ip_output.c revision 111119
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 111119 2003-02-19 05:47:46Z imp $ 35 */ 36 37#include "opt_ipfw.h" 38#include "opt_ipdn.h" 39#include "opt_ipdivert.h" 40#include "opt_ipfilter.h" 41#include "opt_ipsec.h" 42#include "opt_mac.h" 43#include "opt_pfil_hooks.h" 44#include "opt_random_ip_id.h" 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/kernel.h> 49#include <sys/mac.h> 50#include <sys/malloc.h> 51#include <sys/mbuf.h> 52#include <sys/protosw.h> 53#include <sys/socket.h> 54#include <sys/socketvar.h> 55 56#include <net/if.h> 57#include <net/route.h> 58 59#include <netinet/in.h> 60#include <netinet/in_systm.h> 61#include <netinet/ip.h> 62#include <netinet/in_pcb.h> 63#include <netinet/in_var.h> 64#include <netinet/ip_var.h> 65 66#include <machine/in_cksum.h> 67 68static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 69 70#ifdef IPSEC 71#include <netinet6/ipsec.h> 72#include <netkey/key.h> 73#ifdef IPSEC_DEBUG 74#include <netkey/key_debug.h> 75#else 76#define KEYDEBUG(lev,arg) 77#endif 78#endif /*IPSEC*/ 79 80#ifdef FAST_IPSEC 81#include <netipsec/ipsec.h> 82#include <netipsec/xform.h> 83#include <netipsec/key.h> 84#endif /*FAST_IPSEC*/ 85 86#include <netinet/ip_fw.h> 87#include <netinet/ip_dummynet.h> 88 89#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 90 x, (ntohl(a.s_addr)>>24)&0xFF,\ 91 (ntohl(a.s_addr)>>16)&0xFF,\ 92 (ntohl(a.s_addr)>>8)&0xFF,\ 93 (ntohl(a.s_addr))&0xFF, y); 94 95u_short ip_id; 96 97static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 98static struct ifnet *ip_multicast_if(struct in_addr *, int *); 99static void ip_mloopback 100 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 101static int ip_getmoptions 102 (struct sockopt *, struct ip_moptions *); 103static int ip_pcbopts(int, struct mbuf **, struct mbuf *); 104static int ip_setmoptions 105 (struct sockopt *, struct ip_moptions **); 106 107int ip_optcopy(struct ip *, struct ip *); 108 109 110extern struct protosw inetsw[]; 111 112/* 113 * IP output. The packet in mbuf chain m contains a skeletal IP 114 * header (with len, off, ttl, proto, tos, src, dst). 115 * The mbuf chain containing the packet will be freed. 116 * The mbuf opt, if present, will not be freed. 117 */ 118int 119ip_output(m0, opt, ro, flags, imo, inp) 120 struct mbuf *m0; 121 struct mbuf *opt; 122 struct route *ro; 123 int flags; 124 struct ip_moptions *imo; 125 struct inpcb *inp; 126{ 127 struct ip *ip, *mhip; 128 struct ifnet *ifp = NULL; /* keep compiler happy */ 129 struct mbuf *m; 130 int hlen = sizeof (struct ip); 131 int len, off, error = 0; 132 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 133 struct in_ifaddr *ia = NULL; 134 int isbroadcast, sw_csum; 135 struct in_addr pkt_dst; 136#ifdef IPSEC 137 struct route iproute; 138 struct secpolicy *sp = NULL; 139 struct socket *so = inp ? inp->inp_socket : NULL; 140#endif 141#ifdef FAST_IPSEC 142 struct route iproute; 143 struct m_tag *mtag; 144 struct secpolicy *sp = NULL; 145 struct tdb_ident *tdbi; 146 int s; 147#endif /* FAST_IPSEC */ 148 struct ip_fw_args args; 149 int src_was_INADDR_ANY = 0; /* as the name says... */ 150#ifdef PFIL_HOOKS 151 struct packet_filter_hook *pfh; 152 struct mbuf *m1; 153 int rv; 154#endif /* PFIL_HOOKS */ 155 156 args.eh = NULL; 157 args.rule = NULL; 158 args.next_hop = NULL; 159 args.divert_rule = 0; /* divert cookie */ 160 161 /* Grab info from MT_TAG mbufs prepended to the chain. */ 162 for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) { 163 switch(m0->_m_tag_id) { 164 default: 165 printf("ip_output: unrecognised MT_TAG tag %d\n", 166 m0->_m_tag_id); 167 break; 168 169 case PACKET_TAG_DUMMYNET: 170 /* 171 * the packet was already tagged, so part of the 172 * processing was already done, and we need to go down. 173 * Get parameters from the header. 174 */ 175 args.rule = ((struct dn_pkt *)m0)->rule; 176 opt = NULL ; 177 ro = & ( ((struct dn_pkt *)m0)->ro ) ; 178 imo = NULL ; 179 dst = ((struct dn_pkt *)m0)->dn_dst ; 180 ifp = ((struct dn_pkt *)m0)->ifp ; 181 flags = ((struct dn_pkt *)m0)->flags ; 182 break; 183 184 case PACKET_TAG_DIVERT: 185 args.divert_rule = (intptr_t)m0->m_data & 0xffff; 186 break; 187 188 case PACKET_TAG_IPFORWARD: 189 args.next_hop = (struct sockaddr_in *)m0->m_data; 190 break; 191 } 192 } 193 m = m0; 194 195 KASSERT(!m || (m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR")); 196#ifndef FAST_IPSEC 197 KASSERT(ro != NULL, ("ip_output: no route, proto %d", 198 mtod(m, struct ip *)->ip_p)); 199#endif 200 201 if (args.rule != NULL) { /* dummynet already saw us */ 202 ip = mtod(m, struct ip *); 203 hlen = ip->ip_hl << 2 ; 204 if (ro->ro_rt) 205 ia = ifatoia(ro->ro_rt->rt_ifa); 206 goto sendit; 207 } 208 209 if (opt) { 210 len = 0; 211 m = ip_insertoptions(m, opt, &len); 212 if (len != 0) 213 hlen = len; 214 } 215 ip = mtod(m, struct ip *); 216 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 217 218 /* 219 * Fill in IP header. 220 */ 221 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 222 ip->ip_v = IPVERSION; 223 ip->ip_hl = hlen >> 2; 224 ip->ip_off &= IP_DF; 225#ifdef RANDOM_IP_ID 226 ip->ip_id = ip_randomid(); 227#else 228 ip->ip_id = htons(ip_id++); 229#endif 230 ipstat.ips_localout++; 231 } else { 232 hlen = ip->ip_hl << 2; 233 } 234 235#ifdef FAST_IPSEC 236 if (ro == NULL) { 237 ro = &iproute; 238 bzero(ro, sizeof (*ro)); 239 } 240#endif /* FAST_IPSEC */ 241 dst = (struct sockaddr_in *)&ro->ro_dst; 242 /* 243 * If there is a cached route, 244 * check that it is to the same destination 245 * and is still up. If not, free it and try again. 246 * The address family should also be checked in case of sharing the 247 * cache with IPv6. 248 */ 249 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 250 dst->sin_family != AF_INET || 251 dst->sin_addr.s_addr != pkt_dst.s_addr)) { 252 RTFREE(ro->ro_rt); 253 ro->ro_rt = (struct rtentry *)0; 254 } 255 if (ro->ro_rt == 0) { 256 bzero(dst, sizeof(*dst)); 257 dst->sin_family = AF_INET; 258 dst->sin_len = sizeof(*dst); 259 dst->sin_addr = pkt_dst; 260 } 261 /* 262 * If routing to interface only, 263 * short circuit routing lookup. 264 */ 265 if (flags & IP_ROUTETOIF) { 266 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 267 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 268 ipstat.ips_noroute++; 269 error = ENETUNREACH; 270 goto bad; 271 } 272 ifp = ia->ia_ifp; 273 ip->ip_ttl = 1; 274 isbroadcast = in_broadcast(dst->sin_addr, ifp); 275 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 276 imo != NULL && imo->imo_multicast_ifp != NULL) { 277 /* 278 * Bypass the normal routing lookup for multicast 279 * packets if the interface is specified. 280 */ 281 ifp = imo->imo_multicast_ifp; 282 IFP_TO_IA(ifp, ia); 283 isbroadcast = 0; /* fool gcc */ 284 } else { 285 /* 286 * If this is the case, we probably don't want to allocate 287 * a protocol-cloned route since we didn't get one from the 288 * ULP. This lets TCP do its thing, while not burdening 289 * forwarding or ICMP with the overhead of cloning a route. 290 * Of course, we still want to do any cloning requested by 291 * the link layer, as this is probably required in all cases 292 * for correct operation (as it is for ARP). 