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 1.33 1996/07/12 04:11:48 bde Exp $ 13 */ 14 15 #include "opt_mrouting.h" 16 17 #include <sys/param.h> 18 #include <sys/queue.h> 19 #include <sys/systm.h> 20 #include <sys/mbuf.h> 21 #include <sys/socket.h> 22 #include <sys/socketvar.h> 23 #include <sys/protosw.h> 24 #include <sys/errno.h> 25 #include <sys/time.h> 26 #include <sys/kernel.h> 27 #include <sys/ioctl.h> 28 #include <sys/syslog.h> 29 #include <net/if.h> 30 #include <net/route.h> 31 #include <netinet/in.h> 32 #include <netinet/in_systm.h> 33 #include <netinet/ip.h> 34 #include <netinet/ip_var.h> 35 #include <netinet/in_pcb.h> 36 #include <netinet/in_var.h> 37 #include <netinet/igmp.h> 38 #include <netinet/igmp_var.h> 39 #include <netinet/ip_mroute.h> 40 #include <netinet/udp.h> 41 42 #ifndef NTOHL 43 #if BYTE_ORDER != BIG_ENDIAN 44 #define NTOHL(d) ((d) = ntohl((d))) 45 #define NTOHS(d) ((d) = ntohs((u_short)(d))) 46 #define HTONL(d) ((d) = htonl((d))) 47 #define HTONS(d) ((d) = htons((u_short)(d))) 48 #else 49 #define NTOHL(d) 50 #define NTOHS(d) 51 #define HTONL(d) 52 #define HTONS(d) 53 #endif 54 #endif 55 56 #ifndef MROUTING 57 extern u_long _ip_mcast_src __P((int vifi)); 58 extern int _ip_mforward __P((struct ip *ip, struct ifnet *ifp, 59 struct mbuf *m, struct ip_moptions *imo)); 60 extern int _ip_mrouter_done __P((void)); 61 extern int _ip_mrouter_get __P((int cmd, struct socket *so, 62 struct mbuf **m)); 63 extern int _ip_mrouter_set __P((int cmd, struct socket *so, 64 struct mbuf *m)); 65 extern int _mrt_ioctl __P((int req, caddr_t data, struct proc *p)); 66 67 /* 68 * Dummy routines and globals used when multicast routing is not compiled in. 69 */ 70 71 struct socket *ip_mrouter = NULL; 72 static u_int ip_mrtproto = 0; 73 static struct mrtstat mrtstat; 74 u_int rsvpdebug = 0; 75 76 int 77 _ip_mrouter_set(cmd, so, m) 78 int cmd; 79 struct socket *so; 80 struct mbuf *m; 81 { 82 return(EOPNOTSUPP); 83 } 84 85 int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = _ip_mrouter_set; 86 87 88 int 89 _ip_mrouter_get(cmd, so, m) 90 int cmd; 91 struct socket *so; 92 struct mbuf **m; 93 { 94 return(EOPNOTSUPP); 95 } 96 97 int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = _ip_mrouter_get; 98 99 int 100 _ip_mrouter_done() 101 { 102 return(0); 103 } 104 105 int (*ip_mrouter_done)(void) = _ip_mrouter_done; 106 107 int 108 _ip_mforward(ip, ifp, m, imo) 109 struct ip *ip; 110 struct ifnet *ifp; 111 struct mbuf *m; 112 struct ip_moptions *imo; 113 { 114 return(0); 115 } 116 117 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 118 struct ip_moptions *) = _ip_mforward; 119 120 int 121 _mrt_ioctl(int req, caddr_t data, struct proc *p) 122 { 123 return EOPNOTSUPP; 124 } 125 126 int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl; 127 128 void 129 rsvp_input(m, iphlen) /* XXX must fixup manually */ 130 struct mbuf *m; 131 int iphlen; 132 { 133 /* Can still get packets with rsvp_on = 0 if there is a local member 134 * of the group to which the RSVP packet is addressed. But in this 135 * case we want to throw the packet away. 136 */ 137 if (!rsvp_on) { 138 m_freem(m); 139 return; 140 } 141 142 if (ip_rsvpd != NULL) { 143 if (rsvpdebug) 144 printf("rsvp_input: Sending packet up old-style socket\n"); 145 rip_input(m, iphlen); 146 return; 147 } 148 /* Drop the packet */ 149 m_freem(m); 150 } 151 152 void ipip_input(struct mbuf *m, int iphlen) { /* XXX must fixup manually */ 153 rip_input(m, iphlen); 154 } 155 156 int (*legal_vif_num)(int) = 0; 157 158 /* 159 * This should never be called, since IP_MULTICAST_VIF should fail, but 160 * just in case it does get called, the code a little lower in ip_output 161 * will assign the packet a local address. 162 */ 163 u_long 164 _ip_mcast_src(int vifi) { return INADDR_ANY; } 165 u_long (*ip_mcast_src)(int) = _ip_mcast_src; 166 167 int 168 ip_rsvp_vif_init(so, m) 169 struct socket *so; 170 struct mbuf *m; 171 { 172 return(EINVAL); 173 } 174 175 int 176 ip_rsvp_vif_done(so, m) 177 struct socket *so; 178 struct mbuf *m; 179 { 180 return(EINVAL); 181 } 182 183 void 184 ip_rsvp_force_done(so) 185 struct socket *so; 186 { 187 return; 188 } 189 190 #else /* MROUTING */ 191 192 #define M_HASCL(m) ((m)->m_flags & M_EXT) 193 194 #define INSIZ sizeof(struct in_addr) 195 #define same(a1, a2) \ 196 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0) 197 198 #define MT_MRTABLE MT_RTABLE /* since nothing else uses it */ 199 200 /* 201 * Globals. All but ip_mrouter and ip_mrtproto could be static, 202 * except for netstat or debugging purposes. 203 */ 204 #ifndef MROUTE_LKM 205 struct socket *ip_mrouter = NULL; 206 struct mrtstat mrtstat; 207 208 int ip_mrtproto = IGMP_DVMRP; /* for netstat only */ 209 #else /* MROUTE_LKM */ 210 extern void X_ipip_input __P((struct mbuf *m, int iphlen)); 211 extern struct mrtstat mrtstat; 212 static int ip_mrtproto; 213 #endif 214 215 #define NO_RTE_FOUND 0x1 216 #define RTE_FOUND 0x2 217 218 static struct mbuf *mfctable[MFCTBLSIZ]; 219 static u_char nexpire[MFCTBLSIZ]; 220 static struct vif viftable[MAXVIFS]; 221 static u_int mrtdebug = 0; /* debug level */ 222 #define DEBUG_MFC 0x02 223 #define DEBUG_FORWARD 0x04 224 #define DEBUG_EXPIRE 0x08 225 #define DEBUG_XMIT 0x10 226 static u_int tbfdebug = 0; /* tbf debug level */ 227 static u_int rsvpdebug = 0; /* rsvp debug level */ 228 229 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 230 #define UPCALL_EXPIRE 6 /* number of timeouts */ 231 232 /* 233 * Define the token bucket filter structures 234 * tbftable -> each vif has one of these for storing info 235 */ 236 237 static struct tbf tbftable[MAXVIFS]; 238 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 239 240 /* 241 * 'Interfaces' associated with decapsulator (so we can tell 242 * packets that went through it from ones that get reflected 243 * by a broken gateway). These interfaces are never linked into 244 * the system ifnet list & no routes point to them. I.e., packets 245 * can't be sent this way. They only exist as a placeholder for 246 * multicast source verification. 247 */ 248 static struct ifnet multicast_decap_if[MAXVIFS]; 249 250 #define ENCAP_TTL 64 251 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 252 253 /* prototype IP hdr for encapsulated packets */ 254 static struct ip multicast_encap_iphdr = { 255 #if BYTE_ORDER == LITTLE_ENDIAN 256 sizeof(struct ip) >> 2, IPVERSION, 257 #else 258 IPVERSION, sizeof(struct ip) >> 2, 259 #endif 260 0, /* tos */ 261 sizeof(struct ip), /* total length */ 262 0, /* id */ 263 0, /* frag offset */ 264 ENCAP_TTL, ENCAP_PROTO, 265 0, /* checksum */ 266 }; 267 268 /* 269 * Private variables. 270 */ 271 static vifi_t numvifs = 0; 272 static int have_encap_tunnel = 0; 273 274 /* 275 * one-back cache used by ipip_input to locate a tunnel's vif 276 * given a datagram's src ip address. 