xref: /freebsd/sys/netinet/ip_mroute.c (revision 3cbb4cc200f8a0ad7ed08233425ea54524a21f1c)
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 VNET_DEFINE_STATIC(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 VNET_PCPUSTAT_DEFINE_STATIC(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 VNET_DEFINE_STATIC(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 VNET_DEFINE_STATIC(u_char *, nexpire);		/* 0..mfchashsize-1 */
168 #define	V_nexpire		VNET(nexpire)
169 VNET_DEFINE_STATIC(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 VNET_DEFINE_STATIC(vifi_t, numvifs);
181 #define	V_numvifs		VNET(numvifs)
182 VNET_DEFINE_STATIC(struct vif *, viftable);
183 #define	V_viftable		VNET(viftable)
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 VNET_DEFINE_STATIC(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 VNET_DEFINE_STATIC(struct bw_meter **, bw_meter_timers);
211 #define	V_bw_meter_timers	VNET(bw_meter_timers)
212 VNET_DEFINE_STATIC(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 VNET_DEFINE_STATIC(struct bw_upcall *, bw_upcalls);
221 #define	V_bw_upcalls		VNET(bw_upcalls)
222 VNET_DEFINE_STATIC(u_int, bw_upcalls_n); /* # of pending upcalls */
223 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
224 VNET_DEFINE_STATIC(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 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
230 VNET_PCPUSTAT_SYSINIT(pimstat);
231 VNET_PCPUSTAT_SYSUNINIT(pimstat);
232 
233 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
234     "PIM");
235 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
236     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
237 
238 static u_long	pim_squelch_wholepkt = 0;
239 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
240     &pim_squelch_wholepkt, 0,
241     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
242 
243 static const struct encaptab *pim_encap_cookie;
244 static int pim_encapcheck(const struct mbuf *, int, int, void *);
245 static int pim_input(struct mbuf *, int, int, void *);
246 
247 static const struct encap_config ipv4_encap_cfg = {
248 	.proto = IPPROTO_PIM,
249 	.min_length = sizeof(struct ip) + PIM_MINLEN,
250 	.exact_match = 8,
251 	.check = pim_encapcheck,
252 	.input = pim_input
253 };
254 
255 /*
256  * Note: the PIM Register encapsulation adds the following in front of a
257  * data packet:
258  *
259  * struct pim_encap_hdr {
260  *    struct ip ip;
261  *    struct pim_encap_pimhdr  pim;
262  * }
263  *
264  */
265 
266 struct pim_encap_pimhdr {
267 	struct pim pim;
268 	uint32_t   flags;
269 };
270 #define		PIM_ENCAP_TTL	64
271 
272 static struct ip pim_encap_iphdr = {
273 #if BYTE_ORDER == LITTLE_ENDIAN
274 	sizeof(struct ip) >> 2,
275 	IPVERSION,
276 #else
277 	IPVERSION,
278 	sizeof(struct ip) >> 2,
279 #endif
280 	0,			/* tos */
281 	sizeof(struct ip),	/* total length */
282 	0,			/* id */
283 	0,			/* frag offset */
284 	PIM_ENCAP_TTL,
285 	IPPROTO_PIM,
286 	0,			/* checksum */
287 };
288 
289 static struct pim_encap_pimhdr pim_encap_pimhdr = {
290     {
291 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
292 	0,			/* reserved */
293 	0,			/* checksum */
294     },
295     0				/* flags */
296 };
297 
298 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
299 #define	V_reg_vif_num		VNET(reg_vif_num)
300 VNET_DEFINE_STATIC(struct ifnet, multicast_register_if);
301 #define	V_multicast_register_if	VNET(multicast_register_if)
302 
303 /*
304  * Private variables.
305  */
306 
307 static u_long	X_ip_mcast_src(int);
308 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
309 		    struct ip_moptions *);
310 static int	X_ip_mrouter_done(void);
311 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
312 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
313 static int	X_legal_vif_num(int);
314 static int	X_mrt_ioctl(u_long, caddr_t, int);
315 
316 static int	add_bw_upcall(struct bw_upcall *);
317 static int	add_mfc(struct mfcctl2 *);
318 static int	add_vif(struct vifctl *);
319 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
320 static void	bw_meter_process(void);
321 static void	bw_meter_receive_packet(struct bw_meter *, int,
322 		    struct timeval *);
323 static void	bw_upcalls_send(void);
324 static int	del_bw_upcall(struct bw_upcall *);
325 static int	del_mfc(struct mfcctl2 *);
326 static int	del_vif(vifi_t);
327 static int	del_vif_locked(vifi_t);
328 static void	expire_bw_meter_process(void *);
329 static void	expire_bw_upcalls_send(void *);
330 static void	expire_mfc(struct mfc *);
331 static void	expire_upcalls(void *);
332 static void	free_bw_list(struct bw_meter *);
333 static int	get_sg_cnt(struct sioc_sg_req *);
334 static int	get_vif_cnt(struct sioc_vif_req *);
335 static void	if_detached_event(void *, struct ifnet *);
336 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
337 static int	ip_mrouter_init(struct socket *, int);
338 static __inline struct mfc *
339 		mfc_find(struct in_addr *, struct in_addr *);
340 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
341 static struct mbuf *
342 		pim_register_prepare(struct ip *, struct mbuf *);
343 static int	pim_register_send(struct ip *, struct vif *,
344 		    struct mbuf *, struct mfc *);
345 static int	pim_register_send_rp(struct ip *, struct vif *,
346 		    struct mbuf *, struct mfc *);
347 static int	pim_register_send_upcall(struct ip *, struct vif *,
348 		    struct mbuf *, struct mfc *);
349 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
350 static void	send_packet(struct vif *, struct mbuf *);
351 static int	set_api_config(uint32_t *);
352 static int	set_assert(int);
353 static int	socket_send(struct socket *, struct mbuf *,
354 		    struct sockaddr_in *);
355 static void	unschedule_bw_meter(struct bw_meter *);
356 
357 /*
358  * Kernel multicast forwarding API capabilities and setup.
359  * If more API capabilities are added to the kernel, they should be
360  * recorded in `mrt_api_support'.
361  */
362 #define MRT_API_VERSION		0x0305
363 
364 static const int mrt_api_version = MRT_API_VERSION;
365 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
366 					 MRT_MFC_FLAGS_BORDER_VIF |
367 					 MRT_MFC_RP |
368 					 MRT_MFC_BW_UPCALL);
369 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
370 #define	V_mrt_api_config	VNET(mrt_api_config)
371 VNET_DEFINE_STATIC(int, pim_assert_enabled);
372 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
373 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
374 
375 /*
376  * Find a route for a given origin IP address and multicast group address.
377  * Statistics must be updated by the caller.
378  */
379 static __inline struct mfc *
380 mfc_find(struct in_addr *o, struct in_addr *g)
381 {
382 	struct mfc *rt;
383 
384 	MFC_LOCK_ASSERT();
385 
386 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
387 		if (in_hosteq(rt->mfc_origin, *o) &&
388 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
389 		    TAILQ_EMPTY(&rt->mfc_stall))
390 			break;
391 	}
392 
393 	return (rt);
394 }
395 
396 /*
397  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
398  */
399 static int
400 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
401 {
402     int	error, optval;
403     vifi_t	vifi;
404     struct	vifctl vifc;
405     struct	mfcctl2 mfc;
406     struct	bw_upcall bw_upcall;
407     uint32_t	i;
408 
409     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
410 	return EPERM;
411 
412     error = 0;
413     switch (sopt->sopt_name) {
414     case MRT_INIT:
415 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
416 	if (error)
417 	    break;
418 	error = ip_mrouter_init(so, optval);
419 	break;
420 
421     case MRT_DONE:
422 	error = ip_mrouter_done();
423 	break;
424 
425     case MRT_ADD_VIF:
426 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
427 	if (error)
428 	    break;
429 	error = add_vif(&vifc);
430 	break;
431 
432     case MRT_DEL_VIF:
433 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
434 	if (error)
435 	    break;
436 	error = del_vif(vifi);
437 	break;
438 
439     case MRT_ADD_MFC:
440     case MRT_DEL_MFC:
441 	/*
442 	 * select data size depending on API version.
443 	 */
444 	if (sopt->sopt_name == MRT_ADD_MFC &&
445 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
446 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
447 				sizeof(struct mfcctl2));
448 	} else {
449 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
450 				sizeof(struct mfcctl));
451 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
452 			sizeof(mfc) - sizeof(struct mfcctl));
453 	}
454 	if (error)
455 	    break;
456 	if (sopt->sopt_name == MRT_ADD_MFC)
457 	    error = add_mfc(&mfc);
458 	else
459 	    error = del_mfc(&mfc);
460 	break;
461 
462     case MRT_ASSERT:
463 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
464 	if (error)
465 	    break;
466 	set_assert(optval);
467 	break;
468 
469     case MRT_API_CONFIG:
470 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
471 	if (!error)
472 	    error = set_api_config(&i);
473 	if (!error)
474 	    error = sooptcopyout(sopt, &i, sizeof i);
475 	break;
476 
477     case MRT_ADD_BW_UPCALL:
478     case MRT_DEL_BW_UPCALL:
479 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
480 				sizeof bw_upcall);
481 	if (error)
482 	    break;
483 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
484 	    error = add_bw_upcall(&bw_upcall);
485 	else
486 	    error = del_bw_upcall(&bw_upcall);
487 	break;
488 
489     default:
490 	error = EOPNOTSUPP;
491 	break;
492     }
493     return error;
494 }
495 
496 /*
497  * Handle MRT getsockopt commands
498  */
499 static int
500 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
501 {
502     int error;
503 
504     switch (sopt->sopt_name) {
505     case MRT_VERSION:
506 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
507 	break;
508 
509     case MRT_ASSERT:
510 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
511 	    sizeof V_pim_assert_enabled);
512 	break;
513 
514     case MRT_API_SUPPORT:
515 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
516 	break;
517 
518     case MRT_API_CONFIG:
519 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
520 	break;
521 
522     default:
523 	error = EOPNOTSUPP;
524 	break;
525     }
526     return error;
527 }
528 
529 /*
530  * Handle ioctl commands to obtain information from the cache
531  */
532 static int
533 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
534 {
535     int error = 0;
536 
537     /*
538      * Currently the only function calling this ioctl routine is rtioctl_fib().