293 */ 294 if (ro->ro_rt == 0) 295 rtalloc_ign(ro, RTF_PRCLONING); 296 if (ro->ro_rt == 0) { 297 ipstat.ips_noroute++; 298 error = EHOSTUNREACH; 299 goto bad; 300 } 301 ia = ifatoia(ro->ro_rt->rt_ifa); 302 ifp = ro->ro_rt->rt_ifp; 303 ro->ro_rt->rt_use++; 304 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 305 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 306 if (ro->ro_rt->rt_flags & RTF_HOST) 307 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 308 else 309 isbroadcast = in_broadcast(dst->sin_addr, ifp); 310 } 311 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 312 struct in_multi *inm; 313 314 m->m_flags |= M_MCAST; 315 /* 316 * IP destination address is multicast. Make sure "dst" 317 * still points to the address in "ro". (It may have been 318 * changed to point to a gateway address, above.) 319 */ 320 dst = (struct sockaddr_in *)&ro->ro_dst; 321 /* 322 * See if the caller provided any multicast options 323 */ 324 if (imo != NULL) { 325 ip->ip_ttl = imo->imo_multicast_ttl; 326 if (imo->imo_multicast_vif != -1) 327 ip->ip_src.s_addr = 328 ip_mcast_src ? 329 ip_mcast_src(imo->imo_multicast_vif) : 330 INADDR_ANY; 331 } else 332 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 333 /* 334 * Confirm that the outgoing interface supports multicast. 335 */ 336 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 337 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 338 ipstat.ips_noroute++; 339 error = ENETUNREACH; 340 goto bad; 341 } 342 } 343 /* 344 * If source address not specified yet, use address 345 * of outgoing interface. 346 */ 347 if (ip->ip_src.s_addr == INADDR_ANY) { 348 /* Interface may have no addresses. */ 349 if (ia != NULL) 350 ip->ip_src = IA_SIN(ia)->sin_addr; 351 } 352 353 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 354 /* 355 * XXX 356 * delayed checksums are not currently 357 * compatible with IP multicast routing 358 */ 359 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 360 in_delayed_cksum(m); 361 m->m_pkthdr.csum_flags &= 362 ~CSUM_DELAY_DATA; 363 } 364 } 365 IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 366 if (inm != NULL && 367 (imo == NULL || imo->imo_multicast_loop)) { 368 /* 369 * If we belong to the destination multicast group 370 * on the outgoing interface, and the caller did not 371 * forbid loopback, loop back a copy. 372 */ 373 ip_mloopback(ifp, m, dst, hlen); 374 } 375 else { 376 /* 377 * If we are acting as a multicast router, perform 378 * multicast forwarding as if the packet had just 379 * arrived on the interface to which we are about 380 * to send. The multicast forwarding function 381 * recursively calls this function, using the 382 * IP_FORWARDING flag to prevent infinite recursion. 383 * 384 * Multicasts that are looped back by ip_mloopback(), 385 * above, will be forwarded by the ip_input() routine, 386 * if necessary. 387 */ 388 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 389 /* 390 * If rsvp daemon is not running, do not 391 * set ip_moptions. This ensures that the packet 392 * is multicast and not just sent down one link 393 * as prescribed by rsvpd. 394 */ 395 if (!rsvp_on) 396 imo = NULL; 397 if (ip_mforward && 398 ip_mforward(ip, ifp, m, imo) != 0) { 399 m_freem(m); 400 goto done; 401 } 402 } 403 } 404 405 /* 406 * Multicasts with a time-to-live of zero may be looped- 407 * back, above, but must not be transmitted on a network. 408 * Also, multicasts addressed to the loopback interface 409 * are not sent -- the above call to ip_mloopback() will 410 * loop back a copy if this host actually belongs to the 411 * destination group on the loopback interface. 412 */ 413 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 414 m_freem(m); 415 goto done; 416 } 417 418 goto sendit; 419 } 420#ifndef notdef 421 /* 422 * If the source address is not specified yet, use the address 423 * of the outoing interface. In case, keep note we did that, so 424 * if the the firewall changes the next-hop causing the output 425 * interface to change, we can fix that. 426 */ 427 if (ip->ip_src.s_addr == INADDR_ANY) { 428 /* Interface may have no addresses. */ 429 if (ia != NULL) { 430 ip->ip_src = IA_SIN(ia)->sin_addr; 431 src_was_INADDR_ANY = 1; 432 } 433 } 434#endif /* notdef */ 435 /* 436 * Verify that we have any chance at all of being able to queue 437 * the packet or packet fragments 438 */ 439 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 440 ifp->if_snd.ifq_maxlen) { 441 error = ENOBUFS; 442 ipstat.ips_odropped++; 443 goto bad; 444 } 445 446 /* 447 * Look for broadcast address and 448 * verify user is allowed to send 449 * such a packet. 450 */ 451 if (isbroadcast) { 452 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 453 error = EADDRNOTAVAIL; 454 goto bad; 455 } 456 if ((flags & IP_ALLOWBROADCAST) == 0) { 457 error = EACCES; 458 goto bad; 459 } 460 /* don't allow broadcast messages to be fragmented */ 461 if ((u_short)ip->ip_len > ifp->if_mtu) { 462 error = EMSGSIZE; 463 goto bad; 464 } 465 m->m_flags |= M_BCAST; 466 } else { 467 m->m_flags &= ~M_BCAST; 468 } 469 470sendit: 471#ifdef IPSEC 472 /* get SP for this packet */ 473 if (so == NULL) 474 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 475 else 476 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 477 478 if (sp == NULL) { 479 ipsecstat.out_inval++; 480 goto bad; 481 } 482 483 error = 0; 484 485 /* check policy */ 486 switch (sp->policy) { 487 case IPSEC_POLICY_DISCARD: 488 /* 489 * This packet is just discarded. 490 */ 491 ipsecstat.out_polvio++; 492 goto bad; 493 494 case IPSEC_POLICY_BYPASS: 495 case IPSEC_POLICY_NONE: 496 /* no need to do IPsec. */ 497 goto skip_ipsec; 498 499 case IPSEC_POLICY_IPSEC: 500 if (sp->req == NULL) { 501 /* acquire a policy */ 502 error = key_spdacquire(sp); 503 goto bad; 504 } 505 break; 506 507 case IPSEC_POLICY_ENTRUST: 508 default: 509 printf("ip_output: Invalid policy found. %d\n", sp->policy); 510 } 511 { 512 struct ipsec_output_state state; 513 bzero(&state, sizeof(state)); 514 state.m = m; 515 if (flags & IP_ROUTETOIF) { 516 state.ro = &iproute; 517 bzero(&iproute, sizeof(iproute)); 518 } else 519 state.ro = ro; 520 state.dst = (struct sockaddr *)dst; 521 522 ip->ip_sum = 0; 523 524 /* 525 * XXX 526 * delayed checksums are not currently compatible with IPsec 527 */ 528 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 529 in_delayed_cksum(m); 530 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 531 } 532 533 ip->ip_len = htons(ip->ip_len); 534 ip->ip_off = htons(ip->ip_off); 535 536 error = ipsec4_output(&state, sp, flags); 537 538 m = state.m; 539 if (flags & IP_ROUTETOIF) { 540 /* 541 * if we have tunnel mode SA, we may need to ignore 542 * IP_ROUTETOIF. 543 */ 544 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 545 flags &= ~IP_ROUTETOIF; 546 ro = state.ro; 547 } 548 } else 549 ro = state.ro; 550 dst = (struct sockaddr_in *)state.dst; 551 if (error) { 552 /* mbuf is already reclaimed in ipsec4_output. */ 553 m0 = NULL; 554 switch (error) { 555 case EHOSTUNREACH: 556 case ENETUNREACH: 557 case EMSGSIZE: 558 case ENOBUFS: 559 case ENOMEM: 560 break; 561 default: 562 printf("ip4_output (ipsec): error code %d\n", error); 563 /*fall through*/ 564 case ENOENT: 565 /* don't show these error codes to the user */ 566 error = 0; 567 break; 568 } 569 goto bad; 570 } 571 } 572 573 /* be sure to update variables that are affected by ipsec4_output() */ 574 ip = mtod(m, struct ip *); 575 hlen = ip->ip_hl << 2; 576 if (ro->ro_rt == NULL) { 577 if ((flags & IP_ROUTETOIF) == 0) { 578 printf("ip_output: " 579 "can't update route after IPsec processing\n"); 580 error = EHOSTUNREACH; /*XXX*/ 581 goto bad; 582 } 583 } else { 584 ia = ifatoia(ro->ro_rt->rt_ifa); 585 ifp = ro->ro_rt->rt_ifp; 586 } 587 588 /* make it flipped, again. */ 589 ip->ip_len = ntohs(ip->ip_len); 590 ip->ip_off = ntohs(ip->ip_off); 591skip_ipsec: 592#endif /*IPSEC*/ 593#ifdef FAST_IPSEC 594 /* 595 * Check the security policy (SP) for the packet and, if 596 * required, do IPsec-related processing. There are two 597 * cases here; the first time a packet is sent through 598 * it will be untagged and handled by ipsec4_checkpolicy. 