1 /* $NetBSD: ip_mroute.c,v 1.71 2003/05/14 17:28:31 itojun 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 1.71 2003/05/14 17:28:31 itojun 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) do { \ 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 } while (/*CONSTCOND*/ 0) 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) do { \ 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 } while (/*CONSTCOND*/ 0) 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, 994 (struct mbuf *)0) != 0) { 995 sorwakeup(s); 996 return (0); 997 } 998 } 999 m_freem(mm); 1000 return (-1); 1001 } 1002 1003 /* 1004 * IP multicast forwarding function. This function assumes that the packet 1005 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1006 * pointed to by "ifp", and the packet is to be relayed to other networks 1007 * that have members of the packet's destination IP multicast group. 1008 * 1009 * The packet is returned unscathed to the caller, unless it is 1010 * erroneous, in which case a non-zero return value tells the caller to 1011 * discard it. 1012 */ 1013 1014 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1015 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1016 1017 int 1018 #ifdef RSVP_ISI 1019 ip_mforward(m, ifp, imo) 1020 #else 1021 ip_mforward(m, ifp) 1022 #endif /* RSVP_ISI */ 1023 struct mbuf *m; 1024 struct ifnet *ifp; 1025 #ifdef RSVP_ISI 1026 struct ip_moptions *imo; 1027 #endif /* RSVP_ISI */ 1028 { 1029 struct ip *ip = mtod(m, struct ip *); 1030 struct mfc *rt; 1031 static int srctun = 0; 1032 struct mbuf *mm; 1033 int s; 1034 #ifdef RSVP_ISI 1035 struct vif *vifp; 1036 vifi_t vifi; 1037 #endif /* RSVP_ISI */ 1038 1039 /* 1040 * Clear any in-bound checksum flags for this packet. 1041 */ 1042 m->m_pkthdr.csum_flags = 0; 1043 1044 if (mrtdebug & DEBUG_FORWARD) 1045 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n", 1046 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1047 1048 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1049 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) { 1050 /* 1051 * Packet arrived via a physical interface or 1052 * an encapuslated tunnel. 1053 */ 1054 } else { 1055 /* 1056 * Packet arrived through a source-route tunnel. 1057 * Source-route tunnels are no longer supported. 1058 */ 1059 if ((srctun++ % 1000) == 0) 1060 log(LOG_ERR, 1061 "ip_mforward: received source-routed packet from %x\n", 1062 ntohl(ip->ip_src.s_addr)); 1063 1064 return (1); 1065 } 1066 1067 #ifdef RSVP_ISI 1068 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1069 if (ip->ip_ttl < 255) 1070 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1071 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1072 vifp = viftable + vifi; 1073 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n", 1074 ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi, 1075 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1076 vifp->v_ifp->if_xname); 1077 } 1078 return (ip_mdq(m, ifp, (struct mfc *)0, vifi)); 1079 } 1080 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1081 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n", 1082 ntohl(ip->ip_src), ntohl(ip->ip_dst)); 1083 } 1084 #endif /* RSVP_ISI */ 1085 1086 /* 1087 * Don't forward a packet with time-to-live of zero or one, 1088 * or a packet destined to a local-only group. 