xref: /freebsd/sys/netinet/ip_mroute.c (revision 4f29da19bd44f0e99f021510460a81bf754c21d2)
1 /*-
2  * Copyright (c) 1989 Stephen Deering
3  * Copyright (c) 1992, 1993
4  *      The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Stephen Deering of Stanford University.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
34  */
35 
36 /*
37  * IP multicast forwarding procedures
38  *
39  * Written by David Waitzman, BBN Labs, August 1988.
40  * Modified by Steve Deering, Stanford, February 1989.
41  * Modified by Mark J. Steiglitz, Stanford, May, 1991
42  * Modified by Van Jacobson, LBL, January 1993
43  * Modified by Ajit Thyagarajan, PARC, August 1993
44  * Modified by Bill Fenner, PARC, April 1995
45  * Modified by Ahmed Helmy, SGI, June 1996
46  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
47  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
48  * Modified by Hitoshi Asaeda, WIDE, August 2000
49  * Modified by Pavlin Radoslavov, ICSI, October 2002
50  *
51  * MROUTING Revision: 3.5
52  * and PIM-SMv2 and PIM-DM support, advanced API support,
53  * bandwidth metering and signaling
54  *
55  * $FreeBSD$
56  */
57 
58 #include "opt_mac.h"
59 #include "opt_mrouting.h"
60 
61 #ifdef PIM
62 #define _PIM_VT 1
63 #endif
64 
65 #include <sys/param.h>
66 #include <sys/kernel.h>
67 #include <sys/lock.h>
68 #include <sys/mac.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/module.h>
72 #include <sys/protosw.h>
73 #include <sys/signalvar.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/sockio.h>
77 #include <sys/sx.h>
78 #include <sys/sysctl.h>
79 #include <sys/syslog.h>
80 #include <sys/systm.h>
81 #include <sys/time.h>
82 #include <net/if.h>
83 #include <net/netisr.h>
84 #include <net/route.h>
85 #include <netinet/in.h>
86 #include <netinet/igmp.h>
87 #include <netinet/in_systm.h>
88 #include <netinet/in_var.h>
89 #include <netinet/ip.h>
90 #include <netinet/ip_encap.h>
91 #include <netinet/ip_mroute.h>
92 #include <netinet/ip_var.h>
93 #include <netinet/ip_options.h>
94 #ifdef PIM
95 #include <netinet/pim.h>
96 #include <netinet/pim_var.h>
97 #endif
98 #include <netinet/udp.h>
99 #include <machine/in_cksum.h>
100 
101 /*
102  * Control debugging code for rsvp and multicast routing code.
103  * Can only set them with the debugger.
104  */
105 static u_int    rsvpdebug;		/* non-zero enables debugging	*/
106 
107 static u_int	mrtdebug;		/* any set of the flags below	*/
108 #define		DEBUG_MFC	0x02
109 #define		DEBUG_FORWARD	0x04
110 #define		DEBUG_EXPIRE	0x08
111 #define		DEBUG_XMIT	0x10
112 #define		DEBUG_PIM	0x20
113 
114 #define		VIFI_INVALID	((vifi_t) -1)
115 
116 #define M_HASCL(m)	((m)->m_flags & M_EXT)
117 
118 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables");
119 
120 /*
121  * Locking.  We use two locks: one for the virtual interface table and
122  * one for the forwarding table.  These locks may be nested in which case
123  * the VIF lock must always be taken first.  Note that each lock is used
124  * to cover not only the specific data structure but also related data
125  * structures.  It may be better to add more fine-grained locking later;
126  * it's not clear how performance-critical this code is.
127  */
128 
129 static struct mrtstat	mrtstat;
130 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
131     &mrtstat, mrtstat,
132     "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)");
133 
134 static struct mfc	*mfctable[MFCTBLSIZ];
135 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
136     &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]",
137     "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)");
138 
139 static struct mtx mfc_mtx;
140 #define	MFC_LOCK()	mtx_lock(&mfc_mtx)
141 #define	MFC_UNLOCK()	mtx_unlock(&mfc_mtx)
142 #define	MFC_LOCK_ASSERT()	do {					\
143 	mtx_assert(&mfc_mtx, MA_OWNED);					\
144 	NET_ASSERT_GIANT();						\
145 } while (0)
146 #define	MFC_LOCK_INIT()	mtx_init(&mfc_mtx, "mroute mfc table", NULL, MTX_DEF)
147 #define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
148 
149 static struct vif	viftable[MAXVIFS];
150 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
151     &viftable, sizeof(viftable), "S,vif[MAXVIFS]",
152     "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
153 
154 static struct mtx vif_mtx;
155 #define	VIF_LOCK()	mtx_lock(&vif_mtx)
156 #define	VIF_UNLOCK()	mtx_unlock(&vif_mtx)
157 #define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
158 #define	VIF_LOCK_INIT()	mtx_init(&vif_mtx, "mroute vif table", NULL, MTX_DEF)
159 #define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
160 
161 static u_char		nexpire[MFCTBLSIZ];
162 
163 static struct callout expire_upcalls_ch;
164 
165 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
166 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
167 
168 /*
169  * Define the token bucket filter structures
170  * tbftable -> each vif has one of these for storing info
171  */
172 
173 static struct tbf tbftable[MAXVIFS];
174 #define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
175 
176 /*
177  * 'Interfaces' associated with decapsulator (so we can tell
178  * packets that went through it from ones that get reflected
179  * by a broken gateway).  These interfaces are never linked into
180  * the system ifnet list & no routes point to them.  I.e., packets
181  * can't be sent this way.  They only exist as a placeholder for
182  * multicast source verification.
183  */
184 static struct ifnet multicast_decap_if[MAXVIFS];
185 
186 #define ENCAP_TTL 64
187 #define ENCAP_PROTO IPPROTO_IPIP	/* 4 */
188 
189 /* prototype IP hdr for encapsulated packets */
190 static struct ip multicast_encap_iphdr = {
191 #if BYTE_ORDER == LITTLE_ENDIAN
192 	sizeof(struct ip) >> 2, IPVERSION,
193 #else
194 	IPVERSION, sizeof(struct ip) >> 2,
195 #endif
196 	0,				/* tos */
197 	sizeof(struct ip),		/* total length */
198 	0,				/* id */
199 	0,				/* frag offset */
200 	ENCAP_TTL, ENCAP_PROTO,
201 	0,				/* checksum */
202 };
203 
204 /*
205  * Bandwidth meter variables and constants
206  */
207 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
208 /*
209  * Pending timeouts are stored in a hash table, the key being the
210  * expiration time. Periodically, the entries are analysed and processed.
211  */
212 #define BW_METER_BUCKETS	1024
213 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
214 static struct callout bw_meter_ch;
215 #define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
216 
217 /*
218  * Pending upcalls are stored in a vector which is flushed when
219  * full, or periodically
220  */
221 static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
222 static u_int	bw_upcalls_n; /* # of pending upcalls */
223 static struct callout bw_upcalls_ch;
224 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
225 
226 #ifdef PIM
227 static struct pimstat pimstat;
228 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
229     &pimstat, pimstat,
230     "PIM Statistics (struct pimstat, netinet/pim_var.h)");
231 
232 /*
233  * Note: the PIM Register encapsulation adds the following in front of a
234  * data packet:
235  *
236  * struct pim_encap_hdr {
237  *    struct ip ip;
238  *    struct pim_encap_pimhdr  pim;
239  * }
240  *
241  */
242 
243 struct pim_encap_pimhdr {
244 	struct pim pim;
245 	uint32_t   flags;
246 };
247 
248 static struct ip pim_encap_iphdr = {
249 #if BYTE_ORDER == LITTLE_ENDIAN
250 	sizeof(struct ip) >> 2,
251 	IPVERSION,
252 #else
253 	IPVERSION,
254 	sizeof(struct ip) >> 2,
255 #endif
256 	0,			/* tos */
257 	sizeof(struct ip),	/* total length */
258 	0,			/* id */
259 	0,			/* frag offset */
260 	ENCAP_TTL,
261 	IPPROTO_PIM,
262 	0,			/* checksum */
263 };
264 
265 static struct pim_encap_pimhdr pim_encap_pimhdr = {
266     {
267 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
268 	0,			/* reserved */
269 	0,			/* checksum */
270     },
271     0				/* flags */
272 };
273 
274 static struct ifnet multicast_register_if;
275 static vifi_t reg_vif_num = VIFI_INVALID;
276 #endif /* PIM */
277 
278 /*
279  * Private variables.
280  */
281 static vifi_t	   numvifs;
282 static const struct encaptab *encap_cookie;
283 
284 /*
285  * one-back cache used by mroute_encapcheck to locate a tunnel's vif
286  * given a datagram's src ip address.
287  */
288 static u_long last_encap_src;
289 static struct vif *last_encap_vif;
290 
291 /*
292  * Callout for queue processing.
293  */
294 static struct callout tbf_reprocess_ch;
295 
296 static u_long	X_ip_mcast_src(int vifi);
297 static int	X_ip_mforward(struct ip *ip, struct ifnet *ifp,
298 			struct mbuf *m, struct ip_moptions *imo);
299 static int	X_ip_mrouter_done(void);
300 static int	X_ip_mrouter_get(struct socket *so, struct sockopt *m);
301 static int	X_ip_mrouter_set(struct socket *so, struct sockopt *m);
302 static int	X_legal_vif_num(int vif);
303 static int	X_mrt_ioctl(int cmd, caddr_t data);
304 
305 static int get_sg_cnt(struct sioc_sg_req *);
306 static int get_vif_cnt(struct sioc_vif_req *);
307 static int ip_mrouter_init(struct socket *, int);
308 static int add_vif(struct vifctl *);
309 static int del_vif(vifi_t);
310 static int add_mfc(struct mfcctl2 *);
311 static int del_mfc(struct mfcctl2 *);
312 static int set_api_config(uint32_t *); /* chose API capabilities */
313 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
314 static int set_assert(int);
315 static void expire_upcalls(void *);
316 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
317 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
318 static void encap_send(struct ip *, struct vif *, struct mbuf *);
319 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
320 static void tbf_queue(struct vif *, struct mbuf *);
321 static void tbf_process_q(struct vif *);
322 static void tbf_reprocess_q(void *);
323 static int tbf_dq_sel(struct vif *, struct ip *);
324 static void tbf_send_packet(struct vif *, struct mbuf *);
325 static void tbf_update_tokens(struct vif *);
326 static int priority(struct vif *, struct ip *);
327 
328 /*
329  * Bandwidth monitoring
330  */
331 static void free_bw_list(struct bw_meter *list);
332 static int add_bw_upcall(struct bw_upcall *);
333 static int del_bw_upcall(struct bw_upcall *);
334 static void bw_meter_receive_packet(struct bw_meter *x, int plen,
335 		struct timeval *nowp);
336 static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp);
337 static void bw_upcalls_send(void);
338 static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp);
339 static void unschedule_bw_meter(struct bw_meter *x);
340 static void bw_meter_process(void);
341 static void expire_bw_upcalls_send(void *);
342 static void expire_bw_meter_process(void *);
343 
344 #ifdef PIM
345 static int pim_register_send(struct ip *, struct vif *,
346 		struct mbuf *, struct mfc *);
347 static int pim_register_send_rp(struct ip *, struct vif *,
348 		struct mbuf *, struct mfc *);
349 static int pim_register_send_upcall(struct ip *, struct vif *,
350 		struct mbuf *, struct mfc *);
351 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
352 #endif
353 
354 /*
355  * whether or not special PIM assert processing is enabled.
356  */
357 static int pim_assert;
358 /*
359  * Rate limit for assert notification messages, in usec
360  */
361 #define ASSERT_MSG_TIME		3000000
362 
363 /*
364  * Kernel multicast routing 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 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
369 					 MRT_MFC_FLAGS_BORDER_VIF |
370 					 MRT_MFC_RP |
371 					 MRT_MFC_BW_UPCALL);
372 static uint32_t mrt_api_config = 0;
373 
374 /*
375  * Hash function for a source, group entry
376  */
377 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
378 			((g) >> 20) ^ ((g) >> 10) ^ (g))
379 
380 /*
381  * Find a route for a given origin IP address and Multicast group address
382  * Type of service parameter to be added in the future!!!
383  * Statistics are updated by the caller if needed
384  * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
385  */
386 static struct mfc *
387 mfc_find(in_addr_t o, in_addr_t g)
388 {
389     struct mfc *rt;
390 
391     MFC_LOCK_ASSERT();
392 
393     for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next)
394 	if ((rt->mfc_origin.s_addr == o) &&
395 		(rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL))
396 	    break;
397     return rt;
398 }
399 
400 /*
401  * Macros to compute elapsed time efficiently
402  * Borrowed from Van Jacobson's scheduling code
403  */
404 #define TV_DELTA(a, b, delta) {					\
405 	int xxs;						\
406 	delta = (a).tv_usec - (b).tv_usec;			\
407 	if ((xxs = (a).tv_sec - (b).tv_sec)) {			\
408 		switch (xxs) {					\
409 		case 2:						\
410 		      delta += 1000000;				\
411 		      /* FALLTHROUGH */				\
412 		case 1:						\
413 		      delta += 1000000;				\
414 		      break;					\
415 		default:					\
416 		      delta += (1000000 * xxs);			\
417 		}						\
418 	}							\
419 }
420 
421 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
422 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
423 
424 /*
425  * Handle MRT setsockopt commands to modify the multicast routing tables.
