1 /* $NetBSD: ip_mroute.c,v 1.64 2002/11/05 02:07:25 fair Exp $ */ 2 3 /* 4 * Copyright (c) 1989 Stephen Deering 5 * Copyright (c) 1992, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * This code is derived from software contributed to Berkeley by 9 * Stephen Deering of Stanford University. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the University of 22 * California, Berkeley and its contributors. 23 * 4. Neither the name of the University nor the names of its contributors 24 * may be used to endorse or promote products derived from this software 25 * without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 37 * SUCH DAMAGE. 38 * 39 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93 40 */ 41 42 /* 43 * IP multicast forwarding procedures 44 * 45 * Written by David Waitzman, BBN Labs, August 1988. 46 * Modified by Steve Deering, Stanford, February 1989. 47 * Modified by Mark J. Steiglitz, Stanford, May, 1991 48 * Modified by Van Jacobson, LBL, January 1993 49 * Modified by Ajit Thyagarajan, PARC, August 1993 50 * Modified by Bill Fenner, PARC, April 1994 51 * Modified by Charles M. Hannum, NetBSD, May 1995. 52 * 53 * MROUTING Revision: 1.2 54 */ 55 56 #include <sys/cdefs.h> 57 __KERNEL_RCSID(0, "$NetBSD: ip_mroute.c,v 1.64 2002/11/05 02:07:25 fair Exp $"); 58 59 #include "opt_ipsec.h" 60 61 #include <sys/param.h> 62 #include <sys/systm.h> 63 #include <sys/callout.h> 64 #include <sys/mbuf.h> 65 #include <sys/socket.h> 66 #include <sys/socketvar.h> 67 #include <sys/protosw.h> 68 #include <sys/errno.h> 69 #include <sys/time.h> 70 #include <sys/kernel.h> 71 #include <sys/ioctl.h> 72 #include <sys/syslog.h> 73 #include <net/if.h> 74 #include <net/route.h> 75 #include <net/raw_cb.h> 76 #include <netinet/in.h> 77 #include <netinet/in_var.h> 78 #include <netinet/in_systm.h> 79 #include <netinet/ip.h> 80 #include <netinet/ip_var.h> 81 #include <netinet/in_pcb.h> 82 #include <netinet/udp.h> 83 #include <netinet/igmp.h> 84 #include <netinet/igmp_var.h> 85 #include <netinet/ip_mroute.h> 86 #include <netinet/ip_encap.h> 87 88 #ifdef IPSEC 89 #include <netinet6/ipsec.h> 90 #include <netkey/key.h> 91 #endif 92 93 #include <machine/stdarg.h> 94 95 #define IP_MULTICASTOPTS 0 96 #define M_PULLUP(m, len) \ 97 do { \ 98 if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \ 99 (m) = m_pullup((m), (len)); \ 100 } while (/*CONSTCOND*/ 0) 101 102 /* 103 * Globals. All but ip_mrouter and ip_mrtproto could be static, 104 * except for netstat or debugging purposes. 105 */ 106 struct socket *ip_mrouter = 0; 107 int ip_mrtproto = IGMP_DVMRP; /* for netstat only */ 108 109 #define NO_RTE_FOUND 0x1 110 #define RTE_FOUND 0x2 111 112 #define MFCHASH(a, g) \ 113 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \ 114 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash) 115 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl; 116 u_long mfchash; 117 118 u_char nexpire[MFCTBLSIZ]; 119 struct vif viftable[MAXVIFS]; 120 struct mrtstat mrtstat; 121 u_int mrtdebug = 0; /* debug level */ 122 #define DEBUG_MFC 0x02 123 #define DEBUG_FORWARD 0x04 124 #define DEBUG_EXPIRE 0x08 125 #define DEBUG_XMIT 0x10 126 u_int tbfdebug = 0; /* tbf debug level */ 127 #ifdef RSVP_ISI 128 u_int rsvpdebug = 0; /* rsvp debug level */ 129 extern struct socket *ip_rsvpd; 130 extern int rsvp_on; 131 #endif /* RSVP_ISI */ 132 133 /* vif attachment using sys/netinet/ip_encap.c */ 134 extern struct domain inetdomain; 135 static void vif_input __P((struct mbuf *, ...)); 136 static int vif_encapcheck __P((const struct mbuf *, int, int, void *)); 137 static struct protosw vif_protosw = 138 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR, 139 vif_input, rip_output, 0, rip_ctloutput, 140 rip_usrreq, 141 0, 0, 0, 0, 142 }; 143 144 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 145 #define UPCALL_EXPIRE 6 /* number of timeouts */ 146 147 /* 148 * Define the token bucket filter structures 149 */ 150 151 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 152 153 static int get_sg_cnt __P((struct sioc_sg_req *)); 154 static int get_vif_cnt __P((struct sioc_vif_req *)); 155 static int ip_mrouter_init __P((struct socket *, struct mbuf *)); 156 static int get_version __P((struct mbuf *)); 157 static int set_assert __P((struct mbuf *)); 158 static int get_assert __P((struct mbuf *)); 159 static int add_vif __P((struct mbuf *)); 160 static int del_vif __P((struct mbuf *)); 161 static void update_mfc __P((struct mfcctl *, struct mfc *)); 162 static void expire_mfc __P((struct mfc *)); 163 static int add_mfc __P((struct mbuf *)); 164 #ifdef UPCALL_TIMING 165 static void collate __P((struct timeval *)); 166 #endif 167 static int del_mfc __P((struct mbuf *)); 168 static int socket_send __P((struct socket *, struct mbuf *, 169 struct sockaddr_in *)); 170 static void expire_upcalls __P((void *)); 171 #ifdef RSVP_ISI 172 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *, vifi_t)); 173 #else 174 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *)); 175 #endif 176 static void phyint_send __P((struct ip *, struct vif *, struct mbuf *)); 177 static void encap_send __P((struct ip *, struct vif *, struct mbuf *)); 178 static void tbf_control __P((struct vif *, struct mbuf *, struct ip *, 179 u_int32_t)); 180 static void tbf_queue __P((struct vif *, struct mbuf *)); 181 static void tbf_process_q __P((struct vif *)); 182 static void tbf_reprocess_q __P((void *)); 183 static int tbf_dq_sel __P((struct vif *, struct ip *)); 184 static void tbf_send_packet __P((struct vif *, struct mbuf *)); 185 static void tbf_update_tokens __P((struct vif *)); 186 static int priority __P((struct vif *, struct ip *)); 187 188 /* 189 * 'Interfaces' associated with decapsulator (so we can tell 190 * packets that went through it from ones that get reflected 191 * by a broken gateway). These interfaces are never linked into 192 * the system ifnet list & no routes point to them. I.e., packets 193 * can't be sent this way. They only exist as a placeholder for 194 * multicast source verification. 195 */ 196 #if 0 197 struct ifnet multicast_decap_if[MAXVIFS]; 198 #endif 199 200 #define ENCAP_TTL 64 201 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 202 203 /* prototype IP hdr for encapsulated packets */ 204 struct ip multicast_encap_iphdr = { 205 #if BYTE_ORDER == LITTLE_ENDIAN 206 sizeof(struct ip) >> 2, IPVERSION, 207 #else 208 IPVERSION, sizeof(struct ip) >> 2, 209 #endif 210 0, /* tos */ 211 sizeof(struct ip), /* total length */ 212 0, /* id */ 213 0, /* frag offset */ 214 ENCAP_TTL, ENCAP_PROTO, 215 0, /* checksum */ 216 }; 217 218 /* 219 * Private variables. 