ip6_mroute.c revision 96972
1/* $FreeBSD: head/sys/netinet6/ip6_mroute.c 96972 2002-05-20 05:41:09Z tanimura $ */ 2/* $KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $ */ 3 4/* 5 * Copyright (C) 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33/* BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp */ 34 35/* 36 * IP multicast forwarding procedures 37 * 38 * Written by David Waitzman, BBN Labs, August 1988. 39 * Modified by Steve Deering, Stanford, February 1989. 40 * Modified by Mark J. Steiglitz, Stanford, May, 1991 41 * Modified by Van Jacobson, LBL, January 1993 42 * Modified by Ajit Thyagarajan, PARC, August 1993 43 * Modified by Bill Fenenr, PARC, April 1994 44 * 45 * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support 46 */ 47 48#include "opt_inet.h" 49#include "opt_inet6.h" 50 51#include <sys/param.h> 52#include <sys/callout.h> 53#include <sys/errno.h> 54#include <sys/kernel.h> 55#include <sys/lock.h> 56#include <sys/malloc.h> 57#include <sys/mbuf.h> 58#include <sys/protosw.h> 59#include <sys/signalvar.h> 60#include <sys/socket.h> 61#include <sys/socketvar.h> 62#include <sys/sockio.h> 63#include <sys/sx.h> 64#include <sys/syslog.h> 65#include <sys/systm.h> 66#include <sys/time.h> 67 68#include <net/if.h> 69#include <net/raw_cb.h> 70#include <net/route.h> 71 72#include <netinet/in.h> 73#include <netinet/in_var.h> 74 75#include <netinet/ip6.h> 76#include <netinet6/ip6_var.h> 77#include <netinet6/ip6_mroute.h> 78#include <netinet6/pim6.h> 79#include <netinet6/pim6_var.h> 80 81#include <net/net_osdep.h> 82 83static MALLOC_DEFINE(M_MRTABLE, "mf6c", "multicast forwarding cache entry"); 84 85#define M_HASCL(m) ((m)->m_flags & M_EXT) 86 87static int ip6_mdq __P((struct mbuf *, struct ifnet *, struct mf6c *)); 88static void phyint_send __P((struct ip6_hdr *, struct mif6 *, struct mbuf *)); 89 90static int set_pim6 __P((int *)); 91static int socket_send __P((struct socket *, struct mbuf *, 92 struct sockaddr_in6 *)); 93static int register_send __P((struct ip6_hdr *, struct mif6 *, 94 struct mbuf *)); 95 96/* 97 * Globals. All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static, 98 * except for netstat or debugging purposes. 99 */ 100struct socket *ip6_mrouter = NULL; 101int ip6_mrouter_ver = 0; 102int ip6_mrtproto = IPPROTO_PIM; /* for netstat only */ 103struct mrt6stat mrt6stat; 104 105#define NO_RTE_FOUND 0x1 106#define RTE_FOUND 0x2 107 108struct mf6c *mf6ctable[MF6CTBLSIZ]; 109u_char n6expire[MF6CTBLSIZ]; 110static struct mif6 mif6table[MAXMIFS]; 111#ifdef MRT6DEBUG 112u_int mrt6debug = 0; /* debug level */ 113#define DEBUG_MFC 0x02 114#define DEBUG_FORWARD 0x04 115#define DEBUG_EXPIRE 0x08 116#define DEBUG_XMIT 0x10 117#define DEBUG_REG 0x20 118#define DEBUG_PIM 0x40 119#endif 120 121static void expire_upcalls __P((void *)); 122#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 123#define UPCALL_EXPIRE 6 /* number of timeouts */ 124 125#ifdef INET 126#ifdef MROUTING 127extern struct socket *ip_mrouter; 128#endif 129#endif 130 131/* 132 * 'Interfaces' associated with decapsulator (so we can tell 133 * packets that went through it from ones that get reflected 134 * by a broken gateway). These interfaces are never linked into 135 * the system ifnet list & no routes point to them. I.e., packets 136 * can't be sent this way. They only exist as a placeholder for 137 * multicast source verification. 138 */ 139struct ifnet multicast_register_if; 140 141#define ENCAP_HOPS 64 142 143/* 144 * Private variables. 145 */ 146static mifi_t nummifs = 0; 147static mifi_t reg_mif_num = (mifi_t)-1; 148 149static struct pim6stat pim6stat; 150static int pim6; 151 152/* 153 * Hash function for a source, group entry 154 */ 155#define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \ 156 (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \ 157 (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \ 158 (g).s6_addr32[2] ^ (g).s6_addr32[3]) 159 160/* 161 * Find a route for a given origin IPv6 address and Multicast group address. 162 * Quality of service parameter to be added in the future!!! 163 */ 164 165#define MF6CFIND(o, g, rt) do { \ 166 struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \ 167 rt = NULL; \ 168 mrt6stat.mrt6s_mfc_lookups++; \ 169 while (_rt) { \ 170 if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \ 171 IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \ 172 (_rt->mf6c_stall == NULL)) { \ 173 rt = _rt; \ 174 break; \ 175 } \ 176 _rt = _rt->mf6c_next; \ 177 } \ 178 if (rt == NULL) { \ 179 mrt6stat.mrt6s_mfc_misses++; \ 180 } \ 181} while (0) 182 183/* 184 * Macros to compute elapsed time efficiently 185 * Borrowed from Van Jacobson's scheduling code 186 */ 187#define TV_DELTA(a, b, delta) do { \ 188 int xxs; \ 189 \ 190 delta = (a).tv_usec - (b).tv_usec; \ 191 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 192 switch (xxs) { \ 193 case 2: \ 194 delta += 1000000; \ 195 /* fall through */ \ 196 case 1: \ 197 delta += 1000000; \ 198 break; \ 199 default: \ 200 delta += (1000000 * xxs); \ 201 } \ 202 } \ 203} while (0) 204 205#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 206 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 207 208#ifdef UPCALL_TIMING 209#define UPCALL_MAX 50 210u_long upcall_data[UPCALL_MAX + 1]; 211static void collate(); 212#endif /* UPCALL_TIMING */ 213 214static int get_sg_cnt __P((struct sioc_sg_req6 *)); 215static int get_mif6_cnt __P((struct sioc_mif_req6 *)); 216static int ip6_mrouter_init __P((struct socket *, struct mbuf *, int)); 217static int add_m6if __P((struct mif6ctl *)); 218static int del_m6if __P((mifi_t *)); 219static int add_m6fc __P((struct mf6cctl *)); 220static int del_m6fc __P((struct mf6cctl *)); 221 222static struct callout expire_upcalls_ch; 223 224/* 225 * Handle MRT setsockopt commands to modify the multicast routing tables. 