13 */
14
15#include "opt_mrouting.h"
16#include "opt_random_ip_id.h"
17
18#include <sys/param.h>
19#include <sys/kernel.h>
20#include <sys/lock.h>
21#include <sys/malloc.h>
22#include <sys/mbuf.h>
23#include <sys/protosw.h>
24#include <sys/signalvar.h>
25#include <sys/socket.h>
26#include <sys/socketvar.h>
27#include <sys/sockio.h>
28#include <sys/sx.h>
29#include <sys/sysctl.h>
30#include <sys/syslog.h>
31#include <sys/systm.h>
32#include <sys/time.h>
33#include <net/if.h>
34#include <net/route.h>
35#include <netinet/in.h>
36#include <netinet/igmp.h>
37#include <netinet/in_systm.h>
38#include <netinet/in_var.h>
39#include <netinet/ip.h>
40#include <netinet/ip_encap.h>
41#include <netinet/ip_mroute.h>
42#include <netinet/ip_var.h>
43#include <netinet/udp.h>
44#include <machine/in_cksum.h>
45
46#ifndef MROUTING
47extern u_long _ip_mcast_src(int vifi);
48extern int _ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
49 struct ip_moptions *imo);
50extern int _ip_mrouter_done(void);
51extern int _ip_mrouter_get(struct socket *so, struct sockopt *sopt);
52extern int _ip_mrouter_set(struct socket *so, struct sockopt *sopt);
53extern int _mrt_ioctl(int req, caddr_t data);
54
55/*
56 * Dummy routines and globals used when multicast routing is not compiled in.
57 */
58
59struct socket *ip_mrouter = NULL;
60u_int rsvpdebug = 0;
61
62int
63_ip_mrouter_set(so, sopt)
64 struct socket *so;
65 struct sockopt *sopt;
66{
67 return(EOPNOTSUPP);
68}
69
70int (*ip_mrouter_set)(struct socket *, struct sockopt *) = _ip_mrouter_set;
71
72
73int
74_ip_mrouter_get(so, sopt)
75 struct socket *so;
76 struct sockopt *sopt;
77{
78 return(EOPNOTSUPP);
79}
80
81int (*ip_mrouter_get)(struct socket *, struct sockopt *) = _ip_mrouter_get;
82
83int
84_ip_mrouter_done()
85{
86 return(0);
87}
88
89int (*ip_mrouter_done)(void) = _ip_mrouter_done;
90
91int
92_ip_mforward(ip, ifp, m, imo)
93 struct ip *ip;
94 struct ifnet *ifp;
95 struct mbuf *m;
96 struct ip_moptions *imo;
97{
98 return(0);
99}
100
101int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
102 struct ip_moptions *) = _ip_mforward;
103
104int
105_mrt_ioctl(int req, caddr_t data)
106{
107 return EOPNOTSUPP;
108}
109
110int (*mrt_ioctl)(int, caddr_t) = _mrt_ioctl;
111
112void
113rsvp_input(m, off) /* XXX must fixup manually */
114 struct mbuf *m;
115 int off;
116{
117 /* Can still get packets with rsvp_on = 0 if there is a local member
118 * of the group to which the RSVP packet is addressed. But in this
119 * case we want to throw the packet away.
120 */
121 if (!rsvp_on) {
122 m_freem(m);
123 return;
124 }
125
126 if (ip_rsvpd != NULL) {
127 if (rsvpdebug)
128 printf("rsvp_input: Sending packet up old-style socket\n");
129 rip_input(m, off);
130 return;
131 }
132 /* Drop the packet */
133 m_freem(m);
134}
135
136int (*legal_vif_num)(int) = 0;
137
138/*
139 * This should never be called, since IP_MULTICAST_VIF should fail, but
140 * just in case it does get called, the code a little lower in ip_output
141 * will assign the packet a local address.
142 */
143u_long
144_ip_mcast_src(int vifi) { return INADDR_ANY; }
145u_long (*ip_mcast_src)(int) = _ip_mcast_src;
146
147int
148ip_rsvp_vif_init(so, sopt)
149 struct socket *so;
150 struct sockopt *sopt;
151{
152 return(EINVAL);
153}
154
155int
156ip_rsvp_vif_done(so, sopt)
157 struct socket *so;
158 struct sockopt *sopt;
159{
160 return(EINVAL);
161}
162
163void
164ip_rsvp_force_done(so)
165 struct socket *so;
166{
167 return;
168}
169
170#else /* MROUTING */
171
172#define M_HASCL(m) ((m)->m_flags & M_EXT)
173
174static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables");
175
176#ifndef MROUTE_KLD
177/* The socket used to communicate with the multicast routing daemon. */
178struct socket *ip_mrouter = NULL;
179#endif
180
181#if defined(MROUTING) || defined(MROUTE_KLD)
182static struct mrtstat mrtstat;
183SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
184 &mrtstat, mrtstat, "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)");
185#endif
186
187static struct mfc *mfctable[MFCTBLSIZ];
188static u_char nexpire[MFCTBLSIZ];
189static struct vif viftable[MAXVIFS];
190static u_int mrtdebug = 0; /* debug level */
191#define DEBUG_MFC 0x02
192#define DEBUG_FORWARD 0x04
193#define DEBUG_EXPIRE 0x08
194#define DEBUG_XMIT 0x10
195static u_int tbfdebug = 0; /* tbf debug level */
196static u_int rsvpdebug = 0; /* rsvp debug level */
197
198static struct callout_handle expire_upcalls_ch;
199
200#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
201#define UPCALL_EXPIRE 6 /* number of timeouts */
202
203/*
204 * Define the token bucket filter structures
205 * tbftable -> each vif has one of these for storing info
206 */
207
208static struct tbf tbftable[MAXVIFS];
209#define TBF_REPROCESS (hz / 100) /* 100x / second */
210
211/*
212 * 'Interfaces' associated with decapsulator (so we can tell
213 * packets that went through it from ones that get reflected
214 * by a broken gateway). These interfaces are never linked into
215 * the system ifnet list & no routes point to them. I.e., packets
216 * can't be sent this way. They only exist as a placeholder for
217 * multicast source verification.
218 */
219static struct ifnet multicast_decap_if[MAXVIFS];
220
221#define ENCAP_TTL 64
222#define ENCAP_PROTO IPPROTO_IPIP /* 4 */
223
224/* prototype IP hdr for encapsulated packets */
225static struct ip multicast_encap_iphdr = {
226#if BYTE_ORDER == LITTLE_ENDIAN
227 sizeof(struct ip) >> 2, IPVERSION,
228#else
229 IPVERSION, sizeof(struct ip) >> 2,
230#endif
231 0, /* tos */
232 sizeof(struct ip), /* total length */
233 0, /* id */
234 0, /* frag offset */
235 ENCAP_TTL, ENCAP_PROTO,
236 0, /* checksum */
237};
238
239/*
240 * Private variables.
241 */
242static vifi_t numvifs = 0;
243static const struct encaptab *encap_cookie = NULL;
244
245/*
246 * one-back cache used by mroute_encapcheck to locate a tunnel's vif
247 * given a datagram's src ip address.
248 */
249static u_long last_encap_src;
250static struct vif *last_encap_vif;
251
252static u_long X_ip_mcast_src(int vifi);
253static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo);
254static int X_ip_mrouter_done(void);
255static int X_ip_mrouter_get(struct socket *so, struct sockopt *m);
256static int X_ip_mrouter_set(struct socket *so, struct sockopt *m);
257static int X_legal_vif_num(int vif);
258static int X_mrt_ioctl(int cmd, caddr_t data);
259
260static int get_sg_cnt(struct sioc_sg_req *);
261static int get_vif_cnt(struct sioc_vif_req *);
262static int ip_mrouter_init(struct socket *, int);
263static int add_vif(struct vifctl *);
264static int del_vif(vifi_t);
265static int add_mfc(struct mfcctl *);
266static int del_mfc(struct mfcctl *);
267static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
268static int set_assert(int);
269static void expire_upcalls(void *);
270static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *,
271 vifi_t);
272static void phyint_send(struct ip *, struct vif *, struct mbuf *);
273static void encap_send(struct ip *, struct vif *, struct mbuf *);
274static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
275static void tbf_queue(struct vif *, struct mbuf *);
276static void tbf_process_q(struct vif *);
277static void tbf_reprocess_q(void *);
278static int tbf_dq_sel(struct vif *, struct ip *);
279static void tbf_send_packet(struct vif *, struct mbuf *);
280static void tbf_update_tokens(struct vif *);
281static int priority(struct vif *, struct ip *);
282
283/*
284 * whether or not special PIM assert processing is enabled.
285 */
286static int pim_assert;
287/*
288 * Rate limit for assert notification messages, in usec
289 */
290#define ASSERT_MSG_TIME 3000000
291
292/*
293 * Hash function for a source, group entry
294 */
295#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
296 ((g) >> 20) ^ ((g) >> 10) ^ (g))
297
298/*
299 * Find a route for a given origin IP address and Multicast group address
300 * Type of service parameter to be added in the future!!!
301 */
302
303#define MFCFIND(o, g, rt) { \
304 register struct mfc *_rt = mfctable[MFCHASH(o,g)]; \
305 rt = NULL; \
306 ++mrtstat.mrts_mfc_lookups; \
307 while (_rt) { \
308 if ((_rt->mfc_origin.s_addr == o) && \
309 (_rt->mfc_mcastgrp.s_addr == g) && \
310 (_rt->mfc_stall == NULL)) { \
311 rt = _rt; \
312 break; \
313 } \
314 _rt = _rt->mfc_next; \
315 } \
316 if (rt == NULL) { \
317 ++mrtstat.mrts_mfc_misses; \
318 } \
319}
320
321
322/*
323 * Macros to compute elapsed time efficiently
324 * Borrowed from Van Jacobson's scheduling code
325 */
326#define TV_DELTA(a, b, delta) { \
327 register int xxs; \
328 \
329 delta = (a).tv_usec - (b).tv_usec; \
330 if ((xxs = (a).tv_sec - (b).tv_sec)) { \
331 switch (xxs) { \
332 case 2: \
333 delta += 1000000; \
334 /* FALLTHROUGH */ \
335 case 1: \
336 delta += 1000000; \
337 break; \
338 default: \
339 delta += (1000000 * xxs); \
340 } \
341 } \
342}
343
344#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
345 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
346
347#ifdef UPCALL_TIMING
348u_long upcall_data[51];
349static void collate(struct timeval *);
350#endif /* UPCALL_TIMING */
351
352
353/*
354 * Handle MRT setsockopt commands to modify the multicast routing tables.
355 */
356static int
357X_ip_mrouter_set(so, sopt)
358 struct socket *so;
359 struct sockopt *sopt;
360{
361 int error, optval;
362 vifi_t vifi;
363 struct vifctl vifc;
364 struct mfcctl mfc;
365
366 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT)
367 return (EPERM);
368
369 error = 0;
370 switch (sopt->sopt_name) {
371 case MRT_INIT:
372 error = sooptcopyin(sopt, &optval, sizeof optval,
373 sizeof optval);
374 if (error)
375 break;
376 error = ip_mrouter_init(so, optval);
377 break;
378
379 case MRT_DONE:
380 error = ip_mrouter_done();
381 break;
382
383 case MRT_ADD_VIF:
384 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
385 if (error)
386 break;
387 error = add_vif(&vifc);
388 break;
389
390 case MRT_DEL_VIF:
391 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
392 if (error)
393 break;
394 error = del_vif(vifi);
395 break;
396
397 case MRT_ADD_MFC:
398 case MRT_DEL_MFC:
399 error = sooptcopyin(sopt, &mfc, sizeof mfc, sizeof mfc);
400 if (error)
401 break;
402 if (sopt->sopt_name == MRT_ADD_MFC)
403 error = add_mfc(&mfc);
404 else
405 error = del_mfc(&mfc);
406 break;
407
408 case MRT_ASSERT:
409 error = sooptcopyin(sopt, &optval, sizeof optval,
410 sizeof optval);
411 if (error)
412 break;
413 set_assert(optval);
414 break;
415
416 default:
417 error = EOPNOTSUPP;
418 break;
419 }
420 return (error);
421}
422
423#ifndef MROUTE_KLD
424int (*ip_mrouter_set)(struct socket *, struct sockopt *) = X_ip_mrouter_set;
425#endif
426
427/*
428 * Handle MRT getsockopt commands
429 */
430static int
431X_ip_mrouter_get(so, sopt)
432 struct socket *so;
433 struct sockopt *sopt;
434{
435 int error;
436 static int version = 0x0305; /* !!! why is this here? XXX */
437
438 switch (sopt->sopt_name) {
439 case MRT_VERSION:
440 error = sooptcopyout(sopt, &version, sizeof version);
441 break;
442
443 case MRT_ASSERT:
444 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert);
445 break;
446 default:
447 error = EOPNOTSUPP;
448 break;
449 }
450 return (error);
451}
452
453#ifndef MROUTE_KLD
454int (*ip_mrouter_get)(struct socket *, struct sockopt *) = X_ip_mrouter_get;
455#endif
456
457/*
458 * Handle ioctl commands to obtain information from the cache
459 */
460static int
461X_mrt_ioctl(cmd, data)
462 int cmd;
463 caddr_t data;
464{
465 int error = 0;
466
467 switch (cmd) {
468 case (SIOCGETVIFCNT):
469 return (get_vif_cnt((struct sioc_vif_req *)data));
470 break;
471 case (SIOCGETSGCNT):
472 return (get_sg_cnt((struct sioc_sg_req *)data));
473 break;
474 default:
475 return (EINVAL);
476 break;
477 }
478 return error;
479}
480
481#ifndef MROUTE_KLD
482int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl;
483#endif
484
485/*
486 * returns the packet, byte, rpf-failure count for the source group provided
487 */
488static int
489get_sg_cnt(req)
490 register struct sioc_sg_req *req;
491{
492 register struct mfc *rt;
493 int s;
494
495 s = splnet();
496 MFCFIND(req->src.s_addr, req->grp.s_addr, rt);
497 splx(s);
498 if (rt != NULL) {
499 req->pktcnt = rt->mfc_pkt_cnt;
500 req->bytecnt = rt->mfc_byte_cnt;
501 req->wrong_if = rt->mfc_wrong_if;
502 } else
503 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
504
505 return 0;
506}
507
508/*
509 * returns the input and output packet and byte counts on the vif provided
510 */
511static int
512get_vif_cnt(req)
513 register struct sioc_vif_req *req;
514{
515 register vifi_t vifi = req->vifi;
516
517 if (vifi >= numvifs) return EINVAL;
518
519 req->icount = viftable[vifi].v_pkt_in;
520 req->ocount = viftable[vifi].v_pkt_out;
521 req->ibytes = viftable[vifi].v_bytes_in;
522 req->obytes = viftable[vifi].v_bytes_out;
523
524 return 0;
525}
526
527/*
528 * Enable multicast routing
529 */
530static int
531ip_mrouter_init(so, version)
532 struct socket *so;
533 int version;
534{
535 if (mrtdebug)
536 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
537 so->so_type, so->so_proto->pr_protocol);
538
539 if (so->so_type != SOCK_RAW ||
540 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP;
541
542 if (version != 1)
543 return ENOPROTOOPT;
544
545 if (ip_mrouter != NULL) return EADDRINUSE;
546
547 ip_mrouter = so;
548
549 bzero((caddr_t)mfctable, sizeof(mfctable));
550 bzero((caddr_t)nexpire, sizeof(nexpire));
551
552 pim_assert = 0;
553
554 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
555
556 if (mrtdebug)
557 log(LOG_DEBUG, "ip_mrouter_init\n");
558
559 return 0;
560}
561
562/*
563 * Disable multicast routing
564 */
565static int
566X_ip_mrouter_done()
567{
568 vifi_t vifi;
569 int i;
570 struct ifnet *ifp;
571 struct ifreq ifr;
572 struct mfc *rt;
573 struct rtdetq *rte;
574 int s;
575
576 s = splnet();
577
578 /*
579 * For each phyint in use, disable promiscuous reception of all IP
580 * multicasts.
