xref: /freebsd/sys/netinet/ip_mroute.c (revision ab40f58ccfe6c07ebefddc72f4661a52fe746353)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989 Stephen Deering
5  * Copyright (c) 1992, 1993
6  *      The Regents of the University of California.  All rights reserved.
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Stephen Deering of Stanford University.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
36  */
37 
38 /*
39  * IP multicast forwarding procedures
40  *
41  * Written by David Waitzman, BBN Labs, August 1988.
42  * Modified by Steve Deering, Stanford, February 1989.
43  * Modified by Mark J. Steiglitz, Stanford, May, 1991
44  * Modified by Van Jacobson, LBL, January 1993
45  * Modified by Ajit Thyagarajan, PARC, August 1993
46  * Modified by Bill Fenner, PARC, April 1995
47  * Modified by Ahmed Helmy, SGI, June 1996
48  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
49  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
50  * Modified by Hitoshi Asaeda, WIDE, August 2000
51  * Modified by Pavlin Radoslavov, ICSI, October 2002
52  *
53  * MROUTING Revision: 3.5
54  * and PIM-SMv2 and PIM-DM support, advanced API support,
55  * bandwidth metering and signaling
56  */
57 
58 /*
59  * TODO: Prefix functions with ipmf_.
60  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
61  * domain attachment (if_afdata) so we can track consumers of that service.
62  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
63  * move it to socket options.
64  * TODO: Cleanup LSRR removal further.
65  * TODO: Push RSVP stubs into raw_ip.c.
66  * TODO: Use bitstring.h for vif set.
67  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
68  * TODO: Sync ip6_mroute.c with this file.
69  */
70 
71 #include <sys/cdefs.h>
72 __FBSDID("$FreeBSD$");
73 
74 #include "opt_inet.h"
75 #include "opt_mrouting.h"
76 
77 #define _PIM_VT 1
78 
79 #include <sys/param.h>
80 #include <sys/kernel.h>
81 #include <sys/stddef.h>
82 #include <sys/eventhandler.h>
83 #include <sys/lock.h>
84 #include <sys/ktr.h>
85 #include <sys/malloc.h>
86 #include <sys/mbuf.h>
87 #include <sys/module.h>
88 #include <sys/priv.h>
89 #include <sys/protosw.h>
90 #include <sys/signalvar.h>
91 #include <sys/socket.h>
92 #include <sys/socketvar.h>
93 #include <sys/sockio.h>
94 #include <sys/sx.h>
95 #include <sys/sysctl.h>
96 #include <sys/syslog.h>
97 #include <sys/systm.h>
98 #include <sys/time.h>
99 #include <sys/counter.h>
100 
101 #include <net/if.h>
102 #include <net/if_var.h>
103 #include <net/netisr.h>
104 #include <net/route.h>
105 #include <net/vnet.h>
106 
107 #include <netinet/in.h>
108 #include <netinet/igmp.h>
109 #include <netinet/in_systm.h>
110 #include <netinet/in_var.h>
111 #include <netinet/ip.h>
112 #include <netinet/ip_encap.h>
113 #include <netinet/ip_mroute.h>
114 #include <netinet/ip_var.h>
115 #include <netinet/ip_options.h>
116 #include <netinet/pim.h>
117 #include <netinet/pim_var.h>
118 #include <netinet/udp.h>
119 
120 #include <machine/in_cksum.h>
121 
122 #ifndef KTR_IPMF
123 #define KTR_IPMF KTR_INET
124 #endif
125 
126 #define		VIFI_INVALID	((vifi_t) -1)
127 
128 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
129 #define	V_last_tv_sec	VNET(last_tv_sec)
130 
131 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
132 
133 /*
134  * Locking.  We use two locks: one for the virtual interface table and
135  * one for the forwarding table.  These locks may be nested in which case
136  * the VIF lock must always be taken first.  Note that each lock is used
137  * to cover not only the specific data structure but also related data
138  * structures.
139  */
140 
141 static struct mtx mrouter_mtx;
142 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
143 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
144 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
145 #define	MROUTER_LOCK_INIT()						\
146 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
147 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
148 
149 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
150 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
151 
152 static VNET_PCPUSTAT_DEFINE(struct mrtstat, mrtstat);
153 VNET_PCPUSTAT_SYSINIT(mrtstat);
154 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
155 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
156     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
157     "netinet/ip_mroute.h)");
158 
159 static VNET_DEFINE(u_long, mfchash);
160 #define	V_mfchash		VNET(mfchash)
161 #define	MFCHASH(a, g)							\
162 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
163 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
164 #define	MFCHASHSIZE	256
165 
166 static u_long mfchashsize;			/* Hash size */
167 static VNET_DEFINE(u_char *, nexpire);		/* 0..mfchashsize-1 */
168 #define	V_nexpire		VNET(nexpire)
169 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
170 #define	V_mfchashtbl		VNET(mfchashtbl)
171 
172 static struct mtx mfc_mtx;
173 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
174 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
175 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
176 #define	MFC_LOCK_INIT()							\
177 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
178 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
179 
180 static VNET_DEFINE(vifi_t, numvifs);
181 #define	V_numvifs		VNET(numvifs)
182 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
183 #define	V_viftable		VNET(viftable)
184 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_VNET | CTLFLAG_RD,
185     &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
186     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
187 
188 static struct mtx vif_mtx;
189 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
190 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
191 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
192 #define	VIF_LOCK_INIT()							\
193 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
194 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
195 
196 static eventhandler_tag if_detach_event_tag = NULL;
197 
198 static VNET_DEFINE(struct callout, expire_upcalls_ch);
199 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
200 
201 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
202 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
203 
204 /*
205  * Bandwidth meter variables and constants
206  */
207 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
208 /*
209  * Pending timeouts are stored in a hash table, the key being the
210  * expiration time. Periodically, the entries are analysed and processed.
211  */
212 #define	BW_METER_BUCKETS	1024
213 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
214 #define	V_bw_meter_timers	VNET(bw_meter_timers)
215 static VNET_DEFINE(struct callout, bw_meter_ch);
216 #define	V_bw_meter_ch		VNET(bw_meter_ch)
217 #define	BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
218 
219 /*
220  * Pending upcalls are stored in a vector which is flushed when
221  * full, or periodically
222  */
223 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
224 #define	V_bw_upcalls		VNET(bw_upcalls)
225 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
226 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
227 static VNET_DEFINE(struct callout, bw_upcalls_ch);
228 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
229 
230 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
231 
232 static VNET_PCPUSTAT_DEFINE(struct pimstat, pimstat);
233 VNET_PCPUSTAT_SYSINIT(pimstat);
234 VNET_PCPUSTAT_SYSUNINIT(pimstat);
235 
236 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
237 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
238     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
239 
240 static u_long	pim_squelch_wholepkt = 0;
241 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
242     &pim_squelch_wholepkt, 0,
243     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
244 
245 static const struct encaptab *pim_encap_cookie;
246 static int pim_encapcheck(const struct mbuf *, int, int, void *);
247 static int pim_input(struct mbuf *, int, int, void *);
248 
249 static const struct encap_config ipv4_encap_cfg = {
250 	.proto = IPPROTO_PIM,
251 	.min_length = sizeof(struct ip) + PIM_MINLEN,
252 	.exact_match = 8,
253 	.check = pim_encapcheck,
254 	.input = pim_input
255 };
256 
257 /*
258  * Note: the PIM Register encapsulation adds the following in front of a
259  * data packet:
260  *
261  * struct pim_encap_hdr {
262  *    struct ip ip;
263  *    struct pim_encap_pimhdr  pim;
264  * }
265  *
266  */
267 
268 struct pim_encap_pimhdr {
269 	struct pim pim;
270 	uint32_t   flags;
271 };
272 #define		PIM_ENCAP_TTL	64
273 
274 static struct ip pim_encap_iphdr = {
275 #if BYTE_ORDER == LITTLE_ENDIAN
276 	sizeof(struct ip) >> 2,
277 	IPVERSION,
278 #else
279 	IPVERSION,
280 	sizeof(struct ip) >> 2,
281 #endif
282 	0,			/* tos */
283 	sizeof(struct ip),	/* total length */
284 	0,			/* id */
285 	0,			/* frag offset */
286 	PIM_ENCAP_TTL,
287 	IPPROTO_PIM,
288 	0,			/* checksum */
289 };
290 
291 static struct pim_encap_pimhdr pim_encap_pimhdr = {
292     {
293 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
294 	0,			/* reserved */
295 	0,			/* checksum */
296     },
297     0				/* flags */
298 };
299 
300 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
301 #define	V_reg_vif_num		VNET(reg_vif_num)
302 static VNET_DEFINE(struct ifnet, multicast_register_if);
303 #define	V_multicast_register_if	VNET(multicast_register_if)
304 
305 /*
306  * Private variables.
307  */
308 
309 static u_long	X_ip_mcast_src(int);
310 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
311 		    struct ip_moptions *);
312 static int	X_ip_mrouter_done(void);
313 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
314 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
315 static int	X_legal_vif_num(int);
316 static int	X_mrt_ioctl(u_long, caddr_t, int);
317 
318 static int	add_bw_upcall(struct bw_upcall *);
319 static int	add_mfc(struct mfcctl2 *);
320 static int	add_vif(struct vifctl *);
321 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
322 static void	bw_meter_process(void);
323 static void	bw_meter_receive_packet(struct bw_meter *, int,
324 		    struct timeval *);
325 static void	bw_upcalls_send(void);
326 static int	del_bw_upcall(struct bw_upcall *);
327 static int	del_mfc(struct mfcctl2 *);
328 static int	del_vif(vifi_t);
329 static int	del_vif_locked(vifi_t);
330 static void	expire_bw_meter_process(void *);
331 static void	expire_bw_upcalls_send(void *);
332 static void	expire_mfc(struct mfc *);
333 static void	expire_upcalls(void *);
334 static void	free_bw_list(struct bw_meter *);
335 static int	get_sg_cnt(struct sioc_sg_req *);
336 static int	get_vif_cnt(struct sioc_vif_req *);
337 static void	if_detached_event(void *, struct ifnet *);
338 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
339 static int	ip_mrouter_init(struct socket *, int);
340 static __inline struct mfc *
341 		mfc_find(struct in_addr *, struct in_addr *);
342 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
343 static struct mbuf *
344 		pim_register_prepare(struct ip *, struct mbuf *);
345 static int	pim_register_send(struct ip *, struct vif *,
346 		    struct mbuf *, struct mfc *);
347 static int	pim_register_send_rp(struct ip *, struct vif *,
348 		    struct mbuf *, struct mfc *);
349 static int	pim_register_send_upcall(struct ip *, struct vif *,
350 		    struct mbuf *, struct mfc *);
351 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
352 static void	send_packet(struct vif *, struct mbuf *);
353 static int	set_api_config(uint32_t *);
354 static int	set_assert(int);
355 static int	socket_send(struct socket *, struct mbuf *,
356 		    struct sockaddr_in *);
357 static void	unschedule_bw_meter(struct bw_meter *);
358 
359 /*
360  * Kernel multicast forwarding API capabilities and setup.
