xref: /freebsd/sys/netinet/ip_mroute.c (revision 81e0e7b9e36d6a25b3af6482811318e085537d2f)
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  * Modified by Wojciech Macek, Semihalf, May 2021
53  *
54  * MROUTING Revision: 3.5
55  * and PIM-SMv2 and PIM-DM support, advanced API support,
56  * bandwidth metering and signaling
57  */
58 
59 /*
60  * TODO: Prefix functions with ipmf_.
61  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
62  * domain attachment (if_afdata) so we can track consumers of that service.
63  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
64  * move it to socket options.
65  * TODO: Cleanup LSRR removal further.
66  * TODO: Push RSVP stubs into raw_ip.c.
67  * TODO: Use bitstring.h for vif set.
68  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
69  * TODO: Sync ip6_mroute.c with this file.
70  */
71 
72 #include <sys/cdefs.h>
73 __FBSDID("$FreeBSD$");
74 
75 #include "opt_inet.h"
76 #include "opt_mrouting.h"
77 
78 #define _PIM_VT 1
79 
80 #include <sys/types.h>
81 #include <sys/param.h>
82 #include <sys/kernel.h>
83 #include <sys/stddef.h>
84 #include <sys/condvar.h>
85 #include <sys/eventhandler.h>
86 #include <sys/lock.h>
87 #include <sys/kthread.h>
88 #include <sys/ktr.h>
89 #include <sys/malloc.h>
90 #include <sys/mbuf.h>
91 #include <sys/module.h>
92 #include <sys/priv.h>
93 #include <sys/protosw.h>
94 #include <sys/signalvar.h>
95 #include <sys/socket.h>
96 #include <sys/socketvar.h>
97 #include <sys/sockio.h>
98 #include <sys/sx.h>
99 #include <sys/sysctl.h>
100 #include <sys/syslog.h>
101 #include <sys/systm.h>
102 #include <sys/taskqueue.h>
103 #include <sys/time.h>
104 #include <sys/counter.h>
105 #include <machine/atomic.h>
106 
107 #include <net/if.h>
108 #include <net/if_var.h>
109 #include <net/if_types.h>
110 #include <net/netisr.h>
111 #include <net/route.h>
112 #include <net/vnet.h>
113 
114 #include <netinet/in.h>
115 #include <netinet/igmp.h>
116 #include <netinet/in_systm.h>
117 #include <netinet/in_var.h>
118 #include <netinet/ip.h>
119 #include <netinet/ip_encap.h>
120 #include <netinet/ip_mroute.h>
121 #include <netinet/ip_var.h>
122 #include <netinet/ip_options.h>
123 #include <netinet/pim.h>
124 #include <netinet/pim_var.h>
125 #include <netinet/udp.h>
126 
127 #include <machine/in_cksum.h>
128 
129 #ifndef KTR_IPMF
130 #define KTR_IPMF KTR_INET
131 #endif
132 
133 #define		VIFI_INVALID	((vifi_t) -1)
134 
135 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
136 
137 /*
138  * Locking.  We use two locks: one for the virtual interface table and
139  * one for the forwarding table.  These locks may be nested in which case
140  * the VIF lock must always be taken first.  Note that each lock is used
141  * to cover not only the specific data structure but also related data
142  * structures.
143  */
144 
145 static struct rwlock mrouter_mtx;
146 #define	MRW_RLOCK()		rw_rlock(&mrouter_mtx)
147 #define	MRW_WLOCK()		rw_wlock(&mrouter_mtx)
148 #define	MRW_RUNLOCK()	rw_runlock(&mrouter_mtx)
149 #define	MRW_WUNLOCK()	rw_wunlock(&mrouter_mtx)
150 #define	MRW_UNLOCK()	rw_unlock(&mrouter_mtx)
151 #define	MRW_LOCK_ASSERT()	rw_assert(&mrouter_mtx, RA_LOCKED)
152 #define	MRW_WLOCK_ASSERT()	rw_assert(&mrouter_mtx, RA_WLOCKED)
153 #define	MRW_LOCK_TRY_UPGRADE()	rw_try_upgrade(&mrouter_mtx)
154 #define	MRW_WOWNED()	rw_wowned(&mrouter_mtx)
155 #define	MRW_LOCK_INIT()						\
156 	rw_init(&mrouter_mtx, "IPv4 multicast forwarding")
157 #define	MRW_LOCK_DESTROY()	rw_destroy(&mrouter_mtx)
158 
159 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
160 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
161 
162 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
163 VNET_PCPUSTAT_SYSINIT(mrtstat);
164 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
165 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
166     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
167     "netinet/ip_mroute.h)");
168 
169 VNET_DEFINE_STATIC(u_long, mfchash);
170 #define	V_mfchash		VNET(mfchash)
171 #define	MFCHASH(a, g)							\
172 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
173 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
174 #define	MFCHASHSIZE	256
175 
176 static u_long mfchashsize;			/* Hash size */
177 VNET_DEFINE_STATIC(u_char *, nexpire);		/* 0..mfchashsize-1 */
178 #define	V_nexpire		VNET(nexpire)
179 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
180 #define	V_mfchashtbl		VNET(mfchashtbl)
181 VNET_DEFINE_STATIC(struct taskqueue *, task_queue);
182 #define	V_task_queue		VNET(task_queue)
183 VNET_DEFINE_STATIC(struct task, task);
184 #define	V_task		VNET(task)
185 
186 VNET_DEFINE_STATIC(vifi_t, numvifs);
187 #define	V_numvifs		VNET(numvifs)
188 VNET_DEFINE_STATIC(struct vif *, viftable);
189 #define	V_viftable		VNET(viftable)
190 
191 static eventhandler_tag if_detach_event_tag = NULL;
192 
193 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
194 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
195 
196 VNET_DEFINE_STATIC(struct mtx, buf_ring_mtx);
197 #define	V_buf_ring_mtx	VNET(buf_ring_mtx)
198 
199 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
200 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
201 
202 /*
203  * Bandwidth meter variables and constants
204  */
205 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
206 
207 /*
208  * Pending upcalls are stored in a ring which is flushed when
209  * full, or periodically
210  */
211 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
212 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
213 VNET_DEFINE_STATIC(struct buf_ring *, bw_upcalls_ring);
214 #define	V_bw_upcalls_ring    	VNET(bw_upcalls_ring)
215 VNET_DEFINE_STATIC(struct mtx, bw_upcalls_ring_mtx);
216 #define	V_bw_upcalls_ring_mtx    	VNET(bw_upcalls_ring_mtx)
217 
218 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
219 
220 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
221 VNET_PCPUSTAT_SYSINIT(pimstat);
222 VNET_PCPUSTAT_SYSUNINIT(pimstat);
223 
224 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
225     "PIM");
226 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
227     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
228 
229 static u_long	pim_squelch_wholepkt = 0;
230 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
231     &pim_squelch_wholepkt, 0,
232     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
233 
234 static const struct encaptab *pim_encap_cookie;
235 static int pim_encapcheck(const struct mbuf *, int, int, void *);
236 static int pim_input(struct mbuf *, int, int, void *);
237 
238 extern int in_mcast_loop;
239 
240 static const struct encap_config ipv4_encap_cfg = {
241 	.proto = IPPROTO_PIM,
242 	.min_length = sizeof(struct ip) + PIM_MINLEN,
243 	.exact_match = 8,
244 	.check = pim_encapcheck,
245 	.input = pim_input
246 };
247 
248 /*
249  * Note: the PIM Register encapsulation adds the following in front of a
250  * data packet:
251  *
252  * struct pim_encap_hdr {
253  *    struct ip ip;
254  *    struct pim_encap_pimhdr  pim;
255  * }
256  *
257  */
258 
259 struct pim_encap_pimhdr {
260 	struct pim pim;
261 	uint32_t   flags;
262 };
263 #define		PIM_ENCAP_TTL	64
264 
265 static struct ip pim_encap_iphdr = {
266 #if BYTE_ORDER == LITTLE_ENDIAN
267 	sizeof(struct ip) >> 2,
268 	IPVERSION,
269 #else
270 	IPVERSION,
271 	sizeof(struct ip) >> 2,
272 #endif
273 	0,			/* tos */
274 	sizeof(struct ip),	/* total length */
275 	0,			/* id */
276 	0,			/* frag offset */
277 	PIM_ENCAP_TTL,
278 	IPPROTO_PIM,
279 	0,			/* checksum */
280 };
281 
282 static struct pim_encap_pimhdr pim_encap_pimhdr = {
283     {
284 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
285 	0,			/* reserved */
286 	0,			/* checksum */
287     },
288     0				/* flags */
289 };
290 
291 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
292 #define	V_reg_vif_num		VNET(reg_vif_num)
293 VNET_DEFINE_STATIC(struct ifnet *, multicast_register_if);
294 #define	V_multicast_register_if	VNET(multicast_register_if)
295 
296 /*
297  * Private variables.
298  */
299 
300 static u_long	X_ip_mcast_src(int);
301 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
302 		    struct ip_moptions *);
303 static int	X_ip_mrouter_done(void);
304 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
305 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
306 static int	X_legal_vif_num(int);
307 static int	X_mrt_ioctl(u_long, caddr_t, int);
308 
309 static int	add_bw_upcall(struct bw_upcall *);
310 static int	add_mfc(struct mfcctl2 *);
311 static int	add_vif(struct vifctl *);
312 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
313 static void	bw_meter_geq_receive_packet(struct bw_meter *, int,
314 		    struct timeval *);
315 static void	bw_upcalls_send(void);
316 static int	del_bw_upcall(struct bw_upcall *);
317 static int	del_mfc(struct mfcctl2 *);
318 static int	del_vif(vifi_t);
319 static int	del_vif_locked(vifi_t, struct ifnet **);
320 static void	expire_bw_upcalls_send(void *);
321 static void	expire_mfc(struct mfc *);
322 static void	expire_upcalls(void *);
323 static void	free_bw_list(struct bw_meter *);
324 static int	get_sg_cnt(struct sioc_sg_req *);
325 static int	get_vif_cnt(struct sioc_vif_req *);
326 static void	if_detached_event(void *, struct ifnet *);
327 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
328 static int	ip_mrouter_init(struct socket *, int);
329 static __inline struct mfc *
330 		mfc_find(struct in_addr *, struct in_addr *);
331 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
332 static struct mbuf *
333 		pim_register_prepare(struct ip *, struct mbuf *);
334 static int	pim_register_send(struct ip *, struct vif *,
335 		    struct mbuf *, struct mfc *);
336 static int	pim_register_send_rp(struct ip *, struct vif *,
337 		    struct mbuf *, struct mfc *);
338 static int	pim_register_send_upcall(struct ip *, struct vif *,
339 		    struct mbuf *, struct mfc *);
340 static void	send_packet(struct vif *, struct mbuf *);
341 static int	set_api_config(uint32_t *);
342 static int	set_assert(int);
343 static int	socket_send(struct socket *, struct mbuf *,
344 		    struct sockaddr_in *);
345 
346 /*
347  * Kernel multicast forwarding API capabilities and setup.
348  * If more API capabilities are added to the kernel, they should be
349  * recorded in `mrt_api_support'.
