1 /* $NetBSD: ip_mroute.c,v 1.66 2003/02/26 06:31:15 matt 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 1.66 2003/02/26 06:31:15 matt 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 MCLAIM(*m, so->so_mowner); 351 352 switch (optname) { 353 case MRT_VERSION: 354 error = get_version(*m); 355 break; 356 case MRT_ASSERT: 357 error = get_assert(*m); 358 break; 359 default: 360 error = ENOPROTOOPT; 361 break; 362 } 363 364 if (error) 365 m_free(*m); 366 } 367 368 return (error); 369 } 370 371 /* 372 * Handle ioctl commands to obtain information from the cache 373 */ 374 int 375 mrt_ioctl(so, cmd, data) 376 struct socket *so; 377 u_long cmd; 378 caddr_t data; 379 { 380 int error; 381 382 if (so != ip_mrouter) 383 error = EINVAL; 384 else 385 switch (cmd) { 386 case SIOCGETVIFCNT: 387 error = get_vif_cnt((struct sioc_vif_req *)data); 388 break; 389 case SIOCGETSGCNT: 390 error = get_sg_cnt((struct sioc_sg_req *)data); 391 break; 392 default: 393 error = EINVAL; 394 break; 395 } 396 397 return (error); 398 } 399 400 /* 401 * returns the packet, byte, rpf-failure count for the source group provided 402 */ 403 static int 404 get_sg_cnt(req) 405 struct sioc_sg_req *req; 406 { 407 struct mfc *rt; 408 int s; 409 410 s = splsoftnet(); 411 MFCFIND(req->src, req->grp, rt); 412 splx(s); 413 if (rt != 0) { 414 req->pktcnt = rt->mfc_pkt_cnt; 415 req->bytecnt = rt->mfc_byte_cnt; 416 req->wrong_if = rt->mfc_wrong_if; 417 } else 418 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 419 420 return (0); 421 } 422 423 /* 424 * returns the input and output packet and byte counts on the vif provided 425 */ 426 static int 427 get_vif_cnt(req) 428 struct sioc_vif_req *req; 429 { 430 vifi_t vifi = req->vifi; 431 432 if (vifi >= numvifs) 433 return (EINVAL); 434 435 req->icount = viftable[vifi].v_pkt_in; 436 req->ocount = viftable[vifi].v_pkt_out; 437 req->ibytes = viftable[vifi].v_bytes_in; 438 req->obytes = viftable[vifi].v_bytes_out; 439 440 return (0); 441 } 442 443 /* 444 * Enable multicast routing 445 */ 446 static int 447 ip_mrouter_init(so, m) 448 struct socket *so; 449 struct mbuf *m; 450 { 451 int *v; 452 453 if (mrtdebug) 454 log(LOG_DEBUG, 455 "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 456 so->so_type, so->so_proto->pr_protocol); 457 458 if (so->so_type != SOCK_RAW || 459 so->so_proto->pr_protocol != IPPROTO_IGMP) 460 return (EOPNOTSUPP); 461 462 if (m == 0 || m->m_len < sizeof(int)) 463 return (EINVAL); 464 465 v = mtod(m, int *); 466 if (*v != 1) 467 return (EINVAL); 468 469 if (ip_mrouter != 0) 470 return (EADDRINUSE); 471 472 ip_mrouter = so; 473 474 mfchashtbl = 475 hashinit(MFCTBLSIZ, HASH_LIST, M_MRTABLE, M_WAITOK, &mfchash); 476 bzero((caddr_t)nexpire, sizeof(nexpire)); 477 478 pim_assert = 0; 479 480 callout_init(&expire_upcalls_ch); 481 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 482 expire_upcalls, NULL); 483 484 if (mrtdebug) 485 log(LOG_DEBUG, "ip_mrouter_init\n"); 486 487 return (0); 488 } 489 490 /* 491 * Disable multicast routing 492 */ 493 int 494 ip_mrouter_done() 495 { 496 vifi_t vifi; 497 struct vif *vifp; 498 int i; 499 int s; 500 501 s = splsoftnet(); 502 503 /* Clear out all the vifs currently in use. */ 504 for (vifi = 0; vifi < numvifs; vifi++) { 505 vifp = &viftable[vifi]; 506 if (!in_nullhost(vifp->v_lcl_addr)) 507 reset_vif(vifp); 508 } 509 510 numvifs = 0; 511 pim_assert = 0; 512 513 callout_stop(&expire_upcalls_ch); 514 515 /* 516 * Free all multicast forwarding cache entries. 