ip_mroute.c revision 38482
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 1.48 1998/08/17 01:05:24 bde 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); 432 433 case MRT_ASSERT: 434 error = sooptcopyin(sopt, &optval, sizeof optval, 435 sizeof optval); 436 if (error) 437 break; 438 set_assert(optval); 439 440 default: 441 error = EOPNOTSUPP; 442 break; 443 } 444 return (error); 445} 446 447#ifndef MROUTE_LKM 448int (*ip_mrouter_set)(struct socket *, struct sockopt *) = X_ip_mrouter_set; 449#endif 450 451/* 452 * Handle MRT getsockopt commands 453 */ 454static int 455X_ip_mrouter_get(so, sopt) 456 struct socket *so; 457 struct sockopt *sopt; 458{ 459 int error; 460 static int version = 0x0305; /* !!! why is this here? XXX */ 461 462 switch (sopt->sopt_name) { 463 case MRT_VERSION: 464 error = sooptcopyout(sopt, &version, sizeof version); 465 break; 466 467 case MRT_ASSERT: 468 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 469 break; 470 default: 471 error = EOPNOTSUPP; 472 break; 473 } 474 return (error); 475} 476 477#ifndef MROUTE_LKM 478int (*ip_mrouter_get)(struct socket *, struct sockopt *) = X_ip_mrouter_get; 479#endif 480 481/* 482 * Handle ioctl commands to obtain information from the cache 483 */ 484static int 485X_mrt_ioctl(cmd, data) 486 int cmd; 487 caddr_t data; 488{ 489 int error = 0; 490 491 switch (cmd) { 492 case (SIOCGETVIFCNT): 493 return (get_vif_cnt((struct sioc_vif_req *)data)); 494 break; 495 case (SIOCGETSGCNT): 496 return (get_sg_cnt((struct sioc_sg_req *)data)); 497 break; 498 default: 499 return (EINVAL); 500 break; 501 } 502 return error; 503} 504 505#ifndef MROUTE_LKM 506int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl; 507#endif 508 509/* 510 * returns the packet, byte, rpf-failure count for the source group provided 511 */ 512static int 513get_sg_cnt(req) 514 register struct sioc_sg_req *req; 515{ 516 register struct mfc *rt; 517 int s; 518 519 s = splnet(); 520 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 521 splx(s); 522 if (rt != NULL) { 523 req->pktcnt = rt->mfc_pkt_cnt; 524 req->bytecnt = rt->mfc_byte_cnt; 525 req->wrong_if = rt->mfc_wrong_if; 526 } else 527 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 528 529 return 0; 530} 531 532/* 533 * returns the input and output packet and byte counts on the vif provided 534 */ 535static int 536get_vif_cnt(req) 537 register struct sioc_vif_req *req; 538{ 539 register vifi_t vifi = req->vifi; 540 541 if (vifi >= numvifs) return EINVAL; 542 543 req->icount = viftable[vifi].v_pkt_in; 544 req->ocount = viftable[vifi].v_pkt_out; 545 req->ibytes = viftable[vifi].v_bytes_in; 546 req->obytes = viftable[vifi].v_bytes_out; 547 548 return 0; 549} 550 551/* 552 * Enable multicast routing 553 */ 554static int 555ip_mrouter_init(so, version) 556 struct socket *so; 557 int version; 558{ 559 int *v; 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 struct ifreq ifr; 695 int error, s; 696 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 697 698 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL; 699 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE; 700 701 /* Find the interface with an address in AF_INET family */ 702 sin.sin_addr = vifcp->vifc_lcl_addr; 703 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 704 if (ifa == 0) return EADDRNOTAVAIL; 705 ifp = ifa->ifa_ifp; 706 707 if (vifcp->vifc_flags & VIFF_TUNNEL) { 708 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 709 /* 710 * An encapsulating tunnel is wanted. Tell ipip_input() to 711 * start paying attention to encapsulated packets. 712 */ 713 if (have_encap_tunnel == 0) { 714 have_encap_tunnel = 1; 715 for (s = 0; s < MAXVIFS; ++s) { 716 multicast_decap_if[s].if_name = "mdecap"; 717 multicast_decap_if[s].if_unit = s; 718 } 719 } 720 /* 721 * Set interface to fake encapsulator interface 722 */ 723 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 724 /* 725 * Prepare cached route entry 726 */ 727 bzero(&vifp->v_route, sizeof(vifp->v_route)); 728 } else { 729 log(LOG_ERR, "source routed tunnels not supported\n"); 730 return EOPNOTSUPP; 731 } 732 } else { 733 /* Make sure the interface supports multicast */ 734 if ((ifp->if_flags & IFF_MULTICAST) == 0) 735 return EOPNOTSUPP; 736 737 /* Enable promiscuous reception of all IP multicasts from the if */ 738 s = splnet(); 739 error = if_allmulti(ifp, 1); 740 splx(s); 741 if (error) 742 return error; 743 } 744 745 s = splnet(); 746 /* define parameters for the tbf structure */ 747 vifp->v_tbf = v_tbf; 748 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 749 vifp->v_tbf->tbf_n_tok = 0; 750 vifp->v_tbf->tbf_q_len = 0; 751 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 752 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 753 754 vifp->v_flags = vifcp->vifc_flags; 755 vifp->v_threshold = vifcp->vifc_threshold; 756 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 757 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 758 vifp->v_ifp = ifp; 759 /* scaling up here allows division by 1024 in critical code */ 760 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 761 vifp->v_rsvp_on = 0; 762 vifp->v_rsvpd = NULL; 763 /* initialize per vif pkt counters */ 764 vifp->v_pkt_in = 0; 765 vifp->v_pkt_out = 0; 766 vifp->v_bytes_in = 0; 767 vifp->v_bytes_out = 0; 768 splx(s); 769 770 /* Adjust numvifs up if the vifi is higher than numvifs */ 771 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 772 773 if (mrtdebug) 774 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n", 775 vifcp->vifc_vifi, 776 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 777 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 778 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 779 vifcp->vifc_threshold, 780 vifcp->vifc_rate_limit); 781 782 return 0; 783} 784 785/* 786 * Delete a vif from the vif table 787 */ 788static int 789del_vif(vifi) 790 vifi_t vifi; 791{ 792 register struct vif *vifp = &viftable[vifi]; 793 register struct mbuf *m; 794 struct ifnet *ifp; 795 struct ifreq ifr; 796 int s; 797 798 if (vifi >= numvifs) return EINVAL; 799 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL; 800 801 s = splnet(); 802 803 if (!