ip_mroute.c revision 152242
1/*-
2 * Copyright (c) 1989 Stephen Deering
3 * Copyright (c) 1992, 1993
4 *      The Regents of the University of California.  All rights reserved.
5 *
6 * This code is derived from software contributed to Berkeley by
7 * Stephen Deering of Stanford University.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 *    notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 *    notice, this list of conditions and the following disclaimer in the
16 *    documentation and/or other materials provided with the distribution.
17 * 4. Neither the name of the University nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
34 */
35
36/*
37 * IP multicast forwarding procedures
38 *
39 * Written by David Waitzman, BBN Labs, August 1988.
40 * Modified by Steve Deering, Stanford, February 1989.
41 * Modified by Mark J. Steiglitz, Stanford, May, 1991
42 * Modified by Van Jacobson, LBL, January 1993
43 * Modified by Ajit Thyagarajan, PARC, August 1993
44 * Modified by Bill Fenner, PARC, April 1995
45 * Modified by Ahmed Helmy, SGI, June 1996
46 * Modified by George Edmond Eddy (Rusty), ISI, February 1998
47 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
48 * Modified by Hitoshi Asaeda, WIDE, August 2000
49 * Modified by Pavlin Radoslavov, ICSI, October 2002
50 *
51 * MROUTING Revision: 3.5
52 * and PIM-SMv2 and PIM-DM support, advanced API support,
53 * bandwidth metering and signaling
54 *
55 * $FreeBSD: head/sys/netinet/ip_mroute.c 152242 2005-11-09 13:29:16Z ru $
56 */
57
58#include "opt_mac.h"
59#include "opt_mrouting.h"
60
61#ifdef PIM
62#define _PIM_VT 1
63#endif
64
65#include <sys/param.h>
66#include <sys/kernel.h>
67#include <sys/lock.h>
68#include <sys/mac.h>
69#include <sys/malloc.h>
70#include <sys/mbuf.h>
71#include <sys/module.h>
72#include <sys/protosw.h>
73#include <sys/signalvar.h>
74#include <sys/socket.h>
75#include <sys/socketvar.h>
76#include <sys/sockio.h>
77#include <sys/sx.h>
78#include <sys/sysctl.h>
79#include <sys/syslog.h>
80#include <sys/systm.h>
81#include <sys/time.h>
82#include <net/if.h>
83#include <net/netisr.h>
84#include <net/route.h>
85#include <netinet/in.h>
86#include <netinet/igmp.h>
87#include <netinet/in_systm.h>
88#include <netinet/in_var.h>
89#include <netinet/ip.h>
90#include <netinet/ip_encap.h>
91#include <netinet/ip_mroute.h>
92#include <netinet/ip_var.h>
93#ifdef PIM
94#include <netinet/pim.h>
95#include <netinet/pim_var.h>
96#endif
97#include <netinet/udp.h>
98#include <machine/in_cksum.h>
99
100/*
101 * Control debugging code for rsvp and multicast routing code.
102 * Can only set them with the debugger.
103 */
104static u_int    rsvpdebug;		/* non-zero enables debugging	*/
105
106static u_int	mrtdebug;		/* any set of the flags below	*/
107#define		DEBUG_MFC	0x02
108#define		DEBUG_FORWARD	0x04
109#define		DEBUG_EXPIRE	0x08
110#define		DEBUG_XMIT	0x10
111#define		DEBUG_PIM	0x20
112
113#define		VIFI_INVALID	((vifi_t) -1)
114
115#define M_HASCL(m)	((m)->m_flags & M_EXT)
116
117static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables");
118
119/*
120 * Locking.  We use two locks: one for the virtual interface table and
121 * one for the forwarding table.  These locks may be nested in which case
122 * the VIF lock must always be taken first.  Note that each lock is used
123 * to cover not only the specific data structure but also related data
124 * structures.  It may be better to add more fine-grained locking later;
125 * it's not clear how performance-critical this code is.
126 */
127
128static struct mrtstat	mrtstat;
129SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
130    &mrtstat, mrtstat,
131    "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)");
132
133static struct mfc	*mfctable[MFCTBLSIZ];
134SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
135    &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]",
136    "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)");
137
138static struct mtx mfc_mtx;
139#define	MFC_LOCK()	mtx_lock(&mfc_mtx)
140#define	MFC_UNLOCK()	mtx_unlock(&mfc_mtx)
141#define	MFC_LOCK_ASSERT()	do {					\
142	mtx_assert(&mfc_mtx, MA_OWNED);					\
143	NET_ASSERT_GIANT();						\
144} while (0)
145#define	MFC_LOCK_INIT()	mtx_init(&mfc_mtx, "mroute mfc table", NULL, MTX_DEF)
146#define	MFC_LOCK_DESTROY()	mtx_destroy(&mfc_mtx)
147
148static struct vif	viftable[MAXVIFS];
149SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
150    &viftable, sizeof(viftable), "S,vif[MAXVIFS]",
151    "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
152
153static struct mtx vif_mtx;
154#define	VIF_LOCK()	mtx_lock(&vif_mtx)
155#define	VIF_UNLOCK()	mtx_unlock(&vif_mtx)
156#define	VIF_LOCK_ASSERT()	mtx_assert(&vif_mtx, MA_OWNED)
157#define	VIF_LOCK_INIT()	mtx_init(&vif_mtx, "mroute vif table", NULL, MTX_DEF)
158#define	VIF_LOCK_DESTROY()	mtx_destroy(&vif_mtx)
159
160static u_char		nexpire[MFCTBLSIZ];
161
162static struct callout expire_upcalls_ch;
163
164#define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
165#define		UPCALL_EXPIRE	6		/* number of timeouts	*/
166
167/*
168 * Define the token bucket filter structures
169 * tbftable -> each vif has one of these for storing info
170 */
171
172static struct tbf tbftable[MAXVIFS];
173#define		TBF_REPROCESS	(hz / 100)	/* 100x / second */
174
175/*
176 * 'Interfaces' associated with decapsulator (so we can tell
177 * packets that went through it from ones that get reflected
178 * by a broken gateway).  These interfaces are never linked into
179 * the system ifnet list & no routes point to them.  I.e., packets
180 * can't be sent this way.  They only exist as a placeholder for
181 * multicast source verification.
182 */
183static struct ifnet multicast_decap_if[MAXVIFS];
184
185#define ENCAP_TTL 64
186#define ENCAP_PROTO IPPROTO_IPIP	/* 4 */
187
188/* prototype IP hdr for encapsulated packets */
189static struct ip multicast_encap_iphdr = {
190#if BYTE_ORDER == LITTLE_ENDIAN
191	sizeof(struct ip) >> 2, IPVERSION,
192#else
193	IPVERSION, sizeof(struct ip) >> 2,
194#endif
195	0,				/* tos */
196	sizeof(struct ip),		/* total length */
197	0,				/* id */
198	0,				/* frag offset */
199	ENCAP_TTL, ENCAP_PROTO,
200	0,				/* checksum */
201};
202
203/*
204 * Bandwidth meter variables and constants
205 */
206static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
207/*
208 * Pending timeouts are stored in a hash table, the key being the
209 * expiration time. Periodically, the entries are analysed and processed.
210 */
211#define BW_METER_BUCKETS	1024
212static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
213static struct callout bw_meter_ch;
214#define BW_METER_PERIOD (hz)		/* periodical handling of bw meters */
215
216/*
217 * Pending upcalls are stored in a vector which is flushed when
218 * full, or periodically
219 */
220static struct bw_upcall	bw_upcalls[BW_UPCALLS_MAX];
221static u_int	bw_upcalls_n; /* # of pending upcalls */
222static struct callout bw_upcalls_ch;
223#define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
224
225#ifdef PIM
226static struct pimstat pimstat;
227SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
228    &pimstat, pimstat,
229    "PIM Statistics (struct pimstat, netinet/pim_var.h)");
230
231/*
232 * Note: the PIM Register encapsulation adds the following in front of a
233 * data packet:
234 *
235 * struct pim_encap_hdr {
236 *    struct ip ip;
237 *    struct pim_encap_pimhdr  pim;
238 * }
239 *
240 */
241
242struct pim_encap_pimhdr {
243	struct pim pim;
244	uint32_t   flags;
245};
246
247static struct ip pim_encap_iphdr = {
248#if BYTE_ORDER == LITTLE_ENDIAN
249	sizeof(struct ip) >> 2,
250	IPVERSION,
251#else
252	IPVERSION,
253	sizeof(struct ip) >> 2,
254#endif
255	0,			/* tos */
256	sizeof(struct ip),	/* total length */
257	0,			/* id */
258	0,			/* frag offset */
259	ENCAP_TTL,
260	IPPROTO_PIM,
261	0,			/* checksum */
262};
263
264static struct pim_encap_pimhdr pim_encap_pimhdr = {
265    {
266	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
267	0,			/* reserved */
268	0,			/* checksum */
269    },
270    0				/* flags */
271};
272
273static struct ifnet multicast_register_if;
274static vifi_t reg_vif_num = VIFI_INVALID;
275#endif /* PIM */
276
277/*
278 * Private variables.
279 */
280static vifi_t	   numvifs;
281static const struct encaptab *encap_cookie;
282
283/*
284 * one-back cache used by mroute_encapcheck to locate a tunnel's vif
285 * given a datagram's src ip address.
286 */
287static u_long last_encap_src;
288static struct vif *last_encap_vif;
289
290/*
291 * Callout for queue processing.
292 */
293static struct callout tbf_reprocess_ch;
294
295static u_long	X_ip_mcast_src(int vifi);
296static int	X_ip_mforward(struct ip *ip, struct ifnet *ifp,
297			struct mbuf *m, struct ip_moptions *imo);
298static int	X_ip_mrouter_done(void);
299static int	X_ip_mrouter_get(struct socket *so, struct sockopt *m);
300static int	X_ip_mrouter_set(struct socket *so, struct sockopt *m);
301static int	X_legal_vif_num(int vif);
302static int	X_mrt_ioctl(int cmd, caddr_t data);
303
304static int get_sg_cnt(struct sioc_sg_req *);
305static int get_vif_cnt(struct sioc_vif_req *);
306static int ip_mrouter_init(struct socket *, int);
307static int add_vif(struct vifctl *);
308static int del_vif(vifi_t);
309static int add_mfc(struct mfcctl2 *);
310static int del_mfc(struct mfcctl2 *);
311static int set_api_config(uint32_t *); /* chose API capabilities */
312static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
313static int set_assert(int);
314static void expire_upcalls(void *);
315static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
316static void phyint_send(struct ip *, struct vif *, struct mbuf *);
317static void encap_send(struct ip *, struct vif *, struct mbuf *);
318static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
319static void tbf_queue(struct vif *, struct mbuf *);
320static void tbf_process_q(struct vif *);
321static void tbf_reprocess_q(void *);
322static int tbf_dq_sel(struct vif *, struct ip *);
323static void tbf_send_packet(struct vif *, struct mbuf *);
324static void tbf_update_tokens(struct vif *);
325static int priority(struct vif *, struct ip *);
326
327/*
328 * Bandwidth monitoring
329 */
330static void free_bw_list(struct bw_meter *list);
331static int add_bw_upcall(struct bw_upcall *);
332static int del_bw_upcall(struct bw_upcall *);
333static void bw_meter_receive_packet(struct bw_meter *x, int plen,
334		struct timeval *nowp);
335static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp);
336static void bw_upcalls_send(void);
337static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp);
338static void unschedule_bw_meter(struct bw_meter *x);
339static void bw_meter_process(void);
340static void expire_bw_upcalls_send(void *);
341static void expire_bw_meter_process(void *);
342
343#ifdef PIM
344static int pim_register_send(struct ip *, struct vif *,
345		struct mbuf *, struct mfc *);
346static int pim_register_send_rp(struct ip *, struct vif *,
347		struct mbuf *, struct mfc *);
348static int pim_register_send_upcall(struct ip *, struct vif *,
349		struct mbuf *, struct mfc *);
350static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
351#endif
352
353/*
354 * whether or not special PIM assert processing is enabled.
355 */
356static int pim_assert;
357/*
358 * Rate limit for assert notification messages, in usec
359 */
360#define ASSERT_MSG_TIME		3000000
361
362/*
363 * Kernel multicast routing API capabilities and setup.
364 * If more API capabilities are added to the kernel, they should be
365 * recorded in `mrt_api_support'.
366 */
367static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
368					 MRT_MFC_FLAGS_BORDER_VIF |
369					 MRT_MFC_RP |
370					 MRT_MFC_BW_UPCALL);
371static uint32_t mrt_api_config = 0;
372
373/*
374 * Hash function for a source, group entry
375 */
376#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
377			((g) >> 20) ^ ((g) >> 10) ^ (g))
378
379/*
380 * Find a route for a given origin IP address and Multicast group address
381 * Type of service parameter to be added in the future!!!
382 * Statistics are updated by the caller if needed
383 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
384 */
385static struct mfc *
386mfc_find(in_addr_t o, in_addr_t g)
387{
388    struct mfc *rt;
389
390    MFC_LOCK_ASSERT();
391
392    for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next)
393	if ((rt->mfc_origin.s_addr == o) &&
394		(rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL))
395	    break;
396    return rt;
397}
398
399/*
400 * Macros to compute elapsed time efficiently
401 * Borrowed from Van Jacobson's scheduling code
402 */
403#define TV_DELTA(a, b, delta) {					\
404	int xxs;						\
405	delta = (a).tv_usec - (b).tv_usec;			\
406	if ((xxs = (a).tv_sec - (b).tv_sec)) {			\
407		switch (xxs) {					\
408		case 2:						\
409		      delta += 1000000;				\
410		      /* FALLTHROUGH */				\
411		case 1:						\
412		      delta += 1000000;				\
413		      break;					\
414		default:					\
415		      delta += (1000000 * xxs);			\
416		}						\
417	}							\
418}
419
420#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
421	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
422
423/*
424 * Handle MRT setsockopt commands to modify the multicast routing tables.