220 */ 221 static vifi_t numvifs = 0; 222 223 static struct callout expire_upcalls_ch; 224 225 /* 226 * one-back cache used by vif_encapcheck to locate a tunnel's vif 227 * given a datagram's src ip address. 228 */ 229 static struct in_addr last_encap_src; 230 static struct vif *last_encap_vif; 231 232 /* 233 * whether or not special PIM assert processing is enabled. 234 */ 235 static int pim_assert; 236 /* 237 * Rate limit for assert notification messages, in usec 238 */ 239 #define ASSERT_MSG_TIME 3000000 240 241 /* 242 * Find a route for a given origin IP address and Multicast group address 243 * Type of service parameter to be added in the future!!! 244 */ 245 246 #define MFCFIND(o, g, rt) { \ 247 struct mfc *_rt; \ 248 (rt) = 0; \ 249 ++mrtstat.mrts_mfc_lookups; \ 250 LIST_FOREACH(_rt, &mfchashtbl[MFCHASH(o, g)], mfc_hash) { \ 251 if (in_hosteq(_rt->mfc_origin, (o)) && \ 252 in_hosteq(_rt->mfc_mcastgrp, (g)) && \ 253 _rt->mfc_stall == 0) { \ 254 (rt) = _rt; \ 255 break; \ 256 } \ 257 } \ 258 if ((rt) == 0) \ 259 ++mrtstat.mrts_mfc_misses; \ 260 } 261 262 /* 263 * Macros to compute elapsed time efficiently 264 * Borrowed from Van Jacobson's scheduling code 265 */ 266 #define TV_DELTA(a, b, delta) { \ 267 int xxs; \ 268 delta = (a).tv_usec - (b).tv_usec; \ 269 xxs = (a).tv_sec - (b).tv_sec; \ 270 switch (xxs) { \ 271 case 2: \ 272 delta += 1000000; \ 273 /* fall through */ \ 274 case 1: \ 275 delta += 1000000; \ 276 /* fall through */ \ 277 case 0: \ 278 break; \ 279 default: \ 280 delta += (1000000 * xxs); \ 281 break; \ 282 } \ 283 } 284 285 #ifdef UPCALL_TIMING 286 u_int32_t upcall_data[51]; 287 #endif /* UPCALL_TIMING */ 288 289 /* 290 * Handle MRT setsockopt commands to modify the multicast routing tables. 291 */ 292 int 293 ip_mrouter_set(so, optname, m) 294 struct socket *so; 295 int optname; 296 struct mbuf **m; 297 { 298 int error; 299 300 if (optname != MRT_INIT && so != ip_mrouter) 301 error = ENOPROTOOPT; 302 else 303 switch (optname) { 304 case MRT_INIT: 305 error = ip_mrouter_init(so, *m); 306 break; 307 case MRT_DONE: 308 error = ip_mrouter_done(); 309 break; 310 case MRT_ADD_VIF: 311 error = add_vif(*m); 312 break; 313 case MRT_DEL_VIF: 314 error = del_vif(*m); 315 break; 316 case MRT_ADD_MFC: 317 error = add_mfc(*m); 318 break; 319 case MRT_DEL_MFC: 320 error = del_mfc(*m); 321 break; 322 case MRT_ASSERT: 323 error = set_assert(*m); 324 break; 325 default: 326 error = ENOPROTOOPT; 327 break; 328 } 329 330 if (*m) 331 m_free(*m); 332 return (error); 333 } 334 335 /* 336 * Handle MRT getsockopt commands 337 */ 338 int 339 ip_mrouter_get(so, optname, m) 340 struct socket *so; 341 int optname; 342 struct mbuf **m; 343 { 344 int error; 345 346 if (so != ip_mrouter) 347 error = ENOPROTOOPT; 348 else { 349 *m = m_get(M_WAIT, MT_SOOPTS); 350 351 switch (optname) { 352 case MRT_VERSION: 353 error = get_version(*m); 354 break; 355 case MRT_ASSERT: 356 error = get_assert(*m); 357 break; 358 default: 359 error = ENOPROTOOPT; 360 break; 361 } 362 363 if (error) 364 m_free(*m); 365 } 366 367 return (error); 368 } 369 370 /* 371 * Handle ioctl commands to obtain information from the cache 372 */ 373 int 374 mrt_ioctl(so, cmd, data) 375 struct socket *so; 376 u_long cmd; 377 caddr_t data; 378 { 379 int error; 380 381 if (so != ip_mrouter) 382 error = EINVAL; 383 else 384 switch (cmd) { 385 case SIOCGETVIFCNT: 386 error = get_vif_cnt((struct sioc_vif_req *)data); 387 break; 388 case SIOCGETSGCNT: 389 error = get_sg_cnt((struct sioc_sg_req *)data); 390 break; 391 default: 392 error = EINVAL; 393 break; 394 } 395 396 return (error); 397 } 398 399 /* 400 * returns the packet, byte, rpf-failure count for the source group provided 401 */ 402 static int 403 get_sg_cnt(req) 404 struct sioc_sg_req *req; 405 { 406 struct mfc *rt; 407 int s; 408 409 s = splsoftnet(); 410 MFCFIND(req->src, req->grp, rt); 411 splx(s); 412 if (rt != 0) { 413 req->pktcnt = rt->mfc_pkt_cnt; 414 req->bytecnt = rt->mfc_byte_cnt; 415 req->wrong_if = rt->mfc_wrong_if; 416 } else 417 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 418 419 return (0); 420 } 421 422 /* 423 * returns the input and output packet and byte counts on the vif provided 424 */ 425 static int 426 get_vif_cnt(req) 427 struct sioc_vif_req *req; 428 { 429 vifi_t vifi = req->vifi; 430 431 if (vifi >= numvifs) 432 return (EINVAL); 433 434 req->icount = viftable[vifi].v_pkt_in; 435 req->ocount = viftable[vifi].v_pkt_out; 436 req->ibytes = viftable[vifi].v_bytes_in; 437 req->obytes = viftable[vifi].v_bytes_out; 438 439 return (0); 440 } 441 442 /* 443 * Enable multicast routing 444 */ 445 static int 446 ip_mrouter_init(so, m) 447 struct socket *so; 448 struct mbuf *m; 449 { 450 int *v; 451 452 if (mrtdebug) 453 log(LOG_DEBUG, 454 "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 455 so->so_type, so->so_proto->pr_protocol); 456 457 if (so->so_type != SOCK_RAW || 458 so->so_proto->pr_protocol != IPPROTO_IGMP) 459 return (EOPNOTSUPP); 460 461 if (m == 0 || m->m_len < sizeof(int)) 462 return (EINVAL); 463 464 v = mtod(m, int *); 465 if (*v != 1) 466 return (EINVAL); 467 468 if (ip_mrouter != 0) 469 return (EADDRINUSE); 470 471 ip_mrouter = so; 472 473 mfchashtbl = 474 hashinit(MFCTBLSIZ, HASH_LIST, M_MRTABLE, M_WAITOK, &mfchash); 475 bzero((caddr_t)nexpire, sizeof(nexpire)); 476 477 pim_assert = 0; 478 479 callout_init(&expire_upcalls_ch); 480 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 481 expire_upcalls, NULL); 482 483 if (mrtdebug) 484 log(LOG_DEBUG, "ip_mrouter_init\n"); 485 486 return (0); 487 } 488 489 /* 490 * Disable multicast routing 491 */ 492 int 493 ip_mrouter_done() 494 { 495 vifi_t vifi; 496 struct vif *vifp; 497 int i; 498 int s; 499 500 s = splsoftnet(); 501 502 /* Clear out all the vifs currently in use. */ 503 for (vifi = 0; vifi < numvifs; vifi++) { 504 vifp = &viftable[vifi]; 505 if (!in_nullhost(vifp->v_lcl_addr)) 506 reset_vif(vifp); 507 } 508 509 numvifs = 0; 510 pim_assert = 0; 511 512 callout_stop(&expire_upcalls_ch); 513 514 /* 515 * Free all multicast forwarding cache entries. 