xref: /freebsd/sys/netinet/ip_mroute.c (revision 4928135658a9d0eaee37003df6137ab363fcb0b4)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989 Stephen Deering
5  * Copyright (c) 1992, 1993
6  *      The Regents of the University of California.  All rights reserved.
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Stephen Deering of Stanford University.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
36  */
37 
38 /*
39  * IP multicast forwarding procedures
40  *
41  * Written by David Waitzman, BBN Labs, August 1988.
42  * Modified by Steve Deering, Stanford, February 1989.
43  * Modified by Mark J. Steiglitz, Stanford, May, 1991
44  * Modified by Van Jacobson, LBL, January 1993
45  * Modified by Ajit Thyagarajan, PARC, August 1993
46  * Modified by Bill Fenner, PARC, April 1995
47  * Modified by Ahmed Helmy, SGI, June 1996
48  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
49  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
50  * Modified by Hitoshi Asaeda, WIDE, August 2000
51  * Modified by Pavlin Radoslavov, ICSI, October 2002
52  *
53  * MROUTING Revision: 3.5
54  * and PIM-SMv2 and PIM-DM support, advanced API support,
55  * bandwidth metering and signaling
56  */
57 
58 /*
59  * TODO: Prefix functions with ipmf_.
60  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
61  * domain attachment (if_afdata) so we can track consumers of that service.
62  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
63  * move it to socket options.
64  * TODO: Cleanup LSRR removal further.
65  * TODO: Push RSVP stubs into raw_ip.c.
66  * TODO: Use bitstring.h for vif set.
67  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
68  * TODO: Sync ip6_mroute.c with this file.
69  */
70 
71 #include <sys/cdefs.h>
72 __FBSDID("$FreeBSD$");
73 
74 #include "opt_inet.h"
75 #include "opt_mrouting.h"
76 
77 #define _PIM_VT 1
78 
79 #include <sys/param.h>
80 #include <sys/kernel.h>
81 #include <sys/stddef.h>
82 #include <sys/eventhandler.h>
83 #include <sys/lock.h>
84 #include <sys/ktr.h>
85 #include <sys/malloc.h>
86 #include <sys/mbuf.h>
87 #include <sys/module.h>
88 #include <sys/priv.h>
89 #include <sys/protosw.h>
90 #include <sys/signalvar.h>
91 #include <sys/socket.h>
92 #include <sys/socketvar.h>
93 #include <sys/sockio.h>
94 #include <sys/sx.h>
95 #include <sys/sysctl.h>
96 #include <sys/syslog.h>
97 #include <sys/systm.h>
98 #include <sys/time.h>
99 #include <sys/counter.h>
100 
101 #include <net/if.h>
102 #include <net/if_var.h>
103 #include <net/netisr.h>
104 #include <net/route.h>
105 #include <net/vnet.h>
106 
107 #include <netinet/in.h>
108 #include <netinet/igmp.h>
109 #include <netinet/in_systm.h>
110 #include <netinet/in_var.h>
111 #include <netinet/ip.h>
112 #include <netinet/ip_encap.h>
113 #include <netinet/ip_mroute.h>
114 #include <netinet/ip_var.h>
115 #include <netinet/ip_options.h>
116 #include <netinet/pim.h>
117 #include <netinet/pim_var.h>
118 #include <netinet/udp.h>
119 
120 #include <machine/in_cksum.h>
121 
122 #ifndef KTR_IPMF
123 #define KTR_IPMF KTR_INET
124 #endif
125 
126 #define		VIFI_INVALID	((vifi_t) -1)
127 
128 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
129 #define	V_last_tv_sec	VNET(last_tv_sec)
130 
131 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
132 
133 /*
134  * Locking.  We use two locks: one for the virtual interface table and
135  * one for the forwarding table.  These locks may be nested in which case
136  * the VIF lock must always be taken first.  Note that each lock is used
137  * to cover not only the specific data structure but also related data
138  * structures.
139  */
140 
141 static struct mtx mrouter_mtx;
142 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
143 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
144 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
145 #define	MROUTER_LOCK_INIT()						\
146 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
147 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
148 
149 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
150 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
151 
152 static VNET_PCPUSTAT_DEFINE(struct mrtstat, mrtstat);
153 VNET_PCPUSTAT_SYSINIT(mrtstat);
154 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
155 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
156     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
157     "netinet/ip_mroute.h)");
158 
159 static VNET_DEFINE(u_long, mfchash);
160 #define	V_mfchash		VNET(mfchash)
161 #define	MFCHASH(a, g)							\
162 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
163 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
164 #define	MFCHASHSIZE	256
165 
166 static u_long mfchashsize;			/* Hash size */
167 static VNET_DEFINE(u_char *, nexpire);		/* 0..mfchashsize-1 */
168 #define	V_nexpire		VNET(nexpire)
169 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
170 #define	V_mfchashtbl		VNET(mfchashtbl)
171 
172 static struct mtx mfc_mtx;
173 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
174 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
175 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
176 #define	MFC_LOCK_INIT()							\
177 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
178 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
179 
180 static VNET_DEFINE(vifi_t, numvifs);
181 #define	V_numvifs		VNET(numvifs)
182 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
183 #define	V_viftable		VNET(viftable)
184 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_VNET | CTLFLAG_RD,
185     &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
186     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
187 
188 static struct mtx vif_mtx;
189 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
190 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
191 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
192 #define	VIF_LOCK_INIT()							\
193 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
194 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
195 
196 static eventhandler_tag if_detach_event_tag = NULL;
197 
198 static VNET_DEFINE(struct callout, expire_upcalls_ch);
199 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
200 
201 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
202 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
203 
204 /*
205  * Bandwidth meter variables and constants
206  */
207 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
208 /*
209  * Pending timeouts are stored in a hash table, the key being the
210  * expiration time. Periodically, the entries are analysed and processed.
211  */
212 #define	BW_METER_BUCKETS	1024
213 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
214 #define	V_bw_meter_timers	VNET(bw_meter_timers)
215 static VNET_DEFINE(struct callout, bw_meter_ch);
216 #define	V_bw_meter_ch		VNET(bw_meter_ch)
217 #define	BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
218 
219 /*
220  * Pending upcalls are stored in a vector which is flushed when
221  * full, or periodically
222  */
223 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
224 #define	V_bw_upcalls		VNET(bw_upcalls)
225 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
226 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
227 static VNET_DEFINE(struct callout, bw_upcalls_ch);
228 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
229 
230 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
231 
232 static VNET_PCPUSTAT_DEFINE(struct pimstat, pimstat);
233 VNET_PCPUSTAT_SYSINIT(pimstat);
234 VNET_PCPUSTAT_SYSUNINIT(pimstat);
235 
236 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
237 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
238     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
239 
240 static u_long	pim_squelch_wholepkt = 0;
241 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
242     &pim_squelch_wholepkt, 0,
243     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
244 
245 extern  struct domain inetdomain;
246 static const struct protosw in_pim_protosw = {
247 	.pr_type =		SOCK_RAW,
248 	.pr_domain =		&inetdomain,
249 	.pr_protocol =		IPPROTO_PIM,
250 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
251 	.pr_input =		pim_input,
252 	.pr_output =		rip_output,
253 	.pr_ctloutput =		rip_ctloutput,
254 	.pr_usrreqs =		&rip_usrreqs
255 };
256 static const struct encaptab *pim_encap_cookie;
257 
258 static int pim_encapcheck(const struct mbuf *, int, int, void *);
259 
260 /*
261  * Note: the PIM Register encapsulation adds the following in front of a
262  * data packet:
263  *
264  * struct pim_encap_hdr {
265  *    struct ip ip;
266  *    struct pim_encap_pimhdr  pim;
267  * }
268  *
269  */
270 
271 struct pim_encap_pimhdr {
272 	struct pim pim;
273 	uint32_t   flags;
274 };
275 #define		PIM_ENCAP_TTL	64
276 
277 static struct ip pim_encap_iphdr = {
278 #if BYTE_ORDER == LITTLE_ENDIAN
279 	sizeof(struct ip) >> 2,
280 	IPVERSION,
281 #else
282 	IPVERSION,
283 	sizeof(struct ip) >> 2,
284 #endif
285 	0,			/* tos */
286 	sizeof(struct ip),	/* total length */
287 	0,			/* id */
288 	0,			/* frag offset */
289 	PIM_ENCAP_TTL,
290 	IPPROTO_PIM,
291 	0,			/* checksum */
292 };
293 
294 static struct pim_encap_pimhdr pim_encap_pimhdr = {
295     {
296 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
297 	0,			/* reserved */
298 	0,			/* checksum */
299     },
300     0				/* flags */
301 };
302 
303 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
304 #define	V_reg_vif_num		VNET(reg_vif_num)
305 static VNET_DEFINE(struct ifnet, multicast_register_if);
306 #define	V_multicast_register_if	VNET(multicast_register_if)
307 
308 /*
309  * Private variables.
310  */
311 
312 static u_long	X_ip_mcast_src(int);
313 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
314 		    struct ip_moptions *);
315 static int	X_ip_mrouter_done(void);
316 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
317 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
318 static int	X_legal_vif_num(int);
319 static int	X_mrt_ioctl(u_long, caddr_t, int);
320 
321 static int	add_bw_upcall(struct bw_upcall *);
322 static int	add_mfc(struct mfcctl2 *);
323 static int	add_vif(struct vifctl *);
324 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
325 static void	bw_meter_process(void);
326 static void	bw_meter_receive_packet(struct bw_meter *, int,
327 		    struct timeval *);
328 static void	bw_upcalls_send(void);
329 static int	del_bw_upcall(struct bw_upcall *);
330 static int	del_mfc(struct mfcctl2 *);
331 static int	del_vif(vifi_t);
332 static int	del_vif_locked(vifi_t);
333 static void	expire_bw_meter_process(void *);
334 static void	expire_bw_upcalls_send(void *);
335 static void	expire_mfc(struct mfc *);
336 static void	expire_upcalls(void *);
337 static void	free_bw_list(struct bw_meter *);
338 static int	get_sg_cnt(struct sioc_sg_req *);
339 static int	get_vif_cnt(struct sioc_vif_req *);
340 static void	if_detached_event(void *, struct ifnet *);
341 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
342 static int	ip_mrouter_init(struct socket *, int);
343 static __inline struct mfc *
344 		mfc_find(struct in_addr *, struct in_addr *);
345 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
346 static struct mbuf *
347 		pim_register_prepare(struct ip *, struct mbuf *);
348 static int	pim_register_send(struct ip *, struct vif *,
349 		    struct mbuf *, struct mfc *);
350 static int	pim_register_send_rp(struct ip *, struct vif *,
351 		    struct mbuf *, struct mfc *);
352 static int	pim_register_send_upcall(struct ip *, struct vif *,
353 		    struct mbuf *, struct mfc *);
354 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
355 static void	send_packet(struct vif *, struct mbuf *);
356 static int	set_api_config(uint32_t *);
357 static int	set_assert(int);
358 static int	socket_send(struct socket *, struct mbuf *,
359 		    struct sockaddr_in *);
360 static void	unschedule_bw_meter(struct bw_meter *);
361 
362 /*
363  * Kernel multicast forwarding API capabilities and setup.
