13 */
14
15
16#include <sys/param.h>
17#include <sys/systm.h>
18#include <sys/mbuf.h>
19#include <sys/socket.h>
20#include <sys/socketvar.h>
21#include <sys/protosw.h>
22#include <sys/errno.h>
23#include <sys/time.h>
24#include <sys/kernel.h>
25#include <sys/ioctl.h>
26#include <sys/syslog.h>
27#include <sys/queue.h>
28#include <net/if.h>
29#include <net/route.h>
30#include <netinet/in.h>
31#include <netinet/in_systm.h>
32#include <netinet/ip.h>
33#include <netinet/ip_var.h>
34#include <netinet/in_pcb.h>
35#include <netinet/in_var.h>
36#include <netinet/igmp.h>
37#include <netinet/igmp_var.h>
38#include <netinet/ip_mroute.h>
39#include <netinet/udp.h>
40
41#ifndef NTOHL
42#if BYTE_ORDER != BIG_ENDIAN
43#define NTOHL(d) ((d) = ntohl((d)))
44#define NTOHS(d) ((d) = ntohs((u_short)(d)))
45#define HTONL(d) ((d) = htonl((d)))
46#define HTONS(d) ((d) = htons((u_short)(d)))
47#else
48#define NTOHL(d)
49#define NTOHS(d)
50#define HTONL(d)
51#define HTONS(d)
52#endif
53#endif
54
55extern int rsvp_on;
56
57#ifndef MROUTING
58/*
59 * Dummy routines and globals used when multicast routing is not compiled in.
60 */
61
62struct socket *ip_mrouter = NULL;
63u_int ip_mrtproto = 0;
64struct mrtstat mrtstat;
65u_int rsvpdebug = 0;
66
67int
68_ip_mrouter_set(cmd, so, m)
69 int cmd;
70 struct socket *so;
71 struct mbuf *m;
72{
73 return(EOPNOTSUPP);
74}
75
76int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = _ip_mrouter_set;
77
78
79int
80_ip_mrouter_get(cmd, so, m)
81 int cmd;
82 struct socket *so;
83 struct mbuf **m;
84{
85 return(EOPNOTSUPP);
86}
87
88int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = _ip_mrouter_get;
89
90int
91_ip_mrouter_done()
92{
93 return(0);
94}
95
96int (*ip_mrouter_done)(void) = _ip_mrouter_done;
97
98int
99_ip_mforward(ip, ifp, m, imo)
100 struct ip *ip;
101 struct ifnet *ifp;
102 struct mbuf *m;
103 struct ip_moptions *imo;
104{
105 return(0);
106}
107
108int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
109 struct ip_moptions *) = _ip_mforward;
110
111int
112_mrt_ioctl(int req, caddr_t data, struct proc *p)
113{
114 return EOPNOTSUPP;
115}
116
117int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl;
118
119void
120rsvp_input(m, iphlen) /* XXX must fixup manually */
121 struct mbuf *m;
122 int iphlen;
123{
124 /* Can still get packets with rsvp_on = 0 if there is a local member
125 * of the group to which the RSVP packet is addressed. But in this
126 * case we want to throw the packet away.
127 */
128 if (!rsvp_on) {
129 m_freem(m);
130 return;
131 }
132
133 if (ip_rsvpd != NULL) {
134 if (rsvpdebug)
135 printf("rsvp_input: Sending packet up old-style socket\n");
136 rip_input(m);
137 return;
138 }
139 /* Drop the packet */
140 m_freem(m);
141}
142
143void ipip_input(struct mbuf *m) { /* XXX must fixup manually */
144 rip_input(m);
145}
146
147int (*legal_vif_num)(int) = 0;
148
149/*
150 * This should never be called, since IP_MULTICAST_VIF should fail, but
151 * just in case it does get called, the code a little lower in ip_output
152 * will assign the packet a local address.
153 */
154u_long
155_ip_mcast_src(int vifi) { return INADDR_ANY; }
156u_long (*ip_mcast_src)(int) = _ip_mcast_src;
157
158int
159ip_rsvp_vif_init(so, m)
160 struct socket *so;
161 struct mbuf *m;
162{
163 return(EINVAL);
164}
165
166int
167ip_rsvp_vif_done(so, m)
168 struct socket *so;
169 struct mbuf *m;
170{
171 return(EINVAL);
172}
173
174void
175ip_rsvp_force_done(so)
176 struct socket *so;
177{
178 return;
179}
180
181#else /* MROUTING */
182
183#define M_HASCL(m) ((m)->m_flags & M_EXT)
184
185#define INSIZ sizeof(struct in_addr)
186#define same(a1, a2) \
187 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0)
188
189#define MT_MRTABLE MT_RTABLE /* since nothing else uses it */
190
191/*
192 * Globals. All but ip_mrouter and ip_mrtproto could be static,
193 * except for netstat or debugging purposes.
194 */
195#ifndef MROUTE_LKM
196struct socket *ip_mrouter = NULL;
197struct mrtstat mrtstat;
198
199int ip_mrtproto = IGMP_DVMRP; /* for netstat only */
200#else /* MROUTE_LKM */
201extern struct mrtstat mrtstat;
202extern int ip_mrtproto;
203#endif
204
205#define NO_RTE_FOUND 0x1
206#define RTE_FOUND 0x2
207
208struct mbuf *mfctable[MFCTBLSIZ];
209u_char nexpire[MFCTBLSIZ];
210struct vif viftable[MAXVIFS];
211u_int mrtdebug = 0; /* debug level */
212#define DEBUG_MFC 0x02
213#define DEBUG_FORWARD 0x04
214#define DEBUG_EXPIRE 0x08
215#define DEBUG_XMIT 0x10
216u_int tbfdebug = 0; /* tbf debug level */
217u_int rsvpdebug = 0; /* rsvp debug level */
218
219#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
220#define UPCALL_EXPIRE 6 /* number of timeouts */
221
222/*
223 * Define the token bucket filter structures
224 * tbftable -> each vif has one of these for storing info
225 * qtable -> each interface has an associated queue of pkts
226 */
227
228struct tbf tbftable[MAXVIFS];
229struct pkt_queue qtable[MAXVIFS][MAXQSIZE];
230
231/*
232 * 'Interfaces' associated with decapsulator (so we can tell
233 * packets that went through it from ones that get reflected
234 * by a broken gateway). These interfaces are never linked into
235 * the system ifnet list & no routes point to them. I.e., packets
236 * can't be sent this way. They only exist as a placeholder for
237 * multicast source verification.
238 */
239struct ifnet multicast_decap_if[MAXVIFS];
240
241#define ENCAP_TTL 64
242#define ENCAP_PROTO IPPROTO_IPIP /* 4 */
243
244/* prototype IP hdr for encapsulated packets */
245struct ip multicast_encap_iphdr = {
246#if BYTE_ORDER == LITTLE_ENDIAN
247 sizeof(struct ip) >> 2, IPVERSION,
248#else
249 IPVERSION, sizeof(struct ip) >> 2,
250#endif
251 0, /* tos */
252 sizeof(struct ip), /* total length */
253 0, /* id */
254 0, /* frag offset */
255 ENCAP_TTL, ENCAP_PROTO,
256 0, /* checksum */
257};
258
259/*
260 * Private variables.
261 */
262static vifi_t numvifs = 0;
263static void (*encap_oldrawip)() = 0;
264static int have_encap_tunnel = 0;
265
266/*
267 * one-back cache used by ipip_input to locate a tunnel's vif
268 * given a datagram's src ip address.
269 */
270static u_long last_encap_src;
271static struct vif *last_encap_vif;
272
273static int get_sg_cnt(struct sioc_sg_req *);
274static int get_vif_cnt(struct sioc_vif_req *);
275int ip_mrouter_init(struct socket *, struct mbuf *);
276static int add_vif(struct vifctl *);
277static int del_vif(vifi_t *);
278static int add_mfc(struct mfcctl *);
279static int del_mfc(struct mfcctl *);
280static int get_version(struct mbuf *);
281static int get_assert(struct mbuf *);
282static int set_assert(int *);
283static void expire_upcalls(void *);
284static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *,
285 vifi_t);
286static void phyint_send(struct ip *, struct vif *, struct mbuf *);
287static void encap_send(struct ip *, struct vif *, struct mbuf *);
288static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long,
289 struct ip_moptions *);
290static void tbf_queue(struct vif *, struct mbuf *, struct ip *, struct ip_moptions *);
291static void tbf_process_q(struct vif *);
292static void tbf_dequeue(struct vif *, int);
293static void tbf_reprocess_q(void *);
294static int tbf_dq_sel(struct vif *, struct ip *);
295static void tbf_send_packet(struct vif *, struct mbuf *, struct ip_moptions *);
296static void tbf_update_tokens(struct vif *);
297static int priority(struct vif *, struct ip *);
298void multiencap_decap(struct mbuf *);
299
300/*
301 * whether or not special PIM assert processing is enabled.
302 */
303static int pim_assert;
304/*
305 * Rate limit for assert notification messages, in usec
306 */
307#define ASSERT_MSG_TIME 3000000
308
309/*
310 * Hash function for a source, group entry
311 */
312#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
313 ((g) >> 20) ^ ((g) >> 10) ^ (g))
314
315/*
316 * Find a route for a given origin IP address and Multicast group address
317 * Type of service parameter to be added in the future!!!
318 */
319
320#define MFCFIND(o, g, rt) { \
321 register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \
322 register struct mfc *_rt = NULL; \
323 rt = NULL; \
324 ++mrtstat.mrts_mfc_lookups; \
325 while (_mb_rt) { \
326 _rt = mtod(_mb_rt, struct mfc *); \
327 if ((_rt->mfc_origin.s_addr == o) && \
328 (_rt->mfc_mcastgrp.s_addr == g) && \
329 (_mb_rt->m_act == NULL)) { \
330 rt = _rt; \
331 break; \
332 } \
333 _mb_rt = _mb_rt->m_next; \
334 } \
335 if (rt == NULL) { \
336 ++mrtstat.mrts_mfc_misses; \
337 } \
338}
339
340
341/*
342 * Macros to compute elapsed time efficiently
343 * Borrowed from Van Jacobson's scheduling code
344 */
345#define TV_DELTA(a, b, delta) { \
346 register int xxs; \
347 \
348 delta = (a).tv_usec - (b).tv_usec; \
349 if ((xxs = (a).tv_sec - (b).tv_sec)) { \
350 switch (xxs) { \
351 case 2: \
352 delta += 1000000; \
353 /* fall through */ \
354 case 1: \
355 delta += 1000000; \
356 break; \
357 default: \
358 delta += (1000000 * xxs); \
359 } \
360 } \
361}
362
363#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
364 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
365
366#ifdef UPCALL_TIMING
367u_long upcall_data[51];
368static void collate(struct timeval *);
369#endif /* UPCALL_TIMING */
370
371
372/*
373 * Handle MRT setsockopt commands to modify the multicast routing tables.
374 */
375int
376X_ip_mrouter_set(cmd, so, m)
377 int cmd;
378 struct socket *so;
379 struct mbuf *m;
380{
381 if (cmd != MRT_INIT && so != ip_mrouter) return EACCES;
382
383 switch (cmd) {
384 case MRT_INIT: return ip_mrouter_init(so, m);
385 case MRT_DONE: return ip_mrouter_done();
386 case MRT_ADD_VIF: return add_vif (mtod(m, struct vifctl *));
387 case MRT_DEL_VIF: return del_vif (mtod(m, vifi_t *));
388 case MRT_ADD_MFC: return add_mfc (mtod(m, struct mfcctl *));
389 case MRT_DEL_MFC: return del_mfc (mtod(m, struct mfcctl *));
390 case MRT_ASSERT: return set_assert(mtod(m, int *));
391 default: return EOPNOTSUPP;
392 }
393}
394
395#ifndef MROUTE_LKM
396int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = X_ip_mrouter_set;
397#endif
398
399/*
400 * Handle MRT getsockopt commands
401 */
402int
403X_ip_mrouter_get(cmd, so, m)
404 int cmd;
405 struct socket *so;
406 struct mbuf **m;
407{
408 struct mbuf *mb;
409
410 if (so != ip_mrouter) return EACCES;
411
412 *m = mb = m_get(M_WAIT, MT_SOOPTS);
413
414 switch (cmd) {
415 case MRT_VERSION: return get_version(mb);
416 case MRT_ASSERT: return get_assert(mb);
417 default: return EOPNOTSUPP;
418 }
419}
420
421#ifndef MROUTE_LKM
422int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = X_ip_mrouter_get;
423#endif
424
425/*
426 * Handle ioctl commands to obtain information from the cache
427 */
428int
429X_mrt_ioctl(cmd, data)
430 int cmd;
431 caddr_t data;
432{
433 int error = 0;
434
435 switch (cmd) {
436 case (SIOCGETVIFCNT):
437 return (get_vif_cnt((struct sioc_vif_req *)data));
438 break;
439 case (SIOCGETSGCNT):
440 return (get_sg_cnt((struct sioc_sg_req *)data));
441 break;
442 default:
443 return (EINVAL);
444 break;
445 }
446 return error;
447}
448
449#ifndef MROUTE_LKM
450int (*mrt_ioctl)(int, caddr_t, struct proc *) = X_mrt_ioctl;
451#endif
452
453/*
454 * returns the packet, byte, rpf-failure count for the source group provided
455 */
456static int
457get_sg_cnt(req)
458 register struct sioc_sg_req *req;
459{
460 register struct mfc *rt;
461 int s;
462
463 s = splnet();
464 MFCFIND(req->src.s_addr, req->grp.s_addr, rt);
465 splx(s);
466 if (rt != NULL) {
467 req->pktcnt = rt->mfc_pkt_cnt;
468 req->bytecnt = rt->mfc_byte_cnt;
469 req->wrong_if = rt->mfc_wrong_if;
470 } else
471 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
472
473 return 0;
474}
475
476/*
477 * returns the input and output packet and byte counts on the vif provided
478 */
479static int
480get_vif_cnt(req)
481 register struct sioc_vif_req *req;
482{
483 register vifi_t vifi = req->vifi;
484
485 if (vifi >= numvifs) return EINVAL;
486
487 req->icount = viftable[vifi].v_pkt_in;
488 req->ocount = viftable[vifi].v_pkt_out;
489 req->ibytes = viftable[vifi].v_bytes_in;
490 req->obytes = viftable[vifi].v_bytes_out;
491
492 return 0;
493}
494
495/*
496 * Enable multicast routing
497 */
498int
499ip_mrouter_init(so, m)
500 struct socket *so;
501 struct mbuf *m;
502{
503 int *v;
504 int i;
505
506 if (mrtdebug)
507 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d",
508 so->so_type, so->so_proto->pr_protocol);
509
510 if (so->so_type != SOCK_RAW ||
511 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP;
512
513 if (!m || (m->m_len != sizeof(int *)))
514 return ENOPROTOOPT;
515
516 v = mtod(m, int *);
517 if (*v != 1)
518 return ENOPROTOOPT;
519
520 if (ip_mrouter != NULL) return EADDRINUSE;
521
522 ip_mrouter = so;
523
524 bzero((caddr_t)mfctable, sizeof(mfctable));
525 bzero((caddr_t)nexpire, sizeof(nexpire));
526
527 pim_assert = 0;
528
529 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
530
531 if (mrtdebug)
532 log(LOG_DEBUG, "ip_mrouter_init");
533
534 return 0;
535}
536
537/*
538 * Disable multicast routing
539 */
540int
541X_ip_mrouter_done()
542{
543 vifi_t vifi;
544 int i;
545 struct ifnet *ifp;
546 struct ifreq ifr;
547 struct mbuf *mb_rt;
548 struct mfc *rt;
549 struct mbuf *m;
550 struct rtdetq *rte;
551 int s;
552
553 s = splnet();
554
555 /*
556 * For each phyint in use, disable promiscuous reception of all IP
557 * multicasts.