350  */
351 #define MRT_API_VERSION		0x0305
352 
353 static const int mrt_api_version = MRT_API_VERSION;
354 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
355 					 MRT_MFC_FLAGS_BORDER_VIF |
356 					 MRT_MFC_RP |
357 					 MRT_MFC_BW_UPCALL);
358 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
359 #define	V_mrt_api_config	VNET(mrt_api_config)
360 VNET_DEFINE_STATIC(int, pim_assert_enabled);
361 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
362 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
363 
364 /*
365  * Find a route for a given origin IP address and multicast group address.
366  * Statistics must be updated by the caller.
367  */
368 static __inline struct mfc *
369 mfc_find(struct in_addr *o, struct in_addr *g)
370 {
371 	struct mfc *rt;
372 
373 	/*
374 	 * Might be called both RLOCK and WLOCK.
375 	 * Check if any, it's caller responsibility
376 	 * to choose correct option.
377 	 */
378 	MRW_LOCK_ASSERT();
379 
380 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
381 		if (in_hosteq(rt->mfc_origin, *o) &&
382 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
383 		    buf_ring_empty(rt->mfc_stall_ring))
384 			break;
385 	}
386 
387 	return (rt);
388 }
389 
390 static __inline struct mfc *
391 mfc_alloc(void)
392 {
393 	struct mfc *rt;
394 	rt = (struct mfc*) malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT | M_ZERO);
395 	if (rt == NULL)
396 		return rt;
397 
398 	rt->mfc_stall_ring = buf_ring_alloc(MAX_UPQ, M_MRTABLE,
399 	    M_NOWAIT, &V_buf_ring_mtx);
400 	if (rt->mfc_stall_ring == NULL) {
401 		free(rt, M_MRTABLE);
402 		return NULL;
403 	}
404 
405 	return rt;
406 }
407 
408 /*
409  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
410  */
411 static int
412 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
413 {
414     int	error, optval;
415     vifi_t	vifi;
416     struct	vifctl vifc;
417     struct	mfcctl2 mfc;
418     struct	bw_upcall bw_upcall;
419     uint32_t	i;
420 
421     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
422 	return EPERM;
423 
424     error = 0;
425     switch (sopt->sopt_name) {
426     case MRT_INIT:
427 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
428 	if (error)
429 	    break;
430 	error = ip_mrouter_init(so, optval);
431 	break;
432 
433     case MRT_DONE:
434 	error = ip_mrouter_done();
435 	break;
436 
437     case MRT_ADD_VIF:
438 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
439 	if (error)
440 	    break;
441 	error = add_vif(&vifc);
442 	break;
443 
444     case MRT_DEL_VIF:
445 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
446 	if (error)
447 	    break;
448 	error = del_vif(vifi);
449 	break;
450 
451     case MRT_ADD_MFC:
452     case MRT_DEL_MFC:
453 	/*
454 	 * select data size depending on API version.
455 	 */
456 	if (sopt->sopt_name == MRT_ADD_MFC &&
457 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
458 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
459 				sizeof(struct mfcctl2));
460 	} else {
461 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
462 				sizeof(struct mfcctl));
463 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
464 			sizeof(mfc) - sizeof(struct mfcctl));
465 	}
466 	if (error)
467 	    break;
468 	if (sopt->sopt_name == MRT_ADD_MFC)
469 	    error = add_mfc(&mfc);
470 	else
471 	    error = del_mfc(&mfc);
472 	break;
473 
474     case MRT_ASSERT:
475 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
476 	if (error)
477 	    break;
478 	set_assert(optval);
479 	break;
480 
481     case MRT_API_CONFIG:
482 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
483 	if (!error)
484 	    error = set_api_config(&i);
485 	if (!error)
486 	    error = sooptcopyout(sopt, &i, sizeof i);
487 	break;
488 
489     case MRT_ADD_BW_UPCALL:
490     case MRT_DEL_BW_UPCALL:
491 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
492 				sizeof bw_upcall);
493 	if (error)
494 	    break;
495 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
496 	    error = add_bw_upcall(&bw_upcall);
497 	else
498 	    error = del_bw_upcall(&bw_upcall);
499 	break;
500 
501     default:
502 	error = EOPNOTSUPP;
503 	break;
504     }
505     return error;
506 }
507 
508 /*
509  * Handle MRT getsockopt commands
510  */
511 static int
512 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
513 {
514     int error;
515 
516     switch (sopt->sopt_name) {
517     case MRT_VERSION:
518 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
519 	break;
520 
521     case MRT_ASSERT:
522 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
523 	    sizeof V_pim_assert_enabled);
524 	break;
525 
526     case MRT_API_SUPPORT:
527 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
528 	break;
529 
530     case MRT_API_CONFIG:
531 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
532 	break;
533 
534     default:
535 	error = EOPNOTSUPP;
536 	break;
537     }
538     return error;
539 }
540 
541 /*
542  * Handle ioctl commands to obtain information from the cache
543  */
544 static int
545 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
546 {
547     int error = 0;
548 
549     /*
550      * Currently the only function calling this ioctl routine is rtioctl_fib().
551      * Typically, only root can create the raw socket in order to execute
552      * this ioctl method, however the request might be coming from a prison
553      */
554     error = priv_check(curthread, PRIV_NETINET_MROUTE);
555     if (error)
556 	return (error);
557     switch (cmd) {
558     case (SIOCGETVIFCNT):
559 	error = get_vif_cnt((struct sioc_vif_req *)data);
560 	break;
561 
562     case (SIOCGETSGCNT):
563 	error = get_sg_cnt((struct sioc_sg_req *)data);
564 	break;
565 
566     default:
567 	error = EINVAL;
568 	break;
569     }
570     return error;
571 }
572 
573 /*
574  * returns the packet, byte, rpf-failure count for the source group provided
575  */
576 static int
577 get_sg_cnt(struct sioc_sg_req *req)
578 {
579     struct mfc *rt;
580 
581     MRW_RLOCK();
582     rt = mfc_find(&req->src, &req->grp);
583     if (rt == NULL) {
584 	    MRW_RUNLOCK();
585 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
586 	return EADDRNOTAVAIL;
587     }
588     req->pktcnt = rt->mfc_pkt_cnt;
589     req->bytecnt = rt->mfc_byte_cnt;
590     req->wrong_if = rt->mfc_wrong_if;
591     MRW_RUNLOCK();
592     return 0;
593 }
594 
595 /*
596  * returns the input and output packet and byte counts on the vif provided
597  */
598 static int
599 get_vif_cnt(struct sioc_vif_req *req)
600 {
601     vifi_t vifi = req->vifi;
602 
603     MRW_RLOCK();
604     if (vifi >= V_numvifs) {
605 	MRW_RUNLOCK();
606 	return EINVAL;
607     }
608 
609     mtx_lock_spin(&V_viftable[vifi].v_spin);
610     req->icount = V_viftable[vifi].v_pkt_in;
611     req->ocount = V_viftable[vifi].v_pkt_out;
612     req->ibytes = V_viftable[vifi].v_bytes_in;
613     req->obytes = V_viftable[vifi].v_bytes_out;
614     mtx_unlock_spin(&V_viftable[vifi].v_spin);
615     MRW_RUNLOCK();
616 
617     return 0;
618 }
619 
620 static void
621 if_detached_event(void *arg __unused, struct ifnet *ifp)
622 {
623     vifi_t vifi;
624     u_long i, vifi_cnt = 0;
625     struct ifnet *free_ptr;
626 
627     MRW_WLOCK();
628 
629     if (V_ip_mrouter == NULL) {
630 	MRW_WUNLOCK();
631 	return;
632     }
633 
634     /*
635      * Tear down multicast forwarder state associated with this ifnet.
636      * 1. Walk the vif list, matching vifs against this ifnet.
637      * 2. Walk the multicast forwarding cache (mfc) looking for
638      *    inner matches with this vif's index.
639      * 3. Expire any matching multicast forwarding cache entries.
640      * 4. Free vif state. This should disable ALLMULTI on the interface.
641      */
642     for (vifi = 0; vifi < V_numvifs; vifi++) {
643 	if (V_viftable[vifi].v_ifp != ifp)
644 		continue;
645 	for (i = 0; i < mfchashsize; i++) {
646 		struct mfc *rt, *nrt;
647 
648 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
649 			if (rt->mfc_parent == vifi) {
650 				expire_mfc(rt);
651 			}
652 		}
653 	}
654 	del_vif_locked(vifi, &free_ptr);
655 	if (free_ptr != NULL)
656 		vifi_cnt++;
657     }
658 
659     MRW_WUNLOCK();
660 
661     /*
662      * Free IFP. We don't have to use free_ptr here as it is the same
663      * that ifp. Perform free as many times as required in case
664      * refcount is greater than 1.
665      */
666     for (i = 0; i < vifi_cnt; i++)
667 	    if_free(ifp);
668 }
669 
670 static void
671 ip_mrouter_upcall_thread(void *arg, int pending __unused)
672 {
673 	CURVNET_SET((struct vnet *) arg);
674 
675 	MRW_WLOCK();
676 	bw_upcalls_send();
677 	MRW_WUNLOCK();
678 
679 	CURVNET_RESTORE();
680 }
681 
682 /*
683  * Enable multicast forwarding.
684  */
685 static int
686 ip_mrouter_init(struct socket *so, int version)
687 {
688 
689     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
690         so->so_type, so->so_proto->pr_protocol);
691 
692     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
693 	return EOPNOTSUPP;
694 
695     if (version != 1)
696 	return ENOPROTOOPT;
697 
698     MRW_WLOCK();
699 
700     if (ip_mrouter_unloading) {
701 	MRW_WUNLOCK();
702 	return ENOPROTOOPT;
703     }
704 
705     if (V_ip_mrouter != NULL) {
706 	MRW_WUNLOCK();
707 	return EADDRINUSE;
708     }
709 
710     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
711 	HASH_NOWAIT);
712 
713     /* Create upcall ring */
714     mtx_init(&V_bw_upcalls_ring_mtx, "mroute upcall buf_ring mtx", NULL, MTX_DEF);
715     V_bw_upcalls_ring = buf_ring_alloc(BW_UPCALLS_MAX, M_MRTABLE,
716 	M_NOWAIT, &V_bw_upcalls_ring_mtx);
717     if (!V_bw_upcalls_ring) {
718 	MRW_WUNLOCK();
719 	return (ENOMEM);
720     }
721 
722     TASK_INIT(&V_task, 0, ip_mrouter_upcall_thread, curvnet);
723     taskqueue_cancel(V_task_queue, &V_task, NULL);
724     taskqueue_unblock(V_task_queue);
725 
726     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
727 	curvnet);
728     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
729 	curvnet);
730 
731     V_ip_mrouter = so;
732     atomic_add_int(&ip_mrouter_cnt, 1);
733 
734     /* This is a mutex required by buf_ring init, but not used internally */
735     mtx_init(&V_buf_ring_mtx, "mroute buf_ring mtx", NULL, MTX_DEF);
736 
737     MRW_WUNLOCK();
738 
739     CTR1(KTR_IPMF, "%s: done", __func__);
740 
741     return 0;
742 }
743 
744 /*
745  * Disable multicast forwarding.
