xref: /openbsd-src/sys/netinet/ip_mroute.c (revision a28daedfc357b214be5c701aa8ba8adb29a7f1c2)
1 /*	$OpenBSD: ip_mroute.c,v 1.52 2008/09/16 21:33:37 chl Exp $	*/
2 /*	$NetBSD: ip_mroute.c,v 1.85 2004/04/26 01:31:57 matt Exp $	*/
3 
4 /*
5  * Copyright (c) 1989 Stephen Deering
6  * Copyright (c) 1992, 1993
7  *      The Regents of the University of California.  All rights reserved.
8  *
9  * This code is derived from software contributed to Berkeley by
10  * Stephen Deering of Stanford University.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
37  */
38 
39 /*
40  * IP multicast forwarding procedures
41  *
42  * Written by David Waitzman, BBN Labs, August 1988.
43  * Modified by Steve Deering, Stanford, February 1989.
44  * Modified by Mark J. Steiglitz, Stanford, May, 1991
45  * Modified by Van Jacobson, LBL, January 1993
46  * Modified by Ajit Thyagarajan, PARC, August 1993
47  * Modified by Bill Fenner, PARC, April 1994
48  * Modified by Charles M. Hannum, NetBSD, May 1995.
49  * Modified by Ahmed Helmy, SGI, June 1996
50  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
51  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
52  * Modified by Hitoshi Asaeda, WIDE, August 2000
53  * Modified by Pavlin Radoslavov, ICSI, October 2002
54  *
55  * MROUTING Revision: 1.2
56  * and PIM-SMv2 and PIM-DM support, advanced API support,
57  * bandwidth metering and signaling
58  */
59 
60 #ifdef PIM
61 #define _PIM_VT 1
62 #endif
63 
64 #include <sys/param.h>
65 #include <sys/systm.h>
66 #include <sys/mbuf.h>
67 #include <sys/socket.h>
68 #include <sys/socketvar.h>
69 #include <sys/protosw.h>
70 #include <sys/errno.h>
71 #include <sys/time.h>
72 #include <sys/kernel.h>
73 #include <sys/ioctl.h>
74 #include <sys/syslog.h>
75 #include <sys/sysctl.h>
76 #include <sys/timeout.h>
77 
78 #include <net/if.h>
79 #include <net/route.h>
80 #include <net/raw_cb.h>
81 
82 #include <netinet/in.h>
83 #include <netinet/in_var.h>
84 #include <netinet/in_systm.h>
85 #include <netinet/ip.h>
86 #include <netinet/ip_var.h>
87 #include <netinet/in_pcb.h>
88 #include <netinet/udp.h>
89 #include <netinet/igmp.h>
90 #include <netinet/igmp_var.h>
91 #include <netinet/ip_mroute.h>
92 #ifdef PIM
93 #include <netinet/pim.h>
94 #include <netinet/pim_var.h>
95 #endif
96 
97 #include <sys/stdarg.h>
98 
99 #define IP_MULTICASTOPTS 0
100 #define	M_PULLUP(m, len)						 \
101 	do {								 \
102 		if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
103 			(m) = m_pullup((m), (len));			 \
104 	} while (/*CONSTCOND*/ 0)
105 
106 /*
107  * Globals.  All but ip_mrouter and ip_mrtproto could be static,
108  * except for netstat or debugging purposes.
109  */
110 struct socket  *ip_mrouter  = NULL;
111 int		ip_mrtproto = IGMP_DVMRP;    /* for netstat only */
112 
113 #define NO_RTE_FOUND	0x1
114 #define RTE_FOUND	0x2
115 
116 #define	MFCHASH(a, g)							\
117 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^	\
118 	    ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
119 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
120 u_long	mfchash;
121 
122 u_char		nexpire[MFCTBLSIZ];
123 struct vif	viftable[MAXVIFS];
124 struct mrtstat	mrtstat;
125 u_int		mrtdebug = 0;	  /* debug level 	*/
126 #define		DEBUG_MFC	0x02
127 #define		DEBUG_FORWARD	0x04
128 #define		DEBUG_EXPIRE	0x08
129 #define		DEBUG_XMIT	0x10
130 #define		DEBUG_PIM	0x20
131 
132 #define		VIFI_INVALID	((vifi_t) -1)
133 
134 u_int		tbfdebug = 0;     /* tbf debug level 	*/
135 #ifdef RSVP_ISI
136 u_int		rsvpdebug = 0;	  /* rsvp debug level   */
137 extern struct socket *ip_rsvpd;
138 extern int rsvp_on;
139 #endif /* RSVP_ISI */
140 
141 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
142 #define		UPCALL_EXPIRE	6		/* number of timeouts */
143 struct timeout	expire_upcalls_ch;
144 
145 /*
146  * Define the token bucket filter structures
147  */
148 
149 #define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
150 
151 static int get_sg_cnt(struct sioc_sg_req *);
152 static int get_vif_cnt(struct sioc_vif_req *);
153 static int ip_mrouter_init(struct socket *, struct mbuf *);
154 static int get_version(struct mbuf *);
155 static int set_assert(struct mbuf *);
156 static int get_assert(struct mbuf *);
157 static int add_vif(struct mbuf *);
158 static int del_vif(struct mbuf *);
159 static void update_mfc_params(struct mfc *, struct mfcctl2 *);
160 static void init_mfc_params(struct mfc *, struct mfcctl2 *);
161 static void expire_mfc(struct mfc *);
162 static int add_mfc(struct mbuf *);
163 #ifdef UPCALL_TIMING
164 static void collate(struct timeval *);
165 #endif
166 static int del_mfc(struct mbuf *);
167 static int set_api_config(struct mbuf *); /* chose API capabilities */
168 static int get_api_support(struct mbuf *);
169 static int get_api_config(struct mbuf *);
170 static int socket_send(struct socket *, struct mbuf *,
171 			    struct sockaddr_in *);
172 static void expire_upcalls(void *);
173 #ifdef RSVP_ISI
174 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
175 #else
176 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *);
177 #endif
178 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
179 static void encap_send(struct ip *, struct vif *, struct mbuf *);
180 static void tbf_control(struct vif *, struct mbuf *, struct ip *,
181 			     u_int32_t);
182 static void tbf_queue(struct vif *, struct mbuf *);
183 static void tbf_process_q(struct vif *);
184 static void tbf_reprocess_q(void *);
185 static int tbf_dq_sel(struct vif *, struct ip *);
186 static void tbf_send_packet(struct vif *, struct mbuf *);
187 static void tbf_update_tokens(struct vif *);
188 static int priority(struct vif *, struct ip *);
189 
190 /*
191  * Bandwidth monitoring
192  */
193 static void free_bw_list(struct bw_meter *);
194 static int add_bw_upcall(struct mbuf *);
195 static int del_bw_upcall(struct mbuf *);
196 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *);
197 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
198 static void bw_upcalls_send(void);
199 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
200 static void unschedule_bw_meter(struct bw_meter *);
201 static void bw_meter_process(void);
202 static void expire_bw_upcalls_send(void *);
203 static void expire_bw_meter_process(void *);
204 
205 #ifdef PIM
206 static int pim_register_send(struct ip *, struct vif *,
207 		struct mbuf *, struct mfc *);
208 static int pim_register_send_rp(struct ip *, struct vif *,
209 		struct mbuf *, struct mfc *);
210 static int pim_register_send_upcall(struct ip *, struct vif *,
211 		struct mbuf *, struct mfc *);
212 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
213 #endif
214 
215 /*
216  * 'Interfaces' associated with decapsulator (so we can tell
217  * packets that went through it from ones that get reflected
218  * by a broken gateway).  These interfaces are never linked into
219  * the system ifnet list & no routes point to them.  I.e., packets
220  * can't be sent this way.  They only exist as a placeholder for
221  * multicast source verification.
222  */
223 #if 0
224 struct ifnet multicast_decap_if[MAXVIFS];
225 #endif
226 
227 #define	ENCAP_TTL	64
228 #define	ENCAP_PROTO	IPPROTO_IPIP	/* 4 */
229 
230 /* prototype IP hdr for encapsulated packets */
231 struct ip multicast_encap_iphdr = {
232 #if BYTE_ORDER == LITTLE_ENDIAN
233 	sizeof(struct ip) >> 2, IPVERSION,
234 #else
235 	IPVERSION, sizeof(struct ip) >> 2,
236 #endif
237 	0,				/* tos */
238 	sizeof(struct ip),		/* total length */
239 	0,				/* id */
240 	0,				/* frag offset */
241 	ENCAP_TTL, ENCAP_PROTO,
242 	0,				/* checksum */
243 };
244 
245 /*
246  * Bandwidth meter variables and constants
247  */
248 
249 /*
250  * Pending timeouts are stored in a hash table, the key being the
251  * expiration time. Periodically, the entries are analysed and processed.
252  */
253 #define BW_METER_BUCKETS	1024
254 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
255 struct timeout bw_meter_ch;
256 #define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
257 
258 /*
259  * Pending upcalls are stored in a vector which is flushed when
260  * full, or periodically
261  */
262 static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
263 static u_int	bw_upcalls_n; /* # of pending upcalls */
264 struct timeout	bw_upcalls_ch;
265 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
266 
267 #ifdef PIM
268 struct pimstat pimstat;
269 
270 /*
271  * Note: the PIM Register encapsulation adds the following in front of a
272  * data packet:
273  *
274  * struct pim_encap_hdr {
275  *    struct ip ip;
276  *    struct pim_encap_pimhdr  pim;
277  * }
278  *
279  */
280 
281 struct pim_encap_pimhdr {
282 	struct pim pim;
283 	uint32_t   flags;
284 };
285 
286 static struct ip pim_encap_iphdr = {
287 #if BYTE_ORDER == LITTLE_ENDIAN
288 	sizeof(struct ip) >> 2,
289 	IPVERSION,
290 #else
291 	IPVERSION,
292 	sizeof(struct ip) >> 2,
293 #endif
294 	0,			/* tos */
295 	sizeof(struct ip),	/* total length */
296 	0,			/* id */
297 	0,			/* frag offset */
298 	ENCAP_TTL,
299 	IPPROTO_PIM,
300 	0,			/* checksum */
301 };
302 
303 static struct pim_encap_pimhdr pim_encap_pimhdr = {
304     {
305 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
306 	0,			/* reserved */
307 	0,			/* checksum */
308     },
309     0				/* flags */
310 };
311 
312 static struct ifnet multicast_register_if;
313 static vifi_t reg_vif_num = VIFI_INVALID;
314 #endif /* PIM */
315 
316 
317 /*
318  * Private variables.
319  */
320 static vifi_t	   numvifs = 0;
321 static int have_encap_tunnel = 0;
322 
323 /*
324  * whether or not special PIM assert processing is enabled.
325  */
326 static int pim_assert;
327 /*
328  * Rate limit for assert notification messages, in usec
329  */
330 #define ASSERT_MSG_TIME		3000000
331 
332 /*
333  * Kernel multicast routing API capabilities and setup.
334  * If more API capabilities are added to the kernel, they should be
335  * recorded in `mrt_api_support'.
336  */
337 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
338 					  MRT_MFC_FLAGS_BORDER_VIF |
339 					  MRT_MFC_RP |
340 					  MRT_MFC_BW_UPCALL);
341 static u_int32_t mrt_api_config = 0;
342 
343 /*
344  * Find a route for a given origin IP address and Multicast group address
345  * Type of service parameter to be added in the future!!!
346  * Statistics are updated by the caller if needed
347  * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
348  */
349 static struct mfc *
350 mfc_find(struct in_addr *o, struct in_addr *g)
351 {
352 	struct mfc *rt;
353 
354 	LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
355 		if (in_hosteq(rt->mfc_origin, *o) &&
356 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
357 		    (rt->mfc_stall == NULL))
358 			break;
359 	}
360 
361 	return (rt);
362 }
363 
364 /*
365  * Macros to compute elapsed time efficiently
366  * Borrowed from Van Jacobson's scheduling code
367  */
368 #define TV_DELTA(a, b, delta) do {					\
369 	int xxs;							\
370 	delta = (a).tv_usec - (b).tv_usec;				\
371 	xxs = (a).tv_sec - (b).tv_sec;					\
372 	switch (xxs) {							\
373 	case 2:								\
374 		delta += 1000000;					\
375 		/* FALLTHROUGH */					\
376 	case 1:								\
377 		delta += 1000000;					\
378 		/* FALLTHROUGH */					\
379 	case 0:								\
380 		break;							\
381 	default:							\
382 		delta += (1000000 * xxs);				\
383 		break;							\
384 	}								\
385 } while (/*CONSTCOND*/ 0)
386 
387 #ifdef UPCALL_TIMING
388 u_int32_t upcall_data[51];
389 #endif /* UPCALL_TIMING */
390 
391 /*
392  * Handle MRT setsockopt commands to modify the multicast routing tables.