277 */ 278 static u_long last_encap_src; 279 static struct vif *last_encap_vif; 280 281 static u_long X_ip_mcast_src __P((int vifi)); 282 static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo)); 283 static int X_ip_mrouter_done __P((void)); 284 static int X_ip_mrouter_get __P((int cmd, struct socket *so, struct mbuf **m)); 285 static int X_ip_mrouter_set __P((int cmd, struct socket *so, struct mbuf *m)); 286 static int X_legal_vif_num __P((int vif)); 287 static int X_mrt_ioctl __P((int cmd, caddr_t data)); 288 289 static int get_sg_cnt(struct sioc_sg_req *); 290 static int get_vif_cnt(struct sioc_vif_req *); 291 static int ip_mrouter_init(struct socket *, struct mbuf *); 292 static int add_vif(struct vifctl *); 293 static int del_vif(vifi_t *); 294 static int add_mfc(struct mfcctl *); 295 static int del_mfc(struct mfcctl *); 296 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 297 static int get_version(struct mbuf *); 298 static int get_assert(struct mbuf *); 299 static int set_assert(int *); 300 static void expire_upcalls(void *); 301 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, 302 vifi_t); 303 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 304 static void encap_send(struct ip *, struct vif *, struct mbuf *); 305 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 306 static void tbf_queue(struct vif *, struct mbuf *); 307 static void tbf_process_q(struct vif *); 308 static void tbf_reprocess_q(void *); 309 static int tbf_dq_sel(struct vif *, struct ip *); 310 static void tbf_send_packet(struct vif *, struct mbuf *); 311 static void tbf_update_tokens(struct vif *); 312 static int priority(struct vif *, struct ip *); 313 void multiencap_decap(struct mbuf *); 314 315 /* 316 * whether or not special PIM assert processing is enabled. 317 */ 318 static int pim_assert; 319 /* 320 * Rate limit for assert notification messages, in usec 321 */ 322 #define ASSERT_MSG_TIME 3000000 323 324 /* 325 * Hash function for a source, group entry 326 */ 327 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 328 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 329 330 /* 331 * Find a route for a given origin IP address and Multicast group address 332 * Type of service parameter to be added in the future!!! 333 */ 334 335 #define MFCFIND(o, g, rt) { \ 336 register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \ 337 register struct mfc *_rt = NULL; \ 338 rt = NULL; \ 339 ++mrtstat.mrts_mfc_lookups; \ 340 while (_mb_rt) { \ 341 _rt = mtod(_mb_rt, struct mfc *); \ 342 if ((_rt->mfc_origin.s_addr == o) && \ 343 (_rt->mfc_mcastgrp.s_addr == g) && \ 344 (_mb_rt->m_act == NULL)) { \ 345 rt = _rt; \ 346 break; \ 347 } \ 348 _mb_rt = _mb_rt->m_next; \ 349 } \ 350 if (rt == NULL) { \ 351 ++mrtstat.mrts_mfc_misses; \ 352 } \ 353 } 354 355 356 /* 357 * Macros to compute elapsed time efficiently 358 * Borrowed from Van Jacobson's scheduling code 359 */ 360 #define TV_DELTA(a, b, delta) { \ 361 register int xxs; \ 362 \ 363 delta = (a).tv_usec - (b).tv_usec; \ 364 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 365 switch (xxs) { \ 366 case 2: \ 367 delta += 1000000; \ 368 /* fall through */ \ 369 case 1: \ 370 delta += 1000000; \ 371 break; \ 372 default: \ 373 delta += (1000000 * xxs); \ 374 } \ 375 } \ 376 } 377 378 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 379 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 380 381 #ifdef UPCALL_TIMING 382 u_long upcall_data[51]; 383 static void collate(struct timeval *); 384 #endif /* UPCALL_TIMING */ 385 386 387 /* 388 * Handle MRT setsockopt commands to modify the multicast routing tables. 389 */ 390 static int 391 X_ip_mrouter_set(cmd, so, m) 392 int cmd; 393 struct socket *so; 394 struct mbuf *m; 395 { 396 if (cmd != MRT_INIT && so != ip_mrouter) return EACCES; 397 398 switch (cmd) { 399 case MRT_INIT: return ip_mrouter_init(so, m); 400 case MRT_DONE: return ip_mrouter_done(); 401 case MRT_ADD_VIF: return add_vif (mtod(m, struct vifctl *)); 402 case MRT_DEL_VIF: return del_vif (mtod(m, vifi_t *)); 403 case MRT_ADD_MFC: return add_mfc (mtod(m, struct mfcctl *)); 404 case MRT_DEL_MFC: return del_mfc (mtod(m, struct mfcctl *)); 405 case MRT_ASSERT: return set_assert(mtod(m, int *)); 406 default: return EOPNOTSUPP; 407 } 408 } 409 410 #ifndef MROUTE_LKM 411 int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = X_ip_mrouter_set; 412 #endif 413 414 /* 415 * Handle MRT getsockopt commands 416 */ 417 static int 418 X_ip_mrouter_get(cmd, so, m) 419 int cmd; 420 struct socket *so; 421 struct mbuf **m; 422 { 423 struct mbuf *mb; 424 425 if (so != ip_mrouter) return EACCES; 426 427 *m = mb = m_get(M_WAIT, MT_SOOPTS); 428 429 switch (cmd) { 430 case MRT_VERSION: return get_version(mb); 431 case MRT_ASSERT: return get_assert(mb); 432 default: return EOPNOTSUPP; 433 } 434 } 435 436 #ifndef MROUTE_LKM 437 int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = X_ip_mrouter_get; 438 #endif 439 440 /* 441 * Handle ioctl commands to obtain information from the cache 442 */ 443 static int 444 X_mrt_ioctl(cmd, data) 445 int cmd; 446 caddr_t data; 447 { 448 int error = 0; 449 450 switch (cmd) { 451 case (SIOCGETVIFCNT): 452 return (get_vif_cnt((struct sioc_vif_req *)data)); 453 break; 454 case (SIOCGETSGCNT): 455 return (get_sg_cnt((struct sioc_sg_req *)data)); 456 break; 457 default: 458 return (EINVAL); 459 break; 460 } 461 return error; 462 } 463 464 #ifndef MROUTE_LKM 465 int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl; 466 #endif 467 468 /* 469 * returns the packet, byte, rpf-failure count for the source group provided 470 */ 471 static int 472 get_sg_cnt(req) 473 register struct sioc_sg_req *req; 474 { 475 register struct mfc *rt; 476 int s; 477 478 s = splnet(); 479 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 480 splx(s); 481 if (rt != NULL) { 482 req->pktcnt = rt->mfc_pkt_cnt; 483 req->bytecnt = rt->mfc_byte_cnt; 484 req->wrong_if = rt->mfc_wrong_if; 485 } else 486 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 487 488 return 0; 489 } 490 491 /* 492 * returns the input and output packet and byte counts on the vif provided 493 */ 494 static int 495 get_vif_cnt(req) 496 register struct sioc_vif_req *req; 497 { 498 register vifi_t vifi = req->vifi; 499 500 if (vifi >= numvifs) return EINVAL; 501 502 req->icount = viftable[vifi].v_pkt_in; 503 req->ocount = viftable[vifi].v_pkt_out; 504 req->ibytes = viftable[vifi].v_bytes_in; 505 req->obytes = viftable[vifi].v_bytes_out; 506 507 return 0; 508 } 509 510 /* 511 * Enable multicast routing 512 */ 513 static int 514 ip_mrouter_init(so, m) 515 struct socket *so; 516 struct mbuf *m; 517 { 518 int *v; 519 520 if (mrtdebug) 521 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 522 so->so_type, so->so_proto->pr_protocol); 523 524 if (so->so_type != SOCK_RAW || 525 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP; 526 527 if (!m || (m->m_len != sizeof(int *))) 528 return ENOPROTOOPT; 529 530 v = mtod(m, int *); 531 if (*v != 1) 532 return ENOPROTOOPT; 533 534 if (ip_mrouter != NULL) return EADDRINUSE; 535 536 ip_mrouter = so; 537 538 bzero((caddr_t)mfctable, sizeof(mfctable)); 539 bzero((caddr_t)nexpire, sizeof(nexpire)); 540 541 pim_assert = 0; 542 543 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 544 545 if (mrtdebug) 546 log(LOG_DEBUG, "ip_mrouter_init\n"); 547 548 return 0; 549 } 550 551 /* 552 * Disable multicast routing 553 */ 554 static int 555 X_ip_mrouter_done() 556 { 557 vifi_t vifi; 558 int i; 559 struct ifnet *ifp; 560 struct ifreq ifr; 561 struct mbuf *mb_rt; 562 struct mbuf *m; 563 struct rtdetq *rte; 564 int s; 565 566 s = splnet(); 567 568 /* 569 * For each phyint in use, disable promiscuous reception of all IP 570 * multicasts. 