517 */ 518 for (i = 0; i < MFCTBLSIZ; i++) { 519 struct mfc *rt, *nrt; 520 521 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 522 nrt = LIST_NEXT(rt, mfc_hash); 523 524 expire_mfc(rt); 525 } 526 } 527 528 free(mfchashtbl, M_MRTABLE); 529 mfchashtbl = 0; 530 531 /* Reset de-encapsulation cache. */ 532 533 ip_mrouter = 0; 534 535 splx(s); 536 537 if (mrtdebug) 538 log(LOG_DEBUG, "ip_mrouter_done\n"); 539 540 return (0); 541 } 542 543 static int 544 get_version(m) 545 struct mbuf *m; 546 { 547 int *v = mtod(m, int *); 548 549 *v = 0x0305; /* XXX !!!! */ 550 m->m_len = sizeof(int); 551 return (0); 552 } 553 554 /* 555 * Set PIM assert processing global 556 */ 557 static int 558 set_assert(m) 559 struct mbuf *m; 560 { 561 int *i; 562 563 if (m == 0 || m->m_len < sizeof(int)) 564 return (EINVAL); 565 566 i = mtod(m, int *); 567 pim_assert = !!*i; 568 return (0); 569 } 570 571 /* 572 * Get PIM assert processing global 573 */ 574 static int 575 get_assert(m) 576 struct mbuf *m; 577 { 578 int *i = mtod(m, int *); 579 580 *i = pim_assert; 581 m->m_len = sizeof(int); 582 return (0); 583 } 584 585 static struct sockaddr_in sin = { sizeof(sin), AF_INET }; 586 587 /* 588 * Add a vif to the vif table 589 */ 590 static int 591 add_vif(m) 592 struct mbuf *m; 593 { 594 struct vifctl *vifcp; 595 struct vif *vifp; 596 struct ifaddr *ifa; 597 struct ifnet *ifp; 598 struct ifreq ifr; 599 int error, s; 600 601 if (m == 0 || m->m_len < sizeof(struct vifctl)) 602 return (EINVAL); 603 604 vifcp = mtod(m, struct vifctl *); 605 if (vifcp->vifc_vifi >= MAXVIFS) 606 return (EINVAL); 607 608 vifp = &viftable[vifcp->vifc_vifi]; 609 if (!in_nullhost(vifp->v_lcl_addr)) 610 return (EADDRINUSE); 611 612 /* Find the interface with an address in AF_INET family. */ 613 sin.sin_addr = vifcp->vifc_lcl_addr; 614 ifa = ifa_ifwithaddr(sintosa(&sin)); 615 if (ifa == 0) 616 return (EADDRNOTAVAIL); 617 618 if (vifcp->vifc_flags & VIFF_TUNNEL) { 619 if (vifcp->vifc_flags & VIFF_SRCRT) { 620 log(LOG_ERR, "Source routed tunnels not supported\n"); 621 return (EOPNOTSUPP); 622 } 623 624 /* attach this vif to decapsulator dispatch table */ 625 vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4, 626 vif_encapcheck, &vif_protosw, vifp); 627 if (!vifp->v_encap_cookie) 628 return (EINVAL); 629 630 /* Create a fake encapsulation interface. */ 631 ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK); 632 bzero(ifp, sizeof(*ifp)); 633 sprintf(ifp->if_xname, "mdecap%d", vifcp->vifc_vifi); 634 635 /* Prepare cached route entry. */ 636 bzero(&vifp->v_route, sizeof(vifp->v_route)); 637 } else { 638 /* Use the physical interface associated with the address. */ 639 ifp = ifa->ifa_ifp; 640 641 /* Make sure the interface supports multicast. */ 642 if ((ifp->if_flags & IFF_MULTICAST) == 0) 643 return (EOPNOTSUPP); 644 645 /* Enable promiscuous reception of all IP multicasts. */ 646 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 647 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 648 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr; 649 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr); 650 if (error) 651 return (error); 652 } 653 654 s = splsoftnet(); 655 656 /* Define parameters for the tbf structure. */ 657 vifp->tbf_q = 0; 658 vifp->tbf_t = &vifp->tbf_q; 659 microtime(&vifp->tbf_last_pkt_t); 660 vifp->tbf_n_tok = 0; 661 vifp->tbf_q_len = 0; 662 vifp->tbf_max_q_len = MAXQSIZE; 663 664 vifp->v_flags = vifcp->vifc_flags; 665 vifp->v_threshold = vifcp->vifc_threshold; 666 /* scaling up here allows division by 1024 in critical code */ 667 vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000; 668 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 669 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 670 vifp->v_ifp = ifp; 671 /* Initialize per vif pkt counters. */ 672 vifp->v_pkt_in = 0; 673 vifp->v_pkt_out = 0; 674 