558 */
559 for (vifi = 0; vifi < numvifs; vifi++) {
560 if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
561 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
562 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
563 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr
564 = INADDR_ANY;
565 ifp = viftable[vifi].v_ifp;
566 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
567 }
568 }
569 bzero((caddr_t)qtable, sizeof(qtable));
570 bzero((caddr_t)tbftable, sizeof(tbftable));
571 bzero((caddr_t)viftable, sizeof(viftable));
572 numvifs = 0;
573 pim_assert = 0;
574
575 untimeout(expire_upcalls, (caddr_t)NULL);
576
577 /*
578 * Free all multicast forwarding cache entries.
579 */
580 for (i = 0; i < MFCTBLSIZ; i++) {
581 mb_rt = mfctable[i];
582 while (mb_rt) {
583 if (mb_rt->m_act != NULL) {
584 while (mb_rt->m_act) {
585 m = mb_rt->m_act;
586 mb_rt->m_act = m->m_act;
587 rte = mtod(m, struct rtdetq *);
588 m_freem(rte->m);
589 m_free(m);
590 }
591 }
592 mb_rt = m_free(mb_rt);
593 }
594 }
595
596 bzero((caddr_t)mfctable, sizeof(mfctable));
597
598 /*
599 * Reset de-encapsulation cache
600 */
601 last_encap_src = NULL;
602 last_encap_vif = NULL;
603 have_encap_tunnel = 0;
604
605 ip_mrouter = NULL;
606
607 splx(s);
608
609 if (mrtdebug)
610 log(LOG_DEBUG, "ip_mrouter_done");
611
612 return 0;
613}
614
615#ifndef MROUTE_LKM
616int (*ip_mrouter_done)(void) = X_ip_mrouter_done;
617#endif
618
619static int
620get_version(mb)
621 struct mbuf *mb;
622{
623 int *v;
624
625 v = mtod(mb, int *);
626
627 *v = 0x0305; /* XXX !!!! */
628 mb->m_len = sizeof(int);
629
630 return 0;
631}
632
633/*
634 * Set PIM assert processing global
635 */
636static int
637set_assert(i)
638 int *i;
639{
640 if ((*i != 1) && (*i != 0))
641 return EINVAL;
642
643 pim_assert = *i;
644
645 return 0;
646}
647
648/*
649 * Get PIM assert processing global
650 */
651static int
652get_assert(m)
653 struct mbuf *m;
654{
655 int *i;
656
657 i = mtod(m, int *);
658
659 *i = pim_assert;
660
661 return 0;
662}
663
664/*
665 * Add a vif to the vif table
666 */
667static int
668add_vif(vifcp)
669 register struct vifctl *vifcp;
670{
671 register struct vif *vifp = viftable + vifcp->vifc_vifi;
672 static struct sockaddr_in sin = {sizeof sin, AF_INET};
673 struct ifaddr *ifa;
674 struct ifnet *ifp;
675 struct ifreq ifr;
676 int error, s;
677 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
678
679 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL;
680 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE;
681
682 /* Find the interface with an address in AF_INET family */
683 sin.sin_addr = vifcp->vifc_lcl_addr;
684 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
685 if (ifa == 0) return EADDRNOTAVAIL;
686 ifp = ifa->ifa_ifp;
687
688 if (vifcp->vifc_flags & VIFF_TUNNEL) {
689 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
690 /*
691 * An encapsulating tunnel is wanted. Tell ipip_input() to
692 * start paying attention to encapsulated packets.
693 */
694 if (have_encap_tunnel == 0) {
695 have_encap_tunnel = 1;
696 for (s = 0; s < MAXVIFS; ++s) {
697 multicast_decap_if[s].if_name = "mdecap";
698 multicast_decap_if[s].if_unit = s;
699 }
700 }
701 /*
702 * Set interface to fake encapsulator interface
703 */
704 ifp = &multicast_decap_if[vifcp->vifc_vifi];
705 /*
706 * Prepare cached route entry
707 */
708 bzero(&vifp->v_route, sizeof(vifp->v_route));
709 } else {
710 log(LOG_ERR, "Source routed tunnels not supported.");
711 return EOPNOTSUPP;
712 }
713 } else {
714 /* Make sure the interface supports multicast */
715 if ((ifp->if_flags & IFF_MULTICAST) == 0)
716 return EOPNOTSUPP;
717
718 /* Enable promiscuous reception of all IP multicasts from the if */
719 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
720 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
721 s = splnet();
722 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
723 splx(s);
724 if (error)
725 return error;
726 }
727
728 s = splnet();
729 /* define parameters for the tbf structure */
730 vifp->v_tbf = v_tbf;
731 vifp->v_tbf->q_len = 0;
732 vifp->v_tbf->n_tok = 0;
733 vifp->v_tbf->last_pkt_t = 0;
734
735 vifp->v_flags = vifcp->vifc_flags;
736 vifp->v_threshold = vifcp->vifc_threshold;
737 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
738 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
739 vifp->v_ifp = ifp;
740 vifp->v_rate_limit= vifcp->vifc_rate_limit;
741 vifp->v_rsvp_on = 0;
742 vifp->v_rsvpd = NULL;
743 /* initialize per vif pkt counters */
744 vifp->v_pkt_in = 0;
745 vifp->v_pkt_out = 0;
746 vifp->v_bytes_in = 0;
747 vifp->v_bytes_out = 0;
748 splx(s);
749
750 /* Adjust numvifs up if the vifi is higher than numvifs */
751 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
752
753 if (mrtdebug)
754 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d",
755 vifcp->vifc_vifi,
756 ntohl(vifcp->vifc_lcl_addr.s_addr),
757 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
758 ntohl(vifcp->vifc_rmt_addr.s_addr),
759 vifcp->vifc_threshold,
760 vifcp->vifc_rate_limit);
761
762 return 0;
763}
764
765/*
766 * Delete a vif from the vif table
767 */
768static int
769del_vif(vifip)
770 vifi_t *vifip;
771{
772 register struct vif *vifp = viftable + *vifip;
773 register vifi_t vifi;
774 struct ifnet *ifp;
775 struct ifreq ifr;
776 int s;
777
778 if (*vifip >= numvifs) return EINVAL;
779 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL;
780
781 s = splnet();
782
783 if (!(vifp->v_flags & VIFF_TUNNEL)) {
784 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
785 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
786 ifp = vifp->v_ifp;
787 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
788 }
789
790 if (vifp == last_encap_vif) {
791 last_encap_vif = 0;
792 last_encap_src = 0;
793 }
794
795 bzero((caddr_t)qtable[*vifip],
796 sizeof(qtable[*vifip]));
797 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
798 bzero((caddr_t)vifp, sizeof (*vifp));
799
800 /* Adjust numvifs down */
801 for (vifi = numvifs; vifi > 0; vifi--)
802 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break;
803 numvifs = vifi;
804
805 splx(s);
806
807 if (mrtdebug)
808 log(LOG_DEBUG, "del_vif %d, numvifs %d", *vifip, numvifs);
809
810 return 0;
811}
812
813/*
814 * Add an mfc entry
815 */
816static int
817add_mfc(mfccp)
818 struct mfcctl *mfccp;
819{
820 struct mfc *rt;
821 register struct mbuf *mb_rt;
822 u_long hash;
823 struct mbuf *mb_ntry;
824 struct rtdetq *rte;
825 register u_short nstl;
826 int s;
827 int i;
828
829 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt);
830
831 /* If an entry already exists, just update the fields */
832 if (rt) {
833 if (mrtdebug & DEBUG_MFC)
834 log(LOG_DEBUG,"add_mfc update o %x g %x p %x",
835 ntohl(mfccp->mfcc_origin.s_addr),
836 ntohl(mfccp->mfcc_mcastgrp.s_addr),
837 mfccp->mfcc_parent);
838
839 s = splnet();
840 rt->mfc_parent = mfccp->mfcc_parent;
841 for (i = 0; i < numvifs; i++)
842 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
843 splx(s);
844 return 0;
845 }
846
847 /*
848 * Find the entry for which the upcall was made and update
849 */
850 s = splnet();
851 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
852 for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) {
853
854 rt = mtod(mb_rt, struct mfc *);
855 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
856 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
857 (mb_rt->m_act != NULL)) {
858
859 if (nstl++)
860 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %x",
861 "multiple kernel entries",
862 ntohl(mfccp->mfcc_origin.s_addr),
863 ntohl(mfccp->mfcc_mcastgrp.s_addr),
864 mfccp->mfcc_parent, mb_rt->m_act);
865
866 if (mrtdebug & DEBUG_MFC)
867 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %x",
868 ntohl(mfccp->mfcc_origin.s_addr),
869 ntohl(mfccp->mfcc_mcastgrp.s_addr),
870 mfccp->mfcc_parent, mb_rt->m_act);
871
872 rt->mfc_origin = mfccp->mfcc_origin;
873 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
874 rt->mfc_parent = mfccp->mfcc_parent;
875 for (i = 0; i < numvifs; i++)
876 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
877 /* initialize pkt counters per src-grp */
878 rt->mfc_pkt_cnt = 0;
879 rt->mfc_byte_cnt = 0;
880 rt->mfc_wrong_if = 0;
881 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
882
883 rt->mfc_expire = 0; /* Don't clean this guy up */
884 nexpire[hash]--;
885
886 /* free packets Qed at the end of this entry */
887 while (mb_rt->m_act) {
888 mb_ntry = mb_rt->m_act;
889 rte = mtod(mb_ntry, struct rtdetq *);
890/* #ifdef RSVP_ISI */
891 ip_mdq(rte->m, rte->ifp, rt, -1);
892/* #endif */
893 mb_rt->m_act = mb_ntry->m_act;
894 m_freem(rte->m);
895#ifdef UPCALL_TIMING
896 collate(&(rte->t));
897#endif /* UPCALL_TIMING */
898 m_free(mb_ntry);
899 }
900 }
901 }
902
903 /*
904 * It is possible that an entry is being inserted without an upcall
905 */
906 if (nstl == 0) {
907 if (mrtdebug & DEBUG_MFC)
908 log(LOG_DEBUG,"add_mfc no upcall h %d o %x g %x p %x",
909 hash, ntohl(mfccp->mfcc_origin.s_addr),
910 ntohl(mfccp->mfcc_mcastgrp.s_addr),
911 mfccp->mfcc_parent);
912
913 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
914
915 rt = mtod(mb_rt, struct mfc *);
916 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
917 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
918
919 rt->mfc_origin = mfccp->mfcc_origin;
920 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
921 rt->mfc_parent = mfccp->mfcc_parent;
922 for (i = 0; i < numvifs; i++)
923 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
924 /* initialize pkt counters per src-grp */
925 rt->mfc_pkt_cnt = 0;
926 rt->mfc_byte_cnt = 0;
927 rt->mfc_wrong_if = 0;
928 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
929 if (rt->mfc_expire)
930 nexpire[hash]--;
931 rt->mfc_expire = 0;
932 }
933 }
934 if (mb_rt == NULL) {
935 /* no upcall, so make a new entry */
936 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
937 if (mb_rt == NULL) {
938 splx(s);
939 return ENOBUFS;
940 }
941
942 rt = mtod(mb_rt, struct mfc *);
943
944 /* insert new entry at head of hash chain */
945 rt->mfc_origin = mfccp->mfcc_origin;
946 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
947 rt->mfc_parent = mfccp->mfcc_parent;
948 for (i = 0; i < numvifs; i++)
949 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
950 /* initialize pkt counters per src-grp */
951 rt->mfc_pkt_cnt = 0;
952 rt->mfc_byte_cnt = 0;
953 rt->mfc_wrong_if = 0;
954 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
955 rt->mfc_expire = 0;
956
957 /* link into table */
958 mb_rt->m_next = mfctable[hash];
959 mfctable[hash] = mb_rt;
960 mb_rt->m_act = NULL;
961 }
962 }
963 splx(s);
964 return 0;
965}
966
967#ifdef UPCALL_TIMING
968/*
969 * collect delay statistics on the upcalls
970 */
971static void collate(t)
972register struct timeval *t;
973{
974 register u_long d;
975 register struct timeval tp;
976 register u_long delta;
977
978 GET_TIME(tp);
979
980 if (TV_LT(*t, tp))
981 {
982 TV_DELTA(tp, *t, delta);
983
984 d = delta >> 10;
985 if (d > 50)
986 d = 50;
987
988 ++upcall_data[d];
989 }
990}
991#endif /* UPCALL_TIMING */
992
993/*
994 * Delete an mfc entry
995 */
996static int
997del_mfc(mfccp)
998 struct mfcctl *mfccp;
999{
1000 struct in_addr origin;
1001 struct in_addr mcastgrp;
1002 struct mfc *rt;
1003 struct mbuf *mb_rt;
1004 struct mbuf **nptr;
1005 u_long hash;
1006 int s, i;
1007
1008 origin = mfccp->mfcc_origin;
1009 mcastgrp = mfccp->mfcc_mcastgrp;
1010 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1011
1012 if (mrtdebug & DEBUG_MFC)
1013 log(LOG_DEBUG,"del_mfc orig %x mcastgrp %x",
1014 ntohl(origin.s_addr), ntohl(mcastgrp.s_addr));
1015
1016 s = splnet();
1017
1018 nptr = &mfctable[hash];
1019 while ((mb_rt = *nptr) != NULL) {
1020 rt = mtod(mb_rt, struct mfc *);
1021 if (origin.s_addr == rt->mfc_origin.s_addr &&
1022 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1023 mb_rt->m_act == NULL)
1024 break;
1025
1026 nptr = &mb_rt->m_next;
1027 }
1028 if (mb_rt == NULL) {
1029 splx(s);
1030 return EADDRNOTAVAIL;
1031 }
1032
1033 MFREE(mb_rt, *nptr);
1034
1035 splx(s);
1036
1037 return 0;
1038}
1039
1040/*
1041 * Send a message to mrouted on the multicast routing socket
1042 */
1043static int
1044socket_send(s, mm, src)
1045 struct socket *s;
1046 struct mbuf *mm;
1047 struct sockaddr_in *src;
1048{
1049 if (s) {
1050 if (sbappendaddr(&s->so_rcv,
1051 (struct sockaddr *)src,
1052 mm, (struct mbuf *)0) != 0) {
1053 sorwakeup(s);
1054 return 0;
1055 }
1056 }
1057 m_freem(mm);
1058 return -1;
1059}
1060
1061/*
1062 * IP multicast forwarding function. This function assumes that the packet
1063 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1064 * pointed to by "ifp", and the packet is to be relayed to other networks
1065 * that have members of the packet's destination IP multicast group.
1066 *
1067 * The packet is returned unscathed to the caller, unless it is
1068 * erroneous, in which case a non-zero return value tells the caller to
1069 * discard it.
1070 */
1071
1072#define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */
1073#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1074
1075int
1076X_ip_mforward(ip, ifp, m, imo)
1077 register struct ip *ip;
1078 struct ifnet *ifp;
1079 struct mbuf *m;
1080 struct ip_moptions *imo;
1081{
1082 register struct mfc *rt = 0; /* XXX uninit warning */
1083 register u_char *ipoptions;
1084 static struct sockproto k_igmpproto = { AF_INET, IPPROTO_IGMP };
1085 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1086 static int srctun = 0;
1087 register struct mbuf *mm;
1088 int s;
1089 vifi_t vifi;
1090 struct vif *vifp;
1091
1092 if (mrtdebug & DEBUG_FORWARD)
1093 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %x",
1094 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1095
1096 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1097 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1098 /*
1099 * Packet arrived via a physical interface or
1100 * an encapsulated tunnel.
1101 */
1102 } else {
1103 /*
1104 * Packet arrived through a source-route tunnel.
1105 * Source-route tunnels are no longer supported.