425 */
426static int
427X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
428{
429    int	error, optval;
430    vifi_t	vifi;
431    struct	vifctl vifc;
432    struct	mfcctl2 mfc;
433    struct	bw_upcall bw_upcall;
434    uint32_t	i;
435
436    if (so != ip_mrouter && sopt->sopt_name != MRT_INIT)
437	return EPERM;
438
439    error = 0;
440    switch (sopt->sopt_name) {
441    case MRT_INIT:
442	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
443	if (error)
444	    break;
445	error = ip_mrouter_init(so, optval);
446	break;
447
448    case MRT_DONE:
449	error = ip_mrouter_done();
450	break;
451
452    case MRT_ADD_VIF:
453	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
454	if (error)
455	    break;
456	error = add_vif(&vifc);
457	break;
458
459    case MRT_DEL_VIF:
460	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
461	if (error)
462	    break;
463	error = del_vif(vifi);
464	break;
465
466    case MRT_ADD_MFC:
467    case MRT_DEL_MFC:
468	/*
469	 * select data size depending on API version.
470	 */
471	if (sopt->sopt_name == MRT_ADD_MFC &&
472		mrt_api_config & MRT_API_FLAGS_ALL) {
473	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
474				sizeof(struct mfcctl2));
475	} else {
476	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
477				sizeof(struct mfcctl));
478	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
479			sizeof(mfc) - sizeof(struct mfcctl));
480	}
481	if (error)
482	    break;
483	if (sopt->sopt_name == MRT_ADD_MFC)
484	    error = add_mfc(&mfc);
485	else
486	    error = del_mfc(&mfc);
487	break;
488
489    case MRT_ASSERT:
490	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
491	if (error)
492	    break;
493	set_assert(optval);
494	break;
495
496    case MRT_API_CONFIG:
497	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
498	if (!error)
499	    error = set_api_config(&i);
500	if (!error)
501	    error = sooptcopyout(sopt, &i, sizeof i);
502	break;
503
504    case MRT_ADD_BW_UPCALL:
505    case MRT_DEL_BW_UPCALL:
506	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
507				sizeof bw_upcall);
508	if (error)
509	    break;
510	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
511	    error = add_bw_upcall(&bw_upcall);
512	else
513	    error = del_bw_upcall(&bw_upcall);
514	break;
515
516    default:
517	error = EOPNOTSUPP;
518	break;
519    }
520    return error;
521}
522
523/*
524 * Handle MRT getsockopt commands
525 */
526static int
527X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
528{
529    int error;
530    static int version = 0x0305; /* !!! why is this here? XXX */
531
532    switch (sopt->sopt_name) {
533    case MRT_VERSION:
534	error = sooptcopyout(sopt, &version, sizeof version);
535	break;
536
537    case MRT_ASSERT:
538	error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert);
539	break;
540
541    case MRT_API_SUPPORT:
542	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
543	break;
544
545    case MRT_API_CONFIG:
546	error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config);
547	break;
548
549    default:
550	error = EOPNOTSUPP;
551	break;
552    }
553    return error;
554}
555
556/*
557 * Handle ioctl commands to obtain information from the cache
558 */
559static int
560X_mrt_ioctl(int cmd, caddr_t data)
561{
562    int error = 0;
563
564    /*
565     * Currently the only function calling this ioctl routine is rtioctl().
566     * Typically, only root can create the raw socket in order to execute
567     * this ioctl method, however the request might be coming from a prison
568     */
569    error = suser(curthread);
570    if (error)
571	return (error);
572    switch (cmd) {
573    case (SIOCGETVIFCNT):
574	error = get_vif_cnt((struct sioc_vif_req *)data);
575	break;
576
577    case (SIOCGETSGCNT):
578	error = get_sg_cnt((struct sioc_sg_req *)data);
579	break;
580
581    default:
582	error = EINVAL;
583	break;
584    }
585    return error;
586}
587
588/*
589 * returns the packet, byte, rpf-failure count for the source group provided
590 */
591static int
592get_sg_cnt(struct sioc_sg_req *req)
593{
594    struct mfc *rt;
595
596    MFC_LOCK();
597    rt = mfc_find(req->src.s_addr, req->grp.s_addr);
598    if (rt == NULL) {
599	MFC_UNLOCK();
600	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
601	return EADDRNOTAVAIL;
602    }
603    req->pktcnt = rt->mfc_pkt_cnt;
604    req->bytecnt = rt->mfc_byte_cnt;
605    req->wrong_if = rt->mfc_wrong_if;
606    MFC_UNLOCK();
607    return 0;
608}
609
610/*
611 * returns the input and output packet and byte counts on the vif provided
612 */
613static int
614get_vif_cnt(struct sioc_vif_req *req)
615{
616    vifi_t vifi = req->vifi;
617
618    VIF_LOCK();
619    if (vifi >= numvifs) {
620	VIF_UNLOCK();
621	return EINVAL;
622    }
623
624    req->icount = viftable[vifi].v_pkt_in;
625    req->ocount = viftable[vifi].v_pkt_out;
626    req->ibytes = viftable[vifi].v_bytes_in;
627    req->obytes = viftable[vifi].v_bytes_out;
628    VIF_UNLOCK();
629
630    return 0;
631}
632
633static void
634ip_mrouter_reset(void)
635{
636    bzero((caddr_t)mfctable, sizeof(mfctable));
637    bzero((caddr_t)nexpire, sizeof(nexpire));
638
639    pim_assert = 0;
640    mrt_api_config = 0;
641
642    callout_init(&expire_upcalls_ch, NET_CALLOUT_MPSAFE);
643
644    bw_upcalls_n = 0;
645    bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers));
646    callout_init(&bw_upcalls_ch, NET_CALLOUT_MPSAFE);
647    callout_init(&bw_meter_ch, NET_CALLOUT_MPSAFE);
648
649    callout_init(&tbf_reprocess_ch, NET_CALLOUT_MPSAFE);
650}
651
652static struct mtx mrouter_mtx;		/* used to synch init/done work */
653
654/*
655 * Enable multicast routing
656 */
657static int
658ip_mrouter_init(struct socket *so, int version)
659{
660    if (mrtdebug)
661	log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
662	    so->so_type, so->so_proto->pr_protocol);
663
664    if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
665	return EOPNOTSUPP;
666
667    if (version != 1)
668	return ENOPROTOOPT;
669
670    mtx_lock(&mrouter_mtx);
671
672    if (ip_mrouter != NULL) {
673	mtx_unlock(&mrouter_mtx);
674	return EADDRINUSE;
675    }
676
677    callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
678
679    callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
680	expire_bw_upcalls_send, NULL);
681    callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
682
683    ip_mrouter = so;
684
685    mtx_unlock(&mrouter_mtx);
686
687    if (mrtdebug)
688	log(LOG_DEBUG, "ip_mrouter_init\n");
689
690    return 0;
691}
692
693/*
694 * Disable multicast routing
695 */
696static int
697X_ip_mrouter_done(void)
698{
699    vifi_t vifi;
700    int i;
701    struct ifnet *ifp;
702    struct ifreq ifr;
703    struct mfc *rt;
704    struct rtdetq *rte;
705
706    mtx_lock(&mrouter_mtx);
707
708    if (ip_mrouter == NULL) {
709	mtx_unlock(&mrouter_mtx);
710	return EINVAL;
711    }
712
713    /*
714     * Detach/disable hooks to the reset of the system.
715     */
716    ip_mrouter = NULL;
717    mrt_api_config = 0;
718
719    VIF_LOCK();
720    if (encap_cookie) {
721	const struct encaptab *c = encap_cookie;
722	encap_cookie = NULL;
723	encap_detach(c);
724    }
725    VIF_UNLOCK();
726
727    callout_stop(&tbf_reprocess_ch);
728
729    VIF_LOCK();
730    /*
731     * For each phyint in use, disable promiscuous reception of all IP
732     * multicasts.
733     */
734    for (vifi = 0; vifi < numvifs; vifi++) {
735	if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
736		!(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
737	    struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
738
739	    so->sin_len = sizeof(struct sockaddr_in);
740	    so->sin_family = AF_INET;
741	    so->sin_addr.s_addr = INADDR_ANY;
742	    ifp = viftable[vifi].v_ifp;
743	    if_allmulti(ifp, 0);
744	}
745    }
746    bzero((caddr_t)tbftable, sizeof(tbftable));
747    bzero((caddr_t)viftable, sizeof(viftable));
748    numvifs = 0;
749    pim_assert = 0;
750    VIF_UNLOCK();
751
752    /*
753     * Free all multicast forwarding cache entries.
754     */
755    callout_stop(&expire_upcalls_ch);
756    callout_stop(&bw_upcalls_ch);
757    callout_stop(&bw_meter_ch);
758
759    MFC_LOCK();
760    for (i = 0; i < MFCTBLSIZ; i++) {
761	for (rt = mfctable[i]; rt != NULL; ) {
762	    struct mfc *nr = rt->mfc_next;
763
764	    for (rte = rt->mfc_stall; rte != NULL; ) {
765		struct rtdetq *n = rte->next;
766
767		m_freem(rte->m);
768		free(rte, M_MRTABLE);
769		rte = n;
770	    }
771	    free_bw_list(rt->mfc_bw_meter);
772	    free(rt, M_MRTABLE);
773	    rt = nr;
774	}
775    }
776    bzero((caddr_t)mfctable, sizeof(mfctable));
777    bzero((caddr_t)nexpire, sizeof(nexpire));
778    bw_upcalls_n = 0;
779    bzero(bw_meter_timers, sizeof(bw_meter_timers));
780    MFC_UNLOCK();
781
782    /*
783     * Reset de-encapsulation cache
784     */
785    last_encap_src = INADDR_ANY;
786    last_encap_vif = NULL;
787#ifdef PIM
788    reg_vif_num = VIFI_INVALID;
789#endif
790
791    mtx_unlock(&mrouter_mtx);
792
793    if (mrtdebug)
794	log(LOG_DEBUG, "ip_mrouter_done\n");
795
796    return 0;
797}
798
799/*
800 * Set PIM assert processing global
801 */
802static int
803set_assert(int i)
804{
805    if ((i != 1) && (i != 0))
806	return EINVAL;
807
808    pim_assert = i;
809
810    return 0;
811}
812
813/*
814 * Configure API capabilities
815 */
816int
817set_api_config(uint32_t *apival)
818{
819    int i;
820
821    /*
822     * We can set the API capabilities only if it is the first operation
823     * after MRT_INIT. I.e.:
824     *  - there are no vifs installed
825     *  - pim_assert is not enabled
826     *  - the MFC table is empty
827     */
828    if (numvifs > 0) {
829	*apival = 0;
830	return EPERM;
831    }
832    if (pim_assert) {
833	*apival = 0;
834	return EPERM;
835    }
836    for (i = 0; i < MFCTBLSIZ; i++) {
837	if (mfctable[i] != NULL) {
838	    *apival = 0;
839	    return EPERM;
840	}
841    }
842
843    mrt_api_config = *apival & mrt_api_support;
844    *apival = mrt_api_config;
845
846    return 0;
847}
848
849/*
850 * Decide if a packet is from a tunnelled peer.
851 * Return 0 if not, 64 if so.  XXX yuck.. 64 ???
852 */
853static int
854mroute_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
855{
856    struct ip *ip = mtod(m, struct ip *);
857    int hlen = ip->ip_hl << 2;
858
859    /*
860     * don't claim the packet if it's not to a multicast destination or if
861     * we don't have an encapsulating tunnel with the source.
862     * Note:  This code assumes that the remote site IP address
863     * uniquely identifies the tunnel (i.e., that this site has
864     * at most one tunnel with the remote site).
865     */
866    if (!IN_MULTICAST(ntohl(((struct ip *)((char *)ip+hlen))->ip_dst.s_addr)))
867	return 0;
868    if (ip->ip_src.s_addr != last_encap_src) {
869	struct vif *vifp = viftable;
870	struct vif *vife = vifp + numvifs;
871
872	last_encap_src = ip->ip_src.s_addr;
873	last_encap_vif = NULL;
874	for ( ; vifp < vife; ++vifp)
875	    if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
876		if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) == VIFF_TUNNEL)
877		    last_encap_vif = vifp;
878		break;
879	    }
880    }
881    if (last_encap_vif == NULL) {
882	last_encap_src = INADDR_ANY;
883	return 0;
884    }
885    return 64;
886}
887
888/*
889 * De-encapsulate a packet and feed it back through ip input (this
890 * routine is called whenever IP gets a packet that mroute_encap_func()
891 * claimed).