539      * Typically, only root can create the raw socket in order to execute
540      * this ioctl method, however the request might be coming from a prison
541      */
542     error = priv_check(curthread, PRIV_NETINET_MROUTE);
543     if (error)
544 	return (error);
545     switch (cmd) {
546     case (SIOCGETVIFCNT):
547 	error = get_vif_cnt((struct sioc_vif_req *)data);
548 	break;
549 
550     case (SIOCGETSGCNT):
551 	error = get_sg_cnt((struct sioc_sg_req *)data);
552 	break;
553 
554     default:
555 	error = EINVAL;
556 	break;
557     }
558     return error;
559 }
560 
561 /*
562  * returns the packet, byte, rpf-failure count for the source group provided
563  */
564 static int
565 get_sg_cnt(struct sioc_sg_req *req)
566 {
567     struct mfc *rt;
568 
569     MFC_LOCK();
570     rt = mfc_find(&req->src, &req->grp);
571     if (rt == NULL) {
572 	MFC_UNLOCK();
573 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
574 	return EADDRNOTAVAIL;
575     }
576     req->pktcnt = rt->mfc_pkt_cnt;
577     req->bytecnt = rt->mfc_byte_cnt;
578     req->wrong_if = rt->mfc_wrong_if;
579     MFC_UNLOCK();
580     return 0;
581 }
582 
583 /*
584  * returns the input and output packet and byte counts on the vif provided
585  */
586 static int
587 get_vif_cnt(struct sioc_vif_req *req)
588 {
589     vifi_t vifi = req->vifi;
590 
591     VIF_LOCK();
592     if (vifi >= V_numvifs) {
593 	VIF_UNLOCK();
594 	return EINVAL;
595     }
596 
597     req->icount = V_viftable[vifi].v_pkt_in;
598     req->ocount = V_viftable[vifi].v_pkt_out;
599     req->ibytes = V_viftable[vifi].v_bytes_in;
600     req->obytes = V_viftable[vifi].v_bytes_out;
601     VIF_UNLOCK();
602 
603     return 0;
604 }
605 
606 static void
607 if_detached_event(void *arg __unused, struct ifnet *ifp)
608 {
609     vifi_t vifi;
610     u_long i;
611 
612     MROUTER_LOCK();
613 
614     if (V_ip_mrouter == NULL) {
615 	MROUTER_UNLOCK();
616 	return;
617     }
618 
619     VIF_LOCK();
620     MFC_LOCK();
621 
622     /*
623      * Tear down multicast forwarder state associated with this ifnet.
624      * 1. Walk the vif list, matching vifs against this ifnet.
625      * 2. Walk the multicast forwarding cache (mfc) looking for
626      *    inner matches with this vif's index.
627      * 3. Expire any matching multicast forwarding cache entries.
628      * 4. Free vif state. This should disable ALLMULTI on the interface.
629      */
630     for (vifi = 0; vifi < V_numvifs; vifi++) {
631 	if (V_viftable[vifi].v_ifp != ifp)
632 		continue;
633 	for (i = 0; i < mfchashsize; i++) {
634 		struct mfc *rt, *nrt;
635 
636 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
637 			if (rt->mfc_parent == vifi) {
638 				expire_mfc(rt);
639 			}
640 		}
641 	}
642 	del_vif_locked(vifi);
643     }
644 
645     MFC_UNLOCK();
646     VIF_UNLOCK();
647 
648     MROUTER_UNLOCK();
649 }
650 
651 /*
652  * Enable multicast forwarding.
653  */
654 static int
655 ip_mrouter_init(struct socket *so, int version)
656 {
657 
658     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
659         so->so_type, so->so_proto->pr_protocol);
660 
661     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
662 	return EOPNOTSUPP;
663 
664     if (version != 1)
665 	return ENOPROTOOPT;
666 
667     MROUTER_LOCK();
668 
669     if (ip_mrouter_unloading) {
670 	MROUTER_UNLOCK();
671 	return ENOPROTOOPT;
672     }
673 
674     if (V_ip_mrouter != NULL) {
675 	MROUTER_UNLOCK();
676 	return EADDRINUSE;
677     }
678 
679     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
680 	HASH_NOWAIT);
681 
682     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
683 	curvnet);
684     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
685 	curvnet);
686     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
687 	curvnet);
688 
689     V_ip_mrouter = so;
690     ip_mrouter_cnt++;
691 
692     MROUTER_UNLOCK();
693 
694     CTR1(KTR_IPMF, "%s: done", __func__);
695 
696     return 0;
697 }
698 
699 /*
700  * Disable multicast forwarding.
701  */
702 static int
703 X_ip_mrouter_done(void)
704 {
705     struct ifnet *ifp;
706     u_long i;
707     vifi_t vifi;
708 
709     MROUTER_LOCK();
710 
711     if (V_ip_mrouter == NULL) {
712 	MROUTER_UNLOCK();
713 	return EINVAL;
714     }
715 
716     /*
717      * Detach/disable hooks to the reset of the system.
718      */
719     V_ip_mrouter = NULL;
720     ip_mrouter_cnt--;
721     V_mrt_api_config = 0;
722 
723     VIF_LOCK();
724 
725     /*
726      * For each phyint in use, disable promiscuous reception of all IP
727      * multicasts.
728      */
729     for (vifi = 0; vifi < V_numvifs; vifi++) {
730 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
731 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
732 	    ifp = V_viftable[vifi].v_ifp;
733 	    if_allmulti(ifp, 0);
734 	}
735     }
736     bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
737     V_numvifs = 0;
738     V_pim_assert_enabled = 0;
739 
740     VIF_UNLOCK();
741 
742     callout_stop(&V_expire_upcalls_ch);
743     callout_stop(&V_bw_upcalls_ch);
744     callout_stop(&V_bw_meter_ch);
745 
746     MFC_LOCK();
747 
748     /*
749      * Free all multicast forwarding cache entries.
750      * Do not use hashdestroy(), as we must perform other cleanup.
751      */
752     for (i = 0; i < mfchashsize; i++) {
753 	struct mfc *rt, *nrt;
754 
755 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
756 		expire_mfc(rt);
757 	}
758     }
759     free(V_mfchashtbl, M_MRTABLE);
760     V_mfchashtbl = NULL;
761 
762     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
763 
764     V_bw_upcalls_n = 0;
765     bzero(V_bw_meter_timers, BW_METER_BUCKETS * sizeof(*V_bw_meter_timers));
766 
767     MFC_UNLOCK();
768 
769     V_reg_vif_num = VIFI_INVALID;
770 
771     MROUTER_UNLOCK();
772 
773     CTR1(KTR_IPMF, "%s: done", __func__);
774 
775     return 0;
776 }
777 
778 /*
779  * Set PIM assert processing global
780  */
781 static int
782 set_assert(int i)
783 {
784     if ((i != 1) && (i != 0))
785 	return EINVAL;
786 
787     V_pim_assert_enabled = i;
788 
789     return 0;
790 }
791 
792 /*
793  * Configure API capabilities
794  */
795 int
796 set_api_config(uint32_t *apival)
797 {
798     u_long i;
799 
800     /*
801      * We can set the API capabilities only if it is the first operation
802      * after MRT_INIT. I.e.:
803      *  - there are no vifs installed
804      *  - pim_assert is not enabled
805      *  - the MFC table is empty
806      */
807     if (V_numvifs > 0) {
808 	*apival = 0;
809 	return EPERM;
810     }
811     if (V_pim_assert_enabled) {
812 	*apival = 0;
813 	return EPERM;
814     }
815 
816     MFC_LOCK();
817 
818     for (i = 0; i < mfchashsize; i++) {
819 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
820 	    MFC_UNLOCK();
821 	    *apival = 0;
822 	    return EPERM;
823 	}
824     }
825 
826     MFC_UNLOCK();
827 
828     V_mrt_api_config = *apival & mrt_api_support;
829     *apival = V_mrt_api_config;
830 
831     return 0;
832 }
833 
834 /*
835  * Add a vif to the vif table
836  */
837 static int
838 add_vif(struct vifctl *vifcp)
839 {
840     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
841     struct sockaddr_in sin = {sizeof sin, AF_INET};
842     struct ifaddr *ifa;
843     struct ifnet *ifp;
844     int error;
845 
846     VIF_LOCK();
847     if (vifcp->vifc_vifi >= MAXVIFS) {
848 	VIF_UNLOCK();
849 	return EINVAL;
850     }
851     /* rate limiting is no longer supported by this code */
852     if (vifcp->vifc_rate_limit != 0) {
853 	log(LOG_ERR, "rate limiting is no longer supported\n");
854 	VIF_UNLOCK();
855 	return EINVAL;
856     }
857     if (!in_nullhost(vifp->v_lcl_addr)) {
858 	VIF_UNLOCK();
859 	return EADDRINUSE;
860     }
861     if (in_nullhost(vifcp->vifc_lcl_addr)) {
862 	VIF_UNLOCK();
863 	return EADDRNOTAVAIL;
864     }
865 
866     /* Find the interface with an address in AF_INET family */
867     if (vifcp->vifc_flags & VIFF_REGISTER) {
868 	/*
869 	 * XXX: Because VIFF_REGISTER does not really need a valid
870 	 * local interface (e.g. it could be 127.0.0.2), we don't
871 	 * check its address.
