xref: /freebsd/sys/netinet/ip_mroute.c (revision 1f4bcc459a76b7aa664f3fd557684cd0ba6da352)
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/eventhandler.h>
81 #include <sys/lock.h>
82 #include <sys/ktr.h>
83 #include <sys/malloc.h>
84 #include <sys/mbuf.h>
85 #include <sys/module.h>
86 #include <sys/priv.h>
87 #include <sys/protosw.h>
88 #include <sys/signalvar.h>
89 #include <sys/socket.h>
90 #include <sys/socketvar.h>
91 #include <sys/sockio.h>
92 #include <sys/sx.h>
93 #include <sys/sysctl.h>
94 #include <sys/syslog.h>
95 #include <sys/systm.h>
96 #include <sys/time.h>
97 #include <sys/counter.h>
98 
99 #include <net/if.h>
100 #include <net/if_var.h>
101 #include <net/netisr.h>
102 #include <net/route.h>
103 #include <net/vnet.h>
104 
105 #include <netinet/in.h>
106 #include <netinet/igmp.h>
107 #include <netinet/in_systm.h>
108 #include <netinet/in_var.h>
109 #include <netinet/ip.h>
110 #include <netinet/ip_encap.h>
111 #include <netinet/ip_mroute.h>
112 #include <netinet/ip_var.h>
113 #include <netinet/ip_options.h>
114 #include <netinet/pim.h>
115 #include <netinet/pim_var.h>
116 #include <netinet/udp.h>
117 
118 #include <machine/in_cksum.h>
119 
120 #ifndef KTR_IPMF
121 #define KTR_IPMF KTR_INET
122 #endif
123 
124 #define		VIFI_INVALID	((vifi_t) -1)
125 
126 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
127 #define	V_last_tv_sec	VNET(last_tv_sec)
128 
129 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
130 
131 /*
132  * Locking.  We use two locks: one for the virtual interface table and
133  * one for the forwarding table.  These locks may be nested in which case
134  * the VIF lock must always be taken first.  Note that each lock is used
135  * to cover not only the specific data structure but also related data
136  * structures.
137  */
138 
139 static struct mtx mrouter_mtx;
140 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
141 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
142 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
143 #define	MROUTER_LOCK_INIT()						\
144 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
145 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
146 
147 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
148 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
149 
150 static VNET_PCPUSTAT_DEFINE(struct mrtstat, mrtstat);
151 VNET_PCPUSTAT_SYSINIT(mrtstat);
152 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
153 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
154     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
155     "netinet/ip_mroute.h)");
156 
157 static VNET_DEFINE(u_long, mfchash);
158 #define	V_mfchash		VNET(mfchash)
159 #define	MFCHASH(a, g)							\
160 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
161 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
162 #define	MFCHASHSIZE	256
163 
164 static u_long mfchashsize;			/* Hash size */
165 static VNET_DEFINE(u_char *, nexpire);		/* 0..mfchashsize-1 */
166 #define	V_nexpire		VNET(nexpire)
167 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
168 #define	V_mfchashtbl		VNET(mfchashtbl)
169 
170 static struct mtx mfc_mtx;
171 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
172 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
173 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
174 #define	MFC_LOCK_INIT()							\
175 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
176 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
177 
178 static VNET_DEFINE(vifi_t, numvifs);
179 #define	V_numvifs		VNET(numvifs)
180 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
181 #define	V_viftable		VNET(viftable)
182 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_VNET | CTLFLAG_RD,
183     &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
184     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
185 
186 static struct mtx vif_mtx;
187 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
188 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
189 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
190 #define	VIF_LOCK_INIT()							\
191 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
192 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
193 
194 static eventhandler_tag if_detach_event_tag = NULL;
195 
196 static VNET_DEFINE(struct callout, expire_upcalls_ch);
197 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
198 
199 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
200 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
201 
202 /*
203  * Bandwidth meter variables and constants
204  */
205 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
206 /*
207  * Pending timeouts are stored in a hash table, the key being the
208  * expiration time. Periodically, the entries are analysed and processed.
209  */
210 #define	BW_METER_BUCKETS	1024
211 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
212 #define	V_bw_meter_timers	VNET(bw_meter_timers)
213 static VNET_DEFINE(struct callout, bw_meter_ch);
214 #define	V_bw_meter_ch		VNET(bw_meter_ch)
215 #define	BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
216 
217 /*
218  * Pending upcalls are stored in a vector which is flushed when
219  * full, or periodically
220  */
221 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
222 #define	V_bw_upcalls		VNET(bw_upcalls)
223 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
224 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
225 static VNET_DEFINE(struct callout, bw_upcalls_ch);
226 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
227 
228 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
229 
230 static VNET_PCPUSTAT_DEFINE(struct pimstat, pimstat);
231 VNET_PCPUSTAT_SYSINIT(pimstat);
232 VNET_PCPUSTAT_SYSUNINIT(pimstat);
233 
234 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
235 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
236     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
237 
238 static u_long	pim_squelch_wholepkt = 0;
239 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
240     &pim_squelch_wholepkt, 0,
241     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
242 
243 extern  struct domain inetdomain;
244 static const struct protosw in_pim_protosw = {
245 	.pr_type =		SOCK_RAW,
246 	.pr_domain =		&inetdomain,
247 	.pr_protocol =		IPPROTO_PIM,
248 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
249 	.pr_input =		pim_input,
250 	.pr_output =		rip_output,
251 	.pr_ctloutput =		rip_ctloutput,
252 	.pr_usrreqs =		&rip_usrreqs
253 };
254 static const struct encaptab *pim_encap_cookie;
255 
256 static int pim_encapcheck(const struct mbuf *, int, int, void *);
257 
258 /*
259  * Note: the PIM Register encapsulation adds the following in front of a
260  * data packet:
261  *
262  * struct pim_encap_hdr {
263  *    struct ip ip;
264  *    struct pim_encap_pimhdr  pim;
265  * }
266  *
267  */
268 
269 struct pim_encap_pimhdr {
270 	struct pim pim;
271 	uint32_t   flags;
272 };
273 #define		PIM_ENCAP_TTL	64
274 
275 static struct ip pim_encap_iphdr = {
276 #if BYTE_ORDER == LITTLE_ENDIAN
277 	sizeof(struct ip) >> 2,
278 	IPVERSION,
279 #else
280 	IPVERSION,
281 	sizeof(struct ip) >> 2,
282 #endif
283 	0,			/* tos */
284 	sizeof(struct ip),	/* total length */
285 	0,			/* id */
286 	0,			/* frag offset */
287 	PIM_ENCAP_TTL,
288 	IPPROTO_PIM,
289 	0,			/* checksum */
290 };
291 
292 static struct pim_encap_pimhdr pim_encap_pimhdr = {
293     {
294 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
295 	0,			/* reserved */
296 	0,			/* checksum */
297     },
298     0				/* flags */
299 };
300 
301 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
302 #define	V_reg_vif_num		VNET(reg_vif_num)
303 static VNET_DEFINE(struct ifnet, multicast_register_if);
304 #define	V_multicast_register_if	VNET(multicast_register_if)
305 
306 /*
307  * Private variables.
308  */
309 
310 static u_long	X_ip_mcast_src(int);
311 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
312 		    struct ip_moptions *);
313 static int	X_ip_mrouter_done(void);
314 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
315 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
316 static int	X_legal_vif_num(int);
317 static int	X_mrt_ioctl(u_long, caddr_t, int);
318 
319 static int	add_bw_upcall(struct bw_upcall *);
320 static int	add_mfc(struct mfcctl2 *);
321 static int	add_vif(struct vifctl *);
322 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
323 static void	bw_meter_process(void);
324 static void	bw_meter_receive_packet(struct bw_meter *, int,
325 		    struct timeval *);
326 static void	bw_upcalls_send(void);
327 static int	del_bw_upcall(struct bw_upcall *);
328 static int	del_mfc(struct mfcctl2 *);
329 static int	del_vif(vifi_t);
330 static int	del_vif_locked(vifi_t);
331 static void	expire_bw_meter_process(void *);
332 static void	expire_bw_upcalls_send(void *);
333 static void	expire_mfc(struct mfc *);
334 static void	expire_upcalls(void *);
335 static void	free_bw_list(struct bw_meter *);
336 static int	get_sg_cnt(struct sioc_sg_req *);
337 static int	get_vif_cnt(struct sioc_vif_req *);
338 static void	if_detached_event(void *, struct ifnet *);
339 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
340 static int	ip_mrouter_init(struct socket *, int);
341 static __inline struct mfc *
342 		mfc_find(struct in_addr *, struct in_addr *);
343 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
344 static struct mbuf *
345 		pim_register_prepare(struct ip *, struct mbuf *);
346 static int	pim_register_send(struct ip *, struct vif *,
347 		    struct mbuf *, struct mfc *);
348 static int	pim_register_send_rp(struct ip *, struct vif *,
349 		    struct mbuf *, struct mfc *);
350 static int	pim_register_send_upcall(struct ip *, struct vif *,
351 		    struct mbuf *, struct mfc *);
352 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
353 static void	send_packet(struct vif *, struct mbuf *);
354 static int	set_api_config(uint32_t *);
355 static int	set_assert(int);
356 static int	socket_send(struct socket *, struct mbuf *,
357 		    struct sockaddr_in *);
358 static void	unschedule_bw_meter(struct bw_meter *);
359 
360 /*
361  * Kernel multicast forwarding API capabilities and setup.
362  * If more API capabilities are added to the kernel, they should be
363  * recorded in `mrt_api_support'.
364  */
365 #define MRT_API_VERSION		0x0305
366 
367 static const int mrt_api_version = MRT_API_VERSION;
368 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
369 					 MRT_MFC_FLAGS_BORDER_VIF |
370 					 MRT_MFC_RP |
371 					 MRT_MFC_BW_UPCALL);
372 static VNET_DEFINE(uint32_t, mrt_api_config);
373 #define	V_mrt_api_config	VNET(mrt_api_config)
374 static VNET_DEFINE(int, pim_assert_enabled);
375 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
376 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
377 
378 /*
379  * Find a route for a given origin IP address and multicast group address.
380  * Statistics must be updated by the caller.
