ip_mroute.c revision 33108
1/* 2 * IP multicast forwarding procedures 3 * 4 * Written by David Waitzman, BBN Labs, August 1988. 5 * Modified by Steve Deering, Stanford, February 1989. 6 * Modified by Mark J. Steiglitz, Stanford, May, 1991 7 * Modified by Van Jacobson, LBL, January 1993 8 * Modified by Ajit Thyagarajan, PARC, August 1993 9 * Modified by Bill Fenner, PARC, April 1995 10 * 11 * MROUTING Revision: 3.5 12 * $Id: ip_mroute.c,v 1.42 1997/10/28 15:58:50 bde Exp $ 13 */ 14 15#include "opt_diagnostic.h" 16#include "opt_mrouting.h" 17 18#include <sys/param.h> 19#include <sys/systm.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 39#ifndef NTOHL 40#if BYTE_ORDER != BIG_ENDIAN 41#define NTOHL(d) ((d) = ntohl((d))) 42#define NTOHS(d) ((d) = ntohs((u_short)(d))) 43#define HTONL(d) ((d) = htonl((d))) 44#define HTONS(d) ((d) = htons((u_short)(d))) 45#else 46#define NTOHL(d) 47#define NTOHS(d) 48#define HTONL(d) 49#define HTONS(d) 50#endif 51#endif 52 53#ifndef MROUTING 54extern u_long _ip_mcast_src __P((int vifi)); 55extern int _ip_mforward __P((struct ip *ip, struct ifnet *ifp, 56 struct mbuf *m, struct ip_moptions *imo)); 57extern int _ip_mrouter_done __P((void)); 58extern int _ip_mrouter_get __P((int cmd, struct socket *so, 59 struct mbuf **m)); 60extern int _ip_mrouter_set __P((int cmd, struct socket *so, 61 struct mbuf *m)); 62extern int _mrt_ioctl __P((int req, caddr_t data, struct proc *p)); 63 64/* 65 * Dummy routines and globals used when multicast routing is not compiled in. 66 */ 67 68struct socket *ip_mrouter = NULL; 69static u_int ip_mrtproto = 0; 70static struct mrtstat mrtstat; 71u_int rsvpdebug = 0; 72 73int 74_ip_mrouter_set(cmd, so, m) 75 int cmd; 76 struct socket *so; 77 struct mbuf *m; 78{ 79 return(EOPNOTSUPP); 80} 81 82int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = _ip_mrouter_set; 83 84 85int 86_ip_mrouter_get(cmd, so, m) 87 int cmd; 88 struct socket *so; 89 struct mbuf **m; 90{ 91 return(EOPNOTSUPP); 92} 93 94int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = _ip_mrouter_get; 95 96int 97_ip_mrouter_done() 98{ 99 return(0); 100} 101 102int (*ip_mrouter_done)(void) = _ip_mrouter_done; 103 104int 105_ip_mforward(ip, ifp, m, imo) 106 struct ip *ip; 107 struct ifnet *ifp; 108 struct mbuf *m; 109 struct ip_moptions *imo; 110{ 111 return(0); 112} 113 114int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 115 struct ip_moptions *) = _ip_mforward; 116 117int 118_mrt_ioctl(int req, caddr_t data, struct proc *p) 119{ 120 return EOPNOTSUPP; 121} 122 123int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl; 124 125void 126rsvp_input(m, iphlen) /* XXX must fixup manually */ 127 struct mbuf *m; 128 int iphlen; 129{ 130 /* Can still get packets with rsvp_on = 0 if there is a local member 131 * of the group to which the RSVP packet is addressed. But in this 132 * case we want to throw the packet away. 133 */ 134 if (!rsvp_on) { 135 m_freem(m); 136 return; 137 } 138 139 if (ip_rsvpd != NULL) { 140 if (rsvpdebug) 141 printf("rsvp_input: Sending packet up old-style socket\n"); 142 rip_input(m, iphlen); 143 return; 144 } 145 /* Drop the packet */ 146 m_freem(m); 147} 148 149void ipip_input(struct mbuf *m, int iphlen) { /* XXX must fixup manually */ 150 rip_input(m, iphlen); 151} 152 153int (*legal_vif_num)(int) = 0; 154 155/* 156 * This should never be called, since IP_MULTICAST_VIF should fail, but 157 * just in case it does get called, the code a little lower in ip_output 158 * will assign the packet a local address. 159 */ 160u_long 161_ip_mcast_src(int vifi) { return INADDR_ANY; } 162u_long (*ip_mcast_src)(int) = _ip_mcast_src; 163 164int 165ip_rsvp_vif_init(so, m) 166 struct socket *so; 167 struct mbuf *m; 168{ 169 return(EINVAL); 170} 171 172int 173ip_rsvp_vif_done(so, m) 174 struct socket *so; 175 struct mbuf *m; 176{ 177 return(EINVAL); 178} 179 180void 181ip_rsvp_force_done(so) 182 struct socket *so; 183{ 184 return; 185} 186 187#else /* MROUTING */ 188 189#define M_HASCL(m) ((m)->m_flags & M_EXT) 190 191#define INSIZ sizeof(struct in_addr) 192#define same(a1, a2) \ 193 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0) 194 195#define MT_MRTABLE MT_RTABLE /* since nothing else uses it */ 196 197/* 198 * Globals. All but ip_mrouter and ip_mrtproto could be static, 199 * except for netstat or debugging purposes. 200 */ 201#ifndef MROUTE_LKM 202struct socket *ip_mrouter = NULL; 203struct mrtstat mrtstat; 204 205int ip_mrtproto = IGMP_DVMRP; /* for netstat only */ 206#else /* MROUTE_LKM */ 207extern void X_ipip_input __P((struct mbuf *m, int iphlen)); 208extern struct mrtstat mrtstat; 209static int ip_mrtproto; 210#endif 211 212#define NO_RTE_FOUND 0x1 213#define RTE_FOUND 0x2 214 215static struct mbuf *mfctable[MFCTBLSIZ]; 216static u_char nexpire[MFCTBLSIZ]; 217static struct vif viftable[MAXVIFS]; 218static u_int mrtdebug = 0; /* debug level */ 219#define DEBUG_MFC 0x02 220#define DEBUG_FORWARD 0x04 221#define DEBUG_EXPIRE 0x08 222#define DEBUG_XMIT 0x10 223static u_int tbfdebug = 0; /* tbf debug level */ 224static u_int rsvpdebug = 0; /* rsvp debug level */ 225 226static struct callout_handle expire_upcalls_ch; 227 228#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 229#define UPCALL_EXPIRE 6 /* number of timeouts */ 230 231/* 232 * Define the token bucket filter structures 233 * tbftable -> each vif has one of these for storing info 234 */ 235 236static struct tbf tbftable[MAXVIFS]; 237#define TBF_REPROCESS (hz / 100) /* 100x / second */ 238 239/* 240 * 'Interfaces' associated with decapsulator (so we can tell 241 * packets that went through it from ones that get reflected 242 * by a broken gateway). These interfaces are never linked into 243 * the system ifnet list & no routes point to them. I.e., packets 244 * can't be sent this way. They only exist as a placeholder for 245 * multicast source verification. 246 */ 247static struct ifnet multicast_decap_if[MAXVIFS]; 248 249#define ENCAP_TTL 64 250#define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 251 252/* prototype IP hdr for encapsulated packets */ 253static struct ip multicast_encap_iphdr = { 254#if BYTE_ORDER == LITTLE_ENDIAN 255 sizeof(struct ip) >> 2, IPVERSION, 256#else 257 IPVERSION, sizeof(struct ip) >> 2, 258#endif 259 0, /* tos */ 260 sizeof(struct ip), /* total length */ 261 0, /* id */ 262 0, /* frag offset */ 263 ENCAP_TTL, ENCAP_PROTO, 264 0, /* checksum */ 265}; 266 267/* 268 * Private variables. 269 */ 270static vifi_t numvifs = 0; 271static int have_encap_tunnel = 0; 272 273/* 274 * one-back cache used by ipip_input to locate a tunnel's vif 275 * given a datagram's src ip address. 