1 /* 2 * IP multicast forwarding procedures 3 * 4 * Written by David Waitzman, BBN Labs, August 1988. 5 * Modified by Steve Deering, Stanford, February 1989. 6 * Modified by Mark J. Steiglitz, Stanford, May, 1991 7 * Modified by Van Jacobson, LBL, January 1993 8 * Modified by Ajit Thyagarajan, PARC, August 1993 9 * Modified by Bill Fenner, PARC, April 1995 10 * 11 * MROUTING Revision: 3.5 12 * $FreeBSD$ 13 */ 14 15 #include "opt_mrouting.h" 16 #include "opt_random_ip_id.h" 17 18 #include <sys/param.h> 19 #include <sys/systm.h> 20 #include <sys/malloc.h> 21 #include <sys/mbuf.h> 22 #include <sys/socket.h> 23 #include <sys/socketvar.h> 24 #include <sys/protosw.h> 25 #include <sys/time.h> 26 #include <sys/kernel.h> 27 #include <sys/sysctl.h> 28 #include <sys/sockio.h> 29 #include <sys/syslog.h> 30 #include <net/if.h> 31 #include <net/route.h> 32 #include <netinet/in.h> 33 #include <netinet/in_systm.h> 34 #include <netinet/ip.h> 35 #include <netinet/ip_var.h> 36 #include <netinet/in_var.h> 37 #include <netinet/igmp.h> 38 #include <netinet/ip_encap.h> 39 #include <netinet/ip_mroute.h> 40 #include <netinet/udp.h> 41 #include <machine/in_cksum.h> 42 43 #ifndef MROUTING 44 extern u_long _ip_mcast_src __P((int vifi)); 45 extern int _ip_mforward __P((struct ip *ip, struct ifnet *ifp, 46 struct mbuf *m, struct ip_moptions *imo)); 47 extern int _ip_mrouter_done __P((void)); 48 extern int _ip_mrouter_get __P((struct socket *so, struct sockopt *sopt)); 49 extern int _ip_mrouter_set __P((struct socket *so, struct sockopt *sopt)); 50 extern int _mrt_ioctl __P((int req, caddr_t data)); 51 52 /* 53 * Dummy routines and globals used when multicast routing is not compiled in. 54 */ 55 56 struct socket *ip_mrouter = NULL; 57 u_int rsvpdebug = 0; 58 59 int 60 _ip_mrouter_set(so, sopt) 61 struct socket *so; 62 struct sockopt *sopt; 63 { 64 return(EOPNOTSUPP); 65 } 66 67 int (*ip_mrouter_set)(struct socket *, struct sockopt *) = _ip_mrouter_set; 68 69 70 int 71 _ip_mrouter_get(so, sopt) 72 struct socket *so; 73 struct sockopt *sopt; 74 { 75 return(EOPNOTSUPP); 76 } 77 78 int (*ip_mrouter_get)(struct socket *, struct sockopt *) = _ip_mrouter_get; 79 80 int 81 _ip_mrouter_done() 82 { 83 return(0); 84 } 85 86 int (*ip_mrouter_done)(void) = _ip_mrouter_done; 87 88 int 89 _ip_mforward(ip, ifp, m, imo) 90 struct ip *ip; 91 struct ifnet *ifp; 92 struct mbuf *m; 93 struct ip_moptions *imo; 94 { 95 return(0); 96 } 97 98 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 99 struct ip_moptions *) = _ip_mforward; 100 101 int 102 _mrt_ioctl(int req, caddr_t data) 103 { 104 return EOPNOTSUPP; 105 } 106 107 int (*mrt_ioctl)(int, caddr_t) = _mrt_ioctl; 108 109 void 110 rsvp_input(m, off) /* XXX must fixup manually */ 111 struct mbuf *m; 112 int off; 113 { 114 /* Can still get packets with rsvp_on = 0 if there is a local member 115 * of the group to which the RSVP packet is addressed. But in this 116 * case we want to throw the packet away. 117 */ 118 if (!rsvp_on) { 119 m_freem(m); 120 return; 121 } 122 123 if (ip_rsvpd != NULL) { 124 if (rsvpdebug) 125 printf("rsvp_input: Sending packet up old-style socket\n"); 126 rip_input(m, off); 127 return; 128 } 129 /* Drop the packet */ 130 m_freem(m); 131 } 132 133 int (*legal_vif_num)(int) = 0; 134 135 /* 136 * This should never be called, since IP_MULTICAST_VIF should fail, but 137 * just in case it does get called, the code a little lower in ip_output 138 * will assign the packet a local address. 139 */ 140 u_long 141 _ip_mcast_src(int vifi) { return INADDR_ANY; } 142 u_long (*ip_mcast_src)(int) = _ip_mcast_src; 143 144 int 145 ip_rsvp_vif_init(so, sopt) 146 struct socket *so; 147 struct sockopt *sopt; 148 { 149 return(EINVAL); 150 } 151 152 int 153 ip_rsvp_vif_done(so, sopt) 154 struct socket *so; 155 struct sockopt *sopt; 156 { 157 return(EINVAL); 158 } 159 160 void 161 ip_rsvp_force_done(so) 162 struct socket *so; 163 { 164 return; 165 } 166 167 #else /* MROUTING */ 168 169 #define M_HASCL(m) ((m)->m_flags & M_EXT) 170 171 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 172 173 #ifndef MROUTE_KLD 174 /* The socket used to communicate with the multicast routing daemon. */ 175 struct socket *ip_mrouter = NULL; 176 #endif 177 178 #if defined(MROUTING) || defined(MROUTE_KLD) 179 static struct mrtstat mrtstat; 180 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 181 &mrtstat, mrtstat, "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)"); 182 #endif 183 184 static struct mfc *mfctable[MFCTBLSIZ]; 185 static u_char nexpire[MFCTBLSIZ]; 186 static struct vif viftable[MAXVIFS]; 187 static u_int mrtdebug = 0; /* debug level */ 188 #define DEBUG_MFC 0x02 189 #define DEBUG_FORWARD 0x04 190 #define DEBUG_EXPIRE 0x08 191 #define DEBUG_XMIT 0x10 192 static u_int tbfdebug = 0; /* tbf debug level */ 193 static u_int rsvpdebug = 0; /* rsvp debug level */ 194 195 static struct callout_handle expire_upcalls_ch; 196 197 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 198 #define UPCALL_EXPIRE 6 /* number of timeouts */ 199 200 /* 201 * Define the token bucket filter structures 202 * tbftable -> each vif has one of these for storing info 203 */ 204 205 static struct tbf tbftable[MAXVIFS]; 206 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 207 208 /* 209 * 'Interfaces' associated with decapsulator (so we can tell 210 * packets that went through it from ones that get reflected 211 * by a broken gateway). These interfaces are never linked into 212 * the system ifnet list & no routes point to them. I.e., packets 213 * can't be sent this way. They only exist as a placeholder for 214 * multicast source verification. 215 */ 216 static struct ifnet multicast_decap_if[MAXVIFS]; 217 218 #define ENCAP_TTL 64 219 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 220 221 /* prototype IP hdr for encapsulated packets */ 222 static struct ip multicast_encap_iphdr = { 223 #if BYTE_ORDER == LITTLE_ENDIAN 224 sizeof(struct ip) >> 2, IPVERSION, 225 #else 226 IPVERSION, sizeof(struct ip) >> 2, 227 #endif 228 0, /* tos */ 229 sizeof(struct ip), /* total length */ 230 0, /* id */ 231 0, /* frag offset */ 232 ENCAP_TTL, ENCAP_PROTO, 233 0, /* checksum */ 234 }; 235 236 /* 237 * Private variables. 238 */ 239 static vifi_t numvifs = 0; 240 static const struct encaptab *encap_cookie = NULL; 241 242 /* 243 * one-back cache used by mroute_encapcheck to locate a tunnel's vif 244 * given a datagram's src ip address. 245 */ 246 static u_long last_encap_src; 247 static struct vif *last_encap_vif; 248 249 static u_long X_ip_mcast_src __P((int vifi)); 250 static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo)); 251 static int X_ip_mrouter_done __P((void)); 252 static int X_ip_mrouter_get __P((struct socket *so, struct sockopt *m)); 253 static int X_ip_mrouter_set __P((struct socket *so, struct sockopt *m)); 254 static int X_legal_vif_num __P((int vif)); 255 static int X_mrt_ioctl __P((int cmd, caddr_t data)); 256 257 static int get_sg_cnt(struct sioc_sg_req *); 258 static int get_vif_cnt(struct sioc_vif_req *); 259 static int ip_mrouter_init(struct socket *, int); 260 static int add_vif(struct vifctl *); 261 static int del_vif(vifi_t); 262 static int add_mfc(struct mfcctl *); 263 static int del_mfc(struct mfcctl *); 264 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 265 static int set_assert(int); 266 static void expire_upcalls(void *); 267 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, 268 vifi_t); 269 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 270 static void encap_send(struct ip *, struct vif *, struct mbuf *); 271 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 272 static void tbf_queue(struct vif *, struct mbuf *); 273 static void tbf_process_q(struct vif *); 274 static void tbf_reprocess_q(void *); 275 static int tbf_dq_sel(struct vif *, struct ip *); 276 static void tbf_send_packet(struct vif *, struct mbuf *); 277 static void tbf_update_tokens(struct vif *); 278 static int priority(struct vif *, struct ip *); 279 280 /* 281 * whether or not special PIM assert processing is enabled. 