1 /* $OpenBSD: ip_mroute.c,v 1.4 1996/05/10 12:31:19 deraadt Exp $ */ 2 /* $NetBSD: ip_mroute.c,v 1.27 1996/05/07 02:40:50 thorpej Exp $ */ 3 4 /* 5 * IP multicast forwarding procedures 6 * 7 * Written by David Waitzman, BBN Labs, August 1988. 8 * Modified by Steve Deering, Stanford, February 1989. 9 * Modified by Mark J. Steiglitz, Stanford, May, 1991 10 * Modified by Van Jacobson, LBL, January 1993 11 * Modified by Ajit Thyagarajan, PARC, August 1993 12 * Modified by Bill Fenner, PARC, April 1994 13 * Modified by Charles M. Hannum, NetBSD, May 1995. 14 * 15 * MROUTING Revision: 1.2 16 */ 17 18 #include <sys/param.h> 19 #include <sys/systm.h> 20 #include <sys/mbuf.h> 21 #include <sys/socket.h> 22 #include <sys/socketvar.h> 23 #include <sys/protosw.h> 24 #include <sys/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 <net/raw_cb.h> 32 #include <netinet/in.h> 33 #include <netinet/in_var.h> 34 #include <netinet/in_systm.h> 35 #include <netinet/ip.h> 36 #include <netinet/ip_var.h> 37 #include <netinet/in_pcb.h> 38 #include <netinet/udp.h> 39 #include <netinet/igmp.h> 40 #include <netinet/igmp_var.h> 41 #include <netinet/ip_mroute.h> 42 43 #include <machine/stdarg.h> 44 45 #define IP_MULTICASTOPTS 0 46 #define M_PULLUP(m, len) \ 47 do { \ 48 if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \ 49 (m) = m_pullup((m), (len)); \ 50 } while (0) 51 52 /* 53 * Globals. All but ip_mrouter and ip_mrtproto could be static, 54 * except for netstat or debugging purposes. 55 */ 56 struct socket *ip_mrouter = NULL; 57 int ip_mrtproto = IGMP_DVMRP; /* for netstat only */ 58 59 #define NO_RTE_FOUND 0x1 60 #define RTE_FOUND 0x2 61 62 #define MFCHASH(a, g) \ 63 ((((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 64 ((g) >> 20) ^ ((g) >> 10) ^ (g)) & mfchash) 65 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl; 66 u_long mfchash; 67 68 u_char nexpire[MFCTBLSIZ]; 69 struct vif viftable[MAXVIFS]; 70 struct mrtstat mrtstat; 71 u_int mrtdebug = 0; /* debug level */ 72 #define DEBUG_MFC 0x02 73 #define DEBUG_FORWARD 0x04 74 #define DEBUG_EXPIRE 0x08 75 #define DEBUG_XMIT 0x10 76 u_int tbfdebug = 0; /* tbf debug level */ 77 #ifdef RSVP_ISI 78 u_int rsvpdebug = 0; /* rsvp debug level */ 79 extern struct socket *ip_rsvpd; 80 extern int rsvp_on; 81 #endif /* RSVP_ISI */ 82 83 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 84 #define UPCALL_EXPIRE 6 /* number of timeouts */ 85 86 /* 87 * Define the token bucket filter structures 88 * qtable -> each interface has an associated queue of pkts 89 */ 90 91 struct pkt_queue qtable[MAXVIFS][MAXQSIZE]; 92 93 static int get_sg_cnt __P((struct sioc_sg_req *)); 94 static int get_vif_cnt __P((struct sioc_vif_req *)); 95 static int ip_mrouter_init __P((struct socket *, struct mbuf *)); 96 static int get_version __P((struct mbuf *)); 97 static int set_assert __P((struct mbuf *)); 98 static int get_assert __P((struct mbuf *)); 99 static int add_vif __P((struct mbuf *)); 100 static int del_vif __P((struct mbuf *)); 101 static void update_mfc __P((struct mfcctl *, struct mfc *)); 102 static void expire_mfc __P((struct mfc *)); 103 static int add_mfc __P((struct mbuf *)); 104 #ifdef UPCALL_TIMING 105 static void collate __P((struct timeval *)); 106 #endif 107 static int del_mfc __P((struct mbuf *)); 108 static int socket_send __P((struct socket *, struct mbuf *, 109 struct sockaddr_in *)); 110 static void expire_upcalls __P((void *)); 111 #ifdef RSVP_ISI 112 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *, vifi_t)); 113 #else 114 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *)); 115 #endif 116 static void phyint_send __P((struct ip *, struct vif *, struct mbuf *)); 117 static void encap_send __P((struct ip *, struct vif *, struct mbuf *)); 118 static void tbf_control __P((struct vif *, struct mbuf *, struct ip *, 119 u_int32_t)); 120 static void tbf_queue __P((struct vif *, struct mbuf *, struct ip *)); 121 static void tbf_process_q __P((struct vif *)); 122 static void tbf_dequeue __P((struct vif *, int)); 123 static void tbf_reprocess_q __P((void *)); 124 static int tbf_dq_sel __P((struct vif *, struct ip *)); 125 static void tbf_send_packet __P((struct vif *, struct mbuf *)); 126 static void tbf_update_tokens __P((struct vif *)); 127 static int priority __P((struct vif *, struct ip *)); 128 129 /* 130 * 'Interfaces' associated with decapsulator (so we can tell 131 * packets that went through it from ones that get reflected 132 * by a broken gateway). These interfaces are never linked into 133 * the system ifnet list & no routes point to them. I.e., packets 134 * can't be sent this way. They only exist as a placeholder for 135 * multicast source verification. 136 */ 137 #if 0 138 struct ifnet multicast_decap_if[MAXVIFS]; 139 #endif 140 141 #define ENCAP_TTL 64 142 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 143 144 /* prototype IP hdr for encapsulated packets */ 145 struct ip multicast_encap_iphdr = { 146 #if BYTE_ORDER == LITTLE_ENDIAN 147 sizeof(struct ip) >> 2, IPVERSION, 148 #else 149 IPVERSION, sizeof(struct ip) >> 2, 150 #endif 151 0, /* tos */ 152 sizeof(struct ip), /* total length */ 153 0, /* id */ 154 0, /* frag offset */ 155 ENCAP_TTL, ENCAP_PROTO, 156 0, /* checksum */ 157 }; 158 159 /* 160 * Private variables. 161 */ 162 static vifi_t numvifs = 0; 163 static int have_encap_tunnel = 0; 164 165 /* 166 * one-back cache used by ipip_input to locate a tunnel's vif 167 * given a datagram's src ip address. 168 */ 169 static u_int32_t last_encap_src; 170 static struct vif *last_encap_vif; 171 172 /* 173 * whether or not special PIM assert processing is enabled. 174 */ 175 static int pim_assert; 176 /* 177 * Rate limit for assert notification messages, in usec 178 */ 179 #define ASSERT_MSG_TIME 3000000 180 181 /* 182 * Find a route for a given origin IP address and Multicast group address 183 * Type of service parameter to be added in the future!!! 