xref: /freebsd/sys/netinet/ip_mroute.c (revision 595e514d0df2bac5b813d35f83e32875dbf16a83)
1 /*-
2  * Copyright (c) 1989 Stephen Deering
3  * Copyright (c) 1992, 1993
4  *      The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Stephen Deering of Stanford University.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
34  */
35 
36 /*
37  * IP multicast forwarding procedures
38  *
39  * Written by David Waitzman, BBN Labs, August 1988.
40  * Modified by Steve Deering, Stanford, February 1989.
41  * Modified by Mark J. Steiglitz, Stanford, May, 1991
42  * Modified by Van Jacobson, LBL, January 1993
43  * Modified by Ajit Thyagarajan, PARC, August 1993
44  * Modified by Bill Fenner, PARC, April 1995
45  * Modified by Ahmed Helmy, SGI, June 1996
46  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
47  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
48  * Modified by Hitoshi Asaeda, WIDE, August 2000
49  * Modified by Pavlin Radoslavov, ICSI, October 2002
50  *
51  * MROUTING Revision: 3.5
52  * and PIM-SMv2 and PIM-DM support, advanced API support,
53  * bandwidth metering and signaling
54  */
55 
56 /*
57  * TODO: Prefix functions with ipmf_.
58  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
59  * domain attachment (if_afdata) so we can track consumers of that service.
60  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
61  * move it to socket options.
62  * TODO: Cleanup LSRR removal further.
63  * TODO: Push RSVP stubs into raw_ip.c.
64  * TODO: Use bitstring.h for vif set.
65  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
66  * TODO: Sync ip6_mroute.c with this file.
67  */
68 
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
71 
72 #include "opt_inet.h"
73 #include "opt_mrouting.h"
74 
75 #define _PIM_VT 1
76 
77 #include <sys/param.h>
78 #include <sys/kernel.h>
79 #include <sys/stddef.h>
80 #include <sys/lock.h>
81 #include <sys/ktr.h>
82 #include <sys/malloc.h>
83 #include <sys/mbuf.h>
84 #include <sys/module.h>
85 #include <sys/priv.h>
86 #include <sys/protosw.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/socketvar.h>
90 #include <sys/sockio.h>
91 #include <sys/sx.h>
92 #include <sys/sysctl.h>
93 #include <sys/syslog.h>
94 #include <sys/systm.h>
95 #include <sys/time.h>
96 
97 #include <net/if.h>
98 #include <net/netisr.h>
99 #include <net/route.h>
100 #include <net/vnet.h>
101 
102 #include <netinet/in.h>
103 #include <netinet/igmp.h>
104 #include <netinet/in_systm.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip.h>
107 #include <netinet/ip_encap.h>
108 #include <netinet/ip_mroute.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/ip_options.h>
111 #include <netinet/pim.h>
112 #include <netinet/pim_var.h>
113 #include <netinet/udp.h>
114 
115 #include <machine/in_cksum.h>
116 
117 #ifndef KTR_IPMF
118 #define KTR_IPMF KTR_INET
119 #endif
120 
121 #define		VIFI_INVALID	((vifi_t) -1)
122 #define		M_HASCL(m)	((m)->m_flags & M_EXT)
123 
124 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
125 #define	V_last_tv_sec	VNET(last_tv_sec)
126 
127 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
128 
129 /*
130  * Locking.  We use two locks: one for the virtual interface table and
131  * one for the forwarding table.  These locks may be nested in which case
132  * the VIF lock must always be taken first.  Note that each lock is used
133  * to cover not only the specific data structure but also related data
134  * structures.
135  */
136 
137 static struct mtx mrouter_mtx;
138 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
139 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
140 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
141 #define	MROUTER_LOCK_INIT()						\
142 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
143 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
144 
145 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
146 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
147 
148 static VNET_DEFINE(struct mrtstat, mrtstat);
149 #define	V_mrtstat		VNET(mrtstat)
150 SYSCTL_VNET_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
151     &VNET_NAME(mrtstat), mrtstat,
152     "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
153     "netinet/ip_mroute.h)");
154 
155 static VNET_DEFINE(u_long, mfchash);
156 #define	V_mfchash		VNET(mfchash)
157 #define	MFCHASH(a, g)							\
158 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
159 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
160 #define	MFCHASHSIZE	256
161 
162 static u_long mfchashsize;			/* Hash size */
163 static VNET_DEFINE(u_char *, nexpire);		/* 0..mfchashsize-1 */
164 #define	V_nexpire		VNET(nexpire)
165 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
166 #define	V_mfchashtbl		VNET(mfchashtbl)
167 
168 static struct mtx mfc_mtx;
169 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
170 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
171 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
172 #define	MFC_LOCK_INIT()							\
173 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
174 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
175 
176 static VNET_DEFINE(vifi_t, numvifs);
177 #define	V_numvifs		VNET(numvifs)
178 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
179 #define	V_viftable		VNET(viftable)
180 SYSCTL_VNET_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
181     &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
182     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
183 
184 static struct mtx vif_mtx;
185 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
186 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
187 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
188 #define	VIF_LOCK_INIT()							\
189 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
190 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
191 
192 static eventhandler_tag if_detach_event_tag = NULL;
193 
194 static VNET_DEFINE(struct callout, expire_upcalls_ch);
195 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
196 
197 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
198 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
199 
200 /*
201  * Bandwidth meter variables and constants
202  */
203 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
204 /*
205  * Pending timeouts are stored in a hash table, the key being the
206  * expiration time. Periodically, the entries are analysed and processed.
207  */
208 #define	BW_METER_BUCKETS	1024
209 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
210 #define	V_bw_meter_timers	VNET(bw_meter_timers)
211 static VNET_DEFINE(struct callout, bw_meter_ch);
212 #define	V_bw_meter_ch		VNET(bw_meter_ch)
213 #define	BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
214 
215 /*
216  * Pending upcalls are stored in a vector which is flushed when
217  * full, or periodically
218  */
219 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
220 #define	V_bw_upcalls		VNET(bw_upcalls)
221 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
222 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
223 static VNET_DEFINE(struct callout, bw_upcalls_ch);
224 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
225 
226 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
227 
228 static VNET_DEFINE(struct pimstat, pimstat);
229 #define	V_pimstat		VNET(pimstat)
230 
231 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
232 SYSCTL_VNET_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
233     &VNET_NAME(pimstat), pimstat,
234     "PIM Statistics (struct pimstat, netinet/pim_var.h)");
235 
236 static u_long	pim_squelch_wholepkt = 0;
237 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
238     &pim_squelch_wholepkt, 0,
239     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
240 
241 extern  struct domain inetdomain;
242 static const struct protosw in_pim_protosw = {
243 	.pr_type =		SOCK_RAW,
244 	.pr_domain =		&inetdomain,
245 	.pr_protocol =		IPPROTO_PIM,
246 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_LASTHDR,
247 	.pr_input =		pim_input,
248 	.pr_output =		(pr_output_t*)rip_output,
249 	.pr_ctloutput =		rip_ctloutput,
250 	.pr_usrreqs =		&rip_usrreqs
251 };
252 static const struct encaptab *pim_encap_cookie;
253 
254 static int pim_encapcheck(const struct mbuf *, int, int, void *);
255 
256 /*
257  * Note: the PIM Register encapsulation adds the following in front of a
258  * data packet:
259  *
260  * struct pim_encap_hdr {
261  *    struct ip ip;
262  *    struct pim_encap_pimhdr  pim;
263  * }
264  *
265  */
266 
267 struct pim_encap_pimhdr {
268 	struct pim pim;
269 	uint32_t   flags;
270 };
271 #define		PIM_ENCAP_TTL	64
272 
273 static struct ip pim_encap_iphdr = {
274 #if BYTE_ORDER == LITTLE_ENDIAN
275 	sizeof(struct ip) >> 2,
276 	IPVERSION,
277 #else
278 	IPVERSION,
279 	sizeof(struct ip) >> 2,
280 #endif
281 	0,			/* tos */
282 	sizeof(struct ip),	/* total length */
283 	0,			/* id */
284 	0,			/* frag offset */
285 	PIM_ENCAP_TTL,
286 	IPPROTO_PIM,
287 	0,			/* checksum */
288 };
289 
290 static struct pim_encap_pimhdr pim_encap_pimhdr = {
291     {
292 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
293 	0,			/* reserved */
294 	0,			/* checksum */
295     },
296     0				/* flags */
297 };
298 
299 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
300 #define	V_reg_vif_num		VNET(reg_vif_num)
301 static VNET_DEFINE(struct ifnet, multicast_register_if);
302 #define	V_multicast_register_if	VNET(multicast_register_if)
303 
304 /*
305  * Private variables.
