ip_mroute.c revision 42572
1174993Srafan/* 2176187Srafan * IP multicast forwarding procedures 3174993Srafan * 4174993Srafan * Written by David Waitzman, BBN Labs, August 1988. 5174993Srafan * Modified by Steve Deering, Stanford, February 1989. 6174993Srafan * Modified by Mark J. Steiglitz, Stanford, May, 1991 7174993Srafan * Modified by Van Jacobson, LBL, January 1993 8174993Srafan * Modified by Ajit Thyagarajan, PARC, August 1993 9174993Srafan * Modified by Bill Fenner, PARC, April 1995 10174993Srafan * 11174993Srafan * MROUTING Revision: 3.5 12174993Srafan * $Id: ip_mroute.c,v 1.51 1998/12/16 18:07:11 fenner Exp $ 13174993Srafan */ 14174993Srafan 15174993Srafan#include "opt_mrouting.h" 16174993Srafan 17174993Srafan#include <sys/param.h> 18174993Srafan#include <sys/systm.h> 19174993Srafan#include <sys/mbuf.h> 20174993Srafan#include <sys/socket.h> 21174993Srafan#include <sys/socketvar.h> 22174993Srafan#include <sys/protosw.h> 23174993Srafan#include <sys/time.h> 24174993Srafan#include <sys/kernel.h> 25174993Srafan#include <sys/sockio.h> 26174993Srafan#include <sys/syslog.h> 27174993Srafan#include <net/if.h> 28174993Srafan#include <net/route.h> 29174993Srafan#include <netinet/in.h> 30174993Srafan#include <netinet/in_systm.h> 31174993Srafan#include <netinet/ip.h> 32174993Srafan#include <netinet/ip_var.h> 33174993Srafan#include <netinet/in_var.h> 34174993Srafan#include <netinet/igmp.h> 35184989Srafan#include <netinet/ip_mroute.h> 36174993Srafan#include <netinet/udp.h> 37174993Srafan 38178866Srafan#ifndef NTOHL 39174993Srafan#if BYTE_ORDER != BIG_ENDIAN 40174993Srafan#define NTOHL(d) ((d) = ntohl((d))) 41174993Srafan#define NTOHS(d) ((d) = ntohs((u_short)(d))) 42174993Srafan#define HTONL(d) ((d) = htonl((d))) 43184989Srafan#define HTONS(d) ((d) = htons((u_short)(d))) 44174993Srafan#else 45184989Srafan#define NTOHL(d) 46174993Srafan#define NTOHS(d) 47174993Srafan#define HTONL(d) 48174993Srafan#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; 66u_int rsvpdebug = 0; 67 68int 69_ip_mrouter_set(so, sopt) 70 struct socket *so; 71 struct sockopt *sopt; 72{ 73 return(EOPNOTSUPP); 74} 75 76int (*ip_mrouter_set)(struct socket *, struct sockopt *) = _ip_mrouter_set; 77 78 79int 80_ip_mrouter_get(so, sopt) 81 struct socket *so; 82 struct sockopt *sopt; 83{ 84 return(EOPNOTSUPP); 85} 86 87int (*ip_mrouter_get)(struct socket *, struct sockopt *) = _ip_mrouter_get; 88 89int 90_ip_mrouter_done() 91{ 92 return(0); 93} 94 95int (*ip_mrouter_done)(void) = _ip_mrouter_done; 96 97int 98_ip_mforward(ip, ifp, m, imo) 99 struct ip *ip; 100 struct ifnet *ifp; 101 struct mbuf *m; 102 struct ip_moptions *imo; 103{ 104 return(0); 105} 106 107int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 108 struct ip_moptions *) = _ip_mforward; 109 110int 111_mrt_ioctl(int req, caddr_t data, struct proc *p) 112{ 113 return EOPNOTSUPP; 114} 115 116int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl; 117 118void 119rsvp_input(m, iphlen) /* XXX must fixup manually */ 120 struct mbuf *m; 121 int iphlen; 122{ 123 /* Can still get packets with rsvp_on = 0 if there is a local member 124 * of the group to which the RSVP packet is addressed. But in this 125 * case we want to throw the packet away. 126 */ 127 if (!rsvp_on) { 128 m_freem(m); 129 return; 130 } 131 132 if (ip_rsvpd != NULL) { 133 if (rsvpdebug) 134 printf("rsvp_input: Sending packet up old-style socket\n"); 135 rip_input(m, iphlen); 136 return; 137 } 138 /* Drop the packet */ 139 m_freem(m); 140} 141 142void ipip_input(struct mbuf *m, int iphlen) { /* XXX must fixup manually */ 143 rip_input(m, iphlen); 144} 145 146int (*legal_vif_num)(int) = 0; 147 148/* 149 * This should never be called, since IP_MULTICAST_VIF should fail, but 150 * just in case it does get called, the code a little lower in ip_output 151 * will assign the packet a local address. 152 */ 153u_long 154_ip_mcast_src(int vifi) { return INADDR_ANY; } 155u_long (*ip_mcast_src)(int) = _ip_mcast_src; 156 157int 158ip_rsvp_vif_init(so, sopt) 159 struct socket *so; 160 struct sockopt *sopt; 161{ 162 return(EINVAL); 163} 164 165int 166ip_rsvp_vif_done(so, sopt) 167 struct socket *so; 168 struct sockopt *sopt; 169{ 170 return(EINVAL); 171} 172 173void 174ip_rsvp_force_done(so) 175 struct socket *so; 176{ 177 return; 178} 179 180#else /* MROUTING */ 181 182#define M_HASCL(m) ((m)->m_flags & M_EXT) 183 184#define INSIZ sizeof(struct in_addr) 185#define same(a1, a2) \ 186 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0) 187 188#define MT_MRTABLE MT_RTABLE /* since nothing else uses it */ 189 190/* 191 * Globals. All but ip_mrouter and ip_mrtproto could be static, 192 * except for netstat or debugging purposes. 193 */ 194#ifndef MROUTE_LKM 195struct socket *ip_mrouter = NULL; 196static struct mrtstat mrtstat; 197#else /* MROUTE_LKM */ 198extern void X_ipip_input __P((struct mbuf *m, int iphlen)); 199extern struct mrtstat mrtstat; 200static int ip_mrtproto; 201#endif 202 203#define NO_RTE_FOUND 0x1 204#define RTE_FOUND 0x2 205 206static struct mbuf *mfctable[MFCTBLSIZ]; 207static u_char nexpire[MFCTBLSIZ]; 208static struct vif viftable[MAXVIFS]; 209static u_int mrtdebug = 0; /* debug level */ 210#define DEBUG_MFC 0x02 211#define DEBUG_FORWARD 0x04 212#define DEBUG_EXPIRE 0x08 213#define DEBUG_XMIT 0x10 214static u_int tbfdebug = 0; /* tbf debug level */ 215static u_int rsvpdebug = 0; /* rsvp debug level */ 216 217static struct callout_handle expire_upcalls_ch; 218 219#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 220#define UPCALL_EXPIRE 6 /* number of timeouts */ 221 222/* 223 * Define the token bucket filter structures 224 * tbftable -> each vif has one of these for storing info 225 */ 226 227static struct tbf tbftable[MAXVIFS]; 228#define TBF_REPROCESS (hz / 100) /* 100x / second */ 229 230/* 231 * 'Interfaces' associated with decapsulator (so we can tell 232 * packets that went through it from ones that get reflected 233 * by a broken gateway). These interfaces are never linked into 234 * the system ifnet list & no routes point to them. I.e., packets 235 * can't be sent this way. They only exist as a placeholder for 236 * multicast source verification. 237 */ 238static struct ifnet multicast_decap_if[MAXVIFS]; 239 240#define ENCAP_TTL 64 241#define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 242 243/* prototype IP hdr for encapsulated packets */ 244static struct ip multicast_encap_iphdr = { 245#if BYTE_ORDER == LITTLE_ENDIAN 246 sizeof(struct ip) >> 2, IPVERSION, 247#else 248 IPVERSION, sizeof(struct ip) >> 2, 249#endif 250 0, /* tos */ 251 sizeof(struct ip), /* total length */ 252 0, /* id */ 253 0, /* frag offset */ 254 ENCAP_TTL, ENCAP_PROTO, 255 0, /* checksum */ 256}; 257 258/* 259 * Private variables. 