xref: /freebsd/sys/netinet/ip_mroute.c (revision eb69d1f144a6fcc765d1b9d44a5ae8082353e70b)
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 extern  struct domain inetdomain;
246 static const struct protosw in_pim_protosw = {
247 	.pr_type =		SOCK_RAW,
248 	.pr_domain =		&inetdomain,
249 	.pr_protocol =		IPPROTO_PIM,
250 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
251 	.pr_input =		pim_input,
252 	.pr_output =		rip_output,
253 	.pr_ctloutput =		rip_ctloutput,
254 	.pr_usrreqs =		&rip_usrreqs
255 };
256 static const struct encaptab *pim_encap_cookie;
257 
258 static int pim_encapcheck(const struct mbuf *, int, int, void *);
259 
260 /*
261  * Note: the PIM Register encapsulation adds the following in front of a
262  * data packet:
263  *
264  * struct pim_encap_hdr {
265  *    struct ip ip;
266  *    struct pim_encap_pimhdr  pim;
267  * }
268  *
269  */
270 
271 struct pim_encap_pimhdr {
272 	struct pim pim;
273 	uint32_t   flags;
274 };
275 #define		PIM_ENCAP_TTL	64
276 
277 static struct ip pim_encap_iphdr = {
278 #if BYTE_ORDER == LITTLE_ENDIAN
279 	sizeof(struct ip) >> 2,
280 	IPVERSION,
281 #else
282 	IPVERSION,
283 	sizeof(struct ip) >> 2,
284 #endif
285 	0,			/* tos */
286 	sizeof(struct ip),	/* total length */
287 	0,			/* id */
288 	0,			/* frag offset */
289 	PIM_ENCAP_TTL,
290 	IPPROTO_PIM,
291 	0,			/* checksum */
292 };
293 
294 static struct pim_encap_pimhdr pim_encap_pimhdr = {
295     {
296 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
297 	0,			/* reserved */
298 	0,			/* checksum */
299     },
300     0				/* flags */
301 };
302 
303 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
304 #define	V_reg_vif_num		VNET(reg_vif_num)
305 static VNET_DEFINE(struct ifnet, multicast_register_if);
306 #define	V_multicast_register_if	VNET(multicast_register_if)
307 
308 /*
309  * Private variables.
310  */
311 
312 static u_long	X_ip_mcast_src(int);
313 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
314 		    struct ip_moptions *);
315 static int	X_ip_mrouter_done(void);
316 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
317 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
318 static int	X_legal_vif_num(int);
319 static int	X_mrt_ioctl(u_long, caddr_t, int);
320 
321 static int	add_bw_upcall(struct bw_upcall *);
322 static int	add_mfc(struct mfcctl2 *);
323 static int	add_vif(struct vifctl *);
324 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
325 static void	bw_meter_process(void);
326 static void	bw_meter_receive_packet(struct bw_meter *, int,
327 		    struct timeval *);
328 static void	bw_upcalls_send(void);
329 static int	del_bw_upcall(struct bw_upcall *);
330 static int	del_mfc(struct mfcctl2 *);
331 static int	del_vif(vifi_t);
332 static int	del_vif_locked(vifi_t);
333 static void	expire_bw_meter_process(void *);
334 static void	expire_bw_upcalls_send(void *);
335 static void	expire_mfc(struct mfc *);
336 static void	expire_upcalls(void *);
337 static void	free_bw_list(struct bw_meter *);
338 static int	get_sg_cnt(struct sioc_sg_req *);
339 static int	get_vif_cnt(struct sioc_vif_req *);
340 static void	if_detached_event(void *, struct ifnet *);
341 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
342 static int	ip_mrouter_init(struct socket *, int);
343 static __inline struct mfc *
344 		mfc_find(struct in_addr *, struct in_addr *);
345 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
346 static struct mbuf *
347 		pim_register_prepare(struct ip *, struct mbuf *);
348 static int	pim_register_send(struct ip *, struct vif *,
349 		    struct mbuf *, struct mfc *);
350 static int	pim_register_send_rp(struct ip *, struct vif *,
351 		    struct mbuf *, struct mfc *);
352 static int	pim_register_send_upcall(struct ip *, struct vif *,
353 		    struct mbuf *, struct mfc *);
354 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
355 static void	send_packet(struct vif *, struct mbuf *);
356 static int	set_api_config(uint32_t *);
357 static int	set_assert(int);
358 static int	socket_send(struct socket *, struct mbuf *,
359 		    struct sockaddr_in *);
360 static void	unschedule_bw_meter(struct bw_meter *);
361 
362 /*
363  * Kernel multicast forwarding API capabilities and setup.
364  * If more API capabilities are added to the kernel, they should be
365  * recorded in `mrt_api_support'.
366  */
367 #define MRT_API_VERSION		0x0305
368 
369 static const int mrt_api_version = MRT_API_VERSION;
370 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
371 					 MRT_MFC_FLAGS_BORDER_VIF |
372 					 MRT_MFC_RP |
373 					 MRT_MFC_BW_UPCALL);
374 static VNET_DEFINE(uint32_t, mrt_api_config);
375 #define	V_mrt_api_config	VNET(mrt_api_config)
376 static VNET_DEFINE(int, pim_assert_enabled);
377 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
378 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
379 
380 /*
381  * Find a route for a given origin IP address and multicast group address.
382  * Statistics must be updated by the caller.
383  */
384 static __inline struct mfc *
385 mfc_find(struct in_addr *o, struct in_addr *g)
386 {
387 	struct mfc *rt;
388 
389 	MFC_LOCK_ASSERT();
390 
391 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
392 		if (in_hosteq(rt->mfc_origin, *o) &&
393 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
394 		    TAILQ_EMPTY(&rt->mfc_stall))
395 			break;
396 	}
397 
398 	return (rt);
399 }
400 
401 /*
402  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
403  */
404 static int
405 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
406 {
407     int	error, optval;
408     vifi_t	vifi;
409     struct	vifctl vifc;
410     struct	mfcctl2 mfc;
411     struct	bw_upcall bw_upcall;
412     uint32_t	i;
413 
414     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
415 	return EPERM;
416 
417     error = 0;
418     switch (sopt->sopt_name) {
419     case MRT_INIT:
420 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
421 	if (error)
422 	    break;
423 	error = ip_mrouter_init(so, optval);
424 	break;
425 
426     case MRT_DONE:
427 	error = ip_mrouter_done();
428 	break;
429 
430     case MRT_ADD_VIF:
431 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
432 	if (error)
433 	    break;
434 	error = add_vif(&vifc);
435 	break;
436 
437     case MRT_DEL_VIF:
438 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
439 	if (error)
440 	    break;
441 	error = del_vif(vifi);
442 	break;
443 
444     case MRT_ADD_MFC:
445     case MRT_DEL_MFC:
446 	/*
447 	 * select data size depending on API version.
448 	 */
449 	if (sopt->sopt_name == MRT_ADD_MFC &&
450 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
451 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
452 				sizeof(struct mfcctl2));
453 	} else {
454 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
455 				sizeof(struct mfcctl));
456 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
457 			sizeof(mfc) - sizeof(struct mfcctl));
458 	}
459 	if (error)
460 	    break;
461 	if (sopt->sopt_name == MRT_ADD_MFC)
462 	    error = add_mfc(&mfc);
463 	else
464 	    error = del_mfc(&mfc);
465 	break;
466 
467     case MRT_ASSERT:
468 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
469 	if (error)
470 	    break;
471 	set_assert(optval);
472 	break;
473 
474     case MRT_API_CONFIG:
475 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
476 	if (!error)
477 	    error = set_api_config(&i);
478 	if (!error)
479 	    error = sooptcopyout(sopt, &i, sizeof i);
480 	break;
481 
482     case MRT_ADD_BW_UPCALL:
483     case MRT_DEL_BW_UPCALL:
484 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
485 				sizeof bw_upcall);
486 	if (error)
487 	    break;
488 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
489 	    error = add_bw_upcall(&bw_upcall);
490 	else
491 	    error = del_bw_upcall(&bw_upcall);
492 	break;
493 
494     default:
495 	error = EOPNOTSUPP;
496 	break;
497     }
498     return error;
499 }
500 
501 /*
502  * Handle MRT getsockopt commands
503  */
504 static int
505 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
506 {
507     int error;
508 
509     switch (sopt->sopt_name) {
510     case MRT_VERSION:
511 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
512 	break;
513 
514     case MRT_ASSERT:
515 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
516 	    sizeof V_pim_assert_enabled);
517 	break;
518 
519     case MRT_API_SUPPORT:
520 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
521 	break;
522 
523     case MRT_API_CONFIG:
524 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
525 	break;
526 
527     default:
528 	error = EOPNOTSUPP;
529 	break;
530     }
531     return error;
532 }
533 
534 /*
535  * Handle ioctl commands to obtain information from the cache
536  */
537 static int
538 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
539 {
540     int error = 0;
541 
542     /*
543      * Currently the only function calling this ioctl routine is rtioctl_fib().
