xref: /freebsd/sys/netinet/ip_mroute.c (revision 88640c0e8b6f503426cce9ea1337098c241d3801)
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[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 VNET_DEFINE_STATIC(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 VNET_DEFINE_STATIC(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
214 #define	V_bw_meter_timers	VNET(bw_meter_timers)
215 VNET_DEFINE_STATIC(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 VNET_DEFINE_STATIC(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
224 #define	V_bw_upcalls		VNET(bw_upcalls)
225 VNET_DEFINE_STATIC(u_int, bw_upcalls_n); /* # of pending upcalls */
226 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
227 VNET_DEFINE_STATIC(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 VNET_PCPUSTAT_DEFINE_STATIC(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 static const struct encaptab *pim_encap_cookie;
246 static int pim_encapcheck(const struct mbuf *, int, int, void *);
247 static int pim_input(struct mbuf *, int, int, void *);
248 
249 static const struct encap_config ipv4_encap_cfg = {
250 	.proto = IPPROTO_PIM,
251 	.min_length = sizeof(struct ip) + PIM_MINLEN,
252 	.exact_match = 8,
253 	.check = pim_encapcheck,
254 	.input = pim_input
255 };
256 
257 /*
258  * Note: the PIM Register encapsulation adds the following in front of a
259  * data packet:
260  *
261  * struct pim_encap_hdr {
262  *    struct ip ip;
263  *    struct pim_encap_pimhdr  pim;
264  * }
265  *
266  */
267 
268 struct pim_encap_pimhdr {
269 	struct pim pim;
270 	uint32_t   flags;
271 };
272 #define		PIM_ENCAP_TTL	64
273 
274 static struct ip pim_encap_iphdr = {
275 #if BYTE_ORDER == LITTLE_ENDIAN
276 	sizeof(struct ip) >> 2,
277 	IPVERSION,
278 #else
279 	IPVERSION,
280 	sizeof(struct ip) >> 2,
281 #endif
282 	0,			/* tos */
283 	sizeof(struct ip),	/* total length */
284 	0,			/* id */
285 	0,			/* frag offset */
286 	PIM_ENCAP_TTL,
287 	IPPROTO_PIM,
288 	0,			/* checksum */
289 };
290 
291 static struct pim_encap_pimhdr pim_encap_pimhdr = {
292     {
293 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
294 	0,			/* reserved */
295 	0,			/* checksum */
296     },
297     0				/* flags */
298 };
299 
300 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
301 #define	V_reg_vif_num		VNET(reg_vif_num)
302 VNET_DEFINE_STATIC(struct ifnet, multicast_register_if);
303 #define	V_multicast_register_if	VNET(multicast_register_if)
304 
305 /*
306  * Private variables.
307  */
308 
309 static u_long	X_ip_mcast_src(int);
310 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
311 		    struct ip_moptions *);
312 static int	X_ip_mrouter_done(void);
313 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
314 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
315 static int	X_legal_vif_num(int);
316 static int	X_mrt_ioctl(u_long, caddr_t, int);
317 
318 static int	add_bw_upcall(struct bw_upcall *);
319 static int	add_mfc(struct mfcctl2 *);
320 static int	add_vif(struct vifctl *);
321 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
322 static void	bw_meter_process(void);
323 static void	bw_meter_receive_packet(struct bw_meter *, int,
324 		    struct timeval *);
325 static void	bw_upcalls_send(void);
326 static int	del_bw_upcall(struct bw_upcall *);
327 static int	del_mfc(struct mfcctl2 *);
328 static int	del_vif(vifi_t);
329 static int	del_vif_locked(vifi_t);
330 static void	expire_bw_meter_process(void *);
331 static void	expire_bw_upcalls_send(void *);
332 static void	expire_mfc(struct mfc *);
333 static void	expire_upcalls(void *);
334 static void	free_bw_list(struct bw_meter *);
335 static int	get_sg_cnt(struct sioc_sg_req *);
336 static int	get_vif_cnt(struct sioc_vif_req *);
337 static void	if_detached_event(void *, struct ifnet *);
338 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
339 static int	ip_mrouter_init(struct socket *, int);
340 static __inline struct mfc *
341 		mfc_find(struct in_addr *, struct in_addr *);
342 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
343 static struct mbuf *
344 		pim_register_prepare(struct ip *, struct mbuf *);
345 static int	pim_register_send(struct ip *, struct vif *,
346 		    struct mbuf *, struct mfc *);
347 static int	pim_register_send_rp(struct ip *, struct vif *,
348 		    struct mbuf *, struct mfc *);
349 static int	pim_register_send_upcall(struct ip *, struct vif *,
350 		    struct mbuf *, struct mfc *);
351 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
352 static void	send_packet(struct vif *, struct mbuf *);
353 static int	set_api_config(uint32_t *);
354 static int	set_assert(int);
355 static int	socket_send(struct socket *, struct mbuf *,
356 		    struct sockaddr_in *);
357 static void	unschedule_bw_meter(struct bw_meter *);
358 
359 /*
360  * Kernel multicast forwarding API capabilities and setup.
361  * If more API capabilities are added to the kernel, they should be
362  * recorded in `mrt_api_support'.
363  */
364 #define MRT_API_VERSION		0x0305
365 
366 static const int mrt_api_version = MRT_API_VERSION;
367 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
368 					 MRT_MFC_FLAGS_BORDER_VIF |
369 					 MRT_MFC_RP |
370 					 MRT_MFC_BW_UPCALL);
371 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
372 #define	V_mrt_api_config	VNET(mrt_api_config)
373 VNET_DEFINE_STATIC(int, pim_assert_enabled);
374 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
375 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
376 
377 /*
378  * Find a route for a given origin IP address and multicast group address.
379  * Statistics must be updated by the caller.
380  */
381 static __inline struct mfc *
382 mfc_find(struct in_addr *o, struct in_addr *g)
383 {
384 	struct mfc *rt;
385 
386 	MFC_LOCK_ASSERT();
387 
388 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
389 		if (in_hosteq(rt->mfc_origin, *o) &&
390 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
391 		    TAILQ_EMPTY(&rt->mfc_stall))
392 			break;
393 	}
394 
395 	return (rt);
396 }
397 
398 /*
399  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
400  */
401 static int
402 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
403 {
404     int	error, optval;
405     vifi_t	vifi;
406     struct	vifctl vifc;
407     struct	mfcctl2 mfc;
408     struct	bw_upcall bw_upcall;
409     uint32_t	i;
410 
411     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
412 	return EPERM;
413 
414     error = 0;
415     switch (sopt->sopt_name) {
416     case MRT_INIT:
417 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
418 	if (error)
419 	    break;
420 	error = ip_mrouter_init(so, optval);
421 	break;
422 
423     case MRT_DONE:
424 	error = ip_mrouter_done();
425 	break;
426 
427     case MRT_ADD_VIF:
428 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
429 	if (error)
430 	    break;
431 	error = add_vif(&vifc);
432 	break;
433 
434     case MRT_DEL_VIF:
435 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
436 	if (error)
437 	    break;
438 	error = del_vif(vifi);
439 	break;
440 
441     case MRT_ADD_MFC:
442     case MRT_DEL_MFC:
443 	/*
444 	 * select data size depending on API version.
445 	 */
446 	if (sopt->sopt_name == MRT_ADD_MFC &&
447 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
448 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
449 				sizeof(struct mfcctl2));
450 	} else {
451 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
452 				sizeof(struct mfcctl));
453 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
454 			sizeof(mfc) - sizeof(struct mfcctl));
455 	}
456 	if (error)
457 	    break;
458 	if (sopt->sopt_name == MRT_ADD_MFC)
459 	    error = add_mfc(&mfc);
460 	else
461 	    error = del_mfc(&mfc);
462 	break;
463 
464     case MRT_ASSERT:
465 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
466 	if (error)
467 	    break;
468 	set_assert(optval);
469 	break;
470 
471     case MRT_API_CONFIG:
472 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
473 	if (!error)
474 	    error = set_api_config(&i);
475 	if (!error)
476 	    error = sooptcopyout(sopt, &i, sizeof i);
477 	break;
478 
479     case MRT_ADD_BW_UPCALL:
480     case MRT_DEL_BW_UPCALL:
481 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
482 				sizeof bw_upcall);
483 	if (error)
484 	    break;
485 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
486 	    error = add_bw_upcall(&bw_upcall);
487 	else
488 	    error = del_bw_upcall(&bw_upcall);
489 	break;
490 
491     default:
492 	error = EOPNOTSUPP;
493 	break;
494     }
495     return error;
496 }
497 
498 /*
499  * Handle MRT getsockopt commands
500  */
501 static int
502 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
503 {
504     int error;
505 
506     switch (sopt->sopt_name) {
507     case MRT_VERSION:
508 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
509 	break;
510 
511     case MRT_ASSERT:
512 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
513 	    sizeof V_pim_assert_enabled);
514 	break;
515 
516     case MRT_API_SUPPORT:
517 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
518 	break;
519 
520     case MRT_API_CONFIG:
521 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
522 	break;
523 
524     default:
525 	error = EOPNOTSUPP;
526 	break;
527     }
528     return error;
529 }
530 
531 /*
532  * Handle ioctl commands to obtain information from the cache
533  */
534 static int
535 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
536 {
537     int error = 0;
538 
539     /*
540      * Currently the only function calling this ioctl routine is rtioctl_fib().
