xref: /freebsd/sys/netinet/ip_mroute.c (revision 25408c853d9ecb2e76b9e38407338f86ecb8a55c)
1 /*-
2  * Copyright (c) 1989 Stephen Deering
3  * Copyright (c) 1992, 1993
4  *      The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Stephen Deering of Stanford University.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
34  */
35 
36 /*
37  * IP multicast forwarding procedures
38  *
39  * Written by David Waitzman, BBN Labs, August 1988.
40  * Modified by Steve Deering, Stanford, February 1989.
41  * Modified by Mark J. Steiglitz, Stanford, May, 1991
42  * Modified by Van Jacobson, LBL, January 1993
43  * Modified by Ajit Thyagarajan, PARC, August 1993
44  * Modified by Bill Fenner, PARC, April 1995
45  * Modified by Ahmed Helmy, SGI, June 1996
46  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
47  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
48  * Modified by Hitoshi Asaeda, WIDE, August 2000
49  * Modified by Pavlin Radoslavov, ICSI, October 2002
50  *
51  * MROUTING Revision: 3.5
52  * and PIM-SMv2 and PIM-DM support, advanced API support,
53  * bandwidth metering and signaling
54  */
55 
56 /*
57  * TODO: Prefix functions with ipmf_.
58  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
59  * domain attachment (if_afdata) so we can track consumers of that service.
60  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
61  * move it to socket options.
62  * TODO: Cleanup LSRR removal further.
63  * TODO: Push RSVP stubs into raw_ip.c.
64  * TODO: Use bitstring.h for vif set.
65  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
66  * TODO: Sync ip6_mroute.c with this file.
67  */
68 
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
71 
72 #include "opt_inet.h"
73 #include "opt_mrouting.h"
74 
75 #define _PIM_VT 1
76 
77 #include <sys/param.h>
78 #include <sys/kernel.h>
79 #include <sys/stddef.h>
80 #include <sys/lock.h>
81 #include <sys/ktr.h>
82 #include <sys/malloc.h>
83 #include <sys/mbuf.h>
84 #include <sys/module.h>
85 #include <sys/priv.h>
86 #include <sys/protosw.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/socketvar.h>
90 #include <sys/sockio.h>
91 #include <sys/sx.h>
92 #include <sys/sysctl.h>
93 #include <sys/syslog.h>
94 #include <sys/systm.h>
95 #include <sys/time.h>
96 
97 #include <net/if.h>
98 #include <net/netisr.h>
99 #include <net/route.h>
100 #include <net/vnet.h>
101 
102 #include <netinet/in.h>
103 #include <netinet/igmp.h>
104 #include <netinet/in_systm.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip.h>
107 #include <netinet/ip_encap.h>
108 #include <netinet/ip_mroute.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/ip_options.h>
111 #include <netinet/pim.h>
112 #include <netinet/pim_var.h>
113 #include <netinet/udp.h>
114 
115 #include <machine/in_cksum.h>
116 
117 #ifndef KTR_IPMF
118 #define KTR_IPMF KTR_INET
119 #endif
120 
121 #define		VIFI_INVALID	((vifi_t) -1)
122 #define		M_HASCL(m)	((m)->m_flags & M_EXT)
123 
124 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
125 #define	V_last_tv_sec	VNET(last_tv_sec)
126 
127 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
128 
129 /*
130  * Locking.  We use two locks: one for the virtual interface table and
131  * one for the forwarding table.  These locks may be nested in which case
132  * the VIF lock must always be taken first.  Note that each lock is used
133  * to cover not only the specific data structure but also related data
134  * structures.
135  */
136 
137 static struct mtx mrouter_mtx;
138 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
139 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
140 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
141 #define	MROUTER_LOCK_INIT()						\
142 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
143 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
144 
145 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
146 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
147 
148 static VNET_DEFINE(struct mrtstat, mrtstat);
149 #define	V_mrtstat		VNET(mrtstat)
150 SYSCTL_VNET_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
151     &VNET_NAME(mrtstat), mrtstat,
152     "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
153     "netinet/ip_mroute.h)");
154 
155 static VNET_DEFINE(u_long, mfchash);
156 #define	V_mfchash		VNET(mfchash)
157 #define	MFCHASH(a, g)							\
158 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
159 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
160 #define	MFCHASHSIZE	256
161 
162 static u_long mfchashsize;			/* Hash size */
163 static VNET_DEFINE(u_char *, nexpire);		/* 0..mfchashsize-1 */
164 #define	V_nexpire		VNET(nexpire)
165 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
166 #define	V_mfchashtbl		VNET(mfchashtbl)
167 
168 static struct mtx mfc_mtx;
169 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
170 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
171 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
172 #define	MFC_LOCK_INIT()							\
173 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
174 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
175 
176 static VNET_DEFINE(vifi_t, numvifs);
177 #define	V_numvifs		VNET(numvifs)
178 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
179 #define	V_viftable		VNET(viftable)
180 SYSCTL_VNET_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
181     &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
182     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
183 
184 static struct mtx vif_mtx;
185 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
186 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
187 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
188 #define	VIF_LOCK_INIT()							\
189 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
190 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
191 
192 static eventhandler_tag if_detach_event_tag = NULL;
193 
194 static VNET_DEFINE(struct callout, expire_upcalls_ch);
195 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
196 
197 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
198 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
199 
200 /*
201  * Bandwidth meter variables and constants
202  */
203 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
204 /*
205  * Pending timeouts are stored in a hash table, the key being the
206  * expiration time. Periodically, the entries are analysed and processed.
207  */
208 #define	BW_METER_BUCKETS	1024
209 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
210 #define	V_bw_meter_timers	VNET(bw_meter_timers)
211 static VNET_DEFINE(struct callout, bw_meter_ch);
212 #define	V_bw_meter_ch		VNET(bw_meter_ch)
213 #define	BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
214 
215 /*
216  * Pending upcalls are stored in a vector which is flushed when
217  * full, or periodically
218  */
219 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
220 #define	V_bw_upcalls		VNET(bw_upcalls)
221 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
222 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
223 static VNET_DEFINE(struct callout, bw_upcalls_ch);
224 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
225 
226 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
227 
228 static VNET_DEFINE(struct pimstat, pimstat);
229 #define	V_pimstat		VNET(pimstat)
230 
231 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
232 SYSCTL_VNET_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
233     &VNET_NAME(pimstat), pimstat,
234     "PIM Statistics (struct pimstat, netinet/pim_var.h)");
235 
236 static u_long	pim_squelch_wholepkt = 0;
237 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
238     &pim_squelch_wholepkt, 0,
239     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
240 
241 extern  struct domain inetdomain;
242 static const struct protosw in_pim_protosw = {
243 	.pr_type =		SOCK_RAW,
244 	.pr_domain =		&inetdomain,
245 	.pr_protocol =		IPPROTO_PIM,
246 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
247 	.pr_input =		pim_input,
248 	.pr_output =		(pr_output_t*)rip_output,
249 	.pr_ctloutput =		rip_ctloutput,
250 	.pr_usrreqs =		&rip_usrreqs
251 };
252 static const struct encaptab *pim_encap_cookie;
253 
254 static int pim_encapcheck(const struct mbuf *, int, int, void *);
255 
256 /*
257  * Note: the PIM Register encapsulation adds the following in front of a
258  * data packet:
259  *
260  * struct pim_encap_hdr {
261  *    struct ip ip;
262  *    struct pim_encap_pimhdr  pim;
263  * }
264  *
265  */
266 
267 struct pim_encap_pimhdr {
268 	struct pim pim;
269 	uint32_t   flags;
270 };
271 #define		PIM_ENCAP_TTL	64
272 
273 static struct ip pim_encap_iphdr = {
274 #if BYTE_ORDER == LITTLE_ENDIAN
275 	sizeof(struct ip) >> 2,
276 	IPVERSION,
277 #else
278 	IPVERSION,
279 	sizeof(struct ip) >> 2,
280 #endif
281 	0,			/* tos */
282 	sizeof(struct ip),	/* total length */
283 	0,			/* id */
284 	0,			/* frag offset */
285 	PIM_ENCAP_TTL,
286 	IPPROTO_PIM,
287 	0,			/* checksum */
288 };
289 
290 static struct pim_encap_pimhdr pim_encap_pimhdr = {
291     {
292 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
293 	0,			/* reserved */
294 	0,			/* checksum */
295     },
296     0				/* flags */
297 };
298 
299 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
300 #define	V_reg_vif_num		VNET(reg_vif_num)
301 static VNET_DEFINE(struct ifnet, multicast_register_if);
302 #define	V_multicast_register_if	VNET(multicast_register_if)
303 
304 /*
305  * Private variables.