1089 */ 1090 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr)) 1091 return (0); 1092 1093 /* 1094 * Determine forwarding vifs from the forwarding cache table 1095 */ 1096 s = splsoftnet(); 1097 MFCFIND(ip->ip_src, ip->ip_dst, rt); 1098 1099 /* Entry exists, so forward if necessary */ 1100 if (rt != 0) { 1101 splx(s); 1102 #ifdef RSVP_ISI 1103 return (ip_mdq(m, ifp, rt, -1)); 1104 #else 1105 return (ip_mdq(m, ifp, rt)); 1106 #endif /* RSVP_ISI */ 1107 } else { 1108 /* 1109 * If we don't have a route for packet's origin, 1110 * Make a copy of the packet & 1111 * send message to routing daemon 1112 */ 1113 1114 struct mbuf *mb0; 1115 struct rtdetq *rte; 1116 u_int32_t hash; 1117 int hlen = ip->ip_hl << 2; 1118 #ifdef UPCALL_TIMING 1119 struct timeval tp; 1120 1121 microtime(&tp); 1122 #endif /* UPCALL_TIMING */ 1123 1124 mrtstat.mrts_no_route++; 1125 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1126 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1127 ntohl(ip->ip_src.s_addr), 1128 ntohl(ip->ip_dst.s_addr)); 1129 1130 /* 1131 * Allocate mbufs early so that we don't do extra work if we are 1132 * just going to fail anyway. Make sure to pullup the header so 1133 * that other people can't step on it. 1134 */ 1135 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, 1136 M_NOWAIT); 1137 if (rte == 0) { 1138 splx(s); 1139 return (ENOBUFS); 1140 } 1141 mb0 = m_copy(m, 0, M_COPYALL); 1142 M_PULLUP(mb0, hlen); 1143 if (mb0 == 0) { 1144 free(rte, M_MRTABLE); 1145 splx(s); 1146 return (ENOBUFS); 1147 } 1148 1149 /* is there an upcall waiting for this packet? */ 1150 hash = MFCHASH(ip->ip_src, ip->ip_dst); 1151 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1152 if (in_hosteq(ip->ip_src, rt->mfc_origin) && 1153 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) && 1154 rt->mfc_stall != 0) 1155 break; 1156 } 1157 1158 if (rt == 0) { 1159 int i; 1160 struct igmpmsg *im; 1161 1162 /* no upcall, so make a new entry */ 1163 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, 1164 M_NOWAIT); 1165 if (rt == 0) { 1166 free(rte, M_MRTABLE); 1167 m_freem(mb0); 1168 splx(s); 1169 return (ENOBUFS); 1170 } 1171 /* 1172 * Make a copy of the header to send to the user level 1173 * process 1174 */ 1175 mm = m_copy(m, 0, hlen); 1176 M_PULLUP(mm, hlen); 1177 if (mm == 0) { 1178 free(rte, M_MRTABLE); 1179 m_freem(mb0); 1180 free(rt, M_MRTABLE); 1181 splx(s); 1182 return (ENOBUFS); 1183 } 1184 1185 /* 1186 * Send message to routing daemon to install 1187 * a route into the kernel table 1188 */ 1189 sin.sin_addr = ip->ip_src; 1190 1191 im = mtod(mm, struct igmpmsg *); 1192 im->im_msgtype = IGMPMSG_NOCACHE; 1193 im->im_mbz = 0; 1194 1195 mrtstat.mrts_upcalls++; 1196 1197 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1198 log(LOG_WARNING, 1199 "ip_mforward: ip_mrouter socket queue full\n"); 1200 ++mrtstat.mrts_upq_sockfull; 1201 free(rte, M_MRTABLE); 1202 m_freem(mb0); 1203 free(rt, M_MRTABLE); 1204 splx(s); 1205 return (ENOBUFS); 1206 } 1207 1208 /* insert new entry at head of hash chain */ 1209 rt->mfc_origin = ip->ip_src; 1210 rt->mfc_mcastgrp = ip->ip_dst; 1211 rt->mfc_pkt_cnt = 0; 1212 rt->mfc_byte_cnt = 0; 1213 rt->mfc_wrong_if = 0; 1214 rt->mfc_expire = UPCALL_EXPIRE; 1215 nexpire[hash]++; 1216 for (i = 0; i < numvifs; i++) 1217 rt->mfc_ttls[i] = 0; 1218 rt->mfc_parent = -1; 1219 1220 /* link into table */ 1221 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1222 /* Add this entry to the end of the queue */ 1223 rt->mfc_stall = rte; 1224 } else { 1225 /* determine if q has overflowed */ 1226 struct rtdetq **p; 1227 int npkts = 0; 1228 1229 for (p = &rt->mfc_stall; *p != 0; p = &(*p)->next) 1230 if (++npkts > MAX_UPQ) { 1231 mrtstat.mrts_upq_ovflw++; 1232 free(rte, M_MRTABLE); 