1 /* $FreeBSD: src/sys/netinet6/ip6_mroute.c,v 1.2.2.9 2003/01/23 21:06:47 sam Exp $ */ 2 /* $DragonFly: src/sys/netinet6/ip6_mroute.c,v 1.3 2003/08/23 11:02:45 rob Exp $ */ 3 /* $KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $ */ 4 5 /* 6 * Copyright (C) 1998 WIDE Project. 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. Neither the name of the project nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 */ 33 34 /* BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp */ 35 36 /* 37 * IP multicast forwarding procedures 38 * 39 * Written by David Waitzman, BBN Labs, August 1988. 40 * Modified by Steve Deering, Stanford, February 1989. 41 * Modified by Mark J. Steiglitz, Stanford, May, 1991 42 * Modified by Van Jacobson, LBL, January 1993 43 * Modified by Ajit Thyagarajan, PARC, August 1993 44 * Modified by Bill Fenenr, PARC, April 1994 45 * 46 * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support 47 */ 48 49 #include "opt_inet.h" 50 #include "opt_inet6.h" 51 52 #include <sys/param.h> 53 #include <sys/systm.h> 54 #include <sys/callout.h> 55 #include <sys/malloc.h> 56 #include <sys/mbuf.h> 57 #include <sys/socket.h> 58 #include <sys/socketvar.h> 59 #include <sys/sockio.h> 60 #include <sys/protosw.h> 61 #include <sys/errno.h> 62 #include <sys/time.h> 63 #include <sys/kernel.h> 64 #include <sys/syslog.h> 65 66 #include <net/if.h> 67 #include <net/route.h> 68 #include <net/raw_cb.h> 69 70 #include <netinet/in.h> 71 #include <netinet/in_var.h> 72 73 #include <netinet/ip6.h> 74 #include <netinet6/ip6_var.h> 75 #include <netinet6/ip6_mroute.h> 76 #include <netinet6/pim6.h> 77 #include <netinet6/pim6_var.h> 78 79 #include <net/net_osdep.h> 80 81 static MALLOC_DEFINE(M_MRTABLE, "mf6c", "multicast forwarding cache entry"); 82 83 #define M_HASCL(m) ((m)->m_flags & M_EXT) 84 85 static int ip6_mdq (struct mbuf *, struct ifnet *, struct mf6c *); 86 static void phyint_send (struct ip6_hdr *, struct mif6 *, struct mbuf *); 87 88 static int set_pim6 (int *); 89 static int socket_send (struct socket *, struct mbuf *, 90 struct sockaddr_in6 *); 91 static int register_send (struct ip6_hdr *, struct mif6 *, 92 struct mbuf *); 93 94 /* 95 * Globals. All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static, 96 * except for netstat or debugging purposes. 97 */ 98 struct socket *ip6_mrouter = NULL; 99 int ip6_mrouter_ver = 0; 100 int ip6_mrtproto = IPPROTO_PIM; /* for netstat only */ 101 struct mrt6stat mrt6stat; 102 103 #define NO_RTE_FOUND 0x1 104 #define RTE_FOUND 0x2 105 106 struct mf6c *mf6ctable[MF6CTBLSIZ]; 107 u_char n6expire[MF6CTBLSIZ]; 108 static struct mif6 mif6table[MAXMIFS]; 109 #ifdef MRT6DEBUG 110 u_int mrt6debug = 0; /* debug level */ 111 #define DEBUG_MFC 0x02 112 #define DEBUG_FORWARD 0x04 113 #define DEBUG_EXPIRE 0x08 114 #define DEBUG_XMIT 0x10 115 #define DEBUG_REG 0x20 116 #define DEBUG_PIM 0x40 117 #endif 118 119 static void expire_upcalls (void *); 120 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 121 #define UPCALL_EXPIRE 6 /* number of timeouts */ 122 123 #ifdef INET 124 #ifdef MROUTING 125 extern struct socket *ip_mrouter; 126 #endif 127 #endif 128 129 /* 130 * 'Interfaces' associated with decapsulator (so we can tell 131 * packets that went through it from ones that get reflected 132 * by a broken gateway). These interfaces are never linked into 133 * the system ifnet list & no routes point to them. I.e., packets 134 * can't be sent this way. They only exist as a placeholder for 135 * multicast source verification. 136 */ 137 struct ifnet multicast_register_if; 138 139 #define ENCAP_HOPS 64 140 141 /* 142 * Private variables. 143 */ 144 static mifi_t nummifs = 0; 145 static mifi_t reg_mif_num = (mifi_t)-1; 146 147 static struct pim6stat pim6stat; 148 static int pim6; 149 150 /* 151 * Hash function for a source, group entry 152 */ 153 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \ 154 (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \ 155 (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \ 156 (g).s6_addr32[2] ^ (g).s6_addr32[3]) 157 158 /* 159 * Find a route for a given origin IPv6 address and Multicast group address. 160 * Quality of service parameter to be added in the future!!! 161 */ 162 163 #define MF6CFIND(o, g, rt) do { \ 164 struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \ 165 rt = NULL; \ 166 mrt6stat.mrt6s_mfc_lookups++; \ 167 while (_rt) { \ 168 if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \ 169 IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \ 170 (_rt->mf6c_stall == NULL)) { \ 171 rt = _rt; \ 172 break; \ 173 } \ 174 _rt = _rt->mf6c_next; \ 175 } \ 176 if (rt == NULL) { \ 177 mrt6stat.mrt6s_mfc_misses++; \ 178 } \ 179 } while (0) 180 181 /* 182 * Macros to compute elapsed time efficiently 183 * Borrowed from Van Jacobson's scheduling code 184 */ 185 #define TV_DELTA(a, b, delta) do { \ 186 int xxs; \ 187 \ 188 delta = (a).tv_usec - (b).tv_usec; \ 189 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 190 switch (xxs) { \ 191 case 2: \ 192 delta += 1000000; \ 193 /* fall through */ \ 194 case 1: \ 195 delta += 1000000; \ 196 break; \ 197 default: \ 198 delta += (1000000 * xxs); \ 199 } \ 200 } \ 201 } while (0) 202 203 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 204 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 205 206 #ifdef UPCALL_TIMING 207 #define UPCALL_MAX 50 208 u_long upcall_data[UPCALL_MAX + 1]; 209 static void collate(); 210 #endif /* UPCALL_TIMING */ 211 212 static int get_sg_cnt (struct sioc_sg_req6 *); 213 static int get_mif6_cnt (struct sioc_mif_req6 *); 214 static int ip6_mrouter_init (struct socket *, struct mbuf *, int); 215 static int add_m6if (struct mif6ctl *); 216 static int del_m6if (mifi_t *); 217 static int add_m6fc (struct mf6cctl *); 218 static int del_m6fc (struct mf6cctl *); 219 220 static struct callout expire_upcalls_ch; 221 222 /* 223 * Handle MRT setsockopt commands to modify the multicast routing tables. 