306  */
307 
308 static u_long	X_ip_mcast_src(int);
309 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
310 		    struct ip_moptions *);
311 static int	X_ip_mrouter_done(void);
312 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
313 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
314 static int	X_legal_vif_num(int);
315 static int	X_mrt_ioctl(u_long, caddr_t, int);
316 
317 static int	add_bw_upcall(struct bw_upcall *);
318 static int	add_mfc(struct mfcctl2 *);
319 static int	add_vif(struct vifctl *);
320 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
321 static void	bw_meter_process(void);
322 static void	bw_meter_receive_packet(struct bw_meter *, int,
323 		    struct timeval *);
324 static void	bw_upcalls_send(void);
325 static int	del_bw_upcall(struct bw_upcall *);
326 static int	del_mfc(struct mfcctl2 *);
327 static int	del_vif(vifi_t);
328 static int	del_vif_locked(vifi_t);
329 static void	expire_bw_meter_process(void *);
330 static void	expire_bw_upcalls_send(void *);
331 static void	expire_mfc(struct mfc *);
332 static void	expire_upcalls(void *);
333 static void	free_bw_list(struct bw_meter *);
334 static int	get_sg_cnt(struct sioc_sg_req *);
335 static int	get_vif_cnt(struct sioc_vif_req *);
336 static void	if_detached_event(void *, struct ifnet *);
337 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
338 static int	ip_mrouter_init(struct socket *, int);
339 static __inline struct mfc *
340 		mfc_find(struct in_addr *, struct in_addr *);
341 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
342 static struct mbuf *
343 		pim_register_prepare(struct ip *, struct mbuf *);
344 static int	pim_register_send(struct ip *, struct vif *,
345 		    struct mbuf *, struct mfc *);
346 static int	pim_register_send_rp(struct ip *, struct vif *,
347 		    struct mbuf *, struct mfc *);
348 static int	pim_register_send_upcall(struct ip *, struct vif *,
349 		    struct mbuf *, struct mfc *);
350 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
351 static void	send_packet(struct vif *, struct mbuf *);
352 static int	set_api_config(uint32_t *);
353 static int	set_assert(int);
354 static int	socket_send(struct socket *, struct mbuf *,
355 		    struct sockaddr_in *);
356 static void	unschedule_bw_meter(struct bw_meter *);
357 
358 /*
359  * Kernel multicast forwarding API capabilities and setup.
360  * If more API capabilities are added to the kernel, they should be
361  * recorded in `mrt_api_support'.
362  */
363 #define MRT_API_VERSION		0x0305
364 
365 static const int mrt_api_version = MRT_API_VERSION;
366 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
367 					 MRT_MFC_FLAGS_BORDER_VIF |
368 					 MRT_MFC_RP |
369 					 MRT_MFC_BW_UPCALL);
370 static VNET_DEFINE(uint32_t, mrt_api_config);
371 #define	V_mrt_api_config	VNET(mrt_api_config)
372 static VNET_DEFINE(int, pim_assert_enabled);
373 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
374 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
375 
376 /*
377  * Find a route for a given origin IP address and multicast group address.
378  * Statistics must be updated by the caller.
379  */
380 static __inline struct mfc *
381 mfc_find(struct in_addr *o, struct in_addr *g)
382 {
383 	struct mfc *rt;
384 
385 	MFC_LOCK_ASSERT();
386 
387 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
388 		if (in_hosteq(rt->mfc_origin, *o) &&
389 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
390 		    TAILQ_EMPTY(&rt->mfc_stall))
391 			break;
392 	}
393 
394 	return (rt);
395 }
396 
397 /*
398  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
399  */
400 static int
401 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
402 {
403     int	error, optval;
404     vifi_t	vifi;
405     struct	vifctl vifc;
406     struct	mfcctl2 mfc;
407     struct	bw_upcall bw_upcall;
408     uint32_t	i;
409 
410     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
411 	return EPERM;
412 
413     error = 0;
414     switch (sopt->sopt_name) {
415     case MRT_INIT:
416 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
417 	if (error)
418 	    break;
419 	error = ip_mrouter_init(so, optval);
420 	break;
421 
422     case MRT_DONE:
423 	error = ip_mrouter_done();
424 	break;
425 
426     case MRT_ADD_VIF:
427 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
428 	if (error)
429 	    break;
430 	error = add_vif(&vifc);
431 	break;
432 
433     case MRT_DEL_VIF:
434 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
435 	if (error)
436 	    break;
437 	error = del_vif(vifi);
438 	break;
439 
440     case MRT_ADD_MFC:
441     case MRT_DEL_MFC:
442 	/*
443 	 * select data size depending on API version.