892 */
893static void
894mroute_encap_input(struct mbuf *m, int off)
895{
896    struct ip *ip = mtod(m, struct ip *);
897    int hlen = ip->ip_hl << 2;
898
899    if (hlen > sizeof(struct ip))
900	ip_stripoptions(m, (struct mbuf *) 0);
901    m->m_data += sizeof(struct ip);
902    m->m_len -= sizeof(struct ip);
903    m->m_pkthdr.len -= sizeof(struct ip);
904
905    m->m_pkthdr.rcvif = last_encap_vif->v_ifp;
906
907    netisr_queue(NETISR_IP, m);		/* mbuf is free'd on failure. */
908    /*
909     * normally we would need a "schednetisr(NETISR_IP)"
910     * here but we were called by ip_input and it is going
911     * to loop back & try to dequeue the packet we just
912     * queued as soon as we return so we avoid the
913     * unnecessary software interrrupt.
914     *
915     * XXX
916     * This no longer holds - we may have direct-dispatched the packet,
917     * or there may be a queue processing limit.
918     */
919}
920
921extern struct domain inetdomain;
922static struct protosw mroute_encap_protosw =
923{
924	.pr_type =		SOCK_RAW,
925	.pr_domain =		&inetdomain,
926	.pr_protocol =		IPPROTO_IPV4,
927	.pr_flags =		PR_ATOMIC|PR_ADDR,
928	.pr_input =		mroute_encap_input,
929	.pr_ctloutput =		rip_ctloutput,
930	.pr_usrreqs =		&rip_usrreqs
931};
932
933/*
934 * Add a vif to the vif table
935 */
936static int
937add_vif(struct vifctl *vifcp)
938{
939    struct vif *vifp = viftable + vifcp->vifc_vifi;
940    struct sockaddr_in sin = {sizeof sin, AF_INET};
941    struct ifaddr *ifa;
942    struct ifnet *ifp;
943    int error;
944    struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
945
946    VIF_LOCK();
947    if (vifcp->vifc_vifi >= MAXVIFS) {
948	VIF_UNLOCK();
949	return EINVAL;
950    }
951    if (vifp->v_lcl_addr.s_addr != INADDR_ANY) {
952	VIF_UNLOCK();
953	return EADDRINUSE;
954    }
955    if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) {
956	VIF_UNLOCK();
957	return EADDRNOTAVAIL;
958    }
959
960    /* Find the interface with an address in AF_INET family */
961#ifdef PIM
962    if (vifcp->vifc_flags & VIFF_REGISTER) {
963	/*
964	 * XXX: Because VIFF_REGISTER does not really need a valid
965	 * local interface (e.g. it could be 127.0.0.2), we don't
966	 * check its address.
967	 */
968	ifp = NULL;
969    } else
970#endif
971    {
972	sin.sin_addr = vifcp->vifc_lcl_addr;
973	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
974	if (ifa == NULL) {
975	    VIF_UNLOCK();
976	    return EADDRNOTAVAIL;
977	}
978	ifp = ifa->ifa_ifp;
979    }
980
981    if (vifcp->vifc_flags & VIFF_TUNNEL) {
982	if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
983	    /*
984	     * An encapsulating tunnel is wanted.  Tell
985	     * mroute_encap_input() to start paying attention
986	     * to encapsulated packets.
987	     */
988	    if (encap_cookie == NULL) {
989		int i;
990
991		encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
992				mroute_encapcheck,
993				(struct protosw *)&mroute_encap_protosw, NULL);
994
995		if (encap_cookie == NULL) {
996		    printf("ip_mroute: unable to attach encap\n");
997		    VIF_UNLOCK();
998		    return EIO;	/* XXX */
999		}
1000		for (i = 0; i < MAXVIFS; ++i) {
1001		    if_initname(&multicast_decap_if[i], "mdecap", i);
1002		}
1003	    }
1004	    /*
1005	     * Set interface to fake encapsulator interface
1006	     */
1007	    ifp = &multicast_decap_if[vifcp->vifc_vifi];
1008	    /*
1009	     * Prepare cached route entry
1010	     */
1011	    bzero(&vifp->v_route, sizeof(vifp->v_route));
1012	} else {
1013	    log(LOG_ERR, "source routed tunnels not supported\n");
1014	    VIF_UNLOCK();
1015	    return EOPNOTSUPP;
1016	}
1017#ifdef PIM
1018    } else if (vifcp->vifc_flags & VIFF_REGISTER) {
1019	ifp = &multicast_register_if;
1020	if (mrtdebug)
1021	    log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
1022		    (void *)&multicast_register_if);
1023	if (reg_vif_num == VIFI_INVALID) {
1024	    if_initname(&multicast_register_if, "register_vif", 0);
1025	    multicast_register_if.if_flags = IFF_LOOPBACK;
1026	    bzero(&vifp->v_route, sizeof(vifp->v_route));
1027	    reg_vif_num = vifcp->vifc_vifi;
1028	}
1029#endif
1030    } else {		/* Make sure the interface supports multicast */
1031	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
1032	    VIF_UNLOCK();
1033	    return EOPNOTSUPP;
1034	}
1035
1036	/* Enable promiscuous reception of all IP multicasts from the if */
1037	error = if_allmulti(ifp, 1);
1038	if (error) {
1039	    VIF_UNLOCK();
1040	    return error;
1041	}
1042    }
1043
1044    /* define parameters for the tbf structure */
1045    vifp->v_tbf = v_tbf;
1046    GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
1047    vifp->v_tbf->tbf_n_tok = 0;
1048    vifp->v_tbf->tbf_q_len = 0;
1049    vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
1050    vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
1051
1052    vifp->v_flags     = vifcp->vifc_flags;
1053    vifp->v_threshold = vifcp->vifc_threshold;
1054    vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
1055    vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
1056    vifp->v_ifp       = ifp;
1057    /* scaling up here allows division by 1024 in critical code */
1058    vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
1059    vifp->v_rsvp_on   = 0;
1060    vifp->v_rsvpd     = NULL;
1061    /* initialize per vif pkt counters */
1062    vifp->v_pkt_in    = 0;
1063    vifp->v_pkt_out   = 0;
1064    vifp->v_bytes_in  = 0;
1065    vifp->v_bytes_out = 0;
1066
1067    /* Adjust numvifs up if the vifi is higher than numvifs */
1068    if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
1069
1070    VIF_UNLOCK();
1071
1072    if (mrtdebug)
1073	log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n",
1074	    vifcp->vifc_vifi,
1075	    (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr),
1076	    (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
1077	    (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr),
1078	    vifcp->vifc_threshold,
1079	    vifcp->vifc_rate_limit);
1080
1081    return 0;
1082}
1083
1084/*
1085 * Delete a vif from the vif table
1086 */
1087static int
1088del_vif(vifi_t vifi)
1089{
1090    struct vif *vifp;
1091
1092    VIF_LOCK();
1093
1094    if (vifi >= numvifs) {
1095	VIF_UNLOCK();
1096	return EINVAL;
1097    }
1098    vifp = &viftable[vifi];
1099    if (vifp->v_lcl_addr.s_addr == INADDR_ANY) {
1100	VIF_UNLOCK();
1101	return EADDRNOTAVAIL;
1102    }
1103
1104    if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1105	if_allmulti(vifp->v_ifp, 0);
1106
1107    if (vifp == last_encap_vif) {
1108	last_encap_vif = NULL;
1109	last_encap_src = INADDR_ANY;
1110    }
1111
1112    /*
1113     * Free packets queued at the interface
1114     */
1115    while (vifp->v_tbf->tbf_q) {
1116	struct mbuf *m = vifp->v_tbf->tbf_q;
1117
1118	vifp->v_tbf->tbf_q = m->m_act;
1119	m_freem(m);
1120    }
1121
1122#ifdef PIM
1123    if (vifp->v_flags & VIFF_REGISTER)
1124	reg_vif_num = VIFI_INVALID;
1125#endif
1126
1127    bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
1128    bzero((caddr_t)vifp, sizeof (*vifp));
1129
1130    if (mrtdebug)
1131	log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs);
1132
1133    /* Adjust numvifs down */
1134    for (vifi = numvifs; vifi > 0; vifi--)
1135	if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY)
1136	    break;
1137    numvifs = vifi;
1138
1139    VIF_UNLOCK();
1140
1141    return 0;
1142}
1143
1144/*
1145 * update an mfc entry without resetting counters and S,G addresses.
1146 */
1147static void
1148update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1149{
1150    int i;
1151
1152    rt->mfc_parent = mfccp->mfcc_parent;
1153    for (i = 0; i < numvifs; i++) {
1154	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1155	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1156	    MRT_MFC_FLAGS_ALL;
1157    }
1158    /* set the RP address */
1159    if (mrt_api_config & MRT_MFC_RP)
1160	rt->mfc_rp = mfccp->mfcc_rp;
1161    else
1162	rt->mfc_rp.s_addr = INADDR_ANY;
1163}
1164
1165/*
1166 * fully initialize an mfc entry from the parameter.
1167 */
1168static void
1169init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1170{
1171    rt->mfc_origin     = mfccp->mfcc_origin;
1172    rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1173
1174    update_mfc_params(rt, mfccp);
1175
1176    /* initialize pkt counters per src-grp */
1177    rt->mfc_pkt_cnt    = 0;
1178    rt->mfc_byte_cnt   = 0;
1179    rt->mfc_wrong_if   = 0;
1180    rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
1181}
1182
1183
1184/*
1185 * Add an mfc entry
1186 */
1187static int
1188add_mfc(struct mfcctl2 *mfccp)
1189{
1190    struct mfc *rt;
1191    u_long hash;
1192    struct rtdetq *rte;
1193    u_short nstl;
1194
1195    VIF_LOCK();
1196    MFC_LOCK();
1197
1198    rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
1199
1200    /* If an entry already exists, just update the fields */
1201    if (rt) {
1202	if (mrtdebug & DEBUG_MFC)
1203	    log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n",
1204		(u_long)ntohl(mfccp->mfcc_origin.s_addr),
1205		(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1206		mfccp->mfcc_parent);
1207
1208	update_mfc_params(rt, mfccp);
1209	MFC_UNLOCK();
1210	VIF_UNLOCK();
1211	return 0;
1212    }
1213
1214    /*
1215     * Find the entry for which the upcall was made and update
1216     */
1217    hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
1218    for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) {
1219
1220	if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1221		(rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
1222		(rt->mfc_stall != NULL)) {
1223
1224	    if (nstl++)
1225		log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n",
1226		    "multiple kernel entries",
1227		    (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1228		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1229		    mfccp->mfcc_parent, (void *)rt->mfc_stall);
1230
1231	    if (mrtdebug & DEBUG_MFC)
1232		log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n",
1233		    (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1234		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1235		    mfccp->mfcc_parent, (void *)rt->mfc_stall);
1236
1237	    init_mfc_params(rt, mfccp);
1238
1239	    rt->mfc_expire = 0;	/* Don't clean this guy up */
1240	    nexpire[hash]--;
1241
1242	    /* free packets Qed at the end of this entry */
1243	    for (rte = rt->mfc_stall; rte != NULL; ) {
1244		struct rtdetq *n = rte->next;
1245
1246		ip_mdq(rte->m, rte->ifp, rt, -1);
1247		m_freem(rte->m);
1248		free(rte, M_MRTABLE);
1249		rte = n;
1250	    }
1251	    rt->mfc_stall = NULL;
1252	}
1253    }
1254
1255    /*
1256     * It is possible that an entry is being inserted without an upcall
1257     */
1258    if (nstl == 0) {
1259	if (mrtdebug & DEBUG_MFC)
1260	    log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n",
1261		hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1262		(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1263		mfccp->mfcc_parent);
1264
1265	for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) {
1266	    if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1267		    (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
1268		init_mfc_params(rt, mfccp);
1269		if (rt->mfc_expire)
1270		    nexpire[hash]--;
1271		rt->mfc_expire = 0;
1272		break; /* XXX */
1273	    }
1274	}
1275	if (rt == NULL) {		/* no upcall, so make a new entry */
1276	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1277	    if (rt == NULL) {
1278		MFC_UNLOCK();
1279		VIF_UNLOCK();
1280		return ENOBUFS;
1281	    }
1282
1283	    init_mfc_params(rt, mfccp);
1284	    rt->mfc_expire     = 0;
1285	    rt->mfc_stall      = NULL;
1286
1287	    rt->mfc_bw_meter = NULL;
1288	    /* insert new entry at head of hash chain */
1289	    rt->mfc_next = mfctable[hash];
1290	    mfctable[hash] = rt;
1291	}
1292    }
1293    MFC_UNLOCK();
1294    VIF_UNLOCK();
1295    return 0;
1296}
1297
1298/*
1299 * Delete an mfc entry
1300 */
1301static int
1302del_mfc(struct mfcctl2 *mfccp)
1303{
1304    struct in_addr	origin;
1305    struct in_addr	mcastgrp;
1306    struct mfc		*rt;
1307    struct mfc		**nptr;
1308    u_long		hash;
1309    struct bw_meter	*list;
1310
1311    origin = mfccp->mfcc_origin;
1312    mcastgrp = mfccp->mfcc_mcastgrp;
1313
1314    if (mrtdebug & DEBUG_MFC)
1315	log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n",
1316	    (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1317
1318    MFC_LOCK();
1319
1320    hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1321    for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next)
1322	if (origin.s_addr == rt->mfc_origin.s_addr &&
1323		mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1324		rt->mfc_stall == NULL)
1325	    break;
1326    if (rt == NULL) {
1327	MFC_UNLOCK();
1328	return EADDRNOTAVAIL;
1329    }
1330
1331    *nptr = rt->mfc_next;
1332
1333    /*
1334     * free the bw_meter entries
1335     */
1336    list = rt->mfc_bw_meter;
1337    rt->mfc_bw_meter = NULL;
1338
1339    free(rt, M_MRTABLE);
1340
1341    free_bw_list(list);
1342
1343    MFC_UNLOCK();
1344
1345    return 0;
1346}
1347
1348/*
1349 * Send a message to mrouted on the multicast routing socket
1350 */
1351static int
1352socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1353{
1354    if (s) {
1355	SOCKBUF_LOCK(&s->so_rcv);
1356	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1357	    NULL) != 0) {
1358	    sorwakeup_locked(s);
1359	    return 0;
1360	}
1361	SOCKBUF_UNLOCK(&s->so_rcv);
1362    }
1363    m_freem(mm);
1364    return -1;
1365}
1366
1367/*
1368 * IP multicast forwarding function. This function assumes that the packet
1369 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1370 * pointed to by "ifp", and the packet is to be relayed to other networks
1371 * that have members of the packet's destination IP multicast group.
