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 1.26 1995/12/02 19:38:03 bde Exp $ 13 */ 14 15 16 #include <sys/param.h> 17 #include <sys/systm.h> 18 #include <sys/mbuf.h> 19 #include <sys/socket.h> 20 #include <sys/socketvar.h> 21 #include <sys/protosw.h> 22 #include <sys/errno.h> 23 #include <sys/time.h> 24 #include <sys/kernel.h> 25 #include <sys/ioctl.h> 26 #include <sys/syslog.h> 27 #include <sys/queue.h> 28 #include <net/if.h> 29 #include <net/route.h> 30 #include <netinet/in.h> 31 #include <netinet/in_systm.h> 32 #include <netinet/ip.h> 33 #include <netinet/ip_var.h> 34 #include <netinet/in_pcb.h> 35 #include <netinet/in_var.h> 36 #include <netinet/igmp.h> 37 #include <netinet/igmp_var.h> 38 #include <netinet/ip_mroute.h> 39 #include <netinet/udp.h> 40 41 extern void rsvp_input __P((struct mbuf *m, int iphlen)); 42 43 #ifndef NTOHL 44 #if BYTE_ORDER != BIG_ENDIAN 45 #define NTOHL(d) ((d) = ntohl((d))) 46 #define NTOHS(d) ((d) = ntohs((u_short)(d))) 47 #define HTONL(d) ((d) = htonl((d))) 48 #define HTONS(d) ((d) = htons((u_short)(d))) 49 #else 50 #define NTOHL(d) 51 #define NTOHS(d) 52 #define HTONL(d) 53 #define HTONS(d) 54 #endif 55 #endif 56 57 #ifndef MROUTING 58 extern void ipip_input __P((struct mbuf *m)); 59 extern u_long _ip_mcast_src __P((int vifi)); 60 extern int _ip_mforward __P((struct ip *ip, struct ifnet *ifp, 61 struct mbuf *m, struct ip_moptions *imo)); 62 extern int _ip_mrouter_done __P((void)); 63 extern int _ip_mrouter_get __P((int cmd, struct socket *so, 64 struct mbuf **m)); 65 extern int _ip_mrouter_set __P((int cmd, struct socket *so, 66 struct mbuf *m)); 67 extern int _mrt_ioctl __P((int req, caddr_t data, struct proc *p)); 68 69 /* 70 * Dummy routines and globals used when multicast routing is not compiled in. 71 */ 72 73 struct socket *ip_mrouter = NULL; 74 static u_int ip_mrtproto = 0; 75 static struct mrtstat mrtstat; 76 u_int rsvpdebug = 0; 77 78 int 79 _ip_mrouter_set(cmd, so, m) 80 int cmd; 81 struct socket *so; 82 struct mbuf *m; 83 { 84 return(EOPNOTSUPP); 85 } 86 87 int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = _ip_mrouter_set; 88 89 90 int 91 _ip_mrouter_get(cmd, so, m) 92 int cmd; 93 struct socket *so; 94 struct mbuf **m; 95 { 96 return(EOPNOTSUPP); 97 } 98 99 int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = _ip_mrouter_get; 100 101 int 102 _ip_mrouter_done() 103 { 104 return(0); 105 } 106 107 int (*ip_mrouter_done)(void) = _ip_mrouter_done; 108 109 int 110 _ip_mforward(ip, ifp, m, imo) 111 struct ip *ip; 112 struct ifnet *ifp; 113 struct mbuf *m; 114 struct ip_moptions *imo; 115 { 116 return(0); 117 } 118 119 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 120 struct ip_moptions *) = _ip_mforward; 121 122 int 123 _mrt_ioctl(int req, caddr_t data, struct proc *p) 124 { 125 return EOPNOTSUPP; 126 } 127 128 int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl; 129 130 void 131 rsvp_input(m, iphlen) /* XXX must fixup manually */ 132 struct mbuf *m; 133 int iphlen; 134 { 135 /* Can still get packets with rsvp_on = 0 if there is a local member 136 * of the group to which the RSVP packet is addressed. But in this 137 * case we want to throw the packet away. 138 */ 139 if (!rsvp_on) { 140 m_freem(m); 141 return; 142 } 143 144 if (ip_rsvpd != NULL) { 145 if (rsvpdebug) 146 printf("rsvp_input: Sending packet up old-style socket\n"); 147 rip_input(m); 148 return; 149 } 150 /* Drop the packet */ 151 m_freem(m); 152 } 153 154 void ipip_input(struct mbuf *m) { /* XXX must fixup manually */ 155 rip_input(m); 156 } 157 158 int (*legal_vif_num)(int) = 0; 159 160 /* 161 * This should never be called, since IP_MULTICAST_VIF should fail, but 162 * just in case it does get called, the code a little lower in ip_output 163 * will assign the packet a local address. 164 */ 165 u_long 166 _ip_mcast_src(int vifi) { return INADDR_ANY; } 167 u_long (*ip_mcast_src)(int) = _ip_mcast_src; 168 169 int 170 ip_rsvp_vif_init(so, m) 171 struct socket *so; 172 struct mbuf *m; 173 { 174 return(EINVAL); 175 } 176 177 int 178 ip_rsvp_vif_done(so, m) 179 struct socket *so; 180 struct mbuf *m; 181 { 182 return(EINVAL); 183 } 184 185 void 186 ip_rsvp_force_done(so) 187 struct socket *so; 188 { 189 return; 190 } 191 192 #else /* MROUTING */ 193 194 #define M_HASCL(m) ((m)->m_flags & M_EXT) 195 196 #define INSIZ sizeof(struct in_addr) 197 #define same(a1, a2) \ 198 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0) 199 200 #define MT_MRTABLE MT_RTABLE /* since nothing else uses it */ 201 202 /* 203 * Globals. All but ip_mrouter and ip_mrtproto could be static, 204 * except for netstat or debugging purposes. 205 */ 206 #ifndef MROUTE_LKM 207 extern void ipip_input __P((struct mbuf *m, int iphlen)); 208 struct socket *ip_mrouter = NULL; 209 struct mrtstat mrtstat; 210 211 int ip_mrtproto = IGMP_DVMRP; /* for netstat only */ 212 #else /* MROUTE_LKM */ 213 #error /* the function definition will have a syntax error */ 214 extern void X_ipip_input __P((struct mbuf *m)); 215 extern struct mrtstat mrtstat; 216 static int ip_mrtproto; 217 #endif 218 219 #define NO_RTE_FOUND 0x1 220 #define RTE_FOUND 0x2 221 222 static struct mbuf *mfctable[MFCTBLSIZ]; 223 static u_char nexpire[MFCTBLSIZ]; 224 static struct vif viftable[MAXVIFS]; 225 static u_int mrtdebug = 0; /* debug level */ 226 #define DEBUG_MFC 0x02 227 #define DEBUG_FORWARD 0x04 228 #define DEBUG_EXPIRE 0x08 229 #define DEBUG_XMIT 0x10 230 static u_int tbfdebug = 0; /* tbf debug level */ 231 static u_int rsvpdebug = 0; /* rsvp debug level */ 232 233 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 234 #define UPCALL_EXPIRE 6 /* number of timeouts */ 235 236 /* 237 * Define the token bucket filter structures 238 * tbftable -> each vif has one of these for storing info 239 */ 240 241 static struct tbf tbftable[MAXVIFS]; 242 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 243 244 /* 245 * 'Interfaces' associated with decapsulator (so we can tell 246 * packets that went through it from ones that get reflected 247 * by a broken gateway). These interfaces are never linked into 248 * the system ifnet list & no routes point to them. I.e., packets 249 * can't be sent this way. They only exist as a placeholder for 250 * multicast source verification. 251 */ 252 static struct ifnet multicast_decap_if[MAXVIFS]; 253 254 #define ENCAP_TTL 64 255 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 256 257 /* prototype IP hdr for encapsulated packets */ 258 static struct ip multicast_encap_iphdr = { 259 #if BYTE_ORDER == LITTLE_ENDIAN 260 sizeof(struct ip) >> 2, IPVERSION, 261 #else 262 IPVERSION, sizeof(struct ip) >> 2, 263 #endif 264 0, /* tos */ 265 sizeof(struct ip), /* total length */ 266 0, /* id */ 267 0, /* frag offset */ 268 ENCAP_TTL, ENCAP_PROTO, 269 0, /* checksum */ 270 }; 271 272 /* 273 * Private variables. 274 */ 275 static vifi_t numvifs = 0; 276 static int have_encap_tunnel = 0; 277 278 /* 279 * one-back cache used by ipip_input to locate a tunnel's vif 280 * given a datagram's src ip address. 