xref: /netbsd-src/sys/netinet/ip_mroute.c (revision 23c8222edbfb0f0932d88a8351d3a0cf817dfb9e)
1 /*	$NetBSD: ip_mroute.c,v 1.86 2004/09/04 23:30:07 manu Exp $	*/
2 
3 /*
4  * Copyright (c) 1992, 1993
5  *      The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Stephen Deering of Stanford University.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
35  */
36 
37 /*
38  * Copyright (c) 1989 Stephen Deering
39  *
40  * This code is derived from software contributed to Berkeley by
41  * Stephen Deering of Stanford University.
42  *
43  * Redistribution and use in source and binary forms, with or without
44  * modification, are permitted provided that the following conditions
45  * are met:
46  * 1. Redistributions of source code must retain the above copyright
47  *    notice, this list of conditions and the following disclaimer.
48  * 2. Redistributions in binary form must reproduce the above copyright
49  *    notice, this list of conditions and the following disclaimer in the
50  *    documentation and/or other materials provided with the distribution.
51  * 3. All advertising materials mentioning features or use of this software
52  *    must display the following acknowledgement:
53  *      This product includes software developed by the University of
54  *      California, Berkeley and its contributors.
55  * 4. Neither the name of the University nor the names of its contributors
56  *    may be used to endorse or promote products derived from this software
57  *    without specific prior written permission.
58  *
59  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
60  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
61  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
62  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
63  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
64  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
65  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
66  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
67  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
68  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
69  * SUCH DAMAGE.
70  *
71  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
72  */
73 
74 /*
75  * IP multicast forwarding procedures
76  *
77  * Written by David Waitzman, BBN Labs, August 1988.
78  * Modified by Steve Deering, Stanford, February 1989.
79  * Modified by Mark J. Steiglitz, Stanford, May, 1991
80  * Modified by Van Jacobson, LBL, January 1993
81  * Modified by Ajit Thyagarajan, PARC, August 1993
82  * Modified by Bill Fenner, PARC, April 1994
83  * Modified by Charles M. Hannum, NetBSD, May 1995.
84  * Modified by Ahmed Helmy, SGI, June 1996
85  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
86  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
87  * Modified by Hitoshi Asaeda, WIDE, August 2000
88  * Modified by Pavlin Radoslavov, ICSI, October 2002
89  *
90  * MROUTING Revision: 1.2
91  * and PIM-SMv2 and PIM-DM support, advanced API support,
92  * bandwidth metering and signaling
93  */
94 
95 #include <sys/cdefs.h>
96 __KERNEL_RCSID(0, "$NetBSD: ip_mroute.c,v 1.86 2004/09/04 23:30:07 manu Exp $");
97 
98 #include "opt_inet.h"
99 #include "opt_ipsec.h"
100 #include "opt_pim.h"
101 
102 #ifdef PIM
103 #define _PIM_VT 1
104 #endif
105 
106 #include <sys/param.h>
107 #include <sys/systm.h>
108 #include <sys/callout.h>
109 #include <sys/mbuf.h>
110 #include <sys/socket.h>
111 #include <sys/socketvar.h>
112 #include <sys/protosw.h>
113 #include <sys/errno.h>
114 #include <sys/time.h>
115 #include <sys/kernel.h>
116 #include <sys/ioctl.h>
117 #include <sys/syslog.h>
118 
119 #include <net/if.h>
120 #include <net/route.h>
121 #include <net/raw_cb.h>
122 
123 #include <netinet/in.h>
124 #include <netinet/in_var.h>
125 #include <netinet/in_systm.h>
126 #include <netinet/ip.h>
127 #include <netinet/ip_var.h>
128 #include <netinet/in_pcb.h>
129 #include <netinet/udp.h>
130 #include <netinet/igmp.h>
131 #include <netinet/igmp_var.h>
132 #include <netinet/ip_mroute.h>
133 #ifdef PIM
134 #include <netinet/pim.h>
135 #include <netinet/pim_var.h>
136 #endif
137 #include <netinet/ip_encap.h>
138 
139 #ifdef IPSEC
140 #include <netinet6/ipsec.h>
141 #include <netkey/key.h>
142 #endif
143 
144 #ifdef FAST_IPSEC
145 #include <netipsec/ipsec.h>
146 #include <netipsec/key.h>
147 #endif
148 
149 #include <machine/stdarg.h>
150 
151 #define IP_MULTICASTOPTS 0
152 #define	M_PULLUP(m, len)						 \
153 	do {								 \
154 		if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
155 			(m) = m_pullup((m), (len));			 \
156 	} while (/*CONSTCOND*/ 0)
157 
158 /*
159  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
160  * except for netstat or debugging purposes.
161  */
162 struct socket  *ip_mrouter  = NULL;
163 int		ip_mrtproto = IGMP_DVMRP;    /* for netstat only */
164 
165 #define NO_RTE_FOUND 	0x1
166 #define RTE_FOUND	0x2
167 
168 #define	MFCHASH(a, g)							\
169 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^	\
170 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
171 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
172 u_long	mfchash;
173 
174 u_char		nexpire[MFCTBLSIZ];
175 struct vif	viftable[MAXVIFS];
176 struct mrtstat	mrtstat;
177 u_int		mrtdebug = 0;	  /* debug level 	*/
178 #define		DEBUG_MFC	0x02
179 #define		DEBUG_FORWARD	0x04
180 #define		DEBUG_EXPIRE	0x08
181 #define		DEBUG_XMIT	0x10
182 #define		DEBUG_PIM	0x20
183 
184 #define		VIFI_INVALID	((vifi_t) -1)
185 
186 u_int       	tbfdebug = 0;     /* tbf debug level 	*/
187 #ifdef RSVP_ISI
188 u_int		rsvpdebug = 0;	  /* rsvp debug level   */
189 extern struct socket *ip_rsvpd;
190 extern int rsvp_on;
191 #endif /* RSVP_ISI */
192 
193 /* vif attachment using sys/netinet/ip_encap.c */
194 static void vif_input __P((struct mbuf *, ...));
195 static int vif_encapcheck __P((const struct mbuf *, int, int, void *));
196 
197 static const struct protosw vif_protosw =
198 { SOCK_RAW,	&inetdomain,	IPPROTO_IPV4,	PR_ATOMIC|PR_ADDR,
199   vif_input,	rip_output,	0,		rip_ctloutput,
200   rip_usrreq,
201   0,            0,              0,              0,
202 };
203 
204 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
205 #define		UPCALL_EXPIRE	6		/* number of timeouts */
206 
207 /*
208  * Define the token bucket filter structures
209  */
210 
211 #define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
212 
213 static int get_sg_cnt __P((struct sioc_sg_req *));
214 static int get_vif_cnt __P((struct sioc_vif_req *));
215 static int ip_mrouter_init __P((struct socket *, struct mbuf *));
216 static int get_version __P((struct mbuf *));
217 static int set_assert __P((struct mbuf *));
218 static int get_assert __P((struct mbuf *));
219 static int add_vif __P((struct mbuf *));
220 static int del_vif __P((struct mbuf *));
221 static void update_mfc_params __P((struct mfc *, struct mfcctl2 *));
222 static void init_mfc_params __P((struct mfc *, struct mfcctl2 *));
223 static void expire_mfc __P((struct mfc *));
224 static int add_mfc __P((struct mbuf *));
225 #ifdef UPCALL_TIMING
226 static void collate __P((struct timeval *));
227 #endif
228 static int del_mfc __P((struct mbuf *));
229 static int set_api_config __P((struct mbuf *)); /* chose API capabilities */
230 static int get_api_support __P((struct mbuf *));
231 static int get_api_config __P((struct mbuf *));
232 static int socket_send __P((struct socket *, struct mbuf *,
233 			    struct sockaddr_in *));
234 static void expire_upcalls __P((void *));
235 #ifdef RSVP_ISI
236 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *, vifi_t));
237 #else
238 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *));
239 #endif
240 static void phyint_send __P((struct ip *, struct vif *, struct mbuf *));
241 static void encap_send __P((struct ip *, struct vif *, struct mbuf *));
242 static void tbf_control __P((struct vif *, struct mbuf *, struct ip *,
243 			     u_int32_t));
244 static void tbf_queue __P((struct vif *, struct mbuf *));
245 static void tbf_process_q __P((struct vif *));
246 static void tbf_reprocess_q __P((void *));
247 static int tbf_dq_sel __P((struct vif *, struct ip *));
248 static void tbf_send_packet __P((struct vif *, struct mbuf *));
249 static void tbf_update_tokens __P((struct vif *));
250 static int priority __P((struct vif *, struct ip *));
251 
252 /*
253  * Bandwidth monitoring
254  */
255 static void free_bw_list __P((struct bw_meter *));
256 static int add_bw_upcall __P((struct mbuf *));
257 static int del_bw_upcall __P((struct mbuf *));
258 static void bw_meter_receive_packet __P((struct bw_meter *, int , struct timeval *));
259 static void bw_meter_prepare_upcall __P((struct bw_meter *, struct timeval *));
260 static void bw_upcalls_send __P((void));
261 static void schedule_bw_meter __P((struct bw_meter *, struct timeval *));
262 static void unschedule_bw_meter __P((struct bw_meter *));
263 static void bw_meter_process __P((void));
264 static void expire_bw_upcalls_send __P((void *));
265 static void expire_bw_meter_process __P((void *));
266 
267 #ifdef PIM
268 static int pim_register_send __P((struct ip *, struct vif *,
269 		struct mbuf *, struct mfc *));
270 static int pim_register_send_rp __P((struct ip *, struct vif *,
271 		struct mbuf *, struct mfc *));
272 static int pim_register_send_upcall __P((struct ip *, struct vif *,
273 		struct mbuf *, struct mfc *));
274 static struct mbuf *pim_register_prepare __P((struct ip *, struct mbuf *));
275 #endif
276 
277 /*
278  * 'Interfaces' associated with decapsulator (so we can tell
279  * packets that went through it from ones that get reflected
280  * by a broken gateway).  These interfaces are never linked into
281  * the system ifnet list & no routes point to them.  I.e., packets
282  * can't be sent this way.  They only exist as a placeholder for
283  * multicast source verification.
284  */
285 #if 0
286 struct ifnet multicast_decap_if[MAXVIFS];
287 #endif
288 
289 #define	ENCAP_TTL	64
290 #define	ENCAP_PROTO	IPPROTO_IPIP	/* 4 */
291 
292 /* prototype IP hdr for encapsulated packets */
293 struct ip multicast_encap_iphdr = {
294 #if BYTE_ORDER == LITTLE_ENDIAN
295 	sizeof(struct ip) >> 2, IPVERSION,
296 #else
297 	IPVERSION, sizeof(struct ip) >> 2,
298 #endif
299 	0,				/* tos */
300 	sizeof(struct ip),		/* total length */
301 	0,				/* id */
302 	0,				/* frag offset */
303 	ENCAP_TTL, ENCAP_PROTO,
304 	0,				/* checksum */
305 };
306 
307 /*
308  * Bandwidth meter variables and constants
309  */
310 
311 /*
312  * Pending timeouts are stored in a hash table, the key being the
313  * expiration time. Periodically, the entries are analysed and processed.
314  */
315 #define BW_METER_BUCKETS	1024
316 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
317 struct callout bw_meter_ch;
318 #define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
319 
320 /*
321  * Pending upcalls are stored in a vector which is flushed when
322  * full, or periodically
323  */
324 static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
325 static u_int	bw_upcalls_n; /* # of pending upcalls */
326 struct callout	bw_upcalls_ch;
327 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
328 
329 #ifdef PIM
330 struct pimstat pimstat;
331 
332 /*
333  * Note: the PIM Register encapsulation adds the following in front of a
334  * data packet:
335  *
336  * struct pim_encap_hdr {
337  *    struct ip ip;
338  *    struct pim_encap_pimhdr  pim;
339  * }
340  *
341  */
342 
343 struct pim_encap_pimhdr {
344 	struct pim pim;
345 	uint32_t   flags;
346 };
347 
348 static struct ip pim_encap_iphdr = {
349 #if BYTE_ORDER == LITTLE_ENDIAN
350 	sizeof(struct ip) >> 2,
351 	IPVERSION,
352 #else
353 	IPVERSION,
354 	sizeof(struct ip) >> 2,
355 #endif
356 	0,			/* tos */
357 	sizeof(struct ip),	/* total length */
358 	0,			/* id */
359 	0,			/* frag offset */
360 	ENCAP_TTL,
361 	IPPROTO_PIM,
362 	0,			/* checksum */
363 };
364 
365 static struct pim_encap_pimhdr pim_encap_pimhdr = {
366     {
367 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
368 	0,			/* reserved */
369 	0,			/* checksum */
370     },
371     0				/* flags */
372 };
373 
374 static struct ifnet multicast_register_if;
375 static vifi_t reg_vif_num = VIFI_INVALID;
376 #endif /* PIM */
377 
378 
379 /*
380  * Private variables.
381  */
382 static vifi_t	   numvifs = 0;
383 
384 static struct callout expire_upcalls_ch;
385 
386 /*
387  * one-back cache used by vif_encapcheck to locate a tunnel's vif
388  * given a datagram's src ip address.
389  */
390 static struct in_addr last_encap_src;
391 static struct vif *last_encap_vif;
392 
393 /*
394  * whether or not special PIM assert processing is enabled.
395  */
396 static int pim_assert;
397 /*
398  * Rate limit for assert notification messages, in usec
399  */
400 #define ASSERT_MSG_TIME		3000000
401 
402 /*
403  * Kernel multicast routing API capabilities and setup.
404  * If more API capabilities are added to the kernel, they should be
405  * recorded in `mrt_api_support'.
406  */
407 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
408 					  MRT_MFC_FLAGS_BORDER_VIF |
409 					  MRT_MFC_RP |
410 					  MRT_MFC_BW_UPCALL);
411 static u_int32_t mrt_api_config = 0;
412 
413 /*
414  * Find a route for a given origin IP address and Multicast group address
415  * Type of service parameter to be added in the future!!!