571 */ 572 for (vifi = 0; vifi < numvifs; vifi++) { 573 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 574 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 575 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 576 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr 577 = INADDR_ANY; 578 ifp = viftable[vifi].v_ifp; 579 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 580 } 581 } 582 bzero((caddr_t)tbftable, sizeof(tbftable)); 583 bzero((caddr_t)viftable, sizeof(viftable)); 584 numvifs = 0; 585 pim_assert = 0; 586 587 untimeout(expire_upcalls, (caddr_t)NULL); 588 589 /* 590 * Free all multicast forwarding cache entries. 591 */ 592 for (i = 0; i < MFCTBLSIZ; i++) { 593 mb_rt = mfctable[i]; 594 while (mb_rt) { 595 if (mb_rt->m_act != NULL) { 596 while (mb_rt->m_act) { 597 m = mb_rt->m_act; 598 mb_rt->m_act = m->m_act; 599 rte = mtod(m, struct rtdetq *); 600 m_freem(rte->m); 601 m_free(m); 602 } 603 } 604 mb_rt = m_free(mb_rt); 605 } 606 } 607 608 bzero((caddr_t)mfctable, sizeof(mfctable)); 609 610 /* 611 * Reset de-encapsulation cache 612 */ 613 last_encap_src = 0; 614 last_encap_vif = NULL; 615 have_encap_tunnel = 0; 616 617 ip_mrouter = NULL; 618 619 splx(s); 620 621 if (mrtdebug) 622 log(LOG_DEBUG, "ip_mrouter_done\n"); 623 624 return 0; 625 } 626 627 #ifndef MROUTE_LKM 628 int (*ip_mrouter_done)(void) = X_ip_mrouter_done; 629 #endif 630 631 static int 632 get_version(mb) 633 struct mbuf *mb; 634 { 635 int *v; 636 637 v = mtod(mb, int *); 638 639 *v = 0x0305; /* XXX !!!! */ 640 mb->m_len = sizeof(int); 641 642 return 0; 643 } 644 645 /* 646 * Set PIM assert processing global 647 */ 648 static int 649 set_assert(i) 650 int *i; 651 { 652 if ((*i != 1) && (*i != 0)) 653 return EINVAL; 654 655 pim_assert = *i; 656 657 return 0; 658 } 659 660 /* 661 * Get PIM assert processing global 662 */ 663 static int 664 get_assert(m) 665 struct mbuf *m; 666 { 667 int *i; 668 669 i = mtod(m, int *); 670 671 *i = pim_assert; 672 673 return 0; 674 } 675 676 /* 677 * Add a vif to the vif table 678 */ 679 static int 680 add_vif(vifcp) 681 register struct vifctl *vifcp; 682 { 683 register struct vif *vifp = viftable + vifcp->vifc_vifi; 684 static struct sockaddr_in sin = {sizeof sin, AF_INET}; 685 struct ifaddr *ifa; 686 struct ifnet *ifp; 687 struct ifreq ifr; 688 int error, s; 689 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 690 691 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL; 692 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE; 693 694 /* Find the interface with an address in AF_INET family */ 695 sin.sin_addr = vifcp->vifc_lcl_addr; 696 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 697 if (ifa == 0) return EADDRNOTAVAIL; 698 ifp = ifa->ifa_ifp; 699 700 if (vifcp->vifc_flags & VIFF_TUNNEL) { 701 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 702 /* 703 * An encapsulating tunnel is wanted. Tell ipip_input() to 704 * start paying attention to encapsulated packets. 705 */ 706 if (have_encap_tunnel == 0) { 707 have_encap_tunnel = 1; 708 for (s = 0; s < MAXVIFS; ++s) { 709 multicast_decap_if[s].if_name = "mdecap"; 710 multicast_decap_if[s].if_unit = s; 711 } 712 } 713 /* 714 * Set interface to fake encapsulator interface 715 */ 716 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 717 /* 718 * Prepare cached route entry 719 */ 720 bzero(&vifp->v_route, sizeof(vifp->v_route)); 721 } else { 722 log(LOG_ERR, "source routed tunnels not supported\n"); 723 return EOPNOTSUPP; 724 } 725 } else { 726 /* Make sure the interface supports multicast */ 727 if ((ifp->if_flags & IFF_MULTICAST) == 0) 728 return EOPNOTSUPP; 729 730 /* Enable promiscuous reception of all IP multicasts from the if */ 731 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 732 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 733 s = splnet(); 734 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr); 735 splx(s); 736 if (error) 737 return error; 738 } 739 740 s = splnet(); 741 /* define parameters for the tbf structure */ 742 vifp->v_tbf = v_tbf; 743 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 744 vifp->v_tbf->tbf_n_tok = 0; 745 vifp->v_tbf->tbf_q_len = 0; 746 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 747 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 748 749 vifp->v_flags = vifcp->vifc_flags; 750 vifp->v_threshold = vifcp->vifc_threshold; 751 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 752 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 753 vifp->v_ifp = ifp; 754 /* scaling up here allows division by 1024 in critical code */ 755 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 756 vifp->v_rsvp_on = 0; 757 vifp->v_rsvpd = NULL; 758 /* initialize per vif pkt counters */ 759 vifp->v_pkt_in = 0; 760 vifp->v_pkt_out = 0; 761 vifp->v_bytes_in = 0; 762 vifp->v_bytes_out = 0; 763 splx(s); 764 765 /* Adjust numvifs up if the vifi is higher than numvifs */ 766 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 767 768 if (mrtdebug) 769 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 770 vifcp->vifc_vifi, 771 ntohl(vifcp->vifc_lcl_addr.s_addr), 772 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 773 ntohl(vifcp->vifc_rmt_addr.s_addr), 774 vifcp->vifc_threshold, 775 vifcp->vifc_rate_limit); 776 777 return 0; 778 } 779 780 /* 781 * Delete a vif from the vif table 782 */ 783 static int 784 del_vif(vifip) 785 vifi_t *vifip; 786 { 787 register struct vif *vifp = viftable + *vifip; 788 register vifi_t vifi; 789 register struct mbuf *m; 790 struct ifnet *ifp; 791 struct ifreq ifr; 792 int s; 793 794 if (*vifip >= numvifs) return EINVAL; 795 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL; 796 797 s = splnet(); 798 799 if (!(vifp->v_flags & VIFF_TUNNEL)) { 800 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 801 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 802 ifp = vifp->v_ifp; 803 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 804 } 805 806 if (vifp == last_encap_vif) { 807 last_encap_vif = 0; 808 last_encap_src = 0; 809 } 810 811 /* 812 * Free packets queued at the interface 813 */ 814 while (vifp->v_tbf->tbf_q) { 815 m = vifp->v_tbf->tbf_q; 816 vifp->v_tbf->tbf_q = m->m_act; 817 m_freem(m); 818 } 819 820 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 821 bzero((caddr_t)vifp, sizeof (*vifp)); 822 823 /* Adjust numvifs down */ 824 for (vifi = numvifs; vifi > 0; vifi--) 825 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break; 826 numvifs = vifi; 827 828 splx(s); 829 830 if (mrtdebug) 831 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 832 833 return 0; 834 } 835 836 /* 837 * Add an mfc entry 838 */ 839 static int 840 add_mfc(mfccp) 841 struct mfcctl *mfccp; 842 { 843 struct mfc *rt; 844 register struct mbuf *mb_rt; 845 u_long hash; 846 struct mbuf *mb_ntry; 847 struct rtdetq *rte; 848 register u_short nstl; 849 int s; 850 int i; 851 852 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 853 854 /* If an entry already exists, just update the fields */ 855 if (rt) { 856 if (mrtdebug & DEBUG_MFC) 857 log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n", 858 ntohl(mfccp->mfcc_origin.s_addr), 859 ntohl(mfccp->mfcc_mcastgrp.s_addr), 860 mfccp->mfcc_parent); 861 862 s = splnet(); 863 rt->mfc_parent = mfccp->mfcc_parent; 864 for (i = 0; i < numvifs; i++) 865 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 866 splx(s); 867 return 0; 868 } 869 870 /* 871 * Find the entry for which the upcall was made and update 872 */ 873 s = splnet(); 874 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 875 for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) { 