ip_output.c revision 105199
1/* 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University 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 REGENTS 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 REGENTS 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 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 34 * $FreeBSD: head/sys/netinet/ip_output.c 105199 2002-10-16 02:25:05Z sam $ 35 */ 36 37#define _IP_VHL 38 39#include "opt_ipfw.h" 40#include "opt_ipdn.h" 41#include "opt_ipdivert.h" 42#include "opt_ipfilter.h" 43#include "opt_ipsec.h" 44#include "opt_mac.h" 45#include "opt_pfil_hooks.h" 46#include "opt_random_ip_id.h" 47 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/kernel.h> 51#include <sys/mac.h> 52#include <sys/malloc.h> 53#include <sys/mbuf.h> 54#include <sys/protosw.h> 55#include <sys/socket.h> 56#include <sys/socketvar.h> 57 58#include <net/if.h> 59#include <net/route.h> 60 61#include <netinet/in.h> 62#include <netinet/in_systm.h> 63#include <netinet/ip.h> 64#include <netinet/in_pcb.h> 65#include <netinet/in_var.h> 66#include <netinet/ip_var.h> 67 68#include <machine/in_cksum.h> 69 70static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 71 72#ifdef IPSEC 73#include <netinet6/ipsec.h> 74#include <netkey/key.h> 75#ifdef IPSEC_DEBUG 76#include <netkey/key_debug.h> 77#else 78#define KEYDEBUG(lev,arg) 79#endif 80#endif /*IPSEC*/ 81 82#ifdef FAST_IPSEC 83#include <netipsec/ipsec.h> 84#include <netipsec/xform.h> 85#include <netipsec/key.h> 86#endif /*FAST_IPSEC*/ 87 88#include <netinet/ip_fw.h> 89#include <netinet/ip_dummynet.h> 90 91#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 92 x, (ntohl(a.s_addr)>>24)&0xFF,\ 93 (ntohl(a.s_addr)>>16)&0xFF,\ 94 (ntohl(a.s_addr)>>8)&0xFF,\ 95 (ntohl(a.s_addr))&0xFF, y); 96 97u_short ip_id; 98 99static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 100static struct ifnet *ip_multicast_if(struct in_addr *, int *); 101static void ip_mloopback 102 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 103static int ip_getmoptions 104 (struct sockopt *, struct ip_moptions *); 105static int ip_pcbopts(int, struct mbuf **, struct mbuf *); 106static int ip_setmoptions 107 (struct sockopt *, struct ip_moptions **); 108 109int ip_optcopy(struct ip *, struct ip *); 110 111 112extern struct protosw inetsw[]; 113 114/* 115 * IP output. The packet in mbuf chain m contains a skeletal IP 116 * header (with len, off, ttl, proto, tos, src, dst). 117 * The mbuf chain containing the packet will be freed. 118 * The mbuf opt, if present, will not be freed. 119 */ 120int 121ip_output(m0, opt, ro, flags, imo, inp) 122 struct mbuf *m0; 123 struct mbuf *opt; 124 struct route *ro; 125 int flags; 126 struct ip_moptions *imo; 127 struct inpcb *inp; 128{ 129 struct ip *ip, *mhip; 130 struct ifnet *ifp = NULL; /* keep compiler happy */ 131 struct mbuf *m; 132 int hlen = sizeof (struct ip); 133 int len, off, error = 0; 134 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 135 struct in_ifaddr *ia = NULL; 136 int isbroadcast, sw_csum; 137 struct in_addr pkt_dst; 138#ifdef IPSEC 139 struct route iproute; 140 struct secpolicy *sp = NULL; 141 struct socket *so = inp ? inp->inp_socket : NULL; 142#endif 143#ifdef FAST_IPSEC 144 struct route iproute; 145 struct m_tag *mtag; 146 struct secpolicy *sp = NULL; 147 struct tdb_ident *tdbi; 148 int s; 149#endif /* FAST_IPSEC */ 150 struct ip_fw_args args; 151 int src_was_INADDR_ANY = 0; /* as the name says... */ 152#ifdef PFIL_HOOKS 153 struct packet_filter_hook *pfh; 154 struct mbuf *m1; 155 int rv; 156#endif /* PFIL_HOOKS */ 157 158 args.eh = NULL; 159 args.rule = NULL; 160 args.next_hop = NULL; 161 args.divert_rule = 0; /* divert cookie */ 162 163 /* Grab info from MT_TAG mbufs prepended to the chain. */ 164 for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) { 165 switch(m0->_m_tag_id) { 166 default: 167 printf("ip_output: unrecognised MT_TAG tag %d\n", 168 m0->_m_tag_id); 169 break; 170 171 case PACKET_TAG_DUMMYNET: 172 /* 173 * the packet was already tagged, so part of the 174 * processing was already done, and we need to go down. 175 * Get parameters from the header. 176 */ 177 args.rule = ((struct dn_pkt *)m0)->rule; 178 opt = NULL ; 179 ro = & ( ((struct dn_pkt *)m0)->ro ) ; 180 imo = NULL ; 181 dst = ((struct dn_pkt *)m0)->dn_dst ; 182 ifp = ((struct dn_pkt *)m0)->ifp ; 183 flags = ((struct dn_pkt *)m0)->flags ; 184 break; 185 186 case PACKET_TAG_DIVERT: 187 args.divert_rule = (intptr_t)m0->m_data & 0xffff; 188 break; 189 190 case PACKET_TAG_IPFORWARD: 191 args.next_hop = (struct sockaddr_in *)m0->m_data; 192 break; 193 } 194 } 195 m = m0; 196 197 KASSERT(!m || (m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR")); 198#ifndef FAST_IPSEC 199 KASSERT(ro != NULL, ("ip_output: no route, proto %d", 200 mtod(m, struct ip *)->ip_p)); 201#endif 202 203 if (args.rule != NULL) { /* dummynet already saw us */ 204 ip = mtod(m, struct ip *); 205 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ; 206 if (ro->ro_rt) 207 ia = ifatoia(ro->ro_rt->rt_ifa); 208 goto sendit; 209 } 210 211 if (opt) { 212 len = 0; 213 m = ip_insertoptions(m, opt, &len); 214 if (len != 0) 215 hlen = len; 216 } 217 ip = mtod(m, struct ip *); 218 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 219 220 /* 221 * Fill in IP header. 222 */ 223 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 224 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2); 225 ip->ip_off &= IP_DF; 226#ifdef RANDOM_IP_ID 227 ip->ip_id = ip_randomid(); 228#else 229 ip->ip_id = htons(ip_id++); 230#endif 231 ipstat.ips_localout++; 232 } else { 233 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 234 } 235 236#ifdef FAST_IPSEC 237 if (ro == NULL) { 238 ro = &iproute; 239 bzero(ro, sizeof (*ro)); 240 } 241#endif /* FAST_IPSEC */ 242 dst = (struct sockaddr_in *)&ro->ro_dst; 243 /* 244 * If there is a cached route, 245 * check that it is to the same destination 246 * and is still up. If not, free it and try again. 247 * The address family should also be checked in case of sharing the 248 * cache with IPv6. 249 */ 250 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 251 dst->sin_family != AF_INET || 252 dst->sin_addr.s_addr != pkt_dst.s_addr)) { 253 RTFREE(ro->ro_rt); 254 ro->ro_rt = (struct rtentry *)0; 255 } 256 if (ro->ro_rt == 0) { 257 bzero(dst, sizeof(*dst)); 258 dst->sin_family = AF_INET; 259 dst->sin_len = sizeof(*dst); 260 dst->sin_addr = pkt_dst; 261 } 262 /* 263 * If routing to interface only, 264 * short circuit routing lookup. 265 */ 266 if (flags & IP_ROUTETOIF) { 267 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 268 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 269 ipstat.ips_noroute++; 270 error = ENETUNREACH; 271 goto bad; 272 } 273 ifp = ia->ia_ifp; 274 ip->ip_ttl = 1; 275 isbroadcast = in_broadcast(dst->sin_addr, ifp); 276 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 277 imo != NULL && imo->imo_multicast_ifp != NULL) { 278 /* 279 * Bypass the normal routing lookup for multicast 280 * packets if the interface is specified. 281 */ 282 ifp = imo->imo_multicast_ifp; 283 IFP_TO_IA(ifp, ia); 284 isbroadcast = 0; /* fool gcc */ 285 } else { 286 /* 287 * If this is the case, we probably don't want to allocate 288 * a protocol-cloned route since we didn't get one from the 289 * ULP. This lets TCP do its thing, while not burdening 290 * forwarding or ICMP with the overhead of cloning a route. 