361  * If more API capabilities are added to the kernel, they should be
362  * recorded in `mrt_api_support'.
363  */
364 #define MRT_API_VERSION		0x0305
365 
366 static const int mrt_api_version = MRT_API_VERSION;
367 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
368 					 MRT_MFC_FLAGS_BORDER_VIF |
369 					 MRT_MFC_RP |
370 					 MRT_MFC_BW_UPCALL);
371 static VNET_DEFINE(uint32_t, mrt_api_config);
372 #define	V_mrt_api_config	VNET(mrt_api_config)
373 static VNET_DEFINE(int, pim_assert_enabled);
374 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
375 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
376 
377 /*
378  * Find a route for a given origin IP address and multicast group address.
379  * Statistics must be updated by the caller.
380  */
381 static __inline struct mfc *
382 mfc_find(struct in_addr *o, struct in_addr *g)
383 {
384 	struct mfc *rt;
385 
386 	MFC_LOCK_ASSERT();
387 
388 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
389 		if (in_hosteq(rt->mfc_origin, *o) &&
390 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
391 		    TAILQ_EMPTY(&rt->mfc_stall))
392 			break;
393 	}
394 
395 	return (rt);
396 }
397 
398 /*
399  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
400  */
401 static int
402 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
403 {
404     int	error, optval;
405     vifi_t	vifi;
406     struct	vifctl vifc;
407     struct	mfcctl2 mfc;
408     struct	bw_upcall bw_upcall;
409     uint32_t	i;
410 
411     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
412 	return EPERM;
413 
414     error = 0;
415     switch (sopt->sopt_name) {
416     case MRT_INIT:
417 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
418 	if (error)
419 	    break;
420 	error = ip_mrouter_init(so, optval);
421 	break;
422 
423     case MRT_DONE:
424 	error = ip_mrouter_done();
425 	break;
426 
427     case MRT_ADD_VIF:
428 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
429 	if (error)
430 	    break;
431 	error = add_vif(&vifc);
432 	break;
433 
434     case MRT_DEL_VIF:
435 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
436 	if (error)
437 	    break;
438 	error = del_vif(vifi);
439 	break;
440 
441     case MRT_ADD_MFC:
442     case MRT_DEL_MFC:
443 	/*
444 	 * select data size depending on API version.
445 	 */
446 	if (sopt->sopt_name == MRT_ADD_MFC &&
447 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
448 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
449 				sizeof(struct mfcctl2));
450 	} else {
451 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
452 				sizeof(struct mfcctl));
453 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
454 			sizeof(mfc) - sizeof(struct mfcctl));
455 	}
456 	if (error)
457 	    break;
458 	if (sopt->sopt_name == MRT_ADD_MFC)
459 	    error = add_mfc(&mfc);
460 	else
461 	    error = del_mfc(&mfc);
462 	break;
463 
464     case MRT_ASSERT:
465 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
466 	if (error)
467 	    break;
468 	set_assert(optval);
469 	break;
470 
471     case MRT_API_CONFIG:
472 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
473 	if (!error)
474 	    error = set_api_config(&i);
475 	if (!error)
476 	    error = sooptcopyout(sopt, &i, sizeof i);
477 	break;
478 
479     case MRT_ADD_BW_UPCALL:
480     case MRT_DEL_BW_UPCALL:
481 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
482 				sizeof bw_upcall);
483 	if (error)
484 	    break;
485 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
486 	    error = add_bw_upcall(&bw_upcall);
487 	else
488 	    error = del_bw_upcall(&bw_upcall);
489 	break;
490 
491     default:
492 	error = EOPNOTSUPP;
493 	break;
494     }
495     return error;
496 }
497 
498 /*
499  * Handle MRT getsockopt commands
500  */
501 static int
502 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
503 {
504     int error;
505 
506     switch (sopt->sopt_name) {
507     case MRT_VERSION:
508 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
509 	break;
510 
511     case MRT_ASSERT:
512 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
513 	    sizeof V_pim_assert_enabled);
514 	break;
515 
516     case MRT_API_SUPPORT:
517 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
518 	break;
519 
520     case MRT_API_CONFIG:
521 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
522 	break;
523 
524     default:
525 	error = EOPNOTSUPP;
526 	break;
527     }
528     return error;
529 }
530 
531 /*
532  * Handle ioctl commands to obtain information from the cache
533  */
534 static int
535 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
536 {
537     int error = 0;
538 
539     /*
540      * Currently the only function calling this ioctl routine is rtioctl_fib().
541      * Typically, only root can create the raw socket in order to execute
542      * this ioctl method, however the request might be coming from a prison
543      */
544     error = priv_check(curthread, PRIV_NETINET_MROUTE);
545     if (error)
546 	return (error);
547     switch (cmd) {
548     case (SIOCGETVIFCNT):
549 	error = get_vif_cnt((struct sioc_vif_req *)data);
550 	break;
551 
552     case (SIOCGETSGCNT):
553 	error = get_sg_cnt((struct sioc_sg_req *)data);
554 	break;
555 
556     default:
557 	error = EINVAL;
558 	break;
559     }
560     return error;
561 }
562 
563 /*
564  * returns the packet, byte, rpf-failure count for the source group provided
565  */
566 static int
567 get_sg_cnt(struct sioc_sg_req *req)
568 {
569     struct mfc *rt;
570 
571     MFC_LOCK();
572     rt = mfc_find(&req->src, &req->grp);
573     if (rt == NULL) {
574 	MFC_UNLOCK();
575 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
576 	return EADDRNOTAVAIL;
577     }
578     req->pktcnt = rt->mfc_pkt_cnt;
579     req->bytecnt = rt->mfc_byte_cnt;
580     req->wrong_if = rt->mfc_wrong_if;
581     MFC_UNLOCK();
582     return 0;
583 }
584 
585 /*
586  * returns the input and output packet and byte counts on the vif provided
587  */
588 static int
589 get_vif_cnt(struct sioc_vif_req *req)
590 {
591     vifi_t vifi = req->vifi;
592 
593     VIF_LOCK();
594     if (vifi >= V_numvifs) {
595 	VIF_UNLOCK();
596 	return EINVAL;
597     }
598 
599     req->icount = V_viftable[vifi].v_pkt_in;
600     req->ocount = V_viftable[vifi].v_pkt_out;
601     req->ibytes = V_viftable[vifi].v_bytes_in;
602     req->obytes = V_viftable[vifi].v_bytes_out;
603     VIF_UNLOCK();
604 
605     return 0;
606 }
607 
608 static void
609 if_detached_event(void *arg __unused, struct ifnet *ifp)
610 {
611     vifi_t vifi;
612     u_long i;
613 
614     MROUTER_LOCK();
615 
616     if (V_ip_mrouter == NULL) {
617 	MROUTER_UNLOCK();
618 	return;
619     }
620 
621     VIF_LOCK();
622     MFC_LOCK();
623 
624     /*
625      * Tear down multicast forwarder state associated with this ifnet.
626      * 1. Walk the vif list, matching vifs against this ifnet.
627      * 2. Walk the multicast forwarding cache (mfc) looking for
628      *    inner matches with this vif's index.
629      * 3. Expire any matching multicast forwarding cache entries.
630      * 4. Free vif state. This should disable ALLMULTI on the interface.
631      */
632     for (vifi = 0; vifi < V_numvifs; vifi++) {
633 	if (V_viftable[vifi].v_ifp != ifp)
634 		continue;
635 	for (i = 0; i < mfchashsize; i++) {
636 		struct mfc *rt, *nrt;
637 
638 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
639 			if (rt->mfc_parent == vifi) {
640 				expire_mfc(rt);
641 			}
642 		}
643 	}
644 	del_vif_locked(vifi);
645     }
646 
647     MFC_UNLOCK();
648     VIF_UNLOCK();
649 
650     MROUTER_UNLOCK();
651 }
652 
653 /*
654  * Enable multicast forwarding.
655  */
656 static int
657 ip_mrouter_init(struct socket *so, int version)
658 {
659 
660     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
661         so->so_type, so->so_proto->pr_protocol);
662 
663     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
664 	return EOPNOTSUPP;
665 
666     if (version != 1)
667 	return ENOPROTOOPT;
668 
669     MROUTER_LOCK();
670 
671     if (ip_mrouter_unloading) {
672 	MROUTER_UNLOCK();
673 	return ENOPROTOOPT;
674     }
675 
676     if (V_ip_mrouter != NULL) {
677 	MROUTER_UNLOCK();
678 	return EADDRINUSE;
679     }
680 
681     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
682 	HASH_NOWAIT);
683 
684     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
685 	curvnet);
686     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
687 	curvnet);
688     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
689 	curvnet);
690 
691     V_ip_mrouter = so;
692     ip_mrouter_cnt++;
693 
694     MROUTER_UNLOCK();
695 
696     CTR1(KTR_IPMF, "%s: done", __func__);
697 
698     return 0;
699 }
700 
701 /*
702  * Disable multicast forwarding.
703  */
704 static int
705 X_ip_mrouter_done(void)
706 {
707     struct ifnet *ifp;
708     u_long i;
709     vifi_t vifi;
710 
711     MROUTER_LOCK();
712 
713     if (V_ip_mrouter == NULL) {
714 	MROUTER_UNLOCK();
715 	return EINVAL;
716     }
717 
718     /*
719      * Detach/disable hooks to the reset of the system.
720      */
721     V_ip_mrouter = NULL;
722     ip_mrouter_cnt--;
723     V_mrt_api_config = 0;
724 
725     VIF_LOCK();
726 
727     /*
728      * For each phyint in use, disable promiscuous reception of all IP
729      * multicasts.
730      */
731     for (vifi = 0; vifi < V_numvifs; vifi++) {
732 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
733 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
734 	    ifp = V_viftable[vifi].v_ifp;
735 	    if_allmulti(ifp, 0);
736 	}
737     }
738     bzero((caddr_t)V_viftable, sizeof(V_viftable));
739     V_numvifs = 0;
740     V_pim_assert_enabled = 0;
741 
742     VIF_UNLOCK();
743 
744     callout_stop(&V_expire_upcalls_ch);
745     callout_stop(&V_bw_upcalls_ch);
746     callout_stop(&V_bw_meter_ch);
747 
748     MFC_LOCK();
749 
750     /*
751      * Free all multicast forwarding cache entries.
752      * Do not use hashdestroy(), as we must perform other cleanup.
