xref: /freebsd/sys/netinet/ip_mroute.c (revision ec4deee4e4f2aef1b97d9424f25d04e91fd7dc10)
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 VNET_DEFINE_STATIC(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 VNET_PCPUSTAT_DEFINE_STATIC(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 VNET_DEFINE_STATIC(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 VNET_DEFINE_STATIC(u_char *, nexpire);		/* 0..mfchashsize-1 */
168 #define	V_nexpire		VNET(nexpire)
169 VNET_DEFINE_STATIC(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 VNET_DEFINE_STATIC(vifi_t, numvifs);
181 #define	V_numvifs		VNET(numvifs)
182 VNET_DEFINE_STATIC(struct vif *, viftable);
183 #define	V_viftable		VNET(viftable)
184 /*
185  * No one should be able to "query" this before initialisation happened in
186  * vnet_mroute_init(), so we should still be fine.
187  */
188 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_VNET | CTLFLAG_RD,
189     &VNET_NAME(viftable), sizeof(*V_viftable) * MAXVIFS, "S,vif[MAXVIFS]",
190     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
191 
192 static struct mtx vif_mtx;
193 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
194 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
195 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
196 #define	VIF_LOCK_INIT()							\
197 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
198 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
199 
200 static eventhandler_tag if_detach_event_tag = NULL;
201 
202 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
203 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
204 
205 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
206 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
207 
208 /*
209  * Bandwidth meter variables and constants
210  */
211 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
212 /*
213  * Pending timeouts are stored in a hash table, the key being the
214  * expiration time. Periodically, the entries are analysed and processed.
215  */
216 #define	BW_METER_BUCKETS	1024
217 VNET_DEFINE_STATIC(struct bw_meter **, bw_meter_timers);
218 #define	V_bw_meter_timers	VNET(bw_meter_timers)
219 VNET_DEFINE_STATIC(struct callout, bw_meter_ch);
220 #define	V_bw_meter_ch		VNET(bw_meter_ch)
221 #define	BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
222 
223 /*
224  * Pending upcalls are stored in a vector which is flushed when
225  * full, or periodically
226  */
227 VNET_DEFINE_STATIC(struct bw_upcall *, bw_upcalls);
228 #define	V_bw_upcalls		VNET(bw_upcalls)
229 VNET_DEFINE_STATIC(u_int, bw_upcalls_n); /* # of pending upcalls */
230 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
231 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
232 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
233 
234 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
235 
236 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
237 VNET_PCPUSTAT_SYSINIT(pimstat);
238 VNET_PCPUSTAT_SYSUNINIT(pimstat);
239 
240 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
241     "PIM");
242 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
243     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
244 
245 static u_long	pim_squelch_wholepkt = 0;
246 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
247     &pim_squelch_wholepkt, 0,
248     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
249 
250 static const struct encaptab *pim_encap_cookie;
251 static int pim_encapcheck(const struct mbuf *, int, int, void *);
252 static int pim_input(struct mbuf *, int, int, void *);
253 
254 static const struct encap_config ipv4_encap_cfg = {
255 	.proto = IPPROTO_PIM,
256 	.min_length = sizeof(struct ip) + PIM_MINLEN,
257 	.exact_match = 8,
258 	.check = pim_encapcheck,
259 	.input = pim_input
260 };
261 
262 /*
263  * Note: the PIM Register encapsulation adds the following in front of a
264  * data packet:
265  *
266  * struct pim_encap_hdr {
267  *    struct ip ip;
268  *    struct pim_encap_pimhdr  pim;
269  * }
270  *
271  */
272 
273 struct pim_encap_pimhdr {
274 	struct pim pim;
275 	uint32_t   flags;
276 };
277 #define		PIM_ENCAP_TTL	64
278 
279 static struct ip pim_encap_iphdr = {
280 #if BYTE_ORDER == LITTLE_ENDIAN
281 	sizeof(struct ip) >> 2,
282 	IPVERSION,
283 #else
284 	IPVERSION,
285 	sizeof(struct ip) >> 2,
286 #endif
287 	0,			/* tos */
288 	sizeof(struct ip),	/* total length */
289 	0,			/* id */
290 	0,			/* frag offset */
291 	PIM_ENCAP_TTL,
292 	IPPROTO_PIM,
293 	0,			/* checksum */
294 };
295 
296 static struct pim_encap_pimhdr pim_encap_pimhdr = {
297     {
298 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
299 	0,			/* reserved */
300 	0,			/* checksum */
301     },
302     0				/* flags */
303 };
304 
305 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
306 #define	V_reg_vif_num		VNET(reg_vif_num)
307 VNET_DEFINE_STATIC(struct ifnet, multicast_register_if);
308 #define	V_multicast_register_if	VNET(multicast_register_if)
309 
310 /*
311  * Private variables.
312  */
313 
314 static u_long	X_ip_mcast_src(int);
315 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
316 		    struct ip_moptions *);
317 static int	X_ip_mrouter_done(void);
318 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
319 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
320 static int	X_legal_vif_num(int);
321 static int	X_mrt_ioctl(u_long, caddr_t, int);
322 
323 static int	add_bw_upcall(struct bw_upcall *);
324 static int	add_mfc(struct mfcctl2 *);
325 static int	add_vif(struct vifctl *);
326 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
327 static void	bw_meter_process(void);
328 static void	bw_meter_receive_packet(struct bw_meter *, int,
329 		    struct timeval *);
330 static void	bw_upcalls_send(void);
331 static int	del_bw_upcall(struct bw_upcall *);
332 static int	del_mfc(struct mfcctl2 *);
333 static int	del_vif(vifi_t);
334 static int	del_vif_locked(vifi_t);
335 static void	expire_bw_meter_process(void *);
336 static void	expire_bw_upcalls_send(void *);
337 static void	expire_mfc(struct mfc *);
338 static void	expire_upcalls(void *);
339 static void	free_bw_list(struct bw_meter *);
340 static int	get_sg_cnt(struct sioc_sg_req *);
341 static int	get_vif_cnt(struct sioc_vif_req *);
342 static void	if_detached_event(void *, struct ifnet *);
343 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
344 static int	ip_mrouter_init(struct socket *, int);
345 static __inline struct mfc *
346 		mfc_find(struct in_addr *, struct in_addr *);
347 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
348 static struct mbuf *
349 		pim_register_prepare(struct ip *, struct mbuf *);
350 static int	pim_register_send(struct ip *, struct vif *,
351 		    struct mbuf *, struct mfc *);
352 static int	pim_register_send_rp(struct ip *, struct vif *,
353 		    struct mbuf *, struct mfc *);
354 static int	pim_register_send_upcall(struct ip *, struct vif *,
355 		    struct mbuf *, struct mfc *);
356 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
357 static void	send_packet(struct vif *, struct mbuf *);
358 static int	set_api_config(uint32_t *);
359 static int	set_assert(int);
360 static int	socket_send(struct socket *, struct mbuf *,
361 		    struct sockaddr_in *);
362 static void	unschedule_bw_meter(struct bw_meter *);
363 
364 /*
365  * Kernel multicast forwarding API capabilities and setup.
366  * If more API capabilities are added to the kernel, they should be
367  * recorded in `mrt_api_support'.
368  */
369 #define MRT_API_VERSION		0x0305
370 
371 static const int mrt_api_version = MRT_API_VERSION;
372 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
373 					 MRT_MFC_FLAGS_BORDER_VIF |
374 					 MRT_MFC_RP |
375 					 MRT_MFC_BW_UPCALL);
376 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
377 #define	V_mrt_api_config	VNET(mrt_api_config)
378 VNET_DEFINE_STATIC(int, pim_assert_enabled);
379 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
380 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
381 
382 /*
383  * Find a route for a given origin IP address and multicast group address.
384  * Statistics must be updated by the caller.
385  */
386 static __inline struct mfc *
387 mfc_find(struct in_addr *o, struct in_addr *g)
388 {
389 	struct mfc *rt;
390 
391 	MFC_LOCK_ASSERT();
392 
393 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
394 		if (in_hosteq(rt->mfc_origin, *o) &&
395 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
396 		    TAILQ_EMPTY(&rt->mfc_stall))
397 			break;
398 	}
399 
400 	return (rt);
401 }
402 
403 /*
404  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
405  */
406 static int
407 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
408 {
409     int	error, optval;
410     vifi_t	vifi;
411     struct	vifctl vifc;
412     struct	mfcctl2 mfc;
413     struct	bw_upcall bw_upcall;
414     uint32_t	i;
415 
416     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
417 	return EPERM;
418 
419     error = 0;
420     switch (sopt->sopt_name) {
421     case MRT_INIT:
422 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
423 	if (error)
424 	    break;
425 	error = ip_mrouter_init(so, optval);
426 	break;
427 
428     case MRT_DONE:
429 	error = ip_mrouter_done();
430 	break;
431 
432     case MRT_ADD_VIF:
433 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
434 	if (error)
435 	    break;
436 	error = add_vif(&vifc);
437 	break;
438 
439     case MRT_DEL_VIF:
440 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
441 	if (error)
442 	    break;
443 	error = del_vif(vifi);
444 	break;
445 
446     case MRT_ADD_MFC:
447     case MRT_DEL_MFC:
448 	/*
449 	 * select data size depending on API version.
450 	 */
451 	if (sopt->sopt_name == MRT_ADD_MFC &&
452 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
453 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
454 				sizeof(struct mfcctl2));
455 	} else {
456 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
457 				sizeof(struct mfcctl));
458 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
459 			sizeof(mfc) - sizeof(struct mfcctl));
460 	}
461 	if (error)
462 	    break;
463 	if (sopt->sopt_name == MRT_ADD_MFC)
464 	    error = add_mfc(&mfc);
465 	else
466 	    error = del_mfc(&mfc);
467 	break;
468 
469     case MRT_ASSERT:
470 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
471 	if (error)
472 	    break;
473 	set_assert(optval);
474 	break;
475 
476     case MRT_API_CONFIG:
477 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
478 	if (!error)
479 	    error = set_api_config(&i);
480 	if (!error)
481 	    error = sooptcopyout(sopt, &i, sizeof i);
482 	break;
483 
484     case MRT_ADD_BW_UPCALL:
485     case MRT_DEL_BW_UPCALL:
486 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
487 				sizeof bw_upcall);
488 	if (error)
489 	    break;
490 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
491 	    error = add_bw_upcall(&bw_upcall);
492 	else
493 	    error = del_bw_upcall(&bw_upcall);
494 	break;
495 
496     default:
497 	error = EOPNOTSUPP;
498 	break;
499     }
500     return error;
501 }
502 
503 /*
504  * Handle MRT getsockopt commands
505  */
506 static int
507 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
508 {
509     int error;
510 
511     switch (sopt->sopt_name) {
512     case MRT_VERSION:
513 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
514 	break;
515 
516     case MRT_ASSERT:
517 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
518 	    sizeof V_pim_assert_enabled);
519 	break;
520 
521     case MRT_API_SUPPORT:
522 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
523 	break;
524 
525     case MRT_API_CONFIG:
526 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
527 	break;
528 
529     default:
530 	error = EOPNOTSUPP;
531 	break;
532     }
533     return error;
534 }
535 
536 /*
537  * Handle ioctl commands to obtain information from the cache
538  */
539 static int
540 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
541 {
542     int error = 0;
543 
544     /*
545      * Currently the only function calling this ioctl routine is rtioctl_fib().