(vifp->v_flags & VIFF_TUNNEL)) { 804 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 805 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 806 ifp = vifp->v_ifp; 807 if_allmulti(ifp, 0); 808 } 809 810 if (vifp == last_encap_vif) { 811 last_encap_vif = 0; 812 last_encap_src = 0; 813 } 814 815 /* 816 * Free packets queued at the interface 817 */ 818 while (vifp->v_tbf->tbf_q) { 819 m = vifp->v_tbf->tbf_q; 820 vifp->v_tbf->tbf_q = m->m_act; 821 m_freem(m); 822 } 823 824 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 825 bzero((caddr_t)vifp, sizeof (*vifp)); 826 827 if (mrtdebug) 828 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 829 830 /* Adjust numvifs down */ 831 for (vifi = numvifs; vifi > 0; vifi--) 832 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break; 833 numvifs = vifi; 834 835 splx(s); 836 837 return 0; 838} 839 840/* 841 * Add an mfc entry 842 */ 843static int 844add_mfc(mfccp) 845 struct mfcctl *mfccp; 846{ 847 struct mfc *rt; 848 register struct mbuf *mb_rt; 849 u_long hash; 850 struct mbuf *mb_ntry; 851 struct rtdetq *rte; 852 register u_short nstl; 853 int s; 854 int i; 855 856 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 857 858 /* If an entry already exists, just update the fields */ 859 if (rt) { 860 if (mrtdebug & DEBUG_MFC) 861 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 862 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 863 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 864 mfccp->mfcc_parent); 865 866 s = splnet(); 867 rt->mfc_parent = mfccp->mfcc_parent; 868 for (i = 0; i < numvifs; i++) 869 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 870 splx(s); 871 return 0; 872 } 873 874 /* 875 * Find the entry for which the upcall was made and update 876 */ 877 s = splnet(); 878 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 879 for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) { 880 881 rt = mtod(mb_rt, struct mfc *); 882 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 883 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 884 (mb_rt->m_act != NULL)) { 885 886 if (nstl++) 887 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 888 "multiple kernel entries", 889 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 890 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 891 mfccp->mfcc_parent, (void *)mb_rt->m_act); 892 893 if (mrtdebug & DEBUG_MFC) 894 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 895 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 896 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 897 mfccp->mfcc_parent, (void *)mb_rt->m_act); 898 899 rt->mfc_origin = mfccp->mfcc_origin; 900 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 901 rt->mfc_parent = mfccp->mfcc_parent; 902 for (i = 0; i < numvifs; i++) 903 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 904 /* initialize pkt counters per src-grp */ 905 rt->mfc_pkt_cnt = 0; 906 rt->mfc_byte_cnt = 0; 907 rt->mfc_wrong_if = 0; 908 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 909 910 rt->mfc_expire = 0; /* Don't clean this guy up */ 911 nexpire[hash]--; 912 913 /* free packets Qed at the end of this entry */ 914 while (mb_rt->m_act) { 915 mb_ntry = mb_rt->m_act; 916 rte = mtod(mb_ntry, struct rtdetq *); 917/* #ifdef RSVP_ISI */ 918 ip_mdq(rte->m, rte->ifp, rt, -1); 919/* #endif */ 920 mb_rt->m_act = mb_ntry->m_act; 921 m_freem(rte->m); 922#ifdef UPCALL_TIMING 923 collate(&(rte->t)); 924#endif /* UPCALL_TIMING */ 925 m_free(mb_ntry); 926 } 927 } 928 } 929 930 /* 931 * It is possible that an entry is being inserted without an upcall 932 */ 933 if (nstl == 0) { 934 if (mrtdebug & DEBUG_MFC) 935 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 936 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 937 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 938 mfccp->mfcc_parent); 939 940 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 941 942 rt = mtod(mb_rt, struct mfc *); 943 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 944 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 945 946 rt->mfc_origin = mfccp->mfcc_origin; 947 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 948 rt->mfc_parent = mfccp->mfcc_parent; 949 for (i = 0; i < numvifs; i++) 950 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 951 /* initialize pkt counters per src-grp */ 952 rt->mfc_pkt_cnt = 0; 953 rt->mfc_byte_cnt = 0; 954 rt->mfc_wrong_if = 0; 955 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 956 if (rt->mfc_expire) 957 nexpire[hash]--; 958 rt->mfc_expire = 0; 959 } 960 } 961 if (mb_rt == NULL) { 962 /* no upcall, so make a new entry */ 963 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 964 if (mb_rt == NULL) { 965 splx(s); 966 return ENOBUFS; 967 } 968 969 rt = mtod(mb_rt, struct mfc *); 970 971 /* insert new entry at head of hash chain */ 972 rt->mfc_origin = mfccp->mfcc_origin; 973 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 974 rt->mfc_parent = mfccp->mfcc_parent; 975 for (i = 0; i < numvifs; i++) 976 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 977 /* initialize pkt counters per src-grp */ 978 rt->mfc_pkt_cnt = 0; 979 rt->mfc_byte_cnt = 