426  */
427 static int
428 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
429 {
430     int	error, optval;
431     vifi_t	vifi;
432     struct	vifctl vifc;
433     struct	mfcctl2 mfc;
434     struct	bw_upcall bw_upcall;
435     uint32_t	i;
436 
437     if (so != ip_mrouter && sopt->sopt_name != MRT_INIT)
438 	return EPERM;
439 
440     error = 0;
441     switch (sopt->sopt_name) {
442     case MRT_INIT:
443 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
444 	if (error)
445 	    break;
446 	error = ip_mrouter_init(so, optval);
447 	break;
448 
449     case MRT_DONE:
450 	error = ip_mrouter_done();
451 	break;
452 
453     case MRT_ADD_VIF:
454 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
455 	if (error)
456 	    break;
457 	error = add_vif(&vifc);
458 	break;
459 
460     case MRT_DEL_VIF:
461 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
462 	if (error)
463 	    break;
464 	error = del_vif(vifi);
465 	break;
466 
467     case MRT_ADD_MFC:
468     case MRT_DEL_MFC:
469 	/*
470 	 * select data size depending on API version.
471 	 */
472 	if (sopt->sopt_name == MRT_ADD_MFC &&
473 		mrt_api_config & MRT_API_FLAGS_ALL) {
474 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
475 				sizeof(struct mfcctl2));
476 	} else {
477 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
478 				sizeof(struct mfcctl));
479 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
480 			sizeof(mfc) - sizeof(struct mfcctl));
481 	}
482 	if (error)
483 	    break;
484 	if (sopt->sopt_name == MRT_ADD_MFC)
485 	    error = add_mfc(&mfc);
486 	else
487 	    error = del_mfc(&mfc);
488 	break;
489 
490     case MRT_ASSERT:
491 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
492 	if (error)
493 	    break;
494 	set_assert(optval);
495 	break;
496 
497     case MRT_API_CONFIG:
498 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
499 	if (!error)
500 	    error = set_api_config(&i);
501 	if (!error)
502 	    error = sooptcopyout(sopt, &i, sizeof i);
503 	break;
504 
505     case MRT_ADD_BW_UPCALL:
506     case MRT_DEL_BW_UPCALL:
507 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
508 				sizeof bw_upcall);
509 	if (error)
510 	    break;
511 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
512 	    error = add_bw_upcall(&bw_upcall);
513 	else
514 	    error = del_bw_upcall(&bw_upcall);
515 	break;
516 
517     default:
518 	error = EOPNOTSUPP;
519 	break;
520     }
521     return error;
522 }
523 
524 /*
525  * Handle MRT getsockopt commands
526  */
527 static int
528 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
529 {
530     int error;
531     static int version = 0x0305; /* !!! why is this here? XXX */
532 
533     switch (sopt->sopt_name) {
534     case MRT_VERSION:
535 	error = sooptcopyout(sopt, &version, sizeof version);
536 	break;
537 
538     case MRT_ASSERT:
539 	error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert);
540 	break;
541 
542     case MRT_API_SUPPORT:
543 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
544 	break;
545 
546     case MRT_API_CONFIG:
547 	error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config);
548 	break;
549 
550     default:
551 	error = EOPNOTSUPP;
552 	break;
553     }
554     return error;
555 }
556 
557 /*
558  * Handle ioctl commands to obtain information from the cache
559  */
560 static int
561 X_mrt_ioctl(int cmd, caddr_t data)
562 {
563     int error = 0;
564 
565     /*
566      * Currently the only function calling this ioctl routine is rtioctl().
567      * Typically, only root can create the raw socket in order to execute
568      * this ioctl method, however the request might be coming from a prison
569      */
570     error = suser(curthread);
571     if (error)
572 	return (error);
573     switch (cmd) {
574     case (SIOCGETVIFCNT):
575 	error = get_vif_cnt((struct sioc_vif_req *)data);
576 	break;
577 
578     case (SIOCGETSGCNT):
579 	error = get_sg_cnt((struct sioc_sg_req *)data);
580 	break;
581 
582     default:
583 	error = EINVAL;
584 	break;
585     }
586     return error;
587 }
588 
589 /*
590  * returns the packet, byte, rpf-failure count for the source group provided
591  */
592 static int
593 get_sg_cnt(struct sioc_sg_req *req)
594 {
595     struct mfc *rt;
596 
597     MFC_LOCK();
598     rt = mfc_find(req->src.s_addr, req->grp.s_addr);
599     if (rt == NULL) {
600 	MFC_UNLOCK();
601 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
602 	return EADDRNOTAVAIL;
603     }
604     req->pktcnt = rt->mfc_pkt_cnt;
605     req->bytecnt = rt->mfc_byte_cnt;
606     req->wrong_if = rt->mfc_wrong_if;
607     MFC_UNLOCK();
608     return 0;
609 }
610 
611 /*
612  * returns the input and output packet and byte counts on the vif provided
613  */
614 static int
615 get_vif_cnt(struct sioc_vif_req *req)
616 {
617     vifi_t vifi = req->vifi;
618 
619     VIF_LOCK();
620     if (vifi >= numvifs) {
621 	VIF_UNLOCK();
622 	return EINVAL;
623     }
624 
625     req->icount = viftable[vifi].v_pkt_in;
626     req->ocount = viftable[vifi].v_pkt_out;
627     req->ibytes = viftable[vifi].v_bytes_in;
628     req->obytes = viftable[vifi].v_bytes_out;
629     VIF_UNLOCK();
630 
631     return 0;
632 }
633 
634 static void
635 ip_mrouter_reset(void)
636 {
637     bzero((caddr_t)mfctable, sizeof(mfctable));
638     bzero((caddr_t)nexpire, sizeof(nexpire));
639 
640     pim_assert = 0;
641     mrt_api_config = 0;
642 
643     callout_init(&expire_upcalls_ch, NET_CALLOUT_MPSAFE);
644 
645     bw_upcalls_n = 0;
646     bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers));
647     callout_init(&bw_upcalls_ch, NET_CALLOUT_MPSAFE);
648     callout_init(&bw_meter_ch, NET_CALLOUT_MPSAFE);
649 
650     callout_init(&tbf_reprocess_ch, NET_CALLOUT_MPSAFE);
651 }
652 
653 static struct mtx mrouter_mtx;		/* used to synch init/done work */
654 
655 /*
656  * Enable multicast routing
657  */
658 static int
659 ip_mrouter_init(struct socket *so, int version)
660 {
661     if (mrtdebug)
662 	log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
663 	    so->so_type, so->so_proto->pr_protocol);
664 
665     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
666 	return EOPNOTSUPP;
667 
668     if (version != 1)
669 	return ENOPROTOOPT;
670 
671     mtx_lock(&mrouter_mtx);
672 
673     if (ip_mrouter != NULL) {
674 	mtx_unlock(&mrouter_mtx);
675 	return EADDRINUSE;
676     }
677 
678     callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
679 
680     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
681 	expire_bw_upcalls_send, NULL);
682     callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
683 
684     ip_mrouter = so;
685 
686     mtx_unlock(&mrouter_mtx);
687 
688     if (mrtdebug)
689 	log(LOG_DEBUG, "ip_mrouter_init\n");
690 
691     return 0;
692 }
693 
694 /*
695  * Disable multicast routing
696  */
697 static int
698 X_ip_mrouter_done(void)
699 {
700     vifi_t vifi;
701     int i;
702     struct ifnet *ifp;
703     struct ifreq ifr;
704     struct mfc *rt;
705     struct rtdetq *rte;
706 
707     mtx_lock(&mrouter_mtx);
708 
709     if (ip_mrouter == NULL) {
710 	mtx_unlock(&mrouter_mtx);
711 	return EINVAL;
712     }
713 
714     /*
715      * Detach/disable hooks to the reset of the system.
716      */
717     ip_mrouter = NULL;
718     mrt_api_config = 0;
719 
720     VIF_LOCK();
721     if (encap_cookie) {
722 	const struct encaptab *c = encap_cookie;
723 	encap_cookie = NULL;
724 	encap_detach(c);
725     }
726     VIF_UNLOCK();
727 
728     callout_stop(&tbf_reprocess_ch);
729 
730     VIF_LOCK();
731     /*
732      * For each phyint in use, disable promiscuous reception of all IP
733      * multicasts.
734      */
735     for (vifi = 0; vifi < numvifs; vifi++) {
736 	if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
737 		!(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
738 	    struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
739 
740 	    so->sin_len = sizeof(struct sockaddr_in);
741 	    so->sin_family = AF_INET;
742 	    so->sin_addr.s_addr = INADDR_ANY;
743 	    ifp = viftable[vifi].v_ifp;
744 	    if_allmulti(ifp, 0);
745 	}
746     }
747     bzero((caddr_t)tbftable, sizeof(tbftable));
748     bzero((caddr_t)viftable, sizeof(viftable));
749     numvifs = 0;
750     pim_assert = 0;
751     VIF_UNLOCK();
752 
753     /*
754      * Free all multicast forwarding cache entries.
755      */
756     callout_stop(&expire_upcalls_ch);
757     callout_stop(&bw_upcalls_ch);
758     callout_stop(&bw_meter_ch);
759 
760     MFC_LOCK();
761     for (i = 0; i < MFCTBLSIZ; i++) {
762 	for (rt = mfctable[i]; rt != NULL; ) {
763 	    struct mfc *nr = rt->mfc_next;
764 
765 	    for (rte = rt->mfc_stall; rte != NULL; ) {
766 		struct rtdetq *n = rte->next;
767 
768 		m_freem(rte->m);
769 		free(rte, M_MRTABLE);
770 		rte = n;
771 	    }
772 	    free_bw_list(rt->mfc_bw_meter);
773 	    free(rt, M_MRTABLE);
774 	    rt = nr;
775 	}
776     }
777     bzero((caddr_t)mfctable, sizeof(mfctable));
778     bzero((caddr_t)nexpire, sizeof(nexpire));
779     bw_upcalls_n = 0;
780     bzero(bw_meter_timers, sizeof(bw_meter_timers));
781     MFC_UNLOCK();
782 
783     /*
784      * Reset de-encapsulation cache
785      */
786     last_encap_src = INADDR_ANY;
787     last_encap_vif = NULL;
788 #ifdef PIM
789     reg_vif_num = VIFI_INVALID;
790 #endif
791 
792     mtx_unlock(&mrouter_mtx);
793 
794     if (mrtdebug)
795 	log(LOG_DEBUG, "ip_mrouter_done\n");
796 
797     return 0;
798 }
799 
800 /*
801  * Set PIM assert processing global
802  */
803 static int
804 set_assert(int i)
805 {
806     if ((i != 1) && (i != 0))
807 	return EINVAL;
808 
809     pim_assert = i;
810 
811     return 0;
812 }
813 
814 /*
815  * Configure API capabilities
816  */
817 int
818 set_api_config(uint32_t *apival)
819 {
820     int i;
821 
822     /*
823      * We can set the API capabilities only if it is the first operation
824      * after MRT_INIT. I.e.:
825      *  - there are no vifs installed
826      *  - pim_assert is not enabled
827      *  - the MFC table is empty
828      */
829     if (numvifs > 0) {
830 	*apival = 0;
831 	return EPERM;
832     }
833     if (pim_assert) {
834 	*apival = 0;
835 	return EPERM;
836     }
837     for (i = 0; i < MFCTBLSIZ; i++) {
838 	if (mfctable[i] != NULL) {
839 	    *apival = 0;
840 	    return EPERM;
841 	}
842     }
843 
844     mrt_api_config = *apival & mrt_api_support;
845     *apival = mrt_api_config;
846 
847     return 0;
848 }
849 
850 /*
851  * Decide if a packet is from a tunnelled peer.
852  * Return 0 if not, 64 if so.  XXX yuck.. 64 ???
853  */
854 static int
855 mroute_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
856 {
857     struct ip *ip = mtod(m, struct ip *);
858     int hlen = ip->ip_hl << 2;
859 
860     /*
861      * don't claim the packet if it's not to a multicast destination or if
862      * we don't have an encapsulating tunnel with the source.
863      * Note:  This code assumes that the remote site IP address
864      * uniquely identifies the tunnel (i.e., that this site has
865      * at most one tunnel with the remote site).
866      */
867     if (!IN_MULTICAST(ntohl(((struct ip *)((char *)ip+hlen))->ip_dst.s_addr)))
868 	return 0;
869     if (ip->ip_src.s_addr != last_encap_src) {
870 	struct vif *vifp = viftable;
871 	struct vif *vife = vifp + numvifs;
872 
873 	last_encap_src = ip->ip_src.s_addr;
874 	last_encap_vif = NULL;
875 	for ( ; vifp < vife; ++vifp)
876 	    if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
877 		if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) == VIFF_TUNNEL)
878 		    last_encap_vif = vifp;
879 		break;
880 	    }
881     }
882     if (last_encap_vif == NULL) {
883 	last_encap_src = INADDR_ANY;
884 	return 0;
885     }
886     return 64;
887 }
888 
889 /*
890  * De-encapsulate a packet and feed it back through ip input (this
891  * routine is called whenever IP gets a packet that mroute_encap_func()
892  * claimed).
893  */
894 static void
895 mroute_encap_input(struct mbuf *m, int off)
896 {
897     struct ip *ip = mtod(m, struct ip *);
898     int hlen = ip->ip_hl << 2;
899 
900     if (hlen > sizeof(struct ip))
901 	ip_stripoptions(m, (struct mbuf *) 0);
902     m->m_data += sizeof(struct ip);
903     m->m_len -= sizeof(struct ip);
904     m->m_pkthdr.len -= sizeof(struct ip);
905 
906     m->m_pkthdr.rcvif = last_encap_vif->v_ifp;
907 
908     netisr_queue(NETISR_IP, m);		/* mbuf is free'd on failure. */
909     /*
910      * normally we would need a "schednetisr(NETISR_IP)"
911      * here but we were called by ip_input and it is going
912      * to loop back & try to dequeue the packet we just
913      * queued as soon as we return so we avoid the
914      * unnecessary software interrrupt.
915      *
916      * XXX
917      * This no longer holds - we may have direct-dispatched the packet,
918      * or there may be a queue processing limit.