581 */
582 for (vifi = 0; vifi < numvifs; vifi++) {
583 if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
584 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
585 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
586 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr
587 = INADDR_ANY;
588 ifp = viftable[vifi].v_ifp;
589 if_allmulti(ifp, 0);
590 }
591 }
592 bzero((caddr_t)tbftable, sizeof(tbftable));
593 bzero((caddr_t)viftable, sizeof(viftable));
594 numvifs = 0;
595 pim_assert = 0;
596
597 untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch);
598
599 /*
600 * Free all multicast forwarding cache entries.
601 */
602 for (i = 0; i < MFCTBLSIZ; i++) {
603 for (rt = mfctable[i]; rt != NULL; ) {
604 struct mfc *nr = rt->mfc_next;
605
606 for (rte = rt->mfc_stall; rte != NULL; ) {
607 struct rtdetq *n = rte->next;
608
609 m_freem(rte->m);
610 free(rte, M_MRTABLE);
611 rte = n;
612 }
613 free(rt, M_MRTABLE);
614 rt = nr;
615 }
616 }
617
618 bzero((caddr_t)mfctable, sizeof(mfctable));
619
620 /*
621 * Reset de-encapsulation cache
622 */
623 last_encap_src = 0;
624 last_encap_vif = NULL;
625 if (encap_cookie) {
626 encap_detach(encap_cookie);
627 encap_cookie = NULL;
628 }
629
630 ip_mrouter = NULL;
631
632 splx(s);
633
634 if (mrtdebug)
635 log(LOG_DEBUG, "ip_mrouter_done\n");
636
637 return 0;
638}
639
640#ifndef MROUTE_KLD
641int (*ip_mrouter_done)(void) = X_ip_mrouter_done;
642#endif
643
644/*
645 * Set PIM assert processing global
646 */
647static int
648set_assert(i)
649 int i;
650{
651 if ((i != 1) && (i != 0))
652 return EINVAL;
653
654 pim_assert = i;
655
656 return 0;
657}
658
659/*
660 * Decide if a packet is from a tunnelled peer.
661 * Return 0 if not, 64 if so.
662 */
663static int
664mroute_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
665{
666 struct ip *ip = mtod(m, struct ip *);
667 int hlen = ip->ip_hl << 2;
668 register struct vif *vifp;
669
670 /*
671 * don't claim the packet if it's not to a multicast destination or if
672 * we don't have an encapsulating tunnel with the source.
673 * Note: This code assumes that the remote site IP address
674 * uniquely identifies the tunnel (i.e., that this site has
675 * at most one tunnel with the remote site).
676 */
677 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) {
678 return 0;
679 }
680 if (ip->ip_src.s_addr != last_encap_src) {
681 register struct vif *vife;
682
683 vifp = viftable;
684 vife = vifp + numvifs;
685 last_encap_src = ip->ip_src.s_addr;
686 last_encap_vif = 0;
687 for ( ; vifp < vife; ++vifp)
688 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
689 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT))
690 == VIFF_TUNNEL)
691 last_encap_vif = vifp;
692 break;
693 }
694 }
695 if ((vifp = last_encap_vif) == 0) {
696 last_encap_src = 0;
697 return 0;
698 }
699 return 64;
700}
701
702/*
703 * De-encapsulate a packet and feed it back through ip input (this
704 * routine is called whenever IP gets a packet that mroute_encap_func()
705 * claimed).
706 */
707static void
708mroute_encap_input(struct mbuf *m, int off)
709{
710 struct ip *ip = mtod(m, struct ip *);
711 int hlen = ip->ip_hl << 2;
712
713 if (hlen > sizeof(struct ip))
714 ip_stripoptions(m, (struct mbuf *) 0);
715 m->m_data += sizeof(struct ip);
716 m->m_len -= sizeof(struct ip);
717 m->m_pkthdr.len -= sizeof(struct ip);
718
719 m->m_pkthdr.rcvif = last_encap_vif->v_ifp;
720
721 (void) IF_HANDOFF(&ipintrq, m, NULL);
722 /*
723 * normally we would need a "schednetisr(NETISR_IP)"
724 * here but we were called by ip_input and it is going
725 * to loop back & try to dequeue the packet we just
726 * queued as soon as we return so we avoid the
727 * unnecessary software interrrupt.
728 */
729}
730
731extern struct domain inetdomain;
732static struct protosw mroute_encap_protosw =
733{ SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR,
734 mroute_encap_input, 0, 0, rip_ctloutput,
735 0,
736 0, 0, 0, 0,
737 &rip_usrreqs
738};
739
740/*
741 * Add a vif to the vif table
742 */
743static int
744add_vif(vifcp)
745 register struct vifctl *vifcp;
746{
747 register struct vif *vifp = viftable + vifcp->vifc_vifi;
748 static struct sockaddr_in sin = {sizeof sin, AF_INET};
749 struct ifaddr *ifa;
750 struct ifnet *ifp;
751 int error, s;
752 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
753
754 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL;
755 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE;
756
757 /* Find the interface with an address in AF_INET family */
758 sin.sin_addr = vifcp->vifc_lcl_addr;
759 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
760 if (ifa == 0) return EADDRNOTAVAIL;
761 ifp = ifa->ifa_ifp;
762
763 if (vifcp->vifc_flags & VIFF_TUNNEL) {
764 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
765 /*
766 * An encapsulating tunnel is wanted. Tell
767 * mroute_encap_input() to start paying attention
768 * to encapsulated packets.
769 */
770 if (encap_cookie == NULL) {
771 encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
772 mroute_encapcheck,
773 (struct protosw *)&mroute_encap_protosw, NULL);
774
775 if (encap_cookie == NULL) {
776 printf("ip_mroute: unable to attach encap\n");
777 return (EIO); /* XXX */
778 }
779 for (s = 0; s < MAXVIFS; ++s) {
780 multicast_decap_if[s].if_name = "mdecap";
781 multicast_decap_if[s].if_unit = s;
782 }
783 }
784 /*
785 * Set interface to fake encapsulator interface
786 */
787 ifp = &multicast_decap_if[vifcp->vifc_vifi];
788 /*
789 * Prepare cached route entry
790 */
791 bzero(&vifp->v_route, sizeof(vifp->v_route));
792 } else {
793 log(LOG_ERR, "source routed tunnels not supported\n");
794 return EOPNOTSUPP;
795 }
796 } else {
797 /* Make sure the interface supports multicast */
798 if ((ifp->if_flags & IFF_MULTICAST) == 0)
799 return EOPNOTSUPP;
800
801 /* Enable promiscuous reception of all IP multicasts from the if */
802 s = splnet();
803 error = if_allmulti(ifp, 1);
804 splx(s);
805 if (error)
806 return error;
807 }
808
809 s = splnet();
810 /* define parameters for the tbf structure */
811 vifp->v_tbf = v_tbf;
812 GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
813 vifp->v_tbf->tbf_n_tok = 0;
814 vifp->v_tbf->tbf_q_len = 0;
815 vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
816 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
817
818 vifp->v_flags = vifcp->vifc_flags;
819 vifp->v_threshold = vifcp->vifc_threshold;
820 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
821 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
822 vifp->v_ifp = ifp;
823 /* scaling up here allows division by 1024 in critical code */
824 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
825 vifp->v_rsvp_on = 0;
826 vifp->v_rsvpd = NULL;
827 /* initialize per vif pkt counters */
828 vifp->v_pkt_in = 0;
829 vifp->v_pkt_out = 0;
830 vifp->v_bytes_in = 0;
831 vifp->v_bytes_out = 0;
832 splx(s);
833
834 /* Adjust numvifs up if the vifi is higher than numvifs */
835 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
836
837 if (mrtdebug)
838 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n",
839 vifcp->vifc_vifi,
840 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr),
841 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
842 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr),
843 vifcp->vifc_threshold,
844 vifcp->vifc_rate_limit);
845
846 return 0;
847}
848
849/*
850 * Delete a vif from the vif table
851 */
852static int
853del_vif(vifi)
854 vifi_t vifi;
855{
856 register struct vif *vifp = &viftable[vifi];
857 register struct mbuf *m;
858 struct ifnet *ifp;
859 struct ifreq ifr;
860 int s;
861
862 if (vifi >= numvifs) return EINVAL;
863 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL;
864
865 s = splnet();
866
867 if (!(vifp->v_flags & VIFF_TUNNEL)) {
868 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
869 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
870 ifp = vifp->v_ifp;
871 if_allmulti(ifp, 0);
872 }
873
874 if (vifp == last_encap_vif) {
875 last_encap_vif = 0;
876 last_encap_src = 0;
877 }
878
879 /*
880 * Free packets queued at the interface
881 */
882 while (vifp->v_tbf->tbf_q) {
883 m = vifp->v_tbf->tbf_q;
884 vifp->v_tbf->tbf_q = m->m_act;
885 m_freem(m);
886 }
887
888 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
889 bzero((caddr_t)vifp, sizeof (*vifp));
890
891 if (mrtdebug)
892 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs);
893
894 /* Adjust numvifs down */
895 for (vifi = numvifs; vifi > 0; vifi--)
896 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break;
897 numvifs = vifi;
898
899 splx(s);
900
901 return 0;
902}
903
904/*
905 * Add an mfc entry
906 */
907static int
908add_mfc(mfccp)
909 struct mfcctl *mfccp;
910{
911 struct mfc *rt;
912 u_long hash;
913 struct rtdetq *rte;
914 register u_short nstl;
915 int s;
916 int i;
917
918 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt);
919
920 /* If an entry already exists, just update the fields */
921 if (rt) {
922 if (mrtdebug & DEBUG_MFC)
923 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n",
924 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
925 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
926 mfccp->mfcc_parent);
927
928 s = splnet();
929 rt->mfc_parent = mfccp->mfcc_parent;
930 for (i = 0; i < numvifs; i++)
931 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
932 splx(s);
933 return 0;
934 }
935
936 /*
937 * Find the entry for which the upcall was made and update
938 */
939 s = splnet();
940 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
941 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) {
942
943 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
944 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
945 (rt->mfc_stall != NULL)) {
946
947 if (nstl++)
948 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n",
949 "multiple kernel entries",
950 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
951 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
952 mfccp->mfcc_parent, (void *)rt->mfc_stall);
953
954 if (mrtdebug & DEBUG_MFC)
955 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n",
956 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
957 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
958 mfccp->mfcc_parent, (void *)rt->mfc_stall);
959
960 rt->mfc_origin = mfccp->mfcc_origin;
961 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
962 rt->mfc_parent = mfccp->mfcc_parent;
963 for (i = 0; i < numvifs; i++)
964 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
965 /* initialize pkt counters per src-grp */
966 rt->mfc_pkt_cnt = 0;
967 rt->mfc_byte_cnt = 0;
968 rt->mfc_wrong_if = 0;
969 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
970
971 rt->mfc_expire = 0; /* Don't clean this guy up */
972 nexpire[hash]--;
973
974 /* free packets Qed at the end of this entry */
975 for (rte = rt->mfc_stall; rte != NULL; ) {
976 struct rtdetq *n = rte->next;
977
978 ip_mdq(rte->m, rte->ifp, rt, -1);
979 m_freem(rte->m);
980#ifdef UPCALL_TIMING
981 collate(&(rte->t));
982#endif /* UPCALL_TIMING */
983 free(rte, M_MRTABLE);
984 rte = n;
985 }
986 rt->mfc_stall = NULL;
987 }
988 }
989
990 /*
991 * It is possible that an entry is being inserted without an upcall
992 */
993 if (nstl == 0) {
994 if (mrtdebug & DEBUG_MFC)
995 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n",
996 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr),
997 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
998 mfccp->mfcc_parent);
999
1000 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) {
1001
1002 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1003 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
1004
1005 rt->mfc_origin = mfccp->mfcc_origin;
