xref: /freebsd/sys/netinet/ip_mroute.c (revision af23369a6deaaeb612ab266eb88b8bb8d560c322)
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_private.h>
110 #include <net/if_types.h>
111 #include <net/netisr.h>
112 #include <net/route.h>
113 #include <net/vnet.h>
114 
115 #include <netinet/in.h>
116 #include <netinet/igmp.h>
117 #include <netinet/in_systm.h>
118 #include <netinet/in_var.h>
119 #include <netinet/ip.h>
120 #include <netinet/ip_encap.h>
121 #include <netinet/ip_mroute.h>
122 #include <netinet/ip_var.h>
123 #include <netinet/ip_options.h>
124 #include <netinet/pim.h>
125 #include <netinet/pim_var.h>
126 #include <netinet/udp.h>
127 
128 #include <machine/in_cksum.h>
129 
130 #ifndef KTR_IPMF
131 #define KTR_IPMF KTR_INET
132 #endif
133 
134 #define		VIFI_INVALID	((vifi_t) -1)
135 
136 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
137 
138 /*
139  * Locking.  We use two locks: one for the virtual interface table and
140  * one for the forwarding table.  These locks may be nested in which case
141  * the VIF lock must always be taken first.  Note that each lock is used
142  * to cover not only the specific data structure but also related data
143  * structures.
144  */
145 
146 static struct rwlock mrouter_lock;
147 #define	MRW_RLOCK()		rw_rlock(&mrouter_lock)
148 #define	MRW_WLOCK()		rw_wlock(&mrouter_lock)
149 #define	MRW_RUNLOCK()		rw_runlock(&mrouter_lock)
150 #define	MRW_WUNLOCK()		rw_wunlock(&mrouter_lock)
151 #define	MRW_UNLOCK()		rw_unlock(&mrouter_lock)
152 #define	MRW_LOCK_ASSERT()	rw_assert(&mrouter_lock, RA_LOCKED)
153 #define	MRW_WLOCK_ASSERT()	rw_assert(&mrouter_lock, RA_WLOCKED)
154 #define	MRW_LOCK_TRY_UPGRADE()	rw_try_upgrade(&mrouter_lock)
155 #define	MRW_WOWNED()		rw_wowned(&mrouter_lock)
156 #define	MRW_LOCK_INIT()						\
157 	rw_init(&mrouter_lock, "IPv4 multicast forwarding")
158 #define	MRW_LOCK_DESTROY()	rw_destroy(&mrouter_lock)
159 
160 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
161 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
162 
163 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
164 VNET_PCPUSTAT_SYSINIT(mrtstat);
165 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
166 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
167     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
168     "netinet/ip_mroute.h)");
169 
170 VNET_DEFINE_STATIC(u_long, mfchash);
171 #define	V_mfchash		VNET(mfchash)
172 #define	MFCHASH(a, g)							\
173 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
174 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
175 #define	MFCHASHSIZE	256
176 
177 static u_long mfchashsize;			/* Hash size */
178 VNET_DEFINE_STATIC(u_char *, nexpire);		/* 0..mfchashsize-1 */
179 #define	V_nexpire		VNET(nexpire)
180 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
181 #define	V_mfchashtbl		VNET(mfchashtbl)
182 VNET_DEFINE_STATIC(struct taskqueue *, task_queue);
183 #define	V_task_queue		VNET(task_queue)
184 VNET_DEFINE_STATIC(struct task, task);
185 #define	V_task		VNET(task)
186 
187 VNET_DEFINE_STATIC(vifi_t, numvifs);
188 #define	V_numvifs		VNET(numvifs)
189 VNET_DEFINE_STATIC(struct vif *, viftable);
190 #define	V_viftable		VNET(viftable)
191 
192 static eventhandler_tag if_detach_event_tag = NULL;
193 
194 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
195 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
196 
197 VNET_DEFINE_STATIC(struct mtx, buf_ring_mtx);
198 #define	V_buf_ring_mtx	VNET(buf_ring_mtx)
199 
200 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
201 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
202 
203 /*
204  * Bandwidth meter variables and constants
205  */
206 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
207 
208 /*
209  * Pending upcalls are stored in a ring which is flushed when
210  * full, or periodically
211  */
212 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
213 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
214 VNET_DEFINE_STATIC(struct buf_ring *, bw_upcalls_ring);
215 #define	V_bw_upcalls_ring    	VNET(bw_upcalls_ring)
216 VNET_DEFINE_STATIC(struct mtx, bw_upcalls_ring_mtx);
217 #define	V_bw_upcalls_ring_mtx    	VNET(bw_upcalls_ring_mtx)
218 
219 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
220 
221 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
222 VNET_PCPUSTAT_SYSINIT(pimstat);
223 VNET_PCPUSTAT_SYSUNINIT(pimstat);
224 
225 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
226     "PIM");
227 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
228     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
229 
230 static u_long	pim_squelch_wholepkt = 0;
231 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
232     &pim_squelch_wholepkt, 0,
233     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
234 
235 static const struct encaptab *pim_encap_cookie;
236 static int pim_encapcheck(const struct mbuf *, int, int, void *);
237 static int pim_input(struct mbuf *, int, int, void *);
238 
239 extern int in_mcast_loop;
240 
241 static const struct encap_config ipv4_encap_cfg = {
242 	.proto = IPPROTO_PIM,
243 	.min_length = sizeof(struct ip) + PIM_MINLEN,
244 	.exact_match = 8,
245 	.check = pim_encapcheck,
246 	.input = pim_input
247 };
248 
249 /*
250  * Note: the PIM Register encapsulation adds the following in front of a
251  * data packet:
252  *
253  * struct pim_encap_hdr {
254  *    struct ip ip;
255  *    struct pim_encap_pimhdr  pim;
256  * }
257  *
258  */
259 
260 struct pim_encap_pimhdr {
261 	struct pim pim;
262 	uint32_t   flags;
263 };
264 #define		PIM_ENCAP_TTL	64
265 
266 static struct ip pim_encap_iphdr = {
267 #if BYTE_ORDER == LITTLE_ENDIAN
268 	sizeof(struct ip) >> 2,
269 	IPVERSION,
270 #else
271 	IPVERSION,
272 	sizeof(struct ip) >> 2,
273 #endif
274 	0,			/* tos */
275 	sizeof(struct ip),	/* total length */
276 	0,			/* id */
277 	0,			/* frag offset */
278 	PIM_ENCAP_TTL,
279 	IPPROTO_PIM,
280 	0,			/* checksum */
281 };
282 
283 static struct pim_encap_pimhdr pim_encap_pimhdr = {
284     {
285 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
286 	0,			/* reserved */
287 	0,			/* checksum */
288     },
289     0				/* flags */
290 };
291 
292 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
293 #define	V_reg_vif_num		VNET(reg_vif_num)
294 VNET_DEFINE_STATIC(struct ifnet *, multicast_register_if);
295 #define	V_multicast_register_if	VNET(multicast_register_if)
296 
297 /*
298  * Private variables.
299  */
300 
301 static u_long	X_ip_mcast_src(int);
302 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
303 		    struct ip_moptions *);
304 static int	X_ip_mrouter_done(void);
305 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
306 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
307 static int	X_legal_vif_num(int);
308 static int	X_mrt_ioctl(u_long, caddr_t, int);
309 
310 static int	add_bw_upcall(struct bw_upcall *);
311 static int	add_mfc(struct mfcctl2 *);
312 static int	add_vif(struct vifctl *);
313 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
314 static void	bw_meter_geq_receive_packet(struct bw_meter *, int,
315 		    struct timeval *);
316 static void	bw_upcalls_send(void);
317 static int	del_bw_upcall(struct bw_upcall *);
318 static int	del_mfc(struct mfcctl2 *);
319 static int	del_vif(vifi_t);
320 static int	del_vif_locked(vifi_t, struct ifnet **);
321 static void	expire_bw_upcalls_send(void *);
322 static void	expire_mfc(struct mfc *);
323 static void	expire_upcalls(void *);
324 static void	free_bw_list(struct bw_meter *);
325 static int	get_sg_cnt(struct sioc_sg_req *);
326 static int	get_vif_cnt(struct sioc_vif_req *);
327 static void	if_detached_event(void *, struct ifnet *);
328 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
329 static int	ip_mrouter_init(struct socket *, int);
330 static __inline struct mfc *
331 		mfc_find(struct in_addr *, struct in_addr *);
332 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
333 static struct mbuf *
334 		pim_register_prepare(struct ip *, struct mbuf *);
335 static int	pim_register_send(struct ip *, struct vif *,
336 		    struct mbuf *, struct mfc *);
337 static int	pim_register_send_rp(struct ip *, struct vif *,
338 		    struct mbuf *, struct mfc *);
339 static int	pim_register_send_upcall(struct ip *, struct vif *,
340 		    struct mbuf *, struct mfc *);
341 static void	send_packet(struct vif *, struct mbuf *);
342 static int	set_api_config(uint32_t *);
343 static int	set_assert(int);
344 static int	socket_send(struct socket *, struct mbuf *,
345 		    struct sockaddr_in *);
346 
347 /*
348  * Kernel multicast forwarding API capabilities and setup.
349  * If more API capabilities are added to the kernel, they should be
350  * recorded in `mrt_api_support'.
351  */
352 #define MRT_API_VERSION		0x0305
353 
354 static const int mrt_api_version = MRT_API_VERSION;
355 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
356 					 MRT_MFC_FLAGS_BORDER_VIF |
357 					 MRT_MFC_RP |
358 					 MRT_MFC_BW_UPCALL);
359 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
360 #define	V_mrt_api_config	VNET(mrt_api_config)
361 VNET_DEFINE_STATIC(int, pim_assert_enabled);
362 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
363 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
364 
365 /*
366  * Find a route for a given origin IP address and multicast group address.
367  * Statistics must be updated by the caller.
368  */
369 static __inline struct mfc *
370 mfc_find(struct in_addr *o, struct in_addr *g)
371 {
372 	struct mfc *rt;
373 
374 	/*
375 	 * Might be called both RLOCK and WLOCK.
376 	 * Check if any, it's caller responsibility
377 	 * to choose correct option.
378 	 */
379 	MRW_LOCK_ASSERT();
380 
381 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
382 		if (in_hosteq(rt->mfc_origin, *o) &&
383 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
384 		    buf_ring_empty(rt->mfc_stall_ring))
385 			break;
386 	}
387 
388 	return (rt);
389 }
390 
391 static __inline struct mfc *
392 mfc_alloc(void)
393 {
394 	struct mfc *rt;
395 	rt = malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT | M_ZERO);
396 	if (rt == NULL)
397 		return rt;
398 
399 	rt->mfc_stall_ring = buf_ring_alloc(MAX_UPQ, M_MRTABLE,
400 	    M_NOWAIT, &V_buf_ring_mtx);
401 	if (rt->mfc_stall_ring == NULL) {
402 		free(rt, M_MRTABLE);
403 		return NULL;
404 	}
405 
406 	return rt;
407 }
408 
409 /*
410  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
411  */
412 static int
413 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
414 {
415 	int error, optval;
416 	vifi_t vifi;
417 	struct vifctl vifc;
418 	struct mfcctl2 mfc;
419 	struct bw_upcall bw_upcall;
420 	uint32_t i;
421 
422 	if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
423 		return EPERM;
424 
425 	error = 0;
426 	switch (sopt->sopt_name) {
427 	case MRT_INIT:
428 		error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
429 		if (error)
430 			break;
431 		error = ip_mrouter_init(so, optval);
432 		break;
433 	case MRT_DONE:
434 		error = ip_mrouter_done();
435 		break;
436 	case MRT_ADD_VIF:
437 		error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
438 		if (error)
439 			break;
440 		error = add_vif(&vifc);
441 		break;
442 	case MRT_DEL_VIF:
443 		error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
444 		if (error)
445 			break;
446 		error = del_vif(vifi);
447 		break;
448 	case MRT_ADD_MFC:
449 	case MRT_DEL_MFC:
450 		/*
451 		 * select data size depending on API version.