276 */ 277static u_long last_encap_src; 278static struct vif *last_encap_vif; 279 280static u_long X_ip_mcast_src __P((int vifi)); 281static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo)); 282static int X_ip_mrouter_done __P((void)); 283static int X_ip_mrouter_get __P((int cmd, struct socket *so, struct mbuf **m)); 284static int X_ip_mrouter_set __P((int cmd, struct socket *so, struct mbuf *m)); 285static int X_legal_vif_num __P((int vif)); 286static int X_mrt_ioctl __P((int cmd, caddr_t data)); 287 288static int get_sg_cnt(struct sioc_sg_req *); 289static int get_vif_cnt(struct sioc_vif_req *); 290static int ip_mrouter_init(struct socket *, struct mbuf *); 291static int add_vif(struct vifctl *); 292static int del_vif(vifi_t *); 293static int add_mfc(struct mfcctl *); 294static int del_mfc(struct mfcctl *); 295static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 296static int get_version(struct mbuf *); 297static int get_assert(struct mbuf *); 298static int set_assert(int *); 299static void expire_upcalls(void *); 300static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, 301 vifi_t); 302static void phyint_send(struct ip *, struct vif *, struct mbuf *); 303static void encap_send(struct ip *, struct vif *, struct mbuf *); 304static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 305static void tbf_queue(struct vif *, struct mbuf *); 306static void tbf_process_q(struct vif *); 307static void tbf_reprocess_q(void *); 308static int tbf_dq_sel(struct vif *, struct ip *); 309static void tbf_send_packet(struct vif *, struct mbuf *); 310static void tbf_update_tokens(struct vif *); 311static int priority(struct vif *, struct ip *); 312void multiencap_decap(struct mbuf *); 313 314/* 315 * whether or not special PIM assert processing is enabled. 316 */ 317static int pim_assert; 318/* 319 * Rate limit for assert notification messages, in usec 320 */ 321#define ASSERT_MSG_TIME 3000000 322 323/* 324 * Hash function for a source, group entry 325 */ 326#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 327 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 328 329/* 330 * Find a route for a given origin IP address and Multicast group address 331 * Type of service parameter to be added in the future!!! 332 */ 333 334#define MFCFIND(o, g, rt) { \ 335 register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \ 336 register struct mfc *_rt = NULL; \ 337 rt = NULL; \ 338 ++mrtstat.mrts_mfc_lookups; \ 339 while (_mb_rt) { \ 340 _rt = mtod(_mb_rt, struct mfc *); \ 341 if ((_rt->mfc_origin.s_addr == o) && \ 342 (_rt->mfc_mcastgrp.s_addr == g) && \ 343 (_mb_rt->m_act == NULL)) { \ 344 rt = _rt; \ 345 break; \ 346 } \ 347 _mb_rt = _mb_rt->m_next; \ 348 } \ 349 if (rt == NULL) { \ 350 ++mrtstat.mrts_mfc_misses; \ 351 } \ 352} 353 354 355/* 356 * Macros to compute elapsed time efficiently 357 * Borrowed from Van Jacobson's scheduling code 358 */ 359#define TV_DELTA(a, b, delta) { \ 360 register int xxs; \ 361 \ 362 delta = (a).tv_usec - (b).tv_usec; \ 363 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 364 switch (xxs) { \ 365 case 2: \ 366 delta += 1000000; \ 367 /* fall through */ \ 368 case 1: \ 369 delta += 1000000; \ 370 break; \ 371 default: \ 372 delta += (1000000 * xxs); \ 373 } \ 374 } \ 375} 376 377#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 378 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 379 380#ifdef UPCALL_TIMING 381u_long upcall_data[51]; 382static void collate(struct timeval *); 383#endif /* UPCALL_TIMING */ 384 385 386/* 387 * Handle MRT setsockopt commands to modify the multicast routing tables. 388 */ 389static int 390X_ip_mrouter_set(cmd, so, m) 391 int cmd; 392 struct socket *so; 393 struct mbuf *m; 394{ 395 if (cmd != MRT_INIT && so != ip_mrouter) return EACCES; 396 397 switch (cmd) { 398 case MRT_INIT: return ip_mrouter_init(so, m); 399 case MRT_DONE: return ip_mrouter_done(); 400 case MRT_ADD_VIF: return add_vif (mtod(m, struct vifctl *)); 401 case MRT_DEL_VIF: return del_vif (mtod(m, vifi_t *)); 402 case MRT_ADD_MFC: return add_mfc (mtod(m, struct mfcctl *)); 403 case MRT_DEL_MFC: return del_mfc (mtod(m, struct mfcctl *)); 404 case MRT_ASSERT: return set_assert(mtod(m, int *)); 405 default: return EOPNOTSUPP; 406 } 407} 408 409#ifndef MROUTE_LKM 410int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = X_ip_mrouter_set; 411#endif 412 413/* 414 * Handle MRT getsockopt commands 415 */ 416static int 417X_ip_mrouter_get(cmd, so, m) 418 int cmd; 419 struct socket *so; 420 struct mbuf **m; 421{ 422 struct mbuf *mb; 423 424 if (so != ip_mrouter) return EACCES; 425 426 *m = mb = m_get(M_WAIT, MT_SOOPTS); 427 428 switch (cmd) { 429 case MRT_VERSION: return get_version(mb); 430 case MRT_ASSERT: return get_assert(mb); 431 default: return EOPNOTSUPP; 432 } 433} 434 435#ifndef MROUTE_LKM 436int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = X_ip_mrouter_get; 437#endif 438 439/* 440 * Handle ioctl commands to obtain information from the cache 441 */ 442static int 443X_mrt_ioctl(cmd, data) 444 int cmd; 445 caddr_t data; 446{ 447 int error = 0; 448 449 switch (cmd) { 450 case (SIOCGETVIFCNT): 451 return (get_vif_cnt((struct sioc_vif_req *)data)); 452 break; 453 case (SIOCGETSGCNT): 454 return (get_sg_cnt((struct sioc_sg_req *)data)); 455 break; 456 default: 457 return (EINVAL); 458 break; 459 } 460 return error; 461} 462 463#ifndef MROUTE_LKM 464int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl; 465#endif 466 467/* 468 * returns the packet, byte, rpf-failure count for the source group provided 469 */ 470static int 471get_sg_cnt(req) 472 register struct sioc_sg_req *req; 473{ 474 register struct mfc *rt; 475 int s; 476 477 s = splnet(); 478 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 479 splx(s); 480 if (rt != NULL) { 481 req->pktcnt = rt->mfc_pkt_cnt; 482 req->bytecnt = rt->mfc_byte_cnt; 483 req->wrong_if = rt->mfc_wrong_if; 484 } else 485 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 486 487 return 0; 488} 489 490/* 491 * returns the input and output packet and byte counts on the vif provided 492 */ 493static int 494get_vif_cnt(req) 495 register struct sioc_vif_req *req; 496{ 497 register vifi_t vifi = req->vifi; 498 499 if (vifi >= numvifs) return EINVAL; 500 501 req->icount = viftable[vifi].v_pkt_in; 502 req->ocount = viftable[vifi].v_pkt_out; 503 req->ibytes = viftable[vifi].v_bytes_in; 504 req->obytes = viftable[vifi].v_bytes_out; 505 506 return 0; 507} 508 509/* 510 * Enable multicast routing 511 */ 512static int 513ip_mrouter_init(so, m) 514 struct socket *so; 515 struct mbuf *m; 516{ 517 int *v; 518 519 if (mrtdebug) 520 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 521 so->so_type, so->so_proto->pr_protocol); 522 523 if (so->so_type != SOCK_RAW || 524 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP; 525 526 if (!m || (m->m_len != sizeof(int *))) 527 return ENOPROTOOPT; 528 529 v = mtod(m, int *); 530 if (*v != 1) 531 return ENOPROTOOPT; 532 533 if (ip_mrouter != NULL) return EADDRINUSE; 534 535 ip_mrouter = so; 536 537 bzero((caddr_t)mfctable, sizeof(mfctable)); 538 bzero((caddr_t)nexpire, sizeof(nexpire)); 539 540 pim_assert = 0; 541 542 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 543 544 if (mrtdebug) 545 log(LOG_DEBUG, "ip_mrouter_init\n"); 546 547 return 0; 548} 549 550/* 551 * Disable multicast routing 552 */ 553static int 554X_ip_mrouter_done() 555{ 556 vifi_t vifi; 557 int i; 558 struct ifnet *ifp; 559 struct ifreq ifr; 560 struct mbuf *mb_rt; 561 struct mbuf *m; 562 struct rtdetq *rte; 563 int s; 564 565 s = splnet(); 566 567 /* 568 * For each phyint in use, disable promiscuous reception of all IP 569 * multicasts. 