544      * Typically, only root can create the raw socket in order to execute
545      * this ioctl method, however the request might be coming from a prison
546      */
547     error = priv_check(curthread, PRIV_NETINET_MROUTE);
548     if (error)
549 	return (error);
550     switch (cmd) {
551     case (SIOCGETVIFCNT):
552 	error = get_vif_cnt((struct sioc_vif_req *)data);
553 	break;
554 
555     case (SIOCGETSGCNT):
556 	error = get_sg_cnt((struct sioc_sg_req *)data);
557 	break;
558 
559     default:
560 	error = EINVAL;
561 	break;
562     }
563     return error;
564 }
565 
566 /*
567  * returns the packet, byte, rpf-failure count for the source group provided
568  */
569 static int
570 get_sg_cnt(struct sioc_sg_req *req)
571 {
572     struct mfc *rt;
573 
574     MFC_LOCK();
575     rt = mfc_find(&req->src, &req->grp);
576     if (rt == NULL) {
577 	MFC_UNLOCK();
578 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
579 	return EADDRNOTAVAIL;
580     }
581     req->pktcnt = rt->mfc_pkt_cnt;
582     req->bytecnt = rt->mfc_byte_cnt;
583     req->wrong_if = rt->mfc_wrong_if;
584     MFC_UNLOCK();
585     return 0;
586 }
587 
588 /*
589  * returns the input and output packet and byte counts on the vif provided
590  */
591 static int
592 get_vif_cnt(struct sioc_vif_req *req)
593 {
594     vifi_t vifi = req->vifi;
595 
596     VIF_LOCK();
597     if (vifi >= V_numvifs) {
598 	VIF_UNLOCK();
599 	return EINVAL;
600     }
601 
602     req->icount = V_viftable[vifi].v_pkt_in;
603     req->ocount = V_viftable[vifi].v_pkt_out;
604     req->ibytes = V_viftable[vifi].v_bytes_in;
605     req->obytes = V_viftable[vifi].v_bytes_out;
606     VIF_UNLOCK();
607 
608     return 0;
609 }
610 
611 static void
612 if_detached_event(void *arg __unused, struct ifnet *ifp)
613 {
614     vifi_t vifi;
615     u_long i;
616 
617     MROUTER_LOCK();
618 
619     if (V_ip_mrouter == NULL) {
620 	MROUTER_UNLOCK();
621 	return;
622     }
623 
624     VIF_LOCK();
625     MFC_LOCK();
626 
627     /*
628      * Tear down multicast forwarder state associated with this ifnet.
629      * 1. Walk the vif list, matching vifs against this ifnet.
630      * 2. Walk the multicast forwarding cache (mfc) looking for
631      *    inner matches with this vif's index.
632      * 3. Expire any matching multicast forwarding cache entries.
633      * 4. Free vif state. This should disable ALLMULTI on the interface.
634      */
635     for (vifi = 0; vifi < V_numvifs; vifi++) {
636 	if (V_viftable[vifi].v_ifp != ifp)
637 		continue;
638 	for (i = 0; i < mfchashsize; i++) {
639 		struct mfc *rt, *nrt;
640 
641 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
642 			if (rt->mfc_parent == vifi) {
643 				expire_mfc(rt);
644 			}
645 		}
646 	}
647 	del_vif_locked(vifi);
648     }
649 
650     MFC_UNLOCK();
651     VIF_UNLOCK();
652 
653     MROUTER_UNLOCK();
654 }
655 
656 /*
657  * Enable multicast forwarding.
658  */
659 static int
660 ip_mrouter_init(struct socket *so, int version)
661 {
662 
663     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
664         so->so_type, so->so_proto->pr_protocol);
665 
666     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
667 	return EOPNOTSUPP;
668 
669     if (version != 1)
670 	return ENOPROTOOPT;
671 
672     MROUTER_LOCK();
673 
674     if (ip_mrouter_unloading) {
675 	MROUTER_UNLOCK();
676 	return ENOPROTOOPT;
677     }
678 
679     if (V_ip_mrouter != NULL) {
680 	MROUTER_UNLOCK();
681 	return EADDRINUSE;
682     }
683 
684     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
685 	HASH_NOWAIT);
686 
687     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
688 	curvnet);
689     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
690 	curvnet);
691     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
692 	curvnet);
693 
694     V_ip_mrouter = so;
695     ip_mrouter_cnt++;
696 
697     MROUTER_UNLOCK();
698 
699     CTR1(KTR_IPMF, "%s: done", __func__);
700 
701     return 0;
702 }
703 
704 /*
705  * Disable multicast forwarding.
706  */
707 static int
708 X_ip_mrouter_done(void)
709 {
710     struct ifnet *ifp;
711     u_long i;
712     vifi_t vifi;
713 
714     MROUTER_LOCK();
715 
716     if (V_ip_mrouter == NULL) {
717 	MROUTER_UNLOCK();
718 	return EINVAL;
719     }
720 
721     /*
722      * Detach/disable hooks to the reset of the system.
723      */
724     V_ip_mrouter = NULL;
725     ip_mrouter_cnt--;
726     V_mrt_api_config = 0;
727 
728     VIF_LOCK();
729 
730     /*
731      * For each phyint in use, disable promiscuous reception of all IP
732      * multicasts.
733      */
734     for (vifi = 0; vifi < V_numvifs; vifi++) {
735 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
736 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
737 	    ifp = V_viftable[vifi].v_ifp;
738 	    if_allmulti(ifp, 0);
739 	}
740     }
741     bzero((caddr_t)V_viftable, sizeof(V_viftable));
742     V_numvifs = 0;
743     V_pim_assert_enabled = 0;
744 
745     VIF_UNLOCK();
746 
747     callout_stop(&V_expire_upcalls_ch);
748     callout_stop(&V_bw_upcalls_ch);
749     callout_stop(&V_bw_meter_ch);
750 
751     MFC_LOCK();
752 
753     /*
754      * Free all multicast forwarding cache entries.
755      * Do not use hashdestroy(), as we must perform other cleanup.
756      */
757     for (i = 0; i < mfchashsize; i++) {
758 	struct mfc *rt, *nrt;
759 
760 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
761 		expire_mfc(rt);
762 	}
763     }
764     free(V_mfchashtbl, M_MRTABLE);
765     V_mfchashtbl = NULL;
766 
767     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
768 
769     V_bw_upcalls_n = 0;
770     bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
771 
772     MFC_UNLOCK();
773 
774     V_reg_vif_num = VIFI_INVALID;
775 
776     MROUTER_UNLOCK();
777 
778     CTR1(KTR_IPMF, "%s: done", __func__);
779 
780     return 0;
781 }
782 
783 /*
784  * Set PIM assert processing global
785  */
786 static int
787 set_assert(int i)
788 {
789     if ((i != 1) && (i != 0))
790 	return EINVAL;
791 
792     V_pim_assert_enabled = i;
793 
794     return 0;
795 }
796 
797 /*
798  * Configure API capabilities
799  */
800 int
801 set_api_config(uint32_t *apival)
802 {
803     u_long i;
804 
805     /*
806      * We can set the API capabilities only if it is the first operation
807      * after MRT_INIT. I.e.:
808      *  - there are no vifs installed
809      *  - pim_assert is not enabled
810      *  - the MFC table is empty
811      */
812     if (V_numvifs > 0) {
813 	*apival = 0;
814 	return EPERM;
815     }
816     if (V_pim_assert_enabled) {
817 	*apival = 0;
818 	return EPERM;
819     }
820 
821     MFC_LOCK();
822 
823     for (i = 0; i < mfchashsize; i++) {
824 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
825 	    MFC_UNLOCK();
826 	    *apival = 0;
827 	    return EPERM;
828 	}
829     }
830 
831     MFC_UNLOCK();
832 
833     V_mrt_api_config = *apival & mrt_api_support;
834     *apival = V_mrt_api_config;
835 
836     return 0;
837 }
838 
839 /*
840  * Add a vif to the vif table
841  */
842 static int
843 add_vif(struct vifctl *vifcp)
844 {
845     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
846     struct sockaddr_in sin = {sizeof sin, AF_INET};
847     struct ifaddr *ifa;
848     struct ifnet *ifp;
849     int error;
850 
851     VIF_LOCK();
852     if (vifcp->vifc_vifi >= MAXVIFS) {
853 	VIF_UNLOCK();
854 	return EINVAL;
855     }
856     /* rate limiting is no longer supported by this code */
857     if (vifcp->vifc_rate_limit != 0) {
858 	log(LOG_ERR, "rate limiting is no longer supported\n");
859 	VIF_UNLOCK();
860 	return EINVAL;
861     }
862     if (!in_nullhost(vifp->v_lcl_addr)) {
863 	VIF_UNLOCK();
864 	return EADDRINUSE;
865     }
866     if (in_nullhost(vifcp->vifc_lcl_addr)) {
867 	VIF_UNLOCK();
868 	return EADDRNOTAVAIL;
869     }
870 
871     /* Find the interface with an address in AF_INET family */
872     if (vifcp->vifc_flags & VIFF_REGISTER) {
873 	/*
874 	 * XXX: Because VIFF_REGISTER does not really need a valid
875 	 * local interface (e.g. it could be 127.0.0.2), we don't
876 	 * check its address.
877 	 */
878 	ifp = NULL;
879     } else {
880 	sin.sin_addr = vifcp->vifc_lcl_addr;
881 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
882 	if (ifa == NULL) {
883 	    VIF_UNLOCK();
884 	    return EADDRNOTAVAIL;
885 	}
886 	ifp = ifa->ifa_ifp;
887 	ifa_free(ifa);
888     }
889 
890     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
891 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
892 	VIF_UNLOCK();
893 	return EOPNOTSUPP;
894     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
895 	ifp = &V_multicast_register_if;
896 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
897 	if (V_reg_vif_num == VIFI_INVALID) {
898 	    if_initname(&V_multicast_register_if, "register_vif", 0);
899 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
900 	    V_reg_vif_num = vifcp->vifc_vifi;
901 	}
902     } else {		/* Make sure the interface supports multicast */
903 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
904 	    VIF_UNLOCK();
905 	    return EOPNOTSUPP;
906 	}
907 
908 	/* Enable promiscuous reception of all IP multicasts from the if */
909 	error = if_allmulti(ifp, 1);
910 	if (error) {
911 	    VIF_UNLOCK();
912 	    return error;
913 	}
914     }
915 
916     vifp->v_flags     = vifcp->vifc_flags;
917     vifp->v_threshold = vifcp->vifc_threshold;
918     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
919     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
920     vifp->v_ifp       = ifp;
921     /* initialize per vif pkt counters */
922     vifp->v_pkt_in    = 0;
923     vifp->v_pkt_out   = 0;
924     vifp->v_bytes_in  = 0;
925     vifp->v_bytes_out = 0;
926 
927     /* Adjust numvifs up if the vifi is higher than numvifs */
928     if (V_numvifs <= vifcp->vifc_vifi)
929 	V_numvifs = vifcp->vifc_vifi + 1;
930 
931     VIF_UNLOCK();
932 
933     CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
934 	(int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
935 	(int)vifcp->vifc_threshold);
936 
937     return 0;
938 }
939 
940 /*
941  * Delete a vif from the vif table
942  */
943 static int
944 del_vif_locked(vifi_t vifi)
945 {
946     struct vif *vifp;
947 
948     VIF_LOCK_ASSERT();
949 
950     if (vifi >= V_numvifs) {
951 	return EINVAL;
952     }
953     vifp = &V_viftable[vifi];
954     if (in_nullhost(vifp->v_lcl_addr)) {
955 	return EADDRNOTAVAIL;
956     }
957 
958     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
959 	if_allmulti(vifp->v_ifp, 0);
960 
961     if (vifp->v_flags & VIFF_REGISTER)
962 	V_reg_vif_num = VIFI_INVALID;
963 
964     bzero((caddr_t)vifp, sizeof (*vifp));
965 
966     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
967 
968     /* Adjust numvifs down */
969     for (vifi = V_numvifs; vifi > 0; vifi--)
970 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
971 	    break;
972     V_numvifs = vifi;
973 
974     return 0;
975 }
976 
977 static int
978 del_vif(vifi_t vifi)
979 {
980     int cc;
981 
982     VIF_LOCK();
983     cc = del_vif_locked(vifi);
984     VIF_UNLOCK();
985 
986     return cc;
987 }
988 
989 /*
990  * update an mfc entry without resetting counters and S,G addresses.