872 	 */
873 	ifp = NULL;
874     } else {
875 	struct epoch_tracker et;
876 
877 	sin.sin_addr = vifcp->vifc_lcl_addr;
878 	NET_EPOCH_ENTER(et);
879 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
880 	if (ifa == NULL) {
881 	    NET_EPOCH_EXIT(et);
882 	    VIF_UNLOCK();
883 	    return EADDRNOTAVAIL;
884 	}
885 	ifp = ifa->ifa_ifp;
886 	/* XXX FIXME we need to take a ref on ifp and cleanup properly! */
887 	NET_EPOCH_EXIT(et);
888     }
889 
890     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
891 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
892 	VIF_UNLOCK();
893 	return EOPNOTSUPP;
894     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
895 	ifp = &V_multicast_register_if;
896 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
897 	if (V_reg_vif_num == VIFI_INVALID) {
898 	    if_initname(&V_multicast_register_if, "register_vif", 0);
899 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
900 	    V_reg_vif_num = vifcp->vifc_vifi;
901 	}
902     } else {		/* Make sure the interface supports multicast */
903 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
904 	    VIF_UNLOCK();
905 	    return EOPNOTSUPP;
906 	}
907 
908 	/* Enable promiscuous reception of all IP multicasts from the if */
909 	error = if_allmulti(ifp, 1);
910 	if (error) {
911 	    VIF_UNLOCK();
912 	    return error;
913 	}
914     }
915 
916     vifp->v_flags     = vifcp->vifc_flags;
917     vifp->v_threshold = vifcp->vifc_threshold;
918     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
919     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
920     vifp->v_ifp       = ifp;
921     /* initialize per vif pkt counters */
922     vifp->v_pkt_in    = 0;
923     vifp->v_pkt_out   = 0;
924     vifp->v_bytes_in  = 0;
925     vifp->v_bytes_out = 0;
926 
927     /* Adjust numvifs up if the vifi is higher than numvifs */
928     if (V_numvifs <= vifcp->vifc_vifi)
929 	V_numvifs = vifcp->vifc_vifi + 1;
930 
931     VIF_UNLOCK();
932 
933     CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
934 	(int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
935 	(int)vifcp->vifc_threshold);
936 
937     return 0;
938 }
939 
940 /*
941  * Delete a vif from the vif table
942  */
943 static int
944 del_vif_locked(vifi_t vifi)
945 {
946     struct vif *vifp;
947 
948     VIF_LOCK_ASSERT();
949 
950     if (vifi >= V_numvifs) {
951 	return EINVAL;
952     }
953     vifp = &V_viftable[vifi];
954     if (in_nullhost(vifp->v_lcl_addr)) {
955 	return EADDRNOTAVAIL;
956     }
957 
958     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
959 	if_allmulti(vifp->v_ifp, 0);
960 
961     if (vifp->v_flags & VIFF_REGISTER)
962 	V_reg_vif_num = VIFI_INVALID;
963 
964     bzero((caddr_t)vifp, sizeof (*vifp));
965 
966     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
967 
968     /* Adjust numvifs down */
969     for (vifi = V_numvifs; vifi > 0; vifi--)
970 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
971 	    break;
972     V_numvifs = vifi;
973 
974     return 0;
975 }
976 
977 static int
978 del_vif(vifi_t vifi)
979 {
980     int cc;
981 
982     VIF_LOCK();
983     cc = del_vif_locked(vifi);
984     VIF_UNLOCK();
985 
986     return cc;
987 }
988 
989 /*
990  * update an mfc entry without resetting counters and S,G addresses.
991  */
992 static void
993 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
994 {
995     int i;
996 
997     rt->mfc_parent = mfccp->mfcc_parent;
998     for (i = 0; i < V_numvifs; i++) {
999 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1000 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1001 	    MRT_MFC_FLAGS_ALL;
1002     }
1003     /* set the RP address */
1004     if (V_mrt_api_config & MRT_MFC_RP)
1005 	rt->mfc_rp = mfccp->mfcc_rp;
1006     else
1007 	rt->mfc_rp.s_addr = INADDR_ANY;
1008 }
1009 
1010 /*
1011  * fully initialize an mfc entry from the parameter.
1012  */
1013 static void
1014 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1015 {
1016     rt->mfc_origin     = mfccp->mfcc_origin;
1017     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1018 
1019     update_mfc_params(rt, mfccp);
1020 
1021     /* initialize pkt counters per src-grp */
1022     rt->mfc_pkt_cnt    = 0;
1023     rt->mfc_byte_cnt   = 0;
1024     rt->mfc_wrong_if   = 0;
1025     timevalclear(&rt->mfc_last_assert);
1026 }
1027 
1028 static void
1029 expire_mfc(struct mfc *rt)
1030 {
1031 	struct rtdetq *rte, *nrte;
1032 
1033 	MFC_LOCK_ASSERT();
1034 
1035 	free_bw_list(rt->mfc_bw_meter);
1036 
1037 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1038 		m_freem(rte->m);
1039 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1040 		free(rte, M_MRTABLE);
1041 	}
1042 
1043 	LIST_REMOVE(rt, mfc_hash);
1044 	free(rt, M_MRTABLE);
1045 }
1046 
1047 /*
1048  * Add an mfc entry
1049  */
1050 static int
1051 add_mfc(struct mfcctl2 *mfccp)
1052 {
1053     struct mfc *rt;
1054     struct rtdetq *rte, *nrte;
1055     u_long hash = 0;
1056     u_short nstl;
1057 
1058     VIF_LOCK();
1059     MFC_LOCK();
1060 
1061     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1062 
1063     /* If an entry already exists, just update the fields */
1064     if (rt) {
1065 	CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1066 	    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1067 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1068 	    mfccp->mfcc_parent);
1069 	update_mfc_params(rt, mfccp);
1070 	MFC_UNLOCK();
1071 	VIF_UNLOCK();
1072 	return (0);
1073     }
1074 
1075     /*
1076      * Find the entry for which the upcall was made and update
1077      */
1078     nstl = 0;
1079     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1080     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1081 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1082 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1083 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1084 		CTR5(KTR_IPMF,
1085 		    "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1086 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1087 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1088 		    mfccp->mfcc_parent,
1089 		    TAILQ_FIRST(&rt->mfc_stall));
1090 		if (nstl++)
1091 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1092 
1093 		init_mfc_params(rt, mfccp);
1094 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1095 		V_nexpire[hash]--;
1096 
1097 		/* Free queued packets, but attempt to forward them first. */
1098 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1099 			if (rte->ifp != NULL)
1100 				ip_mdq(rte->m, rte->ifp, rt, -1);
1101 			m_freem(rte->m);
1102 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1103 			rt->mfc_nstall--;
1104 			free(rte, M_MRTABLE);
1105 		}
1106 	}
1107     }
1108 
1109     /*
1110      * It is possible that an entry is being inserted without an upcall
1111      */
1112     if (nstl == 0) {
1113 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1114 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1115 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1116 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1117 			init_mfc_params(rt, mfccp);
1118 			if (rt->mfc_expire)
1119 			    V_nexpire[hash]--;
1120 			rt->mfc_expire = 0;
1121 			break; /* XXX */
1122 		}
1123 	}
1124 
1125 	if (rt == NULL) {		/* no upcall, so make a new entry */
1126 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1127 	    if (rt == NULL) {
1128 		MFC_UNLOCK();
1129 		VIF_UNLOCK();
1130 		return (ENOBUFS);
1131 	    }
1132 
1133 	    init_mfc_params(rt, mfccp);
1134 	    TAILQ_INIT(&rt->mfc_stall);
1135 	    rt->mfc_nstall = 0;
1136 
1137 	    rt->mfc_expire     = 0;
1138 	    rt->mfc_bw_meter = NULL;
1139 
1140 	    /* insert new entry at head of hash chain */
1141 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1142 	}
1143     }
1144 
1145     MFC_UNLOCK();
1146     VIF_UNLOCK();
1147 
1148     return (0);
1149 }
1150 
1151 /*
1152  * Delete an mfc entry
1153  */
1154 static int
1155 del_mfc(struct mfcctl2 *mfccp)
1156 {
1157     struct in_addr	origin;
1158     struct in_addr	mcastgrp;
1159     struct mfc		*rt;
1160 
1161     origin = mfccp->mfcc_origin;
1162     mcastgrp = mfccp->mfcc_mcastgrp;
1163 
1164     CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1165 	ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1166 
1167     MFC_LOCK();
1168 
1169     rt = mfc_find(&origin, &mcastgrp);
1170     if (rt == NULL) {
1171 	MFC_UNLOCK();
1172 	return EADDRNOTAVAIL;
1173     }
1174 
1175     /*
1176      * free the bw_meter entries
1177      */
1178     free_bw_list(rt->mfc_bw_meter);
1179     rt->mfc_bw_meter = NULL;
1180 
1181     LIST_REMOVE(rt, mfc_hash);
1182     free(rt, M_MRTABLE);
1183 
1184     MFC_UNLOCK();
1185 
1186     return (0);
1187 }
1188 
1189 /*
1190  * Send a message to the routing daemon on the multicast routing socket.
1191  */
1192 static int
1193 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1194 {
1195     if (s) {
1196 	SOCKBUF_LOCK(&s->so_rcv);
1197 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1198 	    NULL) != 0) {
1199 	    sorwakeup_locked(s);
1200 	    return 0;
1201 	}
1202 	SOCKBUF_UNLOCK(&s->so_rcv);
1203     }
1204     m_freem(mm);
1205     return -1;
1206 }
1207 
1208 /*
1209  * IP multicast forwarding function. This function assumes that the packet
1210  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1211  * pointed to by "ifp", and the packet is to be relayed to other networks
1212  * that have members of the packet's destination IP multicast group.
1213  *
1214  * The packet is returned unscathed to the caller, unless it is
1215  * erroneous, in which case a non-zero return value tells the caller to
1216  * discard it.
1217  */
1218 
1219 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1220 
1221 static int
1222 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1223     struct ip_moptions *imo)
1224 {
1225     struct mfc *rt;
1226     int error;
1227     vifi_t vifi;
1228 
1229     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1230 	ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1231 
1232     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1233 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1234 	/*
1235 	 * Packet arrived via a physical interface or
1236 	 * an encapsulated tunnel or a register_vif.
1237 	 */
1238     } else {
1239 	/*
1240 	 * Packet arrived through a source-route tunnel.
1241 	 * Source-route tunnels are no longer supported.
1242 	 */
1243 	return (1);
1244     }
1245 
1246     VIF_LOCK();
1247     MFC_LOCK();
1248     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1249 	if (ip->ip_ttl < MAXTTL)
1250 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1251 	error = ip_mdq(m, ifp, NULL, vifi);
1252 	MFC_UNLOCK();
1253 	VIF_UNLOCK();
1254 	return error;
1255     }
1256 
1257     /*
1258      * Don't forward a packet with time-to-live of zero or one,
1259      * or a packet destined to a local-only group.
