ip_mroute.c revision 14549
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 1.29 1996/03/11 15:13:17 davidg Exp $ 13 */ 14 15#include "opt_mrouting.h" 16 17#include <sys/param.h> 18#include <sys/queue.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/errno.h> 25#include <sys/time.h> 26#include <sys/kernel.h> 27#include <sys/ioctl.h> 28#include <sys/syslog.h> 29#include <net/if.h> 30#include <net/route.h> 31#include <netinet/in.h> 32#include <netinet/in_systm.h> 33#include <netinet/ip.h> 34#include <netinet/ip_var.h> 35#include <netinet/in_pcb.h> 36#include <netinet/in_var.h> 37#include <netinet/igmp.h> 38#include <netinet/igmp_var.h> 39#include <netinet/ip_mroute.h> 40#include <netinet/udp.h> 41 42extern void rsvp_input __P((struct mbuf *m, int iphlen)); 43 44#ifndef NTOHL 45#if BYTE_ORDER != BIG_ENDIAN 46#define NTOHL(d) ((d) = ntohl((d))) 47#define NTOHS(d) ((d) = ntohs((u_short)(d))) 48#define HTONL(d) ((d) = htonl((d))) 49#define HTONS(d) ((d) = htons((u_short)(d))) 50#else 51#define NTOHL(d) 52#define NTOHS(d) 53#define HTONL(d) 54#define HTONS(d) 55#endif 56#endif 57 58#ifndef MROUTING 59extern void ipip_input __P((struct mbuf *m)); 60extern u_long _ip_mcast_src __P((int vifi)); 61extern int _ip_mforward __P((struct ip *ip, struct ifnet *ifp, 62 struct mbuf *m, struct ip_moptions *imo)); 63extern int _ip_mrouter_done __P((void)); 64extern int _ip_mrouter_get __P((int cmd, struct socket *so, 65 struct mbuf **m)); 66extern int _ip_mrouter_set __P((int cmd, struct socket *so, 67 struct mbuf *m)); 68extern int _mrt_ioctl __P((int req, caddr_t data, struct proc *p)); 69 70/* 71 * Dummy routines and globals used when multicast routing is not compiled in. 72 */ 73 74struct socket *ip_mrouter = NULL; 75static u_int ip_mrtproto = 0; 76static struct mrtstat mrtstat; 77u_int rsvpdebug = 0; 78 79int 80_ip_mrouter_set(cmd, so, m) 81 int cmd; 82 struct socket *so; 83 struct mbuf *m; 84{ 85 return(EOPNOTSUPP); 86} 87 88int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = _ip_mrouter_set; 89 90 91int 92_ip_mrouter_get(cmd, so, m) 93 int cmd; 94 struct socket *so; 95 struct mbuf **m; 96{ 97 return(EOPNOTSUPP); 98} 99 100int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = _ip_mrouter_get; 101 102int 103_ip_mrouter_done() 104{ 105 return(0); 106} 107 108int (*ip_mrouter_done)(void) = _ip_mrouter_done; 109 110int 111_ip_mforward(ip, ifp, m, imo) 112 struct ip *ip; 113 struct ifnet *ifp; 114 struct mbuf *m; 115 struct ip_moptions *imo; 116{ 117 return(0); 118} 119 120int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 121 struct ip_moptions *) = _ip_mforward; 122 123int 124_mrt_ioctl(int req, caddr_t data, struct proc *p) 125{ 126 return EOPNOTSUPP; 127} 128 129int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl; 130 131void 132rsvp_input(m, iphlen) /* XXX must fixup manually */ 133 struct mbuf *m; 134 int iphlen; 135{ 136 /* Can still get packets with rsvp_on = 0 if there is a local member 137 * of the group to which the RSVP packet is addressed. But in this 138 * case we want to throw the packet away. 139 */ 140 if (!rsvp_on) { 141 m_freem(m); 142 return; 143 } 144 145 if (ip_rsvpd != NULL) { 146 if (rsvpdebug) 147 printf("rsvp_input: Sending packet up old-style socket\n"); 148 rip_input(m); 149 return; 150 } 151 /* Drop the packet */ 152 m_freem(m); 153} 154 155void ipip_input(struct mbuf *m) { /* XXX must fixup manually */ 156 rip_input(m); 157} 158 159int (*legal_vif_num)(int) = 0; 160 161/* 162 * This should never be called, since IP_MULTICAST_VIF should fail, but 163 * just in case it does get called, the code a little lower in ip_output 164 * will assign the packet a local address. 165 */ 166u_long 167_ip_mcast_src(int vifi) { return INADDR_ANY; } 168u_long (*ip_mcast_src)(int) = _ip_mcast_src; 169 170int 171ip_rsvp_vif_init(so, m) 172 struct socket *so; 173 struct mbuf *m; 174{ 175 return(EINVAL); 176} 177 178int 179ip_rsvp_vif_done(so, m) 180 struct socket *so; 181 struct mbuf *m; 182{ 183 return(EINVAL); 184} 185 186void 187ip_rsvp_force_done(so) 188 struct socket *so; 189{ 190 return; 191} 192 193#else /* MROUTING */ 194 195#define M_HASCL(m) ((m)->m_flags & M_EXT) 196 197#define INSIZ sizeof(struct in_addr) 198#define same(a1, a2) \ 199 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0) 200 201#define MT_MRTABLE MT_RTABLE /* since nothing else uses it */ 202 203/* 204 * Globals. All but ip_mrouter and ip_mrtproto could be static, 205 * except for netstat or debugging purposes. 206 */ 207#ifndef MROUTE_LKM 208extern void ipip_input __P((struct mbuf *m, int iphlen)); 209struct socket *ip_mrouter = NULL; 210struct mrtstat mrtstat; 211 212int ip_mrtproto = IGMP_DVMRP; /* for netstat only */ 213#else /* MROUTE_LKM */ 214#error /* the function definition will have a syntax error */ 215extern void X_ipip_input __P((struct mbuf *m)); 216extern struct mrtstat mrtstat; 217static int ip_mrtproto; 218#endif 219 220#define NO_RTE_FOUND 0x1 221#define RTE_FOUND 0x2 222 223static struct mbuf *mfctable[MFCTBLSIZ]; 224static u_char nexpire[MFCTBLSIZ]; 225static struct vif viftable[MAXVIFS]; 226static u_int mrtdebug = 0; /* debug level */ 227#define DEBUG_MFC 0x02 228#define DEBUG_FORWARD 0x04 229#define DEBUG_EXPIRE 0x08 230#define DEBUG_XMIT 0x10 231static u_int tbfdebug = 0; /* tbf debug level */ 232static u_int rsvpdebug = 0; /* rsvp debug level */ 233 234#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 235#define UPCALL_EXPIRE 6 /* number of timeouts */ 236 237/* 238 * Define the token bucket filter structures 239 * tbftable -> each vif has one of these for storing info 240 */ 241 242static struct tbf tbftable[MAXVIFS]; 243#define TBF_REPROCESS (hz / 100) /* 100x / second */ 244 245/* 246 * 'Interfaces' associated with decapsulator (so we can tell 247 * packets that went through it from ones that get reflected 248 * by a broken gateway). These interfaces are never linked into 249 * the system ifnet list & no routes point to them. I.e., packets 250 * can't be sent this way. They only exist as a placeholder for 251 * multicast source verification. 252 */ 253static struct ifnet multicast_decap_if[MAXVIFS]; 254 255#define ENCAP_TTL 64 256#define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 257 258/* prototype IP hdr for encapsulated packets */ 259static struct ip multicast_encap_iphdr = { 260#if BYTE_ORDER == LITTLE_ENDIAN 261 sizeof(struct ip) >> 2, IPVERSION, 262#else 263 IPVERSION, sizeof(struct ip) >> 2, 264#endif 265 0, /* tos */ 266 sizeof(struct ip), /* total length */ 267 0, /* id */ 268 0, /* frag offset */ 269 ENCAP_TTL, ENCAP_PROTO, 270 0, /* checksum */ 271}; 272 273/* 274 * Private variables. 275 */ 276static vifi_t numvifs = 0; 277static int have_encap_tunnel = 0; 278 279/* 280 * one-back cache used by ipip_input to locate a tunnel's vif 281 * given a datagram's src ip address. 