xref: /freebsd/sys/netinet/ip_mroute.c (revision 5c4aa6257210502c93ad65882a8a4842d984bae2)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989 Stephen Deering
5  * Copyright (c) 1992, 1993
6  *      The Regents of the University of California.  All rights reserved.
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Stephen Deering of Stanford University.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
36  */
37 
38 /*
39  * IP multicast forwarding procedures
40  *
41  * Written by David Waitzman, BBN Labs, August 1988.
42  * Modified by Steve Deering, Stanford, February 1989.
43  * Modified by Mark J. Steiglitz, Stanford, May, 1991
44  * Modified by Van Jacobson, LBL, January 1993
45  * Modified by Ajit Thyagarajan, PARC, August 1993
46  * Modified by Bill Fenner, PARC, April 1995
47  * Modified by Ahmed Helmy, SGI, June 1996
48  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
49  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
50  * Modified by Hitoshi Asaeda, WIDE, August 2000
51  * Modified by Pavlin Radoslavov, ICSI, October 2002
52  * Modified by Wojciech Macek, Semihalf, May 2021
53  *
54  * MROUTING Revision: 3.5
55  * and PIM-SMv2 and PIM-DM support, advanced API support,
56  * bandwidth metering and signaling
57  */
58 
59 /*
60  * TODO: Prefix functions with ipmf_.
61  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
62  * domain attachment (if_afdata) so we can track consumers of that service.
63  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
64  * move it to socket options.
65  * TODO: Cleanup LSRR removal further.
66  * TODO: Push RSVP stubs into raw_ip.c.
67  * TODO: Use bitstring.h for vif set.
68  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
69  * TODO: Sync ip6_mroute.c with this file.
70  */
71 
72 #include <sys/cdefs.h>
73 __FBSDID("$FreeBSD$");
74 
75 #include "opt_inet.h"
76 #include "opt_mrouting.h"
77 
78 #define _PIM_VT 1
79 
80 #include <sys/types.h>
81 #include <sys/param.h>
82 #include <sys/kernel.h>
83 #include <sys/stddef.h>
84 #include <sys/condvar.h>
85 #include <sys/eventhandler.h>
86 #include <sys/lock.h>
87 #include <sys/kthread.h>
88 #include <sys/ktr.h>
89 #include <sys/malloc.h>
90 #include <sys/mbuf.h>
91 #include <sys/module.h>
92 #include <sys/priv.h>
93 #include <sys/protosw.h>
94 #include <sys/signalvar.h>
95 #include <sys/socket.h>
96 #include <sys/socketvar.h>
97 #include <sys/sockio.h>
98 #include <sys/sx.h>
99 #include <sys/sysctl.h>
100 #include <sys/syslog.h>
101 #include <sys/systm.h>
102 #include <sys/taskqueue.h>
103 #include <sys/time.h>
104 #include <sys/counter.h>
105 #include <machine/atomic.h>
106 
107 #include <net/if.h>
108 #include <net/if_var.h>
109 #include <net/if_types.h>
110 #include <net/netisr.h>
111 #include <net/route.h>
112 #include <net/vnet.h>
113 
114 #include <netinet/in.h>
115 #include <netinet/igmp.h>
116 #include <netinet/in_systm.h>
117 #include <netinet/in_var.h>
118 #include <netinet/ip.h>
119 #include <netinet/ip_encap.h>
120 #include <netinet/ip_mroute.h>
121 #include <netinet/ip_var.h>
122 #include <netinet/ip_options.h>
123 #include <netinet/pim.h>
124 #include <netinet/pim_var.h>
125 #include <netinet/udp.h>
126 
127 #include <machine/in_cksum.h>
128 
129 #ifndef KTR_IPMF
130 #define KTR_IPMF KTR_INET
131 #endif
132 
133 #define		VIFI_INVALID	((vifi_t) -1)
134 
135 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
136 
137 /*
138  * Locking.  We use two locks: one for the virtual interface table and
139  * one for the forwarding table.  These locks may be nested in which case
140  * the VIF lock must always be taken first.  Note that each lock is used
141  * to cover not only the specific data structure but also related data
142  * structures.
143  */
144 
145 static struct rwlock mrouter_mtx;
146 #define	MRW_RLOCK()		rw_rlock(&mrouter_mtx)
147 #define	MRW_WLOCK()		rw_wlock(&mrouter_mtx)
148 #define	MRW_RUNLOCK()	rw_runlock(&mrouter_mtx)
149 #define	MRW_WUNLOCK()	rw_wunlock(&mrouter_mtx)
150 #define	MRW_UNLOCK()	rw_unlock(&mrouter_mtx)
151 #define	MRW_LOCK_ASSERT()	rw_assert(&mrouter_mtx, RA_LOCKED)
152 #define	MRW_WLOCK_ASSERT()	rw_assert(&mrouter_mtx, RA_WLOCKED)
153 #define	MRW_LOCK_TRY_UPGRADE()	rw_try_upgrade(&mrouter_mtx)
154 #define	MRW_WOWNED()	rw_wowned(&mrouter_mtx)
155 #define	MRW_LOCK_INIT()						\
156 	rw_init(&mrouter_mtx, "IPv4 multicast forwarding")
157 #define	MRW_LOCK_DESTROY()	rw_destroy(&mrouter_mtx)
158 
159 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
160 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
161 
162 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
163 VNET_PCPUSTAT_SYSINIT(mrtstat);
164 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
165 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
166     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
167     "netinet/ip_mroute.h)");
168 
169 VNET_DEFINE_STATIC(u_long, mfchash);
170 #define	V_mfchash		VNET(mfchash)
171 #define	MFCHASH(a, g)							\
172 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
173 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
174 #define	MFCHASHSIZE	256
175 
176 static u_long mfchashsize;			/* Hash size */
177 VNET_DEFINE_STATIC(u_char *, nexpire);		/* 0..mfchashsize-1 */
178 #define	V_nexpire		VNET(nexpire)
179 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
180 #define	V_mfchashtbl		VNET(mfchashtbl)
181 VNET_DEFINE_STATIC(struct taskqueue *, task_queue);
182 #define	V_task_queue		VNET(task_queue)
183 VNET_DEFINE_STATIC(struct task, task);
184 #define	V_task		VNET(task)
185 
186 VNET_DEFINE_STATIC(vifi_t, numvifs);
187 #define	V_numvifs		VNET(numvifs)
188 VNET_DEFINE_STATIC(struct vif *, viftable);
189 #define	V_viftable		VNET(viftable)
190 
191 static eventhandler_tag if_detach_event_tag = NULL;
192 
193 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
194 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
195 
196 VNET_DEFINE_STATIC(struct mtx, buf_ring_mtx);
197 #define	V_buf_ring_mtx	VNET(buf_ring_mtx)
198 
199 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
200 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
201 
202 /*
203  * Bandwidth meter variables and constants
204  */
205 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
206 
207 /*
208  * Pending upcalls are stored in a ring which is flushed when
209  * full, or periodically
210  */
211 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
212 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
213 VNET_DEFINE_STATIC(struct buf_ring *, bw_upcalls_ring);
214 #define	V_bw_upcalls_ring    	VNET(bw_upcalls_ring)
215 VNET_DEFINE_STATIC(struct mtx, bw_upcalls_ring_mtx);
216 #define	V_bw_upcalls_ring_mtx    	VNET(bw_upcalls_ring_mtx)
217 
218 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
219 
220 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
221 VNET_PCPUSTAT_SYSINIT(pimstat);
222 VNET_PCPUSTAT_SYSUNINIT(pimstat);
223 
224 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
225     "PIM");
226 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
227     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
228 
229 static u_long	pim_squelch_wholepkt = 0;
230 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
231     &pim_squelch_wholepkt, 0,
232     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
233 
234 static const struct encaptab *pim_encap_cookie;
235 static int pim_encapcheck(const struct mbuf *, int, int, void *);
236 static int pim_input(struct mbuf *, int, int, void *);
237 
238 extern int in_mcast_loop;
239 
240 static const struct encap_config ipv4_encap_cfg = {
241 	.proto = IPPROTO_PIM,
242 	.min_length = sizeof(struct ip) + PIM_MINLEN,
243 	.exact_match = 8,
244 	.check = pim_encapcheck,
245 	.input = pim_input
246 };
247 
248 /*
249  * Note: the PIM Register encapsulation adds the following in front of a
250  * data packet:
251  *
252  * struct pim_encap_hdr {
253  *    struct ip ip;
254  *    struct pim_encap_pimhdr  pim;
255  * }
256  *
257  */
258 
259 struct pim_encap_pimhdr {
260 	struct pim pim;
261 	uint32_t   flags;
262 };
263 #define		PIM_ENCAP_TTL	64
264 
265 static struct ip pim_encap_iphdr = {
266 #if BYTE_ORDER == LITTLE_ENDIAN
267 	sizeof(struct ip) >> 2,
268 	IPVERSION,
269 #else
270 	IPVERSION,
271 	sizeof(struct ip) >> 2,
272 #endif
273 	0,			/* tos */
274 	sizeof(struct ip),	/* total length */
275 	0,			/* id */
276 	0,			/* frag offset */
277 	PIM_ENCAP_TTL,
278 	IPPROTO_PIM,
279 	0,			/* checksum */
280 };
281 
282 static struct pim_encap_pimhdr pim_encap_pimhdr = {
283     {
284 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
285 	0,			/* reserved */
286 	0,			/* checksum */
287     },
288     0				/* flags */
289 };
290 
291 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
292 #define	V_reg_vif_num		VNET(reg_vif_num)
293 VNET_DEFINE_STATIC(struct ifnet *, multicast_register_if);
294 #define	V_multicast_register_if	VNET(multicast_register_if)
295 
296 /*
297  * Private variables.