753      */
754     for (i = 0; i < mfchashsize; i++) {
755 	struct mfc *rt, *nrt;
756 
757 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
758 		expire_mfc(rt);
759 	}
760     }
761     free(V_mfchashtbl, M_MRTABLE);
762     V_mfchashtbl = NULL;
763 
764     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
765 
766     V_bw_upcalls_n = 0;
767     bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
768 
769     MFC_UNLOCK();
770 
771     V_reg_vif_num = VIFI_INVALID;
772 
773     MROUTER_UNLOCK();
774 
775     CTR1(KTR_IPMF, "%s: done", __func__);
776 
777     return 0;
778 }
779 
780 /*
781  * Set PIM assert processing global
782  */
783 static int
784 set_assert(int i)
785 {
786     if ((i != 1) && (i != 0))
787 	return EINVAL;
788 
789     V_pim_assert_enabled = i;
790 
791     return 0;
792 }
793 
794 /*
795  * Configure API capabilities
796  */
797 int
798 set_api_config(uint32_t *apival)
799 {
800     u_long i;
801 
802     /*
803      * We can set the API capabilities only if it is the first operation
804      * after MRT_INIT. I.e.:
805      *  - there are no vifs installed
806      *  - pim_assert is not enabled
807      *  - the MFC table is empty
808      */
809     if (V_numvifs > 0) {
810 	*apival = 0;
811 	return EPERM;
812     }
813     if (V_pim_assert_enabled) {
814 	*apival = 0;
815 	return EPERM;
816     }
817 
818     MFC_LOCK();
819 
820     for (i = 0; i < mfchashsize; i++) {
821 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
822 	    MFC_UNLOCK();
823 	    *apival = 0;
824 	    return EPERM;
825 	}
826     }
827 
828     MFC_UNLOCK();
829 
830     V_mrt_api_config = *apival & mrt_api_support;
831     *apival = V_mrt_api_config;
832 
833     return 0;
834 }
835 
836 /*
837  * Add a vif to the vif table
838  */
839 static int
840 add_vif(struct vifctl *vifcp)
841 {
842     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
843     struct sockaddr_in sin = {sizeof sin, AF_INET};
844     struct ifaddr *ifa;
845     struct ifnet *ifp;
846     int error;
847 
848     VIF_LOCK();
849     if (vifcp->vifc_vifi >= MAXVIFS) {
850 	VIF_UNLOCK();
851 	return EINVAL;
852     }
853     /* rate limiting is no longer supported by this code */
854     if (vifcp->vifc_rate_limit != 0) {
855 	log(LOG_ERR, "rate limiting is no longer supported\n");
856 	VIF_UNLOCK();
857 	return EINVAL;
858     }
859     if (!in_nullhost(vifp->v_lcl_addr)) {
860 	VIF_UNLOCK();
861 	return EADDRINUSE;
862     }
863     if (in_nullhost(vifcp->vifc_lcl_addr)) {
864 	VIF_UNLOCK();
865 	return EADDRNOTAVAIL;
866     }
867 
868     /* Find the interface with an address in AF_INET family */
869     if (vifcp->vifc_flags & VIFF_REGISTER) {
870 	/*
871 	 * XXX: Because VIFF_REGISTER does not really need a valid
872 	 * local interface (e.g. it could be 127.0.0.2), we don't
873 	 * check its address.
874 	 */
875 	ifp = NULL;
876     } else {
877 	sin.sin_addr = vifcp->vifc_lcl_addr;
878 	NET_EPOCH_ENTER();
879 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
880 	if (ifa == NULL) {
881 		NET_EPOCH_EXIT();
882 	    VIF_UNLOCK();
883 	    return EADDRNOTAVAIL;
884 	}
885 	ifp = ifa->ifa_ifp;
886 	NET_EPOCH_EXIT();
887     }
888 
889     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
890 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
891 	VIF_UNLOCK();
892 	return EOPNOTSUPP;
893     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
894 	ifp = &V_multicast_register_if;
895 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
896 	if (V_reg_vif_num == VIFI_INVALID) {
897 	    if_initname(&V_multicast_register_if, "register_vif", 0);
898 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
899 	    V_reg_vif_num = vifcp->vifc_vifi;
900 	}
901     } else {		/* Make sure the interface supports multicast */
902 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
903 	    VIF_UNLOCK();
904 	    return EOPNOTSUPP;
905 	}
906 
907 	/* Enable promiscuous reception of all IP multicasts from the if */
908 	error = if_allmulti(ifp, 1);
909 	if (error) {
910 	    VIF_UNLOCK();
911 	    return error;
912 	}
913     }
914 
915     vifp->v_flags     = vifcp->vifc_flags;
916     vifp->v_threshold = vifcp->vifc_threshold;
917     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
918     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
919     vifp->v_ifp       = ifp;
920     /* initialize per vif pkt counters */
921     vifp->v_pkt_in    = 0;
922     vifp->v_pkt_out   = 0;
923     vifp->v_bytes_in  = 0;
924     vifp->v_bytes_out = 0;
925 
926     /* Adjust numvifs up if the vifi is higher than numvifs */
927     if (V_numvifs <= vifcp->vifc_vifi)
928 	V_numvifs = vifcp->vifc_vifi + 1;
929 
930     VIF_UNLOCK();
931 
932     CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
933 	(int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
934 	(int)vifcp->vifc_threshold);
935 
936     return 0;
937 }
938 
939 /*
940  * Delete a vif from the vif table
941  */
942 static int
943 del_vif_locked(vifi_t vifi)
944 {
945     struct vif *vifp;
946 
947     VIF_LOCK_ASSERT();
948 
949     if (vifi >= V_numvifs) {
950 	return EINVAL;
951     }
952     vifp = &V_viftable[vifi];
953     if (in_nullhost(vifp->v_lcl_addr)) {
954 	return EADDRNOTAVAIL;
955     }
956 
957     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
958 	if_allmulti(vifp->v_ifp, 0);
959 
960     if (vifp->v_flags & VIFF_REGISTER)
961 	V_reg_vif_num = VIFI_INVALID;
962 
963     bzero((caddr_t)vifp, sizeof (*vifp));
964 
965     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
966 
967     /* Adjust numvifs down */
968     for (vifi = V_numvifs; vifi > 0; vifi--)
969 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
970 	    break;
971     V_numvifs = vifi;
972 
973     return 0;
974 }
975 
976 static int
977 del_vif(vifi_t vifi)
978 {
979     int cc;
980 
981     VIF_LOCK();
982     cc = del_vif_locked(vifi);
983     VIF_UNLOCK();
984 
985     return cc;
986 }
987 
988 /*
989  * update an mfc entry without resetting counters and S,G addresses.
990  */
991 static void
992 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
993 {
994     int i;
995 
996     rt->mfc_parent = mfccp->mfcc_parent;
997     for (i = 0; i < V_numvifs; i++) {
998 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
999 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1000 	    MRT_MFC_FLAGS_ALL;
1001     }
1002     /* set the RP address */
1003     if (V_mrt_api_config & MRT_MFC_RP)
1004 	rt->mfc_rp = mfccp->mfcc_rp;
1005     else
1006 	rt->mfc_rp.s_addr = INADDR_ANY;
1007 }
1008 
1009 /*
1010  * fully initialize an mfc entry from the parameter.
1011  */
1012 static void
1013 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1014 {
1015     rt->mfc_origin     = mfccp->mfcc_origin;
1016     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1017 
1018     update_mfc_params(rt, mfccp);
1019 
1020     /* initialize pkt counters per src-grp */
1021     rt->mfc_pkt_cnt    = 0;
1022     rt->mfc_byte_cnt   = 0;
1023     rt->mfc_wrong_if   = 0;
1024     timevalclear(&rt->mfc_last_assert);
1025 }
1026 
1027 static void
1028 expire_mfc(struct mfc *rt)
1029 {
1030 	struct rtdetq *rte, *nrte;
1031 
1032 	MFC_LOCK_ASSERT();
1033 
1034 	free_bw_list(rt->mfc_bw_meter);
1035 
1036 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1037 		m_freem(rte->m);
1038 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1039 		free(rte, M_MRTABLE);
1040 	}
1041 
1042 	LIST_REMOVE(rt, mfc_hash);
1043 	free(rt, M_MRTABLE);
1044 }
1045 
1046 /*
1047  * Add an mfc entry
1048  */
1049 static int
1050 add_mfc(struct mfcctl2 *mfccp)
1051 {
1052     struct mfc *rt;
1053     struct rtdetq *rte, *nrte;
1054     u_long hash = 0;
1055     u_short nstl;
1056 
1057     VIF_LOCK();
1058     MFC_LOCK();
1059 
1060     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1061 
1062     /* If an entry already exists, just update the fields */
1063     if (rt) {
1064 	CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1065 	    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1066 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1067 	    mfccp->mfcc_parent);
1068 	update_mfc_params(rt, mfccp);
1069 	MFC_UNLOCK();
1070 	VIF_UNLOCK();
1071 	return (0);
1072     }
1073 
1074     /*
1075      * Find the entry for which the upcall was made and update
1076      */
1077     nstl = 0;
1078     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1079     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1080 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1081 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1082 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1083 		CTR5(KTR_IPMF,
1084 		    "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1085 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1086 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1087 		    mfccp->mfcc_parent,
1088 		    TAILQ_FIRST(&rt->mfc_stall));
1089 		if (nstl++)
1090 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1091 
1092 		init_mfc_params(rt, mfccp);
1093 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1094 		V_nexpire[hash]--;
1095 
1096 		/* Free queued packets, but attempt to forward them first. */
1097 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1098 			if (rte->ifp != NULL)
1099 				ip_mdq(rte->m, rte->ifp, rt, -1);
1100 			m_freem(rte->m);
1101 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1102 			rt->mfc_nstall--;
1103 			free(rte, M_MRTABLE);
1104 		}
1105 	}
1106     }
1107 
1108     /*
1109      * It is possible that an entry is being inserted without an upcall
1110      */
1111     if (nstl == 0) {
1112 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1113 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1114 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1115 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1116 			init_mfc_params(rt, mfccp);
1117 			if (rt->mfc_expire)
1118 			    V_nexpire[hash]--;
1119 			rt->mfc_expire = 0;
1120 			break; /* XXX */
1121 		}
1122 	}
1123 
1124 	if (rt == NULL) {		/* no upcall, so make a new entry */
1125 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1126 	    if (rt == NULL) {
1127 		MFC_UNLOCK();
1128 		VIF_UNLOCK();
1129 		return (ENOBUFS);
1130 	    }
1131 
1132 	    init_mfc_params(rt, mfccp);
1133 	    TAILQ_INIT(&rt->mfc_stall);
1134 	    rt->mfc_nstall = 0;
1135 
1136 	    rt->mfc_expire     = 0;
1137 	    rt->mfc_bw_meter = NULL;
1138 
1139 	    /* insert new entry at head of hash chain */
1140 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1141 	}
1142     }
1143 
1144     MFC_UNLOCK();
1145     VIF_UNLOCK();
1146 
1147     return (0);
1148 }
1149 
1150 /*
1151  * Delete an mfc entry
1152  */
1153 static int
1154 del_mfc(struct mfcctl2 *mfccp)
1155 {
1156     struct in_addr	origin;
1157     struct in_addr	mcastgrp;
1158     struct mfc		*rt;
1159 
1160     origin = mfccp->mfcc_origin;
1161     mcastgrp = mfccp->mfcc_mcastgrp;
1162 
1163     CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1164 	ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1165 
1166     MFC_LOCK();
1167 
1168     rt = mfc_find(&origin, &mcastgrp);
1169     if (rt == NULL) {
1170 	MFC_UNLOCK();
1171 	return EADDRNOTAVAIL;
1172     }
1173 
1174     /*
1175      * free the bw_meter entries
1176      */
1177     free_bw_list(rt->mfc_bw_meter);
1178     rt->mfc_bw_meter = NULL;
1179 
1180     LIST_REMOVE(rt, mfc_hash);
1181     free(rt, M_MRTABLE);
1182 
1183     MFC_UNLOCK();
1184 
1185     return (0);
1186 }
1187 
1188 /*
1189  * Send a message to the routing daemon on the multicast routing socket.