306  */
307 
308 static u_long	X_ip_mcast_src(int);
309 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
310 		    struct ip_moptions *);
311 static int	X_ip_mrouter_done(void);
312 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
313 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
314 static int	X_legal_vif_num(int);
315 static int	X_mrt_ioctl(u_long, caddr_t, int);
316 
317 static int	add_bw_upcall(struct bw_upcall *);
318 static int	add_mfc(struct mfcctl2 *);
319 static int	add_vif(struct vifctl *);
320 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
321 static void	bw_meter_process(void);
322 static void	bw_meter_receive_packet(struct bw_meter *, int,
323 		    struct timeval *);
324 static void	bw_upcalls_send(void);
325 static int	del_bw_upcall(struct bw_upcall *);
326 static int	del_mfc(struct mfcctl2 *);
327 static int	del_vif(vifi_t);
328 static int	del_vif_locked(vifi_t);
329 static void	expire_bw_meter_process(void *);
330 static void	expire_bw_upcalls_send(void *);
331 static void	expire_mfc(struct mfc *);
332 static void	expire_upcalls(void *);
333 static void	free_bw_list(struct bw_meter *);
334 static int	get_sg_cnt(struct sioc_sg_req *);
335 static int	get_vif_cnt(struct sioc_vif_req *);
336 static void	if_detached_event(void *, struct ifnet *);
337 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
338 static int	ip_mrouter_init(struct socket *, int);
339 static __inline struct mfc *
340 		mfc_find(struct in_addr *, struct in_addr *);
341 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
342 static struct mbuf *
343 		pim_register_prepare(struct ip *, struct mbuf *);
344 static int	pim_register_send(struct ip *, struct vif *,
345 		    struct mbuf *, struct mfc *);
346 static int	pim_register_send_rp(struct ip *, struct vif *,
347 		    struct mbuf *, struct mfc *);
348 static int	pim_register_send_upcall(struct ip *, struct vif *,
349 		    struct mbuf *, struct mfc *);
350 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
351 static void	send_packet(struct vif *, struct mbuf *);
352 static int	set_api_config(uint32_t *);
353 static int	set_assert(int);
354 static int	socket_send(struct socket *, struct mbuf *,
355 		    struct sockaddr_in *);
356 static void	unschedule_bw_meter(struct bw_meter *);
357 
358 /*
359  * Kernel multicast forwarding API capabilities and setup.
360  * If more API capabilities are added to the kernel, they should be
361  * recorded in `mrt_api_support'.
362  */
363 #define MRT_API_VERSION		0x0305
364 
365 static const int mrt_api_version = MRT_API_VERSION;
366 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
367 					 MRT_MFC_FLAGS_BORDER_VIF |
368 					 MRT_MFC_RP |
369 					 MRT_MFC_BW_UPCALL);
370 static VNET_DEFINE(uint32_t, mrt_api_config);
371 #define	V_mrt_api_config	VNET(mrt_api_config)
372 static VNET_DEFINE(int, pim_assert_enabled);
373 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
374 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
375 
376 /*
377  * Find a route for a given origin IP address and multicast group address.
378  * Statistics must be updated by the caller.
379  */
380 static __inline struct mfc *
381 mfc_find(struct in_addr *o, struct in_addr *g)
382 {
383 	struct mfc *rt;
384 
385 	MFC_LOCK_ASSERT();
386 
387 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
388 		if (in_hosteq(rt->mfc_origin, *o) &&
389 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
390 		    TAILQ_EMPTY(&rt->mfc_stall))
391 			break;
392 	}
393 
394 	return (rt);
395 }
396 
397 /*
398  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
399  */
400 static int
401 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
402 {
403     int	error, optval;
404     vifi_t	vifi;
405     struct	vifctl vifc;
406     struct	mfcctl2 mfc;
407     struct	bw_upcall bw_upcall;
408     uint32_t	i;
409 
410     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
411 	return EPERM;
412 
413     error = 0;
414     switch (sopt->sopt_name) {
415     case MRT_INIT:
416 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
417 	if (error)
418 	    break;
419 	error = ip_mrouter_init(so, optval);
420 	break;
421 
422     case MRT_DONE:
423 	error = ip_mrouter_done();
424 	break;
425 
426     case MRT_ADD_VIF:
427 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
428 	if (error)
429 	    break;
430 	error = add_vif(&vifc);
431 	break;
432 
433     case MRT_DEL_VIF:
434 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
435 	if (error)
436 	    break;
437 	error = del_vif(vifi);
438 	break;
439 
440     case MRT_ADD_MFC:
441     case MRT_DEL_MFC:
442 	/*
443 	 * select data size depending on API version.
444 	 */
445 	if (sopt->sopt_name == MRT_ADD_MFC &&
446 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
447 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
448 				sizeof(struct mfcctl2));
449 	} else {
450 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
451 				sizeof(struct mfcctl));
452 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
453 			sizeof(mfc) - sizeof(struct mfcctl));
454 	}
455 	if (error)
456 	    break;
457 	if (sopt->sopt_name == MRT_ADD_MFC)
458 	    error = add_mfc(&mfc);
459 	else
460 	    error = del_mfc(&mfc);
461 	break;
462 
463     case MRT_ASSERT:
464 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
465 	if (error)
466 	    break;
467 	set_assert(optval);
468 	break;
469 
470     case MRT_API_CONFIG:
471 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
472 	if (!error)
473 	    error = set_api_config(&i);
474 	if (!error)
475 	    error = sooptcopyout(sopt, &i, sizeof i);
476 	break;
477 
478     case MRT_ADD_BW_UPCALL:
479     case MRT_DEL_BW_UPCALL:
480 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
481 				sizeof bw_upcall);
482 	if (error)
483 	    break;
484 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
485 	    error = add_bw_upcall(&bw_upcall);
486 	else
487 	    error = del_bw_upcall(&bw_upcall);
488 	break;
489 
490     default:
491 	error = EOPNOTSUPP;
492 	break;
493     }
494     return error;
495 }
496 
497 /*
498  * Handle MRT getsockopt commands
499  */
500 static int
501 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
502 {
503     int error;
504 
505     switch (sopt->sopt_name) {
506     case MRT_VERSION:
507 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
508 	break;
509 
510     case MRT_ASSERT:
511 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
512 	    sizeof V_pim_assert_enabled);
513 	break;
514 
515     case MRT_API_SUPPORT:
516 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
517 	break;
518 
519     case MRT_API_CONFIG:
520 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
521 	break;
522 
523     default:
524 	error = EOPNOTSUPP;
525 	break;
526     }
527     return error;
528 }
529 
530 /*
531  * Handle ioctl commands to obtain information from the cache
532  */
533 static int
534 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
535 {
536     int error = 0;
537 
538     /*
539      * Currently the only function calling this ioctl routine is rtioctl().
540      * Typically, only root can create the raw socket in order to execute
541      * this ioctl method, however the request might be coming from a prison
542      */
543     error = priv_check(curthread, PRIV_NETINET_MROUTE);
544     if (error)
545 	return (error);
546     switch (cmd) {
547     case (SIOCGETVIFCNT):
548 	error = get_vif_cnt((struct sioc_vif_req *)data);
549 	break;
550 
551     case (SIOCGETSGCNT):
552 	error = get_sg_cnt((struct sioc_sg_req *)data);
553 	break;
554 
555     default:
556 	error = EINVAL;
557 	break;
558     }
559     return error;
560 }
561 
562 /*
563  * returns the packet, byte, rpf-failure count for the source group provided
564  */
565 static int
566 get_sg_cnt(struct sioc_sg_req *req)
567 {
568     struct mfc *rt;
569 
570     MFC_LOCK();
571     rt = mfc_find(&req->src, &req->grp);
572     if (rt == NULL) {
573 	MFC_UNLOCK();
574 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
575 	return EADDRNOTAVAIL;
576     }
577     req->pktcnt = rt->mfc_pkt_cnt;
578     req->bytecnt = rt->mfc_byte_cnt;
579     req->wrong_if = rt->mfc_wrong_if;
580     MFC_UNLOCK();
581     return 0;
582 }
583 
584 /*
585  * returns the input and output packet and byte counts on the vif provided
586  */
587 static int
588 get_vif_cnt(struct sioc_vif_req *req)
589 {
590     vifi_t vifi = req->vifi;
591 
592     VIF_LOCK();
593     if (vifi >= V_numvifs) {
594 	VIF_UNLOCK();
595 	return EINVAL;
596     }
597 
598     req->icount = V_viftable[vifi].v_pkt_in;
599     req->ocount = V_viftable[vifi].v_pkt_out;
600     req->ibytes = V_viftable[vifi].v_bytes_in;
601     req->obytes = V_viftable[vifi].v_bytes_out;
602     VIF_UNLOCK();
603 
604     return 0;
605 }
606 
607 static void
608 if_detached_event(void *arg __unused, struct ifnet *ifp)
609 {
610     vifi_t vifi;
611     int i;
612 
613     MROUTER_LOCK();
614 
615     if (V_ip_mrouter == NULL) {
616 	MROUTER_UNLOCK();
617 	return;
618     }
619 
620     VIF_LOCK();
621     MFC_LOCK();
622 
623     /*
624      * Tear down multicast forwarder state associated with this ifnet.