364  * If more API capabilities are added to the kernel, they should be
365  * recorded in `mrt_api_support'.
366  */
367 #define MRT_API_VERSION		0x0305
368 
369 static const int mrt_api_version = MRT_API_VERSION;
370 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
371 					 MRT_MFC_FLAGS_BORDER_VIF |
372 					 MRT_MFC_RP |
373 					 MRT_MFC_BW_UPCALL);
374 static VNET_DEFINE(uint32_t, mrt_api_config);
375 #define	V_mrt_api_config	VNET(mrt_api_config)
376 static VNET_DEFINE(int, pim_assert_enabled);
377 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
378 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
379 
380 /*
381  * Find a route for a given origin IP address and multicast group address.
382  * Statistics must be updated by the caller.
383  */
384 static __inline struct mfc *
385 mfc_find(struct in_addr *o, struct in_addr *g)
386 {
387 	struct mfc *rt;
388 
389 	MFC_LOCK_ASSERT();
390 
391 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
392 		if (in_hosteq(rt->mfc_origin, *o) &&
393 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
394 		    TAILQ_EMPTY(&rt->mfc_stall))
395 			break;
396 	}
397 
398 	return (rt);
399 }
400 
401 /*
402  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
403  */
404 static int
405 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
406 {
407     int	error, optval;
408     vifi_t	vifi;
409     struct	vifctl vifc;
410     struct	mfcctl2 mfc;
411     struct	bw_upcall bw_upcall;
412     uint32_t	i;
413 
414     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
415 	return EPERM;
416 
417     error = 0;
418     switch (sopt->sopt_name) {
419     case MRT_INIT:
420 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
421 	if (error)
422 	    break;
423 	error = ip_mrouter_init(so, optval);
424 	break;
425 
426     case MRT_DONE:
427 	error = ip_mrouter_done();
428 	break;
429 
430     case MRT_ADD_VIF:
431 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
432 	if (error)
433 	    break;
434 	error = add_vif(&vifc);
435 	break;
436 
437     case MRT_DEL_VIF:
438 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
439 	if (error)
440 	    break;
441 	error = del_vif(vifi);
442 	break;
443 
444     case MRT_ADD_MFC:
445     case MRT_DEL_MFC:
446 	/*
447 	 * select data size depending on API version.
448 	 */
449 	if (sopt->sopt_name == MRT_ADD_MFC &&
450 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
451 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
452 				sizeof(struct mfcctl2));
453 	} else {
454 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
455 				sizeof(struct mfcctl));
456 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
457 			sizeof(mfc) - sizeof(struct mfcctl));
458 	}
459 	if (error)
460 	    break;
461 	if (sopt->sopt_name == MRT_ADD_MFC)
462 	    error = add_mfc(&mfc);
463 	else
464 	    error = del_mfc(&mfc);
465 	break;
466 
467     case MRT_ASSERT:
468 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
469 	if (error)
470 	    break;
471 	set_assert(optval);
472 	break;
473 
474     case MRT_API_CONFIG:
475 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
476 	if (!error)
477 	    error = set_api_config(&i);
478 	if (!error)
479 	    error = sooptcopyout(sopt, &i, sizeof i);
480 	break;
481 
482     case MRT_ADD_BW_UPCALL:
483     case MRT_DEL_BW_UPCALL:
484 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
485 				sizeof bw_upcall);
486 	if (error)
487 	    break;
488 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
489 	    error = add_bw_upcall(&bw_upcall);
490 	else
491 	    error = del_bw_upcall(&bw_upcall);
492 	break;
493 
494     default:
495 	error = EOPNOTSUPP;
496 	break;
497     }
498     return error;
499 }
500 
501 /*
502  * Handle MRT getsockopt commands
503  */
504 static int
505 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
506 {
507     int error;
508 
509     switch (sopt->sopt_name) {
510     case MRT_VERSION:
511 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
512 	break;
513 
514     case MRT_ASSERT:
515 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
516 	    sizeof V_pim_assert_enabled);
517 	break;
518 
519     case MRT_API_SUPPORT:
520 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
521 	break;
522 
523     case MRT_API_CONFIG:
524 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
525 	break;
526 
527     default:
528 	error = EOPNOTSUPP;
529 	break;
530     }
531     return error;
532 }
533 
534 /*
535  * Handle ioctl commands to obtain information from the cache
536  */
537 static int
538 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
539 {
540     int error = 0;
541 
542     /*
543      * Currently the only function calling this ioctl routine is rtioctl_fib().
544      * Typically, only root can create the raw socket in order to execute
545      * this ioctl method, however the request might be coming from a prison
546      */
547     error = priv_check(curthread, PRIV_NETINET_MROUTE);
548     if (error)
549 	return (error);
550     switch (cmd) {
551     case (SIOCGETVIFCNT):
552 	error = get_vif_cnt((struct sioc_vif_req *)data);
553 	break;
554 
555     case (SIOCGETSGCNT):
556 	error = get_sg_cnt((struct sioc_sg_req *)data);
557 	break;
558 
559     default:
560 	error = EINVAL;
561 	break;
562     }
563     return error;
564 }
565 
566 /*
567  * returns the packet, byte, rpf-failure count for the source group provided
568  */
569 static int
570 get_sg_cnt(struct sioc_sg_req *req)
571 {
572     struct mfc *rt;
573 
574     MFC_LOCK();
575     rt = mfc_find(&req->src, &req->grp);
576     if (rt == NULL) {
577 	MFC_UNLOCK();
578 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
579 	return EADDRNOTAVAIL;
580     }
581     req->pktcnt = rt->mfc_pkt_cnt;
582     req->bytecnt = rt->mfc_byte_cnt;
583     req->wrong_if = rt->mfc_wrong_if;
584     MFC_UNLOCK();
585     return 0;
586 }
587 
588 /*
589  * returns the input and output packet and byte counts on the vif provided
590  */
591 static int
592 get_vif_cnt(struct sioc_vif_req *req)
593 {
594     vifi_t vifi = req->vifi;
595 
596     VIF_LOCK();
597     if (vifi >= V_numvifs) {
598 	VIF_UNLOCK();
599 	return EINVAL;
600     }
601 
602     req->icount = V_viftable[vifi].v_pkt_in;
603     req->ocount = V_viftable[vifi].v_pkt_out;
604     req->ibytes = V_viftable[vifi].v_bytes_in;
605     req->obytes = V_viftable[vifi].v_bytes_out;
606     VIF_UNLOCK();
607 
608     return 0;
609 }
610 
611 static void
612 if_detached_event(void *arg __unused, struct ifnet *ifp)
613 {
614     vifi_t vifi;
615     u_long i;
616 
617     MROUTER_LOCK();
618 
619     if (V_ip_mrouter == NULL) {
620 	MROUTER_UNLOCK();
621 	return;
622     }
623 
624     VIF_LOCK();
625     MFC_LOCK();
626 
627     /*
628      * Tear down multicast forwarder state associated with this ifnet.
629      * 1. Walk the vif list, matching vifs against this ifnet.
630      * 2. Walk the multicast forwarding cache (mfc) looking for
631      *    inner matches with this vif's index.
632      * 3. Expire any matching multicast forwarding cache entries.
633      * 4. Free vif state. This should disable ALLMULTI on the interface.
634      */
635     for (vifi = 0; vifi < V_numvifs; vifi++) {
636 	if (V_viftable[vifi].v_ifp != ifp)
637 		continue;
638 	for (i = 0; i < mfchashsize; i++) {
639 		struct mfc *rt, *nrt;
640 
641 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
642 			if (rt->mfc_parent == vifi) {
643 				expire_mfc(rt);
644 			}
645 		}
646 	}
647 	del_vif_locked(vifi);
648     }
649 
650     MFC_UNLOCK();
651     VIF_UNLOCK();
652 
653     MROUTER_UNLOCK();
654 }
655 
656 /*
657  * Enable multicast forwarding.
658  */
659 static int
660 ip_mrouter_init(struct socket *so, int version)
661 {
662 
663     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
664         so->so_type, so->so_proto->pr_protocol);
665 
666     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
667 	return EOPNOTSUPP;
668 
669     if (version != 1)
670 	return ENOPROTOOPT;
671 
672     MROUTER_LOCK();
673 
674     if (ip_mrouter_unloading) {
675 	MROUTER_UNLOCK();
676 	return ENOPROTOOPT;
677     }
678 
679     if (V_ip_mrouter != NULL) {
680 	MROUTER_UNLOCK();
681 	return EADDRINUSE;
682     }
683 
684     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
685 	HASH_NOWAIT);
686 
687     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
688 	curvnet);
689     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
690 	curvnet);
691     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
692 	curvnet);
693 
694     V_ip_mrouter = so;
695     ip_mrouter_cnt++;
696 
697     MROUTER_UNLOCK();
698 
699     CTR1(KTR_IPMF, "%s: done", __func__);
700 
701     return 0;
702 }
703 
704 /*
705  * Disable multicast forwarding.
706  */
707 static int
708 X_ip_mrouter_done(void)
709 {
710     struct ifnet *ifp;
711     u_long i;
712     vifi_t vifi;
713 
714     MROUTER_LOCK();
715 
716     if (V_ip_mrouter == NULL) {
717 	MROUTER_UNLOCK();
718 	return EINVAL;
719     }
720 
721     /*
722      * Detach/disable hooks to the reset of the system.
723      */
724     V_ip_mrouter = NULL;
725     ip_mrouter_cnt--;
726     V_mrt_api_config = 0;
727 
728     VIF_LOCK();
729 
730     /*
731      * For each phyint in use, disable promiscuous reception of all IP
732      * multicasts.
733      */
734     for (vifi = 0; vifi < V_numvifs; vifi++) {
735 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
736 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
737 	    ifp = V_viftable[vifi].v_ifp;
738 	    if_allmulti(ifp, 0);
739 	}
740     }
741     bzero((caddr_t)V_viftable, sizeof(V_viftable));
742     V_numvifs = 0;
743     V_pim_assert_enabled = 0;
744 
745     VIF_UNLOCK();
746 
747     callout_stop(&V_expire_upcalls_ch);
748     callout_stop(&V_bw_upcalls_ch);
749     callout_stop(&V_bw_meter_ch);
750 
751     MFC_LOCK();
752 
753     /*
754      * Free all multicast forwarding cache entries.
755      * Do not use hashdestroy(), as we must perform other cleanup.
