xref: /freebsd/sys/netinet/ip_mroute.c (revision 882f88ff77194ef2eea005232780468404438788)
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  *
53  * MROUTING Revision: 3.5
54  * and PIM-SMv2 and PIM-DM support, advanced API support,
55  * bandwidth metering and signaling
56  */
57 
58 /*
59  * TODO: Prefix functions with ipmf_.
60  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
61  * domain attachment (if_afdata) so we can track consumers of that service.
62  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
63  * move it to socket options.
64  * TODO: Cleanup LSRR removal further.
65  * TODO: Push RSVP stubs into raw_ip.c.
66  * TODO: Use bitstring.h for vif set.
67  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
68  * TODO: Sync ip6_mroute.c with this file.
69  */
70 
71 #include <sys/cdefs.h>
72 __FBSDID("$FreeBSD$");
73 
74 #include "opt_inet.h"
75 #include "opt_mrouting.h"
76 
77 #define _PIM_VT 1
78 
79 #include <sys/param.h>
80 #include <sys/kernel.h>
81 #include <sys/stddef.h>
82 #include <sys/eventhandler.h>
83 #include <sys/lock.h>
84 #include <sys/ktr.h>
85 #include <sys/malloc.h>
86 #include <sys/mbuf.h>
87 #include <sys/module.h>
88 #include <sys/priv.h>
89 #include <sys/protosw.h>
90 #include <sys/signalvar.h>
91 #include <sys/socket.h>
92 #include <sys/socketvar.h>
93 #include <sys/sockio.h>
94 #include <sys/sx.h>
95 #include <sys/sysctl.h>
96 #include <sys/syslog.h>
97 #include <sys/systm.h>
98 #include <sys/time.h>
99 #include <sys/counter.h>
100 
101 #include <net/if.h>
102 #include <net/if_var.h>
103 #include <net/netisr.h>
104 #include <net/route.h>
105 #include <net/vnet.h>
106 
107 #include <netinet/in.h>
108 #include <netinet/igmp.h>
109 #include <netinet/in_systm.h>
110 #include <netinet/in_var.h>
111 #include <netinet/ip.h>
112 #include <netinet/ip_encap.h>
113 #include <netinet/ip_mroute.h>
114 #include <netinet/ip_var.h>
115 #include <netinet/ip_options.h>
116 #include <netinet/pim.h>
117 #include <netinet/pim_var.h>
118 #include <netinet/udp.h>
119 
120 #include <machine/in_cksum.h>
121 
122 #ifndef KTR_IPMF
123 #define KTR_IPMF KTR_INET
124 #endif
125 
126 #define		VIFI_INVALID	((vifi_t) -1)
127 
128 VNET_DEFINE_STATIC(uint32_t, last_tv_sec); /* last time we processed this */
129 #define	V_last_tv_sec	VNET(last_tv_sec)
130 
131 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
132 
133 /*
134  * Locking.  We use two locks: one for the virtual interface table and
135  * one for the forwarding table.  These locks may be nested in which case
136  * the VIF lock must always be taken first.  Note that each lock is used
137  * to cover not only the specific data structure but also related data
138  * structures.
139  */
140 
141 static struct mtx mrouter_mtx;
142 #define	MROUTER_LOCK()		mtx_lock(&mrouter_mtx)
143 #define	MROUTER_UNLOCK()	mtx_unlock(&mrouter_mtx)
144 #define	MROUTER_LOCK_ASSERT()	mtx_assert(&mrouter_mtx, MA_OWNED)
145 #define	MROUTER_LOCK_INIT()						\
146 	mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
147 #define	MROUTER_LOCK_DESTROY()	mtx_destroy(&mrouter_mtx)
148 
149 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
150 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
151 
152 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
153 VNET_PCPUSTAT_SYSINIT(mrtstat);
154 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
155 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
156     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
157     "netinet/ip_mroute.h)");
158 
159 VNET_DEFINE_STATIC(u_long, mfchash);
160 #define	V_mfchash		VNET(mfchash)
161 #define	MFCHASH(a, g)							\
162 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
163 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
164 #define	MFCHASHSIZE	256
165 
166 static u_long mfchashsize;			/* Hash size */
167 VNET_DEFINE_STATIC(u_char *, nexpire);		/* 0..mfchashsize-1 */
168 #define	V_nexpire		VNET(nexpire)
169 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
170 #define	V_mfchashtbl		VNET(mfchashtbl)
171 
172 static struct mtx mfc_mtx;
173 #define	MFC_LOCK()		mtx_lock(&mfc_mtx)
174 #define	MFC_UNLOCK()		mtx_unlock(&mfc_mtx)
175 #define	MFC_LOCK_ASSERT()	mtx_assert(&mfc_mtx, MA_OWNED)
176 #define	MFC_LOCK_INIT()							\
177 	mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
178 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
179 
180 VNET_DEFINE_STATIC(vifi_t, numvifs);
181 #define	V_numvifs		VNET(numvifs)
182 VNET_DEFINE_STATIC(struct vif *, viftable);
183 #define	V_viftable		VNET(viftable)
184 /*
185  * No one should be able to "query" this before initialisation happened in
186  * vnet_mroute_init(), so we should still be fine.
187  */
188 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_VNET | CTLFLAG_RD,
189     &VNET_NAME(viftable), sizeof(*V_viftable) * MAXVIFS, "S,vif[MAXVIFS]",
190     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
191 
192 static struct mtx vif_mtx;
193 #define	VIF_LOCK()		mtx_lock(&vif_mtx)
194 #define	VIF_UNLOCK()		mtx_unlock(&vif_mtx)
195 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
196 #define	VIF_LOCK_INIT()							\
197 	mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
198 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
199 
200 static eventhandler_tag if_detach_event_tag = NULL;
201 
202 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
203 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
204 
205 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
206 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
207 
208 /*
209  * Bandwidth meter variables and constants
210  */
211 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
212 /*
213  * Pending timeouts are stored in a hash table, the key being the
214  * expiration time. Periodically, the entries are analysed and processed.
215  */
216 #define	BW_METER_BUCKETS	1024
217 VNET_DEFINE_STATIC(struct bw_meter **, bw_meter_timers);
218 #define	V_bw_meter_timers	VNET(bw_meter_timers)
219 VNET_DEFINE_STATIC(struct callout, bw_meter_ch);
220 #define	V_bw_meter_ch		VNET(bw_meter_ch)
221 #define	BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
222 
223 /*
224  * Pending upcalls are stored in a vector which is flushed when
225  * full, or periodically
226  */
227 VNET_DEFINE_STATIC(struct bw_upcall *, bw_upcalls);
228 #define	V_bw_upcalls		VNET(bw_upcalls)
229 VNET_DEFINE_STATIC(u_int, bw_upcalls_n); /* # of pending upcalls */
230 #define	V_bw_upcalls_n    	VNET(bw_upcalls_n)
231 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
232 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
233 
234 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
235 
236 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
237 VNET_PCPUSTAT_SYSINIT(pimstat);
238 VNET_PCPUSTAT_SYSUNINIT(pimstat);
239 
240 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
241 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
242     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
243 
244 static u_long	pim_squelch_wholepkt = 0;
245 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
246     &pim_squelch_wholepkt, 0,
247     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
248 
249 static const struct encaptab *pim_encap_cookie;
250 static int pim_encapcheck(const struct mbuf *, int, int, void *);
251 static int pim_input(struct mbuf *, int, int, void *);
252 
253 static const struct encap_config ipv4_encap_cfg = {
254 	.proto = IPPROTO_PIM,
255 	.min_length = sizeof(struct ip) + PIM_MINLEN,
256 	.exact_match = 8,
257 	.check = pim_encapcheck,
258 	.input = pim_input
259 };
260 
261 /*
262  * Note: the PIM Register encapsulation adds the following in front of a
263  * data packet:
264  *
265  * struct pim_encap_hdr {
266  *    struct ip ip;
267  *    struct pim_encap_pimhdr  pim;
268  * }
269  *
270  */
271 
272 struct pim_encap_pimhdr {
273 	struct pim pim;
274 	uint32_t   flags;
275 };
276 #define		PIM_ENCAP_TTL	64
277 
278 static struct ip pim_encap_iphdr = {
279 #if BYTE_ORDER == LITTLE_ENDIAN
280 	sizeof(struct ip) >> 2,
281 	IPVERSION,
282 #else
283 	IPVERSION,
284 	sizeof(struct ip) >> 2,
285 #endif
286 	0,			/* tos */
287 	sizeof(struct ip),	/* total length */
288 	0,			/* id */
289 	0,			/* frag offset */
290 	PIM_ENCAP_TTL,
291 	IPPROTO_PIM,
292 	0,			/* checksum */
293 };
294 
295 static struct pim_encap_pimhdr pim_encap_pimhdr = {
296     {
297 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
298 	0,			/* reserved */
299 	0,			/* checksum */
300     },
301     0				/* flags */
302 };
303 
304 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
305 #define	V_reg_vif_num		VNET(reg_vif_num)
306 VNET_DEFINE_STATIC(struct ifnet, multicast_register_if);
307 #define	V_multicast_register_if	VNET(multicast_register_if)
308 
309 /*
310  * Private variables.
311  */
312 
313 static u_long	X_ip_mcast_src(int);
314 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
315 		    struct ip_moptions *);
316 static int	X_ip_mrouter_done(void);
317 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
318 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
319 static int	X_legal_vif_num(int);
320 static int	X_mrt_ioctl(u_long, caddr_t, int);
321 
322 static int	add_bw_upcall(struct bw_upcall *);
323 static int	add_mfc(struct mfcctl2 *);
324 static int	add_vif(struct vifctl *);
325 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
326 static void	bw_meter_process(void);
327 static void	bw_meter_receive_packet(struct bw_meter *, int,
328 		    struct timeval *);
329 static void	bw_upcalls_send(void);
330 static int	del_bw_upcall(struct bw_upcall *);
331 static int	del_mfc(struct mfcctl2 *);
332 static int	del_vif(vifi_t);
333 static int	del_vif_locked(vifi_t);
334 static void	expire_bw_meter_process(void *);
335 static void	expire_bw_upcalls_send(void *);
336 static void	expire_mfc(struct mfc *);
337 static void	expire_upcalls(void *);
338 static void	free_bw_list(struct bw_meter *);
339 static int	get_sg_cnt(struct sioc_sg_req *);
340 static int	get_vif_cnt(struct sioc_vif_req *);
341 static void	if_detached_event(void *, struct ifnet *);
342 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
343 static int	ip_mrouter_init(struct socket *, int);
344 static __inline struct mfc *
345 		mfc_find(struct in_addr *, struct in_addr *);
346 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
347 static struct mbuf *
348 		pim_register_prepare(struct ip *, struct mbuf *);
349 static int	pim_register_send(struct ip *, struct vif *,
350 		    struct mbuf *, struct mfc *);
351 static int	pim_register_send_rp(struct ip *, struct vif *,
352 		    struct mbuf *, struct mfc *);
353 static int	pim_register_send_upcall(struct ip *, struct vif *,
354 		    struct mbuf *, struct mfc *);
355 static void	schedule_bw_meter(struct bw_meter *, struct timeval *);
356 static void	send_packet(struct vif *, struct mbuf *);
357 static int	set_api_config(uint32_t *);
358 static int	set_assert(int);
359 static int	socket_send(struct socket *, struct mbuf *,
360 		    struct sockaddr_in *);
361 static void	unschedule_bw_meter(struct bw_meter *);
362 
363 /*
364  * Kernel multicast forwarding API capabilities and setup.