1006 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1007 rt->mfc_parent = mfccp->mfcc_parent;
1008 for (i = 0; i < numvifs; i++)
1009 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1010 /* initialize pkt counters per src-grp */
1011 rt->mfc_pkt_cnt = 0;
1012 rt->mfc_byte_cnt = 0;
1013 rt->mfc_wrong_if = 0;
1014 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
1015 if (rt->mfc_expire)
1016 nexpire[hash]--;
1017 rt->mfc_expire = 0;
1018 }
1019 }
1020 if (rt == NULL) {
1021 /* no upcall, so make a new entry */
1022 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1023 if (rt == NULL) {
1024 splx(s);
1025 return ENOBUFS;
1026 }
1027
1028 /* insert new entry at head of hash chain */
1029 rt->mfc_origin = mfccp->mfcc_origin;
1030 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1031 rt->mfc_parent = mfccp->mfcc_parent;
1032 for (i = 0; i < numvifs; i++)
1033 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1034 /* initialize pkt counters per src-grp */
1035 rt->mfc_pkt_cnt = 0;
1036 rt->mfc_byte_cnt = 0;
1037 rt->mfc_wrong_if = 0;
1038 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
1039 rt->mfc_expire = 0;
1040 rt->mfc_stall = NULL;
1041
1042 /* link into table */
1043 rt->mfc_next = mfctable[hash];
1044 mfctable[hash] = rt;
1045 }
1046 }
1047 splx(s);
1048 return 0;
1049}
1050
1051#ifdef UPCALL_TIMING
1052/*
1053 * collect delay statistics on the upcalls
1054 */
1055static void collate(t)
1056register struct timeval *t;
1057{
1058 register u_long d;
1059 register struct timeval tp;
1060 register u_long delta;
1061
1062 GET_TIME(tp);
1063
1064 if (TV_LT(*t, tp))
1065 {
1066 TV_DELTA(tp, *t, delta);
1067
1068 d = delta >> 10;
1069 if (d > 50)
1070 d = 50;
1071
1072 ++upcall_data[d];
1073 }
1074}
1075#endif /* UPCALL_TIMING */
1076
1077/*
1078 * Delete an mfc entry
1079 */
1080static int
1081del_mfc(mfccp)
1082 struct mfcctl *mfccp;
1083{
1084 struct in_addr origin;
1085 struct in_addr mcastgrp;
1086 struct mfc *rt;
1087 struct mfc **nptr;
1088 u_long hash;
1089 int s;
1090
1091 origin = mfccp->mfcc_origin;
1092 mcastgrp = mfccp->mfcc_mcastgrp;
1093 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1094
1095 if (mrtdebug & DEBUG_MFC)
1096 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n",
1097 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1098
1099 s = splnet();
1100
1101 nptr = &mfctable[hash];
1102 while ((rt = *nptr) != NULL) {
1103 if (origin.s_addr == rt->mfc_origin.s_addr &&
1104 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1105 rt->mfc_stall == NULL)
1106 break;
1107
1108 nptr = &rt->mfc_next;
1109 }
1110 if (rt == NULL) {
1111 splx(s);
1112 return EADDRNOTAVAIL;
1113 }
1114
1115 *nptr = rt->mfc_next;
1116 free(rt, M_MRTABLE);
1117
1118 splx(s);
1119
1120 return 0;
1121}
1122
1123/*
1124 * Send a message to mrouted on the multicast routing socket
1125 */
1126static int
1127socket_send(s, mm, src)
1128 struct socket *s;
1129 struct mbuf *mm;
1130 struct sockaddr_in *src;
1131{
1132 if (s) {
1133 if (sbappendaddr(&s->so_rcv,
1134 (struct sockaddr *)src,
1135 mm, (struct mbuf *)0) != 0) {
1136 sorwakeup(s);
1137 return 0;
1138 }
1139 }
1140 m_freem(mm);
1141 return -1;
1142}
1143
1144/*
1145 * IP multicast forwarding function. This function assumes that the packet
1146 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1147 * pointed to by "ifp", and the packet is to be relayed to other networks
1148 * that have members of the packet's destination IP multicast group.
1149 *
1150 * The packet is returned unscathed to the caller, unless it is
1151 * erroneous, in which case a non-zero return value tells the caller to
1152 * discard it.
1153 */
1154
1155#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1156
1157static int
1158X_ip_mforward(ip, ifp, m, imo)
1159 register struct ip *ip;
1160 struct ifnet *ifp;
1161 struct mbuf *m;
1162 struct ip_moptions *imo;
1163{
1164 register struct mfc *rt;
1165 register u_char *ipoptions;
1166 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1167 static int srctun = 0;
1168 register struct mbuf *mm;
1169 int s;
1170 vifi_t vifi;
1171 struct vif *vifp;
1172
1173 if (mrtdebug & DEBUG_FORWARD)
1174 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n",
1175 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr),
1176 (void *)ifp);
1177
1178 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1179 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1180 /*
1181 * Packet arrived via a physical interface or
1182 * an encapsulated tunnel.
1183 */
1184 } else {
1185 /*
1186 * Packet arrived through a source-route tunnel.
1187 * Source-route tunnels are no longer supported.
1188 */
1189 if ((srctun++ % 1000) == 0)
1190 log(LOG_ERR,
1191 "ip_mforward: received source-routed packet from %lx\n",
1192 (u_long)ntohl(ip->ip_src.s_addr));
1193
1194 return 1;
1195 }
1196
1197 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1198 if (ip->ip_ttl < 255)
1199 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1200 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1201 vifp = viftable + vifi;
1202 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n",
1203 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr),
1204 vifi,
1205 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1206 vifp->v_ifp->if_name, vifp->v_ifp->if_unit);
1207 }
1208 return (ip_mdq(m, ifp, NULL, vifi));
1209 }
1210 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1211 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1212 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr));
1213 if(!imo)
1214 printf("In fact, no options were specified at all\n");
1215 }
1216
1217 /*
1218 * Don't forward a packet with time-to-live of zero or one,
1219 * or a packet destined to a local-only group.
1220 */
1221 if (ip->ip_ttl <= 1 ||
1222 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
1223 return 0;
1224
1225 /*
1226 * Determine forwarding vifs from the forwarding cache table
1227 */
1228 s = splnet();
1229 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt);
1230
1231 /* Entry exists, so forward if necessary */
1232 if (rt != NULL) {
1233 splx(s);
1234 return (ip_mdq(m, ifp, rt, -1));
1235 } else {
1236 /*
1237 * If we don't have a route for packet's origin,
1238 * Make a copy of the packet &
1239 * send message to routing daemon
1240 */
1241
1242 register struct mbuf *mb0;
1243 register struct rtdetq *rte;
1244 register u_long hash;
1245 int hlen = ip->ip_hl << 2;
1246#ifdef UPCALL_TIMING
1247 struct timeval tp;
1248
1249 GET_TIME(tp);
1250#endif
1251
1252 mrtstat.mrts_no_route++;
1253 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1254 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n",
1255 (u_long)ntohl(ip->ip_src.s_addr),
1256 (u_long)ntohl(ip->ip_dst.s_addr));
1257
1258 /*
1259 * Allocate mbufs early so that we don't do extra work if we are
1260 * just going to fail anyway. Make sure to pullup the header so
1261 * that other people can't step on it.
1262 */
1263 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT);
1264 if (rte == NULL) {
1265 splx(s);
1266 return ENOBUFS;
1267 }
1268 mb0 = m_copy(m, 0, M_COPYALL);
1269 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1270 mb0 = m_pullup(mb0, hlen);
1271 if (mb0 == NULL) {
1272 free(rte, M_MRTABLE);
1273 splx(s);
1274 return ENOBUFS;
1275 }
1276
1277 /* is there an upcall waiting for this packet? */
1278 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1279 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) {
1280 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1281 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1282 (rt->mfc_stall != NULL))
1283 break;
1284 }
1285
1286 if (rt == NULL) {
1287 int i;
1288 struct igmpmsg *im;
1289
1290 /* no upcall, so make a new entry */
1291 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1292 if (rt == NULL) {
1293 free(rte, M_MRTABLE);
1294 m_freem(mb0);
1295 splx(s);
1296 return ENOBUFS;
1297 }
1298 /* Make a copy of the header to send to the user level process */
1299 mm = m_copy(mb0, 0, hlen);
1300 if (mm == NULL) {
1301 free(rte, M_MRTABLE);
1302 m_freem(mb0);
1303 free(rt, M_MRTABLE);
1304 splx(s);
1305 return ENOBUFS;
1306 }
1307
1308 /*
1309 * Send message to routing daemon to install
1310 * a route into the kernel table
1311 */
1312 k_igmpsrc.sin_addr = ip->ip_src;
1313
1314 im = mtod(mm, struct igmpmsg *);
1315 im->im_msgtype = IGMPMSG_NOCACHE;
1316 im->im_mbz = 0;
1317
1318 mrtstat.mrts_upcalls++;
1319
1320 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1321 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1322 ++mrtstat.mrts_upq_sockfull;
1323 free(rte, M_MRTABLE);
1324 m_freem(mb0);
1325 free(rt, M_MRTABLE);
1326 splx(s);
1327 return ENOBUFS;
1328 }
1329
1330 /* insert new entry at head of hash chain */
1331 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1332 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1333 rt->mfc_expire = UPCALL_EXPIRE;
1334 nexpire[hash]++;
1335 for (i = 0; i < numvifs; i++)
1336 rt->mfc_ttls[i] = 0;
1337 rt->mfc_parent = -1;
1338
1339 /* link into table */
1340 rt->mfc_next = mfctable[hash];
1341 mfctable[hash] = rt;
1342 rt->mfc_stall = rte;
1343
1344 } else {
1345 /* determine if q has overflowed */
1346 int npkts = 0;
1347 struct rtdetq **p;
1348
1349 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1350 npkts++;
1351
1352 if (npkts > MAX_UPQ) {
1353 mrtstat.mrts_upq_ovflw++;
1354 free(rte, M_MRTABLE);
1355 m_freem(mb0);
1356 splx(s);
1357 return 0;
1358 }
1359
1360 /* Add this entry to the end of the queue */
1361 *p = rte;
1362 }
1363
1364 rte->m = mb0;
1365 rte->ifp = ifp;
1366#ifdef UPCALL_TIMING
1367 rte->t = tp;
1368#endif
1369 rte->next = NULL;
1370
1371 splx(s);
1372
1373 return 0;
1374 }
1375}
1376
1377#ifndef MROUTE_KLD
1378int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
1379 struct ip_moptions *) = X_ip_mforward;
1380#endif
1381
1382/*
1383 * Clean up the cache entry if upcall is not serviced
1384 */
1385static void
1386expire_upcalls(void *unused)
1387{
1388 struct rtdetq *rte;
1389 struct mfc *mfc, **nptr;
1390 int i;
1391 int s;
1392
1393 s = splnet();
1394 for (i = 0; i < MFCTBLSIZ; i++) {
1395 if (nexpire[i] == 0)
1396 continue;
1397 nptr = &mfctable[i];
1398 for (mfc = *nptr; mfc != NULL; mfc = *nptr) {
1399 /*
1400 * Skip real cache entries
1401 * Make sure it wasn't marked to not expire (shouldn't happen)
1402 * If it expires now
1403 */
1404 if (mfc->mfc_stall != NULL &&
1405 mfc->mfc_expire != 0 &&
1406 --mfc->mfc_expire == 0) {
1407 if (mrtdebug & DEBUG_EXPIRE)
1408 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n",
1409 (u_long)ntohl(mfc->mfc_origin.s_addr),
1410 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr));
1411 /*
1412 * drop all the packets
1413 * free the mbuf with the pkt, if, timing info
1414 */
1415 for (rte = mfc->mfc_stall; rte; ) {
1416 struct rtdetq *n = rte->next;
1417
1418 m_freem(rte->m);
1419 free(rte, M_MRTABLE);
1420 rte = n;
1421 }
1422 ++mrtstat.mrts_cache_cleanups;
1423 nexpire[i]--;
1424
1425 *nptr = mfc->mfc_next;
1426 free(mfc, M_MRTABLE);
1427 } else {
1428 nptr = &mfc->mfc_next;
1429 }
1430 }
1431 }
1432 splx(s);
1433 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
1434}
1435
1436/*
1437 * Packet forwarding routine once entry in the cache is made
1438 */
1439static int
1440ip_mdq(m, ifp, rt, xmt_vif)
1441 register struct mbuf *m;
1442 register struct ifnet *ifp;
1443 register struct mfc *rt;
1444 register vifi_t xmt_vif;
1445{
1446 register struct ip *ip = mtod(m, struct ip *);
1447 register vifi_t vifi;
1448 register struct vif *vifp;
1449 register int plen = ip->ip_len;
1450
1451/*
1452 * Macro to send packet on vif. Since RSVP packets don't get counted on
1453 * input, they shouldn't get counted on output, so statistics keeping is
1454 * separate.