184 */ 185 186 #define MFCFIND(o, g, rt) { \ 187 register struct mfc *_rt; \ 188 (rt) = NULL; \ 189 ++mrtstat.mrts_mfc_lookups; \ 190 for (_rt = mfchashtbl[MFCHASH(o, g)].lh_first; \ 191 _rt; _rt = _rt->mfc_hash.le_next) { \ 192 if (_rt->mfc_origin.s_addr == (o) && \ 193 _rt->mfc_mcastgrp.s_addr == (g) && \ 194 _rt->mfc_stall == NULL) { \ 195 (rt) = _rt; \ 196 break; \ 197 } \ 198 } \ 199 if ((rt) == NULL) \ 200 ++mrtstat.mrts_mfc_misses; \ 201 } 202 203 /* 204 * Macros to compute elapsed time efficiently 205 * Borrowed from Van Jacobson's scheduling code 206 */ 207 #define TV_DELTA(a, b, delta) { \ 208 register int xxs; \ 209 delta = (a).tv_usec - (b).tv_usec; \ 210 xxs = (a).tv_sec - (b).tv_sec; \ 211 switch (xxs) { \ 212 case 2: \ 213 delta += 1000000; \ 214 /* fall through */ \ 215 case 1: \ 216 delta += 1000000; \ 217 /* fall through */ \ 218 case 0: \ 219 break; \ 220 default: \ 221 delta += (1000000 * xxs); \ 222 break; \ 223 } \ 224 } 225 226 #ifdef UPCALL_TIMING 227 u_int32_t upcall_data[51]; 228 #endif /* UPCALL_TIMING */ 229 230 /* 231 * Handle MRT setsockopt commands to modify the multicast routing tables. 232 */ 233 int 234 ip_mrouter_set(cmd, so, m) 235 int cmd; 236 struct socket *so; 237 struct mbuf **m; 238 { 239 int error; 240 241 if (cmd != MRT_INIT && so != ip_mrouter) 242 error = EACCES; 243 else 244 switch (cmd) { 245 case MRT_INIT: 246 error = ip_mrouter_init(so, *m); 247 break; 248 case MRT_DONE: 249 error = ip_mrouter_done(); 250 break; 251 case MRT_ADD_VIF: 252 error = add_vif(*m); 253 break; 254 case MRT_DEL_VIF: 255 error = del_vif(*m); 256 break; 257 case MRT_ADD_MFC: 258 error = add_mfc(*m); 259 break; 260 case MRT_DEL_MFC: 261 error = del_mfc(*m); 262 break; 263 case MRT_ASSERT: 264 error = set_assert(*m); 265 break; 266 default: 267 error = EOPNOTSUPP; 268 break; 269 } 270 271 if (*m) 272 m_free(*m); 273 return (error); 274 } 275 276 /* 277 * Handle MRT getsockopt commands 278 */ 279 int 280 ip_mrouter_get(cmd, so, m) 281 int cmd; 282 struct socket *so; 283 struct mbuf **m; 284 { 285 struct mbuf *mb; 286 int error; 287 288 if (so != ip_mrouter) 289 error = EACCES; 290 else { 291 *m = mb = m_get(M_WAIT, MT_SOOPTS); 292 293 switch (cmd) { 294 case MRT_VERSION: 295 error = get_version(mb); 296 break; 297 case MRT_ASSERT: 298 error = get_assert(mb); 299 break; 300 default: 301 error = EOPNOTSUPP; 302 break; 303 } 304 305 if (error) 306 m_free(mb); 307 } 308 309 return (error); 310 } 311 312 /* 313 * Handle ioctl commands to obtain information from the cache 314 */ 315 int 316 mrt_ioctl(cmd, data) 317 u_long cmd; 318 caddr_t data; 319 { 320 int error; 321 322 switch (cmd) { 323 case SIOCGETVIFCNT: 324 error = get_vif_cnt((struct sioc_vif_req *)data); 325 break; 326 case SIOCGETSGCNT: 327 error = get_sg_cnt((struct sioc_sg_req *)data); 328 break; 329 default: 330 error = EINVAL; 331 break; 332 } 333 334 return (error); 335 } 336 337 /* 338 * returns the packet, byte, rpf-failure count for the source group provided 339 */ 340 static int 341 get_sg_cnt(req) 342 register struct sioc_sg_req *req; 343 { 344 register struct mfc *rt; 345 int s; 346 347 s = splsoftnet(); 348 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 349 splx(s); 350 if (rt != NULL) { 351 req->pktcnt = rt->mfc_pkt_cnt; 352 req->bytecnt = rt->mfc_byte_cnt; 353 req->wrong_if = rt->mfc_wrong_if; 354 } else 355 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 356 357 return (0); 358 } 359 360 /* 361 * returns the input and output packet and byte counts on the vif provided 362 */ 363 static int 364 get_vif_cnt(req) 365 register struct sioc_vif_req *req; 366 { 367 register vifi_t vifi = req->vifi; 368 369 if (vifi >= numvifs) 370 return (EINVAL); 371 372 req->icount = viftable[vifi].v_pkt_in; 373 req->ocount = viftable[vifi].v_pkt_out; 374 req->ibytes = viftable[vifi].v_bytes_in; 375 req->obytes = viftable[vifi].v_bytes_out; 376 377 return (0); 378 } 379 380 /* 381 * Enable multicast routing 382 */ 383 static int 384 ip_mrouter_init(so, m) 385 struct socket *so; 386 struct mbuf *m; 387 { 388 int *v; 389 390 if (mrtdebug) 391 log(LOG_DEBUG, 392 "ip_mrouter_init: so_type = %d, pr_protocol = %d", 393 so->so_type, so->so_proto->pr_protocol); 394 395 if (so->so_type != SOCK_RAW || 396 so->so_proto->pr_protocol != IPPROTO_IGMP) 397 return (EOPNOTSUPP); 398 399 if (m == 0 || m->m_len < sizeof(int)) 400 return (EINVAL); 401 402 v = mtod(m, int *); 403 if (*v != 1) 404 return (EINVAL); 405 406 if (ip_mrouter != NULL) 407 return (EADDRINUSE); 408 409 ip_mrouter = so; 410 411 mfchashtbl = hashinit(MFCTBLSIZ, M_MRTABLE, &mfchash); 412 bzero((caddr_t)nexpire, sizeof(nexpire)); 413 414 pim_assert = 0; 415 416 timeout(expire_upcalls, (caddr_t)0, EXPIRE_TIMEOUT); 417 418 if (mrtdebug) 419 log(LOG_DEBUG, "ip_mrouter_init"); 420 421 return (0); 422 } 423 424 /* 425 * Disable multicast routing 426 */ 427 int 428 ip_mrouter_done() 429 { 430 vifi_t vifi; 431 register struct vif *vifp; 432 int i; 433 int s; 434 435 s = splsoftnet(); 436 437 /* Clear out all the vifs currently in use. */ 438 for (vifi = 0; vifi < numvifs; vifi++) { 439 vifp = &viftable[vifi]; 440 if (vifp->v_lcl_addr.s_addr != 0) 441 reset_vif(vifp); 442 } 443 444 bzero((caddr_t)qtable, sizeof(qtable)); 445 numvifs = 0; 446 pim_assert = 0; 447 448 untimeout(expire_upcalls, (caddr_t)NULL); 449 450 /* 451 * Free all multicast forwarding cache entries. 452 */ 453 for (i = 0; i < MFCTBLSIZ; i++) { 454 register struct mfc *rt, *nrt; 455 456 for (rt = mfchashtbl[i].lh_first; rt; rt = nrt) { 457 nrt = rt->mfc_hash.le_next; 458 459 expire_mfc(rt); 460 } 461 } 462 free(mfchashtbl, M_MRTABLE); 463 464 /* Reset de-encapsulation cache. */ 465 have_encap_tunnel = 0; 466 467 ip_mrouter = NULL; 468 469 splx(s); 470 471 if (mrtdebug) 472 log(LOG_DEBUG, "ip_mrouter_done"); 473 474 return (0); 475 } 476 477 static int 478 get_version(m) 479 struct mbuf *m; 480 { 481 int *v = mtod(m, int *); 482 483 *v = 0x0305; /* XXX !!!! */ 484 m->m_len = sizeof(int); 485 return (0); 486 } 487 488 /* 489 * Set PIM assert processing global 490 */ 491 static int 492 set_assert(m) 493 struct mbuf *m; 494 { 495 int *i; 496 497 if (m == 0 || m->m_len < sizeof(int)) 498 return (EINVAL); 499 500 i = mtod(m, int *); 501 pim_assert = !!