1190  */
1191 static int
1192 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1193 {
1194     if (s) {
1195 	SOCKBUF_LOCK(&s->so_rcv);
1196 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1197 	    NULL) != 0) {
1198 	    sorwakeup_locked(s);
1199 	    return 0;
1200 	}
1201 	SOCKBUF_UNLOCK(&s->so_rcv);
1202     }
1203     m_freem(mm);
1204     return -1;
1205 }
1206 
1207 /*
1208  * IP multicast forwarding function. This function assumes that the packet
1209  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1210  * pointed to by "ifp", and the packet is to be relayed to other networks
1211  * that have members of the packet's destination IP multicast group.
1212  *
1213  * The packet is returned unscathed to the caller, unless it is
1214  * erroneous, in which case a non-zero return value tells the caller to
1215  * discard it.
1216  */
1217 
1218 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1219 
1220 static int
1221 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1222     struct ip_moptions *imo)
1223 {
1224     struct mfc *rt;
1225     int error;
1226     vifi_t vifi;
1227 
1228     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1229 	ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1230 
1231     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1232 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1233 	/*
1234 	 * Packet arrived via a physical interface or
1235 	 * an encapsulated tunnel or a register_vif.
1236 	 */
1237     } else {
1238 	/*
1239 	 * Packet arrived through a source-route tunnel.
1240 	 * Source-route tunnels are no longer supported.
1241 	 */
1242 	return (1);
1243     }
1244 
1245     VIF_LOCK();
1246     MFC_LOCK();
1247     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1248 	if (ip->ip_ttl < MAXTTL)
1249 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1250 	error = ip_mdq(m, ifp, NULL, vifi);
1251 	MFC_UNLOCK();
1252 	VIF_UNLOCK();
1253 	return error;
1254     }
1255 
1256     /*
1257      * Don't forward a packet with time-to-live of zero or one,
1258      * or a packet destined to a local-only group.
1259      */
1260     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1261 	MFC_UNLOCK();
1262 	VIF_UNLOCK();
1263 	return 0;
1264     }
1265 
1266     /*
1267      * Determine forwarding vifs from the forwarding cache table
1268      */
1269     MRTSTAT_INC(mrts_mfc_lookups);
1270     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1271 
1272     /* Entry exists, so forward if necessary */
1273     if (rt != NULL) {
1274 	error = ip_mdq(m, ifp, rt, -1);
1275 	MFC_UNLOCK();
1276 	VIF_UNLOCK();
1277 	return error;
1278     } else {
1279 	/*
1280 	 * If we don't have a route for packet's origin,
1281 	 * Make a copy of the packet & send message to routing daemon
1282 	 */
1283 
1284 	struct mbuf *mb0;
1285 	struct rtdetq *rte;
1286 	u_long hash;
1287 	int hlen = ip->ip_hl << 2;
1288 
1289 	MRTSTAT_INC(mrts_mfc_misses);
1290 	MRTSTAT_INC(mrts_no_route);
1291 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1292 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1293 
1294 	/*
1295 	 * Allocate mbufs early so that we don't do extra work if we are
1296 	 * just going to fail anyway.  Make sure to pullup the header so
1297 	 * that other people can't step on it.
1298 	 */
1299 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1300 	    M_NOWAIT|M_ZERO);
1301 	if (rte == NULL) {
1302 	    MFC_UNLOCK();
1303 	    VIF_UNLOCK();
1304 	    return ENOBUFS;
1305 	}
1306 
1307 	mb0 = m_copypacket(m, M_NOWAIT);
1308 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1309 	    mb0 = m_pullup(mb0, hlen);
1310 	if (mb0 == NULL) {
1311 	    free(rte, M_MRTABLE);
1312 	    MFC_UNLOCK();
1313 	    VIF_UNLOCK();
1314 	    return ENOBUFS;
1315 	}
1316 
1317 	/* is there an upcall waiting for this flow ? */
1318 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1319 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1320 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1321 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1322 		    !TAILQ_EMPTY(&rt->mfc_stall))
1323 			break;
1324 	}
1325 
1326 	if (rt == NULL) {
1327 	    int i;
1328 	    struct igmpmsg *im;
1329 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1330 	    struct mbuf *mm;
1331 
1332 	    /*
1333 	     * Locate the vifi for the incoming interface for this packet.
1334 	     * If none found, drop packet.
1335 	     */
1336 	    for (vifi = 0; vifi < V_numvifs &&
1337 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1338 		;
1339 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1340 		goto non_fatal;
1341 
1342 	    /* no upcall, so make a new entry */
1343 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1344 	    if (rt == NULL)
1345 		goto fail;
1346 
1347 	    /* Make a copy of the header to send to the user level process */
1348 	    mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1349 	    if (mm == NULL)
1350 		goto fail1;
1351 
1352 	    /*
1353 	     * Send message to routing daemon to install
1354 	     * a route into the kernel table
1355 	     */
1356 
1357 	    im = mtod(mm, struct igmpmsg *);
1358 	    im->im_msgtype = IGMPMSG_NOCACHE;
1359 	    im->im_mbz = 0;
1360 	    im->im_vif = vifi;
1361 
1362 	    MRTSTAT_INC(mrts_upcalls);
1363 
1364 	    k_igmpsrc.sin_addr = ip->ip_src;
1365 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1366 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1367 		MRTSTAT_INC(mrts_upq_sockfull);
1368 fail1:
1369 		free(rt, M_MRTABLE);
1370 fail:
1371 		free(rte, M_MRTABLE);
1372 		m_freem(mb0);
1373 		MFC_UNLOCK();
1374 		VIF_UNLOCK();
1375 		return ENOBUFS;
1376 	    }
1377 
1378 	    /* insert new entry at head of hash chain */
1379 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1380 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1381 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1382 	    V_nexpire[hash]++;
1383 	    for (i = 0; i < V_numvifs; i++) {
1384 		rt->mfc_ttls[i] = 0;
1385 		rt->mfc_flags[i] = 0;
1386 	    }
1387 	    rt->mfc_parent = -1;
1388 
1389 	    /* clear the RP address */
1390 	    rt->mfc_rp.s_addr = INADDR_ANY;
1391 	    rt->mfc_bw_meter = NULL;
1392 
1393 	    /* initialize pkt counters per src-grp */
1394 	    rt->mfc_pkt_cnt = 0;
1395 	    rt->mfc_byte_cnt = 0;
1396 	    rt->mfc_wrong_if = 0;
1397 	    timevalclear(&rt->mfc_last_assert);
1398 
1399 	    TAILQ_INIT(&rt->mfc_stall);
1400 	    rt->mfc_nstall = 0;
1401 
1402 	    /* link into table */
1403 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1404 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1405 	    rt->mfc_nstall++;
1406 
1407 	} else {
1408 	    /* determine if queue has overflowed */
1409 	    if (rt->mfc_nstall > MAX_UPQ) {
1410 		MRTSTAT_INC(mrts_upq_ovflw);
1411 non_fatal:
1412 		free(rte, M_MRTABLE);
1413 		m_freem(mb0);
1414 		MFC_UNLOCK();
1415 		VIF_UNLOCK();
1416 		return (0);
1417 	    }
1418 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1419 	    rt->mfc_nstall++;
1420 	}
1421 
1422 	rte->m			= mb0;
1423 	rte->ifp		= ifp;
1424 
1425 	MFC_UNLOCK();
1426 	VIF_UNLOCK();
1427 
1428 	return 0;
1429     }
1430 }
1431 
1432 /*
1433  * Clean up the cache entry if upcall is not serviced
1434  */
1435 static void
1436 expire_upcalls(void *arg)
1437 {
1438     u_long i;
1439 
1440     CURVNET_SET((struct vnet *) arg);
1441 
1442     MFC_LOCK();
1443 
1444     for (i = 0; i < mfchashsize; i++) {
1445 	struct mfc *rt, *nrt;
1446 
1447 	if (V_nexpire[i] == 0)
1448 	    continue;
1449 
1450 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1451 		if (TAILQ_EMPTY(&rt->mfc_stall))
1452 			continue;
1453 
1454 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1455 			continue;
1456 
1457 		/*
1458 		 * free the bw_meter entries
1459 		 */
1460 		while (rt->mfc_bw_meter != NULL) {
1461 		    struct bw_meter *x = rt->mfc_bw_meter;
1462 
1463 		    rt->mfc_bw_meter = x->bm_mfc_next;
1464 		    free(x, M_BWMETER);
1465 		}
1466 
1467 		MRTSTAT_INC(mrts_cache_cleanups);
1468 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1469 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1470 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1471 
1472 		expire_mfc(rt);
1473 	    }
1474     }
1475 
1476     MFC_UNLOCK();
1477 
1478     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1479 	curvnet);
1480 
1481     CURVNET_RESTORE();
1482 }
1483 
1484 /*
1485  * Packet forwarding routine once entry in the cache is made
1486  */
1487 static int
1488 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1489 {
1490     struct ip  *ip = mtod(m, struct ip *);
1491     vifi_t vifi;
1492     int plen = ntohs(ip->ip_len);
1493 
1494     VIF_LOCK_ASSERT();
1495 
1496     /*
1497      * If xmt_vif is not -1, send on only the requested vif.
1498      *
1499      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1500      */
1501     if (xmt_vif < V_numvifs) {
1502 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1503 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1504 	else
1505 		phyint_send(ip, V_viftable + xmt_vif, m);
1506 	return 1;
1507     }
1508 
1509     /*
1510      * Don't forward if it didn't arrive from the parent vif for its origin.