393  */
394 int
395 ip_mrouter_set(struct socket *so, int optname, struct mbuf **m)
396 {
397 	int error;
398 
399 	if (optname != MRT_INIT && so != ip_mrouter)
400 		error = ENOPROTOOPT;
401 	else
402 		switch (optname) {
403 		case MRT_INIT:
404 			error = ip_mrouter_init(so, *m);
405 			break;
406 		case MRT_DONE:
407 			error = ip_mrouter_done();
408 			break;
409 		case MRT_ADD_VIF:
410 			error = add_vif(*m);
411 			break;
412 		case MRT_DEL_VIF:
413 			error = del_vif(*m);
414 			break;
415 		case MRT_ADD_MFC:
416 			error = add_mfc(*m);
417 			break;
418 		case MRT_DEL_MFC:
419 			error = del_mfc(*m);
420 			break;
421 		case MRT_ASSERT:
422 			error = set_assert(*m);
423 			break;
424 		case MRT_API_CONFIG:
425 			error = set_api_config(*m);
426 			break;
427 		case MRT_ADD_BW_UPCALL:
428 			error = add_bw_upcall(*m);
429 			break;
430 		case MRT_DEL_BW_UPCALL:
431 			error = del_bw_upcall(*m);
432 			break;
433 		default:
434 			error = ENOPROTOOPT;
435 			break;
436 		}
437 
438 	if (*m)
439 		m_free(*m);
440 	return (error);
441 }
442 
443 /*
444  * Handle MRT getsockopt commands
445  */
446 int
447 ip_mrouter_get(struct socket *so, int optname, struct mbuf **m)
448 {
449 	int error;
450 
451 	if (so != ip_mrouter)
452 		error = ENOPROTOOPT;
453 	else {
454 		*m = m_get(M_WAIT, MT_SOOPTS);
455 
456 		switch (optname) {
457 		case MRT_VERSION:
458 			error = get_version(*m);
459 			break;
460 		case MRT_ASSERT:
461 			error = get_assert(*m);
462 			break;
463 		case MRT_API_SUPPORT:
464 			error = get_api_support(*m);
465 			break;
466 		case MRT_API_CONFIG:
467 			error = get_api_config(*m);
468 			break;
469 		default:
470 			error = ENOPROTOOPT;
471 			break;
472 		}
473 
474 		if (error)
475 			m_free(*m);
476 	}
477 
478 	return (error);
479 }
480 
481 /*
482  * Handle ioctl commands to obtain information from the cache
483  */
484 int
485 mrt_ioctl(struct socket *so, u_long cmd, caddr_t data)
486 {
487 	int error;
488 
489 	if (so != ip_mrouter)
490 		error = EINVAL;
491 	else
492 		switch (cmd) {
493 		case SIOCGETVIFCNT:
494 			error = get_vif_cnt((struct sioc_vif_req *)data);
495 			break;
496 		case SIOCGETSGCNT:
497 			error = get_sg_cnt((struct sioc_sg_req *)data);
498 			break;
499 		default:
500 			error = ENOTTY;
501 			break;
502 		}
503 
504 	return (error);
505 }
506 
507 /*
508  * returns the packet, byte, rpf-failure count for the source group provided
509  */
510 static int
511 get_sg_cnt(struct sioc_sg_req *req)
512 {
513 	int s;
514 	struct mfc *rt;
515 
516 	s = splsoftnet();
517 	rt = mfc_find(&req->src, &req->grp);
518 	if (rt == NULL) {
519 		splx(s);
520 		req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
521 		return (EADDRNOTAVAIL);
522 	}
523 	req->pktcnt = rt->mfc_pkt_cnt;
524 	req->bytecnt = rt->mfc_byte_cnt;
525 	req->wrong_if = rt->mfc_wrong_if;
526 	splx(s);
527 
528 	return (0);
529 }
530 
531 /*
532  * returns the input and output packet and byte counts on the vif provided
533  */
534 static int
535 get_vif_cnt(struct sioc_vif_req *req)
536 {
537 	vifi_t vifi = req->vifi;
538 
539 	if (vifi >= numvifs)
540 		return (EINVAL);
541 
542 	req->icount = viftable[vifi].v_pkt_in;
543 	req->ocount = viftable[vifi].v_pkt_out;
544 	req->ibytes = viftable[vifi].v_bytes_in;
545 	req->obytes = viftable[vifi].v_bytes_out;
546 
547 	return (0);
548 }
549 
550 /*
551  * Enable multicast routing
552  */
553 static int
554 ip_mrouter_init(struct socket *so, struct mbuf *m)
555 {
556 	int *v;
557 
558 	if (mrtdebug)
559 		log(LOG_DEBUG,
560 		    "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
561 		    so->so_type, so->so_proto->pr_protocol);
562 
563 	if (so->so_type != SOCK_RAW ||
564 	    so->so_proto->pr_protocol != IPPROTO_IGMP)
565 		return (EOPNOTSUPP);
566 
567 	if (m == NULL || m->m_len < sizeof(int))
568 		return (EINVAL);
569 
570 	v = mtod(m, int *);
571 	if (*v != 1)
572 		return (EINVAL);
573 
574 	if (ip_mrouter != NULL)
575 		return (EADDRINUSE);
576 
577 	ip_mrouter = so;
578 
579 	mfchashtbl = hashinit(MFCTBLSIZ, M_MRTABLE, M_WAITOK, &mfchash);
580 	bzero((caddr_t)nexpire, sizeof(nexpire));
581 
582 	pim_assert = 0;
583 
584 	timeout_set(&expire_upcalls_ch, expire_upcalls, NULL);
585 	timeout_add(&expire_upcalls_ch, EXPIRE_TIMEOUT);
586 
587 	timeout_set(&bw_upcalls_ch, expire_bw_upcalls_send, NULL);
588 	timeout_add(&bw_upcalls_ch, BW_UPCALLS_PERIOD);
589 
590 	timeout_set(&bw_meter_ch, expire_bw_meter_process, NULL);
591 	timeout_add(&bw_meter_ch, BW_METER_PERIOD);
592 
593 	if (mrtdebug)
594 		log(LOG_DEBUG, "ip_mrouter_init\n");
595 
596 	return (0);
597 }
598 
599 /*
600  * Disable multicast routing
601  */
602 int
603 ip_mrouter_done()
604 {
605 	vifi_t vifi;
606 	struct vif *vifp;
607 	int i;
608 	int s;
609 
610 	s = splsoftnet();
611 
612 	/* Clear out all the vifs currently in use. */
613 	for (vifi = 0; vifi < numvifs; vifi++) {
614 		vifp = &viftable[vifi];
615 		if (!in_nullhost(vifp->v_lcl_addr))
616 			reset_vif(vifp);
617 	}
618 
619 	numvifs = 0;
620 	pim_assert = 0;
621 	mrt_api_config = 0;
622 
623 	timeout_del(&expire_upcalls_ch);
624 	timeout_del(&bw_upcalls_ch);
625 	timeout_del(&bw_meter_ch);
626 
627 	/*
628 	 * Free all multicast forwarding cache entries.
629 	 */
630 	for (i = 0; i < MFCTBLSIZ; i++) {
631 		struct mfc *rt, *nrt;
632 
633 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
634 			nrt = LIST_NEXT(rt, mfc_hash);
635 
636 			expire_mfc(rt);
637 		}
638 	}
639 
640 	bzero((caddr_t)nexpire, sizeof(nexpire));
641 	free(mfchashtbl, M_MRTABLE);
642 	mfchashtbl = NULL;
643 
644 	bw_upcalls_n = 0;
645 	bzero(bw_meter_timers, sizeof(bw_meter_timers));
646 
647 	/* Reset de-encapsulation cache. */
648 	have_encap_tunnel = 0;
649 
650 	ip_mrouter = NULL;
651 
652 	splx(s);
653 
654 	if (mrtdebug)
655 		log(LOG_DEBUG, "ip_mrouter_done\n");
656 
657 	return (0);
658 }
659 
660 void
661 ip_mrouter_detach(struct ifnet *ifp)
662 {
663 	int vifi, i;
664 	struct vif *vifp;
665 	struct mfc *rt;
666 	struct rtdetq *rte;
667 
668 	/* XXX not sure about side effect to userland routing daemon */
669 	for (vifi = 0; vifi < numvifs; vifi++) {
670 		vifp = &viftable[vifi];
671 		if (vifp->v_ifp == ifp)
672 			reset_vif(vifp);
673 	}
674 	for (i = 0; i < MFCTBLSIZ; i++) {
675 		if (nexpire[i] == 0)
676 			continue;
677 		LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
678 			for (rte = rt->mfc_stall; rte; rte = rte->next) {
679 				if (rte->ifp == ifp)
680 					rte->ifp = NULL;
681 			}
682 		}
683 	}
684 }
685 
686 static int
687 get_version(struct mbuf *m)
688 {
689 	int *v = mtod(m, int *);
690 
691 	*v = 0x0305;	/* XXX !!!! */
692 	m->m_len = sizeof(int);
693 	return (0);
694 }
695 
696 /*
697  * Set PIM assert processing global
698  */
699 static int
700 set_assert(struct mbuf *m)
701 {
702 	int *i;
703 
704 	if (m == NULL || m->m_len < sizeof(int))
705 		return (EINVAL);
706 
707 	i = mtod(m, int *);
708 	pim_assert = !!*i;
709 	return (0);
710 }
711 
712 /*
713  * Get PIM assert processing global
714  */
715 static int
716 get_assert(struct mbuf *m)
717 {
718 	int *i = mtod(m, int *);
719 
720 	*i = pim_assert;
721 	m->m_len = sizeof(int);
722 	return (0);
723 }
724 
725 /*
726  * Configure API capabilities
727  */
728 static int
729 set_api_config(struct mbuf *m)
730 {
731 	int i;
732 	u_int32_t *apival;
733 
734 	if (m == NULL || m->m_len < sizeof(u_int32_t))
735 		return (EINVAL);
736 
737 	apival = mtod(m, u_int32_t *);
738 
739 	/*
740 	 * We can set the API capabilities only if it is the first operation
741 	 * after MRT_INIT. I.e.:
742 	 *  - there are no vifs installed
743 	 *  - pim_assert is not enabled
744 	 *  - the MFC table is empty
745 	 */
746 	if (numvifs > 0) {
747 		*apival = 0;
748 		return (EPERM);
749 	}
750 	if (pim_assert) {
751 		*apival = 0;
752 		return (EPERM);
753 	}
754 	for (i = 0; i < MFCTBLSIZ; i++) {
755 		if (LIST_FIRST(&mfchashtbl[i]) != NULL) {
756 			*apival = 0;
757 			return (EPERM);
758 		}
759 	}
760 
761 	mrt_api_config = *apival & mrt_api_support;
762 	*apival = mrt_api_config;
763 
764 	return (0);
765 }
766 
767 /*
768  * Get API capabilities
769  */
770 static int
771 get_api_support(struct mbuf *m)
772 {
773 	u_int32_t *apival;
774 
775 	if (m == NULL || m->m_len < sizeof(u_int32_t))
776 		return (EINVAL);
777 
778 	apival = mtod(m, u_int32_t *);
779 
780 	*apival = mrt_api_support;
781 
782 	return (0);
783 }
784 
785 /*
786  * Get API configured capabilities
787  */
788 static int
789 get_api_config(struct mbuf *m)
790 {
791 	u_int32_t *apival;
792 
793 	if (m == NULL || m->m_len < sizeof(u_int32_t))
794 		return (EINVAL);
795 
796 	apival = mtod(m, u_int32_t *);
797 
798 	*apival = mrt_api_config;
799 
800 	return (0);
801 }
802 
803 static struct sockaddr_in sin = { sizeof(sin), AF_INET };
804 
805 /*
806  * Add a vif to the vif table
807  */
808 static int
809 add_vif(struct mbuf *m)
810 {
811 	struct vifctl *vifcp;
812 	struct vif *vifp;
813 	struct ifaddr *ifa;
814 	struct ifnet *ifp;
815 	struct ifreq ifr;
816 	int error, s;
817 
818 	if (m == NULL || m->m_len < sizeof(struct vifctl))
819 		return (EINVAL);
820 
821 	vifcp = mtod(m, struct vifctl *);
822 	if (vifcp->vifc_vifi >= MAXVIFS)
823 		return (EINVAL);
824 	if (in_nullhost(vifcp->vifc_lcl_addr))
825 		return (EADDRNOTAVAIL);
826 
827 	vifp = &viftable[vifcp->vifc_vifi];
828 	if (!in_nullhost(vifp->v_lcl_addr))
829 		return (EADDRINUSE);
830 
831 	/* Find the interface with an address in AF_INET family. */
832 #ifdef PIM
833 	if (vifcp->vifc_flags & VIFF_REGISTER) {
834 		/*
835 		 * XXX: Because VIFF_REGISTER does not really need a valid
836 		 * local interface (e.g. it could be 127.0.0.2), we don't
837 		 * check its address.