991  */
992 static void
993 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
994 {
995     int i;
996 
997     rt->mfc_parent = mfccp->mfcc_parent;
998     for (i = 0; i < V_numvifs; i++) {
999 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1000 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1001 	    MRT_MFC_FLAGS_ALL;
1002     }
1003     /* set the RP address */
1004     if (V_mrt_api_config & MRT_MFC_RP)
1005 	rt->mfc_rp = mfccp->mfcc_rp;
1006     else
1007 	rt->mfc_rp.s_addr = INADDR_ANY;
1008 }
1009 
1010 /*
1011  * fully initialize an mfc entry from the parameter.
1012  */
1013 static void
1014 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1015 {
1016     rt->mfc_origin     = mfccp->mfcc_origin;
1017     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1018 
1019     update_mfc_params(rt, mfccp);
1020 
1021     /* initialize pkt counters per src-grp */
1022     rt->mfc_pkt_cnt    = 0;
1023     rt->mfc_byte_cnt   = 0;
1024     rt->mfc_wrong_if   = 0;
1025     timevalclear(&rt->mfc_last_assert);
1026 }
1027 
1028 static void
1029 expire_mfc(struct mfc *rt)
1030 {
1031 	struct rtdetq *rte, *nrte;
1032 
1033 	MFC_LOCK_ASSERT();
1034 
1035 	free_bw_list(rt->mfc_bw_meter);
1036 
1037 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1038 		m_freem(rte->m);
1039 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1040 		free(rte, M_MRTABLE);
1041 	}
1042 
1043 	LIST_REMOVE(rt, mfc_hash);
1044 	free(rt, M_MRTABLE);
1045 }
1046 
1047 /*
1048  * Add an mfc entry
1049  */
1050 static int
1051 add_mfc(struct mfcctl2 *mfccp)
1052 {
1053     struct mfc *rt;
1054     struct rtdetq *rte, *nrte;
1055     u_long hash = 0;
1056     u_short nstl;
1057 
1058     VIF_LOCK();
1059     MFC_LOCK();
1060 
1061     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1062 
1063     /* If an entry already exists, just update the fields */
1064     if (rt) {
1065 	CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1066 	    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1067 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1068 	    mfccp->mfcc_parent);
1069 	update_mfc_params(rt, mfccp);
1070 	MFC_UNLOCK();
1071 	VIF_UNLOCK();
1072 	return (0);
1073     }
1074 
1075     /*
1076      * Find the entry for which the upcall was made and update
1077      */
1078     nstl = 0;
1079     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1080     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1081 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1082 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1083 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1084 		CTR5(KTR_IPMF,
1085 		    "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1086 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1087 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1088 		    mfccp->mfcc_parent,
1089 		    TAILQ_FIRST(&rt->mfc_stall));
1090 		if (nstl++)
1091 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1092 
1093 		init_mfc_params(rt, mfccp);
1094 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1095 		V_nexpire[hash]--;
1096 
1097 		/* Free queued packets, but attempt to forward them first. */
1098 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1099 			if (rte->ifp != NULL)
1100 				ip_mdq(rte->m, rte->ifp, rt, -1);
1101 			m_freem(rte->m);
1102 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1103 			rt->mfc_nstall--;
1104 			free(rte, M_MRTABLE);
1105 		}
1106 	}
1107     }
1108 
1109     /*
1110      * It is possible that an entry is being inserted without an upcall
1111      */
1112     if (nstl == 0) {
1113 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1114 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1115 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1116 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1117 			init_mfc_params(rt, mfccp);
1118 			if (rt->mfc_expire)
1119 			    V_nexpire[hash]--;
1120 			rt->mfc_expire = 0;
1121 			break; /* XXX */
1122 		}
1123 	}
1124 
1125 	if (rt == NULL) {		/* no upcall, so make a new entry */
1126 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1127 	    if (rt == NULL) {
1128 		MFC_UNLOCK();
1129 		VIF_UNLOCK();
1130 		return (ENOBUFS);
1131 	    }
1132 
1133 	    init_mfc_params(rt, mfccp);
1134 	    TAILQ_INIT(&rt->mfc_stall);
1135 	    rt->mfc_nstall = 0;
1136 
1137 	    rt->mfc_expire     = 0;
1138 	    rt->mfc_bw_meter = NULL;
1139 
1140 	    /* insert new entry at head of hash chain */
1141 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1142 	}
1143     }
1144 
1145     MFC_UNLOCK();
1146     VIF_UNLOCK();
1147 
1148     return (0);
1149 }
1150 
1151 /*
1152  * Delete an mfc entry
1153  */
1154 static int
1155 del_mfc(struct mfcctl2 *mfccp)
1156 {
1157     struct in_addr	origin;
1158     struct in_addr	mcastgrp;
1159     struct mfc		*rt;
1160 
1161     origin = mfccp->mfcc_origin;
1162     mcastgrp = mfccp->mfcc_mcastgrp;
1163 
1164     CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1165 	ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1166 
1167     MFC_LOCK();
1168 
1169     rt = mfc_find(&origin, &mcastgrp);
1170     if (rt == NULL) {
1171 	MFC_UNLOCK();
1172 	return EADDRNOTAVAIL;
1173     }
1174 
1175     /*
1176      * free the bw_meter entries
1177      */
1178     free_bw_list(rt->mfc_bw_meter);
1179     rt->mfc_bw_meter = NULL;
1180 
1181     LIST_REMOVE(rt, mfc_hash);
1182     free(rt, M_MRTABLE);
1183 
1184     MFC_UNLOCK();
1185 
1186     return (0);
1187 }
1188 
1189 /*
1190  * Send a message to the routing daemon on the multicast routing socket.
1191  */
1192 static int
1193 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1194 {
1195     if (s) {
1196 	SOCKBUF_LOCK(&s->so_rcv);
1197 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1198 	    NULL) != 0) {
1199 	    sorwakeup_locked(s);
1200 	    return 0;
1201 	}
1202 	SOCKBUF_UNLOCK(&s->so_rcv);
1203     }
1204     m_freem(mm);
1205     return -1;
1206 }
1207 
1208 /*
1209  * IP multicast forwarding function. This function assumes that the packet
1210  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1211  * pointed to by "ifp", and the packet is to be relayed to other networks
1212  * that have members of the packet's destination IP multicast group.
1213  *
1214  * The packet is returned unscathed to the caller, unless it is
1215  * erroneous, in which case a non-zero return value tells the caller to
1216  * discard it.
1217  */
1218 
1219 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1220 
1221 static int
1222 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1223     struct ip_moptions *imo)
1224 {
1225     struct mfc *rt;
1226     int error;
1227     vifi_t vifi;
1228 
1229     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1230 	ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1231 
1232     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1233 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1234 	/*
1235 	 * Packet arrived via a physical interface or
1236 	 * an encapsulated tunnel or a register_vif.
1237 	 */
1238     } else {
1239 	/*
1240 	 * Packet arrived through a source-route tunnel.
1241 	 * Source-route tunnels are no longer supported.
1242 	 */
1243 	return (1);
1244     }
1245 
1246     VIF_LOCK();
1247     MFC_LOCK();
1248     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1249 	if (ip->ip_ttl < MAXTTL)
1250 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1251 	error = ip_mdq(m, ifp, NULL, vifi);
1252 	MFC_UNLOCK();
1253 	VIF_UNLOCK();
1254 	return error;
1255     }
1256 
1257     /*
1258      * Don't forward a packet with time-to-live of zero or one,
1259      * or a packet destined to a local-only group.
1260      */
1261     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1262 	MFC_UNLOCK();
1263 	VIF_UNLOCK();
1264 	return 0;
1265     }
1266 
1267     /*
1268      * Determine forwarding vifs from the forwarding cache table
1269      */
1270     MRTSTAT_INC(mrts_mfc_lookups);
1271     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1272 
1273     /* Entry exists, so forward if necessary */
1274     if (rt != NULL) {
1275 	error = ip_mdq(m, ifp, rt, -1);
1276 	MFC_UNLOCK();
1277 	VIF_UNLOCK();
1278 	return error;
1279     } else {
1280 	/*
1281 	 * If we don't have a route for packet's origin,
1282 	 * Make a copy of the packet & send message to routing daemon
1283 	 */
1284 
1285 	struct mbuf *mb0;
1286 	struct rtdetq *rte;
1287 	u_long hash;
1288 	int hlen = ip->ip_hl << 2;
1289 
1290 	MRTSTAT_INC(mrts_mfc_misses);
1291 	MRTSTAT_INC(mrts_no_route);
1292 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1293 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1294 
1295 	/*
1296 	 * Allocate mbufs early so that we don't do extra work if we are
1297 	 * just going to fail anyway.  Make sure to pullup the header so
1298 	 * that other people can't step on it.