746  */
747 static int
748 X_ip_mrouter_done(void)
749 {
750     struct ifnet *ifp;
751     u_long i;
752     vifi_t vifi;
753     struct bw_upcall *bu;
754 
755     if (V_ip_mrouter == NULL)
756 	return (EINVAL);
757 
758     /*
759      * Detach/disable hooks to the reset of the system.
760      */
761     V_ip_mrouter = NULL;
762     atomic_subtract_int(&ip_mrouter_cnt, 1);
763     V_mrt_api_config = 0;
764 
765     /*
766      * Wait for all epoch sections to complete to ensure
767      * V_ip_mrouter = NULL is visible to others.
768      */
769     epoch_wait_preempt(net_epoch_preempt);
770 
771     /* Stop and drain task queue */
772     taskqueue_block(V_task_queue);
773     while (taskqueue_cancel(V_task_queue, &V_task, NULL)) {
774     	taskqueue_drain(V_task_queue, &V_task);
775     }
776 
777     MRW_WLOCK();
778     taskqueue_cancel(V_task_queue, &V_task, NULL);
779 
780     /* Destroy upcall ring */
781     while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
782 	free(bu, M_MRTABLE);
783     }
784     buf_ring_free(V_bw_upcalls_ring, M_MRTABLE);
785     mtx_destroy(&V_bw_upcalls_ring_mtx);
786 
787     /*
788      * For each phyint in use, disable promiscuous reception of all IP
789      * multicasts.
790      */
791     for (vifi = 0; vifi < V_numvifs; vifi++) {
792 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
793 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
794 	    ifp = V_viftable[vifi].v_ifp;
795 	    if_allmulti(ifp, 0);
796 	}
797     }
798     bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
799     V_numvifs = 0;
800     V_pim_assert_enabled = 0;
801 
802     callout_stop(&V_expire_upcalls_ch);
803     callout_stop(&V_bw_upcalls_ch);
804 
805     /*
806      * Free all multicast forwarding cache entries.
807      * Do not use hashdestroy(), as we must perform other cleanup.
808      */
809     for (i = 0; i < mfchashsize; i++) {
810 	struct mfc *rt, *nrt;
811 
812 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
813 		expire_mfc(rt);
814 	}
815     }
816     free(V_mfchashtbl, M_MRTABLE);
817     V_mfchashtbl = NULL;
818 
819     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
820 
821     V_reg_vif_num = VIFI_INVALID;
822 
823     mtx_destroy(&V_buf_ring_mtx);
824 
825     MRW_WUNLOCK();
826 
827     CTR1(KTR_IPMF, "%s: done", __func__);
828 
829     return 0;
830 }
831 
832 /*
833  * Set PIM assert processing global
834  */
835 static int
836 set_assert(int i)
837 {
838     if ((i != 1) && (i != 0))
839 	return EINVAL;
840 
841     V_pim_assert_enabled = i;
842 
843     return 0;
844 }
845 
846 /*
847  * Configure API capabilities
848  */
849 int
850 set_api_config(uint32_t *apival)
851 {
852     u_long i;
853 
854     /*
855      * We can set the API capabilities only if it is the first operation
856      * after MRT_INIT. I.e.:
857      *  - there are no vifs installed
858      *  - pim_assert is not enabled
859      *  - the MFC table is empty
860      */
861     if (V_numvifs > 0) {
862 	*apival = 0;
863 	return EPERM;
864     }
865     if (V_pim_assert_enabled) {
866 	*apival = 0;
867 	return EPERM;
868     }
869 
870     MRW_RLOCK();
871 
872     for (i = 0; i < mfchashsize; i++) {
873 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
874 	    MRW_RUNLOCK();
875 	    *apival = 0;
876 	    return EPERM;
877 	}
878     }
879 
880     MRW_RUNLOCK();
881 
882     V_mrt_api_config = *apival & mrt_api_support;
883     *apival = V_mrt_api_config;
884 
885     return 0;
886 }
887 
888 /*
889  * Add a vif to the vif table
890  */
891 static int
892 add_vif(struct vifctl *vifcp)
893 {
894     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
895     struct sockaddr_in sin = {sizeof sin, AF_INET};
896     struct ifaddr *ifa;
897     struct ifnet *ifp;
898     int error;
899 
900 
901     if (vifcp->vifc_vifi >= MAXVIFS)
902 	return EINVAL;
903     /* rate limiting is no longer supported by this code */
904     if (vifcp->vifc_rate_limit != 0) {
905 	log(LOG_ERR, "rate limiting is no longer supported\n");
906 	return EINVAL;
907     }
908 
909     if (in_nullhost(vifcp->vifc_lcl_addr))
910 	return EADDRNOTAVAIL;
911 
912     /* Find the interface with an address in AF_INET family */
913     if (vifcp->vifc_flags & VIFF_REGISTER) {
914 	/*
915 	 * XXX: Because VIFF_REGISTER does not really need a valid
916 	 * local interface (e.g. it could be 127.0.0.2), we don't
917 	 * check its address.
918 	 */
919 	ifp = NULL;
920     } else {
921 	struct epoch_tracker et;
922 
923 	sin.sin_addr = vifcp->vifc_lcl_addr;
924 	NET_EPOCH_ENTER(et);
925 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
926 	if (ifa == NULL) {
927 	    NET_EPOCH_EXIT(et);
928 	    return EADDRNOTAVAIL;
929 	}
930 	ifp = ifa->ifa_ifp;
931 	/* XXX FIXME we need to take a ref on ifp and cleanup properly! */
932 	NET_EPOCH_EXIT(et);
933     }
934 
935     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
936 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
937 	return EOPNOTSUPP;
938     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
939 	ifp = V_multicast_register_if = if_alloc(IFT_LOOP);
940 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
941 	if (V_reg_vif_num == VIFI_INVALID) {
942 	    if_initname(V_multicast_register_if, "register_vif", 0);
943 	    V_reg_vif_num = vifcp->vifc_vifi;
944 	}
945     } else {		/* Make sure the interface supports multicast */
946 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
947 	    return EOPNOTSUPP;
948 
949 	/* Enable promiscuous reception of all IP multicasts from the if */
950 	error = if_allmulti(ifp, 1);
951 	if (error)
952 	    return error;
953     }
954 
955     MRW_WLOCK();
956 
957     if (!in_nullhost(vifp->v_lcl_addr)) {
958 	if (ifp)
959 		V_multicast_register_if = NULL;
960 	MRW_WUNLOCK();
961 	if (ifp)
962 		if_free(ifp);
963 	return EADDRINUSE;
964     }
965 
966     vifp->v_flags     = vifcp->vifc_flags;
967     vifp->v_threshold = vifcp->vifc_threshold;
968     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
969     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
970     vifp->v_ifp       = ifp;
971     /* initialize per vif pkt counters */
972     vifp->v_pkt_in    = 0;
973     vifp->v_pkt_out   = 0;
974     vifp->v_bytes_in  = 0;
975     vifp->v_bytes_out = 0;
976     sprintf(vifp->v_spin_name, "BM[%d] spin", vifcp->vifc_vifi);
977     mtx_init(&vifp->v_spin, vifp->v_spin_name, NULL, MTX_SPIN);
978 
979     /* Adjust numvifs up if the vifi is higher than numvifs */
980     if (V_numvifs <= vifcp->vifc_vifi)
981 	V_numvifs = vifcp->vifc_vifi + 1;
982 
983     MRW_WUNLOCK();
984 
985     CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
986 	(int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
987 	(int)vifcp->vifc_threshold);
988 
989     return 0;
990 }
991 
992 /*
993  * Delete a vif from the vif table
994  */
995 static int
996 del_vif_locked(vifi_t vifi, struct ifnet **ifp_free)
997 {
998     struct vif *vifp;
999 
1000     *ifp_free = NULL;
1001 
1002     MRW_WLOCK_ASSERT();
1003 
1004     if (vifi >= V_numvifs) {
1005 	return EINVAL;
1006     }
1007     vifp = &V_viftable[vifi];
1008     if (in_nullhost(vifp->v_lcl_addr)) {
1009 	return EADDRNOTAVAIL;
1010     }
1011 
1012     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1013 	if_allmulti(vifp->v_ifp, 0);
1014 
1015     if (vifp->v_flags & VIFF_REGISTER) {
1016 	V_reg_vif_num = VIFI_INVALID;
1017 	if (vifp->v_ifp) {
1018 	    if (vifp->v_ifp == V_multicast_register_if)
1019 	        V_multicast_register_if = NULL;
1020 	    *ifp_free = vifp->v_ifp;
1021 	}
1022     }
1023 
1024     mtx_destroy(&vifp->v_spin);
1025 
1026     bzero((caddr_t)vifp, sizeof (*vifp));
1027 
1028     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
1029 
1030     /* Adjust numvifs down */
1031     for (vifi = V_numvifs; vifi > 0; vifi--)
1032 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
1033 	    break;
1034     V_numvifs = vifi;
1035 
1036     return 0;
1037 }
1038 
1039 static int
1040 del_vif(vifi_t vifi)
1041 {
1042     int cc;
1043     struct ifnet *free_ptr;
1044 
1045     MRW_WLOCK();
1046     cc = del_vif_locked(vifi, &free_ptr);
1047     MRW_WUNLOCK();
1048 
1049     if (free_ptr)
1050 	    if_free(free_ptr);
1051 
1052     return cc;
1053 }
1054 
1055 /*
1056  * update an mfc entry without resetting counters and S,G addresses.
1057  */
1058 static void
1059 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1060 {
1061     int i;
1062 
1063     rt->mfc_parent = mfccp->mfcc_parent;
1064     for (i = 0; i < V_numvifs; i++) {
1065 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1066 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1067 	    MRT_MFC_FLAGS_ALL;
1068     }
1069     /* set the RP address */
1070     if (V_mrt_api_config & MRT_MFC_RP)
1071 	rt->mfc_rp = mfccp->mfcc_rp;
1072     else
1073 	rt->mfc_rp.s_addr = INADDR_ANY;
1074 }
1075 
1076 /*
1077  * fully initialize an mfc entry from the parameter.
