struct igmpmsg);
2442     mb_first->m_next = mb_copy;
2443 
2444     /* Send message to routing daemon */
2445     im = mtod(mb_first, struct igmpmsg *);
2446     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2447     im->im_mbz		= 0;
2448     im->im_vif		= vifp - V_viftable;
2449     im->im_src		= ip->ip_src;
2450     im->im_dst		= ip->ip_dst;
2451 
2452     k_igmpsrc.sin_addr	= ip->ip_src;
2453 
2454     MRTSTAT_INC(mrts_upcalls);
2455 
2456     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2457 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2458 	MRTSTAT_INC(mrts_upq_sockfull);
2459 	return ENOBUFS;
2460     }
2461 
2462     /* Keep statistics */
2463     PIMSTAT_INC(pims_snd_registers_msgs);
2464     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2465 
2466     return 0;
2467 }
2468 
2469 /*
2470  * Encapsulate the data packet in PIM Register message and send it to the RP.
2471  */
2472 static int
2473 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2474     struct mfc *rt)
2475 {
2476     struct mbuf *mb_first;
2477     struct ip *ip_outer;
2478     struct pim_encap_pimhdr *pimhdr;
2479     int len = ntohs(ip->ip_len);
2480     vifi_t vifi = rt->mfc_parent;
2481 
2482     VIF_LOCK_ASSERT();
2483 
2484     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2485 	m_freem(mb_copy);
2486 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2487     }
2488 
2489     /*
2490      * Add a new mbuf with the encapsulating header
2491      */
2492     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2493     if (mb_first == NULL) {
2494 	m_freem(mb_copy);
2495 	return ENOBUFS;
2496     }
2497     mb_first->m_data += max_linkhdr;
2498     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2499     mb_first->m_next = mb_copy;
2500 
2501     mb_first->m_pkthdr.len = len + mb_first->m_len;
2502 
2503     /*
2504      * Fill in the encapsulating IP and PIM header
2505      */
2506     ip_outer = mtod(mb_first, struct ip *);
2507     *ip_outer = pim_encap_iphdr;
2508     ip_outer->ip_id = ip_newid();
2509     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2510 	sizeof(pim_encap_pimhdr));
2511     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2512     ip_outer->ip_dst = rt->mfc_rp;
2513     /*
2514      * Copy the inner header TOS to the outer header, and take care of the
2515      * IP_DF bit.
2516      */
2517     ip_outer->ip_tos = ip->ip_tos;
2518     if (ip->ip_off & htons(IP_DF))
2519 	ip_outer->ip_off |= htons(IP_DF);
2520     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2521 					 + sizeof(pim_encap_iphdr));
2522     *pimhdr = pim_encap_pimhdr;
2523     /* If the iif crosses a border, set the Border-bit */
2524     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2525 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2526 
2527     mb_first->m_data += sizeof(pim_encap_iphdr);
2528     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2529     mb_first->m_data -= sizeof(pim_encap_iphdr);
2530 
2531     send_packet(vifp, mb_first);
2532 
2533     /* Keep statistics */
2534     PIMSTAT_INC(pims_snd_registers_msgs);
2535     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2536 
2537     return 0;
2538 }
2539 
2540 /*
2541  * pim_encapcheck() is called by the encap4_input() path at runtime to
2542  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2543  * into the kernel.
2544  */
2545 static int
2546 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2547 {
2548 
2549 #ifdef DIAGNOSTIC
2550     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2551 #endif
2552     if (proto != IPPROTO_PIM)
2553 	return 0;	/* not for us; reject the datagram. */
2554 
2555     return 64;		/* claim the datagram. */
2556 }
2557 
2558 /*
2559  * PIM-SMv2 and PIM-DM messages processing.
2560  * Receives and verifies the PIM control messages, and passes them
2561  * up to the listening socket, using rip_input().
2562  * The only message with special processing is the PIM_REGISTER message
2563  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2564  * is passed to if_simloop().