0; 980 rt->mfc_wrong_if = 0; 981 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 982 rt->mfc_expire = 0; 983 984 /* link into table */ 985 mb_rt->m_next = mfctable[hash]; 986 mfctable[hash] = mb_rt; 987 mb_rt->m_act = NULL; 988 } 989 } 990 splx(s); 991 return 0; 992} 993 994#ifdef UPCALL_TIMING 995/* 996 * collect delay statistics on the upcalls 997 */ 998static void collate(t) 999register struct timeval *t; 1000{ 1001 register u_long d; 1002 register struct timeval tp; 1003 register u_long delta; 1004 1005 GET_TIME(tp); 1006 1007 if (TV_LT(*t, tp)) 1008 { 1009 TV_DELTA(tp, *t, delta); 1010 1011 d = delta >> 10; 1012 if (d > 50) 1013 d = 50; 1014 1015 ++upcall_data[d]; 1016 } 1017} 1018#endif /* UPCALL_TIMING */ 1019 1020/* 1021 * Delete an mfc entry 1022 */ 1023static int 1024del_mfc(mfccp) 1025 struct mfcctl *mfccp; 1026{ 1027 struct in_addr origin; 1028 struct in_addr mcastgrp; 1029 struct mfc *rt; 1030 struct mbuf *mb_rt; 1031 struct mbuf **nptr; 1032 u_long hash; 1033 int s; 1034 1035 origin = mfccp->mfcc_origin; 1036 mcastgrp = mfccp->mfcc_mcastgrp; 1037 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1038 1039 if (mrtdebug & DEBUG_MFC) 1040 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1041 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1042 1043 s = splnet(); 1044 1045 nptr = &mfctable[hash]; 1046 while ((mb_rt = *nptr) != NULL) { 1047 rt = mtod(mb_rt, struct mfc *); 1048 if (origin.s_addr == rt->mfc_origin.s_addr && 1049 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1050 mb_rt->m_act == NULL) 1051 break; 1052 1053 nptr = &mb_rt->m_next; 1054 } 1055 if (mb_rt == NULL) { 1056 splx(s); 1057 return EADDRNOTAVAIL; 1058 } 1059 1060 MFREE(mb_rt, *nptr); 1061 1062 splx(s); 1063 1064 return 0; 1065} 1066 1067/* 1068 * Send a message to mrouted on the multicast routing socket 1069 */ 1070static int 1071socket_send(s, mm, src) 1072 struct socket *s; 1073 struct mbuf *mm; 1074 struct sockaddr_in *src; 1075{ 1076 if (s) { 1077 if (sbappendaddr(&s->so_rcv, 1078 (struct sockaddr *)src, 1079 mm, (struct mbuf *)0) != 0) { 1080 sorwakeup(s); 1081 return 0; 1082 } 1083 } 1084 m_freem(mm); 1085 return -1; 1086} 1087 1088/* 1089 * IP multicast forwarding function. This function assumes that the packet 1090 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1091 * pointed to by "ifp", and the packet is to be relayed to other networks 1092 * that have members of the packet's destination IP multicast group. 1093 * 1094 * The packet is returned unscathed to the caller, unless it is 1095 * erroneous, in which case a non-zero return value tells the caller to 1096 * discard it. 1097 */ 1098 1099#define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1100#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1101 1102static int 1103X_ip_mforward(ip, ifp, m, imo) 1104 register struct ip *ip; 1105 struct ifnet *ifp; 1106 struct mbuf *m; 1107 struct ip_moptions *imo; 1108{ 1109 register struct mfc *rt; 1110 register u_char *ipoptions; 1111 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1112 static int srctun = 0; 1113 register struct mbuf *mm; 1114 int s; 1115 vifi_t vifi; 1116 struct vif *vifp; 1117 1118 if (mrtdebug & DEBUG_FORWARD) 1119 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1120 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1121 (void *)ifp); 1122 1123 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1124 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1125 /* 1126 * Packet arrived via a physical interface or 1127 * an encapsulated tunnel. 1128 */ 1129 } else { 1130 /* 1131 * Packet arrived through a source-route tunnel. 1132 * Source-route tunnels are no longer supported. 1133 */ 1134 if ((srctun++ % 1000) == 0) 1135 log(LOG_ERR, 1136 "ip_mforward: received source-routed packet from %lx\n", 1137 (u_long)ntohl(ip->ip_src.s_addr)); 1138 1139 return 1; 1140 } 1141 1142 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1143 if (ip->ip_ttl < 255) 1144 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1145 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1146 vifp = viftable + vifi; 1147 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1148 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi, 1149 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1150 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1151 } 1152 return (ip_mdq(m, ifp, NULL, vifi)); 1153 } 1154 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1155 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1156 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr)); 1157 if(!imo) 1158 printf("In fact, no options were specified at all\n"); 1159 } 1160 1161 /* 1162 * Don't forward a packet with time-to-live of zero or one, 1163 * or a packet destined to a local-only group. 1164 */ 1165 if (ip->ip_ttl <= 1 || 1166 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1167 return 0; 1168 1169 /* 1170 * Determine forwarding vifs from the forwarding cache table 1171 */ 1172 s = splnet(); 1173 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt); 1174 1175 /* Entry exists, so forward if necessary */ 1176 if (rt != NULL) { 1177 splx(s); 1178 return (ip_mdq(m, ifp, rt, -1)); 1179 } else { 1180 /* 1181 * If we don't have a route for packet's origin, 1182 * Make a copy of the packet & 1183 * send message to routing daemon 1184 */ 1185 1186 register struct mbuf *mb_rt; 1187 register struct mbuf *mb_ntry; 1188 register struct mbuf *mb0; 1189 register struct rtdetq *rte; 1190 register struct mbuf *rte_m; 1191 register u_long hash; 1192 register int npkts; 1193 int hlen = ip->ip_hl << 2; 1194#ifdef UPCALL_TIMING 1195 struct timeval tp; 1196 1197 GET_TIME(tp); 