381  */
382 static __inline struct mfc *
383 mfc_find(struct in_addr *o, struct in_addr *g)
384 {
385 	struct mfc *rt;
386 
387 	MFC_LOCK_ASSERT();
388 
389 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
390 		if (in_hosteq(rt->mfc_origin, *o) &&
391 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
392 		    TAILQ_EMPTY(&rt->mfc_stall))
393 			break;
394 	}
395 
396 	return (rt);
397 }
398 
399 /*
400  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
401  */
402 static int
403 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
404 {
405     int	error, optval;
406     vifi_t	vifi;
407     struct	vifctl vifc;
408     struct	mfcctl2 mfc;
409     struct	bw_upcall bw_upcall;
410     uint32_t	i;
411 
412     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
413 	return EPERM;
414 
415     error = 0;
416     switch (sopt->sopt_name) {
417     case MRT_INIT:
418 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
419 	if (error)
420 	    break;
421 	error = ip_mrouter_init(so, optval);
422 	break;
423 
424     case MRT_DONE:
425 	error = ip_mrouter_done();
426 	break;
427 
428     case MRT_ADD_VIF:
429 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
430 	if (error)
431 	    break;
432 	error = add_vif(&vifc);
433 	break;
434 
435     case MRT_DEL_VIF:
436 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
437 	if (error)
438 	    break;
439 	error = del_vif(vifi);
440 	break;
441 
442     case MRT_ADD_MFC:
443     case MRT_DEL_MFC:
444 	/*
445 	 * select data size depending on API version.
446 	 */
447 	if (sopt->sopt_name == MRT_ADD_MFC &&
448 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
449 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
450 				sizeof(struct mfcctl2));
451 	} else {
452 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
453 				sizeof(struct mfcctl));
454 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
455 			sizeof(mfc) - sizeof(struct mfcctl));
456 	}
457 	if (error)
458 	    break;
459 	if (sopt->sopt_name == MRT_ADD_MFC)
460 	    error = add_mfc(&mfc);
461 	else
462 	    error = del_mfc(&mfc);
463 	break;
464 
465     case MRT_ASSERT:
466 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
467 	if (error)
468 	    break;
469 	set_assert(optval);
470 	break;
471 
472     case MRT_API_CONFIG:
473 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
474 	if (!error)
475 	    error = set_api_config(&i);
476 	if (!error)
477 	    error = sooptcopyout(sopt, &i, sizeof i);
478 	break;
479 
480     case MRT_ADD_BW_UPCALL:
481     case MRT_DEL_BW_UPCALL:
482 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
483 				sizeof bw_upcall);
484 	if (error)
485 	    break;
486 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
487 	    error = add_bw_upcall(&bw_upcall);
488 	else
489 	    error = del_bw_upcall(&bw_upcall);
490 	break;
491 
492     default:
493 	error = EOPNOTSUPP;
494 	break;
495     }
496     return error;
497 }
498 
499 /*
500  * Handle MRT getsockopt commands
501  */
502 static int
503 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
504 {
505     int error;
506 
507     switch (sopt->sopt_name) {
508     case MRT_VERSION:
509 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
510 	break;
511 
512     case MRT_ASSERT:
513 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
514 	    sizeof V_pim_assert_enabled);
515 	break;
516 
517     case MRT_API_SUPPORT:
518 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
519 	break;
520 
521     case MRT_API_CONFIG:
522 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
523 	break;
524 
525     default:
526 	error = EOPNOTSUPP;
527 	break;
528     }
529     return error;
530 }
531 
532 /*
533  * Handle ioctl commands to obtain information from the cache
534  */
535 static int
536 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
537 {
538     int error = 0;
539 
540     /*
541      * Currently the only function calling this ioctl routine is rtioctl_fib().
542      * Typically, only root can create the raw socket in order to execute
543      * this ioctl method, however the request might be coming from a prison
544      */
545     error = priv_check(curthread, PRIV_NETINET_MROUTE);
546     if (error)
547 	return (error);
548     switch (cmd) {
549     case (SIOCGETVIFCNT):
550 	error = get_vif_cnt((struct sioc_vif_req *)data);
551 	break;
552 
553     case (SIOCGETSGCNT):
554 	error = get_sg_cnt((struct sioc_sg_req *)data);
555 	break;
556 
557     default:
558 	error = EINVAL;
559 	break;
560     }
561     return error;
562 }
563 
564 /*
565  * returns the packet, byte, rpf-failure count for the source group provided
566  */
567 static int
568 get_sg_cnt(struct sioc_sg_req *req)
569 {
570     struct mfc *rt;
571 
572     MFC_LOCK();
573     rt = mfc_find(&req->src, &req->grp);
574     if (rt == NULL) {
575 	MFC_UNLOCK();
576 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
577 	return EADDRNOTAVAIL;
578     }
579     req->pktcnt = rt->mfc_pkt_cnt;
580     req->bytecnt = rt->mfc_byte_cnt;
581     req->wrong_if = rt->mfc_wrong_if;
582     MFC_UNLOCK();
583     return 0;
584 }
585 
586 /*
587  * returns the input and output packet and byte counts on the vif provided
588  */
589 static int
590 get_vif_cnt(struct sioc_vif_req *req)
591 {
592     vifi_t vifi = req->vifi;
593 
594     VIF_LOCK();
595     if (vifi >= V_numvifs) {
596 	VIF_UNLOCK();
597 	return EINVAL;
598     }
599 
600     req->icount = V_viftable[vifi].v_pkt_in;
601     req->ocount = V_viftable[vifi].v_pkt_out;
602     req->ibytes = V_viftable[vifi].v_bytes_in;
603     req->obytes = V_viftable[vifi].v_bytes_out;
604     VIF_UNLOCK();
605 
606     return 0;
607 }
608 
609 static void
610 if_detached_event(void *arg __unused, struct ifnet *ifp)
611 {
612     vifi_t vifi;
613     u_long i;
614 
615     MROUTER_LOCK();
616 
617     if (V_ip_mrouter == NULL) {
618 	MROUTER_UNLOCK();
619 	return;
620     }
621 
622     VIF_LOCK();
623     MFC_LOCK();
624 
625     /*
626      * Tear down multicast forwarder state associated with this ifnet.
627      * 1. Walk the vif list, matching vifs against this ifnet.
628      * 2. Walk the multicast forwarding cache (mfc) looking for
629      *    inner matches with this vif's index.
630      * 3. Expire any matching multicast forwarding cache entries.
631      * 4. Free vif state. This should disable ALLMULTI on the interface.
632      */
633     for (vifi = 0; vifi < V_numvifs; vifi++) {
634 	if (V_viftable[vifi].v_ifp != ifp)
635 		continue;
636 	for (i = 0; i < mfchashsize; i++) {
637 		struct mfc *rt, *nrt;
638 
639 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
640 			if (rt->mfc_parent == vifi) {
641 				expire_mfc(rt);
642 			}
643 		}
644 	}
645 	del_vif_locked(vifi);
646     }
647 
648     MFC_UNLOCK();
649     VIF_UNLOCK();
650 
651     MROUTER_UNLOCK();
652 }
653 
654 /*
655  * Enable multicast forwarding.
656  */
657 static int
658 ip_mrouter_init(struct socket *so, int version)
659 {
660 
661     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
662         so->so_type, so->so_proto->pr_protocol);
663 
664     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
665 	return EOPNOTSUPP;
666 
667     if (version != 1)
668 	return ENOPROTOOPT;
669 
670     MROUTER_LOCK();
671 
672     if (ip_mrouter_unloading) {
673 	MROUTER_UNLOCK();
674 	return ENOPROTOOPT;
675     }
676 
677     if (V_ip_mrouter != NULL) {
678 	MROUTER_UNLOCK();
679 	return EADDRINUSE;
680     }
681 
682     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
683 	HASH_NOWAIT);
684 
685     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
686 	curvnet);
687     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
688 	curvnet);
689     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
690 	curvnet);
691 
692     V_ip_mrouter = so;
693     ip_mrouter_cnt++;
694 
695     MROUTER_UNLOCK();
696 
697     CTR1(KTR_IPMF, "%s: done", __func__);
698 
699     return 0;
700 }
701 
702 /*
703  * Disable multicast forwarding.
704  */
705 static int
706 X_ip_mrouter_done(void)
707 {
708     struct ifnet *ifp;
709     u_long i;
710     vifi_t vifi;
711 
712     MROUTER_LOCK();
713 
714     if (V_ip_mrouter == NULL) {
715 	MROUTER_UNLOCK();
716 	return EINVAL;
717     }
718 
719     /*
720      * Detach/disable hooks to the reset of the system.
721      */
722     V_ip_mrouter = NULL;
723     ip_mrouter_cnt--;
724     V_mrt_api_config = 0;
725 
726     VIF_LOCK();
727 
728     /*
729      * For each phyint in use, disable promiscuous reception of all IP
730      * multicasts.
731      */
732     for (vifi = 0; vifi < V_numvifs; vifi++) {
733 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
734 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
735 	    ifp = V_viftable[vifi].v_ifp;
736 	    if_allmulti(ifp, 0);
737 	}
738     }
739     bzero((caddr_t)V_viftable, sizeof(V_viftable));
740     V_numvifs = 0;
741     V_pim_assert_enabled = 0;
742 
743     VIF_UNLOCK();
744 
745     callout_stop(&V_expire_upcalls_ch);
746     callout_stop(&V_bw_upcalls_ch);
747     callout_stop(&V_bw_meter_ch);
748 
749     MFC_LOCK();
750 
751     /*
752      * Free all multicast forwarding cache entries.
753      * Do not use hashdestroy(), as we must perform other cleanup.