2565  */
2566 void
2567 pim_input(struct mbuf *m, int off)
2568 {
2569     struct ip *ip = mtod(m, struct ip *);
2570     struct pim *pim;
2571     int minlen;
2572     int datalen = ntohs(ip->ip_len);
2573     int ip_tos;
2574     int iphlen = off;
2575 
2576     /* Keep statistics */
2577     PIMSTAT_INC(pims_rcv_total_msgs);
2578     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2579 
2580     /*
2581      * Validate lengths
2582      */
2583     if (datalen < PIM_MINLEN) {
2584 	PIMSTAT_INC(pims_rcv_tooshort);
2585 	CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2586 	    __func__, datalen, inet_ntoa(ip->ip_src));
2587 	m_freem(m);
2588 	return;
2589     }
2590 
2591     /*
2592      * If the packet is at least as big as a REGISTER, go agead
2593      * and grab the PIM REGISTER header size, to avoid another
2594      * possible m_pullup() later.
2595      *
2596      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2597      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2598      */
2599     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2600     /*
2601      * Get the IP and PIM headers in contiguous memory, and
2602      * possibly the PIM REGISTER header.
2603      */
2604     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
2605 	(m = m_pullup(m, minlen)) == 0) {
2606 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2607 	return;
2608     }
2609 
2610     /* m_pullup() may have given us a new mbuf so reset ip. */
2611     ip = mtod(m, struct ip *);
2612     ip_tos = ip->ip_tos;
2613 
2614     /* adjust mbuf to point to the PIM header */
2615     m->m_data += iphlen;
2616     m->m_len  -= iphlen;
2617     pim = mtod(m, struct pim *);
2618 
2619     /*
2620      * Validate checksum. If PIM REGISTER, exclude the data packet.
2621      *
2622      * XXX: some older PIMv2 implementations don't make this distinction,
2623      * so for compatibility reason perform the checksum over part of the
2624      * message, and if error, then over the whole message.
2625      */
2626     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2627 	/* do nothing, checksum okay */
2628     } else if (in_cksum(m, datalen)) {
2629 	PIMSTAT_INC(pims_rcv_badsum);
2630 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2631 	m_freem(m);
2632 	return;
2633     }
2634 
2635     /* PIM version check */
2636     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2637 	PIMSTAT_INC(pims_rcv_badversion);
2638 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2639 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2640 	m_freem(m);
2641 	return;
2642     }
2643 
2644     /* restore mbuf back to the outer IP */
2645     m->m_data -= iphlen;
2646     m->m_len  += iphlen;
2647 
2648     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2649 	/*
2650 	 * Since this is a REGISTER, we'll make a copy of the register
2651 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2652 	 * routing daemon.
2653 	 */
2654 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2655 	struct mbuf *mcp;
2656 	struct ip *encap_ip;
2657 	u_int32_t *reghdr;
2658 	struct ifnet *vifp;
2659 
2660 	VIF_LOCK();
2661 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2662 	    VIF_UNLOCK();
2663 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2664 		(int)V_reg_vif_num);
2665 	    m_freem(m);
2666 	    return;
2667 	}
2668 	/* XXX need refcnt? */
2669 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2670 	VIF_UNLOCK();
2671 
2672 	/*
2673 	 * Validate length
2674 	 */
2675 	if (datalen < PIM_REG_MINLEN) {
2676 	    PIMSTAT_INC(pims_rcv_tooshort);
2677 	    PIMSTAT_INC(pims_rcv_badregisters);
2678 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2679 	    m_freem(m);
2680 	    return;
2681 	}
2682 
2683 	reghdr = (u_int32_t *)(pim + 1);
2684 	encap_ip = (struct ip *)(reghdr + 1);
2685 
2686 	CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2687 	    __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2688 
2689 	/* verify the version number of the inner packet */
2690 	if (encap_ip->ip_v != IPVERSION) {
2691 	    PIMSTAT_INC(pims_rcv_badregisters);
2692 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2693 	    m_freem(m);
2694 	    return;
2695 	}
2696 
2697 	/* verify the inner packet is destined to a mcast group */
2698 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2699 	    PIMSTAT_INC(pims_rcv_badregisters);
2700 	    CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2701 		inet_ntoa(encap_ip->ip_dst));
2702 	    m_freem(m);
2703 	    return;
2704 	}
2705 
2706 	/* If a NULL_REGISTER, pass it to the daemon */
2707 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2708 	    goto pim_input_to_daemon;
2709 
2710 	/*
2711 	 * Copy the TOS from the outer IP header to the inner IP header.