541      * Typically, only root can create the raw socket in order to execute
542      * this ioctl method, however the request might be coming from a prison
543      */
544     error = priv_check(curthread, PRIV_NETINET_MROUTE);
545     if (error)
546 	return (error);
547     switch (cmd) {
548     case (SIOCGETVIFCNT):
549 	error = get_vif_cnt((struct sioc_vif_req *)data);
550 	break;
551 
552     case (SIOCGETSGCNT):
553 	error = get_sg_cnt((struct sioc_sg_req *)data);
554 	break;
555 
556     default:
557 	error = EINVAL;
558 	break;
559     }
560     return error;
561 }
562 
563 /*
564  * returns the packet, byte, rpf-failure count for the source group provided
565  */
566 static int
567 get_sg_cnt(struct sioc_sg_req *req)
568 {
569     struct mfc *rt;
570 
571     MFC_LOCK();
572     rt = mfc_find(&req->src, &req->grp);
573     if (rt == NULL) {
574 	MFC_UNLOCK();
575 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
576 	return EADDRNOTAVAIL;
577     }
578     req->pktcnt = rt->mfc_pkt_cnt;
579     req->bytecnt = rt->mfc_byte_cnt;
580     req->wrong_if = rt->mfc_wrong_if;
581     MFC_UNLOCK();
582     return 0;
583 }
584 
585 /*
586  * returns the input and output packet and byte counts on the vif provided
587  */
588 static int
589 get_vif_cnt(struct sioc_vif_req *req)
590 {
591     vifi_t vifi = req->vifi;
592 
593     VIF_LOCK();
594     if (vifi >= V_numvifs) {
595 	VIF_UNLOCK();
596 	return EINVAL;
597     }
598 
599     req->icount = V_viftable[vifi].v_pkt_in;
600     req->ocount = V_viftable[vifi].v_pkt_out;
601     req->ibytes = V_viftable[vifi].v_bytes_in;
602     req->obytes = V_viftable[vifi].v_bytes_out;
603     VIF_UNLOCK();
604 
605     return 0;
606 }
607 
608 static void
609 if_detached_event(void *arg __unused, struct ifnet *ifp)
610 {
611     vifi_t vifi;
612     u_long i;
613 
614     MROUTER_LOCK();
615 
616     if (V_ip_mrouter == NULL) {
617 	MROUTER_UNLOCK();
618 	return;
619     }
620 
621     VIF_LOCK();
622     MFC_LOCK();
623 
624     /*
625      * Tear down multicast forwarder state associated with this ifnet.
626      * 1. Walk the vif list, matching vifs against this ifnet.
627      * 2. Walk the multicast forwarding cache (mfc) looking for
628      *    inner matches with this vif's index.
629      * 3. Expire any matching multicast forwarding cache entries.
630      * 4. Free vif state. This should disable ALLMULTI on the interface.
631      */
632     for (vifi = 0; vifi < V_numvifs; vifi++) {
633 	if (V_viftable[vifi].v_ifp != ifp)
634 		continue;
635 	for (i = 0; i < mfchashsize; i++) {
636 		struct mfc *rt, *nrt;
637 
638 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
639 			if (rt->mfc_parent == vifi) {
640 				expire_mfc(rt);
641 			}
642 		}
643 	}
644 	del_vif_locked(vifi);
645     }
646 
647     MFC_UNLOCK();
648     VIF_UNLOCK();
649 
650     MROUTER_UNLOCK();
651 }
652 
653 /*
654  * Enable multicast forwarding.
655  */
656 static int
657 ip_mrouter_init(struct socket *so, int version)
658 {
659 
660     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
661         so->so_type, so->so_proto->pr_protocol);
662 
663     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
664 	return EOPNOTSUPP;
665 
666     if (version != 1)
667 	return ENOPROTOOPT;
668 
669     MROUTER_LOCK();
670 
671     if (ip_mrouter_unloading) {
672 	MROUTER_UNLOCK();
673 	return ENOPROTOOPT;
674     }
675 
676     if (V_ip_mrouter != NULL) {
677 	MROUTER_UNLOCK();
678 	return EADDRINUSE;
679     }
680 
681     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
682 	HASH_NOWAIT);
683 
684     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
685 	curvnet);
686     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
687 	curvnet);
688     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
689 	curvnet);
690 
691     V_ip_mrouter = so;
692     ip_mrouter_cnt++;
693 
694     MROUTER_UNLOCK();
695 
696     CTR1(KTR_IPMF, "%s: done", __func__);
697 
698     return 0;
699 }
700 
701 /*
702  * Disable multicast forwarding.
703  */
704 static int
705 X_ip_mrouter_done(void)
706 {
707     struct ifnet *ifp;
708     u_long i;
709     vifi_t vifi;
710 
711     MROUTER_LOCK();
712 
713     if (V_ip_mrouter == NULL) {
714 	MROUTER_UNLOCK();
715 	return EINVAL;
716     }
717 
718     /*
719      * Detach/disable hooks to the reset of the system.
720      */
721     V_ip_mrouter = NULL;
722     ip_mrouter_cnt--;
723     V_mrt_api_config = 0;
724 
725     VIF_LOCK();
726 
727     /*
728      * For each phyint in use, disable promiscuous reception of all IP
729      * multicasts.
730      */
731     for (vifi = 0; vifi < V_numvifs; vifi++) {
732 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
733 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
734 	    ifp = V_viftable[vifi].v_ifp;
735 	    if_allmulti(ifp, 0);
736 	}
737     }
738     bzero((caddr_t)V_viftable, sizeof(V_viftable));
739     V_numvifs = 0;
740     V_pim_assert_enabled = 0;
741 
742     VIF_UNLOCK();
743 
744     callout_stop(&V_expire_upcalls_ch);
745     callout_stop(&V_bw_upcalls_ch);
746     callout_stop(&V_bw_meter_ch);
747 
748     MFC_LOCK();
749 
750     /*
751      * Free all multicast forwarding cache entries.
752      * Do not use hashdestroy(), as we must perform other cleanup.
753      */
754     for (i = 0; i < mfchashsize; i++) {
755 	struct mfc *rt, *nrt;
756 
757 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
758 		expire_mfc(rt);
759 	}
760     }
761     free(V_mfchashtbl, M_MRTABLE);
762     V_mfchashtbl = NULL;
763 
764     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
765 
766     V_bw_upcalls_n = 0;
767     bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
768 
769     MFC_UNLOCK();
770 
771     V_reg_vif_num = VIFI_INVALID;
772 
773     MROUTER_UNLOCK();
774 
775     CTR1(KTR_IPMF, "%s: done", __func__);
776 
777     return 0;
778 }
779 
780 /*
781  * Set PIM assert processing global
782  */
783 static int
784 set_assert(int i)
785 {
786     if ((i != 1) && (i != 0))
787 	return EINVAL;
788 
789     V_pim_assert_enabled = i;
790 
791     return 0;
792 }
793 
794 /*
795  * Configure API capabilities
796  */
797 int
798 set_api_config(uint32_t *apival)
799 {
800     u_long i;
801 
802     /*
803      * We can set the API capabilities only if it is the first operation
804      * after MRT_INIT. I.e.:
805      *  - there are no vifs installed
806      *  - pim_assert is not enabled
807      *  - the MFC table is empty
808      */
809     if (V_numvifs > 0) {
810 	*apival = 0;
811 	return EPERM;
812     }
813     if (V_pim_assert_enabled) {
814 	*apival = 0;
815 	return EPERM;
816     }
817 
818     MFC_LOCK();
819 
820     for (i = 0; i < mfchashsize; i++) {
821 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
822 	    MFC_UNLOCK();
823 	    *apival = 0;
824 	    return EPERM;
825 	}
826     }
827 
828     MFC_UNLOCK();
829 
830     V_mrt_api_config = *apival & mrt_api_support;
831     *apival = V_mrt_api_config;
832 
833     return 0;
834 }
835 
836 /*
837  * Add a vif to the vif table
838  */
839 static int
840 add_vif(struct vifctl *vifcp)
841 {
842     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
843     struct sockaddr_in sin = {sizeof sin, AF_INET};
844     struct ifaddr *ifa;
845     struct ifnet *ifp;
846     int error;
847 
848     VIF_LOCK();
849     if (vifcp->vifc_vifi >= MAXVIFS) {
850 	VIF_UNLOCK();
851 	return EINVAL;
852     }
853     /* rate limiting is no longer supported by this code */
854     if (vifcp->vifc_rate_limit != 0) {
855 	log(LOG_ERR, "rate limiting is no longer supported\n");
856 	VIF_UNLOCK();
857 	return EINVAL;
858     }
859     if (!in_nullhost(vifp->v_lcl_addr)) {
860 	VIF_UNLOCK();
861 	return EADDRINUSE;
862     }
863     if (in_nullhost(vifcp->vifc_lcl_addr)) {
864 	VIF_UNLOCK();
865 	return EADDRNOTAVAIL;
866     }
867 
868     /* Find the interface with an address in AF_INET family */
869     if (vifcp->vifc_flags & VIFF_REGISTER) {
870 	/*
871 	 * XXX: Because VIFF_REGISTER does not really need a valid
872 	 * local interface (e.g. it could be 127.0.0.2), we don't
873 	 * check its address.