754      */
755     for (i = 0; i < mfchashsize; i++) {
756 	struct mfc *rt, *nrt;
757 
758 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
759 		expire_mfc(rt);
760 	}
761     }
762     free(V_mfchashtbl, M_MRTABLE);
763     V_mfchashtbl = NULL;
764 
765     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
766 
767     V_bw_upcalls_n = 0;
768     bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
769 
770     MFC_UNLOCK();
771 
772     V_reg_vif_num = VIFI_INVALID;
773 
774     MROUTER_UNLOCK();
775 
776     CTR1(KTR_IPMF, "%s: done", __func__);
777 
778     return 0;
779 }
780 
781 /*
782  * Set PIM assert processing global
783  */
784 static int
785 set_assert(int i)
786 {
787     if ((i != 1) && (i != 0))
788 	return EINVAL;
789 
790     V_pim_assert_enabled = i;
791 
792     return 0;
793 }
794 
795 /*
796  * Configure API capabilities
797  */
798 int
799 set_api_config(uint32_t *apival)
800 {
801     u_long i;
802 
803     /*
804      * We can set the API capabilities only if it is the first operation
805      * after MRT_INIT. I.e.:
806      *  - there are no vifs installed
807      *  - pim_assert is not enabled
808      *  - the MFC table is empty
809      */
810     if (V_numvifs > 0) {
811 	*apival = 0;
812 	return EPERM;
813     }
814     if (V_pim_assert_enabled) {
815 	*apival = 0;
816 	return EPERM;
817     }
818 
819     MFC_LOCK();
820 
821     for (i = 0; i < mfchashsize; i++) {
822 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
823 	    MFC_UNLOCK();
824 	    *apival = 0;
825 	    return EPERM;
826 	}
827     }
828 
829     MFC_UNLOCK();
830 
831     V_mrt_api_config = *apival & mrt_api_support;
832     *apival = V_mrt_api_config;
833 
834     return 0;
835 }
836 
837 /*
838  * Add a vif to the vif table
839  */
840 static int
841 add_vif(struct vifctl *vifcp)
842 {
843     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
844     struct sockaddr_in sin = {sizeof sin, AF_INET};
845     struct ifaddr *ifa;
846     struct ifnet *ifp;
847     int error;
848 
849     VIF_LOCK();
850     if (vifcp->vifc_vifi >= MAXVIFS) {
851 	VIF_UNLOCK();
852 	return EINVAL;
853     }
854     /* rate limiting is no longer supported by this code */
855     if (vifcp->vifc_rate_limit != 0) {
856 	log(LOG_ERR, "rate limiting is no longer supported\n");
857 	VIF_UNLOCK();
858 	return EINVAL;
859     }
860     if (!in_nullhost(vifp->v_lcl_addr)) {
861 	VIF_UNLOCK();
862 	return EADDRINUSE;
863     }
864     if (in_nullhost(vifcp->vifc_lcl_addr)) {
865 	VIF_UNLOCK();
866 	return EADDRNOTAVAIL;
867     }
868 
869     /* Find the interface with an address in AF_INET family */
870     if (vifcp->vifc_flags & VIFF_REGISTER) {
871 	/*
872 	 * XXX: Because VIFF_REGISTER does not really need a valid
873 	 * local interface (e.g. it could be 127.0.0.2), we don't
874 	 * check its address.
875 	 */
876 	ifp = NULL;
877     } else {
878 	sin.sin_addr = vifcp->vifc_lcl_addr;
879 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
880 	if (ifa == NULL) {
881 	    VIF_UNLOCK();
882 	    return EADDRNOTAVAIL;
883 	}
884 	ifp = ifa->ifa_ifp;
885 	ifa_free(ifa);
886     }
887 
888     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
889 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
890 	VIF_UNLOCK();
891 	return EOPNOTSUPP;
892     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
893 	ifp = &V_multicast_register_if;
894 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
895 	if (V_reg_vif_num == VIFI_INVALID) {
896 	    if_initname(&V_multicast_register_if, "register_vif", 0);
897 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
898 	    V_reg_vif_num = vifcp->vifc_vifi;
899 	}
900     } else {		/* Make sure the interface supports multicast */
901 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
902 	    VIF_UNLOCK();
903 	    return EOPNOTSUPP;
904 	}
905 
906 	/* Enable promiscuous reception of all IP multicasts from the if */
907 	error = if_allmulti(ifp, 1);
908 	if (error) {
909 	    VIF_UNLOCK();
910 	    return error;
911 	}
912     }
913 
914     vifp->v_flags     = vifcp->vifc_flags;
915     vifp->v_threshold = vifcp->vifc_threshold;
916     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
917     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
918     vifp->v_ifp       = ifp;
919     /* initialize per vif pkt counters */
920     vifp->v_pkt_in    = 0;
921     vifp->v_pkt_out   = 0;
922     vifp->v_bytes_in  = 0;
923     vifp->v_bytes_out = 0;
924 
925     /* Adjust numvifs up if the vifi is higher than numvifs */
926     if (V_numvifs <= vifcp->vifc_vifi)
927 	V_numvifs = vifcp->vifc_vifi + 1;
928 
929     VIF_UNLOCK();
930 
931     CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
932 	(int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr),
933 	(int)vifcp->vifc_threshold);
934 
935     return 0;
936 }
937 
938 /*
939  * Delete a vif from the vif table
940  */
941 static int
942 del_vif_locked(vifi_t vifi)
943 {
944     struct vif *vifp;
945 
946     VIF_LOCK_ASSERT();
947 
948     if (vifi >= V_numvifs) {
949 	return EINVAL;
950     }
951     vifp = &V_viftable[vifi];
952     if (in_nullhost(vifp->v_lcl_addr)) {
953 	return EADDRNOTAVAIL;
954     }
955 
956     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
957 	if_allmulti(vifp->v_ifp, 0);
958 
959     if (vifp->v_flags & VIFF_REGISTER)
960 	V_reg_vif_num = VIFI_INVALID;
961 
962     bzero((caddr_t)vifp, sizeof (*vifp));
963 
964     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
965 
966     /* Adjust numvifs down */
967     for (vifi = V_numvifs; vifi > 0; vifi--)
968 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
969 	    break;
970     V_numvifs = vifi;
971 
972     return 0;
973 }
974 
975 static int
976 del_vif(vifi_t vifi)
977 {
978     int cc;
979 
980     VIF_LOCK();
981     cc = del_vif_locked(vifi);
982     VIF_UNLOCK();
983 
984     return cc;
985 }
986 
987 /*
988  * update an mfc entry without resetting counters and S,G addresses.
989  */
990 static void
991 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
992 {
993     int i;
994 
995     rt->mfc_parent = mfccp->mfcc_parent;
996     for (i = 0; i < V_numvifs; i++) {
997 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
998 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
999 	    MRT_MFC_FLAGS_ALL;
1000     }
1001     /* set the RP address */
1002     if (V_mrt_api_config & MRT_MFC_RP)
1003 	rt->mfc_rp = mfccp->mfcc_rp;
1004     else
1005 	rt->mfc_rp.s_addr = INADDR_ANY;
1006 }
1007 
1008 /*
1009  * fully initialize an mfc entry from the parameter.
1010  */
1011 static void
1012 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1013 {
1014     rt->mfc_origin     = mfccp->mfcc_origin;
1015     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1016 
1017     update_mfc_params(rt, mfccp);
1018 
1019     /* initialize pkt counters per src-grp */
1020     rt->mfc_pkt_cnt    = 0;
1021     rt->mfc_byte_cnt   = 0;
1022     rt->mfc_wrong_if   = 0;
1023     timevalclear(&rt->mfc_last_assert);
1024 }
1025 
1026 static void
1027 expire_mfc(struct mfc *rt)
1028 {
1029 	struct rtdetq *rte, *nrte;
1030 
1031 	MFC_LOCK_ASSERT();
1032 
1033 	free_bw_list(rt->mfc_bw_meter);
1034 
1035 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1036 		m_freem(rte->m);
1037 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1038 		free(rte, M_MRTABLE);
1039 	}
1040 
1041 	LIST_REMOVE(rt, mfc_hash);
1042 	free(rt, M_MRTABLE);
1043 }
1044 
1045 /*
1046  * Add an mfc entry
1047  */
1048 static int
1049 add_mfc(struct mfcctl2 *mfccp)
1050 {
1051     struct mfc *rt;
1052     struct rtdetq *rte, *nrte;
1053     u_long hash = 0;
1054     u_short nstl;
1055 
1056     VIF_LOCK();
1057     MFC_LOCK();
1058 
1059     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1060 
1061     /* If an entry already exists, just update the fields */
1062     if (rt) {
1063 	CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x",
1064 	    __func__, inet_ntoa(mfccp->mfcc_origin),
1065 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1066 	    mfccp->mfcc_parent);
1067 	update_mfc_params(rt, mfccp);
1068 	MFC_UNLOCK();
1069 	VIF_UNLOCK();
1070 	return (0);
1071     }
1072 
1073     /*
1074      * Find the entry for which the upcall was made and update
1075      */
1076     nstl = 0;
1077     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1078     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1079 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1080 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1081 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1082 		CTR5(KTR_IPMF,
1083 		    "%s: add mfc orig %s group %lx parent %x qh %p",
1084 		    __func__, inet_ntoa(mfccp->mfcc_origin),
1085 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1086 		    mfccp->mfcc_parent,
1087 		    TAILQ_FIRST(&rt->mfc_stall));
1088 		if (nstl++)
1089 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1090 
1091 		init_mfc_params(rt, mfccp);
1092 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1093 		V_nexpire[hash]--;
1094 
1095 		/* Free queued packets, but attempt to forward them first. */
1096 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1097 			if (rte->ifp != NULL)
1098 				ip_mdq(rte->m, rte->ifp, rt, -1);
1099 			m_freem(rte->m);
1100 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1101 			rt->mfc_nstall--;
1102 			free(rte, M_MRTABLE);
1103 		}
1104 	}
1105     }
1106 
1107     /*
1108      * It is possible that an entry is being inserted without an upcall
1109      */
1110     if (nstl == 0) {
1111 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1112 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1113 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1114 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1115 			init_mfc_params(rt, mfccp);
1116 			if (rt->mfc_expire)
1117 			    V_nexpire[hash]--;
1118 			rt->mfc_expire = 0;
1119 			break; /* XXX */
1120 		}
1121 	}
1122 
1123 	if (rt == NULL) {		/* no upcall, so make a new entry */
1124 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1125 	    if (rt == NULL) {
1126 		MFC_UNLOCK();
1127 		VIF_UNLOCK();
1128 		return (ENOBUFS);
1129 	    }
1130 
1131 	    init_mfc_params(rt, mfccp);
1132 	    TAILQ_INIT(&rt->mfc_stall);
1133 	    rt->mfc_nstall = 0;
1134 
1135 	    rt->mfc_expire     = 0;
1136 	    rt->mfc_bw_meter = NULL;
1137 
1138 	    /* insert new entry at head of hash chain */
1139 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1140 	}
1141     }
1142 
1143     MFC_UNLOCK();
1144     VIF_UNLOCK();
1145 
1146     return (0);
1147 }
1148 
1149 /*
1150  * Delete an mfc entry
1151  */
1152 static int
1153 del_mfc(struct mfcctl2 *mfccp)
1154 {
1155     struct in_addr	origin;
1156     struct in_addr	mcastgrp;
1157     struct mfc		*rt;
1158 
1159     origin = mfccp->mfcc_origin;
1160     mcastgrp = mfccp->mfcc_mcastgrp;
1161 
1162     CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__,
1163 	inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr));
1164 
1165     MFC_LOCK();
1166 
1167     rt = mfc_find(&origin, &mcastgrp);
1168     if (rt == NULL) {
1169 	MFC_UNLOCK();
1170 	return EADDRNOTAVAIL;
1171     }
1172 
1173     /*
1174      * free the bw_meter entries
1175      */
1176     free_bw_list(rt->mfc_bw_meter);
1177     rt->mfc_bw_meter = NULL;
1178 
1179     LIST_REMOVE(rt, mfc_hash);
1180     free(rt, M_MRTABLE);
1181 
1182     MFC_UNLOCK();
1183 
1184     return (0);
1185 }
1186 
1187 /*
1188  * Send a message to the routing daemon on the multicast routing socket.