416  * Statistics are updated by the caller if needed
417  * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
418  */
419 static struct mfc *
420 mfc_find(struct in_addr *o, struct in_addr *g)
421 {
422 	struct mfc *rt;
423 
424 	LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
425 		if (in_hosteq(rt->mfc_origin, *o) &&
426 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
427 		    (rt->mfc_stall == NULL))
428 			break;
429 	}
430 
431 	return (rt);
432 }
433 
434 /*
435  * Macros to compute elapsed time efficiently
436  * Borrowed from Van Jacobson's scheduling code
437  */
438 #define TV_DELTA(a, b, delta) do {					\
439 	int xxs;							\
440 	delta = (a).tv_usec - (b).tv_usec;				\
441 	xxs = (a).tv_sec - (b).tv_sec;					\
442 	switch (xxs) {							\
443 	case 2:								\
444 		delta += 1000000;					\
445 		/* fall through */					\
446 	case 1:								\
447 		delta += 1000000;					\
448 		/* fall through */					\
449 	case 0:								\
450 		break;							\
451 	default:							\
452 		delta += (1000000 * xxs);				\
453 		break;							\
454 	}								\
455 } while (/*CONSTCOND*/ 0)
456 
457 #ifdef UPCALL_TIMING
458 u_int32_t upcall_data[51];
459 #endif /* UPCALL_TIMING */
460 
461 /*
462  * Handle MRT setsockopt commands to modify the multicast routing tables.
463  */
464 int
465 ip_mrouter_set(so, optname, m)
466 	struct socket *so;
467 	int optname;
468 	struct mbuf **m;
469 {
470 	int error;
471 
472 	if (optname != MRT_INIT && so != ip_mrouter)
473 		error = ENOPROTOOPT;
474 	else
475 		switch (optname) {
476 		case MRT_INIT:
477 			error = ip_mrouter_init(so, *m);
478 			break;
479 		case MRT_DONE:
480 			error = ip_mrouter_done();
481 			break;
482 		case MRT_ADD_VIF:
483 			error = add_vif(*m);
484 			break;
485 		case MRT_DEL_VIF:
486 			error = del_vif(*m);
487 			break;
488 		case MRT_ADD_MFC:
489 			error = add_mfc(*m);
490 			break;
491 		case MRT_DEL_MFC:
492 			error = del_mfc(*m);
493 			break;
494 		case MRT_ASSERT:
495 			error = set_assert(*m);
496 			break;
497 		case MRT_API_CONFIG:
498 			error = set_api_config(*m);
499 			break;
500 		case MRT_ADD_BW_UPCALL:
501 			error = add_bw_upcall(*m);
502 			break;
503 		case MRT_DEL_BW_UPCALL:
504 			error = del_bw_upcall(*m);
505 			break;
506 		default:
507 			error = ENOPROTOOPT;
508 			break;
509 		}
510 
511 	if (*m)
512 		m_free(*m);
513 	return (error);
514 }
515 
516 /*
517  * Handle MRT getsockopt commands
518  */
519 int
520 ip_mrouter_get(so, optname, m)
521 	struct socket *so;
522 	int optname;
523 	struct mbuf **m;
524 {
525 	int error;
526 
527 	if (so != ip_mrouter)
528 		error = ENOPROTOOPT;
529 	else {
530 		*m = m_get(M_WAIT, MT_SOOPTS);
531 		MCLAIM(*m, so->so_mowner);
532 
533 		switch (optname) {
534 		case MRT_VERSION:
535 			error = get_version(*m);
536 			break;
537 		case MRT_ASSERT:
538 			error = get_assert(*m);
539 			break;
540 		case MRT_API_SUPPORT:
541 			error = get_api_support(*m);
542 			break;
543 		case MRT_API_CONFIG:
544 			error = get_api_config(*m);
545 			break;
546 		default:
547 			error = ENOPROTOOPT;
548 			break;
549 		}
550 
551 		if (error)
552 			m_free(*m);
553 	}
554 
555 	return (error);
556 }
557 
558 /*
559  * Handle ioctl commands to obtain information from the cache
560  */
561 int
562 mrt_ioctl(so, cmd, data)
563 	struct socket *so;
564 	u_long cmd;
565 	caddr_t data;
566 {
567 	int error;
568 
569 	if (so != ip_mrouter)
570 		error = EINVAL;
571 	else
572 		switch (cmd) {
573 		case SIOCGETVIFCNT:
574 			error = get_vif_cnt((struct sioc_vif_req *)data);
575 			break;
576 		case SIOCGETSGCNT:
577 			error = get_sg_cnt((struct sioc_sg_req *)data);
578 			break;
579 		default:
580 			error = EINVAL;
581 			break;
582 		}
583 
584 	return (error);
585 }
586 
587 /*
588  * returns the packet, byte, rpf-failure count for the source group provided
589  */
590 static int
591 get_sg_cnt(req)
592 	struct sioc_sg_req *req;
593 {
594 	int s;
595 	struct mfc *rt;
596 
597 	s = splsoftnet();
598 	rt = mfc_find(&req->src, &req->grp);
599 	if (rt == NULL) {
600 		splx(s);
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 	splx(s);
608 
609 	return (0);
610 }
611 
612 /*
613  * returns the input and output packet and byte counts on the vif provided
614  */
615 static int
616 get_vif_cnt(req)
617 	struct sioc_vif_req *req;
618 {
619 	vifi_t vifi = req->vifi;
620 
621 	if (vifi >= numvifs)
622 		return (EINVAL);
623 
624 	req->icount = viftable[vifi].v_pkt_in;
625 	req->ocount = viftable[vifi].v_pkt_out;
626 	req->ibytes = viftable[vifi].v_bytes_in;
627 	req->obytes = viftable[vifi].v_bytes_out;
628 
629 	return (0);
630 }
631 
632 /*
633  * Enable multicast routing
634  */
635 static int
636 ip_mrouter_init(so, m)
637 	struct socket *so;
638 	struct mbuf *m;
639 {
640 	int *v;
641 
642 	if (mrtdebug)
643 		log(LOG_DEBUG,
644 		    "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
645 		    so->so_type, so->so_proto->pr_protocol);
646 
647 	if (so->so_type != SOCK_RAW ||
648 	    so->so_proto->pr_protocol != IPPROTO_IGMP)
649 		return (EOPNOTSUPP);
650 
651 	if (m == NULL || m->m_len < sizeof(int))
652 		return (EINVAL);
653 
654 	v = mtod(m, int *);
655 	if (*v != 1)
656 		return (EINVAL);
657 
658 	if (ip_mrouter != NULL)
659 		return (EADDRINUSE);
660 
661 	ip_mrouter = so;
662 
663 	mfchashtbl =
664 	    hashinit(MFCTBLSIZ, HASH_LIST, M_MRTABLE, M_WAITOK, &mfchash);
665 	bzero((caddr_t)nexpire, sizeof(nexpire));
666 
667 	pim_assert = 0;
668 
669 	callout_init(&expire_upcalls_ch);
670 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
671 		      expire_upcalls, NULL);
672 
673 	callout_init(&bw_upcalls_ch);
674 	callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
675 		      expire_bw_upcalls_send, NULL);
676 
677 	callout_init(&bw_meter_ch);
678 	callout_reset(&bw_meter_ch, BW_METER_PERIOD,
679 		      expire_bw_meter_process, NULL);
680 
681 	if (mrtdebug)
682 		log(LOG_DEBUG, "ip_mrouter_init\n");
683 
684 	return (0);
685 }
686 
687 /*
688  * Disable multicast routing
689  */
690 int
691 ip_mrouter_done()
692 {
693 	vifi_t vifi;
694 	struct vif *vifp;
695 	int i;
696 	int s;
697 
698 	s = splsoftnet();
699 
700 	/* Clear out all the vifs currently in use. */
701 	for (vifi = 0; vifi < numvifs; vifi++) {
702 		vifp = &viftable[vifi];
703 		if (!in_nullhost(vifp->v_lcl_addr))
704 			reset_vif(vifp);
705 	}
706 
707 	numvifs = 0;
708 	pim_assert = 0;
709 	mrt_api_config = 0;
710 
711 	callout_stop(&expire_upcalls_ch);
712 	callout_stop(&bw_upcalls_ch);
713 	callout_stop(&bw_meter_ch);
714 
715 	/*
716 	 * Free all multicast forwarding cache entries.
717 	 */
718 	for (i = 0; i < MFCTBLSIZ; i++) {
719 		struct mfc *rt, *nrt;
720 
721 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
722 			nrt = LIST_NEXT(rt, mfc_hash);
723 
724 			expire_mfc(rt);
725 		}
726 	}
727 
728 	bzero((caddr_t)nexpire, sizeof(nexpire));
729 	free(mfchashtbl, M_MRTABLE);
730 	mfchashtbl = NULL;
731 
732 	bw_upcalls_n = 0;
733 	bzero(bw_meter_timers, sizeof(bw_meter_timers));
734 
735 	/* Reset de-encapsulation cache. */
736 
737 	ip_mrouter = NULL;
738 
739 	splx(s);
740 
741 	if (mrtdebug)
742 		log(LOG_DEBUG, "ip_mrouter_done\n");
743 
744 	return (0);
745 }
746 
747 void
748 ip_mrouter_detach(ifp)
749 	struct ifnet *ifp;
750 {
751 	int vifi, i;
752 	struct vif *vifp;
753 	struct mfc *rt;
754 	struct rtdetq *rte;
755 
756 	/* XXX not sure about side effect to userland routing daemon */
757 	for (vifi = 0; vifi < numvifs; vifi++) {
758 		vifp = &viftable[vifi];
759 		if (vifp->v_ifp == ifp)
760 			reset_vif(vifp);
761 	}
762 	for (i = 0; i < MFCTBLSIZ; i++) {
763 		if (nexpire[i] == 0)
764 			continue;
765 		LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
766 			for (rte = rt->mfc_stall; rte; rte = rte->next) {
767 				if (rte->ifp == ifp)
768 					rte->ifp = NULL;
769 			}
770 		}
771 	}
772 }
773 
774 static int
775 get_version(m)
776 	struct mbuf *m;
777 {
778 	int *v = mtod(m, int *);
779 
780 	*v = 0x0305;	/* XXX !!!! */
781 	m->m_len = sizeof(int);
782 	return (0);
783 }
784 
785 /*
786  * Set PIM assert processing global
787  */
788 static int
789 set_assert(m)
790 	struct mbuf *m;
791 {
792 	int *i;
793 
794 	if (m == NULL || m->m_len < sizeof(int))
795 		return (EINVAL);
796 
797 	i = mtod(m, int *);
798 	pim_assert = !!*i;
799 	return (0);
800 }
801 
802 /*
803  * Get PIM assert processing global
804  */
805 static int
806 get_assert(m)
807 	struct mbuf *m;
808 {
809 	int *i = mtod(m, int *);
810 
811 	*i = pim_assert;
812 	m->m_len = sizeof(int);
813 	return (0);
814 }
815 
816 /*
817  * Configure API capabilities
818  */
819 static int
820 set_api_config(struct mbuf *m)
821 {
822 	int i;
823 	u_int32_t *apival;
824 
825 	if (m == NULL || m->m_len < sizeof(u_int32_t))
826 		return (EINVAL);
827 
828 	apival = mtod(m, u_int32_t *);
829 
830 	/*
831 	 * We can set the API capabilities only if it is the first operation
832 	 * after MRT_INIT. I.e.:
833 	 *  - there are no vifs installed
834 	 *  - pim_assert is not enabled
835 	 *  - the MFC table is empty
836 	 */
837 	if (numvifs > 0) {
838 		*apival = 0;
839 		return (EPERM);
840 	}
841 	if (pim_assert) {
842 		*apival = 0;
843 		return (EPERM);
844 	}
845 	for (i = 0; i < MFCTBLSIZ; i++) {
846 		if (LIST_FIRST(&mfchashtbl[i]) != NULL) {
847 			*apival = 0;
848 			return (EPERM);
849 		}
850 	}
851 
852 	mrt_api_config = *apival & mrt_api_support;
853 	*apival = mrt_api_config;
854 
855 	return (0);
856 }
857 
858 /*
859  * Get API capabilities
860  */
861 static int
862 get_api_support(struct mbuf *m)
863 {
864 	u_int32_t *apival;
865 
866 	if (m == NULL || m->m_len < sizeof(u_int32_t))
867 		return (EINVAL);
868 
869 	apival = mtod(m, u_int32_t *);
870 
871 	*apival = mrt_api_support;
872 
873 	return (0);
874 }
875 
876 /*
877  * Get API configured capabilities
878  */
879 static int
880 get_api_config(struct mbuf *m)
881 {
882 	u_int32_t *apival;
883 
884 	if (m == NULL || m->m_len < sizeof(u_int32_t))
885 		return (EINVAL);
886 
887 	apival = mtod(m, u_int32_t *);
888 
889 	*apival = mrt_api_config;
890 
891 	return (0);
892 }
893 
894 static struct sockaddr_in sin = { sizeof(sin), AF_INET };
895 
896 /*
897  * Add a vif to the vif table
898  */
899 static int
900 add_vif(m)
901 	struct mbuf *m;
902 {
903 	struct vifctl *vifcp;
904 	struct vif *vifp;
905 	struct ifaddr *ifa;
906 	struct ifnet *ifp;
907 	struct ifreq ifr;
908 	int error, s;
909 
910 	if (m == NULL || m->m_len < sizeof(struct vifctl))
911 		return (EINVAL);
912 
913 	vifcp = mtod(m, struct vifctl *);
914 	if (vifcp->vifc_vifi >= MAXVIFS)
915 		return (EINVAL);
916 	if (in_nullhost(vifcp->vifc_lcl_addr))
917 		return (EADDRNOTAVAIL);
918 
919 	vifp = &viftable[vifcp->vifc_vifi];
920 	if (!in_nullhost(vifp->v_lcl_addr))
921 		return (EADDRINUSE);
922 
923 	/* Find the interface with an address in AF_INET family. */
924 #ifdef PIM
925 	if (vifcp->vifc_flags & VIFF_REGISTER) {
926 		/*
927 		 * XXX: Because VIFF_REGISTER does not really need a valid
928 		 * local interface (e.g. it could be 127.0.0.2), we don't
929 		 * check its address.
