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 * $FreeBSD$ 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 #include "opt_mbuf_stress_test.h" 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/kernel.h> 50 #include <sys/mac.h> 51 #include <sys/malloc.h> 52 #include <sys/mbuf.h> 53 #include <sys/protosw.h> 54 #include <sys/socket.h> 55 #include <sys/socketvar.h> 56 #include <sys/sysctl.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 #ifdef PFIL_HOOKS 69 #include <net/pfil.h> 70 #endif 71 72 #include <machine/in_cksum.h> 73 74 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 75 76 #ifdef IPSEC 77 #include <netinet6/ipsec.h> 78 #include <netkey/key.h> 79 #ifdef IPSEC_DEBUG 80 #include <netkey/key_debug.h> 81 #else 82 #define KEYDEBUG(lev,arg) 83 #endif 84 #endif /*IPSEC*/ 85 86 #ifdef FAST_IPSEC 87 #include <netipsec/ipsec.h> 88 #include <netipsec/xform.h> 89 #include <netipsec/key.h> 90 #endif /*FAST_IPSEC*/ 91 92 #include <netinet/ip_fw.h> 93 #include <netinet/ip_dummynet.h> 94 95 #define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 96 x, (ntohl(a.s_addr)>>24)&0xFF,\ 97 (ntohl(a.s_addr)>>16)&0xFF,\ 98 (ntohl(a.s_addr)>>8)&0xFF,\ 99 (ntohl(a.s_addr))&0xFF, y); 100 101 u_short ip_id; 102 103 #ifdef MBUF_STRESS_TEST 104 int mbuf_frag_size = 0; 105 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 106 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 107 #endif 108 109 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 110 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 111 static void ip_mloopback 112 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 113 static int ip_getmoptions 114 (struct sockopt *, struct ip_moptions *); 115 static int ip_pcbopts(int, struct mbuf **, struct mbuf *); 116 static int ip_setmoptions 117 (struct sockopt *, struct ip_moptions **); 118 119 int ip_optcopy(struct ip *, struct ip *); 120 121 122 extern struct protosw inetsw[]; 123 124 /* 125 * IP output. The packet in mbuf chain m contains a skeletal IP 126 * header (with len, off, ttl, proto, tos, src, dst). 127 * The mbuf chain containing the packet will be freed. 128 * The mbuf opt, if present, will not be freed. 129 * In the IP forwarding case, the packet will arrive with options already 130 * inserted, so must have a NULL opt pointer. 131 */ 132 int 133 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, 134 int flags, struct ip_moptions *imo, struct inpcb *inp) 135 { 136 struct ip *ip; 137 struct ifnet *ifp = NULL; /* keep compiler happy */ 138 struct mbuf *m; 139 int hlen = sizeof (struct ip); 140 int len, off, error = 0; 141 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 142 struct in_ifaddr *ia = NULL; 143 int isbroadcast, sw_csum; 144 struct in_addr pkt_dst; 145 #ifdef IPSEC 146 struct route iproute; 147 struct socket *so; 148 struct secpolicy *sp = NULL; 149 #endif 150 #ifdef FAST_IPSEC 151 struct route iproute; 152 struct m_tag *mtag; 153 struct secpolicy *sp = NULL; 154 struct tdb_ident *tdbi; 155 int s; 156 #endif /* FAST_IPSEC */ 157 struct ip_fw_args args; 158 int src_was_INADDR_ANY = 0; /* as the name says... */ 159 160 args.eh = NULL; 161 args.rule = NULL; 162 args.next_hop = NULL; 163 args.divert_rule = 0; /* divert cookie */ 164 165 /* Grab info from MT_TAG mbufs prepended to the chain. */ 166 for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) { 167 switch(m0->_m_tag_id) { 168 default: 169 printf("ip_output: unrecognised MT_TAG tag %d\n", 170 m0->_m_tag_id); 171 break; 172 173 case PACKET_TAG_DUMMYNET: 174 /* 175 * the packet was already tagged, so part of the 176 * processing was already done, and we need to go down. 177 * Get parameters from the header. 