1189  */
1190 static int
1191 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1192 {
1193     if (s) {
1194 	SOCKBUF_LOCK(&s->so_rcv);
1195 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1196 	    NULL) != 0) {
1197 	    sorwakeup_locked(s);
1198 	    return 0;
1199 	}
1200 	SOCKBUF_UNLOCK(&s->so_rcv);
1201     }
1202     m_freem(mm);
1203     return -1;
1204 }
1205 
1206 /*
1207  * IP multicast forwarding function. This function assumes that the packet
1208  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1209  * pointed to by "ifp", and the packet is to be relayed to other networks
1210  * that have members of the packet's destination IP multicast group.
1211  *
1212  * The packet is returned unscathed to the caller, unless it is
1213  * erroneous, in which case a non-zero return value tells the caller to
1214  * discard it.
1215  */
1216 
1217 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1218 
1219 static int
1220 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1221     struct ip_moptions *imo)
1222 {
1223     struct mfc *rt;
1224     int error;
1225     vifi_t vifi;
1226 
1227     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p",
1228 	inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1229 
1230     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1231 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1232 	/*
1233 	 * Packet arrived via a physical interface or
1234 	 * an encapsulated tunnel or a register_vif.
1235 	 */
1236     } else {
1237 	/*
1238 	 * Packet arrived through a source-route tunnel.
1239 	 * Source-route tunnels are no longer supported.
1240 	 */
1241 	return (1);
1242     }
1243 
1244     VIF_LOCK();
1245     MFC_LOCK();
1246     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1247 	if (ip->ip_ttl < MAXTTL)
1248 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1249 	error = ip_mdq(m, ifp, NULL, vifi);
1250 	MFC_UNLOCK();
1251 	VIF_UNLOCK();
1252 	return error;
1253     }
1254 
1255     /*
1256      * Don't forward a packet with time-to-live of zero or one,
1257      * or a packet destined to a local-only group.
1258      */
1259     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1260 	MFC_UNLOCK();
1261 	VIF_UNLOCK();
1262 	return 0;
1263     }
1264 
1265     /*
1266      * Determine forwarding vifs from the forwarding cache table
1267      */
1268     MRTSTAT_INC(mrts_mfc_lookups);
1269     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1270 
1271     /* Entry exists, so forward if necessary */
1272     if (rt != NULL) {
1273 	error = ip_mdq(m, ifp, rt, -1);
1274 	MFC_UNLOCK();
1275 	VIF_UNLOCK();
1276 	return error;
1277     } else {
1278 	/*
1279 	 * If we don't have a route for packet's origin,
1280 	 * Make a copy of the packet & send message to routing daemon
1281 	 */
1282 
1283 	struct mbuf *mb0;
1284 	struct rtdetq *rte;
1285 	u_long hash;
1286 	int hlen = ip->ip_hl << 2;
1287 
1288 	MRTSTAT_INC(mrts_mfc_misses);
1289 	MRTSTAT_INC(mrts_no_route);
1290 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)",
1291 	    inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr));
1292 
1293 	/*
1294 	 * Allocate mbufs early so that we don't do extra work if we are
1295 	 * just going to fail anyway.  Make sure to pullup the header so
1296 	 * that other people can't step on it.
1297 	 */
1298 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1299 	    M_NOWAIT|M_ZERO);
1300 	if (rte == NULL) {
1301 	    MFC_UNLOCK();
1302 	    VIF_UNLOCK();
1303 	    return ENOBUFS;
1304 	}
1305 
1306 	mb0 = m_copypacket(m, M_NOWAIT);
1307 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1308 	    mb0 = m_pullup(mb0, hlen);
1309 	if (mb0 == NULL) {
1310 	    free(rte, M_MRTABLE);
1311 	    MFC_UNLOCK();
1312 	    VIF_UNLOCK();
1313 	    return ENOBUFS;
1314 	}
1315 
1316 	/* is there an upcall waiting for this flow ? */
1317 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1318 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1319 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1320 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1321 		    !TAILQ_EMPTY(&rt->mfc_stall))
1322 			break;
1323 	}
1324 
1325 	if (rt == NULL) {
1326 	    int i;
1327 	    struct igmpmsg *im;
1328 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1329 	    struct mbuf *mm;
1330 
1331 	    /*
1332 	     * Locate the vifi for the incoming interface for this packet.
1333 	     * If none found, drop packet.
1334 	     */
1335 	    for (vifi = 0; vifi < V_numvifs &&
1336 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1337 		;
1338 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1339 		goto non_fatal;
1340 
1341 	    /* no upcall, so make a new entry */
1342 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1343 	    if (rt == NULL)
1344 		goto fail;
1345 
1346 	    /* Make a copy of the header to send to the user level process */
1347 	    mm = m_copy(mb0, 0, hlen);
1348 	    if (mm == NULL)
1349 		goto fail1;
1350 
1351 	    /*
1352 	     * Send message to routing daemon to install
1353 	     * a route into the kernel table
1354 	     */
1355 
1356 	    im = mtod(mm, struct igmpmsg *);
1357 	    im->im_msgtype = IGMPMSG_NOCACHE;
1358 	    im->im_mbz = 0;
1359 	    im->im_vif = vifi;
1360 
1361 	    MRTSTAT_INC(mrts_upcalls);
1362 
1363 	    k_igmpsrc.sin_addr = ip->ip_src;
1364 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1365 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1366 		MRTSTAT_INC(mrts_upq_sockfull);
1367 fail1:
1368 		free(rt, M_MRTABLE);
1369 fail:
1370 		free(rte, M_MRTABLE);
1371 		m_freem(mb0);
1372 		MFC_UNLOCK();
1373 		VIF_UNLOCK();
1374 		return ENOBUFS;
1375 	    }
1376 
1377 	    /* insert new entry at head of hash chain */
1378 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1379 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1380 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1381 	    V_nexpire[hash]++;
1382 	    for (i = 0; i < V_numvifs; i++) {
1383 		rt->mfc_ttls[i] = 0;
1384 		rt->mfc_flags[i] = 0;
1385 	    }
1386 	    rt->mfc_parent = -1;
1387 
1388 	    /* clear the RP address */
1389 	    rt->mfc_rp.s_addr = INADDR_ANY;
1390 	    rt->mfc_bw_meter = NULL;
1391 
1392 	    /* initialize pkt counters per src-grp */
1393 	    rt->mfc_pkt_cnt = 0;
1394 	    rt->mfc_byte_cnt = 0;
1395 	    rt->mfc_wrong_if = 0;
1396 	    timevalclear(&rt->mfc_last_assert);
1397 
1398 	    TAILQ_INIT(&rt->mfc_stall);
1399 	    rt->mfc_nstall = 0;
1400 
1401 	    /* link into table */
1402 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1403 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1404 	    rt->mfc_nstall++;
1405 
1406 	} else {
1407 	    /* determine if queue has overflowed */
1408 	    if (rt->mfc_nstall > MAX_UPQ) {
1409 		MRTSTAT_INC(mrts_upq_ovflw);
1410 non_fatal:
1411 		free(rte, M_MRTABLE);
1412 		m_freem(mb0);
1413 		MFC_UNLOCK();
1414 		VIF_UNLOCK();
1415 		return (0);
1416 	    }
1417 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1418 	    rt->mfc_nstall++;
1419 	}
1420 
1421 	rte->m			= mb0;
1422 	rte->ifp		= ifp;
1423 
1424 	MFC_UNLOCK();
1425 	VIF_UNLOCK();
1426 
1427 	return 0;
1428     }
1429 }
1430 
1431 /*
1432  * Clean up the cache entry if upcall is not serviced
1433  */
1434 static void
1435 expire_upcalls(void *arg)
1436 {
1437     u_long i;
1438 
1439     CURVNET_SET((struct vnet *) arg);
1440 
1441     MFC_LOCK();
1442 
1443     for (i = 0; i < mfchashsize; i++) {
1444 	struct mfc *rt, *nrt;
1445 
1446 	if (V_nexpire[i] == 0)
1447 	    continue;
1448 
1449 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1450 		if (TAILQ_EMPTY(&rt->mfc_stall))
1451 			continue;
1452 
1453 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1454 			continue;
1455 
1456 		/*
1457 		 * free the bw_meter entries
1458 		 */
1459 		while (rt->mfc_bw_meter != NULL) {
1460 		    struct bw_meter *x = rt->mfc_bw_meter;
1461 
1462 		    rt->mfc_bw_meter = x->bm_mfc_next;
1463 		    free(x, M_BWMETER);
1464 		}
1465 
1466 		MRTSTAT_INC(mrts_cache_cleanups);
1467 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1468 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1469 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1470 
1471 		expire_mfc(rt);
1472 	    }
1473     }
1474 
1475     MFC_UNLOCK();
1476 
1477     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1478 	curvnet);
1479 
1480     CURVNET_RESTORE();
1481 }
1482 
1483 /*
1484  * Packet forwarding routine once entry in the cache is made
1485  */
1486 static int
1487 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1488 {
1489     struct ip  *ip = mtod(m, struct ip *);
1490     vifi_t vifi;
1491     int plen = ntohs(ip->ip_len);
1492 
1493     VIF_LOCK_ASSERT();
1494 
1495     /*
1496      * If xmt_vif is not -1, send on only the requested vif.
1497      *
1498      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1499      */
1500     if (xmt_vif < V_numvifs) {
1501 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1502 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1503 	else
1504 		phyint_send(ip, V_viftable + xmt_vif, m);
1505 	return 1;
1506     }
1507 
1508     /*
1509      * Don't forward if it didn't arrive from the parent vif for its origin.
1510      */
1511     vifi = rt->mfc_parent;
1512     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1513 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1514 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1515 	MRTSTAT_INC(mrts_wrong_if);
1516 	++rt->mfc_wrong_if;
1517 	/*
1518 	 * If we are doing PIM assert processing, send a message
1519 	 * to the routing daemon.
1520 	 *
1521 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1522 	 * can complete the SPT switch, regardless of the type
1523 	 * of the iif (broadcast media, GRE tunnel, etc).