1455 */
1456#define MC_SEND(ip,vifp,m) { \
1457 if ((vifp)->v_flags & VIFF_TUNNEL) \
1458 encap_send((ip), (vifp), (m)); \
1459 else \
1460 phyint_send((ip), (vifp), (m)); \
1461}
1462
1463 /*
1464 * If xmt_vif is not -1, send on only the requested vif.
1465 *
1466 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1467 */
1468 if (xmt_vif < numvifs) {
1469 MC_SEND(ip, viftable + xmt_vif, m);
1470 return 1;
1471 }
1472
1473 /*
1474 * Don't forward if it didn't arrive from the parent vif for its origin.
1475 */
1476 vifi = rt->mfc_parent;
1477 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1478 /* came in the wrong interface */
1479 if (mrtdebug & DEBUG_FORWARD)
1480 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1481 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp);
1482 ++mrtstat.mrts_wrong_if;
1483 ++rt->mfc_wrong_if;
1484 /*
1485 * If we are doing PIM assert processing, and we are forwarding
1486 * packets on this interface, and it is a broadcast medium
1487 * interface (and not a tunnel), send a message to the routing daemon.
1488 */
1489 if (pim_assert && rt->mfc_ttls[vifi] &&
1490 (ifp->if_flags & IFF_BROADCAST) &&
1491 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1492 struct sockaddr_in k_igmpsrc;
1493 struct mbuf *mm;
1494 struct igmpmsg *im;
1495 int hlen = ip->ip_hl << 2;
1496 struct timeval now;
1497 register u_long delta;
1498
1499 GET_TIME(now);
1500
1501 TV_DELTA(rt->mfc_last_assert, now, delta);
1502
1503 if (delta > ASSERT_MSG_TIME) {
1504 mm = m_copy(m, 0, hlen);
1505 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1506 mm = m_pullup(mm, hlen);
1507 if (mm == NULL) {
1508 return ENOBUFS;
1509 }
1510
1511 rt->mfc_last_assert = now;
1512
1513 im = mtod(mm, struct igmpmsg *);
1514 im->im_msgtype = IGMPMSG_WRONGVIF;
1515 im->im_mbz = 0;
1516 im->im_vif = vifi;
1517
1518 k_igmpsrc.sin_addr = im->im_src;
1519
1520 socket_send(ip_mrouter, mm, &k_igmpsrc);
1521 }
1522 }
1523 return 0;
1524 }
1525
1526 /* If I sourced this packet, it counts as output, else it was input. */
1527 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1528 viftable[vifi].v_pkt_out++;
1529 viftable[vifi].v_bytes_out += plen;
1530 } else {
1531 viftable[vifi].v_pkt_in++;
1532 viftable[vifi].v_bytes_in += plen;
1533 }
1534 rt->mfc_pkt_cnt++;
1535 rt->mfc_byte_cnt += plen;
1536
1537 /*
1538 * For each vif, decide if a copy of the packet should be forwarded.
1539 * Forward if:
1540 * - the ttl exceeds the vif's threshold
1541 * - there are group members downstream on interface
1542 */
1543 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1544 if ((rt->mfc_ttls[vifi] > 0) &&
1545 (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1546 vifp->v_pkt_out++;
1547 vifp->v_bytes_out += plen;
1548 MC_SEND(ip, vifp, m);
1549 }
1550
1551 return 0;
1552}
1553
1554/*
1555 * check if a vif number is legal/ok. This is used by ip_output, to export
1556 * numvifs there,
1557 */
1558static int
1559X_legal_vif_num(vif)
1560 int vif;
1561{
1562 if (vif >= 0 && vif < numvifs)
1563 return(1);
1564 else
1565 return(0);
1566}
1567
1568#ifndef MROUTE_KLD
1569int (*legal_vif_num)(int) = X_legal_vif_num;
1570#endif
1571
1572/*
1573 * Return the local address used by this vif
1574 */
1575static u_long
1576X_ip_mcast_src(vifi)
1577 int vifi;
1578{
1579 if (vifi >= 0 && vifi < numvifs)
1580 return viftable[vifi].v_lcl_addr.s_addr;
1581 else
1582 return INADDR_ANY;
1583}
1584
1585#ifndef MROUTE_KLD
1586u_long (*ip_mcast_src)(int) = X_ip_mcast_src;
1587#endif
1588
1589static void
1590phyint_send(ip, vifp, m)
1591 struct ip *ip;
1592 struct vif *vifp;
1593 struct mbuf *m;
1594{
1595 register struct mbuf *mb_copy;
1596 register int hlen = ip->ip_hl << 2;
1597
1598 /*
1599 * Make a new reference to the packet; make sure that
1600 * the IP header is actually copied, not just referenced,
1601 * so that ip_output() only scribbles on the copy.
1602 */
1603 mb_copy = m_copy(m, 0, M_COPYALL);
1604 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1605 mb_copy = m_pullup(mb_copy, hlen);
1606 if (mb_copy == NULL)
1607 return;
1608
1609 if (vifp->v_rate_limit == 0)
1610 tbf_send_packet(vifp, mb_copy);
1611 else
1612 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1613}
1614
1615static void
1616encap_send(ip, vifp, m)
1617 register struct ip *ip;
1618 register struct vif *vifp;
1619 register struct mbuf *m;
1620{
1621 register struct mbuf *mb_copy;
1622 register struct ip *ip_copy;
1623 register int i, len = ip->ip_len;
1624
1625 /*
1626 * copy the old packet & pullup its IP header into the
1627 * new mbuf so we can modify it. Try to fill the new
1628 * mbuf since if we don't the ethernet driver will.
1629 */
1630 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER);
1631 if (mb_copy == NULL)
1632 return;
1633 mb_copy->m_data += max_linkhdr;
1634 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1635
1636 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1637 m_freem(mb_copy);
1638 return;
1639 }
1640 i = MHLEN - M_LEADINGSPACE(mb_copy);
1641 if (i > len)
1642 i = len;
1643 mb_copy = m_pullup(mb_copy, i);
1644 if (mb_copy == NULL)
1645 return;
1646 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1647
1648 /*
1649 * fill in the encapsulating IP header.
1650 */
1651 ip_copy = mtod(mb_copy, struct ip *);
1652 *ip_copy = multicast_encap_iphdr;
1653#ifdef RANDOM_IP_ID
1654 ip_copy->ip_id = ip_randomid();
1655#else
1656 ip_copy->ip_id = htons(ip_id++);
1657#endif
1658 ip_copy->ip_len += len;
1659 ip_copy->ip_src = vifp->v_lcl_addr;
1660 ip_copy->ip_dst = vifp->v_rmt_addr;
1661
1662 /*
1663 * turn the encapsulated IP header back into a valid one.
1664 */
1665 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1666 --ip->ip_ttl;
1667 ip->ip_len = htons(ip->ip_len);
1668 ip->ip_off = htons(ip->ip_off);
1669 ip->ip_sum = 0;
1670 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1671 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1672 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1673
1674 if (vifp->v_rate_limit == 0)
1675 tbf_send_packet(vifp, mb_copy);
1676 else
1677 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1678}
1679
1680/*
1681 * Token bucket filter module
1682 */
1683
1684static void
1685tbf_control(vifp, m, ip, p_len)
1686 register struct vif *vifp;
1687 register struct mbuf *m;
1688 register struct ip *ip;
1689 register u_long p_len;
1690{
1691 register struct tbf *t = vifp->v_tbf;
1692
1693 if (p_len > MAX_BKT_SIZE) {
1694 /* drop if packet is too large */
1695 mrtstat.mrts_pkt2large++;
1696 m_freem(m);
1697 return;
1698 }
1699
1700 tbf_update_tokens(vifp);
1701
1702 /* if there are enough tokens,
1703 * and the queue is empty,
1704 * send this packet out
1705 */
1706
1707 if (t->tbf_q_len == 0) {
1708 /* queue empty, send packet if enough tokens */
1709 if (p_len <= t->tbf_n_tok) {
1710 t->tbf_n_tok -= p_len;
1711 tbf_send_packet(vifp, m);
1712 } else {
1713 /* queue packet and timeout till later */
1714 tbf_queue(vifp, m);
1715 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1716 }
1717 } else if (t->tbf_q_len < t->tbf_max_q_len) {
1718 /* finite queue length, so queue pkts and process queue */
1719 tbf_queue(vifp, m);
1720 tbf_process_q(vifp);
1721 } else {
1722 /* queue length too much, try to dq and queue and process */
1723 if (!tbf_dq_sel(vifp, ip)) {
1724 mrtstat.mrts_q_overflow++;
1725 m_freem(m);
1726 return;
1727 } else {
1728 tbf_queue(vifp, m);
1729 tbf_process_q(vifp);
1730 }
1731 }
1732 return;
1733}
1734
1735/*
1736 * adds a packet to the queue at the interface
1737 */
1738static void
1739tbf_queue(vifp, m)
1740 register struct vif *vifp;
1741 register struct mbuf *m;
1742{
1743 register int s = splnet();
1744 register struct tbf *t = vifp->v_tbf;
1745
1746 if (t->tbf_t == NULL) {
1747 /* Queue was empty */
1748 t->tbf_q = m;
1749 } else {
1750 /* Insert at tail */
1751 t->tbf_t->m_act = m;
1752 }
1753
1754 /* Set new tail pointer */
1755 t->tbf_t = m;
1756
1757#ifdef DIAGNOSTIC
1758 /* Make sure we didn't get fed a bogus mbuf */
1759 if (m->m_act)
1760 panic("tbf_queue: m_act");
1761#endif
1762 m->m_act = NULL;
1763
1764 t->tbf_q_len++;
1765
1766 splx(s);
1767}
1768
1769
1770/*
1771 * processes the queue at the interface
1772 */
1773static void
1774tbf_process_q(vifp)
1775 register struct vif *vifp;
1776{
1777 register struct mbuf *m;
1778 register int len;
1779 register int s = splnet();
1780 register struct tbf *t = vifp->v_tbf;
1781
1782 /* loop through the queue at the interface and send as many packets
1783 * as possible
1784 */
1785 while (t->tbf_q_len > 0) {
1786 m = t->tbf_q;
1787
1788 len = mtod(m, struct ip *)->ip_len;
1789
1790 /* determine if the packet can be sent */
1791 if (len <= t->tbf_n_tok) {
1792 /* if so,
1793 * reduce no of tokens, dequeue the packet,
1794 * send the packet.
1795 */
1796 t->tbf_n_tok -= len;
1797
1798 t->tbf_q = m->m_act;
1799 if (--t->tbf_q_len == 0)
1800 t->tbf_t = NULL;
1801
1802 m->m_act = NULL;
1803 tbf_send_packet(vifp, m);
1804
1805 } else break;
1806 }
1807 splx(s);
1808}
1809
1810static void
1811tbf_reprocess_q(xvifp)
1812 void *xvifp;
1813{
1814 register struct vif *vifp = xvifp;
1815 if (ip_mrouter == NULL)
1816 return;
1817
1818 tbf_update_tokens(vifp);
1819
1820 tbf_process_q(vifp);
1821
1822 if (vifp->v_tbf->tbf_q_len)
1823 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1824}
1825
1826/* function that will selectively discard a member of the queue
1827 * based on the precedence value and the priority
1828 */
1829static int
1830tbf_dq_sel(vifp, ip)
1831 register struct vif *vifp;
1832 register struct ip *ip;
1833{
1834 register int s = splnet();
1835 register u_int p;
1836 register struct mbuf *m, *last;
1837 register struct mbuf **np;
1838 register struct tbf *t = vifp->v_tbf;
1839
1840 p = priority(vifp, ip);
1841
1842 np = &t->tbf_q;
1843 last = NULL;
1844 while ((m = *np) != NULL) {
1845 if (p > priority(vifp, mtod(m, struct ip *))) {
1846 *np = m->m_act;
1847 /* If we're removing the last packet, fix the tail pointer */
1848 if (m == t->tbf_t)
1849 t->tbf_t = last;
1850 m_freem(m);
1851 /* it's impossible for the queue to be empty, but
1852 * we check anyway. */
1853 if (--t->tbf_q_len == 0)
1854 t->tbf_t = NULL;
1855 splx(s);
1856 mrtstat.mrts_drop_sel++;
1857 return(1);
1858 }
1859 np = &m->m_act;
1860 last = m;
1861 }
1862 splx(s);
1863 return(0);
1864}
1865
1866static void
1867tbf_send_packet(vifp, m)
1868 register struct vif *vifp;
1869 register struct mbuf *m;
1870{
1871 struct ip_moptions imo;
1872 int error;
1873 static struct route ro;
1874 int s = splnet();
1875
1876 if (vifp->v_flags & VIFF_TUNNEL) {
1877 /* If tunnel options */
1878 ip_output(m, (struct mbuf *)0, &vifp->v_route,
| 13 */
14
15#include "opt_mrouting.h"
16#include "opt_random_ip_id.h"
17
18#include <sys/param.h>
19#include <sys/kernel.h>
20#include <sys/lock.h>
21#include <sys/malloc.h>
22#include <sys/mbuf.h>
23#include <sys/protosw.h>
24#include <sys/signalvar.h>
25#include <sys/socket.h>
26#include <sys/socketvar.h>
27#include <sys/sockio.h>
28#include <sys/sx.h>
29#include <sys/sysctl.h>
30#include <sys/syslog.h>
31#include <sys/systm.h>
32#include <sys/time.h>
33#include <net/if.h>
34#include <net/route.h>
35#include <netinet/in.h>
36#include <netinet/igmp.h>
37#include <netinet/in_systm.h>
38#include <netinet/in_var.h>
39#include <netinet/ip.h>
40#include <netinet/ip_encap.h>
41#include <netinet/ip_mroute.h>
42#include <netinet/ip_var.h>
43#include <netinet/udp.h>
44#include <machine/in_cksum.h>
45
46#ifndef MROUTING
47extern u_long _ip_mcast_src(int vifi);
48extern int _ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
49 struct ip_moptions *imo);
50extern int _ip_mrouter_done(void);
51extern int _ip_mrouter_get(struct socket *so, struct sockopt *sopt);
52extern int _ip_mrouter_set(struct socket *so, struct sockopt *sopt);
53extern int _mrt_ioctl(int req, caddr_t data);
54
55/*
56 * Dummy routines and globals used when multicast routing is not compiled in.