365  * If more API capabilities are added to the kernel, they should be
366  * recorded in `mrt_api_support'.
367  */
368 #define MRT_API_VERSION		0x0305
369 
370 static const int mrt_api_version = MRT_API_VERSION;
371 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
372 					 MRT_MFC_FLAGS_BORDER_VIF |
373 					 MRT_MFC_RP |
374 					 MRT_MFC_BW_UPCALL);
375 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
376 #define	V_mrt_api_config	VNET(mrt_api_config)
377 VNET_DEFINE_STATIC(int, pim_assert_enabled);
378 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
379 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
380 
381 /*
382  * Find a route for a given origin IP address and multicast group address.
383  * Statistics must be updated by the caller.
384  */
385 static __inline struct mfc *
386 mfc_find(struct in_addr *o, struct in_addr *g)
387 {
388 	struct mfc *rt;
389 
390 	MFC_LOCK_ASSERT();
391 
392 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
393 		if (in_hosteq(rt->mfc_origin, *o) &&
394 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
395 		    TAILQ_EMPTY(&rt->mfc_stall))
396 			break;
397 	}
398 
399 	return (rt);
400 }
401 
402 /*
403  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
404  */
405 static int
406 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
407 {
408     int	error, optval;
409     vifi_t	vifi;
410     struct	vifctl vifc;
411     struct	mfcctl2 mfc;
412     struct	bw_upcall bw_upcall;
413     uint32_t	i;
414 
415     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
416 	return EPERM;
417 
418     error = 0;
419     switch (sopt->sopt_name) {
420     case MRT_INIT:
421 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
422 	if (error)
423 	    break;
424 	error = ip_mrouter_init(so, optval);
425 	break;
426 
427     case MRT_DONE:
428 	error = ip_mrouter_done();
429 	break;
430 
431     case MRT_ADD_VIF:
432 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
433 	if (error)
434 	    break;
435 	error = add_vif(&vifc);
436 	break;
437 
438     case MRT_DEL_VIF:
439 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
440 	if (error)
441 	    break;
442 	error = del_vif(vifi);
443 	break;
444 
445     case MRT_ADD_MFC:
446     case MRT_DEL_MFC:
447 	/*
448 	 * select data size depending on API version.
449 	 */
450 	if (sopt->sopt_name == MRT_ADD_MFC &&
451 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
452 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
453 				sizeof(struct mfcctl2));
454 	} else {
455 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
456 				sizeof(struct mfcctl));
457 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
458 			sizeof(mfc) - sizeof(struct mfcctl));
459 	}
460 	if (error)
461 	    break;
462 	if (sopt->sopt_name == MRT_ADD_MFC)
463 	    error = add_mfc(&mfc);
464 	else
465 	    error = del_mfc(&mfc);
466 	break;
467 
468     case MRT_ASSERT:
469 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
470 	if (error)
471 	    break;
472 	set_assert(optval);
473 	break;
474 
475     case MRT_API_CONFIG:
476 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
477 	if (!error)
478 	    error = set_api_config(&i);
479 	if (!error)
480 	    error = sooptcopyout(sopt, &i, sizeof i);
481 	break;
482 
483     case MRT_ADD_BW_UPCALL:
484     case MRT_DEL_BW_UPCALL:
485 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
486 				sizeof bw_upcall);
487 	if (error)
488 	    break;
489 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
490 	    error = add_bw_upcall(&bw_upcall);
491 	else
492 	    error = del_bw_upcall(&bw_upcall);
493 	break;
494 
495     default:
496 	error = EOPNOTSUPP;
497 	break;
498     }
499     return error;
500 }
501 
502 /*
503  * Handle MRT getsockopt commands
504  */
505 static int
506 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
507 {
508     int error;
509 
510     switch (sopt->sopt_name) {
511     case MRT_VERSION:
512 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
513 	break;
514 
515     case MRT_ASSERT:
516 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
517 	    sizeof V_pim_assert_enabled);
518 	break;
519 
520     case MRT_API_SUPPORT:
521 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
522 	break;
523 
524     case MRT_API_CONFIG:
525 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
526 	break;
527 
528     default:
529 	error = EOPNOTSUPP;
530 	break;
531     }
532     return error;
533 }
534 
535 /*
536  * Handle ioctl commands to obtain information from the cache
537  */
538 static int
539 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
540 {
541     int error = 0;
542 
543     /*
544      * Currently the only function calling this ioctl routine is rtioctl_fib().
545      * Typically, only root can create the raw socket in order to execute
546      * this ioctl method, however the request might be coming from a prison
547      */
548     error = priv_check(curthread, PRIV_NETINET_MROUTE);
549     if (error)
550 	return (error);
551     switch (cmd) {
552     case (SIOCGETVIFCNT):
553 	error = get_vif_cnt((struct sioc_vif_req *)data);
554 	break;
555 
556     case (SIOCGETSGCNT):
557 	error = get_sg_cnt((struct sioc_sg_req *)data);
558 	break;
559 
560     default:
561 	error = EINVAL;
562 	break;
563     }
564     return error;
565 }
566 
567 /*
568  * returns the packet, byte, rpf-failure count for the source group provided
569  */
570 static int
571 get_sg_cnt(struct sioc_sg_req *req)
572 {
573     struct mfc *rt;
574 
575     MFC_LOCK();
576     rt = mfc_find(&req->src, &req->grp);
577     if (rt == NULL) {
578 	MFC_UNLOCK();
579 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
580 	return EADDRNOTAVAIL;
581     }
582     req->pktcnt = rt->mfc_pkt_cnt;
583     req->bytecnt = rt->mfc_byte_cnt;
584     req->wrong_if = rt->mfc_wrong_if;
585     MFC_UNLOCK();
586     return 0;
587 }
588 
589 /*
590  * returns the input and output packet and byte counts on the vif provided
591  */
592 static int
593 get_vif_cnt(struct sioc_vif_req *req)
594 {
595     vifi_t vifi = req->vifi;
596 
597     VIF_LOCK();
598     if (vifi >= V_numvifs) {
599 	VIF_UNLOCK();
600 	return EINVAL;
601     }
602 
603     req->icount = V_viftable[vifi].v_pkt_in;
604     req->ocount = V_viftable[vifi].v_pkt_out;
605     req->ibytes = V_viftable[vifi].v_bytes_in;
606     req->obytes = V_viftable[vifi].v_bytes_out;
607     VIF_UNLOCK();
608 
609     return 0;
610 }
611 
612 static void
613 if_detached_event(void *arg __unused, struct ifnet *ifp)
614 {
615     vifi_t vifi;
616     u_long i;
617 
618     MROUTER_LOCK();
619 
620     if (V_ip_mrouter == NULL) {
621 	MROUTER_UNLOCK();
622 	return;
623     }
624 
625     VIF_LOCK();
626     MFC_LOCK();
627 
628     /*
629      * Tear down multicast forwarder state associated with this ifnet.
630      * 1. Walk the vif list, matching vifs against this ifnet.
631      * 2. Walk the multicast forwarding cache (mfc) looking for
632      *    inner matches with this vif's index.
633      * 3. Expire any matching multicast forwarding cache entries.
634      * 4. Free vif state. This should disable ALLMULTI on the interface.
635      */
636     for (vifi = 0; vifi < V_numvifs; vifi++) {
637 	if (V_viftable[vifi].v_ifp != ifp)
638 		continue;
639 	for (i = 0; i < mfchashsize; i++) {
640 		struct mfc *rt, *nrt;
641 
642 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
643 			if (rt->mfc_parent == vifi) {
644 				expire_mfc(rt);
645 			}
646 		}
647 	}
648 	del_vif_locked(vifi);
649     }
650 
651     MFC_UNLOCK();
652     VIF_UNLOCK();
653 
654     MROUTER_UNLOCK();
655 }
656 
657 /*
658  * Enable multicast forwarding.
659  */
660 static int
661 ip_mrouter_init(struct socket *so, int version)
662 {
663 
664     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
665         so->so_type, so->so_proto->pr_protocol);
666 
667     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
668 	return EOPNOTSUPP;
669 
670     if (version != 1)
671 	return ENOPROTOOPT;
672 
673     MROUTER_LOCK();
674 
675     if (ip_mrouter_unloading) {
676 	MROUTER_UNLOCK();
677 	return ENOPROTOOPT;
678     }
679 
680     if (V_ip_mrouter != NULL) {
681 	MROUTER_UNLOCK();
682 	return EADDRINUSE;
683     }
684 
685     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
686 	HASH_NOWAIT);
687 
688     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
689 	curvnet);
690     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
691 	curvnet);
692     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
693 	curvnet);
694 
695     V_ip_mrouter = so;
696     ip_mrouter_cnt++;
697 
698     MROUTER_UNLOCK();
699 
700     CTR1(KTR_IPMF, "%s: done", __func__);
701 
702     return 0;
703 }
704 
705 /*
706  * Disable multicast forwarding.
707  */
708 static int
709 X_ip_mrouter_done(void)
710 {
711     struct ifnet *ifp;
712     u_long i;
713     vifi_t vifi;
714 
715     MROUTER_LOCK();
716 
717     if (V_ip_mrouter == NULL) {
718 	MROUTER_UNLOCK();
719 	return EINVAL;
720     }
721 
722     /*
723      * Detach/disable hooks to the reset of the system.
724      */
725     V_ip_mrouter = NULL;
726     ip_mrouter_cnt--;
727     V_mrt_api_config = 0;
728 
729     VIF_LOCK();
730 
731     /*
732      * For each phyint in use, disable promiscuous reception of all IP
733      * multicasts.
734      */
735     for (vifi = 0; vifi < V_numvifs; vifi++) {
736 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
737 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
738 	    ifp = V_viftable[vifi].v_ifp;
739 	    if_allmulti(ifp, 0);
740 	}
741     }
742     bzero((caddr_t)V_viftable, sizeof(V_viftable));
743     V_numvifs = 0;
744     V_pim_assert_enabled = 0;
745 
746     VIF_UNLOCK();
747 
748     callout_stop(&V_expire_upcalls_ch);
749     callout_stop(&V_bw_upcalls_ch);
750     callout_stop(&V_bw_meter_ch);
751 
752     MFC_LOCK();
753 
754     /*
755      * Free all multicast forwarding cache entries.
756      * Do not use hashdestroy(), as we must perform other cleanup.