1299 	 */
1300 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1301 	    M_NOWAIT|M_ZERO);
1302 	if (rte == NULL) {
1303 	    MFC_UNLOCK();
1304 	    VIF_UNLOCK();
1305 	    return ENOBUFS;
1306 	}
1307 
1308 	mb0 = m_copypacket(m, M_NOWAIT);
1309 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1310 	    mb0 = m_pullup(mb0, hlen);
1311 	if (mb0 == NULL) {
1312 	    free(rte, M_MRTABLE);
1313 	    MFC_UNLOCK();
1314 	    VIF_UNLOCK();
1315 	    return ENOBUFS;
1316 	}
1317 
1318 	/* is there an upcall waiting for this flow ? */
1319 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1320 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1321 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1322 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1323 		    !TAILQ_EMPTY(&rt->mfc_stall))
1324 			break;
1325 	}
1326 
1327 	if (rt == NULL) {
1328 	    int i;
1329 	    struct igmpmsg *im;
1330 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1331 	    struct mbuf *mm;
1332 
1333 	    /*
1334 	     * Locate the vifi for the incoming interface for this packet.
1335 	     * If none found, drop packet.
1336 	     */
1337 	    for (vifi = 0; vifi < V_numvifs &&
1338 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1339 		;
1340 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1341 		goto non_fatal;
1342 
1343 	    /* no upcall, so make a new entry */
1344 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1345 	    if (rt == NULL)
1346 		goto fail;
1347 
1348 	    /* Make a copy of the header to send to the user level process */
1349 	    mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1350 	    if (mm == NULL)
1351 		goto fail1;
1352 
1353 	    /*
1354 	     * Send message to routing daemon to install
1355 	     * a route into the kernel table
1356 	     */
1357 
1358 	    im = mtod(mm, struct igmpmsg *);
1359 	    im->im_msgtype = IGMPMSG_NOCACHE;
1360 	    im->im_mbz = 0;
1361 	    im->im_vif = vifi;
1362 
1363 	    MRTSTAT_INC(mrts_upcalls);
1364 
1365 	    k_igmpsrc.sin_addr = ip->ip_src;
1366 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1367 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1368 		MRTSTAT_INC(mrts_upq_sockfull);
1369 fail1:
1370 		free(rt, M_MRTABLE);
1371 fail:
1372 		free(rte, M_MRTABLE);
1373 		m_freem(mb0);
1374 		MFC_UNLOCK();
1375 		VIF_UNLOCK();
1376 		return ENOBUFS;
1377 	    }
1378 
1379 	    /* insert new entry at head of hash chain */
1380 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1381 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1382 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1383 	    V_nexpire[hash]++;
1384 	    for (i = 0; i < V_numvifs; i++) {
1385 		rt->mfc_ttls[i] = 0;
1386 		rt->mfc_flags[i] = 0;
1387 	    }
1388 	    rt->mfc_parent = -1;
1389 
1390 	    /* clear the RP address */
1391 	    rt->mfc_rp.s_addr = INADDR_ANY;
1392 	    rt->mfc_bw_meter = NULL;
1393 
1394 	    /* initialize pkt counters per src-grp */
1395 	    rt->mfc_pkt_cnt = 0;
1396 	    rt->mfc_byte_cnt = 0;
1397 	    rt->mfc_wrong_if = 0;
1398 	    timevalclear(&rt->mfc_last_assert);
1399 
1400 	    TAILQ_INIT(&rt->mfc_stall);
1401 	    rt->mfc_nstall = 0;
1402 
1403 	    /* link into table */
1404 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1405 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1406 	    rt->mfc_nstall++;
1407 
1408 	} else {
1409 	    /* determine if queue has overflowed */
1410 	    if (rt->mfc_nstall > MAX_UPQ) {
1411 		MRTSTAT_INC(mrts_upq_ovflw);
1412 non_fatal:
1413 		free(rte, M_MRTABLE);
1414 		m_freem(mb0);
1415 		MFC_UNLOCK();
1416 		VIF_UNLOCK();
1417 		return (0);
1418 	    }
1419 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1420 	    rt->mfc_nstall++;
1421 	}
1422 
1423 	rte->m			= mb0;
1424 	rte->ifp		= ifp;
1425 
1426 	MFC_UNLOCK();
1427 	VIF_UNLOCK();
1428 
1429 	return 0;
1430     }
1431 }
1432 
1433 /*
1434  * Clean up the cache entry if upcall is not serviced
1435  */
1436 static void
1437 expire_upcalls(void *arg)
1438 {
1439     u_long i;
1440 
1441     CURVNET_SET((struct vnet *) arg);
1442 
1443     MFC_LOCK();
1444 
1445     for (i = 0; i < mfchashsize; i++) {
1446 	struct mfc *rt, *nrt;
1447 
1448 	if (V_nexpire[i] == 0)
1449 	    continue;
1450 
1451 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1452 		if (TAILQ_EMPTY(&rt->mfc_stall))
1453 			continue;
1454 
1455 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1456 			continue;
1457 
1458 		/*
1459 		 * free the bw_meter entries
1460 		 */
1461 		while (rt->mfc_bw_meter != NULL) {
1462 		    struct bw_meter *x = rt->mfc_bw_meter;
1463 
1464 		    rt->mfc_bw_meter = x->bm_mfc_next;
1465 		    free(x, M_BWMETER);
1466 		}
1467 
1468 		MRTSTAT_INC(mrts_cache_cleanups);
1469 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1470 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1471 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1472 
1473 		expire_mfc(rt);
1474 	    }
1475     }
1476 
1477     MFC_UNLOCK();
1478 
1479     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1480 	curvnet);
1481 
1482     CURVNET_RESTORE();
1483 }
1484 
1485 /*
1486  * Packet forwarding routine once entry in the cache is made
1487  */
1488 static int
1489 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1490 {
1491     struct ip  *ip = mtod(m, struct ip *);
1492     vifi_t vifi;
1493     int plen = ntohs(ip->ip_len);
1494 
1495     VIF_LOCK_ASSERT();
1496 
1497     /*
1498      * If xmt_vif is not -1, send on only the requested vif.
1499      *
1500      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1501      */
1502     if (xmt_vif < V_numvifs) {
1503 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1504 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1505 	else
1506 		phyint_send(ip, V_viftable + xmt_vif, m);
1507 	return 1;
1508     }
1509 
1510     /*
1511      * Don't forward if it didn't arrive from the parent vif for its origin.
1512      */
1513     vifi = rt->mfc_parent;
1514     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1515 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1516 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1517 	MRTSTAT_INC(mrts_wrong_if);
1518 	++rt->mfc_wrong_if;
1519 	/*
1520 	 * If we are doing PIM assert processing, send a message
1521 	 * to the routing daemon.
1522 	 *
1523 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1524 	 * can complete the SPT switch, regardless of the type
1525 	 * of the iif (broadcast media, GRE tunnel, etc).
1526 	 */
1527 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1528 	    V_viftable[vifi].v_ifp) {
1529 
1530 	    if (ifp == &V_multicast_register_if)
1531 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1532 
1533 	    /* Get vifi for the incoming packet */
1534 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1535 		vifi++)
1536 		;
1537 	    if (vifi >= V_numvifs)
1538 		return 0;	/* The iif is not found: ignore the packet. */
1539 
1540 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1541 		return 0;	/* WRONGVIF disabled: ignore the packet */
1542 
1543 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1544 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1545 		struct igmpmsg *im;
1546 		int hlen = ip->ip_hl << 2;
1547 		struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1548 
1549 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1550 		    mm = m_pullup(mm, hlen);
1551 		if (mm == NULL)
1552 		    return ENOBUFS;
1553 
1554 		im = mtod(mm, struct igmpmsg *);
1555 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1556 		im->im_mbz		= 0;
1557 		im->im_vif		= vifi;
1558 
1559 		MRTSTAT_INC(mrts_upcalls);
1560 
1561 		k_igmpsrc.sin_addr = im->im_src;
1562 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1563 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1564 		    MRTSTAT_INC(mrts_upq_sockfull);
1565 		    return ENOBUFS;
1566 		}
1567 	    }
1568 	}
1569 	return 0;
1570     }
1571 
1572 
1573     /* If I sourced this packet, it counts as output, else it was input. */
1574     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1575 	V_viftable[vifi].v_pkt_out++;
1576 	V_viftable[vifi].v_bytes_out += plen;
1577     } else {
1578 	V_viftable[vifi].v_pkt_in++;
1579 	V_viftable[vifi].v_bytes_in += plen;
1580     }
1581     rt->mfc_pkt_cnt++;
1582     rt->mfc_byte_cnt += plen;
1583 
1584     /*
1585      * For each vif, decide if a copy of the packet should be forwarded.
1586      * Forward if:
1587      *		- the ttl exceeds the vif's threshold
1588      *		- there are group members downstream on interface
1589      */
1590     for (vifi = 0; vifi < V_numvifs; vifi++)
1591 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1592 	    V_viftable[vifi].v_pkt_out++;
1593 	    V_viftable[vifi].v_bytes_out += plen;
1594 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1595 		pim_register_send(ip, V_viftable + vifi, m, rt);
1596 	    else
1597 		phyint_send(ip, V_viftable + vifi, m);
1598 	}
1599 
1600     /*
1601      * Perform upcall-related bw measuring.