919      */
920 }
921 
922 extern struct domain inetdomain;
923 static struct protosw mroute_encap_protosw =
924 {
925 	.pr_type =		SOCK_RAW,
926 	.pr_domain =		&inetdomain,
927 	.pr_protocol =		IPPROTO_IPV4,
928 	.pr_flags =		PR_ATOMIC|PR_ADDR,
929 	.pr_input =		mroute_encap_input,
930 	.pr_ctloutput =		rip_ctloutput,
931 	.pr_usrreqs =		&rip_usrreqs
932 };
933 
934 /*
935  * Add a vif to the vif table
936  */
937 static int
938 add_vif(struct vifctl *vifcp)
939 {
940     struct vif *vifp = viftable + vifcp->vifc_vifi;
941     struct sockaddr_in sin = {sizeof sin, AF_INET};
942     struct ifaddr *ifa;
943     struct ifnet *ifp;
944     int error;
945     struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
946 
947     VIF_LOCK();
948     if (vifcp->vifc_vifi >= MAXVIFS) {
949 	VIF_UNLOCK();
950 	return EINVAL;
951     }
952     if (vifp->v_lcl_addr.s_addr != INADDR_ANY) {
953 	VIF_UNLOCK();
954 	return EADDRINUSE;
955     }
956     if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) {
957 	VIF_UNLOCK();
958 	return EADDRNOTAVAIL;
959     }
960 
961     /* Find the interface with an address in AF_INET family */
962 #ifdef PIM
963     if (vifcp->vifc_flags & VIFF_REGISTER) {
964 	/*
965 	 * XXX: Because VIFF_REGISTER does not really need a valid
966 	 * local interface (e.g. it could be 127.0.0.2), we don't
967 	 * check its address.
968 	 */
969 	ifp = NULL;
970     } else
971 #endif
972     {
973 	sin.sin_addr = vifcp->vifc_lcl_addr;
974 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
975 	if (ifa == NULL) {
976 	    VIF_UNLOCK();
977 	    return EADDRNOTAVAIL;
978 	}
979 	ifp = ifa->ifa_ifp;
980     }
981 
982     if (vifcp->vifc_flags & VIFF_TUNNEL) {
983 	if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
984 	    /*
985 	     * An encapsulating tunnel is wanted.  Tell
986 	     * mroute_encap_input() to start paying attention
987 	     * to encapsulated packets.
988 	     */
989 	    if (encap_cookie == NULL) {
990 		int i;
991 
992 		encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
993 				mroute_encapcheck,
994 				(struct protosw *)&mroute_encap_protosw, NULL);
995 
996 		if (encap_cookie == NULL) {
997 		    printf("ip_mroute: unable to attach encap\n");
998 		    VIF_UNLOCK();
999 		    return EIO;	/* XXX */
1000 		}
1001 		for (i = 0; i < MAXVIFS; ++i) {
1002 		    if_initname(&multicast_decap_if[i], "mdecap", i);
1003 		}
1004 	    }
1005 	    /*
1006 	     * Set interface to fake encapsulator interface
1007 	     */
1008 	    ifp = &multicast_decap_if[vifcp->vifc_vifi];
1009 	    /*
1010 	     * Prepare cached route entry
1011 	     */
1012 	    bzero(&vifp->v_route, sizeof(vifp->v_route));
1013 	} else {
1014 	    log(LOG_ERR, "source routed tunnels not supported\n");
1015 	    VIF_UNLOCK();
1016 	    return EOPNOTSUPP;
1017 	}
1018 #ifdef PIM
1019     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
1020 	ifp = &multicast_register_if;
1021 	if (mrtdebug)
1022 	    log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
1023 		    (void *)&multicast_register_if);
1024 	if (reg_vif_num == VIFI_INVALID) {
1025 	    if_initname(&multicast_register_if, "register_vif", 0);
1026 	    multicast_register_if.if_flags = IFF_LOOPBACK;
1027 	    bzero(&vifp->v_route, sizeof(vifp->v_route));
1028 	    reg_vif_num = vifcp->vifc_vifi;
1029 	}
1030 #endif
1031     } else {		/* Make sure the interface supports multicast */
1032 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
1033 	    VIF_UNLOCK();
1034 	    return EOPNOTSUPP;
1035 	}
1036 
1037 	/* Enable promiscuous reception of all IP multicasts from the if */
1038 	error = if_allmulti(ifp, 1);
1039 	if (error) {
1040 	    VIF_UNLOCK();
1041 	    return error;
1042 	}
1043     }
1044 
1045     /* define parameters for the tbf structure */
1046     vifp->v_tbf = v_tbf;
1047     GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
1048     vifp->v_tbf->tbf_n_tok = 0;
1049     vifp->v_tbf->tbf_q_len = 0;
1050     vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
1051     vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
1052 
1053     vifp->v_flags     = vifcp->vifc_flags;
1054     vifp->v_threshold = vifcp->vifc_threshold;
1055     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
1056     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
1057     vifp->v_ifp       = ifp;
1058     /* scaling up here allows division by 1024 in critical code */
1059     vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
1060     vifp->v_rsvp_on   = 0;
1061     vifp->v_rsvpd     = NULL;
1062     /* initialize per vif pkt counters */
1063     vifp->v_pkt_in    = 0;
1064     vifp->v_pkt_out   = 0;
1065     vifp->v_bytes_in  = 0;
1066     vifp->v_bytes_out = 0;
1067 
1068     /* Adjust numvifs up if the vifi is higher than numvifs */
1069     if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
1070 
1071     VIF_UNLOCK();
1072 
1073     if (mrtdebug)
1074 	log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n",
1075 	    vifcp->vifc_vifi,
1076 	    (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr),
1077 	    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
1078 	    (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr),
1079 	    vifcp->vifc_threshold,
1080 	    vifcp->vifc_rate_limit);
1081 
1082     return 0;
1083 }
1084 
1085 /*
1086  * Delete a vif from the vif table
1087  */
1088 static int
1089 del_vif(vifi_t vifi)
1090 {
1091     struct vif *vifp;
1092 
1093     VIF_LOCK();
1094 
1095     if (vifi >= numvifs) {
1096 	VIF_UNLOCK();
1097 	return EINVAL;
1098     }
1099     vifp = &viftable[vifi];
1100     if (vifp->v_lcl_addr.s_addr == INADDR_ANY) {
1101 	VIF_UNLOCK();
1102 	return EADDRNOTAVAIL;
1103     }
1104 
1105     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1106 	if_allmulti(vifp->v_ifp, 0);
1107 
1108     if (vifp == last_encap_vif) {
1109 	last_encap_vif = NULL;
1110 	last_encap_src = INADDR_ANY;
1111     }
1112 
1113     /*
1114      * Free packets queued at the interface
1115      */
1116     while (vifp->v_tbf->tbf_q) {
1117 	struct mbuf *m = vifp->v_tbf->tbf_q;
1118 
1119 	vifp->v_tbf->tbf_q = m->m_act;
1120 	m_freem(m);
1121     }
1122 
1123 #ifdef PIM
1124     if (vifp->v_flags & VIFF_REGISTER)
1125 	reg_vif_num = VIFI_INVALID;
1126 #endif
1127 
1128     bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
1129     bzero((caddr_t)vifp, sizeof (*vifp));
1130 
1131     if (mrtdebug)
1132 	log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs);
1133 
1134     /* Adjust numvifs down */
1135     for (vifi = numvifs; vifi > 0; vifi--)
1136 	if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY)
1137 	    break;
1138     numvifs = vifi;
1139 
1140     VIF_UNLOCK();
1141 
1142     return 0;
1143 }
1144 
1145 /*
1146  * update an mfc entry without resetting counters and S,G addresses.
1147  */
1148 static void
1149 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1150 {
1151     int i;
1152 
1153     rt->mfc_parent = mfccp->mfcc_parent;
1154     for (i = 0; i < numvifs; i++) {
1155 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1156 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1157 	    MRT_MFC_FLAGS_ALL;
1158     }
1159     /* set the RP address */
1160     if (mrt_api_config & MRT_MFC_RP)
1161 	rt->mfc_rp = mfccp->mfcc_rp;
1162     else
1163 	rt->mfc_rp.s_addr = INADDR_ANY;
1164 }
1165 
1166 /*
1167  * fully initialize an mfc entry from the parameter.
1168  */
1169 static void
1170 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1171 {
1172     rt->mfc_origin     = mfccp->mfcc_origin;
1173     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1174 
1175     update_mfc_params(rt, mfccp);
1176 
1177     /* initialize pkt counters per src-grp */
1178     rt->mfc_pkt_cnt    = 0;
1179     rt->mfc_byte_cnt   = 0;
1180     rt->mfc_wrong_if   = 0;
1181     rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
1182 }
1183 
1184 
1185 /*
1186  * Add an mfc entry
1187  */
1188 static int
1189 add_mfc(struct mfcctl2 *mfccp)
1190 {
1191     struct mfc *rt;
1192     u_long hash;
1193     struct rtdetq *rte;
1194     u_short nstl;
1195 
1196     VIF_LOCK();
1197     MFC_LOCK();
1198 
1199     rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
1200 
1201     /* If an entry already exists, just update the fields */
1202     if (rt) {
1203 	if (mrtdebug & DEBUG_MFC)
1204 	    log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n",
1205 		(u_long)ntohl(mfccp->mfcc_origin.s_addr),
1206 		(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1207 		mfccp->mfcc_parent);
1208 
1209 	update_mfc_params(rt, mfccp);
1210 	MFC_UNLOCK();
1211 	VIF_UNLOCK();
1212 	return 0;
1213     }
1214 
1215     /*
1216      * Find the entry for which the upcall was made and update
1217      */
1218     hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
1219     for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) {
1220 
1221 	if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1222 		(rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
1223 		(rt->mfc_stall != NULL)) {
1224 
1225 	    if (nstl++)
1226 		log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n",
1227 		    "multiple kernel entries",
1228 		    (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1229 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1230 		    mfccp->mfcc_parent, (void *)rt->mfc_stall);
1231 
1232 	    if (mrtdebug & DEBUG_MFC)
1233 		log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n",
1234 		    (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1235 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1236 		    mfccp->mfcc_parent, (void *)rt->mfc_stall);
1237 
1238 	    init_mfc_params(rt, mfccp);
1239 
1240 	    rt->mfc_expire = 0;	/* Don't clean this guy up */
1241 	    nexpire[hash]--;
1242 
1243 	    /* free packets Qed at the end of this entry */
1244 	    for (rte = rt->mfc_stall; rte != NULL; ) {
1245 		struct rtdetq *n = rte->next;
1246 
1247 		ip_mdq(rte->m, rte->ifp, rt, -1);
1248 		m_freem(rte->m);
1249 		free(rte, M_MRTABLE);
1250 		rte = n;
1251 	    }
1252 	    rt->mfc_stall = NULL;
1253 	}
1254     }
1255 
1256     /*
1257      * It is possible that an entry is being inserted without an upcall
1258      */
1259     if (nstl == 0) {
1260 	if (mrtdebug & DEBUG_MFC)
1261 	    log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n",
1262 		hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1263 		(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1264 		mfccp->mfcc_parent);
1265 
1266 	for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) {
1267 	    if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1268 		    (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
1269 		init_mfc_params(rt, mfccp);
1270 		if (rt->mfc_expire)
1271 		    nexpire[hash]--;
1272 		rt->mfc_expire = 0;
1273 		break; /* XXX */
1274 	    }
1275 	}
1276 	if (rt == NULL) {		/* no upcall, so make a new entry */
1277 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1278 	    if (rt == NULL) {
1279 		MFC_UNLOCK();
1280 		VIF_UNLOCK();
1281 		return ENOBUFS;
1282 	    }
1283 
1284 	    init_mfc_params(rt, mfccp);
1285 	    rt->mfc_expire     = 0;
1286 	    rt->mfc_stall      = NULL;
1287 
1288 	    rt->mfc_bw_meter = NULL;
1289 	    /* insert new entry at head of hash chain */
1290 	    rt->mfc_next = mfctable[hash];
1291 	    mfctable[hash] = rt;
1292 	}
1293     }
1294     MFC_UNLOCK();
1295     VIF_UNLOCK();
1296     return 0;
1297 }
1298 
1299 /*
1300  * Delete an mfc entry
1301  */
1302 static int
1303 del_mfc(struct mfcctl2 *mfccp)
1304 {
1305     struct in_addr	origin;
1306     struct in_addr	mcastgrp;
1307     struct mfc		*rt;
1308     struct mfc		**nptr;
1309     u_long		hash;
1310     struct bw_meter	*list;
1311 
1312     origin = mfccp->mfcc_origin;
1313     mcastgrp = mfccp->mfcc_mcastgrp;
1314 
1315     if (mrtdebug & DEBUG_MFC)
1316 	log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n",
1317 	    (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1318 
1319     MFC_LOCK();
1320 
1321     hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1322     for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next)
1323 	if (origin.s_addr == rt->mfc_origin.s_addr &&
1324 		mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1325 		rt->mfc_stall == NULL)
1326 	    break;
1327     if (rt == NULL) {
1328 	MFC_UNLOCK();
1329 	return EADDRNOTAVAIL;
1330     }
1331 
1332     *nptr = rt->mfc_next;
1333 
1334     /*
1335      * free the bw_meter entries
1336      */
1337     list = rt->mfc_bw_meter;
1338     rt->mfc_bw_meter = NULL;
1339 
1340     free(rt, M_MRTABLE);
1341 
1342     free_bw_list(list);
1343 
1344     MFC_UNLOCK();
1345 
1346     return 0;
1347 }
1348 
1349 /*
1350  * Send a message to mrouted on the multicast routing socket
1351  */
1352 static int
1353 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1354 {
1355     if (s) {
1356 	SOCKBUF_LOCK(&s->so_rcv);
1357 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1358 	    NULL) != 0) {
1359 	    sorwakeup_locked(s);
1360 	    return 0;
1361 	}
1362 	SOCKBUF_UNLOCK(&s->so_rcv);
1363     }
1364     m_freem(mm);
1365     return -1;
1366 }
1367 
1368 /*
1369  * IP multicast forwarding function. This function assumes that the packet
1370  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1371  * pointed to by "ifp", and the packet is to be relayed to other networks
1372  * that have members of the packet's destination IP multicast group.