57 */
58
59struct socket *ip_mrouter = NULL;
60u_int rsvpdebug = 0;
61
62int
63_ip_mrouter_set(so, sopt)
64 struct socket *so;
65 struct sockopt *sopt;
66{
67 return(EOPNOTSUPP);
68}
69
70int (*ip_mrouter_set)(struct socket *, struct sockopt *) = _ip_mrouter_set;
71
72
73int
74_ip_mrouter_get(so, sopt)
75 struct socket *so;
76 struct sockopt *sopt;
77{
78 return(EOPNOTSUPP);
79}
80
81int (*ip_mrouter_get)(struct socket *, struct sockopt *) = _ip_mrouter_get;
82
83int
84_ip_mrouter_done()
85{
86 return(0);
87}
88
89int (*ip_mrouter_done)(void) = _ip_mrouter_done;
90
91int
92_ip_mforward(ip, ifp, m, imo)
93 struct ip *ip;
94 struct ifnet *ifp;
95 struct mbuf *m;
96 struct ip_moptions *imo;
97{
98 return(0);
99}
100
101int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
102 struct ip_moptions *) = _ip_mforward;
103
104int
105_mrt_ioctl(int req, caddr_t data)
106{
107 return EOPNOTSUPP;
108}
109
110int (*mrt_ioctl)(int, caddr_t) = _mrt_ioctl;
111
112void
113rsvp_input(m, off) /* XXX must fixup manually */
114 struct mbuf *m;
115 int off;
116{
117 /* Can still get packets with rsvp_on = 0 if there is a local member
118 * of the group to which the RSVP packet is addressed. But in this
119 * case we want to throw the packet away.
120 */
121 if (!rsvp_on) {
122 m_freem(m);
123 return;
124 }
125
126 if (ip_rsvpd != NULL) {
127 if (rsvpdebug)
128 printf("rsvp_input: Sending packet up old-style socket\n");
129 rip_input(m, off);
130 return;
131 }
132 /* Drop the packet */
133 m_freem(m);
134}
135
136int (*legal_vif_num)(int) = 0;
137
138/*
139 * This should never be called, since IP_MULTICAST_VIF should fail, but
140 * just in case it does get called, the code a little lower in ip_output
141 * will assign the packet a local address.
142 */
143u_long
144_ip_mcast_src(int vifi) { return INADDR_ANY; }
145u_long (*ip_mcast_src)(int) = _ip_mcast_src;
146
147int
148ip_rsvp_vif_init(so, sopt)
149 struct socket *so;
150 struct sockopt *sopt;
151{
152 return(EINVAL);
153}
154
155int
156ip_rsvp_vif_done(so, sopt)
157 struct socket *so;
158 struct sockopt *sopt;
159{
160 return(EINVAL);
161}
162
163void
164ip_rsvp_force_done(so)
165 struct socket *so;
166{
167 return;
168}
169
170#else /* MROUTING */
171
172#define M_HASCL(m) ((m)->m_flags & M_EXT)
173
174static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables");
175
176#ifndef MROUTE_KLD
177/* The socket used to communicate with the multicast routing daemon. */
178struct socket *ip_mrouter = NULL;
179#endif
180
181#if defined(MROUTING) || defined(MROUTE_KLD)
182static struct mrtstat mrtstat;
183SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
184 &mrtstat, mrtstat, "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)");
185#endif
186
187static struct mfc *mfctable[MFCTBLSIZ];
188static u_char nexpire[MFCTBLSIZ];
189static struct vif viftable[MAXVIFS];
190static u_int mrtdebug = 0; /* debug level */
191#define DEBUG_MFC 0x02
192#define DEBUG_FORWARD 0x04
193#define DEBUG_EXPIRE 0x08
194#define DEBUG_XMIT 0x10
195static u_int tbfdebug = 0; /* tbf debug level */
196static u_int rsvpdebug = 0; /* rsvp debug level */
197
198static struct callout_handle expire_upcalls_ch;
199
200#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
201#define UPCALL_EXPIRE 6 /* number of timeouts */
202
203/*
204 * Define the token bucket filter structures
205 * tbftable -> each vif has one of these for storing info
206 */
207
208static struct tbf tbftable[MAXVIFS];
209#define TBF_REPROCESS (hz / 100) /* 100x / second */
210
211/*
212 * 'Interfaces' associated with decapsulator (so we can tell
213 * packets that went through it from ones that get reflected
214 * by a broken gateway). These interfaces are never linked into
215 * the system ifnet list & no routes point to them. I.e., packets
216 * can't be sent this way. They only exist as a placeholder for
217 * multicast source verification.
218 */
219static struct ifnet multicast_decap_if[MAXVIFS];
220
221#define ENCAP_TTL 64
222#define ENCAP_PROTO IPPROTO_IPIP /* 4 */
223
224/* prototype IP hdr for encapsulated packets */
225static struct ip multicast_encap_iphdr = {
226#if BYTE_ORDER == LITTLE_ENDIAN
227 sizeof(struct ip) >> 2, IPVERSION,
228#else
229 IPVERSION, sizeof(struct ip) >> 2,
230#endif
231 0, /* tos */
232 sizeof(struct ip), /* total length */
233 0, /* id */
234 0, /* frag offset */
235 ENCAP_TTL, ENCAP_PROTO,
236 0, /* checksum */
237};
238
239/*
240 * Private variables.
241 */
242static vifi_t numvifs = 0;
243static const struct encaptab *encap_cookie = NULL;
244
245/*
246 * one-back cache used by mroute_encapcheck to locate a tunnel's vif
247 * given a datagram's src ip address.
248 */
249static u_long last_encap_src;
250static struct vif *last_encap_vif;
251
252static u_long X_ip_mcast_src(int vifi);
253static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo);
254static int X_ip_mrouter_done(void);
255static int X_ip_mrouter_get(struct socket *so, struct sockopt *m);
256static int X_ip_mrouter_set(struct socket *so, struct sockopt *m);
257static int X_legal_vif_num(int vif);
258static int X_mrt_ioctl(int cmd, caddr_t data);
259
260static int get_sg_cnt(struct sioc_sg_req *);
261static int get_vif_cnt(struct sioc_vif_req *);
262static int ip_mrouter_init(struct socket *, int);
263static int add_vif(struct vifctl *);
264static int del_vif(vifi_t);
265static int add_mfc(struct mfcctl *);
266static int del_mfc(struct mfcctl *);
267static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
268static int set_assert(int);
269static void expire_upcalls(void *);
270static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *,
271 vifi_t);
272static void phyint_send(struct ip *, struct vif *, struct mbuf *);
273static void encap_send(struct ip *, struct vif *, struct mbuf *);
274static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
275static void tbf_queue(struct vif *, struct mbuf *);
276static void tbf_process_q(struct vif *);
277static void tbf_reprocess_q(void *);
278static int tbf_dq_sel(struct vif *, struct ip *);
279static void tbf_send_packet(struct vif *, struct mbuf *);
280static void tbf_update_tokens(struct vif *);
281static int priority(struct vif *, struct ip *);
282
283/*
284 * whether or not special PIM assert processing is enabled.
285 */
286static int pim_assert;
287/*
288 * Rate limit for assert notification messages, in usec
289 */
290#define ASSERT_MSG_TIME 3000000
291
292/*
293 * Hash function for a source, group entry
294 */
295#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
296 ((g) >> 20) ^ ((g) >> 10) ^ (g))
297
298/*
299 * Find a route for a given origin IP address and Multicast group address
300 * Type of service parameter to be added in the future!!!
301 */
302
303#define MFCFIND(o, g, rt) { \
304 register struct mfc *_rt = mfctable[MFCHASH(o,g)]; \
305 rt = NULL; \
306 ++mrtstat.mrts_mfc_lookups; \
307 while (_rt) { \
308 if ((_rt->mfc_origin.s_addr == o) && \
309 (_rt->mfc_mcastgrp.s_addr == g) && \
310 (_rt->mfc_stall == NULL)) { \
311 rt = _rt; \
312 break; \
313 } \
314 _rt = _rt->mfc_next; \
315 } \
316 if (rt == NULL) { \
317 ++mrtstat.mrts_mfc_misses; \
318 } \
319}
320
321
322/*
323 * Macros to compute elapsed time efficiently
324 * Borrowed from Van Jacobson's scheduling code
325 */
326#define TV_DELTA(a, b, delta) { \
327 register int xxs; \
328 \
329 delta = (a).tv_usec - (b).tv_usec; \
330 if ((xxs = (a).tv_sec - (b).tv_sec)) { \
331 switch (xxs) { \
332 case 2: \
333 delta += 1000000; \
334 /* FALLTHROUGH */ \
335 case 1: \
336 delta += 1000000; \
337 break; \
338 default: \
339 delta += (1000000 * xxs); \
340 } \
341 } \
342}
343
344#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
345 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
346
347#ifdef UPCALL_TIMING
348u_long upcall_data[51];
349static void collate(struct timeval *);
350#endif /* UPCALL_TIMING */
351
352
353/*
354 * Handle MRT setsockopt commands to modify the multicast routing tables.
355 */
356static int
357X_ip_mrouter_set(so, sopt)
358 struct socket *so;
359 struct sockopt *sopt;
360{
361 int error, optval;
362 vifi_t vifi;
363 struct vifctl vifc;
364 struct mfcctl mfc;
365
366 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT)
367 return (EPERM);
368
369 error = 0;
370 switch (sopt->sopt_name) {
371 case MRT_INIT:
372 error = sooptcopyin(sopt, &optval, sizeof optval,
373 sizeof optval);
374 if (error)
375 break;
376 error = ip_mrouter_init(so, optval);
377 break;
378
379 case MRT_DONE:
380 error = ip_mrouter_done();
381 break;
382
383 case MRT_ADD_VIF:
384 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
385 if (error)
386 break;
387 error = add_vif(&vifc);
388 break;
389
390 case MRT_DEL_VIF:
391 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
392 if (error)
393 break;
394 error = del_vif(vifi);
395 break;
396
397 case MRT_ADD_MFC:
398 case MRT_DEL_MFC:
399 error = sooptcopyin(sopt, &mfc, sizeof mfc, sizeof mfc);
400 if (error)
401 break;
402 if (sopt->sopt_name == MRT_ADD_MFC)
403 error = add_mfc(&mfc);
404 else
405 error = del_mfc(&mfc);
406 break;
407
408 case MRT_ASSERT:
409 error = sooptcopyin(sopt, &optval, sizeof optval,
410 sizeof optval);
411 if (error)
412 break;
413 set_assert(optval);
414 break;
415
416 default:
417 error = EOPNOTSUPP;
418 break;
419 }
420 return (error);
421}
422
423#ifndef MROUTE_KLD
424int (*ip_mrouter_set)(struct socket *, struct sockopt *) = X_ip_mrouter_set;
425#endif
426
427/*
428 * Handle MRT getsockopt commands
429 */
430static int
431X_ip_mrouter_get(so, sopt)
432 struct socket *so;
433 struct sockopt *sopt;
434{
435 int error;
436 static int version = 0x0305; /* !!! why is this here? XXX */
437
438 switch (sopt->sopt_name) {
439 case MRT_VERSION:
440 error = sooptcopyout(sopt, &version, sizeof version);
441 break;
442
443 case MRT_ASSERT:
444 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert);
445 break;
446 default:
447 error = EOPNOTSUPP;
448 break;
449 }
450 return (error);
451}
452
453#ifndef MROUTE_KLD
454int (*ip_mrouter_get)(struct socket *, struct sockopt *) = X_ip_mrouter_get;
455#endif
456
457/*
458 * Handle ioctl commands to obtain information from the cache
459 */
460static int
461X_mrt_ioctl(cmd, data)
462 int cmd;
463 caddr_t data;
464{
465 int error = 0;
466
467 switch (cmd) {
468 case (SIOCGETVIFCNT):
469 return (get_vif_cnt((struct sioc_vif_req *)data));
470 break;
471 case (SIOCGETSGCNT):
472 return (get_sg_cnt((struct sioc_sg_req *)data));
473 break;
474 default:
475 return (EINVAL);
476 break;
477 }
478 return error;
479}
480
481#ifndef MROUTE_KLD
482int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl;
483#endif
484
485/*
486 * returns the packet, byte, rpf-failure count for the source group provided
487 */
488static int
489get_sg_cnt(req)
490 register struct sioc_sg_req *req;
491{
492 register struct mfc *rt;
493 int s;
494
495 s = splnet();
496 MFCFIND(req->src.s_addr, req->grp.s_addr, rt);
497 splx(s);
498 if (rt != NULL) {
499 req->pktcnt = rt->mfc_pkt_cnt;
500 req->bytecnt = rt->mfc_byte_cnt;
501 req->wrong_if = rt->mfc_wrong_if;
502 } else
503 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
504
505 return 0;
506}
507
508/*
509 * returns the input and output packet and byte counts on the vif provided
510 */
511static int
512get_vif_cnt(req)
513 register struct sioc_vif_req *req;
514{
515 register vifi_t vifi = req->vifi;
516
517 if (vifi >= numvifs) return EINVAL;
518
519 req->icount = viftable[vifi].v_pkt_in;
520 req->ocount = viftable[vifi].v_pkt_out;
521 req->ibytes = viftable[vifi].v_bytes_in;
522 req->obytes = viftable[vifi].v_bytes_out;
523
524 return 0;
525}
526
527/*
528 * Enable multicast routing
529 */
530static int
531ip_mrouter_init(so, version)
532 struct socket *so;
533 int version;
534{
535 if (mrtdebug)
536 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
537 so->so_type, so->so_proto->pr_protocol);
538
539 if (so->so_type != SOCK_RAW ||
540 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP;
541
542 if (version != 1)
543 return ENOPROTOOPT;
544
545 if (ip_mrouter != NULL) return EADDRINUSE;
546
547 ip_mrouter = so;
548
549 bzero((caddr_t)mfctable, sizeof(mfctable));
550 bzero((caddr_t)nexpire, sizeof(nexpire));
551
552 pim_assert = 0;
553
554 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
555
556 if (mrtdebug)
557 log(LOG_DEBUG, "ip_mrouter_init\n");
558
559 return 0;
560}
561
562/*
563 * Disable multicast routing
564 */
565static int
566X_ip_mrouter_done()
567{
568 vifi_t vifi;
569 int i;
570 struct ifnet *ifp;
571 struct ifreq ifr;
572 struct mfc *rt;
573 struct rtdetq *rte;
574 int s;
575
576 s = splnet();
577
578 /*
579 * For each phyint in use, disable promiscuous reception of all IP
580 * multicasts.