291 * Of course, we still want to do any cloning requested by 292 * the link layer, as this is probably required in all cases 293 * for correct operation (as it is for ARP). 294 */ 295 if (ro->ro_rt == 0) 296 rtalloc_ign(ro, RTF_PRCLONING); 297 if (ro->ro_rt == 0) { 298 ipstat.ips_noroute++; 299 error = EHOSTUNREACH; 300 goto bad; 301 } 302 ia = ifatoia(ro->ro_rt->rt_ifa); 303 ifp = ro->ro_rt->rt_ifp; 304 ro->ro_rt->rt_use++; 305 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 306 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 307 if (ro->ro_rt->rt_flags & RTF_HOST) 308 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 309 else 310 isbroadcast = in_broadcast(dst->sin_addr, ifp); 311 } 312 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 313 struct in_multi *inm; 314 315 m->m_flags |= M_MCAST; 316 /* 317 * IP destination address is multicast. Make sure "dst" 318 * still points to the address in "ro". (It may have been 319 * changed to point to a gateway address, above.) 320 */ 321 dst = (struct sockaddr_in *)&ro->ro_dst; 322 /* 323 * See if the caller provided any multicast options 324 */ 325 if (imo != NULL) { 326 ip->ip_ttl = imo->imo_multicast_ttl; 327 if (imo->imo_multicast_vif != -1) 328 ip->ip_src.s_addr = 329 ip_mcast_src(imo->imo_multicast_vif); 330 } else 331 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 332 /* 333 * Confirm that the outgoing interface supports multicast. 334 */ 335 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 336 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 337 ipstat.ips_noroute++; 338 error = ENETUNREACH; 339 goto bad; 340 } 341 } 342 /* 343 * If source address not specified yet, use address 344 * of outgoing interface. 345 */ 346 if (ip->ip_src.s_addr == INADDR_ANY) { 347 /* Interface may have no addresses. */ 348 if (ia != NULL) 349 ip->ip_src = IA_SIN(ia)->sin_addr; 350 } 351 352 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 353 /* 354 * XXX 355 * delayed checksums are not currently 356 * compatible with IP multicast routing 357 */ 358 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 359 in_delayed_cksum(m); 360 m->m_pkthdr.csum_flags &= 361 ~CSUM_DELAY_DATA; 362 } 363 } 364 IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 365 if (inm != NULL && 366 (imo == NULL || imo->imo_multicast_loop)) { 367 /* 368 * If we belong to the destination multicast group 369 * on the outgoing interface, and the caller did not 370 * forbid loopback, loop back a copy. 371 */ 372 ip_mloopback(ifp, m, dst, hlen); 373 } 374 else { 375 /* 376 * If we are acting as a multicast router, perform 377 * multicast forwarding as if the packet had just 378 * arrived on the interface to which we are about 379 * to send. The multicast forwarding function 380 * recursively calls this function, using the 381 * IP_FORWARDING flag to prevent infinite recursion. 382 * 383 * Multicasts that are looped back by ip_mloopback(), 384 * above, will be forwarded by the ip_input() routine, 385 * if necessary. 386 */ 387 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 388 /* 389 * Check if rsvp daemon is running. If not, don't 390 * set ip_moptions. This ensures that the packet 391 * is multicast and not just sent down one link 392 * as prescribed by rsvpd. 393 */ 394 if (!rsvp_on) 395 imo = NULL; 396 if (ip_mforward(ip, ifp, m, imo) != 0) { 397 m_freem(m); 398 goto done; 399 } 400 } 401 } 402 403 /* 404 * Multicasts with a time-to-live of zero may be looped- 405 * back, above, but must not be transmitted on a network. 406 * Also, multicasts addressed to the loopback interface 407 * are not sent -- the above call to ip_mloopback() will 408 * loop back a copy if this host actually belongs to the 409 * destination group on the loopback interface. 410 */ 411 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 412 m_freem(m); 413 goto done; 414 } 415 416 goto sendit; 417 } 418#ifndef notdef 419 /* 420 * If the source address is not specified yet, use the address 421 * of the outoing interface. In case, keep note we did that, so 422 * if the the firewall changes the next-hop causing the output 423 * interface to change, we can fix that. 424 */ 425 if (ip->ip_src.s_addr == INADDR_ANY) { 426 /* Interface may have no addresses. */ 427 if (ia != NULL) { 428 ip->ip_src = IA_SIN(ia)->sin_addr; 429 src_was_INADDR_ANY = 1; 430 } 431 } 432#endif /* notdef */ 433 /* 434 * Verify that we have any chance at all of being able to queue 435 * the packet or packet fragments 436 */ 437 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 438 ifp->if_snd.ifq_maxlen) { 439 error = ENOBUFS; 440 ipstat.ips_odropped++; 441 goto bad; 442 } 443 444 /* 445 * Look for broadcast address and 446 * verify user is allowed to send 447 * such a packet. 448 */ 449 if (isbroadcast) { 450 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 451 error = EADDRNOTAVAIL; 452 goto bad; 453 } 454 if ((flags & IP_ALLOWBROADCAST) == 0) { 455 error = EACCES; 456 goto bad; 457 } 458 /* don't allow broadcast messages to be fragmented */ 459 if ((u_short)ip->ip_len > ifp->if_mtu) { 460 error = EMSGSIZE; 461 goto bad; 462 } 463 m->m_flags |= M_BCAST; 464 } else { 465 m->m_flags &= ~M_BCAST; 466 } 467 468sendit: 469#ifdef IPSEC 470 /* get SP for this packet */ 471 if (so == NULL) 472 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 473 else 474 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 475 476 if (sp == NULL) { 477 ipsecstat.out_inval++; 478 goto bad; 479 } 480 481 error = 0; 482 483 /* check policy */ 484 switch (sp->policy) { 485 case IPSEC_POLICY_DISCARD: 486 /* 487 * This packet is just discarded. 488 */ 489 ipsecstat.out_polvio++; 490 goto bad; 491 492 case IPSEC_POLICY_BYPASS: 493 case IPSEC_POLICY_NONE: 494 /* no need to do IPsec. */ 495 goto skip_ipsec; 496 497 case IPSEC_POLICY_IPSEC: 498 if (sp->req == NULL) { 499 /* acquire a policy */ 500 error = key_spdacquire(sp); 501 goto bad; 502 } 503 break; 504 505 case IPSEC_POLICY_ENTRUST: 506 default: 507 printf("ip_output: Invalid policy found. %d\n", sp->policy); 508 } 509 { 510 struct ipsec_output_state state; 511 bzero(&state, sizeof(state)); 512 state.m = m; 513 if (flags & IP_ROUTETOIF) { 514 state.ro = &iproute; 515 bzero(&iproute, sizeof(iproute)); 516 } else 517 state.ro = ro; 518 state.dst = (struct sockaddr *)dst; 519 520 ip->ip_sum = 0; 521 522 /* 523 * XXX 524 * delayed checksums are not currently compatible with IPsec 525 */ 526 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 527 in_delayed_cksum(m); 528 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 529 } 530 531 ip->ip_len = htons(ip->ip_len); 532 ip->ip_off = htons(ip->ip_off); 533 534 error = ipsec4_output(&state, sp, flags); 535 536 m = state.m; 537 if (flags & IP_ROUTETOIF) { 538 /* 539 * if we have tunnel mode SA, we may need to ignore 540 * IP_ROUTETOIF. 