224 */ 225 int 226 ip6_mrouter_set(so, sopt) 227 struct socket *so; 228 struct sockopt *sopt; 229 { 230 int error = 0; 231 struct mbuf *m; 232 233 if (so != ip6_mrouter && sopt->sopt_name != MRT6_INIT) 234 return (EACCES); 235 236 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 237 return (error); 238 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 239 return (error); 240 241 switch (sopt->sopt_name) { 242 case MRT6_INIT: 243 #ifdef MRT6_OINIT 244 case MRT6_OINIT: 245 #endif 246 error = ip6_mrouter_init(so, m, sopt->sopt_name); 247 break; 248 case MRT6_DONE: 249 error = ip6_mrouter_done(); 250 break; 251 case MRT6_ADD_MIF: 252 error = add_m6if(mtod(m, struct mif6ctl *)); 253 break; 254 case MRT6_DEL_MIF: 255 error = del_m6if(mtod(m, mifi_t *)); 256 break; 257 case MRT6_ADD_MFC: 258 error = add_m6fc(mtod(m, struct mf6cctl *)); 259 break; 260 case MRT6_DEL_MFC: 261 error = del_m6fc(mtod(m, struct mf6cctl *)); 262 break; 263 case MRT6_PIM: 264 error = set_pim6(mtod(m, int *)); 265 break; 266 default: 267 error = EOPNOTSUPP; 268 break; 269 } 270 271 (void)m_freem(m); 272 return(error); 273 } 274 275 /* 276 * Handle MRT getsockopt commands 277 */ 278 int 279 ip6_mrouter_get(so, sopt) 280 struct socket *so; 281 struct sockopt *sopt; 282 { 283 int error = 0; 284 285 if (so != ip6_mrouter) return EACCES; 286 287 switch (sopt->sopt_name) { 288 case MRT6_PIM: 289 error = sooptcopyout(sopt, &pim6, sizeof(pim6)); 290 break; 291 } 292 return (error); 293 } 294 295 /* 296 * Handle ioctl commands to obtain information from the cache 297 */ 298 int 299 mrt6_ioctl(cmd, data) 300 int cmd; 301 caddr_t data; 302 { 303 int error = 0; 304 305 switch (cmd) { 306 case SIOCGETSGCNT_IN6: 307 return(get_sg_cnt((struct sioc_sg_req6 *)data)); 308 break; /* for safety */ 309 case SIOCGETMIFCNT_IN6: 310 return(get_mif6_cnt((struct sioc_mif_req6 *)data)); 311 break; /* for safety */ 312 default: 313 return (EINVAL); 314 break; 315 } 316 return error; 317 } 318 319 /* 320 * returns the packet, byte, rpf-failure count for the source group provided 321 */ 322 static int 323 get_sg_cnt(req) 324 struct sioc_sg_req6 *req; 325 { 326 struct mf6c *rt; 327 int s; 328 329 s = splnet(); 330 MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt); 331 splx(s); 332 if (rt != NULL) { 333 req->pktcnt = rt->mf6c_pkt_cnt; 334 req->bytecnt = rt->mf6c_byte_cnt; 335 req->wrong_if = rt->mf6c_wrong_if; 336 } else 337 return(ESRCH); 338 #if 0 339 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 340 #endif 341 342 return 0; 343 } 344 345 /* 346 * returns the input and output packet and byte counts on the mif provided 347 */ 348 static int 349 get_mif6_cnt(req) 350 struct sioc_mif_req6 *req; 351 { 352 mifi_t mifi = req->mifi; 353 354 if (mifi >= nummifs) 355 return EINVAL; 356 357 req->icount = mif6table[mifi].m6_pkt_in; 358 req->ocount = mif6table[mifi].m6_pkt_out; 359 req->ibytes = mif6table[mifi].m6_bytes_in; 360 req->obytes = mif6table[mifi].m6_bytes_out; 361 362 return 0; 363 } 364 365 static int 366 set_pim6(i) 367 int *i; 368 { 369 if ((*i != 1) && (*i != 0)) 370 return EINVAL; 371 372 pim6 = *i; 373 374 return 0; 375 } 376 377 /* 378 * Enable multicast routing 379 */ 380 static int 381 ip6_mrouter_init(so, m, cmd) 382 struct socket *so; 383 struct mbuf *m; 384 int cmd; 385 { 386 int *v; 387 388 #ifdef MRT6DEBUG 389 if (mrt6debug) 390 log(LOG_DEBUG, 391 "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n", 392 so->so_type, so->so_proto->pr_protocol); 393 #endif 394 395 if (so->so_type != SOCK_RAW || 396 so->so_proto->pr_protocol != IPPROTO_ICMPV6) 397 return EOPNOTSUPP; 398 399 if (!m || (m->m_len != sizeof(int *))) 400 return ENOPROTOOPT; 401 402 v = mtod(m, int *); 403 if (*v != 1) 404 return ENOPROTOOPT; 405 406 if (ip6_mrouter != NULL) return EADDRINUSE; 407 408 ip6_mrouter = so; 409 ip6_mrouter_ver = cmd; 410 411 bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); 412 bzero((caddr_t)n6expire, sizeof(n6expire)); 413 414 pim6 = 0;/* used for stubbing out/in pim stuff */ 415 416 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 417 expire_upcalls, NULL); 418 419 #ifdef MRT6DEBUG 420 if (mrt6debug) 421 log(LOG_DEBUG, "ip6_mrouter_init\n"); 422 #endif 423 424 return 0; 425 } 426 427 /* 428 * Disable multicast routing 429 */ 430 int 431 ip6_mrouter_done() 432 { 433 mifi_t mifi; 434 int i; 435 struct ifnet *ifp; 436 struct in6_ifreq ifr; 437 struct mf6c *rt; 438 struct rtdetq *rte; 439 int s; 440 441 s = splnet(); 442 443 /* 444 * For each phyint in use, disable promiscuous reception of all IPv6 445 * multicasts. 446 */ 447 #ifdef INET 448 #ifdef MROUTING 449 /* 450 * If there is still IPv4 multicast routing daemon, 451 * we remain interfaces to receive all muliticasted packets. 452 * XXX: there may be an interface in which the IPv4 multicast 453 * daemon is not interested... 454 */ 455 if (!ip_mrouter) 456 #endif 457 #endif 458 { 459 for (mifi = 0; mifi < nummifs; mifi++) { 460 if (mif6table[mifi].m6_ifp && 461 !(mif6table[mifi].m6_flags & MIFF_REGISTER)) { 462 ifr.ifr_addr.sin6_family = AF_INET6; 463 ifr.ifr_addr.sin6_addr= in6addr_any; 464 ifp = mif6table[mifi].m6_ifp; 465 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, 466 (caddr_t)&ifr); 467 } 468 } 469 } 470 #ifdef notyet 471 bzero((caddr_t)qtable, sizeof(qtable)); 472 bzero((caddr_t)tbftable, sizeof(tbftable)); 473 #endif 474 bzero((caddr_t)mif6table, sizeof(mif6table)); 475 nummifs = 0; 476 477 pim6 = 0; /* used to stub out/in pim specific code */ 478 479 callout_stop(&expire_upcalls_ch); 480 481 /* 482 * Free all multicast forwarding cache entries. 