xref: /freebsd/sys/netinet/ip_mroute.c (revision 830940567b49bb0c08dfaed40418999e76616909)
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 #include <security/mac/mac_framework.h>
118 
119 #ifndef KTR_IPMF
120 #define KTR_IPMF KTR_INET
121 #endif
122 
123 #define		VIFI_INVALID	((vifi_t) -1)
124 #define		M_HASCL(m)	((m)->m_flags & M_EXT)
125 
126 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
127 
128 /*
129  * Locking.  We use two locks: one for the virtual interface table and
130  * one for the forwarding table.  These locks may be nested in which case
131  * the VIF lock must always be taken first.  Note that each lock is used
132  * to cover not only the specific data structure but also related data
133  * structures.
134  */
135 
136 static struct mtx mrouter_mtx;
137 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
138 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
139 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
140 #define	MROUTER_LOCK_INIT()						\
141 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
142 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
143 
144 static struct mrtstat	mrtstat;
145 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
146     &mrtstat, mrtstat,
147     "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
148     "netinet/ip_mroute.h)");
149 
150 static u_long			 mfchash;
151 #define MFCHASH(a, g)							\
152 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
153 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
154 #define MFCHASHSIZE	256
155 
156 static u_char			*nexpire;	/* 0..mfchashsize-1 */
157 static u_long			 mfchashsize;	/* Hash size */
158 LIST_HEAD(mfchashhdr, mfc)	*mfchashtbl;
159 
160 static struct mtx mfc_mtx;
161 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
162 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
163 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
164 #define	MFC_LOCK_INIT()							\
165 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
166 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
167 
168 static vifi_t		numvifs;
169 static struct vif	viftable[MAXVIFS];
170 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
171     &viftable, sizeof(viftable), "S,vif[MAXVIFS]",
172     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
173 
174 static struct mtx vif_mtx;
175 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
176 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
177 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
178 #define	VIF_LOCK_INIT()							\
179 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
180 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
181 
182 static eventhandler_tag if_detach_event_tag = NULL;
183 
184 static struct callout expire_upcalls_ch;
185 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
186 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
187 
188 /*
189  * Bandwidth meter variables and constants
190  */
191 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
192 /*
193  * Pending timeouts are stored in a hash table, the key being the
194  * expiration time. Periodically, the entries are analysed and processed.
195  */
196 #define BW_METER_BUCKETS	1024
197 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
198 static struct callout bw_meter_ch;
199 #define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
200 
201 /*
202  * Pending upcalls are stored in a vector which is flushed when
203  * full, or periodically
204  */
205 static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
206 static u_int	bw_upcalls_n; /* # of pending upcalls */
207 static struct callout bw_upcalls_ch;
208 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
209 
210 static struct pimstat pimstat;
211 
212 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
213 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
214     &pimstat, pimstat,
215     "PIM Statistics (struct pimstat, netinet/pim_var.h)");
216 
217 static u_long	pim_squelch_wholepkt = 0;
218 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
219     &pim_squelch_wholepkt, 0,
220     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
221 
222 extern  struct domain inetdomain;
223 static const struct protosw in_pim_protosw = {
224 	.pr_type =		SOCK_RAW,
225 	.pr_domain =		&inetdomain,
226 	.pr_protocol =		IPPROTO_PIM,
227 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
228 	.pr_input =		pim_input,
229 	.pr_output =		(pr_output_t*)rip_output,
230 	.pr_ctloutput =		rip_ctloutput,
231 	.pr_usrreqs =		&rip_usrreqs
232 };
233 static const struct encaptab *pim_encap_cookie;
234 
235 static int pim_encapcheck(const struct mbuf *, int, int, void *);
236 
237 /*
238  * Note: the PIM Register encapsulation adds the following in front of a
239  * data packet:
240  *
241  * struct pim_encap_hdr {
242  *    struct ip ip;
243  *    struct pim_encap_pimhdr  pim;
244  * }
245  *
246  */
247 
248 struct pim_encap_pimhdr {
249 	struct pim pim;
250 	uint32_t   flags;
251 };
252 #define		PIM_ENCAP_TTL	64
253 
254 static struct ip pim_encap_iphdr = {
255 #if BYTE_ORDER == LITTLE_ENDIAN
256 	sizeof(struct ip) >> 2,
257 	IPVERSION,
258 #else
259 	IPVERSION,
260 	sizeof(struct ip) >> 2,
261 #endif
262 	0,			/* tos */
263 	sizeof(struct ip),	/* total length */
264 	0,			/* id */
265 	0,			/* frag offset */
266 	PIM_ENCAP_TTL,
267 	IPPROTO_PIM,
268 	0,			/* checksum */
269 };
270 
271 static struct pim_encap_pimhdr pim_encap_pimhdr = {
272     {
273 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
274 	0,			/* reserved */
275 	0,			/* checksum */
276     },
277     0				/* flags */
278 };
279 
280 static struct ifnet multicast_register_if;
281 static vifi_t reg_vif_num = VIFI_INVALID;
282 
283 /*
284  * Private variables.
285  */
286 
287 static u_long	X_ip_mcast_src(int);
288 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
289 		    struct ip_moptions *);
290 static int	X_ip_mrouter_done(void);
291 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
292 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
293 static int	X_legal_vif_num(int);
294 static int	X_mrt_ioctl(u_long, caddr_t, int);
295 
296 static int	add_bw_upcall(struct bw_upcall *);
297 static int	add_mfc(struct mfcctl2 *);
298 static int	add_vif(struct vifctl *);
299 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
300 static void	bw_meter_process(void);
301 static void	bw_meter_receive_packet(struct bw_meter *, int,
302 		    struct timeval *);
303 static void	bw_upcalls_send(void);
304 static int	del_bw_upcall(struct bw_upcall *);
305 static int	del_mfc(struct mfcctl2 *);
306 static int	del_vif(vifi_t);
307 static int	del_vif_locked(vifi_t);
308 static void	expire_bw_meter_process(void *);
309 static void	expire_bw_upcalls_send(void *);
310 static void	expire_mfc(struct mfc *);
311 static void	expire_upcalls(void *);
312 static void	free_bw_list(struct bw_meter *);
313 static int	get_sg_cnt(struct sioc_sg_req *);
314 static int	get_vif_cnt(struct sioc_vif_req *);
315 static void	if_detached_event(void *, struct ifnet *);
316 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
317 static int	ip_mrouter_init(struct socket *, int);
318 static __inline struct mfc *
319 		mfc_find(struct in_addr *, struct in_addr *);
320 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
321 static struct mbuf *
322 		pim_register_prepare(struct ip *, struct mbuf *);
323 static int	pim_register_send(struct ip *, struct vif *,
324 		    struct mbuf *, struct mfc *);
325 static int	pim_register_send_rp(struct ip *, struct vif *,
326 		    struct mbuf *, struct mfc *);
327 static int	pim_register_send_upcall(struct ip *, struct vif *,
328 		    struct mbuf *, struct mfc *);
329 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
330 static void	send_packet(struct vif *, struct mbuf *);
331 static int	set_api_config(uint32_t *);
332 static int	set_assert(int);
333 static int	socket_send(struct socket *, struct mbuf *,
334 		    struct sockaddr_in *);
335 static void	unschedule_bw_meter(struct bw_meter *);
336 
337 /*
338  * Kernel multicast forwarding API capabilities and setup.
339  * If more API capabilities are added to the kernel, they should be
340  * recorded in `mrt_api_support'.
341  */
342 #define MRT_API_VERSION		0x0305
343 
344 static const int mrt_api_version = MRT_API_VERSION;
345 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
346 					 MRT_MFC_FLAGS_BORDER_VIF |
347 					 MRT_MFC_RP |
348 					 MRT_MFC_BW_UPCALL);
349 static uint32_t mrt_api_config = 0;
350 
351 static int pim_assert_enabled;
352 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
353 
354 /*
355  * Find a route for a given origin IP address and multicast group address.
356  * Statistics must be updated by the caller.
357  */
358 static __inline struct mfc *
359 mfc_find(struct in_addr *o, struct in_addr *g)
360 {
361 	struct mfc *rt;
362 
363 	MFC_LOCK_ASSERT();
364 
365 	LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
366 		if (in_hosteq(rt->mfc_origin, *o) &&
367 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
368 		    TAILQ_EMPTY(&rt->mfc_stall))
369 			break;
370 	}
371 
372 	return (rt);
373 }
374 
375 /*
376  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
377  */
378 static int
379 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
380 {
381     int	error, optval;
382     vifi_t	vifi;
383     struct	vifctl vifc;
384     struct	mfcctl2 mfc;
385     struct	bw_upcall bw_upcall;
386     uint32_t	i;
387 
388     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
389 	return EPERM;
390 
391     error = 0;
392     switch (sopt->sopt_name) {
393     case MRT_INIT:
394 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
395 	if (error)
396 	    break;
397 	error = ip_mrouter_init(so, optval);
398 	break;
399 
400     case MRT_DONE:
401 	error = ip_mrouter_done();
402 	break;
403 
404     case MRT_ADD_VIF:
405 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
406 	if (error)
407 	    break;
408 	error = add_vif(&vifc);
409 	break;
410 
411     case MRT_DEL_VIF:
412 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
413 	if (error)
414 	    break;
415 	error = del_vif(vifi);
416 	break;
417 
418     case MRT_ADD_MFC:
419     case MRT_DEL_MFC:
420 	/*
421 	 * select data size depending on API version.