*i; 502 return (0); 503 } 504 505 /* 506 * Get PIM assert processing global 507 */ 508 static int 509 get_assert(m) 510 struct mbuf *m; 511 { 512 int *i = mtod(m, int *); 513 514 *i = pim_assert; 515 m->m_len = sizeof(int); 516 return (0); 517 } 518 519 static struct sockaddr_in sin = { sizeof(sin), AF_INET }; 520 521 /* 522 * Add a vif to the vif table 523 */ 524 static int 525 add_vif(m) 526 struct mbuf *m; 527 { 528 register struct vifctl *vifcp; 529 register struct vif *vifp; 530 struct ifaddr *ifa; 531 struct ifnet *ifp; 532 struct ifreq ifr; 533 int error, s; 534 535 if (m == 0 || m->m_len < sizeof(struct vifctl)) 536 return (EINVAL); 537 538 vifcp = mtod(m, struct vifctl *); 539 if (vifcp->vifc_vifi >= MAXVIFS) 540 return (EINVAL); 541 542 vifp = &viftable[vifcp->vifc_vifi]; 543 if (vifp->v_lcl_addr.s_addr != 0) 544 return (EADDRINUSE); 545 546 /* Find the interface with an address in AF_INET family. */ 547 sin.sin_addr = vifcp->vifc_lcl_addr; 548 ifa = ifa_ifwithaddr(sintosa(&sin)); 549 if (ifa == 0) 550 return (EADDRNOTAVAIL); 551 552 if (vifcp->vifc_flags & VIFF_TUNNEL) { 553 if (vifcp->vifc_flags & VIFF_SRCRT) { 554 log(LOG_ERR, "Source routed tunnels not supported."); 555 return (EOPNOTSUPP); 556 } 557 558 /* Create a fake encapsulation interface. */ 559 ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK); 560 bzero(ifp, sizeof(*ifp)); 561 sprintf(ifp->if_xname, "mdecap%d", vifcp->vifc_vifi); 562 563 /* Prepare cached route entry. */ 564 bzero(&vifp->v_route, sizeof(vifp->v_route)); 565 566 /* Tell ipip_input() to start looking at encapsulated packets. */ 567 have_encap_tunnel = 1; 568 } else { 569 /* Use the physical interface associated with the address. */ 570 ifp = ifa->ifa_ifp; 571 572 /* Make sure the interface supports multicast. */ 573 if ((ifp->if_flags & IFF_MULTICAST) == 0) 574 return (EOPNOTSUPP); 575 576 /* Enable promiscuous reception of all IP multicasts. */ 577 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 578 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 579 satosin(&ifr.ifr_addr)->sin_addr.s_addr = INADDR_ANY; 580 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr); 581 if (error) 582 return (error); 583 } 584 585 s = splsoftnet(); 586 /* Define parameters for the tbf structure. */ 587 vifp->v_tbf.q_len = 0; 588 vifp->v_tbf.n_tok = 0; 589 vifp->v_tbf.last_pkt_t = 0; 590 591 vifp->v_flags = vifcp->vifc_flags; 592 vifp->v_threshold = vifcp->vifc_threshold; 593 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 594 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 595 vifp->v_ifp = ifp; 596 vifp->v_rate_limit = vifcp->vifc_rate_limit; 597 #ifdef RSVP_ISI 598 vifp->v_rsvp_on = 0; 599 vifp->v_rsvpd = NULL; 600 #endif /* RSVP_ISI */ 601 /* Initialize per vif pkt counters. */ 602 vifp->v_pkt_in = 0; 603 vifp->v_pkt_out = 0; 604 vifp->v_bytes_in = 0; 605 vifp->v_bytes_out = 0; 606 splx(s); 607 608 /* Adjust numvifs up if the vifi is higher than numvifs. */ 609 if (numvifs <= vifcp->vifc_vifi) 610 numvifs = vifcp->vifc_vifi + 1; 611 612 if (mrtdebug) 613 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d", 614 vifcp->vifc_vifi, 615 ntohl(vifcp->vifc_lcl_addr.s_addr), 616 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 617 ntohl(vifcp->vifc_rmt_addr.s_addr), 618 vifcp->vifc_threshold, 619 vifcp->vifc_rate_limit); 620 621 return (0); 622 } 623 624 void 625 reset_vif(vifp) 626 register struct vif *vifp; 627 { 628 struct ifnet *ifp; 629 struct ifreq ifr; 630 631 if (vifp->v_flags & VIFF_TUNNEL) { 632 free(vifp->v_ifp, M_MRTABLE); 633 if (vifp == last_encap_vif) { 634 last_encap_vif = 0; 635 last_encap_src = 0; 636 } 637 } else { 638 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 639 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 640 satosin(&ifr.ifr_addr)->sin_addr.s_addr = INADDR_ANY; 641 ifp = vifp->v_ifp; 642 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 643 } 644 bzero((caddr_t)vifp, sizeof(*vifp)); 645 } 646 647 /* 648 * Delete a vif from the vif table 649 */ 650 static int 651 del_vif(m) 652 struct mbuf *m; 653 { 654 vifi_t *vifip; 655 register struct vif *vifp; 656 register vifi_t vifi; 657 int s; 658 659 if (m == 0 || m->m_len < sizeof(vifi_t)) 660 return (EINVAL); 661 662 vifip = mtod(m, vifi_t *); 663 if (*vifip >= numvifs) 664 return (EINVAL); 665 666 vifp = &viftable[*vifip]; 667 if (vifp->v_lcl_addr.s_addr == 0) 668 return (EADDRNOTAVAIL); 669 670 s = splsoftnet(); 671 672 reset_vif(vifp); 673 674 bzero((caddr_t)qtable[*vifip], sizeof(qtable[*vifip])); 675 676 /* Adjust numvifs down */ 677 for (vifi = numvifs; vifi > 0; vifi--) 678 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) 679 break; 680 numvifs = vifi; 681 682 splx(s); 683 684 if (mrtdebug) 685 log(LOG_DEBUG, "del_vif %d, numvifs %d", *vifip, numvifs); 686 687 return (0); 688 } 689 690 static void 691 update_mfc(mfccp, rt) 692 struct mfcctl *mfccp; 693 struct mfc *rt; 694 { 695 vifi_t vifi; 696 697 rt->mfc_parent = mfccp->mfcc_parent; 698 for (vifi = 0; vifi < numvifs; vifi++) 699 rt->mfc_ttls[vifi] = mfccp->mfcc_ttls[vifi]; 700 rt->mfc_expire = 0; 701 rt->mfc_stall = 0; 702 } 703 704 static void 705 expire_mfc(rt) 706 struct mfc *rt; 707 { 708 struct rtdetq *rte, *nrte; 709 710 for (rte = rt->mfc_stall; rte != NULL; rte = nrte) { 711 nrte = rte->next; 712 m_freem(rte->m); 713 free(rte, M_MRTABLE); 714 } 715 716 LIST_REMOVE(rt, mfc_hash); 717 free(rt, M_MRTABLE); 718 } 719 720 /* 721 * Add an mfc entry 722 */ 723 static int 724 add_mfc(m) 725 struct mbuf *m; 726 { 727 struct mfcctl *mfccp; 728 struct mfc *rt; 729 u_int32_t hash = 0; 730 struct rtdetq *rte, *nrte; 731 register u_short nstl; 732 int s; 733 734 if (m == 0 || m->m_len < sizeof(struct mfcctl)) 735 return (EINVAL); 736 737 mfccp = mtod(m, struct mfcctl *); 738 739 s = splsoftnet(); 740 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 741 742 /* If an entry already exists, just update the fields */ 743 if (rt) { 744 if (mrtdebug & DEBUG_MFC) 745 log(LOG_DEBUG,"add_mfc update o %x g %x p %x", 746 ntohl(mfccp->mfcc_origin.s_addr), 747 ntohl(mfccp->mfcc_mcastgrp.s_addr), 748 mfccp->mfcc_parent); 749 750 if (rt->mfc_expire) 751 nexpire[hash]--; 752 753 update_mfc(mfccp, rt); 754 755 splx(s); 756 return (0); 757 } 758 759 /* 760 * Find the entry for which the upcall was made and update 761 */ 762 nstl = 0; 763 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 764 for (rt = mfchashtbl[hash].lh_first; rt; rt = rt->mfc_hash.le_next) { 765 if (rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr && 766 rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr && 767 rt->mfc_stall != NULL) { 768 if (nstl++) 769 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p", 