1511      */
1512     vifi = rt->mfc_parent;
1513     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1514 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1515 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1516 	MRTSTAT_INC(mrts_wrong_if);
1517 	++rt->mfc_wrong_if;
1518 	/*
1519 	 * If we are doing PIM assert processing, send a message
1520 	 * to the routing daemon.
1521 	 *
1522 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1523 	 * can complete the SPT switch, regardless of the type
1524 	 * of the iif (broadcast media, GRE tunnel, etc).
1525 	 */
1526 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1527 	    V_viftable[vifi].v_ifp) {
1528 
1529 	    if (ifp == &V_multicast_register_if)
1530 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1531 
1532 	    /* Get vifi for the incoming packet */
1533 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1534 		vifi++)
1535 		;
1536 	    if (vifi >= V_numvifs)
1537 		return 0;	/* The iif is not found: ignore the packet. */
1538 
1539 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1540 		return 0;	/* WRONGVIF disabled: ignore the packet */
1541 
1542 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1543 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1544 		struct igmpmsg *im;
1545 		int hlen = ip->ip_hl << 2;
1546 		struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1547 
1548 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1549 		    mm = m_pullup(mm, hlen);
1550 		if (mm == NULL)
1551 		    return ENOBUFS;
1552 
1553 		im = mtod(mm, struct igmpmsg *);
1554 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1555 		im->im_mbz		= 0;
1556 		im->im_vif		= vifi;
1557 
1558 		MRTSTAT_INC(mrts_upcalls);
1559 
1560 		k_igmpsrc.sin_addr = im->im_src;
1561 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1562 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1563 		    MRTSTAT_INC(mrts_upq_sockfull);
1564 		    return ENOBUFS;
1565 		}
1566 	    }
1567 	}
1568 	return 0;
1569     }
1570 
1571 
1572     /* If I sourced this packet, it counts as output, else it was input. */
1573     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1574 	V_viftable[vifi].v_pkt_out++;
1575 	V_viftable[vifi].v_bytes_out += plen;
1576     } else {
1577 	V_viftable[vifi].v_pkt_in++;
1578 	V_viftable[vifi].v_bytes_in += plen;
1579     }
1580     rt->mfc_pkt_cnt++;
1581     rt->mfc_byte_cnt += plen;
1582 
1583     /*
1584      * For each vif, decide if a copy of the packet should be forwarded.
1585      * Forward if:
1586      *		- the ttl exceeds the vif's threshold
1587      *		- there are group members downstream on interface
1588      */
1589     for (vifi = 0; vifi < V_numvifs; vifi++)
1590 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1591 	    V_viftable[vifi].v_pkt_out++;
1592 	    V_viftable[vifi].v_bytes_out += plen;
1593 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1594 		pim_register_send(ip, V_viftable + vifi, m, rt);
1595 	    else
1596 		phyint_send(ip, V_viftable + vifi, m);
1597 	}
1598 
1599     /*
1600      * Perform upcall-related bw measuring.
1601      */
1602     if (rt->mfc_bw_meter != NULL) {
1603 	struct bw_meter *x;
1604 	struct timeval now;
1605 
1606 	microtime(&now);
1607 	MFC_LOCK_ASSERT();
1608 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1609 	    bw_meter_receive_packet(x, plen, &now);
1610     }
1611 
1612     return 0;
1613 }
1614 
1615 /*
1616  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1617  */
1618 static int
1619 X_legal_vif_num(int vif)
1620 {
1621 	int ret;
1622 
1623 	ret = 0;
1624 	if (vif < 0)
1625 		return (ret);
1626 
1627 	VIF_LOCK();
1628 	if (vif < V_numvifs)
1629 		ret = 1;
1630 	VIF_UNLOCK();
1631 
1632 	return (ret);
1633 }
1634 
1635 /*
1636  * Return the local address used by this vif
1637  */
1638 static u_long
1639 X_ip_mcast_src(int vifi)
1640 {
1641 	in_addr_t addr;
1642 
1643 	addr = INADDR_ANY;
1644 	if (vifi < 0)
1645 		return (addr);
1646 
1647 	VIF_LOCK();
1648 	if (vifi < V_numvifs)
1649 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1650 	VIF_UNLOCK();
1651 
1652 	return (addr);
1653 }
1654 
1655 static void
1656 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1657 {
1658     struct mbuf *mb_copy;
1659     int hlen = ip->ip_hl << 2;
1660 
1661     VIF_LOCK_ASSERT();
1662 
1663     /*
1664      * Make a new reference to the packet; make sure that
1665      * the IP header is actually copied, not just referenced,
1666      * so that ip_output() only scribbles on the copy.
1667      */
1668     mb_copy = m_copypacket(m, M_NOWAIT);
1669     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1670 	mb_copy = m_pullup(mb_copy, hlen);
1671     if (mb_copy == NULL)
1672 	return;
1673 
1674     send_packet(vifp, mb_copy);
1675 }
1676 
1677 static void
1678 send_packet(struct vif *vifp, struct mbuf *m)
1679 {
1680 	struct ip_moptions imo;
1681 	struct in_multi *imm[2];
1682 	int error __unused;
1683 
1684 	VIF_LOCK_ASSERT();
1685 
1686 	imo.imo_multicast_ifp  = vifp->v_ifp;
1687 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1688 	imo.imo_multicast_loop = 1;
1689 	imo.imo_multicast_vif  = -1;
1690 	imo.imo_num_memberships = 0;
1691 	imo.imo_max_memberships = 2;
1692 	imo.imo_membership  = &imm[0];
1693 
1694 	/*
1695 	 * Re-entrancy should not be a problem here, because
1696 	 * the packets that we send out and are looped back at us
1697 	 * should get rejected because they appear to come from
1698 	 * the loopback interface, thus preventing looping.
1699 	 */
1700 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1701 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1702 	    (ptrdiff_t)(vifp - V_viftable), error);
1703 }
1704 
1705 /*
1706  * Stubs for old RSVP socket shim implementation.
1707  */
1708 
1709 static int
1710 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1711 {
1712 
1713 	return (EOPNOTSUPP);
1714 }
1715 
1716 static void
1717 X_ip_rsvp_force_done(struct socket *so __unused)
1718 {
1719 
1720 }
1721 
1722 static int
1723 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1724 {
1725 	struct mbuf *m;
1726 
1727 	m = *mp;
1728 	*mp = NULL;
1729 	if (!V_rsvp_on)
1730 		m_freem(m);
1731 	return (IPPROTO_DONE);
1732 }
1733 
1734 /*
1735  * Code for bandwidth monitors
1736  */
1737 
1738 /*
1739  * Define common interface for timeval-related methods
1740  */
1741 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1742 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1743 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1744 
1745 static uint32_t
1746 compute_bw_meter_flags(struct bw_upcall *req)
1747 {
1748     uint32_t flags = 0;
1749 
1750     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1751 	flags |= BW_METER_UNIT_PACKETS;
1752     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1753 	flags |= BW_METER_UNIT_BYTES;
1754     if (req->bu_flags & BW_UPCALL_GEQ)
1755 	flags |= BW_METER_GEQ;
1756     if (req->bu_flags & BW_UPCALL_LEQ)
1757 	flags |= BW_METER_LEQ;
1758 
1759     return flags;
1760 }
1761 
1762 /*
1763  * Add a bw_meter entry
1764  */
1765 static int
1766 add_bw_upcall(struct bw_upcall *req)
1767 {
1768     struct mfc *mfc;
1769     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1770 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1771     struct timeval now;
1772     struct bw_meter *x;
1773     uint32_t flags;
1774 
1775     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1776 	return EOPNOTSUPP;
1777 
1778     /* Test if the flags are valid */
1779     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1780 	return EINVAL;
1781     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1782 	return EINVAL;
1783     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1784 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1785 	return EINVAL;
1786 
1787     /* Test if the threshold time interval is valid */
1788     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1789 	return EINVAL;
1790 
1791     flags = compute_bw_meter_flags(req);
1792 
1793     /*
1794      * Find if we have already same bw_meter entry
1795      */
1796     MFC_LOCK();
1797     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1798     if (mfc == NULL) {
1799 	MFC_UNLOCK();
1800 	return EADDRNOTAVAIL;
1801     }
1802     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1803 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1804 			   &req->bu_threshold.b_time, ==)) &&
1805 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1806 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1807 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1808 	    MFC_UNLOCK();
1809 	    return 0;		/* XXX Already installed */
1810 	}
1811     }
1812 
1813     /* Allocate the new bw_meter entry */
1814     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1815     if (x == NULL) {
1816 	MFC_UNLOCK();
1817 	return ENOBUFS;
1818     }
1819 
1820     /* Set the new bw_meter entry */
1821     x->bm_threshold.b_time = req->bu_threshold.b_time;
1822     microtime(&now);
1823     x->bm_start_time = now;
1824     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1825     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1826     x->bm_measured.b_packets = 0;
1827     x->bm_measured.b_bytes = 0;
1828     x->bm_flags = flags;
1829     x->bm_time_next = NULL;
1830     x->bm_time_hash = BW_METER_BUCKETS;
1831 
1832     /* Add the new bw_meter entry to the front of entries for this MFC */
1833     x->bm_mfc = mfc;
1834     x->bm_mfc_next = mfc->mfc_bw_meter;
1835     mfc->mfc_bw_meter = x;
1836     schedule_bw_meter(x, &now);
1837     MFC_UNLOCK();
1838 
1839     return 0;
1840 }
1841 
1842 static void
1843 free_bw_list(struct bw_meter *list)
1844 {
1845     while (list != NULL) {
1846 	struct bw_meter *x = list;
1847 
1848 	list = list->bm_mfc_next;
1849 	unschedule_bw_meter(x);
1850 	free(x, M_BWMETER);
1851     }
1852 }
1853 
1854 /*
1855  * Delete one or multiple bw_meter entries
1856  */
1857 static int
1858 del_bw_upcall(struct bw_upcall *req)
1859 {
1860     struct mfc *mfc;
1861     struct bw_meter *x;
1862 
1863     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1864 	return EOPNOTSUPP;
1865 
1866     MFC_LOCK();
1867 
1868     /* Find the corresponding MFC entry */
1869     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1870     if (mfc == NULL) {
1871 	MFC_UNLOCK();
1872 	return EADDRNOTAVAIL;
1873     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1874 	/*
1875 	 * Delete all bw_meter entries for this mfc
1876 	 */
1877 	struct bw_meter *list;
1878 
1879 	list = mfc->mfc_bw_meter;
1880 	mfc->mfc_bw_meter = NULL;
1881 	free_bw_list(list);
1882 	MFC_UNLOCK();
1883 	return 0;
1884     } else {			/* Delete a single bw_meter entry */
1885 	struct bw_meter *prev;
1886 	uint32_t flags = 0;
1887 
1888 	flags = compute_bw_meter_flags(req);
1889 
1890 	/* Find the bw_meter entry to delete */
1891 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1892 	     prev = x, x = x->bm_mfc_next) {
1893 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1894 			       &req->bu_threshold.b_time, ==)) &&
1895 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1896 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1897 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1898 		break;
1899 	}
1900 	if (x != NULL) { /* Delete entry from the list for this MFC */
1901 	    if (prev != NULL)
1902 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1903 	    else
1904 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1905 
1906 	    unschedule_bw_meter(x);
1907 	    MFC_UNLOCK();
1908 	    /* Free the bw_meter entry */
1909 	    free(x, M_BWMETER);
1910 	    return 0;
1911 	} else {
1912 	    MFC_UNLOCK();
1913 	    return EINVAL;
1914 	}
1915     }
1916     /* NOTREACHED */
1917 }
1918 
1919 /*
1920  * Perform bandwidth measurement processing that may result in an upcall
1921  */
1922 static void
1923 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1924 {
1925     struct timeval delta;
1926 
1927     MFC_LOCK_ASSERT();
1928 
1929     delta = *nowp;
1930     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1931 
1932     if (x->bm_flags & BW_METER_GEQ) {
1933 	/*
1934 	 * Processing for ">=" type of bw_meter entry
1935 	 */
1936 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1937 	    /* Reset the bw_meter entry */
1938 	    x->bm_start_time = *nowp;
1939 	    x->bm_measured.b_packets = 0;
1940 	    x->bm_measured.b_bytes = 0;
1941 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1942 	}
1943 
1944 	/* Record that a packet is received */
1945 	x->bm_measured.b_packets++;
1946 	x->bm_measured.b_bytes += plen;
1947 
1948 	/*
1949 	 * Test if we should deliver an upcall
1950 	 */
1951 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1952 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1953 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1954 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1955 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1956 		/* Prepare an upcall for delivery */
1957 		bw_meter_prepare_upcall(x, nowp);
1958 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1959 	    }
1960 	}
1961     } else if (x->bm_flags & BW_METER_LEQ) {
1962 	/*
1963 	 * Processing for "<=" type of bw_meter entry
1964 	 */
1965 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1966 	    /*
1967 	     * We are behind time with the multicast forwarding table
1968 	     * scanning for "<=" type of bw_meter entries, so test now
1969 	     * if we should deliver an upcall.