vifp->v_bytes_in = 0; 675 vifp->v_bytes_out = 0; 676 677 callout_init(&vifp->v_repq_ch); 678 679 #ifdef RSVP_ISI 680 vifp->v_rsvp_on = 0; 681 vifp->v_rsvpd = 0; 682 #endif /* RSVP_ISI */ 683 684 splx(s); 685 686 /* Adjust numvifs up if the vifi is higher than numvifs. */ 687 if (numvifs <= vifcp->vifc_vifi) 688 numvifs = vifcp->vifc_vifi + 1; 689 690 if (mrtdebug) 691 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 692 vifcp->vifc_vifi, 693 ntohl(vifcp->vifc_lcl_addr.s_addr), 694 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 695 ntohl(vifcp->vifc_rmt_addr.s_addr), 696 vifcp->vifc_threshold, 697 vifcp->vifc_rate_limit); 698 699 return (0); 700 } 701 702 void 703 reset_vif(vifp) 704 struct vif *vifp; 705 { 706 struct mbuf *m, *n; 707 struct ifnet *ifp; 708 struct ifreq ifr; 709 710 callout_stop(&vifp->v_repq_ch); 711 712 /* detach this vif from decapsulator dispatch table */ 713 encap_detach(vifp->v_encap_cookie); 714 vifp->v_encap_cookie = NULL; 715 716 for (m = vifp->tbf_q; m != 0; m = n) { 717 n = m->m_nextpkt; 718 m_freem(m); 719 } 720 721 if (vifp->v_flags & VIFF_TUNNEL) { 722 free(vifp->v_ifp, M_MRTABLE); 723 if (vifp == last_encap_vif) { 724 last_encap_vif = 0; 725 last_encap_src = zeroin_addr; 726 } 727 } else { 728 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 729 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 730 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr; 731 ifp = vifp->v_ifp; 732 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 733 } 734 bzero((caddr_t)vifp, sizeof(*vifp)); 735 } 736 737 /* 738 * Delete a vif from the vif table 739 */ 740 static int 741 del_vif(m) 742 struct mbuf *m; 743 { 744 vifi_t *vifip; 745 struct vif *vifp; 746 vifi_t vifi; 747 int s; 748 749 if (m == 0 || m->m_len < sizeof(vifi_t)) 750 return (EINVAL); 751 752 vifip = mtod(m, vifi_t *); 753 if (*vifip >= numvifs) 754 return (EINVAL); 755 756 vifp = &viftable[*vifip]; 757 if (in_nullhost(vifp->v_lcl_addr)) 758 return (EADDRNOTAVAIL); 759 760 s = splsoftnet(); 761 762 reset_vif(vifp); 763 764 /* Adjust numvifs down */ 765 for (vifi = numvifs; vifi > 0; vifi--) 766 if (!in_nullhost(viftable[vifi-1].v_lcl_addr)) 767 break; 768 numvifs = vifi; 769 770 splx(s); 771 772 if (mrtdebug) 773 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 774 775 return (0); 776 } 777 778 static void 779 update_mfc(mfccp, rt) 780 struct mfcctl *mfccp; 781 struct mfc *rt; 782 { 783 vifi_t vifi; 784 785 rt->mfc_parent = mfccp->mfcc_parent; 786 for (vifi = 0; vifi < numvifs; vifi++) 787 rt->mfc_ttls[vifi] = mfccp->mfcc_ttls[vifi]; 788 rt->mfc_expire = 0; 789 rt->mfc_stall = 0; 790 } 791 792 static void 793 expire_mfc(rt) 794 struct mfc *rt; 795 { 796 struct rtdetq *rte, *nrte; 797 798 for (rte = rt->mfc_stall; rte != 0; rte = nrte) { 799 nrte = rte->next; 800 m_freem(rte->m); 801 free(rte, M_MRTABLE); 802 } 803 804 LIST_REMOVE(rt, mfc_hash); 805 free(rt, M_MRTABLE); 806 } 807 808 /* 809 * Add an mfc entry 810 */ 811 static int 812 add_mfc(m) 813 struct mbuf *m; 814 { 815 struct mfcctl *mfccp; 816 struct mfc *rt; 817 u_int32_t hash = 0; 818 struct rtdetq *rte, *nrte; 819 u_short nstl; 820 int s; 821 822 if (m == 0 || m->m_len < sizeof(struct mfcctl)) 823 return (EINVAL); 824 825 mfccp = mtod(m, struct mfcctl *); 826 827 s = splsoftnet(); 828 MFCFIND(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp, rt); 829 830 /* If an entry already exists, just update the fields */ 831 if (rt) { 832 if (mrtdebug & DEBUG_MFC) 833 log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n", 834 ntohl(mfccp->mfcc_origin.s_addr), 835 ntohl(mfccp->mfcc_mcastgrp.s_addr), 836 mfccp->mfcc_parent); 837 838 if (rt->mfc_expire) 