1078  */
1079 static void
1080 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1081 {
1082     rt->mfc_origin     = mfccp->mfcc_origin;
1083     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1084 
1085     update_mfc_params(rt, mfccp);
1086 
1087     /* initialize pkt counters per src-grp */
1088     rt->mfc_pkt_cnt    = 0;
1089     rt->mfc_byte_cnt   = 0;
1090     rt->mfc_wrong_if   = 0;
1091     timevalclear(&rt->mfc_last_assert);
1092 }
1093 
1094 static void
1095 expire_mfc(struct mfc *rt)
1096 {
1097 	struct rtdetq *rte;
1098 
1099 	MRW_WLOCK_ASSERT();
1100 
1101 	free_bw_list(rt->mfc_bw_meter_leq);
1102 	free_bw_list(rt->mfc_bw_meter_geq);
1103 
1104 	while (!buf_ring_empty(rt->mfc_stall_ring)) {
1105 		rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1106 		if (rte) {
1107 			m_freem(rte->m);
1108 			free(rte, M_MRTABLE);
1109 		}
1110 	}
1111 	buf_ring_free(rt->mfc_stall_ring, M_MRTABLE);
1112 
1113 	LIST_REMOVE(rt, mfc_hash);
1114 	free(rt, M_MRTABLE);
1115 }
1116 
1117 /*
1118  * Add an mfc entry
1119  */
1120 static int
1121 add_mfc(struct mfcctl2 *mfccp)
1122 {
1123     struct mfc *rt;
1124     struct rtdetq *rte;
1125     u_long hash = 0;
1126     u_short nstl;
1127 
1128     MRW_WLOCK();
1129     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1130 
1131     /* If an entry already exists, just update the fields */
1132     if (rt) {
1133 	CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1134 	    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1135 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1136 	    mfccp->mfcc_parent);
1137 	update_mfc_params(rt, mfccp);
1138 	MRW_WUNLOCK();
1139 	return (0);
1140     }
1141 
1142     /*
1143      * Find the entry for which the upcall was made and update
1144      */
1145     nstl = 0;
1146     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1147     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1148 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1149 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1150 	    !buf_ring_empty(rt->mfc_stall_ring)) {
1151 		CTR5(KTR_IPMF,
1152 		    "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1153 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1154 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1155 		    mfccp->mfcc_parent,
1156 		    rt->mfc_stall_ring);
1157 		if (nstl++)
1158 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1159 
1160 		init_mfc_params(rt, mfccp);
1161 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1162 		V_nexpire[hash]--;
1163 
1164 		/* Free queued packets, but attempt to forward them first. */
1165 		while (!buf_ring_empty(rt->mfc_stall_ring)) {
1166 			rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1167 			if (rte->ifp != NULL)
1168 				ip_mdq(rte->m, rte->ifp, rt, -1);
1169 			m_freem(rte->m);
1170 			free(rte, M_MRTABLE);
1171 		}
1172 	}
1173     }
1174 
1175     /*
1176      * It is possible that an entry is being inserted without an upcall
1177      */
1178     if (nstl == 0) {
1179 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1180 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1181 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1182 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1183 			init_mfc_params(rt, mfccp);
1184 			if (rt->mfc_expire)
1185 			    V_nexpire[hash]--;
1186 			rt->mfc_expire = 0;
1187 			break; /* XXX */
1188 		}
1189 	}
1190 
1191 	if (rt == NULL) {		/* no upcall, so make a new entry */
1192 	    rt = mfc_alloc();
1193 	    if (rt == NULL) {
1194 		MRW_WUNLOCK();
1195 		return (ENOBUFS);
1196 	    }
1197 
1198 	    init_mfc_params(rt, mfccp);
1199 
1200 	    rt->mfc_expire     = 0;
1201 	    rt->mfc_bw_meter_leq = NULL;
1202 	    rt->mfc_bw_meter_geq = NULL;
1203 
1204 	    /* insert new entry at head of hash chain */
1205 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1206 	}
1207     }
1208 
1209     MRW_WUNLOCK();
1210 
1211     return (0);
1212 }
1213 
1214 /*
1215  * Delete an mfc entry
1216  */
1217 static int
1218 del_mfc(struct mfcctl2 *mfccp)
1219 {
1220     struct in_addr	origin;
1221     struct in_addr	mcastgrp;
1222     struct mfc		*rt;
1223 
1224     origin = mfccp->mfcc_origin;
1225     mcastgrp = mfccp->mfcc_mcastgrp;
1226 
1227     CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1228 	ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1229 
1230     MRW_WLOCK();
1231 
1232     rt = mfc_find(&origin, &mcastgrp);
1233     if (rt == NULL) {
1234 	MRW_WUNLOCK();
1235 	return EADDRNOTAVAIL;
1236     }
1237 
1238     /*
1239      * free the bw_meter entries
1240      */
1241     free_bw_list(rt->mfc_bw_meter_leq);
1242     rt->mfc_bw_meter_leq = NULL;
1243     free_bw_list(rt->mfc_bw_meter_geq);
1244     rt->mfc_bw_meter_geq = NULL;
1245 
1246     LIST_REMOVE(rt, mfc_hash);
1247     free(rt, M_MRTABLE);
1248 
1249     MRW_WUNLOCK();
1250 
1251     return (0);
1252 }
1253 
1254 /*
1255  * Send a message to the routing daemon on the multicast routing socket.
1256  */
1257 static int
1258 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1259 {
1260     if (s) {
1261 	SOCKBUF_LOCK(&s->so_rcv);
1262 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1263 	    NULL) != 0) {
1264 	    sorwakeup_locked(s);
1265 	    return 0;
1266 	}
1267 	soroverflow_locked(s);
1268     }
1269     m_freem(mm);
1270     return -1;
1271 }
1272 
1273 /*
1274  * IP multicast forwarding function. This function assumes that the packet
1275  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1276  * pointed to by "ifp", and the packet is to be relayed to other networks
1277  * that have members of the packet's destination IP multicast group.
1278  *
1279  * The packet is returned unscathed to the caller, unless it is
1280  * erroneous, in which case a non-zero return value tells the caller to
1281  * discard it.
1282  */
1283 
1284 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1285 
1286 static int
1287 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1288     struct ip_moptions *imo)
1289 {
1290 	struct mfc *rt;
1291 	int error;
1292 	vifi_t vifi;
1293 	struct mbuf *mb0;
1294 	struct rtdetq *rte;
1295 	u_long hash;
1296 	int hlen;
1297 
1298 	CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1299 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1300 
1301 	if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1302 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1303 		/*
1304 		 * Packet arrived via a physical interface or
1305 		 * an encapsulated tunnel or a register_vif.
1306 		 */
1307 	} else {
1308 		/*
1309 		 * Packet arrived through a source-route tunnel.
1310 		 * Source-route tunnels are no longer supported.
1311 		 */
1312 		return (1);
1313 	}
1314 
1315 	/*
1316 	 * BEGIN: MCAST ROUTING HOT PATH
1317 	 */
1318 	MRW_RLOCK();
1319 	if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1320 		if (ip->ip_ttl < MAXTTL)
1321 			ip->ip_ttl++; /* compensate for -1 in *_send routines */
1322 		error = ip_mdq(m, ifp, NULL, vifi);
1323 		MRW_RUNLOCK();
1324 		return error;
1325 	}
1326 
1327 	/*
1328 	 * Don't forward a packet with time-to-live of zero or one,
1329 	 * or a packet destined to a local-only group.
1330 	 */
1331 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1332 		MRW_RUNLOCK();
1333 		return 0;
1334 	}
1335 
1336 	mfc_find_retry:
1337 	/*
1338 	 * Determine forwarding vifs from the forwarding cache table
1339 	 */
1340 	MRTSTAT_INC(mrts_mfc_lookups);
1341 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1342 
1343 	/* Entry exists, so forward if necessary */
1344 	if (rt != NULL) {
1345 		error = ip_mdq(m, ifp, rt, -1);
1346 		/* Generic unlock here as we might release R or W lock */
1347 		MRW_UNLOCK();
1348 		return error;
1349 	}
1350 
1351 	/*
1352 	 * END: MCAST ROUTING HOT PATH
1353 	 */
1354 
1355 	/* Further processing must be done with WLOCK taken */
1356 	if ((MRW_WOWNED() == 0) && (MRW_LOCK_TRY_UPGRADE() == 0)) {
1357 		MRW_RUNLOCK();
1358 		MRW_WLOCK();
1359 		goto mfc_find_retry;
1360 	}
1361 
1362 	/*
1363 	 * If we don't have a route for packet's origin,
1364 	 * Make a copy of the packet & send message to routing daemon
1365 	 */
1366 	hlen = ip->ip_hl << 2;
1367 
1368 	MRTSTAT_INC(mrts_mfc_misses);
1369 	MRTSTAT_INC(mrts_no_route);
1370 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1371 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1372 
1373 	/*
1374 	 * Allocate mbufs early so that we don't do extra work if we are
1375 	 * just going to fail anyway.  Make sure to pullup the header so
1376 	 * that other people can't step on it.
1377 	 */
1378 	rte = (struct rtdetq*) malloc((sizeof *rte), M_MRTABLE,
1379 	    M_NOWAIT|M_ZERO);
1380 	if (rte == NULL) {
1381 		MRW_WUNLOCK();
1382 		return ENOBUFS;
1383 	}
1384 
1385 	mb0 = m_copypacket(m, M_NOWAIT);
1386 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1387 		mb0 = m_pullup(mb0, hlen);
1388 	if (mb0 == NULL) {
1389 		free(rte, M_MRTABLE);
1390 		MRW_WUNLOCK();
1391 		return ENOBUFS;
1392 	}
1393 
1394 	/* is there an upcall waiting for this flow ? */
1395 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1396 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash)
1397 	{
1398 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1399 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1400 		    !buf_ring_empty(rt->mfc_stall_ring))
1401 			break;
1402 	}
1403 
1404 	if (rt == NULL) {
1405 		int i;
1406 		struct igmpmsg *im;
1407 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1408 		struct mbuf *mm;
1409 
1410 		/*
1411 		 * Locate the vifi for the incoming interface for this packet.
1412 		 * If none found, drop packet.