226 */ 227int 228ip6_mrouter_set(so, sopt) 229 struct socket *so; 230 struct sockopt *sopt; 231{ 232 int error = 0; 233 struct mbuf *m; 234 235 if (so != ip6_mrouter && sopt->sopt_name != MRT6_INIT) 236 return (EACCES); 237 238 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 239 return (error); 240 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 241 return (error); 242 243 switch (sopt->sopt_name) { 244 case MRT6_INIT: 245#ifdef MRT6_OINIT 246 case MRT6_OINIT: 247#endif 248 error = ip6_mrouter_init(so, m, sopt->sopt_name); 249 break; 250 case MRT6_DONE: 251 error = ip6_mrouter_done(); 252 break; 253 case MRT6_ADD_MIF: 254 error = add_m6if(mtod(m, struct mif6ctl *)); 255 break; 256 case MRT6_DEL_MIF: 257 error = del_m6if(mtod(m, mifi_t *)); 258 break; 259 case MRT6_ADD_MFC: 260 error = add_m6fc(mtod(m, struct mf6cctl *)); 261 break; 262 case MRT6_DEL_MFC: 263 error = del_m6fc(mtod(m, struct mf6cctl *)); 264 break; 265 case MRT6_PIM: 266 error = set_pim6(mtod(m, int *)); 267 break; 268 default: 269 error = EOPNOTSUPP; 270 break; 271 } 272 273 (void)m_freem(m); 274 return(error); 275} 276 277/* 278 * Handle MRT getsockopt commands 279 */ 280int 281ip6_mrouter_get(so, sopt) 282 struct socket *so; 283 struct sockopt *sopt; 284{ 285 int error = 0; 286 287 if (so != ip6_mrouter) return EACCES; 288 289 switch (sopt->sopt_name) { 290 case MRT6_PIM: 291 error = sooptcopyout(sopt, &pim6, sizeof(pim6)); 292 break; 293 } 294 return (error); 295} 296 297/* 298 * Handle ioctl commands to obtain information from the cache 299 */ 300int 301mrt6_ioctl(cmd, data) 302 int cmd; 303 caddr_t data; 304{ 305 int error = 0; 306 307 switch (cmd) { 308 case SIOCGETSGCNT_IN6: 309 return(get_sg_cnt((struct sioc_sg_req6 *)data)); 310 break; /* for safety */ 311 case SIOCGETMIFCNT_IN6: 312 return(get_mif6_cnt((struct sioc_mif_req6 *)data)); 313 break; /* for safety */ 314 default: 315 return (EINVAL); 316 break; 317 } 318 return error; 319} 320 321/* 322 * returns the packet, byte, rpf-failure count for the source group provided 323 */ 324static int 325get_sg_cnt(req) 326 struct sioc_sg_req6 *req; 327{ 328 struct mf6c *rt; 329 int s; 330 331 s = splnet(); 332 MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt); 333 splx(s); 334 if (rt != NULL) { 335 req->pktcnt = rt->mf6c_pkt_cnt; 336 req->bytecnt = rt->mf6c_byte_cnt; 337 req->wrong_if = rt->mf6c_wrong_if; 338 } else 339 return(ESRCH); 340#if 0 341 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 342#endif 343 344 return 0; 345} 346 347/* 348 * returns the input and output packet and byte counts on the mif provided 349 */ 350static int 351get_mif6_cnt(req) 352 struct sioc_mif_req6 *req; 353{ 354 mifi_t mifi = req->mifi; 355 356 if (mifi >= nummifs) 357 return EINVAL; 358 359 req->icount = mif6table[mifi].m6_pkt_in; 360 req->ocount = mif6table[mifi].m6_pkt_out; 361 req->ibytes = mif6table[mifi].m6_bytes_in; 362 req->obytes = mif6table[mifi].m6_bytes_out; 363 364 return 0; 365} 366 367static int 368set_pim6(i) 369 int *i; 370{ 371 if ((*i != 1) && (*i != 0)) 372 return EINVAL; 373 374 pim6 = *i; 375 376 return 0; 377} 378 379/* 380 * Enable multicast routing 381 */ 382static int 383ip6_mrouter_init(so, m, cmd) 384 struct socket *so; 385 struct mbuf *m; 386 int cmd; 387{ 388 int *v; 389 390#ifdef MRT6DEBUG 391 if (mrt6debug) 392 log(LOG_DEBUG, 393 "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n", 394 so->so_type, so->so_proto->pr_protocol); 395#endif 396 397 if (so->so_type != SOCK_RAW || 398 so->so_proto->pr_protocol != IPPROTO_ICMPV6) 399 return EOPNOTSUPP; 400 401 if (!m || (m->m_len != sizeof(int *))) 402 return ENOPROTOOPT; 403 404 v = mtod(m, int *); 405 if (*v != 1) 406 return ENOPROTOOPT; 407 408 if (ip6_mrouter != NULL) return EADDRINUSE; 409 410 ip6_mrouter = so; 411 ip6_mrouter_ver = cmd; 412 413 bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); 414 bzero((caddr_t)n6expire, sizeof(n6expire)); 415 416 pim6 = 0;/* used for stubbing out/in pim stuff */ 417 418 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 419 expire_upcalls, NULL); 420 421#ifdef MRT6DEBUG 422 if (mrt6debug) 423 log(LOG_DEBUG, "ip6_mrouter_init\n"); 424#endif 425 426 return 0; 427} 428 429/* 430 * Disable multicast routing 431 */ 432int 433ip6_mrouter_done() 434{ 435 mifi_t mifi; 436 int i; 437 struct ifnet *ifp; 438 struct in6_ifreq ifr; 439 struct mf6c *rt; 440 struct rtdetq *rte; 441 int s; 442 443 s = splnet(); 444 445 /* 446 * For each phyint in use, disable promiscuous reception of all IPv6 447 * multicasts. 448 */ 449#ifdef INET 450#ifdef MROUTING 451 /* 452 * If there is still IPv4 multicast routing daemon, 453 * we remain interfaces to receive all muliticasted packets. 454 * XXX: there may be an interface in which the IPv4 multicast 455 * daemon is not interested... 456 */ 457 if (!ip_mrouter) 458#endif 459#endif 460 { 461 for (mifi = 0; mifi < nummifs; mifi++) { 462 if (mif6table[mifi].m6_ifp && 463 !(mif6table[mifi].m6_flags & MIFF_REGISTER)) { 464 ifr.ifr_addr.sin6_family = AF_INET6; 465 ifr.ifr_addr.sin6_addr= in6addr_any; 466 ifp = mif6table[mifi].m6_ifp; 467 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, 468 (caddr_t)&ifr); 469 } 470 } 471 } 472#ifdef notyet 473 bzero((caddr_t)qtable, sizeof(qtable)); 474 bzero((caddr_t)tbftable, sizeof(tbftable)); 475#endif 476 bzero((caddr_t)mif6table, sizeof(mif6table)); 477 nummifs = 0; 478 479 pim6 = 0; /* used to stub out/in pim specific code */ 480 481 callout_stop(&expire_upcalls_ch); 482 483 /* 484 * Free all multicast forwarding cache entries. 