452 		 */
453 		if (sopt->sopt_name == MRT_ADD_MFC &&
454 		    V_mrt_api_config & MRT_API_FLAGS_ALL) {
455 			error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
456 			    sizeof(struct mfcctl2));
457 		} else {
458 			error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
459 			    sizeof(struct mfcctl));
460 			bzero((caddr_t)&mfc + sizeof(struct mfcctl),
461 			    sizeof(mfc) - sizeof(struct mfcctl));
462 		}
463 		if (error)
464 			break;
465 		if (sopt->sopt_name == MRT_ADD_MFC)
466 			error = add_mfc(&mfc);
467 		else
468 			error = del_mfc(&mfc);
469 		break;
470 
471 	case MRT_ASSERT:
472 		error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
473 		if (error)
474 			break;
475 		set_assert(optval);
476 		break;
477 
478 	case MRT_API_CONFIG:
479 		error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
480 		if (!error)
481 			error = set_api_config(&i);
482 		if (!error)
483 			error = sooptcopyout(sopt, &i, sizeof i);
484 		break;
485 
486 	case MRT_ADD_BW_UPCALL:
487 	case MRT_DEL_BW_UPCALL:
488 		error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
489 		    sizeof bw_upcall);
490 		if (error)
491 			break;
492 		if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
493 			error = add_bw_upcall(&bw_upcall);
494 		else
495 			error = del_bw_upcall(&bw_upcall);
496 		break;
497 
498 	default:
499 		error = EOPNOTSUPP;
500 		break;
501 	}
502 	return error;
503 }
504 
505 /*
506  * Handle MRT getsockopt commands
507  */
508 static int
509 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
510 {
511 	int error;
512 
513 	switch (sopt->sopt_name) {
514 	case MRT_VERSION:
515 		error = sooptcopyout(sopt, &mrt_api_version,
516 		    sizeof mrt_api_version);
517 		break;
518 	case MRT_ASSERT:
519 		error = sooptcopyout(sopt, &V_pim_assert_enabled,
520 		    sizeof V_pim_assert_enabled);
521 		break;
522 	case MRT_API_SUPPORT:
523 		error = sooptcopyout(sopt, &mrt_api_support,
524 		    sizeof mrt_api_support);
525 		break;
526 	case MRT_API_CONFIG:
527 		error = sooptcopyout(sopt, &V_mrt_api_config,
528 		    sizeof V_mrt_api_config);
529 		break;
530 	default:
531 		error = EOPNOTSUPP;
532 		break;
533 	}
534 	return error;
535 }
536 
537 /*
538  * Handle ioctl commands to obtain information from the cache
539  */
540 static int
541 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
542 {
543 	int error;
544 
545 	/*
546 	 * Currently the only function calling this ioctl routine is rtioctl_fib().
547 	 * Typically, only root can create the raw socket in order to execute
548 	 * this ioctl method, however the request might be coming from a prison
549 	 */
550 	error = priv_check(curthread, PRIV_NETINET_MROUTE);
551 	if (error)
552 		return (error);
553 	switch (cmd) {
554 	case (SIOCGETVIFCNT):
555 		error = get_vif_cnt((struct sioc_vif_req *)data);
556 		break;
557 
558 	case (SIOCGETSGCNT):
559 		error = get_sg_cnt((struct sioc_sg_req *)data);
560 		break;
561 
562 	default:
563 		error = EINVAL;
564 		break;
565 	}
566 	return error;
567 }
568 
569 /*
570  * returns the packet, byte, rpf-failure count for the source group provided
571  */
572 static int
573 get_sg_cnt(struct sioc_sg_req *req)
574 {
575 	struct mfc *rt;
576 
577 	MRW_RLOCK();
578 	rt = mfc_find(&req->src, &req->grp);
579 	if (rt == NULL) {
580 		MRW_RUNLOCK();
581 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
582 		return EADDRNOTAVAIL;
583 	}
584 	req->pktcnt = rt->mfc_pkt_cnt;
585 	req->bytecnt = rt->mfc_byte_cnt;
586 	req->wrong_if = rt->mfc_wrong_if;
587 	MRW_RUNLOCK();
588 	return 0;
589 }
590 
591 /*
592  * returns the input and output packet and byte counts on the vif provided
593  */
594 static int
595 get_vif_cnt(struct sioc_vif_req *req)
596 {
597 	vifi_t vifi = req->vifi;
598 
599 	MRW_RLOCK();
600 	if (vifi >= V_numvifs) {
601 		MRW_RUNLOCK();
602 		return EINVAL;
603 	}
604 
605 	mtx_lock_spin(&V_viftable[vifi].v_spin);
606 	req->icount = V_viftable[vifi].v_pkt_in;
607 	req->ocount = V_viftable[vifi].v_pkt_out;
608 	req->ibytes = V_viftable[vifi].v_bytes_in;
609 	req->obytes = V_viftable[vifi].v_bytes_out;
610 	mtx_unlock_spin(&V_viftable[vifi].v_spin);
611 	MRW_RUNLOCK();
612 
613 	return 0;
614 }
615 
616 static void
617 if_detached_event(void *arg __unused, struct ifnet *ifp)
618 {
619 	vifi_t vifi;
620 	u_long i, vifi_cnt = 0;
621 	struct ifnet *free_ptr;
622 
623 	MRW_WLOCK();
624 
625 	if (V_ip_mrouter == NULL) {
626 		MRW_WUNLOCK();
627 		return;
628 	}
629 
630 	/*
631 	 * Tear down multicast forwarder state associated with this ifnet.
632 	 * 1. Walk the vif list, matching vifs against this ifnet.
633 	 * 2. Walk the multicast forwarding cache (mfc) looking for
634 	 *    inner matches with this vif's index.
635 	 * 3. Expire any matching multicast forwarding cache entries.
636 	 * 4. Free vif state. This should disable ALLMULTI on the interface.
637 	 */
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, &free_ptr);
651 		if (free_ptr != NULL)
652 			vifi_cnt++;
653 	}
654 
655 	MRW_WUNLOCK();
656 
657 	/*
658 	 * Free IFP. We don't have to use free_ptr here as it is the same
659 	 * that ifp. Perform free as many times as required in case
660 	 * refcount is greater than 1.
661 	 */
662 	for (i = 0; i < vifi_cnt; i++)
663 		if_free(ifp);
664 }
665 
666 static void
667 ip_mrouter_upcall_thread(void *arg, int pending __unused)
668 {
669 	CURVNET_SET((struct vnet *) arg);
670 
671 	MRW_WLOCK();
672 	bw_upcalls_send();
673 	MRW_WUNLOCK();
674 
675 	CURVNET_RESTORE();
676 }
677 
678 /*
679  * Enable multicast forwarding.
680  */
681 static int
682 ip_mrouter_init(struct socket *so, int version)
683 {
684 
685 	CTR2(KTR_IPMF, "%s: so %p", __func__, so);
686 
687 	if (version != 1)
688 		return ENOPROTOOPT;
689 
690 	MRW_WLOCK();
691 
692 	if (ip_mrouter_unloading) {
693 		MRW_WUNLOCK();
694 		return ENOPROTOOPT;
695 	}
696 
697 	if (V_ip_mrouter != NULL) {
698 		MRW_WUNLOCK();
699 		return EADDRINUSE;
700 	}
701 
702 	V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
703 	    HASH_NOWAIT);
704 
705 	/* Create upcall ring */
706 	mtx_init(&V_bw_upcalls_ring_mtx, "mroute upcall buf_ring mtx", NULL, MTX_DEF);
707 	V_bw_upcalls_ring = buf_ring_alloc(BW_UPCALLS_MAX, M_MRTABLE,
708 	    M_NOWAIT, &V_bw_upcalls_ring_mtx);
709 	if (!V_bw_upcalls_ring) {
710 		MRW_WUNLOCK();
711 		return (ENOMEM);
712 	}
713 
714 	TASK_INIT(&V_task, 0, ip_mrouter_upcall_thread, curvnet);
715 	taskqueue_cancel(V_task_queue, &V_task, NULL);
716 	taskqueue_unblock(V_task_queue);
717 
718 	callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
719 	    curvnet);
720 	callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
721 	    curvnet);
722 
723 	V_ip_mrouter = so;
724 	atomic_add_int(&ip_mrouter_cnt, 1);
725 
726 	/* This is a mutex required by buf_ring init, but not used internally */
727 	mtx_init(&V_buf_ring_mtx, "mroute buf_ring mtx", NULL, MTX_DEF);
728 
729 	MRW_WUNLOCK();
730 
731 	CTR1(KTR_IPMF, "%s: done", __func__);
732 
733 	return 0;
734 }
735 
736 /*
737  * Disable multicast forwarding.
738  */
739 static int
740 X_ip_mrouter_done(void)
741 {
742 	struct ifnet **ifps;
743 	int nifp;
744 	u_long i;
745 	vifi_t vifi;
746 	struct bw_upcall *bu;
747 
748 	if (V_ip_mrouter == NULL)
749 		return (EINVAL);
750 
751 	/*
752 	 * Detach/disable hooks to the reset of the system.
753 	 */
754 	V_ip_mrouter = NULL;
755 	atomic_subtract_int(&ip_mrouter_cnt, 1);
756 	V_mrt_api_config = 0;
757 
758 	/*
759 	 * Wait for all epoch sections to complete to ensure
760 	 * V_ip_mrouter = NULL is visible to others.
761 	 */
762 	epoch_wait_preempt(net_epoch_preempt);
763 
764 	/* Stop and drain task queue */
765 	taskqueue_block(V_task_queue);
766 	while (taskqueue_cancel(V_task_queue, &V_task, NULL)) {
767 		taskqueue_drain(V_task_queue, &V_task);
768 	}
769 
770 	ifps = malloc(MAXVIFS * sizeof(*ifps), M_TEMP, M_WAITOK);
771 
772 	MRW_WLOCK();
773 	taskqueue_cancel(V_task_queue, &V_task, NULL);
774 
775 	/* Destroy upcall ring */
776 	while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
777 		free(bu, M_MRTABLE);
778 	}
779 	buf_ring_free(V_bw_upcalls_ring, M_MRTABLE);
780 	mtx_destroy(&V_bw_upcalls_ring_mtx);
781 
782 	/*
783 	 * For each phyint in use, prepare to disable promiscuous reception
784 	 * of all IP multicasts.  Defer the actual call until the lock is released;
785 	 * just record the list of interfaces while locked.  Some interfaces use
786 	 * sx locks in their ioctl routines, which is not allowed while holding
787 	 * a non-sleepable lock.
788 	 */
789 	KASSERT(V_numvifs <= MAXVIFS, ("More vifs than possible"));
790 	for (vifi = 0, nifp = 0; vifi < V_numvifs; vifi++) {
791 		if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
792 		    !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
793 			ifps[nifp++] = V_viftable[vifi].v_ifp;
794 		}
795 	}
796 	bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
797 	V_numvifs = 0;
798 	V_pim_assert_enabled = 0;
799 
800 	callout_stop(&V_expire_upcalls_ch);
801 	callout_stop(&V_bw_upcalls_ch);
802 
803 	/*
804 	 * Free all multicast forwarding cache entries.