625      * 1. Walk the vif list, matching vifs against this ifnet.
626      * 2. Walk the multicast forwarding cache (mfc) looking for
627      *    inner matches with this vif's index.
628      * 3. Expire any matching multicast forwarding cache entries.
629      * 4. Free vif state. This should disable ALLMULTI on the interface.
630      */
631     for (vifi = 0; vifi < V_numvifs; vifi++) {
632 	if (V_viftable[vifi].v_ifp != ifp)
633 		continue;
634 	for (i = 0; i < mfchashsize; i++) {
635 		struct mfc *rt, *nrt;
636 		for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
637 			nrt = LIST_NEXT(rt, mfc_hash);
638 			if (rt->mfc_parent == vifi) {
639 				expire_mfc(rt);
640 			}
641 		}
642 	}
643 	del_vif_locked(vifi);
644     }
645 
646     MFC_UNLOCK();
647     VIF_UNLOCK();
648 
649     MROUTER_UNLOCK();
650 }
651 
652 /*
653  * Enable multicast forwarding.
654  */
655 static int
656 ip_mrouter_init(struct socket *so, int version)
657 {
658 
659     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
660         so->so_type, so->so_proto->pr_protocol);
661 
662     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
663 	return EOPNOTSUPP;
664 
665     if (version != 1)
666 	return ENOPROTOOPT;
667 
668     MROUTER_LOCK();
669 
670     if (ip_mrouter_unloading) {
671 	MROUTER_UNLOCK();
672 	return ENOPROTOOPT;
673     }
674 
675     if (V_ip_mrouter != NULL) {
676 	MROUTER_UNLOCK();
677 	return EADDRINUSE;
678     }
679 
680     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
681 	HASH_NOWAIT);
682 
683     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
684 	curvnet);
685     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
686 	curvnet);
687     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
688 	curvnet);
689 
690     V_ip_mrouter = so;
691     ip_mrouter_cnt++;
692 
693     MROUTER_UNLOCK();
694 
695     CTR1(KTR_IPMF, "%s: done", __func__);
696 
697     return 0;
698 }
699 
700 /*
701  * Disable multicast forwarding.
702  */
703 static int
704 X_ip_mrouter_done(void)
705 {
706     vifi_t vifi;
707     int i;
708     struct ifnet *ifp;
709     struct ifreq ifr;
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 	    struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
735 
736 	    so->sin_len = sizeof(struct sockaddr_in);
737 	    so->sin_family = AF_INET;
738 	    so->sin_addr.s_addr = INADDR_ANY;
739 	    ifp = V_viftable[vifi].v_ifp;
740 	    if_allmulti(ifp, 0);
741 	}
742     }
743     bzero((caddr_t)V_viftable, sizeof(V_viftable));
744     V_numvifs = 0;
745     V_pim_assert_enabled = 0;
746 
747     VIF_UNLOCK();
748 
749     callout_stop(&V_expire_upcalls_ch);
750     callout_stop(&V_bw_upcalls_ch);
751     callout_stop(&V_bw_meter_ch);
752 
753     MFC_LOCK();
754 
755     /*
756      * Free all multicast forwarding cache entries.
757      * Do not use hashdestroy(), as we must perform other cleanup.
758      */
759     for (i = 0; i < mfchashsize; i++) {
760 	struct mfc *rt, *nrt;
761 	for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
762 		nrt = LIST_NEXT(rt, mfc_hash);
763 		expire_mfc(rt);
764 	}
765     }
766     free(V_mfchashtbl, M_MRTABLE);
767     V_mfchashtbl = NULL;
768 
769     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
770 
771     V_bw_upcalls_n = 0;
772     bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
773 
774     MFC_UNLOCK();
775 
776     V_reg_vif_num = VIFI_INVALID;
777 
778     MROUTER_UNLOCK();
779 
780     CTR1(KTR_IPMF, "%s: done", __func__);
781 
782     return 0;
783 }
784 
785 /*
786  * Set PIM assert processing global
787  */
788 static int
789 set_assert(int i)
790 {
791     if ((i != 1) && (i != 0))
792 	return EINVAL;
793 
794     V_pim_assert_enabled = i;
795 
796     return 0;
797 }
798 
799 /*
800  * Configure API capabilities
801  */
802 int
803 set_api_config(uint32_t *apival)
804 {
805     int i;
806 
807     /*
808      * We can set the API capabilities only if it is the first operation
809      * after MRT_INIT. I.e.:
810      *  - there are no vifs installed
811      *  - pim_assert is not enabled
812      *  - the MFC table is empty
813      */
814     if (V_numvifs > 0) {
815 	*apival = 0;
816 	return EPERM;
817     }
818     if (V_pim_assert_enabled) {
819 	*apival = 0;
820 	return EPERM;
821     }
822 
823     MFC_LOCK();
824 
825     for (i = 0; i < mfchashsize; i++) {
826 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
827 	    *apival = 0;
828 	    return EPERM;
829 	}
830     }
831 
832     MFC_UNLOCK();
833 
834     V_mrt_api_config = *apival & mrt_api_support;
835     *apival = V_mrt_api_config;
836 
837     return 0;
838 }
839 
840 /*
841  * Add a vif to the vif table
842  */
843 static int
844 add_vif(struct vifctl *vifcp)
845 {
846     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
847     struct sockaddr_in sin = {sizeof sin, AF_INET};
848     struct ifaddr *ifa;
849     struct ifnet *ifp;
850     int error;
851 
852     VIF_LOCK();
853     if (vifcp->vifc_vifi >= MAXVIFS) {
854 	VIF_UNLOCK();
855 	return EINVAL;
856     }
857     /* rate limiting is no longer supported by this code */
858     if (vifcp->vifc_rate_limit != 0) {
859 	log(LOG_ERR, "rate limiting is no longer supported\n");
860 	VIF_UNLOCK();
861 	return EINVAL;
862     }
863     if (!in_nullhost(vifp->v_lcl_addr)) {
864 	VIF_UNLOCK();
865 	return EADDRINUSE;
866     }
867     if (in_nullhost(vifcp->vifc_lcl_addr)) {
868 	VIF_UNLOCK();
869 	return EADDRNOTAVAIL;
870     }
871 
872     /* Find the interface with an address in AF_INET family */
873     if (vifcp->vifc_flags & VIFF_REGISTER) {
874 	/*
875 	 * XXX: Because VIFF_REGISTER does not really need a valid
876 	 * local interface (e.g. it could be 127.0.0.2), we don't
877 	 * check its address.
878 	 */
879 	ifp = NULL;
880     } else {
881 	sin.sin_addr = vifcp->vifc_lcl_addr;
882 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
883 	if (ifa == NULL) {
884 	    VIF_UNLOCK();
885 	    return EADDRNOTAVAIL;
886 	}
887 	ifp = ifa->ifa_ifp;
888 	ifa_free(ifa);
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 %s thresh %x", __func__,
935 	(int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_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 %s group %lx parent %x",
1067 	    __func__, inet_ntoa(mfccp->mfcc_origin),
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 %s group %lx parent %x qh %p",
1087 		    __func__, inet_ntoa(mfccp->mfcc_origin),
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 %s group %lx", __func__,
1166 	inet_ntoa(origin), (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 %s group %lx ifp %p",
1231 	inet_ntoa(ip->ip_src), (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 (%s,%lx)",
1294 	    inet_ntoa(ip->ip_src), (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_DONTWAIT);
1310 	if (mb0 && (M_HASCL(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_copy(mb0, 0, hlen);
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     int 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 	for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
1453 		nrt = LIST_NEXT(rt, mfc_hash);
1454 
1455 		if (TAILQ_EMPTY(&rt->mfc_stall))
1456 			continue;
1457 
1458 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1459 			continue;
1460 
1461 		/*
1462 		 * free the bw_meter entries
1463 		 */
1464 		while (rt->mfc_bw_meter != NULL) {
1465 		    struct bw_meter *x = rt->mfc_bw_meter;
1466 
1467 		    rt->mfc_bw_meter = x->bm_mfc_next;
1468 		    free(x, M_BWMETER);
1469 		}
1470 
1471 		MRTSTAT_INC(mrts_cache_cleanups);
1472 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1473 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1474 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1475 
1476 		expire_mfc(rt);
1477 	    }
1478     }
1479 
1480     MFC_UNLOCK();
1481 
1482     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1483 	curvnet);
1484 
1485     CURVNET_RESTORE();
1486 }
1487 
1488 /*
1489  * Packet forwarding routine once entry in the cache is made
1490  */
1491 static int
1492 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1493 {
1494     struct ip  *ip = mtod(m, struct ip *);
1495     vifi_t vifi;
1496     int plen = ip->ip_len;
1497 
1498     VIF_LOCK_ASSERT();
1499 
1500     /*
1501      * If xmt_vif is not -1, send on only the requested vif.