485 */ 486 for (i = 0; i < MF6CTBLSIZ; i++) { 487 rt = mf6ctable[i]; 488 while (rt) { 489 struct mf6c *frt; 490 491 for (rte = rt->mf6c_stall; rte != NULL; ) { 492 struct rtdetq *n = rte->next; 493 494 m_free(rte->m); 495 free(rte, M_MRTABLE); 496 rte = n; 497 } 498 frt = rt; 499 rt = rt->mf6c_next; 500 free(frt, M_MRTABLE); 501 } 502 } 503 504 bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); 505 506 /* 507 * Reset de-encapsulation cache 508 */ 509 reg_mif_num = -1; 510 511 ip6_mrouter = NULL; 512 ip6_mrouter_ver = 0; 513 514 splx(s); 515 516#ifdef MRT6DEBUG 517 if (mrt6debug) 518 log(LOG_DEBUG, "ip6_mrouter_done\n"); 519#endif 520 521 return 0; 522} 523 524static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; 525 526/* 527 * Add a mif to the mif table 528 */ 529static int 530add_m6if(mifcp) 531 struct mif6ctl *mifcp; 532{ 533 struct mif6 *mifp; 534 struct ifnet *ifp; 535 int error, s; 536#ifdef notyet 537 struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi; 538#endif 539 540 if (mifcp->mif6c_mifi >= MAXMIFS) 541 return EINVAL; 542 mifp = mif6table + mifcp->mif6c_mifi; 543 if (mifp->m6_ifp) 544 return EADDRINUSE; /* XXX: is it appropriate? */ 545 if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index) 546 return ENXIO; 547 ifp = ifnet_byindex(mifcp->mif6c_pifi); 548 549 if (mifcp->mif6c_flags & MIFF_REGISTER) { 550 if (reg_mif_num == (mifi_t)-1) { 551 multicast_register_if.if_name = "register_mif"; 552 multicast_register_if.if_flags |= IFF_LOOPBACK; 553 multicast_register_if.if_index = mifcp->mif6c_mifi; 554 reg_mif_num = mifcp->mif6c_mifi; 555 } 556 557 ifp = &multicast_register_if; 558 559 } /* if REGISTER */ 560 else { 561 /* Make sure the interface supports multicast */ 562 if ((ifp->if_flags & IFF_MULTICAST) == 0) 563 return EOPNOTSUPP; 564 565 s = splnet(); 566 error = if_allmulti(ifp, 1); 567 splx(s); 568 if (error) 569 return error; 570 } 571 572 s = splnet(); 573 mifp->m6_flags = mifcp->mif6c_flags; 574 mifp->m6_ifp = ifp; 575#ifdef notyet 576 /* scaling up here allows division by 1024 in critical code */ 577 mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000; 578#endif 579 /* initialize per mif pkt counters */ 580 mifp->m6_pkt_in = 0; 581 mifp->m6_pkt_out = 0; 582 mifp->m6_bytes_in = 0; 583 mifp->m6_bytes_out = 0; 584 splx(s); 585 586 /* Adjust nummifs up if the mifi is higher than nummifs */ 587 if (nummifs <= mifcp->mif6c_mifi) 588 nummifs = mifcp->mif6c_mifi + 1; 589 590#ifdef MRT6DEBUG 591 if (mrt6debug) 592 log(LOG_DEBUG, 593 "add_mif #%d, phyint %s%d\n", 594 mifcp->mif6c_mifi, 595 ifp->if_name, ifp->if_unit); 596#endif 597 598 return 0; 599} 600 601/* 602 * Delete a mif from the mif table 603 */ 604static int 605del_m6if(mifip) 606 mifi_t *mifip; 607{ 608 struct mif6 *mifp = mif6table + *mifip; 609 mifi_t mifi; 610 struct ifnet *ifp; 611 int s; 612 613 if (*mifip >= nummifs) 614 return EINVAL; 615 if (mifp->m6_ifp == NULL) 616 return EINVAL; 617 618 s = splnet(); 619 620 if (!(mifp->m6_flags & MIFF_REGISTER)) { 621 /* 622 * XXX: what if there is yet IPv4 multicast daemon 623 * using the interface? 624 */ 625 ifp = mifp->m6_ifp; 626 627 if_allmulti(ifp, 0); 628 } 629 630#ifdef notyet 631 bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip])); 632 bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf))); 633#endif 634 bzero((caddr_t)mifp, sizeof (*mifp)); 635 636 /* Adjust nummifs down */ 637 for (mifi = nummifs; mifi > 0; mifi--) 638 if (mif6table[mifi - 1].m6_ifp) 639 break; 640 nummifs = mifi; 641 642 splx(s); 643 644#ifdef MRT6DEBUG 645 if (mrt6debug) 646 log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs); 647#endif 648 649 return 0; 650} 651 652/* 653 * Add an mfc entry 654 */ 655static int 656add_m6fc(mfccp) 657 struct mf6cctl *mfccp; 658{ 659 struct mf6c *rt; 660 u_long hash; 661 struct rtdetq *rte; 662 u_short nstl; 663 int s; 664 665 MF6CFIND(mfccp->mf6cc_origin.sin6_addr, 666 mfccp->mf6cc_mcastgrp.sin6_addr, rt); 667 668 /* If an entry already exists, just update the fields */ 669 if (rt) { 670#ifdef MRT6DEBUG 671 if (mrt6debug & DEBUG_MFC) 672 log(LOG_DEBUG, 673 "add_m6fc no upcall h %d o %s g %s p %x\n", 674 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 675 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 676 mfccp->mf6cc_parent); 677#endif 678 679 s = splnet(); 680 rt->mf6c_parent = mfccp->mf6cc_parent; 681 rt->mf6c_ifset = mfccp->mf6cc_ifset; 682 splx(s); 683 return 0; 684 } 685 686 /* 687 * Find the entry for which the upcall was made and update 688 */ 689 s = splnet(); 690 hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr, 691 mfccp->mf6cc_mcastgrp.sin6_addr); 692 for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) { 693 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, 694 &mfccp->mf6cc_origin.sin6_addr) && 695 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, 696 &mfccp->mf6cc_mcastgrp.sin6_addr) && 697 (rt->mf6c_stall != NULL)) { 698 699 if (nstl++) 700 log(LOG_ERR, 701 "add_m6fc: %s o %s g %s p %x dbx %p\n", 702 "multiple kernel entries", 703 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 704 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 705 mfccp->mf6cc_parent, rt->mf6c_stall); 706 707#ifdef MRT6DEBUG 708 if (mrt6debug & DEBUG_MFC) 709 log(LOG_DEBUG, 710 "add_m6fc o %s g %s p %x dbg %x\n", 711 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 712 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 713 mfccp->mf6cc_parent, rt->mf6c_stall); 714#endif 715 716 rt->mf6c_origin = mfccp->mf6cc_origin; 717 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 718 rt->mf6c_parent = mfccp->mf6cc_parent; 719 rt->mf6c_ifset = mfccp->mf6cc_ifset; 720 /* initialize pkt counters per src-grp */ 721 rt->mf6c_pkt_cnt = 0; 722 rt->mf6c_byte_cnt = 0; 723 rt->mf6c_wrong_if = 0; 724 725 rt->mf6c_expire = 0; /* Don't clean this guy up */ 726 n6expire[hash]--; 727 728 /* free packets Qed