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 1.11 2006/01/14 11:44:25 swildner 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 #include <sys/thread2.h> 66 67 #include <net/if.h> 68 #include <net/route.h> 69 #include <net/raw_cb.h> 70 71 #include <netinet/in.h> 72 #include <netinet/in_var.h> 73 74 #include <netinet/ip6.h> 75 #include <netinet6/ip6_var.h> 76 #include <netinet6/ip6_mroute.h> 77 #include <netinet6/pim6.h> 78 #include <netinet6/pim6_var.h> 79 80 #include <net/net_osdep.h> 81 82 static MALLOC_DEFINE(M_MRTABLE, "mf6c", "multicast forwarding cache entry"); 83 84 #define M_HASCL(m) ((m)->m_flags & M_EXT) 85 86 static int ip6_mdq (struct mbuf *, struct ifnet *, struct mf6c *); 87 static void phyint_send (struct ip6_hdr *, struct mif6 *, struct mbuf *); 88 89 static int set_pim6 (int *); 90 static int socket_send (struct socket *, struct mbuf *, 91 struct sockaddr_in6 *); 92 static int register_send (struct ip6_hdr *, struct mif6 *, 93 struct mbuf *); 94 95 /* 96 * Globals. All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static, 97 * except for netstat or debugging purposes. 98 */ 99 struct socket *ip6_mrouter = NULL; 100 int ip6_mrouter_ver = 0; 101 int ip6_mrtproto = IPPROTO_PIM; /* for netstat only */ 102 struct mrt6stat mrt6stat; 103 104 #define NO_RTE_FOUND 0x1 105 #define RTE_FOUND 0x2 106 107 struct mf6c *mf6ctable[MF6CTBLSIZ]; 108 u_char n6expire[MF6CTBLSIZ]; 109 static struct mif6 mif6table[MAXMIFS]; 110 #ifdef MRT6DEBUG 111 u_int mrt6debug = 0; /* debug level */ 112 #define DEBUG_MFC 0x02 113 #define DEBUG_FORWARD 0x04 114 #define DEBUG_EXPIRE 0x08 115 #define DEBUG_XMIT 0x10 116 #define DEBUG_REG 0x20 117 #define DEBUG_PIM 0x40 118 #endif 119 120 static void expire_upcalls (void *); 121 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 122 #define UPCALL_EXPIRE 6 /* number of timeouts */ 123 124 #ifdef INET 125 #ifdef MROUTING 126 extern struct socket *ip_mrouter; 127 #endif 128 #endif 129 130 /* 131 * 'Interfaces' associated with decapsulator (so we can tell 132 * packets that went through it from ones that get reflected 133 * by a broken gateway). These interfaces are never linked into 134 * the system ifnet list & no routes point to them. I.e., packets 135 * can't be sent this way. They only exist as a placeholder for 136 * multicast source verification. 137 */ 138 struct ifnet multicast_register_if; 139 140 #define ENCAP_HOPS 64 141 142 /* 143 * Private variables. 144 */ 145 static mifi_t nummifs = 0; 146 static mifi_t reg_mif_num = (mifi_t)-1; 147 148 static struct pim6stat pim6stat; 149 static int pim6; 150 151 /* 152 * Hash function for a source, group entry 153 */ 154 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \ 155 (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \ 156 (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \ 157 (g).s6_addr32[2] ^ (g).s6_addr32[3]) 158 159 /* 160 * Find a route for a given origin IPv6 address and Multicast group address. 161 * Quality of service parameter to be added in the future!!! 162 */ 163 164 #define MF6CFIND(o, g, rt) do { \ 165 struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \ 166 rt = NULL; \ 167 mrt6stat.mrt6s_mfc_lookups++; \ 168 while (_rt) { \ 169 if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \ 170 IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \ 171 (_rt->mf6c_stall == NULL)) { \ 172 rt = _rt; \ 173 break; \ 174 } \ 175 _rt = _rt->mf6c_next; \ 176 } \ 177 if (rt == NULL) { \ 178 mrt6stat.mrt6s_mfc_misses++; \ 179 } \ 180 } while (0) 181 182 /* 183 * Macros to compute elapsed time efficiently 184 * Borrowed from Van Jacobson's scheduling code 185 */ 186 #define TV_DELTA(a, b, delta) do { \ 187 int xxs; \ 188 \ 189 delta = (a).tv_usec - (b).tv_usec; \ 190 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 191 switch (xxs) { \ 192 case 2: \ 193 delta += 1000000; \ 194 /* fall through */ \ 195 case 1: \ 196 delta += 1000000; \ 197 break; \ 198 default: \ 199 delta += (1000000 * xxs); \ 200 } \ 201 } \ 202 } while (0) 203 204 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 205 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 206 207 #ifdef UPCALL_TIMING 208 #define UPCALL_MAX 50 209 u_long upcall_data[UPCALL_MAX + 1]; 210 static void collate(); 211 #endif /* UPCALL_TIMING */ 212 213 static int get_sg_cnt (struct sioc_sg_req6 *); 214 static int get_mif6_cnt (struct sioc_mif_req6 *); 215 static int ip6_mrouter_init (struct socket *, struct mbuf *, int); 216 static int add_m6if (struct mif6ctl *); 217 static int del_m6if (mifi_t *); 218 static int add_m6fc (struct mf6cctl *); 219 static int del_m6fc (struct mf6cctl *); 220 221 static struct callout expire_upcalls_ch; 222 223 /* 224 * Handle MRT setsockopt commands to modify the multicast routing tables. 