483 */ 484 for (i = 0; i < MF6CTBLSIZ; i++) { 485 rt = mf6ctable[i]; 486 while (rt) { 487 struct mf6c *frt; 488 489 for (rte = rt->mf6c_stall; rte != NULL; ) { 490 struct rtdetq *n = rte->next; 491 492 m_free(rte->m); 493 free(rte, M_MRTABLE); 494 rte = n; 495 } 496 frt = rt; 497 rt = rt->mf6c_next; 498 free(frt, M_MRTABLE); 499 } 500 } 501 502 bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); 503 504 /* 505 * Reset de-encapsulation cache 506 */ 507 reg_mif_num = -1; 508 509 ip6_mrouter = NULL; 510 ip6_mrouter_ver = 0; 511 512 splx(s); 513 514 #ifdef MRT6DEBUG 515 if (mrt6debug) 516 log(LOG_DEBUG, "ip6_mrouter_done\n"); 517 #endif 518 519 return 0; 520 } 521 522 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; 523 524 /* 525 * Add a mif to the mif table 526 */ 527 static int 528 add_m6if(mifcp) 529 struct mif6ctl *mifcp; 530 { 531 struct mif6 *mifp; 532 struct ifnet *ifp; 533 int error, s; 534 #ifdef notyet 535 struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi; 536 #endif 537 538 if (mifcp->mif6c_mifi >= MAXMIFS) 539 return EINVAL; 540 mifp = mif6table + mifcp->mif6c_mifi; 541 if (mifp->m6_ifp) 542 return EADDRINUSE; /* XXX: is it appropriate? */ 543 if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index) 544 return ENXIO; 545 ifp = ifindex2ifnet[mifcp->mif6c_pifi]; 546 547 if (mifcp->mif6c_flags & MIFF_REGISTER) { 548 if (reg_mif_num == (mifi_t)-1) { 549 multicast_register_if.if_name = "register_mif"; 550 multicast_register_if.if_flags |= IFF_LOOPBACK; 551 multicast_register_if.if_index = mifcp->mif6c_mifi; 552 reg_mif_num = mifcp->mif6c_mifi; 553 } 554 555 ifp = &multicast_register_if; 556 557 } /* if REGISTER */ 558 else { 559 /* Make sure the interface supports multicast */ 560 if ((ifp->if_flags & IFF_MULTICAST) == 0) 561 return EOPNOTSUPP; 562 563 s = splnet(); 564 error = if_allmulti(ifp, 1); 565 splx(s); 566 if (error) 567 return error; 568 } 569 570 s = splnet(); 571 mifp->m6_flags = mifcp->mif6c_flags; 572 mifp->m6_ifp = ifp; 573 #ifdef notyet 574 /* scaling up here allows division by 1024 in critical code */ 575 mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000; 576 #endif 577 /* initialize per mif pkt counters */ 578 mifp->m6_pkt_in = 0; 579 mifp->m6_pkt_out = 0; 580 mifp->m6_bytes_in = 0; 581 mifp->m6_bytes_out = 0; 582 splx(s); 583 584 /* Adjust nummifs up if the mifi is higher than nummifs */ 585 if (nummifs <= mifcp->mif6c_mifi) 586 nummifs = mifcp->mif6c_mifi + 1; 587 588 #ifdef MRT6DEBUG 589 if (mrt6debug) 590 log(LOG_DEBUG, 591 "add_mif #%d, phyint %s%d\n", 592 mifcp->mif6c_mifi, 593 ifp->if_name, ifp->if_unit); 594 #endif 595 596 return 0; 597 } 598 599 /* 600 * Delete a mif from the mif table 601 */ 602 static int 603 del_m6if(mifip) 604 mifi_t *mifip; 605 { 606 struct mif6 *mifp = mif6table + *mifip; 607 mifi_t mifi; 608 struct ifnet *ifp; 609 int s; 610 611 if (*mifip >= nummifs) 612 return EINVAL; 613 if (mifp->m6_ifp == NULL) 614 return EINVAL; 615 616 s = splnet(); 617 618 if (!(mifp->m6_flags & MIFF_REGISTER)) { 619 /* 620 * XXX: what if there is yet IPv4 multicast daemon 621 * using the interface? 622 */ 623 ifp = mifp->m6_ifp; 624 625 if_allmulti(ifp, 0); 626 } 627 628 #ifdef notyet 629 bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip])); 630 bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf))); 631 #endif 632 bzero((caddr_t)mifp, sizeof (*mifp)); 633 634 /* Adjust nummifs down */ 635 for (mifi = nummifs; mifi > 0; mifi--) 636 if (mif6table[mifi - 1].m6_ifp) 637 break; 638 nummifs = mifi; 639 640 splx(s); 641 642 #ifdef MRT6DEBUG 643 if (mrt6debug) 644 log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs); 645 #endif 646 647 return 0; 648 } 649 650 /* 651 * Add an mfc entry 652 */ 653 static int 654 add_m6fc(mfccp) 655 struct mf6cctl *mfccp; 656 { 657 struct mf6c *rt; 658 u_long hash; 659 struct rtdetq *rte; 660 u_short nstl; 661 int s; 662 663 MF6CFIND(mfccp->mf6cc_origin.sin6_addr, 664 mfccp->mf6cc_mcastgrp.sin6_addr, rt); 665 666 /* If an entry already exists, just update the fields */ 667 if (rt) { 668 #ifdef MRT6DEBUG 669 if (mrt6debug & DEBUG_MFC) 670 log(LOG_DEBUG, 671 "add_m6fc no upcall h %d o %s g %s p %x\n", 672 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 673 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 674 mfccp->mf6cc_parent); 675 #endif 676 677 s = splnet(); 678 rt->mf6c_parent = mfccp->mf6cc_parent; 679 rt->mf6c_ifset = mfccp->mf6cc_ifset; 680 splx(s); 681 return 0; 682 } 683 684 /* 685 * Find the entry for which the upcall was made and update 686 */ 687 s = splnet(); 688 hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr, 689 mfccp->mf6cc_mcastgrp.sin6_addr); 690 for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) { 691 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, 692 &mfccp->mf6cc_origin.sin6_addr) && 693 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, 694 &mfccp->mf6cc_mcastgrp.sin6_addr) && 695 (rt->mf6c_stall != NULL)) { 696 697 if (nstl++) 698 log(LOG_ERR, 699 "add_m6fc: %s o %s g %s p %x dbx %p\n", 700 "multiple kernel entries", 701 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 702 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 703 mfccp->mf6cc_parent, rt->mf6c_stall); 704 705 #ifdef MRT6DEBUG 706 if (mrt6debug & DEBUG_MFC) 707 log(LOG_DEBUG, 708 "add_m6fc o %s g %s p %x dbg %x\n", 709 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 710 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 711 mfccp->mf6cc_parent, rt->mf6c_stall); 712 #endif 713 714 rt->mf6c_origin = mfccp->mf6cc_origin; 715 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 716 rt->mf6c_parent = mfccp->mf6cc_parent; 717 rt->mf6c_ifset = mfccp->mf6cc_ifset; 718 /* initialize pkt counters per src-grp */ 719 rt->mf6c_pkt_cnt = 0; 720 rt->mf6c_byte_cnt = 0; 721 rt->mf6c_wrong_if = 0; 722 723 rt->mf6c_expire = 0; /* Don't clean this guy up */ 724 n6expire[hash]--; 725 726 /* free packets Qed at the end of this entry */ 727 for (rte = rt->mf6c_stall; rte != NULL; ) { 728 struct rtdetq *n = rte->next; 729 ip6_mdq(rte->m, rte->ifp, rt); 730 m_freem(rte->m); 731 #ifdef UPCALL_TIMING 732 collate(&(rte->t)); 733 #endif /* UPCALL_TIMING */ 734 free(rte, M_MRTABLE); 735 rte = n; 736 } 737 rt->mf6c_stall = NULL; 738 } 739 } 740 741 /* 742 * It is possible that an entry is being inserted without an upcall 743 */ 744 if (nstl == 0) { 745 #ifdef MRT6DEBUG 746 if (mrt6debug & DEBUG_MFC) 747 log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n", 748 hash, 749 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 750 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 751 mfccp->mf6cc_parent); 752 #endif 753 754 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { 755 756 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, 757 &mfccp->mf6cc_origin.sin6_addr)&& 758 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, 759 &mfccp->mf6cc_mcastgrp.sin6_addr)) { 760 761 rt->mf6c_origin = mfccp->mf6cc_origin; 762 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 763 rt->mf6c_parent = mfccp->mf6cc_parent; 764 rt->mf6c_ifset = mfccp->mf6cc_ifset; 765 /* initialize pkt counters per src-grp */ 766 rt->mf6c_pkt_cnt = 0; 767 rt->mf6c_byte_cnt = 0; 768 rt->mf6c_wrong_if = 0; 769 770 if (rt->mf6c_expire) 771 n6expire[hash]--; 772 rt->mf6c_expire = 0; 773 } 774 } 775 if (rt == NULL) { 776 /* no upcall, so make a new entry */ 777 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE, 778 M_NOWAIT); 779 if (rt == NULL) { 780 splx(s); 781 return ENOBUFS; 782 } 783 784 /* insert new entry at head of hash chain */ 785 rt->mf6c_origin = mfccp->mf6cc_origin; 786 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 787 rt->mf6c_parent = mfccp->mf6cc_parent; 788 rt->mf6c_ifset = mfccp->mf6cc_ifset; 789 /* initialize pkt counters per src-grp */ 790 rt->mf6c_pkt_cnt = 0; 791 rt->mf6c_byte_cnt = 0; 792 rt->mf6c_wrong_if = 0; 793 rt->mf6c_expire = 0; 794 rt->mf6c_stall = NULL; 795 796 /* link into table */ 797 rt->mf6c_next = mf6ctable[hash]; 798 mf6ctable[hash] = rt; 799 } 800 } 801 splx(s); 802 return 0; 803 } 804 805 #ifdef UPCALL_TIMING 806 /* 807 * collect delay statistics on the upcalls 808 */ 809 static void 810 collate(t) 811 struct timeval *t; 812 { 813 u_long d; 814 struct timeval tp; 815 u_long delta; 816 817 GET_TIME(tp); 818 819 if (TV_LT(*t, tp)) 820 { 821 TV_DELTA(tp, *t, delta); 822 823 d = delta >> 10; 824 if (d > UPCALL_MAX) 825 d = UPCALL_MAX; 826 827 ++upcall_data[d]; 828 } 829 } 830 #endif /* UPCALL_TIMING */ 831 832 /* 833 * Delete an mfc entry 834 */ 835 static int 836 del_m6fc(mfccp) 837 struct mf6cctl *mfccp; 838 { 839 struct sockaddr_in6 origin; 840 struct sockaddr_in6 mcastgrp; 841 struct mf6c *rt; 842 struct mf6c **nptr; 843 u_long hash; 844 int s; 845 846 origin = mfccp->mf6cc_origin; 847 mcastgrp = mfccp->mf6cc_mcastgrp; 848 hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr); 849 850 #ifdef MRT6DEBUG 851 if (mrt6debug & DEBUG_MFC) 852 log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n", 853 ip6_sprintf(&origin.sin6_addr), 854 ip6_sprintf(&mcastgrp.sin6_addr)); 855 #endif 856 857 s = splnet(); 858 859 nptr = &mf6ctable[hash]; 860 while ((rt = *nptr) != NULL) { 861 if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr, 862 &rt->mf6c_origin.sin6_addr) && 863 IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr, 864 &rt->mf6c_mcastgrp.sin6_addr) && 865 rt->mf6c_stall == NULL) 866 break; 867 868 nptr = &rt->mf6c_next; 869 } 870 if (rt == NULL) { 871 splx(s); 872 return EADDRNOTAVAIL; 873 } 874 875 *nptr = rt->mf6c_next; 876 free(rt, M_MRTABLE); 877 878 splx(s); 879 880 return 0; 881 } 882 883 static int 884 socket_send(s, mm, src) 885 struct socket *s; 886 struct mbuf *mm; 887 struct sockaddr_in6 *src; 888 { 889 if (s) { 890 if (sbappendaddr(&s->so_rcv, 891 (struct sockaddr *)src, 892 mm, (struct mbuf *)0) != 0) { 893 sorwakeup(s); 894 return 0; 895 } 896 } 897 m_freem(mm); 898 return -1; 899 } 900 901 /* 902 * IPv6 multicast forwarding function. This function assumes that the packet 903 * pointed to by "ip6" has arrived on (or is about to be sent to) the interface 904 * pointed to by "ifp", and the packet is to be relayed to other networks 905 * that have members of the packet's destination IPv6 multicast group. 906 * 907 * The packet is returned unscathed to the caller, unless it is 908 * erroneous, in which case a non-zero return value tells the caller to 909 * discard it. 910 */ 911 912 int 913 ip6_mforward(ip6, ifp, m) 914 struct ip6_hdr *ip6; 915 struct ifnet *ifp; 916 struct mbuf *m; 917 { 918 struct mf6c *rt; 919 struct mif6 *mifp; 920 struct mbuf *mm; 921 int s; 922 mifi_t mifi; 923 924 #ifdef MRT6DEBUG 925 if (mrt6debug & DEBUG_FORWARD) 926 log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n", 927 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst), 928 ifp->if_index); 929 #endif 930 931 /* 932 * Don't forward a packet with Hop limit of zero or one, 933 * or a packet destined to a local-only group. 934 */ 935 if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) || 936 IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst)) 937 return 0; 938 ip6->ip6_hlim--; 939 940 /* 941 * Source address check: do not forward packets with unspecified 942 * source. It was discussed in July 2000, on ipngwg mailing list. 943 * This is rather more serious than unicast cases, because some 944 * MLD packets can be sent with the unspecified source address 945 * (although such packets must normally set 1 to the hop limit field). 