516 */ 517 for (i = 0; i < MFCTBLSIZ; i++) { 518 struct mfc *rt, *nrt; 519 520 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 521 nrt = LIST_NEXT(rt, mfc_hash); 522 523 expire_mfc(rt); 524 } 525 } 526 527 free(mfchashtbl, M_MRTABLE); 528 mfchashtbl = 0; 529 530 /* Reset de-encapsulation cache. */ 531 532 ip_mrouter = 0; 533 534 splx(s); 535 536 if (mrtdebug) 537 log(LOG_DEBUG, "ip_mrouter_done\n"); 538 539 return (0); 540 } 541 542 static int 543 get_version(m) 544 struct mbuf *m; 545 { 546 int *v = mtod(m, int *); 547 548 *v = 0x0305; /* XXX !!!! */ 549 m->m_len = sizeof(int); 550 return (0); 551 } 552 553 /* 554 * Set PIM assert processing global 555 */ 556 static int 557 set_assert(m) 558 struct mbuf *m; 559 { 560 int *i; 561 562 if (m == 0 || m->m_len < sizeof(int)) 563 return (EINVAL); 564 565 i = mtod(m, int *); 566 pim_assert = !!*i; 567 return (0); 568 } 569 570 /* 571 * Get PIM assert processing global 572 */ 573 static int 574 get_assert(m) 575 struct mbuf *m; 576 { 577 int *i = mtod(m, int *); 578 579 *i = pim_assert; 580 m->m_len = sizeof(int); 581 return (0); 582 } 583 584 static struct sockaddr_in sin = { sizeof(sin), AF_INET }; 585 586 /* 587 * Add a vif to the vif table 588 */ 589 static int 590 add_vif(m) 591 struct mbuf *m; 592 { 593 struct vifctl *vifcp; 594 struct vif *vifp; 595 struct ifaddr *ifa; 596 struct ifnet *ifp; 597 struct ifreq ifr; 598 int error, s; 599 600 if (m == 0 || m->m_len < sizeof(struct vifctl)) 601 return (EINVAL); 602 603 vifcp = mtod(m, struct vifctl *); 604 if (vifcp->vifc_vifi >= MAXVIFS) 605 return (EINVAL); 606 607 vifp = &viftable[vifcp->vifc_vifi]; 608 if (!in_nullhost(vifp->v_lcl_addr)) 609 return (EADDRINUSE); 610 611 /* Find the interface with an address in AF_INET family. */ 612 sin.sin_addr = vifcp->vifc_lcl_addr; 613 ifa = ifa_ifwithaddr(sintosa(&sin)); 614 if (ifa == 0) 615 return (EADDRNOTAVAIL); 616 617 if (vifcp->vifc_flags & VIFF_TUNNEL) { 618 if (vifcp->vifc_flags & VIFF_SRCRT) { 619 log(LOG_ERR, "Source routed tunnels not supported\n"); 620 return (EOPNOTSUPP); 621 } 622 623 /* attach this vif to decapsulator dispatch table */ 624 vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4, 625 vif_encapcheck, &vif_protosw, vifp); 626 if (!vifp->v_encap_cookie) 627 return (EINVAL); 628 629 /* Create a fake encapsulation interface. */ 630 ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK); 631 bzero(ifp, sizeof(*ifp)); 632 sprintf(ifp->if_xname, "mdecap%d", vifcp->vifc_vifi); 633 634 /* Prepare cached route entry. */ 635 bzero(&vifp->v_route, sizeof(vifp->v_route)); 636 } else { 637 /* Use the physical interface associated with the address. */ 638 ifp = ifa->ifa_ifp; 639 640 /* Make sure the interface supports multicast. */ 641 if ((ifp->if_flags & IFF_MULTICAST) == 0) 642 return (EOPNOTSUPP); 643 644 /* Enable promiscuous reception of all IP multicasts. */ 645 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 646 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 647 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr; 648 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr); 649 if (error) 650 return (error); 651 } 652 653 s = splsoftnet(); 654 655 /* Define parameters for the tbf structure. */ 656 vifp->tbf_q = 0; 657 vifp->tbf_t = &vifp->tbf_q; 658 microtime(&vifp->tbf_last_pkt_t); 659 vifp->tbf_n_tok = 0; 660 vifp->tbf_q_len = 0; 661 vifp->tbf_max_q_len = MAXQSIZE; 662 663 vifp->v_flags = vifcp->vifc_flags; 664 vifp->v_threshold = vifcp->vifc_threshold; 665 /* scaling up here allows division by 1024 in critical code */ 666 vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000; 667 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 668 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 669 vifp->v_ifp = ifp; 670 /* Initialize per vif pkt counters. */ 671 vifp->v_pkt_in = 0; 672 vifp->v_pkt_out = 0; 673 vifp->v_bytes_in = 0; 674 vifp->v_bytes_out = 0; 675 676 callout_init(&vifp->v_repq_ch); 677 678 #ifdef RSVP_ISI 679 vifp->v_rsvp_on = 0; 680 vifp->v_rsvpd = 0; 681 #endif /* RSVP_ISI */ 682 683 splx(s); 684 685 /* Adjust numvifs up if the vifi is higher than numvifs. */ 686 if (numvifs <= vifcp->vifc_vifi) 687 numvifs = vifcp->vifc_vifi + 1; 688 689 if (mrtdebug) 690 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 691 vifcp->vifc_vifi, 692 ntohl(vifcp->vifc_lcl_addr.s_addr), 693 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 694 ntohl(vifcp->vifc_rmt_addr.s_addr), 695 vifcp->vifc_threshold, 696 vifcp->vifc_rate_limit); 697 698 return (0); 699 } 700 701 void 702 reset_vif(vifp) 703 struct vif *vifp; 704 { 705 struct mbuf *m, *n; 706 struct ifnet *ifp; 707 struct ifreq ifr; 708 709 callout_stop(&vifp->v_repq_ch); 710 711 /* detach this vif from decapsulator dispatch table */ 712 encap_detach(vifp->v_encap_cookie); 713 vifp->v_encap_cookie = NULL; 714 715 for (m = vifp->tbf_q; m != 0; m = n) { 716 n = m->m_nextpkt; 717 m_freem(m); 718 } 719 720 if (vifp->v_flags & VIFF_TUNNEL) { 721 free(vifp->v_ifp, M_MRTABLE); 722 if (vifp == last_encap_vif) { 723 last_encap_vif = 0; 724 last_encap_src = zeroin_addr; 725 } 726 } else { 727 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 728 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 729 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr; 730 ifp = vifp->v_ifp; 731 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 732 } 733 bzero((caddr_t)vifp, sizeof(*vifp)); 734 } 735 736 /* 737 * Delete a vif from the vif table 738 */ 739 static int 740 del_vif(m) 741 struct mbuf *m; 742 { 743 vifi_t *vifip; 744 struct vif *vifp; 745 vifi_t vifi; 746 int s; 747 748 if (m == 0 || m->m_len < sizeof(vifi_t)) 749 return (EINVAL); 750 751 vifip = mtod(m, vifi_t *); 752 if (*vifip >= numvifs) 753 return (EINVAL); 754 755 vifp = &viftable[*vifip]; 756 if (in_nullhost(vifp->v_lcl_addr)) 757 return (EADDRNOTAVAIL); 758 759 s = splsoftnet(); 760 761 reset_vif(vifp); 762 763 /* Adjust numvifs down */ 764 for (vifi = numvifs; vifi > 0; vifi--) 765 if (!in_nullhost(viftable[vifi-1].v_lcl_addr)) 766 break; 767 numvifs = vifi; 768 769 splx(s); 770 771 if (mrtdebug) 772 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 773 774 return (0); 775 } 776 777 static void 778 update_mfc(mfccp, rt) 779 struct mfcctl *mfccp; 780 struct mfc *rt; 781 { 782 vifi_t vifi; 783 784 rt->mfc_parent = mfccp->mfcc_parent; 785 for (vifi = 0; vifi < numvifs; vifi++) 786 rt->mfc_ttls[vifi] = mfccp->mfcc_ttls[vifi]; 787 rt->mfc_expire = 0; 788 rt->mfc_stall = 0; 789 } 790 791 static void 792 expire_mfc(rt) 793 struct mfc *rt; 794 { 795 struct rtdetq *rte, *nrte; 796 797 for (rte = rt->mfc_stall; rte != 0; rte = nrte) { 798 nrte = rte->next; 799 