1413 		 */
1414 		for (vifi = 0; vifi < V_numvifs &&
1415 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1416 			;
1417 		if (vifi >= V_numvifs) /* vif not found, drop packet */
1418 			goto non_fatal;
1419 
1420 		/* no upcall, so make a new entry */
1421 		rt = mfc_alloc();
1422 		if (rt == NULL)
1423 			goto fail;
1424 
1425 		/* Make a copy of the header to send to the user level process */
1426 		mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1427 		if (mm == NULL)
1428 			goto fail1;
1429 
1430 		/*
1431 		 * Send message to routing daemon to install
1432 		 * a route into the kernel table
1433 		 */
1434 
1435 		im = mtod(mm, struct igmpmsg*);
1436 		im->im_msgtype = IGMPMSG_NOCACHE;
1437 		im->im_mbz = 0;
1438 		im->im_vif = vifi;
1439 
1440 		MRTSTAT_INC(mrts_upcalls);
1441 
1442 		k_igmpsrc.sin_addr = ip->ip_src;
1443 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1444 			CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1445 			MRTSTAT_INC(mrts_upq_sockfull);
1446 			fail1: free(rt, M_MRTABLE);
1447 			fail: free(rte, M_MRTABLE);
1448 			m_freem(mb0);
1449 			MRW_WUNLOCK();
1450 			return ENOBUFS;
1451 		}
1452 
1453 		/* insert new entry at head of hash chain */
1454 		rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1455 		rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1456 		rt->mfc_expire = UPCALL_EXPIRE;
1457 		V_nexpire[hash]++;
1458 		for (i = 0; i < V_numvifs; i++) {
1459 			rt->mfc_ttls[i] = 0;
1460 			rt->mfc_flags[i] = 0;
1461 		}
1462 		rt->mfc_parent = -1;
1463 
1464 		/* clear the RP address */
1465 		rt->mfc_rp.s_addr = INADDR_ANY;
1466 		rt->mfc_bw_meter_leq = NULL;
1467 		rt->mfc_bw_meter_geq = NULL;
1468 
1469 		/* initialize pkt counters per src-grp */
1470 		rt->mfc_pkt_cnt = 0;
1471 		rt->mfc_byte_cnt = 0;
1472 		rt->mfc_wrong_if = 0;
1473 		timevalclear(&rt->mfc_last_assert);
1474 
1475 		buf_ring_enqueue(rt->mfc_stall_ring, rte);
1476 
1477 		/* Add RT to hashtable as it didn't exist before */
1478 		LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1479 	} else {
1480 		/* determine if queue has overflowed */
1481 		if (buf_ring_full(rt->mfc_stall_ring)) {
1482 			MRTSTAT_INC(mrts_upq_ovflw);
1483 			non_fatal: free(rte, M_MRTABLE);
1484 			m_freem(mb0);
1485 			MRW_WUNLOCK();
1486 			return (0);
1487 		}
1488 
1489 		buf_ring_enqueue(rt->mfc_stall_ring, rte);
1490 	}
1491 
1492 	rte->m = mb0;
1493 	rte->ifp = ifp;
1494 
1495 	MRW_WUNLOCK();
1496 
1497 	return 0;
1498 }
1499 
1500 /*
1501  * Clean up the cache entry if upcall is not serviced
1502  */
1503 static void
1504 expire_upcalls(void *arg)
1505 {
1506     u_long i;
1507 
1508     CURVNET_SET((struct vnet *) arg);
1509 
1510     /*This callout is always run with MRW_WLOCK taken. */
1511 
1512     for (i = 0; i < mfchashsize; i++) {
1513 	struct mfc *rt, *nrt;
1514 
1515 	if (V_nexpire[i] == 0)
1516 	    continue;
1517 
1518 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1519 		if (buf_ring_empty(rt->mfc_stall_ring))
1520 			continue;
1521 
1522 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1523 			continue;
1524 
1525 		MRTSTAT_INC(mrts_cache_cleanups);
1526 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1527 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1528 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1529 
1530 		expire_mfc(rt);
1531 	    }
1532     }
1533 
1534     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1535 	curvnet);
1536 
1537     CURVNET_RESTORE();
1538 }
1539 
1540 /*
1541  * Packet forwarding routine once entry in the cache is made
1542  */
1543 static int
1544 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1545 {
1546     struct ip  *ip = mtod(m, struct ip *);
1547     vifi_t vifi;
1548     int plen = ntohs(ip->ip_len);
1549 
1550     MRW_LOCK_ASSERT();
1551 
1552     /*
1553      * If xmt_vif is not -1, send on only the requested vif.
1554      *
1555      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1556      */
1557     if (xmt_vif < V_numvifs) {
1558 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1559 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1560 	else
1561 		phyint_send(ip, V_viftable + xmt_vif, m);
1562 	return 1;
1563     }
1564 
1565     /*
1566      * Don't forward if it didn't arrive from the parent vif for its origin.
1567      */
1568     vifi = rt->mfc_parent;
1569     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1570 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1571 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1572 	MRTSTAT_INC(mrts_wrong_if);
1573 	++rt->mfc_wrong_if;
1574 	/*
1575 	 * If we are doing PIM assert processing, send a message
1576 	 * to the routing daemon.
1577 	 *
1578 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1579 	 * can complete the SPT switch, regardless of the type
1580 	 * of the iif (broadcast media, GRE tunnel, etc).
1581 	 */
1582 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1583 	    V_viftable[vifi].v_ifp) {
1584 	    if (ifp == V_multicast_register_if)
1585 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1586 
1587 	    /* Get vifi for the incoming packet */
1588 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1589 		vifi++)
1590 		;
1591 	    if (vifi >= V_numvifs)
1592 		return 0;	/* The iif is not found: ignore the packet. */
1593 
1594 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1595 		return 0;	/* WRONGVIF disabled: ignore the packet */
1596 
1597 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1598 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1599 		struct igmpmsg *im;
1600 		int hlen = ip->ip_hl << 2;
1601 		struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1602 
1603 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1604 		    mm = m_pullup(mm, hlen);
1605 		if (mm == NULL)
1606 		    return ENOBUFS;
1607 
1608 		im = mtod(mm, struct igmpmsg *);
1609 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1610 		im->im_mbz		= 0;
1611 		im->im_vif		= vifi;
1612 
1613 		MRTSTAT_INC(mrts_upcalls);
1614 
1615 		k_igmpsrc.sin_addr = im->im_src;
1616 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1617 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1618 		    MRTSTAT_INC(mrts_upq_sockfull);
1619 		    return ENOBUFS;
1620 		}
1621 	    }
1622 	}
1623 	return 0;
1624     }
1625 
1626     /* If I sourced this packet, it counts as output, else it was input. */
1627     mtx_lock_spin(&V_viftable[vifi].v_spin);
1628     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1629 	V_viftable[vifi].v_pkt_out++;
1630 	V_viftable[vifi].v_bytes_out += plen;
1631     } else {
1632 	V_viftable[vifi].v_pkt_in++;
1633 	V_viftable[vifi].v_bytes_in += plen;
1634     }
1635     mtx_unlock_spin(&V_viftable[vifi].v_spin);
1636 
1637     rt->mfc_pkt_cnt++;
1638     rt->mfc_byte_cnt += plen;
1639 
1640     /*
1641      * For each vif, decide if a copy of the packet should be forwarded.
1642      * Forward if:
1643      *		- the ttl exceeds the vif's threshold
1644      *		- there are group members downstream on interface
1645      */
1646     for (vifi = 0; vifi < V_numvifs; vifi++)
1647 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1648 	    V_viftable[vifi].v_pkt_out++;
1649 	    V_viftable[vifi].v_bytes_out += plen;
1650 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1651 		pim_register_send(ip, V_viftable + vifi, m, rt);
1652 	    else
1653 		phyint_send(ip, V_viftable + vifi, m);
1654 	}
1655 
1656     /*
1657      * Perform upcall-related bw measuring.
1658      */
1659     if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1660 	struct bw_meter *x;
1661 	struct timeval now;
1662 
1663 	microtime(&now);
1664 	/* Process meters for Greater-or-EQual case */
1665 	for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1666 		bw_meter_geq_receive_packet(x, plen, &now);
1667 
1668 	/* Process meters for Lower-or-EQual case */
1669 	for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1670 		/*
1671 		 * Record that a packet is received.
1672 		 * Spin lock has to be taken as callout context
1673 		 * (expire_bw_meter_leq) might modify these fields
1674 		 * as well
1675 		 */
1676 		mtx_lock_spin(&x->bm_spin);
1677 		x->bm_measured.b_packets++;
1678 		x->bm_measured.b_bytes += plen;
1679 		mtx_unlock_spin(&x->bm_spin);
1680 	}
1681     }
1682 
1683     return 0;
1684 }
1685 
1686 /*
1687  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1688  */
1689 static int
1690 X_legal_vif_num(int vif)
1691 {
1692 	int ret;
1693 
1694 	ret = 0;
1695 	if (vif < 0)
1696 		return (ret);
1697 
1698 	MRW_RLOCK();
1699 	if (vif < V_numvifs)
1700 		ret = 1;
1701 	MRW_RUNLOCK();
1702 
1703 	return (ret);
1704 }
1705 
1706 /*
1707  * Return the local address used by this vif
1708  */
1709 static u_long
1710 X_ip_mcast_src(int vifi)
1711 {
1712 	in_addr_t addr;
1713 
1714 	addr = INADDR_ANY;
1715 	if (vifi < 0)
1716 		return (addr);
1717 
1718 	MRW_RLOCK();
1719 	if (vifi < V_numvifs)
1720 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1721 	MRW_RUNLOCK();
1722 
1723 	return (addr);
1724 }
1725 
1726 static void
1727 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1728 {
1729     struct mbuf *mb_copy;
1730     int hlen = ip->ip_hl << 2;
1731 
1732     MRW_LOCK_ASSERT();
1733 
1734     /*
1735      * Make a new reference to the packet; make sure that
1736      * the IP header is actually copied, not just referenced,
1737      * so that ip_output() only scribbles on the copy.
1738      */
1739     mb_copy = m_copypacket(m, M_NOWAIT);
1740     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1741 	mb_copy = m_pullup(mb_copy, hlen);
1742     if (mb_copy == NULL)
1743 	return;
1744 
1745     send_packet(vifp, mb_copy);
1746 }
1747 
1748 static void
1749 send_packet(struct vif *vifp, struct mbuf *m)
1750 {
1751 	struct ip_moptions imo;
1752 	int error __unused;
1753 
1754 	MRW_LOCK_ASSERT();
1755 
1756 	imo.imo_multicast_ifp  = vifp->v_ifp;
1757 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1758 	imo.imo_multicast_loop = !!in_mcast_loop;
1759 	imo.imo_multicast_vif  = -1;
1760 	STAILQ_INIT(&imo.imo_head);
1761 
1762 	/*
1763 	 * Re-entrancy should not be a problem here, because
1764 	 * the packets that we send out and are looped back at us
1765 	 * should get rejected because they appear to come from
1766 	 * the loopback interface, thus preventing looping.
1767 	 */
1768 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1769 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1770 	    (ptrdiff_t)(vifp - V_viftable), error);
1771 }
1772 
1773 /*
1774  * Stubs for old RSVP socket shim implementation.