422 	 */
423 	if (sopt->sopt_name == MRT_ADD_MFC &&
424 		mrt_api_config & MRT_API_FLAGS_ALL) {
425 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
426 				sizeof(struct mfcctl2));
427 	} else {
428 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
429 				sizeof(struct mfcctl));
430 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
431 			sizeof(mfc) - sizeof(struct mfcctl));
432 	}
433 	if (error)
434 	    break;
435 	if (sopt->sopt_name == MRT_ADD_MFC)
436 	    error = add_mfc(&mfc);
437 	else
438 	    error = del_mfc(&mfc);
439 	break;
440 
441     case MRT_ASSERT:
442 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
443 	if (error)
444 	    break;
445 	set_assert(optval);
446 	break;
447 
448     case MRT_API_CONFIG:
449 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
450 	if (!error)
451 	    error = set_api_config(&i);
452 	if (!error)
453 	    error = sooptcopyout(sopt, &i, sizeof i);
454 	break;
455 
456     case MRT_ADD_BW_UPCALL:
457     case MRT_DEL_BW_UPCALL:
458 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
459 				sizeof bw_upcall);
460 	if (error)
461 	    break;
462 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
463 	    error = add_bw_upcall(&bw_upcall);
464 	else
465 	    error = del_bw_upcall(&bw_upcall);
466 	break;
467 
468     default:
469 	error = EOPNOTSUPP;
470 	break;
471     }
472     return error;
473 }
474 
475 /*
476  * Handle MRT getsockopt commands
477  */
478 static int
479 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
480 {
481     int error;
482 
483     switch (sopt->sopt_name) {
484     case MRT_VERSION:
485 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
486 	break;
487 
488     case MRT_ASSERT:
489 	error = sooptcopyout(sopt, &pim_assert_enabled,
490 	    sizeof pim_assert_enabled);
491 	break;
492 
493     case MRT_API_SUPPORT:
494 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
495 	break;
496 
497     case MRT_API_CONFIG:
498 	error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config);
499 	break;
500 
501     default:
502 	error = EOPNOTSUPP;
503 	break;
504     }
505     return error;
506 }
507 
508 /*
509  * Handle ioctl commands to obtain information from the cache
510  */
511 static int
512 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
513 {
514     int error = 0;
515 
516     /*
517      * Currently the only function calling this ioctl routine is rtioctl().
518      * Typically, only root can create the raw socket in order to execute
519      * this ioctl method, however the request might be coming from a prison
520      */
521     error = priv_check(curthread, PRIV_NETINET_MROUTE);
522     if (error)
523 	return (error);
524     switch (cmd) {
525     case (SIOCGETVIFCNT):
526 	error = get_vif_cnt((struct sioc_vif_req *)data);
527 	break;
528 
529     case (SIOCGETSGCNT):
530 	error = get_sg_cnt((struct sioc_sg_req *)data);
531 	break;
532 
533     default:
534 	error = EINVAL;
535 	break;
536     }
537     return error;
538 }
539 
540 /*
541  * returns the packet, byte, rpf-failure count for the source group provided
542  */
543 static int
544 get_sg_cnt(struct sioc_sg_req *req)
545 {
546     struct mfc *rt;
547 
548     MFC_LOCK();
549     rt = mfc_find(&req->src, &req->grp);
550     if (rt == NULL) {
551 	MFC_UNLOCK();
552 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
553 	return EADDRNOTAVAIL;
554     }
555     req->pktcnt = rt->mfc_pkt_cnt;
556     req->bytecnt = rt->mfc_byte_cnt;
557     req->wrong_if = rt->mfc_wrong_if;
558     MFC_UNLOCK();
559     return 0;
560 }
561 
562 /*
563  * returns the input and output packet and byte counts on the vif provided
564  */
565 static int
566 get_vif_cnt(struct sioc_vif_req *req)
567 {
568     vifi_t vifi = req->vifi;
569 
570     VIF_LOCK();
571     if (vifi >= numvifs) {
572 	VIF_UNLOCK();
573 	return EINVAL;
574     }
575 
576     req->icount = viftable[vifi].v_pkt_in;
577     req->ocount = viftable[vifi].v_pkt_out;
578     req->ibytes = viftable[vifi].v_bytes_in;
579     req->obytes = viftable[vifi].v_bytes_out;
580     VIF_UNLOCK();
581 
582     return 0;
583 }
584 
585 static void
586 ip_mrouter_reset(void)
587 {
588 
589     pim_assert_enabled = 0;
590     mrt_api_config = 0;
591 
592     callout_init(&expire_upcalls_ch, CALLOUT_MPSAFE);
593 
594     bw_upcalls_n = 0;
595     bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers));
596     callout_init(&bw_upcalls_ch, CALLOUT_MPSAFE);
597     callout_init(&bw_meter_ch, CALLOUT_MPSAFE);
598 }
599 
600 static void
601 if_detached_event(void *arg __unused, struct ifnet *ifp)
602 {
603     vifi_t vifi;
604     int i;
605 
606     MROUTER_LOCK();
607 
608     if (V_ip_mrouter == NULL) {
609 	MROUTER_UNLOCK();
610 	return;
611     }
612 
613     VIF_LOCK();
614     MFC_LOCK();
615 
616     /*
617      * Tear down multicast forwarder state associated with this ifnet.
618      * 1. Walk the vif list, matching vifs against this ifnet.
619      * 2. Walk the multicast forwarding cache (mfc) looking for
620      *    inner matches with this vif's index.
621      * 3. Expire any matching multicast forwarding cache entries.
622      * 4. Free vif state. This should disable ALLMULTI on the interface.
623      */
624     for (vifi = 0; vifi < numvifs; vifi++) {
625 	if (viftable[vifi].v_ifp != ifp)
626 		continue;
627 	for (i = 0; i < mfchashsize; i++) {
628 		struct mfc *rt, *nrt;
629 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
630 			nrt = LIST_NEXT(rt, mfc_hash);
631 			if (rt->mfc_parent == vifi) {
632 				expire_mfc(rt);
633 			}
634 		}
635 	}
636 	del_vif_locked(vifi);
637     }
638 
639     MFC_UNLOCK();
640     VIF_UNLOCK();
641 
642     MROUTER_UNLOCK();
643 }
644 
645 /*
646  * Enable multicast forwarding.
647  */
648 static int
649 ip_mrouter_init(struct socket *so, int version)
650 {
651 
652     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
653         so->so_type, so->so_proto->pr_protocol);
654 
655     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
656 	return EOPNOTSUPP;
657 
658     if (version != 1)
659 	return ENOPROTOOPT;
660 
661     MROUTER_LOCK();
662 
663     if (V_ip_mrouter != NULL) {
664 	MROUTER_UNLOCK();
665 	return EADDRINUSE;
666     }
667 
668     if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
669         if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
670     if (if_detach_event_tag == NULL) {
671 	MROUTER_UNLOCK();
672 	return (ENOMEM);
673     }
674 
675     mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &mfchash, HASH_NOWAIT);
676 
677     callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
678 
679     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
680 	expire_bw_upcalls_send, NULL);
681     callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
682 
683     V_ip_mrouter = so;
684 
685     MROUTER_UNLOCK();
686 
687     CTR1(KTR_IPMF, "%s: done", __func__);
688 
689     return 0;
690 }
691 
692 /*
693  * Disable multicast forwarding.
694  */
695 static int
696 X_ip_mrouter_done(void)
697 {
698     vifi_t vifi;
699     int i;
700     struct ifnet *ifp;
701     struct ifreq ifr;
702 
703     MROUTER_LOCK();
704 
705     if (V_ip_mrouter == NULL) {
706 	MROUTER_UNLOCK();
707 	return EINVAL;
708     }
709 
710     /*
711      * Detach/disable hooks to the reset of the system.
712      */
713     V_ip_mrouter = NULL;
714     mrt_api_config = 0;
715 
716     VIF_LOCK();
717 
718     /*
719      * For each phyint in use, disable promiscuous reception of all IP
720      * multicasts.
721      */
722     for (vifi = 0; vifi < numvifs; vifi++) {
723 	if (!in_nullhost(viftable[vifi].v_lcl_addr) &&
724 		!(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
725 	    struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
726 
727 	    so->sin_len = sizeof(struct sockaddr_in);
728 	    so->sin_family = AF_INET;
729 	    so->sin_addr.s_addr = INADDR_ANY;
730 	    ifp = viftable[vifi].v_ifp;
731 	    if_allmulti(ifp, 0);
732 	}
733     }
734     bzero((caddr_t)viftable, sizeof(viftable));
735     numvifs = 0;
736     pim_assert_enabled = 0;
737 
738     VIF_UNLOCK();
739 
740     EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
741 
742     callout_stop(&expire_upcalls_ch);
743     callout_stop(&bw_upcalls_ch);
744     callout_stop(&bw_meter_ch);
745 
746     MFC_LOCK();
747 
748     /*
749      * Free all multicast forwarding cache entries.
750      * Do not use hashdestroy(), as we must perform other cleanup.
751      */
752     for (i = 0; i < mfchashsize; i++) {
753 	struct mfc *rt, *nrt;
754 	for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
755 		nrt = LIST_NEXT(rt, mfc_hash);
756 		expire_mfc(rt);
757 	}
758     }
759     free(mfchashtbl, M_MRTABLE);
760     mfchashtbl = NULL;
761 
762     bzero(nexpire, sizeof(nexpire[0]) * mfchashsize);
763 
764     bw_upcalls_n = 0;
765     bzero(bw_meter_timers, sizeof(bw_meter_timers));
766 
767     MFC_UNLOCK();
768 
769     reg_vif_num = VIFI_INVALID;
770 
771     MROUTER_UNLOCK();
772 
773     CTR1(KTR_IPMF, "%s: done", __func__);
774 
775     return 0;
776 }
777 
778 /*
779  * Set PIM assert processing global
780  */
781 static int
782 set_assert(int i)
783 {
784     if ((i != 1) && (i != 0))
785 	return EINVAL;
786 
787     pim_assert_enabled = i;
788 
789     return 0;
790 }
791 
792 /*
793  * Configure API capabilities
794  */
795 int
796 set_api_config(uint32_t *apival)
797 {
798     int i;
799 
800     /*
801      * We can set the API capabilities only if it is the first operation
802      * after MRT_INIT. I.e.:
803      *  - there are no vifs installed
804      *  - pim_assert is not enabled
805      *  - the MFC table is empty
806      */
807     if (numvifs > 0) {
808 	*apival = 0;
809 	return EPERM;
810     }
811     if (pim_assert_enabled) {
812 	*apival = 0;
813 	return EPERM;
814     }
815 
816     MFC_LOCK();
817 
818     for (i = 0; i < mfchashsize; i++) {
819 	if (LIST_FIRST(&mfchashtbl[i]) != NULL) {
820 	    *apival = 0;
821 	    return EPERM;
822 	}
823     }
824 
825     MFC_UNLOCK();
826 
827     mrt_api_config = *apival & mrt_api_support;
828     *apival = mrt_api_config;
829 
830     return 0;
831 }
832 
833 /*
834  * Add a vif to the vif table
835  */
836 static int
837 add_vif(struct vifctl *vifcp)
838 {
839     struct vif *vifp = viftable + vifcp->vifc_vifi;
840     struct sockaddr_in sin = {sizeof sin, AF_INET};
841     struct ifaddr *ifa;
842     struct ifnet *ifp;
843     int error;
844 
845     VIF_LOCK();
846     if (vifcp->vifc_vifi >= MAXVIFS) {
847 	VIF_UNLOCK();
848 	return EINVAL;
849     }
850     /* rate limiting is no longer supported by this code */
851     if (vifcp->vifc_rate_limit != 0) {
852 	log(LOG_ERR, "rate limiting is no longer supported\n");
853 	VIF_UNLOCK();
854 	return EINVAL;
855     }
856     if (!in_nullhost(vifp->v_lcl_addr)) {
857 	VIF_UNLOCK();
858 	return EADDRINUSE;
859     }
860     if (in_nullhost(vifcp->vifc_lcl_addr)) {
861 	VIF_UNLOCK();
862 	return EADDRNOTAVAIL;
863     }
864 
865     /* Find the interface with an address in AF_INET family */
866     if (vifcp->vifc_flags & VIFF_REGISTER) {
867 	/*
868 	 * XXX: Because VIFF_REGISTER does not really need a valid
869 	 * local interface (e.g. it could be 127.0.0.2), we don't
870 	 * check its address.