570 */ 571 for (vifi = 0; vifi < numvifs; vifi++) { 572 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 573 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 574 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 575 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr 576 = INADDR_ANY; 577 ifp = viftable[vifi].v_ifp; 578 if_allmulti(ifp, 0); 579 } 580 } 581 bzero((caddr_t)tbftable, sizeof(tbftable)); 582 bzero((caddr_t)viftable, sizeof(viftable)); 583 numvifs = 0; 584 pim_assert = 0; 585 586 untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch); 587 588 /* 589 * Free all multicast forwarding cache entries. 590 */ 591 for (i = 0; i < MFCTBLSIZ; i++) { 592 mb_rt = mfctable[i]; 593 while (mb_rt) { 594 if (mb_rt->m_act != NULL) { 595 while (mb_rt->m_act) { 596 m = mb_rt->m_act; 597 mb_rt->m_act = m->m_act; 598 rte = mtod(m, struct rtdetq *); 599 m_freem(rte->m); 600 m_free(m); 601 } 602 } 603 mb_rt = m_free(mb_rt); 604 } 605 } 606 607 bzero((caddr_t)mfctable, sizeof(mfctable)); 608 609 /* 610 * Reset de-encapsulation cache 611 */ 612 last_encap_src = 0; 613 last_encap_vif = NULL; 614 have_encap_tunnel = 0; 615 616 ip_mrouter = NULL; 617 618 splx(s); 619 620 if (mrtdebug) 621 log(LOG_DEBUG, "ip_mrouter_done\n"); 622 623 return 0; 624} 625 626#ifndef MROUTE_LKM 627int (*ip_mrouter_done)(void) = X_ip_mrouter_done; 628#endif 629 630static int 631get_version(mb) 632 struct mbuf *mb; 633{ 634 int *v; 635 636 v = mtod(mb, int *); 637 638 *v = 0x0305; /* XXX !!!! */ 639 mb->m_len = sizeof(int); 640 641 return 0; 642} 643 644/* 645 * Set PIM assert processing global 646 */ 647static int 648set_assert(i) 649 int *i; 650{ 651 if ((*i != 1) && (*i != 0)) 652 return EINVAL; 653 654 pim_assert = *i; 655 656 return 0; 657} 658 659/* 660 * Get PIM assert processing global 661 */ 662static int 663get_assert(m) 664 struct mbuf *m; 665{ 666 int *i; 667 668 i = mtod(m, int *); 669 670 *i = pim_assert; 671 672 return 0; 673} 674 675/* 676 * Add a vif to the vif table 677 */ 678static int 679add_vif(vifcp) 680 register struct vifctl *vifcp; 681{ 682 register struct vif *vifp = viftable + vifcp->vifc_vifi; 683 static struct sockaddr_in sin = {sizeof sin, AF_INET}; 684 struct ifaddr *ifa; 685 struct ifnet *ifp; 686 struct ifreq ifr; 687 int error, s; 688 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 689 690 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL; 691 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE; 692 693 /* Find the interface with an address in AF_INET family */ 694 sin.sin_addr = vifcp->vifc_lcl_addr; 695 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 696 if (ifa == 0) return EADDRNOTAVAIL; 697 ifp = ifa->ifa_ifp; 698 699 if (vifcp->vifc_flags & VIFF_TUNNEL) { 700 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 701 /* 702 * An encapsulating tunnel is wanted. Tell ipip_input() to 703 * start paying attention to encapsulated packets. 704 */ 705 if (have_encap_tunnel == 0) { 706 have_encap_tunnel = 1; 707 for (s = 0; s < MAXVIFS; ++s) { 708 multicast_decap_if[s].if_name = "mdecap"; 709 multicast_decap_if[s].if_unit = s; 710 } 711 } 712 /* 713 * Set interface to fake encapsulator interface 714 */ 715 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 716 /* 717 * Prepare cached route entry 718 */ 719 bzero(&vifp->v_route, sizeof(vifp->v_route)); 720 } else { 721 log(LOG_ERR, "source routed tunnels not supported\n"); 722 return EOPNOTSUPP; 723 } 724 } else { 725 /* Make sure the interface supports multicast */ 726 if ((ifp->if_flags & IFF_MULTICAST) == 0) 727 return EOPNOTSUPP; 728 729 /* Enable promiscuous reception of all IP multicasts from the if */ 730 s = splnet(); 731 error = if_allmulti(ifp, 1); 732 splx(s); 733 if (error) 734 return error; 735 } 736 737 s = splnet(); 738 /* define parameters for the tbf structure */ 739 vifp->v_tbf = v_tbf; 740 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 741 vifp->v_tbf->tbf_n_tok = 0; 742 vifp->v_tbf->tbf_q_len = 0; 743 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 744 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 745 746 vifp->v_flags = vifcp->vifc_flags; 747 vifp->v_threshold = vifcp->vifc_threshold; 748 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 749 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 750 vifp->v_ifp = ifp; 751 /* scaling up here allows division by 1024 in critical code */ 752 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 753 vifp->v_rsvp_on = 0; 754 vifp->v_rsvpd = NULL; 755 /* initialize per vif pkt counters */ 756 vifp->v_pkt_in = 0; 757 vifp->v_pkt_out = 0; 758 vifp->v_bytes_in = 0; 759 vifp->v_bytes_out = 0; 760 splx(s); 761 762 /* Adjust numvifs up if the vifi is higher than numvifs */ 763 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 764 765 if (mrtdebug) 766 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 767 vifcp->vifc_vifi, 768 ntohl(vifcp->vifc_lcl_addr.s_addr), 769 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 770 ntohl(vifcp->vifc_rmt_addr.s_addr), 771 vifcp->vifc_threshold, 772 vifcp->vifc_rate_limit); 773 774 return 0; 775} 776 777/* 778 * Delete a vif from the vif table 779 */ 780static int 781del_vif(vifip) 782 vifi_t *vifip; 783{ 784 register struct vif *vifp = viftable + *vifip; 785 register vifi_t vifi; 786 register struct mbuf *m; 787 struct ifnet *ifp; 788 struct ifreq ifr; 789 int s; 790 791 if (*vifip >= numvifs) return EINVAL; 792 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL; 793 794 s = splnet(); 795 796 if (!(vifp->v_flags & VIFF_TUNNEL)) { 797 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 798 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 799 ifp = vifp->v_ifp; 800 if_allmulti(ifp, 0); 801 } 802 803 if (vifp == last_encap_vif) { 804 last_encap_vif = 0; 805 last_encap_src = 0; 806 } 807 808 /* 809 * Free packets queued at the interface 810 */ 811 while (vifp->v_tbf->tbf_q) { 812 m = vifp->v_tbf->tbf_q; 813 vifp->v_tbf->tbf_q = m->m_act; 814 m_freem(m); 815 } 816 817 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 818 bzero((caddr_t)vifp, sizeof (*vifp)); 819 820 /* Adjust numvifs down */ 821 for (vifi = numvifs; vifi > 0; vifi--) 822 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break; 823 numvifs = vifi; 824 825 splx(s); 826 827 if (mrtdebug) 828 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 829 830 return 0; 831} 832 833/* 834 * Add an mfc entry 835 */ 836static int 837add_mfc(mfccp) 838 struct mfcctl *mfccp; 839{ 840 struct mfc *rt; 841 register struct mbuf *mb_rt; 842 u_long hash; 843 struct mbuf *mb_ntry; 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 %x g %x p %x\n", 855 ntohl(mfccp->mfcc_origin.s_addr), 856 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 (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) { 873 874 rt = mtod(mb_rt, struct mfc *); 875 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 876 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 877 (mb_rt->m_act != NULL)) { 878 879 if (nstl++) 880 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %x\n", 881 "multiple kernel entries", 882 ntohl(mfccp->mfcc_origin.s_addr), 883 ntohl(mfccp->mfcc_mcastgrp.s_addr), 884 mfccp->mfcc_parent, mb_rt->m_act); 885 886 if (mrtdebug & DEBUG_MFC) 887 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %x\n", 888 ntohl(mfccp->mfcc_origin.s_addr), 889 ntohl(mfccp->mfcc_mcastgrp.s_addr), 890 mfccp->mfcc_parent, mb_rt->m_act); 891 892 