m_freem(rte->m); 800 free(rte, M_MRTABLE); 801 } 802 803 LIST_REMOVE(rt, mfc_hash); 804 free(rt, M_MRTABLE); 805 } 806 807 /* 808 * Add an mfc entry 809 */ 810 static int 811 add_mfc(m) 812 struct mbuf *m; 813 { 814 struct mfcctl *mfccp; 815 struct mfc *rt; 816 u_int32_t hash = 0; 817 struct rtdetq *rte, *nrte; 818 u_short nstl; 819 int s; 820 821 if (m == 0 || m->m_len < sizeof(struct mfcctl)) 822 return (EINVAL); 823 824 mfccp = mtod(m, struct mfcctl *); 825 826 s = splsoftnet(); 827 MFCFIND(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp, rt); 828 829 /* If an entry already exists, just update the fields */ 830 if (rt) { 831 if (mrtdebug & DEBUG_MFC) 832 log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n", 833 ntohl(mfccp->mfcc_origin.s_addr), 834 ntohl(mfccp->mfcc_mcastgrp.s_addr), 835 mfccp->mfcc_parent); 836 837 if (rt->mfc_expire) 838 nexpire[hash]--; 839 840 update_mfc(mfccp, rt); 841 842 splx(s); 843 return (0); 844 } 845 846 /* 847 * Find the entry for which the upcall was made and update 848 */ 849 nstl = 0; 850 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp); 851 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 852 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 853 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) && 854 rt->mfc_stall != 0) { 855 if (nstl++) 856 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n", 857 "multiple kernel entries", 858 ntohl(mfccp->mfcc_origin.s_addr), 859 ntohl(mfccp->mfcc_mcastgrp.s_addr), 860 mfccp->mfcc_parent, rt->mfc_stall); 861 862 if (mrtdebug & DEBUG_MFC) 863 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %p\n", 864 ntohl(mfccp->mfcc_origin.s_addr), 865 ntohl(mfccp->mfcc_mcastgrp.s_addr), 866 mfccp->mfcc_parent, rt->mfc_stall); 867 868 if (rt->mfc_expire) 869 nexpire[hash]--; 870 871 rte = rt->mfc_stall; 872 update_mfc(mfccp, rt); 873 874 /* free packets Qed at the end of this entry */ 875 for (; rte != 0; rte = nrte) { 876 nrte = rte->next; 877 #ifdef RSVP_ISI 878 ip_mdq(rte->m, rte->ifp, rt, -1); 879 #else 880 ip_mdq(rte->m, rte->ifp, rt); 881 #endif /* RSVP_ISI */ 882 m_freem(rte->m); 883 #ifdef UPCALL_TIMING 884 collate(&rte->t); 885 #endif /* UPCALL_TIMING */ 886 free(rte, M_MRTABLE); 887 } 888 } 889 } 890 891 if (nstl == 0) { 892 /* 893 * No mfc; make a new one 894 */ 895 if (mrtdebug & DEBUG_MFC) 896 log(LOG_DEBUG,"add_mfc no upcall o %x g %x p %x\n", 897 ntohl(mfccp->mfcc_origin.s_addr), 898 ntohl(mfccp->mfcc_mcastgrp.s_addr), 899 mfccp->mfcc_parent); 900 901 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 902 if (rt == 0) { 903 splx(s); 904 return (ENOBUFS); 905 } 906 907 rt->mfc_origin = mfccp->mfcc_origin; 908 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 909 /* initialize pkt counters per src-grp */ 910 rt->mfc_pkt_cnt = 0; 911 rt->mfc_byte_cnt = 0; 912 rt->mfc_wrong_if = 0; 913 timerclear(&rt->mfc_last_assert); 914 update_mfc(mfccp, rt); 915 916 /* insert new entry at head of hash chain */ 917 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 918 } 919 920 splx(s); 921 return (0); 922 } 923 924 #ifdef UPCALL_TIMING 925 /* 926 * collect delay statistics on the upcalls 927 */ 928 static void collate(t) 929 struct timeval *t; 930 { 931 u_int32_t d; 932 struct timeval tp; 933 u_int32_t delta; 934 935 microtime(&tp); 936 937 if (timercmp(t, &tp, <)) { 938 TV_DELTA(tp, *t, delta); 939 940 d = delta >> 10; 941 if (d > 50) 942 d = 50; 943 944 ++upcall_data[d]; 945 } 946 } 947 #endif /* UPCALL_TIMING */ 948 949 /* 950 * Delete an mfc entry 951 */ 952 static int 953 del_mfc(m) 954 struct mbuf *m; 955 { 956 struct mfcctl *mfccp; 957 struct mfc *rt; 958 int s; 959 960 if (m == 0 || m->m_len < sizeof(struct mfcctl)) 961 return (EINVAL); 962 963 mfccp = mtod(m, struct mfcctl *); 964 965 if (mrtdebug & DEBUG_MFC) 966 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n", 967 ntohl(mfccp->mfcc_origin.s_addr), 968 ntohl(mfccp->mfcc_mcastgrp.s_addr)); 969 970 s = splsoftnet(); 971 972 MFCFIND(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp, rt); 973 if (rt == 0) { 974 splx(s); 975 return (EADDRNOTAVAIL); 976 } 977 978 LIST_REMOVE(rt, mfc_hash); 979 free(rt, M_MRTABLE); 980 981 splx(s); 982 return (0); 983 } 984 985 static int 986 socket_send(s, mm, src) 987 struct socket *s; 988 struct mbuf *mm; 989 struct sockaddr_in *src; 990 { 991 if (s) { 992 if (sbappendaddr(&s->so_rcv, sintosa(src), mm, (struct mbuf *)0) != 0) { 993 sorwakeup(s); 994 return (0); 995 } 996 } 997 m_freem(mm); 998 return (-1); 999 } 1000 1001 /* 1002 * IP multicast forwarding function. This function assumes that the packet 1003 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1004 * pointed to by "ifp", and the packet is to be relayed to other networks 1005 * that have members of the packet's destination IP multicast group. 1006 * 1007 * The packet is returned unscathed to the caller, unless it is 1008 * erroneous, in which case a non-zero return value tells the caller to 1009 * discard it. 1010 */ 1011 1012 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1013 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1014 1015 int 1016 #ifdef RSVP_ISI 1017 ip_mforward(m, ifp, imo) 1018 #else 1019 ip_mforward(m, ifp) 1020 #endif /* RSVP_ISI */ 1021 struct mbuf *m; 1022 struct ifnet *ifp; 1023 #ifdef RSVP_ISI 1024 struct ip_moptions *imo; 1025 #endif /* RSVP_ISI */ 1026 { 1027 struct ip *ip = mtod(m, struct ip *); 1028 struct mfc *rt; 1029 u_char *ipoptions; 1030 static int srctun = 0; 1031 struct mbuf *mm; 1032 int s; 1033 #ifdef RSVP_ISI 1034 struct vif *vifp; 1035 vifi_t vifi; 1036 #endif /* RSVP_ISI */ 1037 1038 /* 1039 * Clear any in-bound checksum flags for this packet. 1040 */ 1041 m->m_pkthdr.csum_flags = 0; 1042 1043 if (mrtdebug & DEBUG_FORWARD) 1044 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n", 1045 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1046 1047 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1048 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR) { 1049 /* 1050 * Packet arrived via a physical interface or 1051 * an encapuslated tunnel. 1052 */ 1053 } else { 1054 /* 1055 * Packet arrived through a source-route tunnel. 1056 * Source-route tunnels are no longer supported. 