1372 *
1373 * The packet is returned unscathed to the caller, unless it is
1374 * erroneous, in which case a non-zero return value tells the caller to
1375 * discard it.
1376 */
1377
1378#define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1379
1380static int
1381X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1382    struct ip_moptions *imo)
1383{
1384    struct mfc *rt;
1385    int error;
1386    vifi_t vifi;
1387
1388    if (mrtdebug & DEBUG_FORWARD)
1389	log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n",
1390	    (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr),
1391	    (void *)ifp);
1392
1393    if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1394		((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1395	/*
1396	 * Packet arrived via a physical interface or
1397	 * an encapsulated tunnel or a register_vif.
1398	 */
1399    } else {
1400	/*
1401	 * Packet arrived through a source-route tunnel.
1402	 * Source-route tunnels are no longer supported.
1403	 */
1404	static int last_log;
1405	if (last_log != time_uptime) {
1406	    last_log = time_uptime;
1407	    log(LOG_ERR,
1408		"ip_mforward: received source-routed packet from %lx\n",
1409		(u_long)ntohl(ip->ip_src.s_addr));
1410	}
1411	return 1;
1412    }
1413
1414    VIF_LOCK();
1415    MFC_LOCK();
1416    if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1417	if (ip->ip_ttl < 255)
1418	    ip->ip_ttl++;	/* compensate for -1 in *_send routines */
1419	if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1420	    struct vif *vifp = viftable + vifi;
1421
1422	    printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n",
1423		(long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr),
1424		vifi,
1425		(vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1426		vifp->v_ifp->if_xname);
1427	}
1428	error = ip_mdq(m, ifp, NULL, vifi);
1429	MFC_UNLOCK();
1430	VIF_UNLOCK();
1431	return error;
1432    }
1433    if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1434	printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1435	    (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr));
1436	if (!imo)
1437	    printf("In fact, no options were specified at all\n");
1438    }
1439
1440    /*
1441     * Don't forward a packet with time-to-live of zero or one,
1442     * or a packet destined to a local-only group.
1443     */
1444    if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) {
1445	MFC_UNLOCK();
1446	VIF_UNLOCK();
1447	return 0;
1448    }
1449
1450    /*
1451     * Determine forwarding vifs from the forwarding cache table
1452     */
1453    ++mrtstat.mrts_mfc_lookups;
1454    rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1455
1456    /* Entry exists, so forward if necessary */
1457    if (rt != NULL) {
1458	error = ip_mdq(m, ifp, rt, -1);
1459	MFC_UNLOCK();
1460	VIF_UNLOCK();
1461	return error;
1462    } else {
1463	/*
1464	 * If we don't have a route for packet's origin,
1465	 * Make a copy of the packet & send message to routing daemon
1466	 */
1467
1468	struct mbuf *mb0;
1469	struct rtdetq *rte;
1470	u_long hash;
1471	int hlen = ip->ip_hl << 2;
1472
1473	++mrtstat.mrts_mfc_misses;
1474
1475	mrtstat.mrts_no_route++;
1476	if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1477	    log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n",
1478		(u_long)ntohl(ip->ip_src.s_addr),
1479		(u_long)ntohl(ip->ip_dst.s_addr));
1480
1481	/*
1482	 * Allocate mbufs early so that we don't do extra work if we are
1483	 * just going to fail anyway.  Make sure to pullup the header so
1484	 * that other people can't step on it.
1485	 */
1486	rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT);
1487	if (rte == NULL) {
1488	    MFC_UNLOCK();
1489	    VIF_UNLOCK();
1490	    return ENOBUFS;
1491	}
1492	mb0 = m_copypacket(m, M_DONTWAIT);
1493	if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1494	    mb0 = m_pullup(mb0, hlen);
1495	if (mb0 == NULL) {
1496	    free(rte, M_MRTABLE);
1497	    MFC_UNLOCK();
1498	    VIF_UNLOCK();
1499	    return ENOBUFS;
1500	}
1501
1502	/* is there an upcall waiting for this flow ? */
1503	hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1504	for (rt = mfctable[hash]; rt; rt = rt->mfc_next) {
1505	    if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1506		    (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1507		    (rt->mfc_stall != NULL))
1508		break;
1509	}
1510
1511	if (rt == NULL) {
1512	    int i;
1513	    struct igmpmsg *im;
1514	    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1515	    struct mbuf *mm;
1516
1517	    /*
1518	     * Locate the vifi for the incoming interface for this packet.
1519	     * If none found, drop packet.
1520	     */
1521	    for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1522		;
1523	    if (vifi >= numvifs)	/* vif not found, drop packet */
1524		goto non_fatal;
1525
1526	    /* no upcall, so make a new entry */
1527	    rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1528	    if (rt == NULL)
1529		goto fail;
1530	    /* Make a copy of the header to send to the user level process */
1531	    mm = m_copy(mb0, 0, hlen);
1532	    if (mm == NULL)
1533		goto fail1;
1534
1535	    /*
1536	     * Send message to routing daemon to install
1537	     * a route into the kernel table
1538	     */
1539
1540	    im = mtod(mm, struct igmpmsg *);
1541	    im->im_msgtype = IGMPMSG_NOCACHE;
1542	    im->im_mbz = 0;
1543	    im->im_vif = vifi;
1544
1545	    mrtstat.mrts_upcalls++;
1546
1547	    k_igmpsrc.sin_addr = ip->ip_src;
1548	    if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1549		log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1550		++mrtstat.mrts_upq_sockfull;
1551fail1:
1552		free(rt, M_MRTABLE);
1553fail:
1554		free(rte, M_MRTABLE);
1555		m_freem(mb0);
1556		MFC_UNLOCK();
1557		VIF_UNLOCK();
1558		return ENOBUFS;
1559	    }
1560
1561	    /* insert new entry at head of hash chain */
1562	    rt->mfc_origin.s_addr     = ip->ip_src.s_addr;
1563	    rt->mfc_mcastgrp.s_addr   = ip->ip_dst.s_addr;
1564	    rt->mfc_expire	      = UPCALL_EXPIRE;
1565	    nexpire[hash]++;
1566	    for (i = 0; i < numvifs; i++) {
1567		rt->mfc_ttls[i] = 0;
1568		rt->mfc_flags[i] = 0;
1569	    }
1570	    rt->mfc_parent = -1;
1571
1572	    rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */
1573
1574	    rt->mfc_bw_meter = NULL;
1575
1576	    /* link into table */
1577	    rt->mfc_next   = mfctable[hash];
1578	    mfctable[hash] = rt;
1579	    rt->mfc_stall = rte;
1580
1581	} else {
1582	    /* determine if q has overflowed */
1583	    int npkts = 0;
1584	    struct rtdetq **p;
1585
1586	    /*
1587	     * XXX ouch! we need to append to the list, but we
1588	     * only have a pointer to the front, so we have to
1589	     * scan the entire list every time.
1590	     */
1591	    for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1592		npkts++;
1593
1594	    if (npkts > MAX_UPQ) {
1595		mrtstat.mrts_upq_ovflw++;
1596non_fatal:
1597		free(rte, M_MRTABLE);
1598		m_freem(mb0);
1599		MFC_UNLOCK();
1600		VIF_UNLOCK();
1601		return 0;
1602	    }
1603
1604	    /* Add this entry to the end of the queue */
1605	    *p = rte;
1606	}
1607
1608	rte->m			= mb0;
1609	rte->ifp		= ifp;
1610	rte->next		= NULL;
1611
1612	MFC_UNLOCK();
1613	VIF_UNLOCK();
1614
1615	return 0;
1616    }
1617}
1618
1619/*
1620 * Clean up the cache entry if upcall is not serviced
1621 */
1622static void
1623expire_upcalls(void *unused)
1624{
1625    struct rtdetq *rte;
1626    struct mfc *mfc, **nptr;
1627    int i;
1628
1629    MFC_LOCK();
1630    for (i = 0; i < MFCTBLSIZ; i++) {
1631	if (nexpire[i] == 0)
1632	    continue;
1633	nptr = &mfctable[i];
1634	for (mfc = *nptr; mfc != NULL; mfc = *nptr) {
1635	    /*
1636	     * Skip real cache entries
1637	     * Make sure it wasn't marked to not expire (shouldn't happen)
1638	     * If it expires now
1639	     */
1640	    if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 &&
1641		    --mfc->mfc_expire == 0) {
1642		if (mrtdebug & DEBUG_EXPIRE)
1643		    log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n",
1644			(u_long)ntohl(mfc->mfc_origin.s_addr),
1645			(u_long)ntohl(mfc->mfc_mcastgrp.s_addr));
1646		/*
1647		 * drop all the packets
1648		 * free the mbuf with the pkt, if, timing info
1649		 */
1650		for (rte = mfc->mfc_stall; rte; ) {
1651		    struct rtdetq *n = rte->next;
1652
1653		    m_freem(rte->m);
1654		    free(rte, M_MRTABLE);
1655		    rte = n;
1656		}
1657		++mrtstat.mrts_cache_cleanups;
1658		nexpire[i]--;
1659
1660		/*
1661		 * free the bw_meter entries
1662		 */
1663		while (mfc->mfc_bw_meter != NULL) {
1664		    struct bw_meter *x = mfc->mfc_bw_meter;
1665
1666		    mfc->mfc_bw_meter = x->bm_mfc_next;
1667		    free(x, M_BWMETER);
1668		}
1669
1670		*nptr = mfc->mfc_next;
1671		free(mfc, M_MRTABLE);
1672	    } else {
1673		nptr = &mfc->mfc_next;
1674	    }
1675	}
1676    }
1677    MFC_UNLOCK();
1678
1679    callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
1680}
1681
1682/*
1683 * Packet forwarding routine once entry in the cache is made
1684 */
1685static int
1686ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1687{
1688    struct ip  *ip = mtod(m, struct ip *);
1689    vifi_t vifi;
1690    int plen = ip->ip_len;
1691
1692    VIF_LOCK_ASSERT();
1693/*
1694 * Macro to send packet on vif.  Since RSVP packets don't get counted on
1695 * input, they shouldn't get counted on output, so statistics keeping is
1696 * separate.
1697 */
1698#define MC_SEND(ip,vifp,m) {				\
1699		if ((vifp)->v_flags & VIFF_TUNNEL)	\
1700		    encap_send((ip), (vifp), (m));	\
1701		else					\
1702		    phyint_send((ip), (vifp), (m));	\
1703}
1704
1705    /*
1706     * If xmt_vif is not -1, send on only the requested vif.
1707     *
1708     * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1709     */
1710    if (xmt_vif < numvifs) {
1711#ifdef PIM
1712	if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1713	    pim_register_send(ip, viftable + xmt_vif, m, rt);
1714	else
1715#endif
1716	MC_SEND(ip, viftable + xmt_vif, m);
1717	return 1;
1718    }
1719
1720    /*
1721     * Don't forward if it didn't arrive from the parent vif for its origin.
1722     */
1723    vifi = rt->mfc_parent;
1724    if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1725	/* came in the wrong interface */
1726	if (mrtdebug & DEBUG_FORWARD)
1727	    log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1728		(void *)ifp, vifi, (void *)viftable[vifi].v_ifp);
1729	++mrtstat.mrts_wrong_if;
1730	++rt->mfc_wrong_if;
1731	/*
1732	 * If we are doing PIM assert processing, send a message
1733	 * to the routing daemon.
1734	 *
1735	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1736	 * can complete the SPT switch, regardless of the type
1737	 * of the iif (broadcast media, GRE tunnel, etc).