282 */ 283static u_long last_encap_src; 284static struct vif *last_encap_vif; 285 286static u_long X_ip_mcast_src __P((int vifi)); 287static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo)); 288static int X_ip_mrouter_done __P((void)); 289static int X_ip_mrouter_get __P((int cmd, struct socket *so, struct mbuf **m)); 290static int X_ip_mrouter_set __P((int cmd, struct socket *so, struct mbuf *m)); 291static int X_legal_vif_num __P((int vif)); 292static int X_mrt_ioctl __P((int cmd, caddr_t data)); 293 294static int get_sg_cnt(struct sioc_sg_req *); 295static int get_vif_cnt(struct sioc_vif_req *); 296static int ip_mrouter_init(struct socket *, struct mbuf *); 297static int add_vif(struct vifctl *); 298static int del_vif(vifi_t *); 299static int add_mfc(struct mfcctl *); 300static int del_mfc(struct mfcctl *); 301static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 302static int get_version(struct mbuf *); 303static int get_assert(struct mbuf *); 304static int set_assert(int *); 305static void expire_upcalls(void *); 306static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, 307 vifi_t); 308static void phyint_send(struct ip *, struct vif *, struct mbuf *); 309static void encap_send(struct ip *, struct vif *, struct mbuf *); 310static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 311static void tbf_queue(struct vif *, struct mbuf *); 312static void tbf_process_q(struct vif *); 313static void tbf_reprocess_q(void *); 314static int tbf_dq_sel(struct vif *, struct ip *); 315static void tbf_send_packet(struct vif *, struct mbuf *); 316static void tbf_update_tokens(struct vif *); 317static int priority(struct vif *, struct ip *); 318void multiencap_decap(struct mbuf *); 319 320/* 321 * whether or not special PIM assert processing is enabled. 322 */ 323static int pim_assert; 324/* 325 * Rate limit for assert notification messages, in usec 326 */ 327#define ASSERT_MSG_TIME 3000000 328 329/* 330 * Hash function for a source, group entry 331 */ 332#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 333 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 334 335/* 336 * Find a route for a given origin IP address and Multicast group address 337 * Type of service parameter to be added in the future!!! 338 */ 339 340#define MFCFIND(o, g, rt) { \ 341 register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \ 342 register struct mfc *_rt = NULL; \ 343 rt = NULL; \ 344 ++mrtstat.mrts_mfc_lookups; \ 345 while (_mb_rt) { \ 346 _rt = mtod(_mb_rt, struct mfc *); \ 347 if ((_rt->mfc_origin.s_addr == o) && \ 348 (_rt->mfc_mcastgrp.s_addr == g) && \ 349 (_mb_rt->m_act == NULL)) { \ 350 rt = _rt; \ 351 break; \ 352 } \ 353 _mb_rt = _mb_rt->m_next; \ 354 } \ 355 if (rt == NULL) { \ 356 ++mrtstat.mrts_mfc_misses; \ 357 } \ 358} 359 360 361/* 362 * Macros to compute elapsed time efficiently 363 * Borrowed from Van Jacobson's scheduling code 364 */ 365#define TV_DELTA(a, b, delta) { \ 366 register int xxs; \ 367 \ 368 delta = (a).tv_usec - (b).tv_usec; \ 369 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 370 switch (xxs) { \ 371 case 2: \ 372 delta += 1000000; \ 373 /* fall through */ \ 374 case 1: \ 375 delta += 1000000; \ 376 break; \ 377 default: \ 378 delta += (1000000 * xxs); \ 379 } \ 380 } \ 381} 382 383#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 384 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 385 386#ifdef UPCALL_TIMING 387u_long upcall_data[51]; 388static void collate(struct timeval *); 389#endif /* UPCALL_TIMING */ 390 391 392/* 393 * Handle MRT setsockopt commands to modify the multicast routing tables. 394 */ 395static int 396X_ip_mrouter_set(cmd, so, m) 397 int cmd; 398 struct socket *so; 399 struct mbuf *m; 400{ 401 if (cmd != MRT_INIT && so != ip_mrouter) return EACCES; 402 403 switch (cmd) { 404 case MRT_INIT: return ip_mrouter_init(so, m); 405 case MRT_DONE: return ip_mrouter_done(); 406 case MRT_ADD_VIF: return add_vif (mtod(m, struct vifctl *)); 407 case MRT_DEL_VIF: return del_vif (mtod(m, vifi_t *)); 408 case MRT_ADD_MFC: return add_mfc (mtod(m, struct mfcctl *)); 409 case MRT_DEL_MFC: return del_mfc (mtod(m, struct mfcctl *)); 410 case MRT_ASSERT: return set_assert(mtod(m, int *)); 411 default: return EOPNOTSUPP; 412 } 413} 414 415#ifndef MROUTE_LKM 416int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = X_ip_mrouter_set; 417#endif 418 419/* 420 * Handle MRT getsockopt commands 421 */ 422static int 423X_ip_mrouter_get(cmd, so, m) 424 int cmd; 425 struct socket *so; 426 struct mbuf **m; 427{ 428 struct mbuf *mb; 429 430 if (so != ip_mrouter) return EACCES; 431 432 *m = mb = m_get(M_WAIT, MT_SOOPTS); 433 434 switch (cmd) { 435 case MRT_VERSION: return get_version(mb); 436 case MRT_ASSERT: return get_assert(mb); 437 default: return EOPNOTSUPP; 438 } 439} 440 441#ifndef MROUTE_LKM 442int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = X_ip_mrouter_get; 443#endif 444 445/* 446 * Handle ioctl commands to obtain information from the cache 447 */ 448static int 449X_mrt_ioctl(cmd, data) 450 int cmd; 451 caddr_t data; 452{ 453 int error = 0; 454 455 switch (cmd) { 456 case (SIOCGETVIFCNT): 457 return (get_vif_cnt((struct sioc_vif_req *)data)); 458 break; 459 case (SIOCGETSGCNT): 460 return (get_sg_cnt((struct sioc_sg_req *)data)); 461 break; 462 default: 463 return (EINVAL); 464 break; 465 } 466 return error; 467} 468 469#ifndef MROUTE_LKM 470int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl; 471#endif 472 473/* 474 * returns the packet, byte, rpf-failure count for the source group provided 475 */ 476static int 477get_sg_cnt(req) 478 register struct sioc_sg_req *req; 479{ 480 register struct mfc *rt; 481 int s; 482 483 s = splnet(); 484 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 485 splx(s); 486 if (rt != NULL) { 487 req->pktcnt = rt->mfc_pkt_cnt; 488 req->bytecnt = rt->mfc_byte_cnt; 489 req->wrong_if = rt->mfc_wrong_if; 490 } else 491 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 492 493 return 0; 494} 495 496/* 497 * returns the input and output packet and byte counts on the vif provided 498 */ 499static int 500get_vif_cnt(req) 501 register struct sioc_vif_req *req; 502{ 503 register vifi_t vifi = req->vifi; 504 505 if (vifi >= numvifs) return EINVAL; 506 507 req->icount = viftable[vifi].v_pkt_in; 508 req->ocount = viftable[vifi].v_pkt_out; 509 req->ibytes = viftable[vifi].v_bytes_in; 510 req->obytes = viftable[vifi].v_bytes_out; 511 512 return 0; 513} 514 515/* 516 * Enable multicast routing 517 */ 518static int 519ip_mrouter_init(so, m) 520 struct socket *so; 521 struct mbuf *m; 522{ 523 int *v; 524 525 if (mrtdebug) 526 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 527 so->so_type, so->so_proto->pr_protocol); 528 529 if (so->so_type != SOCK_RAW || 530 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP; 531 532 if (!m || (m->m_len != sizeof(int *))) 533 return ENOPROTOOPT; 534 535 v = mtod(m, int *); 536 if (*v != 1) 537 return ENOPROTOOPT; 538 539 if (ip_mrouter != NULL) return EADDRINUSE; 540 541 ip_mrouter = so; 542 543 bzero((caddr_t)mfctable, sizeof(mfctable)); 544 bzero((caddr_t)nexpire, sizeof(nexpire)); 545 546 pim_assert = 0; 547 548 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 549 550 if (mrtdebug) 551 log(LOG_DEBUG, "ip_mrouter_init\n"); 552 553 return 0; 554} 555 556/* 557 * Disable multicast routing 558 */ 559static int 560X_ip_mrouter_done() 561{ 562 vifi_t vifi; 563 int i; 564 struct ifnet *ifp; 565 struct ifreq ifr; 566 struct mbuf *mb_rt; 567 struct mbuf *m; 568 struct rtdetq *rte; 569 int s; 570 571 s = splnet(); 572 573 /* 574 * For each phyint in use, disable promiscuous reception of all IP 575 * multicasts. 