546      * Typically, only root can create the raw socket in order to execute
547      * this ioctl method, however the request might be coming from a prison
548      */
549     error = priv_check(curthread, PRIV_NETINET_MROUTE);
550     if (error)
551 	return (error);
552     switch (cmd) {
553     case (SIOCGETVIFCNT):
554 	error = get_vif_cnt((struct sioc_vif_req *)data);
555 	break;
556 
557     case (SIOCGETSGCNT):
558 	error = get_sg_cnt((struct sioc_sg_req *)data);
559 	break;
560 
561     default:
562 	error = EINVAL;
563 	break;
564     }
565     return error;
566 }
567 
568 /*
569  * returns the packet, byte, rpf-failure count for the source group provided
570  */
571 static int
572 get_sg_cnt(struct sioc_sg_req *req)
573 {
574     struct mfc *rt;
575 
576     MFC_LOCK();
577     rt = mfc_find(&req->src, &req->grp);
578     if (rt == NULL) {
579 	MFC_UNLOCK();
580 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
581 	return EADDRNOTAVAIL;
582     }
583     req->pktcnt = rt->mfc_pkt_cnt;
584     req->bytecnt = rt->mfc_byte_cnt;
585     req->wrong_if = rt->mfc_wrong_if;
586     MFC_UNLOCK();
587     return 0;
588 }
589 
590 /*
591  * returns the input and output packet and byte counts on the vif provided
592  */
593 static int
594 get_vif_cnt(struct sioc_vif_req *req)
595 {
596     vifi_t vifi = req->vifi;
597 
598     VIF_LOCK();
599     if (vifi >= V_numvifs) {
600 	VIF_UNLOCK();
601 	return EINVAL;
602     }
603 
604     req->icount = V_viftable[vifi].v_pkt_in;
605     req->ocount = V_viftable[vifi].v_pkt_out;
606     req->ibytes = V_viftable[vifi].v_bytes_in;
607     req->obytes = V_viftable[vifi].v_bytes_out;
608     VIF_UNLOCK();
609 
610     return 0;
611 }
612 
613 static void
614 if_detached_event(void *arg __unused, struct ifnet *ifp)
615 {
616     vifi_t vifi;
617     u_long i;
618 
619     MROUTER_LOCK();
620 
621     if (V_ip_mrouter == NULL) {
622 	MROUTER_UNLOCK();
623 	return;
624     }
625 
626     VIF_LOCK();
627     MFC_LOCK();
628 
629     /*
630      * Tear down multicast forwarder state associated with this ifnet.
631      * 1. Walk the vif list, matching vifs against this ifnet.
632      * 2. Walk the multicast forwarding cache (mfc) looking for
633      *    inner matches with this vif's index.
634      * 3. Expire any matching multicast forwarding cache entries.
635      * 4. Free vif state. This should disable ALLMULTI on the interface.
636      */
637     for (vifi = 0; vifi < V_numvifs; vifi++) {
638 	if (V_viftable[vifi].v_ifp != ifp)
639 		continue;
640 	for (i = 0; i < mfchashsize; i++) {
641 		struct mfc *rt, *nrt;
642 
643 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
644 			if (rt->mfc_parent == vifi) {
645 				expire_mfc(rt);
646 			}
647 		}
648 	}
649 	del_vif_locked(vifi);
650     }
651 
652     MFC_UNLOCK();
653     VIF_UNLOCK();
654 
655     MROUTER_UNLOCK();
656 }
657 
658 /*
659  * Enable multicast forwarding.
660  */
661 static int
662 ip_mrouter_init(struct socket *so, int version)
663 {
664 
665     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
666         so->so_type, so->so_proto->pr_protocol);
667 
668     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
669 	return EOPNOTSUPP;
670 
671     if (version != 1)
672 	return ENOPROTOOPT;
673 
674     MROUTER_LOCK();
675 
676     if (ip_mrouter_unloading) {
677 	MROUTER_UNLOCK();
678 	return ENOPROTOOPT;
679     }
680 
681     if (V_ip_mrouter != NULL) {
682 	MROUTER_UNLOCK();
683 	return EADDRINUSE;
684     }
685 
686     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
687 	HASH_NOWAIT);
688 
689     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
690 	curvnet);
691     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
692 	curvnet);
693     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
694 	curvnet);
695 
696     V_ip_mrouter = so;
697     ip_mrouter_cnt++;
698 
699     MROUTER_UNLOCK();
700 
701     CTR1(KTR_IPMF, "%s: done", __func__);
702 
703     return 0;
704 }
705 
706 /*
707  * Disable multicast forwarding.
708  */
709 static int
710 X_ip_mrouter_done(void)
711 {
712     struct ifnet *ifp;
713     u_long i;
714     vifi_t vifi;
715 
716     MROUTER_LOCK();
717 
718     if (V_ip_mrouter == NULL) {
719 	MROUTER_UNLOCK();
720 	return EINVAL;
721     }
722 
723     /*
724      * Detach/disable hooks to the reset of the system.
725      */
726     V_ip_mrouter = NULL;
727     ip_mrouter_cnt--;
728     V_mrt_api_config = 0;
729 
730     VIF_LOCK();
731 
732     /*
733      * For each phyint in use, disable promiscuous reception of all IP
734      * multicasts.
735      */
736     for (vifi = 0; vifi < V_numvifs; vifi++) {
737 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
738 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
739 	    ifp = V_viftable[vifi].v_ifp;
740 	    if_allmulti(ifp, 0);
741 	}
742     }
743     bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
744     V_numvifs = 0;
745     V_pim_assert_enabled = 0;
746 
747     VIF_UNLOCK();
748 
749     callout_stop(&V_expire_upcalls_ch);
750     callout_stop(&V_bw_upcalls_ch);
751     callout_stop(&V_bw_meter_ch);
752 
753     MFC_LOCK();
754 
755     /*
756      * Free all multicast forwarding cache entries.
757      * Do not use hashdestroy(), as we must perform other cleanup.
758      */
759     for (i = 0; i < mfchashsize; i++) {
760 	struct mfc *rt, *nrt;
761 
762 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
763 		expire_mfc(rt);
764 	}
765     }
766     free(V_mfchashtbl, M_MRTABLE);
767     V_mfchashtbl = NULL;
768 
769     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
770 
771     V_bw_upcalls_n = 0;
772     bzero(V_bw_meter_timers, BW_METER_BUCKETS * sizeof(*V_bw_meter_timers));
773 
774     MFC_UNLOCK();
775 
776     V_reg_vif_num = VIFI_INVALID;
777 
778     MROUTER_UNLOCK();
779 
780     CTR1(KTR_IPMF, "%s: done", __func__);
781 
782     return 0;
783 }
784 
785 /*
786  * Set PIM assert processing global
787  */
788 static int
789 set_assert(int i)
790 {
791     if ((i != 1) && (i != 0))
792 	return EINVAL;
793 
794     V_pim_assert_enabled = i;
795 
796     return 0;
797 }
798 
799 /*
800  * Configure API capabilities
801  */
802 int
803 set_api_config(uint32_t *apival)
804 {
805     u_long i;
806 
807     /*
808      * We can set the API capabilities only if it is the first operation
809      * after MRT_INIT. I.e.:
810      *  - there are no vifs installed
811      *  - pim_assert is not enabled
812      *  - the MFC table is empty
813      */
814     if (V_numvifs > 0) {
815 	*apival = 0;
816 	return EPERM;
817     }
818     if (V_pim_assert_enabled) {
819 	*apival = 0;
820 	return EPERM;
821     }
822 
823     MFC_LOCK();
824 
825     for (i = 0; i < mfchashsize; i++) {
826 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
827 	    MFC_UNLOCK();
828 	    *apival = 0;
829 	    return EPERM;
830 	}
831     }
832 
833     MFC_UNLOCK();
834 
835     V_mrt_api_config = *apival & mrt_api_support;
836     *apival = V_mrt_api_config;
837 
838     return 0;
839 }
840 
841 /*
842  * Add a vif to the vif table
843  */
844 static int
845 add_vif(struct vifctl *vifcp)
846 {
847     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
848     struct sockaddr_in sin = {sizeof sin, AF_INET};
849     struct ifaddr *ifa;
850     struct ifnet *ifp;
851     int error;
852 
853     VIF_LOCK();
854     if (vifcp->vifc_vifi >= MAXVIFS) {
855 	VIF_UNLOCK();
856 	return EINVAL;
857     }
858     /* rate limiting is no longer supported by this code */
859     if (vifcp->vifc_rate_limit != 0) {
860 	log(LOG_ERR, "rate limiting is no longer supported\n");
861 	VIF_UNLOCK();
862 	return EINVAL;
863     }
864     if (!in_nullhost(vifp->v_lcl_addr)) {
865 	VIF_UNLOCK();
866 	return EADDRINUSE;
867     }
868     if (in_nullhost(vifcp->vifc_lcl_addr)) {
869 	VIF_UNLOCK();
870 	return EADDRNOTAVAIL;
871     }
872 
873     /* Find the interface with an address in AF_INET family */
874     if (vifcp->vifc_flags & VIFF_REGISTER) {
875 	/*
876 	 * XXX: Because VIFF_REGISTER does not really need a valid
877 	 * local interface (e.g. it could be 127.0.0.2), we don't
878 	 * check its address.