298  */
299 
300 static u_long	X_ip_mcast_src(int);
301 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
302 		    struct ip_moptions *);
303 static int	X_ip_mrouter_done(void);
304 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
305 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
306 static int	X_legal_vif_num(int);
307 static int	X_mrt_ioctl(u_long, caddr_t, int);
308 
309 static int	add_bw_upcall(struct bw_upcall *);
310 static int	add_mfc(struct mfcctl2 *);
311 static int	add_vif(struct vifctl *);
312 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
313 static void	bw_meter_geq_receive_packet(struct bw_meter *, int,
314 		    struct timeval *);
315 static void	bw_upcalls_send(void);
316 static int	del_bw_upcall(struct bw_upcall *);
317 static int	del_mfc(struct mfcctl2 *);
318 static int	del_vif(vifi_t);
319 static int	del_vif_locked(vifi_t, struct ifnet **);
320 static void	expire_bw_upcalls_send(void *);
321 static void	expire_mfc(struct mfc *);
322 static void	expire_upcalls(void *);
323 static void	free_bw_list(struct bw_meter *);
324 static int	get_sg_cnt(struct sioc_sg_req *);
325 static int	get_vif_cnt(struct sioc_vif_req *);
326 static void	if_detached_event(void *, struct ifnet *);
327 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
328 static int	ip_mrouter_init(struct socket *, int);
329 static __inline struct mfc *
330 		mfc_find(struct in_addr *, struct in_addr *);
331 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
332 static struct mbuf *
333 		pim_register_prepare(struct ip *, struct mbuf *);
334 static int	pim_register_send(struct ip *, struct vif *,
335 		    struct mbuf *, struct mfc *);
336 static int	pim_register_send_rp(struct ip *, struct vif *,
337 		    struct mbuf *, struct mfc *);
338 static int	pim_register_send_upcall(struct ip *, struct vif *,
339 		    struct mbuf *, struct mfc *);
340 static void	send_packet(struct vif *, struct mbuf *);
341 static int	set_api_config(uint32_t *);
342 static int	set_assert(int);
343 static int	socket_send(struct socket *, struct mbuf *,
344 		    struct sockaddr_in *);
345 
346 /*
347  * Kernel multicast forwarding API capabilities and setup.
348  * If more API capabilities are added to the kernel, they should be
349  * recorded in `mrt_api_support'.
350  */
351 #define MRT_API_VERSION		0x0305
352 
353 static const int mrt_api_version = MRT_API_VERSION;
354 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
355 					 MRT_MFC_FLAGS_BORDER_VIF |
356 					 MRT_MFC_RP |
357 					 MRT_MFC_BW_UPCALL);
358 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
359 #define	V_mrt_api_config	VNET(mrt_api_config)
360 VNET_DEFINE_STATIC(int, pim_assert_enabled);
361 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
362 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
363 
364 /*
365  * Find a route for a given origin IP address and multicast group address.
366  * Statistics must be updated by the caller.
367  */
368 static __inline struct mfc *
369 mfc_find(struct in_addr *o, struct in_addr *g)
370 {
371 	struct mfc *rt;
372 
373 	/*
374 	 * Might be called both RLOCK and WLOCK.
375 	 * Check if any, it's caller responsibility
376 	 * to choose correct option.
377 	 */
378 	MRW_LOCK_ASSERT();
379 
380 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
381 		if (in_hosteq(rt->mfc_origin, *o) &&
382 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
383 		    buf_ring_empty(rt->mfc_stall_ring))
384 			break;
385 	}
386 
387 	return (rt);
388 }
389 
390 static __inline struct mfc *
391 mfc_alloc(void)
392 {
393 	struct mfc *rt;
394 	rt = (struct mfc*) malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT | M_ZERO);
395 	if (rt == NULL)
396 		return rt;
397 
398 	rt->mfc_stall_ring = buf_ring_alloc(MAX_UPQ, M_MRTABLE,
399 	    M_NOWAIT, &V_buf_ring_mtx);
400 	if (rt->mfc_stall_ring == NULL) {
401 		free(rt, M_MRTABLE);
402 		return NULL;
403 	}
404 
405 	return rt;
406 }
407 
408 /*
409  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
410  */
411 static int
412 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
413 {
414     int	error, optval;
415     vifi_t	vifi;
416     struct	vifctl vifc;
417     struct	mfcctl2 mfc;
418     struct	bw_upcall bw_upcall;
419     uint32_t	i;
420 
421     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
422 	return EPERM;
423 
424     error = 0;
425     switch (sopt->sopt_name) {
426     case MRT_INIT:
427 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
428 	if (error)
429 	    break;
430 	error = ip_mrouter_init(so, optval);
431 	break;
432 
433     case MRT_DONE:
434 	error = ip_mrouter_done();
435 	break;
436 
437     case MRT_ADD_VIF:
438 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
439 	if (error)
440 	    break;
441 	error = add_vif(&vifc);
442 	break;
443 
444     case MRT_DEL_VIF:
445 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
446 	if (error)
447 	    break;
448 	error = del_vif(vifi);
449 	break;
450 
451     case MRT_ADD_MFC:
452     case MRT_DEL_MFC:
453 	/*
454 	 * select data size depending on API version.
455 	 */
456 	if (sopt->sopt_name == MRT_ADD_MFC &&
457 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
458 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
459 				sizeof(struct mfcctl2));
460 	} else {
461 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
462 				sizeof(struct mfcctl));
463 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
464 			sizeof(mfc) - sizeof(struct mfcctl));
465 	}
466 	if (error)
467 	    break;
468 	if (sopt->sopt_name == MRT_ADD_MFC)
469 	    error = add_mfc(&mfc);
470 	else
471 	    error = del_mfc(&mfc);
472 	break;
473 
474     case MRT_ASSERT:
475 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
476 	if (error)
477 	    break;
478 	set_assert(optval);
479 	break;
480 
481     case MRT_API_CONFIG:
482 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
483 	if (!error)
484 	    error = set_api_config(&i);
485 	if (!error)
486 	    error = sooptcopyout(sopt, &i, sizeof i);
487 	break;
488 
489     case MRT_ADD_BW_UPCALL:
490     case MRT_DEL_BW_UPCALL:
491 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
492 				sizeof bw_upcall);
493 	if (error)
494 	    break;
495 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
496 	    error = add_bw_upcall(&bw_upcall);
497 	else
498 	    error = del_bw_upcall(&bw_upcall);
499 	break;
500 
501     default:
502 	error = EOPNOTSUPP;
503 	break;
504     }
505     return error;
506 }
507 
508 /*
509  * Handle MRT getsockopt commands
510  */
511 static int
512 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
513 {
514     int error;
515 
516     switch (sopt->sopt_name) {
517     case MRT_VERSION:
518 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
519 	break;
520 
521     case MRT_ASSERT:
522 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
523 	    sizeof V_pim_assert_enabled);
524 	break;
525 
526     case MRT_API_SUPPORT:
527 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
528 	break;
529 
530     case MRT_API_CONFIG:
531 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
532 	break;
533 
534     default:
535 	error = EOPNOTSUPP;
536 	break;
537     }
538     return error;
539 }
540 
541 /*
542  * Handle ioctl commands to obtain information from the cache
543  */
544 static int
545 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
546 {
547     int error = 0;
548 
549     /*
550      * Currently the only function calling this ioctl routine is rtioctl_fib().
551      * Typically, only root can create the raw socket in order to execute
552      * this ioctl method, however the request might be coming from a prison
553      */
554     error = priv_check(curthread, PRIV_NETINET_MROUTE);
555     if (error)
556 	return (error);
557     switch (cmd) {
558     case (SIOCGETVIFCNT):
559 	error = get_vif_cnt((struct sioc_vif_req *)data);
560 	break;
561 
562     case (SIOCGETSGCNT):
563 	error = get_sg_cnt((struct sioc_sg_req *)data);
564 	break;
565 
566     default:
567 	error = EINVAL;
568 	break;
569     }
570     return error;
571 }
572 
573 /*
574  * returns the packet, byte, rpf-failure count for the source group provided
575  */
576 static int
577 get_sg_cnt(struct sioc_sg_req *req)
578 {
579     struct mfc *rt;
580 
581     MRW_RLOCK();
582     rt = mfc_find(&req->src, &req->grp);
583     if (rt == NULL) {
584 	    MRW_RUNLOCK();
585 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
586 	return EADDRNOTAVAIL;
587     }
588     req->pktcnt = rt->mfc_pkt_cnt;
589     req->bytecnt = rt->mfc_byte_cnt;
590     req->wrong_if = rt->mfc_wrong_if;
591     MRW_RUNLOCK();
592     return 0;
593 }
594 
595 /*
596  * returns the input and output packet and byte counts on the vif provided
597  */
598 static int
599 get_vif_cnt(struct sioc_vif_req *req)
600 {
601     vifi_t vifi = req->vifi;
602 
603     MRW_RLOCK();
604     if (vifi >= V_numvifs) {
605 	MRW_RUNLOCK();
606 	return EINVAL;
607     }
608 
609     mtx_lock_spin(&V_viftable[vifi].v_spin);
610     req->icount = V_viftable[vifi].v_pkt_in;
611     req->ocount = V_viftable[vifi].v_pkt_out;
612     req->ibytes = V_viftable[vifi].v_bytes_in;
613     req->obytes = V_viftable[vifi].v_bytes_out;
614     mtx_unlock_spin(&V_viftable[vifi].v_spin);
615     MRW_RUNLOCK();
616 
617     return 0;
618 }
619 
620 static void
621 if_detached_event(void *arg __unused, struct ifnet *ifp)
622 {
623     vifi_t vifi;
624     u_long i, vifi_cnt = 0;
625     struct ifnet *free_ptr;
626 
627     MRW_WLOCK();
628 
629     if (V_ip_mrouter == NULL) {
630 	MRW_WUNLOCK();
631 	return;
632     }
633 
634     /*
635      * Tear down multicast forwarder state associated with this ifnet.
636      * 1. Walk the vif list, matching vifs against this ifnet.
637      * 2. Walk the multicast forwarding cache (mfc) looking for
638      *    inner matches with this vif's index.
639      * 3. Expire any matching multicast forwarding cache entries.
640      * 4. Free vif state. This should disable ALLMULTI on the interface.
641      */
642     for (vifi = 0; vifi < V_numvifs; vifi++) {
643 	if (V_viftable[vifi].v_ifp != ifp)
644 		continue;
645 	for (i = 0; i < mfchashsize; i++) {
646 		struct mfc *rt, *nrt;
647 
648 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
649 			if (rt->mfc_parent == vifi) {
650 				expire_mfc(rt);
651 			}
652 		}
653 	}
654 	del_vif_locked(vifi, &free_ptr);
655 	if (free_ptr != NULL)
656 		vifi_cnt++;
657     }
658 
659     MRW_WUNLOCK();
660 
661     /*
662      * Free IFP. We don't have to use free_ptr here as it is the same
663      * that ifp. Perform free as many times as required in case
664      * refcount is greater than 1.
665      */
666     for (i = 0; i < vifi_cnt; i++)
667 	    if_free(ifp);
668 }
669 
670 static void
671 ip_mrouter_upcall_thread(void *arg, int pending __unused)
672 {
673 	CURVNET_SET((struct vnet *) arg);
674 
675 	MRW_WLOCK();
676 	bw_upcalls_send();
677 	MRW_WUNLOCK();
678 
679 	CURVNET_RESTORE();
680 }
681 
682 /*
683  * Enable multicast forwarding.