599 * If the packet is resubmitted to ip_output (e.g. after 600 * AH, ESP, etc. processing), there will be a tag to bypass 601 * the lookup and related policy checking. 602 */ 603 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 604 s = splnet(); 605 if (mtag != NULL) { 606 tdbi = (struct tdb_ident *)(mtag + 1); 607 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 608 if (sp == NULL) 609 error = -EINVAL; /* force silent drop */ 610 m_tag_delete(m, mtag); 611 } else { 612 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 613 &error, inp); 614 } 615 /* 616 * There are four return cases: 617 * sp != NULL apply IPsec policy 618 * sp == NULL, error == 0 no IPsec handling needed 619 * sp == NULL, error == -EINVAL discard packet w/o error 620 * sp == NULL, error != 0 discard packet, report error 621 */ 622 if (sp != NULL) { 623 /* Loop detection, check if ipsec processing already done */ 624 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 625 for (mtag = m_tag_first(m); mtag != NULL; 626 mtag = m_tag_next(m, mtag)) { 627 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 628 continue; 629 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 630 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 631 continue; 632 /* 633 * Check if policy has an SA associated with it. 634 * This can happen when an SP has yet to acquire 635 * an SA; e.g. on first reference. If it occurs, 636 * then we let ipsec4_process_packet do its thing. 637 */ 638 if (sp->req->sav == NULL) 639 break; 640 tdbi = (struct tdb_ident *)(mtag + 1); 641 if (tdbi->spi == sp->req->sav->spi && 642 tdbi->proto == sp->req->sav->sah->saidx.proto && 643 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 644 sizeof (union sockaddr_union)) == 0) { 645 /* 646 * No IPsec processing is needed, free 647 * reference to SP. 648 * 649 * NB: null pointer to avoid free at 650 * done: below. 651 */ 652 KEY_FREESP(&sp), sp = NULL; 653 splx(s); 654 goto spd_done; 655 } 656 } 657 658 /* 659 * Do delayed checksums now because we send before 660 * this is done in the normal processing path. 661 */ 662 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 663 in_delayed_cksum(m); 664 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 665 } 666 667 ip->ip_len = htons(ip->ip_len); 668 ip->ip_off = htons(ip->ip_off); 669 670 /* NB: callee frees mbuf */ 671 error = ipsec4_process_packet(m, sp->req, flags, 0); 672 /* 673 * Preserve KAME behaviour: ENOENT can be returned 674 * when an SA acquire is in progress. Don't propagate 675 * this to user-level; it confuses applications. 676 * 677 * XXX this will go away when the SADB is redone. 678 */ 679 if (error == ENOENT) 680 error = 0; 681 splx(s); 682 goto done; 683 } else { 684 splx(s); 685 686 if (error != 0) { 687 /* 688 * Hack: -EINVAL is used to signal that a packet 689 * should be silently discarded. This is typically 690 * because we asked key management for an SA and 691 * it was delayed (e.g. kicked up to IKE). 692 */ 693 if (error == -EINVAL) 694 error = 0; 695 goto bad; 696 } else { 697 /* No IPsec processing for this packet. */ 698 } 699#ifdef notyet 700 /* 701 * If deferred crypto processing is needed, check that 702 * the interface supports it. 703 */ 704 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 705 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 706 /* notify IPsec to do its own crypto */ 707 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 708 error = EHOSTUNREACH; 709 goto bad; 710 } 711#endif 712 } 713spd_done: 714#endif /* FAST_IPSEC */ 715 716 /* 717 * IpHack's section. 718 * - Xlate: translate packet's addr/port (NAT). 719 * - Firewall: deny/allow/etc. 720 * - Wrap: fake packet's addr/port <unimpl.> 721 * - Encapsulate: put it in another IP and send out. <unimp.> 722 */ 723#ifdef PFIL_HOOKS 724 /* 725 * Run through list of hooks for output packets. 726 */ 727 m1 = m; 728 pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 729 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 730 if (pfh->pfil_func) { 731 rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1); 732 if (rv) { 733 error = EHOSTUNREACH; 734 goto done; 735 } 736 m = m1; 737 if (m == NULL) 738 goto done; 739 ip = mtod(m, struct ip *); 740 } 741#endif /* PFIL_HOOKS */ 742 743 /* 744 * Check with the firewall... 745 * but not if we are already being fwd'd from a firewall. 746 */ 747 if (fw_enable && IPFW_LOADED && !args.next_hop) { 748 struct sockaddr_in *old = dst; 749 750 args.m = m; 751 args.next_hop = dst; 752 args.oif = ifp; 753 off = ip_fw_chk_ptr(&args); 754 m = args.m; 755 dst = args.next_hop; 756 757 /* 758 * On return we must do the following: 759 * m == NULL -> drop the pkt (old interface, deprecated) 760 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 761 * 1<=off<= 0xffff -> DIVERT 762 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 763 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 764 * dst != old -> IPFIREWALL_FORWARD 765 * off==0, dst==old -> accept 766 * If some of the above modules are not compiled in, then 767 * we should't have to check the corresponding condition 768 * (because the ipfw control socket should not accept 769 * unsupported rules), but better play safe and drop 770 * packets in case of doubt. 771 */ 772 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) { 773 if (m) 774 m_freem(m); 775 error = EACCES; 776 goto done; 777 } 778 ip = mtod(m, struct ip *); 779 if (off == 0 && dst == old) /* common case */ 780 goto pass; 781 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 782 /* 783 * pass the pkt to dummynet. Need to include 784 * pipe number, m, ifp, ro, dst because these are 785 * not recomputed in the next pass. 786 * All other parameters have been already used and 787 * so they are not needed anymore. 788 * XXX note: if the ifp or ro entry are deleted 789 * while a pkt is in dummynet, we are in trouble! 790 */ 791 args.ro = ro; 792 args.dst = dst; 793 args.flags = flags; 794 795 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, 796 &args); 797 goto done; 798 } 799#ifdef IPDIVERT 800 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 801 struct mbuf *clone = NULL; 802 803 /* Clone packet if we're doing a 'tee' */ 804 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 805 clone = m_dup(m, M_DONTWAIT); 806 807 /* 808 * XXX 809 * delayed checksums are not currently compatible 810 * with divert sockets. 