ip_mroute.c revision 82884
1/* 2 * IP multicast forwarding procedures 3 * 4 * Written by David Waitzman, BBN Labs, August 1988. 5 * Modified by Steve Deering, Stanford, February 1989. 6 * Modified by Mark J. Steiglitz, Stanford, May, 1991 7 * Modified by Van Jacobson, LBL, January 1993 8 * Modified by Ajit Thyagarajan, PARC, August 1993 9 * Modified by Bill Fenner, PARC, April 1995 10 * 11 * MROUTING Revision: 3.5 12 * $FreeBSD: head/sys/netinet/ip_mroute.c 82884 2001-09-03 20:03:55Z julian $ 13 */ 14 15#include "opt_mrouting.h" 16#include "opt_random_ip_id.h" 17 18#include <sys/param.h> 19#include <sys/systm.h> 20#include <sys/malloc.h> 21#include <sys/mbuf.h> 22#include <sys/socket.h> 23#include <sys/socketvar.h> 24#include <sys/protosw.h> 25#include <sys/time.h> 26#include <sys/kernel.h> 27#include <sys/sysctl.h> 28#include <sys/sockio.h> 29#include <sys/syslog.h> 30#include <net/if.h> 31#include <net/route.h> 32#include <netinet/in.h> 33#include <netinet/in_systm.h> 34#include <netinet/ip.h> 35#include <netinet/ip_var.h> 36#include <netinet/in_var.h> 37#include <netinet/igmp.h> 38#include <netinet/ip_encap.h> 39#include <netinet/ip_mroute.h> 40#include <netinet/udp.h> 41#include <machine/in_cksum.h> 42 43#ifndef MROUTING 44extern u_long _ip_mcast_src __P((int vifi)); 45extern int _ip_mforward __P((struct ip *ip, struct ifnet *ifp, 46 struct mbuf *m, struct ip_moptions *imo)); 47extern int _ip_mrouter_done __P((void)); 48extern int _ip_mrouter_get __P((struct socket *so, struct sockopt *sopt)); 49extern int _ip_mrouter_set __P((struct socket *so, struct sockopt *sopt)); 50extern int _mrt_ioctl __P((int req, caddr_t data)); 51 52/* 53 * Dummy routines and globals used when multicast routing is not compiled in. 54 */ 55 56struct socket *ip_mrouter = NULL; 57u_int rsvpdebug = 0; 58 59int 60_ip_mrouter_set(so, sopt) 61 struct socket *so; 62 struct sockopt *sopt; 63{ 64 return(EOPNOTSUPP); 65} 66 67int (*ip_mrouter_set)(struct socket *, struct sockopt *) = _ip_mrouter_set; 68 69 70int 71_ip_mrouter_get(so, sopt) 72 struct socket *so; 73 struct sockopt *sopt; 74{ 75 return(EOPNOTSUPP); 76} 77 78int (*ip_mrouter_get)(struct socket *, struct sockopt *) = _ip_mrouter_get; 79 80int 81_ip_mrouter_done() 82{ 83 return(0); 84} 85 86int (*ip_mrouter_done)(void) = _ip_mrouter_done; 87 88int 89_ip_mforward(ip, ifp, m, imo) 90 struct ip *ip; 91 struct ifnet *ifp; 92 struct mbuf *m; 93 struct ip_moptions *imo; 94{ 95 return(0); 96} 97 98int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 99 struct ip_moptions *) = _ip_mforward; 100 101int 102_mrt_ioctl(int req, caddr_t data) 103{ 104 return EOPNOTSUPP; 105} 106 107int (*mrt_ioctl)(int, caddr_t) = _mrt_ioctl; 108 109void 110rsvp_input(m, off) /* XXX must fixup manually */ 111 struct mbuf *m; 112 int off; 113{ 114 /* Can still get packets with rsvp_on = 0 if there is a local member 115 * of the group to which the RSVP packet is addressed. But in this 116 * case we want to throw the packet away. 117 */ 118 if (!rsvp_on) { 119 m_freem(m); 120 return; 121 } 122 123 if (ip_rsvpd != NULL) { 124 if (rsvpdebug) 125 printf("rsvp_input: Sending packet up old-style socket\n"); 126 rip_input(m, off); 127 return; 128 } 129 /* Drop the packet */ 130 m_freem(m); 131} 132 133void ipip_input(struct mbuf *m, int off) { /* XXX must fixup manually */ 134 rip_input(m, off); 135} 136 137int (*legal_vif_num)(int) = 0; 138 139/* 140 * This should never be called, since IP_MULTICAST_VIF should fail, but 141 * just in case it does get called, the code a little lower in ip_output 142 * will assign the packet a local address. 143 */ 144u_long 145_ip_mcast_src(int vifi) { return INADDR_ANY; } 146u_long (*ip_mcast_src)(int) = _ip_mcast_src; 147 148int 149ip_rsvp_vif_init(so, sopt) 150 struct socket *so; 151 struct sockopt *sopt; 152{ 153 return(EINVAL); 154} 155 156int 157ip_rsvp_vif_done(so, sopt) 158 struct socket *so; 159 struct sockopt *sopt; 160{ 161 return(EINVAL); 162} 163 164void 165ip_rsvp_force_done(so) 166 struct socket *so; 167{ 168 return; 169} 170 171#else /* MROUTING */ 172 173#define M_HASCL(m) ((m)->m_flags & M_EXT) 174 175static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 176 177#ifndef MROUTE_KLD 178/* The socket used to communicate with the multicast routing daemon. */ 179struct socket *ip_mrouter = NULL; 180#endif 181 182#if defined(MROUTING) || defined(MROUTE_KLD) 183static struct mrtstat mrtstat; 184SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 185 &mrtstat, mrtstat, "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)"); 186#endif 187 188static struct mfc *mfctable[MFCTBLSIZ]; 189static u_char nexpire[MFCTBLSIZ]; 190static struct vif viftable[MAXVIFS]; 191static u_int mrtdebug = 0; /* debug level */ 192#define DEBUG_MFC 0x02 193#define DEBUG_FORWARD 0x04 194#define DEBUG_EXPIRE 0x08 195#define DEBUG_XMIT 0x10 196static u_int tbfdebug = 0; /* tbf debug level */ 197static u_int rsvpdebug = 0; /* rsvp debug level */ 198 199static struct callout_handle expire_upcalls_ch; 200 201#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 202#define UPCALL_EXPIRE 6 /* number of timeouts */ 203 204/* 205 * Define the token bucket filter structures 206 * tbftable -> each vif has one of these for storing info 207 */ 208 209static struct tbf tbftable[MAXVIFS]; 210#define TBF_REPROCESS (hz / 100) /* 100x / second */ 211 212/* 213 * 'Interfaces' associated with decapsulator (so we can tell 214 * packets that went through it from ones that get reflected 215 * by a broken gateway). These interfaces are never linked into 216 * the system ifnet list & no routes point to them. I.e., packets 217 * can't be sent this way. They only exist as a placeholder for 218 * multicast source verification. 