1502      *
1503      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1504      */
1505     if (xmt_vif < V_numvifs) {
1506 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1507 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1508 	else
1509 		phyint_send(ip, V_viftable + xmt_vif, m);
1510 	return 1;
1511     }
1512 
1513     /*
1514      * Don't forward if it didn't arrive from the parent vif for its origin.
1515      */
1516     vifi = rt->mfc_parent;
1517     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1518 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1519 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1520 	MRTSTAT_INC(mrts_wrong_if);
1521 	++rt->mfc_wrong_if;
1522 	/*
1523 	 * If we are doing PIM assert processing, send a message
1524 	 * to the routing daemon.
1525 	 *
1526 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1527 	 * can complete the SPT switch, regardless of the type
1528 	 * of the iif (broadcast media, GRE tunnel, etc).
1529 	 */
1530 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1531 	    V_viftable[vifi].v_ifp) {
1532 
1533 	    if (ifp == &V_multicast_register_if)
1534 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1535 
1536 	    /* Get vifi for the incoming packet */
1537 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1538 		vifi++)
1539 		;
1540 	    if (vifi >= V_numvifs)
1541 		return 0;	/* The iif is not found: ignore the packet. */
1542 
1543 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1544 		return 0;	/* WRONGVIF disabled: ignore the packet */
1545 
1546 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1547 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1548 		struct igmpmsg *im;
1549 		int hlen = ip->ip_hl << 2;
1550 		struct mbuf *mm = m_copy(m, 0, hlen);
1551 
1552 		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1553 		    mm = m_pullup(mm, hlen);
1554 		if (mm == NULL)
1555 		    return ENOBUFS;
1556 
1557 		im = mtod(mm, struct igmpmsg *);
1558 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1559 		im->im_mbz		= 0;
1560 		im->im_vif		= vifi;
1561 
1562 		MRTSTAT_INC(mrts_upcalls);
1563 
1564 		k_igmpsrc.sin_addr = im->im_src;
1565 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1566 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1567 		    MRTSTAT_INC(mrts_upq_sockfull);
1568 		    return ENOBUFS;
1569 		}
1570 	    }
1571 	}
1572 	return 0;
1573     }
1574 
1575 
1576     /* If I sourced this packet, it counts as output, else it was input. */
1577     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1578 	V_viftable[vifi].v_pkt_out++;
1579 	V_viftable[vifi].v_bytes_out += plen;
1580     } else {
1581 	V_viftable[vifi].v_pkt_in++;
1582 	V_viftable[vifi].v_bytes_in += plen;
1583     }
1584     rt->mfc_pkt_cnt++;
1585     rt->mfc_byte_cnt += plen;
1586 
1587     /*
1588      * For each vif, decide if a copy of the packet should be forwarded.
1589      * Forward if:
1590      *		- the ttl exceeds the vif's threshold
1591      *		- there are group members downstream on interface
1592      */
1593     for (vifi = 0; vifi < V_numvifs; vifi++)
1594 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1595 	    V_viftable[vifi].v_pkt_out++;
1596 	    V_viftable[vifi].v_bytes_out += plen;
1597 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1598 		pim_register_send(ip, V_viftable + vifi, m, rt);
1599 	    else
1600 		phyint_send(ip, V_viftable + vifi, m);
1601 	}
1602 
1603     /*
1604      * Perform upcall-related bw measuring.
1605      */
1606     if (rt->mfc_bw_meter != NULL) {
1607 	struct bw_meter *x;
1608 	struct timeval now;
1609 
1610 	microtime(&now);
1611 	MFC_LOCK_ASSERT();
1612 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1613 	    bw_meter_receive_packet(x, plen, &now);
1614     }
1615 
1616     return 0;
1617 }
1618 
1619 /*
1620  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1621  */
1622 static int
1623 X_legal_vif_num(int vif)
1624 {
1625 	int ret;
1626 
1627 	ret = 0;
1628 	if (vif < 0)
1629 		return (ret);
1630 
1631 	VIF_LOCK();
1632 	if (vif < V_numvifs)
1633 		ret = 1;
1634 	VIF_UNLOCK();
1635 
1636 	return (ret);
1637 }
1638 
1639 /*
1640  * Return the local address used by this vif
1641  */
1642 static u_long
1643 X_ip_mcast_src(int vifi)
1644 {
1645 	in_addr_t addr;
1646 
1647 	addr = INADDR_ANY;
1648 	if (vifi < 0)
1649 		return (addr);
1650 
1651 	VIF_LOCK();
1652 	if (vifi < V_numvifs)
1653 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1654 	VIF_UNLOCK();
1655 
1656 	return (addr);
1657 }
1658 
1659 static void
1660 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1661 {
1662     struct mbuf *mb_copy;
1663     int hlen = ip->ip_hl << 2;
1664 
1665     VIF_LOCK_ASSERT();
1666 
1667     /*
1668      * Make a new reference to the packet; make sure that
1669      * the IP header is actually copied, not just referenced,
1670      * so that ip_output() only scribbles on the copy.
1671      */
1672     mb_copy = m_copypacket(m, M_DONTWAIT);
1673     if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1674 	mb_copy = m_pullup(mb_copy, hlen);
1675     if (mb_copy == NULL)
1676 	return;
1677 
1678     send_packet(vifp, mb_copy);
1679 }
1680 
1681 static void
1682 send_packet(struct vif *vifp, struct mbuf *m)
1683 {
1684 	struct ip_moptions imo;
1685 	struct in_multi *imm[2];
1686 	int error;
1687 
1688 	VIF_LOCK_ASSERT();
1689 
1690 	imo.imo_multicast_ifp  = vifp->v_ifp;
1691 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1692 	imo.imo_multicast_loop = 1;
1693 	imo.imo_multicast_vif  = -1;
1694 	imo.imo_num_memberships = 0;
1695 	imo.imo_max_memberships = 2;
1696 	imo.imo_membership  = &imm[0];
1697 
1698 	/*
1699 	 * Re-entrancy should not be a problem here, because
1700 	 * the packets that we send out and are looped back at us
1701 	 * should get rejected because they appear to come from
1702 	 * the loopback interface, thus preventing looping.
1703 	 */
1704 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1705 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1706 	    (ptrdiff_t)(vifp - V_viftable), error);
1707 }
1708 
1709 /*
1710  * Stubs for old RSVP socket shim implementation.