282 */ 283 static int pim_assert; 284 /* 285 * Rate limit for assert notification messages, in usec 286 */ 287 #define ASSERT_MSG_TIME 3000000 288 289 /* 290 * Hash function for a source, group entry 291 */ 292 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 293 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 294 295 /* 296 * Find a route for a given origin IP address and Multicast group address 297 * Type of service parameter to be added in the future!!! 298 */ 299 300 #define MFCFIND(o, g, rt) { \ 301 register struct mfc *_rt = mfctable[MFCHASH(o,g)]; \ 302 rt = NULL; \ 303 ++mrtstat.mrts_mfc_lookups; \ 304 while (_rt) { \ 305 if ((_rt->mfc_origin.s_addr == o) && \ 306 (_rt->mfc_mcastgrp.s_addr == g) && \ 307 (_rt->mfc_stall == NULL)) { \ 308 rt = _rt; \ 309 break; \ 310 } \ 311 _rt = _rt->mfc_next; \ 312 } \ 313 if (rt == NULL) { \ 314 ++mrtstat.mrts_mfc_misses; \ 315 } \ 316 } 317 318 319 /* 320 * Macros to compute elapsed time efficiently 321 * Borrowed from Van Jacobson's scheduling code 322 */ 323 #define TV_DELTA(a, b, delta) { \ 324 register int xxs; \ 325 \ 326 delta = (a).tv_usec - (b).tv_usec; \ 327 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 328 switch (xxs) { \ 329 case 2: \ 330 delta += 1000000; \ 331 /* fall through */ \ 332 case 1: \ 333 delta += 1000000; \ 334 break; \ 335 default: \ 336 delta += (1000000 * xxs); \ 337 } \ 338 } \ 339 } 340 341 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 342 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 343 344 #ifdef UPCALL_TIMING 345 u_long upcall_data[51]; 346 static void collate(struct timeval *); 347 #endif /* UPCALL_TIMING */ 348 349 350 /* 351 * Handle MRT setsockopt commands to modify the multicast routing tables. 352 */ 353 static int 354 X_ip_mrouter_set(so, sopt) 355 struct socket *so; 356 struct sockopt *sopt; 357 { 358 int error, optval; 359 vifi_t vifi; 360 struct vifctl vifc; 361 struct mfcctl mfc; 362 363 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 364 return (EPERM); 365 366 error = 0; 367 switch (sopt->sopt_name) { 368 case MRT_INIT: 369 error = sooptcopyin(sopt, &optval, sizeof optval, 370 sizeof optval); 371 if (error) 372 break; 373 error = ip_mrouter_init(so, optval); 374 break; 375 376 case MRT_DONE: 377 error = ip_mrouter_done(); 378 break; 379 380 case MRT_ADD_VIF: 381 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 382 if (error) 383 break; 384 error = add_vif(&vifc); 385 break; 386 387 case MRT_DEL_VIF: 388 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 389 if (error) 390 break; 391 error = del_vif(vifi); 392 break; 393 394 case MRT_ADD_MFC: 395 case MRT_DEL_MFC: 396 error = sooptcopyin(sopt, &mfc, sizeof mfc, sizeof mfc); 397 if (error) 398 break; 399 if (sopt->sopt_name == MRT_ADD_MFC) 400 error = add_mfc(&mfc); 401 else 402 error = del_mfc(&mfc); 403 break; 404 405 case MRT_ASSERT: 406 error = sooptcopyin(sopt, &optval, sizeof optval, 407 sizeof optval); 408 if (error) 409 break; 410 set_assert(optval); 411 break; 412 413 default: 414 error = EOPNOTSUPP; 415 break; 416 } 417 return (error); 418 } 419 420 #ifndef MROUTE_KLD 421 int (*ip_mrouter_set)(struct socket *, struct sockopt *) = X_ip_mrouter_set; 422 #endif 423 424 /* 425 * Handle MRT getsockopt commands 426 */ 427 static int 428 X_ip_mrouter_get(so, sopt) 429 struct socket *so; 430 struct sockopt *sopt; 431 { 432 int error; 433 static int version = 0x0305; /* !!! why is this here? XXX */ 434 435 switch (sopt->sopt_name) { 436 case MRT_VERSION: 437 error = sooptcopyout(sopt, &version, sizeof version); 438 break; 439 440 case MRT_ASSERT: 441 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 442 break; 443 default: 444 error = EOPNOTSUPP; 445 break; 446 } 447 return (error); 448 } 449 450 #ifndef MROUTE_KLD 451 int (*ip_mrouter_get)(struct socket *, struct sockopt *) = X_ip_mrouter_get; 452 #endif 453 454 /* 455 * Handle ioctl commands to obtain information from the cache 456 */ 457 static int 458 X_mrt_ioctl(cmd, data) 459 int cmd; 460 caddr_t data; 461 { 462 int error = 0; 463 464 switch (cmd) { 465 case (SIOCGETVIFCNT): 466 return (get_vif_cnt((struct sioc_vif_req *)data)); 467 break; 468 case (SIOCGETSGCNT): 469 return (get_sg_cnt((struct sioc_sg_req *)data)); 470 break; 471 default: 472 return (EINVAL); 473 break; 474 } 475 return error; 476 } 477 478 #ifndef MROUTE_KLD 479 int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl; 480 #endif 481 482 /* 483 * returns the packet, byte, rpf-failure count for the source group provided 484 */ 485 static int 486 get_sg_cnt(req) 487 register struct sioc_sg_req *req; 488 { 489 register struct mfc *rt; 490 int s; 491 492 s = splnet(); 493 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 494 splx(s); 495 if (rt != NULL) { 496 req->pktcnt = rt->mfc_pkt_cnt; 497 req->bytecnt = rt->mfc_byte_cnt; 498 req->wrong_if = rt->mfc_wrong_if; 499 } else 500 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 501 502 return 0; 503 } 504 505 /* 506 * returns the input and output packet and byte counts on the vif provided 507 */ 508 static int 509 get_vif_cnt(req) 510 register struct sioc_vif_req *req; 511 { 512 register vifi_t vifi = req->vifi; 513 514 if (vifi >= numvifs) return EINVAL; 515 516 req->icount = viftable[vifi].v_pkt_in; 517 req->ocount = viftable[vifi].v_pkt_out; 518 req->ibytes = viftable[vifi].v_bytes_in; 519 req->obytes = viftable[vifi].v_bytes_out; 520 521 return 0; 522 } 523 524 /* 525 * Enable multicast routing 526 */ 527 static int 528 ip_mrouter_init(so, version) 529 struct socket *so; 530 int version; 531 { 532 if (mrtdebug) 533 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 534 so->so_type, so->so_proto->pr_protocol); 535 536 if (so->so_type != SOCK_RAW || 537 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP; 538 539 if (version != 1) 540 return ENOPROTOOPT; 541 542 if (ip_mrouter != NULL) return EADDRINUSE; 543 544 ip_mrouter = so; 545 546 bzero((caddr_t)mfctable, sizeof(mfctable)); 547 bzero((caddr_t)nexpire, sizeof(nexpire)); 548 549 pim_assert = 0; 550 551 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 552 553 if (mrtdebug) 554 log(LOG_DEBUG, "ip_mrouter_init\n"); 555 556 return 0; 557 } 558 559 /* 560 * Disable multicast routing 561 */ 562 static int 563 X_ip_mrouter_done() 564 { 565 vifi_t vifi; 566 int i; 567 struct ifnet *ifp; 568 struct ifreq ifr; 569 struct mfc *rt; 570 struct rtdetq *rte; 571 int s; 572 573 s = splnet(); 574 575 /* 576 * For each phyint in use, disable promiscuous reception of all IP 577 * multicasts. 