1 /* $NetBSD: ip6_mroute.c,v 1.30 2002/05/30 04:39:15 itojun Exp $ */ 2 /* $KAME: ip6_mroute.c,v 1.49 2001/07/25 09:21:18 jinmei 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 <sys/cdefs.h> 49 __KERNEL_RCSID(0, "$NetBSD: ip6_mroute.c,v 1.30 2002/05/30 04:39:15 itojun Exp $"); 50 51 #include "opt_inet.h" 52 53 #ifndef _KERNEL 54 # ifdef KERNEL 55 # define _KERNEL 56 # endif 57 #endif 58 59 #include <sys/param.h> 60 #include <sys/systm.h> 61 #include <sys/callout.h> 62 #include <sys/mbuf.h> 63 #include <sys/socket.h> 64 #include <sys/socketvar.h> 65 #include <sys/sockio.h> 66 #include <sys/protosw.h> 67 #include <sys/errno.h> 68 #include <sys/time.h> 69 #include <sys/kernel.h> 70 #include <sys/ioctl.h> 71 #include <sys/syslog.h> 72 73 #include <net/if.h> 74 #include <net/route.h> 75 #include <net/raw_cb.h> 76 77 #include <netinet/in.h> 78 #include <netinet/in_var.h> 79 80 #include <netinet/ip6.h> 81 #include <netinet6/ip6_var.h> 82 #include <netinet6/ip6_mroute.h> 83 #include <netinet6/pim6.h> 84 #include <netinet6/pim6_var.h> 85 #include <netinet6/nd6.h> 86 87 #include <net/net_osdep.h> 88 89 static int ip6_mdq __P((struct mbuf *, struct ifnet *, struct mf6c *)); 90 static void phyint_send __P((struct ip6_hdr *, struct mif6 *, struct mbuf *)); 91 92 static int set_pim6 __P((int *)); 93 static int get_pim6 __P((struct mbuf *)); 94 static int socket_send __P((struct socket *, struct mbuf *, 95 struct sockaddr_in6 *)); 96 static int register_send __P((struct ip6_hdr *, struct mif6 *, 97 struct mbuf *)); 98 99 /* 100 * Globals. All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static, 101 * except for netstat or debugging purposes. 102 */ 103 struct socket *ip6_mrouter = NULL; 104 int ip6_mrouter_ver = 0; 105 int ip6_mrtproto = IPPROTO_PIM; /* for netstat only */ 106 struct mrt6stat mrt6stat; 107 108 #define NO_RTE_FOUND 0x1 109 #define RTE_FOUND 0x2 110 111 struct mf6c *mf6ctable[MF6CTBLSIZ]; 112 u_char n6expire[MF6CTBLSIZ]; 113 static struct mif6 mif6table[MAXMIFS]; 114 #ifdef MRT6DEBUG 115 u_int mrt6debug = 0; /* debug level */ 116 #define DEBUG_MFC 0x02 117 #define DEBUG_FORWARD 0x04 118 #define DEBUG_EXPIRE 0x08 119 #define DEBUG_XMIT 0x10 120 #define DEBUG_REG 0x20 121 #define DEBUG_PIM 0x40 122 #endif 123 124 static void expire_upcalls __P((void *)); 125 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 126 #define UPCALL_EXPIRE 6 /* number of timeouts */ 127 128 #ifdef INET 129 #ifdef MROUTING 130 extern struct socket *ip_mrouter; 131 #endif 132 #endif 133 134 /* 135 * 'Interfaces' associated with decapsulator (so we can tell 136 * packets that went through it from ones that get reflected 137 * by a broken gateway). These interfaces are never linked into 138 * the system ifnet list & no routes point to them. I.e., packets 139 * can't be sent this way. They only exist as a placeholder for 140 * multicast source verification. 141 */ 142 struct ifnet multicast_register_if; 143 144 #define ENCAP_HOPS 64 145 146 /* 147 * Private variables. 148 */ 149 static mifi_t nummifs = 0; 150 static mifi_t reg_mif_num = (mifi_t)-1; 151 152 static struct pim6stat pim6stat; 153 static int pim6; 154 155 /* 156 * Hash function for a source, group entry 157 */ 158 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \ 159 (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \ 160 (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \ 161 (g).s6_addr32[2] ^ (g).s6_addr32[3]) 162 163 /* 164 * Find a route for a given origin IPv6 address and Multicast group address. 165 * Quality of service parameter to be added in the future!!! 166 */ 167 168 #define MF6CFIND(o, g, rt) do { \ 169 struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \ 170 rt = NULL; \ 171 mrt6stat.mrt6s_mfc_lookups++; \ 172 while (_rt) { \ 173 if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \ 174 IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \ 175 (_rt->mf6c_stall == NULL)) { \ 176 rt = _rt; \ 177 break; \ 178 } \ 179 _rt = _rt->mf6c_next; \ 180 } \ 181 if (rt == NULL) { \ 182 mrt6stat.mrt6s_mfc_misses++; \ 183 } \ 184 } while (0) 185 186 /* 187 * Macros to compute elapsed time efficiently 188 * Borrowed from Van Jacobson's scheduling code 189 */ 190 #define TV_DELTA(a, b, delta) do { \ 191 int xxs; \ 192 \ 193 delta = (a).tv_usec - (b).tv_usec; \ 194 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 195 switch (xxs) { \ 196 case 2: \ 197 delta += 1000000; \ 198 /* fall through */ \ 199 case 1: \ 200 delta += 1000000; \ 201 break; \ 202 default: \ 203 delta += (1000000 * xxs); \ 204 } \ 205 } \ 206 } while (0) 207 208 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 209 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 210 211 #ifdef UPCALL_TIMING 212 #define UPCALL_MAX 50 213 u_long upcall_data[UPCALL_MAX + 1]; 214 static void collate(); 215 #endif /* UPCALL_TIMING */ 216 217 static int get_sg_cnt __P((struct sioc_sg_req6 *)); 218 static int get_mif6_cnt __P((struct sioc_mif_req6 *)); 219 static int ip6_mrouter_init __P((struct socket *, struct mbuf *, int)); 220 static int add_m6if __P((struct mif6ctl *)); 221 static int del_m6if __P((mifi_t *)); 222 static int add_m6fc __P((struct mf6cctl *)); 223 static int del_m6fc __P((struct mf6cctl *)); 224 225 static struct callout expire_upcalls_ch = CALLOUT_INITIALIZER; 226 227 /* 228 * Handle MRT setsockopt commands to modify the multicast routing tables. 