1775  */
1776 
1777 static int
1778 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1779 {
1780 
1781 	return (EOPNOTSUPP);
1782 }
1783 
1784 static void
1785 X_ip_rsvp_force_done(struct socket *so __unused)
1786 {
1787 
1788 }
1789 
1790 static int
1791 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1792 {
1793 	struct mbuf *m;
1794 
1795 	m = *mp;
1796 	*mp = NULL;
1797 	if (!V_rsvp_on)
1798 		m_freem(m);
1799 	return (IPPROTO_DONE);
1800 }
1801 
1802 /*
1803  * Code for bandwidth monitors
1804  */
1805 
1806 /*
1807  * Define common interface for timeval-related methods
1808  */
1809 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1810 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1811 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1812 
1813 static uint32_t
1814 compute_bw_meter_flags(struct bw_upcall *req)
1815 {
1816     uint32_t flags = 0;
1817 
1818     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1819 	flags |= BW_METER_UNIT_PACKETS;
1820     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1821 	flags |= BW_METER_UNIT_BYTES;
1822     if (req->bu_flags & BW_UPCALL_GEQ)
1823 	flags |= BW_METER_GEQ;
1824     if (req->bu_flags & BW_UPCALL_LEQ)
1825 	flags |= BW_METER_LEQ;
1826 
1827     return flags;
1828 }
1829 
1830 static void
1831 expire_bw_meter_leq(void *arg)
1832 {
1833 	struct bw_meter *x = arg;
1834 	struct timeval now;
1835 	/*
1836 	 * INFO:
1837 	 * callout is always executed with MRW_WLOCK taken
1838 	 */
1839 
1840 	CURVNET_SET((struct vnet *)x->arg);
1841 
1842 	microtime(&now);
1843 
1844 	/*
1845 	 * Test if we should deliver an upcall
1846 	 */
1847 	if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1848 	    (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1849 	    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1850 	    (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1851 		/* Prepare an upcall for delivery */
1852 		bw_meter_prepare_upcall(x, &now);
1853 	}
1854 
1855 	/* Send all upcalls that are pending delivery */
1856 	taskqueue_enqueue(V_task_queue, &V_task);
1857 
1858 	/* Reset counters */
1859 	x->bm_start_time = now;
1860 	/* Spin lock has to be taken as ip_forward context
1861 	 * might modify these fields as well
1862 	 */
1863 	mtx_lock_spin(&x->bm_spin);
1864 	x->bm_measured.b_bytes = 0;
1865 	x->bm_measured.b_packets = 0;
1866 	mtx_unlock_spin(&x->bm_spin);
1867 
1868 	callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1869 
1870 	CURVNET_RESTORE();
1871 }
1872 
1873 /*
1874  * Add a bw_meter entry
1875  */
1876 static int
1877 add_bw_upcall(struct bw_upcall *req)
1878 {
1879 	struct mfc *mfc;
1880 	struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1881 	BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1882 	struct timeval now;
1883 	struct bw_meter *x, **bwm_ptr;
1884 	uint32_t flags;
1885 
1886 	if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1887 		return EOPNOTSUPP;
1888 
1889 	/* Test if the flags are valid */
1890 	if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1891 		return EINVAL;
1892 	if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1893 		return EINVAL;
1894 	if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1895 			== (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1896 		return EINVAL;
1897 
1898 	/* Test if the threshold time interval is valid */
1899 	if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1900 		return EINVAL;
1901 
1902 	flags = compute_bw_meter_flags(req);
1903 
1904 	/*
1905 	 * Find if we have already same bw_meter entry
1906 	 */
1907 	MRW_WLOCK();
1908 	mfc = mfc_find(&req->bu_src, &req->bu_dst);
1909 	if (mfc == NULL) {
1910 		MRW_WUNLOCK();
1911 		return EADDRNOTAVAIL;
1912 	}
1913 
1914 	/* Choose an appropriate bw_meter list */
1915 	if (req->bu_flags & BW_UPCALL_GEQ)
1916 		bwm_ptr = &mfc->mfc_bw_meter_geq;
1917 	else
1918 		bwm_ptr = &mfc->mfc_bw_meter_leq;
1919 
1920 	for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
1921 		if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1922 		    &req->bu_threshold.b_time, ==))
1923 		    && (x->bm_threshold.b_packets
1924 		    == req->bu_threshold.b_packets)
1925 		    && (x->bm_threshold.b_bytes
1926 		    == req->bu_threshold.b_bytes)
1927 		    && (x->bm_flags & BW_METER_USER_FLAGS)
1928 		    == flags) {
1929 			MRW_WUNLOCK();
1930 			return 0; /* XXX Already installed */
1931 		}
1932 	}
1933 
1934 	/* Allocate the new bw_meter entry */
1935 	x = (struct bw_meter*) malloc(sizeof(*x), M_BWMETER,
1936 	    M_ZERO | M_NOWAIT);
1937 	if (x == NULL) {
1938 		MRW_WUNLOCK();
1939 		return ENOBUFS;
1940 	}
1941 
1942 	/* Set the new bw_meter entry */
1943 	x->bm_threshold.b_time = req->bu_threshold.b_time;
1944 	microtime(&now);
1945 	x->bm_start_time = now;
1946 	x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1947 	x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1948 	x->bm_measured.b_packets = 0;
1949 	x->bm_measured.b_bytes = 0;
1950 	x->bm_flags = flags;
1951 	x->bm_time_next = NULL;
1952 	x->bm_mfc = mfc;
1953 	x->arg = curvnet;
1954 	sprintf(x->bm_spin_name, "BM spin %p", x);
1955 	mtx_init(&x->bm_spin, x->bm_spin_name, NULL, MTX_SPIN);
1956 
1957 	/* For LEQ case create periodic callout */
1958 	if (req->bu_flags & BW_UPCALL_LEQ) {
1959 		callout_init_rw(&x->bm_meter_callout, &mrouter_mtx, CALLOUT_SHAREDLOCK);
1960 		callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
1961 		    expire_bw_meter_leq, x);
1962 	}
1963 
1964 	/* Add the new bw_meter entry to the front of entries for this MFC */
1965 	x->bm_mfc_next = *bwm_ptr;
1966 	*bwm_ptr = x;
1967 
1968 	MRW_WUNLOCK();
1969 
1970 	return 0;
1971 }
1972 
1973 static void
1974 free_bw_list(struct bw_meter *list)
1975 {
1976     while (list != NULL) {
1977 	struct bw_meter *x = list;
1978 
1979 	/* MRW_WLOCK must be held here */
1980 	if (x->bm_flags & BW_METER_LEQ) {
1981 		callout_drain(&x->bm_meter_callout);
1982 		mtx_destroy(&x->bm_spin);
1983 	}
1984 
1985 	list = list->bm_mfc_next;
1986 	free(x, M_BWMETER);
1987     }
1988 }
1989 
1990 /*
1991  * Delete one or multiple bw_meter entries
1992  */
1993 static int
1994 del_bw_upcall(struct bw_upcall *req)
1995 {
1996     struct mfc *mfc;
1997     struct bw_meter *x, **bwm_ptr;
1998 
1999     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
2000 	return EOPNOTSUPP;
2001 
2002     MRW_WLOCK();
2003 
2004     /* Find the corresponding MFC entry */
2005     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2006     if (mfc == NULL) {
2007 	MRW_WUNLOCK();
2008 	return EADDRNOTAVAIL;
2009     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2010 	/*
2011 	 * Delete all bw_meter entries for this mfc
2012 	 */
2013 	struct bw_meter *list;
2014 
2015 	/* Free LEQ list */
2016 	list = mfc->mfc_bw_meter_leq;
2017 	mfc->mfc_bw_meter_leq = NULL;
2018 	free_bw_list(list);
2019 
2020 	/* Free GEQ list */
2021 	list = mfc->mfc_bw_meter_geq;
2022 	mfc->mfc_bw_meter_geq = NULL;
2023 	free_bw_list(list);
2024 	MRW_WUNLOCK();
2025 	return 0;
2026     } else {			/* Delete a single bw_meter entry */
2027 	struct bw_meter *prev;
2028 	uint32_t flags = 0;
2029 
2030 	flags = compute_bw_meter_flags(req);
2031 
2032 	/* Choose an appropriate bw_meter list */
2033 	if (req->bu_flags & BW_UPCALL_GEQ)
2034 		bwm_ptr = &mfc->mfc_bw_meter_geq;
2035 	else
2036 		bwm_ptr = &mfc->mfc_bw_meter_leq;
2037 
2038 	/* Find the bw_meter entry to delete */
2039 	for (prev = NULL, x = *bwm_ptr; x != NULL;
2040 	     prev = x, x = x->bm_mfc_next) {
2041 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2042 			       &req->bu_threshold.b_time, ==)) &&
2043 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2044 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2045 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2046 		break;
2047 	}
2048 	if (x != NULL) { /* Delete entry from the list for this MFC */
2049 	    if (prev != NULL)
2050 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2051 	    else
2052 		*bwm_ptr = x->bm_mfc_next;/* new head of list */
2053 
2054 	    if (req->bu_flags & BW_UPCALL_LEQ)
2055 		    callout_stop(&x->bm_meter_callout);
2056 
2057 	    MRW_WUNLOCK();
2058 	    /* Free the bw_meter entry */
2059 	    free(x, M_BWMETER);
2060 	    return 0;
2061 	} else {
2062 	    MRW_WUNLOCK();
2063 	    return EINVAL;
2064 	}
2065     }
2066     /* NOTREACHED */
2067 }
2068 
2069 /*
2070  * Perform bandwidth measurement processing that may result in an upcall
2071  */
2072 static void
2073 bw_meter_geq_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2074 {
2075 	struct timeval delta;
2076 
2077 	MRW_LOCK_ASSERT();
2078 
2079 	delta = *nowp;
2080 	BW_TIMEVALDECR(&delta, &x->bm_start_time);
2081 
2082 	/*
2083 	 * Processing for ">=" type of bw_meter entry.
2084 	 * bm_spin does not have to be hold here as in GEQ
2085 	 * case this is the only context accessing bm_measured.
2086 	 */
2087 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2088 	    /* Reset the bw_meter entry */
2089 	    x->bm_start_time = *nowp;
2090 	    x->bm_measured.b_packets = 0;
2091 	    x->bm_measured.b_bytes = 0;
2092 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2093 	}
2094 
2095 	/* Record that a packet is received */
2096 	x->bm_measured.b_packets++;
2097 	x->bm_measured.b_bytes += plen;
2098 
2099 	/*
2100 	 * Test if we should deliver an upcall
2101 	 */
2102 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2103 		if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2104 		    (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2105 		    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2106 		    (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2107 			/* Prepare an upcall for delivery */
2108 			bw_meter_prepare_upcall(x, nowp);
2109 			x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2110 		}
2111 	}
2112 }
2113 
2114 /*
2115  * Prepare a bandwidth-related upcall
2116  */
2117 static void
2118 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2119 {
2120 	struct timeval delta;
2121 	struct bw_upcall *u;
2122 
2123 	MRW_LOCK_ASSERT();
2124 
2125 	/*
2126 	 * Compute the measured time interval
2127 	 */
2128 	delta = *nowp;
2129 	BW_TIMEVALDECR(&delta, &x->bm_start_time);
2130 
2131 	/*
2132 	 * Set the bw_upcall entry
2133 	 */
2134 	u = malloc(sizeof(struct bw_upcall), M_MRTABLE, M_NOWAIT | M_ZERO);
2135 	if (!u) {
2136 		log(LOG_WARNING, "bw_meter_prepare_upcall: cannot allocate entry\n");
2137 		return;
2138 	}
2139 	u->bu_src = x->bm_mfc->mfc_origin;
2140 	u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2141 	u->bu_threshold.b_time = x->bm_threshold.b_time;
2142 	u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2143 	u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2144 	u->bu_measured.b_time = delta;
2145 	u->bu_measured.b_packets = x->bm_measured.b_packets;
2146 	u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2147 	u->bu_flags = 0;
2148 	if (x->bm_flags & BW_METER_UNIT_PACKETS)
2149 		u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2150 	if (x->bm_flags & BW_METER_UNIT_BYTES)
2151 		u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2152 	if (x->bm_flags & BW_METER_GEQ)
2153 		u->bu_flags |= BW_UPCALL_GEQ;
2154 	if (x->bm_flags & BW_METER_LEQ)
2155 		u->bu_flags |= BW_UPCALL_LEQ;
2156 
2157 	if (buf_ring_enqueue(V_bw_upcalls_ring, u))
2158 		log(LOG_WARNING, "bw_meter_prepare_upcall: cannot enqueue upcall\n");
2159 	if (buf_ring_count(V_bw_upcalls_ring) > (BW_UPCALLS_MAX / 2)) {
2160 		taskqueue_enqueue(V_task_queue, &V_task);
2161 	}
2162 }
2163 /*
2164  * Send the pending bandwidth-related upcalls
2165  */
2166 static void
2167 bw_upcalls_send(void)
2168 {
2169     struct mbuf *m;
2170     int len = 0;
2171     struct bw_upcall *bu;
2172     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2173     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2174 				      0,		/* unused2 */
2175 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2176 				      0,		/* im_mbz  */
2177 				      0,		/* im_vif  */
2178 				      0,		/* unused3 */
2179 				      { 0 },		/* im_src  */
2180 				      { 0 } };		/* im_dst  */
2181 
2182     MRW_LOCK_ASSERT();
2183 
2184     if (buf_ring_empty(V_bw_upcalls_ring))
2185 	return;
2186 
2187     /*
2188      * Allocate a new mbuf, initialize it with the header and
2189      * the payload for the pending calls.