1198#endif 1199 1200 mrtstat.mrts_no_route++; 1201 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1202 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1203 (u_long)ntohl(ip->ip_src.s_addr), 1204 (u_long)ntohl(ip->ip_dst.s_addr)); 1205 1206 /* 1207 * Allocate mbufs early so that we don't do extra work if we are 1208 * just going to fail anyway. Make sure to pullup the header so 1209 * that other people can't step on it. 1210 */ 1211 MGET(mb_ntry, M_DONTWAIT, MT_DATA); 1212 if (mb_ntry == NULL) { 1213 splx(s); 1214 return ENOBUFS; 1215 } 1216 mb0 = m_copy(m, 0, M_COPYALL); 1217 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1218 mb0 = m_pullup(mb0, hlen); 1219 if (mb0 == NULL) { 1220 m_free(mb_ntry); 1221 splx(s); 1222 return ENOBUFS; 1223 } 1224 1225 /* is there an upcall waiting for this packet? */ 1226 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1227 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 1228 rt = mtod(mb_rt, struct mfc *); 1229 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1230 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1231 (mb_rt->m_act != NULL)) 1232 break; 1233 } 1234 1235 if (mb_rt == NULL) { 1236 int i; 1237 struct igmpmsg *im; 1238 1239 /* no upcall, so make a new entry */ 1240 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 1241 if (mb_rt == NULL) { 1242 m_free(mb_ntry); 1243 m_freem(mb0); 1244 splx(s); 1245 return ENOBUFS; 1246 } 1247 /* Make a copy of the header to send to the user level process */ 1248 mm = m_copy(mb0, 0, hlen); 1249 if (mm == NULL) { 1250 m_free(mb_ntry); 1251 m_freem(mb0); 1252 m_free(mb_rt); 1253 splx(s); 1254 return ENOBUFS; 1255 } 1256 1257 /* 1258 * Send message to routing daemon to install 1259 * a route into the kernel table 1260 */ 1261 k_igmpsrc.sin_addr = ip->ip_src; 1262 1263 im = mtod(mm, struct igmpmsg *); 1264 im->im_msgtype = IGMPMSG_NOCACHE; 1265 im->im_mbz = 0; 1266 1267 mrtstat.mrts_upcalls++; 1268 1269 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1270 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1271 ++mrtstat.mrts_upq_sockfull; 1272 m_free(mb_ntry); 1273 m_freem(mb0); 1274 m_free(mb_rt); 1275 splx(s); 1276 return ENOBUFS; 1277 } 1278 1279 rt = mtod(mb_rt, struct mfc *); 1280 1281 /* insert new entry at head of hash chain */ 1282 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1283 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1284 rt->mfc_expire = UPCALL_EXPIRE; 1285 nexpire[hash]++; 1286 for (i = 0; i < numvifs; i++) 1287 rt->mfc_ttls[i] = 0; 1288 rt->mfc_parent = -1; 1289 1290 /* link into table */ 1291 mb_rt->m_next = mfctable[hash]; 1292 mfctable[hash] = mb_rt; 1293 mb_rt->m_act = NULL; 1294 1295 rte_m = mb_rt; 1296 } else { 1297 /* determine if q has overflowed */ 1298 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act) 1299 npkts++; 1300 1301 if (npkts > MAX_UPQ) { 1302 mrtstat.mrts_upq_ovflw++; 1303 m_free(mb_ntry); 1304 m_freem(mb0); 1305 splx(s); 1306 return 0; 1307 } 1308 } 1309 1310 mb_ntry->m_act = NULL; 1311 rte = mtod(mb_ntry, struct rtdetq *); 1312 1313 rte->m = mb0; 1314 rte->ifp = ifp; 1315#ifdef UPCALL_TIMING 1316 rte->t = tp; 1317#endif 1318 1319 /* Add this entry to the end of the queue */ 1320 rte_m->m_act = mb_ntry; 1321 1322 splx(s); 1323 1324 return 0; 1325 } 1326} 1327 1328#ifndef MROUTE_LKM 1329int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 1330 struct ip_moptions *) = X_ip_mforward; 1331#endif 1332 1333/* 1334 * Clean up the cache entry if upcall is not serviced 1335 */ 1336static void 1337expire_upcalls(void *unused) 1338{ 1339 struct mbuf *mb_rt, *m, **nptr; 1340 struct rtdetq *rte; 1341 struct mfc *mfc; 1342 int i; 1343 int s; 1344 1345 s = splnet(); 1346 for (i = 0; i < MFCTBLSIZ; i++) { 1347 if (nexpire[i] == 0) 1348 continue; 1349 nptr = &mfctable[i]; 1350 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) { 1351 mfc = mtod(mb_rt, struct mfc *); 1352 1353 /* 1354 * Skip real cache entries 1355 * Make sure it wasn't marked to not expire (shouldn't happen) 1356 * If it expires now 1357 */ 1358 if (mb_rt->m_act != NULL && 1359 mfc->mfc_expire != 0 && 1360 --mfc->mfc_expire == 0) { 1361 if (mrtdebug & DEBUG_EXPIRE) 1362 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1363 (u_long)ntohl(mfc->mfc_origin.s_addr), 1364 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1365 /* 1366 * drop all the packets 1367 * free the mbuf with the pkt, if, timing info 1368 */ 1369 while (mb_rt->m_act) { 1370 m = mb_rt->m_act; 1371 mb_rt->m_act = m->m_act; 1372 1373 rte = mtod(m, struct rtdetq *); 1374 m_freem(rte->m); 1375 m_free(m); 1376 } 1377 ++mrtstat.mrts_cache_cleanups; 1378 nexpire[i]--; 1379 1380 MFREE(mb_rt, *nptr); 1381 } else { 1382 nptr = &mb_rt->m_next; 1383 } 1384 } 1385 } 1386 splx(s); 1387 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 1388} 1389 1390/* 1391 * Packet forwarding routine once entry in the cache is made 1392 */ 1393static int 1394ip_mdq(m, ifp, rt, xmt_vif) 1395 register struct mbuf *m; 1396 register struct ifnet *ifp; 1397 register struct mfc *rt; 1398 register vifi_t xmt_vif; 1399{ 1400 register struct ip *ip = mtod(m, struct ip *); 1401 register vifi_t vifi; 1402 register struct vif *vifp; 1403 register int plen = ip->ip_len; 1404 1405/* 1406 * Macro to send packet on vif. Since RSVP packets don't get counted on 1407 * input, they shouldn't get counted on output, so statistics keeping is 1408 * seperate. 