xref: /freebsd/sys/netinet/ip_mroute.c (revision b773da180e02c0c27de48dbba72f3fb7a64656da)
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 
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  * Modified by Wojciech Macek, Semihalf, May 2021
51  *
52  * MROUTING Revision: 3.5
53  * and PIM-SMv2 and PIM-DM support, advanced API support,
54  * bandwidth metering and signaling
55  */
56 
57 /*
58  * TODO: Prefix functions with ipmf_.
59  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
60  * domain attachment (if_afdata) so we can track consumers of that service.
61  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
62  * move it to socket options.
63  * TODO: Cleanup LSRR removal further.
64  * TODO: Push RSVP stubs into raw_ip.c.
65  * TODO: Use bitstring.h for vif set.
66  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
67  * TODO: Sync ip6_mroute.c with this file.
68  */
69 
70 #include "opt_inet.h"
71 #include "opt_mrouting.h"
72 
73 #define _PIM_VT 1
74 
75 #include <sys/types.h>
76 #include <sys/param.h>
77 #include <sys/kernel.h>
78 #include <sys/stddef.h>
79 #include <sys/condvar.h>
80 #include <sys/eventhandler.h>
81 #include <sys/lock.h>
82 #include <sys/kthread.h>
83 #include <sys/ktr.h>
84 #include <sys/malloc.h>
85 #include <sys/mbuf.h>
86 #include <sys/module.h>
87 #include <sys/priv.h>
88 #include <sys/protosw.h>
89 #include <sys/signalvar.h>
90 #include <sys/socket.h>
91 #include <sys/socketvar.h>
92 #include <sys/sockio.h>
93 #include <sys/sx.h>
94 #include <sys/sysctl.h>
95 #include <sys/syslog.h>
96 #include <sys/systm.h>
97 #include <sys/taskqueue.h>
98 #include <sys/time.h>
99 #include <sys/counter.h>
100 #include <machine/atomic.h>
101 
102 #include <net/if.h>
103 #include <net/if_var.h>
104 #include <net/if_private.h>
105 #include <net/if_types.h>
106 #include <net/netisr.h>
107 #include <net/route.h>
108 #include <net/vnet.h>
109 
110 #include <netinet/in.h>
111 #include <netinet/igmp.h>
112 #include <netinet/in_systm.h>
113 #include <netinet/in_var.h>
114 #include <netinet/ip.h>
115 #include <netinet/ip_encap.h>
116 #include <netinet/ip_mroute.h>
117 #include <netinet/ip_var.h>
118 #include <netinet/ip_options.h>
119 #include <netinet/pim.h>
120 #include <netinet/pim_var.h>
121 #include <netinet/udp.h>
122 
123 #include <machine/in_cksum.h>
124 
125 #ifndef KTR_IPMF
126 #define KTR_IPMF KTR_INET
127 #endif
128 
129 #define		VIFI_INVALID	((vifi_t) -1)
130 
131 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
132 
133 /*
134  * Locking.  We use two locks: one for the virtual interface table and
135  * one for the forwarding table.  These locks may be nested in which case
136  * the VIF lock must always be taken first.  Note that each lock is used
137  * to cover not only the specific data structure but also related data
138  * structures.
139  */
140 
141 static struct sx __exclusive_cache_line mrouter_teardown;
142 #define	MRW_TEARDOWN_WLOCK()	sx_xlock(&mrouter_teardown)
143 #define	MRW_TEARDOWN_WUNLOCK()	sx_xunlock(&mrouter_teardown)
144 #define	MRW_TEARDOWN_LOCK_INIT()				\
145 	sx_init(&mrouter_teardown, "IPv4 multicast forwarding teardown")
146 #define	MRW_TEARDOWN_LOCK_DESTROY()	sx_destroy(&mrouter_teardown)
147 
148 static struct rwlock mrouter_lock;
149 #define	MRW_RLOCK()		rw_rlock(&mrouter_lock)
150 #define	MRW_WLOCK()		rw_wlock(&mrouter_lock)
151 #define	MRW_RUNLOCK()		rw_runlock(&mrouter_lock)
152 #define	MRW_WUNLOCK()		rw_wunlock(&mrouter_lock)
153 #define	MRW_UNLOCK()		rw_unlock(&mrouter_lock)
154 #define	MRW_LOCK_ASSERT()	rw_assert(&mrouter_lock, RA_LOCKED)
155 #define	MRW_WLOCK_ASSERT()	rw_assert(&mrouter_lock, RA_WLOCKED)
156 #define	MRW_LOCK_TRY_UPGRADE()	rw_try_upgrade(&mrouter_lock)
157 #define	MRW_WOWNED()		rw_wowned(&mrouter_lock)
158 #define	MRW_LOCK_INIT()						\
159 	rw_init(&mrouter_lock, "IPv4 multicast forwarding")
160 #define	MRW_LOCK_DESTROY()	rw_destroy(&mrouter_lock)
161 
162 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
163 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
164 
165 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
166 VNET_PCPUSTAT_SYSINIT(mrtstat);
167 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
168 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
169     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
170     "netinet/ip_mroute.h)");
171 
172 VNET_DEFINE_STATIC(u_long, mfchash);
173 #define	V_mfchash		VNET(mfchash)
174 #define	MFCHASH(a, g)							\
175 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
176 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
177 #define	MFCHASHSIZE	256
178 
179 static u_long mfchashsize = MFCHASHSIZE;	/* Hash size */
180 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, mfchashsize, CTLFLAG_RDTUN,
181     &mfchashsize, 0, "IPv4 Multicast Forwarding Table hash size");
182 VNET_DEFINE_STATIC(u_char *, nexpire);		/* 0..mfchashsize-1 */
183 #define	V_nexpire		VNET(nexpire)
184 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
185 #define	V_mfchashtbl		VNET(mfchashtbl)
186 VNET_DEFINE_STATIC(struct taskqueue *, task_queue);
187 #define	V_task_queue		VNET(task_queue)
188 VNET_DEFINE_STATIC(struct task, task);
189 #define	V_task		VNET(task)
190 
191 VNET_DEFINE_STATIC(vifi_t, numvifs);
192 #define	V_numvifs		VNET(numvifs)
193 VNET_DEFINE_STATIC(struct vif *, viftable);
194 #define	V_viftable		VNET(viftable)
195 
196 static eventhandler_tag if_detach_event_tag = NULL;
197 
198 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
199 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
200 
201 VNET_DEFINE_STATIC(struct mtx, buf_ring_mtx);
202 #define	V_buf_ring_mtx	VNET(buf_ring_mtx)
203 
204 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
205 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
206 
207 /*
208  * Bandwidth meter variables and constants
209  */
210 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
211 
212 /*
213  * Pending upcalls are stored in a ring which is flushed when
214  * full, or periodically
215  */
216 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
217 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
218 VNET_DEFINE_STATIC(struct buf_ring *, bw_upcalls_ring);
219 #define	V_bw_upcalls_ring    	VNET(bw_upcalls_ring)
220 VNET_DEFINE_STATIC(struct mtx, bw_upcalls_ring_mtx);
221 #define	V_bw_upcalls_ring_mtx    	VNET(bw_upcalls_ring_mtx)
222 
223 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
224 
225 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
226 VNET_PCPUSTAT_SYSINIT(pimstat);
227 VNET_PCPUSTAT_SYSUNINIT(pimstat);
228 
229 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
230     "PIM");
231 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
232     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
233 
234 static u_long	pim_squelch_wholepkt = 0;
235 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RWTUN,
236     &pim_squelch_wholepkt, 0,
237     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
238 
239 static const struct encaptab *pim_encap_cookie;
240 static int pim_encapcheck(const struct mbuf *, int, int, void *);
241 static int pim_input(struct mbuf *, int, int, void *);
242 
243 extern int in_mcast_loop;
244 
245 static const struct encap_config ipv4_encap_cfg = {
246 	.proto = IPPROTO_PIM,
247 	.min_length = sizeof(struct ip) + PIM_MINLEN,
248 	.exact_match = 8,
249 	.check = pim_encapcheck,
250 	.input = pim_input
251 };
252 
253 /*
254  * Note: the PIM Register encapsulation adds the following in front of a
255  * data packet:
256  *
257  * struct pim_encap_hdr {
258  *    struct ip ip;
259  *    struct pim_encap_pimhdr  pim;
260  * }
261  *
262  */
263 
264 struct pim_encap_pimhdr {
265 	struct pim pim;
266 	uint32_t   flags;
267 };
268 #define		PIM_ENCAP_TTL	64
269 
270 static struct ip pim_encap_iphdr = {
271 #if BYTE_ORDER == LITTLE_ENDIAN
272 	sizeof(struct ip) >> 2,
273 	IPVERSION,
274 #else
275 	IPVERSION,
276 	sizeof(struct ip) >> 2,
277 #endif
278 	0,			/* tos */
279 	sizeof(struct ip),	/* total length */
280 	0,			/* id */
281 	0,			/* frag offset */
282 	PIM_ENCAP_TTL,
283 	IPPROTO_PIM,
284 	0,			/* checksum */
285 };
286 
287 static struct pim_encap_pimhdr pim_encap_pimhdr = {
288     {
289 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
290 	0,			/* reserved */
291 	0,			/* checksum */
292     },
293     0				/* flags */
294 };
295 
296 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
297 #define	V_reg_vif_num		VNET(reg_vif_num)
298 VNET_DEFINE_STATIC(struct ifnet *, multicast_register_if);
299 #define	V_multicast_register_if	VNET(multicast_register_if)
300 
301 /*
302  * Private variables.
303  */
304 
305 static u_long	X_ip_mcast_src(int);
306 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
307 		    struct ip_moptions *);
308 static int	X_ip_mrouter_done(void);
309 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
310 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
311 static int	X_legal_vif_num(int);
312 static int	X_mrt_ioctl(u_long, caddr_t, int);
313 
314 static int	add_bw_upcall(struct bw_upcall *);
315 static int	add_mfc(struct mfcctl2 *);
316 static int	add_vif(struct vifctl *);
317 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
318 static void	bw_meter_geq_receive_packet(struct bw_meter *, int,
319 		    struct timeval *);
320 static void	bw_upcalls_send(void);
321 static int	del_bw_upcall(struct bw_upcall *);
322 static int	del_mfc(struct mfcctl2 *);
323 static int	del_vif(vifi_t);
324 static int	del_vif_locked(vifi_t, struct ifnet **, struct ifnet **);
325 static void	expire_bw_upcalls_send(void *);
326 static void	expire_mfc(struct mfc *);
327 static void	expire_upcalls(void *);
328 static void	free_bw_list(struct bw_meter *);
329 static int	get_sg_cnt(struct sioc_sg_req *);
330 static int	get_vif_cnt(struct sioc_vif_req *);
331 static void	if_detached_event(void *, struct ifnet *);
332 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
333 static int	ip_mrouter_init(struct socket *, int);
334 static __inline struct mfc *
335 		mfc_find(struct in_addr *, struct in_addr *);
336 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
337 static struct mbuf *
338 		pim_register_prepare(struct ip *, struct mbuf *);
339 static int	pim_register_send(struct ip *, struct vif *,
340 		    struct mbuf *, struct mfc *);
341 static int	pim_register_send_rp(struct ip *, struct vif *,
342 		    struct mbuf *, struct mfc *);
343 static int	pim_register_send_upcall(struct ip *, struct vif *,
344 		    struct mbuf *, struct mfc *);
345 static void	send_packet(struct vif *, struct mbuf *);
346 static int	set_api_config(uint32_t *);
347 static int	set_assert(int);
348 static int	socket_send(struct socket *, struct mbuf *,
349 		    struct sockaddr_in *);
350 
351 /*
352  * Kernel multicast forwarding API capabilities and setup.
353  * If more API capabilities are added to the kernel, they should be
354  * recorded in `mrt_api_support'.