770 "multiple kernel entries", 771 ntohl(mfccp->mfcc_origin.s_addr), 772 ntohl(mfccp->mfcc_mcastgrp.s_addr), 773 mfccp->mfcc_parent, rt->mfc_stall); 774 775 if (mrtdebug & DEBUG_MFC) 776 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %p", 777 ntohl(mfccp->mfcc_origin.s_addr), 778 ntohl(mfccp->mfcc_mcastgrp.s_addr), 779 mfccp->mfcc_parent, rt->mfc_stall); 780 781 if (rt->mfc_expire) 782 nexpire[hash]--; 783 784 /* free packets Qed at the end of this entry */ 785 for (rte = rt->mfc_stall; rte != NULL; rte = nrte) { 786 nrte = rte->next; 787 #ifdef RSVP_ISI 788 ip_mdq(rte->m, rte->ifp, rt, -1); 789 #else 790 ip_mdq(rte->m, rte->ifp, rt); 791 #endif /* RSVP_ISI */ 792 m_freem(rte->m); 793 #ifdef UPCALL_TIMING 794 collate(&rte->t); 795 #endif /* UPCALL_TIMING */ 796 free(rte, M_MRTABLE); 797 } 798 799 update_mfc(mfccp, rt); 800 } 801 } 802 803 if (nstl == 0) { 804 /* 805 * No mfc; make a new one 806 */ 807 if (mrtdebug & DEBUG_MFC) 808 log(LOG_DEBUG,"add_mfc no upcall o %x g %x p %x", 809 ntohl(mfccp->mfcc_origin.s_addr), 810 ntohl(mfccp->mfcc_mcastgrp.s_addr), 811 mfccp->mfcc_parent); 812 813 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 814 if (rt == NULL) { 815 splx(s); 816 return (ENOBUFS); 817 } 818 819 rt->mfc_origin = mfccp->mfcc_origin; 820 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 821 /* initialize pkt counters per src-grp */ 822 rt->mfc_pkt_cnt = 0; 823 rt->mfc_byte_cnt = 0; 824 rt->mfc_wrong_if = 0; 825 timerclear(&rt->mfc_last_assert); 826 update_mfc(mfccp, rt); 827 828 /* insert new entry at head of hash chain */ 829 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 830 } 831 832 splx(s); 833 return (0); 834 } 835 836 #ifdef UPCALL_TIMING 837 /* 838 * collect delay statistics on the upcalls 839 */ 840 static void collate(t) 841 register struct timeval *t; 842 { 843 register u_int32_t d; 844 register struct timeval tp; 845 register u_int32_t delta; 846 847 microtime(&tp); 848 849 if (timercmp(t, &tp, <)) { 850 TV_DELTA(tp, *t, delta); 851 852 d = delta >> 10; 853 if (d > 50) 854 d = 50; 855 856 ++upcall_data[d]; 857 } 858 } 859 #endif /* UPCALL_TIMING */ 860 861 /* 862 * Delete an mfc entry 863 */ 864 static int 865 del_mfc(m) 866 struct mbuf *m; 867 { 868 struct mfcctl *mfccp; 869 struct mfc *rt; 870 int s; 871 872 if (m == 0 || m->m_len < sizeof(struct mfcctl)) 873 return (EINVAL); 874 875 mfccp = mtod(m, struct mfcctl *); 876 877 if (mrtdebug & DEBUG_MFC) 878 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x", 879 ntohl(mfccp->mfcc_origin.s_addr), ntohl(mfccp->mfcc_mcastgrp.s_addr)); 880 881 s = splsoftnet(); 882 883 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 884 if (rt == NULL) { 885 splx(s); 886 return (EADDRNOTAVAIL); 887 } 888 889 LIST_REMOVE(rt, mfc_hash); 890 free(rt, M_MRTABLE); 891 892 splx(s); 893 return (0); 894 } 895 896 static int 897 socket_send(s, mm, src) 898 struct socket *s; 899 struct mbuf *mm; 900 struct sockaddr_in *src; 901 { 902 if (s) { 903 if (sbappendaddr(&s->so_rcv, sintosa(src), mm, (struct mbuf *)0) != 0) { 904 sorwakeup(s); 905 return (0); 906 } 907 } 908 m_freem(mm); 909 return (-1); 910 } 911 912 /* 913 * IP multicast forwarding function. This function assumes that the packet 914 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 915 * pointed to by "ifp", and the packet is to be relayed to other networks 916 * that have members of the packet's destination IP multicast group. 917 * 918 * The packet is returned unscathed to the caller, unless it is 919 * erroneous, in which case a non-zero return value tells the caller to 920 * discard it. 921 */ 922 923 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 924 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 925 926 int 927 #ifdef RSVP_ISI 928 ip_mforward(m, ifp, imo) 929 #else 930 ip_mforward(m, ifp) 931 #endif /* RSVP_ISI */ 932 struct mbuf *m; 933 struct ifnet *ifp; 934 #ifdef RSVP_ISI 935 struct ip_moptions *imo; 936 #endif /* RSVP_ISI */ 937 { 938 register struct ip *ip = mtod(m, struct ip *); 939 register struct mfc *rt; 940 register u_char *ipoptions; 941 static int srctun = 0; 942 register struct mbuf *mm; 943 int s; 944 #ifdef RSVP_ISI 945 register struct vif *vifp; 946 vifi_t vifi; 947 #endif /* RSVP_ISI */ 948 949 if (mrtdebug & DEBUG_FORWARD) 950 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p", 951 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 952 953 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 954 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR) { 955 /* 956 * Packet arrived via a physical interface or 957 * an encapuslated tunnel. 958 */ 959 } else { 960 /* 961 * Packet arrived through a source-route tunnel. 962 * Source-route tunnels are no longer supported. 963 */ 964 if ((srctun++ % 1000) == 0) 965 log(LOG_ERR, "ip_mforward: received source-routed packet from %x", 966 ntohl(ip->ip_src.s_addr)); 967 968 return (1); 969 } 970 971 #ifdef RSVP_ISI 972 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 973 if (ip->ip_ttl < 255) 974 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 975 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 976 vifp = viftable + vifi; 977 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n", 978 ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi, 979 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 980 vifp->v_ifp->if_xname); 981 } 982 return (ip_mdq(m, ifp, rt, vifi)); 983 } 984 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 985 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n", 986 ntohl(ip->ip_src), ntohl(ip->ip_dst)); 987 } 988 #endif /* RSVP_ISI */ 989 990 /* 991 * Don't forward a packet with time-to-live of zero or one, 992 * or a packet destined to a local-only group. 