576 */ 577 for (vifi = 0; vifi < numvifs; vifi++) { 578 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 579 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 580 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 581 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr 582 = INADDR_ANY; 583 ifp = viftable[vifi].v_ifp; 584 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 585 } 586 } 587 bzero((caddr_t)tbftable, sizeof(tbftable)); 588 bzero((caddr_t)viftable, sizeof(viftable)); 589 numvifs = 0; 590 pim_assert = 0; 591 592 untimeout(expire_upcalls, (caddr_t)NULL); 593 594 /* 595 * Free all multicast forwarding cache entries. 596 */ 597 for (i = 0; i < MFCTBLSIZ; i++) { 598 mb_rt = mfctable[i]; 599 while (mb_rt) { 600 if (mb_rt->m_act != NULL) { 601 while (mb_rt->m_act) { 602 m = mb_rt->m_act; 603 mb_rt->m_act = m->m_act; 604 rte = mtod(m, struct rtdetq *); 605 m_freem(rte->m); 606 m_free(m); 607 } 608 } 609 mb_rt = m_free(mb_rt); 610 } 611 } 612 613 bzero((caddr_t)mfctable, sizeof(mfctable)); 614 615 /* 616 * Reset de-encapsulation cache 617 */ 618 last_encap_src = NULL; 619 last_encap_vif = NULL; 620 have_encap_tunnel = 0; 621 622 ip_mrouter = NULL; 623 624 splx(s); 625 626 if (mrtdebug) 627 log(LOG_DEBUG, "ip_mrouter_done\n"); 628 629 return 0; 630} 631 632#ifndef MROUTE_LKM 633int (*ip_mrouter_done)(void) = X_ip_mrouter_done; 634#endif 635 636static int 637get_version(mb) 638 struct mbuf *mb; 639{ 640 int *v; 641 642 v = mtod(mb, int *); 643 644 *v = 0x0305; /* XXX !!!! */ 645 mb->m_len = sizeof(int); 646 647 return 0; 648} 649 650/* 651 * Set PIM assert processing global 652 */ 653static int 654set_assert(i) 655 int *i; 656{ 657 if ((*i != 1) && (*i != 0)) 658 return EINVAL; 659 660 pim_assert = *i; 661 662 return 0; 663} 664 665/* 666 * Get PIM assert processing global 667 */ 668static int 669get_assert(m) 670 struct mbuf *m; 671{ 672 int *i; 673 674 i = mtod(m, int *); 675 676 *i = pim_assert; 677 678 return 0; 679} 680 681/* 682 * Add a vif to the vif table 683 */ 684static int 685add_vif(vifcp) 686 register struct vifctl *vifcp; 687{ 688 register struct vif *vifp = viftable + vifcp->vifc_vifi; 689 static struct sockaddr_in sin = {sizeof sin, AF_INET}; 690 struct ifaddr *ifa; 691 struct ifnet *ifp; 692 struct ifreq ifr; 693 int error, s; 694 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 695 696 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL; 697 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE; 698 699 /* Find the interface with an address in AF_INET family */ 700 sin.sin_addr = vifcp->vifc_lcl_addr; 701 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 702 if (ifa == 0) return EADDRNOTAVAIL; 703 ifp = ifa->ifa_ifp; 704 705 if (vifcp->vifc_flags & VIFF_TUNNEL) { 706 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 707 /* 708 * An encapsulating tunnel is wanted. Tell ipip_input() to 709 * start paying attention to encapsulated packets. 710 */ 711 if (have_encap_tunnel == 0) { 712 have_encap_tunnel = 1; 713 for (s = 0; s < MAXVIFS; ++s) { 714 multicast_decap_if[s].if_name = "mdecap"; 715 multicast_decap_if[s].if_unit = s; 716 } 717 } 718 /* 719 * Set interface to fake encapsulator interface 720 */ 721 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 722 /* 723 * Prepare cached route entry 724 */ 725 bzero(&vifp->v_route, sizeof(vifp->v_route)); 726 } else { 727 log(LOG_ERR, "source routed tunnels not supported\n"); 728 return EOPNOTSUPP; 729 } 730 } else { 731 /* Make sure the interface supports multicast */ 732 if ((ifp->if_flags & IFF_MULTICAST) == 0) 733 return EOPNOTSUPP; 734 735 /* Enable promiscuous reception of all IP multicasts from the if */ 736 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 737 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 738 s = splnet(); 739 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr); 740 splx(s); 741 if (error) 742 return error; 743 } 744 745 s = splnet(); 746 /* define parameters for the tbf structure */ 747 vifp->v_tbf = v_tbf; 748 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 749 vifp->v_tbf->tbf_n_tok = 0; 750 vifp->v_tbf->tbf_q_len = 0; 751 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 752 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 753 754 vifp->v_flags = vifcp->vifc_flags; 755 vifp->v_threshold = vifcp->vifc_threshold; 756 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 757 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 758 vifp->v_ifp = ifp; 759 /* scaling up here allows division by 1024 in critical code */ 760 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 761 vifp->v_rsvp_on = 0; 762 vifp->v_rsvpd = NULL; 763 /* initialize per vif pkt counters */ 764 vifp->v_pkt_in = 0; 765 vifp->v_pkt_out = 0; 766 vifp->v_bytes_in = 0; 767 vifp->v_bytes_out = 0; 768 splx(s); 769 770 /* Adjust numvifs up if the vifi is higher than numvifs */ 771 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 772 773 if (mrtdebug) 774 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 775 vifcp->vifc_vifi, 776 ntohl(vifcp->vifc_lcl_addr.s_addr), 777 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 778 ntohl(vifcp->vifc_rmt_addr.s_addr), 779 vifcp->vifc_threshold, 780 vifcp->vifc_rate_limit); 781 782 return 0; 783} 784 785/* 786 * Delete a vif from the vif table 787 */ 788static int 789del_vif(vifip) 790 vifi_t *vifip; 791{ 792 register struct vif *vifp = viftable + *vifip; 793 register vifi_t vifi; 794 register struct mbuf *m; 795 struct ifnet *ifp; 796 struct ifreq ifr; 797 int s; 798 799 if (*vifip >= numvifs) return EINVAL; 800 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL; 801 802 s = splnet(); 803 804 if (!(vifp->v_flags & VIFF_TUNNEL)) { 805 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 806 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 807 ifp = vifp->v_ifp; 808 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 809 } 810 811 if (vifp == last_encap_vif) { 812 last_encap_vif = 0; 813 last_encap_src = 0; 814 } 815 816 /* 817 * Free packets queued at the interface 818 */ 819 while (vifp->v_tbf->tbf_q) { 820 m = vifp->v_tbf->tbf_q; 821 vifp->v_tbf->tbf_q = m->m_act; 822 m_freem(m); 823 } 824 825 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 826 bzero((caddr_t)vifp, sizeof (*vifp)); 827 828 /* Adjust numvifs down */ 829 for (vifi = numvifs; vifi > 0; vifi--) 830 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break; 831 numvifs = vifi; 832 833 splx(s); 834 835 if (mrtdebug) 836 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 837 838 return 0; 839} 840 841/* 842 * Add an mfc entry 843 */ 844static int 845add_mfc(mfccp) 846 struct mfcctl *mfccp; 847{ 848 struct mfc *rt; 849 register struct mbuf *mb_rt; 850 u_long hash; 851 struct mbuf *mb_ntry; 852 struct rtdetq *rte; 853 register u_short nstl; 854 int s; 855 int i; 856 857 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 858 859 /* If an entry already exists, just update the fields */ 860 if (rt) { 861 if (mrtdebug & DEBUG_MFC) 862 log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n", 863 ntohl(mfccp->mfcc_origin.s_addr), 864 ntohl(mfccp->mfcc_mcastgrp.s_addr), 865 mfccp->mfcc_parent); 866 867 s = splnet(); 868 rt->mfc_parent = mfccp->mfcc_parent; 869 for (i = 0; i < numvifs; i++) 870 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 871 splx(s); 872 return 0; 873 } 874 875 /* 876 * Find the entry for which the upcall was made and update 877 */ 878 s = splnet(); 879 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 880 for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) { 881 882 rt = mtod(mb_rt, struct