1602      */
1603     if (rt->mfc_bw_meter != NULL) {
1604 	struct bw_meter *x;
1605 	struct timeval now;
1606 
1607 	microtime(&now);
1608 	MFC_LOCK_ASSERT();
1609 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1610 	    bw_meter_receive_packet(x, plen, &now);
1611     }
1612 
1613     return 0;
1614 }
1615 
1616 /*
1617  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1618  */
1619 static int
1620 X_legal_vif_num(int vif)
1621 {
1622 	int ret;
1623 
1624 	ret = 0;
1625 	if (vif < 0)
1626 		return (ret);
1627 
1628 	VIF_LOCK();
1629 	if (vif < V_numvifs)
1630 		ret = 1;
1631 	VIF_UNLOCK();
1632 
1633 	return (ret);
1634 }
1635 
1636 /*
1637  * Return the local address used by this vif
1638  */
1639 static u_long
1640 X_ip_mcast_src(int vifi)
1641 {
1642 	in_addr_t addr;
1643 
1644 	addr = INADDR_ANY;
1645 	if (vifi < 0)
1646 		return (addr);
1647 
1648 	VIF_LOCK();
1649 	if (vifi < V_numvifs)
1650 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1651 	VIF_UNLOCK();
1652 
1653 	return (addr);
1654 }
1655 
1656 static void
1657 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1658 {
1659     struct mbuf *mb_copy;
1660     int hlen = ip->ip_hl << 2;
1661 
1662     VIF_LOCK_ASSERT();
1663 
1664     /*
1665      * Make a new reference to the packet; make sure that
1666      * the IP header is actually copied, not just referenced,
1667      * so that ip_output() only scribbles on the copy.
1668      */
1669     mb_copy = m_copypacket(m, M_NOWAIT);
1670     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1671 	mb_copy = m_pullup(mb_copy, hlen);
1672     if (mb_copy == NULL)
1673 	return;
1674 
1675     send_packet(vifp, mb_copy);
1676 }
1677 
1678 static void
1679 send_packet(struct vif *vifp, struct mbuf *m)
1680 {
1681 	struct ip_moptions imo;
1682 	struct in_multi *imm[2];
1683 	int error;
1684 
1685 	VIF_LOCK_ASSERT();
1686 
1687 	imo.imo_multicast_ifp  = vifp->v_ifp;
1688 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1689 	imo.imo_multicast_loop = 1;
1690 	imo.imo_multicast_vif  = -1;
1691 	imo.imo_num_memberships = 0;
1692 	imo.imo_max_memberships = 2;
1693 	imo.imo_membership  = &imm[0];
1694 
1695 	/*
1696 	 * Re-entrancy should not be a problem here, because
1697 	 * the packets that we send out and are looped back at us
1698 	 * should get rejected because they appear to come from
1699 	 * the loopback interface, thus preventing looping.
1700 	 */
1701 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1702 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1703 	    (ptrdiff_t)(vifp - V_viftable), error);
1704 }
1705 
1706 /*
1707  * Stubs for old RSVP socket shim implementation.
1708  */
1709 
1710 static int
1711 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1712 {
1713 
1714 	return (EOPNOTSUPP);
1715 }
1716 
1717 static void
1718 X_ip_rsvp_force_done(struct socket *so __unused)
1719 {
1720 
1721 }
1722 
1723 static int
1724 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1725 {
1726 	struct mbuf *m;
1727 
1728 	m = *mp;
1729 	*mp = NULL;
1730 	if (!V_rsvp_on)
1731 		m_freem(m);
1732 	return (IPPROTO_DONE);
1733 }
1734 
1735 /*
1736  * Code for bandwidth monitors
1737  */
1738 
1739 /*
1740  * Define common interface for timeval-related methods
1741  */
1742 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1743 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1744 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1745 
1746 static uint32_t
1747 compute_bw_meter_flags(struct bw_upcall *req)
1748 {
1749     uint32_t flags = 0;
1750 
1751     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1752 	flags |= BW_METER_UNIT_PACKETS;
1753     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1754 	flags |= BW_METER_UNIT_BYTES;
1755     if (req->bu_flags & BW_UPCALL_GEQ)
1756 	flags |= BW_METER_GEQ;
1757     if (req->bu_flags & BW_UPCALL_LEQ)
1758 	flags |= BW_METER_LEQ;
1759 
1760     return flags;
1761 }
1762 
1763 /*
1764  * Add a bw_meter entry
1765  */
1766 static int
1767 add_bw_upcall(struct bw_upcall *req)
1768 {
1769     struct mfc *mfc;
1770     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1771 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1772     struct timeval now;
1773     struct bw_meter *x;
1774     uint32_t flags;
1775 
1776     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1777 	return EOPNOTSUPP;
1778 
1779     /* Test if the flags are valid */
1780     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1781 	return EINVAL;
1782     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1783 	return EINVAL;
1784     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1785 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1786 	return EINVAL;
1787 
1788     /* Test if the threshold time interval is valid */
1789     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1790 	return EINVAL;
1791 
1792     flags = compute_bw_meter_flags(req);
1793 
1794     /*
1795      * Find if we have already same bw_meter entry
1796      */
1797     MFC_LOCK();
1798     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1799     if (mfc == NULL) {
1800 	MFC_UNLOCK();
1801 	return EADDRNOTAVAIL;
1802     }
1803     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1804 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1805 			   &req->bu_threshold.b_time, ==)) &&
1806 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1807 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1808 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1809 	    MFC_UNLOCK();
1810 	    return 0;		/* XXX Already installed */
1811 	}
1812     }
1813 
1814     /* Allocate the new bw_meter entry */
1815     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1816     if (x == NULL) {
1817 	MFC_UNLOCK();
1818 	return ENOBUFS;
1819     }
1820 
1821     /* Set the new bw_meter entry */
1822     x->bm_threshold.b_time = req->bu_threshold.b_time;
1823     microtime(&now);
1824     x->bm_start_time = now;
1825     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1826     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1827     x->bm_measured.b_packets = 0;
1828     x->bm_measured.b_bytes = 0;
1829     x->bm_flags = flags;
1830     x->bm_time_next = NULL;
1831     x->bm_time_hash = BW_METER_BUCKETS;
1832 
1833     /* Add the new bw_meter entry to the front of entries for this MFC */
1834     x->bm_mfc = mfc;
1835     x->bm_mfc_next = mfc->mfc_bw_meter;
1836     mfc->mfc_bw_meter = x;
1837     schedule_bw_meter(x, &now);
1838     MFC_UNLOCK();
1839 
1840     return 0;
1841 }
1842 
1843 static void
1844 free_bw_list(struct bw_meter *list)
1845 {
1846     while (list != NULL) {
1847 	struct bw_meter *x = list;
1848 
1849 	list = list->bm_mfc_next;
1850 	unschedule_bw_meter(x);
1851 	free(x, M_BWMETER);
1852     }
1853 }
1854 
1855 /*
1856  * Delete one or multiple bw_meter entries
1857  */
1858 static int
1859 del_bw_upcall(struct bw_upcall *req)
1860 {
1861     struct mfc *mfc;
1862     struct bw_meter *x;
1863 
1864     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1865 	return EOPNOTSUPP;
1866 
1867     MFC_LOCK();
1868 
1869     /* Find the corresponding MFC entry */
1870     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1871     if (mfc == NULL) {
1872 	MFC_UNLOCK();
1873 	return EADDRNOTAVAIL;
1874     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1875 	/*
1876 	 * Delete all bw_meter entries for this mfc
1877 	 */
1878 	struct bw_meter *list;
1879 
1880 	list = mfc->mfc_bw_meter;
1881 	mfc->mfc_bw_meter = NULL;
1882 	free_bw_list(list);
1883 	MFC_UNLOCK();
1884 	return 0;
1885     } else {			/* Delete a single bw_meter entry */
1886 	struct bw_meter *prev;
1887 	uint32_t flags = 0;
1888 
1889 	flags = compute_bw_meter_flags(req);
1890 
1891 	/* Find the bw_meter entry to delete */
1892 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1893 	     prev = x, x = x->bm_mfc_next) {
1894 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1895 			       &req->bu_threshold.b_time, ==)) &&
1896 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1897 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1898 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1899 		break;
1900 	}
1901 	if (x != NULL) { /* Delete entry from the list for this MFC */
1902 	    if (prev != NULL)
1903 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1904 	    else
1905 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1906 
1907 	    unschedule_bw_meter(x);
1908 	    MFC_UNLOCK();
1909 	    /* Free the bw_meter entry */
1910 	    free(x, M_BWMETER);
1911 	    return 0;
1912 	} else {
1913 	    MFC_UNLOCK();
1914 	    return EINVAL;
1915 	}
1916     }
1917     /* NOTREACHED */
1918 }
1919 
1920 /*
1921  * Perform bandwidth measurement processing that may result in an upcall
1922  */
1923 static void
1924 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1925 {
1926     struct timeval delta;
1927 
1928     MFC_LOCK_ASSERT();
1929 
1930     delta = *nowp;
1931     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1932 
1933     if (x->bm_flags & BW_METER_GEQ) {
1934 	/*
1935 	 * Processing for ">=" type of bw_meter entry
1936 	 */
1937 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1938 	    /* Reset the bw_meter entry */
1939 	    x->bm_start_time = *nowp;
1940 	    x->bm_measured.b_packets = 0;
1941 	    x->bm_measured.b_bytes = 0;
1942 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1943 	}
1944 
1945 	/* Record that a packet is received */
1946 	x->bm_measured.b_packets++;
1947 	x->bm_measured.b_bytes += plen;
1948 
1949 	/*
1950 	 * Test if we should deliver an upcall
1951 	 */
1952 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1953 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1954 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1955 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1956 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1957 		/* Prepare an upcall for delivery */
1958 		bw_meter_prepare_upcall(x, nowp);
1959 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1960 	    }
1961 	}
1962     } else if (x->bm_flags & BW_METER_LEQ) {
1963 	/*
1964 	 * Processing for "<=" type of bw_meter entry
1965 	 */
1966 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1967 	    /*
1968 	     * We are behind time with the multicast forwarding table
1969 	     * scanning for "<=" type of bw_meter entries, so test now
1970 	     * if we should deliver an upcall.