444 	 */
445 	if (sopt->sopt_name == MRT_ADD_MFC &&
446 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
447 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
448 				sizeof(struct mfcctl2));
449 	} else {
450 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
451 				sizeof(struct mfcctl));
452 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
453 			sizeof(mfc) - sizeof(struct mfcctl));
454 	}
455 	if (error)
456 	    break;
457 	if (sopt->sopt_name == MRT_ADD_MFC)
458 	    error = add_mfc(&mfc);
459 	else
460 	    error = del_mfc(&mfc);
461 	break;
462 
463     case MRT_ASSERT:
464 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
465 	if (error)
466 	    break;
467 	set_assert(optval);
468 	break;
469 
470     case MRT_API_CONFIG:
471 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
472 	if (!error)
473 	    error = set_api_config(&i);
474 	if (!error)
475 	    error = sooptcopyout(sopt, &i, sizeof i);
476 	break;
477 
478     case MRT_ADD_BW_UPCALL:
479     case MRT_DEL_BW_UPCALL:
480 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
481 				sizeof bw_upcall);
482 	if (error)
483 	    break;
484 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
485 	    error = add_bw_upcall(&bw_upcall);
486 	else
487 	    error = del_bw_upcall(&bw_upcall);
488 	break;
489 
490     default:
491 	error = EOPNOTSUPP;
492 	break;
493     }
494     return error;
495 }
496 
497 /*
498  * Handle MRT getsockopt commands
499  */
500 static int
501 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
502 {
503     int error;
504 
505     switch (sopt->sopt_name) {
506     case MRT_VERSION:
507 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
508 	break;
509 
510     case MRT_ASSERT:
511 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
512 	    sizeof V_pim_assert_enabled);
513 	break;
514 
515     case MRT_API_SUPPORT:
516 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
517 	break;
518 
519     case MRT_API_CONFIG:
520 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
521 	break;
522 
523     default:
524 	error = EOPNOTSUPP;
525 	break;
526     }
527     return error;
528 }
529 
530 /*
531  * Handle ioctl commands to obtain information from the cache
532  */
533 static int
534 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
535 {
536     int error = 0;
537 
538     /*
539      * Currently the only function calling this ioctl routine is rtioctl().
540      * Typically, only root can create the raw socket in order to execute
541      * this ioctl method, however the request might be coming from a prison
542      */
543     error = priv_check(curthread, PRIV_NETINET_MROUTE);
544     if (error)
545 	return (error);
546     switch (cmd) {
547     case (SIOCGETVIFCNT):
548 	error = get_vif_cnt((struct sioc_vif_req *)data);
549 	break;
550 
551     case (SIOCGETSGCNT):
552 	error = get_sg_cnt((struct sioc_sg_req *)data);
553 	break;
554 
555     default:
556 	error = EINVAL;
557 	break;
558     }
559     return error;
560 }
561 
562 /*
563  * returns the packet, byte, rpf-failure count for the source group provided
564  */
565 static int
566 get_sg_cnt(struct sioc_sg_req *req)
567 {
568     struct mfc *rt;
569 
570     MFC_LOCK();
571     rt = mfc_find(&req->src, &req->grp);
572     if (rt == NULL) {
573 	MFC_UNLOCK();
574 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
575 	return EADDRNOTAVAIL;
576     }
577     req->pktcnt = rt->mfc_pkt_cnt;
578     req->bytecnt = rt->mfc_byte_cnt;
579     req->wrong_if = rt->mfc_wrong_if;
580     MFC_UNLOCK();
581     return 0;
582 }
583 
584 /*
585  * returns the input and output packet and byte counts on the vif provided
586  */
587 static int
588 get_vif_cnt(struct sioc_vif_req *req)
589 {
590     vifi_t vifi = req->vifi;
591 
592     VIF_LOCK();
593     if (vifi >= V_numvifs) {
594 	VIF_UNLOCK();
595 	return EINVAL;
596     }
597 
598     req->icount = V_viftable[vifi].v_pkt_in;
599     req->ocount = V_viftable[vifi].v_pkt_out;
600     req->ibytes = V_viftable[vifi].v_bytes_in;
601     req->obytes = V_viftable[vifi].v_bytes_out;
602     VIF_UNLOCK();
603 
604     return 0;
605 }
606 
607 static void
608 if_detached_event(void *arg __unused, struct ifnet *ifp)
609 {
610     vifi_t vifi;
611     int i;
612 
613     MROUTER_LOCK();
614 
615     if (V_ip_mrouter == NULL) {
616 	MROUTER_UNLOCK();
617 	return;
618     }
619 
620     VIF_LOCK();
621     MFC_LOCK();
622 
623     /*
624      * Tear down multicast forwarder state associated with this ifnet.
625      * 1. Walk the vif list, matching vifs against this ifnet.
626      * 2. Walk the multicast forwarding cache (mfc) looking for
627      *    inner matches with this vif's index.
628      * 3. Expire any matching multicast forwarding cache entries.
629      * 4. Free vif state. This should disable ALLMULTI on the interface.
630      */
631     for (vifi = 0; vifi < V_numvifs; vifi++) {
632 	if (V_viftable[vifi].v_ifp != ifp)
633 		continue;
634 	for (i = 0; i < mfchashsize; i++) {
635 		struct mfc *rt, *nrt;
636 		for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
637 			nrt = LIST_NEXT(rt, mfc_hash);
638 			if (rt->mfc_parent == vifi) {
639 				expire_mfc(rt);
640 			}
641 		}
642 	}
643 	del_vif_locked(vifi);
644     }
645 
646     MFC_UNLOCK();
647     VIF_UNLOCK();
648 
649     MROUTER_UNLOCK();
650 }
651 
652 /*
653  * Enable multicast forwarding.
654  */
655 static int
656 ip_mrouter_init(struct socket *so, int version)
657 {
658 
659     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
660         so->so_type, so->so_proto->pr_protocol);
661 
662     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
663 	return EOPNOTSUPP;
664 
665     if (version != 1)
666 	return ENOPROTOOPT;
667 
668     MROUTER_LOCK();
669 
670     if (ip_mrouter_unloading) {
671 	MROUTER_UNLOCK();
672 	return ENOPROTOOPT;
673     }
674 
675     if (V_ip_mrouter != NULL) {
676 	MROUTER_UNLOCK();
677 	return EADDRINUSE;
678     }
679 
680     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
681 	HASH_NOWAIT);
682 
683     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
684 	curvnet);
685     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
686 	curvnet);
687     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
688 	curvnet);
689 
690     V_ip_mrouter = so;
691     ip_mrouter_cnt++;
692 
693     MROUTER_UNLOCK();
694 
695     CTR1(KTR_IPMF, "%s: done", __func__);
696 
697     return 0;
698 }
699 
700 /*
701  * Disable multicast forwarding.
702  */
703 static int
704 X_ip_mrouter_done(void)
705 {
706     vifi_t vifi;
707     int i;
708     struct ifnet *ifp;
709     struct ifreq ifr;
710 
711     MROUTER_LOCK();
712 
713     if (V_ip_mrouter == NULL) {
714 	MROUTER_UNLOCK();
715 	return EINVAL;
716     }
717 
718     /*
719      * Detach/disable hooks to the reset of the system.
720      */
721     V_ip_mrouter = NULL;
722     ip_mrouter_cnt--;
723     V_mrt_api_config = 0;
724 
725     VIF_LOCK();
726 
727     /*
728      * For each phyint in use, disable promiscuous reception of all IP
729      * multicasts.