281 */ 282 static u_long last_encap_src; 283 static struct vif *last_encap_vif; 284 285 static u_long X_ip_mcast_src __P((int vifi)); 286 static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo)); 287 static int X_ip_mrouter_done __P((void)); 288 static int X_ip_mrouter_get __P((int cmd, struct socket *so, struct mbuf **m)); 289 static int X_ip_mrouter_set __P((int cmd, struct socket *so, struct mbuf *m)); 290 static int X_legal_vif_num __P((int vif)); 291 static int X_mrt_ioctl __P((int cmd, caddr_t data)); 292 293 static int get_sg_cnt(struct sioc_sg_req *); 294 static int get_vif_cnt(struct sioc_vif_req *); 295 static int ip_mrouter_init(struct socket *, struct mbuf *); 296 static int add_vif(struct vifctl *); 297 static int del_vif(vifi_t *); 298 static int add_mfc(struct mfcctl *); 299 static int del_mfc(struct mfcctl *); 300 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 301 static int get_version(struct mbuf *); 302 static int get_assert(struct mbuf *); 303 static int set_assert(int *); 304 static void expire_upcalls(void *); 305 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, 306 vifi_t); 307 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 308 static void encap_send(struct ip *, struct vif *, struct mbuf *); 309 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 310 static void tbf_queue(struct vif *, struct mbuf *); 311 static void tbf_process_q(struct vif *); 312 static void tbf_reprocess_q(void *); 313 static int tbf_dq_sel(struct vif *, struct ip *); 314 static void tbf_send_packet(struct vif *, struct mbuf *); 315 static void tbf_update_tokens(struct vif *); 316 static int priority(struct vif *, struct ip *); 317 void multiencap_decap(struct mbuf *); 318 319 /* 320 * whether or not special PIM assert processing is enabled. 321 */ 322 static int pim_assert; 323 /* 324 * Rate limit for assert notification messages, in usec 325 */ 326 #define ASSERT_MSG_TIME 3000000 327 328 /* 329 * Hash function for a source, group entry 330 */ 331 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 332 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 333 334 /* 335 * Find a route for a given origin IP address and Multicast group address 336 * Type of service parameter to be added in the future!!! 337 */ 338 339 #define MFCFIND(o, g, rt) { \ 340 register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \ 341 register struct mfc *_rt = NULL; \ 342 rt = NULL; \ 343 ++mrtstat.mrts_mfc_lookups; \ 344 while (_mb_rt) { \ 345 _rt = mtod(_mb_rt, struct mfc *); \ 346 if ((_rt->mfc_origin.s_addr == o) && \ 347 (_rt->mfc_mcastgrp.s_addr == g) && \ 348 (_mb_rt->m_act == NULL)) { \ 349 rt = _rt; \ 350 break; \ 351 } \ 352 _mb_rt = _mb_rt->m_next; \ 353 } \ 354 if (rt == NULL) { \ 355 ++mrtstat.mrts_mfc_misses; \ 356 } \ 357 } 358 359 360 /* 361 * Macros to compute elapsed time efficiently 362 * Borrowed from Van Jacobson's scheduling code 363 */ 364 #define TV_DELTA(a, b, delta) { \ 365 register int xxs; \ 366 \ 367 delta = (a).tv_usec - (b).tv_usec; \ 368 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 369 switch (xxs) { \ 370 case 2: \ 371 delta += 1000000; \ 372 /* fall through */ \ 373 case 1: \ 374 delta += 1000000; \ 375 break; \ 376 default: \ 377 delta += (1000000 * xxs); \ 378 } \ 379 } \ 380 } 381 382 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 383 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 384 385 #ifdef UPCALL_TIMING 386 u_long upcall_data[51]; 387 static void collate(struct timeval *); 388 #endif /* UPCALL_TIMING */ 389 390 391 /* 392 * Handle MRT setsockopt commands to modify the multicast routing tables. 393 */ 394 static int 395 X_ip_mrouter_set(cmd, so, m) 396 int cmd; 397 struct socket *so; 398 struct mbuf *m; 399 { 400 if (cmd != MRT_INIT && so != ip_mrouter) return EACCES; 401 402 switch (cmd) { 403 case MRT_INIT: return ip_mrouter_init(so, m); 404 case MRT_DONE: return ip_mrouter_done(); 405 case MRT_ADD_VIF: return add_vif (mtod(m, struct vifctl *)); 406 case MRT_DEL_VIF: return del_vif (mtod(m, vifi_t *)); 407 case MRT_ADD_MFC: return add_mfc (mtod(m, struct mfcctl *)); 408 case MRT_DEL_MFC: return del_mfc (mtod(m, struct mfcctl *)); 409 case MRT_ASSERT: return set_assert(mtod(m, int *)); 410 default: return EOPNOTSUPP; 411 } 412 } 413 414 #ifndef MROUTE_LKM 415 int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = X_ip_mrouter_set; 416 #endif 417 418 /* 419 * Handle MRT getsockopt commands 420 */ 421 static int 422 X_ip_mrouter_get(cmd, so, m) 423 int cmd; 424 struct socket *so; 425 struct mbuf **m; 426 { 427 struct mbuf *mb; 428 429 if (so != ip_mrouter) return EACCES; 430 431 *m = mb = m_get(M_WAIT, MT_SOOPTS); 432 433 switch (cmd) { 434 case MRT_VERSION: return get_version(mb); 435 case MRT_ASSERT: return get_assert(mb); 436 default: return EOPNOTSUPP; 437 } 438 } 439 440 #ifndef MROUTE_LKM 441 int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = X_ip_mrouter_get; 442 #endif 443 444 /* 445 * Handle ioctl commands to obtain information from the cache 446 */ 447 static int 448 X_mrt_ioctl(cmd, data) 449 int cmd; 450 caddr_t data; 451 { 452 int error = 0; 453 454 switch (cmd) { 455 case (SIOCGETVIFCNT): 456 return (get_vif_cnt((struct sioc_vif_req *)data)); 457 break; 458 case (SIOCGETSGCNT): 459 return (get_sg_cnt((struct sioc_sg_req *)data)); 460 break; 461 default: 462 return (EINVAL); 463 break; 464 } 465 return error; 466 } 467 468 #ifndef MROUTE_LKM 469 int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl; 470 #endif 471 472 /* 473 * returns the packet, byte, rpf-failure count for the source group provided 474 */ 475 static int 476 get_sg_cnt(req) 477 register struct sioc_sg_req *req; 478 { 479 register struct mfc *rt; 480 int s; 481 482 s = splnet(); 483 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 484 splx(s); 485 if (rt != NULL) { 486 req->pktcnt = rt->mfc_pkt_cnt; 487 req->bytecnt = rt->mfc_byte_cnt; 488 req->wrong_if = rt->mfc_wrong_if; 489 } else 490 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 491 492 return 0; 493 } 494 495 /* 496 * returns the input and output packet and byte counts on the vif provided 497 */ 498 static int 499 get_vif_cnt(req) 500 register struct sioc_vif_req *req; 501 { 502 register vifi_t vifi = req->vifi; 503 504 if (vifi >= numvifs) return EINVAL; 505 506 req->icount = viftable[vifi].v_pkt_in; 507 req->ocount = viftable[vifi].v_pkt_out; 508 req->ibytes = viftable[vifi].v_bytes_in; 509 req->obytes = viftable[vifi].v_bytes_out; 510 511 return 0; 512 } 513 514 /* 515 * Enable multicast routing 516 */ 517 static int 518 ip_mrouter_init(so, m) 519 struct socket *so; 520 struct mbuf *m; 521 { 522 int *v; 523 524 if (mrtdebug) 525 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 526 so->so_type, so->so_proto->pr_protocol); 527 528 if (so->so_type != SOCK_RAW || 529 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP; 530 531 if (!m || (m->m_len != sizeof(int *))) 532 return ENOPROTOOPT; 533 534 v = mtod(m, int *); 535 if (*v != 1) 536 return ENOPROTOOPT; 537 538 if (ip_mrouter != NULL) return EADDRINUSE; 539 540 ip_mrouter = so; 541 542 bzero((caddr_t)mfctable, sizeof(mfctable)); 543 bzero((caddr_t)nexpire, sizeof(nexpire)); 544 545 pim_assert = 0; 546 547 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 548 549 if (mrtdebug) 550 log(LOG_DEBUG, "ip_mrouter_init\n"); 551 552 return 0; 553 } 554 555 /* 556 * Disable multicast routing 557 */ 558 static int 559 X_ip_mrouter_done() 560 { 561 vifi_t vifi; 562 int i; 563 struct ifnet *ifp; 564 struct ifreq ifr; 565 struct mbuf *mb_rt; 566 struct mbuf *m; 567 struct rtdetq *rte; 568 int s; 569 570 s = splnet(); 571 572 /* 573 * For each phyint in use, disable promiscuous reception of all IP 574 * multicasts. 