874 	 */
875 	ifp = NULL;
876     } else {
877 	struct epoch_tracker et;
878 
879 	sin.sin_addr = vifcp->vifc_lcl_addr;
880 	NET_EPOCH_ENTER(et);
881 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
882 	if (ifa == NULL) {
883 		NET_EPOCH_EXIT(et);
884 	    VIF_UNLOCK();
885 	    return EADDRNOTAVAIL;
886 	}
887 	ifp = ifa->ifa_ifp;
888 	NET_EPOCH_EXIT(et);
889     }
890 
891     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
892 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
893 	VIF_UNLOCK();
894 	return EOPNOTSUPP;
895     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
896 	ifp = &V_multicast_register_if;
897 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
898 	if (V_reg_vif_num == VIFI_INVALID) {
899 	    if_initname(&V_multicast_register_if, "register_vif", 0);
900 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
901 	    V_reg_vif_num = vifcp->vifc_vifi;
902 	}
903     } else {		/* Make sure the interface supports multicast */
904 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
905 	    VIF_UNLOCK();
906 	    return EOPNOTSUPP;
907 	}
908 
909 	/* Enable promiscuous reception of all IP multicasts from the if */
910 	error = if_allmulti(ifp, 1);
911 	if (error) {
912 	    VIF_UNLOCK();
913 	    return error;
914 	}
915     }
916 
917     vifp->v_flags     = vifcp->vifc_flags;
918     vifp->v_threshold = vifcp->vifc_threshold;
919     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
920     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
921     vifp->v_ifp       = ifp;
922     /* initialize per vif pkt counters */
923     vifp->v_pkt_in    = 0;
924     vifp->v_pkt_out   = 0;
925     vifp->v_bytes_in  = 0;
926     vifp->v_bytes_out = 0;
927 
928     /* Adjust numvifs up if the vifi is higher than numvifs */
929     if (V_numvifs <= vifcp->vifc_vifi)
930 	V_numvifs = vifcp->vifc_vifi + 1;
931 
932     VIF_UNLOCK();
933 
934     CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
935 	(int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
936 	(int)vifcp->vifc_threshold);
937 
938     return 0;
939 }
940 
941 /*
942  * Delete a vif from the vif table
943  */
944 static int
945 del_vif_locked(vifi_t vifi)
946 {
947     struct vif *vifp;
948 
949     VIF_LOCK_ASSERT();
950 
951     if (vifi >= V_numvifs) {
952 	return EINVAL;
953     }
954     vifp = &V_viftable[vifi];
955     if (in_nullhost(vifp->v_lcl_addr)) {
956 	return EADDRNOTAVAIL;
957     }
958 
959     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
960 	if_allmulti(vifp->v_ifp, 0);
961 
962     if (vifp->v_flags & VIFF_REGISTER)
963 	V_reg_vif_num = VIFI_INVALID;
964 
965     bzero((caddr_t)vifp, sizeof (*vifp));
966 
967     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
968 
969     /* Adjust numvifs down */
970     for (vifi = V_numvifs; vifi > 0; vifi--)
971 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
972 	    break;
973     V_numvifs = vifi;
974 
975     return 0;
976 }
977 
978 static int
979 del_vif(vifi_t vifi)
980 {
981     int cc;
982 
983     VIF_LOCK();
984     cc = del_vif_locked(vifi);
985     VIF_UNLOCK();
986 
987     return cc;
988 }
989 
990 /*
991  * update an mfc entry without resetting counters and S,G addresses.
992  */
993 static void
994 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
995 {
996     int i;
997 
998     rt->mfc_parent = mfccp->mfcc_parent;
999     for (i = 0; i < V_numvifs; i++) {
1000 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1001 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1002 	    MRT_MFC_FLAGS_ALL;
1003     }
1004     /* set the RP address */
1005     if (V_mrt_api_config & MRT_MFC_RP)
1006 	rt->mfc_rp = mfccp->mfcc_rp;
1007     else
1008 	rt->mfc_rp.s_addr = INADDR_ANY;
1009 }
1010 
1011 /*
1012  * fully initialize an mfc entry from the parameter.
1013  */
1014 static void
1015 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1016 {
1017     rt->mfc_origin     = mfccp->mfcc_origin;
1018     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1019 
1020     update_mfc_params(rt, mfccp);
1021 
1022     /* initialize pkt counters per src-grp */
1023     rt->mfc_pkt_cnt    = 0;
1024     rt->mfc_byte_cnt   = 0;
1025     rt->mfc_wrong_if   = 0;
1026     timevalclear(&rt->mfc_last_assert);
1027 }
1028 
1029 static void
1030 expire_mfc(struct mfc *rt)
1031 {
1032 	struct rtdetq *rte, *nrte;
1033 
1034 	MFC_LOCK_ASSERT();
1035 
1036 	free_bw_list(rt->mfc_bw_meter);
1037 
1038 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1039 		m_freem(rte->m);
1040 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1041 		free(rte, M_MRTABLE);
1042 	}
1043 
1044 	LIST_REMOVE(rt, mfc_hash);
1045 	free(rt, M_MRTABLE);
1046 }
1047 
1048 /*
1049  * Add an mfc entry
1050  */
1051 static int
1052 add_mfc(struct mfcctl2 *mfccp)
1053 {
1054     struct mfc *rt;
1055     struct rtdetq *rte, *nrte;
1056     u_long hash = 0;
1057     u_short nstl;
1058 
1059     VIF_LOCK();
1060     MFC_LOCK();
1061 
1062     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1063 
1064     /* If an entry already exists, just update the fields */
1065     if (rt) {
1066 	CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1067 	    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1068 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1069 	    mfccp->mfcc_parent);
1070 	update_mfc_params(rt, mfccp);
1071 	MFC_UNLOCK();
1072 	VIF_UNLOCK();
1073 	return (0);
1074     }
1075 
1076     /*
1077      * Find the entry for which the upcall was made and update
1078      */
1079     nstl = 0;
1080     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1081     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1082 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1083 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1084 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1085 		CTR5(KTR_IPMF,
1086 		    "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1087 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1088 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1089 		    mfccp->mfcc_parent,
1090 		    TAILQ_FIRST(&rt->mfc_stall));
1091 		if (nstl++)
1092 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1093 
1094 		init_mfc_params(rt, mfccp);
1095 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1096 		V_nexpire[hash]--;
1097 
1098 		/* Free queued packets, but attempt to forward them first. */
1099 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1100 			if (rte->ifp != NULL)
1101 				ip_mdq(rte->m, rte->ifp, rt, -1);
1102 			m_freem(rte->m);
1103 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1104 			rt->mfc_nstall--;
1105 			free(rte, M_MRTABLE);
1106 		}
1107 	}
1108     }
1109 
1110     /*
1111      * It is possible that an entry is being inserted without an upcall
1112      */
1113     if (nstl == 0) {
1114 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1115 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1116 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1117 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1118 			init_mfc_params(rt, mfccp);
1119 			if (rt->mfc_expire)
1120 			    V_nexpire[hash]--;
1121 			rt->mfc_expire = 0;
1122 			break; /* XXX */
1123 		}
1124 	}
1125 
1126 	if (rt == NULL) {		/* no upcall, so make a new entry */
1127 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1128 	    if (rt == NULL) {
1129 		MFC_UNLOCK();
1130 		VIF_UNLOCK();
1131 		return (ENOBUFS);
1132 	    }
1133 
1134 	    init_mfc_params(rt, mfccp);
1135 	    TAILQ_INIT(&rt->mfc_stall);
1136 	    rt->mfc_nstall = 0;
1137 
1138 	    rt->mfc_expire     = 0;
1139 	    rt->mfc_bw_meter = NULL;
1140 
1141 	    /* insert new entry at head of hash chain */
1142 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1143 	}
1144     }
1145 
1146     MFC_UNLOCK();
1147     VIF_UNLOCK();
1148 
1149     return (0);
1150 }
1151 
1152 /*
1153  * Delete an mfc entry
1154  */
1155 static int
1156 del_mfc(struct mfcctl2 *mfccp)
1157 {
1158     struct in_addr	origin;
1159     struct in_addr	mcastgrp;
1160     struct mfc		*rt;
1161 
1162     origin = mfccp->mfcc_origin;
1163     mcastgrp = mfccp->mfcc_mcastgrp;
1164 
1165     CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1166 	ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1167 
1168     MFC_LOCK();
1169 
1170     rt = mfc_find(&origin, &mcastgrp);
1171     if (rt == NULL) {
1172 	MFC_UNLOCK();
1173 	return EADDRNOTAVAIL;
1174     }
1175 
1176     /*
1177      * free the bw_meter entries
1178      */
1179     free_bw_list(rt->mfc_bw_meter);
1180     rt->mfc_bw_meter = NULL;
1181 
1182     LIST_REMOVE(rt, mfc_hash);
1183     free(rt, M_MRTABLE);
1184 
1185     MFC_UNLOCK();
1186 
1187     return (0);
1188 }
1189 
1190 /*
1191  * Send a message to the routing daemon on the multicast routing socket.
1192  */
1193 static int
1194 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1195 {
1196     if (s) {
1197 	SOCKBUF_LOCK(&s->so_rcv);
1198 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1199 	    NULL) != 0) {
1200 	    sorwakeup_locked(s);
1201 	    return 0;
1202 	}
1203 	SOCKBUF_UNLOCK(&s->so_rcv);
1204     }
1205     m_freem(mm);
1206     return -1;
1207 }
1208 
1209 /*
1210  * IP multicast forwarding function. This function assumes that the packet
1211  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1212  * pointed to by "ifp", and the packet is to be relayed to other networks
1213  * that have members of the packet's destination IP multicast group.
1214  *
1215  * The packet is returned unscathed to the caller, unless it is
1216  * erroneous, in which case a non-zero return value tells the caller to
1217  * discard it.
1218  */
1219 
1220 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1221 
1222 static int
1223 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1224     struct ip_moptions *imo)
1225 {
1226     struct mfc *rt;
1227     int error;
1228     vifi_t vifi;
1229 
1230     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1231 	ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1232 
1233     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1234 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1235 	/*
1236 	 * Packet arrived via a physical interface or
1237 	 * an encapsulated tunnel or a register_vif.
1238 	 */
1239     } else {
1240 	/*
1241 	 * Packet arrived through a source-route tunnel.
1242 	 * Source-route tunnels are no longer supported.
1243 	 */
1244 	return (1);
1245     }
1246 
1247     VIF_LOCK();
1248     MFC_LOCK();
1249     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1250 	if (ip->ip_ttl < MAXTTL)
1251 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1252 	error = ip_mdq(m, ifp, NULL, vifi);
1253 	MFC_UNLOCK();
1254 	VIF_UNLOCK();
1255 	return error;
1256     }
1257 
1258     /*
1259      * Don't forward a packet with time-to-live of zero or one,
1260      * or a packet destined to a local-only group.
1261      */
1262     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1263 	MFC_UNLOCK();
1264 	VIF_UNLOCK();
1265 	return 0;
1266     }
1267 
1268     /*
1269      * Determine forwarding vifs from the forwarding cache table
1270      */
1271     MRTSTAT_INC(mrts_mfc_lookups);
1272     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1273 
1274     /* Entry exists, so forward if necessary */
1275     if (rt != NULL) {
1276 	error = ip_mdq(m, ifp, rt, -1);
1277 	MFC_UNLOCK();
1278 	VIF_UNLOCK();
1279 	return error;
1280     } else {
1281 	/*
1282 	 * If we don't have a route for packet's origin,
1283 	 * Make a copy of the packet & send message to routing daemon
1284 	 */
1285 
1286 	struct mbuf *mb0;
1287 	struct rtdetq *rte;
1288 	u_long hash;
1289 	int hlen = ip->ip_hl << 2;
1290 
1291 	MRTSTAT_INC(mrts_mfc_misses);
1292 	MRTSTAT_INC(mrts_no_route);
1293 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1294 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1295 
1296 	/*
1297 	 * Allocate mbufs early so that we don't do extra work if we are
1298 	 * just going to fail anyway.  Make sure to pullup the header so
1299 	 * that other people can't step on it.