229 */ 230 int 231 ip6_mrouter_set(cmd, so, m) 232 int cmd; 233 struct socket *so; 234 struct mbuf *m; 235 { 236 if (cmd != MRT6_INIT && so != ip6_mrouter) 237 return EACCES; 238 239 switch (cmd) { 240 case MRT6_OINIT: return ip6_mrouter_init(so, m, cmd); 241 case MRT6_INIT: return ip6_mrouter_init(so, m, cmd); 242 case MRT6_DONE: return ip6_mrouter_done(); 243 case MRT6_ADD_MIF: return add_m6if(mtod(m, struct mif6ctl *)); 244 case MRT6_DEL_MIF: return del_m6if(mtod(m, mifi_t *)); 245 case MRT6_ADD_MFC: return add_m6fc(mtod(m, struct mf6cctl *)); 246 case MRT6_DEL_MFC: return del_m6fc(mtod(m, struct mf6cctl *)); 247 case MRT6_PIM: return set_pim6(mtod(m, int *)); 248 default: return EOPNOTSUPP; 249 } 250 } 251 252 /* 253 * Handle MRT getsockopt commands 254 */ 255 int 256 ip6_mrouter_get(cmd, so, m) 257 int cmd; 258 struct socket *so; 259 struct mbuf **m; 260 { 261 struct mbuf *mb; 262 263 if (so != ip6_mrouter) return EACCES; 264 265 *m = mb = m_get(M_WAIT, MT_SOOPTS); 266 267 switch (cmd) { 268 case MRT6_PIM: 269 return get_pim6(mb); 270 default: 271 m_free(mb); 272 return EOPNOTSUPP; 273 } 274 } 275 276 /* 277 * Handle ioctl commands to obtain information from the cache 278 */ 279 int 280 mrt6_ioctl(cmd, data) 281 int cmd; 282 caddr_t data; 283 { 284 int error = 0; 285 286 switch (cmd) { 287 case SIOCGETSGCNT_IN6: 288 return(get_sg_cnt((struct sioc_sg_req6 *)data)); 289 break; /* for safety */ 290 case SIOCGETMIFCNT_IN6: 291 return(get_mif6_cnt((struct sioc_mif_req6 *)data)); 292 break; /* for safety */ 293 default: 294 return (EINVAL); 295 break; 296 } 297 return error; 298 } 299 300 /* 301 * returns the packet, byte, rpf-failure count for the source group provided 302 */ 303 static int 304 get_sg_cnt(req) 305 struct sioc_sg_req6 *req; 306 { 307 struct mf6c *rt; 308 int s; 309 310 s = splsoftnet(); 311 MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt); 312 splx(s); 313 if (rt != NULL) { 314 req->pktcnt = rt->mf6c_pkt_cnt; 315 req->bytecnt = rt->mf6c_byte_cnt; 316 req->wrong_if = rt->mf6c_wrong_if; 317 } else 318 return(ESRCH); 319 #if 0 320 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 321 #endif 322 323 return 0; 324 } 325 326 /* 327 * returns the input and output packet and byte counts on the mif provided 328 */ 329 static int 330 get_mif6_cnt(req) 331 struct sioc_mif_req6 *req; 332 { 333 mifi_t mifi = req->mifi; 334 335 if (mifi >= nummifs) 336 return EINVAL; 337 338 req->icount = mif6table[mifi].m6_pkt_in; 339 req->ocount = mif6table[mifi].m6_pkt_out; 340 req->ibytes = mif6table[mifi].m6_bytes_in; 341 req->obytes = mif6table[mifi].m6_bytes_out; 342 343 return 0; 344 } 345 346 /* 347 * Get PIM processiong global 348 */ 349 static int 350 get_pim6(m) 351 struct mbuf *m; 352 { 353 int *i; 354 355 i = mtod(m, int *); 356 357 *i = pim6; 358 359 return 0; 360 } 361 362 static int 363 set_pim6(i) 364 int *i; 365 { 366 if ((*i != 1) && (*i != 0)) 367 return EINVAL; 368 369 pim6 = *i; 370 371 return 0; 372 } 373 374 /* 375 * Enable multicast routing 376 */ 377 static int 378 ip6_mrouter_init(so, m, cmd) 379 struct socket *so; 380 struct mbuf *m; 381 int cmd; 382 { 383 int *v; 384 385 #ifdef MRT6DEBUG 386 if (mrt6debug) 387 log(LOG_DEBUG, 388 "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n", 389 so->so_type, so->so_proto->pr_protocol); 390 #endif 391 392 if (so->so_type != SOCK_RAW || 393 so->so_proto->pr_protocol != IPPROTO_ICMPV6) 394 return EOPNOTSUPP; 395 396 if (!m || (m->m_len != sizeof(int *))) 397 return ENOPROTOOPT; 398 399 v = mtod(m, int *); 400 if (*v != 1) 401 return ENOPROTOOPT; 402 403 if (ip6_mrouter != NULL) return EADDRINUSE; 404 405 ip6_mrouter = so; 406 ip6_mrouter_ver = cmd; 407 408 bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); 409 bzero((caddr_t)n6expire, sizeof(n6expire)); 410 411 pim6 = 0;/* used for stubbing out/in pim stuff */ 412 413 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 414 expire_upcalls, NULL); 415 416 #ifdef MRT6DEBUG 417 if (mrt6debug) 418 log(LOG_DEBUG, "ip6_mrouter_init\n"); 419 #endif 420 421 return 0; 422 } 423 424 /* 425 * Disable multicast routing 426 */ 427 int 428 ip6_mrouter_done() 429 { 430 mifi_t mifi; 431 int i; 432 struct ifnet *ifp; 433 struct in6_ifreq ifr; 434 struct mf6c *rt; 435 struct rtdetq *rte; 436 int s; 437 438 s = splsoftnet(); 439 440 /* 441 * For each phyint in use, disable promiscuous reception of all IPv6 442 * multicasts. 443 */ 444 #ifdef INET 445 #ifdef MROUTING 446 /* 447 * If there is still IPv4 multicast routing daemon, 448 * we remain interfaces to receive all muliticasted packets. 449 * XXX: there may be an interface in which the IPv4 multicast 450 * daemon is not interested... 451 */ 452 if (!ip_mrouter) 453 #endif 454 #endif 455 { 456 for (mifi = 0; mifi < nummifs; mifi++) { 457 if (mif6table[mifi].m6_ifp && 458 !(mif6table[mifi].m6_flags & MIFF_REGISTER)) { 459 ifr.ifr_addr.sin6_family = AF_INET6; 460 ifr.ifr_addr.sin6_addr= in6addr_any; 461 ifp = mif6table[mifi].m6_ifp; 462 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, 463 (caddr_t)&ifr); 464 } 465 } 466 } 467 #ifdef notyet 468 bzero((caddr_t)qtable, sizeof(qtable)); 469 bzero((caddr_t)tbftable, sizeof(tbftable)); 470 #endif 471 bzero((caddr_t)mif6table, sizeof(mif6table)); 472 nummifs = 0; 473 474 pim6 = 0; /* used to stub out/in pim specific code */ 475 476 callout_stop(&expire_upcalls_ch); 477 478 /* 479 * Free all multicast forwarding cache entries. 