rt->mfc_origin = mfccp->mfcc_origin; 893 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 894 rt->mfc_parent = mfccp->mfcc_parent; 895 for (i = 0; i < numvifs; i++) 896 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 897 /* initialize pkt counters per src-grp */ 898 rt->mfc_pkt_cnt = 0; 899 rt->mfc_byte_cnt = 0; 900 rt->mfc_wrong_if = 0; 901 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 902 903 rt->mfc_expire = 0; /* Don't clean this guy up */ 904 nexpire[hash]--; 905 906 /* free packets Qed at the end of this entry */ 907 while (mb_rt->m_act) { 908 mb_ntry = mb_rt->m_act; 909 rte = mtod(mb_ntry, struct rtdetq *); 910/* #ifdef RSVP_ISI */ 911 ip_mdq(rte->m, rte->ifp, rt, -1); 912/* #endif */ 913 mb_rt->m_act = mb_ntry->m_act; 914 m_freem(rte->m); 915#ifdef UPCALL_TIMING 916 collate(&(rte->t)); 917#endif /* UPCALL_TIMING */ 918 m_free(mb_ntry); 919 } 920 } 921 } 922 923 /* 924 * It is possible that an entry is being inserted without an upcall 925 */ 926 if (nstl == 0) { 927 if (mrtdebug & DEBUG_MFC) 928 log(LOG_DEBUG,"add_mfc no upcall h %d o %x g %x p %x\n", 929 hash, ntohl(mfccp->mfcc_origin.s_addr), 930 ntohl(mfccp->mfcc_mcastgrp.s_addr), 931 mfccp->mfcc_parent); 932 933 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 934 935 rt = mtod(mb_rt, struct mfc *); 936 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 937 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 938 939 rt->mfc_origin = mfccp->mfcc_origin; 940 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 941 rt->mfc_parent = mfccp->mfcc_parent; 942 for (i = 0; i < numvifs; i++) 943 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 944 /* initialize pkt counters per src-grp */ 945 rt->mfc_pkt_cnt = 0; 946 rt->mfc_byte_cnt = 0; 947 rt->mfc_wrong_if = 0; 948 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 949 if (rt->mfc_expire) 950 nexpire[hash]--; 951 rt->mfc_expire = 0; 952 } 953 } 954 if (mb_rt == NULL) { 955 /* no upcall, so make a new entry */ 956 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 957 if (mb_rt == NULL) { 958 splx(s); 959 return ENOBUFS; 960 } 961 962 rt = mtod(mb_rt, struct mfc *); 963 964 /* insert new entry at head of hash chain */ 965 rt->mfc_origin = mfccp->mfcc_origin; 966 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 967 rt->mfc_parent = mfccp->mfcc_parent; 968 for (i = 0; i < numvifs; i++) 969 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 970 /* initialize pkt counters per src-grp */ 971 rt->mfc_pkt_cnt = 0; 972 rt->mfc_byte_cnt = 0; 973 rt->mfc_wrong_if = 0; 974 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 975 rt->mfc_expire = 0; 976 977 /* link into table */ 978 mb_rt->m_next = mfctable[hash]; 979 mfctable[hash] = mb_rt; 980 mb_rt->m_act = NULL; 981 } 982 } 983 splx(s); 984 return 0; 985} 986 987#ifdef UPCALL_TIMING 988/* 989 * collect delay statistics on the upcalls 990 */ 991static void collate(t) 992register struct timeval *t; 993{ 994 register u_long d; 995 register struct timeval tp; 996 register u_long delta; 997 998 GET_TIME(tp); 999 1000 if (TV_LT(*t, tp)) 1001 { 1002 TV_DELTA(tp, *t, delta); 1003 1004 d = delta >> 10; 1005 if (d > 50) 1006 d = 50; 1007 1008 ++upcall_data[d]; 1009 } 1010} 1011#endif /* UPCALL_TIMING */ 1012 1013/* 1014 * Delete an mfc entry 1015 */ 1016static int 1017del_mfc(mfccp) 1018 struct mfcctl *mfccp; 1019{ 1020 struct in_addr origin; 1021 struct in_addr mcastgrp; 1022 struct mfc *rt; 1023 struct mbuf *mb_rt; 1024 struct mbuf **nptr; 1025 u_long hash; 1026 int s; 1027 1028 origin = mfccp->mfcc_origin; 1029 mcastgrp = mfccp->mfcc_mcastgrp; 1030 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1031 1032 if (mrtdebug & DEBUG_MFC) 1033 log(LOG_DEBUG,"del_mfc orig %x mcastgrp %x\n", 1034 ntohl(origin.s_addr), ntohl(mcastgrp.s_addr)); 1035 1036 s = splnet(); 1037 1038 nptr = &mfctable[hash]; 1039 while ((mb_rt = *nptr) != NULL) { 1040 rt = mtod(mb_rt, struct mfc *); 1041 if (origin.s_addr == rt->mfc_origin.s_addr && 1042 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1043 mb_rt->m_act == NULL) 1044 break; 1045 1046 nptr = &mb_rt->m_next; 1047 } 1048 if (mb_rt == NULL) { 1049 splx(s); 1050 return EADDRNOTAVAIL; 1051 } 1052 1053 MFREE(mb_rt, *nptr); 1054 1055 splx(s); 1056 1057 return 0; 1058} 1059 1060/* 1061 * Send a message to mrouted on the multicast routing socket 1062 */ 1063static int 1064socket_send(s, mm, src) 1065 struct socket *s; 1066 struct mbuf *mm; 1067 struct sockaddr_in *src; 1068{ 1069 if (s) { 1070 if (sbappendaddr(&s->so_rcv, 1071 (struct sockaddr *)src, 1072 mm, (struct mbuf *)0) != 0) { 1073 sorwakeup(s); 1074 return 0; 1075 } 1076 } 1077 m_freem(mm); 1078 return -1; 1079} 1080 1081/* 1082 * IP multicast forwarding function. This function assumes that the packet 1083 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1084 * pointed to by "ifp", and the packet is to be relayed to other networks 1085 * that have members of the packet's destination IP multicast group. 1086 * 1087 * The packet is returned unscathed to the caller, unless it is 1088 * erroneous, in which case a non-zero return value tells the caller to 1089 * discard it. 1090 */ 1091 1092#define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1093#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1094 1095static int 1096X_ip_mforward(ip, ifp, m, imo) 1097 register struct ip *ip; 1098 struct ifnet *ifp; 1099 struct mbuf *m; 1100 struct ip_moptions *imo; 1101{ 1102 register struct mfc *rt; 1103 register u_char *ipoptions; 1104 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1105 static int srctun = 0; 1106 register struct mbuf *mm; 1107 int s; 1108 vifi_t vifi; 1109 struct vif *vifp; 1110 1111 if (mrtdebug & DEBUG_FORWARD) 1112 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %x\n", 1113 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1114 1115 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1116 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1117 /* 1118 * Packet arrived via a physical interface or 1119 * an encapsulated tunnel. 1120 */ 1121 } else { 1122 /* 1123 * Packet arrived through a source-route tunnel. 1124 * Source-route tunnels are no longer supported. 1125 */ 1126 if ((srctun++ % 1000) == 0) 1127 log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n", 1128 ntohl(ip->ip_src.s_addr)); 1129 1130 return 1; 1131 } 1132 1133 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1134 if (ip->ip_ttl < 255) 1135 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1136 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1137 vifp = viftable + vifi; 1138 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1139 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi, 1140 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1141 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1142 } 1143 return (ip_mdq(m, ifp, NULL, vifi)); 1144 } 1145 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1146 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1147 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr)); 1148 if(!imo) 1149 printf("In fact, no options were specified at all\n"); 1150 } 1151 1152 /* 1153 * Don't forward a packet with time-to-live of zero or one, 1154 * or a packet destined to a local-only group. 