756      */
757     for (i = 0; i < mfchashsize; i++) {
758 	struct mfc *rt, *nrt;
759 
760 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
761 		expire_mfc(rt);
762 	}
763     }
764     free(V_mfchashtbl, M_MRTABLE);
765     V_mfchashtbl = NULL;
766 
767     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
768 
769     V_bw_upcalls_n = 0;
770     bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
771 
772     MFC_UNLOCK();
773 
774     V_reg_vif_num = VIFI_INVALID;
775 
776     MROUTER_UNLOCK();
777 
778     CTR1(KTR_IPMF, "%s: done", __func__);
779 
780     return 0;
781 }
782 
783 /*
784  * Set PIM assert processing global
785  */
786 static int
787 set_assert(int i)
788 {
789     if ((i != 1) && (i != 0))
790 	return EINVAL;
791 
792     V_pim_assert_enabled = i;
793 
794     return 0;
795 }
796 
797 /*
798  * Configure API capabilities
799  */
800 int
801 set_api_config(uint32_t *apival)
802 {
803     u_long i;
804 
805     /*
806      * We can set the API capabilities only if it is the first operation
807      * after MRT_INIT. I.e.:
808      *  - there are no vifs installed
809      *  - pim_assert is not enabled
810      *  - the MFC table is empty
811      */
812     if (V_numvifs > 0) {
813 	*apival = 0;
814 	return EPERM;
815     }
816     if (V_pim_assert_enabled) {
817 	*apival = 0;
818 	return EPERM;
819     }
820 
821     MFC_LOCK();
822 
823     for (i = 0; i < mfchashsize; i++) {
824 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
825 	    MFC_UNLOCK();
826 	    *apival = 0;
827 	    return EPERM;
828 	}
829     }
830 
831     MFC_UNLOCK();
832 
833     V_mrt_api_config = *apival & mrt_api_support;
834     *apival = V_mrt_api_config;
835 
836     return 0;
837 }
838 
839 /*
840  * Add a vif to the vif table
841  */
842 static int
843 add_vif(struct vifctl *vifcp)
844 {
845     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
846     struct sockaddr_in sin = {sizeof sin, AF_INET};
847     struct ifaddr *ifa;
848     struct ifnet *ifp;
849     int error;
850 
851     VIF_LOCK();
852     if (vifcp->vifc_vifi >= MAXVIFS) {
853 	VIF_UNLOCK();
854 	return EINVAL;
855     }
856     /* rate limiting is no longer supported by this code */
857     if (vifcp->vifc_rate_limit != 0) {
858 	log(LOG_ERR, "rate limiting is no longer supported\n");
859 	VIF_UNLOCK();
860 	return EINVAL;
861     }
862     if (!in_nullhost(vifp->v_lcl_addr)) {
863 	VIF_UNLOCK();
864 	return EADDRINUSE;
865     }
866     if (in_nullhost(vifcp->vifc_lcl_addr)) {
867 	VIF_UNLOCK();
868 	return EADDRNOTAVAIL;
869     }
870 
871     /* Find the interface with an address in AF_INET family */
872     if (vifcp->vifc_flags & VIFF_REGISTER) {
873 	/*
874 	 * XXX: Because VIFF_REGISTER does not really need a valid
875 	 * local interface (e.g. it could be 127.0.0.2), we don't
876 	 * check its address.
877 	 */
878 	ifp = NULL;
879     } else {
880 	sin.sin_addr = vifcp->vifc_lcl_addr;
881 	NET_EPOCH_ENTER();
882 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
883 	if (ifa == NULL) {
884 		NET_EPOCH_EXIT();
885 	    VIF_UNLOCK();
886 	    return EADDRNOTAVAIL;
887 	}
888 	ifp = ifa->ifa_ifp;
889 	NET_EPOCH_EXIT();
890     }
891 
892     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
893 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
894 	VIF_UNLOCK();
895 	return EOPNOTSUPP;
896     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
897 	ifp = &V_multicast_register_if;
898 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
899 	if (V_reg_vif_num == VIFI_INVALID) {
900 	    if_initname(&V_multicast_register_if, "register_vif", 0);
901 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
902 	    V_reg_vif_num = vifcp->vifc_vifi;
903 	}
904     } else {		/* Make sure the interface supports multicast */
905 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
906 	    VIF_UNLOCK();
907 	    return EOPNOTSUPP;
908 	}
909 
910 	/* Enable promiscuous reception of all IP multicasts from the if */
911 	error = if_allmulti(ifp, 1);
912 	if (error) {
913 	    VIF_UNLOCK();
914 	    return error;
915 	}
916     }
917 
918     vifp->v_flags     = vifcp->vifc_flags;
919     vifp->v_threshold = vifcp->vifc_threshold;
920     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
921     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
922     vifp->v_ifp       = ifp;
923     /* initialize per vif pkt counters */
924     vifp->v_pkt_in    = 0;
925     vifp->v_pkt_out   = 0;
926     vifp->v_bytes_in  = 0;
927     vifp->v_bytes_out = 0;
928 
929     /* Adjust numvifs up if the vifi is higher than numvifs */
930     if (V_numvifs <= vifcp->vifc_vifi)
931 	V_numvifs = vifcp->vifc_vifi + 1;
932 
933     VIF_UNLOCK();
934 
935     CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
936 	(int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
937 	(int)vifcp->vifc_threshold);
938 
939     return 0;
940 }
941 
942 /*
943  * Delete a vif from the vif table
944  */
945 static int
946 del_vif_locked(vifi_t vifi)
947 {
948     struct vif *vifp;
949 
950     VIF_LOCK_ASSERT();
951 
952     if (vifi >= V_numvifs) {
953 	return EINVAL;
954     }
955     vifp = &V_viftable[vifi];
956     if (in_nullhost(vifp->v_lcl_addr)) {
957 	return EADDRNOTAVAIL;
958     }
959 
960     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
961 	if_allmulti(vifp->v_ifp, 0);
962 
963     if (vifp->v_flags & VIFF_REGISTER)
964 	V_reg_vif_num = VIFI_INVALID;
965 
966     bzero((caddr_t)vifp, sizeof (*vifp));
967 
968     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
969 
970     /* Adjust numvifs down */
971     for (vifi = V_numvifs; vifi > 0; vifi--)
972 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
973 	    break;
974     V_numvifs = vifi;
975 
976     return 0;
977 }
978 
979 static int
980 del_vif(vifi_t vifi)
981 {
982     int cc;
983 
984     VIF_LOCK();
985     cc = del_vif_locked(vifi);
986     VIF_UNLOCK();
987 
988     return cc;
989 }
990 
991 /*
992  * update an mfc entry without resetting counters and S,G addresses.
993  */
994 static void
995 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
996 {
997     int i;
998 
999     rt->mfc_parent = mfccp->mfcc_parent;
1000     for (i = 0; i < V_numvifs; i++) {
1001 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1002 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1003 	    MRT_MFC_FLAGS_ALL;
1004     }
1005     /* set the RP address */
1006     if (V_mrt_api_config & MRT_MFC_RP)
1007 	rt->mfc_rp = mfccp->mfcc_rp;
1008     else
1009 	rt->mfc_rp.s_addr = INADDR_ANY;
1010 }
1011 
1012 /*
1013  * fully initialize an mfc entry from the parameter.
1014  */
1015 static void
1016 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1017 {
1018     rt->mfc_origin     = mfccp->mfcc_origin;
1019     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1020 
1021     update_mfc_params(rt, mfccp);
1022 
1023     /* initialize pkt counters per src-grp */
1024     rt->mfc_pkt_cnt    = 0;
1025     rt->mfc_byte_cnt   = 0;
1026     rt->mfc_wrong_if   = 0;
1027     timevalclear(&rt->mfc_last_assert);
1028 }
1029 
1030 static void
1031 expire_mfc(struct mfc *rt)
1032 {
1033 	struct rtdetq *rte, *nrte;
1034 
1035 	MFC_LOCK_ASSERT();
1036 
1037 	free_bw_list(rt->mfc_bw_meter);
1038 
1039 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1040 		m_freem(rte->m);
1041 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1042 		free(rte, M_MRTABLE);
1043 	}
1044 
1045 	LIST_REMOVE(rt, mfc_hash);
1046 	free(rt, M_MRTABLE);
1047 }
1048 
1049 /*
1050  * Add an mfc entry
1051  */
1052 static int
1053 add_mfc(struct mfcctl2 *mfccp)
1054 {
1055     struct mfc *rt;
1056     struct rtdetq *rte, *nrte;
1057     u_long hash = 0;
1058     u_short nstl;
1059 
1060     VIF_LOCK();
1061     MFC_LOCK();
1062 
1063     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1064 
1065     /* If an entry already exists, just update the fields */
1066     if (rt) {
1067 	CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1068 	    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1069 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1070 	    mfccp->mfcc_parent);
1071 	update_mfc_params(rt, mfccp);
1072 	MFC_UNLOCK();
1073 	VIF_UNLOCK();
1074 	return (0);
1075     }
1076 
1077     /*
1078      * Find the entry for which the upcall was made and update
1079      */
1080     nstl = 0;
1081     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1082     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1083 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1084 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1085 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1086 		CTR5(KTR_IPMF,
1087 		    "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1088 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1089 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1090 		    mfccp->mfcc_parent,
1091 		    TAILQ_FIRST(&rt->mfc_stall));
1092 		if (nstl++)
1093 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1094 
1095 		init_mfc_params(rt, mfccp);
1096 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1097 		V_nexpire[hash]--;
1098 
1099 		/* Free queued packets, but attempt to forward them first. */
1100 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1101 			if (rte->ifp != NULL)
1102 				ip_mdq(rte->m, rte->ifp, rt, -1);
1103 			m_freem(rte->m);
1104 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1105 			rt->mfc_nstall--;
1106 			free(rte, M_MRTABLE);
1107 		}
1108 	}
1109     }
1110 
1111     /*
1112      * It is possible that an entry is being inserted without an upcall
1113      */
1114     if (nstl == 0) {
1115 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1116 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1117 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1118 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1119 			init_mfc_params(rt, mfccp);
1120 			if (rt->mfc_expire)
1121 			    V_nexpire[hash]--;
1122 			rt->mfc_expire = 0;
1123 			break; /* XXX */
1124 		}
1125 	}
1126 
1127 	if (rt == NULL) {		/* no upcall, so make a new entry */
1128 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1129 	    if (rt == NULL) {
1130 		MFC_UNLOCK();
1131 		VIF_UNLOCK();
1132 		return (ENOBUFS);
1133 	    }
1134 
1135 	    init_mfc_params(rt, mfccp);
1136 	    TAILQ_INIT(&rt->mfc_stall);
1137 	    rt->mfc_nstall = 0;
1138 
1139 	    rt->mfc_expire     = 0;
1140 	    rt->mfc_bw_meter = NULL;
1141 
1142 	    /* insert new entry at head of hash chain */
1143 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1144 	}
1145     }
1146 
1147     MFC_UNLOCK();
1148     VIF_UNLOCK();
1149 
1150     return (0);
1151 }
1152 
1153 /*
1154  * Delete an mfc entry
1155  */
1156 static int
1157 del_mfc(struct mfcctl2 *mfccp)
1158 {
1159     struct in_addr	origin;
1160     struct in_addr	mcastgrp;
1161     struct mfc		*rt;
1162 
1163     origin = mfccp->mfcc_origin;
1164     mcastgrp = mfccp->mfcc_mcastgrp;
1165 
1166     CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1167 	ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1168 
1169     MFC_LOCK();
1170 
1171     rt = mfc_find(&origin, &mcastgrp);
1172     if (rt == NULL) {
1173 	MFC_UNLOCK();
1174 	return EADDRNOTAVAIL;
1175     }
1176 
1177     /*
1178      * free the bw_meter entries
1179      */
1180     free_bw_list(rt->mfc_bw_meter);
1181     rt->mfc_bw_meter = NULL;
1182 
1183     LIST_REMOVE(rt, mfc_hash);
1184     free(rt, M_MRTABLE);
1185 
1186     MFC_UNLOCK();
1187 
1188     return (0);
1189 }
1190 
1191 /*
1192  * Send a message to the routing daemon on the multicast routing socket.
