ip_mroute.c revision 41878
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 * $Id: ip_mroute.c,v 1.50 1998/12/07 21:58:41 archie Exp $ 13 */ 14 15#include "opt_mrouting.h" 16 17#include <sys/param.h> 18#include <sys/systm.h> 19#include <sys/mbuf.h> 20#include <sys/socket.h> 21#include <sys/socketvar.h> 22#include <sys/protosw.h> 23#include <sys/time.h> 24#include <sys/kernel.h> 25#include <sys/sockio.h> 26#include <sys/syslog.h> 27#include <net/if.h> 28#include <net/route.h> 29#include <netinet/in.h> 30#include <netinet/in_systm.h> 31#include <netinet/ip.h> 32#include <netinet/ip_var.h> 33#include <netinet/in_var.h> 34#include <netinet/igmp.h> 35#include <netinet/ip_mroute.h> 36#include <netinet/udp.h> 37 38#ifndef NTOHL 39#if BYTE_ORDER != BIG_ENDIAN 40#define NTOHL(d) ((d) = ntohl((d))) 41#define NTOHS(d) ((d) = ntohs((u_short)(d))) 42#define HTONL(d) ((d) = htonl((d))) 43#define HTONS(d) ((d) = htons((u_short)(d))) 44#else 45#define NTOHL(d) 46#define NTOHS(d) 47#define HTONL(d) 48#define HTONS(d) 49#endif 50#endif 51 52#ifndef MROUTING 53extern u_long _ip_mcast_src __P((int vifi)); 54extern int _ip_mforward __P((struct ip *ip, struct ifnet *ifp, 55 struct mbuf *m, struct ip_moptions *imo)); 56extern int _ip_mrouter_done __P((void)); 57extern int _ip_mrouter_get __P((struct socket *so, struct sockopt *sopt)); 58extern int _ip_mrouter_set __P((struct socket *so, struct sockopt *sopt)); 59extern int _mrt_ioctl __P((int req, caddr_t data, struct proc *p)); 60 61/* 62 * Dummy routines and globals used when multicast routing is not compiled in. 63 */ 64 65struct socket *ip_mrouter = NULL; 66static u_int ip_mrtproto = 0; 67static struct mrtstat mrtstat; 68u_int rsvpdebug = 0; 69 70int 71_ip_mrouter_set(so, sopt) 72 struct socket *so; 73 struct sockopt *sopt; 74{ 75 return(EOPNOTSUPP); 76} 77 78int (*ip_mrouter_set)(struct socket *, struct sockopt *) = _ip_mrouter_set; 79 80 81int 82_ip_mrouter_get(so, sopt) 83 struct socket *so; 84 struct sockopt *sopt; 85{ 86 return(EOPNOTSUPP); 87} 88 89int (*ip_mrouter_get)(struct socket *, struct sockopt *) = _ip_mrouter_get; 90 91int 92_ip_mrouter_done() 93{ 94 return(0); 95} 96 97int (*ip_mrouter_done)(void) = _ip_mrouter_done; 98 99int 100_ip_mforward(ip, ifp, m, imo) 101 struct ip *ip; 102 struct ifnet *ifp; 103 struct mbuf *m; 104 struct ip_moptions *imo; 105{ 106 return(0); 107} 108 109int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 110 struct ip_moptions *) = _ip_mforward; 111 112int 113_mrt_ioctl(int req, caddr_t data, struct proc *p) 114{ 115 return EOPNOTSUPP; 116} 117 118int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl; 119 120void 121rsvp_input(m, iphlen) /* XXX must fixup manually */ 122 struct mbuf *m; 123 int iphlen; 124{ 125 /* Can still get packets with rsvp_on = 0 if there is a local member 126 * of the group to which the RSVP packet is addressed. But in this 127 * case we want to throw the packet away. 128 */ 129 if (!rsvp_on) { 130 m_freem(m); 131 return; 132 } 133 134 if (ip_rsvpd != NULL) { 135 if (rsvpdebug) 136 printf("rsvp_input: Sending packet up old-style socket\n"); 137 rip_input(m, iphlen); 138 return; 139 } 140 /* Drop the packet */ 141 m_freem(m); 142} 143 144void ipip_input(struct mbuf *m, int iphlen) { /* XXX must fixup manually */ 145 rip_input(m, iphlen); 146} 147 148int (*legal_vif_num)(int) = 0; 149 150/* 151 * This should never be called, since IP_MULTICAST_VIF should fail, but 152 * just in case it does get called, the code a little lower in ip_output 153 * will assign the packet a local address. 154 */ 155u_long 156_ip_mcast_src(int vifi) { return INADDR_ANY; } 157u_long (*ip_mcast_src)(int) = _ip_mcast_src; 158 159int 160ip_rsvp_vif_init(so, sopt) 161 struct socket *so; 162 struct sockopt *sopt; 163{ 164 return(EINVAL); 165} 166 167int 168ip_rsvp_vif_done(so, sopt) 169 struct socket *so; 170 struct sockopt *sopt; 171{ 172 return(EINVAL); 173} 174 175void 176ip_rsvp_force_done(so) 177 struct socket *so; 178{ 179 return; 180} 181 182#else /* MROUTING */ 183 184#define M_HASCL(m) ((m)->m_flags & M_EXT) 185 186#define INSIZ sizeof(struct in_addr) 187#define same(a1, a2) \ 188 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0) 189 190#define MT_MRTABLE MT_RTABLE /* since nothing else uses it */ 191 192/* 193 * Globals. All but ip_mrouter and ip_mrtproto could be static, 194 * except for netstat or debugging purposes. 195 */ 196#ifndef MROUTE_LKM 197struct socket *ip_mrouter = NULL; 198static struct mrtstat mrtstat; 199 200static int ip_mrtproto = IGMP_DVMRP; /* for netstat only */ 201#else /* MROUTE_LKM */ 202extern void X_ipip_input __P((struct mbuf *m, int iphlen)); 203extern struct mrtstat mrtstat; 204static int ip_mrtproto; 205#endif 206 207#define NO_RTE_FOUND 0x1 208#define RTE_FOUND 0x2 209 210static struct mbuf *mfctable[MFCTBLSIZ]; 211static u_char nexpire[MFCTBLSIZ]; 212static struct vif viftable[MAXVIFS]; 213static u_int mrtdebug = 0; /* debug level */ 214#define DEBUG_MFC 0x02 215#define DEBUG_FORWARD 0x04 216#define DEBUG_EXPIRE 0x08 217#define DEBUG_XMIT 0x10 218static u_int tbfdebug = 0; /* tbf debug level */ 219static u_int rsvpdebug = 0; /* rsvp debug level */ 220 221static struct callout_handle expire_upcalls_ch; 222 223#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 224#define UPCALL_EXPIRE 6 /* number of timeouts */ 225 226/* 227 * Define the token bucket filter structures 228 * tbftable -> each vif has one of these for storing info 229 */ 230 231static struct tbf tbftable[MAXVIFS]; 232#define TBF_REPROCESS (hz / 100) /* 100x / second */ 233 234/* 235 * 'Interfaces' associated with decapsulator (so we can tell 236 * packets that went through it from ones that get reflected 237 * by a broken gateway). These interfaces are never linked into 238 * the system ifnet list & no routes point to them. I.e., packets 239 * can't be sent this way. They only exist as a placeholder for 240 * multicast source verification. 241 */ 242static struct ifnet multicast_decap_if[MAXVIFS]; 243 244#define ENCAP_TTL 64 245#define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 246 247/* prototype IP hdr for encapsulated packets */ 248static struct ip multicast_encap_iphdr = { 249#if BYTE_ORDER == LITTLE_ENDIAN 250 sizeof(struct ip) >> 2, IPVERSION, 251#else 252 IPVERSION, sizeof(struct ip) >> 2, 253#endif 254 0, /* tos */ 255 sizeof(struct ip), /* total length */ 256 0, /* id */ 257 0, /* frag offset */ 258 ENCAP_TTL, ENCAP_PROTO, 259 0, /* checksum */ 260}; 261 262/* 263 * Private variables. 264 */ 265static vifi_t numvifs = 0; 266static int have_encap_tunnel = 0; 267 268/* 269 * one-back cache used by ipip_input to locate a tunnel's vif 270 * given a datagram's src ip address. 