684  */
685 static int
686 ip_mrouter_init(struct socket *so, int version)
687 {
688 
689     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
690         so->so_type, so->so_proto->pr_protocol);
691 
692     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
693 	return EOPNOTSUPP;
694 
695     if (version != 1)
696 	return ENOPROTOOPT;
697 
698     MRW_WLOCK();
699 
700     if (ip_mrouter_unloading) {
701 	MRW_WUNLOCK();
702 	return ENOPROTOOPT;
703     }
704 
705     if (V_ip_mrouter != NULL) {
706 	MRW_WUNLOCK();
707 	return EADDRINUSE;
708     }
709 
710     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
711 	HASH_NOWAIT);
712 
713     /* Create upcall ring */
714     mtx_init(&V_bw_upcalls_ring_mtx, "mroute upcall buf_ring mtx", NULL, MTX_DEF);
715     V_bw_upcalls_ring = buf_ring_alloc(BW_UPCALLS_MAX, M_MRTABLE,
716 	M_NOWAIT, &V_bw_upcalls_ring_mtx);
717     if (!V_bw_upcalls_ring) {
718 	MRW_WUNLOCK();
719 	return (ENOMEM);
720     }
721 
722     TASK_INIT(&V_task, 0, ip_mrouter_upcall_thread, curvnet);
723     taskqueue_cancel(V_task_queue, &V_task, NULL);
724     taskqueue_unblock(V_task_queue);
725 
726     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
727 	curvnet);
728     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
729 	curvnet);
730 
731     V_ip_mrouter = so;
732     atomic_add_int(&ip_mrouter_cnt, 1);
733 
734     /* This is a mutex required by buf_ring init, but not used internally */
735     mtx_init(&V_buf_ring_mtx, "mroute buf_ring mtx", NULL, MTX_DEF);
736 
737     MRW_WUNLOCK();
738 
739     CTR1(KTR_IPMF, "%s: done", __func__);
740 
741     return 0;
742 }
743 
744 /*
745  * Disable multicast forwarding.
746  */
747 static int
748 X_ip_mrouter_done(void)
749 {
750     struct ifnet *ifp;
751     u_long i;
752     vifi_t vifi;
753     struct bw_upcall *bu;
754 
755     if (V_ip_mrouter == NULL)
756 	return (EINVAL);
757 
758     /*
759      * Detach/disable hooks to the reset of the system.
760      */
761     V_ip_mrouter = NULL;
762     atomic_subtract_int(&ip_mrouter_cnt, 1);
763     V_mrt_api_config = 0;
764 
765     /*
766      * Wait for all epoch sections to complete to ensure
767      * V_ip_mrouter = NULL is visible to others.
768      */
769     epoch_wait_preempt(net_epoch_preempt);
770 
771     /* Stop and drain task queue */
772     taskqueue_block(V_task_queue);
773     while (taskqueue_cancel(V_task_queue, &V_task, NULL)) {
774     	taskqueue_drain(V_task_queue, &V_task);
775     }
776 
777     MRW_WLOCK();
778     taskqueue_cancel(V_task_queue, &V_task, NULL);
779 
780     /* Destroy upcall ring */
781     while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
782 	free(bu, M_MRTABLE);
783     }
784     buf_ring_free(V_bw_upcalls_ring, M_MRTABLE);
785     mtx_destroy(&V_bw_upcalls_ring_mtx);
786 
787     /*
788      * For each phyint in use, disable promiscuous reception of all IP
789      * multicasts.
790      */
791     for (vifi = 0; vifi < V_numvifs; vifi++) {
792 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
793 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
794 	    ifp = V_viftable[vifi].v_ifp;
795 	    if_allmulti(ifp, 0);
796 	}
797     }
798     bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
799     V_numvifs = 0;
800     V_pim_assert_enabled = 0;
801 
802     callout_stop(&V_expire_upcalls_ch);
803     callout_stop(&V_bw_upcalls_ch);
804 
805     /*
806      * Free all multicast forwarding cache entries.
807      * Do not use hashdestroy(), as we must perform other cleanup.
808      */
809     for (i = 0; i < mfchashsize; i++) {
810 	struct mfc *rt, *nrt;
811 
812 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
813 		expire_mfc(rt);
814 	}
815     }
816     free(V_mfchashtbl, M_MRTABLE);
817     V_mfchashtbl = NULL;
818 
819     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
820 
821     V_reg_vif_num = VIFI_INVALID;
822 
823     mtx_destroy(&V_buf_ring_mtx);
824 
825     MRW_WUNLOCK();
826 
827     CTR1(KTR_IPMF, "%s: done", __func__);
828 
829     return 0;
830 }
831 
832 /*
833  * Set PIM assert processing global
834  */
835 static int
836 set_assert(int i)
837 {
838     if ((i != 1) && (i != 0))
839 	return EINVAL;
840 
841     V_pim_assert_enabled = i;
842 
843     return 0;
844 }
845 
846 /*
847  * Configure API capabilities
848  */
849 int
850 set_api_config(uint32_t *apival)
851 {
852     u_long i;
853 
854     /*
855      * We can set the API capabilities only if it is the first operation
856      * after MRT_INIT. I.e.:
857      *  - there are no vifs installed
858      *  - pim_assert is not enabled
859      *  - the MFC table is empty
860      */
861     if (V_numvifs > 0) {
862 	*apival = 0;
863 	return EPERM;
864     }
865     if (V_pim_assert_enabled) {
866 	*apival = 0;
867 	return EPERM;
868     }
869 
870     MRW_RLOCK();
871 
872     for (i = 0; i < mfchashsize; i++) {
873 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
874 	    MRW_RUNLOCK();
875 	    *apival = 0;
876 	    return EPERM;
877 	}
878     }
879 
880     MRW_RUNLOCK();
881 
882     V_mrt_api_config = *apival & mrt_api_support;
883     *apival = V_mrt_api_config;
884 
885     return 0;
886 }
887 
888 /*
889  * Add a vif to the vif table
890  */
891 static int
892 add_vif(struct vifctl *vifcp)
893 {
894     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
895     struct sockaddr_in sin = {sizeof sin, AF_INET};
896     struct ifaddr *ifa;
897     struct ifnet *ifp;
898     int error;
899 
900 
901     if (vifcp->vifc_vifi >= MAXVIFS)
902 	return EINVAL;
903     /* rate limiting is no longer supported by this code */
904     if (vifcp->vifc_rate_limit != 0) {
905 	log(LOG_ERR, "rate limiting is no longer supported\n");
906 	return EINVAL;
907     }
908 
909     if (in_nullhost(vifcp->vifc_lcl_addr))
910 	return EADDRNOTAVAIL;
911 
912     /* Find the interface with an address in AF_INET family */
913     if (vifcp->vifc_flags & VIFF_REGISTER) {
914 	/*
915 	 * XXX: Because VIFF_REGISTER does not really need a valid
916 	 * local interface (e.g. it could be 127.0.0.2), we don't
917 	 * check its address.
918 	 */
919 	ifp = NULL;
920     } else {
921 	struct epoch_tracker et;
922 
923 	sin.sin_addr = vifcp->vifc_lcl_addr;
924 	NET_EPOCH_ENTER(et);
925 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
926 	if (ifa == NULL) {
927 	    NET_EPOCH_EXIT(et);
928 	    return EADDRNOTAVAIL;
929 	}
930 	ifp = ifa->ifa_ifp;
931 	/* XXX FIXME we need to take a ref on ifp and cleanup properly! */
932 	NET_EPOCH_EXIT(et);
933     }
934 
935     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
936 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
937 	return EOPNOTSUPP;
938     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
939 	ifp = V_multicast_register_if = if_alloc(IFT_LOOP);
940 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
941 	if (V_reg_vif_num == VIFI_INVALID) {
942 	    if_initname(V_multicast_register_if, "register_vif", 0);
943 	    V_reg_vif_num = vifcp->vifc_vifi;
944 	}
945     } else {		/* Make sure the interface supports multicast */
946 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
947 	    return EOPNOTSUPP;
948 
949 	/* Enable promiscuous reception of all IP multicasts from the if */
950 	error = if_allmulti(ifp, 1);
951 	if (error)
952 	    return error;
953     }
954 
955     MRW_WLOCK();
956 
957     if (!in_nullhost(vifp->v_lcl_addr)) {
958 	if (ifp)
959 		V_multicast_register_if = NULL;
960 	MRW_WUNLOCK();
961 	if (ifp)
962 		if_free(ifp);
963 	return EADDRINUSE;
964     }
965 
966     vifp->v_flags     = vifcp->vifc_flags;
967     vifp->v_threshold = vifcp->vifc_threshold;
968     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
969     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
970     vifp->v_ifp       = ifp;
971     /* initialize per vif pkt counters */
972     vifp->v_pkt_in    = 0;
973     vifp->v_pkt_out   = 0;
974     vifp->v_bytes_in  = 0;
975     vifp->v_bytes_out = 0;
976     sprintf(vifp->v_spin_name, "BM[%d] spin", vifcp->vifc_vifi);
977     mtx_init(&vifp->v_spin, vifp->v_spin_name, NULL, MTX_SPIN);
978 
979     /* Adjust numvifs up if the vifi is higher than numvifs */
980     if (V_numvifs <= vifcp->vifc_vifi)
981 	V_numvifs = vifcp->vifc_vifi + 1;
982 
983     MRW_WUNLOCK();
984 
985     CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
986 	(int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
987 	(int)vifcp->vifc_threshold);
988 
989     return 0;
990 }
991 
992 /*
993  * Delete a vif from the vif table
994  */
995 static int
996 del_vif_locked(vifi_t vifi, struct ifnet **ifp_free)
997 {
998     struct vif *vifp;
999 
1000     *ifp_free = NULL;
1001 
1002     MRW_WLOCK_ASSERT();
1003 
1004     if (vifi >= V_numvifs) {
1005 	return EINVAL;
1006     }
1007     vifp = &V_viftable[vifi];
1008     if (in_nullhost(vifp->v_lcl_addr)) {
1009 	return EADDRNOTAVAIL;
1010     }
1011 
1012     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1013 	if_allmulti(vifp->v_ifp, 0);
1014 
1015     if (vifp->v_flags & VIFF_REGISTER) {
1016 	V_reg_vif_num = VIFI_INVALID;
1017 	if (vifp->v_ifp) {
1018 	    if (vifp->v_ifp == V_multicast_register_if)
1019 	        V_multicast_register_if = NULL;
1020 	    *ifp_free = vifp->v_ifp;
1021 	}
1022     }
1023 
1024     mtx_destroy(&vifp->v_spin);
1025 
1026     bzero((caddr_t)vifp, sizeof (*vifp));
1027 
1028     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
1029 
1030     /* Adjust numvifs down */
1031     for (vifi = V_numvifs; vifi > 0; vifi--)
1032 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
1033 	    break;
1034     V_numvifs = vifi;
1035 
1036     return 0;
1037 }
1038 
1039 static int
1040 del_vif(vifi_t vifi)
1041 {
1042     int cc;
1043     struct ifnet *free_ptr;
1044 
1045     MRW_WLOCK();
1046     cc = del_vif_locked(vifi, &free_ptr);
1047     MRW_WUNLOCK();
1048 
1049     if (free_ptr)
1050 	    if_free(free_ptr);
1051 
1052     return cc;
1053 }
1054 
1055 /*
1056  * update an mfc entry without resetting counters and S,G addresses.