1233 m_freem(mb0); 1234 splx(s); 1235 return (0); 1236 } 1237 1238 /* Add this entry to the end of the queue */ 1239 *p = rte; 1240 } 1241 1242 rte->next = 0; 1243 rte->m = mb0; 1244 rte->ifp = ifp; 1245 #ifdef UPCALL_TIMING 1246 rte->t = tp; 1247 #endif /* UPCALL_TIMING */ 1248 1249 splx(s); 1250 1251 return (0); 1252 } 1253 } 1254 1255 1256 /*ARGSUSED*/ 1257 static void 1258 expire_upcalls(v) 1259 void *v; 1260 { 1261 int i; 1262 int s; 1263 1264 s = splsoftnet(); 1265 1266 for (i = 0; i < MFCTBLSIZ; i++) { 1267 struct mfc *rt, *nrt; 1268 1269 if (nexpire[i] == 0) 1270 continue; 1271 1272 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 1273 nrt = LIST_NEXT(rt, mfc_hash); 1274 1275 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0) 1276 continue; 1277 nexpire[i]--; 1278 1279 ++mrtstat.mrts_cache_cleanups; 1280 if (mrtdebug & DEBUG_EXPIRE) 1281 log(LOG_DEBUG, 1282 "expire_upcalls: expiring (%x %x)\n", 1283 ntohl(rt->mfc_origin.s_addr), 1284 ntohl(rt->mfc_mcastgrp.s_addr)); 1285 1286 expire_mfc(rt); 1287 } 1288 } 1289 1290 splx(s); 1291 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1292 expire_upcalls, NULL); 1293 } 1294 1295 /* 1296 * Packet forwarding routine once entry in the cache is made 1297 */ 1298 static int 1299 #ifdef RSVP_ISI 1300 ip_mdq(m, ifp, rt, xmt_vif) 1301 #else 1302 ip_mdq(m, ifp, rt) 1303 #endif /* RSVP_ISI */ 1304 struct mbuf *m; 1305 struct ifnet *ifp; 1306 struct mfc *rt; 1307 #ifdef RSVP_ISI 1308 vifi_t xmt_vif; 1309 #endif /* RSVP_ISI */ 1310 { 1311 struct ip *ip = mtod(m, struct ip *); 1312 vifi_t vifi; 1313 struct vif *vifp; 1314 int plen = ntohs(ip->ip_len); 1315 1316 /* 1317 * Macro to send packet on vif. Since RSVP packets don't get counted on 1318 * input, they shouldn't get counted on output, so statistics keeping is 1319 * separate. 1320 */ 1321 #define MC_SEND(ip, vifp, m) do { \ 1322 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1323 encap_send((ip), (vifp), (m)); \ 1324 else \ 1325 phyint_send((ip), (vifp), (m)); \ 1326 } while (/*CONSTCOND*/ 0) 1327 1328 #ifdef RSVP_ISI 1329 /* 1330 * If xmt_vif is not -1, send on only the requested vif. 1331 * 1332 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs. 1333 */ 1334 if (xmt_vif < numvifs) { 1335 MC_SEND(ip, viftable + xmt_vif, m); 1336 return (1); 1337 } 1338 #endif /* RSVP_ISI */ 1339 1340 /* 1341 * Don't forward if it didn't arrive from the parent vif for its origin. 1342 */ 1343 vifi = rt->mfc_parent; 1344 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1345 /* came in the wrong interface */ 1346 if (mrtdebug & DEBUG_FORWARD) 1347 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1348 ifp, vifi, 1349 vifi >= numvifs ? 0 : viftable[vifi].v_ifp); 1350 ++mrtstat.mrts_wrong_if; 1351 ++rt->mfc_wrong_if; 1352 /* 1353 * If we are doing PIM assert processing, and we are forwarding 1354 * packets on this interface, and it is a broadcast medium 1355 * interface (and not a tunnel), send a message to the routing 1356 * daemon. 1357 */ 1358 if (pim_assert && rt->mfc_ttls[vifi] && 1359 (ifp->if_flags & IFF_BROADCAST) && 1360 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1361 struct mbuf *mm; 1362 struct igmpmsg *im; 1363 int hlen = ip->ip_hl << 2; 1364 struct timeval now; 1365 u_int32_t delta; 1366 1367 microtime(&now); 1368 1369 TV_DELTA(rt->mfc_last_assert, now, delta); 1370 1371 if (delta > ASSERT_MSG_TIME) { 1372 mm = m_copy(m, 0, hlen); 1373 M_PULLUP(mm, hlen); 1374 if (mm == 0) { 1375 return (ENOBUFS); 1376 } 1377 1378 rt->mfc_last_assert = now; 1379 1380 im = mtod(mm, struct igmpmsg *); 1381 im->im_msgtype = IGMPMSG_WRONGVIF; 1382 im->im_mbz = 0; 1383 im->im_vif = vifi; 1384 1385 sin.sin_addr = im->im_src; 1386 1387 socket_send(ip_mrouter, mm, &sin); 1388 } 1389 } 1390 return (0); 1391 } 1392 1393 /* If I sourced this packet, it counts as output, else it was input. */ 1394 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) { 1395 viftable[vifi].v_pkt_out++; 1396 viftable[vifi].v_bytes_out += plen; 1397 } else { 1398 viftable[vifi].v_pkt_in++; 1399 viftable[vifi].v_bytes_in += plen; 1400 } 1401 rt->mfc_pkt_cnt++; 1402 rt->mfc_byte_cnt += plen; 1403 1404 /* 1405 * For each vif, decide if a copy of the packet should be forwarded. 1406 * Forward if: 1407 * - the ttl exceeds the vif's threshold 1408 * - there are group members downstream on interface 1409 */ 1410 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1411 if ((rt->mfc_ttls[vifi] > 0) && 1412 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1413 vifp->v_pkt_out++; 1414 vifp->v_bytes_out += plen; 1415 MC_SEND(ip, vifp, m); 1416 } 1417 1418 return (0); 1419 } 1420 1421 #ifdef RSVP_ISI 1422 /* 1423 * check if a vif number is legal/ok. This is used by ip_output, to export 1424 * numvifs there, 1425 */ 1426 int 1427 legal_vif_num(vif) 1428 int vif; 1429 { 1430 if (vif >= 0 && vif < numvifs) 1431 return (1); 1432 else 1433 return (0); 1434 } 1435 #endif /* RSVP_ISI */ 1436 1437 static void 1438 phyint_send(ip, vifp, m) 1439 struct ip *ip; 1440 struct vif *vifp; 1441 struct mbuf *m; 1442 { 1443 struct mbuf *mb_copy; 1444 int hlen = ip->ip_hl << 2; 1445 1446 /* 1447 * Make a new reference to the packet; make sure that 1448 * the IP header is actually copied, not just referenced, 1449 * so that ip_output() only scribbles on the copy. 1450 */ 1451 mb_copy = m_copy(m, 0, M_COPYALL); 1452 M_PULLUP(mb_copy, hlen); 1453 if (mb_copy == 0) 1454 return; 1455 1456 if (vifp->v_rate_limit <= 0) 1457 tbf_send_packet(vifp, mb_copy); 1458 else 1459 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), 1460 ntohs(ip->ip_len)); 1461 } 1462 1463 static void 1464 encap_send(ip, vifp, m) 1465 struct ip *ip; 1466 struct vif *vifp; 1467 struct mbuf *m; 1468 { 1469 struct mbuf *mb_copy; 1470 struct ip *ip_copy; 1471 int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr); 1472 1473 /* 1474 * copy the old packet & pullup it's IP header into the 1475 * new mbuf so we can modify it. Try to fill the new 1476 * mbuf since if we don't the ethernet driver will. 1477 */ 1478 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA); 1479 if (mb_copy == 0) 1480 return; 1481 mb_copy->m_data += max_linkhdr; 1482 mb_copy->m_pkthdr.len = len; 1483 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1484 1485 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == 0) { 1486 m_freem(mb_copy); 1487 return; 1488 } 1489 i = MHLEN - max_linkhdr; 1490 if (i > len) 1491 i = len; 1492 mb_copy = m_pullup(mb_copy, i); 1493 if (mb_copy == 0) 1494 return; 1495 1496 /* 1497 * fill in the encapsulating IP header. 1498 */ 1499 ip_copy = mtod(mb_copy, struct ip *); 1500 *ip_copy = multicast_encap_iphdr; 1501 ip_copy->ip_id = htons(ip_id++); 1502 ip_copy->ip_len = htons(len); 1503 ip_copy->ip_src = vifp->v_lcl_addr; 1504 ip_copy->ip_dst = vifp->v_rmt_addr; 1505 1506 /* 1507 * turn the encapsulated IP header back into a valid one. 1508 */ 1509 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1510 --ip->ip_ttl; 1511 ip->ip_sum = 0; 1512 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1513 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1514 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1515 1516 if (vifp->v_rate_limit <= 0) 1517 tbf_send_packet(vifp, mb_copy); 1518 else 1519 tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len)); 1520 } 1521 1522 /* 1523 * De-encapsulate a packet and feed it back through ip input. 