1711  */
1712 
1713 static int
1714 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1715 {
1716 
1717 	return (EOPNOTSUPP);
1718 }
1719 
1720 static void
1721 X_ip_rsvp_force_done(struct socket *so __unused)
1722 {
1723 
1724 }
1725 
1726 static void
1727 X_rsvp_input(struct mbuf *m, int off __unused)
1728 {
1729 
1730 	if (!V_rsvp_on)
1731 		m_freem(m);
1732 }
1733 
1734 /*
1735  * Code for bandwidth monitors
1736  */
1737 
1738 /*
1739  * Define common interface for timeval-related methods
1740  */
1741 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1742 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1743 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1744 
1745 static uint32_t
1746 compute_bw_meter_flags(struct bw_upcall *req)
1747 {
1748     uint32_t flags = 0;
1749 
1750     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1751 	flags |= BW_METER_UNIT_PACKETS;
1752     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1753 	flags |= BW_METER_UNIT_BYTES;
1754     if (req->bu_flags & BW_UPCALL_GEQ)
1755 	flags |= BW_METER_GEQ;
1756     if (req->bu_flags & BW_UPCALL_LEQ)
1757 	flags |= BW_METER_LEQ;
1758 
1759     return flags;
1760 }
1761 
1762 /*
1763  * Add a bw_meter entry
1764  */
1765 static int
1766 add_bw_upcall(struct bw_upcall *req)
1767 {
1768     struct mfc *mfc;
1769     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1770 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1771     struct timeval now;
1772     struct bw_meter *x;
1773     uint32_t flags;
1774 
1775     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1776 	return EOPNOTSUPP;
1777 
1778     /* Test if the flags are valid */
1779     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1780 	return EINVAL;
1781     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1782 	return EINVAL;
1783     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1784 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1785 	return EINVAL;
1786 
1787     /* Test if the threshold time interval is valid */
1788     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1789 	return EINVAL;
1790 
1791     flags = compute_bw_meter_flags(req);
1792 
1793     /*
1794      * Find if we have already same bw_meter entry
1795      */
1796     MFC_LOCK();
1797     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1798     if (mfc == NULL) {
1799 	MFC_UNLOCK();
1800 	return EADDRNOTAVAIL;
1801     }
1802     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1803 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1804 			   &req->bu_threshold.b_time, ==)) &&
1805 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1806 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1807 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1808 	    MFC_UNLOCK();
1809 	    return 0;		/* XXX Already installed */
1810 	}
1811     }
1812 
1813     /* Allocate the new bw_meter entry */
1814     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1815     if (x == NULL) {
1816 	MFC_UNLOCK();
1817 	return ENOBUFS;
1818     }
1819 
1820     /* Set the new bw_meter entry */
1821     x->bm_threshold.b_time = req->bu_threshold.b_time;
1822     microtime(&now);
1823     x->bm_start_time = now;
1824     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1825     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1826     x->bm_measured.b_packets = 0;
1827     x->bm_measured.b_bytes = 0;
1828     x->bm_flags = flags;
1829     x->bm_time_next = NULL;
1830     x->bm_time_hash = BW_METER_BUCKETS;
1831 
1832     /* Add the new bw_meter entry to the front of entries for this MFC */
1833     x->bm_mfc = mfc;
1834     x->bm_mfc_next = mfc->mfc_bw_meter;
1835     mfc->mfc_bw_meter = x;
1836     schedule_bw_meter(x, &now);
1837     MFC_UNLOCK();
1838 
1839     return 0;
1840 }
1841 
1842 static void
1843 free_bw_list(struct bw_meter *list)
1844 {
1845     while (list != NULL) {
1846 	struct bw_meter *x = list;
1847 
1848 	list = list->bm_mfc_next;
1849 	unschedule_bw_meter(x);
1850 	free(x, M_BWMETER);
1851     }
1852 }
1853 
1854 /*
1855  * Delete one or multiple bw_meter entries
1856  */
1857 static int
1858 del_bw_upcall(struct bw_upcall *req)
1859 {
1860     struct mfc *mfc;
1861     struct bw_meter *x;
1862 
1863     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1864 	return EOPNOTSUPP;
1865 
1866     MFC_LOCK();
1867 
1868     /* Find the corresponding MFC entry */
1869     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1870     if (mfc == NULL) {
1871 	MFC_UNLOCK();
1872 	return EADDRNOTAVAIL;
1873     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1874 	/*
1875 	 * Delete all bw_meter entries for this mfc
1876 	 */
1877 	struct bw_meter *list;
1878 
1879 	list = mfc->mfc_bw_meter;
1880 	mfc->mfc_bw_meter = NULL;
1881 	free_bw_list(list);
1882 	MFC_UNLOCK();
1883 	return 0;
1884     } else {			/* Delete a single bw_meter entry */
1885 	struct bw_meter *prev;
1886 	uint32_t flags = 0;
1887 
1888 	flags = compute_bw_meter_flags(req);
1889 
1890 	/* Find the bw_meter entry to delete */
1891 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1892 	     prev = x, x = x->bm_mfc_next) {
1893 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1894 			       &req->bu_threshold.b_time, ==)) &&
1895 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1896 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1897 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1898 		break;
1899 	}
1900 	if (x != NULL) { /* Delete entry from the list for this MFC */
1901 	    if (prev != NULL)
1902 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1903 	    else
1904 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1905 
1906 	    unschedule_bw_meter(x);
1907 	    MFC_UNLOCK();
1908 	    /* Free the bw_meter entry */
1909 	    free(x, M_BWMETER);
1910 	    return 0;
1911 	} else {
1912 	    MFC_UNLOCK();
1913 	    return EINVAL;
1914 	}
1915     }
1916     /* NOTREACHED */
1917 }
1918 
1919 /*
1920  * Perform bandwidth measurement processing that may result in an upcall
1921  */
1922 static void
1923 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1924 {
1925     struct timeval delta;
1926 
1927     MFC_LOCK_ASSERT();
1928 
1929     delta = *nowp;
1930     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1931 
1932     if (x->bm_flags & BW_METER_GEQ) {
1933 	/*
1934 	 * Processing for ">=" type of bw_meter entry
1935 	 */
1936 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1937 	    /* Reset the bw_meter entry */
1938 	    x->bm_start_time = *nowp;
1939 	    x->bm_measured.b_packets = 0;
1940 	    x->bm_measured.b_bytes = 0;
1941 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1942 	}
1943 
1944 	/* Record that a packet is received */
1945 	x->bm_measured.b_packets++;
1946 	x->bm_measured.b_bytes += plen;
1947 
1948 	/*
1949 	 * Test if we should deliver an upcall
1950 	 */
1951 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1952 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1953 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1954 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1955 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1956 		/* Prepare an upcall for delivery */
1957 		bw_meter_prepare_upcall(x, nowp);
1958 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1959 	    }
1960 	}
1961     } else if (x->bm_flags & BW_METER_LEQ) {
1962 	/*
1963 	 * Processing for "<=" type of bw_meter entry
1964 	 */
1965 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1966 	    /*
1967 	     * We are behind time with the multicast forwarding table
1968 	     * scanning for "<=" type of bw_meter entries, so test now
1969 	     * if we should deliver an upcall.
1970 	     */
1971 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1972 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1973 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1974 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1975 		/* Prepare an upcall for delivery */
1976 		bw_meter_prepare_upcall(x, nowp);
1977 	    }
1978 	    /* Reschedule the bw_meter entry */
1979 	    unschedule_bw_meter(x);
1980 	    schedule_bw_meter(x, nowp);
1981 	}
1982 
1983 	/* Record that a packet is received */
1984 	x->bm_measured.b_packets++;
1985 	x->bm_measured.b_bytes += plen;
1986 
1987 	/*
1988 	 * Test if we should restart the measuring interval
1989 	 */
1990 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1991 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1992 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1993 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1994 	    /* Don't restart the measuring interval */
1995 	} else {
1996 	    /* Do restart the measuring interval */
1997 	    /*
1998 	     * XXX: note that we don't unschedule and schedule, because this
1999 	     * might be too much overhead per packet. Instead, when we process
2000 	     * all entries for a given timer hash bin, we check whether it is
2001 	     * really a timeout. If not, we reschedule at that time.
2002 	     */
2003 	    x->bm_start_time = *nowp;
2004 	    x->bm_measured.b_packets = 0;
2005 	    x->bm_measured.b_bytes = 0;
2006 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2007 	}
2008     }
2009 }
2010 
2011 /*
2012  * Prepare a bandwidth-related upcall
2013  */
2014 static void
2015 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2016 {
2017     struct timeval delta;
2018     struct bw_upcall *u;
2019 
2020     MFC_LOCK_ASSERT();
2021 
2022     /*
2023      * Compute the measured time interval
2024      */
2025     delta = *nowp;
2026     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2027 
2028     /*
2029      * If there are too many pending upcalls, deliver them now
2030      */
2031     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2032 	bw_upcalls_send();
2033 
2034     /*
2035      * Set the bw_upcall entry
2036      */
2037     u = &V_bw_upcalls[V_bw_upcalls_n++];
2038     u->bu_src = x->bm_mfc->mfc_origin;
2039     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2040     u->bu_threshold.b_time = x->bm_threshold.b_time;
2041     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2042     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2043     u->bu_measured.b_time = delta;
2044     u->bu_measured.b_packets = x->bm_measured.b_packets;
2045     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2046     u->bu_flags = 0;
2047     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2048 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2049     if (x->bm_flags & BW_METER_UNIT_BYTES)
2050 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2051     if (x->bm_flags & BW_METER_GEQ)
2052 	u->bu_flags |= BW_UPCALL_GEQ;
2053     if (x->bm_flags & BW_METER_LEQ)
2054 	u->bu_flags |= BW_UPCALL_LEQ;
2055 }
2056 
2057 /*
2058  * Send the pending bandwidth-related upcalls
2059  */
2060 static void
2061 bw_upcalls_send(void)
2062 {
2063     struct mbuf *m;
2064     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2065     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2066     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2067 				      0,		/* unused2 */
2068 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2069 				      0,		/* im_mbz  */
2070 				      0,		/* im_vif  */
2071 				      0,		/* unused3 */
2072 				      { 0 },		/* im_src  */
2073 				      { 0 } };		/* im_dst  */
2074 
2075     MFC_LOCK_ASSERT();
2076 
2077     if (V_bw_upcalls_n == 0)
2078 	return;			/* No pending upcalls */
2079 
2080     V_bw_upcalls_n = 0;
2081 
2082     /*
2083      * Allocate a new mbuf, initialize it with the header and
2084      * the payload for the pending calls.
2085      */
2086     MGETHDR(m, M_DONTWAIT, MT_DATA);
2087     if (m == NULL) {
2088 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2089 	return;
2090     }
2091 
2092     m->m_len = m->m_pkthdr.len = 0;
2093     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2094     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2095 
2096     /*
2097      * Send the upcalls
2098      * XXX do we need to set the address in k_igmpsrc ?