993 */ 994 if (ip->ip_ttl <= 1 || 995 IN_LOCAL_GROUP(ip->ip_dst.s_addr)) 996 return (0); 997 998 /* 999 * Determine forwarding vifs from the forwarding cache table 1000 */ 1001 s = splsoftnet(); 1002 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt); 1003 1004 /* Entry exists, so forward if necessary */ 1005 if (rt != NULL) { 1006 splx(s); 1007 #ifdef RSVP_ISI 1008 return (ip_mdq(m, ifp, rt, -1)); 1009 #else 1010 return (ip_mdq(m, ifp, rt)); 1011 #endif /* RSVP_ISI */ 1012 } else { 1013 /* 1014 * If we don't have a route for packet's origin, 1015 * Make a copy of the packet & 1016 * send message to routing daemon 1017 */ 1018 1019 register struct mbuf *mb0; 1020 register struct rtdetq *rte; 1021 register u_int32_t hash; 1022 #ifdef UPCALL_TIMING 1023 struct timeval tp; 1024 1025 microtime(&tp); 1026 #endif /* UPCALL_TIMING */ 1027 1028 mrtstat.mrts_no_route++; 1029 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1030 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x", 1031 ntohl(ip->ip_src.s_addr), 1032 ntohl(ip->ip_dst.s_addr)); 1033 1034 /* 1035 * Allocate mbufs early so that we don't do extra work if we are 1036 * just going to fail anyway. 1037 */ 1038 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, M_NOWAIT); 1039 if (rte == NULL) { 1040 splx(s); 1041 return (ENOBUFS); 1042 } 1043 mb0 = m_copy(m, 0, M_COPYALL); 1044 if (mb0 == NULL) { 1045 free(rte, M_MRTABLE); 1046 splx(s); 1047 return (ENOBUFS); 1048 } 1049 1050 /* is there an upcall waiting for this packet? */ 1051 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1052 for (rt = mfchashtbl[hash].lh_first; rt; rt = rt->mfc_hash.le_next) { 1053 if (ip->ip_src.s_addr == rt->mfc_origin.s_addr && 1054 ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr && 1055 rt->mfc_stall != NULL) 1056 break; 1057 } 1058 1059 if (rt == NULL) { 1060 int hlen = ip->ip_hl << 2; 1061 int i; 1062 struct igmpmsg *im; 1063 1064 /* no upcall, so make a new entry */ 1065 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1066 if (rt == NULL) { 1067 free(rte, M_MRTABLE); 1068 m_free(mb0); 1069 splx(s); 1070 return (ENOBUFS); 1071 } 1072 /* Make a copy of the header to send to the user level process */ 1073 mm = m_copy(m, 0, hlen); 1074 M_PULLUP(mm, hlen); 1075 if (mm == NULL) { 1076 free(rte, M_MRTABLE); 1077 m_free(mb0); 1078 free(rt, M_MRTABLE); 1079 splx(s); 1080 return (ENOBUFS); 1081 } 1082 1083 /* 1084 * Send message to routing daemon to install 1085 * a route into the kernel table 1086 */ 1087 sin.sin_addr = ip->ip_src; 1088 1089 im = mtod(mm, struct igmpmsg *); 1090 im->im_msgtype = IGMPMSG_NOCACHE; 1091 im->im_mbz = 0; 1092 1093 mrtstat.mrts_upcalls++; 1094 1095 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1096 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full"); 1097 ++mrtstat.mrts_upq_sockfull; 1098 free(rte, M_MRTABLE); 1099 m_free(mb0); 1100 free(rt, M_MRTABLE); 1101 splx(s); 1102 return (ENOBUFS); 1103 } 1104 1105 /* insert new entry at head of hash chain */ 1106 rt->mfc_origin = ip->ip_src; 1107 rt->mfc_mcastgrp = ip->ip_dst; 1108 rt->mfc_pkt_cnt = 0; 1109 rt->mfc_byte_cnt = 0; 1110 rt->mfc_wrong_if = 0; 1111 rt->mfc_expire = UPCALL_EXPIRE; 1112 nexpire[hash]++; 1113 for (i = 0; i < numvifs; i++) 1114 rt->mfc_ttls[i] = 0; 1115 rt->mfc_parent = -1; 1116 1117 /* link into table */ 1118 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1119 /* Add this entry to the end of the queue */ 1120 rt->mfc_stall = rte; 1121 } else { 1122 /* determine if q has overflowed */ 1123 struct rtdetq **p; 1124 register int npkts = 0; 1125 1126 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1127 if (++npkts > MAX_UPQ) { 1128 mrtstat.mrts_upq_ovflw++; 1129 free(rte, M_MRTABLE); 1130 m_free(mb0); 1131 splx(s); 1132 return (0); 1133 } 1134 1135 /* Add this entry to the end of the queue */ 1136 *p = rte; 1137 } 1138 1139 rte->next = NULL; 1140 rte->m = mb0; 1141 rte->ifp = ifp; 1142 #ifdef UPCALL_TIMING 1143 rte->t = tp; 1144 #endif /* UPCALL_TIMING */ 1145 1146 1147 splx(s); 1148 1149 return (0); 1150 } 1151 } 1152 1153 1154 /*ARGSUSED*/ 1155 static void 1156 expire_upcalls(v) 1157 void *v; 1158 { 1159 int i; 1160 int s; 1161 1162 s = splsoftnet(); 1163 1164 for (i = 0; i < MFCTBLSIZ; i++) { 1165 register struct mfc *rt, *nrt; 1166 1167 if (nexpire[i] == 0) 1168 continue; 1169 1170 for (rt = mfchashtbl[i].lh_first; rt; rt = nrt) { 1171 nrt = rt->mfc_hash.le_next; 1172 1173 if (rt->mfc_expire == 0 || 1174 --rt->mfc_expire > 0) 1175 continue; 1176 nexpire[i]--; 1177 1178 ++mrtstat.mrts_cache_cleanups; 1179 if (mrtdebug & DEBUG_EXPIRE) 1180 log(LOG_DEBUG, 1181 "expire_upcalls: expiring (%x %x)", 1182 ntohl(rt->mfc_origin.s_addr), 1183 ntohl(rt->mfc_mcastgrp.s_addr)); 1184 1185 expire_mfc(rt); 1186 } 1187 } 1188 1189 splx(s); 1190 timeout(expire_upcalls, (caddr_t)0, EXPIRE_TIMEOUT); 1191 } 1192 1193 /* 1194 * Packet forwarding routine once entry in the cache is made 1195 */ 1196 static int 1197 #ifdef RSVP_ISI 1198 ip_mdq(m, ifp, rt, xmt_vif) 1199 #else 1200 ip_mdq(m, ifp, rt) 1201 #endif /* RSVP_ISI */ 1202 register struct mbuf *m; 1203 register struct ifnet *ifp; 1204 register struct mfc *rt; 1205 #ifdef RSVP_ISI 1206 register vifi_t xmt_vif; 1207 #endif /* RSVP_ISI */ 1208 { 1209 register struct ip *ip = mtod(m, struct ip *); 1210 register vifi_t vifi; 1211 register struct vif *vifp; 1212 register int plen = ntohs(ip->ip_len); 1213 1214 /* 1215 * Macro to send packet on vif. Since RSVP packets don't get counted on 1216 * input, they shouldn't get counted on output, so statistics keeping is 1217 * seperate. 1218 */ 1219 #define MC_SEND(ip,vifp,m) { \ 1220 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1221 encap_send((ip), (vifp), (m)); \ 1222 else \ 1223 phyint_send((ip), (vifp), (m)); \ 1224 } 1225 1226 #ifdef RSVP_ISI 1227 /* 1228 * If xmt_vif is not -1, send on only the requested vif. 1229 * 1230 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs. 1231 */ 1232 if (xmt_vif < numvifs) { 1233 MC_SEND(ip, viftable + xmt_vif, m); 1234 return (1); 1235 } 1236 #endif /* RSVP_ISI */ 1237 1238 /* 1239 * Don't forward if it didn't arrive from the parent vif for its origin. 1240 */ 1241 vifi = rt->mfc_parent; 1242 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1243 /* came in the wrong interface */ 1244 if (mrtdebug & DEBUG_FORWARD) 1245 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p", 1246 ifp, vifi, viftable[vifi].v_ifp); 1247 ++mrtstat.mrts_wrong_if; 1248 ++rt->mfc_wrong_if; 1249 /* 1250 * If we are doing PIM assert processing, and we are forwarding 1251 * packets on this interface, and it is a broadcast medium 1252 * interface (and not a tunnel), send a message to the routing daemon. 