1409 */ 1410#define MC_SEND(ip,vifp,m) { \ 1411 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1412 encap_send((ip), (vifp), (m)); \ 1413 else \ 1414 phyint_send((ip), (vifp), (m)); \ 1415} 1416 1417 /* 1418 * If xmt_vif is not -1, send on only the requested vif. 1419 * 1420 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1421 */ 1422 if (xmt_vif < numvifs) { 1423 MC_SEND(ip, viftable + xmt_vif, m); 1424 return 1; 1425 } 1426 1427 /* 1428 * Don't forward if it didn't arrive from the parent vif for its origin. 1429 */ 1430 vifi = rt->mfc_parent; 1431 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1432 /* came in the wrong interface */ 1433 if (mrtdebug & DEBUG_FORWARD) 1434 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1435 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1436 ++mrtstat.mrts_wrong_if; 1437 ++rt->mfc_wrong_if; 1438 /* 1439 * If we are doing PIM assert processing, and we are forwarding 1440 * packets on this interface, and it is a broadcast medium 1441 * interface (and not a tunnel), send a message to the routing daemon. 1442 */ 1443 if (pim_assert && rt->mfc_ttls[vifi] && 1444 (ifp->if_flags & IFF_BROADCAST) && 1445 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1446 struct sockaddr_in k_igmpsrc; 1447 struct mbuf *mm; 1448 struct igmpmsg *im; 1449 int hlen = ip->ip_hl << 2; 1450 struct timeval now; 1451 register u_long delta; 1452 1453 GET_TIME(now); 1454 1455 TV_DELTA(rt->mfc_last_assert, now, delta); 1456 1457 if (delta > ASSERT_MSG_TIME) { 1458 mm = m_copy(m, 0, hlen); 1459 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1460 mm = m_pullup(mm, hlen); 1461 if (mm == NULL) { 1462 return ENOBUFS; 1463 } 1464 1465 rt->mfc_last_assert = now; 1466 1467 im = mtod(mm, struct igmpmsg *); 1468 im->im_msgtype = IGMPMSG_WRONGVIF; 1469 im->im_mbz = 0; 1470 im->im_vif = vifi; 1471 1472 k_igmpsrc.sin_addr = im->im_src; 1473 1474 socket_send(ip_mrouter, mm, &k_igmpsrc); 1475 } 1476 } 1477 return 0; 1478 } 1479 1480 /* If I sourced this packet, it counts as output, else it was input. */ 1481 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1482 viftable[vifi].v_pkt_out++; 1483 viftable[vifi].v_bytes_out += plen; 1484 } else { 1485 viftable[vifi].v_pkt_in++; 1486 viftable[vifi].v_bytes_in += plen; 1487 } 1488 rt->mfc_pkt_cnt++; 1489 rt->mfc_byte_cnt += plen; 1490 1491 /* 1492 * For each vif, decide if a copy of the packet should be forwarded. 1493 * Forward if: 1494 * - the ttl exceeds the vif's threshold 1495 * - there are group members downstream on interface 1496 */ 1497 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1498 if ((rt->mfc_ttls[vifi] > 0) && 1499 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1500 vifp->v_pkt_out++; 1501 vifp->v_bytes_out += plen; 1502 MC_SEND(ip, vifp, m); 1503 } 1504 1505 return 0; 1506} 1507 1508/* 1509 * check if a vif number is legal/ok. This is used by ip_output, to export 1510 * numvifs there, 1511 */ 1512static int 1513X_legal_vif_num(vif) 1514 int vif; 1515{ 1516 if (vif >= 0 && vif < numvifs) 1517 return(1); 1518 else 1519 return(0); 1520} 1521 1522#ifndef MROUTE_LKM 1523int (*legal_vif_num)(int) = X_legal_vif_num; 1524#endif 1525 1526/* 1527 * Return the local address used by this vif 1528 */ 1529static u_long 1530X_ip_mcast_src(vifi) 1531 int vifi; 1532{ 1533 if (vifi >= 0 && vifi < numvifs) 1534 return viftable[vifi].v_lcl_addr.s_addr; 1535 else 1536 return INADDR_ANY; 1537} 1538 1539#ifndef MROUTE_LKM 1540u_long (*ip_mcast_src)(int) = X_ip_mcast_src; 1541#endif 1542 1543static void 1544phyint_send(ip, vifp, m) 1545 struct ip *ip; 1546 struct vif *vifp; 1547 struct mbuf *m; 1548{ 1549 register struct mbuf *mb_copy; 1550 register int hlen = ip->ip_hl << 2; 1551 1552 /* 1553 * Make a new reference to the packet; make sure that 1554 * the IP header is actually copied, not just referenced, 1555 * so that ip_output() only scribbles on the copy. 1556 */ 1557 mb_copy = m_copy(m, 0, M_COPYALL); 1558 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1559 mb_copy = m_pullup(mb_copy, hlen); 1560 if (mb_copy == NULL) 1561 return; 1562 1563 if (vifp->v_rate_limit <= 0) 1564 tbf_send_packet(vifp, mb_copy); 1565 else 1566 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1567} 1568 1569static void 1570encap_send(ip, vifp, m) 1571 register struct ip *ip; 1572 register struct vif *vifp; 1573 register struct mbuf *m; 1574{ 1575 register struct mbuf *mb_copy; 1576 register struct ip *ip_copy; 1577 register int i, len = ip->ip_len; 1578 1579 /* 1580 * copy the old packet & pullup its IP header into the 1581 * new mbuf so we can modify it. Try to fill the new 1582 * mbuf since if we don't the ethernet driver will. 1583 */ 1584 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER); 1585 if (mb_copy == NULL) 1586 return; 1587 mb_copy->m_data += max_linkhdr; 1588 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1589 1590 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1591 m_freem(mb_copy); 1592 return; 1593 } 1594 i = MHLEN - M_LEADINGSPACE(mb_copy); 1595 if (i > len) 1596 i = len; 1597 mb_copy = m_pullup(mb_copy, i); 1598 if (mb_copy == NULL) 1599 return; 1600 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1601 1602 /* 1603 * fill in the encapsulating IP header. 1604 */ 1605 ip_copy = mtod(mb_copy, struct ip *); 1606 *ip_copy = multicast_encap_iphdr; 1607 ip_copy->ip_id = htons(ip_id++); 1608 ip_copy->ip_len += len; 1609 ip_copy->ip_src = vifp->v_lcl_addr; 1610 ip_copy->ip_dst = vifp->v_rmt_addr; 1611 1612 /* 1613 * turn the encapsulated IP header back into a valid one. 1614 */ 1615 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1616 --ip->ip_ttl; 1617 HTONS(ip->ip_len); 1618 HTONS(ip->ip_off); 1619 ip->ip_sum = 0; 1620 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1621 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1622 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1623 1624 if (vifp->v_rate_limit <= 0) 1625 tbf_send_packet(vifp, mb_copy); 1626 else 1627 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1628} 1629 1630/* 1631 * De-encapsulate a packet and feed it back through ip input (this 1632 * routine is called whenever IP gets a packet with proto type 1633 * ENCAP_PROTO and a local destination address). 