2190      */
2191     m = m_gethdr(M_NOWAIT, MT_DATA);
2192     if (m == NULL) {
2193 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2194 	return;
2195     }
2196 
2197     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2198     len += sizeof(struct igmpmsg);
2199     while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
2200 	m_copyback(m, len, sizeof(struct bw_upcall), (caddr_t)bu);
2201 	len += sizeof(struct bw_upcall);
2202 	free(bu, M_MRTABLE);
2203     }
2204 
2205     /*
2206      * Send the upcalls
2207      * XXX do we need to set the address in k_igmpsrc ?
2208      */
2209     MRTSTAT_INC(mrts_upcalls);
2210     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2211 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2212 	MRTSTAT_INC(mrts_upq_sockfull);
2213     }
2214 }
2215 
2216 /*
2217  * A periodic function for sending all upcalls that are pending delivery
2218  */
2219 static void
2220 expire_bw_upcalls_send(void *arg)
2221 {
2222     CURVNET_SET((struct vnet *) arg);
2223 
2224     /* This callout is run with MRW_RLOCK taken */
2225 
2226     bw_upcalls_send();
2227 
2228     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2229 	curvnet);
2230     CURVNET_RESTORE();
2231 }
2232 
2233 /*
2234  * End of bandwidth monitoring code
2235  */
2236 
2237 /*
2238  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2239  *
2240  */
2241 static int
2242 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2243     struct mfc *rt)
2244 {
2245     struct mbuf *mb_copy, *mm;
2246 
2247     /*
2248      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2249      * rendezvous point was unspecified, and we were told not to.
2250      */
2251     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2252 	in_nullhost(rt->mfc_rp))
2253 	return 0;
2254 
2255     mb_copy = pim_register_prepare(ip, m);
2256     if (mb_copy == NULL)
2257 	return ENOBUFS;
2258 
2259     /*
2260      * Send all the fragments. Note that the mbuf for each fragment
2261      * is freed by the sending machinery.
2262      */
2263     for (mm = mb_copy; mm; mm = mb_copy) {
2264 	mb_copy = mm->m_nextpkt;
2265 	mm->m_nextpkt = 0;
2266 	mm = m_pullup(mm, sizeof(struct ip));
2267 	if (mm != NULL) {
2268 	    ip = mtod(mm, struct ip *);
2269 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2270 		pim_register_send_rp(ip, vifp, mm, rt);
2271 	    } else {
2272 		pim_register_send_upcall(ip, vifp, mm, rt);
2273 	    }
2274 	}
2275     }
2276 
2277     return 0;
2278 }
2279 
2280 /*
2281  * Return a copy of the data packet that is ready for PIM Register
2282  * encapsulation.
2283  * XXX: Note that in the returned copy the IP header is a valid one.
2284  */
2285 static struct mbuf *
2286 pim_register_prepare(struct ip *ip, struct mbuf *m)
2287 {
2288     struct mbuf *mb_copy = NULL;
2289     int mtu;
2290 
2291     /* Take care of delayed checksums */
2292     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2293 	in_delayed_cksum(m);
2294 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2295     }
2296 
2297     /*
2298      * Copy the old packet & pullup its IP header into the
2299      * new mbuf so we can modify it.
2300      */
2301     mb_copy = m_copypacket(m, M_NOWAIT);
2302     if (mb_copy == NULL)
2303 	return NULL;
2304     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2305     if (mb_copy == NULL)
2306 	return NULL;
2307 
2308     /* take care of the TTL */
2309     ip = mtod(mb_copy, struct ip *);
2310     --ip->ip_ttl;
2311 
2312     /* Compute the MTU after the PIM Register encapsulation */
2313     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2314 
2315     if (ntohs(ip->ip_len) <= mtu) {
2316 	/* Turn the IP header into a valid one */
2317 	ip->ip_sum = 0;
2318 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2319     } else {
2320 	/* Fragment the packet */
2321 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2322 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2323 	    m_freem(mb_copy);
2324 	    return NULL;
2325 	}
2326     }
2327     return mb_copy;
2328 }
2329 
2330 /*
2331  * Send an upcall with the data packet to the user-level process.
2332  */
2333 static int
2334 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2335     struct mbuf *mb_copy, struct mfc *rt)
2336 {
2337     struct mbuf *mb_first;
2338     int len = ntohs(ip->ip_len);
2339     struct igmpmsg *im;
2340     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2341 
2342     MRW_LOCK_ASSERT();
2343 
2344     /*
2345      * Add a new mbuf with an upcall header
2346      */
2347     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2348     if (mb_first == NULL) {
2349 	m_freem(mb_copy);
2350 	return ENOBUFS;
2351     }
2352     mb_first->m_data += max_linkhdr;
2353     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2354     mb_first->m_len = sizeof(struct igmpmsg);
2355     mb_first->m_next = mb_copy;
2356 
2357     /* Send message to routing daemon */
2358     im = mtod(mb_first, struct igmpmsg *);
2359     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2360     im->im_mbz		= 0;
2361     im->im_vif		= vifp - V_viftable;
2362     im->im_src		= ip->ip_src;
2363     im->im_dst		= ip->ip_dst;
2364 
2365     k_igmpsrc.sin_addr	= ip->ip_src;
2366 
2367     MRTSTAT_INC(mrts_upcalls);
2368 
2369     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2370 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2371 	MRTSTAT_INC(mrts_upq_sockfull);
2372 	return ENOBUFS;
2373     }
2374 
2375     /* Keep statistics */
2376     PIMSTAT_INC(pims_snd_registers_msgs);
2377     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2378 
2379     return 0;
2380 }
2381 
2382 /*
2383  * Encapsulate the data packet in PIM Register message and send it to the RP.
2384  */
2385 static int
2386 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2387     struct mfc *rt)
2388 {
2389     struct mbuf *mb_first;
2390     struct ip *ip_outer;
2391     struct pim_encap_pimhdr *pimhdr;
2392     int len = ntohs(ip->ip_len);
2393     vifi_t vifi = rt->mfc_parent;
2394 
2395     MRW_LOCK_ASSERT();
2396 
2397     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2398 	m_freem(mb_copy);
2399 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2400     }
2401 
2402     /*
2403      * Add a new mbuf with the encapsulating header
2404      */
2405     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2406     if (mb_first == NULL) {
2407 	m_freem(mb_copy);
2408 	return ENOBUFS;
2409     }
2410     mb_first->m_data += max_linkhdr;
2411     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2412     mb_first->m_next = mb_copy;
2413 
2414     mb_first->m_pkthdr.len = len + mb_first->m_len;
2415 
2416     /*
2417      * Fill in the encapsulating IP and PIM header
2418      */
2419     ip_outer = mtod(mb_first, struct ip *);
2420     *ip_outer = pim_encap_iphdr;
2421     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2422 	sizeof(pim_encap_pimhdr));
2423     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2424     ip_outer->ip_dst = rt->mfc_rp;
2425     /*
2426      * Copy the inner header TOS to the outer header, and take care of the
2427      * IP_DF bit.
2428      */
2429     ip_outer->ip_tos = ip->ip_tos;
2430     if (ip->ip_off & htons(IP_DF))
2431 	ip_outer->ip_off |= htons(IP_DF);
2432     ip_fillid(ip_outer);
2433     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2434 					 + sizeof(pim_encap_iphdr));
2435     *pimhdr = pim_encap_pimhdr;
2436     /* If the iif crosses a border, set the Border-bit */
2437     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2438 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2439 
2440     mb_first->m_data += sizeof(pim_encap_iphdr);
2441     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2442     mb_first->m_data -= sizeof(pim_encap_iphdr);
2443 
2444     send_packet(vifp, mb_first);
2445 
2446     /* Keep statistics */
2447     PIMSTAT_INC(pims_snd_registers_msgs);
2448     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2449 
2450     return 0;
2451 }
2452 
2453 /*
2454  * pim_encapcheck() is called by the encap4_input() path at runtime to
2455  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2456  * into the kernel.
2457  */
2458 static int
2459 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2460     int proto __unused, void *arg __unused)
2461 {
2462 
2463     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2464     return (8);		/* claim the datagram. */
2465 }
2466 
2467 /*
2468  * PIM-SMv2 and PIM-DM messages processing.
2469  * Receives and verifies the PIM control messages, and passes them
2470  * up to the listening socket, using rip_input().
2471  * The only message with special processing is the PIM_REGISTER message
2472  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2473  * is passed to if_simloop().
2474  */
2475 static int
2476 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2477 {
2478     struct ip *ip = mtod(m, struct ip *);
2479     struct pim *pim;
2480     int iphlen = off;
2481     int minlen;
2482     int datalen = ntohs(ip->ip_len) - iphlen;
2483     int ip_tos;
2484 
2485     /* Keep statistics */
2486     PIMSTAT_INC(pims_rcv_total_msgs);
2487     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2488 
2489     /*
2490      * Validate lengths
2491      */
2492     if (datalen < PIM_MINLEN) {
2493 	PIMSTAT_INC(pims_rcv_tooshort);
2494 	CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2495 	    __func__, datalen, ntohl(ip->ip_src.s_addr));
2496 	m_freem(m);
2497 	return (IPPROTO_DONE);
2498     }
2499 
2500     /*
2501      * If the packet is at least as big as a REGISTER, go agead
2502      * and grab the PIM REGISTER header size, to avoid another
2503      * possible m_pullup() later.
2504      *
2505      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2506      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2507      */
2508     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2509     /*
2510      * Get the IP and PIM headers in contiguous memory, and
2511      * possibly the PIM REGISTER header.
2512      */
2513     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2514 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2515 	return (IPPROTO_DONE);
2516     }
2517 
2518     /* m_pullup() may have given us a new mbuf so reset ip. */
2519     ip = mtod(m, struct ip *);
2520     ip_tos = ip->ip_tos;
2521 
2522     /* adjust mbuf to point to the PIM header */
2523     m->m_data += iphlen;
2524     m->m_len  -= iphlen;
2525     pim = mtod(m, struct pim *);
2526 
2527     /*
2528      * Validate checksum. If PIM REGISTER, exclude the data packet.
2529      *
2530      * XXX: some older PIMv2 implementations don't make this distinction,
2531      * so for compatibility reason perform the checksum over part of the
2532      * message, and if error, then over the whole message.
2533      */
2534     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2535 	/* do nothing, checksum okay */
2536     } else if (in_cksum(m, datalen)) {
2537 	PIMSTAT_INC(pims_rcv_badsum);
2538 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2539 	m_freem(m);
2540 	return (IPPROTO_DONE);
2541     }
2542 
2543     /* PIM version check */
2544     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2545 	PIMSTAT_INC(pims_rcv_badversion);
2546 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2547 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2548 	m_freem(m);
2549 	return (IPPROTO_DONE);
2550     }
2551 
2552     /* restore mbuf back to the outer IP */
2553     m->m_data -= iphlen;
2554     m->m_len  += iphlen;
2555 
2556     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2557 	/*
2558 	 * Since this is a REGISTER, we'll make a copy of the register
2559 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2560 	 * routing daemon.