578 */ 579 for (vifi = 0; vifi < numvifs; vifi++) { 580 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 581 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 582 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 583 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr 584 = INADDR_ANY; 585 ifp = viftable[vifi].v_ifp; 586 if_allmulti(ifp, 0); 587 } 588 } 589 bzero((caddr_t)tbftable, sizeof(tbftable)); 590 bzero((caddr_t)viftable, sizeof(viftable)); 591 numvifs = 0; 592 pim_assert = 0; 593 594 untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch); 595 596 /* 597 * Free all multicast forwarding cache entries. 598 */ 599 for (i = 0; i < MFCTBLSIZ; i++) { 600 for (rt = mfctable[i]; rt != NULL; ) { 601 struct mfc *nr = rt->mfc_next; 602 603 for (rte = rt->mfc_stall; rte != NULL; ) { 604 struct rtdetq *n = rte->next; 605 606 m_freem(rte->m); 607 free(rte, M_MRTABLE); 608 rte = n; 609 } 610 free(rt, M_MRTABLE); 611 rt = nr; 612 } 613 } 614 615 bzero((caddr_t)mfctable, sizeof(mfctable)); 616 617 /* 618 * Reset de-encapsulation cache 619 */ 620 last_encap_src = 0; 621 last_encap_vif = NULL; 622 if (encap_cookie) { 623 encap_detach(encap_cookie); 624 encap_cookie = NULL; 625 } 626 627 ip_mrouter = NULL; 628 629 splx(s); 630 631 if (mrtdebug) 632 log(LOG_DEBUG, "ip_mrouter_done\n"); 633 634 return 0; 635 } 636 637 #ifndef MROUTE_KLD 638 int (*ip_mrouter_done)(void) = X_ip_mrouter_done; 639 #endif 640 641 /* 642 * Set PIM assert processing global 643 */ 644 static int 645 set_assert(i) 646 int i; 647 { 648 if ((i != 1) && (i != 0)) 649 return EINVAL; 650 651 pim_assert = i; 652 653 return 0; 654 } 655 656 /* 657 * Decide if a packet is from a tunnelled peer. 658 * Return 0 if not, 64 if so. 659 */ 660 static int 661 mroute_encapcheck(const struct mbuf *m, int off, int proto, void *arg) 662 { 663 struct ip *ip = mtod(m, struct ip *); 664 int hlen = ip->ip_hl << 2; 665 register struct vif *vifp; 666 667 /* 668 * don't claim the packet if it's not to a multicast destination or if 669 * we don't have an encapsulating tunnel with the source. 670 * Note: This code assumes that the remote site IP address 671 * uniquely identifies the tunnel (i.e., that this site has 672 * at most one tunnel with the remote site). 673 */ 674 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) { 675 return 0; 676 } 677 if (ip->ip_src.s_addr != last_encap_src) { 678 register struct vif *vife; 679 680 vifp = viftable; 681 vife = vifp + numvifs; 682 last_encap_src = ip->ip_src.s_addr; 683 last_encap_vif = 0; 684 for ( ; vifp < vife; ++vifp) 685 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 686 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 687 == VIFF_TUNNEL) 688 last_encap_vif = vifp; 689 break; 690 } 691 } 692 if ((vifp = last_encap_vif) == 0) { 693 last_encap_src = 0; 694 return 0; 695 } 696 return 64; 697 } 698 699 /* 700 * De-encapsulate a packet and feed it back through ip input (this 701 * routine is called whenever IP gets a packet that mroute_encap_func() 702 * claimed). 703 */ 704 static void 705 mroute_encap_input(struct mbuf *m, int off) 706 { 707 struct ip *ip = mtod(m, struct ip *); 708 int hlen = ip->ip_hl << 2; 709 710 if (hlen > sizeof(struct ip)) 711 ip_stripoptions(m, (struct mbuf *) 0); 712 m->m_data += sizeof(struct ip); 713 m->m_len -= sizeof(struct ip); 714 m->m_pkthdr.len -= sizeof(struct ip); 715 716 m->m_pkthdr.rcvif = last_encap_vif->v_ifp; 717 718 (void) IF_HANDOFF(&ipintrq, m, NULL); 719 /* 720 * normally we would need a "schednetisr(NETISR_IP)" 721 * here but we were called by ip_input and it is going 722 * to loop back & try to dequeue the packet we just 723 * queued as soon as we return so we avoid the 724 * unnecessary software interrrupt. 725 */ 726 } 727 728 extern struct domain inetdomain; 729 static struct protosw mroute_encap_protosw = 730 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR, 731 mroute_encap_input, 0, 0, rip_ctloutput, 732 0, 733 0, 0, 0, 0, 734 &rip_usrreqs 735 }; 736 737 /* 738 * Add a vif to the vif table 739 */ 740 static int 741 add_vif(vifcp) 742 register struct vifctl *vifcp; 743 { 744 register struct vif *vifp = viftable + vifcp->vifc_vifi; 745 static struct sockaddr_in sin = {sizeof sin, AF_INET}; 746 struct ifaddr *ifa; 747 struct ifnet *ifp; 748 int error, s; 749 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 750 751 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL; 752 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE; 753 754 /* Find the interface with an address in AF_INET family */ 755 sin.sin_addr = vifcp->vifc_lcl_addr; 756 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 757 if (ifa == 0) return EADDRNOTAVAIL; 758 ifp = ifa->ifa_ifp; 759 760 if (vifcp->vifc_flags & VIFF_TUNNEL) { 761 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 762 /* 763 * An encapsulating tunnel is wanted. Tell 764 * mroute_encap_input() to start paying attention 765 * to encapsulated packets. 766 */ 767 if (encap_cookie == NULL) { 768 encap_cookie = encap_attach_func(AF_INET, -1, 769 mroute_encapcheck, 770 (struct protosw *)&mroute_encap_protosw, NULL); 771 772 if (encap_cookie == NULL) { 773 printf("ip_mroute: unable to attach encap\n"); 774 return (EIO); /* XXX */ 775 } 776 for (s = 0; s < MAXVIFS; ++s) { 777 multicast_decap_if[s].if_name = "mdecap"; 778 multicast_decap_if[s].if_unit = s; 779 } 780 } 781 /* 782 * Set interface to fake encapsulator interface 783 */ 784 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 785 /* 786 * Prepare cached route entry 787 */ 788 bzero(&vifp->v_route, sizeof(vifp->v_route)); 789 } else { 790 log(LOG_ERR, "source routed tunnels not supported\n"); 791 return EOPNOTSUPP; 792 } 793 } else { 794 /* Make sure the interface supports multicast */ 795 if ((ifp->if_flags & IFF_MULTICAST) == 0) 796 return EOPNOTSUPP; 797 798 /* Enable promiscuous reception of all IP multicasts from the if */ 799 s = splnet(); 800 error = if_allmulti(ifp, 1); 801 splx(s); 802 if (error) 803 return error; 804 } 805 806 s = splnet(); 807 /* define parameters for the tbf structure */ 808 vifp->v_tbf = v_tbf; 809 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 810 vifp->v_tbf->tbf_n_tok = 0; 811 vifp->v_tbf->tbf_q_len = 0; 812 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 813 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 814 815 vifp->v_flags = vifcp->vifc_flags; 816 vifp->v_threshold = vifcp->vifc_threshold; 817 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 818 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 819 vifp->v_ifp = ifp; 820 /* scaling up here allows division by 1024 in critical code */ 821 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 822 vifp->v_rsvp_on = 0; 823 vifp->v_rsvpd = NULL; 824 /* initialize per vif pkt counters */ 825 vifp->v_pkt_in = 0; 826 vifp->v_pkt_out = 0; 827 vifp->v_bytes_in = 0; 828 vifp->v_bytes_out = 0; 829 splx(s); 830 831 /* Adjust numvifs up if the vifi is higher than numvifs */ 832 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 833 834 if (mrtdebug) 835 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n", 836 vifcp->vifc_vifi, 837 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 838 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 839 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 840 vifcp->vifc_threshold, 841 vifcp->vifc_rate_limit); 842 843 return 0; 844 } 845 846 /* 847 * Delete a vif from the vif table 848 */ 849 static int 850 del_vif(vifi) 851 vifi_t vifi; 852 { 853 register struct vif *vifp = &viftable[vifi]; 854 register struct mbuf *m; 855 struct ifnet *ifp; 856 struct ifreq ifr; 857 int s; 858 859 if (vifi >= numvifs) return EINVAL; 860 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL; 861 862 s = splnet(); 863 864 if (!