1634 */ 1635void 1636#ifdef MROUTE_LKM 1637X_ipip_input(m, iphlen) 1638#else 1639ipip_input(m, iphlen) 1640#endif 1641 register struct mbuf *m; 1642 int iphlen; 1643{ 1644 struct ifnet *ifp = m->m_pkthdr.rcvif; 1645 register struct ip *ip = mtod(m, struct ip *); 1646 register int hlen = ip->ip_hl << 2; 1647 register int s; 1648 register struct ifqueue *ifq; 1649 register struct vif *vifp; 1650 1651 if (!have_encap_tunnel) { 1652 rip_input(m, iphlen); 1653 return; 1654 } 1655 /* 1656 * dump the packet if it's not to a multicast destination or if 1657 * we don't have an encapsulating tunnel with the source. 1658 * Note: This code assumes that the remote site IP address 1659 * uniquely identifies the tunnel (i.e., that this site has 1660 * at most one tunnel with the remote site). 1661 */ 1662 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) { 1663 ++mrtstat.mrts_bad_tunnel; 1664 m_freem(m); 1665 return; 1666 } 1667 if (ip->ip_src.s_addr != last_encap_src) { 1668 register struct vif *vife; 1669 1670 vifp = viftable; 1671 vife = vifp + numvifs; 1672 last_encap_src = ip->ip_src.s_addr; 1673 last_encap_vif = 0; 1674 for ( ; vifp < vife; ++vifp) 1675 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1676 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1677 == VIFF_TUNNEL) 1678 last_encap_vif = vifp; 1679 break; 1680 } 1681 } 1682 if ((vifp = last_encap_vif) == 0) { 1683 last_encap_src = 0; 1684 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1685 m_freem(m); 1686 if (mrtdebug) 1687 log(LOG_DEBUG, "ip_mforward: no tunnel with %lx\n", 1688 (u_long)ntohl(ip->ip_src.s_addr)); 1689 return; 1690 } 1691 ifp = vifp->v_ifp; 1692 1693 if (hlen > IP_HDR_LEN) 1694 ip_stripoptions(m, (struct mbuf *) 0); 1695 m->m_data += IP_HDR_LEN; 1696 m->m_len -= IP_HDR_LEN; 1697 m->m_pkthdr.len -= IP_HDR_LEN; 1698 m->m_pkthdr.rcvif = ifp; 1699 1700 ifq = &ipintrq; 1701 s = splimp(); 1702 if (IF_QFULL(ifq)) { 1703 IF_DROP(ifq); 1704 m_freem(m); 1705 } else { 1706 IF_ENQUEUE(ifq, m); 1707 /* 1708 * normally we would need a "schednetisr(NETISR_IP)" 1709 * here but we were called by ip_input and it is going 1710 * to loop back & try to dequeue the packet we just 1711 * queued as soon as we return so we avoid the 1712 * unnecessary software interrrupt. 1713 */ 1714 } 1715 splx(s); 1716} 1717 1718/* 1719 * Token bucket filter module 1720 */ 1721 1722static void 1723tbf_control(vifp, m, ip, p_len) 1724 register struct vif *vifp; 1725 register struct mbuf *m; 1726 register struct ip *ip; 1727 register u_long p_len; 1728{ 1729 register struct tbf *t = vifp->v_tbf; 1730 1731 if (p_len > MAX_BKT_SIZE) { 1732 /* drop if packet is too large */ 1733 mrtstat.mrts_pkt2large++; 1734 m_freem(m); 1735 return; 1736 } 1737 1738 tbf_update_tokens(vifp); 1739 1740 /* if there are enough tokens, 1741 * and the queue is empty, 1742 * send this packet out 1743 */ 1744 1745 if (t->tbf_q_len == 0) { 1746 /* queue empty, send packet if enough tokens */ 1747 if (p_len <= t->tbf_n_tok) { 1748 t->tbf_n_tok -= p_len; 1749 tbf_send_packet(vifp, m); 1750 } else { 1751 /* queue packet and timeout till later */ 1752 tbf_queue(vifp, m); 1753 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1754 } 1755 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1756 /* finite queue length, so queue pkts and process queue */ 1757 tbf_queue(vifp, m); 1758 tbf_process_q(vifp); 1759 } else { 1760 /* queue length too much, try to dq and queue and process */ 1761 if (!tbf_dq_sel(vifp, ip)) { 1762 mrtstat.mrts_q_overflow++; 1763 m_freem(m); 1764 return; 1765 } else { 1766 tbf_queue(vifp, m); 1767 tbf_process_q(vifp); 1768 } 1769 } 1770 return; 1771} 1772 1773/* 1774 * adds a packet to the queue at the interface 1775 */ 1776static void 1777tbf_queue(vifp, m) 1778 register struct vif *vifp; 1779 register struct mbuf *m; 1780{ 1781 register int s = splnet(); 1782 register struct tbf *t = vifp->v_tbf; 1783 1784 if (t->tbf_t == NULL) { 1785 /* Queue was empty */ 1786 t->tbf_q = m; 1787 } else { 1788 /* Insert at tail */ 1789 t->tbf_t->m_act = m; 1790 } 1791 1792 /* Set new tail pointer */ 1793 t->tbf_t = m; 1794 1795#ifdef DIAGNOSTIC 1796 /* Make sure we didn't get fed a bogus mbuf */ 1797 if (m->m_act) 1798 panic("tbf_queue: m_act"); 1799#endif 1800 m->m_act = NULL; 1801 1802 t->tbf_q_len++; 1803 1804 splx(s); 1805} 1806 1807 1808/* 1809 * processes the queue at the interface 1810 */ 1811static void 1812tbf_process_q(vifp) 1813 register struct vif *vifp; 1814{ 1815 register struct mbuf *m; 1816 register int len; 1817 register int s = splnet(); 1818 register struct tbf *t = vifp->v_tbf; 1819 1820 /* loop through the queue at the interface and send as many packets 1821 * as possible 1822 */ 1823 while (t->tbf_q_len > 0) { 1824 m = t->tbf_q; 1825 1826 len = mtod(m, struct ip *)->ip_len; 1827 1828 /* determine if the packet can be sent */ 1829 if (len <= t->tbf_n_tok) { 1830 /* if so, 1831 * reduce no of tokens, dequeue the packet, 1832 * send the packet. 