1106 */
1107 if ((srctun++ % 1000) == 0)
1108 log(LOG_ERR, "ip_mforward: received source-routed packet from %x",
1109 ntohl(ip->ip_src.s_addr));
1110
1111 return 1;
1112 }
1113
1114 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1115 if (ip->ip_ttl < 255)
1116 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1117 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1118 vifp = viftable + vifi;
1119 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s%d)\n",
1120 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi,
1121 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1122 vifp->v_ifp->if_name, vifp->v_ifp->if_unit);
1123 }
1124 return (ip_mdq(m, ifp, rt, vifi));
1125 }
1126 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1127 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
1128 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr));
1129 }
1130
1131 /*
1132 * Don't forward a packet with time-to-live of zero or one,
1133 * or a packet destined to a local-only group.
1134 */
1135 if (ip->ip_ttl <= 1 ||
1136 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
1137 return 0;
1138
1139 /*
1140 * Determine forwarding vifs from the forwarding cache table
1141 */
1142 s = splnet();
1143 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt);
1144
1145 /* Entry exists, so forward if necessary */
1146 if (rt != NULL) {
1147 splx(s);
1148 return (ip_mdq(m, ifp, rt, -1));
1149 } else {
1150 /*
1151 * If we don't have a route for packet's origin,
1152 * Make a copy of the packet &
1153 * send message to routing daemon
1154 */
1155
1156 register struct mbuf *mb_rt;
1157 register struct mbuf *mb_ntry;
1158 register struct mbuf *mb0;
1159 register struct rtdetq *rte;
1160 register struct mbuf *rte_m;
1161 register u_long hash;
1162 register int npkts;
1163#ifdef UPCALL_TIMING
1164 struct timeval tp;
1165
1166 GET_TIME(tp);
1167#endif
1168
1169 mrtstat.mrts_no_route++;
1170 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1171 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x",
1172 ntohl(ip->ip_src.s_addr),
1173 ntohl(ip->ip_dst.s_addr));
1174
1175 /*
1176 * Allocate mbufs early so that we don't do extra work if we are
1177 * just going to fail anyway.
1178 */
1179 MGET(mb_ntry, M_DONTWAIT, MT_DATA);
1180 if (mb_ntry == NULL) {
1181 splx(s);
1182 return ENOBUFS;
1183 }
1184 mb0 = m_copy(m, 0, M_COPYALL);
1185 if (mb0 == NULL) {
1186 m_free(mb_ntry);
1187 splx(s);
1188 return ENOBUFS;
1189 }
1190
1191 /* is there an upcall waiting for this packet? */
1192 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1193 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
1194 rt = mtod(mb_rt, struct mfc *);
1195 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1196 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1197 (mb_rt->m_act != NULL))
1198 break;
1199 }
1200
1201 if (mb_rt == NULL) {
1202 int hlen = ip->ip_hl << 2;
1203 int i;
1204 struct igmpmsg *im;
1205
1206 /* no upcall, so make a new entry */
1207 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
1208 if (mb_rt == NULL) {
1209 m_free(mb_ntry);
1210 m_freem(mb0);
1211 splx(s);
1212 return ENOBUFS;
1213 }
1214 /* Make a copy of the header to send to the user level process */
1215 mm = m_copy(m, 0, hlen);
1216 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1217 mm = m_pullup(mm, hlen);
1218 if (mm == NULL) {
1219 m_free(mb_ntry);
1220 m_freem(mb0);
1221 m_free(mb_rt);
1222 splx(s);
1223 return ENOBUFS;
1224 }
1225
1226 /*
1227 * Send message to routing daemon to install
1228 * a route into the kernel table
1229 */
1230 k_igmpsrc.sin_addr = ip->ip_src;
1231
1232 im = mtod(mm, struct igmpmsg *);
1233 im->im_msgtype = IGMPMSG_NOCACHE;
1234 im->im_mbz = 0;
1235
1236 mrtstat.mrts_upcalls++;
1237
1238 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1239 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full");
1240 ++mrtstat.mrts_upq_sockfull;
1241 m_free(mb_ntry);
1242 m_freem(mb0);
1243 m_free(mb_rt);
1244 splx(s);
1245 return ENOBUFS;
1246 }
1247
1248 rt = mtod(mb_rt, struct mfc *);
1249
1250 /* insert new entry at head of hash chain */
1251 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1252 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1253 rt->mfc_expire = UPCALL_EXPIRE;
1254 nexpire[hash]++;
1255 for (i = 0; i < numvifs; i++)
1256 rt->mfc_ttls[i] = 0;
1257 rt->mfc_parent = -1;
1258
1259 /* link into table */
1260 mb_rt->m_next = mfctable[hash];
1261 mfctable[hash] = mb_rt;
1262 mb_rt->m_act = NULL;
1263
1264 rte_m = mb_rt;
1265 } else {
1266 /* determine if q has overflowed */
1267 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act)
1268 npkts++;
1269
1270 if (npkts > MAX_UPQ) {
1271 mrtstat.mrts_upq_ovflw++;
1272 m_free(mb_ntry);
1273 m_freem(mb0);
1274 splx(s);
1275 return 0;
1276 }
1277 }
1278
1279 mb_ntry->m_act = NULL;
1280 rte = mtod(mb_ntry, struct rtdetq *);
1281
1282 rte->m = mb0;
1283 rte->ifp = ifp;
1284#ifdef UPCALL_TIMING
1285 rte->t = tp;
1286#endif
1287
1288 /* Add this entry to the end of the queue */
1289 rte_m->m_act = mb_ntry;
1290
1291 splx(s);
1292
1293 return 0;
1294 }
1295}
1296
1297#ifndef MROUTE_LKM
1298int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
1299 struct ip_moptions *) = X_ip_mforward;
1300#endif
1301
1302/*
1303 * Clean up the cache entry if upcall is not serviced
1304 */
1305static void
1306expire_upcalls(void *unused)
1307{
1308 struct mbuf *mb_rt, *m, **nptr;
1309 struct rtdetq *rte;
1310 struct mfc *mfc;
1311 int i;
1312 int s;
1313
1314 s = splnet();
1315 for (i = 0; i < MFCTBLSIZ; i++) {
1316 if (nexpire[i] == 0)
1317 continue;
1318 nptr = &mfctable[i];
1319 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) {
1320 mfc = mtod(mb_rt, struct mfc *);
1321
1322 /*
1323 * Skip real cache entries
1324 * Make sure it wasn't marked to not expire (shouldn't happen)
1325 * If it expires now
1326 */
1327 if (mb_rt->m_act != NULL &&
1328 mfc->mfc_expire != 0 &&
1329 --mfc->mfc_expire == 0) {
1330 if (mrtdebug & DEBUG_EXPIRE)
1331 log(LOG_DEBUG, "expire_upcalls: expiring (%x %x)",
1332 ntohl(mfc->mfc_origin.s_addr),
1333 ntohl(mfc->mfc_mcastgrp.s_addr));
1334 /*
1335 * drop all the packets
1336 * free the mbuf with the pkt, if, timing info
1337 */
1338 while (mb_rt->m_act) {
1339 m = mb_rt->m_act;
1340 mb_rt->m_act = m->m_act;
1341
1342 rte = mtod(m, struct rtdetq *);
1343 m_freem(rte->m);
1344 m_free(m);
1345 }
1346 ++mrtstat.mrts_cache_cleanups;
1347 nexpire[i]--;
1348
1349 MFREE(mb_rt, *nptr);
1350 } else {
1351 nptr = &mb_rt->m_next;
1352 }
1353 }
1354 }
1355 splx(s);
1356 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
1357}
1358
1359/*
1360 * Packet forwarding routine once entry in the cache is made
1361 */
1362static int
1363ip_mdq(m, ifp, rt, xmt_vif)
1364 register struct mbuf *m;
1365 register struct ifnet *ifp;
1366 register struct mfc *rt;
1367 register vifi_t xmt_vif;
1368{
1369 register struct ip *ip = mtod(m, struct ip *);
1370 register vifi_t vifi;
1371 register struct vif *vifp;
1372 register struct mbuf *tmp;
1373 register int plen = ntohs(ip->ip_len);
1374
1375/*
1376 * Macro to send packet on vif. Since RSVP packets don't get counted on
1377 * input, they shouldn't get counted on output, so statistics keeping is
1378 * seperate.
1379 */
1380#define MC_SEND(ip,vifp,m) { \
1381 if ((vifp)->v_flags & VIFF_TUNNEL) \
1382 encap_send((ip), (vifp), (m)); \
1383 else \
1384 phyint_send((ip), (vifp), (m)); \
1385}
1386
1387 /*
1388 * If xmt_vif is not -1, send on only the requested vif.
1389 *
1390 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1391 */
1392 if (xmt_vif < numvifs) {
1393 MC_SEND(ip, viftable + xmt_vif, m);
1394 return 1;
1395 }
1396
1397 /*
1398 * Don't forward if it didn't arrive from the parent vif for its origin.
1399 */
1400 vifi = rt->mfc_parent;
1401 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1402 /* came in the wrong interface */
1403 if (mrtdebug & DEBUG_FORWARD)
1404 log(LOG_DEBUG, "wrong if: ifp %x vifi %d vififp %x",
1405 ifp, vifi, viftable[vifi].v_ifp);
1406 ++mrtstat.mrts_wrong_if;
1407 ++rt->mfc_wrong_if;
1408 /*
1409 * If we are doing PIM assert processing, and we are forwarding
1410 * packets on this interface, and it is a broadcast medium
1411 * interface (and not a tunnel), send a message to the routing daemon.
1412 */
1413 if (pim_assert && rt->mfc_ttls[vifi] &&
1414 (ifp->if_flags & IFF_BROADCAST) &&
1415 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1416 struct sockaddr_in k_igmpsrc;
1417 struct mbuf *mm;
1418 struct igmpmsg *im;
1419 int hlen = ip->ip_hl << 2;
1420 struct timeval now;
1421 register u_long delta;
1422
1423 GET_TIME(now);
1424
1425 TV_DELTA(rt->mfc_last_assert, now, delta);
1426
1427 if (delta > ASSERT_MSG_TIME) {
1428 mm = m_copy(m, 0, hlen);
1429 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1430 mm = m_pullup(mm, hlen);
1431 if (mm == NULL) {
1432 return ENOBUFS;
1433 }
1434
1435 rt->mfc_last_assert = now;
1436
1437 im = mtod(mm, struct igmpmsg *);
1438 im->im_msgtype = IGMPMSG_WRONGVIF;
1439 im->im_mbz = 0;
1440 im->im_vif = vifi;
1441
1442 k_igmpsrc.sin_addr = im->im_src;
1443
1444 socket_send(ip_mrouter, mm, &k_igmpsrc);
1445 }
1446 }
1447 return 0;
1448 }
1449
1450 /* If I sourced this packet, it counts as output, else it was input. */
1451 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1452 viftable[vifi].v_pkt_out++;
1453 viftable[vifi].v_bytes_out += plen;
1454 } else {
1455 viftable[vifi].v_pkt_in++;
1456 viftable[vifi].v_bytes_in += plen;
1457 }
1458 rt->mfc_pkt_cnt++;
1459 rt->mfc_byte_cnt += plen;
1460
1461 /*
1462 * For each vif, decide if a copy of the packet should be forwarded.
1463 * Forward if:
1464 * - the ttl exceeds the vif's threshold
1465 * - there are group members downstream on interface
1466 */
1467 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1468 if ((rt->mfc_ttls[vifi] > 0) &&
1469 (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1470 vifp->v_pkt_out++;
1471 vifp->v_bytes_out += plen;
1472 MC_SEND(ip, vifp, m);
1473 }
1474
1475 return 0;
1476}
1477
1478/*
1479 * check if a vif number is legal/ok. This is used by ip_output, to export
1480 * numvifs there,
1481 */
1482int
1483X_legal_vif_num(vif)
1484 int vif;
1485{
1486 if (vif >= 0 && vif < numvifs)
1487 return(1);
1488 else
1489 return(0);
1490}
1491
1492#ifndef MROUTE_LKM
1493int (*legal_vif_num)(int) = X_legal_vif_num;
1494#endif
1495
1496/*
1497 * Return the local address used by this vif
1498 */
1499u_long
1500X_ip_mcast_src(vifi)
1501 int vifi;
1502{
1503 if (vifi >= 0 && vifi < numvifs)
1504 return viftable[vifi].v_lcl_addr.s_addr;
1505 else
1506 return INADDR_ANY;
1507}
1508
1509#ifndef MROUTE_LKM
1510u_long (*ip_mcast_src)(int) = X_ip_mcast_src;
1511#endif
1512
1513static void
1514phyint_send(ip, vifp, m)
1515 struct ip *ip;
1516 struct vif *vifp;
1517 struct mbuf *m;
1518{
1519 register struct mbuf *mb_copy;
1520 register int hlen = ip->ip_hl << 2;
1521 register struct ip_moptions *imo;
1522
1523 /*
1524 * Make a new reference to the packet; make sure that
1525 * the IP header is actually copied, not just referenced,
1526 * so that ip_output() only scribbles on the copy.
1527 */
1528 mb_copy = m_copy(m, 0, M_COPYALL);
1529 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1530 mb_copy = m_pullup(mb_copy, hlen);
1531 if (mb_copy == NULL)
1532 return;
1533
1534 MALLOC(imo, struct ip_moptions *, sizeof *imo, M_IPMOPTS, M_NOWAIT);
1535 if (imo == NULL) {
1536 m_freem(mb_copy);
1537 return;
1538 }
1539
1540 imo->imo_multicast_ifp = vifp->v_ifp;
1541 imo->imo_multicast_ttl = ip->ip_ttl - 1;
1542 imo->imo_multicast_loop = 1;
1543 imo->imo_multicast_vif = -1;
1544
1545 if (vifp->v_rate_limit <= 0)
1546 tbf_send_packet(vifp, mb_copy, imo);
1547 else
1548 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len,
1549 imo);
1550}
1551
1552static void
1553encap_send(ip, vifp, m)
1554 register struct ip *ip;
1555 register struct vif *vifp;
1556 register struct mbuf *m;
1557{
1558 register struct mbuf *mb_copy;
1559 register struct ip *ip_copy;
1560 int hlen = ip->ip_hl << 2;
1561 register int i, len = ip->ip_len;
1562
1563 /*
1564 * copy the old packet & pullup it's IP header into the
1565 * new mbuf so we can modify it. Try to fill the new
1566 * mbuf since if we don't the ethernet driver will.
1567 */
1568 MGET(mb_copy, M_DONTWAIT, MT_DATA);
1569 if (mb_copy == NULL)
1570 return;
1571 mb_copy->m_data += 16;
1572 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1573
1574 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1575 m_freem(mb_copy);
1576 return;
1577 }
1578 i = MHLEN - M_LEADINGSPACE(mb_copy);
1579 if (i > len)
1580 i = len;
1581 mb_copy = m_pullup(mb_copy, i);
1582 if (mb_copy == NULL)
1583 return;
1584 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1585
1586 /*
1587 * fill in the encapsulating IP header.
1588 */
1589 ip_copy = mtod(mb_copy, struct ip *);
1590 *ip_copy = multicast_encap_iphdr;
1591 ip_copy->ip_id = htons(ip_id++);
1592 ip_copy->ip_len += len;
1593 ip_copy->ip_src = vifp->v_lcl_addr;
1594 ip_copy->ip_dst = vifp->v_rmt_addr;
1595
1596 /*
1597 * turn the encapsulated IP header back into a valid one.
1598 */
1599 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1600 --ip->ip_ttl;
1601 HTONS(ip->ip_len);
1602 HTONS(ip->ip_off);
1603 ip->ip_sum = 0;
1604#if defined(LBL) && !defined(ultrix)
1605 ip->ip_sum = ~oc_cksum((caddr_t)ip, ip->ip_hl << 2, 0);
1606#else
1607 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1608 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1609 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1610#endif
1611
1612 if (vifp->v_rate_limit <= 0)
1613 tbf_send_packet(vifp, mb_copy, 0);
1614 else
1615 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len, 0);
1616}
1617
1618/*
1619 * De-encapsulate a packet and feed it back through ip input (this
1620 * routine is called whenever IP gets a packet with proto type
1621 * ENCAP_PROTO and a local destination address).
