ip_mroute.c revision 65986
1/* 2 * IP multicast forwarding procedures 3 * 4 * Written by David Waitzman, BBN Labs, August 1988. 5 * Modified by Steve Deering, Stanford, February 1989. 6 * Modified by Mark J. Steiglitz, Stanford, May, 1991 7 * Modified by Van Jacobson, LBL, January 1993 8 * Modified by Ajit Thyagarajan, PARC, August 1993 9 * Modified by Bill Fenner, PARC, April 1995 10 * 11 * MROUTING Revision: 3.5 12 * $FreeBSD: head/sys/netinet/ip_mroute.c 65986 2000-09-17 13:50:12Z kjc $ 13 */ 14 15#include "opt_mrouting.h" 16 17#include <sys/param.h> 18#include <sys/systm.h> 19#include <sys/malloc.h> 20#include <sys/mbuf.h> 21#include <sys/socket.h> 22#include <sys/socketvar.h> 23#include <sys/protosw.h> 24#include <sys/time.h> 25#include <sys/kernel.h> 26#include <sys/sockio.h> 27#include <sys/syslog.h> 28#include <net/if.h> 29#include <net/route.h> 30#include <netinet/in.h> 31#include <netinet/in_systm.h> 32#include <netinet/ip.h> 33#include <netinet/ip_var.h> 34#include <netinet/in_var.h> 35#include <netinet/igmp.h> 36#include <netinet/ip_mroute.h> 37#include <netinet/udp.h> 38#include <machine/in_cksum.h> 39 40#ifndef NTOHL 41#if BYTE_ORDER != BIG_ENDIAN 42#define NTOHL(d) ((d) = ntohl((d))) 43#define NTOHS(d) ((d) = ntohs((u_short)(d))) 44#define HTONL(d) ((d) = htonl((d))) 45#define HTONS(d) ((d) = htons((u_short)(d))) 46#else 47#define NTOHL(d) 48#define NTOHS(d) 49#define HTONL(d) 50#define HTONS(d) 51#endif 52#endif 53 54#ifndef MROUTING 55extern u_long _ip_mcast_src __P((int vifi)); 56extern int _ip_mforward __P((struct ip *ip, struct ifnet *ifp, 57 struct mbuf *m, struct ip_moptions *imo)); 58extern int _ip_mrouter_done __P((void)); 59extern int _ip_mrouter_get __P((struct socket *so, struct sockopt *sopt)); 60extern int _ip_mrouter_set __P((struct socket *so, struct sockopt *sopt)); 61extern int _mrt_ioctl __P((int req, caddr_t data, struct proc *p)); 62 63/* 64 * Dummy routines and globals used when multicast routing is not compiled in. 65 */ 66 67struct socket *ip_mrouter = NULL; 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, off, proto) /* XXX must fixup manually */ 122 struct mbuf *m; 123 int off; 124 int proto; 125{ 126 /* Can still get packets with rsvp_on = 0 if there is a local member 127 * of the group to which the RSVP packet is addressed. But in this 128 * case we want to throw the packet away. 129 */ 130 if (!rsvp_on) { 131 m_freem(m); 132 return; 133 } 134 135 if (ip_rsvpd != NULL) { 136 if (rsvpdebug) 137 printf("rsvp_input: Sending packet up old-style socket\n"); 138 rip_input(m, off, proto); 139 return; 140 } 141 /* Drop the packet */ 142 m_freem(m); 143} 144 145void ipip_input(struct mbuf *m, int off, int proto) { /* XXX must fixup manually */ 146 rip_input(m, off, proto); 147} 148 149int (*legal_vif_num)(int) = 0; 150 151/* 152 * This should never be called, since IP_MULTICAST_VIF should fail, but 153 * just in case it does get called, the code a little lower in ip_output 154 * will assign the packet a local address. 155 */ 156u_long 157_ip_mcast_src(int vifi) { return INADDR_ANY; } 158u_long (*ip_mcast_src)(int) = _ip_mcast_src; 159 160int 161ip_rsvp_vif_init(so, sopt) 162 struct socket *so; 163 struct sockopt *sopt; 164{ 165 return(EINVAL); 166} 167 168int 169ip_rsvp_vif_done(so, sopt) 170 struct socket *so; 171 struct sockopt *sopt; 172{ 173 return(EINVAL); 174} 175 176void 177ip_rsvp_force_done(so) 178 struct socket *so; 179{ 180 return; 181} 182 183#else /* MROUTING */ 184 185#define M_HASCL(m) ((m)->m_flags & M_EXT) 186 187#define INSIZ sizeof(struct in_addr) 188#define same(a1, a2) \ 189 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0) 190 191static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 192 193/* 194 * Globals. All but ip_mrouter and ip_mrtproto could be static, 195 * except for netstat or debugging purposes. 196 */ 197#ifndef MROUTE_LKM 198struct socket *ip_mrouter = NULL; 199static struct mrtstat mrtstat; 200#else /* MROUTE_LKM */ 201extern void X_ipip_input __P((struct mbuf *m, int iphlen)); 202extern struct mrtstat mrtstat; 203static int ip_mrtproto; 204#endif 205 206#define NO_RTE_FOUND 0x1 207#define RTE_FOUND 0x2 208 209static struct mfc *mfctable[MFCTBLSIZ]; 210static u_char nexpire[MFCTBLSIZ]; 211static struct vif viftable[MAXVIFS]; 212static u_int mrtdebug = 0; /* debug level */ 213#define DEBUG_MFC 0x02 214#define DEBUG_FORWARD 0x04 215#define DEBUG_EXPIRE 0x08 216#define DEBUG_XMIT 0x10 217static u_int tbfdebug = 0; /* tbf debug level */ 218static u_int rsvpdebug = 0; /* rsvp debug level */ 219 220static struct callout_handle expire_upcalls_ch; 221 222#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 223#define UPCALL_EXPIRE 6 /* number of timeouts */ 224 225/* 226 * Define the token bucket filter structures 227 * tbftable -> each vif has one of these for storing info 228 */ 229 230static struct tbf tbftable[MAXVIFS]; 231#define TBF_REPROCESS (hz / 100) /* 100x / second */ 232 233/* 234 * 'Interfaces' associated with decapsulator (so we can tell 235 * packets that went through it from ones that get reflected 236 * by a broken gateway). These interfaces are never linked into 237 * the system ifnet list & no routes point to them. I.e., packets 238 * can't be sent this way. They only exist as a placeholder for 239 * multicast source verification. 240 */ 241static struct ifnet multicast_decap_if[MAXVIFS]; 242 243#define ENCAP_TTL 64 244#define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 245 246/* prototype IP hdr for encapsulated packets */ 247static struct ip multicast_encap_iphdr = { 248#if BYTE_ORDER == LITTLE_ENDIAN 249 sizeof(struct ip) >> 2, IPVERSION, 250#else 251 IPVERSION, sizeof(struct ip) >> 2, 252#endif 253 0, /* tos */ 254 sizeof(struct ip), /* total length */ 255 0, /* id */ 256 0, /* frag offset */ 257 ENCAP_TTL, ENCAP_PROTO, 258 0, /* checksum */ 259}; 260 261/* 262 * Private variables. 263 */ 264static vifi_t numvifs = 0; 265static int have_encap_tunnel = 0; 266 267/* 268 * one-back cache used by ipip_input to locate a tunnel's vif 269 * given a datagram's src ip address. 270 */ 271static u_long last_encap_src; 272static struct vif *last_encap_vif; 273 274static u_long X_ip_mcast_src __P((int vifi)); 275static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo)); 276static int X_ip_mrouter_done __P((void)); 277static int X_ip_mrouter_get __P((struct socket *so, struct sockopt *m)); 278static int X_ip_mrouter_set __P((struct socket *so, struct sockopt *m)); 279static int X_legal_vif_num __P((int vif)); 280static int X_mrt_ioctl __P((int cmd, caddr_t data)); 281 282static int get_sg_cnt(struct sioc_sg_req *); 283static int get_vif_cnt(struct sioc_vif_req *); 284static int ip_mrouter_init(struct socket *, int); 285static int add_vif(struct vifctl *); 286static int del_vif(vifi_t); 287static int add_mfc(struct mfcctl *); 288static int del_mfc(struct mfcctl *); 289static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 290static int set_assert(int); 291static void expire_upcalls(void *); 292static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, 293 vifi_t); 294static void phyint_send(struct ip *, struct vif *, struct mbuf *); 295static void encap_send(struct ip *, struct vif *, struct mbuf *); 296static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 297static void tbf_queue(struct vif *, struct mbuf *); 298static void tbf_process_q(struct vif *); 299static void tbf_reprocess_q(void *); 300static int tbf_dq_sel(struct vif *, struct ip *); 301static void tbf_send_packet(struct vif *, struct mbuf *); 302static void tbf_update_tokens(struct vif *); 303static int priority(struct vif *, struct ip *); 304void multiencap_decap(struct mbuf *); 305 306/* 307 * whether or not special PIM assert processing is enabled. 