271 */ 272static u_long last_encap_src; 273static struct vif *last_encap_vif; 274 275static u_long X_ip_mcast_src __P((int vifi)); 276static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo)); 277static int X_ip_mrouter_done __P((void)); 278static int X_ip_mrouter_get __P((struct socket *so, struct sockopt *m)); 279static int X_ip_mrouter_set __P((struct socket *so, struct sockopt *m)); 280static int X_legal_vif_num __P((int vif)); 281static int X_mrt_ioctl __P((int cmd, caddr_t data)); 282 283static int get_sg_cnt(struct sioc_sg_req *); 284static int get_vif_cnt(struct sioc_vif_req *); 285static int ip_mrouter_init(struct socket *, int); 286static int add_vif(struct vifctl *); 287static int del_vif(vifi_t); 288static int add_mfc(struct mfcctl *); 289static int del_mfc(struct mfcctl *); 290static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 291static int set_assert(int); 292static void expire_upcalls(void *); 293static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, 294 vifi_t); 295static void phyint_send(struct ip *, struct vif *, struct mbuf *); 296static void encap_send(struct ip *, struct vif *, struct mbuf *); 297static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 298static void tbf_queue(struct vif *, struct mbuf *); 299static void tbf_process_q(struct vif *); 300static void tbf_reprocess_q(void *); 301static int tbf_dq_sel(struct vif *, struct ip *); 302static void tbf_send_packet(struct vif *, struct mbuf *); 303static void tbf_update_tokens(struct vif *); 304static int priority(struct vif *, struct ip *); 305void multiencap_decap(struct mbuf *); 306 307/* 308 * whether or not special PIM assert processing is enabled. 309 */ 310static int pim_assert; 311/* 312 * Rate limit for assert notification messages, in usec 313 */ 314#define ASSERT_MSG_TIME 3000000 315 316/* 317 * Hash function for a source, group entry 318 */ 319#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 320 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 321 322/* 323 * Find a route for a given origin IP address and Multicast group address 324 * Type of service parameter to be added in the future!!! 325 */ 326 327#define MFCFIND(o, g, rt) { \ 328 register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \ 329 register struct mfc *_rt = NULL; \ 330 rt = NULL; \ 331 ++mrtstat.mrts_mfc_lookups; \ 332 while (_mb_rt) { \ 333 _rt = mtod(_mb_rt, struct mfc *); \ 334 if ((_rt->mfc_origin.s_addr == o) && \ 335 (_rt->mfc_mcastgrp.s_addr == g) && \ 336 (_mb_rt->m_act == NULL)) { \ 337 rt = _rt; \ 338 break; \ 339 } \ 340 _mb_rt = _mb_rt->m_next; \ 341 } \ 342 if (rt == NULL) { \ 343 ++mrtstat.mrts_mfc_misses; \ 344 } \ 345} 346 347 348/* 349 * Macros to compute elapsed time efficiently 350 * Borrowed from Van Jacobson's scheduling code 351 */ 352#define TV_DELTA(a, b, delta) { \ 353 register int xxs; \ 354 \ 355 delta = (a).tv_usec - (b).tv_usec; \ 356 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 357 switch (xxs) { \ 358 case 2: \ 359 delta += 1000000; \ 360 /* fall through */ \ 361 case 1: \ 362 delta += 1000000; \ 363 break; \ 364 default: \ 365 delta += (1000000 * xxs); \ 366 } \ 367 } \ 368} 369 370#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 371 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 372 373#ifdef UPCALL_TIMING 374u_long upcall_data[51]; 375static void collate(struct timeval *); 376#endif /* UPCALL_TIMING */ 377 378 379/* 380 * Handle MRT setsockopt commands to modify the multicast routing tables. 381 */ 382static int 383X_ip_mrouter_set(so, sopt) 384 struct socket *so; 385 struct sockopt *sopt; 386{ 387 int error, optval; 388 vifi_t vifi; 389 struct vifctl vifc; 390 struct mfcctl mfc; 391 392 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 393 return (EPERM); 394 395 error = 0; 396 switch (sopt->sopt_name) { 397 case MRT_INIT: 398 error = sooptcopyin(sopt, &optval, sizeof optval, 399 sizeof optval); 400 if (error) 401 break; 402 error = ip_mrouter_init(so, optval); 403 break; 404 405 case MRT_DONE: 406 error = ip_mrouter_done(); 407 break; 408 409 case MRT_ADD_VIF: 410 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 411 if (error) 412 break; 413 error = add_vif(&vifc); 414 break; 415 416 case MRT_DEL_VIF: 417 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 418 if (error) 419 break; 420 error = del_vif(vifi); 421 break; 422 423 case MRT_ADD_MFC: 424 case MRT_DEL_MFC: 425 error = sooptcopyin(sopt, &mfc, sizeof mfc, sizeof mfc); 426 if (error) 427 break; 428 if (sopt->sopt_name == MRT_ADD_MFC) 429 error = add_mfc(&mfc); 430 else 431 error = del_mfc(&mfc); 432 break; 433 434 case MRT_ASSERT: 435 error = sooptcopyin(sopt, &optval, sizeof optval, 436 sizeof optval); 437 if (error) 438 break; 439 set_assert(optval); 440 break; 441 442 default: 443 error = EOPNOTSUPP; 444 break; 445 } 446 return (error); 447} 448 449#ifndef MROUTE_LKM 450int (*ip_mrouter_set)(struct socket *, struct sockopt *) = X_ip_mrouter_set; 451#endif 452 453/* 454 * Handle MRT getsockopt commands 455 */ 456static int 457X_ip_mrouter_get(so, sopt) 458 struct socket *so; 459 struct sockopt *sopt; 460{ 461 int error; 462 static int version = 0x0305; /* !!! why is this here? XXX */ 463 464 switch (sopt->sopt_name) { 465 case MRT_VERSION: 466 error = sooptcopyout(sopt, &version, sizeof version); 467 break; 468 469 case MRT_ASSERT: 470 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 471 break; 472 default: 473 error = EOPNOTSUPP; 474 break; 475 } 476 return (error); 477} 478 479#ifndef MROUTE_LKM 480int (*ip_mrouter_get)(struct socket *, struct sockopt *) = X_ip_mrouter_get; 481#endif 482 483/* 484 * Handle ioctl commands to obtain information from the cache 485 */ 486static int 487X_mrt_ioctl(cmd, data) 488 int cmd; 489 caddr_t data; 490{ 491 int error = 0; 492 493 switch (cmd) { 494 case (SIOCGETVIFCNT): 495 return (get_vif_cnt((struct sioc_vif_req *)data)); 496 break; 497 case (SIOCGETSGCNT): 498 return (get_sg_cnt((struct sioc_sg_req *)data)); 499 break; 500 default: 501 return (EINVAL); 502 break; 503 } 504 return error; 505} 506 507#ifndef MROUTE_LKM 508int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl; 509#endif 510 511/* 512 * returns the packet, byte, rpf-failure count for the source group provided 513 */ 514static int 515get_sg_cnt(req) 516 register struct sioc_sg_req *req; 517{ 518 register struct mfc *rt; 519 int s; 520 521 s = splnet(); 522 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 523 splx(s); 524 if (rt != NULL) { 525 req->pktcnt = rt->mfc_pkt_cnt; 526 req->bytecnt = rt->mfc_byte_cnt; 527 req->wrong_if = rt->mfc_wrong_if; 528 } else 529 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 530 531 return 0; 532} 533 534/* 535 * returns the input and output packet and byte counts on the vif provided 536 */ 537static int 538get_vif_cnt(req) 539 register struct sioc_vif_req *req; 540{ 541 register vifi_t vifi = req->vifi; 542 543 if (vifi >= numvifs) return EINVAL; 544 545 req->icount = viftable[vifi].v_pkt_in; 546 req->ocount = viftable[vifi].v_pkt_out; 547 req->ibytes = viftable[vifi].v_bytes_in; 548 req->obytes = viftable[vifi].v_bytes_out; 549 550 return 0; 551} 552 553/* 554 * Enable multicast routing 555 */ 556static int 557ip_mrouter_init(so, version) 558 struct socket *so; 559 int version; 560{ 561 if (mrtdebug) 562 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 563 so->so_type, so->so_proto->pr_protocol); 564 565 if (so->so_type != SOCK_RAW || 566 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP; 567 568 if (version != 1) 569 return ENOPROTOOPT; 570 571 if (ip_mrouter != NULL) return EADDRINUSE; 572 573 ip_mrouter = so; 574 575 bzero((caddr_t)mfctable, sizeof(mfctable)); 576 bzero((caddr_t)nexpire, sizeof(nexpire)); 577 578 pim_assert = 0; 579 580 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 581 582 if (mrtdebug) 583 log(LOG_DEBUG, "ip_mrouter_init\n"); 584 585 return 0; 586} 587 588/* 589 * Disable multicast routing 590 */ 591static int 592X_ip_mrouter_done() 593{ 594 vifi_t vifi; 595 int i; 596 struct ifnet *ifp; 597 struct ifreq ifr; 598 struct mbuf *mb_rt; 599 struct mbuf *m; 600 struct rtdetq *rte; 601 int s; 602 603 s = splnet(); 604 605 /* 606 * For each phyint in use, disable promiscuous reception of all IP 607 * multicasts. 