1260      */
1261     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1262 	MFC_UNLOCK();
1263 	VIF_UNLOCK();
1264 	return 0;
1265     }
1266 
1267     /*
1268      * Determine forwarding vifs from the forwarding cache table
1269      */
1270     MRTSTAT_INC(mrts_mfc_lookups);
1271     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1272 
1273     /* Entry exists, so forward if necessary */
1274     if (rt != NULL) {
1275 	error = ip_mdq(m, ifp, rt, -1);
1276 	MFC_UNLOCK();
1277 	VIF_UNLOCK();
1278 	return error;
1279     } else {
1280 	/*
1281 	 * If we don't have a route for packet's origin,
1282 	 * Make a copy of the packet & send message to routing daemon
1283 	 */
1284 
1285 	struct mbuf *mb0;
1286 	struct rtdetq *rte;
1287 	u_long hash;
1288 	int hlen = ip->ip_hl << 2;
1289 
1290 	MRTSTAT_INC(mrts_mfc_misses);
1291 	MRTSTAT_INC(mrts_no_route);
1292 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1293 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1294 
1295 	/*
1296 	 * Allocate mbufs early so that we don't do extra work if we are
1297 	 * just going to fail anyway.  Make sure to pullup the header so
1298 	 * that other people can't step on it.
1299 	 */
1300 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1301 	    M_NOWAIT|M_ZERO);
1302 	if (rte == NULL) {
1303 	    MFC_UNLOCK();
1304 	    VIF_UNLOCK();
1305 	    return ENOBUFS;
1306 	}
1307 
1308 	mb0 = m_copypacket(m, M_NOWAIT);
1309 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1310 	    mb0 = m_pullup(mb0, hlen);
1311 	if (mb0 == NULL) {
1312 	    free(rte, M_MRTABLE);
1313 	    MFC_UNLOCK();
1314 	    VIF_UNLOCK();
1315 	    return ENOBUFS;
1316 	}
1317 
1318 	/* is there an upcall waiting for this flow ? */
1319 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1320 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1321 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1322 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1323 		    !TAILQ_EMPTY(&rt->mfc_stall))
1324 			break;
1325 	}
1326 
1327 	if (rt == NULL) {
1328 	    int i;
1329 	    struct igmpmsg *im;
1330 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1331 	    struct mbuf *mm;
1332 
1333 	    /*
1334 	     * Locate the vifi for the incoming interface for this packet.
1335 	     * If none found, drop packet.
1336 	     */
1337 	    for (vifi = 0; vifi < V_numvifs &&
1338 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1339 		;
1340 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1341 		goto non_fatal;
1342 
1343 	    /* no upcall, so make a new entry */
1344 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1345 	    if (rt == NULL)
1346 		goto fail;
1347 
1348 	    /* Make a copy of the header to send to the user level process */
1349 	    mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1350 	    if (mm == NULL)
1351 		goto fail1;
1352 
1353 	    /*
1354 	     * Send message to routing daemon to install
1355 	     * a route into the kernel table
1356 	     */
1357 
1358 	    im = mtod(mm, struct igmpmsg *);
1359 	    im->im_msgtype = IGMPMSG_NOCACHE;
1360 	    im->im_mbz = 0;
1361 	    im->im_vif = vifi;
1362 
1363 	    MRTSTAT_INC(mrts_upcalls);
1364 
1365 	    k_igmpsrc.sin_addr = ip->ip_src;
1366 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1367 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1368 		MRTSTAT_INC(mrts_upq_sockfull);
1369 fail1:
1370 		free(rt, M_MRTABLE);
1371 fail:
1372 		free(rte, M_MRTABLE);
1373 		m_freem(mb0);
1374 		MFC_UNLOCK();
1375 		VIF_UNLOCK();
1376 		return ENOBUFS;
1377 	    }
1378 
1379 	    /* insert new entry at head of hash chain */
1380 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1381 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1382 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1383 	    V_nexpire[hash]++;
1384 	    for (i = 0; i < V_numvifs; i++) {
1385 		rt->mfc_ttls[i] = 0;
1386 		rt->mfc_flags[i] = 0;
1387 	    }
1388 	    rt->mfc_parent = -1;
1389 
1390 	    /* clear the RP address */
1391 	    rt->mfc_rp.s_addr = INADDR_ANY;
1392 	    rt->mfc_bw_meter = NULL;
1393 
1394 	    /* initialize pkt counters per src-grp */
1395 	    rt->mfc_pkt_cnt = 0;
1396 	    rt->mfc_byte_cnt = 0;
1397 	    rt->mfc_wrong_if = 0;
1398 	    timevalclear(&rt->mfc_last_assert);
1399 
1400 	    TAILQ_INIT(&rt->mfc_stall);
1401 	    rt->mfc_nstall = 0;
1402 
1403 	    /* link into table */
1404 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1405 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1406 	    rt->mfc_nstall++;
1407 
1408 	} else {
1409 	    /* determine if queue has overflowed */
1410 	    if (rt->mfc_nstall > MAX_UPQ) {
1411 		MRTSTAT_INC(mrts_upq_ovflw);
1412 non_fatal:
1413 		free(rte, M_MRTABLE);
1414 		m_freem(mb0);
1415 		MFC_UNLOCK();
1416 		VIF_UNLOCK();
1417 		return (0);
1418 	    }
1419 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1420 	    rt->mfc_nstall++;
1421 	}
1422 
1423 	rte->m			= mb0;
1424 	rte->ifp		= ifp;
1425 
1426 	MFC_UNLOCK();
1427 	VIF_UNLOCK();
1428 
1429 	return 0;
1430     }
1431 }
1432 
1433 /*
1434  * Clean up the cache entry if upcall is not serviced
1435  */
1436 static void
1437 expire_upcalls(void *arg)
1438 {
1439     u_long i;
1440 
1441     CURVNET_SET((struct vnet *) arg);
1442 
1443     MFC_LOCK();
1444 
1445     for (i = 0; i < mfchashsize; i++) {
1446 	struct mfc *rt, *nrt;
1447 
1448 	if (V_nexpire[i] == 0)
1449 	    continue;
1450 
1451 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1452 		if (TAILQ_EMPTY(&rt->mfc_stall))
1453 			continue;
1454 
1455 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1456 			continue;
1457 
1458 		/*
1459 		 * free the bw_meter entries
1460 		 */
1461 		while (rt->mfc_bw_meter != NULL) {
1462 		    struct bw_meter *x = rt->mfc_bw_meter;
1463 
1464 		    rt->mfc_bw_meter = x->bm_mfc_next;
1465 		    free(x, M_BWMETER);
1466 		}
1467 
1468 		MRTSTAT_INC(mrts_cache_cleanups);
1469 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1470 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1471 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1472 
1473 		expire_mfc(rt);
1474 	    }
1475     }
1476 
1477     MFC_UNLOCK();
1478 
1479     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1480 	curvnet);
1481 
1482     CURVNET_RESTORE();
1483 }
1484 
1485 /*
1486  * Packet forwarding routine once entry in the cache is made
1487  */
1488 static int
1489 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1490 {
1491     struct ip  *ip = mtod(m, struct ip *);
1492     vifi_t vifi;
1493     int plen = ntohs(ip->ip_len);
1494 
1495     VIF_LOCK_ASSERT();
1496 
1497     /*
1498      * If xmt_vif is not -1, send on only the requested vif.
1499      *
1500      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1501      */
1502     if (xmt_vif < V_numvifs) {
1503 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1504 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1505 	else
1506 		phyint_send(ip, V_viftable + xmt_vif, m);
1507 	return 1;
1508     }
1509 
1510     /*
1511      * Don't forward if it didn't arrive from the parent vif for its origin.
1512      */
1513     vifi = rt->mfc_parent;
1514     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1515 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1516 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1517 	MRTSTAT_INC(mrts_wrong_if);
1518 	++rt->mfc_wrong_if;
1519 	/*
1520 	 * If we are doing PIM assert processing, send a message
1521 	 * to the routing daemon.
1522 	 *
1523 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1524 	 * can complete the SPT switch, regardless of the type
1525 	 * of the iif (broadcast media, GRE tunnel, etc).
1526 	 */
1527 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1528 	    V_viftable[vifi].v_ifp) {
1529 	    if (ifp == &V_multicast_register_if)
1530 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1531 
1532 	    /* Get vifi for the incoming packet */
1533 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1534 		vifi++)
1535 		;
1536 	    if (vifi >= V_numvifs)
1537 		return 0;	/* The iif is not found: ignore the packet. */
1538 
1539 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1540 		return 0;	/* WRONGVIF disabled: ignore the packet */
1541 
1542 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1543 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1544 		struct igmpmsg *im;
1545 		int hlen = ip->ip_hl << 2;
1546 		struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1547 
1548 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1549 		    mm = m_pullup(mm, hlen);
1550 		if (mm == NULL)
1551 		    return ENOBUFS;
1552 
1553 		im = mtod(mm, struct igmpmsg *);
1554 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1555 		im->im_mbz		= 0;
1556 		im->im_vif		= vifi;
1557 
1558 		MRTSTAT_INC(mrts_upcalls);
1559 
1560 		k_igmpsrc.sin_addr = im->im_src;
1561 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1562 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1563 		    MRTSTAT_INC(mrts_upq_sockfull);
1564 		    return ENOBUFS;
1565 		}
1566 	    }
1567 	}
1568 	return 0;
1569     }
1570 
1571     /* If I sourced this packet, it counts as output, else it was input. */
1572     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1573 	V_viftable[vifi].v_pkt_out++;
1574 	V_viftable[vifi].v_bytes_out += plen;
1575     } else {
1576 	V_viftable[vifi].v_pkt_in++;
1577 	V_viftable[vifi].v_bytes_in += plen;
1578     }
1579     rt->mfc_pkt_cnt++;
1580     rt->mfc_byte_cnt += plen;
1581 
1582     /*
1583      * For each vif, decide if a copy of the packet should be forwarded.
1584      * Forward if:
1585      *		- the ttl exceeds the vif's threshold
1586      *		- there are group members downstream on interface
1587      */
1588     for (vifi = 0; vifi < V_numvifs; vifi++)
1589 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1590 	    V_viftable[vifi].v_pkt_out++;
1591 	    V_viftable[vifi].v_bytes_out += plen;
1592 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1593 		pim_register_send(ip, V_viftable + vifi, m, rt);
1594 	    else
1595 		phyint_send(ip, V_viftable + vifi, m);
1596 	}
1597 
1598     /*
1599      * Perform upcall-related bw measuring.