308 */ 309static int pim_assert; 310/* 311 * Rate limit for assert notification messages, in usec 312 */ 313#define ASSERT_MSG_TIME 3000000 314 315/* 316 * Hash function for a source, group entry 317 */ 318#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 319 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 320 321/* 322 * Find a route for a given origin IP address and Multicast group address 323 * Type of service parameter to be added in the future!!! 324 */ 325 326#define MFCFIND(o, g, rt) { \ 327 register struct mfc *_rt = mfctable[MFCHASH(o,g)]; \ 328 rt = NULL; \ 329 ++mrtstat.mrts_mfc_lookups; \ 330 while (_rt) { \ 331 if ((_rt->mfc_origin.s_addr == o) && \ 332 (_rt->mfc_mcastgrp.s_addr == g) && \ 333 (_rt->mfc_stall == NULL)) { \ 334 rt = _rt; \ 335 break; \ 336 } \ 337 _rt = _rt->mfc_next; \ 338 } \ 339 if (rt == NULL) { \ 340 ++mrtstat.mrts_mfc_misses; \ 341 } \ 342} 343 344 345/* 346 * Macros to compute elapsed time efficiently 347 * Borrowed from Van Jacobson's scheduling code 348 */ 349#define TV_DELTA(a, b, delta) { \ 350 register int xxs; \ 351 \ 352 delta = (a).tv_usec - (b).tv_usec; \ 353 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 354 switch (xxs) { \ 355 case 2: \ 356 delta += 1000000; \ 357 /* fall through */ \ 358 case 1: \ 359 delta += 1000000; \ 360 break; \ 361 default: \ 362 delta += (1000000 * xxs); \ 363 } \ 364 } \ 365} 366 367#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 368 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 369 370#ifdef UPCALL_TIMING 371u_long upcall_data[51]; 372static void collate(struct timeval *); 373#endif /* UPCALL_TIMING */ 374 375 376/* 377 * Handle MRT setsockopt commands to modify the multicast routing tables. 378 */ 379static int 380X_ip_mrouter_set(so, sopt) 381 struct socket *so; 382 struct sockopt *sopt; 383{ 384 int error, optval; 385 vifi_t vifi; 386 struct vifctl vifc; 387 struct mfcctl mfc; 388 389 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 390 return (EPERM); 391 392 error = 0; 393 switch (sopt->sopt_name) { 394 case MRT_INIT: 395 error = sooptcopyin(sopt, &optval, sizeof optval, 396 sizeof optval); 397 if (error) 398 break; 399 error = ip_mrouter_init(so, optval); 400 break; 401 402 case MRT_DONE: 403 error = ip_mrouter_done(); 404 break; 405 406 case MRT_ADD_VIF: 407 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 408 if (error) 409 break; 410 error = add_vif(&vifc); 411 break; 412 413 case MRT_DEL_VIF: 414 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 415 if (error) 416 break; 417 error = del_vif(vifi); 418 break; 419 420 case MRT_ADD_MFC: 421 case MRT_DEL_MFC: 422 error = sooptcopyin(sopt, &mfc, sizeof mfc, sizeof mfc); 423 if (error) 424 break; 425 if (sopt->sopt_name == MRT_ADD_MFC) 426 error = add_mfc(&mfc); 427 else 428 error = del_mfc(&mfc); 429 break; 430 431 case MRT_ASSERT: 432 error = sooptcopyin(sopt, &optval, sizeof optval, 433 sizeof optval); 434 if (error) 435 break; 436 set_assert(optval); 437 break; 438 439 default: 440 error = EOPNOTSUPP; 441 break; 442 } 443 return (error); 444} 445 446#ifndef MROUTE_LKM 447int (*ip_mrouter_set)(struct socket *, struct sockopt *) = X_ip_mrouter_set; 448#endif 449 450/* 451 * Handle MRT getsockopt commands 452 */ 453static int 454X_ip_mrouter_get(so, sopt) 455 struct socket *so; 456 struct sockopt *sopt; 457{ 458 int error; 459 static int version = 0x0305; /* !!! why is this here? XXX */ 460 461 switch (sopt->sopt_name) { 462 case MRT_VERSION: 463 error = sooptcopyout(sopt, &version, sizeof version); 464 break; 465 466 case MRT_ASSERT: 467 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 468 break; 469 default: 470 error = EOPNOTSUPP; 471 break; 472 } 473 return (error); 474} 475 476#ifndef MROUTE_LKM 477int (*ip_mrouter_get)(struct socket *, struct sockopt *) = X_ip_mrouter_get; 478#endif 479 480/* 481 * Handle ioctl commands to obtain information from the cache 482 */ 483static int 484X_mrt_ioctl(cmd, data) 485 int cmd; 486 caddr_t data; 487{ 488 int error = 0; 489 490 switch (cmd) { 491 case (SIOCGETVIFCNT): 492 return (get_vif_cnt((struct sioc_vif_req *)data)); 493 break; 494 case (SIOCGETSGCNT): 495 return (get_sg_cnt((struct sioc_sg_req *)data)); 496 break; 497 default: 498 return (EINVAL); 499 break; 500 } 501 return error; 502} 503 504#ifndef MROUTE_LKM 505int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl; 506#endif 507 508/* 509 * returns the packet, byte, rpf-failure count for the source group provided 510 */ 511static int 512get_sg_cnt(req) 513 register struct sioc_sg_req *req; 514{ 515 register struct mfc *rt; 516 int s; 517 518 s = splnet(); 519 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 520 splx(s); 521 if (rt != NULL) { 522 req->pktcnt = rt->mfc_pkt_cnt; 523 req->bytecnt = rt->mfc_byte_cnt; 524 req->wrong_if = rt->mfc_wrong_if; 525 } else 526 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 527 528 return 0; 529} 530 531/* 532 * returns the input and output packet and byte counts on the vif provided 533 */ 534static int 535get_vif_cnt(req) 536 register struct sioc_vif_req *req; 537{ 538 register vifi_t vifi = req->vifi; 539 540 if (vifi >= numvifs) return EINVAL; 541 542 req->icount = viftable[vifi].v_pkt_in; 543 req->ocount = viftable[vifi].v_pkt_out; 544 req->ibytes = viftable[vifi].v_bytes_in; 545 req->obytes = viftable[vifi].v_bytes_out; 546 547 return 0; 548} 549 550/* 551 * Enable multicast routing 552 */ 553static int 554ip_mrouter_init(so, version) 555 struct socket *so; 556 int version; 557{ 558 if (mrtdebug) 559 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 560 so->so_type, so->so_proto->pr_protocol); 561 562 if (so->so_type != SOCK_RAW || 563 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP; 564 565 if (version != 1) 566 return ENOPROTOOPT; 567 568 if (ip_mrouter != NULL) return EADDRINUSE; 569 570 ip_mrouter = so; 571 572 bzero((caddr_t)mfctable, sizeof(mfctable)); 573 bzero((caddr_t)nexpire, sizeof(nexpire)); 574 575 pim_assert = 0; 576 577 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 578 579 if (mrtdebug) 580 log(LOG_DEBUG, "ip_mrouter_init\n"); 581 582 return 0; 583} 584 585/* 586 * Disable multicast routing 587 */ 588static int 589X_ip_mrouter_done() 590{ 591 vifi_t vifi; 592 int i; 593 struct ifnet *ifp; 594 struct ifreq ifr; 595 struct mfc *rt; 596 struct rtdetq *rte; 597 int s; 598 599 s = splnet(); 600 601 /* 602 * For each phyint in use, disable promiscuous reception of all IP 603 * multicasts. 