838 		 */
839 	} else
840 #endif
841 	{
842 		sin.sin_addr = vifcp->vifc_lcl_addr;
843 		ifa = ifa_ifwithaddr(sintosa(&sin));
844 		if (ifa == NULL)
845 			return (EADDRNOTAVAIL);
846 	}
847 
848 	if (vifcp->vifc_flags & VIFF_TUNNEL) {
849 		/* tunnels are no longer supported use gif(4) instead */
850 		return (EOPNOTSUPP);
851 #ifdef PIM
852 	} else if (vifcp->vifc_flags & VIFF_REGISTER) {
853 		ifp = &multicast_register_if;
854 		if (mrtdebug)
855 			log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
856 			    (void *)ifp);
857 		if (reg_vif_num == VIFI_INVALID) {
858 			bzero(ifp, sizeof(*ifp));
859 			snprintf(ifp->if_xname, sizeof ifp->if_xname,
860 				 "register_vif");
861 			ifp->if_flags = IFF_LOOPBACK;
862 			bzero(&vifp->v_route, sizeof(vifp->v_route));
863 			reg_vif_num = vifcp->vifc_vifi;
864 		}
865 #endif
866 	} else {
867 		/* Use the physical interface associated with the address. */
868 		ifp = ifa->ifa_ifp;
869 
870 		/* Make sure the interface supports multicast. */
871 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
872 			return (EOPNOTSUPP);
873 
874 		/* Enable promiscuous reception of all IP multicasts. */
875 		satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
876 		satosin(&ifr.ifr_addr)->sin_family = AF_INET;
877 		satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
878 		error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
879 		if (error)
880 			return (error);
881 	}
882 
883 	s = splsoftnet();
884 
885 	/* Define parameters for the tbf structure. */
886 	vifp->tbf_q = NULL;
887 	vifp->tbf_t = &vifp->tbf_q;
888 	microtime(&vifp->tbf_last_pkt_t);
889 	vifp->tbf_n_tok = 0;
890 	vifp->tbf_q_len = 0;
891 	vifp->tbf_max_q_len = MAXQSIZE;
892 
893 	vifp->v_flags = vifcp->vifc_flags;
894 	vifp->v_threshold = vifcp->vifc_threshold;
895 	/* scaling up here allows division by 1024 in critical code */
896 	vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
897 	vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
898 	vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
899 	vifp->v_ifp = ifp;
900 	/* Initialize per vif pkt counters. */
901 	vifp->v_pkt_in = 0;
902 	vifp->v_pkt_out = 0;
903 	vifp->v_bytes_in = 0;
904 	vifp->v_bytes_out = 0;
905 
906 	timeout_del(&vifp->v_repq_ch);
907 
908 #ifdef RSVP_ISI
909 	vifp->v_rsvp_on = 0;
910 	vifp->v_rsvpd = NULL;
911 #endif /* RSVP_ISI */
912 
913 	splx(s);
914 
915 	/* Adjust numvifs up if the vifi is higher than numvifs. */
916 	if (numvifs <= vifcp->vifc_vifi)
917 		numvifs = vifcp->vifc_vifi + 1;
918 
919 	if (mrtdebug)
920 		log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, "
921 		    "thresh %x, rate %d\n",
922 		    vifcp->vifc_vifi,
923 		    ntohl(vifcp->vifc_lcl_addr.s_addr),
924 		    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
925 		    ntohl(vifcp->vifc_rmt_addr.s_addr),
926 		    vifcp->vifc_threshold,
927 		    vifcp->vifc_rate_limit);
928 
929 	return (0);
930 }
931 
932 void
933 reset_vif(struct vif *vifp)
934 {
935 	struct mbuf *m, *n;
936 	struct ifnet *ifp;
937 	struct ifreq ifr;
938 
939 	timeout_set(&vifp->v_repq_ch, tbf_reprocess_q, vifp);
940 
941 	/*
942 	 * Free packets queued at the interface
943 	 */
944 	for (m = vifp->tbf_q; m != NULL; m = n) {
945 		n = m->m_nextpkt;
946 		m_freem(m);
947 	}
948 
949 	if (vifp->v_flags & VIFF_TUNNEL) {
950 		/* empty */
951 	} else if (vifp->v_flags & VIFF_REGISTER) {
952 #ifdef PIM
953 		reg_vif_num = VIFI_INVALID;
954 #endif
955 	} else {
956 		satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
957 		satosin(&ifr.ifr_addr)->sin_family = AF_INET;
958 		satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
959 		ifp = vifp->v_ifp;
960 		(*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
961 	}
962 	bzero((caddr_t)vifp, sizeof(*vifp));
963 }
964 
965 /*
966  * Delete a vif from the vif table
967  */
968 static int
969 del_vif(struct mbuf *m)
970 {
971 	vifi_t *vifip;
972 	struct vif *vifp;
973 	vifi_t vifi;
974 	int s;
975 
976 	if (m == NULL || m->m_len < sizeof(vifi_t))
977 		return (EINVAL);
978 
979 	vifip = mtod(m, vifi_t *);
980 	if (*vifip >= numvifs)
981 		return (EINVAL);
982 
983 	vifp = &viftable[*vifip];
984 	if (in_nullhost(vifp->v_lcl_addr))
985 		return (EADDRNOTAVAIL);
986 
987 	s = splsoftnet();
988 
989 	reset_vif(vifp);
990 
991 	/* Adjust numvifs down */
992 	for (vifi = numvifs; vifi > 0; vifi--)
993 		if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
994 			break;
995 	numvifs = vifi;
996 
997 	splx(s);
998 
999 	if (mrtdebug)
1000 		log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
1001 
1002 	return (0);
1003 }
1004 
1005 void
1006 vif_delete(struct ifnet *ifp)
1007 {
1008 	int i;
1009 	struct vif *vifp;
1010 	struct mfc *rt;
1011 	struct rtdetq *rte;
1012 
1013 	for (i = 0; i < numvifs; i++) {
1014 		vifp = &viftable[i];
1015 		if (vifp->v_ifp == ifp)
1016 			bzero((caddr_t)vifp, sizeof *vifp);
1017 	}
1018 
1019 	for (i = numvifs; i > 0; i--)
1020 		if (!in_nullhost(viftable[i - 1].v_lcl_addr))
1021 			break;
1022 	numvifs = i;
1023 
1024 	for (i = 0; i < MFCTBLSIZ; i++) {
1025 		if (nexpire[i] == 0)
1026 			continue;
1027 		LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
1028 			for (rte = rt->mfc_stall; rte; rte = rte->next) {
1029 				if (rte->ifp == ifp)
1030 					rte->ifp = NULL;
1031 			}
1032 		}
1033 	}
1034 }
1035 
1036 /*
1037  * update an mfc entry without resetting counters and S,G addresses.
1038  */
1039 static void
1040 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1041 {
1042 	int i;
1043 
1044 	rt->mfc_parent = mfccp->mfcc_parent;
1045 	for (i = 0; i < numvifs; i++) {
1046 		rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1047 		rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1048 		    MRT_MFC_FLAGS_ALL;
1049 	}
1050 	/* set the RP address */
1051 	if (mrt_api_config & MRT_MFC_RP)
1052 		rt->mfc_rp = mfccp->mfcc_rp;
1053 	else
1054 		rt->mfc_rp = zeroin_addr;
1055 }
1056 
1057 /*
1058  * fully initialize an mfc entry from the parameter.
1059  */
1060 static void
1061 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1062 {
1063 	rt->mfc_origin     = mfccp->mfcc_origin;
1064 	rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1065 
1066 	update_mfc_params(rt, mfccp);
1067 
1068 	/* initialize pkt counters per src-grp */
1069 	rt->mfc_pkt_cnt    = 0;
1070 	rt->mfc_byte_cnt   = 0;
1071 	rt->mfc_wrong_if   = 0;
1072 	timerclear(&rt->mfc_last_assert);
1073 }
1074 
1075 static void
1076 expire_mfc(struct mfc *rt)
1077 {
1078 	struct rtdetq *rte, *nrte;
1079 
1080 	free_bw_list(rt->mfc_bw_meter);
1081 
1082 	for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1083 		nrte = rte->next;
1084 		m_freem(rte->m);
1085 		free(rte, M_MRTABLE);
1086 	}
1087 
1088 	LIST_REMOVE(rt, mfc_hash);
1089 	free(rt, M_MRTABLE);
1090 }
1091 
1092 /*
1093  * Add an mfc entry
1094  */
1095 static int
1096 add_mfc(struct mbuf *m)
1097 {
1098 	struct mfcctl2 mfcctl2;
1099 	struct mfcctl2 *mfccp;
1100 	struct mfc *rt;
1101 	u_int32_t hash = 0;
1102 	struct rtdetq *rte, *nrte;
1103 	u_short nstl;
1104 	int s;
1105 	int mfcctl_size = sizeof(struct mfcctl);
1106 
1107 	if (mrt_api_config & MRT_API_FLAGS_ALL)
1108 		mfcctl_size = sizeof(struct mfcctl2);
1109 
1110 	if (m == NULL || m->m_len < mfcctl_size)
1111 		return (EINVAL);
1112 
1113 	/*
1114 	 * select data size depending on API version.
1115 	 */
1116 	if (mrt_api_config & MRT_API_FLAGS_ALL) {
1117 		struct mfcctl2 *mp2 = mtod(m, struct mfcctl2 *);
1118 		bcopy(mp2, (caddr_t)&mfcctl2, sizeof(*mp2));
1119 	} else {
1120 		struct mfcctl *mp = mtod(m, struct mfcctl *);
1121 		bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp));
1122 		bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl),
1123 		    sizeof(mfcctl2) - sizeof(struct mfcctl));
1124 	}
1125 	mfccp = &mfcctl2;
1126 
1127 	s = splsoftnet();
1128 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1129 
1130 	/* If an entry already exists, just update the fields */
1131 	if (rt) {
1132 		if (mrtdebug & DEBUG_MFC)
1133 			log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1134 			    ntohl(mfccp->mfcc_origin.s_addr),
1135 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1136 			    mfccp->mfcc_parent);
1137 
1138 		update_mfc_params(rt, mfccp);
1139 
1140 		splx(s);
1141 		return (0);
1142 	}
1143 
1144 	/*
1145 	 * Find the entry for which the upcall was made and update
1146 	 */
1147 	nstl = 0;
1148 	hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1149 	LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1150 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1151 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1152 		    rt->mfc_stall != NULL) {
1153 			if (nstl++)
1154 				log(LOG_ERR, "add_mfc %s o %x g %x "
1155 				    "p %x dbx %p\n",
1156 				    "multiple kernel entries",
1157 				    ntohl(mfccp->mfcc_origin.s_addr),
1158 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1159 				    mfccp->mfcc_parent, rt->mfc_stall);
1160 
1161 			if (mrtdebug & DEBUG_MFC)
1162 				log(LOG_DEBUG, "add_mfc o %x g %x "
1163 				    "p %x dbg %p\n",
1164 				    ntohl(mfccp->mfcc_origin.s_addr),
1165 				    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1166 				    mfccp->mfcc_parent, rt->mfc_stall);
1167 
1168 			rte = rt->mfc_stall;
1169 			init_mfc_params(rt, mfccp);
1170 			rt->mfc_stall = NULL;
1171 
1172 			rt->mfc_expire = 0; /* Don't clean this guy up */
1173 			nexpire[hash]--;
1174 
1175 			/* free packets Qed at the end of this entry */
1176 			for (; rte != NULL; rte = nrte) {
1177 				nrte = rte->next;
1178 				if (rte->ifp) {
1179 #ifdef RSVP_ISI
1180 					ip_mdq(rte->m, rte->ifp, rt, -1);
1181 #else
1182 					ip_mdq(rte->m, rte->ifp, rt);
1183 #endif /* RSVP_ISI */
1184 				}
1185 				m_freem(rte->m);
1186 #ifdef UPCALL_TIMING
1187 				collate(&rte->t);
1188 #endif /* UPCALL_TIMING */
1189 				free(rte, M_MRTABLE);
1190 			}
1191 		}
1192 	}
1193 
1194 	/*
1195 	 * It is possible that an entry is being inserted without an upcall
1196 	 */
1197 	if (nstl == 0) {
1198 		/*
1199 		 * No mfc; make a new one
1200 		 */
1201 		if (mrtdebug & DEBUG_MFC)
1202 			log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1203 			    ntohl(mfccp->mfcc_origin.s_addr),
1204 			    ntohl(mfccp->mfcc_mcastgrp.s_addr),
1205 			    mfccp->mfcc_parent);
1206 
1207 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1208 			if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1209 			    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1210 				init_mfc_params(rt, mfccp);
1211 				if (rt->mfc_expire)
1212 					nexpire[hash]--;
1213 				rt->mfc_expire = 0;
1214 				break; /* XXX */
1215 			}
1216 		}
1217 		if (rt == NULL) {	/* no upcall, so make a new entry */
1218 			rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1219 			    M_NOWAIT);
1220 			if (rt == NULL) {
1221 				splx(s);
1222 				return (ENOBUFS);
1223 			}
1224 
1225 			init_mfc_params(rt, mfccp);
1226 			rt->mfc_expire	= 0;
1227 			rt->mfc_stall	= NULL;
1228 			rt->mfc_bw_meter = NULL;
1229 
1230 			/* insert new entry at head of hash chain */
1231 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1232 		}
1233 	}
1234 
1235 	splx(s);
1236 	return (0);
1237 }
1238 
1239 #ifdef UPCALL_TIMING
1240 /*
1241  * collect delay statistics on the upcalls
1242  */
1243 static void
1244 collate(struct timeval *t)
1245 {
1246 	u_int32_t d;
1247 	struct timeval tp;
1248 	u_int32_t delta;
1249 
1250 	microtime(&tp);
1251 
1252 	if (timercmp(t, &tp, <)) {
1253 		TV_DELTA(tp, *t, delta);
1254 
1255 		d = delta >> 10;
1256 		if (d > 50)
1257 			d = 50;
1258 
1259 		++upcall_data[d];
1260 	}
1261 }
1262 #endif /* UPCALL_TIMING */
1263 
1264 /*
1265  * Delete an mfc entry
1266  */
1267 static int
1268 del_mfc(struct mbuf *m)
1269 {
1270 	struct mfcctl2 mfcctl2;
1271 	struct mfcctl2 *mfccp;
1272 	struct mfc *rt;
1273 	int s;
1274 	int mfcctl_size = sizeof(struct mfcctl);
1275 	struct mfcctl *mp = mtod(m, struct mfcctl *);
1276 
1277 	/*
1278 	 * XXX: for deleting MFC entries the information in entries
1279 	 * of size "struct mfcctl" is sufficient.
1280 	 */
1281 
1282 	if (m == NULL || m->m_len < mfcctl_size)
1283 		return (EINVAL);
1284 
1285 	bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp));
1286 	bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl),
1287 	    sizeof(mfcctl2) - sizeof(struct mfcctl));
1288 
1289 	mfccp = &mfcctl2;
1290 
1291 	if (mrtdebug & DEBUG_MFC)
1292 		log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1293 		    ntohl(mfccp->mfcc_origin.s_addr),
1294 		    ntohl(mfccp->mfcc_mcastgrp.s_addr));
1295 
1296 	s = splsoftnet();
1297 
1298 	rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1299 	if (rt == NULL) {
1300 		splx(s);
1301 		return (EADDRNOTAVAIL);
1302 	}
1303 
1304 	/*
1305 	 * free the bw_meter entries
1306 	 */
1307 	free_bw_list(rt->mfc_bw_meter);
1308 	rt->mfc_bw_meter = NULL;
1309 
1310 	LIST_REMOVE(rt, mfc_hash);
1311 	free(rt, M_MRTABLE);
1312 
1313 	splx(s);
1314 	return (0);
1315 }
1316 
1317 static int
1318 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1319 {
1320 	if (s != NULL) {
1321 		if (sbappendaddr(&s->so_rcv, sintosa(src), mm,
1322 		    (struct mbuf *)NULL) != 0) {
1323 			sorwakeup(s);
1324 			return (0);
1325 		}
1326 	}
1327 	m_freem(mm);
1328 	return (-1);
1329 }
1330 
1331 /*
1332  * IP multicast forwarding function. This function assumes that the packet
1333  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1334  * pointed to by "ifp", and the packet is to be relayed to other networks
1335  * that have members of the packet's destination IP multicast group.
1336  *
1337  * The packet is returned unscathed to the caller, unless it is
1338  * erroneous, in which case a non-zero return value tells the caller to
1339  * discard it.
1340  */
1341 
1342 #define IP_HDR_LEN  20	/* # bytes of fixed IP header (excluding options) */
1343 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1344 
1345 int
1346 #ifdef RSVP_ISI
1347 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo)
1348 #else
1349 ip_mforward(struct mbuf *m, struct ifnet *ifp)
1350 #endif /* RSVP_ISI */
1351 {
1352 	struct ip *ip = mtod(m, struct ip *);
1353 	struct mfc *rt;
1354 	static int srctun = 0;
1355 	struct mbuf *mm;
1356 	int s;
1357 	vifi_t vifi;
1358 
1359 	if (mrtdebug & DEBUG_FORWARD)
1360 		log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1361 		    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1362 
1363 	if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1364 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1365 		/*
1366 		 * Packet arrived via a physical interface or
1367 		 * an encapsulated tunnel or a register_vif.