930 		 */
931 	    ifp = NULL;
932 	} else
933 #endif
934 	{
935 		sin.sin_addr = vifcp->vifc_lcl_addr;
936 		ifa = ifa_ifwithaddr(sintosa(&sin));
937 		if (ifa == NULL)
938 			return (EADDRNOTAVAIL);
939 		ifp = ifa->ifa_ifp;
940 	}
941 
942 	if (vifcp->vifc_flags & VIFF_TUNNEL) {
943 		if (vifcp->vifc_flags & VIFF_SRCRT) {
944 			log(LOG_ERR, "source routed tunnels not supported\n");
945 			return (EOPNOTSUPP);
946 		}
947 
948 		/* attach this vif to decapsulator dispatch table */
949 		vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
950 		    vif_encapcheck, &vif_protosw, vifp);
951 		if (!vifp->v_encap_cookie)
952 			return (EINVAL);
953 
954 		/* Create a fake encapsulation interface. */
955 		ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK);
956 		bzero(ifp, sizeof(*ifp));
957 		snprintf(ifp->if_xname, sizeof(ifp->if_xname),
958 			 "mdecap%d", vifcp->vifc_vifi);
959 
960 		/* Prepare cached route entry. */
961 		bzero(&vifp->v_route, sizeof(vifp->v_route));
962 #ifdef PIM
963 	} else if (vifcp->vifc_flags & VIFF_REGISTER) {
964 		ifp = &multicast_register_if;
965 		if (mrtdebug)
966 			log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
967 			    (void *)ifp);
968 		if (reg_vif_num == VIFI_INVALID) {
969 			bzero(ifp, sizeof(*ifp));
970 			snprintf(ifp->if_xname, sizeof(ifp->if_xname),
971 				 "register_vif");
972 			ifp->if_flags = IFF_LOOPBACK;
973 			bzero(&vifp->v_route, sizeof(vifp->v_route));
974 			reg_vif_num = vifcp->vifc_vifi;
975 		}
976 #endif
977 	} else {
978 		/* Make sure the interface supports multicast. */
979 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
980 			return (EOPNOTSUPP);
981 
982 		/* Enable promiscuous reception of all IP multicasts. */
983 		satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
984 		satosin(&ifr.ifr_addr)->sin_family = AF_INET;
985 		satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
986 		error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
987 		if (error)
988 			return (error);
989 	}
990 
991 	s = splsoftnet();
992 
993 	/* Define parameters for the tbf structure. */
994 	vifp->tbf_q = NULL;
995 	vifp->tbf_t = &vifp->tbf_q;
996 	microtime(&vifp->tbf_last_pkt_t);
997 	vifp->tbf_n_tok = 0;
998 	vifp->tbf_q_len = 0;
999 	vifp->tbf_max_q_len = MAXQSIZE;
1000 
1001 	vifp->v_flags = vifcp->vifc_flags;
1002 	vifp->v_threshold = vifcp->vifc_threshold;
1003 	/* scaling up here allows division by 1024 in critical code */
1004 	vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
1005 	vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
1006 	vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
1007 	vifp->v_ifp = ifp;
1008 	/* Initialize per vif pkt counters. */
1009 	vifp->v_pkt_in = 0;
1010 	vifp->v_pkt_out = 0;
1011 	vifp->v_bytes_in = 0;
1012 	vifp->v_bytes_out = 0;
1013 
1014 	callout_init(&vifp->v_repq_ch);
1015 
1016 #ifdef RSVP_ISI
1017 	vifp->v_rsvp_on = 0;
1018 	vifp->v_rsvpd = NULL;
1019 #endif /* RSVP_ISI */
1020 
1021 	splx(s);
1022 
1023 	/* Adjust numvifs up if the vifi is higher than numvifs. */
1024 	if (numvifs <= vifcp->vifc_vifi)
1025 		numvifs = vifcp->vifc_vifi + 1;
1026 
1027 	if (mrtdebug)
1028 		log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
1029 		    vifcp->vifc_vifi,
1030 		    ntohl(vifcp->vifc_lcl_addr.s_addr),
1031 		    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
1032 		    ntohl(vifcp->vifc_rmt_addr.s_addr),
1033 		    vifcp->vifc_threshold,
1034 		    vifcp->vifc_rate_limit);
1035 
1036 	return (0);
1037 }
1038 
1039 void
1040 reset_vif(vifp)
1041 	struct vif *vifp;
1042 {
1043 	struct mbuf *m, *n;
1044 	struct ifnet *ifp;
1045 	struct ifreq ifr;
1046 
1047 	callout_stop(&vifp->v_repq_ch);
1048 
1049 	/* detach this vif from decapsulator dispatch table */
1050 	encap_detach(vifp->v_encap_cookie);
1051 	vifp->v_encap_cookie = NULL;
1052 
1053 	/*
1054 	 * Free packets queued at the interface
1055 	 */
1056 	for (m = vifp->tbf_q; m != NULL; m = n) {
1057 		n = m->m_nextpkt;
1058 		m_freem(m);
1059 	}
1060 
1061 	if (vifp->v_flags & VIFF_TUNNEL) {
1062 		free(vifp->v_ifp, M_MRTABLE);
1063 		if (vifp == last_encap_vif) {
1064 			last_encap_vif = NULL;
1065 			last_encap_src = zeroin_addr;
1066 		}
1067 	} else if (vifp->v_flags & VIFF_REGISTER) {
1068 #ifdef PIM
1069 		if (vifp->v_flags & VIFF_REGISTER)
1070 			reg_vif_num = VIFI_INVALID;
1071 #endif
1072 	} else {
1073 		satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
1074 		satosin(&ifr.ifr_addr)->sin_family = AF_INET;
1075 		satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
1076 		ifp = vifp->v_ifp;
1077 		(*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
1078 	}
1079 	bzero((caddr_t)vifp, sizeof(*vifp));
1080 }
1081 
1082 /*
1083  * Delete a vif from the vif table
1084  */
1085 static int
1086 del_vif(m)
1087 	struct mbuf *m;
1088 {
1089 	vifi_t *vifip;
1090 	struct vif *vifp;
1091 	vifi_t vifi;
1092 	int s;
1093 
1094 	if (m == NULL || m->m_len < sizeof(vifi_t))
1095 		return (EINVAL);
1096 
1097 	vifip = mtod(m, vifi_t *);
1098 	if (*vifip >= numvifs)
1099 		return (EINVAL);
1100 
1101 	vifp = &viftable[*vifip];
1102 	if (in_nullhost(vifp->v_lcl_addr))
1103 		return (EADDRNOTAVAIL);
1104 
1105 	s = splsoftnet();
1106 
1107 	reset_vif(vifp);
1108 
1109 	/* Adjust numvifs down */
1110 	for (vifi = numvifs; vifi > 0; vifi--)
1111 		if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
1112 			break;
1113 	numvifs = vifi;
1114 
1115 	splx(s);
1116 
1117 	if (mrtdebug)
1118 		log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
1119 
1120 	return (0);
1121 }
1122 
1123 /*
1124  * update an mfc entry without resetting counters and S,G addresses.
1125  */
1126 static void
1127 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1128 {
1129 	int i;
1130 
1131 	rt->mfc_parent = mfccp->mfcc_parent;
1132 	for (i = 0; i < numvifs; i++) {
1133 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1134 		rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1135 			MRT_MFC_FLAGS_ALL;
1136 	}
1137 	/* set the RP address */
1138 	if (mrt_api_config & MRT_MFC_RP)
1139 		rt->mfc_rp = mfccp->mfcc_rp;
1140 	else
1141 		rt->mfc_rp = zeroin_addr;
1142 }
1143 
1144 /*
1145  * fully initialize an mfc entry from the parameter.
1146  */
1147 static void
1148 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1149 {
1150 	rt->mfc_origin     = mfccp->mfcc_origin;
1151 	rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1152 
1153 	update_mfc_params(rt, mfccp);
1154 
1155 	/* initialize pkt counters per src-grp */
1156 	rt->mfc_pkt_cnt    = 0;
1157 	rt->mfc_byte_cnt   = 0;
1158 	rt->mfc_wrong_if   = 0;
1159 	timerclear(&rt->mfc_last_assert);
1160 }
1161 
1162 static void
1163 expire_mfc(rt)
1164 	struct mfc *rt;
1165 {
1166 	struct rtdetq *rte, *nrte;
1167 
1168 	free_bw_list(rt->mfc_bw_meter);
1169 
1170 	for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1171 		nrte = rte->next;
1172 		m_freem(rte->m);
1173 		free(rte, M_MRTABLE);
1174 	}
1175 
1176 	LIST_REMOVE(rt, mfc_hash);
1177 	free(rt, M_MRTABLE);
1178 }
1179 
1180 /*
1181  * Add an mfc entry
1182  */
1183 static int
1184 add_mfc(m)
1185 	struct mbuf *m;
1186 {
1187 	struct mfcctl2 mfcctl2;
1188 	struct mfcctl2 *mfccp;
1189 	struct mfc *rt;
1190 	u_int32_t hash = 0;
1191 	struct rtdetq *rte, *nrte;
1192 	u_short nstl;
1193 	int s;
1194 	int mfcctl_size = sizeof(struct mfcctl);
1195 
1196 	if (mrt_api_config & MRT_API_FLAGS_ALL)
1197 		mfcctl_size = sizeof(struct mfcctl2);
1198 
1199 	if (m == NULL || m->m_len < mfcctl_size)
1200 		return (EINVAL);
1201 
1202 	/*
1203 	 * select data size depending on API version.
1204 	 */
1205 	if (mrt_api_config & MRT_API_FLAGS_ALL) {
1206 		struct mfcctl2 *mp2 = mtod(m, struct mfcctl2 *);
1207 		bcopy(mp2, (caddr_t)&mfcctl2, sizeof(*mp2));
1208 	} else {
1209 		struct mfcctl *mp = mtod(m, struct mfcctl *);
1210 		bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp));
1211 		bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl),
1212 		      sizeof(mfcctl2) - sizeof(struct mfcctl));
1213 	}
1214 	mfccp = &mfcctl2;
1215 
1216 	s = splsoftnet();
1217 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1218 
1219 	/* If an entry already exists, just update the fields */
1220 	if (rt) {
1221 		if (mrtdebug & DEBUG_MFC)
1222 			log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1223 			    ntohl(mfccp->mfcc_origin.s_addr),
1224 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1225 			    mfccp->mfcc_parent);
1226 
1227 		update_mfc_params(rt, mfccp);
1228 
1229 		splx(s);
1230 		return (0);
1231 	}
1232 
1233 	/*
1234 	 * Find the entry for which the upcall was made and update
1235 	 */
1236 	nstl = 0;
1237 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1238 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1239 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1240 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1241 		    rt->mfc_stall != NULL) {
1242 			if (nstl++)
1243 				log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
1244 				    "multiple kernel entries",
1245 				    ntohl(mfccp->mfcc_origin.s_addr),
1246 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1247 				    mfccp->mfcc_parent, rt->mfc_stall);
1248 
1249 			if (mrtdebug & DEBUG_MFC)
1250 				log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n",
1251 				    ntohl(mfccp->mfcc_origin.s_addr),
1252 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1253 				    mfccp->mfcc_parent, rt->mfc_stall);
1254 
1255 			rte = rt->mfc_stall;
1256 			init_mfc_params(rt, mfccp);
1257 			rt->mfc_stall = NULL;
1258 
1259 			rt->mfc_expire = 0; /* Don't clean this guy up */
1260 			nexpire[hash]--;
1261 
1262 			/* free packets Qed at the end of this entry */
1263 			for (; rte != NULL; rte = nrte) {
1264 				nrte = rte->next;
1265 				if (rte->ifp) {
1266 #ifdef RSVP_ISI
1267 					ip_mdq(rte->m, rte->ifp, rt, -1);
1268 #else
1269 					ip_mdq(rte->m, rte->ifp, rt);
1270 #endif /* RSVP_ISI */
1271 				}
1272 				m_freem(rte->m);
1273 #ifdef UPCALL_TIMING
1274 				collate(&rte->t);
1275 #endif /* UPCALL_TIMING */
1276 				free(rte, M_MRTABLE);
1277 			}
1278 		}
1279 	}
1280 
1281 	/*
1282 	 * It is possible that an entry is being inserted without an upcall
1283 	 */
1284 	if (nstl == 0) {
1285 		/*
1286 		 * No mfc; make a new one
1287 		 */
1288 		if (mrtdebug & DEBUG_MFC)
1289 			log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1290 			    ntohl(mfccp->mfcc_origin.s_addr),
1291 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1292 			    mfccp->mfcc_parent);
1293 
1294 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1295 			if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1296 			    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1297 				init_mfc_params(rt, mfccp);
1298 				if (rt->mfc_expire)
1299 					nexpire[hash]--;
1300 				rt->mfc_expire = 0;
1301 				break; /* XXX */
1302 			}
1303 		}
1304 		if (rt == NULL) {	/* no upcall, so make a new entry */
1305 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1306 						  M_NOWAIT);
1307 			if (rt == NULL) {
1308 				splx(s);
1309 				return (ENOBUFS);
1310 			}
1311 
1312 			init_mfc_params(rt, mfccp);
1313 			rt->mfc_expire	= 0;
1314 			rt->mfc_stall	= NULL;
1315 			rt->mfc_bw_meter = NULL;
1316 
1317 			/* insert new entry at head of hash chain */
1318 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1319 		}
1320 	}
1321 
1322 	splx(s);
1323 	return (0);
1324 }
1325 
1326 #ifdef UPCALL_TIMING
1327 /*
1328  * collect delay statistics on the upcalls
1329  */
1330 static void
1331 collate(t)
1332 	struct timeval *t;
1333 {
1334 	u_int32_t d;
1335 	struct timeval tp;
1336 	u_int32_t delta;
1337 
1338 	microtime(&tp);
1339 
1340 	if (timercmp(t, &tp, <)) {
1341 		TV_DELTA(tp, *t, delta);
1342 
1343 		d = delta >> 10;
1344 		if (d > 50)
1345 			d = 50;
1346 
1347 		++upcall_data[d];
1348 	}
1349 }
1350 #endif /* UPCALL_TIMING */
1351 
1352 /*
1353  * Delete an mfc entry
1354  */
1355 static int
1356 del_mfc(m)
1357 	struct mbuf *m;
1358 {
1359 	struct mfcctl2 mfcctl2;
1360 	struct mfcctl2 *mfccp;
1361 	struct mfc *rt;
1362 	int s;
1363 	int mfcctl_size = sizeof(struct mfcctl);
1364 	struct mfcctl *mp = mtod(m, struct mfcctl *);
1365 
1366 	/*
1367 	 * XXX: for deleting MFC entries the information in entries
1368 	 * of size "struct mfcctl" is sufficient.
1369 	 */
1370 
1371 	if (m == NULL || m->m_len < mfcctl_size)
1372 		return (EINVAL);
1373 
1374 	bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp));
1375 	bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl),
1376 	      sizeof(mfcctl2) - sizeof(struct mfcctl));
1377 
1378 	mfccp = &mfcctl2;
1379 
1380 	if (mrtdebug & DEBUG_MFC)
1381 		log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1382 		    ntohl(mfccp->mfcc_origin.s_addr),
1383 		    ntohl(mfccp->mfcc_mcastgrp.s_addr));
1384 
1385 	s = splsoftnet();
1386 
1387 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1388 	if (rt == NULL) {
1389 		splx(s);
1390 		return (EADDRNOTAVAIL);
1391 	}
1392 
1393 	/*
1394 	 * free the bw_meter entries
1395 	 */
1396 	free_bw_list(rt->mfc_bw_meter);
1397 	rt->mfc_bw_meter = NULL;
1398 
1399 	LIST_REMOVE(rt, mfc_hash);
1400 	free(rt, M_MRTABLE);
1401 
1402 	splx(s);
1403 	return (0);
1404 }
1405 
1406 static int
1407 socket_send(s, mm, src)
1408 	struct socket *s;
1409 	struct mbuf *mm;
1410 	struct sockaddr_in *src;
1411 {
1412 	if (s) {
1413 		if (sbappendaddr(&s->so_rcv, sintosa(src), mm,
1414 		    (struct mbuf *)NULL) != 0) {
1415 			sorwakeup(s);
1416 			return (0);
1417 		}
1418 	}
1419 	m_freem(mm);
1420 	return (-1);
1421 }
1422 
1423 /*
1424  * IP multicast forwarding function. This function assumes that the packet
1425  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1426  * pointed to by "ifp", and the packet is to be relayed to other networks
1427  * that have members of the packet's destination IP multicast group.