757      */
758     for (i = 0; i < mfchashsize; i++) {
759 	struct mfc *rt, *nrt;
760 
761 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
762 		expire_mfc(rt);
763 	}
764     }
765     free(V_mfchashtbl, M_MRTABLE);
766     V_mfchashtbl = NULL;
767 
768     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
769 
770     V_bw_upcalls_n = 0;
771     bzero(V_bw_meter_timers, BW_METER_BUCKETS * sizeof(*V_bw_meter_timers));
772 
773     MFC_UNLOCK();
774 
775     V_reg_vif_num = VIFI_INVALID;
776 
777     MROUTER_UNLOCK();
778 
779     CTR1(KTR_IPMF, "%s: done", __func__);
780 
781     return 0;
782 }
783 
784 /*
785  * Set PIM assert processing global
786  */
787 static int
788 set_assert(int i)
789 {
790     if ((i != 1) && (i != 0))
791 	return EINVAL;
792 
793     V_pim_assert_enabled = i;
794 
795     return 0;
796 }
797 
798 /*
799  * Configure API capabilities
800  */
801 int
802 set_api_config(uint32_t *apival)
803 {
804     u_long i;
805 
806     /*
807      * We can set the API capabilities only if it is the first operation
808      * after MRT_INIT. I.e.:
809      *  - there are no vifs installed
810      *  - pim_assert is not enabled
811      *  - the MFC table is empty
812      */
813     if (V_numvifs > 0) {
814 	*apival = 0;
815 	return EPERM;
816     }
817     if (V_pim_assert_enabled) {
818 	*apival = 0;
819 	return EPERM;
820     }
821 
822     MFC_LOCK();
823 
824     for (i = 0; i < mfchashsize; i++) {
825 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
826 	    MFC_UNLOCK();
827 	    *apival = 0;
828 	    return EPERM;
829 	}
830     }
831 
832     MFC_UNLOCK();
833 
834     V_mrt_api_config = *apival & mrt_api_support;
835     *apival = V_mrt_api_config;
836 
837     return 0;
838 }
839 
840 /*
841  * Add a vif to the vif table
842  */
843 static int
844 add_vif(struct vifctl *vifcp)
845 {
846     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
847     struct sockaddr_in sin = {sizeof sin, AF_INET};
848     struct ifaddr *ifa;
849     struct ifnet *ifp;
850     int error;
851 
852     VIF_LOCK();
853     if (vifcp->vifc_vifi >= MAXVIFS) {
854 	VIF_UNLOCK();
855 	return EINVAL;
856     }
857     /* rate limiting is no longer supported by this code */
858     if (vifcp->vifc_rate_limit != 0) {
859 	log(LOG_ERR, "rate limiting is no longer supported\n");
860 	VIF_UNLOCK();
861 	return EINVAL;
862     }
863     if (!in_nullhost(vifp->v_lcl_addr)) {
864 	VIF_UNLOCK();
865 	return EADDRINUSE;
866     }
867     if (in_nullhost(vifcp->vifc_lcl_addr)) {
868 	VIF_UNLOCK();
869 	return EADDRNOTAVAIL;
870     }
871 
872     /* Find the interface with an address in AF_INET family */
873     if (vifcp->vifc_flags & VIFF_REGISTER) {
874 	/*
875 	 * XXX: Because VIFF_REGISTER does not really need a valid
876 	 * local interface (e.g. it could be 127.0.0.2), we don't
877 	 * check its address.
878 	 */
879 	ifp = NULL;
880     } else {
881 	sin.sin_addr = vifcp->vifc_lcl_addr;
882 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
883 	if (ifa == NULL) {
884 	    VIF_UNLOCK();
885 	    return EADDRNOTAVAIL;
886 	}
887 	ifp = ifa->ifa_ifp;
888     }
889 
890     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
891 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
892 	VIF_UNLOCK();
893 	return EOPNOTSUPP;
894     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
895 	ifp = &V_multicast_register_if;
896 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
897 	if (V_reg_vif_num == VIFI_INVALID) {
898 	    if_initname(&V_multicast_register_if, "register_vif", 0);
899 	    V_multicast_register_if.if_flags = IFF_LOOPBACK;
900 	    V_reg_vif_num = vifcp->vifc_vifi;
901 	}
902     } else {		/* Make sure the interface supports multicast */
903 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
904 	    VIF_UNLOCK();
905 	    return EOPNOTSUPP;
906 	}
907 
908 	/* Enable promiscuous reception of all IP multicasts from the if */
909 	error = if_allmulti(ifp, 1);
910 	if (error) {
911 	    VIF_UNLOCK();
912 	    return error;
913 	}
914     }
915 
916     vifp->v_flags     = vifcp->vifc_flags;
917     vifp->v_threshold = vifcp->vifc_threshold;
918     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
919     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
920     vifp->v_ifp       = ifp;
921     /* initialize per vif pkt counters */
922     vifp->v_pkt_in    = 0;
923     vifp->v_pkt_out   = 0;
924     vifp->v_bytes_in  = 0;
925     vifp->v_bytes_out = 0;
926 
927     /* Adjust numvifs up if the vifi is higher than numvifs */
928     if (V_numvifs <= vifcp->vifc_vifi)
929 	V_numvifs = vifcp->vifc_vifi + 1;
930 
931     VIF_UNLOCK();
932 
933     CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
934 	(int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
935 	(int)vifcp->vifc_threshold);
936 
937     return 0;
938 }
939 
940 /*
941  * Delete a vif from the vif table
942  */
943 static int
944 del_vif_locked(vifi_t vifi)
945 {
946     struct vif *vifp;
947 
948     VIF_LOCK_ASSERT();
949 
950     if (vifi >= V_numvifs) {
951 	return EINVAL;
952     }
953     vifp = &V_viftable[vifi];
954     if (in_nullhost(vifp->v_lcl_addr)) {
955 	return EADDRNOTAVAIL;
956     }
957 
958     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
959 	if_allmulti(vifp->v_ifp, 0);
960 
961     if (vifp->v_flags & VIFF_REGISTER)
962 	V_reg_vif_num = VIFI_INVALID;
963 
964     bzero((caddr_t)vifp, sizeof (*vifp));
965 
966     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
967 
968     /* Adjust numvifs down */
969     for (vifi = V_numvifs; vifi > 0; vifi--)
970 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
971 	    break;
972     V_numvifs = vifi;
973 
974     return 0;
975 }
976 
977 static int
978 del_vif(vifi_t vifi)
979 {
980     int cc;
981 
982     VIF_LOCK();
983     cc = del_vif_locked(vifi);
984     VIF_UNLOCK();
985 
986     return cc;
987 }
988 
989 /*
990  * update an mfc entry without resetting counters and S,G addresses.
991  */
992 static void
993 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
994 {
995     int i;
996 
997     rt->mfc_parent = mfccp->mfcc_parent;
998     for (i = 0; i < V_numvifs; i++) {
999 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1000 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1001 	    MRT_MFC_FLAGS_ALL;
1002     }
1003     /* set the RP address */
1004     if (V_mrt_api_config & MRT_MFC_RP)
1005 	rt->mfc_rp = mfccp->mfcc_rp;
1006     else
1007 	rt->mfc_rp.s_addr = INADDR_ANY;
1008 }
1009 
1010 /*
1011  * fully initialize an mfc entry from the parameter.
1012  */
1013 static void
1014 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1015 {
1016     rt->mfc_origin     = mfccp->mfcc_origin;
1017     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1018 
1019     update_mfc_params(rt, mfccp);
1020 
1021     /* initialize pkt counters per src-grp */
1022     rt->mfc_pkt_cnt    = 0;
1023     rt->mfc_byte_cnt   = 0;
1024     rt->mfc_wrong_if   = 0;
1025     timevalclear(&rt->mfc_last_assert);
1026 }
1027 
1028 static void
1029 expire_mfc(struct mfc *rt)
1030 {
1031 	struct rtdetq *rte, *nrte;
1032 
1033 	MFC_LOCK_ASSERT();
1034 
1035 	free_bw_list(rt->mfc_bw_meter);
1036 
1037 	TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1038 		m_freem(rte->m);
1039 		TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1040 		free(rte, M_MRTABLE);
1041 	}
1042 
1043 	LIST_REMOVE(rt, mfc_hash);
1044 	free(rt, M_MRTABLE);
1045 }
1046 
1047 /*
1048  * Add an mfc entry
1049  */
1050 static int
1051 add_mfc(struct mfcctl2 *mfccp)
1052 {
1053     struct mfc *rt;
1054     struct rtdetq *rte, *nrte;
1055     u_long hash = 0;
1056     u_short nstl;
1057 
1058     VIF_LOCK();
1059     MFC_LOCK();
1060 
1061     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1062 
1063     /* If an entry already exists, just update the fields */
1064     if (rt) {
1065 	CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1066 	    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1067 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1068 	    mfccp->mfcc_parent);
1069 	update_mfc_params(rt, mfccp);
1070 	MFC_UNLOCK();
1071 	VIF_UNLOCK();
1072 	return (0);
1073     }
1074 
1075     /*
1076      * Find the entry for which the upcall was made and update
1077      */
1078     nstl = 0;
1079     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1080     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1081 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1082 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1083 	    !TAILQ_EMPTY(&rt->mfc_stall)) {
1084 		CTR5(KTR_IPMF,
1085 		    "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1086 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1087 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1088 		    mfccp->mfcc_parent,
1089 		    TAILQ_FIRST(&rt->mfc_stall));
1090 		if (nstl++)
1091 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1092 
1093 		init_mfc_params(rt, mfccp);
1094 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1095 		V_nexpire[hash]--;
1096 
1097 		/* Free queued packets, but attempt to forward them first. */
1098 		TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1099 			if (rte->ifp != NULL)
1100 				ip_mdq(rte->m, rte->ifp, rt, -1);
1101 			m_freem(rte->m);
1102 			TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1103 			rt->mfc_nstall--;
1104 			free(rte, M_MRTABLE);
1105 		}
1106 	}
1107     }
1108 
1109     /*
1110      * It is possible that an entry is being inserted without an upcall
1111      */
1112     if (nstl == 0) {
1113 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1114 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1115 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1116 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1117 			init_mfc_params(rt, mfccp);
1118 			if (rt->mfc_expire)
1119 			    V_nexpire[hash]--;
1120 			rt->mfc_expire = 0;
1121 			break; /* XXX */
1122 		}
1123 	}
1124 
1125 	if (rt == NULL) {		/* no upcall, so make a new entry */
1126 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1127 	    if (rt == NULL) {
1128 		MFC_UNLOCK();
1129 		VIF_UNLOCK();
1130 		return (ENOBUFS);
1131 	    }
1132 
1133 	    init_mfc_params(rt, mfccp);
1134 	    TAILQ_INIT(&rt->mfc_stall);
1135 	    rt->mfc_nstall = 0;
1136 
1137 	    rt->mfc_expire     = 0;
1138 	    rt->mfc_bw_meter = NULL;
1139 
1140 	    /* insert new entry at head of hash chain */
1141 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1142 	}
1143     }
1144 
1145     MFC_UNLOCK();
1146     VIF_UNLOCK();
1147 
1148     return (0);
1149 }
1150 
1151 /*
1152  * Delete an mfc entry
1153  */
1154 static int
1155 del_mfc(struct mfcctl2 *mfccp)
1156 {
1157     struct in_addr	origin;
1158     struct in_addr	mcastgrp;
1159     struct mfc		*rt;
1160 
1161     origin = mfccp->mfcc_origin;
1162     mcastgrp = mfccp->mfcc_mcastgrp;
1163 
1164     CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1165 	ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1166 
1167     MFC_LOCK();
1168 
1169     rt = mfc_find(&origin, &mcastgrp);
1170     if (rt == NULL) {
1171 	MFC_UNLOCK();
1172 	return EADDRNOTAVAIL;
1173     }
1174 
1175     /*
1176      * free the bw_meter entries
1177      */
1178     free_bw_list(rt->mfc_bw_meter);
1179     rt->mfc_bw_meter = NULL;
1180 
1181     LIST_REMOVE(rt, mfc_hash);
1182     free(rt, M_MRTABLE);
1183 
1184     MFC_UNLOCK();
1185 
1186     return (0);
1187 }
1188 
1189 /*
1190  * Send a message to the routing daemon on the multicast routing socket.