219 */ 220static struct ifnet multicast_decap_if[MAXVIFS]; 221 222#define ENCAP_TTL 64 223#define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 224 225/* prototype IP hdr for encapsulated packets */ 226static struct ip multicast_encap_iphdr = { 227#if BYTE_ORDER == LITTLE_ENDIAN 228 sizeof(struct ip) >> 2, IPVERSION, 229#else 230 IPVERSION, sizeof(struct ip) >> 2, 231#endif 232 0, /* tos */ 233 sizeof(struct ip), /* total length */ 234 0, /* id */ 235 0, /* frag offset */ 236 ENCAP_TTL, ENCAP_PROTO, 237 0, /* checksum */ 238}; 239 240/* 241 * Private variables. 242 */ 243static vifi_t numvifs = 0; 244static const struct encaptab *encap_cookie = NULL; 245 246/* 247 * one-back cache used by ipip_input to locate a tunnel's vif 248 * given a datagram's src ip address. 249 */ 250static u_long last_encap_src; 251static struct vif *last_encap_vif; 252 253static u_long X_ip_mcast_src __P((int vifi)); 254static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo)); 255static int X_ip_mrouter_done __P((void)); 256static int X_ip_mrouter_get __P((struct socket *so, struct sockopt *m)); 257static int X_ip_mrouter_set __P((struct socket *so, struct sockopt *m)); 258static int X_legal_vif_num __P((int vif)); 259static int X_mrt_ioctl __P((int cmd, caddr_t data)); 260 261static int get_sg_cnt(struct sioc_sg_req *); 262static int get_vif_cnt(struct sioc_vif_req *); 263static int ip_mrouter_init(struct socket *, int); 264static int add_vif(struct vifctl *); 265static int del_vif(vifi_t); 266static int add_mfc(struct mfcctl *); 267static int del_mfc(struct mfcctl *); 268static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 269static int set_assert(int); 270static void expire_upcalls(void *); 271static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, 272 vifi_t); 273static void phyint_send(struct ip *, struct vif *, struct mbuf *); 274static void encap_send(struct ip *, struct vif *, struct mbuf *); 275static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 276static void tbf_queue(struct vif *, struct mbuf *); 277static void tbf_process_q(struct vif *); 278static void tbf_reprocess_q(void *); 279static int tbf_dq_sel(struct vif *, struct ip *); 280static void tbf_send_packet(struct vif *, struct mbuf *); 281static void tbf_update_tokens(struct vif *); 282static int priority(struct vif *, struct ip *); 283 284/* 285 * whether or not special PIM assert processing is enabled. 286 */ 287static int pim_assert; 288/* 289 * Rate limit for assert notification messages, in usec 290 */ 291#define ASSERT_MSG_TIME 3000000 292 293/* 294 * Hash function for a source, group entry 295 */ 296#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 297 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 298 299/* 300 * Find a route for a given origin IP address and Multicast group address 301 * Type of service parameter to be added in the future!!! 302 */ 303 304#define MFCFIND(o, g, rt) { \ 305 register struct mfc *_rt = mfctable[MFCHASH(o,g)]; \ 306 rt = NULL; \ 307 ++mrtstat.mrts_mfc_lookups; \ 308 while (_rt) { \ 309 if ((_rt->mfc_origin.s_addr == o) && \ 310 (_rt->mfc_mcastgrp.s_addr == g) && \ 311 (_rt->mfc_stall == NULL)) { \ 312 rt = _rt; \ 313 break; \ 314 } \ 315 _rt = _rt->mfc_next; \ 316 } \ 317 if (rt == NULL) { \ 318 ++mrtstat.mrts_mfc_misses; \ 319 } \ 320} 321 322 323/* 324 * Macros to compute elapsed time efficiently 325 * Borrowed from Van Jacobson's scheduling code 326 */ 327#define TV_DELTA(a, b, delta) { \ 328 register int xxs; \ 329 \ 330 delta = (a).tv_usec - (b).tv_usec; \ 331 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 332 switch (xxs) { \ 333 case 2: \ 334 delta += 1000000; \ 335 /* fall through */ \ 336 case 1: \ 337 delta += 1000000; \ 338 break; \ 339 default: \ 340 delta += (1000000 * xxs); \ 341 } \ 342 } \ 343} 344 345#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 346 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 347 348#ifdef UPCALL_TIMING 349u_long upcall_data[51]; 350static void collate(struct timeval *); 351#endif /* UPCALL_TIMING */ 352 353 354/* 355 * Handle MRT setsockopt commands to modify the multicast routing tables. 