575 */ 576 for (vifi = 0; vifi < numvifs; vifi++) { 577 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 578 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 579 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 580 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr 581 = INADDR_ANY; 582 ifp = viftable[vifi].v_ifp; 583 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 584 } 585 } 586 bzero((caddr_t)tbftable, sizeof(tbftable)); 587 bzero((caddr_t)viftable, sizeof(viftable)); 588 numvifs = 0; 589 pim_assert = 0; 590 591 untimeout(expire_upcalls, (caddr_t)NULL); 592 593 /* 594 * Free all multicast forwarding cache entries. 595 */ 596 for (i = 0; i < MFCTBLSIZ; i++) { 597 mb_rt = mfctable[i]; 598 while (mb_rt) { 599 if (mb_rt->m_act != NULL) { 600 while (mb_rt->m_act) { 601 m = mb_rt->m_act; 602 mb_rt->m_act = m->m_act; 603 rte = mtod(m, struct rtdetq *); 604 m_freem(rte->m); 605 m_free(m); 606 } 607 } 608 mb_rt = m_free(mb_rt); 609 } 610 } 611 612 bzero((caddr_t)mfctable, sizeof(mfctable)); 613 614 /* 615 * Reset de-encapsulation cache 616 */ 617 last_encap_src = NULL; 618 last_encap_vif = NULL; 619 have_encap_tunnel = 0; 620 621 ip_mrouter = NULL; 622 623 splx(s); 624 625 if (mrtdebug) 626 log(LOG_DEBUG, "ip_mrouter_done\n"); 627 628 return 0; 629 } 630 631 #ifndef MROUTE_LKM 632 int (*ip_mrouter_done)(void) = X_ip_mrouter_done; 633 #endif 634 635 static int 636 get_version(mb) 637 struct mbuf *mb; 638 { 639 int *v; 640 641 v = mtod(mb, int *); 642 643 *v = 0x0305; /* XXX !!!! */ 644 mb->m_len = sizeof(int); 645 646 return 0; 647 } 648 649 /* 650 * Set PIM assert processing global 651 */ 652 static int 653 set_assert(i) 654 int *i; 655 { 656 if ((*i != 1) && (*i != 0)) 657 return EINVAL; 658 659 pim_assert = *i; 660 661 return 0; 662 } 663 664 /* 665 * Get PIM assert processing global 666 */ 667 static int 668 get_assert(m) 669 struct mbuf *m; 670 { 671 int *i; 672 673 i = mtod(m, int *); 674 675 *i = pim_assert; 676 677 return 0; 678 } 679 680 /* 681 * Add a vif to the vif table 682 */ 683 static int 684 add_vif(vifcp) 685 register struct vifctl *vifcp; 686 { 687 register struct vif *vifp = viftable + vifcp->vifc_vifi; 688 static struct sockaddr_in sin = {sizeof sin, AF_INET}; 689 struct ifaddr *ifa; 690 struct ifnet *ifp; 691 struct ifreq ifr; 692 int error, s; 693 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 694 695 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL; 696 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE; 697 698 /* Find the interface with an address in AF_INET family */ 699 sin.sin_addr = vifcp->vifc_lcl_addr; 700 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 701 if (ifa == 0) return EADDRNOTAVAIL; 702 ifp = ifa->ifa_ifp; 703 704 if (vifcp->vifc_flags & VIFF_TUNNEL) { 705 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 706 /* 707 * An encapsulating tunnel is wanted. Tell ipip_input() to 708 * start paying attention to encapsulated packets. 709 */ 710 if (have_encap_tunnel == 0) { 711 have_encap_tunnel = 1; 712 for (s = 0; s < MAXVIFS; ++s) { 713 multicast_decap_if[s].if_name = "mdecap"; 714 multicast_decap_if[s].if_unit = s; 715 } 716 } 717 /* 718 * Set interface to fake encapsulator interface 719 */ 720 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 721 /* 722 * Prepare cached route entry 723 */ 724 bzero(&vifp->v_route, sizeof(vifp->v_route)); 725 } else { 726 log(LOG_ERR, "source routed tunnels not supported\n"); 727 return EOPNOTSUPP; 728 } 729 } else { 730 /* Make sure the interface supports multicast */ 731 if ((ifp->if_flags & IFF_MULTICAST) == 0) 732 return EOPNOTSUPP; 733 734 /* Enable promiscuous reception of all IP multicasts from the if */ 735 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 736 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 737 s = splnet(); 738 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr); 739 splx(s); 740 if (error) 741 return error; 742 } 743 744 s = splnet(); 745 /* define parameters for the tbf structure */ 746 vifp->v_tbf = v_tbf; 747 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 748 vifp->v_tbf->tbf_n_tok = 0; 749 vifp->v_tbf->tbf_q_len = 0; 750 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 751 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 752 753 vifp->v_flags = vifcp->vifc_flags; 754 vifp->v_threshold = vifcp->vifc_threshold; 755 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 756 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 757 vifp->v_ifp = ifp; 758 /* scaling up here allows division by 1024 in critical code */ 759 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 760 vifp->v_rsvp_on = 0; 761 vifp->v_rsvpd = NULL; 762 /* initialize per vif pkt counters */ 763 vifp->v_pkt_in = 0; 764 vifp->v_pkt_out = 0; 765 vifp->v_bytes_in = 0; 766 vifp->v_bytes_out = 0; 767 splx(s); 768 769 /* Adjust numvifs up if the vifi is higher than numvifs */ 770 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 771 772 if (mrtdebug) 773 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 774 vifcp->vifc_vifi, 775 ntohl(vifcp->vifc_lcl_addr.s_addr), 776 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 777 ntohl(vifcp->vifc_rmt_addr.s_addr), 778 vifcp->vifc_threshold, 779 vifcp->vifc_rate_limit); 780 781 return 0; 782 } 783 784 /* 785 * Delete a vif from the vif table 786 */ 787 static int 788 del_vif(vifip) 789 vifi_t *vifip; 790 { 791 register struct vif *vifp = viftable + *vifip; 792 register vifi_t vifi; 793 register struct mbuf *m; 794 struct ifnet *ifp; 795 struct ifreq ifr; 796 int s; 797 798 if (*vifip >= numvifs) return EINVAL; 799 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL; 800 801 s = splnet(); 802 803 if (!(vifp->v_flags & VIFF_TUNNEL)) { 804 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 805 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 806 ifp = vifp->v_ifp; 807 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 808 } 809 810 if (vifp == last_encap_vif) { 811 last_encap_vif = 0; 812 last_encap_src = 0; 813 } 814 815 /* 816 * Free packets queued at the interface 817 */ 818 while (vifp->v_tbf->tbf_q) { 819 m = vifp->v_tbf->tbf_q; 820 vifp->v_tbf->tbf_q = m->m_act; 821 m_freem(m); 822 } 823 824 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 825 bzero((caddr_t)vifp, sizeof (*vifp)); 826 827 /* Adjust numvifs down */ 828 for (vifi = numvifs; vifi > 0; vifi--) 829 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break; 830 numvifs = vifi; 831 832 splx(s); 833 834 if (mrtdebug) 835 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 836 837 return 0; 838 } 839 840 /* 841 * Add an mfc entry 842 */ 843 static int 844 add_mfc(mfccp) 845 struct mfcctl *mfccp; 846 { 847 struct mfc *rt; 848 register struct mbuf *mb_rt; 849 u_long hash; 850 struct mbuf *mb_ntry; 851 struct rtdetq *rte; 852 register u_short nstl; 853 int s; 854 int i; 855 856 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 857 858 /* If an entry already exists, just update the fields */ 859 if (rt) { 860 if (mrtdebug & DEBUG_MFC) 861 log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n", 862 ntohl(mfccp->mfcc_origin.s_addr), 863 ntohl(mfccp->mfcc_mcastgrp.s_addr), 864 mfccp->mfcc_parent); 865 866 s = splnet(); 867 rt->mfc_parent = mfccp->mfcc_parent; 868 for (i = 0; i < numvifs; i++) 869 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 870 splx(s); 871 return 0; 872 } 873 874 /* 875 * Find the entry for which the upcall was made and update 876 */ 877 s = splnet(); 878 