805 	 * Do not use hashdestroy(), as we must perform other cleanup.
806 	 */
807 	for (i = 0; i < mfchashsize; i++) {
808 		struct mfc *rt, *nrt;
809 
810 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
811 			expire_mfc(rt);
812 		}
813 	}
814 	free(V_mfchashtbl, M_MRTABLE);
815 	V_mfchashtbl = NULL;
816 
817 	bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
818 
819 	V_reg_vif_num = VIFI_INVALID;
820 
821 	mtx_destroy(&V_buf_ring_mtx);
822 
823 	MRW_WUNLOCK();
824 
825 	/*
826 	 * Now drop our claim on promiscuous multicast on the interfaces recorded
827 	 * above.  This is safe to do now because ALLMULTI is reference counted.
828 	 */
829 	for (vifi = 0; vifi < nifp; vifi++)
830 		if_allmulti(ifps[vifi], 0);
831 	free(ifps, M_TEMP);
832 
833 	CTR1(KTR_IPMF, "%s: done", __func__);
834 
835 	return 0;
836 }
837 
838 /*
839  * Set PIM assert processing global
840  */
841 static int
842 set_assert(int i)
843 {
844 	if ((i != 1) && (i != 0))
845 		return EINVAL;
846 
847 	V_pim_assert_enabled = i;
848 
849 	return 0;
850 }
851 
852 /*
853  * Configure API capabilities
854  */
855 int
856 set_api_config(uint32_t *apival)
857 {
858 	u_long i;
859 
860 	/*
861 	 * We can set the API capabilities only if it is the first operation
862 	 * after MRT_INIT. I.e.:
863 	 *  - there are no vifs installed
864 	 *  - pim_assert is not enabled
865 	 *  - the MFC table is empty
866 	 */
867 	if (V_numvifs > 0) {
868 		*apival = 0;
869 		return EPERM;
870 	}
871 	if (V_pim_assert_enabled) {
872 		*apival = 0;
873 		return EPERM;
874 	}
875 
876 	MRW_RLOCK();
877 
878 	for (i = 0; i < mfchashsize; i++) {
879 		if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
880 			MRW_RUNLOCK();
881 			*apival = 0;
882 			return EPERM;
883 		}
884 	}
885 
886 	MRW_RUNLOCK();
887 
888 	V_mrt_api_config = *apival & mrt_api_support;
889 	*apival = V_mrt_api_config;
890 
891 	return 0;
892 }
893 
894 /*
895  * Add a vif to the vif table
896  */
897 static int
898 add_vif(struct vifctl *vifcp)
899 {
900 	struct vif *vifp = V_viftable + vifcp->vifc_vifi;
901 	struct sockaddr_in sin = {sizeof sin, AF_INET};
902 	struct ifaddr *ifa;
903 	struct ifnet *ifp;
904 	int error;
905 
906 	if (vifcp->vifc_vifi >= MAXVIFS)
907 		return EINVAL;
908 	/* rate limiting is no longer supported by this code */
909 	if (vifcp->vifc_rate_limit != 0) {
910 		log(LOG_ERR, "rate limiting is no longer supported\n");
911 		return EINVAL;
912 	}
913 
914 	if (in_nullhost(vifcp->vifc_lcl_addr))
915 		return EADDRNOTAVAIL;
916 
917 	/* Find the interface with an address in AF_INET family */
918 	if (vifcp->vifc_flags & VIFF_REGISTER) {
919 		/*
920 		 * XXX: Because VIFF_REGISTER does not really need a valid
921 		 * local interface (e.g. it could be 127.0.0.2), we don't
922 		 * check its address.
923 		 */
924 		ifp = NULL;
925 	} else {
926 		struct epoch_tracker et;
927 
928 		sin.sin_addr = vifcp->vifc_lcl_addr;
929 		NET_EPOCH_ENTER(et);
930 		ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
931 		if (ifa == NULL) {
932 			NET_EPOCH_EXIT(et);
933 			return EADDRNOTAVAIL;
934 		}
935 		ifp = ifa->ifa_ifp;
936 		/* XXX FIXME we need to take a ref on ifp and cleanup properly! */
937 		NET_EPOCH_EXIT(et);
938 	}
939 
940 	if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
941 		CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
942 		return EOPNOTSUPP;
943 	} else if (vifcp->vifc_flags & VIFF_REGISTER) {
944 		ifp = V_multicast_register_if = if_alloc(IFT_LOOP);
945 		CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
946 		if (V_reg_vif_num == VIFI_INVALID) {
947 			if_initname(V_multicast_register_if, "register_vif", 0);
948 			V_reg_vif_num = vifcp->vifc_vifi;
949 		}
950 	} else {		/* Make sure the interface supports multicast */
951 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
952 			return EOPNOTSUPP;
953 
954 		/* Enable promiscuous reception of all IP multicasts from the if */
955 		error = if_allmulti(ifp, 1);
956 		if (error)
957 			return error;
958 	}
959 
960 	MRW_WLOCK();
961 
962 	if (!in_nullhost(vifp->v_lcl_addr)) {
963 		if (ifp)
964 			V_multicast_register_if = NULL;
965 		MRW_WUNLOCK();
966 		if (ifp)
967 			if_free(ifp);
968 		return EADDRINUSE;
969 	}
970 
971 	vifp->v_flags     = vifcp->vifc_flags;
972 	vifp->v_threshold = vifcp->vifc_threshold;
973 	vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
974 	vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
975 	vifp->v_ifp       = ifp;
976 	/* initialize per vif pkt counters */
977 	vifp->v_pkt_in    = 0;
978 	vifp->v_pkt_out   = 0;
979 	vifp->v_bytes_in  = 0;
980 	vifp->v_bytes_out = 0;
981 	sprintf(vifp->v_spin_name, "BM[%d] spin", vifcp->vifc_vifi);
982 	mtx_init(&vifp->v_spin, vifp->v_spin_name, NULL, MTX_SPIN);
983 
984 	/* Adjust numvifs up if the vifi is higher than numvifs */
985 	if (V_numvifs <= vifcp->vifc_vifi)
986 		V_numvifs = vifcp->vifc_vifi + 1;
987 
988 	MRW_WUNLOCK();
989 
990 	CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
991 	    (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
992 	    (int)vifcp->vifc_threshold);
993 
994 	return 0;
995 }
996 
997 /*
998  * Delete a vif from the vif table
999  */
1000 static int
1001 del_vif_locked(vifi_t vifi, struct ifnet **ifp_free)
1002 {
1003 	struct vif *vifp;
1004 
1005 	*ifp_free = NULL;
1006 
1007 	MRW_WLOCK_ASSERT();
1008 
1009 	if (vifi >= V_numvifs) {
1010 		return EINVAL;
1011 	}
1012 	vifp = &V_viftable[vifi];
1013 	if (in_nullhost(vifp->v_lcl_addr)) {
1014 		return EADDRNOTAVAIL;
1015 	}
1016 
1017 	if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1018 		if_allmulti(vifp->v_ifp, 0);
1019 
1020 	if (vifp->v_flags & VIFF_REGISTER) {
1021 		V_reg_vif_num = VIFI_INVALID;
1022 		if (vifp->v_ifp) {
1023 			if (vifp->v_ifp == V_multicast_register_if)
1024 				V_multicast_register_if = NULL;
1025 			*ifp_free = vifp->v_ifp;
1026 		}
1027 	}
1028 
1029 	mtx_destroy(&vifp->v_spin);
1030 
1031 	bzero((caddr_t)vifp, sizeof (*vifp));
1032 
1033 	CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
1034 
1035 	/* Adjust numvifs down */
1036 	for (vifi = V_numvifs; vifi > 0; vifi--)
1037 		if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
1038 			break;
1039 	V_numvifs = vifi;
1040 
1041 	return 0;
1042 }
1043 
1044 static int
1045 del_vif(vifi_t vifi)
1046 {
1047 	int cc;
1048 	struct ifnet *free_ptr;
1049 
1050 	MRW_WLOCK();
1051 	cc = del_vif_locked(vifi, &free_ptr);
1052 	MRW_WUNLOCK();
1053 
1054 	if (free_ptr)
1055 		if_free(free_ptr);
1056 
1057 	return cc;
1058 }
1059 
1060 /*
1061  * update an mfc entry without resetting counters and S,G addresses.
1062  */
1063 static void
1064 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1065 {
1066 	int i;
1067 
1068 	rt->mfc_parent = mfccp->mfcc_parent;
1069 	for (i = 0; i < V_numvifs; i++) {
1070 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1071 		rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1072 			MRT_MFC_FLAGS_ALL;
1073 	}
1074 	/* set the RP address */
1075 	if (V_mrt_api_config & MRT_MFC_RP)
1076 		rt->mfc_rp = mfccp->mfcc_rp;
1077 	else
1078 		rt->mfc_rp.s_addr = INADDR_ANY;
1079 }
1080 
1081 /*
1082  * fully initialize an mfc entry from the parameter.