1057 */ 1058 if ((srctun++ % 1000) == 0) 1059 log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n", 1060 ntohl(ip->ip_src.s_addr)); 1061 1062 return (1); 1063 } 1064 1065 #ifdef RSVP_ISI 1066 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1067 if (ip->ip_ttl < 255) 1068 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1069 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1070 vifp = viftable + vifi; 1071 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n", 1072 ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi, 1073 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1074 vifp->v_ifp->if_xname); 1075 } 1076 return (ip_mdq(m, ifp, (struct mfc *)0, vifi)); 1077 } 1078 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1079 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n", 1080 ntohl(ip->ip_src), ntohl(ip->ip_dst)); 1081 } 1082 #endif /* RSVP_ISI */ 1083 1084 /* 1085 * Don't forward a packet with time-to-live of zero or one, 1086 * or a packet destined to a local-only group. 1087 */ 1088 if (ip->ip_ttl <= 1 || 1089 IN_LOCAL_GROUP(ip->ip_dst.s_addr)) 1090 return (0); 1091 1092 /* 1093 * Determine forwarding vifs from the forwarding cache table 1094 */ 1095 s = splsoftnet(); 1096 MFCFIND(ip->ip_src, ip->ip_dst, rt); 1097 1098 /* Entry exists, so forward if necessary */ 1099 if (rt != 0) { 1100 splx(s); 1101 #ifdef RSVP_ISI 1102 return (ip_mdq(m, ifp, rt, -1)); 1103 #else 1104 return (ip_mdq(m, ifp, rt)); 1105 #endif /* RSVP_ISI */ 1106 } else { 1107 /* 1108 * If we don't have a route for packet's origin, 1109 * Make a copy of the packet & 1110 * send message to routing daemon 1111 */ 1112 1113 struct mbuf *mb0; 1114 struct rtdetq *rte; 1115 u_int32_t hash; 1116 int hlen = ip->ip_hl << 2; 1117 #ifdef UPCALL_TIMING 1118 struct timeval tp; 1119 1120 microtime(&tp); 1121 #endif /* UPCALL_TIMING */ 1122 1123 mrtstat.mrts_no_route++; 1124 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1125 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1126 ntohl(ip->ip_src.s_addr), 1127 ntohl(ip->ip_dst.s_addr)); 1128 1129 /* 1130 * Allocate mbufs early so that we don't do extra work if we are 1131 * just going to fail anyway. Make sure to pullup the header so 1132 * that other people can't step on it. 1133 */ 1134 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, M_NOWAIT); 1135 if (rte == 0) { 1136 splx(s); 1137 return (ENOBUFS); 1138 } 1139 mb0 = m_copy(m, 0, M_COPYALL); 1140 M_PULLUP(mb0, hlen); 1141 if (mb0 == 0) { 1142 free(rte, M_MRTABLE); 1143 splx(s); 1144 return (ENOBUFS); 1145 } 1146 1147 /* is there an upcall waiting for this packet? */ 1148 hash = MFCHASH(ip->ip_src, ip->ip_dst); 1149 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1150 if (in_hosteq(ip->ip_src, rt->mfc_origin) && 1151 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) && 1152 rt->mfc_stall != 0) 1153 break; 1154 } 1155 1156 if (rt == 0) { 1157 int i; 1158 struct igmpmsg *im; 1159 1160 /* no upcall, so make a new entry */ 1161 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1162 if (rt == 0) { 1163 free(rte, M_MRTABLE); 1164 m_freem(mb0); 1165 splx(s); 1166 return (ENOBUFS); 1167 } 1168 /* Make a copy of the header to send to the user level process */ 1169 mm = m_copy(m, 0, hlen); 1170 M_PULLUP(mm, hlen); 1171 if (mm == 0) { 1172 free(rte, M_MRTABLE); 1173 m_freem(mb0); 1174 free(rt, M_MRTABLE); 1175 splx(s); 1176 return (ENOBUFS); 1177 } 1178 1179 /* 1180 * Send message to routing daemon to install 1181 * a route into the kernel table 1182 */ 1183 sin.sin_addr = ip->ip_src; 1184 1185 im = mtod(mm, struct igmpmsg *); 1186 im->im_msgtype = IGMPMSG_NOCACHE; 1187 im->im_mbz = 0; 1188 1189 mrtstat.mrts_upcalls++; 1190 1191 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1192 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1193 ++mrtstat.mrts_upq_sockfull; 1194 free(rte, M_MRTABLE); 1195 m_freem(mb0); 1196 free(rt, M_MRTABLE); 1197 splx(s); 1198 return (ENOBUFS); 1199 } 1200 1201 /* insert new entry at head of hash chain */ 1202 rt->mfc_origin = ip->ip_src; 1203 rt->mfc_mcastgrp = ip->ip_dst; 1204 rt->mfc_pkt_cnt = 0; 1205 rt->mfc_byte_cnt = 0; 1206 rt->mfc_wrong_if = 0; 1207 rt->mfc_expire = UPCALL_EXPIRE; 1208 nexpire[hash]++; 1209 for (i = 0; i < numvifs; i++) 1210 rt->mfc_ttls[i] = 0; 1211 rt->mfc_parent = -1; 1212 1213 /* link into table */ 1214 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1215 /* Add this entry to the end of the queue */ 1216 rt->mfc_stall = rte; 1217 } else { 1218 /* determine if q has overflowed */ 1219 struct rtdetq **p; 1220 int npkts = 0; 1221 1222 for (p = &rt->mfc_stall; *p != 0; p = &(*p)->next) 1223 if (++npkts > MAX_UPQ) { 1224 mrtstat.mrts_upq_ovflw++; 1225 free(rte, M_MRTABLE); 1226 m_freem(mb0); 1227 splx(s); 1228 return (0); 1229 } 1230 1231 /* Add this entry to the end of the queue */ 1232 *p = rte; 1233 } 1234 1235 rte->next = 0; 1236 rte->m = mb0; 1237 rte->ifp = ifp; 1238 #ifdef UPCALL_TIMING 1239 rte->t = tp; 1240 #endif /* UPCALL_TIMING */ 1241 1242 1243 splx(s); 1244 1245 return (0); 1246 } 1247 } 1248 1249 1250 /*ARGSUSED*/ 1251 static void 1252 expire_upcalls(v) 1253 void *v; 1254 { 1255 int i; 1256 int s; 1257 1258 s = splsoftnet(); 1259 1260 for (i = 0; i < MFCTBLSIZ; i++) { 1261 struct mfc *rt, *nrt; 1262 1263 if (nexpire[i] == 0) 1264 continue; 1265 1266 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 1267 nrt = LIST_NEXT(rt, mfc_hash); 1268 1269 if (rt->mfc_expire == 0 || 1270 --rt->mfc_expire > 0) 1271 continue; 1272 nexpire[i]--; 1273 1274 ++mrtstat.mrts_cache_cleanups; 1275 if (mrtdebug & DEBUG_EXPIRE) 1276 log(LOG_DEBUG, 1277 "expire_upcalls: expiring (%x %x)\n", 1278 ntohl(rt->mfc_origin.s_addr), 1279 ntohl(rt->mfc_mcastgrp.s_addr)); 1280 1281 expire_mfc(rt); 1282 } 1283 } 1284 1285 splx(s); 1286 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1287 expire_upcalls, NULL); 1288 } 1289 1290 /* 1291 * Packet forwarding routine once entry in the cache is made 1292 */ 1293 static int 1294 #ifdef RSVP_ISI 1295 ip_mdq(m, ifp, rt, xmt_vif) 1296 #else 1297 ip_mdq(m, ifp, rt) 1298 #endif /* RSVP_ISI */ 1299 struct mbuf *m; 1300 struct ifnet *ifp; 1301 struct mfc *rt; 1302 #ifdef RSVP_ISI 1303 vifi_t xmt_vif; 1304 #endif /* RSVP_ISI */ 1305 { 1306 struct ip *ip = mtod(m, struct ip *); 1307 vifi_t vifi; 1308 struct vif *vifp; 1309 int plen = ntohs(ip->ip_len); 1310 1311 /* 1312 * Macro to send packet on vif. Since RSVP packets don't get counted on 1313 * input, they shouldn't get counted on output, so statistics keeping is 1314 * separate. 