(vifp->v_flags & VIFF_TUNNEL)) { 865 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 866 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 867 ifp = vifp->v_ifp; 868 if_allmulti(ifp, 0); 869 } 870 871 if (vifp == last_encap_vif) { 872 last_encap_vif = 0; 873 last_encap_src = 0; 874 } 875 876 /* 877 * Free packets queued at the interface 878 */ 879 while (vifp->v_tbf->tbf_q) { 880 m = vifp->v_tbf->tbf_q; 881 vifp->v_tbf->tbf_q = m->m_act; 882 m_freem(m); 883 } 884 885 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 886 bzero((caddr_t)vifp, sizeof (*vifp)); 887 888 if (mrtdebug) 889 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 890 891 /* Adjust numvifs down */ 892 for (vifi = numvifs; vifi > 0; vifi--) 893 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break; 894 numvifs = vifi; 895 896 splx(s); 897 898 return 0; 899 } 900 901 /* 902 * Add an mfc entry 903 */ 904 static int 905 add_mfc(mfccp) 906 struct mfcctl *mfccp; 907 { 908 struct mfc *rt; 909 u_long hash; 910 struct rtdetq *rte; 911 register u_short nstl; 912 int s; 913 int i; 914 915 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 916 917 /* If an entry already exists, just update the fields */ 918 if (rt) { 919 if (mrtdebug & DEBUG_MFC) 920 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 921 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 922 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 923 mfccp->mfcc_parent); 924 925 s = splnet(); 926 rt->mfc_parent = mfccp->mfcc_parent; 927 for (i = 0; i < numvifs; i++) 928 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 929 splx(s); 930 return 0; 931 } 932 933 /* 934 * Find the entry for which the upcall was made and update 935 */ 936 s = splnet(); 937 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 938 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 939 940 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 941 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 942 (rt->mfc_stall != NULL)) { 943 944 if (nstl++) 945 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 946 "multiple kernel entries", 947 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 948 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 949 mfccp->mfcc_parent, (void *)rt->mfc_stall); 950 951 if (mrtdebug & DEBUG_MFC) 952 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 953 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 954 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 955 mfccp->mfcc_parent, (void *)rt->mfc_stall); 956 957 rt->mfc_origin = mfccp->mfcc_origin; 958 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 959 rt->mfc_parent = mfccp->mfcc_parent; 960 for (i = 0; i < numvifs; i++) 961 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 962 /* initialize pkt counters per src-grp */ 963 rt->mfc_pkt_cnt = 0; 964 rt->mfc_byte_cnt = 0; 965 rt->mfc_wrong_if = 0; 966 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 967 968 rt->mfc_expire = 0; /* Don't clean this guy up */ 969 nexpire[hash]--; 970 971 /* free packets Qed at the end of this entry */ 972 for (rte = rt->mfc_stall; rte != NULL; ) { 973 struct rtdetq *n = rte->next; 974 975 ip_mdq(rte->m, rte->ifp, rt, -1); 976 m_freem(rte->m); 977 #ifdef UPCALL_TIMING 978 collate(&(rte->t)); 979 #endif /* UPCALL_TIMING */ 980 free(rte, M_MRTABLE); 981 rte = n; 982 } 983 rt->mfc_stall = NULL; 984 } 985 } 986 987 /* 988 * It is possible that an entry is being inserted without an upcall 989 */ 990 if (nstl == 0) { 991 if (mrtdebug & DEBUG_MFC) 992 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 993 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 994 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 995 mfccp->mfcc_parent); 996 997 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 998 999 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1000 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 1001 1002 rt->mfc_origin = mfccp->mfcc_origin; 1003 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1004 rt->mfc_parent = mfccp->mfcc_parent; 1005 for (i = 0; i < numvifs; i++) 1006 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1007 /* initialize pkt counters per src-grp */ 1008 rt->mfc_pkt_cnt = 0; 1009 rt->mfc_byte_cnt = 0; 1010 rt->mfc_wrong_if = 0; 1011 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 1012 if (rt->mfc_expire) 1013 nexpire[hash]--; 1014 rt->mfc_expire = 0; 1015 } 1016 } 1017 if (rt == NULL) { 1018 /* no upcall, so make a new entry */ 1019 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1020 if (rt == NULL) { 1021 splx(s); 1022 return ENOBUFS; 1023 } 1024 1025 /* insert new entry at head of hash chain */ 1026 rt->mfc_origin = mfccp->mfcc_origin; 1027 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1028 rt->mfc_parent = mfccp->mfcc_parent; 1029 for (i = 0; i < numvifs; i++) 1030 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1031 /* initialize pkt counters per src-grp */ 1032 rt->mfc_pkt_cnt = 0; 1033 rt->mfc_byte_cnt = 0; 1034 rt->mfc_wrong_if = 0; 1035 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 1036 rt->mfc_expire = 0; 1037 rt->mfc_stall = NULL; 1038 1039 /* link into table */ 1040 rt->mfc_next = mfctable[hash]; 1041 mfctable[hash] = rt; 1042 } 1043 } 1044 splx(s); 1045 return 0; 1046 } 1047 1048 #ifdef UPCALL_TIMING 1049 /* 1050 * collect delay statistics on the upcalls 1051 */ 1052 static void collate(t) 1053 register struct timeval *t; 1054 { 1055 register u_long d; 1056 register struct timeval tp; 1057 register u_long delta; 1058 1059 GET_TIME(tp); 1060 1061 if (TV_LT(*t, tp)) 1062 { 1063 TV_DELTA(tp, *t, delta); 1064 1065 d = delta >> 10; 1066 if (d > 50) 1067 d = 50; 1068 1069 ++upcall_data[d]; 1070 } 1071 } 1072 #endif /* UPCALL_TIMING */ 1073 1074 /* 1075 * Delete an mfc entry 1076 */ 1077 static int 1078 del_mfc(mfccp) 1079 struct mfcctl *mfccp; 1080 { 1081 struct in_addr origin; 1082 struct in_addr mcastgrp; 1083 struct mfc *rt; 1084 struct mfc **nptr; 1085 u_long hash; 1086 int s; 1087 1088 origin = mfccp->mfcc_origin; 1089 mcastgrp = mfccp->mfcc_mcastgrp; 1090 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1091 1092 if (mrtdebug & DEBUG_MFC) 1093 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1094 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1095 1096 s = splnet(); 1097 1098 nptr = &mfctable[hash]; 1099 while ((rt = *nptr) != NULL) { 1100 if (origin.s_addr == rt->mfc_origin.s_addr && 1101 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1102 rt->mfc_stall == NULL) 1103 break; 1104 1105 nptr = &rt->mfc_next; 1106 } 1107 if (rt == NULL) { 1108 splx(s); 1109 return EADDRNOTAVAIL; 1110 } 1111 1112 *nptr = rt->mfc_next; 1113 free(rt, M_MRTABLE); 1114 1115 splx(s); 1116 1117 return 0; 1118 } 1119 1120 /* 1121 * Send a message to mrouted on the multicast routing socket 1122 */ 1123 static int 1124 socket_send(s, mm, src) 1125 struct socket *s; 1126 struct mbuf *mm; 1127 struct sockaddr_in *src; 1128 { 1129 if (s) { 1130 if (sbappendaddr(&s->so_rcv, 1131 (struct sockaddr *)src, 1132 mm, (struct mbuf *)0) != 0) { 1133 sorwakeup(s); 1134 return 0; 1135 } 1136 } 1137 m_freem(mm); 1138 return -1; 1139 } 1140 1141 /* 1142 * IP multicast forwarding function. This function assumes that the packet 1143 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1144 * pointed to by "ifp", and the packet is to be relayed to other networks 1145 * that have members of the packet's destination IP multicast group. 