1368 		 */
1369 	} else {
1370 		/*
1371 		 * Packet arrived through a source-route tunnel.
1372 		 * Source-route tunnels are no longer supported.
1373 		 */
1374 		if ((srctun++ % 1000) == 0)
1375 			log(LOG_ERR, "ip_mforward: received source-routed "
1376 			    "packet from %x\n", ntohl(ip->ip_src.s_addr));
1377 
1378 		return (1);
1379 	}
1380 
1381 #ifdef RSVP_ISI
1382 	if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1383 		if (ip->ip_ttl < 255) {
1384 			/* compensate for -1 in *_send routines */
1385 			ip->ip_ttl++;
1386 		}
1387 		if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1388 			struct vif *vifp = viftable + vifi;
1389 			printf("Sending IPPROTO_RSVP from %x to %x on "
1390 			    "vif %d (%s%s)\n",
1391 			    ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1392 			    (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1393 			    vifp->v_ifp->if_xname);
1394 		}
1395 		return (ip_mdq(m, ifp, (struct mfc *)NULL, vifi));
1396 	}
1397 	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1398 		printf("Warning: IPPROTO_RSVP from %x to %x without "
1399 		    "vif option\n", ntohl(ip->ip_src), ntohl(ip->ip_dst));
1400 	}
1401 #endif /* RSVP_ISI */
1402 
1403 	/*
1404 	 * Don't forward a packet with time-to-live of zero or one,
1405 	 * or a packet destined to a local-only group.
1406 	 */
1407 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1408 		return (0);
1409 
1410 	/*
1411 	 * Determine forwarding vifs from the forwarding cache table
1412 	 */
1413 	s = splsoftnet();
1414 	++mrtstat.mrts_mfc_lookups;
1415 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1416 
1417 	/* Entry exists, so forward if necessary */
1418 	if (rt != NULL) {
1419 		splx(s);
1420 #ifdef RSVP_ISI
1421 		return (ip_mdq(m, ifp, rt, -1));
1422 #else
1423 		return (ip_mdq(m, ifp, rt));
1424 #endif /* RSVP_ISI */
1425 	} else {
1426 		/*
1427 		 * If we don't have a route for packet's origin,
1428 		 * Make a copy of the packet & send message to routing daemon
1429 		 */
1430 
1431 		struct mbuf *mb0;
1432 		struct rtdetq *rte;
1433 		u_int32_t hash;
1434 		int hlen = ip->ip_hl << 2;
1435 #ifdef UPCALL_TIMING
1436 		struct timeval tp;
1437 
1438 		microtime(&tp);
1439 #endif /* UPCALL_TIMING */
1440 
1441 		++mrtstat.mrts_mfc_misses;
1442 
1443 		mrtstat.mrts_no_route++;
1444 		if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1445 			log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1446 			    ntohl(ip->ip_src.s_addr),
1447 			    ntohl(ip->ip_dst.s_addr));
1448 
1449 		/*
1450 		 * Allocate mbufs early so that we don't do extra work if we are
1451 		 * just going to fail anyway.  Make sure to pullup the header so
1452 		 * that other people can't step on it.
1453 		 */
1454 		rte = (struct rtdetq *)malloc(sizeof(*rte),
1455 		    M_MRTABLE, M_NOWAIT);
1456 		if (rte == NULL) {
1457 			splx(s);
1458 			return (ENOBUFS);
1459 		}
1460 		mb0 = m_copy(m, 0, M_COPYALL);
1461 		M_PULLUP(mb0, hlen);
1462 		if (mb0 == NULL) {
1463 			free(rte, M_MRTABLE);
1464 			splx(s);
1465 			return (ENOBUFS);
1466 		}
1467 
1468 		/* is there an upcall waiting for this flow? */
1469 		hash = MFCHASH(ip->ip_src, ip->ip_dst);
1470 		LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1471 			if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1472 			    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1473 			    rt->mfc_stall != NULL)
1474 				break;
1475 		}
1476 
1477 		if (rt == NULL) {
1478 			int i;
1479 			struct igmpmsg *im;
1480 
1481 			/*
1482 			 * Locate the vifi for the incoming interface for
1483 			 * this packet.
1484 			 * If none found, drop packet.
1485 			 */
1486 			for (vifi = 0; vifi < numvifs &&
1487 				 viftable[vifi].v_ifp != ifp; vifi++)
1488 				;
1489 			if (vifi >= numvifs) /* vif not found, drop packet */
1490 				goto non_fatal;
1491 
1492 			/* no upcall, so make a new entry */
1493 			rt = (struct mfc *)malloc(sizeof(*rt),
1494 			    M_MRTABLE, M_NOWAIT);
1495 			if (rt == NULL)
1496 				goto fail;
1497 			/*
1498 			 * Make a copy of the header to send to the user level
1499 			 * process
1500 			 */
1501 			mm = m_copy(m, 0, hlen);
1502 			M_PULLUP(mm, hlen);
1503 			if (mm == NULL)
1504 				goto fail1;
1505 
1506 			/*
1507 			 * Send message to routing daemon to install
1508 			 * a route into the kernel table
1509 			 */
1510 
1511 			im = mtod(mm, struct igmpmsg *);
1512 			im->im_msgtype = IGMPMSG_NOCACHE;
1513 			im->im_mbz = 0;
1514 			im->im_vif = vifi;
1515 
1516 			mrtstat.mrts_upcalls++;
1517 
1518 			sin.sin_addr = ip->ip_src;
1519 			if (socket_send(ip_mrouter, mm, &sin) < 0) {
1520 				log(LOG_WARNING, "ip_mforward: ip_mrouter "
1521 				    "socket queue full\n");
1522 				++mrtstat.mrts_upq_sockfull;
1523 			fail1:
1524 				free(rt, M_MRTABLE);
1525 			fail:
1526 				free(rte, M_MRTABLE);
1527 				m_freem(mb0);
1528 				splx(s);
1529 				return (ENOBUFS);
1530 			}
1531 
1532 			/* insert new entry at head of hash chain */
1533 			rt->mfc_origin = ip->ip_src;
1534 			rt->mfc_mcastgrp = ip->ip_dst;
1535 			rt->mfc_pkt_cnt = 0;
1536 			rt->mfc_byte_cnt = 0;
1537 			rt->mfc_wrong_if = 0;
1538 			rt->mfc_expire = UPCALL_EXPIRE;
1539 			nexpire[hash]++;
1540 			for (i = 0; i < numvifs; i++) {
1541 				rt->mfc_ttls[i] = 0;
1542 				rt->mfc_flags[i] = 0;
1543 			}
1544 			rt->mfc_parent = -1;
1545 
1546 			/* clear the RP address */
1547 			rt->mfc_rp = zeroin_addr;
1548 
1549 			rt->mfc_bw_meter = NULL;
1550 
1551 			/* link into table */
1552 			LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1553 			/* Add this entry to the end of the queue */
1554 			rt->mfc_stall = rte;
1555 		} else {
1556 			/* determine if q has overflowed */
1557 			struct rtdetq **p;
1558 			int npkts = 0;
1559 
1560 			/*
1561 			 * XXX ouch! we need to append to the list, but we
1562 			 * only have a pointer to the front, so we have to
1563 			 * scan the entire list every time.
1564 			 */
1565 			for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1566 				if (++npkts > MAX_UPQ) {
1567 					mrtstat.mrts_upq_ovflw++;
1568 				non_fatal:
1569 					free(rte, M_MRTABLE);
1570 					m_freem(mb0);
1571 					splx(s);
1572 					return (0);
1573 				}
1574 
1575 			/* Add this entry to the end of the queue */
1576 			*p = rte;
1577 		}
1578 
1579 		rte->next = NULL;
1580 		rte->m = mb0;
1581 		rte->ifp = ifp;
1582 	#ifdef UPCALL_TIMING
1583 		rte->t = tp;
1584 	#endif /* UPCALL_TIMING */
1585 
1586 		splx(s);
1587 
1588 		return (0);
1589 	}
1590 }
1591 
1592 
1593 /*ARGSUSED*/
1594 static void
1595 expire_upcalls(void *v)
1596 {
1597 	int i;
1598 	int s;
1599 
1600 	s = splsoftnet();
1601 
1602 	for (i = 0; i < MFCTBLSIZ; i++) {
1603 		struct mfc *rt, *nrt;
1604 
1605 		if (nexpire[i] == 0)
1606 			continue;
1607 
1608 		for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1609 			nrt = LIST_NEXT(rt, mfc_hash);
1610 
1611 			if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1612 				continue;
1613 			nexpire[i]--;
1614 
1615 			/*
1616 			 * free the bw_meter entries
1617 			 */
1618 			while (rt->mfc_bw_meter != NULL) {
1619 				struct bw_meter *x = rt->mfc_bw_meter;
1620 
1621 				rt->mfc_bw_meter = x->bm_mfc_next;
1622 				free(x, M_BWMETER);
1623 			}
1624 
1625 			++mrtstat.mrts_cache_cleanups;
1626 			if (mrtdebug & DEBUG_EXPIRE)
1627 				log(LOG_DEBUG,
1628 				    "expire_upcalls: expiring (%x %x)\n",
1629 				    ntohl(rt->mfc_origin.s_addr),
1630 				    ntohl(rt->mfc_mcastgrp.s_addr));
1631 
1632 			expire_mfc(rt);
1633 		}
1634 	}
1635 
1636 	splx(s);
1637 	timeout_add(&expire_upcalls_ch, EXPIRE_TIMEOUT);
1638 }
1639 
1640 /*
1641  * Packet forwarding routine once entry in the cache is made
1642  */
1643 static int
1644 #ifdef RSVP_ISI
1645 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1646 #else
1647 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1648 #endif /* RSVP_ISI */
1649 {
1650 	struct ip  *ip = mtod(m, struct ip *);
1651 	vifi_t vifi;
1652 	struct vif *vifp;
1653 	int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1654 
1655 /*
1656  * Macro to send packet on vif.  Since RSVP packets don't get counted on
1657  * input, they shouldn't get counted on output, so statistics keeping is
1658  * separate.
1659  */
1660 #define MC_SEND(ip, vifp, m) do {					\
1661 	if ((vifp)->v_flags & VIFF_TUNNEL)				\
1662 		encap_send((ip), (vifp), (m));				\
1663 	else								\
1664 		phyint_send((ip), (vifp), (m));				\
1665 } while (/*CONSTCOND*/ 0)
1666 
1667 #ifdef RSVP_ISI
1668 	/*
1669 	 * If xmt_vif is not -1, send on only the requested vif.
1670 	 *
1671 	 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1672 	 */
1673 	if (xmt_vif < numvifs) {
1674 #ifdef PIM
1675 		if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1676 			pim_register_send(ip, viftable + xmt_vif, m, rt);
1677 		else
1678 #endif
1679 		MC_SEND(ip, viftable + xmt_vif, m);
1680 		return (1);
1681 	}
1682 #endif /* RSVP_ISI */
1683 
1684 	/*
1685 	 * Don't forward if it didn't arrive from the parent vif for its origin.
1686 	 */
1687 	vifi = rt->mfc_parent;
1688 	if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1689 		/* came in the wrong interface */
1690 		if (mrtdebug & DEBUG_FORWARD)
1691 			log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1692 			    ifp, vifi,
1693 			    vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1694 		++mrtstat.mrts_wrong_if;
1695 		++rt->mfc_wrong_if;
1696 		/*
1697 		 * If we are doing PIM assert processing, send a message
1698 		 * to the routing daemon.
1699 		 *
1700 		 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1701 		 * can complete the SPT switch, regardless of the type
1702 		 * of interface (broadcast media, GRE tunnel, etc).
1703 		 */
1704 		if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1705 			struct timeval now;
1706 			u_int32_t delta;
1707 
1708 #ifdef PIM
1709 			if (ifp == &multicast_register_if)
1710 				pimstat.pims_rcv_registers_wrongiif++;
1711 #endif
1712 
1713 			/* Get vifi for the incoming packet */
1714 			for (vifi = 0;
1715 			     vifi < numvifs && viftable[vifi].v_ifp != ifp;
1716 			     vifi++)
1717 			    ;
1718 			if (vifi >= numvifs) {
1719 				/* The iif is not found: ignore the packet. */
1720 				return (0);
1721 			}
1722 
1723 			if (rt->mfc_flags[vifi] &
1724 			    MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1725 				/* WRONGVIF disabled: ignore the packet */
1726 				return (0);
1727 			}
1728 
1729 			microtime(&now);
1730 
1731 			TV_DELTA(rt->mfc_last_assert, now, delta);
1732 
1733 			if (delta > ASSERT_MSG_TIME) {
1734 				struct igmpmsg *im;
1735 				int hlen = ip->ip_hl << 2;
1736 				struct mbuf *mm = m_copy(m, 0, hlen);
1737 
1738 				M_PULLUP(mm, hlen);
1739 				if (mm == NULL)
1740 					return (ENOBUFS);
1741 
1742 				rt->mfc_last_assert = now;
1743 
1744 				im = mtod(mm, struct igmpmsg *);
1745 				im->im_msgtype	= IGMPMSG_WRONGVIF;
1746 				im->im_mbz	= 0;
1747 				im->im_vif	= vifi;
1748 
1749 				mrtstat.mrts_upcalls++;
1750 
1751 				sin.sin_addr = im->im_src;
1752 				if (socket_send(ip_mrouter, mm, &sin) < 0) {
1753 					log(LOG_WARNING, "ip_mforward: "
1754 					    "ip_mrouter socket queue full\n");
1755 					++mrtstat.mrts_upq_sockfull;
1756 					return (ENOBUFS);
1757 				}
1758 			}
1759 		}
1760 		return (0);
1761 	}
1762 
1763 	/* If I sourced this packet, it counts as output, else it was input. */
1764 	if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1765 		viftable[vifi].v_pkt_out++;
1766 		viftable[vifi].v_bytes_out += plen;
1767 	} else {
1768 		viftable[vifi].v_pkt_in++;
1769 		viftable[vifi].v_bytes_in += plen;
1770 	}
1771 	rt->mfc_pkt_cnt++;
1772 	rt->mfc_byte_cnt += plen;
1773 
1774 	/*
1775 	 * For each vif, decide if a copy of the packet should be forwarded.