730      */
731     for (vifi = 0; vifi < V_numvifs; vifi++) {
732 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
733 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
734 	    struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
735 
736 	    so->sin_len = sizeof(struct sockaddr_in);
737 	    so->sin_family = AF_INET;
738 	    so->sin_addr.s_addr = INADDR_ANY;
739 	    ifp = V_viftable[vifi].v_ifp;
740 	    if_allmulti(ifp, 0);
741 	}
742     }
743     bzero((caddr_t)V_viftable, sizeof(V_viftable));
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 	for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
762 		nrt = LIST_NEXT(rt, mfc_hash);
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, 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     int 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 	sin.sin_addr = vifcp->vifc_lcl_addr;
883 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
884 	if (ifa == NULL) {
885 	    VIF_UNLOCK();
886 	    return EADDRNOTAVAIL;
887 	}
888 	ifp = ifa->ifa_ifp;
889 	ifa_free(ifa);
890     }
891 
892     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
893 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
894 	VIF_UNLOCK();
895 	return EOPNOTSUPP;
896     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
897 	ifp = &V_multicast_register_if;
898 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
899 	if (V_reg_vif_num == VIFI_INVALID) {
900 	    if_initname(&V_multicast_register_if, "register_vif", 0);
901 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
902 	    V_reg_vif_num = vifcp->vifc_vifi;
903 	}
904     } else {		/* Make sure the interface supports multicast */
905 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
906 	    VIF_UNLOCK();
907 	    return EOPNOTSUPP;
908 	}
909 
910 	/* Enable promiscuous reception of all IP multicasts from the if */
911 	error = if_allmulti(ifp, 1);
912 	if (error) {
913 	    VIF_UNLOCK();
914 	    return error;
915 	}
916     }
917 
918     vifp->v_flags     = vifcp->vifc_flags;
919     vifp->v_threshold = vifcp->vifc_threshold;
920     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
921     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
922     vifp->v_ifp       = ifp;
923     /* initialize per vif pkt counters */
924     vifp->v_pkt_in    = 0;
925     vifp->v_pkt_out   = 0;
926     vifp->v_bytes_in  = 0;
927     vifp->v_bytes_out = 0;
928 
929     /* Adjust numvifs up if the vifi is higher than numvifs */
930     if (V_numvifs <= vifcp->vifc_vifi)
931 	V_numvifs = vifcp->vifc_vifi + 1;
932 
933     VIF_UNLOCK();
934 
935     CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
936 	(int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr),
937 	(int)vifcp->vifc_threshold);
938 
939     return 0;
940 }
941 
942 /*
943  * Delete a vif from the vif table
944  */
945 static int
946 del_vif_locked(vifi_t vifi)
947 {
948     struct vif *vifp;
949 
950     VIF_LOCK_ASSERT();
951 
952     if (vifi >= V_numvifs) {
953 	return EINVAL;
954     }
955     vifp = &V_viftable[vifi];
956     if (in_nullhost(vifp->v_lcl_addr)) {
957 	return EADDRNOTAVAIL;
958     }
959 
960     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
961 	if_allmulti(vifp->v_ifp, 0);
962 
963     if (vifp->v_flags & VIFF_REGISTER)
964 	V_reg_vif_num = VIFI_INVALID;
965 
966     bzero((caddr_t)vifp, sizeof (*vifp));
967 
968     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
969 
970     /* Adjust numvifs down */
971     for (vifi = V_numvifs; vifi > 0; vifi--)
972 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
973 	    break;
974     V_numvifs = vifi;
975 
976     return 0;
977 }
978 
979 static int
980 del_vif(vifi_t vifi)
981 {
982     int cc;
983 
984     VIF_LOCK();
985     cc = del_vif_locked(vifi);
986     VIF_UNLOCK();
987 
988     return cc;
989 }
990 
991 /*
992  * update an mfc entry without resetting counters and S,G addresses.
993  */
994 static void
995 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
996 {
997     int i;
998 
999     rt->mfc_parent = mfccp->mfcc_parent;
1000     for (i = 0; i < V_numvifs; i++) {
1001 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1002 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1003 	    MRT_MFC_FLAGS_ALL;
1004     }
1005     /* set the RP address */
1006     if (V_mrt_api_config & MRT_MFC_RP)
1007 	rt->mfc_rp = mfccp->mfcc_rp;
1008     else
1009 	rt->mfc_rp.s_addr = INADDR_ANY;
1010 }
1011 
1012 /*
1013  * fully initialize an mfc entry from the parameter.
1014  */
1015 static void
1016 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1017 {
1018     rt->mfc_origin     = mfccp->mfcc_origin;
1019     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1020 
1021     update_mfc_params(rt, mfccp);
1022 
1023     /* initialize pkt counters per src-grp */
1024     rt->mfc_pkt_cnt    = 0;
1025     rt->mfc_byte_cnt   = 0;
1026     rt->mfc_wrong_if   = 0;
1027     timevalclear(&rt->mfc_last_assert);
1028 }
1029 
1030 static void
1031 expire_mfc(struct mfc *rt)
1032 {
1033 	struct rtdetq *rte, *nrte;
1034 
1035 	MFC_LOCK_ASSERT();
1036 
1037 	free_bw_list(rt->mfc_bw_meter);
1038 
1039 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1040 		m_freem(rte->m);
1041 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1042 		free(rte, M_MRTABLE);
1043 	}
1044 
1045 	LIST_REMOVE(rt, mfc_hash);
1046 	free(rt, M_MRTABLE);
1047 }
1048 
1049 /*
1050  * Add an mfc entry
1051  */
1052 static int
1053 add_mfc(struct mfcctl2 *mfccp)
1054 {
1055     struct mfc *rt;
1056     struct rtdetq *rte, *nrte;
1057     u_long hash = 0;
1058     u_short nstl;
1059 
1060     VIF_LOCK();
1061     MFC_LOCK();
1062 
1063     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1064 
1065     /* If an entry already exists, just update the fields */
1066     if (rt) {
1067 	CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x",
1068 	    __func__, inet_ntoa(mfccp->mfcc_origin),
1069 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1070 	    mfccp->mfcc_parent);
1071 	update_mfc_params(rt, mfccp);
1072 	MFC_UNLOCK();
1073 	VIF_UNLOCK();
1074 	return (0);
1075     }
1076 
1077     /*
1078      * Find the entry for which the upcall was made and update
1079      */
1080     nstl = 0;
1081     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1082     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1083 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1084 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1085 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1086 		CTR5(KTR_IPMF,
1087 		    "%s: add mfc orig %s group %lx parent %x qh %p",
1088 		    __func__, inet_ntoa(mfccp->mfcc_origin),
1089 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1090 		    mfccp->mfcc_parent,
1091 		    TAILQ_FIRST(&rt->mfc_stall));
1092 		if (nstl++)
1093 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1094 
1095 		init_mfc_params(rt, mfccp);
1096 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1097 		V_nexpire[hash]--;
1098 
1099 		/* Free queued packets, but attempt to forward them first. */
1100 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1101 			if (rte->ifp != NULL)
1102 				ip_mdq(rte->m, rte->ifp, rt, -1);
1103 			m_freem(rte->m);
1104 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1105 			rt->mfc_nstall--;
1106 			free(rte, M_MRTABLE);
1107 		}
1108 	}
1109     }
1110 
1111     /*
1112      * It is possible that an entry is being inserted without an upcall
1113      */
1114     if (nstl == 0) {
1115 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1116 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1117 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1118 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1119 			init_mfc_params(rt, mfccp);
1120 			if (rt->mfc_expire)
1121 			    V_nexpire[hash]--;
1122 			rt->mfc_expire = 0;
1123 			break; /* XXX */
1124 		}
1125 	}
1126 
1127 	if (rt == NULL) {		/* no upcall, so make a new entry */
1128 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1129 	    if (rt == NULL) {
1130 		MFC_UNLOCK();
1131 		VIF_UNLOCK();
1132 		return (ENOBUFS);
1133 	    }
1134 
1135 	    init_mfc_params(rt, mfccp);
1136 	    TAILQ_INIT(&rt->mfc_stall);
1137 	    rt->mfc_nstall = 0;
1138 
1139 	    rt->mfc_expire     = 0;
1140 	    rt->mfc_bw_meter = NULL;
1141 
1142 	    /* insert new entry at head of hash chain */
1143 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1144 	}
1145     }
1146 
1147     MFC_UNLOCK();
1148     VIF_UNLOCK();
1149 
1150     return (0);
1151 }
1152 
1153 /*
1154  * Delete an mfc entry
1155  */
1156 static int
1157 del_mfc(struct mfcctl2 *mfccp)
1158 {
1159     struct in_addr	origin;
1160     struct in_addr	mcastgrp;
1161     struct mfc		*rt;
1162 
1163     origin = mfccp->mfcc_origin;
1164     mcastgrp = mfccp->mfcc_mcastgrp;
1165 
1166     CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__,
1167 	inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr));
1168 
1169     MFC_LOCK();
1170 
1171     rt = mfc_find(&origin, &mcastgrp);
1172     if (rt == NULL) {
1173 	MFC_UNLOCK();
1174 	return EADDRNOTAVAIL;
1175     }
1176 
1177     /*
1178      * free the bw_meter entries
1179      */
1180     free_bw_list(rt->mfc_bw_meter);
1181     rt->mfc_bw_meter = NULL;
1182 
1183     LIST_REMOVE(rt, mfc_hash);
1184     free(rt, M_MRTABLE);
1185 
1186     MFC_UNLOCK();
1187 
1188     return (0);
1189 }
1190 
1191 /*
1192  * Send a message to the routing daemon on the multicast routing socket.