879 	 */
880 	ifp = NULL;
881     } else {
882 	struct epoch_tracker et;
883 
884 	sin.sin_addr = vifcp->vifc_lcl_addr;
885 	NET_EPOCH_ENTER(et);
886 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
887 	if (ifa == NULL) {
888 	    NET_EPOCH_EXIT(et);
889 	    VIF_UNLOCK();
890 	    return EADDRNOTAVAIL;
891 	}
892 	ifp = ifa->ifa_ifp;
893 	/* XXX FIXME we need to take a ref on ifp and cleanup properly! */
894 	NET_EPOCH_EXIT(et);
895     }
896 
897     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
898 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
899 	VIF_UNLOCK();
900 	return EOPNOTSUPP;
901     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
902 	ifp = &V_multicast_register_if;
903 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
904 	if (V_reg_vif_num == VIFI_INVALID) {
905 	    if_initname(&V_multicast_register_if, "register_vif", 0);
906 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
907 	    V_reg_vif_num = vifcp->vifc_vifi;
908 	}
909     } else {		/* Make sure the interface supports multicast */
910 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
911 	    VIF_UNLOCK();
912 	    return EOPNOTSUPP;
913 	}
914 
915 	/* Enable promiscuous reception of all IP multicasts from the if */
916 	error = if_allmulti(ifp, 1);
917 	if (error) {
918 	    VIF_UNLOCK();
919 	    return error;
920 	}
921     }
922 
923     vifp->v_flags     = vifcp->vifc_flags;
924     vifp->v_threshold = vifcp->vifc_threshold;
925     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
926     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
927     vifp->v_ifp       = ifp;
928     /* initialize per vif pkt counters */
929     vifp->v_pkt_in    = 0;
930     vifp->v_pkt_out   = 0;
931     vifp->v_bytes_in  = 0;
932     vifp->v_bytes_out = 0;
933 
934     /* Adjust numvifs up if the vifi is higher than numvifs */
935     if (V_numvifs <= vifcp->vifc_vifi)
936 	V_numvifs = vifcp->vifc_vifi + 1;
937 
938     VIF_UNLOCK();
939 
940     CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
941 	(int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
942 	(int)vifcp->vifc_threshold);
943 
944     return 0;
945 }
946 
947 /*
948  * Delete a vif from the vif table
949  */
950 static int
951 del_vif_locked(vifi_t vifi)
952 {
953     struct vif *vifp;
954 
955     VIF_LOCK_ASSERT();
956 
957     if (vifi >= V_numvifs) {
958 	return EINVAL;
959     }
960     vifp = &V_viftable[vifi];
961     if (in_nullhost(vifp->v_lcl_addr)) {
962 	return EADDRNOTAVAIL;
963     }
964 
965     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
966 	if_allmulti(vifp->v_ifp, 0);
967 
968     if (vifp->v_flags & VIFF_REGISTER)
969 	V_reg_vif_num = VIFI_INVALID;
970 
971     bzero((caddr_t)vifp, sizeof (*vifp));
972 
973     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
974 
975     /* Adjust numvifs down */
976     for (vifi = V_numvifs; vifi > 0; vifi--)
977 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
978 	    break;
979     V_numvifs = vifi;
980 
981     return 0;
982 }
983 
984 static int
985 del_vif(vifi_t vifi)
986 {
987     int cc;
988 
989     VIF_LOCK();
990     cc = del_vif_locked(vifi);
991     VIF_UNLOCK();
992 
993     return cc;
994 }
995 
996 /*
997  * update an mfc entry without resetting counters and S,G addresses.
998  */
999 static void
1000 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1001 {
1002     int i;
1003 
1004     rt->mfc_parent = mfccp->mfcc_parent;
1005     for (i = 0; i < V_numvifs; i++) {
1006 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1007 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1008 	    MRT_MFC_FLAGS_ALL;
1009     }
1010     /* set the RP address */
1011     if (V_mrt_api_config & MRT_MFC_RP)
1012 	rt->mfc_rp = mfccp->mfcc_rp;
1013     else
1014 	rt->mfc_rp.s_addr = INADDR_ANY;
1015 }
1016 
1017 /*
1018  * fully initialize an mfc entry from the parameter.
1019  */
1020 static void
1021 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1022 {
1023     rt->mfc_origin     = mfccp->mfcc_origin;
1024     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1025 
1026     update_mfc_params(rt, mfccp);
1027 
1028     /* initialize pkt counters per src-grp */
1029     rt->mfc_pkt_cnt    = 0;
1030     rt->mfc_byte_cnt   = 0;
1031     rt->mfc_wrong_if   = 0;
1032     timevalclear(&rt->mfc_last_assert);
1033 }
1034 
1035 static void
1036 expire_mfc(struct mfc *rt)
1037 {
1038 	struct rtdetq *rte, *nrte;
1039 
1040 	MFC_LOCK_ASSERT();
1041 
1042 	free_bw_list(rt->mfc_bw_meter);
1043 
1044 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1045 		m_freem(rte->m);
1046 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1047 		free(rte, M_MRTABLE);
1048 	}
1049 
1050 	LIST_REMOVE(rt, mfc_hash);
1051 	free(rt, M_MRTABLE);
1052 }
1053 
1054 /*
1055  * Add an mfc entry
1056  */
1057 static int
1058 add_mfc(struct mfcctl2 *mfccp)
1059 {
1060     struct mfc *rt;
1061     struct rtdetq *rte, *nrte;
1062     u_long hash = 0;
1063     u_short nstl;
1064 
1065     VIF_LOCK();
1066     MFC_LOCK();
1067 
1068     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1069 
1070     /* If an entry already exists, just update the fields */
1071     if (rt) {
1072 	CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1073 	    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1074 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1075 	    mfccp->mfcc_parent);
1076 	update_mfc_params(rt, mfccp);
1077 	MFC_UNLOCK();
1078 	VIF_UNLOCK();
1079 	return (0);
1080     }
1081 
1082     /*
1083      * Find the entry for which the upcall was made and update
1084      */
1085     nstl = 0;
1086     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1087     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1088 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1089 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1090 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1091 		CTR5(KTR_IPMF,
1092 		    "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1093 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1094 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1095 		    mfccp->mfcc_parent,
1096 		    TAILQ_FIRST(&rt->mfc_stall));
1097 		if (nstl++)
1098 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1099 
1100 		init_mfc_params(rt, mfccp);
1101 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1102 		V_nexpire[hash]--;
1103 
1104 		/* Free queued packets, but attempt to forward them first. */
1105 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1106 			if (rte->ifp != NULL)
1107 				ip_mdq(rte->m, rte->ifp, rt, -1);
1108 			m_freem(rte->m);
1109 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1110 			rt->mfc_nstall--;
1111 			free(rte, M_MRTABLE);
1112 		}
1113 	}
1114     }
1115 
1116     /*
1117      * It is possible that an entry is being inserted without an upcall
1118      */
1119     if (nstl == 0) {
1120 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1121 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1122 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1123 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1124 			init_mfc_params(rt, mfccp);
1125 			if (rt->mfc_expire)
1126 			    V_nexpire[hash]--;
1127 			rt->mfc_expire = 0;
1128 			break; /* XXX */
1129 		}
1130 	}
1131 
1132 	if (rt == NULL) {		/* no upcall, so make a new entry */
1133 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1134 	    if (rt == NULL) {
1135 		MFC_UNLOCK();
1136 		VIF_UNLOCK();
1137 		return (ENOBUFS);
1138 	    }
1139 
1140 	    init_mfc_params(rt, mfccp);
1141 	    TAILQ_INIT(&rt->mfc_stall);
1142 	    rt->mfc_nstall = 0;
1143 
1144 	    rt->mfc_expire     = 0;
1145 	    rt->mfc_bw_meter = NULL;
1146 
1147 	    /* insert new entry at head of hash chain */
1148 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1149 	}
1150     }
1151 
1152     MFC_UNLOCK();
1153     VIF_UNLOCK();
1154 
1155     return (0);
1156 }
1157 
1158 /*
1159  * Delete an mfc entry
1160  */
1161 static int
1162 del_mfc(struct mfcctl2 *mfccp)
1163 {
1164     struct in_addr	origin;
1165     struct in_addr	mcastgrp;
1166     struct mfc		*rt;
1167 
1168     origin = mfccp->mfcc_origin;
1169     mcastgrp = mfccp->mfcc_mcastgrp;
1170 
1171     CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1172 	ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1173 
1174     MFC_LOCK();
1175 
1176     rt = mfc_find(&origin, &mcastgrp);
1177     if (rt == NULL) {
1178 	MFC_UNLOCK();
1179 	return EADDRNOTAVAIL;
1180     }
1181 
1182     /*
1183      * free the bw_meter entries
1184      */
1185     free_bw_list(rt->mfc_bw_meter);
1186     rt->mfc_bw_meter = NULL;
1187 
1188     LIST_REMOVE(rt, mfc_hash);
1189     free(rt, M_MRTABLE);
1190 
1191     MFC_UNLOCK();
1192 
1193     return (0);
1194 }
1195 
1196 /*
1197  * Send a message to the routing daemon on the multicast routing socket.
1198  */
1199 static int
1200 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1201 {
1202     if (s) {
1203 	SOCKBUF_LOCK(&s->so_rcv);
1204 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1205 	    NULL) != 0) {
1206 	    sorwakeup_locked(s);
1207 	    return 0;
1208 	}
1209 	SOCKBUF_UNLOCK(&s->so_rcv);
1210     }
1211     m_freem(mm);
1212     return -1;
1213 }
1214 
1215 /*
1216  * IP multicast forwarding function. This function assumes that the packet
1217  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1218  * pointed to by "ifp", and the packet is to be relayed to other networks
1219  * that have members of the packet's destination IP multicast group.
1220  *
1221  * The packet is returned unscathed to the caller, unless it is
1222  * erroneous, in which case a non-zero return value tells the caller to
1223  * discard it.
1224  */
1225 
1226 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1227 
1228 static int
1229 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1230     struct ip_moptions *imo)
1231 {
1232     struct mfc *rt;
1233     int error;
1234     vifi_t vifi;
1235 
1236     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1237 	ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1238 
1239     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1240 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1241 	/*
1242 	 * Packet arrived via a physical interface or
1243 	 * an encapsulated tunnel or a register_vif.
1244 	 */
1245     } else {
1246 	/*
1247 	 * Packet arrived through a source-route tunnel.