1776 	 * Forward if:
1777 	 *		- the ttl exceeds the vif's threshold
1778 	 *		- there are group members downstream on interface
1779 	 */
1780 	for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1781 		if ((rt->mfc_ttls[vifi] > 0) &&
1782 			(ip->ip_ttl > rt->mfc_ttls[vifi])) {
1783 			vifp->v_pkt_out++;
1784 			vifp->v_bytes_out += plen;
1785 #ifdef PIM
1786 			if (vifp->v_flags & VIFF_REGISTER)
1787 				pim_register_send(ip, vifp, m, rt);
1788 			else
1789 #endif
1790 			MC_SEND(ip, vifp, m);
1791 		}
1792 
1793 	/*
1794 	 * Perform upcall-related bw measuring.
1795 	 */
1796 	if (rt->mfc_bw_meter != NULL) {
1797 		struct bw_meter *x;
1798 		struct timeval now;
1799 
1800 		microtime(&now);
1801 		for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1802 			bw_meter_receive_packet(x, plen, &now);
1803 	}
1804 
1805 	return (0);
1806 }
1807 
1808 #ifdef RSVP_ISI
1809 /*
1810  * check if a vif number is legal/ok. This is used by ip_output.
1811  */
1812 int
1813 legal_vif_num(int vif)
1814 {
1815 	if (vif >= 0 && vif < numvifs)
1816 		return (1);
1817 	else
1818 		return (0);
1819 }
1820 #endif /* RSVP_ISI */
1821 
1822 static void
1823 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1824 {
1825 	struct mbuf *mb_copy;
1826 	int hlen = ip->ip_hl << 2;
1827 
1828 	/*
1829 	 * Make a new reference to the packet; make sure that
1830 	 * the IP header is actually copied, not just referenced,
1831 	 * so that ip_output() only scribbles on the copy.
1832 	 */
1833 	mb_copy = m_copy(m, 0, M_COPYALL);
1834 	M_PULLUP(mb_copy, hlen);
1835 	if (mb_copy == NULL)
1836 		return;
1837 
1838 	if (vifp->v_rate_limit <= 0)
1839 		tbf_send_packet(vifp, mb_copy);
1840 	else
1841 		tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1842 		    ntohs(ip->ip_len));
1843 }
1844 
1845 static void
1846 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1847 {
1848 	struct mbuf *mb_copy;
1849 	struct ip *ip_copy;
1850 	int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1851 
1852 	/* Take care of delayed checksums */
1853 	if (m->m_pkthdr.csum_flags & (M_TCPV4_CSUM_OUT | M_UDPV4_CSUM_OUT)) {
1854 		in_delayed_cksum(m);
1855 		m->m_pkthdr.csum_flags &=
1856 		    ~(M_UDPV4_CSUM_OUT | M_TCPV4_CSUM_OUT);
1857 	}
1858 
1859 	/*
1860 	 * copy the old packet & pullup its IP header into the
1861 	 * new mbuf so we can modify it.  Try to fill the new
1862 	 * mbuf since if we don't the ethernet driver will.
1863 	 */
1864 	MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1865 	if (mb_copy == NULL)
1866 		return;
1867 	mb_copy->m_data += max_linkhdr;
1868 	mb_copy->m_pkthdr.len = len;
1869 	mb_copy->m_len = sizeof(multicast_encap_iphdr);
1870 
1871 	if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1872 		m_freem(mb_copy);
1873 		return;
1874 	}
1875 	i = MHLEN - max_linkhdr;
1876 	if (i > len)
1877 		i = len;
1878 	mb_copy = m_pullup(mb_copy, i);
1879 	if (mb_copy == NULL)
1880 		return;
1881 
1882 	/*
1883 	 * fill in the encapsulating IP header.
1884 	 */
1885 	ip_copy = mtod(mb_copy, struct ip *);
1886 	*ip_copy = multicast_encap_iphdr;
1887 	ip_copy->ip_id = htons(ip_randomid());
1888 	ip_copy->ip_len = htons(len);
1889 	ip_copy->ip_src = vifp->v_lcl_addr;
1890 	ip_copy->ip_dst = vifp->v_rmt_addr;
1891 
1892 	/*
1893 	 * turn the encapsulated IP header back into a valid one.
1894 	 */
1895 	ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1896 	--ip->ip_ttl;
1897 	ip->ip_sum = 0;
1898 	mb_copy->m_data += sizeof(multicast_encap_iphdr);
1899 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1900 	mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1901 
1902 	if (vifp->v_rate_limit <= 0)
1903 		tbf_send_packet(vifp, mb_copy);
1904 	else
1905 		tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1906 }
1907 
1908 /*
1909  * Token bucket filter module
1910  */
1911 static void
1912 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
1913 {
1914 
1915 	if (len > MAX_BKT_SIZE) {
1916 		/* drop if packet is too large */
1917 		mrtstat.mrts_pkt2large++;
1918 		m_freem(m);
1919 		return;
1920 	}
1921 
1922 	tbf_update_tokens(vifp);
1923 
1924 	/*
1925 	 * If there are enough tokens, and the queue is empty, send this packet
1926 	 * out immediately.  Otherwise, try to insert it on this vif's queue.
1927 	 */
1928 	if (vifp->tbf_q_len == 0) {
1929 		if (len <= vifp->tbf_n_tok) {
1930 			vifp->tbf_n_tok -= len;
1931 			tbf_send_packet(vifp, m);
1932 		} else {
1933 			/* queue packet and timeout till later */
1934 			tbf_queue(vifp, m);
1935 			timeout_add(&vifp->v_repq_ch, TBF_REPROCESS);
1936 		}
1937 	} else {
1938 		if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
1939 		    !tbf_dq_sel(vifp, ip)) {
1940 			/* queue full, and couldn't make room */
1941 			mrtstat.mrts_q_overflow++;
1942 			m_freem(m);
1943 		} else {
1944 			/* queue length low enough, or made room */
1945 			tbf_queue(vifp, m);
1946 			tbf_process_q(vifp);
1947 		}
1948 	}
1949 }
1950 
1951 /*
1952  * adds a packet to the queue at the interface
1953  */
1954 static void
1955 tbf_queue(struct vif *vifp, struct mbuf *m)
1956 {
1957 	int s = splsoftnet();
1958 
1959 	/* insert at tail */
1960 	*vifp->tbf_t = m;
1961 	vifp->tbf_t = &m->m_nextpkt;
1962 	vifp->tbf_q_len++;
1963 
1964 	splx(s);
1965 }
1966 
1967 
1968 /*
1969  * processes the queue at the interface
1970  */
1971 static void
1972 tbf_process_q(struct vif *vifp)
1973 {
1974 	struct mbuf *m;
1975 	int len;
1976 	int s = splsoftnet();
1977 
1978 	/*
1979 	 * Loop through the queue at the interface and send as many packets
1980 	 * as possible.
1981 	 */
1982 	for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
1983 		len = ntohs(mtod(m, struct ip *)->ip_len);
1984 
1985 		/* determine if the packet can be sent */
1986 		if (len <= vifp->tbf_n_tok) {
1987 			/* if so,
1988 			 * reduce no of tokens, dequeue the packet,
1989 			 * send the packet.
1990 			 */
1991 			if ((vifp->tbf_q = m->m_nextpkt) == NULL)
1992 				vifp->tbf_t = &vifp->tbf_q;
1993 			--vifp->tbf_q_len;
1994 
1995 			m->m_nextpkt = NULL;
1996 			vifp->tbf_n_tok -= len;
1997 			tbf_send_packet(vifp, m);
1998 		} else
1999 			break;
2000 	}
2001 	splx(s);
2002 }
2003 
2004 static void
2005 tbf_reprocess_q(void *arg)
2006 {
2007 	struct vif *vifp = arg;
2008 
2009 	if (ip_mrouter == NULL)
2010 		return;
2011 
2012 	tbf_update_tokens(vifp);
2013 	tbf_process_q(vifp);
2014 
2015 	if (vifp->tbf_q_len != 0)
2016 		timeout_add(&vifp->v_repq_ch, TBF_REPROCESS);
2017 }
2018 
2019 /* function that will selectively discard a member of the queue
2020  * based on the precedence value and the priority
2021  */
2022 static int
2023 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2024 {
2025 	u_int p;
2026 	struct mbuf **mp, *m;
2027 	int s = splsoftnet();
2028 
2029 	p = priority(vifp, ip);
2030 
2031 	for (mp = &vifp->tbf_q, m = *mp;
2032 	    m != NULL;
2033 	    mp = &m->m_nextpkt, m = *mp) {
2034 		if (p > priority(vifp, mtod(m, struct ip *))) {
2035 			if ((*mp = m->m_nextpkt) == NULL)
2036 				vifp->tbf_t = mp;
2037 			--vifp->tbf_q_len;
2038 
2039 			m_freem(m);
2040 			mrtstat.mrts_drop_sel++;
2041 			splx(s);
2042 			return (1);
2043 		}
2044 	}
2045 	splx(s);
2046 	return (0);
2047 }
2048 
2049 static void
2050 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2051 {
2052 	int error;
2053 	int s = splsoftnet();
2054 
2055 	if (vifp->v_flags & VIFF_TUNNEL) {
2056 		/* If tunnel options */
2057 		ip_output(m, (struct mbuf *)NULL, &vifp->v_route,
2058 		    IP_FORWARDING, (struct ip_moptions *)NULL,
2059 		    (struct inpcb *)NULL);
2060 	} else {
2061 		/*
2062 		 * if physical interface option, extract the options
2063 		 * and then send
2064 		 */
2065 		struct ip_moptions imo;
2066 
2067 		imo.imo_multicast_ifp = vifp->v_ifp;
2068 		imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2069 		imo.imo_multicast_loop = 1;
2070 #ifdef RSVP_ISI
2071 		imo.imo_multicast_vif = -1;
2072 #endif
2073 
2074 		error = ip_output(m, (struct mbuf *)NULL, (struct route *)NULL,
2075 		    IP_FORWARDING|IP_MULTICASTOPTS, &imo,
2076 		    (struct inpcb *)NULL);
2077 
2078 		if (mrtdebug & DEBUG_XMIT)
2079 			log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2080 			    (long)(vifp - viftable), error);
2081 	}
2082 	splx(s);
2083 }
2084 
2085 /* determine the current time and then
2086  * the elapsed time (between the last time and time now)
2087  * in milliseconds & update the no. of tokens in the bucket
2088  */
2089 static void
2090 tbf_update_tokens(struct vif *vifp)
2091 {
2092 	struct timeval tp;
2093 	u_int32_t tm;
2094 	int s = splsoftnet();
2095 
2096 	microtime(&tp);
2097 
2098 	TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2099 
2100 	/*
2101 	 * This formula is actually
2102 	 * "time in seconds" * "bytes/second".
2103 	 *
2104 	 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2105 	 *
2106 	 * The (1000/1024) was introduced in add_vif to optimize
2107 	 * this divide into a shift.
2108 	 */
2109 	vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2110 	vifp->tbf_last_pkt_t = tp;
2111 
2112 	if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2113 		vifp->tbf_n_tok = MAX_BKT_SIZE;
2114 
2115 	splx(s);
2116 }
2117 
2118 static int
2119 priority(struct vif *vifp, struct ip *ip)
2120 {
2121 	int prio = 50;	/* the lowest priority -- default case */
2122 
2123 	/* temporary hack; may add general packet classifier some day */
2124 
2125 	/*
2126 	 * The UDP port space is divided up into four priority ranges:
2127 	 * [0, 16384)     : unclassified - lowest priority
2128 	 * [16384, 32768) : audio - highest priority
2129 	 * [32768, 49152) : whiteboard - medium priority
2130 	 * [49152, 65536) : video - low priority
2131 	 */
2132 	if (ip->ip_p == IPPROTO_UDP) {
2133 		struct udphdr *udp =
2134 		    (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2135 
2136 		switch (ntohs(udp->uh_dport) & 0xc000) {
2137 		case 0x4000:
2138 			prio = 70;
2139 			break;
2140 		case 0x8000:
2141 			prio = 60;
2142 			break;
2143 		case 0xc000:
2144 			prio = 55;
2145 			break;
2146 		}
2147 
2148 		if (tbfdebug > 1)
2149 			log(LOG_DEBUG, "port %x prio %d\n",
2150 			    ntohs(udp->uh_dport), prio);
2151 	}
2152 
2153 	return (prio);
2154 }
2155 
2156 /*
2157  * End of token bucket filter modifications
2158  */
2159 #ifdef RSVP_ISI
2160 int
2161 ip_rsvp_vif_init(struct socket *so, struct mbuf *m)
2162 {
2163 	int vifi, s;
2164 
2165 	if (rsvpdebug)
2166 		printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2167 		    so->so_type, so->so_proto->pr_protocol);
2168 
2169 	if (so->so_type != SOCK_RAW ||
2170 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2171 		return (EOPNOTSUPP);
2172 
2173 	/* Check mbuf. */
2174 	if (m == NULL || m->m_len != sizeof(int)) {
2175 		return (EINVAL);
2176 	}
2177 	vifi = *(mtod(m, int *));
2178 
2179 	if (rsvpdebug)
2180 		printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",
2181 		    vifi, rsvp_on);
2182 
2183 	s = splsoftnet();
2184 
2185 	/* Check vif. */
2186 	if (!legal_vif_num(vifi)) {
2187 		splx(s);
2188 		return (EADDRNOTAVAIL);
2189 	}
2190 
2191 	/* Check if socket is available. */
2192 	if (viftable[vifi].v_rsvpd != NULL) {
2193 		splx(s);
2194 		return (EADDRINUSE);
2195 	}
2196 
2197 	viftable[vifi].v_rsvpd = so;
2198 	/* This may seem silly, but we need to be sure we don't over-increment
2199 	 * the RSVP counter, in case something slips up.