at the end of this entry */ 729 for (rte = rt->mf6c_stall; rte != NULL; ) { 730 struct rtdetq *n = rte->next; 731 ip6_mdq(rte->m, rte->ifp, rt); 732 m_freem(rte->m); 733#ifdef UPCALL_TIMING 734 collate(&(rte->t)); 735#endif /* UPCALL_TIMING */ 736 free(rte, M_MRTABLE); 737 rte = n; 738 } 739 rt->mf6c_stall = NULL; 740 } 741 } 742 743 /* 744 * It is possible that an entry is being inserted without an upcall 745 */ 746 if (nstl == 0) { 747#ifdef MRT6DEBUG 748 if (mrt6debug & DEBUG_MFC) 749 log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n", 750 hash, 751 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 752 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 753 mfccp->mf6cc_parent); 754#endif 755 756 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { 757 758 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, 759 &mfccp->mf6cc_origin.sin6_addr)&& 760 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, 761 &mfccp->mf6cc_mcastgrp.sin6_addr)) { 762 763 rt->mf6c_origin = mfccp->mf6cc_origin; 764 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 765 rt->mf6c_parent = mfccp->mf6cc_parent; 766 rt->mf6c_ifset = mfccp->mf6cc_ifset; 767 /* initialize pkt counters per src-grp */ 768 rt->mf6c_pkt_cnt = 0; 769 rt->mf6c_byte_cnt = 0; 770 rt->mf6c_wrong_if = 0; 771 772 if (rt->mf6c_expire) 773 n6expire[hash]--; 774 rt->mf6c_expire = 0; 775 } 776 } 777 if (rt == NULL) { 778 /* no upcall, so make a new entry */ 779 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE, 780 M_NOWAIT); 781 if (rt == NULL) { 782 splx(s); 783 return ENOBUFS; 784 } 785 786 /* insert new entry at head of hash chain */ 787 rt->mf6c_origin = mfccp->mf6cc_origin; 788 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 789 rt->mf6c_parent = mfccp->mf6cc_parent; 790 rt->mf6c_ifset = mfccp->mf6cc_ifset; 791 /* initialize pkt counters per src-grp */ 792 rt->mf6c_pkt_cnt = 0; 793 rt->mf6c_byte_cnt = 0; 794 rt->mf6c_wrong_if = 0; 795 rt->mf6c_expire = 0; 796 rt->mf6c_stall = NULL; 797 798 /* link into table */ 799 rt->mf6c_next = mf6ctable[hash]; 800 mf6ctable[hash] = rt; 801 } 802 } 803 splx(s); 804 return 0; 805} 806 807#ifdef UPCALL_TIMING 808/* 809 * collect delay statistics on the upcalls 810 */ 811static void 812collate(t) 813 struct timeval *t; 814{ 815 u_long d; 816 struct timeval tp; 817 u_long delta; 818 819 GET_TIME(tp); 820 821 if (TV_LT(*t, tp)) 822 { 823 TV_DELTA(tp, *t, delta); 824 825 d = delta >> 10; 826 if (d > UPCALL_MAX) 827 d = UPCALL_MAX; 828 829 ++upcall_data[d]; 830 } 831} 832#endif /* UPCALL_TIMING */ 833 834/* 835 * Delete an mfc entry 836 */ 837static int 838del_m6fc(mfccp) 839 struct mf6cctl *mfccp; 840{ 841 struct sockaddr_in6 origin; 842 struct sockaddr_in6 mcastgrp; 843 struct mf6c *rt; 844 struct mf6c **nptr; 845 u_long hash; 846 int s; 847 848 origin = mfccp->mf6cc_origin; 849 mcastgrp = mfccp->mf6cc_mcastgrp; 850 hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr); 851 852#ifdef MRT6DEBUG 853 if (mrt6debug & DEBUG_MFC) 854 log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n", 855 ip6_sprintf(&origin.sin6_addr), 856 ip6_sprintf(&mcastgrp.sin6_addr)); 857#endif 858 859 s = splnet(); 860 861 nptr = &mf6ctable[hash]; 862 while ((rt = *nptr) != NULL) { 863 if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr, 864 &rt->mf6c_origin.sin6_addr) && 865 IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr, 866 &rt->mf6c_mcastgrp.sin6_addr) && 867 rt->mf6c_stall == NULL) 868 break; 869 870 nptr = &rt->mf6c_next; 871 } 872 if (rt == NULL) { 873 splx(s); 874 return EADDRNOTAVAIL; 875 } 876 877 *nptr = rt->mf6c_next; 878 free(rt, M_MRTABLE); 879 880 splx(s); 881 882 return 0; 883} 884 885static int 886socket_send(s, mm, src) 887 struct socket *s; 888 struct mbuf *mm; 889 struct sockaddr_in6 *src; 890{ 891 if (s) { 892 if (sbappendaddr(&s->so_rcv, 893 (struct sockaddr *)src, 894 mm, (struct mbuf *)0) != 0) { 895 SOCK_LOCK(s); 896 sorwakeup(s); 897 SOCK_UNLOCK(s); 898 return 0; 899 } 900 } 901 m_freem(mm); 902 return -1; 903} 904 905/* 906 * IPv6 multicast forwarding function. This function assumes that the packet 907 * pointed to by "ip6" has arrived on (or is about to be sent to) the interface 908 * pointed to by "ifp", and the packet is to be relayed to other networks 909 * that have members of the packet's destination IPv6 multicast group. 910 * 911 * The packet is returned unscathed to the caller, unless it is 912 * erroneous, in which case a non-zero return value tells the caller to 913 * discard it. 914 */ 915 916int 917ip6_mforward(ip6, ifp, m) 918 struct ip6_hdr *ip6; 919 struct ifnet *ifp; 920 struct mbuf *m; 921{ 922 struct mf6c *rt; 923 struct mif6 *mifp; 924 struct mbuf *mm; 925 int s; 926 mifi_t mifi; 927 928#ifdef MRT6DEBUG 929 if (mrt6debug & DEBUG_FORWARD) 930 log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n", 931 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst), 932 ifp->if_index); 933#endif 934 935 /* 936 * Don't forward a packet with Hop limit of zero or one, 937 * or a packet destined to a local-only group. 938 */ 939 if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) || 940 IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst)) 941 return 0; 942 ip6->ip6_hlim--; 943 944 /* 945 * Source address check: do not forward packets with unspecified 946 * source. It was discussed in July 2000, on ipngwg mailing list. 947 * This is rather more serious than unicast cases, because some 948 * MLD packets can be sent with the unspecified source address 949 * (although such packets must normally set 1 to the hop limit field). 