946 */ 947 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 948 ip6stat.ip6s_cantforward++; 949 if (ip6_log_time + ip6_log_interval < time_second) { 950 ip6_log_time = time_second; 951 log(LOG_DEBUG, 952 "cannot forward " 953 "from %s to %s nxt %d received on %s\n", 954 ip6_sprintf(&ip6->ip6_src), 955 ip6_sprintf(&ip6->ip6_dst), 956 ip6->ip6_nxt, 957 if_name(m->m_pkthdr.rcvif)); 958 } 959 return 0; 960 } 961 962 /* 963 * Determine forwarding mifs from the forwarding cache table 964 */ 965 s = splnet(); 966 MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt); 967 968 /* Entry exists, so forward if necessary */ 969 if (rt) { 970 splx(s); 971 return (ip6_mdq(m, ifp, rt)); 972 } else { 973 /* 974 * If we don't have a route for packet's origin, 975 * Make a copy of the packet & 976 * send message to routing daemon 977 */ 978 979 struct mbuf *mb0; 980 struct rtdetq *rte; 981 u_long hash; 982 /* int i, npkts;*/ 983 #ifdef UPCALL_TIMING 984 struct timeval tp; 985 986 GET_TIME(tp); 987 #endif /* UPCALL_TIMING */ 988 989 mrt6stat.mrt6s_no_route++; 990 #ifdef MRT6DEBUG 991 if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC)) 992 log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n", 993 ip6_sprintf(&ip6->ip6_src), 994 ip6_sprintf(&ip6->ip6_dst)); 995 #endif 996 997 /* 998 * Allocate mbufs early so that we don't do extra work if we 999 * are just going to fail anyway. 1000 */ 1001 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, 1002 M_NOWAIT); 1003 if (rte == NULL) { 1004 splx(s); 1005 return ENOBUFS; 1006 } 1007 mb0 = m_copy(m, 0, M_COPYALL); 1008 /* 1009 * Pullup packet header if needed before storing it, 1010 * as other references may modify it in the meantime. 1011 */ 1012 if (mb0 && 1013 (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr))) 1014 mb0 = m_pullup(mb0, sizeof(struct ip6_hdr)); 1015 if (mb0 == NULL) { 1016 free(rte, M_MRTABLE); 1017 splx(s); 1018 return ENOBUFS; 1019 } 1020 1021 /* is there an upcall waiting for this packet? */ 1022 hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst); 1023 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { 1024 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, 1025 &rt->mf6c_origin.sin6_addr) && 1026 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 1027 &rt->mf6c_mcastgrp.sin6_addr) && 1028 (rt->mf6c_stall != NULL)) 1029 break; 1030 } 1031 1032 if (rt == NULL) { 1033 struct mrt6msg *im; 1034 #ifdef MRT6_OINIT 1035 struct omrt6msg *oim; 1036 #endif 1037 1038 /* no upcall, so make a new entry */ 1039 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE, 1040 M_NOWAIT); 1041 if (rt == NULL) { 1042 free(rte, M_MRTABLE); 1043 m_freem(mb0); 1044 splx(s); 1045 return ENOBUFS; 1046 } 1047 /* 1048 * Make a copy of the header to send to the user 1049 * level process 1050 */ 1051 mm = m_copy(mb0, 0, sizeof(struct ip6_hdr)); 1052 1053 if (mm == NULL) { 1054 free(rte, M_MRTABLE); 1055 m_freem(mb0); 1056 free(rt, M_MRTABLE); 1057 splx(s); 1058 return ENOBUFS; 1059 } 1060 1061 /* 1062 * Send message to routing daemon 1063 */ 1064 sin6.sin6_addr = ip6->ip6_src; 1065 1066 im = NULL; 1067 #ifdef MRT6_OINIT 1068 oim = NULL; 1069 #endif 1070 switch (ip6_mrouter_ver) { 1071 #ifdef MRT6_OINIT 1072 case MRT6_OINIT: 1073 oim = mtod(mm, struct omrt6msg *); 1074 oim->im6_msgtype = MRT6MSG_NOCACHE; 1075 oim->im6_mbz = 0; 1076 break; 1077 #endif 1078 case MRT6_INIT: 1079 im = mtod(mm, struct mrt6msg *); 1080 im->im6_msgtype = MRT6MSG_NOCACHE; 1081 im->im6_mbz = 0; 1082 break; 1083 default: 1084 free(rte, M_MRTABLE); 1085 m_freem(mb0); 1086 free(rt, M_MRTABLE); 1087 splx(s); 1088 return EINVAL; 1089 } 1090 1091 #ifdef MRT6DEBUG 1092 if (mrt6debug & DEBUG_FORWARD) 1093 log(LOG_DEBUG, 1094 "getting the iif info in the kernel\n"); 1095 #endif 1096 1097 for (mifp = mif6table, mifi = 0; 1098 mifi < nummifs && mifp->m6_ifp != ifp; 1099 mifp++, mifi++) 1100 ; 1101 1102 switch (ip6_mrouter_ver) { 1103 #ifdef MRT6_OINIT 1104 case MRT6_OINIT: 1105 oim->im6_mif = mifi; 1106 break; 1107 #endif 1108 case MRT6_INIT: 1109 im->im6_mif = mifi; 1110 break; 1111 } 1112 1113 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1114 log(LOG_WARNING, "ip6_mforward: ip6_mrouter " 1115 "socket queue full\n"); 1116 mrt6stat.mrt6s_upq_sockfull++; 1117 free(rte, M_MRTABLE); 1118 m_freem(mb0); 1119 free(rt, M_MRTABLE); 1120 splx(s); 1121 return ENOBUFS; 1122 } 1123 1124 mrt6stat.mrt6s_upcalls++; 1125 1126 /* insert new entry at head of hash chain */ 1127 bzero(rt, sizeof(*rt)); 1128 rt->mf6c_origin.sin6_family = AF_INET6; 1129 rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6); 1130 rt->mf6c_origin.sin6_addr = ip6->ip6_src; 1131 rt->mf6c_mcastgrp.sin6_family = AF_INET6; 1132 rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6); 1133 rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst; 1134 rt->mf6c_expire = UPCALL_EXPIRE; 1135 n6expire[hash]++; 1136 rt->mf6c_parent = MF6C_INCOMPLETE_PARENT; 1137 1138 /* link into table */ 1139 rt->mf6c_next = mf6ctable[hash]; 1140 mf6ctable[hash] = rt; 1141 /* Add this entry to the end of the queue */ 1142 rt->mf6c_stall = rte; 1143 } else { 1144 /* determine if q has overflowed */ 1145 struct rtdetq **p; 1146 int npkts = 0; 1147 1148 for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next) 1149 if (++npkts > MAX_UPQ6) { 1150 mrt6stat.mrt6s_upq_ovflw++; 1151 free(rte, M_MRTABLE); 1152 m_freem(mb0); 1153 splx(s); 1154 return 0; 1155 } 1156 1157 /* Add this entry to the end of the queue */ 1158 *p = rte; 1159 } 1160 1161 rte->next = NULL; 1162 rte->m = mb0; 1163 rte->ifp = ifp; 1164 #ifdef UPCALL_TIMING 1165 rte->t = tp; 1166 #endif /* UPCALL_TIMING */ 1167 1168 splx(s); 1169 1170 return 0; 1171 } 1172 } 1173 1174 /* 1175 * Clean up cache entries if upcalls are not serviced 1176 * Call from the Slow Timeout mechanism, every half second. 