480 */ 481 for (i = 0; i < MF6CTBLSIZ; i++) { 482 rt = mf6ctable[i]; 483 while (rt) { 484 struct mf6c *frt; 485 486 for (rte = rt->mf6c_stall; rte != NULL; ) { 487 struct rtdetq *n = rte->next; 488 489 m_free(rte->m); 490 free(rte, M_MRTABLE); 491 rte = n; 492 } 493 frt = rt; 494 rt = rt->mf6c_next; 495 free(frt, M_MRTABLE); 496 } 497 } 498 499 bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); 500 501 /* 502 * Reset de-encapsulation cache 503 */ 504 reg_mif_num = -1; 505 506 ip6_mrouter = NULL; 507 ip6_mrouter_ver = 0; 508 509 splx(s); 510 511 #ifdef MRT6DEBUG 512 if (mrt6debug) 513 log(LOG_DEBUG, "ip6_mrouter_done\n"); 514 #endif 515 516 return 0; 517 } 518 519 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; 520 521 /* 522 * Add a mif to the mif table 523 */ 524 static int 525 add_m6if(mifcp) 526 struct mif6ctl *mifcp; 527 { 528 struct mif6 *mifp; 529 struct ifnet *ifp; 530 struct in6_ifreq ifr; 531 int error, s; 532 #ifdef notyet 533 struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi; 534 #endif 535 536 if (mifcp->mif6c_mifi >= MAXMIFS) 537 return EINVAL; 538 mifp = mif6table + mifcp->mif6c_mifi; 539 if (mifp->m6_ifp) 540 return EADDRINUSE; /* XXX: is it appropriate? */ 541 if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index) 542 return ENXIO; 543 /* 544 * XXX: some OSes can remove ifp and clear ifindex2ifnet[id] 545 * even for id between 0 and if_index. 546 */ 547 if ((ifp = ifindex2ifnet[mifcp->mif6c_pifi]) == NULL) 548 return ENXIO; 549 550 if (mifcp->mif6c_flags & MIFF_REGISTER) { 551 if (reg_mif_num == (mifi_t)-1) { 552 strcpy(multicast_register_if.if_xname, 553 "register_mif"); /* XXX */ 554 multicast_register_if.if_flags |= IFF_LOOPBACK; 555 multicast_register_if.if_index = mifcp->mif6c_mifi; 556 reg_mif_num = mifcp->mif6c_mifi; 557 } 558 559 ifp = &multicast_register_if; 560 561 } /* if REGISTER */ 562 else { 563 /* Make sure the interface supports multicast */ 564 if ((ifp->if_flags & IFF_MULTICAST) == 0) 565 return EOPNOTSUPP; 566 567 s = splsoftnet(); 568 /* 569 * Enable promiscuous reception of all IPv6 multicasts 570 * from the interface. 571 */ 572 ifr.ifr_addr.sin6_family = AF_INET6; 573 ifr.ifr_addr.sin6_addr = in6addr_any; 574 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr); 575 splx(s); 576 if (error) 577 return error; 578 } 579 580 s = splsoftnet(); 581 mifp->m6_flags = mifcp->mif6c_flags; 582 mifp->m6_ifp = ifp; 583 #ifdef notyet 584 /* scaling up here allows division by 1024 in critical code */ 585 mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000; 586 #endif 587 /* initialize per mif pkt counters */ 588 mifp->m6_pkt_in = 0; 589 mifp->m6_pkt_out = 0; 590 mifp->m6_bytes_in = 0; 591 mifp->m6_bytes_out = 0; 592 splx(s); 593 594 /* Adjust nummifs up if the mifi is higher than nummifs */ 595 if (nummifs <= mifcp->mif6c_mifi) 596 nummifs = mifcp->mif6c_mifi + 1; 597 598 #ifdef MRT6DEBUG 599 if (mrt6debug) 600 log(LOG_DEBUG, 601 "add_mif #%d, phyint %s%d\n", 602 mifcp->mif6c_mifi, 603 ifp->if_name, ifp->if_unit); 604 #endif 605 606 return 0; 607 } 608 609 /* 610 * Delete a mif from the mif table 611 */ 612 static int 613 del_m6if(mifip) 614 mifi_t *mifip; 615 { 616 struct mif6 *mifp = mif6table + *mifip; 617 mifi_t mifi; 618 struct ifnet *ifp; 619 struct in6_ifreq ifr; 620 int s; 621 622 if (*mifip >= nummifs) 623 return EINVAL; 624 if (mifp->m6_ifp == NULL) 625 return EINVAL; 626 627 s = splsoftnet(); 628 629 if (!(mifp->m6_flags & MIFF_REGISTER)) { 630 /* 631 * XXX: what if there is yet IPv4 multicast daemon 632 * using the interface? 633 */ 634 ifp = mifp->m6_ifp; 635 636 ifr.ifr_addr.sin6_family = AF_INET6; 637 ifr.ifr_addr.sin6_addr = in6addr_any; 638 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 639 } 640 641 #ifdef notyet 642 bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip])); 643 bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf))); 644 #endif 645 bzero((caddr_t)mifp, sizeof (*mifp)); 646 647 /* Adjust nummifs down */ 648 for (mifi = nummifs; mifi > 0; mifi--) 649 if (mif6table[mifi - 1].m6_ifp) 650 break; 651 nummifs = mifi; 652 653 splx(s); 654 655 #ifdef MRT6DEBUG 656 if (mrt6debug) 657 log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs); 658 #endif 659 660 return 0; 661 } 662 663 /* 664 * Add an mfc entry 665 */ 666 static int 667 add_m6fc(mfccp) 668 struct mf6cctl *mfccp; 669 { 670 struct mf6c *rt; 671 u_long hash; 672 struct rtdetq *rte; 673 u_short nstl; 674 int s; 675 676 MF6CFIND(mfccp->mf6cc_origin.sin6_addr, 677 mfccp->mf6cc_mcastgrp.sin6_addr, rt); 678 679 /* If an entry already exists, just update the fields */ 680 if (rt) { 681 #ifdef MRT6DEBUG 682 if (mrt6debug & DEBUG_MFC) 683 log(LOG_DEBUG,"add_m6fc update o %s g %s p %x\n", 684 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 685 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 686 mfccp->mf6cc_parent); 687 #endif 688 689 s = splsoftnet(); 690 rt->mf6c_parent = mfccp->mf6cc_parent; 691 rt->mf6c_ifset = mfccp->mf6cc_ifset; 692 splx(s); 693 return 0; 694 } 695 696 /* 697 * Find the entry for which the upcall was made and update 698 */ 699 s = splsoftnet(); 700 hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr, 701 mfccp->mf6cc_mcastgrp.sin6_addr); 702 for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) { 703 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, 704 &mfccp->mf6cc_origin.sin6_addr) && 705 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, 706 &mfccp->mf6cc_mcastgrp.sin6_addr) && 707 (rt->mf6c_stall != NULL)) { 708 709 if (nstl++) 710 log(LOG_ERR, 711 "add_m6fc: %s o %s g %s p %x dbx %p\n", 712 "multiple kernel entries", 713 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 714 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 715 mfccp->mf6cc_parent, rt->mf6c_stall); 716 717 #ifdef MRT6DEBUG 718 if (mrt6debug & DEBUG_MFC) 719 log(LOG_DEBUG, 720 "add_m6fc o %s g %s p %x dbg %x\n", 721 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 722 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 723 mfccp->mf6cc_parent, rt->mf6c_stall); 724 #endif 725 726 rt->mf6c_origin = mfccp->mf6cc_origin; 727 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 728 rt->mf6c_parent = mfccp->mf6cc_parent; 729 rt->mf6c_ifset = mfccp->mf6cc_ifset; 730 /* initialize pkt counters per src-grp */ 731 rt->mf6c_pkt_cnt = 0; 732 rt->mf6c_byte_cnt = 0; 733 rt->mf6c_wrong_if = 0; 734 735 rt->mf6c_expire = 0; /* Don't clean this guy up */ 736 n6expire[hash]--; 737 738 /* free packets Qed at the end of this entry */ 739 for (rte = rt->mf6c_stall; rte != NULL; ) { 740 struct rtdetq *n = rte->next; 741 ip6_mdq(rte->m, rte->ifp, rt); 742 m_freem(rte->m); 743 #ifdef UPCALL_TIMING 744 collate(&(rte->t)); 745 #endif /* UPCALL_TIMING */ 746 free(rte, M_MRTABLE); 747 rte = n; 748 } 749 rt->mf6c_stall = NULL; 750 } 751 } 752 753 /* 754 * It is possible that an entry is being inserted without an upcall 755 */ 756 if (nstl == 0) { 757 #ifdef MRT6DEBUG 758 if (mrt6debug & DEBUG_MFC) 759 log(LOG_DEBUG, 760 "add_mfc no upcall h %d o %s g %s p %x\n", 761 hash, 762 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 763 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 764 mfccp->mf6cc_parent); 765 #endif 766 767 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { 768 769 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, 770 &mfccp->mf6cc_origin.sin6_addr)&& 771 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, 772 &mfccp->mf6cc_mcastgrp.sin6_addr)) { 773 774 rt->mf6c_origin = mfccp->mf6cc_origin; 775 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 776 rt->mf6c_parent = mfccp->mf6cc_parent; 777 rt->mf6c_ifset = mfccp->mf6cc_ifset; 778 /* initialize pkt counters per src-grp */ 779 rt->mf6c_pkt_cnt = 0; 780 rt->mf6c_byte_cnt = 0; 781 rt->mf6c_wrong_if = 0; 782 783 if (rt->mf6c_expire) 784 n6expire[hash]--; 785 rt->mf6c_expire = 0; 786 } 787 } 788 if (rt == NULL) { 789 /* no upcall, so make a new entry */ 790 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE, 791 M_NOWAIT); 792 if (rt == NULL) { 793 splx(s); 794 return ENOBUFS; 795 } 796 797 /* insert new entry at head of hash chain */ 798 rt->mf6c_origin = mfccp->mf6cc_origin; 799 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 800 rt->mf6c_parent = mfccp->mf6cc_parent; 801 rt->mf6c_ifset = mfccp->mf6cc_ifset; 802 /* initialize pkt counters per src-grp */ 803 rt->mf6c_pkt_cnt = 0; 804 rt->mf6c_byte_cnt = 0; 805 rt->mf6c_wrong_if = 0; 806 rt->mf6c_expire = 0; 807 rt->mf6c_stall = NULL; 808 809 /* link into table */ 810 rt->mf6c_next = mf6ctable[hash]; 811 mf6ctable[hash] = rt; 812 } 813 } 814 splx(s); 815 return 0; 816 } 817 818 #ifdef UPCALL_TIMING 819 /* 820 * collect delay statistics on the upcalls 821 */ 822 static void 823 collate(t) 824 struct timeval *t; 825 { 826 u_long d; 827 struct timeval tp; 828 u_long delta; 829 830 GET_TIME(tp); 831 832 if (TV_LT(*t, tp)) 833 { 834 TV_DELTA(tp, *t, delta); 835 836 d = delta >> 10; 837 if (d > UPCALL_MAX) 838 d = UPCALL_MAX; 839 840 ++upcall_data[d]; 841 } 842 } 843 #endif /* UPCALL_TIMING */ 844 845 /* 846 * Delete an mfc entry 847 */ 848 static int 849 del_m6fc(mfccp) 850 struct mf6cctl *mfccp; 851 { 852 struct sockaddr_in6 origin; 853 struct sockaddr_in6 mcastgrp; 854 struct mf6c *rt; 855 struct mf6c **nptr; 856 u_long hash; 857 int s; 858 859 origin = mfccp->mf6cc_origin; 860 mcastgrp = mfccp->mf6cc_mcastgrp; 861 hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr); 862 863 #ifdef MRT6DEBUG 864 if (mrt6debug & DEBUG_MFC) 865 log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n", 866 ip6_sprintf(&origin.sin6_addr), 867 ip6_sprintf(&mcastgrp.sin6_addr)); 868 #endif 869 870 s = splsoftnet(); 871 872 nptr = &mf6ctable[hash]; 873 while ((rt = *nptr) != NULL) { 874 if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr, 875 &rt->mf6c_origin.sin6_addr) && 876 IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr, 877 &rt->mf6c_mcastgrp.sin6_addr) && 878 rt->mf6c_stall == NULL) 879 break; 880 881 nptr = &rt->mf6c_next; 882 } 883 if (rt == NULL) { 884 splx(s); 885 return EADDRNOTAVAIL; 886 } 887 888 *nptr = rt->mf6c_next; 889 free(rt, M_MRTABLE); 890 891 splx(s); 892 893 return 0; 894 } 895 896 static int 897 socket_send(s, mm, src) 898 struct socket *s; 899 struct mbuf *mm; 900 struct sockaddr_in6 *src; 901 { 902 if (s) { 903 if (sbappendaddr(&s->so_rcv, 904 (struct sockaddr *)src, 905 mm, (struct mbuf *)0) != 0) { 906 sorwakeup(s); 907 return 0; 908 } 909 } 910 m_freem(mm); 911 return -1; 912 } 913 914 /* 915 * IPv6 multicast forwarding function. This function assumes that the packet 916 * pointed to by "ip6" has arrived on (or is about to be sent to) the interface 917 * pointed to by "ifp", and the packet is to be relayed to other networks 918 * that have members of the packet's destination IPv6 multicast group. 919 * 920 * The packet is returned unscathed to the caller, unless it is 921 * erroneous, in which case a non-zero return value tells the caller to 922 * discard it. 923 */ 924 925 int 926 ip6_mforward(ip6, ifp, m) 927 struct ip6_hdr *ip6; 928 struct ifnet *ifp; 929 struct mbuf *m; 930 { 931 struct mf6c *rt; 932 struct mif6 *mifp; 933 struct mbuf *mm; 934 int s; 935 mifi_t mifi; 936 long time_second = time.tv_sec; 937 938 #ifdef MRT6DEBUG 939 if (mrt6debug & DEBUG_FORWARD) 940 log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n", 941 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst), 942 ifp->if_index); 943 #endif 944 945 /* 946 * Don't forward a packet with Hop limit of zero or one, 947 * or a packet destined to a local-only group. 948 */ 949 if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) || 950 IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst)) 951 return 0; 952 ip6->ip6_hlim--; 953 954 /* 955 * Source address check: do not forward packets with unspecified 956 * source. It was discussed in July 2000, on ipngwg mailing list. 957 * This is rather more serious than unicast cases, because some 958 * MLD packets can be sent with the unspecified source address 959 * (although such packets must normally set 1 to the hop limit field). 