1622 */
1623void
1624#ifdef MROUTE_LKM
1625X_ipip_input(m)
1626#else
1627ipip_input(m, iphlen)
1628#endif
1629 register struct mbuf *m;
1630 int iphlen;
1631{
1632 struct ifnet *ifp = m->m_pkthdr.rcvif;
1633 register struct ip *ip = mtod(m, struct ip *);
1634 register int hlen = ip->ip_hl << 2;
1635 register int s;
1636 register struct ifqueue *ifq;
1637 register struct vif *vifp;
1638
1639 if (!have_encap_tunnel) {
1640 rip_input(m);
1641 return;
1642 }
1643 /*
1644 * dump the packet if it's not to a multicast destination or if
1645 * we don't have an encapsulating tunnel with the source.
1646 * Note: This code assumes that the remote site IP address
1647 * uniquely identifies the tunnel (i.e., that this site has
1648 * at most one tunnel with the remote site).
1649 */
1650 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) {
1651 ++mrtstat.mrts_bad_tunnel;
1652 m_freem(m);
1653 return;
1654 }
1655 if (ip->ip_src.s_addr != last_encap_src) {
1656 register struct vif *vife;
1657
1658 vifp = viftable;
1659 vife = vifp + numvifs;
1660 last_encap_src = ip->ip_src.s_addr;
1661 last_encap_vif = 0;
1662 for ( ; vifp < vife; ++vifp)
1663 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
1664 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT))
1665 == VIFF_TUNNEL)
1666 last_encap_vif = vifp;
1667 break;
1668 }
1669 }
1670 if ((vifp = last_encap_vif) == 0) {
1671 last_encap_src = 0;
1672 mrtstat.mrts_cant_tunnel++; /*XXX*/
1673 m_freem(m);
1674 if (mrtdebug)
1675 log(LOG_DEBUG, "ip_mforward: no tunnel with %x",
1676 ntohl(ip->ip_src.s_addr));
1677 return;
1678 }
1679 ifp = vifp->v_ifp;
1680
1681 if (hlen > IP_HDR_LEN)
1682 ip_stripoptions(m, (struct mbuf *) 0);
1683 m->m_data += IP_HDR_LEN;
1684 m->m_len -= IP_HDR_LEN;
1685 m->m_pkthdr.len -= IP_HDR_LEN;
1686 m->m_pkthdr.rcvif = ifp;
1687
1688 ifq = &ipintrq;
1689 s = splimp();
1690 if (IF_QFULL(ifq)) {
1691 IF_DROP(ifq);
1692 m_freem(m);
1693 } else {
1694 IF_ENQUEUE(ifq, m);
1695 /*
1696 * normally we would need a "schednetisr(NETISR_IP)"
1697 * here but we were called by ip_input and it is going
1698 * to loop back & try to dequeue the packet we just
1699 * queued as soon as we return so we avoid the
1700 * unnecessary software interrrupt.
1701 */
1702 }
1703 splx(s);
1704}
1705
1706/*
1707 * Token bucket filter module
1708 */
1709static void
1710tbf_control(vifp, m, ip, p_len, imo)
1711 register struct vif *vifp;
1712 register struct mbuf *m;
1713 register struct ip *ip;
1714 register u_long p_len;
1715 struct ip_moptions *imo;
1716{
1717 tbf_update_tokens(vifp);
1718
1719 /* if there are enough tokens,
1720 * and the queue is empty,
1721 * send this packet out
1722 */
1723
1724 if (vifp->v_tbf->q_len == 0) {
1725 if (p_len <= vifp->v_tbf->n_tok) {
1726 vifp->v_tbf->n_tok -= p_len;
1727 tbf_send_packet(vifp, m, imo);
1728 } else if (p_len > MAX_BKT_SIZE) {
1729 /* drop if packet is too large */
1730 mrtstat.mrts_pkt2large++;
1731 m_freem(m);
1732 return;
1733 } else {
1734 /* queue packet and timeout till later */
1735 tbf_queue(vifp, m, ip, imo);
1736 timeout(tbf_reprocess_q, (caddr_t)vifp, 1);
1737 }
1738 } else if (vifp->v_tbf->q_len < MAXQSIZE) {
1739 /* finite queue length, so queue pkts and process queue */
1740 tbf_queue(vifp, m, ip, imo);
1741 tbf_process_q(vifp);
1742 } else {
1743 /* queue length too much, try to dq and queue and process */
1744 if (!tbf_dq_sel(vifp, ip)) {
1745 mrtstat.mrts_q_overflow++;
1746 m_freem(m);
1747 return;
1748 } else {
1749 tbf_queue(vifp, m, ip, imo);
1750 tbf_process_q(vifp);
1751 }
1752 }
1753 return;
1754}
1755
1756/*
1757 * adds a packet to the queue at the interface
1758 */
1759static void
1760tbf_queue(vifp, m, ip, imo)
1761 register struct vif *vifp;
1762 register struct mbuf *m;
1763 register struct ip *ip;
1764 struct ip_moptions *imo;
1765{
1766 register u_long ql;
1767 register int index = (vifp - viftable);
1768 register int s = splnet();
1769
1770 ql = vifp->v_tbf->q_len;
1771
1772 qtable[index][ql].pkt_m = m;
1773 qtable[index][ql].pkt_len = (mtod(m, struct ip *))->ip_len;
1774 qtable[index][ql].pkt_ip = ip;
1775 qtable[index][ql].pkt_imo = imo;
1776
1777 vifp->v_tbf->q_len++;
1778 splx(s);
1779}
1780
1781
1782/*
1783 * processes the queue at the interface
1784 */
1785static void
1786tbf_process_q(vifp)
1787 register struct vif *vifp;
1788{
1789 register struct pkt_queue pkt_1;
1790 register int index = (vifp - viftable);
1791 register int s = splnet();
1792
1793 /* loop through the queue at the interface and send as many packets
1794 * as possible
1795 */
1796 while (vifp->v_tbf->q_len > 0) {
1797 /* locate the first packet */
1798 pkt_1.pkt_len = (qtable[index][0]).pkt_len;
1799 pkt_1.pkt_m = (qtable[index][0]).pkt_m;
1800 pkt_1.pkt_ip = (qtable[index][0]).pkt_ip;
1801 pkt_1.pkt_imo = (qtable[index][0]).pkt_imo;
1802
1803 /* determine if the packet can be sent */
1804 if (pkt_1.pkt_len <= vifp->v_tbf->n_tok) {
1805 /* if so,
1806 * reduce no of tokens, dequeue the queue,
1807 * send the packet.
1808 */
1809 vifp->v_tbf->n_tok -= pkt_1.pkt_len;
1810
1811 tbf_dequeue(vifp, 0);
1812
1813 tbf_send_packet(vifp, pkt_1.pkt_m, pkt_1.pkt_imo);
1814
1815 } else break;
1816 }
1817 splx(s);
1818}
1819
1820/*
1821 * removes the jth packet from the queue at the interface
1822 */
1823static void
1824tbf_dequeue(vifp,j)
1825 register struct vif *vifp;
1826 register int j;
1827{
1828 register u_long index = vifp - viftable;
1829 register int i;
1830
1831 for (i=j+1; i <= vifp->v_tbf->q_len - 1; i++) {
1832 qtable[index][i-1].pkt_m = qtable[index][i].pkt_m;
1833 qtable[index][i-1].pkt_len = qtable[index][i].pkt_len;
1834 qtable[index][i-1].pkt_ip = qtable[index][i].pkt_ip;
1835 qtable[index][i-1].pkt_imo = qtable[index][i].pkt_imo;
1836 }
1837 qtable[index][i-1].pkt_m = NULL;
1838 qtable[index][i-1].pkt_len = NULL;
1839 qtable[index][i-1].pkt_ip = NULL;
1840 qtable[index][i-1].pkt_imo = NULL;
1841
1842 vifp->v_tbf->q_len--;
1843
1844 if (tbfdebug > 1)
1845 log(LOG_DEBUG, "tbf_dequeue: vif# %d qlen %d",vifp-viftable, i-1);
1846}
1847
1848static void
1849tbf_reprocess_q(xvifp)
1850 void *xvifp;
1851{
1852 register struct vif *vifp = xvifp;
1853 if (ip_mrouter == NULL)
1854 return;
1855
1856 tbf_update_tokens(vifp);
1857
1858 tbf_process_q(vifp);
1859
1860 if (vifp->v_tbf->q_len)
1861 timeout(tbf_reprocess_q, (caddr_t)vifp, 1);
1862}
1863
1864/* function that will selectively discard a member of the queue
1865 * based on the precedence value and the priority obtained through
1866 * a lookup table - not yet implemented accurately!
1867 */
1868static int
1869tbf_dq_sel(vifp, ip)
1870 register struct vif *vifp;
1871 register struct ip *ip;
1872{
1873 register int i;
1874 register int s = splnet();
1875 register u_int p;
1876
1877 p = priority(vifp, ip);
1878
1879 for(i=vifp->v_tbf->q_len-1;i >= 0;i--) {
1880 if (p > priority(vifp, qtable[vifp-viftable][i].pkt_ip)) {
1881 m_freem(qtable[vifp-viftable][i].pkt_m);
1882 tbf_dequeue(vifp,i);
1883 splx(s);
1884 mrtstat.mrts_drop_sel++;
1885 return(1);
1886 }
1887 }
1888 splx(s);
1889 return(0);
1890}
1891
1892static void
1893tbf_send_packet(vifp, m, imo)
1894 register struct vif *vifp;
1895 register struct mbuf *m;
1896 struct ip_moptions *imo;
1897{
1898 int error;
1899 int s = splnet();
1900
1901 if (vifp->v_flags & VIFF_TUNNEL) {
1902 /* If tunnel options */
1903 ip_output(m, (struct mbuf *)0, (struct route *)0,
1904 IP_FORWARDING, imo);
1905 } else {
1906 /* if physical interface option, extract the options and then send */
1907 error = ip_output(m, (struct mbuf *)0, (struct route *)0,
1908 IP_FORWARDING, imo);
1909 FREE(imo, M_IPMOPTS);
1910
1911 if (mrtdebug & DEBUG_XMIT)
1912 log(LOG_DEBUG, "phyint_send on vif %d err %d", vifp-viftable, error);
1913 }
1914 splx(s);
1915}
1916
1917/* determine the current time and then
1918 * the elapsed time (between the last time and time now)
1919 * in milliseconds & update the no. of tokens in the bucket
1920 */
1921static void
1922tbf_update_tokens(vifp)
1923 register struct vif *vifp;
1924{
1925 struct timeval tp;
1926 register u_long t;
1927 register u_long elapsed;
1928 register int s = splnet();
1929
1930 GET_TIME(tp);
1931
1932 t = tp.tv_sec*1000 + tp.tv_usec/1000;
1933
1934 elapsed = (t - vifp->v_tbf->last_pkt_t) * vifp->v_rate_limit /8;
1935 vifp->v_tbf->n_tok += elapsed;
1936 vifp->v_tbf->last_pkt_t = t;
1937
1938 if (vifp->v_tbf->n_tok > MAX_BKT_SIZE)
1939 vifp->v_tbf->n_tok = MAX_BKT_SIZE;
1940
1941 splx(s);
1942}
1943
1944static int
1945priority(vifp, ip)
1946 register struct vif *vifp;
1947 register struct ip *ip;
1948{
1949 register int prio;
1950
1951 /* temporary hack; may add general packet classifier some day */
1952
1953 /*
1954 * The UDP port space is divided up into four priority ranges:
1955 * [0, 16384) : unclassified - lowest priority
1956 * [16384, 32768) : audio - highest priority
1957 * [32768, 49152) : whiteboard - medium priority
1958 * [49152, 65536) : video - low priority
1959 */
1960 if (ip->ip_p == IPPROTO_UDP) {
1961 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
1962 switch (ntohs(udp->uh_dport) & 0xc000) {
1963 case 0x4000:
1964 prio = 70;
1965 break;
1966 case 0x8000:
1967 prio = 60;
1968 break;
1969 case 0xc000:
1970 prio = 55;
1971 break;
1972 default:
1973 prio = 50;
1974 break;
1975 }
1976 if (tbfdebug > 1)
1977 log(LOG_DEBUG, "port %x prio%d", ntohs(udp->uh_dport), prio);
1978 } else {
1979 prio = 50;
1980 }
1981 return prio;
1982}
1983
1984/*
1985 * End of token bucket filter modifications
1986 */
1987
1988int
1989ip_rsvp_vif_init(so, m)
1990 struct socket *so;
1991 struct mbuf *m;
1992{
1993 int i;
1994 register int s;
1995
1996 if (rsvpdebug)
1997 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
1998 so->so_type, so->so_proto->pr_protocol);
1999
2000 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2001 return EOPNOTSUPP;
2002
2003 /* Check mbuf. */
2004 if (m == NULL || m->m_len != sizeof(int)) {
2005 return EINVAL;
2006 }
2007 i = *(mtod(m, int *));
2008
2009 if (rsvpdebug)
2010 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on);
2011
2012 s = splnet();
2013
2014 /* Check vif. */
2015 if (!legal_vif_num(i)) {
2016 splx(s);
2017 return EADDRNOTAVAIL;
2018 }
2019
2020 /* Check if socket is available. */
2021 if (viftable[i].v_rsvpd != NULL) {
2022 splx(s);
2023 return EADDRINUSE;
2024 }
2025
2026 viftable[i].v_rsvpd = so;
2027 /* This may seem silly, but we need to be sure we don't over-increment
2028 * the RSVP counter, in case something slips up.
2029 */
2030 if (!viftable[i].v_rsvp_on) {
2031 viftable[i].v_rsvp_on = 1;
2032 rsvp_on++;
2033 }
2034
2035 splx(s);
2036 return 0;
2037}
2038
2039int
2040ip_rsvp_vif_done(so, m)
2041 struct socket *so;
2042 struct mbuf *m;
2043{
2044 int i;
2045 register int s;
2046
2047 if (rsvpdebug)
2048 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2049 so->so_type, so->so_proto->pr_protocol);
2050
2051 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2052 return EOPNOTSUPP;
2053
2054 /* Check mbuf. */
2055 if (m == NULL || m->m_len != sizeof(int)) {
2056 return EINVAL;
2057 }
2058 i = *(mtod(m, int *));
2059
2060 s = splnet();
2061
2062 /* Check vif. */
2063 if (!legal_vif_num(i)) {
2064 splx(s);
2065 return EADDRNOTAVAIL;
2066 }
2067
2068 if (rsvpdebug)
2069 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2070 viftable[i].v_rsvpd, so);
2071
2072 viftable[i].v_rsvpd = NULL;
2073 /* This may seem silly, but we need to be sure we don't over-decrement
2074 * the RSVP counter, in case something slips up.
2075 */
2076 if (viftable[i].v_rsvp_on) {
2077 viftable[i].v_rsvp_on = 0;
2078 rsvp_on--;
2079 }
2080
2081 splx(s);
2082 return 0;
2083}
2084
2085void
2086ip_rsvp_force_done(so)
2087 struct socket *so;
2088{
2089 int vifi;
2090 register int s;
2091
2092 /* Don't bother if it is not the right type of socket. */
2093 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2094 return;
2095
2096 s = splnet();
2097
2098 /* The socket may be attached to more than one vif...this
2099 * is perfectly legal.
2100 */
2101 for (vifi = 0; vifi < numvifs; vifi++) {
2102 if (viftable[vifi].v_rsvpd == so) {
2103 viftable[vifi].v_rsvpd = NULL;
2104 /* This may seem silly, but we need to be sure we don't
2105 * over-decrement the RSVP counter, in case something slips up.
