1 /* $FreeBSD: src/sys/netinet6/ip6_mroute.c,v 1.2.2.9 2003/01/23 21:06:47 sam Exp $ */ 2 /* $KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $ */ 3 4 /* 5 * Copyright (C) 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp */ 34 35 /* 36 * IP multicast forwarding procedures 37 * 38 * Written by David Waitzman, BBN Labs, August 1988. 39 * Modified by Steve Deering, Stanford, February 1989. 40 * Modified by Mark J. Steiglitz, Stanford, May, 1991 41 * Modified by Van Jacobson, LBL, January 1993 42 * Modified by Ajit Thyagarajan, PARC, August 1993 43 * Modified by Bill Fenenr, PARC, April 1994 44 * 45 * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support 46 */ 47 48 #include "opt_inet.h" 49 #include "opt_inet6.h" 50 51 #include <sys/param.h> 52 #include <sys/systm.h> 53 #include <sys/callout.h> 54 #include <sys/malloc.h> 55 #include <sys/mbuf.h> 56 #include <sys/socket.h> 57 #include <sys/socketvar.h> 58 #include <sys/sockio.h> 59 #include <sys/protosw.h> 60 #include <sys/errno.h> 61 #include <sys/time.h> 62 #include <sys/kernel.h> 63 #include <sys/syslog.h> 64 #include <sys/thread2.h> 65 66 #include <net/if.h> 67 #include <net/route.h> 68 #include <net/raw_cb.h> 69 70 #include <netinet/in.h> 71 #include <netinet/in_var.h> 72 73 #include <netinet/ip6.h> 74 #include <netinet6/ip6_var.h> 75 #include <netinet6/ip6_mroute.h> 76 #include <netinet6/pim6.h> 77 #include <netinet6/pim6_var.h> 78 79 #include <net/net_osdep.h> 80 81 static MALLOC_DEFINE(M_MRTABLE, "mf6c", "multicast forwarding cache entry"); 82 83 #define M_HASCL(m) ((m)->m_flags & M_EXT) 84 85 static int ip6_mdq (struct mbuf *, struct ifnet *, struct mf6c *); 86 static void phyint_send (struct ip6_hdr *, struct mif6 *, struct mbuf *); 87 88 static int set_pim6 (int *); 89 static int socket_send (struct socket *, struct mbuf *, 90 struct sockaddr_in6 *); 91 static int register_send (struct ip6_hdr *, struct mif6 *, 92 struct mbuf *); 93 94 /* 95 * Globals. All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static, 96 * except for netstat or debugging purposes. 97 */ 98 struct socket *ip6_mrouter = NULL; 99 int ip6_mrouter_ver = 0; 100 int ip6_mrtproto = IPPROTO_PIM; /* for netstat only */ 101 struct mrt6stat mrt6stat; 102 103 #define NO_RTE_FOUND 0x1 104 #define RTE_FOUND 0x2 105 106 struct mf6c *mf6ctable[MF6CTBLSIZ]; 107 u_char n6expire[MF6CTBLSIZ]; 108 static struct mif6 mif6table[MAXMIFS]; 109 #ifdef MRT6DEBUG 110 u_int mrt6debug = 0; /* debug level */ 111 #define DEBUG_MFC 0x02 112 #define DEBUG_FORWARD 0x04 113 #define DEBUG_EXPIRE 0x08 114 #define DEBUG_XMIT 0x10 115 #define DEBUG_REG 0x20 116 #define DEBUG_PIM 0x40 117 #endif 118 119 static void expire_upcalls (void *); 120 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 121 #define UPCALL_EXPIRE 6 /* number of timeouts */ 122 123 #ifdef INET 124 #ifdef MROUTING 125 extern struct socket *ip_mrouter; 126 #endif 127 #endif 128 129 /* 130 * 'Interfaces' associated with decapsulator (so we can tell 131 * packets that went through it from ones that get reflected 132 * by a broken gateway). These interfaces are never linked into 133 * the system ifnet list & no routes point to them. I.e., packets 134 * can't be sent this way. They only exist as a placeholder for 135 * multicast source verification. 136 */ 137 struct ifnet multicast_register_if; 138 139 #define ENCAP_HOPS 64 140 141 /* 142 * Private variables. 143 */ 144 static mifi_t nummifs = 0; 145 static mifi_t reg_mif_num = (mifi_t)-1; 146 147 static struct pim6stat pim6stat; 148 static int pim6; 149 150 /* 151 * Hash function for a source, group entry 152 */ 153 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \ 154 (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \ 155 (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \ 156 (g).s6_addr32[2] ^ (g).s6_addr32[3]) 157 158 /* 159 * Find a route for a given origin IPv6 address and Multicast group address. 160 * Quality of service parameter to be added in the future!!! 161 */ 162 163 #define MF6CFIND(o, g, rt) do { \ 164 struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \ 165 rt = NULL; \ 166 mrt6stat.mrt6s_mfc_lookups++; \ 167 while (_rt) { \ 168 if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \ 169 IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \ 170 (_rt->mf6c_stall == NULL)) { \ 171 rt = _rt; \ 172 break; \ 173 } \ 174 _rt = _rt->mf6c_next; \ 175 } \ 176 if (rt == NULL) { \ 177 mrt6stat.mrt6s_mfc_misses++; \ 178 } \ 179 } while (0) 180 181 /* 182 * Macros to compute elapsed time efficiently 183 * Borrowed from Van Jacobson's scheduling code 184 */ 185 #define TV_DELTA(a, b, delta) do { \ 186 int xxs; \ 187 \ 188 delta = (a).tv_usec - (b).tv_usec; \ 189 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 190 switch (xxs) { \ 191 case 2: \ 192 delta += 1000000; \ 193 /* fall through */ \ 194 case 1: \ 195 delta += 1000000; \ 196 break; \ 197 default: \ 198 delta += (1000000 * xxs); \ 199 } \ 200 } \ 201 } while (0) 202 203 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 204 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 205 206 #ifdef UPCALL_TIMING 207 #define UPCALL_MAX 50 208 u_long upcall_data[UPCALL_MAX + 1]; 209 static void collate(); 210 #endif /* UPCALL_TIMING */ 211 212 static int get_sg_cnt (struct sioc_sg_req6 *); 213 static int get_mif6_cnt (struct sioc_mif_req6 *); 214 static int ip6_mrouter_init (struct socket *, struct mbuf *, int); 215 static int add_m6if (struct mif6ctl *); 216 static int del_m6if (mifi_t *); 217 static int add_m6fc (struct mf6cctl *); 218 static int del_m6fc (struct mf6cctl *); 219 220 static struct callout expire_upcalls_ch; 221 222 /* 223 * Handle MRT setsockopt commands to modify the multicast routing tables. 