356 */ 357static int 358X_ip_mrouter_set(so, sopt) 359 struct socket *so; 360 struct sockopt *sopt; 361{ 362 int error, optval; 363 vifi_t vifi; 364 struct vifctl vifc; 365 struct mfcctl mfc; 366 367 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 368 return (EPERM); 369 370 error = 0; 371 switch (sopt->sopt_name) { 372 case MRT_INIT: 373 error = sooptcopyin(sopt, &optval, sizeof optval, 374 sizeof optval); 375 if (error) 376 break; 377 error = ip_mrouter_init(so, optval); 378 break; 379 380 case MRT_DONE: 381 error = ip_mrouter_done(); 382 break; 383 384 case MRT_ADD_VIF: 385 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 386 if (error) 387 break; 388 error = add_vif(&vifc); 389 break; 390 391 case MRT_DEL_VIF: 392 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 393 if (error) 394 break; 395 error = del_vif(vifi); 396 break; 397 398 case MRT_ADD_MFC: 399 case MRT_DEL_MFC: 400 error = sooptcopyin(sopt, &mfc, sizeof mfc, sizeof mfc); 401 if (error) 402 break; 403 if (sopt->sopt_name == MRT_ADD_MFC) 404 error = add_mfc(&mfc); 405 else 406 error = del_mfc(&mfc); 407 break; 408 409 case MRT_ASSERT: 410 error = sooptcopyin(sopt, &optval, sizeof optval, 411 sizeof optval); 412 if (error) 413 break; 414 set_assert(optval); 415 break; 416 417 default: 418 error = EOPNOTSUPP; 419 break; 420 } 421 return (error); 422} 423 424#ifndef MROUTE_KLD 425int (*ip_mrouter_set)(struct socket *, struct sockopt *) = X_ip_mrouter_set; 426#endif 427 428/* 429 * Handle MRT getsockopt commands 430 */ 431static int 432X_ip_mrouter_get(so, sopt) 433 struct socket *so; 434 struct sockopt *sopt; 435{ 436 int error; 437 static int version = 0x0305; /* !!! why is this here? XXX */ 438 439 switch (sopt->sopt_name) { 440 case MRT_VERSION: 441 error = sooptcopyout(sopt, &version, sizeof version); 442 break; 443 444 case MRT_ASSERT: 445 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 446 break; 447 default: 448 error = EOPNOTSUPP; 449 break; 450 } 451 return (error); 452} 453 454#ifndef MROUTE_KLD 455int (*ip_mrouter_get)(struct socket *, struct sockopt *) = X_ip_mrouter_get; 456#endif 457 458/* 459 * Handle ioctl commands to obtain information from the cache 460 */ 461static int 462X_mrt_ioctl(cmd, data) 463 int cmd; 464 caddr_t data; 465{ 466 int error = 0; 467 468 switch (cmd) { 469 case (SIOCGETVIFCNT): 470 return (get_vif_cnt((struct sioc_vif_req *)data)); 471 break; 472 case (SIOCGETSGCNT): 473 return (get_sg_cnt((struct sioc_sg_req *)data)); 474 break; 475 default: 476 return (EINVAL); 477 break; 478 } 479 return error; 480} 481 482#ifndef MROUTE_KLD 483int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl; 484#endif 485 486/* 487 * returns the packet, byte, rpf-failure count for the source group provided 488 */ 489static int 490get_sg_cnt(req) 491 register struct sioc_sg_req *req; 492{ 493 register struct mfc *rt; 494 int s; 495 496 s = splnet(); 497 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 498 splx(s); 499 if (rt != NULL) { 500 req->pktcnt = rt->mfc_pkt_cnt; 501 req->bytecnt = rt->mfc_byte_cnt; 502 req->wrong_if = rt->mfc_wrong_if; 503 } else 504 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 505 506 return 0; 507} 508 509/* 510 * returns the input and output packet and byte counts on the vif provided 511 */ 512static int 513get_vif_cnt(req) 514 register struct sioc_vif_req *req; 515{ 516 register vifi_t vifi = req->vifi; 517 518 if (vifi >= numvifs) return EINVAL; 519 520 req->icount = viftable[vifi].v_pkt_in; 521 req->ocount = viftable[vifi].v_pkt_out; 522 req->ibytes = viftable[vifi].v_bytes_in; 523 req->obytes = viftable[vifi].v_bytes_out; 524 525 return 0; 526} 527 528/* 529 * Enable multicast routing 530 */ 531static int 532ip_mrouter_init(so, version) 533 struct socket *so; 534 int version; 535{ 536 if (mrtdebug) 537 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 538 so->so_type, so->so_proto->pr_protocol); 539 540 if (so->so_type != SOCK_RAW || 541 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP; 542 543 if (version != 1) 544 return ENOPROTOOPT; 545 546 if (ip_mrouter != NULL) return EADDRINUSE; 547 548 ip_mrouter = so; 549 550 bzero((caddr_t)mfctable, sizeof(mfctable)); 551 bzero((caddr_t)nexpire, sizeof(nexpire)); 552 553 pim_assert = 0; 554 555 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 556 557 if (mrtdebug) 558 log(LOG_DEBUG, "ip_mrouter_init\n"); 559 560 return 0; 561} 562 563/* 564 * Disable multicast routing 565 */ 566static int 567X_ip_mrouter_done() 568{ 569 vifi_t vifi; 570 int i; 571 struct ifnet *ifp; 572 struct ifreq ifr; 573 struct mfc *rt; 574 struct rtdetq *rte; 575 int s; 576 577 s = splnet(); 578 579 /* 580 * For each phyint in use, disable promiscuous reception of all IP 581 * multicasts. 582 */ 583 for (vifi = 0; vifi < numvifs; vifi++) { 584 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 585 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 586 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 587 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr 588 = INADDR_ANY; 589 ifp = viftable[vifi].v_ifp; 590 if_allmulti(ifp, 0); 591 } 592 } 593 bzero((caddr_t)tbftable, sizeof(tbftable)); 594 bzero((caddr_t)viftable, sizeof(viftable)); 595 numvifs = 0; 596 pim_assert = 0; 597 598 untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch); 599 600 /* 601 * Free all multicast forwarding cache entries. 