1155 */ 1156 if (ip->ip_ttl <= 1 || 1157 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1158 return 0; 1159 1160 /* 1161 * Determine forwarding vifs from the forwarding cache table 1162 */ 1163 s = splnet(); 1164 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt); 1165 1166 /* Entry exists, so forward if necessary */ 1167 if (rt != NULL) { 1168 splx(s); 1169 return (ip_mdq(m, ifp, rt, -1)); 1170 } else { 1171 /* 1172 * If we don't have a route for packet's origin, 1173 * Make a copy of the packet & 1174 * send message to routing daemon 1175 */ 1176 1177 register struct mbuf *mb_rt; 1178 register struct mbuf *mb_ntry; 1179 register struct mbuf *mb0; 1180 register struct rtdetq *rte; 1181 register struct mbuf *rte_m; 1182 register u_long hash; 1183 register int npkts; 1184 int hlen = ip->ip_hl << 2; 1185#ifdef UPCALL_TIMING 1186 struct timeval tp; 1187 1188 GET_TIME(tp); 1189#endif 1190 1191 mrtstat.mrts_no_route++; 1192 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1193 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1194 ntohl(ip->ip_src.s_addr), 1195 ntohl(ip->ip_dst.s_addr)); 1196 1197 /* 1198 * Allocate mbufs early so that we don't do extra work if we are 1199 * just going to fail anyway. Make sure to pullup the header so 1200 * that other people can't step on it. 1201 */ 1202 MGET(mb_ntry, M_DONTWAIT, MT_DATA); 1203 if (mb_ntry == NULL) { 1204 splx(s); 1205 return ENOBUFS; 1206 } 1207 mb0 = m_copy(m, 0, M_COPYALL); 1208 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1209 mb0 = m_pullup(mb0, hlen); 1210 if (mb0 == NULL) { 1211 m_free(mb_ntry); 1212 splx(s); 1213 return ENOBUFS; 1214 } 1215 1216 /* is there an upcall waiting for this packet? */ 1217 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1218 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 1219 rt = mtod(mb_rt, struct mfc *); 1220 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1221 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1222 (mb_rt->m_act != NULL)) 1223 break; 1224 } 1225 1226 if (mb_rt == NULL) { 1227 int i; 1228 struct igmpmsg *im; 1229 1230 /* no upcall, so make a new entry */ 1231 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 1232 if (mb_rt == NULL) { 1233 m_free(mb_ntry); 1234 m_freem(mb0); 1235 splx(s); 1236 return ENOBUFS; 1237 } 1238 /* Make a copy of the header to send to the user level process */ 1239 mm = m_copy(mb0, 0, hlen); 1240 if (mm == NULL) { 1241 m_free(mb_ntry); 1242 m_freem(mb0); 1243 m_free(mb_rt); 1244 splx(s); 1245 return ENOBUFS; 1246 } 1247 1248 /* 1249 * Send message to routing daemon to install 1250 * a route into the kernel table 1251 */ 1252 k_igmpsrc.sin_addr = ip->ip_src; 1253 1254 im = mtod(mm, struct igmpmsg *); 1255 im->im_msgtype = IGMPMSG_NOCACHE; 1256 im->im_mbz = 0; 1257 1258 mrtstat.mrts_upcalls++; 1259 1260 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1261 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1262 ++mrtstat.mrts_upq_sockfull; 1263 m_free(mb_ntry); 1264 m_freem(mb0); 1265 m_free(mb_rt); 1266 splx(s); 1267 return ENOBUFS; 1268 } 1269 1270 rt = mtod(mb_rt, struct mfc *); 1271 1272 /* insert new entry at head of hash chain */ 1273 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1274 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1275 rt->mfc_expire = UPCALL_EXPIRE; 1276 nexpire[hash]++; 1277 for (i = 0; i < numvifs; i++) 1278 rt->mfc_ttls[i] = 0; 1279 rt->mfc_parent = -1; 1280 1281 /* link into table */ 1282 mb_rt->m_next = mfctable[hash]; 1283 mfctable[hash] = mb_rt; 1284 mb_rt->m_act = NULL; 1285 1286 rte_m = mb_rt; 1287 } else { 1288 /* determine if q has overflowed */ 1289 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act) 1290 npkts++; 1291 1292 if (npkts > MAX_UPQ) { 1293 mrtstat.mrts_upq_ovflw++; 1294 m_free(mb_ntry); 1295 m_freem(mb0); 1296 splx(s); 1297 return 0; 1298 } 1299 } 1300 1301 mb_ntry->m_act = NULL; 1302 rte = mtod(mb_ntry, struct rtdetq *); 1303 1304 rte->m = mb0; 1305 rte->ifp = ifp; 1306#ifdef UPCALL_TIMING 1307 rte->t = tp; 1308#endif 1309 1310 /* Add this entry to the end of the queue */ 1311 rte_m->m_act = mb_ntry; 1312 1313 splx(s); 1314 1315 return 0; 1316 } 1317} 1318 1319#ifndef MROUTE_LKM 1320int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 1321 struct ip_moptions *) = X_ip_mforward; 1322#endif 1323 1324/* 1325 * Clean up the cache entry if upcall is not serviced 1326 */ 1327static void 1328expire_upcalls(void *unused) 1329{ 1330 struct mbuf *mb_rt, *m, **nptr; 1331 struct rtdetq *rte; 1332 struct mfc *mfc; 1333 int i; 1334 int s; 1335 1336 s = splnet(); 1337 for (i = 0; i < MFCTBLSIZ; i++) { 1338 if (nexpire[i] == 0) 1339 continue; 1340 nptr = &mfctable[i]; 1341 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) { 1342 mfc = mtod(mb_rt, struct mfc *); 1343 1344 /* 1345 * Skip real cache entries 1346 * Make sure it wasn't marked to not expire (shouldn't happen) 1347 * If it expires now 1348 */ 1349 if (mb_rt->m_act != NULL && 1350 mfc->mfc_expire != 0 && 1351 --mfc->mfc_expire == 0) { 1352 if (mrtdebug & DEBUG_EXPIRE) 1353 log(LOG_DEBUG, "expire_upcalls: expiring (%x %x)\n", 1354 ntohl(mfc->mfc_origin.s_addr), 1355 ntohl(mfc->mfc_mcastgrp.s_addr)); 1356 /* 1357 * drop all the packets 1358 * free the mbuf with the pkt, if, timing info 1359 */ 1360 while (mb_rt->m_act) { 1361 m = mb_rt->m_act; 1362 mb_rt->m_act = m->m_act; 1363 1364 rte = mtod(m, struct rtdetq *); 1365 m_freem(rte->m); 1366 m_free(m); 1367 } 1368 ++mrtstat.mrts_cache_cleanups; 1369 nexpire[i]--; 1370 1371 MFREE(mb_rt, *nptr); 1372 } else { 1373 nptr = &mb_rt->m_next; 1374 } 1375 } 1376 } 1377 splx(s); 1378 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 1379} 1380 1381/* 1382 * Packet forwarding routine once entry in the cache is made 1383 */ 1384static int 1385ip_mdq(m, ifp, rt, xmt_vif) 1386 register struct mbuf *m; 1387 register struct ifnet *ifp; 1388 register struct mfc *rt; 1389 register vifi_t xmt_vif; 1390{ 1391 register struct ip *ip = mtod(m, struct ip *); 1392 register vifi_t vifi; 1393 register struct vif *vifp; 1394 register int plen = ntohs(ip->ip_len); 1395 1396/* 1397 * Macro to send packet on vif. Since RSVP packets don't get counted on 1398 * input, they shouldn't get counted on output, so statistics keeping is 1399 * seperate. 1400 */ 1401#define MC_SEND(ip,vifp,m) { \ 1402 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1403 encap_send((ip), (vifp), (m)); \ 1404 else \ 1405 phyint_send((ip), (vifp), (m)); \ 1406} 1407 1408 /* 1409 * If xmt_vif is not -1, send on only the requested vif. 1410 * 1411 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1412 */ 1413 if (xmt_vif < numvifs) { 1414 MC_SEND(ip, viftable + xmt_vif, m); 1415 return 1; 1416 } 1417 1418 /* 1419 * Don't forward if it didn't arrive from the parent vif for its origin. 1420 */ 1421 vifi = rt->mfc_parent; 1422 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1423 /* came in the wrong interface */ 1424 if (mrtdebug & DEBUG_FORWARD) 1425 log(LOG_DEBUG, "wrong if: ifp %x vifi %d vififp %x\n", 1426 ifp, vifi, viftable[vifi].v_ifp); 1427 ++mrtstat.mrts_wrong_if; 1428 ++rt->mfc_wrong_if; 1429 /* 1430 * If we are doing PIM assert processing, and we are forwarding 1431 * packets on this interface, and it is a broadcast medium 1432 * interface (and not a tunnel), send a message to the routing daemon. 