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 879 for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) { 880 881 rt = mtod(mb_rt, struct mfc *); 882 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 883 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 884 (mb_rt->m_act != NULL)) { 885 886 if (nstl++) 887 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %x\n", 888 "multiple kernel entries", 889 ntohl(mfccp->mfcc_origin.s_addr), 890 ntohl(mfccp->mfcc_mcastgrp.s_addr), 891 mfccp->mfcc_parent, mb_rt->m_act); 892 893 if (mrtdebug & DEBUG_MFC) 894 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %x\n", 895 ntohl(mfccp->mfcc_origin.s_addr), 896 ntohl(mfccp->mfcc_mcastgrp.s_addr), 897 mfccp->mfcc_parent, mb_rt->m_act); 898 899 rt->mfc_origin = mfccp->mfcc_origin; 900 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 901 rt->mfc_parent = mfccp->mfcc_parent; 902 for (i = 0; i < numvifs; i++) 903 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 904 /* initialize pkt counters per src-grp */ 905 rt->mfc_pkt_cnt = 0; 906 rt->mfc_byte_cnt = 0; 907 rt->mfc_wrong_if = 0; 908 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 909 910 rt->mfc_expire = 0; /* Don't clean this guy up */ 911 nexpire[hash]--; 912 913 /* free packets Qed at the end of this entry */ 914 while (mb_rt->m_act) { 915 mb_ntry = mb_rt->m_act; 916 rte = mtod(mb_ntry, struct rtdetq *); 917 /* #ifdef RSVP_ISI */ 918 ip_mdq(rte->m, rte->ifp, rt, -1); 919 /* #endif */ 920 mb_rt->m_act = mb_ntry->m_act; 921 m_freem(rte->m); 922 #ifdef UPCALL_TIMING 923 collate(&(rte->t)); 924 #endif /* UPCALL_TIMING */ 925 m_free(mb_ntry); 926 } 927 } 928 } 929 930 /* 931 * It is possible that an entry is being inserted without an upcall 932 */ 933 if (nstl == 0) { 934 if (mrtdebug & DEBUG_MFC) 935 log(LOG_DEBUG,"add_mfc no upcall h %d o %x g %x p %x\n", 936 hash, ntohl(mfccp->mfcc_origin.s_addr), 937 ntohl(mfccp->mfcc_mcastgrp.s_addr), 938 mfccp->mfcc_parent); 939 940 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 941 942 rt = mtod(mb_rt, struct mfc *); 943 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 944 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 945 946 rt->mfc_origin = mfccp->mfcc_origin; 947 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 948 rt->mfc_parent = mfccp->mfcc_parent; 949 for (i = 0; i < numvifs; i++) 950 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 951 /* initialize pkt counters per src-grp */ 952 rt->mfc_pkt_cnt = 0; 953 rt->mfc_byte_cnt = 0; 954 rt->mfc_wrong_if = 0; 955 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 956 if (rt->mfc_expire) 957 nexpire[hash]--; 958 rt->mfc_expire = 0; 959 } 960 } 961 if (mb_rt == NULL) { 962 /* no upcall, so make a new entry */ 963 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 964 if (mb_rt == NULL) { 965 splx(s); 966 return ENOBUFS; 967 } 968 969 rt = mtod(mb_rt, struct mfc *); 970 971 /* insert new entry at head of hash chain */ 972 rt->mfc_origin = mfccp->mfcc_origin; 973 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 974 rt->mfc_parent = mfccp->mfcc_parent; 975 for (i = 0; i < numvifs; i++) 976 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 977 /* initialize pkt counters per src-grp */ 978 rt->mfc_pkt_cnt = 0; 979 rt->mfc_byte_cnt = 0; 980 rt->mfc_wrong_if = 0; 981 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 982 rt->mfc_expire = 0; 983 984 /* link into table */ 985 mb_rt->m_next = mfctable[hash]; 986 mfctable[hash] = mb_rt; 987 mb_rt->m_act = NULL; 988 } 989 } 990 splx(s); 991 return 0; 992 } 993 994 #ifdef UPCALL_TIMING 995 /* 996 * collect delay statistics on the upcalls 997 */ 998 static void collate(t) 999 register struct timeval *t; 1000 { 1001 register u_long d; 1002 register struct timeval tp; 1003 register u_long delta; 1004 1005 GET_TIME(tp); 1006 1007 if (TV_LT(*t, tp)) 1008 { 1009 TV_DELTA(tp, *t, delta); 1010 1011 d = delta >> 10; 1012 if (d > 50) 1013 d = 50; 1014 1015 ++upcall_data[d]; 1016 } 1017 } 1018 #endif /* UPCALL_TIMING */ 1019 1020 /* 1021 * Delete an mfc entry 1022 */ 1023 static int 1024 del_mfc(mfccp) 1025 struct mfcctl *mfccp; 1026 { 1027 struct in_addr origin; 1028 struct in_addr mcastgrp; 1029 struct mfc *rt; 1030 struct mbuf *mb_rt; 1031 struct mbuf **nptr; 1032 u_long hash; 1033 int s; 1034 1035 origin = mfccp->mfcc_origin; 1036 mcastgrp = mfccp->mfcc_mcastgrp; 1037 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1038 1039 if (mrtdebug & DEBUG_MFC) 1040 log(LOG_DEBUG,"del_mfc orig %x mcastgrp %x\n", 1041 ntohl(origin.s_addr), ntohl(mcastgrp.s_addr)); 1042 1043 s = splnet(); 1044 1045 nptr = &mfctable[hash]; 1046 while ((mb_rt = *nptr) != NULL) { 1047 rt = mtod(mb_rt, struct mfc *); 1048 if (origin.s_addr == rt->mfc_origin.s_addr && 1049 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1050 mb_rt->m_act == NULL) 1051 break; 1052 1053 nptr = &mb_rt->m_next; 1054 } 1055 if (mb_rt == NULL) { 1056 splx(s); 1057 return EADDRNOTAVAIL; 1058 } 1059 1060 MFREE(mb_rt, *nptr); 1061 1062 splx(s); 1063 1064 return 0; 1065 } 1066 1067 /* 1068 * Send a message to mrouted on the multicast routing socket 1069 */ 1070 static int 1071 socket_send(s, mm, src) 1072 struct socket *s; 1073 struct mbuf *mm; 1074 struct sockaddr_in *src; 1075 { 1076 if (s) { 1077 if (sbappendaddr(&s->so_rcv, 1078 (struct sockaddr *)src, 1079 mm, (struct mbuf *)0) != 0) { 1080 sorwakeup(s); 1081 return 0; 1082 } 1083 } 1084 m_freem(mm); 1085 return -1; 1086 } 1087 1088 /* 1089 * IP multicast forwarding function. This function assumes that the packet 1090 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1091 * pointed to by "ifp", and the packet is to be relayed to other networks 1092 * that have members of the packet's destination IP multicast group. 1093 * 1094 * The packet is returned unscathed to the caller, unless it is 1095 * erroneous, in which case a non-zero return value tells the caller to 1096 * discard it. 1097 */ 1098 1099 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1100 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1101 1102 static int 1103 X_ip_mforward(ip, ifp, m, imo) 1104 register struct ip *ip; 1105 struct ifnet *ifp; 1106 struct mbuf *m; 1107 struct ip_moptions *imo; 1108 { 1109 register struct mfc *rt = 0; /* XXX uninit warning */ 1110 register u_char *ipoptions; 1111 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1112 static int srctun = 0; 1113 register struct mbuf *mm; 1114 int s; 1115 vifi_t vifi; 1116 struct vif *vifp; 1117 1118 if (mrtdebug & DEBUG_FORWARD) 1119 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %x\n", 1120 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1121 1122 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1123 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1124 /* 1125 * Packet arrived via a physical interface or 1126 * an encapsulated tunnel. 1127 */ 1128 } else { 1129 /* 1130 * Packet arrived through a source-route tunnel. 1131 * Source-route tunnels are no longer supported. 