602 */ 603 for (i = 0; i < MFCTBLSIZ; i++) { 604 for (rt = mfctable[i]; rt != NULL; ) { 605 struct mfc *nr = rt->mfc_next; 606 607 for (rte = rt->mfc_stall; rte != NULL; ) { 608 struct rtdetq *n = rte->next; 609 610 m_freem(rte->m); 611 free(rte, M_MRTABLE); 612 rte = n; 613 } 614 free(rt, M_MRTABLE); 615 rt = nr; 616 } 617 } 618 619 bzero((caddr_t)mfctable, sizeof(mfctable)); 620 621 /* 622 * Reset de-encapsulation cache 623 */ 624 last_encap_src = 0; 625 last_encap_vif = NULL; 626 if (encap_cookie) { 627 encap_detach(encap_cookie); 628 encap_cookie = NULL; 629 } 630 631 ip_mrouter = NULL; 632 633 splx(s); 634 635 if (mrtdebug) 636 log(LOG_DEBUG, "ip_mrouter_done\n"); 637 638 return 0; 639} 640 641#ifndef MROUTE_KLD 642int (*ip_mrouter_done)(void) = X_ip_mrouter_done; 643#endif 644 645/* 646 * Set PIM assert processing global 647 */ 648static int 649set_assert(i) 650 int i; 651{ 652 if ((i != 1) && (i != 0)) 653 return EINVAL; 654 655 pim_assert = i; 656 657 return 0; 658} 659 660/* 661 * Decide if a packet is from a tunnelled peer. 662 * Return 0 if not, 64 if so. 663 */ 664static int 665mroute_encapcheck(const struct mbuf *m, int off, int proto, void *arg) 666{ 667 struct ip *ip = mtod(m, struct ip *); 668 int hlen = ip->ip_hl << 2; 669 register struct vif *vifp; 670 671 /* 672 * don't claim the packet if it's not to a multicast destination or if 673 * we don't have an encapsulating tunnel with the source. 674 * Note: This code assumes that the remote site IP address 675 * uniquely identifies the tunnel (i.e., that this site has 676 * at most one tunnel with the remote site). 677 */ 678 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) { 679 return 0; 680 } 681 if (ip->ip_src.s_addr != last_encap_src) { 682 register struct vif *vife; 683 684 vifp = viftable; 685 vife = vifp + numvifs; 686 last_encap_src = ip->ip_src.s_addr; 687 last_encap_vif = 0; 688 for ( ; vifp < vife; ++vifp) 689 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 690 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 691 == VIFF_TUNNEL) 692 last_encap_vif = vifp; 693 break; 694 } 695 } 696 if ((vifp = last_encap_vif) == 0) { 697 last_encap_src = 0; 698 return 0; 699 } 700 return 64; 701} 702 703/* 704 * De-encapsulate a packet and feed it back through ip input (this 705 * routine is called whenever IP gets a packet that mroute_encap_func() 706 * claimed). 707 */ 708static void 709mroute_encap_input(struct mbuf *m, int off) 710{ 711 struct ip *ip = mtod(m, struct ip *); 712 int hlen = ip->ip_hl << 2; 713 714 if (hlen > sizeof(struct ip)) 715 ip_stripoptions(m, (struct mbuf *) 0); 716 m->m_data += sizeof(struct ip); 717 m->m_len -= sizeof(struct ip); 718 m->m_pkthdr.len -= sizeof(struct ip); 719 720 m->m_pkthdr.rcvif = last_encap_vif->v_ifp; 721 722 (void) IF_HANDOFF(&ipintrq, m, NULL); 723 /* 724 * normally we would need a "schednetisr(NETISR_IP)" 725 * here but we were called by ip_input and it is going 726 * to loop back & try to dequeue the packet we just 727 * queued as soon as we return so we avoid the 728 * unnecessary software interrrupt. 729 */ 730} 731 732extern struct domain inetdomain; 733static struct protosw mroute_encap_protosw = 734{ SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR, 735 mroute_encap_input, 0, 0, rip_ctloutput, 736 0, 737 0, 0, 0, 0, 738 &rip_usrreqs 739}; 740 741/* 742 * Add a vif to the vif table 743 */ 744static int 745add_vif(vifcp) 746 register struct vifctl *vifcp; 747{ 748 register struct vif *vifp = viftable + vifcp->vifc_vifi; 749 static struct sockaddr_in sin = {sizeof sin, AF_INET}; 750 struct ifaddr *ifa; 751 struct ifnet *ifp; 752 int error, s; 753 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 754 755 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL; 756 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE; 757 758 /* Find the interface with an address in AF_INET family */ 759 sin.sin_addr = vifcp->vifc_lcl_addr; 760 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 761 if (ifa == 0) return EADDRNOTAVAIL; 762 ifp = ifa->ifa_ifp; 763 764 if (vifcp->vifc_flags & VIFF_TUNNEL) { 765 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 766 /* 767 * An encapsulating tunnel is wanted. Tell 768 * mroute_encap_input() to start paying attention 769 * to encapsulated packets. 770 */ 771 if (encap_cookie == NULL) { 772 encap_cookie = encap_attach_func(AF_INET, -1, 773 mroute_encapcheck, 774 (struct protosw *)&mroute_encap_protosw, NULL); 775 776 if (encap_cookie == NULL) { 777 printf("ip_mroute: unable to attach encap\n"); 778 return (EIO); /* XXX */ 779 } 780 for (s = 0; s < MAXVIFS; ++s) { 781 multicast_decap_if[s].if_name = "mdecap"; 782 multicast_decap_if[s].if_unit = s; 783 } 784 } 785 /* 786 * Set interface to fake encapsulator interface 787 */ 788 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 789 /* 790 * Prepare cached route entry 791 */ 792 bzero(&vifp->v_route, sizeof(vifp->v_route)); 793 } else { 794 log(LOG_ERR, "source routed tunnels not supported\n"); 795 return EOPNOTSUPP; 796 } 797 } else { 798 /* Make sure the interface supports multicast */ 799 if ((ifp->if_flags & IFF_MULTICAST) == 0) 800 return EOPNOTSUPP; 801 802 /* Enable promiscuous reception of all IP multicasts from the if */ 803 s = splnet(); 804 error = if_allmulti(ifp, 1); 805 splx(s); 806 if (error) 807 return error; 808 } 809 810 s = splnet(); 811 /* define