178 */ 179 args.rule = ((struct dn_pkt *)m0)->rule; 180 opt = NULL ; 181 ro = & ( ((struct dn_pkt *)m0)->ro ) ; 182 imo = NULL ; 183 dst = ((struct dn_pkt *)m0)->dn_dst ; 184 ifp = ((struct dn_pkt *)m0)->ifp ; 185 flags = ((struct dn_pkt *)m0)->flags ; 186 break; 187 188 case PACKET_TAG_DIVERT: 189 args.divert_rule = (intptr_t)m0->m_data & 0xffff; 190 break; 191 192 case PACKET_TAG_IPFORWARD: 193 args.next_hop = (struct sockaddr_in *)m0->m_data; 194 break; 195 } 196 } 197 m = m0; 198 199 #ifdef IPSEC 200 so = ipsec_getsocket(m); 201 (void)ipsec_setsocket(m, NULL); 202 #endif /*IPSEC*/ 203 204 M_ASSERTPKTHDR(m); 205 #ifndef FAST_IPSEC 206 KASSERT(ro != NULL, ("ip_output: no route, proto %d", 207 mtod(m, struct ip *)->ip_p)); 208 #endif 209 210 if (args.rule != NULL) { /* dummynet already saw us */ 211 ip = mtod(m, struct ip *); 212 hlen = ip->ip_hl << 2 ; 213 if (ro->ro_rt) 214 ia = ifatoia(ro->ro_rt->rt_ifa); 215 goto sendit; 216 } 217 218 if (opt) { 219 len = 0; 220 m = ip_insertoptions(m, opt, &len); 221 if (len != 0) 222 hlen = len; 223 } 224 ip = mtod(m, struct ip *); 225 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 226 227 /* 228 * Fill in IP header. If we are not allowing fragmentation, 229 * then the ip_id field is meaningless, so send it as zero 230 * to reduce information leakage. Otherwise, if we are not 231 * randomizing ip_id, then don't bother to convert it to network 232 * byte order -- it's just a nonce. Note that a 16-bit counter 233 * will wrap around in less than 10 seconds at 100 Mbit/s on a 234 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 235 * for Counting NATted Hosts", Proc. IMW'02, available at 236 * <http://www.research.att.com/~smb/papers/fnat.pdf>. 237 */ 238 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 239 ip->ip_v = IPVERSION; 240 ip->ip_hl = hlen >> 2; 241 if ((ip->ip_off & IP_DF) == 0) { 242 ip->ip_off = 0; 243 #ifdef RANDOM_IP_ID 244 ip->ip_id = ip_randomid(); 245 #else 246 ip->ip_id = ip_id++; 247 #endif 248 } else { 249 ip->ip_off = IP_DF; 250 ip->ip_id = 0; 251 } 252 ipstat.ips_localout++; 253 } else { 254 hlen = ip->ip_hl << 2; 255 } 256 257 #ifdef FAST_IPSEC 258 if (ro == NULL) { 259 ro = &iproute; 260 bzero(ro, sizeof (*ro)); 261 } 262 #endif /* FAST_IPSEC */ 263 dst = (struct sockaddr_in *)&ro->ro_dst; 264 /* 265 * If there is a cached route, 266 * check that it is to the same destination 267 * and is still up. If not, free it and try again. 268 * The address family should also be checked in case of sharing the 269 * cache with IPv6. 270 */ 271 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 272 dst->sin_family != AF_INET || 273 dst->sin_addr.s_addr != pkt_dst.s_addr)) { 274 RTFREE(ro->ro_rt); 275 ro->ro_rt = (struct rtentry *)0; 276 } 277 if (ro->ro_rt == 0) { 278 bzero(dst, sizeof(*dst)); 279 dst->sin_family = AF_INET; 280 dst->sin_len = sizeof(*dst); 281 dst->sin_addr = pkt_dst; 282 } 283 /* 284 * If routing to interface only, 285 * short circuit routing lookup. 