960 */ 961 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 962 ip6stat.ip6s_cantforward++; 963 if (ip6_log_time + ip6_log_interval < time_second) { 964 ip6_log_time = time_second; 965 log(LOG_DEBUG, 966 "cannot forward " 967 "from %s to %s nxt %d received on %s\n", 968 ip6_sprintf(&ip6->ip6_src), 969 ip6_sprintf(&ip6->ip6_dst), 970 ip6->ip6_nxt, 971 m->m_pkthdr.rcvif ? 972 if_name(m->m_pkthdr.rcvif) : "?"); 973 } 974 return 0; 975 } 976 977 /* 978 * Determine forwarding mifs from the forwarding cache table 979 */ 980 s = splsoftnet(); 981 MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt); 982 983 /* Entry exists, so forward if necessary */ 984 if (rt) { 985 splx(s); 986 return (ip6_mdq(m, ifp, rt)); 987 } else { 988 /* 989 * If we don't have a route for packet's origin, 990 * Make a copy of the packet & 991 * send message to routing daemon 992 */ 993 994 struct mbuf *mb0; 995 struct rtdetq *rte; 996 u_long hash; 997 /* int i, npkts;*/ 998 #ifdef UPCALL_TIMING 999 struct timeval tp; 1000 1001 GET_TIME(tp); 1002 #endif /* UPCALL_TIMING */ 1003 1004 mrt6stat.mrt6s_no_route++; 1005 #ifdef MRT6DEBUG 1006 if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC)) 1007 log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n", 1008 ip6_sprintf(&ip6->ip6_src), 1009 ip6_sprintf(&ip6->ip6_dst)); 1010 #endif 1011 1012 /* 1013 * Allocate mbufs early so that we don't do extra work if we 1014 * are just going to fail anyway. 1015 */ 1016 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, 1017 M_NOWAIT); 1018 if (rte == NULL) { 1019 splx(s); 1020 return ENOBUFS; 1021 } 1022 mb0 = m_copy(m, 0, M_COPYALL); 1023 /* 1024 * Pullup packet header if needed before storing it, 1025 * as other references may modify it in the meantime. 1026 */ 1027 if (mb0 && 1028 (M_READONLY(mb0) || mb0->m_len < sizeof(struct ip6_hdr))) 1029 mb0 = m_pullup(mb0, sizeof(struct ip6_hdr)); 1030 if (mb0 == NULL) { 1031 free(rte, M_MRTABLE); 1032 splx(s); 1033 return ENOBUFS; 1034 } 1035 1036 /* is there an upcall waiting for this packet? */ 1037 hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst); 1038 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { 1039 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, 1040 &rt->mf6c_origin.sin6_addr) && 1041 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 1042 &rt->mf6c_mcastgrp.sin6_addr) && 1043 (rt->mf6c_stall != NULL)) 1044 break; 1045 } 1046 1047 if (rt == NULL) { 1048 struct mrt6msg *im; 1049 struct omrt6msg *oim; 1050 1051 /* no upcall, so make a new entry */ 1052 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE, 1053 M_NOWAIT); 1054 if (rt == NULL) { 1055 free(rte, M_MRTABLE); 1056 m_freem(mb0); 1057 splx(s); 1058 return ENOBUFS; 1059 } 1060 /* 1061 * Make a copy of the header to send to the user 1062 * level process 1063 */ 1064 mm = m_copy(mb0, 0, sizeof(struct ip6_hdr)); 1065 1066 if (mm == NULL) { 1067 free(rte, M_MRTABLE); 1068 m_freem(mb0); 1069 free(rt, M_MRTABLE); 1070 splx(s); 1071 return ENOBUFS; 1072 } 1073 1074 /* 1075 * Send message to routing daemon 1076 */ 1077 sin6.sin6_addr = ip6->ip6_src; 1078 1079 im = NULL; 1080 oim = NULL; 1081 switch (ip6_mrouter_ver) { 1082 case MRT6_OINIT: 1083 oim = mtod(mm, struct omrt6msg *); 1084 oim->im6_msgtype = MRT6MSG_NOCACHE; 1085 oim->im6_mbz = 0; 1086 break; 1087 case MRT6_INIT: 1088 im = mtod(mm, struct mrt6msg *); 1089 im->im6_msgtype = MRT6MSG_NOCACHE; 1090 im->im6_mbz = 0; 1091 break; 1092 default: 1093 free(rte, M_MRTABLE); 1094 m_freem(mb0); 1095 free(rt, M_MRTABLE); 1096 splx(s); 1097 return EINVAL; 1098 } 1099 1100 #ifdef MRT6DEBUG 1101 if (mrt6debug & DEBUG_FORWARD) 1102 log(LOG_DEBUG, 1103 "getting the iif info in the kernel\n"); 1104 #endif 1105 1106 for (mifp = mif6table, mifi = 0; 1107 mifi < nummifs && mifp->m6_ifp != ifp; 1108 mifp++, mifi++) 1109 ; 1110 1111 switch (ip6_mrouter_ver) { 1112 case MRT6_OINIT: 1113 oim->im6_mif = mifi; 1114 break; 1115 case MRT6_INIT: 1116 im->im6_mif = mifi; 1117 break; 1118 } 1119 1120 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1121 log(LOG_WARNING, "ip6_mforward: ip6_mrouter " 1122 "socket queue full\n"); 1123 mrt6stat.mrt6s_upq_sockfull++; 1124 free(rte, M_MRTABLE); 1125 m_freem(mb0); 1126 free(rt, M_MRTABLE); 1127 splx(s); 1128 return ENOBUFS; 1129 } 1130 1131 mrt6stat.mrt6s_upcalls++; 1132 1133 /* insert new entry at head of hash chain */ 1134 bzero(rt, sizeof(*rt)); 1135 rt->mf6c_origin.sin6_family = AF_INET6; 1136 rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6); 1137 rt->mf6c_origin.sin6_addr = ip6->ip6_src; 1138 rt->mf6c_mcastgrp.sin6_family = AF_INET6; 1139 rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6); 1140 rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst; 1141 rt->mf6c_expire = UPCALL_EXPIRE; 1142 n6expire[hash]++; 1143 rt->mf6c_parent = MF6C_INCOMPLETE_PARENT; 1144 1145 /* link into table */ 1146 rt->mf6c_next = mf6ctable[hash]; 1147 mf6ctable[hash] = rt; 1148 /* Add this entry to the end of the queue */ 1149 rt->mf6c_stall = rte; 1150 } else { 1151 /* determine if q has overflowed */ 1152 struct rtdetq **p; 1153 int npkts = 0; 1154 1155 for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next) 1156 if (++npkts > MAX_UPQ6) { 1157 mrt6stat.mrt6s_upq_ovflw++; 1158 free(rte, M_MRTABLE); 1159 m_freem(mb0); 1160 splx(s); 1161 return 0; 1162 } 1163 1164 /* Add this entry to the end of the queue */ 1165 *p = rte; 1166 } 1167 1168 rte->next = NULL; 1169 rte->m = mb0; 1170 rte->ifp = ifp; 1171 #ifdef UPCALL_TIMING 1172 rte->t = tp; 1173 #endif /* UPCALL_TIMING */ 1174 1175 splx(s); 1176 1177 return 0; 1178 } 1179 } 1180 1181 /* 1182 * Clean up cache entries if upcalls are not serviced 1183 * Call from the Slow Timeout mechanism, every half second. 