parameters for the tbf structure */ 812 vifp->v_tbf = v_tbf; 813 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 814 vifp->v_tbf->tbf_n_tok = 0; 815 vifp->v_tbf->tbf_q_len = 0; 816 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 817 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 818 819 vifp->v_flags = vifcp->vifc_flags; 820 vifp->v_threshold = vifcp->vifc_threshold; 821 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 822 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 823 vifp->v_ifp = ifp; 824 /* scaling up here allows division by 1024 in critical code */ 825 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 826 vifp->v_rsvp_on = 0; 827 vifp->v_rsvpd = NULL; 828 /* initialize per vif pkt counters */ 829 vifp->v_pkt_in = 0; 830 vifp->v_pkt_out = 0; 831 vifp->v_bytes_in = 0; 832 vifp->v_bytes_out = 0; 833 splx(s); 834 835 /* Adjust numvifs up if the vifi is higher than numvifs */ 836 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 837 838 if (mrtdebug) 839 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n", 840 vifcp->vifc_vifi, 841 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 842 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 843 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 844 vifcp->vifc_threshold, 845 vifcp->vifc_rate_limit); 846 847 return 0; 848} 849 850/* 851 * Delete a vif from the vif table 852 */ 853static int 854del_vif(vifi) 855 vifi_t vifi; 856{ 857 register struct vif *vifp = &viftable[vifi]; 858 register struct mbuf *m; 859 struct ifnet *ifp; 860 struct ifreq ifr; 861 int s; 862 863 if (vifi >= numvifs) return EINVAL; 864 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL; 865 866 s = splnet(); 867 868 if (!(vifp->v_flags & VIFF_TUNNEL)) { 869 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 870 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 871 ifp = vifp->v_ifp; 872 if_allmulti(ifp, 0); 873 } 874 875 if (vifp == last_encap_vif) { 876 last_encap_vif = 0; 877 last_encap_src = 0; 878 } 879 880 /* 881 * Free packets queued at the interface 882 */ 883 while (vifp->v_tbf->tbf_q) { 884 m = vifp->v_tbf->tbf_q; 885 vifp->v_tbf->tbf_q = m->m_act; 886 m_freem(m); 887 } 888 889 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 890 bzero((caddr_t)vifp, sizeof (*vifp)); 891 892 if (mrtdebug) 893 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 894 895 /* Adjust numvifs down */ 896 for (vifi = numvifs; vifi > 0; vifi--) 897 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break; 898 numvifs = vifi; 899 900 splx(s); 901 902 return 0; 903} 904 905/* 906 * Add an mfc entry 907 */ 908static int 909add_mfc(mfccp) 910 struct mfcctl *mfccp; 911{ 912 struct mfc *rt; 913 u_long hash; 914 struct rtdetq *rte; 915 register u_short nstl; 916 int s; 917 int i; 918 919 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 920 921 /* If an entry already exists, just update the fields */ 922 if (rt) { 923 if (mrtdebug & DEBUG_MFC) 924 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 925 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 926 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 927 mfccp->mfcc_parent); 928 929 s = splnet(); 930 rt->mfc_parent = mfccp->mfcc_parent; 931 for (i = 0; i < numvifs; i++) 932 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 933 splx(s); 934 return 0; 935 } 936 937 /* 938 * Find the entry for which the upcall was made and update 939 */ 940 s = splnet(); 941 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 942 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 943 944 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 945 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 946 (rt->mfc_stall != NULL)) { 947 948 if (nstl++) 949 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 950 "multiple kernel entries", 951 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 952 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 953 mfccp->mfcc_parent, (void *)rt->mfc_stall); 954 955 if (mrtdebug & DEBUG_MFC) 956 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 957 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 958 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 959 mfccp->mfcc_parent, (void *)rt->mfc_stall); 960 961 rt->mfc_origin = mfccp->mfcc_origin; 962 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 963 rt->mfc_parent = mfccp->mfcc_parent; 964 for (i = 0; i < numvifs; i++) 965 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 966 /* initialize pkt counters per src-grp */ 967 rt->mfc_pkt_cnt = 0; 968 rt->mfc_byte_cnt = 0; 969 rt->mfc_wrong_if = 0; 970 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 971 972 rt->mfc_expire = 0; /* Don't clean this guy up */ 973 nexpire[hash]--; 974 975 /* free packets Qed at the end of this entry */ 976 for (rte = rt->mfc_stall; rte != NULL; ) { 977 struct rtdetq *n = rte->next; 978 979 ip_mdq(rte->m, rte->ifp, rt, -1); 980 m_freem(rte->m); 981#ifdef UPCALL_TIMING 982 collate(&(rte->t)); 983#endif /* UPCALL_TIMING */ 984 free(rte, M_MRTABLE); 985 rte = n; 986 } 987 rt->mfc_stall = NULL; 988 } 989 } 990 991 /* 992 * It is possible that an entry is being