mfc *); 883 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 884 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 885 (mb_rt->m_act != NULL)) { 886 887 if (nstl++) 888 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %x\n", 889 "multiple kernel entries", 890 ntohl(mfccp->mfcc_origin.s_addr), 891 ntohl(mfccp->mfcc_mcastgrp.s_addr), 892 mfccp->mfcc_parent, mb_rt->m_act); 893 894 if (mrtdebug & DEBUG_MFC) 895 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %x\n", 896 ntohl(mfccp->mfcc_origin.s_addr), 897 ntohl(mfccp->mfcc_mcastgrp.s_addr), 898 mfccp->mfcc_parent, mb_rt->m_act); 899 900 rt->mfc_origin = mfccp->mfcc_origin; 901 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 902 rt->mfc_parent = mfccp->mfcc_parent; 903 for (i = 0; i < numvifs; i++) 904 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 905 /* initialize pkt counters per src-grp */ 906 rt->mfc_pkt_cnt = 0; 907 rt->mfc_byte_cnt = 0; 908 rt->mfc_wrong_if = 0; 909 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 910 911 rt->mfc_expire = 0; /* Don't clean this guy up */ 912 nexpire[hash]--; 913 914 /* free packets Qed at the end of this entry */ 915 while (mb_rt->m_act) { 916 mb_ntry = mb_rt->m_act; 917 rte = mtod(mb_ntry, struct rtdetq *); 918/* #ifdef RSVP_ISI */ 919 ip_mdq(rte->m, rte->ifp, rt, -1); 920/* #endif */ 921 mb_rt->m_act = mb_ntry->m_act; 922 m_freem(rte->m); 923#ifdef UPCALL_TIMING 924 collate(&(rte->t)); 925#endif /* UPCALL_TIMING */ 926 m_free(mb_ntry); 927 } 928 } 929 } 930 931 /* 932 * It is possible that an entry is being inserted without an upcall 933 */ 934 if (nstl == 0) { 935 if (mrtdebug & DEBUG_MFC) 936 log(LOG_DEBUG,"add_mfc no upcall h %d o %x g %x p %x\n", 937 hash, ntohl(mfccp->mfcc_origin.s_addr), 938 ntohl(mfccp->mfcc_mcastgrp.s_addr), 939 mfccp->mfcc_parent); 940 941 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 942 943 rt = mtod(mb_rt, struct mfc *); 944 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 945 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 946 947 rt->mfc_origin = mfccp->mfcc_origin; 948 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 949 rt->mfc_parent = mfccp->mfcc_parent; 950 for (i = 0; i < numvifs; i++) 951 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 952 /* initialize pkt counters per src-grp */ 953 rt->mfc_pkt_cnt = 0; 954 rt->mfc_byte_cnt = 0; 955 rt->mfc_wrong_if = 0; 956 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 957 if (rt->mfc_expire) 958 nexpire[hash]--; 959 rt->mfc_expire = 0; 960 } 961 } 962 if (mb_rt == NULL) { 963 /* no upcall, so make a new entry */ 964 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 965 if (mb_rt == NULL) { 966 splx(s); 967 return ENOBUFS; 968 } 969 970 rt = mtod(mb_rt, struct mfc *); 971 972 /* insert new entry at head of hash chain */ 973 rt->mfc_origin = mfccp->mfcc_origin; 974 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 975 rt->mfc_parent = mfccp->mfcc_parent; 976 for (i = 0; i < numvifs; i++) 977 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 978 /* initialize pkt counters per src-grp */ 979 rt->mfc_pkt_cnt = 0; 980 rt->mfc_byte_cnt = 0; 981 rt->mfc_wrong_if = 0; 982 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 983 rt->mfc_expire = 0; 984 985 /* link into table */ 986 mb_rt->m_next = mfctable[hash]; 987 mfctable[hash] = mb_rt; 988 mb_rt->m_act = NULL; 989 } 990 } 991 splx(s); 992 return 0; 993} 994 995#ifdef UPCALL_TIMING 996/* 997 * collect delay statistics on the upcalls 998 */ 999static void collate(t) 1000register struct timeval *t; 1001{ 1002 register u_long d; 1003 register struct timeval tp; 1004 register u_long delta; 1005 1006 GET_TIME(tp); 1007 1008 if (TV_LT(*t, tp)) 1009 { 1010 TV_DELTA(tp, *t, delta); 1011 1012 d = delta >> 10; 1013 if (d > 50) 1014 d = 50; 1015 1016 ++upcall_data[d]; 1017 } 1018} 1019#endif /* UPCALL_TIMING */ 1020 1021/* 1022 * Delete an mfc entry 1023 */ 1024static int 1025del_mfc(mfccp) 1026 struct mfcctl *mfccp; 1027{ 1028 struct in_addr origin; 1029 struct in_addr mcastgrp; 1030 struct mfc *rt; 1031 struct mbuf *mb_rt; 1032 struct mbuf **nptr; 1033 u_long hash; 1034 int s; 1035 1036 origin = mfccp->mfcc_origin; 1037 mcastgrp = mfccp->mfcc_mcastgrp; 1038 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1039 1040 if (mrtdebug & DEBUG_MFC) 1041 log(LOG_DEBUG,"del_mfc orig %x mcastgrp %x\n", 1042 ntohl(origin.s_addr), ntohl(mcastgrp.s_addr)); 1043 1044 s = splnet(); 1045 1046 nptr = &mfctable[hash]; 1047 while ((mb_rt = *nptr) != NULL) { 1048 rt = mtod(mb_rt, struct mfc *); 1049 if (origin.s_addr == rt->mfc_origin.s_addr && 1050 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1051 mb_rt->m_act == NULL) 1052 break; 1053 1054 nptr = &mb_rt->m_next; 1055 } 1056 if (mb_rt == NULL) { 1057 splx(s); 1058 return EADDRNOTAVAIL; 1059 } 1060 1061 MFREE(mb_rt, *nptr); 1062 1063 splx(s); 1064 1065 return 0; 1066} 1067 1068/* 1069 * Send a message to mrouted on the multicast routing socket 1070 */ 1071static int 1072socket_send(s, mm, src) 1073 struct socket *s; 1074 struct mbuf *mm; 1075 struct sockaddr_in *src; 1076{ 1077 if (s) { 1078 if (sbappendaddr(&s->so_rcv, 1079 (struct sockaddr *)src, 1080 mm, (struct mbuf *)0) != 0) { 1081 sorwakeup(s); 1082 return 0; 1083 } 1084 } 1085 m_freem(mm); 1086 return -1; 1087} 1088 1089/* 1090 * IP multicast forwarding function. This function assumes that the packet 1091 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1092 * pointed to by "ifp", and the packet is to be relayed to other networks 1093 * that have members of the packet's destination IP multicast group. 1094 * 1095 * The packet is returned unscathed to the caller, unless it is 1096 * erroneous, in which case a non-zero return value tells the caller to 1097 * discard it. 1098 */ 1099 1100#define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1101#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1102 1103static int 1104X_ip_mforward(ip, ifp, m, imo) 1105 register struct ip *ip; 1106 struct ifnet *ifp; 1107 struct mbuf *m; 1108 struct ip_moptions *imo; 1109{ 1110 register struct mfc *rt; 1111 register u_char *ipoptions; 1112 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1113 static int srctun = 0; 1114 register struct mbuf *mm; 1115 int s; 1116 vifi_t vifi; 1117 struct vif *vifp; 1118 1119 if (mrtdebug & DEBUG_FORWARD) 1120 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %x\n", 1121 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1122 1123 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1124 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1125 /* 1126 * Packet arrived via a physical interface or 1127 * an encapsulated tunnel. 1128 */ 1129 } else { 1130 /* 1131 * Packet arrived through a source-route tunnel. 1132 * Source-route tunnels are no longer supported. 1133 */ 1134 if ((srctun++ % 1000) == 0) 1135 log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n", 1136 ntohl(ip->ip_src.s_addr)); 1137 1138 return 1; 1139 } 1140 1141 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1142 if (ip->ip_ttl < 255) 1143 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1144 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1145 vifp = viftable + vifi; 1146 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1147 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi, 1148 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1149 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1150 } 1151 return (ip_mdq(m, ifp, NULL, vifi)); 1152 } 1153 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1154 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1155 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr)); 1156 if(!imo) 1157 printf("In fact, no options were specified at all\n"); 1158 } 1159 1160 /* 1161 * Don't forward a packet with time-to-live of zero or one, 1162 * or a packet destined to a local-only group. 