2712 	 */
2713 	if (encap_ip->ip_tos != ip_tos) {
2714 	    /* Outer TOS -> inner TOS */
2715 	    encap_ip->ip_tos = ip_tos;
2716 	    /* Recompute the inner header checksum. Sigh... */
2717 
2718 	    /* adjust mbuf to point to the inner IP header */
2719 	    m->m_data += (iphlen + PIM_MINLEN);
2720 	    m->m_len  -= (iphlen + PIM_MINLEN);
2721 
2722 	    encap_ip->ip_sum = 0;
2723 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2724 
2725 	    /* restore mbuf to point back to the outer IP header */
2726 	    m->m_data -= (iphlen + PIM_MINLEN);
2727 	    m->m_len  += (iphlen + PIM_MINLEN);
2728 	}
2729 
2730 	/*
2731 	 * Decapsulate the inner IP packet and loopback to forward it
2732 	 * as a normal multicast packet. Also, make a copy of the
2733 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2734 	 * to pass to the daemon later, so it can take the appropriate
2735 	 * actions (e.g., send back PIM_REGISTER_STOP).
2736 	 * XXX: here m->m_data points to the outer IP header.
2737 	 */
2738 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2739 	if (mcp == NULL) {
2740 	    CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2741 	    m_freem(m);
2742 	    return;
2743 	}
2744 
2745 	/* Keep statistics */
2746 	/* XXX: registers_bytes include only the encap. mcast pkt */
2747 	PIMSTAT_INC(pims_rcv_registers_msgs);
2748 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2749 
2750 	/*
2751 	 * forward the inner ip packet; point m_data at the inner ip.
2752 	 */
2753 	m_adj(m, iphlen + PIM_MINLEN);
2754 
2755 	CTR4(KTR_IPMF,
2756 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2757 	    __func__,
2758 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2759 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2760 	    (int)V_reg_vif_num);
2761 
2762 	/* NB: vifp was collected above; can it change on us? */
2763 	if_simloop(vifp, m, dst.sin_family, 0);
2764 
2765 	/* prepare the register head to send to the mrouting daemon */
2766 	m = mcp;
2767     }
2768 
2769 pim_input_to_daemon:
2770     /*
2771      * Pass the PIM message up to the daemon; if it is a Register message,
2772      * pass the 'head' only up to the daemon. This includes the
2773      * outer IP header, PIM header, PIM-Register header and the
2774      * inner IP header.
2775      * XXX: the outer IP header pkt size of a Register is not adjust to
2776      * reflect the fact that the inner multicast data is truncated.
2777      */
2778     rip_input(m, iphlen);
2779 
2780     return;
2781 }
2782 
2783 static int
2784 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2785 {
2786 	struct mfc	*rt;
2787 	int		 error, i;
2788 
2789 	if (req->newptr)
2790 		return (EPERM);
2791 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2792 		return (0);
2793 	error = sysctl_wire_old_buffer(req, 0);
2794 	if (error)
2795 		return (error);
2796 
2797 	MFC_LOCK();
2798 	for (i = 0; i < mfchashsize; i++) {
2799 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2800 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2801 			if (error)
2802 				goto out_locked;
2803 		}
2804 	}
2805 out_locked:
2806 	MFC_UNLOCK();
2807 	return (error);
2808 }
2809 
2810 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2811     sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2812     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2813 
2814 static void
2815 vnet_mroute_init(const void *unused __unused)
2816 {
2817 
2818 	MALLOC(V_nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2819 	bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2820 	callout_init(&V_expire_upcalls_ch, CALLOUT_MPSAFE);