1191  */
1192 static int
1193 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1194 {
1195     if (s) {
1196 	SOCKBUF_LOCK(&s->so_rcv);
1197 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1198 	    NULL) != 0) {
1199 	    sorwakeup_locked(s);
1200 	    return 0;
1201 	}
1202 	SOCKBUF_UNLOCK(&s->so_rcv);
1203     }
1204     m_freem(mm);
1205     return -1;
1206 }
1207 
1208 /*
1209  * IP multicast forwarding function. This function assumes that the packet
1210  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1211  * pointed to by "ifp", and the packet is to be relayed to other networks
1212  * that have members of the packet's destination IP multicast group.
1213  *
1214  * The packet is returned unscathed to the caller, unless it is
1215  * erroneous, in which case a non-zero return value tells the caller to
1216  * discard it.
1217  */
1218 
1219 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1220 
1221 static int
1222 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1223     struct ip_moptions *imo)
1224 {
1225     struct mfc *rt;
1226     int error;
1227     vifi_t vifi;
1228 
1229     CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1230 	ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1231 
1232     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1233 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1234 	/*
1235 	 * Packet arrived via a physical interface or
1236 	 * an encapsulated tunnel or a register_vif.
1237 	 */
1238     } else {
1239 	/*
1240 	 * Packet arrived through a source-route tunnel.
1241 	 * Source-route tunnels are no longer supported.
1242 	 */
1243 	return (1);
1244     }
1245 
1246     VIF_LOCK();
1247     MFC_LOCK();
1248     if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1249 	if (ip->ip_ttl < MAXTTL)
1250 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1251 	error = ip_mdq(m, ifp, NULL, vifi);
1252 	MFC_UNLOCK();
1253 	VIF_UNLOCK();
1254 	return error;
1255     }
1256 
1257     /*
1258      * Don't forward a packet with time-to-live of zero or one,
1259      * or a packet destined to a local-only group.
1260      */
1261     if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1262 	MFC_UNLOCK();
1263 	VIF_UNLOCK();
1264 	return 0;
1265     }
1266 
1267     /*
1268      * Determine forwarding vifs from the forwarding cache table
1269      */
1270     MRTSTAT_INC(mrts_mfc_lookups);
1271     rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1272 
1273     /* Entry exists, so forward if necessary */
1274     if (rt != NULL) {
1275 	error = ip_mdq(m, ifp, rt, -1);
1276 	MFC_UNLOCK();
1277 	VIF_UNLOCK();
1278 	return error;
1279     } else {
1280 	/*
1281 	 * If we don't have a route for packet's origin,
1282 	 * Make a copy of the packet & send message to routing daemon
1283 	 */
1284 
1285 	struct mbuf *mb0;
1286 	struct rtdetq *rte;
1287 	u_long hash;
1288 	int hlen = ip->ip_hl << 2;
1289 
1290 	MRTSTAT_INC(mrts_mfc_misses);
1291 	MRTSTAT_INC(mrts_no_route);
1292 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1293 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1294 
1295 	/*
1296 	 * Allocate mbufs early so that we don't do extra work if we are
1297 	 * just going to fail anyway.  Make sure to pullup the header so
1298 	 * that other people can't step on it.
1299 	 */
1300 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1301 	    M_NOWAIT|M_ZERO);
1302 	if (rte == NULL) {
1303 	    MFC_UNLOCK();
1304 	    VIF_UNLOCK();
1305 	    return ENOBUFS;
1306 	}
1307 
1308 	mb0 = m_copypacket(m, M_NOWAIT);
1309 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1310 	    mb0 = m_pullup(mb0, hlen);
1311 	if (mb0 == NULL) {
1312 	    free(rte, M_MRTABLE);
1313 	    MFC_UNLOCK();
1314 	    VIF_UNLOCK();
1315 	    return ENOBUFS;
1316 	}
1317 
1318 	/* is there an upcall waiting for this flow ? */
1319 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1320 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1321 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1322 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1323 		    !TAILQ_EMPTY(&rt->mfc_stall))
1324 			break;
1325 	}
1326 
1327 	if (rt == NULL) {
1328 	    int i;
1329 	    struct igmpmsg *im;
1330 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1331 	    struct mbuf *mm;
1332 
1333 	    /*
1334 	     * Locate the vifi for the incoming interface for this packet.
1335 	     * If none found, drop packet.
1336 	     */
1337 	    for (vifi = 0; vifi < V_numvifs &&
1338 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1339 		;
1340 	    if (vifi >= V_numvifs)	/* vif not found, drop packet */
1341 		goto non_fatal;
1342 
1343 	    /* no upcall, so make a new entry */
1344 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1345 	    if (rt == NULL)
1346 		goto fail;
1347 
1348 	    /* Make a copy of the header to send to the user level process */
1349 	    mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1350 	    if (mm == NULL)
1351 		goto fail1;
1352 
1353 	    /*
1354 	     * Send message to routing daemon to install
1355 	     * a route into the kernel table
1356 	     */
1357 
1358 	    im = mtod(mm, struct igmpmsg *);
1359 	    im->im_msgtype = IGMPMSG_NOCACHE;
1360 	    im->im_mbz = 0;
1361 	    im->im_vif = vifi;
1362 
1363 	    MRTSTAT_INC(mrts_upcalls);
1364 
1365 	    k_igmpsrc.sin_addr = ip->ip_src;
1366 	    if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1367 		CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1368 		MRTSTAT_INC(mrts_upq_sockfull);
1369 fail1:
1370 		free(rt, M_MRTABLE);
1371 fail:
1372 		free(rte, M_MRTABLE);
1373 		m_freem(mb0);
1374 		MFC_UNLOCK();
1375 		VIF_UNLOCK();
1376 		return ENOBUFS;
1377 	    }
1378 
1379 	    /* insert new entry at head of hash chain */
1380 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1381 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1382 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1383 	    V_nexpire[hash]++;
1384 	    for (i = 0; i < V_numvifs; i++) {
1385 		rt->mfc_ttls[i] = 0;
1386 		rt->mfc_flags[i] = 0;
1387 	    }
1388 	    rt->mfc_parent = -1;
1389 
1390 	    /* clear the RP address */
1391 	    rt->mfc_rp.s_addr = INADDR_ANY;
1392 	    rt->mfc_bw_meter = NULL;
1393 
1394 	    /* initialize pkt counters per src-grp */
1395 	    rt->mfc_pkt_cnt = 0;
1396 	    rt->mfc_byte_cnt = 0;
1397 	    rt->mfc_wrong_if = 0;
1398 	    timevalclear(&rt->mfc_last_assert);
1399 
1400 	    TAILQ_INIT(&rt->mfc_stall);
1401 	    rt->mfc_nstall = 0;
1402 
1403 	    /* link into table */
1404 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1405 	    TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1406 	    rt->mfc_nstall++;
1407 
1408 	} else {
1409 	    /* determine if queue has overflowed */
1410 	    if (rt->mfc_nstall > MAX_UPQ) {
1411 		MRTSTAT_INC(mrts_upq_ovflw);
1412 non_fatal:
1413 		free(rte, M_MRTABLE);
1414 		m_freem(mb0);
1415 		MFC_UNLOCK();
1416 		VIF_UNLOCK();
1417 		return (0);
1418 	    }
1419 	    TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1420 	    rt->mfc_nstall++;
1421 	}
1422 
1423 	rte->m			= mb0;
1424 	rte->ifp		= ifp;
1425 
1426 	MFC_UNLOCK();
1427 	VIF_UNLOCK();
1428 
1429 	return 0;
1430     }
1431 }
1432 
1433 /*
1434  * Clean up the cache entry if upcall is not serviced
1435  */
1436 static void
1437 expire_upcalls(void *arg)
1438 {
1439     u_long i;
1440 
1441     CURVNET_SET((struct vnet *) arg);
1442 
1443     MFC_LOCK();
1444 
1445     for (i = 0; i < mfchashsize; i++) {
1446 	struct mfc *rt, *nrt;
1447 
1448 	if (V_nexpire[i] == 0)
1449 	    continue;
1450 
1451 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1452 		if (TAILQ_EMPTY(&rt->mfc_stall))
1453 			continue;
1454 
1455 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1456 			continue;
1457 
1458 		/*
1459 		 * free the bw_meter entries
1460 		 */
1461 		while (rt->mfc_bw_meter != NULL) {
1462 		    struct bw_meter *x = rt->mfc_bw_meter;
1463 
1464 		    rt->mfc_bw_meter = x->bm_mfc_next;
1465 		    free(x, M_BWMETER);
1466 		}
1467 
1468 		MRTSTAT_INC(mrts_cache_cleanups);
1469 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1470 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1471 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1472 
1473 		expire_mfc(rt);
1474 	    }
1475     }
1476 
1477     MFC_UNLOCK();
1478 
1479     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1480 	curvnet);
1481 
1482     CURVNET_RESTORE();
1483 }
1484 
1485 /*
1486  * Packet forwarding routine once entry in the cache is made
1487  */
1488 static int
1489 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1490 {
1491     struct ip  *ip = mtod(m, struct ip *);
1492     vifi_t vifi;
1493     int plen = ntohs(ip->ip_len);
1494 
1495     VIF_LOCK_ASSERT();
1496 
1497     /*
1498      * If xmt_vif is not -1, send on only the requested vif.
1499      *
1500      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1501      */
1502     if (xmt_vif < V_numvifs) {
1503 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1504 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1505 	else
1506 		phyint_send(ip, V_viftable + xmt_vif, m);
1507 	return 1;
1508     }
1509 
1510     /*
1511      * Don't forward if it didn't arrive from the parent vif for its origin.