608 */ 609 for (vifi = 0; vifi < numvifs; vifi++) { 610 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 611 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 612 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 613 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr 614 = INADDR_ANY; 615 ifp = viftable[vifi].v_ifp; 616 if_allmulti(ifp, 0); 617 } 618 } 619 bzero((caddr_t)tbftable, sizeof(tbftable)); 620 bzero((caddr_t)viftable, sizeof(viftable)); 621 numvifs = 0; 622 pim_assert = 0; 623 624 untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch); 625 626 /* 627 * Free all multicast forwarding cache entries. 628 */ 629 for (i = 0; i < MFCTBLSIZ; i++) { 630 mb_rt = mfctable[i]; 631 while (mb_rt) { 632 if (mb_rt->m_act != NULL) { 633 while (mb_rt->m_act) { 634 m = mb_rt->m_act; 635 mb_rt->m_act = m->m_act; 636 rte = mtod(m, struct rtdetq *); 637 m_freem(rte->m); 638 m_free(m); 639 } 640 } 641 mb_rt = m_free(mb_rt); 642 } 643 } 644 645 bzero((caddr_t)mfctable, sizeof(mfctable)); 646 647 /* 648 * Reset de-encapsulation cache 649 */ 650 last_encap_src = 0; 651 last_encap_vif = NULL; 652 have_encap_tunnel = 0; 653 654 ip_mrouter = NULL; 655 656 splx(s); 657 658 if (mrtdebug) 659 log(LOG_DEBUG, "ip_mrouter_done\n"); 660 661 return 0; 662} 663 664#ifndef MROUTE_LKM 665int (*ip_mrouter_done)(void) = X_ip_mrouter_done; 666#endif 667 668/* 669 * Set PIM assert processing global 670 */ 671static int 672set_assert(i) 673 int i; 674{ 675 if ((i != 1) && (i != 0)) 676 return EINVAL; 677 678 pim_assert = i; 679 680 return 0; 681} 682 683/* 684 * Add a vif to the vif table 685 */ 686static int 687add_vif(vifcp) 688 register struct vifctl *vifcp; 689{ 690 register struct vif *vifp = viftable + vifcp->vifc_vifi; 691 static struct sockaddr_in sin = {sizeof sin, AF_INET}; 692 struct ifaddr *ifa; 693 struct ifnet *ifp; 694 int error, s; 695 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 696 697 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL; 698 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE; 699 700 /* Find the interface with an address in AF_INET family */ 701 sin.sin_addr = vifcp->vifc_lcl_addr; 702 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 703 if (ifa == 0) return EADDRNOTAVAIL; 704 ifp = ifa->ifa_ifp; 705 706 if (vifcp->vifc_flags & VIFF_TUNNEL) { 707 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 708 /* 709 * An encapsulating tunnel is wanted. Tell ipip_input() to 710 * start paying attention to encapsulated packets. 711 */ 712 if (have_encap_tunnel == 0) { 713 have_encap_tunnel = 1; 714 for (s = 0; s < MAXVIFS; ++s) { 715 multicast_decap_if[s].if_name = "mdecap"; 716 multicast_decap_if[s].if_unit = s; 717 } 718 } 719 /* 720 * Set interface to fake encapsulator interface 721 */ 722 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 723 /* 724 * Prepare cached route entry 725 */ 726 bzero(&vifp->v_route, sizeof(vifp->v_route)); 727 } else { 728 log(LOG_ERR, "source routed tunnels not supported\n"); 729 return EOPNOTSUPP; 730 } 731 } else { 732 /* Make sure the interface supports multicast */ 733 if ((ifp->if_flags & IFF_MULTICAST) == 0) 734 return EOPNOTSUPP; 735 736 /* Enable promiscuous reception of all IP multicasts from the if */ 737 s = splnet(); 738 error = if_allmulti(ifp, 1); 739 splx(s); 740 if (error) 741 return error; 742 } 743 744 s = splnet(); 745 /* define parameters for the tbf structure */ 746 vifp->v_tbf = v_tbf; 747 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 748 vifp->v_tbf->tbf_n_tok = 0; 749 vifp->v_tbf->tbf_q_len = 0; 750 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 751 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 752 753 vifp->v_flags = vifcp->vifc_flags; 754 vifp->v_threshold = vifcp->vifc_threshold; 755 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 756 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 757 vifp->v_ifp = ifp; 758 /* scaling up here allows division by 1024 in critical code */ 759 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 760 vifp->v_rsvp_on = 0; 761 vifp->v_rsvpd = NULL; 762 /* initialize per vif pkt counters */ 763 vifp->v_pkt_in = 0; 764 vifp->v_pkt_out = 0; 765 vifp->v_bytes_in = 0; 766 vifp->v_bytes_out = 0; 767 splx(s); 768 769 /* Adjust numvifs up if the vifi is higher than numvifs */ 770 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 771 772 if (mrtdebug) 773 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n", 774 vifcp->vifc_vifi, 775 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 776 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 777 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 778 vifcp->vifc_threshold, 779 vifcp->vifc_rate_limit); 780 781 return 0; 782} 783 784/* 785 * Delete a vif from the vif table 786 */ 787static int 788del_vif(vifi) 789 vifi_t vifi; 790{ 791 register struct vif *vifp = &viftable[vifi]; 792 register struct mbuf *m; 793 struct ifnet *ifp; 794 struct ifreq ifr; 795 int s; 796 797 if (vifi >= numvifs) return EINVAL; 798 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL; 799 800 s = splnet(); 801 802 if (!(vifp->v_flags & VIFF_TUNNEL)) { 803 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 804 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 805 ifp = vifp->v_ifp; 806 if_allmulti(ifp, 0); 807 } 808 809 if (vifp == last_encap_vif) { 810 last_encap_vif = 0; 811 last_encap_src = 0; 812 } 813 814 /* 815 * Free packets queued at the interface 816 */ 817 while (vifp->v_tbf->tbf_q) { 818 m = vifp->v_tbf->tbf_q; 819 vifp->v_tbf->tbf_q = m->m_act; 820 m_freem(m); 821 } 822 823 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 824 bzero((caddr_t)vifp, sizeof (*vifp)); 825 826 if (mrtdebug) 827 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 828 829 /* Adjust numvifs down */ 830 for (vifi = numvifs; vifi > 0; vifi--) 831 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break; 832 numvifs = vifi; 833 834 splx(s); 835 836 return 0; 837} 838 839/* 840 * Add an mfc entry 841 */ 842static int 843add_mfc(mfccp) 844 struct mfcctl *mfccp; 845{ 846 struct mfc *rt; 847 register struct mbuf *mb_rt; 848 u_long hash; 849 struct mbuf *mb_ntry; 850 struct rtdetq *rte; 851 register u_short nstl; 852 int s; 853 int i; 854 855 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 856 857 /* If an entry already exists, just update the fields */ 858 if (rt) { 859 if (mrtdebug & DEBUG_MFC) 860 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 861 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 862 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 863 mfccp->mfcc_parent); 864 865 s = splnet(); 866 rt->mfc_parent = mfccp->mfcc_parent; 867 for (i = 0; i < numvifs; i++) 868 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 869 splx(s); 870 return 0; 871 } 872 873 /* 874 * Find the entry for which the upcall was made and update 875 */ 876 s = splnet(); 877 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 878 for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) { 879 880 rt = mtod(mb_rt, struct mfc *); 881 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 882 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 883 (mb_rt->m_act != NULL)) { 884 885 if (nstl++) 886 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 