839 nexpire[hash]--; 840 841 update_mfc(mfccp, rt); 842 843 splx(s); 844 return (0); 845 } 846 847 /* 848 * Find the entry for which the upcall was made and update 849 */ 850 nstl = 0; 851 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp); 852 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 853 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 854 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) && 855 rt->mfc_stall != 0) { 856 if (nstl++) 857 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n", 858 "multiple kernel entries", 859 ntohl(mfccp->mfcc_origin.s_addr), 860 ntohl(mfccp->mfcc_mcastgrp.s_addr), 861 mfccp->mfcc_parent, rt->mfc_stall); 862 863 if (mrtdebug & DEBUG_MFC) 864 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %p\n", 865 ntohl(mfccp->mfcc_origin.s_addr), 866 ntohl(mfccp->mfcc_mcastgrp.s_addr), 867 mfccp->mfcc_parent, rt->mfc_stall); 868 869 if (rt->mfc_expire) 870 nexpire[hash]--; 871 872 rte = rt->mfc_stall; 873 update_mfc(mfccp, rt); 874 875 /* free packets Qed at the end of this entry */ 876 for (; rte != 0; rte = nrte) { 877 nrte = rte->next; 878 #ifdef RSVP_ISI 879 ip_mdq(rte->m, rte->ifp, rt, -1); 880 #else 881 ip_mdq(rte->m, rte->ifp, rt); 882 #endif /* RSVP_ISI */ 883 m_freem(rte->m); 884 #ifdef UPCALL_TIMING 885 collate(&rte->t); 886 #endif /* UPCALL_TIMING */ 887 free(rte, M_MRTABLE); 888 } 889 } 890 } 891 892 if (nstl == 0) { 893 /* 894 * No mfc; make a new one 895 */ 896 if (mrtdebug & DEBUG_MFC) 897 log(LOG_DEBUG,"add_mfc no upcall o %x g %x p %x\n", 898 ntohl(mfccp->mfcc_origin.s_addr), 899 ntohl(mfccp->mfcc_mcastgrp.s_addr), 900 mfccp->mfcc_parent); 901 902 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 903 if (rt == 0) { 904 splx(s); 905 return (ENOBUFS); 906 } 907 908 rt->mfc_origin = mfccp->mfcc_origin; 909 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 910 /* initialize pkt counters per src-grp */ 911 rt->mfc_pkt_cnt = 0; 912 rt->mfc_byte_cnt = 0; 913 rt->mfc_wrong_if = 0; 914 timerclear(&rt->mfc_last_assert); 915 update_mfc(mfccp, rt); 916 917 /* insert new entry at head of hash chain */ 918 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 919 } 920 921 splx(s); 922 return (0); 923 } 924 925 #ifdef UPCALL_TIMING 926 /* 927 * collect delay statistics on the upcalls 928 */ 929 static void collate(t) 930 struct timeval *t; 931 { 932 u_int32_t d; 933 struct timeval tp; 934 u_int32_t delta; 935 936 microtime(&tp); 937 938 if (timercmp(t, &tp, <)) { 939 TV_DELTA(tp, *t, delta); 940 941 d = delta >> 10; 942 if (d > 50) 943 d = 50; 944 945 ++upcall_data[d]; 946 } 947 } 948 #endif /* UPCALL_TIMING */ 949 950 /* 951 * Delete an mfc entry 952 */ 953 static int 954 del_mfc(m) 955 struct mbuf *m; 956 { 957 struct mfcctl *mfccp; 958 struct mfc *rt; 959 int s; 960 961 if (m == 0 || m->m_len < sizeof(struct mfcctl)) 962 return (EINVAL); 963 964 mfccp = mtod(m, struct mfcctl *); 965 966 if (mrtdebug & DEBUG_MFC) 967 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n", 968 ntohl(mfccp->mfcc_origin.s_addr), 969 ntohl(mfccp->mfcc_mcastgrp.s_addr)); 970 971 s = splsoftnet(); 972 973 MFCFIND(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp, rt); 974 if (rt == 0) { 975 splx(s); 976 return (EADDRNOTAVAIL); 977 } 978 979 LIST_REMOVE(rt, mfc_hash); 980 free(rt, M_MRTABLE); 981 982 splx(s); 983 return (0); 984 } 985 986 static int 987 socket_send(s, mm, src) 988 struct socket *s; 989 struct mbuf *mm; 990 struct sockaddr_in *src; 991 { 992 if (s) { 993 if (sbappendaddr(&s->so_rcv, sintosa(src), mm, (struct mbuf *)0) != 0) { 994 sorwakeup(s); 995 return (0); 996 } 997 } 998 m_freem(mm); 999 return (-1); 1000 } 1001 1002 /* 1003 * IP multicast forwarding function. This function assumes that the packet 1004 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1005 * pointed to by "ifp", and the packet is to be relayed to other networks 1006 * that have members of the packet's destination IP multicast group. 