1428  *
1429  * The packet is returned unscathed to the caller, unless it is
1430  * erroneous, in which case a non-zero return value tells the caller to
1431  * discard it.
1432  */
1433 
1434 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1435 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1436 
1437 int
1438 #ifdef RSVP_ISI
1439 ip_mforward(m, ifp, imo)
1440 #else
1441 ip_mforward(m, ifp)
1442 #endif /* RSVP_ISI */
1443 	struct mbuf *m;
1444 	struct ifnet *ifp;
1445 #ifdef RSVP_ISI
1446 	struct ip_moptions *imo;
1447 #endif /* RSVP_ISI */
1448 {
1449 	struct ip *ip = mtod(m, struct ip *);
1450 	struct mfc *rt;
1451 	static int srctun = 0;
1452 	struct mbuf *mm;
1453 	int s;
1454 	vifi_t vifi;
1455 
1456 	if (mrtdebug & DEBUG_FORWARD)
1457 		log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1458 		    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1459 
1460 	if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1461 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1462 		/*
1463 		 * Packet arrived via a physical interface or
1464 		 * an encapsulated tunnel or a register_vif.
1465 		 */
1466 	} else {
1467 		/*
1468 		 * Packet arrived through a source-route tunnel.
1469 		 * Source-route tunnels are no longer supported.
1470 		 */
1471 		if ((srctun++ % 1000) == 0)
1472 			log(LOG_ERR,
1473 			    "ip_mforward: received source-routed packet from %x\n",
1474 			    ntohl(ip->ip_src.s_addr));
1475 
1476 		return (1);
1477 	}
1478 
1479 #ifdef RSVP_ISI
1480 	if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1481 		if (ip->ip_ttl < 255)
1482 			ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1483 		if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1484 			struct vif *vifp = viftable + vifi;
1485 			printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n",
1486 			    ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1487 			    (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1488 			    vifp->v_ifp->if_xname);
1489 		}
1490 		return (ip_mdq(m, ifp, (struct mfc *)NULL, vifi));
1491 	}
1492 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1493 		printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
1494 		    ntohl(ip->ip_src), ntohl(ip->ip_dst));
1495 	}
1496 #endif /* RSVP_ISI */
1497 
1498 	/*
1499 	 * Don't forward a packet with time-to-live of zero or one,
1500 	 * or a packet destined to a local-only group.
1501 	 */
1502 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1503 		return (0);
1504 
1505 	/*
1506 	 * Determine forwarding vifs from the forwarding cache table
1507 	 */
1508 	s = splsoftnet();
1509 	++mrtstat.mrts_mfc_lookups;
1510 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1511 
1512 	/* Entry exists, so forward if necessary */
1513 	if (rt != NULL) {
1514 		splx(s);
1515 #ifdef RSVP_ISI
1516 		return (ip_mdq(m, ifp, rt, -1));
1517 #else
1518 		return (ip_mdq(m, ifp, rt));
1519 #endif /* RSVP_ISI */
1520 	} else {
1521 		/*
1522 		 * If we don't have a route for packet's origin,
1523 		 * Make a copy of the packet & send message to routing daemon
1524 		 */
1525 
1526 		struct mbuf *mb0;
1527 		struct rtdetq *rte;
1528 		u_int32_t hash;
1529 		int hlen = ip->ip_hl << 2;
1530 #ifdef UPCALL_TIMING
1531 		struct timeval tp;
1532 
1533 		microtime(&tp);
1534 #endif /* UPCALL_TIMING */
1535 
1536 		++mrtstat.mrts_mfc_misses;
1537 
1538 		mrtstat.mrts_no_route++;
1539 		if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1540 			log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1541 			    ntohl(ip->ip_src.s_addr),
1542 			    ntohl(ip->ip_dst.s_addr));
1543 
1544 		/*
1545 		 * Allocate mbufs early so that we don't do extra work if we are
1546 		 * just going to fail anyway.  Make sure to pullup the header so
1547 		 * that other people can't step on it.
1548 		 */
1549 		rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1550 					      M_NOWAIT);
1551 		if (rte == NULL) {
1552 			splx(s);
1553 			return (ENOBUFS);
1554 		}
1555 		mb0 = m_copy(m, 0, M_COPYALL);
1556 		M_PULLUP(mb0, hlen);
1557 		if (mb0 == NULL) {
1558 			free(rte, M_MRTABLE);
1559 			splx(s);
1560 			return (ENOBUFS);
1561 		}
1562 
1563 		/* is there an upcall waiting for this flow? */
1564 		hash = MFCHASH(ip->ip_src, ip->ip_dst);
1565 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1566 			if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1567 			    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1568 			    rt->mfc_stall != NULL)
1569 				break;
1570 		}
1571 
1572 		if (rt == NULL) {
1573 			int i;
1574 			struct igmpmsg *im;
1575 
1576 			/*
1577 			 * Locate the vifi for the incoming interface for
1578 			 * this packet.
1579 			 * If none found, drop packet.
1580 			 */
1581 			for (vifi = 0; vifi < numvifs &&
1582 				 viftable[vifi].v_ifp != ifp; vifi++)
1583 				;
1584 			if (vifi >= numvifs) /* vif not found, drop packet */
1585 				goto non_fatal;
1586 
1587 			/* no upcall, so make a new entry */
1588 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1589 						  M_NOWAIT);
1590 			if (rt == NULL)
1591 				goto fail;
1592 
1593 			/*
1594 			 * Make a copy of the header to send to the user level
1595 			 * process
1596 			 */
1597 			mm = m_copy(m, 0, hlen);
1598 			M_PULLUP(mm, hlen);
1599 			if (mm == NULL)
1600 				goto fail1;
1601 
1602 			/*
1603 			 * Send message to routing daemon to install
1604 			 * a route into the kernel table
1605 			 */
1606 
1607 			im = mtod(mm, struct igmpmsg *);
1608 			im->im_msgtype = IGMPMSG_NOCACHE;
1609 			im->im_mbz = 0;
1610 			im->im_vif = vifi;
1611 
1612 			mrtstat.mrts_upcalls++;
1613 
1614 			sin.sin_addr = ip->ip_src;
1615 			if (socket_send(ip_mrouter, mm, &sin) < 0) {
1616 				log(LOG_WARNING,
1617 				    "ip_mforward: ip_mrouter socket queue full\n");
1618 				++mrtstat.mrts_upq_sockfull;
1619 			fail1:
1620 				free(rt, M_MRTABLE);
1621 			fail:
1622 				free(rte, M_MRTABLE);
1623 				m_freem(mb0);
1624 				splx(s);
1625 				return (ENOBUFS);
1626 			}
1627 
1628 			/* insert new entry at head of hash chain */
1629 			rt->mfc_origin = ip->ip_src;
1630 			rt->mfc_mcastgrp = ip->ip_dst;
1631 			rt->mfc_pkt_cnt = 0;
1632 			rt->mfc_byte_cnt = 0;
1633 			rt->mfc_wrong_if = 0;
1634 			rt->mfc_expire = UPCALL_EXPIRE;
1635 			nexpire[hash]++;
1636 			for (i = 0; i < numvifs; i++) {
1637 				rt->mfc_ttls[i] = 0;
1638 				rt->mfc_flags[i] = 0;
1639 			}
1640 			rt->mfc_parent = -1;
1641 
1642 			/* clear the RP address */
1643 			rt->mfc_rp = zeroin_addr;
1644 
1645 			rt->mfc_bw_meter = NULL;
1646 
1647 			/* link into table */
1648 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1649 			/* Add this entry to the end of the queue */
1650 			rt->mfc_stall = rte;
1651 		} else {
1652 			/* determine if q has overflowed */
1653 			struct rtdetq **p;
1654 			int npkts = 0;
1655 
1656 			/*
1657 			 * XXX ouch! we need to append to the list, but we
1658 			 * only have a pointer to the front, so we have to
1659 			 * scan the entire list every time.
1660 			 */
1661 			for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1662 				if (++npkts > MAX_UPQ) {
1663 					mrtstat.mrts_upq_ovflw++;
1664 				non_fatal:
1665 					free(rte, M_MRTABLE);
1666 					m_freem(mb0);
1667 					splx(s);
1668 					return (0);
1669 				}
1670 
1671 			/* Add this entry to the end of the queue */
1672 			*p = rte;
1673 		}
1674 
1675 		rte->next = NULL;
1676 		rte->m = mb0;
1677 		rte->ifp = ifp;
1678 #ifdef UPCALL_TIMING
1679 		rte->t = tp;
1680 #endif /* UPCALL_TIMING */
1681 
1682 		splx(s);
1683 
1684 		return (0);
1685 	}
1686 }
1687 
1688 
1689 /*ARGSUSED*/
1690 static void
1691 expire_upcalls(v)
1692 	void *v;
1693 {
1694 	int i;
1695 	int s;
1696 
1697 	s = splsoftnet();
1698 
1699 	for (i = 0; i < MFCTBLSIZ; i++) {
1700 		struct mfc *rt, *nrt;
1701 
1702 		if (nexpire[i] == 0)
1703 			continue;
1704 
1705 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1706 			nrt = LIST_NEXT(rt, mfc_hash);
1707 
1708 			if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1709 				continue;
1710 			nexpire[i]--;
1711 
1712 			/*
1713 			 * free the bw_meter entries
1714 			 */
1715 			while (rt->mfc_bw_meter != NULL) {
1716 				struct bw_meter *x = rt->mfc_bw_meter;
1717 
1718 				rt->mfc_bw_meter = x->bm_mfc_next;
1719 				free(x, M_BWMETER);
1720 			}
1721 
1722 			++mrtstat.mrts_cache_cleanups;
1723 			if (mrtdebug & DEBUG_EXPIRE)
1724 				log(LOG_DEBUG,
1725 				    "expire_upcalls: expiring (%x %x)\n",
1726 				    ntohl(rt->mfc_origin.s_addr),
1727 				    ntohl(rt->mfc_mcastgrp.s_addr));
1728 
1729 			expire_mfc(rt);
1730 		}
1731 	}
1732 
1733 	splx(s);
1734 	callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1735 	    expire_upcalls, NULL);
1736 }
1737 
1738 /*
1739  * Packet forwarding routine once entry in the cache is made
1740  */
1741 static int
1742 #ifdef RSVP_ISI
1743 ip_mdq(m, ifp, rt, xmt_vif)
1744 #else
1745 ip_mdq(m, ifp, rt)
1746 #endif /* RSVP_ISI */
1747 	struct mbuf *m;
1748 	struct ifnet *ifp;
1749 	struct mfc *rt;
1750 #ifdef RSVP_ISI
1751 	vifi_t xmt_vif;
1752 #endif /* RSVP_ISI */
1753 {
1754 	struct ip  *ip = mtod(m, struct ip *);
1755 	vifi_t vifi;
1756 	struct vif *vifp;
1757 	int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1758 
1759 /*
1760  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1761  * input, they shouldn't get counted on output, so statistics keeping is
1762  * separate.
1763  */
1764 #define MC_SEND(ip, vifp, m) do {					\
1765 	if ((vifp)->v_flags & VIFF_TUNNEL)				\
1766 		encap_send((ip), (vifp), (m));				\
1767 	else								\
1768 		phyint_send((ip), (vifp), (m));				\
1769 } while (/*CONSTCOND*/ 0)
1770 
1771 #ifdef RSVP_ISI
1772 	/*
1773 	 * If xmt_vif is not -1, send on only the requested vif.
1774 	 *
1775 	 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1776 	 */
1777 	if (xmt_vif < numvifs) {
1778 #ifdef PIM
1779 		if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1780 			pim_register_send(ip, viftable + xmt_vif, m, rt);
1781 		else
1782 #endif
1783 		MC_SEND(ip, viftable + xmt_vif, m);
1784 		return (1);
1785 	}
1786 #endif /* RSVP_ISI */
1787 
1788 	/*
1789 	 * Don't forward if it didn't arrive from the parent vif for its origin.
1790 	 */
1791 	vifi = rt->mfc_parent;
1792 	if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1793 		/* came in the wrong interface */
1794 		if (mrtdebug & DEBUG_FORWARD)
1795 			log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1796 			    ifp, vifi,
1797 			    vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1798 		++mrtstat.mrts_wrong_if;
1799 		++rt->mfc_wrong_if;
1800 		/*
1801 		 * If we are doing PIM assert processing, send a message
1802 		 * to the routing daemon.
1803 		 *
1804 		 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1805 		 * can complete the SPT switch, regardless of the type
1806 		 * of the iif (broadcast media, GRE tunnel, etc).