1193  */
1194 static int
1195 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1196 {
1197     if (s) {
1198 	SOCKBUF_LOCK(&s->so_rcv);
1199 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1200 	    NULL) != 0) {
1201 	    sorwakeup_locked(s);
1202 	    return 0;
1203 	}
1204 	SOCKBUF_UNLOCK(&s->so_rcv);
1205     }
1206     m_freem(mm);
1207     return -1;
1208 }
1209 
1210 /*
1211  * IP multicast forwarding function. This function assumes that the packet
1212  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1213  * pointed to by "ifp", and the packet is to be relayed to other networks
1214  * that have members of the packet's destination IP multicast group.
1215  *
1216  * The packet is returned unscathed to the caller, unless it is
1217  * erroneous, in which case a non-zero return value tells the caller to
1218  * discard it.
1219  */
1220 
1221 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1222 
1223 static int
1224 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1225     struct ip_moptions *imo)
1226 {
1227     struct mfc *rt;
1228     int error;
1229     vifi_t vifi;
1230 
1231     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p",
1232 	inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1233 
1234     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1235 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1236 	/*
1237 	 * Packet arrived via a physical interface or
1238 	 * an encapsulated tunnel or a register_vif.
1239 	 */
1240     } else {
1241 	/*
1242 	 * Packet arrived through a source-route tunnel.
1243 	 * Source-route tunnels are no longer supported.
1244 	 */
1245 	return (1);
1246     }
1247 
1248     VIF_LOCK();
1249     MFC_LOCK();
1250     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1251 	if (ip->ip_ttl < MAXTTL)
1252 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1253 	error = ip_mdq(m, ifp, NULL, vifi);
1254 	MFC_UNLOCK();
1255 	VIF_UNLOCK();
1256 	return error;
1257     }
1258 
1259     /*
1260      * Don't forward a packet with time-to-live of zero or one,
1261      * or a packet destined to a local-only group.
1262      */
1263     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1264 	MFC_UNLOCK();
1265 	VIF_UNLOCK();
1266 	return 0;
1267     }
1268 
1269     /*
1270      * Determine forwarding vifs from the forwarding cache table
1271      */
1272     MRTSTAT_INC(mrts_mfc_lookups);
1273     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1274 
1275     /* Entry exists, so forward if necessary */
1276     if (rt != NULL) {
1277 	error = ip_mdq(m, ifp, rt, -1);
1278 	MFC_UNLOCK();
1279 	VIF_UNLOCK();
1280 	return error;
1281     } else {
1282 	/*
1283 	 * If we don't have a route for packet's origin,
1284 	 * Make a copy of the packet & send message to routing daemon
1285 	 */
1286 
1287 	struct mbuf *mb0;
1288 	struct rtdetq *rte;
1289 	u_long hash;
1290 	int hlen = ip->ip_hl << 2;
1291 
1292 	MRTSTAT_INC(mrts_mfc_misses);
1293 	MRTSTAT_INC(mrts_no_route);
1294 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)",
1295 	    inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr));
1296 
1297 	/*
1298 	 * Allocate mbufs early so that we don't do extra work if we are
1299 	 * just going to fail anyway.  Make sure to pullup the header so
1300 	 * that other people can't step on it.
1301 	 */
1302 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1303 	    M_NOWAIT|M_ZERO);
1304 	if (rte == NULL) {
1305 	    MFC_UNLOCK();
1306 	    VIF_UNLOCK();
1307 	    return ENOBUFS;
1308 	}
1309 
1310 	mb0 = m_copypacket(m, M_NOWAIT);
1311 	if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1312 	    mb0 = m_pullup(mb0, hlen);
1313 	if (mb0 == NULL) {
1314 	    free(rte, M_MRTABLE);
1315 	    MFC_UNLOCK();
1316 	    VIF_UNLOCK();
1317 	    return ENOBUFS;
1318 	}
1319 
1320 	/* is there an upcall waiting for this flow ? */
1321 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1322 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1323 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1324 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1325 		    !TAILQ_EMPTY(&rt->mfc_stall))
1326 			break;
1327 	}
1328 
1329 	if (rt == NULL) {
1330 	    int i;
1331 	    struct igmpmsg *im;
1332 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1333 	    struct mbuf *mm;
1334 
1335 	    /*
1336 	     * Locate the vifi for the incoming interface for this packet.
1337 	     * If none found, drop packet.