1315 */ 1316 #define MC_SEND(ip,vifp,m) { \ 1317 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1318 encap_send((ip), (vifp), (m)); \ 1319 else \ 1320 phyint_send((ip), (vifp), (m)); \ 1321 } 1322 1323 #ifdef RSVP_ISI 1324 /* 1325 * If xmt_vif is not -1, send on only the requested vif. 1326 * 1327 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs. 1328 */ 1329 if (xmt_vif < numvifs) { 1330 MC_SEND(ip, viftable + xmt_vif, m); 1331 return (1); 1332 } 1333 #endif /* RSVP_ISI */ 1334 1335 /* 1336 * Don't forward if it didn't arrive from the parent vif for its origin. 1337 */ 1338 vifi = rt->mfc_parent; 1339 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1340 /* came in the wrong interface */ 1341 if (mrtdebug & DEBUG_FORWARD) 1342 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1343 ifp, vifi, viftable[vifi].v_ifp); 1344 ++mrtstat.mrts_wrong_if; 1345 ++rt->mfc_wrong_if; 1346 /* 1347 * If we are doing PIM assert processing, and we are forwarding 1348 * packets on this interface, and it is a broadcast medium 1349 * interface (and not a tunnel), send a message to the routing daemon. 1350 */ 1351 if (pim_assert && rt->mfc_ttls[vifi] && 1352 (ifp->if_flags & IFF_BROADCAST) && 1353 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1354 struct mbuf *mm; 1355 struct igmpmsg *im; 1356 int hlen = ip->ip_hl << 2; 1357 struct timeval now; 1358 u_int32_t delta; 1359 1360 microtime(&now); 1361 1362 TV_DELTA(rt->mfc_last_assert, now, delta); 1363 1364 if (delta > ASSERT_MSG_TIME) { 1365 mm = m_copy(m, 0, hlen); 1366 M_PULLUP(mm, hlen); 1367 if (mm == 0) { 1368 return (ENOBUFS); 1369 } 1370 1371 rt->mfc_last_assert = now; 1372 1373 im = mtod(mm, struct igmpmsg *); 1374 im->im_msgtype = IGMPMSG_WRONGVIF; 1375 im->im_mbz = 0; 1376 im->im_vif = vifi; 1377 1378 sin.sin_addr = im->im_src; 1379 1380 socket_send(ip_mrouter, mm, &sin); 1381 } 1382 } 1383 return (0); 1384 } 1385 1386 /* If I sourced this packet, it counts as output, else it was input. */ 1387 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) { 1388 viftable[vifi].v_pkt_out++; 1389 viftable[vifi].v_bytes_out += plen; 1390 } else { 1391 viftable[vifi].v_pkt_in++; 1392 viftable[vifi].v_bytes_in += plen; 1393 } 1394 rt->mfc_pkt_cnt++; 1395 rt->mfc_byte_cnt += plen; 1396 1397 /* 1398 * For each vif, decide if a copy of the packet should be forwarded. 1399 * Forward if: 1400 * - the ttl exceeds the vif's threshold 1401 * - there are group members downstream on interface 1402 */ 1403 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1404 if ((rt->mfc_ttls[vifi] > 0) && 1405 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1406 vifp->v_pkt_out++; 1407 vifp->v_bytes_out += plen; 1408 MC_SEND(ip, vifp, m); 1409 } 1410 1411 return (0); 1412 } 1413 1414 #ifdef RSVP_ISI 1415 /* 1416 * check if a vif number is legal/ok. This is used by ip_output, to export 1417 * numvifs there, 1418 */ 1419 int 1420 legal_vif_num(vif) 1421 int vif; 1422 { 1423 if (vif >= 0 && vif < numvifs) 1424 return (1); 1425 else 1426 return (0); 1427 } 1428 #endif /* RSVP_ISI */ 1429 1430 static void 1431 phyint_send(ip, vifp, m) 1432 struct ip *ip; 1433 struct vif *vifp; 1434 struct mbuf *m; 1435 { 1436 struct mbuf *mb_copy; 1437 int hlen = ip->ip_hl << 2; 1438 1439 /* 1440 * Make a new reference to the packet; make sure that 1441 * the IP header is actually copied, not just referenced, 1442 * so that ip_output() only scribbles on the copy. 1443 */ 1444 mb_copy = m_copy(m, 0, M_COPYALL); 1445 M_PULLUP(mb_copy, hlen); 1446 if (mb_copy == 0) 1447 return; 1448 1449 if (vifp->v_rate_limit <= 0) 1450 tbf_send_packet(vifp, mb_copy); 1451 else 1452 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), 1453 ntohs(ip->ip_len)); 1454 } 1455 1456 static void 1457 encap_send(ip, vifp, m) 1458 struct ip *ip; 1459 struct vif *vifp; 1460 struct mbuf *m; 1461 { 1462 struct mbuf *mb_copy; 1463 struct ip *ip_copy; 1464 int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr); 1465 1466 /* 1467 * copy the old packet & pullup it's IP header into the 1468 * new mbuf so we can modify it. Try to fill the new 1469 * mbuf since if we don't the ethernet driver will. 1470 */ 1471 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA); 1472 if (mb_copy == 0) 1473 return; 1474 mb_copy->m_data += max_linkhdr; 1475 mb_copy->m_pkthdr.len = len; 1476 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1477 1478 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == 0) { 1479 m_freem(mb_copy); 1480 return; 1481 } 1482 i = MHLEN - max_linkhdr; 1483 if (i > len) 1484 i = len; 1485 mb_copy = m_pullup(mb_copy, i); 1486 if (mb_copy == 0) 1487 return; 1488 1489 /* 1490 * fill in the encapsulating IP header. 1491 */ 1492 ip_copy = mtod(mb_copy, struct ip *); 1493 *ip_copy = multicast_encap_iphdr; 1494 ip_copy->ip_id = htons(ip_id++); 1495 ip_copy->ip_len = htons(len); 1496 ip_copy->ip_src = vifp->v_lcl_addr; 1497 ip_copy->ip_dst = vifp->v_rmt_addr; 1498 1499 /* 1500 * turn the encapsulated IP header back into a valid one. 1501 */ 1502 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1503 --ip->ip_ttl; 1504 ip->ip_sum = 0; 1505 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1506 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1507 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1508 1509 if (vifp->v_rate_limit <= 0) 1510 tbf_send_packet(vifp, mb_copy); 1511 else 1512 tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len)); 1513 } 1514 1515 /* 1516 * De-encapsulate a packet and feed it back through ip input. 1517 */ 1518 static void 1519 #if __STDC__ 1520 vif_input(struct mbuf *m, ...) 1521 #else 1522 vif_input(m, va_alist) 1523 struct mbuf *m; 1524 va_dcl 1525 #endif 1526 { 1527 int off, proto; 1528 va_list ap; 1529 struct ip *ip; 1530 struct vif *vifp; 1531 int s; 1532 struct ifqueue *ifq; 1533 1534 va_start(ap, m); 1535 off = va_arg(ap, int); 1536 proto = va_arg(ap, int); 1537 va_end(ap); 1538 1539 vifp = (struct vif *)encap_getarg(m); 1540 if (!vifp || proto != AF_INET) { 1541 m_freem(m); 1542 mrtstat.mrts_bad_tunnel++; 1543 return; 1544 } 1545 1546 ip = mtod(m, struct ip *); 1547 1548 m_adj(m, off); 1549 m->m_pkthdr.rcvif = vifp->v_ifp; 1550 ifq = &ipintrq; 1551 s = splnet(); 1552 if (IF_QFULL(ifq)) { 1553 IF_DROP(ifq); 1554 m_freem(m); 1555 } else { 1556 IF_ENQUEUE(ifq, m); 1557 /* 1558 * normally we would need a "schednetisr(NETISR_IP)" 1559 * here but we were called by ip_input and it is going 1560 * to loop back & try to dequeue the packet we just 1561 * queued as soon as we return so we avoid the 1562 * unnecessary software interrrupt. 