1512      */
1513     vifi = rt->mfc_parent;
1514     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1515 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1516 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1517 	MRTSTAT_INC(mrts_wrong_if);
1518 	++rt->mfc_wrong_if;
1519 	/*
1520 	 * If we are doing PIM assert processing, send a message
1521 	 * to the routing daemon.
1522 	 *
1523 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1524 	 * can complete the SPT switch, regardless of the type
1525 	 * of the iif (broadcast media, GRE tunnel, etc).
1526 	 */
1527 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1528 	    V_viftable[vifi].v_ifp) {
1529 
1530 	    if (ifp == &V_multicast_register_if)
1531 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1532 
1533 	    /* Get vifi for the incoming packet */
1534 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1535 		vifi++)
1536 		;
1537 	    if (vifi >= V_numvifs)
1538 		return 0;	/* The iif is not found: ignore the packet. */
1539 
1540 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1541 		return 0;	/* WRONGVIF disabled: ignore the packet */
1542 
1543 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1544 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1545 		struct igmpmsg *im;
1546 		int hlen = ip->ip_hl << 2;
1547 		struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1548 
1549 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1550 		    mm = m_pullup(mm, hlen);
1551 		if (mm == NULL)
1552 		    return ENOBUFS;
1553 
1554 		im = mtod(mm, struct igmpmsg *);
1555 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1556 		im->im_mbz		= 0;
1557 		im->im_vif		= vifi;
1558 
1559 		MRTSTAT_INC(mrts_upcalls);
1560 
1561 		k_igmpsrc.sin_addr = im->im_src;
1562 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1563 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1564 		    MRTSTAT_INC(mrts_upq_sockfull);
1565 		    return ENOBUFS;
1566 		}
1567 	    }
1568 	}
1569 	return 0;
1570     }
1571 
1572 
1573     /* If I sourced this packet, it counts as output, else it was input. */
1574     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1575 	V_viftable[vifi].v_pkt_out++;
1576 	V_viftable[vifi].v_bytes_out += plen;
1577     } else {
1578 	V_viftable[vifi].v_pkt_in++;
1579 	V_viftable[vifi].v_bytes_in += plen;
1580     }
1581     rt->mfc_pkt_cnt++;
1582     rt->mfc_byte_cnt += plen;
1583 
1584     /*
1585      * For each vif, decide if a copy of the packet should be forwarded.
1586      * Forward if:
1587      *		- the ttl exceeds the vif's threshold
1588      *		- there are group members downstream on interface
1589      */
1590     for (vifi = 0; vifi < V_numvifs; vifi++)
1591 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1592 	    V_viftable[vifi].v_pkt_out++;
1593 	    V_viftable[vifi].v_bytes_out += plen;
1594 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1595 		pim_register_send(ip, V_viftable + vifi, m, rt);
1596 	    else
1597 		phyint_send(ip, V_viftable + vifi, m);
1598 	}
1599 
1600     /*
1601      * Perform upcall-related bw measuring.
1602      */
1603     if (rt->mfc_bw_meter != NULL) {
1604 	struct bw_meter *x;
1605 	struct timeval now;
1606 
1607 	microtime(&now);
1608 	MFC_LOCK_ASSERT();
1609 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1610 	    bw_meter_receive_packet(x, plen, &now);
1611     }
1612 
1613     return 0;
1614 }
1615 
1616 /*
1617  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1618  */
1619 static int
1620 X_legal_vif_num(int vif)
1621 {
1622 	int ret;
1623 
1624 	ret = 0;
1625 	if (vif < 0)
1626 		return (ret);
1627 
1628 	VIF_LOCK();
1629 	if (vif < V_numvifs)
1630 		ret = 1;
1631 	VIF_UNLOCK();
1632 
1633 	return (ret);
1634 }
1635 
1636 /*
1637  * Return the local address used by this vif
1638  */
1639 static u_long
1640 X_ip_mcast_src(int vifi)
1641 {
1642 	in_addr_t addr;
1643 
1644 	addr = INADDR_ANY;
1645 	if (vifi < 0)
1646 		return (addr);
1647 
1648 	VIF_LOCK();
1649 	if (vifi < V_numvifs)
1650 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1651 	VIF_UNLOCK();
1652 
1653 	return (addr);
1654 }
1655 
1656 static void
1657 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1658 {
1659     struct mbuf *mb_copy;
1660     int hlen = ip->ip_hl << 2;
1661 
1662     VIF_LOCK_ASSERT();
1663 
1664     /*
1665      * Make a new reference to the packet; make sure that
1666      * the IP header is actually copied, not just referenced,
1667      * so that ip_output() only scribbles on the copy.
1668      */
1669     mb_copy = m_copypacket(m, M_NOWAIT);
1670     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1671 	mb_copy = m_pullup(mb_copy, hlen);
1672     if (mb_copy == NULL)
1673 	return;
1674 
1675     send_packet(vifp, mb_copy);
1676 }
1677 
1678 static void
1679 send_packet(struct vif *vifp, struct mbuf *m)
1680 {
1681 	struct ip_moptions imo;
1682 	int error __unused;
1683 
1684 	VIF_LOCK_ASSERT();
1685 
1686 	imo.imo_multicast_ifp  = vifp->v_ifp;
1687 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1688 	imo.imo_multicast_loop = 1;
1689 	imo.imo_multicast_vif  = -1;
1690 	STAILQ_INIT(&imo.imo_head);
1691 
1692 	/*
1693 	 * Re-entrancy should not be a problem here, because
1694 	 * the packets that we send out and are looped back at us
1695 	 * should get rejected because they appear to come from
1696 	 * the loopback interface, thus preventing looping.
1697 	 */
1698 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1699 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1700 	    (ptrdiff_t)(vifp - V_viftable), error);
1701 }
1702 
1703 /*
1704  * Stubs for old RSVP socket shim implementation.
1705  */
1706 
1707 static int
1708 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1709 {
1710 
1711 	return (EOPNOTSUPP);
1712 }
1713 
1714 static void
1715 X_ip_rsvp_force_done(struct socket *so __unused)
1716 {
1717 
1718 }
1719 
1720 static int
1721 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1722 {
1723 	struct mbuf *m;
1724 
1725 	m = *mp;
1726 	*mp = NULL;
1727 	if (!V_rsvp_on)
1728 		m_freem(m);
1729 	return (IPPROTO_DONE);
1730 }
1731 
1732 /*
1733  * Code for bandwidth monitors
1734  */
1735 
1736 /*
1737  * Define common interface for timeval-related methods
1738  */
1739 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1740 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1741 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1742 
1743 static uint32_t
1744 compute_bw_meter_flags(struct bw_upcall *req)
1745 {
1746     uint32_t flags = 0;
1747 
1748     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1749 	flags |= BW_METER_UNIT_PACKETS;
1750     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1751 	flags |= BW_METER_UNIT_BYTES;
1752     if (req->bu_flags & BW_UPCALL_GEQ)
1753 	flags |= BW_METER_GEQ;
1754     if (req->bu_flags & BW_UPCALL_LEQ)
1755 	flags |= BW_METER_LEQ;
1756 
1757     return flags;
1758 }
1759 
1760 /*
1761  * Add a bw_meter entry
1762  */
1763 static int
1764 add_bw_upcall(struct bw_upcall *req)
1765 {
1766     struct mfc *mfc;
1767     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1768 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1769     struct timeval now;
1770     struct bw_meter *x;
1771     uint32_t flags;
1772 
1773     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1774 	return EOPNOTSUPP;
1775 
1776     /* Test if the flags are valid */
1777     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1778 	return EINVAL;
1779     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1780 	return EINVAL;
1781     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1782 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1783 	return EINVAL;
1784 
1785     /* Test if the threshold time interval is valid */
1786     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1787 	return EINVAL;
1788 
1789     flags = compute_bw_meter_flags(req);
1790 
1791     /*
1792      * Find if we have already same bw_meter entry
1793      */
1794     MFC_LOCK();
1795     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1796     if (mfc == NULL) {
1797 	MFC_UNLOCK();
1798 	return EADDRNOTAVAIL;
1799     }
1800     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1801 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1802 			   &req->bu_threshold.b_time, ==)) &&
1803 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1804 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1805 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
1806 	    MFC_UNLOCK();
1807 	    return 0;		/* XXX Already installed */
1808 	}
1809     }
1810 
1811     /* Allocate the new bw_meter entry */
1812     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1813     if (x == NULL) {
1814 	MFC_UNLOCK();
1815 	return ENOBUFS;
1816     }
1817 
1818     /* Set the new bw_meter entry */
1819     x->bm_threshold.b_time = req->bu_threshold.b_time;
1820     microtime(&now);
1821     x->bm_start_time = now;
1822     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1823     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1824     x->bm_measured.b_packets = 0;
1825     x->bm_measured.b_bytes = 0;
1826     x->bm_flags = flags;
1827     x->bm_time_next = NULL;
1828     x->bm_time_hash = BW_METER_BUCKETS;
1829 
1830     /* Add the new bw_meter entry to the front of entries for this MFC */
1831     x->bm_mfc = mfc;
1832     x->bm_mfc_next = mfc->mfc_bw_meter;
1833     mfc->mfc_bw_meter = x;
1834     schedule_bw_meter(x, &now);
1835     MFC_UNLOCK();
1836 
1837     return 0;
1838 }
1839 
1840 static void
1841 free_bw_list(struct bw_meter *list)
1842 {
1843     while (list != NULL) {
1844 	struct bw_meter *x = list;
1845 
1846 	list = list->bm_mfc_next;
1847 	unschedule_bw_meter(x);
1848 	free(x, M_BWMETER);
1849     }
1850 }
1851 
1852 /*
1853  * Delete one or multiple bw_meter entries
1854  */
1855 static int
1856 del_bw_upcall(struct bw_upcall *req)
1857 {
1858     struct mfc *mfc;
1859     struct bw_meter *x;
1860 
1861     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1862 	return EOPNOTSUPP;
1863 
1864     MFC_LOCK();
1865 
1866     /* Find the corresponding MFC entry */
1867     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1868     if (mfc == NULL) {
1869 	MFC_UNLOCK();
1870 	return EADDRNOTAVAIL;
1871     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1872 	/*
1873 	 * Delete all bw_meter entries for this mfc
1874 	 */
1875 	struct bw_meter *list;
1876 
1877 	list = mfc->mfc_bw_meter;
1878 	mfc->mfc_bw_meter = NULL;
1879 	free_bw_list(list);
1880 	MFC_UNLOCK();
1881 	return 0;
1882     } else {			/* Delete a single bw_meter entry */
1883 	struct bw_meter *prev;
1884 	uint32_t flags = 0;
1885 
1886 	flags = compute_bw_meter_flags(req);
1887 
1888 	/* Find the bw_meter entry to delete */
1889 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1890 	     prev = x, x = x->bm_mfc_next) {
1891 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1892 			       &req->bu_threshold.b_time, ==)) &&
1893 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1894 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1895 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
1896 		break;
1897 	}
1898 	if (x != NULL) { /* Delete entry from the list for this MFC */
1899 	    if (prev != NULL)
1900 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
1901 	    else
1902 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1903 
1904 	    unschedule_bw_meter(x);
1905 	    MFC_UNLOCK();
1906 	    /* Free the bw_meter entry */
1907 	    free(x, M_BWMETER);
1908 	    return 0;
1909 	} else {
1910 	    MFC_UNLOCK();
1911 	    return EINVAL;
1912 	}
1913     }
1914     /* NOTREACHED */
1915 }
1916 
1917 /*
1918  * Perform bandwidth measurement processing that may result in an upcall
1919  */
1920 static void
1921 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1922 {
1923     struct timeval delta;
1924 
1925     MFC_LOCK_ASSERT();
1926 
1927     delta = *nowp;
1928     BW_TIMEVALDECR(&delta, &x->bm_start_time);
1929 
1930     if (x->bm_flags & BW_METER_GEQ) {
1931 	/*
1932 	 * Processing for ">=" type of bw_meter entry
1933 	 */
1934 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1935 	    /* Reset the bw_meter entry */
1936 	    x->bm_start_time = *nowp;
1937 	    x->bm_measured.b_packets = 0;
1938 	    x->bm_measured.b_bytes = 0;
1939 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1940 	}
1941 
1942 	/* Record that a packet is received */
1943 	x->bm_measured.b_packets++;
1944 	x->bm_measured.b_bytes += plen;
1945 
1946 	/*
1947 	 * Test if we should deliver an upcall
1948 	 */
1949 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1950 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1951 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1952 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1953 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1954 		/* Prepare an upcall for delivery */
1955 		bw_meter_prepare_upcall(x, nowp);
1956 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1957 	    }
1958 	}
1959     } else if (x->bm_flags & BW_METER_LEQ) {
1960 	/*
1961 	 * Processing for "<=" type of bw_meter entry
1962 	 */
1963 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1964 	    /*
1965 	     * We are behind time with the multicast forwarding table
1966 	     * scanning for "<=" type of bw_meter entries, so test now
1967 	     * if we should deliver an upcall.