1177 */ 1178 static void 1179 expire_upcalls(unused) 1180 void *unused; 1181 { 1182 struct rtdetq *rte; 1183 struct mf6c *mfc, **nptr; 1184 int i; 1185 int s; 1186 1187 s = splnet(); 1188 for (i = 0; i < MF6CTBLSIZ; i++) { 1189 if (n6expire[i] == 0) 1190 continue; 1191 nptr = &mf6ctable[i]; 1192 while ((mfc = *nptr) != NULL) { 1193 rte = mfc->mf6c_stall; 1194 /* 1195 * Skip real cache entries 1196 * Make sure it wasn't marked to not expire (shouldn't happen) 1197 * If it expires now 1198 */ 1199 if (rte != NULL && 1200 mfc->mf6c_expire != 0 && 1201 --mfc->mf6c_expire == 0) { 1202 #ifdef MRT6DEBUG 1203 if (mrt6debug & DEBUG_EXPIRE) 1204 log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n", 1205 ip6_sprintf(&mfc->mf6c_origin.sin6_addr), 1206 ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr)); 1207 #endif 1208 /* 1209 * drop all the packets 1210 * free the mbuf with the pkt, if, timing info 1211 */ 1212 do { 1213 struct rtdetq *n = rte->next; 1214 m_freem(rte->m); 1215 free(rte, M_MRTABLE); 1216 rte = n; 1217 } while (rte != NULL); 1218 mrt6stat.mrt6s_cache_cleanups++; 1219 n6expire[i]--; 1220 1221 *nptr = mfc->mf6c_next; 1222 free(mfc, M_MRTABLE); 1223 } else { 1224 nptr = &mfc->mf6c_next; 1225 } 1226 } 1227 } 1228 splx(s); 1229 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1230 expire_upcalls, NULL); 1231 } 1232 1233 /* 1234 * Packet forwarding routine once entry in the cache is made 1235 */ 1236 static int 1237 ip6_mdq(m, ifp, rt) 1238 struct mbuf *m; 1239 struct ifnet *ifp; 1240 struct mf6c *rt; 1241 { 1242 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1243 mifi_t mifi, iif; 1244 struct mif6 *mifp; 1245 int plen = m->m_pkthdr.len; 1246 1247 /* 1248 * Macro to send packet on mif. Since RSVP packets don't get counted on 1249 * input, they shouldn't get counted on output, so statistics keeping is 1250 * seperate. 1251 */ 1252 1253 #define MC6_SEND(ip6, mifp, m) do { \ 1254 if ((mifp)->m6_flags & MIFF_REGISTER) \ 1255 register_send((ip6), (mifp), (m)); \ 1256 else \ 1257 phyint_send((ip6), (mifp), (m)); \ 1258 } while (0) 1259 1260 /* 1261 * Don't forward if it didn't arrive from the parent mif 1262 * for its origin. 1263 */ 1264 mifi = rt->mf6c_parent; 1265 if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) { 1266 /* came in the wrong interface */ 1267 #ifdef MRT6DEBUG 1268 if (mrt6debug & DEBUG_FORWARD) 1269 log(LOG_DEBUG, 1270 "wrong if: ifid %d mifi %d mififid %x\n", 1271 ifp->if_index, mifi, 1272 mif6table[mifi].m6_ifp->if_index); 1273 #endif 1274 mrt6stat.mrt6s_wrong_if++; 1275 rt->mf6c_wrong_if++; 1276 /* 1277 * If we are doing PIM processing, and we are forwarding 1278 * packets on this interface, send a message to the 1279 * routing daemon. 1280 */ 1281 /* have to make sure this is a valid mif */ 1282 if (mifi < nummifs && mif6table[mifi].m6_ifp) 1283 if (pim6 && (m->m_flags & M_LOOP) == 0) { 1284 /* 1285 * Check the M_LOOP flag to avoid an 1286 * unnecessary PIM assert. 1287 * XXX: M_LOOP is an ad-hoc hack... 1288 */ 1289 static struct sockaddr_in6 sin6 = 1290 { sizeof(sin6), AF_INET6 }; 1291 1292 struct mbuf *mm; 1293 struct mrt6msg *im; 1294 #ifdef MRT6_OINIT 1295 struct omrt6msg *oim; 1296 #endif 1297 1298 mm = m_copy(m, 0, sizeof(struct ip6_hdr)); 1299 if (mm && 1300 (M_HASCL(mm) || 1301 mm->m_len < sizeof(struct ip6_hdr))) 1302 mm = m_pullup(mm, sizeof(struct ip6_hdr)); 1303 if (mm == NULL) 1304 return ENOBUFS; 1305 1306 #ifdef MRT6_OINIT 1307 oim = NULL; 1308 #endif 1309 im = NULL; 1310 switch (ip6_mrouter_ver) { 1311 #ifdef MRT6_OINIT 1312 case MRT6_OINIT: 1313 oim = mtod(mm, struct omrt6msg *); 1314 oim->im6_msgtype = MRT6MSG_WRONGMIF; 1315 oim->im6_mbz = 0; 1316 break; 1317 #endif 1318 case MRT6_INIT: 1319 im = mtod(mm, struct mrt6msg *); 1320 im->im6_msgtype = MRT6MSG_WRONGMIF; 1321 im->im6_mbz = 0; 1322 break; 1323 default: 1324 m_freem(mm); 1325 return EINVAL; 1326 } 1327 1328 for (mifp = mif6table, iif = 0; 1329 iif < nummifs && mifp && 1330 mifp->m6_ifp != ifp; 1331 mifp++, iif++) 1332 ; 1333 1334 switch (ip6_mrouter_ver) { 1335 #ifdef MRT6_OINIT 1336 case MRT6_OINIT: 1337 oim->im6_mif = iif; 1338 sin6.sin6_addr = oim->im6_src; 1339 break; 1340 #endif 1341 case MRT6_INIT: 1342 im->im6_mif = iif; 1343 sin6.sin6_addr = im->im6_src; 1344 break; 1345 } 1346 1347 mrt6stat.mrt6s_upcalls++; 1348 1349 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1350 #ifdef MRT6DEBUG 1351 if (mrt6debug) 1352 log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n"); 1353 #endif 1354 ++mrt6stat.mrt6s_upq_sockfull; 1355 return ENOBUFS; 1356 } /* if socket Q full */ 1357 } /* if PIM */ 1358 return 0; 1359 } /* if wrong iif */ 1360 1361 /* If I sourced this packet, it counts as output, else it was input. */ 1362 if (m->m_pkthdr.rcvif == NULL) { 1363 /* XXX: is rcvif really NULL when output?? */ 1364 mif6table[mifi].m6_pkt_out++; 1365 mif6table[mifi].m6_bytes_out += plen; 1366 } else { 1367 mif6table[mifi].m6_pkt_in++; 1368 mif6table[mifi].m6_bytes_in += plen; 1369 } 1370 rt->mf6c_pkt_cnt++; 1371 rt->mf6c_byte_cnt += plen; 1372 1373 /* 1374 * For each mif, forward a copy of the packet if there are group 1375 * members downstream on the interface. 1376 */ 1377 for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) 1378 if (IF_ISSET(mifi, &rt->mf6c_ifset)) { 1379 /* 1380 * check if the outgoing packet is going to break 1381 * a scope boundary. 1382 * XXX For packets through PIM register tunnel 1383 * interface, we believe a routing daemon. 