950 */ 951 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 952 ip6stat.ip6s_cantforward++; 953 if (ip6_log_time + ip6_log_interval < time_second) { 954 ip6_log_time = time_second; 955 log(LOG_DEBUG, 956 "cannot forward " 957 "from %s to %s nxt %d received on %s\n", 958 ip6_sprintf(&ip6->ip6_src), 959 ip6_sprintf(&ip6->ip6_dst), 960 ip6->ip6_nxt, 961 if_name(m->m_pkthdr.rcvif)); 962 } 963 return 0; 964 } 965 966 /* 967 * Determine forwarding mifs from the forwarding cache table 968 */ 969 s = splnet(); 970 MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt); 971 972 /* Entry exists, so forward if necessary */ 973 if (rt) { 974 splx(s); 975 return (ip6_mdq(m, ifp, rt)); 976 } else { 977 /* 978 * If we don't have a route for packet's origin, 979 * Make a copy of the packet & 980 * send message to routing daemon 981 */ 982 983 struct mbuf *mb0; 984 struct rtdetq *rte; 985 u_long hash; 986/* int i, npkts;*/ 987#ifdef UPCALL_TIMING 988 struct timeval tp; 989 990 GET_TIME(tp); 991#endif /* UPCALL_TIMING */ 992 993 mrt6stat.mrt6s_no_route++; 994#ifdef MRT6DEBUG 995 if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC)) 996 log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n", 997 ip6_sprintf(&ip6->ip6_src), 998 ip6_sprintf(&ip6->ip6_dst)); 999#endif 1000 1001 /* 1002 * Allocate mbufs early so that we don't do extra work if we 1003 * are just going to fail anyway. 1004 */ 1005 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, 1006 M_NOWAIT); 1007 if (rte == NULL) { 1008 splx(s); 1009 return ENOBUFS; 1010 } 1011 mb0 = m_copy(m, 0, M_COPYALL); 1012 /* 1013 * Pullup packet header if needed before storing it, 1014 * as other references may modify it in the meantime. 1015 */ 1016 if (mb0 && 1017 (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr))) 1018 mb0 = m_pullup(mb0, sizeof(struct ip6_hdr)); 1019 if (mb0 == NULL) { 1020 free(rte, M_MRTABLE); 1021 splx(s); 1022 return ENOBUFS; 1023 } 1024 1025 /* is there an upcall waiting for this packet? */ 1026 hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst); 1027 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { 1028 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, 1029 &rt->mf6c_origin.sin6_addr) && 1030 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 1031 &rt->mf6c_mcastgrp.sin6_addr) && 1032 (rt->mf6c_stall != NULL)) 1033 break; 1034 } 1035 1036 if (rt == NULL) { 1037 struct mrt6msg *im; 1038#ifdef MRT6_OINIT 1039 struct omrt6msg *oim; 1040#endif 1041 1042 /* no upcall, so make a new entry */ 1043 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE, 1044 M_NOWAIT); 1045 if (rt == NULL) { 1046 free(rte, M_MRTABLE); 1047 m_freem(mb0); 1048 splx(s); 1049 return ENOBUFS; 1050 } 1051 /* 1052 * Make a copy of the header to send to the user 1053 * level process 1054 */ 1055 mm = m_copy(mb0, 0, sizeof(struct ip6_hdr)); 1056 1057 if (mm == NULL) { 1058 free(rte, M_MRTABLE); 1059 m_freem(mb0); 1060 free(rt, M_MRTABLE); 1061 splx(s); 1062 return ENOBUFS; 1063 } 1064 1065 /* 1066 * Send message to routing daemon 1067 */ 1068 sin6.sin6_addr = ip6->ip6_src; 1069 1070 im = NULL; 1071#ifdef MRT6_OINIT 1072 oim = NULL; 1073#endif 1074 switch (ip6_mrouter_ver) { 1075#ifdef MRT6_OINIT 1076 case MRT6_OINIT: 1077 oim = mtod(mm, struct omrt6msg *); 1078 oim->im6_msgtype = MRT6MSG_NOCACHE; 1079 oim->im6_mbz = 0; 1080 break; 1081#endif 1082 case MRT6_INIT: 1083 im = mtod(mm, struct mrt6msg *); 1084 im->im6_msgtype = MRT6MSG_NOCACHE; 1085 im->im6_mbz = 0; 1086 break; 1087 default: 1088 free(rte, M_MRTABLE); 1089 m_freem(mb0); 1090 free(rt, M_MRTABLE); 1091 splx(s); 1092 return EINVAL; 1093 } 1094 1095#ifdef MRT6DEBUG 1096 if (mrt6debug & DEBUG_FORWARD) 1097 log(LOG_DEBUG, 1098 "getting the iif info in the kernel\n"); 1099#endif 1100 1101 for (mifp = mif6table, mifi = 0; 1102 mifi < nummifs && mifp->m6_ifp != ifp; 1103 mifp++, mifi++) 1104 ; 1105 1106 switch (ip6_mrouter_ver) { 1107#ifdef MRT6_OINIT 1108 case MRT6_OINIT: 1109 oim->im6_mif = mifi; 1110 break; 1111#endif 1112 case MRT6_INIT: 1113 im->im6_mif = mifi; 1114 break; 1115 } 1116 1117 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1118 log(LOG_WARNING, "ip6_mforward: ip6_mrouter " 1119 "socket queue full\n"); 1120 mrt6stat.mrt6s_upq_sockfull++; 1121 free(rte, M_MRTABLE); 1122 m_freem(mb0); 1123 free(rt, M_MRTABLE); 1124 splx(s); 1125 return ENOBUFS; 1126 } 1127 1128 mrt6stat.mrt6s_upcalls++; 1129 1130 /* insert new entry at head of hash chain */ 1131 bzero(rt, sizeof(*rt)); 1132 rt->mf6c_origin.sin6_family = AF_INET6; 1133 rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6); 1134 rt->mf6c_origin.sin6_addr = ip6->ip6_src; 1135 rt->mf6c_mcastgrp.sin6_family = AF_INET6; 1136 rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6); 1137 rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst; 1138 rt->mf6c_expire = UPCALL_EXPIRE; 1139 n6expire[hash]++; 1140 rt->mf6c_parent = MF6C_INCOMPLETE_PARENT; 1141 1142 /* link into table */ 1143 rt->mf6c_next = mf6ctable[hash]; 1144 mf6ctable[hash] = rt; 1145 /* Add this entry to the end of the queue */ 1146 rt->mf6c_stall = rte; 1147 } else { 1148 /* determine if q has overflowed */ 1149 struct rtdetq **p; 1150 int npkts = 0; 1151 1152 for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next) 1153 if (++npkts > MAX_UPQ6) { 1154 mrt6stat.mrt6s_upq_ovflw++; 1155 free(rte, M_MRTABLE); 1156 m_freem(mb0); 1157 splx(s); 1158 return 0; 1159 } 1160 1161 /* Add this entry to the end of the queue */ 1162 *p = rte; 1163 } 1164 1165 rte->next = NULL; 1166 rte->m = mb0; 1167 rte->ifp = ifp; 1168#ifdef UPCALL_TIMING 1169 rte->t = tp; 1170#endif /* UPCALL_TIMING */ 1171 1172 splx(s); 1173 1174 return 0; 1175 } 1176} 1177 1178/* 1179 * Clean up cache entries if upcalls are not serviced 1180 * Call from the Slow Timeout mechanism, every half second. 