1968 	     */
1969 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1970 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1971 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
1972 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1973 		/* Prepare an upcall for delivery */
1974 		bw_meter_prepare_upcall(x, nowp);
1975 	    }
1976 	    /* Reschedule the bw_meter entry */
1977 	    unschedule_bw_meter(x);
1978 	    schedule_bw_meter(x, nowp);
1979 	}
1980 
1981 	/* Record that a packet is received */
1982 	x->bm_measured.b_packets++;
1983 	x->bm_measured.b_bytes += plen;
1984 
1985 	/*
1986 	 * Test if we should restart the measuring interval
1987 	 */
1988 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1989 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1990 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
1991 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1992 	    /* Don't restart the measuring interval */
1993 	} else {
1994 	    /* Do restart the measuring interval */
1995 	    /*
1996 	     * XXX: note that we don't unschedule and schedule, because this
1997 	     * might be too much overhead per packet. Instead, when we process
1998 	     * all entries for a given timer hash bin, we check whether it is
1999 	     * really a timeout. If not, we reschedule at that time.
2000 	     */
2001 	    x->bm_start_time = *nowp;
2002 	    x->bm_measured.b_packets = 0;
2003 	    x->bm_measured.b_bytes = 0;
2004 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2005 	}
2006     }
2007 }
2008 
2009 /*
2010  * Prepare a bandwidth-related upcall
2011  */
2012 static void
2013 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2014 {
2015     struct timeval delta;
2016     struct bw_upcall *u;
2017 
2018     MFC_LOCK_ASSERT();
2019 
2020     /*
2021      * Compute the measured time interval
2022      */
2023     delta = *nowp;
2024     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2025 
2026     /*
2027      * If there are too many pending upcalls, deliver them now
2028      */
2029     if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2030 	bw_upcalls_send();
2031 
2032     /*
2033      * Set the bw_upcall entry
2034      */
2035     u = &V_bw_upcalls[V_bw_upcalls_n++];
2036     u->bu_src = x->bm_mfc->mfc_origin;
2037     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2038     u->bu_threshold.b_time = x->bm_threshold.b_time;
2039     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2040     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2041     u->bu_measured.b_time = delta;
2042     u->bu_measured.b_packets = x->bm_measured.b_packets;
2043     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2044     u->bu_flags = 0;
2045     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2046 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2047     if (x->bm_flags & BW_METER_UNIT_BYTES)
2048 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2049     if (x->bm_flags & BW_METER_GEQ)
2050 	u->bu_flags |= BW_UPCALL_GEQ;
2051     if (x->bm_flags & BW_METER_LEQ)
2052 	u->bu_flags |= BW_UPCALL_LEQ;
2053 }
2054 
2055 /*
2056  * Send the pending bandwidth-related upcalls
2057  */
2058 static void
2059 bw_upcalls_send(void)
2060 {
2061     struct mbuf *m;
2062     int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2063     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2064     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2065 				      0,		/* unused2 */
2066 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2067 				      0,		/* im_mbz  */
2068 				      0,		/* im_vif  */
2069 				      0,		/* unused3 */
2070 				      { 0 },		/* im_src  */
2071 				      { 0 } };		/* im_dst  */
2072 
2073     MFC_LOCK_ASSERT();
2074 
2075     if (V_bw_upcalls_n == 0)
2076 	return;			/* No pending upcalls */
2077 
2078     V_bw_upcalls_n = 0;
2079 
2080     /*
2081      * Allocate a new mbuf, initialize it with the header and
2082      * the payload for the pending calls.
2083      */
2084     m = m_gethdr(M_NOWAIT, MT_DATA);
2085     if (m == NULL) {
2086 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2087 	return;
2088     }
2089 
2090     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2091     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2092 
2093     /*
2094      * Send the upcalls
2095      * XXX do we need to set the address in k_igmpsrc ?
2096      */
2097     MRTSTAT_INC(mrts_upcalls);
2098     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2099 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2100 	MRTSTAT_INC(mrts_upq_sockfull);
2101     }
2102 }
2103 
2104 /*
2105  * Compute the timeout hash value for the bw_meter entries
2106  */
2107 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2108     do {								\
2109 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2110 									\
2111 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2112 	(hash) = next_timeval.tv_sec;					\
2113 	if (next_timeval.tv_usec)					\
2114 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2115 	(hash) %= BW_METER_BUCKETS;					\
2116     } while (0)
2117 
2118 /*
2119  * Schedule a timer to process periodically bw_meter entry of type "<="
2120  * by linking the entry in the proper hash bucket.
2121  */
2122 static void
2123 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2124 {
2125     int time_hash;
2126 
2127     MFC_LOCK_ASSERT();
2128 
2129     if (!(x->bm_flags & BW_METER_LEQ))
2130 	return;		/* XXX: we schedule timers only for "<=" entries */
2131 
2132     /*
2133      * Reset the bw_meter entry
2134      */
2135     x->bm_start_time = *nowp;
2136     x->bm_measured.b_packets = 0;
2137     x->bm_measured.b_bytes = 0;
2138     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2139 
2140     /*
2141      * Compute the timeout hash value and insert the entry
2142      */
2143     BW_METER_TIMEHASH(x, time_hash);
2144     x->bm_time_next = V_bw_meter_timers[time_hash];
2145     V_bw_meter_timers[time_hash] = x;
2146     x->bm_time_hash = time_hash;
2147 }
2148 
2149 /*
2150  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2151  * by removing the entry from the proper hash bucket.
2152  */
2153 static void
2154 unschedule_bw_meter(struct bw_meter *x)
2155 {
2156     int time_hash;
2157     struct bw_meter *prev, *tmp;
2158 
2159     MFC_LOCK_ASSERT();
2160 
2161     if (!(x->bm_flags & BW_METER_LEQ))
2162 	return;		/* XXX: we schedule timers only for "<=" entries */
2163 
2164     /*
2165      * Compute the timeout hash value and delete the entry
2166      */
2167     time_hash = x->bm_time_hash;
2168     if (time_hash >= BW_METER_BUCKETS)
2169 	return;		/* Entry was not scheduled */
2170 
2171     for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2172 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2173 	if (tmp == x)
2174 	    break;
2175 
2176     if (tmp == NULL)
2177 	panic("unschedule_bw_meter: bw_meter entry not found");
2178 
2179     if (prev != NULL)
2180 	prev->bm_time_next = x->bm_time_next;
2181     else
2182 	V_bw_meter_timers[time_hash] = x->bm_time_next;
2183 
2184     x->bm_time_next = NULL;
2185     x->bm_time_hash = BW_METER_BUCKETS;
2186 }
2187 
2188 
2189 /*
2190  * Process all "<=" type of bw_meter that should be processed now,
2191  * and for each entry prepare an upcall if necessary. Each processed
2192  * entry is rescheduled again for the (periodic) processing.
2193  *
2194  * This is run periodically (once per second normally). On each round,
2195  * all the potentially matching entries are in the hash slot that we are
2196  * looking at.
2197  */
2198 static void
2199 bw_meter_process()
2200 {
2201     uint32_t loops;
2202     int i;
2203     struct timeval now, process_endtime;
2204 
2205     microtime(&now);
2206     if (V_last_tv_sec == now.tv_sec)
2207 	return;		/* nothing to do */
2208 
2209     loops = now.tv_sec - V_last_tv_sec;
2210     V_last_tv_sec = now.tv_sec;
2211     if (loops > BW_METER_BUCKETS)
2212 	loops = BW_METER_BUCKETS;
2213 
2214     MFC_LOCK();
2215     /*
2216      * Process all bins of bw_meter entries from the one after the last
2217      * processed to the current one. On entry, i points to the last bucket
2218      * visited, so we need to increment i at the beginning of the loop.
2219      */
2220     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2221 	struct bw_meter *x, *tmp_list;
2222 
2223 	if (++i >= BW_METER_BUCKETS)
2224 	    i = 0;
2225 
2226 	/* Disconnect the list of bw_meter entries from the bin */
2227 	tmp_list = V_bw_meter_timers[i];
2228 	V_bw_meter_timers[i] = NULL;
2229 
2230 	/* Process the list of bw_meter entries */
2231 	while (tmp_list != NULL) {
2232 	    x = tmp_list;
2233 	    tmp_list = tmp_list->bm_time_next;
2234 
2235 	    /* Test if the time interval is over */
2236 	    process_endtime = x->bm_start_time;
2237 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2238 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2239 		/* Not yet: reschedule, but don't reset */
2240 		int time_hash;
2241 
2242 		BW_METER_TIMEHASH(x, time_hash);
2243 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2244 		    /*
2245 		     * XXX: somehow the bin processing is a bit ahead of time.
2246 		     * Put the entry in the next bin.