1373  *
1374  * The packet is returned unscathed to the caller, unless it is
1375  * erroneous, in which case a non-zero return value tells the caller to
1376  * discard it.
1377  */
1378 
1379 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1380 
1381 static int
1382 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1383     struct ip_moptions *imo)
1384 {
1385     struct mfc *rt;
1386     int error;
1387     vifi_t vifi;
1388 
1389     if (mrtdebug & DEBUG_FORWARD)
1390 	log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n",
1391 	    (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr),
1392 	    (void *)ifp);
1393 
1394     if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1395 		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1396 	/*
1397 	 * Packet arrived via a physical interface or
1398 	 * an encapsulated tunnel or a register_vif.
1399 	 */
1400     } else {
1401 	/*
1402 	 * Packet arrived through a source-route tunnel.
1403 	 * Source-route tunnels are no longer supported.
1404 	 */
1405 	static int last_log;
1406 	if (last_log != time_uptime) {
1407 	    last_log = time_uptime;
1408 	    log(LOG_ERR,
1409 		"ip_mforward: received source-routed packet from %lx\n",
1410 		(u_long)ntohl(ip->ip_src.s_addr));
1411 	}
1412 	return 1;
1413     }
1414 
1415     VIF_LOCK();
1416     MFC_LOCK();
1417     if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1418 	if (ip->ip_ttl < 255)
1419 	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1420 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1421 	    struct vif *vifp = viftable + vifi;
1422 
1423 	    printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n",
1424 		(long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr),
1425 		vifi,
1426 		(vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1427 		vifp->v_ifp->if_xname);
1428 	}
1429 	error = ip_mdq(m, ifp, NULL, vifi);
1430 	MFC_UNLOCK();
1431 	VIF_UNLOCK();
1432 	return error;
1433     }
1434     if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1435 	printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1436 	    (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr));
1437 	if (!imo)
1438 	    printf("In fact, no options were specified at all\n");
1439     }
1440 
1441     /*
1442      * Don't forward a packet with time-to-live of zero or one,
1443      * or a packet destined to a local-only group.
1444      */
1445     if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) {
1446 	MFC_UNLOCK();
1447 	VIF_UNLOCK();
1448 	return 0;
1449     }
1450 
1451     /*
1452      * Determine forwarding vifs from the forwarding cache table
1453      */
1454     ++mrtstat.mrts_mfc_lookups;
1455     rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1456 
1457     /* Entry exists, so forward if necessary */
1458     if (rt != NULL) {
1459 	error = ip_mdq(m, ifp, rt, -1);
1460 	MFC_UNLOCK();
1461 	VIF_UNLOCK();
1462 	return error;
1463     } else {
1464 	/*
1465 	 * If we don't have a route for packet's origin,
1466 	 * Make a copy of the packet & send message to routing daemon
1467 	 */
1468 
1469 	struct mbuf *mb0;
1470 	struct rtdetq *rte;
1471 	u_long hash;
1472 	int hlen = ip->ip_hl << 2;
1473 
1474 	++mrtstat.mrts_mfc_misses;
1475 
1476 	mrtstat.mrts_no_route++;
1477 	if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1478 	    log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n",
1479 		(u_long)ntohl(ip->ip_src.s_addr),
1480 		(u_long)ntohl(ip->ip_dst.s_addr));
1481 
1482 	/*
1483 	 * Allocate mbufs early so that we don't do extra work if we are
1484 	 * just going to fail anyway.  Make sure to pullup the header so
1485 	 * that other people can't step on it.
1486 	 */
1487 	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT);
1488 	if (rte == NULL) {
1489 	    MFC_UNLOCK();
1490 	    VIF_UNLOCK();
1491 	    return ENOBUFS;
1492 	}
1493 	mb0 = m_copypacket(m, M_DONTWAIT);
1494 	if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1495 	    mb0 = m_pullup(mb0, hlen);
1496 	if (mb0 == NULL) {
1497 	    free(rte, M_MRTABLE);
1498 	    MFC_UNLOCK();
1499 	    VIF_UNLOCK();
1500 	    return ENOBUFS;
1501 	}
1502 
1503 	/* is there an upcall waiting for this flow ? */
1504 	hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1505 	for (rt = mfctable[hash]; rt; rt = rt->mfc_next) {
1506 	    if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1507 		    (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1508 		    (rt->mfc_stall != NULL))
1509 		break;
1510 	}
1511 
1512 	if (rt == NULL) {
1513 	    int i;
1514 	    struct igmpmsg *im;
1515 	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1516 	    struct mbuf *mm;
1517 
1518 	    /*
1519 	     * Locate the vifi for the incoming interface for this packet.
1520 	     * If none found, drop packet.
1521 	     */
1522 	    for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1523 		;
1524 	    if (vifi >= numvifs)	/* vif not found, drop packet */
1525 		goto non_fatal;
1526 
1527 	    /* no upcall, so make a new entry */
1528 	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1529 	    if (rt == NULL)
1530 		goto fail;
1531 	    /* Make a copy of the header to send to the user level process */
1532 	    mm = m_copy(mb0, 0, hlen);
1533 	    if (mm == NULL)
1534 		goto fail1;
1535 
1536 	    /*
1537 	     * Send message to routing daemon to install
1538 	     * a route into the kernel table
1539 	     */
1540 
1541 	    im = mtod(mm, struct igmpmsg *);
1542 	    im->im_msgtype = IGMPMSG_NOCACHE;
1543 	    im->im_mbz = 0;
1544 	    im->im_vif = vifi;
1545 
1546 	    mrtstat.mrts_upcalls++;
1547 
1548 	    k_igmpsrc.sin_addr = ip->ip_src;
1549 	    if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1550 		log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1551 		++mrtstat.mrts_upq_sockfull;
1552 fail1:
1553 		free(rt, M_MRTABLE);
1554 fail:
1555 		free(rte, M_MRTABLE);
1556 		m_freem(mb0);
1557 		MFC_UNLOCK();
1558 		VIF_UNLOCK();
1559 		return ENOBUFS;
1560 	    }
1561 
1562 	    /* insert new entry at head of hash chain */
1563 	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1564 	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1565 	    rt->mfc_expire	      = UPCALL_EXPIRE;
1566 	    nexpire[hash]++;
1567 	    for (i = 0; i < numvifs; i++) {
1568 		rt->mfc_ttls[i] = 0;
1569 		rt->mfc_flags[i] = 0;
1570 	    }
1571 	    rt->mfc_parent = -1;
1572 
1573 	    rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */
1574 
1575 	    rt->mfc_bw_meter = NULL;
1576 
1577 	    /* link into table */
1578 	    rt->mfc_next   = mfctable[hash];
1579 	    mfctable[hash] = rt;
1580 	    rt->mfc_stall = rte;
1581 
1582 	} else {
1583 	    /* determine if q has overflowed */
1584 	    int npkts = 0;
1585 	    struct rtdetq **p;
1586 
1587 	    /*
1588 	     * XXX ouch! we need to append to the list, but we
1589 	     * only have a pointer to the front, so we have to
1590 	     * scan the entire list every time.
1591 	     */
1592 	    for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1593 		npkts++;
1594 
1595 	    if (npkts > MAX_UPQ) {
1596 		mrtstat.mrts_upq_ovflw++;
1597 non_fatal:
1598 		free(rte, M_MRTABLE);
1599 		m_freem(mb0);
1600 		MFC_UNLOCK();
1601 		VIF_UNLOCK();
1602 		return 0;
1603 	    }
1604 
1605 	    /* Add this entry to the end of the queue */
1606 	    *p = rte;
1607 	}
1608 
1609 	rte->m			= mb0;
1610 	rte->ifp		= ifp;
1611 	rte->next		= NULL;
1612 
1613 	MFC_UNLOCK();
1614 	VIF_UNLOCK();
1615 
1616 	return 0;
1617     }
1618 }
1619 
1620 /*
1621  * Clean up the cache entry if upcall is not serviced
1622  */
1623 static void
1624 expire_upcalls(void *unused)
1625 {
1626     struct rtdetq *rte;
1627     struct mfc *mfc, **nptr;
1628     int i;
1629 
1630     MFC_LOCK();
1631     for (i = 0; i < MFCTBLSIZ; i++) {
1632 	if (nexpire[i] == 0)
1633 	    continue;
1634 	nptr = &mfctable[i];
1635 	for (mfc = *nptr; mfc != NULL; mfc = *nptr) {
1636 	    /*
1637 	     * Skip real cache entries
1638 	     * Make sure it wasn't marked to not expire (shouldn't happen)
1639 	     * If it expires now
1640 	     */
1641 	    if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 &&
1642 		    --mfc->mfc_expire == 0) {
1643 		if (mrtdebug & DEBUG_EXPIRE)
1644 		    log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n",
1645 			(u_long)ntohl(mfc->mfc_origin.s_addr),
1646 			(u_long)ntohl(mfc->mfc_mcastgrp.s_addr));
1647 		/*
1648 		 * drop all the packets
1649 		 * free the mbuf with the pkt, if, timing info
1650 		 */
1651 		for (rte = mfc->mfc_stall; rte; ) {
1652 		    struct rtdetq *n = rte->next;
1653 
1654 		    m_freem(rte->m);
1655 		    free(rte, M_MRTABLE);
1656 		    rte = n;
1657 		}
1658 		++mrtstat.mrts_cache_cleanups;
1659 		nexpire[i]--;
1660 
1661 		/*
1662 		 * free the bw_meter entries
1663 		 */
1664 		while (mfc->mfc_bw_meter != NULL) {
1665 		    struct bw_meter *x = mfc->mfc_bw_meter;
1666 
1667 		    mfc->mfc_bw_meter = x->bm_mfc_next;
1668 		    free(x, M_BWMETER);
1669 		}
1670 
1671 		*nptr = mfc->mfc_next;
1672 		free(mfc, M_MRTABLE);
1673 	    } else {
1674 		nptr = &mfc->mfc_next;
1675 	    }
1676 	}
1677     }
1678     MFC_UNLOCK();
1679 
1680     callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
1681 }
1682 
1683 /*
1684  * Packet forwarding routine once entry in the cache is made
1685  */
1686 static int
1687 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1688 {
1689     struct ip  *ip = mtod(m, struct ip *);
1690     vifi_t vifi;
1691     int plen = ip->ip_len;
1692 
1693     VIF_LOCK_ASSERT();
1694 /*
1695  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1696  * input, they shouldn't get counted on output, so statistics keeping is
1697  * separate.
1698  */
1699 #define MC_SEND(ip,vifp,m) {				\
1700 		if ((vifp)->v_flags & VIFF_TUNNEL)	\
1701 		    encap_send((ip), (vifp), (m));	\
1702 		else					\
1703 		    phyint_send((ip), (vifp), (m));	\
1704 }
1705 
1706     /*
1707      * If xmt_vif is not -1, send on only the requested vif.
1708      *
1709      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1710      */
1711     if (xmt_vif < numvifs) {
1712 #ifdef PIM
1713 	if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1714 	    pim_register_send(ip, viftable + xmt_vif, m, rt);
1715 	else
1716 #endif
1717 	MC_SEND(ip, viftable + xmt_vif, m);
1718 	return 1;
1719     }
1720 
1721     /*
1722      * Don't forward if it didn't arrive from the parent vif for its origin.
1723      */
1724     vifi = rt->mfc_parent;
1725     if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1726 	/* came in the wrong interface */
1727 	if (mrtdebug & DEBUG_FORWARD)
1728 	    log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1729 		(void *)ifp, vifi, (void *)viftable[vifi].v_ifp);
1730 	++mrtstat.mrts_wrong_if;
1731 	++rt->mfc_wrong_if;
1732 	/*
1733 	 * If we are doing PIM assert processing, send a message
1734 	 * to the routing daemon.
1735 	 *
1736 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1737 	 * can complete the SPT switch, regardless of the type
1738 	 * of the iif (broadcast media, GRE tunnel, etc).
1739 	 */
1740 	if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1741 	    struct timeval now;
1742 	    u_long delta;
1743 
1744 #ifdef PIM
1745 	    if (ifp == &multicast_register_if)
1746 		pimstat.pims_rcv_registers_wrongiif++;
1747 #endif
1748 
1749 	    /* Get vifi for the incoming packet */
1750 	    for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1751 		;
1752 	    if (vifi >= numvifs)
1753 		return 0;	/* The iif is not found: ignore the packet. */
1754 
1755 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1756 		return 0;	/* WRONGVIF disabled: ignore the packet */
1757 
1758 	    GET_TIME(now);
1759 
1760 	    TV_DELTA(now, rt->mfc_last_assert, delta);
1761 
1762 	    if (delta > ASSERT_MSG_TIME) {
1763 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1764 		struct igmpmsg *im;
1765 		int hlen = ip->ip_hl << 2;
1766 		struct mbuf *mm = m_copy(m, 0, hlen);
1767 
1768 		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1769 		    mm = m_pullup(mm, hlen);
1770 		if (mm == NULL)
1771 		    return ENOBUFS;
1772 
1773 		rt->mfc_last_assert = now;
1774 
1775 		im = mtod(mm, struct igmpmsg *);
1776 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1777 		im->im_mbz		= 0;
1778 		im->im_vif		= vifi;
1779 
1780 		mrtstat.mrts_upcalls++;
1781 
1782 		k_igmpsrc.sin_addr = im->im_src;
1783 		if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1784 		    log(LOG_WARNING,
1785 			"ip_mforward: ip_mrouter socket queue full\n");
1786 		    ++mrtstat.mrts_upq_sockfull;
1787 		    return ENOBUFS;
1788 		}
1789 	    }
1790 	}
1791 	return 0;
1792     }
1793 
1794     /* If I sourced this packet, it counts as output, else it was input. */
1795     if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1796 	viftable[vifi].v_pkt_out++;
1797 	viftable[vifi].v_bytes_out += plen;
1798     } else {
1799 	viftable[vifi].v_pkt_in++;
1800 	viftable[vifi].v_bytes_in += plen;
1801     }
1802     rt->mfc_pkt_cnt++;
1803     rt->mfc_byte_cnt += plen;
1804 
1805     /*
1806      * For each vif, decide if a copy of the packet should be forwarded.