1300 	 */
1301 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1302 	    M_NOWAIT|M_ZERO);
1303 	if (rte == NULL) {
1304 	    MFC_UNLOCK();
1305 	    VIF_UNLOCK();
1306 	    return ENOBUFS;
1307 	}
1308 
1309 	mb0 = m_copypacket(m, M_NOWAIT);
1310 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1311 	    mb0 = m_pullup(mb0, hlen);
1312 	if (mb0 == NULL) {
1313 	    free(rte, M_MRTABLE);
1314 	    MFC_UNLOCK();
1315 	    VIF_UNLOCK();
1316 	    return ENOBUFS;
1317 	}
1318 
1319 	/* is there an upcall waiting for this flow ? */
1320 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1321 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1322 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1323 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1324 		    !TAILQ_EMPTY(&rt->mfc_stall))
1325 			break;
1326 	}
1327 
1328 	if (rt == NULL) {
1329 	    int i;
1330 	    struct igmpmsg *im;
1331 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1332 	    struct mbuf *mm;
1333 
1334 	    /*
1335 	     * Locate the vifi for the incoming interface for this packet.
1336 	     * If none found, drop packet.
1337 	     */
1338 	    for (vifi = 0; vifi < V_numvifs &&
1339 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1340 		;
1341 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1342 		goto non_fatal;
1343 
1344 	    /* no upcall, so make a new entry */
1345 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1346 	    if (rt == NULL)
1347 		goto fail;
1348 
1349 	    /* Make a copy of the header to send to the user level process */
1350 	    mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1351 	    if (mm == NULL)
1352 		goto fail1;
1353 
1354 	    /*
1355 	     * Send message to routing daemon to install
1356 	     * a route into the kernel table
1357 	     */
1358 
1359 	    im = mtod(mm, struct igmpmsg *);
1360 	    im->im_msgtype = IGMPMSG_NOCACHE;
1361 	    im->im_mbz = 0;
1362 	    im->im_vif = vifi;
1363 
1364 	    MRTSTAT_INC(mrts_upcalls);
1365 
1366 	    k_igmpsrc.sin_addr = ip->ip_src;
1367 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1368 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1369 		MRTSTAT_INC(mrts_upq_sockfull);
1370 fail1:
1371 		free(rt, M_MRTABLE);
1372 fail:
1373 		free(rte, M_MRTABLE);
1374 		m_freem(mb0);
1375 		MFC_UNLOCK();
1376 		VIF_UNLOCK();
1377 		return ENOBUFS;
1378 	    }
1379 
1380 	    /* insert new entry at head of hash chain */
1381 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1382 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1383 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1384 	    V_nexpire[hash]++;
1385 	    for (i = 0; i < V_numvifs; i++) {
1386 		rt->mfc_ttls[i] = 0;
1387 		rt->mfc_flags[i] = 0;
1388 	    }
1389 	    rt->mfc_parent = -1;
1390 
1391 	    /* clear the RP address */
1392 	    rt->mfc_rp.s_addr = INADDR_ANY;
1393 	    rt->mfc_bw_meter = NULL;
1394 
1395 	    /* initialize pkt counters per src-grp */
1396 	    rt->mfc_pkt_cnt = 0;
1397 	    rt->mfc_byte_cnt = 0;
1398 	    rt->mfc_wrong_if = 0;
1399 	    timevalclear(&rt->mfc_last_assert);
1400 
1401 	    TAILQ_INIT(&rt->mfc_stall);
1402 	    rt->mfc_nstall = 0;
1403 
1404 	    /* link into table */
1405 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1406 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1407 	    rt->mfc_nstall++;
1408 
1409 	} else {
1410 	    /* determine if queue has overflowed */
1411 	    if (rt->mfc_nstall > MAX_UPQ) {
1412 		MRTSTAT_INC(mrts_upq_ovflw);
1413 non_fatal:
1414 		free(rte, M_MRTABLE);
1415 		m_freem(mb0);
1416 		MFC_UNLOCK();
1417 		VIF_UNLOCK();
1418 		return (0);
1419 	    }
1420 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1421 	    rt->mfc_nstall++;
1422 	}
1423 
1424 	rte->m			= mb0;
1425 	rte->ifp		= ifp;
1426 
1427 	MFC_UNLOCK();
1428 	VIF_UNLOCK();
1429 
1430 	return 0;
1431     }
1432 }
1433 
1434 /*
1435  * Clean up the cache entry if upcall is not serviced
1436  */
1437 static void
1438 expire_upcalls(void *arg)
1439 {
1440     u_long i;
1441 
1442     CURVNET_SET((struct vnet *) arg);
1443 
1444     MFC_LOCK();
1445 
1446     for (i = 0; i < mfchashsize; i++) {
1447 	struct mfc *rt, *nrt;
1448 
1449 	if (V_nexpire[i] == 0)
1450 	    continue;
1451 
1452 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1453 		if (TAILQ_EMPTY(&rt->mfc_stall))
1454 			continue;
1455 
1456 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1457 			continue;
1458 
1459 		/*
1460 		 * free the bw_meter entries
1461 		 */
1462 		while (rt->mfc_bw_meter != NULL) {
1463 		    struct bw_meter *x = rt->mfc_bw_meter;
1464 
1465 		    rt->mfc_bw_meter = x->bm_mfc_next;
1466 		    free(x, M_BWMETER);
1467 		}
1468 
1469 		MRTSTAT_INC(mrts_cache_cleanups);
1470 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1471 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1472 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1473 
1474 		expire_mfc(rt);
1475 	    }
1476     }
1477 
1478     MFC_UNLOCK();
1479 
1480     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1481 	curvnet);
1482 
1483     CURVNET_RESTORE();
1484 }
1485 
1486 /*
1487  * Packet forwarding routine once entry in the cache is made
1488  */
1489 static int
1490 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1491 {
1492     struct ip  *ip = mtod(m, struct ip *);
1493     vifi_t vifi;
1494     int plen = ntohs(ip->ip_len);
1495 
1496     VIF_LOCK_ASSERT();
1497 
1498     /*
1499      * If xmt_vif is not -1, send on only the requested vif.
1500      *
1501      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1502      */
1503     if (xmt_vif < V_numvifs) {
1504 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1505 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1506 	else
1507 		phyint_send(ip, V_viftable + xmt_vif, m);
1508 	return 1;
1509     }
1510 
1511     /*
1512      * Don't forward if it didn't arrive from the parent vif for its origin.
1513      */
1514     vifi = rt->mfc_parent;
1515     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1516 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1517 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1518 	MRTSTAT_INC(mrts_wrong_if);
1519 	++rt->mfc_wrong_if;
1520 	/*
1521 	 * If we are doing PIM assert processing, send a message
1522 	 * to the routing daemon.
1523 	 *
1524 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1525 	 * can complete the SPT switch, regardless of the type
1526 	 * of the iif (broadcast media, GRE tunnel, etc).
1527 	 */
1528 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1529 	    V_viftable[vifi].v_ifp) {
1530 
1531 	    if (ifp == &V_multicast_register_if)
1532 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1533 
1534 	    /* Get vifi for the incoming packet */
1535 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1536 		vifi++)
1537 		;
1538 	    if (vifi >= V_numvifs)
1539 		return 0;	/* The iif is not found: ignore the packet. */
1540 
1541 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1542 		return 0;	/* WRONGVIF disabled: ignore the packet */
1543 
1544 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1545 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1546 		struct igmpmsg *im;
1547 		int hlen = ip->ip_hl << 2;
1548 		struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1549 
1550 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1551 		    mm = m_pullup(mm, hlen);
1552 		if (mm == NULL)
1553 		    return ENOBUFS;
1554 
1555 		im = mtod(mm, struct igmpmsg *);
1556 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1557 		im->im_mbz		= 0;
1558 		im->im_vif		= vifi;
1559 
1560 		MRTSTAT_INC(mrts_upcalls);
1561 
1562 		k_igmpsrc.sin_addr = im->im_src;
1563 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1564 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1565 		    MRTSTAT_INC(mrts_upq_sockfull);
1566 		    return ENOBUFS;
1567 		}
1568 	    }
1569 	}
1570 	return 0;
1571     }
1572 
1573 
1574     /* If I sourced this packet, it counts as output, else it was input. */
1575     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1576 	V_viftable[vifi].v_pkt_out++;
1577 	V_viftable[vifi].v_bytes_out += plen;
1578     } else {
1579 	V_viftable[vifi].v_pkt_in++;
1580 	V_viftable[vifi].v_bytes_in += plen;
1581     }
1582     rt->mfc_pkt_cnt++;
1583     rt->mfc_byte_cnt += plen;
1584 
1585     /*
1586      * For each vif, decide if a copy of the packet should be forwarded.
1587      * Forward if:
1588      *		- the ttl exceeds the vif's threshold
1589      *		- there are group members downstream on interface
1590      */
1591     for (vifi = 0; vifi < V_numvifs; vifi++)
1592 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1593 	    V_viftable[vifi].v_pkt_out++;
1594 	    V_viftable[vifi].v_bytes_out += plen;
1595 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1596 		pim_register_send(ip, V_viftable + vifi, m, rt);
1597 	    else
1598 		phyint_send(ip, V_viftable + vifi, m);
1599 	}
1600 
1601     /*
1602      * Perform upcall-related bw measuring.