2106 */
2107 if (viftable[vifi].v_rsvp_on) {
2108 viftable[vifi].v_rsvp_on = 0;
2109 rsvp_on--;
2110 }
2111 }
2112 }
2113
2114 splx(s);
2115 return;
2116}
2117
2118void
| 13 */
14
15
16#include <sys/param.h>
17#include <sys/systm.h>
18#include <sys/mbuf.h>
19#include <sys/socket.h>
20#include <sys/socketvar.h>
21#include <sys/protosw.h>
22#include <sys/errno.h>
23#include <sys/time.h>
24#include <sys/kernel.h>
25#include <sys/ioctl.h>
26#include <sys/syslog.h>
27#include <sys/queue.h>
28#include <net/if.h>
29#include <net/route.h>
30#include <netinet/in.h>
31#include <netinet/in_systm.h>
32#include <netinet/ip.h>
33#include <netinet/ip_var.h>
34#include <netinet/in_pcb.h>
35#include <netinet/in_var.h>
36#include <netinet/igmp.h>
37#include <netinet/igmp_var.h>
38#include <netinet/ip_mroute.h>
39#include <netinet/udp.h>
40
41#ifndef NTOHL
42#if BYTE_ORDER != BIG_ENDIAN
43#define NTOHL(d) ((d) = ntohl((d)))
44#define NTOHS(d) ((d) = ntohs((u_short)(d)))
45#define HTONL(d) ((d) = htonl((d)))
46#define HTONS(d) ((d) = htons((u_short)(d)))
47#else
48#define NTOHL(d)
49#define NTOHS(d)
50#define HTONL(d)
51#define HTONS(d)
52#endif
53#endif
54
55extern int rsvp_on;
56
57#ifndef MROUTING
58/*
59 * Dummy routines and globals used when multicast routing is not compiled in.
60 */
61
62struct socket *ip_mrouter = NULL;
63u_int ip_mrtproto = 0;
64struct mrtstat mrtstat;
65u_int rsvpdebug = 0;
66
67int
68_ip_mrouter_set(cmd, so, m)
69 int cmd;
70 struct socket *so;
71 struct mbuf *m;
72{
73 return(EOPNOTSUPP);
74}
75
76int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = _ip_mrouter_set;
77
78
79int
80_ip_mrouter_get(cmd, so, m)
81 int cmd;
82 struct socket *so;
83 struct mbuf **m;
84{
85 return(EOPNOTSUPP);
86}
87
88int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = _ip_mrouter_get;
89
90int
91_ip_mrouter_done()
92{
93 return(0);
94}
95
96int (*ip_mrouter_done)(void) = _ip_mrouter_done;
97
98int
99_ip_mforward(ip, ifp, m, imo)
100 struct ip *ip;
101 struct ifnet *ifp;
102 struct mbuf *m;
103 struct ip_moptions *imo;
104{
105 return(0);
106}
107
108int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
109 struct ip_moptions *) = _ip_mforward;
110
111int
112_mrt_ioctl(int req, caddr_t data, struct proc *p)
113{
114 return EOPNOTSUPP;
115}
116
117int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl;
118
119void
120rsvp_input(m, iphlen) /* XXX must fixup manually */
121 struct mbuf *m;
122 int iphlen;
123{
124 /* Can still get packets with rsvp_on = 0 if there is a local member
125 * of the group to which the RSVP packet is addressed. But in this
126 * case we want to throw the packet away.
127 */
128 if (!rsvp_on) {
129 m_freem(m);
130 return;
131 }
132
133 if (ip_rsvpd != NULL) {
134 if (rsvpdebug)
135 printf("rsvp_input: Sending packet up old-style socket\n");
136 rip_input(m);
137 return;
138 }
139 /* Drop the packet */
140 m_freem(m);
141}
142
143void ipip_input(struct mbuf *m) { /* XXX must fixup manually */
144 rip_input(m);
145}
146
147int (*legal_vif_num)(int) = 0;
148
149/*
150 * This should never be called, since IP_MULTICAST_VIF should fail, but
151 * just in case it does get called, the code a little lower in ip_output
152 * will assign the packet a local address.
153 */
154u_long
155_ip_mcast_src(int vifi) { return INADDR_ANY; }
156u_long (*ip_mcast_src)(int) = _ip_mcast_src;
157
158int
159ip_rsvp_vif_init(so, m)
160 struct socket *so;
161 struct mbuf *m;
162{
163 return(EINVAL);
164}
165
166int
167ip_rsvp_vif_done(so, m)
168 struct socket *so;
169 struct mbuf *m;
170{
171 return(EINVAL);
172}
173
174void
175ip_rsvp_force_done(so)
176 struct socket *so;
177{
178 return;
179}
180
181#else /* MROUTING */
182
183#define M_HASCL(m) ((m)->m_flags & M_EXT)
184
185#define INSIZ sizeof(struct in_addr)
186#define same(a1, a2) \
187 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0)
188
189#define MT_MRTABLE MT_RTABLE /* since nothing else uses it */
190
191/*
192 * Globals. All but ip_mrouter and ip_mrtproto could be static,
193 * except for netstat or debugging purposes.
194 */
195#ifndef MROUTE_LKM
196struct socket *ip_mrouter = NULL;
197struct mrtstat mrtstat;
198
199int ip_mrtproto = IGMP_DVMRP; /* for netstat only */
200#else /* MROUTE_LKM */
201extern struct mrtstat mrtstat;
202extern int ip_mrtproto;
203#endif
204
205#define NO_RTE_FOUND 0x1
206#define RTE_FOUND 0x2
207
208struct mbuf *mfctable[MFCTBLSIZ];
209u_char nexpire[MFCTBLSIZ];
210struct vif viftable[MAXVIFS];
211u_int mrtdebug = 0; /* debug level */
212#define DEBUG_MFC 0x02
213#define DEBUG_FORWARD 0x04
214#define DEBUG_EXPIRE 0x08
215#define DEBUG_XMIT 0x10
216u_int tbfdebug = 0; /* tbf debug level */
217u_int rsvpdebug = 0; /* rsvp debug level */
218
219#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
220#define UPCALL_EXPIRE 6 /* number of timeouts */
221
222/*
223 * Define the token bucket filter structures
224 * tbftable -> each vif has one of these for storing info
225 * qtable -> each interface has an associated queue of pkts
226 */
227
228struct tbf tbftable[MAXVIFS];
229struct pkt_queue qtable[MAXVIFS][MAXQSIZE];
230
231/*
232 * 'Interfaces' associated with decapsulator (so we can tell
233 * packets that went through it from ones that get reflected
234 * by a broken gateway). These interfaces are never linked into
235 * the system ifnet list & no routes point to them. I.e., packets
236 * can't be sent this way. They only exist as a placeholder for
237 * multicast source verification.
238 */
239struct ifnet multicast_decap_if[MAXVIFS];
240
241#define ENCAP_TTL 64
242#define ENCAP_PROTO IPPROTO_IPIP /* 4 */
243
244/* prototype IP hdr for encapsulated packets */
245struct ip multicast_encap_iphdr = {
246#if BYTE_ORDER == LITTLE_ENDIAN
247 sizeof(struct ip) >> 2, IPVERSION,
248#else
249 IPVERSION, sizeof(struct ip) >> 2,
250#endif
251 0, /* tos */
252 sizeof(struct ip), /* total length */
253 0, /* id */
254 0, /* frag offset */
255 ENCAP_TTL, ENCAP_PROTO,
256 0, /* checksum */
257};
258
259/*
260 * Private variables.
261 */
262static vifi_t numvifs = 0;
263static void (*encap_oldrawip)() = 0;
264static int have_encap_tunnel = 0;
265
266/*
267 * one-back cache used by ipip_input to locate a tunnel's vif
268 * given a datagram's src ip address.
269 */
270static u_long last_encap_src;
271static struct vif *last_encap_vif;
272
273static int get_sg_cnt(struct sioc_sg_req *);
274static int get_vif_cnt(struct sioc_vif_req *);
275int ip_mrouter_init(struct socket *, struct mbuf *);
276static int add_vif(struct vifctl *);
277static int del_vif(vifi_t *);
278static int add_mfc(struct mfcctl *);
279static int del_mfc(struct mfcctl *);
280static int get_version(struct mbuf *);
281static int get_assert(struct mbuf *);
282static int set_assert(int *);
283static void expire_upcalls(void *);
284static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *,
285 vifi_t);
286static void phyint_send(struct ip *, struct vif *, struct mbuf *);
287static void encap_send(struct ip *, struct vif *, struct mbuf *);
288static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long,
289 struct ip_moptions *);
290static void tbf_queue(struct vif *, struct mbuf *, struct ip *, struct ip_moptions *);
291static void tbf_process_q(struct vif *);
292static void tbf_dequeue(struct vif *, int);
293static void tbf_reprocess_q(void *);
294static int tbf_dq_sel(struct vif *, struct ip *);
295static void tbf_send_packet(struct vif *, struct mbuf *, struct ip_moptions *);
296static void tbf_update_tokens(struct vif *);
297static int priority(struct vif *, struct ip *);
298void multiencap_decap(struct mbuf *);
299
300/*
301 * whether or not special PIM assert processing is enabled.
302 */
303static int pim_assert;
304/*
305 * Rate limit for assert notification messages, in usec
306 */
307#define ASSERT_MSG_TIME 3000000
308
309/*
310 * Hash function for a source, group entry
311 */
312#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
313 ((g) >> 20) ^ ((g) >> 10) ^ (g))
314
315/*
316 * Find a route for a given origin IP address and Multicast group address
317 * Type of service parameter to be added in the future!!!
318 */
319
320#define MFCFIND(o, g, rt) { \
321 register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \
322 register struct mfc *_rt = NULL; \
323 rt = NULL; \
324 ++mrtstat.mrts_mfc_lookups; \
325 while (_mb_rt) { \
326 _rt = mtod(_mb_rt, struct mfc *); \
327 if ((_rt->mfc_origin.s_addr == o) && \
328 (_rt->mfc_mcastgrp.s_addr == g) && \
329 (_mb_rt->m_act == NULL)) { \
330 rt = _rt; \
331 break; \
332 } \
333 _mb_rt = _mb_rt->m_next; \
334 } \
335 if (rt == NULL) { \
336 ++mrtstat.mrts_mfc_misses; \
337 } \
338}
339
340
341/*
342 * Macros to compute elapsed time efficiently
343 * Borrowed from Van Jacobson's scheduling code
344 */
345#define TV_DELTA(a, b, delta) { \
346 register int xxs; \
347 \
348 delta = (a).tv_usec - (b).tv_usec; \
349 if ((xxs = (a).tv_sec - (b).tv_sec)) { \
350 switch (xxs) { \
351 case 2: \
352 delta += 1000000; \
353 /* fall through */ \
354 case 1: \
355 delta += 1000000; \
356 break; \
357 default: \
358 delta += (1000000 * xxs); \
359 } \
360 } \
361}
362
363#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
364 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
365
366#ifdef UPCALL_TIMING
367u_long upcall_data[51];
368static void collate(struct timeval *);
369#endif /* UPCALL_TIMING */
370
371
372/*
373 * Handle MRT setsockopt commands to modify the multicast routing tables.
374 */
375int
376X_ip_mrouter_set(cmd, so, m)
377 int cmd;
378 struct socket *so;
379 struct mbuf *m;
380{
381 if (cmd != MRT_INIT && so != ip_mrouter) return EACCES;
382
383 switch (cmd) {
384 case MRT_INIT: return ip_mrouter_init(so, m);
385 case MRT_DONE: return ip_mrouter_done();
386 case MRT_ADD_VIF: return add_vif (mtod(m, struct vifctl *));
387 case MRT_DEL_VIF: return del_vif (mtod(m, vifi_t *));
388 case MRT_ADD_MFC: return add_mfc (mtod(m, struct mfcctl *));
389 case MRT_DEL_MFC: return del_mfc (mtod(m, struct mfcctl *));
390 case MRT_ASSERT: return set_assert(mtod(m, int *));
391 default: return EOPNOTSUPP;
392 }
393}
394
395#ifndef MROUTE_LKM
396int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = X_ip_mrouter_set;
397#endif
398
399/*
400 * Handle MRT getsockopt commands
401 */
402int
403X_ip_mrouter_get(cmd, so, m)
404 int cmd;
405 struct socket *so;
406 struct mbuf **m;
407{
408 struct mbuf *mb;
409
410 if (so != ip_mrouter) return EACCES;
411
412 *m = mb = m_get(M_WAIT, MT_SOOPTS);
413
414 switch (cmd) {
415 case MRT_VERSION: return get_version(mb);
416 case MRT_ASSERT: return get_assert(mb);
417 default: return EOPNOTSUPP;
418 }
419}
420
421#ifndef MROUTE_LKM
422int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = X_ip_mrouter_get;
423#endif
424
425/*
426 * Handle ioctl commands to obtain information from the cache
427 */
428int
429X_mrt_ioctl(cmd, data)
430 int cmd;
431 caddr_t data;
432{
433 int error = 0;
434
435 switch (cmd) {
436 case (SIOCGETVIFCNT):
437 return (get_vif_cnt((struct sioc_vif_req *)data));
438 break;
439 case (SIOCGETSGCNT):
440 return (get_sg_cnt((struct sioc_sg_req *)data));
441 break;
442 default:
443 return (EINVAL);
444 break;
445 }
446 return error;
447}
448
449#ifndef MROUTE_LKM
450int (*mrt_ioctl)(int, caddr_t, struct proc *) = X_mrt_ioctl;
451#endif
452
453/*
454 * returns the packet, byte, rpf-failure count for the source group provided
455 */
456static int
457get_sg_cnt(req)
458 register struct sioc_sg_req *req;
459{
460 register struct mfc *rt;
461 int s;
462
463 s = splnet();
464 MFCFIND(req->src.s_addr, req->grp.s_addr, rt);
465 splx(s);
466 if (rt != NULL) {
467 req->pktcnt = rt->mfc_pkt_cnt;
468 req->bytecnt = rt->mfc_byte_cnt;
469 req->wrong_if = rt->mfc_wrong_if;
470 } else
471 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
472
473 return 0;
474}
475
476/*
477 * returns the input and output packet and byte counts on the vif provided
478 */
479static int
480get_vif_cnt(req)
481 register struct sioc_vif_req *req;
482{
483 register vifi_t vifi = req->vifi;
484
485 if (vifi >= numvifs) return EINVAL;
486
487 req->icount = viftable[vifi].v_pkt_in;
488 req->ocount = viftable[vifi].v_pkt_out;
489 req->ibytes = viftable[vifi].v_bytes_in;
490 req->obytes = viftable[vifi].v_bytes_out;
491
492 return 0;
493}
494
495/*
496 * Enable multicast routing
497 */
498int
499ip_mrouter_init(so, m)
500 struct socket *so;
501 struct mbuf *m;
502{
503 int *v;
504 int i;
505
506 if (mrtdebug)
507 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d",
508 so->so_type, so->so_proto->pr_protocol);
509
510 if (so->so_type != SOCK_RAW ||
511 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP;
512
513 if (!m || (m->m_len != sizeof(int *)))
514 return ENOPROTOOPT;
515
516 v = mtod(m, int *);
517 if (*v != 1)
518 return ENOPROTOOPT;
519
520 if (ip_mrouter != NULL) return EADDRINUSE;
521
522 ip_mrouter = so;
523
524 bzero((caddr_t)mfctable, sizeof(mfctable));
525 bzero((caddr_t)nexpire, sizeof(nexpire));
526
527 pim_assert = 0;
528
529 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
530
531 if (mrtdebug)
532 log(LOG_DEBUG, "ip_mrouter_init");
533
534 return 0;
535}
536
537/*
538 * Disable multicast routing
539 */
540int
541X_ip_mrouter_done()
542{
543 vifi_t vifi;
544 int i;
545 struct ifnet *ifp;
546 struct ifreq ifr;
547 struct mbuf *mb_rt;
548 struct mfc *rt;
549 struct mbuf *m;
550 struct rtdetq *rte;
551 int s;
552
553 s = splnet();
554
555 /*
556 * For each phyint in use, disable promiscuous reception of all IP
557 * multicasts.