1193  */
1194 static int
1195 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1196 {
1197     if (s) {
1198 	SOCKBUF_LOCK(&s->so_rcv);
1199 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1200 	    NULL) != 0) {
1201 	    sorwakeup_locked(s);
1202 	    return 0;
1203 	}
1204 	SOCKBUF_UNLOCK(&s->so_rcv);
1205     }
1206     m_freem(mm);
1207     return -1;
1208 }
1209 
1210 /*
1211  * IP multicast forwarding function. This function assumes that the packet
1212  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1213  * pointed to by "ifp", and the packet is to be relayed to other networks
1214  * that have members of the packet's destination IP multicast group.
1215  *
1216  * The packet is returned unscathed to the caller, unless it is
1217  * erroneous, in which case a non-zero return value tells the caller to
1218  * discard it.
1219  */
1220 
1221 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1222 
1223 static int
1224 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1225     struct ip_moptions *imo)
1226 {
1227     struct mfc *rt;
1228     int error;
1229     vifi_t vifi;
1230 
1231     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1232 	ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1233 
1234     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1235 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1236 	/*
1237 	 * Packet arrived via a physical interface or
1238 	 * an encapsulated tunnel or a register_vif.
1239 	 */
1240     } else {
1241 	/*
1242 	 * Packet arrived through a source-route tunnel.
1243 	 * Source-route tunnels are no longer supported.
1244 	 */
1245 	return (1);
1246     }
1247 
1248     VIF_LOCK();
1249     MFC_LOCK();
1250     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1251 	if (ip->ip_ttl < MAXTTL)
1252 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1253 	error = ip_mdq(m, ifp, NULL, vifi);
1254 	MFC_UNLOCK();
1255 	VIF_UNLOCK();
1256 	return error;
1257     }
1258 
1259     /*
1260      * Don't forward a packet with time-to-live of zero or one,
1261      * or a packet destined to a local-only group.
1262      */
1263     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1264 	MFC_UNLOCK();
1265 	VIF_UNLOCK();
1266 	return 0;
1267     }
1268 
1269     /*
1270      * Determine forwarding vifs from the forwarding cache table
1271      */
1272     MRTSTAT_INC(mrts_mfc_lookups);
1273     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1274 
1275     /* Entry exists, so forward if necessary */
1276     if (rt != NULL) {
1277 	error = ip_mdq(m, ifp, rt, -1);
1278 	MFC_UNLOCK();
1279 	VIF_UNLOCK();
1280 	return error;
1281     } else {
1282 	/*
1283 	 * If we don't have a route for packet's origin,
1284 	 * Make a copy of the packet & send message to routing daemon
1285 	 */
1286 
1287 	struct mbuf *mb0;
1288 	struct rtdetq *rte;
1289 	u_long hash;
1290 	int hlen = ip->ip_hl << 2;
1291 
1292 	MRTSTAT_INC(mrts_mfc_misses);
1293 	MRTSTAT_INC(mrts_no_route);
1294 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1295 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1296 
1297 	/*
1298 	 * Allocate mbufs early so that we don't do extra work if we are
1299 	 * just going to fail anyway.  Make sure to pullup the header so
1300 	 * that other people can't step on it.
1301 	 */
1302 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1303 	    M_NOWAIT|M_ZERO);
1304 	if (rte == NULL) {
1305 	    MFC_UNLOCK();
1306 	    VIF_UNLOCK();
1307 	    return ENOBUFS;
1308 	}
1309 
1310 	mb0 = m_copypacket(m, M_NOWAIT);
1311 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1312 	    mb0 = m_pullup(mb0, hlen);
1313 	if (mb0 == NULL) {
1314 	    free(rte, M_MRTABLE);
1315 	    MFC_UNLOCK();
1316 	    VIF_UNLOCK();
1317 	    return ENOBUFS;
1318 	}
1319 
1320 	/* is there an upcall waiting for this flow ? */
1321 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1322 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1323 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1324 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1325 		    !TAILQ_EMPTY(&rt->mfc_stall))
1326 			break;
1327 	}
1328 
1329 	if (rt == NULL) {
1330 	    int i;
1331 	    struct igmpmsg *im;
1332 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1333 	    struct mbuf *mm;
1334 
1335 	    /*
1336 	     * Locate the vifi for the incoming interface for this packet.
1337 	     * If none found, drop packet.
1338 	     */
1339 	    for (vifi = 0; vifi < V_numvifs &&
1340 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1341 		;
1342 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1343 		goto non_fatal;
1344 
1345 	    /* no upcall, so make a new entry */
1346 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1347 	    if (rt == NULL)
1348 		goto fail;
1349 
1350 	    /* Make a copy of the header to send to the user level process */
1351 	    mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1352 	    if (mm == NULL)
1353 		goto fail1;
1354 
1355 	    /*
1356 	     * Send message to routing daemon to install
1357 	     * a route into the kernel table
1358 	     */
1359 
1360 	    im = mtod(mm, struct igmpmsg *);
1361 	    im->im_msgtype = IGMPMSG_NOCACHE;
1362 	    im->im_mbz = 0;
1363 	    im->im_vif = vifi;
1364 
1365 	    MRTSTAT_INC(mrts_upcalls);
1366 
1367 	    k_igmpsrc.sin_addr = ip->ip_src;
1368 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1369 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1370 		MRTSTAT_INC(mrts_upq_sockfull);
1371 fail1:
1372 		free(rt, M_MRTABLE);
1373 fail:
1374 		free(rte, M_MRTABLE);
1375 		m_freem(mb0);
1376 		MFC_UNLOCK();
1377 		VIF_UNLOCK();
1378 		return ENOBUFS;
1379 	    }
1380 
1381 	    /* insert new entry at head of hash chain */
1382 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1383 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1384 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1385 	    V_nexpire[hash]++;
1386 	    for (i = 0; i < V_numvifs; i++) {
1387 		rt->mfc_ttls[i] = 0;
1388 		rt->mfc_flags[i] = 0;
1389 	    }
1390 	    rt->mfc_parent = -1;
1391 
1392 	    /* clear the RP address */
1393 	    rt->mfc_rp.s_addr = INADDR_ANY;
1394 	    rt->mfc_bw_meter = NULL;
1395 
1396 	    /* initialize pkt counters per src-grp */
1397 	    rt->mfc_pkt_cnt = 0;
1398 	    rt->mfc_byte_cnt = 0;
1399 	    rt->mfc_wrong_if = 0;
1400 	    timevalclear(&rt->mfc_last_assert);
1401 
1402 	    TAILQ_INIT(&rt->mfc_stall);
1403 	    rt->mfc_nstall = 0;
1404 
1405 	    /* link into table */
1406 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1407 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1408 	    rt->mfc_nstall++;
1409 
1410 	} else {
1411 	    /* determine if queue has overflowed */
1412 	    if (rt->mfc_nstall > MAX_UPQ) {
1413 		MRTSTAT_INC(mrts_upq_ovflw);
1414 non_fatal:
1415 		free(rte, M_MRTABLE);
1416 		m_freem(mb0);
1417 		MFC_UNLOCK();
1418 		VIF_UNLOCK();
1419 		return (0);
1420 	    }
1421 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1422 	    rt->mfc_nstall++;
1423 	}
1424 
1425 	rte->m			= mb0;
1426 	rte->ifp		= ifp;
1427 
1428 	MFC_UNLOCK();
1429 	VIF_UNLOCK();
1430 
1431 	return 0;
1432     }
1433 }
1434 
1435 /*
1436  * Clean up the cache entry if upcall is not serviced
1437  */
1438 static void
1439 expire_upcalls(void *arg)
1440 {
1441     u_long i;
1442 
1443     CURVNET_SET((struct vnet *) arg);
1444 
1445     MFC_LOCK();
1446 
1447     for (i = 0; i < mfchashsize; i++) {
1448 	struct mfc *rt, *nrt;
1449 
1450 	if (V_nexpire[i] == 0)
1451 	    continue;
1452 
1453 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1454 		if (TAILQ_EMPTY(&rt->mfc_stall))
1455 			continue;
1456 
1457 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1458 			continue;
1459 
1460 		/*
1461 		 * free the bw_meter entries
1462 		 */
1463 		while (rt->mfc_bw_meter != NULL) {
1464 		    struct bw_meter *x = rt->mfc_bw_meter;
1465 
1466 		    rt->mfc_bw_meter = x->bm_mfc_next;
1467 		    free(x, M_BWMETER);
1468 		}
1469 
1470 		MRTSTAT_INC(mrts_cache_cleanups);
1471 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1472 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1473 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1474 
1475 		expire_mfc(rt);
1476 	    }
1477     }
1478 
1479     MFC_UNLOCK();
1480 
1481     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1482 	curvnet);
1483 
1484     CURVNET_RESTORE();
1485 }
1486 
1487 /*
1488  * Packet forwarding routine once entry in the cache is made
1489  */
1490 static int
1491 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1492 {
1493     struct ip  *ip = mtod(m, struct ip *);
1494     vifi_t vifi;
1495     int plen = ntohs(ip->ip_len);
1496 
1497     VIF_LOCK_ASSERT();
1498 
1499     /*
1500      * If xmt_vif is not -1, send on only the requested vif.
1501      *
1502      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1503      */
1504     if (xmt_vif < V_numvifs) {
1505 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1506 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1507 	else
1508 		phyint_send(ip, V_viftable + xmt_vif, m);
1509 	return 1;
1510     }
1511 
1512     /*
1513      * Don't forward if it didn't arrive from the parent vif for its origin.