871 	 */
872 	ifp = NULL;
873     } else {
874 	sin.sin_addr = vifcp->vifc_lcl_addr;
875 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
876 	if (ifa == NULL) {
877 	    VIF_UNLOCK();
878 	    return EADDRNOTAVAIL;
879 	}
880 	ifp = ifa->ifa_ifp;
881 	ifa_free(ifa);
882     }
883 
884     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
885 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
886 	VIF_UNLOCK();
887 	return EOPNOTSUPP;
888     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
889 	ifp = &multicast_register_if;
890 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
891 	if (reg_vif_num == VIFI_INVALID) {
892 	    if_initname(&multicast_register_if, "register_vif", 0);
893 	    multicast_register_if.if_flags = IFF_LOOPBACK;
894 	    reg_vif_num = vifcp->vifc_vifi;
895 	}
896     } else {		/* Make sure the interface supports multicast */
897 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
898 	    VIF_UNLOCK();
899 	    return EOPNOTSUPP;
900 	}
901 
902 	/* Enable promiscuous reception of all IP multicasts from the if */
903 	error = if_allmulti(ifp, 1);
904 	if (error) {
905 	    VIF_UNLOCK();
906 	    return error;
907 	}
908     }
909 
910     vifp->v_flags     = vifcp->vifc_flags;
911     vifp->v_threshold = vifcp->vifc_threshold;
912     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
913     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
914     vifp->v_ifp       = ifp;
915     /* initialize per vif pkt counters */
916     vifp->v_pkt_in    = 0;
917     vifp->v_pkt_out   = 0;
918     vifp->v_bytes_in  = 0;
919     vifp->v_bytes_out = 0;
920     bzero(&vifp->v_route, sizeof(vifp->v_route));
921 
922     /* Adjust numvifs up if the vifi is higher than numvifs */
923     if (numvifs <= vifcp->vifc_vifi)
924 	numvifs = vifcp->vifc_vifi + 1;
925 
926     VIF_UNLOCK();
927 
928     CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
929 	(int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr),
930 	(int)vifcp->vifc_threshold);
931 
932     return 0;
933 }
934 
935 /*
936  * Delete a vif from the vif table
937  */
938 static int
939 del_vif_locked(vifi_t vifi)
940 {
941     struct vif *vifp;
942 
943     VIF_LOCK_ASSERT();
944 
945     if (vifi >= numvifs) {
946 	return EINVAL;
947     }
948     vifp = &viftable[vifi];
949     if (in_nullhost(vifp->v_lcl_addr)) {
950 	return EADDRNOTAVAIL;
951     }
952 
953     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
954 	if_allmulti(vifp->v_ifp, 0);
955 
956     if (vifp->v_flags & VIFF_REGISTER)
957 	reg_vif_num = VIFI_INVALID;
958 
959     bzero((caddr_t)vifp, sizeof (*vifp));
960 
961     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
962 
963     /* Adjust numvifs down */
964     for (vifi = numvifs; vifi > 0; vifi--)
965 	if (!in_nullhost(viftable[vifi-1].v_lcl_addr))
966 	    break;
967     numvifs = vifi;
968 
969     return 0;
970 }
971 
972 static int
973 del_vif(vifi_t vifi)
974 {
975     int cc;
976 
977     VIF_LOCK();
978     cc = del_vif_locked(vifi);
979     VIF_UNLOCK();
980 
981     return cc;
982 }
983 
984 /*
985  * update an mfc entry without resetting counters and S,G addresses.
986  */
987 static void
988 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
989 {
990     int i;
991 
992     rt->mfc_parent = mfccp->mfcc_parent;
993     for (i = 0; i < numvifs; i++) {
994 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
995 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
996 	    MRT_MFC_FLAGS_ALL;
997     }
998     /* set the RP address */
999     if (mrt_api_config & MRT_MFC_RP)
1000 	rt->mfc_rp = mfccp->mfcc_rp;
1001     else
1002 	rt->mfc_rp.s_addr = INADDR_ANY;
1003 }
1004 
1005 /*
1006  * fully initialize an mfc entry from the parameter.
1007  */
1008 static void
1009 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1010 {
1011     rt->mfc_origin     = mfccp->mfcc_origin;
1012     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1013 
1014     update_mfc_params(rt, mfccp);
1015 
1016     /* initialize pkt counters per src-grp */
1017     rt->mfc_pkt_cnt    = 0;
1018     rt->mfc_byte_cnt   = 0;
1019     rt->mfc_wrong_if   = 0;
1020     timevalclear(&rt->mfc_last_assert);
1021 }
1022 
1023 static void
1024 expire_mfc(struct mfc *rt)
1025 {
1026 	struct rtdetq *rte, *nrte;
1027 
1028 	free_bw_list(rt->mfc_bw_meter);
1029 
1030 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1031 		m_freem(rte->m);
1032 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1033 		free(rte, M_MRTABLE);
1034 	}
1035 
1036 	LIST_REMOVE(rt, mfc_hash);
1037 	free(rt, M_MRTABLE);
1038 }
1039 
1040 /*
1041  * Add an mfc entry
1042  */
1043 static int
1044 add_mfc(struct mfcctl2 *mfccp)
1045 {
1046     struct mfc *rt;
1047     struct rtdetq *rte, *nrte;
1048     u_long hash = 0;
1049     u_short nstl;
1050 
1051     VIF_LOCK();
1052     MFC_LOCK();
1053 
1054     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1055 
1056     /* If an entry already exists, just update the fields */
1057     if (rt) {
1058 	CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x",
1059 	    __func__, inet_ntoa(mfccp->mfcc_origin),
1060 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1061 	    mfccp->mfcc_parent);
1062 	update_mfc_params(rt, mfccp);
1063 	MFC_UNLOCK();
1064 	VIF_UNLOCK();
1065 	return (0);
1066     }
1067 
1068     /*
1069      * Find the entry for which the upcall was made and update
1070      */
1071     nstl = 0;
1072     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1073     LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1074 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1075 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1076 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1077 		CTR5(KTR_IPMF,
1078 		    "%s: add mfc orig %s group %lx parent %x qh %p",
1079 		    __func__, inet_ntoa(mfccp->mfcc_origin),
1080 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1081 		    mfccp->mfcc_parent,
1082 		    TAILQ_FIRST(&rt->mfc_stall));
1083 		if (nstl++)
1084 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1085 
1086 		init_mfc_params(rt, mfccp);
1087 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1088 		nexpire[hash]--;
1089 
1090 		/* Free queued packets, but attempt to forward them first. */
1091 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1092 			if (rte->ifp != NULL)
1093 				ip_mdq(rte->m, rte->ifp, rt, -1);
1094 			m_freem(rte->m);
1095 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1096 			rt->mfc_nstall--;
1097 			free(rte, M_MRTABLE);
1098 		}
1099 	}
1100     }
1101 
1102     /*
1103      * It is possible that an entry is being inserted without an upcall
1104      */
1105     if (nstl == 0) {
1106 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1107 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1108 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1109 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1110 			init_mfc_params(rt, mfccp);
1111 			if (rt->mfc_expire)
1112 			    nexpire[hash]--;
1113 			rt->mfc_expire = 0;
1114 			break; /* XXX */
1115 		}
1116 	}
1117 
1118 	if (rt == NULL) {		/* no upcall, so make a new entry */
1119 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1120 	    if (rt == NULL) {
1121 		MFC_UNLOCK();
1122 		VIF_UNLOCK();
1123 		return (ENOBUFS);
1124 	    }
1125 
1126 	    init_mfc_params(rt, mfccp);
1127 	    TAILQ_INIT(&rt->mfc_stall);
1128 	    rt->mfc_nstall = 0;
1129 
1130 	    rt->mfc_expire     = 0;
1131 	    rt->mfc_bw_meter = NULL;
1132 
1133 	    /* insert new entry at head of hash chain */
1134 	    LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1135 	}
1136     }
1137 
1138     MFC_UNLOCK();
1139     VIF_UNLOCK();
1140 
1141     return (0);
1142 }
1143 
1144 /*
1145  * Delete an mfc entry
1146  */
1147 static int
1148 del_mfc(struct mfcctl2 *mfccp)
1149 {
1150     struct in_addr	origin;
1151     struct in_addr	mcastgrp;
1152     struct mfc		*rt;
1153 
1154     origin = mfccp->mfcc_origin;
1155     mcastgrp = mfccp->mfcc_mcastgrp;
1156 
1157     CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__,
1158 	inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr));
1159 
1160     MFC_LOCK();
1161 
1162     rt = mfc_find(&origin, &mcastgrp);
1163     if (rt == NULL) {
1164 	MFC_UNLOCK();
1165 	return EADDRNOTAVAIL;
1166     }
1167 
1168     /*
1169      * free the bw_meter entries
1170      */
1171     free_bw_list(rt->mfc_bw_meter);
1172     rt->mfc_bw_meter = NULL;
1173 
1174     LIST_REMOVE(rt, mfc_hash);
1175     free(rt, M_MRTABLE);
1176 
1177     MFC_UNLOCK();
1178 
1179     return (0);
1180 }
1181 
1182 /*
1183  * Send a message to the routing daemon on the multicast routing socket.