1433 */ 1434 if (pim_assert && rt->mfc_ttls[vifi] && 1435 (ifp->if_flags & IFF_BROADCAST) && 1436 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1437 struct sockaddr_in k_igmpsrc; 1438 struct mbuf *mm; 1439 struct igmpmsg *im; 1440 int hlen = ip->ip_hl << 2; 1441 struct timeval now; 1442 register u_long delta; 1443 1444 GET_TIME(now); 1445 1446 TV_DELTA(rt->mfc_last_assert, now, delta); 1447 1448 if (delta > ASSERT_MSG_TIME) { 1449 mm = m_copy(m, 0, hlen); 1450 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1451 mm = m_pullup(mm, hlen); 1452 if (mm == NULL) { 1453 return ENOBUFS; 1454 } 1455 1456 rt->mfc_last_assert = now; 1457 1458 im = mtod(mm, struct igmpmsg *); 1459 im->im_msgtype = IGMPMSG_WRONGVIF; 1460 im->im_mbz = 0; 1461 im->im_vif = vifi; 1462 1463 k_igmpsrc.sin_addr = im->im_src; 1464 1465 socket_send(ip_mrouter, mm, &k_igmpsrc); 1466 } 1467 } 1468 return 0; 1469 } 1470 1471 /* If I sourced this packet, it counts as output, else it was input. */ 1472 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1473 viftable[vifi].v_pkt_out++; 1474 viftable[vifi].v_bytes_out += plen; 1475 } else { 1476 viftable[vifi].v_pkt_in++; 1477 viftable[vifi].v_bytes_in += plen; 1478 } 1479 rt->mfc_pkt_cnt++; 1480 rt->mfc_byte_cnt += plen; 1481 1482 /* 1483 * For each vif, decide if a copy of the packet should be forwarded. 1484 * Forward if: 1485 * - the ttl exceeds the vif's threshold 1486 * - there are group members downstream on interface 1487 */ 1488 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1489 if ((rt->mfc_ttls[vifi] > 0) && 1490 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1491 vifp->v_pkt_out++; 1492 vifp->v_bytes_out += plen; 1493 MC_SEND(ip, vifp, m); 1494 } 1495 1496 return 0; 1497} 1498 1499/* 1500 * check if a vif number is legal/ok. This is used by ip_output, to export 1501 * numvifs there, 1502 */ 1503static int 1504X_legal_vif_num(vif) 1505 int vif; 1506{ 1507 if (vif >= 0 && vif < numvifs) 1508 return(1); 1509 else 1510 return(0); 1511} 1512 1513#ifndef MROUTE_LKM 1514int (*legal_vif_num)(int) = X_legal_vif_num; 1515#endif 1516 1517/* 1518 * Return the local address used by this vif 1519 */ 1520static u_long 1521X_ip_mcast_src(vifi) 1522 int vifi; 1523{ 1524 if (vifi >= 0 && vifi < numvifs) 1525 return viftable[vifi].v_lcl_addr.s_addr; 1526 else 1527 return INADDR_ANY; 1528} 1529 1530#ifndef MROUTE_LKM 1531u_long (*ip_mcast_src)(int) = X_ip_mcast_src; 1532#endif 1533 1534static void 1535phyint_send(ip, vifp, m) 1536 struct ip *ip; 1537 struct vif *vifp; 1538 struct mbuf *m; 1539{ 1540 register struct mbuf *mb_copy; 1541 register int hlen = ip->ip_hl << 2; 1542 1543 /* 1544 * Make a new reference to the packet; make sure that 1545 * the IP header is actually copied, not just referenced, 1546 * so that ip_output() only scribbles on the copy. 1547 */ 1548 mb_copy = m_copy(m, 0, M_COPYALL); 1549 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1550 mb_copy = m_pullup(mb_copy, hlen); 1551 if (mb_copy == NULL) 1552 return; 1553 1554 if (vifp->v_rate_limit <= 0) 1555 tbf_send_packet(vifp, mb_copy); 1556 else 1557 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1558} 1559 1560static void 1561encap_send(ip, vifp, m) 1562 register struct ip *ip; 1563 register struct vif *vifp; 1564 register struct mbuf *m; 1565{ 1566 register struct mbuf *mb_copy; 1567 register struct ip *ip_copy; 1568 register int i, len = ip->ip_len; 1569 1570 /* 1571 * copy the old packet & pullup it's IP header into the 1572 * new mbuf so we can modify it. Try to fill the new 1573 * mbuf since if we don't the ethernet driver will. 1574 */ 1575 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER); 1576 if (mb_copy == NULL) 1577 return; 1578 mb_copy->m_data += max_linkhdr; 1579 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1580 1581 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1582 m_freem(mb_copy); 1583 return; 1584 } 1585 i = MHLEN - M_LEADINGSPACE(mb_copy); 1586 if (i > len) 1587 i = len; 1588 mb_copy = m_pullup(mb_copy, i); 1589 if (mb_copy == NULL) 1590 return; 1591 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1592 1593 /* 1594 * fill in the encapsulating IP header. 1595 */ 1596 ip_copy = mtod(mb_copy, struct ip *); 1597 *ip_copy = multicast_encap_iphdr; 1598 ip_copy->ip_id = htons(ip_id++); 1599 ip_copy->ip_len += len; 1600 ip_copy->ip_src = vifp->v_lcl_addr; 1601 ip_copy->ip_dst = vifp->v_rmt_addr; 1602 1603 /* 1604 * turn the encapsulated IP header back into a valid one. 1605 */ 1606 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1607 --ip->ip_ttl; 1608 HTONS(ip->ip_len); 1609 HTONS(ip->ip_off); 1610 ip->ip_sum = 0; 1611 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1612 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1613 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1614 1615 if (vifp->v_rate_limit <= 0) 1616 tbf_send_packet(vifp, mb_copy); 1617 else 1618 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1619} 1620 1621/* 1622 * De-encapsulate a packet and feed it back through ip input (this 1623 * routine is called whenever IP gets a packet with proto type 1624 * ENCAP_PROTO and a local destination address). 1625 */ 1626void 1627#ifdef MROUTE_LKM 1628X_ipip_input(m, iphlen) 1629#else 1630ipip_input(m, iphlen) 1631#endif 1632 register struct mbuf *m; 1633 int iphlen; 1634{ 1635 struct ifnet *ifp = m->m_pkthdr.rcvif; 1636 register struct ip *ip = mtod(m, struct ip *); 1637 register int hlen = ip->ip_hl << 2; 1638 register int s; 1639 register struct ifqueue *ifq; 1640 register struct vif *vifp; 1641 1642 if (!have_encap_tunnel) { 1643 rip_input(m, iphlen); 1644 return; 1645 } 1646 /* 1647 * dump the packet if it's not to a multicast destination or if 1648 * we don't have an encapsulating tunnel with the source. 1649 * Note: This code assumes that the remote site IP address 1650 * uniquely identifies the tunnel (i.e., that this site has 1651 * at most one tunnel with the remote site). 1652 */ 1653 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) { 1654 ++mrtstat.mrts_bad_tunnel; 1655 m_freem(m); 1656 return; 1657 } 1658 if (ip->ip_src.s_addr != last_encap_src) { 1659 register struct vif *vife; 1660 1661 vifp = viftable; 1662 vife = vifp + numvifs; 1663 last_encap_src = ip->ip_src.s_addr; 1664 last_encap_vif = 0; 1665 for ( ; vifp < vife; ++vifp) 1666 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1667 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1668 == VIFF_TUNNEL) 1669 last_encap_vif = vifp; 1670 break; 1671 } 1672 } 1673 if ((vifp = last_encap_vif) == 0) { 1674 last_encap_src = 0; 1675 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1676 m_freem(m); 1677 if (mrtdebug) 1678 log(LOG_DEBUG, "ip_mforward: no tunnel with %x\n", 1679 ntohl(ip->ip_src.s_addr)); 1680 return; 1681 } 1682 ifp = vifp->v_ifp; 1683 1684 if (hlen > IP_HDR_LEN) 1685 ip_stripoptions(m, (struct mbuf *) 0); 1686 m->m_data += IP_HDR_LEN; 1687 m->m_len -= IP_HDR_LEN; 1688 m->m_pkthdr.len -= IP_HDR_LEN; 1689 m->m_pkthdr.rcvif = ifp; 1690 1691 ifq = &ipintrq; 1692 s = splimp(); 1693 if (IF_QFULL(ifq)) { 1694 IF_DROP(ifq); 1695 m_freem(m); 1696 } else { 1697 IF_ENQUEUE(ifq, m); 1698 /* 1699 * normally we would need a "schednetisr(NETISR_IP)" 1700 * here but we were called by ip_input and it is going 1701 * to loop back & try to dequeue the packet we just 1702 * queued as soon as we return so we avoid the 1703 * unnecessary software interrrupt. 