355  */
356 #define MRT_API_VERSION		0x0305
357 
358 static const int mrt_api_version = MRT_API_VERSION;
359 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
360 					 MRT_MFC_FLAGS_BORDER_VIF |
361 					 MRT_MFC_RP |
362 					 MRT_MFC_BW_UPCALL);
363 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
364 #define	V_mrt_api_config	VNET(mrt_api_config)
365 VNET_DEFINE_STATIC(int, pim_assert_enabled);
366 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
367 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
368 
369 /*
370  * Find a route for a given origin IP address and multicast group address.
371  * Statistics must be updated by the caller.
372  */
373 static __inline struct mfc *
374 mfc_find(struct in_addr *o, struct in_addr *g)
375 {
376 	struct mfc *rt;
377 
378 	/*
379 	 * Might be called both RLOCK and WLOCK.
380 	 * Check if any, it's caller responsibility
381 	 * to choose correct option.
382 	 */
383 	MRW_LOCK_ASSERT();
384 
385 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
386 		if (in_hosteq(rt->mfc_origin, *o) &&
387 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
388 		    buf_ring_empty(rt->mfc_stall_ring))
389 			break;
390 	}
391 
392 	return (rt);
393 }
394 
395 static __inline struct mfc *
396 mfc_alloc(void)
397 {
398 	struct mfc *rt;
399 	rt = malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT | M_ZERO);
400 	if (rt == NULL)
401 		return rt;
402 
403 	rt->mfc_stall_ring = buf_ring_alloc(MAX_UPQ, M_MRTABLE,
404 	    M_NOWAIT, &V_buf_ring_mtx);
405 	if (rt->mfc_stall_ring == NULL) {
406 		free(rt, M_MRTABLE);
407 		return NULL;
408 	}
409 
410 	return rt;
411 }
412 
413 /*
414  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
415  */
416 static int
417 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
418 {
419 	int error, optval;
420 	vifi_t vifi;
421 	struct vifctl vifc;
422 	struct mfcctl2 mfc;
423 	struct bw_upcall bw_upcall;
424 	uint32_t i;
425 
426 	if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
427 		return EPERM;
428 
429 	error = 0;
430 	switch (sopt->sopt_name) {
431 	case MRT_INIT:
432 		error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
433 		if (error)
434 			break;
435 		error = ip_mrouter_init(so, optval);
436 		break;
437 	case MRT_DONE:
438 		error = ip_mrouter_done();
439 		break;
440 	case MRT_ADD_VIF:
441 		error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
442 		if (error)
443 			break;
444 		error = add_vif(&vifc);
445 		break;
446 	case MRT_DEL_VIF:
447 		error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
448 		if (error)
449 			break;
450 		error = del_vif(vifi);
451 		break;
452 	case MRT_ADD_MFC:
453 	case MRT_DEL_MFC:
454 		/*
455 		 * select data size depending on API version.
456 		 */
457 		if (sopt->sopt_name == MRT_ADD_MFC &&
458 		    V_mrt_api_config & MRT_API_FLAGS_ALL) {
459 			error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
460 			    sizeof(struct mfcctl2));
461 		} else {
462 			error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
463 			    sizeof(struct mfcctl));
464 			bzero((caddr_t)&mfc + sizeof(struct mfcctl),
465 			    sizeof(mfc) - sizeof(struct mfcctl));
466 		}
467 		if (error)
468 			break;
469 		if (sopt->sopt_name == MRT_ADD_MFC)
470 			error = add_mfc(&mfc);
471 		else
472 			error = del_mfc(&mfc);
473 		break;
474 
475 	case MRT_ASSERT:
476 		error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
477 		if (error)
478 			break;
479 		set_assert(optval);
480 		break;
481 
482 	case MRT_API_CONFIG:
483 		error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
484 		if (!error)
485 			error = set_api_config(&i);
486 		if (!error)
487 			error = sooptcopyout(sopt, &i, sizeof i);
488 		break;
489 
490 	case MRT_ADD_BW_UPCALL:
491 	case MRT_DEL_BW_UPCALL:
492 		error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
493 		    sizeof bw_upcall);
494 		if (error)
495 			break;
496 		if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
497 			error = add_bw_upcall(&bw_upcall);
498 		else
499 			error = del_bw_upcall(&bw_upcall);
500 		break;
501 
502 	default:
503 		error = EOPNOTSUPP;
504 		break;
505 	}
506 	return error;
507 }
508 
509 /*
510  * Handle MRT getsockopt commands
511  */
512 static int
513 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
514 {
515 	int error;
516 
517 	switch (sopt->sopt_name) {
518 	case MRT_VERSION:
519 		error = sooptcopyout(sopt, &mrt_api_version,
520 		    sizeof mrt_api_version);
521 		break;
522 	case MRT_ASSERT:
523 		error = sooptcopyout(sopt, &V_pim_assert_enabled,
524 		    sizeof V_pim_assert_enabled);
525 		break;
526 	case MRT_API_SUPPORT:
527 		error = sooptcopyout(sopt, &mrt_api_support,
528 		    sizeof mrt_api_support);
529 		break;
530 	case MRT_API_CONFIG:
531 		error = sooptcopyout(sopt, &V_mrt_api_config,
532 		    sizeof V_mrt_api_config);
533 		break;
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;
548 
549 	error = priv_check(curthread, PRIV_NETINET_MROUTE);
550 	if (error)
551 		return (error);
552 	switch (cmd) {
553 	case (SIOCGETVIFCNT):
554 		error = get_vif_cnt((struct sioc_vif_req *)data);
555 		break;
556 
557 	case (SIOCGETSGCNT):
558 		error = get_sg_cnt((struct sioc_sg_req *)data);
559 		break;
560 
561 	default:
562 		error = EINVAL;
563 		break;
564 	}
565 	return error;
566 }
567 
568 /*
569  * returns the packet, byte, rpf-failure count for the source group provided
570  */
571 static int
572 get_sg_cnt(struct sioc_sg_req *req)
573 {
574 	struct mfc *rt;
575 
576 	MRW_RLOCK();
577 	rt = mfc_find(&req->src, &req->grp);
578 	if (rt == NULL) {
579 		MRW_RUNLOCK();
580 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
581 		return EADDRNOTAVAIL;
582 	}
583 	req->pktcnt = rt->mfc_pkt_cnt;
584 	req->bytecnt = rt->mfc_byte_cnt;
585 	req->wrong_if = rt->mfc_wrong_if;
586 	MRW_RUNLOCK();
587 	return 0;
588 }
589 
590 /*
591  * returns the input and output packet and byte counts on the vif provided
592  */
593 static int
594 get_vif_cnt(struct sioc_vif_req *req)
595 {
596 	vifi_t vifi = req->vifi;
597 
598 	MRW_RLOCK();
599 	if (vifi >= V_numvifs) {
600 		MRW_RUNLOCK();
601 		return EINVAL;
602 	}
603 
604 	mtx_lock(&V_viftable[vifi].v_mtx);
605 	req->icount = V_viftable[vifi].v_pkt_in;
606 	req->ocount = V_viftable[vifi].v_pkt_out;
607 	req->ibytes = V_viftable[vifi].v_bytes_in;
608 	req->obytes = V_viftable[vifi].v_bytes_out;
609 	mtx_unlock(&V_viftable[vifi].v_mtx);
610 	MRW_RUNLOCK();
611 
612 	return 0;
613 }
614 
615 static void
616 if_detached_event(void *arg __unused, struct ifnet *ifp)
617 {
618 	vifi_t vifi;
619 	u_long i, vifi_cnt = 0;
620 	struct ifnet *free_ptr, *multi_leave;
621 
622 	MRW_WLOCK();
623 
624 	if (V_ip_mrouter == NULL) {
625 		MRW_WUNLOCK();
626 		return;
627 	}
628 
629 	/*
630 	 * Tear down multicast forwarder state associated with this ifnet.
631 	 * 1. Walk the vif list, matching vifs against this ifnet.
632 	 * 2. Walk the multicast forwarding cache (mfc) looking for
633 	 *    inner matches with this vif's index.
634 	 * 3. Expire any matching multicast forwarding cache entries.
635 	 * 4. Free vif state. This should disable ALLMULTI on the interface.
636 	 */
637 restart:
638 	for (vifi = 0; vifi < V_numvifs; vifi++) {
639 		if (V_viftable[vifi].v_ifp != ifp)
640 			continue;
641 		for (i = 0; i < mfchashsize; i++) {
642 			struct mfc *rt, *nrt;
643 
644 			LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
645 				if (rt->mfc_parent == vifi) {
646 					expire_mfc(rt);
647 				}
648 			}
649 		}
650 		del_vif_locked(vifi, &multi_leave, &free_ptr);
651 		if (free_ptr != NULL)
652 			vifi_cnt++;
653 		if (multi_leave) {
654 			MRW_WUNLOCK();
655 			if_allmulti(multi_leave, 0);
656 			MRW_WLOCK();
657 			goto restart;
658 		}
659 	}
660 
661 	MRW_WUNLOCK();
662 
663 	/*
664 	 * Free IFP. We don't have to use free_ptr here as it is the same
665 	 * that ifp. Perform free as many times as required in case
666 	 * refcount is greater than 1.
667 	 */
668 	for (i = 0; i < vifi_cnt; i++)
669 		if_free(ifp);
670 }
671 
672 static void
673 ip_mrouter_upcall_thread(void *arg, int pending __unused)
674 {
675 	CURVNET_SET((struct vnet *) arg);
676 
677 	MRW_WLOCK();
678 	bw_upcalls_send();
679 	MRW_WUNLOCK();
680 
681 	CURVNET_RESTORE();
682 }
683 
684 /*
685  * Enable multicast forwarding.
686  */
687 static int
688 ip_mrouter_init(struct socket *so, int version)
689 {
690 
691 	CTR2(KTR_IPMF, "%s: so %p", __func__, so);
692 
693 	if (version != 1)
694 		return ENOPROTOOPT;
695 
696 	MRW_TEARDOWN_WLOCK();
697 	MRW_WLOCK();
698 
699 	if (ip_mrouter_unloading) {
700 		MRW_WUNLOCK();
701 		MRW_TEARDOWN_WUNLOCK();
702 		return ENOPROTOOPT;
703 	}
704 
705 	if (V_ip_mrouter != NULL) {
706 		MRW_WUNLOCK();
707 		MRW_TEARDOWN_WUNLOCK();
708 		return EADDRINUSE;
709 	}
710 
711 	V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
712 	    HASH_NOWAIT);
713 	if (V_mfchashtbl == NULL) {
714 		MRW_WUNLOCK();
715 		MRW_TEARDOWN_WUNLOCK();
716 		return (ENOMEM);
717 	}
718 
719 	/* Create upcall ring */
720 	mtx_init(&V_bw_upcalls_ring_mtx, "mroute upcall buf_ring mtx", NULL, MTX_DEF);
721 	V_bw_upcalls_ring = buf_ring_alloc(BW_UPCALLS_MAX, M_MRTABLE,
722 	    M_NOWAIT, &V_bw_upcalls_ring_mtx);
723 	if (!V_bw_upcalls_ring) {
724 		MRW_WUNLOCK();
725 		MRW_TEARDOWN_WUNLOCK();
726 		return (ENOMEM);
727 	}
728 
729 	TASK_INIT(&V_task, 0, ip_mrouter_upcall_thread, curvnet);
730 	taskqueue_cancel(V_task_queue, &V_task, NULL);
731 	taskqueue_unblock(V_task_queue);
732 
733 	callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
734 	    curvnet);
735 	callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
736 	    curvnet);
737 
738 	V_ip_mrouter = so;
739 	atomic_add_int(&ip_mrouter_cnt, 1);
740 
741 	/* This is a mutex required by buf_ring init, but not used internally */
742 	mtx_init(&V_buf_ring_mtx, "mroute buf_ring mtx", NULL, MTX_DEF);
743 
744 	MRW_WUNLOCK();
745 	MRW_TEARDOWN_WUNLOCK();
746 
747 	CTR1(KTR_IPMF, "%s: done", __func__);
748 
749 	return 0;
750 }
751 
752 /*
753  * Disable multicast forwarding.