558 */
559 for (vifi = 0; vifi < numvifs; vifi++) {
560 if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
561 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
562 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
563 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr
564 = INADDR_ANY;
565 ifp = viftable[vifi].v_ifp;
566 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
567 }
568 }
569 bzero((caddr_t)qtable, sizeof(qtable));
570 bzero((caddr_t)tbftable, sizeof(tbftable));
571 bzero((caddr_t)viftable, sizeof(viftable));
572 numvifs = 0;
573 pim_assert = 0;
574
575 untimeout(expire_upcalls, (caddr_t)NULL);
576
577 /*
578 * Free all multicast forwarding cache entries.
579 */
580 for (i = 0; i < MFCTBLSIZ; i++) {
581 mb_rt = mfctable[i];
582 while (mb_rt) {
583 if (mb_rt->m_act != NULL) {
584 while (mb_rt->m_act) {
585 m = mb_rt->m_act;
586 mb_rt->m_act = m->m_act;
587 rte = mtod(m, struct rtdetq *);
588 m_freem(rte->m);
589 m_free(m);
590 }
591 }
592 mb_rt = m_free(mb_rt);
593 }
594 }
595
596 bzero((caddr_t)mfctable, sizeof(mfctable));
597
598 /*
599 * Reset de-encapsulation cache
600 */
601 last_encap_src = NULL;
602 last_encap_vif = NULL;
603 have_encap_tunnel = 0;
604
605 ip_mrouter = NULL;
606
607 splx(s);
608
609 if (mrtdebug)
610 log(LOG_DEBUG, "ip_mrouter_done");
611
612 return 0;
613}
614
615#ifndef MROUTE_LKM
616int (*ip_mrouter_done)(void) = X_ip_mrouter_done;
617#endif
618
619static int
620get_version(mb)
621 struct mbuf *mb;
622{
623 int *v;
624
625 v = mtod(mb, int *);
626
627 *v = 0x0305; /* XXX !!!! */
628 mb->m_len = sizeof(int);
629
630 return 0;
631}
632
633/*
634 * Set PIM assert processing global
635 */
636static int
637set_assert(i)
638 int *i;
639{
640 if ((*i != 1) && (*i != 0))
641 return EINVAL;
642
643 pim_assert = *i;
644
645 return 0;
646}
647
648/*
649 * Get PIM assert processing global
650 */
651static int
652get_assert(m)
653 struct mbuf *m;
654{
655 int *i;
656
657 i = mtod(m, int *);
658
659 *i = pim_assert;
660
661 return 0;
662}
663
664/*
665 * Add a vif to the vif table
666 */
667static int
668add_vif(vifcp)
669 register struct vifctl *vifcp;
670{
671 register struct vif *vifp = viftable + vifcp->vifc_vifi;
672 static struct sockaddr_in sin = {sizeof sin, AF_INET};
673 struct ifaddr *ifa;
674 struct ifnet *ifp;
675 struct ifreq ifr;
676 int error, s;
677 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
678
679 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL;
680 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE;
681
682 /* Find the interface with an address in AF_INET family */
683 sin.sin_addr = vifcp->vifc_lcl_addr;
684 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
685 if (ifa == 0) return EADDRNOTAVAIL;
686 ifp = ifa->ifa_ifp;
687
688 if (vifcp->vifc_flags & VIFF_TUNNEL) {
689 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
690 /*
691 * An encapsulating tunnel is wanted. Tell ipip_input() to
692 * start paying attention to encapsulated packets.
693 */
694 if (have_encap_tunnel == 0) {
695 have_encap_tunnel = 1;
696 for (s = 0; s < MAXVIFS; ++s) {
697 multicast_decap_if[s].if_name = "mdecap";
698 multicast_decap_if[s].if_unit = s;
699 }
700 }
701 /*
702 * Set interface to fake encapsulator interface
703 */
704 ifp = &multicast_decap_if[vifcp->vifc_vifi];
705 /*
706 * Prepare cached route entry
707 */
708 bzero(&vifp->v_route, sizeof(vifp->v_route));
709 } else {
710 log(LOG_ERR, "Source routed tunnels not supported.");
711 return EOPNOTSUPP;
712 }
713 } else {
714 /* Make sure the interface supports multicast */
715 if ((ifp->if_flags & IFF_MULTICAST) == 0)
716 return EOPNOTSUPP;
717
718 /* Enable promiscuous reception of all IP multicasts from the if */
719 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
720 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
721 s = splnet();
722 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
723 splx(s);
724 if (error)
725 return error;
726 }
727
728 s = splnet();
729 /* define parameters for the tbf structure */
730 vifp->v_tbf = v_tbf;
731 vifp->v_tbf->q_len = 0;
732 vifp->v_tbf->n_tok = 0;
733 vifp->v_tbf->last_pkt_t = 0;
734
735 vifp->v_flags = vifcp->vifc_flags;
736 vifp->v_threshold = vifcp->vifc_threshold;
737 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
738 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
739 vifp->v_ifp = ifp;
740 vifp->v_rate_limit= vifcp->vifc_rate_limit;
741 vifp->v_rsvp_on = 0;
742 vifp->v_rsvpd = NULL;
743 /* initialize per vif pkt counters */
744 vifp->v_pkt_in = 0;
745 vifp->v_pkt_out = 0;
746 vifp->v_bytes_in = 0;
747 vifp->v_bytes_out = 0;
748 splx(s);
749
750 /* Adjust numvifs up if the vifi is higher than numvifs */
751 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
752
753 if (mrtdebug)
754 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d",
755 vifcp->vifc_vifi,
756 ntohl(vifcp->vifc_lcl_addr.s_addr),
757 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
758 ntohl(vifcp->vifc_rmt_addr.s_addr),
759 vifcp->vifc_threshold,
760 vifcp->vifc_rate_limit);
761
762 return 0;
763}
764
765/*
766 * Delete a vif from the vif table
767 */
768static int
769del_vif(vifip)
770 vifi_t *vifip;
771{
772 register struct vif *vifp = viftable + *vifip;
773 register vifi_t vifi;
774 struct ifnet *ifp;
775 struct ifreq ifr;
776 int s;
777
778 if (*vifip >= numvifs) return EINVAL;
779 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL;
780
781 s = splnet();
782
783 if (!(vifp->v_flags & VIFF_TUNNEL)) {
784 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
785 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
786 ifp = vifp->v_ifp;
787 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
788 }
789
790 if (vifp == last_encap_vif) {
791 last_encap_vif = 0;
792 last_encap_src = 0;
793 }
794
795 bzero((caddr_t)qtable[*vifip],
796 sizeof(qtable[*vifip]));
797 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
798 bzero((caddr_t)vifp, sizeof (*vifp));
799
800 /* Adjust numvifs down */
801 for (vifi = numvifs; vifi > 0; vifi--)
802 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break;
803 numvifs = vifi;
804
805 splx(s);
806
807 if (mrtdebug)
808 log(LOG_DEBUG, "del_vif %d, numvifs %d", *vifip, numvifs);
809
810 return 0;
811}
812
813/*
814 * Add an mfc entry
815 */
816static int
817add_mfc(mfccp)
818 struct mfcctl *mfccp;
819{
820 struct mfc *rt;
821 register struct mbuf *mb_rt;
822 u_long hash;
823 struct mbuf *mb_ntry;
824 struct rtdetq *rte;
825 register u_short nstl;
826 int s;
827 int i;
828
829 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt);
830
831 /* If an entry already exists, just update the fields */
832 if (rt) {
833 if (mrtdebug & DEBUG_MFC)
834 log(LOG_DEBUG,"add_mfc update o %x g %x p %x",
835 ntohl(mfccp->mfcc_origin.s_addr),
836 ntohl(mfccp->mfcc_mcastgrp.s_addr),
837 mfccp->mfcc_parent);
838
839 s = splnet();
840 rt->mfc_parent = mfccp->mfcc_parent;
841 for (i = 0; i < numvifs; i++)
842 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
843 splx(s);
844 return 0;
845 }
846
847 /*
848 * Find the entry for which the upcall was made and update
849 */
850 s = splnet();
851 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
852 for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) {
853
854 rt = mtod(mb_rt, struct mfc *);
855 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
856 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
857 (mb_rt->m_act != NULL)) {
858
859 if (nstl++)
860 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %x",
861 "multiple kernel entries",
862 ntohl(mfccp->mfcc_origin.s_addr),
863 ntohl(mfccp->mfcc_mcastgrp.s_addr),
864 mfccp->mfcc_parent, mb_rt->m_act);
865
866 if (mrtdebug & DEBUG_MFC)
867 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %x",
868 ntohl(mfccp->mfcc_origin.s_addr),
869 ntohl(mfccp->mfcc_mcastgrp.s_addr),
870 mfccp->mfcc_parent, mb_rt->m_act);
871
872 rt->mfc_origin = mfccp->mfcc_origin;
873 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
874 rt->mfc_parent = mfccp->mfcc_parent;
875 for (i = 0; i < numvifs; i++)
876 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
877 /* initialize pkt counters per src-grp */
878 rt->mfc_pkt_cnt = 0;
879 rt->mfc_byte_cnt = 0;
880 rt->mfc_wrong_if = 0;
881 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
882
883 rt->mfc_expire = 0; /* Don't clean this guy up */
884 nexpire[hash]--;
885
886 /* free packets Qed at the end of this entry */
887 while (mb_rt->m_act) {
888 mb_ntry = mb_rt->m_act;
889 rte = mtod(mb_ntry, struct rtdetq *);
890/* #ifdef RSVP_ISI */
891 ip_mdq(rte->m, rte->ifp, rt, -1);
892/* #endif */
893 mb_rt->m_act = mb_ntry->m_act;
894 m_freem(rte->m);
895#ifdef UPCALL_TIMING
896 collate(&(rte->t));
897#endif /* UPCALL_TIMING */
898 m_free(mb_ntry);
899 }
900 }
901 }
902
903 /*
904 * It is possible that an entry is being inserted without an upcall
905 */
906 if (nstl == 0) {
907 if (mrtdebug & DEBUG_MFC)
908 log(LOG_DEBUG,"add_mfc no upcall h %d o %x g %x p %x",
909 hash, ntohl(mfccp->mfcc_origin.s_addr),
910 ntohl(mfccp->mfcc_mcastgrp.s_addr),
911 mfccp->mfcc_parent);
912
913 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
914
915 rt = mtod(mb_rt, struct mfc *);
916 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
917 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
918
919 rt->mfc_origin = mfccp->mfcc_origin;
920 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
921 rt->mfc_parent = mfccp->mfcc_parent;
922 for (i = 0; i < numvifs; i++)
923 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
924 /* initialize pkt counters per src-grp */
925 rt->mfc_pkt_cnt = 0;
926 rt->mfc_byte_cnt = 0;
927 rt->mfc_wrong_if = 0;
928 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
929 if (rt->mfc_expire)
930 nexpire[hash]--;
931 rt->mfc_expire = 0;
932 }
933 }
934 if (mb_rt == NULL) {
935 /* no upcall, so make a new entry */
936 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
937 if (mb_rt == NULL) {
938 splx(s);
939 return ENOBUFS;
940 }
941
942 rt = mtod(mb_rt, struct mfc *);
943
944 /* insert new entry at head of hash chain */
945 rt->mfc_origin = mfccp->mfcc_origin;
946 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
947 rt->mfc_parent = mfccp->mfcc_parent;
948 for (i = 0; i < numvifs; i++)
949 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
950 /* initialize pkt counters per src-grp */
951 rt->mfc_pkt_cnt = 0;
952 rt->mfc_byte_cnt = 0;
953 rt->mfc_wrong_if = 0;
954 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
955 rt->mfc_expire = 0;
956
957 /* link into table */
958 mb_rt->m_next = mfctable[hash];
959 mfctable[hash] = mb_rt;
960 mb_rt->m_act = NULL;
961 }
962 }
963 splx(s);
964 return 0;
965}
966
967#ifdef UPCALL_TIMING
968/*
969 * collect delay statistics on the upcalls
970 */
971static void collate(t)
972register struct timeval *t;
973{
974 register u_long d;
975 register struct timeval tp;
976 register u_long delta;
977
978 GET_TIME(tp);
979
980 if (TV_LT(*t, tp))
981 {
982 TV_DELTA(tp, *t, delta);
983
984 d = delta >> 10;
985 if (d > 50)
986 d = 50;
987
988 ++upcall_data[d];
989 }
990}
991#endif /* UPCALL_TIMING */
992
993/*
994 * Delete an mfc entry
995 */
996static int
997del_mfc(mfccp)
998 struct mfcctl *mfccp;
999{
1000 struct in_addr origin;
1001 struct in_addr mcastgrp;
1002 struct mfc *rt;
1003 struct mbuf *mb_rt;
1004 struct mbuf **nptr;
1005 u_long hash;
1006 int s, i;
1007
1008 origin = mfccp->mfcc_origin;
1009 mcastgrp = mfccp->mfcc_mcastgrp;
1010 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1011
1012 if (mrtdebug & DEBUG_MFC)
1013 log(LOG_DEBUG,"del_mfc orig %x mcastgrp %x",
1014 ntohl(origin.s_addr), ntohl(mcastgrp.s_addr));
1015
1016 s = splnet();
1017
1018 nptr = &mfctable[hash];
1019 while ((mb_rt = *nptr) != NULL) {
1020 rt = mtod(mb_rt, struct mfc *);
1021 if (origin.s_addr == rt->mfc_origin.s_addr &&
1022 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1023 mb_rt->m_act == NULL)
1024 break;
1025
1026 nptr = &mb_rt->m_next;
1027 }
1028 if (mb_rt == NULL) {
1029 splx(s);
1030 return EADDRNOTAVAIL;
1031 }
1032
1033 MFREE(mb_rt, *nptr);
1034
1035 splx(s);
1036
1037 return 0;
1038}
1039
1040/*
1041 * Send a message to mrouted on the multicast routing socket
1042 */
1043static int
1044socket_send(s, mm, src)
1045 struct socket *s;
1046 struct mbuf *mm;
1047 struct sockaddr_in *src;
1048{
1049 if (s) {
1050 if (sbappendaddr(&s->so_rcv,
1051 (struct sockaddr *)src,
1052 mm, (struct mbuf *)0) != 0) {
1053 sorwakeup(s);
1054 return 0;
1055 }
1056 }
1057 m_freem(mm);
1058 return -1;
1059}
1060
1061/*
1062 * IP multicast forwarding function. This function assumes that the packet
1063 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1064 * pointed to by "ifp", and the packet is to be relayed to other networks
1065 * that have members of the packet's destination IP multicast group.
1066 *
1067 * The packet is returned unscathed to the caller, unless it is
1068 * erroneous, in which case a non-zero return value tells the caller to
1069 * discard it.
1070 */
1071
1072#define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */
1073#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1074
1075int
1076X_ip_mforward(ip, ifp, m, imo)
1077 register struct ip *ip;
1078 struct ifnet *ifp;
1079 struct mbuf *m;
1080 struct ip_moptions *imo;
1081{
1082 register struct mfc *rt = 0; /* XXX uninit warning */
1083 register u_char *ipoptions;
1084 static struct sockproto k_igmpproto = { AF_INET, IPPROTO_IGMP };
1085 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1086 static int srctun = 0;
1087 register struct mbuf *mm;
1088 int s;
1089 vifi_t vifi;
1090 struct vif *vifp;
1091
1092 if (mrtdebug & DEBUG_FORWARD)
1093 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %x",
1094 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1095
1096 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1097 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1098 /*
1099 * Packet arrived via a physical interface or
1100 * an encapsulated tunnel.
1101 */
1102 } else {
1103 /*
1104 * Packet arrived through a source-route tunnel.
1105 * Source-route tunnels are no longer supported.