1970 	     */
1971 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1972 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1973 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1974 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1975 		/* Prepare an upcall for delivery */
1976 		bw_meter_prepare_upcall(x, nowp);
1977 	    }
1978 	    /* Reschedule the bw_meter entry */
1979 	    unschedule_bw_meter(x);
1980 	    schedule_bw_meter(x, nowp);
1981 	}
1982 
1983 	/* Record that a packet is received */
1984 	x->bm_measured.b_packets++;
1985 	x->bm_measured.b_bytes += plen;
1986 
1987 	/*
1988 	 * Test if we should restart the measuring interval
1989 	 */
1990 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1991 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1992 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1993 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1994 	    /* Don't restart the measuring interval */
1995 	} else {
1996 	    /* Do restart the measuring interval */
1997 	    /*
1998 	     * XXX: note that we don't unschedule and schedule, because this
1999 	     * might be too much overhead per packet. Instead, when we process
2000 	     * all entries for a given timer hash bin, we check whether it is
2001 	     * really a timeout. If not, we reschedule at that time.
2002 	     */
2003 	    x->bm_start_time = *nowp;
2004 	    x->bm_measured.b_packets = 0;
2005 	    x->bm_measured.b_bytes = 0;
2006 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2007 	}
2008     }
2009 }
2010 
2011 /*
2012  * Prepare a bandwidth-related upcall
2013  */
2014 static void
2015 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2016 {
2017     struct timeval delta;
2018     struct bw_upcall *u;
2019 
2020     MFC_LOCK_ASSERT();
2021 
2022     /*
2023      * Compute the measured time interval
2024      */
2025     delta = *nowp;
2026     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2027 
2028     /*
2029      * If there are too many pending upcalls, deliver them now
2030      */
2031     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2032 	bw_upcalls_send();
2033 
2034     /*
2035      * Set the bw_upcall entry
2036      */
2037     u = &V_bw_upcalls[V_bw_upcalls_n++];
2038     u->bu_src = x->bm_mfc->mfc_origin;
2039     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2040     u->bu_threshold.b_time = x->bm_threshold.b_time;
2041     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2042     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2043     u->bu_measured.b_time = delta;
2044     u->bu_measured.b_packets = x->bm_measured.b_packets;
2045     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2046     u->bu_flags = 0;
2047     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2048 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2049     if (x->bm_flags & BW_METER_UNIT_BYTES)
2050 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2051     if (x->bm_flags & BW_METER_GEQ)
2052 	u->bu_flags |= BW_UPCALL_GEQ;
2053     if (x->bm_flags & BW_METER_LEQ)
2054 	u->bu_flags |= BW_UPCALL_LEQ;
2055 }
2056 
2057 /*
2058  * Send the pending bandwidth-related upcalls
2059  */
2060 static void
2061 bw_upcalls_send(void)
2062 {
2063     struct mbuf *m;
2064     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2065     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2066     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2067 				      0,		/* unused2 */
2068 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2069 				      0,		/* im_mbz  */
2070 				      0,		/* im_vif  */
2071 				      0,		/* unused3 */
2072 				      { 0 },		/* im_src  */
2073 				      { 0 } };		/* im_dst  */
2074 
2075     MFC_LOCK_ASSERT();
2076 
2077     if (V_bw_upcalls_n == 0)
2078 	return;			/* No pending upcalls */
2079 
2080     V_bw_upcalls_n = 0;
2081 
2082     /*
2083      * Allocate a new mbuf, initialize it with the header and
2084      * the payload for the pending calls.
2085      */
2086     m = m_gethdr(M_NOWAIT, MT_DATA);
2087     if (m == NULL) {
2088 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2089 	return;
2090     }
2091 
2092     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2093     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2094 
2095     /*
2096      * Send the upcalls
2097      * XXX do we need to set the address in k_igmpsrc ?
2098      */
2099     MRTSTAT_INC(mrts_upcalls);
2100     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2101 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2102 	MRTSTAT_INC(mrts_upq_sockfull);
2103     }
2104 }
2105 
2106 /*
2107  * Compute the timeout hash value for the bw_meter entries
2108  */
2109 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2110     do {								\
2111 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2112 									\
2113 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2114 	(hash) = next_timeval.tv_sec;					\
2115 	if (next_timeval.tv_usec)					\
2116 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2117 	(hash) %= BW_METER_BUCKETS;					\
2118     } while (0)
2119 
2120 /*
2121  * Schedule a timer to process periodically bw_meter entry of type "<="
2122  * by linking the entry in the proper hash bucket.
2123  */
2124 static void
2125 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2126 {
2127     int time_hash;
2128 
2129     MFC_LOCK_ASSERT();
2130 
2131     if (!(x->bm_flags & BW_METER_LEQ))
2132 	return;		/* XXX: we schedule timers only for "<=" entries */
2133 
2134     /*
2135      * Reset the bw_meter entry
2136      */
2137     x->bm_start_time = *nowp;
2138     x->bm_measured.b_packets = 0;
2139     x->bm_measured.b_bytes = 0;
2140     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2141 
2142     /*
2143      * Compute the timeout hash value and insert the entry
2144      */
2145     BW_METER_TIMEHASH(x, time_hash);
2146     x->bm_time_next = V_bw_meter_timers[time_hash];
2147     V_bw_meter_timers[time_hash] = x;
2148     x->bm_time_hash = time_hash;
2149 }
2150 
2151 /*
2152  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2153  * by removing the entry from the proper hash bucket.
2154  */
2155 static void
2156 unschedule_bw_meter(struct bw_meter *x)
2157 {
2158     int time_hash;
2159     struct bw_meter *prev, *tmp;
2160 
2161     MFC_LOCK_ASSERT();
2162 
2163     if (!(x->bm_flags & BW_METER_LEQ))
2164 	return;		/* XXX: we schedule timers only for "<=" entries */
2165 
2166     /*
2167      * Compute the timeout hash value and delete the entry
2168      */
2169     time_hash = x->bm_time_hash;
2170     if (time_hash >= BW_METER_BUCKETS)
2171 	return;		/* Entry was not scheduled */
2172 
2173     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2174 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2175 	if (tmp == x)
2176 	    break;
2177 
2178     if (tmp == NULL)
2179 	panic("unschedule_bw_meter: bw_meter entry not found");
2180 
2181     if (prev != NULL)
2182 	prev->bm_time_next = x->bm_time_next;
2183     else
2184 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2185 
2186     x->bm_time_next = NULL;
2187     x->bm_time_hash = BW_METER_BUCKETS;
2188 }
2189 
2190 
2191 /*
2192  * Process all "<=" type of bw_meter that should be processed now,
2193  * and for each entry prepare an upcall if necessary. Each processed
2194  * entry is rescheduled again for the (periodic) processing.
2195  *
2196  * This is run periodically (once per second normally). On each round,
2197  * all the potentially matching entries are in the hash slot that we are
2198  * looking at.
2199  */
2200 static void
2201 bw_meter_process()
2202 {
2203     uint32_t loops;
2204     int i;
2205     struct timeval now, process_endtime;
2206 
2207     microtime(&now);
2208     if (V_last_tv_sec == now.tv_sec)
2209 	return;		/* nothing to do */
2210 
2211     loops = now.tv_sec - V_last_tv_sec;
2212     V_last_tv_sec = now.tv_sec;
2213     if (loops > BW_METER_BUCKETS)
2214 	loops = BW_METER_BUCKETS;
2215 
2216     MFC_LOCK();
2217     /*
2218      * Process all bins of bw_meter entries from the one after the last
2219      * processed to the current one. On entry, i points to the last bucket
2220      * visited, so we need to increment i at the beginning of the loop.
2221      */
2222     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2223 	struct bw_meter *x, *tmp_list;
2224 
2225 	if (++i >= BW_METER_BUCKETS)
2226 	    i = 0;
2227 
2228 	/* Disconnect the list of bw_meter entries from the bin */
2229 	tmp_list = V_bw_meter_timers[i];
2230 	V_bw_meter_timers[i] = NULL;
2231 
2232 	/* Process the list of bw_meter entries */
2233 	while (tmp_list != NULL) {
2234 	    x = tmp_list;
2235 	    tmp_list = tmp_list->bm_time_next;
2236 
2237 	    /* Test if the time interval is over */
2238 	    process_endtime = x->bm_start_time;
2239 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2240 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2241 		/* Not yet: reschedule, but don't reset */
2242 		int time_hash;
2243 
2244 		BW_METER_TIMEHASH(x, time_hash);
2245 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2246 		    /*
2247 		     * XXX: somehow the bin processing is a bit ahead of time.