260 */ 261static vifi_t numvifs = 0; 262static int have_encap_tunnel = 0; 263 264/* 265 * one-back cache used by ipip_input to locate a tunnel's vif 266 * given a datagram's src ip address. 267 */ 268static u_long last_encap_src; 269static struct vif *last_encap_vif; 270 271static u_long X_ip_mcast_src __P((int vifi)); 272static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo)); 273static int X_ip_mrouter_done __P((void)); 274static int X_ip_mrouter_get __P((struct socket *so, struct sockopt *m)); 275static int X_ip_mrouter_set __P((struct socket *so, struct sockopt *m)); 276static int X_legal_vif_num __P((int vif)); 277static int X_mrt_ioctl __P((int cmd, caddr_t data)); 278 279static int get_sg_cnt(struct sioc_sg_req *); 280static int get_vif_cnt(struct sioc_vif_req *); 281static int ip_mrouter_init(struct socket *, int); 282static int add_vif(struct vifctl *); 283static int del_vif(vifi_t); 284static int add_mfc(struct mfcctl *); 285static int del_mfc(struct mfcctl *); 286static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 287static int set_assert(int); 288static void expire_upcalls(void *); 289static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, 290 vifi_t); 291static void phyint_send(struct ip *, struct vif *, struct mbuf *); 292static void encap_send(struct ip *, struct vif *, struct mbuf *); 293static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 294static void tbf_queue(struct vif *, struct mbuf *); 295static void tbf_process_q(struct vif *); 296static void tbf_reprocess_q(void *); 297static int tbf_dq_sel(struct vif *, struct ip *); 298static void tbf_send_packet(struct vif *, struct mbuf *); 299static void tbf_update_tokens(struct vif *); 300static int priority(struct vif *, struct ip *); 301void multiencap_decap(struct mbuf *); 302 303/* 304 * whether or not special PIM assert processing is enabled. 305 */ 306static int pim_assert; 307/* 308 * Rate limit for assert notification messages, in usec 309 */ 310#define ASSERT_MSG_TIME 3000000 311 312/* 313 * Hash function for a source, group entry 314 */ 315#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 316 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 317 318/* 319 * Find a route for a given origin IP address and Multicast group address 320 * Type of service parameter to be added in the future!!! 321 */ 322 323#define MFCFIND(o, g, rt) { \ 324 register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \ 325 register struct mfc *_rt = NULL; \ 326 rt = NULL; \ 327 ++mrtstat.mrts_mfc_lookups; \ 328 while (_mb_rt) { \ 329 _rt = mtod(_mb_rt, struct mfc *); \ 330 if ((_rt->mfc_origin.s_addr == o) && \ 331 (_rt->mfc_mcastgrp.s_addr == g) && \ 332 (_mb_rt->m_act == NULL)) { \ 333 rt = _rt; \ 334 break; \ 335 } \ 336 _mb_rt = _mb_rt->m_next; \ 337 } \ 338 if (rt == NULL) { \ 339 ++mrtstat.mrts_mfc_misses; \ 340 } \ 341} 342 343 344/* 345 * Macros to compute elapsed time efficiently 346 * Borrowed from Van Jacobson's scheduling code 347 */ 348#define TV_DELTA(a, b, delta) { \ 349 register int xxs; \ 350 \ 351 delta = (a).tv_usec - (b).tv_usec; \ 352 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 353 switch (xxs) { \ 354 case 2: \ 355 delta += 1000000; \ 356 /* fall through */ \ 357 case 1: \ 358 delta += 1000000; \ 359 break; \ 360 default: \ 361 delta += (1000000 * xxs); \ 362 } \ 363 } \ 364} 365 366#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 367 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 368 369#ifdef UPCALL_TIMING 370u_long upcall_data[51]; 371static void collate(struct timeval *); 372#endif /* UPCALL_TIMING */ 373 374 375/* 376 * Handle MRT setsockopt commands to modify the multicast routing tables. 377 */ 378static int 379X_ip_mrouter_set(so, sopt) 380 struct socket *so; 381 struct sockopt *sopt; 382{ 383 int error, optval; 384 vifi_t vifi; 385 struct vifctl vifc; 386 struct mfcctl mfc; 387 388 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 389 return (EPERM); 390 391 error = 0; 392 switch (sopt->sopt_name) { 393 case MRT_INIT: 394 error = sooptcopyin(sopt, &optval, sizeof optval, 395 sizeof optval); 396 if (error) 397 break; 398 error = ip_mrouter_init(so, optval); 399 break; 400 401 case MRT_DONE: 402 error = ip_mrouter_done(); 403 break; 404 405 case MRT_ADD_VIF: 406 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 407 if (error) 408 break; 409 error = add_vif(&vifc); 410 break; 411 412 case MRT_DEL_VIF: 413 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 414 if (error) 415 break; 416 error = del_vif(vifi); 417 break; 418 419 case MRT_ADD_MFC: 420 case MRT_DEL_MFC: 421 error = sooptcopyin(sopt, &mfc, sizeof mfc, sizeof mfc); 422 if (error) 423 break; 424 if (sopt->sopt_name == MRT_ADD_MFC) 425 error = add_mfc(&mfc); 426 else 427 error = del_mfc(&mfc); 428 break; 429 430 case MRT_ASSERT: 431 error = sooptcopyin(sopt, &optval, sizeof optval, 432 sizeof optval); 433 if (error) 434 break; 435 set_assert(optval); 436 break; 437 438 default: 439 error = EOPNOTSUPP; 440 break; 441 } 442 return (error); 443} 444 445#ifndef MROUTE_LKM 446int (*ip_mrouter_set)(struct socket *, struct sockopt *) = X_ip_mrouter_set; 447#endif 448 449/* 450 * Handle MRT getsockopt commands 451 */ 452static int 453X_ip_mrouter_get(so, sopt) 454 struct socket *so; 455 struct sockopt *sopt; 456{ 457 int error; 458 static int version = 0x0305; /* !!! why is this here? XXX */ 459 460 switch (sopt->sopt_name) { 461 case MRT_VERSION: 462 error = sooptcopyout(sopt, &version, sizeof version); 463 break; 464 465 case MRT_ASSERT: 466 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 467 break; 468 default: 469 error = EOPNOTSUPP; 470 break; 471 } 472 return (error); 473} 474 475#ifndef MROUTE_LKM 476int (*ip_mrouter_get)(struct socket *, struct sockopt *) = X_ip_mrouter_get; 477#endif 478 479/* 480 * Handle ioctl commands to obtain information from the cache 481 */ 482static int 483X_mrt_ioctl(cmd, data) 484 int cmd; 485 caddr_t data; 486{ 487 int error = 0; 488 489 switch (cmd) { 490 case (SIOCGETVIFCNT): 491 return (get_vif_cnt((struct sioc_vif_req *)data)); 492 break; 493 case (SIOCGETSGCNT): 494 return (get_sg_cnt((struct sioc_sg_req *)data)); 495 break; 496 default: 497 return (EINVAL); 498 break; 499 } 500 return error; 501} 502 503#ifndef MROUTE_LKM 504int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl; 505#endif 506 507/* 508 * returns the packet, byte, rpf-failure count for the source group provided 509 */ 510static int 511get_sg_cnt(req) 512 register struct sioc_sg_req *req; 513{ 514 register struct mfc *rt; 515 int s; 516 517 s = splnet(); 518 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 519 splx(s); 520 if (rt != NULL) { 