754  */
755 static int
756 X_ip_mrouter_done(void)
757 {
758 	struct ifnet **ifps;
759 	int nifp;
760 	u_long i;
761 	vifi_t vifi;
762 	struct bw_upcall *bu;
763 
764 	MRW_TEARDOWN_WLOCK();
765 
766 	if (V_ip_mrouter == NULL) {
767 		MRW_TEARDOWN_WUNLOCK();
768 		return (EINVAL);
769 	}
770 
771 	/*
772 	 * Detach/disable hooks to the reset of the system.
773 	 */
774 	V_ip_mrouter = NULL;
775 	atomic_subtract_int(&ip_mrouter_cnt, 1);
776 	V_mrt_api_config = 0;
777 
778 	/*
779 	 * Wait for all epoch sections to complete to ensure
780 	 * V_ip_mrouter = NULL is visible to others.
781 	 */
782 	NET_EPOCH_WAIT();
783 
784 	/* Stop and drain task queue */
785 	taskqueue_block(V_task_queue);
786 	while (taskqueue_cancel(V_task_queue, &V_task, NULL)) {
787 		taskqueue_drain(V_task_queue, &V_task);
788 	}
789 
790 	ifps = malloc(MAXVIFS * sizeof(*ifps), M_TEMP, M_WAITOK);
791 
792 	MRW_WLOCK();
793 	taskqueue_cancel(V_task_queue, &V_task, NULL);
794 
795 	/* Destroy upcall ring */
796 	while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
797 		free(bu, M_MRTABLE);
798 	}
799 	buf_ring_free(V_bw_upcalls_ring, M_MRTABLE);
800 	mtx_destroy(&V_bw_upcalls_ring_mtx);
801 
802 	/*
803 	 * For each phyint in use, prepare to disable promiscuous reception
804 	 * of all IP multicasts.  Defer the actual call until the lock is released;
805 	 * just record the list of interfaces while locked.  Some interfaces use
806 	 * sx locks in their ioctl routines, which is not allowed while holding
807 	 * a non-sleepable lock.
808 	 */
809 	KASSERT(V_numvifs <= MAXVIFS, ("More vifs than possible"));
810 	for (vifi = 0, nifp = 0; vifi < V_numvifs; vifi++) {
811 		if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
812 		    !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
813 			ifps[nifp++] = V_viftable[vifi].v_ifp;
814 		}
815 	}
816 	bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
817 	V_numvifs = 0;
818 	V_pim_assert_enabled = 0;
819 
820 	callout_stop(&V_expire_upcalls_ch);
821 	callout_stop(&V_bw_upcalls_ch);
822 
823 	/*
824 	 * Free all multicast forwarding cache entries.
825 	 * Do not use hashdestroy(), as we must perform other cleanup.
826 	 */
827 	for (i = 0; i < mfchashsize; i++) {
828 		struct mfc *rt, *nrt;
829 
830 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
831 			expire_mfc(rt);
832 		}
833 	}
834 	free(V_mfchashtbl, M_MRTABLE);
835 	V_mfchashtbl = NULL;
836 
837 	bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
838 
839 	V_reg_vif_num = VIFI_INVALID;
840 
841 	mtx_destroy(&V_buf_ring_mtx);
842 
843 	MRW_WUNLOCK();
844 	MRW_TEARDOWN_WUNLOCK();
845 
846 	/*
847 	 * Now drop our claim on promiscuous multicast on the interfaces recorded
848 	 * above.  This is safe to do now because ALLMULTI is reference counted.
849 	 */
850 	for (vifi = 0; vifi < nifp; vifi++)
851 		if_allmulti(ifps[vifi], 0);
852 	free(ifps, M_TEMP);
853 
854 	CTR1(KTR_IPMF, "%s: done", __func__);
855 
856 	return 0;
857 }
858 
859 /*
860  * Set PIM assert processing global
861  */
862 static int
863 set_assert(int i)
864 {
865 	if ((i != 1) && (i != 0))
866 		return EINVAL;
867 
868 	V_pim_assert_enabled = i;
869 
870 	return 0;
871 }
872 
873 /*
874  * Configure API capabilities
875  */
876 int
877 set_api_config(uint32_t *apival)
878 {
879 	u_long i;
880 
881 	/*
882 	 * We can set the API capabilities only if it is the first operation
883 	 * after MRT_INIT. I.e.:
884 	 *  - there are no vifs installed
885 	 *  - pim_assert is not enabled
886 	 *  - the MFC table is empty
887 	 */
888 	if (V_numvifs > 0) {
889 		*apival = 0;
890 		return EPERM;
891 	}
892 	if (V_pim_assert_enabled) {
893 		*apival = 0;
894 		return EPERM;
895 	}
896 
897 	MRW_RLOCK();
898 
899 	for (i = 0; i < mfchashsize; i++) {
900 		if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
901 			MRW_RUNLOCK();
902 			*apival = 0;
903 			return EPERM;
904 		}
905 	}
906 
907 	MRW_RUNLOCK();
908 
909 	V_mrt_api_config = *apival & mrt_api_support;
910 	*apival = V_mrt_api_config;
911 
912 	return 0;
913 }
914 
915 /*
916  * Add a vif to the vif table
917  */
918 static int
919 add_vif(struct vifctl *vifcp)
920 {
921 	struct vif *vifp = V_viftable + vifcp->vifc_vifi;
922 	struct sockaddr_in sin = {sizeof sin, AF_INET};
923 	struct ifaddr *ifa;
924 	struct ifnet *ifp;
925 	int error;
926 
927 	if (vifcp->vifc_vifi >= MAXVIFS)
928 		return EINVAL;
929 	/* rate limiting is no longer supported by this code */
930 	if (vifcp->vifc_rate_limit != 0) {
931 		log(LOG_ERR, "rate limiting is no longer supported\n");
932 		return EINVAL;
933 	}
934 
935 	if (in_nullhost(vifcp->vifc_lcl_addr))
936 		return EADDRNOTAVAIL;
937 
938 	/* Find the interface with an address in AF_INET family */
939 	if (vifcp->vifc_flags & VIFF_REGISTER) {
940 		/*
941 		 * XXX: Because VIFF_REGISTER does not really need a valid
942 		 * local interface (e.g. it could be 127.0.0.2), we don't
943 		 * check its address.
944 		 */
945 		ifp = NULL;
946 	} else {
947 		struct epoch_tracker et;
948 
949 		sin.sin_addr = vifcp->vifc_lcl_addr;
950 		NET_EPOCH_ENTER(et);
951 		ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
952 		if (ifa == NULL) {
953 			NET_EPOCH_EXIT(et);
954 			return EADDRNOTAVAIL;
955 		}
956 		ifp = ifa->ifa_ifp;
957 		/* XXX FIXME we need to take a ref on ifp and cleanup properly! */
958 		NET_EPOCH_EXIT(et);
959 	}
960 
961 	if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
962 		CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
963 		return EOPNOTSUPP;
964 	} else if (vifcp->vifc_flags & VIFF_REGISTER) {
965 		ifp = V_multicast_register_if = if_alloc(IFT_LOOP);
966 		CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
967 		if (V_reg_vif_num == VIFI_INVALID) {
968 			if_initname(V_multicast_register_if, "register_vif", 0);
969 			V_reg_vif_num = vifcp->vifc_vifi;
970 		}
971 	} else {		/* Make sure the interface supports multicast */
972 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
973 			return EOPNOTSUPP;
974 
975 		/* Enable promiscuous reception of all IP multicasts from the if */
976 		error = if_allmulti(ifp, 1);
977 		if (error)
978 			return error;
979 	}
980 
981 	MRW_WLOCK();
982 
983 	if (!in_nullhost(vifp->v_lcl_addr)) {
984 		if (ifp)
985 			V_multicast_register_if = NULL;
986 		MRW_WUNLOCK();
987 		if (ifp)
988 			if_free(ifp);
989 		return EADDRINUSE;
990 	}
991 
992 	vifp->v_flags     = vifcp->vifc_flags;
993 	vifp->v_threshold = vifcp->vifc_threshold;
994 	vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
995 	vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
996 	vifp->v_ifp       = ifp;
997 	/* initialize per vif pkt counters */
998 	vifp->v_pkt_in    = 0;
999 	vifp->v_pkt_out   = 0;
1000 	vifp->v_bytes_in  = 0;
1001 	vifp->v_bytes_out = 0;
1002 	sprintf(vifp->v_mtx_name, "BM[%d] mtx", vifcp->vifc_vifi);
1003 	mtx_init(&vifp->v_mtx, vifp->v_mtx_name, NULL, MTX_DEF);
1004 
1005 	/* Adjust numvifs up if the vifi is higher than numvifs */
1006 	if (V_numvifs <= vifcp->vifc_vifi)
1007 		V_numvifs = vifcp->vifc_vifi + 1;
1008 
1009 	MRW_WUNLOCK();
1010 
1011 	CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
1012 	    (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
1013 	    (int)vifcp->vifc_threshold);
1014 
1015 	return 0;
1016 }
1017 
1018 /*
1019  * Delete a vif from the vif table
1020  */
1021 static int
1022 del_vif_locked(vifi_t vifi, struct ifnet **ifp_multi_leave, struct ifnet **ifp_free)
1023 {
1024 	struct vif *vifp;
1025 
1026 	*ifp_free = NULL;
1027 	*ifp_multi_leave = NULL;
1028 
1029 	MRW_WLOCK_ASSERT();
1030 
1031 	if (vifi >= V_numvifs) {
1032 		return EINVAL;
1033 	}
1034 	vifp = &V_viftable[vifi];
1035 	if (in_nullhost(vifp->v_lcl_addr)) {
1036 		return EADDRNOTAVAIL;
1037 	}
1038 
1039 	if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1040 		*ifp_multi_leave = vifp->v_ifp;
1041 
1042 	if (vifp->v_flags & VIFF_REGISTER) {
1043 		V_reg_vif_num = VIFI_INVALID;
1044 		if (vifp->v_ifp) {
1045 			if (vifp->v_ifp == V_multicast_register_if)
1046 				V_multicast_register_if = NULL;
1047 			*ifp_free = vifp->v_ifp;
1048 		}
1049 	}
1050 
1051 	mtx_destroy(&vifp->v_mtx);
1052 
1053 	bzero((caddr_t)vifp, sizeof (*vifp));
1054 
1055 	CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
1056 
1057 	/* Adjust numvifs down */
1058 	for (vifi = V_numvifs; vifi > 0; vifi--)
1059 		if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
1060 			break;
1061 	V_numvifs = vifi;
1062 
1063 	return 0;
1064 }
1065 
1066 static int
1067 del_vif(vifi_t vifi)
1068 {
1069 	int cc;
1070 	struct ifnet *free_ptr, *multi_leave;
1071 
1072 	MRW_WLOCK();
1073 	cc = del_vif_locked(vifi, &multi_leave, &free_ptr);
1074 	MRW_WUNLOCK();
1075 
1076 	if (multi_leave)
1077 		if_allmulti(multi_leave, 0);
1078 	if (free_ptr) {
1079 		if_free(free_ptr);
1080 	}
1081 
1082 	return cc;
1083 }
1084 
1085 /*
1086  * update an mfc entry without resetting counters and S,G addresses.
1087  */
1088 static void
1089 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1090 {
1091 	int i;
1092 
1093 	rt->mfc_parent = mfccp->mfcc_parent;
1094 	for (i = 0; i < V_numvifs; i++) {
1095 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1096 		rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1097 			MRT_MFC_FLAGS_ALL;
1098 	}
1099 	/* set the RP address */
1100 	if (V_mrt_api_config & MRT_MFC_RP)
1101 		rt->mfc_rp = mfccp->mfcc_rp;
1102 	else
1103 		rt->mfc_rp.s_addr = INADDR_ANY;
1104 }
1105 
1106 /*
1107  * fully initialize an mfc entry from the parameter.