1057  */
1058 static void
1059 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1060 {
1061     int i;
1062 
1063     rt->mfc_parent = mfccp->mfcc_parent;
1064     for (i = 0; i < V_numvifs; i++) {
1065 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1066 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1067 	    MRT_MFC_FLAGS_ALL;
1068     }
1069     /* set the RP address */
1070     if (V_mrt_api_config & MRT_MFC_RP)
1071 	rt->mfc_rp = mfccp->mfcc_rp;
1072     else
1073 	rt->mfc_rp.s_addr = INADDR_ANY;
1074 }
1075 
1076 /*
1077  * fully initialize an mfc entry from the parameter.
1078  */
1079 static void
1080 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1081 {
1082     rt->mfc_origin     = mfccp->mfcc_origin;
1083     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1084 
1085     update_mfc_params(rt, mfccp);
1086 
1087     /* initialize pkt counters per src-grp */
1088     rt->mfc_pkt_cnt    = 0;
1089     rt->mfc_byte_cnt   = 0;
1090     rt->mfc_wrong_if   = 0;
1091     timevalclear(&rt->mfc_last_assert);
1092 }
1093 
1094 static void
1095 expire_mfc(struct mfc *rt)
1096 {
1097 	struct rtdetq *rte;
1098 
1099 	MRW_WLOCK_ASSERT();
1100 
1101 	free_bw_list(rt->mfc_bw_meter_leq);
1102 	free_bw_list(rt->mfc_bw_meter_geq);
1103 
1104 	while (!buf_ring_empty(rt->mfc_stall_ring)) {
1105 		rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1106 		if (rte) {
1107 			m_freem(rte->m);
1108 			free(rte, M_MRTABLE);
1109 		}
1110 	}
1111 	buf_ring_free(rt->mfc_stall_ring, M_MRTABLE);
1112 
1113 	LIST_REMOVE(rt, mfc_hash);
1114 	free(rt, M_MRTABLE);
1115 }
1116 
1117 /*
1118  * Add an mfc entry
1119  */
1120 static int
1121 add_mfc(struct mfcctl2 *mfccp)
1122 {
1123     struct mfc *rt;
1124     struct rtdetq *rte;
1125     u_long hash = 0;
1126     u_short nstl;
1127     struct epoch_tracker et;
1128 
1129     MRW_WLOCK();
1130     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1131 
1132     /* If an entry already exists, just update the fields */
1133     if (rt) {
1134 	CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1135 	    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1136 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1137 	    mfccp->mfcc_parent);
1138 	update_mfc_params(rt, mfccp);
1139 	MRW_WUNLOCK();
1140 	return (0);
1141     }
1142 
1143     /*
1144      * Find the entry for which the upcall was made and update
1145      */
1146     nstl = 0;
1147     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1148     NET_EPOCH_ENTER(et);
1149     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1150 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1151 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1152 	    !buf_ring_empty(rt->mfc_stall_ring)) {
1153 		CTR5(KTR_IPMF,
1154 		    "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1155 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1156 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1157 		    mfccp->mfcc_parent,
1158 		    rt->mfc_stall_ring);
1159 		if (nstl++)
1160 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1161 
1162 		init_mfc_params(rt, mfccp);
1163 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1164 		V_nexpire[hash]--;
1165 
1166 		/* Free queued packets, but attempt to forward them first. */
1167 		while (!buf_ring_empty(rt->mfc_stall_ring)) {
1168 			rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1169 			if (rte->ifp != NULL)
1170 				ip_mdq(rte->m, rte->ifp, rt, -1);
1171 			m_freem(rte->m);
1172 			free(rte, M_MRTABLE);
1173 		}
1174 	}
1175     }
1176     NET_EPOCH_EXIT(et);
1177 
1178     /*
1179      * It is possible that an entry is being inserted without an upcall
1180      */
1181     if (nstl == 0) {
1182 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1183 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1184 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1185 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1186 			init_mfc_params(rt, mfccp);
1187 			if (rt->mfc_expire)
1188 			    V_nexpire[hash]--;
1189 			rt->mfc_expire = 0;
1190 			break; /* XXX */
1191 		}
1192 	}
1193 
1194 	if (rt == NULL) {		/* no upcall, so make a new entry */
1195 	    rt = mfc_alloc();
1196 	    if (rt == NULL) {
1197 		MRW_WUNLOCK();
1198 		return (ENOBUFS);
1199 	    }
1200 
1201 	    init_mfc_params(rt, mfccp);
1202 
1203 	    rt->mfc_expire     = 0;
1204 	    rt->mfc_bw_meter_leq = NULL;
1205 	    rt->mfc_bw_meter_geq = NULL;
1206 
1207 	    /* insert new entry at head of hash chain */
1208 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1209 	}
1210     }
1211 
1212     MRW_WUNLOCK();
1213 
1214     return (0);
1215 }
1216 
1217 /*
1218  * Delete an mfc entry
1219  */
1220 static int
1221 del_mfc(struct mfcctl2 *mfccp)
1222 {
1223     struct in_addr	origin;
1224     struct in_addr	mcastgrp;
1225     struct mfc		*rt;
1226 
1227     origin = mfccp->mfcc_origin;
1228     mcastgrp = mfccp->mfcc_mcastgrp;
1229 
1230     CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1231 	ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1232 
1233     MRW_WLOCK();
1234 
1235     rt = mfc_find(&origin, &mcastgrp);
1236     if (rt == NULL) {
1237 	MRW_WUNLOCK();
1238 	return EADDRNOTAVAIL;
1239     }
1240 
1241     /*
1242      * free the bw_meter entries
1243      */
1244     free_bw_list(rt->mfc_bw_meter_leq);
1245     rt->mfc_bw_meter_leq = NULL;
1246     free_bw_list(rt->mfc_bw_meter_geq);
1247     rt->mfc_bw_meter_geq = NULL;
1248 
1249     LIST_REMOVE(rt, mfc_hash);
1250     free(rt, M_MRTABLE);
1251 
1252     MRW_WUNLOCK();
1253 
1254     return (0);
1255 }
1256 
1257 /*
1258  * Send a message to the routing daemon on the multicast routing socket.
1259  */
1260 static int
1261 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1262 {
1263     if (s) {
1264 	SOCKBUF_LOCK(&s->so_rcv);
1265 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1266 	    NULL) != 0) {
1267 	    sorwakeup_locked(s);
1268 	    return 0;
1269 	}
1270 	soroverflow_locked(s);
1271     }
1272     m_freem(mm);
1273     return -1;
1274 }
1275 
1276 /*
1277  * IP multicast forwarding function. This function assumes that the packet
1278  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1279  * pointed to by "ifp", and the packet is to be relayed to other networks
1280  * that have members of the packet's destination IP multicast group.
1281  *
1282  * The packet is returned unscathed to the caller, unless it is
1283  * erroneous, in which case a non-zero return value tells the caller to
1284  * discard it.
1285  */
1286 
1287 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1288 
1289 static int
1290 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1291     struct ip_moptions *imo)
1292 {
1293 	struct mfc *rt;
1294 	int error;
1295 	vifi_t vifi;
1296 	struct mbuf *mb0;
1297 	struct rtdetq *rte;
1298 	u_long hash;
1299 	int hlen;
1300 
1301 	CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1302 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1303 
1304 	if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1305 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1306 		/*
1307 		 * Packet arrived via a physical interface or
1308 		 * an encapsulated tunnel or a register_vif.
1309 		 */
1310 	} else {
1311 		/*
1312 		 * Packet arrived through a source-route tunnel.
1313 		 * Source-route tunnels are no longer supported.
1314 		 */
1315 		return (1);
1316 	}
1317 
1318 	/*
1319 	 * BEGIN: MCAST ROUTING HOT PATH
1320 	 */
1321 	MRW_RLOCK();
1322 	if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1323 		if (ip->ip_ttl < MAXTTL)
1324 			ip->ip_ttl++; /* compensate for -1 in *_send routines */
1325 		error = ip_mdq(m, ifp, NULL, vifi);
1326 		MRW_RUNLOCK();
1327 		return error;
1328 	}
1329 
1330 	/*
1331 	 * Don't forward a packet with time-to-live of zero or one,
1332 	 * or a packet destined to a local-only group.
1333 	 */
1334 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1335 		MRW_RUNLOCK();
1336 		return 0;
1337 	}
1338 
1339 	mfc_find_retry:
1340 	/*
1341 	 * Determine forwarding vifs from the forwarding cache table
1342 	 */
1343 	MRTSTAT_INC(mrts_mfc_lookups);
1344 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1345 
1346 	/* Entry exists, so forward if necessary */
1347 	if (rt != NULL) {
1348 		error = ip_mdq(m, ifp, rt, -1);
1349 		/* Generic unlock here as we might release R or W lock */
1350 		MRW_UNLOCK();
1351 		return error;
1352 	}
1353 
1354 	/*
1355 	 * END: MCAST ROUTING HOT PATH
1356 	 */
1357 
1358 	/* Further processing must be done with WLOCK taken */
1359 	if ((MRW_WOWNED() == 0) && (MRW_LOCK_TRY_UPGRADE() == 0)) {
1360 		MRW_RUNLOCK();
1361 		MRW_WLOCK();
1362 		goto mfc_find_retry;
1363 	}
1364 
1365 	/*
1366 	 * If we don't have a route for packet's origin,
1367 	 * Make a copy of the packet & send message to routing daemon
1368 	 */
1369 	hlen = ip->ip_hl << 2;
1370 
1371 	MRTSTAT_INC(mrts_mfc_misses);
1372 	MRTSTAT_INC(mrts_no_route);
1373 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1374 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1375 
1376 	/*
1377 	 * Allocate mbufs early so that we don't do extra work if we are
1378 	 * just going to fail anyway.  Make sure to pullup the header so
1379 	 * that other people can't step on it.
1380 	 */
1381 	rte = (struct rtdetq*) malloc((sizeof *rte), M_MRTABLE,
1382 	    M_NOWAIT|M_ZERO);
1383 	if (rte == NULL) {
1384 		MRW_WUNLOCK();
1385 		return ENOBUFS;
1386 	}
1387 
1388 	mb0 = m_copypacket(m, M_NOWAIT);
1389 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1390 		mb0 = m_pullup(mb0, hlen);
1391 	if (mb0 == NULL) {
1392 		free(rte, M_MRTABLE);
1393 		MRW_WUNLOCK();
1394 		return ENOBUFS;
1395 	}
1396 
1397 	/* is there an upcall waiting for this flow ? */
1398 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1399 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash)
1400 	{
1401 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1402 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1403 		    !buf_ring_empty(rt->mfc_stall_ring))
1404 			break;
1405 	}
1406 
1407 	if (rt == NULL) {
1408 		int i;
1409 		struct igmpmsg *im;
1410 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1411 		struct mbuf *mm;
1412 
1413 		/*
1414 		 * Locate the vifi for the incoming interface for this packet.