521 req->pktcnt = rt->mfc_pkt_cnt; 522 req->bytecnt = rt->mfc_byte_cnt; 523 req->wrong_if = rt->mfc_wrong_if; 524 } else 525 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 526 527 return 0; 528} 529 530/* 531 * returns the input and output packet and byte counts on the vif provided 532 */ 533static int 534get_vif_cnt(req) 535 register struct sioc_vif_req *req; 536{ 537 register vifi_t vifi = req->vifi; 538 539 if (vifi >= numvifs) return EINVAL; 540 541 req->icount = viftable[vifi].v_pkt_in; 542 req->ocount = viftable[vifi].v_pkt_out; 543 req->ibytes = viftable[vifi].v_bytes_in; 544 req->obytes = viftable[vifi].v_bytes_out; 545 546 return 0; 547} 548 549/* 550 * Enable multicast routing 551 */ 552static int 553ip_mrouter_init(so, version) 554 struct socket *so; 555 int version; 556{ 557 if (mrtdebug) 558 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 559 so->so_type, so->so_proto->pr_protocol); 560 561 if (so->so_type != SOCK_RAW || 562 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP; 563 564 if (version != 1) 565 return ENOPROTOOPT; 566 567 if (ip_mrouter != NULL) return EADDRINUSE; 568 569 ip_mrouter = so; 570 571 bzero((caddr_t)mfctable, sizeof(mfctable)); 572 bzero((caddr_t)nexpire, sizeof(nexpire)); 573 574 pim_assert = 0; 575 576 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 577 578 if (mrtdebug) 579 log(LOG_DEBUG, "ip_mrouter_init\n"); 580 581 return 0; 582} 583 584/* 585 * Disable multicast routing 586 */ 587static int 588X_ip_mrouter_done() 589{ 590 vifi_t vifi; 591 int i; 592 struct ifnet *ifp; 593 struct ifreq ifr; 594 struct mbuf *mb_rt; 595 struct mbuf *m; 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 mb_rt = mfctable[i]; 627 while (mb_rt) { 628 if (mb_rt->m_act != NULL) { 629 while (mb_rt->m_act) { 630 m = mb_rt->m_act; 631 mb_rt->m_act = m->m_act; 632 rte = mtod(m, struct rtdetq *); 633 m_freem(rte->m); 634 m_free(m); 635 } 636 } 637 mb_rt = m_free(mb_rt); 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 register struct mbuf *mb_rt; 844 u_long hash; 845 struct mbuf *mb_ntry; 846 struct rtdetq *rte; 847 register u_short nstl; 848 int s; 849 int i; 850 851 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 852 853 /* If an entry already exists, just update the fields */ 854 if (rt) { 855 if (mrtdebug & DEBUG_MFC) 856 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 857 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 858 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 859 mfccp->mfcc_parent); 860 861 s = splnet(); 862 rt->mfc_parent = mfccp->mfcc_parent; 863 for (i = 0; i < numvifs; i++) 864 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 865 splx(s); 866 return 0; 867 } 868 869 /* 870 * Find the entry for which the upcall was made and update 871 */ 872 s = splnet(); 873 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 874 for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) { 875 876 rt = mtod(mb_rt, struct mfc *); 877 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 878 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 879 (mb_rt->m_act != NULL)) { 880 881 if (nstl++) 882 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 883 "multiple kernel entries", 884 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 885 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 886 mfccp->mfcc_parent, (void *)mb_rt->m_act); 887 888 if (mrtdebug & DEBUG_MFC) 889 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 890 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 891 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 892 mfccp->mfcc_parent, (void *)mb_rt->m_act); 893 894 rt->mfc_origin = mfccp->mfcc_origin; 895 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 896 rt->mfc_parent = mfccp->mfcc_parent; 897 for (i = 0; i < numvifs; i++) 898 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 899 /* initialize pkt counters per src-grp */ 900 rt->mfc_pkt_cnt = 0; 901 rt->mfc_byte_cnt = 0; 902 rt->mfc_wrong_if = 0; 903 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 904 905 rt->mfc_expire = 0; /* Don't clean this guy up */ 906 nexpire[hash]--; 907 908 /* free packets Qed at the end of this entry */ 909 while (mb_rt->m_act) { 910 mb_ntry = mb_rt->m_act; 911 rte = mtod(mb_ntry, struct rtdetq *); 912/* #ifdef RSVP_ISI */ 913 ip_mdq(rte->m, rte->ifp, rt, -1); 914/* #endif */ 915 mb_rt->m_act = mb_ntry->m_act; 916 m_freem(rte->m); 917#ifdef UPCALL_TIMING 918 collate(&(rte->t)); 919#endif /* UPCALL_TIMING */ 920 m_free(mb_ntry); 921 } 922 } 923 } 924 925 /* 926 * It is possible that an entry is being inserted without an upcall 927 */ 928 if (nstl == 0) { 929 if (mrtdebug & DEBUG_MFC) 930 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 931 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 932 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 933 mfccp->mfcc_parent); 934 935 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 936 937 rt = mtod(mb_rt, struct mfc *); 938 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 939 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 940 941 rt->mfc_origin = mfccp->mfcc_origin; 942 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 943 rt->mfc_parent = mfccp->mfcc_parent; 944 for (i = 0; i < numvifs; i++) 945 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 946 /* initialize pkt counters per src-grp */ 947 rt->mfc_pkt_cnt = 0; 948 rt->mfc_byte_cnt = 0; 949 rt->mfc_wrong_if = 0; 950 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 951 if (rt->mfc_expire) 952 nexpire[hash]--; 953 rt->mfc_expire = 0; 954 } 955 } 956 if (mb_rt == NULL) { 957 /* no upcall, so make a new entry */ 958 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 959 if (mb_rt == NULL) { 960 splx(s); 961 return ENOBUFS; 962 } 963 964 rt = mtod(mb_rt, struct mfc *); 965 966 /* insert new entry at head of hash chain */ 967 rt->mfc_origin = mfccp->mfcc_origin; 968 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 969 rt->mfc_parent = mfccp->mfcc_parent; 970 for (i = 0; i < numvifs; i++) 971 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 972 /* initialize pkt counters per src-grp */ 973 rt->mfc_pkt_cnt = 0; 974 rt->mfc_byte_cnt = 0; 975 rt->mfc_wrong_if = 0; 976 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 977 rt->mfc_expire = 0; 978 979 /* link into table */ 980 mb_rt->m_next = mfctable[hash]; 981 mfctable[hash] = mb_rt; 982 mb_rt->m_act = NULL; 983 } 984 } 985 splx(s); 986 return 0; 987} 988 989#ifdef UPCALL_TIMING 990/* 991 * collect delay statistics on the upcalls 992 */ 993static void collate(t) 994register struct timeval *t; 995{ 996 register u_long d; 997 register struct timeval tp; 998 register u_long delta; 999 1000 GET_TIME(tp); 