1248 	 * Source-route tunnels are no longer supported.
1249 	 */
1250 	return (1);
1251     }
1252 
1253     VIF_LOCK();
1254     MFC_LOCK();
1255     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1256 	if (ip->ip_ttl < MAXTTL)
1257 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1258 	error = ip_mdq(m, ifp, NULL, vifi);
1259 	MFC_UNLOCK();
1260 	VIF_UNLOCK();
1261 	return error;
1262     }
1263 
1264     /*
1265      * Don't forward a packet with time-to-live of zero or one,
1266      * or a packet destined to a local-only group.
1267      */
1268     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1269 	MFC_UNLOCK();
1270 	VIF_UNLOCK();
1271 	return 0;
1272     }
1273 
1274     /*
1275      * Determine forwarding vifs from the forwarding cache table
1276      */
1277     MRTSTAT_INC(mrts_mfc_lookups);
1278     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1279 
1280     /* Entry exists, so forward if necessary */
1281     if (rt != NULL) {
1282 	error = ip_mdq(m, ifp, rt, -1);
1283 	MFC_UNLOCK();
1284 	VIF_UNLOCK();
1285 	return error;
1286     } else {
1287 	/*
1288 	 * If we don't have a route for packet's origin,
1289 	 * Make a copy of the packet & send message to routing daemon
1290 	 */
1291 
1292 	struct mbuf *mb0;
1293 	struct rtdetq *rte;
1294 	u_long hash;
1295 	int hlen = ip->ip_hl << 2;
1296 
1297 	MRTSTAT_INC(mrts_mfc_misses);
1298 	MRTSTAT_INC(mrts_no_route);
1299 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1300 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1301 
1302 	/*
1303 	 * Allocate mbufs early so that we don't do extra work if we are
1304 	 * just going to fail anyway.  Make sure to pullup the header so
1305 	 * that other people can't step on it.
1306 	 */
1307 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1308 	    M_NOWAIT|M_ZERO);
1309 	if (rte == NULL) {
1310 	    MFC_UNLOCK();
1311 	    VIF_UNLOCK();
1312 	    return ENOBUFS;
1313 	}
1314 
1315 	mb0 = m_copypacket(m, M_NOWAIT);
1316 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1317 	    mb0 = m_pullup(mb0, hlen);
1318 	if (mb0 == NULL) {
1319 	    free(rte, M_MRTABLE);
1320 	    MFC_UNLOCK();
1321 	    VIF_UNLOCK();
1322 	    return ENOBUFS;
1323 	}
1324 
1325 	/* is there an upcall waiting for this flow ? */
1326 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1327 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1328 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1329 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1330 		    !TAILQ_EMPTY(&rt->mfc_stall))
1331 			break;
1332 	}
1333 
1334 	if (rt == NULL) {
1335 	    int i;
1336 	    struct igmpmsg *im;
1337 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1338 	    struct mbuf *mm;
1339 
1340 	    /*
1341 	     * Locate the vifi for the incoming interface for this packet.
1342 	     * If none found, drop packet.
1343 	     */
1344 	    for (vifi = 0; vifi < V_numvifs &&
1345 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1346 		;
1347 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1348 		goto non_fatal;
1349 
1350 	    /* no upcall, so make a new entry */
1351 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1352 	    if (rt == NULL)
1353 		goto fail;
1354 
1355 	    /* Make a copy of the header to send to the user level process */
1356 	    mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1357 	    if (mm == NULL)
1358 		goto fail1;
1359 
1360 	    /*
1361 	     * Send message to routing daemon to install
1362 	     * a route into the kernel table
1363 	     */
1364 
1365 	    im = mtod(mm, struct igmpmsg *);
1366 	    im->im_msgtype = IGMPMSG_NOCACHE;
1367 	    im->im_mbz = 0;
1368 	    im->im_vif = vifi;
1369 
1370 	    MRTSTAT_INC(mrts_upcalls);
1371 
1372 	    k_igmpsrc.sin_addr = ip->ip_src;
1373 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1374 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1375 		MRTSTAT_INC(mrts_upq_sockfull);
1376 fail1:
1377 		free(rt, M_MRTABLE);
1378 fail:
1379 		free(rte, M_MRTABLE);
1380 		m_freem(mb0);
1381 		MFC_UNLOCK();
1382 		VIF_UNLOCK();
1383 		return ENOBUFS;
1384 	    }
1385 
1386 	    /* insert new entry at head of hash chain */
1387 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1388 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1389 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1390 	    V_nexpire[hash]++;
1391 	    for (i = 0; i < V_numvifs; i++) {
1392 		rt->mfc_ttls[i] = 0;
1393 		rt->mfc_flags[i] = 0;
1394 	    }
1395 	    rt->mfc_parent = -1;
1396 
1397 	    /* clear the RP address */
1398 	    rt->mfc_rp.s_addr = INADDR_ANY;
1399 	    rt->mfc_bw_meter = NULL;
1400 
1401 	    /* initialize pkt counters per src-grp */
1402 	    rt->mfc_pkt_cnt = 0;
1403 	    rt->mfc_byte_cnt = 0;
1404 	    rt->mfc_wrong_if = 0;
1405 	    timevalclear(&rt->mfc_last_assert);
1406 
1407 	    TAILQ_INIT(&rt->mfc_stall);
1408 	    rt->mfc_nstall = 0;
1409 
1410 	    /* link into table */
1411 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1412 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1413 	    rt->mfc_nstall++;
1414 
1415 	} else {
1416 	    /* determine if queue has overflowed */
1417 	    if (rt->mfc_nstall > MAX_UPQ) {
1418 		MRTSTAT_INC(mrts_upq_ovflw);
1419 non_fatal:
1420 		free(rte, M_MRTABLE);
1421 		m_freem(mb0);
1422 		MFC_UNLOCK();
1423 		VIF_UNLOCK();
1424 		return (0);
1425 	    }
1426 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1427 	    rt->mfc_nstall++;
1428 	}
1429 
1430 	rte->m			= mb0;
1431 	rte->ifp		= ifp;
1432 
1433 	MFC_UNLOCK();
1434 	VIF_UNLOCK();
1435 
1436 	return 0;
1437     }
1438 }
1439 
1440 /*
1441  * Clean up the cache entry if upcall is not serviced
1442  */
1443 static void
1444 expire_upcalls(void *arg)
1445 {
1446     u_long i;
1447 
1448     CURVNET_SET((struct vnet *) arg);
1449 
1450     MFC_LOCK();
1451 
1452     for (i = 0; i < mfchashsize; i++) {
1453 	struct mfc *rt, *nrt;
1454 
1455 	if (V_nexpire[i] == 0)
1456 	    continue;
1457 
1458 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1459 		if (TAILQ_EMPTY(&rt->mfc_stall))
1460 			continue;
1461 
1462 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1463 			continue;
1464 
1465 		/*
1466 		 * free the bw_meter entries
1467 		 */
1468 		while (rt->mfc_bw_meter != NULL) {
1469 		    struct bw_meter *x = rt->mfc_bw_meter;
1470 
1471 		    rt->mfc_bw_meter = x->bm_mfc_next;
1472 		    free(x, M_BWMETER);
1473 		}
1474 
1475 		MRTSTAT_INC(mrts_cache_cleanups);
1476 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1477 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1478 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1479 
1480 		expire_mfc(rt);
1481 	    }
1482     }
1483 
1484     MFC_UNLOCK();
1485 
1486     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1487 	curvnet);
1488 
1489     CURVNET_RESTORE();
1490 }
1491 
1492 /*
1493  * Packet forwarding routine once entry in the cache is made
1494  */
1495 static int
1496 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1497 {
1498     struct ip  *ip = mtod(m, struct ip *);
1499     vifi_t vifi;
1500     int plen = ntohs(ip->ip_len);
1501 
1502     VIF_LOCK_ASSERT();
1503 
1504     /*
1505      * If xmt_vif is not -1, send on only the requested vif.
1506      *
1507      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1508      */
1509     if (xmt_vif < V_numvifs) {
1510 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1511 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1512 	else
1513 		phyint_send(ip, V_viftable + xmt_vif, m);
1514 	return 1;
1515     }
1516 
1517     /*
1518      * Don't forward if it didn't arrive from the parent vif for its origin.
1519      */
1520     vifi = rt->mfc_parent;
1521     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1522 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1523 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1524 	MRTSTAT_INC(mrts_wrong_if);
1525 	++rt->mfc_wrong_if;
1526 	/*
1527 	 * If we are doing PIM assert processing, send a message
1528 	 * to the routing daemon.
1529 	 *
1530 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1531 	 * can complete the SPT switch, regardless of the type
1532 	 * of the iif (broadcast media, GRE tunnel, etc).
1533 	 */
1534 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1535 	    V_viftable[vifi].v_ifp) {
1536 
1537 	    if (ifp == &V_multicast_register_if)
1538 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1539 
1540 	    /* Get vifi for the incoming packet */
1541 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1542 		vifi++)
1543 		;
1544 	    if (vifi >= V_numvifs)
1545 		return 0;	/* The iif is not found: ignore the packet. */
1546 
1547 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1548 		return 0;	/* WRONGVIF disabled: ignore the packet */
1549 
1550 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1551 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1552 		struct igmpmsg *im;
1553 		int hlen = ip->ip_hl << 2;
1554 		struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1555 
1556 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1557 		    mm = m_pullup(mm, hlen);
1558 		if (mm == NULL)
1559 		    return ENOBUFS;
1560 
1561 		im = mtod(mm, struct igmpmsg *);
1562 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1563 		im->im_mbz		= 0;
1564 		im->im_vif		= vifi;
1565 
1566 		MRTSTAT_INC(mrts_upcalls);
1567 
1568 		k_igmpsrc.sin_addr = im->im_src;
1569 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1570 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1571 		    MRTSTAT_INC(mrts_upq_sockfull);
1572 		    return ENOBUFS;
1573 		}
1574 	    }
1575 	}
1576 	return 0;
1577     }
1578 
1579 
1580     /* If I sourced this packet, it counts as output, else it was input. */
1581     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1582 	V_viftable[vifi].v_pkt_out++;
1583 	V_viftable[vifi].v_bytes_out += plen;
1584     } else {
1585 	V_viftable[vifi].v_pkt_in++;
1586 	V_viftable[vifi].v_bytes_in += plen;
1587     }
1588     rt->mfc_pkt_cnt++;
1589     rt->mfc_byte_cnt += plen;
1590 
1591     /*
1592      * For each vif, decide if a copy of the packet should be forwarded.