1083  */
1084 static void
1085 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1086 {
1087 	rt->mfc_origin     = mfccp->mfcc_origin;
1088 	rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1089 
1090 	update_mfc_params(rt, mfccp);
1091 
1092 	/* initialize pkt counters per src-grp */
1093 	rt->mfc_pkt_cnt    = 0;
1094 	rt->mfc_byte_cnt   = 0;
1095 	rt->mfc_wrong_if   = 0;
1096 	timevalclear(&rt->mfc_last_assert);
1097 }
1098 
1099 static void
1100 expire_mfc(struct mfc *rt)
1101 {
1102 	struct rtdetq *rte;
1103 
1104 	MRW_WLOCK_ASSERT();
1105 
1106 	free_bw_list(rt->mfc_bw_meter_leq);
1107 	free_bw_list(rt->mfc_bw_meter_geq);
1108 
1109 	while (!buf_ring_empty(rt->mfc_stall_ring)) {
1110 		rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1111 		if (rte) {
1112 			m_freem(rte->m);
1113 			free(rte, M_MRTABLE);
1114 		}
1115 	}
1116 	buf_ring_free(rt->mfc_stall_ring, M_MRTABLE);
1117 
1118 	LIST_REMOVE(rt, mfc_hash);
1119 	free(rt, M_MRTABLE);
1120 }
1121 
1122 /*
1123  * Add an mfc entry
1124  */
1125 static int
1126 add_mfc(struct mfcctl2 *mfccp)
1127 {
1128 	struct mfc *rt;
1129 	struct rtdetq *rte;
1130 	u_long hash = 0;
1131 	u_short nstl;
1132 	struct epoch_tracker et;
1133 
1134 	MRW_WLOCK();
1135 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1136 
1137 	/* If an entry already exists, just update the fields */
1138 	if (rt) {
1139 		CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1140 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1141 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1142 		    mfccp->mfcc_parent);
1143 		update_mfc_params(rt, mfccp);
1144 		MRW_WUNLOCK();
1145 		return (0);
1146 	}
1147 
1148 	/*
1149 	 * Find the entry for which the upcall was made and update
1150 	 */
1151 	nstl = 0;
1152 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1153 	NET_EPOCH_ENTER(et);
1154 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1155 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1156 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1157 		    !buf_ring_empty(rt->mfc_stall_ring)) {
1158 			CTR5(KTR_IPMF,
1159 			   "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1160 			    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1161 			    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1162 			    mfccp->mfcc_parent,
1163 			    rt->mfc_stall_ring);
1164 			if (nstl++)
1165 				CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1166 
1167 			init_mfc_params(rt, mfccp);
1168 			rt->mfc_expire = 0;	/* Don't clean this guy up */
1169 			V_nexpire[hash]--;
1170 
1171 			/* Free queued packets, but attempt to forward them first. */
1172 			while (!buf_ring_empty(rt->mfc_stall_ring)) {
1173 				rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1174 				if (rte->ifp != NULL)
1175 					ip_mdq(rte->m, rte->ifp, rt, -1);
1176 				m_freem(rte->m);
1177 				free(rte, M_MRTABLE);
1178 			}
1179 		}
1180 	}
1181 	NET_EPOCH_EXIT(et);
1182 
1183 	/*
1184 	 * It is possible that an entry is being inserted without an upcall
1185 	 */
1186 	if (nstl == 0) {
1187 		CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1188 		LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1189 			if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1190 			    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1191 				init_mfc_params(rt, mfccp);
1192 				if (rt->mfc_expire)
1193 					V_nexpire[hash]--;
1194 				rt->mfc_expire = 0;
1195 				break; /* XXX */
1196 			}
1197 		}
1198 
1199 		if (rt == NULL) {		/* no upcall, so make a new entry */
1200 			rt = mfc_alloc();
1201 			if (rt == NULL) {
1202 				MRW_WUNLOCK();
1203 				return (ENOBUFS);
1204 			}
1205 
1206 			init_mfc_params(rt, mfccp);
1207 
1208 			rt->mfc_expire     = 0;
1209 			rt->mfc_bw_meter_leq = NULL;
1210 			rt->mfc_bw_meter_geq = NULL;
1211 
1212 			/* insert new entry at head of hash chain */
1213 			LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1214 		}
1215 	}
1216 
1217 	MRW_WUNLOCK();
1218 
1219 	return (0);
1220 }
1221 
1222 /*
1223  * Delete an mfc entry
1224  */
1225 static int
1226 del_mfc(struct mfcctl2 *mfccp)
1227 {
1228 	struct in_addr origin;
1229 	struct in_addr mcastgrp;
1230 	struct mfc *rt;
1231 
1232 	origin = mfccp->mfcc_origin;
1233 	mcastgrp = mfccp->mfcc_mcastgrp;
1234 
1235 	CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1236 			ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1237 
1238 	MRW_WLOCK();
1239 
1240 	rt = mfc_find(&origin, &mcastgrp);
1241 	if (rt == NULL) {
1242 		MRW_WUNLOCK();
1243 		return EADDRNOTAVAIL;
1244 	}
1245 
1246 	/*
1247 	 * free the bw_meter entries
1248 	 */
1249 	free_bw_list(rt->mfc_bw_meter_leq);
1250 	rt->mfc_bw_meter_leq = NULL;
1251 	free_bw_list(rt->mfc_bw_meter_geq);
1252 	rt->mfc_bw_meter_geq = NULL;
1253 
1254 	LIST_REMOVE(rt, mfc_hash);
1255 	free(rt, M_MRTABLE);
1256 
1257 	MRW_WUNLOCK();
1258 
1259 	return (0);
1260 }
1261 
1262 /*
1263  * Send a message to the routing daemon on the multicast routing socket.
1264  */
1265 static int
1266 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1267 {
1268 	if (s) {
1269 		SOCKBUF_LOCK(&s->so_rcv);
1270 		if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1271 		    NULL) != 0) {
1272 			sorwakeup_locked(s);
1273 			return 0;
1274 		}
1275 		soroverflow_locked(s);
1276 	}
1277 	m_freem(mm);
1278 	return -1;
1279 }
1280 
1281 /*
1282  * IP multicast forwarding function. This function assumes that the packet
1283  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1284  * pointed to by "ifp", and the packet is to be relayed to other networks
1285  * that have members of the packet's destination IP multicast group.
1286  *
1287  * The packet is returned unscathed to the caller, unless it is
1288  * erroneous, in which case a non-zero return value tells the caller to
1289  * discard it.
1290  */
1291 
1292 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1293 
1294 static int
1295 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1296     struct ip_moptions *imo)
1297 {
1298 	struct mfc *rt;
1299 	int error;
1300 	vifi_t vifi;
1301 	struct mbuf *mb0;
1302 	struct rtdetq *rte;
1303 	u_long hash;
1304 	int hlen;
1305 
1306 	CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1307 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1308 
1309 	if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1310 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1311 		/*
1312 		 * Packet arrived via a physical interface or
1313 		 * an encapsulated tunnel or a register_vif.
1314 		 */
1315 	} else {
1316 		/*
1317 		 * Packet arrived through a source-route tunnel.
1318 		 * Source-route tunnels are no longer supported.
1319 		 */
1320 		return (1);
1321 	}
1322 
1323 	/*
1324 	 * BEGIN: MCAST ROUTING HOT PATH
1325 	 */
1326 	MRW_RLOCK();
1327 	if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1328 		if (ip->ip_ttl < MAXTTL)
1329 			ip->ip_ttl++; /* compensate for -1 in *_send routines */
1330 		error = ip_mdq(m, ifp, NULL, vifi);
1331 		MRW_RUNLOCK();
1332 		return error;
1333 	}
1334 
1335 	/*
1336 	 * Don't forward a packet with time-to-live of zero or one,
1337 	 * or a packet destined to a local-only group.
1338 	 */
1339 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1340 		MRW_RUNLOCK();
1341 		return 0;
1342 	}
1343 
1344 mfc_find_retry:
1345 	/*
1346 	 * Determine forwarding vifs from the forwarding cache table
1347 	 */
1348 	MRTSTAT_INC(mrts_mfc_lookups);
1349 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1350 
1351 	/* Entry exists, so forward if necessary */
1352 	if (rt != NULL) {
1353 		error = ip_mdq(m, ifp, rt, -1);
1354 		/* Generic unlock here as we might release R or W lock */
1355 		MRW_UNLOCK();
1356 		return error;
1357 	}
1358 
1359 	/*
1360 	 * END: MCAST ROUTING HOT PATH
1361 	 */
1362 
1363 	/* Further processing must be done with WLOCK taken */
1364 	if ((MRW_WOWNED() == 0) && (MRW_LOCK_TRY_UPGRADE() == 0)) {
1365 		MRW_RUNLOCK();
1366 		MRW_WLOCK();
1367 		goto mfc_find_retry;
1368 	}
1369 
1370 	/*
1371 	 * If we don't have a route for packet's origin,
1372 	 * Make a copy of the packet & send message to routing daemon
1373 	 */
1374 	hlen = ip->ip_hl << 2;
1375 
1376 	MRTSTAT_INC(mrts_mfc_misses);
1377 	MRTSTAT_INC(mrts_no_route);
1378 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1379 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1380 
1381 	/*
1382 	 * Allocate mbufs early so that we don't do extra work if we are
1383 	 * just going to fail anyway.  Make sure to pullup the header so
1384 	 * that other people can't step on it.
1385 	 */
1386 	rte = malloc((sizeof *rte), M_MRTABLE, M_NOWAIT|M_ZERO);
1387 	if (rte == NULL) {
1388 		MRW_WUNLOCK();
1389 		return ENOBUFS;
1390 	}
1391 
1392 	mb0 = m_copypacket(m, M_NOWAIT);
1393 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1394 		mb0 = m_pullup(mb0, hlen);
1395 	if (mb0 == NULL) {
1396 		free(rte, M_MRTABLE);
1397 		MRW_WUNLOCK();
1398 		return ENOBUFS;
1399 	}
1400 
1401 	/* is there an upcall waiting for this flow ? */
1402 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1403 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash)
1404 	{
1405 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1406 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1407 		    !buf_ring_empty(rt->mfc_stall_ring))
1408 			break;
1409 	}
1410 
1411 	if (rt == NULL) {
1412 		int i;
1413 		struct igmpmsg *im;
1414 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1415 		struct mbuf *mm;
1416 
1417 		/*
1418 		 * Locate the vifi for the incoming interface for this packet.
1419 		 * If none found, drop packet.
1420 		 */
1421 		for (vifi = 0; vifi < V_numvifs &&
1422 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1423 			;
1424 		if (vifi >= V_numvifs) /* vif not found, drop packet */
1425 			goto non_fatal;
1426 
1427 		/* no upcall, so make a new entry */
1428 		rt = mfc_alloc();
1429 		if (rt == NULL)
1430 			goto fail;
1431 
1432 		/* Make a copy of the header to send to the user level process */
1433 		mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1434 		if (mm == NULL)
1435 			goto fail1;
1436 
1437 		/*
1438 		 * Send message to routing daemon to install
1439 		 * a route into the kernel table
1440 		 */
1441 
1442 		im = mtod(mm, struct igmpmsg*);
1443 		im->im_msgtype = IGMPMSG_NOCACHE;
1444 		im->im_mbz = 0;
1445 		im->im_vif = vifi;
1446 
1447 		MRTSTAT_INC(mrts_upcalls);
1448 
1449 		k_igmpsrc.sin_addr = ip->ip_src;
1450 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1451 			CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1452 			MRTSTAT_INC(mrts_upq_sockfull);
1453 			fail1: free(rt, M_MRTABLE);
1454 			fail: free(rte, M_MRTABLE);
1455 			m_freem(mb0);
1456 			MRW_WUNLOCK();
1457 			return ENOBUFS;
1458 		}
1459 
1460 		/* insert new entry at head of hash chain */
1461 		rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1462 		rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1463 		rt->mfc_expire = UPCALL_EXPIRE;
1464 		V_nexpire[hash]++;
1465 		for (i = 0; i < V_numvifs; i++) {
1466 			rt->mfc_ttls[i] = 0;
1467 			rt->mfc_flags[i] = 0;
1468 		}
1469 		rt->mfc_parent = -1;
1470 
1471 		/* clear the RP address */
1472 		rt->mfc_rp.s_addr = INADDR_ANY;
1473 		rt->mfc_bw_meter_leq = NULL;
1474 		rt->mfc_bw_meter_geq = NULL;
1475 
1476 		/* initialize pkt counters per src-grp */
1477 		rt->mfc_pkt_cnt = 0;
1478 		rt->mfc_byte_cnt = 0;
1479 		rt->mfc_wrong_if = 0;
1480 		timevalclear(&rt->mfc_last_assert);
1481 
1482 		buf_ring_enqueue(rt->mfc_stall_ring, rte);
1483 
1484 		/* Add RT to hashtable as it didn't exist before */
1485 		LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1486 	} else {
1487 		/* determine if queue has overflowed */
1488 		if (buf_ring_full(rt->mfc_stall_ring)) {
1489 			MRTSTAT_INC(mrts_upq_ovflw);
1490 			non_fatal: free(rte, M_MRTABLE);
1491 			m_freem(mb0);
1492 			MRW_WUNLOCK();
1493 			return (0);
1494 		}
1495 
1496 		buf_ring_enqueue(rt->mfc_stall_ring, rte);
1497 	}
1498 
1499 	rte->m = mb0;
1500 	rte->ifp = ifp;
1501 
1502 	MRW_WUNLOCK();
1503 
1504 	return 0;
1505 }
1506 
1507 /*
1508  * Clean up the cache entry if upcall is not serviced
1509  */
1510 static void
1511 expire_upcalls(void *arg)
1512 {
1513 	u_long i;
1514 
1515 	CURVNET_SET((struct vnet *) arg);
1516 
1517 	/*This callout is always run with MRW_WLOCK taken. */
1518 
1519 	for (i = 0; i < mfchashsize; i++) {
1520 		struct mfc *rt, *nrt;
1521 
1522 		if (V_nexpire[i] == 0)
1523 			continue;
1524 
1525 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1526 			if (buf_ring_empty(rt->mfc_stall_ring))
1527 				continue;
1528 
1529 			if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1530 				continue;
1531 
1532 			MRTSTAT_INC(mrts_cache_cleanups);
1533 			CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1534 			    (u_long)ntohl(rt->mfc_origin.s_addr),
1535 			    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1536 
1537 			expire_mfc(rt);
1538 		}
1539 	}
1540 
1541 	callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1542 	    curvnet);
1543 
1544 	CURVNET_RESTORE();
1545 }
1546 
1547 /*
1548  * Packet forwarding routine once entry in the cache is made
1549  */
1550 static int
1551 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1552 {
1553 	struct ip *ip = mtod(m, struct ip *);
1554 	vifi_t vifi;
1555 	int plen = ntohs(ip->ip_len);
1556 
1557 	MRW_LOCK_ASSERT();
1558 	NET_EPOCH_ASSERT();
1559 
1560 	/*
1561 	 * If xmt_vif is not -1, send on only the requested vif.