1971 	     */
1972 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1973 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1974 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1975 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1976 		/* Prepare an upcall for delivery */
1977 		bw_meter_prepare_upcall(x, nowp);
1978 	    }
1979 	    /* Reschedule the bw_meter entry */
1980 	    unschedule_bw_meter(x);
1981 	    schedule_bw_meter(x, nowp);
1982 	}
1983 
1984 	/* Record that a packet is received */
1985 	x->bm_measured.b_packets++;
1986 	x->bm_measured.b_bytes += plen;
1987 
1988 	/*
1989 	 * Test if we should restart the measuring interval
1990 	 */
1991 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1992 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1993 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1994 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1995 	    /* Don't restart the measuring interval */
1996 	} else {
1997 	    /* Do restart the measuring interval */
1998 	    /*
1999 	     * XXX: note that we don't unschedule and schedule, because this
2000 	     * might be too much overhead per packet. Instead, when we process
2001 	     * all entries for a given timer hash bin, we check whether it is
2002 	     * really a timeout. If not, we reschedule at that time.
2003 	     */
2004 	    x->bm_start_time = *nowp;
2005 	    x->bm_measured.b_packets = 0;
2006 	    x->bm_measured.b_bytes = 0;
2007 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2008 	}
2009     }
2010 }
2011 
2012 /*
2013  * Prepare a bandwidth-related upcall
2014  */
2015 static void
2016 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2017 {
2018     struct timeval delta;
2019     struct bw_upcall *u;
2020 
2021     MFC_LOCK_ASSERT();
2022 
2023     /*
2024      * Compute the measured time interval
2025      */
2026     delta = *nowp;
2027     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2028 
2029     /*
2030      * If there are too many pending upcalls, deliver them now
2031      */
2032     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2033 	bw_upcalls_send();
2034 
2035     /*
2036      * Set the bw_upcall entry
2037      */
2038     u = &V_bw_upcalls[V_bw_upcalls_n++];
2039     u->bu_src = x->bm_mfc->mfc_origin;
2040     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2041     u->bu_threshold.b_time = x->bm_threshold.b_time;
2042     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2043     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2044     u->bu_measured.b_time = delta;
2045     u->bu_measured.b_packets = x->bm_measured.b_packets;
2046     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2047     u->bu_flags = 0;
2048     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2049 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2050     if (x->bm_flags & BW_METER_UNIT_BYTES)
2051 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2052     if (x->bm_flags & BW_METER_GEQ)
2053 	u->bu_flags |= BW_UPCALL_GEQ;
2054     if (x->bm_flags & BW_METER_LEQ)
2055 	u->bu_flags |= BW_UPCALL_LEQ;
2056 }
2057 
2058 /*
2059  * Send the pending bandwidth-related upcalls
2060  */
2061 static void
2062 bw_upcalls_send(void)
2063 {
2064     struct mbuf *m;
2065     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2066     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2067     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2068 				      0,		/* unused2 */
2069 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2070 				      0,		/* im_mbz  */
2071 				      0,		/* im_vif  */
2072 				      0,		/* unused3 */
2073 				      { 0 },		/* im_src  */
2074 				      { 0 } };		/* im_dst  */
2075 
2076     MFC_LOCK_ASSERT();
2077 
2078     if (V_bw_upcalls_n == 0)
2079 	return;			/* No pending upcalls */
2080 
2081     V_bw_upcalls_n = 0;
2082 
2083     /*
2084      * Allocate a new mbuf, initialize it with the header and
2085      * the payload for the pending calls.
2086      */
2087     m = m_gethdr(M_NOWAIT, MT_DATA);
2088     if (m == NULL) {
2089 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2090 	return;
2091     }
2092 
2093     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2094     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2095 
2096     /*
2097      * Send the upcalls
2098      * XXX do we need to set the address in k_igmpsrc ?
2099      */
2100     MRTSTAT_INC(mrts_upcalls);
2101     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2102 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2103 	MRTSTAT_INC(mrts_upq_sockfull);
2104     }
2105 }
2106 
2107 /*
2108  * Compute the timeout hash value for the bw_meter entries
2109  */
2110 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2111     do {								\
2112 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2113 									\
2114 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2115 	(hash) = next_timeval.tv_sec;					\
2116 	if (next_timeval.tv_usec)					\
2117 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2118 	(hash) %= BW_METER_BUCKETS;					\
2119     } while (0)
2120 
2121 /*
2122  * Schedule a timer to process periodically bw_meter entry of type "<="
2123  * by linking the entry in the proper hash bucket.
2124  */
2125 static void
2126 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2127 {
2128     int time_hash;
2129 
2130     MFC_LOCK_ASSERT();
2131 
2132     if (!(x->bm_flags & BW_METER_LEQ))
2133 	return;		/* XXX: we schedule timers only for "<=" entries */
2134 
2135     /*
2136      * Reset the bw_meter entry
2137      */
2138     x->bm_start_time = *nowp;
2139     x->bm_measured.b_packets = 0;
2140     x->bm_measured.b_bytes = 0;
2141     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2142 
2143     /*
2144      * Compute the timeout hash value and insert the entry
2145      */
2146     BW_METER_TIMEHASH(x, time_hash);
2147     x->bm_time_next = V_bw_meter_timers[time_hash];
2148     V_bw_meter_timers[time_hash] = x;
2149     x->bm_time_hash = time_hash;
2150 }
2151 
2152 /*
2153  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2154  * by removing the entry from the proper hash bucket.
2155  */
2156 static void
2157 unschedule_bw_meter(struct bw_meter *x)
2158 {
2159     int time_hash;
2160     struct bw_meter *prev, *tmp;
2161 
2162     MFC_LOCK_ASSERT();
2163 
2164     if (!(x->bm_flags & BW_METER_LEQ))
2165 	return;		/* XXX: we schedule timers only for "<=" entries */
2166 
2167     /*
2168      * Compute the timeout hash value and delete the entry
2169      */
2170     time_hash = x->bm_time_hash;
2171     if (time_hash >= BW_METER_BUCKETS)
2172 	return;		/* Entry was not scheduled */
2173 
2174     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2175 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2176 	if (tmp == x)
2177 	    break;
2178 
2179     if (tmp == NULL)
2180 	panic("unschedule_bw_meter: bw_meter entry not found");
2181 
2182     if (prev != NULL)
2183 	prev->bm_time_next = x->bm_time_next;
2184     else
2185 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2186 
2187     x->bm_time_next = NULL;
2188     x->bm_time_hash = BW_METER_BUCKETS;
2189 }
2190 
2191 
2192 /*
2193  * Process all "<=" type of bw_meter that should be processed now,
2194  * and for each entry prepare an upcall if necessary. Each processed
2195  * entry is rescheduled again for the (periodic) processing.
2196  *
2197  * This is run periodically (once per second normally). On each round,
2198  * all the potentially matching entries are in the hash slot that we are
2199  * looking at.
2200  */
2201 static void
2202 bw_meter_process()
2203 {
2204     uint32_t loops;
2205     int i;
2206     struct timeval now, process_endtime;
2207 
2208     microtime(&now);
2209     if (V_last_tv_sec == now.tv_sec)
2210 	return;		/* nothing to do */
2211 
2212     loops = now.tv_sec - V_last_tv_sec;
2213     V_last_tv_sec = now.tv_sec;
2214     if (loops > BW_METER_BUCKETS)
2215 	loops = BW_METER_BUCKETS;
2216 
2217     MFC_LOCK();
2218     /*
2219      * Process all bins of bw_meter entries from the one after the last
2220      * processed to the current one. On entry, i points to the last bucket
2221      * visited, so we need to increment i at the beginning of the loop.
2222      */
2223     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2224 	struct bw_meter *x, *tmp_list;
2225 
2226 	if (++i >= BW_METER_BUCKETS)
2227 	    i = 0;
2228 
2229 	/* Disconnect the list of bw_meter entries from the bin */
2230 	tmp_list = V_bw_meter_timers[i];
2231 	V_bw_meter_timers[i] = NULL;
2232 
2233 	/* Process the list of bw_meter entries */
2234 	while (tmp_list != NULL) {
2235 	    x = tmp_list;
2236 	    tmp_list = tmp_list->bm_time_next;
2237 
2238 	    /* Test if the time interval is over */
2239 	    process_endtime = x->bm_start_time;
2240 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2241 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2242 		/* Not yet: reschedule, but don't reset */
2243 		int time_hash;
2244 
2245 		BW_METER_TIMEHASH(x, time_hash);
2246 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2247 		    /*
2248 		     * XXX: somehow the bin processing is a bit ahead of time.
2249 		     * Put the entry in the next bin.
2250 		     */
2251 		    if (++time_hash >= BW_METER_BUCKETS)
2252 			time_hash = 0;
2253 		}
2254 		x->bm_time_next = V_bw_meter_timers[time_hash];
2255 		V_bw_meter_timers[time_hash] = x;
2256 		x->bm_time_hash = time_hash;
2257 
2258 		continue;
2259 	    }
2260 
2261 	    /*
2262 	     * Test if we should deliver an upcall
2263 	     */
2264 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2265 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2266 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2267 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2268 		/* Prepare an upcall for delivery */
2269 		bw_meter_prepare_upcall(x, &now);
2270 	    }
2271 
2272 	    /*
2273 	     * Reschedule for next processing
2274 	     */
2275 	    schedule_bw_meter(x, &now);
2276 	}
2277     }
2278 
2279     /* Send all upcalls that are pending delivery */
2280     bw_upcalls_send();
2281 
2282     MFC_UNLOCK();
2283 }
2284 
2285 /*
2286  * A periodic function for sending all upcalls that are pending delivery
2287  */
2288 static void
2289 expire_bw_upcalls_send(void *arg)
2290 {
2291     CURVNET_SET((struct vnet *) arg);
2292 
2293     MFC_LOCK();
2294     bw_upcalls_send();
2295     MFC_UNLOCK();
2296 
2297     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2298 	curvnet);
2299     CURVNET_RESTORE();
2300 }
2301 
2302 /*
2303  * A periodic function for periodic scanning of the multicast forwarding
2304  * table for processing all "<=" bw_meter entries.
2305  */
2306 static void
2307 expire_bw_meter_process(void *arg)
2308 {
2309     CURVNET_SET((struct vnet *) arg);
2310 
2311     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2312 	bw_meter_process();
2313 
2314     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2315 	curvnet);
2316     CURVNET_RESTORE();
2317 }
2318 
2319 /*
2320  * End of bandwidth monitoring code
2321  */
2322 
2323 /*
2324  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2325  *
2326  */
2327 static int
2328 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2329     struct mfc *rt)
2330 {
2331     struct mbuf *mb_copy, *mm;
2332 
2333     /*
2334      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2335      * rendezvous point was unspecified, and we were told not to.