1108  */
1109 static void
1110 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1111 {
1112 	rt->mfc_origin     = mfccp->mfcc_origin;
1113 	rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1114 
1115 	update_mfc_params(rt, mfccp);
1116 
1117 	/* initialize pkt counters per src-grp */
1118 	rt->mfc_pkt_cnt    = 0;
1119 	rt->mfc_byte_cnt   = 0;
1120 	rt->mfc_wrong_if   = 0;
1121 	timevalclear(&rt->mfc_last_assert);
1122 }
1123 
1124 static void
1125 expire_mfc(struct mfc *rt)
1126 {
1127 	struct rtdetq *rte;
1128 
1129 	MRW_WLOCK_ASSERT();
1130 
1131 	free_bw_list(rt->mfc_bw_meter_leq);
1132 	free_bw_list(rt->mfc_bw_meter_geq);
1133 
1134 	while (!buf_ring_empty(rt->mfc_stall_ring)) {
1135 		rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1136 		if (rte) {
1137 			m_freem(rte->m);
1138 			free(rte, M_MRTABLE);
1139 		}
1140 	}
1141 	buf_ring_free(rt->mfc_stall_ring, M_MRTABLE);
1142 
1143 	LIST_REMOVE(rt, mfc_hash);
1144 	free(rt, M_MRTABLE);
1145 }
1146 
1147 /*
1148  * Add an mfc entry
1149  */
1150 static int
1151 add_mfc(struct mfcctl2 *mfccp)
1152 {
1153 	struct mfc *rt;
1154 	struct rtdetq *rte;
1155 	u_long hash = 0;
1156 	u_short nstl;
1157 	struct epoch_tracker et;
1158 
1159 	MRW_WLOCK();
1160 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1161 
1162 	/* If an entry already exists, just update the fields */
1163 	if (rt) {
1164 		CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1165 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1166 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1167 		    mfccp->mfcc_parent);
1168 		update_mfc_params(rt, mfccp);
1169 		MRW_WUNLOCK();
1170 		return (0);
1171 	}
1172 
1173 	/*
1174 	 * Find the entry for which the upcall was made and update
1175 	 */
1176 	nstl = 0;
1177 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1178 	NET_EPOCH_ENTER(et);
1179 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1180 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1181 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1182 		    !buf_ring_empty(rt->mfc_stall_ring)) {
1183 			CTR5(KTR_IPMF,
1184 			   "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1185 			    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1186 			    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1187 			    mfccp->mfcc_parent,
1188 			    rt->mfc_stall_ring);
1189 			if (nstl++)
1190 				CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1191 
1192 			init_mfc_params(rt, mfccp);
1193 			rt->mfc_expire = 0;	/* Don't clean this guy up */
1194 			V_nexpire[hash]--;
1195 
1196 			/* Free queued packets, but attempt to forward them first. */
1197 			while (!buf_ring_empty(rt->mfc_stall_ring)) {
1198 				rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1199 				if (rte->ifp != NULL)
1200 					ip_mdq(rte->m, rte->ifp, rt, -1);
1201 				m_freem(rte->m);
1202 				free(rte, M_MRTABLE);
1203 			}
1204 		}
1205 	}
1206 	NET_EPOCH_EXIT(et);
1207 
1208 	/*
1209 	 * It is possible that an entry is being inserted without an upcall
1210 	 */
1211 	if (nstl == 0) {
1212 		CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1213 		LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1214 			if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1215 			    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1216 				init_mfc_params(rt, mfccp);
1217 				if (rt->mfc_expire)
1218 					V_nexpire[hash]--;
1219 				rt->mfc_expire = 0;
1220 				break; /* XXX */
1221 			}
1222 		}
1223 
1224 		if (rt == NULL) {		/* no upcall, so make a new entry */
1225 			rt = mfc_alloc();
1226 			if (rt == NULL) {
1227 				MRW_WUNLOCK();
1228 				return (ENOBUFS);
1229 			}
1230 
1231 			init_mfc_params(rt, mfccp);
1232 
1233 			rt->mfc_expire     = 0;
1234 			rt->mfc_bw_meter_leq = NULL;
1235 			rt->mfc_bw_meter_geq = NULL;
1236 
1237 			/* insert new entry at head of hash chain */
1238 			LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1239 		}
1240 	}
1241 
1242 	MRW_WUNLOCK();
1243 
1244 	return (0);
1245 }
1246 
1247 /*
1248  * Delete an mfc entry
1249  */
1250 static int
1251 del_mfc(struct mfcctl2 *mfccp)
1252 {
1253 	struct in_addr origin;
1254 	struct in_addr mcastgrp;
1255 	struct mfc *rt;
1256 
1257 	origin = mfccp->mfcc_origin;
1258 	mcastgrp = mfccp->mfcc_mcastgrp;
1259 
1260 	CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1261 			ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1262 
1263 	MRW_WLOCK();
1264 
1265 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(origin, mcastgrp)], mfc_hash) {
1266 		if (in_hosteq(rt->mfc_origin, origin) &&
1267 		    in_hosteq(rt->mfc_mcastgrp, mcastgrp))
1268 			break;
1269 	}
1270 	if (rt == NULL) {
1271 		MRW_WUNLOCK();
1272 		return EADDRNOTAVAIL;
1273 	}
1274 
1275 	expire_mfc(rt);
1276 
1277 	MRW_WUNLOCK();
1278 
1279 	return (0);
1280 }
1281 
1282 /*
1283  * Send a message to the routing daemon on the multicast routing socket.
1284  */
1285 static int
1286 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1287 {
1288 	if (s) {
1289 		SOCKBUF_LOCK(&s->so_rcv);
1290 		if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1291 		    NULL) != 0) {
1292 			sorwakeup_locked(s);
1293 			return 0;
1294 		}
1295 		soroverflow_locked(s);
1296 	}
1297 	m_freem(mm);
1298 	return -1;
1299 }
1300 
1301 /*
1302  * IP multicast forwarding function. This function assumes that the packet
1303  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1304  * pointed to by "ifp", and the packet is to be relayed to other networks
1305  * that have members of the packet's destination IP multicast group.
1306  *
1307  * The packet is returned unscathed to the caller, unless it is
1308  * erroneous, in which case a non-zero return value tells the caller to
1309  * discard it.
1310  */
1311 
1312 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1313 
1314 static int
1315 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1316     struct ip_moptions *imo)
1317 {
1318 	struct mfc *rt;
1319 	int error;
1320 	vifi_t vifi;
1321 	struct mbuf *mb0;
1322 	struct rtdetq *rte;
1323 	u_long hash;
1324 	int hlen;
1325 
1326 	M_ASSERTMAPPED(m);
1327 
1328 	CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1329 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1330 
1331 	if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1332 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1333 		/*
1334 		 * Packet arrived via a physical interface or
1335 		 * an encapsulated tunnel or a register_vif.
1336 		 */
1337 	} else {
1338 		/*
1339 		 * Packet arrived through a source-route tunnel.
1340 		 * Source-route tunnels are no longer supported.
1341 		 */
1342 		return (1);
1343 	}
1344 
1345 	/*
1346 	 * BEGIN: MCAST ROUTING HOT PATH
1347 	 */
1348 	MRW_RLOCK();
1349 	if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1350 		if (ip->ip_ttl < MAXTTL)
1351 			ip->ip_ttl++; /* compensate for -1 in *_send routines */
1352 		error = ip_mdq(m, ifp, NULL, vifi);
1353 		MRW_RUNLOCK();
1354 		return error;
1355 	}
1356 
1357 	/*
1358 	 * Don't forward a packet with time-to-live of zero or one,
1359 	 * or a packet destined to a local-only group.
1360 	 */
1361 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1362 		MRW_RUNLOCK();
1363 		return 0;
1364 	}
1365 
1366 mfc_find_retry:
1367 	/*
1368 	 * Determine forwarding vifs from the forwarding cache table
1369 	 */
1370 	MRTSTAT_INC(mrts_mfc_lookups);
1371 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1372 
1373 	/* Entry exists, so forward if necessary */
1374 	if (rt != NULL) {
1375 		error = ip_mdq(m, ifp, rt, -1);
1376 		/* Generic unlock here as we might release R or W lock */
1377 		MRW_UNLOCK();
1378 		return error;
1379 	}
1380 
1381 	/*
1382 	 * END: MCAST ROUTING HOT PATH
1383 	 */
1384 
1385 	/* Further processing must be done with WLOCK taken */
1386 	if ((MRW_WOWNED() == 0) && (MRW_LOCK_TRY_UPGRADE() == 0)) {
1387 		MRW_RUNLOCK();
1388 		MRW_WLOCK();
1389 		goto mfc_find_retry;
1390 	}
1391 
1392 	/*
1393 	 * If we don't have a route for packet's origin,
1394 	 * Make a copy of the packet & send message to routing daemon
1395 	 */
1396 	hlen = ip->ip_hl << 2;
1397 
1398 	MRTSTAT_INC(mrts_mfc_misses);
1399 	MRTSTAT_INC(mrts_no_route);
1400 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1401 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1402 
1403 	/*
1404 	 * Allocate mbufs early so that we don't do extra work if we are
1405 	 * just going to fail anyway.  Make sure to pullup the header so
1406 	 * that other people can't step on it.
1407 	 */
1408 	rte = malloc((sizeof *rte), M_MRTABLE, M_NOWAIT|M_ZERO);
1409 	if (rte == NULL) {
1410 		MRW_WUNLOCK();
1411 		return ENOBUFS;
1412 	}
1413 
1414 	mb0 = m_copypacket(m, M_NOWAIT);
1415 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1416 		mb0 = m_pullup(mb0, hlen);
1417 	if (mb0 == NULL) {
1418 		free(rte, M_MRTABLE);
1419 		MRW_WUNLOCK();
1420 		return ENOBUFS;
1421 	}
1422 
1423 	/* is there an upcall waiting for this flow ? */
1424 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1425 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash)
1426 	{
1427 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1428 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1429 		    !buf_ring_empty(rt->mfc_stall_ring))
1430 			break;
1431 	}
1432 
1433 	if (rt == NULL) {
1434 		int i;
1435 		struct igmpmsg *im;
1436 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1437 		struct mbuf *mm;
1438 
1439 		/*
1440 		 * Locate the vifi for the incoming interface for this packet.
1441 		 * If none found, drop packet.