1253 */ 1254 if (pim_assert && rt->mfc_ttls[vifi] && 1255 (ifp->if_flags & IFF_BROADCAST) && 1256 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1257 struct mbuf *mm; 1258 struct igmpmsg *im; 1259 int hlen = ip->ip_hl << 2; 1260 struct timeval now; 1261 register u_int32_t delta; 1262 1263 microtime(&now); 1264 1265 TV_DELTA(rt->mfc_last_assert, now, delta); 1266 1267 if (delta > ASSERT_MSG_TIME) { 1268 mm = m_copy(m, 0, hlen); 1269 M_PULLUP(mm, hlen); 1270 if (mm == NULL) { 1271 return (ENOBUFS); 1272 } 1273 1274 rt->mfc_last_assert = now; 1275 1276 im = mtod(mm, struct igmpmsg *); 1277 im->im_msgtype = IGMPMSG_WRONGVIF; 1278 im->im_mbz = 0; 1279 im->im_vif = vifi; 1280 1281 sin.sin_addr = im->im_src; 1282 1283 socket_send(ip_mrouter, m, &sin); 1284 } 1285 } 1286 return (0); 1287 } 1288 1289 /* If I sourced this packet, it counts as output, else it was input. */ 1290 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1291 viftable[vifi].v_pkt_out++; 1292 viftable[vifi].v_bytes_out += plen; 1293 } else { 1294 viftable[vifi].v_pkt_in++; 1295 viftable[vifi].v_bytes_in += plen; 1296 } 1297 rt->mfc_pkt_cnt++; 1298 rt->mfc_byte_cnt += plen; 1299 1300 /* 1301 * For each vif, decide if a copy of the packet should be forwarded. 1302 * Forward if: 1303 * - the ttl exceeds the vif's threshold 1304 * - there are group members downstream on interface 1305 */ 1306 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1307 if ((rt->mfc_ttls[vifi] > 0) && 1308 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1309 vifp->v_pkt_out++; 1310 vifp->v_bytes_out += plen; 1311 MC_SEND(ip, vifp, m); 1312 } 1313 1314 return (0); 1315 } 1316 1317 #ifdef RSVP_ISI 1318 /* 1319 * check if a vif number is legal/ok. This is used by ip_output, to export 1320 * numvifs there, 1321 */ 1322 int 1323 legal_vif_num(vif) 1324 int vif; 1325 { 1326 if (vif >= 0 && vif < numvifs) 1327 return (1); 1328 else 1329 return (0); 1330 } 1331 #endif /* RSVP_ISI */ 1332 1333 static void 1334 phyint_send(ip, vifp, m) 1335 struct ip *ip; 1336 struct vif *vifp; 1337 struct mbuf *m; 1338 { 1339 register struct mbuf *mb_copy; 1340 register int hlen = ip->ip_hl << 2; 1341 1342 /* 1343 * Make a new reference to the packet; make sure that 1344 * the IP header is actually copied, not just referenced, 1345 * so that ip_output() only scribbles on the copy. 1346 */ 1347 mb_copy = m_copy(m, 0, M_COPYALL); 1348 M_PULLUP(mb_copy, hlen); 1349 if (mb_copy == NULL) 1350 return; 1351 1352 if (vifp->v_rate_limit <= 0) 1353 tbf_send_packet(vifp, mb_copy); 1354 else 1355 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1356 } 1357 1358 static void 1359 encap_send(ip, vifp, m) 1360 register struct ip *ip; 1361 register struct vif *vifp; 1362 register struct mbuf *m; 1363 { 1364 register struct mbuf *mb_copy; 1365 register struct ip *ip_copy; 1366 register int i, len = ip->ip_len + sizeof(multicast_encap_iphdr); 1367 1368 /* 1369 * copy the old packet & pullup it's IP header into the 1370 * new mbuf so we can modify it. Try to fill the new 1371 * mbuf since if we don't the ethernet driver will. 1372 */ 1373 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA); 1374 if (mb_copy == NULL) 1375 return; 1376 mb_copy->m_data += max_linkhdr; 1377 mb_copy->m_pkthdr.len = len; 1378 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1379 1380 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1381 m_freem(mb_copy); 1382 return; 1383 } 1384 i = MHLEN - max_linkhdr; 1385 if (i > len) 1386 i = len; 1387 mb_copy = m_pullup(mb_copy, i); 1388 if (mb_copy == NULL) 1389 return; 1390 1391 /* 1392 * fill in the encapsulating IP header. 1393 */ 1394 ip_copy = mtod(mb_copy, struct ip *); 1395 *ip_copy = multicast_encap_iphdr; 1396 ip_copy->ip_id = htons(ip_id++); 1397 ip_copy->ip_len = len; 1398 ip_copy->ip_src = vifp->v_lcl_addr; 1399 ip_copy->ip_dst = vifp->v_rmt_addr; 1400 1401 /* 1402 * turn the encapsulated IP header back into a valid one. 1403 */ 1404 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1405 --ip->ip_ttl; 1406 HTONS(ip->ip_len); 1407 HTONS(ip->ip_off); 1408 ip->ip_sum = 0; 1409 #if defined(LBL) && !defined(ultrix) && !defined(i386) 1410 ip->ip_sum = ~oc_cksum((caddr_t)ip, ip->ip_hl << 2, 0); 1411 #else 1412 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1413 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1414 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1415 #endif 1416 1417 if (vifp->v_rate_limit <= 0) 1418 tbf_send_packet(vifp, mb_copy); 1419 else 1420 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1421 } 1422 1423 /* 1424 * De-encapsulate a packet and feed it back through ip input (this 1425 * routine is called whenever IP gets a packet with proto type 1426 * ENCAP_PROTO and a local destination address). 1427 */ 1428 void 1429 #if __STDC__ 1430 ipip_input(struct mbuf *m, ...) 1431 #else 1432 ipip_input(m, va_alist) 1433 struct mbuf *m; 1434 va_dcl 1435 #endif 1436 { 1437 register int hlen; 1438 register struct ip *ip = mtod(m, struct ip *); 1439 register int s; 1440 register struct ifqueue *ifq; 1441 register struct vif *vifp; 1442 va_list ap; 1443 1444 va_start(ap, m); 1445 hlen = va_arg(ap, int); 1446 va_end(ap); 1447 1448 if (!have_encap_tunnel) { 1449 rip_input(m); 1450 return; 1451 } 1452 1453 /* 1454 * dump the packet if it's not to a multicast destination or if 1455 * we don't have an encapsulating tunnel with the source. 1456 * Note: This code assumes that the remote site IP address 1457 * uniquely identifies the tunnel (i.e., that this site has 1458 * at most one tunnel with the remote site). 1459 */ 1460 if (!IN_MULTICAST(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr)) { 1461 ++mrtstat.mrts_bad_tunnel; 1462 m_freem(m); 1463 return; 1464 } 1465 1466 if (ip->ip_src.s_addr != last_encap_src) { 1467 register struct vif *vife; 1468 1469 vifp = viftable; 1470 vife = vifp + numvifs; 1471 for (; vifp < vife; vifp++) 1472 if (vifp->v_flags & VIFF_TUNNEL && 1473 vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) 1474 break; 1475 if (vifp == vife) { 1476 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1477 m_freem(m); 1478 if (mrtdebug) 1479 log(LOG_DEBUG, "ip_mforward: no tunnel with %x", 1480 ntohl(ip->ip_src.s_addr)); 1481 return; 1482 } 1483 last_encap_vif = vifp; 1484 last_encap_src = ip->ip_src.s_addr; 1485 } else 1486 vifp = last_encap_vif; 1487 1488 m->m_data += hlen; 1489 m->m_len -= hlen; 1490 m->m_pkthdr.len -= hlen; 1491 m->m_pkthdr.rcvif = vifp->v_ifp; 1492 ifq = &ipintrq; 1493 s = splimp(); 1494 if (IF_QFULL(ifq)) { 1495 IF_DROP(ifq); 1496 m_freem(m); 1497 } else { 1498 IF_ENQUEUE(ifq, m); 1499 /* 1500 * normally we would need a "schednetisr(NETISR_IP)" 1501 * here but we were called by ip_input and it is going 1502 * to loop back & try to dequeue the packet we just 1503 * queued as soon as we return so we avoid the 1504 * unnecessary software interrrupt. 