604 */ 605 for (vifi = 0; vifi < numvifs; vifi++) { 606 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 607 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 608 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 609 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr 610 = INADDR_ANY; 611 ifp = viftable[vifi].v_ifp; 612 if_allmulti(ifp, 0); 613 } 614 } 615 bzero((caddr_t)tbftable, sizeof(tbftable)); 616 bzero((caddr_t)viftable, sizeof(viftable)); 617 numvifs = 0; 618 pim_assert = 0; 619 620 untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch); 621 622 /* 623 * Free all multicast forwarding cache entries. 624 */ 625 for (i = 0; i < MFCTBLSIZ; i++) { 626 for (rt = mfctable[i]; rt != NULL; ) { 627 struct mfc *nr = rt->mfc_next; 628 629 for (rte = rt->mfc_stall; rte != NULL; ) { 630 struct rtdetq *n = rte->next; 631 632 m_freem(rte->m); 633 free(rte, M_MRTABLE); 634 rte = n; 635 } 636 free(rt, M_MRTABLE); 637 rt = nr; 638 } 639 } 640 641 bzero((caddr_t)mfctable, sizeof(mfctable)); 642 643 /* 644 * Reset de-encapsulation cache 645 */ 646 last_encap_src = 0; 647 last_encap_vif = NULL; 648 have_encap_tunnel = 0; 649 650 ip_mrouter = NULL; 651 652 splx(s); 653 654 if (mrtdebug) 655 log(LOG_DEBUG, "ip_mrouter_done\n"); 656 657 return 0; 658} 659 660#ifndef MROUTE_LKM 661int (*ip_mrouter_done)(void) = X_ip_mrouter_done; 662#endif 663 664/* 665 * Set PIM assert processing global 666 */ 667static int 668set_assert(i) 669 int i; 670{ 671 if ((i != 1) && (i != 0)) 672 return EINVAL; 673 674 pim_assert = i; 675 676 return 0; 677} 678 679/* 680 * Add a vif to the vif table 681 */ 682static int 683add_vif(vifcp) 684 register struct vifctl *vifcp; 685{ 686 register struct vif *vifp = viftable + vifcp->vifc_vifi; 687 static struct sockaddr_in sin = {sizeof sin, AF_INET}; 688 struct ifaddr *ifa; 689 struct ifnet *ifp; 690 int error, s; 691 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 692 693 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL; 694 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE; 695 696 /* Find the interface with an address in AF_INET family */ 697 sin.sin_addr = vifcp->vifc_lcl_addr; 698 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 699 if (ifa == 0) return EADDRNOTAVAIL; 700 ifp = ifa->ifa_ifp; 701 702 if (vifcp->vifc_flags & VIFF_TUNNEL) { 703 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 704 /* 705 * An encapsulating tunnel is wanted. Tell ipip_input() to 706 * start paying attention to encapsulated packets. 707 */ 708 if (have_encap_tunnel == 0) { 709 have_encap_tunnel = 1; 710 for (s = 0; s < MAXVIFS; ++s) { 711 multicast_decap_if[s].if_name = "mdecap"; 712 multicast_decap_if[s].if_unit = s; 713 } 714 } 715 /* 716 * Set interface to fake encapsulator interface 717 */ 718 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 719 /* 720 * Prepare cached route entry 721 */ 722 bzero(&vifp->v_route, sizeof(vifp->v_route)); 723 } else { 724 log(LOG_ERR, "source routed tunnels not supported\n"); 725 return EOPNOTSUPP; 726 } 727 } else { 728 /* Make sure the interface supports multicast */ 729 if ((ifp->if_flags & IFF_MULTICAST) == 0) 730 return EOPNOTSUPP; 731 732 /* Enable promiscuous reception of all IP multicasts from the if */ 733 s = splnet(); 734 error = if_allmulti(ifp, 1); 735 splx(s); 736 if (error) 737 return error; 738 } 739 740 s = splnet(); 741 /* define parameters for the tbf structure */ 742 vifp->v_tbf = v_tbf; 743 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 744 vifp->v_tbf->tbf_n_tok = 0; 745 vifp->v_tbf->tbf_q_len = 0; 746 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 747 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 748 749 vifp->v_flags = vifcp->vifc_flags; 750 vifp->v_threshold = vifcp->vifc_threshold; 751 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 752 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 753 vifp->v_ifp = ifp; 754 /* scaling up here allows division by 1024 in critical code */ 755 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 756 vifp->v_rsvp_on = 0; 757 vifp->v_rsvpd = NULL; 758 /* initialize per vif pkt counters */ 759 vifp->v_pkt_in = 0; 760 vifp->v_pkt_out = 0; 761 vifp->v_bytes_in = 0; 762 vifp->v_bytes_out = 0; 763 splx(s); 764 765 /* Adjust numvifs up if the vifi is higher than numvifs */ 766 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 767 768 if (mrtdebug) 769 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n", 770 vifcp->vifc_vifi, 771 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 772 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 773 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 774 vifcp->vifc_threshold, 775 vifcp->vifc_rate_limit); 776 777 return 0; 778} 779 780/* 781 * Delete a vif from the vif table 782 */ 783static int 784del_vif(vifi) 785 vifi_t vifi; 786{ 787 register struct vif *vifp = &viftable[vifi]; 788 register struct mbuf *m; 789 struct ifnet *ifp; 790 struct ifreq ifr; 791 int s; 792 793 if (vifi >= numvifs) return EINVAL; 794 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL; 795 796 s = splnet(); 797 798 if (!(vifp->v_flags & VIFF_TUNNEL)) { 799 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 800 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 801 ifp = vifp->v_ifp; 802 if_allmulti(ifp, 0); 803 } 804 805 if (vifp == last_encap_vif) { 806 last_encap_vif = 0; 807 last_encap_src = 0; 808 } 809 810 /* 811 * Free packets queued at the interface 812 */ 813 while (vifp->v_tbf->tbf_q) { 814 m = vifp->v_tbf->tbf_q; 815 vifp->v_tbf->tbf_q = m->m_act; 816 m_freem(m); 817 } 818 819 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 820 bzero((caddr_t)vifp, sizeof (*vifp)); 821 822 if (mrtdebug) 823 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 824 825 /* Adjust numvifs down */ 826 for (vifi = numvifs; vifi > 0; vifi--) 827 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break; 828 numvifs = vifi; 829 830 splx(s); 831 832 return 0; 833} 834 835/* 836 * Add an mfc entry 837 */ 838static int 839add_mfc(mfccp) 840 struct mfcctl *mfccp; 841{ 842 struct mfc *rt; 843 u_long hash; 844 struct rtdetq *rte; 845 register u_short nstl; 846 int s; 847 int i; 848 849 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 850 851 /* If an entry already exists, just update the fields */ 852 if (rt) { 853 if (mrtdebug & DEBUG_MFC) 854 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 855 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 856 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 857 mfccp->mfcc_parent); 858 859 s = splnet(); 860 rt->mfc_parent = mfccp->mfcc_parent; 861 for (i = 0; i < numvifs; i++) 862 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 863 splx(s); 864 return 0; 865 } 866 867 /* 868 * Find the entry for which the upcall was made and update 869 */ 870 s = splnet(); 871 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 872 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 873 874 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 875 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 876 (rt->mfc_stall != NULL)) { 877 878 if (nstl++) 879 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 880 "multiple kernel entries", 881 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 882 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 883 mfccp->mfcc_parent, (void *)rt->mfc_stall); 884 885 if (mrtdebug & DEBUG_MFC) 886 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 887 