1338 	     */
1339 	    for (vifi = 0; vifi < V_numvifs &&
1340 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1341 		;
1342 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1343 		goto non_fatal;
1344 
1345 	    /* no upcall, so make a new entry */
1346 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1347 	    if (rt == NULL)
1348 		goto fail;
1349 
1350 	    /* Make a copy of the header to send to the user level process */
1351 	    mm = m_copy(mb0, 0, hlen);
1352 	    if (mm == NULL)
1353 		goto fail1;
1354 
1355 	    /*
1356 	     * Send message to routing daemon to install
1357 	     * a route into the kernel table
1358 	     */
1359 
1360 	    im = mtod(mm, struct igmpmsg *);
1361 	    im->im_msgtype = IGMPMSG_NOCACHE;
1362 	    im->im_mbz = 0;
1363 	    im->im_vif = vifi;
1364 
1365 	    MRTSTAT_INC(mrts_upcalls);
1366 
1367 	    k_igmpsrc.sin_addr = ip->ip_src;
1368 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1369 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1370 		MRTSTAT_INC(mrts_upq_sockfull);
1371 fail1:
1372 		free(rt, M_MRTABLE);
1373 fail:
1374 		free(rte, M_MRTABLE);
1375 		m_freem(mb0);
1376 		MFC_UNLOCK();
1377 		VIF_UNLOCK();
1378 		return ENOBUFS;
1379 	    }
1380 
1381 	    /* insert new entry at head of hash chain */
1382 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1383 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1384 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1385 	    V_nexpire[hash]++;
1386 	    for (i = 0; i < V_numvifs; i++) {
1387 		rt->mfc_ttls[i] = 0;
1388 		rt->mfc_flags[i] = 0;
1389 	    }
1390 	    rt->mfc_parent = -1;
1391 
1392 	    /* clear the RP address */
1393 	    rt->mfc_rp.s_addr = INADDR_ANY;
1394 	    rt->mfc_bw_meter = NULL;
1395 
1396 	    /* initialize pkt counters per src-grp */
1397 	    rt->mfc_pkt_cnt = 0;
1398 	    rt->mfc_byte_cnt = 0;
1399 	    rt->mfc_wrong_if = 0;
1400 	    timevalclear(&rt->mfc_last_assert);
1401 
1402 	    TAILQ_INIT(&rt->mfc_stall);
1403 	    rt->mfc_nstall = 0;
1404 
1405 	    /* link into table */
1406 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1407 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1408 	    rt->mfc_nstall++;
1409 
1410 	} else {
1411 	    /* determine if queue has overflowed */
1412 	    if (rt->mfc_nstall > MAX_UPQ) {
1413 		MRTSTAT_INC(mrts_upq_ovflw);
1414 non_fatal:
1415 		free(rte, M_MRTABLE);
1416 		m_freem(mb0);
1417 		MFC_UNLOCK();
1418 		VIF_UNLOCK();
1419 		return (0);
1420 	    }
1421 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1422 	    rt->mfc_nstall++;
1423 	}
1424 
1425 	rte->m			= mb0;
1426 	rte->ifp		= ifp;
1427 
1428 	MFC_UNLOCK();
1429 	VIF_UNLOCK();
1430 
1431 	return 0;
1432     }
1433 }
1434 
1435 /*
1436  * Clean up the cache entry if upcall is not serviced
1437  */
1438 static void
1439 expire_upcalls(void *arg)
1440 {
1441     int i;
1442 
1443     CURVNET_SET((struct vnet *) arg);
1444 
1445     MFC_LOCK();
1446 
1447     for (i = 0; i < mfchashsize; i++) {
1448 	struct mfc *rt, *nrt;
1449 
1450 	if (V_nexpire[i] == 0)
1451 	    continue;
1452 
1453 	for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
1454 		nrt = LIST_NEXT(rt, mfc_hash);
1455 
1456 		if (TAILQ_EMPTY(&rt->mfc_stall))
1457 			continue;
1458 
1459 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1460 			continue;
1461 
1462 		/*
1463 		 * free the bw_meter entries
1464 		 */
1465 		while (rt->mfc_bw_meter != NULL) {
1466 		    struct bw_meter *x = rt->mfc_bw_meter;
1467 
1468 		    rt->mfc_bw_meter = x->bm_mfc_next;
1469 		    free(x, M_BWMETER);
1470 		}
1471 
1472 		MRTSTAT_INC(mrts_cache_cleanups);
1473 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1474 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1475 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1476 
1477 		expire_mfc(rt);
1478 	    }
1479     }
1480 
1481     MFC_UNLOCK();
1482 
1483     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1484 	curvnet);
1485 
1486     CURVNET_RESTORE();
1487 }
1488 
1489 /*
1490  * Packet forwarding routine once entry in the cache is made
1491  */
1492 static int
1493 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1494 {
1495     struct ip  *ip = mtod(m, struct ip *);
1496     vifi_t vifi;
1497     int plen = ntohs(ip->ip_len);
1498 
1499     VIF_LOCK_ASSERT();
1500 
1501     /*
1502      * If xmt_vif is not -1, send on only the requested vif.
1503      *
1504      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1505      */
1506     if (xmt_vif < V_numvifs) {
1507 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1508 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1509 	else
1510 		phyint_send(ip, V_viftable + xmt_vif, m);
1511 	return 1;
1512     }
1513 
1514     /*
1515      * Don't forward if it didn't arrive from the parent vif for its origin.
1516      */
1517     vifi = rt->mfc_parent;
1518     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1519 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1520 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1521 	MRTSTAT_INC(mrts_wrong_if);
1522 	++rt->mfc_wrong_if;
1523 	/*
1524 	 * If we are doing PIM assert processing, send a message
1525 	 * to the routing daemon.
1526 	 *
1527 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1528 	 * can complete the SPT switch, regardless of the type
1529 	 * of the iif (broadcast media, GRE tunnel, etc).
1530 	 */
1531 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1532 	    V_viftable[vifi].v_ifp) {
1533 
1534 	    if (ifp == &V_multicast_register_if)
1535 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1536 
1537 	    /* Get vifi for the incoming packet */
1538 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1539 		vifi++)
1540 		;
1541 	    if (vifi >= V_numvifs)
1542 		return 0;	/* The iif is not found: ignore the packet. */
1543 
1544 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1545 		return 0;	/* WRONGVIF disabled: ignore the packet */
1546 
1547 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1548 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1549 		struct igmpmsg *im;
1550 		int hlen = ip->ip_hl << 2;
1551 		struct mbuf *mm = m_copy(m, 0, hlen);
1552 
1553 		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1554 		    mm = m_pullup(mm, hlen);
1555 		if (mm == NULL)
1556 		    return ENOBUFS;
1557 
1558 		im = mtod(mm, struct igmpmsg *);
1559 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1560 		im->im_mbz		= 0;
1561 		im->im_vif		= vifi;
1562 
1563 		MRTSTAT_INC(mrts_upcalls);
1564 
1565 		k_igmpsrc.sin_addr = im->im_src;
1566 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1567 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1568 		    MRTSTAT_INC(mrts_upq_sockfull);
1569 		    return ENOBUFS;
1570 		}
1571 	    }
1572 	}
1573 	return 0;
1574     }
1575 
1576 
1577     /* If I sourced this packet, it counts as output, else it was input. */
1578     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1579 	V_viftable[vifi].v_pkt_out++;
1580 	V_viftable[vifi].v_bytes_out += plen;
1581     } else {
1582 	V_viftable[vifi].v_pkt_in++;
1583 	V_viftable[vifi].v_bytes_in += plen;
1584     }
1585     rt->mfc_pkt_cnt++;
1586     rt->mfc_byte_cnt += plen;
1587 
1588     /*
1589      * For each vif, decide if a copy of the packet should be forwarded.