1146 * 1147 * The packet is returned unscathed to the caller, unless it is 1148 * erroneous, in which case a non-zero return value tells the caller to 1149 * discard it. 1150 */ 1151 1152 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1153 1154 static int 1155 X_ip_mforward(ip, ifp, m, imo) 1156 register struct ip *ip; 1157 struct ifnet *ifp; 1158 struct mbuf *m; 1159 struct ip_moptions *imo; 1160 { 1161 register struct mfc *rt; 1162 register u_char *ipoptions; 1163 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1164 static int srctun = 0; 1165 register struct mbuf *mm; 1166 int s; 1167 vifi_t vifi; 1168 struct vif *vifp; 1169 1170 if (mrtdebug & DEBUG_FORWARD) 1171 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1172 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1173 (void *)ifp); 1174 1175 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1176 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1177 /* 1178 * Packet arrived via a physical interface or 1179 * an encapsulated tunnel. 1180 */ 1181 } else { 1182 /* 1183 * Packet arrived through a source-route tunnel. 1184 * Source-route tunnels are no longer supported. 1185 */ 1186 if ((srctun++ % 1000) == 0) 1187 log(LOG_ERR, 1188 "ip_mforward: received source-routed packet from %lx\n", 1189 (u_long)ntohl(ip->ip_src.s_addr)); 1190 1191 return 1; 1192 } 1193 1194 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1195 if (ip->ip_ttl < 255) 1196 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1197 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1198 vifp = viftable + vifi; 1199 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1200 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr), 1201 vifi, 1202 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1203 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1204 } 1205 return (ip_mdq(m, ifp, NULL, vifi)); 1206 } 1207 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1208 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1209 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr)); 1210 if(!imo) 1211 printf("In fact, no options were specified at all\n"); 1212 } 1213 1214 /* 1215 * Don't forward a packet with time-to-live of zero or one, 1216 * or a packet destined to a local-only group. 1217 */ 1218 if (ip->ip_ttl <= 1 || 1219 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1220 return 0; 1221 1222 /* 1223 * Determine forwarding vifs from the forwarding cache table 1224 */ 1225 s = splnet(); 1226 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt); 1227 1228 /* Entry exists, so forward if necessary */ 1229 if (rt != NULL) { 1230 splx(s); 1231 return (ip_mdq(m, ifp, rt, -1)); 1232 } else { 1233 /* 1234 * If we don't have a route for packet's origin, 1235 * Make a copy of the packet & 1236 * send message to routing daemon 1237 */ 1238 1239 register struct mbuf *mb0; 1240 register struct rtdetq *rte; 1241 register u_long hash; 1242 int hlen = ip->ip_hl << 2; 1243 #ifdef UPCALL_TIMING 1244 struct timeval tp; 1245 1246 GET_TIME(tp); 1247 #endif 1248 1249 mrtstat.mrts_no_route++; 1250 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1251 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1252 (u_long)ntohl(ip->ip_src.s_addr), 1253 (u_long)ntohl(ip->ip_dst.s_addr)); 1254 1255 /* 1256 * Allocate mbufs early so that we don't do extra work if we are 1257 * just going to fail anyway. Make sure to pullup the header so 1258 * that other people can't step on it. 1259 */ 1260 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT); 1261 if (rte == NULL) { 1262 splx(s); 1263 return ENOBUFS; 1264 } 1265 mb0 = m_copy(m, 0, M_COPYALL); 1266 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1267 mb0 = m_pullup(mb0, hlen); 1268 if (mb0 == NULL) { 1269 free(rte, M_MRTABLE); 1270 splx(s); 1271 return ENOBUFS; 1272 } 1273 1274 /* is there an upcall waiting for this packet? */ 1275 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1276 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) { 1277 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1278 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1279 (rt->mfc_stall != NULL)) 1280 break; 1281 } 1282 1283 if (rt == NULL) { 1284 int i; 1285 struct igmpmsg *im; 1286 1287 /* no upcall, so make a new entry */ 1288 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1289 if (rt == NULL) { 1290 free(rte, M_MRTABLE); 1291 m_freem(mb0); 1292 splx(s); 1293 return ENOBUFS; 1294 } 1295 /* Make a copy of the header to send to the user level process */ 1296 mm = m_copy(mb0, 0, hlen); 1297 if (mm == NULL) { 1298 free(rte, M_MRTABLE); 1299 m_freem(mb0); 1300 free(rt, M_MRTABLE); 1301 splx(s); 1302 return ENOBUFS; 1303 } 1304 1305 /* 1306 * Send message to routing daemon to install 1307 * a route into the kernel table 1308 */ 1309 k_igmpsrc.sin_addr = ip->ip_src; 1310 1311 im = mtod(mm, struct igmpmsg *); 1312 im->im_msgtype = IGMPMSG_NOCACHE; 1313 im->im_mbz = 0; 1314 1315 mrtstat.mrts_upcalls++; 1316 1317 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1318 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1319 ++mrtstat.mrts_upq_sockfull; 1320 free(rte, M_MRTABLE); 1321 m_freem(mb0); 1322 free(rt, M_MRTABLE); 1323 splx(s); 1324 return ENOBUFS; 1325 } 1326 1327 /* insert new entry at head of hash chain */ 1328 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1329 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1330 rt->mfc_expire = UPCALL_EXPIRE; 1331 nexpire[hash]++; 1332 for (i = 0; i < numvifs; i++) 1333 rt->mfc_ttls[i] = 0; 1334 rt->mfc_parent = -1; 1335 1336 /* link into table */ 1337 rt->mfc_next = mfctable[hash]; 1338 mfctable[hash] = rt; 1339 rt->mfc_stall = rte; 1340 1341 } else { 1342 /* determine if q has overflowed */ 1343 int npkts = 0; 1344 struct rtdetq **p; 1345 1346 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1347 npkts++; 1348 1349 if (npkts > MAX_UPQ) { 1350 mrtstat.mrts_upq_ovflw++; 1351 free(rte, M_MRTABLE); 1352 m_freem(mb0); 1353 splx(s); 1354 return 0; 1355 } 1356 1357 /* Add this entry to the end of the queue */ 1358 *p = rte; 1359 } 1360 1361 rte->m = mb0; 1362 rte->ifp = ifp; 1363 #ifdef UPCALL_TIMING 1364 rte->t = tp; 1365 #endif 1366 rte->next = NULL; 1367 1368 splx(s); 1369 1370 return 0; 1371 } 1372 } 1373 1374 #ifndef MROUTE_KLD 1375 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 1376 struct ip_moptions *) = X_ip_mforward; 1377 #endif 1378 1379 /* 1380 * Clean up the cache entry if upcall is not serviced 1381 */ 1382 static void 1383 expire_upcalls(void *unused) 1384 { 1385 struct rtdetq *rte; 1386 struct mfc *mfc, **nptr; 1387 int i; 1388 int s; 1389 1390 s = splnet(); 1391 for (i = 0; i < MFCTBLSIZ; i++) { 1392 if (nexpire[i] == 0) 1393 continue; 1394 nptr = &mfctable[i]; 1395 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1396 /* 1397 * Skip real cache entries 1398 * Make sure it wasn't marked to not expire (shouldn't happen) 1399 * If it expires now 1400 */ 1401 if (mfc->mfc_stall != NULL && 1402 mfc->mfc_expire != 0 && 1403 --mfc->mfc_expire == 0) { 1404 if (mrtdebug & DEBUG_EXPIRE) 1405 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1406 (u_long)ntohl(mfc->mfc_origin.s_addr), 1407 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1408 /* 1409 * drop all the packets 1410 * free the mbuf with the pkt, if, timing info 1411 */ 1412 for (rte = mfc->mfc_stall; rte; ) { 1413 struct rtdetq *n = rte->next; 1414 1415 m_freem(rte->m); 1416 free(rte, M_MRTABLE); 1417 rte = n; 1418 } 1419 ++mrtstat.mrts_cache_cleanups; 1420 nexpire[i]--; 1421 1422 *nptr = mfc->mfc_next; 1423 free(mfc, M_MRTABLE); 1424 } else { 1425 nptr = &mfc->mfc_next; 1426 } 1427 } 1428 } 1429 splx(s); 1430 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 1431 } 1432 1433 /* 1434 * Packet forwarding routine once entry in the cache is made 1435 */ 1436 static int 1437 ip_mdq(m, ifp, rt, xmt_vif) 1438 register struct mbuf *m; 1439 register struct ifnet *ifp; 1440 register struct mfc *rt; 1441 register vifi_t xmt_vif; 1442 { 1443 register struct ip *ip = mtod(m, struct ip *); 1444 register vifi_t vifi; 1445 register struct vif *vifp; 1446 register int plen = ip->ip_len; 1447 1448 /* 1449 * Macro to send packet on vif. Since RSVP packets don't get counted on 1450 * input, they shouldn't get counted on output, so statistics keeping is 1451 * separate. 1452 */ 1453 #define MC_SEND(ip,vifp,m) { \ 1454 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1455 encap_send((ip), (vifp), (m)); \ 1456 else \ 1457 phyint_send((ip), (vifp), (m)); \ 1458 } 1459 1460 /* 1461 * If xmt_vif is not -1, send on only the requested vif. 1462 * 1463 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1464 */ 1465 if (xmt_vif < numvifs) { 1466 MC_SEND(ip, viftable + xmt_vif, m); 1467 return 1; 1468 } 1469 1470 /* 1471 * Don't forward if it didn't arrive from the parent vif for its origin. 1472 */ 1473 vifi = rt->mfc_parent; 1474 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1475 /* came in the wrong interface */ 1476 if (mrtdebug & DEBUG_FORWARD) 1477 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1478 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1479 ++mrtstat.mrts_wrong_if; 1480 ++rt->mfc_wrong_if; 1481 /* 1482 * If we are doing PIM assert processing, and we are forwarding 1483 * packets on this interface, and it is a broadcast medium 1484 * interface (and not a tunnel), send a message to the routing daemon. 1485 */ 1486 if (pim_assert && rt->mfc_ttls[vifi] && 1487 (ifp->if_flags & IFF_BROADCAST) && 1488 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1489 struct sockaddr_in k_igmpsrc; 1490 struct mbuf *mm; 1491 struct igmpmsg *im; 1492 int hlen = ip->ip_hl << 2; 1493 struct timeval now; 1494 register u_long delta; 1495 1496 GET_TIME(now); 1497 1498 TV_DELTA(rt->mfc_last_assert, now, delta); 1499 1500 if (delta > ASSERT_MSG_TIME) { 1501 mm = m_copy(m, 0, hlen); 1502 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1503 mm = m_pullup(mm, hlen); 1504 if (mm == NULL) { 1505 return ENOBUFS; 1506 } 1507 1508 rt->mfc_last_assert = now; 1509 1510 im = mtod(mm, struct igmpmsg *); 1511 im->im_msgtype = IGMPMSG_WRONGVIF; 1512 im->im_mbz = 0; 1513 im->im_vif = vifi; 1514 1515 k_igmpsrc.sin_addr = im->im_src; 1516 1517 socket_send(ip_mrouter, mm, &k_igmpsrc); 1518 } 1519 } 1520 return 0; 1521 } 1522 1523 /* If I sourced this packet, it counts as output, else it was input. */ 1524 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1525 viftable[vifi].v_pkt_out++; 1526 viftable[vifi].v_bytes_out += plen; 1527 } else { 1528 viftable[vifi].v_pkt_in++; 1529 viftable[vifi].v_bytes_in += plen; 1530 } 1531 rt->mfc_pkt_cnt++; 1532 rt->mfc_byte_cnt += plen; 1533 1534 /* 1535 * For each vif, decide if a copy of the packet should be forwarded. 1536 * Forward if: 1537 * - the ttl exceeds the vif's threshold 1538 * - there are group members downstream on interface 1539 */ 1540 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1541 if ((rt->mfc_ttls[vifi] > 0) && 1542 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1543 vifp->v_pkt_out++; 1544 vifp->v_bytes_out += plen; 1545 MC_SEND(ip, vifp, m); 1546 } 1547 1548 return 0; 1549 } 1550 1551 /* 1552 * check if a vif number is legal/ok. This is used by ip_output, to export 1553 * numvifs there, 1554 */ 1555 static int 1556 X_legal_vif_num(vif) 1557 int vif; 1558 { 1559 if (vif >= 0 && vif < numvifs) 1560 return(1); 1561 else 1562 return(0); 1563 } 1564 1565 #ifndef MROUTE_KLD 1566 int (*legal_vif_num)(int) = X_legal_vif_num; 1567 #endif 1568 1569 /* 1570 * Return the local address used by this vif 1571 */ 1572 static u_long 1573 X_ip_mcast_src(vifi) 1574 int vifi; 1575 { 1576 if (vifi >= 0 && vifi < numvifs) 1577 return viftable[vifi].v_lcl_addr.s_addr; 1578 else 1579 return INADDR_ANY; 1580 } 1581 1582 #ifndef MROUTE_KLD 1583 u_long (*ip_mcast_src)(int) = X_ip_mcast_src; 1584 #endif 1585 1586 static void 1587 phyint_send(ip, vifp, m) 1588 struct ip *ip; 1589 struct vif *vifp; 1590 struct mbuf *m; 1591 { 1592 register struct mbuf *mb_copy; 1593 register int hlen = ip->ip_hl << 2; 1594 1595 /* 1596 * Make a new reference to the packet; make sure that 1597 * the IP header is actually copied, not just referenced, 1598 * so that ip_output() only scribbles on the copy. 1599 */ 1600 mb_copy = m_copy(m, 0, M_COPYALL); 1601 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1602 mb_copy = m_pullup(mb_copy, hlen); 1603 if (mb_copy == NULL) 1604 return; 1605 1606 if (vifp->v_rate_limit == 0) 1607 tbf_send_packet(vifp, mb_copy); 1608 else 1609 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1610 } 1611 1612 static void 1613 encap_send(ip, vifp, m) 1614 register struct ip *ip; 1615 register struct vif *vifp; 1616 register struct mbuf *m; 1617 { 1618 register struct mbuf *mb_copy; 1619 register struct ip *ip_copy; 1620 register int i, len = ip->ip_len; 1621 1622 /* 1623 * copy the old packet & pullup its IP header into the 1624 * new mbuf so we can modify it. Try to fill the new 1625 * mbuf since if we don't the ethernet driver will. 1626 */ 1627 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER); 1628 if (mb_copy == NULL) 1629 return; 1630 mb_copy->m_data += max_linkhdr; 1631 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1632 1633 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1634 m_freem(mb_copy); 1635 return; 1636 } 1637 i = MHLEN - M_LEADINGSPACE(mb_copy); 1638 if (i > len) 1639 i = len; 1640 mb_copy = m_pullup(mb_copy, i); 1641 if (mb_copy == NULL) 1642 return; 1643 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1644 1645 /* 1646 * fill in the encapsulating IP header. 