1603      */
1604     if (rt->mfc_bw_meter != NULL) {
1605 	struct bw_meter *x;
1606 	struct timeval now;
1607 
1608 	microtime(&now);
1609 	MFC_LOCK_ASSERT();
1610 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1611 	    bw_meter_receive_packet(x, plen, &now);
1612     }
1613 
1614     return 0;
1615 }
1616 
1617 /*
1618  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1619  */
1620 static int
1621 X_legal_vif_num(int vif)
1622 {
1623 	int ret;
1624 
1625 	ret = 0;
1626 	if (vif < 0)
1627 		return (ret);
1628 
1629 	VIF_LOCK();
1630 	if (vif < V_numvifs)
1631 		ret = 1;
1632 	VIF_UNLOCK();
1633 
1634 	return (ret);
1635 }
1636 
1637 /*
1638  * Return the local address used by this vif
1639  */
1640 static u_long
1641 X_ip_mcast_src(int vifi)
1642 {
1643 	in_addr_t addr;
1644 
1645 	addr = INADDR_ANY;
1646 	if (vifi < 0)
1647 		return (addr);
1648 
1649 	VIF_LOCK();
1650 	if (vifi < V_numvifs)
1651 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1652 	VIF_UNLOCK();
1653 
1654 	return (addr);
1655 }
1656 
1657 static void
1658 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1659 {
1660     struct mbuf *mb_copy;
1661     int hlen = ip->ip_hl << 2;
1662 
1663     VIF_LOCK_ASSERT();
1664 
1665     /*
1666      * Make a new reference to the packet; make sure that
1667      * the IP header is actually copied, not just referenced,
1668      * so that ip_output() only scribbles on the copy.
1669      */
1670     mb_copy = m_copypacket(m, M_NOWAIT);
1671     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1672 	mb_copy = m_pullup(mb_copy, hlen);
1673     if (mb_copy == NULL)
1674 	return;
1675 
1676     send_packet(vifp, mb_copy);
1677 }
1678 
1679 static void
1680 send_packet(struct vif *vifp, struct mbuf *m)
1681 {
1682 	struct ip_moptions imo;
1683 	struct in_multi *imm[2];
1684 	int error __unused;
1685 
1686 	VIF_LOCK_ASSERT();
1687 
1688 	imo.imo_multicast_ifp  = vifp->v_ifp;
1689 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1690 	imo.imo_multicast_loop = 1;
1691 	imo.imo_multicast_vif  = -1;
1692 	imo.imo_num_memberships = 0;
1693 	imo.imo_max_memberships = 2;
1694 	imo.imo_membership  = &imm[0];
1695 
1696 	/*
1697 	 * Re-entrancy should not be a problem here, because
1698 	 * the packets that we send out and are looped back at us
1699 	 * should get rejected because they appear to come from
1700 	 * the loopback interface, thus preventing looping.
1701 	 */
1702 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1703 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1704 	    (ptrdiff_t)(vifp - V_viftable), error);
1705 }
1706 
1707 /*
1708  * Stubs for old RSVP socket shim implementation.
1709  */
1710 
1711 static int
1712 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1713 {
1714 
1715 	return (EOPNOTSUPP);
1716 }
1717 
1718 static void
1719 X_ip_rsvp_force_done(struct socket *so __unused)
1720 {
1721 
1722 }
1723 
1724 static int
1725 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1726 {
1727 	struct mbuf *m;
1728 
1729 	m = *mp;
1730 	*mp = NULL;
1731 	if (!V_rsvp_on)
1732 		m_freem(m);
1733 	return (IPPROTO_DONE);
1734 }
1735 
1736 /*
1737  * Code for bandwidth monitors
1738  */
1739 
1740 /*
1741  * Define common interface for timeval-related methods
1742  */
1743 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1744 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1745 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1746 
1747 static uint32_t
1748 compute_bw_meter_flags(struct bw_upcall *req)
1749 {
1750     uint32_t flags = 0;
1751 
1752     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1753 	flags |= BW_METER_UNIT_PACKETS;
1754     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1755 	flags |= BW_METER_UNIT_BYTES;
1756     if (req->bu_flags & BW_UPCALL_GEQ)
1757 	flags |= BW_METER_GEQ;
1758     if (req->bu_flags & BW_UPCALL_LEQ)
1759 	flags |= BW_METER_LEQ;
1760 
1761     return flags;
1762 }
1763 
1764 /*
1765  * Add a bw_meter entry
1766  */
1767 static int
1768 add_bw_upcall(struct bw_upcall *req)
1769 {
1770     struct mfc *mfc;
1771     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1772 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1773     struct timeval now;
1774     struct bw_meter *x;
1775     uint32_t flags;
1776 
1777     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1778 	return EOPNOTSUPP;
1779 
1780     /* Test if the flags are valid */
1781     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1782 	return EINVAL;
1783     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1784 	return EINVAL;
1785     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1786 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1787 	return EINVAL;
1788 
1789     /* Test if the threshold time interval is valid */
1790     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1791 	return EINVAL;
1792 
1793     flags = compute_bw_meter_flags(req);
1794 
1795     /*
1796      * Find if we have already same bw_meter entry
1797      */
1798     MFC_LOCK();
1799     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1800     if (mfc == NULL) {
1801 	MFC_UNLOCK();
1802 	return EADDRNOTAVAIL;
1803     }
1804     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1805 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1806 			   &req->bu_threshold.b_time, ==)) &&
1807 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1808 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1809 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1810 	    MFC_UNLOCK();
1811 	    return 0;		/* XXX Already installed */
1812 	}
1813     }
1814 
1815     /* Allocate the new bw_meter entry */
1816     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1817     if (x == NULL) {
1818 	MFC_UNLOCK();
1819 	return ENOBUFS;
1820     }
1821 
1822     /* Set the new bw_meter entry */
1823     x->bm_threshold.b_time = req->bu_threshold.b_time;
1824     microtime(&now);
1825     x->bm_start_time = now;
1826     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1827     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1828     x->bm_measured.b_packets = 0;
1829     x->bm_measured.b_bytes = 0;
1830     x->bm_flags = flags;
1831     x->bm_time_next = NULL;
1832     x->bm_time_hash = BW_METER_BUCKETS;
1833 
1834     /* Add the new bw_meter entry to the front of entries for this MFC */
1835     x->bm_mfc = mfc;
1836     x->bm_mfc_next = mfc->mfc_bw_meter;
1837     mfc->mfc_bw_meter = x;
1838     schedule_bw_meter(x, &now);
1839     MFC_UNLOCK();
1840 
1841     return 0;
1842 }
1843 
1844 static void
1845 free_bw_list(struct bw_meter *list)
1846 {
1847     while (list != NULL) {
1848 	struct bw_meter *x = list;
1849 
1850 	list = list->bm_mfc_next;
1851 	unschedule_bw_meter(x);
1852 	free(x, M_BWMETER);
1853     }
1854 }
1855 
1856 /*
1857  * Delete one or multiple bw_meter entries
1858  */
1859 static int
1860 del_bw_upcall(struct bw_upcall *req)
1861 {
1862     struct mfc *mfc;
1863     struct bw_meter *x;
1864 
1865     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1866 	return EOPNOTSUPP;
1867 
1868     MFC_LOCK();
1869 
1870     /* Find the corresponding MFC entry */
1871     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1872     if (mfc == NULL) {
1873 	MFC_UNLOCK();
1874 	return EADDRNOTAVAIL;
1875     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1876 	/*
1877 	 * Delete all bw_meter entries for this mfc
1878 	 */
1879 	struct bw_meter *list;
1880 
1881 	list = mfc->mfc_bw_meter;
1882 	mfc->mfc_bw_meter = NULL;
1883 	free_bw_list(list);
1884 	MFC_UNLOCK();
1885 	return 0;
1886     } else {			/* Delete a single bw_meter entry */
1887 	struct bw_meter *prev;
1888 	uint32_t flags = 0;
1889 
1890 	flags = compute_bw_meter_flags(req);
1891 
1892 	/* Find the bw_meter entry to delete */
1893 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1894 	     prev = x, x = x->bm_mfc_next) {
1895 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1896 			       &req->bu_threshold.b_time, ==)) &&
1897 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1898 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1899 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1900 		break;
1901 	}
1902 	if (x != NULL) { /* Delete entry from the list for this MFC */
1903 	    if (prev != NULL)
1904 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1905 	    else
1906 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1907 
1908 	    unschedule_bw_meter(x);
1909 	    MFC_UNLOCK();
1910 	    /* Free the bw_meter entry */
1911 	    free(x, M_BWMETER);
1912 	    return 0;
1913 	} else {
1914 	    MFC_UNLOCK();
1915 	    return EINVAL;
1916 	}
1917     }
1918     /* NOTREACHED */
1919 }
1920 
1921 /*
1922  * Perform bandwidth measurement processing that may result in an upcall
1923  */
1924 static void
1925 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1926 {
1927     struct timeval delta;
1928 
1929     MFC_LOCK_ASSERT();
1930 
1931     delta = *nowp;
1932     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1933 
1934     if (x->bm_flags & BW_METER_GEQ) {
1935 	/*
1936 	 * Processing for ">=" type of bw_meter entry
1937 	 */
1938 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1939 	    /* Reset the bw_meter entry */
1940 	    x->bm_start_time = *nowp;
1941 	    x->bm_measured.b_packets = 0;
1942 	    x->bm_measured.b_bytes = 0;
1943 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1944 	}
1945 
1946 	/* Record that a packet is received */
1947 	x->bm_measured.b_packets++;
1948 	x->bm_measured.b_bytes += plen;
1949 
1950 	/*
1951 	 * Test if we should deliver an upcall
1952 	 */
1953 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1954 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1955 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1956 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1957 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1958 		/* Prepare an upcall for delivery */
1959 		bw_meter_prepare_upcall(x, nowp);
1960 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1961 	    }
1962 	}
1963     } else if (x->bm_flags & BW_METER_LEQ) {
1964 	/*
1965 	 * Processing for "<=" type of bw_meter entry
1966 	 */
1967 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1968 	    /*
1969 	     * We are behind time with the multicast forwarding table
1970 	     * scanning for "<=" type of bw_meter entries, so test now
1971 	     * if we should deliver an upcall.