1007 * 1008 * The packet is returned unscathed to the caller, unless it is 1009 * erroneous, in which case a non-zero return value tells the caller to 1010 * discard it. 1011 */ 1012 1013 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1014 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1015 1016 int 1017 #ifdef RSVP_ISI 1018 ip_mforward(m, ifp, imo) 1019 #else 1020 ip_mforward(m, ifp) 1021 #endif /* RSVP_ISI */ 1022 struct mbuf *m; 1023 struct ifnet *ifp; 1024 #ifdef RSVP_ISI 1025 struct ip_moptions *imo; 1026 #endif /* RSVP_ISI */ 1027 { 1028 struct ip *ip = mtod(m, struct ip *); 1029 struct mfc *rt; 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 ((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 vif *vifp; 1530 int s; 1531 struct ifqueue *ifq; 1532 1533 va_start(ap, m); 1534 off = va_arg(ap, int); 1535 proto = va_arg(ap, int); 1536 va_end(ap); 1537 1538 vifp = (struct vif *)encap_getarg(m); 1539 if (!vifp || proto != AF_INET) { 1540 m_freem(m); 1541 mrtstat.mrts_bad_tunnel++; 1542 return; 1543 } 1544 1545 m_adj(m, off); 1546 m->m_pkthdr.rcvif = vifp->v_ifp; 1547 ifq = &ipintrq; 1548 s = splnet(); 1549 if (IF_QFULL(ifq)) { 1550 IF_DROP(ifq); 1551 m_freem(m); 1552 } else { 1553 IF_ENQUEUE(ifq, m); 1554 /* 1555 * normally we would need a "schednetisr(NETISR_IP)" 1556 * here but we were called by ip_input and it is going 1557 * to loop back & try to dequeue the packet we just 1558 * queued as soon as we return so we avoid the 1559 * unnecessary software interrrupt. 1560 */ 1561 } 1562 splx(s); 1563 } 1564 1565 /* 1566 * Check if the packet should be grabbed by us. 1567 */ 1568 static int 1569 vif_encapcheck(m, off, proto, arg) 1570 const struct mbuf *m; 1571 int off; 1572 int proto; 1573 void *arg; 1574 { 1575 struct vif *vifp; 1576 struct ip ip; 1577 1578 #ifdef DIAGNOSTIC 1579 if (!arg || proto != IPPROTO_IPV4) 1580 panic("unexpected arg in vif_encapcheck"); 1581 #endif 1582 1583 /* 1584 * do not grab the packet if it's not to a multicast destination or if 1585 * we don't have an encapsulating tunnel with the source. 1586 * Note: This code assumes that the remote site IP address 1587 * uniquely identifies the tunnel (i.e., that this site has 1588 * at most one tunnel with the remote site). 1589 */ 1590 1591 /* LINTED const cast */ 1592 m_copydata((struct mbuf *)m, off, sizeof(ip), (caddr_t)&ip); 1593 if (!IN_MULTICAST(ip.ip_dst.s_addr)) 1594 return 0; 1595 1596 /* LINTED const cast */ 1597 m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip); 1598 if (!in_hosteq(ip.ip_src, last_encap_src)) { 1599 vifp = (struct vif *)arg; 1600 if (vifp->v_flags & VIFF_TUNNEL && 1601 in_hosteq(vifp->v_rmt_addr, ip.ip_src)) 1602 ; 1603 else 1604 return 0; 1605 last_encap_vif = vifp; 1606 last_encap_src = ip.ip_src; 1607 } else 1608 vifp = last_encap_vif; 1609 1610 /* 32bit match, since we have checked ip_src only */ 1611 return 32; 1612 } 1613 1614 /* 1615 * Token bucket filter module 1616 */ 1617 static void 1618 tbf_control(vifp, m, ip, len) 1619 struct vif *vifp; 1620 struct mbuf *m; 1621 struct ip *ip; 1622 u_int32_t len; 1623 { 1624 1625 if (len > MAX_BKT_SIZE) { 1626 /* drop if packet is too large */ 1627 mrtstat.mrts_pkt2large++; 1628 m_freem(m); 1629 return; 1630 } 1631 1632 tbf_update_tokens(vifp); 1633 1634 /* 1635 * If there are enough tokens, and the queue is empty, send this packet 1636 * out immediately. Otherwise, try to insert it on this vif's queue. 