1001 1002 if (TV_LT(*t, tp)) 1003 { 1004 TV_DELTA(tp, *t, delta); 1005 1006 d = delta >> 10; 1007 if (d > 50) 1008 d = 50; 1009 1010 ++upcall_data[d]; 1011 } 1012} 1013#endif /* UPCALL_TIMING */ 1014 1015/* 1016 * Delete an mfc entry 1017 */ 1018static int 1019del_mfc(mfccp) 1020 struct mfcctl *mfccp; 1021{ 1022 struct in_addr origin; 1023 struct in_addr mcastgrp; 1024 struct mfc *rt; 1025 struct mbuf *mb_rt; 1026 struct mbuf **nptr; 1027 u_long hash; 1028 int s; 1029 1030 origin = mfccp->mfcc_origin; 1031 mcastgrp = mfccp->mfcc_mcastgrp; 1032 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1033 1034 if (mrtdebug & DEBUG_MFC) 1035 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1036 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1037 1038 s = splnet(); 1039 1040 nptr = &mfctable[hash]; 1041 while ((mb_rt = *nptr) != NULL) { 1042 rt = mtod(mb_rt, struct mfc *); 1043 if (origin.s_addr == rt->mfc_origin.s_addr && 1044 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1045 mb_rt->m_act == NULL) 1046 break; 1047 1048 nptr = &mb_rt->m_next; 1049 } 1050 if (mb_rt == NULL) { 1051 splx(s); 1052 return EADDRNOTAVAIL; 1053 } 1054 1055 MFREE(mb_rt, *nptr); 1056 1057 splx(s); 1058 1059 return 0; 1060} 1061 1062/* 1063 * Send a message to mrouted on the multicast routing socket 1064 */ 1065static int 1066socket_send(s, mm, src) 1067 struct socket *s; 1068 struct mbuf *mm; 1069 struct sockaddr_in *src; 1070{ 1071 if (s) { 1072 if (sbappendaddr(&s->so_rcv, 1073 (struct sockaddr *)src, 1074 mm, (struct mbuf *)0) != 0) { 1075 sorwakeup(s); 1076 return 0; 1077 } 1078 } 1079 m_freem(mm); 1080 return -1; 1081} 1082 1083/* 1084 * IP multicast forwarding function. This function assumes that the packet 1085 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1086 * pointed to by "ifp", and the packet is to be relayed to other networks 1087 * that have members of the packet's destination IP multicast group. 1088 * 1089 * The packet is returned unscathed to the caller, unless it is 1090 * erroneous, in which case a non-zero return value tells the caller to 1091 * discard it. 1092 */ 1093 1094#define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1095#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1096 1097static int 1098X_ip_mforward(ip, ifp, m, imo) 1099 register struct ip *ip; 1100 struct ifnet *ifp; 1101 struct mbuf *m; 1102 struct ip_moptions *imo; 1103{ 1104 register struct mfc *rt; 1105 register u_char *ipoptions; 1106 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1107 static int srctun = 0; 1108 register struct mbuf *mm; 1109 int s; 1110 vifi_t vifi; 1111 struct vif *vifp; 1112 1113 if (mrtdebug & DEBUG_FORWARD) 1114 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1115 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1116 (void *)ifp); 1117 1118 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1119 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1120 /* 1121 * Packet arrived via a physical interface or 1122 * an encapsulated tunnel. 1123 */ 1124 } else { 1125 /* 1126 * Packet arrived through a source-route tunnel. 1127 * Source-route tunnels are no longer supported. 1128 */ 1129 if ((srctun++ % 1000) == 0) 1130 log(LOG_ERR, 1131 "ip_mforward: received source-routed packet from %lx\n", 1132 (u_long)ntohl(ip->ip_src.s_addr)); 1133 1134 return 1; 1135 } 1136 1137 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1138 if (ip->ip_ttl < 255) 1139 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1140 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1141 vifp = viftable + vifi; 1142 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1143 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi, 1144 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1145 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1146 } 1147 return (ip_mdq(m, ifp, NULL, vifi)); 1148 } 1149 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1150 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1151 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr)); 1152 if(!imo) 1153 printf("In fact, no options were specified at all\n"); 1154 } 1155 1156 /* 1157 * Don't forward a packet with time-to-live of zero or one, 1158 * or a packet destined to a local-only group. 1159 */ 1160 if (ip->ip_ttl <= 1 || 1161 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1162 return 0; 1163 1164 /* 1165 * Determine forwarding vifs from the forwarding cache table 1166 */ 1167 s = splnet(); 1168 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt); 1169 1170 /* Entry exists, so forward if necessary */ 1171 if (rt != NULL) { 1172 splx(s); 1173 return (ip_mdq(m, ifp, rt, -1)); 1174 } else { 1175 /* 1176 * If we don't have a route for packet's origin, 1177 * Make a copy of the packet & 1178 * send message to routing daemon 1179 */ 1180 1181 register struct mbuf *mb_rt; 1182 register struct mbuf *mb_ntry; 1183 register struct mbuf *mb0; 1184 register struct rtdetq *rte; 1185 register struct mbuf *rte_m; 1186 register u_long hash; 1187 register int npkts; 1188 int hlen = ip->ip_hl << 2; 1189#ifdef UPCALL_TIMING 1190 struct timeval tp; 1191 1192 GET_TIME(tp); 1193#endif 1194 1195 mrtstat.mrts_no_route++; 1196 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1197 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1198 (u_long)ntohl(ip->ip_src.s_addr), 1199 (u_long)ntohl(ip->ip_dst.s_addr)); 1200 1201 /* 1202 * Allocate mbufs early so that we don't do extra work if we are 1203 * just going to fail anyway. Make sure to pullup the header so 1204 * that other people can't step on it. 1205 */ 1206 MGET(mb_ntry, M_DONTWAIT, MT_DATA); 1207 if (mb_ntry == NULL) { 1208 splx(s); 1209 return ENOBUFS; 1210 } 1211 mb0 = m_copy(m, 0, M_COPYALL); 1212 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1213 mb0 = m_pullup(mb0, hlen); 1214 if (mb0 == NULL) { 1215 m_free(mb_ntry); 1216 splx(s); 1217 return ENOBUFS; 1218 } 1219 1220 /* is there an upcall waiting for this packet? */ 1221 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1222 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 1223 rt = mtod(mb_rt, struct mfc *); 1224 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1225 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1226 (mb_rt->m_act != NULL)) 1227 break; 1228 } 1229 1230 if (mb_rt == NULL) { 1231 int i; 1232 struct igmpmsg *im; 1233 1234 /* no upcall, so make a new entry */ 1235 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 1236 if (mb_rt == NULL) { 1237 m_free(mb_ntry); 1238 m_freem(mb0); 1239 splx(s); 1240 return ENOBUFS; 1241 } 1242 /* Make a copy of the header to send to the user level process */ 1243 mm = m_copy(mb0, 0, hlen); 1244 if (mm == NULL) { 