1524 */ 1525 static void 1526 #if __STDC__ 1527 vif_input(struct mbuf *m, ...) 1528 #else 1529 vif_input(m, va_alist) 1530 struct mbuf *m; 1531 va_dcl 1532 #endif 1533 { 1534 int off, proto; 1535 va_list ap; 1536 struct vif *vifp; 1537 int s; 1538 struct ifqueue *ifq; 1539 1540 va_start(ap, m); 1541 off = va_arg(ap, int); 1542 proto = va_arg(ap, int); 1543 va_end(ap); 1544 1545 vifp = (struct vif *)encap_getarg(m); 1546 if (!vifp || proto != AF_INET) { 1547 m_freem(m); 1548 mrtstat.mrts_bad_tunnel++; 1549 return; 1550 } 1551 1552 m_adj(m, off); 1553 m->m_pkthdr.rcvif = vifp->v_ifp; 1554 ifq = &ipintrq; 1555 s = splnet(); 1556 if (IF_QFULL(ifq)) { 1557 IF_DROP(ifq); 1558 m_freem(m); 1559 } else { 1560 IF_ENQUEUE(ifq, m); 1561 /* 1562 * normally we would need a "schednetisr(NETISR_IP)" 1563 * here but we were called by ip_input and it is going 1564 * to loop back & try to dequeue the packet we just 1565 * queued as soon as we return so we avoid the 1566 * unnecessary software interrrupt. 1567 */ 1568 } 1569 splx(s); 1570 } 1571 1572 /* 1573 * Check if the packet should be grabbed by us. 1574 */ 1575 static int 1576 vif_encapcheck(m, off, proto, arg) 1577 const struct mbuf *m; 1578 int off; 1579 int proto; 1580 void *arg; 1581 { 1582 struct vif *vifp; 1583 struct ip ip; 1584 1585 #ifdef DIAGNOSTIC 1586 if (!arg || proto != IPPROTO_IPV4) 1587 panic("unexpected arg in vif_encapcheck"); 1588 #endif 1589 1590 /* 1591 * do not grab the packet if it's not to a multicast destination or if 1592 * we don't have an encapsulating tunnel with the source. 1593 * Note: This code assumes that the remote site IP address 1594 * uniquely identifies the tunnel (i.e., that this site has 1595 * at most one tunnel with the remote site). 1596 */ 1597 1598 /* LINTED const cast */ 1599 m_copydata((struct mbuf *)m, off, sizeof(ip), (caddr_t)&ip); 1600 if (!IN_MULTICAST(ip.ip_dst.s_addr)) 1601 return 0; 1602 1603 /* LINTED const cast */ 1604 m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip); 1605 if (!in_hosteq(ip.ip_src, last_encap_src)) { 1606 vifp = (struct vif *)arg; 1607 if (vifp->v_flags & VIFF_TUNNEL && 1608 in_hosteq(vifp->v_rmt_addr, ip.ip_src)) 1609 ; 1610 else 1611 return 0; 1612 last_encap_vif = vifp; 1613 last_encap_src = ip.ip_src; 1614 } else 1615 vifp = last_encap_vif; 1616 1617 /* 32bit match, since we have checked ip_src only */ 1618 return 32; 1619 } 1620 1621 /* 1622 * Token bucket filter module 1623 */ 1624 static void 1625 tbf_control(vifp, m, ip, len) 1626 struct vif *vifp; 1627 struct mbuf *m; 1628 struct ip *ip; 1629 u_int32_t len; 1630 { 1631 1632 if (len > MAX_BKT_SIZE) { 1633 /* drop if packet is too large */ 1634 mrtstat.mrts_pkt2large++; 1635 m_freem(m); 1636 return; 1637 } 1638 1639 tbf_update_tokens(vifp); 1640 1641 /* 1642 * If there are enough tokens, and the queue is empty, send this packet 1643 * out immediately. Otherwise, try to insert it on this vif's queue. 