225 */ 226 int 227 ip6_mrouter_set(struct socket *so, struct sockopt *sopt) 228 { 229 int error = 0; 230 struct mbuf *m; 231 232 if (so != ip6_mrouter && sopt->sopt_name != MRT6_INIT) 233 return (EACCES); 234 235 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 236 return (error); 237 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 238 return (error); 239 240 switch (sopt->sopt_name) { 241 case MRT6_INIT: 242 #ifdef MRT6_OINIT 243 case MRT6_OINIT: 244 #endif 245 error = ip6_mrouter_init(so, m, sopt->sopt_name); 246 break; 247 case MRT6_DONE: 248 error = ip6_mrouter_done(); 249 break; 250 case MRT6_ADD_MIF: 251 error = add_m6if(mtod(m, struct mif6ctl *)); 252 break; 253 case MRT6_DEL_MIF: 254 error = del_m6if(mtod(m, mifi_t *)); 255 break; 256 case MRT6_ADD_MFC: 257 error = add_m6fc(mtod(m, struct mf6cctl *)); 258 break; 259 case MRT6_DEL_MFC: 260 error = del_m6fc(mtod(m, struct mf6cctl *)); 261 break; 262 case MRT6_PIM: 263 error = set_pim6(mtod(m, int *)); 264 break; 265 default: 266 error = EOPNOTSUPP; 267 break; 268 } 269 270 m_freem(m); 271 return(error); 272 } 273 274 /* 275 * Handle MRT getsockopt commands 276 */ 277 int 278 ip6_mrouter_get(struct socket *so, struct sockopt *sopt) 279 { 280 int error = 0; 281 282 if (so != ip6_mrouter) return EACCES; 283 284 switch (sopt->sopt_name) { 285 case MRT6_PIM: 286 error = sooptcopyout(sopt, &pim6, sizeof(pim6)); 287 break; 288 } 289 return (error); 290 } 291 292 /* 293 * Handle ioctl commands to obtain information from the cache 294 */ 295 int 296 mrt6_ioctl(int cmd, caddr_t data) 297 { 298 int error = 0; 299 300 switch (cmd) { 301 case SIOCGETSGCNT_IN6: 302 return(get_sg_cnt((struct sioc_sg_req6 *)data)); 303 break; /* for safety */ 304 case SIOCGETMIFCNT_IN6: 305 return(get_mif6_cnt((struct sioc_mif_req6 *)data)); 306 break; /* for safety */ 307 default: 308 return (EINVAL); 309 break; 310 } 311 return error; 312 } 313 314 /* 315 * returns the packet, byte, rpf-failure count for the source group provided 316 */ 317 static int 318 get_sg_cnt(struct sioc_sg_req6 *req) 319 { 320 struct mf6c *rt; 321 322 crit_enter(); 323 MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt); 324 crit_exit(); 325 if (rt != NULL) { 326 req->pktcnt = rt->mf6c_pkt_cnt; 327 req->bytecnt = rt->mf6c_byte_cnt; 328 req->wrong_if = rt->mf6c_wrong_if; 329 } else 330 return(ESRCH); 331 #if 0 332 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 333 #endif 334 335 return 0; 336 } 337 338 /* 339 * returns the input and output packet and byte counts on the mif provided 340 */ 341 static int 342 get_mif6_cnt(struct sioc_mif_req6 *req) 343 { 344 mifi_t mifi = req->mifi; 345 346 if (mifi >= nummifs) 347 return EINVAL; 348 349 req->icount = mif6table[mifi].m6_pkt_in; 350 req->ocount = mif6table[mifi].m6_pkt_out; 351 req->ibytes = mif6table[mifi].m6_bytes_in; 352 req->obytes = mif6table[mifi].m6_bytes_out; 353 354 return 0; 355 } 356 357 static int 358 set_pim6(int *i) 359 { 360 if ((*i != 1) && (*i != 0)) 361 return EINVAL; 362 363 pim6 = *i; 364 365 return 0; 366 } 367 368 /* 369 * Enable multicast routing 370 */ 371 static int 372 ip6_mrouter_init(struct socket *so, struct mbuf *m, int cmd) 373 { 374 int *v; 375 376 #ifdef MRT6DEBUG 377 if (mrt6debug) 378 log(LOG_DEBUG, 379 "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n", 380 so->so_type, so->so_proto->pr_protocol); 381 #endif 382 383 if (so->so_type != SOCK_RAW || 384 so->so_proto->pr_protocol != IPPROTO_ICMPV6) 385 return EOPNOTSUPP; 386 387 if (!m || (m->m_len != sizeof(int *))) 388 return ENOPROTOOPT; 389 390 v = mtod(m, int *); 391 if (*v != 1) 392 return ENOPROTOOPT; 393 394 if (ip6_mrouter != NULL) return EADDRINUSE; 395 396 ip6_mrouter = so; 397 ip6_mrouter_ver = cmd; 398 399 bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); 400 bzero((caddr_t)n6expire, sizeof(n6expire)); 401 402 pim6 = 0;/* used for stubbing out/in pim stuff */ 403 404 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 405 expire_upcalls, NULL); 406 407 #ifdef MRT6DEBUG 408 if (mrt6debug) 409 log(LOG_DEBUG, "ip6_mrouter_init\n"); 410 #endif 411 412 return 0; 413 } 414 415 /* 416 * Disable multicast routing 417 */ 418 int 419 ip6_mrouter_done(void) 420 { 421 mifi_t mifi; 422 int i; 423 struct ifnet *ifp; 424 struct in6_ifreq ifr; 425 struct mf6c *rt; 426 struct rtdetq *rte; 427 428 /* 429 * For each phyint in use, disable promiscuous reception of all IPv6 430 * multicasts. 431 */ 432 #ifdef INET 433 #ifdef MROUTING 434 /* 435 * If there is still IPv4 multicast routing daemon, 436 * we remain interfaces to receive all muliticasted packets. 437 * XXX: there may be an interface in which the IPv4 multicast 438 * daemon is not interested... 439 */ 440 if (!ip_mrouter) 441 #endif 442 #endif 443 { 444 for (mifi = 0; mifi < nummifs; mifi++) { 445 if (mif6table[mifi].m6_ifp && 446 !(mif6table[mifi].m6_flags & MIFF_REGISTER)) { 447 ifr.ifr_addr.sin6_family = AF_INET6; 448 ifr.ifr_addr.sin6_addr= in6addr_any; 449 ifp = mif6table[mifi].m6_ifp; 450 lwkt_serialize_enter(ifp->if_serializer); 451 ifp->if_ioctl(ifp, SIOCDELMULTI, 452 (caddr_t)&ifr, NULL); 453 lwkt_serialize_exit(ifp->if_serializer); 454 } 455 } 456 } 457 #ifdef notyet 458 bzero((caddr_t)qtable, sizeof(qtable)); 459 bzero((caddr_t)tbftable, sizeof(tbftable)); 460 #endif 461 bzero((caddr_t)mif6table, sizeof(mif6table)); 462 nummifs = 0; 463 464 pim6 = 0; /* used to stub out/in pim specific code */ 465 466 callout_stop(&expire_upcalls_ch); 467 468 /* 469 * Free all multicast forwarding cache entries. 470 */ 471 for (i = 0; i < MF6CTBLSIZ; i++) { 472 rt = mf6ctable[i]; 473 while (rt) { 474 struct mf6c *frt; 475 476 for (rte = rt->mf6c_stall; rte != NULL; ) { 477 struct rtdetq *n = rte->next; 478 479 m_free(rte->m); 480 free(rte, M_MRTABLE); 481 rte = n; 482 } 483 frt = rt; 484 rt = rt->mf6c_next; 485 free(frt, M_MRTABLE); 486 } 487 } 488 489 bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); 490 491 /* 492 * Reset de-encapsulation cache 493 */ 494 reg_mif_num = -1; 495 496 ip6_mrouter = NULL; 497 ip6_mrouter_ver = 0; 498 499 #ifdef MRT6DEBUG 500 if (mrt6debug) 501 log(LOG_DEBUG, "ip6_mrouter_done\n"); 502 #endif 503 504 return 0; 505 } 506 507 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; 508 509 /* 510 * Add a mif to the mif table 511 */ 512 static int 513 add_m6if(struct mif6ctl *mifcp) 514 { 515 struct mif6 *mifp; 516 struct ifnet *ifp; 517 int error; 518 #ifdef notyet 519 struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi; 520 #endif 521 522 if (mifcp->mif6c_mifi >= MAXMIFS) 523 return EINVAL; 524 mifp = mif6table + mifcp->mif6c_mifi; 525 if (mifp->m6_ifp) 526 return EADDRINUSE; /* XXX: is it appropriate? */ 527 if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index) 528 return ENXIO; 529 ifp = ifindex2ifnet[mifcp->mif6c_pifi]; 530 531 if (mifcp->mif6c_flags & MIFF_REGISTER) { 532 if (reg_mif_num == (mifi_t)-1) { 533 strlcpy(multicast_register_if.if_xname, "register_mif", 534 IFNAMSIZ); 535 multicast_register_if.if_flags |= IFF_LOOPBACK; 536 multicast_register_if.if_index = mifcp->mif6c_mifi; 537 reg_mif_num = mifcp->mif6c_mifi; 538 } 539 540 ifp = &multicast_register_if; 541 542 } /* if REGISTER */ 543 else { 544 /* Make sure the interface supports multicast */ 545 if ((ifp->if_flags & IFF_MULTICAST) == 0) 546 return EOPNOTSUPP; 547 548 crit_enter(); 549 error = if_allmulti(ifp, 1); 550 crit_exit(); 551 if (error) 552 return error; 553 } 554 555 crit_enter(); 556 mifp->m6_flags = mifcp->mif6c_flags; 557 mifp->m6_ifp = ifp; 558 #ifdef notyet 559 /* scaling up here allows division by 1024 in critical code */ 560 mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000; 561 #endif 562 /* initialize per mif pkt counters */ 563 mifp->m6_pkt_in = 0; 564 mifp->m6_pkt_out = 0; 565 mifp->m6_bytes_in = 0; 566 mifp->m6_bytes_out = 0; 567 crit_exit(); 568 569 /* Adjust nummifs up if the mifi is higher than nummifs */ 570 if (nummifs <= mifcp->mif6c_mifi) 571 nummifs = mifcp->mif6c_mifi + 1; 572 573 #ifdef MRT6DEBUG 574 if (mrt6debug) 575 log(LOG_DEBUG, 576 "add_mif #%d, phyint %s\n", 577 mifcp->mif6c_mifi, 578 ifp->if_xname); 579 #endif 580 581 return 0; 582 } 583 584 /* 585 * Delete a mif from the mif table 586 */ 587 static int 588 del_m6if(mifi_t *mifip) 589 { 590 struct mif6 *mifp = mif6table + *mifip; 591 mifi_t mifi; 592 struct ifnet *ifp; 593 594 if (*mifip >= nummifs) 595 return EINVAL; 596 if (mifp->m6_ifp == NULL) 597 return EINVAL; 598 599 crit_enter(); 600 601 if (!(mifp->m6_flags & MIFF_REGISTER)) { 602 /* 603 * XXX: what if there is yet IPv4 multicast daemon 604 * using the interface? 605 */ 606 ifp = mifp->m6_ifp; 607 608 if_allmulti(ifp, 0); 609 } 610 611 #ifdef notyet 612 bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip])); 613 bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf))); 614 #endif 615 bzero((caddr_t)mifp, sizeof (*mifp)); 616 617 /* Adjust nummifs down */ 618 for (mifi = nummifs; mifi > 0; mifi--) 619 if (mif6table[mifi - 1].m6_ifp) 620 break; 621 nummifs = mifi; 622 623 crit_exit(); 624 625 #ifdef MRT6DEBUG 626 if (mrt6debug) 627 log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs); 628 #endif 629 630 return 0; 631 } 632 633 /* 634 * Add an mfc entry 635 */ 636 static int 637 add_m6fc(struct mf6cctl *mfccp) 638 { 639 struct mf6c *rt; 640 u_long hash; 641 struct rtdetq *rte; 642 u_short nstl; 643 644 MF6CFIND(mfccp->mf6cc_origin.sin6_addr, 645 mfccp->mf6cc_mcastgrp.sin6_addr, rt); 646 647 /* If an entry already exists, just update the fields */ 648 if (rt) { 649 #ifdef MRT6DEBUG 650 if (mrt6debug & DEBUG_MFC) 651 log(LOG_DEBUG, 652 "add_m6fc no upcall h %d o %s g %s p %x\n", 653 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 654 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 655 mfccp->mf6cc_parent); 656 #endif 657 658 crit_enter(); 659 rt->mf6c_parent = mfccp->mf6cc_parent; 660 rt->mf6c_ifset = mfccp->mf6cc_ifset; 661 crit_exit(); 662 return 0; 663 } 664 665 /* 666 * Find the entry for which the upcall was made and update 667 */ 668 crit_enter(); 669 hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr, 670 mfccp->mf6cc_mcastgrp.sin6_addr); 671 for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) { 672 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, 673 &mfccp->mf6cc_origin.sin6_addr) && 674 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, 675 &mfccp->mf6cc_mcastgrp.sin6_addr) && 676 (rt->mf6c_stall != NULL)) { 677 678 if (nstl++) 679 log(LOG_ERR, 680 "add_m6fc: %s o %s g %s p %x dbx %p\n", 681 "multiple kernel entries", 682 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 683 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 684 mfccp->mf6cc_parent, rt->mf6c_stall); 685 686 #ifdef MRT6DEBUG 687 if (mrt6debug & DEBUG_MFC) 688 log(LOG_DEBUG, 689 "add_m6fc o %s g %s p %x dbg %x\n", 690 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 691 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 692 mfccp->mf6cc_parent, rt->mf6c_stall); 693 #endif 694 695 rt->mf6c_origin = mfccp->mf6cc_origin; 696 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 697 rt->mf6c_parent = mfccp->mf6cc_parent; 698 rt->mf6c_ifset = mfccp->mf6cc_ifset; 699 /* initialize pkt counters per src-grp */ 700 rt->mf6c_pkt_cnt = 0; 701 rt->mf6c_byte_cnt = 0; 702 rt->mf6c_wrong_if = 0; 703 704 rt->mf6c_expire = 0; /* Don't clean this guy up */ 705 n6expire[hash]--; 706 707 /* free packets Qed at the end of this entry */ 708 for (rte = rt->mf6c_stall; rte != NULL; ) { 709 struct rtdetq *n = rte->next; 710 ip6_mdq(rte->m, rte->ifp, rt); 711 m_freem(rte->m); 712 #ifdef UPCALL_TIMING 713 collate(&(rte->t)); 714 #endif /* UPCALL_TIMING */ 715 free(rte, M_MRTABLE); 716 rte = n; 717 } 718 rt->mf6c_stall = NULL; 719 } 720 } 721 722 /* 723 * It is possible that an entry is being inserted without an upcall 724 */ 725 if (nstl == 0) { 726 #ifdef MRT6DEBUG 727 if (mrt6debug & DEBUG_MFC) 728 log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n", 729 hash, 730 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 731 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 