2099      */
2100     MRTSTAT_INC(mrts_upcalls);
2101     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2102 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2103 	MRTSTAT_INC(mrts_upq_sockfull);
2104     }
2105 }
2106 
2107 /*
2108  * Compute the timeout hash value for the bw_meter entries
2109  */
2110 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2111     do {								\
2112 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2113 									\
2114 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2115 	(hash) = next_timeval.tv_sec;					\
2116 	if (next_timeval.tv_usec)					\
2117 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2118 	(hash) %= BW_METER_BUCKETS;					\
2119     } while (0)
2120 
2121 /*
2122  * Schedule a timer to process periodically bw_meter entry of type "<="
2123  * by linking the entry in the proper hash bucket.
2124  */
2125 static void
2126 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2127 {
2128     int time_hash;
2129 
2130     MFC_LOCK_ASSERT();
2131 
2132     if (!(x->bm_flags & BW_METER_LEQ))
2133 	return;		/* XXX: we schedule timers only for "<=" entries */
2134 
2135     /*
2136      * Reset the bw_meter entry
2137      */
2138     x->bm_start_time = *nowp;
2139     x->bm_measured.b_packets = 0;
2140     x->bm_measured.b_bytes = 0;
2141     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2142 
2143     /*
2144      * Compute the timeout hash value and insert the entry
2145      */
2146     BW_METER_TIMEHASH(x, time_hash);
2147     x->bm_time_next = V_bw_meter_timers[time_hash];
2148     V_bw_meter_timers[time_hash] = x;
2149     x->bm_time_hash = time_hash;
2150 }
2151 
2152 /*
2153  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2154  * by removing the entry from the proper hash bucket.
2155  */
2156 static void
2157 unschedule_bw_meter(struct bw_meter *x)
2158 {
2159     int time_hash;
2160     struct bw_meter *prev, *tmp;
2161 
2162     MFC_LOCK_ASSERT();
2163 
2164     if (!(x->bm_flags & BW_METER_LEQ))
2165 	return;		/* XXX: we schedule timers only for "<=" entries */
2166 
2167     /*
2168      * Compute the timeout hash value and delete the entry
2169      */
2170     time_hash = x->bm_time_hash;
2171     if (time_hash >= BW_METER_BUCKETS)
2172 	return;		/* Entry was not scheduled */
2173 
2174     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2175 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2176 	if (tmp == x)
2177 	    break;
2178 
2179     if (tmp == NULL)
2180 	panic("unschedule_bw_meter: bw_meter entry not found");
2181 
2182     if (prev != NULL)
2183 	prev->bm_time_next = x->bm_time_next;
2184     else
2185 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2186 
2187     x->bm_time_next = NULL;
2188     x->bm_time_hash = BW_METER_BUCKETS;
2189 }
2190 
2191 
2192 /*
2193  * Process all "<=" type of bw_meter that should be processed now,
2194  * and for each entry prepare an upcall if necessary. Each processed
2195  * entry is rescheduled again for the (periodic) processing.
2196  *
2197  * This is run periodically (once per second normally). On each round,
2198  * all the potentially matching entries are in the hash slot that we are
2199  * looking at.
2200  */
2201 static void
2202 bw_meter_process()
2203 {
2204     uint32_t loops;
2205     int i;
2206     struct timeval now, process_endtime;
2207 
2208     microtime(&now);
2209     if (V_last_tv_sec == now.tv_sec)
2210 	return;		/* nothing to do */
2211 
2212     loops = now.tv_sec - V_last_tv_sec;
2213     V_last_tv_sec = now.tv_sec;
2214     if (loops > BW_METER_BUCKETS)
2215 	loops = BW_METER_BUCKETS;
2216 
2217     MFC_LOCK();
2218     /*
2219      * Process all bins of bw_meter entries from the one after the last
2220      * processed to the current one. On entry, i points to the last bucket
2221      * visited, so we need to increment i at the beginning of the loop.
2222      */
2223     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2224 	struct bw_meter *x, *tmp_list;
2225 
2226 	if (++i >= BW_METER_BUCKETS)
2227 	    i = 0;
2228 
2229 	/* Disconnect the list of bw_meter entries from the bin */
2230 	tmp_list = V_bw_meter_timers[i];
2231 	V_bw_meter_timers[i] = NULL;
2232 
2233 	/* Process the list of bw_meter entries */
2234 	while (tmp_list != NULL) {
2235 	    x = tmp_list;
2236 	    tmp_list = tmp_list->bm_time_next;
2237 
2238 	    /* Test if the time interval is over */
2239 	    process_endtime = x->bm_start_time;
2240 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2241 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2242 		/* Not yet: reschedule, but don't reset */
2243 		int time_hash;
2244 
2245 		BW_METER_TIMEHASH(x, time_hash);
2246 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2247 		    /*
2248 		     * XXX: somehow the bin processing is a bit ahead of time.
2249 		     * Put the entry in the next bin.
2250 		     */
2251 		    if (++time_hash >= BW_METER_BUCKETS)
2252 			time_hash = 0;
2253 		}
2254 		x->bm_time_next = V_bw_meter_timers[time_hash];
2255 		V_bw_meter_timers[time_hash] = x;
2256 		x->bm_time_hash = time_hash;
2257 
2258 		continue;
2259 	    }
2260 
2261 	    /*
2262 	     * Test if we should deliver an upcall
2263 	     */
2264 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2265 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2266 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2267 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2268 		/* Prepare an upcall for delivery */
2269 		bw_meter_prepare_upcall(x, &now);
2270 	    }
2271 
2272 	    /*
2273 	     * Reschedule for next processing
2274 	     */
2275 	    schedule_bw_meter(x, &now);
2276 	}
2277     }
2278 
2279     /* Send all upcalls that are pending delivery */
2280     bw_upcalls_send();
2281 
2282     MFC_UNLOCK();
2283 }
2284 
2285 /*
2286  * A periodic function for sending all upcalls that are pending delivery
2287  */
2288 static void
2289 expire_bw_upcalls_send(void *arg)
2290 {
2291     CURVNET_SET((struct vnet *) arg);
2292 
2293     MFC_LOCK();
2294     bw_upcalls_send();
2295     MFC_UNLOCK();
2296 
2297     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2298 	curvnet);
2299     CURVNET_RESTORE();
2300 }
2301 
2302 /*
2303  * A periodic function for periodic scanning of the multicast forwarding
2304  * table for processing all "<=" bw_meter entries.
2305  */
2306 static void
2307 expire_bw_meter_process(void *arg)
2308 {
2309     CURVNET_SET((struct vnet *) arg);
2310 
2311     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2312 	bw_meter_process();
2313 
2314     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2315 	curvnet);
2316     CURVNET_RESTORE();
2317 }
2318 
2319 /*
2320  * End of bandwidth monitoring code
2321  */
2322 
2323 /*
2324  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2325  *
2326  */
2327 static int
2328 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2329     struct mfc *rt)
2330 {
2331     struct mbuf *mb_copy, *mm;
2332 
2333     /*
2334      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2335      * rendezvous point was unspecified, and we were told not to.
2336      */
2337     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2338 	in_nullhost(rt->mfc_rp))
2339 	return 0;
2340 
2341     mb_copy = pim_register_prepare(ip, m);
2342     if (mb_copy == NULL)
2343 	return ENOBUFS;
2344 
2345     /*
2346      * Send all the fragments. Note that the mbuf for each fragment
2347      * is freed by the sending machinery.
2348      */
2349     for (mm = mb_copy; mm; mm = mb_copy) {
2350 	mb_copy = mm->m_nextpkt;
2351 	mm->m_nextpkt = 0;
2352 	mm = m_pullup(mm, sizeof(struct ip));
2353 	if (mm != NULL) {
2354 	    ip = mtod(mm, struct ip *);
2355 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2356 		pim_register_send_rp(ip, vifp, mm, rt);
2357 	    } else {
2358 		pim_register_send_upcall(ip, vifp, mm, rt);
2359 	    }
2360 	}
2361     }
2362 
2363     return 0;
2364 }
2365 
2366 /*
2367  * Return a copy of the data packet that is ready for PIM Register
2368  * encapsulation.
2369  * XXX: Note that in the returned copy the IP header is a valid one.
2370  */
2371 static struct mbuf *
2372 pim_register_prepare(struct ip *ip, struct mbuf *m)
2373 {
2374     struct mbuf *mb_copy = NULL;
2375     int mtu;
2376 
2377     /* Take care of delayed checksums */
2378     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2379 	/* XXX: in_delayed_cksum() expects net byte order */
2380 	ip->ip_len = htons(ip->ip_len);
2381 	in_delayed_cksum(m);
2382 	ip->ip_len = ntohs(ip->ip_len);
2383 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2384     }
2385 
2386     /*
2387      * Copy the old packet & pullup its IP header into the
2388      * new mbuf so we can modify it.