1562 	 *
1563 	 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1564 	 */
1565 	if (xmt_vif < V_numvifs) {
1566 		if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1567 			pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1568 		else
1569 			phyint_send(ip, V_viftable + xmt_vif, m);
1570 		return 1;
1571 	}
1572 
1573 	/*
1574 	 * Don't forward if it didn't arrive from the parent vif for its origin.
1575 	 */
1576 	vifi = rt->mfc_parent;
1577 	if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1578 		CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1579 				__func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1580 		MRTSTAT_INC(mrts_wrong_if);
1581 		++rt->mfc_wrong_if;
1582 		/*
1583 		 * If we are doing PIM assert processing, send a message
1584 		 * to the routing daemon.
1585 		 *
1586 		 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1587 		 * can complete the SPT switch, regardless of the type
1588 		 * of the iif (broadcast media, GRE tunnel, etc).
1589 		 */
1590 		if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1591 		    V_viftable[vifi].v_ifp) {
1592 			if (ifp == V_multicast_register_if)
1593 				PIMSTAT_INC(pims_rcv_registers_wrongiif);
1594 
1595 			/* Get vifi for the incoming packet */
1596 			for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp; vifi++)
1597 				;
1598 			if (vifi >= V_numvifs)
1599 				return 0;	/* The iif is not found: ignore the packet. */
1600 
1601 			if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1602 				return 0;	/* WRONGVIF disabled: ignore the packet */
1603 
1604 			if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1605 				struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1606 				struct igmpmsg *im;
1607 				int hlen = ip->ip_hl << 2;
1608 				struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1609 
1610 				if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1611 					mm = m_pullup(mm, hlen);
1612 				if (mm == NULL)
1613 					return ENOBUFS;
1614 
1615 				im = mtod(mm, struct igmpmsg *);
1616 				im->im_msgtype = IGMPMSG_WRONGVIF;
1617 				im->im_mbz = 0;
1618 				im->im_vif = vifi;
1619 
1620 				MRTSTAT_INC(mrts_upcalls);
1621 
1622 				k_igmpsrc.sin_addr = im->im_src;
1623 				if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1624 					CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1625 					MRTSTAT_INC(mrts_upq_sockfull);
1626 					return ENOBUFS;
1627 				}
1628 			}
1629 		}
1630 		return 0;
1631 	}
1632 
1633 	/* If I sourced this packet, it counts as output, else it was input. */
1634 	mtx_lock_spin(&V_viftable[vifi].v_spin);
1635 	if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1636 		V_viftable[vifi].v_pkt_out++;
1637 		V_viftable[vifi].v_bytes_out += plen;
1638 	} else {
1639 		V_viftable[vifi].v_pkt_in++;
1640 		V_viftable[vifi].v_bytes_in += plen;
1641 	}
1642 	mtx_unlock_spin(&V_viftable[vifi].v_spin);
1643 
1644 	rt->mfc_pkt_cnt++;
1645 	rt->mfc_byte_cnt += plen;
1646 
1647 	/*
1648 	 * For each vif, decide if a copy of the packet should be forwarded.
1649 	 * Forward if:
1650 	 *		- the ttl exceeds the vif's threshold
1651 	 *		- there are group members downstream on interface
1652 	 */
1653 	for (vifi = 0; vifi < V_numvifs; vifi++)
1654 		if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1655 			V_viftable[vifi].v_pkt_out++;
1656 			V_viftable[vifi].v_bytes_out += plen;
1657 			if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1658 				pim_register_send(ip, V_viftable + vifi, m, rt);
1659 			else
1660 				phyint_send(ip, V_viftable + vifi, m);
1661 		}
1662 
1663 	/*
1664 	 * Perform upcall-related bw measuring.
1665 	 */
1666 	if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1667 		struct bw_meter *x;
1668 		struct timeval now;
1669 
1670 		microtime(&now);
1671 		/* Process meters for Greater-or-EQual case */
1672 		for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1673 			bw_meter_geq_receive_packet(x, plen, &now);
1674 
1675 		/* Process meters for Lower-or-EQual case */
1676 		for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1677 			/*
1678 			 * Record that a packet is received.
1679 			 * Spin lock has to be taken as callout context
1680 			 * (expire_bw_meter_leq) might modify these fields
1681 			 * as well
1682 			 */
1683 			mtx_lock_spin(&x->bm_spin);
1684 			x->bm_measured.b_packets++;
1685 			x->bm_measured.b_bytes += plen;
1686 			mtx_unlock_spin(&x->bm_spin);
1687 		}
1688 	}
1689 
1690 	return 0;
1691 }
1692 
1693 /*
1694  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1695  */
1696 static int
1697 X_legal_vif_num(int vif)
1698 {
1699 	int ret;
1700 
1701 	ret = 0;
1702 	if (vif < 0)
1703 		return (ret);
1704 
1705 	MRW_RLOCK();
1706 	if (vif < V_numvifs)
1707 		ret = 1;
1708 	MRW_RUNLOCK();
1709 
1710 	return (ret);
1711 }
1712 
1713 /*
1714  * Return the local address used by this vif
1715  */
1716 static u_long
1717 X_ip_mcast_src(int vifi)
1718 {
1719 	in_addr_t addr;
1720 
1721 	addr = INADDR_ANY;
1722 	if (vifi < 0)
1723 		return (addr);
1724 
1725 	MRW_RLOCK();
1726 	if (vifi < V_numvifs)
1727 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1728 	MRW_RUNLOCK();
1729 
1730 	return (addr);
1731 }
1732 
1733 static void
1734 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1735 {
1736 	struct mbuf *mb_copy;
1737 	int hlen = ip->ip_hl << 2;
1738 
1739 	MRW_LOCK_ASSERT();
1740 
1741 	/*
1742 	 * Make a new reference to the packet; make sure that
1743 	 * the IP header is actually copied, not just referenced,
1744 	 * so that ip_output() only scribbles on the copy.
1745 	 */
1746 	mb_copy = m_copypacket(m, M_NOWAIT);
1747 	if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1748 		mb_copy = m_pullup(mb_copy, hlen);
1749 	if (mb_copy == NULL)
1750 		return;
1751 
1752 	send_packet(vifp, mb_copy);
1753 }
1754 
1755 static void
1756 send_packet(struct vif *vifp, struct mbuf *m)
1757 {
1758 	struct ip_moptions imo;
1759 	int error __unused;
1760 
1761 	MRW_LOCK_ASSERT();
1762 	NET_EPOCH_ASSERT();
1763 
1764 	imo.imo_multicast_ifp  = vifp->v_ifp;
1765 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1766 	imo.imo_multicast_loop = !!in_mcast_loop;
1767 	imo.imo_multicast_vif  = -1;
1768 	STAILQ_INIT(&imo.imo_head);
1769 
1770 	/*
1771 	 * Re-entrancy should not be a problem here, because
1772 	 * the packets that we send out and are looped back at us
1773 	 * should get rejected because they appear to come from
1774 	 * the loopback interface, thus preventing looping.
1775 	 */
1776 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1777 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1778 	    (ptrdiff_t)(vifp - V_viftable), error);
1779 }
1780 
1781 /*
1782  * Stubs for old RSVP socket shim implementation.