2247 		     */
2248 		    if (++time_hash >= BW_METER_BUCKETS)
2249 			time_hash = 0;
2250 		}
2251 		x->bm_time_next = V_bw_meter_timers[time_hash];
2252 		V_bw_meter_timers[time_hash] = x;
2253 		x->bm_time_hash = time_hash;
2254 
2255 		continue;
2256 	    }
2257 
2258 	    /*
2259 	     * Test if we should deliver an upcall
2260 	     */
2261 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2262 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2263 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2264 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2265 		/* Prepare an upcall for delivery */
2266 		bw_meter_prepare_upcall(x, &now);
2267 	    }
2268 
2269 	    /*
2270 	     * Reschedule for next processing
2271 	     */
2272 	    schedule_bw_meter(x, &now);
2273 	}
2274     }
2275 
2276     /* Send all upcalls that are pending delivery */
2277     bw_upcalls_send();
2278 
2279     MFC_UNLOCK();
2280 }
2281 
2282 /*
2283  * A periodic function for sending all upcalls that are pending delivery
2284  */
2285 static void
2286 expire_bw_upcalls_send(void *arg)
2287 {
2288     CURVNET_SET((struct vnet *) arg);
2289 
2290     MFC_LOCK();
2291     bw_upcalls_send();
2292     MFC_UNLOCK();
2293 
2294     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2295 	curvnet);
2296     CURVNET_RESTORE();
2297 }
2298 
2299 /*
2300  * A periodic function for periodic scanning of the multicast forwarding
2301  * table for processing all "<=" bw_meter entries.
2302  */
2303 static void
2304 expire_bw_meter_process(void *arg)
2305 {
2306     CURVNET_SET((struct vnet *) arg);
2307 
2308     if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2309 	bw_meter_process();
2310 
2311     callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2312 	curvnet);
2313     CURVNET_RESTORE();
2314 }
2315 
2316 /*
2317  * End of bandwidth monitoring code
2318  */
2319 
2320 /*
2321  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2322  *
2323  */
2324 static int
2325 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2326     struct mfc *rt)
2327 {
2328     struct mbuf *mb_copy, *mm;
2329 
2330     /*
2331      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2332      * rendezvous point was unspecified, and we were told not to.
2333      */
2334     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2335 	in_nullhost(rt->mfc_rp))
2336 	return 0;
2337 
2338     mb_copy = pim_register_prepare(ip, m);
2339     if (mb_copy == NULL)
2340 	return ENOBUFS;
2341 
2342     /*
2343      * Send all the fragments. Note that the mbuf for each fragment
2344      * is freed by the sending machinery.
2345      */
2346     for (mm = mb_copy; mm; mm = mb_copy) {
2347 	mb_copy = mm->m_nextpkt;
2348 	mm->m_nextpkt = 0;
2349 	mm = m_pullup(mm, sizeof(struct ip));
2350 	if (mm != NULL) {
2351 	    ip = mtod(mm, struct ip *);
2352 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2353 		pim_register_send_rp(ip, vifp, mm, rt);
2354 	    } else {
2355 		pim_register_send_upcall(ip, vifp, mm, rt);
2356 	    }
2357 	}
2358     }
2359 
2360     return 0;
2361 }
2362 
2363 /*
2364  * Return a copy of the data packet that is ready for PIM Register
2365  * encapsulation.
2366  * XXX: Note that in the returned copy the IP header is a valid one.
2367  */
2368 static struct mbuf *
2369 pim_register_prepare(struct ip *ip, struct mbuf *m)
2370 {
2371     struct mbuf *mb_copy = NULL;
2372     int mtu;
2373 
2374     /* Take care of delayed checksums */
2375     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2376 	in_delayed_cksum(m);
2377 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2378     }
2379 
2380     /*
2381      * Copy the old packet & pullup its IP header into the
2382      * new mbuf so we can modify it.
2383      */
2384     mb_copy = m_copypacket(m, M_NOWAIT);
2385     if (mb_copy == NULL)
2386 	return NULL;
2387     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2388     if (mb_copy == NULL)
2389 	return NULL;
2390 
2391     /* take care of the TTL */
2392     ip = mtod(mb_copy, struct ip *);
2393     --ip->ip_ttl;
2394 
2395     /* Compute the MTU after the PIM Register encapsulation */
2396     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2397 
2398     if (ntohs(ip->ip_len) <= mtu) {
2399 	/* Turn the IP header into a valid one */
2400 	ip->ip_sum = 0;
2401 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2402     } else {
2403 	/* Fragment the packet */
2404 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2405 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2406 	    m_freem(mb_copy);
2407 	    return NULL;
2408 	}
2409     }
2410     return mb_copy;
2411 }
2412 
2413 /*
2414  * Send an upcall with the data packet to the user-level process.
2415  */
2416 static int
2417 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2418     struct mbuf *mb_copy, struct mfc *rt)
2419 {
2420     struct mbuf *mb_first;
2421     int len = ntohs(ip->ip_len);
2422     struct igmpmsg *im;
2423     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2424 
2425     VIF_LOCK_ASSERT();
2426 
2427     /*
2428      * Add a new mbuf with an upcall header
2429      */
2430     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2431     if (mb_first == NULL) {
2432 	m_freem(mb_copy);
2433 	return ENOBUFS;
2434     }
2435     mb_first->m_data += max_linkhdr;
2436     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2437     mb_first->m_len = sizeof(struct igmpmsg);
2438     mb_first->m_next = mb_copy;
2439 
2440     /* Send message to routing daemon */
2441     im = mtod(mb_first, struct igmpmsg *);
2442     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2443     im->im_mbz		= 0;
2444     im->im_vif		= vifp - V_viftable;
2445     im->im_src		= ip->ip_src;
2446     im->im_dst		= ip->ip_dst;
2447 
2448     k_igmpsrc.sin_addr	= ip->ip_src;
2449 
2450     MRTSTAT_INC(mrts_upcalls);
2451 
2452     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2453 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2454 	MRTSTAT_INC(mrts_upq_sockfull);
2455 	return ENOBUFS;
2456     }
2457 
2458     /* Keep statistics */
2459     PIMSTAT_INC(pims_snd_registers_msgs);
2460     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2461 
2462     return 0;
2463 }
2464 
2465 /*
2466  * Encapsulate the data packet in PIM Register message and send it to the RP.
2467  */
2468 static int
2469 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2470     struct mfc *rt)
2471 {
2472     struct mbuf *mb_first;
2473     struct ip *ip_outer;
2474     struct pim_encap_pimhdr *pimhdr;
2475     int len = ntohs(ip->ip_len);
2476     vifi_t vifi = rt->mfc_parent;
2477 
2478     VIF_LOCK_ASSERT();
2479 
2480     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2481 	m_freem(mb_copy);
2482 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2483     }
2484 
2485     /*
2486      * Add a new mbuf with the encapsulating header
2487      */
2488     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2489     if (mb_first == NULL) {
2490 	m_freem(mb_copy);
2491 	return ENOBUFS;
2492     }
2493     mb_first->m_data += max_linkhdr;
2494     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2495     mb_first->m_next = mb_copy;
2496 
2497     mb_first->m_pkthdr.len = len + mb_first->m_len;
2498 
2499     /*
2500      * Fill in the encapsulating IP and PIM header
2501      */
2502     ip_outer = mtod(mb_first, struct ip *);
2503     *ip_outer = pim_encap_iphdr;
2504     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2505 	sizeof(pim_encap_pimhdr));
2506     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2507     ip_outer->ip_dst = rt->mfc_rp;
2508     /*
2509      * Copy the inner header TOS to the outer header, and take care of the
2510      * IP_DF bit.
2511      */
2512     ip_outer->ip_tos = ip->ip_tos;
2513     if (ip->ip_off & htons(IP_DF))
2514 	ip_outer->ip_off |= htons(IP_DF);
2515     ip_fillid(ip_outer);
2516     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2517 					 + sizeof(pim_encap_iphdr));
2518     *pimhdr = pim_encap_pimhdr;
2519     /* If the iif crosses a border, set the Border-bit */
2520     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2521 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2522 
2523     mb_first->m_data += sizeof(pim_encap_iphdr);
2524     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2525     mb_first->m_data -= sizeof(pim_encap_iphdr);
2526 
2527     send_packet(vifp, mb_first);
2528 
2529     /* Keep statistics */
2530     PIMSTAT_INC(pims_snd_registers_msgs);
2531     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2532 
2533     return 0;
2534 }
2535 
2536 /*
2537  * pim_encapcheck() is called by the encap4_input() path at runtime to
2538  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2539  * into the kernel.
2540  */
2541 static int
2542 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2543     int proto __unused, void *arg __unused)
2544 {
2545 
2546     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2547     return (8);		/* claim the datagram. */
2548 }
2549 
2550 /*
2551  * PIM-SMv2 and PIM-DM messages processing.
2552  * Receives and verifies the PIM control messages, and passes them
2553  * up to the listening socket, using rip_input().
2554  * The only message with special processing is the PIM_REGISTER message
2555  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2556  * is passed to if_simloop().
2557  */
2558 static int
2559 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2560 {
2561     struct ip *ip = mtod(m, struct ip *);
2562     struct pim *pim;
2563     int iphlen = off;
2564     int minlen;
2565     int datalen = ntohs(ip->ip_len) - iphlen;
2566     int ip_tos;
2567 
2568     /* Keep statistics */
2569     PIMSTAT_INC(pims_rcv_total_msgs);
2570     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2571 
2572     /*
2573      * Validate lengths
2574      */
2575     if (datalen < PIM_MINLEN) {
2576 	PIMSTAT_INC(pims_rcv_tooshort);
2577 	CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2578 	    __func__, datalen, ntohl(ip->ip_src.s_addr));
2579 	m_freem(m);
2580 	return (IPPROTO_DONE);
2581     }
2582 
2583     /*
2584      * If the packet is at least as big as a REGISTER, go agead
2585      * and grab the PIM REGISTER header size, to avoid another
2586      * possible m_pullup() later.
2587      *
2588      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2589      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2590      */
2591     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2592     /*
2593      * Get the IP and PIM headers in contiguous memory, and
2594      * possibly the PIM REGISTER header.
2595      */
2596     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2597 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2598 	return (IPPROTO_DONE);
2599     }
2600 
2601     /* m_pullup() may have given us a new mbuf so reset ip. */
2602     ip = mtod(m, struct ip *);
2603     ip_tos = ip->ip_tos;
2604 
2605     /* adjust mbuf to point to the PIM header */
2606     m->m_data += iphlen;
2607     m->m_len  -= iphlen;
2608     pim = mtod(m, struct pim *);
2609 
2610     /*
2611      * Validate checksum. If PIM REGISTER, exclude the data packet.
2612      *
2613      * XXX: some older PIMv2 implementations don't make this distinction,
2614      * so for compatibility reason perform the checksum over part of the
2615      * message, and if error, then over the whole message.