1600      */
1601     if (rt->mfc_bw_meter != NULL) {
1602 	struct bw_meter *x;
1603 	struct timeval now;
1604 
1605 	microtime(&now);
1606 	MFC_LOCK_ASSERT();
1607 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1608 	    bw_meter_receive_packet(x, plen, &now);
1609     }
1610 
1611     return 0;
1612 }
1613 
1614 /*
1615  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1616  */
1617 static int
1618 X_legal_vif_num(int vif)
1619 {
1620 	int ret;
1621 
1622 	ret = 0;
1623 	if (vif < 0)
1624 		return (ret);
1625 
1626 	VIF_LOCK();
1627 	if (vif < V_numvifs)
1628 		ret = 1;
1629 	VIF_UNLOCK();
1630 
1631 	return (ret);
1632 }
1633 
1634 /*
1635  * Return the local address used by this vif
1636  */
1637 static u_long
1638 X_ip_mcast_src(int vifi)
1639 {
1640 	in_addr_t addr;
1641 
1642 	addr = INADDR_ANY;
1643 	if (vifi < 0)
1644 		return (addr);
1645 
1646 	VIF_LOCK();
1647 	if (vifi < V_numvifs)
1648 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1649 	VIF_UNLOCK();
1650 
1651 	return (addr);
1652 }
1653 
1654 static void
1655 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1656 {
1657     struct mbuf *mb_copy;
1658     int hlen = ip->ip_hl << 2;
1659 
1660     VIF_LOCK_ASSERT();
1661 
1662     /*
1663      * Make a new reference to the packet; make sure that
1664      * the IP header is actually copied, not just referenced,
1665      * so that ip_output() only scribbles on the copy.
1666      */
1667     mb_copy = m_copypacket(m, M_NOWAIT);
1668     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1669 	mb_copy = m_pullup(mb_copy, hlen);
1670     if (mb_copy == NULL)
1671 	return;
1672 
1673     send_packet(vifp, mb_copy);
1674 }
1675 
1676 static void
1677 send_packet(struct vif *vifp, struct mbuf *m)
1678 {
1679 	struct ip_moptions imo;
1680 	int error __unused;
1681 
1682 	VIF_LOCK_ASSERT();
1683 
1684 	imo.imo_multicast_ifp  = vifp->v_ifp;
1685 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1686 	imo.imo_multicast_loop = 1;
1687 	imo.imo_multicast_vif  = -1;
1688 	STAILQ_INIT(&imo.imo_head);
1689 
1690 	/*
1691 	 * Re-entrancy should not be a problem here, because
1692 	 * the packets that we send out and are looped back at us
1693 	 * should get rejected because they appear to come from
1694 	 * the loopback interface, thus preventing looping.
1695 	 */
1696 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1697 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1698 	    (ptrdiff_t)(vifp - V_viftable), error);
1699 }
1700 
1701 /*
1702  * Stubs for old RSVP socket shim implementation.
1703  */
1704 
1705 static int
1706 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1707 {
1708 
1709 	return (EOPNOTSUPP);
1710 }
1711 
1712 static void
1713 X_ip_rsvp_force_done(struct socket *so __unused)
1714 {
1715 
1716 }
1717 
1718 static int
1719 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1720 {
1721 	struct mbuf *m;
1722 
1723 	m = *mp;
1724 	*mp = NULL;
1725 	if (!V_rsvp_on)
1726 		m_freem(m);
1727 	return (IPPROTO_DONE);
1728 }
1729 
1730 /*
1731  * Code for bandwidth monitors
1732  */
1733 
1734 /*
1735  * Define common interface for timeval-related methods
1736  */
1737 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1738 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1739 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1740 
1741 static uint32_t
1742 compute_bw_meter_flags(struct bw_upcall *req)
1743 {
1744     uint32_t flags = 0;
1745 
1746     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1747 	flags |= BW_METER_UNIT_PACKETS;
1748     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1749 	flags |= BW_METER_UNIT_BYTES;
1750     if (req->bu_flags & BW_UPCALL_GEQ)
1751 	flags |= BW_METER_GEQ;
1752     if (req->bu_flags & BW_UPCALL_LEQ)
1753 	flags |= BW_METER_LEQ;
1754 
1755     return flags;
1756 }
1757 
1758 /*
1759  * Add a bw_meter entry
1760  */
1761 static int
1762 add_bw_upcall(struct bw_upcall *req)
1763 {
1764     struct mfc *mfc;
1765     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1766 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1767     struct timeval now;
1768     struct bw_meter *x;
1769     uint32_t flags;
1770 
1771     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1772 	return EOPNOTSUPP;
1773 
1774     /* Test if the flags are valid */
1775     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1776 	return EINVAL;
1777     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1778 	return EINVAL;
1779     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1780 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1781 	return EINVAL;
1782 
1783     /* Test if the threshold time interval is valid */
1784     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1785 	return EINVAL;
1786 
1787     flags = compute_bw_meter_flags(req);
1788 
1789     /*
1790      * Find if we have already same bw_meter entry
1791      */
1792     MFC_LOCK();
1793     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1794     if (mfc == NULL) {
1795 	MFC_UNLOCK();
1796 	return EADDRNOTAVAIL;
1797     }
1798     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1799 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1800 			   &req->bu_threshold.b_time, ==)) &&
1801 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1802 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1803 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1804 	    MFC_UNLOCK();
1805 	    return 0;		/* XXX Already installed */
1806 	}
1807     }
1808 
1809     /* Allocate the new bw_meter entry */
1810     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1811     if (x == NULL) {
1812 	MFC_UNLOCK();
1813 	return ENOBUFS;
1814     }
1815 
1816     /* Set the new bw_meter entry */
1817     x->bm_threshold.b_time = req->bu_threshold.b_time;
1818     microtime(&now);
1819     x->bm_start_time = now;
1820     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1821     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1822     x->bm_measured.b_packets = 0;
1823     x->bm_measured.b_bytes = 0;
1824     x->bm_flags = flags;
1825     x->bm_time_next = NULL;
1826     x->bm_time_hash = BW_METER_BUCKETS;
1827 
1828     /* Add the new bw_meter entry to the front of entries for this MFC */
1829     x->bm_mfc = mfc;
1830     x->bm_mfc_next = mfc->mfc_bw_meter;
1831     mfc->mfc_bw_meter = x;
1832     schedule_bw_meter(x, &now);
1833     MFC_UNLOCK();
1834 
1835     return 0;
1836 }
1837 
1838 static void
1839 free_bw_list(struct bw_meter *list)
1840 {
1841     while (list != NULL) {
1842 	struct bw_meter *x = list;
1843 
1844 	list = list->bm_mfc_next;
1845 	unschedule_bw_meter(x);
1846 	free(x, M_BWMETER);
1847     }
1848 }
1849 
1850 /*
1851  * Delete one or multiple bw_meter entries
1852  */
1853 static int
1854 del_bw_upcall(struct bw_upcall *req)
1855 {
1856     struct mfc *mfc;
1857     struct bw_meter *x;
1858 
1859     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1860 	return EOPNOTSUPP;
1861 
1862     MFC_LOCK();
1863 
1864     /* Find the corresponding MFC entry */
1865     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1866     if (mfc == NULL) {
1867 	MFC_UNLOCK();
1868 	return EADDRNOTAVAIL;
1869     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1870 	/*
1871 	 * Delete all bw_meter entries for this mfc
1872 	 */
1873 	struct bw_meter *list;
1874 
1875 	list = mfc->mfc_bw_meter;
1876 	mfc->mfc_bw_meter = NULL;
1877 	free_bw_list(list);
1878 	MFC_UNLOCK();
1879 	return 0;
1880     } else {			/* Delete a single bw_meter entry */
1881 	struct bw_meter *prev;
1882 	uint32_t flags = 0;
1883 
1884 	flags = compute_bw_meter_flags(req);
1885 
1886 	/* Find the bw_meter entry to delete */
1887 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1888 	     prev = x, x = x->bm_mfc_next) {
1889 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1890 			       &req->bu_threshold.b_time, ==)) &&
1891 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1892 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1893 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1894 		break;
1895 	}
1896 	if (x != NULL) { /* Delete entry from the list for this MFC */
1897 	    if (prev != NULL)
1898 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1899 	    else
1900 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1901 
1902 	    unschedule_bw_meter(x);
1903 	    MFC_UNLOCK();
1904 	    /* Free the bw_meter entry */
1905 	    free(x, M_BWMETER);
1906 	    return 0;
1907 	} else {
1908 	    MFC_UNLOCK();
1909 	    return EINVAL;
1910 	}
1911     }
1912     /* NOTREACHED */
1913 }
1914 
1915 /*
1916  * Perform bandwidth measurement processing that may result in an upcall
1917  */
1918 static void
1919 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1920 {
1921     struct timeval delta;
1922 
1923     MFC_LOCK_ASSERT();
1924 
1925     delta = *nowp;
1926     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1927 
1928     if (x->bm_flags & BW_METER_GEQ) {
1929 	/*
1930 	 * Processing for ">=" type of bw_meter entry
1931 	 */
1932 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1933 	    /* Reset the bw_meter entry */
1934 	    x->bm_start_time = *nowp;
1935 	    x->bm_measured.b_packets = 0;
1936 	    x->bm_measured.b_bytes = 0;
1937 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1938 	}
1939 
1940 	/* Record that a packet is received */
1941 	x->bm_measured.b_packets++;
1942 	x->bm_measured.b_bytes += plen;
1943 
1944 	/*
1945 	 * Test if we should deliver an upcall
1946 	 */
1947 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1948 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1949 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1950 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1951 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1952 		/* Prepare an upcall for delivery */
1953 		bw_meter_prepare_upcall(x, nowp);
1954 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1955 	    }
1956 	}
1957     } else if (x->bm_flags & BW_METER_LEQ) {
1958 	/*
1959 	 * Processing for "<=" type of bw_meter entry
1960 	 */
1961 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1962 	    /*
1963 	     * We are behind time with the multicast forwarding table
1964 	     * scanning for "<=" type of bw_meter entries, so test now
1965 	     * if we should deliver an upcall.
1966 	     */
1967 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1968 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1969 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1970 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1971 		/* Prepare an upcall for delivery */
1972 		bw_meter_prepare_upcall(x, nowp);
1973 	    }
1974 	    /* Reschedule the bw_meter entry */
1975 	    unschedule_bw_meter(x);
1976 	    schedule_bw_meter(x, nowp);
1977 	}
1978 
1979 	/* Record that a packet is received */
1980 	x->bm_measured.b_packets++;
1981 	x->bm_measured.b_bytes += plen;
1982 
1983 	/*
1984 	 * Test if we should restart the measuring interval
1985 	 */
1986 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1987 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1988 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1989 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1990 	    /* Don't restart the measuring interval */
1991 	} else {
1992 	    /* Do restart the measuring interval */
1993 	    /*
1994 	     * XXX: note that we don't unschedule and schedule, because this
1995 	     * might be too much overhead per packet. Instead, when we process
1996 	     * all entries for a given timer hash bin, we check whether it is
1997 	     * really a timeout. If not, we reschedule at that time.