1132 */ 1133 if ((srctun++ % 1000) == 0) 1134 log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n", 1135 ntohl(ip->ip_src.s_addr)); 1136 1137 return 1; 1138 } 1139 1140 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1141 if (ip->ip_ttl < 255) 1142 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1143 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1144 vifp = viftable + vifi; 1145 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1146 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi, 1147 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1148 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1149 } 1150 return (ip_mdq(m, ifp, rt, vifi)); 1151 } 1152 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1153 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1154 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr)); 1155 if(!imo) 1156 printf("In fact, no options were specified at all\n"); 1157 } 1158 1159 /* 1160 * Don't forward a packet with time-to-live of zero or one, 1161 * or a packet destined to a local-only group. 1162 */ 1163 if (ip->ip_ttl <= 1 || 1164 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1165 return 0; 1166 1167 /* 1168 * Determine forwarding vifs from the forwarding cache table 1169 */ 1170 s = splnet(); 1171 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt); 1172 1173 /* Entry exists, so forward if necessary */ 1174 if (rt != NULL) { 1175 splx(s); 1176 return (ip_mdq(m, ifp, rt, -1)); 1177 } else { 1178 /* 1179 * If we don't have a route for packet's origin, 1180 * Make a copy of the packet & 1181 * send message to routing daemon 1182 */ 1183 1184 register struct mbuf *mb_rt; 1185 register struct mbuf *mb_ntry; 1186 register struct mbuf *mb0; 1187 register struct rtdetq *rte; 1188 register struct mbuf *rte_m; 1189 register u_long hash; 1190 register int npkts; 1191 #ifdef UPCALL_TIMING 1192 struct timeval tp; 1193 1194 GET_TIME(tp); 1195 #endif 1196 1197 mrtstat.mrts_no_route++; 1198 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1199 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1200 ntohl(ip->ip_src.s_addr), 1201 ntohl(ip->ip_dst.s_addr)); 1202 1203 /* 1204 * Allocate mbufs early so that we don't do extra work if we are 1205 * just going to fail anyway. 1206 */ 1207 MGET(mb_ntry, M_DONTWAIT, MT_DATA); 1208 if (mb_ntry == NULL) { 1209 splx(s); 1210 return ENOBUFS; 1211 } 1212 mb0 = m_copy(m, 0, M_COPYALL); 1213 if (mb0 == NULL) { 1214 m_free(mb_ntry); 1215 splx(s); 1216 return ENOBUFS; 1217 } 1218 1219 /* is there an upcall waiting for this packet? */ 1220 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1221 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 1222 rt = mtod(mb_rt, struct mfc *); 1223 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1224 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1225 (mb_rt->m_act != NULL)) 1226 break; 1227 } 1228 1229 if (mb_rt == NULL) { 1230 int hlen = ip->ip_hl << 2; 1231 int i; 1232 struct igmpmsg *im; 1233 1234 /* no upcall, so make a new entry */ 1235 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 1236 if (mb_rt == NULL) { 1237 m_free(mb_ntry); 1238 m_freem(mb0); 1239 splx(s); 1240 return ENOBUFS; 1241 } 1242 /* Make a copy of the header to send to the user level process */ 1243 mm = m_copy(m, 0, hlen); 1244 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1245 mm = m_pullup(mm, hlen); 1246 if (mm == NULL) { 1247 m_free(mb_ntry); 1248 m_freem(mb0); 1249 m_free(mb_rt); 1250 splx(s); 1251 return ENOBUFS; 1252 } 1253 1254 /* 1255 * Send message to routing daemon to install 1256 * a route into the kernel table 1257 */ 1258 k_igmpsrc.sin_addr = ip->ip_src; 1259 1260 im = mtod(mm, struct igmpmsg *); 1261 im->im_msgtype = IGMPMSG_NOCACHE; 1262 im->im_mbz = 0; 1263 1264 mrtstat.mrts_upcalls++; 1265 1266 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1267 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1268 ++mrtstat.mrts_upq_sockfull; 1269 m_free(mb_ntry); 1270 m_freem(mb0); 1271 m_free(mb_rt); 1272 splx(s); 1273 return ENOBUFS; 1274 } 1275 1276 rt = mtod(mb_rt, struct mfc *); 1277 1278 /* insert new entry at head of hash chain */ 1279 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1280 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1281 rt->mfc_expire = UPCALL_EXPIRE; 1282 nexpire[hash]++; 1283 for (i = 0; i < numvifs; i++) 1284 rt->mfc_ttls[i] = 0; 1285 rt->mfc_parent = -1; 1286 1287 /* link into table */ 1288 mb_rt->m_next = mfctable[hash]; 1289 mfctable[hash] = mb_rt; 1290 mb_rt->m_act = NULL; 1291 1292 rte_m = mb_rt; 1293 } else { 1294 /* determine if q has overflowed */ 1295 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act) 1296 npkts++; 1297 1298 if (npkts > MAX_UPQ) { 1299 mrtstat.mrts_upq_ovflw++; 1300 m_free(mb_ntry); 1301 m_freem(mb0); 1302 splx(s); 1303 return 0; 1304 } 1305 } 1306 1307 mb_ntry->m_act = NULL; 1308 rte = mtod(mb_ntry, struct rtdetq *); 1309 1310 rte->m = mb0; 1311 rte->ifp = ifp; 1312 #ifdef UPCALL_TIMING 1313 rte->t = tp; 1314 #endif 1315 1316 /* Add this entry to the end of the queue */ 1317 rte_m->m_act = mb_ntry; 1318 1319 splx(s); 1320 1321 return 0; 1322 } 1323 } 1324 1325 #ifndef MROUTE_LKM 1326 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 1327 struct ip_moptions *) = X_ip_mforward; 1328 #endif 1329 1330 /* 1331 * Clean up the cache entry if upcall is not serviced 1332 */ 1333 static void 1334 expire_upcalls(void *unused) 1335 { 1336 struct mbuf *mb_rt, *m, **nptr; 1337 struct rtdetq *rte; 1338 struct mfc *mfc; 1339 int i; 1340 int s; 1341 1342 s = splnet(); 1343 for (i = 0; i < MFCTBLSIZ; i++) { 1344 if (nexpire[i] == 0) 1345 continue; 1346 nptr = &mfctable[i]; 1347 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) { 1348 mfc = mtod(mb_rt, struct mfc *); 1349 1350 /* 1351 * Skip real cache entries 1352 * Make sure it wasn't marked to not expire (shouldn't happen) 1353 * If it expires now 1354 */ 1355 if (mb_rt->m_act != NULL && 1356 mfc->mfc_expire != 0 && 1357 --mfc->mfc_expire == 0) { 1358 if (mrtdebug & DEBUG_EXPIRE) 1359 log(LOG_DEBUG, "expire_upcalls: expiring (%x %x)\n", 1360 ntohl(mfc->mfc_origin.s_addr), 1361 ntohl(mfc->mfc_mcastgrp.s_addr)); 1362 /* 1363 * drop all the packets 1364 * free the mbuf with the pkt, if, timing info 1365 */ 1366 while (mb_rt->m_act) { 1367 m = mb_rt->m_act; 1368 mb_rt->m_act = m->m_act; 1369 1370 rte = mtod(m, struct rtdetq *); 1371 m_freem(rte->m); 1372 m_free(m); 1373 } 1374 ++mrtstat.mrts_cache_cleanups; 1375 nexpire[i]--; 1376 1377 MFREE(mb_rt, *nptr); 1378 } else { 1379 nptr = &mb_rt->m_next; 1380 } 1381 } 1382 } 1383 splx(s); 1384 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 1385 } 1386 1387 /* 1388 * Packet forwarding routine once entry in the cache is made 1389 */ 1390 static int 1391 ip_mdq(m, ifp, rt, xmt_vif) 1392 register struct mbuf *m; 1393 register struct ifnet *ifp; 1394 register struct mfc *rt; 1395 register vifi_t xmt_vif; 1396 { 1397 register struct ip *ip = mtod(m, struct ip *); 1398 register vifi_t vifi; 1399 register struct vif *vifp; 1400 register int plen = ntohs(ip->ip_len); 1401 1402 /* 1403 * Macro to send packet on vif. Since RSVP packets don't get counted on 1404 * input, they shouldn't get counted on output, so statistics keeping is 1405 * seperate. 1406 */ 1407 #define MC_SEND(ip,vifp,m) { \ 1408 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1409 encap_send((ip), (vifp), (m)); \ 1410 else \ 1411 phyint_send((ip), (vifp), (m)); \ 1412 } 1413 1414 /* 1415 * If xmt_vif is not -1, send on only the requested vif. 1416 * 1417 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1418 */ 1419 if (xmt_vif < numvifs) { 1420 MC_SEND(ip, viftable + xmt_vif, m); 1421 return 1; 1422 } 1423 1424 /* 1425 * Don't forward if it didn't arrive from the parent vif for its origin. 