887 "multiple kernel entries", 888 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 889 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 890 mfccp->mfcc_parent, (void *)mb_rt->m_act); 891 892 if (mrtdebug & DEBUG_MFC) 893 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 894 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 895 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 896 mfccp->mfcc_parent, (void *)mb_rt->m_act); 897 898 rt->mfc_origin = mfccp->mfcc_origin; 899 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 900 rt->mfc_parent = mfccp->mfcc_parent; 901 for (i = 0; i < numvifs; i++) 902 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 903 /* initialize pkt counters per src-grp */ 904 rt->mfc_pkt_cnt = 0; 905 rt->mfc_byte_cnt = 0; 906 rt->mfc_wrong_if = 0; 907 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 908 909 rt->mfc_expire = 0; /* Don't clean this guy up */ 910 nexpire[hash]--; 911 912 /* free packets Qed at the end of this entry */ 913 while (mb_rt->m_act) { 914 mb_ntry = mb_rt->m_act; 915 rte = mtod(mb_ntry, struct rtdetq *); 916/* #ifdef RSVP_ISI */ 917 ip_mdq(rte->m, rte->ifp, rt, -1); 918/* #endif */ 919 mb_rt->m_act = mb_ntry->m_act; 920 m_freem(rte->m); 921#ifdef UPCALL_TIMING 922 collate(&(rte->t)); 923#endif /* UPCALL_TIMING */ 924 m_free(mb_ntry); 925 } 926 } 927 } 928 929 /* 930 * It is possible that an entry is being inserted without an upcall 931 */ 932 if (nstl == 0) { 933 if (mrtdebug & DEBUG_MFC) 934 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 935 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 936 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 937 mfccp->mfcc_parent); 938 939 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 940 941 rt = mtod(mb_rt, struct mfc *); 942 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 943 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 944 945 rt->mfc_origin = mfccp->mfcc_origin; 946 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 947 rt->mfc_parent = mfccp->mfcc_parent; 948 for (i = 0; i < numvifs; i++) 949 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 950 /* initialize pkt counters per src-grp */ 951 rt->mfc_pkt_cnt = 0; 952 rt->mfc_byte_cnt = 0; 953 rt->mfc_wrong_if = 0; 954 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 955 if (rt->mfc_expire) 956 nexpire[hash]--; 957 rt->mfc_expire = 0; 958 } 959 } 960 if (mb_rt == NULL) { 961 /* no upcall, so make a new entry */ 962 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 963 if (mb_rt == NULL) { 964 splx(s); 965 return ENOBUFS; 966 } 967 968 rt = mtod(mb_rt, struct mfc *); 969 970 /* insert new entry at head of hash chain */ 971 rt->mfc_origin = mfccp->mfcc_origin; 972 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 973 rt->mfc_parent = mfccp->mfcc_parent; 974 for (i = 0; i < numvifs; i++) 975 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 976 /* initialize pkt counters per src-grp */ 977 rt->mfc_pkt_cnt = 0; 978 rt->mfc_byte_cnt = 0; 979 rt->mfc_wrong_if = 0; 980 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 981 rt->mfc_expire = 0; 982 983 /* link into table */ 984 mb_rt->m_next = mfctable[hash]; 985 mfctable[hash] = mb_rt; 986 mb_rt->m_act = NULL; 987 } 988 } 989 splx(s); 990 return 0; 991} 992 993#ifdef UPCALL_TIMING 994/* 995 * collect delay statistics on the upcalls 996 */ 997static void collate(t) 998register struct timeval *t; 999{ 1000 register u_long d; 1001 register struct timeval tp; 1002 register u_long delta; 1003 1004 GET_TIME(tp); 1005 1006 if (TV_LT(*t, tp)) 1007 { 1008 TV_DELTA(tp, *t, delta); 1009 1010 d = delta >> 10; 1011 if (d > 50) 1012 d = 50; 1013 1014 ++upcall_data[d]; 1015 } 1016} 1017#endif /* UPCALL_TIMING */ 1018 1019/* 1020 * Delete an mfc entry 1021 */ 1022static int 1023del_mfc(mfccp) 1024 struct mfcctl *mfccp; 1025{ 1026 struct in_addr origin; 1027 struct in_addr mcastgrp; 1028 struct mfc *rt; 1029 struct mbuf *mb_rt; 1030 struct mbuf **nptr; 1031 u_long hash; 1032 int s; 1033 1034 origin = mfccp->mfcc_origin; 1035 mcastgrp = mfccp->mfcc_mcastgrp; 1036 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1037 1038 if (mrtdebug & DEBUG_MFC) 1039 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1040 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1041 1042 s = splnet(); 1043 1044 nptr = &mfctable[hash]; 1045 while ((mb_rt = *nptr) != NULL) { 1046 rt = mtod(mb_rt, struct mfc *); 1047 if (origin.s_addr == rt->mfc_origin.s_addr && 1048 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1049 mb_rt->m_act == NULL) 1050 break; 1051 1052 nptr = &mb_rt->m_next; 1053 } 1054 if (mb_rt == NULL) { 1055 splx(s); 1056 return EADDRNOTAVAIL; 1057 } 1058 1059 MFREE(mb_rt, *nptr); 1060 1061 splx(s); 1062 1063 return 0; 1064} 1065 1066/* 1067 * Send a message to mrouted on the multicast routing socket 1068 */ 1069static int 1070socket_send(s, mm, src) 1071 struct socket *s; 1072 struct mbuf *mm; 1073 struct sockaddr_in *src; 1074{ 1075 if (s) { 1076 if (sbappendaddr(&s->so_rcv, 1077 (struct sockaddr *)src, 1078 mm, (struct mbuf *)0) != 0) { 1079 sorwakeup(s); 1080 return 0; 1081 } 1082 } 1083 m_freem(mm); 1084 return -1; 1085} 1086 1087/* 1088 * IP multicast forwarding function. This function assumes that the packet 1089 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1090 * pointed to by "ifp", and the packet is to be relayed to other networks 1091 * that have members of the packet's destination IP multicast group. 1092 * 1093 * The packet is returned unscathed to the caller, unless it is 1094 * erroneous, in which case a non-zero return value tells the caller to 1095 * discard it. 1096 */ 1097 1098#define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1099#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1100 1101static int 1102X_ip_mforward(ip, ifp, m, imo) 1103 register struct ip *ip; 1104 struct ifnet *ifp; 1105 struct mbuf *m; 1106 struct ip_moptions *imo; 1107{ 1108 register struct mfc *rt; 1109 register u_char *ipoptions; 1110 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1111 static int srctun = 0; 1112 register struct mbuf *mm; 1113 int s; 1114 vifi_t vifi; 1115 struct vif *vifp; 1116 1117 if (mrtdebug & DEBUG_FORWARD) 1118 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1119 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1120 (void *)ifp); 1121 1122 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1123 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1124 /* 1125 * Packet arrived via a physical interface or 1126 * an encapsulated tunnel. 1127 */ 1128 } else { 1129 /* 1130 * Packet arrived through a source-route tunnel. 1131 * Source-route tunnels are no longer supported. 1132 */ 1133 if ((srctun++ % 1000) == 0) 1134 log(LOG_ERR, 1135 "ip_mforward: received source-routed packet from %lx\n", 1136 (u_long)ntohl(ip->ip_src.s_addr)); 1137 1138 return 1; 1139 } 1140 1141 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1142 if (ip->ip_ttl < 255) 1143 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1144 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1145 vifp = viftable + vifi; 1146 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1147 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi, 1148 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1149 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1150 } 1151 return (ip_mdq(m, ifp, NULL, vifi)); 1152 } 1153 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1154 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1155 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr)); 1156 if(!imo) 1157 printf("In fact, no options were specified at all\n"); 1158 } 1159 1160 /* 1161 * Don't forward a packet with time-to-live of zero or one, 1162 * or a packet destined to a local-only group. 