2336      */
2337     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2338 	in_nullhost(rt->mfc_rp))
2339 	return 0;
2340 
2341     mb_copy = pim_register_prepare(ip, m);
2342     if (mb_copy == NULL)
2343 	return ENOBUFS;
2344 
2345     /*
2346      * Send all the fragments. Note that the mbuf for each fragment
2347      * is freed by the sending machinery.
2348      */
2349     for (mm = mb_copy; mm; mm = mb_copy) {
2350 	mb_copy = mm->m_nextpkt;
2351 	mm->m_nextpkt = 0;
2352 	mm = m_pullup(mm, sizeof(struct ip));
2353 	if (mm != NULL) {
2354 	    ip = mtod(mm, struct ip *);
2355 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2356 		pim_register_send_rp(ip, vifp, mm, rt);
2357 	    } else {
2358 		pim_register_send_upcall(ip, vifp, mm, rt);
2359 	    }
2360 	}
2361     }
2362 
2363     return 0;
2364 }
2365 
2366 /*
2367  * Return a copy of the data packet that is ready for PIM Register
2368  * encapsulation.
2369  * XXX: Note that in the returned copy the IP header is a valid one.
2370  */
2371 static struct mbuf *
2372 pim_register_prepare(struct ip *ip, struct mbuf *m)
2373 {
2374     struct mbuf *mb_copy = NULL;
2375     int mtu;
2376 
2377     /* Take care of delayed checksums */
2378     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2379 	in_delayed_cksum(m);
2380 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2381     }
2382 
2383     /*
2384      * Copy the old packet & pullup its IP header into the
2385      * new mbuf so we can modify it.
2386      */
2387     mb_copy = m_copypacket(m, M_NOWAIT);
2388     if (mb_copy == NULL)
2389 	return NULL;
2390     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2391     if (mb_copy == NULL)
2392 	return NULL;
2393 
2394     /* take care of the TTL */
2395     ip = mtod(mb_copy, struct ip *);
2396     --ip->ip_ttl;
2397 
2398     /* Compute the MTU after the PIM Register encapsulation */
2399     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2400 
2401     if (ntohs(ip->ip_len) <= mtu) {
2402 	/* Turn the IP header into a valid one */
2403 	ip->ip_sum = 0;
2404 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2405     } else {
2406 	/* Fragment the packet */
2407 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2408 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2409 	    m_freem(mb_copy);
2410 	    return NULL;
2411 	}
2412     }
2413     return mb_copy;
2414 }
2415 
2416 /*
2417  * Send an upcall with the data packet to the user-level process.
2418  */
2419 static int
2420 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2421     struct mbuf *mb_copy, struct mfc *rt)
2422 {
2423     struct mbuf *mb_first;
2424     int len = ntohs(ip->ip_len);
2425     struct igmpmsg *im;
2426     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2427 
2428     VIF_LOCK_ASSERT();
2429 
2430     /*
2431      * Add a new mbuf with an upcall header
2432      */
2433     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2434     if (mb_first == NULL) {
2435 	m_freem(mb_copy);
2436 	return ENOBUFS;
2437     }
2438     mb_first->m_data += max_linkhdr;
2439     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2440     mb_first->m_len = sizeof(struct igmpmsg);
2441     mb_first->m_next = mb_copy;
2442 
2443     /* Send message to routing daemon */
2444     im = mtod(mb_first, struct igmpmsg *);
2445     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2446     im->im_mbz		= 0;
2447     im->im_vif		= vifp - V_viftable;
2448     im->im_src		= ip->ip_src;
2449     im->im_dst		= ip->ip_dst;
2450 
2451     k_igmpsrc.sin_addr	= ip->ip_src;
2452 
2453     MRTSTAT_INC(mrts_upcalls);
2454 
2455     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2456 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2457 	MRTSTAT_INC(mrts_upq_sockfull);
2458 	return ENOBUFS;
2459     }
2460 
2461     /* Keep statistics */
2462     PIMSTAT_INC(pims_snd_registers_msgs);
2463     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2464 
2465     return 0;
2466 }
2467 
2468 /*
2469  * Encapsulate the data packet in PIM Register message and send it to the RP.
2470  */
2471 static int
2472 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2473     struct mfc *rt)
2474 {
2475     struct mbuf *mb_first;
2476     struct ip *ip_outer;
2477     struct pim_encap_pimhdr *pimhdr;
2478     int len = ntohs(ip->ip_len);
2479     vifi_t vifi = rt->mfc_parent;
2480 
2481     VIF_LOCK_ASSERT();
2482 
2483     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2484 	m_freem(mb_copy);
2485 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2486     }
2487 
2488     /*
2489      * Add a new mbuf with the encapsulating header
2490      */
2491     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2492     if (mb_first == NULL) {
2493 	m_freem(mb_copy);
2494 	return ENOBUFS;
2495     }
2496     mb_first->m_data += max_linkhdr;
2497     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2498     mb_first->m_next = mb_copy;
2499 
2500     mb_first->m_pkthdr.len = len + mb_first->m_len;
2501 
2502     /*
2503      * Fill in the encapsulating IP and PIM header
2504      */
2505     ip_outer = mtod(mb_first, struct ip *);
2506     *ip_outer = pim_encap_iphdr;
2507     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2508 	sizeof(pim_encap_pimhdr));
2509     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2510     ip_outer->ip_dst = rt->mfc_rp;
2511     /*
2512      * Copy the inner header TOS to the outer header, and take care of the
2513      * IP_DF bit.
2514      */
2515     ip_outer->ip_tos = ip->ip_tos;
2516     if (ip->ip_off & htons(IP_DF))
2517 	ip_outer->ip_off |= htons(IP_DF);
2518     ip_fillid(ip_outer);
2519     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2520 					 + sizeof(pim_encap_iphdr));
2521     *pimhdr = pim_encap_pimhdr;
2522     /* If the iif crosses a border, set the Border-bit */
2523     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2524 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2525 
2526     mb_first->m_data += sizeof(pim_encap_iphdr);
2527     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2528     mb_first->m_data -= sizeof(pim_encap_iphdr);
2529 
2530     send_packet(vifp, mb_first);
2531 
2532     /* Keep statistics */
2533     PIMSTAT_INC(pims_snd_registers_msgs);
2534     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2535 
2536     return 0;
2537 }
2538 
2539 /*
2540  * pim_encapcheck() is called by the encap4_input() path at runtime to
2541  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2542  * into the kernel.
2543  */
2544 static int
2545 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2546 {
2547 
2548 #ifdef DIAGNOSTIC
2549     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2550 #endif
2551     if (proto != IPPROTO_PIM)
2552 	return 0;	/* not for us; reject the datagram. */
2553 
2554     return 64;		/* claim the datagram. */
2555 }
2556 
2557 /*
2558  * PIM-SMv2 and PIM-DM messages processing.
2559  * Receives and verifies the PIM control messages, and passes them
2560  * up to the listening socket, using rip_input().
2561  * The only message with special processing is the PIM_REGISTER message
2562  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2563  * is passed to if_simloop().
2564  */
2565 int
2566 pim_input(struct mbuf **mp, int *offp, int proto)
2567 {
2568     struct mbuf *m = *mp;
2569     struct ip *ip = mtod(m, struct ip *);
2570     struct pim *pim;
2571     int iphlen = *offp;
2572     int minlen;
2573     int datalen = ntohs(ip->ip_len) - iphlen;
2574     int ip_tos;
2575 
2576     *mp = NULL;
2577 
2578     /* Keep statistics */
2579     PIMSTAT_INC(pims_rcv_total_msgs);
2580     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2581 
2582     /*
2583      * Validate lengths
2584      */
2585     if (datalen < PIM_MINLEN) {
2586 	PIMSTAT_INC(pims_rcv_tooshort);
2587 	CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2588 	    __func__, datalen, ntohl(ip->ip_src.s_addr));
2589 	m_freem(m);
2590 	return (IPPROTO_DONE);
2591     }
2592 
2593     /*
2594      * If the packet is at least as big as a REGISTER, go agead
2595      * and grab the PIM REGISTER header size, to avoid another
2596      * possible m_pullup() later.
2597      *
2598      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2599      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2600      */
2601     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2602     /*
2603      * Get the IP and PIM headers in contiguous memory, and
2604      * possibly the PIM REGISTER header.
2605      */
2606     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2607 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2608 	return (IPPROTO_DONE);
2609     }
2610 
2611     /* m_pullup() may have given us a new mbuf so reset ip. */
2612     ip = mtod(m, struct ip *);
2613     ip_tos = ip->ip_tos;
2614 
2615     /* adjust mbuf to point to the PIM header */
2616     m->m_data += iphlen;
2617     m->m_len  -= iphlen;
2618     pim = mtod(m, struct pim *);
2619 
2620     /*
2621      * Validate checksum. If PIM REGISTER, exclude the data packet.
2622      *
2623      * XXX: some older PIMv2 implementations don't make this distinction,
2624      * so for compatibility reason perform the checksum over part of the
2625      * message, and if error, then over the whole message.
2626      */
2627     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2628 	/* do nothing, checksum okay */
2629     } else if (in_cksum(m, datalen)) {
2630 	PIMSTAT_INC(pims_rcv_badsum);
2631 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2632 	m_freem(m);
2633 	return (IPPROTO_DONE);
2634     }
2635 
2636     /* PIM version check */
2637     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2638 	PIMSTAT_INC(pims_rcv_badversion);
2639 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2640 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2641 	m_freem(m);
2642 	return (IPPROTO_DONE);
2643     }
2644 
2645     /* restore mbuf back to the outer IP */
2646     m->m_data -= iphlen;
2647     m->m_len  += iphlen;
2648 
2649     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2650 	/*
2651 	 * Since this is a REGISTER, we'll make a copy of the register
2652 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2653 	 * routing daemon.