1163 */ 1164 if (ip->ip_ttl <= 1 || 1165 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1166 return 0; 1167 1168 /* 1169 * Determine forwarding vifs from the forwarding cache table 1170 */ 1171 s = splnet(); 1172 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt); 1173 1174 /* Entry exists, so forward if necessary */ 1175 if (rt != NULL) { 1176 splx(s); 1177 return (ip_mdq(m, ifp, rt, -1)); 1178 } else { 1179 /* 1180 * If we don't have a route for packet's origin, 1181 * Make a copy of the packet & 1182 * send message to routing daemon 1183 */ 1184 1185 register struct mbuf *mb_rt; 1186 register struct mbuf *mb_ntry; 1187 register struct mbuf *mb0; 1188 register struct rtdetq *rte; 1189 register struct mbuf *rte_m; 1190 register u_long hash; 1191 register int npkts; 1192 int hlen = ip->ip_hl << 2; 1193#ifdef UPCALL_TIMING 1194 struct timeval tp; 1195 1196 GET_TIME(tp); 1197#endif 1198 1199 mrtstat.mrts_no_route++; 1200 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1201 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1202 ntohl(ip->ip_src.s_addr), 1203 ntohl(ip->ip_dst.s_addr)); 1204 1205 /* 1206 * Allocate mbufs early so that we don't do extra work if we are 1207 * just going to fail anyway. Make sure to pullup the header so 1208 * that other people can't step on it. 1209 */ 1210 MGET(mb_ntry, M_DONTWAIT, MT_DATA); 1211 if (mb_ntry == NULL) { 1212 splx(s); 1213 return ENOBUFS; 1214 } 1215 mb0 = m_copy(m, 0, M_COPYALL); 1216 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1217 mb0 = m_pullup(mb0, hlen); 1218 if (mb0 == NULL) { 1219 m_free(mb_ntry); 1220 splx(s); 1221 return ENOBUFS; 1222 } 1223 1224 /* is there an upcall waiting for this packet? */ 1225 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1226 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 1227 rt = mtod(mb_rt, struct mfc *); 1228 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1229 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1230 (mb_rt->m_act != NULL)) 1231 break; 1232 } 1233 1234 if (mb_rt == NULL) { 1235 int i; 1236 struct igmpmsg *im; 1237 1238 /* no upcall, so make a new entry */ 1239 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 1240 if (mb_rt == NULL) { 1241 m_free(mb_ntry); 1242 m_freem(mb0); 1243 splx(s); 1244 return ENOBUFS; 1245 } 1246 /* Make a copy of the header to send to the user level process */ 1247 mm = m_copy(m, 0, hlen); 1248 if (mm == NULL) { 1249 m_free(mb_ntry); 1250 m_freem(mb0); 1251 m_free(mb_rt); 1252 splx(s); 1253 return ENOBUFS; 1254 } 1255 1256 /* 1257 * Send message to routing daemon to install 1258 * a route into the kernel table 1259 */ 1260 k_igmpsrc.sin_addr = ip->ip_src; 1261 1262 im = mtod(mm, struct igmpmsg *); 1263 im->im_msgtype = IGMPMSG_NOCACHE; 1264 im->im_mbz = 0; 1265 1266 mrtstat.mrts_upcalls++; 1267 1268 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1269 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1270 ++mrtstat.mrts_upq_sockfull; 1271 m_free(mb_ntry); 1272 m_freem(mb0); 1273 m_free(mb_rt); 1274 splx(s); 1275 return ENOBUFS; 1276 } 1277 1278 rt = mtod(mb_rt, struct mfc *); 1279 1280 /* insert new entry at head of hash chain */ 1281 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1282 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1283 rt->mfc_expire = UPCALL_EXPIRE; 1284 nexpire[hash]++; 1285 for (i = 0; i < numvifs; i++) 1286 rt->mfc_ttls[i] = 0; 1287 rt->mfc_parent = -1; 1288 1289 /* link into table */ 1290 mb_rt->m_next = mfctable[hash]; 1291 mfctable[hash] = mb_rt; 1292 mb_rt->m_act = NULL; 1293 1294 rte_m = mb_rt; 1295 } else { 1296 /* determine if q has overflowed */ 1297 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act) 1298 npkts++; 1299 1300 if (npkts > MAX_UPQ) { 1301 mrtstat.mrts_upq_ovflw++; 1302 m_free(mb_ntry); 1303 m_freem(mb0); 1304 splx(s); 1305 return 0; 1306 } 1307 } 1308 1309 mb_ntry->m_act = NULL; 1310 rte = mtod(mb_ntry, struct rtdetq *); 1311 1312 rte->m = mb0; 1313 rte->ifp = ifp; 1314#ifdef UPCALL_TIMING 1315 rte->t = tp; 1316#endif 1317 1318 /* Add this entry to the end of the queue */ 1319 rte_m->m_act = mb_ntry; 1320 1321 splx(s); 1322 1323 return 0; 1324 } 1325} 1326 1327#ifndef MROUTE_LKM 1328int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 1329 struct ip_moptions *) = X_ip_mforward; 1330#endif 1331 1332/* 1333 * Clean up the cache entry if upcall is not serviced 1334 */ 1335static void 1336expire_upcalls(void *unused) 1337{ 1338 struct mbuf *mb_rt, *m, **nptr; 1339 struct rtdetq *rte; 1340 struct mfc *mfc; 1341 int i; 1342 int s; 1343 1344 s = splnet(); 1345 for (i = 0; i < MFCTBLSIZ; i++) { 1346 if (nexpire[i] == 0) 1347 continue; 1348 nptr = &mfctable[i]; 1349 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) { 1350 mfc = mtod(mb_rt, struct mfc *); 1351 1352 /* 1353 * Skip real cache entries 1354 * Make sure it wasn't marked to not expire (shouldn't happen) 1355 * If it expires now 1356 */ 1357 if (mb_rt->m_act != NULL && 1358 mfc->mfc_expire != 0 && 1359 --mfc->mfc_expire == 0) { 1360 if (mrtdebug & DEBUG_EXPIRE) 1361 log(LOG_DEBUG, "expire_upcalls: expiring (%x %x)\n", 1362 ntohl(mfc->mfc_origin.s_addr), 1363 ntohl(mfc->mfc_mcastgrp.s_addr)); 1364 /* 1365 * drop all the packets 1366 * free the mbuf with the pkt, if, timing info 1367 */ 1368 while (mb_rt->m_act) { 1369 m = mb_rt->m_act; 1370 mb_rt->m_act = m->m_act; 1371 1372 rte = mtod(m, struct rtdetq *); 1373 m_freem(rte->m); 1374 m_free(m); 1375 } 1376 ++mrtstat.mrts_cache_cleanups; 1377 nexpire[i]--; 1378 1379 MFREE(mb_rt, *nptr); 1380 } else { 1381 nptr = &mb_rt->m_next; 1382 } 1383 } 1384 } 1385 splx(s); 1386 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 1387} 1388 1389/* 1390 * Packet forwarding routine once entry in the cache is made 1391 */ 1392static int 1393ip_mdq(m, ifp, rt, xmt_vif) 1394 register struct mbuf *m; 1395 register struct ifnet *ifp; 1396 register struct mfc *rt; 1397 register vifi_t xmt_vif; 1398{ 1399 register struct ip *ip = mtod(m, struct ip *); 1400 register vifi_t vifi; 1401 register struct vif *vifp; 1402 register int plen = ntohs(ip->ip_len); 1403 1404/* 1405 * Macro to send packet on vif. Since RSVP packets don't get counted on 1406 * input, they shouldn't get counted on output, so statistics keeping is 1407 * seperate. 1408 */ 1409#define MC_SEND(ip,vifp,m) { \ 1410 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1411 encap_send((ip), (vifp), (m)); \ 1412 else \ 1413 phyint_send((ip), (vifp), (m)); \ 1414} 1415 1416 /* 1417 * If xmt_vif is not -1, send on only the requested vif. 1418 * 1419 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1420 */ 1421 if (xmt_vif < numvifs) { 1422 MC_SEND(ip, viftable + xmt_vif, m); 1423 return 1; 1424 } 1425 1426 /* 1427 * Don't forward if it didn't arrive from the parent vif for its origin. 