1426 */ 1427 vifi = rt->mfc_parent; 1428 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1429 /* came in the wrong interface */ 1430 if (mrtdebug & DEBUG_FORWARD) 1431 log(LOG_DEBUG, "wrong if: ifp %x vifi %d vififp %x\n", 1432 ifp, vifi, viftable[vifi].v_ifp); 1433 ++mrtstat.mrts_wrong_if; 1434 ++rt->mfc_wrong_if; 1435 /* 1436 * If we are doing PIM assert processing, and we are forwarding 1437 * packets on this interface, and it is a broadcast medium 1438 * interface (and not a tunnel), send a message to the routing daemon. 1439 */ 1440 if (pim_assert && rt->mfc_ttls[vifi] && 1441 (ifp->if_flags & IFF_BROADCAST) && 1442 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1443 struct sockaddr_in k_igmpsrc; 1444 struct mbuf *mm; 1445 struct igmpmsg *im; 1446 int hlen = ip->ip_hl << 2; 1447 struct timeval now; 1448 register u_long delta; 1449 1450 GET_TIME(now); 1451 1452 TV_DELTA(rt->mfc_last_assert, now, delta); 1453 1454 if (delta > ASSERT_MSG_TIME) { 1455 mm = m_copy(m, 0, hlen); 1456 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1457 mm = m_pullup(mm, hlen); 1458 if (mm == NULL) { 1459 return ENOBUFS; 1460 } 1461 1462 rt->mfc_last_assert = now; 1463 1464 im = mtod(mm, struct igmpmsg *); 1465 im->im_msgtype = IGMPMSG_WRONGVIF; 1466 im->im_mbz = 0; 1467 im->im_vif = vifi; 1468 1469 k_igmpsrc.sin_addr = im->im_src; 1470 1471 socket_send(ip_mrouter, mm, &k_igmpsrc); 1472 } 1473 } 1474 return 0; 1475 } 1476 1477 /* If I sourced this packet, it counts as output, else it was input. */ 1478 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1479 viftable[vifi].v_pkt_out++; 1480 viftable[vifi].v_bytes_out += plen; 1481 } else { 1482 viftable[vifi].v_pkt_in++; 1483 viftable[vifi].v_bytes_in += plen; 1484 } 1485 rt->mfc_pkt_cnt++; 1486 rt->mfc_byte_cnt += plen; 1487 1488 /* 1489 * For each vif, decide if a copy of the packet should be forwarded. 1490 * Forward if: 1491 * - the ttl exceeds the vif's threshold 1492 * - there are group members downstream on interface 1493 */ 1494 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1495 if ((rt->mfc_ttls[vifi] > 0) && 1496 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1497 vifp->v_pkt_out++; 1498 vifp->v_bytes_out += plen; 1499 MC_SEND(ip, vifp, m); 1500 } 1501 1502 return 0; 1503 } 1504 1505 /* 1506 * check if a vif number is legal/ok. This is used by ip_output, to export 1507 * numvifs there, 1508 */ 1509 static int 1510 X_legal_vif_num(vif) 1511 int vif; 1512 { 1513 if (vif >= 0 && vif < numvifs) 1514 return(1); 1515 else 1516 return(0); 1517 } 1518 1519 #ifndef MROUTE_LKM 1520 int (*legal_vif_num)(int) = X_legal_vif_num; 1521 #endif 1522 1523 /* 1524 * Return the local address used by this vif 1525 */ 1526 static u_long 1527 X_ip_mcast_src(vifi) 1528 int vifi; 1529 { 1530 if (vifi >= 0 && vifi < numvifs) 1531 return viftable[vifi].v_lcl_addr.s_addr; 1532 else 1533 return INADDR_ANY; 1534 } 1535 1536 #ifndef MROUTE_LKM 1537 u_long (*ip_mcast_src)(int) = X_ip_mcast_src; 1538 #endif 1539 1540 static void 1541 phyint_send(ip, vifp, m) 1542 struct ip *ip; 1543 struct vif *vifp; 1544 struct mbuf *m; 1545 { 1546 register struct mbuf *mb_copy; 1547 register int hlen = ip->ip_hl << 2; 1548 1549 /* 1550 * Make a new reference to the packet; make sure that 1551 * the IP header is actually copied, not just referenced, 1552 * so that ip_output() only scribbles on the copy. 1553 */ 1554 mb_copy = m_copy(m, 0, M_COPYALL); 1555 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1556 mb_copy = m_pullup(mb_copy, hlen); 1557 if (mb_copy == NULL) 1558 return; 1559 1560 if (vifp->v_rate_limit <= 0) 1561 tbf_send_packet(vifp, mb_copy); 1562 else 1563 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1564 } 1565 1566 static void 1567 encap_send(ip, vifp, m) 1568 register struct ip *ip; 1569 register struct vif *vifp; 1570 register struct mbuf *m; 1571 { 1572 register struct mbuf *mb_copy; 1573 register struct ip *ip_copy; 1574 register int i, len = ip->ip_len; 1575 1576 /* 1577 * copy the old packet & pullup it's IP header into the 1578 * new mbuf so we can modify it. Try to fill the new 1579 * mbuf since if we don't the ethernet driver will. 1580 */ 1581 MGET(mb_copy, M_DONTWAIT, MT_DATA); 1582 if (mb_copy == NULL) 1583 return; 1584 mb_copy->m_data += 16; 1585 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1586 1587 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1588 m_freem(mb_copy); 1589 return; 1590 } 1591 i = MHLEN - M_LEADINGSPACE(mb_copy); 1592 if (i > len) 1593 i = len; 1594 mb_copy = m_pullup(mb_copy, i); 1595 if (mb_copy == NULL) 1596 return; 1597 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1598 1599 /* 1600 * fill in the encapsulating IP header. 1601 */ 1602 ip_copy = mtod(mb_copy, struct ip *); 1603 *ip_copy = multicast_encap_iphdr; 1604 ip_copy->ip_id = htons(ip_id++); 1605 ip_copy->ip_len += len; 1606 ip_copy->ip_src = vifp->v_lcl_addr; 1607 ip_copy->ip_dst = vifp->v_rmt_addr; 1608 1609 /* 1610 * turn the encapsulated IP header back into a valid one. 1611 */ 1612 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1613 --ip->ip_ttl; 1614 HTONS(ip->ip_len); 1615 HTONS(ip->ip_off); 1616 ip->ip_sum = 0; 1617 #if defined(LBL) && !defined(ultrix) 1618 ip->ip_sum = ~oc_cksum((caddr_t)ip, ip->ip_hl << 2, 0); 1619 #else 1620 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1621 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1622 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1623 #endif 1624 1625 if (vifp->v_rate_limit <= 0) 1626 tbf_send_packet(vifp, mb_copy); 1627 else 1628 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1629 } 1630 1631 /* 1632 * De-encapsulate a packet and feed it back through ip input (this 1633 * routine is called whenever IP gets a packet with proto type 1634 * ENCAP_PROTO and a local destination address). 1635 */ 1636 void 1637 #ifdef MROUTE_LKM 1638 X_ipip_input(m) 1639 #else 1640 ipip_input(m, iphlen) 1641 #endif 1642 register struct mbuf *m; 1643 int iphlen; 1644 { 1645 struct ifnet *ifp = m->m_pkthdr.rcvif; 1646 register struct ip *ip = mtod(m, struct ip *); 1647 register int hlen = ip->ip_hl << 2; 1648 register int s; 1649 register struct ifqueue *ifq; 1650 register struct vif *vifp; 1651 1652 if (!have_encap_tunnel) { 1653 rip_input(m); 1654 return; 1655 } 1656 /* 1657 * dump the packet if it's not to a multicast destination or if 1658 * we don't have an encapsulating tunnel with the source. 1659 * Note: This code assumes that the remote site IP address 1660 * uniquely identifies the tunnel (i.e., that this site has 1661 * at most one tunnel with the remote site). 1662 */ 1663 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) { 1664 ++mrtstat.mrts_bad_tunnel; 1665 m_freem(m); 1666 return; 1667 } 1668 if (ip->ip_src.s_addr != last_encap_src) { 1669 register struct vif *vife; 1670 1671 vifp = viftable; 1672 vife = vifp + numvifs; 1673 last_encap_src = ip->ip_src.s_addr; 1674 last_encap_vif = 0; 1675 for ( ; vifp < vife; ++vifp) 1676 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1677 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1678 == VIFF_TUNNEL) 1679 last_encap_vif = vifp; 1680 break; 1681 } 1682 } 1683 if ((vifp = last_encap_vif) == 0) { 1684 last_encap_src = 0; 1685 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1686 m_freem(m); 1687 if (mrtdebug) 1688 log(LOG_DEBUG, "ip_mforward: no tunnel with %x\n", 1689 ntohl(ip->ip_src.s_addr)); 1690 return; 1691 } 1692 ifp = vifp->v_ifp; 1693 1694 if (hlen > IP_HDR_LEN) 1695 ip_stripoptions(m, (struct mbuf *) 0); 1696 m->m_data += IP_HDR_LEN; 1697 m->m_len -= IP_HDR_LEN; 1698 m->m_pkthdr.len -= IP_HDR_LEN; 1699 m->m_pkthdr.rcvif = ifp; 1700 1701 ifq = &ipintrq; 1702 s = splimp(); 1703 if (IF_QFULL(ifq)) { 1704 IF_DROP(ifq); 1705 m_freem(m); 1706 } else { 1707 IF_ENQUEUE(ifq, m); 1708 /* 1709 * normally we would need a "schednetisr(NETISR_IP)" 1710 * here but we were called by ip_input and it is going 1711 * to loop back & try to dequeue the packet we just 1712 * queued as soon as we return so we avoid the 1713 * unnecessary software interrrupt. 