1505 */ 1506 } 1507 splx(s); 1508 } 1509 1510 /* 1511 * Token bucket filter module 1512 */ 1513 static void 1514 tbf_control(vifp, m, ip, p_len) 1515 register struct vif *vifp; 1516 register struct mbuf *m; 1517 register struct ip *ip; 1518 register u_int32_t p_len; 1519 { 1520 1521 tbf_update_tokens(vifp); 1522 1523 /* 1524 * If there are enough tokens, and the queue is empty, send this packet 1525 * out immediately. Otherwise, try to insert it on this vif's queue. 1526 */ 1527 if (vifp->v_tbf.q_len == 0) { 1528 if (p_len <= vifp->v_tbf.n_tok) { 1529 vifp->v_tbf.n_tok -= p_len; 1530 tbf_send_packet(vifp, m); 1531 } else if (p_len > MAX_BKT_SIZE) { 1532 /* drop if packet is too large */ 1533 mrtstat.mrts_pkt2large++; 1534 m_freem(m); 1535 } else { 1536 /* queue packet and timeout till later */ 1537 tbf_queue(vifp, m, ip); 1538 timeout(tbf_reprocess_q, vifp, 1); 1539 } 1540 } else { 1541 if (vifp->v_tbf.q_len >= MAXQSIZE && 1542 !tbf_dq_sel(vifp, ip)) { 1543 /* queue length too much, and couldn't make room */ 1544 mrtstat.mrts_q_overflow++; 1545 m_freem(m); 1546 } else { 1547 /* queue length low enough, or made room */ 1548 tbf_queue(vifp, m, ip); 1549 tbf_process_q(vifp); 1550 } 1551 } 1552 } 1553 1554 /* 1555 * adds a packet to the queue at the interface 1556 */ 1557 static void 1558 tbf_queue(vifp, m, ip) 1559 register struct vif *vifp; 1560 register struct mbuf *m; 1561 register struct ip *ip; 1562 { 1563 register u_int32_t ql; 1564 register int index = (vifp - viftable); 1565 register int s = splsoftnet(); 1566 1567 ql = vifp->v_tbf.q_len; 1568 1569 qtable[index][ql].pkt_m = m; 1570 qtable[index][ql].pkt_len = (mtod(m, struct ip *))->ip_len; 1571 qtable[index][ql].pkt_ip = ip; 1572 1573 vifp->v_tbf.q_len++; 1574 splx(s); 1575 } 1576 1577 1578 /* 1579 * processes the queue at the interface 1580 */ 1581 static void 1582 tbf_process_q(vifp) 1583 register struct vif *vifp; 1584 { 1585 register struct pkt_queue pkt_1; 1586 register int index = (vifp - viftable); 1587 register int s = splsoftnet(); 1588 1589 /* loop through the queue at the interface and send as many packets 1590 * as possible 1591 */ 1592 while (vifp->v_tbf.q_len > 0) { 1593 /* locate the first packet */ 1594 pkt_1 = qtable[index][0]; 1595 1596 /* determine if the packet can be sent */ 1597 if (pkt_1.pkt_len <= vifp->v_tbf.n_tok) { 1598 /* if so, 1599 * reduce no of tokens, dequeue the queue, 1600 * send the packet. 1601 */ 1602 vifp->v_tbf.n_tok -= pkt_1.pkt_len; 1603 1604 tbf_dequeue(vifp, 0); 1605 tbf_send_packet(vifp, pkt_1.pkt_m); 1606 } else 1607 break; 1608 } 1609 splx(s); 1610 } 1611 1612 /* 1613 * removes the jth packet from the queue at the interface 1614 */ 1615 static void 1616 tbf_dequeue(vifp, j) 1617 register struct vif *vifp; 1618 register int j; 1619 { 1620 register u_int32_t index = vifp - viftable; 1621 register int i; 1622 1623 for (i=j+1; i <= vifp->v_tbf.q_len - 1; i++) { 1624 qtable[index][i-1] = qtable[index][i]; 1625 } 1626 qtable[index][i-1].pkt_m = NULL; 1627 qtable[index][i-1].pkt_len = NULL; 1628 qtable[index][i-1].pkt_ip = NULL; 1629 1630 vifp->v_tbf.q_len--; 1631 1632 if (tbfdebug > 1) 1633 log(LOG_DEBUG, "tbf_dequeue: vif# %d qlen %d",vifp-viftable, i-1); 1634 } 1635 1636 static void 1637 tbf_reprocess_q(arg) 1638 void *arg; 1639 { 1640 register struct vif *vifp = arg; 1641 1642 if (ip_mrouter == NULL) 1643 return; 1644 1645 tbf_update_tokens(vifp); 1646 tbf_process_q(vifp); 1647 1648 if (vifp->v_tbf.q_len) 1649 timeout(tbf_reprocess_q, vifp, 1); 1650 } 1651 1652 /* function that will selectively discard a member of the queue 1653 * based on the precedence value and the priority obtained through 1654 * a lookup table - not yet implemented accurately! 1655 */ 1656 static int 1657 tbf_dq_sel(vifp, ip) 1658 register struct vif *vifp; 1659 register struct ip *ip; 1660 { 1661 register int i; 1662 register int s = splsoftnet(); 1663 register u_int p; 1664 1665 p = priority(vifp, ip); 1666 1667 for(i=vifp->v_tbf.q_len-1;i >= 0;i--) { 1668 if (p > priority(vifp, qtable[vifp-viftable][i].pkt_ip)) { 1669 m_freem(qtable[vifp-viftable][i].pkt_m); 1670 tbf_dequeue(vifp, i); 1671 splx(s); 1672 mrtstat.mrts_drop_sel++; 1673 return (1); 1674 } 1675 } 1676 splx(s); 1677 return (0); 1678 } 1679 1680 static void 1681 tbf_send_packet(vifp,m) 1682 register struct vif *vifp; 1683 register struct mbuf *m; 1684 { 1685 int error; 1686 int s = splsoftnet(); 1687 1688 if (vifp->v_flags & VIFF_TUNNEL) { 1689 /* If tunnel options */ 1690 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1691 IP_FORWARDING, NULL); 1692 } else { 1693 /* if physical interface option, extract the options and then send */ 1694 struct ip *ip = mtod(m, struct ip *); 1695 struct ip_moptions imo; 1696 imo.imo_multicast_ifp = vifp->v_ifp; 1697 imo.imo_multicast_ttl = ip->ip_ttl - 1; 1698 imo.imo_multicast_loop = 1; 1699 #ifdef RSVP_ISI 1700 imo.imo_multicast_vif = -1; 1701 #endif 1702 1703 error = ip_output(m, (struct mbuf *)0, (struct route *)0, 1704 IP_FORWARDING|IP_MULTICASTOPTS, &imo); 1705 if (mrtdebug & DEBUG_XMIT) 1706 log(LOG_DEBUG, "phyint_send on vif %d err %d", vifp-viftable, error); 1707 } 1708 splx(s); 1709 } 1710 1711 /* determine the current time and then 1712 * the elapsed time (between the last time and time now) 1713 * in milliseconds & update the no. of tokens in the bucket 1714 */ 1715 static void 1716 tbf_update_tokens(vifp) 1717 register struct vif *vifp; 1718 { 1719 struct timeval tp; 1720 register u_int32_t t; 1721 register u_int32_t elapsed; 1722 register int s = splsoftnet(); 1723 1724 microtime(&tp); 1725 1726 t = tp.tv_sec*1000 + tp.tv_usec/1000; 1727 1728 elapsed = (t - vifp->v_tbf.last_pkt_t) * vifp->v_rate_limit /8; 1729 vifp->v_tbf.n_tok += elapsed; 1730 vifp->v_tbf.last_pkt_t = t; 1731 1732 if (vifp->v_tbf.n_tok > MAX_BKT_SIZE) 1733 vifp->v_tbf.n_tok = MAX_BKT_SIZE; 1734 1735 splx(s); 1736 } 1737 1738 static int 1739 priority(vifp, ip) 1740 register struct vif *vifp; 1741 register struct ip *ip; 1742 { 1743 register int prio; 1744 1745 /* temporary hack; may add general packet classifier some day */ 1746 1747 /* 1748 * The UDP port space is divided up into four priority ranges: 1749 * [0, 16384) : unclassified - lowest priority 1750 * [16384, 32768) : audio - highest priority 1751 * [32768, 49152) : whiteboard - medium priority 1752 * [49152, 65536) : video - low priority 1753 */ 1754 if (ip->ip_p == IPPROTO_UDP) { 1755 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1756 1757 switch (ntohs(udp->uh_dport) & 0xc000) { 1758 case 0x4000: 1759 prio = 70; 1760 break; 1761 case 0x8000: 1762 prio = 60; 1763 break; 1764 case 0xc000: 1765 prio = 55; 1766 break; 1767 default: 1768 prio = 50; 1769 break; 1770 } 1771 1772 if (tbfdebug > 1) log(LOG_DEBUG, "port %x prio %d", ntohs(udp->uh_dport), prio); 1773 } else 1774 prio = 50; 1775 1776 1777 return (prio); 1778 } 1779 1780 /* 1781 * End of token bucket filter modifications 1782 */ 1783 1784 #ifdef RSVP_ISI 1785 1786 int 1787 ip_rsvp_vif_init(so, m) 1788 struct socket *so; 1789 struct mbuf *m; 1790 { 1791 int i; 1792 register int s; 1793 1794 if (rsvpdebug) 1795 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 1796 so->so_type, so->so_proto->pr_protocol); 1797 1798 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1799 return (EOPNOTSUPP); 1800 1801 /* Check mbuf. */ 1802 if (m == NULL || m->m_len != sizeof(int)) { 1803 return (EINVAL); 1804 } 1805 i = *(mtod(m, int *)); 1806 1807 if (rsvpdebug) 1808 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on); 1809 1810 s = splsoftnet(); 1811 1812 /* Check vif. */ 1813 if (!legal_vif_num(i)) { 1814 splx(s); 1815 return (EADDRNOTAVAIL); 1816 } 1817 1818 /* Check if socket is available. */ 1819 if (viftable[i].v_rsvpd != NULL) { 1820 splx(s); 1821 return (EADDRINUSE); 1822 } 1823 1824 viftable[i].v_rsvpd = so; 1825 /* This may seem silly, but we need to be sure we don't over-increment 1826 * the RSVP counter, in case something slips up. 1827 */ 1828 if (!viftable[i].v_rsvp_on) { 1829 viftable[i].v_rsvp_on = 1; 1830 rsvp_on++; 1831 } 1832 1833 splx(s); 1834 return (0); 1835 } 1836 1837 int 1838 ip_rsvp_vif_done(so, m) 1839 struct socket *so; 1840 struct mbuf *m; 1841 { 1842 int i; 1843 register int s; 1844 1845 if (rsvpdebug) 1846 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 1847 so->so_type, so->so_proto->pr_protocol); 1848 1849 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1850 return (EOPNOTSUPP); 1851 1852 /* Check mbuf. */ 1853 if (m == NULL || m->m_len != sizeof(int)) { 1854 return (EINVAL); 1855 } 1856 i = *(mtod(m, int *)); 1857 1858 s = splsoftnet(); 1859 1860 /* Check vif. */ 1861 if (!legal_vif_num(i)) { 1862 splx(s); 1863 return (EADDRNOTAVAIL); 1864 } 1865 1866 if (rsvpdebug) 1867 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n", 1868 viftable[i].v_rsvpd, so); 1869 1870 viftable[i].v_rsvpd = NULL; 1871 /* This may seem silly, but we need to be sure we don't over-decrement 1872 * the RSVP counter, in case something slips up. 1873 */ 1874 if (viftable[i].v_rsvp_on) { 1875 viftable[i].v_rsvp_on = 0; 1876 rsvp_on--; 1877 } 1878 1879 splx(s); 1880 return (0); 1881 } 1882 1883 void 1884 ip_rsvp_force_done(so) 1885 struct socket *so; 1886 { 1887 int vifi; 1888 register int s; 1889 1890 /* Don't bother if it is not the right type of socket. */ 1891 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1892 return; 1893 1894 s = splsoftnet(); 1895 1896 /* The socket may be attached to more than one vif...this 1897 * is perfectly legal. 1898 */ 1899 for (vifi = 0; vifi < numvifs; vifi++) { 1900 if (viftable[vifi].v_rsvpd == so) { 1901 viftable[vifi].v_rsvpd = NULL; 1902 /* This may seem silly, but we need to be sure we don't 1903 * over-decrement the RSVP counter, in case something slips up. 1904 */ 1905 if (viftable[vifi].v_rsvp_on) { 1906 viftable[vifi].v_rsvp_on = 0; 1907 rsvp_on--; 1908 } 1909 } 1910 } 1911 1912 splx(s); 1913 return; 1914 } 1915 1916 void 1917 rsvp_input(m, ifp) 1918 struct mbuf *m; 1919 struct ifnet *ifp; 1920 { 1921 int vifi; 1922 register struct ip *ip = mtod(m, struct ip *); 1923 static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET }; 1924 register int s; 1925 1926 if (rsvpdebug) 1927 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 1928 1929 /* Can still get packets with rsvp_on = 0 if there is a local member 1930 * of the group to which the RSVP packet is addressed. But in this 1931 * case we want to throw the packet away. 1932 */ 1933 if (!rsvp_on) { 1934 m_freem(m); 1935 return; 1936 } 1937 1938 /* If the old-style non-vif-associated socket is set, then use 1939 * it and ignore the new ones. 1940 */ 1941 if (ip_rsvpd != NULL) { 1942 if (rsvpdebug) 1943 printf("rsvp_input: Sending packet up old-style socket\n"); 1944 rip_input(m); 1945 return; 1946 } 1947 1948 s = splsoftnet(); 1949 1950 if (rsvpdebug) 1951 printf("rsvp_input: check vifs\n"); 1952 1953 /* Find which vif the packet arrived on. */ 1954 for (vifi = 0; vifi < numvifs; vifi++) { 1955 if (viftable[vifi].v_ifp == ifp) 1956 break; 1957 } 1958 1959 if (vifi == numvifs) { 1960 /* Can't find vif packet arrived on. Drop packet. */ 1961 if (rsvpdebug) 1962 printf("rsvp_input: Can't find vif for packet...dropping it.\n"); 1963 m_freem(m); 1964 splx(s); 1965 return; 1966 } 1967 1968 if (rsvpdebug) 1969 printf("rsvp_input: check socket\n"); 1970 1971 if (viftable[vifi].v_rsvpd == NULL) { 1972 /* drop packet, since there is no specific socket for this 1973 * interface */ 1974 if (rsvpdebug) 1975 printf("rsvp_input: No socket defined for vif %d\n",vifi); 1976 m_freem(m); 1977 splx(s); 1978 return; 1979 } 1980 1981 rsvp_src.sin_addr = ip->ip_src; 1982 1983 if (rsvpdebug && m) 1984 printf("rsvp_input: m->m_len = %d, sbspace() = %d\n", 1985 m->m_len,sbspace(&viftable[vifi].v_rsvpd->so_rcv)); 1986 1987 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 1988 if (rsvpdebug) 1989 printf("rsvp_input: Failed to append to socket\n"); 1990 else 1991 if (rsvpdebug) 1992 printf("rsvp_input: send packet up\n"); 1993 1994 splx(s); 1995 } 1996 #endif /* RSVP_ISI */ 1997