1524 	 */
1525 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1526 	    V_viftable[vifi].v_ifp) {
1527 
1528 	    if (ifp == &V_multicast_register_if)
1529 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1530 
1531 	    /* Get vifi for the incoming packet */
1532 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1533 		vifi++)
1534 		;
1535 	    if (vifi >= V_numvifs)
1536 		return 0;	/* The iif is not found: ignore the packet. */
1537 
1538 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1539 		return 0;	/* WRONGVIF disabled: ignore the packet */
1540 
1541 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1542 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1543 		struct igmpmsg *im;
1544 		int hlen = ip->ip_hl << 2;
1545 		struct mbuf *mm = m_copy(m, 0, hlen);
1546 
1547 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1548 		    mm = m_pullup(mm, hlen);
1549 		if (mm == NULL)
1550 		    return ENOBUFS;
1551 
1552 		im = mtod(mm, struct igmpmsg *);
1553 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1554 		im->im_mbz		= 0;
1555 		im->im_vif		= vifi;
1556 
1557 		MRTSTAT_INC(mrts_upcalls);
1558 
1559 		k_igmpsrc.sin_addr = im->im_src;
1560 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1561 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1562 		    MRTSTAT_INC(mrts_upq_sockfull);
1563 		    return ENOBUFS;
1564 		}
1565 	    }
1566 	}
1567 	return 0;
1568     }
1569 
1570 
1571     /* If I sourced this packet, it counts as output, else it was input. */
1572     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1573 	V_viftable[vifi].v_pkt_out++;
1574 	V_viftable[vifi].v_bytes_out += plen;
1575     } else {
1576 	V_viftable[vifi].v_pkt_in++;
1577 	V_viftable[vifi].v_bytes_in += plen;
1578     }
1579     rt->mfc_pkt_cnt++;
1580     rt->mfc_byte_cnt += plen;
1581 
1582     /*
1583      * For each vif, decide if a copy of the packet should be forwarded.
1584      * Forward if:
1585      *		- the ttl exceeds the vif's threshold
1586      *		- there are group members downstream on interface
1587      */
1588     for (vifi = 0; vifi < V_numvifs; vifi++)
1589 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1590 	    V_viftable[vifi].v_pkt_out++;
1591 	    V_viftable[vifi].v_bytes_out += plen;
1592 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1593 		pim_register_send(ip, V_viftable + vifi, m, rt);
1594 	    else
1595 		phyint_send(ip, V_viftable + vifi, m);
1596 	}
1597 
1598     /*
1599      * Perform upcall-related bw measuring.
1600      */
1601     if (rt->mfc_bw_meter != NULL) {
1602 	struct bw_meter *x;
1603 	struct timeval now;
1604 
1605 	microtime(&now);
1606 	MFC_LOCK_ASSERT();
1607 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1608 	    bw_meter_receive_packet(x, plen, &now);
1609     }
1610 
1611     return 0;
1612 }
1613 
1614 /*
1615  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1616  */
1617 static int
1618 X_legal_vif_num(int vif)
1619 {
1620 	int ret;
1621 
1622 	ret = 0;
1623 	if (vif < 0)
1624 		return (ret);
1625 
1626 	VIF_LOCK();
1627 	if (vif < V_numvifs)
1628 		ret = 1;
1629 	VIF_UNLOCK();
1630 
1631 	return (ret);
1632 }
1633 
1634 /*
1635  * Return the local address used by this vif
1636  */
1637 static u_long
1638 X_ip_mcast_src(int vifi)
1639 {
1640 	in_addr_t addr;
1641 
1642 	addr = INADDR_ANY;
1643 	if (vifi < 0)
1644 		return (addr);
1645 
1646 	VIF_LOCK();
1647 	if (vifi < V_numvifs)
1648 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1649 	VIF_UNLOCK();
1650 
1651 	return (addr);
1652 }
1653 
1654 static void
1655 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1656 {
1657     struct mbuf *mb_copy;
1658     int hlen = ip->ip_hl << 2;
1659 
1660     VIF_LOCK_ASSERT();
1661 
1662     /*
1663      * Make a new reference to the packet; make sure that
1664      * the IP header is actually copied, not just referenced,
1665      * so that ip_output() only scribbles on the copy.
1666      */
1667     mb_copy = m_copypacket(m, M_NOWAIT);
1668     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1669 	mb_copy = m_pullup(mb_copy, hlen);
1670     if (mb_copy == NULL)
1671 	return;
1672 
1673     send_packet(vifp, mb_copy);
1674 }
1675 
1676 static void
1677 send_packet(struct vif *vifp, struct mbuf *m)
1678 {
1679 	struct ip_moptions imo;
1680 	struct in_multi *imm[2];
1681 	int error;
1682 
1683 	VIF_LOCK_ASSERT();
1684 
1685 	imo.imo_multicast_ifp  = vifp->v_ifp;
1686 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1687 	imo.imo_multicast_loop = 1;
1688 	imo.imo_multicast_vif  = -1;
1689 	imo.imo_num_memberships = 0;
1690 	imo.imo_max_memberships = 2;
1691 	imo.imo_membership  = &imm[0];
1692 
1693 	/*
1694 	 * Re-entrancy should not be a problem here, because
1695 	 * the packets that we send out and are looped back at us
1696 	 * should get rejected because they appear to come from
1697 	 * the loopback interface, thus preventing looping.
1698 	 */
1699 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1700 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1701 	    (ptrdiff_t)(vifp - V_viftable), error);
1702 }
1703 
1704 /*
1705  * Stubs for old RSVP socket shim implementation.
1706  */
1707 
1708 static int
1709 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1710 {
1711 
1712 	return (EOPNOTSUPP);
1713 }
1714 
1715 static void
1716 X_ip_rsvp_force_done(struct socket *so __unused)
1717 {
1718 
1719 }
1720 
1721 static int
1722 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1723 {
1724 	struct mbuf *m;
1725 
1726 	m = *mp;
1727 	*mp = NULL;
1728 	if (!V_rsvp_on)
1729 		m_freem(m);
1730 	return (IPPROTO_DONE);
1731 }
1732 
1733 /*
1734  * Code for bandwidth monitors
1735  */
1736 
1737 /*
1738  * Define common interface for timeval-related methods
1739  */
1740 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1741 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1742 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1743 
1744 static uint32_t
1745 compute_bw_meter_flags(struct bw_upcall *req)
1746 {
1747     uint32_t flags = 0;
1748 
1749     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1750 	flags |= BW_METER_UNIT_PACKETS;
1751     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1752 	flags |= BW_METER_UNIT_BYTES;
1753     if (req->bu_flags & BW_UPCALL_GEQ)
1754 	flags |= BW_METER_GEQ;
1755     if (req->bu_flags & BW_UPCALL_LEQ)
1756 	flags |= BW_METER_LEQ;
1757 
1758     return flags;
1759 }
1760 
1761 /*
1762  * Add a bw_meter entry
1763  */
1764 static int
1765 add_bw_upcall(struct bw_upcall *req)
1766 {
1767     struct mfc *mfc;
1768     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1769 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1770     struct timeval now;
1771     struct bw_meter *x;
1772     uint32_t flags;
1773 
1774     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1775 	return EOPNOTSUPP;
1776 
1777     /* Test if the flags are valid */
1778     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1779 	return EINVAL;
1780     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1781 	return EINVAL;
1782     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1783 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1784 	return EINVAL;
1785 
1786     /* Test if the threshold time interval is valid */
1787     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1788 	return EINVAL;
1789 
1790     flags = compute_bw_meter_flags(req);
1791 
1792     /*
1793      * Find if we have already same bw_meter entry
1794      */
1795     MFC_LOCK();
1796     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1797     if (mfc == NULL) {
1798 	MFC_UNLOCK();
1799 	return EADDRNOTAVAIL;
1800     }
1801     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1802 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1803 			   &req->bu_threshold.b_time, ==)) &&
1804 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1805 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1806 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1807 	    MFC_UNLOCK();
1808 	    return 0;		/* XXX Already installed */
1809 	}
1810     }
1811 
1812     /* Allocate the new bw_meter entry */
1813     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1814     if (x == NULL) {
1815 	MFC_UNLOCK();
1816 	return ENOBUFS;
1817     }
1818 
1819     /* Set the new bw_meter entry */
1820     x->bm_threshold.b_time = req->bu_threshold.b_time;
1821     microtime(&now);
1822     x->bm_start_time = now;
1823     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1824     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1825     x->bm_measured.b_packets = 0;
1826     x->bm_measured.b_bytes = 0;
1827     x->bm_flags = flags;
1828     x->bm_time_next = NULL;
1829     x->bm_time_hash = BW_METER_BUCKETS;
1830 
1831     /* Add the new bw_meter entry to the front of entries for this MFC */
1832     x->bm_mfc = mfc;
1833     x->bm_mfc_next = mfc->mfc_bw_meter;
1834     mfc->mfc_bw_meter = x;
1835     schedule_bw_meter(x, &now);
1836     MFC_UNLOCK();
1837 
1838     return 0;
1839 }
1840 
1841 static void
1842 free_bw_list(struct bw_meter *list)
1843 {
1844     while (list != NULL) {
1845 	struct bw_meter *x = list;
1846 
1847 	list = list->bm_mfc_next;
1848 	unschedule_bw_meter(x);
1849 	free(x, M_BWMETER);
1850     }
1851 }
1852 
1853 /*
1854  * Delete one or multiple bw_meter entries
1855  */
1856 static int
1857 del_bw_upcall(struct bw_upcall *req)
1858 {
1859     struct mfc *mfc;
1860     struct bw_meter *x;
1861 
1862     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1863 	return EOPNOTSUPP;
1864 
1865     MFC_LOCK();
1866 
1867     /* Find the corresponding MFC entry */
1868     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1869     if (mfc == NULL) {
1870 	MFC_UNLOCK();
1871 	return EADDRNOTAVAIL;
1872     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1873 	/*
1874 	 * Delete all bw_meter entries for this mfc
1875 	 */
1876 	struct bw_meter *list;
1877 
1878 	list = mfc->mfc_bw_meter;
1879 	mfc->mfc_bw_meter = NULL;
1880 	free_bw_list(list);
1881 	MFC_UNLOCK();
1882 	return 0;
1883     } else {			/* Delete a single bw_meter entry */
1884 	struct bw_meter *prev;
1885 	uint32_t flags = 0;
1886 
1887 	flags = compute_bw_meter_flags(req);
1888 
1889 	/* Find the bw_meter entry to delete */
1890 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1891 	     prev = x, x = x->bm_mfc_next) {
1892 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1893 			       &req->bu_threshold.b_time, ==)) &&
1894 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1895 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1896 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1897 		break;
1898 	}
1899 	if (x != NULL) { /* Delete entry from the list for this MFC */
1900 	    if (prev != NULL)
1901 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1902 	    else
1903 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1904 
1905 	    unschedule_bw_meter(x);
1906 	    MFC_UNLOCK();
1907 	    /* Free the bw_meter entry */
1908 	    free(x, M_BWMETER);
1909 	    return 0;
1910 	} else {
1911 	    MFC_UNLOCK();
1912 	    return EINVAL;
1913 	}
1914     }
1915     /* NOTREACHED */
1916 }
1917 
1918 /*
1919  * Perform bandwidth measurement processing that may result in an upcall
1920  */
1921 static void
1922 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1923 {
1924     struct timeval delta;
1925 
1926     MFC_LOCK_ASSERT();
1927 
1928     delta = *nowp;
1929     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1930 
1931     if (x->bm_flags & BW_METER_GEQ) {
1932 	/*
1933 	 * Processing for ">=" type of bw_meter entry
1934 	 */
1935 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1936 	    /* Reset the bw_meter entry */
1937 	    x->bm_start_time = *nowp;
1938 	    x->bm_measured.b_packets = 0;
1939 	    x->bm_measured.b_bytes = 0;
1940 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1941 	}
1942 
1943 	/* Record that a packet is received */
1944 	x->bm_measured.b_packets++;
1945 	x->bm_measured.b_bytes += plen;
1946 
1947 	/*
1948 	 * Test if we should deliver an upcall
1949 	 */
1950 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1951 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1952 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1953 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1954 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1955 		/* Prepare an upcall for delivery */
1956 		bw_meter_prepare_upcall(x, nowp);
1957 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1958 	    }
1959 	}
1960     } else if (x->bm_flags & BW_METER_LEQ) {
1961 	/*
1962 	 * Processing for "<=" type of bw_meter entry
1963 	 */
1964 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1965 	    /*
1966 	     * We are behind time with the multicast forwarding table
1967 	     * scanning for "<=" type of bw_meter entries, so test now
1968 	     * if we should deliver an upcall.