876 877 rt = mtod(mb_rt, struct mfc *); 878 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 879 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 880 (mb_rt->m_act != NULL)) { 881 882 if (nstl++) 883 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %x\n", 884 "multiple kernel entries", 885 ntohl(mfccp->mfcc_origin.s_addr), 886 ntohl(mfccp->mfcc_mcastgrp.s_addr), 887 mfccp->mfcc_parent, mb_rt->m_act); 888 889 if (mrtdebug & DEBUG_MFC) 890 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %x\n", 891 ntohl(mfccp->mfcc_origin.s_addr), 892 ntohl(mfccp->mfcc_mcastgrp.s_addr), 893 mfccp->mfcc_parent, mb_rt->m_act); 894 895 rt->mfc_origin = mfccp->mfcc_origin; 896 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 897 rt->mfc_parent = mfccp->mfcc_parent; 898 for (i = 0; i < numvifs; i++) 899 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 900 /* initialize pkt counters per src-grp */ 901 rt->mfc_pkt_cnt = 0; 902 rt->mfc_byte_cnt = 0; 903 rt->mfc_wrong_if = 0; 904 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 905 906 rt->mfc_expire = 0; /* Don't clean this guy up */ 907 nexpire[hash]--; 908 909 /* free packets Qed at the end of this entry */ 910 while (mb_rt->m_act) { 911 mb_ntry = mb_rt->m_act; 912 rte = mtod(mb_ntry, struct rtdetq *); 913 /* #ifdef RSVP_ISI */ 914 ip_mdq(rte->m, rte->ifp, rt, -1); 915 /* #endif */ 916 mb_rt->m_act = mb_ntry->m_act; 917 m_freem(rte->m); 918 #ifdef UPCALL_TIMING 919 collate(&(rte->t)); 920 #endif /* UPCALL_TIMING */ 921 m_free(mb_ntry); 922 } 923 } 924 } 925 926 /* 927 * It is possible that an entry is being inserted without an upcall 928 */ 929 if (nstl == 0) { 930 if (mrtdebug & DEBUG_MFC) 931 log(LOG_DEBUG,"add_mfc no upcall h %d o %x g %x p %x\n", 932 hash, ntohl(mfccp->mfcc_origin.s_addr), 933 ntohl(mfccp->mfcc_mcastgrp.s_addr), 934 mfccp->mfcc_parent); 935 936 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 937 938 rt = mtod(mb_rt, struct mfc *); 939 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 940 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 941 942 rt->mfc_origin = mfccp->mfcc_origin; 943 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 944 rt->mfc_parent = mfccp->mfcc_parent; 945 for (i = 0; i < numvifs; i++) 946 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 947 /* initialize pkt counters per src-grp */ 948 rt->mfc_pkt_cnt = 0; 949 rt->mfc_byte_cnt = 0; 950 rt->mfc_wrong_if = 0; 951 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 952 if (rt->mfc_expire) 953 nexpire[hash]--; 954 rt->mfc_expire = 0; 955 } 956 } 957 if (mb_rt == NULL) { 958 /* no upcall, so make a new entry */ 959 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 960 if (mb_rt == NULL) { 961 splx(s); 962 return ENOBUFS; 963 } 964 965 rt = mtod(mb_rt, struct mfc *); 966 967 /* insert new entry at head of hash chain */ 968 rt->mfc_origin = mfccp->mfcc_origin; 969 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 970 rt->mfc_parent = mfccp->mfcc_parent; 971 for (i = 0; i < numvifs; i++) 972 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 973 /* initialize pkt counters per src-grp */ 974 rt->mfc_pkt_cnt = 0; 975 rt->mfc_byte_cnt = 0; 976 rt->mfc_wrong_if = 0; 977 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 978 rt->mfc_expire = 0; 979 980 /* link into table */ 981 mb_rt->m_next = mfctable[hash]; 982 mfctable[hash] = mb_rt; 983 mb_rt->m_act = NULL; 984 } 985 } 986 splx(s); 987 return 0; 988 } 989 990 #ifdef UPCALL_TIMING 991 /* 992 * collect delay statistics on the upcalls 993 */ 994 static void collate(t) 995 register struct timeval *t; 996 { 997 register u_long d; 998 register struct timeval tp; 999 register u_long delta; 1000 1001 GET_TIME(tp); 1002 1003 if (TV_LT(*t, tp)) 1004 { 1005 TV_DELTA(tp, *t, delta); 1006 1007 d = delta >> 10; 1008 if (d > 50) 1009 d = 50; 1010 1011 ++upcall_data[d]; 1012 } 1013 } 1014 #endif /* UPCALL_TIMING */ 1015 1016 /* 1017 * Delete an mfc entry 1018 */ 1019 static int 1020 del_mfc(mfccp) 1021 struct mfcctl *mfccp; 1022 { 1023 struct in_addr origin; 1024 struct in_addr mcastgrp; 1025 struct mfc *rt; 1026 struct mbuf *mb_rt; 1027 struct mbuf **nptr; 1028 u_long hash; 1029 int s; 1030 1031 origin = mfccp->mfcc_origin; 1032 mcastgrp = mfccp->mfcc_mcastgrp; 1033 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1034 1035 if (mrtdebug & DEBUG_MFC) 1036 log(LOG_DEBUG,"del_mfc orig %x mcastgrp %x\n", 1037 ntohl(origin.s_addr), ntohl(mcastgrp.s_addr)); 1038 1039 s = splnet(); 1040 1041 nptr = &mfctable[hash]; 1042 while ((mb_rt = *nptr) != NULL) { 1043 rt = mtod(mb_rt, struct mfc *); 1044 if (origin.s_addr == rt->mfc_origin.s_addr && 1045 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1046 mb_rt->m_act == NULL) 1047 break; 1048 1049 nptr = &mb_rt->m_next; 1050 } 1051 if (mb_rt == NULL) { 1052 splx(s); 1053 return EADDRNOTAVAIL; 1054 } 1055 1056 MFREE(mb_rt, *nptr); 1057 1058 splx(s); 1059 1060 return 0; 1061 } 1062 1063 /* 1064 * Send a message to mrouted on the multicast routing socket 1065 */ 1066 static int 1067 socket_send(s, mm, src) 1068 struct socket *s; 1069 struct mbuf *mm; 1070 struct sockaddr_in *src; 1071 { 1072 if (s) { 1073 if (sbappendaddr(&s->so_rcv, 1074 (struct sockaddr *)src, 1075 mm, (struct mbuf *)0) != 0) { 1076 sorwakeup(s); 1077 return 0; 1078 } 1079 } 1080 m_freem(mm); 1081 return -1; 1082 } 1083 1084 /* 1085 * IP multicast forwarding function. This function assumes that the packet 1086 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1087 * pointed to by "ifp", and the packet is to be relayed to other networks 1088 * that have members of the packet's destination IP multicast group. 1089 * 1090 * The packet is returned unscathed to the caller, unless it is 1091 * erroneous, in which case a non-zero return value tells the caller to 1092 * discard it. 1093 */ 1094 1095 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1096 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1097 1098 static int 1099 X_ip_mforward(ip, ifp, m, imo) 1100 register struct ip *ip; 1101 struct ifnet *ifp; 1102 struct mbuf *m; 1103 struct ip_moptions *imo; 1104 { 1105 register struct mfc *rt; 1106 register u_char *ipoptions; 1107 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1108 static int srctun = 0; 1109 register struct mbuf *mm; 1110 int s; 1111 vifi_t vifi; 1112 struct vif *vifp; 1113 1114 if (mrtdebug & DEBUG_FORWARD) 1115 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %x\n", 1116 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1117 1118 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1119 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1120 /* 1121 * Packet arrived via a physical interface or 1122 * an encapsulated tunnel. 1123 */ 1124 } else { 1125 /* 1126 * Packet arrived through a source-route tunnel. 1127 * Source-route tunnels are no longer supported. 