286 */ 287 if (flags & IP_ROUTETOIF) { 288 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 289 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 290 ipstat.ips_noroute++; 291 error = ENETUNREACH; 292 goto bad; 293 } 294 ifp = ia->ia_ifp; 295 ip->ip_ttl = 1; 296 isbroadcast = in_broadcast(dst->sin_addr, ifp); 297 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 298 imo != NULL && imo->imo_multicast_ifp != NULL) { 299 /* 300 * Bypass the normal routing lookup for multicast 301 * packets if the interface is specified. 302 */ 303 ifp = imo->imo_multicast_ifp; 304 IFP_TO_IA(ifp, ia); 305 isbroadcast = 0; /* fool gcc */ 306 } else { 307 /* 308 * If this is the case, we probably don't want to allocate 309 * a protocol-cloned route since we didn't get one from the 310 * ULP. This lets TCP do its thing, while not burdening 311 * forwarding or ICMP with the overhead of cloning a route. 312 * Of course, we still want to do any cloning requested by 313 * the link layer, as this is probably required in all cases 314 * for correct operation (as it is for ARP). 315 */ 316 if (ro->ro_rt == 0) 317 rtalloc_ign(ro, RTF_PRCLONING); 318 if (ro->ro_rt == 0) { 319 ipstat.ips_noroute++; 320 error = EHOSTUNREACH; 321 goto bad; 322 } 323 ia = ifatoia(ro->ro_rt->rt_ifa); 324 ifp = ro->ro_rt->rt_ifp; 325 ro->ro_rt->rt_use++; 326 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 327 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 328 if (ro->ro_rt->rt_flags & RTF_HOST) 329 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 330 else 331 isbroadcast = in_broadcast(dst->sin_addr, ifp); 332 } 333 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 334 struct in_multi *inm; 335 336 m->m_flags |= M_MCAST; 337 /* 338 * IP destination address is multicast. Make sure "dst" 339 * still points to the address in "ro". (It may have been 340 * changed to point to a gateway address, above.) 341 */ 342 dst = (struct sockaddr_in *)&ro->ro_dst; 343 /* 344 * See if the caller provided any multicast options 345 */ 346 if (imo != NULL) { 347 ip->ip_ttl = imo->imo_multicast_ttl; 348 if (imo->imo_multicast_vif != -1) 349 ip->ip_src.s_addr = 350 ip_mcast_src ? 351 ip_mcast_src(imo->imo_multicast_vif) : 352 INADDR_ANY; 353 } else 354 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 355 /* 356 * Confirm that the outgoing interface supports multicast. 357 */ 358 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 359 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 360 ipstat.ips_noroute++; 361 error = ENETUNREACH; 362 goto bad; 363 } 364 } 365 /* 366 * If source address not specified yet, use address 367 * of outgoing interface. 368 */ 369 if (ip->ip_src.s_addr == INADDR_ANY) { 370 /* Interface may have no addresses. */ 371 if (ia != NULL) 372 ip->ip_src = IA_SIN(ia)->sin_addr; 373 } 374 375 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 376 /* 377 * XXX 378 * delayed checksums are not currently 379 * compatible with IP multicast routing 380 */ 381 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 382 in_delayed_cksum(m); 383 m->m_pkthdr.csum_flags &= 384 ~CSUM_DELAY_DATA; 385 } 386 } 387 IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 388 if (inm != NULL && 389 (imo == NULL || imo->imo_multicast_loop)) { 390 /* 391 * If we belong to the destination multicast group 392 * on the outgoing interface, and the caller did not 393 * forbid loopback, loop back a copy. 394 */ 395 ip_mloopback(ifp, m, dst, hlen); 396 } 397 else { 398 /* 399 * If we are acting as a multicast router, perform 400 * multicast forwarding as if the packet had just 401 * arrived on the interface to which we are about 402 * to send. The multicast forwarding function 403 * recursively calls this function, using the 404 * IP_FORWARDING flag to prevent infinite recursion. 