1807      * Forward if:
1808      *		- the ttl exceeds the vif's threshold
1809      *		- there are group members downstream on interface
1810      */
1811     for (vifi = 0; vifi < numvifs; vifi++)
1812 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1813 	    viftable[vifi].v_pkt_out++;
1814 	    viftable[vifi].v_bytes_out += plen;
1815 #ifdef PIM
1816 	    if (viftable[vifi].v_flags & VIFF_REGISTER)
1817 		pim_register_send(ip, viftable + vifi, m, rt);
1818 	    else
1819 #endif
1820 	    MC_SEND(ip, viftable+vifi, m);
1821 	}
1822 
1823     /*
1824      * Perform upcall-related bw measuring.
1825      */
1826     if (rt->mfc_bw_meter != NULL) {
1827 	struct bw_meter *x;
1828 	struct timeval now;
1829 
1830 	GET_TIME(now);
1831 	MFC_LOCK_ASSERT();
1832 	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1833 	    bw_meter_receive_packet(x, plen, &now);
1834     }
1835 
1836     return 0;
1837 }
1838 
1839 /*
1840  * check if a vif number is legal/ok. This is used by ip_output.
1841  */
1842 static int
1843 X_legal_vif_num(int vif)
1844 {
1845     /* XXX unlocked, matter? */
1846     return (vif >= 0 && vif < numvifs);
1847 }
1848 
1849 /*
1850  * Return the local address used by this vif
1851  */
1852 static u_long
1853 X_ip_mcast_src(int vifi)
1854 {
1855     /* XXX unlocked, matter? */
1856     if (vifi >= 0 && vifi < numvifs)
1857 	return viftable[vifi].v_lcl_addr.s_addr;
1858     else
1859 	return INADDR_ANY;
1860 }
1861 
1862 static void
1863 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1864 {
1865     struct mbuf *mb_copy;
1866     int hlen = ip->ip_hl << 2;
1867 
1868     VIF_LOCK_ASSERT();
1869 
1870     /*
1871      * Make a new reference to the packet; make sure that
1872      * the IP header is actually copied, not just referenced,
1873      * so that ip_output() only scribbles on the copy.
1874      */
1875     mb_copy = m_copypacket(m, M_DONTWAIT);
1876     if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1877 	mb_copy = m_pullup(mb_copy, hlen);
1878     if (mb_copy == NULL)
1879 	return;
1880 
1881     if (vifp->v_rate_limit == 0)
1882 	tbf_send_packet(vifp, mb_copy);
1883     else
1884 	tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1885 }
1886 
1887 static void
1888 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1889 {
1890     struct mbuf *mb_copy;
1891     struct ip *ip_copy;
1892     int i, len = ip->ip_len;
1893 
1894     VIF_LOCK_ASSERT();
1895 
1896     /* Take care of delayed checksums */
1897     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1898 	in_delayed_cksum(m);
1899 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1900     }
1901 
1902     /*
1903      * copy the old packet & pullup its IP header into the
1904      * new mbuf so we can modify it.  Try to fill the new
1905      * mbuf since if we don't the ethernet driver will.
1906      */
1907     MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1908     if (mb_copy == NULL)
1909 	return;
1910 #ifdef MAC
1911     mac_create_mbuf_multicast_encap(m, vifp->v_ifp, mb_copy);
1912 #endif
1913     mb_copy->m_data += max_linkhdr;
1914     mb_copy->m_len = sizeof(multicast_encap_iphdr);
1915 
1916     if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
1917 	m_freem(mb_copy);
1918 	return;
1919     }
1920     i = MHLEN - M_LEADINGSPACE(mb_copy);
1921     if (i > len)
1922 	i = len;
1923     mb_copy = m_pullup(mb_copy, i);
1924     if (mb_copy == NULL)
1925 	return;
1926     mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1927 
1928     /*
1929      * fill in the encapsulating IP header.
1930      */
1931     ip_copy = mtod(mb_copy, struct ip *);
1932     *ip_copy = multicast_encap_iphdr;
1933     ip_copy->ip_id = ip_newid();
1934     ip_copy->ip_len += len;
1935     ip_copy->ip_src = vifp->v_lcl_addr;
1936     ip_copy->ip_dst = vifp->v_rmt_addr;
1937 
1938     /*
1939      * turn the encapsulated IP header back into a valid one.
1940      */
1941     ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1942     --ip->ip_ttl;
1943     ip->ip_len = htons(ip->ip_len);
1944     ip->ip_off = htons(ip->ip_off);
1945     ip->ip_sum = 0;
1946     mb_copy->m_data += sizeof(multicast_encap_iphdr);
1947     ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1948     mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1949 
1950     if (vifp->v_rate_limit == 0)
1951 	tbf_send_packet(vifp, mb_copy);
1952     else
1953 	tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1954 }
1955 
1956 /*
1957  * Token bucket filter module
1958  */
1959 
1960 static void
1961 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_long p_len)
1962 {
1963     struct tbf *t = vifp->v_tbf;
1964 
1965     VIF_LOCK_ASSERT();
1966 
1967     if (p_len > MAX_BKT_SIZE) {		/* drop if packet is too large */
1968 	mrtstat.mrts_pkt2large++;
1969 	m_freem(m);
1970 	return;
1971     }
1972 
1973     tbf_update_tokens(vifp);
1974 
1975     if (t->tbf_q_len == 0) {		/* queue empty...		*/
1976 	if (p_len <= t->tbf_n_tok) {	/* send packet if enough tokens */
1977 	    t->tbf_n_tok -= p_len;
1978 	    tbf_send_packet(vifp, m);
1979 	} else {			/* no, queue packet and try later */
1980 	    tbf_queue(vifp, m);
1981 	    callout_reset(&tbf_reprocess_ch, TBF_REPROCESS,
1982 		tbf_reprocess_q, vifp);
1983 	}
1984     } else if (t->tbf_q_len < t->tbf_max_q_len) {
1985 	/* finite queue length, so queue pkts and process queue */
1986 	tbf_queue(vifp, m);
1987 	tbf_process_q(vifp);
1988     } else {
1989 	/* queue full, try to dq and queue and process */
1990 	if (!tbf_dq_sel(vifp, ip)) {
1991 	    mrtstat.mrts_q_overflow++;
1992 	    m_freem(m);
1993 	} else {
1994 	    tbf_queue(vifp, m);
1995 	    tbf_process_q(vifp);
1996 	}
1997     }
1998 }
1999 
2000 /*
2001  * adds a packet to the queue at the interface
2002  */
2003 static void
2004 tbf_queue(struct vif *vifp, struct mbuf *m)
2005 {
2006     struct tbf *t = vifp->v_tbf;
2007 
2008     VIF_LOCK_ASSERT();
2009 
2010     if (t->tbf_t == NULL)	/* Queue was empty */
2011 	t->tbf_q = m;
2012     else			/* Insert at tail */
2013 	t->tbf_t->m_act = m;
2014 
2015     t->tbf_t = m;		/* Set new tail pointer */
2016 
2017 #ifdef DIAGNOSTIC
2018     /* Make sure we didn't get fed a bogus mbuf */
2019     if (m->m_act)
2020 	panic("tbf_queue: m_act");
2021 #endif
2022     m->m_act = NULL;
2023 
2024     t->tbf_q_len++;
2025 }
2026 
2027 /*
2028  * processes the queue at the interface
2029  */
2030 static void
2031 tbf_process_q(struct vif *vifp)
2032 {
2033     struct tbf *t = vifp->v_tbf;
2034 
2035     VIF_LOCK_ASSERT();
2036 
2037     /* loop through the queue at the interface and send as many packets
2038      * as possible
2039      */
2040     while (t->tbf_q_len > 0) {
2041 	struct mbuf *m = t->tbf_q;
2042 	int len = mtod(m, struct ip *)->ip_len;
2043 
2044 	/* determine if the packet can be sent */
2045 	if (len > t->tbf_n_tok)	/* not enough tokens, we are done */
2046 	    break;
2047 	/* ok, reduce no of tokens, dequeue and send the packet. */
2048 	t->tbf_n_tok -= len;
2049 
2050 	t->tbf_q = m->m_act;
2051 	if (--t->tbf_q_len == 0)
2052 	    t->tbf_t = NULL;
2053 
2054 	m->m_act = NULL;
2055 	tbf_send_packet(vifp, m);
2056     }
2057 }
2058 
2059 static void
2060 tbf_reprocess_q(void *xvifp)
2061 {
2062     struct vif *vifp = xvifp;
2063 
2064     if (ip_mrouter == NULL)
2065 	return;
2066     VIF_LOCK();
2067     tbf_update_tokens(vifp);
2068     tbf_process_q(vifp);
2069     if (vifp->v_tbf->tbf_q_len)
2070 	callout_reset(&tbf_reprocess_ch, TBF_REPROCESS, tbf_reprocess_q, vifp);
2071     VIF_UNLOCK();
2072 }
2073 
2074 /* function that will selectively discard a member of the queue
2075  * based on the precedence value and the priority
2076  */
2077 static int
2078 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2079 {
2080     u_int p;
2081     struct mbuf *m, *last;
2082     struct mbuf **np;
2083     struct tbf *t = vifp->v_tbf;
2084 
2085     VIF_LOCK_ASSERT();
2086 
2087     p = priority(vifp, ip);
2088 
2089     np = &t->tbf_q;
2090     last = NULL;
2091     while ((m = *np) != NULL) {
2092 	if (p > priority(vifp, mtod(m, struct ip *))) {
2093 	    *np = m->m_act;
2094 	    /* If we're removing the last packet, fix the tail pointer */
2095 	    if (m == t->tbf_t)
2096 		t->tbf_t = last;
2097 	    m_freem(m);
2098 	    /* It's impossible for the queue to be empty, but check anyways. */
2099 	    if (--t->tbf_q_len == 0)
2100 		t->tbf_t = NULL;
2101 	    mrtstat.mrts_drop_sel++;
2102 	    return 1;
2103 	}
2104 	np = &m->m_act;
2105 	last = m;
2106     }
2107     return 0;
2108 }
2109 
2110 static void
2111 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2112 {
2113     VIF_LOCK_ASSERT();
2114 
2115     if (vifp->v_flags & VIFF_TUNNEL)	/* If tunnel options */
2116 	ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL);
2117     else {
2118 	struct ip_moptions imo;
2119 	int error;
2120 	static struct route ro; /* XXX check this */
2121 
2122 	imo.imo_multicast_ifp  = vifp->v_ifp;
2123 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
2124 	imo.imo_multicast_loop = 1;
2125 	imo.imo_multicast_vif  = -1;
2126 
2127 	/*
2128 	 * Re-entrancy should not be a problem here, because
2129 	 * the packets that we send out and are looped back at us
2130 	 * should get rejected because they appear to come from
2131 	 * the loopback interface, thus preventing looping.
2132 	 */
2133 	error = ip_output(m, NULL, &ro, IP_FORWARDING, &imo, NULL);
2134 
2135 	if (mrtdebug & DEBUG_XMIT)
2136 	    log(LOG_DEBUG, "phyint_send on vif %td err %d\n",
2137 		vifp - viftable, error);
2138     }
2139 }
2140 
2141 /* determine the current time and then
2142  * the elapsed time (between the last time and time now)
2143  * in milliseconds & update the no. of tokens in the bucket
2144  */
2145 static void
2146 tbf_update_tokens(struct vif *vifp)
2147 {
2148     struct timeval tp;
2149     u_long tm;
2150     struct tbf *t = vifp->v_tbf;
2151 
2152     VIF_LOCK_ASSERT();
2153 
2154     GET_TIME(tp);
2155 
2156     TV_DELTA(tp, t->tbf_last_pkt_t, tm);
2157 
2158     /*
2159      * This formula is actually
2160      * "time in seconds" * "bytes/second".
2161      *
2162      * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2163      *
2164      * The (1000/1024) was introduced in add_vif to optimize
2165      * this divide into a shift.
2166      */
2167     t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8;
2168     t->tbf_last_pkt_t = tp;
2169 
2170     if (t->tbf_n_tok > MAX_BKT_SIZE)
2171 	t->tbf_n_tok = MAX_BKT_SIZE;
2172 }
2173 
2174 static int
2175 priority(struct vif *vifp, struct ip *ip)
2176 {
2177     int prio = 50; /* the lowest priority -- default case */
2178 
2179     /* temporary hack; may add general packet classifier some day */
2180 
2181     /*
2182      * The UDP port space is divided up into four priority ranges:
2183      * [0, 16384)     : unclassified - lowest priority
2184      * [16384, 32768) : audio - highest priority
2185      * [32768, 49152) : whiteboard - medium priority
2186      * [49152, 65536) : video - low priority
2187      *
2188      * Everything else gets lowest priority.