1128 */ 1129 if ((srctun++ % 1000) == 0) 1130 log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n", 1131 ntohl(ip->ip_src.s_addr)); 1132 1133 return 1; 1134 } 1135 1136 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1137 if (ip->ip_ttl < 255) 1138 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1139 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1140 vifp = viftable + vifi; 1141 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1142 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi, 1143 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1144 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1145 } 1146 return (ip_mdq(m, ifp, NULL, vifi)); 1147 } 1148 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1149 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1150 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr)); 1151 if(!imo) 1152 printf("In fact, no options were specified at all\n"); 1153 } 1154 1155 /* 1156 * Don't forward a packet with time-to-live of zero or one, 1157 * or a packet destined to a local-only group. 1158 */ 1159 if (ip->ip_ttl <= 1 || 1160 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1161 return 0; 1162 1163 /* 1164 * Determine forwarding vifs from the forwarding cache table 1165 */ 1166 s = splnet(); 1167 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt); 1168 1169 /* Entry exists, so forward if necessary */ 1170 if (rt != NULL) { 1171 splx(s); 1172 return (ip_mdq(m, ifp, rt, -1)); 1173 } else { 1174 /* 1175 * If we don't have a route for packet's origin, 1176 * Make a copy of the packet & 1177 * send message to routing daemon 1178 */ 1179 1180 register struct mbuf *mb_rt; 1181 register struct mbuf *mb_ntry; 1182 register struct mbuf *mb0; 1183 register struct rtdetq *rte; 1184 register struct mbuf *rte_m; 1185 register u_long hash; 1186 register int npkts; 1187 int hlen = ip->ip_hl << 2; 1188 #ifdef UPCALL_TIMING 1189 struct timeval tp; 1190 1191 GET_TIME(tp); 1192 #endif 1193 1194 mrtstat.mrts_no_route++; 1195 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1196 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1197 ntohl(ip->ip_src.s_addr), 1198 ntohl(ip->ip_dst.s_addr)); 1199 1200 /* 1201 * Allocate mbufs early so that we don't do extra work if we are 1202 * just going to fail anyway. Make sure to pullup the header so 1203 * that other people can't step on it. 1204 */ 1205 MGET(mb_ntry, M_DONTWAIT, MT_DATA); 1206 if (mb_ntry == NULL) { 1207 splx(s); 1208 return ENOBUFS; 1209 } 1210 mb0 = m_copy(m, 0, M_COPYALL); 1211 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1212 mb0 = m_pullup(mb0, hlen); 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 i; 1231 struct igmpmsg *im; 1232 1233 /* no upcall, so make a new entry */ 1234 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 1235 if (mb_rt == NULL) { 1236 m_free(mb_ntry); 1237 m_freem(mb0); 1238 splx(s); 1239 return ENOBUFS; 1240 } 1241 /* Make a copy of the header to send to the user level process */ 1242 mm = m_copy(mb0, 0, hlen); 1243 if (mm == NULL) { 1244 m_free(mb_ntry); 1245 m_freem(mb0); 1246 m_free(mb_rt); 1247 splx(s); 1248 return ENOBUFS; 1249 } 1250 1251 /* 1252 * Send message to routing daemon to install 1253 * a route into the kernel table 1254 */ 1255 k_igmpsrc.sin_addr = ip->ip_src; 1256 1257 im = mtod(mm, struct igmpmsg *); 1258 im->im_msgtype = IGMPMSG_NOCACHE; 1259 im->im_mbz = 0; 1260 1261 mrtstat.mrts_upcalls++; 1262 1263 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1264 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1265 ++mrtstat.mrts_upq_sockfull; 1266 m_free(mb_ntry); 1267 m_freem(mb0); 1268 m_free(mb_rt); 1269 splx(s); 1270 return ENOBUFS; 1271 } 1272 1273 rt = mtod(mb_rt, struct mfc *); 1274 1275 /* insert new entry at head of hash chain */ 1276 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1277 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1278 rt->mfc_expire = UPCALL_EXPIRE; 1279 nexpire[hash]++; 1280 for (i = 0; i < numvifs; i++) 1281 rt->mfc_ttls[i] = 0; 1282 rt->mfc_parent = -1; 1283 1284 /* link into table */ 1285 mb_rt->m_next = mfctable[hash]; 1286 mfctable[hash] = mb_rt; 1287 mb_rt->m_act = NULL; 1288 1289 rte_m = mb_rt; 1290 } else { 1291 /* determine if q has overflowed */ 1292 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act) 1293 npkts++; 1294 1295 if (npkts > MAX_UPQ) { 1296 mrtstat.mrts_upq_ovflw++; 1297 m_free(mb_ntry); 1298 m_freem(mb0); 1299 splx(s); 1300 return 0; 1301 } 1302 } 1303 1304 mb_ntry->m_act = NULL; 1305 rte = mtod(mb_ntry, struct rtdetq *); 1306 1307 rte->m = mb0; 1308 rte->ifp = ifp; 1309 #ifdef UPCALL_TIMING 1310 rte->t = tp; 1311 #endif 1312 1313 /* Add this entry to the end of the queue */ 1314 rte_m->m_act = mb_ntry; 1315 1316 splx(s); 1317 1318 return 0; 1319 } 1320 } 1321 1322 #ifndef MROUTE_LKM 1323 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 1324 struct ip_moptions *) = X_ip_mforward; 1325 #endif 1326 1327 /* 1328 * Clean up the cache entry if upcall is not serviced 1329 */ 1330 static void 1331 expire_upcalls(void *unused) 1332 { 1333 struct mbuf *mb_rt, *m, **nptr; 1334 struct rtdetq *rte; 1335 struct mfc *mfc; 1336 int i; 1337 int s; 1338 1339 s = splnet(); 1340 for (i = 0; i < MFCTBLSIZ; i++) { 1341 if (nexpire[i] == 0) 1342 continue; 1343 nptr = &mfctable[i]; 1344 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) { 1345 mfc = mtod(mb_rt, struct mfc *); 1346 1347 /* 1348 * Skip real cache entries 1349 * Make sure it wasn't marked to not expire (shouldn't happen) 1350 * If it expires now 1351 */ 1352 if (mb_rt->m_act != NULL && 1353 mfc->mfc_expire != 0 && 1354 --mfc->mfc_expire == 0) { 1355 if (mrtdebug & DEBUG_EXPIRE) 1356 log(LOG_DEBUG, "expire_upcalls: expiring (%x %x)\n", 1357 ntohl(mfc->mfc_origin.s_addr), 1358 ntohl(mfc->mfc_mcastgrp.s_addr)); 1359 /* 1360 * drop all the packets 1361 * free the mbuf with the pkt, if, timing info 1362 */ 1363 while (mb_rt->m_act) { 1364 m = mb_rt->m_act; 1365 mb_rt->m_act = m->m_act; 1366 1367 rte = mtod(m, struct rtdetq *); 1368 m_freem(rte->m); 1369 m_free(m); 1370 } 1371 ++mrtstat.mrts_cache_cleanups; 1372 nexpire[i]--; 1373 1374 MFREE(mb_rt, *nptr); 1375 } else { 1376 nptr = &mb_rt->m_next; 1377 } 1378 } 1379 } 1380 splx(s); 1381 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 1382 } 1383 1384 /* 1385 * Packet forwarding routine once entry in the cache is made 1386 */ 1387 static int 1388 ip_mdq(m, ifp, rt, xmt_vif) 1389 register struct mbuf *m; 1390 register struct ifnet *ifp; 1391 register struct mfc *rt; 1392 register vifi_t xmt_vif; 1393 { 1394 register struct ip *ip = mtod(m, struct ip *); 1395 register vifi_t vifi; 1396 register struct vif *vifp; 1397 register int plen = ntohs(ip->ip_len); 1398 1399 /* 1400 * Macro to send packet on vif. Since RSVP packets don't get counted on 1401 * input, they shouldn't get counted on output, so statistics keeping is 1402 * seperate. 1403 */ 1404 #define MC_SEND(ip,vifp,m) { \ 1405 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1406 encap_send((ip), (vifp), (m)); \ 1407 else \ 1408 phyint_send((ip), (vifp), (m)); \ 1409 } 1410 1411 /* 1412 * If xmt_vif is not -1, send on only the requested vif. 1413 * 1414 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1415 */ 1416 if (xmt_vif < numvifs) { 1417 MC_SEND(ip, viftable + xmt_vif, m); 1418 return 1; 1419 } 1420 1421 /* 1422 * Don't forward if it didn't arrive from the parent vif for its origin. 1423 */ 1424 vifi = rt->mfc_parent; 1425 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1426 /* came in the wrong interface */ 1427 if (mrtdebug & DEBUG_FORWARD) 1428 log(LOG_DEBUG, "wrong if: ifp %x vifi %d vififp %x\n", 1429 ifp, vifi, viftable[vifi].v_ifp); 1430 ++mrtstat.mrts_wrong_if; 1431 ++rt->mfc_wrong_if; 1432 /* 1433 * If we are doing PIM assert processing, and we are forwarding 1434 * packets on this interface, and it is a broadcast medium 1435 * interface (and not a tunnel), send a message to the routing daemon. 