1384 */ 1385 if ((mif6table[rt->mf6c_parent].m6_flags & 1386 MIFF_REGISTER) == 0 && 1387 (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 && 1388 (in6_addr2scopeid(ifp, &ip6->ip6_dst) != 1389 in6_addr2scopeid(mif6table[mifi].m6_ifp, 1390 &ip6->ip6_dst) || 1391 in6_addr2scopeid(ifp, &ip6->ip6_src) != 1392 in6_addr2scopeid(mif6table[mifi].m6_ifp, 1393 &ip6->ip6_src))) { 1394 ip6stat.ip6s_badscope++; 1395 continue; 1396 } 1397 1398 mifp->m6_pkt_out++; 1399 mifp->m6_bytes_out += plen; 1400 MC6_SEND(ip6, mifp, m); 1401 } 1402 return 0; 1403 } 1404 1405 static void 1406 phyint_send(ip6, mifp, m) 1407 struct ip6_hdr *ip6; 1408 struct mif6 *mifp; 1409 struct mbuf *m; 1410 { 1411 struct mbuf *mb_copy; 1412 struct ifnet *ifp = mifp->m6_ifp; 1413 int error = 0; 1414 int s = splnet(); /* needs to protect static "ro" below. */ 1415 static struct route_in6 ro; 1416 struct in6_multi *in6m; 1417 struct sockaddr_in6 *dst6; 1418 1419 /* 1420 * Make a new reference to the packet; make sure that 1421 * the IPv6 header is actually copied, not just referenced, 1422 * so that ip6_output() only scribbles on the copy. 1423 */ 1424 mb_copy = m_copy(m, 0, M_COPYALL); 1425 if (mb_copy && 1426 (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr))) 1427 mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr)); 1428 if (mb_copy == NULL) { 1429 splx(s); 1430 return; 1431 } 1432 /* set MCAST flag to the outgoing packet */ 1433 mb_copy->m_flags |= M_MCAST; 1434 1435 /* 1436 * If we sourced the packet, call ip6_output since we may devide 1437 * the packet into fragments when the packet is too big for the 1438 * outgoing interface. 1439 * Otherwise, we can simply send the packet to the interface 1440 * sending queue. 1441 */ 1442 if (m->m_pkthdr.rcvif == NULL) { 1443 struct ip6_moptions im6o; 1444 1445 im6o.im6o_multicast_ifp = ifp; 1446 /* XXX: ip6_output will override ip6->ip6_hlim */ 1447 im6o.im6o_multicast_hlim = ip6->ip6_hlim; 1448 im6o.im6o_multicast_loop = 1; 1449 error = ip6_output(mb_copy, NULL, &ro, 1450 IPV6_FORWARDING, &im6o, NULL, NULL); 1451 1452 #ifdef MRT6DEBUG 1453 if (mrt6debug & DEBUG_XMIT) 1454 log(LOG_DEBUG, "phyint_send on mif %d err %d\n", 1455 mifp - mif6table, error); 1456 #endif 1457 splx(s); 1458 return; 1459 } 1460 1461 /* 1462 * If we belong to the destination multicast group 1463 * on the outgoing interface, loop back a copy. 1464 */ 1465 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 1466 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 1467 if (in6m != NULL) { 1468 dst6->sin6_len = sizeof(struct sockaddr_in6); 1469 dst6->sin6_family = AF_INET6; 1470 dst6->sin6_addr = ip6->ip6_dst; 1471 ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst); 1472 } 1473 /* 1474 * Put the packet into the sending queue of the outgoing interface 1475 * if it would fit in the MTU of the interface. 1476 */ 1477 if (mb_copy->m_pkthdr.len <= ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) { 1478 dst6->sin6_len = sizeof(struct sockaddr_in6); 1479 dst6->sin6_family = AF_INET6; 1480 dst6->sin6_addr = ip6->ip6_dst; 1481 /* 1482 * We just call if_output instead of nd6_output here, since 1483 * we need no ND for a multicast forwarded packet...right? 1484 */ 1485 error = (*ifp->if_output)(ifp, mb_copy, 1486 (struct sockaddr *)&ro.ro_dst, NULL); 1487 #ifdef MRT6DEBUG 1488 if (mrt6debug & DEBUG_XMIT) 1489 log(LOG_DEBUG, "phyint_send on mif %d err %d\n", 1490 mifp - mif6table, error); 1491 #endif 1492 } else { 1493 #ifdef MULTICAST_PMTUD 1494 icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 1495 #else 1496 #ifdef MRT6DEBUG 1497 if (mrt6debug & DEBUG_XMIT) 1498 log(LOG_DEBUG, 1499 "phyint_send: packet too big on %s o %s g %s" 1500 " size %d(discarded)\n", 1501 if_name(ifp), 1502 ip6_sprintf(&ip6->ip6_src), 1503 ip6_sprintf(&ip6->ip6_dst), 1504 mb_copy->m_pkthdr.len); 1505 #endif /* MRT6DEBUG */ 1506 m_freem(mb_copy); /* simply discard the packet */ 1507 #endif 1508 } 1509 1510 splx(s); 1511 } 1512 1513 static int 1514 register_send(ip6, mif, m) 1515 struct ip6_hdr *ip6; 1516 struct mif6 *mif; 1517 struct mbuf *m; 1518 { 1519 struct mbuf *mm; 1520 int i, len = m->m_pkthdr.len; 1521 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; 1522 struct mrt6msg *im6; 1523 1524 #ifdef MRT6DEBUG 1525 if (mrt6debug) 1526 log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n", 1527 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst)); 1528 #endif 1529 ++pim6stat.pim6s_snd_registers; 1530 1531 /* Make a copy of the packet to send to the user level process */ 1532 MGETHDR(mm, M_DONTWAIT, MT_HEADER); 1533 if (mm == NULL) 1534 return ENOBUFS; 1535 mm->m_pkthdr.rcvif = NULL; 1536 mm->m_data += max_linkhdr; 1537 mm->m_len = sizeof(struct ip6_hdr); 1538 1539 if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1540 m_freem(mm); 1541 return ENOBUFS; 1542 } 1543 i = MHLEN - M_LEADINGSPACE(mm); 1544 if (i > len) 1545 i = len; 1546 mm = m_pullup(mm, i); 1547 if (mm == NULL) 1548 return ENOBUFS; 1549 /* TODO: check it! */ 1550 mm->m_pkthdr.len = len + sizeof(struct ip6_hdr); 1551 1552 /* 1553 * Send message to routing daemon 1554 */ 1555 sin6.sin6_addr = ip6->ip6_src; 1556 1557 im6 = mtod(mm, struct mrt6msg *); 1558 im6->im6_msgtype = MRT6MSG_WHOLEPKT; 1559 im6->im6_mbz = 0; 1560 1561 im6->im6_mif = mif - mif6table; 1562 1563 /* iif info is not given for reg. encap.n */ 1564 mrt6stat.mrt6s_upcalls++; 1565 1566 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1567 #ifdef MRT6DEBUG 1568 if (mrt6debug) 1569 log(LOG_WARNING, 1570 "register_send: ip6_mrouter socket queue full\n"); 1571 #endif 1572 ++mrt6stat.mrt6s_upq_sockfull; 1573 return ENOBUFS; 1574 } 1575 return 0; 1576 } 1577 1578 /* 1579 * PIM sparse mode hook 1580 * Receives the pim control messages, and passes them up to the listening 1581 * socket, using rip6_input. 1582 * The only message processed is the REGISTER pim message; the pim header 1583 * is stripped off, and the inner packet is passed to register_mforward. 