inserted without an upcall 993 */ 994 if (nstl == 0) { 995 if (mrtdebug & DEBUG_MFC) 996 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 997 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 998 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 999 mfccp->mfcc_parent); 1000 1001 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 1002 1003 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1004 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 1005 1006 rt->mfc_origin = mfccp->mfcc_origin; 1007 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1008 rt->mfc_parent = mfccp->mfcc_parent; 1009 for (i = 0; i < numvifs; i++) 1010 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1011 /* initialize pkt counters per src-grp */ 1012 rt->mfc_pkt_cnt = 0; 1013 rt->mfc_byte_cnt = 0; 1014 rt->mfc_wrong_if = 0; 1015 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 1016 if (rt->mfc_expire) 1017 nexpire[hash]--; 1018 rt->mfc_expire = 0; 1019 } 1020 } 1021 if (rt == NULL) { 1022 /* no upcall, so make a new entry */ 1023 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1024 if (rt == NULL) { 1025 splx(s); 1026 return ENOBUFS; 1027 } 1028 1029 /* insert new entry at head of hash chain */ 1030 rt->mfc_origin = mfccp->mfcc_origin; 1031 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1032 rt->mfc_parent = mfccp->mfcc_parent; 1033 for (i = 0; i < numvifs; i++) 1034 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1035 /* initialize pkt counters per src-grp */ 1036 rt->mfc_pkt_cnt = 0; 1037 rt->mfc_byte_cnt = 0; 1038 rt->mfc_wrong_if = 0; 1039 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 1040 rt->mfc_expire = 0; 1041 rt->mfc_stall = NULL; 1042 1043 /* link into table */ 1044 rt->mfc_next = mfctable[hash]; 1045 mfctable[hash] = rt; 1046 } 1047 } 1048 splx(s); 1049 return 0; 1050} 1051 1052#ifdef UPCALL_TIMING 1053/* 1054 * collect delay statistics on the upcalls 1055 */ 1056static void collate(t) 1057register struct timeval *t; 1058{ 1059 register u_long d; 1060 register struct timeval tp; 1061 register u_long delta; 1062 1063 GET_TIME(tp); 1064 1065 if (TV_LT(*t, tp)) 1066 { 1067 TV_DELTA(tp, *t, delta); 1068 1069 d = delta >> 10; 1070 if (d > 50) 1071 d = 50; 1072 1073 ++upcall_data[d]; 1074 } 1075} 1076#endif /* UPCALL_TIMING */ 1077 1078/* 1079 * Delete an mfc entry 1080 */ 1081static int 1082del_mfc(mfccp) 1083 struct mfcctl *mfccp; 1084{ 1085 struct in_addr origin; 1086 struct in_addr mcastgrp; 1087 struct mfc *rt; 1088 struct mfc **nptr; 1089 u_long hash; 1090 int s; 1091 1092 origin = mfccp->mfcc_origin; 1093 mcastgrp = mfccp->mfcc_mcastgrp; 1094 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1095 1096 if (mrtdebug & DEBUG_MFC) 1097 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1098 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1099 1100 s = splnet(); 1101 1102 nptr = &mfctable[hash]; 1103 while ((rt = *nptr) != NULL) { 1104 if (origin.s_addr == rt->mfc_origin.s_addr && 1105 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1106 rt->mfc_stall == NULL) 1107 break; 1108 1109 nptr = &rt->mfc_next; 1110 } 1111 if (rt == NULL) { 1112 splx(s); 1113 return EADDRNOTAVAIL; 1114 } 1115 1116 *nptr = rt->mfc_next; 1117 free(rt, M_MRTABLE); 1118 1119 splx(s); 1120 1121 return 0; 1122} 1123 1124/* 1125 * Send a message to mrouted on the multicast routing socket 1126 */ 1127static int 1128socket_send(s, mm, src) 1129 struct socket *s; 1130 struct mbuf *mm; 1131 struct sockaddr_in *src; 1132{ 1133 if (s) { 1134 if (sbappendaddr(&s->so_rcv, 1135 (struct sockaddr *)src, 1136 mm, (struct mbuf *)0) != 0) { 1137 sorwakeup(s); 1138 return 0; 1139 } 1140 } 1141 m_freem(mm); 1142 return -1; 1143} 1144 1145/* 1146 * IP multicast forwarding function. This function assumes that the packet 1147 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1148 * pointed to by "ifp", and the packet is to be relayed to other networks 1149 * that have members of the packet's destination IP multicast group. 1150 * 1151 * The packet is returned unscathed to the caller, unless it is 1152 * erroneous, in which case a non-zero return value tells the caller to 1153 * discard it. 1154 */ 1155 1156#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1157 1158static int 1159X_ip_mforward(ip, ifp, m, imo) 1160 register struct ip *ip; 1161 struct ifnet *ifp; 1162 struct mbuf *m; 1163 struct ip_moptions *imo; 1164{ 1165 register struct mfc *rt; 1166 register u_char *ipoptions; 1167 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1168 static int srctun = 0; 1169 register struct mbuf *mm; 1170 int s; 1171 vifi_t vifi; 1172 struct vif *vifp; 1173 1174 if (mrtdebug & DEBUG_FORWARD) 1175 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1176 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1177 (void *)ifp); 1178 1179 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1180 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1181 /* 1182 * Packet arrived via a physical interface or 1183 * an encapsulated tunnel. 1184 */ 1185 } else { 1186 /* 1187 * Packet arrived through a source-route tunnel. 1188 * Source-route tunnels are no longer supported. 