1184  */
1185 static int
1186 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1187 {
1188     if (s) {
1189 	SOCKBUF_LOCK(&s->so_rcv);
1190 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1191 	    NULL) != 0) {
1192 	    sorwakeup_locked(s);
1193 	    return 0;
1194 	}
1195 	SOCKBUF_UNLOCK(&s->so_rcv);
1196     }
1197     m_freem(mm);
1198     return -1;
1199 }
1200 
1201 /*
1202  * IP multicast forwarding function. This function assumes that the packet
1203  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1204  * pointed to by "ifp", and the packet is to be relayed to other networks
1205  * that have members of the packet's destination IP multicast group.
1206  *
1207  * The packet is returned unscathed to the caller, unless it is
1208  * erroneous, in which case a non-zero return value tells the caller to
1209  * discard it.
1210  */
1211 
1212 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1213 
1214 static int
1215 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1216     struct ip_moptions *imo)
1217 {
1218     struct mfc *rt;
1219     int error;
1220     vifi_t vifi;
1221 
1222     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p",
1223 	inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1224 
1225     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1226 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1227 	/*
1228 	 * Packet arrived via a physical interface or
1229 	 * an encapsulated tunnel or a register_vif.
1230 	 */
1231     } else {
1232 	/*
1233 	 * Packet arrived through a source-route tunnel.
1234 	 * Source-route tunnels are no longer supported.
1235 	 */
1236 	return (1);
1237     }
1238 
1239     VIF_LOCK();
1240     MFC_LOCK();
1241     if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1242 	if (ip->ip_ttl < MAXTTL)
1243 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1244 	error = ip_mdq(m, ifp, NULL, vifi);
1245 	MFC_UNLOCK();
1246 	VIF_UNLOCK();
1247 	return error;
1248     }
1249 
1250     /*
1251      * Don't forward a packet with time-to-live of zero or one,
1252      * or a packet destined to a local-only group.
1253      */
1254     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1255 	MFC_UNLOCK();
1256 	VIF_UNLOCK();
1257 	return 0;
1258     }
1259 
1260     /*
1261      * Determine forwarding vifs from the forwarding cache table
1262      */
1263     MRTSTAT_INC(mrts_mfc_lookups);
1264     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1265 
1266     /* Entry exists, so forward if necessary */
1267     if (rt != NULL) {
1268 	error = ip_mdq(m, ifp, rt, -1);
1269 	MFC_UNLOCK();
1270 	VIF_UNLOCK();
1271 	return error;
1272     } else {
1273 	/*
1274 	 * If we don't have a route for packet's origin,
1275 	 * Make a copy of the packet & send message to routing daemon
1276 	 */
1277 
1278 	struct mbuf *mb0;
1279 	struct rtdetq *rte;
1280 	u_long hash;
1281 	int hlen = ip->ip_hl << 2;
1282 
1283 	MRTSTAT_INC(mrts_mfc_misses);
1284 	MRTSTAT_INC(mrts_no_route);
1285 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)",
1286 	    inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr));
1287 
1288 	/*
1289 	 * Allocate mbufs early so that we don't do extra work if we are
1290 	 * just going to fail anyway.  Make sure to pullup the header so
1291 	 * that other people can't step on it.
1292 	 */
1293 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1294 	    M_NOWAIT|M_ZERO);
1295 	if (rte == NULL) {
1296 	    MFC_UNLOCK();
1297 	    VIF_UNLOCK();
1298 	    return ENOBUFS;
1299 	}
1300 
1301 	mb0 = m_copypacket(m, M_DONTWAIT);
1302 	if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1303 	    mb0 = m_pullup(mb0, hlen);
1304 	if (mb0 == NULL) {
1305 	    free(rte, M_MRTABLE);
1306 	    MFC_UNLOCK();
1307 	    VIF_UNLOCK();
1308 	    return ENOBUFS;
1309 	}
1310 
1311 	/* is there an upcall waiting for this flow ? */
1312 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1313 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1314 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1315 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1316 		    !TAILQ_EMPTY(&rt->mfc_stall))
1317 			break;
1318 	}
1319 
1320 	if (rt == NULL) {
1321 	    int i;
1322 	    struct igmpmsg *im;
1323 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1324 	    struct mbuf *mm;
1325 
1326 	    /*
1327 	     * Locate the vifi for the incoming interface for this packet.
1328 	     * If none found, drop packet.
1329 	     */
1330 	    for (vifi = 0; vifi < numvifs &&
1331 		    viftable[vifi].v_ifp != ifp; vifi++)
1332 		;
1333 	    if (vifi >= numvifs)	/* vif not found, drop packet */
1334 		goto non_fatal;
1335 
1336 	    /* no upcall, so make a new entry */
1337 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1338 	    if (rt == NULL)
1339 		goto fail;
1340 
1341 	    /* Make a copy of the header to send to the user level process */
1342 	    mm = m_copy(mb0, 0, hlen);
1343 	    if (mm == NULL)
1344 		goto fail1;
1345 
1346 	    /*
1347 	     * Send message to routing daemon to install
1348 	     * a route into the kernel table
1349 	     */
1350 
1351 	    im = mtod(mm, struct igmpmsg *);
1352 	    im->im_msgtype = IGMPMSG_NOCACHE;
1353 	    im->im_mbz = 0;
1354 	    im->im_vif = vifi;
1355 
1356 	    MRTSTAT_INC(mrts_upcalls);
1357 
1358 	    k_igmpsrc.sin_addr = ip->ip_src;
1359 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1360 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1361 		MRTSTAT_INC(mrts_upq_sockfull);
1362 fail1:
1363 		free(rt, M_MRTABLE);
1364 fail:
1365 		free(rte, M_MRTABLE);
1366 		m_freem(mb0);
1367 		MFC_UNLOCK();
1368 		VIF_UNLOCK();
1369 		return ENOBUFS;
1370 	    }
1371 
1372 	    /* insert new entry at head of hash chain */
1373 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1374 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1375 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1376 	    nexpire[hash]++;
1377 	    for (i = 0; i < numvifs; i++) {
1378 		rt->mfc_ttls[i] = 0;
1379 		rt->mfc_flags[i] = 0;
1380 	    }
1381 	    rt->mfc_parent = -1;
1382 
1383 	    /* clear the RP address */
1384 	    rt->mfc_rp.s_addr = INADDR_ANY;
1385 	    rt->mfc_bw_meter = NULL;
1386 
1387 	    /* link into table */
1388 	    LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1389 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1390 	    rt->mfc_nstall++;
1391 
1392 	} else {
1393 	    /* determine if queue has overflowed */
1394 	    if (rt->mfc_nstall > MAX_UPQ) {
1395 		MRTSTAT_INC(mrts_upq_ovflw);
1396 non_fatal:
1397 		free(rte, M_MRTABLE);
1398 		m_freem(mb0);
1399 		MFC_UNLOCK();
1400 		VIF_UNLOCK();
1401 		return (0);
1402 	    }
1403 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1404 	    rt->mfc_nstall++;
1405 	}
1406 
1407 	rte->m			= mb0;
1408 	rte->ifp		= ifp;
1409 
1410 	MFC_UNLOCK();
1411 	VIF_UNLOCK();
1412 
1413 	return 0;
1414     }
1415 }
1416 
1417 /*
1418  * Clean up the cache entry if upcall is not serviced
1419  */
1420 static void
1421 expire_upcalls(void *unused)
1422 {
1423     int i;
1424 
1425     MFC_LOCK();
1426 
1427     for (i = 0; i < mfchashsize; i++) {
1428 	struct mfc *rt, *nrt;
1429 
1430 	if (nexpire[i] == 0)
1431 	    continue;
1432 
1433 	for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1434 		nrt = LIST_NEXT(rt, mfc_hash);
1435 
1436 		if (TAILQ_EMPTY(&rt->mfc_stall))
1437 			continue;
1438 
1439 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1440 			continue;
1441 
1442 		/*
1443 		 * free the bw_meter entries
1444 		 */
1445 		while (rt->mfc_bw_meter != NULL) {
1446 		    struct bw_meter *x = rt->mfc_bw_meter;
1447 
1448 		    rt->mfc_bw_meter = x->bm_mfc_next;
1449 		    free(x, M_BWMETER);
1450 		}
1451 
1452 		MRTSTAT_INC(mrts_cache_cleanups);
1453 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1454 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1455 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1456 
1457 		expire_mfc(rt);
1458 	    }
1459     }
1460 
1461     MFC_UNLOCK();
1462 
1463     callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
1464 }
1465 
1466 /*
1467  * Packet forwarding routine once entry in the cache is made
1468  */
1469 static int
1470 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1471 {
1472     struct ip  *ip = mtod(m, struct ip *);
1473     vifi_t vifi;
1474     int plen = ip->ip_len;
1475 
1476     VIF_LOCK_ASSERT();
1477 
1478     /*
1479      * If xmt_vif is not -1, send on only the requested vif.
1480      *
1481      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1482      */
1483     if (xmt_vif < numvifs) {
1484 	if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1485 		pim_register_send(ip, viftable + xmt_vif, m, rt);
1486 	else
1487 		phyint_send(ip, viftable + xmt_vif, m);
1488 	return 1;
1489     }
1490 
1491     /*
1492      * Don't forward if it didn't arrive from the parent vif for its origin.
1493      */
1494     vifi = rt->mfc_parent;
1495     if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1496 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1497 	    __func__, ifp, (int)vifi, viftable[vifi].v_ifp);
1498 	MRTSTAT_INC(mrts_wrong_if);
1499 	++rt->mfc_wrong_if;
1500 	/*
1501 	 * If we are doing PIM assert processing, send a message
1502 	 * to the routing daemon.
1503 	 *
1504 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1505 	 * can complete the SPT switch, regardless of the type
1506 	 * of the iif (broadcast media, GRE tunnel, etc).