1783  */
1784 
1785 static int
1786 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1787 {
1788 
1789 	return (EOPNOTSUPP);
1790 }
1791 
1792 static void
1793 X_ip_rsvp_force_done(struct socket *so __unused)
1794 {
1795 
1796 }
1797 
1798 static int
1799 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1800 {
1801 	struct mbuf *m;
1802 
1803 	m = *mp;
1804 	*mp = NULL;
1805 	if (!V_rsvp_on)
1806 		m_freem(m);
1807 	return (IPPROTO_DONE);
1808 }
1809 
1810 /*
1811  * Code for bandwidth monitors
1812  */
1813 
1814 /*
1815  * Define common interface for timeval-related methods
1816  */
1817 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1818 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1819 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1820 
1821 static uint32_t
1822 compute_bw_meter_flags(struct bw_upcall *req)
1823 {
1824 	uint32_t flags = 0;
1825 
1826 	if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1827 		flags |= BW_METER_UNIT_PACKETS;
1828 	if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1829 		flags |= BW_METER_UNIT_BYTES;
1830 	if (req->bu_flags & BW_UPCALL_GEQ)
1831 		flags |= BW_METER_GEQ;
1832 	if (req->bu_flags & BW_UPCALL_LEQ)
1833 		flags |= BW_METER_LEQ;
1834 
1835 	return flags;
1836 }
1837 
1838 static void
1839 expire_bw_meter_leq(void *arg)
1840 {
1841 	struct bw_meter *x = arg;
1842 	struct timeval now;
1843 	/*
1844 	 * INFO:
1845 	 * callout is always executed with MRW_WLOCK taken
1846 	 */
1847 
1848 	CURVNET_SET((struct vnet *)x->arg);
1849 
1850 	microtime(&now);
1851 
1852 	/*
1853 	 * Test if we should deliver an upcall
1854 	 */
1855 	if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1856 	    (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1857 	    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1858 	    (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1859 		/* Prepare an upcall for delivery */
1860 		bw_meter_prepare_upcall(x, &now);
1861 	}
1862 
1863 	/* Send all upcalls that are pending delivery */
1864 	taskqueue_enqueue(V_task_queue, &V_task);
1865 
1866 	/* Reset counters */
1867 	x->bm_start_time = now;
1868 	/* Spin lock has to be taken as ip_forward context
1869 	 * might modify these fields as well
1870 	 */
1871 	mtx_lock_spin(&x->bm_spin);
1872 	x->bm_measured.b_bytes = 0;
1873 	x->bm_measured.b_packets = 0;
1874 	mtx_unlock_spin(&x->bm_spin);
1875 
1876 	callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1877 
1878 	CURVNET_RESTORE();
1879 }
1880 
1881 /*
1882  * Add a bw_meter entry
1883  */
1884 static int
1885 add_bw_upcall(struct bw_upcall *req)
1886 {
1887 	struct mfc *mfc;
1888 	struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1889 	BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1890 	struct timeval now;
1891 	struct bw_meter *x, **bwm_ptr;
1892 	uint32_t flags;
1893 
1894 	if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1895 		return EOPNOTSUPP;
1896 
1897 	/* Test if the flags are valid */
1898 	if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1899 		return EINVAL;
1900 	if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1901 		return EINVAL;
1902 	if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1903 		return EINVAL;
1904 
1905 	/* Test if the threshold time interval is valid */
1906 	if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1907 		return EINVAL;
1908 
1909 	flags = compute_bw_meter_flags(req);
1910 
1911 	/*
1912 	 * Find if we have already same bw_meter entry
1913 	 */
1914 	MRW_WLOCK();
1915 	mfc = mfc_find(&req->bu_src, &req->bu_dst);
1916 	if (mfc == NULL) {
1917 		MRW_WUNLOCK();
1918 		return EADDRNOTAVAIL;
1919 	}
1920 
1921 	/* Choose an appropriate bw_meter list */
1922 	if (req->bu_flags & BW_UPCALL_GEQ)
1923 		bwm_ptr = &mfc->mfc_bw_meter_geq;
1924 	else
1925 		bwm_ptr = &mfc->mfc_bw_meter_leq;
1926 
1927 	for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
1928 		if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1929 		    &req->bu_threshold.b_time, ==))
1930 		    && (x->bm_threshold.b_packets
1931 		    == req->bu_threshold.b_packets)
1932 		    && (x->bm_threshold.b_bytes
1933 		    == req->bu_threshold.b_bytes)
1934 		    && (x->bm_flags & BW_METER_USER_FLAGS)
1935 		    == flags) {
1936 			MRW_WUNLOCK();
1937 			return 0; /* XXX Already installed */
1938 		}
1939 	}
1940 
1941 	/* Allocate the new bw_meter entry */
1942 	x = malloc(sizeof(*x), M_BWMETER, M_ZERO | M_NOWAIT);
1943 	if (x == NULL) {
1944 		MRW_WUNLOCK();
1945 		return ENOBUFS;
1946 	}
1947 
1948 	/* Set the new bw_meter entry */
1949 	x->bm_threshold.b_time = req->bu_threshold.b_time;
1950 	microtime(&now);
1951 	x->bm_start_time = now;
1952 	x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1953 	x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1954 	x->bm_measured.b_packets = 0;
1955 	x->bm_measured.b_bytes = 0;
1956 	x->bm_flags = flags;
1957 	x->bm_time_next = NULL;
1958 	x->bm_mfc = mfc;
1959 	x->arg = curvnet;
1960 	sprintf(x->bm_spin_name, "BM spin %p", x);
1961 	mtx_init(&x->bm_spin, x->bm_spin_name, NULL, MTX_SPIN);
1962 
1963 	/* For LEQ case create periodic callout */
1964 	if (req->bu_flags & BW_UPCALL_LEQ) {
1965 		callout_init_rw(&x->bm_meter_callout, &mrouter_lock, CALLOUT_SHAREDLOCK);
1966 		callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
1967 		    expire_bw_meter_leq, x);
1968 	}
1969 
1970 	/* Add the new bw_meter entry to the front of entries for this MFC */
1971 	x->bm_mfc_next = *bwm_ptr;
1972 	*bwm_ptr = x;
1973 
1974 	MRW_WUNLOCK();
1975 
1976 	return 0;
1977 }
1978 
1979 static void
1980 free_bw_list(struct bw_meter *list)
1981 {
1982 	while (list != NULL) {
1983 		struct bw_meter *x = list;
1984 
1985 		/* MRW_WLOCK must be held here */
1986 		if (x->bm_flags & BW_METER_LEQ) {
1987 			callout_drain(&x->bm_meter_callout);
1988 			mtx_destroy(&x->bm_spin);
1989 		}
1990 
1991 		list = list->bm_mfc_next;
1992 		free(x, M_BWMETER);
1993 	}
1994 }
1995 
1996 /*
1997  * Delete one or multiple bw_meter entries
1998  */
1999 static int
2000 del_bw_upcall(struct bw_upcall *req)
2001 {
2002 	struct mfc *mfc;
2003 	struct bw_meter *x, **bwm_ptr;
2004 
2005 	if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
2006 		return EOPNOTSUPP;
2007 
2008 	MRW_WLOCK();
2009 
2010 	/* Find the corresponding MFC entry */
2011 	mfc = mfc_find(&req->bu_src, &req->bu_dst);
2012 	if (mfc == NULL) {
2013 		MRW_WUNLOCK();
2014 		return EADDRNOTAVAIL;
2015 	} else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2016 		/*
2017 		 * Delete all bw_meter entries for this mfc
2018 		 */
2019 		struct bw_meter *list;
2020 
2021 		/* Free LEQ list */
2022 		list = mfc->mfc_bw_meter_leq;
2023 		mfc->mfc_bw_meter_leq = NULL;
2024 		free_bw_list(list);
2025 
2026 		/* Free GEQ list */
2027 		list = mfc->mfc_bw_meter_geq;
2028 		mfc->mfc_bw_meter_geq = NULL;
2029 		free_bw_list(list);
2030 		MRW_WUNLOCK();
2031 		return 0;
2032 	} else {			/* Delete a single bw_meter entry */
2033 		struct bw_meter *prev;
2034 		uint32_t flags = 0;
2035 
2036 		flags = compute_bw_meter_flags(req);
2037 
2038 		/* Choose an appropriate bw_meter list */
2039 		if (req->bu_flags & BW_UPCALL_GEQ)
2040 			bwm_ptr = &mfc->mfc_bw_meter_geq;
2041 		else
2042 			bwm_ptr = &mfc->mfc_bw_meter_leq;
2043 
2044 		/* Find the bw_meter entry to delete */
2045 		for (prev = NULL, x = *bwm_ptr; x != NULL;
2046 				prev = x, x = x->bm_mfc_next) {
2047 			if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, &req->bu_threshold.b_time, ==)) &&
2048 			    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2049 			    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2050 			    (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2051 				break;
2052 		}
2053 		if (x != NULL) { /* Delete entry from the list for this MFC */
2054 			if (prev != NULL)
2055 				prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2056 			else
2057 				*bwm_ptr = x->bm_mfc_next;/* new head of list */
2058 
2059 			if (req->bu_flags & BW_UPCALL_LEQ)
2060 				callout_stop(&x->bm_meter_callout);
2061 
2062 			MRW_WUNLOCK();
2063 			/* Free the bw_meter entry */
2064 			free(x, M_BWMETER);
2065 			return 0;
2066 		} else {
2067 			MRW_WUNLOCK();
2068 			return EINVAL;
2069 		}
2070 	}
2071 	__assert_unreachable();
2072 }
2073 
2074 /*
2075  * Perform bandwidth measurement processing that may result in an upcall
2076  */
2077 static void
2078 bw_meter_geq_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2079 {
2080 	struct timeval delta;
2081 
2082 	MRW_LOCK_ASSERT();
2083 
2084 	delta = *nowp;
2085 	BW_TIMEVALDECR(&delta, &x->bm_start_time);
2086 
2087 	/*
2088 	 * Processing for ">=" type of bw_meter entry.
2089 	 * bm_spin does not have to be hold here as in GEQ
2090 	 * case this is the only context accessing bm_measured.
2091 	 */
2092 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2093 	    /* Reset the bw_meter entry */
2094 	    x->bm_start_time = *nowp;
2095 	    x->bm_measured.b_packets = 0;
2096 	    x->bm_measured.b_bytes = 0;
2097 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2098 	}
2099 
2100 	/* Record that a packet is received */
2101 	x->bm_measured.b_packets++;
2102 	x->bm_measured.b_bytes += plen;
2103 
2104 	/*
2105 	 * Test if we should deliver an upcall
2106 	 */
2107 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2108 		if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2109 		    (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2110 		    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2111 		    (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2112 			/* Prepare an upcall for delivery */
2113 			bw_meter_prepare_upcall(x, nowp);
2114 			x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2115 		}
2116 	}
2117 }
2118 
2119 /*
2120  * Prepare a bandwidth-related upcall
2121  */
2122 static void
2123 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2124 {
2125 	struct timeval delta;
2126 	struct bw_upcall *u;
2127 
2128 	MRW_LOCK_ASSERT();
2129 
2130 	/*
2131 	 * Compute the measured time interval
2132 	 */
2133 	delta = *nowp;
2134 	BW_TIMEVALDECR(&delta, &x->bm_start_time);
2135 
2136 	/*
2137 	 * Set the bw_upcall entry
2138 	 */
2139 	u = malloc(sizeof(struct bw_upcall), M_MRTABLE, M_NOWAIT | M_ZERO);
2140 	if (!u) {
2141 		log(LOG_WARNING, "bw_meter_prepare_upcall: cannot allocate entry\n");
2142 		return;
2143 	}
2144 	u->bu_src = x->bm_mfc->mfc_origin;
2145 	u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2146 	u->bu_threshold.b_time = x->bm_threshold.b_time;
2147 	u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2148 	u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2149 	u->bu_measured.b_time = delta;
2150 	u->bu_measured.b_packets = x->bm_measured.b_packets;
2151 	u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2152 	u->bu_flags = 0;
2153 	if (x->bm_flags & BW_METER_UNIT_PACKETS)
2154 		u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2155 	if (x->bm_flags & BW_METER_UNIT_BYTES)
2156 		u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2157 	if (x->bm_flags & BW_METER_GEQ)
2158 		u->bu_flags |= BW_UPCALL_GEQ;
2159 	if (x->bm_flags & BW_METER_LEQ)
2160 		u->bu_flags |= BW_UPCALL_LEQ;
2161 
2162 	if (buf_ring_enqueue(V_bw_upcalls_ring, u))
2163 		log(LOG_WARNING, "bw_meter_prepare_upcall: cannot enqueue upcall\n");
2164 	if (buf_ring_count(V_bw_upcalls_ring) > (BW_UPCALLS_MAX / 2)) {
2165 		taskqueue_enqueue(V_task_queue, &V_task);
2166 	}
2167 }
2168 /*
2169  * Send the pending bandwidth-related upcalls
2170  */
2171 static void
2172 bw_upcalls_send(void)
2173 {
2174 	struct mbuf *m;
2175 	int len = 0;
2176 	struct bw_upcall *bu;
2177 	struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2178 	static struct igmpmsg igmpmsg = {
2179 		0,		/* unused1 */
2180 		0,		/* unused2 */
2181 		IGMPMSG_BW_UPCALL,/* im_msgtype */
2182 		0,		/* im_mbz  */
2183 		0,		/* im_vif  */
2184 		0,		/* unused3 */
2185 		{ 0 },		/* im_src  */
2186 		{ 0 }		/* im_dst  */
2187 	};
2188 
2189 	MRW_LOCK_ASSERT();
2190 
2191 	if (buf_ring_empty(V_bw_upcalls_ring))
2192 		return;
2193 
2194 	/*
2195 	 * Allocate a new mbuf, initialize it with the header and
2196 	 * the payload for the pending calls.
2197 	 */
2198 	m = m_gethdr(M_NOWAIT, MT_DATA);
2199 	if (m == NULL) {
2200 		log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2201 		return;
2202 	}
2203 
2204 	m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2205 	len += sizeof(struct igmpmsg);
2206 	while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
2207 		m_copyback(m, len, sizeof(struct bw_upcall), (caddr_t)bu);
2208 		len += sizeof(struct bw_upcall);
2209 		free(bu, M_MRTABLE);
2210 	}
2211 
2212 	/*
2213 	 * Send the upcalls
2214 	 * XXX do we need to set the address in k_igmpsrc ?