1106 */
1107 if ((srctun++ % 1000) == 0)
1108 log(LOG_ERR, "ip_mforward: received source-routed packet from %x",
1109 ntohl(ip->ip_src.s_addr));
1110
1111 return 1;
1112 }
1113
1114 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1115 if (ip->ip_ttl < 255)
1116 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1117 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1118 vifp = viftable + vifi;
1119 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s%d)\n",
1120 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi,
1121 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1122 vifp->v_ifp->if_name, vifp->v_ifp->if_unit);
1123 }
1124 return (ip_mdq(m, ifp, rt, vifi));
1125 }
1126 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1127 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
1128 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr));
1129 }
1130
1131 /*
1132 * Don't forward a packet with time-to-live of zero or one,
1133 * or a packet destined to a local-only group.
1134 */
1135 if (ip->ip_ttl <= 1 ||
1136 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
1137 return 0;
1138
1139 /*
1140 * Determine forwarding vifs from the forwarding cache table
1141 */
1142 s = splnet();
1143 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt);
1144
1145 /* Entry exists, so forward if necessary */
1146 if (rt != NULL) {
1147 splx(s);
1148 return (ip_mdq(m, ifp, rt, -1));
1149 } else {
1150 /*
1151 * If we don't have a route for packet's origin,
1152 * Make a copy of the packet &
1153 * send message to routing daemon
1154 */
1155
1156 register struct mbuf *mb_rt;
1157 register struct mbuf *mb_ntry;
1158 register struct mbuf *mb0;
1159 register struct rtdetq *rte;
1160 register struct mbuf *rte_m;
1161 register u_long hash;
1162 register int npkts;
1163#ifdef UPCALL_TIMING
1164 struct timeval tp;
1165
1166 GET_TIME(tp);
1167#endif
1168
1169 mrtstat.mrts_no_route++;
1170 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1171 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x",
1172 ntohl(ip->ip_src.s_addr),
1173 ntohl(ip->ip_dst.s_addr));
1174
1175 /*
1176 * Allocate mbufs early so that we don't do extra work if we are
1177 * just going to fail anyway.
1178 */
1179 MGET(mb_ntry, M_DONTWAIT, MT_DATA);
1180 if (mb_ntry == NULL) {
1181 splx(s);
1182 return ENOBUFS;
1183 }
1184 mb0 = m_copy(m, 0, M_COPYALL);
1185 if (mb0 == NULL) {
1186 m_free(mb_ntry);
1187 splx(s);
1188 return ENOBUFS;
1189 }
1190
1191 /* is there an upcall waiting for this packet? */
1192 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1193 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
1194 rt = mtod(mb_rt, struct mfc *);
1195 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1196 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1197 (mb_rt->m_act != NULL))
1198 break;
1199 }
1200
1201 if (mb_rt == NULL) {
1202 int hlen = ip->ip_hl << 2;
1203 int i;
1204 struct igmpmsg *im;
1205
1206 /* no upcall, so make a new entry */
1207 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
1208 if (mb_rt == NULL) {
1209 m_free(mb_ntry);
1210 m_freem(mb0);
1211 splx(s);
1212 return ENOBUFS;
1213 }
1214 /* Make a copy of the header to send to the user level process */
1215 mm = m_copy(m, 0, hlen);
1216 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1217 mm = m_pullup(mm, hlen);
1218 if (mm == NULL) {
1219 m_free(mb_ntry);
1220 m_freem(mb0);
1221 m_free(mb_rt);
1222 splx(s);
1223 return ENOBUFS;
1224 }
1225
1226 /*
1227 * Send message to routing daemon to install
1228 * a route into the kernel table
1229 */
1230 k_igmpsrc.sin_addr = ip->ip_src;
1231
1232 im = mtod(mm, struct igmpmsg *);
1233 im->im_msgtype = IGMPMSG_NOCACHE;
1234 im->im_mbz = 0;
1235
1236 mrtstat.mrts_upcalls++;
1237
1238 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1239 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full");
1240 ++mrtstat.mrts_upq_sockfull;
1241 m_free(mb_ntry);
1242 m_freem(mb0);
1243 m_free(mb_rt);
1244 splx(s);
1245 return ENOBUFS;
1246 }
1247
1248 rt = mtod(mb_rt, struct mfc *);
1249
1250 /* insert new entry at head of hash chain */
1251 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1252 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1253 rt->mfc_expire = UPCALL_EXPIRE;
1254 nexpire[hash]++;
1255 for (i = 0; i < numvifs; i++)
1256 rt->mfc_ttls[i] = 0;
1257 rt->mfc_parent = -1;
1258
1259 /* link into table */
1260 mb_rt->m_next = mfctable[hash];
1261 mfctable[hash] = mb_rt;
1262 mb_rt->m_act = NULL;
1263
1264 rte_m = mb_rt;
1265 } else {
1266 /* determine if q has overflowed */
1267 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act)
1268 npkts++;
1269
1270 if (npkts > MAX_UPQ) {
1271 mrtstat.mrts_upq_ovflw++;
1272 m_free(mb_ntry);
1273 m_freem(mb0);
1274 splx(s);
1275 return 0;
1276 }
1277 }
1278
1279 mb_ntry->m_act = NULL;
1280 rte = mtod(mb_ntry, struct rtdetq *);
1281
1282 rte->m = mb0;
1283 rte->ifp = ifp;
1284#ifdef UPCALL_TIMING
1285 rte->t = tp;
1286#endif
1287
1288 /* Add this entry to the end of the queue */
1289 rte_m->m_act = mb_ntry;
1290
1291 splx(s);
1292
1293 return 0;
1294 }
1295}
1296
1297#ifndef MROUTE_LKM
1298int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
1299 struct ip_moptions *) = X_ip_mforward;
1300#endif
1301
1302/*
1303 * Clean up the cache entry if upcall is not serviced
1304 */
1305static void
1306expire_upcalls(void *unused)
1307{
1308 struct mbuf *mb_rt, *m, **nptr;
1309 struct rtdetq *rte;
1310 struct mfc *mfc;
1311 int i;
1312 int s;
1313
1314 s = splnet();
1315 for (i = 0; i < MFCTBLSIZ; i++) {
1316 if (nexpire[i] == 0)
1317 continue;
1318 nptr = &mfctable[i];
1319 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) {
1320 mfc = mtod(mb_rt, struct mfc *);
1321
1322 /*
1323 * Skip real cache entries
1324 * Make sure it wasn't marked to not expire (shouldn't happen)
1325 * If it expires now
1326 */
1327 if (mb_rt->m_act != NULL &&
1328 mfc->mfc_expire != 0 &&
1329 --mfc->mfc_expire == 0) {
1330 if (mrtdebug & DEBUG_EXPIRE)
1331 log(LOG_DEBUG, "expire_upcalls: expiring (%x %x)",
1332 ntohl(mfc->mfc_origin.s_addr),
1333 ntohl(mfc->mfc_mcastgrp.s_addr));
1334 /*
1335 * drop all the packets
1336 * free the mbuf with the pkt, if, timing info
1337 */
1338 while (mb_rt->m_act) {
1339 m = mb_rt->m_act;
1340 mb_rt->m_act = m->m_act;
1341
1342 rte = mtod(m, struct rtdetq *);
1343 m_freem(rte->m);
1344 m_free(m);
1345 }
1346 ++mrtstat.mrts_cache_cleanups;
1347 nexpire[i]--;
1348
1349 MFREE(mb_rt, *nptr);
1350 } else {
1351 nptr = &mb_rt->m_next;
1352 }
1353 }
1354 }
1355 splx(s);
1356 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
1357}
1358
1359/*
1360 * Packet forwarding routine once entry in the cache is made
1361 */
1362static int
1363ip_mdq(m, ifp, rt, xmt_vif)
1364 register struct mbuf *m;
1365 register struct ifnet *ifp;
1366 register struct mfc *rt;
1367 register vifi_t xmt_vif;
1368{
1369 register struct ip *ip = mtod(m, struct ip *);
1370 register vifi_t vifi;
1371 register struct vif *vifp;
1372 register struct mbuf *tmp;
1373 register int plen = ntohs(ip->ip_len);
1374
1375/*
1376 * Macro to send packet on vif. Since RSVP packets don't get counted on
1377 * input, they shouldn't get counted on output, so statistics keeping is
1378 * seperate.
1379 */
1380#define MC_SEND(ip,vifp,m) { \
1381 if ((vifp)->v_flags & VIFF_TUNNEL) \
1382 encap_send((ip), (vifp), (m)); \
1383 else \
1384 phyint_send((ip), (vifp), (m)); \
1385}
1386
1387 /*
1388 * If xmt_vif is not -1, send on only the requested vif.
1389 *
1390 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1391 */
1392 if (xmt_vif < numvifs) {
1393 MC_SEND(ip, viftable + xmt_vif, m);
1394 return 1;
1395 }
1396
1397 /*
1398 * Don't forward if it didn't arrive from the parent vif for its origin.
1399 */
1400 vifi = rt->mfc_parent;
1401 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1402 /* came in the wrong interface */
1403 if (mrtdebug & DEBUG_FORWARD)
1404 log(LOG_DEBUG, "wrong if: ifp %x vifi %d vififp %x",
1405 ifp, vifi, viftable[vifi].v_ifp);
1406 ++mrtstat.mrts_wrong_if;
1407 ++rt->mfc_wrong_if;
1408 /*
1409 * If we are doing PIM assert processing, and we are forwarding
1410 * packets on this interface, and it is a broadcast medium
1411 * interface (and not a tunnel), send a message to the routing daemon.
1412 */
1413 if (pim_assert && rt->mfc_ttls[vifi] &&
1414 (ifp->if_flags & IFF_BROADCAST) &&
1415 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1416 struct sockaddr_in k_igmpsrc;
1417 struct mbuf *mm;
1418 struct igmpmsg *im;
1419 int hlen = ip->ip_hl << 2;
1420 struct timeval now;
1421 register u_long delta;
1422
1423 GET_TIME(now);
1424
1425 TV_DELTA(rt->mfc_last_assert, now, delta);
1426
1427 if (delta > ASSERT_MSG_TIME) {
1428 mm = m_copy(m, 0, hlen);
1429 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1430 mm = m_pullup(mm, hlen);
1431 if (mm == NULL) {
1432 return ENOBUFS;
1433 }
1434
1435 rt->mfc_last_assert = now;
1436
1437 im = mtod(mm, struct igmpmsg *);
1438 im->im_msgtype = IGMPMSG_WRONGVIF;
1439 im->im_mbz = 0;
1440 im->im_vif = vifi;
1441
1442 k_igmpsrc.sin_addr = im->im_src;
1443
1444 socket_send(ip_mrouter, mm, &k_igmpsrc);
1445 }
1446 }
1447 return 0;
1448 }
1449
1450 /* If I sourced this packet, it counts as output, else it was input. */
1451 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1452 viftable[vifi].v_pkt_out++;
1453 viftable[vifi].v_bytes_out += plen;
1454 } else {
1455 viftable[vifi].v_pkt_in++;
1456 viftable[vifi].v_bytes_in += plen;
1457 }
1458 rt->mfc_pkt_cnt++;
1459 rt->mfc_byte_cnt += plen;
1460
1461 /*
1462 * For each vif, decide if a copy of the packet should be forwarded.
1463 * Forward if:
1464 * - the ttl exceeds the vif's threshold
1465 * - there are group members downstream on interface
1466 */
1467 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1468 if ((rt->mfc_ttls[vifi] > 0) &&
1469 (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1470 vifp->v_pkt_out++;
1471 vifp->v_bytes_out += plen;
1472 MC_SEND(ip, vifp, m);
1473 }
1474
1475 return 0;
1476}
1477
1478/*
1479 * check if a vif number is legal/ok. This is used by ip_output, to export
1480 * numvifs there,
1481 */
1482int
1483X_legal_vif_num(vif)
1484 int vif;
1485{
1486 if (vif >= 0 && vif < numvifs)
1487 return(1);
1488 else
1489 return(0);
1490}
1491
1492#ifndef MROUTE_LKM
1493int (*legal_vif_num)(int) = X_legal_vif_num;
1494#endif
1495
1496/*
1497 * Return the local address used by this vif
1498 */
1499u_long
1500X_ip_mcast_src(vifi)
1501 int vifi;
1502{
1503 if (vifi >= 0 && vifi < numvifs)
1504 return viftable[vifi].v_lcl_addr.s_addr;
1505 else
1506 return INADDR_ANY;
1507}
1508
1509#ifndef MROUTE_LKM
1510u_long (*ip_mcast_src)(int) = X_ip_mcast_src;
1511#endif
1512
1513static void
1514phyint_send(ip, vifp, m)
1515 struct ip *ip;
1516 struct vif *vifp;
1517 struct mbuf *m;
1518{
1519 register struct mbuf *mb_copy;
1520 register int hlen = ip->ip_hl << 2;
1521 register struct ip_moptions *imo;
1522
1523 /*
1524 * Make a new reference to the packet; make sure that
1525 * the IP header is actually copied, not just referenced,
1526 * so that ip_output() only scribbles on the copy.
1527 */
1528 mb_copy = m_copy(m, 0, M_COPYALL);
1529 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1530 mb_copy = m_pullup(mb_copy, hlen);
1531 if (mb_copy == NULL)
1532 return;
1533
1534 MALLOC(imo, struct ip_moptions *, sizeof *imo, M_IPMOPTS, M_NOWAIT);
1535 if (imo == NULL) {
1536 m_freem(mb_copy);
1537 return;
1538 }
1539
1540 imo->imo_multicast_ifp = vifp->v_ifp;
1541 imo->imo_multicast_ttl = ip->ip_ttl - 1;
1542 imo->imo_multicast_loop = 1;
1543 imo->imo_multicast_vif = -1;
1544
1545 if (vifp->v_rate_limit <= 0)
1546 tbf_send_packet(vifp, mb_copy, imo);
1547 else
1548 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len,
1549 imo);
1550}
1551
1552static void
1553encap_send(ip, vifp, m)
1554 register struct ip *ip;
1555 register struct vif *vifp;
1556 register struct mbuf *m;
1557{
1558 register struct mbuf *mb_copy;
1559 register struct ip *ip_copy;
1560 int hlen = ip->ip_hl << 2;
1561 register int i, len = ip->ip_len;
1562
1563 /*
1564 * copy the old packet & pullup it's IP header into the
1565 * new mbuf so we can modify it. Try to fill the new
1566 * mbuf since if we don't the ethernet driver will.
1567 */
1568 MGET(mb_copy, M_DONTWAIT, MT_DATA);
1569 if (mb_copy == NULL)
1570 return;
1571 mb_copy->m_data += 16;
1572 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1573
1574 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1575 m_freem(mb_copy);
1576 return;
1577 }
1578 i = MHLEN - M_LEADINGSPACE(mb_copy);
1579 if (i > len)
1580 i = len;
1581 mb_copy = m_pullup(mb_copy, i);
1582 if (mb_copy == NULL)
1583 return;
1584 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1585
1586 /*
1587 * fill in the encapsulating IP header.
1588 */
1589 ip_copy = mtod(mb_copy, struct ip *);
1590 *ip_copy = multicast_encap_iphdr;
1591 ip_copy->ip_id = htons(ip_id++);
1592 ip_copy->ip_len += len;
1593 ip_copy->ip_src = vifp->v_lcl_addr;
1594 ip_copy->ip_dst = vifp->v_rmt_addr;
1595
1596 /*
1597 * turn the encapsulated IP header back into a valid one.
1598 */
1599 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1600 --ip->ip_ttl;
1601 HTONS(ip->ip_len);
1602 HTONS(ip->ip_off);
1603 ip->ip_sum = 0;
1604#if defined(LBL) && !defined(ultrix)
1605 ip->ip_sum = ~oc_cksum((caddr_t)ip, ip->ip_hl << 2, 0);
1606#else
1607 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1608 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1609 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1610#endif
1611
1612 if (vifp->v_rate_limit <= 0)
1613 tbf_send_packet(vifp, mb_copy, 0);
1614 else
1615 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len, 0);
1616}
1617
1618/*
1619 * De-encapsulate a packet and feed it back through ip input (this
1620 * routine is called whenever IP gets a packet with proto type
1621 * ENCAP_PROTO and a local destination address).