811 */ 812 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 813 in_delayed_cksum(m); 814 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 815 } 816 817 /* Restore packet header fields to original values */ 818 ip->ip_len = htons(ip->ip_len); 819 ip->ip_off = htons(ip->ip_off); 820 821 /* Deliver packet to divert input routine */ 822 divert_packet(m, 0, off & 0xffff, args.divert_rule); 823 824 /* If 'tee', continue with original packet */ 825 if (clone != NULL) { 826 m = clone; 827 ip = mtod(m, struct ip *); 828 goto pass; 829 } 830 goto done; 831 } 832#endif 833 834 /* IPFIREWALL_FORWARD */ 835 /* 836 * Check dst to make sure it is directly reachable on the 837 * interface we previously thought it was. 838 * If it isn't (which may be likely in some situations) we have 839 * to re-route it (ie, find a route for the next-hop and the 840 * associated interface) and set them here. This is nested 841 * forwarding which in most cases is undesirable, except where 842 * such control is nigh impossible. So we do it here. 843 * And I'm babbling. 844 */ 845 if (off == 0 && old != dst) { /* FORWARD, dst has changed */ 846#if 0 847 /* 848 * XXX To improve readability, this block should be 849 * changed into a function call as below: 850 */ 851 error = ip_ipforward(&m, &dst, &ifp); 852 if (error) 853 goto bad; 854 if (m == NULL) /* ip_input consumed the mbuf */ 855 goto done; 856#else 857 struct in_ifaddr *ia; 858 859 /* 860 * XXX sro_fwd below is static, and a pointer 861 * to it gets passed to routines downstream. 862 * This could have surprisingly bad results in 863 * practice, because its content is overwritten 864 * by subsequent packets. 865 */ 866 /* There must be a better way to do this next line... */ 867 static struct route sro_fwd; 868 struct route *ro_fwd = &sro_fwd; 869 870#if 0 871 print_ip("IPFIREWALL_FORWARD: New dst ip: ", 872 dst->sin_addr, "\n"); 873#endif 874 875 /* 876 * We need to figure out if we have been forwarded 877 * to a local socket. If so, then we should somehow 878 * "loop back" to ip_input, and get directed to the 879 * PCB as if we had received this packet. This is 880 * because it may be dificult to identify the packets 881 * you want to forward until they are being output 882 * and have selected an interface. (e.g. locally 883 * initiated packets) If we used the loopback inteface, 884 * we would not be able to control what happens 885 * as the packet runs through ip_input() as 886 * it is done through an ISR. 887 */ 888 LIST_FOREACH(ia, 889 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 890 /* 891 * If the addr to forward to is one 892 * of ours, we pretend to 893 * be the destination for this packet. 894 */ 895 if (IA_SIN(ia)->sin_addr.s_addr == 896 dst->sin_addr.s_addr) 897 break; 898 } 899 if (ia) { /* tell ip_input "dont filter" */ 900 struct m_hdr tag; 901 902 tag.mh_type = MT_TAG; 903 tag.mh_flags = PACKET_TAG_IPFORWARD; 904 tag.mh_data = (caddr_t)args.next_hop; 905 tag.mh_next = m; 906 907 if (m->m_pkthdr.rcvif == NULL) 908 m->m_pkthdr.rcvif = ifunit("lo0"); 909 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 910 m->m_pkthdr.csum_flags |= 911 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 912 m0->m_pkthdr.csum_data = 0xffff; 913 } 914 m->m_pkthdr.csum_flags |= 915 CSUM_IP_CHECKED | CSUM_IP_VALID; 916 ip->ip_len = htons(ip->ip_len); 917 ip->ip_off = htons(ip->ip_off); 918 ip_input((struct mbuf *)&tag); 919 goto done; 920 } 921 /* Some of the logic for this was 922 * nicked from above. 923 * 924 * This rewrites the cached route in a local PCB. 925 * Is this what we want to do? 926 */ 927 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 928 929 ro_fwd->ro_rt = 0; 930 rtalloc_ign(ro_fwd, RTF_PRCLONING); 931 932 if (ro_fwd->ro_rt == 0) { 933 ipstat.ips_noroute++; 934 error = EHOSTUNREACH; 935 goto bad; 936 } 937 938 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 939 ifp = ro_fwd->ro_rt->rt_ifp; 940 ro_fwd->ro_rt->rt_use++; 941 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 942 dst = (struct sockaddr_in *) 943 ro_fwd->ro_rt->rt_gateway; 944 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 945 isbroadcast = 946 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 947 else 948 isbroadcast = in_broadcast(dst->sin_addr, ifp); 949 if (ro->ro_rt) 950 RTFREE(ro->ro_rt); 951 ro->ro_rt = ro_fwd->ro_rt; 952 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 953 954#endif /* ... block to be put into a function */ 955 /* 956 * If we added a default src ip earlier, 957 * which would have been gotten from the-then 958 * interface, do it again, from the new one. 959 */ 960 if (src_was_INADDR_ANY) 961 ip->ip_src = IA_SIN(ia)->sin_addr; 962 goto pass ; 963 } 964 965 /* 966 * if we get here, none of the above matches, and 967 * we have to drop the pkt 968 */ 969 m_freem(m); 970 error = EACCES; /* not sure this is the right error msg */ 971 goto done; 972 } 973 974pass: 975 /* 127/8 must not appear on wire - RFC1122. */ 976 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 977 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 978 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 979 ipstat.ips_badaddr++; 980 error = EADDRNOTAVAIL; 981 goto bad; 982 } 983 } 984 985 m->m_pkthdr.csum_flags |= CSUM_IP; 986 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 987 if (sw_csum & CSUM_DELAY_DATA) { 988 in_delayed_cksum(m); 989 sw_csum &= ~CSUM_DELAY_DATA; 990 } 991 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 992 993 /* 994 * If small enough for interface, or the interface will take 995 * care of the fragmentation for us, can just send directly. 996 */ 997 if ((u_short)ip->ip_len <= ifp->if_mtu || 998 ifp->if_hwassist & CSUM_FRAGMENT) { 999 ip->ip_len = htons(ip->ip_len); 1000 ip->ip_off = htons(ip->ip_off); 1001 ip->ip_sum = 0; 1002 if (sw_csum & CSUM_DELAY_IP) 1003 ip->ip_sum = in_cksum(m, hlen); 1004 1005 /* Record statistics for this interface address. */ 1006 if (!(flags & IP_FORWARDING) && ia) { 1007 ia->ia_ifa.if_opackets++; 1008 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1009 } 1010 1011#ifdef IPSEC 1012 /* clean ipsec history once it goes out of the node */ 1013 ipsec_delaux(m); 1014#endif 1015 1016 error = (*ifp->if_output)(ifp, m, 1017 (struct sockaddr *)dst, ro->ro_rt); 1018 goto done; 1019 } 1020 /* 1021 * Too large for interface; fragment if possible. 1022 * Must be able to put at least 8 bytes per fragment. 1023 */ 1024 if (ip->ip_off & IP_DF) { 1025 error = EMSGSIZE; 1026 /* 1027 * This case can happen if the user changed the MTU 1028 * of an interface after enabling IP on it. Because 1029 * most netifs don't keep track of routes pointing to 1030 * them, there is no way for one to update all its 1031 * routes when the MTU is changed. 1032 */ 1033 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 1034 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 1035 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 1036 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 1037 } 1038 ipstat.ips_cantfrag++; 1039 goto bad; 1040 } 1041 len = (ifp->if_mtu - hlen) &~ 7; 1042 if (len < 8) { 1043 error = EMSGSIZE; 1044 goto bad; 1045 } 1046 1047 /* 1048 * if the interface will not calculate checksums on 1049 * fragmented packets, then do it here. 1050 */ 1051 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 1052 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 1053 in_delayed_cksum(m); 1054 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1055 } 1056 1057 if (len > PAGE_SIZE) { 1058 /* 1059 * Fragement large datagrams such that each segment 1060 * contains a multiple of PAGE_SIZE amount of data, 1061 * plus headers. This enables a receiver to perform 1062 * page-flipping zero-copy optimizations. 