2389      */
2390     mb_copy = m_copypacket(m, M_DONTWAIT);
2391     if (mb_copy == NULL)
2392 	return NULL;
2393     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2394     if (mb_copy == NULL)
2395 	return NULL;
2396 
2397     /* take care of the TTL */
2398     ip = mtod(mb_copy, struct ip *);
2399     --ip->ip_ttl;
2400 
2401     /* Compute the MTU after the PIM Register encapsulation */
2402     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2403 
2404     if (ip->ip_len <= mtu) {
2405 	/* Turn the IP header into a valid one */
2406 	ip->ip_len = htons(ip->ip_len);
2407 	ip->ip_off = htons(ip->ip_off);
2408 	ip->ip_sum = 0;
2409 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2410     } else {
2411 	/* Fragment the packet */
2412 	if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
2413 	    m_freem(mb_copy);
2414 	    return NULL;
2415 	}
2416     }
2417     return mb_copy;
2418 }
2419 
2420 /*
2421  * Send an upcall with the data packet to the user-level process.
2422  */
2423 static int
2424 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2425     struct mbuf *mb_copy, struct mfc *rt)
2426 {
2427     struct mbuf *mb_first;
2428     int len = ntohs(ip->ip_len);
2429     struct igmpmsg *im;
2430     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2431 
2432     VIF_LOCK_ASSERT();
2433 
2434     /*
2435      * Add a new mbuf with an upcall header
2436      */
2437     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2438     if (mb_first == NULL) {
2439 	m_freem(mb_copy);
2440 	return ENOBUFS;
2441     }
2442     mb_first->m_data += max_linkhdr;
2443     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2444     mb_first->m_len = sizeof(struct igmpmsg);
2445     mb_first->m_next = mb_copy;
2446 
2447     /* Send message to routing daemon */
2448     im = mtod(mb_first, struct igmpmsg *);
2449     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2450     im->im_mbz		= 0;
2451     im->im_vif		= vifp - V_viftable;
2452     im->im_src		= ip->ip_src;
2453     im->im_dst		= ip->ip_dst;
2454 
2455     k_igmpsrc.sin_addr	= ip->ip_src;
2456 
2457     MRTSTAT_INC(mrts_upcalls);
2458 
2459     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2460 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2461 	MRTSTAT_INC(mrts_upq_sockfull);
2462 	return ENOBUFS;
2463     }
2464 
2465     /* Keep statistics */
2466     PIMSTAT_INC(pims_snd_registers_msgs);
2467     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2468 
2469     return 0;
2470 }
2471 
2472 /*
2473  * Encapsulate the data packet in PIM Register message and send it to the RP.
2474  */
2475 static int
2476 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2477     struct mfc *rt)
2478 {
2479     struct mbuf *mb_first;
2480     struct ip *ip_outer;
2481     struct pim_encap_pimhdr *pimhdr;
2482     int len = ntohs(ip->ip_len);
2483     vifi_t vifi = rt->mfc_parent;
2484 
2485     VIF_LOCK_ASSERT();
2486 
2487     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2488 	m_freem(mb_copy);
2489 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2490     }
2491 
2492     /*
2493      * Add a new mbuf with the encapsulating header
2494      */
2495     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2496     if (mb_first == NULL) {
2497 	m_freem(mb_copy);
2498 	return ENOBUFS;
2499     }
2500     mb_first->m_data += max_linkhdr;
2501     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2502     mb_first->m_next = mb_copy;
2503 
2504     mb_first->m_pkthdr.len = len + mb_first->m_len;
2505 
2506     /*
2507      * Fill in the encapsulating IP and PIM header
2508      */
2509     ip_outer = mtod(mb_first, struct ip *);
2510     *ip_outer = pim_encap_iphdr;
2511     ip_outer->ip_id = ip_newid();
2512     ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2513     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2514     ip_outer->ip_dst = rt->mfc_rp;
2515     /*
2516      * Copy the inner header TOS to the outer header, and take care of the
2517      * IP_DF bit.
2518      */
2519     ip_outer->ip_tos = ip->ip_tos;
2520     if (ntohs(ip->ip_off) & IP_DF)
2521 	ip_outer->ip_off |= IP_DF;
2522     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2523 					 + sizeof(pim_encap_iphdr));
2524     *pimhdr = pim_encap_pimhdr;
2525     /* If the iif crosses a border, set the Border-bit */
2526     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2527 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2528 
2529     mb_first->m_data += sizeof(pim_encap_iphdr);
2530     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2531     mb_first->m_data -= sizeof(pim_encap_iphdr);
2532 
2533     send_packet(vifp, mb_first);
2534 
2535     /* Keep statistics */
2536     PIMSTAT_INC(pims_snd_registers_msgs);
2537     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2538 
2539     return 0;
2540 }
2541 
2542 /*
2543  * pim_encapcheck() is called by the encap4_input() path at runtime to
2544  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2545  * into the kernel.
2546  */
2547 static int
2548 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2549 {
2550 
2551 #ifdef DIAGNOSTIC
2552     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2553 #endif
2554     if (proto != IPPROTO_PIM)
2555 	return 0;	/* not for us; reject the datagram. */
2556 
2557     return 64;		/* claim the datagram. */
2558 }
2559 
2560 /*
2561  * PIM-SMv2 and PIM-DM messages processing.
2562  * Receives and verifies the PIM control messages, and passes them
2563  * up to the listening socket, using rip_input().
2564  * The only message with special processing is the PIM_REGISTER message
2565  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2566  * is passed to if_simloop().
2567  */
2568 void
2569 pim_input(struct mbuf *m, int off)
2570 {
2571     struct ip *ip = mtod(m, struct ip *);
2572     struct pim *pim;
2573     int minlen;
2574     int datalen = ip->ip_len;
2575     int ip_tos;
2576     int iphlen = off;
2577 
2578     /* Keep statistics */
2579     PIMSTAT_INC(pims_rcv_total_msgs);
2580     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2581 
2582     /*
2583      * Validate lengths
2584      */
2585     if (datalen < PIM_MINLEN) {
2586 	PIMSTAT_INC(pims_rcv_tooshort);
2587 	CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2588 	    __func__, datalen, inet_ntoa(ip->ip_src));
2589 	m_freem(m);
2590 	return;
2591     }
2592 
2593     /*
2594      * If the packet is at least as big as a REGISTER, go agead
2595      * and grab the PIM REGISTER header size, to avoid another
2596      * possible m_pullup() later.
2597      *
2598      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2599      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2600      */
2601     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2602     /*
2603      * Get the IP and PIM headers in contiguous memory, and
2604      * possibly the PIM REGISTER header.
2605      */
2606     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
2607 	(m = m_pullup(m, minlen)) == 0) {
2608 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2609 	return;
2610     }
2611 
2612     /* m_pullup() may have given us a new mbuf so reset ip. */
2613     ip = mtod(m, struct ip *);
2614     ip_tos = ip->ip_tos;
2615 
2616     /* adjust mbuf to point to the PIM header */
2617     m->m_data += iphlen;
2618     m->m_len  -= iphlen;
2619     pim = mtod(m, struct pim *);
2620 
2621     /*
2622      * Validate checksum. If PIM REGISTER, exclude the data packet.
2623      *
2624      * XXX: some older PIMv2 implementations don't make this distinction,
2625      * so for compatibility reason perform the checksum over part of the
2626      * message, and if error, then over the whole message.
2627      */
2628     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2629 	/* do nothing, checksum okay */
2630     } else if (in_cksum(m, datalen)) {
2631 	PIMSTAT_INC(pims_rcv_badsum);
2632 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2633 	m_freem(m);
2634 	return;
2635     }
2636 
2637     /* PIM version check */
2638     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2639 	PIMSTAT_INC(pims_rcv_badversion);
2640 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2641 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2642 	m_freem(m);
2643 	return;
2644     }
2645 
2646     /* restore mbuf back to the outer IP */
2647     m->m_data -= iphlen;
2648     m->m_len  += iphlen;
2649 
2650     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2651 	/*
2652 	 * Since this is a REGISTER, we'll make a copy of the register
2653 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2654 	 * routing daemon.
2655 	 */
2656 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2657 	struct mbuf *mcp;
2658 	struct ip *encap_ip;
2659 	u_int32_t *reghdr;
2660 	struct ifnet *vifp;
2661 
2662 	VIF_LOCK();
2663 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2664 	    VIF_UNLOCK();
2665 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2666 		(int)V_reg_vif_num);
2667 	    m_freem(m);
2668 	    return;
2669 	}
2670 	/* XXX need refcnt? */
2671 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2672 	VIF_UNLOCK();
2673 
2674 	/*
2675 	 * Validate length
2676 	 */
2677 	if (datalen < PIM_REG_MINLEN) {
2678 	    PIMSTAT_INC(pims_rcv_tooshort);
2679 	    PIMSTAT_INC(pims_rcv_badregisters);
2680 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2681 	    m_freem(m);
2682 	    return;
2683 	}
2684 
2685 	reghdr = (u_int32_t *)(pim + 1);
2686 	encap_ip = (struct ip *)(reghdr + 1);
2687 
2688 	CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2689 	    __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2690 
2691 	/* verify the version number of the inner packet */
2692 	if (encap_ip->ip_v != IPVERSION) {
2693 	    PIMSTAT_INC(pims_rcv_badregisters);
2694 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2695 	    m_freem(m);
2696 	    return;
2697 	}
2698 
2699 	/* verify the inner packet is destined to a mcast group */
2700 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2701 	    PIMSTAT_INC(pims_rcv_badregisters);
2702 	    CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2703 		inet_ntoa(encap_ip->ip_dst));
2704 	    m_freem(m);
2705 	    return;
2706 	}
2707 
2708 	/* If a NULL_REGISTER, pass it to the daemon */
2709 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2710 	    goto pim_input_to_daemon;
2711 
2712 	/*
2713 	 * Copy the TOS from the outer IP header to the inner IP header.