1714 */ 1715 } 1716 splx(s); 1717 } 1718 1719 /* 1720 * Token bucket filter module 1721 */ 1722 1723 static void 1724 tbf_control(vifp, m, ip, p_len) 1725 register struct vif *vifp; 1726 register struct mbuf *m; 1727 register struct ip *ip; 1728 register u_long p_len; 1729 { 1730 register struct tbf *t = vifp->v_tbf; 1731 1732 if (p_len > MAX_BKT_SIZE) { 1733 /* drop if packet is too large */ 1734 mrtstat.mrts_pkt2large++; 1735 m_freem(m); 1736 return; 1737 } 1738 1739 tbf_update_tokens(vifp); 1740 1741 /* if there are enough tokens, 1742 * and the queue is empty, 1743 * send this packet out 1744 */ 1745 1746 if (t->tbf_q_len == 0) { 1747 /* queue empty, send packet if enough tokens */ 1748 if (p_len <= t->tbf_n_tok) { 1749 t->tbf_n_tok -= p_len; 1750 tbf_send_packet(vifp, m); 1751 } else { 1752 /* queue packet and timeout till later */ 1753 tbf_queue(vifp, m); 1754 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1755 } 1756 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1757 /* finite queue length, so queue pkts and process queue */ 1758 tbf_queue(vifp, m); 1759 tbf_process_q(vifp); 1760 } else { 1761 /* queue length too much, try to dq and queue and process */ 1762 if (!tbf_dq_sel(vifp, ip)) { 1763 mrtstat.mrts_q_overflow++; 1764 m_freem(m); 1765 return; 1766 } else { 1767 tbf_queue(vifp, m); 1768 tbf_process_q(vifp); 1769 } 1770 } 1771 return; 1772 } 1773 1774 /* 1775 * adds a packet to the queue at the interface 1776 */ 1777 static void 1778 tbf_queue(vifp, m) 1779 register struct vif *vifp; 1780 register struct mbuf *m; 1781 { 1782 register int s = splnet(); 1783 register struct tbf *t = vifp->v_tbf; 1784 1785 if (t->tbf_t == NULL) { 1786 /* Queue was empty */ 1787 t->tbf_q = m; 1788 } else { 1789 /* Insert at tail */ 1790 t->tbf_t->m_act = m; 1791 } 1792 1793 /* Set new tail pointer */ 1794 t->tbf_t = m; 1795 1796 #ifdef DIAGNOSTIC 1797 /* Make sure we didn't get fed a bogus mbuf */ 1798 if (m->m_act) 1799 panic("tbf_queue: m_act"); 1800 #endif 1801 m->m_act = NULL; 1802 1803 t->tbf_q_len++; 1804 1805 splx(s); 1806 } 1807 1808 1809 /* 1810 * processes the queue at the interface 1811 */ 1812 static void 1813 tbf_process_q(vifp) 1814 register struct vif *vifp; 1815 { 1816 register struct mbuf *m; 1817 register int len; 1818 register int s = splnet(); 1819 register struct tbf *t = vifp->v_tbf; 1820 1821 /* loop through the queue at the interface and send as many packets 1822 * as possible 1823 */ 1824 while (t->tbf_q_len > 0) { 1825 m = t->tbf_q; 1826 1827 len = mtod(m, struct ip *)->ip_len; 1828 1829 /* determine if the packet can be sent */ 1830 if (len <= t->tbf_n_tok) { 1831 /* if so, 1832 * reduce no of tokens, dequeue the packet, 1833 * send the packet. 1834 */ 1835 t->tbf_n_tok -= len; 1836 1837 t->tbf_q = m->m_act; 1838 if (--t->tbf_q_len == 0) 1839 t->tbf_t = NULL; 1840 1841 m->m_act = NULL; 1842 tbf_send_packet(vifp, m); 1843 1844 } else break; 1845 } 1846 splx(s); 1847 } 1848 1849 static void 1850 tbf_reprocess_q(xvifp) 1851 void *xvifp; 1852 { 1853 register struct vif *vifp = xvifp; 1854 if (ip_mrouter == NULL) 1855 return; 1856 1857 tbf_update_tokens(vifp); 1858 1859 tbf_process_q(vifp); 1860 1861 if (vifp->v_tbf->tbf_q_len) 1862 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1863 } 1864 1865 /* function that will selectively discard a member of the queue 1866 * based on the precedence value and the priority 1867 */ 1868 static int 1869 tbf_dq_sel(vifp, ip) 1870 register struct vif *vifp; 1871 register struct ip *ip; 1872 { 1873 register int s = splnet(); 1874 register u_int p; 1875 register struct mbuf *m, *last; 1876 register struct mbuf **np; 1877 register struct tbf *t = vifp->v_tbf; 1878 1879 p = priority(vifp, ip); 1880 1881 np = &t->tbf_q; 1882 last = NULL; 1883 while ((m = *np) != NULL) { 1884 if (p > priority(vifp, mtod(m, struct ip *))) { 1885 *np = m->m_act; 1886 /* If we're removing the last packet, fix the tail pointer */ 1887 if (m == t->tbf_t) 1888 t->tbf_t = last; 1889 m_freem(m); 1890 /* it's impossible for the queue to be empty, but 1891 * we check anyway. */ 1892 if (--t->tbf_q_len == 0) 1893 t->tbf_t = NULL; 1894 splx(s); 1895 mrtstat.mrts_drop_sel++; 1896 return(1); 1897 } 1898 np = &m->m_act; 1899 last = m; 1900 } 1901 splx(s); 1902 return(0); 1903 } 1904 1905 static void 1906 tbf_send_packet(vifp, m) 1907 register struct vif *vifp; 1908 register struct mbuf *m; 1909 { 1910 struct ip_moptions imo; 1911 int error; 1912 int s = splnet(); 1913 1914 if (vifp->v_flags & VIFF_TUNNEL) { 1915 /* If tunnel options */ 1916 ip_output(m, (struct mbuf *)0, (struct route *)0, 1917 IP_FORWARDING, (struct ip_moptions *)0); 1918 } else { 1919 imo.imo_multicast_ifp = vifp->v_ifp; 1920 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1921 imo.imo_multicast_loop = 1; 1922 imo.imo_multicast_vif = -1; 1923 1924 error = ip_output(m, (struct mbuf *)0, (struct route *)0, 1925 IP_FORWARDING, &imo); 1926 1927 if (mrtdebug & DEBUG_XMIT) 1928 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1929 vifp - viftable, error); 1930 } 1931 splx(s); 1932 } 1933 1934 /* determine the current time and then 1935 * the elapsed time (between the last time and time now) 1936 * in milliseconds & update the no. of tokens in the bucket 1937 */ 1938 static void 1939 tbf_update_tokens(vifp) 1940 register struct vif *vifp; 1941 { 1942 struct timeval tp; 1943 register u_long tm; 1944 register int s = splnet(); 1945 register struct tbf *t = vifp->v_tbf; 1946 1947 GET_TIME(tp); 1948 1949 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1950 1951 /* 1952 * This formula is actually 1953 * "time in seconds" * "bytes/second". 1954 * 1955 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1956 * 1957 * The (1000/1024) was introduced in add_vif to optimize 1958 * this divide into a shift. 