581 */
582 for (vifi = 0; vifi < numvifs; vifi++) {
583 if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
584 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
585 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
586 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr
587 = INADDR_ANY;
588 ifp = viftable[vifi].v_ifp;
589 if_allmulti(ifp, 0);
590 }
591 }
592 bzero((caddr_t)tbftable, sizeof(tbftable));
593 bzero((caddr_t)viftable, sizeof(viftable));
594 numvifs = 0;
595 pim_assert = 0;
596
597 untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch);
598
599 /*
600 * Free all multicast forwarding cache entries.
601 */
602 for (i = 0; i < MFCTBLSIZ; i++) {
603 for (rt = mfctable[i]; rt != NULL; ) {
604 struct mfc *nr = rt->mfc_next;
605
606 for (rte = rt->mfc_stall; rte != NULL; ) {
607 struct rtdetq *n = rte->next;
608
609 m_freem(rte->m);
610 free(rte, M_MRTABLE);
611 rte = n;
612 }
613 free(rt, M_MRTABLE);
614 rt = nr;
615 }
616 }
617
618 bzero((caddr_t)mfctable, sizeof(mfctable));
619
620 /*
621 * Reset de-encapsulation cache
622 */
623 last_encap_src = 0;
624 last_encap_vif = NULL;
625 if (encap_cookie) {
626 encap_detach(encap_cookie);
627 encap_cookie = NULL;
628 }
629
630 ip_mrouter = NULL;
631
632 splx(s);
633
634 if (mrtdebug)
635 log(LOG_DEBUG, "ip_mrouter_done\n");
636
637 return 0;
638}
639
640#ifndef MROUTE_KLD
641int (*ip_mrouter_done)(void) = X_ip_mrouter_done;
642#endif
643
644/*
645 * Set PIM assert processing global
646 */
647static int
648set_assert(i)
649 int i;
650{
651 if ((i != 1) && (i != 0))
652 return EINVAL;
653
654 pim_assert = i;
655
656 return 0;
657}
658
659/*
660 * Decide if a packet is from a tunnelled peer.
661 * Return 0 if not, 64 if so.
662 */
663static int
664mroute_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
665{
666 struct ip *ip = mtod(m, struct ip *);
667 int hlen = ip->ip_hl << 2;
668 register struct vif *vifp;
669
670 /*
671 * don't claim the packet if it's not to a multicast destination or if
672 * we don't have an encapsulating tunnel with the source.
673 * Note: This code assumes that the remote site IP address
674 * uniquely identifies the tunnel (i.e., that this site has
675 * at most one tunnel with the remote site).
676 */
677 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) {
678 return 0;
679 }
680 if (ip->ip_src.s_addr != last_encap_src) {
681 register struct vif *vife;
682
683 vifp = viftable;
684 vife = vifp + numvifs;
685 last_encap_src = ip->ip_src.s_addr;
686 last_encap_vif = 0;
687 for ( ; vifp < vife; ++vifp)
688 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
689 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT))
690 == VIFF_TUNNEL)
691 last_encap_vif = vifp;
692 break;
693 }
694 }
695 if ((vifp = last_encap_vif) == 0) {
696 last_encap_src = 0;
697 return 0;
698 }
699 return 64;
700}
701
702/*
703 * De-encapsulate a packet and feed it back through ip input (this
704 * routine is called whenever IP gets a packet that mroute_encap_func()
705 * claimed).
706 */
707static void
708mroute_encap_input(struct mbuf *m, int off)
709{
710 struct ip *ip = mtod(m, struct ip *);
711 int hlen = ip->ip_hl << 2;
712
713 if (hlen > sizeof(struct ip))
714 ip_stripoptions(m, (struct mbuf *) 0);
715 m->m_data += sizeof(struct ip);
716 m->m_len -= sizeof(struct ip);
717 m->m_pkthdr.len -= sizeof(struct ip);
718
719 m->m_pkthdr.rcvif = last_encap_vif->v_ifp;
720
721 (void) IF_HANDOFF(&ipintrq, m, NULL);
722 /*
723 * normally we would need a "schednetisr(NETISR_IP)"
724 * here but we were called by ip_input and it is going
725 * to loop back & try to dequeue the packet we just
726 * queued as soon as we return so we avoid the
727 * unnecessary software interrrupt.
728 */
729}
730
731extern struct domain inetdomain;
732static struct protosw mroute_encap_protosw =
733{ SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR,
734 mroute_encap_input, 0, 0, rip_ctloutput,
735 0,
736 0, 0, 0, 0,
737 &rip_usrreqs
738};
739
740/*
741 * Add a vif to the vif table
742 */
743static int
744add_vif(vifcp)
745 register struct vifctl *vifcp;
746{
747 register struct vif *vifp = viftable + vifcp->vifc_vifi;
748 static struct sockaddr_in sin = {sizeof sin, AF_INET};
749 struct ifaddr *ifa;
750 struct ifnet *ifp;
751 int error, s;
752 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
753
754 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL;
755 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE;
756
757 /* Find the interface with an address in AF_INET family */
758 sin.sin_addr = vifcp->vifc_lcl_addr;
759 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
760 if (ifa == 0) return EADDRNOTAVAIL;
761 ifp = ifa->ifa_ifp;
762
763 if (vifcp->vifc_flags & VIFF_TUNNEL) {
764 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
765 /*
766 * An encapsulating tunnel is wanted. Tell
767 * mroute_encap_input() to start paying attention
768 * to encapsulated packets.
769 */
770 if (encap_cookie == NULL) {
771 encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
772 mroute_encapcheck,
773 (struct protosw *)&mroute_encap_protosw, NULL);
774
775 if (encap_cookie == NULL) {
776 printf("ip_mroute: unable to attach encap\n");
777 return (EIO); /* XXX */
778 }
779 for (s = 0; s < MAXVIFS; ++s) {
780 multicast_decap_if[s].if_name = "mdecap";
781 multicast_decap_if[s].if_unit = s;
782 }
783 }
784 /*
785 * Set interface to fake encapsulator interface
786 */
787 ifp = &multicast_decap_if[vifcp->vifc_vifi];
788 /*
789 * Prepare cached route entry
790 */
791 bzero(&vifp->v_route, sizeof(vifp->v_route));
792 } else {
793 log(LOG_ERR, "source routed tunnels not supported\n");
794 return EOPNOTSUPP;
795 }
796 } else {
797 /* Make sure the interface supports multicast */
798 if ((ifp->if_flags & IFF_MULTICAST) == 0)
799 return EOPNOTSUPP;
800
801 /* Enable promiscuous reception of all IP multicasts from the if */
802 s = splnet();
803 error = if_allmulti(ifp, 1);
804 splx(s);
805 if (error)
806 return error;
807 }
808
809 s = splnet();
810 /* define parameters for the tbf structure */
811 vifp->v_tbf = v_tbf;
812 GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
813 vifp->v_tbf->tbf_n_tok = 0;
814 vifp->v_tbf->tbf_q_len = 0;
815 vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
816 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
817
818 vifp->v_flags = vifcp->vifc_flags;
819 vifp->v_threshold = vifcp->vifc_threshold;
820 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
821 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
822 vifp->v_ifp = ifp;
823 /* scaling up here allows division by 1024 in critical code */
824 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
825 vifp->v_rsvp_on = 0;
826 vifp->v_rsvpd = NULL;
827 /* initialize per vif pkt counters */
828 vifp->v_pkt_in = 0;
829 vifp->v_pkt_out = 0;
830 vifp->v_bytes_in = 0;
831 vifp->v_bytes_out = 0;
832 splx(s);
833
834 /* Adjust numvifs up if the vifi is higher than numvifs */
835 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
836
837 if (mrtdebug)
838 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n",
839 vifcp->vifc_vifi,
840 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr),
841 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
842 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr),
843 vifcp->vifc_threshold,
844 vifcp->vifc_rate_limit);
845
846 return 0;
847}
848
849/*
850 * Delete a vif from the vif table
851 */
852static int
853del_vif(vifi)
854 vifi_t vifi;
855{
856 register struct vif *vifp = &viftable[vifi];
857 register struct mbuf *m;
858 struct ifnet *ifp;
859 struct ifreq ifr;
860 int s;
861
862 if (vifi >= numvifs) return EINVAL;
863 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL;
864
865 s = splnet();
866
867 if (!(vifp->v_flags & VIFF_TUNNEL)) {
868 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
869 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
870 ifp = vifp->v_ifp;
871 if_allmulti(ifp, 0);
872 }
873
874 if (vifp == last_encap_vif) {
875 last_encap_vif = 0;
876 last_encap_src = 0;
877 }
878
879 /*
880 * Free packets queued at the interface
881 */
882 while (vifp->v_tbf->tbf_q) {
883 m = vifp->v_tbf->tbf_q;
884 vifp->v_tbf->tbf_q = m->m_act;
885 m_freem(m);
886 }
887
888 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
889 bzero((caddr_t)vifp, sizeof (*vifp));
890
891 if (mrtdebug)
892 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs);
893
894 /* Adjust numvifs down */
895 for (vifi = numvifs; vifi > 0; vifi--)
896 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break;
897 numvifs = vifi;
898
899 splx(s);
900
901 return 0;
902}
903
904/*
905 * Add an mfc entry
906 */
907static int
908add_mfc(mfccp)
909 struct mfcctl *mfccp;
910{
911 struct mfc *rt;
912 u_long hash;
913 struct rtdetq *rte;
914 register u_short nstl;
915 int s;
916 int i;
917
918 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt);
919
920 /* If an entry already exists, just update the fields */
921 if (rt) {
922 if (mrtdebug & DEBUG_MFC)
923 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n",
924 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
925 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
926 mfccp->mfcc_parent);
927
928 s = splnet();
929 rt->mfc_parent = mfccp->mfcc_parent;
930 for (i = 0; i < numvifs; i++)
931 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
932 splx(s);
933 return 0;
934 }
935
936 /*
937 * Find the entry for which the upcall was made and update
938 */
939 s = splnet();
940 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
941 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) {
942
943 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
944 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
945 (rt->mfc_stall != NULL)) {
946
947 if (nstl++)
948 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n",
949 "multiple kernel entries",
950 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
951 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
952 mfccp->mfcc_parent, (void *)rt->mfc_stall);
953
954 if (mrtdebug & DEBUG_MFC)
955 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n",
956 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
957 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
958 mfccp->mfcc_parent, (void *)rt->mfc_stall);
959
960 rt->mfc_origin = mfccp->mfcc_origin;
961 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
962 rt->mfc_parent = mfccp->mfcc_parent;
963 for (i = 0; i < numvifs; i++)
964 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
965 /* initialize pkt counters per src-grp */
966 rt->mfc_pkt_cnt = 0;
967 rt->mfc_byte_cnt = 0;
968 rt->mfc_wrong_if = 0;
969 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
970
971 rt->mfc_expire = 0; /* Don't clean this guy up */
972 nexpire[hash]--;
973
974 /* free packets Qed at the end of this entry */
975 for (rte = rt->mfc_stall; rte != NULL; ) {
976 struct rtdetq *n = rte->next;
977
978 ip_mdq(rte->m, rte->ifp, rt, -1);
979 m_freem(rte->m);
980#ifdef UPCALL_TIMING
981 collate(&(rte->t));
982#endif /* UPCALL_TIMING */
983 free(rte, M_MRTABLE);
984 rte = n;
985 }
986 rt->mfc_stall = NULL;
987 }
988 }
989
990 /*
991 * It is possible that an entry is being inserted without an upcall
992 */
993 if (nstl == 0) {
994 if (mrtdebug & DEBUG_MFC)
995 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n",
996 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr),
997 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
998 mfccp->mfcc_parent);
999
1000 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) {
1001
1002 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1003 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
1004
1005 rt->mfc_origin = mfccp->mfcc_origin;