1181 */ 1182static void 1183expire_upcalls(unused) 1184 void *unused; 1185{ 1186 struct rtdetq *rte; 1187 struct mf6c *mfc, **nptr; 1188 int i; 1189 int s; 1190 1191 s = splnet(); 1192 for (i = 0; i < MF6CTBLSIZ; i++) { 1193 if (n6expire[i] == 0) 1194 continue; 1195 nptr = &mf6ctable[i]; 1196 while ((mfc = *nptr) != NULL) { 1197 rte = mfc->mf6c_stall; 1198 /* 1199 * Skip real cache entries 1200 * Make sure it wasn't marked to not expire (shouldn't happen) 1201 * If it expires now 1202 */ 1203 if (rte != NULL && 1204 mfc->mf6c_expire != 0 && 1205 --mfc->mf6c_expire == 0) { 1206#ifdef MRT6DEBUG 1207 if (mrt6debug & DEBUG_EXPIRE) 1208 log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n", 1209 ip6_sprintf(&mfc->mf6c_origin.sin6_addr), 1210 ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr)); 1211#endif 1212 /* 1213 * drop all the packets 1214 * free the mbuf with the pkt, if, timing info 1215 */ 1216 do { 1217 struct rtdetq *n = rte->next; 1218 m_freem(rte->m); 1219 free(rte, M_MRTABLE); 1220 rte = n; 1221 } while (rte != NULL); 1222 mrt6stat.mrt6s_cache_cleanups++; 1223 n6expire[i]--; 1224 1225 *nptr = mfc->mf6c_next; 1226 free(mfc, M_MRTABLE); 1227 } else { 1228 nptr = &mfc->mf6c_next; 1229 } 1230 } 1231 } 1232 splx(s); 1233 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1234 expire_upcalls, NULL); 1235} 1236 1237/* 1238 * Packet forwarding routine once entry in the cache is made 1239 */ 1240static int 1241ip6_mdq(m, ifp, rt) 1242 struct mbuf *m; 1243 struct ifnet *ifp; 1244 struct mf6c *rt; 1245{ 1246 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1247 mifi_t mifi, iif; 1248 struct mif6 *mifp; 1249 int plen = m->m_pkthdr.len; 1250 1251/* 1252 * Macro to send packet on mif. Since RSVP packets don't get counted on 1253 * input, they shouldn't get counted on output, so statistics keeping is 1254 * separate. 1255 */ 1256 1257#define MC6_SEND(ip6, mifp, m) do { \ 1258 if ((mifp)->m6_flags & MIFF_REGISTER) \ 1259 register_send((ip6), (mifp), (m)); \ 1260 else \ 1261 phyint_send((ip6), (mifp), (m)); \ 1262} while (0) 1263 1264 /* 1265 * Don't forward if it didn't arrive from the parent mif 1266 * for its origin. 1267 */ 1268 mifi = rt->mf6c_parent; 1269 if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) { 1270 /* came in the wrong interface */ 1271#ifdef MRT6DEBUG 1272 if (mrt6debug & DEBUG_FORWARD) 1273 log(LOG_DEBUG, 1274 "wrong if: ifid %d mifi %d mififid %x\n", 1275 ifp->if_index, mifi, 1276 mif6table[mifi].m6_ifp->if_index); 1277#endif 1278 mrt6stat.mrt6s_wrong_if++; 1279 rt->mf6c_wrong_if++; 1280 /* 1281 * If we are doing PIM processing, and we are forwarding 1282 * packets on this interface, send a message to the 1283 * routing daemon. 1284 */ 1285 /* have to make sure this is a valid mif */ 1286 if (mifi < nummifs && mif6table[mifi].m6_ifp) 1287 if (pim6 && (m->m_flags & M_LOOP) == 0) { 1288 /* 1289 * Check the M_LOOP flag to avoid an 1290 * unnecessary PIM assert. 1291 * XXX: M_LOOP is an ad-hoc hack... 1292 */ 1293 static struct sockaddr_in6 sin6 = 1294 { sizeof(sin6), AF_INET6 }; 1295 1296 struct mbuf *mm; 1297 struct mrt6msg *im; 1298#ifdef MRT6_OINIT 1299 struct omrt6msg *oim; 1300#endif 1301 1302 mm = m_copy(m, 0, sizeof(struct ip6_hdr)); 1303 if (mm && 1304 (M_HASCL(mm) || 1305 mm->m_len < sizeof(struct ip6_hdr))) 1306 mm = m_pullup(mm, sizeof(struct ip6_hdr)); 1307 if (mm == NULL) 1308 return ENOBUFS; 1309 1310#ifdef MRT6_OINIT 1311 oim = NULL; 1312#endif 1313 im = NULL; 1314 switch (ip6_mrouter_ver) { 1315#ifdef MRT6_OINIT 1316 case MRT6_OINIT: 1317 oim = mtod(mm, struct omrt6msg *); 1318 oim->im6_msgtype = MRT6MSG_WRONGMIF; 1319 oim->im6_mbz = 0; 1320 break; 1321#endif 1322 case MRT6_INIT: 1323 im = mtod(mm, struct mrt6msg *); 1324 im->im6_msgtype = MRT6MSG_WRONGMIF; 1325 im->im6_mbz = 0; 1326 break; 1327 default: 1328 m_freem(mm); 1329 return EINVAL; 1330 } 1331 1332 for (mifp = mif6table, iif = 0; 1333 iif < nummifs && mifp && 1334 mifp->m6_ifp != ifp; 1335 mifp++, iif++) 1336 ; 1337 1338 switch (ip6_mrouter_ver) { 1339#ifdef MRT6_OINIT 1340 case MRT6_OINIT: 1341 oim->im6_mif = iif; 1342 sin6.sin6_addr = oim->im6_src; 1343 break; 1344#endif 1345 case MRT6_INIT: 1346 im->im6_mif = iif; 1347 sin6.sin6_addr = im->im6_src; 1348 break; 1349 } 1350 1351 mrt6stat.mrt6s_upcalls++; 1352 1353 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1354#ifdef MRT6DEBUG 1355 if (mrt6debug) 1356 log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n"); 1357#endif 1358 ++mrt6stat.mrt6s_upq_sockfull; 1359 return ENOBUFS; 1360 } /* if socket Q full */ 1361 } /* if PIM */ 1362 return 0; 1363 } /* if wrong iif */ 1364 1365 /* If I sourced this packet, it counts as output, else it was input. */ 1366 if (m->m_pkthdr.rcvif == NULL) { 1367 /* XXX: is rcvif really NULL when output?? */ 1368 mif6table[mifi].m6_pkt_out++; 1369 mif6table[mifi].m6_bytes_out += plen; 1370 } else { 1371 mif6table[mifi].m6_pkt_in++; 1372 mif6table[mifi].m6_bytes_in += plen; 1373 } 1374 rt->mf6c_pkt_cnt++; 1375 rt->mf6c_byte_cnt += plen; 1376 1377 /* 1378 * For each mif, forward a copy of the packet if there are group 1379 * members downstream on the interface. 1380 */ 1381 for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) 1382 if (IF_ISSET(mifi, &rt->mf6c_ifset)) { 1383 /* 1384 * check if the outgoing packet is going to break 1385 * a scope boundary. 1386 * XXX For packets through PIM register tunnel 1387 * interface, we believe a routing daemon. 