1184 */ 1185 static void 1186 expire_upcalls(unused) 1187 void *unused; 1188 { 1189 struct rtdetq *rte; 1190 struct mf6c *mfc, **nptr; 1191 int i; 1192 int s; 1193 1194 s = splsoftnet(); 1195 for (i = 0; i < MF6CTBLSIZ; i++) { 1196 if (n6expire[i] == 0) 1197 continue; 1198 nptr = &mf6ctable[i]; 1199 while ((mfc = *nptr) != NULL) { 1200 rte = mfc->mf6c_stall; 1201 /* 1202 * Skip real cache entries 1203 * Make sure it wasn't marked to not expire (shouldn't happen) 1204 * If it expires now 1205 */ 1206 if (rte != NULL && 1207 mfc->mf6c_expire != 0 && 1208 --mfc->mf6c_expire == 0) { 1209 #ifdef MRT6DEBUG 1210 if (mrt6debug & DEBUG_EXPIRE) 1211 log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n", 1212 ip6_sprintf(&mfc->mf6c_origin.sin6_addr), 1213 ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr)); 1214 #endif 1215 /* 1216 * drop all the packets 1217 * free the mbuf with the pkt, if, timing info 1218 */ 1219 do { 1220 struct rtdetq *n = rte->next; 1221 m_freem(rte->m); 1222 free(rte, M_MRTABLE); 1223 rte = n; 1224 } while (rte != NULL); 1225 mrt6stat.mrt6s_cache_cleanups++; 1226 n6expire[i]--; 1227 1228 *nptr = mfc->mf6c_next; 1229 free(mfc, M_MRTABLE); 1230 } else { 1231 nptr = &mfc->mf6c_next; 1232 } 1233 } 1234 } 1235 splx(s); 1236 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1237 expire_upcalls, NULL); 1238 } 1239 1240 /* 1241 * Packet forwarding routine once entry in the cache is made 1242 */ 1243 static int 1244 ip6_mdq(m, ifp, rt) 1245 struct mbuf *m; 1246 struct ifnet *ifp; 1247 struct mf6c *rt; 1248 { 1249 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1250 mifi_t mifi, iif; 1251 struct mif6 *mifp; 1252 int plen = m->m_pkthdr.len; 1253 1254 /* 1255 * Macro to send packet on mif. Since RSVP packets don't get counted on 1256 * input, they shouldn't get counted on output, so statistics keeping is 1257 * separate. 1258 */ 1259 1260 #define MC6_SEND(ip6, mifp, m) do { \ 1261 if ((mifp)->m6_flags & MIFF_REGISTER) \ 1262 register_send((ip6), (mifp), (m)); \ 1263 else \ 1264 phyint_send((ip6), (mifp), (m)); \ 1265 } while (0) 1266 1267 /* 1268 * Don't forward if it didn't arrive from the parent mif 1269 * for its origin. 1270 */ 1271 mifi = rt->mf6c_parent; 1272 if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) { 1273 /* came in the wrong interface */ 1274 #ifdef MRT6DEBUG 1275 if (mrt6debug & DEBUG_FORWARD) 1276 log(LOG_DEBUG, 1277 "wrong if: ifid %d mifi %d mififid %x\n", 1278 ifp->if_index, mifi, 1279 mif6table[mifi].m6_ifp->if_index); 1280 #endif 1281 mrt6stat.mrt6s_wrong_if++; 1282 rt->mf6c_wrong_if++; 1283 /* 1284 * If we are doing PIM processing, and we are forwarding 1285 * packets on this interface, send a message to the 1286 * routing daemon. 1287 */ 1288 /* have to make sure this is a valid mif */ 1289 if (mifi < nummifs && mif6table[mifi].m6_ifp) 1290 if (pim6 && (m->m_flags & M_LOOP) == 0) { 1291 /* 1292 * Check the M_LOOP flag to avoid an 1293 * unnecessary PIM assert. 1294 * XXX: M_LOOP is an ad-hoc hack... 1295 */ 1296 static struct sockaddr_in6 sin6 = 1297 { sizeof(sin6), AF_INET6 }; 1298 1299 struct mbuf *mm; 1300 struct mrt6msg *im; 1301 struct omrt6msg *oim; 1302 1303 mm = m_copy(m, 0, sizeof(struct ip6_hdr)); 1304 if (mm && 1305 (M_READONLY(mm) || 1306 mm->m_len < sizeof(struct ip6_hdr))) 1307 mm = m_pullup(mm, sizeof(struct ip6_hdr)); 1308 if (mm == NULL) 1309 return ENOBUFS; 1310 1311 oim = NULL; 1312 im = NULL; 1313 switch (ip6_mrouter_ver) { 1314 case MRT6_OINIT: 1315 oim = mtod(mm, struct omrt6msg *); 1316 oim->im6_msgtype = MRT6MSG_WRONGMIF; 1317 oim->im6_mbz = 0; 1318 break; 1319 case MRT6_INIT: 1320 im = mtod(mm, struct mrt6msg *); 1321 im->im6_msgtype = MRT6MSG_WRONGMIF; 1322 im->im6_mbz = 0; 1323 break; 1324 default: 1325 m_freem(mm); 1326 return EINVAL; 1327 } 1328 1329 for (mifp = mif6table, iif = 0; 1330 iif < nummifs && mifp && 1331 mifp->m6_ifp != ifp; 1332 mifp++, iif++) 1333 ; 1334 1335 switch (ip6_mrouter_ver) { 1336 case MRT6_OINIT: 1337 oim->im6_mif = iif; 1338 sin6.sin6_addr = oim->im6_src; 1339 break; 1340 case MRT6_INIT: 1341 im->im6_mif = iif; 1342 sin6.sin6_addr = im->im6_src; 1343 break; 1344 } 1345 1346 mrt6stat.mrt6s_upcalls++; 1347 1348 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1349 #ifdef MRT6DEBUG 1350 if (mrt6debug) 1351 log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n"); 1352 #endif 1353 ++mrt6stat.mrt6s_upq_sockfull; 1354 return ENOBUFS; 1355 } /* if socket Q full */ 1356 } /* if PIM */ 1357 return 0; 1358 } /* if wrong iif */ 1359 1360 /* If I sourced this packet, it counts as output, else it was input. */ 1361 if (m->m_pkthdr.rcvif == NULL) { 1362 /* XXX: is rcvif really NULL when output?? */ 1363 mif6table[mifi].m6_pkt_out++; 1364 mif6table[mifi].m6_bytes_out += plen; 1365 } else { 1366 mif6table[mifi].m6_pkt_in++; 1367 mif6table[mifi].m6_bytes_in += plen; 1368 } 1369 rt->mf6c_pkt_cnt++; 1370 rt->mf6c_byte_cnt += plen; 1371 1372 /* 1373 * For each mif, forward a copy of the packet if there are group 1374 * members downstream on the interface. 