1647 */ 1648 ip_copy = mtod(mb_copy, struct ip *); 1649 *ip_copy = multicast_encap_iphdr; 1650 #ifdef RANDOM_IP_ID 1651 ip_copy->ip_id = ip_randomid(); 1652 #else 1653 ip_copy->ip_id = htons(ip_id++); 1654 #endif 1655 ip_copy->ip_len += len; 1656 ip_copy->ip_src = vifp->v_lcl_addr; 1657 ip_copy->ip_dst = vifp->v_rmt_addr; 1658 1659 /* 1660 * turn the encapsulated IP header back into a valid one. 1661 */ 1662 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1663 --ip->ip_ttl; 1664 HTONS(ip->ip_len); 1665 HTONS(ip->ip_off); 1666 ip->ip_sum = 0; 1667 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1668 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1669 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1670 1671 if (vifp->v_rate_limit == 0) 1672 tbf_send_packet(vifp, mb_copy); 1673 else 1674 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1675 } 1676 1677 /* 1678 * Token bucket filter module 1679 */ 1680 1681 static void 1682 tbf_control(vifp, m, ip, p_len) 1683 register struct vif *vifp; 1684 register struct mbuf *m; 1685 register struct ip *ip; 1686 register u_long p_len; 1687 { 1688 register struct tbf *t = vifp->v_tbf; 1689 1690 if (p_len > MAX_BKT_SIZE) { 1691 /* drop if packet is too large */ 1692 mrtstat.mrts_pkt2large++; 1693 m_freem(m); 1694 return; 1695 } 1696 1697 tbf_update_tokens(vifp); 1698 1699 /* if there are enough tokens, 1700 * and the queue is empty, 1701 * send this packet out 1702 */ 1703 1704 if (t->tbf_q_len == 0) { 1705 /* queue empty, send packet if enough tokens */ 1706 if (p_len <= t->tbf_n_tok) { 1707 t->tbf_n_tok -= p_len; 1708 tbf_send_packet(vifp, m); 1709 } else { 1710 /* queue packet and timeout till later */ 1711 tbf_queue(vifp, m); 1712 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1713 } 1714 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1715 /* finite queue length, so queue pkts and process queue */ 1716 tbf_queue(vifp, m); 1717 tbf_process_q(vifp); 1718 } else { 1719 /* queue length too much, try to dq and queue and process */ 1720 if (!tbf_dq_sel(vifp, ip)) { 1721 mrtstat.mrts_q_overflow++; 1722 m_freem(m); 1723 return; 1724 } else { 1725 tbf_queue(vifp, m); 1726 tbf_process_q(vifp); 1727 } 1728 } 1729 return; 1730 } 1731 1732 /* 1733 * adds a packet to the queue at the interface 1734 */ 1735 static void 1736 tbf_queue(vifp, m) 1737 register struct vif *vifp; 1738 register struct mbuf *m; 1739 { 1740 register int s = splnet(); 1741 register struct tbf *t = vifp->v_tbf; 1742 1743 if (t->tbf_t == NULL) { 1744 /* Queue was empty */ 1745 t->tbf_q = m; 1746 } else { 1747 /* Insert at tail */ 1748 t->tbf_t->m_act = m; 1749 } 1750 1751 /* Set new tail pointer */ 1752 t->tbf_t = m; 1753 1754 #ifdef DIAGNOSTIC 1755 /* Make sure we didn't get fed a bogus mbuf */ 1756 if (m->m_act) 1757 panic("tbf_queue: m_act"); 1758 #endif 1759 m->m_act = NULL; 1760 1761 t->tbf_q_len++; 1762 1763 splx(s); 1764 } 1765 1766 1767 /* 1768 * processes the queue at the interface 1769 */ 1770 static void 1771 tbf_process_q(vifp) 1772 register struct vif *vifp; 1773 { 1774 register struct mbuf *m; 1775 register int len; 1776 register int s = splnet(); 1777 register struct tbf *t = vifp->v_tbf; 1778 1779 /* loop through the queue at the interface and send as many packets 1780 * as possible 1781 */ 1782 while (t->tbf_q_len > 0) { 1783 m = t->tbf_q; 1784 1785 len = mtod(m, struct ip *)->ip_len; 1786 1787 /* determine if the packet can be sent */ 1788 if (len <= t->tbf_n_tok) { 1789 /* if so, 1790 * reduce no of tokens, dequeue the packet, 1791 * send the packet. 1792 */ 1793 t->tbf_n_tok -= len; 1794 1795 t->tbf_q = m->m_act; 1796 if (--t->tbf_q_len == 0) 1797 t->tbf_t = NULL; 1798 1799 m->m_act = NULL; 1800 tbf_send_packet(vifp, m); 1801 1802 } else break; 1803 } 1804 splx(s); 1805 } 1806 1807 static void 1808 tbf_reprocess_q(xvifp) 1809 void *xvifp; 1810 { 1811 register struct vif *vifp = xvifp; 1812 if (ip_mrouter == NULL) 1813 return; 1814 1815 tbf_update_tokens(vifp); 1816 1817 tbf_process_q(vifp); 1818 1819 if (vifp->v_tbf->tbf_q_len) 1820 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1821 } 1822 1823 /* function that will selectively discard a member of the queue 1824 * based on the precedence value and the priority 1825 */ 1826 static int 1827 tbf_dq_sel(vifp, ip) 1828 register struct vif *vifp; 1829 register struct ip *ip; 1830 { 1831 register int s = splnet(); 1832 register u_int p; 1833 register struct mbuf *m, *last; 1834 register struct mbuf **np; 1835 register struct tbf *t = vifp->v_tbf; 1836 1837 p = priority(vifp, ip); 1838 1839 np = &t->tbf_q; 1840 last = NULL; 1841 while ((m = *np) != NULL) { 1842 if (p > priority(vifp, mtod(m, struct ip *))) { 1843 *np = m->m_act; 1844 /* If we're removing the last packet, fix the tail pointer */ 1845 if (m == t->tbf_t) 1846 t->tbf_t = last; 1847 m_freem(m); 1848 /* it's impossible for the queue to be empty, but 1849 * we check anyway. */ 1850 if (--t->tbf_q_len == 0) 1851 t->tbf_t = NULL; 1852 splx(s); 1853 mrtstat.mrts_drop_sel++; 1854 return(1); 1855 } 1856 np = &m->m_act; 1857 last = m; 1858 } 1859 splx(s); 1860 return(0); 1861 } 1862 1863 static void 1864 tbf_send_packet(vifp, m) 1865 register struct vif *vifp; 1866 register struct mbuf *m; 1867 { 1868 struct ip_moptions imo; 1869 int error; 1870 int s = splnet(); 1871 1872 if (vifp->v_flags & VIFF_TUNNEL) { 1873 /* If tunnel options */ 1874 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1875 IP_FORWARDING, (struct ip_moptions *)0); 1876 } else { 1877 imo.imo_multicast_ifp = vifp->v_ifp; 1878 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1879 imo.imo_multicast_loop = 1; 1880 imo.imo_multicast_vif = -1; 1881 1882 /* 1883 * Re-entrancy should not be a problem here, because 1884 * the packets that we send out and are looped back at us 1885 * should get rejected because they appear to come from 1886 * the loopback interface, thus preventing looping. 1887 */ 1888 error = ip_output(m, (struct mbuf *)0, NULL, 1889 IP_FORWARDING, &imo); 1890 1891 if (mrtdebug & DEBUG_XMIT) 1892 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1893 vifp - viftable, error); 1894 } 1895 splx(s); 1896 } 1897 1898 /* determine the current time and then 1899 * the elapsed time (between the last time and time now) 1900 * in milliseconds & update the no. of tokens in the bucket 1901 */ 1902 static void 1903 tbf_update_tokens(vifp) 1904 register struct vif *vifp; 1905 { 1906 struct timeval tp; 1907 register u_long tm; 1908 register int s = splnet(); 1909 register struct tbf *t = vifp->v_tbf; 1910 1911 GET_TIME(tp); 1912 1913 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1914 1915 /* 1916 * This formula is actually 1917 * "time in seconds" * "bytes/second". 1918 * 1919 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1920 * 1921 * The (1000/1024) was introduced in add_vif to optimize 1922 * this divide into a shift. 