2200 	 */
2201 	if (!viftable[vifi].v_rsvp_on) {
2202 		viftable[vifi].v_rsvp_on = 1;
2203 		rsvp_on++;
2204 	}
2205 
2206 	splx(s);
2207 	return (0);
2208 }
2209 
2210 int
2211 ip_rsvp_vif_done(struct socket *so, struct mbuf *m)
2212 {
2213 	int vifi, s;
2214 
2215 	if (rsvpdebug)
2216 		printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2217 		    so->so_type, so->so_proto->pr_protocol);
2218 
2219 	if (so->so_type != SOCK_RAW ||
2220 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2221 		return (EOPNOTSUPP);
2222 
2223 	/* Check mbuf. */
2224 	if (m == NULL || m->m_len != sizeof(int)) {
2225 		return (EINVAL);
2226 	}
2227 	vifi = *(mtod(m, int *));
2228 
2229 	s = splsoftnet();
2230 
2231 	/* Check vif. */
2232 	if (!legal_vif_num(vifi)) {
2233 		splx(s);
2234 		return (EADDRNOTAVAIL);
2235 	}
2236 
2237 	if (rsvpdebug)
2238 		printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2239 		    viftable[vifi].v_rsvpd, so);
2240 
2241 	viftable[vifi].v_rsvpd = NULL;
2242 	/*
2243 	 * This may seem silly, but we need to be sure we don't over-decrement
2244 	 * the RSVP counter, in case something slips up.
2245 	 */
2246 	if (viftable[vifi].v_rsvp_on) {
2247 		viftable[vifi].v_rsvp_on = 0;
2248 		rsvp_on--;
2249 	}
2250 
2251 	splx(s);
2252 	return (0);
2253 }
2254 
2255 void
2256 ip_rsvp_force_done(struct socket *so)
2257 {
2258 	int vifi, s;
2259 
2260 	/* Don't bother if it is not the right type of socket. */
2261 	if (so->so_type != SOCK_RAW ||
2262 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2263 		return;
2264 
2265 	s = splsoftnet();
2266 
2267 	/*
2268 	 * The socket may be attached to more than one vif...this
2269 	 * is perfectly legal.
2270 	 */
2271 	for (vifi = 0; vifi < numvifs; vifi++) {
2272 		if (viftable[vifi].v_rsvpd == so) {
2273 			viftable[vifi].v_rsvpd = NULL;
2274 			/*
2275 			 * This may seem silly, but we need to be sure we don't
2276 			 * over-decrement the RSVP counter, in case something
2277 			 * slips up.
2278 			 */
2279 			if (viftable[vifi].v_rsvp_on) {
2280 				viftable[vifi].v_rsvp_on = 0;
2281 				rsvp_on--;
2282 			}
2283 		}
2284 	}
2285 
2286 	splx(s);
2287 	return;
2288 }
2289 
2290 void
2291 rsvp_input(struct mbuf *m, struct ifnet *ifp)
2292 {
2293 	int vifi, s;
2294 	struct ip *ip = mtod(m, struct ip *);
2295 	static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET };
2296 
2297 	if (rsvpdebug)
2298 		printf("rsvp_input: rsvp_on %d\n", rsvp_on);
2299 
2300 	/*
2301 	 * Can still get packets with rsvp_on = 0 if there is a local member
2302 	 * of the group to which the RSVP packet is addressed.  But in this
2303 	 * case we want to throw the packet away.
2304 	 */
2305 	if (!rsvp_on) {
2306 		m_freem(m);
2307 		return;
2308 	}
2309 
2310 	/*
2311 	 * If the old-style non-vif-associated socket is set, then use
2312 	 * it and ignore the new ones.
2313 	 */
2314 	if (ip_rsvpd != NULL) {
2315 		if (rsvpdebug)
2316 			printf("rsvp_input: "
2317 			    "Sending packet up old-style socket\n");
2318 		rip_input(m, 0);	/*XXX*/
2319 		return;
2320 	}
2321 
2322 	s = splsoftnet();
2323 
2324 	if (rsvpdebug)
2325 		printf("rsvp_input: check vifs\n");
2326 
2327 	/* Find which vif the packet arrived on. */
2328 	for (vifi = 0; vifi < numvifs; vifi++) {
2329 		if (viftable[vifi].v_ifp == ifp)
2330 			break;
2331 	}
2332 
2333 	if (vifi == numvifs) {
2334 		/* Can't find vif packet arrived on. Drop packet. */
2335 		if (rsvpdebug)
2336 			printf("rsvp_input: "
2337 			    "Can't find vif for packet...dropping it.\n");
2338 		m_freem(m);
2339 		splx(s);
2340 		return;
2341 	}
2342 
2343 	if (rsvpdebug)
2344 		printf("rsvp_input: check socket\n");
2345 
2346 	if (viftable[vifi].v_rsvpd == NULL) {
2347 		/*
2348 	 	 * drop packet, since there is no specific socket for this
2349 		 * interface
2350 		 */
2351 		if (rsvpdebug)
2352 			printf("rsvp_input: No socket defined for vif %d\n",
2353 			    vifi);
2354 		m_freem(m);
2355 		splx(s);
2356 		return;
2357 	}
2358 
2359 	rsvp_src.sin_addr = ip->ip_src;
2360 
2361 	if (rsvpdebug && m)
2362 		printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
2363 		    m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv));
2364 
2365 	if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2366 		if (rsvpdebug)
2367 			printf("rsvp_input: Failed to append to socket\n");
2368 	else
2369 		if (rsvpdebug)
2370 			printf("rsvp_input: send packet up\n");
2371 
2372 	splx(s);
2373 }
2374 #endif /* RSVP_ISI */
2375 
2376 /*
2377  * Code for bandwidth monitors
2378  */
2379 
2380 /*
2381  * Define common interface for timeval-related methods
2382  */
2383 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2384 #define	BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2385 #define	BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2386 
2387 static uint32_t
2388 compute_bw_meter_flags(struct bw_upcall *req)
2389 {
2390 	uint32_t flags = 0;
2391 
2392 	if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2393 		flags |= BW_METER_UNIT_PACKETS;
2394 	if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2395 		flags |= BW_METER_UNIT_BYTES;
2396 	if (req->bu_flags & BW_UPCALL_GEQ)
2397 		flags |= BW_METER_GEQ;
2398 	if (req->bu_flags & BW_UPCALL_LEQ)
2399 		flags |= BW_METER_LEQ;
2400 
2401 	return (flags);
2402 }
2403 
2404 /*
2405  * Add a bw_meter entry
2406  */
2407 static int
2408 add_bw_upcall(struct mbuf *m)
2409 {
2410 	int s;
2411 	struct mfc *mfc;
2412 	struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2413 	    BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2414 	struct timeval now;
2415 	struct bw_meter *x;
2416 	uint32_t flags;
2417 	struct bw_upcall *req;
2418 
2419 	if (m == NULL || m->m_len < sizeof(struct bw_upcall))
2420 		return (EINVAL);
2421 
2422 	req = mtod(m, struct bw_upcall *);
2423 
2424 	if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2425 		return (EOPNOTSUPP);
2426 
2427 	/* Test if the flags are valid */
2428 	if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2429 		return (EINVAL);
2430 	if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2431 		return (EINVAL);
2432 	if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2433 	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2434 		return (EINVAL);
2435 
2436 	/* Test if the threshold time interval is valid */
2437 	if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2438 		return (EINVAL);
2439 
2440 	flags = compute_bw_meter_flags(req);
2441 
2442 	/* Find if we have already same bw_meter entry */
2443 	s = splsoftnet();
2444 	mfc = mfc_find(&req->bu_src, &req->bu_dst);
2445 	if (mfc == NULL) {
2446 		splx(s);
2447 		return (EADDRNOTAVAIL);
2448 	}
2449 	for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2450 		if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2451 		    &req->bu_threshold.b_time, ==)) &&
2452 		    (x->bm_threshold.b_packets ==
2453 		    req->bu_threshold.b_packets) &&
2454 		    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2455 		    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
2456 			splx(s);
2457 			return (0);	/* XXX Already installed */
2458 		}
2459 	}
2460 
2461 	/* Allocate the new bw_meter entry */
2462 	x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2463 	if (x == NULL) {
2464 		splx(s);
2465 		return (ENOBUFS);
2466 	}
2467 
2468 	/* Set the new bw_meter entry */
2469 	x->bm_threshold.b_time = req->bu_threshold.b_time;
2470 	microtime(&now);
2471 	x->bm_start_time = now;
2472 	x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2473 	x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2474 	x->bm_measured.b_packets = 0;
2475 	x->bm_measured.b_bytes = 0;
2476 	x->bm_flags = flags;
2477 	x->bm_time_next = NULL;
2478 	x->bm_time_hash = BW_METER_BUCKETS;
2479 
2480 	/* Add the new bw_meter entry to the front of entries for this MFC */
2481 	x->bm_mfc = mfc;
2482 	x->bm_mfc_next = mfc->mfc_bw_meter;
2483 	mfc->mfc_bw_meter = x;
2484 	schedule_bw_meter(x, &now);
2485 	splx(s);
2486 
2487 	return (0);
2488 }
2489 
2490 static void
2491 free_bw_list(struct bw_meter *list)
2492 {
2493 	while (list != NULL) {
2494 		struct bw_meter *x = list;
2495 
2496 		list = list->bm_mfc_next;
2497 		unschedule_bw_meter(x);
2498 		free(x, M_BWMETER);
2499 	}
2500 }
2501 
2502 /*
2503  * Delete one or multiple bw_meter entries
2504  */
2505 static int
2506 del_bw_upcall(struct mbuf *m)
2507 {
2508 	int s;
2509 	struct mfc *mfc;
2510 	struct bw_meter *x;
2511 	struct bw_upcall *req;
2512 
2513 	if (m == NULL || m->m_len < sizeof(struct bw_upcall))
2514 		return (EINVAL);
2515 
2516 	req = mtod(m, struct bw_upcall *);
2517 
2518 	if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2519 		return (EOPNOTSUPP);
2520 
2521 	s = splsoftnet();
2522 	/* Find the corresponding MFC entry */
2523 	mfc = mfc_find(&req->bu_src, &req->bu_dst);
2524 	if (mfc == NULL) {
2525 		splx(s);
2526 		return (EADDRNOTAVAIL);
2527 	} else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2528 		/* Delete all bw_meter entries for this mfc */
2529 		struct bw_meter *list;
2530 
2531 		list = mfc->mfc_bw_meter;
2532 		mfc->mfc_bw_meter = NULL;
2533 		free_bw_list(list);
2534 		splx(s);
2535 		return (0);
2536 	} else {	/* Delete a single bw_meter entry */
2537 		struct bw_meter *prev;
2538 		uint32_t flags = 0;
2539 
2540 		flags = compute_bw_meter_flags(req);
2541 
2542 		/* Find the bw_meter entry to delete */
2543 		for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2544 		    prev = x, x = x->bm_mfc_next) {
2545 			if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2546 			    &req->bu_threshold.b_time, ==)) &&
2547 			    (x->bm_threshold.b_packets ==
2548 			    req->bu_threshold.b_packets) &&
2549 			    (x->bm_threshold.b_bytes ==
2550 			    req->bu_threshold.b_bytes) &&
2551 			    (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2552 				break;
2553 		}
2554 		if (x != NULL) { /* Delete entry from the list for this MFC */
2555 			if (prev != NULL) {
2556 				/* remove from middle */
2557 				prev->bm_mfc_next = x->bm_mfc_next;
2558 			} else {
2559 				/* new head of list */
2560 				x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;
2561 			}
2562 
2563 			unschedule_bw_meter(x);
2564 			splx(s);
2565 			/* Free the bw_meter entry */
2566 			free(x, M_BWMETER);
2567 			return (0);
2568 		} else {
2569 			splx(s);
2570 			return (EINVAL);
2571 		}
2572 	}
2573 	/* NOTREACHED */
2574 }
2575 
2576 /*
2577  * Perform bandwidth measurement processing that may result in an upcall
2578  */
2579 static void
2580 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2581 {
2582 	struct timeval delta;
2583 
2584 	delta = *nowp;
2585 	BW_TIMEVALDECR(&delta, &x->bm_start_time);
2586 
2587 	if (x->bm_flags & BW_METER_GEQ) {
2588 		/* Processing for ">=" type of bw_meter entry */
2589 		if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2590 			/* Reset the bw_meter entry */
2591 			x->bm_start_time = *nowp;
2592 			x->bm_measured.b_packets = 0;
2593 			x->bm_measured.b_bytes = 0;
2594 			x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2595 		}
2596 
2597 		/* Record that a packet is received */
2598 		x->bm_measured.b_packets++;
2599 		x->bm_measured.b_bytes += plen;
2600 
2601 		/* Test if we should deliver an upcall */
2602 		if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2603 			if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2604 			    (x->bm_measured.b_packets >=
2605 			    x->bm_threshold.b_packets)) ||
2606 			    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2607 			    (x->bm_measured.b_bytes >=
2608 			    x->bm_threshold.b_bytes))) {
2609 				/* Prepare an upcall for delivery */
2610 				bw_meter_prepare_upcall(x, nowp);
2611 				x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2612 			}
2613 		}
2614 	} else if (x->bm_flags & BW_METER_LEQ) {
2615 		/* Processing for "<=" type of bw_meter entry */
2616 		if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2617 			/*
2618 			 * We are behind time with the multicast forwarding
2619 			 * table scanning for "<=" type of bw_meter entries,
2620 			 * so test now if we should deliver an upcall.
2621 			 */
2622 			if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2623 			    (x->bm_measured.b_packets <=
2624 			    x->bm_threshold.b_packets)) ||
2625 			    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2626 			    (x->bm_measured.b_bytes <=
2627 			    x->bm_threshold.b_bytes))) {
2628 				/* Prepare an upcall for delivery */
2629 				bw_meter_prepare_upcall(x, nowp);
2630 			}
2631 			/* Reschedule the bw_meter entry */
2632 			unschedule_bw_meter(x);
2633 			schedule_bw_meter(x, nowp);
2634 		}
2635 
2636 		/* Record that a packet is received */
2637 		x->bm_measured.b_packets++;
2638 		x->bm_measured.b_bytes += plen;
2639 
2640 		/* Test if we should restart the measuring interval */
2641 		if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2642 		    x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2643 		    (x->bm_flags & BW_METER_UNIT_BYTES &&
2644 		    x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2645 			/* Don't restart the measuring interval */
2646 		} else {
2647 			/* Do restart the measuring interval */
2648 			/*
2649 			 * XXX: note that we don't unschedule and schedule,
2650 			 * because this might be too much overhead per packet.