1593      * Forward if:
1594      *		- the ttl exceeds the vif's threshold
1595      *		- there are group members downstream on interface
1596      */
1597     for (vifi = 0; vifi < V_numvifs; vifi++)
1598 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1599 	    V_viftable[vifi].v_pkt_out++;
1600 	    V_viftable[vifi].v_bytes_out += plen;
1601 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1602 		pim_register_send(ip, V_viftable + vifi, m, rt);
1603 	    else
1604 		phyint_send(ip, V_viftable + vifi, m);
1605 	}
1606 
1607     /*
1608      * Perform upcall-related bw measuring.
1609      */
1610     if (rt->mfc_bw_meter != NULL) {
1611 	struct bw_meter *x;
1612 	struct timeval now;
1613 
1614 	microtime(&now);
1615 	MFC_LOCK_ASSERT();
1616 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1617 	    bw_meter_receive_packet(x, plen, &now);
1618     }
1619 
1620     return 0;
1621 }
1622 
1623 /*
1624  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1625  */
1626 static int
1627 X_legal_vif_num(int vif)
1628 {
1629 	int ret;
1630 
1631 	ret = 0;
1632 	if (vif < 0)
1633 		return (ret);
1634 
1635 	VIF_LOCK();
1636 	if (vif < V_numvifs)
1637 		ret = 1;
1638 	VIF_UNLOCK();
1639 
1640 	return (ret);
1641 }
1642 
1643 /*
1644  * Return the local address used by this vif
1645  */
1646 static u_long
1647 X_ip_mcast_src(int vifi)
1648 {
1649 	in_addr_t addr;
1650 
1651 	addr = INADDR_ANY;
1652 	if (vifi < 0)
1653 		return (addr);
1654 
1655 	VIF_LOCK();
1656 	if (vifi < V_numvifs)
1657 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1658 	VIF_UNLOCK();
1659 
1660 	return (addr);
1661 }
1662 
1663 static void
1664 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1665 {
1666     struct mbuf *mb_copy;
1667     int hlen = ip->ip_hl << 2;
1668 
1669     VIF_LOCK_ASSERT();
1670 
1671     /*
1672      * Make a new reference to the packet; make sure that
1673      * the IP header is actually copied, not just referenced,
1674      * so that ip_output() only scribbles on the copy.
1675      */
1676     mb_copy = m_copypacket(m, M_NOWAIT);
1677     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1678 	mb_copy = m_pullup(mb_copy, hlen);
1679     if (mb_copy == NULL)
1680 	return;
1681 
1682     send_packet(vifp, mb_copy);
1683 }
1684 
1685 static void
1686 send_packet(struct vif *vifp, struct mbuf *m)
1687 {
1688 	struct ip_moptions imo;
1689 	int error __unused;
1690 
1691 	VIF_LOCK_ASSERT();
1692 
1693 	imo.imo_multicast_ifp  = vifp->v_ifp;
1694 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1695 	imo.imo_multicast_loop = 1;
1696 	imo.imo_multicast_vif  = -1;
1697 	STAILQ_INIT(&imo.imo_head);
1698 
1699 	/*
1700 	 * Re-entrancy should not be a problem here, because
1701 	 * the packets that we send out and are looped back at us
1702 	 * should get rejected because they appear to come from
1703 	 * the loopback interface, thus preventing looping.
1704 	 */
1705 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1706 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1707 	    (ptrdiff_t)(vifp - V_viftable), error);
1708 }
1709 
1710 /*
1711  * Stubs for old RSVP socket shim implementation.
1712  */
1713 
1714 static int
1715 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1716 {
1717 
1718 	return (EOPNOTSUPP);
1719 }
1720 
1721 static void
1722 X_ip_rsvp_force_done(struct socket *so __unused)
1723 {
1724 
1725 }
1726 
1727 static int
1728 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1729 {
1730 	struct mbuf *m;
1731 
1732 	m = *mp;
1733 	*mp = NULL;
1734 	if (!V_rsvp_on)
1735 		m_freem(m);
1736 	return (IPPROTO_DONE);
1737 }
1738 
1739 /*
1740  * Code for bandwidth monitors
1741  */
1742 
1743 /*
1744  * Define common interface for timeval-related methods
1745  */
1746 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1747 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1748 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1749 
1750 static uint32_t
1751 compute_bw_meter_flags(struct bw_upcall *req)
1752 {
1753     uint32_t flags = 0;
1754 
1755     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1756 	flags |= BW_METER_UNIT_PACKETS;
1757     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1758 	flags |= BW_METER_UNIT_BYTES;
1759     if (req->bu_flags & BW_UPCALL_GEQ)
1760 	flags |= BW_METER_GEQ;
1761     if (req->bu_flags & BW_UPCALL_LEQ)
1762 	flags |= BW_METER_LEQ;
1763 
1764     return flags;
1765 }
1766 
1767 /*
1768  * Add a bw_meter entry
1769  */
1770 static int
1771 add_bw_upcall(struct bw_upcall *req)
1772 {
1773     struct mfc *mfc;
1774     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1775 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1776     struct timeval now;
1777     struct bw_meter *x;
1778     uint32_t flags;
1779 
1780     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1781 	return EOPNOTSUPP;
1782 
1783     /* Test if the flags are valid */
1784     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1785 	return EINVAL;
1786     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1787 	return EINVAL;
1788     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1789 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1790 	return EINVAL;
1791 
1792     /* Test if the threshold time interval is valid */
1793     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1794 	return EINVAL;
1795 
1796     flags = compute_bw_meter_flags(req);
1797 
1798     /*
1799      * Find if we have already same bw_meter entry
1800      */
1801     MFC_LOCK();
1802     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1803     if (mfc == NULL) {
1804 	MFC_UNLOCK();
1805 	return EADDRNOTAVAIL;
1806     }
1807     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1808 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1809 			   &req->bu_threshold.b_time, ==)) &&
1810 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1811 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1812 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1813 	    MFC_UNLOCK();
1814 	    return 0;		/* XXX Already installed */
1815 	}
1816     }
1817 
1818     /* Allocate the new bw_meter entry */
1819     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1820     if (x == NULL) {
1821 	MFC_UNLOCK();
1822 	return ENOBUFS;
1823     }
1824 
1825     /* Set the new bw_meter entry */
1826     x->bm_threshold.b_time = req->bu_threshold.b_time;
1827     microtime(&now);
1828     x->bm_start_time = now;
1829     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1830     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1831     x->bm_measured.b_packets = 0;
1832     x->bm_measured.b_bytes = 0;
1833     x->bm_flags = flags;
1834     x->bm_time_next = NULL;
1835     x->bm_time_hash = BW_METER_BUCKETS;
1836 
1837     /* Add the new bw_meter entry to the front of entries for this MFC */
1838     x->bm_mfc = mfc;
1839     x->bm_mfc_next = mfc->mfc_bw_meter;
1840     mfc->mfc_bw_meter = x;
1841     schedule_bw_meter(x, &now);
1842     MFC_UNLOCK();
1843 
1844     return 0;
1845 }
1846 
1847 static void
1848 free_bw_list(struct bw_meter *list)
1849 {
1850     while (list != NULL) {
1851 	struct bw_meter *x = list;
1852 
1853 	list = list->bm_mfc_next;
1854 	unschedule_bw_meter(x);
1855 	free(x, M_BWMETER);
1856     }
1857 }
1858 
1859 /*
1860  * Delete one or multiple bw_meter entries
1861  */
1862 static int
1863 del_bw_upcall(struct bw_upcall *req)
1864 {
1865     struct mfc *mfc;
1866     struct bw_meter *x;
1867 
1868     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1869 	return EOPNOTSUPP;
1870 
1871     MFC_LOCK();
1872 
1873     /* Find the corresponding MFC entry */
1874     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1875     if (mfc == NULL) {
1876 	MFC_UNLOCK();
1877 	return EADDRNOTAVAIL;
1878     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1879 	/*
1880 	 * Delete all bw_meter entries for this mfc
1881 	 */
1882 	struct bw_meter *list;
1883 
1884 	list = mfc->mfc_bw_meter;
1885 	mfc->mfc_bw_meter = NULL;
1886 	free_bw_list(list);
1887 	MFC_UNLOCK();
1888 	return 0;
1889     } else {			/* Delete a single bw_meter entry */
1890 	struct bw_meter *prev;
1891 	uint32_t flags = 0;
1892 
1893 	flags = compute_bw_meter_flags(req);
1894 
1895 	/* Find the bw_meter entry to delete */
1896 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1897 	     prev = x, x = x->bm_mfc_next) {
1898 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1899 			       &req->bu_threshold.b_time, ==)) &&
1900 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1901 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1902 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1903 		break;
1904 	}
1905 	if (x != NULL) { /* Delete entry from the list for this MFC */
1906 	    if (prev != NULL)
1907 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1908 	    else
1909 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1910 
1911 	    unschedule_bw_meter(x);
1912 	    MFC_UNLOCK();
1913 	    /* Free the bw_meter entry */
1914 	    free(x, M_BWMETER);
1915 	    return 0;
1916 	} else {
1917 	    MFC_UNLOCK();
1918 	    return EINVAL;
1919 	}
1920     }
1921     /* NOTREACHED */
1922 }
1923 
1924 /*
1925  * Perform bandwidth measurement processing that may result in an upcall
1926  */
1927 static void
1928 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1929 {
1930     struct timeval delta;
1931 
1932     MFC_LOCK_ASSERT();
1933 
1934     delta = *nowp;
1935     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1936 
1937     if (x->bm_flags & BW_METER_GEQ) {
1938 	/*
1939 	 * Processing for ">=" type of bw_meter entry
1940 	 */
1941 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1942 	    /* Reset the bw_meter entry */
1943 	    x->bm_start_time = *nowp;
1944 	    x->bm_measured.b_packets = 0;
1945 	    x->bm_measured.b_bytes = 0;
1946 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1947 	}
1948 
1949 	/* Record that a packet is received */
1950 	x->bm_measured.b_packets++;
1951 	x->bm_measured.b_bytes += plen;
1952 
1953 	/*
1954 	 * Test if we should deliver an upcall
1955 	 */
1956 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1957 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1958 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1959 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1960 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1961 		/* Prepare an upcall for delivery */
1962 		bw_meter_prepare_upcall(x, nowp);
1963 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1964 	    }
1965 	}
1966     } else if (x->bm_flags & BW_METER_LEQ) {
1967 	/*
1968 	 * Processing for "<=" type of bw_meter entry
1969 	 */
1970 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1971 	    /*
1972 	     * We are behind time with the multicast forwarding table
1973 	     * scanning for "<=" type of bw_meter entries, so test now
1974 	     * if we should deliver an upcall.