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 888 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 889 mfccp->mfcc_parent, (void *)rt->mfc_stall); 890 891 rt->mfc_origin = mfccp->mfcc_origin; 892 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 893 rt->mfc_parent = mfccp->mfcc_parent; 894 for (i = 0; i < numvifs; i++) 895 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 896 /* initialize pkt counters per src-grp */ 897 rt->mfc_pkt_cnt = 0; 898 rt->mfc_byte_cnt = 0; 899 rt->mfc_wrong_if = 0; 900 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 901 902 rt->mfc_expire = 0; /* Don't clean this guy up */ 903 nexpire[hash]--; 904 905 /* free packets Qed at the end of this entry */ 906 for (rte = rt->mfc_stall; rte != NULL; ) { 907 struct rtdetq *n = rte->next; 908 909 ip_mdq(rte->m, rte->ifp, rt, -1); 910 m_freem(rte->m); 911#ifdef UPCALL_TIMING 912 collate(&(rte->t)); 913#endif /* UPCALL_TIMING */ 914 free(rte, M_MRTABLE); 915 rte = n; 916 } 917 rt->mfc_stall = NULL; 918 } 919 } 920 921 /* 922 * It is possible that an entry is being inserted without an upcall 923 */ 924 if (nstl == 0) { 925 if (mrtdebug & DEBUG_MFC) 926 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 927 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 928 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 929 mfccp->mfcc_parent); 930 931 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 932 933 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 934 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 935 936 rt->mfc_origin = mfccp->mfcc_origin; 937 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 938 rt->mfc_parent = mfccp->mfcc_parent; 939 for (i = 0; i < numvifs; i++) 940 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 941 /* initialize pkt counters per src-grp */ 942 rt->mfc_pkt_cnt = 0; 943 rt->mfc_byte_cnt = 0; 944 rt->mfc_wrong_if = 0; 945 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 946 if (rt->mfc_expire) 947 nexpire[hash]--; 948 rt->mfc_expire = 0; 949 } 950 } 951 if (rt == NULL) { 952 /* no upcall, so make a new entry */ 953 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 954 if (rt == NULL) { 955 splx(s); 956 return ENOBUFS; 957 } 958 959 /* insert new entry at head of hash chain */ 960 rt->mfc_origin = mfccp->mfcc_origin; 961 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 962 rt->mfc_parent = mfccp->mfcc_parent; 963 for (i = 0; i < numvifs; i++) 964 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 965 /* initialize pkt counters per src-grp */ 966 rt->mfc_pkt_cnt = 0; 967 rt->mfc_byte_cnt = 0; 968 rt->mfc_wrong_if = 0; 969 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 970 rt->mfc_expire = 0; 971 rt->mfc_stall = NULL; 972 973 /* link into table */ 974 rt->mfc_next = mfctable[hash]; 975 mfctable[hash] = rt; 976 } 977 } 978 splx(s); 979 return 0; 980} 981 982#ifdef UPCALL_TIMING 983/* 984 * collect delay statistics on the upcalls 985 */ 986static void collate(t) 987register struct timeval *t; 988{ 989 register u_long d; 990 register struct timeval tp; 991 register u_long delta; 992 993 GET_TIME(tp); 994 995 if (TV_LT(*t, tp)) 996 { 997 TV_DELTA(tp, *t, delta); 998 999 d = delta >> 10; 1000 if (d > 50) 1001 d = 50; 1002 1003 ++upcall_data[d]; 1004 } 1005} 1006#endif /* UPCALL_TIMING */ 1007 1008/* 1009 * Delete an mfc entry 1010 */ 1011static int 1012del_mfc(mfccp) 1013 struct mfcctl *mfccp; 1014{ 1015 struct in_addr origin; 1016 struct in_addr mcastgrp; 1017 struct mfc *rt; 1018 struct mfc **nptr; 1019 u_long hash; 1020 int s; 1021 1022 origin = mfccp->mfcc_origin; 1023 mcastgrp = mfccp->mfcc_mcastgrp; 1024 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1025 1026 if (mrtdebug & DEBUG_MFC) 1027 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1028 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1029 1030 s = splnet(); 1031 1032 nptr = &mfctable[hash]; 1033 while ((rt = *nptr) != NULL) { 1034 if (origin.s_addr == rt->mfc_origin.s_addr && 1035 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1036 rt->mfc_stall == NULL) 1037 break; 1038 1039 nptr = &rt->mfc_next; 1040 } 1041 if (rt == NULL) { 1042 splx(s); 1043 return EADDRNOTAVAIL; 1044 } 1045 1046 *nptr = rt->mfc_next; 1047 free(rt, M_MRTABLE); 1048 1049 splx(s); 1050 1051 return 0; 1052} 1053 1054/* 1055 * Send a message to mrouted on the multicast routing socket 1056 */ 1057static int 1058socket_send(s, mm, src) 1059 struct socket *s; 1060 struct mbuf *mm; 1061 struct sockaddr_in *src; 1062{ 1063 if (s) { 1064 if (sbappendaddr(&s->so_rcv, 1065 (struct sockaddr *)src, 1066 mm, (struct mbuf *)0) != 0) { 1067 sorwakeup(s); 1068 return 0; 1069 } 1070 } 1071 m_freem(mm); 1072 return -1; 1073} 1074 1075/* 1076 * IP multicast forwarding function. This function assumes that the packet 1077 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1078 * pointed to by "ifp", and the packet is to be relayed to other networks 1079 * that have members of the packet's destination IP multicast group. 1080 * 1081 * The packet is returned unscathed to the caller, unless it is 1082 * erroneous, in which case a non-zero return value tells the caller to 1083 * discard it. 1084 */ 1085 1086#define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1087#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1088 1089static int 1090X_ip_mforward(ip, ifp, m, imo) 1091 register struct ip *ip; 1092 struct ifnet *ifp; 1093 struct mbuf *m; 1094 struct ip_moptions *imo; 1095{ 1096 register struct mfc *rt; 1097 register u_char *ipoptions; 1098 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1099 static int srctun = 0; 1100 register struct mbuf *mm; 1101 int s; 1102 vifi_t vifi; 1103 struct vif *vifp; 1104 1105 if (mrtdebug & DEBUG_FORWARD) 1106 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1107 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1108 (void *)ifp); 1109 1110 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1111 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1112 /* 1113 * Packet arrived via a physical interface or 1114 * an encapsulated tunnel. 1115 */ 1116 } else { 1117 /* 1118 * Packet arrived through a source-route tunnel. 1119 * Source-route tunnels are no longer supported. 1120 */ 1121 if ((srctun++ % 1000) == 0) 1122 log(LOG_ERR, 1123 "ip_mforward: received source-routed packet from %lx\n", 1124 (u_long)ntohl(ip->ip_src.s_addr)); 1125 1126 return 1; 1127 } 1128 1129 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1130 if (ip->ip_ttl < 255) 1131 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1132 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1133 vifp = viftable + vifi; 1134 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1135 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi, 1136 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1137 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1138 } 1139 return (ip_mdq(m, ifp, NULL, vifi)); 1140 } 1141 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1142 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1143 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr)); 1144 if(!imo) 1145 printf("In fact, no options were specified at all\n"); 1146 } 1147 1148 /* 1149 * Don't forward a packet with time-to-live of zero or one, 1150 * or a packet destined to a local-only group. 