1388 */ 1389 if ((mif6table[rt->mf6c_parent].m6_flags & 1390 MIFF_REGISTER) == 0 && 1391 (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 && 1392 (in6_addr2scopeid(ifp, &ip6->ip6_dst) != 1393 in6_addr2scopeid(mif6table[mifi].m6_ifp, 1394 &ip6->ip6_dst) || 1395 in6_addr2scopeid(ifp, &ip6->ip6_src) != 1396 in6_addr2scopeid(mif6table[mifi].m6_ifp, 1397 &ip6->ip6_src))) { 1398 ip6stat.ip6s_badscope++; 1399 continue; 1400 } 1401 1402 mifp->m6_pkt_out++; 1403 mifp->m6_bytes_out += plen; 1404 MC6_SEND(ip6, mifp, m); 1405 } 1406 return 0; 1407} 1408 1409static void 1410phyint_send(ip6, mifp, m) 1411 struct ip6_hdr *ip6; 1412 struct mif6 *mifp; 1413 struct mbuf *m; 1414{ 1415 struct mbuf *mb_copy; 1416 struct ifnet *ifp = mifp->m6_ifp; 1417 int error = 0; 1418 int s = splnet(); /* needs to protect static "ro" below. */ 1419 static struct route_in6 ro; 1420 struct in6_multi *in6m; 1421 struct sockaddr_in6 *dst6; 1422 1423 /* 1424 * Make a new reference to the packet; make sure that 1425 * the IPv6 header is actually copied, not just referenced, 1426 * so that ip6_output() only scribbles on the copy. 1427 */ 1428 mb_copy = m_copy(m, 0, M_COPYALL); 1429 if (mb_copy && 1430 (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr))) 1431 mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr)); 1432 if (mb_copy == NULL) { 1433 splx(s); 1434 return; 1435 } 1436 /* set MCAST flag to the outgoing packet */ 1437 mb_copy->m_flags |= M_MCAST; 1438 1439 /* 1440 * If we sourced the packet, call ip6_output since we may devide 1441 * the packet into fragments when the packet is too big for the 1442 * outgoing interface. 1443 * Otherwise, we can simply send the packet to the interface 1444 * sending queue. 1445 */ 1446 if (m->m_pkthdr.rcvif == NULL) { 1447 struct ip6_moptions im6o; 1448 1449 im6o.im6o_multicast_ifp = ifp; 1450 /* XXX: ip6_output will override ip6->ip6_hlim */ 1451 im6o.im6o_multicast_hlim = ip6->ip6_hlim; 1452 im6o.im6o_multicast_loop = 1; 1453 error = ip6_output(mb_copy, NULL, &ro, 1454 IPV6_FORWARDING, &im6o, NULL); 1455 1456#ifdef MRT6DEBUG 1457 if (mrt6debug & DEBUG_XMIT) 1458 log(LOG_DEBUG, "phyint_send on mif %d err %d\n", 1459 mifp - mif6table, error); 1460#endif 1461 splx(s); 1462 return; 1463 } 1464 1465 /* 1466 * If we belong to the destination multicast group 1467 * on the outgoing interface, loop back a copy. 1468 */ 1469 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 1470 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 1471 if (in6m != NULL) { 1472 dst6->sin6_len = sizeof(struct sockaddr_in6); 1473 dst6->sin6_family = AF_INET6; 1474 dst6->sin6_addr = ip6->ip6_dst; 1475 ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst); 1476 } 1477 /* 1478 * Put the packet into the sending queue of the outgoing interface 1479 * if it would fit in the MTU of the interface. 1480 */ 1481 if (mb_copy->m_pkthdr.len < ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) { 1482 dst6->sin6_len = sizeof(struct sockaddr_in6); 1483 dst6->sin6_family = AF_INET6; 1484 dst6->sin6_addr = ip6->ip6_dst; 1485 /* 1486 * We just call if_output instead of nd6_output here, since 1487 * we need no ND for a multicast forwarded packet...right? 1488 */ 1489 error = (*ifp->if_output)(ifp, mb_copy, 1490 (struct sockaddr *)&ro.ro_dst, NULL); 1491#ifdef MRT6DEBUG 1492 if (mrt6debug & DEBUG_XMIT) 1493 log(LOG_DEBUG, "phyint_send on mif %d err %d\n", 1494 mifp - mif6table, error); 1495#endif 1496 } else { 1497#ifdef MULTICAST_PMTUD 1498 icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 1499#else 1500#ifdef MRT6DEBUG 1501 if (mrt6debug & DEBUG_XMIT) 1502 log(LOG_DEBUG, 1503 "phyint_send: packet too big on %s o %s g %s" 1504 " size %d(discarded)\n", 1505 if_name(ifp), 1506 ip6_sprintf(&ip6->ip6_src), 1507 ip6_sprintf(&ip6->ip6_dst), 1508 mb_copy->m_pkthdr.len); 1509#endif /* MRT6DEBUG */ 1510 m_freem(mb_copy); /* simply discard the packet */ 1511#endif 1512 } 1513 1514 splx(s); 1515} 1516 1517static int 1518register_send(ip6, mif, m) 1519 struct ip6_hdr *ip6; 1520 struct mif6 *mif; 1521 struct mbuf *m; 1522{ 1523 struct mbuf *mm; 1524 int i, len = m->m_pkthdr.len; 1525 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; 1526 struct mrt6msg *im6; 1527 1528#ifdef MRT6DEBUG 1529 if (mrt6debug) 1530 log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n", 1531 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst)); 1532#endif 1533 ++pim6stat.pim6s_snd_registers; 1534 1535 /* Make a copy of the packet to send to the user level process */ 1536 MGETHDR(mm, M_DONTWAIT, MT_HEADER); 1537 if (mm == NULL) 1538 return ENOBUFS; 1539 mm->m_pkthdr.rcvif = NULL; 1540 mm->m_data += max_linkhdr; 1541 mm->m_len = sizeof(struct ip6_hdr); 1542 1543 if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1544 m_freem(mm); 1545 return ENOBUFS; 1546 } 1547 i = MHLEN - M_LEADINGSPACE(mm); 1548 if (i > len) 1549 i = len; 1550 mm = m_pullup(mm, i); 1551 if (mm == NULL) 1552 return ENOBUFS; 1553/* TODO: check it! */ 1554 mm->m_pkthdr.len = len + sizeof(struct ip6_hdr); 1555 1556 /* 1557 * Send message to routing daemon 1558 */ 1559 sin6.sin6_addr = ip6->ip6_src; 1560 1561 im6 = mtod(mm, struct mrt6msg *); 1562 im6->im6_msgtype = MRT6MSG_WHOLEPKT; 1563 im6->im6_mbz = 0; 1564 1565 im6->im6_mif = mif - mif6table; 1566 1567 /* iif info is not given for reg. encap.n */ 1568 mrt6stat.mrt6s_upcalls++; 1569 1570 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1571#ifdef MRT6DEBUG 1572 if (mrt6debug) 1573 log(LOG_WARNING, 1574 "register_send: ip6_mrouter socket queue full\n"); 1575#endif 1576 ++mrt6stat.mrt6s_upq_sockfull; 1577 return ENOBUFS; 1578 } 1579 return 0; 1580} 1581 1582/* 1583 * PIM sparse mode hook 1584 * Receives the pim control messages, and passes them up to the listening 1585 * socket, using rip6_input. 1586 * The only message processed is the REGISTER pim message; the pim header 1587 * is stripped off, and the inner packet is passed to register_mforward. 