1972 	     */
1973 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1974 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1975 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1976 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1977 		/* Prepare an upcall for delivery */
1978 		bw_meter_prepare_upcall(x, nowp);
1979 	    }
1980 	    /* Reschedule the bw_meter entry */
1981 	    unschedule_bw_meter(x);
1982 	    schedule_bw_meter(x, nowp);
1983 	}
1984 
1985 	/* Record that a packet is received */
1986 	x->bm_measured.b_packets++;
1987 	x->bm_measured.b_bytes += plen;
1988 
1989 	/*
1990 	 * Test if we should restart the measuring interval
1991 	 */
1992 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1993 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1994 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1995 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1996 	    /* Don't restart the measuring interval */
1997 	} else {
1998 	    /* Do restart the measuring interval */
1999 	    /*
2000 	     * XXX: note that we don't unschedule and schedule, because this
2001 	     * might be too much overhead per packet. Instead, when we process
2002 	     * all entries for a given timer hash bin, we check whether it is
2003 	     * really a timeout. If not, we reschedule at that time.
2004 	     */
2005 	    x->bm_start_time = *nowp;
2006 	    x->bm_measured.b_packets = 0;
2007 	    x->bm_measured.b_bytes = 0;
2008 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2009 	}
2010     }
2011 }
2012 
2013 /*
2014  * Prepare a bandwidth-related upcall
2015  */
2016 static void
2017 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2018 {
2019     struct timeval delta;
2020     struct bw_upcall *u;
2021 
2022     MFC_LOCK_ASSERT();
2023 
2024     /*
2025      * Compute the measured time interval
2026      */
2027     delta = *nowp;
2028     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2029 
2030     /*
2031      * If there are too many pending upcalls, deliver them now
2032      */
2033     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2034 	bw_upcalls_send();
2035 
2036     /*
2037      * Set the bw_upcall entry
2038      */
2039     u = &V_bw_upcalls[V_bw_upcalls_n++];
2040     u->bu_src = x->bm_mfc->mfc_origin;
2041     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2042     u->bu_threshold.b_time = x->bm_threshold.b_time;
2043     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2044     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2045     u->bu_measured.b_time = delta;
2046     u->bu_measured.b_packets = x->bm_measured.b_packets;
2047     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2048     u->bu_flags = 0;
2049     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2050 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2051     if (x->bm_flags & BW_METER_UNIT_BYTES)
2052 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2053     if (x->bm_flags & BW_METER_GEQ)
2054 	u->bu_flags |= BW_UPCALL_GEQ;
2055     if (x->bm_flags & BW_METER_LEQ)
2056 	u->bu_flags |= BW_UPCALL_LEQ;
2057 }
2058 
2059 /*
2060  * Send the pending bandwidth-related upcalls
2061  */
2062 static void
2063 bw_upcalls_send(void)
2064 {
2065     struct mbuf *m;
2066     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2067     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2068     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2069 				      0,		/* unused2 */
2070 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2071 				      0,		/* im_mbz  */
2072 				      0,		/* im_vif  */
2073 				      0,		/* unused3 */
2074 				      { 0 },		/* im_src  */
2075 				      { 0 } };		/* im_dst  */
2076 
2077     MFC_LOCK_ASSERT();
2078 
2079     if (V_bw_upcalls_n == 0)
2080 	return;			/* No pending upcalls */
2081 
2082     V_bw_upcalls_n = 0;
2083 
2084     /*
2085      * Allocate a new mbuf, initialize it with the header and
2086      * the payload for the pending calls.
2087      */
2088     m = m_gethdr(M_NOWAIT, MT_DATA);
2089     if (m == NULL) {
2090 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2091 	return;
2092     }
2093 
2094     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2095     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2096 
2097     /*
2098      * Send the upcalls
2099      * XXX do we need to set the address in k_igmpsrc ?
2100      */
2101     MRTSTAT_INC(mrts_upcalls);
2102     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2103 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2104 	MRTSTAT_INC(mrts_upq_sockfull);
2105     }
2106 }
2107 
2108 /*
2109  * Compute the timeout hash value for the bw_meter entries
2110  */
2111 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2112     do {								\
2113 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2114 									\
2115 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2116 	(hash) = next_timeval.tv_sec;					\
2117 	if (next_timeval.tv_usec)					\
2118 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2119 	(hash) %= BW_METER_BUCKETS;					\
2120     } while (0)
2121 
2122 /*
2123  * Schedule a timer to process periodically bw_meter entry of type "<="
2124  * by linking the entry in the proper hash bucket.
2125  */
2126 static void
2127 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2128 {
2129     int time_hash;
2130 
2131     MFC_LOCK_ASSERT();
2132 
2133     if (!(x->bm_flags & BW_METER_LEQ))
2134 	return;		/* XXX: we schedule timers only for "<=" entries */
2135 
2136     /*
2137      * Reset the bw_meter entry
2138      */
2139     x->bm_start_time = *nowp;
2140     x->bm_measured.b_packets = 0;
2141     x->bm_measured.b_bytes = 0;
2142     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2143 
2144     /*
2145      * Compute the timeout hash value and insert the entry
2146      */
2147     BW_METER_TIMEHASH(x, time_hash);
2148     x->bm_time_next = V_bw_meter_timers[time_hash];
2149     V_bw_meter_timers[time_hash] = x;
2150     x->bm_time_hash = time_hash;
2151 }
2152 
2153 /*
2154  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2155  * by removing the entry from the proper hash bucket.
2156  */
2157 static void
2158 unschedule_bw_meter(struct bw_meter *x)
2159 {
2160     int time_hash;
2161     struct bw_meter *prev, *tmp;
2162 
2163     MFC_LOCK_ASSERT();
2164 
2165     if (!(x->bm_flags & BW_METER_LEQ))
2166 	return;		/* XXX: we schedule timers only for "<=" entries */
2167 
2168     /*
2169      * Compute the timeout hash value and delete the entry
2170      */
2171     time_hash = x->bm_time_hash;
2172     if (time_hash >= BW_METER_BUCKETS)
2173 	return;		/* Entry was not scheduled */
2174 
2175     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2176 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2177 	if (tmp == x)
2178 	    break;
2179 
2180     if (tmp == NULL)
2181 	panic("unschedule_bw_meter: bw_meter entry not found");
2182 
2183     if (prev != NULL)
2184 	prev->bm_time_next = x->bm_time_next;
2185     else
2186 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2187 
2188     x->bm_time_next = NULL;
2189     x->bm_time_hash = BW_METER_BUCKETS;
2190 }
2191 
2192 
2193 /*
2194  * Process all "<=" type of bw_meter that should be processed now,
2195  * and for each entry prepare an upcall if necessary. Each processed
2196  * entry is rescheduled again for the (periodic) processing.
2197  *
2198  * This is run periodically (once per second normally). On each round,
2199  * all the potentially matching entries are in the hash slot that we are
2200  * looking at.
2201  */
2202 static void
2203 bw_meter_process()
2204 {
2205     uint32_t loops;
2206     int i;
2207     struct timeval now, process_endtime;
2208 
2209     microtime(&now);
2210     if (V_last_tv_sec == now.tv_sec)
2211 	return;		/* nothing to do */
2212 
2213     loops = now.tv_sec - V_last_tv_sec;
2214     V_last_tv_sec = now.tv_sec;
2215     if (loops > BW_METER_BUCKETS)
2216 	loops = BW_METER_BUCKETS;
2217 
2218     MFC_LOCK();
2219     /*
2220      * Process all bins of bw_meter entries from the one after the last
2221      * processed to the current one. On entry, i points to the last bucket
2222      * visited, so we need to increment i at the beginning of the loop.
2223      */
2224     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2225 	struct bw_meter *x, *tmp_list;
2226 
2227 	if (++i >= BW_METER_BUCKETS)
2228 	    i = 0;
2229 
2230 	/* Disconnect the list of bw_meter entries from the bin */
2231 	tmp_list = V_bw_meter_timers[i];
2232 	V_bw_meter_timers[i] = NULL;
2233 
2234 	/* Process the list of bw_meter entries */
2235 	while (tmp_list != NULL) {
2236 	    x = tmp_list;
2237 	    tmp_list = tmp_list->bm_time_next;
2238 
2239 	    /* Test if the time interval is over */
2240 	    process_endtime = x->bm_start_time;
2241 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2242 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2243 		/* Not yet: reschedule, but don't reset */
2244 		int time_hash;
2245 
2246 		BW_METER_TIMEHASH(x, time_hash);
2247 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2248 		    /*
2249 		     * XXX: somehow the bin processing is a bit ahead of time.
2250 		     * Put the entry in the next bin.
2251 		     */
2252 		    if (++time_hash >= BW_METER_BUCKETS)
2253 			time_hash = 0;
2254 		}
2255 		x->bm_time_next = V_bw_meter_timers[time_hash];
2256 		V_bw_meter_timers[time_hash] = x;
2257 		x->bm_time_hash = time_hash;
2258 
2259 		continue;
2260 	    }
2261 
2262 	    /*
2263 	     * Test if we should deliver an upcall
2264 	     */
2265 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2266 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2267 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2268 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2269 		/* Prepare an upcall for delivery */
2270 		bw_meter_prepare_upcall(x, &now);
2271 	    }
2272 
2273 	    /*
2274 	     * Reschedule for next processing
2275 	     */
2276 	    schedule_bw_meter(x, &now);
2277 	}
2278     }
2279 
2280     /* Send all upcalls that are pending delivery */
2281     bw_upcalls_send();
2282 
2283     MFC_UNLOCK();
2284 }
2285 
2286 /*
2287  * A periodic function for sending all upcalls that are pending delivery
2288  */
2289 static void
2290 expire_bw_upcalls_send(void *arg)
2291 {
2292     CURVNET_SET((struct vnet *) arg);
2293 
2294     MFC_LOCK();
2295     bw_upcalls_send();
2296     MFC_UNLOCK();
2297 
2298     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2299 	curvnet);
2300     CURVNET_RESTORE();
2301 }
2302 
2303 /*
2304  * A periodic function for periodic scanning of the multicast forwarding
2305  * table for processing all "<=" bw_meter entries.