1644 */ 1645 if (vifp->tbf_q_len == 0) { 1646 if (len <= vifp->tbf_n_tok) { 1647 vifp->tbf_n_tok -= len; 1648 tbf_send_packet(vifp, m); 1649 } else { 1650 /* queue packet and timeout till later */ 1651 tbf_queue(vifp, m); 1652 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 1653 tbf_reprocess_q, vifp); 1654 } 1655 } else { 1656 if (vifp->tbf_q_len >= vifp->tbf_max_q_len && 1657 !tbf_dq_sel(vifp, ip)) { 1658 /* queue length too much, and couldn't make room */ 1659 mrtstat.mrts_q_overflow++; 1660 m_freem(m); 1661 } else { 1662 /* queue length low enough, or made room */ 1663 tbf_queue(vifp, m); 1664 tbf_process_q(vifp); 1665 } 1666 } 1667 } 1668 1669 /* 1670 * adds a packet to the queue at the interface 1671 */ 1672 static void 1673 tbf_queue(vifp, m) 1674 struct vif *vifp; 1675 struct mbuf *m; 1676 { 1677 int s = splsoftnet(); 1678 1679 /* insert at tail */ 1680 *vifp->tbf_t = m; 1681 vifp->tbf_t = &m->m_nextpkt; 1682 vifp->tbf_q_len++; 1683 1684 splx(s); 1685 } 1686 1687 1688 /* 1689 * processes the queue at the interface 1690 */ 1691 static void 1692 tbf_process_q(vifp) 1693 struct vif *vifp; 1694 { 1695 struct mbuf *m; 1696 int len; 1697 int s = splsoftnet(); 1698 1699 /* 1700 * Loop through the queue at the interface and send as many packets 1701 * as possible. 1702 */ 1703 for (m = vifp->tbf_q; m != 0; m = vifp->tbf_q) { 1704 len = ntohs(mtod(m, struct ip *)->ip_len); 1705 1706 /* determine if the packet can be sent */ 1707 if (len <= vifp->tbf_n_tok) { 1708 /* if so, 1709 * reduce no of tokens, dequeue the packet, 1710 * send the packet. 1711 */ 1712 if ((vifp->tbf_q = m->m_nextpkt) == 0) 1713 vifp->tbf_t = &vifp->tbf_q; 1714 --vifp->tbf_q_len; 1715 1716 m->m_nextpkt = 0; 1717 vifp->tbf_n_tok -= len; 1718 tbf_send_packet(vifp, m); 1719 } else 1720 break; 1721 } 1722 splx(s); 1723 } 1724 1725 static void 1726 tbf_reprocess_q(arg) 1727 void *arg; 1728 { 1729 struct vif *vifp = arg; 1730 1731 if (ip_mrouter == 0) 1732 return; 1733 1734 tbf_update_tokens(vifp); 1735 tbf_process_q(vifp); 1736 1737 if (vifp->tbf_q_len != 0) 1738 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 1739 tbf_reprocess_q, vifp); 1740 } 1741 1742 /* function that will selectively discard a member of the queue 1743 * based on the precedence value and the priority 1744 */ 1745 static int 1746 tbf_dq_sel(vifp, ip) 1747 struct vif *vifp; 1748 struct ip *ip; 1749 { 1750 u_int p; 1751 struct mbuf **mp, *m; 1752 int s = splsoftnet(); 1753 1754 p = priority(vifp, ip); 1755 1756 for (mp = &vifp->tbf_q, m = *mp; 1757 m != 0; 1758 mp = &m->m_nextpkt, m = *mp) { 1759 if (p > priority(vifp, mtod(m, struct ip *))) { 1760 if ((*mp = m->m_nextpkt) == 0) 1761 vifp->tbf_t = mp; 1762 --vifp->tbf_q_len; 1763 1764 m_freem(m); 1765 mrtstat.mrts_drop_sel++; 1766 splx(s); 1767 return (1); 1768 } 1769 } 1770 splx(s); 1771 return (0); 1772 } 1773 1774 static void 1775 tbf_send_packet(vifp, m) 1776 struct vif *vifp; 1777 struct mbuf *m; 1778 { 1779 int error; 1780 int s = splsoftnet(); 1781 1782 if (vifp->v_flags & VIFF_TUNNEL) { 1783 /* If tunnel options */ 1784 #ifdef IPSEC 1785 /* Don't lookup socket in forwading case */ 1786 (void)ipsec_setsocket(m, NULL); 1787 #endif 1788 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1789 IP_FORWARDING, (struct ip_moptions *)0); 1790 } else { 1791 /* if physical interface option, extract the options and then send */ 1792 struct ip_moptions imo; 1793 1794 imo.imo_multicast_ifp = vifp->v_ifp; 1795 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1796 imo.imo_multicast_loop = 1; 1797 #ifdef RSVP_ISI 1798 imo.imo_multicast_vif = -1; 1799 #endif 1800 1801 #ifdef IPSEC 1802 /* Don't lookup socket in forwading case */ 1803 (void)ipsec_setsocket(m, NULL); 1804 #endif 1805 error = ip_output(m, (struct mbuf *)0, (struct route *)0, 1806 IP_FORWARDING|IP_MULTICASTOPTS, &imo); 1807 