1637 */ 1638 if (vifp->tbf_q_len == 0) { 1639 if (len <= vifp->tbf_n_tok) { 1640 vifp->tbf_n_tok -= len; 1641 tbf_send_packet(vifp, m); 1642 } else { 1643 /* queue packet and timeout till later */ 1644 tbf_queue(vifp, m); 1645 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 1646 tbf_reprocess_q, vifp); 1647 } 1648 } else { 1649 if (vifp->tbf_q_len >= vifp->tbf_max_q_len && 1650 !tbf_dq_sel(vifp, ip)) { 1651 /* queue length too much, and couldn't make room */ 1652 mrtstat.mrts_q_overflow++; 1653 m_freem(m); 1654 } else { 1655 /* queue length low enough, or made room */ 1656 tbf_queue(vifp, m); 1657 tbf_process_q(vifp); 1658 } 1659 } 1660 } 1661 1662 /* 1663 * adds a packet to the queue at the interface 1664 */ 1665 static void 1666 tbf_queue(vifp, m) 1667 struct vif *vifp; 1668 struct mbuf *m; 1669 { 1670 int s = splsoftnet(); 1671 1672 /* insert at tail */ 1673 *vifp->tbf_t = m; 1674 vifp->tbf_t = &m->m_nextpkt; 1675 vifp->tbf_q_len++; 1676 1677 splx(s); 1678 } 1679 1680 1681 /* 1682 * processes the queue at the interface 1683 */ 1684 static void 1685 tbf_process_q(vifp) 1686 struct vif *vifp; 1687 { 1688 struct mbuf *m; 1689 int len; 1690 int s = splsoftnet(); 1691 1692 /* 1693 * Loop through the queue at the interface and send as many packets 1694 * as possible. 1695 */ 1696 for (m = vifp->tbf_q; 1697 m != 0; 1698 m = vifp->tbf_q) { 1699 len = ntohs(mtod(m, struct ip *)->ip_len); 1700 1701 /* determine if the packet can be sent */ 1702 if (len <= vifp->tbf_n_tok) { 1703 /* if so, 1704 * reduce no of tokens, dequeue the packet, 1705 * send the packet. 1706 */ 1707 if ((vifp->tbf_q = m->m_nextpkt) == 0) 1708 vifp->tbf_t = &vifp->tbf_q; 1709 --vifp->tbf_q_len; 1710 1711 m->m_nextpkt = 0; 1712 vifp->tbf_n_tok -= len; 1713 tbf_send_packet(vifp, m); 1714 } else 1715 break; 1716 } 1717 splx(s); 1718 } 1719 1720 static void 1721 tbf_reprocess_q(arg) 1722 void *arg; 1723 { 1724 struct vif *vifp = arg; 1725 1726 if (ip_mrouter == 0) 1727 return; 1728 1729 tbf_update_tokens(vifp); 1730 tbf_process_q(vifp); 1731 1732 if (vifp->tbf_q_len != 0) 1733 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 1734 tbf_reprocess_q, vifp); 1735 } 1736 1737 /* function that will selectively discard a member of the queue 1738 * based on the precedence value and the priority 1739 */ 1740 static int 1741 tbf_dq_sel(vifp, ip) 1742 struct vif *vifp; 1743 struct ip *ip; 1744 { 1745 u_int p; 1746 struct mbuf **mp, *m; 1747 int s = splsoftnet(); 1748 1749 p = priority(vifp, ip); 1750 1751 for (mp = &vifp->tbf_q, m = *mp; 1752 m != 0; 1753 mp = &m->m_nextpkt, m = *mp) { 1754 if (p > priority(vifp, mtod(m, struct ip *))) { 1755 if ((*mp = m->m_nextpkt) == 0) 1756 vifp->tbf_t = mp; 1757 --vifp->tbf_q_len; 1758 1759 m_freem(m); 1760 mrtstat.mrts_drop_sel++; 1761 splx(s); 1762 return (1); 1763 } 1764 } 1765 splx(s); 1766 return (0); 1767 } 1768 1769 static void 1770 tbf_send_packet(vifp, m) 1771 struct vif *vifp; 1772 struct mbuf *m; 1773 { 1774 int error; 1775 int s = splsoftnet(); 1776 1777 if (vifp->v_flags & VIFF_TUNNEL) { 1778 /* If tunnel options */ 1779 #ifdef IPSEC 1780 /* Don't lookup socket in forwading case */ 1781 (void)ipsec_setsocket(m, NULL); 1782 #endif 1783 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1784 IP_FORWARDING, (struct ip_moptions *)0); 1785 } else { 1786 /* if physical interface option, extract the options and then send */ 1787 struct ip_moptions imo; 1788 1789 imo.imo_multicast_ifp = vifp->v_ifp; 1790 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1791 imo.imo_multicast_loop = 1; 1792 #ifdef RSVP_ISI 1793 imo.imo_multicast_vif = -1; 1794 #endif 1795 1796 #ifdef IPSEC 1797 /* Don't lookup socket in forwading case */ 1798 (void)ipsec_setsocket(m, NULL); 1799 #endif 1800 error = ip_output(m, (struct mbuf *)0, (struct route *)0, 1801 IP_FORWARDING|IP_MULTICASTOPTS, &imo); 1802 1803 if (mrtdebug & DEBUG_XMIT) 1804 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n", 1805 (long)(vifp-viftable), error); 1806 } 1807 splx(s); 1808 } 1809 1810 /* determine the current time and then 1811 * the elapsed time (between the last time and time now) 1812 * in milliseconds & update the no. of tokens in the bucket 1813 */ 1814 static void 1815 tbf_update_tokens(vifp) 1816 struct vif *vifp; 1817 { 1818 struct timeval tp; 1819 u_int32_t tm; 1820 int s = splsoftnet(); 1821 1822 microtime(&tp); 1823 1824 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm); 1825 1826 /* 1827 * This formula is actually 1828 * "time in seconds" * "bytes/second". 1829 * 1830 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1831 * 1832 * The (1000/1024) was introduced in add_vif to optimize 1833 * this divide into a shift. 1834 */ 1835 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192; 1836 vifp->tbf_last_pkt_t = tp; 1837 1838 if (vifp->tbf_n_tok > MAX_BKT_SIZE) 1839 vifp->tbf_n_tok = MAX_BKT_SIZE; 1840 1841 splx(s); 1842 } 1843 1844 static int 1845 priority(vifp, ip) 1846 struct vif *vifp; 1847 struct ip *ip; 1848 { 1849 int prio; 1850 1851 /* temporary hack; may add general packet classifier some day */ 1852 1853 /* 1854 * The UDP port space is divided up into four priority ranges: 1855 * [0, 16384) : unclassified - lowest priority 1856 * [16384, 32768) : audio - highest priority 1857 * [32768, 49152) : whiteboard - medium priority 1858 * [49152, 65536) : video - low priority 1859 */ 1860 if (ip->ip_p == IPPROTO_UDP) { 1861 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1862 1863 switch (ntohs(udp->uh_dport) & 0xc000) { 1864 case 0x4000: 1865 prio = 70; 1866 break; 1867 case 0x8000: 1868 prio = 60; 1869 break; 1870 case 0xc000: 1871 prio = 55; 1872 break; 1873 default: 1874 prio = 50; 1875 break; 1876 } 1877 1878 if (tbfdebug > 1) 1879 log(LOG_DEBUG, "port %x prio %d\n", ntohs(udp->uh_dport), prio); 1880 } else 1881 prio = 50; 1882 1883 1884 return (prio); 1885 } 1886 1887 /* 1888 * End of token bucket filter modifications 1889 */ 1890 1891 #ifdef RSVP_ISI 1892 1893 int 1894 ip_rsvp_vif_init(so, m) 1895 struct socket *so; 1896 struct mbuf *m; 1897 { 1898 int i; 1899 int s; 1900 1901 if (rsvpdebug) 1902 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 1903 so->so_type, so->so_proto->pr_protocol); 1904 1905 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1906 return (EOPNOTSUPP); 1907 1908 /* Check mbuf. */ 1909 if (m == 0 || m->m_len != sizeof(int)) { 1910 return (EINVAL); 1911 } 1912 i = *(mtod(m, int *)); 1913 1914 if (rsvpdebug) 1915 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on); 1916 1917 s = splsoftnet(); 1918 1919 /* Check vif. */ 1920 if (!legal_vif_num(i)) { 1921 splx(s); 1922 return (EADDRNOTAVAIL); 1923 } 1924 1925 /* Check if socket is available. */ 1926 if (viftable[i].v_rsvpd != 0) { 1927 splx(s); 1928 return (EADDRINUSE); 1929 } 1930 1931 viftable[i].v_rsvpd = so; 1932 /* This may seem silly, but we need to be sure we don't over-increment 1933 * the RSVP counter, in case something slips up. 1934 */ 1935 if (!viftable[i].v_rsvp_on) { 1936 viftable[i].v_rsvp_on = 1; 1937 rsvp_on++; 1938 } 1939 1940 splx(s); 1941 return (0); 1942 } 1943 1944 int 1945 ip_rsvp_vif_done(so, m) 1946 struct socket *so; 1947 struct mbuf *m; 1948 { 1949 int i; 1950 int s; 1951 1952 if (rsvpdebug) 1953 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 1954 so->so_type, so->so_proto->pr_protocol); 