1063 */ 1064 1065 int newlen; 1066 struct mbuf *mtmp; 1067 1068 for (mtmp = m, off = 0; 1069 mtmp && ((off + mtmp->m_len) <= ifp->if_mtu); 1070 mtmp = mtmp->m_next) { 1071 off += mtmp->m_len; 1072 } 1073 /* 1074 * firstlen (off - hlen) must be aligned on an 1075 * 8-byte boundary 1076 */ 1077 if (off < hlen) 1078 goto smart_frag_failure; 1079 off = ((off - hlen) & ~7) + hlen; 1080 newlen = (~PAGE_MASK) & ifp->if_mtu; 1081 if ((newlen + sizeof (struct ip)) > ifp->if_mtu) { 1082 /* we failed, go back the default */ 1083smart_frag_failure: 1084 newlen = len; 1085 off = hlen + len; 1086 } 1087 1088/* printf("ipfrag: len = %d, hlen = %d, mhlen = %d, newlen = %d, off = %d\n", 1089 len, hlen, sizeof (struct ip), newlen, off);*/ 1090 1091 len = newlen; 1092 1093 } else { 1094 off = hlen + len; 1095 } 1096 1097 1098 1099 { 1100 int mhlen, firstlen = off - hlen; 1101 struct mbuf **mnext = &m->m_nextpkt; 1102 int nfrags = 1; 1103 1104 /* 1105 * Loop through length of segment after first fragment, 1106 * make new header and copy data of each part and link onto chain. 1107 */ 1108 m0 = m; 1109 mhlen = sizeof (struct ip); 1110 for (; off < (u_short)ip->ip_len; off += len) { 1111 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1112 if (m == 0) { 1113 error = ENOBUFS; 1114 ipstat.ips_odropped++; 1115 goto sendorfree; 1116 } 1117 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 1118 m->m_data += max_linkhdr; 1119 mhip = mtod(m, struct ip *); 1120 *mhip = *ip; 1121 if (hlen > sizeof (struct ip)) { 1122 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 1123 mhip->ip_v = IPVERSION; 1124 mhip->ip_hl = mhlen >> 2; 1125 } 1126 m->m_len = mhlen; 1127 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 1128 if (off + len >= (u_short)ip->ip_len) 1129 len = (u_short)ip->ip_len - off; 1130 else 1131 mhip->ip_off |= IP_MF; 1132 mhip->ip_len = htons((u_short)(len + mhlen)); 1133 m->m_next = m_copy(m0, off, len); 1134 if (m->m_next == 0) { 1135 (void) m_free(m); 1136 error = ENOBUFS; /* ??? */ 1137 ipstat.ips_odropped++; 1138 goto sendorfree; 1139 } 1140 m->m_pkthdr.len = mhlen + len; 1141 m->m_pkthdr.rcvif = (struct ifnet *)0; 1142#ifdef MAC 1143 mac_create_fragment(m0, m); 1144#endif 1145 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 1146 mhip->ip_off = htons(mhip->ip_off); 1147 mhip->ip_sum = 0; 1148 if (sw_csum & CSUM_DELAY_IP) 1149 mhip->ip_sum = in_cksum(m, mhlen); 1150 *mnext = m; 1151 mnext = &m->m_nextpkt; 1152 nfrags++; 1153 } 1154 ipstat.ips_ofragments += nfrags; 1155 1156 /* set first/last markers for fragment chain */ 1157 m->m_flags |= M_LASTFRAG; 1158 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 1159 m0->m_pkthdr.csum_data = nfrags; 1160 1161 /* 1162 * Update first fragment by trimming what's been copied out 1163 * and updating header, then send each fragment (in order). 1164 */ 1165 m = m0; 1166 m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 1167 m->m_pkthdr.len = hlen + firstlen; 1168 ip->ip_len = htons((u_short)m->m_pkthdr.len); 1169 ip->ip_off |= IP_MF; 1170 ip->ip_off = htons(ip->ip_off); 1171 ip->ip_sum = 0; 1172 if (sw_csum & CSUM_DELAY_IP) 1173 ip->ip_sum = in_cksum(m, hlen); 1174sendorfree: 1175 for (m = m0; m; m = m0) { 1176 m0 = m->m_nextpkt; 1177 m->m_nextpkt = 0; 1178#ifdef IPSEC 1179 /* clean ipsec history once it goes out of the node */ 1180 ipsec_delaux(m); 1181#endif 1182 if (error == 0) { 1183 /* Record statistics for this interface address. */ 1184 if (ia != NULL) { 1185 ia->ia_ifa.if_opackets++; 1186 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1187 } 1188 1189 error = (*ifp->if_output)(ifp, m, 1190 (struct sockaddr *)dst, ro->ro_rt); 1191 } else 1192 m_freem(m); 1193 } 1194 1195 if (error == 0) 1196 ipstat.ips_fragmented++; 1197 } 1198done: 1199#ifdef IPSEC 1200 if (ro == &iproute && ro->ro_rt) { 1201 RTFREE(ro->ro_rt); 1202 ro->ro_rt = NULL; 1203 } 1204 if (sp != NULL) { 1205 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1206 printf("DP ip_output call free SP:%p\n", sp)); 1207 key_freesp(sp); 1208 } 1209#endif /* IPSEC */ 1210#ifdef FAST_IPSEC 1211 if (ro == &iproute && ro->ro_rt) { 1212 RTFREE(ro->ro_rt); 1213 ro->ro_rt = NULL; 1214 } 1215 if (sp != NULL) 1216 KEY_FREESP(&sp); 1217#endif /* FAST_IPSEC */ 1218 return (error); 1219bad: 1220 m_freem(m); 1221 goto done; 1222} 1223 1224void 1225in_delayed_cksum(struct mbuf *m) 1226{ 1227 struct ip *ip; 1228 u_short csum, offset; 1229 1230 ip = mtod(m, struct ip *); 1231 offset = ip->ip_hl << 2 ; 1232 csum = in_cksum_skip(m, ip->ip_len, offset); 1233 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1234 csum = 0xffff; 1235 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1236 1237 if (offset + sizeof(u_short) > m->m_len) { 1238 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1239 m->m_len, offset, ip->ip_p); 1240 /* 1241 * XXX 1242 * this shouldn't happen, but if it does, the 1243 * correct behavior may be to insert the checksum 1244 * in the existing chain instead of rearranging it. 1245 */ 1246 m = m_pullup(m, offset + sizeof(u_short)); 1247 } 1248 *(u_short *)(m->m_data + offset) = csum; 1249} 1250 1251/* 1252 * Insert IP options into preformed packet. 1253 * Adjust IP destination as required for IP source routing, 1254 * as indicated by a non-zero in_addr at the start of the options. 1255 * 1256 * XXX This routine assumes that the packet has no options in place. 1257 */ 1258static struct mbuf * 1259ip_insertoptions(m, opt, phlen) 1260 register struct mbuf *m; 1261 struct mbuf *opt; 1262 int *phlen; 1263{ 1264 register struct ipoption *p = mtod(opt, struct ipoption *); 1265 struct mbuf *n; 1266 register struct ip *ip = mtod(m, struct ip *); 1267 unsigned optlen; 1268 1269 optlen = opt->m_len - sizeof(p->ipopt_dst); 1270 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) { 1271 *phlen = 0; 1272 return (m); /* XXX should fail */ 1273 } 1274 if (p->ipopt_dst.s_addr) 1275 ip->ip_dst = p->ipopt_dst; 1276 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1277 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1278 if (n == 0) { 1279 *phlen = 0; 1280 return (m); 1281 } 1282 n->m_pkthdr.rcvif = (struct ifnet *)0; 1283#ifdef MAC 1284 mac_create_mbuf_from_mbuf(m, n); 1285#endif 1286 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1287 m->m_len -= sizeof(struct ip); 1288 m->m_data += sizeof(struct ip); 1289 n->m_next = m; 1290 m = n; 1291 m->m_len = optlen + sizeof(struct ip); 1292 m->m_data += max_linkhdr; 1293 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1294 } else { 1295 m->m_data -= optlen; 1296 m->m_len += optlen; 1297 m->m_pkthdr.len += optlen; 1298 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1299 } 1300 ip = mtod(m, struct ip *); 1301 bcopy(p->ipopt_list, ip + 1, optlen); 1302 *phlen = sizeof(struct ip) + optlen; 1303 ip->ip_v = IPVERSION; 1304 ip->ip_hl = *phlen >> 2; 1305 ip->ip_len += optlen; 1306 return (m); 1307} 1308 1309/* 1310 * Copy options from ip to jp, 1311 * omitting those not copied during fragmentation. 