541 */ 542 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 543 flags &= ~IP_ROUTETOIF; 544 ro = state.ro; 545 } 546 } else 547 ro = state.ro; 548 dst = (struct sockaddr_in *)state.dst; 549 if (error) { 550 /* mbuf is already reclaimed in ipsec4_output. */ 551 m0 = NULL; 552 switch (error) { 553 case EHOSTUNREACH: 554 case ENETUNREACH: 555 case EMSGSIZE: 556 case ENOBUFS: 557 case ENOMEM: 558 break; 559 default: 560 printf("ip4_output (ipsec): error code %d\n", error); 561 /*fall through*/ 562 case ENOENT: 563 /* don't show these error codes to the user */ 564 error = 0; 565 break; 566 } 567 goto bad; 568 } 569 } 570 571 /* be sure to update variables that are affected by ipsec4_output() */ 572 ip = mtod(m, struct ip *); 573#ifdef _IP_VHL 574 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 575#else 576 hlen = ip->ip_hl << 2; 577#endif 578 if (ro->ro_rt == NULL) { 579 if ((flags & IP_ROUTETOIF) == 0) { 580 printf("ip_output: " 581 "can't update route after IPsec processing\n"); 582 error = EHOSTUNREACH; /*XXX*/ 583 goto bad; 584 } 585 } else { 586 ia = ifatoia(ro->ro_rt->rt_ifa); 587 ifp = ro->ro_rt->rt_ifp; 588 } 589 590 /* make it flipped, again. */ 591 ip->ip_len = ntohs(ip->ip_len); 592 ip->ip_off = ntohs(ip->ip_off); 593skip_ipsec: 594#endif /*IPSEC*/ 595#ifdef FAST_IPSEC 596 /* 597 * Check the security policy (SP) for the packet and, if 598 * required, do IPsec-related processing. There are two 599 * cases here; the first time a packet is sent through 600 * it will be untagged and handled by ipsec4_checkpolicy. 601 * If the packet is resubmitted to ip_output (e.g. after 602 * AH, ESP, etc. processing), there will be a tag to bypass 603 * the lookup and related policy checking. 604 */ 605 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 606 s = splnet(); 607 if (mtag != NULL) { 608 tdbi = (struct tdb_ident *)(mtag + 1); 609 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 610 if (sp == NULL) 611 error = -EINVAL; /* force silent drop */ 612 m_tag_delete(m, mtag); 613 } else { 614 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 615 &error, inp); 616 } 617 /* 618 * There are four return cases: 619 * sp != NULL apply IPsec policy 620 * sp == NULL, error == 0 no IPsec handling needed 621 * sp == NULL, error == -EINVAL discard packet w/o error 622 * sp == NULL, error != 0 discard packet, report error 623 */ 624 if (sp != NULL) { 625 /* Loop detection, check if ipsec processing already done */ 626 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 627 for (mtag = m_tag_first(m); mtag != NULL; 628 mtag = m_tag_next(m, mtag)) { 629 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 630 continue; 631 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 632 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 633 continue; 634 /* 635 * Check if policy has an SA associated with it. 636 * This can happen when an SP has yet to acquire 637 * an SA; e.g. on first reference. If it occurs, 638 * then we let ipsec4_process_packet do its thing. 639 */ 640 if (sp->req->sav == NULL) 641 break; 642 tdbi = (struct tdb_ident *)(mtag + 1); 643 if (tdbi->spi == sp->req->sav->spi && 644 tdbi->proto == sp->req->sav->sah->saidx.proto && 645 bcmp(&tdbi->dst, &sp->spidx.dst, 646 sizeof (union sockaddr_union)) == 0) { 647 /* 648 * No IPsec processing is needed, free 649 * reference to SP. 650 * 651 * NB: null pointer to avoid free at 652 * done: below. 653 */ 654 KEY_FREESP(&sp), sp = NULL; 655 splx(s); 656 goto spd_done; 657 } 658 } 659 660 /* 661 * Do delayed checksums now because we send before 662 * this is done in the normal processing path. 663 */ 664 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 665 in_delayed_cksum(m); 666 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 667 } 668 669 ip->ip_len = htons(ip->ip_len); 670 ip->ip_off = htons(ip->ip_off); 671 672 /* NB: callee frees mbuf */ 673 error = ipsec4_process_packet(m, sp->req, flags, 0); 674 splx(s); 675 goto done; 676 } else { 677 splx(s); 678 679 if (error != 0) { 680 /* 681 * Hack: -EINVAL is used to signal that a packet 682 * should be silently discarded. This is typically 683 * because we asked key management for an SA and 684 * it was delayed (e.g. kicked up to IKE). 685 */ 686 if (error == -EINVAL) 687 error = 0; 688 goto bad; 689 } else { 690 /* No IPsec processing for this packet. */ 691 } 692#ifdef notyet 693 /* 694 * If deferred crypto processing is needed, check that 695 * the interface supports it. 696 */ 697 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 698 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 699 /* notify IPsec to do its own crypto */ 700 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 701 error = EHOSTUNREACH; 702 goto bad; 703 } 704#endif 705 } 706spd_done: 707#endif /* FAST_IPSEC */ 708 709 /* 710 * IpHack's section. 711 * - Xlate: translate packet's addr/port (NAT). 712 * - Firewall: deny/allow/etc. 713 * - Wrap: fake packet's addr/port <unimpl.> 714 * - Encapsulate: put it in another IP and send out. <unimp.> 715 */ 716#ifdef PFIL_HOOKS 717 /* 718 * Run through list of hooks for output packets. 719 */ 720 m1 = m; 721 pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 722 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 723 if (pfh->pfil_func) { 724 rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1); 725 if (rv) { 726 error = EHOSTUNREACH; 727 goto done; 728 } 729 m = m1; 730 if (m == NULL) 731 goto done; 732 ip = mtod(m, struct ip *); 733 } 734#endif /* PFIL_HOOKS */ 735 736 /* 737 * Check with the firewall... 738 * but not if we are already being fwd'd from a firewall. 739 */ 740 if (fw_enable && IPFW_LOADED && !args.next_hop) { 741 struct sockaddr_in *old = dst; 742 743 args.m = m; 744 args.next_hop = dst; 745 args.oif = ifp; 746 off = ip_fw_chk_ptr(&args); 747 m = args.m; 748 dst = args.next_hop; 749 750 /* 751 * On return we must do the following: 752 * m == NULL -> drop the pkt (old interface, deprecated) 753 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 754 * 1<=off<= 0xffff -> DIVERT 755 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 756 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 757 * dst != old -> IPFIREWALL_FORWARD 758 * off==0, dst==old -> accept 759 * If some of the above modules are not compiled in, then 760 * we should't have to check the corresponding condition 761 * (because the ipfw control socket should not accept 762 * unsupported rules), but better play safe and drop 763 * packets in case of doubt. 764 */ 765 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) { 766 if (m) 767 m_freem(m); 768 error = EACCES; 769 goto done; 770 } 771 ip = mtod(m, struct ip *); 772 if (off == 0 && dst == old) /* common case */ 773 goto pass; 774 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 775 /* 776 * pass the pkt to dummynet. Need to include 777 * pipe number, m, ifp, ro, dst because these are 778 * not recomputed in the next pass. 779 * All other parameters have been already used and 780 * so they are not needed anymore. 781 * XXX note: if the ifp or ro entry are deleted 782 * while a pkt is in dummynet, we are in trouble! 783 */ 784 args.ro = ro; 785 args.dst = dst; 786 args.flags = flags; 787 788 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, 789 &args); 790 goto done; 791 } 792#ifdef IPDIVERT 793 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 794 struct mbuf *clone = NULL; 795 796 /* Clone packet if we're doing a 'tee' */ 797 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 798 clone = m_dup(m, M_DONTWAIT); 799 800 /* 801 * XXX 802 * delayed checksums are not currently compatible 803 * with divert sockets. 