1584 */ 1585 int 1586 pim6_input(mp, offp, proto) 1587 struct mbuf **mp; 1588 int *offp, proto; 1589 { 1590 struct pim *pim; /* pointer to a pim struct */ 1591 struct ip6_hdr *ip6; 1592 int pimlen; 1593 struct mbuf *m = *mp; 1594 int minlen; 1595 int off = *offp; 1596 1597 ++pim6stat.pim6s_rcv_total; 1598 1599 ip6 = mtod(m, struct ip6_hdr *); 1600 pimlen = m->m_pkthdr.len - *offp; 1601 1602 /* 1603 * Validate lengths 1604 */ 1605 if (pimlen < PIM_MINLEN) { 1606 ++pim6stat.pim6s_rcv_tooshort; 1607 #ifdef MRT6DEBUG 1608 if (mrt6debug & DEBUG_PIM) 1609 log(LOG_DEBUG,"pim6_input: PIM packet too short\n"); 1610 #endif 1611 m_freem(m); 1612 return(IPPROTO_DONE); 1613 } 1614 1615 /* 1616 * if the packet is at least as big as a REGISTER, go ahead 1617 * and grab the PIM REGISTER header size, to avoid another 1618 * possible m_pullup() later. 1619 * 1620 * PIM_MINLEN == pimhdr + u_int32 == 8 1621 * PIM6_REG_MINLEN == pimhdr + reghdr + eip6hdr == 4 + 4 + 40 1622 */ 1623 minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN; 1624 1625 /* 1626 * Make sure that the IP6 and PIM headers in contiguous memory, and 1627 * possibly the PIM REGISTER header 1628 */ 1629 #ifndef PULLDOWN_TEST 1630 IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE); 1631 /* adjust pointer */ 1632 ip6 = mtod(m, struct ip6_hdr *); 1633 1634 /* adjust mbuf to point to the PIM header */ 1635 pim = (struct pim *)((caddr_t)ip6 + off); 1636 #else 1637 IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen); 1638 if (pim == NULL) { 1639 pim6stat.pim6s_rcv_tooshort++; 1640 return IPPROTO_DONE; 1641 } 1642 #endif 1643 1644 #define PIM6_CHECKSUM 1645 #ifdef PIM6_CHECKSUM 1646 { 1647 int cksumlen; 1648 1649 /* 1650 * Validate checksum. 1651 * If PIM REGISTER, exclude the data packet 1652 */ 1653 if (pim->pim_type == PIM_REGISTER) 1654 cksumlen = PIM_MINLEN; 1655 else 1656 cksumlen = pimlen; 1657 1658 if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) { 1659 ++pim6stat.pim6s_rcv_badsum; 1660 #ifdef MRT6DEBUG 1661 if (mrt6debug & DEBUG_PIM) 1662 log(LOG_DEBUG, 1663 "pim6_input: invalid checksum\n"); 1664 #endif 1665 m_freem(m); 1666 return(IPPROTO_DONE); 1667 } 1668 } 1669 #endif /* PIM_CHECKSUM */ 1670 1671 /* PIM version check */ 1672 if (pim->pim_ver != PIM_VERSION) { 1673 ++pim6stat.pim6s_rcv_badversion; 1674 #ifdef MRT6DEBUG 1675 log(LOG_ERR, 1676 "pim6_input: incorrect version %d, expecting %d\n", 1677 pim->pim_ver, PIM_VERSION); 1678 #endif 1679 m_freem(m); 1680 return(IPPROTO_DONE); 1681 } 1682 1683 if (pim->pim_type == PIM_REGISTER) { 1684 /* 1685 * since this is a REGISTER, we'll make a copy of the register 1686 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the 1687 * routing daemon. 1688 */ 1689 static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 }; 1690 1691 struct mbuf *mcp; 1692 struct ip6_hdr *eip6; 1693 u_int32_t *reghdr; 1694 int rc; 1695 1696 ++pim6stat.pim6s_rcv_registers; 1697 1698 if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) { 1699 #ifdef MRT6DEBUG 1700 if (mrt6debug & DEBUG_PIM) 1701 log(LOG_DEBUG, 1702 "pim6_input: register mif not set: %d\n", 1703 reg_mif_num); 1704 #endif 1705 m_freem(m); 1706 return(IPPROTO_DONE); 1707 } 1708 1709 reghdr = (u_int32_t *)(pim + 1); 1710 1711 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 1712 goto pim6_input_to_daemon; 1713 1714 /* 1715 * Validate length 1716 */ 1717 if (pimlen < PIM6_REG_MINLEN) { 1718 ++pim6stat.pim6s_rcv_tooshort; 1719 ++pim6stat.pim6s_rcv_badregisters; 1720 #ifdef MRT6DEBUG 1721 log(LOG_ERR, 1722 "pim6_input: register packet size too " 1723 "small %d from %s\n", 1724 pimlen, ip6_sprintf(&ip6->ip6_src)); 1725 #endif 1726 m_freem(m); 1727 return(IPPROTO_DONE); 1728 } 1729 1730 eip6 = (struct ip6_hdr *) (reghdr + 1); 1731 #ifdef MRT6DEBUG 1732 if (mrt6debug & DEBUG_PIM) 1733 log(LOG_DEBUG, 1734 "pim6_input[register], eip6: %s -> %s, " 1735 "eip6 plen %d\n", 1736 ip6_sprintf(&eip6->ip6_src), 1737 ip6_sprintf(&eip6->ip6_dst), 1738 ntohs(eip6->ip6_plen)); 1739 #endif 1740 1741 /* verify the version number of the inner packet */ 1742 if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1743 ++pim6stat.pim6s_rcv_badregisters; 1744 #ifdef MRT6DEBUG 1745 log(LOG_DEBUG, "pim6_input: invalid IP version (%d) " 1746 "of the inner packet\n", 1747 (eip6->ip6_vfc & IPV6_VERSION)); 1748 #endif 1749 m_freem(m); 1750 return(IPPROTO_NONE); 1751 } 1752 1753 /* verify the inner packet is destined to a mcast group */ 1754 if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) { 1755 ++pim6stat.pim6s_rcv_badregisters; 1756 #ifdef MRT6DEBUG 1757 if (mrt6debug & DEBUG_PIM) 1758 log(LOG_DEBUG, 1759 "pim6_input: inner packet of register " 1760 "is not multicast %s\n", 1761 ip6_sprintf(&eip6->ip6_dst)); 1762 #endif 1763 m_freem(m); 1764 return(IPPROTO_DONE); 1765 } 1766 1767 /* 1768 * make a copy of the whole header to pass to the daemon later. 1769 */ 1770 mcp = m_copy(m, 0, off + PIM6_REG_MINLEN); 1771 if (mcp == NULL) { 1772 #ifdef MRT6DEBUG 1773 log(LOG_ERR, 1774 "pim6_input: pim register: " 1775 "could not copy register head\n"); 1776 #endif 1777 m_freem(m); 1778 return(IPPROTO_DONE); 1779 } 1780 1781 /* 1782 * forward the inner ip6 packet; point m_data at the inner ip6. 1783 */ 1784 m_adj(m, off + PIM_MINLEN); 1785 #ifdef MRT6DEBUG 1786 if (mrt6debug & DEBUG_PIM) { 1787 log(LOG_DEBUG, 1788 "pim6_input: forwarding decapsulated register: " 1789 "src %s, dst %s, mif %d\n", 1790 ip6_sprintf(&eip6->ip6_src), 1791 ip6_sprintf(&eip6->ip6_dst), 1792 reg_mif_num); 1793 } 1794 #endif 1795 1796 rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m, 1797 dst.sin6_family, NULL); 1798 1799 /* prepare the register head to send to the mrouting daemon */ 1800 m = mcp; 1801 } 1802 1803 /* 1804 * Pass the PIM message up to the daemon; if it is a register message 1805 * pass the 'head' only up to the daemon. This includes the 1806 * encapsulator ip6 header, pim header, register header and the 1807 * encapsulated ip6 header. 1808 */ 1809 pim6_input_to_daemon: 1810 rip6_input(&m, offp, proto); 1811 return(IPPROTO_DONE); 1812 } 1813