1312 */ 1313int 1314ip_optcopy(ip, jp) 1315 struct ip *ip, *jp; 1316{ 1317 register u_char *cp, *dp; 1318 int opt, optlen, cnt; 1319 1320 cp = (u_char *)(ip + 1); 1321 dp = (u_char *)(jp + 1); 1322 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1323 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1324 opt = cp[0]; 1325 if (opt == IPOPT_EOL) 1326 break; 1327 if (opt == IPOPT_NOP) { 1328 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1329 *dp++ = IPOPT_NOP; 1330 optlen = 1; 1331 continue; 1332 } 1333 1334 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1335 ("ip_optcopy: malformed ipv4 option")); 1336 optlen = cp[IPOPT_OLEN]; 1337 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1338 ("ip_optcopy: malformed ipv4 option")); 1339 1340 /* bogus lengths should have been caught by ip_dooptions */ 1341 if (optlen > cnt) 1342 optlen = cnt; 1343 if (IPOPT_COPIED(opt)) { 1344 bcopy(cp, dp, optlen); 1345 dp += optlen; 1346 } 1347 } 1348 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1349 *dp++ = IPOPT_EOL; 1350 return (optlen); 1351} 1352 1353/* 1354 * IP socket option processing. 1355 */ 1356int 1357ip_ctloutput(so, sopt) 1358 struct socket *so; 1359 struct sockopt *sopt; 1360{ 1361 struct inpcb *inp = sotoinpcb(so); 1362 int error, optval; 1363 1364 error = optval = 0; 1365 if (sopt->sopt_level != IPPROTO_IP) { 1366 return (EINVAL); 1367 } 1368 1369 switch (sopt->sopt_dir) { 1370 case SOPT_SET: 1371 switch (sopt->sopt_name) { 1372 case IP_OPTIONS: 1373#ifdef notyet 1374 case IP_RETOPTS: 1375#endif 1376 { 1377 struct mbuf *m; 1378 if (sopt->sopt_valsize > MLEN) { 1379 error = EMSGSIZE; 1380 break; 1381 } 1382 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1383 if (m == 0) { 1384 error = ENOBUFS; 1385 break; 1386 } 1387 m->m_len = sopt->sopt_valsize; 1388 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1389 m->m_len); 1390 1391 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1392 m)); 1393 } 1394 1395 case IP_TOS: 1396 case IP_TTL: 1397 case IP_RECVOPTS: 1398 case IP_RECVRETOPTS: 1399 case IP_RECVDSTADDR: 1400 case IP_RECVIF: 1401 case IP_FAITH: 1402 error = sooptcopyin(sopt, &optval, sizeof optval, 1403 sizeof optval); 1404 if (error) 1405 break; 1406 1407 switch (sopt->sopt_name) { 1408 case IP_TOS: 1409 inp->inp_ip_tos = optval; 1410 break; 1411 1412 case IP_TTL: 1413 inp->inp_ip_ttl = optval; 1414 break; 1415#define OPTSET(bit) \ 1416 if (optval) \ 1417 inp->inp_flags |= bit; \ 1418 else \ 1419 inp->inp_flags &= ~bit; 1420 1421 case IP_RECVOPTS: 1422 OPTSET(INP_RECVOPTS); 1423 break; 1424 1425 case IP_RECVRETOPTS: 1426 OPTSET(INP_RECVRETOPTS); 1427 break; 1428 1429 case IP_RECVDSTADDR: 1430 OPTSET(INP_RECVDSTADDR); 1431 break; 1432 1433 case IP_RECVIF: 1434 OPTSET(INP_RECVIF); 1435 break; 1436 1437 case IP_FAITH: 1438 OPTSET(INP_FAITH); 1439 break; 1440 } 1441 break; 1442#undef OPTSET 1443 1444 case IP_MULTICAST_IF: 1445 case IP_MULTICAST_VIF: 1446 case IP_MULTICAST_TTL: 1447 case IP_MULTICAST_LOOP: 1448 case IP_ADD_MEMBERSHIP: 1449 case IP_DROP_MEMBERSHIP: 1450 error = ip_setmoptions(sopt, &inp->inp_moptions); 1451 break; 1452 1453 case IP_PORTRANGE: 1454 error = sooptcopyin(sopt, &optval, sizeof optval, 1455 sizeof optval); 1456 if (error) 1457 break; 1458 1459 switch (optval) { 1460 case IP_PORTRANGE_DEFAULT: 1461 inp->inp_flags &= ~(INP_LOWPORT); 1462 inp->inp_flags &= ~(INP_HIGHPORT); 1463 break; 1464 1465 case IP_PORTRANGE_HIGH: 1466 inp->inp_flags &= ~(INP_LOWPORT); 1467 inp->inp_flags |= INP_HIGHPORT; 1468 break; 1469 1470 case IP_PORTRANGE_LOW: 1471 inp->inp_flags &= ~(INP_HIGHPORT); 1472 inp->inp_flags |= INP_LOWPORT; 1473 break; 1474 1475 default: 1476 error = EINVAL; 1477 break; 1478 } 1479 break; 1480 1481#if defined(IPSEC) || defined(FAST_IPSEC) 1482 case IP_IPSEC_POLICY: 1483 { 1484 caddr_t req; 1485 size_t len = 0; 1486 int priv; 1487 struct mbuf *m; 1488 int optname; 1489 1490 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1491 break; 1492 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1493 break; 1494 priv = (sopt->sopt_td != NULL && 1495 suser(sopt->sopt_td) != 0) ? 0 : 1; 1496 req = mtod(m, caddr_t); 1497 len = m->m_len; 1498 optname = sopt->sopt_name; 1499 error = ipsec4_set_policy(inp, optname, req, len, priv); 1500 m_freem(m); 1501 break; 1502 } 1503#endif /*IPSEC*/ 1504 1505 default: 1506 error = ENOPROTOOPT; 1507 break; 1508 } 1509 break; 1510 1511 case SOPT_GET: 1512 switch (sopt->sopt_name) { 1513 case IP_OPTIONS: 1514 case IP_RETOPTS: 1515 if (inp->inp_options) 1516 error = sooptcopyout(sopt, 1517 mtod(inp->inp_options, 1518 char *), 1519 inp->inp_options->m_len); 1520 else 1521 sopt->sopt_valsize = 0; 1522 break; 1523 1524 case IP_TOS: 1525 case IP_TTL: 1526 case IP_RECVOPTS: 1527 case IP_RECVRETOPTS: 1528 case IP_RECVDSTADDR: 1529 case IP_RECVIF: 1530 case IP_PORTRANGE: 1531 case IP_FAITH: 1532 switch (sopt->sopt_name) { 1533 1534 case IP_TOS: 1535 optval = inp->inp_ip_tos; 1536 break; 1537 1538 case IP_TTL: 1539 optval = inp->inp_ip_ttl; 1540 break; 1541 1542#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1543 1544 case IP_RECVOPTS: 1545 optval = OPTBIT(INP_RECVOPTS); 1546 break; 1547 1548 case IP_RECVRETOPTS: 1549 optval = OPTBIT(INP_RECVRETOPTS); 1550 break; 1551 1552 case IP_RECVDSTADDR: 1553 optval = OPTBIT(INP_RECVDSTADDR); 1554 break; 1555 1556 case IP_RECVIF: 1557 optval = OPTBIT(INP_RECVIF); 1558 break; 1559 1560 case IP_PORTRANGE: 1561 if (inp->inp_flags & INP_HIGHPORT) 1562 optval = IP_PORTRANGE_HIGH; 1563 else if (inp->inp_flags & INP_LOWPORT) 1564 optval = IP_PORTRANGE_LOW; 1565 else 1566 optval = 0; 1567 break; 1568 1569 case IP_FAITH: 1570 optval = OPTBIT(INP_FAITH); 1571 break; 1572 } 1573 error = sooptcopyout(sopt, &optval, sizeof optval); 1574 break; 1575 1576 case IP_MULTICAST_IF: 1577 case IP_MULTICAST_VIF: 1578 case IP_MULTICAST_TTL: 1579 case IP_MULTICAST_LOOP: 1580 case IP_ADD_MEMBERSHIP: 1581 case IP_DROP_MEMBERSHIP: 1582 error = ip_getmoptions(sopt, inp->inp_moptions); 1583 break; 1584 1585#if defined(IPSEC) || defined(FAST_IPSEC) 1586 case IP_IPSEC_POLICY: 1587 { 1588 struct mbuf *m = NULL; 1589 caddr_t req = NULL; 1590 size_t len = 0; 1591 1592 if (m != 0) { 1593 req = mtod(m, caddr_t); 1594 len = m->m_len; 1595 } 1596 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1597 if (error == 0) 1598 error = soopt_mcopyout(sopt, m); /* XXX */ 1599 if (error == 0) 1600 m_freem(m); 1601 break; 1602 } 1603#endif /*IPSEC*/ 1604 1605 default: 1606 error = ENOPROTOOPT; 1607 break; 1608 } 1609 break; 1610 } 1611 return (error); 1612} 1613 1614/* 1615 * Set up IP options in pcb for insertion in output packets. 1616 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1617 * with destination address if source routed. 1618 */ 1619static int 1620ip_pcbopts(optname, pcbopt, m) 1621 int optname; 1622 struct mbuf **pcbopt; 1623 register struct mbuf *m; 1624{ 1625 register int cnt, optlen; 1626 register u_char *cp; 1627 u_char opt; 1628 1629 /* turn off any old options */ 1630 if (*pcbopt) 1631 (void)m_free(*pcbopt); 1632 *pcbopt = 0; 1633 if (m == (struct mbuf *)0 || m->m_len == 0) { 1634 /* 1635 * Only turning off any previous options. 1636 */ 1637 if (m) 1638 (void)m_free(m); 1639 return (0); 1640 } 1641 1642 if (m->m_len % sizeof(int32_t)) 1643 goto bad; 1644 /* 1645 * IP first-hop destination address will be stored before 1646 * actual options; move other options back 1647 * and clear it when none present. 