2561 	 */
2562 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2563 	struct mbuf *mcp;
2564 	struct ip *encap_ip;
2565 	u_int32_t *reghdr;
2566 	struct ifnet *vifp;
2567 
2568 	MRW_RLOCK();
2569 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2570 	    MRW_RUNLOCK();
2571 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2572 		(int)V_reg_vif_num);
2573 	    m_freem(m);
2574 	    return (IPPROTO_DONE);
2575 	}
2576 	/* XXX need refcnt? */
2577 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2578 	MRW_RUNLOCK();
2579 
2580 	/*
2581 	 * Validate length
2582 	 */
2583 	if (datalen < PIM_REG_MINLEN) {
2584 	    PIMSTAT_INC(pims_rcv_tooshort);
2585 	    PIMSTAT_INC(pims_rcv_badregisters);
2586 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2587 	    m_freem(m);
2588 	    return (IPPROTO_DONE);
2589 	}
2590 
2591 	reghdr = (u_int32_t *)(pim + 1);
2592 	encap_ip = (struct ip *)(reghdr + 1);
2593 
2594 	CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2595 	    __func__, ntohl(encap_ip->ip_src.s_addr),
2596 	    ntohs(encap_ip->ip_len));
2597 
2598 	/* verify the version number of the inner packet */
2599 	if (encap_ip->ip_v != IPVERSION) {
2600 	    PIMSTAT_INC(pims_rcv_badregisters);
2601 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2602 	    m_freem(m);
2603 	    return (IPPROTO_DONE);
2604 	}
2605 
2606 	/* verify the inner packet is destined to a mcast group */
2607 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2608 	    PIMSTAT_INC(pims_rcv_badregisters);
2609 	    CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2610 		ntohl(encap_ip->ip_dst.s_addr));
2611 	    m_freem(m);
2612 	    return (IPPROTO_DONE);
2613 	}
2614 
2615 	/* If a NULL_REGISTER, pass it to the daemon */
2616 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2617 	    goto pim_input_to_daemon;
2618 
2619 	/*
2620 	 * Copy the TOS from the outer IP header to the inner IP header.
2621 	 */
2622 	if (encap_ip->ip_tos != ip_tos) {
2623 	    /* Outer TOS -> inner TOS */
2624 	    encap_ip->ip_tos = ip_tos;
2625 	    /* Recompute the inner header checksum. Sigh... */
2626 
2627 	    /* adjust mbuf to point to the inner IP header */
2628 	    m->m_data += (iphlen + PIM_MINLEN);
2629 	    m->m_len  -= (iphlen + PIM_MINLEN);
2630 
2631 	    encap_ip->ip_sum = 0;
2632 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2633 
2634 	    /* restore mbuf to point back to the outer IP header */
2635 	    m->m_data -= (iphlen + PIM_MINLEN);
2636 	    m->m_len  += (iphlen + PIM_MINLEN);
2637 	}
2638 
2639 	/*
2640 	 * Decapsulate the inner IP packet and loopback to forward it
2641 	 * as a normal multicast packet. Also, make a copy of the
2642 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2643 	 * to pass to the daemon later, so it can take the appropriate
2644 	 * actions (e.g., send back PIM_REGISTER_STOP).
2645 	 * XXX: here m->m_data points to the outer IP header.
2646 	 */
2647 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2648 	if (mcp == NULL) {
2649 	    CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2650 	    m_freem(m);
2651 	    return (IPPROTO_DONE);
2652 	}
2653 
2654 	/* Keep statistics */
2655 	/* XXX: registers_bytes include only the encap. mcast pkt */
2656 	PIMSTAT_INC(pims_rcv_registers_msgs);
2657 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2658 
2659 	/*
2660 	 * forward the inner ip packet; point m_data at the inner ip.
2661 	 */
2662 	m_adj(m, iphlen + PIM_MINLEN);
2663 
2664 	CTR4(KTR_IPMF,
2665 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2666 	    __func__,
2667 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2668 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2669 	    (int)V_reg_vif_num);
2670 
2671 	/* NB: vifp was collected above; can it change on us? */
2672 	if_simloop(vifp, m, dst.sin_family, 0);
2673 
2674 	/* prepare the register head to send to the mrouting daemon */
2675 	m = mcp;
2676     }
2677 
2678 pim_input_to_daemon:
2679     /*
2680      * Pass the PIM message up to the daemon; if it is a Register message,
2681      * pass the 'head' only up to the daemon. This includes the
2682      * outer IP header, PIM header, PIM-Register header and the
2683      * inner IP header.
2684      * XXX: the outer IP header pkt size of a Register is not adjust to
2685      * reflect the fact that the inner multicast data is truncated.
2686      */
2687     return (rip_input(&m, &off, proto));
2688 }
2689 
2690 static int
2691 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2692 {
2693 	struct mfc	*rt;
2694 	int		 error, i;
2695 
2696 	if (req->newptr)
2697 		return (EPERM);
2698 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2699 		return (0);
2700 	error = sysctl_wire_old_buffer(req, 0);
2701 	if (error)
2702 		return (error);
2703 
2704 	MRW_RLOCK();
2705 	for (i = 0; i < mfchashsize; i++) {
2706 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2707 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2708 			if (error)
2709 				goto out_locked;
2710 		}
2711 	}
2712 out_locked:
2713 	MRW_RUNLOCK();
2714 	return (error);
2715 }
2716 
2717 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2718     CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2719     "IPv4 Multicast Forwarding Table "
2720     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2721 
2722 static int
2723 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2724 {
2725 	int error;
2726 
2727 	if (req->newptr)
2728 		return (EPERM);
2729 	if (V_viftable == NULL)		/* XXX unlocked */
2730 		return (0);
2731 	error = sysctl_wire_old_buffer(req, sizeof(*V_viftable) * MAXVIFS);
2732 	if (error)
2733 		return (error);
2734 
2735 	MRW_RLOCK();
2736 	error = SYSCTL_OUT(req, V_viftable, sizeof(*V_viftable) * MAXVIFS);
2737 	MRW_RUNLOCK();
2738 	return (error);
2739 }
2740 
2741 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2742     CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2743     sysctl_viflist, "S,vif[MAXVIFS]",
2744     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2745 
2746 static void
2747 vnet_mroute_init(const void *unused __unused)
2748 {
2749 
2750 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2751 
2752 	V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2753 	    M_MRTABLE, M_WAITOK|M_ZERO);
2754 
2755 	callout_init_rw(&V_expire_upcalls_ch, &mrouter_mtx, 0);
2756 	callout_init_rw(&V_bw_upcalls_ch, &mrouter_mtx, 0);
2757 
2758 	/* Prepare taskqueue */
2759 	V_task_queue = taskqueue_create_fast("ip_mroute_tskq", M_NOWAIT,
2760 		    taskqueue_thread_enqueue, &V_task_queue);
2761 	taskqueue_start_threads(&V_task_queue, 1, PI_NET, "ip_mroute_tskq task");
2762 }
2763 
2764 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2765 	NULL);
2766 
2767 static void
2768 vnet_mroute_uninit(const void *unused __unused)
2769 {
2770 
2771 	/* Taskqueue should be cancelled and drained before freeing */
2772 	taskqueue_free(V_task_queue);
2773 
2774 	free(V_viftable, M_MRTABLE);
2775 	free(V_nexpire, M_MRTABLE);
2776 	V_nexpire = NULL;
2777 }
2778 
2779 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2780 	vnet_mroute_uninit, NULL);
2781 
2782 static int
2783 ip_mroute_modevent(module_t mod, int type, void *unused)
2784 {
2785 
2786     switch (type) {
2787     case MOD_LOAD:
2788 	MRW_LOCK_INIT();
2789 
2790 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2791 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2792 	if (if_detach_event_tag == NULL) {
2793 		printf("ip_mroute: unable to register "
2794 		    "ifnet_departure_event handler\n");
2795 		MRW_LOCK_DESTROY();
2796 		return (EINVAL);
2797 	}
2798 
2799 	mfchashsize = MFCHASHSIZE;
2800 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2801 	    !powerof2(mfchashsize)) {
2802 		printf("WARNING: %s not a power of 2; using default\n",
2803 		    "net.inet.ip.mfchashsize");
2804 		mfchashsize = MFCHASHSIZE;
2805 	}
2806 
2807 	pim_squelch_wholepkt = 0;
2808 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2809 	    &pim_squelch_wholepkt);
2810 
2811 	pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2812 	if (pim_encap_cookie == NULL) {
2813 		printf("ip_mroute: unable to attach pim encap\n");
2814 		MRW_LOCK_DESTROY();
2815 		return (EINVAL);
2816 	}
2817 
2818 	ip_mcast_src = X_ip_mcast_src;
2819 	ip_mforward = X_ip_mforward;
2820 	ip_mrouter_done = X_ip_mrouter_done;
2821 	ip_mrouter_get = X_ip_mrouter_get;
2822 	ip_mrouter_set = X_ip_mrouter_set;
2823 
2824 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2825 	ip_rsvp_vif = X_ip_rsvp_vif;
2826 
2827 	legal_vif_num = X_legal_vif_num;
2828 	mrt_ioctl = X_mrt_ioctl;
2829 	rsvp_input_p = X_rsvp_input;
2830 	break;
2831 
2832     case MOD_UNLOAD:
2833 	/*
2834 	 * Typically module unload happens after the user-level
2835 	 * process has shutdown the kernel services (the check
2836 	 * below insures someone can't just yank the module out
2837 	 * from under a running process).  But if the module is
2838 	 * just loaded and then unloaded w/o starting up a user
2839 	 * process we still need to cleanup.
2840 	 */
2841 	MRW_WLOCK();
2842 	if (ip_mrouter_cnt != 0) {
2843 	    MRW_WUNLOCK();
2844 	    return (EINVAL);
2845 	}
2846 	ip_mrouter_unloading = 1;
2847 	MRW_WUNLOCK();
2848 
2849 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2850 
2851 	if (pim_encap_cookie) {
2852 	    ip_encap_detach(pim_encap_cookie);
2853 	    pim_encap_cookie = NULL;
2854 	}
2855 
2856 	ip_mcast_src = NULL;
2857 	ip_mforward = NULL;
2858 	ip_mrouter_done = NULL;
2859 	ip_mrouter_get = NULL;
2860 	ip_mrouter_set = NULL;
2861 
2862 	ip_rsvp_force_done = NULL;
2863 	ip_rsvp_vif = NULL;
2864 
2865 	legal_vif_num = NULL;
2866 	mrt_ioctl = NULL;
2867 	rsvp_input_p = NULL;
2868 
2869 	MRW_LOCK_DESTROY();
2870 	break;
2871 
2872     default:
2873 	return EOPNOTSUPP;
2874     }
2875     return 0;
2876 }
2877 
2878 static moduledata_t ip_mroutemod = {
2879     "ip_mroute",
2880     ip_mroute_modevent,
2881     0
2882 };
2883 
2884 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2885