2651 			 * Instead, when we process all entries for a given
2652 			 * timer hash bin, we check whether it is really a
2653 			 * timeout. If not, we reschedule at that time.
2654 			 */
2655 			x->bm_start_time = *nowp;
2656 			x->bm_measured.b_packets = 0;
2657 			x->bm_measured.b_bytes = 0;
2658 			x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2659 		}
2660 	}
2661 }
2662 
2663 /*
2664  * Prepare a bandwidth-related upcall
2665  */
2666 static void
2667 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2668 {
2669 	struct timeval delta;
2670 	struct bw_upcall *u;
2671 
2672 	/* Compute the measured time interval */
2673 	delta = *nowp;
2674 	BW_TIMEVALDECR(&delta, &x->bm_start_time);
2675 
2676 	/* If there are too many pending upcalls, deliver them now */
2677 	if (bw_upcalls_n >= BW_UPCALLS_MAX)
2678 		bw_upcalls_send();
2679 
2680 	/* Set the bw_upcall entry */
2681 	u = &bw_upcalls[bw_upcalls_n++];
2682 	u->bu_src = x->bm_mfc->mfc_origin;
2683 	u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2684 	u->bu_threshold.b_time = x->bm_threshold.b_time;
2685 	u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2686 	u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2687 	u->bu_measured.b_time = delta;
2688 	u->bu_measured.b_packets = x->bm_measured.b_packets;
2689 	u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2690 	u->bu_flags = 0;
2691 	if (x->bm_flags & BW_METER_UNIT_PACKETS)
2692 		u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2693 	if (x->bm_flags & BW_METER_UNIT_BYTES)
2694 		u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2695 	if (x->bm_flags & BW_METER_GEQ)
2696 		u->bu_flags |= BW_UPCALL_GEQ;
2697 	if (x->bm_flags & BW_METER_LEQ)
2698 		u->bu_flags |= BW_UPCALL_LEQ;
2699 }
2700 
2701 /*
2702  * Send the pending bandwidth-related upcalls
2703  */
2704 static void
2705 bw_upcalls_send(void)
2706 {
2707 	struct mbuf *m;
2708 	int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2709 	struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2710 	static struct igmpmsg igmpmsg = {
2711 	    0,			/* unused1 */
2712 	    0,			/* unused2 */
2713 	    IGMPMSG_BW_UPCALL,	/* im_msgtype */
2714 	    0,			/* im_mbz  */
2715 	    0,			/* im_vif  */
2716 	    0,			/* unused3 */
2717 	    { 0 },		/* im_src  */
2718 	    { 0 } };		/* im_dst  */
2719 
2720 	if (bw_upcalls_n == 0)
2721 		return;		/* No pending upcalls */
2722 
2723 	bw_upcalls_n = 0;
2724 
2725 	/*
2726 	 * Allocate a new mbuf, initialize it with the header and
2727 	 * the payload for the pending calls.
2728 	 */
2729 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2730 	if (m == NULL) {
2731 		log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2732 		return;
2733 	}
2734 
2735 	m->m_len = m->m_pkthdr.len = 0;
2736 	m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2737 	m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2738 
2739 	/*
2740 	 * Send the upcalls
2741 	 * XXX do we need to set the address in k_igmpsrc ?
2742 	 */
2743 	mrtstat.mrts_upcalls++;
2744 	if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2745 		log(LOG_WARNING,
2746 		    "bw_upcalls_send: ip_mrouter socket queue full\n");
2747 		++mrtstat.mrts_upq_sockfull;
2748 	}
2749 }
2750 
2751 /*
2752  * Compute the timeout hash value for the bw_meter entries
2753  */
2754 #define	BW_METER_TIMEHASH(bw_meter, hash) do {				\
2755 	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2756 									\
2757 	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2758 	(hash) = next_timeval.tv_sec;					\
2759 	if (next_timeval.tv_usec)					\
2760 		(hash)++; /* XXX: make sure we don't timeout early */	\
2761 	(hash) %= BW_METER_BUCKETS;					\
2762 } while (/*CONSTCOND*/ 0)
2763 
2764 /*
2765  * Schedule a timer to process periodically bw_meter entry of type "<="
2766  * by linking the entry in the proper hash bucket.
2767  */
2768 static void
2769 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2770 {
2771 	int time_hash;
2772 
2773 	if (!(x->bm_flags & BW_METER_LEQ))
2774 		return;	/* XXX: we schedule timers only for "<=" entries */
2775 
2776 	/* Reset the bw_meter entry */
2777 	x->bm_start_time = *nowp;
2778 	x->bm_measured.b_packets = 0;
2779 	x->bm_measured.b_bytes = 0;
2780 	x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2781 
2782 	/* Compute the timeout hash value and insert the entry */
2783 	BW_METER_TIMEHASH(x, time_hash);
2784 	x->bm_time_next = bw_meter_timers[time_hash];
2785 	bw_meter_timers[time_hash] = x;
2786 	x->bm_time_hash = time_hash;
2787 }
2788 
2789 /*
2790  * Unschedule the periodic timer that processes bw_meter entry of type "<="
2791  * by removing the entry from the proper hash bucket.
2792  */
2793 static void
2794 unschedule_bw_meter(struct bw_meter *x)
2795 {
2796 	int time_hash;
2797 	struct bw_meter *prev, *tmp;
2798 
2799 	if (!(x->bm_flags & BW_METER_LEQ))
2800 		return;	/* XXX: we schedule timers only for "<=" entries */
2801 
2802 	/* Compute the timeout hash value and delete the entry */
2803 	time_hash = x->bm_time_hash;
2804 	if (time_hash >= BW_METER_BUCKETS)
2805 		return;		/* Entry was not scheduled */
2806 
2807 	for (prev = NULL, tmp = bw_meter_timers[time_hash];
2808 	    tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2809 		if (tmp == x)
2810 			break;
2811 
2812 	if (tmp == NULL)
2813 		panic("unschedule_bw_meter: bw_meter entry not found");
2814 
2815 	if (prev != NULL)
2816 		prev->bm_time_next = x->bm_time_next;
2817 	else
2818 		bw_meter_timers[time_hash] = x->bm_time_next;
2819 
2820 	x->bm_time_next = NULL;
2821 	x->bm_time_hash = BW_METER_BUCKETS;
2822 }
2823 
2824 /*
2825  * Process all "<=" type of bw_meter that should be processed now,
2826  * and for each entry prepare an upcall if necessary. Each processed
2827  * entry is rescheduled again for the (periodic) processing.
2828  *
2829  * This is run periodically (once per second normally). On each round,
2830  * all the potentially matching entries are in the hash slot that we are
2831  * looking at.
2832  */
2833 static void
2834 bw_meter_process()
2835 {
2836 	int s;
2837 	static uint32_t last_tv_sec;	/* last time we processed this */
2838 
2839 	uint32_t loops;
2840 	int i;
2841 	struct timeval now, process_endtime;
2842 
2843 	microtime(&now);
2844 	if (last_tv_sec == now.tv_sec)
2845 		return;		/* nothing to do */
2846 
2847 	loops = now.tv_sec - last_tv_sec;
2848 	last_tv_sec = now.tv_sec;
2849 	if (loops > BW_METER_BUCKETS)
2850 		loops = BW_METER_BUCKETS;
2851 
2852 	s = splsoftnet();
2853 	/*
2854 	 * Process all bins of bw_meter entries from the one after the last
2855 	 * processed to the current one. On entry, i points to the last bucket
2856 	 * visited, so we need to increment i at the beginning of the loop.
2857 	 */
2858 	for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2859 		struct bw_meter *x, *tmp_list;
2860 
2861 		if (++i >= BW_METER_BUCKETS)
2862 			i = 0;
2863 
2864 		/* Disconnect the list of bw_meter entries from the bin */
2865 		tmp_list = bw_meter_timers[i];
2866 		bw_meter_timers[i] = NULL;
2867 
2868 		/* Process the list of bw_meter entries */
2869 		while (tmp_list != NULL) {
2870 			x = tmp_list;
2871 			tmp_list = tmp_list->bm_time_next;
2872 
2873 			/* Test if the time interval is over */
2874 			process_endtime = x->bm_start_time;
2875 			BW_TIMEVALADD(&process_endtime,
2876 			    &x->bm_threshold.b_time);
2877 			if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2878 				/* Not yet: reschedule, but don't reset */
2879 				int time_hash;
2880 
2881 				BW_METER_TIMEHASH(x, time_hash);
2882 				if (time_hash == i &&
2883 				    process_endtime.tv_sec == now.tv_sec) {
2884 					/*
2885 					 * XXX: somehow the bin processing is
2886 					 * a bit ahead of time. Put the entry
2887 					 * in the next bin.
2888 					 */
2889 					if (++time_hash >= BW_METER_BUCKETS)
2890 						time_hash = 0;
2891 				}
2892 				x->bm_time_next = bw_meter_timers[time_hash];
2893 				bw_meter_timers[time_hash] = x;
2894 				x->bm_time_hash = time_hash;
2895 
2896 				continue;
2897 			}
2898 
2899 			/* Test if we should deliver an upcall */
2900 			if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2901 			    (x->bm_measured.b_packets <=
2902 			    x->bm_threshold.b_packets)) ||
2903 			    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2904 			    (x->bm_measured.b_bytes <=
2905 			    x->bm_threshold.b_bytes))) {
2906 				/* Prepare an upcall for delivery */
2907 				bw_meter_prepare_upcall(x, &now);
2908 			}
2909 
2910 			/* Reschedule for next processing */
2911 			schedule_bw_meter(x, &now);
2912 		}
2913 	}
2914 
2915 	/* Send all upcalls that are pending delivery */
2916 	bw_upcalls_send();
2917 
2918 	splx(s);
2919 }
2920 
2921 /*
2922  * A periodic function for sending all upcalls that are pending delivery
2923  */
2924 static void
2925 expire_bw_upcalls_send(void *unused)
2926 {
2927 	int s;
2928 
2929 	s = splsoftnet();
2930 	bw_upcalls_send();
2931 	splx(s);
2932 
2933 	timeout_add(&bw_upcalls_ch, BW_UPCALLS_PERIOD);
2934 }
2935 
2936 /*
2937  * A periodic function for periodic scanning of the multicast forwarding
2938  * table for processing all "<=" bw_meter entries.
2939  */
2940 static void
2941 expire_bw_meter_process(void *unused)
2942 {
2943 	if (mrt_api_config & MRT_MFC_BW_UPCALL)
2944 		bw_meter_process();
2945 
2946 	timeout_add(&bw_meter_ch, BW_METER_PERIOD);
2947 }
2948 
2949 /*
2950  * End of bandwidth monitoring code
2951  */
2952 
2953 #ifdef PIM
2954 /*
2955  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2956  */
2957 static int
2958 pim_register_send(struct ip *ip, struct vif *vifp,
2959 	struct mbuf *m, struct mfc *rt)
2960 {
2961 	struct mbuf *mb_copy, *mm;
2962 
2963 	if (mrtdebug & DEBUG_PIM)
2964 		log(LOG_DEBUG, "pim_register_send: ");
2965 
2966 	mb_copy = pim_register_prepare(ip, m);
2967 	if (mb_copy == NULL)
2968 		return (ENOBUFS);
2969 
2970 	/*
2971 	 * Send all the fragments. Note that the mbuf for each fragment
2972 	 * is freed by the sending machinery.
2973 	 */
2974 	for (mm = mb_copy; mm; mm = mb_copy) {
2975 		mb_copy = mm->m_nextpkt;
2976 		mm->m_nextpkt = NULL;
2977 		mm = m_pullup(mm, sizeof(struct ip));
2978 		if (mm != NULL) {
2979 			ip = mtod(mm, struct ip *);
2980 			if ((mrt_api_config & MRT_MFC_RP) &&
2981 			    !in_nullhost(rt->mfc_rp)) {
2982 				pim_register_send_rp(ip, vifp, mm, rt);
2983 			} else {
2984 				pim_register_send_upcall(ip, vifp, mm, rt);
2985 			}
2986 		}
2987 	}
2988 
2989 	return (0);
2990 }
2991 
2992 /*
2993  * Return a copy of the data packet that is ready for PIM Register
2994  * encapsulation.
2995  * XXX: Note that in the returned copy the IP header is a valid one.
2996  */
2997 static struct mbuf *
2998 pim_register_prepare(struct ip *ip, struct mbuf *m)
2999 {
3000 	struct mbuf *mb_copy = NULL;
3001 	int mtu;
3002 
3003 	/* Take care of delayed checksums */
3004 	if (m->m_pkthdr.csum_flags & (M_TCPV4_CSUM_OUT | M_UDPV4_CSUM_OUT)) {
3005 		in_delayed_cksum(m);
3006 		m->m_pkthdr.csum_flags &=
3007 		    ~(M_UDPV4_CSUM_OUT | M_TCPV4_CSUM_OUT);
3008 	}
3009 
3010 	/*
3011 	 * Copy the old packet & pullup its IP header into the
3012 	 * new mbuf so we can modify it.
3013 	 */
3014 	mb_copy = m_copy(m, 0, M_COPYALL);
3015 	if (mb_copy == NULL)
3016 		return (NULL);
3017 	mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3018 	if (mb_copy == NULL)
3019 		return (NULL);
3020 
3021 	/* take care of the TTL */
3022 	ip = mtod(mb_copy, struct ip *);
3023 	--ip->ip_ttl;
3024 
3025 	/* Compute the MTU after the PIM Register encapsulation */
3026 	mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3027 
3028 	if (ntohs(ip->ip_len) <= mtu) {
3029 		/* Turn the IP header into a valid one */
3030 		ip->ip_sum = 0;
3031 		ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3032 	} else {
3033 		/* Fragment the packet */
3034 		if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3035 			/* XXX: mb_copy was freed by ip_fragment() */
3036 			return (NULL);
3037 		}
3038 	}
3039 	return (mb_copy);
3040 }
3041 
3042 /*
3043  * Send an upcall with the data packet to the user-level process.