732 mfccp->mf6cc_parent); 733 #endif 734 735 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { 736 737 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, 738 &mfccp->mf6cc_origin.sin6_addr)&& 739 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, 740 &mfccp->mf6cc_mcastgrp.sin6_addr)) { 741 742 rt->mf6c_origin = mfccp->mf6cc_origin; 743 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 744 rt->mf6c_parent = mfccp->mf6cc_parent; 745 rt->mf6c_ifset = mfccp->mf6cc_ifset; 746 /* initialize pkt counters per src-grp */ 747 rt->mf6c_pkt_cnt = 0; 748 rt->mf6c_byte_cnt = 0; 749 rt->mf6c_wrong_if = 0; 750 751 if (rt->mf6c_expire) 752 n6expire[hash]--; 753 rt->mf6c_expire = 0; 754 } 755 } 756 if (rt == NULL) { 757 /* no upcall, so make a new entry */ 758 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE, 759 M_NOWAIT); 760 if (rt == NULL) { 761 crit_exit(); 762 return ENOBUFS; 763 } 764 765 /* insert new entry at head of hash chain */ 766 rt->mf6c_origin = mfccp->mf6cc_origin; 767 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 768 rt->mf6c_parent = mfccp->mf6cc_parent; 769 rt->mf6c_ifset = mfccp->mf6cc_ifset; 770 /* initialize pkt counters per src-grp */ 771 rt->mf6c_pkt_cnt = 0; 772 rt->mf6c_byte_cnt = 0; 773 rt->mf6c_wrong_if = 0; 774 rt->mf6c_expire = 0; 775 rt->mf6c_stall = NULL; 776 777 /* link into table */ 778 rt->mf6c_next = mf6ctable[hash]; 779 mf6ctable[hash] = rt; 780 } 781 } 782 crit_exit(); 783 return 0; 784 } 785 786 #ifdef UPCALL_TIMING 787 /* 788 * collect delay statistics on the upcalls 789 */ 790 static void 791 collate(struct timeval *t) 792 { 793 u_long d; 794 struct timeval tp; 795 u_long delta; 796 797 GET_TIME(tp); 798 799 if (TV_LT(*t, tp)) 800 { 801 TV_DELTA(tp, *t, delta); 802 803 d = delta >> 10; 804 if (d > UPCALL_MAX) 805 d = UPCALL_MAX; 806 807 ++upcall_data[d]; 808 } 809 } 810 #endif /* UPCALL_TIMING */ 811 812 /* 813 * Delete an mfc entry 814 */ 815 static int 816 del_m6fc(struct mf6cctl *mfccp) 817 { 818 struct sockaddr_in6 origin; 819 struct sockaddr_in6 mcastgrp; 820 struct mf6c *rt; 821 struct mf6c **nptr; 822 u_long hash; 823 824 origin = mfccp->mf6cc_origin; 825 mcastgrp = mfccp->mf6cc_mcastgrp; 826 hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr); 827 828 #ifdef MRT6DEBUG 829 if (mrt6debug & DEBUG_MFC) 830 log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n", 831 ip6_sprintf(&origin.sin6_addr), 832 ip6_sprintf(&mcastgrp.sin6_addr)); 833 #endif 834 835 crit_enter(); 836 837 nptr = &mf6ctable[hash]; 838 while ((rt = *nptr) != NULL) { 839 if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr, 840 &rt->mf6c_origin.sin6_addr) && 841 IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr, 842 &rt->mf6c_mcastgrp.sin6_addr) && 843 rt->mf6c_stall == NULL) 844 break; 845 846 nptr = &rt->mf6c_next; 847 } 848 if (rt == NULL) { 849 crit_exit(); 850 return EADDRNOTAVAIL; 851 } 852 853 *nptr = rt->mf6c_next; 854 free(rt, M_MRTABLE); 855 856 crit_exit(); 857 858 return 0; 859 } 860 861 static int 862 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src) 863 { 864 if (s) { 865 if (sbappendaddr(&s->so_rcv, 866 (struct sockaddr *)src, 867 mm, (struct mbuf *)0) != 0) { 868 sorwakeup(s); 869 return 0; 870 } 871 } 872 m_freem(mm); 873 return -1; 874 } 875 876 /* 877 * IPv6 multicast forwarding function. This function assumes that the packet 878 * pointed to by "ip6" has arrived on (or is about to be sent to) the interface 879 * pointed to by "ifp", and the packet is to be relayed to other networks 880 * that have members of the packet's destination IPv6 multicast group. 881 * 882 * The packet is returned unscathed to the caller, unless it is 883 * erroneous, in which case a non-zero return value tells the caller to 884 * discard it. 885 */ 886 887 int 888 ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m) 889 { 890 struct mf6c *rt; 891 struct mif6 *mifp; 892 struct mbuf *mm; 893 mifi_t mifi; 894 895 #ifdef MRT6DEBUG 896 if (mrt6debug & DEBUG_FORWARD) 897 log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n", 898 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst), 899 ifp->if_index); 900 #endif 901 902 /* 903 * Don't forward a packet with Hop limit of zero or one, 904 * or a packet destined to a local-only group. 905 */ 906 if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) || 907 IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst)) 908 return 0; 909 ip6->ip6_hlim--; 910 911 /* 912 * Source address check: do not forward packets with unspecified 913 * source. It was discussed in July 2000, on ipngwg mailing list. 914 * This is rather more serious than unicast cases, because some 915 * MLD packets can be sent with the unspecified source address 916 * (although such packets must normally set 1 to the hop limit field). 917 */ 918 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 919 ip6stat.ip6s_cantforward++; 920 if (ip6_log_time + ip6_log_interval < time_second) { 921 ip6_log_time = time_second; 922 log(LOG_DEBUG, 923 "cannot forward " 924 "from %s to %s nxt %d received on %s\n", 925 ip6_sprintf(&ip6->ip6_src), 926 ip6_sprintf(&ip6->ip6_dst), 927 ip6->ip6_nxt, 928 if_name(m->m_pkthdr.rcvif)); 929 } 930 return 0; 931 } 932 933 /* 934 * Determine forwarding mifs from the forwarding cache table 935 */ 936 crit_enter(); 937 MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt); 938 939 /* Entry exists, so forward if necessary */ 940 if (rt) { 941 crit_exit(); 942 return (ip6_mdq(m, ifp, rt)); 943 } else { 944 /* 945 * If we don't have a route for packet's origin, 946 * Make a copy of the packet & 947 * send message to routing daemon 948 */ 949 950 struct mbuf *mb0; 951 struct rtdetq *rte; 952 u_long hash; 953 /* int i, npkts;*/ 954 #ifdef UPCALL_TIMING 955 struct timeval tp; 956 957 GET_TIME(tp); 958 #endif /* UPCALL_TIMING */ 959 960 mrt6stat.mrt6s_no_route++; 961 #ifdef MRT6DEBUG 962 if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC)) 963 log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n", 964 ip6_sprintf(&ip6->ip6_src), 965 ip6_sprintf(&ip6->ip6_dst)); 966 #endif 967 968 /* 969 * Allocate mbufs early so that we don't do extra work if we 970 * are just going to fail anyway. 