1969 	     */
1970 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1971 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1972 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1973 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1974 		/* Prepare an upcall for delivery */
1975 		bw_meter_prepare_upcall(x, nowp);
1976 	    }
1977 	    /* Reschedule the bw_meter entry */
1978 	    unschedule_bw_meter(x);
1979 	    schedule_bw_meter(x, nowp);
1980 	}
1981 
1982 	/* Record that a packet is received */
1983 	x->bm_measured.b_packets++;
1984 	x->bm_measured.b_bytes += plen;
1985 
1986 	/*
1987 	 * Test if we should restart the measuring interval
1988 	 */
1989 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1990 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1991 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1992 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1993 	    /* Don't restart the measuring interval */
1994 	} else {
1995 	    /* Do restart the measuring interval */
1996 	    /*
1997 	     * XXX: note that we don't unschedule and schedule, because this
1998 	     * might be too much overhead per packet. Instead, when we process
1999 	     * all entries for a given timer hash bin, we check whether it is
2000 	     * really a timeout. If not, we reschedule at that time.
2001 	     */
2002 	    x->bm_start_time = *nowp;
2003 	    x->bm_measured.b_packets = 0;
2004 	    x->bm_measured.b_bytes = 0;
2005 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2006 	}
2007     }
2008 }
2009 
2010 /*
2011  * Prepare a bandwidth-related upcall
2012  */
2013 static void
2014 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2015 {
2016     struct timeval delta;
2017     struct bw_upcall *u;
2018 
2019     MFC_LOCK_ASSERT();
2020 
2021     /*
2022      * Compute the measured time interval
2023      */
2024     delta = *nowp;
2025     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2026 
2027     /*
2028      * If there are too many pending upcalls, deliver them now
2029      */
2030     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2031 	bw_upcalls_send();
2032 
2033     /*
2034      * Set the bw_upcall entry
2035      */
2036     u = &V_bw_upcalls[V_bw_upcalls_n++];
2037     u->bu_src = x->bm_mfc->mfc_origin;
2038     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2039     u->bu_threshold.b_time = x->bm_threshold.b_time;
2040     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2041     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2042     u->bu_measured.b_time = delta;
2043     u->bu_measured.b_packets = x->bm_measured.b_packets;
2044     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2045     u->bu_flags = 0;
2046     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2047 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2048     if (x->bm_flags & BW_METER_UNIT_BYTES)
2049 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2050     if (x->bm_flags & BW_METER_GEQ)
2051 	u->bu_flags |= BW_UPCALL_GEQ;
2052     if (x->bm_flags & BW_METER_LEQ)
2053 	u->bu_flags |= BW_UPCALL_LEQ;
2054 }
2055 
2056 /*
2057  * Send the pending bandwidth-related upcalls
2058  */
2059 static void
2060 bw_upcalls_send(void)
2061 {
2062     struct mbuf *m;
2063     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2064     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2065     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2066 				      0,		/* unused2 */
2067 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2068 				      0,		/* im_mbz  */
2069 				      0,		/* im_vif  */
2070 				      0,		/* unused3 */
2071 				      { 0 },		/* im_src  */
2072 				      { 0 } };		/* im_dst  */
2073 
2074     MFC_LOCK_ASSERT();
2075 
2076     if (V_bw_upcalls_n == 0)
2077 	return;			/* No pending upcalls */
2078 
2079     V_bw_upcalls_n = 0;
2080 
2081     /*
2082      * Allocate a new mbuf, initialize it with the header and
2083      * the payload for the pending calls.
2084      */
2085     m = m_gethdr(M_NOWAIT, MT_DATA);
2086     if (m == NULL) {
2087 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2088 	return;
2089     }
2090 
2091     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2092     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2093 
2094     /*
2095      * Send the upcalls
2096      * XXX do we need to set the address in k_igmpsrc ?
2097      */
2098     MRTSTAT_INC(mrts_upcalls);
2099     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2100 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2101 	MRTSTAT_INC(mrts_upq_sockfull);
2102     }
2103 }
2104 
2105 /*
2106  * Compute the timeout hash value for the bw_meter entries
2107  */
2108 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2109     do {								\
2110 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2111 									\
2112 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2113 	(hash) = next_timeval.tv_sec;					\
2114 	if (next_timeval.tv_usec)					\
2115 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2116 	(hash) %= BW_METER_BUCKETS;					\
2117     } while (0)
2118 
2119 /*
2120  * Schedule a timer to process periodically bw_meter entry of type "<="
2121  * by linking the entry in the proper hash bucket.
2122  */
2123 static void
2124 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2125 {
2126     int time_hash;
2127 
2128     MFC_LOCK_ASSERT();
2129 
2130     if (!(x->bm_flags & BW_METER_LEQ))
2131 	return;		/* XXX: we schedule timers only for "<=" entries */
2132 
2133     /*
2134      * Reset the bw_meter entry
2135      */
2136     x->bm_start_time = *nowp;
2137     x->bm_measured.b_packets = 0;
2138     x->bm_measured.b_bytes = 0;
2139     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2140 
2141     /*
2142      * Compute the timeout hash value and insert the entry
2143      */
2144     BW_METER_TIMEHASH(x, time_hash);
2145     x->bm_time_next = V_bw_meter_timers[time_hash];
2146     V_bw_meter_timers[time_hash] = x;
2147     x->bm_time_hash = time_hash;
2148 }
2149 
2150 /*
2151  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2152  * by removing the entry from the proper hash bucket.
2153  */
2154 static void
2155 unschedule_bw_meter(struct bw_meter *x)
2156 {
2157     int time_hash;
2158     struct bw_meter *prev, *tmp;
2159 
2160     MFC_LOCK_ASSERT();
2161 
2162     if (!(x->bm_flags & BW_METER_LEQ))
2163 	return;		/* XXX: we schedule timers only for "<=" entries */
2164 
2165     /*
2166      * Compute the timeout hash value and delete the entry
2167      */
2168     time_hash = x->bm_time_hash;
2169     if (time_hash >= BW_METER_BUCKETS)
2170 	return;		/* Entry was not scheduled */
2171 
2172     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2173 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2174 	if (tmp == x)
2175 	    break;
2176 
2177     if (tmp == NULL)
2178 	panic("unschedule_bw_meter: bw_meter entry not found");
2179 
2180     if (prev != NULL)
2181 	prev->bm_time_next = x->bm_time_next;
2182     else
2183 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2184 
2185     x->bm_time_next = NULL;
2186     x->bm_time_hash = BW_METER_BUCKETS;
2187 }
2188 
2189 
2190 /*
2191  * Process all "<=" type of bw_meter that should be processed now,
2192  * and for each entry prepare an upcall if necessary. Each processed
2193  * entry is rescheduled again for the (periodic) processing.
2194  *
2195  * This is run periodically (once per second normally). On each round,
2196  * all the potentially matching entries are in the hash slot that we are
2197  * looking at.
2198  */
2199 static void
2200 bw_meter_process()
2201 {
2202     uint32_t loops;
2203     int i;
2204     struct timeval now, process_endtime;
2205 
2206     microtime(&now);
2207     if (V_last_tv_sec == now.tv_sec)
2208 	return;		/* nothing to do */
2209 
2210     loops = now.tv_sec - V_last_tv_sec;
2211     V_last_tv_sec = now.tv_sec;
2212     if (loops > BW_METER_BUCKETS)
2213 	loops = BW_METER_BUCKETS;
2214 
2215     MFC_LOCK();
2216     /*
2217      * Process all bins of bw_meter entries from the one after the last
2218      * processed to the current one. On entry, i points to the last bucket
2219      * visited, so we need to increment i at the beginning of the loop.
2220      */
2221     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2222 	struct bw_meter *x, *tmp_list;
2223 
2224 	if (++i >= BW_METER_BUCKETS)
2225 	    i = 0;
2226 
2227 	/* Disconnect the list of bw_meter entries from the bin */
2228 	tmp_list = V_bw_meter_timers[i];
2229 	V_bw_meter_timers[i] = NULL;
2230 
2231 	/* Process the list of bw_meter entries */
2232 	while (tmp_list != NULL) {
2233 	    x = tmp_list;
2234 	    tmp_list = tmp_list->bm_time_next;
2235 
2236 	    /* Test if the time interval is over */
2237 	    process_endtime = x->bm_start_time;
2238 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2239 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2240 		/* Not yet: reschedule, but don't reset */
2241 		int time_hash;
2242 
2243 		BW_METER_TIMEHASH(x, time_hash);
2244 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2245 		    /*
2246 		     * XXX: somehow the bin processing is a bit ahead of time.
2247 		     * Put the entry in the next bin.