1163 */ 1164 if (ip->ip_ttl <= 1 || 1165 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1166 return 0; 1167 1168 /* 1169 * Determine forwarding vifs from the forwarding cache table 1170 */ 1171 s = splnet(); 1172 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt); 1173 1174 /* Entry exists, so forward if necessary */ 1175 if (rt != NULL) { 1176 splx(s); 1177 return (ip_mdq(m, ifp, rt, -1)); 1178 } else { 1179 /* 1180 * If we don't have a route for packet's origin, 1181 * Make a copy of the packet & 1182 * send message to routing daemon 1183 */ 1184 1185 register struct mbuf *mb_rt; 1186 register struct mbuf *mb_ntry; 1187 register struct mbuf *mb0; 1188 register struct rtdetq *rte; 1189 register struct mbuf *rte_m; 1190 register u_long hash; 1191 register int npkts; 1192 int hlen = ip->ip_hl << 2; 1193#ifdef UPCALL_TIMING 1194 struct timeval tp; 1195 1196 GET_TIME(tp); 1197#endif 1198 1199 mrtstat.mrts_no_route++; 1200 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1201 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1202 (u_long)ntohl(ip->ip_src.s_addr), 1203 (u_long)ntohl(ip->ip_dst.s_addr)); 1204 1205 /* 1206 * Allocate mbufs early so that we don't do extra work if we are 1207 * just going to fail anyway. Make sure to pullup the header so 1208 * that other people can't step on it. 1209 */ 1210 MGET(mb_ntry, M_DONTWAIT, MT_DATA); 1211 if (mb_ntry == NULL) { 1212 splx(s); 1213 return ENOBUFS; 1214 } 1215 mb0 = m_copy(m, 0, M_COPYALL); 1216 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1217 mb0 = m_pullup(mb0, hlen); 1218 if (mb0 == NULL) { 1219 m_free(mb_ntry); 1220 splx(s); 1221 return ENOBUFS; 1222 } 1223 1224 /* is there an upcall waiting for this packet? */ 1225 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1226 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 1227 rt = mtod(mb_rt, struct mfc *); 1228 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1229 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1230 (mb_rt->m_act != NULL)) 1231 break; 1232 } 1233 1234 if (mb_rt == NULL) { 1235 int i; 1236 struct igmpmsg *im; 1237 1238 /* no upcall, so make a new entry */ 1239 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 1240 if (mb_rt == NULL) { 1241 m_free(mb_ntry); 1242 m_freem(mb0); 1243 splx(s); 1244 return ENOBUFS; 1245 } 1246 /* Make a copy of the header to send to the user level process */ 1247 mm = m_copy(mb0, 0, hlen); 1248 if (mm == NULL) { 1249 m_free(mb_ntry); 1250 m_freem(mb0); 1251 m_free(mb_rt); 1252 splx(s); 1253 return ENOBUFS; 1254 } 1255 1256 /* 1257 * Send message to routing daemon to install 1258 * a route into the kernel table 1259 */ 1260 k_igmpsrc.sin_addr = ip->ip_src; 1261 1262 im = mtod(mm, struct igmpmsg *); 1263 im->im_msgtype = IGMPMSG_NOCACHE; 1264 im->im_mbz = 0; 1265 1266 mrtstat.mrts_upcalls++; 1267 1268 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1269 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1270 ++mrtstat.mrts_upq_sockfull; 1271 m_free(mb_ntry); 1272 m_freem(mb0); 1273 m_free(mb_rt); 1274 splx(s); 1275 return ENOBUFS; 1276 } 1277 1278 rt = mtod(mb_rt, struct mfc *); 1279 1280 /* insert new entry at head of hash chain */ 1281 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1282 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1283 rt->mfc_expire = UPCALL_EXPIRE; 1284 nexpire[hash]++; 1285 for (i = 0; i < numvifs; i++) 1286 rt->mfc_ttls[i] = 0; 1287 rt->mfc_parent = -1; 1288 1289 /* link into table */ 1290 mb_rt->m_next = mfctable[hash]; 1291 mfctable[hash] = mb_rt; 1292 mb_rt->m_act = NULL; 1293 1294 rte_m = mb_rt; 1295 } else { 1296 /* determine if q has overflowed */ 1297 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act) 1298 npkts++; 1299 1300 if (npkts > MAX_UPQ) { 1301 mrtstat.mrts_upq_ovflw++; 1302 m_free(mb_ntry); 1303 m_freem(mb0); 1304 splx(s); 1305 return 0; 1306 } 1307 } 1308 1309 mb_ntry->m_act = NULL; 1310 rte = mtod(mb_ntry, struct rtdetq *); 1311 1312 rte->m = mb0; 1313 rte->ifp = ifp; 1314#ifdef UPCALL_TIMING 1315 rte->t = tp; 1316#endif 1317 1318 /* Add this entry to the end of the queue */ 1319 rte_m->m_act = mb_ntry; 1320 1321 splx(s); 1322 1323 return 0; 1324 } 1325} 1326 1327#ifndef MROUTE_LKM 1328int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 1329 struct ip_moptions *) = X_ip_mforward; 1330#endif 1331 1332/* 1333 * Clean up the cache entry if upcall is not serviced 1334 */ 1335static void 1336expire_upcalls(void *unused) 1337{ 1338 struct mbuf *mb_rt, *m, **nptr; 1339 struct rtdetq *rte; 1340 struct mfc *mfc; 1341 int i; 1342 int s; 1343 1344 s = splnet(); 1345 for (i = 0; i < MFCTBLSIZ; i++) { 1346 if (nexpire[i] == 0) 1347 continue; 1348 nptr = &mfctable[i]; 1349 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) { 1350 mfc = mtod(mb_rt, struct mfc *); 1351 1352 /* 1353 * Skip real cache entries 1354 * Make sure it wasn't marked to not expire (shouldn't happen) 1355 * If it expires now 1356 */ 1357 if (mb_rt->m_act != NULL && 1358 mfc->mfc_expire != 0 && 1359 --mfc->mfc_expire == 0) { 1360 if (mrtdebug & DEBUG_EXPIRE) 1361 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1362 (u_long)ntohl(mfc->mfc_origin.s_addr), 1363 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1364 /* 1365 * drop all the packets 1366 * free the mbuf with the pkt, if, timing info 1367 */ 1368 while (mb_rt->m_act) { 1369 m = mb_rt->m_act; 1370 mb_rt->m_act = m->m_act; 1371 1372 rte = mtod(m, struct rtdetq *); 1373 m_freem(rte->m); 1374 m_free(m); 1375 } 1376 ++mrtstat.mrts_cache_cleanups; 1377 nexpire[i]--; 1378 1379 MFREE(mb_rt, *nptr); 1380 } else { 1381 nptr = &mb_rt->m_next; 1382 } 1383 } 1384 } 1385 splx(s); 1386 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 1387} 1388 1389/* 1390 * Packet forwarding routine once entry in the cache is made 1391 */ 1392static int 1393ip_mdq(m, ifp, rt, xmt_vif) 1394 register struct mbuf *m; 1395 register struct ifnet *ifp; 1396 register struct mfc *rt; 1397 register vifi_t xmt_vif; 1398{ 1399 register struct ip *ip = mtod(m, struct ip *); 1400 register vifi_t vifi; 1401 register struct vif *vifp; 1402 register int plen = ip->ip_len; 1403 1404/* 1405 * Macro to send packet on vif. Since RSVP packets don't get counted on 1406 * input, they shouldn't get counted on output, so statistics keeping is 1407 * seperate. 1408 */ 1409#define MC_SEND(ip,vifp,m) { \ 1410 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1411 encap_send((ip), (vifp), (m)); \ 1412 else \ 1413 phyint_send((ip), (vifp), (m)); \ 1414} 1415 1416 /* 1417 * If xmt_vif is not -1, send on only the requested vif. 1418 * 1419 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1420 */ 1421 if (xmt_vif < numvifs) { 1422 MC_SEND(ip, viftable + xmt_vif, m); 1423 return 1; 1424 } 1425 1426 /* 1427 * Don't forward if it didn't arrive from the parent vif for its origin. 