405 * 406 * Multicasts that are looped back by ip_mloopback(), 407 * above, will be forwarded by the ip_input() routine, 408 * if necessary. 409 */ 410 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 411 /* 412 * If rsvp daemon is not running, do not 413 * set ip_moptions. This ensures that the packet 414 * is multicast and not just sent down one link 415 * as prescribed by rsvpd. 416 */ 417 if (!rsvp_on) 418 imo = NULL; 419 if (ip_mforward && 420 ip_mforward(ip, ifp, m, imo) != 0) { 421 m_freem(m); 422 goto done; 423 } 424 } 425 } 426 427 /* 428 * Multicasts with a time-to-live of zero may be looped- 429 * back, above, but must not be transmitted on a network. 430 * Also, multicasts addressed to the loopback interface 431 * are not sent -- the above call to ip_mloopback() will 432 * loop back a copy if this host actually belongs to the 433 * destination group on the loopback interface. 434 */ 435 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 436 m_freem(m); 437 goto done; 438 } 439 440 goto sendit; 441 } 442 #ifndef notdef 443 /* 444 * If the source address is not specified yet, use the address 445 * of the outoing interface. In case, keep note we did that, so 446 * if the the firewall changes the next-hop causing the output 447 * interface to change, we can fix that. 448 */ 449 if (ip->ip_src.s_addr == INADDR_ANY) { 450 /* Interface may have no addresses. */ 451 if (ia != NULL) { 452 ip->ip_src = IA_SIN(ia)->sin_addr; 453 src_was_INADDR_ANY = 1; 454 } 455 } 456 #endif /* notdef */ 457 /* 458 * Verify that we have any chance at all of being able to queue 459 * the packet or packet fragments 460 */ 461 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 462 ifp->if_snd.ifq_maxlen) { 463 error = ENOBUFS; 464 ipstat.ips_odropped++; 465 goto bad; 466 } 467 468 /* 469 * Look for broadcast address and 470 * verify user is allowed to send 471 * such a packet. 472 */ 473 if (isbroadcast) { 474 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 475 error = EADDRNOTAVAIL; 476 goto bad; 477 } 478 if ((flags & IP_ALLOWBROADCAST) == 0) { 479 error = EACCES; 480 goto bad; 481 } 482 /* don't allow broadcast messages to be fragmented */ 483 if (ip->ip_len > ifp->if_mtu) { 484 error = EMSGSIZE; 485 goto bad; 486 } 487 if (flags & IP_SENDONES) 488 ip->ip_dst.s_addr = INADDR_BROADCAST; 489 m->m_flags |= M_BCAST; 490 } else { 491 m->m_flags &= ~M_BCAST; 492 } 493 494 sendit: 495 #ifdef IPSEC 496 /* get SP for this packet */ 497 if (so == NULL) 498 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 499 flags, &error); 500 else 501 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 502 503 if (sp == NULL) { 504 ipsecstat.out_inval++; 505 goto bad; 506 } 507 508 error = 0; 509 510 /* check policy */ 511 switch (sp->policy) { 512 case IPSEC_POLICY_DISCARD: 513 /* 514 * This packet is just discarded. 515 */ 516 ipsecstat.out_polvio++; 517 goto bad; 518 519 case IPSEC_POLICY_BYPASS: 520 case IPSEC_POLICY_NONE: 521 /* no need to do IPsec. */ 522 goto skip_ipsec; 523 524 case IPSEC_POLICY_IPSEC: 525 if (sp->req == NULL) { 526 /* acquire a policy */ 527 error = key_spdacquire(sp); 528 goto bad; 529 } 530 break; 531 532 case IPSEC_POLICY_ENTRUST: 533 default: 534 printf("ip_output: Invalid policy found. %d\n", sp->policy); 535 } 536 { 537 struct ipsec_output_state state; 538 bzero(&state, sizeof(state)); 539 state.m = m; 540 if (flags & IP_ROUTETOIF) { 541 state.ro = &iproute; 542 bzero(&iproute, sizeof(iproute)); 543 } else 544 state.ro = ro; 545 state.dst = (struct sockaddr *)dst; 546 547 ip->ip_sum = 0; 548 549 /* 550 * XXX 551 * delayed checksums are not currently compatible with IPsec 552 */ 553 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 554 in_delayed_cksum(m); 555 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 556 } 557 558 ip->ip_len = htons(ip->ip_len); 559 ip->ip_off = htons(ip->ip_off); 560 561 error = ipsec4_output(&state, sp, flags); 562 563 m = state.m; 564 if (flags & IP_ROUTETOIF) { 565 /* 566 * if we have tunnel mode SA, we may need to ignore 567 * IP_ROUTETOIF. 568 */ 569 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 570 flags &= ~IP_ROUTETOIF; 571 ro = state.ro; 572 } 573 } else 574 ro = state.ro; 575 dst = (struct sockaddr_in *)state.dst; 576 if (error) { 577 /* mbuf is already reclaimed in ipsec4_output. */ 578 m0 = NULL; 579 switch (error) { 580 case EHOSTUNREACH: 581 case ENETUNREACH: 582 case EMSGSIZE: 583 case ENOBUFS: 584 case ENOMEM: 585 break; 586 default: 587 printf("ip4_output (ipsec): error code %d\n", error); 588 /*fall through*/ 589 case ENOENT: 590 /* don't show these error codes to the user */ 591 error = 0; 592 break; 593 } 594 goto bad; 595 } 596 } 597 598 /* be sure to update variables that are affected by ipsec4_output() */ 599 ip = mtod(m, struct ip *); 600 hlen = ip->ip_hl << 2; 601 if (ro->ro_rt == NULL) { 602 if ((flags & IP_ROUTETOIF) == 0) { 603 printf("ip_output: " 604 "can't update route after IPsec processing\n"); 605 error = EHOSTUNREACH; /*XXX*/ 606 goto bad; 607 } 608 } else { 609 ia = ifatoia(ro->ro_rt->rt_ifa); 610 ifp = ro->ro_rt->rt_ifp; 611 } 612 613 /* make it flipped, again. */ 614 ip->ip_len = ntohs(ip->ip_len); 615 ip->ip_off = ntohs(ip->ip_off); 616 skip_ipsec: 617 #endif /*IPSEC*/ 618 #ifdef FAST_IPSEC 619 /* 620 * Check the security policy (SP) for the packet and, if 621 * required, do IPsec-related processing. There are two 622 * cases here; the first time a packet is sent through 623 * it will be untagged and handled by ipsec4_checkpolicy. 624 * If the packet is resubmitted to ip_output (e.g. after 625 * AH, ESP, etc. processing), there will be a tag to bypass 626 * the lookup and related policy checking. 627 */ 628 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 629 s = splnet(); 630 if (mtag != NULL) { 631 tdbi = (struct tdb_ident *)(mtag + 1); 632 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 633 if (sp == NULL) 634 error = -EINVAL; /* force silent drop */ 635 m_tag_delete(m, mtag); 636 } else { 637 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 638 &error, inp); 639 } 640 /* 641 * There are four return cases: 642 * sp != NULL apply IPsec policy 643 * sp == NULL, error == 0 no IPsec handling needed 644 * sp == NULL, error == -EINVAL discard packet w/o error 645 * sp == NULL, error != 0 discard packet, report error 646 */ 647 if (sp != NULL) { 648 /* Loop detection, check if ipsec processing already done */ 649 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 650 for (mtag = m_tag_first(m); mtag != NULL; 651 mtag = m_tag_next(m, mtag)) { 652 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 653 continue; 654 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 655 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 656 continue; 657 /* 658 * Check if policy has an SA associated with it. 659 * This can happen when an SP has yet to acquire 660 * an SA; e.g. on first reference. If it occurs, 661 * then we let ipsec4_process_packet do its thing. 