1428 */ 1429 vifi = rt->mfc_parent; 1430 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1431 /* came in the wrong interface */ 1432 if (mrtdebug & DEBUG_FORWARD) 1433 log(LOG_DEBUG, "wrong if: ifp %x vifi %d vififp %x\n", 1434 ifp, vifi, viftable[vifi].v_ifp); 1435 ++mrtstat.mrts_wrong_if; 1436 ++rt->mfc_wrong_if; 1437 /* 1438 * If we are doing PIM assert processing, and we are forwarding 1439 * packets on this interface, and it is a broadcast medium 1440 * interface (and not a tunnel), send a message to the routing daemon. 1441 */ 1442 if (pim_assert && rt->mfc_ttls[vifi] && 1443 (ifp->if_flags & IFF_BROADCAST) && 1444 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1445 struct sockaddr_in k_igmpsrc; 1446 struct mbuf *mm; 1447 struct igmpmsg *im; 1448 int hlen = ip->ip_hl << 2; 1449 struct timeval now; 1450 register u_long delta; 1451 1452 GET_TIME(now); 1453 1454 TV_DELTA(rt->mfc_last_assert, now, delta); 1455 1456 if (delta > ASSERT_MSG_TIME) { 1457 mm = m_copy(m, 0, hlen); 1458 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1459 mm = m_pullup(mm, hlen); 1460 if (mm == NULL) { 1461 return ENOBUFS; 1462 } 1463 1464 rt->mfc_last_assert = now; 1465 1466 im = mtod(mm, struct igmpmsg *); 1467 im->im_msgtype = IGMPMSG_WRONGVIF; 1468 im->im_mbz = 0; 1469 im->im_vif = vifi; 1470 1471 k_igmpsrc.sin_addr = im->im_src; 1472 1473 socket_send(ip_mrouter, mm, &k_igmpsrc); 1474 } 1475 } 1476 return 0; 1477 } 1478 1479 /* If I sourced this packet, it counts as output, else it was input. */ 1480 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1481 viftable[vifi].v_pkt_out++; 1482 viftable[vifi].v_bytes_out += plen; 1483 } else { 1484 viftable[vifi].v_pkt_in++; 1485 viftable[vifi].v_bytes_in += plen; 1486 } 1487 rt->mfc_pkt_cnt++; 1488 rt->mfc_byte_cnt += plen; 1489 1490 /* 1491 * For each vif, decide if a copy of the packet should be forwarded. 1492 * Forward if: 1493 * - the ttl exceeds the vif's threshold 1494 * - there are group members downstream on interface 1495 */ 1496 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1497 if ((rt->mfc_ttls[vifi] > 0) && 1498 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1499 vifp->v_pkt_out++; 1500 vifp->v_bytes_out += plen; 1501 MC_SEND(ip, vifp, m); 1502 } 1503 1504 return 0; 1505} 1506 1507/* 1508 * check if a vif number is legal/ok. This is used by ip_output, to export 1509 * numvifs there, 1510 */ 1511static int 1512X_legal_vif_num(vif) 1513 int vif; 1514{ 1515 if (vif >= 0 && vif < numvifs) 1516 return(1); 1517 else 1518 return(0); 1519} 1520 1521#ifndef MROUTE_LKM 1522int (*legal_vif_num)(int) = X_legal_vif_num; 1523#endif 1524 1525/* 1526 * Return the local address used by this vif 1527 */ 1528static u_long 1529X_ip_mcast_src(vifi) 1530 int vifi; 1531{ 1532 if (vifi >= 0 && vifi < numvifs) 1533 return viftable[vifi].v_lcl_addr.s_addr; 1534 else 1535 return INADDR_ANY; 1536} 1537 1538#ifndef MROUTE_LKM 1539u_long (*ip_mcast_src)(int) = X_ip_mcast_src; 1540#endif 1541 1542static void 1543phyint_send(ip, vifp, m) 1544 struct ip *ip; 1545 struct vif *vifp; 1546 struct mbuf *m; 1547{ 1548 register struct mbuf *mb_copy; 1549 register int hlen = ip->ip_hl << 2; 1550 1551 /* 1552 * Make a new reference to the packet; make sure that 1553 * the IP header is actually copied, not just referenced, 1554 * so that ip_output() only scribbles on the copy. 1555 */ 1556 mb_copy = m_copy(m, 0, M_COPYALL); 1557 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1558 mb_copy = m_pullup(mb_copy, hlen); 1559 if (mb_copy == NULL) 1560 return; 1561 1562 if (vifp->v_rate_limit <= 0) 1563 tbf_send_packet(vifp, mb_copy); 1564 else 1565 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1566} 1567 1568static void 1569encap_send(ip, vifp, m) 1570 register struct ip *ip; 1571 register struct vif *vifp; 1572 register struct mbuf *m; 1573{ 1574 register struct mbuf *mb_copy; 1575 register struct ip *ip_copy; 1576 register int i, len = ip->ip_len; 1577 1578 /* 1579 * copy the old packet & pullup it's IP header into the 1580 * new mbuf so we can modify it. Try to fill the new 1581 * mbuf since if we don't the ethernet driver will. 1582 */ 1583 MGET(mb_copy, M_DONTWAIT, MT_DATA); 1584 if (mb_copy == NULL) 1585 return; 1586 mb_copy->m_data += 16; 1587 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1588 1589 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1590 m_freem(mb_copy); 1591 return; 1592 } 1593 i = MHLEN - M_LEADINGSPACE(mb_copy); 1594 if (i > len) 1595 i = len; 1596 mb_copy = m_pullup(mb_copy, i); 1597 if (mb_copy == NULL) 1598 return; 1599 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1600 1601 /* 1602 * fill in the encapsulating IP header. 1603 */ 1604 ip_copy = mtod(mb_copy, struct ip *); 1605 *ip_copy = multicast_encap_iphdr; 1606 ip_copy->ip_id = htons(ip_id++); 1607 ip_copy->ip_len += len; 1608 ip_copy->ip_src = vifp->v_lcl_addr; 1609 ip_copy->ip_dst = vifp->v_rmt_addr; 1610 1611 /* 1612 * turn the encapsulated IP header back into a valid one. 1613 */ 1614 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1615 --ip->ip_ttl; 1616 HTONS(ip->ip_len); 1617 HTONS(ip->ip_off); 1618 ip->ip_sum = 0; 1619#if defined(LBL) && !defined(ultrix) 1620 ip->ip_sum = ~oc_cksum((caddr_t)ip, ip->ip_hl << 2, 0); 1621#else 1622 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1623 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1624 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1625#endif 1626 1627 if (vifp->v_rate_limit <= 0) 1628 tbf_send_packet(vifp, mb_copy); 1629 else 1630 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1631} 1632 1633/* 1634 * De-encapsulate a packet and feed it back through ip input (this 1635 * routine is called whenever IP gets a packet with proto type 1636 * ENCAP_PROTO and a local destination address). 1637 */ 1638void 1639#ifdef MROUTE_LKM 1640X_ipip_input(m) 1641#else 1642ipip_input(m, iphlen) 1643#endif 1644 register struct mbuf *m; 1645 int iphlen; 1646{ 1647 struct ifnet *ifp = m->m_pkthdr.rcvif; 1648 register struct ip *ip = mtod(m, struct ip *); 1649 register int hlen = ip->ip_hl << 2; 1650 register int s; 1651 register struct ifqueue *ifq; 1652 register struct vif *vifp; 1653 1654 if (!have_encap_tunnel) { 1655 rip_input(m); 1656 return; 1657 } 1658 /* 1659 * dump the packet if it's not to a multicast destination or if 1660 * we don't have an encapsulating tunnel with the source. 1661 * Note: This code assumes that the remote site IP address 1662 * uniquely identifies the tunnel (i.e., that this site has 1663 * at most one tunnel with the remote site). 1664 */ 1665 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) { 1666 ++mrtstat.mrts_bad_tunnel; 1667 m_freem(m); 1668 return; 1669 } 1670 if (ip->ip_src.s_addr != last_encap_src) { 1671 register struct vif *vife; 1672 1673 vifp = viftable; 1674 vife = vifp + numvifs; 1675 last_encap_src = ip->ip_src.s_addr; 1676 last_encap_vif = 0; 1677 for ( ; vifp < vife; ++vifp) 1678 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1679 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1680 == VIFF_TUNNEL) 1681 last_encap_vif = vifp; 1682 break; 1683 } 1684 } 1685 if ((vifp = last_encap_vif) == 0) { 1686 last_encap_src = 0; 1687 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1688 m_freem(m); 1689 if (mrtdebug) 1690 log(LOG_DEBUG, "ip_mforward: no tunnel with %x\n", 1691 ntohl(ip->ip_src.s_addr)); 1692 return; 1693 } 1694 ifp = vifp->v_ifp; 1695 1696 if (hlen > IP_HDR_LEN) 1697 ip_stripoptions(m, (struct mbuf *) 0); 1698 m->m_data += IP_HDR_LEN; 1699 m->m_len -= IP_HDR_LEN; 1700 m->m_pkthdr.len -= IP_HDR_LEN; 1701 m->m_pkthdr.rcvif = ifp; 1702 1703 ifq = &ipintrq; 1704 s = splimp(); 1705 if (IF_QFULL(ifq)) { 1706 IF_DROP(ifq); 1707 m_freem(m); 1708 } else { 1709 IF_ENQUEUE(ifq, m); 1710 /* 1711 * normally we would need a "schednetisr(NETISR_IP)" 1712 * here but we were called by ip_input and it is going 1713 * to loop back & try to dequeue the packet we just 1714 * queued as soon as we return so we avoid the 1715 * unnecessary software interrrupt. 