2215 	 */
2216 	MRTSTAT_INC(mrts_upcalls);
2217 	if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2218 		log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2219 		MRTSTAT_INC(mrts_upq_sockfull);
2220 	}
2221 }
2222 
2223 /*
2224  * A periodic function for sending all upcalls that are pending delivery
2225  */
2226 static void
2227 expire_bw_upcalls_send(void *arg)
2228 {
2229 	CURVNET_SET((struct vnet *) arg);
2230 
2231 	/* This callout is run with MRW_RLOCK taken */
2232 
2233 	bw_upcalls_send();
2234 
2235 	callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2236 	    curvnet);
2237 	CURVNET_RESTORE();
2238 }
2239 
2240 /*
2241  * End of bandwidth monitoring code
2242  */
2243 
2244 /*
2245  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2246  *
2247  */
2248 static int
2249 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2250     struct mfc *rt)
2251 {
2252 	struct mbuf *mb_copy, *mm;
2253 
2254 	/*
2255 	 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2256 	 * rendezvous point was unspecified, and we were told not to.
2257 	 */
2258 	if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2259 	    in_nullhost(rt->mfc_rp))
2260 		return 0;
2261 
2262 	mb_copy = pim_register_prepare(ip, m);
2263 	if (mb_copy == NULL)
2264 		return ENOBUFS;
2265 
2266 	/*
2267 	 * Send all the fragments. Note that the mbuf for each fragment
2268 	 * is freed by the sending machinery.
2269 	 */
2270 	for (mm = mb_copy; mm; mm = mb_copy) {
2271 		mb_copy = mm->m_nextpkt;
2272 		mm->m_nextpkt = 0;
2273 		mm = m_pullup(mm, sizeof(struct ip));
2274 		if (mm != NULL) {
2275 			ip = mtod(mm, struct ip *);
2276 			if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2277 				pim_register_send_rp(ip, vifp, mm, rt);
2278 			} else {
2279 				pim_register_send_upcall(ip, vifp, mm, rt);
2280 			}
2281 		}
2282 	}
2283 
2284 	return 0;
2285 }
2286 
2287 /*
2288  * Return a copy of the data packet that is ready for PIM Register
2289  * encapsulation.
2290  * XXX: Note that in the returned copy the IP header is a valid one.
2291  */
2292 static struct mbuf *
2293 pim_register_prepare(struct ip *ip, struct mbuf *m)
2294 {
2295 	struct mbuf *mb_copy = NULL;
2296 	int mtu;
2297 
2298 	/* Take care of delayed checksums */
2299 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2300 		in_delayed_cksum(m);
2301 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2302 	}
2303 
2304 	/*
2305 	 * Copy the old packet & pullup its IP header into the
2306 	 * new mbuf so we can modify it.
2307 	 */
2308 	mb_copy = m_copypacket(m, M_NOWAIT);
2309 	if (mb_copy == NULL)
2310 		return NULL;
2311 	mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2312 	if (mb_copy == NULL)
2313 		return NULL;
2314 
2315 	/* take care of the TTL */
2316 	ip = mtod(mb_copy, struct ip *);
2317 	--ip->ip_ttl;
2318 
2319 	/* Compute the MTU after the PIM Register encapsulation */
2320 	mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2321 
2322 	if (ntohs(ip->ip_len) <= mtu) {
2323 		/* Turn the IP header into a valid one */
2324 		ip->ip_sum = 0;
2325 		ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2326 	} else {
2327 		/* Fragment the packet */
2328 		mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2329 		if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2330 			m_freem(mb_copy);
2331 			return NULL;
2332 		}
2333 	}
2334 	return mb_copy;
2335 }
2336 
2337 /*
2338  * Send an upcall with the data packet to the user-level process.
2339  */
2340 static int
2341 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2342     struct mbuf *mb_copy, struct mfc *rt)
2343 {
2344 	struct mbuf *mb_first;
2345 	int len = ntohs(ip->ip_len);
2346 	struct igmpmsg *im;
2347 	struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2348 
2349 	MRW_LOCK_ASSERT();
2350 
2351 	/*
2352 	 * Add a new mbuf with an upcall header
2353 	 */
2354 	mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2355 	if (mb_first == NULL) {
2356 		m_freem(mb_copy);
2357 		return ENOBUFS;
2358 	}
2359 	mb_first->m_data += max_linkhdr;
2360 	mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2361 	mb_first->m_len = sizeof(struct igmpmsg);
2362 	mb_first->m_next = mb_copy;
2363 
2364 	/* Send message to routing daemon */
2365 	im = mtod(mb_first, struct igmpmsg *);
2366 	im->im_msgtype	= IGMPMSG_WHOLEPKT;
2367 	im->im_mbz		= 0;
2368 	im->im_vif		= vifp - V_viftable;
2369 	im->im_src		= ip->ip_src;
2370 	im->im_dst		= ip->ip_dst;
2371 
2372 	k_igmpsrc.sin_addr	= ip->ip_src;
2373 
2374 	MRTSTAT_INC(mrts_upcalls);
2375 
2376 	if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2377 		CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2378 		MRTSTAT_INC(mrts_upq_sockfull);
2379 		return ENOBUFS;
2380 	}
2381 
2382 	/* Keep statistics */
2383 	PIMSTAT_INC(pims_snd_registers_msgs);
2384 	PIMSTAT_ADD(pims_snd_registers_bytes, len);
2385 
2386 	return 0;
2387 }
2388 
2389 /*
2390  * Encapsulate the data packet in PIM Register message and send it to the RP.
2391  */
2392 static int
2393 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2394     struct mfc *rt)
2395 {
2396 	struct mbuf *mb_first;
2397 	struct ip *ip_outer;
2398 	struct pim_encap_pimhdr *pimhdr;
2399 	int len = ntohs(ip->ip_len);
2400 	vifi_t vifi = rt->mfc_parent;
2401 
2402 	MRW_LOCK_ASSERT();
2403 
2404 	if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2405 		m_freem(mb_copy);
2406 		return EADDRNOTAVAIL;		/* The iif vif is invalid */
2407 	}
2408 
2409 	/*
2410 	 * Add a new mbuf with the encapsulating header
2411 	 */
2412 	mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2413 	if (mb_first == NULL) {
2414 		m_freem(mb_copy);
2415 		return ENOBUFS;
2416 	}
2417 	mb_first->m_data += max_linkhdr;
2418 	mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2419 	mb_first->m_next = mb_copy;
2420 
2421 	mb_first->m_pkthdr.len = len + mb_first->m_len;
2422 
2423 	/*
2424 	 * Fill in the encapsulating IP and PIM header
2425 	 */
2426 	ip_outer = mtod(mb_first, struct ip *);
2427 	*ip_outer = pim_encap_iphdr;
2428 	ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2429 			sizeof(pim_encap_pimhdr));
2430 	ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2431 	ip_outer->ip_dst = rt->mfc_rp;
2432 	/*
2433 	 * Copy the inner header TOS to the outer header, and take care of the
2434 	 * IP_DF bit.
2435 	 */
2436 	ip_outer->ip_tos = ip->ip_tos;
2437 	if (ip->ip_off & htons(IP_DF))
2438 		ip_outer->ip_off |= htons(IP_DF);
2439 	ip_fillid(ip_outer);
2440 	pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2441 			+ sizeof(pim_encap_iphdr));
2442 	*pimhdr = pim_encap_pimhdr;
2443 	/* If the iif crosses a border, set the Border-bit */
2444 	if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2445 		pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2446 
2447 	mb_first->m_data += sizeof(pim_encap_iphdr);
2448 	pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2449 	mb_first->m_data -= sizeof(pim_encap_iphdr);
2450 
2451 	send_packet(vifp, mb_first);
2452 
2453 	/* Keep statistics */
2454 	PIMSTAT_INC(pims_snd_registers_msgs);
2455 	PIMSTAT_ADD(pims_snd_registers_bytes, len);
2456 
2457 	return 0;
2458 }
2459 
2460 /*
2461  * pim_encapcheck() is called by the encap4_input() path at runtime to
2462  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2463  * into the kernel.
2464  */
2465 static int
2466 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2467     int proto __unused, void *arg __unused)
2468 {
2469 
2470 	KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2471 	return (8);		/* claim the datagram. */
2472 }
2473 
2474 /*
2475  * PIM-SMv2 and PIM-DM messages processing.
2476  * Receives and verifies the PIM control messages, and passes them
2477  * up to the listening socket, using rip_input().
2478  * The only message with special processing is the PIM_REGISTER message
2479  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2480  * is passed to if_simloop().
2481  */
2482 static int
2483 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2484 {
2485 	struct ip *ip = mtod(m, struct ip *);
2486 	struct pim *pim;
2487 	int iphlen = off;
2488 	int minlen;
2489 	int datalen = ntohs(ip->ip_len) - iphlen;
2490 	int ip_tos;
2491 
2492 	/* Keep statistics */
2493 	PIMSTAT_INC(pims_rcv_total_msgs);
2494 	PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2495 
2496 	/*
2497 	 * Validate lengths
2498 	 */
2499 	if (datalen < PIM_MINLEN) {
2500 		PIMSTAT_INC(pims_rcv_tooshort);
2501 		CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2502 		    __func__, datalen, ntohl(ip->ip_src.s_addr));
2503 		m_freem(m);
2504 		return (IPPROTO_DONE);
2505 	}
2506 
2507 	/*
2508 	 * If the packet is at least as big as a REGISTER, go agead
2509 	 * and grab the PIM REGISTER header size, to avoid another
2510 	 * possible m_pullup() later.
2511 	 *
2512 	 * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2513 	 * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2514 	 */
2515 	minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2516 	/*
2517 	 * Get the IP and PIM headers in contiguous memory, and
2518 	 * possibly the PIM REGISTER header.
2519 	 */
2520 	if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2521 		CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2522 		return (IPPROTO_DONE);
2523 	}
2524 
2525 	/* m_pullup() may have given us a new mbuf so reset ip. */
2526 	ip = mtod(m, struct ip *);
2527 	ip_tos = ip->ip_tos;
2528 
2529 	/* adjust mbuf to point to the PIM header */
2530 	m->m_data += iphlen;
2531 	m->m_len  -= iphlen;
2532 	pim = mtod(m, struct pim *);
2533 
2534 	/*
2535 	 * Validate checksum. If PIM REGISTER, exclude the data packet.
2536 	 *
2537 	 * XXX: some older PIMv2 implementations don't make this distinction,
2538 	 * so for compatibility reason perform the checksum over part of the
2539 	 * message, and if error, then over the whole message.
2540 	 */
2541 	if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2542 		/* do nothing, checksum okay */
2543 	} else if (in_cksum(m, datalen)) {
2544 		PIMSTAT_INC(pims_rcv_badsum);
2545 		CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2546 		m_freem(m);
2547 		return (IPPROTO_DONE);
2548 	}
2549 
2550 	/* PIM version check */
2551 	if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2552 		PIMSTAT_INC(pims_rcv_badversion);
2553 		CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2554 		    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2555 		m_freem(m);
2556 		return (IPPROTO_DONE);
2557 	}
2558 
2559 	/* restore mbuf back to the outer IP */
2560 	m->m_data -= iphlen;
2561 	m->m_len  += iphlen;
2562 
2563 	if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2564 		/*
2565 		 * Since this is a REGISTER, we'll make a copy of the register
2566 		 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2567 		 * routing daemon.