1738	 */
1739	if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1740	    struct timeval now;
1741	    u_long delta;
1742
1743#ifdef PIM
1744	    if (ifp == &multicast_register_if)
1745		pimstat.pims_rcv_registers_wrongiif++;
1746#endif
1747
1748	    /* Get vifi for the incoming packet */
1749	    for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1750		;
1751	    if (vifi >= numvifs)
1752		return 0;	/* The iif is not found: ignore the packet. */
1753
1754	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1755		return 0;	/* WRONGVIF disabled: ignore the packet */
1756
1757	    GET_TIME(now);
1758
1759	    TV_DELTA(rt->mfc_last_assert, now, delta);
1760
1761	    if (delta > ASSERT_MSG_TIME) {
1762		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1763		struct igmpmsg *im;
1764		int hlen = ip->ip_hl << 2;
1765		struct mbuf *mm = m_copy(m, 0, hlen);
1766
1767		if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1768		    mm = m_pullup(mm, hlen);
1769		if (mm == NULL)
1770		    return ENOBUFS;
1771
1772		rt->mfc_last_assert = now;
1773
1774		im = mtod(mm, struct igmpmsg *);
1775		im->im_msgtype	= IGMPMSG_WRONGVIF;
1776		im->im_mbz		= 0;
1777		im->im_vif		= vifi;
1778
1779		mrtstat.mrts_upcalls++;
1780
1781		k_igmpsrc.sin_addr = im->im_src;
1782		if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1783		    log(LOG_WARNING,
1784			"ip_mforward: ip_mrouter socket queue full\n");
1785		    ++mrtstat.mrts_upq_sockfull;
1786		    return ENOBUFS;
1787		}
1788	    }
1789	}
1790	return 0;
1791    }
1792
1793    /* If I sourced this packet, it counts as output, else it was input. */
1794    if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1795	viftable[vifi].v_pkt_out++;
1796	viftable[vifi].v_bytes_out += plen;
1797    } else {
1798	viftable[vifi].v_pkt_in++;
1799	viftable[vifi].v_bytes_in += plen;
1800    }
1801    rt->mfc_pkt_cnt++;
1802    rt->mfc_byte_cnt += plen;
1803
1804    /*
1805     * For each vif, decide if a copy of the packet should be forwarded.
1806     * Forward if:
1807     *		- the ttl exceeds the vif's threshold
1808     *		- there are group members downstream on interface
1809     */
1810    for (vifi = 0; vifi < numvifs; vifi++)
1811	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1812	    viftable[vifi].v_pkt_out++;
1813	    viftable[vifi].v_bytes_out += plen;
1814#ifdef PIM
1815	    if (viftable[vifi].v_flags & VIFF_REGISTER)
1816		pim_register_send(ip, viftable + vifi, m, rt);
1817	    else
1818#endif
1819	    MC_SEND(ip, viftable+vifi, m);
1820	}
1821
1822    /*
1823     * Perform upcall-related bw measuring.
1824     */
1825    if (rt->mfc_bw_meter != NULL) {
1826	struct bw_meter *x;
1827	struct timeval now;
1828
1829	GET_TIME(now);
1830	MFC_LOCK_ASSERT();
1831	for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1832	    bw_meter_receive_packet(x, plen, &now);
1833    }
1834
1835    return 0;
1836}
1837
1838/*
1839 * check if a vif number is legal/ok. This is used by ip_output.
1840 */
1841static int
1842X_legal_vif_num(int vif)
1843{
1844    /* XXX unlocked, matter? */
1845    return (vif >= 0 && vif < numvifs);
1846}
1847
1848/*
1849 * Return the local address used by this vif
1850 */
1851static u_long
1852X_ip_mcast_src(int vifi)
1853{
1854    /* XXX unlocked, matter? */
1855    if (vifi >= 0 && vifi < numvifs)
1856	return viftable[vifi].v_lcl_addr.s_addr;
1857    else
1858	return INADDR_ANY;
1859}
1860
1861static void
1862phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1863{
1864    struct mbuf *mb_copy;
1865    int hlen = ip->ip_hl << 2;
1866
1867    VIF_LOCK_ASSERT();
1868
1869    /*
1870     * Make a new reference to the packet; make sure that
1871     * the IP header is actually copied, not just referenced,
1872     * so that ip_output() only scribbles on the copy.
1873     */
1874    mb_copy = m_copypacket(m, M_DONTWAIT);
1875    if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1876	mb_copy = m_pullup(mb_copy, hlen);
1877    if (mb_copy == NULL)
1878	return;
1879
1880    if (vifp->v_rate_limit == 0)
1881	tbf_send_packet(vifp, mb_copy);
1882    else
1883	tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1884}
1885
1886static void
1887encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1888{
1889    struct mbuf *mb_copy;
1890    struct ip *ip_copy;
1891    int i, len = ip->ip_len;
1892
1893    VIF_LOCK_ASSERT();
1894
1895    /* Take care of delayed checksums */
1896    if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1897	in_delayed_cksum(m);
1898	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1899    }
1900
1901    /*
1902     * copy the old packet & pullup its IP header into the
1903     * new mbuf so we can modify it.  Try to fill the new
1904     * mbuf since if we don't the ethernet driver will.
1905     */
1906    MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1907    if (mb_copy == NULL)
1908	return;
1909#ifdef MAC
1910    mac_create_mbuf_multicast_encap(m, vifp->v_ifp, mb_copy);
1911#endif
1912    mb_copy->m_data += max_linkhdr;
1913    mb_copy->m_len = sizeof(multicast_encap_iphdr);
1914
1915    if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
1916	m_freem(mb_copy);
1917	return;
1918    }
1919    i = MHLEN - M_LEADINGSPACE(mb_copy);
1920    if (i > len)
1921	i = len;
1922    mb_copy = m_pullup(mb_copy, i);
1923    if (mb_copy == NULL)
1924	return;
1925    mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1926
1927    /*
1928     * fill in the encapsulating IP header.
1929     */
1930    ip_copy = mtod(mb_copy, struct ip *);
1931    *ip_copy = multicast_encap_iphdr;
1932    ip_copy->ip_id = ip_newid();
1933    ip_copy->ip_len += len;
1934    ip_copy->ip_src = vifp->v_lcl_addr;
1935    ip_copy->ip_dst = vifp->v_rmt_addr;
1936
1937    /*
1938     * turn the encapsulated IP header back into a valid one.
1939     */
1940    ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1941    --ip->ip_ttl;
1942    ip->ip_len = htons(ip->ip_len);
1943    ip->ip_off = htons(ip->ip_off);
1944    ip->ip_sum = 0;
1945    mb_copy->m_data += sizeof(multicast_encap_iphdr);
1946    ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1947    mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1948
1949    if (vifp->v_rate_limit == 0)
1950	tbf_send_packet(vifp, mb_copy);
1951    else
1952	tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1953}
1954
1955/*
1956 * Token bucket filter module
1957 */
1958
1959static void
1960tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_long p_len)
1961{
1962    struct tbf *t = vifp->v_tbf;
1963
1964    VIF_LOCK_ASSERT();
1965
1966    if (p_len > MAX_BKT_SIZE) {		/* drop if packet is too large */
1967	mrtstat.mrts_pkt2large++;
1968	m_freem(m);
1969	return;
1970    }
1971
1972    tbf_update_tokens(vifp);
1973
1974    if (t->tbf_q_len == 0) {		/* queue empty...		*/
1975	if (p_len <= t->tbf_n_tok) {	/* send packet if enough tokens */
1976	    t->tbf_n_tok -= p_len;
1977	    tbf_send_packet(vifp, m);
1978	} else {			/* no, queue packet and try later */
1979	    tbf_queue(vifp, m);
1980	    callout_reset(&tbf_reprocess_ch, TBF_REPROCESS,
1981		tbf_reprocess_q, vifp);
1982	}
1983    } else if (t->tbf_q_len < t->tbf_max_q_len) {
1984	/* finite queue length, so queue pkts and process queue */
1985	tbf_queue(vifp, m);
1986	tbf_process_q(vifp);
1987    } else {
1988	/* queue full, try to dq and queue and process */
1989	if (!tbf_dq_sel(vifp, ip)) {
1990	    mrtstat.mrts_q_overflow++;
1991	    m_freem(m);
1992	} else {
1993	    tbf_queue(vifp, m);
1994	    tbf_process_q(vifp);
1995	}
1996    }
1997}
1998
1999/*
2000 * adds a packet to the queue at the interface
2001 */
2002static void
2003tbf_queue(struct vif *vifp, struct mbuf *m)
2004{
2005    struct tbf *t = vifp->v_tbf;
2006
2007    VIF_LOCK_ASSERT();
2008
2009    if (t->tbf_t == NULL)	/* Queue was empty */
2010	t->tbf_q = m;
2011    else			/* Insert at tail */
2012	t->tbf_t->m_act = m;
2013
2014    t->tbf_t = m;		/* Set new tail pointer */
2015
2016#ifdef DIAGNOSTIC
2017    /* Make sure we didn't get fed a bogus mbuf */
2018    if (m->m_act)
2019	panic("tbf_queue: m_act");
2020#endif
2021    m->m_act = NULL;
2022
2023    t->tbf_q_len++;
2024}
2025
2026/*
2027 * processes the queue at the interface
2028 */
2029static void
2030tbf_process_q(struct vif *vifp)
2031{
2032    struct tbf *t = vifp->v_tbf;
2033
2034    VIF_LOCK_ASSERT();
2035
2036    /* loop through the queue at the interface and send as many packets
2037     * as possible
2038     */
2039    while (t->tbf_q_len > 0) {
2040	struct mbuf *m = t->tbf_q;
2041	int len = mtod(m, struct ip *)->ip_len;
2042
2043	/* determine if the packet can be sent */
2044	if (len > t->tbf_n_tok)	/* not enough tokens, we are done */
2045	    break;
2046	/* ok, reduce no of tokens, dequeue and send the packet. */
2047	t->tbf_n_tok -= len;
2048
2049	t->tbf_q = m->m_act;
2050	if (--t->tbf_q_len == 0)
2051	    t->tbf_t = NULL;
2052
2053	m->m_act = NULL;
2054	tbf_send_packet(vifp, m);
2055    }
2056}
2057
2058static void
2059tbf_reprocess_q(void *xvifp)
2060{
2061    struct vif *vifp = xvifp;
2062
2063    if (ip_mrouter == NULL)
2064	return;
2065    VIF_LOCK();
2066    tbf_update_tokens(vifp);
2067    tbf_process_q(vifp);
2068    if (vifp->v_tbf->tbf_q_len)
2069	callout_reset(&tbf_reprocess_ch, TBF_REPROCESS, tbf_reprocess_q, vifp);
2070    VIF_UNLOCK();
2071}
2072
2073/* function that will selectively discard a member of the queue
2074 * based on the precedence value and the priority
2075 */
2076static int
2077tbf_dq_sel(struct vif *vifp, struct ip *ip)
2078{
2079    u_int p;
2080    struct mbuf *m, *last;
2081    struct mbuf **np;
2082    struct tbf *t = vifp->v_tbf;
2083
2084    VIF_LOCK_ASSERT();
2085
2086    p = priority(vifp, ip);
2087
2088    np = &t->tbf_q;
2089    last = NULL;
2090    while ((m = *np) != NULL) {
2091	if (p > priority(vifp, mtod(m, struct ip *))) {
2092	    *np = m->m_act;
2093	    /* If we're removing the last packet, fix the tail pointer */
2094	    if (m == t->tbf_t)
2095		t->tbf_t = last;
2096	    m_freem(m);
2097	    /* It's impossible for the queue to be empty, but check anyways. */
2098	    if (--t->tbf_q_len == 0)
2099		t->tbf_t = NULL;
2100	    mrtstat.mrts_drop_sel++;
2101	    return 1;
2102	}
2103	np = &m->m_act;
2104	last = m;
2105    }
2106    return 0;
2107}
2108
2109static void
2110tbf_send_packet(struct vif *vifp, struct mbuf *m)
2111{
2112    VIF_LOCK_ASSERT();
2113
2114    if (vifp->v_flags & VIFF_TUNNEL)	/* If tunnel options */
2115	ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL);
2116    else {
2117	struct ip_moptions imo;
2118	int error;
2119	static struct route ro; /* XXX check this */
2120
2121	imo.imo_multicast_ifp  = vifp->v_ifp;
2122	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
2123	imo.imo_multicast_loop = 1;
2124	imo.imo_multicast_vif  = -1;
2125
2126	/*
2127	 * Re-entrancy should not be a problem here, because
2128	 * the packets that we send out and are looped back at us
2129	 * should get rejected because they appear to come from
2130	 * the loopback interface, thus preventing looping.
2131	 */
2132	error = ip_output(m, NULL, &ro, IP_FORWARDING, &imo, NULL);
2133
2134	if (mrtdebug & DEBUG_XMIT)
2135	    log(LOG_DEBUG, "phyint_send on vif %d err %d\n",
2136		(int)(vifp - viftable), error);
2137    }
2138}
2139
2140/* determine the current time and then
2141 * the elapsed time (between the last time and time now)
2142 * in milliseconds & update the no. of tokens in the bucket
2143 */
2144static void
2145tbf_update_tokens(struct vif *vifp)
2146{
2147    struct timeval tp;
2148    u_long tm;
2149    struct tbf *t = vifp->v_tbf;
2150
2151    VIF_LOCK_ASSERT();
2152
2153    GET_TIME(tp);
2154
2155    TV_DELTA(tp, t->tbf_last_pkt_t, tm);
2156
2157    /*
2158     * This formula is actually
2159     * "time in seconds" * "bytes/second".
2160     *
2161     * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2162     *
2163     * The (1000/1024) was introduced in add_vif to optimize
2164     * this divide into a shift.
2165     */
2166    t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8;
2167    t->tbf_last_pkt_t = tp;
2168
2169    if (t->tbf_n_tok > MAX_BKT_SIZE)
2170	t->tbf_n_tok = MAX_BKT_SIZE;
2171}
2172
2173static int
2174priority(struct vif *vifp, struct ip *ip)
2175{
2176    int prio = 50; /* the lowest priority -- default case */
2177
2178    /* temporary hack; may add general packet classifier some day */
2179
2180    /*
2181     * The UDP port space is divided up into four priority ranges:
2182     * [0, 16384)     : unclassified - lowest priority
2183     * [16384, 32768) : audio - highest priority
2184     * [32768, 49152) : whiteboard - medium priority
2185     * [49152, 65536) : video - low priority
2186     *
2187     * Everything else gets lowest priority.