2248 		     */
2249 		    if (++time_hash >= BW_METER_BUCKETS)
2250 			time_hash = 0;
2251 		}
2252 		x->bm_time_next = V_bw_meter_timers[time_hash];
2253 		V_bw_meter_timers[time_hash] = x;
2254 		x->bm_time_hash = time_hash;
2255 
2256 		continue;
2257 	    }
2258 
2259 	    /*
2260 	     * Test if we should deliver an upcall
2261 	     */
2262 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2263 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2264 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2265 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2266 		/* Prepare an upcall for delivery */
2267 		bw_meter_prepare_upcall(x, &now);
2268 	    }
2269 
2270 	    /*
2271 	     * Reschedule for next processing
2272 	     */
2273 	    schedule_bw_meter(x, &now);
2274 	}
2275     }
2276 
2277     /* Send all upcalls that are pending delivery */
2278     bw_upcalls_send();
2279 
2280     MFC_UNLOCK();
2281 }
2282 
2283 /*
2284  * A periodic function for sending all upcalls that are pending delivery
2285  */
2286 static void
2287 expire_bw_upcalls_send(void *arg)
2288 {
2289     CURVNET_SET((struct vnet *) arg);
2290 
2291     MFC_LOCK();
2292     bw_upcalls_send();
2293     MFC_UNLOCK();
2294 
2295     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2296 	curvnet);
2297     CURVNET_RESTORE();
2298 }
2299 
2300 /*
2301  * A periodic function for periodic scanning of the multicast forwarding
2302  * table for processing all "<=" bw_meter entries.
2303  */
2304 static void
2305 expire_bw_meter_process(void *arg)
2306 {
2307     CURVNET_SET((struct vnet *) arg);
2308 
2309     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2310 	bw_meter_process();
2311 
2312     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2313 	curvnet);
2314     CURVNET_RESTORE();
2315 }
2316 
2317 /*
2318  * End of bandwidth monitoring code
2319  */
2320 
2321 /*
2322  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2323  *
2324  */
2325 static int
2326 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2327     struct mfc *rt)
2328 {
2329     struct mbuf *mb_copy, *mm;
2330 
2331     /*
2332      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2333      * rendezvous point was unspecified, and we were told not to.
2334      */
2335     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2336 	in_nullhost(rt->mfc_rp))
2337 	return 0;
2338 
2339     mb_copy = pim_register_prepare(ip, m);
2340     if (mb_copy == NULL)
2341 	return ENOBUFS;
2342 
2343     /*
2344      * Send all the fragments. Note that the mbuf for each fragment
2345      * is freed by the sending machinery.
2346      */
2347     for (mm = mb_copy; mm; mm = mb_copy) {
2348 	mb_copy = mm->m_nextpkt;
2349 	mm->m_nextpkt = 0;
2350 	mm = m_pullup(mm, sizeof(struct ip));
2351 	if (mm != NULL) {
2352 	    ip = mtod(mm, struct ip *);
2353 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2354 		pim_register_send_rp(ip, vifp, mm, rt);
2355 	    } else {
2356 		pim_register_send_upcall(ip, vifp, mm, rt);
2357 	    }
2358 	}
2359     }
2360 
2361     return 0;
2362 }
2363 
2364 /*
2365  * Return a copy of the data packet that is ready for PIM Register
2366  * encapsulation.
2367  * XXX: Note that in the returned copy the IP header is a valid one.
2368  */
2369 static struct mbuf *
2370 pim_register_prepare(struct ip *ip, struct mbuf *m)
2371 {
2372     struct mbuf *mb_copy = NULL;
2373     int mtu;
2374 
2375     /* Take care of delayed checksums */
2376     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2377 	in_delayed_cksum(m);
2378 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2379     }
2380 
2381     /*
2382      * Copy the old packet & pullup its IP header into the
2383      * new mbuf so we can modify it.
2384      */
2385     mb_copy = m_copypacket(m, M_NOWAIT);
2386     if (mb_copy == NULL)
2387 	return NULL;
2388     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2389     if (mb_copy == NULL)
2390 	return NULL;
2391 
2392     /* take care of the TTL */
2393     ip = mtod(mb_copy, struct ip *);
2394     --ip->ip_ttl;
2395 
2396     /* Compute the MTU after the PIM Register encapsulation */
2397     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2398 
2399     if (ntohs(ip->ip_len) <= mtu) {
2400 	/* Turn the IP header into a valid one */
2401 	ip->ip_sum = 0;
2402 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2403     } else {
2404 	/* Fragment the packet */
2405 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2406 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2407 	    m_freem(mb_copy);
2408 	    return NULL;
2409 	}
2410     }
2411     return mb_copy;
2412 }
2413 
2414 /*
2415  * Send an upcall with the data packet to the user-level process.
2416  */
2417 static int
2418 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2419     struct mbuf *mb_copy, struct mfc *rt)
2420 {
2421     struct mbuf *mb_first;
2422     int len = ntohs(ip->ip_len);
2423     struct igmpmsg *im;
2424     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2425 
2426     VIF_LOCK_ASSERT();
2427 
2428     /*
2429      * Add a new mbuf with an upcall header
2430      */
2431     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2432     if (mb_first == NULL) {
2433 	m_freem(mb_copy);
2434 	return ENOBUFS;
2435     }
2436     mb_first->m_data += max_linkhdr;
2437     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2438     mb_first->m_len = sizeof(struct igmpmsg);
2439     mb_first->m_next = mb_copy;
2440 
2441     /* Send message to routing daemon */
2442     im = mtod(mb_first, struct igmpmsg *);
2443     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2444     im->im_mbz		= 0;
2445     im->im_vif		= vifp - V_viftable;
2446     im->im_src		= ip->ip_src;
2447     im->im_dst		= ip->ip_dst;
2448 
2449     k_igmpsrc.sin_addr	= ip->ip_src;
2450 
2451     MRTSTAT_INC(mrts_upcalls);
2452 
2453     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2454 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2455 	MRTSTAT_INC(mrts_upq_sockfull);
2456 	return ENOBUFS;
2457     }
2458 
2459     /* Keep statistics */
2460     PIMSTAT_INC(pims_snd_registers_msgs);
2461     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2462 
2463     return 0;
2464 }
2465 
2466 /*
2467  * Encapsulate the data packet in PIM Register message and send it to the RP.
2468  */
2469 static int
2470 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2471     struct mfc *rt)
2472 {
2473     struct mbuf *mb_first;
2474     struct ip *ip_outer;
2475     struct pim_encap_pimhdr *pimhdr;
2476     int len = ntohs(ip->ip_len);
2477     vifi_t vifi = rt->mfc_parent;
2478 
2479     VIF_LOCK_ASSERT();
2480 
2481     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2482 	m_freem(mb_copy);
2483 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2484     }
2485 
2486     /*
2487      * Add a new mbuf with the encapsulating header
2488      */
2489     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2490     if (mb_first == NULL) {
2491 	m_freem(mb_copy);
2492 	return ENOBUFS;
2493     }
2494     mb_first->m_data += max_linkhdr;
2495     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2496     mb_first->m_next = mb_copy;
2497 
2498     mb_first->m_pkthdr.len = len + mb_first->m_len;
2499 
2500     /*
2501      * Fill in the encapsulating IP and PIM header
2502      */
2503     ip_outer = mtod(mb_first, struct ip *);
2504     *ip_outer = pim_encap_iphdr;
2505     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2506 	sizeof(pim_encap_pimhdr));
2507     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2508     ip_outer->ip_dst = rt->mfc_rp;
2509     /*
2510      * Copy the inner header TOS to the outer header, and take care of the
2511      * IP_DF bit.
2512      */
2513     ip_outer->ip_tos = ip->ip_tos;
2514     if (ip->ip_off & htons(IP_DF))
2515 	ip_outer->ip_off |= htons(IP_DF);
2516     ip_fillid(ip_outer);
2517     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2518 					 + sizeof(pim_encap_iphdr));
2519     *pimhdr = pim_encap_pimhdr;
2520     /* If the iif crosses a border, set the Border-bit */
2521     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2522 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2523 
2524     mb_first->m_data += sizeof(pim_encap_iphdr);
2525     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2526     mb_first->m_data -= sizeof(pim_encap_iphdr);
2527 
2528     send_packet(vifp, mb_first);
2529 
2530     /* Keep statistics */
2531     PIMSTAT_INC(pims_snd_registers_msgs);
2532     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2533 
2534     return 0;
2535 }
2536 
2537 /*
2538  * pim_encapcheck() is called by the encap4_input() path at runtime to
2539  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2540  * into the kernel.
2541  */
2542 static int
2543 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2544 {
2545 
2546 #ifdef DIAGNOSTIC
2547     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2548 #endif
2549     if (proto != IPPROTO_PIM)
2550 	return 0;	/* not for us; reject the datagram. */
2551 
2552     return 64;		/* claim the datagram. */
2553 }
2554 
2555 /*
2556  * PIM-SMv2 and PIM-DM messages processing.
2557  * Receives and verifies the PIM control messages, and passes them
2558  * up to the listening socket, using rip_input().
2559  * The only message with special processing is the PIM_REGISTER message
2560  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2561  * is passed to if_simloop().
2562  */
2563 int
2564 pim_input(struct mbuf **mp, int *offp, int proto)
2565 {
2566     struct mbuf *m = *mp;
2567     struct ip *ip = mtod(m, struct ip *);
2568     struct pim *pim;
2569     int iphlen = *offp;
2570     int minlen;
2571     int datalen = ntohs(ip->ip_len) - iphlen;
2572     int ip_tos;
2573 
2574     *mp = NULL;
2575 
2576     /* Keep statistics */
2577     PIMSTAT_INC(pims_rcv_total_msgs);
2578     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2579 
2580     /*
2581      * Validate lengths
2582      */
2583     if (datalen < PIM_MINLEN) {
2584 	PIMSTAT_INC(pims_rcv_tooshort);
2585 	CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2586 	    __func__, datalen, inet_ntoa(ip->ip_src));
2587 	m_freem(m);
2588 	return (IPPROTO_DONE);
2589     }
2590 
2591     /*
2592      * If the packet is at least as big as a REGISTER, go agead
2593      * and grab the PIM REGISTER header size, to avoid another
2594      * possible m_pullup() later.
2595      *
2596      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2597      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2598      */
2599     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2600     /*
2601      * Get the IP and PIM headers in contiguous memory, and
2602      * possibly the PIM REGISTER header.
2603      */
2604     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == 0) {
2605 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2606 	return (IPPROTO_DONE);
2607     }
2608 
2609     /* m_pullup() may have given us a new mbuf so reset ip. */
2610     ip = mtod(m, struct ip *);
2611     ip_tos = ip->ip_tos;
2612 
2613     /* adjust mbuf to point to the PIM header */
2614     m->m_data += iphlen;
2615     m->m_len  -= iphlen;
2616     pim = mtod(m, struct pim *);
2617 
2618     /*
2619      * Validate checksum. If PIM REGISTER, exclude the data packet.
2620      *
2621      * XXX: some older PIMv2 implementations don't make this distinction,
2622      * so for compatibility reason perform the checksum over part of the
2623      * message, and if error, then over the whole message.