1590      * Forward if:
1591      *		- the ttl exceeds the vif's threshold
1592      *		- there are group members downstream on interface
1593      */
1594     for (vifi = 0; vifi < V_numvifs; vifi++)
1595 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1596 	    V_viftable[vifi].v_pkt_out++;
1597 	    V_viftable[vifi].v_bytes_out += plen;
1598 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1599 		pim_register_send(ip, V_viftable + vifi, m, rt);
1600 	    else
1601 		phyint_send(ip, V_viftable + vifi, m);
1602 	}
1603 
1604     /*
1605      * Perform upcall-related bw measuring.
1606      */
1607     if (rt->mfc_bw_meter != NULL) {
1608 	struct bw_meter *x;
1609 	struct timeval now;
1610 
1611 	microtime(&now);
1612 	MFC_LOCK_ASSERT();
1613 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1614 	    bw_meter_receive_packet(x, plen, &now);
1615     }
1616 
1617     return 0;
1618 }
1619 
1620 /*
1621  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1622  */
1623 static int
1624 X_legal_vif_num(int vif)
1625 {
1626 	int ret;
1627 
1628 	ret = 0;
1629 	if (vif < 0)
1630 		return (ret);
1631 
1632 	VIF_LOCK();
1633 	if (vif < V_numvifs)
1634 		ret = 1;
1635 	VIF_UNLOCK();
1636 
1637 	return (ret);
1638 }
1639 
1640 /*
1641  * Return the local address used by this vif
1642  */
1643 static u_long
1644 X_ip_mcast_src(int vifi)
1645 {
1646 	in_addr_t addr;
1647 
1648 	addr = INADDR_ANY;
1649 	if (vifi < 0)
1650 		return (addr);
1651 
1652 	VIF_LOCK();
1653 	if (vifi < V_numvifs)
1654 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1655 	VIF_UNLOCK();
1656 
1657 	return (addr);
1658 }
1659 
1660 static void
1661 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1662 {
1663     struct mbuf *mb_copy;
1664     int hlen = ip->ip_hl << 2;
1665 
1666     VIF_LOCK_ASSERT();
1667 
1668     /*
1669      * Make a new reference to the packet; make sure that
1670      * the IP header is actually copied, not just referenced,
1671      * so that ip_output() only scribbles on the copy.
1672      */
1673     mb_copy = m_copypacket(m, M_NOWAIT);
1674     if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1675 	mb_copy = m_pullup(mb_copy, hlen);
1676     if (mb_copy == NULL)
1677 	return;
1678 
1679     send_packet(vifp, mb_copy);
1680 }
1681 
1682 static void
1683 send_packet(struct vif *vifp, struct mbuf *m)
1684 {
1685 	struct ip_moptions imo;
1686 	struct in_multi *imm[2];
1687 	int error;
1688 
1689 	VIF_LOCK_ASSERT();
1690 
1691 	imo.imo_multicast_ifp  = vifp->v_ifp;
1692 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1693 	imo.imo_multicast_loop = 1;
1694 	imo.imo_multicast_vif  = -1;
1695 	imo.imo_num_memberships = 0;
1696 	imo.imo_max_memberships = 2;
1697 	imo.imo_membership  = &imm[0];
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 void
1728 X_rsvp_input(struct mbuf *m, int off __unused)
1729 {
1730 
1731 	if (!V_rsvp_on)
1732 		m_freem(m);
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_id = ip_newid();
2508     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2509 	sizeof(pim_encap_pimhdr));
2510     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2511     ip_outer->ip_dst = rt->mfc_rp;
2512     /*
2513      * Copy the inner header TOS to the outer header, and take care of the
2514      * IP_DF bit.
2515      */
2516     ip_outer->ip_tos = ip->ip_tos;
2517     if (ip->ip_off & htons(IP_DF))
2518 	ip_outer->ip_off |= htons(IP_DF);
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 void
2566 pim_input(struct mbuf *m, int off)
2567 {
2568     struct ip *ip = mtod(m, struct ip *);
2569     struct pim *pim;
2570     int minlen;
2571     int datalen = ntohs(ip->ip_len);
2572     int ip_tos;
2573     int iphlen = off;
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 %s",
2585 	    __func__, datalen, inet_ntoa(ip->ip_src));
2586 	m_freem(m);
2587 	return;
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_flags & M_EXT || m->m_len < minlen) &&
2604 	(m = m_pullup(m, minlen)) == 0) {
2605 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2606 	return;
2607     }
2608 
2609     /* m_pullup() may have given us a new mbuf so reset ip. */
2610     ip = mtod(m, struct ip *);
2611     ip_tos = ip->ip_tos;
2612 
2613     /* adjust mbuf to point to the PIM header */
2614     m->m_data += iphlen;
2615     m->m_len  -= iphlen;
2616     pim = mtod(m, struct pim *);
2617 
2618     /*
2619      * Validate checksum. If PIM REGISTER, exclude the data packet.
2620      *
2621      * XXX: some older PIMv2 implementations don't make this distinction,
2622      * so for compatibility reason perform the checksum over part of the
2623      * message, and if error, then over the whole message.
2624      */
2625     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2626 	/* do nothing, checksum okay */
2627     } else if (in_cksum(m, datalen)) {
2628 	PIMSTAT_INC(pims_rcv_badsum);
2629 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2630 	m_freem(m);
2631 	return;
2632     }
2633 
2634     /* PIM version check */
2635     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2636 	PIMSTAT_INC(pims_rcv_badversion);
2637 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2638 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2639 	m_freem(m);
2640 	return;
2641     }
2642 
2643     /* restore mbuf back to the outer IP */
2644     m->m_data -= iphlen;
2645     m->m_len  += iphlen;
2646 
2647     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2648 	/*
2649 	 * Since this is a REGISTER, we'll make a copy of the register
2650 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2651 	 * routing daemon.
2652 	 */
2653 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2654 	struct mbuf *mcp;
2655 	struct ip *encap_ip;
2656 	u_int32_t *reghdr;
2657 	struct ifnet *vifp;
2658 
2659 	VIF_LOCK();
2660 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2661 	    VIF_UNLOCK();
2662 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2663 		(int)V_reg_vif_num);
2664 	    m_freem(m);
2665 	    return;
2666 	}
2667 	/* XXX need refcnt? */
2668 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2669 	VIF_UNLOCK();
2670 
2671 	/*
2672 	 * Validate length
2673 	 */
2674 	if (datalen < PIM_REG_MINLEN) {
2675 	    PIMSTAT_INC(pims_rcv_tooshort);
2676 	    PIMSTAT_INC(pims_rcv_badregisters);
2677 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2678 	    m_freem(m);
2679 	    return;
2680 	}
2681 
2682 	reghdr = (u_int32_t *)(pim + 1);
2683 	encap_ip = (struct ip *)(reghdr + 1);
2684 
2685 	CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2686 	    __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2687 
2688 	/* verify the version number of the inner packet */
2689 	if (encap_ip->ip_v != IPVERSION) {
2690 	    PIMSTAT_INC(pims_rcv_badregisters);
2691 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2692 	    m_freem(m);
2693 	    return;
2694 	}
2695 
2696 	/* verify the inner packet is destined to a mcast group */
2697 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2698 	    PIMSTAT_INC(pims_rcv_badregisters);
2699 	    CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2700 		inet_ntoa(encap_ip->ip_dst));
2701 	    m_freem(m);
2702 	    return;
2703 	}
2704 
2705 	/* If a NULL_REGISTER, pass it to the daemon */
2706 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2707 	    goto pim_input_to_daemon;
2708 
2709 	/*
2710 	 * Copy the TOS from the outer IP header to the inner IP header.