2188     */
2189    if (ip->ip_p == IPPROTO_UDP) {
2190	struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2191	switch (ntohs(udp->uh_dport) & 0xc000) {
2192	case 0x4000:
2193	    prio = 70;
2194	    break;
2195	case 0x8000:
2196	    prio = 60;
2197	    break;
2198	case 0xc000:
2199	    prio = 55;
2200	    break;
2201	}
2202    }
2203    return prio;
2204}
2205
2206/*
2207 * End of token bucket filter modifications
2208 */
2209
2210static int
2211X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt)
2212{
2213    int error, vifi;
2214
2215    if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2216	return EOPNOTSUPP;
2217
2218    error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
2219    if (error)
2220	return error;
2221
2222    VIF_LOCK();
2223
2224    if (vifi < 0 || vifi >= numvifs) {	/* Error if vif is invalid */
2225	VIF_UNLOCK();
2226	return EADDRNOTAVAIL;
2227    }
2228
2229    if (sopt->sopt_name == IP_RSVP_VIF_ON) {
2230	/* Check if socket is available. */
2231	if (viftable[vifi].v_rsvpd != NULL) {
2232	    VIF_UNLOCK();
2233	    return EADDRINUSE;
2234	}
2235
2236	viftable[vifi].v_rsvpd = so;
2237	/* This may seem silly, but we need to be sure we don't over-increment
2238	 * the RSVP counter, in case something slips up.
2239	 */
2240	if (!viftable[vifi].v_rsvp_on) {
2241	    viftable[vifi].v_rsvp_on = 1;
2242	    rsvp_on++;
2243	}
2244    } else { /* must be VIF_OFF */
2245	/*
2246	 * XXX as an additional consistency check, one could make sure
2247	 * that viftable[vifi].v_rsvpd == so, otherwise passing so as
2248	 * first parameter is pretty useless.
2249	 */
2250	viftable[vifi].v_rsvpd = NULL;
2251	/*
2252	 * This may seem silly, but we need to be sure we don't over-decrement
2253	 * the RSVP counter, in case something slips up.
2254	 */
2255	if (viftable[vifi].v_rsvp_on) {
2256	    viftable[vifi].v_rsvp_on = 0;
2257	    rsvp_on--;
2258	}
2259    }
2260    VIF_UNLOCK();
2261    return 0;
2262}
2263
2264static void
2265X_ip_rsvp_force_done(struct socket *so)
2266{
2267    int vifi;
2268
2269    /* Don't bother if it is not the right type of socket. */
2270    if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2271	return;
2272
2273    VIF_LOCK();
2274
2275    /* The socket may be attached to more than one vif...this
2276     * is perfectly legal.
2277     */
2278    for (vifi = 0; vifi < numvifs; vifi++) {
2279	if (viftable[vifi].v_rsvpd == so) {
2280	    viftable[vifi].v_rsvpd = NULL;
2281	    /* This may seem silly, but we need to be sure we don't
2282	     * over-decrement the RSVP counter, in case something slips up.
2283	     */
2284	    if (viftable[vifi].v_rsvp_on) {
2285		viftable[vifi].v_rsvp_on = 0;
2286		rsvp_on--;
2287	    }
2288	}
2289    }
2290
2291    VIF_UNLOCK();
2292}
2293
2294static void
2295X_rsvp_input(struct mbuf *m, int off)
2296{
2297    int vifi;
2298    struct ip *ip = mtod(m, struct ip *);
2299    struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
2300    struct ifnet *ifp;
2301
2302    if (rsvpdebug)
2303	printf("rsvp_input: rsvp_on %d\n",rsvp_on);
2304
2305    /* Can still get packets with rsvp_on = 0 if there is a local member
2306     * of the group to which the RSVP packet is addressed.  But in this
2307     * case we want to throw the packet away.
2308     */
2309    if (!rsvp_on) {
2310	m_freem(m);
2311	return;
2312    }
2313
2314    if (rsvpdebug)
2315	printf("rsvp_input: check vifs\n");
2316
2317#ifdef DIAGNOSTIC
2318    M_ASSERTPKTHDR(m);
2319#endif
2320
2321    ifp = m->m_pkthdr.rcvif;
2322
2323    VIF_LOCK();
2324    /* Find which vif the packet arrived on. */
2325    for (vifi = 0; vifi < numvifs; vifi++)
2326	if (viftable[vifi].v_ifp == ifp)
2327	    break;
2328
2329    if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) {
2330	/*
2331	 * Drop the lock here to avoid holding it across rip_input.
2332	 * This could make rsvpdebug printfs wrong.  If you care,
2333	 * record the state of stuff before dropping the lock.
2334	 */
2335	VIF_UNLOCK();
2336	/*
2337	 * If the old-style non-vif-associated socket is set,
2338	 * then use it.  Otherwise, drop packet since there
2339	 * is no specific socket for this vif.
2340	 */
2341	if (ip_rsvpd != NULL) {
2342	    if (rsvpdebug)
2343		printf("rsvp_input: Sending packet up old-style socket\n");
2344	    rip_input(m, off);  /* xxx */
2345	} else {
2346	    if (rsvpdebug && vifi == numvifs)
2347		printf("rsvp_input: Can't find vif for packet.\n");
2348	    else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL)
2349		printf("rsvp_input: No socket defined for vif %d\n",vifi);
2350	    m_freem(m);
2351	}
2352	return;
2353    }
2354    rsvp_src.sin_addr = ip->ip_src;
2355
2356    if (rsvpdebug && m)
2357	printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n",
2358	       m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
2359
2360    if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) {
2361	if (rsvpdebug)
2362	    printf("rsvp_input: Failed to append to socket\n");
2363    } else {
2364	if (rsvpdebug)
2365	    printf("rsvp_input: send packet up\n");
2366    }
2367    VIF_UNLOCK();
2368}
2369
2370/*
2371 * Code for bandwidth monitors
2372 */
2373
2374/*
2375 * Define common interface for timeval-related methods
2376 */
2377#define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
2378#define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
2379#define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
2380
2381static uint32_t
2382compute_bw_meter_flags(struct bw_upcall *req)
2383{
2384    uint32_t flags = 0;
2385
2386    if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2387	flags |= BW_METER_UNIT_PACKETS;
2388    if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2389	flags |= BW_METER_UNIT_BYTES;
2390    if (req->bu_flags & BW_UPCALL_GEQ)
2391	flags |= BW_METER_GEQ;
2392    if (req->bu_flags & BW_UPCALL_LEQ)
2393	flags |= BW_METER_LEQ;
2394
2395    return flags;
2396}
2397
2398/*
2399 * Add a bw_meter entry
2400 */
2401static int
2402add_bw_upcall(struct bw_upcall *req)
2403{
2404    struct mfc *mfc;
2405    struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2406		BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2407    struct timeval now;
2408    struct bw_meter *x;
2409    uint32_t flags;
2410
2411    if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2412	return EOPNOTSUPP;
2413
2414    /* Test if the flags are valid */
2415    if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2416	return EINVAL;
2417    if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2418	return EINVAL;
2419    if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2420	    == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2421	return EINVAL;
2422
2423    /* Test if the threshold time interval is valid */
2424    if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2425	return EINVAL;
2426
2427    flags = compute_bw_meter_flags(req);
2428
2429    /*
2430     * Find if we have already same bw_meter entry
2431     */
2432    MFC_LOCK();
2433    mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
2434    if (mfc == NULL) {
2435	MFC_UNLOCK();
2436	return EADDRNOTAVAIL;
2437    }
2438    for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2439	if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2440			   &req->bu_threshold.b_time, ==)) &&
2441	    (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2442	    (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2443	    (x->bm_flags & BW_METER_USER_FLAGS) == flags)  {
2444	    MFC_UNLOCK();
2445	    return 0;		/* XXX Already installed */
2446	}
2447    }
2448
2449    /* Allocate the new bw_meter entry */
2450    x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2451    if (x == NULL) {
2452	MFC_UNLOCK();
2453	return ENOBUFS;
2454    }
2455
2456    /* Set the new bw_meter entry */
2457    x->bm_threshold.b_time = req->bu_threshold.b_time;
2458    GET_TIME(now);
2459    x->bm_start_time = now;
2460    x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2461    x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2462    x->bm_measured.b_packets = 0;
2463    x->bm_measured.b_bytes = 0;
2464    x->bm_flags = flags;
2465    x->bm_time_next = NULL;
2466    x->bm_time_hash = BW_METER_BUCKETS;
2467
2468    /* Add the new bw_meter entry to the front of entries for this MFC */
2469    x->bm_mfc = mfc;
2470    x->bm_mfc_next = mfc->mfc_bw_meter;
2471    mfc->mfc_bw_meter = x;
2472    schedule_bw_meter(x, &now);
2473    MFC_UNLOCK();
2474
2475    return 0;
2476}
2477
2478static void
2479free_bw_list(struct bw_meter *list)
2480{
2481    while (list != NULL) {
2482	struct bw_meter *x = list;
2483
2484	list = list->bm_mfc_next;
2485	unschedule_bw_meter(x);
2486	free(x, M_BWMETER);
2487    }
2488}
2489
2490/*
2491 * Delete one or multiple bw_meter entries
2492 */
2493static int
2494del_bw_upcall(struct bw_upcall *req)
2495{
2496    struct mfc *mfc;
2497    struct bw_meter *x;
2498
2499    if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2500	return EOPNOTSUPP;
2501
2502    MFC_LOCK();
2503    /* Find the corresponding MFC entry */
2504    mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
2505    if (mfc == NULL) {
2506	MFC_UNLOCK();
2507	return EADDRNOTAVAIL;
2508    } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2509	/*
2510	 * Delete all bw_meter entries for this mfc
2511	 */
2512	struct bw_meter *list;
2513
2514	list = mfc->mfc_bw_meter;
2515	mfc->mfc_bw_meter = NULL;
2516	free_bw_list(list);
2517	MFC_UNLOCK();
2518	return 0;
2519    } else {			/* Delete a single bw_meter entry */
2520	struct bw_meter *prev;
2521	uint32_t flags = 0;
2522
2523	flags = compute_bw_meter_flags(req);
2524
2525	/* Find the bw_meter entry to delete */
2526	for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2527	     prev = x, x = x->bm_mfc_next) {
2528	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2529			       &req->bu_threshold.b_time, ==)) &&
2530		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2531		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2532		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2533		break;
2534	}
2535	if (x != NULL) { /* Delete entry from the list for this MFC */
2536	    if (prev != NULL)
2537		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2538	    else
2539		x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2540
2541	    unschedule_bw_meter(x);
2542	    MFC_UNLOCK();
2543	    /* Free the bw_meter entry */
2544	    free(x, M_BWMETER);
2545	    return 0;
2546	} else {
2547	    MFC_UNLOCK();
2548	    return EINVAL;
2549	}
2550    }
2551    /* NOTREACHED */
2552}
2553
2554/*
2555 * Perform bandwidth measurement processing that may result in an upcall
2556 */
2557static void
2558bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2559{
2560    struct timeval delta;
2561
2562    MFC_LOCK_ASSERT();
2563
2564    delta = *nowp;
2565    BW_TIMEVALDECR(&delta, &x->bm_start_time);
2566
2567    if (x->bm_flags & BW_METER_GEQ) {
2568	/*
2569	 * Processing for ">=" type of bw_meter entry
2570	 */
2571	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2572	    /* Reset the bw_meter entry */
2573	    x->bm_start_time = *nowp;
2574	    x->bm_measured.b_packets = 0;
2575	    x->bm_measured.b_bytes = 0;
2576	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2577	}
2578
2579	/* Record that a packet is received */
2580	x->bm_measured.b_packets++;
2581	x->bm_measured.b_bytes += plen;
2582
2583	/*
2584	 * Test if we should deliver an upcall
2585	 */
2586	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2587	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2588		 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2589		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2590		 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2591		/* Prepare an upcall for delivery */
2592		bw_meter_prepare_upcall(x, nowp);
2593		x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2594	    }
2595	}
2596    } else if (x->bm_flags & BW_METER_LEQ) {
2597	/*
2598	 * Processing for "<=" type of bw_meter entry
2599	 */
2600	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2601	    /*
2602	     * We are behind time with the multicast forwarding table
2603	     * scanning for "<=" type of bw_meter entries, so test now
2604	     * if we should deliver an upcall.
2605	     */
2606	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2607		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2608		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2609		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2610		/* Prepare an upcall for delivery */
2611		bw_meter_prepare_upcall(x, nowp);
2612	    }
2613	    /* Reschedule the bw_meter entry */
2614	    unschedule_bw_meter(x);
2615	    schedule_bw_meter(x, nowp);
2616	}
2617
2618	/* Record that a packet is received */
2619	x->bm_measured.b_packets++;
2620	x->bm_measured.b_bytes += plen;
2621
2622	/*
2623	 * Test if we should restart the measuring interval
2624	 */
2625	if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2626	     x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2627	    (x->bm_flags & BW_METER_UNIT_BYTES &&
2628	     x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2629	    /* Don't restart the measuring interval */
2630	} else {
2631	    /* Do restart the measuring interval */
2632	    /*
2633	     * XXX: note that we don't unschedule and schedule, because this
2634	     * might be too much overhead per packet. Instead, when we process
2635	     * all entries for a given timer hash bin, we check whether it is
2636	     * really a timeout. If not, we reschedule at that time.