1245 m_free(mb_ntry); 1246 m_freem(mb0); 1247 m_free(mb_rt); 1248 splx(s); 1249 return ENOBUFS; 1250 } 1251 1252 /* 1253 * Send message to routing daemon to install 1254 * a route into the kernel table 1255 */ 1256 k_igmpsrc.sin_addr = ip->ip_src; 1257 1258 im = mtod(mm, struct igmpmsg *); 1259 im->im_msgtype = IGMPMSG_NOCACHE; 1260 im->im_mbz = 0; 1261 1262 mrtstat.mrts_upcalls++; 1263 1264 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1265 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1266 ++mrtstat.mrts_upq_sockfull; 1267 m_free(mb_ntry); 1268 m_freem(mb0); 1269 m_free(mb_rt); 1270 splx(s); 1271 return ENOBUFS; 1272 } 1273 1274 rt = mtod(mb_rt, struct mfc *); 1275 1276 /* insert new entry at head of hash chain */ 1277 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1278 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1279 rt->mfc_expire = UPCALL_EXPIRE; 1280 nexpire[hash]++; 1281 for (i = 0; i < numvifs; i++) 1282 rt->mfc_ttls[i] = 0; 1283 rt->mfc_parent = -1; 1284 1285 /* link into table */ 1286 mb_rt->m_next = mfctable[hash]; 1287 mfctable[hash] = mb_rt; 1288 mb_rt->m_act = NULL; 1289 1290 rte_m = mb_rt; 1291 } else { 1292 /* determine if q has overflowed */ 1293 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act) 1294 npkts++; 1295 1296 if (npkts > MAX_UPQ) { 1297 mrtstat.mrts_upq_ovflw++; 1298 m_free(mb_ntry); 1299 m_freem(mb0); 1300 splx(s); 1301 return 0; 1302 } 1303 } 1304 1305 mb_ntry->m_act = NULL; 1306 rte = mtod(mb_ntry, struct rtdetq *); 1307 1308 rte->m = mb0; 1309 rte->ifp = ifp; 1310#ifdef UPCALL_TIMING 1311 rte->t = tp; 1312#endif 1313 1314 /* Add this entry to the end of the queue */ 1315 rte_m->m_act = mb_ntry; 1316 1317 splx(s); 1318 1319 return 0; 1320 } 1321} 1322 1323#ifndef MROUTE_LKM 1324int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 1325 struct ip_moptions *) = X_ip_mforward; 1326#endif 1327 1328/* 1329 * Clean up the cache entry if upcall is not serviced 1330 */ 1331static void 1332expire_upcalls(void *unused) 1333{ 1334 struct mbuf *mb_rt, *m, **nptr; 1335 struct rtdetq *rte; 1336 struct mfc *mfc; 1337 int i; 1338 int s; 1339 1340 s = splnet(); 1341 for (i = 0; i < MFCTBLSIZ; i++) { 1342 if (nexpire[i] == 0) 1343 continue; 1344 nptr = &mfctable[i]; 1345 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) { 1346 mfc = mtod(mb_rt, struct mfc *); 1347 1348 /* 1349 * Skip real cache entries 1350 * Make sure it wasn't marked to not expire (shouldn't happen) 1351 * If it expires now 1352 */ 1353 if (mb_rt->m_act != NULL && 1354 mfc->mfc_expire != 0 && 1355 --mfc->mfc_expire == 0) { 1356 if (mrtdebug & DEBUG_EXPIRE) 1357 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1358 (u_long)ntohl(mfc->mfc_origin.s_addr), 1359 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1360 /* 1361 * drop all the packets 1362 * free the mbuf with the pkt, if, timing info 1363 */ 1364 while (mb_rt->m_act) { 1365 m = mb_rt->m_act; 1366 mb_rt->m_act = m->m_act; 1367 1368 rte = mtod(m, struct rtdetq *); 1369 m_freem(rte->m); 1370 m_free(m); 1371 } 1372 ++mrtstat.mrts_cache_cleanups; 1373 nexpire[i]--; 1374 1375 MFREE(mb_rt, *nptr); 1376 } else { 1377 nptr = &mb_rt->m_next; 1378 } 1379 } 1380 } 1381 splx(s); 1382 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 1383} 1384 1385/* 1386 * Packet forwarding routine once entry in the cache is made 1387 */ 1388static int 1389ip_mdq(m, ifp, rt, xmt_vif) 1390 register struct mbuf *m; 1391 register struct ifnet *ifp; 1392 register struct mfc *rt; 1393 register vifi_t xmt_vif; 1394{ 1395 register struct ip *ip = mtod(m, struct ip *); 1396 register vifi_t vifi; 1397 register struct vif *vifp; 1398 register int plen = ip->ip_len; 1399 1400/* 1401 * Macro to send packet on vif. Since RSVP packets don't get counted on 1402 * input, they shouldn't get counted on output, so statistics keeping is 1403 * seperate. 1404 */ 1405#define MC_SEND(ip,vifp,m) { \ 1406 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1407 encap_send((ip), (vifp), (m)); \ 1408 else \ 1409 phyint_send((ip), (vifp), (m)); \ 1410} 1411 1412 /* 1413 * If xmt_vif is not -1, send on only the requested vif. 1414 * 1415 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1416 */ 1417 if (xmt_vif < numvifs) { 1418 MC_SEND(ip, viftable + xmt_vif, m); 1419 return 1; 1420 } 1421 1422 /* 1423 * Don't forward if it didn't arrive from the parent vif for its origin. 1424 */ 1425 vifi = rt->mfc_parent; 1426 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1427 /* came in the wrong interface */ 1428 if (mrtdebug & DEBUG_FORWARD) 1429 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1430 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1431 ++mrtstat.mrts_wrong_if; 1432 ++rt->mfc_wrong_if; 1433 /* 1434 * If we are doing PIM assert processing, and we are forwarding 1435 * packets on this interface, and it is a broadcast medium 1436 * interface (and not a tunnel), send a message to the routing daemon. 1437 */ 1438 if (pim_assert && rt->mfc_ttls[vifi] && 1439 (ifp->if_flags & IFF_BROADCAST) && 1440 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1441 struct sockaddr_in k_igmpsrc; 1442 struct mbuf *mm; 1443 struct igmpmsg *im; 1444 int hlen = ip->ip_hl << 2; 1445 struct timeval now; 1446 register u_long delta; 1447 1448 GET_TIME(now); 1449 1450 TV_DELTA(rt->mfc_last_assert, now, delta); 1451 1452 if (delta > ASSERT_MSG_TIME) { 1453 mm = m_copy(m, 0, hlen); 1454 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1455 mm = m_pullup(mm, hlen); 1456 if (mm == NULL) { 1457 return ENOBUFS; 1458 } 1459 1460 rt->mfc_last_assert = now; 1461 1462 im = mtod(mm, struct igmpmsg *); 1463 im->im_msgtype = IGMPMSG_WRONGVIF; 1464 im->im_mbz = 0; 1465 im->im_vif = vifi; 1466 1467 k_igmpsrc.sin_addr = im->im_src; 1468 1469 socket_send(ip_mrouter, mm, &k_igmpsrc); 1470 } 1471 } 1472 return 0; 1473 } 1474 1475 /* If I sourced this packet, it counts as output, else it was input. */ 1476 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1477 viftable[vifi].v_pkt_out++; 1478 viftable[vifi].v_bytes_out += plen; 1479 } else { 1480 viftable[vifi].v_pkt_in++; 1481 viftable[vifi].v_bytes_in += plen; 1482 } 1483 rt->mfc_pkt_cnt++; 1484 rt->mfc_byte_cnt += plen; 1485 1486 /* 1487 * For each vif, decide if a copy of the packet should be forwarded. 