1923 */ 1924 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1925 t->tbf_last_pkt_t = tp; 1926 1927 if (t->tbf_n_tok > MAX_BKT_SIZE) 1928 t->tbf_n_tok = MAX_BKT_SIZE; 1929 1930 splx(s); 1931 } 1932 1933 static int 1934 priority(vifp, ip) 1935 register struct vif *vifp; 1936 register struct ip *ip; 1937 { 1938 register int prio; 1939 1940 /* temporary hack; may add general packet classifier some day */ 1941 1942 /* 1943 * The UDP port space is divided up into four priority ranges: 1944 * [0, 16384) : unclassified - lowest priority 1945 * [16384, 32768) : audio - highest priority 1946 * [32768, 49152) : whiteboard - medium priority 1947 * [49152, 65536) : video - low priority 1948 */ 1949 if (ip->ip_p == IPPROTO_UDP) { 1950 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1951 switch (ntohs(udp->uh_dport) & 0xc000) { 1952 case 0x4000: 1953 prio = 70; 1954 break; 1955 case 0x8000: 1956 prio = 60; 1957 break; 1958 case 0xc000: 1959 prio = 55; 1960 break; 1961 default: 1962 prio = 50; 1963 break; 1964 } 1965 if (tbfdebug > 1) 1966 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio); 1967 } else { 1968 prio = 50; 1969 } 1970 return prio; 1971 } 1972 1973 /* 1974 * End of token bucket filter modifications 1975 */ 1976 1977 int 1978 ip_rsvp_vif_init(so, sopt) 1979 struct socket *so; 1980 struct sockopt *sopt; 1981 { 1982 int error, i, s; 1983 1984 if (rsvpdebug) 1985 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 1986 so->so_type, so->so_proto->pr_protocol); 1987 1988 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1989 return EOPNOTSUPP; 1990 1991 /* Check mbuf. */ 1992 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1993 if (error) 1994 return (error); 1995 1996 if (rsvpdebug) 1997 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", i, rsvp_on); 1998 1999 s = splnet(); 2000 2001 /* Check vif. */ 2002 if (!legal_vif_num(i)) { 2003 splx(s); 2004 return EADDRNOTAVAIL; 2005 } 2006 2007 /* Check if socket is available. */ 2008 if (viftable[i].v_rsvpd != NULL) { 2009 splx(s); 2010 return EADDRINUSE; 2011 } 2012 2013 viftable[i].v_rsvpd = so; 2014 /* This may seem silly, but we need to be sure we don't over-increment 2015 * the RSVP counter, in case something slips up. 2016 */ 2017 if (!viftable[i].v_rsvp_on) { 2018 viftable[i].v_rsvp_on = 1; 2019 rsvp_on++; 2020 } 2021 2022 splx(s); 2023 return 0; 2024 } 2025 2026 int 2027 ip_rsvp_vif_done(so, sopt) 2028 struct socket *so; 2029 struct sockopt *sopt; 2030 { 2031 int error, i, s; 2032 2033 if (rsvpdebug) 2034 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2035 so->so_type, so->so_proto->pr_protocol); 2036 2037 if (so->so_type != SOCK_RAW || 2038 so->so_proto->pr_protocol != IPPROTO_RSVP) 2039 return EOPNOTSUPP; 2040 2041 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 2042 if (error) 2043 return (error); 2044 2045 s = splnet(); 2046 2047 /* Check vif. */ 2048 if (!legal_vif_num(i)) { 2049 splx(s); 2050 return EADDRNOTAVAIL; 2051 } 2052 2053 if (rsvpdebug) 2054 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n", 2055 viftable[i].v_rsvpd, so); 2056 2057 viftable[i].v_rsvpd = NULL; 2058 /* 2059 * This may seem silly, but we need to be sure we don't over-decrement 2060 * the RSVP counter, in case something slips up. 2061 */ 2062 if (viftable[i].v_rsvp_on) { 2063 viftable[i].v_rsvp_on = 0; 2064 rsvp_on--; 2065 } 2066 2067 splx(s); 2068 return 0; 2069 } 2070 2071 void 2072 ip_rsvp_force_done(so) 2073 struct socket *so; 2074 { 2075 int vifi; 2076 register int s; 2077 2078 /* Don't bother if it is not the right type of socket. */ 2079 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2080 return; 2081 2082 s = splnet(); 2083 2084 /* The socket may be attached to more than one vif...this 2085 * is perfectly legal. 2086 */ 2087 for (vifi = 0; vifi < numvifs; vifi++) { 2088 if (viftable[vifi].v_rsvpd == so) { 2089 viftable[vifi].v_rsvpd = NULL; 2090 /* This may seem silly, but we need to be sure we don't 2091 * over-decrement the RSVP counter, in case something slips up. 2092 */ 2093 if (viftable[vifi].v_rsvp_on) { 2094 viftable[vifi].v_rsvp_on = 0; 2095 rsvp_on--; 2096 } 2097 } 2098 } 2099 2100 splx(s); 2101 return; 2102 } 2103 2104 void 2105 rsvp_input(m, off) 2106 struct mbuf *m; 2107 int off; 2108 { 2109 int vifi; 2110 register struct ip *ip = mtod(m, struct ip *); 2111 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2112 register int s; 2113 struct ifnet *ifp; 2114 2115 if (rsvpdebug) 2116 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2117 2118 /* Can still get packets with rsvp_on = 0 if there is a local member 2119 * of the group to which the RSVP packet is addressed. But in this 2120 * case we want to throw the packet away. 2121 */ 2122 if (!rsvp_on) { 2123 m_freem(m); 2124 return; 2125 } 2126 2127 s = splnet(); 2128 2129 if (rsvpdebug) 2130 printf("rsvp_input: check vifs\n"); 2131 2132 #ifdef DIAGNOSTIC 2133 if (!(m->m_flags & M_PKTHDR)) 2134 panic("rsvp_input no hdr"); 2135 #endif 2136 2137 ifp = m->m_pkthdr.rcvif; 2138 /* Find which vif the packet arrived on. */ 2139 for (vifi = 0; vifi < numvifs; vifi++) 2140 if (viftable[vifi].v_ifp == ifp) 2141 break; 2142 2143 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 2144 /* 2145 * If the old-style non-vif-associated socket is set, 2146 * then use it. Otherwise, drop packet since there 2147 * is no specific socket for this vif. 2148 */ 2149 if (ip_rsvpd != NULL) { 2150 if (rsvpdebug) 2151 printf("rsvp_input: Sending packet up old-style socket\n"); 2152 rip_input(m, off); /* xxx */ 2153 } else { 2154 if (rsvpdebug && vifi == numvifs) 2155 printf("rsvp_input: Can't find vif for packet.\n"); 2156 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 2157 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2158 m_freem(m); 2159 } 2160 splx(s); 2161 return; 2162 } 2163 rsvp_src.sin_addr = ip->ip_src; 2164 2165 if (rsvpdebug && m) 2166 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2167 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2168 2169 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 2170 if (rsvpdebug) 2171 printf("rsvp_input: Failed to append to socket\n"); 2172 } else { 2173 if (rsvpdebug) 2174 printf("rsvp_input: send packet up\n"); 2175 } 2176 2177 splx(s); 2178 } 2179 2180 #ifdef MROUTE_KLD 2181 2182 static int 2183 ip_mroute_modevent(module_t mod, int type, void *unused) 2184 { 2185 int s; 2186 2187 switch (type) { 2188 static u_long (*old_ip_mcast_src)(int); 2189 static int (*old_ip_mrouter_set)(struct socket *, 2190 struct sockopt *); 2191 static int (*old_ip_mrouter_get)(struct socket *, 2192 struct sockopt *); 2193 static int (*old_ip_mrouter_done)(void); 2194 static int (*old_ip_mforward)(struct ip *, struct ifnet *, 2195 struct mbuf *, struct ip_moptions *); 2196 static int (*old_mrt_ioctl)(int, caddr_t); 2197 static int (*old_legal_vif_num)(int); 2198 2199 case MOD_LOAD: 2200 s = splnet(); 2201 /* XXX Protect against multiple loading */ 2202 old_ip_mcast_src = ip_mcast_src; 2203 ip_mcast_src = X_ip_mcast_src; 2204 old_ip_mrouter_get = ip_mrouter_get; 2205 ip_mrouter_get = X_ip_mrouter_get; 2206 old_ip_mrouter_set = ip_mrouter_set; 2207 ip_mrouter_set = X_ip_mrouter_set; 2208 old_ip_mrouter_done = ip_mrouter_done; 2209 ip_mrouter_done = X_ip_mrouter_done; 2210 old_ip_mforward = ip_mforward; 2211 ip_mforward = X_ip_mforward; 2212 old_mrt_ioctl = mrt_ioctl; 2213 mrt_ioctl = X_mrt_ioctl; 2214 old_legal_vif_num = legal_vif_num; 2215 legal_vif_num = X_legal_vif_num; 2216 2217 splx(s); 2218 return 0; 2219 2220 case MOD_UNLOAD: 2221 if (ip_mrouter) 2222 return EINVAL; 2223 2224 s = splnet(); 2225 ip_mrouter_get = old_ip_mrouter_get; 2226 ip_mrouter_set = old_ip_mrouter_set; 2227 ip_mrouter_done = old_ip_mrouter_done; 2228 ip_mforward = old_ip_mforward; 2229 mrt_ioctl = old_mrt_ioctl; 2230 legal_vif_num = old_legal_vif_num; 2231 splx(s); 2232 return 0; 2233 2234 default: 2235 break; 2236 } 2237 return 0; 2238 } 2239 2240 static moduledata_t ip_mroutemod = { 2241 "ip_mroute", 2242 ip_mroute_modevent, 2243 0 2244 }; 2245 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 2246 2247 #endif /* MROUTE_KLD */ 2248 #endif /* MROUTING */ 2249