224 */ 225 int 226 ip6_mrouter_set(struct socket *so, struct sockopt *sopt) 227 { 228 int error = 0; 229 struct mbuf *m; 230 231 if (so != ip6_mrouter && sopt->sopt_name != MRT6_INIT) 232 return (EACCES); 233 234 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 235 return (error); 236 soopt_to_mbuf(sopt, m); /* XXX */ 237 238 switch (sopt->sopt_name) { 239 case MRT6_INIT: 240 #ifdef MRT6_OINIT 241 case MRT6_OINIT: 242 #endif 243 error = ip6_mrouter_init(so, m, sopt->sopt_name); 244 break; 245 case MRT6_DONE: 246 error = ip6_mrouter_done(); 247 break; 248 case MRT6_ADD_MIF: 249 error = add_m6if(mtod(m, struct mif6ctl *)); 250 break; 251 case MRT6_DEL_MIF: 252 error = del_m6if(mtod(m, mifi_t *)); 253 break; 254 case MRT6_ADD_MFC: 255 error = add_m6fc(mtod(m, struct mf6cctl *)); 256 break; 257 case MRT6_DEL_MFC: 258 error = del_m6fc(mtod(m, struct mf6cctl *)); 259 break; 260 case MRT6_PIM: 261 error = set_pim6(mtod(m, int *)); 262 break; 263 default: 264 error = EOPNOTSUPP; 265 break; 266 } 267 268 m_freem(m); 269 return (error); 270 } 271 272 /* 273 * Handle MRT getsockopt commands 274 */ 275 int 276 ip6_mrouter_get(struct socket *so, struct sockopt *sopt) 277 { 278 int error = 0; 279 280 if (so != ip6_mrouter) return EACCES; 281 282 switch (sopt->sopt_name) { 283 case MRT6_PIM: 284 soopt_from_kbuf(sopt, &pim6, sizeof(pim6)); 285 break; 286 } 287 return (error); 288 } 289 290 /* 291 * Handle ioctl commands to obtain information from the cache 292 */ 293 int 294 mrt6_ioctl(int cmd, caddr_t data) 295 { 296 int error = 0; 297 298 switch (cmd) { 299 case SIOCGETSGCNT_IN6: 300 return (get_sg_cnt((struct sioc_sg_req6 *)data)); 301 break; /* for safety */ 302 case SIOCGETMIFCNT_IN6: 303 return (get_mif6_cnt((struct sioc_mif_req6 *)data)); 304 break; /* for safety */ 305 default: 306 return (EINVAL); 307 break; 308 } 309 return error; 310 } 311 312 /* 313 * returns the packet, byte, rpf-failure count for the source group provided 314 */ 315 static int 316 get_sg_cnt(struct sioc_sg_req6 *req) 317 { 318 struct mf6c *rt; 319 320 crit_enter(); 321 MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt); 322 crit_exit(); 323 if (rt != NULL) { 324 req->pktcnt = rt->mf6c_pkt_cnt; 325 req->bytecnt = rt->mf6c_byte_cnt; 326 req->wrong_if = rt->mf6c_wrong_if; 327 } else 328 return (ESRCH); 329 #if 0 330 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 331 #endif 332 333 return 0; 334 } 335 336 /* 337 * returns the input and output packet and byte counts on the mif provided 338 */ 339 static int 340 get_mif6_cnt(struct sioc_mif_req6 *req) 341 { 342 mifi_t mifi = req->mifi; 343 344 if (mifi >= nummifs) 345 return EINVAL; 346 347 req->icount = mif6table[mifi].m6_pkt_in; 348 req->ocount = mif6table[mifi].m6_pkt_out; 349 req->ibytes = mif6table[mifi].m6_bytes_in; 350 req->obytes = mif6table[mifi].m6_bytes_out; 351 352 return 0; 353 } 354 355 static int 356 set_pim6(int *i) 357 { 358 if ((*i != 1) && (*i != 0)) 359 return EINVAL; 360 361 pim6 = *i; 362 363 return 0; 364 } 365 366 /* 367 * Enable multicast routing 368 */ 369 static int 370 ip6_mrouter_init(struct socket *so, struct mbuf *m, int cmd) 371 { 372 int *v; 373 374 #ifdef MRT6DEBUG 375 if (mrt6debug) 376 log(LOG_DEBUG, 377 "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n", 378 so->so_type, so->so_proto->pr_protocol); 379 #endif 380 381 if (so->so_type != SOCK_RAW || 382 so->so_proto->pr_protocol != IPPROTO_ICMPV6) 383 return EOPNOTSUPP; 384 385 if (!m || (m->m_len != sizeof(int *))) 386 return ENOPROTOOPT; 387 388 v = mtod(m, int *); 389 if (*v != 1) 390 return ENOPROTOOPT; 391 392 if (ip6_mrouter != NULL) return EADDRINUSE; 393 394 ip6_mrouter = so; 395 ip6_mrouter_ver = cmd; 396 397 bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); 398 bzero((caddr_t)n6expire, sizeof(n6expire)); 399 400 pim6 = 0;/* used for stubbing out/in pim stuff */ 401 402 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 403 expire_upcalls, NULL); 404 405 #ifdef MRT6DEBUG 406 if (mrt6debug) 407 log(LOG_DEBUG, "ip6_mrouter_init\n"); 408 #endif 409 410 return 0; 411 } 412 413 /* 414 * Disable multicast routing 415 */ 416 int 417 ip6_mrouter_done(void) 418 { 419 mifi_t mifi; 420 int i; 421 struct ifnet *ifp; 422 struct in6_ifreq ifr; 423 struct mf6c *rt; 424 struct rtdetq *rte; 425 426 /* 427 * For each phyint in use, disable promiscuous reception of all IPv6 428 * multicasts. 429 */ 430 #ifdef INET 431 #ifdef MROUTING 432 /* 433 * If there is still IPv4 multicast routing daemon, 434 * we remain interfaces to receive all muliticasted packets. 435 * XXX: there may be an interface in which the IPv4 multicast 436 * daemon is not interested... 437 */ 438 if (!ip_mrouter) 439 #endif 440 #endif 441 { 442 for (mifi = 0; mifi < nummifs; mifi++) { 443 if (mif6table[mifi].m6_ifp && 444 !(mif6table[mifi].m6_flags & MIFF_REGISTER)) { 445 ifr.ifr_addr.sin6_family = AF_INET6; 446 ifr.ifr_addr.sin6_addr= kin6addr_any; 447 ifp = mif6table[mifi].m6_ifp; 448 ifnet_serialize_all(ifp); 449 ifp->if_ioctl(ifp, SIOCDELMULTI, 450 (caddr_t)&ifr, NULL); 451 ifnet_deserialize_all(ifp); 452 } 453 } 454 } 455 #ifdef notyet 456 bzero((caddr_t)qtable, sizeof(qtable)); 457 bzero((caddr_t)tbftable, sizeof(tbftable)); 458 #endif 459 bzero((caddr_t)mif6table, sizeof(mif6table)); 460 nummifs = 0; 461 462 pim6 = 0; /* used to stub out/in pim specific code */ 463 464 callout_stop(&expire_upcalls_ch); 465 466 /* 467 * Free all multicast forwarding cache entries. 