1807 		 */
1808 		if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1809 			struct timeval now;
1810 			u_int32_t delta;
1811 
1812 #ifdef PIM
1813 			if (ifp == &multicast_register_if)
1814 				pimstat.pims_rcv_registers_wrongiif++;
1815 #endif
1816 
1817 			/* Get vifi for the incoming packet */
1818 			for (vifi = 0;
1819 			     vifi < numvifs && viftable[vifi].v_ifp != ifp;
1820 			     vifi++)
1821 			    ;
1822 			if (vifi >= numvifs) {
1823 				/* The iif is not found: ignore the packet. */
1824 				return (0);
1825 			}
1826 
1827 			if (rt->mfc_flags[vifi] &
1828 			    MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1829 				/* WRONGVIF disabled: ignore the packet */
1830 				return (0);
1831 			}
1832 
1833 			microtime(&now);
1834 
1835 			TV_DELTA(rt->mfc_last_assert, now, delta);
1836 
1837 			if (delta > ASSERT_MSG_TIME) {
1838 				struct igmpmsg *im;
1839 				int hlen = ip->ip_hl << 2;
1840 				struct mbuf *mm = m_copy(m, 0, hlen);
1841 
1842 				M_PULLUP(mm, hlen);
1843 				if (mm == NULL)
1844 					return (ENOBUFS);
1845 
1846 				rt->mfc_last_assert = now;
1847 
1848 				im = mtod(mm, struct igmpmsg *);
1849 				im->im_msgtype	= IGMPMSG_WRONGVIF;
1850 				im->im_mbz	= 0;
1851 				im->im_vif	= vifi;
1852 
1853 				mrtstat.mrts_upcalls++;
1854 
1855 				sin.sin_addr = im->im_src;
1856 				if (socket_send(ip_mrouter, mm, &sin) < 0) {
1857 					log(LOG_WARNING,
1858 					    "ip_mforward: ip_mrouter socket queue full\n");
1859 					++mrtstat.mrts_upq_sockfull;
1860 					return (ENOBUFS);
1861 				}
1862 			}
1863 		}
1864 		return (0);
1865 	}
1866 
1867 	/* If I sourced this packet, it counts as output, else it was input. */
1868 	if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1869 		viftable[vifi].v_pkt_out++;
1870 		viftable[vifi].v_bytes_out += plen;
1871 	} else {
1872 		viftable[vifi].v_pkt_in++;
1873 		viftable[vifi].v_bytes_in += plen;
1874 	}
1875 	rt->mfc_pkt_cnt++;
1876 	rt->mfc_byte_cnt += plen;
1877 
1878 	/*
1879 	 * For each vif, decide if a copy of the packet should be forwarded.
1880 	 * Forward if:
1881 	 *		- the ttl exceeds the vif's threshold
1882 	 *		- there are group members downstream on interface
1883 	 */
1884 	for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1885 		if ((rt->mfc_ttls[vifi] > 0) &&
1886 			(ip->ip_ttl > rt->mfc_ttls[vifi])) {
1887 			vifp->v_pkt_out++;
1888 			vifp->v_bytes_out += plen;
1889 #ifdef PIM
1890 			if (vifp->v_flags & VIFF_REGISTER)
1891 				pim_register_send(ip, vifp, m, rt);
1892 			else
1893 #endif
1894 			MC_SEND(ip, vifp, m);
1895 		}
1896 
1897 	/*
1898 	 * Perform upcall-related bw measuring.
1899 	 */
1900 	if (rt->mfc_bw_meter != NULL) {
1901 		struct bw_meter *x;
1902 		struct timeval now;
1903 
1904 		microtime(&now);
1905 		for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1906 			bw_meter_receive_packet(x, plen, &now);
1907 	}
1908 
1909 	return (0);
1910 }
1911 
1912 #ifdef RSVP_ISI
1913 /*
1914  * check if a vif number is legal/ok. This is used by ip_output.
1915  */
1916 int
1917 legal_vif_num(vif)
1918 	int vif;
1919 {
1920 	if (vif >= 0 && vif < numvifs)
1921 		return (1);
1922 	else
1923 		return (0);
1924 }
1925 #endif /* RSVP_ISI */
1926 
1927 static void
1928 phyint_send(ip, vifp, m)
1929 	struct ip *ip;
1930 	struct vif *vifp;
1931 	struct mbuf *m;
1932 {
1933 	struct mbuf *mb_copy;
1934 	int hlen = ip->ip_hl << 2;
1935 
1936 	/*
1937 	 * Make a new reference to the packet; make sure that
1938 	 * the IP header is actually copied, not just referenced,
1939 	 * so that ip_output() only scribbles on the copy.
1940 	 */
1941 	mb_copy = m_copy(m, 0, M_COPYALL);
1942 	M_PULLUP(mb_copy, hlen);
1943 	if (mb_copy == NULL)
1944 		return;
1945 
1946 	if (vifp->v_rate_limit <= 0)
1947 		tbf_send_packet(vifp, mb_copy);
1948 	else
1949 		tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1950 		    ntohs(ip->ip_len));
1951 }
1952 
1953 static void
1954 encap_send(ip, vifp, m)
1955 	struct ip *ip;
1956 	struct vif *vifp;
1957 	struct mbuf *m;
1958 {
1959 	struct mbuf *mb_copy;
1960 	struct ip *ip_copy;
1961 	int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1962 
1963 	/* Take care of delayed checksums */
1964 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1965 		in_delayed_cksum(m);
1966 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1967 	}
1968 
1969 	/*
1970 	 * copy the old packet & pullup it's IP header into the
1971 	 * new mbuf so we can modify it.  Try to fill the new
1972 	 * mbuf since if we don't the ethernet driver will.
1973 	 */
1974 	MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1975 	if (mb_copy == NULL)
1976 		return;
1977 	mb_copy->m_data += max_linkhdr;
1978 	mb_copy->m_pkthdr.len = len;
1979 	mb_copy->m_len = sizeof(multicast_encap_iphdr);
1980 
1981 	if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1982 		m_freem(mb_copy);
1983 		return;
1984 	}
1985 	i = MHLEN - max_linkhdr;
1986 	if (i > len)
1987 		i = len;
1988 	mb_copy = m_pullup(mb_copy, i);
1989 	if (mb_copy == NULL)
1990 		return;
1991 
1992 	/*
1993 	 * fill in the encapsulating IP header.
1994 	 */
1995 	ip_copy = mtod(mb_copy, struct ip *);
1996 	*ip_copy = multicast_encap_iphdr;
1997 	ip_copy->ip_id = ip_newid();
1998 	ip_copy->ip_len = htons(len);
1999 	ip_copy->ip_src = vifp->v_lcl_addr;
2000 	ip_copy->ip_dst = vifp->v_rmt_addr;
2001 
2002 	/*
2003 	 * turn the encapsulated IP header back into a valid one.
2004 	 */
2005 	ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
2006 	--ip->ip_ttl;
2007 	ip->ip_sum = 0;
2008 	mb_copy->m_data += sizeof(multicast_encap_iphdr);
2009 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2010 	mb_copy->m_data -= sizeof(multicast_encap_iphdr);
2011 
2012 	if (vifp->v_rate_limit <= 0)
2013 		tbf_send_packet(vifp, mb_copy);
2014 	else
2015 		tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
2016 }
2017 
2018 /*
2019  * De-encapsulate a packet and feed it back through ip input.
2020  */
2021 static void
2022 vif_input(struct mbuf *m, ...)
2023 {
2024 	int off, proto;
2025 	va_list ap;
2026 	struct vif *vifp;
2027 	int s;
2028 	struct ifqueue *ifq;
2029 
2030 	va_start(ap, m);
2031 	off = va_arg(ap, int);
2032 	proto = va_arg(ap, int);
2033 	va_end(ap);
2034 
2035 	vifp = (struct vif *)encap_getarg(m);
2036 	if (!vifp || proto != AF_INET) {
2037 		m_freem(m);
2038 		mrtstat.mrts_bad_tunnel++;
2039 		return;
2040 	}
2041 
2042 	m_adj(m, off);
2043 	m->m_pkthdr.rcvif = vifp->v_ifp;
2044 	ifq = &ipintrq;
2045 	s = splnet();
2046 	if (IF_QFULL(ifq)) {
2047 		IF_DROP(ifq);
2048 		m_freem(m);
2049 	} else {
2050 		IF_ENQUEUE(ifq, m);
2051 		/*
2052 		 * normally we would need a "schednetisr(NETISR_IP)"
2053 		 * here but we were called by ip_input and it is going
2054 		 * to loop back & try to dequeue the packet we just
2055 		 * queued as soon as we return so we avoid the
2056 		 * unnecessary software interrrupt.
2057 		 */
2058 	}
2059 	splx(s);
2060 }
2061 
2062 /*
2063  * Check if the packet should be grabbed by us.
2064  */
2065 static int
2066 vif_encapcheck(m, off, proto, arg)
2067 	const struct mbuf *m;
2068 	int off;
2069 	int proto;
2070 	void *arg;
2071 {
2072 	struct vif *vifp;
2073 	struct ip ip;
2074 
2075 #ifdef DIAGNOSTIC
2076 	if (!arg || proto != IPPROTO_IPV4)
2077 		panic("unexpected arg in vif_encapcheck");
2078 #endif
2079 
2080 	/*
2081 	 * do not grab the packet if it's not to a multicast destination or if
2082 	 * we don't have an encapsulating tunnel with the source.
2083 	 * Note:  This code assumes that the remote site IP address
2084 	 * uniquely identifies the tunnel (i.e., that this site has
2085 	 * at most one tunnel with the remote site).
2086 	 */
2087 
2088 	/* LINTED const cast */
2089 	m_copydata((struct mbuf *)m, off, sizeof(ip), (caddr_t)&ip);
2090 	if (!IN_MULTICAST(ip.ip_dst.s_addr))
2091 		return 0;
2092 
2093 	/* LINTED const cast */
2094 	m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip);
2095 	if (!in_hosteq(ip.ip_src, last_encap_src)) {
2096 		vifp = (struct vif *)arg;
2097 		if (vifp->v_flags & VIFF_TUNNEL &&
2098 		    in_hosteq(vifp->v_rmt_addr, ip.ip_src))
2099 			;
2100 		else
2101 			return 0;
2102 		last_encap_vif = vifp;
2103 		last_encap_src = ip.ip_src;
2104 	} else
2105 		vifp = last_encap_vif;
2106 
2107 	/* 32bit match, since we have checked ip_src only */
2108 	return 32;
2109 }
2110 
2111 /*
2112  * Token bucket filter module
2113  */
2114 static void
2115 tbf_control(vifp, m, ip, len)
2116 	struct vif *vifp;
2117 	struct mbuf *m;
2118 	struct ip *ip;
2119 	u_int32_t len;
2120 {
2121 
2122 	if (len > MAX_BKT_SIZE) {
2123 		/* drop if packet is too large */
2124 		mrtstat.mrts_pkt2large++;
2125 		m_freem(m);
2126 		return;
2127 	}
2128 
2129 	tbf_update_tokens(vifp);
2130 
2131 	/*
2132 	 * If there are enough tokens, and the queue is empty, send this packet
2133 	 * out immediately.  Otherwise, try to insert it on this vif's queue.
2134 	 */
2135 	if (vifp->tbf_q_len == 0) {
2136 		if (len <= vifp->tbf_n_tok) {
2137 			vifp->tbf_n_tok -= len;
2138 			tbf_send_packet(vifp, m);
2139 		} else {
2140 			/* queue packet and timeout till later */
2141 			tbf_queue(vifp, m);
2142 			callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2143 			    tbf_reprocess_q, vifp);
2144 		}
2145 	} else {
2146 		if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
2147 		    !tbf_dq_sel(vifp, ip)) {
2148 			/* queue full, and couldn't make room */
2149 			mrtstat.mrts_q_overflow++;
2150 			m_freem(m);
2151 		} else {
2152 			/* queue length low enough, or made room */
2153 			tbf_queue(vifp, m);
2154 			tbf_process_q(vifp);
2155 		}
2156 	}
2157 }
2158 
2159 /*
2160  * adds a packet to the queue at the interface
2161  */
2162 static void
2163 tbf_queue(vifp, m)
2164 	struct vif *vifp;
2165 	struct mbuf *m;
2166 {
2167 	int s = splsoftnet();
2168 
2169 	/* insert at tail */
2170 	*vifp->tbf_t = m;
2171 	vifp->tbf_t = &m->m_nextpkt;
2172 	vifp->tbf_q_len++;
2173 
2174 	splx(s);
2175 }
2176 
2177 
2178 /*
2179  * processes the queue at the interface
2180  */
2181 static void
2182 tbf_process_q(vifp)
2183 	struct vif *vifp;
2184 {
2185 	struct mbuf *m;
2186 	int len;
2187 	int s = splsoftnet();
2188 
2189 	/*
2190 	 * Loop through the queue at the interface and send as many packets
2191 	 * as possible.
2192 	 */
2193 	for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
2194 		len = ntohs(mtod(m, struct ip *)->ip_len);
2195 
2196 		/* determine if the packet can be sent */
2197 		if (len <= vifp->tbf_n_tok) {
2198 			/* if so,
2199 			 * reduce no of tokens, dequeue the packet,
2200 			 * send the packet.
2201 			 */
2202 			if ((vifp->tbf_q = m->m_nextpkt) == NULL)
2203 				vifp->tbf_t = &vifp->tbf_q;
2204 			--vifp->tbf_q_len;
2205 
2206 			m->m_nextpkt = NULL;
2207 			vifp->tbf_n_tok -= len;
2208 			tbf_send_packet(vifp, m);
2209 		} else
2210 			break;
2211 	}
2212 	splx(s);
2213 }
2214 
2215 static void
2216 tbf_reprocess_q(arg)
2217 	void *arg;
2218 {
2219 	struct vif *vifp = arg;
2220 
2221 	if (ip_mrouter == NULL)
2222 		return;
2223 
2224 	tbf_update_tokens(vifp);
2225 	tbf_process_q(vifp);
2226 
2227 	if (vifp->tbf_q_len != 0)
2228 		callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2229 		    tbf_reprocess_q, vifp);
2230 }
2231 
2232 /* function that will selectively discard a member of the queue
2233  * based on the precedence value and the priority
2234  */
2235 static int
2236 tbf_dq_sel(vifp, ip)
2237 	struct vif *vifp;
2238 	struct ip *ip;
2239 {
2240 	u_int p;
2241 	struct mbuf **mp, *m;
2242 	int s = splsoftnet();
2243 
2244 	p = priority(vifp, ip);
2245 
2246 	for (mp = &vifp->tbf_q, m = *mp;
2247 	    m != NULL;
2248 	    mp = &m->m_nextpkt, m = *mp) {
2249 		if (p > priority(vifp, mtod(m, struct ip *))) {
2250 			if ((*mp = m->m_nextpkt) == NULL)
2251 				vifp->tbf_t = mp;
2252 			--vifp->tbf_q_len;
2253 
2254 			m_freem(m);
2255 			mrtstat.mrts_drop_sel++;
2256 			splx(s);
2257 			return (1);
2258 		}
2259 	}
2260 	splx(s);
2261 	return (0);
2262 }
2263 
2264 static void
2265 tbf_send_packet(vifp, m)
2266 	struct vif *vifp;
2267 	struct mbuf *m;
2268 {
2269 	int error;
2270 	int s = splsoftnet();
2271 
2272 	if (vifp->v_flags & VIFF_TUNNEL) {
2273 		/* If tunnel options */
2274 		ip_output(m, (struct mbuf *)NULL, &vifp->v_route,
2275 		    IP_FORWARDING, (struct ip_moptions *)NULL,
2276 		    (struct socket *)NULL);
2277 	} else {
2278 		/* if physical interface option, extract the options and then send */
2279 		struct ip_moptions imo;
2280 
2281 		imo.imo_multicast_ifp = vifp->v_ifp;
2282 		imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2283 		imo.imo_multicast_loop = 1;
2284 #ifdef RSVP_ISI
2285 		imo.imo_multicast_vif = -1;
2286 #endif
2287 
2288 		error = ip_output(m, (struct mbuf *)NULL, (struct route *)NULL,
2289 		    IP_FORWARDING|IP_MULTICASTOPTS, &imo,
2290 		    (struct socket *)NULL);
2291 
2292 		if (mrtdebug & DEBUG_XMIT)
2293 			log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2294 			    (long)(vifp - viftable), error);
2295 	}
2296 	splx(s);
2297 }
2298 
2299 /* determine the current time and then
2300  * the elapsed time (between the last time and time now)
2301  * in milliseconds & update the no. of tokens in the bucket
2302  */
2303 static void
2304 tbf_update_tokens(vifp)
2305 	struct vif *vifp;
2306 {
2307 	struct timeval tp;
2308 	u_int32_t tm;
2309 	int s = splsoftnet();
2310 
2311 	microtime(&tp);
2312 
2313 	TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2314 
2315 	/*
2316 	 * This formula is actually
2317 	 * "time in seconds" * "bytes/second".