1375 */ 1376 for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) 1377 if (IF_ISSET(mifi, &rt->mf6c_ifset)) { 1378 #ifdef notyet 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 #endif 1398 1399 mifp->m6_pkt_out++; 1400 mifp->m6_bytes_out += plen; 1401 MC6_SEND(ip6, mifp, m); 1402 } 1403 return 0; 1404 } 1405 1406 static void 1407 phyint_send(ip6, mifp, m) 1408 struct ip6_hdr *ip6; 1409 struct mif6 *mifp; 1410 struct mbuf *m; 1411 { 1412 struct mbuf *mb_copy; 1413 struct ifnet *ifp = mifp->m6_ifp; 1414 int error = 0; 1415 int s = splsoftnet(); 1416 static struct route_in6 ro; 1417 struct in6_multi *in6m; 1418 struct sockaddr_in6 *dst6; 1419 u_long linkmtu; 1420 1421 /* 1422 * Make a new reference to the packet; make sure that 1423 * the IPv6 header is actually copied, not just referenced, 1424 * so that ip6_output() only scribbles on the copy. 1425 */ 1426 mb_copy = m_copy(m, 0, M_COPYALL); 1427 if (mb_copy && 1428 (M_READONLY(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr))) 1429 mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr)); 1430 if (mb_copy == NULL) { 1431 splx(s); 1432 return; 1433 } 1434 /* set MCAST flag to the outgoing packet */ 1435 mb_copy->m_flags |= M_MCAST; 1436 1437 /* 1438 * If we sourced the packet, call ip6_output since we may devide 1439 * the packet into fragments when the packet is too big for the 1440 * outgoing interface. 1441 * Otherwise, we can simply send the packet to the interface 1442 * sending queue. 1443 */ 1444 if (m->m_pkthdr.rcvif == NULL) { 1445 struct ip6_moptions im6o; 1446 1447 im6o.im6o_multicast_ifp = ifp; 1448 /* XXX: ip6_output will override ip6->ip6_hlim */ 1449 im6o.im6o_multicast_hlim = ip6->ip6_hlim; 1450 im6o.im6o_multicast_loop = 1; 1451 error = ip6_output(mb_copy, NULL, &ro, 1452 IPV6_FORWARDING, &im6o, NULL); 1453 1454 #ifdef MRT6DEBUG 1455 if (mrt6debug & DEBUG_XMIT) 1456 log(LOG_DEBUG, "phyint_send on mif %d err %d\n", 1457 mifp - mif6table, error); 1458 #endif 1459 splx(s); 1460 return; 1461 } 1462 1463 /* 1464 * If we belong to the destination multicast group 1465 * on the outgoing interface, loop back a copy. 1466 */ 1467 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 1468 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 1469 if (in6m != NULL) { 1470 dst6->sin6_len = sizeof(struct sockaddr_in6); 1471 dst6->sin6_family = AF_INET6; 1472 dst6->sin6_addr = ip6->ip6_dst; 1473 ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst); 1474 } 1475 /* 1476 * Put the packet into the sending queue of the outgoing interface 1477 * if it would fit in the MTU of the interface. 1478 */ 1479 linkmtu = IN6_LINKMTU(ifp); 1480 if (mb_copy->m_pkthdr.len < linkmtu || linkmtu < IPV6_MMTU) { 1481 dst6->sin6_len = sizeof(struct sockaddr_in6); 1482 dst6->sin6_family = AF_INET6; 1483 dst6->sin6_addr = ip6->ip6_dst; 1484 /* 1485 * We just call if_output instead of nd6_output here, since 1486 * we need no ND for a multicast forwarded packet...right? 1487 */ 1488 error = (*ifp->if_output)(ifp, mb_copy, 1489 (struct sockaddr *)&ro.ro_dst, NULL); 1490 #ifdef MRT6DEBUG 1491 if (mrt6debug & DEBUG_XMIT) 1492 log(LOG_DEBUG, "phyint_send on mif %d err %d\n", 1493 mifp - mif6table, error); 1494 #endif 1495 } else { 1496 #ifdef MULTICAST_PMTUD 1497 icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu); 1498 #else 1499 #ifdef MRT6DEBUG 1500 if (mrt6debug & DEBUG_XMIT) 1501 log(LOG_DEBUG, 1502 "phyint_send: packet too big on %s o %s g %s" 1503 " size %d(discarded)\n", 1504 if_name(ifp), 1505 ip6_sprintf(&ip6->ip6_src), 1506 ip6_sprintf(&ip6->ip6_dst), 1507 mb_copy->m_pkthdr.len); 1508 #endif /* MRT6DEBUG */ 1509 m_freem(mb_copy); /* simply discard the packet */ 1510 #endif 1511 } 1512 1513 splx(s); 1514 } 1515 1516 static int 1517 register_send(ip6, mif, m) 1518 struct ip6_hdr *ip6; 1519 struct mif6 *mif; 1520 struct mbuf *m; 1521 { 1522 struct mbuf *mm; 1523 int i, len = m->m_pkthdr.len; 1524 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; 1525 struct mrt6msg *im6; 1526 1527 #ifdef MRT6DEBUG 1528 if (mrt6debug) 1529 log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n", 1530 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst)); 1531 #endif 1532 ++pim6stat.pim6s_snd_registers; 1533 1534 /* Make a copy of the packet to send to the user level process */ 1535 MGETHDR(mm, M_DONTWAIT, MT_HEADER); 1536 if (mm == NULL) 1537 return ENOBUFS; 1538 mm->m_data += max_linkhdr; 1539 mm->m_len = sizeof(struct ip6_hdr); 1540 1541 if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1542 m_freem(mm); 1543 return ENOBUFS; 1544 } 1545 i = MHLEN - M_LEADINGSPACE(mm); 1546 if (i > len) 1547 i = len; 1548 mm = m_pullup(mm, i); 1549 if (mm == NULL) 1550 return ENOBUFS; 1551 /* TODO: check it! */ 1552 mm->m_pkthdr.len = len + sizeof(struct ip6_hdr); 1553 1554 /* 1555 * Send message to routing daemon 1556 */ 1557 sin6.sin6_addr = ip6->ip6_src; 1558 1559 im6 = mtod(mm, struct mrt6msg *); 1560 im6->im6_msgtype = MRT6MSG_WHOLEPKT; 1561 im6->im6_mbz = 0; 1562 1563 im6->im6_mif = mif - mif6table; 1564 1565 /* iif info is not given for reg. encap.n */ 1566 mrt6stat.mrt6s_upcalls++; 1567 1568 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1569 #ifdef MRT6DEBUG 1570 if (mrt6debug) 1571 log(LOG_WARNING, 1572 "register_send: ip_mrouter socket queue full\n"); 1573 #endif 1574 ++mrt6stat.mrt6s_upq_sockfull; 1575 return ENOBUFS; 1576 } 1577 return 0; 1578 } 1579 1580 /* 1581 * PIM sparse mode hook 1582 * Receives the pim control messages, and passes them up to the listening 1583 * socket, using rip6_input. 1584 * The only message processed is the REGISTER pim message; the pim header 1585 * is stripped off, and the inner packet is passed to register_mforward. 