971 */ 972 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, 973 M_NOWAIT); 974 if (rte == NULL) { 975 crit_exit(); 976 return ENOBUFS; 977 } 978 mb0 = m_copy(m, 0, M_COPYALL); 979 /* 980 * Pullup packet header if needed before storing it, 981 * as other references may modify it in the meantime. 982 */ 983 if (mb0 && 984 (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr))) 985 mb0 = m_pullup(mb0, sizeof(struct ip6_hdr)); 986 if (mb0 == NULL) { 987 free(rte, M_MRTABLE); 988 crit_exit(); 989 return ENOBUFS; 990 } 991 992 /* is there an upcall waiting for this packet? */ 993 hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst); 994 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { 995 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, 996 &rt->mf6c_origin.sin6_addr) && 997 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 998 &rt->mf6c_mcastgrp.sin6_addr) && 999 (rt->mf6c_stall != NULL)) 1000 break; 1001 } 1002 1003 if (rt == NULL) { 1004 struct mrt6msg *im; 1005 #ifdef MRT6_OINIT 1006 struct omrt6msg *oim; 1007 #endif 1008 1009 /* no upcall, so make a new entry */ 1010 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE, 1011 M_NOWAIT); 1012 if (rt == NULL) { 1013 free(rte, M_MRTABLE); 1014 m_freem(mb0); 1015 crit_exit(); 1016 return ENOBUFS; 1017 } 1018 /* 1019 * Make a copy of the header to send to the user 1020 * level process 1021 */ 1022 mm = m_copy(mb0, 0, sizeof(struct ip6_hdr)); 1023 1024 if (mm == NULL) { 1025 free(rte, M_MRTABLE); 1026 m_freem(mb0); 1027 free(rt, M_MRTABLE); 1028 crit_exit(); 1029 return ENOBUFS; 1030 } 1031 1032 /* 1033 * Send message to routing daemon 1034 */ 1035 sin6.sin6_addr = ip6->ip6_src; 1036 1037 im = NULL; 1038 #ifdef MRT6_OINIT 1039 oim = NULL; 1040 #endif 1041 switch (ip6_mrouter_ver) { 1042 #ifdef MRT6_OINIT 1043 case MRT6_OINIT: 1044 oim = mtod(mm, struct omrt6msg *); 1045 oim->im6_msgtype = MRT6MSG_NOCACHE; 1046 oim->im6_mbz = 0; 1047 break; 1048 #endif 1049 case MRT6_INIT: 1050 im = mtod(mm, struct mrt6msg *); 1051 im->im6_msgtype = MRT6MSG_NOCACHE; 1052 im->im6_mbz = 0; 1053 break; 1054 default: 1055 free(rte, M_MRTABLE); 1056 m_freem(mb0); 1057 free(rt, M_MRTABLE); 1058 crit_exit(); 1059 return EINVAL; 1060 } 1061 1062 #ifdef MRT6DEBUG 1063 if (mrt6debug & DEBUG_FORWARD) 1064 log(LOG_DEBUG, 1065 "getting the iif info in the kernel\n"); 1066 #endif 1067 1068 for (mifp = mif6table, mifi = 0; 1069 mifi < nummifs && mifp->m6_ifp != ifp; 1070 mifp++, mifi++) 1071 ; 1072 1073 switch (ip6_mrouter_ver) { 1074 #ifdef MRT6_OINIT 1075 case MRT6_OINIT: 1076 oim->im6_mif = mifi; 1077 break; 1078 #endif 1079 case MRT6_INIT: 1080 im->im6_mif = mifi; 1081 break; 1082 } 1083 1084 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1085 log(LOG_WARNING, "ip6_mforward: ip6_mrouter " 1086 "socket queue full\n"); 1087 mrt6stat.mrt6s_upq_sockfull++; 1088 free(rte, M_MRTABLE); 1089 m_freem(mb0); 1090 free(rt, M_MRTABLE); 1091 crit_exit(); 1092 return ENOBUFS; 1093 } 1094 1095 mrt6stat.mrt6s_upcalls++; 1096 1097 /* insert new entry at head of hash chain */ 1098 bzero(rt, sizeof(*rt)); 1099 rt->mf6c_origin.sin6_family = AF_INET6; 1100 rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6); 1101 rt->mf6c_origin.sin6_addr = ip6->ip6_src; 1102 rt->mf6c_mcastgrp.sin6_family = AF_INET6; 1103 rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6); 1104 rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst; 1105 rt->mf6c_expire = UPCALL_EXPIRE; 1106 n6expire[hash]++; 1107 rt->mf6c_parent = MF6C_INCOMPLETE_PARENT; 1108 1109 /* link into table */ 1110 rt->mf6c_next = mf6ctable[hash]; 1111 mf6ctable[hash] = rt; 1112 /* Add this entry to the end of the queue */ 1113 rt->mf6c_stall = rte; 1114 } else { 1115 /* determine if q has overflowed */ 1116 struct rtdetq **p; 1117 int npkts = 0; 1118 1119 for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next) 1120 if (++npkts > MAX_UPQ6) { 1121 mrt6stat.mrt6s_upq_ovflw++; 1122 free(rte, M_MRTABLE); 1123 m_freem(mb0); 1124 crit_exit(); 1125 return 0; 1126 } 1127 1128 /* Add this entry to the end of the queue */ 1129 *p = rte; 1130 } 1131 1132 rte->next = NULL; 1133 rte->m = mb0; 1134 rte->ifp = ifp; 1135 #ifdef UPCALL_TIMING 1136 rte->t = tp; 1137 #endif /* UPCALL_TIMING */ 1138 1139 crit_exit(); 1140 1141 return 0; 1142 } 1143 } 1144 1145 /* 1146 * Clean up cache entries if upcalls are not serviced 1147 * Call from the Slow Timeout mechanism, every half second. 