1415 		 * If none found, drop packet.
1416 		 */
1417 		for (vifi = 0; vifi < V_numvifs &&
1418 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1419 			;
1420 		if (vifi >= V_numvifs) /* vif not found, drop packet */
1421 			goto non_fatal;
1422 
1423 		/* no upcall, so make a new entry */
1424 		rt = mfc_alloc();
1425 		if (rt == NULL)
1426 			goto fail;
1427 
1428 		/* Make a copy of the header to send to the user level process */
1429 		mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1430 		if (mm == NULL)
1431 			goto fail1;
1432 
1433 		/*
1434 		 * Send message to routing daemon to install
1435 		 * a route into the kernel table
1436 		 */
1437 
1438 		im = mtod(mm, struct igmpmsg*);
1439 		im->im_msgtype = IGMPMSG_NOCACHE;
1440 		im->im_mbz = 0;
1441 		im->im_vif = vifi;
1442 
1443 		MRTSTAT_INC(mrts_upcalls);
1444 
1445 		k_igmpsrc.sin_addr = ip->ip_src;
1446 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1447 			CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1448 			MRTSTAT_INC(mrts_upq_sockfull);
1449 			fail1: free(rt, M_MRTABLE);
1450 			fail: free(rte, M_MRTABLE);
1451 			m_freem(mb0);
1452 			MRW_WUNLOCK();
1453 			return ENOBUFS;
1454 		}
1455 
1456 		/* insert new entry at head of hash chain */
1457 		rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1458 		rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1459 		rt->mfc_expire = UPCALL_EXPIRE;
1460 		V_nexpire[hash]++;
1461 		for (i = 0; i < V_numvifs; i++) {
1462 			rt->mfc_ttls[i] = 0;
1463 			rt->mfc_flags[i] = 0;
1464 		}
1465 		rt->mfc_parent = -1;
1466 
1467 		/* clear the RP address */
1468 		rt->mfc_rp.s_addr = INADDR_ANY;
1469 		rt->mfc_bw_meter_leq = NULL;
1470 		rt->mfc_bw_meter_geq = NULL;
1471 
1472 		/* initialize pkt counters per src-grp */
1473 		rt->mfc_pkt_cnt = 0;
1474 		rt->mfc_byte_cnt = 0;
1475 		rt->mfc_wrong_if = 0;
1476 		timevalclear(&rt->mfc_last_assert);
1477 
1478 		buf_ring_enqueue(rt->mfc_stall_ring, rte);
1479 
1480 		/* Add RT to hashtable as it didn't exist before */
1481 		LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1482 	} else {
1483 		/* determine if queue has overflowed */
1484 		if (buf_ring_full(rt->mfc_stall_ring)) {
1485 			MRTSTAT_INC(mrts_upq_ovflw);
1486 			non_fatal: free(rte, M_MRTABLE);
1487 			m_freem(mb0);
1488 			MRW_WUNLOCK();
1489 			return (0);
1490 		}
1491 
1492 		buf_ring_enqueue(rt->mfc_stall_ring, rte);
1493 	}
1494 
1495 	rte->m = mb0;
1496 	rte->ifp = ifp;
1497 
1498 	MRW_WUNLOCK();
1499 
1500 	return 0;
1501 }
1502 
1503 /*
1504  * Clean up the cache entry if upcall is not serviced
1505  */
1506 static void
1507 expire_upcalls(void *arg)
1508 {
1509     u_long i;
1510 
1511     CURVNET_SET((struct vnet *) arg);
1512 
1513     /*This callout is always run with MRW_WLOCK taken. */
1514 
1515     for (i = 0; i < mfchashsize; i++) {
1516 	struct mfc *rt, *nrt;
1517 
1518 	if (V_nexpire[i] == 0)
1519 	    continue;
1520 
1521 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1522 		if (buf_ring_empty(rt->mfc_stall_ring))
1523 			continue;
1524 
1525 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1526 			continue;
1527 
1528 		MRTSTAT_INC(mrts_cache_cleanups);
1529 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1530 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1531 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1532 
1533 		expire_mfc(rt);
1534 	    }
1535     }
1536 
1537     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1538 	curvnet);
1539 
1540     CURVNET_RESTORE();
1541 }
1542 
1543 /*
1544  * Packet forwarding routine once entry in the cache is made
1545  */
1546 static int
1547 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1548 {
1549     struct ip  *ip = mtod(m, struct ip *);
1550     vifi_t vifi;
1551     int plen = ntohs(ip->ip_len);
1552 
1553     MRW_LOCK_ASSERT();
1554     NET_EPOCH_ASSERT();
1555 
1556     /*
1557      * If xmt_vif is not -1, send on only the requested vif.
1558      *
1559      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1560      */
1561     if (xmt_vif < V_numvifs) {
1562 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1563 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1564 	else
1565 		phyint_send(ip, V_viftable + xmt_vif, m);
1566 	return 1;
1567     }
1568 
1569     /*
1570      * Don't forward if it didn't arrive from the parent vif for its origin.
1571      */
1572     vifi = rt->mfc_parent;
1573     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1574 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1575 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1576 	MRTSTAT_INC(mrts_wrong_if);
1577 	++rt->mfc_wrong_if;
1578 	/*
1579 	 * If we are doing PIM assert processing, send a message
1580 	 * to the routing daemon.
1581 	 *
1582 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1583 	 * can complete the SPT switch, regardless of the type
1584 	 * of the iif (broadcast media, GRE tunnel, etc).
1585 	 */
1586 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1587 	    V_viftable[vifi].v_ifp) {
1588 	    if (ifp == V_multicast_register_if)
1589 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1590 
1591 	    /* Get vifi for the incoming packet */
1592 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1593 		vifi++)
1594 		;
1595 	    if (vifi >= V_numvifs)
1596 		return 0;	/* The iif is not found: ignore the packet. */
1597 
1598 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1599 		return 0;	/* WRONGVIF disabled: ignore the packet */
1600 
1601 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1602 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1603 		struct igmpmsg *im;
1604 		int hlen = ip->ip_hl << 2;
1605 		struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1606 
1607 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1608 		    mm = m_pullup(mm, hlen);
1609 		if (mm == NULL)
1610 		    return ENOBUFS;
1611 
1612 		im = mtod(mm, struct igmpmsg *);
1613 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1614 		im->im_mbz		= 0;
1615 		im->im_vif		= vifi;
1616 
1617 		MRTSTAT_INC(mrts_upcalls);
1618 
1619 		k_igmpsrc.sin_addr = im->im_src;
1620 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1621 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1622 		    MRTSTAT_INC(mrts_upq_sockfull);
1623 		    return ENOBUFS;
1624 		}
1625 	    }
1626 	}
1627 	return 0;
1628     }
1629 
1630     /* If I sourced this packet, it counts as output, else it was input. */
1631     mtx_lock_spin(&V_viftable[vifi].v_spin);
1632     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1633 	V_viftable[vifi].v_pkt_out++;
1634 	V_viftable[vifi].v_bytes_out += plen;
1635     } else {
1636 	V_viftable[vifi].v_pkt_in++;
1637 	V_viftable[vifi].v_bytes_in += plen;
1638     }
1639     mtx_unlock_spin(&V_viftable[vifi].v_spin);
1640 
1641     rt->mfc_pkt_cnt++;
1642     rt->mfc_byte_cnt += plen;
1643 
1644     /*
1645      * For each vif, decide if a copy of the packet should be forwarded.
1646      * Forward if:
1647      *		- the ttl exceeds the vif's threshold
1648      *		- there are group members downstream on interface
1649      */
1650     for (vifi = 0; vifi < V_numvifs; vifi++)
1651 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1652 	    V_viftable[vifi].v_pkt_out++;
1653 	    V_viftable[vifi].v_bytes_out += plen;
1654 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1655 		pim_register_send(ip, V_viftable + vifi, m, rt);
1656 	    else
1657 		phyint_send(ip, V_viftable + vifi, m);
1658 	}
1659 
1660     /*
1661      * Perform upcall-related bw measuring.
1662      */
1663     if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1664 	struct bw_meter *x;
1665 	struct timeval now;
1666 
1667 	microtime(&now);
1668 	/* Process meters for Greater-or-EQual case */
1669 	for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1670 		bw_meter_geq_receive_packet(x, plen, &now);
1671 
1672 	/* Process meters for Lower-or-EQual case */
1673 	for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1674 		/*
1675 		 * Record that a packet is received.
1676 		 * Spin lock has to be taken as callout context
1677 		 * (expire_bw_meter_leq) might modify these fields
1678 		 * as well
1679 		 */
1680 		mtx_lock_spin(&x->bm_spin);
1681 		x->bm_measured.b_packets++;
1682 		x->bm_measured.b_bytes += plen;
1683 		mtx_unlock_spin(&x->bm_spin);
1684 	}
1685     }
1686 
1687     return 0;
1688 }
1689 
1690 /*
1691  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1692  */
1693 static int
1694 X_legal_vif_num(int vif)
1695 {
1696 	int ret;
1697 
1698 	ret = 0;
1699 	if (vif < 0)
1700 		return (ret);
1701 
1702 	MRW_RLOCK();
1703 	if (vif < V_numvifs)
1704 		ret = 1;
1705 	MRW_RUNLOCK();
1706 
1707 	return (ret);
1708 }
1709 
1710 /*
1711  * Return the local address used by this vif
1712  */
1713 static u_long
1714 X_ip_mcast_src(int vifi)
1715 {
1716 	in_addr_t addr;
1717 
1718 	addr = INADDR_ANY;
1719 	if (vifi < 0)
1720 		return (addr);
1721 
1722 	MRW_RLOCK();
1723 	if (vifi < V_numvifs)
1724 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1725 	MRW_RUNLOCK();
1726 
1727 	return (addr);
1728 }
1729 
1730 static void
1731 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1732 {
1733     struct mbuf *mb_copy;
1734     int hlen = ip->ip_hl << 2;
1735 
1736     MRW_LOCK_ASSERT();
1737 
1738     /*
1739      * Make a new reference to the packet; make sure that
1740      * the IP header is actually copied, not just referenced,
1741      * so that ip_output() only scribbles on the copy.