3044  */
3045 static int
3046 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3047 	struct mbuf *mb_copy, struct mfc *rt)
3048 {
3049 	struct mbuf *mb_first;
3050 	int len = ntohs(ip->ip_len);
3051 	struct igmpmsg *im;
3052 	struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
3053 
3054 	/* Add a new mbuf with an upcall header */
3055 	MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3056 	if (mb_first == NULL) {
3057 		m_freem(mb_copy);
3058 		return (ENOBUFS);
3059 	}
3060 	mb_first->m_data += max_linkhdr;
3061 	mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3062 	mb_first->m_len = sizeof(struct igmpmsg);
3063 	mb_first->m_next = mb_copy;
3064 
3065 	/* Send message to routing daemon */
3066 	im = mtod(mb_first, struct igmpmsg *);
3067 	im->im_msgtype = IGMPMSG_WHOLEPKT;
3068 	im->im_mbz = 0;
3069 	im->im_vif = vifp - viftable;
3070 	im->im_src = ip->ip_src;
3071 	im->im_dst = ip->ip_dst;
3072 
3073 	k_igmpsrc.sin_addr = ip->ip_src;
3074 
3075 	mrtstat.mrts_upcalls++;
3076 
3077 	if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3078 		if (mrtdebug & DEBUG_PIM)
3079 			log(LOG_WARNING, "mcast: pim_register_send_upcall: "
3080 			    "ip_mrouter socket queue full");
3081 		++mrtstat.mrts_upq_sockfull;
3082 		return (ENOBUFS);
3083 	}
3084 
3085 	/* Keep statistics */
3086 	pimstat.pims_snd_registers_msgs++;
3087 	pimstat.pims_snd_registers_bytes += len;
3088 
3089 	return (0);
3090 }
3091 
3092 /*
3093  * Encapsulate the data packet in PIM Register message and send it to the RP.
3094  */
3095 static int
3096 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3097 	struct mbuf *mb_copy, struct mfc *rt)
3098 {
3099 	struct mbuf *mb_first;
3100 	struct ip *ip_outer;
3101 	struct pim_encap_pimhdr *pimhdr;
3102 	int len = ntohs(ip->ip_len);
3103 	vifi_t vifi = rt->mfc_parent;
3104 
3105 	if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3106 		m_freem(mb_copy);
3107 		return (EADDRNOTAVAIL);		/* The iif vif is invalid */
3108 	}
3109 
3110 	/* Add a new mbuf with the encapsulating header */
3111 	MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3112 	if (mb_first == NULL) {
3113 		m_freem(mb_copy);
3114 		return (ENOBUFS);
3115 	}
3116 	mb_first->m_data += max_linkhdr;
3117 	mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3118 	mb_first->m_next = mb_copy;
3119 
3120 	mb_first->m_pkthdr.len = len + mb_first->m_len;
3121 
3122 	/* Fill in the encapsulating IP and PIM header */
3123 	ip_outer = mtod(mb_first, struct ip *);
3124 	*ip_outer = pim_encap_iphdr;
3125 	ip_outer->ip_id = htons(ip_randomid());
3126 	ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3127 	    sizeof(pim_encap_pimhdr));
3128 	ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3129 	ip_outer->ip_dst = rt->mfc_rp;
3130 	/*
3131 	 * Copy the inner header TOS to the outer header, and take care of the
3132 	 * IP_DF bit.
3133 	 */
3134 	ip_outer->ip_tos = ip->ip_tos;
3135 	if (ntohs(ip->ip_off) & IP_DF)
3136 		ip_outer->ip_off |= htons(IP_DF);
3137 	pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
3138 	    + sizeof(pim_encap_iphdr));
3139 	*pimhdr = pim_encap_pimhdr;
3140 	/* If the iif crosses a border, set the Border-bit */
3141 	if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3142 		pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3143 
3144 	mb_first->m_data += sizeof(pim_encap_iphdr);
3145 	pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3146 	mb_first->m_data -= sizeof(pim_encap_iphdr);
3147 
3148 	if (vifp->v_rate_limit == 0)
3149 		tbf_send_packet(vifp, mb_first);
3150 	else
3151 		tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3152 
3153 	/* Keep statistics */
3154 	pimstat.pims_snd_registers_msgs++;
3155 	pimstat.pims_snd_registers_bytes += len;
3156 
3157 	return (0);
3158 }
3159 
3160 /*
3161  * PIM-SMv2 and PIM-DM messages processing.
3162  * Receives and verifies the PIM control messages, and passes them
3163  * up to the listening socket, using rip_input().
3164  * The only message with special processing is the PIM_REGISTER message
3165  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3166  * is passed to if_simloop().
3167  */
3168 void
3169 pim_input(struct mbuf *m, ...)
3170 {
3171 	struct ip *ip = mtod(m, struct ip *);
3172 	struct pim *pim;
3173 	int minlen;
3174 	int datalen;
3175 	int ip_tos;
3176 	int iphlen;
3177 	va_list ap;
3178 
3179 	va_start(ap, m);
3180 	iphlen = va_arg(ap, int);
3181 	va_end(ap);
3182 
3183 	datalen = ntohs(ip->ip_len) - iphlen;
3184 
3185 	/* Keep statistics */
3186 	pimstat.pims_rcv_total_msgs++;
3187 	pimstat.pims_rcv_total_bytes += datalen;
3188 
3189 	/* Validate lengths */
3190 	if (datalen < PIM_MINLEN) {
3191 		pimstat.pims_rcv_tooshort++;
3192 		log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3193 		    datalen, (u_long)ip->ip_src.s_addr);
3194 		m_freem(m);
3195 		return;
3196 	}
3197 
3198 	/*
3199 	 * If the packet is at least as big as a REGISTER, go agead
3200 	 * and grab the PIM REGISTER header size, to avoid another
3201 	 * possible m_pullup() later.
3202 	 *
3203 	 * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
3204 	 * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3205 	 */
3206 	minlen = iphlen + (datalen >= PIM_REG_MINLEN ?
3207 	    PIM_REG_MINLEN : PIM_MINLEN);
3208 	/*
3209 	 * Get the IP and PIM headers in contiguous memory, and
3210 	 * possibly the PIM REGISTER header.
3211 	 */
3212 	if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3213 	    (m = m_pullup(m, minlen)) == NULL) {
3214 		log(LOG_ERR, "pim_input: m_pullup failure\n");
3215 		return;
3216 	}
3217 	/* m_pullup() may have given us a new mbuf so reset ip. */
3218 	ip = mtod(m, struct ip *);
3219 	ip_tos = ip->ip_tos;
3220 
3221 	/* adjust mbuf to point to the PIM header */
3222 	m->m_data += iphlen;
3223 	m->m_len  -= iphlen;
3224 	pim = mtod(m, struct pim *);
3225 
3226 	/*
3227 	 * Validate checksum. If PIM REGISTER, exclude the data packet.
3228 	 *
3229 	 * XXX: some older PIMv2 implementations don't make this distinction,
3230 	 * so for compatibility reason perform the checksum over part of the
3231 	 * message, and if error, then over the whole message.
3232 	 */
3233 	if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER &&
3234 	    in_cksum(m, PIM_MINLEN) == 0) {
3235 		/* do nothing, checksum okay */
3236 	} else if (in_cksum(m, datalen)) {
3237 		pimstat.pims_rcv_badsum++;
3238 		if (mrtdebug & DEBUG_PIM)
3239 			log(LOG_DEBUG, "pim_input: invalid checksum");
3240 		m_freem(m);
3241 		return;
3242 	}
3243 
3244 	/* PIM version check */
3245 	if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3246 		pimstat.pims_rcv_badversion++;
3247 		log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3248 		    PIM_VT_V(pim->pim_vt), PIM_VERSION);
3249 		m_freem(m);
3250 		return;
3251 	}
3252 
3253 	/* restore mbuf back to the outer IP */
3254 	m->m_data -= iphlen;
3255 	m->m_len  += iphlen;
3256 
3257 	if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3258 		/*
3259 		 * Since this is a REGISTER, we'll make a copy of the register
3260 		 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3261 		 * routing daemon.
3262 		 */
3263 		int s;
3264 		struct sockaddr_in dst = { sizeof(dst), AF_INET };
3265 		struct mbuf *mcp;
3266 		struct ip *encap_ip;
3267 		u_int32_t *reghdr;
3268 		struct ifnet *vifp;
3269 
3270 		s = splsoftnet();
3271 		if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3272 			splx(s);
3273 			if (mrtdebug & DEBUG_PIM)
3274 				log(LOG_DEBUG, "pim_input: register vif "
3275 				    "not set: %d\n", reg_vif_num);
3276 			m_freem(m);
3277 			return;
3278 		}
3279 		/* XXX need refcnt? */
3280 		vifp = viftable[reg_vif_num].v_ifp;
3281 		splx(s);
3282 
3283 		/* Validate length */
3284 		if (datalen < PIM_REG_MINLEN) {
3285 			pimstat.pims_rcv_tooshort++;
3286 			pimstat.pims_rcv_badregisters++;
3287 			log(LOG_ERR, "pim_input: register packet size "
3288 			    "too small %d from %lx\n",
3289 			    datalen, (u_long)ip->ip_src.s_addr);
3290 			m_freem(m);
3291 			return;
3292 		}
3293 
3294 		reghdr = (u_int32_t *)(pim + 1);
3295 		encap_ip = (struct ip *)(reghdr + 1);
3296 
3297 		if (mrtdebug & DEBUG_PIM) {
3298 			log(LOG_DEBUG, "pim_input[register], encap_ip: "
3299 			    "%lx -> %lx, encap_ip len %d\n",
3300 			    (u_long)ntohl(encap_ip->ip_src.s_addr),
3301 			    (u_long)ntohl(encap_ip->ip_dst.s_addr),
3302 			    ntohs(encap_ip->ip_len));
3303 		}
3304 
3305 		/* verify the version number of the inner packet */
3306 		if (encap_ip->ip_v != IPVERSION) {
3307 			pimstat.pims_rcv_badregisters++;
3308 			if (mrtdebug & DEBUG_PIM) {
3309 				log(LOG_DEBUG, "pim_input: invalid IP version"
3310 				    " (%d) of the inner packet\n",
3311 				    encap_ip->ip_v);
3312 			}
3313 			m_freem(m);
3314 			return;
3315 		}
3316 
3317 		/* verify the inner packet is destined to a mcast group */
3318 		if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3319 			pimstat.pims_rcv_badregisters++;
3320 			if (mrtdebug & DEBUG_PIM)
3321 				log(LOG_DEBUG,
3322 				    "pim_input: inner packet of register is"
3323 				    " not multicast %lx\n",
3324 				    (u_long)ntohl(encap_ip->ip_dst.s_addr));
3325 			m_freem(m);
3326 			return;
3327 		}
3328 
3329 		/* If a NULL_REGISTER, pass it to the daemon */
3330 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3331 			goto pim_input_to_daemon;
3332 
3333 		/*
3334 		 * Copy the TOS from the outer IP header to the inner
3335 		 * IP header.
3336 		 */
3337 		if (encap_ip->ip_tos != ip_tos) {
3338 			/* Outer TOS -> inner TOS */
3339 			encap_ip->ip_tos = ip_tos;
3340 			/* Recompute the inner header checksum. Sigh... */
3341 
3342 			/* adjust mbuf to point to the inner IP header */
3343 			m->m_data += (iphlen + PIM_MINLEN);
3344 			m->m_len  -= (iphlen + PIM_MINLEN);
3345 
3346 			encap_ip->ip_sum = 0;
3347 			encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3348 
3349 			/* restore mbuf to point back to the outer IP header */
3350 			m->m_data -= (iphlen + PIM_MINLEN);
3351 			m->m_len  += (iphlen + PIM_MINLEN);
3352 		}
3353 
3354 		/*
3355 		 * Decapsulate the inner IP packet and loopback to forward it
3356 		 * as a normal multicast packet. Also, make a copy of the
3357 		 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
3358 		 * to pass to the daemon later, so it can take the appropriate
3359 		 * actions (e.g., send back PIM_REGISTER_STOP).
3360 		 * XXX: here m->m_data points to the outer IP header.
3361 		 */
3362 		mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
3363 		if (mcp == NULL) {
3364 			log(LOG_ERR, "pim_input: pim register: could not "
3365 			    "copy register head\n");
3366 			m_freem(m);
3367 			return;
3368 		}
3369 
3370 		/* Keep statistics */
3371 		/* XXX: registers_bytes include only the encap. mcast pkt */
3372 		pimstat.pims_rcv_registers_msgs++;
3373 		pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3374 
3375 		/* forward the inner ip packet; point m_data at the inner ip. */
3376 		m_adj(m, iphlen + PIM_MINLEN);
3377 
3378 		if (mrtdebug & DEBUG_PIM) {
3379 			log(LOG_DEBUG,
3380 			    "pim_input: forwarding decapsulated register: "
3381 			    "src %lx, dst %lx, vif %d\n",
3382 			    (u_long)ntohl(encap_ip->ip_src.s_addr),
3383 			    (u_long)ntohl(encap_ip->ip_dst.s_addr),
3384 			    reg_vif_num);
3385 		}
3386 		/* NB: vifp was collected above; can it change on us? */
3387 		looutput(vifp, m, (struct sockaddr *)&dst,
3388 		    (struct rtentry *)NULL);
3389 
3390 		/* prepare the register head to send to the mrouting daemon */
3391 		m = mcp;
3392 	}
3393 
3394 pim_input_to_daemon:
3395 	/*
3396 	 * Pass the PIM message up to the daemon; if it is a Register message,
3397 	 * pass the 'head' only up to the daemon. This includes the
3398 	 * outer IP header, PIM header, PIM-Register header and the
3399 	 * inner IP header.
3400 	 * XXX: the outer IP header pkt size of a Register is not adjust to
3401 	 * reflect the fact that the inner multicast data is truncated.
3402 	 */
3403 	rip_input(m);
3404 
3405 	return;
3406 }
3407 
3408 /*
3409  * Sysctl for pim variables.
3410  */
3411 int
3412 pim_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
3413     void *newp, size_t newlen)
3414 {
3415 	/* All sysctl names at this level are terminal. */
3416 	if (namelen != 1)
3417 		return (ENOTDIR);
3418 
3419 	switch (name[0]) {
3420 	case PIMCTL_STATS:
3421 		if (newp != NULL)
3422 			return (EPERM);
3423 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
3424 		    &pimstat, sizeof(pimstat)));
3425 
3426 	default:
3427 		return (ENOPROTOOPT);
3428 	}
3429 	/* NOTREACHED */
3430 }
3431 
3432 
3433 #endif /* PIM */
3434