804 */ 805 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 806 in_delayed_cksum(m); 807 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 808 } 809 810 /* Restore packet header fields to original values */ 811 ip->ip_len = htons(ip->ip_len); 812 ip->ip_off = htons(ip->ip_off); 813 814 /* Deliver packet to divert input routine */ 815 divert_packet(m, 0, off & 0xffff, args.divert_rule); 816 817 /* If 'tee', continue with original packet */ 818 if (clone != NULL) { 819 m = clone; 820 ip = mtod(m, struct ip *); 821 goto pass; 822 } 823 goto done; 824 } 825#endif 826 827 /* IPFIREWALL_FORWARD */ 828 /* 829 * Check dst to make sure it is directly reachable on the 830 * interface we previously thought it was. 831 * If it isn't (which may be likely in some situations) we have 832 * to re-route it (ie, find a route for the next-hop and the 833 * associated interface) and set them here. This is nested 834 * forwarding which in most cases is undesirable, except where 835 * such control is nigh impossible. So we do it here. 836 * And I'm babbling. 837 */ 838 if (off == 0 && old != dst) { /* FORWARD, dst has changed */ 839#if 0 840 /* 841 * XXX To improve readability, this block should be 842 * changed into a function call as below: 843 */ 844 error = ip_ipforward(&m, &dst, &ifp); 845 if (error) 846 goto bad; 847 if (m == NULL) /* ip_input consumed the mbuf */ 848 goto done; 849#else 850 struct in_ifaddr *ia; 851 852 /* 853 * XXX sro_fwd below is static, and a pointer 854 * to it gets passed to routines downstream. 855 * This could have surprisingly bad results in 856 * practice, because its content is overwritten 857 * by subsequent packets. 858 */ 859 /* There must be a better way to do this next line... */ 860 static struct route sro_fwd; 861 struct route *ro_fwd = &sro_fwd; 862 863#if 0 864 print_ip("IPFIREWALL_FORWARD: New dst ip: ", 865 dst->sin_addr, "\n"); 866#endif 867 868 /* 869 * We need to figure out if we have been forwarded 870 * to a local socket. If so, then we should somehow 871 * "loop back" to ip_input, and get directed to the 872 * PCB as if we had received this packet. This is 873 * because it may be dificult to identify the packets 874 * you want to forward until they are being output 875 * and have selected an interface. (e.g. locally 876 * initiated packets) If we used the loopback inteface, 877 * we would not be able to control what happens 878 * as the packet runs through ip_input() as 879 * it is done through a ISR. 880 */ 881 LIST_FOREACH(ia, 882 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 883 /* 884 * If the addr to forward to is one 885 * of ours, we pretend to 886 * be the destination for this packet. 887 */ 888 if (IA_SIN(ia)->sin_addr.s_addr == 889 dst->sin_addr.s_addr) 890 break; 891 } 892 if (ia) { /* tell ip_input "dont filter" */ 893 struct m_hdr tag; 894 895 tag.mh_type = MT_TAG; 896 tag.mh_flags = PACKET_TAG_IPFORWARD; 897 tag.mh_data = (caddr_t)args.next_hop; 898 tag.mh_next = m; 899 900 if (m->m_pkthdr.rcvif == NULL) 901 m->m_pkthdr.rcvif = ifunit("lo0"); 902 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 903 m->m_pkthdr.csum_flags |= 904 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 905 m0->m_pkthdr.csum_data = 0xffff; 906 } 907 m->m_pkthdr.csum_flags |= 908 CSUM_IP_CHECKED | CSUM_IP_VALID; 909 ip->ip_len = htons(ip->ip_len); 910 ip->ip_off = htons(ip->ip_off); 911 ip_input((struct mbuf *)&tag); 912 goto done; 913 } 914 /* Some of the logic for this was 915 * nicked from above. 916 * 917 * This rewrites the cached route in a local PCB. 918 * Is this what we want to do? 919 */ 920 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 921 922 ro_fwd->ro_rt = 0; 923 rtalloc_ign(ro_fwd, RTF_PRCLONING); 924 925 if (ro_fwd->ro_rt == 0) { 926 ipstat.ips_noroute++; 927 error = EHOSTUNREACH; 928 goto bad; 929 } 930 931 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 932 ifp = ro_fwd->ro_rt->rt_ifp; 933 ro_fwd->ro_rt->rt_use++; 934 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 935 dst = (struct sockaddr_in *) 936 ro_fwd->ro_rt->rt_gateway; 937 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 938 isbroadcast = 939 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 940 else 941 isbroadcast = in_broadcast(dst->sin_addr, ifp); 942 if (ro->ro_rt) 943 RTFREE(ro->ro_rt); 944 ro->ro_rt = ro_fwd->ro_rt; 945 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 946 947#endif /* ... block to be put into a function */ 948 /* 949 * If we added a default src ip earlier, 950 * which would have been gotten from the-then 951 * interface, do it again, from the new one. 952 */ 953 if (src_was_INADDR_ANY) 954 ip->ip_src = IA_SIN(ia)->sin_addr; 955 goto pass ; 956 } 957 958 /* 959 * if we get here, none of the above matches, and 960 * we have to drop the pkt 961 */ 962 m_freem(m); 963 error = EACCES; /* not sure this is the right error msg */ 964 goto done; 965 } 966 967pass: 968 /* 127/8 must not appear on wire - RFC1122. */ 969 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 970 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 971 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 972 ipstat.ips_badaddr++; 973 error = EADDRNOTAVAIL; 974 goto bad; 975 } 976 } 977 978 m->m_pkthdr.csum_flags |= CSUM_IP; 979 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 980 if (sw_csum & CSUM_DELAY_DATA) { 981 in_delayed_cksum(m); 982 sw_csum &= ~CSUM_DELAY_DATA; 983 } 984 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 985 986 /* 987 * If small enough for interface, or the interface will take 988 * care of the fragmentation for us, can just send directly. 989 */ 990 if ((u_short)ip->ip_len <= ifp->if_mtu || 991 ifp->if_hwassist & CSUM_FRAGMENT) { 992 ip->ip_len = htons(ip->ip_len); 993 ip->ip_off = htons(ip->ip_off); 994 ip->ip_sum = 0; 995 if (sw_csum & CSUM_DELAY_IP) { 996 if (ip->ip_vhl == IP_VHL_BORING) { 997 ip->ip_sum = in_cksum_hdr(ip); 998 } else { 999 ip->ip_sum = in_cksum(m, hlen); 1000 } 1001 } 1002 1003 /* Record statistics for this interface address. */ 1004 if (!(flags & IP_FORWARDING) && ia) { 1005 ia->ia_ifa.if_opackets++; 1006 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1007 } 1008 1009#ifdef IPSEC 1010 /* clean ipsec history once it goes out of the node */ 1011 ipsec_delaux(m); 1012#endif 1013 1014 error = (*ifp->if_output)(ifp, m, 1015 (struct sockaddr *)dst, ro->ro_rt); 1016 goto done; 1017 } 1018 /* 1019 * Too large for interface; fragment if possible. 1020 * Must be able to put at least 8 bytes per fragment. 1021 */ 1022 if (ip->ip_off & IP_DF) { 1023 error = EMSGSIZE; 1024 /* 1025 * This case can happen if the user changed the MTU 1026 * of an interface after enabling IP on it. Because 1027 * most netifs don't keep track of routes pointing to 1028 * them, there is no way for one to update all its 1029 * routes when the MTU is changed. 1030 */ 1031 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 1032 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 1033 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 1034 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 1035 } 1036 ipstat.ips_cantfrag++; 1037 goto bad; 1038 } 1039 len = (ifp->if_mtu - hlen) &~ 7; 1040 if (len < 8) { 1041 error = EMSGSIZE; 1042 goto bad; 1043 } 1044 1045 /* 1046 * if the interface will not calculate checksums on 1047 * fragmented packets, then do it here. 1048 */ 1049 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 1050 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 1051 in_delayed_cksum(m); 1052 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1053 } 1054 1055 if (len > PAGE_SIZE) { 1056 /* 1057 * Fragement large datagrams such that each segment 1058 * contains a multiple of PAGE_SIZE amount of data, 1059 * plus headers. This enables a receiver to perform 1060 * page-flipping zero-copy optimizations. 