2821 	callout_init(&V_bw_upcalls_ch, CALLOUT_MPSAFE);
2822 	callout_init(&V_bw_meter_ch, CALLOUT_MPSAFE);
2823 }
2824 
2825 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mroute_init,
2826 	NULL);
2827 
2828 static void
2829 vnet_mroute_uninit(const void *unused __unused)
2830 {
2831 
2832 	FREE(V_nexpire, M_MRTABLE);
2833 	V_nexpire = NULL;
2834 }
2835 
2836 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE,
2837 	vnet_mroute_uninit, NULL);
2838 
2839 static int
2840 ip_mroute_modevent(module_t mod, int type, void *unused)
2841 {
2842 
2843     switch (type) {
2844     case MOD_LOAD:
2845 	MROUTER_LOCK_INIT();
2846 
2847 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2848 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2849 	if (if_detach_event_tag == NULL) {
2850 		printf("ip_mroute: unable to register "
2851 		    "ifnet_departure_event handler\n");
2852 		MROUTER_LOCK_DESTROY();
2853 		return (EINVAL);
2854 	}
2855 
2856 	MFC_LOCK_INIT();
2857 	VIF_LOCK_INIT();
2858 
2859 	mfchashsize = MFCHASHSIZE;
2860 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2861 	    !powerof2(mfchashsize)) {
2862 		printf("WARNING: %s not a power of 2; using default\n",
2863 		    "net.inet.ip.mfchashsize");
2864 		mfchashsize = MFCHASHSIZE;
2865 	}
2866 
2867 	pim_squelch_wholepkt = 0;
2868 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2869 	    &pim_squelch_wholepkt);
2870 
2871 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2872 	    pim_encapcheck, &in_pim_protosw, NULL);
2873 	if (pim_encap_cookie == NULL) {
2874 		printf("ip_mroute: unable to attach pim encap\n");
2875 		VIF_LOCK_DESTROY();
2876 		MFC_LOCK_DESTROY();
2877 		MROUTER_LOCK_DESTROY();
2878 		return (EINVAL);
2879 	}
2880 
2881 	ip_mcast_src = X_ip_mcast_src;
2882 	ip_mforward = X_ip_mforward;
2883 	ip_mrouter_done = X_ip_mrouter_done;
2884 	ip_mrouter_get = X_ip_mrouter_get;
2885 	ip_mrouter_set = X_ip_mrouter_set;
2886 
2887 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2888 	ip_rsvp_vif = X_ip_rsvp_vif;
2889 
2890 	legal_vif_num = X_legal_vif_num;
2891 	mrt_ioctl = X_mrt_ioctl;
2892 	rsvp_input_p = X_rsvp_input;
2893 	break;
2894 
2895     case MOD_UNLOAD:
2896 	/*
2897 	 * Typically module unload happens after the user-level
2898 	 * process has shutdown the kernel services (the check
2899 	 * below insures someone can't just yank the module out
2900 	 * from under a running process).  But if the module is
2901 	 * just loaded and then unloaded w/o starting up a user
2902 	 * process we still need to cleanup.
2903 	 */
2904 	MROUTER_LOCK();
2905 	if (ip_mrouter_cnt != 0) {
2906 	    MROUTER_UNLOCK();
2907 	    return (EINVAL);
2908 	}
2909 	ip_mrouter_unloading = 1;
2910 	MROUTER_UNLOCK();
2911 
2912 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2913 
2914 	if (pim_encap_cookie) {
2915 	    encap_detach(pim_encap_cookie);
2916 	    pim_encap_cookie = NULL;
2917 	}
2918 
2919 	ip_mcast_src = NULL;
2920 	ip_mforward = NULL;
2921 	ip_mrouter_done = NULL;
2922 	ip_mrouter_get = NULL;
2923 	ip_mrouter_set = NULL;
2924 
2925 	ip_rsvp_force_done = NULL;
2926 	ip_rsvp_vif = NULL;
2927 
2928 	legal_vif_num = NULL;
2929 	mrt_ioctl = NULL;
2930 	rsvp_input_p = NULL;
2931 
2932 	VIF_LOCK_DESTROY();
2933 	MFC_LOCK_DESTROY();
2934 	MROUTER_LOCK_DESTROY();
2935 	break;
2936 
2937     default:
2938 	return EOPNOTSUPP;
2939     }
2940     return 0;
2941 }
2942 
2943 static moduledata_t ip_mroutemod = {
2944     "ip_mroute",
2945     ip_mroute_modevent,
2946     0
2947 };
2948 
2949 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_MIDDLE);
2950