2568 		 */
2569 		struct sockaddr_in dst = { sizeof(dst), AF_INET };
2570 		struct mbuf *mcp;
2571 		struct ip *encap_ip;
2572 		u_int32_t *reghdr;
2573 		struct ifnet *vifp;
2574 
2575 		MRW_RLOCK();
2576 		if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2577 			MRW_RUNLOCK();
2578 			CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2579 			    (int)V_reg_vif_num);
2580 			m_freem(m);
2581 			return (IPPROTO_DONE);
2582 		}
2583 		/* XXX need refcnt? */
2584 		vifp = V_viftable[V_reg_vif_num].v_ifp;
2585 		MRW_RUNLOCK();
2586 
2587 		/*
2588 		 * Validate length
2589 		 */
2590 		if (datalen < PIM_REG_MINLEN) {
2591 			PIMSTAT_INC(pims_rcv_tooshort);
2592 			PIMSTAT_INC(pims_rcv_badregisters);
2593 			CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2594 			m_freem(m);
2595 			return (IPPROTO_DONE);
2596 		}
2597 
2598 		reghdr = (u_int32_t *)(pim + 1);
2599 		encap_ip = (struct ip *)(reghdr + 1);
2600 
2601 		CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2602 		    __func__, ntohl(encap_ip->ip_src.s_addr),
2603 		    ntohs(encap_ip->ip_len));
2604 
2605 		/* verify the version number of the inner packet */
2606 		if (encap_ip->ip_v != IPVERSION) {
2607 			PIMSTAT_INC(pims_rcv_badregisters);
2608 			CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2609 			m_freem(m);
2610 			return (IPPROTO_DONE);
2611 		}
2612 
2613 		/* verify the inner packet is destined to a mcast group */
2614 		if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2615 			PIMSTAT_INC(pims_rcv_badregisters);
2616 			CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2617 			    ntohl(encap_ip->ip_dst.s_addr));
2618 			m_freem(m);
2619 			return (IPPROTO_DONE);
2620 		}
2621 
2622 		/* If a NULL_REGISTER, pass it to the daemon */
2623 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2624 			goto pim_input_to_daemon;
2625 
2626 		/*
2627 		 * Copy the TOS from the outer IP header to the inner IP header.
2628 		 */
2629 		if (encap_ip->ip_tos != ip_tos) {
2630 			/* Outer TOS -> inner TOS */
2631 			encap_ip->ip_tos = ip_tos;
2632 			/* Recompute the inner header checksum. Sigh... */
2633 
2634 			/* adjust mbuf to point to the inner IP header */
2635 			m->m_data += (iphlen + PIM_MINLEN);
2636 			m->m_len  -= (iphlen + PIM_MINLEN);
2637 
2638 			encap_ip->ip_sum = 0;
2639 			encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2640 
2641 			/* restore mbuf to point back to the outer IP header */
2642 			m->m_data -= (iphlen + PIM_MINLEN);
2643 			m->m_len  += (iphlen + PIM_MINLEN);
2644 		}
2645 
2646 		/*
2647 		 * Decapsulate the inner IP packet and loopback to forward it
2648 		 * as a normal multicast packet. Also, make a copy of the
2649 		 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2650 		 * to pass to the daemon later, so it can take the appropriate
2651 		 * actions (e.g., send back PIM_REGISTER_STOP).
2652 		 * XXX: here m->m_data points to the outer IP header.
2653 		 */
2654 		mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2655 		if (mcp == NULL) {
2656 			CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2657 			m_freem(m);
2658 			return (IPPROTO_DONE);
2659 		}
2660 
2661 		/* Keep statistics */
2662 		/* XXX: registers_bytes include only the encap. mcast pkt */
2663 		PIMSTAT_INC(pims_rcv_registers_msgs);
2664 		PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2665 
2666 		/*
2667 		 * forward the inner ip packet; point m_data at the inner ip.
2668 		 */
2669 		m_adj(m, iphlen + PIM_MINLEN);
2670 
2671 		CTR4(KTR_IPMF,
2672 		    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2673 		    __func__,
2674 		    (u_long)ntohl(encap_ip->ip_src.s_addr),
2675 		    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2676 		    (int)V_reg_vif_num);
2677 
2678 		/* NB: vifp was collected above; can it change on us? */
2679 		if_simloop(vifp, m, dst.sin_family, 0);
2680 
2681 		/* prepare the register head to send to the mrouting daemon */
2682 		m = mcp;
2683 	}
2684 
2685 pim_input_to_daemon:
2686 	/*
2687 	 * Pass the PIM message up to the daemon; if it is a Register message,
2688 	 * pass the 'head' only up to the daemon. This includes the
2689 	 * outer IP header, PIM header, PIM-Register header and the
2690 	 * inner IP header.
2691 	 * XXX: the outer IP header pkt size of a Register is not adjust to
2692 	 * reflect the fact that the inner multicast data is truncated.
2693 	 */
2694 	return (rip_input(&m, &off, proto));
2695 }
2696 
2697 static int
2698 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2699 {
2700 	struct mfc	*rt;
2701 	int		 error, i;
2702 
2703 	if (req->newptr)
2704 		return (EPERM);
2705 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2706 		return (0);
2707 	error = sysctl_wire_old_buffer(req, 0);
2708 	if (error)
2709 		return (error);
2710 
2711 	MRW_RLOCK();
2712 	for (i = 0; i < mfchashsize; i++) {
2713 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2714 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2715 			if (error)
2716 				goto out_locked;
2717 		}
2718 	}
2719 out_locked:
2720 	MRW_RUNLOCK();
2721 	return (error);
2722 }
2723 
2724 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2725     CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2726     "IPv4 Multicast Forwarding Table "
2727     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2728 
2729 static int
2730 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2731 {
2732 	int error, i;
2733 
2734 	if (req->newptr)
2735 		return (EPERM);
2736 	if (V_viftable == NULL)		/* XXX unlocked */
2737 		return (0);
2738 	error = sysctl_wire_old_buffer(req, MROUTE_VIF_SYSCTL_LEN * MAXVIFS);
2739 	if (error)
2740 		return (error);
2741 
2742 	MRW_RLOCK();
2743 	/* Copy out user-visible portion of vif entry. */
2744 	for (i = 0; i < MAXVIFS; i++) {
2745 		error = SYSCTL_OUT(req, &V_viftable[i], MROUTE_VIF_SYSCTL_LEN);
2746 		if (error)
2747 			break;
2748 	}
2749 	MRW_RUNLOCK();
2750 	return (error);
2751 }
2752 
2753 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2754     CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2755     sysctl_viflist, "S,vif[MAXVIFS]",
2756     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2757 
2758 static void
2759 vnet_mroute_init(const void *unused __unused)
2760 {
2761 
2762 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2763 
2764 	V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2765 	    M_MRTABLE, M_WAITOK|M_ZERO);
2766 
2767 	callout_init_rw(&V_expire_upcalls_ch, &mrouter_lock, 0);
2768 	callout_init_rw(&V_bw_upcalls_ch, &mrouter_lock, 0);
2769 
2770 	/* Prepare taskqueue */
2771 	V_task_queue = taskqueue_create_fast("ip_mroute_tskq", M_NOWAIT,
2772 		    taskqueue_thread_enqueue, &V_task_queue);
2773 	taskqueue_start_threads(&V_task_queue, 1, PI_NET, "ip_mroute_tskq task");
2774 }
2775 
2776 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2777 	NULL);
2778 
2779 static void
2780 vnet_mroute_uninit(const void *unused __unused)
2781 {
2782 
2783 	/* Taskqueue should be cancelled and drained before freeing */
2784 	taskqueue_free(V_task_queue);
2785 
2786 	free(V_viftable, M_MRTABLE);
2787 	free(V_nexpire, M_MRTABLE);
2788 	V_nexpire = NULL;
2789 }
2790 
2791 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2792     vnet_mroute_uninit, NULL);
2793 
2794 static int
2795 ip_mroute_modevent(module_t mod, int type, void *unused)
2796 {
2797 
2798 	switch (type) {
2799 	case MOD_LOAD:
2800 		MRW_LOCK_INIT();
2801 
2802 		if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2803 		    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2804 		if (if_detach_event_tag == NULL) {
2805 			printf("ip_mroute: unable to register "
2806 					"ifnet_departure_event handler\n");
2807 			MRW_LOCK_DESTROY();
2808 			return (EINVAL);
2809 		}
2810 
2811 		mfchashsize = MFCHASHSIZE;
2812 		if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2813 				!powerof2(mfchashsize)) {
2814 			printf("WARNING: %s not a power of 2; using default\n",
2815 					"net.inet.ip.mfchashsize");
2816 			mfchashsize = MFCHASHSIZE;
2817 		}
2818 
2819 		pim_squelch_wholepkt = 0;
2820 		TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2821 				&pim_squelch_wholepkt);
2822 
2823 		pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2824 		if (pim_encap_cookie == NULL) {
2825 			printf("ip_mroute: unable to attach pim encap\n");
2826 			MRW_LOCK_DESTROY();
2827 			return (EINVAL);
2828 		}
2829 
2830 		ip_mcast_src = X_ip_mcast_src;
2831 		ip_mforward = X_ip_mforward;
2832 		ip_mrouter_done = X_ip_mrouter_done;
2833 		ip_mrouter_get = X_ip_mrouter_get;
2834 		ip_mrouter_set = X_ip_mrouter_set;
2835 
2836 		ip_rsvp_force_done = X_ip_rsvp_force_done;
2837 		ip_rsvp_vif = X_ip_rsvp_vif;
2838 
2839 		legal_vif_num = X_legal_vif_num;
2840 		mrt_ioctl = X_mrt_ioctl;
2841 		rsvp_input_p = X_rsvp_input;
2842 		break;
2843 
2844 	case MOD_UNLOAD:
2845 		/*
2846 		 * Typically module unload happens after the user-level
2847 		 * process has shutdown the kernel services (the check
2848 		 * below insures someone can't just yank the module out
2849 		 * from under a running process).  But if the module is
2850 		 * just loaded and then unloaded w/o starting up a user
2851 		 * process we still need to cleanup.
2852 		 */
2853 		MRW_WLOCK();
2854 		if (ip_mrouter_cnt != 0) {
2855 			MRW_WUNLOCK();
2856 			return (EINVAL);
2857 		}
2858 		ip_mrouter_unloading = 1;
2859 		MRW_WUNLOCK();
2860 
2861 		EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2862 
2863 		if (pim_encap_cookie) {
2864 			ip_encap_detach(pim_encap_cookie);
2865 			pim_encap_cookie = NULL;
2866 		}
2867 
2868 		ip_mcast_src = NULL;
2869 		ip_mforward = NULL;
2870 		ip_mrouter_done = NULL;
2871 		ip_mrouter_get = NULL;
2872 		ip_mrouter_set = NULL;
2873 
2874 		ip_rsvp_force_done = NULL;
2875 		ip_rsvp_vif = NULL;
2876 
2877 		legal_vif_num = NULL;
2878 		mrt_ioctl = NULL;
2879 		rsvp_input_p = NULL;
2880 
2881 		MRW_LOCK_DESTROY();
2882 		break;
2883 
2884 	default:
2885 		return EOPNOTSUPP;
2886 	}
2887 	return 0;
2888 }
2889 
2890 static moduledata_t ip_mroutemod = {
2891 	"ip_mroute",
2892 	ip_mroute_modevent,
2893 	0
2894 };
2895 
2896 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2897