2616      */
2617     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2618 	/* do nothing, checksum okay */
2619     } else if (in_cksum(m, datalen)) {
2620 	PIMSTAT_INC(pims_rcv_badsum);
2621 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2622 	m_freem(m);
2623 	return (IPPROTO_DONE);
2624     }
2625 
2626     /* PIM version check */
2627     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2628 	PIMSTAT_INC(pims_rcv_badversion);
2629 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2630 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2631 	m_freem(m);
2632 	return (IPPROTO_DONE);
2633     }
2634 
2635     /* restore mbuf back to the outer IP */
2636     m->m_data -= iphlen;
2637     m->m_len  += iphlen;
2638 
2639     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2640 	/*
2641 	 * Since this is a REGISTER, we'll make a copy of the register
2642 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2643 	 * routing daemon.
2644 	 */
2645 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2646 	struct mbuf *mcp;
2647 	struct ip *encap_ip;
2648 	u_int32_t *reghdr;
2649 	struct ifnet *vifp;
2650 
2651 	VIF_LOCK();
2652 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2653 	    VIF_UNLOCK();
2654 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2655 		(int)V_reg_vif_num);
2656 	    m_freem(m);
2657 	    return (IPPROTO_DONE);
2658 	}
2659 	/* XXX need refcnt? */
2660 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2661 	VIF_UNLOCK();
2662 
2663 	/*
2664 	 * Validate length
2665 	 */
2666 	if (datalen < PIM_REG_MINLEN) {
2667 	    PIMSTAT_INC(pims_rcv_tooshort);
2668 	    PIMSTAT_INC(pims_rcv_badregisters);
2669 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2670 	    m_freem(m);
2671 	    return (IPPROTO_DONE);
2672 	}
2673 
2674 	reghdr = (u_int32_t *)(pim + 1);
2675 	encap_ip = (struct ip *)(reghdr + 1);
2676 
2677 	CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2678 	    __func__, ntohl(encap_ip->ip_src.s_addr),
2679 	    ntohs(encap_ip->ip_len));
2680 
2681 	/* verify the version number of the inner packet */
2682 	if (encap_ip->ip_v != IPVERSION) {
2683 	    PIMSTAT_INC(pims_rcv_badregisters);
2684 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2685 	    m_freem(m);
2686 	    return (IPPROTO_DONE);
2687 	}
2688 
2689 	/* verify the inner packet is destined to a mcast group */
2690 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2691 	    PIMSTAT_INC(pims_rcv_badregisters);
2692 	    CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2693 		ntohl(encap_ip->ip_dst.s_addr));
2694 	    m_freem(m);
2695 	    return (IPPROTO_DONE);
2696 	}
2697 
2698 	/* If a NULL_REGISTER, pass it to the daemon */
2699 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2700 	    goto pim_input_to_daemon;
2701 
2702 	/*
2703 	 * Copy the TOS from the outer IP header to the inner IP header.
2704 	 */
2705 	if (encap_ip->ip_tos != ip_tos) {
2706 	    /* Outer TOS -> inner TOS */
2707 	    encap_ip->ip_tos = ip_tos;
2708 	    /* Recompute the inner header checksum. Sigh... */
2709 
2710 	    /* adjust mbuf to point to the inner IP header */
2711 	    m->m_data += (iphlen + PIM_MINLEN);
2712 	    m->m_len  -= (iphlen + PIM_MINLEN);
2713 
2714 	    encap_ip->ip_sum = 0;
2715 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2716 
2717 	    /* restore mbuf to point back to the outer IP header */
2718 	    m->m_data -= (iphlen + PIM_MINLEN);
2719 	    m->m_len  += (iphlen + PIM_MINLEN);
2720 	}
2721 
2722 	/*
2723 	 * Decapsulate the inner IP packet and loopback to forward it
2724 	 * as a normal multicast packet. Also, make a copy of the
2725 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2726 	 * to pass to the daemon later, so it can take the appropriate
2727 	 * actions (e.g., send back PIM_REGISTER_STOP).
2728 	 * XXX: here m->m_data points to the outer IP header.
2729 	 */
2730 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2731 	if (mcp == NULL) {
2732 	    CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2733 	    m_freem(m);
2734 	    return (IPPROTO_DONE);
2735 	}
2736 
2737 	/* Keep statistics */
2738 	/* XXX: registers_bytes include only the encap. mcast pkt */
2739 	PIMSTAT_INC(pims_rcv_registers_msgs);
2740 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2741 
2742 	/*
2743 	 * forward the inner ip packet; point m_data at the inner ip.
2744 	 */
2745 	m_adj(m, iphlen + PIM_MINLEN);
2746 
2747 	CTR4(KTR_IPMF,
2748 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2749 	    __func__,
2750 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2751 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2752 	    (int)V_reg_vif_num);
2753 
2754 	/* NB: vifp was collected above; can it change on us? */
2755 	if_simloop(vifp, m, dst.sin_family, 0);
2756 
2757 	/* prepare the register head to send to the mrouting daemon */
2758 	m = mcp;
2759     }
2760 
2761 pim_input_to_daemon:
2762     /*
2763      * Pass the PIM message up to the daemon; if it is a Register message,
2764      * pass the 'head' only up to the daemon. This includes the
2765      * outer IP header, PIM header, PIM-Register header and the
2766      * inner IP header.
2767      * XXX: the outer IP header pkt size of a Register is not adjust to
2768      * reflect the fact that the inner multicast data is truncated.
2769      */
2770     return (rip_input(&m, &off, proto));
2771 }
2772 
2773 static int
2774 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2775 {
2776 	struct mfc	*rt;
2777 	int		 error, i;
2778 
2779 	if (req->newptr)
2780 		return (EPERM);
2781 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2782 		return (0);
2783 	error = sysctl_wire_old_buffer(req, 0);
2784 	if (error)
2785 		return (error);
2786 
2787 	MFC_LOCK();
2788 	for (i = 0; i < mfchashsize; i++) {
2789 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2790 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2791 			if (error)
2792 				goto out_locked;
2793 		}
2794 	}
2795 out_locked:
2796 	MFC_UNLOCK();
2797 	return (error);
2798 }
2799 
2800 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2801     sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2802     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2803 
2804 static void
2805 vnet_mroute_init(const void *unused __unused)
2806 {
2807 
2808 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2809 
2810 	V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2811 	    M_MRTABLE, M_WAITOK|M_ZERO);
2812 	V_bw_meter_timers = mallocarray(BW_METER_BUCKETS,
2813 	    sizeof(*V_bw_meter_timers), M_MRTABLE, M_WAITOK|M_ZERO);
2814 	V_bw_upcalls = mallocarray(BW_UPCALLS_MAX, sizeof(*V_bw_upcalls),
2815 	    M_MRTABLE, M_WAITOK|M_ZERO);
2816 
2817 	callout_init(&V_expire_upcalls_ch, 1);
2818 	callout_init(&V_bw_upcalls_ch, 1);
2819 	callout_init(&V_bw_meter_ch, 1);
2820 }
2821 
2822 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2823 	NULL);
2824 
2825 static void
2826 vnet_mroute_uninit(const void *unused __unused)
2827 {
2828 
2829 	free(V_bw_upcalls, M_MRTABLE);
2830 	free(V_bw_meter_timers, M_MRTABLE);
2831 	free(V_viftable, M_MRTABLE);
2832 	free(V_nexpire, M_MRTABLE);
2833 	V_nexpire = NULL;
2834 }
2835 
2836 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2837 	vnet_mroute_uninit, NULL);
2838 
2839 static int
2840 ip_mroute_modevent(module_t mod, int type, void *unused)
2841 {
2842 
2843     switch (type) {
2844     case MOD_LOAD:
2845 	MROUTER_LOCK_INIT();
2846 
2847 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2848 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2849 	if (if_detach_event_tag == NULL) {
2850 		printf("ip_mroute: unable to register "
2851 		    "ifnet_departure_event handler\n");
2852 		MROUTER_LOCK_DESTROY();
2853 		return (EINVAL);
2854 	}
2855 
2856 	MFC_LOCK_INIT();
2857 	VIF_LOCK_INIT();
2858 
2859 	mfchashsize = MFCHASHSIZE;
2860 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2861 	    !powerof2(mfchashsize)) {
2862 		printf("WARNING: %s not a power of 2; using default\n",
2863 		    "net.inet.ip.mfchashsize");
2864 		mfchashsize = MFCHASHSIZE;
2865 	}
2866 
2867 	pim_squelch_wholepkt = 0;
2868 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2869 	    &pim_squelch_wholepkt);
2870 
2871 	pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2872 	if (pim_encap_cookie == NULL) {
2873 		printf("ip_mroute: unable to attach pim encap\n");
2874 		VIF_LOCK_DESTROY();
2875 		MFC_LOCK_DESTROY();
2876 		MROUTER_LOCK_DESTROY();
2877 		return (EINVAL);
2878 	}
2879 
2880 	ip_mcast_src = X_ip_mcast_src;
2881 	ip_mforward = X_ip_mforward;
2882 	ip_mrouter_done = X_ip_mrouter_done;
2883 	ip_mrouter_get = X_ip_mrouter_get;
2884 	ip_mrouter_set = X_ip_mrouter_set;
2885 
2886 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2887 	ip_rsvp_vif = X_ip_rsvp_vif;
2888 
2889 	legal_vif_num = X_legal_vif_num;
2890 	mrt_ioctl = X_mrt_ioctl;
2891 	rsvp_input_p = X_rsvp_input;
2892 	break;
2893 
2894     case MOD_UNLOAD:
2895 	/*
2896 	 * Typically module unload happens after the user-level
2897 	 * process has shutdown the kernel services (the check
2898 	 * below insures someone can't just yank the module out
2899 	 * from under a running process).  But if the module is
2900 	 * just loaded and then unloaded w/o starting up a user
2901 	 * process we still need to cleanup.
2902 	 */
2903 	MROUTER_LOCK();
2904 	if (ip_mrouter_cnt != 0) {
2905 	    MROUTER_UNLOCK();
2906 	    return (EINVAL);
2907 	}
2908 	ip_mrouter_unloading = 1;
2909 	MROUTER_UNLOCK();
2910 
2911 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2912 
2913 	if (pim_encap_cookie) {
2914 	    ip_encap_detach(pim_encap_cookie);
2915 	    pim_encap_cookie = NULL;
2916 	}
2917 
2918 	ip_mcast_src = NULL;
2919 	ip_mforward = NULL;
2920 	ip_mrouter_done = NULL;
2921 	ip_mrouter_get = NULL;
2922 	ip_mrouter_set = NULL;
2923 
2924 	ip_rsvp_force_done = NULL;
2925 	ip_rsvp_vif = NULL;
2926 
2927 	legal_vif_num = NULL;
2928 	mrt_ioctl = NULL;
2929 	rsvp_input_p = NULL;
2930 
2931 	VIF_LOCK_DESTROY();
2932 	MFC_LOCK_DESTROY();
2933 	MROUTER_LOCK_DESTROY();
2934 	break;
2935 
2936     default:
2937 	return EOPNOTSUPP;
2938     }
2939     return 0;
2940 }
2941 
2942 static moduledata_t ip_mroutemod = {
2943     "ip_mroute",
2944     ip_mroute_modevent,
2945     0
2946 };
2947 
2948 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2949