1742      */
1743     mb_copy = m_copypacket(m, M_NOWAIT);
1744     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1745 	mb_copy = m_pullup(mb_copy, hlen);
1746     if (mb_copy == NULL)
1747 	return;
1748 
1749     send_packet(vifp, mb_copy);
1750 }
1751 
1752 static void
1753 send_packet(struct vif *vifp, struct mbuf *m)
1754 {
1755 	struct ip_moptions imo;
1756 	int error __unused;
1757 
1758 	MRW_LOCK_ASSERT();
1759 	NET_EPOCH_ASSERT();
1760 
1761 	imo.imo_multicast_ifp  = vifp->v_ifp;
1762 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1763 	imo.imo_multicast_loop = !!in_mcast_loop;
1764 	imo.imo_multicast_vif  = -1;
1765 	STAILQ_INIT(&imo.imo_head);
1766 
1767 	/*
1768 	 * Re-entrancy should not be a problem here, because
1769 	 * the packets that we send out and are looped back at us
1770 	 * should get rejected because they appear to come from
1771 	 * the loopback interface, thus preventing looping.
1772 	 */
1773 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1774 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1775 	    (ptrdiff_t)(vifp - V_viftable), error);
1776 }
1777 
1778 /*
1779  * Stubs for old RSVP socket shim implementation.
1780  */
1781 
1782 static int
1783 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1784 {
1785 
1786 	return (EOPNOTSUPP);
1787 }
1788 
1789 static void
1790 X_ip_rsvp_force_done(struct socket *so __unused)
1791 {
1792 
1793 }
1794 
1795 static int
1796 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1797 {
1798 	struct mbuf *m;
1799 
1800 	m = *mp;
1801 	*mp = NULL;
1802 	if (!V_rsvp_on)
1803 		m_freem(m);
1804 	return (IPPROTO_DONE);
1805 }
1806 
1807 /*
1808  * Code for bandwidth monitors
1809  */
1810 
1811 /*
1812  * Define common interface for timeval-related methods
1813  */
1814 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1815 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1816 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1817 
1818 static uint32_t
1819 compute_bw_meter_flags(struct bw_upcall *req)
1820 {
1821     uint32_t flags = 0;
1822 
1823     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1824 	flags |= BW_METER_UNIT_PACKETS;
1825     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1826 	flags |= BW_METER_UNIT_BYTES;
1827     if (req->bu_flags & BW_UPCALL_GEQ)
1828 	flags |= BW_METER_GEQ;
1829     if (req->bu_flags & BW_UPCALL_LEQ)
1830 	flags |= BW_METER_LEQ;
1831 
1832     return flags;
1833 }
1834 
1835 static void
1836 expire_bw_meter_leq(void *arg)
1837 {
1838 	struct bw_meter *x = arg;
1839 	struct timeval now;
1840 	/*
1841 	 * INFO:
1842 	 * callout is always executed with MRW_WLOCK taken
1843 	 */
1844 
1845 	CURVNET_SET((struct vnet *)x->arg);
1846 
1847 	microtime(&now);
1848 
1849 	/*
1850 	 * Test if we should deliver an upcall
1851 	 */
1852 	if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1853 	    (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1854 	    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1855 	    (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1856 		/* Prepare an upcall for delivery */
1857 		bw_meter_prepare_upcall(x, &now);
1858 	}
1859 
1860 	/* Send all upcalls that are pending delivery */
1861 	taskqueue_enqueue(V_task_queue, &V_task);
1862 
1863 	/* Reset counters */
1864 	x->bm_start_time = now;
1865 	/* Spin lock has to be taken as ip_forward context
1866 	 * might modify these fields as well
1867 	 */
1868 	mtx_lock_spin(&x->bm_spin);
1869 	x->bm_measured.b_bytes = 0;
1870 	x->bm_measured.b_packets = 0;
1871 	mtx_unlock_spin(&x->bm_spin);
1872 
1873 	callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1874 
1875 	CURVNET_RESTORE();
1876 }
1877 
1878 /*
1879  * Add a bw_meter entry
1880  */
1881 static int
1882 add_bw_upcall(struct bw_upcall *req)
1883 {
1884 	struct mfc *mfc;
1885 	struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1886 	BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1887 	struct timeval now;
1888 	struct bw_meter *x, **bwm_ptr;
1889 	uint32_t flags;
1890 
1891 	if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1892 		return EOPNOTSUPP;
1893 
1894 	/* Test if the flags are valid */
1895 	if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1896 		return EINVAL;
1897 	if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1898 		return EINVAL;
1899 	if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1900 			== (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1901 		return EINVAL;
1902 
1903 	/* Test if the threshold time interval is valid */
1904 	if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1905 		return EINVAL;
1906 
1907 	flags = compute_bw_meter_flags(req);
1908 
1909 	/*
1910 	 * Find if we have already same bw_meter entry
1911 	 */
1912 	MRW_WLOCK();
1913 	mfc = mfc_find(&req->bu_src, &req->bu_dst);
1914 	if (mfc == NULL) {
1915 		MRW_WUNLOCK();
1916 		return EADDRNOTAVAIL;
1917 	}
1918 
1919 	/* Choose an appropriate bw_meter list */
1920 	if (req->bu_flags & BW_UPCALL_GEQ)
1921 		bwm_ptr = &mfc->mfc_bw_meter_geq;
1922 	else
1923 		bwm_ptr = &mfc->mfc_bw_meter_leq;
1924 
1925 	for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
1926 		if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1927 		    &req->bu_threshold.b_time, ==))
1928 		    && (x->bm_threshold.b_packets
1929 		    == req->bu_threshold.b_packets)
1930 		    && (x->bm_threshold.b_bytes
1931 		    == req->bu_threshold.b_bytes)
1932 		    && (x->bm_flags & BW_METER_USER_FLAGS)
1933 		    == flags) {
1934 			MRW_WUNLOCK();
1935 			return 0; /* XXX Already installed */
1936 		}
1937 	}
1938 
1939 	/* Allocate the new bw_meter entry */
1940 	x = (struct bw_meter*) malloc(sizeof(*x), M_BWMETER,
1941 	    M_ZERO | M_NOWAIT);
1942 	if (x == NULL) {
1943 		MRW_WUNLOCK();
1944 		return ENOBUFS;
1945 	}
1946 
1947 	/* Set the new bw_meter entry */
1948 	x->bm_threshold.b_time = req->bu_threshold.b_time;
1949 	microtime(&now);
1950 	x->bm_start_time = now;
1951 	x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1952 	x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1953 	x->bm_measured.b_packets = 0;
1954 	x->bm_measured.b_bytes = 0;
1955 	x->bm_flags = flags;
1956 	x->bm_time_next = NULL;
1957 	x->bm_mfc = mfc;
1958 	x->arg = curvnet;
1959 	sprintf(x->bm_spin_name, "BM spin %p", x);
1960 	mtx_init(&x->bm_spin, x->bm_spin_name, NULL, MTX_SPIN);
1961 
1962 	/* For LEQ case create periodic callout */
1963 	if (req->bu_flags & BW_UPCALL_LEQ) {
1964 		callout_init_rw(&x->bm_meter_callout, &mrouter_mtx, CALLOUT_SHAREDLOCK);
1965 		callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
1966 		    expire_bw_meter_leq, x);
1967 	}
1968 
1969 	/* Add the new bw_meter entry to the front of entries for this MFC */
1970 	x->bm_mfc_next = *bwm_ptr;
1971 	*bwm_ptr = x;
1972 
1973 	MRW_WUNLOCK();
1974 
1975 	return 0;
1976 }
1977 
1978 static void
1979 free_bw_list(struct bw_meter *list)
1980 {
1981     while (list != NULL) {
1982 	struct bw_meter *x = list;
1983 
1984 	/* MRW_WLOCK must be held here */
1985 	if (x->bm_flags & BW_METER_LEQ) {
1986 		callout_drain(&x->bm_meter_callout);
1987 		mtx_destroy(&x->bm_spin);
1988 	}
1989 
1990 	list = list->bm_mfc_next;
1991 	free(x, M_BWMETER);
1992     }
1993 }
1994 
1995 /*
1996  * Delete one or multiple bw_meter entries
1997  */
1998 static int
1999 del_bw_upcall(struct bw_upcall *req)
2000 {
2001     struct mfc *mfc;
2002     struct bw_meter *x, **bwm_ptr;
2003 
2004     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
2005 	return EOPNOTSUPP;
2006 
2007     MRW_WLOCK();
2008 
2009     /* Find the corresponding MFC entry */
2010     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2011     if (mfc == NULL) {
2012 	MRW_WUNLOCK();
2013 	return EADDRNOTAVAIL;
2014     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2015 	/*
2016 	 * Delete all bw_meter entries for this mfc
2017 	 */
2018 	struct bw_meter *list;
2019 
2020 	/* Free LEQ list */
2021 	list = mfc->mfc_bw_meter_leq;
2022 	mfc->mfc_bw_meter_leq = NULL;
2023 	free_bw_list(list);
2024 
2025 	/* Free GEQ list */
2026 	list = mfc->mfc_bw_meter_geq;
2027 	mfc->mfc_bw_meter_geq = NULL;
2028 	free_bw_list(list);
2029 	MRW_WUNLOCK();
2030 	return 0;
2031     } else {			/* Delete a single bw_meter entry */
2032 	struct bw_meter *prev;
2033 	uint32_t flags = 0;
2034 
2035 	flags = compute_bw_meter_flags(req);
2036 
2037 	/* Choose an appropriate bw_meter list */
2038 	if (req->bu_flags & BW_UPCALL_GEQ)
2039 		bwm_ptr = &mfc->mfc_bw_meter_geq;
2040 	else
2041 		bwm_ptr = &mfc->mfc_bw_meter_leq;
2042 
2043 	/* Find the bw_meter entry to delete */
2044 	for (prev = NULL, x = *bwm_ptr; x != NULL;
2045 	     prev = x, x = x->bm_mfc_next) {
2046 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2047 			       &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     /* NOTREACHED */
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 = { 0,		/* unused1 */
2179 				      0,		/* unused2 */
2180 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2181 				      0,		/* im_mbz  */
2182 				      0,		/* im_vif  */
2183 				      0,		/* unused3 */
2184 				      { 0 },		/* im_src  */
2185 				      { 0 } };		/* im_dst  */
2186 
2187     MRW_LOCK_ASSERT();
2188 
2189     if (buf_ring_empty(V_bw_upcalls_ring))
2190 	return;
2191 
2192     /*
2193      * Allocate a new mbuf, initialize it with the header and
2194      * the payload for the pending calls.
2195      */
2196     m = m_gethdr(M_NOWAIT, MT_DATA);
2197     if (m == NULL) {
2198 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2199 	return;
2200     }
2201 
2202     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2203     len += sizeof(struct igmpmsg);
2204     while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
2205 	m_copyback(m, len, sizeof(struct bw_upcall), (caddr_t)bu);
2206 	len += sizeof(struct bw_upcall);
2207 	free(bu, M_MRTABLE);
2208     }
2209 
2210     /*
2211      * Send the upcalls
2212      * XXX do we need to set the address in k_igmpsrc ?