468 */ 469 for (i = 0; i < MF6CTBLSIZ; i++) { 470 rt = mf6ctable[i]; 471 while (rt) { 472 struct mf6c *frt; 473 474 for (rte = rt->mf6c_stall; rte != NULL; ) { 475 struct rtdetq *n = rte->next; 476 477 m_freem(rte->m); 478 kfree(rte, M_MRTABLE); 479 rte = n; 480 } 481 frt = rt; 482 rt = rt->mf6c_next; 483 kfree(frt, M_MRTABLE); 484 } 485 } 486 487 bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); 488 489 /* 490 * Reset de-encapsulation cache 491 */ 492 reg_mif_num = -1; 493 494 ip6_mrouter = NULL; 495 ip6_mrouter_ver = 0; 496 497 #ifdef MRT6DEBUG 498 if (mrt6debug) 499 log(LOG_DEBUG, "ip6_mrouter_done\n"); 500 #endif 501 502 return 0; 503 } 504 505 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; 506 507 /* 508 * Add a mif to the mif table 509 */ 510 static int 511 add_m6if(struct mif6ctl *mifcp) 512 { 513 struct mif6 *mifp; 514 struct ifnet *ifp; 515 int error; 516 #ifdef notyet 517 struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi; 518 #endif 519 520 if (mifcp->mif6c_mifi >= MAXMIFS) 521 return EINVAL; 522 mifp = mif6table + mifcp->mif6c_mifi; 523 if (mifp->m6_ifp) 524 return EADDRINUSE; /* XXX: is it appropriate? */ 525 if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index) 526 return ENXIO; 527 ifp = ifindex2ifnet[mifcp->mif6c_pifi]; 528 529 if (mifcp->mif6c_flags & MIFF_REGISTER) { 530 if (reg_mif_num == (mifi_t)-1) { 531 strlcpy(multicast_register_if.if_xname, "register_mif", 532 IFNAMSIZ); 533 multicast_register_if.if_flags |= IFF_LOOPBACK; 534 multicast_register_if.if_index = mifcp->mif6c_mifi; 535 reg_mif_num = mifcp->mif6c_mifi; 536 } 537 538 ifp = &multicast_register_if; 539 540 } /* if REGISTER */ 541 else { 542 /* Make sure the interface supports multicast */ 543 if ((ifp->if_flags & IFF_MULTICAST) == 0) 544 return EOPNOTSUPP; 545 546 crit_enter(); 547 error = if_allmulti(ifp, 1); 548 crit_exit(); 549 if (error) 550 return error; 551 } 552 553 crit_enter(); 554 mifp->m6_flags = mifcp->mif6c_flags; 555 mifp->m6_ifp = ifp; 556 #ifdef notyet 557 /* scaling up here allows division by 1024 in critical code */ 558 mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000; 559 #endif 560 /* initialize per mif pkt counters */ 561 mifp->m6_pkt_in = 0; 562 mifp->m6_pkt_out = 0; 563 mifp->m6_bytes_in = 0; 564 mifp->m6_bytes_out = 0; 565 crit_exit(); 566 567 /* Adjust nummifs up if the mifi is higher than nummifs */ 568 if (nummifs <= mifcp->mif6c_mifi) 569 nummifs = mifcp->mif6c_mifi + 1; 570 571 #ifdef MRT6DEBUG 572 if (mrt6debug) 573 log(LOG_DEBUG, 574 "add_mif #%d, phyint %s\n", 575 mifcp->mif6c_mifi, 576 ifp->if_xname); 577 #endif 578 579 return 0; 580 } 581 582 /* 583 * Delete a mif from the mif table 584 */ 585 static int 586 del_m6if(mifi_t *mifip) 587 { 588 struct mif6 *mifp = mif6table + *mifip; 589 mifi_t mifi; 590 struct ifnet *ifp; 591 592 if (*mifip >= nummifs) 593 return EINVAL; 594 if (mifp->m6_ifp == NULL) 595 return EINVAL; 596 597 crit_enter(); 598 599 if (!(mifp->m6_flags & MIFF_REGISTER)) { 600 /* 601 * XXX: what if there is yet IPv4 multicast daemon 602 * using the interface? 603 */ 604 ifp = mifp->m6_ifp; 605 606 if_allmulti(ifp, 0); 607 } 608 609 #ifdef notyet 610 bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip])); 611 bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf))); 612 #endif 613 bzero((caddr_t)mifp, sizeof (*mifp)); 614 615 /* Adjust nummifs down */ 616 for (mifi = nummifs; mifi > 0; mifi--) 617 if (mif6table[mifi - 1].m6_ifp) 618 break; 619 nummifs = mifi; 620 621 crit_exit(); 622 623 #ifdef MRT6DEBUG 624 if (mrt6debug) 625 log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs); 626 #endif 627 628 return 0; 629 } 630 631 /* 632 * Add an mfc entry 633 */ 634 static int 635 add_m6fc(struct mf6cctl *mfccp) 636 { 637 struct mf6c *rt; 638 u_long hash; 639 struct rtdetq *rte; 640 u_short nstl; 641 642 MF6CFIND(mfccp->mf6cc_origin.sin6_addr, 643 mfccp->mf6cc_mcastgrp.sin6_addr, rt); 644 645 /* If an entry already exists, just update the fields */ 646 if (rt) { 647 #ifdef MRT6DEBUG 648 if (mrt6debug & DEBUG_MFC) 649 log(LOG_DEBUG, 650 "add_m6fc no upcall h %d o %s g %s p %x\n", 651 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 652 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 653 mfccp->mf6cc_parent); 654 #endif 655 656 crit_enter(); 657 rt->mf6c_parent = mfccp->mf6cc_parent; 658 rt->mf6c_ifset = mfccp->mf6cc_ifset; 659 crit_exit(); 660 return 0; 661 } 662 663 /* 664 * Find the entry for which the upcall was made and update 665 */ 666 crit_enter(); 667 hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr, 668 mfccp->mf6cc_mcastgrp.sin6_addr); 669 for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) { 670 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, 671 &mfccp->mf6cc_origin.sin6_addr) && 672 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, 673 &mfccp->mf6cc_mcastgrp.sin6_addr) && 674 (rt->mf6c_stall != NULL)) { 675 676 if (nstl++) 677 log(LOG_ERR, 678 "add_m6fc: %s o %s g %s p %x dbx %p\n", 679 "multiple kernel entries", 680 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 681 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 682 mfccp->mf6cc_parent, rt->mf6c_stall); 683 684 #ifdef MRT6DEBUG 685 if (mrt6debug & DEBUG_MFC) 686 log(LOG_DEBUG, 687 "add_m6fc o %s g %s p %x dbg %x\n", 688 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 689 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 690 mfccp->mf6cc_parent, rt->mf6c_stall); 691 #endif 692 693 rt->mf6c_origin = mfccp->mf6cc_origin; 694 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 695 rt->mf6c_parent = mfccp->mf6cc_parent; 696 rt->mf6c_ifset = mfccp->mf6cc_ifset; 697 /* initialize pkt counters per src-grp */ 698 rt->mf6c_pkt_cnt = 0; 699 rt->mf6c_byte_cnt = 0; 700 rt->mf6c_wrong_if = 0; 701 702 rt->mf6c_expire = 0; /* Don't clean this guy up */ 703 n6expire[hash]--; 704 705 /* free packets Qed at the end of this entry */ 706 for (rte = rt->mf6c_stall; rte != NULL; ) { 707 struct rtdetq *n = rte->next; 708 ip6_mdq(rte->m, rte->ifp, rt); 