1507 	 */
1508 	if (pim_assert_enabled && (vifi < numvifs) && viftable[vifi].v_ifp) {
1509 
1510 	    if (ifp == &multicast_register_if)
1511 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1512 
1513 	    /* Get vifi for the incoming packet */
1514 	    for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1515 		;
1516 	    if (vifi >= numvifs)
1517 		return 0;	/* The iif is not found: ignore the packet. */
1518 
1519 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1520 		return 0;	/* WRONGVIF disabled: ignore the packet */
1521 
1522 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1523 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1524 		struct igmpmsg *im;
1525 		int hlen = ip->ip_hl << 2;
1526 		struct mbuf *mm = m_copy(m, 0, hlen);
1527 
1528 		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1529 		    mm = m_pullup(mm, hlen);
1530 		if (mm == NULL)
1531 		    return ENOBUFS;
1532 
1533 		im = mtod(mm, struct igmpmsg *);
1534 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1535 		im->im_mbz		= 0;
1536 		im->im_vif		= vifi;
1537 
1538 		MRTSTAT_INC(mrts_upcalls);
1539 
1540 		k_igmpsrc.sin_addr = im->im_src;
1541 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1542 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1543 		    MRTSTAT_INC(mrts_upq_sockfull);
1544 		    return ENOBUFS;
1545 		}
1546 	    }
1547 	}
1548 	return 0;
1549     }
1550 
1551 
1552     /* If I sourced this packet, it counts as output, else it was input. */
1553     if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1554 	viftable[vifi].v_pkt_out++;
1555 	viftable[vifi].v_bytes_out += plen;
1556     } else {
1557 	viftable[vifi].v_pkt_in++;
1558 	viftable[vifi].v_bytes_in += plen;
1559     }
1560     rt->mfc_pkt_cnt++;
1561     rt->mfc_byte_cnt += plen;
1562 
1563     /*
1564      * For each vif, decide if a copy of the packet should be forwarded.
1565      * Forward if:
1566      *		- the ttl exceeds the vif's threshold
1567      *		- there are group members downstream on interface
1568      */
1569     for (vifi = 0; vifi < numvifs; vifi++)
1570 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1571 	    viftable[vifi].v_pkt_out++;
1572 	    viftable[vifi].v_bytes_out += plen;
1573 	    if (viftable[vifi].v_flags & VIFF_REGISTER)
1574 		pim_register_send(ip, viftable + vifi, m, rt);
1575 	    else
1576 		phyint_send(ip, viftable + vifi, m);
1577 	}
1578 
1579     /*
1580      * Perform upcall-related bw measuring.
1581      */
1582     if (rt->mfc_bw_meter != NULL) {
1583 	struct bw_meter *x;
1584 	struct timeval now;
1585 
1586 	microtime(&now);
1587 	MFC_LOCK_ASSERT();
1588 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1589 	    bw_meter_receive_packet(x, plen, &now);
1590     }
1591 
1592     return 0;
1593 }
1594 
1595 /*
1596  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1597  */
1598 static int
1599 X_legal_vif_num(int vif)
1600 {
1601 	int ret;
1602 
1603 	ret = 0;
1604 	if (vif < 0)
1605 		return (ret);
1606 
1607 	VIF_LOCK();
1608 	if (vif < numvifs)
1609 		ret = 1;
1610 	VIF_UNLOCK();
1611 
1612 	return (ret);
1613 }
1614 
1615 /*
1616  * Return the local address used by this vif
1617  */
1618 static u_long
1619 X_ip_mcast_src(int vifi)
1620 {
1621 	in_addr_t addr;
1622 
1623 	addr = INADDR_ANY;
1624 	if (vifi < 0)
1625 		return (addr);
1626 
1627 	VIF_LOCK();
1628 	if (vifi < numvifs)
1629 		addr = viftable[vifi].v_lcl_addr.s_addr;
1630 	VIF_UNLOCK();
1631 
1632 	return (addr);
1633 }
1634 
1635 static void
1636 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1637 {
1638     struct mbuf *mb_copy;
1639     int hlen = ip->ip_hl << 2;
1640 
1641     VIF_LOCK_ASSERT();
1642 
1643     /*
1644      * Make a new reference to the packet; make sure that
1645      * the IP header is actually copied, not just referenced,
1646      * so that ip_output() only scribbles on the copy.
1647      */
1648     mb_copy = m_copypacket(m, M_DONTWAIT);
1649     if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1650 	mb_copy = m_pullup(mb_copy, hlen);
1651     if (mb_copy == NULL)
1652 	return;
1653 
1654     send_packet(vifp, mb_copy);
1655 }
1656 
1657 static void
1658 send_packet(struct vif *vifp, struct mbuf *m)
1659 {
1660 	struct ip_moptions imo;
1661 	struct in_multi *imm[2];
1662 	int error;
1663 
1664 	VIF_LOCK_ASSERT();
1665 
1666 	imo.imo_multicast_ifp  = vifp->v_ifp;
1667 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1668 	imo.imo_multicast_loop = 1;
1669 	imo.imo_multicast_vif  = -1;
1670 	imo.imo_num_memberships = 0;
1671 	imo.imo_max_memberships = 2;
1672 	imo.imo_membership  = &imm[0];
1673 
1674 	/*
1675 	 * Re-entrancy should not be a problem here, because
1676 	 * the packets that we send out and are looped back at us
1677 	 * should get rejected because they appear to come from
1678 	 * the loopback interface, thus preventing looping.
1679 	 */
1680 	error = ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, &imo, NULL);
1681 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1682 	    (ptrdiff_t)(vifp - viftable), error);
1683 }
1684 
1685 /*
1686  * Stubs for old RSVP socket shim implementation.
1687  */
1688 
1689 static int
1690 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1691 {
1692 
1693 	return (EOPNOTSUPP);
1694 }
1695 
1696 static void
1697 X_ip_rsvp_force_done(struct socket *so __unused)
1698 {
1699 
1700 }
1701 
1702 static void
1703 X_rsvp_input(struct mbuf *m, int off __unused)
1704 {
1705 
1706 	if (!V_rsvp_on)
1707 		m_freem(m);
1708 }
1709 
1710 /*
1711  * Code for bandwidth monitors
1712  */
1713 
1714 /*
1715  * Define common interface for timeval-related methods
1716  */
1717 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1718 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1719 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1720 
1721 static uint32_t
1722 compute_bw_meter_flags(struct bw_upcall *req)
1723 {
1724     uint32_t flags = 0;
1725 
1726     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1727 	flags |= BW_METER_UNIT_PACKETS;
1728     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1729 	flags |= BW_METER_UNIT_BYTES;
1730     if (req->bu_flags & BW_UPCALL_GEQ)
1731 	flags |= BW_METER_GEQ;
1732     if (req->bu_flags & BW_UPCALL_LEQ)
1733 	flags |= BW_METER_LEQ;
1734 
1735     return flags;
1736 }
1737 
1738 /*
1739  * Add a bw_meter entry
1740  */
1741 static int
1742 add_bw_upcall(struct bw_upcall *req)
1743 {
1744     struct mfc *mfc;
1745     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1746 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1747     struct timeval now;
1748     struct bw_meter *x;
1749     uint32_t flags;
1750 
1751     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
1752 	return EOPNOTSUPP;
1753 
1754     /* Test if the flags are valid */
1755     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1756 	return EINVAL;
1757     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1758 	return EINVAL;
1759     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1760 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1761 	return EINVAL;
1762 
1763     /* Test if the threshold time interval is valid */
1764     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1765 	return EINVAL;
1766 
1767     flags = compute_bw_meter_flags(req);
1768 
1769     /*
1770      * Find if we have already same bw_meter entry
1771      */
1772     MFC_LOCK();
1773     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1774     if (mfc == NULL) {
1775 	MFC_UNLOCK();
1776 	return EADDRNOTAVAIL;
1777     }
1778     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1779 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1780 			   &req->bu_threshold.b_time, ==)) &&
1781 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1782 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1783 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1784 	    MFC_UNLOCK();
1785 	    return 0;		/* XXX Already installed */
1786 	}
1787     }
1788 
1789     /* Allocate the new bw_meter entry */
1790     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1791     if (x == NULL) {
1792 	MFC_UNLOCK();
1793 	return ENOBUFS;
1794     }
1795 
1796     /* Set the new bw_meter entry */
1797     x->bm_threshold.b_time = req->bu_threshold.b_time;
1798     microtime(&now);
1799     x->bm_start_time = now;
1800     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1801     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1802     x->bm_measured.b_packets = 0;
1803     x->bm_measured.b_bytes = 0;
1804     x->bm_flags = flags;
1805     x->bm_time_next = NULL;
1806     x->bm_time_hash = BW_METER_BUCKETS;
1807 
1808     /* Add the new bw_meter entry to the front of entries for this MFC */
1809     x->bm_mfc = mfc;
1810     x->bm_mfc_next = mfc->mfc_bw_meter;
1811     mfc->mfc_bw_meter = x;
1812     schedule_bw_meter(x, &now);
1813     MFC_UNLOCK();
1814 
1815     return 0;
1816 }
1817 
1818 static void
1819 free_bw_list(struct bw_meter *list)
1820 {
1821     while (list != NULL) {
1822 	struct bw_meter *x = list;
1823 
1824 	list = list->bm_mfc_next;
1825 	unschedule_bw_meter(x);
1826 	free(x, M_BWMETER);
1827     }
1828 }
1829 
1830 /*
1831  * Delete one or multiple bw_meter entries
1832  */
1833 static int
1834 del_bw_upcall(struct bw_upcall *req)
1835 {
1836     struct mfc *mfc;
1837     struct bw_meter *x;
1838 
1839     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
1840 	return EOPNOTSUPP;
1841 
1842     MFC_LOCK();
1843 
1844     /* Find the corresponding MFC entry */
1845     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1846     if (mfc == NULL) {
1847 	MFC_UNLOCK();
1848 	return EADDRNOTAVAIL;
1849     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1850 	/*
1851 	 * Delete all bw_meter entries for this mfc
1852 	 */
1853 	struct bw_meter *list;
1854 
1855 	list = mfc->mfc_bw_meter;
1856 	mfc->mfc_bw_meter = NULL;
1857 	free_bw_list(list);
1858 	MFC_UNLOCK();
1859 	return 0;
1860     } else {			/* Delete a single bw_meter entry */
1861 	struct bw_meter *prev;
1862 	uint32_t flags = 0;
1863 
1864 	flags = compute_bw_meter_flags(req);
1865 
1866 	/* Find the bw_meter entry to delete */
1867 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1868 	     prev = x, x = x->bm_mfc_next) {
1869 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1870 			       &req->bu_threshold.b_time, ==)) &&
1871 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1872 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1873 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1874 		break;
1875 	}
1876 	if (x != NULL) { /* Delete entry from the list for this MFC */
1877 	    if (prev != NULL)
1878 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1879 	    else
1880 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1881 
1882 	    unschedule_bw_meter(x);
1883 	    MFC_UNLOCK();
1884 	    /* Free the bw_meter entry */
1885 	    free(x, M_BWMETER);
1886 	    return 0;
1887 	} else {
1888 	    MFC_UNLOCK();
1889 	    return EINVAL;
1890 	}
1891     }
1892     /* NOTREACHED */
1893 }
1894 
1895 /*
1896  * Perform bandwidth measurement processing that may result in an upcall
1897  */
1898 static void
1899 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1900 {
1901     struct timeval delta;
1902 
1903     MFC_LOCK_ASSERT();
1904 
1905     delta = *nowp;
1906     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1907 
1908     if (x->bm_flags & BW_METER_GEQ) {
1909 	/*
1910 	 * Processing for ">=" type of bw_meter entry
1911 	 */
1912 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1913 	    /* Reset the bw_meter entry */
1914 	    x->bm_start_time = *nowp;
1915 	    x->bm_measured.b_packets = 0;
1916 	    x->bm_measured.b_bytes = 0;
1917 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1918 	}
1919 
1920 	/* Record that a packet is received */
1921 	x->bm_measured.b_packets++;
1922 	x->bm_measured.b_bytes += plen;
1923 
1924 	/*
1925 	 * Test if we should deliver an upcall
1926 	 */
1927 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1928 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1929 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1930 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1931 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1932 		/* Prepare an upcall for delivery */
1933 		bw_meter_prepare_upcall(x, nowp);
1934 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1935 	    }
1936 	}
1937     } else if (x->bm_flags & BW_METER_LEQ) {
1938 	/*
1939 	 * Processing for "<=" type of bw_meter entry
1940 	 */
1941 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1942 	    /*
1943 	     * We are behind time with the multicast forwarding table
1944 	     * scanning for "<=" type of bw_meter entries, so test now
1945 	     * if we should deliver an upcall.