1975 	     */
1976 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1977 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1978 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1979 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1980 		/* Prepare an upcall for delivery */
1981 		bw_meter_prepare_upcall(x, nowp);
1982 	    }
1983 	    /* Reschedule the bw_meter entry */
1984 	    unschedule_bw_meter(x);
1985 	    schedule_bw_meter(x, nowp);
1986 	}
1987 
1988 	/* Record that a packet is received */
1989 	x->bm_measured.b_packets++;
1990 	x->bm_measured.b_bytes += plen;
1991 
1992 	/*
1993 	 * Test if we should restart the measuring interval
1994 	 */
1995 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1996 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1997 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1998 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1999 	    /* Don't restart the measuring interval */
2000 	} else {
2001 	    /* Do restart the measuring interval */
2002 	    /*
2003 	     * XXX: note that we don't unschedule and schedule, because this
2004 	     * might be too much overhead per packet. Instead, when we process
2005 	     * all entries for a given timer hash bin, we check whether it is
2006 	     * really a timeout. If not, we reschedule at that time.
2007 	     */
2008 	    x->bm_start_time = *nowp;
2009 	    x->bm_measured.b_packets = 0;
2010 	    x->bm_measured.b_bytes = 0;
2011 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2012 	}
2013     }
2014 }
2015 
2016 /*
2017  * Prepare a bandwidth-related upcall
2018  */
2019 static void
2020 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2021 {
2022     struct timeval delta;
2023     struct bw_upcall *u;
2024 
2025     MFC_LOCK_ASSERT();
2026 
2027     /*
2028      * Compute the measured time interval
2029      */
2030     delta = *nowp;
2031     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2032 
2033     /*
2034      * If there are too many pending upcalls, deliver them now
2035      */
2036     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2037 	bw_upcalls_send();
2038 
2039     /*
2040      * Set the bw_upcall entry
2041      */
2042     u = &V_bw_upcalls[V_bw_upcalls_n++];
2043     u->bu_src = x->bm_mfc->mfc_origin;
2044     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2045     u->bu_threshold.b_time = x->bm_threshold.b_time;
2046     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2047     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2048     u->bu_measured.b_time = delta;
2049     u->bu_measured.b_packets = x->bm_measured.b_packets;
2050     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2051     u->bu_flags = 0;
2052     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2053 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2054     if (x->bm_flags & BW_METER_UNIT_BYTES)
2055 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2056     if (x->bm_flags & BW_METER_GEQ)
2057 	u->bu_flags |= BW_UPCALL_GEQ;
2058     if (x->bm_flags & BW_METER_LEQ)
2059 	u->bu_flags |= BW_UPCALL_LEQ;
2060 }
2061 
2062 /*
2063  * Send the pending bandwidth-related upcalls
2064  */
2065 static void
2066 bw_upcalls_send(void)
2067 {
2068     struct mbuf *m;
2069     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2070     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2071     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2072 				      0,		/* unused2 */
2073 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2074 				      0,		/* im_mbz  */
2075 				      0,		/* im_vif  */
2076 				      0,		/* unused3 */
2077 				      { 0 },		/* im_src  */
2078 				      { 0 } };		/* im_dst  */
2079 
2080     MFC_LOCK_ASSERT();
2081 
2082     if (V_bw_upcalls_n == 0)
2083 	return;			/* No pending upcalls */
2084 
2085     V_bw_upcalls_n = 0;
2086 
2087     /*
2088      * Allocate a new mbuf, initialize it with the header and
2089      * the payload for the pending calls.
2090      */
2091     m = m_gethdr(M_NOWAIT, MT_DATA);
2092     if (m == NULL) {
2093 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2094 	return;
2095     }
2096 
2097     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2098     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2099 
2100     /*
2101      * Send the upcalls
2102      * XXX do we need to set the address in k_igmpsrc ?
2103      */
2104     MRTSTAT_INC(mrts_upcalls);
2105     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2106 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2107 	MRTSTAT_INC(mrts_upq_sockfull);
2108     }
2109 }
2110 
2111 /*
2112  * Compute the timeout hash value for the bw_meter entries
2113  */
2114 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2115     do {								\
2116 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2117 									\
2118 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2119 	(hash) = next_timeval.tv_sec;					\
2120 	if (next_timeval.tv_usec)					\
2121 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2122 	(hash) %= BW_METER_BUCKETS;					\
2123     } while (0)
2124 
2125 /*
2126  * Schedule a timer to process periodically bw_meter entry of type "<="
2127  * by linking the entry in the proper hash bucket.
2128  */
2129 static void
2130 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2131 {
2132     int time_hash;
2133 
2134     MFC_LOCK_ASSERT();
2135 
2136     if (!(x->bm_flags & BW_METER_LEQ))
2137 	return;		/* XXX: we schedule timers only for "<=" entries */
2138 
2139     /*
2140      * Reset the bw_meter entry
2141      */
2142     x->bm_start_time = *nowp;
2143     x->bm_measured.b_packets = 0;
2144     x->bm_measured.b_bytes = 0;
2145     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2146 
2147     /*
2148      * Compute the timeout hash value and insert the entry
2149      */
2150     BW_METER_TIMEHASH(x, time_hash);
2151     x->bm_time_next = V_bw_meter_timers[time_hash];
2152     V_bw_meter_timers[time_hash] = x;
2153     x->bm_time_hash = time_hash;
2154 }
2155 
2156 /*
2157  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2158  * by removing the entry from the proper hash bucket.
2159  */
2160 static void
2161 unschedule_bw_meter(struct bw_meter *x)
2162 {
2163     int time_hash;
2164     struct bw_meter *prev, *tmp;
2165 
2166     MFC_LOCK_ASSERT();
2167 
2168     if (!(x->bm_flags & BW_METER_LEQ))
2169 	return;		/* XXX: we schedule timers only for "<=" entries */
2170 
2171     /*
2172      * Compute the timeout hash value and delete the entry
2173      */
2174     time_hash = x->bm_time_hash;
2175     if (time_hash >= BW_METER_BUCKETS)
2176 	return;		/* Entry was not scheduled */
2177 
2178     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2179 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2180 	if (tmp == x)
2181 	    break;
2182 
2183     if (tmp == NULL)
2184 	panic("unschedule_bw_meter: bw_meter entry not found");
2185 
2186     if (prev != NULL)
2187 	prev->bm_time_next = x->bm_time_next;
2188     else
2189 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2190 
2191     x->bm_time_next = NULL;
2192     x->bm_time_hash = BW_METER_BUCKETS;
2193 }
2194 
2195 
2196 /*
2197  * Process all "<=" type of bw_meter that should be processed now,
2198  * and for each entry prepare an upcall if necessary. Each processed
2199  * entry is rescheduled again for the (periodic) processing.
2200  *
2201  * This is run periodically (once per second normally). On each round,
2202  * all the potentially matching entries are in the hash slot that we are
2203  * looking at.
2204  */
2205 static void
2206 bw_meter_process()
2207 {
2208     uint32_t loops;
2209     int i;
2210     struct timeval now, process_endtime;
2211 
2212     microtime(&now);
2213     if (V_last_tv_sec == now.tv_sec)
2214 	return;		/* nothing to do */
2215 
2216     loops = now.tv_sec - V_last_tv_sec;
2217     V_last_tv_sec = now.tv_sec;
2218     if (loops > BW_METER_BUCKETS)
2219 	loops = BW_METER_BUCKETS;
2220 
2221     MFC_LOCK();
2222     /*
2223      * Process all bins of bw_meter entries from the one after the last
2224      * processed to the current one. On entry, i points to the last bucket
2225      * visited, so we need to increment i at the beginning of the loop.
2226      */
2227     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2228 	struct bw_meter *x, *tmp_list;
2229 
2230 	if (++i >= BW_METER_BUCKETS)
2231 	    i = 0;
2232 
2233 	/* Disconnect the list of bw_meter entries from the bin */
2234 	tmp_list = V_bw_meter_timers[i];
2235 	V_bw_meter_timers[i] = NULL;
2236 
2237 	/* Process the list of bw_meter entries */
2238 	while (tmp_list != NULL) {
2239 	    x = tmp_list;
2240 	    tmp_list = tmp_list->bm_time_next;
2241 
2242 	    /* Test if the time interval is over */
2243 	    process_endtime = x->bm_start_time;
2244 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2245 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2246 		/* Not yet: reschedule, but don't reset */
2247 		int time_hash;
2248 
2249 		BW_METER_TIMEHASH(x, time_hash);
2250 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2251 		    /*
2252 		     * XXX: somehow the bin processing is a bit ahead of time.
2253 		     * Put the entry in the next bin.
2254 		     */
2255 		    if (++time_hash >= BW_METER_BUCKETS)
2256 			time_hash = 0;
2257 		}
2258 		x->bm_time_next = V_bw_meter_timers[time_hash];
2259 		V_bw_meter_timers[time_hash] = x;
2260 		x->bm_time_hash = time_hash;
2261 
2262 		continue;
2263 	    }
2264 
2265 	    /*
2266 	     * Test if we should deliver an upcall
2267 	     */
2268 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2269 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2270 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2271 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2272 		/* Prepare an upcall for delivery */
2273 		bw_meter_prepare_upcall(x, &now);
2274 	    }
2275 
2276 	    /*
2277 	     * Reschedule for next processing
2278 	     */
2279 	    schedule_bw_meter(x, &now);
2280 	}
2281     }
2282 
2283     /* Send all upcalls that are pending delivery */
2284     bw_upcalls_send();
2285 
2286     MFC_UNLOCK();
2287 }
2288 
2289 /*
2290  * A periodic function for sending all upcalls that are pending delivery
2291  */
2292 static void
2293 expire_bw_upcalls_send(void *arg)
2294 {
2295     CURVNET_SET((struct vnet *) arg);
2296 
2297     MFC_LOCK();
2298     bw_upcalls_send();
2299     MFC_UNLOCK();
2300 
2301     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2302 	curvnet);
2303     CURVNET_RESTORE();
2304 }
2305 
2306 /*
2307  * A periodic function for periodic scanning of the multicast forwarding
2308  * table for processing all "<=" bw_meter entries.