1151 */ 1152 if (ip->ip_ttl <= 1 || 1153 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1154 return 0; 1155 1156 /* 1157 * Determine forwarding vifs from the forwarding cache table 1158 */ 1159 s = splnet(); 1160 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt); 1161 1162 /* Entry exists, so forward if necessary */ 1163 if (rt != NULL) { 1164 splx(s); 1165 return (ip_mdq(m, ifp, rt, -1)); 1166 } else { 1167 /* 1168 * If we don't have a route for packet's origin, 1169 * Make a copy of the packet & 1170 * send message to routing daemon 1171 */ 1172 1173 register struct mbuf *mb0; 1174 register struct rtdetq *rte; 1175 register u_long hash; 1176 int hlen = ip->ip_hl << 2; 1177#ifdef UPCALL_TIMING 1178 struct timeval tp; 1179 1180 GET_TIME(tp); 1181#endif 1182 1183 mrtstat.mrts_no_route++; 1184 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1185 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1186 (u_long)ntohl(ip->ip_src.s_addr), 1187 (u_long)ntohl(ip->ip_dst.s_addr)); 1188 1189 /* 1190 * Allocate mbufs early so that we don't do extra work if we are 1191 * just going to fail anyway. Make sure to pullup the header so 1192 * that other people can't step on it. 1193 */ 1194 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT); 1195 if (rte == NULL) { 1196 splx(s); 1197 return ENOBUFS; 1198 } 1199 mb0 = m_copy(m, 0, M_COPYALL); 1200 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1201 mb0 = m_pullup(mb0, hlen); 1202 if (mb0 == NULL) { 1203 free(rte, M_MRTABLE); 1204 splx(s); 1205 return ENOBUFS; 1206 } 1207 1208 /* is there an upcall waiting for this packet? */ 1209 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1210 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) { 1211 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1212 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1213 (rt->mfc_stall != NULL)) 1214 break; 1215 } 1216 1217 if (rt == NULL) { 1218 int i; 1219 struct igmpmsg *im; 1220 1221 /* no upcall, so make a new entry */ 1222 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1223 if (rt == NULL) { 1224 free(rte, M_MRTABLE); 1225 m_freem(mb0); 1226 splx(s); 1227 return ENOBUFS; 1228 } 1229 /* Make a copy of the header to send to the user level process */ 1230 mm = m_copy(mb0, 0, hlen); 1231 if (mm == NULL) { 1232 free(rte, M_MRTABLE); 1233 m_freem(mb0); 1234 free(rt, M_MRTABLE); 1235 splx(s); 1236 return ENOBUFS; 1237 } 1238 1239 /* 1240 * Send message to routing daemon to install 1241 * a route into the kernel table 1242 */ 1243 k_igmpsrc.sin_addr = ip->ip_src; 1244 1245 im = mtod(mm, struct igmpmsg *); 1246 im->im_msgtype = IGMPMSG_NOCACHE; 1247 im->im_mbz = 0; 1248 1249 mrtstat.mrts_upcalls++; 1250 1251 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1252 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1253 ++mrtstat.mrts_upq_sockfull; 1254 free(rte, M_MRTABLE); 1255 m_freem(mb0); 1256 free(rt, M_MRTABLE); 1257 splx(s); 1258 return ENOBUFS; 1259 } 1260 1261 /* insert new entry at head of hash chain */ 1262 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1263 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1264 rt->mfc_expire = UPCALL_EXPIRE; 1265 nexpire[hash]++; 1266 for (i = 0; i < numvifs; i++) 1267 rt->mfc_ttls[i] = 0; 1268 rt->mfc_parent = -1; 1269 1270 /* link into table */ 1271 rt->mfc_next = mfctable[hash]; 1272 mfctable[hash] = rt; 1273 rt->mfc_stall = rte; 1274 1275 } else { 1276 /* determine if q has overflowed */ 1277 int npkts = 0; 1278 struct rtdetq **p; 1279 1280 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1281 npkts++; 1282 1283 if (npkts > MAX_UPQ) { 1284 mrtstat.mrts_upq_ovflw++; 1285 free(rte, M_MRTABLE); 1286 m_freem(mb0); 1287 splx(s); 1288 return 0; 1289 } 1290 1291 /* Add this entry to the end of the queue */ 1292 *p = rte; 1293 } 1294 1295 rte->m = mb0; 1296 rte->ifp = ifp; 1297#ifdef UPCALL_TIMING 1298 rte->t = tp; 1299#endif 1300 rte->next = NULL; 1301 1302 splx(s); 1303 1304 return 0; 1305 } 1306} 1307 1308#ifndef MROUTE_LKM 1309int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 1310 struct ip_moptions *) = X_ip_mforward; 1311#endif 1312 1313/* 1314 * Clean up the cache entry if upcall is not serviced 1315 */ 1316static void 1317expire_upcalls(void *unused) 1318{ 1319 struct rtdetq *rte; 1320 struct mfc *mfc, **nptr; 1321 int i; 1322 int s; 1323 1324 s = splnet(); 1325 for (i = 0; i < MFCTBLSIZ; i++) { 1326 if (nexpire[i] == 0) 1327 continue; 1328 nptr = &mfctable[i]; 1329 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1330 /* 1331 * Skip real cache entries 1332 * Make sure it wasn't marked to not expire (shouldn't happen) 1333 * If it expires now 1334 */ 1335 if (mfc->mfc_stall != NULL && 1336 mfc->mfc_expire != 0 && 1337 --mfc->mfc_expire == 0) { 1338 if (mrtdebug & DEBUG_EXPIRE) 1339 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1340 (u_long)ntohl(mfc->mfc_origin.s_addr), 1341 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1342 /* 1343 * drop all the packets 1344 * free the mbuf with the pkt, if, timing info 1345 */ 1346 for (rte = mfc->mfc_stall; rte; ) { 1347 struct rtdetq *n = rte->next; 1348 1349 m_freem(rte->m); 1350 free(rte, M_MRTABLE); 1351 rte = n; 1352 } 1353 ++mrtstat.mrts_cache_cleanups; 1354 nexpire[i]--; 1355 1356 *nptr = mfc->mfc_next; 1357 free(mfc, M_MRTABLE); 1358 } else { 1359 nptr = &mfc->mfc_next; 1360 } 1361 } 1362 } 1363 splx(s); 1364 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 1365} 1366 1367/* 1368 * Packet forwarding routine once entry in the cache is made 1369 */ 1370static int 1371ip_mdq(m, ifp, rt, xmt_vif) 1372 register struct mbuf *m; 1373 register struct ifnet *ifp; 1374 register struct mfc *rt; 1375 register vifi_t xmt_vif; 1376{ 1377 register struct ip *ip = mtod(m, struct ip *); 1378 register vifi_t vifi; 1379 register struct vif *vifp; 1380 register int plen = ip->ip_len; 1381 1382/* 1383 * Macro to send packet on vif. Since RSVP packets don't get counted on 1384 * input, they shouldn't get counted on output, so statistics keeping is 1385 * seperate. 1386 */ 1387#define MC_SEND(ip,vifp,m) { \ 1388 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1389 encap_send((ip), (vifp), (m)); \ 1390 else \ 1391 phyint_send((ip), (vifp), (m)); \ 1392} 1393 1394 /* 1395 * If xmt_vif is not -1, send on only the requested vif. 1396 * 1397 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1398 */ 1399 if (xmt_vif < numvifs) { 1400 MC_SEND(ip, viftable + xmt_vif, m); 1401 return 1; 1402 } 1403 1404 /* 1405 * Don't forward if it didn't arrive from the parent vif for its origin. 1406 */ 1407 vifi = rt->mfc_parent; 1408 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1409 /* came in the wrong interface */ 1410 if (mrtdebug & DEBUG_FORWARD) 1411 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1412 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1413 ++mrtstat.mrts_wrong_if; 1414 ++rt->mfc_wrong_if; 1415 /* 1416 * If we are doing PIM assert processing, and we are forwarding 1417 * packets on this interface, and it is a broadcast medium 1418 * interface (and not a tunnel), send a message to the routing daemon. 