709 m_freem(rte->m); 710 #ifdef UPCALL_TIMING 711 collate(&(rte->t)); 712 #endif /* UPCALL_TIMING */ 713 kfree(rte, M_MRTABLE); 714 rte = n; 715 } 716 rt->mf6c_stall = NULL; 717 } 718 } 719 720 /* 721 * It is possible that an entry is being inserted without an upcall 722 */ 723 if (nstl == 0) { 724 #ifdef MRT6DEBUG 725 if (mrt6debug & DEBUG_MFC) 726 log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n", 727 hash, 728 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr), 729 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr), 730 mfccp->mf6cc_parent); 731 #endif 732 733 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { 734 735 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, 736 &mfccp->mf6cc_origin.sin6_addr)&& 737 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, 738 &mfccp->mf6cc_mcastgrp.sin6_addr)) { 739 740 rt->mf6c_origin = mfccp->mf6cc_origin; 741 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 742 rt->mf6c_parent = mfccp->mf6cc_parent; 743 rt->mf6c_ifset = mfccp->mf6cc_ifset; 744 /* initialize pkt counters per src-grp */ 745 rt->mf6c_pkt_cnt = 0; 746 rt->mf6c_byte_cnt = 0; 747 rt->mf6c_wrong_if = 0; 748 749 if (rt->mf6c_expire) 750 n6expire[hash]--; 751 rt->mf6c_expire = 0; 752 } 753 } 754 if (rt == NULL) { 755 /* no upcall, so make a new entry */ 756 rt = (struct mf6c *)kmalloc(sizeof(*rt), M_MRTABLE, 757 M_NOWAIT); 758 if (rt == NULL) { 759 crit_exit(); 760 return ENOBUFS; 761 } 762 763 /* insert new entry at head of hash chain */ 764 rt->mf6c_origin = mfccp->mf6cc_origin; 765 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; 766 rt->mf6c_parent = mfccp->mf6cc_parent; 767 rt->mf6c_ifset = mfccp->mf6cc_ifset; 768 /* initialize pkt counters per src-grp */ 769 rt->mf6c_pkt_cnt = 0; 770 rt->mf6c_byte_cnt = 0; 771 rt->mf6c_wrong_if = 0; 772 rt->mf6c_expire = 0; 773 rt->mf6c_stall = NULL; 774 775 /* link into table */ 776 rt->mf6c_next = mf6ctable[hash]; 777 mf6ctable[hash] = rt; 778 } 779 } 780 crit_exit(); 781 return 0; 782 } 783 784 #ifdef UPCALL_TIMING 785 /* 786 * collect delay statistics on the upcalls 787 */ 788 static void 789 collate(struct timeval *t) 790 { 791 u_long d; 792 struct timeval tp; 793 u_long delta; 794 795 GET_TIME(tp); 796 797 if (TV_LT(*t, tp)) 798 { 799 TV_DELTA(tp, *t, delta); 800 801 d = delta >> 10; 802 if (d > UPCALL_MAX) 803 d = UPCALL_MAX; 804 805 ++upcall_data[d]; 806 } 807 } 808 #endif /* UPCALL_TIMING */ 809 810 /* 811 * Delete an mfc entry 812 */ 813 static int 814 del_m6fc(struct mf6cctl *mfccp) 815 { 816 struct sockaddr_in6 origin; 817 struct sockaddr_in6 mcastgrp; 818 struct mf6c *rt; 819 struct mf6c **nptr; 820 u_long hash; 821 822 origin = mfccp->mf6cc_origin; 823 mcastgrp = mfccp->mf6cc_mcastgrp; 824 hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr); 825 826 #ifdef MRT6DEBUG 827 if (mrt6debug & DEBUG_MFC) 828 log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n", 829 ip6_sprintf(&origin.sin6_addr), 830 ip6_sprintf(&mcastgrp.sin6_addr)); 831 #endif 832 833 crit_enter(); 834 835 nptr = &mf6ctable[hash]; 836 while ((rt = *nptr) != NULL) { 837 if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr, 838 &rt->mf6c_origin.sin6_addr) && 839 IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr, 840 &rt->mf6c_mcastgrp.sin6_addr) && 841 rt->mf6c_stall == NULL) 842 break; 843 844 nptr = &rt->mf6c_next; 845 } 846 if (rt == NULL) { 847 crit_exit(); 848 return EADDRNOTAVAIL; 849 } 850 851 *nptr = rt->mf6c_next; 852 kfree(rt, M_MRTABLE); 853 854 crit_exit(); 855 856 return 0; 857 } 858 859 static int 860 socket_send(struct socket *so, struct mbuf *mm, struct sockaddr_in6 *src) 861 { 862 if (so) { 863 lwkt_gettoken(&so->so_rcv.ssb_token); 864 if (ssb_appendaddr(&so->so_rcv, 865 (struct sockaddr *)src, 866 mm, NULL) != 0) { 867 sorwakeup(so); 868 lwkt_reltoken(&so->so_rcv.ssb_token); 869 return 0; 870 } 871 lwkt_reltoken(&so->so_rcv.ssb_token); 872 } 873 m_freem(mm); 874 return -1; 875 } 876 877 /* 878 * IPv6 multicast forwarding function. This function assumes that the packet 879 * pointed to by "ip6" has arrived on (or is about to be sent to) the interface 880 * pointed to by "ifp", and the packet is to be relayed to other networks 881 * that have members of the packet's destination IPv6 multicast group. 882 * 883 * The packet is returned unscathed to the caller, unless it is 884 * erroneous, in which case a non-zero return value tells the caller to 885 * discard it. 886 */ 887 888 int 889 ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m) 890 { 891 struct mf6c *rt; 892 struct mif6 *mifp; 893 struct mbuf *mm; 894 mifi_t mifi; 895 896 #ifdef MRT6DEBUG 897 if (mrt6debug & DEBUG_FORWARD) 898 log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n", 899 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst), 900 ifp->if_index); 901 #endif 902 903 /* 904 * Don't forward a packet with Hop limit of zero or one, 905 * or a packet destined to a local-only group. 906 */ 907 if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) || 908 IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst)) 909 return 0; 910 ip6->ip6_hlim--; 911 912 /* 913 * Source address check: do not forward packets with unspecified 914 * source. It was discussed in July 2000, on ipngwg mailing list. 915 * This is rather more serious than unicast cases, because some 916 * MLD packets can be sent with the unspecified source address 917 * (although such packets must normally set 1 to the hop limit field). 