2309  */
2310 static void
2311 expire_bw_meter_process(void *arg)
2312 {
2313     CURVNET_SET((struct vnet *) arg);
2314 
2315     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2316 	bw_meter_process();
2317 
2318     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2319 	curvnet);
2320     CURVNET_RESTORE();
2321 }
2322 
2323 /*
2324  * End of bandwidth monitoring code
2325  */
2326 
2327 /*
2328  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2329  *
2330  */
2331 static int
2332 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2333     struct mfc *rt)
2334 {
2335     struct mbuf *mb_copy, *mm;
2336 
2337     /*
2338      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2339      * rendezvous point was unspecified, and we were told not to.
2340      */
2341     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2342 	in_nullhost(rt->mfc_rp))
2343 	return 0;
2344 
2345     mb_copy = pim_register_prepare(ip, m);
2346     if (mb_copy == NULL)
2347 	return ENOBUFS;
2348 
2349     /*
2350      * Send all the fragments. Note that the mbuf for each fragment
2351      * is freed by the sending machinery.
2352      */
2353     for (mm = mb_copy; mm; mm = mb_copy) {
2354 	mb_copy = mm->m_nextpkt;
2355 	mm->m_nextpkt = 0;
2356 	mm = m_pullup(mm, sizeof(struct ip));
2357 	if (mm != NULL) {
2358 	    ip = mtod(mm, struct ip *);
2359 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2360 		pim_register_send_rp(ip, vifp, mm, rt);
2361 	    } else {
2362 		pim_register_send_upcall(ip, vifp, mm, rt);
2363 	    }
2364 	}
2365     }
2366 
2367     return 0;
2368 }
2369 
2370 /*
2371  * Return a copy of the data packet that is ready for PIM Register
2372  * encapsulation.
2373  * XXX: Note that in the returned copy the IP header is a valid one.
2374  */
2375 static struct mbuf *
2376 pim_register_prepare(struct ip *ip, struct mbuf *m)
2377 {
2378     struct mbuf *mb_copy = NULL;
2379     int mtu;
2380 
2381     /* Take care of delayed checksums */
2382     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2383 	in_delayed_cksum(m);
2384 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2385     }
2386 
2387     /*
2388      * Copy the old packet & pullup its IP header into the
2389      * new mbuf so we can modify it.
2390      */
2391     mb_copy = m_copypacket(m, M_NOWAIT);
2392     if (mb_copy == NULL)
2393 	return NULL;
2394     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2395     if (mb_copy == NULL)
2396 	return NULL;
2397 
2398     /* take care of the TTL */
2399     ip = mtod(mb_copy, struct ip *);
2400     --ip->ip_ttl;
2401 
2402     /* Compute the MTU after the PIM Register encapsulation */
2403     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2404 
2405     if (ntohs(ip->ip_len) <= mtu) {
2406 	/* Turn the IP header into a valid one */
2407 	ip->ip_sum = 0;
2408 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2409     } else {
2410 	/* Fragment the packet */
2411 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2412 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2413 	    m_freem(mb_copy);
2414 	    return NULL;
2415 	}
2416     }
2417     return mb_copy;
2418 }
2419 
2420 /*
2421  * Send an upcall with the data packet to the user-level process.
2422  */
2423 static int
2424 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2425     struct mbuf *mb_copy, struct mfc *rt)
2426 {
2427     struct mbuf *mb_first;
2428     int len = ntohs(ip->ip_len);
2429     struct igmpmsg *im;
2430     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2431 
2432     VIF_LOCK_ASSERT();
2433 
2434     /*
2435      * Add a new mbuf with an upcall header
2436      */
2437     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2438     if (mb_first == NULL) {
2439 	m_freem(mb_copy);
2440 	return ENOBUFS;
2441     }
2442     mb_first->m_data += max_linkhdr;
2443     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2444     mb_first->m_len = sizeof(struct igmpmsg);
2445     mb_first->m_next = mb_copy;
2446 
2447     /* Send message to routing daemon */
2448     im = mtod(mb_first, struct igmpmsg *);
2449     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2450     im->im_mbz		= 0;
2451     im->im_vif		= vifp - V_viftable;
2452     im->im_src		= ip->ip_src;
2453     im->im_dst		= ip->ip_dst;
2454 
2455     k_igmpsrc.sin_addr	= ip->ip_src;
2456 
2457     MRTSTAT_INC(mrts_upcalls);
2458 
2459     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2460 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2461 	MRTSTAT_INC(mrts_upq_sockfull);
2462 	return ENOBUFS;
2463     }
2464 
2465     /* Keep statistics */
2466     PIMSTAT_INC(pims_snd_registers_msgs);
2467     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2468 
2469     return 0;
2470 }
2471 
2472 /*
2473  * Encapsulate the data packet in PIM Register message and send it to the RP.
2474  */
2475 static int
2476 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2477     struct mfc *rt)
2478 {
2479     struct mbuf *mb_first;
2480     struct ip *ip_outer;
2481     struct pim_encap_pimhdr *pimhdr;
2482     int len = ntohs(ip->ip_len);
2483     vifi_t vifi = rt->mfc_parent;
2484 
2485     VIF_LOCK_ASSERT();
2486 
2487     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2488 	m_freem(mb_copy);
2489 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2490     }
2491 
2492     /*
2493      * Add a new mbuf with the encapsulating header
2494      */
2495     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2496     if (mb_first == NULL) {
2497 	m_freem(mb_copy);
2498 	return ENOBUFS;
2499     }
2500     mb_first->m_data += max_linkhdr;
2501     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2502     mb_first->m_next = mb_copy;
2503 
2504     mb_first->m_pkthdr.len = len + mb_first->m_len;
2505 
2506     /*
2507      * Fill in the encapsulating IP and PIM header
2508      */
2509     ip_outer = mtod(mb_first, struct ip *);
2510     *ip_outer = pim_encap_iphdr;
2511     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2512 	sizeof(pim_encap_pimhdr));
2513     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2514     ip_outer->ip_dst = rt->mfc_rp;
2515     /*
2516      * Copy the inner header TOS to the outer header, and take care of the
2517      * IP_DF bit.
2518      */
2519     ip_outer->ip_tos = ip->ip_tos;
2520     if (ip->ip_off & htons(IP_DF))
2521 	ip_outer->ip_off |= htons(IP_DF);
2522     ip_fillid(ip_outer);
2523     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2524 					 + sizeof(pim_encap_iphdr));
2525     *pimhdr = pim_encap_pimhdr;
2526     /* If the iif crosses a border, set the Border-bit */
2527     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2528 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2529 
2530     mb_first->m_data += sizeof(pim_encap_iphdr);
2531     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2532     mb_first->m_data -= sizeof(pim_encap_iphdr);
2533 
2534     send_packet(vifp, mb_first);
2535 
2536     /* Keep statistics */
2537     PIMSTAT_INC(pims_snd_registers_msgs);
2538     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2539 
2540     return 0;
2541 }
2542 
2543 /*
2544  * pim_encapcheck() is called by the encap4_input() path at runtime to
2545  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2546  * into the kernel.
2547  */
2548 static int
2549 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2550     int proto __unused, void *arg __unused)
2551 {
2552 
2553     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2554     return (8);		/* 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 static int
2566 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2567 {
2568     struct ip *ip = mtod(m, struct ip *);
2569     struct pim *pim;
2570     int iphlen = off;
2571     int minlen;
2572     int datalen = ntohs(ip->ip_len) - iphlen;
2573     int ip_tos;
2574 
2575     /* Keep statistics */
2576     PIMSTAT_INC(pims_rcv_total_msgs);
2577     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2578 
2579     /*
2580      * Validate lengths
2581      */
2582     if (datalen < PIM_MINLEN) {
2583 	PIMSTAT_INC(pims_rcv_tooshort);
2584 	CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2585 	    __func__, datalen, ntohl(ip->ip_src.s_addr));
2586 	m_freem(m);
2587 	return (IPPROTO_DONE);
2588     }
2589 
2590     /*
2591      * If the packet is at least as big as a REGISTER, go agead
2592      * and grab the PIM REGISTER header size, to avoid another
2593      * possible m_pullup() later.
2594      *
2595      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2596      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2597      */
2598     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2599     /*
2600      * Get the IP and PIM headers in contiguous memory, and
2601      * possibly the PIM REGISTER header.
2602      */
2603     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2604 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2605 	return (IPPROTO_DONE);
2606     }
2607 
2608     /* m_pullup() may have given us a new mbuf so reset ip. */
2609     ip = mtod(m, struct ip *);
2610     ip_tos = ip->ip_tos;
2611 
2612     /* adjust mbuf to point to the PIM header */
2613     m->m_data += iphlen;
2614     m->m_len  -= iphlen;
2615     pim = mtod(m, struct pim *);
2616 
2617     /*
2618      * Validate checksum. If PIM REGISTER, exclude the data packet.
2619      *
2620      * XXX: some older PIMv2 implementations don't make this distinction,
2621      * so for compatibility reason perform the checksum over part of the
2622      * message, and if error, then over the whole message.
2623      */
2624     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2625 	/* do nothing, checksum okay */
2626     } else if (in_cksum(m, datalen)) {
2627 	PIMSTAT_INC(pims_rcv_badsum);
2628 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2629 	m_freem(m);
2630 	return (IPPROTO_DONE);
2631     }
2632 
2633     /* PIM version check */
2634     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2635 	PIMSTAT_INC(pims_rcv_badversion);
2636 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2637 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2638 	m_freem(m);
2639 	return (IPPROTO_DONE);
2640     }
2641 
2642     /* restore mbuf back to the outer IP */
2643     m->m_data -= iphlen;
2644     m->m_len  += iphlen;
2645 
2646     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2647 	/*
2648 	 * Since this is a REGISTER, we'll make a copy of the register
2649 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2650 	 * routing daemon.