1442 		 */
1443 		for (vifi = 0; vifi < V_numvifs &&
1444 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1445 			;
1446 		if (vifi >= V_numvifs) /* vif not found, drop packet */
1447 			goto non_fatal;
1448 
1449 		/* no upcall, so make a new entry */
1450 		rt = mfc_alloc();
1451 		if (rt == NULL)
1452 			goto fail;
1453 
1454 		/* Make a copy of the header to send to the user level process */
1455 		mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1456 		if (mm == NULL)
1457 			goto fail1;
1458 
1459 		/*
1460 		 * Send message to routing daemon to install
1461 		 * a route into the kernel table
1462 		 */
1463 
1464 		im = mtod(mm, struct igmpmsg*);
1465 		im->im_msgtype = IGMPMSG_NOCACHE;
1466 		im->im_mbz = 0;
1467 		im->im_vif = vifi;
1468 
1469 		MRTSTAT_INC(mrts_upcalls);
1470 
1471 		k_igmpsrc.sin_addr = ip->ip_src;
1472 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1473 			CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1474 			MRTSTAT_INC(mrts_upq_sockfull);
1475 			fail1: free(rt, M_MRTABLE);
1476 			fail: free(rte, M_MRTABLE);
1477 			m_freem(mb0);
1478 			MRW_WUNLOCK();
1479 			return ENOBUFS;
1480 		}
1481 
1482 		/* insert new entry at head of hash chain */
1483 		rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1484 		rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1485 		rt->mfc_expire = UPCALL_EXPIRE;
1486 		V_nexpire[hash]++;
1487 		for (i = 0; i < V_numvifs; i++) {
1488 			rt->mfc_ttls[i] = 0;
1489 			rt->mfc_flags[i] = 0;
1490 		}
1491 		rt->mfc_parent = -1;
1492 
1493 		/* clear the RP address */
1494 		rt->mfc_rp.s_addr = INADDR_ANY;
1495 		rt->mfc_bw_meter_leq = NULL;
1496 		rt->mfc_bw_meter_geq = NULL;
1497 
1498 		/* initialize pkt counters per src-grp */
1499 		rt->mfc_pkt_cnt = 0;
1500 		rt->mfc_byte_cnt = 0;
1501 		rt->mfc_wrong_if = 0;
1502 		timevalclear(&rt->mfc_last_assert);
1503 
1504 		buf_ring_enqueue(rt->mfc_stall_ring, rte);
1505 
1506 		/* Add RT to hashtable as it didn't exist before */
1507 		LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1508 	} else {
1509 		/* determine if queue has overflowed */
1510 		if (buf_ring_full(rt->mfc_stall_ring)) {
1511 			MRTSTAT_INC(mrts_upq_ovflw);
1512 			non_fatal: free(rte, M_MRTABLE);
1513 			m_freem(mb0);
1514 			MRW_WUNLOCK();
1515 			return (0);
1516 		}
1517 
1518 		buf_ring_enqueue(rt->mfc_stall_ring, rte);
1519 	}
1520 
1521 	rte->m = mb0;
1522 	rte->ifp = ifp;
1523 
1524 	MRW_WUNLOCK();
1525 
1526 	return 0;
1527 }
1528 
1529 /*
1530  * Clean up the cache entry if upcall is not serviced
1531  */
1532 static void
1533 expire_upcalls(void *arg)
1534 {
1535 	u_long i;
1536 
1537 	CURVNET_SET((struct vnet *) arg);
1538 
1539 	/*This callout is always run with MRW_WLOCK taken. */
1540 
1541 	for (i = 0; i < mfchashsize; i++) {
1542 		struct mfc *rt, *nrt;
1543 
1544 		if (V_nexpire[i] == 0)
1545 			continue;
1546 
1547 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1548 			if (buf_ring_empty(rt->mfc_stall_ring))
1549 				continue;
1550 
1551 			if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1552 				continue;
1553 
1554 			MRTSTAT_INC(mrts_cache_cleanups);
1555 			CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1556 			    (u_long)ntohl(rt->mfc_origin.s_addr),
1557 			    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1558 
1559 			expire_mfc(rt);
1560 		}
1561 	}
1562 
1563 	callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1564 	    curvnet);
1565 
1566 	CURVNET_RESTORE();
1567 }
1568 
1569 /*
1570  * Packet forwarding routine once entry in the cache is made
1571  */
1572 static int
1573 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1574 {
1575 	struct ip *ip = mtod(m, struct ip *);
1576 	vifi_t vifi;
1577 	int plen = ntohs(ip->ip_len);
1578 
1579 	M_ASSERTMAPPED(m);
1580 	MRW_LOCK_ASSERT();
1581 	NET_EPOCH_ASSERT();
1582 
1583 	/*
1584 	 * If xmt_vif is not -1, send on only the requested vif.
1585 	 *
1586 	 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1587 	 */
1588 	if (xmt_vif < V_numvifs) {
1589 		if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1590 			pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1591 		else
1592 			phyint_send(ip, V_viftable + xmt_vif, m);
1593 		return 1;
1594 	}
1595 
1596 	/*
1597 	 * Don't forward if it didn't arrive from the parent vif for its origin.
1598 	 */
1599 	vifi = rt->mfc_parent;
1600 	if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1601 		CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1602 				__func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1603 		MRTSTAT_INC(mrts_wrong_if);
1604 		++rt->mfc_wrong_if;
1605 		/*
1606 		 * If we are doing PIM assert processing, send a message
1607 		 * to the routing daemon.
1608 		 *
1609 		 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1610 		 * can complete the SPT switch, regardless of the type
1611 		 * of the iif (broadcast media, GRE tunnel, etc).
1612 		 */
1613 		if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1614 		    V_viftable[vifi].v_ifp) {
1615 			if (ifp == V_multicast_register_if)
1616 				PIMSTAT_INC(pims_rcv_registers_wrongiif);
1617 
1618 			/* Get vifi for the incoming packet */
1619 			for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp; vifi++)
1620 				;
1621 			if (vifi >= V_numvifs)
1622 				return 0;	/* The iif is not found: ignore the packet. */
1623 
1624 			if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1625 				return 0;	/* WRONGVIF disabled: ignore the packet */
1626 
1627 			if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1628 				struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1629 				struct igmpmsg *im;
1630 				int hlen = ip->ip_hl << 2;
1631 				struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1632 
1633 				if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1634 					mm = m_pullup(mm, hlen);
1635 				if (mm == NULL)
1636 					return ENOBUFS;
1637 
1638 				im = mtod(mm, struct igmpmsg *);
1639 				im->im_msgtype = IGMPMSG_WRONGVIF;
1640 				im->im_mbz = 0;
1641 				im->im_vif = vifi;
1642 
1643 				MRTSTAT_INC(mrts_upcalls);
1644 
1645 				k_igmpsrc.sin_addr = im->im_src;
1646 				if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1647 					CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1648 					MRTSTAT_INC(mrts_upq_sockfull);
1649 					return ENOBUFS;
1650 				}
1651 			}
1652 		}
1653 		return 0;
1654 	}
1655 
1656 	/* If I sourced this packet, it counts as output, else it was input. */
1657 	mtx_lock(&V_viftable[vifi].v_mtx);
1658 	if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1659 		V_viftable[vifi].v_pkt_out++;
1660 		V_viftable[vifi].v_bytes_out += plen;
1661 	} else {
1662 		V_viftable[vifi].v_pkt_in++;
1663 		V_viftable[vifi].v_bytes_in += plen;
1664 	}
1665 	mtx_unlock(&V_viftable[vifi].v_mtx);
1666 
1667 	rt->mfc_pkt_cnt++;
1668 	rt->mfc_byte_cnt += plen;
1669 
1670 	/*
1671 	 * For each vif, decide if a copy of the packet should be forwarded.
1672 	 * Forward if:
1673 	 *		- the ttl exceeds the vif's threshold
1674 	 *		- there are group members downstream on interface
1675 	 */
1676 	for (vifi = 0; vifi < V_numvifs; vifi++)
1677 		if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1678 			V_viftable[vifi].v_pkt_out++;
1679 			V_viftable[vifi].v_bytes_out += plen;
1680 			if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1681 				pim_register_send(ip, V_viftable + vifi, m, rt);
1682 			else
1683 				phyint_send(ip, V_viftable + vifi, m);
1684 		}
1685 
1686 	/*
1687 	 * Perform upcall-related bw measuring.
1688 	 */
1689 	if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1690 		struct bw_meter *x;
1691 		struct timeval now;
1692 
1693 		microtime(&now);
1694 		/* Process meters for Greater-or-EQual case */
1695 		for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1696 			bw_meter_geq_receive_packet(x, plen, &now);
1697 
1698 		/* Process meters for Lower-or-EQual case */
1699 		for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1700 			/*
1701 			 * Record that a packet is received.
1702 			 * A lock has to be taken as callout context
1703 			 * (expire_bw_meter_leq) might modify these fields
1704 			 * as well
1705 			 */
1706 			mtx_lock(&x->bm_mtx);
1707 			x->bm_measured.b_packets++;
1708 			x->bm_measured.b_bytes += plen;
1709 			mtx_unlock(&x->bm_mtx);
1710 		}
1711 	}
1712 
1713 	return 0;
1714 }
1715 
1716 /*
1717  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1718  */
1719 static int
1720 X_legal_vif_num(int vif)
1721 {
1722 	int ret;
1723 
1724 	ret = 0;
1725 	if (vif < 0)
1726 		return (ret);
1727 
1728 	MRW_RLOCK();
1729 	if (vif < V_numvifs)
1730 		ret = 1;
1731 	MRW_RUNLOCK();
1732 
1733 	return (ret);
1734 }
1735 
1736 /*
1737  * Return the local address used by this vif
1738  */
1739 static u_long
1740 X_ip_mcast_src(int vifi)
1741 {
1742 	in_addr_t addr;
1743 
1744 	addr = INADDR_ANY;
1745 	if (vifi < 0)
1746 		return (addr);
1747 
1748 	MRW_RLOCK();
1749 	if (vifi < V_numvifs)
1750 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1751 	MRW_RUNLOCK();
1752 
1753 	return (addr);
1754 }
1755 
1756 static void
1757 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1758 {
1759 	struct mbuf *mb_copy;
1760 	int hlen = ip->ip_hl << 2;
1761 
1762 	MRW_LOCK_ASSERT();
1763 	M_ASSERTMAPPED(m);
1764 
1765 	/*
1766 	 * Make a new reference to the packet; make sure that
1767 	 * the IP header is actually copied, not just referenced,
1768 	 * so that ip_output() only scribbles on the copy.
1769 	 */
1770 	mb_copy = m_copypacket(m, M_NOWAIT);
1771 	if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1772 		mb_copy = m_pullup(mb_copy, hlen);
1773 	if (mb_copy == NULL)
1774 		return;
1775 
1776 	send_packet(vifp, mb_copy);
1777 }
1778 
1779 static void
1780 send_packet(struct vif *vifp, struct mbuf *m)
1781 {
1782 	struct ip_moptions imo;
1783 	int error __unused;
1784 
1785 	MRW_LOCK_ASSERT();
1786 	NET_EPOCH_ASSERT();
1787 
1788 	imo.imo_multicast_ifp  = vifp->v_ifp;
1789 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1790 	imo.imo_multicast_loop = !!in_mcast_loop;
1791 	imo.imo_multicast_vif  = -1;
1792 	STAILQ_INIT(&imo.imo_head);
1793 
1794 	/*
1795 	 * Re-entrancy should not be a problem here, because
1796 	 * the packets that we send out and are looped back at us
1797 	 * should get rejected because they appear to come from
1798 	 * the loopback interface, thus preventing looping.
1799 	 */
1800 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1801 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1802 	    (ptrdiff_t)(vifp - V_viftable), error);
1803 }
1804 
1805 /*
1806  * Stubs for old RSVP socket shim implementation.