918 */ 919 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 920 ip6stat.ip6s_cantforward++; 921 if (ip6_log_time + ip6_log_interval < time_uptime) { 922 ip6_log_time = time_uptime; 923 log(LOG_DEBUG, 924 "cannot forward " 925 "from %s to %s nxt %d received on %s\n", 926 ip6_sprintf(&ip6->ip6_src), 927 ip6_sprintf(&ip6->ip6_dst), 928 ip6->ip6_nxt, 929 if_name(m->m_pkthdr.rcvif)); 930 } 931 return 0; 932 } 933 934 /* 935 * Determine forwarding mifs from the forwarding cache table 936 */ 937 crit_enter(); 938 MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt); 939 940 /* Entry exists, so forward if necessary */ 941 if (rt) { 942 crit_exit(); 943 return (ip6_mdq(m, ifp, rt)); 944 } else { 945 /* 946 * If we don't have a route for packet's origin, 947 * Make a copy of the packet & 948 * send message to routing daemon 949 */ 950 951 struct mbuf *mb0; 952 struct rtdetq *rte; 953 u_long hash; 954 /* int i, npkts;*/ 955 #ifdef UPCALL_TIMING 956 struct timeval tp; 957 958 GET_TIME(tp); 959 #endif /* UPCALL_TIMING */ 960 961 mrt6stat.mrt6s_no_route++; 962 #ifdef MRT6DEBUG 963 if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC)) 964 log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n", 965 ip6_sprintf(&ip6->ip6_src), 966 ip6_sprintf(&ip6->ip6_dst)); 967 #endif 968 969 /* 970 * Allocate mbufs early so that we don't do extra work if we 971 * are just going to fail anyway. 972 */ 973 rte = (struct rtdetq *)kmalloc(sizeof(*rte), M_MRTABLE, 974 M_NOWAIT); 975 if (rte == NULL) { 976 crit_exit(); 977 return ENOBUFS; 978 } 979 mb0 = m_copy(m, 0, M_COPYALL); 980 /* 981 * Pullup packet header if needed before storing it, 982 * as other references may modify it in the meantime. 983 */ 984 if (mb0 && 985 (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr))) 986 mb0 = m_pullup(mb0, sizeof(struct ip6_hdr)); 987 if (mb0 == NULL) { 988 kfree(rte, M_MRTABLE); 989 crit_exit(); 990 return ENOBUFS; 991 } 992 993 /* is there an upcall waiting for this packet? */ 994 hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst); 995 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { 996 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, 997 &rt->mf6c_origin.sin6_addr) && 998 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 999 &rt->mf6c_mcastgrp.sin6_addr) && 1000 (rt->mf6c_stall != NULL)) 1001 break; 1002 } 1003 1004 if (rt == NULL) { 1005 struct mrt6msg *im; 1006 #ifdef MRT6_OINIT 1007 struct omrt6msg *oim; 1008 #endif 1009 1010 /* no upcall, so make a new entry */ 1011 rt = (struct mf6c *)kmalloc(sizeof(*rt), M_MRTABLE, 1012 M_NOWAIT); 1013 if (rt == NULL) { 1014 kfree(rte, M_MRTABLE); 1015 m_freem(mb0); 1016 crit_exit(); 1017 return ENOBUFS; 1018 } 1019 /* 1020 * Make a copy of the header to send to the user 1021 * level process 1022 */ 1023 mm = m_copy(mb0, 0, sizeof(struct ip6_hdr)); 1024 1025 if (mm == NULL) { 1026 kfree(rte, M_MRTABLE); 1027 m_freem(mb0); 1028 kfree(rt, M_MRTABLE); 1029 crit_exit(); 1030 return ENOBUFS; 1031 } 1032 1033 /* 1034 * Send message to routing daemon 1035 */ 1036 sin6.sin6_addr = ip6->ip6_src; 1037 1038 im = NULL; 1039 #ifdef MRT6_OINIT 1040 oim = NULL; 1041 #endif 1042 switch (ip6_mrouter_ver) { 1043 #ifdef MRT6_OINIT 1044 case MRT6_OINIT: 1045 oim = mtod(mm, struct omrt6msg *); 1046 oim->im6_msgtype = MRT6MSG_NOCACHE; 1047 oim->im6_mbz = 0; 1048 break; 1049 #endif 1050 case MRT6_INIT: 1051 im = mtod(mm, struct mrt6msg *); 1052 im->im6_msgtype = MRT6MSG_NOCACHE; 1053 im->im6_mbz = 0; 1054 break; 1055 default: 1056 kfree(rte, M_MRTABLE); 1057 m_freem(mb0); 1058 kfree(rt, M_MRTABLE); 1059 crit_exit(); 1060 return EINVAL; 1061 } 1062 1063 #ifdef MRT6DEBUG 1064 if (mrt6debug & DEBUG_FORWARD) 1065 log(LOG_DEBUG, 1066 "getting the iif info in the kernel\n"); 1067 #endif 1068 1069 for (mifp = mif6table, mifi = 0; 1070 mifi < nummifs && mifp->m6_ifp != ifp; 1071 mifp++, mifi++) 1072 ; 1073 1074 switch (ip6_mrouter_ver) { 1075 #ifdef MRT6_OINIT 1076 case MRT6_OINIT: 1077 oim->im6_mif = mifi; 1078 break; 1079 #endif 1080 case MRT6_INIT: 1081 im->im6_mif = mifi; 1082 break; 1083 } 1084 1085 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1086 log(LOG_WARNING, "ip6_mforward: ip6_mrouter " 1087 "socket queue full\n"); 1088 mrt6stat.mrt6s_upq_sockfull++; 1089 kfree(rte, M_MRTABLE); 1090 m_freem(mb0); 1091 kfree(rt, M_MRTABLE); 1092 crit_exit(); 1093 return ENOBUFS; 1094 } 1095 1096 mrt6stat.mrt6s_upcalls++; 1097 1098 /* insert new entry at head of hash chain */ 1099 bzero(rt, sizeof(*rt)); 1100 rt->mf6c_origin.sin6_family = AF_INET6; 1101 rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6); 1102 rt->mf6c_origin.sin6_addr = ip6->ip6_src; 1103 rt->mf6c_mcastgrp.sin6_family = AF_INET6; 1104 rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6); 1105 rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst; 1106 rt->mf6c_expire = UPCALL_EXPIRE; 1107 n6expire[hash]++; 1108 rt->mf6c_parent = MF6C_INCOMPLETE_PARENT; 1109 1110 /* link into table */ 1111 rt->mf6c_next = mf6ctable[hash]; 1112 mf6ctable[hash] = rt; 1113 /* Add this entry to the end of the queue */ 1114 rt->mf6c_stall = rte; 1115 } else { 1116 /* determine if q has overflowed */ 1117 struct rtdetq **p; 1118 int npkts = 0; 1119 1120 for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next) 1121 if (++npkts > MAX_UPQ6) { 1122 mrt6stat.mrt6s_upq_ovflw++; 1123 kfree(rte, M_MRTABLE); 1124 m_freem(mb0); 1125 crit_exit(); 1126 return 0; 1127 } 1128 1129 /* Add this entry to the end of the queue */ 1130 *p = rte; 1131 } 1132 1133 rte->next = NULL; 1134 rte->m = mb0; 1135 rte->ifp = ifp; 1136 #ifdef UPCALL_TIMING 1137 rte->t = tp; 1138 #endif /* UPCALL_TIMING */ 1139 1140 crit_exit(); 1141 1142 return 0; 1143 } 1144 } 1145 1146 /* 1147 * Clean up cache entries if upcalls are not serviced 1148 * Call from the Slow Timeout mechanism, every half second. 