2213      */
2214     MRTSTAT_INC(mrts_upcalls);
2215     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2216 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2217 	MRTSTAT_INC(mrts_upq_sockfull);
2218     }
2219 }
2220 
2221 /*
2222  * A periodic function for sending all upcalls that are pending delivery
2223  */
2224 static void
2225 expire_bw_upcalls_send(void *arg)
2226 {
2227     CURVNET_SET((struct vnet *) arg);
2228 
2229     /* This callout is run with MRW_RLOCK taken */
2230 
2231     bw_upcalls_send();
2232 
2233     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2234 	curvnet);
2235     CURVNET_RESTORE();
2236 }
2237 
2238 /*
2239  * End of bandwidth monitoring code
2240  */
2241 
2242 /*
2243  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2244  *
2245  */
2246 static int
2247 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2248     struct mfc *rt)
2249 {
2250     struct mbuf *mb_copy, *mm;
2251 
2252     /*
2253      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2254      * rendezvous point was unspecified, and we were told not to.
2255      */
2256     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2257 	in_nullhost(rt->mfc_rp))
2258 	return 0;
2259 
2260     mb_copy = pim_register_prepare(ip, m);
2261     if (mb_copy == NULL)
2262 	return ENOBUFS;
2263 
2264     /*
2265      * Send all the fragments. Note that the mbuf for each fragment
2266      * is freed by the sending machinery.
2267      */
2268     for (mm = mb_copy; mm; mm = mb_copy) {
2269 	mb_copy = mm->m_nextpkt;
2270 	mm->m_nextpkt = 0;
2271 	mm = m_pullup(mm, sizeof(struct ip));
2272 	if (mm != NULL) {
2273 	    ip = mtod(mm, struct ip *);
2274 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2275 		pim_register_send_rp(ip, vifp, mm, rt);
2276 	    } else {
2277 		pim_register_send_upcall(ip, vifp, mm, rt);
2278 	    }
2279 	}
2280     }
2281 
2282     return 0;
2283 }
2284 
2285 /*
2286  * Return a copy of the data packet that is ready for PIM Register
2287  * encapsulation.
2288  * XXX: Note that in the returned copy the IP header is a valid one.
2289  */
2290 static struct mbuf *
2291 pim_register_prepare(struct ip *ip, struct mbuf *m)
2292 {
2293     struct mbuf *mb_copy = NULL;
2294     int mtu;
2295 
2296     /* Take care of delayed checksums */
2297     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2298 	in_delayed_cksum(m);
2299 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2300     }
2301 
2302     /*
2303      * Copy the old packet & pullup its IP header into the
2304      * new mbuf so we can modify it.
2305      */
2306     mb_copy = m_copypacket(m, M_NOWAIT);
2307     if (mb_copy == NULL)
2308 	return NULL;
2309     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2310     if (mb_copy == NULL)
2311 	return NULL;
2312 
2313     /* take care of the TTL */
2314     ip = mtod(mb_copy, struct ip *);
2315     --ip->ip_ttl;
2316 
2317     /* Compute the MTU after the PIM Register encapsulation */
2318     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2319 
2320     if (ntohs(ip->ip_len) <= mtu) {
2321 	/* Turn the IP header into a valid one */
2322 	ip->ip_sum = 0;
2323 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2324     } else {
2325 	/* Fragment the packet */
2326 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2327 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2328 	    m_freem(mb_copy);
2329 	    return NULL;
2330 	}
2331     }
2332     return mb_copy;
2333 }
2334 
2335 /*
2336  * Send an upcall with the data packet to the user-level process.
2337  */
2338 static int
2339 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2340     struct mbuf *mb_copy, struct mfc *rt)
2341 {
2342     struct mbuf *mb_first;
2343     int len = ntohs(ip->ip_len);
2344     struct igmpmsg *im;
2345     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2346 
2347     MRW_LOCK_ASSERT();
2348 
2349     /*
2350      * Add a new mbuf with an upcall header
2351      */
2352     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2353     if (mb_first == NULL) {
2354 	m_freem(mb_copy);
2355 	return ENOBUFS;
2356     }
2357     mb_first->m_data += max_linkhdr;
2358     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2359     mb_first->m_len = sizeof(struct igmpmsg);
2360     mb_first->m_next = mb_copy;
2361 
2362     /* Send message to routing daemon */
2363     im = mtod(mb_first, struct igmpmsg *);
2364     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2365     im->im_mbz		= 0;
2366     im->im_vif		= vifp - V_viftable;
2367     im->im_src		= ip->ip_src;
2368     im->im_dst		= ip->ip_dst;
2369 
2370     k_igmpsrc.sin_addr	= ip->ip_src;
2371 
2372     MRTSTAT_INC(mrts_upcalls);
2373 
2374     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2375 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2376 	MRTSTAT_INC(mrts_upq_sockfull);
2377 	return ENOBUFS;
2378     }
2379 
2380     /* Keep statistics */
2381     PIMSTAT_INC(pims_snd_registers_msgs);
2382     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2383 
2384     return 0;
2385 }
2386 
2387 /*
2388  * Encapsulate the data packet in PIM Register message and send it to the RP.
2389  */
2390 static int
2391 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2392     struct mfc *rt)
2393 {
2394     struct mbuf *mb_first;
2395     struct ip *ip_outer;
2396     struct pim_encap_pimhdr *pimhdr;
2397     int len = ntohs(ip->ip_len);
2398     vifi_t vifi = rt->mfc_parent;
2399 
2400     MRW_LOCK_ASSERT();
2401 
2402     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2403 	m_freem(mb_copy);
2404 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2405     }
2406 
2407     /*
2408      * Add a new mbuf with the encapsulating header
2409      */
2410     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2411     if (mb_first == NULL) {
2412 	m_freem(mb_copy);
2413 	return ENOBUFS;
2414     }
2415     mb_first->m_data += max_linkhdr;
2416     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2417     mb_first->m_next = mb_copy;
2418 
2419     mb_first->m_pkthdr.len = len + mb_first->m_len;
2420 
2421     /*
2422      * Fill in the encapsulating IP and PIM header
2423      */
2424     ip_outer = mtod(mb_first, struct ip *);
2425     *ip_outer = pim_encap_iphdr;
2426     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2427 	sizeof(pim_encap_pimhdr));
2428     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2429     ip_outer->ip_dst = rt->mfc_rp;
2430     /*
2431      * Copy the inner header TOS to the outer header, and take care of the
2432      * IP_DF bit.
2433      */
2434     ip_outer->ip_tos = ip->ip_tos;
2435     if (ip->ip_off & htons(IP_DF))
2436 	ip_outer->ip_off |= htons(IP_DF);
2437     ip_fillid(ip_outer);
2438     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2439 					 + sizeof(pim_encap_iphdr));
2440     *pimhdr = pim_encap_pimhdr;
2441     /* If the iif crosses a border, set the Border-bit */
2442     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2443 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2444 
2445     mb_first->m_data += sizeof(pim_encap_iphdr);
2446     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2447     mb_first->m_data -= sizeof(pim_encap_iphdr);
2448 
2449     send_packet(vifp, mb_first);
2450 
2451     /* Keep statistics */
2452     PIMSTAT_INC(pims_snd_registers_msgs);
2453     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2454 
2455     return 0;
2456 }
2457 
2458 /*
2459  * pim_encapcheck() is called by the encap4_input() path at runtime to
2460  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2461  * into the kernel.
2462  */
2463 static int
2464 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2465     int proto __unused, void *arg __unused)
2466 {
2467 
2468     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2469     return (8);		/* claim the datagram. */
2470 }
2471 
2472 /*
2473  * PIM-SMv2 and PIM-DM messages processing.
2474  * Receives and verifies the PIM control messages, and passes them
2475  * up to the listening socket, using rip_input().
2476  * The only message with special processing is the PIM_REGISTER message
2477  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2478  * is passed to if_simloop().
2479  */
2480 static int
2481 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2482 {
2483     struct ip *ip = mtod(m, struct ip *);
2484     struct pim *pim;
2485     int iphlen = off;
2486     int minlen;
2487     int datalen = ntohs(ip->ip_len) - iphlen;
2488     int ip_tos;
2489 
2490     /* Keep statistics */
2491     PIMSTAT_INC(pims_rcv_total_msgs);
2492     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2493 
2494     /*
2495      * Validate lengths
2496      */
2497     if (datalen < PIM_MINLEN) {
2498 	PIMSTAT_INC(pims_rcv_tooshort);
2499 	CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2500 	    __func__, datalen, ntohl(ip->ip_src.s_addr));
2501 	m_freem(m);
2502 	return (IPPROTO_DONE);
2503     }
2504 
2505     /*
2506      * If the packet is at least as big as a REGISTER, go agead
2507      * and grab the PIM REGISTER header size, to avoid another
2508      * possible m_pullup() later.
2509      *
2510      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2511      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2512      */
2513     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2514     /*
2515      * Get the IP and PIM headers in contiguous memory, and
2516      * possibly the PIM REGISTER header.
2517      */
2518     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2519 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2520 	return (IPPROTO_DONE);
2521     }
2522 
2523     /* m_pullup() may have given us a new mbuf so reset ip. */
2524     ip = mtod(m, struct ip *);
2525     ip_tos = ip->ip_tos;
2526 
2527     /* adjust mbuf to point to the PIM header */
2528     m->m_data += iphlen;
2529     m->m_len  -= iphlen;
2530     pim = mtod(m, struct pim *);
2531 
2532     /*
2533      * Validate checksum. If PIM REGISTER, exclude the data packet.
2534      *
2535      * XXX: some older PIMv2 implementations don't make this distinction,
2536      * so for compatibility reason perform the checksum over part of the
2537      * message, and if error, then over the whole message.
2538      */
2539     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2540 	/* do nothing, checksum okay */
2541     } else if (in_cksum(m, datalen)) {
2542 	PIMSTAT_INC(pims_rcv_badsum);
2543 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2544 	m_freem(m);
2545 	return (IPPROTO_DONE);
2546     }
2547 
2548     /* PIM version check */
2549     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2550 	PIMSTAT_INC(pims_rcv_badversion);
2551 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2552 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2553 	m_freem(m);
2554 	return (IPPROTO_DONE);
2555     }
2556 
2557     /* restore mbuf back to the outer IP */
2558     m->m_data -= iphlen;
2559     m->m_len  += iphlen;
2560 
2561     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2562 	/*
2563 	 * Since this is a REGISTER, we'll make a copy of the register
2564 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2565 	 * routing daemon.