1428 */ 1429 vifi = rt->mfc_parent; 1430 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1431 /* came in the wrong interface */ 1432 if (mrtdebug & DEBUG_FORWARD) 1433 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1434 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1435 ++mrtstat.mrts_wrong_if; 1436 ++rt->mfc_wrong_if; 1437 /* 1438 * If we are doing PIM assert processing, and we are forwarding 1439 * packets on this interface, and it is a broadcast medium 1440 * interface (and not a tunnel), send a message to the routing daemon. 1441 */ 1442 if (pim_assert && rt->mfc_ttls[vifi] && 1443 (ifp->if_flags & IFF_BROADCAST) && 1444 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1445 struct sockaddr_in k_igmpsrc; 1446 struct mbuf *mm; 1447 struct igmpmsg *im; 1448 int hlen = ip->ip_hl << 2; 1449 struct timeval now; 1450 register u_long delta; 1451 1452 GET_TIME(now); 1453 1454 TV_DELTA(rt->mfc_last_assert, now, delta); 1455 1456 if (delta > ASSERT_MSG_TIME) { 1457 mm = m_copy(m, 0, hlen); 1458 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1459 mm = m_pullup(mm, hlen); 1460 if (mm == NULL) { 1461 return ENOBUFS; 1462 } 1463 1464 rt->mfc_last_assert = now; 1465 1466 im = mtod(mm, struct igmpmsg *); 1467 im->im_msgtype = IGMPMSG_WRONGVIF; 1468 im->im_mbz = 0; 1469 im->im_vif = vifi; 1470 1471 k_igmpsrc.sin_addr = im->im_src; 1472 1473 socket_send(ip_mrouter, mm, &k_igmpsrc); 1474 } 1475 } 1476 return 0; 1477 } 1478 1479 /* If I sourced this packet, it counts as output, else it was input. */ 1480 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1481 viftable[vifi].v_pkt_out++; 1482 viftable[vifi].v_bytes_out += plen; 1483 } else { 1484 viftable[vifi].v_pkt_in++; 1485 viftable[vifi].v_bytes_in += plen; 1486 } 1487 rt->mfc_pkt_cnt++; 1488 rt->mfc_byte_cnt += plen; 1489 1490 /* 1491 * For each vif, decide if a copy of the packet should be forwarded. 1492 * Forward if: 1493 * - the ttl exceeds the vif's threshold 1494 * - there are group members downstream on interface 1495 */ 1496 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1497 if ((rt->mfc_ttls[vifi] > 0) && 1498 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1499 vifp->v_pkt_out++; 1500 vifp->v_bytes_out += plen; 1501 MC_SEND(ip, vifp, m); 1502 } 1503 1504 return 0; 1505} 1506 1507/* 1508 * check if a vif number is legal/ok. This is used by ip_output, to export 1509 * numvifs there, 1510 */ 1511static int 1512X_legal_vif_num(vif) 1513 int vif; 1514{ 1515 if (vif >= 0 && vif < numvifs) 1516 return(1); 1517 else 1518 return(0); 1519} 1520 1521#ifndef MROUTE_LKM 1522int (*legal_vif_num)(int) = X_legal_vif_num; 1523#endif 1524 1525/* 1526 * Return the local address used by this vif 1527 */ 1528static u_long 1529X_ip_mcast_src(vifi) 1530 int vifi; 1531{ 1532 if (vifi >= 0 && vifi < numvifs) 1533 return viftable[vifi].v_lcl_addr.s_addr; 1534 else 1535 return INADDR_ANY; 1536} 1537 1538#ifndef MROUTE_LKM 1539u_long (*ip_mcast_src)(int) = X_ip_mcast_src; 1540#endif 1541 1542static void 1543phyint_send(ip, vifp, m) 1544 struct ip *ip; 1545 struct vif *vifp; 1546 struct mbuf *m; 1547{ 1548 register struct mbuf *mb_copy; 1549 register int hlen = ip->ip_hl << 2; 1550 1551 /* 1552 * Make a new reference to the packet; make sure that 1553 * the IP header is actually copied, not just referenced, 1554 * so that ip_output() only scribbles on the copy. 1555 */ 1556 mb_copy = m_copy(m, 0, M_COPYALL); 1557 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1558 mb_copy = m_pullup(mb_copy, hlen); 1559 if (mb_copy == NULL) 1560 return; 1561 1562 if (vifp->v_rate_limit <= 0) 1563 tbf_send_packet(vifp, mb_copy); 1564 else 1565 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1566} 1567 1568static void 1569encap_send(ip, vifp, m) 1570 register struct ip *ip; 1571 register struct vif *vifp; 1572 register struct mbuf *m; 1573{ 1574 register struct mbuf *mb_copy; 1575 register struct ip *ip_copy; 1576 register int i, len = ip->ip_len; 1577 1578 /* 1579 * copy the old packet & pullup its IP header into the 1580 * new mbuf so we can modify it. Try to fill the new 1581 * mbuf since if we don't the ethernet driver will. 1582 */ 1583 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER); 1584 if (mb_copy == NULL) 1585 return; 1586 mb_copy->m_data += max_linkhdr; 1587 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1588 1589 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1590 m_freem(mb_copy); 1591 return; 1592 } 1593 i = MHLEN - M_LEADINGSPACE(mb_copy); 1594 if (i > len) 1595 i = len; 1596 mb_copy = m_pullup(mb_copy, i); 1597 if (mb_copy == NULL) 1598 return; 1599 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1600 1601 /* 1602 * fill in the encapsulating IP header. 1603 */ 1604 ip_copy = mtod(mb_copy, struct ip *); 1605 *ip_copy = multicast_encap_iphdr; 1606 ip_copy->ip_id = htons(ip_id++); 1607 ip_copy->ip_len += len; 1608 ip_copy->ip_src = vifp->v_lcl_addr; 1609 ip_copy->ip_dst = vifp->v_rmt_addr; 1610 1611 /* 1612 * turn the encapsulated IP header back into a valid one. 1613 */ 1614 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1615 --ip->ip_ttl; 1616 HTONS(ip->ip_len); 1617 HTONS(ip->ip_off); 1618 ip->ip_sum = 0; 1619 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1620 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1621 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1622 1623 if (vifp->v_rate_limit <= 0) 1624 tbf_send_packet(vifp, mb_copy); 1625 else 1626 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1627} 1628 1629/* 1630 * De-encapsulate a packet and feed it back through ip input (this 1631 * routine is called whenever IP gets a packet with proto type 1632 * ENCAP_PROTO and a local destination address). 1633 */ 1634void 1635#ifdef MROUTE_LKM 1636X_ipip_input(m, iphlen) 1637#else 1638ipip_input(m, iphlen) 1639#endif 1640 register struct mbuf *m; 1641 int iphlen; 1642{ 1643 struct ifnet *ifp = m->m_pkthdr.rcvif; 1644 register struct ip *ip = mtod(m, struct ip *); 1645 register int hlen = ip->ip_hl << 2; 1646 register int s; 1647 register struct ifqueue *ifq; 1648 register struct vif *vifp; 1649 1650 if (!have_encap_tunnel) { 1651 rip_input(m, iphlen); 1652 return; 1653 } 1654 /* 1655 * dump the packet if it's not to a multicast destination or if 1656 * we don't have an encapsulating tunnel with the source. 1657 * Note: This code assumes that the remote site IP address 1658 * uniquely identifies the tunnel (i.e., that this site has 1659 * at most one tunnel with the remote site). 1660 */ 1661 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) { 1662 ++mrtstat.mrts_bad_tunnel; 1663 m_freem(m); 1664 return; 1665 } 1666 if (ip->ip_src.s_addr != last_encap_src) { 1667 register struct vif *vife; 1668 1669 vifp = viftable; 1670 vife = vifp + numvifs; 1671 last_encap_src = ip->ip_src.s_addr; 1672 last_encap_vif = 0; 1673 for ( ; vifp < vife; ++vifp) 1674 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1675 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1676 == VIFF_TUNNEL) 1677 last_encap_vif = vifp; 1678 break; 1679 } 1680 } 1681 if ((vifp = last_encap_vif) == 0) { 1682 last_encap_src = 0; 1683 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1684 m_freem(m); 1685 if (mrtdebug) 1686 log(LOG_DEBUG, "ip_mforward: no tunnel with %lx\n", 1687 (u_long)ntohl(ip->ip_src.s_addr)); 1688 return; 1689 } 1690 ifp = vifp->v_ifp; 1691 1692 if (hlen > IP_HDR_LEN) 1693 ip_stripoptions(m, (struct mbuf *) 0); 1694 m->m_data += IP_HDR_LEN; 1695 m->m_len -= IP_HDR_LEN; 1696 m->m_pkthdr.len -= IP_HDR_LEN; 1697 m->m_pkthdr.rcvif = ifp; 1698 1699 ifq = &ipintrq; 1700 s = splimp(); 1701 if (IF_QFULL(ifq)) { 1702 IF_DROP(ifq); 1703 m_freem(m); 1704 } else { 1705 IF_ENQUEUE(ifq, m); 1706 /* 1707 * normally we would need a "schednetisr(NETISR_IP)" 1708 * here but we were called by ip_input and it is going 1709 * to loop back & try to dequeue the packet we just 1710 * queued as soon as we return so we avoid the 1711 * unnecessary software interrrupt. 