1648 */ 1649 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1650 goto bad; 1651 cnt = m->m_len; 1652 m->m_len += sizeof(struct in_addr); 1653 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1654 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1655 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1656 1657 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1658 opt = cp[IPOPT_OPTVAL]; 1659 if (opt == IPOPT_EOL) 1660 break; 1661 if (opt == IPOPT_NOP) 1662 optlen = 1; 1663 else { 1664 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1665 goto bad; 1666 optlen = cp[IPOPT_OLEN]; 1667 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1668 goto bad; 1669 } 1670 switch (opt) { 1671 1672 default: 1673 break; 1674 1675 case IPOPT_LSRR: 1676 case IPOPT_SSRR: 1677 /* 1678 * user process specifies route as: 1679 * ->A->B->C->D 1680 * D must be our final destination (but we can't 1681 * check that since we may not have connected yet). 1682 * A is first hop destination, which doesn't appear in 1683 * actual IP option, but is stored before the options. 1684 */ 1685 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1686 goto bad; 1687 m->m_len -= sizeof(struct in_addr); 1688 cnt -= sizeof(struct in_addr); 1689 optlen -= sizeof(struct in_addr); 1690 cp[IPOPT_OLEN] = optlen; 1691 /* 1692 * Move first hop before start of options. 1693 */ 1694 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1695 sizeof(struct in_addr)); 1696 /* 1697 * Then copy rest of options back 1698 * to close up the deleted entry. 1699 */ 1700 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1701 sizeof(struct in_addr)), 1702 (caddr_t)&cp[IPOPT_OFFSET+1], 1703 (unsigned)cnt + sizeof(struct in_addr)); 1704 break; 1705 } 1706 } 1707 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1708 goto bad; 1709 *pcbopt = m; 1710 return (0); 1711 1712bad: 1713 (void)m_free(m); 1714 return (EINVAL); 1715} 1716 1717/* 1718 * XXX 1719 * The whole multicast option thing needs to be re-thought. 1720 * Several of these options are equally applicable to non-multicast 1721 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1722 * standard option (IP_TTL). 1723 */ 1724 1725/* 1726 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1727 */ 1728static struct ifnet * 1729ip_multicast_if(a, ifindexp) 1730 struct in_addr *a; 1731 int *ifindexp; 1732{ 1733 int ifindex; 1734 struct ifnet *ifp; 1735 1736 if (ifindexp) 1737 *ifindexp = 0; 1738 if (ntohl(a->s_addr) >> 24 == 0) { 1739 ifindex = ntohl(a->s_addr) & 0xffffff; 1740 if (ifindex < 0 || if_index < ifindex) 1741 return NULL; 1742 ifp = ifnet_byindex(ifindex); 1743 if (ifindexp) 1744 *ifindexp = ifindex; 1745 } else { 1746 INADDR_TO_IFP(*a, ifp); 1747 } 1748 return ifp; 1749} 1750 1751/* 1752 * Set the IP multicast options in response to user setsockopt(). 1753 */ 1754static int 1755ip_setmoptions(sopt, imop) 1756 struct sockopt *sopt; 1757 struct ip_moptions **imop; 1758{ 1759 int error = 0; 1760 int i; 1761 struct in_addr addr; 1762 struct ip_mreq mreq; 1763 struct ifnet *ifp; 1764 struct ip_moptions *imo = *imop; 1765 struct route ro; 1766 struct sockaddr_in *dst; 1767 int ifindex; 1768 int s; 1769 1770 if (imo == NULL) { 1771 /* 1772 * No multicast option buffer attached to the pcb; 1773 * allocate one and initialize to default values. 1774 */ 1775 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1776 M_WAITOK); 1777 1778 if (imo == NULL) 1779 return (ENOBUFS); 1780 *imop = imo; 1781 imo->imo_multicast_ifp = NULL; 1782 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1783 imo->imo_multicast_vif = -1; 1784 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1785 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1786 imo->imo_num_memberships = 0; 1787 } 1788 1789 switch (sopt->sopt_name) { 1790 /* store an index number for the vif you wanna use in the send */ 1791 case IP_MULTICAST_VIF: 1792 if (legal_vif_num == 0) { 1793 error = EOPNOTSUPP; 1794 break; 1795 } 1796 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1797 if (error) 1798 break; 1799 if (!legal_vif_num(i) && (i != -1)) { 1800 error = EINVAL; 1801 break; 1802 } 1803 imo->imo_multicast_vif = i; 1804 break; 1805 1806 case IP_MULTICAST_IF: 1807 /* 1808 * Select the interface for outgoing multicast packets. 1809 */ 1810 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1811 if (error) 1812 break; 1813 /* 1814 * INADDR_ANY is used to remove a previous selection. 1815 * When no interface is selected, a default one is 1816 * chosen every time a multicast packet is sent. 1817 */ 1818 if (addr.s_addr == INADDR_ANY) { 1819 imo->imo_multicast_ifp = NULL; 1820 break; 1821 } 1822 /* 1823 * The selected interface is identified by its local 1824 * IP address. Find the interface and confirm that 1825 * it supports multicasting. 1826 */ 1827 s = splimp(); 1828 ifp = ip_multicast_if(&addr, &ifindex); 1829 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1830 splx(s); 1831 error = EADDRNOTAVAIL; 1832 break; 1833 } 1834 imo->imo_multicast_ifp = ifp; 1835 if (ifindex) 1836 imo->imo_multicast_addr = addr; 1837 else 1838 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1839 splx(s); 1840 break; 1841 1842 case IP_MULTICAST_TTL: 1843 /* 1844 * Set the IP time-to-live for outgoing multicast packets. 1845 * The original multicast API required a char argument, 1846 * which is inconsistent with the rest of the socket API. 1847 * We allow either a char or an int. 1848 */ 1849 if (sopt->sopt_valsize == 1) { 1850 u_char ttl; 1851 error = sooptcopyin(sopt, &ttl, 1, 1); 1852 if (error) 1853 break; 1854 imo->imo_multicast_ttl = ttl; 1855 } else { 1856 u_int ttl; 1857 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1858 sizeof ttl); 1859 if (error) 1860 break; 1861 if (ttl > 255) 1862 error = EINVAL; 1863 else 1864 imo->imo_multicast_ttl = ttl; 1865 } 1866 break; 1867 1868 case IP_MULTICAST_LOOP: 1869 /* 1870 * Set the loopback flag for outgoing multicast packets. 1871 * Must be zero or one. The original multicast API required a 1872 * char argument, which is inconsistent with the rest 1873 * of the socket API. We allow either a char or an int. 1874 */ 1875 if (sopt->sopt_valsize == 1) { 1876 u_char loop; 1877 error = sooptcopyin(sopt, &loop, 1, 1); 1878 if (error) 1879 break; 1880 imo->imo_multicast_loop = !!loop; 1881 } else { 1882 u_int loop; 1883 error = sooptcopyin(sopt, &loop, sizeof loop, 1884 sizeof loop); 1885 if (error) 1886 break; 1887 imo->imo_multicast_loop = !!loop; 1888 } 1889 break; 1890 1891 case IP_ADD_MEMBERSHIP: 1892 /* 1893 * Add a multicast group membership. 1894 * Group must be a valid IP multicast address. 1895 */ 1896 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1897 if (error) 1898 break; 1899 1900 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1901 error = EINVAL; 1902 break; 1903 } 1904 s = splimp(); 1905 /* 1906 * If no interface address was provided, use the interface of 1907 * the route to the given multicast address. 1908 */ 1909 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1910 bzero((caddr_t)&ro, sizeof(ro)); 1911 dst = (struct sockaddr_in *)&ro.ro_dst; 1912 dst->sin_len = sizeof(*dst); 1913 dst->sin_family = AF_INET; 1914 dst->sin_addr = mreq.imr_multiaddr; 1915 rtalloc(&ro); 1916 if (ro.ro_rt == NULL) { 1917 error = EADDRNOTAVAIL; 1918 splx(s); 1919 break; 1920 } 1921 ifp = ro.ro_rt->rt_ifp; 1922 rtfree(ro.ro_rt); 1923 } 1924 else { 1925 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1926 } 1927 1928 /* 1929 * See if we found an interface, and confirm that it 1930 * supports multicast. 