1716 */ 1717 } 1718 splx(s); 1719} 1720 1721/* 1722 * Token bucket filter module 1723 */ 1724 1725static void 1726tbf_control(vifp, m, ip, p_len) 1727 register struct vif *vifp; 1728 register struct mbuf *m; 1729 register struct ip *ip; 1730 register u_long p_len; 1731{ 1732 register struct tbf *t = vifp->v_tbf; 1733 1734 if (p_len > MAX_BKT_SIZE) { 1735 /* drop if packet is too large */ 1736 mrtstat.mrts_pkt2large++; 1737 m_freem(m); 1738 return; 1739 } 1740 1741 tbf_update_tokens(vifp); 1742 1743 /* if there are enough tokens, 1744 * and the queue is empty, 1745 * send this packet out 1746 */ 1747 1748 if (t->tbf_q_len == 0) { 1749 /* queue empty, send packet if enough tokens */ 1750 if (p_len <= t->tbf_n_tok) { 1751 t->tbf_n_tok -= p_len; 1752 tbf_send_packet(vifp, m); 1753 } else { 1754 /* queue packet and timeout till later */ 1755 tbf_queue(vifp, m); 1756 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1757 } 1758 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1759 /* finite queue length, so queue pkts and process queue */ 1760 tbf_queue(vifp, m); 1761 tbf_process_q(vifp); 1762 } else { 1763 /* queue length too much, try to dq and queue and process */ 1764 if (!tbf_dq_sel(vifp, ip)) { 1765 mrtstat.mrts_q_overflow++; 1766 m_freem(m); 1767 return; 1768 } else { 1769 tbf_queue(vifp, m); 1770 tbf_process_q(vifp); 1771 } 1772 } 1773 return; 1774} 1775 1776/* 1777 * adds a packet to the queue at the interface 1778 */ 1779static void 1780tbf_queue(vifp, m) 1781 register struct vif *vifp; 1782 register struct mbuf *m; 1783{ 1784 register int s = splnet(); 1785 register struct tbf *t = vifp->v_tbf; 1786 1787 if (t->tbf_t == NULL) { 1788 /* Queue was empty */ 1789 t->tbf_q = m; 1790 } else { 1791 /* Insert at tail */ 1792 t->tbf_t->m_act = m; 1793 } 1794 1795 /* Set new tail pointer */ 1796 t->tbf_t = m; 1797 1798#ifdef DIAGNOSTIC 1799 /* Make sure we didn't get fed a bogus mbuf */ 1800 if (m->m_act) 1801 panic("tbf_queue: m_act"); 1802#endif 1803 m->m_act = NULL; 1804 1805 t->tbf_q_len++; 1806 1807 splx(s); 1808} 1809 1810 1811/* 1812 * processes the queue at the interface 1813 */ 1814static void 1815tbf_process_q(vifp) 1816 register struct vif *vifp; 1817{ 1818 register struct mbuf *m; 1819 register int len; 1820 register int s = splnet(); 1821 register struct tbf *t = vifp->v_tbf; 1822 1823 /* loop through the queue at the interface and send as many packets 1824 * as possible 1825 */ 1826 while (t->tbf_q_len > 0) { 1827 m = t->tbf_q; 1828 1829 len = mtod(m, struct ip *)->ip_len; 1830 1831 /* determine if the packet can be sent */ 1832 if (len <= t->tbf_n_tok) { 1833 /* if so, 1834 * reduce no of tokens, dequeue the packet, 1835 * send the packet. 1836 */ 1837 t->tbf_n_tok -= len; 1838 1839 t->tbf_q = m->m_act; 1840 if (--t->tbf_q_len == 0) 1841 t->tbf_t = NULL; 1842 1843 m->m_act = NULL; 1844 tbf_send_packet(vifp, m); 1845 1846 } else break; 1847 } 1848 splx(s); 1849} 1850 1851static void 1852tbf_reprocess_q(xvifp) 1853 void *xvifp; 1854{ 1855 register struct vif *vifp = xvifp; 1856 if (ip_mrouter == NULL) 1857 return; 1858 1859 tbf_update_tokens(vifp); 1860 1861 tbf_process_q(vifp); 1862 1863 if (vifp->v_tbf->tbf_q_len) 1864 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1865} 1866 1867/* function that will selectively discard a member of the queue 1868 * based on the precedence value and the priority 1869 */ 1870static int 1871tbf_dq_sel(vifp, ip) 1872 register struct vif *vifp; 1873 register struct ip *ip; 1874{ 1875 register int s = splnet(); 1876 register u_int p; 1877 register struct mbuf *m, *last; 1878 register struct mbuf **np; 1879 register struct tbf *t = vifp->v_tbf; 1880 1881 p = priority(vifp, ip); 1882 1883 np = &t->tbf_q; 1884 last = NULL; 1885 while ((m = *np) != NULL) { 1886 if (p > priority(vifp, mtod(m, struct ip *))) { 1887 *np = m->m_act; 1888 /* If we're removing the last packet, fix the tail pointer */ 1889 if (m == t->tbf_t) 1890 t->tbf_t = last; 1891 m_freem(m); 1892 /* it's impossible for the queue to be empty, but 1893 * we check anyway. */ 1894 if (--t->tbf_q_len == 0) 1895 t->tbf_t = NULL; 1896 splx(s); 1897 mrtstat.mrts_drop_sel++; 1898 return(1); 1899 } 1900 np = &m->m_act; 1901 last = m; 1902 } 1903 splx(s); 1904 return(0); 1905} 1906 1907static void 1908tbf_send_packet(vifp, m) 1909 register struct vif *vifp; 1910 register struct mbuf *m; 1911{ 1912 struct ip_moptions imo; 1913 int error; 1914 int s = splnet(); 1915 1916 if (vifp->v_flags & VIFF_TUNNEL) { 1917 /* If tunnel options */ 1918 ip_output(m, (struct mbuf *)0, (struct route *)0, 1919 IP_FORWARDING, (struct ip_moptions *)0); 1920 } else { 1921 imo.imo_multicast_ifp = vifp->v_ifp; 1922 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1923 imo.imo_multicast_loop = 1; 1924 imo.imo_multicast_vif = -1; 1925 1926 error = ip_output(m, (struct mbuf *)0, (struct route *)0, 1927 IP_FORWARDING, &imo); 1928 1929 if (mrtdebug & DEBUG_XMIT) 1930 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1931 vifp - viftable, error); 1932 } 1933 splx(s); 1934} 1935 1936/* determine the current time and then 1937 * the elapsed time (between the last time and time now) 1938 * in milliseconds & update the no. of tokens in the bucket 1939 */ 1940static void 1941tbf_update_tokens(vifp) 1942 register struct vif *vifp; 1943{ 1944 struct timeval tp; 1945 register u_long tm; 1946 register int s = splnet(); 1947 register struct tbf *t = vifp->v_tbf; 1948 1949 GET_TIME(tp); 1950 1951 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1952 1953 /* 1954 * This formula is actually 1955 * "time in seconds" * "bytes/second". 1956 * 1957 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1958 * 1959 * The (1000/1024) was introduced in add_vif to optimize 1960 * this divide into a shift. 