2306  */
2307 static void
2308 expire_bw_meter_process(void *arg)
2309 {
2310     CURVNET_SET((struct vnet *) arg);
2311 
2312     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2313 	bw_meter_process();
2314 
2315     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2316 	curvnet);
2317     CURVNET_RESTORE();
2318 }
2319 
2320 /*
2321  * End of bandwidth monitoring code
2322  */
2323 
2324 /*
2325  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2326  *
2327  */
2328 static int
2329 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2330     struct mfc *rt)
2331 {
2332     struct mbuf *mb_copy, *mm;
2333 
2334     /*
2335      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2336      * rendezvous point was unspecified, and we were told not to.
2337      */
2338     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2339 	in_nullhost(rt->mfc_rp))
2340 	return 0;
2341 
2342     mb_copy = pim_register_prepare(ip, m);
2343     if (mb_copy == NULL)
2344 	return ENOBUFS;
2345 
2346     /*
2347      * Send all the fragments. Note that the mbuf for each fragment
2348      * is freed by the sending machinery.
2349      */
2350     for (mm = mb_copy; mm; mm = mb_copy) {
2351 	mb_copy = mm->m_nextpkt;
2352 	mm->m_nextpkt = 0;
2353 	mm = m_pullup(mm, sizeof(struct ip));
2354 	if (mm != NULL) {
2355 	    ip = mtod(mm, struct ip *);
2356 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2357 		pim_register_send_rp(ip, vifp, mm, rt);
2358 	    } else {
2359 		pim_register_send_upcall(ip, vifp, mm, rt);
2360 	    }
2361 	}
2362     }
2363 
2364     return 0;
2365 }
2366 
2367 /*
2368  * Return a copy of the data packet that is ready for PIM Register
2369  * encapsulation.
2370  * XXX: Note that in the returned copy the IP header is a valid one.
2371  */
2372 static struct mbuf *
2373 pim_register_prepare(struct ip *ip, struct mbuf *m)
2374 {
2375     struct mbuf *mb_copy = NULL;
2376     int mtu;
2377 
2378     /* Take care of delayed checksums */
2379     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2380 	in_delayed_cksum(m);
2381 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2382     }
2383 
2384     /*
2385      * Copy the old packet & pullup its IP header into the
2386      * new mbuf so we can modify it.
2387      */
2388     mb_copy = m_copypacket(m, M_NOWAIT);
2389     if (mb_copy == NULL)
2390 	return NULL;
2391     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2392     if (mb_copy == NULL)
2393 	return NULL;
2394 
2395     /* take care of the TTL */
2396     ip = mtod(mb_copy, struct ip *);
2397     --ip->ip_ttl;
2398 
2399     /* Compute the MTU after the PIM Register encapsulation */
2400     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2401 
2402     if (ntohs(ip->ip_len) <= mtu) {
2403 	/* Turn the IP header into a valid one */
2404 	ip->ip_sum = 0;
2405 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2406     } else {
2407 	/* Fragment the packet */
2408 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2409 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2410 	    m_freem(mb_copy);
2411 	    return NULL;
2412 	}
2413     }
2414     return mb_copy;
2415 }
2416 
2417 /*
2418  * Send an upcall with the data packet to the user-level process.
2419  */
2420 static int
2421 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2422     struct mbuf *mb_copy, struct mfc *rt)
2423 {
2424     struct mbuf *mb_first;
2425     int len = ntohs(ip->ip_len);
2426     struct igmpmsg *im;
2427     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2428 
2429     VIF_LOCK_ASSERT();
2430 
2431     /*
2432      * Add a new mbuf with an upcall header
2433      */
2434     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2435     if (mb_first == NULL) {
2436 	m_freem(mb_copy);
2437 	return ENOBUFS;
2438     }
2439     mb_first->m_data += max_linkhdr;
2440     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2441     mb_first->m_len = sizeof(struct igmpmsg);
2442     mb_first->m_next = mb_copy;
2443 
2444     /* Send message to routing daemon */
2445     im = mtod(mb_first, struct igmpmsg *);
2446     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2447     im->im_mbz		= 0;
2448     im->im_vif		= vifp - V_viftable;
2449     im->im_src		= ip->ip_src;
2450     im->im_dst		= ip->ip_dst;
2451 
2452     k_igmpsrc.sin_addr	= ip->ip_src;
2453 
2454     MRTSTAT_INC(mrts_upcalls);
2455 
2456     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2457 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2458 	MRTSTAT_INC(mrts_upq_sockfull);
2459 	return ENOBUFS;
2460     }
2461 
2462     /* Keep statistics */
2463     PIMSTAT_INC(pims_snd_registers_msgs);
2464     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2465 
2466     return 0;
2467 }
2468 
2469 /*
2470  * Encapsulate the data packet in PIM Register message and send it to the RP.
2471  */
2472 static int
2473 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2474     struct mfc *rt)
2475 {
2476     struct mbuf *mb_first;
2477     struct ip *ip_outer;
2478     struct pim_encap_pimhdr *pimhdr;
2479     int len = ntohs(ip->ip_len);
2480     vifi_t vifi = rt->mfc_parent;
2481 
2482     VIF_LOCK_ASSERT();
2483 
2484     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2485 	m_freem(mb_copy);
2486 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2487     }
2488 
2489     /*
2490      * Add a new mbuf with the encapsulating header
2491      */
2492     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2493     if (mb_first == NULL) {
2494 	m_freem(mb_copy);
2495 	return ENOBUFS;
2496     }
2497     mb_first->m_data += max_linkhdr;
2498     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2499     mb_first->m_next = mb_copy;
2500 
2501     mb_first->m_pkthdr.len = len + mb_first->m_len;
2502 
2503     /*
2504      * Fill in the encapsulating IP and PIM header
2505      */
2506     ip_outer = mtod(mb_first, struct ip *);
2507     *ip_outer = pim_encap_iphdr;
2508     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2509 	sizeof(pim_encap_pimhdr));
2510     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2511     ip_outer->ip_dst = rt->mfc_rp;
2512     /*
2513      * Copy the inner header TOS to the outer header, and take care of the
2514      * IP_DF bit.
2515      */
2516     ip_outer->ip_tos = ip->ip_tos;
2517     if (ip->ip_off & htons(IP_DF))
2518 	ip_outer->ip_off |= htons(IP_DF);
2519     ip_fillid(ip_outer);
2520     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2521 					 + sizeof(pim_encap_iphdr));
2522     *pimhdr = pim_encap_pimhdr;
2523     /* If the iif crosses a border, set the Border-bit */
2524     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2525 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2526 
2527     mb_first->m_data += sizeof(pim_encap_iphdr);
2528     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2529     mb_first->m_data -= sizeof(pim_encap_iphdr);
2530 
2531     send_packet(vifp, mb_first);
2532 
2533     /* Keep statistics */
2534     PIMSTAT_INC(pims_snd_registers_msgs);
2535     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2536 
2537     return 0;
2538 }
2539 
2540 /*
2541  * pim_encapcheck() is called by the encap4_input() path at runtime to
2542  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2543  * into the kernel.
2544  */
2545 static int
2546 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2547     int proto __unused, void *arg __unused)
2548 {
2549 
2550     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2551     return (8);		/* claim the datagram. */
2552 }
2553 
2554 /*
2555  * PIM-SMv2 and PIM-DM messages processing.
2556  * Receives and verifies the PIM control messages, and passes them
2557  * up to the listening socket, using rip_input().
2558  * The only message with special processing is the PIM_REGISTER message
2559  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2560  * is passed to if_simloop().
2561  */
2562 static int
2563 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2564 {
2565     struct ip *ip = mtod(m, struct ip *);
2566     struct pim *pim;
2567     int iphlen = off;
2568     int minlen;
2569     int datalen = ntohs(ip->ip_len) - iphlen;
2570     int ip_tos;
2571 
2572     /* Keep statistics */
2573     PIMSTAT_INC(pims_rcv_total_msgs);
2574     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2575 
2576     /*
2577      * Validate lengths
2578      */
2579     if (datalen < PIM_MINLEN) {
2580 	PIMSTAT_INC(pims_rcv_tooshort);
2581 	CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2582 	    __func__, datalen, ntohl(ip->ip_src.s_addr));
2583 	m_freem(m);
2584 	return (IPPROTO_DONE);
2585     }
2586 
2587     /*
2588      * If the packet is at least as big as a REGISTER, go agead
2589      * and grab the PIM REGISTER header size, to avoid another
2590      * possible m_pullup() later.
2591      *
2592      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2593      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2594      */
2595     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2596     /*
2597      * Get the IP and PIM headers in contiguous memory, and
2598      * possibly the PIM REGISTER header.
2599      */
2600     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2601 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2602 	return (IPPROTO_DONE);
2603     }
2604 
2605     /* m_pullup() may have given us a new mbuf so reset ip. */
2606     ip = mtod(m, struct ip *);
2607     ip_tos = ip->ip_tos;
2608 
2609     /* adjust mbuf to point to the PIM header */
2610     m->m_data += iphlen;
2611     m->m_len  -= iphlen;
2612     pim = mtod(m, struct pim *);
2613 
2614     /*
2615      * Validate checksum. If PIM REGISTER, exclude the data packet.
2616      *
2617      * XXX: some older PIMv2 implementations don't make this distinction,
2618      * so for compatibility reason perform the checksum over part of the
2619      * message, and if error, then over the whole message.
2620      */
2621     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2622 	/* do nothing, checksum okay */
2623     } else if (in_cksum(m, datalen)) {
2624 	PIMSTAT_INC(pims_rcv_badsum);
2625 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2626 	m_freem(m);
2627 	return (IPPROTO_DONE);
2628     }
2629 
2630     /* PIM version check */
2631     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2632 	PIMSTAT_INC(pims_rcv_badversion);
2633 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2634 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2635 	m_freem(m);
2636 	return (IPPROTO_DONE);
2637     }
2638 
2639     /* restore mbuf back to the outer IP */
2640     m->m_data -= iphlen;
2641     m->m_len  += iphlen;
2642 
2643     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2644 	/*
2645 	 * Since this is a REGISTER, we'll make a copy of the register
2646 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2647 	 * routing daemon.