1808 if (mrtdebug & DEBUG_XMIT) 1809 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n", 1810 (long)(vifp - viftable), error); 1811 } 1812 splx(s); 1813 } 1814 1815 /* determine the current time and then 1816 * the elapsed time (between the last time and time now) 1817 * in milliseconds & update the no. of tokens in the bucket 1818 */ 1819 static void 1820 tbf_update_tokens(vifp) 1821 struct vif *vifp; 1822 { 1823 struct timeval tp; 1824 u_int32_t tm; 1825 int s = splsoftnet(); 1826 1827 microtime(&tp); 1828 1829 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm); 1830 1831 /* 1832 * This formula is actually 1833 * "time in seconds" * "bytes/second". 1834 * 1835 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1836 * 1837 * The (1000/1024) was introduced in add_vif to optimize 1838 * this divide into a shift. 1839 */ 1840 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192; 1841 vifp->tbf_last_pkt_t = tp; 1842 1843 if (vifp->tbf_n_tok > MAX_BKT_SIZE) 1844 vifp->tbf_n_tok = MAX_BKT_SIZE; 1845 1846 splx(s); 1847 } 1848 1849 static int 1850 priority(vifp, ip) 1851 struct vif *vifp; 1852 struct ip *ip; 1853 { 1854 int prio; 1855 1856 /* temporary hack; may add general packet classifier some day */ 1857 1858 /* 1859 * The UDP port space is divided up into four priority ranges: 1860 * [0, 16384) : unclassified - lowest priority 1861 * [16384, 32768) : audio - highest priority 1862 * [32768, 49152) : whiteboard - medium priority 1863 * [49152, 65536) : video - low priority 1864 */ 1865 if (ip->ip_p == IPPROTO_UDP) { 1866 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1867 1868 switch (ntohs(udp->uh_dport) & 0xc000) { 1869 case 0x4000: 1870 prio = 70; 1871 break; 1872 case 0x8000: 1873 prio = 60; 1874 break; 1875 case 0xc000: 1876 prio = 55; 1877 break; 1878 default: 1879 prio = 50; 1880 break; 1881 } 1882 1883 if (tbfdebug > 1) 1884 log(LOG_DEBUG, "port %x prio %d\n", 1885 ntohs(udp->uh_dport), prio); 1886 } else 1887 prio = 50; 1888 1889 return (prio); 1890 } 1891 1892 /* 1893 * End of token bucket filter modifications 1894 */ 1895 #ifdef RSVP_ISI 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 || 1909 so->so_proto->pr_protocol != IPPROTO_RSVP) 1910 return (EOPNOTSUPP); 1911 1912 /* Check mbuf. */ 1913 if (m == 0 || m->m_len != sizeof(int)) { 1914 return (EINVAL); 1915 } 1916 i = *(mtod(m, int *)); 1917 1918 if (rsvpdebug) 1919 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", i, rsvp_on); 1920 1921 s = splsoftnet(); 1922 1923 /* Check vif. */ 1924 if (!legal_vif_num(i)) { 1925 splx(s); 1926 return (EADDRNOTAVAIL); 1927 } 1928 1929 /* Check if socket is available. */ 1930 if (viftable[i].v_rsvpd != 0) { 1931 splx(s); 1932 return (EADDRINUSE); 1933 } 1934 1935 viftable[i].v_rsvpd = so; 1936 /* 1937 * This may seem silly, but we need to be sure we don't over-increment 1938 * the RSVP counter, in case something slips up. 1939 */ 1940 if (!viftable[i].v_rsvp_on) { 1941 viftable[i].v_rsvp_on = 1; 1942 rsvp_on++; 1943 } 1944 1945 splx(s); 1946 return (0); 1947 } 1948 1949 int 1950 ip_rsvp_vif_done(so, m) 1951 struct socket *so; 1952 struct mbuf *m; 1953 { 1954 int i; 1955 int s; 1956 1957 if (rsvpdebug) 1958 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 1959 so->so_type, so->so_proto->pr_protocol); 1960 1961 if (so->so_type != SOCK_RAW || 1962 so->so_proto->pr_protocol != IPPROTO_RSVP) 1963 return (EOPNOTSUPP); 1964 1965 /* Check mbuf. */ 1966 if (m == 0 || m->m_len != sizeof(int)) { 1967 return (EINVAL); 1968 } 1969 i = *(mtod(m, int *)); 1970 1971 s = splsoftnet(); 1972 1973 /* Check vif. */ 1974 if (!legal_vif_num(i)) { 1975 splx(s); 1976 return (EADDRNOTAVAIL); 1977 } 1978 1979 if (rsvpdebug) 1980 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n", 1981 viftable[i].v_rsvpd, so); 1982 1983 viftable[i].v_rsvpd = 0; 1984 /* 1985 * This may seem silly, but we need to be sure we don't over-decrement 1986 * the RSVP counter, in case something slips up. 1987 */ 1988 if (viftable[i].v_rsvp_on) { 1989 viftable[i].v_rsvp_on = 0; 1990 rsvp_on--; 1991 } 1992 1993 splx(s); 1994 return (0); 1995 } 1996 1997 void 1998 ip_rsvp_force_done(so) 1999 struct socket *so; 2000 { 2001 int vifi; 2002 int s; 2003 2004 /* Don't bother if it is not the right type of socket. */ 2005 if (so->so_type != SOCK_RAW || 2006 so->so_proto->pr_protocol != IPPROTO_RSVP) 2007 return; 2008 2009 s = splsoftnet(); 2010 2011 /* 2012 * The socket may be attached to more than one vif...this 2013 * is perfectly legal. 2014 */ 2015 for (vifi = 0; vifi < numvifs; vifi++) { 2016 if (viftable[vifi].v_rsvpd == so) { 2017 viftable[vifi].v_rsvpd = 0; 2018 /* 2019 * This may seem silly, but we need to be sure we don't 2020 * over-decrement the RSVP counter, in case something 2021 * slips up. 2022 */ 2023 if (viftable[vifi].v_rsvp_on) { 2024 viftable[vifi].v_rsvp_on = 0; 2025 rsvp_on--; 2026 } 2027 } 2028 } 2029 2030 splx(s); 2031 return; 2032 } 2033 2034 void 2035 rsvp_input(m, ifp) 2036 struct mbuf *m; 2037 struct ifnet *ifp; 2038 { 2039 int vifi; 2040 struct ip *ip = mtod(m, struct ip *); 2041 static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET }; 2042 int s; 2043 2044 if (rsvpdebug) 2045 printf("rsvp_input: rsvp_on %d\n", rsvp_on); 2046 2047 /* 2048 * Can still get packets with rsvp_on = 0 if there is a local member 2049 * of the group to which the RSVP packet is addressed. But in this 2050 * case we want to throw the packet away. 2051 */ 2052 if (!rsvp_on) { 2053 m_freem(m); 2054 return; 2055 } 2056 2057 /* 2058 * If the old-style non-vif-associated socket is set, then use 2059 * it and ignore the new ones. 2060 */ 2061 if (ip_rsvpd != 0) { 2062 if (rsvpdebug) 2063 printf("rsvp_input: " 2064 "Sending packet up old-style socket\n"); 2065 rip_input(m); /*XXX*/ 2066 return; 2067 } 2068 2069 s = splsoftnet(); 2070 2071 if (rsvpdebug) 2072 printf("rsvp_input: check vifs\n"); 2073 2074 /* Find which vif the packet arrived on. */ 2075 for (vifi = 0; vifi < numvifs; vifi++) { 2076 if (viftable[vifi].v_ifp == ifp) 2077 break; 2078 } 2079 2080 if (vifi == numvifs) { 2081 /* Can't find vif packet arrived on. Drop packet. */ 2082 if (rsvpdebug) 2083 printf("rsvp_input: " 2084 "Can't find vif for packet...dropping it.\n"); 2085 m_freem(m); 2086 splx(s); 2087 return; 2088 } 2089 2090 if (rsvpdebug) 2091 printf("rsvp_input: check socket\n"); 2092 2093 if (viftable[vifi].v_rsvpd == 0) { 2094 /* 2095 * drop packet, since there is no specific socket for this 2096 * interface 2097 */ 2098 if (rsvpdebug) 2099 printf("rsvp_input: No socket defined for vif %d\n", 2100 vifi); 2101 m_freem(m); 2102 splx(s); 2103 return; 2104 } 2105 2106 rsvp_src.sin_addr = ip->ip_src; 2107 2108 if (rsvpdebug && m) 2109 printf("rsvp_input: m->m_len = %d, sbspace() = %d\n", 2110 m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv)); 2111 2112 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2113 if (rsvpdebug) 2114 printf("rsvp_input: Failed to append to socket\n"); 2115 else 2116 if (rsvpdebug) 2117 printf("rsvp_input: send packet up\n"); 2118 2119 splx(s); 2120 } 2121 #endif /* RSVP_ISI */ 2122