1061 */ 1062 1063 int newlen; 1064 struct mbuf *mtmp; 1065 1066 for (mtmp = m, off = 0; 1067 mtmp && ((off + mtmp->m_len) <= ifp->if_mtu); 1068 mtmp = mtmp->m_next) { 1069 off += mtmp->m_len; 1070 } 1071 /* 1072 * firstlen (off - hlen) must be aligned on an 1073 * 8-byte boundary 1074 */ 1075 if (off < hlen) 1076 goto smart_frag_failure; 1077 off = ((off - hlen) & ~7) + hlen; 1078 newlen = (~PAGE_MASK) & ifp->if_mtu; 1079 if ((newlen + sizeof (struct ip)) > ifp->if_mtu) { 1080 /* we failed, go back the default */ 1081smart_frag_failure: 1082 newlen = len; 1083 off = hlen + len; 1084 } 1085 1086/* printf("ipfrag: len = %d, hlen = %d, mhlen = %d, newlen = %d, off = %d\n", 1087 len, hlen, sizeof (struct ip), newlen, off);*/ 1088 1089 len = newlen; 1090 1091 } else { 1092 off = hlen + len; 1093 } 1094 1095 1096 1097 { 1098 int mhlen, firstlen = off - hlen; 1099 struct mbuf **mnext = &m->m_nextpkt; 1100 int nfrags = 1; 1101 1102 /* 1103 * Loop through length of segment after first fragment, 1104 * make new header and copy data of each part and link onto chain. 1105 */ 1106 m0 = m; 1107 mhlen = sizeof (struct ip); 1108 for (; off < (u_short)ip->ip_len; off += len) { 1109 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1110 if (m == 0) { 1111 error = ENOBUFS; 1112 ipstat.ips_odropped++; 1113 goto sendorfree; 1114 } 1115 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 1116 m->m_data += max_linkhdr; 1117 mhip = mtod(m, struct ip *); 1118 *mhip = *ip; 1119 if (hlen > sizeof (struct ip)) { 1120 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 1121 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2); 1122 } 1123 m->m_len = mhlen; 1124 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 1125 if (off + len >= (u_short)ip->ip_len) 1126 len = (u_short)ip->ip_len - off; 1127 else 1128 mhip->ip_off |= IP_MF; 1129 mhip->ip_len = htons((u_short)(len + mhlen)); 1130 m->m_next = m_copy(m0, off, len); 1131 if (m->m_next == 0) { 1132 (void) m_free(m); 1133 error = ENOBUFS; /* ??? */ 1134 ipstat.ips_odropped++; 1135 goto sendorfree; 1136 } 1137 m->m_pkthdr.len = mhlen + len; 1138 m->m_pkthdr.rcvif = (struct ifnet *)0; 1139#ifdef MAC 1140 mac_create_fragment(m0, m); 1141#endif 1142 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 1143 mhip->ip_off = htons(mhip->ip_off); 1144 mhip->ip_sum = 0; 1145 if (sw_csum & CSUM_DELAY_IP) { 1146 if (mhip->ip_vhl == IP_VHL_BORING) { 1147 mhip->ip_sum = in_cksum_hdr(mhip); 1148 } else { 1149 mhip->ip_sum = in_cksum(m, mhlen); 1150 } 1151 } 1152 *mnext = m; 1153 mnext = &m->m_nextpkt; 1154 nfrags++; 1155 } 1156 ipstat.ips_ofragments += nfrags; 1157 1158 /* set first/last markers for fragment chain */ 1159 m->m_flags |= M_LASTFRAG; 1160 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 1161 m0->m_pkthdr.csum_data = nfrags; 1162 1163 /* 1164 * Update first fragment by trimming what's been copied out 1165 * and updating header, then send each fragment (in order). 1166 */ 1167 m = m0; 1168 m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 1169 m->m_pkthdr.len = hlen + firstlen; 1170 ip->ip_len = htons((u_short)m->m_pkthdr.len); 1171 ip->ip_off |= IP_MF; 1172 ip->ip_off = htons(ip->ip_off); 1173 ip->ip_sum = 0; 1174 if (sw_csum & CSUM_DELAY_IP) { 1175 if (ip->ip_vhl == IP_VHL_BORING) { 1176 ip->ip_sum = in_cksum_hdr(ip); 1177 } else { 1178 ip->ip_sum = in_cksum(m, hlen); 1179 } 1180 } 1181sendorfree: 1182 for (m = m0; m; m = m0) { 1183 m0 = m->m_nextpkt; 1184 m->m_nextpkt = 0; 1185#ifdef IPSEC 1186 /* clean ipsec history once it goes out of the node */ 1187 ipsec_delaux(m); 1188#endif 1189 if (error == 0) { 1190 /* Record statistics for this interface address. */ 1191 if (ia != NULL) { 1192 ia->ia_ifa.if_opackets++; 1193 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1194 } 1195 1196 error = (*ifp->if_output)(ifp, m, 1197 (struct sockaddr *)dst, ro->ro_rt); 1198 } else 1199 m_freem(m); 1200 } 1201 1202 if (error == 0) 1203 ipstat.ips_fragmented++; 1204 } 1205done: 1206#ifdef IPSEC 1207 if (ro == &iproute && ro->ro_rt) { 1208 RTFREE(ro->ro_rt); 1209 ro->ro_rt = NULL; 1210 } 1211 if (sp != NULL) { 1212 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1213 printf("DP ip_output call free SP:%p\n", sp)); 1214 key_freesp(sp); 1215 } 1216#endif /* IPSEC */ 1217#ifdef FAST_IPSEC 1218 if (ro == &iproute && ro->ro_rt) { 1219 RTFREE(ro->ro_rt); 1220 ro->ro_rt = NULL; 1221 } 1222 if (sp != NULL) 1223 KEY_FREESP(&sp); 1224#endif /* FAST_IPSEC */ 1225 return (error); 1226bad: 1227 m_freem(m); 1228 goto done; 1229} 1230 1231void 1232in_delayed_cksum(struct mbuf *m) 1233{ 1234 struct ip *ip; 1235 u_short csum, offset; 1236 1237 ip = mtod(m, struct ip *); 1238 offset = IP_VHL_HL(ip->ip_vhl) << 2 ; 1239 csum = in_cksum_skip(m, ip->ip_len, offset); 1240 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1241 csum = 0xffff; 1242 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1243 1244 if (offset + sizeof(u_short) > m->m_len) { 1245 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1246 m->m_len, offset, ip->ip_p); 1247 /* 1248 * XXX 1249 * this shouldn't happen, but if it does, the 1250 * correct behavior may be to insert the checksum 1251 * in the existing chain instead of rearranging it. 1252 */ 1253 m = m_pullup(m, offset + sizeof(u_short)); 1254 } 1255 *(u_short *)(m->m_data + offset) = csum; 1256} 1257 1258/* 1259 * Insert IP options into preformed packet. 1260 * Adjust IP destination as required for IP source routing, 1261 * as indicated by a non-zero in_addr at the start of the options. 1262 * 1263 * XXX This routine assumes that the packet has no options in place. 1264 */ 1265static struct mbuf * 1266ip_insertoptions(m, opt, phlen) 1267 register struct mbuf *m; 1268 struct mbuf *opt; 1269 int *phlen; 1270{ 1271 register struct ipoption *p = mtod(opt, struct ipoption *); 1272 struct mbuf *n; 1273 register struct ip *ip = mtod(m, struct ip *); 1274 unsigned optlen; 1275 1276 optlen = opt->m_len - sizeof(p->ipopt_dst); 1277 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) { 1278 *phlen = 0; 1279 return (m); /* XXX should fail */ 1280 } 1281 if (p->ipopt_dst.s_addr) 1282 ip->ip_dst = p->ipopt_dst; 1283 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1284 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1285 if (n == 0) { 1286 *phlen = 0; 1287 return (m); 1288 } 1289 n->m_pkthdr.rcvif = (struct ifnet *)0; 1290#ifdef MAC 1291 mac_create_mbuf_from_mbuf(m, n); 1292#endif 1293 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1294 m->m_len -= sizeof(struct ip); 1295 m->m_data += sizeof(struct ip); 1296 n->m_next = m; 1297 m = n; 1298 m->m_len = optlen + sizeof(struct ip); 1299 m->m_data += max_linkhdr; 1300 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1301 } else { 1302 m->m_data -= optlen; 1303 m->m_len += optlen; 1304 m->m_pkthdr.len += optlen; 1305 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1306 } 1307 ip = mtod(m, struct ip *); 1308 bcopy(p->ipopt_list, ip + 1, optlen); 1309 *phlen = sizeof(struct ip) + optlen; 1310 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2); 1311 ip->ip_len += optlen; 1312 return (m); 1313} 1314 1315/* 1316 * Copy options from ip to jp, 1317 * omitting those not copied during fragmentation. 