2189      */
2190     if (ip->ip_p == IPPROTO_UDP) {
2191 	struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2192 	switch (ntohs(udp->uh_dport) & 0xc000) {
2193 	case 0x4000:
2194 	    prio = 70;
2195 	    break;
2196 	case 0x8000:
2197 	    prio = 60;
2198 	    break;
2199 	case 0xc000:
2200 	    prio = 55;
2201 	    break;
2202 	}
2203     }
2204     return prio;
2205 }
2206 
2207 /*
2208  * End of token bucket filter modifications
2209  */
2210 
2211 static int
2212 X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt)
2213 {
2214     int error, vifi;
2215 
2216     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2217 	return EOPNOTSUPP;
2218 
2219     error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
2220     if (error)
2221 	return error;
2222 
2223     VIF_LOCK();
2224 
2225     if (vifi < 0 || vifi >= numvifs) {	/* Error if vif is invalid */
2226 	VIF_UNLOCK();
2227 	return EADDRNOTAVAIL;
2228     }
2229 
2230     if (sopt->sopt_name == IP_RSVP_VIF_ON) {
2231 	/* Check if socket is available. */
2232 	if (viftable[vifi].v_rsvpd != NULL) {
2233 	    VIF_UNLOCK();
2234 	    return EADDRINUSE;
2235 	}
2236 
2237 	viftable[vifi].v_rsvpd = so;
2238 	/* This may seem silly, but we need to be sure we don't over-increment
2239 	 * the RSVP counter, in case something slips up.
2240 	 */
2241 	if (!viftable[vifi].v_rsvp_on) {
2242 	    viftable[vifi].v_rsvp_on = 1;
2243 	    rsvp_on++;
2244 	}
2245     } else { /* must be VIF_OFF */
2246 	/*
2247 	 * XXX as an additional consistency check, one could make sure
2248 	 * that viftable[vifi].v_rsvpd == so, otherwise passing so as
2249 	 * first parameter is pretty useless.
2250 	 */
2251 	viftable[vifi].v_rsvpd = NULL;
2252 	/*
2253 	 * This may seem silly, but we need to be sure we don't over-decrement
2254 	 * the RSVP counter, in case something slips up.
2255 	 */
2256 	if (viftable[vifi].v_rsvp_on) {
2257 	    viftable[vifi].v_rsvp_on = 0;
2258 	    rsvp_on--;
2259 	}
2260     }
2261     VIF_UNLOCK();
2262     return 0;
2263 }
2264 
2265 static void
2266 X_ip_rsvp_force_done(struct socket *so)
2267 {
2268     int vifi;
2269 
2270     /* Don't bother if it is not the right type of socket. */
2271     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2272 	return;
2273 
2274     VIF_LOCK();
2275 
2276     /* The socket may be attached to more than one vif...this
2277      * is perfectly legal.
2278      */
2279     for (vifi = 0; vifi < numvifs; vifi++) {
2280 	if (viftable[vifi].v_rsvpd == so) {
2281 	    viftable[vifi].v_rsvpd = NULL;
2282 	    /* This may seem silly, but we need to be sure we don't
2283 	     * over-decrement the RSVP counter, in case something slips up.
2284 	     */
2285 	    if (viftable[vifi].v_rsvp_on) {
2286 		viftable[vifi].v_rsvp_on = 0;
2287 		rsvp_on--;
2288 	    }
2289 	}
2290     }
2291 
2292     VIF_UNLOCK();
2293 }
2294 
2295 static void
2296 X_rsvp_input(struct mbuf *m, int off)
2297 {
2298     int vifi;
2299     struct ip *ip = mtod(m, struct ip *);
2300     struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
2301     struct ifnet *ifp;
2302 
2303     if (rsvpdebug)
2304 	printf("rsvp_input: rsvp_on %d\n",rsvp_on);
2305 
2306     /* Can still get packets with rsvp_on = 0 if there is a local member
2307      * of the group to which the RSVP packet is addressed.  But in this
2308      * case we want to throw the packet away.
2309      */
2310     if (!rsvp_on) {
2311 	m_freem(m);
2312 	return;
2313     }
2314 
2315     if (rsvpdebug)
2316 	printf("rsvp_input: check vifs\n");
2317 
2318 #ifdef DIAGNOSTIC
2319     M_ASSERTPKTHDR(m);
2320 #endif
2321 
2322     ifp = m->m_pkthdr.rcvif;
2323 
2324     VIF_LOCK();
2325     /* Find which vif the packet arrived on. */
2326     for (vifi = 0; vifi < numvifs; vifi++)
2327 	if (viftable[vifi].v_ifp == ifp)
2328 	    break;
2329 
2330     if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) {
2331 	/*
2332 	 * Drop the lock here to avoid holding it across rip_input.
2333 	 * This could make rsvpdebug printfs wrong.  If you care,
2334 	 * record the state of stuff before dropping the lock.
2335 	 */
2336 	VIF_UNLOCK();
2337 	/*
2338 	 * If the old-style non-vif-associated socket is set,
2339 	 * then use it.  Otherwise, drop packet since there
2340 	 * is no specific socket for this vif.
2341 	 */
2342 	if (ip_rsvpd != NULL) {
2343 	    if (rsvpdebug)
2344 		printf("rsvp_input: Sending packet up old-style socket\n");
2345 	    rip_input(m, off);  /* xxx */
2346 	} else {
2347 	    if (rsvpdebug && vifi == numvifs)
2348 		printf("rsvp_input: Can't find vif for packet.\n");
2349 	    else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL)
2350 		printf("rsvp_input: No socket defined for vif %d\n",vifi);
2351 	    m_freem(m);
2352 	}
2353 	return;
2354     }
2355     rsvp_src.sin_addr = ip->ip_src;
2356 
2357     if (rsvpdebug && m)
2358 	printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n",
2359 	       m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
2360 
2361     if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) {
2362 	if (rsvpdebug)
2363 	    printf("rsvp_input: Failed to append to socket\n");
2364     } else {
2365 	if (rsvpdebug)
2366 	    printf("rsvp_input: send packet up\n");
2367     }
2368     VIF_UNLOCK();
2369 }
2370 
2371 /*
2372  * Code for bandwidth monitors
2373  */
2374 
2375 /*
2376  * Define common interface for timeval-related methods
2377  */
2378 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
2379 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
2380 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
2381 
2382 static uint32_t
2383 compute_bw_meter_flags(struct bw_upcall *req)
2384 {
2385     uint32_t flags = 0;
2386 
2387     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2388 	flags |= BW_METER_UNIT_PACKETS;
2389     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2390 	flags |= BW_METER_UNIT_BYTES;
2391     if (req->bu_flags & BW_UPCALL_GEQ)
2392 	flags |= BW_METER_GEQ;
2393     if (req->bu_flags & BW_UPCALL_LEQ)
2394 	flags |= BW_METER_LEQ;
2395 
2396     return flags;
2397 }
2398 
2399 /*
2400  * Add a bw_meter entry
2401  */
2402 static int
2403 add_bw_upcall(struct bw_upcall *req)
2404 {
2405     struct mfc *mfc;
2406     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2407 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2408     struct timeval now;
2409     struct bw_meter *x;
2410     uint32_t flags;
2411 
2412     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2413 	return EOPNOTSUPP;
2414 
2415     /* Test if the flags are valid */
2416     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2417 	return EINVAL;
2418     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2419 	return EINVAL;
2420     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2421 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2422 	return EINVAL;
2423 
2424     /* Test if the threshold time interval is valid */
2425     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2426 	return EINVAL;
2427 
2428     flags = compute_bw_meter_flags(req);
2429 
2430     /*
2431      * Find if we have already same bw_meter entry
2432      */
2433     MFC_LOCK();
2434     mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
2435     if (mfc == NULL) {
2436 	MFC_UNLOCK();
2437 	return EADDRNOTAVAIL;
2438     }
2439     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2440 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2441 			   &req->bu_threshold.b_time, ==)) &&
2442 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2443 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2444 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
2445 	    MFC_UNLOCK();
2446 	    return 0;		/* XXX Already installed */
2447 	}
2448     }
2449 
2450     /* Allocate the new bw_meter entry */
2451     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2452     if (x == NULL) {
2453 	MFC_UNLOCK();
2454 	return ENOBUFS;
2455     }
2456 
2457     /* Set the new bw_meter entry */
2458     x->bm_threshold.b_time = req->bu_threshold.b_time;
2459     GET_TIME(now);
2460     x->bm_start_time = now;
2461     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2462     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2463     x->bm_measured.b_packets = 0;
2464     x->bm_measured.b_bytes = 0;
2465     x->bm_flags = flags;
2466     x->bm_time_next = NULL;
2467     x->bm_time_hash = BW_METER_BUCKETS;
2468 
2469     /* Add the new bw_meter entry to the front of entries for this MFC */
2470     x->bm_mfc = mfc;
2471     x->bm_mfc_next = mfc->mfc_bw_meter;
2472     mfc->mfc_bw_meter = x;
2473     schedule_bw_meter(x, &now);
2474     MFC_UNLOCK();
2475 
2476     return 0;
2477 }
2478 
2479 static void
2480 free_bw_list(struct bw_meter *list)
2481 {
2482     while (list != NULL) {
2483 	struct bw_meter *x = list;
2484 
2485 	list = list->bm_mfc_next;
2486 	unschedule_bw_meter(x);
2487 	free(x, M_BWMETER);
2488     }
2489 }
2490 
2491 /*
2492  * Delete one or multiple bw_meter entries
2493  */
2494 static int
2495 del_bw_upcall(struct bw_upcall *req)
2496 {
2497     struct mfc *mfc;
2498     struct bw_meter *x;
2499 
2500     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2501 	return EOPNOTSUPP;
2502 
2503     MFC_LOCK();
2504     /* Find the corresponding MFC entry */
2505     mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
2506     if (mfc == NULL) {
2507 	MFC_UNLOCK();
2508 	return EADDRNOTAVAIL;
2509     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2510 	/*
2511 	 * Delete all bw_meter entries for this mfc
2512 	 */
2513 	struct bw_meter *list;
2514 
2515 	list = mfc->mfc_bw_meter;
2516 	mfc->mfc_bw_meter = NULL;
2517 	free_bw_list(list);
2518 	MFC_UNLOCK();
2519 	return 0;
2520     } else {			/* Delete a single bw_meter entry */
2521 	struct bw_meter *prev;
2522 	uint32_t flags = 0;
2523 
2524 	flags = compute_bw_meter_flags(req);
2525 
2526 	/* Find the bw_meter entry to delete */
2527 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2528 	     prev = x, x = x->bm_mfc_next) {
2529 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2530 			       &req->bu_threshold.b_time, ==)) &&
2531 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2532 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2533 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2534 		break;
2535 	}
2536 	if (x != NULL) { /* Delete entry from the list for this MFC */
2537 	    if (prev != NULL)
2538 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2539 	    else
2540 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2541 
2542 	    unschedule_bw_meter(x);
2543 	    MFC_UNLOCK();
2544 	    /* Free the bw_meter entry */
2545 	    free(x, M_BWMETER);
2546 	    return 0;
2547 	} else {
2548 	    MFC_UNLOCK();
2549 	    return EINVAL;
2550 	}
2551     }
2552     /* NOTREACHED */
2553 }
2554 
2555 /*
2556  * Perform bandwidth measurement processing that may result in an upcall
2557  */
2558 static void
2559 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2560 {
2561     struct timeval delta;
2562 
2563     MFC_LOCK_ASSERT();
2564 
2565     delta = *nowp;
2566     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2567 
2568     if (x->bm_flags & BW_METER_GEQ) {
2569 	/*
2570 	 * Processing for ">=" type of bw_meter entry
2571 	 */
2572 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2573 	    /* Reset the bw_meter entry */
2574 	    x->bm_start_time = *nowp;
2575 	    x->bm_measured.b_packets = 0;
2576 	    x->bm_measured.b_bytes = 0;
2577 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2578 	}
2579 
2580 	/* Record that a packet is received */
2581 	x->bm_measured.b_packets++;
2582 	x->bm_measured.b_bytes += plen;
2583 
2584 	/*
2585 	 * Test if we should deliver an upcall
2586 	 */
2587 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2588 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2589 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2590 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2591 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2592 		/* Prepare an upcall for delivery */
2593 		bw_meter_prepare_upcall(x, nowp);
2594 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2595 	    }
2596 	}
2597     } else if (x->bm_flags & BW_METER_LEQ) {
2598 	/*
2599 	 * Processing for "<=" type of bw_meter entry
2600 	 */
2601 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2602 	    /*
2603 	     * We are behind time with the multicast forwarding table
2604 	     * scanning for "<=" type of bw_meter entries, so test now
2605 	     * if we should deliver an upcall.
2606 	     */
2607 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2608 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2609 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2610 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2611 		/* Prepare an upcall for delivery */
2612 		bw_meter_prepare_upcall(x, nowp);
2613 	    }
2614 	    /* Reschedule the bw_meter entry */
2615 	    unschedule_bw_meter(x);
2616 	    schedule_bw_meter(x, nowp);
2617 	}
2618 
2619 	/* Record that a packet is received */
2620 	x->bm_measured.b_packets++;
2621 	x->bm_measured.b_bytes += plen;
2622 
2623 	/*
2624 	 * Test if we should restart the measuring interval
2625 	 */
2626 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2627 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2628 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
2629 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2630 	    /* Don't restart the measuring interval */
2631 	} else {
2632 	    /* Do restart the measuring interval */
2633 	    /*
2634 	     * XXX: note that we don't unschedule and schedule, because this
2635 	     * might be too much overhead per packet. Instead, when we process
2636 	     * all entries for a given timer hash bin, we check whether it is
2637 	     * really a timeout. If not, we reschedule at that time.