2711 	 */
2712 	if (encap_ip->ip_tos != ip_tos) {
2713 	    /* Outer TOS -> inner TOS */
2714 	    encap_ip->ip_tos = ip_tos;
2715 	    /* Recompute the inner header checksum. Sigh... */
2716 
2717 	    /* adjust mbuf to point to the inner IP header */
2718 	    m->m_data += (iphlen + PIM_MINLEN);
2719 	    m->m_len  -= (iphlen + PIM_MINLEN);
2720 
2721 	    encap_ip->ip_sum = 0;
2722 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2723 
2724 	    /* restore mbuf to point back to the outer IP header */
2725 	    m->m_data -= (iphlen + PIM_MINLEN);
2726 	    m->m_len  += (iphlen + PIM_MINLEN);
2727 	}
2728 
2729 	/*
2730 	 * Decapsulate the inner IP packet and loopback to forward it
2731 	 * as a normal multicast packet. Also, make a copy of the
2732 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2733 	 * to pass to the daemon later, so it can take the appropriate
2734 	 * actions (e.g., send back PIM_REGISTER_STOP).
2735 	 * XXX: here m->m_data points to the outer IP header.
2736 	 */
2737 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2738 	if (mcp == NULL) {
2739 	    CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2740 	    m_freem(m);
2741 	    return;
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     rip_input(m, iphlen);
2778 
2779     return;
2780 }
2781 
2782 static int
2783 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2784 {
2785 	struct mfc	*rt;
2786 	int		 error, i;
2787 
2788 	if (req->newptr)
2789 		return (EPERM);
2790 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2791 		return (0);
2792 	error = sysctl_wire_old_buffer(req, 0);
2793 	if (error)
2794 		return (error);
2795 
2796 	MFC_LOCK();
2797 	for (i = 0; i < mfchashsize; i++) {
2798 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2799 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2800 			if (error)
2801 				goto out_locked;
2802 		}
2803 	}
2804 out_locked:
2805 	MFC_UNLOCK();
2806 	return (error);
2807 }
2808 
2809 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2810     sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2811     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2812 
2813 static void
2814 vnet_mroute_init(const void *unused __unused)
2815 {
2816 
2817 	MALLOC(V_nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2818 	bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2819 	callout_init(&V_expire_upcalls_ch, CALLOUT_MPSAFE);
2820 	callout_init(&V_bw_upcalls_ch, CALLOUT_MPSAFE);
2821 	callout_init(&V_bw_meter_ch, CALLOUT_MPSAFE);
2822 }
2823 
2824 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mroute_init,
2825 	NULL);
2826 
2827 static void
2828 vnet_mroute_uninit(const void *unused __unused)
2829 {
2830 
2831 	FREE(V_nexpire, M_MRTABLE);
2832 	V_nexpire = NULL;
2833 }
2834 
2835 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE,
2836 	vnet_mroute_uninit, NULL);
2837 
2838 static int
2839 ip_mroute_modevent(module_t mod, int type, void *unused)
2840 {
2841 
2842     switch (type) {
2843     case MOD_LOAD:
2844 	MROUTER_LOCK_INIT();
2845 
2846 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2847 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2848 	if (if_detach_event_tag == NULL) {
2849 		printf("ip_mroute: unable to register "
2850 		    "ifnet_departure_event handler\n");
2851 		MROUTER_LOCK_DESTROY();
2852 		return (EINVAL);
2853 	}
2854 
2855 	MFC_LOCK_INIT();
2856 	VIF_LOCK_INIT();
2857 
2858 	mfchashsize = MFCHASHSIZE;
2859 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2860 	    !powerof2(mfchashsize)) {
2861 		printf("WARNING: %s not a power of 2; using default\n",
2862 		    "net.inet.ip.mfchashsize");
2863 		mfchashsize = MFCHASHSIZE;
2864 	}
2865 
2866 	pim_squelch_wholepkt = 0;
2867 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2868 	    &pim_squelch_wholepkt);
2869 
2870 	pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2871 	    pim_encapcheck, &in_pim_protosw, NULL);
2872 	if (pim_encap_cookie == NULL) {
2873 		printf("ip_mroute: unable to attach pim encap\n");
2874 		VIF_LOCK_DESTROY();
2875 		MFC_LOCK_DESTROY();
2876 		MROUTER_LOCK_DESTROY();
2877 		return (EINVAL);
2878 	}
2879 
2880 	ip_mcast_src = X_ip_mcast_src;
2881 	ip_mforward = X_ip_mforward;
2882 	ip_mrouter_done = X_ip_mrouter_done;
2883 	ip_mrouter_get = X_ip_mrouter_get;
2884 	ip_mrouter_set = X_ip_mrouter_set;
2885 
2886 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2887 	ip_rsvp_vif = X_ip_rsvp_vif;
2888 
2889 	legal_vif_num = X_legal_vif_num;
2890 	mrt_ioctl = X_mrt_ioctl;
2891 	rsvp_input_p = X_rsvp_input;
2892 	break;
2893 
2894     case MOD_UNLOAD:
2895 	/*
2896 	 * Typically module unload happens after the user-level
2897 	 * process has shutdown the kernel services (the check
2898 	 * below insures someone can't just yank the module out
2899 	 * from under a running process).  But if the module is
2900 	 * just loaded and then unloaded w/o starting up a user
2901 	 * process we still need to cleanup.
2902 	 */
2903 	MROUTER_LOCK();
2904 	if (ip_mrouter_cnt != 0) {
2905 	    MROUTER_UNLOCK();
2906 	    return (EINVAL);
2907 	}
2908 	ip_mrouter_unloading = 1;
2909 	MROUTER_UNLOCK();
2910 
2911 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2912 
2913 	if (pim_encap_cookie) {
2914 	    encap_detach(pim_encap_cookie);
2915 	    pim_encap_cookie = NULL;
2916 	}
2917 
2918 	ip_mcast_src = NULL;
2919 	ip_mforward = NULL;
2920 	ip_mrouter_done = NULL;
2921 	ip_mrouter_get = NULL;
2922 	ip_mrouter_set = NULL;
2923 
2924 	ip_rsvp_force_done = NULL;
2925 	ip_rsvp_vif = NULL;
2926 
2927 	legal_vif_num = NULL;
2928 	mrt_ioctl = NULL;
2929 	rsvp_input_p = NULL;
2930 
2931 	VIF_LOCK_DESTROY();
2932 	MFC_LOCK_DESTROY();
2933 	MROUTER_LOCK_DESTROY();
2934 	break;
2935 
2936     default:
2937 	return EOPNOTSUPP;
2938     }
2939     return 0;
2940 }
2941 
2942 static moduledata_t ip_mroutemod = {
2943     "ip_mroute",
2944     ip_mroute_modevent,
2945     0
2946 };
2947 
2948 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_MIDDLE);
2949