1419 */ 1420 if (pim_assert && rt->mfc_ttls[vifi] && 1421 (ifp->if_flags & IFF_BROADCAST) && 1422 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1423 struct sockaddr_in k_igmpsrc; 1424 struct mbuf *mm; 1425 struct igmpmsg *im; 1426 int hlen = ip->ip_hl << 2; 1427 struct timeval now; 1428 register u_long delta; 1429 1430 GET_TIME(now); 1431 1432 TV_DELTA(rt->mfc_last_assert, now, delta); 1433 1434 if (delta > ASSERT_MSG_TIME) { 1435 mm = m_copy(m, 0, hlen); 1436 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1437 mm = m_pullup(mm, hlen); 1438 if (mm == NULL) { 1439 return ENOBUFS; 1440 } 1441 1442 rt->mfc_last_assert = now; 1443 1444 im = mtod(mm, struct igmpmsg *); 1445 im->im_msgtype = IGMPMSG_WRONGVIF; 1446 im->im_mbz = 0; 1447 im->im_vif = vifi; 1448 1449 k_igmpsrc.sin_addr = im->im_src; 1450 1451 socket_send(ip_mrouter, mm, &k_igmpsrc); 1452 } 1453 } 1454 return 0; 1455 } 1456 1457 /* If I sourced this packet, it counts as output, else it was input. */ 1458 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1459 viftable[vifi].v_pkt_out++; 1460 viftable[vifi].v_bytes_out += plen; 1461 } else { 1462 viftable[vifi].v_pkt_in++; 1463 viftable[vifi].v_bytes_in += plen; 1464 } 1465 rt->mfc_pkt_cnt++; 1466 rt->mfc_byte_cnt += plen; 1467 1468 /* 1469 * For each vif, decide if a copy of the packet should be forwarded. 1470 * Forward if: 1471 * - the ttl exceeds the vif's threshold 1472 * - there are group members downstream on interface 1473 */ 1474 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1475 if ((rt->mfc_ttls[vifi] > 0) && 1476 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1477 vifp->v_pkt_out++; 1478 vifp->v_bytes_out += plen; 1479 MC_SEND(ip, vifp, m); 1480 } 1481 1482 return 0; 1483} 1484 1485/* 1486 * check if a vif number is legal/ok. This is used by ip_output, to export 1487 * numvifs there, 1488 */ 1489static int 1490X_legal_vif_num(vif) 1491 int vif; 1492{ 1493 if (vif >= 0 && vif < numvifs) 1494 return(1); 1495 else 1496 return(0); 1497} 1498 1499#ifndef MROUTE_LKM 1500int (*legal_vif_num)(int) = X_legal_vif_num; 1501#endif 1502 1503/* 1504 * Return the local address used by this vif 1505 */ 1506static u_long 1507X_ip_mcast_src(vifi) 1508 int vifi; 1509{ 1510 if (vifi >= 0 && vifi < numvifs) 1511 return viftable[vifi].v_lcl_addr.s_addr; 1512 else 1513 return INADDR_ANY; 1514} 1515 1516#ifndef MROUTE_LKM 1517u_long (*ip_mcast_src)(int) = X_ip_mcast_src; 1518#endif 1519 1520static void 1521phyint_send(ip, vifp, m) 1522 struct ip *ip; 1523 struct vif *vifp; 1524 struct mbuf *m; 1525{ 1526 register struct mbuf *mb_copy; 1527 register int hlen = ip->ip_hl << 2; 1528 1529 /* 1530 * Make a new reference to the packet; make sure that 1531 * the IP header is actually copied, not just referenced, 1532 * so that ip_output() only scribbles on the copy. 1533 */ 1534 mb_copy = m_copy(m, 0, M_COPYALL); 1535 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1536 mb_copy = m_pullup(mb_copy, hlen); 1537 if (mb_copy == NULL) 1538 return; 1539 1540 if (vifp->v_rate_limit == 0) 1541 tbf_send_packet(vifp, mb_copy); 1542 else 1543 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1544} 1545 1546static void 1547encap_send(ip, vifp, m) 1548 register struct ip *ip; 1549 register struct vif *vifp; 1550 register struct mbuf *m; 1551{ 1552 register struct mbuf *mb_copy; 1553 register struct ip *ip_copy; 1554 register int i, len = ip->ip_len; 1555 1556 /* 1557 * copy the old packet & pullup its IP header into the 1558 * new mbuf so we can modify it. Try to fill the new 1559 * mbuf since if we don't the ethernet driver will. 1560 */ 1561 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER); 1562 if (mb_copy == NULL) 1563 return; 1564 mb_copy->m_data += max_linkhdr; 1565 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1566 1567 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1568 m_freem(mb_copy); 1569 return; 1570 } 1571 i = MHLEN - M_LEADINGSPACE(mb_copy); 1572 if (i > len) 1573 i = len; 1574 mb_copy = m_pullup(mb_copy, i); 1575 if (mb_copy == NULL) 1576 return; 1577 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1578 1579 /* 1580 * fill in the encapsulating IP header. 1581 */ 1582 ip_copy = mtod(mb_copy, struct ip *); 1583 *ip_copy = multicast_encap_iphdr; 1584 ip_copy->ip_id = htons(ip_id++); 1585 ip_copy->ip_len += len; 1586 ip_copy->ip_src = vifp->v_lcl_addr; 1587 ip_copy->ip_dst = vifp->v_rmt_addr; 1588 1589 /* 1590 * turn the encapsulated IP header back into a valid one. 1591 */ 1592 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1593 --ip->ip_ttl; 1594 HTONS(ip->ip_len); 1595 HTONS(ip->ip_off); 1596 ip->ip_sum = 0; 1597 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1598 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1599 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1600 1601 if (vifp->v_rate_limit == 0) 1602 tbf_send_packet(vifp, mb_copy); 1603 else 1604 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1605} 1606 1607/* 1608 * De-encapsulate a packet and feed it back through ip input (this 1609 * routine is called whenever IP gets a packet with proto type 1610 * ENCAP_PROTO and a local destination address). 1611 */ 1612void 1613#ifdef MROUTE_LKM 1614X_ipip_input(m, off, proto) 1615#else 1616ipip_input(m, off, proto) 1617#endif 1618 register struct mbuf *m; 1619 int off; 1620 int proto; 1621{ 1622 struct ifnet *ifp = m->m_pkthdr.rcvif; 1623 register struct ip *ip = mtod(m, struct ip *); 1624 register int hlen = ip->ip_hl << 2; 1625 register int s; 1626 register struct ifqueue *ifq; 1627 register struct vif *vifp; 1628 1629 if (!have_encap_tunnel) { 1630 rip_input(m, off, proto); 1631 return; 1632 } 1633 /* 1634 * dump the packet if it's not to a multicast destination or if 1635 * we don't have an encapsulating tunnel with the source. 1636 * Note: This code assumes that the remote site IP address 1637 * uniquely identifies the tunnel (i.e., that this site has 1638 * at most one tunnel with the remote site). 1639 */ 1640 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) { 1641 ++mrtstat.mrts_bad_tunnel; 1642 m_freem(m); 1643 return; 1644 } 1645 if (ip->ip_src.s_addr != last_encap_src) { 1646 register struct vif *vife; 1647 1648 vifp = viftable; 1649 vife = vifp + numvifs; 1650 last_encap_src = ip->ip_src.s_addr; 1651 last_encap_vif = 0; 1652 for ( ; vifp < vife; ++vifp) 1653 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1654 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1655 == VIFF_TUNNEL) 1656 last_encap_vif = vifp; 1657 break; 1658 } 1659 } 1660 if ((vifp = last_encap_vif) == 0) { 1661 last_encap_src = 0; 1662 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1663 m_freem(m); 1664 if (mrtdebug) 1665 log(LOG_DEBUG, "ip_mforward: no tunnel with %lx\n", 1666 (u_long)ntohl(ip->ip_src.s_addr)); 1667 return; 1668 } 1669 ifp = vifp->v_ifp; 1670 1671 if (hlen > IP_HDR_LEN) 1672 ip_stripoptions(m, (struct mbuf *) 0); 1673 m->m_data += IP_HDR_LEN; 1674 m->m_len -= IP_HDR_LEN; 1675 m->m_pkthdr.len -= IP_HDR_LEN; 1676 m->m_pkthdr.rcvif = ifp; 1677 1678 ifq = &ipintrq; 1679 s = splimp(); 1680 if (IF_QFULL(ifq)) { 1681 IF_DROP(ifq); 1682 m_freem(m); 1683 } else { 1684 IF_ENQUEUE(ifq, m); 1685 /* 1686 * normally we would need a "schednetisr(NETISR_IP)" 1687 * here but we were called by ip_input and it is going 1688 * to loop back & try to dequeue the packet we just 1689 * queued as soon as we return so we avoid the 1690 * unnecessary software interrrupt. 