2638 	     */
2639 	    x->bm_start_time = *nowp;
2640 	    x->bm_measured.b_packets = 0;
2641 	    x->bm_measured.b_bytes = 0;
2642 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2643 	}
2644     }
2645 }
2646 
2647 /*
2648  * Prepare a bandwidth-related upcall
2649  */
2650 static void
2651 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2652 {
2653     struct timeval delta;
2654     struct bw_upcall *u;
2655 
2656     MFC_LOCK_ASSERT();
2657 
2658     /*
2659      * Compute the measured time interval
2660      */
2661     delta = *nowp;
2662     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2663 
2664     /*
2665      * If there are too many pending upcalls, deliver them now
2666      */
2667     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2668 	bw_upcalls_send();
2669 
2670     /*
2671      * Set the bw_upcall entry
2672      */
2673     u = &bw_upcalls[bw_upcalls_n++];
2674     u->bu_src = x->bm_mfc->mfc_origin;
2675     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2676     u->bu_threshold.b_time = x->bm_threshold.b_time;
2677     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2678     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2679     u->bu_measured.b_time = delta;
2680     u->bu_measured.b_packets = x->bm_measured.b_packets;
2681     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2682     u->bu_flags = 0;
2683     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2684 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2685     if (x->bm_flags & BW_METER_UNIT_BYTES)
2686 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2687     if (x->bm_flags & BW_METER_GEQ)
2688 	u->bu_flags |= BW_UPCALL_GEQ;
2689     if (x->bm_flags & BW_METER_LEQ)
2690 	u->bu_flags |= BW_UPCALL_LEQ;
2691 }
2692 
2693 /*
2694  * Send the pending bandwidth-related upcalls
2695  */
2696 static void
2697 bw_upcalls_send(void)
2698 {
2699     struct mbuf *m;
2700     int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2701     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2702     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2703 				      0,		/* unused2 */
2704 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2705 				      0,		/* im_mbz  */
2706 				      0,		/* im_vif  */
2707 				      0,		/* unused3 */
2708 				      { 0 },		/* im_src  */
2709 				      { 0 } };		/* im_dst  */
2710 
2711     MFC_LOCK_ASSERT();
2712 
2713     if (bw_upcalls_n == 0)
2714 	return;			/* No pending upcalls */
2715 
2716     bw_upcalls_n = 0;
2717 
2718     /*
2719      * Allocate a new mbuf, initialize it with the header and
2720      * the payload for the pending calls.
2721      */
2722     MGETHDR(m, M_DONTWAIT, MT_DATA);
2723     if (m == NULL) {
2724 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2725 	return;
2726     }
2727 
2728     m->m_len = m->m_pkthdr.len = 0;
2729     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2730     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2731 
2732     /*
2733      * Send the upcalls
2734      * XXX do we need to set the address in k_igmpsrc ?
2735      */
2736     mrtstat.mrts_upcalls++;
2737     if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2738 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2739 	++mrtstat.mrts_upq_sockfull;
2740     }
2741 }
2742 
2743 /*
2744  * Compute the timeout hash value for the bw_meter entries
2745  */
2746 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2747     do {								\
2748 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2749 									\
2750 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2751 	(hash) = next_timeval.tv_sec;					\
2752 	if (next_timeval.tv_usec)					\
2753 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2754 	(hash) %= BW_METER_BUCKETS;					\
2755     } while (0)
2756 
2757 /*
2758  * Schedule a timer to process periodically bw_meter entry of type "<="
2759  * by linking the entry in the proper hash bucket.
2760  */
2761 static void
2762 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2763 {
2764     int time_hash;
2765 
2766     MFC_LOCK_ASSERT();
2767 
2768     if (!(x->bm_flags & BW_METER_LEQ))
2769 	return;		/* XXX: we schedule timers only for "<=" entries */
2770 
2771     /*
2772      * Reset the bw_meter entry
2773      */
2774     x->bm_start_time = *nowp;
2775     x->bm_measured.b_packets = 0;
2776     x->bm_measured.b_bytes = 0;
2777     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2778 
2779     /*
2780      * Compute the timeout hash value and insert the entry
2781      */
2782     BW_METER_TIMEHASH(x, time_hash);
2783     x->bm_time_next = bw_meter_timers[time_hash];
2784     bw_meter_timers[time_hash] = x;
2785     x->bm_time_hash = time_hash;
2786 }
2787 
2788 /*
2789  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2790  * by removing the entry from the proper hash bucket.
2791  */
2792 static void
2793 unschedule_bw_meter(struct bw_meter *x)
2794 {
2795     int time_hash;
2796     struct bw_meter *prev, *tmp;
2797 
2798     MFC_LOCK_ASSERT();
2799 
2800     if (!(x->bm_flags & BW_METER_LEQ))
2801 	return;		/* XXX: we schedule timers only for "<=" entries */
2802 
2803     /*
2804      * Compute the timeout hash value and delete the entry
2805      */
2806     time_hash = x->bm_time_hash;
2807     if (time_hash >= BW_METER_BUCKETS)
2808 	return;		/* Entry was not scheduled */
2809 
2810     for (prev = NULL, tmp = bw_meter_timers[time_hash];
2811 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2812 	if (tmp == x)
2813 	    break;
2814 
2815     if (tmp == NULL)
2816 	panic("unschedule_bw_meter: bw_meter entry not found");
2817 
2818     if (prev != NULL)
2819 	prev->bm_time_next = x->bm_time_next;
2820     else
2821 	bw_meter_timers[time_hash] = x->bm_time_next;
2822 
2823     x->bm_time_next = NULL;
2824     x->bm_time_hash = BW_METER_BUCKETS;
2825 }
2826 
2827 
2828 /*
2829  * Process all "<=" type of bw_meter that should be processed now,
2830  * and for each entry prepare an upcall if necessary. Each processed
2831  * entry is rescheduled again for the (periodic) processing.
2832  *
2833  * This is run periodically (once per second normally). On each round,
2834  * all the potentially matching entries are in the hash slot that we are
2835  * looking at.
2836  */
2837 static void
2838 bw_meter_process()
2839 {
2840     static uint32_t last_tv_sec;	/* last time we processed this */
2841 
2842     uint32_t loops;
2843     int i;
2844     struct timeval now, process_endtime;
2845 
2846     GET_TIME(now);
2847     if (last_tv_sec == now.tv_sec)
2848 	return;		/* nothing to do */
2849 
2850     loops = now.tv_sec - last_tv_sec;
2851     last_tv_sec = now.tv_sec;
2852     if (loops > BW_METER_BUCKETS)
2853 	loops = BW_METER_BUCKETS;
2854 
2855     MFC_LOCK();
2856     /*
2857      * Process all bins of bw_meter entries from the one after the last
2858      * processed to the current one. On entry, i points to the last bucket
2859      * visited, so we need to increment i at the beginning of the loop.
2860      */
2861     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2862 	struct bw_meter *x, *tmp_list;
2863 
2864 	if (++i >= BW_METER_BUCKETS)
2865 	    i = 0;
2866 
2867 	/* Disconnect the list of bw_meter entries from the bin */
2868 	tmp_list = bw_meter_timers[i];
2869 	bw_meter_timers[i] = NULL;
2870 
2871 	/* Process the list of bw_meter entries */
2872 	while (tmp_list != NULL) {
2873 	    x = tmp_list;
2874 	    tmp_list = tmp_list->bm_time_next;
2875 
2876 	    /* Test if the time interval is over */
2877 	    process_endtime = x->bm_start_time;
2878 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2879 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2880 		/* Not yet: reschedule, but don't reset */
2881 		int time_hash;
2882 
2883 		BW_METER_TIMEHASH(x, time_hash);
2884 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2885 		    /*
2886 		     * XXX: somehow the bin processing is a bit ahead of time.
2887 		     * Put the entry in the next bin.
2888 		     */
2889 		    if (++time_hash >= BW_METER_BUCKETS)
2890 			time_hash = 0;
2891 		}
2892 		x->bm_time_next = bw_meter_timers[time_hash];
2893 		bw_meter_timers[time_hash] = x;
2894 		x->bm_time_hash = time_hash;
2895 
2896 		continue;
2897 	    }
2898 
2899 	    /*
2900 	     * Test if we should deliver an upcall
2901 	     */
2902 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2903 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2904 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2905 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2906 		/* Prepare an upcall for delivery */
2907 		bw_meter_prepare_upcall(x, &now);
2908 	    }
2909 
2910 	    /*
2911 	     * Reschedule for next processing
2912 	     */
2913 	    schedule_bw_meter(x, &now);
2914 	}
2915     }
2916 
2917     /* Send all upcalls that are pending delivery */
2918     bw_upcalls_send();
2919 
2920     MFC_UNLOCK();
2921 }
2922 
2923 /*
2924  * A periodic function for sending all upcalls that are pending delivery
2925  */
2926 static void
2927 expire_bw_upcalls_send(void *unused)
2928 {
2929     MFC_LOCK();
2930     bw_upcalls_send();
2931     MFC_UNLOCK();
2932 
2933     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2934 	expire_bw_upcalls_send, NULL);
2935 }
2936 
2937 /*
2938  * A periodic function for periodic scanning of the multicast forwarding
2939  * table for processing all "<=" bw_meter entries.
2940  */
2941 static void
2942 expire_bw_meter_process(void *unused)
2943 {
2944     if (mrt_api_config & MRT_MFC_BW_UPCALL)
2945 	bw_meter_process();
2946 
2947     callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
2948 }
2949 
2950 /*
2951  * End of bandwidth monitoring code
2952  */
2953 
2954 #ifdef PIM
2955 /*
2956  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2957  *
2958  */
2959 static int
2960 pim_register_send(struct ip *ip, struct vif *vifp,
2961 	struct mbuf *m, struct mfc *rt)
2962 {
2963     struct mbuf *mb_copy, *mm;
2964 
2965     if (mrtdebug & DEBUG_PIM)
2966 	log(LOG_DEBUG, "pim_register_send: ");
2967 
2968     mb_copy = pim_register_prepare(ip, m);
2969     if (mb_copy == NULL)
2970 	return ENOBUFS;
2971 
2972     /*
2973      * Send all the fragments. Note that the mbuf for each fragment
2974      * is freed by the sending machinery.
2975      */
2976     for (mm = mb_copy; mm; mm = mb_copy) {
2977 	mb_copy = mm->m_nextpkt;
2978 	mm->m_nextpkt = 0;
2979 	mm = m_pullup(mm, sizeof(struct ip));
2980 	if (mm != NULL) {
2981 	    ip = mtod(mm, struct ip *);
2982 	    if ((mrt_api_config & MRT_MFC_RP) &&
2983 		(rt->mfc_rp.s_addr != INADDR_ANY)) {
2984 		pim_register_send_rp(ip, vifp, mm, rt);
2985 	    } else {
2986 		pim_register_send_upcall(ip, vifp, mm, rt);
2987 	    }
2988 	}
2989     }
2990 
2991     return 0;
2992 }
2993 
2994 /*
2995  * Return a copy of the data packet that is ready for PIM Register
2996  * encapsulation.
2997  * XXX: Note that in the returned copy the IP header is a valid one.
2998  */
2999 static struct mbuf *
3000 pim_register_prepare(struct ip *ip, struct mbuf *m)
3001 {
3002     struct mbuf *mb_copy = NULL;
3003     int mtu;
3004 
3005     /* Take care of delayed checksums */
3006     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3007 	in_delayed_cksum(m);
3008 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3009     }
3010 
3011     /*
3012      * Copy the old packet & pullup its IP header into the
3013      * new mbuf so we can modify it.
3014      */
3015     mb_copy = m_copypacket(m, M_DONTWAIT);
3016     if (mb_copy == NULL)
3017 	return NULL;
3018     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3019     if (mb_copy == NULL)
3020 	return NULL;
3021 
3022     /* take care of the TTL */
3023     ip = mtod(mb_copy, struct ip *);
3024     --ip->ip_ttl;
3025 
3026     /* Compute the MTU after the PIM Register encapsulation */
3027     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3028 
3029     if (ip->ip_len <= mtu) {
3030 	/* Turn the IP header into a valid one */
3031 	ip->ip_len = htons(ip->ip_len);
3032 	ip->ip_off = htons(ip->ip_off);
3033 	ip->ip_sum = 0;
3034 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3035     } else {
3036 	/* Fragment the packet */
3037 	if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
3038 	    m_freem(mb_copy);
3039 	    return NULL;
3040 	}
3041     }
3042     return mb_copy;
3043 }
3044 
3045 /*
3046  * Send an upcall with the data packet to the user-level process.
3047  */
3048 static int
3049 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3050 	struct mbuf *mb_copy, struct mfc *rt)
3051 {
3052     struct mbuf *mb_first;
3053     int len = ntohs(ip->ip_len);
3054     struct igmpmsg *im;
3055     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
3056 
3057     VIF_LOCK_ASSERT();
3058 
3059     /*
3060      * Add a new mbuf with an upcall header
3061      */
3062     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
3063     if (mb_first == NULL) {
3064 	m_freem(mb_copy);
3065 	return ENOBUFS;
3066     }
3067     mb_first->m_data += max_linkhdr;
3068     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3069     mb_first->m_len = sizeof(struct igmpmsg);
3070     mb_first->m_next = mb_copy;
3071 
3072     /* Send message to routing daemon */
3073     im = mtod(mb_first, struct igmpmsg *);
3074     im->im_msgtype	= IGMPMSG_WHOLEPKT;
3075     im->im_mbz		= 0;
3076     im->im_vif		= vifp - viftable;
3077     im->im_src		= ip->ip_src;
3078     im->im_dst		= ip->ip_dst;
3079 
3080     k_igmpsrc.sin_addr	= ip->ip_src;
3081 
3082     mrtstat.mrts_upcalls++;
3083 
3084     if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3085 	if (mrtdebug & DEBUG_PIM)
3086 	    log(LOG_WARNING,
3087 		"mcast: pim_register_send_upcall: ip_mrouter socket queue full");
3088 	++mrtstat.mrts_upq_sockfull;
3089 	return ENOBUFS;
3090     }
3091 
3092     /* Keep statistics */
3093     pimstat.pims_snd_registers_msgs++;
3094     pimstat.pims_snd_registers_bytes += len;
3095 
3096     return 0;
3097 }
3098 
3099 /*
3100  * Encapsulate the data packet in PIM Register message and send it to the RP.