1514      */
1515     vifi = rt->mfc_parent;
1516     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1517 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1518 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1519 	MRTSTAT_INC(mrts_wrong_if);
1520 	++rt->mfc_wrong_if;
1521 	/*
1522 	 * If we are doing PIM assert processing, send a message
1523 	 * to the routing daemon.
1524 	 *
1525 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1526 	 * can complete the SPT switch, regardless of the type
1527 	 * of the iif (broadcast media, GRE tunnel, etc).
1528 	 */
1529 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1530 	    V_viftable[vifi].v_ifp) {
1531 
1532 	    if (ifp == &V_multicast_register_if)
1533 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1534 
1535 	    /* Get vifi for the incoming packet */
1536 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1537 		vifi++)
1538 		;
1539 	    if (vifi >= V_numvifs)
1540 		return 0;	/* The iif is not found: ignore the packet. */
1541 
1542 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1543 		return 0;	/* WRONGVIF disabled: ignore the packet */
1544 
1545 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1546 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1547 		struct igmpmsg *im;
1548 		int hlen = ip->ip_hl << 2;
1549 		struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1550 
1551 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1552 		    mm = m_pullup(mm, hlen);
1553 		if (mm == NULL)
1554 		    return ENOBUFS;
1555 
1556 		im = mtod(mm, struct igmpmsg *);
1557 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1558 		im->im_mbz		= 0;
1559 		im->im_vif		= vifi;
1560 
1561 		MRTSTAT_INC(mrts_upcalls);
1562 
1563 		k_igmpsrc.sin_addr = im->im_src;
1564 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1565 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1566 		    MRTSTAT_INC(mrts_upq_sockfull);
1567 		    return ENOBUFS;
1568 		}
1569 	    }
1570 	}
1571 	return 0;
1572     }
1573 
1574 
1575     /* If I sourced this packet, it counts as output, else it was input. */
1576     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1577 	V_viftable[vifi].v_pkt_out++;
1578 	V_viftable[vifi].v_bytes_out += plen;
1579     } else {
1580 	V_viftable[vifi].v_pkt_in++;
1581 	V_viftable[vifi].v_bytes_in += plen;
1582     }
1583     rt->mfc_pkt_cnt++;
1584     rt->mfc_byte_cnt += plen;
1585 
1586     /*
1587      * For each vif, decide if a copy of the packet should be forwarded.
1588      * Forward if:
1589      *		- the ttl exceeds the vif's threshold
1590      *		- there are group members downstream on interface
1591      */
1592     for (vifi = 0; vifi < V_numvifs; vifi++)
1593 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1594 	    V_viftable[vifi].v_pkt_out++;
1595 	    V_viftable[vifi].v_bytes_out += plen;
1596 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1597 		pim_register_send(ip, V_viftable + vifi, m, rt);
1598 	    else
1599 		phyint_send(ip, V_viftable + vifi, m);
1600 	}
1601 
1602     /*
1603      * Perform upcall-related bw measuring.
1604      */
1605     if (rt->mfc_bw_meter != NULL) {
1606 	struct bw_meter *x;
1607 	struct timeval now;
1608 
1609 	microtime(&now);
1610 	MFC_LOCK_ASSERT();
1611 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1612 	    bw_meter_receive_packet(x, plen, &now);
1613     }
1614 
1615     return 0;
1616 }
1617 
1618 /*
1619  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1620  */
1621 static int
1622 X_legal_vif_num(int vif)
1623 {
1624 	int ret;
1625 
1626 	ret = 0;
1627 	if (vif < 0)
1628 		return (ret);
1629 
1630 	VIF_LOCK();
1631 	if (vif < V_numvifs)
1632 		ret = 1;
1633 	VIF_UNLOCK();
1634 
1635 	return (ret);
1636 }
1637 
1638 /*
1639  * Return the local address used by this vif
1640  */
1641 static u_long
1642 X_ip_mcast_src(int vifi)
1643 {
1644 	in_addr_t addr;
1645 
1646 	addr = INADDR_ANY;
1647 	if (vifi < 0)
1648 		return (addr);
1649 
1650 	VIF_LOCK();
1651 	if (vifi < V_numvifs)
1652 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1653 	VIF_UNLOCK();
1654 
1655 	return (addr);
1656 }
1657 
1658 static void
1659 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1660 {
1661     struct mbuf *mb_copy;
1662     int hlen = ip->ip_hl << 2;
1663 
1664     VIF_LOCK_ASSERT();
1665 
1666     /*
1667      * Make a new reference to the packet; make sure that
1668      * the IP header is actually copied, not just referenced,
1669      * so that ip_output() only scribbles on the copy.
1670      */
1671     mb_copy = m_copypacket(m, M_NOWAIT);
1672     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1673 	mb_copy = m_pullup(mb_copy, hlen);
1674     if (mb_copy == NULL)
1675 	return;
1676 
1677     send_packet(vifp, mb_copy);
1678 }
1679 
1680 static void
1681 send_packet(struct vif *vifp, struct mbuf *m)
1682 {
1683 	struct ip_moptions imo;
1684 	struct in_multi *imm[2];
1685 	int error __unused;
1686 
1687 	VIF_LOCK_ASSERT();
1688 
1689 	imo.imo_multicast_ifp  = vifp->v_ifp;
1690 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1691 	imo.imo_multicast_loop = 1;
1692 	imo.imo_multicast_vif  = -1;
1693 	imo.imo_num_memberships = 0;
1694 	imo.imo_max_memberships = 2;
1695 	imo.imo_membership  = &imm[0];
1696 
1697 	/*
1698 	 * Re-entrancy should not be a problem here, because
1699 	 * the packets that we send out and are looped back at us
1700 	 * should get rejected because they appear to come from
1701 	 * the loopback interface, thus preventing looping.
1702 	 */
1703 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1704 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1705 	    (ptrdiff_t)(vifp - V_viftable), error);
1706 }
1707 
1708 /*
1709  * Stubs for old RSVP socket shim implementation.
1710  */
1711 
1712 static int
1713 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1714 {
1715 
1716 	return (EOPNOTSUPP);
1717 }
1718 
1719 static void
1720 X_ip_rsvp_force_done(struct socket *so __unused)
1721 {
1722 
1723 }
1724 
1725 static int
1726 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1727 {
1728 	struct mbuf *m;
1729 
1730 	m = *mp;
1731 	*mp = NULL;
1732 	if (!V_rsvp_on)
1733 		m_freem(m);
1734 	return (IPPROTO_DONE);
1735 }
1736 
1737 /*
1738  * Code for bandwidth monitors
1739  */
1740 
1741 /*
1742  * Define common interface for timeval-related methods
1743  */
1744 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1745 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1746 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1747 
1748 static uint32_t
1749 compute_bw_meter_flags(struct bw_upcall *req)
1750 {
1751     uint32_t flags = 0;
1752 
1753     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1754 	flags |= BW_METER_UNIT_PACKETS;
1755     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1756 	flags |= BW_METER_UNIT_BYTES;
1757     if (req->bu_flags & BW_UPCALL_GEQ)
1758 	flags |= BW_METER_GEQ;
1759     if (req->bu_flags & BW_UPCALL_LEQ)
1760 	flags |= BW_METER_LEQ;
1761 
1762     return flags;
1763 }
1764 
1765 /*
1766  * Add a bw_meter entry
1767  */
1768 static int
1769 add_bw_upcall(struct bw_upcall *req)
1770 {
1771     struct mfc *mfc;
1772     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1773 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1774     struct timeval now;
1775     struct bw_meter *x;
1776     uint32_t flags;
1777 
1778     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1779 	return EOPNOTSUPP;
1780 
1781     /* Test if the flags are valid */
1782     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1783 	return EINVAL;
1784     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1785 	return EINVAL;
1786     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1787 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1788 	return EINVAL;
1789 
1790     /* Test if the threshold time interval is valid */
1791     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1792 	return EINVAL;
1793 
1794     flags = compute_bw_meter_flags(req);
1795 
1796     /*
1797      * Find if we have already same bw_meter entry
1798      */
1799     MFC_LOCK();
1800     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1801     if (mfc == NULL) {
1802 	MFC_UNLOCK();
1803 	return EADDRNOTAVAIL;
1804     }
1805     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1806 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1807 			   &req->bu_threshold.b_time, ==)) &&
1808 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1809 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1810 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1811 	    MFC_UNLOCK();
1812 	    return 0;		/* XXX Already installed */
1813 	}
1814     }
1815 
1816     /* Allocate the new bw_meter entry */
1817     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1818     if (x == NULL) {
1819 	MFC_UNLOCK();
1820 	return ENOBUFS;
1821     }
1822 
1823     /* Set the new bw_meter entry */
1824     x->bm_threshold.b_time = req->bu_threshold.b_time;
1825     microtime(&now);
1826     x->bm_start_time = now;
1827     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1828     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1829     x->bm_measured.b_packets = 0;
1830     x->bm_measured.b_bytes = 0;
1831     x->bm_flags = flags;
1832     x->bm_time_next = NULL;
1833     x->bm_time_hash = BW_METER_BUCKETS;
1834 
1835     /* Add the new bw_meter entry to the front of entries for this MFC */
1836     x->bm_mfc = mfc;
1837     x->bm_mfc_next = mfc->mfc_bw_meter;
1838     mfc->mfc_bw_meter = x;
1839     schedule_bw_meter(x, &now);
1840     MFC_UNLOCK();
1841 
1842     return 0;
1843 }
1844 
1845 static void
1846 free_bw_list(struct bw_meter *list)
1847 {
1848     while (list != NULL) {
1849 	struct bw_meter *x = list;
1850 
1851 	list = list->bm_mfc_next;
1852 	unschedule_bw_meter(x);
1853 	free(x, M_BWMETER);
1854     }
1855 }
1856 
1857 /*
1858  * Delete one or multiple bw_meter entries
1859  */
1860 static int
1861 del_bw_upcall(struct bw_upcall *req)
1862 {
1863     struct mfc *mfc;
1864     struct bw_meter *x;
1865 
1866     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1867 	return EOPNOTSUPP;
1868 
1869     MFC_LOCK();
1870 
1871     /* Find the corresponding MFC entry */
1872     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1873     if (mfc == NULL) {
1874 	MFC_UNLOCK();
1875 	return EADDRNOTAVAIL;
1876     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1877 	/*
1878 	 * Delete all bw_meter entries for this mfc
1879 	 */
1880 	struct bw_meter *list;
1881 
1882 	list = mfc->mfc_bw_meter;
1883 	mfc->mfc_bw_meter = NULL;
1884 	free_bw_list(list);
1885 	MFC_UNLOCK();
1886 	return 0;
1887     } else {			/* Delete a single bw_meter entry */
1888 	struct bw_meter *prev;
1889 	uint32_t flags = 0;
1890 
1891 	flags = compute_bw_meter_flags(req);
1892 
1893 	/* Find the bw_meter entry to delete */
1894 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1895 	     prev = x, x = x->bm_mfc_next) {
1896 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1897 			       &req->bu_threshold.b_time, ==)) &&
1898 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1899 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1900 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1901 		break;
1902 	}
1903 	if (x != NULL) { /* Delete entry from the list for this MFC */
1904 	    if (prev != NULL)
1905 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1906 	    else
1907 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1908 
1909 	    unschedule_bw_meter(x);
1910 	    MFC_UNLOCK();
1911 	    /* Free the bw_meter entry */
1912 	    free(x, M_BWMETER);
1913 	    return 0;
1914 	} else {
1915 	    MFC_UNLOCK();
1916 	    return EINVAL;
1917 	}
1918     }
1919     /* NOTREACHED */
1920 }
1921 
1922 /*
1923  * Perform bandwidth measurement processing that may result in an upcall
1924  */
1925 static void
1926 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1927 {
1928     struct timeval delta;
1929 
1930     MFC_LOCK_ASSERT();
1931 
1932     delta = *nowp;
1933     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1934 
1935     if (x->bm_flags & BW_METER_GEQ) {
1936 	/*
1937 	 * Processing for ">=" type of bw_meter entry
1938 	 */
1939 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1940 	    /* Reset the bw_meter entry */
1941 	    x->bm_start_time = *nowp;
1942 	    x->bm_measured.b_packets = 0;
1943 	    x->bm_measured.b_bytes = 0;
1944 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1945 	}
1946 
1947 	/* Record that a packet is received */
1948 	x->bm_measured.b_packets++;
1949 	x->bm_measured.b_bytes += plen;
1950 
1951 	/*
1952 	 * Test if we should deliver an upcall
1953 	 */
1954 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1955 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1956 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1957 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1958 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1959 		/* Prepare an upcall for delivery */
1960 		bw_meter_prepare_upcall(x, nowp);
1961 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1962 	    }
1963 	}
1964     } else if (x->bm_flags & BW_METER_LEQ) {
1965 	/*
1966 	 * Processing for "<=" type of bw_meter entry
1967 	 */
1968 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1969 	    /*
1970 	     * We are behind time with the multicast forwarding table
1971 	     * scanning for "<=" type of bw_meter entries, so test now
1972 	     * if we should deliver an upcall.