1946 	     */
1947 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1948 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1949 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1950 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1951 		/* Prepare an upcall for delivery */
1952 		bw_meter_prepare_upcall(x, nowp);
1953 	    }
1954 	    /* Reschedule the bw_meter entry */
1955 	    unschedule_bw_meter(x);
1956 	    schedule_bw_meter(x, nowp);
1957 	}
1958 
1959 	/* Record that a packet is received */
1960 	x->bm_measured.b_packets++;
1961 	x->bm_measured.b_bytes += plen;
1962 
1963 	/*
1964 	 * Test if we should restart the measuring interval
1965 	 */
1966 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1967 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1968 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1969 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1970 	    /* Don't restart the measuring interval */
1971 	} else {
1972 	    /* Do restart the measuring interval */
1973 	    /*
1974 	     * XXX: note that we don't unschedule and schedule, because this
1975 	     * might be too much overhead per packet. Instead, when we process
1976 	     * all entries for a given timer hash bin, we check whether it is
1977 	     * really a timeout. If not, we reschedule at that time.
1978 	     */
1979 	    x->bm_start_time = *nowp;
1980 	    x->bm_measured.b_packets = 0;
1981 	    x->bm_measured.b_bytes = 0;
1982 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1983 	}
1984     }
1985 }
1986 
1987 /*
1988  * Prepare a bandwidth-related upcall
1989  */
1990 static void
1991 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
1992 {
1993     struct timeval delta;
1994     struct bw_upcall *u;
1995 
1996     MFC_LOCK_ASSERT();
1997 
1998     /*
1999      * Compute the measured time interval
2000      */
2001     delta = *nowp;
2002     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2003 
2004     /*
2005      * If there are too many pending upcalls, deliver them now
2006      */
2007     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2008 	bw_upcalls_send();
2009 
2010     /*
2011      * Set the bw_upcall entry
2012      */
2013     u = &bw_upcalls[bw_upcalls_n++];
2014     u->bu_src = x->bm_mfc->mfc_origin;
2015     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2016     u->bu_threshold.b_time = x->bm_threshold.b_time;
2017     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2018     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2019     u->bu_measured.b_time = delta;
2020     u->bu_measured.b_packets = x->bm_measured.b_packets;
2021     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2022     u->bu_flags = 0;
2023     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2024 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2025     if (x->bm_flags & BW_METER_UNIT_BYTES)
2026 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2027     if (x->bm_flags & BW_METER_GEQ)
2028 	u->bu_flags |= BW_UPCALL_GEQ;
2029     if (x->bm_flags & BW_METER_LEQ)
2030 	u->bu_flags |= BW_UPCALL_LEQ;
2031 }
2032 
2033 /*
2034  * Send the pending bandwidth-related upcalls
2035  */
2036 static void
2037 bw_upcalls_send(void)
2038 {
2039     struct mbuf *m;
2040     int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2041     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2042     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2043 				      0,		/* unused2 */
2044 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2045 				      0,		/* im_mbz  */
2046 				      0,		/* im_vif  */
2047 				      0,		/* unused3 */
2048 				      { 0 },		/* im_src  */
2049 				      { 0 } };		/* im_dst  */
2050 
2051     MFC_LOCK_ASSERT();
2052 
2053     if (bw_upcalls_n == 0)
2054 	return;			/* No pending upcalls */
2055 
2056     bw_upcalls_n = 0;
2057 
2058     /*
2059      * Allocate a new mbuf, initialize it with the header and
2060      * the payload for the pending calls.
2061      */
2062     MGETHDR(m, M_DONTWAIT, MT_DATA);
2063     if (m == NULL) {
2064 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2065 	return;
2066     }
2067 
2068     m->m_len = m->m_pkthdr.len = 0;
2069     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2070     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2071 
2072     /*
2073      * Send the upcalls
2074      * XXX do we need to set the address in k_igmpsrc ?
2075      */
2076     MRTSTAT_INC(mrts_upcalls);
2077     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2078 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2079 	MRTSTAT_INC(mrts_upq_sockfull);
2080     }
2081 }
2082 
2083 /*
2084  * Compute the timeout hash value for the bw_meter entries
2085  */
2086 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2087     do {								\
2088 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2089 									\
2090 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2091 	(hash) = next_timeval.tv_sec;					\
2092 	if (next_timeval.tv_usec)					\
2093 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2094 	(hash) %= BW_METER_BUCKETS;					\
2095     } while (0)
2096 
2097 /*
2098  * Schedule a timer to process periodically bw_meter entry of type "<="
2099  * by linking the entry in the proper hash bucket.
2100  */
2101 static void
2102 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2103 {
2104     int time_hash;
2105 
2106     MFC_LOCK_ASSERT();
2107 
2108     if (!(x->bm_flags & BW_METER_LEQ))
2109 	return;		/* XXX: we schedule timers only for "<=" entries */
2110 
2111     /*
2112      * Reset the bw_meter entry
2113      */
2114     x->bm_start_time = *nowp;
2115     x->bm_measured.b_packets = 0;
2116     x->bm_measured.b_bytes = 0;
2117     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2118 
2119     /*
2120      * Compute the timeout hash value and insert the entry
2121      */
2122     BW_METER_TIMEHASH(x, time_hash);
2123     x->bm_time_next = bw_meter_timers[time_hash];
2124     bw_meter_timers[time_hash] = x;
2125     x->bm_time_hash = time_hash;
2126 }
2127 
2128 /*
2129  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2130  * by removing the entry from the proper hash bucket.
2131  */
2132 static void
2133 unschedule_bw_meter(struct bw_meter *x)
2134 {
2135     int time_hash;
2136     struct bw_meter *prev, *tmp;
2137 
2138     MFC_LOCK_ASSERT();
2139 
2140     if (!(x->bm_flags & BW_METER_LEQ))
2141 	return;		/* XXX: we schedule timers only for "<=" entries */
2142 
2143     /*
2144      * Compute the timeout hash value and delete the entry
2145      */
2146     time_hash = x->bm_time_hash;
2147     if (time_hash >= BW_METER_BUCKETS)
2148 	return;		/* Entry was not scheduled */
2149 
2150     for (prev = NULL, tmp = bw_meter_timers[time_hash];
2151 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2152 	if (tmp == x)
2153 	    break;
2154 
2155     if (tmp == NULL)
2156 	panic("unschedule_bw_meter: bw_meter entry not found");
2157 
2158     if (prev != NULL)
2159 	prev->bm_time_next = x->bm_time_next;
2160     else
2161 	bw_meter_timers[time_hash] = x->bm_time_next;
2162 
2163     x->bm_time_next = NULL;
2164     x->bm_time_hash = BW_METER_BUCKETS;
2165 }
2166 
2167 
2168 /*
2169  * Process all "<=" type of bw_meter that should be processed now,
2170  * and for each entry prepare an upcall if necessary. Each processed
2171  * entry is rescheduled again for the (periodic) processing.
2172  *
2173  * This is run periodically (once per second normally). On each round,
2174  * all the potentially matching entries are in the hash slot that we are
2175  * looking at.
2176  */
2177 static void
2178 bw_meter_process()
2179 {
2180     static uint32_t last_tv_sec;	/* last time we processed this */
2181 
2182     uint32_t loops;
2183     int i;
2184     struct timeval now, process_endtime;
2185 
2186     microtime(&now);
2187     if (last_tv_sec == now.tv_sec)
2188 	return;		/* nothing to do */
2189 
2190     loops = now.tv_sec - last_tv_sec;
2191     last_tv_sec = now.tv_sec;
2192     if (loops > BW_METER_BUCKETS)
2193 	loops = BW_METER_BUCKETS;
2194 
2195     MFC_LOCK();
2196     /*
2197      * Process all bins of bw_meter entries from the one after the last
2198      * processed to the current one. On entry, i points to the last bucket
2199      * visited, so we need to increment i at the beginning of the loop.
2200      */
2201     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2202 	struct bw_meter *x, *tmp_list;
2203 
2204 	if (++i >= BW_METER_BUCKETS)
2205 	    i = 0;
2206 
2207 	/* Disconnect the list of bw_meter entries from the bin */
2208 	tmp_list = bw_meter_timers[i];
2209 	bw_meter_timers[i] = NULL;
2210 
2211 	/* Process the list of bw_meter entries */
2212 	while (tmp_list != NULL) {
2213 	    x = tmp_list;
2214 	    tmp_list = tmp_list->bm_time_next;
2215 
2216 	    /* Test if the time interval is over */
2217 	    process_endtime = x->bm_start_time;
2218 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2219 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2220 		/* Not yet: reschedule, but don't reset */
2221 		int time_hash;
2222 
2223 		BW_METER_TIMEHASH(x, time_hash);
2224 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2225 		    /*
2226 		     * XXX: somehow the bin processing is a bit ahead of time.