3101  */
3102 static int
3103 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3104 	struct mbuf *mb_copy, struct mfc *rt)
3105 {
3106     struct mbuf *mb_first;
3107     struct ip *ip_outer;
3108     struct pim_encap_pimhdr *pimhdr;
3109     int len = ntohs(ip->ip_len);
3110     vifi_t vifi = rt->mfc_parent;
3111 
3112     VIF_LOCK_ASSERT();
3113 
3114     if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) {
3115 	m_freem(mb_copy);
3116 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
3117     }
3118 
3119     /*
3120      * Add a new mbuf with the encapsulating header
3121      */
3122     MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
3123     if (mb_first == NULL) {
3124 	m_freem(mb_copy);
3125 	return ENOBUFS;
3126     }
3127     mb_first->m_data += max_linkhdr;
3128     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3129     mb_first->m_next = mb_copy;
3130 
3131     mb_first->m_pkthdr.len = len + mb_first->m_len;
3132 
3133     /*
3134      * Fill in the encapsulating IP and PIM header
3135      */
3136     ip_outer = mtod(mb_first, struct ip *);
3137     *ip_outer = pim_encap_iphdr;
3138     ip_outer->ip_id = ip_newid();
3139     ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3140     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3141     ip_outer->ip_dst = rt->mfc_rp;
3142     /*
3143      * Copy the inner header TOS to the outer header, and take care of the
3144      * IP_DF bit.
3145      */
3146     ip_outer->ip_tos = ip->ip_tos;
3147     if (ntohs(ip->ip_off) & IP_DF)
3148 	ip_outer->ip_off |= IP_DF;
3149     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
3150 					 + sizeof(pim_encap_iphdr));
3151     *pimhdr = pim_encap_pimhdr;
3152     /* If the iif crosses a border, set the Border-bit */
3153     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3154 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3155 
3156     mb_first->m_data += sizeof(pim_encap_iphdr);
3157     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3158     mb_first->m_data -= sizeof(pim_encap_iphdr);
3159 
3160     if (vifp->v_rate_limit == 0)
3161 	tbf_send_packet(vifp, mb_first);
3162     else
3163 	tbf_control(vifp, mb_first, ip, ip_outer->ip_len);
3164 
3165     /* Keep statistics */
3166     pimstat.pims_snd_registers_msgs++;
3167     pimstat.pims_snd_registers_bytes += len;
3168 
3169     return 0;
3170 }
3171 
3172 /*
3173  * PIM-SMv2 and PIM-DM messages processing.
3174  * Receives and verifies the PIM control messages, and passes them
3175  * up to the listening socket, using rip_input().
3176  * The only message with special processing is the PIM_REGISTER message
3177  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3178  * is passed to if_simloop().
3179  */
3180 void
3181 pim_input(struct mbuf *m, int off)
3182 {
3183     struct ip *ip = mtod(m, struct ip *);
3184     struct pim *pim;
3185     int minlen;
3186     int datalen = ip->ip_len;
3187     int ip_tos;
3188     int iphlen = off;
3189 
3190     /* Keep statistics */
3191     pimstat.pims_rcv_total_msgs++;
3192     pimstat.pims_rcv_total_bytes += datalen;
3193 
3194     /*
3195      * Validate lengths
3196      */
3197     if (datalen < PIM_MINLEN) {
3198 	pimstat.pims_rcv_tooshort++;
3199 	log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3200 	    datalen, (u_long)ip->ip_src.s_addr);
3201 	m_freem(m);
3202 	return;
3203     }
3204 
3205     /*
3206      * If the packet is at least as big as a REGISTER, go agead
3207      * and grab the PIM REGISTER header size, to avoid another
3208      * possible m_pullup() later.
3209      *
3210      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
3211      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3212      */
3213     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3214     /*
3215      * Get the IP and PIM headers in contiguous memory, and
3216      * possibly the PIM REGISTER header.
3217      */
3218     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3219 	(m = m_pullup(m, minlen)) == 0) {
3220 	log(LOG_ERR, "pim_input: m_pullup failure\n");
3221 	return;
3222     }
3223     /* m_pullup() may have given us a new mbuf so reset ip. */
3224     ip = mtod(m, struct ip *);
3225     ip_tos = ip->ip_tos;
3226 
3227     /* adjust mbuf to point to the PIM header */
3228     m->m_data += iphlen;
3229     m->m_len  -= iphlen;
3230     pim = mtod(m, struct pim *);
3231 
3232     /*
3233      * Validate checksum. If PIM REGISTER, exclude the data packet.
3234      *
3235      * XXX: some older PIMv2 implementations don't make this distinction,
3236      * so for compatibility reason perform the checksum over part of the
3237      * message, and if error, then over the whole message.
3238      */
3239     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3240 	/* do nothing, checksum okay */
3241     } else if (in_cksum(m, datalen)) {
3242 	pimstat.pims_rcv_badsum++;
3243 	if (mrtdebug & DEBUG_PIM)
3244 	    log(LOG_DEBUG, "pim_input: invalid checksum");
3245 	m_freem(m);
3246 	return;
3247     }
3248 
3249     /* PIM version check */
3250     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3251 	pimstat.pims_rcv_badversion++;
3252 	log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3253 	    PIM_VT_V(pim->pim_vt), PIM_VERSION);
3254 	m_freem(m);
3255 	return;
3256     }
3257 
3258     /* restore mbuf back to the outer IP */
3259     m->m_data -= iphlen;
3260     m->m_len  += iphlen;
3261 
3262     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3263 	/*
3264 	 * Since this is a REGISTER, we'll make a copy of the register
3265 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3266 	 * routing daemon.
3267 	 */
3268 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
3269 	struct mbuf *mcp;
3270 	struct ip *encap_ip;
3271 	u_int32_t *reghdr;
3272 	struct ifnet *vifp;
3273 
3274 	VIF_LOCK();
3275 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3276 	    VIF_UNLOCK();
3277 	    if (mrtdebug & DEBUG_PIM)
3278 		log(LOG_DEBUG,
3279 		    "pim_input: register vif not set: %d\n", reg_vif_num);
3280 	    m_freem(m);
3281 	    return;
3282 	}
3283 	/* XXX need refcnt? */
3284 	vifp = viftable[reg_vif_num].v_ifp;
3285 	VIF_UNLOCK();
3286 
3287 	/*
3288 	 * Validate length
3289 	 */
3290 	if (datalen < PIM_REG_MINLEN) {
3291 	    pimstat.pims_rcv_tooshort++;
3292 	    pimstat.pims_rcv_badregisters++;
3293 	    log(LOG_ERR,
3294 		"pim_input: register packet size too small %d from %lx\n",
3295 		datalen, (u_long)ip->ip_src.s_addr);
3296 	    m_freem(m);
3297 	    return;
3298 	}
3299 
3300 	reghdr = (u_int32_t *)(pim + 1);
3301 	encap_ip = (struct ip *)(reghdr + 1);
3302 
3303 	if (mrtdebug & DEBUG_PIM) {
3304 	    log(LOG_DEBUG,
3305 		"pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3306 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3307 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3308 		ntohs(encap_ip->ip_len));
3309 	}
3310 
3311 	/* verify the version number of the inner packet */
3312 	if (encap_ip->ip_v != IPVERSION) {
3313 	    pimstat.pims_rcv_badregisters++;
3314 	    if (mrtdebug & DEBUG_PIM) {
3315 		log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3316 		    "of the inner packet\n", encap_ip->ip_v);
3317 	    }
3318 	    m_freem(m);
3319 	    return;
3320 	}
3321 
3322 	/* verify the inner packet is destined to a mcast group */
3323 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
3324 	    pimstat.pims_rcv_badregisters++;
3325 	    if (mrtdebug & DEBUG_PIM)
3326 		log(LOG_DEBUG,
3327 		    "pim_input: inner packet of register is not "
3328 		    "multicast %lx\n",
3329 		    (u_long)ntohl(encap_ip->ip_dst.s_addr));
3330 	    m_freem(m);
3331 	    return;
3332 	}
3333 
3334 	/* If a NULL_REGISTER, pass it to the daemon */
3335 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3336 	    goto pim_input_to_daemon;
3337 
3338 	/*
3339 	 * Copy the TOS from the outer IP header to the inner IP header.
3340 	 */
3341 	if (encap_ip->ip_tos != ip_tos) {
3342 	    /* Outer TOS -> inner TOS */
3343 	    encap_ip->ip_tos = ip_tos;
3344 	    /* Recompute the inner header checksum. Sigh... */
3345 
3346 	    /* adjust mbuf to point to the inner IP header */
3347 	    m->m_data += (iphlen + PIM_MINLEN);
3348 	    m->m_len  -= (iphlen + PIM_MINLEN);
3349 
3350 	    encap_ip->ip_sum = 0;
3351 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3352 
3353 	    /* restore mbuf to point back to the outer IP header */
3354 	    m->m_data -= (iphlen + PIM_MINLEN);
3355 	    m->m_len  += (iphlen + PIM_MINLEN);
3356 	}
3357 
3358 	/*
3359 	 * Decapsulate the inner IP packet and loopback to forward it
3360 	 * as a normal multicast packet. Also, make a copy of the
3361 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
3362 	 * to pass to the daemon later, so it can take the appropriate
3363 	 * actions (e.g., send back PIM_REGISTER_STOP).
3364 	 * XXX: here m->m_data points to the outer IP header.
3365 	 */
3366 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
3367 	if (mcp == NULL) {
3368 	    log(LOG_ERR,
3369 		"pim_input: pim register: could not copy register head\n");
3370 	    m_freem(m);
3371 	    return;
3372 	}
3373 
3374 	/* Keep statistics */
3375 	/* XXX: registers_bytes include only the encap. mcast pkt */
3376 	pimstat.pims_rcv_registers_msgs++;
3377 	pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3378 
3379 	/*
3380 	 * forward the inner ip packet; point m_data at the inner ip.
3381 	 */
3382 	m_adj(m, iphlen + PIM_MINLEN);
3383 
3384 	if (mrtdebug & DEBUG_PIM) {
3385 	    log(LOG_DEBUG,
3386 		"pim_input: forwarding decapsulated register: "
3387 		"src %lx, dst %lx, vif %d\n",
3388 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3389 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3390 		reg_vif_num);
3391 	}
3392 	/* NB: vifp was collected above; can it change on us? */
3393 	if_simloop(vifp, m, dst.sin_family, 0);
3394 
3395 	/* prepare the register head to send to the mrouting daemon */
3396 	m = mcp;
3397     }
3398 
3399 pim_input_to_daemon:
3400     /*
3401      * Pass the PIM message up to the daemon; if it is a Register message,
3402      * pass the 'head' only up to the daemon. This includes the
3403      * outer IP header, PIM header, PIM-Register header and the
3404      * inner IP header.
3405      * XXX: the outer IP header pkt size of a Register is not adjust to
3406      * reflect the fact that the inner multicast data is truncated.
3407      */
3408     rip_input(m, iphlen);
3409 
3410     return;
3411 }
3412 #endif /* PIM */
3413 
3414 static int
3415 ip_mroute_modevent(module_t mod, int type, void *unused)
3416 {
3417     switch (type) {
3418     case MOD_LOAD:
3419 	mtx_init(&mrouter_mtx, "mrouter initialization", NULL, MTX_DEF);
3420 	MFC_LOCK_INIT();
3421 	VIF_LOCK_INIT();
3422 	ip_mrouter_reset();
3423 	ip_mcast_src = X_ip_mcast_src;
3424 	ip_mforward = X_ip_mforward;
3425 	ip_mrouter_done = X_ip_mrouter_done;
3426 	ip_mrouter_get = X_ip_mrouter_get;
3427 	ip_mrouter_set = X_ip_mrouter_set;
3428 	ip_rsvp_force_done = X_ip_rsvp_force_done;
3429 	ip_rsvp_vif = X_ip_rsvp_vif;
3430 	legal_vif_num = X_legal_vif_num;
3431 	mrt_ioctl = X_mrt_ioctl;
3432 	rsvp_input_p = X_rsvp_input;
3433 	break;
3434 
3435     case MOD_UNLOAD:
3436 	/*
3437 	 * Typically module unload happens after the user-level
3438 	 * process has shutdown the kernel services (the check
3439 	 * below insures someone can't just yank the module out
3440 	 * from under a running process).  But if the module is
3441 	 * just loaded and then unloaded w/o starting up a user
3442 	 * process we still need to cleanup.
3443 	 */
3444 	if (ip_mrouter)
3445 	    return EINVAL;
3446 
3447 	X_ip_mrouter_done();
3448 	ip_mcast_src = NULL;
3449 	ip_mforward = NULL;
3450 	ip_mrouter_done = NULL;
3451 	ip_mrouter_get = NULL;
3452 	ip_mrouter_set = NULL;
3453 	ip_rsvp_force_done = NULL;
3454 	ip_rsvp_vif = NULL;
3455 	legal_vif_num = NULL;
3456 	mrt_ioctl = NULL;
3457 	rsvp_input_p = NULL;
3458 	VIF_LOCK_DESTROY();
3459 	MFC_LOCK_DESTROY();
3460 	mtx_destroy(&mrouter_mtx);
3461 	break;
3462     default:
3463 	return EOPNOTSUPP;
3464     }
3465     return 0;
3466 }
3467 
3468 static moduledata_t ip_mroutemod = {
3469     "ip_mroute",
3470     ip_mroute_modevent,
3471     0
3472 };
3473 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
3474