1691 */ 1692 } 1693 splx(s); 1694} 1695 1696/* 1697 * Token bucket filter module 1698 */ 1699 1700static void 1701tbf_control(vifp, m, ip, p_len) 1702 register struct vif *vifp; 1703 register struct mbuf *m; 1704 register struct ip *ip; 1705 register u_long p_len; 1706{ 1707 register struct tbf *t = vifp->v_tbf; 1708 1709 if (p_len > MAX_BKT_SIZE) { 1710 /* drop if packet is too large */ 1711 mrtstat.mrts_pkt2large++; 1712 m_freem(m); 1713 return; 1714 } 1715 1716 tbf_update_tokens(vifp); 1717 1718 /* if there are enough tokens, 1719 * and the queue is empty, 1720 * send this packet out 1721 */ 1722 1723 if (t->tbf_q_len == 0) { 1724 /* queue empty, send packet if enough tokens */ 1725 if (p_len <= t->tbf_n_tok) { 1726 t->tbf_n_tok -= p_len; 1727 tbf_send_packet(vifp, m); 1728 } else { 1729 /* queue packet and timeout till later */ 1730 tbf_queue(vifp, m); 1731 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1732 } 1733 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1734 /* finite queue length, so queue pkts and process queue */ 1735 tbf_queue(vifp, m); 1736 tbf_process_q(vifp); 1737 } else { 1738 /* queue length too much, try to dq and queue and process */ 1739 if (!tbf_dq_sel(vifp, ip)) { 1740 mrtstat.mrts_q_overflow++; 1741 m_freem(m); 1742 return; 1743 } else { 1744 tbf_queue(vifp, m); 1745 tbf_process_q(vifp); 1746 } 1747 } 1748 return; 1749} 1750 1751/* 1752 * adds a packet to the queue at the interface 1753 */ 1754static void 1755tbf_queue(vifp, m) 1756 register struct vif *vifp; 1757 register struct mbuf *m; 1758{ 1759 register int s = splnet(); 1760 register struct tbf *t = vifp->v_tbf; 1761 1762 if (t->tbf_t == NULL) { 1763 /* Queue was empty */ 1764 t->tbf_q = m; 1765 } else { 1766 /* Insert at tail */ 1767 t->tbf_t->m_act = m; 1768 } 1769 1770 /* Set new tail pointer */ 1771 t->tbf_t = m; 1772 1773#ifdef DIAGNOSTIC 1774 /* Make sure we didn't get fed a bogus mbuf */ 1775 if (m->m_act) 1776 panic("tbf_queue: m_act"); 1777#endif 1778 m->m_act = NULL; 1779 1780 t->tbf_q_len++; 1781 1782 splx(s); 1783} 1784 1785 1786/* 1787 * processes the queue at the interface 1788 */ 1789static void 1790tbf_process_q(vifp) 1791 register struct vif *vifp; 1792{ 1793 register struct mbuf *m; 1794 register int len; 1795 register int s = splnet(); 1796 register struct tbf *t = vifp->v_tbf; 1797 1798 /* loop through the queue at the interface and send as many packets 1799 * as possible 1800 */ 1801 while (t->tbf_q_len > 0) { 1802 m = t->tbf_q; 1803 1804 len = mtod(m, struct ip *)->ip_len; 1805 1806 /* determine if the packet can be sent */ 1807 if (len <= t->tbf_n_tok) { 1808 /* if so, 1809 * reduce no of tokens, dequeue the packet, 1810 * send the packet. 1811 */ 1812 t->tbf_n_tok -= len; 1813 1814 t->tbf_q = m->m_act; 1815 if (--t->tbf_q_len == 0) 1816 t->tbf_t = NULL; 1817 1818 m->m_act = NULL; 1819 tbf_send_packet(vifp, m); 1820 1821 } else break; 1822 } 1823 splx(s); 1824} 1825 1826static void 1827tbf_reprocess_q(xvifp) 1828 void *xvifp; 1829{ 1830 register struct vif *vifp = xvifp; 1831 if (ip_mrouter == NULL) 1832 return; 1833 1834 tbf_update_tokens(vifp); 1835 1836 tbf_process_q(vifp); 1837 1838 if (vifp->v_tbf->tbf_q_len) 1839 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1840} 1841 1842/* function that will selectively discard a member of the queue 1843 * based on the precedence value and the priority 1844 */ 1845static int 1846tbf_dq_sel(vifp, ip) 1847 register struct vif *vifp; 1848 register struct ip *ip; 1849{ 1850 register int s = splnet(); 1851 register u_int p; 1852 register struct mbuf *m, *last; 1853 register struct mbuf **np; 1854 register struct tbf *t = vifp->v_tbf; 1855 1856 p = priority(vifp, ip); 1857 1858 np = &t->tbf_q; 1859 last = NULL; 1860 while ((m = *np) != NULL) { 1861 if (p > priority(vifp, mtod(m, struct ip *))) { 1862 *np = m->m_act; 1863 /* If we're removing the last packet, fix the tail pointer */ 1864 if (m == t->tbf_t) 1865 t->tbf_t = last; 1866 m_freem(m); 1867 /* it's impossible for the queue to be empty, but 1868 * we check anyway. */ 1869 if (--t->tbf_q_len == 0) 1870 t->tbf_t = NULL; 1871 splx(s); 1872 mrtstat.mrts_drop_sel++; 1873 return(1); 1874 } 1875 np = &m->m_act; 1876 last = m; 1877 } 1878 splx(s); 1879 return(0); 1880} 1881 1882static void 1883tbf_send_packet(vifp, m) 1884 register struct vif *vifp; 1885 register struct mbuf *m; 1886{ 1887 struct ip_moptions imo; 1888 int error; 1889 static struct route ro; 1890 int s = splnet(); 1891 1892 if (vifp->v_flags & VIFF_TUNNEL) { 1893 /* If tunnel options */ 1894 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1895 IP_FORWARDING, (struct ip_moptions *)0); 1896 } else { 1897 imo.imo_multicast_ifp = vifp->v_ifp; 1898 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1899 imo.imo_multicast_loop = 1; 1900 imo.imo_multicast_vif = -1; 1901 1902 /* 1903 * Re-entrancy should not be a problem here, because 1904 * the packets that we send out and are looped back at us 1905 * should get rejected because they appear to come from 1906 * the loopback interface, thus preventing looping. 1907 */ 1908 error = ip_output(m, (struct mbuf *)0, &ro, 1909 IP_FORWARDING, &imo); 1910 1911 if (mrtdebug & DEBUG_XMIT) 1912 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1913 vifp - viftable, error); 1914 } 1915 splx(s); 1916} 1917 1918/* determine the current time and then 1919 * the elapsed time (between the last time and time now) 1920 * in milliseconds & update the no. of tokens in the bucket 1921 */ 1922static void 1923tbf_update_tokens(vifp) 1924 register struct vif *vifp; 1925{ 1926 struct timeval tp; 1927 register u_long tm; 1928 register int s = splnet(); 1929 register struct tbf *t = vifp->v_tbf; 1930 1931 GET_TIME(tp); 1932 1933 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1934 1935 /* 1936 * This formula is actually 1937 * "time in seconds" * "bytes/second". 1938 * 1939 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1940 * 1941 * The (1000/1024) was introduced in add_vif to optimize 1942 * this divide into a shift. 