1961 */ 1962 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1963 t->tbf_last_pkt_t = tp; 1964 1965 if (t->tbf_n_tok > MAX_BKT_SIZE) 1966 t->tbf_n_tok = MAX_BKT_SIZE; 1967 1968 splx(s); 1969} 1970 1971static int 1972priority(vifp, ip) 1973 register struct vif *vifp; 1974 register struct ip *ip; 1975{ 1976 register int prio; 1977 1978 /* temporary hack; may add general packet classifier some day */ 1979 1980 /* 1981 * The UDP port space is divided up into four priority ranges: 1982 * [0, 16384) : unclassified - lowest priority 1983 * [16384, 32768) : audio - highest priority 1984 * [32768, 49152) : whiteboard - medium priority 1985 * [49152, 65536) : video - low priority 1986 */ 1987 if (ip->ip_p == IPPROTO_UDP) { 1988 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1989 switch (ntohs(udp->uh_dport) & 0xc000) { 1990 case 0x4000: 1991 prio = 70; 1992 break; 1993 case 0x8000: 1994 prio = 60; 1995 break; 1996 case 0xc000: 1997 prio = 55; 1998 break; 1999 default: 2000 prio = 50; 2001 break; 2002 } 2003 if (tbfdebug > 1) 2004 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio); 2005 } else { 2006 prio = 50; 2007 } 2008 return prio; 2009} 2010 2011/* 2012 * End of token bucket filter modifications 2013 */ 2014 2015int 2016ip_rsvp_vif_init(so, m) 2017 struct socket *so; 2018 struct mbuf *m; 2019{ 2020 int i; 2021 register int s; 2022 2023 if (rsvpdebug) 2024 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2025 so->so_type, so->so_proto->pr_protocol); 2026 2027 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2028 return EOPNOTSUPP; 2029 2030 /* Check mbuf. */ 2031 if (m == NULL || m->m_len != sizeof(int)) { 2032 return EINVAL; 2033 } 2034 i = *(mtod(m, int *)); 2035 2036 if (rsvpdebug) 2037 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on); 2038 2039 s = splnet(); 2040 2041 /* Check vif. */ 2042 if (!legal_vif_num(i)) { 2043 splx(s); 2044 return EADDRNOTAVAIL; 2045 } 2046 2047 /* Check if socket is available. */ 2048 if (viftable[i].v_rsvpd != NULL) { 2049 splx(s); 2050 return EADDRINUSE; 2051 } 2052 2053 viftable[i].v_rsvpd = so; 2054 /* This may seem silly, but we need to be sure we don't over-increment 2055 * the RSVP counter, in case something slips up. 2056 */ 2057 if (!viftable[i].v_rsvp_on) { 2058 viftable[i].v_rsvp_on = 1; 2059 rsvp_on++; 2060 } 2061 2062 splx(s); 2063 return 0; 2064} 2065 2066int 2067ip_rsvp_vif_done(so, m) 2068 struct socket *so; 2069 struct mbuf *m; 2070{ 2071 int i; 2072 register int s; 2073 2074 if (rsvpdebug) 2075 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2076 so->so_type, so->so_proto->pr_protocol); 2077 2078 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2079 return EOPNOTSUPP; 2080 2081 /* Check mbuf. */ 2082 if (m == NULL || m->m_len != sizeof(int)) { 2083 return EINVAL; 2084 } 2085 i = *(mtod(m, int *)); 2086 2087 s = splnet(); 2088 2089 /* Check vif. */ 2090 if (!legal_vif_num(i)) { 2091 splx(s); 2092 return EADDRNOTAVAIL; 2093 } 2094 2095 if (rsvpdebug) 2096 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n", 2097 viftable[i].v_rsvpd, so); 2098 2099 viftable[i].v_rsvpd = NULL; 2100 /* This may seem silly, but we need to be sure we don't over-decrement 2101 * the RSVP counter, in case something slips up. 2102 */ 2103 if (viftable[i].v_rsvp_on) { 2104 viftable[i].v_rsvp_on = 0; 2105 rsvp_on--; 2106 } 2107 2108 splx(s); 2109 return 0; 2110} 2111 2112void 2113ip_rsvp_force_done(so) 2114 struct socket *so; 2115{ 2116 int vifi; 2117 register int s; 2118 2119 /* Don't bother if it is not the right type of socket. */ 2120 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2121 return; 2122 2123 s = splnet(); 2124 2125 /* The socket may be attached to more than one vif...this 2126 * is perfectly legal. 2127 */ 2128 for (vifi = 0; vifi < numvifs; vifi++) { 2129 if (viftable[vifi].v_rsvpd == so) { 2130 viftable[vifi].v_rsvpd = NULL; 2131 /* This may seem silly, but we need to be sure we don't 2132 * over-decrement the RSVP counter, in case something slips up. 2133 */ 2134 if (viftable[vifi].v_rsvp_on) { 2135 viftable[vifi].v_rsvp_on = 0; 2136 rsvp_on--; 2137 } 2138 } 2139 } 2140 2141 splx(s); 2142 return; 2143} 2144 2145void 2146rsvp_input(m, iphlen) 2147 struct mbuf *m; 2148 int iphlen; 2149{ 2150 int vifi; 2151 register struct ip *ip = mtod(m, struct ip *); 2152 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2153 register int s; 2154 struct ifnet *ifp; 2155 2156 if (rsvpdebug) 2157 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2158 2159 /* Can still get packets with rsvp_on = 0 if there is a local member 2160 * of the group to which the RSVP packet is addressed. But in this 2161 * case we want to throw the packet away. 2162 */ 2163 if (!rsvp_on) { 2164 m_freem(m); 2165 return; 2166 } 2167 2168 /* If the old-style non-vif-associated socket is set, then use 2169 * it and ignore the new ones. 2170 */ 2171 if (ip_rsvpd != NULL) { 2172 if (rsvpdebug) 2173 printf("rsvp_input: Sending packet up old-style socket\n"); 2174 rip_input(m); 2175 return; 2176 } 2177 2178 s = splnet(); 2179 2180 if (rsvpdebug) 2181 printf("rsvp_input: check vifs\n"); 2182 2183#ifdef DIAGNOSTIC 2184 if (!(m->m_flags & M_PKTHDR)) 2185 panic("rsvp_input no hdr"); 2186#endif 2187 2188 ifp = m->m_pkthdr.rcvif; 2189 /* Find which vif the packet arrived on. */ 2190 for (vifi = 0; vifi < numvifs; vifi++) { 2191 if (viftable[vifi].v_ifp == ifp) 2192 break; 2193 } 2194 2195 if (vifi == numvifs) { 2196 /* Can't find vif packet arrived on. Drop packet. */ 2197 if (rsvpdebug) 2198 printf("rsvp_input: Can't find vif for packet...dropping it.\n"); 2199 m_freem(m); 2200 splx(s); 2201 return; 2202 } 2203 2204 if (rsvpdebug) 2205 printf("rsvp_input: check socket\n"); 2206 2207 if (viftable[vifi].v_rsvpd == NULL) { 2208 /* drop packet, since there is no specific socket for this 2209 * interface */ 2210 if (rsvpdebug) 2211 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2212 m_freem(m); 2213 splx(s); 2214 return; 2215 } 2216 rsvp_src.sin_addr = ip->ip_src; 2217 2218 if (rsvpdebug && m) 2219 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2220 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2221 2222 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2223 if (rsvpdebug) 2224 printf("rsvp_input: Failed to append to socket\n"); 2225 else 2226 if (rsvpdebug) 2227 printf("rsvp_input: send packet up\n"); 2228 2229 splx(s); 2230} 2231 2232#ifdef MROUTE_LKM 2233#include <sys/conf.h> 2234#include <sys/exec.h> 2235#include <sys/sysent.h> 2236#include <sys/lkm.h> 2237 2238MOD_MISC("ip_mroute_mod") 2239 2240static int 2241ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd) 2242{ 2243 int i; 2244 struct lkm_misc *args = lkmtp->private.lkm_misc; 2245 int err = 0; 2246 2247 switch(cmd) { 2248 static int (*old_ip_mrouter_cmd)(); 2249 static int (*old_ip_mrouter_done)(); 2250 static int (*old_ip_mforward)(); 2251 static int (*old_mrt_ioctl)(); 2252 static void (*old_proto4_input)(); 2253 static int (*old_legal_vif_num)(); 2254 extern struct protosw inetsw[]; 2255 2256 case LKM_E_LOAD: 2257 if(lkmexists(lkmtp) || ip_mrtproto) 2258 return(EEXIST); 2259 old_ip_mrouter_cmd = ip_mrouter_cmd; 2260 ip_mrouter_cmd = X_ip_mrouter_cmd; 2261 old_ip_mrouter_done = ip_mrouter_done; 2262 ip_mrouter_done = X_ip_mrouter_done; 2263 old_ip_mforward = ip_mforward; 2264 ip_mforward = X_ip_mforward; 2265 old_mrt_ioctl = mrt_ioctl; 2266 mrt_ioctl = X_mrt_ioctl; 2267 old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input; 2268 inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input; 2269 old_legal_vif_num = legal_vif_num; 2270 legal_vif_num = X_legal_vif_num; 2271 ip_mrtproto = IGMP_DVMRP; 2272 2273 printf("\nIP multicast routing loaded\n"); 2274 break; 2275 2276 case LKM_E_UNLOAD: 2277 if (ip_mrouter) 2278 return EINVAL; 2279 2280 ip_mrouter_cmd = old_ip_mrouter_cmd; 2281 ip_mrouter_done = old_ip_mrouter_done; 2282 ip_mforward = old_ip_mforward; 2283 mrt_ioctl = old_mrt_ioctl; 2284 inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input; 2285 legal_vif_num = old_legal_vif_num; 2286 ip_mrtproto = 0; 2287 break; 2288 2289 default: 2290 err = EINVAL; 2291 break; 2292 } 2293 2294 return(err); 2295} 2296 2297int 2298ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) { 2299 DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle, 2300 nosys); 2301} 2302 2303#endif /* MROUTE_LKM */ 2304#endif /* MROUTING */ 2305