1712 */ 1713 } 1714 splx(s); 1715} 1716 1717/* 1718 * Token bucket filter module 1719 */ 1720 1721static void 1722tbf_control(vifp, m, ip, p_len) 1723 register struct vif *vifp; 1724 register struct mbuf *m; 1725 register struct ip *ip; 1726 register u_long p_len; 1727{ 1728 register struct tbf *t = vifp->v_tbf; 1729 1730 if (p_len > MAX_BKT_SIZE) { 1731 /* drop if packet is too large */ 1732 mrtstat.mrts_pkt2large++; 1733 m_freem(m); 1734 return; 1735 } 1736 1737 tbf_update_tokens(vifp); 1738 1739 /* if there are enough tokens, 1740 * and the queue is empty, 1741 * send this packet out 1742 */ 1743 1744 if (t->tbf_q_len == 0) { 1745 /* queue empty, send packet if enough tokens */ 1746 if (p_len <= t->tbf_n_tok) { 1747 t->tbf_n_tok -= p_len; 1748 tbf_send_packet(vifp, m); 1749 } else { 1750 /* queue packet and timeout till later */ 1751 tbf_queue(vifp, m); 1752 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1753 } 1754 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1755 /* finite queue length, so queue pkts and process queue */ 1756 tbf_queue(vifp, m); 1757 tbf_process_q(vifp); 1758 } else { 1759 /* queue length too much, try to dq and queue and process */ 1760 if (!tbf_dq_sel(vifp, ip)) { 1761 mrtstat.mrts_q_overflow++; 1762 m_freem(m); 1763 return; 1764 } else { 1765 tbf_queue(vifp, m); 1766 tbf_process_q(vifp); 1767 } 1768 } 1769 return; 1770} 1771 1772/* 1773 * adds a packet to the queue at the interface 1774 */ 1775static void 1776tbf_queue(vifp, m) 1777 register struct vif *vifp; 1778 register struct mbuf *m; 1779{ 1780 register int s = splnet(); 1781 register struct tbf *t = vifp->v_tbf; 1782 1783 if (t->tbf_t == NULL) { 1784 /* Queue was empty */ 1785 t->tbf_q = m; 1786 } else { 1787 /* Insert at tail */ 1788 t->tbf_t->m_act = m; 1789 } 1790 1791 /* Set new tail pointer */ 1792 t->tbf_t = m; 1793 1794#ifdef DIAGNOSTIC 1795 /* Make sure we didn't get fed a bogus mbuf */ 1796 if (m->m_act) 1797 panic("tbf_queue: m_act"); 1798#endif 1799 m->m_act = NULL; 1800 1801 t->tbf_q_len++; 1802 1803 splx(s); 1804} 1805 1806 1807/* 1808 * processes the queue at the interface 1809 */ 1810static void 1811tbf_process_q(vifp) 1812 register struct vif *vifp; 1813{ 1814 register struct mbuf *m; 1815 register int len; 1816 register int s = splnet(); 1817 register struct tbf *t = vifp->v_tbf; 1818 1819 /* loop through the queue at the interface and send as many packets 1820 * as possible 1821 */ 1822 while (t->tbf_q_len > 0) { 1823 m = t->tbf_q; 1824 1825 len = mtod(m, struct ip *)->ip_len; 1826 1827 /* determine if the packet can be sent */ 1828 if (len <= t->tbf_n_tok) { 1829 /* if so, 1830 * reduce no of tokens, dequeue the packet, 1831 * send the packet. 1832 */ 1833 t->tbf_n_tok -= len; 1834 1835 t->tbf_q = m->m_act; 1836 if (--t->tbf_q_len == 0) 1837 t->tbf_t = NULL; 1838 1839 m->m_act = NULL; 1840 tbf_send_packet(vifp, m); 1841 1842 } else break; 1843 } 1844 splx(s); 1845} 1846 1847static void 1848tbf_reprocess_q(xvifp) 1849 void *xvifp; 1850{ 1851 register struct vif *vifp = xvifp; 1852 if (ip_mrouter == NULL) 1853 return; 1854 1855 tbf_update_tokens(vifp); 1856 1857 tbf_process_q(vifp); 1858 1859 if (vifp->v_tbf->tbf_q_len) 1860 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1861} 1862 1863/* function that will selectively discard a member of the queue 1864 * based on the precedence value and the priority 1865 */ 1866static int 1867tbf_dq_sel(vifp, ip) 1868 register struct vif *vifp; 1869 register struct ip *ip; 1870{ 1871 register int s = splnet(); 1872 register u_int p; 1873 register struct mbuf *m, *last; 1874 register struct mbuf **np; 1875 register struct tbf *t = vifp->v_tbf; 1876 1877 p = priority(vifp, ip); 1878 1879 np = &t->tbf_q; 1880 last = NULL; 1881 while ((m = *np) != NULL) { 1882 if (p > priority(vifp, mtod(m, struct ip *))) { 1883 *np = m->m_act; 1884 /* If we're removing the last packet, fix the tail pointer */ 1885 if (m == t->tbf_t) 1886 t->tbf_t = last; 1887 m_freem(m); 1888 /* it's impossible for the queue to be empty, but 1889 * we check anyway. */ 1890 if (--t->tbf_q_len == 0) 1891 t->tbf_t = NULL; 1892 splx(s); 1893 mrtstat.mrts_drop_sel++; 1894 return(1); 1895 } 1896 np = &m->m_act; 1897 last = m; 1898 } 1899 splx(s); 1900 return(0); 1901} 1902 1903static void 1904tbf_send_packet(vifp, m) 1905 register struct vif *vifp; 1906 register struct mbuf *m; 1907{ 1908 struct ip_moptions imo; 1909 int error; 1910 static struct route ro; 1911 int s = splnet(); 1912 1913 if (vifp->v_flags & VIFF_TUNNEL) { 1914 /* If tunnel options */ 1915 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1916 IP_FORWARDING, (struct ip_moptions *)0); 1917 } else { 1918 imo.imo_multicast_ifp = vifp->v_ifp; 1919 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1920 imo.imo_multicast_loop = 1; 1921 imo.imo_multicast_vif = -1; 1922 1923 /* 1924 * Re-entrancy should not be a problem here, because 1925 * the packets that we send out and are looped back at us 1926 * should get rejected because they appear to come from 1927 * the loopback interface, thus preventing looping. 1928 */ 1929 error = ip_output(m, (struct mbuf *)0, &ro, 1930 IP_FORWARDING, &imo); 1931 1932 if (mrtdebug & DEBUG_XMIT) 1933 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1934 vifp - viftable, error); 1935 } 1936 splx(s); 1937} 1938 1939/* determine the current time and then 1940 * the elapsed time (between the last time and time now) 1941 * in milliseconds & update the no. of tokens in the bucket 1942 */ 1943static void 1944tbf_update_tokens(vifp) 1945 register struct vif *vifp; 1946{ 1947 struct timeval tp; 1948 register u_long tm; 1949 register int s = splnet(); 1950 register struct tbf *t = vifp->v_tbf; 1951 1952 GET_TIME(tp); 1953 1954 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1955 1956 /* 1957 * This formula is actually 1958 * "time in seconds" * "bytes/second". 1959 * 1960 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1961 * 1962 * The (1000/1024) was introduced in add_vif to optimize 1963 * this divide into a shift. 