1006 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1007 rt->mfc_parent = mfccp->mfcc_parent;
1008 for (i = 0; i < numvifs; i++)
1009 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1010 /* initialize pkt counters per src-grp */
1011 rt->mfc_pkt_cnt = 0;
1012 rt->mfc_byte_cnt = 0;
1013 rt->mfc_wrong_if = 0;
1014 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
1015 if (rt->mfc_expire)
1016 nexpire[hash]--;
1017 rt->mfc_expire = 0;
1018 }
1019 }
1020 if (rt == NULL) {
1021 /* no upcall, so make a new entry */
1022 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1023 if (rt == NULL) {
1024 splx(s);
1025 return ENOBUFS;
1026 }
1027
1028 /* insert new entry at head of hash chain */
1029 rt->mfc_origin = mfccp->mfcc_origin;
1030 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1031 rt->mfc_parent = mfccp->mfcc_parent;
1032 for (i = 0; i < numvifs; i++)
1033 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1034 /* initialize pkt counters per src-grp */
1035 rt->mfc_pkt_cnt = 0;
1036 rt->mfc_byte_cnt = 0;
1037 rt->mfc_wrong_if = 0;
1038 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
1039 rt->mfc_expire = 0;
1040 rt->mfc_stall = NULL;
1041
1042 /* link into table */
1043 rt->mfc_next = mfctable[hash];
1044 mfctable[hash] = rt;
1045 }
1046 }
1047 splx(s);
1048 return 0;
1049}
1050
1051#ifdef UPCALL_TIMING
1052/*
1053 * collect delay statistics on the upcalls
1054 */
1055static void collate(t)
1056register struct timeval *t;
1057{
1058 register u_long d;
1059 register struct timeval tp;
1060 register u_long delta;
1061
1062 GET_TIME(tp);
1063
1064 if (TV_LT(*t, tp))
1065 {
1066 TV_DELTA(tp, *t, delta);
1067
1068 d = delta >> 10;
1069 if (d > 50)
1070 d = 50;
1071
1072 ++upcall_data[d];
1073 }
1074}
1075#endif /* UPCALL_TIMING */
1076
1077/*
1078 * Delete an mfc entry
1079 */
1080static int
1081del_mfc(mfccp)
1082 struct mfcctl *mfccp;
1083{
1084 struct in_addr origin;
1085 struct in_addr mcastgrp;
1086 struct mfc *rt;
1087 struct mfc **nptr;
1088 u_long hash;
1089 int s;
1090
1091 origin = mfccp->mfcc_origin;
1092 mcastgrp = mfccp->mfcc_mcastgrp;
1093 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1094
1095 if (mrtdebug & DEBUG_MFC)
1096 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n",
1097 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1098
1099 s = splnet();
1100
1101 nptr = &mfctable[hash];
1102 while ((rt = *nptr) != NULL) {
1103 if (origin.s_addr == rt->mfc_origin.s_addr &&
1104 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1105 rt->mfc_stall == NULL)
1106 break;
1107
1108 nptr = &rt->mfc_next;
1109 }
1110 if (rt == NULL) {
1111 splx(s);
1112 return EADDRNOTAVAIL;
1113 }
1114
1115 *nptr = rt->mfc_next;
1116 free(rt, M_MRTABLE);
1117
1118 splx(s);
1119
1120 return 0;
1121}
1122
1123/*
1124 * Send a message to mrouted on the multicast routing socket
1125 */
1126static int
1127socket_send(s, mm, src)
1128 struct socket *s;
1129 struct mbuf *mm;
1130 struct sockaddr_in *src;
1131{
1132 if (s) {
1133 if (sbappendaddr(&s->so_rcv,
1134 (struct sockaddr *)src,
1135 mm, (struct mbuf *)0) != 0) {
1136 sorwakeup(s);
1137 return 0;
1138 }
1139 }
1140 m_freem(mm);
1141 return -1;
1142}
1143
1144/*
1145 * IP multicast forwarding function. This function assumes that the packet
1146 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1147 * pointed to by "ifp", and the packet is to be relayed to other networks
1148 * that have members of the packet's destination IP multicast group.
1149 *
1150 * The packet is returned unscathed to the caller, unless it is
1151 * erroneous, in which case a non-zero return value tells the caller to
1152 * discard it.
1153 */
1154
1155#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1156
1157static int
1158X_ip_mforward(ip, ifp, m, imo)
1159 register struct ip *ip;
1160 struct ifnet *ifp;
1161 struct mbuf *m;
1162 struct ip_moptions *imo;
1163{
1164 register struct mfc *rt;
1165 register u_char *ipoptions;
1166 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1167 static int srctun = 0;
1168 register struct mbuf *mm;
1169 int s;
1170 vifi_t vifi;
1171 struct vif *vifp;
1172
1173 if (mrtdebug & DEBUG_FORWARD)
1174 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n",
1175 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr),
1176 (void *)ifp);
1177
1178 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1179 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1180 /*
1181 * Packet arrived via a physical interface or
1182 * an encapsulated tunnel.
1183 */
1184 } else {
1185 /*
1186 * Packet arrived through a source-route tunnel.
1187 * Source-route tunnels are no longer supported.
1188 */
1189 if ((srctun++ % 1000) == 0)
1190 log(LOG_ERR,
1191 "ip_mforward: received source-routed packet from %lx\n",
1192 (u_long)ntohl(ip->ip_src.s_addr));
1193
1194 return 1;
1195 }
1196
1197 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1198 if (ip->ip_ttl < 255)
1199 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1200 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1201 vifp = viftable + vifi;
1202 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n",
1203 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr),
1204 vifi,
1205 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1206 vifp->v_ifp->if_name, vifp->v_ifp->if_unit);
1207 }
1208 return (ip_mdq(m, ifp, NULL, vifi));
1209 }
1210 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1211 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1212 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr));
1213 if(!imo)
1214 printf("In fact, no options were specified at all\n");
1215 }
1216
1217 /*
1218 * Don't forward a packet with time-to-live of zero or one,
1219 * or a packet destined to a local-only group.
1220 */
1221 if (ip->ip_ttl <= 1 ||
1222 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
1223 return 0;
1224
1225 /*
1226 * Determine forwarding vifs from the forwarding cache table
1227 */
1228 s = splnet();
1229 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt);
1230
1231 /* Entry exists, so forward if necessary */
1232 if (rt != NULL) {
1233 splx(s);
1234 return (ip_mdq(m, ifp, rt, -1));
1235 } else {
1236 /*
1237 * If we don't have a route for packet's origin,
1238 * Make a copy of the packet &
1239 * send message to routing daemon
1240 */
1241
1242 register struct mbuf *mb0;
1243 register struct rtdetq *rte;
1244 register u_long hash;
1245 int hlen = ip->ip_hl << 2;
1246#ifdef UPCALL_TIMING
1247 struct timeval tp;
1248
1249 GET_TIME(tp);
1250#endif
1251
1252 mrtstat.mrts_no_route++;
1253 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1254 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n",
1255 (u_long)ntohl(ip->ip_src.s_addr),
1256 (u_long)ntohl(ip->ip_dst.s_addr));
1257
1258 /*
1259 * Allocate mbufs early so that we don't do extra work if we are
1260 * just going to fail anyway. Make sure to pullup the header so
1261 * that other people can't step on it.
1262 */
1263 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT);
1264 if (rte == NULL) {
1265 splx(s);
1266 return ENOBUFS;
1267 }
1268 mb0 = m_copy(m, 0, M_COPYALL);
1269 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1270 mb0 = m_pullup(mb0, hlen);
1271 if (mb0 == NULL) {
1272 free(rte, M_MRTABLE);
1273 splx(s);
1274 return ENOBUFS;
1275 }
1276
1277 /* is there an upcall waiting for this packet? */
1278 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1279 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) {
1280 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1281 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1282 (rt->mfc_stall != NULL))
1283 break;
1284 }
1285
1286 if (rt == NULL) {
1287 int i;
1288 struct igmpmsg *im;
1289
1290 /* no upcall, so make a new entry */
1291 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1292 if (rt == NULL) {
1293 free(rte, M_MRTABLE);
1294 m_freem(mb0);
1295 splx(s);
1296 return ENOBUFS;
1297 }
1298 /* Make a copy of the header to send to the user level process */
1299 mm = m_copy(mb0, 0, hlen);
1300 if (mm == NULL) {
1301 free(rte, M_MRTABLE);
1302 m_freem(mb0);
1303 free(rt, M_MRTABLE);
1304 splx(s);
1305 return ENOBUFS;
1306 }
1307
1308 /*
1309 * Send message to routing daemon to install
1310 * a route into the kernel table
1311 */
1312 k_igmpsrc.sin_addr = ip->ip_src;
1313
1314 im = mtod(mm, struct igmpmsg *);
1315 im->im_msgtype = IGMPMSG_NOCACHE;
1316 im->im_mbz = 0;
1317
1318 mrtstat.mrts_upcalls++;
1319
1320 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1321 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1322 ++mrtstat.mrts_upq_sockfull;
1323 free(rte, M_MRTABLE);
1324 m_freem(mb0);
1325 free(rt, M_MRTABLE);
1326 splx(s);
1327 return ENOBUFS;
1328 }
1329
1330 /* insert new entry at head of hash chain */
1331 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1332 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1333 rt->mfc_expire = UPCALL_EXPIRE;
1334 nexpire[hash]++;
1335 for (i = 0; i < numvifs; i++)
1336 rt->mfc_ttls[i] = 0;
1337 rt->mfc_parent = -1;
1338
1339 /* link into table */
1340 rt->mfc_next = mfctable[hash];
1341 mfctable[hash] = rt;
1342 rt->mfc_stall = rte;
1343
1344 } else {
1345 /* determine if q has overflowed */
1346 int npkts = 0;
1347 struct rtdetq **p;
1348
1349 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1350 npkts++;
1351
1352 if (npkts > MAX_UPQ) {
1353 mrtstat.mrts_upq_ovflw++;
1354 free(rte, M_MRTABLE);
1355 m_freem(mb0);
1356 splx(s);
1357 return 0;
1358 }
1359
1360 /* Add this entry to the end of the queue */
1361 *p = rte;
1362 }
1363
1364 rte->m = mb0;
1365 rte->ifp = ifp;
1366#ifdef UPCALL_TIMING
1367 rte->t = tp;
1368#endif
1369 rte->next = NULL;
1370
1371 splx(s);
1372
1373 return 0;
1374 }
1375}
1376
1377#ifndef MROUTE_KLD
1378int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
1379 struct ip_moptions *) = X_ip_mforward;
1380#endif
1381
1382/*
1383 * Clean up the cache entry if upcall is not serviced
1384 */
1385static void
1386expire_upcalls(void *unused)
1387{
1388 struct rtdetq *rte;
1389 struct mfc *mfc, **nptr;
1390 int i;
1391 int s;
1392
1393 s = splnet();
1394 for (i = 0; i < MFCTBLSIZ; i++) {
1395 if (nexpire[i] == 0)
1396 continue;
1397 nptr = &mfctable[i];
1398 for (mfc = *nptr; mfc != NULL; mfc = *nptr) {
1399 /*
1400 * Skip real cache entries
1401 * Make sure it wasn't marked to not expire (shouldn't happen)
1402 * If it expires now
1403 */
1404 if (mfc->mfc_stall != NULL &&
1405 mfc->mfc_expire != 0 &&
1406 --mfc->mfc_expire == 0) {
1407 if (mrtdebug & DEBUG_EXPIRE)
1408 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n",
1409 (u_long)ntohl(mfc->mfc_origin.s_addr),
1410 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr));
1411 /*
1412 * drop all the packets
1413 * free the mbuf with the pkt, if, timing info
1414 */
1415 for (rte = mfc->mfc_stall; rte; ) {
1416 struct rtdetq *n = rte->next;
1417
1418 m_freem(rte->m);
1419 free(rte, M_MRTABLE);
1420 rte = n;
1421 }
1422 ++mrtstat.mrts_cache_cleanups;
1423 nexpire[i]--;
1424
1425 *nptr = mfc->mfc_next;
1426 free(mfc, M_MRTABLE);
1427 } else {
1428 nptr = &mfc->mfc_next;
1429 }
1430 }
1431 }
1432 splx(s);
1433 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
1434}
1435
1436/*
1437 * Packet forwarding routine once entry in the cache is made
1438 */
1439static int
1440ip_mdq(m, ifp, rt, xmt_vif)
1441 register struct mbuf *m;
1442 register struct ifnet *ifp;
1443 register struct mfc *rt;
1444 register vifi_t xmt_vif;
1445{
1446 register struct ip *ip = mtod(m, struct ip *);
1447 register vifi_t vifi;
1448 register struct vif *vifp;
1449 register int plen = ip->ip_len;
1450
1451/*
1452 * Macro to send packet on vif. Since RSVP packets don't get counted on
1453 * input, they shouldn't get counted on output, so statistics keeping is
1454 * separate.