2654 	 */
2655 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2656 	struct mbuf *mcp;
2657 	struct ip *encap_ip;
2658 	u_int32_t *reghdr;
2659 	struct ifnet *vifp;
2660 
2661 	VIF_LOCK();
2662 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2663 	    VIF_UNLOCK();
2664 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2665 		(int)V_reg_vif_num);
2666 	    m_freem(m);
2667 	    return (IPPROTO_DONE);
2668 	}
2669 	/* XXX need refcnt? */
2670 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2671 	VIF_UNLOCK();
2672 
2673 	/*
2674 	 * Validate length
2675 	 */
2676 	if (datalen < PIM_REG_MINLEN) {
2677 	    PIMSTAT_INC(pims_rcv_tooshort);
2678 	    PIMSTAT_INC(pims_rcv_badregisters);
2679 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2680 	    m_freem(m);
2681 	    return (IPPROTO_DONE);
2682 	}
2683 
2684 	reghdr = (u_int32_t *)(pim + 1);
2685 	encap_ip = (struct ip *)(reghdr + 1);
2686 
2687 	CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2688 	    __func__, ntohl(encap_ip->ip_src.s_addr),
2689 	    ntohs(encap_ip->ip_len));
2690 
2691 	/* verify the version number of the inner packet */
2692 	if (encap_ip->ip_v != IPVERSION) {
2693 	    PIMSTAT_INC(pims_rcv_badregisters);
2694 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2695 	    m_freem(m);
2696 	    return (IPPROTO_DONE);
2697 	}
2698 
2699 	/* verify the inner packet is destined to a mcast group */
2700 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2701 	    PIMSTAT_INC(pims_rcv_badregisters);
2702 	    CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2703 		ntohl(encap_ip->ip_dst.s_addr));
2704 	    m_freem(m);
2705 	    return (IPPROTO_DONE);
2706 	}
2707 
2708 	/* If a NULL_REGISTER, pass it to the daemon */
2709 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2710 	    goto pim_input_to_daemon;
2711 
2712 	/*
2713 	 * Copy the TOS from the outer IP header to the inner IP header.
2714 	 */
2715 	if (encap_ip->ip_tos != ip_tos) {
2716 	    /* Outer TOS -> inner TOS */
2717 	    encap_ip->ip_tos = ip_tos;
2718 	    /* Recompute the inner header checksum. Sigh... */
2719 
2720 	    /* adjust mbuf to point to the inner IP header */
2721 	    m->m_data += (iphlen + PIM_MINLEN);
2722 	    m->m_len  -= (iphlen + PIM_MINLEN);
2723 
2724 	    encap_ip->ip_sum = 0;
2725 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2726 
2727 	    /* restore mbuf to point back to the outer IP header */
2728 	    m->m_data -= (iphlen + PIM_MINLEN);
2729 	    m->m_len  += (iphlen + PIM_MINLEN);
2730 	}
2731 
2732 	/*
2733 	 * Decapsulate the inner IP packet and loopback to forward it
2734 	 * as a normal multicast packet. Also, make a copy of the
2735 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2736 	 * to pass to the daemon later, so it can take the appropriate
2737 	 * actions (e.g., send back PIM_REGISTER_STOP).
2738 	 * XXX: here m->m_data points to the outer IP header.
2739 	 */
2740 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2741 	if (mcp == NULL) {
2742 	    CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2743 	    m_freem(m);
2744 	    return (IPPROTO_DONE);
2745 	}
2746 
2747 	/* Keep statistics */
2748 	/* XXX: registers_bytes include only the encap. mcast pkt */
2749 	PIMSTAT_INC(pims_rcv_registers_msgs);
2750 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2751 
2752 	/*
2753 	 * forward the inner ip packet; point m_data at the inner ip.
2754 	 */
2755 	m_adj(m, iphlen + PIM_MINLEN);
2756 
2757 	CTR4(KTR_IPMF,
2758 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2759 	    __func__,
2760 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2761 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2762 	    (int)V_reg_vif_num);
2763 
2764 	/* NB: vifp was collected above; can it change on us? */
2765 	if_simloop(vifp, m, dst.sin_family, 0);
2766 
2767 	/* prepare the register head to send to the mrouting daemon */
2768 	m = mcp;
2769     }
2770 
2771 pim_input_to_daemon:
2772     /*
2773      * Pass the PIM message up to the daemon; if it is a Register message,
2774      * pass the 'head' only up to the daemon. This includes the
2775      * outer IP header, PIM header, PIM-Register header and the
2776      * inner IP header.
2777      * XXX: the outer IP header pkt size of a Register is not adjust to
2778      * reflect the fact that the inner multicast data is truncated.
2779      */
2780     *mp = m;
2781     rip_input(mp, offp, proto);
2782 
2783     return (IPPROTO_DONE);
2784 }
2785 
2786 static int
2787 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2788 {
2789 	struct mfc	*rt;
2790 	int		 error, i;
2791 
2792 	if (req->newptr)
2793 		return (EPERM);
2794 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2795 		return (0);
2796 	error = sysctl_wire_old_buffer(req, 0);
2797 	if (error)
2798 		return (error);
2799 
2800 	MFC_LOCK();
2801 	for (i = 0; i < mfchashsize; i++) {
2802 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2803 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2804 			if (error)
2805 				goto out_locked;
2806 		}
2807 	}
2808 out_locked:
2809 	MFC_UNLOCK();
2810 	return (error);
2811 }
2812 
2813 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2814     sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2815     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2816 
2817 static void
2818 vnet_mroute_init(const void *unused __unused)
2819 {
2820 
2821 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2822 	bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2823 	callout_init(&V_expire_upcalls_ch, 1);
2824 	callout_init(&V_bw_upcalls_ch, 1);
2825 	callout_init(&V_bw_meter_ch, 1);
2826 }
2827 
2828 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2829 	NULL);
2830 
2831 static void
2832 vnet_mroute_uninit(const void *unused __unused)
2833 {
2834 
2835 	free(V_nexpire, M_MRTABLE);
2836 	V_nexpire = NULL;
2837 }
2838 
2839 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2840 	vnet_mroute_uninit, NULL);
2841 
2842 static int
2843 ip_mroute_modevent(module_t mod, int type, void *unused)
2844 {
2845 
2846     switch (type) {
2847     case MOD_LOAD:
2848 	MROUTER_LOCK_INIT();
2849 
2850 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2851 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2852 	if (if_detach_event_tag == NULL) {
2853 		printf("ip_mroute: unable to register "
2854 		    "ifnet_departure_event handler\n");
2855 		MROUTER_LOCK_DESTROY();
2856 		return (EINVAL);
2857 	}
2858 
2859 	MFC_LOCK_INIT();
2860 	VIF_LOCK_INIT();
2861 
2862 	mfchashsize = MFCHASHSIZE;
2863 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2864 	    !powerof2(mfchashsize)) {
2865 		printf("WARNING: %s not a power of 2; using default\n",
2866 		    "net.inet.ip.mfchashsize");
2867 		mfchashsize = MFCHASHSIZE;
2868 	}
2869 
2870 	pim_squelch_wholepkt = 0;
2871 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2872 	    &pim_squelch_wholepkt);
2873 
2874 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2875 	    pim_encapcheck, &in_pim_protosw, NULL);
2876 	if (pim_encap_cookie == NULL) {
2877 		printf("ip_mroute: unable to attach pim encap\n");
2878 		VIF_LOCK_DESTROY();
2879 		MFC_LOCK_DESTROY();
2880 		MROUTER_LOCK_DESTROY();
2881 		return (EINVAL);
2882 	}
2883 
2884 	ip_mcast_src = X_ip_mcast_src;
2885 	ip_mforward = X_ip_mforward;
2886 	ip_mrouter_done = X_ip_mrouter_done;
2887 	ip_mrouter_get = X_ip_mrouter_get;
2888 	ip_mrouter_set = X_ip_mrouter_set;
2889 
2890 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2891 	ip_rsvp_vif = X_ip_rsvp_vif;
2892 
2893 	legal_vif_num = X_legal_vif_num;
2894 	mrt_ioctl = X_mrt_ioctl;
2895 	rsvp_input_p = X_rsvp_input;
2896 	break;
2897 
2898     case MOD_UNLOAD:
2899 	/*
2900 	 * Typically module unload happens after the user-level
2901 	 * process has shutdown the kernel services (the check
2902 	 * below insures someone can't just yank the module out
2903 	 * from under a running process).  But if the module is
2904 	 * just loaded and then unloaded w/o starting up a user
2905 	 * process we still need to cleanup.
2906 	 */
2907 	MROUTER_LOCK();
2908 	if (ip_mrouter_cnt != 0) {
2909 	    MROUTER_UNLOCK();
2910 	    return (EINVAL);
2911 	}
2912 	ip_mrouter_unloading = 1;
2913 	MROUTER_UNLOCK();
2914 
2915 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2916 
2917 	if (pim_encap_cookie) {
2918 	    encap_detach(pim_encap_cookie);
2919 	    pim_encap_cookie = NULL;
2920 	}
2921 
2922 	ip_mcast_src = NULL;
2923 	ip_mforward = NULL;
2924 	ip_mrouter_done = NULL;
2925 	ip_mrouter_get = NULL;
2926 	ip_mrouter_set = NULL;
2927 
2928 	ip_rsvp_force_done = NULL;
2929 	ip_rsvp_vif = NULL;
2930 
2931 	legal_vif_num = NULL;
2932 	mrt_ioctl = NULL;
2933 	rsvp_input_p = NULL;
2934 
2935 	VIF_LOCK_DESTROY();
2936 	MFC_LOCK_DESTROY();
2937 	MROUTER_LOCK_DESTROY();
2938 	break;
2939 
2940     default:
2941 	return EOPNOTSUPP;
2942     }
2943     return 0;
2944 }
2945 
2946 static moduledata_t ip_mroutemod = {
2947     "ip_mroute",
2948     ip_mroute_modevent,
2949     0
2950 };
2951 
2952 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2953