2651 	 */
2652 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2653 	struct mbuf *mcp;
2654 	struct ip *encap_ip;
2655 	u_int32_t *reghdr;
2656 	struct ifnet *vifp;
2657 
2658 	VIF_LOCK();
2659 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2660 	    VIF_UNLOCK();
2661 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2662 		(int)V_reg_vif_num);
2663 	    m_freem(m);
2664 	    return (IPPROTO_DONE);
2665 	}
2666 	/* XXX need refcnt? */
2667 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2668 	VIF_UNLOCK();
2669 
2670 	/*
2671 	 * Validate length
2672 	 */
2673 	if (datalen < PIM_REG_MINLEN) {
2674 	    PIMSTAT_INC(pims_rcv_tooshort);
2675 	    PIMSTAT_INC(pims_rcv_badregisters);
2676 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2677 	    m_freem(m);
2678 	    return (IPPROTO_DONE);
2679 	}
2680 
2681 	reghdr = (u_int32_t *)(pim + 1);
2682 	encap_ip = (struct ip *)(reghdr + 1);
2683 
2684 	CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2685 	    __func__, ntohl(encap_ip->ip_src.s_addr),
2686 	    ntohs(encap_ip->ip_len));
2687 
2688 	/* verify the version number of the inner packet */
2689 	if (encap_ip->ip_v != IPVERSION) {
2690 	    PIMSTAT_INC(pims_rcv_badregisters);
2691 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2692 	    m_freem(m);
2693 	    return (IPPROTO_DONE);
2694 	}
2695 
2696 	/* verify the inner packet is destined to a mcast group */
2697 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2698 	    PIMSTAT_INC(pims_rcv_badregisters);
2699 	    CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2700 		ntohl(encap_ip->ip_dst.s_addr));
2701 	    m_freem(m);
2702 	    return (IPPROTO_DONE);
2703 	}
2704 
2705 	/* If a NULL_REGISTER, pass it to the daemon */
2706 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2707 	    goto pim_input_to_daemon;
2708 
2709 	/*
2710 	 * Copy the TOS from the outer IP header to the inner IP header.
2711 	 */
2712 	if (encap_ip->ip_tos != ip_tos) {
2713 	    /* Outer TOS -> inner TOS */
2714 	    encap_ip->ip_tos = ip_tos;
2715 	    /* Recompute the inner header checksum. Sigh... */
2716 
2717 	    /* adjust mbuf to point to the inner IP header */
2718 	    m->m_data += (iphlen + PIM_MINLEN);
2719 	    m->m_len  -= (iphlen + PIM_MINLEN);
2720 
2721 	    encap_ip->ip_sum = 0;
2722 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2723 
2724 	    /* restore mbuf to point back to the outer IP header */
2725 	    m->m_data -= (iphlen + PIM_MINLEN);
2726 	    m->m_len  += (iphlen + PIM_MINLEN);
2727 	}
2728 
2729 	/*
2730 	 * Decapsulate the inner IP packet and loopback to forward it
2731 	 * as a normal multicast packet. Also, make a copy of the
2732 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2733 	 * to pass to the daemon later, so it can take the appropriate
2734 	 * actions (e.g., send back PIM_REGISTER_STOP).
2735 	 * XXX: here m->m_data points to the outer IP header.
2736 	 */
2737 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2738 	if (mcp == NULL) {
2739 	    CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2740 	    m_freem(m);
2741 	    return (IPPROTO_DONE);
2742 	}
2743 
2744 	/* Keep statistics */
2745 	/* XXX: registers_bytes include only the encap. mcast pkt */
2746 	PIMSTAT_INC(pims_rcv_registers_msgs);
2747 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2748 
2749 	/*
2750 	 * forward the inner ip packet; point m_data at the inner ip.
2751 	 */
2752 	m_adj(m, iphlen + PIM_MINLEN);
2753 
2754 	CTR4(KTR_IPMF,
2755 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2756 	    __func__,
2757 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2758 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2759 	    (int)V_reg_vif_num);
2760 
2761 	/* NB: vifp was collected above; can it change on us? */
2762 	if_simloop(vifp, m, dst.sin_family, 0);
2763 
2764 	/* prepare the register head to send to the mrouting daemon */
2765 	m = mcp;
2766     }
2767 
2768 pim_input_to_daemon:
2769     /*
2770      * Pass the PIM message up to the daemon; if it is a Register message,
2771      * pass the 'head' only up to the daemon. This includes the
2772      * outer IP header, PIM header, PIM-Register header and the
2773      * inner IP header.
2774      * XXX: the outer IP header pkt size of a Register is not adjust to
2775      * reflect the fact that the inner multicast data is truncated.
2776      */
2777     return (rip_input(&m, &off, proto));
2778 }
2779 
2780 static int
2781 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2782 {
2783 	struct mfc	*rt;
2784 	int		 error, i;
2785 
2786 	if (req->newptr)
2787 		return (EPERM);
2788 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2789 		return (0);
2790 	error = sysctl_wire_old_buffer(req, 0);
2791 	if (error)
2792 		return (error);
2793 
2794 	MFC_LOCK();
2795 	for (i = 0; i < mfchashsize; i++) {
2796 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2797 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2798 			if (error)
2799 				goto out_locked;
2800 		}
2801 	}
2802 out_locked:
2803 	MFC_UNLOCK();
2804 	return (error);
2805 }
2806 
2807 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2808     CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2809     "IPv4 Multicast Forwarding Table "
2810     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2811 
2812 static void
2813 vnet_mroute_init(const void *unused __unused)
2814 {
2815 
2816 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2817 
2818 	V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2819 	    M_MRTABLE, M_WAITOK|M_ZERO);
2820 	V_bw_meter_timers = mallocarray(BW_METER_BUCKETS,
2821 	    sizeof(*V_bw_meter_timers), M_MRTABLE, M_WAITOK|M_ZERO);
2822 	V_bw_upcalls = mallocarray(BW_UPCALLS_MAX, sizeof(*V_bw_upcalls),
2823 	    M_MRTABLE, M_WAITOK|M_ZERO);
2824 
2825 	callout_init(&V_expire_upcalls_ch, 1);
2826 	callout_init(&V_bw_upcalls_ch, 1);
2827 	callout_init(&V_bw_meter_ch, 1);
2828 }
2829 
2830 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2831 	NULL);
2832 
2833 static void
2834 vnet_mroute_uninit(const void *unused __unused)
2835 {
2836 
2837 	free(V_bw_upcalls, M_MRTABLE);
2838 	free(V_bw_meter_timers, M_MRTABLE);
2839 	free(V_viftable, M_MRTABLE);
2840 	free(V_nexpire, M_MRTABLE);
2841 	V_nexpire = NULL;
2842 }
2843 
2844 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2845 	vnet_mroute_uninit, NULL);
2846 
2847 static int
2848 ip_mroute_modevent(module_t mod, int type, void *unused)
2849 {
2850 
2851     switch (type) {
2852     case MOD_LOAD:
2853 	MROUTER_LOCK_INIT();
2854 
2855 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2856 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2857 	if (if_detach_event_tag == NULL) {
2858 		printf("ip_mroute: unable to register "
2859 		    "ifnet_departure_event handler\n");
2860 		MROUTER_LOCK_DESTROY();
2861 		return (EINVAL);
2862 	}
2863 
2864 	MFC_LOCK_INIT();
2865 	VIF_LOCK_INIT();
2866 
2867 	mfchashsize = MFCHASHSIZE;
2868 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2869 	    !powerof2(mfchashsize)) {
2870 		printf("WARNING: %s not a power of 2; using default\n",
2871 		    "net.inet.ip.mfchashsize");
2872 		mfchashsize = MFCHASHSIZE;
2873 	}
2874 
2875 	pim_squelch_wholepkt = 0;
2876 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2877 	    &pim_squelch_wholepkt);
2878 
2879 	pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2880 	if (pim_encap_cookie == NULL) {
2881 		printf("ip_mroute: unable to attach pim encap\n");
2882 		VIF_LOCK_DESTROY();
2883 		MFC_LOCK_DESTROY();
2884 		MROUTER_LOCK_DESTROY();
2885 		return (EINVAL);
2886 	}
2887 
2888 	ip_mcast_src = X_ip_mcast_src;
2889 	ip_mforward = X_ip_mforward;
2890 	ip_mrouter_done = X_ip_mrouter_done;
2891 	ip_mrouter_get = X_ip_mrouter_get;
2892 	ip_mrouter_set = X_ip_mrouter_set;
2893 
2894 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2895 	ip_rsvp_vif = X_ip_rsvp_vif;
2896 
2897 	legal_vif_num = X_legal_vif_num;
2898 	mrt_ioctl = X_mrt_ioctl;
2899 	rsvp_input_p = X_rsvp_input;
2900 	break;
2901 
2902     case MOD_UNLOAD:
2903 	/*
2904 	 * Typically module unload happens after the user-level
2905 	 * process has shutdown the kernel services (the check
2906 	 * below insures someone can't just yank the module out
2907 	 * from under a running process).  But if the module is
2908 	 * just loaded and then unloaded w/o starting up a user
2909 	 * process we still need to cleanup.
2910 	 */
2911 	MROUTER_LOCK();
2912 	if (ip_mrouter_cnt != 0) {
2913 	    MROUTER_UNLOCK();
2914 	    return (EINVAL);
2915 	}
2916 	ip_mrouter_unloading = 1;
2917 	MROUTER_UNLOCK();
2918 
2919 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2920 
2921 	if (pim_encap_cookie) {
2922 	    ip_encap_detach(pim_encap_cookie);
2923 	    pim_encap_cookie = NULL;
2924 	}
2925 
2926 	ip_mcast_src = NULL;
2927 	ip_mforward = NULL;
2928 	ip_mrouter_done = NULL;
2929 	ip_mrouter_get = NULL;
2930 	ip_mrouter_set = NULL;
2931 
2932 	ip_rsvp_force_done = NULL;
2933 	ip_rsvp_vif = NULL;
2934 
2935 	legal_vif_num = NULL;
2936 	mrt_ioctl = NULL;
2937 	rsvp_input_p = NULL;
2938 
2939 	VIF_LOCK_DESTROY();
2940 	MFC_LOCK_DESTROY();
2941 	MROUTER_LOCK_DESTROY();
2942 	break;
2943 
2944     default:
2945 	return EOPNOTSUPP;
2946     }
2947     return 0;
2948 }
2949 
2950 static moduledata_t ip_mroutemod = {
2951     "ip_mroute",
2952     ip_mroute_modevent,
2953     0
2954 };
2955 
2956 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2957