1959 */ 1960 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1961 t->tbf_last_pkt_t = tp; 1962 1963 if (t->tbf_n_tok > MAX_BKT_SIZE) 1964 t->tbf_n_tok = MAX_BKT_SIZE; 1965 1966 splx(s); 1967 } 1968 1969 static int 1970 priority(vifp, ip) 1971 register struct vif *vifp; 1972 register struct ip *ip; 1973 { 1974 register int prio; 1975 1976 /* temporary hack; may add general packet classifier some day */ 1977 1978 /* 1979 * The UDP port space is divided up into four priority ranges: 1980 * [0, 16384) : unclassified - lowest priority 1981 * [16384, 32768) : audio - highest priority 1982 * [32768, 49152) : whiteboard - medium priority 1983 * [49152, 65536) : video - low priority 1984 */ 1985 if (ip->ip_p == IPPROTO_UDP) { 1986 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1987 switch (ntohs(udp->uh_dport) & 0xc000) { 1988 case 0x4000: 1989 prio = 70; 1990 break; 1991 case 0x8000: 1992 prio = 60; 1993 break; 1994 case 0xc000: 1995 prio = 55; 1996 break; 1997 default: 1998 prio = 50; 1999 break; 2000 } 2001 if (tbfdebug > 1) 2002 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio); 2003 } else { 2004 prio = 50; 2005 } 2006 return prio; 2007 } 2008 2009 /* 2010 * End of token bucket filter modifications 2011 */ 2012 2013 int 2014 ip_rsvp_vif_init(so, m) 2015 struct socket *so; 2016 struct mbuf *m; 2017 { 2018 int i; 2019 register int s; 2020 2021 if (rsvpdebug) 2022 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2023 so->so_type, so->so_proto->pr_protocol); 2024 2025 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2026 return EOPNOTSUPP; 2027 2028 /* Check mbuf. */ 2029 if (m == NULL || m->m_len != sizeof(int)) { 2030 return EINVAL; 2031 } 2032 i = *(mtod(m, int *)); 2033 2034 if (rsvpdebug) 2035 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on); 2036 2037 s = splnet(); 2038 2039 /* Check vif. */ 2040 if (!legal_vif_num(i)) { 2041 splx(s); 2042 return EADDRNOTAVAIL; 2043 } 2044 2045 /* Check if socket is available. */ 2046 if (viftable[i].v_rsvpd != NULL) { 2047 splx(s); 2048 return EADDRINUSE; 2049 } 2050 2051 viftable[i].v_rsvpd = so; 2052 /* This may seem silly, but we need to be sure we don't over-increment 2053 * the RSVP counter, in case something slips up. 2054 */ 2055 if (!viftable[i].v_rsvp_on) { 2056 viftable[i].v_rsvp_on = 1; 2057 rsvp_on++; 2058 } 2059 2060 splx(s); 2061 return 0; 2062 } 2063 2064 int 2065 ip_rsvp_vif_done(so, m) 2066 struct socket *so; 2067 struct mbuf *m; 2068 { 2069 int i; 2070 register int s; 2071 2072 if (rsvpdebug) 2073 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2074 so->so_type, so->so_proto->pr_protocol); 2075 2076 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2077 return EOPNOTSUPP; 2078 2079 /* Check mbuf. */ 2080 if (m == NULL || m->m_len != sizeof(int)) { 2081 return EINVAL; 2082 } 2083 i = *(mtod(m, int *)); 2084 2085 s = splnet(); 2086 2087 /* Check vif. */ 2088 if (!legal_vif_num(i)) { 2089 splx(s); 2090 return EADDRNOTAVAIL; 2091 } 2092 2093 if (rsvpdebug) 2094 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n", 2095 viftable[i].v_rsvpd, so); 2096 2097 viftable[i].v_rsvpd = NULL; 2098 /* This may seem silly, but we need to be sure we don't over-decrement 2099 * the RSVP counter, in case something slips up. 2100 */ 2101 if (viftable[i].v_rsvp_on) { 2102 viftable[i].v_rsvp_on = 0; 2103 rsvp_on--; 2104 } 2105 2106 splx(s); 2107 return 0; 2108 } 2109 2110 void 2111 ip_rsvp_force_done(so) 2112 struct socket *so; 2113 { 2114 int vifi; 2115 register int s; 2116 2117 /* Don't bother if it is not the right type of socket. */ 2118 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2119 return; 2120 2121 s = splnet(); 2122 2123 /* The socket may be attached to more than one vif...this 2124 * is perfectly legal. 2125 */ 2126 for (vifi = 0; vifi < numvifs; vifi++) { 2127 if (viftable[vifi].v_rsvpd == so) { 2128 viftable[vifi].v_rsvpd = NULL; 2129 /* This may seem silly, but we need to be sure we don't 2130 * over-decrement the RSVP counter, in case something slips up. 2131 */ 2132 if (viftable[vifi].v_rsvp_on) { 2133 viftable[vifi].v_rsvp_on = 0; 2134 rsvp_on--; 2135 } 2136 } 2137 } 2138 2139 splx(s); 2140 return; 2141 } 2142 2143 void 2144 rsvp_input(m, iphlen) 2145 struct mbuf *m; 2146 int iphlen; 2147 { 2148 int vifi; 2149 register struct ip *ip = mtod(m, struct ip *); 2150 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2151 register int s; 2152 struct ifnet *ifp; 2153 2154 if (rsvpdebug) 2155 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2156 2157 /* Can still get packets with rsvp_on = 0 if there is a local member 2158 * of the group to which the RSVP packet is addressed. But in this 2159 * case we want to throw the packet away. 2160 */ 2161 if (!rsvp_on) { 2162 m_freem(m); 2163 return; 2164 } 2165 2166 /* If the old-style non-vif-associated socket is set, then use 2167 * it and ignore the new ones. 2168 */ 2169 if (ip_rsvpd != NULL) { 2170 if (rsvpdebug) 2171 printf("rsvp_input: Sending packet up old-style socket\n"); 2172 rip_input(m); 2173 return; 2174 } 2175 2176 s = splnet(); 2177 2178 if (rsvpdebug) 2179 printf("rsvp_input: check vifs\n"); 2180 2181 #ifdef DIAGNOSTIC 2182 if (!(m->m_flags & M_PKTHDR)) 2183 panic("rsvp_input no hdr"); 2184 #endif 2185 2186 ifp = m->m_pkthdr.rcvif; 2187 /* Find which vif the packet arrived on. */ 2188 for (vifi = 0; vifi < numvifs; vifi++) { 2189 if (viftable[vifi].v_ifp == ifp) 2190 break; 2191 } 2192 2193 if (vifi == numvifs) { 2194 /* Can't find vif packet arrived on. Drop packet. */ 2195 if (rsvpdebug) 2196 printf("rsvp_input: Can't find vif for packet...dropping it.\n"); 2197 m_freem(m); 2198 splx(s); 2199 return; 2200 } 2201 2202 if (rsvpdebug) 2203 printf("rsvp_input: check socket\n"); 2204 2205 if (viftable[vifi].v_rsvpd == NULL) { 2206 /* drop packet, since there is no specific socket for this 2207 * interface */ 2208 if (rsvpdebug) 2209 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2210 m_freem(m); 2211 splx(s); 2212 return; 2213 } 2214 rsvp_src.sin_addr = ip->ip_src; 2215 2216 if (rsvpdebug && m) 2217 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2218 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2219 2220 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2221 if (rsvpdebug) 2222 printf("rsvp_input: Failed to append to socket\n"); 2223 else 2224 if (rsvpdebug) 2225 printf("rsvp_input: send packet up\n"); 2226 2227 splx(s); 2228 } 2229 2230 #ifdef MROUTE_LKM 2231 #include <sys/conf.h> 2232 #include <sys/exec.h> 2233 #include <sys/sysent.h> 2234 #include <sys/lkm.h> 2235 2236 MOD_MISC("ip_mroute_mod") 2237 2238 static int 2239 ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd) 2240 { 2241 int i; 2242 struct lkm_misc *args = lkmtp->private.lkm_misc; 2243 int err = 0; 2244 2245 switch(cmd) { 2246 static int (*old_ip_mrouter_cmd)(); 2247 static int (*old_ip_mrouter_done)(); 2248 static int (*old_ip_mforward)(); 2249 static int (*old_mrt_ioctl)(); 2250 static void (*old_proto4_input)(); 2251 static int (*old_legal_vif_num)(); 2252 extern struct protosw inetsw[]; 2253 2254 case LKM_E_LOAD: 2255 if(lkmexists(lkmtp) || ip_mrtproto) 2256 return(EEXIST); 2257 old_ip_mrouter_cmd = ip_mrouter_cmd; 2258 ip_mrouter_cmd = X_ip_mrouter_cmd; 2259 old_ip_mrouter_done = ip_mrouter_done; 2260 ip_mrouter_done = X_ip_mrouter_done; 2261 old_ip_mforward = ip_mforward; 2262 ip_mforward = X_ip_mforward; 2263 old_mrt_ioctl = mrt_ioctl; 2264 mrt_ioctl = X_mrt_ioctl; 2265 old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input; 2266 inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input; 2267 old_legal_vif_num = legal_vif_num; 2268 legal_vif_num = X_legal_vif_num; 2269 ip_mrtproto = IGMP_DVMRP; 2270 2271 printf("\nIP multicast routing loaded\n"); 2272 break; 2273 2274 case LKM_E_UNLOAD: 2275 if (ip_mrouter) 2276 return EINVAL; 2277 2278 ip_mrouter_cmd = old_ip_mrouter_cmd; 2279 ip_mrouter_done = old_ip_mrouter_done; 2280 ip_mforward = old_ip_mforward; 2281 mrt_ioctl = old_mrt_ioctl; 2282 inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input; 2283 legal_vif_num = old_legal_vif_num; 2284 ip_mrtproto = 0; 2285 break; 2286 2287 default: 2288 err = EINVAL; 2289 break; 2290 } 2291 2292 return(err); 2293 } 2294 2295 int 2296 ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) { 2297 DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle, 2298 nosys); 2299 } 2300 2301 #endif /* MROUTE_LKM */ 2302 #endif /* MROUTING */ 2303