1998 	     */
1999 	    x->bm_start_time = *nowp;
2000 	    x->bm_measured.b_packets = 0;
2001 	    x->bm_measured.b_bytes = 0;
2002 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2003 	}
2004     }
2005 }
2006 
2007 /*
2008  * Prepare a bandwidth-related upcall
2009  */
2010 static void
2011 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2012 {
2013     struct timeval delta;
2014     struct bw_upcall *u;
2015 
2016     MFC_LOCK_ASSERT();
2017 
2018     /*
2019      * Compute the measured time interval
2020      */
2021     delta = *nowp;
2022     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2023 
2024     /*
2025      * If there are too many pending upcalls, deliver them now
2026      */
2027     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2028 	bw_upcalls_send();
2029 
2030     /*
2031      * Set the bw_upcall entry
2032      */
2033     u = &V_bw_upcalls[V_bw_upcalls_n++];
2034     u->bu_src = x->bm_mfc->mfc_origin;
2035     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2036     u->bu_threshold.b_time = x->bm_threshold.b_time;
2037     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2038     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2039     u->bu_measured.b_time = delta;
2040     u->bu_measured.b_packets = x->bm_measured.b_packets;
2041     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2042     u->bu_flags = 0;
2043     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2044 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2045     if (x->bm_flags & BW_METER_UNIT_BYTES)
2046 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2047     if (x->bm_flags & BW_METER_GEQ)
2048 	u->bu_flags |= BW_UPCALL_GEQ;
2049     if (x->bm_flags & BW_METER_LEQ)
2050 	u->bu_flags |= BW_UPCALL_LEQ;
2051 }
2052 
2053 /*
2054  * Send the pending bandwidth-related upcalls
2055  */
2056 static void
2057 bw_upcalls_send(void)
2058 {
2059     struct mbuf *m;
2060     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2061     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2062     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2063 				      0,		/* unused2 */
2064 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2065 				      0,		/* im_mbz  */
2066 				      0,		/* im_vif  */
2067 				      0,		/* unused3 */
2068 				      { 0 },		/* im_src  */
2069 				      { 0 } };		/* im_dst  */
2070 
2071     MFC_LOCK_ASSERT();
2072 
2073     if (V_bw_upcalls_n == 0)
2074 	return;			/* No pending upcalls */
2075 
2076     V_bw_upcalls_n = 0;
2077 
2078     /*
2079      * Allocate a new mbuf, initialize it with the header and
2080      * the payload for the pending calls.
2081      */
2082     m = m_gethdr(M_NOWAIT, MT_DATA);
2083     if (m == NULL) {
2084 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2085 	return;
2086     }
2087 
2088     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2089     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2090 
2091     /*
2092      * Send the upcalls
2093      * XXX do we need to set the address in k_igmpsrc ?
2094      */
2095     MRTSTAT_INC(mrts_upcalls);
2096     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2097 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2098 	MRTSTAT_INC(mrts_upq_sockfull);
2099     }
2100 }
2101 
2102 /*
2103  * Compute the timeout hash value for the bw_meter entries
2104  */
2105 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2106     do {								\
2107 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2108 									\
2109 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2110 	(hash) = next_timeval.tv_sec;					\
2111 	if (next_timeval.tv_usec)					\
2112 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2113 	(hash) %= BW_METER_BUCKETS;					\
2114     } while (0)
2115 
2116 /*
2117  * Schedule a timer to process periodically bw_meter entry of type "<="
2118  * by linking the entry in the proper hash bucket.
2119  */
2120 static void
2121 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2122 {
2123     int time_hash;
2124 
2125     MFC_LOCK_ASSERT();
2126 
2127     if (!(x->bm_flags & BW_METER_LEQ))
2128 	return;		/* XXX: we schedule timers only for "<=" entries */
2129 
2130     /*
2131      * Reset the bw_meter entry
2132      */
2133     x->bm_start_time = *nowp;
2134     x->bm_measured.b_packets = 0;
2135     x->bm_measured.b_bytes = 0;
2136     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2137 
2138     /*
2139      * Compute the timeout hash value and insert the entry
2140      */
2141     BW_METER_TIMEHASH(x, time_hash);
2142     x->bm_time_next = V_bw_meter_timers[time_hash];
2143     V_bw_meter_timers[time_hash] = x;
2144     x->bm_time_hash = time_hash;
2145 }
2146 
2147 /*
2148  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2149  * by removing the entry from the proper hash bucket.
2150  */
2151 static void
2152 unschedule_bw_meter(struct bw_meter *x)
2153 {
2154     int time_hash;
2155     struct bw_meter *prev, *tmp;
2156 
2157     MFC_LOCK_ASSERT();
2158 
2159     if (!(x->bm_flags & BW_METER_LEQ))
2160 	return;		/* XXX: we schedule timers only for "<=" entries */
2161 
2162     /*
2163      * Compute the timeout hash value and delete the entry
2164      */
2165     time_hash = x->bm_time_hash;
2166     if (time_hash >= BW_METER_BUCKETS)
2167 	return;		/* Entry was not scheduled */
2168 
2169     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2170 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2171 	if (tmp == x)
2172 	    break;
2173 
2174     if (tmp == NULL)
2175 	panic("unschedule_bw_meter: bw_meter entry not found");
2176 
2177     if (prev != NULL)
2178 	prev->bm_time_next = x->bm_time_next;
2179     else
2180 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2181 
2182     x->bm_time_next = NULL;
2183     x->bm_time_hash = BW_METER_BUCKETS;
2184 }
2185 
2186 /*
2187  * Process all "<=" type of bw_meter that should be processed now,
2188  * and for each entry prepare an upcall if necessary. Each processed
2189  * entry is rescheduled again for the (periodic) processing.
2190  *
2191  * This is run periodically (once per second normally). On each round,
2192  * all the potentially matching entries are in the hash slot that we are
2193  * looking at.
2194  */
2195 static void
2196 bw_meter_process()
2197 {
2198     uint32_t loops;
2199     int i;
2200     struct timeval now, process_endtime;
2201 
2202     microtime(&now);
2203     if (V_last_tv_sec == now.tv_sec)
2204 	return;		/* nothing to do */
2205 
2206     loops = now.tv_sec - V_last_tv_sec;
2207     V_last_tv_sec = now.tv_sec;
2208     if (loops > BW_METER_BUCKETS)
2209 	loops = BW_METER_BUCKETS;
2210 
2211     MFC_LOCK();
2212     /*
2213      * Process all bins of bw_meter entries from the one after the last
2214      * processed to the current one. On entry, i points to the last bucket
2215      * visited, so we need to increment i at the beginning of the loop.
2216      */
2217     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2218 	struct bw_meter *x, *tmp_list;
2219 
2220 	if (++i >= BW_METER_BUCKETS)
2221 	    i = 0;
2222 
2223 	/* Disconnect the list of bw_meter entries from the bin */
2224 	tmp_list = V_bw_meter_timers[i];
2225 	V_bw_meter_timers[i] = NULL;
2226 
2227 	/* Process the list of bw_meter entries */
2228 	while (tmp_list != NULL) {
2229 	    x = tmp_list;
2230 	    tmp_list = tmp_list->bm_time_next;
2231 
2232 	    /* Test if the time interval is over */
2233 	    process_endtime = x->bm_start_time;
2234 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2235 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2236 		/* Not yet: reschedule, but don't reset */
2237 		int time_hash;
2238 
2239 		BW_METER_TIMEHASH(x, time_hash);
2240 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2241 		    /*
2242 		     * XXX: somehow the bin processing is a bit ahead of time.
2243 		     * Put the entry in the next bin.
2244 		     */
2245 		    if (++time_hash >= BW_METER_BUCKETS)
2246 			time_hash = 0;
2247 		}
2248 		x->bm_time_next = V_bw_meter_timers[time_hash];
2249 		V_bw_meter_timers[time_hash] = x;
2250 		x->bm_time_hash = time_hash;
2251 
2252 		continue;
2253 	    }
2254 
2255 	    /*
2256 	     * Test if we should deliver an upcall
2257 	     */
2258 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2259 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2260 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2261 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2262 		/* Prepare an upcall for delivery */
2263 		bw_meter_prepare_upcall(x, &now);
2264 	    }
2265 
2266 	    /*
2267 	     * Reschedule for next processing
2268 	     */
2269 	    schedule_bw_meter(x, &now);
2270 	}
2271     }
2272 
2273     /* Send all upcalls that are pending delivery */
2274     bw_upcalls_send();
2275 
2276     MFC_UNLOCK();
2277 }
2278 
2279 /*
2280  * A periodic function for sending all upcalls that are pending delivery
2281  */
2282 static void
2283 expire_bw_upcalls_send(void *arg)
2284 {
2285     CURVNET_SET((struct vnet *) arg);
2286 
2287     MFC_LOCK();
2288     bw_upcalls_send();
2289     MFC_UNLOCK();
2290 
2291     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2292 	curvnet);
2293     CURVNET_RESTORE();
2294 }
2295 
2296 /*
2297  * A periodic function for periodic scanning of the multicast forwarding
2298  * table for processing all "<=" bw_meter entries.
2299  */
2300 static void
2301 expire_bw_meter_process(void *arg)
2302 {
2303     CURVNET_SET((struct vnet *) arg);
2304 
2305     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2306 	bw_meter_process();
2307 
2308     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2309 	curvnet);
2310     CURVNET_RESTORE();
2311 }
2312 
2313 /*
2314  * End of bandwidth monitoring code
2315  */
2316 
2317 /*
2318  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2319  *
2320  */
2321 static int
2322 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2323     struct mfc *rt)
2324 {
2325     struct mbuf *mb_copy, *mm;
2326 
2327     /*
2328      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2329      * rendezvous point was unspecified, and we were told not to.