1588 */ 1589int 1590pim6_input(mp, offp, proto) 1591 struct mbuf **mp; 1592 int *offp, proto; 1593{ 1594 struct pim *pim; /* pointer to a pim struct */ 1595 struct ip6_hdr *ip6; 1596 int pimlen; 1597 struct mbuf *m = *mp; 1598 int minlen; 1599 int off = *offp; 1600 1601 ++pim6stat.pim6s_rcv_total; 1602 1603 ip6 = mtod(m, struct ip6_hdr *); 1604 pimlen = m->m_pkthdr.len - *offp; 1605 1606 /* 1607 * Validate lengths 1608 */ 1609 if (pimlen < PIM_MINLEN) { 1610 ++pim6stat.pim6s_rcv_tooshort; 1611#ifdef MRT6DEBUG 1612 if (mrt6debug & DEBUG_PIM) 1613 log(LOG_DEBUG,"pim6_input: PIM packet too short\n"); 1614#endif 1615 m_freem(m); 1616 return(IPPROTO_DONE); 1617 } 1618 1619 /* 1620 * if the packet is at least as big as a REGISTER, go ahead 1621 * and grab the PIM REGISTER header size, to avoid another 1622 * possible m_pullup() later. 1623 * 1624 * PIM_MINLEN == pimhdr + u_int32 == 8 1625 * PIM6_REG_MINLEN == pimhdr + reghdr + eip6hdr == 4 + 4 + 40 1626 */ 1627 minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN; 1628 1629 /* 1630 * Make sure that the IP6 and PIM headers in contiguous memory, and 1631 * possibly the PIM REGISTER header 1632 */ 1633#ifndef PULLDOWN_TEST 1634 IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE); 1635 /* adjust pointer */ 1636 ip6 = mtod(m, struct ip6_hdr *); 1637 1638 /* adjust mbuf to point to the PIM header */ 1639 pim = (struct pim *)((caddr_t)ip6 + off); 1640#else 1641 IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen); 1642 if (pim == NULL) { 1643 pim6stat.pim6s_rcv_tooshort++; 1644 return IPPROTO_DONE; 1645 } 1646#endif 1647 1648#define PIM6_CHECKSUM 1649#ifdef PIM6_CHECKSUM 1650 { 1651 int cksumlen; 1652 1653 /* 1654 * Validate checksum. 1655 * If PIM REGISTER, exclude the data packet 1656 */ 1657 if (pim->pim_type == PIM_REGISTER) 1658 cksumlen = PIM_MINLEN; 1659 else 1660 cksumlen = pimlen; 1661 1662 if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) { 1663 ++pim6stat.pim6s_rcv_badsum; 1664#ifdef MRT6DEBUG 1665 if (mrt6debug & DEBUG_PIM) 1666 log(LOG_DEBUG, 1667 "pim6_input: invalid checksum\n"); 1668#endif 1669 m_freem(m); 1670 return(IPPROTO_DONE); 1671 } 1672 } 1673#endif /* PIM_CHECKSUM */ 1674 1675 /* PIM version check */ 1676 if (pim->pim_ver != PIM_VERSION) { 1677 ++pim6stat.pim6s_rcv_badversion; 1678#ifdef MRT6DEBUG 1679 log(LOG_ERR, 1680 "pim6_input: incorrect version %d, expecting %d\n", 1681 pim->pim_ver, PIM_VERSION); 1682#endif 1683 m_freem(m); 1684 return(IPPROTO_DONE); 1685 } 1686 1687 if (pim->pim_type == PIM_REGISTER) { 1688 /* 1689 * since this is a REGISTER, we'll make a copy of the register 1690 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the 1691 * routing daemon. 1692 */ 1693 static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 }; 1694 1695 struct mbuf *mcp; 1696 struct ip6_hdr *eip6; 1697 u_int32_t *reghdr; 1698 int rc; 1699 1700 ++pim6stat.pim6s_rcv_registers; 1701 1702 if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) { 1703#ifdef MRT6DEBUG 1704 if (mrt6debug & DEBUG_PIM) 1705 log(LOG_DEBUG, 1706 "pim6_input: register mif not set: %d\n", 1707 reg_mif_num); 1708#endif 1709 m_freem(m); 1710 return(IPPROTO_DONE); 1711 } 1712 1713 reghdr = (u_int32_t *)(pim + 1); 1714 1715 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 1716 goto pim6_input_to_daemon; 1717 1718 /* 1719 * Validate length 1720 */ 1721 if (pimlen < PIM6_REG_MINLEN) { 1722 ++pim6stat.pim6s_rcv_tooshort; 1723 ++pim6stat.pim6s_rcv_badregisters; 1724#ifdef MRT6DEBUG 1725 log(LOG_ERR, 1726 "pim6_input: register packet size too " 1727 "small %d from %s\n", 1728 pimlen, ip6_sprintf(&ip6->ip6_src)); 1729#endif 1730 m_freem(m); 1731 return(IPPROTO_DONE); 1732 } 1733 1734 eip6 = (struct ip6_hdr *) (reghdr + 1); 1735#ifdef MRT6DEBUG 1736 if (mrt6debug & DEBUG_PIM) 1737 log(LOG_DEBUG, 1738 "pim6_input[register], eip6: %s -> %s, " 1739 "eip6 plen %d\n", 1740 ip6_sprintf(&eip6->ip6_src), 1741 ip6_sprintf(&eip6->ip6_dst), 1742 ntohs(eip6->ip6_plen)); 1743#endif 1744 1745 /* verify the version number of the inner packet */ 1746 if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1747 ++pim6stat.pim6s_rcv_badregisters; 1748#ifdef MRT6DEBUG 1749 log(LOG_DEBUG, "pim6_input: invalid IP version (%d) " 1750 "of the inner packet\n", 1751 (eip6->ip6_vfc & IPV6_VERSION)); 1752#endif 1753 m_freem(m); 1754 return(IPPROTO_NONE); 1755 } 1756 1757 /* verify the inner packet is destined to a mcast group */ 1758 if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) { 1759 ++pim6stat.pim6s_rcv_badregisters; 1760#ifdef MRT6DEBUG 1761 if (mrt6debug & DEBUG_PIM) 1762 log(LOG_DEBUG, 1763 "pim6_input: inner packet of register " 1764 "is not multicast %s\n", 1765 ip6_sprintf(&eip6->ip6_dst)); 1766#endif 1767 m_freem(m); 1768 return(IPPROTO_DONE); 1769 } 1770 1771 /* 1772 * make a copy of the whole header to pass to the daemon later. 1773 */ 1774 mcp = m_copy(m, 0, off + PIM6_REG_MINLEN); 1775 if (mcp == NULL) { 1776#ifdef MRT6DEBUG 1777 log(LOG_ERR, 1778 "pim6_input: pim register: " 1779 "could not copy register head\n"); 1780#endif 1781 m_freem(m); 1782 return(IPPROTO_DONE); 1783 } 1784 1785 /* 1786 * forward the inner ip6 packet; point m_data at the inner ip6. 1787 */ 1788 m_adj(m, off + PIM_MINLEN); 1789#ifdef MRT6DEBUG 1790 if (mrt6debug & DEBUG_PIM) { 1791 log(LOG_DEBUG, 1792 "pim6_input: forwarding decapsulated register: " 1793 "src %s, dst %s, mif %d\n", 1794 ip6_sprintf(&eip6->ip6_src), 1795 ip6_sprintf(&eip6->ip6_dst), 1796 reg_mif_num); 1797 } 1798#endif 1799 1800 rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m, 1801 dst.sin6_family, NULL); 1802 1803 /* prepare the register head to send to the mrouting daemon */ 1804 m = mcp; 1805 } 1806 1807 /* 1808 * Pass the PIM message up to the daemon; if it is a register message 1809 * pass the 'head' only up to the daemon. This includes the 1810 * encapsulator ip6 header, pim header, register header and the 1811 * encapsulated ip6 header. 1812 */ 1813 pim6_input_to_daemon: 1814 rip6_input(&m, offp, proto); 1815 return(IPPROTO_DONE); 1816} 1817