1955 1956 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1957 return (EOPNOTSUPP); 1958 1959 /* Check mbuf. */ 1960 if (m == 0 || m->m_len != sizeof(int)) { 1961 return (EINVAL); 1962 } 1963 i = *(mtod(m, int *)); 1964 1965 s = splsoftnet(); 1966 1967 /* Check vif. */ 1968 if (!legal_vif_num(i)) { 1969 splx(s); 1970 return (EADDRNOTAVAIL); 1971 } 1972 1973 if (rsvpdebug) 1974 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n", 1975 viftable[i].v_rsvpd, so); 1976 1977 viftable[i].v_rsvpd = 0; 1978 /* This may seem silly, but we need to be sure we don't over-decrement 1979 * the RSVP counter, in case something slips up. 1980 */ 1981 if (viftable[i].v_rsvp_on) { 1982 viftable[i].v_rsvp_on = 0; 1983 rsvp_on--; 1984 } 1985 1986 splx(s); 1987 return (0); 1988 } 1989 1990 void 1991 ip_rsvp_force_done(so) 1992 struct socket *so; 1993 { 1994 int vifi; 1995 int s; 1996 1997 /* Don't bother if it is not the right type of socket. */ 1998 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1999 return; 2000 2001 s = splsoftnet(); 2002 2003 /* The socket may be attached to more than one vif...this 2004 * is perfectly legal. 2005 */ 2006 for (vifi = 0; vifi < numvifs; vifi++) { 2007 if (viftable[vifi].v_rsvpd == so) { 2008 viftable[vifi].v_rsvpd = 0; 2009 /* This may seem silly, but we need to be sure we don't 2010 * over-decrement the RSVP counter, in case something slips up. 2011 */ 2012 if (viftable[vifi].v_rsvp_on) { 2013 viftable[vifi].v_rsvp_on = 0; 2014 rsvp_on--; 2015 } 2016 } 2017 } 2018 2019 splx(s); 2020 return; 2021 } 2022 2023 void 2024 rsvp_input(m, ifp) 2025 struct mbuf *m; 2026 struct ifnet *ifp; 2027 { 2028 int vifi; 2029 struct ip *ip = mtod(m, struct ip *); 2030 static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET }; 2031 int s; 2032 2033 if (rsvpdebug) 2034 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2035 2036 /* Can still get packets with rsvp_on = 0 if there is a local member 2037 * of the group to which the RSVP packet is addressed. But in this 2038 * case we want to throw the packet away. 2039 */ 2040 if (!rsvp_on) { 2041 m_freem(m); 2042 return; 2043 } 2044 2045 /* If the old-style non-vif-associated socket is set, then use 2046 * it and ignore the new ones. 2047 */ 2048 if (ip_rsvpd != 0) { 2049 if (rsvpdebug) 2050 printf("rsvp_input: Sending packet up old-style socket\n"); 2051 rip_input(m); /*XXX*/ 2052 return; 2053 } 2054 2055 s = splsoftnet(); 2056 2057 if (rsvpdebug) 2058 printf("rsvp_input: check vifs\n"); 2059 2060 /* Find which vif the packet arrived on. */ 2061 for (vifi = 0; vifi < numvifs; vifi++) { 2062 if (viftable[vifi].v_ifp == ifp) 2063 break; 2064 } 2065 2066 if (vifi == numvifs) { 2067 /* Can't find vif packet arrived on. Drop packet. */ 2068 if (rsvpdebug) 2069 printf("rsvp_input: Can't find vif for packet...dropping it.\n"); 2070 m_freem(m); 2071 splx(s); 2072 return; 2073 } 2074 2075 if (rsvpdebug) 2076 printf("rsvp_input: check socket\n"); 2077 2078 if (viftable[vifi].v_rsvpd == 0) { 2079 /* drop packet, since there is no specific socket for this 2080 * interface */ 2081 if (rsvpdebug) 2082 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2083 m_freem(m); 2084 splx(s); 2085 return; 2086 } 2087 2088 rsvp_src.sin_addr = ip->ip_src; 2089 2090 if (rsvpdebug && m) 2091 printf("rsvp_input: m->m_len = %d, sbspace() = %d\n", 2092 m->m_len,sbspace(&viftable[vifi].v_rsvpd->so_rcv)); 2093 2094 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2095 if (rsvpdebug) 2096 printf("rsvp_input: Failed to append to socket\n"); 2097 else 2098 if (rsvpdebug) 2099 printf("rsvp_input: send packet up\n"); 2100 2101 splx(s); 2102 } 2103 #endif /* RSVP_ISI */ 2104