1807  */
1808 
1809 static int
1810 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1811 {
1812 
1813 	return (EOPNOTSUPP);
1814 }
1815 
1816 static void
1817 X_ip_rsvp_force_done(struct socket *so __unused)
1818 {
1819 
1820 }
1821 
1822 static int
1823 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1824 {
1825 	struct mbuf *m;
1826 
1827 	m = *mp;
1828 	*mp = NULL;
1829 	if (!V_rsvp_on)
1830 		m_freem(m);
1831 	return (IPPROTO_DONE);
1832 }
1833 
1834 /*
1835  * Code for bandwidth monitors
1836  */
1837 
1838 /*
1839  * Define common interface for timeval-related methods
1840  */
1841 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1842 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1843 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1844 
1845 static uint32_t
1846 compute_bw_meter_flags(struct bw_upcall *req)
1847 {
1848 	uint32_t flags = 0;
1849 
1850 	if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1851 		flags |= BW_METER_UNIT_PACKETS;
1852 	if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1853 		flags |= BW_METER_UNIT_BYTES;
1854 	if (req->bu_flags & BW_UPCALL_GEQ)
1855 		flags |= BW_METER_GEQ;
1856 	if (req->bu_flags & BW_UPCALL_LEQ)
1857 		flags |= BW_METER_LEQ;
1858 
1859 	return flags;
1860 }
1861 
1862 static void
1863 expire_bw_meter_leq(void *arg)
1864 {
1865 	struct bw_meter *x = arg;
1866 	struct timeval now;
1867 	/*
1868 	 * INFO:
1869 	 * callout is always executed with MRW_WLOCK taken
1870 	 */
1871 
1872 	CURVNET_SET((struct vnet *)x->arg);
1873 
1874 	microtime(&now);
1875 
1876 	/*
1877 	 * Test if we should deliver an upcall
1878 	 */
1879 	if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1880 	    (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1881 	    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1882 	    (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1883 		/* Prepare an upcall for delivery */
1884 		bw_meter_prepare_upcall(x, &now);
1885 	}
1886 
1887 	/* Send all upcalls that are pending delivery */
1888 	taskqueue_enqueue(V_task_queue, &V_task);
1889 
1890 	/* Reset counters */
1891 	x->bm_start_time = now;
1892 	/*
1893 	 * The lock has to be taken as ip_forward context
1894 	 * might modify these fields as well
1895 	 */
1896 	mtx_lock(&x->bm_mtx);
1897 	x->bm_measured.b_bytes = 0;
1898 	x->bm_measured.b_packets = 0;
1899 	mtx_unlock(&x->bm_mtx);
1900 
1901 	callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1902 
1903 	CURVNET_RESTORE();
1904 }
1905 
1906 /*
1907  * Add a bw_meter entry
1908  */
1909 static int
1910 add_bw_upcall(struct bw_upcall *req)
1911 {
1912 	struct mfc *mfc;
1913 	struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1914 	BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1915 	struct timeval now;
1916 	struct bw_meter *x, **bwm_ptr;
1917 	uint32_t flags;
1918 
1919 	if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1920 		return EOPNOTSUPP;
1921 
1922 	/* Test if the flags are valid */
1923 	if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1924 		return EINVAL;
1925 	if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1926 		return EINVAL;
1927 	if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1928 		return EINVAL;
1929 
1930 	/* Test if the threshold time interval is valid */
1931 	if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1932 		return EINVAL;
1933 
1934 	flags = compute_bw_meter_flags(req);
1935 
1936 	/*
1937 	 * Find if we have already same bw_meter entry
1938 	 */
1939 	MRW_WLOCK();
1940 	mfc = mfc_find(&req->bu_src, &req->bu_dst);
1941 	if (mfc == NULL) {
1942 		MRW_WUNLOCK();
1943 		return EADDRNOTAVAIL;
1944 	}
1945 
1946 	/* Choose an appropriate bw_meter list */
1947 	if (req->bu_flags & BW_UPCALL_GEQ)
1948 		bwm_ptr = &mfc->mfc_bw_meter_geq;
1949 	else
1950 		bwm_ptr = &mfc->mfc_bw_meter_leq;
1951 
1952 	for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
1953 		if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1954 		    &req->bu_threshold.b_time, ==))
1955 		    && (x->bm_threshold.b_packets
1956 		    == req->bu_threshold.b_packets)
1957 		    && (x->bm_threshold.b_bytes
1958 		    == req->bu_threshold.b_bytes)
1959 		    && (x->bm_flags & BW_METER_USER_FLAGS)
1960 		    == flags) {
1961 			MRW_WUNLOCK();
1962 			return 0; /* XXX Already installed */
1963 		}
1964 	}
1965 
1966 	/* Allocate the new bw_meter entry */
1967 	x = malloc(sizeof(*x), M_BWMETER, M_ZERO | M_NOWAIT);
1968 	if (x == NULL) {
1969 		MRW_WUNLOCK();
1970 		return ENOBUFS;
1971 	}
1972 
1973 	/* Set the new bw_meter entry */
1974 	x->bm_threshold.b_time = req->bu_threshold.b_time;
1975 	microtime(&now);
1976 	x->bm_start_time = now;
1977 	x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1978 	x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1979 	x->bm_measured.b_packets = 0;
1980 	x->bm_measured.b_bytes = 0;
1981 	x->bm_flags = flags;
1982 	x->bm_time_next = NULL;
1983 	x->bm_mfc = mfc;
1984 	x->arg = curvnet;
1985 	sprintf(x->bm_mtx_name, "BM mtx %p", x);
1986 	mtx_init(&x->bm_mtx, x->bm_mtx_name, NULL, MTX_DEF);
1987 
1988 	/* For LEQ case create periodic callout */
1989 	if (req->bu_flags & BW_UPCALL_LEQ) {
1990 		callout_init_rw(&x->bm_meter_callout, &mrouter_lock, CALLOUT_SHAREDLOCK);
1991 		callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
1992 		    expire_bw_meter_leq, x);
1993 	}
1994 
1995 	/* Add the new bw_meter entry to the front of entries for this MFC */
1996 	x->bm_mfc_next = *bwm_ptr;
1997 	*bwm_ptr = x;
1998 
1999 	MRW_WUNLOCK();
2000 
2001 	return 0;
2002 }
2003 
2004 static void
2005 free_bw_list(struct bw_meter *list)
2006 {
2007 	while (list != NULL) {
2008 		struct bw_meter *x = list;
2009 
2010 		/* MRW_WLOCK must be held here */
2011 		if (x->bm_flags & BW_METER_LEQ) {
2012 			callout_drain(&x->bm_meter_callout);
2013 			mtx_destroy(&x->bm_mtx);
2014 		}
2015 
2016 		list = list->bm_mfc_next;
2017 		free(x, M_BWMETER);
2018 	}
2019 }
2020 
2021 /*
2022  * Delete one or multiple bw_meter entries
2023  */
2024 static int
2025 del_bw_upcall(struct bw_upcall *req)
2026 {
2027 	struct mfc *mfc;
2028 	struct bw_meter *x, **bwm_ptr;
2029 
2030 	if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
2031 		return EOPNOTSUPP;
2032 
2033 	MRW_WLOCK();
2034 
2035 	/* Find the corresponding MFC entry */
2036 	mfc = mfc_find(&req->bu_src, &req->bu_dst);
2037 	if (mfc == NULL) {
2038 		MRW_WUNLOCK();
2039 		return EADDRNOTAVAIL;
2040 	} else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2041 		/*
2042 		 * Delete all bw_meter entries for this mfc
2043 		 */
2044 		struct bw_meter *list;
2045 
2046 		/* Free LEQ list */
2047 		list = mfc->mfc_bw_meter_leq;
2048 		mfc->mfc_bw_meter_leq = NULL;
2049 		free_bw_list(list);
2050 
2051 		/* Free GEQ list */
2052 		list = mfc->mfc_bw_meter_geq;
2053 		mfc->mfc_bw_meter_geq = NULL;
2054 		free_bw_list(list);
2055 		MRW_WUNLOCK();
2056 		return 0;
2057 	} else {			/* Delete a single bw_meter entry */
2058 		struct bw_meter *prev;
2059 		uint32_t flags = 0;
2060 
2061 		flags = compute_bw_meter_flags(req);
2062 
2063 		/* Choose an appropriate bw_meter list */
2064 		if (req->bu_flags & BW_UPCALL_GEQ)
2065 			bwm_ptr = &mfc->mfc_bw_meter_geq;
2066 		else
2067 			bwm_ptr = &mfc->mfc_bw_meter_leq;
2068 
2069 		/* Find the bw_meter entry to delete */
2070 		for (prev = NULL, x = *bwm_ptr; x != NULL;
2071 				prev = x, x = x->bm_mfc_next) {
2072 			if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, &req->bu_threshold.b_time, ==)) &&
2073 			    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2074 			    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2075 			    (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2076 				break;
2077 		}
2078 		if (x != NULL) { /* Delete entry from the list for this MFC */
2079 			if (prev != NULL)
2080 				prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2081 			else
2082 				*bwm_ptr = x->bm_mfc_next;/* new head of list */
2083 
2084 			if (req->bu_flags & BW_UPCALL_LEQ)
2085 				callout_stop(&x->bm_meter_callout);
2086 
2087 			MRW_WUNLOCK();
2088 			/* Free the bw_meter entry */
2089 			free(x, M_BWMETER);
2090 			return 0;
2091 		} else {
2092 			MRW_WUNLOCK();
2093 			return EINVAL;
2094 		}
2095 	}
2096 	__assert_unreachable();
2097 }
2098 
2099 /*
2100  * Perform bandwidth measurement processing that may result in an upcall
2101  */
2102 static void
2103 bw_meter_geq_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2104 {
2105 	struct timeval delta;
2106 
2107 	MRW_LOCK_ASSERT();
2108 
2109 	delta = *nowp;
2110 	BW_TIMEVALDECR(&delta, &x->bm_start_time);
2111 
2112 	/*
2113 	 * Processing for ">=" type of bw_meter entry.
2114 	 * bm_mtx does not have to be hold here as in GEQ
2115 	 * case this is the only context accessing bm_measured.
2116 	 */
2117 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2118 	    /* Reset the bw_meter entry */
2119 	    x->bm_start_time = *nowp;
2120 	    x->bm_measured.b_packets = 0;
2121 	    x->bm_measured.b_bytes = 0;
2122 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2123 	}
2124 
2125 	/* Record that a packet is received */
2126 	x->bm_measured.b_packets++;
2127 	x->bm_measured.b_bytes += plen;
2128 
2129 	/*
2130 	 * Test if we should deliver an upcall
2131 	 */
2132 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2133 		if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2134 		    (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2135 		    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2136 		    (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2137 			/* Prepare an upcall for delivery */
2138 			bw_meter_prepare_upcall(x, nowp);
2139 			x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2140 		}
2141 	}
2142 }
2143 
2144 /*
2145  * Prepare a bandwidth-related upcall
2146  */
2147 static void
2148 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2149 {
2150 	struct timeval delta;
2151 	struct bw_upcall *u;
2152 
2153 	MRW_LOCK_ASSERT();
2154 
2155 	/*
2156 	 * Compute the measured time interval
2157 	 */
2158 	delta = *nowp;
2159 	BW_TIMEVALDECR(&delta, &x->bm_start_time);
2160 
2161 	/*
2162 	 * Set the bw_upcall entry
2163 	 */
2164 	u = malloc(sizeof(struct bw_upcall), M_MRTABLE, M_NOWAIT | M_ZERO);
2165 	if (!u) {
2166 		log(LOG_WARNING, "bw_meter_prepare_upcall: cannot allocate entry\n");
2167 		return;
2168 	}
2169 	u->bu_src = x->bm_mfc->mfc_origin;
2170 	u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2171 	u->bu_threshold.b_time = x->bm_threshold.b_time;
2172 	u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2173 	u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2174 	u->bu_measured.b_time = delta;
2175 	u->bu_measured.b_packets = x->bm_measured.b_packets;
2176 	u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2177 	u->bu_flags = 0;
2178 	if (x->bm_flags & BW_METER_UNIT_PACKETS)
2179 		u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2180 	if (x->bm_flags & BW_METER_UNIT_BYTES)
2181 		u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2182 	if (x->bm_flags & BW_METER_GEQ)
2183 		u->bu_flags |= BW_UPCALL_GEQ;
2184 	if (x->bm_flags & BW_METER_LEQ)
2185 		u->bu_flags |= BW_UPCALL_LEQ;
2186 
2187 	if (buf_ring_enqueue(V_bw_upcalls_ring, u))
2188 		log(LOG_WARNING, "bw_meter_prepare_upcall: cannot enqueue upcall\n");
2189 	if (buf_ring_count(V_bw_upcalls_ring) > (BW_UPCALLS_MAX / 2)) {
2190 		taskqueue_enqueue(V_task_queue, &V_task);
2191 	}
2192 }
2193 /*
2194  * Send the pending bandwidth-related upcalls
2195  */
2196 static void
2197 bw_upcalls_send(void)
2198 {
2199 	struct mbuf *m;
2200 	int len = 0;
2201 	struct bw_upcall *bu;
2202 	struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2203 	static struct igmpmsg igmpmsg = {
2204 		0,		/* unused1 */
2205 		0,		/* unused2 */
2206 		IGMPMSG_BW_UPCALL,/* im_msgtype */
2207 		0,		/* im_mbz  */
2208 		0,		/* im_vif  */
2209 		0,		/* unused3 */
2210 		{ 0 },		/* im_src  */
2211 		{ 0 }		/* im_dst  */
2212 	};
2213 
2214 	MRW_LOCK_ASSERT();
2215 
2216 	if (buf_ring_empty(V_bw_upcalls_ring))
2217 		return;
2218 
2219 	/*
2220 	 * Allocate a new mbuf, initialize it with the header and
2221 	 * the payload for the pending calls.