1488 * Forward if: 1489 * - the ttl exceeds the vif's threshold 1490 * - there are group members downstream on interface 1491 */ 1492 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1493 if ((rt->mfc_ttls[vifi] > 0) && 1494 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1495 vifp->v_pkt_out++; 1496 vifp->v_bytes_out += plen; 1497 MC_SEND(ip, vifp, m); 1498 } 1499 1500 return 0; 1501} 1502 1503/* 1504 * check if a vif number is legal/ok. This is used by ip_output, to export 1505 * numvifs there, 1506 */ 1507static int 1508X_legal_vif_num(vif) 1509 int vif; 1510{ 1511 if (vif >= 0 && vif < numvifs) 1512 return(1); 1513 else 1514 return(0); 1515} 1516 1517#ifndef MROUTE_LKM 1518int (*legal_vif_num)(int) = X_legal_vif_num; 1519#endif 1520 1521/* 1522 * Return the local address used by this vif 1523 */ 1524static u_long 1525X_ip_mcast_src(vifi) 1526 int vifi; 1527{ 1528 if (vifi >= 0 && vifi < numvifs) 1529 return viftable[vifi].v_lcl_addr.s_addr; 1530 else 1531 return INADDR_ANY; 1532} 1533 1534#ifndef MROUTE_LKM 1535u_long (*ip_mcast_src)(int) = X_ip_mcast_src; 1536#endif 1537 1538static void 1539phyint_send(ip, vifp, m) 1540 struct ip *ip; 1541 struct vif *vifp; 1542 struct mbuf *m; 1543{ 1544 register struct mbuf *mb_copy; 1545 register int hlen = ip->ip_hl << 2; 1546 1547 /* 1548 * Make a new reference to the packet; make sure that 1549 * the IP header is actually copied, not just referenced, 1550 * so that ip_output() only scribbles on the copy. 1551 */ 1552 mb_copy = m_copy(m, 0, M_COPYALL); 1553 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1554 mb_copy = m_pullup(mb_copy, hlen); 1555 if (mb_copy == NULL) 1556 return; 1557 1558 if (vifp->v_rate_limit <= 0) 1559 tbf_send_packet(vifp, mb_copy); 1560 else 1561 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1562} 1563 1564static void 1565encap_send(ip, vifp, m) 1566 register struct ip *ip; 1567 register struct vif *vifp; 1568 register struct mbuf *m; 1569{ 1570 register struct mbuf *mb_copy; 1571 register struct ip *ip_copy; 1572 register int i, len = ip->ip_len; 1573 1574 /* 1575 * copy the old packet & pullup its IP header into the 1576 * new mbuf so we can modify it. Try to fill the new 1577 * mbuf since if we don't the ethernet driver will. 1578 */ 1579 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER); 1580 if (mb_copy == NULL) 1581 return; 1582 mb_copy->m_data += max_linkhdr; 1583 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1584 1585 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1586 m_freem(mb_copy); 1587 return; 1588 } 1589 i = MHLEN - M_LEADINGSPACE(mb_copy); 1590 if (i > len) 1591 i = len; 1592 mb_copy = m_pullup(mb_copy, i); 1593 if (mb_copy == NULL) 1594 return; 1595 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1596 1597 /* 1598 * fill in the encapsulating IP header. 1599 */ 1600 ip_copy = mtod(mb_copy, struct ip *); 1601 *ip_copy = multicast_encap_iphdr; 1602 ip_copy->ip_id = htons(ip_id++); 1603 ip_copy->ip_len += len; 1604 ip_copy->ip_src = vifp->v_lcl_addr; 1605 ip_copy->ip_dst = vifp->v_rmt_addr; 1606 1607 /* 1608 * turn the encapsulated IP header back into a valid one. 1609 */ 1610 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1611 --ip->ip_ttl; 1612 HTONS(ip->ip_len); 1613 HTONS(ip->ip_off); 1614 ip->ip_sum = 0; 1615 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1616 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1617 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1618 1619 if (vifp->v_rate_limit <= 0) 1620 tbf_send_packet(vifp, mb_copy); 1621 else 1622 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1623} 1624 1625/* 1626 * De-encapsulate a packet and feed it back through ip input (this 1627 * routine is called whenever IP gets a packet with proto type 1628 * ENCAP_PROTO and a local destination address). 1629 */ 1630void 1631#ifdef MROUTE_LKM 1632X_ipip_input(m, iphlen) 1633#else 1634ipip_input(m, iphlen) 1635#endif 1636 register struct mbuf *m; 1637 int iphlen; 1638{ 1639 struct ifnet *ifp = m->m_pkthdr.rcvif; 1640 register struct ip *ip = mtod(m, struct ip *); 1641 register int hlen = ip->ip_hl << 2; 1642 register int s; 1643 register struct ifqueue *ifq; 1644 register struct vif *vifp; 1645 1646 if (!have_encap_tunnel) { 1647 rip_input(m, iphlen); 1648 return; 1649 } 1650 /* 1651 * dump the packet if it's not to a multicast destination or if 1652 * we don't have an encapsulating tunnel with the source. 1653 * Note: This code assumes that the remote site IP address 1654 * uniquely identifies the tunnel (i.e., that this site has 1655 * at most one tunnel with the remote site). 1656 */ 1657 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) { 1658 ++mrtstat.mrts_bad_tunnel; 1659 m_freem(m); 1660 return; 1661 } 1662 if (ip->ip_src.s_addr != last_encap_src) { 1663 register struct vif *vife; 1664 1665 vifp = viftable; 1666 vife = vifp + numvifs; 1667 last_encap_src = ip->ip_src.s_addr; 1668 last_encap_vif = 0; 1669 for ( ; vifp < vife; ++vifp) 1670 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1671 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1672 == VIFF_TUNNEL) 1673 last_encap_vif = vifp; 1674 break; 1675 } 1676 } 1677 if ((vifp = last_encap_vif) == 0) { 1678 last_encap_src = 0; 1679 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1680 m_freem(m); 1681 if (mrtdebug) 1682 log(LOG_DEBUG, "ip_mforward: no tunnel with %lx\n", 1683 (u_long)ntohl(ip->ip_src.s_addr)); 1684 return; 1685 } 1686 ifp = vifp->v_ifp; 1687 1688 if (hlen > IP_HDR_LEN) 1689 ip_stripoptions(m, (struct mbuf *) 0); 1690 m->m_data += IP_HDR_LEN; 1691 m->m_len -= IP_HDR_LEN; 1692 m->m_pkthdr.len -= IP_HDR_LEN; 1693 m->m_pkthdr.rcvif = ifp; 1694 1695 ifq = &ipintrq; 1696 s = splimp(); 1697 if (IF_QFULL(ifq)) { 1698 IF_DROP(ifq); 1699 m_freem(m); 1700 } else { 1701 IF_ENQUEUE(ifq, m); 1702 /* 1703 * normally we would need a "schednetisr(NETISR_IP)" 1704 * here but we were called by ip_input and it is going 1705 * to loop back & try to dequeue the packet we just 1706 * queued as soon as we return so we avoid the 1707 * unnecessary software interrrupt. 1708 */ 1709 } 1710 splx(s); 1711} 1712 1713/* 1714 * Token bucket filter module 1715 */ 1716 1717static void 1718tbf_control(vifp, m, ip, p_len) 1719 register struct vif *vifp; 1720 register struct mbuf *m; 1721 register struct ip *ip; 1722 register u_long p_len; 1723{ 1724 register struct tbf *t = vifp->v_tbf; 1725 1726 if (p_len > MAX_BKT_SIZE) { 1727 /* drop if packet is too large */ 1728 mrtstat.mrts_pkt2large++; 1729 m_freem(m); 1730 return; 1731 } 1732 1733 tbf_update_tokens(vifp); 1734 1735 /* if there are enough tokens, 1736 * and the queue is empty, 1737 * send this packet out 1738 */ 1739 1740 if (t->tbf_q_len == 0) { 1741 /* queue empty, send packet if enough tokens */ 1742 if (p_len <= t->tbf_n_tok) { 1743 t->tbf_n_tok -= p_len; 1744 tbf_send_packet(vifp, m); 1745 } else { 1746 /* queue packet and timeout till later */ 1747 tbf_queue(vifp, m); 1748 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1749 } 1750 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1751 /* finite queue length, so queue