1833 */ 1834 t->tbf_n_tok -= len; 1835 1836 t->tbf_q = m->m_act; 1837 if (--t->tbf_q_len == 0) 1838 t->tbf_t = NULL; 1839 1840 m->m_act = NULL; 1841 tbf_send_packet(vifp, m); 1842 1843 } else break; 1844 } 1845 splx(s); 1846} 1847 1848static void 1849tbf_reprocess_q(xvifp) 1850 void *xvifp; 1851{ 1852 register struct vif *vifp = xvifp; 1853 if (ip_mrouter == NULL) 1854 return; 1855 1856 tbf_update_tokens(vifp); 1857 1858 tbf_process_q(vifp); 1859 1860 if (vifp->v_tbf->tbf_q_len) 1861 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1862} 1863 1864/* function that will selectively discard a member of the queue 1865 * based on the precedence value and the priority 1866 */ 1867static int 1868tbf_dq_sel(vifp, ip) 1869 register struct vif *vifp; 1870 register struct ip *ip; 1871{ 1872 register int s = splnet(); 1873 register u_int p; 1874 register struct mbuf *m, *last; 1875 register struct mbuf **np; 1876 register struct tbf *t = vifp->v_tbf; 1877 1878 p = priority(vifp, ip); 1879 1880 np = &t->tbf_q; 1881 last = NULL; 1882 while ((m = *np) != NULL) { 1883 if (p > priority(vifp, mtod(m, struct ip *))) { 1884 *np = m->m_act; 1885 /* If we're removing the last packet, fix the tail pointer */ 1886 if (m == t->tbf_t) 1887 t->tbf_t = last; 1888 m_freem(m); 1889 /* it's impossible for the queue to be empty, but 1890 * we check anyway. */ 1891 if (--t->tbf_q_len == 0) 1892 t->tbf_t = NULL; 1893 splx(s); 1894 mrtstat.mrts_drop_sel++; 1895 return(1); 1896 } 1897 np = &m->m_act; 1898 last = m; 1899 } 1900 splx(s); 1901 return(0); 1902} 1903 1904static void 1905tbf_send_packet(vifp, m) 1906 register struct vif *vifp; 1907 register struct mbuf *m; 1908{ 1909 struct ip_moptions imo; 1910 int error; 1911 static struct route ro; 1912 int s = splnet(); 1913 1914 if (vifp->v_flags & VIFF_TUNNEL) { 1915 /* If tunnel options */ 1916 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1917 IP_FORWARDING, (struct ip_moptions *)0); 1918 } else { 1919 imo.imo_multicast_ifp = vifp->v_ifp; 1920 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1921 imo.imo_multicast_loop = 1; 1922 imo.imo_multicast_vif = -1; 1923 1924 /* 1925 * Re-entrancy should not be a problem here, because 1926 * the packets that we send out and are looped back at us 1927 * should get rejected because they appear to come from 1928 * the loopback interface, thus preventing looping. 1929 */ 1930 error = ip_output(m, (struct mbuf *)0, &ro, 1931 IP_FORWARDING, &imo); 1932 1933 if (mrtdebug & DEBUG_XMIT) 1934 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1935 vifp - viftable, error); 1936 } 1937 splx(s); 1938} 1939 1940/* determine the current time and then 1941 * the elapsed time (between the last time and time now) 1942 * in milliseconds & update the no. of tokens in the bucket 1943 */ 1944static void 1945tbf_update_tokens(vifp) 1946 register struct vif *vifp; 1947{ 1948 struct timeval tp; 1949 register u_long tm; 1950 register int s = splnet(); 1951 register struct tbf *t = vifp->v_tbf; 1952 1953 GET_TIME(tp); 1954 1955 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1956 1957 /* 1958 * This formula is actually 1959 * "time in seconds" * "bytes/second". 1960 * 1961 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1962 * 1963 * The (1000/1024) was introduced in add_vif to optimize 1964 * this divide into a shift. 1965 */ 1966 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1967 t->tbf_last_pkt_t = tp; 1968 1969 if (t->tbf_n_tok > MAX_BKT_SIZE) 1970 t->tbf_n_tok = MAX_BKT_SIZE; 1971 1972 splx(s); 1973} 1974 1975static int 1976priority(vifp, ip) 1977 register struct vif *vifp; 1978 register struct ip *ip; 1979{ 1980 register int prio; 1981 1982 /* temporary hack; may add general packet classifier some day */ 1983 1984 /* 1985 * The UDP port space is divided up into four priority ranges: 1986 * [0, 16384) : unclassified - lowest priority 1987 * [16384, 32768) : audio - highest priority 1988 * [32768, 49152) : whiteboard - medium priority 1989 * [49152, 65536) : video - low priority 1990 */ 1991 if (ip->ip_p == IPPROTO_UDP) { 1992 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1993 switch (ntohs(udp->uh_dport) & 0xc000) { 1994 case 0x4000: 1995 prio = 70; 1996 break; 1997 case 0x8000: 1998 prio = 60; 1999 break; 2000 case 0xc000: 2001 prio = 55; 2002 break; 2003 default: 2004 prio = 50; 2005 break; 2006 } 2007 if (tbfdebug > 1) 2008 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio); 2009 } else { 2010 prio = 50; 2011 } 2012 return prio; 2013} 2014 2015/* 2016 * End of token bucket filter modifications 2017 */ 2018 2019int 2020ip_rsvp_vif_init(so, sopt) 2021 struct socket *so; 2022 struct sockopt *sopt; 2023{ 2024 int error, i, s; 2025 2026 if (rsvpdebug) 2027 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2028 so->so_type, so->so_proto->pr_protocol); 2029 2030 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2031 return EOPNOTSUPP; 2032 2033 /* Check mbuf. */ 2034 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 2035 if (error) 2036 return (error); 2037 2038 if (rsvpdebug) 2039 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", i, rsvp_on); 2040 2041 s = splnet(); 2042 2043 /* Check vif. */ 2044 if (!legal_vif_num(i)) { 2045 splx(s); 2046 return EADDRNOTAVAIL; 2047 } 2048 2049 /* Check if socket is available. */ 2050 if (viftable[i].v_rsvpd != NULL) { 2051 splx(s); 2052 return EADDRINUSE; 2053 } 2054 2055 viftable[i].v_rsvpd = so; 2056 /* This may seem silly, but we need to be sure we don't over-increment 2057 * the RSVP counter, in case something slips up. 2058 */ 2059 if (!viftable[i].v_rsvp_on) { 2060 viftable[i].v_rsvp_on = 1; 2061 rsvp_on++; 2062 } 2063 2064 splx(s); 2065 return 0; 2066} 2067 2068int 2069ip_rsvp_vif_done(so, sopt) 2070 struct socket *so; 2071 struct sockopt *sopt; 2072{ 2073 int error, i, s; 2074 2075 if (rsvpdebug) 2076 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2077 so->so_type, so->so_proto->pr_protocol); 2078 2079 if (so->so_type != SOCK_RAW || 2080 so->so_proto->pr_protocol != IPPROTO_RSVP) 2081 return EOPNOTSUPP; 2082 2083 