1318 */ 1319int 1320ip_optcopy(ip, jp) 1321 struct ip *ip, *jp; 1322{ 1323 register u_char *cp, *dp; 1324 int opt, optlen, cnt; 1325 1326 cp = (u_char *)(ip + 1); 1327 dp = (u_char *)(jp + 1); 1328 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1329 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1330 opt = cp[0]; 1331 if (opt == IPOPT_EOL) 1332 break; 1333 if (opt == IPOPT_NOP) { 1334 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1335 *dp++ = IPOPT_NOP; 1336 optlen = 1; 1337 continue; 1338 } 1339 1340 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1341 ("ip_optcopy: malformed ipv4 option")); 1342 optlen = cp[IPOPT_OLEN]; 1343 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1344 ("ip_optcopy: malformed ipv4 option")); 1345 1346 /* bogus lengths should have been caught by ip_dooptions */ 1347 if (optlen > cnt) 1348 optlen = cnt; 1349 if (IPOPT_COPIED(opt)) { 1350 bcopy(cp, dp, optlen); 1351 dp += optlen; 1352 } 1353 } 1354 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1355 *dp++ = IPOPT_EOL; 1356 return (optlen); 1357} 1358 1359/* 1360 * IP socket option processing. 1361 */ 1362int 1363ip_ctloutput(so, sopt) 1364 struct socket *so; 1365 struct sockopt *sopt; 1366{ 1367 struct inpcb *inp = sotoinpcb(so); 1368 int error, optval; 1369 1370 error = optval = 0; 1371 if (sopt->sopt_level != IPPROTO_IP) { 1372 return (EINVAL); 1373 } 1374 1375 switch (sopt->sopt_dir) { 1376 case SOPT_SET: 1377 switch (sopt->sopt_name) { 1378 case IP_OPTIONS: 1379#ifdef notyet 1380 case IP_RETOPTS: 1381#endif 1382 { 1383 struct mbuf *m; 1384 if (sopt->sopt_valsize > MLEN) { 1385 error = EMSGSIZE; 1386 break; 1387 } 1388 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1389 if (m == 0) { 1390 error = ENOBUFS; 1391 break; 1392 } 1393 m->m_len = sopt->sopt_valsize; 1394 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1395 m->m_len); 1396 1397 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1398 m)); 1399 } 1400 1401 case IP_TOS: 1402 case IP_TTL: 1403 case IP_RECVOPTS: 1404 case IP_RECVRETOPTS: 1405 case IP_RECVDSTADDR: 1406 case IP_RECVIF: 1407 case IP_FAITH: 1408 error = sooptcopyin(sopt, &optval, sizeof optval, 1409 sizeof optval); 1410 if (error) 1411 break; 1412 1413 switch (sopt->sopt_name) { 1414 case IP_TOS: 1415 inp->inp_ip_tos = optval; 1416 break; 1417 1418 case IP_TTL: 1419 inp->inp_ip_ttl = optval; 1420 break; 1421#define OPTSET(bit) \ 1422 if (optval) \ 1423 inp->inp_flags |= bit; \ 1424 else \ 1425 inp->inp_flags &= ~bit; 1426 1427 case IP_RECVOPTS: 1428 OPTSET(INP_RECVOPTS); 1429 break; 1430 1431 case IP_RECVRETOPTS: 1432 OPTSET(INP_RECVRETOPTS); 1433 break; 1434 1435 case IP_RECVDSTADDR: 1436 OPTSET(INP_RECVDSTADDR); 1437 break; 1438 1439 case IP_RECVIF: 1440 OPTSET(INP_RECVIF); 1441 break; 1442 1443 case IP_FAITH: 1444 OPTSET(INP_FAITH); 1445 break; 1446 } 1447 break; 1448#undef OPTSET 1449 1450 case IP_MULTICAST_IF: 1451 case IP_MULTICAST_VIF: 1452 case IP_MULTICAST_TTL: 1453 case IP_MULTICAST_LOOP: 1454 case IP_ADD_MEMBERSHIP: 1455 case IP_DROP_MEMBERSHIP: 1456 error = ip_setmoptions(sopt, &inp->inp_moptions); 1457 break; 1458 1459 case IP_PORTRANGE: 1460 error = sooptcopyin(sopt, &optval, sizeof optval, 1461 sizeof optval); 1462 if (error) 1463 break; 1464 1465 switch (optval) { 1466 case IP_PORTRANGE_DEFAULT: 1467 inp->inp_flags &= ~(INP_LOWPORT); 1468 inp->inp_flags &= ~(INP_HIGHPORT); 1469 break; 1470 1471 case IP_PORTRANGE_HIGH: 1472 inp->inp_flags &= ~(INP_LOWPORT); 1473 inp->inp_flags |= INP_HIGHPORT; 1474 break; 1475 1476 case IP_PORTRANGE_LOW: 1477 inp->inp_flags &= ~(INP_HIGHPORT); 1478 inp->inp_flags |= INP_LOWPORT; 1479 break; 1480 1481 default: 1482 error = EINVAL; 1483 break; 1484 } 1485 break; 1486 1487#if defined(IPSEC) || defined(FAST_IPSEC) 1488 case IP_IPSEC_POLICY: 1489 { 1490 caddr_t req; 1491 size_t len = 0; 1492 int priv; 1493 struct mbuf *m; 1494 int optname; 1495 1496 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1497 break; 1498 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1499 break; 1500 priv = (sopt->sopt_td != NULL && 1501 suser(sopt->sopt_td) != 0) ? 0 : 1; 1502 req = mtod(m, caddr_t); 1503 len = m->m_len; 1504 optname = sopt->sopt_name; 1505 error = ipsec4_set_policy(inp, optname, req, len, priv); 1506 m_freem(m); 1507 break; 1508 } 1509#endif /*IPSEC*/ 1510 1511 default: 1512 error = ENOPROTOOPT; 1513 break; 1514 } 1515 break; 1516 1517 case SOPT_GET: 1518 switch (sopt->sopt_name) { 1519 case IP_OPTIONS: 1520 case IP_RETOPTS: 1521 if (inp->inp_options) 1522 error = sooptcopyout(sopt, 1523 mtod(inp->inp_options, 1524 char *), 1525 inp->inp_options->m_len); 1526 else 1527 sopt->sopt_valsize = 0; 1528 break; 1529 1530 case IP_TOS: 1531 case IP_TTL: 1532 case IP_RECVOPTS: 1533 case IP_RECVRETOPTS: 1534 case IP_RECVDSTADDR: 1535 case IP_RECVIF: 1536 case IP_PORTRANGE: 1537 case IP_FAITH: 1538 switch (sopt->sopt_name) { 1539 1540 case IP_TOS: 1541 optval = inp->inp_ip_tos; 1542 break; 1543 1544 case IP_TTL: 1545 optval = inp->inp_ip_ttl; 1546 break; 1547 1548#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1549 1550 case IP_RECVOPTS: 1551 optval = OPTBIT(INP_RECVOPTS); 1552 break; 1553 1554 case IP_RECVRETOPTS: 1555 optval = OPTBIT(INP_RECVRETOPTS); 1556 break; 1557 1558 case IP_RECVDSTADDR: 1559 optval = OPTBIT(INP_RECVDSTADDR); 1560 break; 1561 1562 case IP_RECVIF: 1563 optval = OPTBIT(INP_RECVIF); 1564 break; 1565 1566 case IP_PORTRANGE: 1567 if (inp->inp_flags & INP_HIGHPORT) 1568 optval = IP_PORTRANGE_HIGH; 1569 else if (inp->inp_flags & INP_LOWPORT) 1570 optval = IP_PORTRANGE_LOW; 1571 else 1572 optval = 0; 1573 break; 1574 1575 case IP_FAITH: 1576 optval = OPTBIT(INP_FAITH); 1577 break; 1578 } 1579 error = sooptcopyout(sopt, &optval, sizeof optval); 1580 break; 1581 1582 case IP_MULTICAST_IF: 1583 case IP_MULTICAST_VIF: 1584 case IP_MULTICAST_TTL: 1585 case IP_MULTICAST_LOOP: 1586 case IP_ADD_MEMBERSHIP: 1587 case IP_DROP_MEMBERSHIP: 1588 error = ip_getmoptions(sopt, inp->inp_moptions); 1589 break; 1590 1591#if defined(IPSEC) || defined(FAST_IPSEC) 1592 case IP_IPSEC_POLICY: 1593 { 1594 struct mbuf *m = NULL; 1595 caddr_t req = NULL; 1596 size_t len = 0; 1597 1598 if (m != 0) { 1599 req = mtod(m, caddr_t); 1600 len = m->m_len; 1601 } 1602 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1603 if (error == 0) 1604 error = soopt_mcopyout(sopt, m); /* XXX */ 1605 if (error == 0) 1606 m_freem(m); 1607 break; 1608 } 1609#endif /*IPSEC*/ 1610 1611 default: 1612 error = ENOPROTOOPT; 1613 break; 1614 } 1615 break; 1616 } 1617 return (error); 1618} 1619 1620/* 1621 * Set up IP options in pcb for insertion in output packets. 1622 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1623 * with destination address if source routed. 1624 */ 1625static int 1626ip_pcbopts(optname, pcbopt, m) 1627 int optname; 1628 struct mbuf **pcbopt; 1629 register struct mbuf *m; 1630{ 1631 register int cnt, optlen; 1632 register u_char *cp; 1633 u_char opt; 1634 1635 /* turn off any old options */ 1636 if (*pcbopt) 1637 (void)m_free(*pcbopt); 1638 *pcbopt = 0; 1639 if (m == (struct mbuf *)0 || m->m_len == 0) { 1640 /* 1641 * Only turning off any previous options. 1642 */ 1643 if (m) 1644 (void)m_free(m); 1645 return (0); 1646 } 1647 1648 if (m->m_len % sizeof(int32_t)) 1649 goto bad; 1650 /* 1651 * IP first-hop destination address will be stored before 1652 * actual options; move other options back 1653 * and clear it when none present. 