pkts and process queue */ 1752 tbf_queue(vifp, m); 1753 tbf_process_q(vifp); 1754 } else { 1755 /* queue length too much, try to dq and queue and process */ 1756 if (!tbf_dq_sel(vifp, ip)) { 1757 mrtstat.mrts_q_overflow++; 1758 m_freem(m); 1759 return; 1760 } else { 1761 tbf_queue(vifp, m); 1762 tbf_process_q(vifp); 1763 } 1764 } 1765 return; 1766} 1767 1768/* 1769 * adds a packet to the queue at the interface 1770 */ 1771static void 1772tbf_queue(vifp, m) 1773 register struct vif *vifp; 1774 register struct mbuf *m; 1775{ 1776 register int s = splnet(); 1777 register struct tbf *t = vifp->v_tbf; 1778 1779 if (t->tbf_t == NULL) { 1780 /* Queue was empty */ 1781 t->tbf_q = m; 1782 } else { 1783 /* Insert at tail */ 1784 t->tbf_t->m_act = m; 1785 } 1786 1787 /* Set new tail pointer */ 1788 t->tbf_t = m; 1789 1790#ifdef DIAGNOSTIC 1791 /* Make sure we didn't get fed a bogus mbuf */ 1792 if (m->m_act) 1793 panic("tbf_queue: m_act"); 1794#endif 1795 m->m_act = NULL; 1796 1797 t->tbf_q_len++; 1798 1799 splx(s); 1800} 1801 1802 1803/* 1804 * processes the queue at the interface 1805 */ 1806static void 1807tbf_process_q(vifp) 1808 register struct vif *vifp; 1809{ 1810 register struct mbuf *m; 1811 register int len; 1812 register int s = splnet(); 1813 register struct tbf *t = vifp->v_tbf; 1814 1815 /* loop through the queue at the interface and send as many packets 1816 * as possible 1817 */ 1818 while (t->tbf_q_len > 0) { 1819 m = t->tbf_q; 1820 1821 len = mtod(m, struct ip *)->ip_len; 1822 1823 /* determine if the packet can be sent */ 1824 if (len <= t->tbf_n_tok) { 1825 /* if so, 1826 * reduce no of tokens, dequeue the packet, 1827 * send the packet. 1828 */ 1829 t->tbf_n_tok -= len; 1830 1831 t->tbf_q = m->m_act; 1832 if (--t->tbf_q_len == 0) 1833 t->tbf_t = NULL; 1834 1835 m->m_act = NULL; 1836 tbf_send_packet(vifp, m); 1837 1838 } else break; 1839 } 1840 splx(s); 1841} 1842 1843static void 1844tbf_reprocess_q(xvifp) 1845 void *xvifp; 1846{ 1847 register struct vif *vifp = xvifp; 1848 if (ip_mrouter == NULL) 1849 return; 1850 1851 tbf_update_tokens(vifp); 1852 1853 tbf_process_q(vifp); 1854 1855 if (vifp->v_tbf->tbf_q_len) 1856 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1857} 1858 1859/* function that will selectively discard a member of the queue 1860 * based on the precedence value and the priority 1861 */ 1862static int 1863tbf_dq_sel(vifp, ip) 1864 register struct vif *vifp; 1865 register struct ip *ip; 1866{ 1867 register int s = splnet(); 1868 register u_int p; 1869 register struct mbuf *m, *last; 1870 register struct mbuf **np; 1871 register struct tbf *t = vifp->v_tbf; 1872 1873 p = priority(vifp, ip); 1874 1875 np = &t->tbf_q; 1876 last = NULL; 1877 while ((m = *np) != NULL) { 1878 if (p > priority(vifp, mtod(m, struct ip *))) { 1879 *np = m->m_act; 1880 /* If we're removing the last packet, fix the tail pointer */ 1881 if (m == t->tbf_t) 1882 t->tbf_t = last; 1883 m_freem(m); 1884 /* it's impossible for the queue to be empty, but 1885 * we check anyway. */ 1886 if (--t->tbf_q_len == 0) 1887 t->tbf_t = NULL; 1888 splx(s); 1889 mrtstat.mrts_drop_sel++; 1890 return(1); 1891 } 1892 np = &m->m_act; 1893 last = m; 1894 } 1895 splx(s); 1896 return(0); 1897} 1898 1899static void 1900tbf_send_packet(vifp, m) 1901 register struct vif *vifp; 1902 register struct mbuf *m; 1903{ 1904 struct ip_moptions imo; 1905 int error; 1906 static struct route ro; 1907 int s = splnet(); 1908 1909 if (vifp->v_flags & VIFF_TUNNEL) { 1910 /* If tunnel options */ 1911 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1912 IP_FORWARDING, (struct ip_moptions *)0); 1913 } else { 1914 imo.imo_multicast_ifp = vifp->v_ifp; 1915 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1916 imo.imo_multicast_loop = 1; 1917 imo.imo_multicast_vif = -1; 1918 1919 /* 1920 * Re-entrancy should not be a problem here, because 1921 * the packets that we send out and are looped back at us 1922 * should get rejected because they appear to come from 1923 * the loopback interface, thus preventing looping. 1924 */ 1925 error = ip_output(m, (struct mbuf *)0, &ro, 1926 IP_FORWARDING, &imo); 1927 1928 if (mrtdebug & DEBUG_XMIT) 1929 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1930 vifp - viftable, error); 1931 } 1932 splx(s); 1933} 1934 1935/* determine the current time and then 1936 * the elapsed time (between the last time and time now) 1937 * in milliseconds & update the no. of tokens in the bucket 1938 */ 1939static void 1940tbf_update_tokens(vifp) 1941 register struct vif *vifp; 1942{ 1943 struct timeval tp; 1944 register u_long tm; 1945 register int s = splnet(); 1946 register struct tbf *t = vifp->v_tbf; 1947 1948 GET_TIME(tp); 1949 1950 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1951 1952 /* 1953 * This formula is actually 1954 * "time in seconds" * "bytes/second". 1955 * 1956 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1957 * 1958 * The (1000/1024) was introduced in add_vif to optimize 1959 * this divide into a shift. 1960 */ 1961 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1962 t->tbf_last_pkt_t = tp; 1963 1964 if (t->tbf_n_tok > MAX_BKT_SIZE) 1965 t->tbf_n_tok = MAX_BKT_SIZE; 1966 1967 splx(s); 1968} 1969 1970static int 1971priority(vifp, ip) 1972 register struct vif *vifp; 1973 register struct ip *ip; 1974{ 1975 register int prio; 1976 1977 /* temporary hack; may add general packet classifier some day */ 1978 1979 /* 1980 * The UDP port space is divided up into four priority ranges: 1981 * [0, 16384) : unclassified - lowest priority 1982 * [16384, 32768) : audio - highest priority 1983 * [32768, 49152) : whiteboard - medium priority 1984 * [49152, 65536) : video - low priority 1985 */ 1986 if (ip->ip_p == IPPROTO_UDP) { 1987 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1988 switch (ntohs(udp->uh_dport) & 0xc000) { 1989 case 0x4000: 1990 prio = 70; 1991 break; 1992 case 0x8000: 1993 prio = 60; 1994 break; 1995 case 0xc000: 1996 prio = 55; 1997 break; 1998 default: 1999 prio = 50; 2000 break; 2001 } 2002 if (tbfdebug > 1) 2003 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio); 2004 } else { 2005 prio = 50; 2006 } 2007 return prio; 2008} 2009 2010/* 2011 * End of token bucket filter modifications 2012 */ 2013 2014int 2015ip_rsvp_vif_init(so, sopt) 2016 struct socket *so; 2017 struct sockopt *sopt; 2018{ 2019 int error, i, s; 2020 2021 if (rsvpdebug) 2022 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2023 so->so_type, so->so_proto->pr_protocol); 2024 2025 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2026 return EOPNOTSUPP; 2027 2028 /* Check mbuf. */ 2029 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 2030 if (error) 2031 return (error); 2032 