2330      */
2331     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2332 	in_nullhost(rt->mfc_rp))
2333 	return 0;
2334 
2335     mb_copy = pim_register_prepare(ip, m);
2336     if (mb_copy == NULL)
2337 	return ENOBUFS;
2338 
2339     /*
2340      * Send all the fragments. Note that the mbuf for each fragment
2341      * is freed by the sending machinery.
2342      */
2343     for (mm = mb_copy; mm; mm = mb_copy) {
2344 	mb_copy = mm->m_nextpkt;
2345 	mm->m_nextpkt = 0;
2346 	mm = m_pullup(mm, sizeof(struct ip));
2347 	if (mm != NULL) {
2348 	    ip = mtod(mm, struct ip *);
2349 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2350 		pim_register_send_rp(ip, vifp, mm, rt);
2351 	    } else {
2352 		pim_register_send_upcall(ip, vifp, mm, rt);
2353 	    }
2354 	}
2355     }
2356 
2357     return 0;
2358 }
2359 
2360 /*
2361  * Return a copy of the data packet that is ready for PIM Register
2362  * encapsulation.
2363  * XXX: Note that in the returned copy the IP header is a valid one.
2364  */
2365 static struct mbuf *
2366 pim_register_prepare(struct ip *ip, struct mbuf *m)
2367 {
2368     struct mbuf *mb_copy = NULL;
2369     int mtu;
2370 
2371     /* Take care of delayed checksums */
2372     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2373 	in_delayed_cksum(m);
2374 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2375     }
2376 
2377     /*
2378      * Copy the old packet & pullup its IP header into the
2379      * new mbuf so we can modify it.
2380      */
2381     mb_copy = m_copypacket(m, M_NOWAIT);
2382     if (mb_copy == NULL)
2383 	return NULL;
2384     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2385     if (mb_copy == NULL)
2386 	return NULL;
2387 
2388     /* take care of the TTL */
2389     ip = mtod(mb_copy, struct ip *);
2390     --ip->ip_ttl;
2391 
2392     /* Compute the MTU after the PIM Register encapsulation */
2393     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2394 
2395     if (ntohs(ip->ip_len) <= mtu) {
2396 	/* Turn the IP header into a valid one */
2397 	ip->ip_sum = 0;
2398 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2399     } else {
2400 	/* Fragment the packet */
2401 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2402 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2403 	    m_freem(mb_copy);
2404 	    return NULL;
2405 	}
2406     }
2407     return mb_copy;
2408 }
2409 
2410 /*
2411  * Send an upcall with the data packet to the user-level process.
2412  */
2413 static int
2414 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2415     struct mbuf *mb_copy, struct mfc *rt)
2416 {
2417     struct mbuf *mb_first;
2418     int len = ntohs(ip->ip_len);
2419     struct igmpmsg *im;
2420     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2421 
2422     VIF_LOCK_ASSERT();
2423 
2424     /*
2425      * Add a new mbuf with an upcall header
2426      */
2427     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2428     if (mb_first == NULL) {
2429 	m_freem(mb_copy);
2430 	return ENOBUFS;
2431     }
2432     mb_first->m_data += max_linkhdr;
2433     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2434     mb_first->m_len = sizeof(struct igmpmsg);
2435     mb_first->m_next = mb_copy;
2436 
2437     /* Send message to routing daemon */
2438     im = mtod(mb_first, struct igmpmsg *);
2439     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2440     im->im_mbz		= 0;
2441     im->im_vif		= vifp - V_viftable;
2442     im->im_src		= ip->ip_src;
2443     im->im_dst		= ip->ip_dst;
2444 
2445     k_igmpsrc.sin_addr	= ip->ip_src;
2446 
2447     MRTSTAT_INC(mrts_upcalls);
2448 
2449     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2450 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2451 	MRTSTAT_INC(mrts_upq_sockfull);
2452 	return ENOBUFS;
2453     }
2454 
2455     /* Keep statistics */
2456     PIMSTAT_INC(pims_snd_registers_msgs);
2457     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2458 
2459     return 0;
2460 }
2461 
2462 /*
2463  * Encapsulate the data packet in PIM Register message and send it to the RP.
2464  */
2465 static int
2466 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2467     struct mfc *rt)
2468 {
2469     struct mbuf *mb_first;
2470     struct ip *ip_outer;
2471     struct pim_encap_pimhdr *pimhdr;
2472     int len = ntohs(ip->ip_len);
2473     vifi_t vifi = rt->mfc_parent;
2474 
2475     VIF_LOCK_ASSERT();
2476 
2477     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2478 	m_freem(mb_copy);
2479 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2480     }
2481 
2482     /*
2483      * Add a new mbuf with the encapsulating header
2484      */
2485     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2486     if (mb_first == NULL) {
2487 	m_freem(mb_copy);
2488 	return ENOBUFS;
2489     }
2490     mb_first->m_data += max_linkhdr;
2491     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2492     mb_first->m_next = mb_copy;
2493 
2494     mb_first->m_pkthdr.len = len + mb_first->m_len;
2495 
2496     /*
2497      * Fill in the encapsulating IP and PIM header
2498      */
2499     ip_outer = mtod(mb_first, struct ip *);
2500     *ip_outer = pim_encap_iphdr;
2501     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2502 	sizeof(pim_encap_pimhdr));
2503     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2504     ip_outer->ip_dst = rt->mfc_rp;
2505     /*
2506      * Copy the inner header TOS to the outer header, and take care of the
2507      * IP_DF bit.
2508      */
2509     ip_outer->ip_tos = ip->ip_tos;
2510     if (ip->ip_off & htons(IP_DF))
2511 	ip_outer->ip_off |= htons(IP_DF);
2512     ip_fillid(ip_outer);
2513     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2514 					 + sizeof(pim_encap_iphdr));
2515     *pimhdr = pim_encap_pimhdr;
2516     /* If the iif crosses a border, set the Border-bit */
2517     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2518 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2519 
2520     mb_first->m_data += sizeof(pim_encap_iphdr);
2521     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2522     mb_first->m_data -= sizeof(pim_encap_iphdr);
2523 
2524     send_packet(vifp, mb_first);
2525 
2526     /* Keep statistics */
2527     PIMSTAT_INC(pims_snd_registers_msgs);
2528     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2529 
2530     return 0;
2531 }
2532 
2533 /*
2534  * pim_encapcheck() is called by the encap4_input() path at runtime to
2535  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2536  * into the kernel.
2537  */
2538 static int
2539 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2540     int proto __unused, void *arg __unused)
2541 {
2542 
2543     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2544     return (8);		/* claim the datagram. */
2545 }
2546 
2547 /*
2548  * PIM-SMv2 and PIM-DM messages processing.
2549  * Receives and verifies the PIM control messages, and passes them
2550  * up to the listening socket, using rip_input().
2551  * The only message with special processing is the PIM_REGISTER message
2552  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2553  * is passed to if_simloop().
2554  */
2555 static int
2556 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2557 {
2558     struct ip *ip = mtod(m, struct ip *);
2559     struct pim *pim;
2560     int iphlen = off;
2561     int minlen;
2562     int datalen = ntohs(ip->ip_len) - iphlen;
2563     int ip_tos;
2564 
2565     /* Keep statistics */
2566     PIMSTAT_INC(pims_rcv_total_msgs);
2567     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2568 
2569     /*
2570      * Validate lengths
2571      */
2572     if (datalen < PIM_MINLEN) {
2573 	PIMSTAT_INC(pims_rcv_tooshort);
2574 	CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2575 	    __func__, datalen, ntohl(ip->ip_src.s_addr));
2576 	m_freem(m);
2577 	return (IPPROTO_DONE);
2578     }
2579 
2580     /*
2581      * If the packet is at least as big as a REGISTER, go agead
2582      * and grab the PIM REGISTER header size, to avoid another
2583      * possible m_pullup() later.
2584      *
2585      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2586      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2587      */
2588     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2589     /*
2590      * Get the IP and PIM headers in contiguous memory, and
2591      * possibly the PIM REGISTER header.
2592      */
2593     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2594 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2595 	return (IPPROTO_DONE);
2596     }
2597 
2598     /* m_pullup() may have given us a new mbuf so reset ip. */
2599     ip = mtod(m, struct ip *);
2600     ip_tos = ip->ip_tos;
2601 
2602     /* adjust mbuf to point to the PIM header */
2603     m->m_data += iphlen;
2604     m->m_len  -= iphlen;
2605     pim = mtod(m, struct pim *);
2606 
2607     /*
2608      * Validate checksum. If PIM REGISTER, exclude the data packet.
2609      *
2610      * XXX: some older PIMv2 implementations don't make this distinction,
2611      * so for compatibility reason perform the checksum over part of the
2612      * message, and if error, then over the whole message.
2613      */
2614     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2615 	/* do nothing, checksum okay */
2616     } else if (in_cksum(m, datalen)) {
2617 	PIMSTAT_INC(pims_rcv_badsum);
2618 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2619 	m_freem(m);
2620 	return (IPPROTO_DONE);
2621     }
2622 
2623     /* PIM version check */
2624     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2625 	PIMSTAT_INC(pims_rcv_badversion);
2626 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2627 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2628 	m_freem(m);
2629 	return (IPPROTO_DONE);
2630     }
2631 
2632     /* restore mbuf back to the outer IP */
2633     m->m_data -= iphlen;
2634     m->m_len  += iphlen;
2635 
2636     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2637 	/*
2638 	 * Since this is a REGISTER, we'll make a copy of the register
2639 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2640 	 * routing daemon.