1586 */ 1587 int 1588 pim6_input(mp, offp, proto) 1589 struct mbuf **mp; 1590 int *offp, proto; 1591 { 1592 struct pim *pim; /* pointer to a pim struct */ 1593 struct ip6_hdr *ip6; 1594 int pimlen; 1595 struct mbuf *m = *mp; 1596 int minlen; 1597 int off = *offp; 1598 1599 ++pim6stat.pim6s_rcv_total; 1600 1601 ip6 = mtod(m, struct ip6_hdr *); 1602 pimlen = m->m_pkthdr.len - *offp; 1603 1604 /* 1605 * Validate lengths 1606 */ 1607 if (pimlen < PIM_MINLEN) { 1608 ++pim6stat.pim6s_rcv_tooshort; 1609 #ifdef MRT6DEBUG 1610 if (mrt6debug & DEBUG_PIM) 1611 log(LOG_DEBUG,"pim6_input: PIM packet too short\n"); 1612 #endif 1613 m_freem(m); 1614 return(IPPROTO_DONE); 1615 } 1616 1617 /* 1618 * if the packet is at least as big as a REGISTER, go ahead 1619 * and grab the PIM REGISTER header size, to avoid another 1620 * possible m_pullup() later. 1621 * 1622 * PIM_MINLEN == pimhdr + u_int32 == 8 1623 * PIM6_REG_MINLEN == pimhdr + reghdr + eip6hdr == 4 + 4 + 40 1624 */ 1625 minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN; 1626 1627 /* 1628 * Make sure that the IP6 and PIM headers in contiguous memory, and 1629 * possibly the PIM REGISTER header 1630 */ 1631 #ifndef PULLDOWN_TEST 1632 IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE); 1633 /* adjust pointer */ 1634 ip6 = mtod(m, struct ip6_hdr *); 1635 1636 /* adjust mbuf to point to the PIM header */ 1637 pim = (struct pim *)((caddr_t)ip6 + off); 1638 #else 1639 IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen); 1640 if (pim == NULL) { 1641 pim6stat.pim6s_rcv_tooshort++; 1642 return IPPROTO_DONE; 1643 } 1644 #endif 1645 1646 #define PIM6_CHECKSUM 1647 #ifdef PIM6_CHECKSUM 1648 { 1649 int cksumlen; 1650 1651 /* 1652 * Validate checksum. 1653 * If PIM REGISTER, exclude the data packet 1654 */ 1655 if (pim->pim_type == PIM_REGISTER) 1656 cksumlen = PIM_MINLEN; 1657 else 1658 cksumlen = pimlen; 1659 1660 if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) { 1661 ++pim6stat.pim6s_rcv_badsum; 1662 #ifdef MRT6DEBUG 1663 if (mrt6debug & DEBUG_PIM) 1664 log(LOG_DEBUG, 1665 "pim6_input: invalid checksum\n"); 1666 #endif 1667 m_freem(m); 1668 return(IPPROTO_DONE); 1669 } 1670 } 1671 #endif /* PIM_CHECKSUM */ 1672 1673 /* PIM version check */ 1674 if (pim->pim_ver != PIM_VERSION) { 1675 ++pim6stat.pim6s_rcv_badversion; 1676 #ifdef MRT6DEBUG 1677 log(LOG_ERR, 1678 "pim6_input: incorrect version %d, expecting %d\n", 1679 pim->pim_ver, PIM_VERSION); 1680 #endif 1681 m_freem(m); 1682 return(IPPROTO_DONE); 1683 } 1684 1685 if (pim->pim_type == PIM_REGISTER) { 1686 /* 1687 * since this is a REGISTER, we'll make a copy of the register 1688 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the 1689 * routing daemon. 1690 */ 1691 static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 }; 1692 1693 struct mbuf *mcp; 1694 struct ip6_hdr *eip6; 1695 u_int32_t *reghdr; 1696 int rc; 1697 1698 ++pim6stat.pim6s_rcv_registers; 1699 1700 if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) { 1701 #ifdef MRT6DEBUG 1702 if (mrt6debug & DEBUG_PIM) 1703 log(LOG_DEBUG, 1704 "pim6_input: register mif not set: %d\n", 1705 reg_mif_num); 1706 #endif 1707 m_freem(m); 1708 return(IPPROTO_DONE); 1709 } 1710 1711 reghdr = (u_int32_t *)(pim + 1); 1712 1713 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 1714 goto pim6_input_to_daemon; 1715 1716 /* 1717 * Validate length 1718 */ 1719 if (pimlen < PIM6_REG_MINLEN) { 1720 ++pim6stat.pim6s_rcv_tooshort; 1721 ++pim6stat.pim6s_rcv_badregisters; 1722 #ifdef MRT6DEBUG 1723 log(LOG_ERR, 1724 "pim6_input: register packet size too " 1725 "small %d from %s\n", 1726 pimlen, ip6_sprintf(&ip6->ip6_src)); 1727 #endif 1728 m_freem(m); 1729 return(IPPROTO_DONE); 1730 } 1731 1732 eip6 = (struct ip6_hdr *) (reghdr + 1); 1733 #ifdef MRT6DEBUG 1734 if (mrt6debug & DEBUG_PIM) 1735 log(LOG_DEBUG, 1736 "pim6_input[register], eip6: %s -> %s, " 1737 "eip6 plen %d\n", 1738 ip6_sprintf(&eip6->ip6_src), 1739 ip6_sprintf(&eip6->ip6_dst), 1740 ntohs(eip6->ip6_plen)); 1741 #endif 1742 1743 /* verify the version number of the inner packet */ 1744 if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1745 ++pim6stat.pim6s_rcv_badregisters; 1746 #ifdef MRT6DEBUG 1747 log(LOG_DEBUG, "pim6_input: invalid IP version (%d) " 1748 "of the inner packet\n", 1749 (eip6->ip6_vfc & IPV6_VERSION)); 1750 #endif 1751 m_freem(m); 1752 return(IPPROTO_NONE); 1753 } 1754 1755 /* verify the inner packet is destined to a mcast group */ 1756 if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) { 1757 ++pim6stat.pim6s_rcv_badregisters; 1758 #ifdef MRT6DEBUG 1759 if (mrt6debug & DEBUG_PIM) 1760 log(LOG_DEBUG, 1761 "pim6_input: inner packet of register " 1762 "is not multicast %s\n", 1763 ip6_sprintf(&eip6->ip6_dst)); 1764 #endif 1765 m_freem(m); 1766 return(IPPROTO_DONE); 1767 } 1768 1769 /* 1770 * make a copy of the whole header to pass to the daemon later. 1771 */ 1772 mcp = m_copy(m, 0, off + PIM6_REG_MINLEN); 1773 if (mcp == NULL) { 1774 #ifdef MRT6DEBUG 1775 log(LOG_ERR, 1776 "pim6_input: pim register: " 1777 "could not copy register head\n"); 1778 #endif 1779 m_freem(m); 1780 return(IPPROTO_DONE); 1781 } 1782 1783 /* 1784 * forward the inner ip6 packet; point m_data at the inner ip6. 1785 */ 1786 m_adj(m, off + PIM_MINLEN); 1787 #ifdef MRT6DEBUG 1788 if (mrt6debug & DEBUG_PIM) { 1789 log(LOG_DEBUG, 1790 "pim6_input: forwarding decapsulated register: " 1791 "src %s, dst %s, mif %d\n", 1792 ip6_sprintf(&eip6->ip6_src), 1793 ip6_sprintf(&eip6->ip6_dst), 1794 reg_mif_num); 1795 } 1796 #endif 1797 1798 rc = looutput(mif6table[reg_mif_num].m6_ifp, m, 1799 (struct sockaddr *) &dst, 1800 (struct rtentry *) NULL); 1801 1802 /* prepare the register head to send to the mrouting daemon */ 1803 m = mcp; 1804 } 1805 1806 /* 1807 * Pass the PIM message up to the daemon; if it is a register message 1808 * pass the 'head' only up to the daemon. This includes the 1809 * encapsulator ip6 header, pim header, register header and the 1810 * encapsulated ip6 header. 1811 */ 1812 pim6_input_to_daemon: 1813 rip6_input(&m, offp, proto); 1814 return(IPPROTO_DONE); 1815 } 1816