2624      */
2625     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2626 	/* do nothing, checksum okay */
2627     } else if (in_cksum(m, datalen)) {
2628 	PIMSTAT_INC(pims_rcv_badsum);
2629 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2630 	m_freem(m);
2631 	return (IPPROTO_DONE);
2632     }
2633 
2634     /* PIM version check */
2635     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2636 	PIMSTAT_INC(pims_rcv_badversion);
2637 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2638 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2639 	m_freem(m);
2640 	return (IPPROTO_DONE);
2641     }
2642 
2643     /* restore mbuf back to the outer IP */
2644     m->m_data -= iphlen;
2645     m->m_len  += iphlen;
2646 
2647     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2648 	/*
2649 	 * Since this is a REGISTER, we'll make a copy of the register
2650 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2651 	 * routing daemon.
2652 	 */
2653 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2654 	struct mbuf *mcp;
2655 	struct ip *encap_ip;
2656 	u_int32_t *reghdr;
2657 	struct ifnet *vifp;
2658 
2659 	VIF_LOCK();
2660 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2661 	    VIF_UNLOCK();
2662 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2663 		(int)V_reg_vif_num);
2664 	    m_freem(m);
2665 	    return (IPPROTO_DONE);
2666 	}
2667 	/* XXX need refcnt? */
2668 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2669 	VIF_UNLOCK();
2670 
2671 	/*
2672 	 * Validate length
2673 	 */
2674 	if (datalen < PIM_REG_MINLEN) {
2675 	    PIMSTAT_INC(pims_rcv_tooshort);
2676 	    PIMSTAT_INC(pims_rcv_badregisters);
2677 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2678 	    m_freem(m);
2679 	    return (IPPROTO_DONE);
2680 	}
2681 
2682 	reghdr = (u_int32_t *)(pim + 1);
2683 	encap_ip = (struct ip *)(reghdr + 1);
2684 
2685 	CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2686 	    __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2687 
2688 	/* verify the version number of the inner packet */
2689 	if (encap_ip->ip_v != IPVERSION) {
2690 	    PIMSTAT_INC(pims_rcv_badregisters);
2691 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2692 	    m_freem(m);
2693 	    return (IPPROTO_DONE);
2694 	}
2695 
2696 	/* verify the inner packet is destined to a mcast group */
2697 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2698 	    PIMSTAT_INC(pims_rcv_badregisters);
2699 	    CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2700 		inet_ntoa(encap_ip->ip_dst));
2701 	    m_freem(m);
2702 	    return (IPPROTO_DONE);
2703 	}
2704 
2705 	/* If a NULL_REGISTER, pass it to the daemon */
2706 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2707 	    goto pim_input_to_daemon;
2708 
2709 	/*
2710 	 * Copy the TOS from the outer IP header to the inner IP header.
2711 	 */
2712 	if (encap_ip->ip_tos != ip_tos) {
2713 	    /* Outer TOS -> inner TOS */
2714 	    encap_ip->ip_tos = ip_tos;
2715 	    /* Recompute the inner header checksum. Sigh... */
2716 
2717 	    /* adjust mbuf to point to the inner IP header */
2718 	    m->m_data += (iphlen + PIM_MINLEN);
2719 	    m->m_len  -= (iphlen + PIM_MINLEN);
2720 
2721 	    encap_ip->ip_sum = 0;
2722 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2723 
2724 	    /* restore mbuf to point back to the outer IP header */
2725 	    m->m_data -= (iphlen + PIM_MINLEN);
2726 	    m->m_len  += (iphlen + PIM_MINLEN);
2727 	}
2728 
2729 	/*
2730 	 * Decapsulate the inner IP packet and loopback to forward it
2731 	 * as a normal multicast packet. Also, make a copy of the
2732 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2733 	 * to pass to the daemon later, so it can take the appropriate
2734 	 * actions (e.g., send back PIM_REGISTER_STOP).
2735 	 * XXX: here m->m_data points to the outer IP header.
2736 	 */
2737 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2738 	if (mcp == NULL) {
2739 	    CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2740 	    m_freem(m);
2741 	    return (IPPROTO_DONE);
2742 	}
2743 
2744 	/* Keep statistics */
2745 	/* XXX: registers_bytes include only the encap. mcast pkt */
2746 	PIMSTAT_INC(pims_rcv_registers_msgs);
2747 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2748 
2749 	/*
2750 	 * forward the inner ip packet; point m_data at the inner ip.
2751 	 */
2752 	m_adj(m, iphlen + PIM_MINLEN);
2753 
2754 	CTR4(KTR_IPMF,
2755 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2756 	    __func__,
2757 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2758 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2759 	    (int)V_reg_vif_num);
2760 
2761 	/* NB: vifp was collected above; can it change on us? */
2762 	if_simloop(vifp, m, dst.sin_family, 0);
2763 
2764 	/* prepare the register head to send to the mrouting daemon */
2765 	m = mcp;
2766     }
2767 
2768 pim_input_to_daemon:
2769     /*
2770      * Pass the PIM message up to the daemon; if it is a Register message,
2771      * pass the 'head' only up to the daemon. This includes the
2772      * outer IP header, PIM header, PIM-Register header and the
2773      * inner IP header.
2774      * XXX: the outer IP header pkt size of a Register is not adjust to
2775      * reflect the fact that the inner multicast data is truncated.
2776      */
2777     *mp = m;
2778     rip_input(mp, offp, proto);
2779 
2780     return (IPPROTO_DONE);
2781 }
2782 
2783 static int
2784 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2785 {
2786 	struct mfc	*rt;
2787 	int		 error, i;
2788 
2789 	if (req->newptr)
2790 		return (EPERM);
2791 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2792 		return (0);
2793 	error = sysctl_wire_old_buffer(req, 0);
2794 	if (error)
2795 		return (error);
2796 
2797 	MFC_LOCK();
2798 	for (i = 0; i < mfchashsize; i++) {
2799 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2800 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2801 			if (error)
2802 				goto out_locked;
2803 		}
2804 	}
2805 out_locked:
2806 	MFC_UNLOCK();
2807 	return (error);
2808 }
2809 
2810 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2811     sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2812     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2813 
2814 static void
2815 vnet_mroute_init(const void *unused __unused)
2816 {
2817 
2818 	MALLOC(V_nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2819 	bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2820 	callout_init(&V_expire_upcalls_ch, 1);
2821 	callout_init(&V_bw_upcalls_ch, 1);
2822 	callout_init(&V_bw_meter_ch, 1);
2823 }
2824 
2825 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mroute_init,
2826 	NULL);
2827 
2828 static void
2829 vnet_mroute_uninit(const void *unused __unused)
2830 {
2831 
2832 	FREE(V_nexpire, M_MRTABLE);
2833 	V_nexpire = NULL;
2834 }
2835 
2836 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE,
2837 	vnet_mroute_uninit, NULL);
2838 
2839 static int
2840 ip_mroute_modevent(module_t mod, int type, void *unused)
2841 {
2842 
2843     switch (type) {
2844     case MOD_LOAD:
2845 	MROUTER_LOCK_INIT();
2846 
2847 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2848 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2849 	if (if_detach_event_tag == NULL) {
2850 		printf("ip_mroute: unable to register "
2851 		    "ifnet_departure_event handler\n");
2852 		MROUTER_LOCK_DESTROY();
2853 		return (EINVAL);
2854 	}
2855 
2856 	MFC_LOCK_INIT();
2857 	VIF_LOCK_INIT();
2858 
2859 	mfchashsize = MFCHASHSIZE;
2860 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2861 	    !powerof2(mfchashsize)) {
2862 		printf("WARNING: %s not a power of 2; using default\n",
2863 		    "net.inet.ip.mfchashsize");
2864 		mfchashsize = MFCHASHSIZE;
2865 	}
2866 
2867 	pim_squelch_wholepkt = 0;
2868 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2869 	    &pim_squelch_wholepkt);
2870 
2871 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2872 	    pim_encapcheck, &in_pim_protosw, NULL);
2873 	if (pim_encap_cookie == NULL) {
2874 		printf("ip_mroute: unable to attach pim encap\n");
2875 		VIF_LOCK_DESTROY();
2876 		MFC_LOCK_DESTROY();
2877 		MROUTER_LOCK_DESTROY();
2878 		return (EINVAL);
2879 	}
2880 
2881 	ip_mcast_src = X_ip_mcast_src;
2882 	ip_mforward = X_ip_mforward;
2883 	ip_mrouter_done = X_ip_mrouter_done;
2884 	ip_mrouter_get = X_ip_mrouter_get;
2885 	ip_mrouter_set = X_ip_mrouter_set;
2886 
2887 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2888 	ip_rsvp_vif = X_ip_rsvp_vif;
2889 
2890 	legal_vif_num = X_legal_vif_num;
2891 	mrt_ioctl = X_mrt_ioctl;
2892 	rsvp_input_p = X_rsvp_input;
2893 	break;
2894 
2895     case MOD_UNLOAD:
2896 	/*
2897 	 * Typically module unload happens after the user-level
2898 	 * process has shutdown the kernel services (the check
2899 	 * below insures someone can't just yank the module out
2900 	 * from under a running process).  But if the module is
2901 	 * just loaded and then unloaded w/o starting up a user
2902 	 * process we still need to cleanup.
2903 	 */
2904 	MROUTER_LOCK();
2905 	if (ip_mrouter_cnt != 0) {
2906 	    MROUTER_UNLOCK();
2907 	    return (EINVAL);
2908 	}
2909 	ip_mrouter_unloading = 1;
2910 	MROUTER_UNLOCK();
2911 
2912 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2913 
2914 	if (pim_encap_cookie) {
2915 	    encap_detach(pim_encap_cookie);
2916 	    pim_encap_cookie = NULL;
2917 	}
2918 
2919 	ip_mcast_src = NULL;
2920 	ip_mforward = NULL;
2921 	ip_mrouter_done = NULL;
2922 	ip_mrouter_get = NULL;
2923 	ip_mrouter_set = NULL;
2924 
2925 	ip_rsvp_force_done = NULL;
2926 	ip_rsvp_vif = NULL;
2927 
2928 	legal_vif_num = NULL;
2929 	mrt_ioctl = NULL;
2930 	rsvp_input_p = NULL;
2931 
2932 	VIF_LOCK_DESTROY();
2933 	MFC_LOCK_DESTROY();
2934 	MROUTER_LOCK_DESTROY();
2935 	break;
2936 
2937     default:
2938 	return EOPNOTSUPP;
2939     }
2940     return 0;
2941 }
2942 
2943 static moduledata_t ip_mroutemod = {
2944     "ip_mroute",
2945     ip_mroute_modevent,
2946     0
2947 };
2948 
2949 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_MIDDLE);
2950