1973 	     */
1974 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1975 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1976 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1977 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1978 		/* Prepare an upcall for delivery */
1979 		bw_meter_prepare_upcall(x, nowp);
1980 	    }
1981 	    /* Reschedule the bw_meter entry */
1982 	    unschedule_bw_meter(x);
1983 	    schedule_bw_meter(x, nowp);
1984 	}
1985 
1986 	/* Record that a packet is received */
1987 	x->bm_measured.b_packets++;
1988 	x->bm_measured.b_bytes += plen;
1989 
1990 	/*
1991 	 * Test if we should restart the measuring interval
1992 	 */
1993 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1994 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1995 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1996 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1997 	    /* Don't restart the measuring interval */
1998 	} else {
1999 	    /* Do restart the measuring interval */
2000 	    /*
2001 	     * XXX: note that we don't unschedule and schedule, because this
2002 	     * might be too much overhead per packet. Instead, when we process
2003 	     * all entries for a given timer hash bin, we check whether it is
2004 	     * really a timeout. If not, we reschedule at that time.
2005 	     */
2006 	    x->bm_start_time = *nowp;
2007 	    x->bm_measured.b_packets = 0;
2008 	    x->bm_measured.b_bytes = 0;
2009 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2010 	}
2011     }
2012 }
2013 
2014 /*
2015  * Prepare a bandwidth-related upcall
2016  */
2017 static void
2018 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2019 {
2020     struct timeval delta;
2021     struct bw_upcall *u;
2022 
2023     MFC_LOCK_ASSERT();
2024 
2025     /*
2026      * Compute the measured time interval
2027      */
2028     delta = *nowp;
2029     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2030 
2031     /*
2032      * If there are too many pending upcalls, deliver them now
2033      */
2034     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2035 	bw_upcalls_send();
2036 
2037     /*
2038      * Set the bw_upcall entry
2039      */
2040     u = &V_bw_upcalls[V_bw_upcalls_n++];
2041     u->bu_src = x->bm_mfc->mfc_origin;
2042     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2043     u->bu_threshold.b_time = x->bm_threshold.b_time;
2044     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2045     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2046     u->bu_measured.b_time = delta;
2047     u->bu_measured.b_packets = x->bm_measured.b_packets;
2048     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2049     u->bu_flags = 0;
2050     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2051 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2052     if (x->bm_flags & BW_METER_UNIT_BYTES)
2053 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2054     if (x->bm_flags & BW_METER_GEQ)
2055 	u->bu_flags |= BW_UPCALL_GEQ;
2056     if (x->bm_flags & BW_METER_LEQ)
2057 	u->bu_flags |= BW_UPCALL_LEQ;
2058 }
2059 
2060 /*
2061  * Send the pending bandwidth-related upcalls
2062  */
2063 static void
2064 bw_upcalls_send(void)
2065 {
2066     struct mbuf *m;
2067     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2068     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2069     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2070 				      0,		/* unused2 */
2071 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2072 				      0,		/* im_mbz  */
2073 				      0,		/* im_vif  */
2074 				      0,		/* unused3 */
2075 				      { 0 },		/* im_src  */
2076 				      { 0 } };		/* im_dst  */
2077 
2078     MFC_LOCK_ASSERT();
2079 
2080     if (V_bw_upcalls_n == 0)
2081 	return;			/* No pending upcalls */
2082 
2083     V_bw_upcalls_n = 0;
2084 
2085     /*
2086      * Allocate a new mbuf, initialize it with the header and
2087      * the payload for the pending calls.
2088      */
2089     m = m_gethdr(M_NOWAIT, MT_DATA);
2090     if (m == NULL) {
2091 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2092 	return;
2093     }
2094 
2095     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2096     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2097 
2098     /*
2099      * Send the upcalls
2100      * XXX do we need to set the address in k_igmpsrc ?
2101      */
2102     MRTSTAT_INC(mrts_upcalls);
2103     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2104 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2105 	MRTSTAT_INC(mrts_upq_sockfull);
2106     }
2107 }
2108 
2109 /*
2110  * Compute the timeout hash value for the bw_meter entries
2111  */
2112 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2113     do {								\
2114 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2115 									\
2116 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2117 	(hash) = next_timeval.tv_sec;					\
2118 	if (next_timeval.tv_usec)					\
2119 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2120 	(hash) %= BW_METER_BUCKETS;					\
2121     } while (0)
2122 
2123 /*
2124  * Schedule a timer to process periodically bw_meter entry of type "<="
2125  * by linking the entry in the proper hash bucket.
2126  */
2127 static void
2128 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2129 {
2130     int time_hash;
2131 
2132     MFC_LOCK_ASSERT();
2133 
2134     if (!(x->bm_flags & BW_METER_LEQ))
2135 	return;		/* XXX: we schedule timers only for "<=" entries */
2136 
2137     /*
2138      * Reset the bw_meter entry
2139      */
2140     x->bm_start_time = *nowp;
2141     x->bm_measured.b_packets = 0;
2142     x->bm_measured.b_bytes = 0;
2143     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2144 
2145     /*
2146      * Compute the timeout hash value and insert the entry
2147      */
2148     BW_METER_TIMEHASH(x, time_hash);
2149     x->bm_time_next = V_bw_meter_timers[time_hash];
2150     V_bw_meter_timers[time_hash] = x;
2151     x->bm_time_hash = time_hash;
2152 }
2153 
2154 /*
2155  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2156  * by removing the entry from the proper hash bucket.
2157  */
2158 static void
2159 unschedule_bw_meter(struct bw_meter *x)
2160 {
2161     int time_hash;
2162     struct bw_meter *prev, *tmp;
2163 
2164     MFC_LOCK_ASSERT();
2165 
2166     if (!(x->bm_flags & BW_METER_LEQ))
2167 	return;		/* XXX: we schedule timers only for "<=" entries */
2168 
2169     /*
2170      * Compute the timeout hash value and delete the entry
2171      */
2172     time_hash = x->bm_time_hash;
2173     if (time_hash >= BW_METER_BUCKETS)
2174 	return;		/* Entry was not scheduled */
2175 
2176     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2177 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2178 	if (tmp == x)
2179 	    break;
2180 
2181     if (tmp == NULL)
2182 	panic("unschedule_bw_meter: bw_meter entry not found");
2183 
2184     if (prev != NULL)
2185 	prev->bm_time_next = x->bm_time_next;
2186     else
2187 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2188 
2189     x->bm_time_next = NULL;
2190     x->bm_time_hash = BW_METER_BUCKETS;
2191 }
2192 
2193 
2194 /*
2195  * Process all "<=" type of bw_meter that should be processed now,
2196  * and for each entry prepare an upcall if necessary. Each processed
2197  * entry is rescheduled again for the (periodic) processing.
2198  *
2199  * This is run periodically (once per second normally). On each round,
2200  * all the potentially matching entries are in the hash slot that we are
2201  * looking at.
2202  */
2203 static void
2204 bw_meter_process()
2205 {
2206     uint32_t loops;
2207     int i;
2208     struct timeval now, process_endtime;
2209 
2210     microtime(&now);
2211     if (V_last_tv_sec == now.tv_sec)
2212 	return;		/* nothing to do */
2213 
2214     loops = now.tv_sec - V_last_tv_sec;
2215     V_last_tv_sec = now.tv_sec;
2216     if (loops > BW_METER_BUCKETS)
2217 	loops = BW_METER_BUCKETS;
2218 
2219     MFC_LOCK();
2220     /*
2221      * Process all bins of bw_meter entries from the one after the last
2222      * processed to the current one. On entry, i points to the last bucket
2223      * visited, so we need to increment i at the beginning of the loop.
2224      */
2225     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2226 	struct bw_meter *x, *tmp_list;
2227 
2228 	if (++i >= BW_METER_BUCKETS)
2229 	    i = 0;
2230 
2231 	/* Disconnect the list of bw_meter entries from the bin */
2232 	tmp_list = V_bw_meter_timers[i];
2233 	V_bw_meter_timers[i] = NULL;
2234 
2235 	/* Process the list of bw_meter entries */
2236 	while (tmp_list != NULL) {
2237 	    x = tmp_list;
2238 	    tmp_list = tmp_list->bm_time_next;
2239 
2240 	    /* Test if the time interval is over */
2241 	    process_endtime = x->bm_start_time;
2242 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2243 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2244 		/* Not yet: reschedule, but don't reset */
2245 		int time_hash;
2246 
2247 		BW_METER_TIMEHASH(x, time_hash);
2248 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2249 		    /*
2250 		     * XXX: somehow the bin processing is a bit ahead of time.