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 2084 if (error) 2085 return (error); 2086 2087 s = splnet(); 2088 2089 /* Check vif. */ 2090 if (!legal_vif_num(i)) { 2091 splx(s); 2092 return EADDRNOTAVAIL; 2093 } 2094 2095 if (rsvpdebug) 2096 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n", 2097 viftable[i].v_rsvpd, so); 2098 2099 viftable[i].v_rsvpd = NULL; 2100 /* 2101 * This may seem silly, but we need to be sure we don't over-decrement 2102 * the RSVP counter, in case something slips up. 2103 */ 2104 if (viftable[i].v_rsvp_on) { 2105 viftable[i].v_rsvp_on = 0; 2106 rsvp_on--; 2107 } 2108 2109 splx(s); 2110 return 0; 2111} 2112 2113void 2114ip_rsvp_force_done(so) 2115 struct socket *so; 2116{ 2117 int vifi; 2118 register int s; 2119 2120 /* Don't bother if it is not the right type of socket. */ 2121 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2122 return; 2123 2124 s = splnet(); 2125 2126 /* The socket may be attached to more than one vif...this 2127 * is perfectly legal. 2128 */ 2129 for (vifi = 0; vifi < numvifs; vifi++) { 2130 if (viftable[vifi].v_rsvpd == so) { 2131 viftable[vifi].v_rsvpd = NULL; 2132 /* This may seem silly, but we need to be sure we don't 2133 * over-decrement the RSVP counter, in case something slips up. 2134 */ 2135 if (viftable[vifi].v_rsvp_on) { 2136 viftable[vifi].v_rsvp_on = 0; 2137 rsvp_on--; 2138 } 2139 } 2140 } 2141 2142 splx(s); 2143 return; 2144} 2145 2146void 2147rsvp_input(m, iphlen) 2148 struct mbuf *m; 2149 int iphlen; 2150{ 2151 int vifi; 2152 register struct ip *ip = mtod(m, struct ip *); 2153 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2154 register int s; 2155 struct ifnet *ifp; 2156 2157 if (rsvpdebug) 2158 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2159 2160 /* Can still get packets with rsvp_on = 0 if there is a local member 2161 * of the group to which the RSVP packet is addressed. But in this 2162 * case we want to throw the packet away. 2163 */ 2164 if (!rsvp_on) { 2165 m_freem(m); 2166 return; 2167 } 2168 2169 /* If the old-style non-vif-associated socket is set, then use 2170 * it and ignore the new ones. 2171 */ 2172 if (ip_rsvpd != NULL) { 2173 if (rsvpdebug) 2174 printf("rsvp_input: Sending packet up old-style socket\n"); 2175 rip_input(m, iphlen); 2176 return; 2177 } 2178 2179 s = splnet(); 2180 2181 if (rsvpdebug) 2182 printf("rsvp_input: check vifs\n"); 2183 2184#ifdef DIAGNOSTIC 2185 if (!(m->m_flags & M_PKTHDR)) 2186 panic("rsvp_input no hdr"); 2187#endif 2188 2189 ifp = m->m_pkthdr.rcvif; 2190 /* Find which vif the packet arrived on. */ 2191 for (vifi = 0; vifi < numvifs; vifi++) { 2192 if (viftable[vifi].v_ifp == ifp) 2193 break; 2194 } 2195 2196 if (vifi == numvifs) { 2197 /* Can't find vif packet arrived on. Drop packet. */ 2198 if (rsvpdebug) 2199 printf("rsvp_input: Can't find vif for packet...dropping it.\n"); 2200 m_freem(m); 2201 splx(s); 2202 return; 2203 } 2204 2205 if (rsvpdebug) 2206 printf("rsvp_input: check socket\n"); 2207 2208 if (viftable[vifi].v_rsvpd == NULL) { 2209 /* drop packet, since there is no specific socket for this 2210 * interface */ 2211 if (rsvpdebug) 2212 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2213 m_freem(m); 2214 splx(s); 2215 return; 2216 } 2217 rsvp_src.sin_addr = ip->ip_src; 2218 2219 if (rsvpdebug && m) 2220 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2221 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2222 2223 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2224 if (rsvpdebug) 2225 printf("rsvp_input: Failed to append to socket\n"); 2226 else 2227 if (rsvpdebug) 2228 printf("rsvp_input: send packet up\n"); 2229 2230 splx(s); 2231} 2232 2233#ifdef MROUTE_LKM 2234#include <sys/conf.h> 2235#include <sys/exec.h> 2236#include <sys/sysent.h> 2237#include <sys/lkm.h> 2238 2239MOD_MISC("ip_mroute_mod") 2240 2241static int 2242ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd) 2243{ 2244 int i; 2245 struct lkm_misc *args = lkmtp->private.lkm_misc; 2246 int err = 0; 2247 2248 switch(cmd) { 2249 static int (*old_ip_mrouter_cmd)(); 2250 static int (*old_ip_mrouter_done)(); 2251 static int (*old_ip_mforward)(); 2252 static int (*old_mrt_ioctl)(); 2253 static void (*old_proto4_input)(); 2254 static int (*old_legal_vif_num)(); 2255 extern struct protosw inetsw[]; 2256 2257 case LKM_E_LOAD: 2258 if(lkmexists(lkmtp) || ip_mrtproto) 2259 return(EEXIST); 2260 old_ip_mrouter_cmd = ip_mrouter_cmd; 2261 ip_mrouter_cmd = X_ip_mrouter_cmd; 2262 old_ip_mrouter_done = ip_mrouter_done; 2263 ip_mrouter_done = X_ip_mrouter_done; 2264 old_ip_mforward = ip_mforward; 2265 ip_mforward = X_ip_mforward; 2266 old_mrt_ioctl = mrt_ioctl; 2267 mrt_ioctl = X_mrt_ioctl; 2268 old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input; 2269 inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input; 2270 old_legal_vif_num = legal_vif_num; 2271 legal_vif_num = X_legal_vif_num; 2272 ip_mrtproto = IGMP_DVMRP; 2273 2274 printf("\nIP multicast routing loaded\n"); 2275 break; 2276 2277 case LKM_E_UNLOAD: 2278 if (ip_mrouter) 2279 return EINVAL; 2280 2281 ip_mrouter_cmd = old_ip_mrouter_cmd; 2282 ip_mrouter_done = old_ip_mrouter_done; 2283 ip_mforward = old_ip_mforward; 2284 mrt_ioctl = old_mrt_ioctl; 2285 inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input; 2286 legal_vif_num = old_legal_vif_num; 2287 ip_mrtproto = 0; 2288 break; 2289 2290 default: 2291 err = EINVAL; 2292 break; 2293 } 2294 2295 return(err); 2296} 2297 2298int 2299ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) { 2300 DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle, 2301 nosys); 2302} 2303 2304#endif /* MROUTE_LKM */ 2305#endif /* MROUTING */ 2306