1149 */ 1150 static void 1151 expire_upcalls(void *unused) 1152 { 1153 struct rtdetq *rte; 1154 struct mf6c *mfc, **nptr; 1155 int i; 1156 1157 crit_enter(); 1158 for (i = 0; i < MF6CTBLSIZ; i++) { 1159 if (n6expire[i] == 0) 1160 continue; 1161 nptr = &mf6ctable[i]; 1162 while ((mfc = *nptr) != NULL) { 1163 rte = mfc->mf6c_stall; 1164 /* 1165 * Skip real cache entries 1166 * Make sure it wasn't marked to not expire (shouldn't happen) 1167 * If it expires now 1168 */ 1169 if (rte != NULL && 1170 mfc->mf6c_expire != 0 && 1171 --mfc->mf6c_expire == 0) { 1172 #ifdef MRT6DEBUG 1173 if (mrt6debug & DEBUG_EXPIRE) 1174 log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n", 1175 ip6_sprintf(&mfc->mf6c_origin.sin6_addr), 1176 ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr)); 1177 #endif 1178 /* 1179 * drop all the packets 1180 * free the mbuf with the pkt, if, timing info 1181 */ 1182 do { 1183 struct rtdetq *n = rte->next; 1184 m_freem(rte->m); 1185 kfree(rte, M_MRTABLE); 1186 rte = n; 1187 } while (rte != NULL); 1188 mrt6stat.mrt6s_cache_cleanups++; 1189 n6expire[i]--; 1190 1191 *nptr = mfc->mf6c_next; 1192 kfree(mfc, M_MRTABLE); 1193 } else { 1194 nptr = &mfc->mf6c_next; 1195 } 1196 } 1197 } 1198 crit_exit(); 1199 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1200 expire_upcalls, NULL); 1201 } 1202 1203 /* 1204 * Packet forwarding routine once entry in the cache is made 1205 */ 1206 static int 1207 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt) 1208 { 1209 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1210 mifi_t mifi, iif; 1211 struct mif6 *mifp; 1212 int plen = m->m_pkthdr.len; 1213 1214 /* 1215 * Macro to send packet on mif. Since RSVP packets don't get counted on 1216 * input, they shouldn't get counted on output, so statistics keeping is 1217 * separate. 1218 */ 1219 1220 #define MC6_SEND(ip6, mifp, m) do { \ 1221 if ((mifp)->m6_flags & MIFF_REGISTER) \ 1222 register_send((ip6), (mifp), (m)); \ 1223 else \ 1224 phyint_send((ip6), (mifp), (m)); \ 1225 } while (0) 1226 1227 /* 1228 * Don't forward if it didn't arrive from the parent mif 1229 * for its origin. 1230 */ 1231 mifi = rt->mf6c_parent; 1232 if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) { 1233 /* came in the wrong interface */ 1234 #ifdef MRT6DEBUG 1235 if (mrt6debug & DEBUG_FORWARD) 1236 log(LOG_DEBUG, 1237 "wrong if: ifid %d mifi %d mififid %x\n", 1238 ifp->if_index, mifi, 1239 mif6table[mifi].m6_ifp->if_index); 1240 #endif 1241 mrt6stat.mrt6s_wrong_if++; 1242 rt->mf6c_wrong_if++; 1243 /* 1244 * If we are doing PIM processing, and we are forwarding 1245 * packets on this interface, send a message to the 1246 * routing daemon. 1247 */ 1248 /* have to make sure this is a valid mif */ 1249 if (mifi < nummifs && mif6table[mifi].m6_ifp) 1250 if (pim6 && (m->m_flags & M_LOOP) == 0) { 1251 /* 1252 * Check the M_LOOP flag to avoid an 1253 * unnecessary PIM assert. 1254 * XXX: M_LOOP is an ad-hoc hack... 1255 */ 1256 static struct sockaddr_in6 sin6 = 1257 { sizeof(sin6), AF_INET6 }; 1258 1259 struct mbuf *mm; 1260 struct mrt6msg *im; 1261 #ifdef MRT6_OINIT 1262 struct omrt6msg *oim; 1263 #endif 1264 1265 mm = m_copy(m, 0, sizeof(struct ip6_hdr)); 1266 if (mm && 1267 (M_HASCL(mm) || 1268 mm->m_len < sizeof(struct ip6_hdr))) 1269 mm = m_pullup(mm, sizeof(struct ip6_hdr)); 1270 if (mm == NULL) 1271 return ENOBUFS; 1272 1273 #ifdef MRT6_OINIT 1274 oim = NULL; 1275 #endif 1276 im = NULL; 1277 switch (ip6_mrouter_ver) { 1278 #ifdef MRT6_OINIT 1279 case MRT6_OINIT: 1280 oim = mtod(mm, struct omrt6msg *); 1281 oim->im6_msgtype = MRT6MSG_WRONGMIF; 1282 oim->im6_mbz = 0; 1283 break; 1284 #endif 1285 case MRT6_INIT: 1286 im = mtod(mm, struct mrt6msg *); 1287 im->im6_msgtype = MRT6MSG_WRONGMIF; 1288 im->im6_mbz = 0; 1289 break; 1290 default: 1291 m_freem(mm); 1292 return EINVAL; 1293 } 1294 1295 for (mifp = mif6table, iif = 0; 1296 iif < nummifs && mifp && 1297 mifp->m6_ifp != ifp; 1298 mifp++, iif++) 1299 ; 1300 1301 switch (ip6_mrouter_ver) { 1302 #ifdef MRT6_OINIT 1303 case MRT6_OINIT: 1304 oim->im6_mif = iif; 1305 sin6.sin6_addr = oim->im6_src; 1306 break; 1307 #endif 1308 case MRT6_INIT: 1309 im->im6_mif = iif; 1310 sin6.sin6_addr = im->im6_src; 1311 break; 1312 } 1313 1314 mrt6stat.mrt6s_upcalls++; 1315 1316 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1317 #ifdef MRT6DEBUG 1318 if (mrt6debug) 1319 log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n"); 1320 #endif 1321 ++mrt6stat.mrt6s_upq_sockfull; 1322 return ENOBUFS; 1323 } /* if socket Q full */ 1324 } /* if PIM */ 1325 return 0; 1326 } /* if wrong iif */ 1327 1328 /* If I sourced this packet, it counts as output, else it was input. */ 1329 if (m->m_pkthdr.rcvif == NULL) { 1330 /* XXX: is rcvif really NULL when output?? */ 1331 mif6table[mifi].m6_pkt_out++; 1332 mif6table[mifi].m6_bytes_out += plen; 1333 } else { 1334 mif6table[mifi].m6_pkt_in++; 1335 mif6table[mifi].m6_bytes_in += plen; 1336 } 1337 rt->mf6c_pkt_cnt++; 1338 rt->mf6c_byte_cnt += plen; 1339 1340 /* 1341 * For each mif, forward a copy of the packet if there are group 1342 * members downstream on the interface. 1343 */ 1344 for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) 1345 if (IF_ISSET(mifi, &rt->mf6c_ifset)) { 1346 /* 1347 * check if the outgoing packet is going to break 1348 * a scope boundary. 1349 * XXX For packets through PIM register tunnel 1350 * interface, we believe a routing daemon. 