2251 		     * Put the entry in the next bin.
2252 		     */
2253 		    if (++time_hash >= BW_METER_BUCKETS)
2254 			time_hash = 0;
2255 		}
2256 		x->bm_time_next = V_bw_meter_timers[time_hash];
2257 		V_bw_meter_timers[time_hash] = x;
2258 		x->bm_time_hash = time_hash;
2259 
2260 		continue;
2261 	    }
2262 
2263 	    /*
2264 	     * Test if we should deliver an upcall
2265 	     */
2266 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2267 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2268 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2269 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2270 		/* Prepare an upcall for delivery */
2271 		bw_meter_prepare_upcall(x, &now);
2272 	    }
2273 
2274 	    /*
2275 	     * Reschedule for next processing
2276 	     */
2277 	    schedule_bw_meter(x, &now);
2278 	}
2279     }
2280 
2281     /* Send all upcalls that are pending delivery */
2282     bw_upcalls_send();
2283 
2284     MFC_UNLOCK();
2285 }
2286 
2287 /*
2288  * A periodic function for sending all upcalls that are pending delivery
2289  */
2290 static void
2291 expire_bw_upcalls_send(void *arg)
2292 {
2293     CURVNET_SET((struct vnet *) arg);
2294 
2295     MFC_LOCK();
2296     bw_upcalls_send();
2297     MFC_UNLOCK();
2298 
2299     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2300 	curvnet);
2301     CURVNET_RESTORE();
2302 }
2303 
2304 /*
2305  * A periodic function for periodic scanning of the multicast forwarding
2306  * table for processing all "<=" bw_meter entries.
2307  */
2308 static void
2309 expire_bw_meter_process(void *arg)
2310 {
2311     CURVNET_SET((struct vnet *) arg);
2312 
2313     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2314 	bw_meter_process();
2315 
2316     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2317 	curvnet);
2318     CURVNET_RESTORE();
2319 }
2320 
2321 /*
2322  * End of bandwidth monitoring code
2323  */
2324 
2325 /*
2326  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2327  *
2328  */
2329 static int
2330 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2331     struct mfc *rt)
2332 {
2333     struct mbuf *mb_copy, *mm;
2334 
2335     /*
2336      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2337      * rendezvous point was unspecified, and we were told not to.
2338      */
2339     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2340 	in_nullhost(rt->mfc_rp))
2341 	return 0;
2342 
2343     mb_copy = pim_register_prepare(ip, m);
2344     if (mb_copy == NULL)
2345 	return ENOBUFS;
2346 
2347     /*
2348      * Send all the fragments. Note that the mbuf for each fragment
2349      * is freed by the sending machinery.
2350      */
2351     for (mm = mb_copy; mm; mm = mb_copy) {
2352 	mb_copy = mm->m_nextpkt;
2353 	mm->m_nextpkt = 0;
2354 	mm = m_pullup(mm, sizeof(struct ip));
2355 	if (mm != NULL) {
2356 	    ip = mtod(mm, struct ip *);
2357 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2358 		pim_register_send_rp(ip, vifp, mm, rt);
2359 	    } else {
2360 		pim_register_send_upcall(ip, vifp, mm, rt);
2361 	    }
2362 	}
2363     }
2364 
2365     return 0;
2366 }
2367 
2368 /*
2369  * Return a copy of the data packet that is ready for PIM Register
2370  * encapsulation.
2371  * XXX: Note that in the returned copy the IP header is a valid one.
2372  */
2373 static struct mbuf *
2374 pim_register_prepare(struct ip *ip, struct mbuf *m)
2375 {
2376     struct mbuf *mb_copy = NULL;
2377     int mtu;
2378 
2379     /* Take care of delayed checksums */
2380     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2381 	in_delayed_cksum(m);
2382 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2383     }
2384 
2385     /*
2386      * Copy the old packet & pullup its IP header into the
2387      * new mbuf so we can modify it.
2388      */
2389     mb_copy = m_copypacket(m, M_NOWAIT);
2390     if (mb_copy == NULL)
2391 	return NULL;
2392     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2393     if (mb_copy == NULL)
2394 	return NULL;
2395 
2396     /* take care of the TTL */
2397     ip = mtod(mb_copy, struct ip *);
2398     --ip->ip_ttl;
2399 
2400     /* Compute the MTU after the PIM Register encapsulation */
2401     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2402 
2403     if (ntohs(ip->ip_len) <= mtu) {
2404 	/* Turn the IP header into a valid one */
2405 	ip->ip_sum = 0;
2406 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2407     } else {
2408 	/* Fragment the packet */
2409 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2410 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2411 	    m_freem(mb_copy);
2412 	    return NULL;
2413 	}
2414     }
2415     return mb_copy;
2416 }
2417 
2418 /*
2419  * Send an upcall with the data packet to the user-level process.
2420  */
2421 static int
2422 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2423     struct mbuf *mb_copy, struct mfc *rt)
2424 {
2425     struct mbuf *mb_first;
2426     int len = ntohs(ip->ip_len);
2427     struct igmpmsg *im;
2428     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2429 
2430     VIF_LOCK_ASSERT();
2431 
2432     /*
2433      * Add a new mbuf with an upcall header
2434      */
2435     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2436     if (mb_first == NULL) {
2437 	m_freem(mb_copy);
2438 	return ENOBUFS;
2439     }
2440     mb_first->m_data += max_linkhdr;
2441     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2442     mb_first->m_len = sizeof(struct igmpmsg);
2443     mb_first->m_next = mb_copy;
2444 
2445     /* Send message to routing daemon */
2446     im = mtod(mb_first, struct igmpmsg *);
2447     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2448     im->im_mbz		= 0;
2449     im->im_vif		= vifp - V_viftable;
2450     im->im_src		= ip->ip_src;
2451     im->im_dst		= ip->ip_dst;
2452 
2453     k_igmpsrc.sin_addr	= ip->ip_src;
2454 
2455     MRTSTAT_INC(mrts_upcalls);
2456 
2457     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2458 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2459 	MRTSTAT_INC(mrts_upq_sockfull);
2460 	return ENOBUFS;
2461     }
2462 
2463     /* Keep statistics */
2464     PIMSTAT_INC(pims_snd_registers_msgs);
2465     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2466 
2467     return 0;
2468 }
2469 
2470 /*
2471  * Encapsulate the data packet in PIM Register message and send it to the RP.
2472  */
2473 static int
2474 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2475     struct mfc *rt)
2476 {
2477     struct mbuf *mb_first;
2478     struct ip *ip_outer;
2479     struct pim_encap_pimhdr *pimhdr;
2480     int len = ntohs(ip->ip_len);
2481     vifi_t vifi = rt->mfc_parent;
2482 
2483     VIF_LOCK_ASSERT();
2484 
2485     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2486 	m_freem(mb_copy);
2487 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2488     }
2489 
2490     /*
2491      * Add a new mbuf with the encapsulating header
2492      */
2493     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2494     if (mb_first == NULL) {
2495 	m_freem(mb_copy);
2496 	return ENOBUFS;
2497     }
2498     mb_first->m_data += max_linkhdr;
2499     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2500     mb_first->m_next = mb_copy;
2501 
2502     mb_first->m_pkthdr.len = len + mb_first->m_len;
2503 
2504     /*
2505      * Fill in the encapsulating IP and PIM header
2506      */
2507     ip_outer = mtod(mb_first, struct ip *);
2508     *ip_outer = pim_encap_iphdr;
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     ip_fillid(ip_outer);
2521     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2522 					 + sizeof(pim_encap_iphdr));
2523     *pimhdr = pim_encap_pimhdr;
2524     /* If the iif crosses a border, set the Border-bit */
2525     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2526 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2527 
2528     mb_first->m_data += sizeof(pim_encap_iphdr);
2529     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2530     mb_first->m_data -= sizeof(pim_encap_iphdr);
2531 
2532     send_packet(vifp, mb_first);
2533 
2534     /* Keep statistics */
2535     PIMSTAT_INC(pims_snd_registers_msgs);
2536     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2537 
2538     return 0;
2539 }
2540 
2541 /*
2542  * pim_encapcheck() is called by the encap4_input() path at runtime to
2543  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2544  * into the kernel.
2545  */
2546 static int
2547 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2548 {
2549 
2550 #ifdef DIAGNOSTIC
2551     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2552 #endif
2553     if (proto != IPPROTO_PIM)
2554 	return 0;	/* not for us; reject the datagram. */
2555 
2556     return 64;		/* claim the datagram. */
2557 }
2558 
2559 /*
2560  * PIM-SMv2 and PIM-DM messages processing.
2561  * Receives and verifies the PIM control messages, and passes them
2562  * up to the listening socket, using rip_input().
2563  * The only message with special processing is the PIM_REGISTER message
2564  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2565  * is passed to if_simloop().
2566  */
2567 int
2568 pim_input(struct mbuf **mp, int *offp, int proto)
2569 {
2570     struct mbuf *m = *mp;
2571     struct ip *ip = mtod(m, struct ip *);
2572     struct pim *pim;
2573     int iphlen = *offp;
2574     int minlen;
2575     int datalen = ntohs(ip->ip_len) - iphlen;
2576     int ip_tos;
2577 
2578     *mp = NULL;
2579 
2580     /* Keep statistics */
2581     PIMSTAT_INC(pims_rcv_total_msgs);
2582     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2583 
2584     /*
2585      * Validate lengths
2586      */
2587     if (datalen < PIM_MINLEN) {
2588 	PIMSTAT_INC(pims_rcv_tooshort);
2589 	CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2590 	    __func__, datalen, ntohl(ip->ip_src.s_addr));
2591 	m_freem(m);
2592 	return (IPPROTO_DONE);
2593     }
2594 
2595     /*
2596      * If the packet is at least as big as a REGISTER, go agead
2597      * and grab the PIM REGISTER header size, to avoid another
2598      * possible m_pullup() later.
2599      *
2600      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2601      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2602      */
2603     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2604     /*
2605      * Get the IP and PIM headers in contiguous memory, and
2606      * possibly the PIM REGISTER header.
2607      */
2608     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2609 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2610 	return (IPPROTO_DONE);
2611     }
2612 
2613     /* m_pullup() may have given us a new mbuf so reset ip. */
2614     ip = mtod(m, struct ip *);
2615     ip_tos = ip->ip_tos;
2616 
2617     /* adjust mbuf to point to the PIM header */
2618     m->m_data += iphlen;
2619     m->m_len  -= iphlen;
2620     pim = mtod(m, struct pim *);
2621 
2622     /*
2623      * Validate checksum. If PIM REGISTER, exclude the data packet.
2624      *
2625      * XXX: some older PIMv2 implementations don't make this distinction,
2626      * so for compatibility reason perform the checksum over part of the
2627      * message, and if error, then over the whole message.