1622 */
1623void
1624#ifdef MROUTE_LKM
1625X_ipip_input(m)
1626#else
1627ipip_input(m, iphlen)
1628#endif
1629 register struct mbuf *m;
1630 int iphlen;
1631{
1632 struct ifnet *ifp = m->m_pkthdr.rcvif;
1633 register struct ip *ip = mtod(m, struct ip *);
1634 register int hlen = ip->ip_hl << 2;
1635 register int s;
1636 register struct ifqueue *ifq;
1637 register struct vif *vifp;
1638
1639 if (!have_encap_tunnel) {
1640 rip_input(m);
1641 return;
1642 }
1643 /*
1644 * dump the packet if it's not to a multicast destination or if
1645 * we don't have an encapsulating tunnel with the source.
1646 * Note: This code assumes that the remote site IP address
1647 * uniquely identifies the tunnel (i.e., that this site has
1648 * at most one tunnel with the remote site).
1649 */
1650 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) {
1651 ++mrtstat.mrts_bad_tunnel;
1652 m_freem(m);
1653 return;
1654 }
1655 if (ip->ip_src.s_addr != last_encap_src) {
1656 register struct vif *vife;
1657
1658 vifp = viftable;
1659 vife = vifp + numvifs;
1660 last_encap_src = ip->ip_src.s_addr;
1661 last_encap_vif = 0;
1662 for ( ; vifp < vife; ++vifp)
1663 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
1664 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT))
1665 == VIFF_TUNNEL)
1666 last_encap_vif = vifp;
1667 break;
1668 }
1669 }
1670 if ((vifp = last_encap_vif) == 0) {
1671 last_encap_src = 0;
1672 mrtstat.mrts_cant_tunnel++; /*XXX*/
1673 m_freem(m);
1674 if (mrtdebug)
1675 log(LOG_DEBUG, "ip_mforward: no tunnel with %x",
1676 ntohl(ip->ip_src.s_addr));
1677 return;
1678 }
1679 ifp = vifp->v_ifp;
1680
1681 if (hlen > IP_HDR_LEN)
1682 ip_stripoptions(m, (struct mbuf *) 0);
1683 m->m_data += IP_HDR_LEN;
1684 m->m_len -= IP_HDR_LEN;
1685 m->m_pkthdr.len -= IP_HDR_LEN;
1686 m->m_pkthdr.rcvif = ifp;
1687
1688 ifq = &ipintrq;
1689 s = splimp();
1690 if (IF_QFULL(ifq)) {
1691 IF_DROP(ifq);
1692 m_freem(m);
1693 } else {
1694 IF_ENQUEUE(ifq, m);
1695 /*
1696 * normally we would need a "schednetisr(NETISR_IP)"
1697 * here but we were called by ip_input and it is going
1698 * to loop back & try to dequeue the packet we just
1699 * queued as soon as we return so we avoid the
1700 * unnecessary software interrrupt.
1701 */
1702 }
1703 splx(s);
1704}
1705
1706/*
1707 * Token bucket filter module
1708 */
1709static void
1710tbf_control(vifp, m, ip, p_len, imo)
1711 register struct vif *vifp;
1712 register struct mbuf *m;
1713 register struct ip *ip;
1714 register u_long p_len;
1715 struct ip_moptions *imo;
1716{
1717 tbf_update_tokens(vifp);
1718
1719 /* if there are enough tokens,
1720 * and the queue is empty,
1721 * send this packet out
1722 */
1723
1724 if (vifp->v_tbf->q_len == 0) {
1725 if (p_len <= vifp->v_tbf->n_tok) {
1726 vifp->v_tbf->n_tok -= p_len;
1727 tbf_send_packet(vifp, m, imo);
1728 } else if (p_len > MAX_BKT_SIZE) {
1729 /* drop if packet is too large */
1730 mrtstat.mrts_pkt2large++;
1731 m_freem(m);
1732 return;
1733 } else {
1734 /* queue packet and timeout till later */
1735 tbf_queue(vifp, m, ip, imo);
1736 timeout(tbf_reprocess_q, (caddr_t)vifp, 1);
1737 }
1738 } else if (vifp->v_tbf->q_len < MAXQSIZE) {
1739 /* finite queue length, so queue pkts and process queue */
1740 tbf_queue(vifp, m, ip, imo);
1741 tbf_process_q(vifp);
1742 } else {
1743 /* queue length too much, try to dq and queue and process */
1744 if (!tbf_dq_sel(vifp, ip)) {
1745 mrtstat.mrts_q_overflow++;
1746 m_freem(m);
1747 return;
1748 } else {
1749 tbf_queue(vifp, m, ip, imo);
1750 tbf_process_q(vifp);
1751 }
1752 }
1753 return;
1754}
1755
1756/*
1757 * adds a packet to the queue at the interface
1758 */
1759static void
1760tbf_queue(vifp, m, ip, imo)
1761 register struct vif *vifp;
1762 register struct mbuf *m;
1763 register struct ip *ip;
1764 struct ip_moptions *imo;
1765{
1766 register u_long ql;
1767 register int index = (vifp - viftable);
1768 register int s = splnet();
1769
1770 ql = vifp->v_tbf->q_len;
1771
1772 qtable[index][ql].pkt_m = m;
1773 qtable[index][ql].pkt_len = (mtod(m, struct ip *))->ip_len;
1774 qtable[index][ql].pkt_ip = ip;
1775 qtable[index][ql].pkt_imo = imo;
1776
1777 vifp->v_tbf->q_len++;
1778 splx(s);
1779}
1780
1781
1782/*
1783 * processes the queue at the interface
1784 */
1785static void
1786tbf_process_q(vifp)
1787 register struct vif *vifp;
1788{
1789 register struct pkt_queue pkt_1;
1790 register int index = (vifp - viftable);
1791 register int s = splnet();
1792
1793 /* loop through the queue at the interface and send as many packets
1794 * as possible
1795 */
1796 while (vifp->v_tbf->q_len > 0) {
1797 /* locate the first packet */
1798 pkt_1.pkt_len = (qtable[index][0]).pkt_len;
1799 pkt_1.pkt_m = (qtable[index][0]).pkt_m;
1800 pkt_1.pkt_ip = (qtable[index][0]).pkt_ip;
1801 pkt_1.pkt_imo = (qtable[index][0]).pkt_imo;
1802
1803 /* determine if the packet can be sent */
1804 if (pkt_1.pkt_len <= vifp->v_tbf->n_tok) {
1805 /* if so,
1806 * reduce no of tokens, dequeue the queue,
1807 * send the packet.
1808 */
1809 vifp->v_tbf->n_tok -= pkt_1.pkt_len;
1810
1811 tbf_dequeue(vifp, 0);
1812
1813 tbf_send_packet(vifp, pkt_1.pkt_m, pkt_1.pkt_imo);
1814
1815 } else break;
1816 }
1817 splx(s);
1818}
1819
1820/*
1821 * removes the jth packet from the queue at the interface
1822 */
1823static void
1824tbf_dequeue(vifp,j)
1825 register struct vif *vifp;
1826 register int j;
1827{
1828 register u_long index = vifp - viftable;
1829 register int i;
1830
1831 for (i=j+1; i <= vifp->v_tbf->q_len - 1; i++) {
1832 qtable[index][i-1].pkt_m = qtable[index][i].pkt_m;
1833 qtable[index][i-1].pkt_len = qtable[index][i].pkt_len;
1834 qtable[index][i-1].pkt_ip = qtable[index][i].pkt_ip;
1835 qtable[index][i-1].pkt_imo = qtable[index][i].pkt_imo;
1836 }
1837 qtable[index][i-1].pkt_m = NULL;
1838 qtable[index][i-1].pkt_len = NULL;
1839 qtable[index][i-1].pkt_ip = NULL;
1840 qtable[index][i-1].pkt_imo = NULL;
1841
1842 vifp->v_tbf->q_len--;
1843
1844 if (tbfdebug > 1)
1845 log(LOG_DEBUG, "tbf_dequeue: vif# %d qlen %d",vifp-viftable, i-1);
1846}
1847
1848static void
1849tbf_reprocess_q(xvifp)
1850 void *xvifp;
1851{
1852 register struct vif *vifp = xvifp;
1853 if (ip_mrouter == NULL)
1854 return;
1855
1856 tbf_update_tokens(vifp);
1857
1858 tbf_process_q(vifp);
1859
1860 if (vifp->v_tbf->q_len)
1861 timeout(tbf_reprocess_q, (caddr_t)vifp, 1);
1862}
1863
1864/* function that will selectively discard a member of the queue
1865 * based on the precedence value and the priority obtained through
1866 * a lookup table - not yet implemented accurately!
1867 */
1868static int
1869tbf_dq_sel(vifp, ip)
1870 register struct vif *vifp;
1871 register struct ip *ip;
1872{
1873 register int i;
1874 register int s = splnet();
1875 register u_int p;
1876
1877 p = priority(vifp, ip);
1878
1879 for(i=vifp->v_tbf->q_len-1;i >= 0;i--) {
1880 if (p > priority(vifp, qtable[vifp-viftable][i].pkt_ip)) {
1881 m_freem(qtable[vifp-viftable][i].pkt_m);
1882 tbf_dequeue(vifp,i);
1883 splx(s);
1884 mrtstat.mrts_drop_sel++;
1885 return(1);
1886 }
1887 }
1888 splx(s);
1889 return(0);
1890}
1891
1892static void
1893tbf_send_packet(vifp, m, imo)
1894 register struct vif *vifp;
1895 register struct mbuf *m;
1896 struct ip_moptions *imo;
1897{
1898 int error;
1899 int s = splnet();
1900
1901 if (vifp->v_flags & VIFF_TUNNEL) {
1902 /* If tunnel options */
1903 ip_output(m, (struct mbuf *)0, (struct route *)0,
1904 IP_FORWARDING, imo);
1905 } else {
1906 /* if physical interface option, extract the options and then send */
1907 error = ip_output(m, (struct mbuf *)0, (struct route *)0,
1908 IP_FORWARDING, imo);
1909 FREE(imo, M_IPMOPTS);
1910
1911 if (mrtdebug & DEBUG_XMIT)
1912 log(LOG_DEBUG, "phyint_send on vif %d err %d", vifp-viftable, error);
1913 }
1914 splx(s);
1915}
1916
1917/* determine the current time and then
1918 * the elapsed time (between the last time and time now)
1919 * in milliseconds & update the no. of tokens in the bucket
1920 */
1921static void
1922tbf_update_tokens(vifp)
1923 register struct vif *vifp;
1924{
1925 struct timeval tp;
1926 register u_long t;
1927 register u_long elapsed;
1928 register int s = splnet();
1929
1930 GET_TIME(tp);
1931
1932 t = tp.tv_sec*1000 + tp.tv_usec/1000;
1933
1934 elapsed = (t - vifp->v_tbf->last_pkt_t) * vifp->v_rate_limit /8;
1935 vifp->v_tbf->n_tok += elapsed;
1936 vifp->v_tbf->last_pkt_t = t;
1937
1938 if (vifp->v_tbf->n_tok > MAX_BKT_SIZE)
1939 vifp->v_tbf->n_tok = MAX_BKT_SIZE;
1940
1941 splx(s);
1942}
1943
1944static int
1945priority(vifp, ip)
1946 register struct vif *vifp;
1947 register struct ip *ip;
1948{
1949 register int prio;
1950
1951 /* temporary hack; may add general packet classifier some day */
1952
1953 /*
1954 * The UDP port space is divided up into four priority ranges:
1955 * [0, 16384) : unclassified - lowest priority
1956 * [16384, 32768) : audio - highest priority
1957 * [32768, 49152) : whiteboard - medium priority
1958 * [49152, 65536) : video - low priority
1959 */
1960 if (ip->ip_p == IPPROTO_UDP) {
1961 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
1962 switch (ntohs(udp->uh_dport) & 0xc000) {
1963 case 0x4000:
1964 prio = 70;
1965 break;
1966 case 0x8000:
1967 prio = 60;
1968 break;
1969 case 0xc000:
1970 prio = 55;
1971 break;
1972 default:
1973 prio = 50;
1974 break;
1975 }
1976 if (tbfdebug > 1)
1977 log(LOG_DEBUG, "port %x prio%d", ntohs(udp->uh_dport), prio);
1978 } else {
1979 prio = 50;
1980 }
1981 return prio;
1982}
1983
1984/*
1985 * End of token bucket filter modifications
1986 */
1987
1988int
1989ip_rsvp_vif_init(so, m)
1990 struct socket *so;
1991 struct mbuf *m;
1992{
1993 int i;
1994 register int s;
1995
1996 if (rsvpdebug)
1997 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
1998 so->so_type, so->so_proto->pr_protocol);
1999
2000 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2001 return EOPNOTSUPP;
2002
2003 /* Check mbuf. */
2004 if (m == NULL || m->m_len != sizeof(int)) {
2005 return EINVAL;
2006 }
2007 i = *(mtod(m, int *));
2008
2009 if (rsvpdebug)
2010 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on);
2011
2012 s = splnet();
2013
2014 /* Check vif. */
2015 if (!legal_vif_num(i)) {
2016 splx(s);
2017 return EADDRNOTAVAIL;
2018 }
2019
2020 /* Check if socket is available. */
2021 if (viftable[i].v_rsvpd != NULL) {
2022 splx(s);
2023 return EADDRINUSE;
2024 }
2025
2026 viftable[i].v_rsvpd = so;
2027 /* This may seem silly, but we need to be sure we don't over-increment
2028 * the RSVP counter, in case something slips up.
2029 */
2030 if (!viftable[i].v_rsvp_on) {
2031 viftable[i].v_rsvp_on = 1;
2032 rsvp_on++;
2033 }
2034
2035 splx(s);
2036 return 0;
2037}
2038
2039int
2040ip_rsvp_vif_done(so, m)
2041 struct socket *so;
2042 struct mbuf *m;
2043{
2044 int i;
2045 register int s;
2046
2047 if (rsvpdebug)
2048 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2049 so->so_type, so->so_proto->pr_protocol);
2050
2051 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2052 return EOPNOTSUPP;
2053
2054 /* Check mbuf. */
2055 if (m == NULL || m->m_len != sizeof(int)) {
2056 return EINVAL;
2057 }
2058 i = *(mtod(m, int *));
2059
2060 s = splnet();
2061
2062 /* Check vif. */
2063 if (!legal_vif_num(i)) {
2064 splx(s);
2065 return EADDRNOTAVAIL;
2066 }
2067
2068 if (rsvpdebug)
2069 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2070 viftable[i].v_rsvpd, so);
2071
2072 viftable[i].v_rsvpd = NULL;
2073 /* This may seem silly, but we need to be sure we don't over-decrement
2074 * the RSVP counter, in case something slips up.
2075 */
2076 if (viftable[i].v_rsvp_on) {
2077 viftable[i].v_rsvp_on = 0;
2078 rsvp_on--;
2079 }
2080
2081 splx(s);
2082 return 0;
2083}
2084
2085void
2086ip_rsvp_force_done(so)
2087 struct socket *so;
2088{
2089 int vifi;
2090 register int s;
2091
2092 /* Don't bother if it is not the right type of socket. */
2093 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2094 return;
2095
2096 s = splnet();
2097
2098 /* The socket may be attached to more than one vif...this
2099 * is perfectly legal.
2100 */
2101 for (vifi = 0; vifi < numvifs; vifi++) {
2102 if (viftable[vifi].v_rsvpd == so) {
2103 viftable[vifi].v_rsvpd = NULL;
2104 /* This may seem silly, but we need to be sure we don't
2105 * over-decrement the RSVP counter, in case something slips up.
2106 */
2107 if (viftable[vifi].v_rsvp_on) {
2108 viftable[vifi].v_rsvp_on = 0;
2109 rsvp_on--;
2110 }
2111 }
2112 }
2113
2114 splx(s);
2115 return;
2116}
2117
2118void
|