1351 */ 1352 if ((mif6table[rt->mf6c_parent].m6_flags & 1353 MIFF_REGISTER) == 0 && 1354 (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 && 1355 (in6_addr2scopeid(ifp, &ip6->ip6_dst) != 1356 in6_addr2scopeid(mif6table[mifi].m6_ifp, 1357 &ip6->ip6_dst) || 1358 in6_addr2scopeid(ifp, &ip6->ip6_src) != 1359 in6_addr2scopeid(mif6table[mifi].m6_ifp, 1360 &ip6->ip6_src))) { 1361 ip6stat.ip6s_badscope++; 1362 continue; 1363 } 1364 1365 mifp->m6_pkt_out++; 1366 mifp->m6_bytes_out += plen; 1367 MC6_SEND(ip6, mifp, m); 1368 } 1369 return 0; 1370 } 1371 1372 static void 1373 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m) 1374 { 1375 struct mbuf *mb_copy; 1376 struct ifnet *ifp = mifp->m6_ifp; 1377 int error = 0; 1378 static struct route_in6 ro; 1379 struct in6_multi *in6m; 1380 struct sockaddr_in6 *dst6; 1381 1382 crit_enter(); /* needs to protect static "ro" below. */ 1383 1384 /* 1385 * Make a new reference to the packet; make sure that 1386 * the IPv6 header is actually copied, not just referenced, 1387 * so that ip6_output() only scribbles on the copy. 1388 */ 1389 mb_copy = m_copy(m, 0, M_COPYALL); 1390 if (mb_copy && 1391 (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr))) 1392 mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr)); 1393 if (mb_copy == NULL) { 1394 crit_exit(); 1395 return; 1396 } 1397 /* set MCAST flag to the outgoing packet */ 1398 mb_copy->m_flags |= M_MCAST; 1399 1400 /* 1401 * If we sourced the packet, call ip6_output since we may devide 1402 * the packet into fragments when the packet is too big for the 1403 * outgoing interface. 1404 * Otherwise, we can simply send the packet to the interface 1405 * sending queue. 1406 */ 1407 if (m->m_pkthdr.rcvif == NULL) { 1408 struct ip6_moptions im6o; 1409 1410 im6o.im6o_multicast_ifp = ifp; 1411 /* XXX: ip6_output will override ip6->ip6_hlim */ 1412 im6o.im6o_multicast_hlim = ip6->ip6_hlim; 1413 im6o.im6o_multicast_loop = 1; 1414 error = ip6_output(mb_copy, NULL, &ro, 1415 IPV6_FORWARDING, &im6o, NULL, NULL); 1416 1417 #ifdef MRT6DEBUG 1418 if (mrt6debug & DEBUG_XMIT) 1419 log(LOG_DEBUG, "phyint_send on mif %d err %d\n", 1420 mifp - mif6table, error); 1421 #endif 1422 crit_exit(); 1423 return; 1424 } 1425 1426 /* 1427 * If we belong to the destination multicast group 1428 * on the outgoing interface, loop back a copy. 1429 */ 1430 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 1431 in6m = IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp); 1432 if (in6m != NULL) { 1433 dst6->sin6_len = sizeof(struct sockaddr_in6); 1434 dst6->sin6_family = AF_INET6; 1435 dst6->sin6_addr = ip6->ip6_dst; 1436 ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst); 1437 } 1438 /* 1439 * Put the packet into the sending queue of the outgoing interface 1440 * if it would fit in the MTU of the interface. 1441 */ 1442 if (mb_copy->m_pkthdr.len <= ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) { 1443 dst6->sin6_len = sizeof(struct sockaddr_in6); 1444 dst6->sin6_family = AF_INET6; 1445 dst6->sin6_addr = ip6->ip6_dst; 1446 /* 1447 * We just call if_output instead of nd6_output here, since 1448 * we need no ND for a multicast forwarded packet...right? 1449 */ 1450 error = ifp->if_output(ifp, mb_copy, 1451 (struct sockaddr *)&ro.ro_dst, NULL); 1452 #ifdef MRT6DEBUG 1453 if (mrt6debug & DEBUG_XMIT) 1454 log(LOG_DEBUG, "phyint_send on mif %d err %d\n", 1455 mifp - mif6table, error); 1456 #endif 1457 } else { 1458 #ifdef MULTICAST_PMTUD 1459 icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 1460 #else 1461 #ifdef MRT6DEBUG 1462 if (mrt6debug & DEBUG_XMIT) 1463 log(LOG_DEBUG, 1464 "phyint_send: packet too big on %s o %s g %s" 1465 " size %d(discarded)\n", 1466 if_name(ifp), 1467 ip6_sprintf(&ip6->ip6_src), 1468 ip6_sprintf(&ip6->ip6_dst), 1469 mb_copy->m_pkthdr.len); 1470 #endif /* MRT6DEBUG */ 1471 m_freem(mb_copy); /* simply discard the packet */ 1472 #endif 1473 } 1474 1475 crit_exit(); 1476 } 1477 1478 static int 1479 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m) 1480 { 1481 struct mbuf *mm; 1482 int i, len = m->m_pkthdr.len; 1483 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; 1484 struct mrt6msg *im6; 1485 1486 #ifdef MRT6DEBUG 1487 if (mrt6debug) 1488 log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n", 1489 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst)); 1490 #endif 1491 ++pim6stat.pim6s_snd_registers; 1492 1493 /* Make a copy of the packet to send to the user level process */ 1494 MGETHDR(mm, M_NOWAIT, MT_HEADER); 1495 if (mm == NULL) 1496 return ENOBUFS; 1497 mm->m_pkthdr.rcvif = NULL; 1498 mm->m_data += max_linkhdr; 1499 mm->m_len = sizeof(struct ip6_hdr); 1500 1501 if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1502 m_freem(mm); 1503 return ENOBUFS; 1504 } 1505 i = MHLEN - M_LEADINGSPACE(mm); 1506 if (i > len) 1507 i = len; 1508 mm = m_pullup(mm, i); 1509 if (mm == NULL) 1510 return ENOBUFS; 1511 /* TODO: check it! */ 1512 mm->m_pkthdr.len = len + sizeof(struct ip6_hdr); 1513 1514 /* 1515 * Send message to routing daemon 1516 */ 1517 sin6.sin6_addr = ip6->ip6_src; 1518 1519 im6 = mtod(mm, struct mrt6msg *); 1520 im6->im6_msgtype = MRT6MSG_WHOLEPKT; 1521 im6->im6_mbz = 0; 1522 1523 im6->im6_mif = mif - mif6table; 1524 1525 /* iif info is not given for reg. encap.n */ 1526 mrt6stat.mrt6s_upcalls++; 1527 1528 if (socket_send(ip6_mrouter, mm, &sin6) < 0) { 1529 #ifdef MRT6DEBUG 1530 if (mrt6debug) 1531 log(LOG_WARNING, 1532 "register_send: ip6_mrouter socket queue full\n"); 1533 #endif 1534 ++mrt6stat.mrt6s_upq_sockfull; 1535 return ENOBUFS; 1536 } 1537 return 0; 1538 } 1539 1540 /* 1541 * PIM sparse mode hook 1542 * Receives the pim control messages, and passes them up to the listening 1543 * socket, using rip6_input. 1544 * The only message processed is the REGISTER pim message; the pim header 1545 * is stripped off, and the inner packet is passed to register_mforward. 