2628      */
2629     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2630 	/* do nothing, checksum okay */
2631     } else if (in_cksum(m, datalen)) {
2632 	PIMSTAT_INC(pims_rcv_badsum);
2633 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2634 	m_freem(m);
2635 	return (IPPROTO_DONE);
2636     }
2637 
2638     /* PIM version check */
2639     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2640 	PIMSTAT_INC(pims_rcv_badversion);
2641 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2642 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2643 	m_freem(m);
2644 	return (IPPROTO_DONE);
2645     }
2646 
2647     /* restore mbuf back to the outer IP */
2648     m->m_data -= iphlen;
2649     m->m_len  += iphlen;
2650 
2651     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2652 	/*
2653 	 * Since this is a REGISTER, we'll make a copy of the register
2654 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2655 	 * routing daemon.
2656 	 */
2657 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2658 	struct mbuf *mcp;
2659 	struct ip *encap_ip;
2660 	u_int32_t *reghdr;
2661 	struct ifnet *vifp;
2662 
2663 	VIF_LOCK();
2664 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2665 	    VIF_UNLOCK();
2666 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2667 		(int)V_reg_vif_num);
2668 	    m_freem(m);
2669 	    return (IPPROTO_DONE);
2670 	}
2671 	/* XXX need refcnt? */
2672 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2673 	VIF_UNLOCK();
2674 
2675 	/*
2676 	 * Validate length
2677 	 */
2678 	if (datalen < PIM_REG_MINLEN) {
2679 	    PIMSTAT_INC(pims_rcv_tooshort);
2680 	    PIMSTAT_INC(pims_rcv_badregisters);
2681 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2682 	    m_freem(m);
2683 	    return (IPPROTO_DONE);
2684 	}
2685 
2686 	reghdr = (u_int32_t *)(pim + 1);
2687 	encap_ip = (struct ip *)(reghdr + 1);
2688 
2689 	CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2690 	    __func__, ntohl(encap_ip->ip_src.s_addr),
2691 	    ntohs(encap_ip->ip_len));
2692 
2693 	/* verify the version number of the inner packet */
2694 	if (encap_ip->ip_v != IPVERSION) {
2695 	    PIMSTAT_INC(pims_rcv_badregisters);
2696 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2697 	    m_freem(m);
2698 	    return (IPPROTO_DONE);
2699 	}
2700 
2701 	/* verify the inner packet is destined to a mcast group */
2702 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2703 	    PIMSTAT_INC(pims_rcv_badregisters);
2704 	    CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2705 		ntohl(encap_ip->ip_dst.s_addr));
2706 	    m_freem(m);
2707 	    return (IPPROTO_DONE);
2708 	}
2709 
2710 	/* If a NULL_REGISTER, pass it to the daemon */
2711 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2712 	    goto pim_input_to_daemon;
2713 
2714 	/*
2715 	 * Copy the TOS from the outer IP header to the inner IP header.
2716 	 */
2717 	if (encap_ip->ip_tos != ip_tos) {
2718 	    /* Outer TOS -> inner TOS */
2719 	    encap_ip->ip_tos = ip_tos;
2720 	    /* Recompute the inner header checksum. Sigh... */
2721 
2722 	    /* adjust mbuf to point to the inner IP header */
2723 	    m->m_data += (iphlen + PIM_MINLEN);
2724 	    m->m_len  -= (iphlen + PIM_MINLEN);
2725 
2726 	    encap_ip->ip_sum = 0;
2727 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2728 
2729 	    /* restore mbuf to point back to the outer IP header */
2730 	    m->m_data -= (iphlen + PIM_MINLEN);
2731 	    m->m_len  += (iphlen + PIM_MINLEN);
2732 	}
2733 
2734 	/*
2735 	 * Decapsulate the inner IP packet and loopback to forward it
2736 	 * as a normal multicast packet. Also, make a copy of the
2737 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2738 	 * to pass to the daemon later, so it can take the appropriate
2739 	 * actions (e.g., send back PIM_REGISTER_STOP).
2740 	 * XXX: here m->m_data points to the outer IP header.
2741 	 */
2742 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2743 	if (mcp == NULL) {
2744 	    CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2745 	    m_freem(m);
2746 	    return (IPPROTO_DONE);
2747 	}
2748 
2749 	/* Keep statistics */
2750 	/* XXX: registers_bytes include only the encap. mcast pkt */
2751 	PIMSTAT_INC(pims_rcv_registers_msgs);
2752 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2753 
2754 	/*
2755 	 * forward the inner ip packet; point m_data at the inner ip.
2756 	 */
2757 	m_adj(m, iphlen + PIM_MINLEN);
2758 
2759 	CTR4(KTR_IPMF,
2760 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2761 	    __func__,
2762 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2763 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2764 	    (int)V_reg_vif_num);
2765 
2766 	/* NB: vifp was collected above; can it change on us? */
2767 	if_simloop(vifp, m, dst.sin_family, 0);
2768 
2769 	/* prepare the register head to send to the mrouting daemon */
2770 	m = mcp;
2771     }
2772 
2773 pim_input_to_daemon:
2774     /*
2775      * Pass the PIM message up to the daemon; if it is a Register message,
2776      * pass the 'head' only up to the daemon. This includes the
2777      * outer IP header, PIM header, PIM-Register header and the
2778      * inner IP header.
2779      * XXX: the outer IP header pkt size of a Register is not adjust to
2780      * reflect the fact that the inner multicast data is truncated.
2781      */
2782     *mp = m;
2783     rip_input(mp, offp, proto);
2784 
2785     return (IPPROTO_DONE);
2786 }
2787 
2788 static int
2789 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2790 {
2791 	struct mfc	*rt;
2792 	int		 error, i;
2793 
2794 	if (req->newptr)
2795 		return (EPERM);
2796 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2797 		return (0);
2798 	error = sysctl_wire_old_buffer(req, 0);
2799 	if (error)
2800 		return (error);
2801 
2802 	MFC_LOCK();
2803 	for (i = 0; i < mfchashsize; i++) {
2804 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2805 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2806 			if (error)
2807 				goto out_locked;
2808 		}
2809 	}
2810 out_locked:
2811 	MFC_UNLOCK();
2812 	return (error);
2813 }
2814 
2815 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2816     sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2817     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2818 
2819 static void
2820 vnet_mroute_init(const void *unused __unused)
2821 {
2822 
2823 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2824 	bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2825 	callout_init(&V_expire_upcalls_ch, 1);
2826 	callout_init(&V_bw_upcalls_ch, 1);
2827 	callout_init(&V_bw_meter_ch, 1);
2828 }
2829 
2830 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2831 	NULL);
2832 
2833 static void
2834 vnet_mroute_uninit(const void *unused __unused)
2835 {
2836 
2837 	free(V_nexpire, M_MRTABLE);
2838 	V_nexpire = NULL;
2839 }
2840 
2841 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2842 	vnet_mroute_uninit, NULL);
2843 
2844 static int
2845 ip_mroute_modevent(module_t mod, int type, void *unused)
2846 {
2847 
2848     switch (type) {
2849     case MOD_LOAD:
2850 	MROUTER_LOCK_INIT();
2851 
2852 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2853 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2854 	if (if_detach_event_tag == NULL) {
2855 		printf("ip_mroute: unable to register "
2856 		    "ifnet_departure_event handler\n");
2857 		MROUTER_LOCK_DESTROY();
2858 		return (EINVAL);
2859 	}
2860 
2861 	MFC_LOCK_INIT();
2862 	VIF_LOCK_INIT();
2863 
2864 	mfchashsize = MFCHASHSIZE;
2865 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2866 	    !powerof2(mfchashsize)) {
2867 		printf("WARNING: %s not a power of 2; using default\n",
2868 		    "net.inet.ip.mfchashsize");
2869 		mfchashsize = MFCHASHSIZE;
2870 	}
2871 
2872 	pim_squelch_wholepkt = 0;
2873 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2874 	    &pim_squelch_wholepkt);
2875 
2876 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2877 	    pim_encapcheck, &in_pim_protosw, NULL);
2878 	if (pim_encap_cookie == NULL) {
2879 		printf("ip_mroute: unable to attach pim encap\n");
2880 		VIF_LOCK_DESTROY();
2881 		MFC_LOCK_DESTROY();
2882 		MROUTER_LOCK_DESTROY();
2883 		return (EINVAL);
2884 	}
2885 
2886 	ip_mcast_src = X_ip_mcast_src;
2887 	ip_mforward = X_ip_mforward;
2888 	ip_mrouter_done = X_ip_mrouter_done;
2889 	ip_mrouter_get = X_ip_mrouter_get;
2890 	ip_mrouter_set = X_ip_mrouter_set;
2891 
2892 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2893 	ip_rsvp_vif = X_ip_rsvp_vif;
2894 
2895 	legal_vif_num = X_legal_vif_num;
2896 	mrt_ioctl = X_mrt_ioctl;
2897 	rsvp_input_p = X_rsvp_input;
2898 	break;
2899 
2900     case MOD_UNLOAD:
2901 	/*
2902 	 * Typically module unload happens after the user-level
2903 	 * process has shutdown the kernel services (the check
2904 	 * below insures someone can't just yank the module out
2905 	 * from under a running process).  But if the module is
2906 	 * just loaded and then unloaded w/o starting up a user
2907 	 * process we still need to cleanup.
2908 	 */
2909 	MROUTER_LOCK();
2910 	if (ip_mrouter_cnt != 0) {
2911 	    MROUTER_UNLOCK();
2912 	    return (EINVAL);
2913 	}
2914 	ip_mrouter_unloading = 1;
2915 	MROUTER_UNLOCK();
2916 
2917 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2918 
2919 	if (pim_encap_cookie) {
2920 	    encap_detach(pim_encap_cookie);
2921 	    pim_encap_cookie = NULL;
2922 	}
2923 
2924 	ip_mcast_src = NULL;
2925 	ip_mforward = NULL;
2926 	ip_mrouter_done = NULL;
2927 	ip_mrouter_get = NULL;
2928 	ip_mrouter_set = NULL;
2929 
2930 	ip_rsvp_force_done = NULL;
2931 	ip_rsvp_vif = NULL;
2932 
2933 	legal_vif_num = NULL;
2934 	mrt_ioctl = NULL;
2935 	rsvp_input_p = NULL;
2936 
2937 	VIF_LOCK_DESTROY();
2938 	MFC_LOCK_DESTROY();
2939 	MROUTER_LOCK_DESTROY();
2940 	break;
2941 
2942     default:
2943 	return EOPNOTSUPP;
2944     }
2945     return 0;
2946 }
2947 
2948 static moduledata_t ip_mroutemod = {
2949     "ip_mroute",
2950     ip_mroute_modevent,
2951     0
2952 };
2953 
2954 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2955