2222 	 */
2223 	m = m_gethdr(M_NOWAIT, MT_DATA);
2224 	if (m == NULL) {
2225 		log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2226 		return;
2227 	}
2228 
2229 	m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2230 	len += sizeof(struct igmpmsg);
2231 	while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
2232 		m_copyback(m, len, sizeof(struct bw_upcall), (caddr_t)bu);
2233 		len += sizeof(struct bw_upcall);
2234 		free(bu, M_MRTABLE);
2235 	}
2236 
2237 	/*
2238 	 * Send the upcalls
2239 	 * XXX do we need to set the address in k_igmpsrc ?
2240 	 */
2241 	MRTSTAT_INC(mrts_upcalls);
2242 	if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2243 		log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2244 		MRTSTAT_INC(mrts_upq_sockfull);
2245 	}
2246 }
2247 
2248 /*
2249  * A periodic function for sending all upcalls that are pending delivery
2250  */
2251 static void
2252 expire_bw_upcalls_send(void *arg)
2253 {
2254 	CURVNET_SET((struct vnet *) arg);
2255 
2256 	/* This callout is run with MRW_RLOCK taken */
2257 
2258 	bw_upcalls_send();
2259 
2260 	callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2261 	    curvnet);
2262 	CURVNET_RESTORE();
2263 }
2264 
2265 /*
2266  * End of bandwidth monitoring code
2267  */
2268 
2269 /*
2270  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2271  *
2272  */
2273 static int
2274 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2275     struct mfc *rt)
2276 {
2277 	struct mbuf *mb_copy, *mm;
2278 
2279 	/*
2280 	 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2281 	 * rendezvous point was unspecified, and we were told not to.
2282 	 */
2283 	if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2284 	    in_nullhost(rt->mfc_rp))
2285 		return 0;
2286 
2287 	mb_copy = pim_register_prepare(ip, m);
2288 	if (mb_copy == NULL)
2289 		return ENOBUFS;
2290 
2291 	/*
2292 	 * Send all the fragments. Note that the mbuf for each fragment
2293 	 * is freed by the sending machinery.
2294 	 */
2295 	for (mm = mb_copy; mm; mm = mb_copy) {
2296 		mb_copy = mm->m_nextpkt;
2297 		mm->m_nextpkt = 0;
2298 		mm = m_pullup(mm, sizeof(struct ip));
2299 		if (mm != NULL) {
2300 			ip = mtod(mm, struct ip *);
2301 			if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2302 				pim_register_send_rp(ip, vifp, mm, rt);
2303 			} else {
2304 				pim_register_send_upcall(ip, vifp, mm, rt);
2305 			}
2306 		}
2307 	}
2308 
2309 	return 0;
2310 }
2311 
2312 /*
2313  * Return a copy of the data packet that is ready for PIM Register
2314  * encapsulation.
2315  * XXX: Note that in the returned copy the IP header is a valid one.
2316  */
2317 static struct mbuf *
2318 pim_register_prepare(struct ip *ip, struct mbuf *m)
2319 {
2320 	struct mbuf *mb_copy = NULL;
2321 	int mtu;
2322 
2323 	/* Take care of delayed checksums */
2324 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2325 		in_delayed_cksum(m);
2326 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2327 	}
2328 
2329 	/*
2330 	 * Copy the old packet & pullup its IP header into the
2331 	 * new mbuf so we can modify it.
2332 	 */
2333 	mb_copy = m_copypacket(m, M_NOWAIT);
2334 	if (mb_copy == NULL)
2335 		return NULL;
2336 	mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2337 	if (mb_copy == NULL)
2338 		return NULL;
2339 
2340 	/* take care of the TTL */
2341 	ip = mtod(mb_copy, struct ip *);
2342 	--ip->ip_ttl;
2343 
2344 	/* Compute the MTU after the PIM Register encapsulation */
2345 	mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2346 
2347 	if (ntohs(ip->ip_len) <= mtu) {
2348 		/* Turn the IP header into a valid one */
2349 		ip->ip_sum = 0;
2350 		ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2351 	} else {
2352 		/* Fragment the packet */
2353 		mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2354 		if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2355 			m_freem(mb_copy);
2356 			return NULL;
2357 		}
2358 	}
2359 	return mb_copy;
2360 }
2361 
2362 /*
2363  * Send an upcall with the data packet to the user-level process.
2364  */
2365 static int
2366 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2367     struct mbuf *mb_copy, struct mfc *rt)
2368 {
2369 	struct mbuf *mb_first;
2370 	int len = ntohs(ip->ip_len);
2371 	struct igmpmsg *im;
2372 	struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2373 
2374 	MRW_LOCK_ASSERT();
2375 
2376 	/*
2377 	 * Add a new mbuf with an upcall header
2378 	 */
2379 	mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2380 	if (mb_first == NULL) {
2381 		m_freem(mb_copy);
2382 		return ENOBUFS;
2383 	}
2384 	mb_first->m_data += max_linkhdr;
2385 	mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2386 	mb_first->m_len = sizeof(struct igmpmsg);
2387 	mb_first->m_next = mb_copy;
2388 
2389 	/* Send message to routing daemon */
2390 	im = mtod(mb_first, struct igmpmsg *);
2391 	im->im_msgtype	= IGMPMSG_WHOLEPKT;
2392 	im->im_mbz		= 0;
2393 	im->im_vif		= vifp - V_viftable;
2394 	im->im_src		= ip->ip_src;
2395 	im->im_dst		= ip->ip_dst;
2396 
2397 	k_igmpsrc.sin_addr	= ip->ip_src;
2398 
2399 	MRTSTAT_INC(mrts_upcalls);
2400 
2401 	if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2402 		CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2403 		MRTSTAT_INC(mrts_upq_sockfull);
2404 		return ENOBUFS;
2405 	}
2406 
2407 	/* Keep statistics */
2408 	PIMSTAT_INC(pims_snd_registers_msgs);
2409 	PIMSTAT_ADD(pims_snd_registers_bytes, len);
2410 
2411 	return 0;
2412 }
2413 
2414 /*
2415  * Encapsulate the data packet in PIM Register message and send it to the RP.
2416  */
2417 static int
2418 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2419     struct mfc *rt)
2420 {
2421 	struct mbuf *mb_first;
2422 	struct ip *ip_outer;
2423 	struct pim_encap_pimhdr *pimhdr;
2424 	int len = ntohs(ip->ip_len);
2425 	vifi_t vifi = rt->mfc_parent;
2426 
2427 	MRW_LOCK_ASSERT();
2428 
2429 	if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2430 		m_freem(mb_copy);
2431 		return EADDRNOTAVAIL;		/* The iif vif is invalid */
2432 	}
2433 
2434 	/*
2435 	 * Add a new mbuf with the encapsulating header
2436 	 */
2437 	mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2438 	if (mb_first == NULL) {
2439 		m_freem(mb_copy);
2440 		return ENOBUFS;
2441 	}
2442 	mb_first->m_data += max_linkhdr;
2443 	mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2444 	mb_first->m_next = mb_copy;
2445 
2446 	mb_first->m_pkthdr.len = len + mb_first->m_len;
2447 
2448 	/*
2449 	 * Fill in the encapsulating IP and PIM header
2450 	 */
2451 	ip_outer = mtod(mb_first, struct ip *);
2452 	*ip_outer = pim_encap_iphdr;
2453 	ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2454 			sizeof(pim_encap_pimhdr));
2455 	ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2456 	ip_outer->ip_dst = rt->mfc_rp;
2457 	/*
2458 	 * Copy the inner header TOS to the outer header, and take care of the
2459 	 * IP_DF bit.
2460 	 */
2461 	ip_outer->ip_tos = ip->ip_tos;
2462 	if (ip->ip_off & htons(IP_DF))
2463 		ip_outer->ip_off |= htons(IP_DF);
2464 	ip_fillid(ip_outer, V_ip_random_id);
2465 	pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2466 			+ sizeof(pim_encap_iphdr));
2467 	*pimhdr = pim_encap_pimhdr;
2468 	/* If the iif crosses a border, set the Border-bit */
2469 	if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2470 		pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2471 
2472 	mb_first->m_data += sizeof(pim_encap_iphdr);
2473 	pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2474 	mb_first->m_data -= sizeof(pim_encap_iphdr);
2475 
2476 	send_packet(vifp, mb_first);
2477 
2478 	/* Keep statistics */
2479 	PIMSTAT_INC(pims_snd_registers_msgs);
2480 	PIMSTAT_ADD(pims_snd_registers_bytes, len);
2481 
2482 	return 0;
2483 }
2484 
2485 /*
2486  * pim_encapcheck() is called by the encap4_input() path at runtime to
2487  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2488  * into the kernel.
2489  */
2490 static int
2491 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2492     int proto __unused, void *arg __unused)
2493 {
2494 
2495 	KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2496 	return (8);		/* claim the datagram. */
2497 }
2498 
2499 /*
2500  * PIM-SMv2 and PIM-DM messages processing.
2501  * Receives and verifies the PIM control messages, and passes them
2502  * up to the listening socket, using rip_input().
2503  * The only message with special processing is the PIM_REGISTER message
2504  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2505  * is passed to if_simloop().
2506  */
2507 static int
2508 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2509 {
2510 	struct ip *ip = mtod(m, struct ip *);
2511 	struct pim *pim;
2512 	int iphlen = off;
2513 	int minlen;
2514 	int datalen = ntohs(ip->ip_len) - iphlen;
2515 	int ip_tos;
2516 
2517 	/* Keep statistics */
2518 	PIMSTAT_INC(pims_rcv_total_msgs);
2519 	PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2520 
2521 	/*
2522 	 * Validate lengths
2523 	 */
2524 	if (datalen < PIM_MINLEN) {
2525 		PIMSTAT_INC(pims_rcv_tooshort);
2526 		CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2527 		    __func__, datalen, ntohl(ip->ip_src.s_addr));
2528 		m_freem(m);
2529 		return (IPPROTO_DONE);
2530 	}
2531 
2532 	/*
2533 	 * If the packet is at least as big as a REGISTER, go agead
2534 	 * and grab the PIM REGISTER header size, to avoid another
2535 	 * possible m_pullup() later.
2536 	 *
2537 	 * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2538 	 * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2539 	 */
2540 	minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2541 	/*
2542 	 * Get the IP and PIM headers in contiguous memory, and
2543 	 * possibly the PIM REGISTER header.
2544 	 */
2545 	if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2546 		CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2547 		return (IPPROTO_DONE);
2548 	}
2549 
2550 	/* m_pullup() may have given us a new mbuf so reset ip. */
2551 	ip = mtod(m, struct ip *);
2552 	ip_tos = ip->ip_tos;
2553 
2554 	/* adjust mbuf to point to the PIM header */
2555 	m->m_data += iphlen;
2556 	m->m_len  -= iphlen;
2557 	pim = mtod(m, struct pim *);
2558 
2559 	/*
2560 	 * Validate checksum. If PIM REGISTER, exclude the data packet.
2561 	 *
2562 	 * XXX: some older PIMv2 implementations don't make this distinction,
2563 	 * so for compatibility reason perform the checksum over part of the
2564 	 * message, and if error, then over the whole message.