2248 		     * Put the entry in the next bin.
2249 		     */
2250 		    if (++time_hash >= BW_METER_BUCKETS)
2251 			time_hash = 0;
2252 		}
2253 		x->bm_time_next = V_bw_meter_timers[time_hash];
2254 		V_bw_meter_timers[time_hash] = x;
2255 		x->bm_time_hash = time_hash;
2256 
2257 		continue;
2258 	    }
2259 
2260 	    /*
2261 	     * Test if we should deliver an upcall
2262 	     */
2263 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2264 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2265 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2266 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2267 		/* Prepare an upcall for delivery */
2268 		bw_meter_prepare_upcall(x, &now);
2269 	    }
2270 
2271 	    /*
2272 	     * Reschedule for next processing
2273 	     */
2274 	    schedule_bw_meter(x, &now);
2275 	}
2276     }
2277 
2278     /* Send all upcalls that are pending delivery */
2279     bw_upcalls_send();
2280 
2281     MFC_UNLOCK();
2282 }
2283 
2284 /*
2285  * A periodic function for sending all upcalls that are pending delivery
2286  */
2287 static void
2288 expire_bw_upcalls_send(void *arg)
2289 {
2290     CURVNET_SET((struct vnet *) arg);
2291 
2292     MFC_LOCK();
2293     bw_upcalls_send();
2294     MFC_UNLOCK();
2295 
2296     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2297 	curvnet);
2298     CURVNET_RESTORE();
2299 }
2300 
2301 /*
2302  * A periodic function for periodic scanning of the multicast forwarding
2303  * table for processing all "<=" bw_meter entries.
2304  */
2305 static void
2306 expire_bw_meter_process(void *arg)
2307 {
2308     CURVNET_SET((struct vnet *) arg);
2309 
2310     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2311 	bw_meter_process();
2312 
2313     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2314 	curvnet);
2315     CURVNET_RESTORE();
2316 }
2317 
2318 /*
2319  * End of bandwidth monitoring code
2320  */
2321 
2322 /*
2323  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2324  *
2325  */
2326 static int
2327 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2328     struct mfc *rt)
2329 {
2330     struct mbuf *mb_copy, *mm;
2331 
2332     /*
2333      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2334      * rendezvous point was unspecified, and we were told not to.
2335      */
2336     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2337 	in_nullhost(rt->mfc_rp))
2338 	return 0;
2339 
2340     mb_copy = pim_register_prepare(ip, m);
2341     if (mb_copy == NULL)
2342 	return ENOBUFS;
2343 
2344     /*
2345      * Send all the fragments. Note that the mbuf for each fragment
2346      * is freed by the sending machinery.
2347      */
2348     for (mm = mb_copy; mm; mm = mb_copy) {
2349 	mb_copy = mm->m_nextpkt;
2350 	mm->m_nextpkt = 0;
2351 	mm = m_pullup(mm, sizeof(struct ip));
2352 	if (mm != NULL) {
2353 	    ip = mtod(mm, struct ip *);
2354 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2355 		pim_register_send_rp(ip, vifp, mm, rt);
2356 	    } else {
2357 		pim_register_send_upcall(ip, vifp, mm, rt);
2358 	    }
2359 	}
2360     }
2361 
2362     return 0;
2363 }
2364 
2365 /*
2366  * Return a copy of the data packet that is ready for PIM Register
2367  * encapsulation.
2368  * XXX: Note that in the returned copy the IP header is a valid one.
2369  */
2370 static struct mbuf *
2371 pim_register_prepare(struct ip *ip, struct mbuf *m)
2372 {
2373     struct mbuf *mb_copy = NULL;
2374     int mtu;
2375 
2376     /* Take care of delayed checksums */
2377     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2378 	in_delayed_cksum(m);
2379 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2380     }
2381 
2382     /*
2383      * Copy the old packet & pullup its IP header into the
2384      * new mbuf so we can modify it.
2385      */
2386     mb_copy = m_copypacket(m, M_NOWAIT);
2387     if (mb_copy == NULL)
2388 	return NULL;
2389     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2390     if (mb_copy == NULL)
2391 	return NULL;
2392 
2393     /* take care of the TTL */
2394     ip = mtod(mb_copy, struct ip *);
2395     --ip->ip_ttl;
2396 
2397     /* Compute the MTU after the PIM Register encapsulation */
2398     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2399 
2400     if (ntohs(ip->ip_len) <= mtu) {
2401 	/* Turn the IP header into a valid one */
2402 	ip->ip_sum = 0;
2403 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2404     } else {
2405 	/* Fragment the packet */
2406 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2407 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2408 	    m_freem(mb_copy);
2409 	    return NULL;
2410 	}
2411     }
2412     return mb_copy;
2413 }
2414 
2415 /*
2416  * Send an upcall with the data packet to the user-level process.
2417  */
2418 static int
2419 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2420     struct mbuf *mb_copy, struct mfc *rt)
2421 {
2422     struct mbuf *mb_first;
2423     int len = ntohs(ip->ip_len);
2424     struct igmpmsg *im;
2425     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2426 
2427     VIF_LOCK_ASSERT();
2428 
2429     /*
2430      * Add a new mbuf with an upcall header
2431      */
2432     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2433     if (mb_first == NULL) {
2434 	m_freem(mb_copy);
2435 	return ENOBUFS;
2436     }
2437     mb_first->m_data += max_linkhdr;
2438     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2439     mb_first->m_len = sizeof(struct igmpmsg);
2440     mb_first->m_next = mb_copy;
2441 
2442     /* Send message to routing daemon */
2443     im = mtod(mb_first, struct igmpmsg *);
2444     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2445     im->im_mbz		= 0;
2446     im->im_vif		= vifp - V_viftable;
2447     im->im_src		= ip->ip_src;
2448     im->im_dst		= ip->ip_dst;
2449 
2450     k_igmpsrc.sin_addr	= ip->ip_src;
2451 
2452     MRTSTAT_INC(mrts_upcalls);
2453 
2454     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2455 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2456 	MRTSTAT_INC(mrts_upq_sockfull);
2457 	return ENOBUFS;
2458     }
2459 
2460     /* Keep statistics */
2461     PIMSTAT_INC(pims_snd_registers_msgs);
2462     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2463 
2464     return 0;
2465 }
2466 
2467 /*
2468  * Encapsulate the data packet in PIM Register message and send it to the RP.
2469  */
2470 static int
2471 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2472     struct mfc *rt)
2473 {
2474     struct mbuf *mb_first;
2475     struct ip *ip_outer;
2476     struct pim_encap_pimhdr *pimhdr;
2477     int len = ntohs(ip->ip_len);
2478     vifi_t vifi = rt->mfc_parent;
2479 
2480     VIF_LOCK_ASSERT();
2481 
2482     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2483 	m_freem(mb_copy);
2484 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2485     }
2486 
2487     /*
2488      * Add a new mbuf with the encapsulating header
2489      */
2490     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2491     if (mb_first == NULL) {
2492 	m_freem(mb_copy);
2493 	return ENOBUFS;
2494     }
2495     mb_first->m_data += max_linkhdr;
2496     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2497     mb_first->m_next = mb_copy;
2498 
2499     mb_first->m_pkthdr.len = len + mb_first->m_len;
2500 
2501     /*
2502      * Fill in the encapsulating IP and PIM header
2503      */
2504     ip_outer = mtod(mb_first, struct ip *);
2505     *ip_outer = pim_encap_iphdr;
2506     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2507 	sizeof(pim_encap_pimhdr));
2508     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2509     ip_outer->ip_dst = rt->mfc_rp;
2510     /*
2511      * Copy the inner header TOS to the outer header, and take care of the
2512      * IP_DF bit.
2513      */
2514     ip_outer->ip_tos = ip->ip_tos;
2515     if (ip->ip_off & htons(IP_DF))
2516 	ip_outer->ip_off |= htons(IP_DF);
2517     ip_fillid(ip_outer);
2518     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2519 					 + sizeof(pim_encap_iphdr));
2520     *pimhdr = pim_encap_pimhdr;
2521     /* If the iif crosses a border, set the Border-bit */
2522     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2523 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2524 
2525     mb_first->m_data += sizeof(pim_encap_iphdr);
2526     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2527     mb_first->m_data -= sizeof(pim_encap_iphdr);
2528 
2529     send_packet(vifp, mb_first);
2530 
2531     /* Keep statistics */
2532     PIMSTAT_INC(pims_snd_registers_msgs);
2533     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2534 
2535     return 0;
2536 }
2537 
2538 /*
2539  * pim_encapcheck() is called by the encap4_input() path at runtime to
2540  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2541  * into the kernel.
2542  */
2543 static int
2544 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2545     int proto __unused, void *arg __unused)
2546 {
2547 
2548     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2549     return (8);		/* claim the datagram. */
2550 }
2551 
2552 /*
2553  * PIM-SMv2 and PIM-DM messages processing.
2554  * Receives and verifies the PIM control messages, and passes them
2555  * up to the listening socket, using rip_input().
2556  * The only message with special processing is the PIM_REGISTER message
2557  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2558  * is passed to if_simloop().
2559  */
2560 static int
2561 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2562 {
2563     struct ip *ip = mtod(m, struct ip *);
2564     struct pim *pim;
2565     int iphlen = off;
2566     int minlen;
2567     int datalen = ntohs(ip->ip_len) - iphlen;
2568     int ip_tos;
2569 
2570     /* Keep statistics */
2571     PIMSTAT_INC(pims_rcv_total_msgs);
2572     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2573 
2574     /*
2575      * Validate lengths
2576      */
2577     if (datalen < PIM_MINLEN) {
2578 	PIMSTAT_INC(pims_rcv_tooshort);
2579 	CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2580 	    __func__, datalen, ntohl(ip->ip_src.s_addr));
2581 	m_freem(m);
2582 	return (IPPROTO_DONE);
2583     }
2584 
2585     /*
2586      * If the packet is at least as big as a REGISTER, go agead
2587      * and grab the PIM REGISTER header size, to avoid another
2588      * possible m_pullup() later.
2589      *
2590      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2591      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2592      */
2593     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2594     /*
2595      * Get the IP and PIM headers in contiguous memory, and
2596      * possibly the PIM REGISTER header.
2597      */
2598     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2599 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2600 	return (IPPROTO_DONE);
2601     }
2602 
2603     /* m_pullup() may have given us a new mbuf so reset ip. */
2604     ip = mtod(m, struct ip *);
2605     ip_tos = ip->ip_tos;
2606 
2607     /* adjust mbuf to point to the PIM header */
2608     m->m_data += iphlen;
2609     m->m_len  -= iphlen;
2610     pim = mtod(m, struct pim *);
2611 
2612     /*
2613      * Validate checksum. If PIM REGISTER, exclude the data packet.