1563 */ 1564 } 1565 splx(s); 1566 } 1567 1568 /* 1569 * Check if the packet should be grabbed by us. 1570 */ 1571 static int 1572 vif_encapcheck(m, off, proto, arg) 1573 const struct mbuf *m; 1574 int off; 1575 int proto; 1576 void *arg; 1577 { 1578 struct vif *vifp; 1579 struct ip ip; 1580 1581 #ifdef DIAGNOSTIC 1582 if (!arg || proto != IPPROTO_IPV4) 1583 panic("unexpected arg in vif_encapcheck"); 1584 #endif 1585 1586 /* 1587 * do not grab the packet if it's not to a multicast destination or if 1588 * we don't have an encapsulating tunnel with the source. 1589 * Note: This code assumes that the remote site IP address 1590 * uniquely identifies the tunnel (i.e., that this site has 1591 * at most one tunnel with the remote site). 1592 */ 1593 1594 /* LINTED const cast */ 1595 m_copydata((struct mbuf *)m, off, sizeof(ip), (caddr_t)&ip); 1596 if (!IN_MULTICAST(ip.ip_dst.s_addr)) 1597 return 0; 1598 1599 /* LINTED const cast */ 1600 m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip); 1601 if (!in_hosteq(ip.ip_src, last_encap_src)) { 1602 vifp = (struct vif *)arg; 1603 if (vifp->v_flags & VIFF_TUNNEL && 1604 in_hosteq(vifp->v_rmt_addr, ip.ip_src)) 1605 ; 1606 else 1607 return 0; 1608 last_encap_vif = vifp; 1609 last_encap_src = ip.ip_src; 1610 } else 1611 vifp = last_encap_vif; 1612 1613 /* 32bit match, since we have checked ip_src only */ 1614 return 32; 1615 } 1616 1617 /* 1618 * Token bucket filter module 1619 */ 1620 static void 1621 tbf_control(vifp, m, ip, len) 1622 struct vif *vifp; 1623 struct mbuf *m; 1624 struct ip *ip; 1625 u_int32_t len; 1626 { 1627 1628 if (len > MAX_BKT_SIZE) { 1629 /* drop if packet is too large */ 1630 mrtstat.mrts_pkt2large++; 1631 m_freem(m); 1632 return; 1633 } 1634 1635 tbf_update_tokens(vifp); 1636 1637 /* 1638 * If there are enough tokens, and the queue is empty, send this packet 1639 * out immediately. Otherwise, try to insert it on this vif's queue. 1640 */ 1641 if (vifp->tbf_q_len == 0) { 1642 if (len <= vifp->tbf_n_tok) { 1643 vifp->tbf_n_tok -= len; 1644 tbf_send_packet(vifp, m); 1645 } else { 1646 /* queue packet and timeout till later */ 1647 tbf_queue(vifp, m); 1648 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 1649 tbf_reprocess_q, vifp); 1650 } 1651 } else { 1652 if (vifp->tbf_q_len >= vifp->tbf_max_q_len && 1653 !tbf_dq_sel(vifp, ip)) { 1654 /* queue length too much, and couldn't make room */ 1655 mrtstat.mrts_q_overflow++; 1656 m_freem(m); 1657 } else { 1658 /* queue length low enough, or made room */ 1659 tbf_queue(vifp, m); 1660 tbf_process_q(vifp); 1661 } 1662 } 1663 } 1664 1665 /* 1666 * adds a packet to the queue at the interface 1667 */ 1668 static void 1669 tbf_queue(vifp, m) 1670 struct vif *vifp; 1671 struct mbuf *m; 1672 { 1673 int s = splsoftnet(); 1674 1675 /* insert at tail */ 1676 *vifp->tbf_t = m; 1677 vifp->tbf_t = &m->m_nextpkt; 1678 vifp->tbf_q_len++; 1679 1680 splx(s); 1681 } 1682 1683 1684 /* 1685 * processes the queue at the interface 1686 */ 1687 static void 1688 tbf_process_q(vifp) 1689 struct vif *vifp; 1690 { 1691 struct mbuf *m; 1692 int len; 1693 int s = splsoftnet(); 1694 1695 /* 1696 * Loop through the queue at the interface and send as many packets 1697 * as possible. 1698 */ 1699 for (m = vifp->tbf_q; 1700 m != 0; 1701 m = vifp->tbf_q) { 1702 len = ntohs(mtod(m, struct ip *)->ip_len); 1703 1704 /* determine if the packet can be sent */ 1705 if (len <= vifp->tbf_n_tok) { 1706 /* if so, 1707 * reduce no of tokens, dequeue the packet, 1708 * send the packet. 1709 */ 1710 if ((vifp->tbf_q = m->m_nextpkt) == 0) 1711 vifp->tbf_t = &vifp->tbf_q; 1712 --vifp->tbf_q_len; 1713 1714 m->m_nextpkt = 0; 1715 vifp->tbf_n_tok -= len; 1716 tbf_send_packet(vifp, m); 1717 } else 1718 break; 1719 } 1720 splx(s); 1721 } 1722 1723 static void 1724 tbf_reprocess_q(arg) 1725 void *arg; 1726 { 1727 struct vif *vifp = arg; 1728 1729 if (ip_mrouter == 0) 1730 return; 1731 1732 tbf_update_tokens(vifp); 1733 tbf_process_q(vifp); 1734 1735 if (vifp->tbf_q_len != 0) 1736 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 1737 tbf_reprocess_q, vifp); 1738 } 1739 1740 /* function that will selectively discard a member of the queue 1741 * based on the precedence value and the priority 1742 */ 1743 static int 1744 tbf_dq_sel(vifp, ip) 1745 struct vif *vifp; 1746 struct ip *ip; 1747 { 1748 u_int p; 1749 struct mbuf **mp, *m; 1750 int s = splsoftnet(); 1751 1752 p = priority(vifp, ip); 1753 1754 for (mp = &vifp->tbf_q, m = *mp; 1755 m != 0; 1756 mp = &m->m_nextpkt, m = *mp) { 1757 if (p > priority(vifp, mtod(m, struct ip *))) { 1758 if ((*mp = m->m_nextpkt) == 0) 1759 vifp->tbf_t = mp; 1760 --vifp->tbf_q_len; 1761 1762 m_freem(m); 1763 mrtstat.mrts_drop_sel++; 1764 splx(s); 1765 return (1); 1766 } 1767 } 1768 splx(s); 1769 return (0); 1770 } 1771 1772 static void 1773 tbf_send_packet(vifp, m) 1774 struct vif *vifp; 1775 struct mbuf *m; 1776 { 1777 int error; 1778 int s = splsoftnet(); 1779 1780 if (vifp->v_flags & VIFF_TUNNEL) { 1781 /* If tunnel options */ 1782 #ifdef IPSEC 1783 /* Don't lookup socket in forwading case */ 1784 (void)ipsec_setsocket(m, NULL); 1785 #endif 1786 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1787 IP_FORWARDING, (struct ip_moptions *)0); 1788 } else { 1789 /* if physical interface option, extract the options and then send */ 1790 struct ip_moptions imo; 1791 1792 imo.imo_multicast_ifp = vifp->v_ifp; 1793 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1794 imo.imo_multicast_loop = 1; 1795 #ifdef RSVP_ISI 1796 imo.imo_multicast_vif = -1; 1797 #endif 1798 1799 #ifdef IPSEC 1800 /* Don't lookup socket in forwading case */ 1801 (void)ipsec_setsocket(m, NULL); 1802 #endif 1803 error = ip_output(m, (struct mbuf *)0, (struct route *)0, 1804 IP_FORWARDING|IP_MULTICASTOPTS, &imo); 1805 1806 if (mrtdebug & DEBUG_XMIT) 1807 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n", 1808 (long)(vifp-viftable), error); 1809 } 1810 splx(s); 1811 } 1812 1813 /* determine the current time and then 1814 * the elapsed time (between the last time and time now) 1815 * in milliseconds & update the no. of tokens in the bucket 1816 */ 1817 static void 1818 tbf_update_tokens(vifp) 1819 struct vif *vifp; 1820 { 1821 struct timeval tp; 1822 u_int32_t tm; 1823 int s = splsoftnet(); 1824 1825 microtime(&tp); 1826 1827 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm); 1828 1829 /* 1830 * This formula is actually 1831 * "time in seconds" * "bytes/second". 