1654 */ 1655 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1656 goto bad; 1657 cnt = m->m_len; 1658 m->m_len += sizeof(struct in_addr); 1659 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1660 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1661 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1662 1663 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1664 opt = cp[IPOPT_OPTVAL]; 1665 if (opt == IPOPT_EOL) 1666 break; 1667 if (opt == IPOPT_NOP) 1668 optlen = 1; 1669 else { 1670 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1671 goto bad; 1672 optlen = cp[IPOPT_OLEN]; 1673 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1674 goto bad; 1675 } 1676 switch (opt) { 1677 1678 default: 1679 break; 1680 1681 case IPOPT_LSRR: 1682 case IPOPT_SSRR: 1683 /* 1684 * user process specifies route as: 1685 * ->A->B->C->D 1686 * D must be our final destination (but we can't 1687 * check that since we may not have connected yet). 1688 * A is first hop destination, which doesn't appear in 1689 * actual IP option, but is stored before the options. 1690 */ 1691 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1692 goto bad; 1693 m->m_len -= sizeof(struct in_addr); 1694 cnt -= sizeof(struct in_addr); 1695 optlen -= sizeof(struct in_addr); 1696 cp[IPOPT_OLEN] = optlen; 1697 /* 1698 * Move first hop before start of options. 1699 */ 1700 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1701 sizeof(struct in_addr)); 1702 /* 1703 * Then copy rest of options back 1704 * to close up the deleted entry. 1705 */ 1706 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1707 sizeof(struct in_addr)), 1708 (caddr_t)&cp[IPOPT_OFFSET+1], 1709 (unsigned)cnt + sizeof(struct in_addr)); 1710 break; 1711 } 1712 } 1713 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1714 goto bad; 1715 *pcbopt = m; 1716 return (0); 1717 1718bad: 1719 (void)m_free(m); 1720 return (EINVAL); 1721} 1722 1723/* 1724 * XXX 1725 * The whole multicast option thing needs to be re-thought. 1726 * Several of these options are equally applicable to non-multicast 1727 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1728 * standard option (IP_TTL). 1729 */ 1730 1731/* 1732 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1733 */ 1734static struct ifnet * 1735ip_multicast_if(a, ifindexp) 1736 struct in_addr *a; 1737 int *ifindexp; 1738{ 1739 int ifindex; 1740 struct ifnet *ifp; 1741 1742 if (ifindexp) 1743 *ifindexp = 0; 1744 if (ntohl(a->s_addr) >> 24 == 0) { 1745 ifindex = ntohl(a->s_addr) & 0xffffff; 1746 if (ifindex < 0 || if_index < ifindex) 1747 return NULL; 1748 ifp = ifnet_byindex(ifindex); 1749 if (ifindexp) 1750 *ifindexp = ifindex; 1751 } else { 1752 INADDR_TO_IFP(*a, ifp); 1753 } 1754 return ifp; 1755} 1756 1757/* 1758 * Set the IP multicast options in response to user setsockopt(). 1759 */ 1760static int 1761ip_setmoptions(sopt, imop) 1762 struct sockopt *sopt; 1763 struct ip_moptions **imop; 1764{ 1765 int error = 0; 1766 int i; 1767 struct in_addr addr; 1768 struct ip_mreq mreq; 1769 struct ifnet *ifp; 1770 struct ip_moptions *imo = *imop; 1771 struct route ro; 1772 struct sockaddr_in *dst; 1773 int ifindex; 1774 int s; 1775 1776 if (imo == NULL) { 1777 /* 1778 * No multicast option buffer attached to the pcb; 1779 * allocate one and initialize to default values. 1780 */ 1781 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1782 M_WAITOK); 1783 1784 if (imo == NULL) 1785 return (ENOBUFS); 1786 *imop = imo; 1787 imo->imo_multicast_ifp = NULL; 1788 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1789 imo->imo_multicast_vif = -1; 1790 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1791 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1792 imo->imo_num_memberships = 0; 1793 } 1794 1795 switch (sopt->sopt_name) { 1796 /* store an index number for the vif you wanna use in the send */ 1797 case IP_MULTICAST_VIF: 1798 if (legal_vif_num == 0) { 1799 error = EOPNOTSUPP; 1800 break; 1801 } 1802 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1803 if (error) 1804 break; 1805 if (!legal_vif_num(i) && (i != -1)) { 1806 error = EINVAL; 1807 break; 1808 } 1809 imo->imo_multicast_vif = i; 1810 break; 1811 1812 case IP_MULTICAST_IF: 1813 /* 1814 * Select the interface for outgoing multicast packets. 1815 */ 1816 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1817 if (error) 1818 break; 1819 /* 1820 * INADDR_ANY is used to remove a previous selection. 1821 * When no interface is selected, a default one is 1822 * chosen every time a multicast packet is sent. 1823 */ 1824 if (addr.s_addr == INADDR_ANY) { 1825 imo->imo_multicast_ifp = NULL; 1826 break; 1827 } 1828 /* 1829 * The selected interface is identified by its local 1830 * IP address. Find the interface and confirm that 1831 * it supports multicasting. 1832 */ 1833 s = splimp(); 1834 ifp = ip_multicast_if(&addr, &ifindex); 1835 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1836 splx(s); 1837 error = EADDRNOTAVAIL; 1838 break; 1839 } 1840 imo->imo_multicast_ifp = ifp; 1841 if (ifindex) 1842 imo->imo_multicast_addr = addr; 1843 else 1844 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1845 splx(s); 1846 break; 1847 1848 case IP_MULTICAST_TTL: 1849 /* 1850 * Set the IP time-to-live for outgoing multicast packets. 1851 * The original multicast API required a char argument, 1852 * which is inconsistent with the rest of the socket API. 1853 * We allow either a char or an int. 1854 */ 1855 if (sopt->sopt_valsize == 1) { 1856 u_char ttl; 1857 error = sooptcopyin(sopt, &ttl, 1, 1); 1858 if (error) 1859 break; 1860 imo->imo_multicast_ttl = ttl; 1861 } else { 1862 u_int ttl; 1863 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1864 sizeof ttl); 1865 if (error) 1866 break; 1867 if (ttl > 255) 1868 error = EINVAL; 1869 else 1870 imo->imo_multicast_ttl = ttl; 1871 } 1872 break; 1873 1874 case IP_MULTICAST_LOOP: 1875 /* 1876 * Set the loopback flag for outgoing multicast packets. 1877 * Must be zero or one. The original multicast API required a 1878 * char argument, which is inconsistent with the rest 1879 * of the socket API. We allow either a char or an int. 1880 */ 1881 if (sopt->sopt_valsize == 1) { 1882 u_char loop; 1883 error = sooptcopyin(sopt, &loop, 1, 1); 1884 if (error) 1885 break; 1886 imo->imo_multicast_loop = !!loop; 1887 } else { 1888 u_int loop; 1889 error = sooptcopyin(sopt, &loop, sizeof loop, 1890 sizeof loop); 1891 if (error) 1892 break; 1893 imo->imo_multicast_loop = !!loop; 1894 } 1895 break; 1896 1897 case IP_ADD_MEMBERSHIP: 1898 /* 1899 * Add a multicast group membership. 1900 * Group must be a valid IP multicast address. 1901 */ 1902 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1903 if (error) 1904 break; 1905 1906 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1907 error = EINVAL; 1908 break; 1909 } 1910 s = splimp(); 1911 /* 1912 * If no interface address was provided, use the interface of 1913 * the route to the given multicast address. 1914 */ 1915 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1916 bzero((caddr_t)&ro, sizeof(ro)); 1917 dst = (struct sockaddr_in *)&ro.ro_dst; 1918 dst->sin_len = sizeof(*dst); 1919 dst->sin_family = AF_INET; 1920 dst->sin_addr = mreq.imr_multiaddr; 1921 rtalloc(&ro); 1922 if (ro.ro_rt == NULL) { 1923 error = EADDRNOTAVAIL; 1924 splx(s); 1925 break; 1926 } 1927 ifp = ro.ro_rt->rt_ifp; 1928 rtfree(ro.ro_rt); 1929 } 1930 else { 1931 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1932 } 1933 1934 /* 1935 * See if we found an interface, and confirm that it 1936 * supports multicast. 