2614      *
2615      * XXX: some older PIMv2 implementations don't make this distinction,
2616      * so for compatibility reason perform the checksum over part of the
2617      * message, and if error, then over the whole message.
2618      */
2619     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2620 	/* do nothing, checksum okay */
2621     } else if (in_cksum(m, datalen)) {
2622 	PIMSTAT_INC(pims_rcv_badsum);
2623 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2624 	m_freem(m);
2625 	return (IPPROTO_DONE);
2626     }
2627 
2628     /* PIM version check */
2629     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2630 	PIMSTAT_INC(pims_rcv_badversion);
2631 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2632 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2633 	m_freem(m);
2634 	return (IPPROTO_DONE);
2635     }
2636 
2637     /* restore mbuf back to the outer IP */
2638     m->m_data -= iphlen;
2639     m->m_len  += iphlen;
2640 
2641     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2642 	/*
2643 	 * Since this is a REGISTER, we'll make a copy of the register
2644 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2645 	 * routing daemon.
2646 	 */
2647 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2648 	struct mbuf *mcp;
2649 	struct ip *encap_ip;
2650 	u_int32_t *reghdr;
2651 	struct ifnet *vifp;
2652 
2653 	VIF_LOCK();
2654 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2655 	    VIF_UNLOCK();
2656 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2657 		(int)V_reg_vif_num);
2658 	    m_freem(m);
2659 	    return (IPPROTO_DONE);
2660 	}
2661 	/* XXX need refcnt? */
2662 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2663 	VIF_UNLOCK();
2664 
2665 	/*
2666 	 * Validate length
2667 	 */
2668 	if (datalen < PIM_REG_MINLEN) {
2669 	    PIMSTAT_INC(pims_rcv_tooshort);
2670 	    PIMSTAT_INC(pims_rcv_badregisters);
2671 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2672 	    m_freem(m);
2673 	    return (IPPROTO_DONE);
2674 	}
2675 
2676 	reghdr = (u_int32_t *)(pim + 1);
2677 	encap_ip = (struct ip *)(reghdr + 1);
2678 
2679 	CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2680 	    __func__, ntohl(encap_ip->ip_src.s_addr),
2681 	    ntohs(encap_ip->ip_len));
2682 
2683 	/* verify the version number of the inner packet */
2684 	if (encap_ip->ip_v != IPVERSION) {
2685 	    PIMSTAT_INC(pims_rcv_badregisters);
2686 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2687 	    m_freem(m);
2688 	    return (IPPROTO_DONE);
2689 	}
2690 
2691 	/* verify the inner packet is destined to a mcast group */
2692 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2693 	    PIMSTAT_INC(pims_rcv_badregisters);
2694 	    CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2695 		ntohl(encap_ip->ip_dst.s_addr));
2696 	    m_freem(m);
2697 	    return (IPPROTO_DONE);
2698 	}
2699 
2700 	/* If a NULL_REGISTER, pass it to the daemon */
2701 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2702 	    goto pim_input_to_daemon;
2703 
2704 	/*
2705 	 * Copy the TOS from the outer IP header to the inner IP header.
2706 	 */
2707 	if (encap_ip->ip_tos != ip_tos) {
2708 	    /* Outer TOS -> inner TOS */
2709 	    encap_ip->ip_tos = ip_tos;
2710 	    /* Recompute the inner header checksum. Sigh... */
2711 
2712 	    /* adjust mbuf to point to the inner IP header */
2713 	    m->m_data += (iphlen + PIM_MINLEN);
2714 	    m->m_len  -= (iphlen + PIM_MINLEN);
2715 
2716 	    encap_ip->ip_sum = 0;
2717 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2718 
2719 	    /* restore mbuf to point back to the outer IP header */
2720 	    m->m_data -= (iphlen + PIM_MINLEN);
2721 	    m->m_len  += (iphlen + PIM_MINLEN);
2722 	}
2723 
2724 	/*
2725 	 * Decapsulate the inner IP packet and loopback to forward it
2726 	 * as a normal multicast packet. Also, make a copy of the
2727 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2728 	 * to pass to the daemon later, so it can take the appropriate
2729 	 * actions (e.g., send back PIM_REGISTER_STOP).
2730 	 * XXX: here m->m_data points to the outer IP header.
2731 	 */
2732 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2733 	if (mcp == NULL) {
2734 	    CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2735 	    m_freem(m);
2736 	    return (IPPROTO_DONE);
2737 	}
2738 
2739 	/* Keep statistics */
2740 	/* XXX: registers_bytes include only the encap. mcast pkt */
2741 	PIMSTAT_INC(pims_rcv_registers_msgs);
2742 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2743 
2744 	/*
2745 	 * forward the inner ip packet; point m_data at the inner ip.
2746 	 */
2747 	m_adj(m, iphlen + PIM_MINLEN);
2748 
2749 	CTR4(KTR_IPMF,
2750 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2751 	    __func__,
2752 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2753 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2754 	    (int)V_reg_vif_num);
2755 
2756 	/* NB: vifp was collected above; can it change on us? */
2757 	if_simloop(vifp, m, dst.sin_family, 0);
2758 
2759 	/* prepare the register head to send to the mrouting daemon */
2760 	m = mcp;
2761     }
2762 
2763 pim_input_to_daemon:
2764     /*
2765      * Pass the PIM message up to the daemon; if it is a Register message,
2766      * pass the 'head' only up to the daemon. This includes the
2767      * outer IP header, PIM header, PIM-Register header and the
2768      * inner IP header.
2769      * XXX: the outer IP header pkt size of a Register is not adjust to
2770      * reflect the fact that the inner multicast data is truncated.
2771      */
2772     return (rip_input(&m, &off, proto));
2773 }
2774 
2775 static int
2776 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2777 {
2778 	struct mfc	*rt;
2779 	int		 error, i;
2780 
2781 	if (req->newptr)
2782 		return (EPERM);
2783 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2784 		return (0);
2785 	error = sysctl_wire_old_buffer(req, 0);
2786 	if (error)
2787 		return (error);
2788 
2789 	MFC_LOCK();
2790 	for (i = 0; i < mfchashsize; i++) {
2791 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2792 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2793 			if (error)
2794 				goto out_locked;
2795 		}
2796 	}
2797 out_locked:
2798 	MFC_UNLOCK();
2799 	return (error);
2800 }
2801 
2802 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2803     sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2804     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2805 
2806 static void
2807 vnet_mroute_init(const void *unused __unused)
2808 {
2809 
2810 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2811 	bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2812 	callout_init(&V_expire_upcalls_ch, 1);
2813 	callout_init(&V_bw_upcalls_ch, 1);
2814 	callout_init(&V_bw_meter_ch, 1);
2815 }
2816 
2817 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2818 	NULL);
2819 
2820 static void
2821 vnet_mroute_uninit(const void *unused __unused)
2822 {
2823 
2824 	free(V_nexpire, M_MRTABLE);
2825 	V_nexpire = NULL;
2826 }
2827 
2828 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2829 	vnet_mroute_uninit, NULL);
2830 
2831 static int
2832 ip_mroute_modevent(module_t mod, int type, void *unused)
2833 {
2834 
2835     switch (type) {
2836     case MOD_LOAD:
2837 	MROUTER_LOCK_INIT();
2838 
2839 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2840 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2841 	if (if_detach_event_tag == NULL) {
2842 		printf("ip_mroute: unable to register "
2843 		    "ifnet_departure_event handler\n");
2844 		MROUTER_LOCK_DESTROY();
2845 		return (EINVAL);
2846 	}
2847 
2848 	MFC_LOCK_INIT();
2849 	VIF_LOCK_INIT();
2850 
2851 	mfchashsize = MFCHASHSIZE;
2852 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2853 	    !powerof2(mfchashsize)) {
2854 		printf("WARNING: %s not a power of 2; using default\n",
2855 		    "net.inet.ip.mfchashsize");
2856 		mfchashsize = MFCHASHSIZE;
2857 	}
2858 
2859 	pim_squelch_wholepkt = 0;
2860 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2861 	    &pim_squelch_wholepkt);
2862 
2863 	pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2864 	if (pim_encap_cookie == NULL) {
2865 		printf("ip_mroute: unable to attach pim encap\n");
2866 		VIF_LOCK_DESTROY();
2867 		MFC_LOCK_DESTROY();
2868 		MROUTER_LOCK_DESTROY();
2869 		return (EINVAL);
2870 	}
2871 
2872 	ip_mcast_src = X_ip_mcast_src;
2873 	ip_mforward = X_ip_mforward;
2874 	ip_mrouter_done = X_ip_mrouter_done;
2875 	ip_mrouter_get = X_ip_mrouter_get;
2876 	ip_mrouter_set = X_ip_mrouter_set;
2877 
2878 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2879 	ip_rsvp_vif = X_ip_rsvp_vif;
2880 
2881 	legal_vif_num = X_legal_vif_num;
2882 	mrt_ioctl = X_mrt_ioctl;
2883 	rsvp_input_p = X_rsvp_input;
2884 	break;
2885 
2886     case MOD_UNLOAD:
2887 	/*
2888 	 * Typically module unload happens after the user-level
2889 	 * process has shutdown the kernel services (the check
2890 	 * below insures someone can't just yank the module out
2891 	 * from under a running process).  But if the module is
2892 	 * just loaded and then unloaded w/o starting up a user
2893 	 * process we still need to cleanup.
2894 	 */
2895 	MROUTER_LOCK();
2896 	if (ip_mrouter_cnt != 0) {
2897 	    MROUTER_UNLOCK();
2898 	    return (EINVAL);
2899 	}
2900 	ip_mrouter_unloading = 1;
2901 	MROUTER_UNLOCK();
2902 
2903 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2904 
2905 	if (pim_encap_cookie) {
2906 	    ip_encap_detach(pim_encap_cookie);
2907 	    pim_encap_cookie = NULL;
2908 	}
2909 
2910 	ip_mcast_src = NULL;
2911 	ip_mforward = NULL;
2912 	ip_mrouter_done = NULL;
2913 	ip_mrouter_get = NULL;
2914 	ip_mrouter_set = NULL;
2915 
2916 	ip_rsvp_force_done = NULL;
2917 	ip_rsvp_vif = NULL;
2918 
2919 	legal_vif_num = NULL;
2920 	mrt_ioctl = NULL;
2921 	rsvp_input_p = NULL;
2922 
2923 	VIF_LOCK_DESTROY();
2924 	MFC_LOCK_DESTROY();
2925 	MROUTER_LOCK_DESTROY();
2926 	break;
2927 
2928     default:
2929 	return EOPNOTSUPP;
2930     }
2931     return 0;
2932 }
2933 
2934 static moduledata_t ip_mroutemod = {
2935     "ip_mroute",
2936     ip_mroute_modevent,
2937     0
2938 };
2939 
2940 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2941