1189 */ 1190 if ((srctun++ % 1000) == 0) 1191 log(LOG_ERR, 1192 "ip_mforward: received source-routed packet from %lx\n", 1193 (u_long)ntohl(ip->ip_src.s_addr)); 1194 1195 return 1; 1196 } 1197 1198 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1199 if (ip->ip_ttl < 255) 1200 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1201 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1202 vifp = viftable + vifi; 1203 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1204 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), 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 ntohl(ip->ip_src.s_addr), 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 HTONS(ip->ip_len); 1668 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, 1879 IP_FORWARDING, (struct ip_moptions *)0); 1880 } else { 1881 imo.imo_multicast_ifp = vifp->v_ifp; 1882 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1883 imo.imo_multicast_loop = 1; 1884 imo.imo_multicast_vif = -1; 1885 1886 /* 1887 * Re-entrancy should not be a problem here, because 1888 * the packets that we send out and are looped back at us 1889 * should get rejected because they appear to come from 1890 * the loopback interface, thus preventing looping. 1891 */ 1892 error = ip_output(m, (struct mbuf *)0, &ro, 1893 IP_FORWARDING, &imo); 1894 1895 if (mrtdebug & DEBUG_XMIT) 1896 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1897 vifp - viftable, error); 1898 } 1899 splx(s); 1900} 1901 1902/* determine the current time and then 1903 * the elapsed time (between the last time and time now) 1904 * in milliseconds & update the no. of tokens in the bucket 1905 */ 1906static void 1907tbf_update_tokens(vifp) 1908 register struct vif *vifp; 1909{ 1910 struct timeval tp; 1911 register u_long tm; 1912 register int s = splnet(); 1913 register struct tbf *t = vifp->v_tbf; 1914 1915 GET_TIME(tp); 1916 1917 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1918 1919 /* 1920 * This formula is actually 1921 * "time in seconds" * "bytes/second". 1922 * 1923 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1924 * 1925 * The (1000/1024) was introduced in add_vif to optimize 1926 * this divide into a shift. 1927 */ 1928 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1929 t->tbf_last_pkt_t = tp; 1930 1931 if (t->tbf_n_tok > MAX_BKT_SIZE) 1932 t->tbf_n_tok = MAX_BKT_SIZE; 1933 1934 splx(s); 1935} 1936 1937static int 1938priority(vifp, ip) 1939 register struct vif *vifp; 1940 register struct ip *ip; 1941{ 1942 register int prio; 1943 1944 /* temporary hack; may add general packet classifier some day */ 1945 1946 /* 1947 * The UDP port space is divided up into four priority ranges: 1948 * [0, 16384) : unclassified - lowest priority 1949 * [16384, 32768) : audio - highest priority 1950 * [32768, 49152) : whiteboard - medium priority 1951 * [49152, 65536) : video - low priority 1952 */ 1953 if (ip->ip_p == IPPROTO_UDP) { 1954 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1955 switch (ntohs(udp->uh_dport) & 0xc000) { 1956 case 0x4000: 1957 prio = 70; 1958 break; 1959 case 0x8000: 1960 prio = 60; 1961 break; 1962 case 0xc000: 1963 prio = 55; 1964 break; 1965 default: 1966 prio = 50; 1967 break; 1968 } 1969 if (tbfdebug > 1) 1970 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio); 1971 } else { 1972 prio = 50; 1973 } 1974 return prio; 1975} 1976 1977/* 1978 * End of token bucket filter modifications 1979 */ 1980 1981int 1982ip_rsvp_vif_init(so, sopt) 1983 struct socket *so; 1984 struct sockopt *sopt; 1985{ 1986 int error, i, s; 1987 1988 if (rsvpdebug) 1989 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 1990 so->so_type, so->so_proto->pr_protocol); 1991 1992 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1993 return EOPNOTSUPP; 1994 1995 /* Check mbuf. */ 1996 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1997 if (error) 1998 return (error); 1999 2000 if (rsvpdebug) 2001 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", i, rsvp_on); 2002 2003 s = splnet(); 2004 2005 /* Check vif. */ 2006 if (!legal_vif_num(i)) { 2007 splx(s); 2008 return EADDRNOTAVAIL; 2009 } 2010 2011 /* Check if socket is available. */ 2012 if (viftable[i].v_rsvpd != NULL) { 2013 splx(s); 2014 return EADDRINUSE; 2015 } 2016 2017 viftable[i].v_rsvpd = so; 2018 /* This may seem silly, but we need to be sure we don't over-increment 2019 * the RSVP counter, in case something slips up. 2020 */ 2021 if (!viftable[i].v_rsvp_on) { 2022 viftable[i].v_rsvp_on = 1; 2023 rsvp_on++; 2024 } 2025 2026 splx(s); 2027 return 0; 2028} 2029 2030int 2031ip_rsvp_vif_done(so, sopt) 2032 struct socket *so; 2033 struct sockopt *sopt; 2034{ 2035 int error, i, s; 2036 2037 if (rsvpdebug) 2038 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2039 so->so_type, so->so_proto->pr_protocol); 2040 2041 if (so->so_type != SOCK_RAW || 2042 so->so_proto->pr_protocol != IPPROTO_RSVP) 2043 return EOPNOTSUPP; 2044 2045 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 2046 if (error) 2047 return (error); 2048 2049 s = splnet(); 2050 2051 /* Check vif. */ 2052 if (!legal_vif_num(i)) { 2053 splx(s); 2054 return EADDRNOTAVAIL; 2055 } 2056 2057 if (rsvpdebug) 2058 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n", 2059 viftable[i].v_rsvpd, so); 2060 2061 viftable[i].v_rsvpd = NULL; 2062 /* 2063 * This may seem silly, but we need to be sure we don't over-decrement 2064 * the RSVP counter, in case something slips up. 2065 */ 2066 if (viftable[i].v_rsvp_on) { 2067 viftable[i].v_rsvp_on = 0; 2068 rsvp_on--; 2069 } 2070 2071 splx(s); 2072 return 0; 2073} 2074 2075void 2076ip_rsvp_force_done(so) 2077 struct socket *so; 2078{ 2079 int vifi; 2080 register int s; 2081 2082 /* Don't bother if it is not the right type of socket. */ 2083 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2084 return; 2085 2086 s = splnet(); 2087 2088 /* The socket may be attached to more than one vif...this 2089 * is perfectly legal. 2090 */ 2091 for (vifi = 0; vifi < numvifs; vifi++) { 2092 if (viftable[vifi].v_rsvpd == so) { 2093 viftable[vifi].v_rsvpd = NULL; 2094 /* This may seem silly, but we need to be sure we don't 2095 * over-decrement the RSVP counter, in case something slips up. 2096 */ 2097 if (viftable[vifi].v_rsvp_on) { 2098 viftable[vifi].v_rsvp_on = 0; 2099 rsvp_on--; 2100 } 2101 } 2102 } 2103 2104 splx(s); 2105 return; 2106} 2107 2108void 2109rsvp_input(m, off) 2110 struct mbuf *m; 2111 int off; 2112{ 2113 int vifi; 2114 register struct ip *ip = mtod(m, struct ip *); 2115 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2116 register int s; 2117 struct ifnet *ifp; 2118 2119 if (rsvpdebug) 2120 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2121 2122 /* Can still get packets with rsvp_on = 0 if there is a local member 2123 * of the group to which the RSVP packet is addressed. But in this 2124 * case we want to throw the packet away. 2125 */ 2126 if (!rsvp_on) { 2127 m_freem(m); 2128 return; 2129 } 2130 2131 s = splnet(); 2132 2133 if (rsvpdebug) 2134 printf("rsvp_input: check vifs\n"); 2135 2136#ifdef DIAGNOSTIC 2137 if (!(m->m_flags & M_PKTHDR)) 2138 panic("rsvp_input no hdr"); 2139#endif 2140 2141 ifp = m->m_pkthdr.rcvif; 2142 /* Find which vif the packet arrived on. */ 2143 for (vifi = 0; vifi < numvifs; vifi++) 2144 if (viftable[vifi].v_ifp == ifp) 2145 break; 2146 2147 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 2148 /* 2149 * If the old-style non-vif-associated socket is set, 2150 * then use it. Otherwise, drop packet since there 2151 * is no specific socket for this vif. 2152 */ 2153 if (ip_rsvpd != NULL) { 2154 if (rsvpdebug) 2155 printf("rsvp_input: Sending packet up old-style socket\n"); 2156 rip_input(m, off); /* xxx */ 2157 } else { 2158 if (rsvpdebug && vifi == numvifs) 2159 printf("rsvp_input: Can't find vif for packet.\n"); 2160 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 2161 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2162 m_freem(m); 2163 } 2164 splx(s); 2165 return; 2166 } 2167 rsvp_src.sin_addr = ip->ip_src; 2168 2169 if (rsvpdebug && m) 2170 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2171 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2172 2173 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 2174 if (rsvpdebug) 2175 printf("rsvp_input: Failed to append to socket\n"); 2176 } else { 2177 if (rsvpdebug) 2178 printf("rsvp_input: send packet up\n"); 2179 } 2180 2181 splx(s); 2182} 2183 2184#ifdef MROUTE_KLD 2185 2186static int 2187ip_mroute_modevent(module_t mod, int type, void *unused) 2188{ 2189 int s; 2190 2191 switch (type) { 2192 static u_long (*old_ip_mcast_src)(int); 2193 static int (*old_ip_mrouter_set)(struct socket *, 2194 struct sockopt *); 2195 static int (*old_ip_mrouter_get)(struct socket *, 2196 struct sockopt *); 2197 static int (*old_ip_mrouter_done)(void); 2198 static int (*old_ip_mforward)(struct ip *, struct ifnet *, 2199 struct mbuf *, struct ip_moptions *); 2200 static int (*old_mrt_ioctl)(int, caddr_t); 2201 static int (*old_legal_vif_num)(int); 2202 2203 case MOD_LOAD: 2204 s = splnet(); 2205 /* XXX Protect against multiple loading */ 2206 old_ip_mcast_src = ip_mcast_src; 2207 ip_mcast_src = X_ip_mcast_src; 2208 old_ip_mrouter_get = ip_mrouter_get; 2209 ip_mrouter_get = X_ip_mrouter_get; 2210 old_ip_mrouter_set = ip_mrouter_set; 2211 ip_mrouter_set = X_ip_mrouter_set; 2212 old_ip_mrouter_done = ip_mrouter_done; 2213 ip_mrouter_done = X_ip_mrouter_done; 2214 old_ip_mforward = ip_mforward; 2215 ip_mforward = X_ip_mforward; 2216 old_mrt_ioctl = mrt_ioctl; 2217 mrt_ioctl = X_mrt_ioctl; 2218 old_legal_vif_num = legal_vif_num; 2219 legal_vif_num = X_legal_vif_num; 2220 2221 splx(s); 2222 return 0; 2223 2224 case MOD_UNLOAD: 2225 if (ip_mrouter) 2226 return EINVAL; 2227 2228 s = splnet(); 2229 ip_mrouter_get = old_ip_mrouter_get; 2230 ip_mrouter_set = old_ip_mrouter_set; 2231 ip_mrouter_done = old_ip_mrouter_done; 2232 ip_mforward = old_ip_mforward; 2233 mrt_ioctl = old_mrt_ioctl; 2234 legal_vif_num = old_legal_vif_num; 2235 splx(s); 2236 return 0; 2237 2238 default: 2239 break; 2240 } 2241 return 0; 2242} 2243 2244static moduledata_t ip_mroutemod = { 2245 "ip_mroute", 2246 ip_mroute_modevent, 2247 0 2248}; 2249DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 2250 2251#endif /* MROUTE_KLD */ 2252#endif /* MROUTING */ 2253