1148 */ 1149 static void 1150 expire_upcalls(void *unused) 1151 { 1152 struct rtdetq *rte; 1153 struct mf6c *mfc, **nptr; 1154 int i; 1155 1156 crit_enter(); 1157 for (i = 0; i < MF6CTBLSIZ; i++) { 1158 if (n6expire[i] == 0) 1159 continue; 1160 nptr = &mf6ctable[i]; 1161 while ((mfc = *nptr) != NULL) { 1162 rte = mfc->mf6c_stall; 1163 /* 1164 * Skip real cache entries 1165 * Make sure it wasn't marked to not expire (shouldn't happen) 1166 * If it expires now 1167 */ 1168 if (rte != NULL && 1169 mfc->mf6c_expire != 0 && 1170 --mfc->mf6c_expire == 0) { 1171 #ifdef MRT6DEBUG 1172 if (mrt6debug & DEBUG_EXPIRE) 1173 log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n", 1174 ip6_sprintf(&mfc->mf6c_origin.sin6_addr), 1175 ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr)); 1176 #endif 1177 /* 1178 * drop all the packets 1179 * free the mbuf with the pkt, if, timing info 1180 */ 1181 do { 1182 struct rtdetq *n = rte->next; 1183 m_freem(rte->m); 1184 free(rte, M_MRTABLE); 1185 rte = n; 1186 } while (rte != NULL); 1187 mrt6stat.mrt6s_cache_cleanups++; 1188 n6expire[i]--; 1189 1190 *nptr = mfc->mf6c_next; 1191 free(mfc, M_MRTABLE); 1192 } else { 1193 nptr = &mfc->mf6c_next; 1194 } 1195 } 1196 } 1197 crit_exit(); 1198 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1199 expire_upcalls, NULL); 1200 } 1201 1202 /* 1203 * Packet forwarding routine once entry in the cache is made 1204 */ 1205 static int 1206 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt) 1207 { 1208 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1209 mifi_t mifi, iif; 1210 struct mif6 *mifp; 1211 int plen = m->m_pkthdr.len; 1212 1213 /* 1214 * Macro to send packet on mif. Since RSVP packets don't get counted on 1215 * input, they shouldn't get counted on output, so statistics keeping is 1216 * separate. 1217 */ 1218 1219 #define MC6_SEND(ip6, mifp, m) do { \ 1220 if ((mifp)->m6_flags & MIFF_REGISTER) \ 1221 register_send((ip6), (mifp), (m)); \ 1222 else \ 1223 phyint_send((ip6), (mifp), (m)); \ 1224 } while (0) 1225 1226 /* 1227 * Don't forward if it didn't arrive from the parent mif 1228 * for its origin. 1229 */ 1230 mifi = rt->mf6c_parent; 1231 if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) { 1232 /* came in the wrong interface */ 1233 #ifdef MRT6DEBUG 1234 if (mrt6debug & DEBUG_FORWARD) 1235 log(LOG_DEBUG, 1236 "wrong if: ifid %d mifi %d mififid %x\n", 1237 ifp->if_index, mifi, 1238 mif6table[mifi].m6_ifp->if_index); 1239 #endif 1240 mrt6stat.mrt6s_wrong_if++; 1241 rt->mf6c_wrong_if++; 1242 /* 1243 * If we are doing PIM processing, and we are forwarding 1244 * packets on this interface, send a message to the 1245 * routing daemon. 1246 */ 1247 /* have to make sure this is a valid mif */ 1248 if (mifi < nummifs && mif6table[mifi].m6_ifp) 1249 if (pim6 && (m->m_flags & M_LOOP) == 0) { 1250 /* 1251 * Check the M_LOOP flag to avoid an 1252 * unnecessary PIM assert. 1253 * XXX: M_LOOP is an ad-hoc hack... 1254 */ 1255 static struct sockaddr_in6 sin6 = 1256 { sizeof(sin6), AF_INET6 }; 1257 1258 struct mbuf *mm; 1259 struct mrt6msg *im; 1260 #ifdef MRT6_OINIT 1261 struct omrt6msg *oim; 1262 #endif 1263 1264 mm = m_copy(m, 0, sizeof(struct ip6_hdr)); 1265 if (mm && 1266 (M_HASCL(mm) || 1267 mm->m_len < sizeof(struct ip6_hdr))) 1268 mm = m_pullup(mm, sizeof(struct ip6_hdr)); 1269 if (mm == NULL) 1270 return ENOBUFS; 1271 1272 #ifdef MRT6_OINIT 1273 oim = NULL; 1274 #endif 1275 im = NULL; 1276 switch (ip6_mrouter_ver) { 1277 #ifdef MRT6_OINIT 1278 case MRT6_OINIT: 1279 oim = mtod(mm, struct omrt6msg *); 1280 oim->im6_msgtype = MRT6MSG_WRONGMIF; 1281 oim->im6_mbz = 0; 1282 break; 1283 #endif 1284 case MRT6_INIT: 1285 im = mtod(mm, struct mrt6msg *); 1286 im->im6_msgtype = MRT6MSG_WRONGMIF; 1287 im->im6_mbz = 0; 1288 break; 1289 default: 1290 m_freem(mm); 1291 return EINVAL; 1292 } 1293 1294 for (mifp = mif6table, iif = 0; 1295 iif < nummifs && mifp && 1296 mifp->m6_ifp != ifp; 1297 mifp++, iif++) 1298 ; 1299 1300 switch (ip6_mrouter_ver) { 1301 #ifdef MRT6_OINIT 1302 case MRT6_OINIT: 1303 oim->im6_mif = iif; 1304 sin6.sin6_addr = oim->im6_src; 1305 break; 1306 #endif 1307 case MRT6_INIT: 1308 im->im6_mif = iif; 1309 sin6.sin6_addr = im->im6_src; 1310 break; 1311 } 1312 1313 mrt6stat.mrt6s_upcalls++; 1314 1315 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1316 #ifdef MRT6DEBUG 1317 if (mrt6debug) 1318 log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n"); 1319 #endif 1320 ++mrt6stat.mrt6s_upq_sockfull; 1321 return ENOBUFS; 1322 } /* if socket Q full */ 1323 } /* if PIM */ 1324 return 0; 1325 } /* if wrong iif */ 1326 1327 /* If I sourced this packet, it counts as output, else it was input. */ 1328 if (m->m_pkthdr.rcvif == NULL) { 1329 /* XXX: is rcvif really NULL when output?? */ 1330 mif6table[mifi].m6_pkt_out++; 1331 mif6table[mifi].m6_bytes_out += plen; 1332 } else { 1333 mif6table[mifi].m6_pkt_in++; 1334 mif6table[mifi].m6_bytes_in += plen; 1335 } 1336 rt->mf6c_pkt_cnt++; 1337 rt->mf6c_byte_cnt += plen; 1338 1339 /* 1340 * For each mif, forward a copy of the packet if there are group 1341 * members downstream on the interface. 1342 */ 1343 for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) 1344 if (IF_ISSET(mifi, &rt->mf6c_ifset)) { 1345 /* 1346 * check if the outgoing packet is going to break 1347 * a scope boundary. 1348 * XXX For packets through PIM register tunnel 1349 * interface, we believe a routing daemon. 1350 */ 1351 if ((mif6table[rt->mf6c_parent].m6_flags & 1352 MIFF_REGISTER) == 0 && 1353 (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 && 1354 (in6_addr2scopeid(ifp, &ip6->ip6_dst) != 1355 in6_addr2scopeid(mif6table[mifi].m6_ifp, 1356 &ip6->ip6_dst) || 1357 in6_addr2scopeid(ifp, &ip6->ip6_src) != 1358 in6_addr2scopeid(mif6table[mifi].m6_ifp, 1359 &ip6->ip6_src))) { 1360 ip6stat.ip6s_badscope++; 1361 continue; 1362 } 1363 1364 mifp->m6_pkt_out++; 1365 mifp->m6_bytes_out += plen; 1366 MC6_SEND(ip6, mifp, m); 1367 } 1368 return 0; 1369 } 1370 1371 static void 1372 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m) 1373 { 1374 struct mbuf *mb_copy; 1375 struct ifnet *ifp = mifp->m6_ifp; 1376 int error = 0; 1377 static struct route_in6 ro; 1378 struct in6_multi *in6m; 1379 struct sockaddr_in6 *dst6; 1380 1381 crit_enter(); /* needs to protect static "ro" below. */ 1382 1383 /* 1384 * Make a new reference to the packet; make sure that 1385 * the IPv6 header is actually copied, not just referenced, 1386 * so that ip6_output() only scribbles on the copy. 1387 */ 1388 mb_copy = m_copy(m, 0, M_COPYALL); 1389 if (mb_copy && 1390 (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr))) 1391 mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr)); 1392 if (mb_copy == NULL) { 1393 crit_exit(); 1394 return; 1395 } 1396 /* set MCAST flag to the outgoing packet */ 1397 mb_copy->m_flags |= M_MCAST; 1398 1399 /* 1400 * If we sourced the packet, call ip6_output since we may devide 1401 * the packet into fragments when the packet is too big for the 1402 * outgoing interface. 