2227 		     * Put the entry in the next bin.
2228 		     */
2229 		    if (++time_hash >= BW_METER_BUCKETS)
2230 			time_hash = 0;
2231 		}
2232 		x->bm_time_next = bw_meter_timers[time_hash];
2233 		bw_meter_timers[time_hash] = x;
2234 		x->bm_time_hash = time_hash;
2235 
2236 		continue;
2237 	    }
2238 
2239 	    /*
2240 	     * Test if we should deliver an upcall
2241 	     */
2242 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2243 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2244 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2245 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2246 		/* Prepare an upcall for delivery */
2247 		bw_meter_prepare_upcall(x, &now);
2248 	    }
2249 
2250 	    /*
2251 	     * Reschedule for next processing
2252 	     */
2253 	    schedule_bw_meter(x, &now);
2254 	}
2255     }
2256 
2257     /* Send all upcalls that are pending delivery */
2258     bw_upcalls_send();
2259 
2260     MFC_UNLOCK();
2261 }
2262 
2263 /*
2264  * A periodic function for sending all upcalls that are pending delivery
2265  */
2266 static void
2267 expire_bw_upcalls_send(void *unused)
2268 {
2269     MFC_LOCK();
2270     bw_upcalls_send();
2271     MFC_UNLOCK();
2272 
2273     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2274 	expire_bw_upcalls_send, NULL);
2275 }
2276 
2277 /*
2278  * A periodic function for periodic scanning of the multicast forwarding
2279  * table for processing all "<=" bw_meter entries.
2280  */
2281 static void
2282 expire_bw_meter_process(void *unused)
2283 {
2284     if (mrt_api_config & MRT_MFC_BW_UPCALL)
2285 	bw_meter_process();
2286 
2287     callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
2288 }
2289 
2290 /*
2291  * End of bandwidth monitoring code
2292  */
2293 
2294 /*
2295  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2296  *
2297  */
2298 static int
2299 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2300     struct mfc *rt)
2301 {
2302     struct mbuf *mb_copy, *mm;
2303 
2304     /*
2305      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2306      * rendezvous point was unspecified, and we were told not to.
2307      */
2308     if (pim_squelch_wholepkt != 0 && (mrt_api_config & MRT_MFC_RP) &&
2309 	in_nullhost(rt->mfc_rp))
2310 	return 0;
2311 
2312     mb_copy = pim_register_prepare(ip, m);
2313     if (mb_copy == NULL)
2314 	return ENOBUFS;
2315 
2316     /*
2317      * Send all the fragments. Note that the mbuf for each fragment
2318      * is freed by the sending machinery.
2319      */
2320     for (mm = mb_copy; mm; mm = mb_copy) {
2321 	mb_copy = mm->m_nextpkt;
2322 	mm->m_nextpkt = 0;
2323 	mm = m_pullup(mm, sizeof(struct ip));
2324 	if (mm != NULL) {
2325 	    ip = mtod(mm, struct ip *);
2326 	    if ((mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2327 		pim_register_send_rp(ip, vifp, mm, rt);
2328 	    } else {
2329 		pim_register_send_upcall(ip, vifp, mm, rt);
2330 	    }
2331 	}
2332     }
2333 
2334     return 0;
2335 }
2336 
2337 /*
2338  * Return a copy of the data packet that is ready for PIM Register
2339  * encapsulation.
2340  * XXX: Note that in the returned copy the IP header is a valid one.
2341  */
2342 static struct mbuf *
2343 pim_register_prepare(struct ip *ip, struct mbuf *m)
2344 {
2345     struct mbuf *mb_copy = NULL;
2346     int mtu;
2347 
2348     /* Take care of delayed checksums */
2349     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2350 	in_delayed_cksum(m);
2351 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2352     }
2353 
2354     /*
2355      * Copy the old packet & pullup its IP header into the
2356      * new mbuf so we can modify it.
2357      */
2358     mb_copy = m_copypacket(m, M_DONTWAIT);
2359     if (mb_copy == NULL)
2360 	return NULL;
2361     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2362     if (mb_copy == NULL)
2363 	return NULL;
2364 
2365     /* take care of the TTL */
2366     ip = mtod(mb_copy, struct ip *);
2367     --ip->ip_ttl;
2368 
2369     /* Compute the MTU after the PIM Register encapsulation */
2370     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2371 
2372     if (ip->ip_len <= mtu) {
2373 	/* Turn the IP header into a valid one */
2374 	ip->ip_len = htons(ip->ip_len);
2375 	ip->ip_off = htons(ip->ip_off);
2376 	ip->ip_sum = 0;
2377 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2378     } else {
2379 	/* Fragment the packet */
2380 	if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
2381 	    m_freem(mb_copy);
2382 	    return NULL;
2383 	}
2384     }
2385     return mb_copy;
2386 }
2387 
2388 /*
2389  * Send an upcall with the data packet to the user-level process.
2390  */
2391 static int
2392 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2393     struct mbuf *mb_copy, struct mfc *rt)
2394 {
2395     struct mbuf *mb_first;
2396     int len = ntohs(ip->ip_len);
2397     struct igmpmsg *im;
2398     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2399 
2400     VIF_LOCK_ASSERT();
2401 
2402     /*
2403      * Add a new mbuf with an upcall header
2404      */
2405     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2406     if (mb_first == NULL) {
2407 	m_freem(mb_copy);
2408 	return ENOBUFS;
2409     }
2410     mb_first->m_data += max_linkhdr;
2411     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2412     mb_first->m_len = sizeof(struct igmpmsg);
2413     mb_first->m_next = mb_copy;
2414 
2415     /* Send message to routing daemon */
2416     im = mtod(mb_first, struct igmpmsg *);
2417     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2418     im->im_mbz		= 0;
2419     im->im_vif		= vifp - viftable;
2420     im->im_src		= ip->ip_src;
2421     im->im_dst		= ip->ip_dst;
2422 
2423     k_igmpsrc.sin_addr	= ip->ip_src;
2424 
2425     MRTSTAT_INC(mrts_upcalls);
2426 
2427     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2428 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2429 	MRTSTAT_INC(mrts_upq_sockfull);
2430 	return ENOBUFS;
2431     }
2432 
2433     /* Keep statistics */
2434     PIMSTAT_INC(pims_snd_registers_msgs);
2435     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2436 
2437     return 0;
2438 }
2439 
2440 /*
2441  * Encapsulate the data packet in PIM Register message and send it to the RP.
2442  */
2443 static int
2444 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2445     struct mfc *rt)
2446 {
2447     struct mbuf *mb_first;
2448     struct ip *ip_outer;
2449     struct pim_encap_pimhdr *pimhdr;
2450     int len = ntohs(ip->ip_len);
2451     vifi_t vifi = rt->mfc_parent;
2452 
2453     VIF_LOCK_ASSERT();
2454 
2455     if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
2456 	m_freem(mb_copy);
2457 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2458     }
2459 
2460     /*
2461      * Add a new mbuf with the encapsulating header
2462      */
2463     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2464     if (mb_first == NULL) {
2465 	m_freem(mb_copy);
2466 	return ENOBUFS;
2467     }
2468     mb_first->m_data += max_linkhdr;
2469     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2470     mb_first->m_next = mb_copy;
2471 
2472     mb_first->m_pkthdr.len = len + mb_first->m_len;
2473 
2474     /*
2475      * Fill in the encapsulating IP and PIM header
2476      */
2477     ip_outer = mtod(mb_first, struct ip *);
2478     *ip_outer = pim_encap_iphdr;
2479     ip_outer->ip_id = ip_newid();
2480     ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2481     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
2482     ip_outer->ip_dst = rt->mfc_rp;
2483     /*
2484      * Copy the inner header TOS to the outer header, and take care of the
2485      * IP_DF bit.
2486      */
2487     ip_outer->ip_tos = ip->ip_tos;
2488     if (ntohs(ip->ip_off) & IP_DF)
2489 	ip_outer->ip_off |= IP_DF;
2490     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2491 					 + sizeof(pim_encap_iphdr));
2492     *pimhdr = pim_encap_pimhdr;
2493     /* If the iif crosses a border, set the Border-bit */
2494     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
2495 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2496 
2497     mb_first->m_data += sizeof(pim_encap_iphdr);
2498     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2499     mb_first->m_data -= sizeof(pim_encap_iphdr);
2500 
2501     send_packet(vifp, mb_first);
2502 
2503     /* Keep statistics */
2504     PIMSTAT_INC(pims_snd_registers_msgs);
2505     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2506 
2507     return 0;
2508 }
2509 
2510 /*
2511  * pim_encapcheck() is called by the encap4_input() path at runtime to
2512  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2513  * into the kernel.
2514  */
2515 static int
2516 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2517 {
2518 
2519 #ifdef DIAGNOSTIC
2520     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2521 #endif
2522     if (proto != IPPROTO_PIM)
2523 	return 0;	/* not for us; reject the datagram. */
2524 
2525     return 64;		/* claim the datagram. */
2526 }
2527 
2528 /*
2529  * PIM-SMv2 and PIM-DM messages processing.
2530  * Receives and verifies the PIM control messages, and passes them
2531  * up to the listening socket, using rip_input().
2532  * The only message with special processing is the PIM_REGISTER message
2533  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2534  * is passed to if_simloop().
2535  */
2536 void
2537 pim_input(struct mbuf *m, int off)
2538 {
2539     struct ip *ip = mtod(m, struct ip *);
2540     struct pim *pim;
2541     int minlen;
2542     int datalen = ip->ip_len;
2543     int ip_tos;
2544     int iphlen = off;
2545 
2546     /* Keep statistics */
2547     PIMSTAT_INC(pims_rcv_total_msgs);
2548     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2549 
2550     /*
2551      * Validate lengths
2552      */
2553     if (datalen < PIM_MINLEN) {
2554 	PIMSTAT_INC(pims_rcv_tooshort);
2555 	CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2556 	    __func__, datalen, inet_ntoa(ip->ip_src));
2557 	m_freem(m);
2558 	return;
2559     }
2560 
2561     /*
2562      * If the packet is at least as big as a REGISTER, go agead
2563      * and grab the PIM REGISTER header size, to avoid another
2564      * possible m_pullup() later.
2565      *
2566      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2567      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2568      */
2569     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2570     /*
2571      * Get the IP and PIM headers in contiguous memory, and
2572      * possibly the PIM REGISTER header.