1436 */ 1437 if (pim_assert && rt->mfc_ttls[vifi] && 1438 (ifp->if_flags & IFF_BROADCAST) && 1439 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1440 struct sockaddr_in k_igmpsrc; 1441 struct mbuf *mm; 1442 struct igmpmsg *im; 1443 int hlen = ip->ip_hl << 2; 1444 struct timeval now; 1445 register u_long delta; 1446 1447 GET_TIME(now); 1448 1449 TV_DELTA(rt->mfc_last_assert, now, delta); 1450 1451 if (delta > ASSERT_MSG_TIME) { 1452 mm = m_copy(m, 0, hlen); 1453 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1454 mm = m_pullup(mm, hlen); 1455 if (mm == NULL) { 1456 return ENOBUFS; 1457 } 1458 1459 rt->mfc_last_assert = now; 1460 1461 im = mtod(mm, struct igmpmsg *); 1462 im->im_msgtype = IGMPMSG_WRONGVIF; 1463 im->im_mbz = 0; 1464 im->im_vif = vifi; 1465 1466 k_igmpsrc.sin_addr = im->im_src; 1467 1468 socket_send(ip_mrouter, mm, &k_igmpsrc); 1469 } 1470 } 1471 return 0; 1472 } 1473 1474 /* If I sourced this packet, it counts as output, else it was input. */ 1475 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1476 viftable[vifi].v_pkt_out++; 1477 viftable[vifi].v_bytes_out += plen; 1478 } else { 1479 viftable[vifi].v_pkt_in++; 1480 viftable[vifi].v_bytes_in += plen; 1481 } 1482 rt->mfc_pkt_cnt++; 1483 rt->mfc_byte_cnt += plen; 1484 1485 /* 1486 * For each vif, decide if a copy of the packet should be forwarded. 1487 * Forward if: 1488 * - the ttl exceeds the vif's threshold 1489 * - there are group members downstream on interface 1490 */ 1491 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1492 if ((rt->mfc_ttls[vifi] > 0) && 1493 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1494 vifp->v_pkt_out++; 1495 vifp->v_bytes_out += plen; 1496 MC_SEND(ip, vifp, m); 1497 } 1498 1499 return 0; 1500 } 1501 1502 /* 1503 * check if a vif number is legal/ok. This is used by ip_output, to export 1504 * numvifs there, 1505 */ 1506 static int 1507 X_legal_vif_num(vif) 1508 int vif; 1509 { 1510 if (vif >= 0 && vif < numvifs) 1511 return(1); 1512 else 1513 return(0); 1514 } 1515 1516 #ifndef MROUTE_LKM 1517 int (*legal_vif_num)(int) = X_legal_vif_num; 1518 #endif 1519 1520 /* 1521 * Return the local address used by this vif 1522 */ 1523 static u_long 1524 X_ip_mcast_src(vifi) 1525 int vifi; 1526 { 1527 if (vifi >= 0 && vifi < numvifs) 1528 return viftable[vifi].v_lcl_addr.s_addr; 1529 else 1530 return INADDR_ANY; 1531 } 1532 1533 #ifndef MROUTE_LKM 1534 u_long (*ip_mcast_src)(int) = X_ip_mcast_src; 1535 #endif 1536 1537 static void 1538 phyint_send(ip, vifp, m) 1539 struct ip *ip; 1540 struct vif *vifp; 1541 struct mbuf *m; 1542 { 1543 register struct mbuf *mb_copy; 1544 register int hlen = ip->ip_hl << 2; 1545 1546 /* 1547 * Make a new reference to the packet; make sure that 1548 * the IP header is actually copied, not just referenced, 1549 * so that ip_output() only scribbles on the copy. 1550 */ 1551 mb_copy = m_copy(m, 0, M_COPYALL); 1552 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1553 mb_copy = m_pullup(mb_copy, hlen); 1554 if (mb_copy == NULL) 1555 return; 1556 1557 if (vifp->v_rate_limit <= 0) 1558 tbf_send_packet(vifp, mb_copy); 1559 else 1560 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1561 } 1562 1563 static void 1564 encap_send(ip, vifp, m) 1565 register struct ip *ip; 1566 register struct vif *vifp; 1567 register struct mbuf *m; 1568 { 1569 register struct mbuf *mb_copy; 1570 register struct ip *ip_copy; 1571 register int i, len = ip->ip_len; 1572 1573 /* 1574 * copy the old packet & pullup it's IP header into the 1575 * new mbuf so we can modify it. Try to fill the new 1576 * mbuf since if we don't the ethernet driver will. 1577 */ 1578 MGET(mb_copy, M_DONTWAIT, MT_DATA); 1579 if (mb_copy == NULL) 1580 return; 1581 mb_copy->m_data += 16; 1582 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1583 1584 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1585 m_freem(mb_copy); 1586 return; 1587 } 1588 i = MHLEN - M_LEADINGSPACE(mb_copy); 1589 if (i > len) 1590 i = len; 1591 mb_copy = m_pullup(mb_copy, i); 1592 if (mb_copy == NULL) 1593 return; 1594 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1595 1596 /* 1597 * fill in the encapsulating IP header. 1598 */ 1599 ip_copy = mtod(mb_copy, struct ip *); 1600 *ip_copy = multicast_encap_iphdr; 1601 ip_copy->ip_id = htons(ip_id++); 1602 ip_copy->ip_len += len; 1603 ip_copy->ip_src = vifp->v_lcl_addr; 1604 ip_copy->ip_dst = vifp->v_rmt_addr; 1605 1606 /* 1607 * turn the encapsulated IP header back into a valid one. 1608 */ 1609 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1610 --ip->ip_ttl; 1611 HTONS(ip->ip_len); 1612 HTONS(ip->ip_off); 1613 ip->ip_sum = 0; 1614 #if defined(LBL) && !defined(ultrix) 1615 ip->ip_sum = ~oc_cksum((caddr_t)ip, ip->ip_hl << 2, 0); 1616 #else 1617 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1618 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1619 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1620 #endif 1621 1622 if (vifp->v_rate_limit <= 0) 1623 tbf_send_packet(vifp, mb_copy); 1624 else 1625 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1626 } 1627 1628 /* 1629 * De-encapsulate a packet and feed it back through ip input (this 1630 * routine is called whenever IP gets a packet with proto type 1631 * ENCAP_PROTO and a local destination address). 1632 */ 1633 void 1634 #ifdef MROUTE_LKM 1635 X_ipip_input(m, iphlen) 1636 #else 1637 ipip_input(m, iphlen) 1638 #endif 1639 register struct mbuf *m; 1640 int iphlen; 1641 { 1642 struct ifnet *ifp = m->m_pkthdr.rcvif; 1643 register struct ip *ip = mtod(m, struct ip *); 1644 register int hlen = ip->ip_hl << 2; 1645 register int s; 1646 register struct ifqueue *ifq; 1647 register struct vif *vifp; 1648 1649 if (!have_encap_tunnel) { 1650 rip_input(m, iphlen); 1651 return; 1652 } 1653 /* 1654 * dump the packet if it's not to a multicast destination or if 1655 * we don't have an encapsulating tunnel with the source. 1656 * Note: This code assumes that the remote site IP address 1657 * uniquely identifies the tunnel (i.e., that this site has 1658 * at most one tunnel with the remote site). 1659 */ 1660 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) { 1661 ++mrtstat.mrts_bad_tunnel; 1662 m_freem(m); 1663 return; 1664 } 1665 if (ip->ip_src.s_addr != last_encap_src) { 1666 register struct vif *vife; 1667 1668 vifp = viftable; 1669 vife = vifp + numvifs; 1670 last_encap_src = ip->ip_src.s_addr; 1671 last_encap_vif = 0; 1672 for ( ; vifp < vife; ++vifp) 1673 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1674 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1675 == VIFF_TUNNEL) 1676 last_encap_vif = vifp; 1677 break; 1678 } 1679 } 1680 if ((vifp = last_encap_vif) == 0) { 1681 last_encap_src = 0; 1682 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1683 m_freem(m); 1684 if (mrtdebug) 1685 log(LOG_DEBUG, "ip_mforward: no tunnel with %x\n", 1686 ntohl(ip->ip_src.s_addr)); 1687 return; 1688 } 1689 ifp = vifp->v_ifp; 1690 1691 if (hlen > IP_HDR_LEN) 1692 ip_stripoptions(m, (struct mbuf *) 0); 1693 m->m_data += IP_HDR_LEN; 1694 m->m_len -= IP_HDR_LEN; 1695 m->m_pkthdr.len -= IP_HDR_LEN; 1696 m->m_pkthdr.rcvif = ifp; 1697 1698 ifq = &ipintrq; 1699 s = splimp(); 1700 if (IF_QFULL(ifq)) { 1701 IF_DROP(ifq); 1702 m_freem(m); 1703 } else { 1704 IF_ENQUEUE(ifq, m); 1705 /* 1706 * normally we would need a "schednetisr(NETISR_IP)" 1707 * here but we were called by ip_input and it is going 1708 * to loop back & try to dequeue the packet we just 1709 * queued as soon as we return so we avoid the 1710 * unnecessary software interrrupt. 