2648 	 */
2649 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2650 	struct mbuf *mcp;
2651 	struct ip *encap_ip;
2652 	u_int32_t *reghdr;
2653 	struct ifnet *vifp;
2654 
2655 	VIF_LOCK();
2656 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2657 	    VIF_UNLOCK();
2658 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2659 		(int)V_reg_vif_num);
2660 	    m_freem(m);
2661 	    return (IPPROTO_DONE);
2662 	}
2663 	/* XXX need refcnt? */
2664 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2665 	VIF_UNLOCK();
2666 
2667 	/*
2668 	 * Validate length
2669 	 */
2670 	if (datalen < PIM_REG_MINLEN) {
2671 	    PIMSTAT_INC(pims_rcv_tooshort);
2672 	    PIMSTAT_INC(pims_rcv_badregisters);
2673 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2674 	    m_freem(m);
2675 	    return (IPPROTO_DONE);
2676 	}
2677 
2678 	reghdr = (u_int32_t *)(pim + 1);
2679 	encap_ip = (struct ip *)(reghdr + 1);
2680 
2681 	CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2682 	    __func__, ntohl(encap_ip->ip_src.s_addr),
2683 	    ntohs(encap_ip->ip_len));
2684 
2685 	/* verify the version number of the inner packet */
2686 	if (encap_ip->ip_v != IPVERSION) {
2687 	    PIMSTAT_INC(pims_rcv_badregisters);
2688 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2689 	    m_freem(m);
2690 	    return (IPPROTO_DONE);
2691 	}
2692 
2693 	/* verify the inner packet is destined to a mcast group */
2694 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2695 	    PIMSTAT_INC(pims_rcv_badregisters);
2696 	    CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2697 		ntohl(encap_ip->ip_dst.s_addr));
2698 	    m_freem(m);
2699 	    return (IPPROTO_DONE);
2700 	}
2701 
2702 	/* If a NULL_REGISTER, pass it to the daemon */
2703 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2704 	    goto pim_input_to_daemon;
2705 
2706 	/*
2707 	 * Copy the TOS from the outer IP header to the inner IP header.
2708 	 */
2709 	if (encap_ip->ip_tos != ip_tos) {
2710 	    /* Outer TOS -> inner TOS */
2711 	    encap_ip->ip_tos = ip_tos;
2712 	    /* Recompute the inner header checksum. Sigh... */
2713 
2714 	    /* adjust mbuf to point to the inner IP header */
2715 	    m->m_data += (iphlen + PIM_MINLEN);
2716 	    m->m_len  -= (iphlen + PIM_MINLEN);
2717 
2718 	    encap_ip->ip_sum = 0;
2719 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2720 
2721 	    /* restore mbuf to point back to the outer IP header */
2722 	    m->m_data -= (iphlen + PIM_MINLEN);
2723 	    m->m_len  += (iphlen + PIM_MINLEN);
2724 	}
2725 
2726 	/*
2727 	 * Decapsulate the inner IP packet and loopback to forward it
2728 	 * as a normal multicast packet. Also, make a copy of the
2729 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2730 	 * to pass to the daemon later, so it can take the appropriate
2731 	 * actions (e.g., send back PIM_REGISTER_STOP).
2732 	 * XXX: here m->m_data points to the outer IP header.
2733 	 */
2734 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2735 	if (mcp == NULL) {
2736 	    CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2737 	    m_freem(m);
2738 	    return (IPPROTO_DONE);
2739 	}
2740 
2741 	/* Keep statistics */
2742 	/* XXX: registers_bytes include only the encap. mcast pkt */
2743 	PIMSTAT_INC(pims_rcv_registers_msgs);
2744 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2745 
2746 	/*
2747 	 * forward the inner ip packet; point m_data at the inner ip.
2748 	 */
2749 	m_adj(m, iphlen + PIM_MINLEN);
2750 
2751 	CTR4(KTR_IPMF,
2752 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2753 	    __func__,
2754 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2755 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2756 	    (int)V_reg_vif_num);
2757 
2758 	/* NB: vifp was collected above; can it change on us? */
2759 	if_simloop(vifp, m, dst.sin_family, 0);
2760 
2761 	/* prepare the register head to send to the mrouting daemon */
2762 	m = mcp;
2763     }
2764 
2765 pim_input_to_daemon:
2766     /*
2767      * Pass the PIM message up to the daemon; if it is a Register message,
2768      * pass the 'head' only up to the daemon. This includes the
2769      * outer IP header, PIM header, PIM-Register header and the
2770      * inner IP header.
2771      * XXX: the outer IP header pkt size of a Register is not adjust to
2772      * reflect the fact that the inner multicast data is truncated.
2773      */
2774     return (rip_input(&m, &off, proto));
2775 }
2776 
2777 static int
2778 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2779 {
2780 	struct mfc	*rt;
2781 	int		 error, i;
2782 
2783 	if (req->newptr)
2784 		return (EPERM);
2785 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2786 		return (0);
2787 	error = sysctl_wire_old_buffer(req, 0);
2788 	if (error)
2789 		return (error);
2790 
2791 	MFC_LOCK();
2792 	for (i = 0; i < mfchashsize; i++) {
2793 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2794 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2795 			if (error)
2796 				goto out_locked;
2797 		}
2798 	}
2799 out_locked:
2800 	MFC_UNLOCK();
2801 	return (error);
2802 }
2803 
2804 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2805     sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2806     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2807 
2808 static void
2809 vnet_mroute_init(const void *unused __unused)
2810 {
2811 
2812 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2813 	bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2814 	callout_init(&V_expire_upcalls_ch, 1);
2815 	callout_init(&V_bw_upcalls_ch, 1);
2816 	callout_init(&V_bw_meter_ch, 1);
2817 }
2818 
2819 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2820 	NULL);
2821 
2822 static void
2823 vnet_mroute_uninit(const void *unused __unused)
2824 {
2825 
2826 	free(V_nexpire, M_MRTABLE);
2827 	V_nexpire = NULL;
2828 }
2829 
2830 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2831 	vnet_mroute_uninit, NULL);
2832 
2833 static int
2834 ip_mroute_modevent(module_t mod, int type, void *unused)
2835 {
2836 
2837     switch (type) {
2838     case MOD_LOAD:
2839 	MROUTER_LOCK_INIT();
2840 
2841 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2842 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2843 	if (if_detach_event_tag == NULL) {
2844 		printf("ip_mroute: unable to register "
2845 		    "ifnet_departure_event handler\n");
2846 		MROUTER_LOCK_DESTROY();
2847 		return (EINVAL);
2848 	}
2849 
2850 	MFC_LOCK_INIT();
2851 	VIF_LOCK_INIT();
2852 
2853 	mfchashsize = MFCHASHSIZE;
2854 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2855 	    !powerof2(mfchashsize)) {
2856 		printf("WARNING: %s not a power of 2; using default\n",
2857 		    "net.inet.ip.mfchashsize");
2858 		mfchashsize = MFCHASHSIZE;
2859 	}
2860 
2861 	pim_squelch_wholepkt = 0;
2862 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2863 	    &pim_squelch_wholepkt);
2864 
2865 	pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2866 	if (pim_encap_cookie == NULL) {
2867 		printf("ip_mroute: unable to attach pim encap\n");
2868 		VIF_LOCK_DESTROY();
2869 		MFC_LOCK_DESTROY();
2870 		MROUTER_LOCK_DESTROY();
2871 		return (EINVAL);
2872 	}
2873 
2874 	ip_mcast_src = X_ip_mcast_src;
2875 	ip_mforward = X_ip_mforward;
2876 	ip_mrouter_done = X_ip_mrouter_done;
2877 	ip_mrouter_get = X_ip_mrouter_get;
2878 	ip_mrouter_set = X_ip_mrouter_set;
2879 
2880 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2881 	ip_rsvp_vif = X_ip_rsvp_vif;
2882 
2883 	legal_vif_num = X_legal_vif_num;
2884 	mrt_ioctl = X_mrt_ioctl;
2885 	rsvp_input_p = X_rsvp_input;
2886 	break;
2887 
2888     case MOD_UNLOAD:
2889 	/*
2890 	 * Typically module unload happens after the user-level
2891 	 * process has shutdown the kernel services (the check
2892 	 * below insures someone can't just yank the module out
2893 	 * from under a running process).  But if the module is
2894 	 * just loaded and then unloaded w/o starting up a user
2895 	 * process we still need to cleanup.
2896 	 */
2897 	MROUTER_LOCK();
2898 	if (ip_mrouter_cnt != 0) {
2899 	    MROUTER_UNLOCK();
2900 	    return (EINVAL);
2901 	}
2902 	ip_mrouter_unloading = 1;
2903 	MROUTER_UNLOCK();
2904 
2905 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2906 
2907 	if (pim_encap_cookie) {
2908 	    ip_encap_detach(pim_encap_cookie);
2909 	    pim_encap_cookie = NULL;
2910 	}
2911 
2912 	ip_mcast_src = NULL;
2913 	ip_mforward = NULL;
2914 	ip_mrouter_done = NULL;
2915 	ip_mrouter_get = NULL;
2916 	ip_mrouter_set = NULL;
2917 
2918 	ip_rsvp_force_done = NULL;
2919 	ip_rsvp_vif = NULL;
2920 
2921 	legal_vif_num = NULL;
2922 	mrt_ioctl = NULL;
2923 	rsvp_input_p = NULL;
2924 
2925 	VIF_LOCK_DESTROY();
2926 	MFC_LOCK_DESTROY();
2927 	MROUTER_LOCK_DESTROY();
2928 	break;
2929 
2930     default:
2931 	return EOPNOTSUPP;
2932     }
2933     return 0;
2934 }
2935 
2936 static moduledata_t ip_mroutemod = {
2937     "ip_mroute",
2938     ip_mroute_modevent,
2939     0
2940 };
2941 
2942 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2943