1937 */ 1938 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1939 error = EADDRNOTAVAIL; 1940 splx(s); 1941 break; 1942 } 1943 /* 1944 * See if the membership already exists or if all the 1945 * membership slots are full. 1946 */ 1947 for (i = 0; i < imo->imo_num_memberships; ++i) { 1948 if (imo->imo_membership[i]->inm_ifp == ifp && 1949 imo->imo_membership[i]->inm_addr.s_addr 1950 == mreq.imr_multiaddr.s_addr) 1951 break; 1952 } 1953 if (i < imo->imo_num_memberships) { 1954 error = EADDRINUSE; 1955 splx(s); 1956 break; 1957 } 1958 if (i == IP_MAX_MEMBERSHIPS) { 1959 error = ETOOMANYREFS; 1960 splx(s); 1961 break; 1962 } 1963 /* 1964 * Everything looks good; add a new record to the multicast 1965 * address list for the given interface. 1966 */ 1967 if ((imo->imo_membership[i] = 1968 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1969 error = ENOBUFS; 1970 splx(s); 1971 break; 1972 } 1973 ++imo->imo_num_memberships; 1974 splx(s); 1975 break; 1976 1977 case IP_DROP_MEMBERSHIP: 1978 /* 1979 * Drop a multicast group membership. 1980 * Group must be a valid IP multicast address. 1981 */ 1982 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1983 if (error) 1984 break; 1985 1986 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1987 error = EINVAL; 1988 break; 1989 } 1990 1991 s = splimp(); 1992 /* 1993 * If an interface address was specified, get a pointer 1994 * to its ifnet structure. 1995 */ 1996 if (mreq.imr_interface.s_addr == INADDR_ANY) 1997 ifp = NULL; 1998 else { 1999 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 2000 if (ifp == NULL) { 2001 error = EADDRNOTAVAIL; 2002 splx(s); 2003 break; 2004 } 2005 } 2006 /* 2007 * Find the membership in the membership array. 2008 */ 2009 for (i = 0; i < imo->imo_num_memberships; ++i) { 2010 if ((ifp == NULL || 2011 imo->imo_membership[i]->inm_ifp == ifp) && 2012 imo->imo_membership[i]->inm_addr.s_addr == 2013 mreq.imr_multiaddr.s_addr) 2014 break; 2015 } 2016 if (i == imo->imo_num_memberships) { 2017 error = EADDRNOTAVAIL; 2018 splx(s); 2019 break; 2020 } 2021 /* 2022 * Give up the multicast address record to which the 2023 * membership points. 2024 */ 2025 in_delmulti(imo->imo_membership[i]); 2026 /* 2027 * Remove the gap in the membership array. 2028 */ 2029 for (++i; i < imo->imo_num_memberships; ++i) 2030 imo->imo_membership[i-1] = imo->imo_membership[i]; 2031 --imo->imo_num_memberships; 2032 splx(s); 2033 break; 2034 2035 default: 2036 error = EOPNOTSUPP; 2037 break; 2038 } 2039 2040 /* 2041 * If all options have default values, no need to keep the mbuf. 2042 */ 2043 if (imo->imo_multicast_ifp == NULL && 2044 imo->imo_multicast_vif == -1 && 2045 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2046 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2047 imo->imo_num_memberships == 0) { 2048 free(*imop, M_IPMOPTS); 2049 *imop = NULL; 2050 } 2051 2052 return (error); 2053} 2054 2055/* 2056 * Return the IP multicast options in response to user getsockopt(). 2057 */ 2058static int 2059ip_getmoptions(sopt, imo) 2060 struct sockopt *sopt; 2061 register struct ip_moptions *imo; 2062{ 2063 struct in_addr addr; 2064 struct in_ifaddr *ia; 2065 int error, optval; 2066 u_char coptval; 2067 2068 error = 0; 2069 switch (sopt->sopt_name) { 2070 case IP_MULTICAST_VIF: 2071 if (imo != NULL) 2072 optval = imo->imo_multicast_vif; 2073 else 2074 optval = -1; 2075 error = sooptcopyout(sopt, &optval, sizeof optval); 2076 break; 2077 2078 case IP_MULTICAST_IF: 2079 if (imo == NULL || imo->imo_multicast_ifp == NULL) 2080 addr.s_addr = INADDR_ANY; 2081 else if (imo->imo_multicast_addr.s_addr) { 2082 /* return the value user has set */ 2083 addr = imo->imo_multicast_addr; 2084 } else { 2085 IFP_TO_IA(imo->imo_multicast_ifp, ia); 2086 addr.s_addr = (ia == NULL) ? INADDR_ANY 2087 : IA_SIN(ia)->sin_addr.s_addr; 2088 } 2089 error = sooptcopyout(sopt, &addr, sizeof addr); 2090 break; 2091 2092 case IP_MULTICAST_TTL: 2093 if (imo == 0) 2094 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 2095 else 2096 optval = coptval = imo->imo_multicast_ttl; 2097 if (sopt->sopt_valsize == 1) 2098 error = sooptcopyout(sopt, &coptval, 1); 2099 else 2100 error = sooptcopyout(sopt, &optval, sizeof optval); 2101 break; 2102 2103 case IP_MULTICAST_LOOP: 2104 if (imo == 0) 2105 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 2106 else 2107 optval = coptval = imo->imo_multicast_loop; 2108 if (sopt->sopt_valsize == 1) 2109 error = sooptcopyout(sopt, &coptval, 1); 2110 else 2111 error = sooptcopyout(sopt, &optval, sizeof optval); 2112 break; 2113 2114 default: 2115 error = ENOPROTOOPT; 2116 break; 2117 } 2118 return (error); 2119} 2120 2121/* 2122 * Discard the IP multicast options. 2123 */ 2124void 2125ip_freemoptions(imo) 2126 register struct ip_moptions *imo; 2127{ 2128 register int i; 2129 2130 if (imo != NULL) { 2131 for (i = 0; i < imo->imo_num_memberships; ++i) 2132 in_delmulti(imo->imo_membership[i]); 2133 free(imo, M_IPMOPTS); 2134 } 2135} 2136 2137/* 2138 * Routine called from ip_output() to loop back a copy of an IP multicast 2139 * packet to the input queue of a specified interface. Note that this 2140 * calls the output routine of the loopback "driver", but with an interface 2141 * pointer that might NOT be a loopback interface -- evil, but easier than 2142 * replicating that code here. 2143 */ 2144static void 2145ip_mloopback(ifp, m, dst, hlen) 2146 struct ifnet *ifp; 2147 register struct mbuf *m; 2148 register struct sockaddr_in *dst; 2149 int hlen; 2150{ 2151 register struct ip *ip; 2152 struct mbuf *copym; 2153 2154 copym = m_copy(m, 0, M_COPYALL); 2155 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 2156 copym = m_pullup(copym, hlen); 2157 if (copym != NULL) { 2158 /* 2159 * We don't bother to fragment if the IP length is greater 2160 * than the interface's MTU. Can this possibly matter? 2161 */ 2162 ip = mtod(copym, struct ip *); 2163 ip->ip_len = htons(ip->ip_len); 2164 ip->ip_off = htons(ip->ip_off); 2165 ip->ip_sum = 0; 2166 if (ip->ip_vhl == IP_VHL_BORING) { 2167 ip->ip_sum = in_cksum_hdr(ip); 2168 } else { 2169 ip->ip_sum = in_cksum(copym, hlen); 2170 } 2171 /* 2172 * NB: 2173 * It's not clear whether there are any lingering 2174 * reentrancy problems in other areas which might 2175 * be exposed by using ip_input directly (in 2176 * particular, everything which modifies the packet 2177 * in-place). Yet another option is using the 2178 * protosw directly to deliver the looped back 2179 * packet. For the moment, we'll err on the side 2180 * of safety by using if_simloop(). 2181 */ 2182#if 1 /* XXX */ 2183 if (dst->sin_family != AF_INET) { 2184 printf("ip_mloopback: bad address family %d\n", 2185 dst->sin_family); 2186 dst->sin_family = AF_INET; 2187 } 2188#endif 2189 2190#ifdef notdef 2191 copym->m_pkthdr.rcvif = ifp; 2192 ip_input(copym); 2193#else 2194 /* if the checksum hasn't been computed, mark it as valid */ 2195 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2196 copym->m_pkthdr.csum_flags |= 2197 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 2198 copym->m_pkthdr.csum_data = 0xffff; 2199 } 2200 if_simloop(ifp, copym, dst->sin_family, 0); 2201#endif 2202 } 2203} 2204