2318 	 *
2319 	 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2320 	 *
2321 	 * The (1000/1024) was introduced in add_vif to optimize
2322 	 * this divide into a shift.
2323 	 */
2324 	vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2325 	vifp->tbf_last_pkt_t = tp;
2326 
2327 	if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2328 		vifp->tbf_n_tok = MAX_BKT_SIZE;
2329 
2330 	splx(s);
2331 }
2332 
2333 static int
2334 priority(vifp, ip)
2335 	struct vif *vifp;
2336 	struct ip *ip;
2337 {
2338 	int prio = 50;	/* the lowest priority -- default case */
2339 
2340 	/* temporary hack; may add general packet classifier some day */
2341 
2342 	/*
2343 	 * The UDP port space is divided up into four priority ranges:
2344 	 * [0, 16384)     : unclassified - lowest priority
2345 	 * [16384, 32768) : audio - highest priority
2346 	 * [32768, 49152) : whiteboard - medium priority
2347 	 * [49152, 65536) : video - low priority
2348 	 */
2349 	if (ip->ip_p == IPPROTO_UDP) {
2350 		struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2351 
2352 		switch (ntohs(udp->uh_dport) & 0xc000) {
2353 		case 0x4000:
2354 			prio = 70;
2355 			break;
2356 		case 0x8000:
2357 			prio = 60;
2358 			break;
2359 		case 0xc000:
2360 			prio = 55;
2361 			break;
2362 		}
2363 
2364 		if (tbfdebug > 1)
2365 			log(LOG_DEBUG, "port %x prio %d\n",
2366 			    ntohs(udp->uh_dport), prio);
2367 	}
2368 
2369 	return (prio);
2370 }
2371 
2372 /*
2373  * End of token bucket filter modifications
2374  */
2375 #ifdef RSVP_ISI
2376 int
2377 ip_rsvp_vif_init(so, m)
2378 	struct socket *so;
2379 	struct mbuf *m;
2380 {
2381 	int vifi, s;
2382 
2383 	if (rsvpdebug)
2384 		printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2385 		    so->so_type, so->so_proto->pr_protocol);
2386 
2387 	if (so->so_type != SOCK_RAW ||
2388 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2389 		return (EOPNOTSUPP);
2390 
2391 	/* Check mbuf. */
2392 	if (m == NULL || m->m_len != sizeof(int)) {
2393 		return (EINVAL);
2394 	}
2395 	vifi = *(mtod(m, int *));
2396 
2397 	if (rsvpdebug)
2398 		printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",
2399 		       vifi, rsvp_on);
2400 
2401 	s = splsoftnet();
2402 
2403 	/* Check vif. */
2404 	if (!legal_vif_num(vifi)) {
2405 		splx(s);
2406 		return (EADDRNOTAVAIL);
2407 	}
2408 
2409 	/* Check if socket is available. */
2410 	if (viftable[vifi].v_rsvpd != NULL) {
2411 		splx(s);
2412 		return (EADDRINUSE);
2413 	}
2414 
2415 	viftable[vifi].v_rsvpd = so;
2416 	/*
2417 	 * This may seem silly, but we need to be sure we don't over-increment
2418 	 * the RSVP counter, in case something slips up.
2419 	 */
2420 	if (!viftable[vifi].v_rsvp_on) {
2421 		viftable[vifi].v_rsvp_on = 1;
2422 		rsvp_on++;
2423 	}
2424 
2425 	splx(s);
2426 	return (0);
2427 }
2428 
2429 int
2430 ip_rsvp_vif_done(so, m)
2431 	struct socket *so;
2432 	struct mbuf *m;
2433 {
2434 	int vifi, s;
2435 
2436 	if (rsvpdebug)
2437 		printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2438 		    so->so_type, so->so_proto->pr_protocol);
2439 
2440 	if (so->so_type != SOCK_RAW ||
2441 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2442 		return (EOPNOTSUPP);
2443 
2444 	/* Check mbuf. */
2445 	if (m == NULL || m->m_len != sizeof(int)) {
2446 		return (EINVAL);
2447 	}
2448 	vifi = *(mtod(m, int *));
2449 
2450 	s = splsoftnet();
2451 
2452 	/* Check vif. */
2453 	if (!legal_vif_num(vifi)) {
2454 		splx(s);
2455 		return (EADDRNOTAVAIL);
2456 	}
2457 
2458 	if (rsvpdebug)
2459 		printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2460 		    viftable[vifi].v_rsvpd, so);
2461 
2462 	viftable[vifi].v_rsvpd = NULL;
2463 	/*
2464 	 * This may seem silly, but we need to be sure we don't over-decrement
2465 	 * the RSVP counter, in case something slips up.
2466 	 */
2467 	if (viftable[vifi].v_rsvp_on) {
2468 		viftable[vifi].v_rsvp_on = 0;
2469 		rsvp_on--;
2470 	}
2471 
2472 	splx(s);
2473 	return (0);
2474 }
2475 
2476 void
2477 ip_rsvp_force_done(so)
2478 	struct socket *so;
2479 {
2480 	int vifi, s;
2481 
2482 	/* Don't bother if it is not the right type of socket. */
2483 	if (so->so_type != SOCK_RAW ||
2484 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2485 		return;
2486 
2487 	s = splsoftnet();
2488 
2489 	/*
2490 	 * The socket may be attached to more than one vif...this
2491 	 * is perfectly legal.
2492 	 */
2493 	for (vifi = 0; vifi < numvifs; vifi++) {
2494 		if (viftable[vifi].v_rsvpd == so) {
2495 			viftable[vifi].v_rsvpd = NULL;
2496 			/*
2497 			 * This may seem silly, but we need to be sure we don't
2498 			 * over-decrement the RSVP counter, in case something
2499 			 * slips up.
2500 			 */
2501 			if (viftable[vifi].v_rsvp_on) {
2502 				viftable[vifi].v_rsvp_on = 0;
2503 				rsvp_on--;
2504 			}
2505 		}
2506 	}
2507 
2508 	splx(s);
2509 	return;
2510 }
2511 
2512 void
2513 rsvp_input(m, ifp)
2514 	struct mbuf *m;
2515 	struct ifnet *ifp;
2516 {
2517 	int vifi, s;
2518 	struct ip *ip = mtod(m, struct ip *);
2519 	static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET };
2520 
2521 	if (rsvpdebug)
2522 		printf("rsvp_input: rsvp_on %d\n", rsvp_on);
2523 
2524 	/*
2525 	 * Can still get packets with rsvp_on = 0 if there is a local member
2526 	 * of the group to which the RSVP packet is addressed.  But in this
2527 	 * case we want to throw the packet away.
2528 	 */
2529 	if (!rsvp_on) {
2530 		m_freem(m);
2531 		return;
2532 	}
2533 
2534 	/*
2535 	 * If the old-style non-vif-associated socket is set, then use
2536 	 * it and ignore the new ones.
2537 	 */
2538 	if (ip_rsvpd != NULL) {
2539 		if (rsvpdebug)
2540 			printf("rsvp_input: "
2541 			    "Sending packet up old-style socket\n");
2542 		rip_input(m);	/*XXX*/
2543 		return;
2544 	}
2545 
2546 	s = splsoftnet();
2547 
2548 	if (rsvpdebug)
2549 		printf("rsvp_input: check vifs\n");
2550 
2551 	/* Find which vif the packet arrived on. */
2552 	for (vifi = 0; vifi < numvifs; vifi++) {
2553 		if (viftable[vifi].v_ifp == ifp)
2554 			break;
2555 	}
2556 
2557 	if (vifi == numvifs) {
2558 		/* Can't find vif packet arrived on. Drop packet. */
2559 		if (rsvpdebug)
2560 			printf("rsvp_input: "
2561 			    "Can't find vif for packet...dropping it.\n");
2562 		m_freem(m);
2563 		splx(s);
2564 		return;
2565 	}
2566 
2567 	if (rsvpdebug)
2568 		printf("rsvp_input: check socket\n");
2569 
2570 	if (viftable[vifi].v_rsvpd == NULL) {
2571 		/*
2572 		 * drop packet, since there is no specific socket for this
2573 		 * interface
2574 		 */
2575 		if (rsvpdebug)
2576 			printf("rsvp_input: No socket defined for vif %d\n",
2577 			    vifi);
2578 		m_freem(m);
2579 		splx(s);
2580 		return;
2581 	}
2582 
2583 	rsvp_src.sin_addr = ip->ip_src;
2584 
2585 	if (rsvpdebug && m)
2586 		printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
2587 		    m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv));
2588 
2589 	if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2590 		if (rsvpdebug)
2591 			printf("rsvp_input: Failed to append to socket\n");
2592 	else
2593 		if (rsvpdebug)
2594 			printf("rsvp_input: send packet up\n");
2595 
2596 	splx(s);
2597 }
2598 #endif /* RSVP_ISI */
2599 
2600 /*
2601  * Code for bandwidth monitors
2602  */
2603 
2604 /*
2605  * Define common interface for timeval-related methods
2606  */
2607 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2608 #define	BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2609 #define	BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2610 
2611 static uint32_t
2612 compute_bw_meter_flags(struct bw_upcall *req)
2613 {
2614     uint32_t flags = 0;
2615 
2616     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2617 	flags |= BW_METER_UNIT_PACKETS;
2618     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2619 	flags |= BW_METER_UNIT_BYTES;
2620     if (req->bu_flags & BW_UPCALL_GEQ)
2621 	flags |= BW_METER_GEQ;
2622     if (req->bu_flags & BW_UPCALL_LEQ)
2623 	flags |= BW_METER_LEQ;
2624 
2625     return flags;
2626 }
2627 
2628 /*
2629  * Add a bw_meter entry
2630  */
2631 static int
2632 add_bw_upcall(struct mbuf *m)
2633 {
2634     int s;
2635     struct mfc *mfc;
2636     struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2637 		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2638     struct timeval now;
2639     struct bw_meter *x;
2640     uint32_t flags;
2641     struct bw_upcall *req;
2642 
2643     if (m == NULL || m->m_len < sizeof(struct bw_upcall))
2644 	return EINVAL;
2645 
2646     req = mtod(m, struct bw_upcall *);
2647 
2648     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2649 	return EOPNOTSUPP;
2650 
2651     /* Test if the flags are valid */
2652     if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2653 	return EINVAL;
2654     if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2655 	return EINVAL;
2656     if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2657 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2658 	return EINVAL;
2659 
2660     /* Test if the threshold time interval is valid */
2661     if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2662 	return EINVAL;
2663 
2664     flags = compute_bw_meter_flags(req);
2665 
2666     /*
2667      * Find if we have already same bw_meter entry
2668      */
2669     s = splsoftnet();
2670     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2671     if (mfc == NULL) {
2672 	splx(s);
2673 	return EADDRNOTAVAIL;
2674     }
2675     for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2676 	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2677 			   &req->bu_threshold.b_time, ==)) &&
2678 	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2679 	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2680 	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
2681 	    splx(s);
2682 	    return 0;		/* XXX Already installed */
2683 	}
2684     }
2685 
2686     /* Allocate the new bw_meter entry */
2687     x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2688     if (x == NULL) {
2689 	splx(s);
2690 	return ENOBUFS;
2691     }
2692 
2693     /* Set the new bw_meter entry */
2694     x->bm_threshold.b_time = req->bu_threshold.b_time;
2695     microtime(&now);
2696     x->bm_start_time = now;
2697     x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2698     x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2699     x->bm_measured.b_packets = 0;
2700     x->bm_measured.b_bytes = 0;
2701     x->bm_flags = flags;
2702     x->bm_time_next = NULL;
2703     x->bm_time_hash = BW_METER_BUCKETS;
2704 
2705     /* Add the new bw_meter entry to the front of entries for this MFC */
2706     x->bm_mfc = mfc;
2707     x->bm_mfc_next = mfc->mfc_bw_meter;
2708     mfc->mfc_bw_meter = x;
2709     schedule_bw_meter(x, &now);
2710     splx(s);
2711 
2712     return 0;
2713 }
2714 
2715 static void
2716 free_bw_list(struct bw_meter *list)
2717 {
2718     while (list != NULL) {
2719 	struct bw_meter *x = list;
2720 
2721 	list = list->bm_mfc_next;
2722 	unschedule_bw_meter(x);
2723 	free(x, M_BWMETER);
2724     }
2725 }
2726 
2727 /*
2728  * Delete one or multiple bw_meter entries
2729  */
2730 static int
2731 del_bw_upcall(struct mbuf *m)
2732 {
2733     int s;
2734     struct mfc *mfc;
2735     struct bw_meter *x;
2736     struct bw_upcall *req;
2737 
2738     if (m == NULL || m->m_len < sizeof(struct bw_upcall))
2739 	return EINVAL;
2740 
2741     req = mtod(m, struct bw_upcall *);
2742 
2743     if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2744 	return EOPNOTSUPP;
2745 
2746     s = splsoftnet();
2747     /* Find the corresponding MFC entry */
2748     mfc = mfc_find(&req->bu_src, &req->bu_dst);
2749     if (mfc == NULL) {
2750 	splx(s);
2751 	return EADDRNOTAVAIL;
2752     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2753 	/*
2754 	 * Delete all bw_meter entries for this mfc
2755 	 */
2756 	struct bw_meter *list;
2757 
2758 	list = mfc->mfc_bw_meter;
2759 	mfc->mfc_bw_meter = NULL;
2760 	free_bw_list(list);
2761 	splx(s);
2762 	return 0;
2763     } else {			/* Delete a single bw_meter entry */
2764 	struct bw_meter *prev;
2765 	uint32_t flags = 0;
2766 
2767 	flags = compute_bw_meter_flags(req);
2768 
2769 	/* Find the bw_meter entry to delete */
2770 	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2771 	     prev = x, x = x->bm_mfc_next) {
2772 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2773 			       &req->bu_threshold.b_time, ==)) &&
2774 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2775 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2776 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2777 		break;
2778 	}
2779 	if (x != NULL) { /* Delete entry from the list for this MFC */
2780 	    if (prev != NULL)
2781 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2782 	    else
2783 		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2784 
2785 	    unschedule_bw_meter(x);
2786 	    splx(s);
2787 	    /* Free the bw_meter entry */
2788 	    free(x, M_BWMETER);
2789 	    return 0;
2790 	} else {
2791 	    splx(s);
2792 	    return EINVAL;
2793 	}
2794     }
2795     /* NOTREACHED */
2796 }
2797 
2798 /*
2799  * Perform bandwidth measurement processing that may result in an upcall
2800  */
2801 static void
2802 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2803 {
2804     struct timeval delta;
2805 
2806     delta = *nowp;
2807     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2808 
2809     if (x->bm_flags & BW_METER_GEQ) {
2810 	/*
2811 	 * Processing for ">=" type of bw_meter entry
2812 	 */
2813 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2814 	    /* Reset the bw_meter entry */
2815 	    x->bm_start_time = *nowp;
2816 	    x->bm_measured.b_packets = 0;
2817 	    x->bm_measured.b_bytes = 0;
2818 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2819 	}
2820 
2821 	/* Record that a packet is received */
2822 	x->bm_measured.b_packets++;
2823 	x->bm_measured.b_bytes += plen;
2824 
2825 	/*
2826 	 * Test if we should deliver an upcall
2827 	 */
2828 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2829 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2830 		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2831 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2832 		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2833 		/* Prepare an upcall for delivery */
2834 		bw_meter_prepare_upcall(x, nowp);
2835 		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2836 	    }
2837 	}
2838     } else if (x->bm_flags & BW_METER_LEQ) {
2839 	/*
2840 	 * Processing for "<=" type of bw_meter entry
2841 	 */
2842 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2843 	    /*
2844 	     * We are behind time with the multicast forwarding table
2845 	     * scanning for "<=" type of bw_meter entries, so test now
2846 	     * if we should deliver an upcall.