1943 */ 1944 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1945 t->tbf_last_pkt_t = tp; 1946 1947 if (t->tbf_n_tok > MAX_BKT_SIZE) 1948 t->tbf_n_tok = MAX_BKT_SIZE; 1949 1950 splx(s); 1951} 1952 1953static int 1954priority(vifp, ip) 1955 register struct vif *vifp; 1956 register struct ip *ip; 1957{ 1958 register int prio; 1959 1960 /* temporary hack; may add general packet classifier some day */ 1961 1962 /* 1963 * The UDP port space is divided up into four priority ranges: 1964 * [0, 16384) : unclassified - lowest priority 1965 * [16384, 32768) : audio - highest priority 1966 * [32768, 49152) : whiteboard - medium priority 1967 * [49152, 65536) : video - low priority 1968 */ 1969 if (ip->ip_p == IPPROTO_UDP) { 1970 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1971 switch (ntohs(udp->uh_dport) & 0xc000) { 1972 case 0x4000: 1973 prio = 70; 1974 break; 1975 case 0x8000: 1976 prio = 60; 1977 break; 1978 case 0xc000: 1979 prio = 55; 1980 break; 1981 default: 1982 prio = 50; 1983 break; 1984 } 1985 if (tbfdebug > 1) 1986 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio); 1987 } else { 1988 prio = 50; 1989 } 1990 return prio; 1991} 1992 1993/* 1994 * End of token bucket filter modifications 1995 */ 1996 1997int 1998ip_rsvp_vif_init(so, sopt) 1999 struct socket *so; 2000 struct sockopt *sopt; 2001{ 2002 int error, i, s; 2003 2004 if (rsvpdebug) 2005 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2006 so->so_type, so->so_proto->pr_protocol); 2007 2008 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2009 return EOPNOTSUPP; 2010 2011 /* Check mbuf. */ 2012 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 2013 if (error) 2014 return (error); 2015 2016 if (rsvpdebug) 2017 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", i, rsvp_on); 2018 2019 s = splnet(); 2020 2021 /* Check vif. */ 2022 if (!legal_vif_num(i)) { 2023 splx(s); 2024 return EADDRNOTAVAIL; 2025 } 2026 2027 /* Check if socket is available. */ 2028 if (viftable[i].v_rsvpd != NULL) { 2029 splx(s); 2030 return EADDRINUSE; 2031 } 2032 2033 viftable[i].v_rsvpd = so; 2034 /* This may seem silly, but we need to be sure we don't over-increment 2035 * the RSVP counter, in case something slips up. 2036 */ 2037 if (!viftable[i].v_rsvp_on) { 2038 viftable[i].v_rsvp_on = 1; 2039 rsvp_on++; 2040 } 2041 2042 splx(s); 2043 return 0; 2044} 2045 2046int 2047ip_rsvp_vif_done(so, sopt) 2048 struct socket *so; 2049 struct sockopt *sopt; 2050{ 2051 int error, i, s; 2052 2053 if (rsvpdebug) 2054 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2055 so->so_type, so->so_proto->pr_protocol); 2056 2057 if (so->so_type != SOCK_RAW || 2058 so->so_proto->pr_protocol != IPPROTO_RSVP) 2059 return EOPNOTSUPP; 2060 2061 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 2062 if (error) 2063 return (error); 2064 2065 s = splnet(); 2066 2067 /* Check vif. */ 2068 if (!legal_vif_num(i)) { 2069 splx(s); 2070 return EADDRNOTAVAIL; 2071 } 2072 2073 if (rsvpdebug) 2074 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n", 2075 viftable[i].v_rsvpd, so); 2076 2077 viftable[i].v_rsvpd = NULL; 2078 /* 2079 * This may seem silly, but we need to be sure we don't over-decrement 2080 * the RSVP counter, in case something slips up. 2081 */ 2082 if (viftable[i].v_rsvp_on) { 2083 viftable[i].v_rsvp_on = 0; 2084 rsvp_on--; 2085 } 2086 2087 splx(s); 2088 return 0; 2089} 2090 2091void 2092ip_rsvp_force_done(so) 2093 struct socket *so; 2094{ 2095 int vifi; 2096 register int s; 2097 2098 /* Don't bother if it is not the right type of socket. */ 2099 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2100 return; 2101 2102 s = splnet(); 2103 2104 /* The socket may be attached to more than one vif...this 2105 * is perfectly legal. 2106 */ 2107 for (vifi = 0; vifi < numvifs; vifi++) { 2108 if (viftable[vifi].v_rsvpd == so) { 2109 viftable[vifi].v_rsvpd = NULL; 2110 /* This may seem silly, but we need to be sure we don't 2111 * over-decrement the RSVP counter, in case something slips up. 2112 */ 2113 if (viftable[vifi].v_rsvp_on) { 2114 viftable[vifi].v_rsvp_on = 0; 2115 rsvp_on--; 2116 } 2117 } 2118 } 2119 2120 splx(s); 2121 return; 2122} 2123 2124void 2125rsvp_input(m, off, proto) 2126 struct mbuf *m; 2127 int off; 2128 int proto; 2129{ 2130 int vifi; 2131 register struct ip *ip = mtod(m, struct ip *); 2132 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2133 register int s; 2134 struct ifnet *ifp; 2135 2136 if (rsvpdebug) 2137 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2138 2139 /* Can still get packets with rsvp_on = 0 if there is a local member 2140 * of the group to which the RSVP packet is addressed. But in this 2141 * case we want to throw the packet away. 2142 */ 2143 if (!rsvp_on) { 2144 m_freem(m); 2145 return; 2146 } 2147 2148 s = splnet(); 2149 2150 if (rsvpdebug) 2151 printf("rsvp_input: check vifs\n"); 2152 2153#ifdef DIAGNOSTIC 2154 if (!(m->m_flags & M_PKTHDR)) 2155 panic("rsvp_input no hdr"); 2156#endif 2157 2158 ifp = m->m_pkthdr.rcvif; 2159 /* Find which vif the packet arrived on. */ 2160 for (vifi = 0; vifi < numvifs; vifi++) 2161 if (viftable[vifi].v_ifp == ifp) 2162 break; 2163 2164 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 2165 /* 2166 * If the old-style non-vif-associated socket is set, 2167 * then use it. Otherwise, drop packet since there 2168 * is no specific socket for this vif. 2169 */ 2170 if (ip_rsvpd != NULL) { 2171 if (rsvpdebug) 2172 printf("rsvp_input: Sending packet up old-style socket\n"); 2173 rip_input(m, off, proto); /* xxx */ 2174 } else { 2175 if (rsvpdebug && vifi == numvifs) 2176 printf("rsvp_input: Can't find vif for packet.\n"); 2177 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 2178 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2179 m_freem(m); 2180 } 2181 splx(s); 2182 return; 2183 } 2184 rsvp_src.sin_addr = ip->ip_src; 2185 2186 if (rsvpdebug && m) 2187 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2188 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2189 2190 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 2191 if (rsvpdebug) 2192 printf("rsvp_input: Failed to append to socket\n"); 2193 } else { 2194 if (rsvpdebug) 2195 printf("rsvp_input: send packet up\n"); 2196 } 2197 2198 splx(s); 2199} 2200 2201#ifdef MROUTE_LKM 2202#include <sys/conf.h> 2203#include <sys/exec.h> 2204#include <sys/sysent.h> 2205#include <sys/lkm.h> 2206 2207MOD_MISC("ip_mroute_mod") 2208 2209static int 2210ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd) 2211{ 2212 int i; 2213 struct lkm_misc *args = lkmtp->private.lkm_misc; 2214 int err = 0; 2215 2216 switch(cmd) { 2217 static int (*old_ip_mrouter_cmd)(); 2218 static int (*old_ip_mrouter_done)(); 2219 static int (*old_ip_mforward)(); 2220 static int (*old_mrt_ioctl)(); 2221 static void (*old_proto4_input)(); 2222 static int (*old_legal_vif_num)(); 2223 extern struct protosw inetsw[]; 2224 2225 case LKM_E_LOAD: 2226 if(lkmexists(lkmtp) || ip_mrtproto) 2227 return(EEXIST); 2228 old_ip_mrouter_cmd = ip_mrouter_cmd; 2229 ip_mrouter_cmd = X_ip_mrouter_cmd; 2230 old_ip_mrouter_done = ip_mrouter_done; 2231 ip_mrouter_done = X_ip_mrouter_done; 2232 old_ip_mforward = ip_mforward; 2233 ip_mforward = X_ip_mforward; 2234 old_mrt_ioctl = mrt_ioctl; 2235 mrt_ioctl = X_mrt_ioctl; 2236 old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input; 2237 inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input; 2238 old_legal_vif_num = legal_vif_num; 2239 legal_vif_num = X_legal_vif_num; 2240 ip_mrtproto = IGMP_DVMRP; 2241 2242 printf("\nIP multicast routing loaded\n"); 2243 break; 2244 2245 case LKM_E_UNLOAD: 2246 if (ip_mrouter) 2247 return EINVAL; 2248 2249 ip_mrouter_cmd = old_ip_mrouter_cmd; 2250 ip_mrouter_done = old_ip_mrouter_done; 2251 ip_mforward = old_ip_mforward; 2252 mrt_ioctl = old_mrt_ioctl; 2253 inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input; 2254 legal_vif_num = old_legal_vif_num; 2255 ip_mrtproto = 0; 2256 break; 2257 2258 default: 2259 err = EINVAL; 2260 break; 2261 } 2262 2263 return(err); 2264} 2265 2266int 2267ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) { 2268 DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle, 2269 nosys); 2270} 2271 2272#endif /* MROUTE_LKM */ 2273#endif /* MROUTING */ 2274