1704 */ 1705 } 1706 splx(s); 1707} 1708 1709/* 1710 * Token bucket filter module 1711 */ 1712 1713static void 1714tbf_control(vifp, m, ip, p_len) 1715 register struct vif *vifp; 1716 register struct mbuf *m; 1717 register struct ip *ip; 1718 register u_long p_len; 1719{ 1720 register struct tbf *t = vifp->v_tbf; 1721 1722 if (p_len > MAX_BKT_SIZE) { 1723 /* drop if packet is too large */ 1724 mrtstat.mrts_pkt2large++; 1725 m_freem(m); 1726 return; 1727 } 1728 1729 tbf_update_tokens(vifp); 1730 1731 /* if there are enough tokens, 1732 * and the queue is empty, 1733 * send this packet out 1734 */ 1735 1736 if (t->tbf_q_len == 0) { 1737 /* queue empty, send packet if enough tokens */ 1738 if (p_len <= t->tbf_n_tok) { 1739 t->tbf_n_tok -= p_len; 1740 tbf_send_packet(vifp, m); 1741 } else { 1742 /* queue packet and timeout till later */ 1743 tbf_queue(vifp, m); 1744 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1745 } 1746 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1747 /* finite queue length, so queue pkts and process queue */ 1748 tbf_queue(vifp, m); 1749 tbf_process_q(vifp); 1750 } else { 1751 /* queue length too much, try to dq and queue and process */ 1752 if (!tbf_dq_sel(vifp, ip)) { 1753 mrtstat.mrts_q_overflow++; 1754 m_freem(m); 1755 return; 1756 } else { 1757 tbf_queue(vifp, m); 1758 tbf_process_q(vifp); 1759 } 1760 } 1761 return; 1762} 1763 1764/* 1765 * adds a packet to the queue at the interface 1766 */ 1767static void 1768tbf_queue(vifp, m) 1769 register struct vif *vifp; 1770 register struct mbuf *m; 1771{ 1772 register int s = splnet(); 1773 register struct tbf *t = vifp->v_tbf; 1774 1775 if (t->tbf_t == NULL) { 1776 /* Queue was empty */ 1777 t->tbf_q = m; 1778 } else { 1779 /* Insert at tail */ 1780 t->tbf_t->m_act = m; 1781 } 1782 1783 /* Set new tail pointer */ 1784 t->tbf_t = m; 1785 1786#ifdef DIAGNOSTIC 1787 /* Make sure we didn't get fed a bogus mbuf */ 1788 if (m->m_act) 1789 panic("tbf_queue: m_act"); 1790#endif 1791 m->m_act = NULL; 1792 1793 t->tbf_q_len++; 1794 1795 splx(s); 1796} 1797 1798 1799/* 1800 * processes the queue at the interface 1801 */ 1802static void 1803tbf_process_q(vifp) 1804 register struct vif *vifp; 1805{ 1806 register struct mbuf *m; 1807 register int len; 1808 register int s = splnet(); 1809 register struct tbf *t = vifp->v_tbf; 1810 1811 /* loop through the queue at the interface and send as many packets 1812 * as possible 1813 */ 1814 while (t->tbf_q_len > 0) { 1815 m = t->tbf_q; 1816 1817 len = mtod(m, struct ip *)->ip_len; 1818 1819 /* determine if the packet can be sent */ 1820 if (len <= t->tbf_n_tok) { 1821 /* if so, 1822 * reduce no of tokens, dequeue the packet, 1823 * send the packet. 1824 */ 1825 t->tbf_n_tok -= len; 1826 1827 t->tbf_q = m->m_act; 1828 if (--t->tbf_q_len == 0) 1829 t->tbf_t = NULL; 1830 1831 m->m_act = NULL; 1832 tbf_send_packet(vifp, m); 1833 1834 } else break; 1835 } 1836 splx(s); 1837} 1838 1839static void 1840tbf_reprocess_q(xvifp) 1841 void *xvifp; 1842{ 1843 register struct vif *vifp = xvifp; 1844 if (ip_mrouter == NULL) 1845 return; 1846 1847 tbf_update_tokens(vifp); 1848 1849 tbf_process_q(vifp); 1850 1851 if (vifp->v_tbf->tbf_q_len) 1852 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1853} 1854 1855/* function that will selectively discard a member of the queue 1856 * based on the precedence value and the priority 1857 */ 1858static int 1859tbf_dq_sel(vifp, ip) 1860 register struct vif *vifp; 1861 register struct ip *ip; 1862{ 1863 register int s = splnet(); 1864 register u_int p; 1865 register struct mbuf *m, *last; 1866 register struct mbuf **np; 1867 register struct tbf *t = vifp->v_tbf; 1868 1869 p = priority(vifp, ip); 1870 1871 np = &t->tbf_q; 1872 last = NULL; 1873 while ((m = *np) != NULL) { 1874 if (p > priority(vifp, mtod(m, struct ip *))) { 1875 *np = m->m_act; 1876 /* If we're removing the last packet, fix the tail pointer */ 1877 if (m == t->tbf_t) 1878 t->tbf_t = last; 1879 m_freem(m); 1880 /* it's impossible for the queue to be empty, but 1881 * we check anyway. */ 1882 if (--t->tbf_q_len == 0) 1883 t->tbf_t = NULL; 1884 splx(s); 1885 mrtstat.mrts_drop_sel++; 1886 return(1); 1887 } 1888 np = &m->m_act; 1889 last = m; 1890 } 1891 splx(s); 1892 return(0); 1893} 1894 1895static void 1896tbf_send_packet(vifp, m) 1897 register struct vif *vifp; 1898 register struct mbuf *m; 1899{ 1900 struct ip_moptions imo; 1901 int error; 1902 static struct route ro; 1903 int s = splnet(); 1904 1905 if (vifp->v_flags & VIFF_TUNNEL) { 1906 /* If tunnel options */ 1907 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1908 IP_FORWARDING, (struct ip_moptions *)0); 1909 } else { 1910 imo.imo_multicast_ifp = vifp->v_ifp; 1911 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1912 imo.imo_multicast_loop = 1; 1913 imo.imo_multicast_vif = -1; 1914 1915 /* 1916 * Re-entrancy should not be a problem here, because 1917 * the packets that we send out and are looped back at us 1918 * should get rejected because they appear to come from 1919 * the loopback interface, thus preventing looping. 1920 */ 1921 error = ip_output(m, (struct mbuf *)0, &ro, 1922 IP_FORWARDING, &imo); 1923 1924 if (mrtdebug & DEBUG_XMIT) 1925 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1926 vifp - viftable, error); 1927 } 1928 splx(s); 1929} 1930 1931/* determine the current time and then 1932 * the elapsed time (between the last time and time now) 1933 * in milliseconds & update the no. of tokens in the bucket 1934 */ 1935static void 1936tbf_update_tokens(vifp) 1937 register struct vif *vifp; 1938{ 1939 struct timeval tp; 1940 register u_long tm; 1941 register int s = splnet(); 1942 register struct tbf *t = vifp->v_tbf; 1943 1944 GET_TIME(tp); 1945 1946 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1947 1948 /* 1949 * This formula is actually 1950 * "time in seconds" * "bytes/second". 1951 * 1952 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1953 * 1954 * The (1000/1024) was introduced in add_vif to optimize 1955 * this divide into a shift. 