1832 * 1833 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1834 * 1835 * The (1000/1024) was introduced in add_vif to optimize 1836 * this divide into a shift. 1837 */ 1838 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192; 1839 vifp->tbf_last_pkt_t = tp; 1840 1841 if (vifp->tbf_n_tok > MAX_BKT_SIZE) 1842 vifp->tbf_n_tok = MAX_BKT_SIZE; 1843 1844 splx(s); 1845 } 1846 1847 static int 1848 priority(vifp, ip) 1849 struct vif *vifp; 1850 struct ip *ip; 1851 { 1852 int prio; 1853 1854 /* temporary hack; may add general packet classifier some day */ 1855 1856 /* 1857 * The UDP port space is divided up into four priority ranges: 1858 * [0, 16384) : unclassified - lowest priority 1859 * [16384, 32768) : audio - highest priority 1860 * [32768, 49152) : whiteboard - medium priority 1861 * [49152, 65536) : video - low priority 1862 */ 1863 if (ip->ip_p == IPPROTO_UDP) { 1864 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1865 1866 switch (ntohs(udp->uh_dport) & 0xc000) { 1867 case 0x4000: 1868 prio = 70; 1869 break; 1870 case 0x8000: 1871 prio = 60; 1872 break; 1873 case 0xc000: 1874 prio = 55; 1875 break; 1876 default: 1877 prio = 50; 1878 break; 1879 } 1880 1881 if (tbfdebug > 1) 1882 log(LOG_DEBUG, "port %x prio %d\n", ntohs(udp->uh_dport), prio); 1883 } else 1884 prio = 50; 1885 1886 1887 return (prio); 1888 } 1889 1890 /* 1891 * End of token bucket filter modifications 1892 */ 1893 1894 #ifdef RSVP_ISI 1895 1896 int 1897 ip_rsvp_vif_init(so, m) 1898 struct socket *so; 1899 struct mbuf *m; 1900 { 1901 int i; 1902 int s; 1903 1904 if (rsvpdebug) 1905 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 1906 so->so_type, so->so_proto->pr_protocol); 1907 1908 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1909 return (EOPNOTSUPP); 1910 1911 /* Check mbuf. */ 1912 if (m == 0 || m->m_len != sizeof(int)) { 1913 return (EINVAL); 1914 } 1915 i = *(mtod(m, int *)); 1916 1917 if (rsvpdebug) 1918 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on); 1919 1920 s = splsoftnet(); 1921 1922 /* Check vif. */ 1923 if (!legal_vif_num(i)) { 1924 splx(s); 1925 return (EADDRNOTAVAIL); 1926 } 1927 1928 /* Check if socket is available. */ 1929 if (viftable[i].v_rsvpd != 0) { 1930 splx(s); 1931 return (EADDRINUSE); 1932 } 1933 1934 viftable[i].v_rsvpd = so; 1935 /* This may seem silly, but we need to be sure we don't over-increment 1936 * the RSVP counter, in case something slips up. 1937 */ 1938 if (!viftable[i].v_rsvp_on) { 1939 viftable[i].v_rsvp_on = 1; 1940 rsvp_on++; 1941 } 1942 1943 splx(s); 1944 return (0); 1945 } 1946 1947 int 1948 ip_rsvp_vif_done(so, m) 1949 struct socket *so; 1950 struct mbuf *m; 1951 { 1952 int i; 1953 int s; 1954 1955 if (rsvpdebug) 1956 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 1957 so->so_type, so->so_proto->pr_protocol); 1958 1959 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1960 return (EOPNOTSUPP); 1961 1962 /* Check mbuf. */ 1963 if (m == 0 || m->m_len != sizeof(int)) { 1964 return (EINVAL); 1965 } 1966 i = *(mtod(m, int *)); 1967 1968 s = splsoftnet(); 1969 1970 /* Check vif. */ 1971 if (!legal_vif_num(i)) { 1972 splx(s); 1973 return (EADDRNOTAVAIL); 1974 } 1975 1976 if (rsvpdebug) 1977 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n", 1978 viftable[i].v_rsvpd, so); 1979 1980 viftable[i].v_rsvpd = 0; 1981 /* This may seem silly, but we need to be sure we don't over-decrement 1982 * the RSVP counter, in case something slips up. 1983 */ 1984 if (viftable[i].v_rsvp_on) { 1985 viftable[i].v_rsvp_on = 0; 1986 rsvp_on--; 1987 } 1988 1989 splx(s); 1990 return (0); 1991 } 1992 1993 void 1994 ip_rsvp_force_done(so) 1995 struct socket *so; 1996 { 1997 int vifi; 1998 int s; 1999 2000 /* Don't bother if it is not the right type of socket. */ 2001 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2002 return; 2003 2004 s = splsoftnet(); 2005 2006 /* The socket may be attached to more than one vif...this 2007 * is perfectly legal. 2008 */ 2009 for (vifi = 0; vifi < numvifs; vifi++) { 2010 if (viftable[vifi].v_rsvpd == so) { 2011 viftable[vifi].v_rsvpd = 0; 2012 /* This may seem silly, but we need to be sure we don't 2013 * over-decrement the RSVP counter, in case something slips up. 2014 */ 2015 if (viftable[vifi].v_rsvp_on) { 2016 viftable[vifi].v_rsvp_on = 0; 2017 rsvp_on--; 2018 } 2019 } 2020 } 2021 2022 splx(s); 2023 return; 2024 } 2025 2026 void 2027 rsvp_input(m, ifp) 2028 struct mbuf *m; 2029 struct ifnet *ifp; 2030 { 2031 int vifi; 2032 struct ip *ip = mtod(m, struct ip *); 2033 static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET }; 2034 int s; 2035 2036 if (rsvpdebug) 2037 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2038 2039 /* Can still get packets with rsvp_on = 0 if there is a local member 2040 * of the group to which the RSVP packet is addressed. But in this 2041 * case we want to throw the packet away. 2042 */ 2043 if (!rsvp_on) { 2044 m_freem(m); 2045 return; 2046 } 2047 2048 /* If the old-style non-vif-associated socket is set, then use 2049 * it and ignore the new ones. 2050 */ 2051 if (ip_rsvpd != 0) { 2052 if (rsvpdebug) 2053 printf("rsvp_input: Sending packet up old-style socket\n"); 2054 rip_input(m); /*XXX*/ 2055 return; 2056 } 2057 2058 s = splsoftnet(); 2059 2060 if (rsvpdebug) 2061 printf("rsvp_input: check vifs\n"); 2062 2063 /* Find which vif the packet arrived on. */ 2064 for (vifi = 0; vifi < numvifs; vifi++) { 2065 if (viftable[vifi].v_ifp == ifp) 2066 break; 2067 } 2068 2069 if (vifi == numvifs) { 2070 /* Can't find vif packet arrived on. Drop packet. */ 2071 if (rsvpdebug) 2072 printf("rsvp_input: Can't find vif for packet...dropping it.\n"); 2073 m_freem(m); 2074 splx(s); 2075 return; 2076 } 2077 2078 if (rsvpdebug) 2079 printf("rsvp_input: check socket\n"); 2080 2081 if (viftable[vifi].v_rsvpd == 0) { 2082 /* drop packet, since there is no specific socket for this 2083 * interface */ 2084 if (rsvpdebug) 2085 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2086 m_freem(m); 2087 splx(s); 2088 return; 2089 } 2090 2091 rsvp_src.sin_addr = ip->ip_src; 2092 2093 if (rsvpdebug && m) 2094 printf("rsvp_input: m->m_len = %d, sbspace() = %d\n", 2095 m->m_len,sbspace(&viftable[vifi].v_rsvpd->so_rcv)); 2096 2097 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2098 if (rsvpdebug) 2099 printf("rsvp_input: Failed to append to socket\n"); 2100 else 2101 if (rsvpdebug) 2102 printf("rsvp_input: send packet up\n"); 2103 2104 splx(s); 2105 } 2106 #endif /* RSVP_ISI */ 2107