1455 */
1456#define MC_SEND(ip,vifp,m) { \
1457 if ((vifp)->v_flags & VIFF_TUNNEL) \
1458 encap_send((ip), (vifp), (m)); \
1459 else \
1460 phyint_send((ip), (vifp), (m)); \
1461}
1462
1463 /*
1464 * If xmt_vif is not -1, send on only the requested vif.
1465 *
1466 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1467 */
1468 if (xmt_vif < numvifs) {
1469 MC_SEND(ip, viftable + xmt_vif, m);
1470 return 1;
1471 }
1472
1473 /*
1474 * Don't forward if it didn't arrive from the parent vif for its origin.
1475 */
1476 vifi = rt->mfc_parent;
1477 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1478 /* came in the wrong interface */
1479 if (mrtdebug & DEBUG_FORWARD)
1480 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1481 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp);
1482 ++mrtstat.mrts_wrong_if;
1483 ++rt->mfc_wrong_if;
1484 /*
1485 * If we are doing PIM assert processing, and we are forwarding
1486 * packets on this interface, and it is a broadcast medium
1487 * interface (and not a tunnel), send a message to the routing daemon.
1488 */
1489 if (pim_assert && rt->mfc_ttls[vifi] &&
1490 (ifp->if_flags & IFF_BROADCAST) &&
1491 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1492 struct sockaddr_in k_igmpsrc;
1493 struct mbuf *mm;
1494 struct igmpmsg *im;
1495 int hlen = ip->ip_hl << 2;
1496 struct timeval now;
1497 register u_long delta;
1498
1499 GET_TIME(now);
1500
1501 TV_DELTA(rt->mfc_last_assert, now, delta);
1502
1503 if (delta > ASSERT_MSG_TIME) {
1504 mm = m_copy(m, 0, hlen);
1505 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1506 mm = m_pullup(mm, hlen);
1507 if (mm == NULL) {
1508 return ENOBUFS;
1509 }
1510
1511 rt->mfc_last_assert = now;
1512
1513 im = mtod(mm, struct igmpmsg *);
1514 im->im_msgtype = IGMPMSG_WRONGVIF;
1515 im->im_mbz = 0;
1516 im->im_vif = vifi;
1517
1518 k_igmpsrc.sin_addr = im->im_src;
1519
1520 socket_send(ip_mrouter, mm, &k_igmpsrc);
1521 }
1522 }
1523 return 0;
1524 }
1525
1526 /* If I sourced this packet, it counts as output, else it was input. */
1527 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1528 viftable[vifi].v_pkt_out++;
1529 viftable[vifi].v_bytes_out += plen;
1530 } else {
1531 viftable[vifi].v_pkt_in++;
1532 viftable[vifi].v_bytes_in += plen;
1533 }
1534 rt->mfc_pkt_cnt++;
1535 rt->mfc_byte_cnt += plen;
1536
1537 /*
1538 * For each vif, decide if a copy of the packet should be forwarded.
1539 * Forward if:
1540 * - the ttl exceeds the vif's threshold
1541 * - there are group members downstream on interface
1542 */
1543 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1544 if ((rt->mfc_ttls[vifi] > 0) &&
1545 (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1546 vifp->v_pkt_out++;
1547 vifp->v_bytes_out += plen;
1548 MC_SEND(ip, vifp, m);
1549 }
1550
1551 return 0;
1552}
1553
1554/*
1555 * check if a vif number is legal/ok. This is used by ip_output, to export
1556 * numvifs there,
1557 */
1558static int
1559X_legal_vif_num(vif)
1560 int vif;
1561{
1562 if (vif >= 0 && vif < numvifs)
1563 return(1);
1564 else
1565 return(0);
1566}
1567
1568#ifndef MROUTE_KLD
1569int (*legal_vif_num)(int) = X_legal_vif_num;
1570#endif
1571
1572/*
1573 * Return the local address used by this vif
1574 */
1575static u_long
1576X_ip_mcast_src(vifi)
1577 int vifi;
1578{
1579 if (vifi >= 0 && vifi < numvifs)
1580 return viftable[vifi].v_lcl_addr.s_addr;
1581 else
1582 return INADDR_ANY;
1583}
1584
1585#ifndef MROUTE_KLD
1586u_long (*ip_mcast_src)(int) = X_ip_mcast_src;
1587#endif
1588
1589static void
1590phyint_send(ip, vifp, m)
1591 struct ip *ip;
1592 struct vif *vifp;
1593 struct mbuf *m;
1594{
1595 register struct mbuf *mb_copy;
1596 register int hlen = ip->ip_hl << 2;
1597
1598 /*
1599 * Make a new reference to the packet; make sure that
1600 * the IP header is actually copied, not just referenced,
1601 * so that ip_output() only scribbles on the copy.
1602 */
1603 mb_copy = m_copy(m, 0, M_COPYALL);
1604 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1605 mb_copy = m_pullup(mb_copy, hlen);
1606 if (mb_copy == NULL)
1607 return;
1608
1609 if (vifp->v_rate_limit == 0)
1610 tbf_send_packet(vifp, mb_copy);
1611 else
1612 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1613}
1614
1615static void
1616encap_send(ip, vifp, m)
1617 register struct ip *ip;
1618 register struct vif *vifp;
1619 register struct mbuf *m;
1620{
1621 register struct mbuf *mb_copy;
1622 register struct ip *ip_copy;
1623 register int i, len = ip->ip_len;
1624
1625 /*
1626 * copy the old packet & pullup its IP header into the
1627 * new mbuf so we can modify it. Try to fill the new
1628 * mbuf since if we don't the ethernet driver will.
1629 */
1630 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER);
1631 if (mb_copy == NULL)
1632 return;
1633 mb_copy->m_data += max_linkhdr;
1634 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1635
1636 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1637 m_freem(mb_copy);
1638 return;
1639 }
1640 i = MHLEN - M_LEADINGSPACE(mb_copy);
1641 if (i > len)
1642 i = len;
1643 mb_copy = m_pullup(mb_copy, i);
1644 if (mb_copy == NULL)
1645 return;
1646 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1647
1648 /*
1649 * fill in the encapsulating IP header.
1650 */
1651 ip_copy = mtod(mb_copy, struct ip *);
1652 *ip_copy = multicast_encap_iphdr;
1653#ifdef RANDOM_IP_ID
1654 ip_copy->ip_id = ip_randomid();
1655#else
1656 ip_copy->ip_id = htons(ip_id++);
1657#endif
1658 ip_copy->ip_len += len;
1659 ip_copy->ip_src = vifp->v_lcl_addr;
1660 ip_copy->ip_dst = vifp->v_rmt_addr;
1661
1662 /*
1663 * turn the encapsulated IP header back into a valid one.
1664 */
1665 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1666 --ip->ip_ttl;
1667 ip->ip_len = htons(ip->ip_len);
1668 ip->ip_off = htons(ip->ip_off);
1669 ip->ip_sum = 0;
1670 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1671 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1672 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1673
1674 if (vifp->v_rate_limit == 0)
1675 tbf_send_packet(vifp, mb_copy);
1676 else
1677 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1678}
1679
1680/*
1681 * Token bucket filter module
1682 */
1683
1684static void
1685tbf_control(vifp, m, ip, p_len)
1686 register struct vif *vifp;
1687 register struct mbuf *m;
1688 register struct ip *ip;
1689 register u_long p_len;
1690{
1691 register struct tbf *t = vifp->v_tbf;
1692
1693 if (p_len > MAX_BKT_SIZE) {
1694 /* drop if packet is too large */
1695 mrtstat.mrts_pkt2large++;
1696 m_freem(m);
1697 return;
1698 }
1699
1700 tbf_update_tokens(vifp);
1701
1702 /* if there are enough tokens,
1703 * and the queue is empty,
1704 * send this packet out
1705 */
1706
1707 if (t->tbf_q_len == 0) {
1708 /* queue empty, send packet if enough tokens */
1709 if (p_len <= t->tbf_n_tok) {
1710 t->tbf_n_tok -= p_len;
1711 tbf_send_packet(vifp, m);
1712 } else {
1713 /* queue packet and timeout till later */
1714 tbf_queue(vifp, m);
1715 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1716 }
1717 } else if (t->tbf_q_len < t->tbf_max_q_len) {
1718 /* finite queue length, so queue pkts and process queue */
1719 tbf_queue(vifp, m);
1720 tbf_process_q(vifp);
1721 } else {
1722 /* queue length too much, try to dq and queue and process */
1723 if (!tbf_dq_sel(vifp, ip)) {
1724 mrtstat.mrts_q_overflow++;
1725 m_freem(m);
1726 return;
1727 } else {
1728 tbf_queue(vifp, m);
1729 tbf_process_q(vifp);
1730 }
1731 }
1732 return;
1733}
1734
1735/*
1736 * adds a packet to the queue at the interface
1737 */
1738static void
1739tbf_queue(vifp, m)
1740 register struct vif *vifp;
1741 register struct mbuf *m;
1742{
1743 register int s = splnet();
1744 register struct tbf *t = vifp->v_tbf;
1745
1746 if (t->tbf_t == NULL) {
1747 /* Queue was empty */
1748 t->tbf_q = m;
1749 } else {
1750 /* Insert at tail */
1751 t->tbf_t->m_act = m;
1752 }
1753
1754 /* Set new tail pointer */
1755 t->tbf_t = m;
1756
1757#ifdef DIAGNOSTIC
1758 /* Make sure we didn't get fed a bogus mbuf */
1759 if (m->m_act)
1760 panic("tbf_queue: m_act");
1761#endif
1762 m->m_act = NULL;
1763
1764 t->tbf_q_len++;
1765
1766 splx(s);
1767}
1768
1769
1770/*
1771 * processes the queue at the interface
1772 */
1773static void
1774tbf_process_q(vifp)
1775 register struct vif *vifp;
1776{
1777 register struct mbuf *m;
1778 register int len;
1779 register int s = splnet();
1780 register struct tbf *t = vifp->v_tbf;
1781
1782 /* loop through the queue at the interface and send as many packets
1783 * as possible
1784 */
1785 while (t->tbf_q_len > 0) {
1786 m = t->tbf_q;
1787
1788 len = mtod(m, struct ip *)->ip_len;
1789
1790 /* determine if the packet can be sent */
1791 if (len <= t->tbf_n_tok) {
1792 /* if so,
1793 * reduce no of tokens, dequeue the packet,
1794 * send the packet.
1795 */
1796 t->tbf_n_tok -= len;
1797
1798 t->tbf_q = m->m_act;
1799 if (--t->tbf_q_len == 0)
1800 t->tbf_t = NULL;
1801
1802 m->m_act = NULL;
1803 tbf_send_packet(vifp, m);
1804
1805 } else break;
1806 }
1807 splx(s);
1808}
1809
1810static void
1811tbf_reprocess_q(xvifp)
1812 void *xvifp;
1813{
1814 register struct vif *vifp = xvifp;
1815 if (ip_mrouter == NULL)
1816 return;
1817
1818 tbf_update_tokens(vifp);
1819
1820 tbf_process_q(vifp);
1821
1822 if (vifp->v_tbf->tbf_q_len)
1823 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1824}
1825
1826/* function that will selectively discard a member of the queue
1827 * based on the precedence value and the priority
1828 */
1829static int
1830tbf_dq_sel(vifp, ip)
1831 register struct vif *vifp;
1832 register struct ip *ip;
1833{
1834 register int s = splnet();
1835 register u_int p;
1836 register struct mbuf *m, *last;
1837 register struct mbuf **np;
1838 register struct tbf *t = vifp->v_tbf;
1839
1840 p = priority(vifp, ip);
1841
1842 np = &t->tbf_q;
1843 last = NULL;
1844 while ((m = *np) != NULL) {
1845 if (p > priority(vifp, mtod(m, struct ip *))) {
1846 *np = m->m_act;
1847 /* If we're removing the last packet, fix the tail pointer */
1848 if (m == t->tbf_t)
1849 t->tbf_t = last;
1850 m_freem(m);
1851 /* it's impossible for the queue to be empty, but
1852 * we check anyway. */
1853 if (--t->tbf_q_len == 0)
1854 t->tbf_t = NULL;
1855 splx(s);
1856 mrtstat.mrts_drop_sel++;
1857 return(1);
1858 }
1859 np = &m->m_act;
1860 last = m;
1861 }
1862 splx(s);
1863 return(0);
1864}
1865
1866static void
1867tbf_send_packet(vifp, m)
1868 register struct vif *vifp;
1869 register struct mbuf *m;
1870{
1871 struct ip_moptions imo;
1872 int error;
1873 static struct route ro;
1874 int s = splnet();
1875
1876 if (vifp->v_flags & VIFF_TUNNEL) {
1877 /* If tunnel options */
1878 ip_output(m, (struct mbuf *)0, &vifp->v_route,
|