2847 	     */
2848 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2849 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2850 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2851 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2852 		/* Prepare an upcall for delivery */
2853 		bw_meter_prepare_upcall(x, nowp);
2854 	    }
2855 	    /* Reschedule the bw_meter entry */
2856 	    unschedule_bw_meter(x);
2857 	    schedule_bw_meter(x, nowp);
2858 	}
2859 
2860 	/* Record that a packet is received */
2861 	x->bm_measured.b_packets++;
2862 	x->bm_measured.b_bytes += plen;
2863 
2864 	/*
2865 	 * Test if we should restart the measuring interval
2866 	 */
2867 	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2868 	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2869 	    (x->bm_flags & BW_METER_UNIT_BYTES &&
2870 	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2871 	    /* Don't restart the measuring interval */
2872 	} else {
2873 	    /* Do restart the measuring interval */
2874 	    /*
2875 	     * XXX: note that we don't unschedule and schedule, because this
2876 	     * might be too much overhead per packet. Instead, when we process
2877 	     * all entries for a given timer hash bin, we check whether it is
2878 	     * really a timeout. If not, we reschedule at that time.
2879 	     */
2880 	    x->bm_start_time = *nowp;
2881 	    x->bm_measured.b_packets = 0;
2882 	    x->bm_measured.b_bytes = 0;
2883 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2884 	}
2885     }
2886 }
2887 
2888 /*
2889  * Prepare a bandwidth-related upcall
2890  */
2891 static void
2892 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2893 {
2894     struct timeval delta;
2895     struct bw_upcall *u;
2896 
2897     /*
2898      * Compute the measured time interval
2899      */
2900     delta = *nowp;
2901     BW_TIMEVALDECR(&delta, &x->bm_start_time);
2902 
2903     /*
2904      * If there are too many pending upcalls, deliver them now
2905      */
2906     if (bw_upcalls_n >= BW_UPCALLS_MAX)
2907 	bw_upcalls_send();
2908 
2909     /*
2910      * Set the bw_upcall entry
2911      */
2912     u = &bw_upcalls[bw_upcalls_n++];
2913     u->bu_src = x->bm_mfc->mfc_origin;
2914     u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2915     u->bu_threshold.b_time = x->bm_threshold.b_time;
2916     u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2917     u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2918     u->bu_measured.b_time = delta;
2919     u->bu_measured.b_packets = x->bm_measured.b_packets;
2920     u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2921     u->bu_flags = 0;
2922     if (x->bm_flags & BW_METER_UNIT_PACKETS)
2923 	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2924     if (x->bm_flags & BW_METER_UNIT_BYTES)
2925 	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2926     if (x->bm_flags & BW_METER_GEQ)
2927 	u->bu_flags |= BW_UPCALL_GEQ;
2928     if (x->bm_flags & BW_METER_LEQ)
2929 	u->bu_flags |= BW_UPCALL_LEQ;
2930 }
2931 
2932 /*
2933  * Send the pending bandwidth-related upcalls
2934  */
2935 static void
2936 bw_upcalls_send(void)
2937 {
2938     struct mbuf *m;
2939     int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2940     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2941     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2942 				      0,		/* unused2 */
2943 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2944 				      0,		/* im_mbz  */
2945 				      0,		/* im_vif  */
2946 				      0,		/* unused3 */
2947 				      { 0 },		/* im_src  */
2948 				      { 0 } };		/* im_dst  */
2949 
2950     if (bw_upcalls_n == 0)
2951 	return;			/* No pending upcalls */
2952 
2953     bw_upcalls_n = 0;
2954 
2955     /*
2956      * Allocate a new mbuf, initialize it with the header and
2957      * the payload for the pending calls.
2958      */
2959     MGETHDR(m, M_DONTWAIT, MT_HEADER);
2960     if (m == NULL) {
2961 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2962 	return;
2963     }
2964 
2965     m->m_len = m->m_pkthdr.len = 0;
2966     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2967     m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2968 
2969     /*
2970      * Send the upcalls
2971      * XXX do we need to set the address in k_igmpsrc ?
2972      */
2973     mrtstat.mrts_upcalls++;
2974     if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2975 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2976 	++mrtstat.mrts_upq_sockfull;
2977     }
2978 }
2979 
2980 /*
2981  * Compute the timeout hash value for the bw_meter entries
2982  */
2983 #define	BW_METER_TIMEHASH(bw_meter, hash)				\
2984     do {								\
2985 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2986 									\
2987 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2988 	(hash) = next_timeval.tv_sec;					\
2989 	if (next_timeval.tv_usec)					\
2990 	    (hash)++; /* XXX: make sure we don't timeout early */	\
2991 	(hash) %= BW_METER_BUCKETS;					\
2992     } while (/*CONSTCOND*/ 0)
2993 
2994 /*
2995  * Schedule a timer to process periodically bw_meter entry of type "<="
2996  * by linking the entry in the proper hash bucket.
2997  */
2998 static void
2999 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
3000 {
3001     int time_hash;
3002 
3003     if (!(x->bm_flags & BW_METER_LEQ))
3004 	return;		/* XXX: we schedule timers only for "<=" entries */
3005 
3006     /*
3007      * Reset the bw_meter entry
3008      */
3009     x->bm_start_time = *nowp;
3010     x->bm_measured.b_packets = 0;
3011     x->bm_measured.b_bytes = 0;
3012     x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
3013 
3014     /*
3015      * Compute the timeout hash value and insert the entry
3016      */
3017     BW_METER_TIMEHASH(x, time_hash);
3018     x->bm_time_next = bw_meter_timers[time_hash];
3019     bw_meter_timers[time_hash] = x;
3020     x->bm_time_hash = time_hash;
3021 }
3022 
3023 /*
3024  * Unschedule the periodic timer that processes bw_meter entry of type "<="
3025  * by removing the entry from the proper hash bucket.
3026  */
3027 static void
3028 unschedule_bw_meter(struct bw_meter *x)
3029 {
3030     int time_hash;
3031     struct bw_meter *prev, *tmp;
3032 
3033     if (!(x->bm_flags & BW_METER_LEQ))
3034 	return;		/* XXX: we schedule timers only for "<=" entries */
3035 
3036     /*
3037      * Compute the timeout hash value and delete the entry
3038      */
3039     time_hash = x->bm_time_hash;
3040     if (time_hash >= BW_METER_BUCKETS)
3041 	return;		/* Entry was not scheduled */
3042 
3043     for (prev = NULL, tmp = bw_meter_timers[time_hash];
3044 	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
3045 	if (tmp == x)
3046 	    break;
3047 
3048     if (tmp == NULL)
3049 	panic("unschedule_bw_meter: bw_meter entry not found");
3050 
3051     if (prev != NULL)
3052 	prev->bm_time_next = x->bm_time_next;
3053     else
3054 	bw_meter_timers[time_hash] = x->bm_time_next;
3055 
3056     x->bm_time_next = NULL;
3057     x->bm_time_hash = BW_METER_BUCKETS;
3058 }
3059 
3060 /*
3061  * Process all "<=" type of bw_meter that should be processed now,
3062  * and for each entry prepare an upcall if necessary. Each processed
3063  * entry is rescheduled again for the (periodic) processing.
3064  *
3065  * This is run periodically (once per second normally). On each round,
3066  * all the potentially matching entries are in the hash slot that we are
3067  * looking at.
3068  */
3069 static void
3070 bw_meter_process()
3071 {
3072     int s;
3073     static uint32_t last_tv_sec;	/* last time we processed this */
3074 
3075     uint32_t loops;
3076     int i;
3077     struct timeval now, process_endtime;
3078 
3079     microtime(&now);
3080     if (last_tv_sec == now.tv_sec)
3081 	return;		/* nothing to do */
3082 
3083     loops = now.tv_sec - last_tv_sec;
3084     last_tv_sec = now.tv_sec;
3085     if (loops > BW_METER_BUCKETS)
3086 	loops = BW_METER_BUCKETS;
3087 
3088     s = splsoftnet();
3089     /*
3090      * Process all bins of bw_meter entries from the one after the last
3091      * processed to the current one. On entry, i points to the last bucket
3092      * visited, so we need to increment i at the beginning of the loop.
3093      */
3094     for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
3095 	struct bw_meter *x, *tmp_list;
3096 
3097 	if (++i >= BW_METER_BUCKETS)
3098 	    i = 0;
3099 
3100 	/* Disconnect the list of bw_meter entries from the bin */
3101 	tmp_list = bw_meter_timers[i];
3102 	bw_meter_timers[i] = NULL;
3103 
3104 	/* Process the list of bw_meter entries */
3105 	while (tmp_list != NULL) {
3106 	    x = tmp_list;
3107 	    tmp_list = tmp_list->bm_time_next;
3108 
3109 	    /* Test if the time interval is over */
3110 	    process_endtime = x->bm_start_time;
3111 	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
3112 	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
3113 		/* Not yet: reschedule, but don't reset */
3114 		int time_hash;
3115 
3116 		BW_METER_TIMEHASH(x, time_hash);
3117 		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
3118 		    /*
3119 		     * XXX: somehow the bin processing is a bit ahead of time.
3120 		     * Put the entry in the next bin.
3121 		     */
3122 		    if (++time_hash >= BW_METER_BUCKETS)
3123 			time_hash = 0;
3124 		}
3125 		x->bm_time_next = bw_meter_timers[time_hash];
3126 		bw_meter_timers[time_hash] = x;
3127 		x->bm_time_hash = time_hash;
3128 
3129 		continue;
3130 	    }
3131 
3132 	    /*
3133 	     * Test if we should deliver an upcall
3134 	     */
3135 	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
3136 		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
3137 		((x->bm_flags & BW_METER_UNIT_BYTES) &&
3138 		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
3139 		/* Prepare an upcall for delivery */
3140 		bw_meter_prepare_upcall(x, &now);
3141 	    }
3142 
3143 	    /*
3144 	     * Reschedule for next processing
3145 	     */
3146 	    schedule_bw_meter(x, &now);
3147 	}
3148     }
3149 
3150     /* Send all upcalls that are pending delivery */
3151     bw_upcalls_send();
3152 
3153     splx(s);
3154 }
3155 
3156 /*
3157  * A periodic function for sending all upcalls that are pending delivery
3158  */
3159 static void
3160 expire_bw_upcalls_send(void *unused)
3161 {
3162     int s;
3163 
3164     s = splsoftnet();
3165     bw_upcalls_send();
3166     splx(s);
3167 
3168     callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
3169 		  expire_bw_upcalls_send, NULL);
3170 }
3171 
3172 /*
3173  * A periodic function for periodic scanning of the multicast forwarding
3174  * table for processing all "<=" bw_meter entries.
3175  */
3176 static void
3177 expire_bw_meter_process(void *unused)
3178 {
3179     if (mrt_api_config & MRT_MFC_BW_UPCALL)
3180 	bw_meter_process();
3181 
3182     callout_reset(&bw_meter_ch, BW_METER_PERIOD,
3183 		  expire_bw_meter_process, NULL);
3184 }
3185 
3186 /*
3187  * End of bandwidth monitoring code
3188  */
3189 
3190 #ifdef PIM
3191 /*
3192  * Send the packet up to the user daemon, or eventually do kernel encapsulation
3193  */
3194 static int
3195 pim_register_send(struct ip *ip, struct vif *vifp,
3196 	struct mbuf *m, struct mfc *rt)
3197 {
3198     struct mbuf *mb_copy, *mm;
3199 
3200     if (mrtdebug & DEBUG_PIM)
3201         log(LOG_DEBUG, "pim_register_send: ");
3202 
3203     mb_copy = pim_register_prepare(ip, m);
3204     if (mb_copy == NULL)
3205 	return ENOBUFS;
3206 
3207     /*
3208      * Send all the fragments. Note that the mbuf for each fragment
3209      * is freed by the sending machinery.
3210      */
3211     for (mm = mb_copy; mm; mm = mb_copy) {
3212 	mb_copy = mm->m_nextpkt;
3213 	mm->m_nextpkt = NULL;
3214 	mm = m_pullup(mm, sizeof(struct ip));
3215 	if (mm != NULL) {
3216 	    ip = mtod(mm, struct ip *);
3217 	    if ((mrt_api_config & MRT_MFC_RP) &&
3218 		!in_nullhost(rt->mfc_rp)) {
3219 		pim_register_send_rp(ip, vifp, mm, rt);
3220 	    } else {
3221 		pim_register_send_upcall(ip, vifp, mm, rt);
3222 	    }
3223 	}
3224     }
3225 
3226     return 0;
3227 }
3228 
3229 /*
3230  * Return a copy of the data packet that is ready for PIM Register
3231  * encapsulation.