1964 */ 1965 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1966 t->tbf_last_pkt_t = tp; 1967 1968 if (t->tbf_n_tok > MAX_BKT_SIZE) 1969 t->tbf_n_tok = MAX_BKT_SIZE; 1970 1971 splx(s); 1972} 1973 1974static int 1975priority(vifp, ip) 1976 register struct vif *vifp; 1977 register struct ip *ip; 1978{ 1979 register int prio; 1980 1981 /* temporary hack; may add general packet classifier some day */ 1982 1983 /* 1984 * The UDP port space is divided up into four priority ranges: 1985 * [0, 16384) : unclassified - lowest priority 1986 * [16384, 32768) : audio - highest priority 1987 * [32768, 49152) : whiteboard - medium priority 1988 * [49152, 65536) : video - low priority 1989 */ 1990 if (ip->ip_p == IPPROTO_UDP) { 1991 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1992 switch (ntohs(udp->uh_dport) & 0xc000) { 1993 case 0x4000: 1994 prio = 70; 1995 break; 1996 case 0x8000: 1997 prio = 60; 1998 break; 1999 case 0xc000: 2000 prio = 55; 2001 break; 2002 default: 2003 prio = 50; 2004 break; 2005 } 2006 if (tbfdebug > 1) 2007 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio); 2008 } else { 2009 prio = 50; 2010 } 2011 return prio; 2012} 2013 2014/* 2015 * End of token bucket filter modifications 2016 */ 2017 2018int 2019ip_rsvp_vif_init(so, sopt) 2020 struct socket *so; 2021 struct sockopt *sopt; 2022{ 2023 int error, i, s; 2024 2025 if (rsvpdebug) 2026 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2027 so->so_type, so->so_proto->pr_protocol); 2028 2029 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2030 return EOPNOTSUPP; 2031 2032 /* Check mbuf. */ 2033 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 2034 if (error) 2035 return (error); 2036 2037 if (rsvpdebug) 2038 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", i, rsvp_on); 2039 2040 s = splnet(); 2041 2042 /* Check vif. */ 2043 if (!legal_vif_num(i)) { 2044 splx(s); 2045 return EADDRNOTAVAIL; 2046 } 2047 2048 /* Check if socket is available. */ 2049 if (viftable[i].v_rsvpd != NULL) { 2050 splx(s); 2051 return EADDRINUSE; 2052 } 2053 2054 viftable[i].v_rsvpd = so; 2055 /* This may seem silly, but we need to be sure we don't over-increment 2056 * the RSVP counter, in case something slips up. 2057 */ 2058 if (!viftable[i].v_rsvp_on) { 2059 viftable[i].v_rsvp_on = 1; 2060 rsvp_on++; 2061 } 2062 2063 splx(s); 2064 return 0; 2065} 2066 2067int 2068ip_rsvp_vif_done(so, sopt) 2069 struct socket *so; 2070 struct sockopt *sopt; 2071{ 2072 int error, i, s; 2073 2074 if (rsvpdebug) 2075 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2076 so->so_type, so->so_proto->pr_protocol); 2077 2078 if (so->so_type != SOCK_RAW || 2079 so->so_proto->pr_protocol != IPPROTO_RSVP) 2080 return EOPNOTSUPP; 2081 2082 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 2083 if (error) 2084 return (error); 2085 2086 s = splnet(); 2087 2088 /* Check vif. */ 2089 if (!legal_vif_num(i)) { 2090 splx(s); 2091 return EADDRNOTAVAIL; 2092 } 2093 2094 if (rsvpdebug) 2095 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n", 2096 viftable[i].v_rsvpd, so); 2097 2098 viftable[i].v_rsvpd = NULL; 2099 /* 2100 * This may seem silly, but we need to be sure we don't over-decrement 2101 * the RSVP counter, in case something slips up. 2102 */ 2103 if (viftable[i].v_rsvp_on) { 2104 viftable[i].v_rsvp_on = 0; 2105 rsvp_on--; 2106 } 2107 2108 splx(s); 2109 return 0; 2110} 2111 2112void 2113ip_rsvp_force_done(so) 2114 struct socket *so; 2115{ 2116 int vifi; 2117 register int s; 2118 2119 /* Don't bother if it is not the right type of socket. */ 2120 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2121 return; 2122 2123 s = splnet(); 2124 2125 /* The socket may be attached to more than one vif...this 2126 * is perfectly legal. 2127 */ 2128 for (vifi = 0; vifi < numvifs; vifi++) { 2129 if (viftable[vifi].v_rsvpd == so) { 2130 viftable[vifi].v_rsvpd = NULL; 2131 /* This may seem silly, but we need to be sure we don't 2132 * over-decrement the RSVP counter, in case something slips up. 2133 */ 2134 if (viftable[vifi].v_rsvp_on) { 2135 viftable[vifi].v_rsvp_on = 0; 2136 rsvp_on--; 2137 } 2138 } 2139 } 2140 2141 splx(s); 2142 return; 2143} 2144 2145void 2146rsvp_input(m, iphlen) 2147 struct mbuf *m; 2148 int iphlen; 2149{ 2150 int vifi; 2151 register struct ip *ip = mtod(m, struct ip *); 2152 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2153 register int s; 2154 struct ifnet *ifp; 2155 2156 if (rsvpdebug) 2157 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2158 2159 /* Can still get packets with rsvp_on = 0 if there is a local member 2160 * of the group to which the RSVP packet is addressed. But in this 2161 * case we want to throw the packet away. 2162 */ 2163 if (!rsvp_on) { 2164 m_freem(m); 2165 return; 2166 } 2167 2168 /* If the old-style non-vif-associated socket is set, then use 2169 * it and ignore the new ones. 2170 */ 2171 if (ip_rsvpd != NULL) { 2172 if (rsvpdebug) 2173 printf("rsvp_input: Sending packet up old-style socket\n"); 2174 rip_input(m, iphlen); 2175 return; 2176 } 2177 2178 s = splnet(); 2179 2180 if (rsvpdebug) 2181 printf("rsvp_input: check vifs\n"); 2182 2183#ifdef DIAGNOSTIC 2184 if (!(m->m_flags & M_PKTHDR)) 2185 panic("rsvp_input no hdr"); 2186#endif 2187 2188 ifp = m->m_pkthdr.rcvif; 2189 /* Find which vif the packet arrived on. */ 2190 for (vifi = 0; vifi < numvifs; vifi++) { 2191 if (viftable[vifi].v_ifp == ifp) 2192 break; 2193 } 2194 2195 if (vifi == numvifs) { 2196 /* Can't find vif packet arrived on. Drop packet. */ 2197 if (rsvpdebug) 2198 printf("rsvp_input: Can't find vif for packet...dropping it.\n"); 2199 m_freem(m); 2200 splx(s); 2201 return; 2202 } 2203 2204 if (rsvpdebug) 2205 printf("rsvp_input: check socket\n"); 2206 2207 if (viftable[vifi].v_rsvpd == NULL) { 2208 /* drop packet, since there is no specific socket for this 2209 * interface */ 2210 if (rsvpdebug) 2211 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2212 m_freem(m); 2213 splx(s); 2214 return; 2215 } 2216 rsvp_src.sin_addr = ip->ip_src; 2217 2218 if (rsvpdebug && m) 2219 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2220 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2221 2222 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2223 if (rsvpdebug) 2224 printf("rsvp_input: Failed to append to socket\n"); 2225 else 2226 if (rsvpdebug) 2227 printf("rsvp_input: send packet up\n"); 2228 2229 splx(s); 2230} 2231 2232#ifdef MROUTE_LKM 2233#include <sys/conf.h> 2234#include <sys/exec.h> 2235#include <sys/sysent.h> 2236#include <sys/lkm.h> 2237 2238MOD_MISC("ip_mroute_mod") 2239 2240static int 2241ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd) 2242{ 2243 int i; 2244 struct lkm_misc *args = lkmtp->private.lkm_misc; 2245 int err = 0; 2246 2247 switch(cmd) { 2248 static int (*old_ip_mrouter_cmd)(); 2249 static int (*old_ip_mrouter_done)(); 2250 static int (*old_ip_mforward)(); 2251 static int (*old_mrt_ioctl)(); 2252 static void (*old_proto4_input)(); 2253 static int (*old_legal_vif_num)(); 2254 extern struct protosw inetsw[]; 2255 2256 case LKM_E_LOAD: 2257 if(lkmexists(lkmtp) || ip_mrtproto) 2258 return(EEXIST); 2259 old_ip_mrouter_cmd = ip_mrouter_cmd; 2260 ip_mrouter_cmd = X_ip_mrouter_cmd; 2261 old_ip_mrouter_done = ip_mrouter_done; 2262 ip_mrouter_done = X_ip_mrouter_done; 2263 old_ip_mforward = ip_mforward; 2264 ip_mforward = X_ip_mforward; 2265 old_mrt_ioctl = mrt_ioctl; 2266 mrt_ioctl = X_mrt_ioctl; 2267 old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input; 2268 inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input; 2269 old_legal_vif_num = legal_vif_num; 2270 legal_vif_num = X_legal_vif_num; 2271 ip_mrtproto = IGMP_DVMRP; 2272 2273 printf("\nIP multicast routing loaded\n"); 2274 break; 2275 2276 case LKM_E_UNLOAD: 2277 if (ip_mrouter) 2278 return EINVAL; 2279 2280 ip_mrouter_cmd = old_ip_mrouter_cmd; 2281 ip_mrouter_done = old_ip_mrouter_done; 2282 ip_mforward = old_ip_mforward; 2283 mrt_ioctl = old_mrt_ioctl; 2284 inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input; 2285 legal_vif_num = old_legal_vif_num; 2286 ip_mrtproto = 0; 2287 break; 2288 2289 default: 2290 err = EINVAL; 2291 break; 2292 } 2293 2294 return(err); 2295} 2296 2297int 2298ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) { 2299 DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle, 2300 nosys); 2301} 2302 2303#endif /* MROUTE_LKM */ 2304#endif /* MROUTING */ 2305