1711 */ 1712 } 1713 splx(s); 1714 } 1715 1716 /* 1717 * Token bucket filter module 1718 */ 1719 1720 static void 1721 tbf_control(vifp, m, ip, p_len) 1722 register struct vif *vifp; 1723 register struct mbuf *m; 1724 register struct ip *ip; 1725 register u_long p_len; 1726 { 1727 register struct tbf *t = vifp->v_tbf; 1728 1729 if (p_len > MAX_BKT_SIZE) { 1730 /* drop if packet is too large */ 1731 mrtstat.mrts_pkt2large++; 1732 m_freem(m); 1733 return; 1734 } 1735 1736 tbf_update_tokens(vifp); 1737 1738 /* if there are enough tokens, 1739 * and the queue is empty, 1740 * send this packet out 1741 */ 1742 1743 if (t->tbf_q_len == 0) { 1744 /* queue empty, send packet if enough tokens */ 1745 if (p_len <= t->tbf_n_tok) { 1746 t->tbf_n_tok -= p_len; 1747 tbf_send_packet(vifp, m); 1748 } else { 1749 /* queue packet and timeout till later */ 1750 tbf_queue(vifp, m); 1751 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1752 } 1753 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1754 /* finite queue length, so queue pkts and process queue */ 1755 tbf_queue(vifp, m); 1756 tbf_process_q(vifp); 1757 } else { 1758 /* queue length too much, try to dq and queue and process */ 1759 if (!tbf_dq_sel(vifp, ip)) { 1760 mrtstat.mrts_q_overflow++; 1761 m_freem(m); 1762 return; 1763 } else { 1764 tbf_queue(vifp, m); 1765 tbf_process_q(vifp); 1766 } 1767 } 1768 return; 1769 } 1770 1771 /* 1772 * adds a packet to the queue at the interface 1773 */ 1774 static void 1775 tbf_queue(vifp, m) 1776 register struct vif *vifp; 1777 register struct mbuf *m; 1778 { 1779 register int s = splnet(); 1780 register struct tbf *t = vifp->v_tbf; 1781 1782 if (t->tbf_t == NULL) { 1783 /* Queue was empty */ 1784 t->tbf_q = m; 1785 } else { 1786 /* Insert at tail */ 1787 t->tbf_t->m_act = m; 1788 } 1789 1790 /* Set new tail pointer */ 1791 t->tbf_t = m; 1792 1793 #ifdef DIAGNOSTIC 1794 /* Make sure we didn't get fed a bogus mbuf */ 1795 if (m->m_act) 1796 panic("tbf_queue: m_act"); 1797 #endif 1798 m->m_act = NULL; 1799 1800 t->tbf_q_len++; 1801 1802 splx(s); 1803 } 1804 1805 1806 /* 1807 * processes the queue at the interface 1808 */ 1809 static void 1810 tbf_process_q(vifp) 1811 register struct vif *vifp; 1812 { 1813 register struct mbuf *m; 1814 register int len; 1815 register int s = splnet(); 1816 register struct tbf *t = vifp->v_tbf; 1817 1818 /* loop through the queue at the interface and send as many packets 1819 * as possible 1820 */ 1821 while (t->tbf_q_len > 0) { 1822 m = t->tbf_q; 1823 1824 len = mtod(m, struct ip *)->ip_len; 1825 1826 /* determine if the packet can be sent */ 1827 if (len <= t->tbf_n_tok) { 1828 /* if so, 1829 * reduce no of tokens, dequeue the packet, 1830 * send the packet. 1831 */ 1832 t->tbf_n_tok -= len; 1833 1834 t->tbf_q = m->m_act; 1835 if (--t->tbf_q_len == 0) 1836 t->tbf_t = NULL; 1837 1838 m->m_act = NULL; 1839 tbf_send_packet(vifp, m); 1840 1841 } else break; 1842 } 1843 splx(s); 1844 } 1845 1846 static void 1847 tbf_reprocess_q(xvifp) 1848 void *xvifp; 1849 { 1850 register struct vif *vifp = xvifp; 1851 if (ip_mrouter == NULL) 1852 return; 1853 1854 tbf_update_tokens(vifp); 1855 1856 tbf_process_q(vifp); 1857 1858 if (vifp->v_tbf->tbf_q_len) 1859 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1860 } 1861 1862 /* function that will selectively discard a member of the queue 1863 * based on the precedence value and the priority 1864 */ 1865 static int 1866 tbf_dq_sel(vifp, ip) 1867 register struct vif *vifp; 1868 register struct ip *ip; 1869 { 1870 register int s = splnet(); 1871 register u_int p; 1872 register struct mbuf *m, *last; 1873 register struct mbuf **np; 1874 register struct tbf *t = vifp->v_tbf; 1875 1876 p = priority(vifp, ip); 1877 1878 np = &t->tbf_q; 1879 last = NULL; 1880 while ((m = *np) != NULL) { 1881 if (p > priority(vifp, mtod(m, struct ip *))) { 1882 *np = m->m_act; 1883 /* If we're removing the last packet, fix the tail pointer */ 1884 if (m == t->tbf_t) 1885 t->tbf_t = last; 1886 m_freem(m); 1887 /* it's impossible for the queue to be empty, but 1888 * we check anyway. */ 1889 if (--t->tbf_q_len == 0) 1890 t->tbf_t = NULL; 1891 splx(s); 1892 mrtstat.mrts_drop_sel++; 1893 return(1); 1894 } 1895 np = &m->m_act; 1896 last = m; 1897 } 1898 splx(s); 1899 return(0); 1900 } 1901 1902 static void 1903 tbf_send_packet(vifp, m) 1904 register struct vif *vifp; 1905 register struct mbuf *m; 1906 { 1907 struct ip_moptions imo; 1908 int error; 1909 static struct route ro; 1910 int s = splnet(); 1911 1912 if (vifp->v_flags & VIFF_TUNNEL) { 1913 /* If tunnel options */ 1914 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1915 IP_FORWARDING, (struct ip_moptions *)0); 1916 } else { 1917 imo.imo_multicast_ifp = vifp->v_ifp; 1918 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1919 imo.imo_multicast_loop = 1; 1920 imo.imo_multicast_vif = -1; 1921 1922 /* 1923 * Re-entrancy should not be a problem here, because 1924 * the packets that we send out and are looped back at us 1925 * should get rejected because they appear to come from 1926 * the loopback interface, thus preventing looping. 1927 */ 1928 error = ip_output(m, (struct mbuf *)0, &ro, 1929 IP_FORWARDING, &imo); 1930 1931 if (mrtdebug & DEBUG_XMIT) 1932 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1933 vifp - viftable, error); 1934 } 1935 splx(s); 1936 } 1937 1938 /* determine the current time and then 1939 * the elapsed time (between the last time and time now) 1940 * in milliseconds & update the no. of tokens in the bucket 1941 */ 1942 static void 1943 tbf_update_tokens(vifp) 1944 register struct vif *vifp; 1945 { 1946 struct timeval tp; 1947 register u_long tm; 1948 register int s = splnet(); 1949 register struct tbf *t = vifp->v_tbf; 1950 1951 GET_TIME(tp); 1952 1953 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1954 1955 /* 1956 * This formula is actually 1957 * "time in seconds" * "bytes/second". 1958 * 1959 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1960 * 1961 * The (1000/1024) was introduced in add_vif to optimize 1962 * this divide into a shift. 