1956 */ 1957 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1958 t->tbf_last_pkt_t = tp; 1959 1960 if (t->tbf_n_tok > MAX_BKT_SIZE) 1961 t->tbf_n_tok = MAX_BKT_SIZE; 1962 1963 splx(s); 1964} 1965 1966static int 1967priority(vifp, ip) 1968 register struct vif *vifp; 1969 register struct ip *ip; 1970{ 1971 register int prio; 1972 1973 /* temporary hack; may add general packet classifier some day */ 1974 1975 /* 1976 * The UDP port space is divided up into four priority ranges: 1977 * [0, 16384) : unclassified - lowest priority 1978 * [16384, 32768) : audio - highest priority 1979 * [32768, 49152) : whiteboard - medium priority 1980 * [49152, 65536) : video - low priority 1981 */ 1982 if (ip->ip_p == IPPROTO_UDP) { 1983 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1984 switch (ntohs(udp->uh_dport) & 0xc000) { 1985 case 0x4000: 1986 prio = 70; 1987 break; 1988 case 0x8000: 1989 prio = 60; 1990 break; 1991 case 0xc000: 1992 prio = 55; 1993 break; 1994 default: 1995 prio = 50; 1996 break; 1997 } 1998 if (tbfdebug > 1) 1999 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio); 2000 } else { 2001 prio = 50; 2002 } 2003 return prio; 2004} 2005 2006/* 2007 * End of token bucket filter modifications 2008 */ 2009 2010int 2011ip_rsvp_vif_init(so, m) 2012 struct socket *so; 2013 struct mbuf *m; 2014{ 2015 int i; 2016 register int s; 2017 2018 if (rsvpdebug) 2019 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2020 so->so_type, so->so_proto->pr_protocol); 2021 2022 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2023 return EOPNOTSUPP; 2024 2025 /* Check mbuf. */ 2026 if (m == NULL || m->m_len != sizeof(int)) { 2027 return EINVAL; 2028 } 2029 i = *(mtod(m, int *)); 2030 2031 if (rsvpdebug) 2032 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on); 2033 2034 s = splnet(); 2035 2036 /* Check vif. */ 2037 if (!legal_vif_num(i)) { 2038 splx(s); 2039 return EADDRNOTAVAIL; 2040 } 2041 2042 /* Check if socket is available. */ 2043 if (viftable[i].v_rsvpd != NULL) { 2044 splx(s); 2045 return EADDRINUSE; 2046 } 2047 2048 viftable[i].v_rsvpd = so; 2049 /* This may seem silly, but we need to be sure we don't over-increment 2050 * the RSVP counter, in case something slips up. 2051 */ 2052 if (!viftable[i].v_rsvp_on) { 2053 viftable[i].v_rsvp_on = 1; 2054 rsvp_on++; 2055 } 2056 2057 splx(s); 2058 return 0; 2059} 2060 2061int 2062ip_rsvp_vif_done(so, m) 2063 struct socket *so; 2064 struct mbuf *m; 2065{ 2066 int i; 2067 register int s; 2068 2069 if (rsvpdebug) 2070 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2071 so->so_type, so->so_proto->pr_protocol); 2072 2073 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2074 return EOPNOTSUPP; 2075 2076 /* Check mbuf. */ 2077 if (m == NULL || m->m_len != sizeof(int)) { 2078 return EINVAL; 2079 } 2080 i = *(mtod(m, int *)); 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 /* This may seem silly, but we need to be sure we don't over-decrement 2096 * the RSVP counter, in case something slips up. 2097 */ 2098 if (viftable[i].v_rsvp_on) { 2099 viftable[i].v_rsvp_on = 0; 2100 rsvp_on--; 2101 } 2102 2103 splx(s); 2104 return 0; 2105} 2106 2107void 2108ip_rsvp_force_done(so) 2109 struct socket *so; 2110{ 2111 int vifi; 2112 register int s; 2113 2114 /* Don't bother if it is not the right type of socket. */ 2115 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2116 return; 2117 2118 s = splnet(); 2119 2120 /* The socket may be attached to more than one vif...this 2121 * is perfectly legal. 2122 */ 2123 for (vifi = 0; vifi < numvifs; vifi++) { 2124 if (viftable[vifi].v_rsvpd == so) { 2125 viftable[vifi].v_rsvpd = NULL; 2126 /* This may seem silly, but we need to be sure we don't 2127 * over-decrement the RSVP counter, in case something slips up. 2128 */ 2129 if (viftable[vifi].v_rsvp_on) { 2130 viftable[vifi].v_rsvp_on = 0; 2131 rsvp_on--; 2132 } 2133 } 2134 } 2135 2136 splx(s); 2137 return; 2138} 2139 2140void 2141rsvp_input(m, iphlen) 2142 struct mbuf *m; 2143 int iphlen; 2144{ 2145 int vifi; 2146 register struct ip *ip = mtod(m, struct ip *); 2147 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2148 register int s; 2149 struct ifnet *ifp; 2150 2151 if (rsvpdebug) 2152 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2153 2154 /* Can still get packets with rsvp_on = 0 if there is a local member 2155 * of the group to which the RSVP packet is addressed. But in this 2156 * case we want to throw the packet away. 2157 */ 2158 if (!rsvp_on) { 2159 m_freem(m); 2160 return; 2161 } 2162 2163 /* If the old-style non-vif-associated socket is set, then use 2164 * it and ignore the new ones. 2165 */ 2166 if (ip_rsvpd != NULL) { 2167 if (rsvpdebug) 2168 printf("rsvp_input: Sending packet up old-style socket\n"); 2169 rip_input(m, iphlen); 2170 return; 2171 } 2172 2173 s = splnet(); 2174 2175 if (rsvpdebug) 2176 printf("rsvp_input: check vifs\n"); 2177 2178#ifdef DIAGNOSTIC 2179 if (!(m->m_flags & M_PKTHDR)) 2180 panic("rsvp_input no hdr"); 2181#endif 2182 2183 ifp = m->m_pkthdr.rcvif; 2184 /* Find which vif the packet arrived on. */ 2185 for (vifi = 0; vifi < numvifs; vifi++) { 2186 if (viftable[vifi].v_ifp == ifp) 2187 break; 2188 } 2189 2190 if (vifi == numvifs) { 2191 /* Can't find vif packet arrived on. Drop packet. */ 2192 if (rsvpdebug) 2193 printf("rsvp_input: Can't find vif for packet...dropping it.\n"); 2194 m_freem(m); 2195 splx(s); 2196 return; 2197 } 2198 2199 if (rsvpdebug) 2200 printf("rsvp_input: check socket\n"); 2201 2202 if (viftable[vifi].v_rsvpd == NULL) { 2203 /* drop packet, since there is no specific socket for this 2204 * interface */ 2205 if (rsvpdebug) 2206 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2207 m_freem(m); 2208 splx(s); 2209 return; 2210 } 2211 rsvp_src.sin_addr = ip->ip_src; 2212 2213 if (rsvpdebug && m) 2214 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2215 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2216 2217 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2218 if (rsvpdebug) 2219 printf("rsvp_input: Failed to append to socket\n"); 2220 else 2221 if (rsvpdebug) 2222 printf("rsvp_input: send packet up\n"); 2223 2224 splx(s); 2225} 2226 2227#ifdef MROUTE_LKM 2228#include <sys/conf.h> 2229#include <sys/exec.h> 2230#include <sys/sysent.h> 2231#include <sys/lkm.h> 2232 2233MOD_MISC("ip_mroute_mod") 2234 2235static int 2236ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd) 2237{ 2238 int i; 2239 struct lkm_misc *args = lkmtp->private.lkm_misc; 2240 int err = 0; 2241 2242 switch(cmd) { 2243 static int (*old_ip_mrouter_cmd)(); 2244 static int (*old_ip_mrouter_done)(); 2245 static int (*old_ip_mforward)(); 2246 static int (*old_mrt_ioctl)(); 2247 static void (*old_proto4_input)(); 2248 static int (*old_legal_vif_num)(); 2249 extern struct protosw inetsw[]; 2250 2251 case LKM_E_LOAD: 2252 if(lkmexists(lkmtp) || ip_mrtproto) 2253 return(EEXIST); 2254 old_ip_mrouter_cmd = ip_mrouter_cmd; 2255 ip_mrouter_cmd = X_ip_mrouter_cmd; 2256 old_ip_mrouter_done = ip_mrouter_done; 2257 ip_mrouter_done = X_ip_mrouter_done; 2258 old_ip_mforward = ip_mforward; 2259 ip_mforward = X_ip_mforward; 2260 old_mrt_ioctl = mrt_ioctl; 2261 mrt_ioctl = X_mrt_ioctl; 2262 old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input; 2263 inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input; 2264 old_legal_vif_num = legal_vif_num; 2265 legal_vif_num = X_legal_vif_num; 2266 ip_mrtproto = IGMP_DVMRP; 2267 2268 printf("\nIP multicast routing loaded\n"); 2269 break; 2270 2271 case LKM_E_UNLOAD: 2272 if (ip_mrouter) 2273 return EINVAL; 2274 2275 ip_mrouter_cmd = old_ip_mrouter_cmd; 2276 ip_mrouter_done = old_ip_mrouter_done; 2277 ip_mforward = old_ip_mforward; 2278 mrt_ioctl = old_mrt_ioctl; 2279 inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input; 2280 legal_vif_num = old_legal_vif_num; 2281 ip_mrtproto = 0; 2282 break; 2283 2284 default: 2285 err = EINVAL; 2286 break; 2287 } 2288 2289 return(err); 2290} 2291 2292int 2293ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) { 2294 DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle, 2295 nosys); 2296} 2297 2298#endif /* MROUTE_LKM */ 2299#endif /* MROUTING */ 2300