2565 	 */
2566 	if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2567 		/* do nothing, checksum okay */
2568 	} else if (in_cksum(m, datalen)) {
2569 		PIMSTAT_INC(pims_rcv_badsum);
2570 		CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2571 		m_freem(m);
2572 		return (IPPROTO_DONE);
2573 	}
2574 
2575 	/* PIM version check */
2576 	if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2577 		PIMSTAT_INC(pims_rcv_badversion);
2578 		CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2579 		    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2580 		m_freem(m);
2581 		return (IPPROTO_DONE);
2582 	}
2583 
2584 	/* restore mbuf back to the outer IP */
2585 	m->m_data -= iphlen;
2586 	m->m_len  += iphlen;
2587 
2588 	if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2589 		/*
2590 		 * Since this is a REGISTER, we'll make a copy of the register
2591 		 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2592 		 * routing daemon.
2593 		 */
2594 		struct sockaddr_in dst = { sizeof(dst), AF_INET };
2595 		struct mbuf *mcp;
2596 		struct ip *encap_ip;
2597 		u_int32_t *reghdr;
2598 		struct ifnet *vifp;
2599 
2600 		MRW_RLOCK();
2601 		if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2602 			MRW_RUNLOCK();
2603 			CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2604 			    (int)V_reg_vif_num);
2605 			m_freem(m);
2606 			return (IPPROTO_DONE);
2607 		}
2608 		/* XXX need refcnt? */
2609 		vifp = V_viftable[V_reg_vif_num].v_ifp;
2610 		MRW_RUNLOCK();
2611 
2612 		/*
2613 		 * Validate length
2614 		 */
2615 		if (datalen < PIM_REG_MINLEN) {
2616 			PIMSTAT_INC(pims_rcv_tooshort);
2617 			PIMSTAT_INC(pims_rcv_badregisters);
2618 			CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2619 			m_freem(m);
2620 			return (IPPROTO_DONE);
2621 		}
2622 
2623 		reghdr = (u_int32_t *)(pim + 1);
2624 		encap_ip = (struct ip *)(reghdr + 1);
2625 
2626 		CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2627 		    __func__, ntohl(encap_ip->ip_src.s_addr),
2628 		    ntohs(encap_ip->ip_len));
2629 
2630 		/* verify the version number of the inner packet */
2631 		if (encap_ip->ip_v != IPVERSION) {
2632 			PIMSTAT_INC(pims_rcv_badregisters);
2633 			CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2634 			m_freem(m);
2635 			return (IPPROTO_DONE);
2636 		}
2637 
2638 		/* verify the inner packet is destined to a mcast group */
2639 		if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2640 			PIMSTAT_INC(pims_rcv_badregisters);
2641 			CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2642 			    ntohl(encap_ip->ip_dst.s_addr));
2643 			m_freem(m);
2644 			return (IPPROTO_DONE);
2645 		}
2646 
2647 		/* If a NULL_REGISTER, pass it to the daemon */
2648 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2649 			goto pim_input_to_daemon;
2650 
2651 		/*
2652 		 * Copy the TOS from the outer IP header to the inner IP header.
2653 		 */
2654 		if (encap_ip->ip_tos != ip_tos) {
2655 			/* Outer TOS -> inner TOS */
2656 			encap_ip->ip_tos = ip_tos;
2657 			/* Recompute the inner header checksum. Sigh... */
2658 
2659 			/* adjust mbuf to point to the inner IP header */
2660 			m->m_data += (iphlen + PIM_MINLEN);
2661 			m->m_len  -= (iphlen + PIM_MINLEN);
2662 
2663 			encap_ip->ip_sum = 0;
2664 			encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2665 
2666 			/* restore mbuf to point back to the outer IP header */
2667 			m->m_data -= (iphlen + PIM_MINLEN);
2668 			m->m_len  += (iphlen + PIM_MINLEN);
2669 		}
2670 
2671 		/*
2672 		 * Decapsulate the inner IP packet and loopback to forward it
2673 		 * as a normal multicast packet. Also, make a copy of the
2674 		 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2675 		 * to pass to the daemon later, so it can take the appropriate
2676 		 * actions (e.g., send back PIM_REGISTER_STOP).
2677 		 * XXX: here m->m_data points to the outer IP header.
2678 		 */
2679 		mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2680 		if (mcp == NULL) {
2681 			CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2682 			m_freem(m);
2683 			return (IPPROTO_DONE);
2684 		}
2685 
2686 		/* Keep statistics */
2687 		/* XXX: registers_bytes include only the encap. mcast pkt */
2688 		PIMSTAT_INC(pims_rcv_registers_msgs);
2689 		PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2690 
2691 		/*
2692 		 * forward the inner ip packet; point m_data at the inner ip.
2693 		 */
2694 		m_adj(m, iphlen + PIM_MINLEN);
2695 
2696 		CTR4(KTR_IPMF,
2697 		    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2698 		    __func__,
2699 		    (u_long)ntohl(encap_ip->ip_src.s_addr),
2700 		    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2701 		    (int)V_reg_vif_num);
2702 
2703 		/* NB: vifp was collected above; can it change on us? */
2704 		if_simloop(vifp, m, dst.sin_family, 0);
2705 
2706 		/* prepare the register head to send to the mrouting daemon */
2707 		m = mcp;
2708 	}
2709 
2710 pim_input_to_daemon:
2711 	/*
2712 	 * Pass the PIM message up to the daemon; if it is a Register message,
2713 	 * pass the 'head' only up to the daemon. This includes the
2714 	 * outer IP header, PIM header, PIM-Register header and the
2715 	 * inner IP header.
2716 	 * XXX: the outer IP header pkt size of a Register is not adjust to
2717 	 * reflect the fact that the inner multicast data is truncated.
2718 	 */
2719 	return (rip_input(&m, &off, proto));
2720 }
2721 
2722 static int
2723 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2724 {
2725 	struct mfc	*rt;
2726 	int		 error, i;
2727 
2728 	if (req->newptr)
2729 		return (EPERM);
2730 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2731 		return (0);
2732 	error = sysctl_wire_old_buffer(req, 0);
2733 	if (error)
2734 		return (error);
2735 
2736 	MRW_RLOCK();
2737 	if (V_mfchashtbl == NULL)
2738 		goto out_locked;
2739 
2740 	for (i = 0; i < mfchashsize; i++) {
2741 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2742 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2743 			if (error)
2744 				goto out_locked;
2745 		}
2746 	}
2747 out_locked:
2748 	MRW_RUNLOCK();
2749 	return (error);
2750 }
2751 
2752 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2753     CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2754     "IPv4 Multicast Forwarding Table "
2755     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2756 
2757 static int
2758 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2759 {
2760 	int error, i;
2761 
2762 	if (req->newptr)
2763 		return (EPERM);
2764 	if (V_viftable == NULL)		/* XXX unlocked */
2765 		return (0);
2766 	error = sysctl_wire_old_buffer(req, MROUTE_VIF_SYSCTL_LEN * MAXVIFS);
2767 	if (error)
2768 		return (error);
2769 
2770 	MRW_RLOCK();
2771 	/* Copy out user-visible portion of vif entry. */
2772 	for (i = 0; i < MAXVIFS; i++) {
2773 		error = SYSCTL_OUT(req, &V_viftable[i], MROUTE_VIF_SYSCTL_LEN);
2774 		if (error)
2775 			break;
2776 	}
2777 	MRW_RUNLOCK();
2778 	return (error);
2779 }
2780 
2781 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2782     CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2783     sysctl_viflist, "S,vif[MAXVIFS]",
2784     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2785 
2786 static void
2787 vnet_mroute_init(const void *unused __unused)
2788 {
2789 
2790 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2791 
2792 	V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2793 	    M_MRTABLE, M_WAITOK|M_ZERO);
2794 
2795 	callout_init_rw(&V_expire_upcalls_ch, &mrouter_lock, 0);
2796 	callout_init_rw(&V_bw_upcalls_ch, &mrouter_lock, 0);
2797 
2798 	/* Prepare taskqueue */
2799 	V_task_queue = taskqueue_create_fast("ip_mroute_tskq", M_NOWAIT,
2800 		    taskqueue_thread_enqueue, &V_task_queue);
2801 	taskqueue_start_threads(&V_task_queue, 1, PI_NET, "ip_mroute_tskq task");
2802 }
2803 
2804 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2805 	NULL);
2806 
2807 static void
2808 vnet_mroute_uninit(const void *unused __unused)
2809 {
2810 
2811 	/* Taskqueue should be cancelled and drained before freeing */
2812 	taskqueue_free(V_task_queue);
2813 
2814 	free(V_viftable, M_MRTABLE);
2815 	free(V_nexpire, M_MRTABLE);
2816 	V_nexpire = NULL;
2817 }
2818 
2819 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2820     vnet_mroute_uninit, NULL);
2821 
2822 static int
2823 ip_mroute_modevent(module_t mod, int type, void *unused)
2824 {
2825 
2826 	switch (type) {
2827 	case MOD_LOAD:
2828 		MRW_TEARDOWN_LOCK_INIT();
2829 		MRW_LOCK_INIT();
2830 
2831 		if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2832 		    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2833 
2834 		if (!powerof2(mfchashsize)) {
2835 			printf("WARNING: %s not a power of 2; using default\n",
2836 					"net.inet.ip.mfchashsize");
2837 			mfchashsize = MFCHASHSIZE;
2838 		}
2839 
2840 		pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2841 
2842 		ip_mcast_src = X_ip_mcast_src;
2843 		ip_mforward = X_ip_mforward;
2844 		ip_mrouter_done = X_ip_mrouter_done;
2845 		ip_mrouter_get = X_ip_mrouter_get;
2846 		ip_mrouter_set = X_ip_mrouter_set;
2847 
2848 		ip_rsvp_force_done = X_ip_rsvp_force_done;
2849 		ip_rsvp_vif = X_ip_rsvp_vif;
2850 
2851 		legal_vif_num = X_legal_vif_num;
2852 		mrt_ioctl = X_mrt_ioctl;
2853 		rsvp_input_p = X_rsvp_input;
2854 		break;
2855 
2856 	case MOD_UNLOAD:
2857 		/*
2858 		 * Typically module unload happens after the user-level
2859 		 * process has shutdown the kernel services (the check
2860 		 * below insures someone can't just yank the module out
2861 		 * from under a running process).  But if the module is
2862 		 * just loaded and then unloaded w/o starting up a user
2863 		 * process we still need to cleanup.
2864 		 */
2865 		MRW_WLOCK();
2866 		if (ip_mrouter_cnt != 0) {
2867 			MRW_WUNLOCK();
2868 			return (EINVAL);
2869 		}
2870 		ip_mrouter_unloading = 1;
2871 		MRW_WUNLOCK();
2872 
2873 		EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2874 
2875 		if (pim_encap_cookie) {
2876 			ip_encap_detach(pim_encap_cookie);
2877 			pim_encap_cookie = NULL;
2878 		}
2879 
2880 		ip_mcast_src = NULL;
2881 		ip_mforward = NULL;
2882 		ip_mrouter_done = NULL;
2883 		ip_mrouter_get = NULL;
2884 		ip_mrouter_set = NULL;
2885 
2886 		ip_rsvp_force_done = NULL;
2887 		ip_rsvp_vif = NULL;
2888 
2889 		legal_vif_num = NULL;
2890 		mrt_ioctl = NULL;
2891 		rsvp_input_p = NULL;
2892 
2893 		MRW_LOCK_DESTROY();
2894 		MRW_TEARDOWN_LOCK_DESTROY();
2895 		break;
2896 
2897 	default:
2898 		return EOPNOTSUPP;
2899 	}
2900 	return 0;
2901 }
2902 
2903 static moduledata_t ip_mroutemod = {
2904 	"ip_mroute",
2905 	ip_mroute_modevent,
2906 	0
2907 };
2908 
2909 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2910