1963 */ 1964 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1965 t->tbf_last_pkt_t = tp; 1966 1967 if (t->tbf_n_tok > MAX_BKT_SIZE) 1968 t->tbf_n_tok = MAX_BKT_SIZE; 1969 1970 splx(s); 1971 } 1972 1973 static int 1974 priority(vifp, ip) 1975 register struct vif *vifp; 1976 register struct ip *ip; 1977 { 1978 register int prio; 1979 1980 /* temporary hack; may add general packet classifier some day */ 1981 1982 /* 1983 * The UDP port space is divided up into four priority ranges: 1984 * [0, 16384) : unclassified - lowest priority 1985 * [16384, 32768) : audio - highest priority 1986 * [32768, 49152) : whiteboard - medium priority 1987 * [49152, 65536) : video - low priority 1988 */ 1989 if (ip->ip_p == IPPROTO_UDP) { 1990 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1991 switch (ntohs(udp->uh_dport) & 0xc000) { 1992 case 0x4000: 1993 prio = 70; 1994 break; 1995 case 0x8000: 1996 prio = 60; 1997 break; 1998 case 0xc000: 1999 prio = 55; 2000 break; 2001 default: 2002 prio = 50; 2003 break; 2004 } 2005 if (tbfdebug > 1) 2006 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio); 2007 } else { 2008 prio = 50; 2009 } 2010 return prio; 2011 } 2012 2013 /* 2014 * End of token bucket filter modifications 2015 */ 2016 2017 int 2018 ip_rsvp_vif_init(so, m) 2019 struct socket *so; 2020 struct mbuf *m; 2021 { 2022 int i; 2023 register int s; 2024 2025 if (rsvpdebug) 2026 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2027 so->so_type, so->so_proto->pr_protocol); 2028 2029 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2030 return EOPNOTSUPP; 2031 2032 /* Check mbuf. */ 2033 if (m == NULL || m->m_len != sizeof(int)) { 2034 return EINVAL; 2035 } 2036 i = *(mtod(m, int *)); 2037 2038 if (rsvpdebug) 2039 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on); 2040 2041 s = splnet(); 2042 2043 /* Check vif. */ 2044 if (!legal_vif_num(i)) { 2045 splx(s); 2046 return EADDRNOTAVAIL; 2047 } 2048 2049 /* Check if socket is available. */ 2050 if (viftable[i].v_rsvpd != NULL) { 2051 splx(s); 2052 return EADDRINUSE; 2053 } 2054 2055 viftable[i].v_rsvpd = so; 2056 /* This may seem silly, but we need to be sure we don't over-increment 2057 * the RSVP counter, in case something slips up. 2058 */ 2059 if (!viftable[i].v_rsvp_on) { 2060 viftable[i].v_rsvp_on = 1; 2061 rsvp_on++; 2062 } 2063 2064 splx(s); 2065 return 0; 2066 } 2067 2068 int 2069 ip_rsvp_vif_done(so, m) 2070 struct socket *so; 2071 struct mbuf *m; 2072 { 2073 int i; 2074 register int s; 2075 2076 if (rsvpdebug) 2077 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2078 so->so_type, so->so_proto->pr_protocol); 2079 2080 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2081 return EOPNOTSUPP; 2082 2083 /* Check mbuf. */ 2084 if (m == NULL || m->m_len != sizeof(int)) { 2085 return EINVAL; 2086 } 2087 i = *(mtod(m, int *)); 2088 2089 s = splnet(); 2090 2091 /* Check vif. */ 2092 if (!legal_vif_num(i)) { 2093 splx(s); 2094 return EADDRNOTAVAIL; 2095 } 2096 2097 if (rsvpdebug) 2098 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n", 2099 viftable[i].v_rsvpd, so); 2100 2101 viftable[i].v_rsvpd = NULL; 2102 /* This may seem silly, but we need to be sure we don't over-decrement 2103 * the RSVP counter, in case something slips up. 2104 */ 2105 if (viftable[i].v_rsvp_on) { 2106 viftable[i].v_rsvp_on = 0; 2107 rsvp_on--; 2108 } 2109 2110 splx(s); 2111 return 0; 2112 } 2113 2114 void 2115 ip_rsvp_force_done(so) 2116 struct socket *so; 2117 { 2118 int vifi; 2119 register int s; 2120 2121 /* Don't bother if it is not the right type of socket. */ 2122 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2123 return; 2124 2125 s = splnet(); 2126 2127 /* The socket may be attached to more than one vif...this 2128 * is perfectly legal. 2129 */ 2130 for (vifi = 0; vifi < numvifs; vifi++) { 2131 if (viftable[vifi].v_rsvpd == so) { 2132 viftable[vifi].v_rsvpd = NULL; 2133 /* This may seem silly, but we need to be sure we don't 2134 * over-decrement the RSVP counter, in case something slips up. 2135 */ 2136 if (viftable[vifi].v_rsvp_on) { 2137 viftable[vifi].v_rsvp_on = 0; 2138 rsvp_on--; 2139 } 2140 } 2141 } 2142 2143 splx(s); 2144 return; 2145 } 2146 2147 void 2148 rsvp_input(m, iphlen) 2149 struct mbuf *m; 2150 int iphlen; 2151 { 2152 int vifi; 2153 register struct ip *ip = mtod(m, struct ip *); 2154 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2155 register int s; 2156 struct ifnet *ifp; 2157 2158 if (rsvpdebug) 2159 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2160 2161 /* Can still get packets with rsvp_on = 0 if there is a local member 2162 * of the group to which the RSVP packet is addressed. But in this 2163 * case we want to throw the packet away. 2164 */ 2165 if (!rsvp_on) { 2166 m_freem(m); 2167 return; 2168 } 2169 2170 /* If the old-style non-vif-associated socket is set, then use 2171 * it and ignore the new ones. 2172 */ 2173 if (ip_rsvpd != NULL) { 2174 if (rsvpdebug) 2175 printf("rsvp_input: Sending packet up old-style socket\n"); 2176 rip_input(m, iphlen); 2177 return; 2178 } 2179 2180 s = splnet(); 2181 2182 if (rsvpdebug) 2183 printf("rsvp_input: check vifs\n"); 2184 2185 #ifdef DIAGNOSTIC 2186 if (!(m->m_flags & M_PKTHDR)) 2187 panic("rsvp_input no hdr"); 2188 #endif 2189 2190 ifp = m->m_pkthdr.rcvif; 2191 /* Find which vif the packet arrived on. */ 2192 for (vifi = 0; vifi < numvifs; vifi++) { 2193 if (viftable[vifi].v_ifp == ifp) 2194 break; 2195 } 2196 2197 if (vifi == numvifs) { 2198 /* Can't find vif packet arrived on. Drop packet. */ 2199 if (rsvpdebug) 2200 printf("rsvp_input: Can't find vif for packet...dropping it.\n"); 2201 m_freem(m); 2202 splx(s); 2203 return; 2204 } 2205 2206 if (rsvpdebug) 2207 printf("rsvp_input: check socket\n"); 2208 2209 if (viftable[vifi].v_rsvpd == NULL) { 2210 /* drop packet, since there is no specific socket for this 2211 * interface */ 2212 if (rsvpdebug) 2213 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2214 m_freem(m); 2215 splx(s); 2216 return; 2217 } 2218 rsvp_src.sin_addr = ip->ip_src; 2219 2220 if (rsvpdebug && m) 2221 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2222 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2223 2224 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2225 if (rsvpdebug) 2226 printf("rsvp_input: Failed to append to socket\n"); 2227 else 2228 if (rsvpdebug) 2229 printf("rsvp_input: send packet up\n"); 2230 2231 splx(s); 2232 } 2233 2234 #ifdef MROUTE_LKM 2235 #include <sys/conf.h> 2236 #include <sys/exec.h> 2237 #include <sys/sysent.h> 2238 #include <sys/lkm.h> 2239 2240 MOD_MISC("ip_mroute_mod") 2241 2242 static int 2243 ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd) 2244 { 2245 int i; 2246 struct lkm_misc *args = lkmtp->private.lkm_misc; 2247 int err = 0; 2248 2249 switch(cmd) { 2250 static int (*old_ip_mrouter_cmd)(); 2251 static int (*old_ip_mrouter_done)(); 2252 static int (*old_ip_mforward)(); 2253 static int (*old_mrt_ioctl)(); 2254 static void (*old_proto4_input)(); 2255 static int (*old_legal_vif_num)(); 2256 extern struct protosw inetsw[]; 2257 2258 case LKM_E_LOAD: 2259 if(lkmexists(lkmtp) || ip_mrtproto) 2260 return(EEXIST); 2261 old_ip_mrouter_cmd = ip_mrouter_cmd; 2262 ip_mrouter_cmd = X_ip_mrouter_cmd; 2263 old_ip_mrouter_done = ip_mrouter_done; 2264 ip_mrouter_done = X_ip_mrouter_done; 2265 old_ip_mforward = ip_mforward; 2266 ip_mforward = X_ip_mforward; 2267 old_mrt_ioctl = mrt_ioctl; 2268 mrt_ioctl = X_mrt_ioctl; 2269 old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input; 2270 inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input; 2271 old_legal_vif_num = legal_vif_num; 2272 legal_vif_num = X_legal_vif_num; 2273 ip_mrtproto = IGMP_DVMRP; 2274 2275 printf("\nIP multicast routing loaded\n"); 2276 break; 2277 2278 case LKM_E_UNLOAD: 2279 if (ip_mrouter) 2280 return EINVAL; 2281 2282 ip_mrouter_cmd = old_ip_mrouter_cmd; 2283 ip_mrouter_done = old_ip_mrouter_done; 2284 ip_mforward = old_ip_mforward; 2285 mrt_ioctl = old_mrt_ioctl; 2286 inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input; 2287 legal_vif_num = old_legal_vif_num; 2288 ip_mrtproto = 0; 2289 break; 2290 2291 default: 2292 err = EINVAL; 2293 break; 2294 } 2295 2296 return(err); 2297 } 2298 2299 int 2300 ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) { 2301 DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle, 2302 nosys); 2303 } 2304 2305 #endif /* MROUTE_LKM */ 2306 #endif /* MROUTING */ 2307