2033 if (rsvpdebug) 2034 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", i, rsvp_on); 2035 2036 s = splnet(); 2037 2038 /* Check vif. */ 2039 if (!legal_vif_num(i)) { 2040 splx(s); 2041 return EADDRNOTAVAIL; 2042 } 2043 2044 /* Check if socket is available. */ 2045 if (viftable[i].v_rsvpd != NULL) { 2046 splx(s); 2047 return EADDRINUSE; 2048 } 2049 2050 viftable[i].v_rsvpd = so; 2051 /* This may seem silly, but we need to be sure we don't over-increment 2052 * the RSVP counter, in case something slips up. 2053 */ 2054 if (!viftable[i].v_rsvp_on) { 2055 viftable[i].v_rsvp_on = 1; 2056 rsvp_on++; 2057 } 2058 2059 splx(s); 2060 return 0; 2061} 2062 2063int 2064ip_rsvp_vif_done(so, sopt) 2065 struct socket *so; 2066 struct sockopt *sopt; 2067{ 2068 int error, i, s; 2069 2070 if (rsvpdebug) 2071 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2072 so->so_type, so->so_proto->pr_protocol); 2073 2074 if (so->so_type != SOCK_RAW || 2075 so->so_proto->pr_protocol != IPPROTO_RSVP) 2076 return EOPNOTSUPP; 2077 2078 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 2079 if (error) 2080 return (error); 2081 2082 s = splnet(); 2083 2084 /* Check vif. */ 2085 if (!legal_vif_num(i)) { 2086 splx(s); 2087 return EADDRNOTAVAIL; 2088 } 2089 2090 if (rsvpdebug) 2091 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n", 2092 viftable[i].v_rsvpd, so); 2093 2094 viftable[i].v_rsvpd = NULL; 2095 /* 2096 * This may seem silly, but we need to be sure we don't over-decrement 2097 * the RSVP counter, in case something slips up. 2098 */ 2099 if (viftable[i].v_rsvp_on) { 2100 viftable[i].v_rsvp_on = 0; 2101 rsvp_on--; 2102 } 2103 2104 splx(s); 2105 return 0; 2106} 2107 2108void 2109ip_rsvp_force_done(so) 2110 struct socket *so; 2111{ 2112 int vifi; 2113 register int s; 2114 2115 /* Don't bother if it is not the right type of socket. */ 2116 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2117 return; 2118 2119 s = splnet(); 2120 2121 /* The socket may be attached to more than one vif...this 2122 * is perfectly legal. 2123 */ 2124 for (vifi = 0; vifi < numvifs; vifi++) { 2125 if (viftable[vifi].v_rsvpd == so) { 2126 viftable[vifi].v_rsvpd = NULL; 2127 /* This may seem silly, but we need to be sure we don't 2128 * over-decrement the RSVP counter, in case something slips up. 2129 */ 2130 if (viftable[vifi].v_rsvp_on) { 2131 viftable[vifi].v_rsvp_on = 0; 2132 rsvp_on--; 2133 } 2134 } 2135 } 2136 2137 splx(s); 2138 return; 2139} 2140 2141void 2142rsvp_input(m, iphlen) 2143 struct mbuf *m; 2144 int iphlen; 2145{ 2146 int vifi; 2147 register struct ip *ip = mtod(m, struct ip *); 2148 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2149 register int s; 2150 struct ifnet *ifp; 2151 2152 if (rsvpdebug) 2153 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2154 2155 /* Can still get packets with rsvp_on = 0 if there is a local member 2156 * of the group to which the RSVP packet is addressed. But in this 2157 * case we want to throw the packet away. 2158 */ 2159 if (!rsvp_on) { 2160 m_freem(m); 2161 return; 2162 } 2163 2164 /* If the old-style non-vif-associated socket is set, then use 2165 * it and ignore the new ones. 2166 */ 2167 if (ip_rsvpd != NULL) { 2168 if (rsvpdebug) 2169 printf("rsvp_input: Sending packet up old-style socket\n"); 2170 rip_input(m, iphlen); 2171 return; 2172 } 2173 2174 s = splnet(); 2175 2176 if (rsvpdebug) 2177 printf("rsvp_input: check vifs\n"); 2178 2179#ifdef DIAGNOSTIC 2180 if (!(m->m_flags & M_PKTHDR)) 2181 panic("rsvp_input no hdr"); 2182#endif 2183 2184 ifp = m->m_pkthdr.rcvif; 2185 /* Find which vif the packet arrived on. */ 2186 for (vifi = 0; vifi < numvifs; vifi++) { 2187 if (viftable[vifi].v_ifp == ifp) 2188 break; 2189 } 2190 2191 if (vifi == numvifs) { 2192 /* Can't find vif packet arrived on. Drop packet. */ 2193 if (rsvpdebug) 2194 printf("rsvp_input: Can't find vif for packet...dropping it.\n"); 2195 m_freem(m); 2196 splx(s); 2197 return; 2198 } 2199 2200 if (rsvpdebug) 2201 printf("rsvp_input: check socket\n"); 2202 2203 if (viftable[vifi].v_rsvpd == NULL) { 2204 /* drop packet, since there is no specific socket for this 2205 * interface */ 2206 if (rsvpdebug) 2207 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2208 m_freem(m); 2209 splx(s); 2210 return; 2211 } 2212 rsvp_src.sin_addr = ip->ip_src; 2213 2214 if (rsvpdebug && m) 2215 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2216 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2217 2218 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2219 if (rsvpdebug) 2220 printf("rsvp_input: Failed to append to socket\n"); 2221 else 2222 if (rsvpdebug) 2223 printf("rsvp_input: send packet up\n"); 2224 2225 splx(s); 2226} 2227 2228#ifdef MROUTE_LKM 2229#include <sys/conf.h> 2230#include <sys/exec.h> 2231#include <sys/sysent.h> 2232#include <sys/lkm.h> 2233 2234MOD_MISC("ip_mroute_mod") 2235 2236static int 2237ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd) 2238{ 2239 int i; 2240 struct lkm_misc *args = lkmtp->private.lkm_misc; 2241 int err = 0; 2242 2243 switch(cmd) { 2244 static int (*old_ip_mrouter_cmd)(); 2245 static int (*old_ip_mrouter_done)(); 2246 static int (*old_ip_mforward)(); 2247 static int (*old_mrt_ioctl)(); 2248 static void (*old_proto4_input)(); 2249 static int (*old_legal_vif_num)(); 2250 extern struct protosw inetsw[]; 2251 2252 case LKM_E_LOAD: 2253 if(lkmexists(lkmtp) || ip_mrtproto) 2254 return(EEXIST); 2255 old_ip_mrouter_cmd = ip_mrouter_cmd; 2256 ip_mrouter_cmd = X_ip_mrouter_cmd; 2257 old_ip_mrouter_done = ip_mrouter_done; 2258 ip_mrouter_done = X_ip_mrouter_done; 2259 old_ip_mforward = ip_mforward; 2260 ip_mforward = X_ip_mforward; 2261 old_mrt_ioctl = mrt_ioctl; 2262 mrt_ioctl = X_mrt_ioctl; 2263 old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input; 2264 inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input; 2265 old_legal_vif_num = legal_vif_num; 2266 legal_vif_num = X_legal_vif_num; 2267 ip_mrtproto = IGMP_DVMRP; 2268 2269 printf("\nIP multicast routing loaded\n"); 2270 break; 2271 2272 case LKM_E_UNLOAD: 2273 if (ip_mrouter) 2274 return EINVAL; 2275 2276 ip_mrouter_cmd = old_ip_mrouter_cmd; 2277 ip_mrouter_done = old_ip_mrouter_done; 2278 ip_mforward = old_ip_mforward; 2279 mrt_ioctl = old_mrt_ioctl; 2280 inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input; 2281 legal_vif_num = old_legal_vif_num; 2282 ip_mrtproto = 0; 2283 break; 2284 2285 default: 2286 err = EINVAL; 2287 break; 2288 } 2289 2290 return(err); 2291} 2292 2293int 2294ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) { 2295 DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle, 2296 nosys); 2297} 2298 2299#endif /* MROUTE_LKM */ 2300#endif /* MROUTING */ 2301