2637	     */
2638	    x->bm_start_time = *nowp;
2639	    x->bm_measured.b_packets = 0;
2640	    x->bm_measured.b_bytes = 0;
2641	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2642	}
2643    }
2644}
2645
2646/*
2647 * Prepare a bandwidth-related upcall
2648 */
2649static void
2650bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2651{
2652    struct timeval delta;
2653    struct bw_upcall *u;
2654
2655    MFC_LOCK_ASSERT();
2656
2657    /*
2658     * Compute the measured time interval
2659     */
2660    delta = *nowp;
2661    BW_TIMEVALDECR(&delta, &x->bm_start_time);
2662
2663    /*
2664     * If there are too many pending upcalls, deliver them now
2665     */
2666    if (bw_upcalls_n >= BW_UPCALLS_MAX)
2667	bw_upcalls_send();
2668
2669    /*
2670     * Set the bw_upcall entry
2671     */
2672    u = &bw_upcalls[bw_upcalls_n++];
2673    u->bu_src = x->bm_mfc->mfc_origin;
2674    u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2675    u->bu_threshold.b_time = x->bm_threshold.b_time;
2676    u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2677    u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2678    u->bu_measured.b_time = delta;
2679    u->bu_measured.b_packets = x->bm_measured.b_packets;
2680    u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2681    u->bu_flags = 0;
2682    if (x->bm_flags & BW_METER_UNIT_PACKETS)
2683	u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2684    if (x->bm_flags & BW_METER_UNIT_BYTES)
2685	u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2686    if (x->bm_flags & BW_METER_GEQ)
2687	u->bu_flags |= BW_UPCALL_GEQ;
2688    if (x->bm_flags & BW_METER_LEQ)
2689	u->bu_flags |= BW_UPCALL_LEQ;
2690}
2691
2692/*
2693 * Send the pending bandwidth-related upcalls
2694 */
2695static void
2696bw_upcalls_send(void)
2697{
2698    struct mbuf *m;
2699    int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2700    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2701    static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2702				      0,		/* unused2 */
2703				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2704				      0,		/* im_mbz  */
2705				      0,		/* im_vif  */
2706				      0,		/* unused3 */
2707				      { 0 },		/* im_src  */
2708				      { 0 } };		/* im_dst  */
2709
2710    MFC_LOCK_ASSERT();
2711
2712    if (bw_upcalls_n == 0)
2713	return;			/* No pending upcalls */
2714
2715    bw_upcalls_n = 0;
2716
2717    /*
2718     * Allocate a new mbuf, initialize it with the header and
2719     * the payload for the pending calls.
2720     */
2721    MGETHDR(m, M_DONTWAIT, MT_DATA);
2722    if (m == NULL) {
2723	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2724	return;
2725    }
2726
2727    m->m_len = m->m_pkthdr.len = 0;
2728    m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2729    m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2730
2731    /*
2732     * Send the upcalls
2733     * XXX do we need to set the address in k_igmpsrc ?
2734     */
2735    mrtstat.mrts_upcalls++;
2736    if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2737	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2738	++mrtstat.mrts_upq_sockfull;
2739    }
2740}
2741
2742/*
2743 * Compute the timeout hash value for the bw_meter entries
2744 */
2745#define	BW_METER_TIMEHASH(bw_meter, hash)				\
2746    do {								\
2747	struct timeval next_timeval = (bw_meter)->bm_start_time;	\
2748									\
2749	BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2750	(hash) = next_timeval.tv_sec;					\
2751	if (next_timeval.tv_usec)					\
2752	    (hash)++; /* XXX: make sure we don't timeout early */	\
2753	(hash) %= BW_METER_BUCKETS;					\
2754    } while (0)
2755
2756/*
2757 * Schedule a timer to process periodically bw_meter entry of type "<="
2758 * by linking the entry in the proper hash bucket.
2759 */
2760static void
2761schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2762{
2763    int time_hash;
2764
2765    MFC_LOCK_ASSERT();
2766
2767    if (!(x->bm_flags & BW_METER_LEQ))
2768	return;		/* XXX: we schedule timers only for "<=" entries */
2769
2770    /*
2771     * Reset the bw_meter entry
2772     */
2773    x->bm_start_time = *nowp;
2774    x->bm_measured.b_packets = 0;
2775    x->bm_measured.b_bytes = 0;
2776    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2777
2778    /*
2779     * Compute the timeout hash value and insert the entry
2780     */
2781    BW_METER_TIMEHASH(x, time_hash);
2782    x->bm_time_next = bw_meter_timers[time_hash];
2783    bw_meter_timers[time_hash] = x;
2784    x->bm_time_hash = time_hash;
2785}
2786
2787/*
2788 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2789 * by removing the entry from the proper hash bucket.
2790 */
2791static void
2792unschedule_bw_meter(struct bw_meter *x)
2793{
2794    int time_hash;
2795    struct bw_meter *prev, *tmp;
2796
2797    MFC_LOCK_ASSERT();
2798
2799    if (!(x->bm_flags & BW_METER_LEQ))
2800	return;		/* XXX: we schedule timers only for "<=" entries */
2801
2802    /*
2803     * Compute the timeout hash value and delete the entry
2804     */
2805    time_hash = x->bm_time_hash;
2806    if (time_hash >= BW_METER_BUCKETS)
2807	return;		/* Entry was not scheduled */
2808
2809    for (prev = NULL, tmp = bw_meter_timers[time_hash];
2810	     tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2811	if (tmp == x)
2812	    break;
2813
2814    if (tmp == NULL)
2815	panic("unschedule_bw_meter: bw_meter entry not found");
2816
2817    if (prev != NULL)
2818	prev->bm_time_next = x->bm_time_next;
2819    else
2820	bw_meter_timers[time_hash] = x->bm_time_next;
2821
2822    x->bm_time_next = NULL;
2823    x->bm_time_hash = BW_METER_BUCKETS;
2824}
2825
2826
2827/*
2828 * Process all "<=" type of bw_meter that should be processed now,
2829 * and for each entry prepare an upcall if necessary. Each processed
2830 * entry is rescheduled again for the (periodic) processing.
2831 *
2832 * This is run periodically (once per second normally). On each round,
2833 * all the potentially matching entries are in the hash slot that we are
2834 * looking at.
2835 */
2836static void
2837bw_meter_process()
2838{
2839    static uint32_t last_tv_sec;	/* last time we processed this */
2840
2841    uint32_t loops;
2842    int i;
2843    struct timeval now, process_endtime;
2844
2845    GET_TIME(now);
2846    if (last_tv_sec == now.tv_sec)
2847	return;		/* nothing to do */
2848
2849    loops = now.tv_sec - last_tv_sec;
2850    last_tv_sec = now.tv_sec;
2851    if (loops > BW_METER_BUCKETS)
2852	loops = BW_METER_BUCKETS;
2853
2854    MFC_LOCK();
2855    /*
2856     * Process all bins of bw_meter entries from the one after the last
2857     * processed to the current one. On entry, i points to the last bucket
2858     * visited, so we need to increment i at the beginning of the loop.
2859     */
2860    for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2861	struct bw_meter *x, *tmp_list;
2862
2863	if (++i >= BW_METER_BUCKETS)
2864	    i = 0;
2865
2866	/* Disconnect the list of bw_meter entries from the bin */
2867	tmp_list = bw_meter_timers[i];
2868	bw_meter_timers[i] = NULL;
2869
2870	/* Process the list of bw_meter entries */
2871	while (tmp_list != NULL) {
2872	    x = tmp_list;
2873	    tmp_list = tmp_list->bm_time_next;
2874
2875	    /* Test if the time interval is over */
2876	    process_endtime = x->bm_start_time;
2877	    BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2878	    if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2879		/* Not yet: reschedule, but don't reset */
2880		int time_hash;
2881
2882		BW_METER_TIMEHASH(x, time_hash);
2883		if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2884		    /*
2885		     * XXX: somehow the bin processing is a bit ahead of time.
2886		     * Put the entry in the next bin.
2887		     */
2888		    if (++time_hash >= BW_METER_BUCKETS)
2889			time_hash = 0;
2890		}
2891		x->bm_time_next = bw_meter_timers[time_hash];
2892		bw_meter_timers[time_hash] = x;
2893		x->bm_time_hash = time_hash;
2894
2895		continue;
2896	    }
2897
2898	    /*
2899	     * Test if we should deliver an upcall
2900	     */
2901	    if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2902		 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2903		((x->bm_flags & BW_METER_UNIT_BYTES) &&
2904		 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2905		/* Prepare an upcall for delivery */
2906		bw_meter_prepare_upcall(x, &now);
2907	    }
2908
2909	    /*
2910	     * Reschedule for next processing
2911	     */
2912	    schedule_bw_meter(x, &now);
2913	}
2914    }
2915
2916    /* Send all upcalls that are pending delivery */
2917    bw_upcalls_send();
2918
2919    MFC_UNLOCK();
2920}
2921
2922/*
2923 * A periodic function for sending all upcalls that are pending delivery
2924 */
2925static void
2926expire_bw_upcalls_send(void *unused)
2927{
2928    MFC_LOCK();
2929    bw_upcalls_send();
2930    MFC_UNLOCK();
2931
2932    callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2933	expire_bw_upcalls_send, NULL);
2934}
2935
2936/*
2937 * A periodic function for periodic scanning of the multicast forwarding
2938 * table for processing all "<=" bw_meter entries.
2939 */
2940static void
2941expire_bw_meter_process(void *unused)
2942{
2943    if (mrt_api_config & MRT_MFC_BW_UPCALL)
2944	bw_meter_process();
2945
2946    callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
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 */
2958static int
2959pim_register_send(struct ip *ip, struct vif *vifp,
2960	struct mbuf *m, struct mfc *rt)
2961{
2962    struct mbuf *mb_copy, *mm;
2963
2964    if (mrtdebug & DEBUG_PIM)
2965	log(LOG_DEBUG, "pim_register_send: ");
2966
2967    mb_copy = pim_register_prepare(ip, m);
2968    if (mb_copy == NULL)
2969	return ENOBUFS;
2970
2971    /*
2972     * Send all the fragments. Note that the mbuf for each fragment
2973     * is freed by the sending machinery.
2974     */
2975    for (mm = mb_copy; mm; mm = mb_copy) {
2976	mb_copy = mm->m_nextpkt;
2977	mm->m_nextpkt = 0;
2978	mm = m_pullup(mm, sizeof(struct ip));
2979	if (mm != NULL) {
2980	    ip = mtod(mm, struct ip *);
2981	    if ((mrt_api_config & MRT_MFC_RP) &&
2982		(rt->mfc_rp.s_addr != INADDR_ANY)) {
2983		pim_register_send_rp(ip, vifp, mm, rt);
2984	    } else {
2985		pim_register_send_upcall(ip, vifp, mm, rt);
2986	    }
2987	}
2988    }
2989
2990    return 0;
2991}
2992
2993/*
2994 * Return a copy of the data packet that is ready for PIM Register
2995 * encapsulation.
2996 * XXX: Note that in the returned copy the IP header is a valid one.
2997 */
2998static struct mbuf *
2999pim_register_prepare(struct ip *ip, struct mbuf *m)
3000{
3001    struct mbuf *mb_copy = NULL;
3002    int mtu;
3003
3004    /* Take care of delayed checksums */
3005    if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3006	in_delayed_cksum(m);
3007	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
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_copypacket(m, M_DONTWAIT);
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 (ip->ip_len <= mtu) {
3029	/* Turn the IP header into a valid one */
3030	ip->ip_len = htons(ip->ip_len);
3031	ip->ip_off = htons(ip->ip_off);
3032	ip->ip_sum = 0;
3033	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3034    } else {
3035	/* Fragment the packet */
3036	if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
3037	    m_freem(mb_copy);
3038	    return NULL;
3039	}
3040    }
3041    return mb_copy;
3042}
3043
3044/*
3045 * Send an upcall with the data packet to the user-level process.
3046 */
3047static int
3048pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3049	struct mbuf *mb_copy, struct mfc *rt)
3050{
3051    struct mbuf *mb_first;
3052    int len = ntohs(ip->ip_len);
3053    struct igmpmsg *im;
3054    struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
3055
3056    VIF_LOCK_ASSERT();
3057
3058    /*
3059     * Add a new mbuf with an upcall header
3060     */
3061    MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
3062    if (mb_first == NULL) {
3063	m_freem(mb_copy);
3064	return ENOBUFS;
3065    }
3066    mb_first->m_data += max_linkhdr;
3067    mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3068    mb_first->m_len = sizeof(struct igmpmsg);
3069    mb_first->m_next = mb_copy;
3070
3071    /* Send message to routing daemon */
3072    im = mtod(mb_first, struct igmpmsg *);
3073    im->im_msgtype	= IGMPMSG_WHOLEPKT;
3074    im->im_mbz		= 0;
3075    im->im_vif		= vifp - viftable;
3076    im->im_src		= ip->ip_src;
3077    im->im_dst		= ip->ip_dst;
3078
3079    k_igmpsrc.sin_addr	= ip->ip_src;
3080
3081    mrtstat.mrts_upcalls++;
3082
3083    if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3084	if (mrtdebug & DEBUG_PIM)
3085	    log(LOG_WARNING,
3086		"mcast: pim_register_send_upcall: ip_mrouter socket queue full");
3087	++mrtstat.mrts_upq_sockfull;
3088	return ENOBUFS;
3089    }
3090
3091    /* Keep statistics */
3092    pimstat.pims_snd_registers_msgs++;
3093    pimstat.pims_snd_registers_bytes += len;
3094
3095    return 0;
3096}
3097
3098/*
3099 * Encapsulate the data packet in PIM Register message and send it to the RP.