1546 */ 1547 int 1548 pim6_input(struct mbuf **mp, int *offp, int proto) 1549 { 1550 struct pim *pim; /* pointer to a pim struct */ 1551 struct ip6_hdr *ip6; 1552 int pimlen; 1553 struct mbuf *m = *mp; 1554 int minlen; 1555 int off = *offp; 1556 1557 ++pim6stat.pim6s_rcv_total; 1558 1559 ip6 = mtod(m, struct ip6_hdr *); 1560 pimlen = m->m_pkthdr.len - *offp; 1561 1562 /* 1563 * Validate lengths 1564 */ 1565 if (pimlen < PIM_MINLEN) { 1566 ++pim6stat.pim6s_rcv_tooshort; 1567 #ifdef MRT6DEBUG 1568 if (mrt6debug & DEBUG_PIM) 1569 log(LOG_DEBUG,"pim6_input: PIM packet too short\n"); 1570 #endif 1571 m_freem(m); 1572 return (IPPROTO_DONE); 1573 } 1574 1575 /* 1576 * if the packet is at least as big as a REGISTER, go ahead 1577 * and grab the PIM REGISTER header size, to avoid another 1578 * possible m_pullup() later. 1579 * 1580 * PIM_MINLEN == pimhdr + u_int32 == 8 1581 * PIM6_REG_MINLEN == pimhdr + reghdr + eip6hdr == 4 + 4 + 40 1582 */ 1583 minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN; 1584 1585 /* 1586 * Make sure that the IP6 and PIM headers in contiguous memory, and 1587 * possibly the PIM REGISTER header 1588 */ 1589 #ifndef PULLDOWN_TEST 1590 IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE); 1591 /* adjust pointer */ 1592 ip6 = mtod(m, struct ip6_hdr *); 1593 1594 /* adjust mbuf to point to the PIM header */ 1595 pim = (struct pim *)((caddr_t)ip6 + off); 1596 #else 1597 IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen); 1598 if (pim == NULL) { 1599 pim6stat.pim6s_rcv_tooshort++; 1600 return IPPROTO_DONE; 1601 } 1602 #endif 1603 1604 #define PIM6_CHECKSUM 1605 #ifdef PIM6_CHECKSUM 1606 { 1607 int cksumlen; 1608 1609 /* 1610 * Validate checksum. 1611 * If PIM REGISTER, exclude the data packet 1612 */ 1613 if (pim->pim_type == PIM_REGISTER) 1614 cksumlen = PIM_MINLEN; 1615 else 1616 cksumlen = pimlen; 1617 1618 if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) { 1619 ++pim6stat.pim6s_rcv_badsum; 1620 #ifdef MRT6DEBUG 1621 if (mrt6debug & DEBUG_PIM) 1622 log(LOG_DEBUG, 1623 "pim6_input: invalid checksum\n"); 1624 #endif 1625 m_freem(m); 1626 return (IPPROTO_DONE); 1627 } 1628 } 1629 #endif /* PIM_CHECKSUM */ 1630 1631 /* PIM version check */ 1632 if (pim->pim_ver != PIM_VERSION) { 1633 ++pim6stat.pim6s_rcv_badversion; 1634 #ifdef MRT6DEBUG 1635 log(LOG_ERR, 1636 "pim6_input: incorrect version %d, expecting %d\n", 1637 pim->pim_ver, PIM_VERSION); 1638 #endif 1639 m_freem(m); 1640 return (IPPROTO_DONE); 1641 } 1642 1643 if (pim->pim_type == PIM_REGISTER) { 1644 /* 1645 * since this is a REGISTER, we'll make a copy of the register 1646 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the 1647 * routing daemon. 1648 */ 1649 static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 }; 1650 1651 struct mbuf *mcp; 1652 struct ip6_hdr *eip6; 1653 u_int32_t *reghdr; 1654 1655 ++pim6stat.pim6s_rcv_registers; 1656 1657 if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) { 1658 #ifdef MRT6DEBUG 1659 if (mrt6debug & DEBUG_PIM) 1660 log(LOG_DEBUG, 1661 "pim6_input: register mif not set: %d\n", 1662 reg_mif_num); 1663 #endif 1664 m_freem(m); 1665 return (IPPROTO_DONE); 1666 } 1667 1668 reghdr = (u_int32_t *)(pim + 1); 1669 1670 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 1671 goto pim6_input_to_daemon; 1672 1673 /* 1674 * Validate length 1675 */ 1676 if (pimlen < PIM6_REG_MINLEN) { 1677 ++pim6stat.pim6s_rcv_tooshort; 1678 ++pim6stat.pim6s_rcv_badregisters; 1679 #ifdef MRT6DEBUG 1680 log(LOG_ERR, 1681 "pim6_input: register packet size too " 1682 "small %d from %s\n", 1683 pimlen, ip6_sprintf(&ip6->ip6_src)); 1684 #endif 1685 m_freem(m); 1686 return (IPPROTO_DONE); 1687 } 1688 1689 eip6 = (struct ip6_hdr *) (reghdr + 1); 1690 #ifdef MRT6DEBUG 1691 if (mrt6debug & DEBUG_PIM) 1692 log(LOG_DEBUG, 1693 "pim6_input[register], eip6: %s -> %s, " 1694 "eip6 plen %d\n", 1695 ip6_sprintf(&eip6->ip6_src), 1696 ip6_sprintf(&eip6->ip6_dst), 1697 ntohs(eip6->ip6_plen)); 1698 #endif 1699 1700 /* verify the version number of the inner packet */ 1701 if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1702 ++pim6stat.pim6s_rcv_badregisters; 1703 #ifdef MRT6DEBUG 1704 log(LOG_DEBUG, "pim6_input: invalid IP version (%d) " 1705 "of the inner packet\n", 1706 (eip6->ip6_vfc & IPV6_VERSION)); 1707 #endif 1708 m_freem(m); 1709 return (IPPROTO_NONE); 1710 } 1711 1712 /* verify the inner packet is destined to a mcast group */ 1713 if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) { 1714 ++pim6stat.pim6s_rcv_badregisters; 1715 #ifdef MRT6DEBUG 1716 if (mrt6debug & DEBUG_PIM) 1717 log(LOG_DEBUG, 1718 "pim6_input: inner packet of register " 1719 "is not multicast %s\n", 1720 ip6_sprintf(&eip6->ip6_dst)); 1721 #endif 1722 m_freem(m); 1723 return (IPPROTO_DONE); 1724 } 1725 1726 /* 1727 * make a copy of the whole header to pass to the daemon later. 1728 */ 1729 mcp = m_copy(m, 0, off + PIM6_REG_MINLEN); 1730 if (mcp == NULL) { 1731 #ifdef MRT6DEBUG 1732 log(LOG_ERR, 1733 "pim6_input: pim register: " 1734 "could not copy register head\n"); 1735 #endif 1736 m_freem(m); 1737 return (IPPROTO_DONE); 1738 } 1739 1740 /* 1741 * forward the inner ip6 packet; point m_data at the inner ip6. 1742 */ 1743 m_adj(m, off + PIM_MINLEN); 1744 #ifdef MRT6DEBUG 1745 if (mrt6debug & DEBUG_PIM) { 1746 log(LOG_DEBUG, 1747 "pim6_input: forwarding decapsulated register: " 1748 "src %s, dst %s, mif %d\n", 1749 ip6_sprintf(&eip6->ip6_src), 1750 ip6_sprintf(&eip6->ip6_dst), 1751 reg_mif_num); 1752 } 1753 #endif 1754 1755 if_simloop(mif6table[reg_mif_num].m6_ifp, m, 1756 dst.sin6_family, 0); 1757 1758 /* prepare the register head to send to the mrouting daemon */ 1759 m = mcp; 1760 } 1761 1762 /* 1763 * Pass the PIM message up to the daemon; if it is a register message 1764 * pass the 'head' only up to the daemon. This includes the 1765 * encapsulator ip6 header, pim header, register header and the 1766 * encapsulated ip6 header. 1767 */ 1768 pim6_input_to_daemon: 1769 rip6_input(&m, offp, proto); 1770 return (IPPROTO_DONE); 1771 } 1772