1403 * Otherwise, we can simply send the packet to the interface 1404 * sending queue. 1405 */ 1406 if (m->m_pkthdr.rcvif == NULL) { 1407 struct ip6_moptions im6o; 1408 1409 im6o.im6o_multicast_ifp = ifp; 1410 /* XXX: ip6_output will override ip6->ip6_hlim */ 1411 im6o.im6o_multicast_hlim = ip6->ip6_hlim; 1412 im6o.im6o_multicast_loop = 1; 1413 error = ip6_output(mb_copy, NULL, &ro, 1414 IPV6_FORWARDING, &im6o, NULL, NULL); 1415 1416 #ifdef MRT6DEBUG 1417 if (mrt6debug & DEBUG_XMIT) 1418 log(LOG_DEBUG, "phyint_send on mif %d err %d\n", 1419 mifp - mif6table, error); 1420 #endif 1421 crit_exit(); 1422 return; 1423 } 1424 1425 /* 1426 * If we belong to the destination multicast group 1427 * on the outgoing interface, loop back a copy. 1428 */ 1429 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 1430 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); 1431 if (in6m != NULL) { 1432 dst6->sin6_len = sizeof(struct sockaddr_in6); 1433 dst6->sin6_family = AF_INET6; 1434 dst6->sin6_addr = ip6->ip6_dst; 1435 ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst); 1436 } 1437 /* 1438 * Put the packet into the sending queue of the outgoing interface 1439 * if it would fit in the MTU of the interface. 1440 */ 1441 if (mb_copy->m_pkthdr.len <= ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) { 1442 dst6->sin6_len = sizeof(struct sockaddr_in6); 1443 dst6->sin6_family = AF_INET6; 1444 dst6->sin6_addr = ip6->ip6_dst; 1445 /* 1446 * We just call if_output instead of nd6_output here, since 1447 * we need no ND for a multicast forwarded packet...right? 1448 */ 1449 lwkt_serialize_enter(ifp->if_serializer); 1450 error = (*ifp->if_output)(ifp, mb_copy, 1451 (struct sockaddr *)&ro.ro_dst, NULL); 1452 lwkt_serialize_exit(ifp->if_serializer); 1453 #ifdef MRT6DEBUG 1454 if (mrt6debug & DEBUG_XMIT) 1455 log(LOG_DEBUG, "phyint_send on mif %d err %d\n", 1456 mifp - mif6table, error); 1457 #endif 1458 } else { 1459 #ifdef MULTICAST_PMTUD 1460 icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 1461 #else 1462 #ifdef MRT6DEBUG 1463 if (mrt6debug & DEBUG_XMIT) 1464 log(LOG_DEBUG, 1465 "phyint_send: packet too big on %s o %s g %s" 1466 " size %d(discarded)\n", 1467 if_name(ifp), 1468 ip6_sprintf(&ip6->ip6_src), 1469 ip6_sprintf(&ip6->ip6_dst), 1470 mb_copy->m_pkthdr.len); 1471 #endif /* MRT6DEBUG */ 1472 m_freem(mb_copy); /* simply discard the packet */ 1473 #endif 1474 } 1475 1476 crit_exit(); 1477 } 1478 1479 static int 1480 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m) 1481 { 1482 struct mbuf *mm; 1483 int i, len = m->m_pkthdr.len; 1484 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; 1485 struct mrt6msg *im6; 1486 1487 #ifdef MRT6DEBUG 1488 if (mrt6debug) 1489 log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n", 1490 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst)); 1491 #endif 1492 ++pim6stat.pim6s_snd_registers; 1493 1494 /* Make a copy of the packet to send to the user level process */ 1495 MGETHDR(mm, MB_DONTWAIT, MT_HEADER); 1496 if (mm == NULL) 1497 return ENOBUFS; 1498 mm->m_pkthdr.rcvif = NULL; 1499 mm->m_data += max_linkhdr; 1500 mm->m_len = sizeof(struct ip6_hdr); 1501 1502 if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1503 m_freem(mm); 1504 return ENOBUFS; 1505 } 1506 i = MHLEN - M_LEADINGSPACE(mm); 1507 if (i > len) 1508 i = len; 1509 mm = m_pullup(mm, i); 1510 if (mm == NULL) 1511 return ENOBUFS; 1512 /* TODO: check it! */ 1513 mm->m_pkthdr.len = len + sizeof(struct ip6_hdr); 1514 1515 /* 1516 * Send message to routing daemon 1517 */ 1518 sin6.sin6_addr = ip6->ip6_src; 1519 1520 im6 = mtod(mm, struct mrt6msg *); 1521 im6->im6_msgtype = MRT6MSG_WHOLEPKT; 1522 im6->im6_mbz = 0; 1523 1524 im6->im6_mif = mif - mif6table; 1525 1526 /* iif info is not given for reg. encap.n */ 1527 mrt6stat.mrt6s_upcalls++; 1528 1529 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1530 #ifdef MRT6DEBUG 1531 if (mrt6debug) 1532 log(LOG_WARNING, 1533 "register_send: ip6_mrouter socket queue full\n"); 1534 #endif 1535 ++mrt6stat.mrt6s_upq_sockfull; 1536 return ENOBUFS; 1537 } 1538 return 0; 1539 } 1540 1541 /* 1542 * PIM sparse mode hook 1543 * Receives the pim control messages, and passes them up to the listening 1544 * socket, using rip6_input. 1545 * The only message processed is the REGISTER pim message; the pim header 1546 * is stripped off, and the inner packet is passed to register_mforward. 1547 */ 1548 int 1549 pim6_input(struct mbuf **mp, int *offp, int proto) 1550 { 1551 struct pim *pim; /* pointer to a pim struct */ 1552 struct ip6_hdr *ip6; 1553 int pimlen; 1554 struct mbuf *m = *mp; 1555 int minlen; 1556 int off = *offp; 1557 1558 ++pim6stat.pim6s_rcv_total; 1559 1560 ip6 = mtod(m, struct ip6_hdr *); 1561 pimlen = m->m_pkthdr.len - *offp; 1562 1563 /* 1564 * Validate lengths 1565 */ 1566 if (pimlen < PIM_MINLEN) { 1567 ++pim6stat.pim6s_rcv_tooshort; 1568 #ifdef MRT6DEBUG 1569 if (mrt6debug & DEBUG_PIM) 1570 log(LOG_DEBUG,"pim6_input: PIM packet too short\n"); 1571 #endif 1572 m_freem(m); 1573 return(IPPROTO_DONE); 1574 } 1575 1576 /* 1577 * if the packet is at least as big as a REGISTER, go ahead 1578 * and grab the PIM REGISTER header size, to avoid another 1579 * possible m_pullup() later. 