3100 */
3101static int
3102pim_register_send_rp(struct ip *ip, struct vif *vifp,
3103	struct mbuf *mb_copy, struct mfc *rt)
3104{
3105    struct mbuf *mb_first;
3106    struct ip *ip_outer;
3107    struct pim_encap_pimhdr *pimhdr;
3108    int len = ntohs(ip->ip_len);
3109    vifi_t vifi = rt->mfc_parent;
3110
3111    VIF_LOCK_ASSERT();
3112
3113    if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) {
3114	m_freem(mb_copy);
3115	return EADDRNOTAVAIL;		/* The iif vif is invalid */
3116    }
3117
3118    /*
3119     * Add a new mbuf with the encapsulating header
3120     */
3121    MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
3122    if (mb_first == NULL) {
3123	m_freem(mb_copy);
3124	return ENOBUFS;
3125    }
3126    mb_first->m_data += max_linkhdr;
3127    mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3128    mb_first->m_next = mb_copy;
3129
3130    mb_first->m_pkthdr.len = len + mb_first->m_len;
3131
3132    /*
3133     * Fill in the encapsulating IP and PIM header
3134     */
3135    ip_outer = mtod(mb_first, struct ip *);
3136    *ip_outer = pim_encap_iphdr;
3137    ip_outer->ip_id = ip_newid();
3138    ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3139    ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3140    ip_outer->ip_dst = rt->mfc_rp;
3141    /*
3142     * Copy the inner header TOS to the outer header, and take care of the
3143     * IP_DF bit.
3144     */
3145    ip_outer->ip_tos = ip->ip_tos;
3146    if (ntohs(ip->ip_off) & IP_DF)
3147	ip_outer->ip_off |= IP_DF;
3148    pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
3149					 + sizeof(pim_encap_iphdr));
3150    *pimhdr = pim_encap_pimhdr;
3151    /* If the iif crosses a border, set the Border-bit */
3152    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3153	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3154
3155    mb_first->m_data += sizeof(pim_encap_iphdr);
3156    pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3157    mb_first->m_data -= sizeof(pim_encap_iphdr);
3158
3159    if (vifp->v_rate_limit == 0)
3160	tbf_send_packet(vifp, mb_first);
3161    else
3162	tbf_control(vifp, mb_first, ip, ip_outer->ip_len);
3163
3164    /* Keep statistics */
3165    pimstat.pims_snd_registers_msgs++;
3166    pimstat.pims_snd_registers_bytes += len;
3167
3168    return 0;
3169}
3170
3171/*
3172 * PIM-SMv2 and PIM-DM messages processing.
3173 * Receives and verifies the PIM control messages, and passes them
3174 * up to the listening socket, using rip_input().
3175 * The only message with special processing is the PIM_REGISTER message
3176 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3177 * is passed to if_simloop().
3178 */
3179void
3180pim_input(struct mbuf *m, int off)
3181{
3182    struct ip *ip = mtod(m, struct ip *);
3183    struct pim *pim;
3184    int minlen;
3185    int datalen = ip->ip_len;
3186    int ip_tos;
3187    int iphlen = off;
3188
3189    /* Keep statistics */
3190    pimstat.pims_rcv_total_msgs++;
3191    pimstat.pims_rcv_total_bytes += datalen;
3192
3193    /*
3194     * Validate lengths
3195     */
3196    if (datalen < PIM_MINLEN) {
3197	pimstat.pims_rcv_tooshort++;
3198	log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3199	    datalen, (u_long)ip->ip_src.s_addr);
3200	m_freem(m);
3201	return;
3202    }
3203
3204    /*
3205     * If the packet is at least as big as a REGISTER, go agead
3206     * and grab the PIM REGISTER header size, to avoid another
3207     * possible m_pullup() later.
3208     *
3209     * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
3210     * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3211     */
3212    minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3213    /*
3214     * Get the IP and PIM headers in contiguous memory, and
3215     * possibly the PIM REGISTER header.
3216     */
3217    if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3218	(m = m_pullup(m, minlen)) == 0) {
3219	log(LOG_ERR, "pim_input: m_pullup failure\n");
3220	return;
3221    }
3222    /* m_pullup() may have given us a new mbuf so reset ip. */
3223    ip = mtod(m, struct ip *);
3224    ip_tos = ip->ip_tos;
3225
3226    /* adjust mbuf to point to the PIM header */
3227    m->m_data += iphlen;
3228    m->m_len  -= iphlen;
3229    pim = mtod(m, struct pim *);
3230
3231    /*
3232     * Validate checksum. If PIM REGISTER, exclude the data packet.
3233     *
3234     * XXX: some older PIMv2 implementations don't make this distinction,
3235     * so for compatibility reason perform the checksum over part of the
3236     * message, and if error, then over the whole message.
3237     */
3238    if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3239	/* do nothing, checksum okay */
3240    } else if (in_cksum(m, datalen)) {
3241	pimstat.pims_rcv_badsum++;
3242	if (mrtdebug & DEBUG_PIM)
3243	    log(LOG_DEBUG, "pim_input: invalid checksum");
3244	m_freem(m);
3245	return;
3246    }
3247
3248    /* PIM version check */
3249    if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3250	pimstat.pims_rcv_badversion++;
3251	log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3252	    PIM_VT_V(pim->pim_vt), PIM_VERSION);
3253	m_freem(m);
3254	return;
3255    }
3256
3257    /* restore mbuf back to the outer IP */
3258    m->m_data -= iphlen;
3259    m->m_len  += iphlen;
3260
3261    if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3262	/*
3263	 * Since this is a REGISTER, we'll make a copy of the register
3264	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3265	 * routing daemon.
3266	 */
3267	struct sockaddr_in dst = { sizeof(dst), AF_INET };
3268	struct mbuf *mcp;
3269	struct ip *encap_ip;
3270	u_int32_t *reghdr;
3271	struct ifnet *vifp;
3272
3273	VIF_LOCK();
3274	if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3275	    VIF_UNLOCK();
3276	    if (mrtdebug & DEBUG_PIM)
3277		log(LOG_DEBUG,
3278		    "pim_input: register vif not set: %d\n", reg_vif_num);
3279	    m_freem(m);
3280	    return;
3281	}
3282	/* XXX need refcnt? */
3283	vifp = viftable[reg_vif_num].v_ifp;
3284	VIF_UNLOCK();
3285
3286	/*
3287	 * Validate length
3288	 */
3289	if (datalen < PIM_REG_MINLEN) {
3290	    pimstat.pims_rcv_tooshort++;
3291	    pimstat.pims_rcv_badregisters++;
3292	    log(LOG_ERR,
3293		"pim_input: register packet size too small %d from %lx\n",
3294		datalen, (u_long)ip->ip_src.s_addr);
3295	    m_freem(m);
3296	    return;
3297	}
3298
3299	reghdr = (u_int32_t *)(pim + 1);
3300	encap_ip = (struct ip *)(reghdr + 1);
3301
3302	if (mrtdebug & DEBUG_PIM) {
3303	    log(LOG_DEBUG,
3304		"pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3305		(u_long)ntohl(encap_ip->ip_src.s_addr),
3306		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3307		ntohs(encap_ip->ip_len));
3308	}
3309
3310	/* verify the version number of the inner packet */
3311	if (encap_ip->ip_v != IPVERSION) {
3312	    pimstat.pims_rcv_badregisters++;
3313	    if (mrtdebug & DEBUG_PIM) {
3314		log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3315		    "of the inner packet\n", encap_ip->ip_v);
3316	    }
3317	    m_freem(m);
3318	    return;
3319	}
3320
3321	/* verify the inner packet is destined to a mcast group */
3322	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
3323	    pimstat.pims_rcv_badregisters++;
3324	    if (mrtdebug & DEBUG_PIM)
3325		log(LOG_DEBUG,
3326		    "pim_input: inner packet of register is not "
3327		    "multicast %lx\n",
3328		    (u_long)ntohl(encap_ip->ip_dst.s_addr));
3329	    m_freem(m);
3330	    return;
3331	}
3332
3333	/* If a NULL_REGISTER, pass it to the daemon */
3334	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3335	    goto pim_input_to_daemon;
3336
3337	/*
3338	 * Copy the TOS from the outer IP header to the inner IP header.
3339	 */
3340	if (encap_ip->ip_tos != ip_tos) {
3341	    /* Outer TOS -> inner TOS */
3342	    encap_ip->ip_tos = ip_tos;
3343	    /* Recompute the inner header checksum. Sigh... */
3344
3345	    /* adjust mbuf to point to the inner IP header */
3346	    m->m_data += (iphlen + PIM_MINLEN);
3347	    m->m_len  -= (iphlen + PIM_MINLEN);
3348
3349	    encap_ip->ip_sum = 0;
3350	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3351
3352	    /* restore mbuf to point back to the outer IP header */
3353	    m->m_data -= (iphlen + PIM_MINLEN);
3354	    m->m_len  += (iphlen + PIM_MINLEN);
3355	}
3356
3357	/*
3358	 * Decapsulate the inner IP packet and loopback to forward it
3359	 * as a normal multicast packet. Also, make a copy of the
3360	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
3361	 * to pass to the daemon later, so it can take the appropriate
3362	 * actions (e.g., send back PIM_REGISTER_STOP).
3363	 * XXX: here m->m_data points to the outer IP header.
3364	 */
3365	mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
3366	if (mcp == NULL) {
3367	    log(LOG_ERR,
3368		"pim_input: pim register: could not copy register head\n");
3369	    m_freem(m);
3370	    return;
3371	}
3372
3373	/* Keep statistics */
3374	/* XXX: registers_bytes include only the encap. mcast pkt */
3375	pimstat.pims_rcv_registers_msgs++;
3376	pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3377
3378	/*
3379	 * forward the inner ip packet; point m_data at the inner ip.
3380	 */
3381	m_adj(m, iphlen + PIM_MINLEN);
3382
3383	if (mrtdebug & DEBUG_PIM) {
3384	    log(LOG_DEBUG,
3385		"pim_input: forwarding decapsulated register: "
3386		"src %lx, dst %lx, vif %d\n",
3387		(u_long)ntohl(encap_ip->ip_src.s_addr),
3388		(u_long)ntohl(encap_ip->ip_dst.s_addr),
3389		reg_vif_num);
3390	}
3391	/* NB: vifp was collected above; can it change on us? */
3392	if_simloop(vifp, m, dst.sin_family, 0);
3393
3394	/* prepare the register head to send to the mrouting daemon */
3395	m = mcp;
3396    }
3397
3398pim_input_to_daemon:
3399    /*
3400     * Pass the PIM message up to the daemon; if it is a Register message,
3401     * pass the 'head' only up to the daemon. This includes the
3402     * outer IP header, PIM header, PIM-Register header and the
3403     * inner IP header.
3404     * XXX: the outer IP header pkt size of a Register is not adjust to
3405     * reflect the fact that the inner multicast data is truncated.
3406     */
3407    rip_input(m, iphlen);
3408
3409    return;
3410}
3411#endif /* PIM */
3412
3413static int
3414ip_mroute_modevent(module_t mod, int type, void *unused)
3415{
3416    switch (type) {
3417    case MOD_LOAD:
3418	mtx_init(&mrouter_mtx, "mrouter initialization", NULL, MTX_DEF);
3419	MFC_LOCK_INIT();
3420	VIF_LOCK_INIT();
3421	ip_mrouter_reset();
3422	ip_mcast_src = X_ip_mcast_src;
3423	ip_mforward = X_ip_mforward;
3424	ip_mrouter_done = X_ip_mrouter_done;
3425	ip_mrouter_get = X_ip_mrouter_get;
3426	ip_mrouter_set = X_ip_mrouter_set;
3427	ip_rsvp_force_done = X_ip_rsvp_force_done;
3428	ip_rsvp_vif = X_ip_rsvp_vif;
3429	legal_vif_num = X_legal_vif_num;
3430	mrt_ioctl = X_mrt_ioctl;
3431	rsvp_input_p = X_rsvp_input;
3432	break;
3433
3434    case MOD_UNLOAD:
3435	/*
3436	 * Typically module unload happens after the user-level
3437	 * process has shutdown the kernel services (the check
3438	 * below insures someone can't just yank the module out
3439	 * from under a running process).  But if the module is
3440	 * just loaded and then unloaded w/o starting up a user
3441	 * process we still need to cleanup.
3442	 */
3443	if (ip_mrouter)
3444	    return EINVAL;
3445
3446	X_ip_mrouter_done();
3447	ip_mcast_src = NULL;
3448	ip_mforward = NULL;
3449	ip_mrouter_done = NULL;
3450	ip_mrouter_get = NULL;
3451	ip_mrouter_set = NULL;
3452	ip_rsvp_force_done = NULL;
3453	ip_rsvp_vif = NULL;
3454	legal_vif_num = NULL;
3455	mrt_ioctl = NULL;
3456	rsvp_input_p = NULL;
3457	VIF_LOCK_DESTROY();
3458	MFC_LOCK_DESTROY();
3459	mtx_destroy(&mrouter_mtx);
3460	break;
3461    default:
3462	return EOPNOTSUPP;
3463    }
3464    return 0;
3465}
3466
3467static moduledata_t ip_mroutemod = {
3468    "ip_mroute",
3469    ip_mroute_modevent,
3470    0
3471};
3472DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
3473