3232  * XXX: Note that in the returned copy the IP header is a valid one.
3233  */
3234 static struct mbuf *
3235 pim_register_prepare(struct ip *ip, struct mbuf *m)
3236 {
3237     struct mbuf *mb_copy = NULL;
3238     int mtu;
3239 
3240     /* Take care of delayed checksums */
3241     if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3242 	in_delayed_cksum(m);
3243 	m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3244     }
3245 
3246     /*
3247      * Copy the old packet & pullup its IP header into the
3248      * new mbuf so we can modify it.
3249      */
3250     mb_copy = m_copy(m, 0, M_COPYALL);
3251     if (mb_copy == NULL)
3252 	return NULL;
3253     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3254     if (mb_copy == NULL)
3255 	return NULL;
3256 
3257     /* take care of the TTL */
3258     ip = mtod(mb_copy, struct ip *);
3259     --ip->ip_ttl;
3260 
3261     /* Compute the MTU after the PIM Register encapsulation */
3262     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3263 
3264     if (ntohs(ip->ip_len) <= mtu) {
3265 	/* Turn the IP header into a valid one */
3266 	ip->ip_sum = 0;
3267 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3268     } else {
3269 	/* Fragment the packet */
3270 	if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3271 	    /* XXX: mb_copy was freed by ip_fragment() */
3272 	    return NULL;
3273 	}
3274     }
3275     return mb_copy;
3276 }
3277 
3278 /*
3279  * Send an upcall with the data packet to the user-level process.
3280  */
3281 static int
3282 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3283 	struct mbuf *mb_copy, struct mfc *rt)
3284 {
3285     struct mbuf *mb_first;
3286     int len = ntohs(ip->ip_len);
3287     struct igmpmsg *im;
3288     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
3289 
3290     /*
3291      * Add a new mbuf with an upcall header
3292      */
3293     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3294     if (mb_first == NULL) {
3295 	m_freem(mb_copy);
3296 	return ENOBUFS;
3297     }
3298     mb_first->m_data += max_linkhdr;
3299     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3300     mb_first->m_len = sizeof(struct igmpmsg);
3301     mb_first->m_next = mb_copy;
3302 
3303     /* Send message to routing daemon */
3304     im = mtod(mb_first, struct igmpmsg *);
3305     im->im_msgtype	= IGMPMSG_WHOLEPKT;
3306     im->im_mbz		= 0;
3307     im->im_vif		= vifp - viftable;
3308     im->im_src		= ip->ip_src;
3309     im->im_dst		= ip->ip_dst;
3310 
3311     k_igmpsrc.sin_addr	= ip->ip_src;
3312 
3313     mrtstat.mrts_upcalls++;
3314 
3315     if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3316 	if (mrtdebug & DEBUG_PIM)
3317 	    log(LOG_WARNING,
3318 		"mcast: pim_register_send_upcall: ip_mrouter socket queue full");
3319 	++mrtstat.mrts_upq_sockfull;
3320 	return ENOBUFS;
3321     }
3322 
3323     /* Keep statistics */
3324     pimstat.pims_snd_registers_msgs++;
3325     pimstat.pims_snd_registers_bytes += len;
3326 
3327     return 0;
3328 }
3329 
3330 /*
3331  * Encapsulate the data packet in PIM Register message and send it to the RP.
3332  */
3333 static int
3334 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3335 	struct mbuf *mb_copy, struct mfc *rt)
3336 {
3337     struct mbuf *mb_first;
3338     struct ip *ip_outer;
3339     struct pim_encap_pimhdr *pimhdr;
3340     int len = ntohs(ip->ip_len);
3341     vifi_t vifi = rt->mfc_parent;
3342 
3343     if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3344 	m_freem(mb_copy);
3345 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
3346     }
3347 
3348     /*
3349      * Add a new mbuf with the encapsulating header
3350      */
3351     MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3352     if (mb_first == NULL) {
3353 	m_freem(mb_copy);
3354 	return ENOBUFS;
3355     }
3356     mb_first->m_data += max_linkhdr;
3357     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3358     mb_first->m_next = mb_copy;
3359 
3360     mb_first->m_pkthdr.len = len + mb_first->m_len;
3361 
3362     /*
3363      * Fill in the encapsulating IP and PIM header
3364      */
3365     ip_outer = mtod(mb_first, struct ip *);
3366     *ip_outer = pim_encap_iphdr;
3367     ip_outer->ip_id = ip_newid();
3368     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3369 			     sizeof(pim_encap_pimhdr));
3370     ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3371     ip_outer->ip_dst = rt->mfc_rp;
3372     /*
3373      * Copy the inner header TOS to the outer header, and take care of the
3374      * IP_DF bit.
3375      */
3376     ip_outer->ip_tos = ip->ip_tos;
3377     if (ntohs(ip->ip_off) & IP_DF)
3378 	ip_outer->ip_off |= IP_DF;
3379     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
3380 					 + sizeof(pim_encap_iphdr));
3381     *pimhdr = pim_encap_pimhdr;
3382     /* If the iif crosses a border, set the Border-bit */
3383     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3384 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3385 
3386     mb_first->m_data += sizeof(pim_encap_iphdr);
3387     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3388     mb_first->m_data -= sizeof(pim_encap_iphdr);
3389 
3390     if (vifp->v_rate_limit == 0)
3391 	tbf_send_packet(vifp, mb_first);
3392     else
3393 	tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3394 
3395     /* Keep statistics */
3396     pimstat.pims_snd_registers_msgs++;
3397     pimstat.pims_snd_registers_bytes += len;
3398 
3399     return 0;
3400 }
3401 
3402 /*
3403  * PIM-SMv2 and PIM-DM messages processing.
3404  * Receives and verifies the PIM control messages, and passes them
3405  * up to the listening socket, using rip_input().
3406  * The only message with special processing is the PIM_REGISTER message
3407  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3408  * is passed to if_simloop().
3409  */
3410 void
3411 pim_input(struct mbuf *m, ...)
3412 {
3413     struct ip *ip = mtod(m, struct ip *);
3414     struct pim *pim;
3415     int minlen;
3416     int datalen;
3417     int ip_tos;
3418     int proto;
3419     int iphlen;
3420     va_list ap;
3421 
3422     va_start(ap, m);
3423     iphlen = va_arg(ap, int);
3424     proto = va_arg(ap, int);
3425     va_end(ap);
3426 
3427     datalen = ntohs(ip->ip_len) - iphlen;
3428 
3429     /* Keep statistics */
3430     pimstat.pims_rcv_total_msgs++;
3431     pimstat.pims_rcv_total_bytes += datalen;
3432 
3433     /*
3434      * Validate lengths
3435      */
3436     if (datalen < PIM_MINLEN) {
3437 	pimstat.pims_rcv_tooshort++;
3438 	log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3439 	    datalen, (u_long)ip->ip_src.s_addr);
3440 	m_freem(m);
3441 	return;
3442     }
3443 
3444     /*
3445      * If the packet is at least as big as a REGISTER, go agead
3446      * and grab the PIM REGISTER header size, to avoid another
3447      * possible m_pullup() later.
3448      *
3449      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
3450      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3451      */
3452     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3453     /*
3454      * Get the IP and PIM headers in contiguous memory, and
3455      * possibly the PIM REGISTER header.
3456      */
3457     if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3458 	(m = m_pullup(m, minlen)) == NULL) {
3459 	log(LOG_ERR, "pim_input: m_pullup failure\n");
3460 	return;
3461     }
3462     /* m_pullup() may have given us a new mbuf so reset ip. */
3463     ip = mtod(m, struct ip *);
3464     ip_tos = ip->ip_tos;
3465 
3466     /* adjust mbuf to point to the PIM header */
3467     m->m_data += iphlen;
3468     m->m_len  -= iphlen;
3469     pim = mtod(m, struct pim *);
3470 
3471     /*
3472      * Validate checksum. If PIM REGISTER, exclude the data packet.
3473      *
3474      * XXX: some older PIMv2 implementations don't make this distinction,
3475      * so for compatibility reason perform the checksum over part of the
3476      * message, and if error, then over the whole message.
3477      */
3478     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3479 	/* do nothing, checksum okay */
3480     } else if (in_cksum(m, datalen)) {
3481 	pimstat.pims_rcv_badsum++;
3482 	if (mrtdebug & DEBUG_PIM)
3483 	    log(LOG_DEBUG, "pim_input: invalid checksum");
3484 	m_freem(m);
3485 	return;
3486     }
3487 
3488     /* PIM version check */
3489     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3490 	pimstat.pims_rcv_badversion++;
3491 	log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3492 	    PIM_VT_V(pim->pim_vt), PIM_VERSION);
3493 	m_freem(m);
3494 	return;
3495     }
3496 
3497     /* restore mbuf back to the outer IP */
3498     m->m_data -= iphlen;
3499     m->m_len  += iphlen;
3500 
3501     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3502 	/*
3503 	 * Since this is a REGISTER, we'll make a copy of the register
3504 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3505 	 * routing daemon.
3506 	 */
3507 	int s;
3508 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
3509 	struct mbuf *mcp;
3510 	struct ip *encap_ip;
3511 	u_int32_t *reghdr;
3512 	struct ifnet *vifp;
3513 
3514 	s = splsoftnet();
3515 	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3516 	    splx(s);
3517 	    if (mrtdebug & DEBUG_PIM)
3518 		log(LOG_DEBUG,
3519 		    "pim_input: register vif not set: %d\n", reg_vif_num);
3520 	    m_freem(m);
3521 	    return;
3522 	}
3523 	/* XXX need refcnt? */
3524 	vifp = viftable[reg_vif_num].v_ifp;
3525 	splx(s);
3526 
3527 	/*
3528 	 * Validate length
3529 	 */
3530 	if (datalen < PIM_REG_MINLEN) {
3531 	    pimstat.pims_rcv_tooshort++;
3532 	    pimstat.pims_rcv_badregisters++;
3533 	    log(LOG_ERR,
3534 		"pim_input: register packet size too small %d from %lx\n",
3535 		datalen, (u_long)ip->ip_src.s_addr);
3536 	    m_freem(m);
3537 	    return;
3538 	}
3539 
3540 	reghdr = (u_int32_t *)(pim + 1);
3541 	encap_ip = (struct ip *)(reghdr + 1);
3542 
3543 	if (mrtdebug & DEBUG_PIM) {
3544 	    log(LOG_DEBUG,
3545 		"pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3546 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3547 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3548 		ntohs(encap_ip->ip_len));
3549 	}
3550 
3551 	/* verify the version number of the inner packet */
3552 	if (encap_ip->ip_v != IPVERSION) {
3553 	    pimstat.pims_rcv_badregisters++;
3554 	    if (mrtdebug & DEBUG_PIM) {
3555 		log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3556 		    "of the inner packet\n", encap_ip->ip_v);
3557 	    }
3558 	    m_freem(m);
3559 	    return;
3560 	}
3561 
3562 	/* verify the inner packet is destined to a mcast group */
3563 	if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3564 	    pimstat.pims_rcv_badregisters++;
3565 	    if (mrtdebug & DEBUG_PIM)
3566 		log(LOG_DEBUG,
3567 		    "pim_input: inner packet of register is not "
3568 		    "multicast %lx\n",
3569 		    (u_long)ntohl(encap_ip->ip_dst.s_addr));
3570 	    m_freem(m);
3571 	    return;
3572 	}
3573 
3574 	/* If a NULL_REGISTER, pass it to the daemon */
3575 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3576 	    goto pim_input_to_daemon;
3577 
3578 	/*
3579 	 * Copy the TOS from the outer IP header to the inner IP header.
3580 	 */
3581 	if (encap_ip->ip_tos != ip_tos) {
3582 	    /* Outer TOS -> inner TOS */
3583 	    encap_ip->ip_tos = ip_tos;
3584 	    /* Recompute the inner header checksum. Sigh... */
3585 
3586 	    /* adjust mbuf to point to the inner IP header */
3587 	    m->m_data += (iphlen + PIM_MINLEN);
3588 	    m->m_len  -= (iphlen + PIM_MINLEN);
3589 
3590 	    encap_ip->ip_sum = 0;
3591 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3592 
3593 	    /* restore mbuf to point back to the outer IP header */
3594 	    m->m_data -= (iphlen + PIM_MINLEN);
3595 	    m->m_len  += (iphlen + PIM_MINLEN);
3596 	}
3597 
3598 	/*
3599 	 * Decapsulate the inner IP packet and loopback to forward it
3600 	 * as a normal multicast packet. Also, make a copy of the
3601 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
3602 	 * to pass to the daemon later, so it can take the appropriate
3603 	 * actions (e.g., send back PIM_REGISTER_STOP).
3604 	 * XXX: here m->m_data points to the outer IP header.
3605 	 */
3606 	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
3607 	if (mcp == NULL) {
3608 	    log(LOG_ERR,
3609 		"pim_input: pim register: could not copy register head\n");
3610 	    m_freem(m);
3611 	    return;
3612 	}
3613 
3614 	/* Keep statistics */
3615 	/* XXX: registers_bytes include only the encap. mcast pkt */
3616 	pimstat.pims_rcv_registers_msgs++;
3617 	pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3618 
3619 	/*
3620 	 * forward the inner ip packet; point m_data at the inner ip.
3621 	 */
3622 	m_adj(m, iphlen + PIM_MINLEN);
3623 
3624 	if (mrtdebug & DEBUG_PIM) {
3625 	    log(LOG_DEBUG,
3626 		"pim_input: forwarding decapsulated register: "
3627 		"src %lx, dst %lx, vif %d\n",
3628 		(u_long)ntohl(encap_ip->ip_src.s_addr),
3629 		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3630 		reg_vif_num);
3631 	}
3632 	/* NB: vifp was collected above; can it change on us? */
3633 	looutput(vifp, m, (struct sockaddr *)&dst, (struct rtentry *)NULL);
3634 
3635 	/* prepare the register head to send to the mrouting daemon */
3636 	m = mcp;
3637     }
3638 
3639 pim_input_to_daemon:
3640     /*
3641      * Pass the PIM message up to the daemon; if it is a Register message,
3642      * pass the 'head' only up to the daemon. This includes the
3643      * outer IP header, PIM header, PIM-Register header and the
3644      * inner IP header.
3645      * XXX: the outer IP header pkt size of a Register is not adjust to
3646      * reflect the fact that the inner multicast data is truncated.
3647      */
3648     rip_input(m, iphlen, proto);
3649 
3650     return;
3651 }
3652 #endif /* PIM */
3653