1580 * 1581 * PIM_MINLEN == pimhdr + u_int32 == 8 1582 * PIM6_REG_MINLEN == pimhdr + reghdr + eip6hdr == 4 + 4 + 40 1583 */ 1584 minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN; 1585 1586 /* 1587 * Make sure that the IP6 and PIM headers in contiguous memory, and 1588 * possibly the PIM REGISTER header 1589 */ 1590 #ifndef PULLDOWN_TEST 1591 IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE); 1592 /* adjust pointer */ 1593 ip6 = mtod(m, struct ip6_hdr *); 1594 1595 /* adjust mbuf to point to the PIM header */ 1596 pim = (struct pim *)((caddr_t)ip6 + off); 1597 #else 1598 IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen); 1599 if (pim == NULL) { 1600 pim6stat.pim6s_rcv_tooshort++; 1601 return IPPROTO_DONE; 1602 } 1603 #endif 1604 1605 #define PIM6_CHECKSUM 1606 #ifdef PIM6_CHECKSUM 1607 { 1608 int cksumlen; 1609 1610 /* 1611 * Validate checksum. 1612 * If PIM REGISTER, exclude the data packet 1613 */ 1614 if (pim->pim_type == PIM_REGISTER) 1615 cksumlen = PIM_MINLEN; 1616 else 1617 cksumlen = pimlen; 1618 1619 if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) { 1620 ++pim6stat.pim6s_rcv_badsum; 1621 #ifdef MRT6DEBUG 1622 if (mrt6debug & DEBUG_PIM) 1623 log(LOG_DEBUG, 1624 "pim6_input: invalid checksum\n"); 1625 #endif 1626 m_freem(m); 1627 return(IPPROTO_DONE); 1628 } 1629 } 1630 #endif /* PIM_CHECKSUM */ 1631 1632 /* PIM version check */ 1633 if (pim->pim_ver != PIM_VERSION) { 1634 ++pim6stat.pim6s_rcv_badversion; 1635 #ifdef MRT6DEBUG 1636 log(LOG_ERR, 1637 "pim6_input: incorrect version %d, expecting %d\n", 1638 pim->pim_ver, PIM_VERSION); 1639 #endif 1640 m_freem(m); 1641 return(IPPROTO_DONE); 1642 } 1643 1644 if (pim->pim_type == PIM_REGISTER) { 1645 /* 1646 * since this is a REGISTER, we'll make a copy of the register 1647 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the 1648 * routing daemon. 1649 */ 1650 static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 }; 1651 1652 struct mbuf *mcp; 1653 struct ip6_hdr *eip6; 1654 u_int32_t *reghdr; 1655 int rc; 1656 1657 ++pim6stat.pim6s_rcv_registers; 1658 1659 if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) { 1660 #ifdef MRT6DEBUG 1661 if (mrt6debug & DEBUG_PIM) 1662 log(LOG_DEBUG, 1663 "pim6_input: register mif not set: %d\n", 1664 reg_mif_num); 1665 #endif 1666 m_freem(m); 1667 return(IPPROTO_DONE); 1668 } 1669 1670 reghdr = (u_int32_t *)(pim + 1); 1671 1672 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 1673 goto pim6_input_to_daemon; 1674 1675 /* 1676 * Validate length 1677 */ 1678 if (pimlen < PIM6_REG_MINLEN) { 1679 ++pim6stat.pim6s_rcv_tooshort; 1680 ++pim6stat.pim6s_rcv_badregisters; 1681 #ifdef MRT6DEBUG 1682 log(LOG_ERR, 1683 "pim6_input: register packet size too " 1684 "small %d from %s\n", 1685 pimlen, ip6_sprintf(&ip6->ip6_src)); 1686 #endif 1687 m_freem(m); 1688 return(IPPROTO_DONE); 1689 } 1690 1691 eip6 = (struct ip6_hdr *) (reghdr + 1); 1692 #ifdef MRT6DEBUG 1693 if (mrt6debug & DEBUG_PIM) 1694 log(LOG_DEBUG, 1695 "pim6_input[register], eip6: %s -> %s, " 1696 "eip6 plen %d\n", 1697 ip6_sprintf(&eip6->ip6_src), 1698 ip6_sprintf(&eip6->ip6_dst), 1699 ntohs(eip6->ip6_plen)); 1700 #endif 1701 1702 /* verify the version number of the inner packet */ 1703 if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1704 ++pim6stat.pim6s_rcv_badregisters; 1705 #ifdef MRT6DEBUG 1706 log(LOG_DEBUG, "pim6_input: invalid IP version (%d) " 1707 "of the inner packet\n", 1708 (eip6->ip6_vfc & IPV6_VERSION)); 1709 #endif 1710 m_freem(m); 1711 return(IPPROTO_NONE); 1712 } 1713 1714 /* verify the inner packet is destined to a mcast group */ 1715 if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) { 1716 ++pim6stat.pim6s_rcv_badregisters; 1717 #ifdef MRT6DEBUG 1718 if (mrt6debug & DEBUG_PIM) 1719 log(LOG_DEBUG, 1720 "pim6_input: inner packet of register " 1721 "is not multicast %s\n", 1722 ip6_sprintf(&eip6->ip6_dst)); 1723 #endif 1724 m_freem(m); 1725 return(IPPROTO_DONE); 1726 } 1727 1728 /* 1729 * make a copy of the whole header to pass to the daemon later. 1730 */ 1731 mcp = m_copy(m, 0, off + PIM6_REG_MINLEN); 1732 if (mcp == NULL) { 1733 #ifdef MRT6DEBUG 1734 log(LOG_ERR, 1735 "pim6_input: pim register: " 1736 "could not copy register head\n"); 1737 #endif 1738 m_freem(m); 1739 return(IPPROTO_DONE); 1740 } 1741 1742 /* 1743 * forward the inner ip6 packet; point m_data at the inner ip6. 1744 */ 1745 m_adj(m, off + PIM_MINLEN); 1746 #ifdef MRT6DEBUG 1747 if (mrt6debug & DEBUG_PIM) { 1748 log(LOG_DEBUG, 1749 "pim6_input: forwarding decapsulated register: " 1750 "src %s, dst %s, mif %d\n", 1751 ip6_sprintf(&eip6->ip6_src), 1752 ip6_sprintf(&eip6->ip6_dst), 1753 reg_mif_num); 1754 } 1755 #endif 1756 1757 rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m, 1758 dst.sin6_family, NULL); 1759 1760 /* prepare the register head to send to the mrouting daemon */ 1761 m = mcp; 1762 } 1763 1764 /* 1765 * Pass the PIM message up to the daemon; if it is a register message 1766 * pass the 'head' only up to the daemon. This includes the 1767 * encapsulator ip6 header, pim header, register header and the 1768 * encapsulated ip6 header. 1769 */ 1770 pim6_input_to_daemon: 1771 rip6_input(&m, offp, proto); 1772 return(IPPROTO_DONE); 1773 } 1774