1931 */ 1932 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1933 error = EADDRNOTAVAIL; 1934 splx(s); 1935 break; 1936 } 1937 /* 1938 * See if the membership already exists or if all the 1939 * membership slots are full. 1940 */ 1941 for (i = 0; i < imo->imo_num_memberships; ++i) { 1942 if (imo->imo_membership[i]->inm_ifp == ifp && 1943 imo->imo_membership[i]->inm_addr.s_addr 1944 == mreq.imr_multiaddr.s_addr) 1945 break; 1946 } 1947 if (i < imo->imo_num_memberships) { 1948 error = EADDRINUSE; 1949 splx(s); 1950 break; 1951 } 1952 if (i == IP_MAX_MEMBERSHIPS) { 1953 error = ETOOMANYREFS; 1954 splx(s); 1955 break; 1956 } 1957 /* 1958 * Everything looks good; add a new record to the multicast 1959 * address list for the given interface. 1960 */ 1961 if ((imo->imo_membership[i] = 1962 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1963 error = ENOBUFS; 1964 splx(s); 1965 break; 1966 } 1967 ++imo->imo_num_memberships; 1968 splx(s); 1969 break; 1970 1971 case IP_DROP_MEMBERSHIP: 1972 /* 1973 * Drop a multicast group membership. 1974 * Group must be a valid IP multicast address. 1975 */ 1976 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1977 if (error) 1978 break; 1979 1980 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1981 error = EINVAL; 1982 break; 1983 } 1984 1985 s = splimp(); 1986 /* 1987 * If an interface address was specified, get a pointer 1988 * to its ifnet structure. 1989 */ 1990 if (mreq.imr_interface.s_addr == INADDR_ANY) 1991 ifp = NULL; 1992 else { 1993 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1994 if (ifp == NULL) { 1995 error = EADDRNOTAVAIL; 1996 splx(s); 1997 break; 1998 } 1999 } 2000 /* 2001 * Find the membership in the membership array. 2002 */ 2003 for (i = 0; i < imo->imo_num_memberships; ++i) { 2004 if ((ifp == NULL || 2005 imo->imo_membership[i]->inm_ifp == ifp) && 2006 imo->imo_membership[i]->inm_addr.s_addr == 2007 mreq.imr_multiaddr.s_addr) 2008 break; 2009 } 2010 if (i == imo->imo_num_memberships) { 2011 error = EADDRNOTAVAIL; 2012 splx(s); 2013 break; 2014 } 2015 /* 2016 * Give up the multicast address record to which the 2017 * membership points. 2018 */ 2019 in_delmulti(imo->imo_membership[i]); 2020 /* 2021 * Remove the gap in the membership array. 2022 */ 2023 for (++i; i < imo->imo_num_memberships; ++i) 2024 imo->imo_membership[i-1] = imo->imo_membership[i]; 2025 --imo->imo_num_memberships; 2026 splx(s); 2027 break; 2028 2029 default: 2030 error = EOPNOTSUPP; 2031 break; 2032 } 2033 2034 /* 2035 * If all options have default values, no need to keep the mbuf. 2036 */ 2037 if (imo->imo_multicast_ifp == NULL && 2038 imo->imo_multicast_vif == -1 && 2039 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2040 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2041 imo->imo_num_memberships == 0) { 2042 free(*imop, M_IPMOPTS); 2043 *imop = NULL; 2044 } 2045 2046 return (error); 2047} 2048 2049/* 2050 * Return the IP multicast options in response to user getsockopt(). 2051 */ 2052static int 2053ip_getmoptions(sopt, imo) 2054 struct sockopt *sopt; 2055 register struct ip_moptions *imo; 2056{ 2057 struct in_addr addr; 2058 struct in_ifaddr *ia; 2059 int error, optval; 2060 u_char coptval; 2061 2062 error = 0; 2063 switch (sopt->sopt_name) { 2064 case IP_MULTICAST_VIF: 2065 if (imo != NULL) 2066 optval = imo->imo_multicast_vif; 2067 else 2068 optval = -1; 2069 error = sooptcopyout(sopt, &optval, sizeof optval); 2070 break; 2071 2072 case IP_MULTICAST_IF: 2073 if (imo == NULL || imo->imo_multicast_ifp == NULL) 2074 addr.s_addr = INADDR_ANY; 2075 else if (imo->imo_multicast_addr.s_addr) { 2076 /* return the value user has set */ 2077 addr = imo->imo_multicast_addr; 2078 } else { 2079 IFP_TO_IA(imo->imo_multicast_ifp, ia); 2080 addr.s_addr = (ia == NULL) ? INADDR_ANY 2081 : IA_SIN(ia)->sin_addr.s_addr; 2082 } 2083 error = sooptcopyout(sopt, &addr, sizeof addr); 2084 break; 2085 2086 case IP_MULTICAST_TTL: 2087 if (imo == 0) 2088 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 2089 else 2090 optval = coptval = imo->imo_multicast_ttl; 2091 if (sopt->sopt_valsize == 1) 2092 error = sooptcopyout(sopt, &coptval, 1); 2093 else 2094 error = sooptcopyout(sopt, &optval, sizeof optval); 2095 break; 2096 2097 case IP_MULTICAST_LOOP: 2098 if (imo == 0) 2099 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 2100 else 2101 optval = coptval = imo->imo_multicast_loop; 2102 if (sopt->sopt_valsize == 1) 2103 error = sooptcopyout(sopt, &coptval, 1); 2104 else 2105 error = sooptcopyout(sopt, &optval, sizeof optval); 2106 break; 2107 2108 default: 2109 error = ENOPROTOOPT; 2110 break; 2111 } 2112 return (error); 2113} 2114 2115/* 2116 * Discard the IP multicast options. 2117 */ 2118void 2119ip_freemoptions(imo) 2120 register struct ip_moptions *imo; 2121{ 2122 register int i; 2123 2124 if (imo != NULL) { 2125 for (i = 0; i < imo->imo_num_memberships; ++i) 2126 in_delmulti(imo->imo_membership[i]); 2127 free(imo, M_IPMOPTS); 2128 } 2129} 2130 2131/* 2132 * Routine called from ip_output() to loop back a copy of an IP multicast 2133 * packet to the input queue of a specified interface. Note that this 2134 * calls the output routine of the loopback "driver", but with an interface 2135 * pointer that might NOT be a loopback interface -- evil, but easier than 2136 * replicating that code here. 2137 */ 2138static void 2139ip_mloopback(ifp, m, dst, hlen) 2140 struct ifnet *ifp; 2141 register struct mbuf *m; 2142 register struct sockaddr_in *dst; 2143 int hlen; 2144{ 2145 register struct ip *ip; 2146 struct mbuf *copym; 2147 2148 copym = m_copy(m, 0, M_COPYALL); 2149 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 2150 copym = m_pullup(copym, hlen); 2151 if (copym != NULL) { 2152 /* 2153 * We don't bother to fragment if the IP length is greater 2154 * than the interface's MTU. Can this possibly matter? 2155 */ 2156 ip = mtod(copym, struct ip *); 2157 ip->ip_len = htons(ip->ip_len); 2158 ip->ip_off = htons(ip->ip_off); 2159 ip->ip_sum = 0; 2160 ip->ip_sum = in_cksum(copym, hlen); 2161 /* 2162 * NB: 2163 * It's not clear whether there are any lingering 2164 * reentrancy problems in other areas which might 2165 * be exposed by using ip_input directly (in 2166 * particular, everything which modifies the packet 2167 * in-place). Yet another option is using the 2168 * protosw directly to deliver the looped back 2169 * packet. For the moment, we'll err on the side 2170 * of safety by using if_simloop(). 2171 */ 2172#if 1 /* XXX */ 2173 if (dst->sin_family != AF_INET) { 2174 printf("ip_mloopback: bad address family %d\n", 2175 dst->sin_family); 2176 dst->sin_family = AF_INET; 2177 } 2178#endif 2179 2180#ifdef notdef 2181 copym->m_pkthdr.rcvif = ifp; 2182 ip_input(copym); 2183#else 2184 /* if the checksum hasn't been computed, mark it as valid */ 2185 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2186 copym->m_pkthdr.csum_flags |= 2187 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 2188 copym->m_pkthdr.csum_data = 0xffff; 2189 } 2190 if_simloop(ifp, copym, dst->sin_family, 0); 2191#endif 2192 } 2193} 2194