2566 	 */
2567 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2568 	struct mbuf *mcp;
2569 	struct ip *encap_ip;
2570 	u_int32_t *reghdr;
2571 	struct ifnet *vifp;
2572 
2573 	MRW_RLOCK();
2574 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2575 	    MRW_RUNLOCK();
2576 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2577 		(int)V_reg_vif_num);
2578 	    m_freem(m);
2579 	    return (IPPROTO_DONE);
2580 	}
2581 	/* XXX need refcnt? */
2582 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2583 	MRW_RUNLOCK();
2584 
2585 	/*
2586 	 * Validate length
2587 	 */
2588 	if (datalen < PIM_REG_MINLEN) {
2589 	    PIMSTAT_INC(pims_rcv_tooshort);
2590 	    PIMSTAT_INC(pims_rcv_badregisters);
2591 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2592 	    m_freem(m);
2593 	    return (IPPROTO_DONE);
2594 	}
2595 
2596 	reghdr = (u_int32_t *)(pim + 1);
2597 	encap_ip = (struct ip *)(reghdr + 1);
2598 
2599 	CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2600 	    __func__, ntohl(encap_ip->ip_src.s_addr),
2601 	    ntohs(encap_ip->ip_len));
2602 
2603 	/* verify the version number of the inner packet */
2604 	if (encap_ip->ip_v != IPVERSION) {
2605 	    PIMSTAT_INC(pims_rcv_badregisters);
2606 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2607 	    m_freem(m);
2608 	    return (IPPROTO_DONE);
2609 	}
2610 
2611 	/* verify the inner packet is destined to a mcast group */
2612 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2613 	    PIMSTAT_INC(pims_rcv_badregisters);
2614 	    CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2615 		ntohl(encap_ip->ip_dst.s_addr));
2616 	    m_freem(m);
2617 	    return (IPPROTO_DONE);
2618 	}
2619 
2620 	/* If a NULL_REGISTER, pass it to the daemon */
2621 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2622 	    goto pim_input_to_daemon;
2623 
2624 	/*
2625 	 * Copy the TOS from the outer IP header to the inner IP header.
2626 	 */
2627 	if (encap_ip->ip_tos != ip_tos) {
2628 	    /* Outer TOS -> inner TOS */
2629 	    encap_ip->ip_tos = ip_tos;
2630 	    /* Recompute the inner header checksum. Sigh... */
2631 
2632 	    /* adjust mbuf to point to the inner IP header */
2633 	    m->m_data += (iphlen + PIM_MINLEN);
2634 	    m->m_len  -= (iphlen + PIM_MINLEN);
2635 
2636 	    encap_ip->ip_sum = 0;
2637 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2638 
2639 	    /* restore mbuf to point back to the outer IP header */
2640 	    m->m_data -= (iphlen + PIM_MINLEN);
2641 	    m->m_len  += (iphlen + PIM_MINLEN);
2642 	}
2643 
2644 	/*
2645 	 * Decapsulate the inner IP packet and loopback to forward it
2646 	 * as a normal multicast packet. Also, make a copy of the
2647 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2648 	 * to pass to the daemon later, so it can take the appropriate
2649 	 * actions (e.g., send back PIM_REGISTER_STOP).
2650 	 * XXX: here m->m_data points to the outer IP header.
2651 	 */
2652 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2653 	if (mcp == NULL) {
2654 	    CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2655 	    m_freem(m);
2656 	    return (IPPROTO_DONE);
2657 	}
2658 
2659 	/* Keep statistics */
2660 	/* XXX: registers_bytes include only the encap. mcast pkt */
2661 	PIMSTAT_INC(pims_rcv_registers_msgs);
2662 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2663 
2664 	/*
2665 	 * forward the inner ip packet; point m_data at the inner ip.
2666 	 */
2667 	m_adj(m, iphlen + PIM_MINLEN);
2668 
2669 	CTR4(KTR_IPMF,
2670 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2671 	    __func__,
2672 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2673 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2674 	    (int)V_reg_vif_num);
2675 
2676 	/* NB: vifp was collected above; can it change on us? */
2677 	if_simloop(vifp, m, dst.sin_family, 0);
2678 
2679 	/* prepare the register head to send to the mrouting daemon */
2680 	m = mcp;
2681     }
2682 
2683 pim_input_to_daemon:
2684     /*
2685      * Pass the PIM message up to the daemon; if it is a Register message,
2686      * pass the 'head' only up to the daemon. This includes the
2687      * outer IP header, PIM header, PIM-Register header and the
2688      * inner IP header.
2689      * XXX: the outer IP header pkt size of a Register is not adjust to
2690      * reflect the fact that the inner multicast data is truncated.
2691      */
2692     return (rip_input(&m, &off, proto));
2693 }
2694 
2695 static int
2696 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2697 {
2698 	struct mfc	*rt;
2699 	int		 error, i;
2700 
2701 	if (req->newptr)
2702 		return (EPERM);
2703 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2704 		return (0);
2705 	error = sysctl_wire_old_buffer(req, 0);
2706 	if (error)
2707 		return (error);
2708 
2709 	MRW_RLOCK();
2710 	for (i = 0; i < mfchashsize; i++) {
2711 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2712 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2713 			if (error)
2714 				goto out_locked;
2715 		}
2716 	}
2717 out_locked:
2718 	MRW_RUNLOCK();
2719 	return (error);
2720 }
2721 
2722 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2723     CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2724     "IPv4 Multicast Forwarding Table "
2725     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2726 
2727 static int
2728 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2729 {
2730 	int error, i;
2731 
2732 	if (req->newptr)
2733 		return (EPERM);
2734 	if (V_viftable == NULL)		/* XXX unlocked */
2735 		return (0);
2736 	error = sysctl_wire_old_buffer(req, MROUTE_VIF_SYSCTL_LEN * MAXVIFS);
2737 	if (error)
2738 		return (error);
2739 
2740 	MRW_RLOCK();
2741 	/* Copy out user-visible portion of vif entry. */
2742 	for (i = 0; i < MAXVIFS; i++) {
2743 		error = SYSCTL_OUT(req, &V_viftable[i], MROUTE_VIF_SYSCTL_LEN);
2744 		if (error)
2745 			break;
2746 	}
2747 	MRW_RUNLOCK();
2748 	return (error);
2749 }
2750 
2751 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2752     CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2753     sysctl_viflist, "S,vif[MAXVIFS]",
2754     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2755 
2756 static void
2757 vnet_mroute_init(const void *unused __unused)
2758 {
2759 
2760 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2761 
2762 	V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2763 	    M_MRTABLE, M_WAITOK|M_ZERO);
2764 
2765 	callout_init_rw(&V_expire_upcalls_ch, &mrouter_mtx, 0);
2766 	callout_init_rw(&V_bw_upcalls_ch, &mrouter_mtx, 0);
2767 
2768 	/* Prepare taskqueue */
2769 	V_task_queue = taskqueue_create_fast("ip_mroute_tskq", M_NOWAIT,
2770 		    taskqueue_thread_enqueue, &V_task_queue);
2771 	taskqueue_start_threads(&V_task_queue, 1, PI_NET, "ip_mroute_tskq task");
2772 }
2773 
2774 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2775 	NULL);
2776 
2777 static void
2778 vnet_mroute_uninit(const void *unused __unused)
2779 {
2780 
2781 	/* Taskqueue should be cancelled and drained before freeing */
2782 	taskqueue_free(V_task_queue);
2783 
2784 	free(V_viftable, M_MRTABLE);
2785 	free(V_nexpire, M_MRTABLE);
2786 	V_nexpire = NULL;
2787 }
2788 
2789 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2790 	vnet_mroute_uninit, NULL);
2791 
2792 static int
2793 ip_mroute_modevent(module_t mod, int type, void *unused)
2794 {
2795 
2796     switch (type) {
2797     case MOD_LOAD:
2798 	MRW_LOCK_INIT();
2799 
2800 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2801 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2802 	if (if_detach_event_tag == NULL) {
2803 		printf("ip_mroute: unable to register "
2804 		    "ifnet_departure_event handler\n");
2805 		MRW_LOCK_DESTROY();
2806 		return (EINVAL);
2807 	}
2808 
2809 	mfchashsize = MFCHASHSIZE;
2810 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2811 	    !powerof2(mfchashsize)) {
2812 		printf("WARNING: %s not a power of 2; using default\n",
2813 		    "net.inet.ip.mfchashsize");
2814 		mfchashsize = MFCHASHSIZE;
2815 	}
2816 
2817 	pim_squelch_wholepkt = 0;
2818 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2819 	    &pim_squelch_wholepkt);
2820 
2821 	pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2822 	if (pim_encap_cookie == NULL) {
2823 		printf("ip_mroute: unable to attach pim encap\n");
2824 		MRW_LOCK_DESTROY();
2825 		return (EINVAL);
2826 	}
2827 
2828 	ip_mcast_src = X_ip_mcast_src;
2829 	ip_mforward = X_ip_mforward;
2830 	ip_mrouter_done = X_ip_mrouter_done;
2831 	ip_mrouter_get = X_ip_mrouter_get;
2832 	ip_mrouter_set = X_ip_mrouter_set;
2833 
2834 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2835 	ip_rsvp_vif = X_ip_rsvp_vif;
2836 
2837 	legal_vif_num = X_legal_vif_num;
2838 	mrt_ioctl = X_mrt_ioctl;
2839 	rsvp_input_p = X_rsvp_input;
2840 	break;
2841 
2842     case MOD_UNLOAD:
2843 	/*
2844 	 * Typically module unload happens after the user-level
2845 	 * process has shutdown the kernel services (the check
2846 	 * below insures someone can't just yank the module out
2847 	 * from under a running process).  But if the module is
2848 	 * just loaded and then unloaded w/o starting up a user
2849 	 * process we still need to cleanup.
2850 	 */
2851 	MRW_WLOCK();
2852 	if (ip_mrouter_cnt != 0) {
2853 	    MRW_WUNLOCK();
2854 	    return (EINVAL);
2855 	}
2856 	ip_mrouter_unloading = 1;
2857 	MRW_WUNLOCK();
2858 
2859 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2860 
2861 	if (pim_encap_cookie) {
2862 	    ip_encap_detach(pim_encap_cookie);
2863 	    pim_encap_cookie = NULL;
2864 	}
2865 
2866 	ip_mcast_src = NULL;
2867 	ip_mforward = NULL;
2868 	ip_mrouter_done = NULL;
2869 	ip_mrouter_get = NULL;
2870 	ip_mrouter_set = NULL;
2871 
2872 	ip_rsvp_force_done = NULL;
2873 	ip_rsvp_vif = NULL;
2874 
2875 	legal_vif_num = NULL;
2876 	mrt_ioctl = NULL;
2877 	rsvp_input_p = NULL;
2878 
2879 	MRW_LOCK_DESTROY();
2880 	break;
2881 
2882     default:
2883 	return EOPNOTSUPP;
2884     }
2885     return 0;
2886 }
2887 
2888 static moduledata_t ip_mroutemod = {
2889     "ip_mroute",
2890     ip_mroute_modevent,
2891     0
2892 };
2893 
2894 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2895