2641 	 */
2642 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2643 	struct mbuf *mcp;
2644 	struct ip *encap_ip;
2645 	u_int32_t *reghdr;
2646 	struct ifnet *vifp;
2647 
2648 	VIF_LOCK();
2649 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2650 	    VIF_UNLOCK();
2651 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2652 		(int)V_reg_vif_num);
2653 	    m_freem(m);
2654 	    return (IPPROTO_DONE);
2655 	}
2656 	/* XXX need refcnt? */
2657 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2658 	VIF_UNLOCK();
2659 
2660 	/*
2661 	 * Validate length
2662 	 */
2663 	if (datalen < PIM_REG_MINLEN) {
2664 	    PIMSTAT_INC(pims_rcv_tooshort);
2665 	    PIMSTAT_INC(pims_rcv_badregisters);
2666 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2667 	    m_freem(m);
2668 	    return (IPPROTO_DONE);
2669 	}
2670 
2671 	reghdr = (u_int32_t *)(pim + 1);
2672 	encap_ip = (struct ip *)(reghdr + 1);
2673 
2674 	CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2675 	    __func__, ntohl(encap_ip->ip_src.s_addr),
2676 	    ntohs(encap_ip->ip_len));
2677 
2678 	/* verify the version number of the inner packet */
2679 	if (encap_ip->ip_v != IPVERSION) {
2680 	    PIMSTAT_INC(pims_rcv_badregisters);
2681 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2682 	    m_freem(m);
2683 	    return (IPPROTO_DONE);
2684 	}
2685 
2686 	/* verify the inner packet is destined to a mcast group */
2687 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2688 	    PIMSTAT_INC(pims_rcv_badregisters);
2689 	    CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2690 		ntohl(encap_ip->ip_dst.s_addr));
2691 	    m_freem(m);
2692 	    return (IPPROTO_DONE);
2693 	}
2694 
2695 	/* If a NULL_REGISTER, pass it to the daemon */
2696 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2697 	    goto pim_input_to_daemon;
2698 
2699 	/*
2700 	 * Copy the TOS from the outer IP header to the inner IP header.
2701 	 */
2702 	if (encap_ip->ip_tos != ip_tos) {
2703 	    /* Outer TOS -> inner TOS */
2704 	    encap_ip->ip_tos = ip_tos;
2705 	    /* Recompute the inner header checksum. Sigh... */
2706 
2707 	    /* adjust mbuf to point to the inner IP header */
2708 	    m->m_data += (iphlen + PIM_MINLEN);
2709 	    m->m_len  -= (iphlen + PIM_MINLEN);
2710 
2711 	    encap_ip->ip_sum = 0;
2712 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2713 
2714 	    /* restore mbuf to point back to the outer IP header */
2715 	    m->m_data -= (iphlen + PIM_MINLEN);
2716 	    m->m_len  += (iphlen + PIM_MINLEN);
2717 	}
2718 
2719 	/*
2720 	 * Decapsulate the inner IP packet and loopback to forward it
2721 	 * as a normal multicast packet. Also, make a copy of the
2722 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2723 	 * to pass to the daemon later, so it can take the appropriate
2724 	 * actions (e.g., send back PIM_REGISTER_STOP).
2725 	 * XXX: here m->m_data points to the outer IP header.
2726 	 */
2727 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2728 	if (mcp == NULL) {
2729 	    CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2730 	    m_freem(m);
2731 	    return (IPPROTO_DONE);
2732 	}
2733 
2734 	/* Keep statistics */
2735 	/* XXX: registers_bytes include only the encap. mcast pkt */
2736 	PIMSTAT_INC(pims_rcv_registers_msgs);
2737 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2738 
2739 	/*
2740 	 * forward the inner ip packet; point m_data at the inner ip.
2741 	 */
2742 	m_adj(m, iphlen + PIM_MINLEN);
2743 
2744 	CTR4(KTR_IPMF,
2745 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2746 	    __func__,
2747 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2748 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2749 	    (int)V_reg_vif_num);
2750 
2751 	/* NB: vifp was collected above; can it change on us? */
2752 	if_simloop(vifp, m, dst.sin_family, 0);
2753 
2754 	/* prepare the register head to send to the mrouting daemon */
2755 	m = mcp;
2756     }
2757 
2758 pim_input_to_daemon:
2759     /*
2760      * Pass the PIM message up to the daemon; if it is a Register message,
2761      * pass the 'head' only up to the daemon. This includes the
2762      * outer IP header, PIM header, PIM-Register header and the
2763      * inner IP header.
2764      * XXX: the outer IP header pkt size of a Register is not adjust to
2765      * reflect the fact that the inner multicast data is truncated.
2766      */
2767     return (rip_input(&m, &off, proto));
2768 }
2769 
2770 static int
2771 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2772 {
2773 	struct mfc	*rt;
2774 	int		 error, i;
2775 
2776 	if (req->newptr)
2777 		return (EPERM);
2778 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2779 		return (0);
2780 	error = sysctl_wire_old_buffer(req, 0);
2781 	if (error)
2782 		return (error);
2783 
2784 	MFC_LOCK();
2785 	for (i = 0; i < mfchashsize; i++) {
2786 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2787 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2788 			if (error)
2789 				goto out_locked;
2790 		}
2791 	}
2792 out_locked:
2793 	MFC_UNLOCK();
2794 	return (error);
2795 }
2796 
2797 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2798     CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2799     "IPv4 Multicast Forwarding Table "
2800     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2801 
2802 static int
2803 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2804 {
2805 	int error;
2806 
2807 	if (req->newptr)
2808 		return (EPERM);
2809 	if (V_viftable == NULL)		/* XXX unlocked */
2810 		return (0);
2811 	error = sysctl_wire_old_buffer(req, sizeof(*V_viftable) * MAXVIFS);
2812 	if (error)
2813 		return (error);
2814 
2815 	VIF_LOCK();
2816 	error = SYSCTL_OUT(req, V_viftable, sizeof(*V_viftable) * MAXVIFS);
2817 	VIF_UNLOCK();
2818 	return (error);
2819 }
2820 
2821 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2822     CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2823     sysctl_viflist, "S,vif[MAXVIFS]",
2824     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2825 
2826 static void
2827 vnet_mroute_init(const void *unused __unused)
2828 {
2829 
2830 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2831 
2832 	V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2833 	    M_MRTABLE, M_WAITOK|M_ZERO);
2834 	V_bw_meter_timers = mallocarray(BW_METER_BUCKETS,
2835 	    sizeof(*V_bw_meter_timers), M_MRTABLE, M_WAITOK|M_ZERO);
2836 	V_bw_upcalls = mallocarray(BW_UPCALLS_MAX, sizeof(*V_bw_upcalls),
2837 	    M_MRTABLE, M_WAITOK|M_ZERO);
2838 
2839 	callout_init(&V_expire_upcalls_ch, 1);
2840 	callout_init(&V_bw_upcalls_ch, 1);
2841 	callout_init(&V_bw_meter_ch, 1);
2842 }
2843 
2844 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2845 	NULL);
2846 
2847 static void
2848 vnet_mroute_uninit(const void *unused __unused)
2849 {
2850 
2851 	free(V_bw_upcalls, M_MRTABLE);
2852 	free(V_bw_meter_timers, M_MRTABLE);
2853 	free(V_viftable, M_MRTABLE);
2854 	free(V_nexpire, M_MRTABLE);
2855 	V_nexpire = NULL;
2856 }
2857 
2858 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2859 	vnet_mroute_uninit, NULL);
2860 
2861 static int
2862 ip_mroute_modevent(module_t mod, int type, void *unused)
2863 {
2864 
2865     switch (type) {
2866     case MOD_LOAD:
2867 	MROUTER_LOCK_INIT();
2868 
2869 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2870 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2871 	if (if_detach_event_tag == NULL) {
2872 		printf("ip_mroute: unable to register "
2873 		    "ifnet_departure_event handler\n");
2874 		MROUTER_LOCK_DESTROY();
2875 		return (EINVAL);
2876 	}
2877 
2878 	MFC_LOCK_INIT();
2879 	VIF_LOCK_INIT();
2880 
2881 	mfchashsize = MFCHASHSIZE;
2882 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2883 	    !powerof2(mfchashsize)) {
2884 		printf("WARNING: %s not a power of 2; using default\n",
2885 		    "net.inet.ip.mfchashsize");
2886 		mfchashsize = MFCHASHSIZE;
2887 	}
2888 
2889 	pim_squelch_wholepkt = 0;
2890 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2891 	    &pim_squelch_wholepkt);
2892 
2893 	pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2894 	if (pim_encap_cookie == NULL) {
2895 		printf("ip_mroute: unable to attach pim encap\n");
2896 		VIF_LOCK_DESTROY();
2897 		MFC_LOCK_DESTROY();
2898 		MROUTER_LOCK_DESTROY();
2899 		return (EINVAL);
2900 	}
2901 
2902 	ip_mcast_src = X_ip_mcast_src;
2903 	ip_mforward = X_ip_mforward;
2904 	ip_mrouter_done = X_ip_mrouter_done;
2905 	ip_mrouter_get = X_ip_mrouter_get;
2906 	ip_mrouter_set = X_ip_mrouter_set;
2907 
2908 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2909 	ip_rsvp_vif = X_ip_rsvp_vif;
2910 
2911 	legal_vif_num = X_legal_vif_num;
2912 	mrt_ioctl = X_mrt_ioctl;
2913 	rsvp_input_p = X_rsvp_input;
2914 	break;
2915 
2916     case MOD_UNLOAD:
2917 	/*
2918 	 * Typically module unload happens after the user-level
2919 	 * process has shutdown the kernel services (the check
2920 	 * below insures someone can't just yank the module out
2921 	 * from under a running process).  But if the module is
2922 	 * just loaded and then unloaded w/o starting up a user
2923 	 * process we still need to cleanup.
2924 	 */
2925 	MROUTER_LOCK();
2926 	if (ip_mrouter_cnt != 0) {
2927 	    MROUTER_UNLOCK();
2928 	    return (EINVAL);
2929 	}
2930 	ip_mrouter_unloading = 1;
2931 	MROUTER_UNLOCK();
2932 
2933 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2934 
2935 	if (pim_encap_cookie) {
2936 	    ip_encap_detach(pim_encap_cookie);
2937 	    pim_encap_cookie = NULL;
2938 	}
2939 
2940 	ip_mcast_src = NULL;
2941 	ip_mforward = NULL;
2942 	ip_mrouter_done = NULL;
2943 	ip_mrouter_get = NULL;
2944 	ip_mrouter_set = NULL;
2945 
2946 	ip_rsvp_force_done = NULL;
2947 	ip_rsvp_vif = NULL;
2948 
2949 	legal_vif_num = NULL;
2950 	mrt_ioctl = NULL;
2951 	rsvp_input_p = NULL;
2952 
2953 	VIF_LOCK_DESTROY();
2954 	MFC_LOCK_DESTROY();
2955 	MROUTER_LOCK_DESTROY();
2956 	break;
2957 
2958     default:
2959 	return EOPNOTSUPP;
2960     }
2961     return 0;
2962 }
2963 
2964 static moduledata_t ip_mroutemod = {
2965     "ip_mroute",
2966     ip_mroute_modevent,
2967     0
2968 };
2969 
2970 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2971