2714 	 */
2715 	if (encap_ip->ip_tos != ip_tos) {
2716 	    /* Outer TOS -> inner TOS */
2717 	    encap_ip->ip_tos = ip_tos;
2718 	    /* Recompute the inner header checksum. Sigh... */
2719 
2720 	    /* adjust mbuf to point to the inner IP header */
2721 	    m->m_data += (iphlen + PIM_MINLEN);
2722 	    m->m_len  -= (iphlen + PIM_MINLEN);
2723 
2724 	    encap_ip->ip_sum = 0;
2725 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2726 
2727 	    /* restore mbuf to point back to the outer IP header */
2728 	    m->m_data -= (iphlen + PIM_MINLEN);
2729 	    m->m_len  += (iphlen + PIM_MINLEN);
2730 	}
2731 
2732 	/*
2733 	 * Decapsulate the inner IP packet and loopback to forward it
2734 	 * as a normal multicast packet. Also, make a copy of the
2735 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2736 	 * to pass to the daemon later, so it can take the appropriate
2737 	 * actions (e.g., send back PIM_REGISTER_STOP).
2738 	 * XXX: here m->m_data points to the outer IP header.
2739 	 */
2740 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2741 	if (mcp == NULL) {
2742 	    CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2743 	    m_freem(m);
2744 	    return;
2745 	}
2746 
2747 	/* Keep statistics */
2748 	/* XXX: registers_bytes include only the encap. mcast pkt */
2749 	PIMSTAT_INC(pims_rcv_registers_msgs);
2750 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2751 
2752 	/*
2753 	 * forward the inner ip packet; point m_data at the inner ip.
2754 	 */
2755 	m_adj(m, iphlen + PIM_MINLEN);
2756 
2757 	CTR4(KTR_IPMF,
2758 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2759 	    __func__,
2760 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2761 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2762 	    (int)V_reg_vif_num);
2763 
2764 	/* NB: vifp was collected above; can it change on us? */
2765 	if_simloop(vifp, m, dst.sin_family, 0);
2766 
2767 	/* prepare the register head to send to the mrouting daemon */
2768 	m = mcp;
2769     }
2770 
2771 pim_input_to_daemon:
2772     /*
2773      * Pass the PIM message up to the daemon; if it is a Register message,
2774      * pass the 'head' only up to the daemon. This includes the
2775      * outer IP header, PIM header, PIM-Register header and the
2776      * inner IP header.
2777      * XXX: the outer IP header pkt size of a Register is not adjust to
2778      * reflect the fact that the inner multicast data is truncated.
2779      */
2780     rip_input(m, iphlen);
2781 
2782     return;
2783 }
2784 
2785 static int
2786 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2787 {
2788 	struct mfc	*rt;
2789 	int		 error, i;
2790 
2791 	if (req->newptr)
2792 		return (EPERM);
2793 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2794 		return (0);
2795 	error = sysctl_wire_old_buffer(req, 0);
2796 	if (error)
2797 		return (error);
2798 
2799 	MFC_LOCK();
2800 	for (i = 0; i < mfchashsize; i++) {
2801 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2802 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2803 			if (error)
2804 				goto out_locked;
2805 		}
2806 	}
2807 out_locked:
2808 	MFC_UNLOCK();
2809 	return (error);
2810 }
2811 
2812 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2813     sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2814     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2815 
2816 static void
2817 vnet_mroute_init(const void *unused __unused)
2818 {
2819 
2820 	MALLOC(V_nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2821 	bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2822 	callout_init(&V_expire_upcalls_ch, CALLOUT_MPSAFE);
2823 	callout_init(&V_bw_upcalls_ch, CALLOUT_MPSAFE);
2824 	callout_init(&V_bw_meter_ch, CALLOUT_MPSAFE);
2825 }
2826 
2827 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mroute_init,
2828 	NULL);
2829 
2830 static void
2831 vnet_mroute_uninit(const void *unused __unused)
2832 {
2833 
2834 	FREE(V_nexpire, M_MRTABLE);
2835 	V_nexpire = NULL;
2836 }
2837 
2838 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE,
2839 	vnet_mroute_uninit, NULL);
2840 
2841 static int
2842 ip_mroute_modevent(module_t mod, int type, void *unused)
2843 {
2844 
2845     switch (type) {
2846     case MOD_LOAD:
2847 	MROUTER_LOCK_INIT();
2848 
2849 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2850 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2851 	if (if_detach_event_tag == NULL) {
2852 		printf("ip_mroute: unable to ifnet_deperture_even handler\n");
2853 		MROUTER_LOCK_DESTROY();
2854 		return (EINVAL);
2855 	}
2856 
2857 	MFC_LOCK_INIT();
2858 	VIF_LOCK_INIT();
2859 
2860 	mfchashsize = MFCHASHSIZE;
2861 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2862 	    !powerof2(mfchashsize)) {
2863 		printf("WARNING: %s not a power of 2; using default\n",
2864 		    "net.inet.ip.mfchashsize");
2865 		mfchashsize = MFCHASHSIZE;
2866 	}
2867 
2868 	pim_squelch_wholepkt = 0;
2869 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2870 	    &pim_squelch_wholepkt);
2871 
2872 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2873 	    pim_encapcheck, &in_pim_protosw, NULL);
2874 	if (pim_encap_cookie == NULL) {
2875 		printf("ip_mroute: unable to attach pim encap\n");
2876 		VIF_LOCK_DESTROY();
2877 		MFC_LOCK_DESTROY();
2878 		MROUTER_LOCK_DESTROY();
2879 		return (EINVAL);
2880 	}
2881 
2882 	ip_mcast_src = X_ip_mcast_src;
2883 	ip_mforward = X_ip_mforward;
2884 	ip_mrouter_done = X_ip_mrouter_done;
2885 	ip_mrouter_get = X_ip_mrouter_get;
2886 	ip_mrouter_set = X_ip_mrouter_set;
2887 
2888 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2889 	ip_rsvp_vif = X_ip_rsvp_vif;
2890 
2891 	legal_vif_num = X_legal_vif_num;
2892 	mrt_ioctl = X_mrt_ioctl;
2893 	rsvp_input_p = X_rsvp_input;
2894 	break;
2895 
2896     case MOD_UNLOAD:
2897 	/*
2898 	 * Typically module unload happens after the user-level
2899 	 * process has shutdown the kernel services (the check
2900 	 * below insures someone can't just yank the module out
2901 	 * from under a running process).  But if the module is
2902 	 * just loaded and then unloaded w/o starting up a user
2903 	 * process we still need to cleanup.
2904 	 */
2905 	MROUTER_LOCK();
2906 	if (ip_mrouter_cnt != 0) {
2907 	    MROUTER_UNLOCK();
2908 	    return (EINVAL);
2909 	}
2910 	ip_mrouter_unloading = 1;
2911 	MROUTER_UNLOCK();
2912 
2913 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2914 
2915 	if (pim_encap_cookie) {
2916 	    encap_detach(pim_encap_cookie);
2917 	    pim_encap_cookie = NULL;
2918 	}
2919 
2920 	ip_mcast_src = NULL;
2921 	ip_mforward = NULL;
2922 	ip_mrouter_done = NULL;
2923 	ip_mrouter_get = NULL;
2924 	ip_mrouter_set = NULL;
2925 
2926 	ip_rsvp_force_done = NULL;
2927 	ip_rsvp_vif = NULL;
2928 
2929 	legal_vif_num = NULL;
2930 	mrt_ioctl = NULL;
2931 	rsvp_input_p = NULL;
2932 
2933 	VIF_LOCK_DESTROY();
2934 	MFC_LOCK_DESTROY();
2935 	MROUTER_LOCK_DESTROY();
2936 	break;
2937 
2938     default:
2939 	return EOPNOTSUPP;
2940     }
2941     return 0;
2942 }
2943 
2944 static moduledata_t ip_mroutemod = {
2945     "ip_mroute",
2946     ip_mroute_modevent,
2947     0
2948 };
2949 
2950 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_MIDDLE);
2951