2573      */
2574     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
2575 	(m = m_pullup(m, minlen)) == 0) {
2576 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2577 	return;
2578     }
2579 
2580     /* m_pullup() may have given us a new mbuf so reset ip. */
2581     ip = mtod(m, struct ip *);
2582     ip_tos = ip->ip_tos;
2583 
2584     /* adjust mbuf to point to the PIM header */
2585     m->m_data += iphlen;
2586     m->m_len  -= iphlen;
2587     pim = mtod(m, struct pim *);
2588 
2589     /*
2590      * Validate checksum. If PIM REGISTER, exclude the data packet.
2591      *
2592      * XXX: some older PIMv2 implementations don't make this distinction,
2593      * so for compatibility reason perform the checksum over part of the
2594      * message, and if error, then over the whole message.
2595      */
2596     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2597 	/* do nothing, checksum okay */
2598     } else if (in_cksum(m, datalen)) {
2599 	PIMSTAT_INC(pims_rcv_badsum);
2600 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2601 	m_freem(m);
2602 	return;
2603     }
2604 
2605     /* PIM version check */
2606     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2607 	PIMSTAT_INC(pims_rcv_badversion);
2608 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2609 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2610 	m_freem(m);
2611 	return;
2612     }
2613 
2614     /* restore mbuf back to the outer IP */
2615     m->m_data -= iphlen;
2616     m->m_len  += iphlen;
2617 
2618     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2619 	/*
2620 	 * Since this is a REGISTER, we'll make a copy of the register
2621 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2622 	 * routing daemon.
2623 	 */
2624 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2625 	struct mbuf *mcp;
2626 	struct ip *encap_ip;
2627 	u_int32_t *reghdr;
2628 	struct ifnet *vifp;
2629 
2630 	VIF_LOCK();
2631 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
2632 	    VIF_UNLOCK();
2633 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2634 		(int)reg_vif_num);
2635 	    m_freem(m);
2636 	    return;
2637 	}
2638 	/* XXX need refcnt? */
2639 	vifp = viftable[reg_vif_num].v_ifp;
2640 	VIF_UNLOCK();
2641 
2642 	/*
2643 	 * Validate length
2644 	 */
2645 	if (datalen < PIM_REG_MINLEN) {
2646 	    PIMSTAT_INC(pims_rcv_tooshort);
2647 	    PIMSTAT_INC(pims_rcv_badregisters);
2648 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2649 	    m_freem(m);
2650 	    return;
2651 	}
2652 
2653 	reghdr = (u_int32_t *)(pim + 1);
2654 	encap_ip = (struct ip *)(reghdr + 1);
2655 
2656 	CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2657 	    __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2658 
2659 	/* verify the version number of the inner packet */
2660 	if (encap_ip->ip_v != IPVERSION) {
2661 	    PIMSTAT_INC(pims_rcv_badregisters);
2662 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2663 	    m_freem(m);
2664 	    return;
2665 	}
2666 
2667 	/* verify the inner packet is destined to a mcast group */
2668 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2669 	    PIMSTAT_INC(pims_rcv_badregisters);
2670 	    CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2671 		inet_ntoa(encap_ip->ip_dst));
2672 	    m_freem(m);
2673 	    return;
2674 	}
2675 
2676 	/* If a NULL_REGISTER, pass it to the daemon */
2677 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2678 	    goto pim_input_to_daemon;
2679 
2680 	/*
2681 	 * Copy the TOS from the outer IP header to the inner IP header.
2682 	 */
2683 	if (encap_ip->ip_tos != ip_tos) {
2684 	    /* Outer TOS -> inner TOS */
2685 	    encap_ip->ip_tos = ip_tos;
2686 	    /* Recompute the inner header checksum. Sigh... */
2687 
2688 	    /* adjust mbuf to point to the inner IP header */
2689 	    m->m_data += (iphlen + PIM_MINLEN);
2690 	    m->m_len  -= (iphlen + PIM_MINLEN);
2691 
2692 	    encap_ip->ip_sum = 0;
2693 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2694 
2695 	    /* restore mbuf to point back to the outer IP header */
2696 	    m->m_data -= (iphlen + PIM_MINLEN);
2697 	    m->m_len  += (iphlen + PIM_MINLEN);
2698 	}
2699 
2700 	/*
2701 	 * Decapsulate the inner IP packet and loopback to forward it
2702 	 * as a normal multicast packet. Also, make a copy of the
2703 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2704 	 * to pass to the daemon later, so it can take the appropriate
2705 	 * actions (e.g., send back PIM_REGISTER_STOP).
2706 	 * XXX: here m->m_data points to the outer IP header.
2707 	 */
2708 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2709 	if (mcp == NULL) {
2710 	    CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2711 	    m_freem(m);
2712 	    return;
2713 	}
2714 
2715 	/* Keep statistics */
2716 	/* XXX: registers_bytes include only the encap. mcast pkt */
2717 	PIMSTAT_INC(pims_rcv_registers_msgs);
2718 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2719 
2720 	/*
2721 	 * forward the inner ip packet; point m_data at the inner ip.
2722 	 */
2723 	m_adj(m, iphlen + PIM_MINLEN);
2724 
2725 	CTR4(KTR_IPMF,
2726 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2727 	    __func__,
2728 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2729 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2730 	    (int)reg_vif_num);
2731 
2732 	/* NB: vifp was collected above; can it change on us? */
2733 	if_simloop(vifp, m, dst.sin_family, 0);
2734 
2735 	/* prepare the register head to send to the mrouting daemon */
2736 	m = mcp;
2737     }
2738 
2739 pim_input_to_daemon:
2740     /*
2741      * Pass the PIM message up to the daemon; if it is a Register message,
2742      * pass the 'head' only up to the daemon. This includes the
2743      * outer IP header, PIM header, PIM-Register header and the
2744      * inner IP header.
2745      * XXX: the outer IP header pkt size of a Register is not adjust to
2746      * reflect the fact that the inner multicast data is truncated.
2747      */
2748     rip_input(m, iphlen);
2749 
2750     return;
2751 }
2752 
2753 static int
2754 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2755 {
2756 	struct mfc	*rt;
2757 	int		 error, i;
2758 
2759 	if (req->newptr)
2760 		return (EPERM);
2761 	if (mfchashtbl == NULL)	/* XXX unlocked */
2762 		return (0);
2763 	error = sysctl_wire_old_buffer(req, 0);
2764 	if (error)
2765 		return (error);
2766 
2767 	MFC_LOCK();
2768 	for (i = 0; i < mfchashsize; i++) {
2769 		LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
2770 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2771 			if (error)
2772 				goto out_locked;
2773 		}
2774 	}
2775 out_locked:
2776 	MFC_UNLOCK();
2777 	return (error);
2778 }
2779 
2780 SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, sysctl_mfctable,
2781     "IPv4 Multicast Forwarding Table (struct *mfc[mfchashsize], "
2782     "netinet/ip_mroute.h)");
2783 
2784 static int
2785 ip_mroute_modevent(module_t mod, int type, void *unused)
2786 {
2787 
2788     switch (type) {
2789     case MOD_LOAD:
2790 	MROUTER_LOCK_INIT();
2791 	MFC_LOCK_INIT();
2792 	VIF_LOCK_INIT();
2793 
2794 	mfchashsize = MFCHASHSIZE;
2795 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2796 	    !powerof2(mfchashsize)) {
2797 		printf("WARNING: %s not a power of 2; using default\n",
2798 		    "net.inet.ip.mfchashsize");
2799 		mfchashsize = MFCHASHSIZE;
2800 	}
2801 	MALLOC(nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2802 
2803 	pim_squelch_wholepkt = 0;
2804 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2805 	    &pim_squelch_wholepkt);
2806 	ip_mrouter_reset();
2807 
2808 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2809 	    pim_encapcheck, &in_pim_protosw, NULL);
2810 	if (pim_encap_cookie == NULL) {
2811 		printf("ip_mroute: unable to attach pim encap\n");
2812 		VIF_LOCK_DESTROY();
2813 		MFC_LOCK_DESTROY();
2814 		MROUTER_LOCK_DESTROY();
2815 		return (EINVAL);
2816 	}
2817 
2818 	ip_mcast_src = X_ip_mcast_src;
2819 	ip_mforward = X_ip_mforward;
2820 	ip_mrouter_done = X_ip_mrouter_done;
2821 	ip_mrouter_get = X_ip_mrouter_get;
2822 	ip_mrouter_set = X_ip_mrouter_set;
2823 
2824 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2825 	ip_rsvp_vif = X_ip_rsvp_vif;
2826 
2827 	legal_vif_num = X_legal_vif_num;
2828 	mrt_ioctl = X_mrt_ioctl;
2829 	rsvp_input_p = X_rsvp_input;
2830 	break;
2831 
2832     case MOD_UNLOAD:
2833 	/*
2834 	 * Typically module unload happens after the user-level
2835 	 * process has shutdown the kernel services (the check
2836 	 * below insures someone can't just yank the module out
2837 	 * from under a running process).  But if the module is
2838 	 * just loaded and then unloaded w/o starting up a user
2839 	 * process we still need to cleanup.
2840 	 */
2841 	if (V_ip_mrouter != NULL)
2842 	    return (EINVAL);
2843 
2844 	if (pim_encap_cookie) {
2845 	    encap_detach(pim_encap_cookie);
2846 	    pim_encap_cookie = NULL;
2847 	}
2848 	X_ip_mrouter_done();
2849 
2850 	FREE(nexpire, M_MRTABLE);
2851 	nexpire = NULL;
2852 
2853 	ip_mcast_src = NULL;
2854 	ip_mforward = NULL;
2855 	ip_mrouter_done = NULL;
2856 	ip_mrouter_get = NULL;
2857 	ip_mrouter_set = NULL;
2858 
2859 	ip_rsvp_force_done = NULL;
2860 	ip_rsvp_vif = NULL;
2861 
2862 	legal_vif_num = NULL;
2863 	mrt_ioctl = NULL;
2864 	rsvp_input_p = NULL;
2865 
2866 	VIF_LOCK_DESTROY();
2867 	MFC_LOCK_DESTROY();
2868 	MROUTER_LOCK_DESTROY();
2869 	break;
2870 
2871     default:
2872 	return EOPNOTSUPP;
2873     }
2874     return 0;
2875 }
2876 
2877 static moduledata_t ip_mroutemod = {
2878     "ip_mroute",
2879     ip_mroute_modevent,
2880     0
2881 };
2882 
2883 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
2884