662 */ 663 if (sp->req->sav == NULL) 664 break; 665 tdbi = (struct tdb_ident *)(mtag + 1); 666 if (tdbi->spi == sp->req->sav->spi && 667 tdbi->proto == sp->req->sav->sah->saidx.proto && 668 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 669 sizeof (union sockaddr_union)) == 0) { 670 /* 671 * No IPsec processing is needed, free 672 * reference to SP. 673 * 674 * NB: null pointer to avoid free at 675 * done: below. 676 */ 677 KEY_FREESP(&sp), sp = NULL; 678 splx(s); 679 goto spd_done; 680 } 681 } 682 683 /* 684 * Do delayed checksums now because we send before 685 * this is done in the normal processing path. 686 */ 687 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 688 in_delayed_cksum(m); 689 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 690 } 691 692 ip->ip_len = htons(ip->ip_len); 693 ip->ip_off = htons(ip->ip_off); 694 695 /* NB: callee frees mbuf */ 696 error = ipsec4_process_packet(m, sp->req, flags, 0); 697 /* 698 * Preserve KAME behaviour: ENOENT can be returned 699 * when an SA acquire is in progress. Don't propagate 700 * this to user-level; it confuses applications. 701 * 702 * XXX this will go away when the SADB is redone. 703 */ 704 if (error == ENOENT) 705 error = 0; 706 splx(s); 707 goto done; 708 } else { 709 splx(s); 710 711 if (error != 0) { 712 /* 713 * Hack: -EINVAL is used to signal that a packet 714 * should be silently discarded. This is typically 715 * because we asked key management for an SA and 716 * it was delayed (e.g. kicked up to IKE). 717 */ 718 if (error == -EINVAL) 719 error = 0; 720 goto bad; 721 } else { 722 /* No IPsec processing for this packet. */ 723 } 724 #ifdef notyet 725 /* 726 * If deferred crypto processing is needed, check that 727 * the interface supports it. 728 */ 729 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 730 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 731 /* notify IPsec to do its own crypto */ 732 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 733 error = EHOSTUNREACH; 734 goto bad; 735 } 736 #endif 737 } 738 spd_done: 739 #endif /* FAST_IPSEC */ 740 741 /* 742 * IpHack's section. 743 * - Xlate: translate packet's addr/port (NAT). 744 * - Firewall: deny/allow/etc. 745 * - Wrap: fake packet's addr/port <unimpl.> 746 * - Encapsulate: put it in another IP and send out. <unimp.> 747 */ 748 #ifdef PFIL_HOOKS 749 /* 750 * Run through list of hooks for output packets. 751 */ 752 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT); 753 if (error != 0 || m == NULL) 754 goto done; 755 ip = mtod(m, struct ip *); 756 #endif /* PFIL_HOOKS */ 757 758 /* 759 * Check with the firewall... 760 * but not if we are already being fwd'd from a firewall. 761 */ 762 if (fw_enable && IPFW_LOADED && !args.next_hop) { 763 struct sockaddr_in *old = dst; 764 765 args.m = m; 766 args.next_hop = dst; 767 args.oif = ifp; 768 off = ip_fw_chk_ptr(&args); 769 m = args.m; 770 dst = args.next_hop; 771 772 /* 773 * On return we must do the following: 774 * m == NULL -> drop the pkt (old interface, deprecated) 775 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 776 * 1<=off<= 0xffff -> DIVERT 777 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 778 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 779 * dst != old -> IPFIREWALL_FORWARD 780 * off==0, dst==old -> accept 781 * If some of the above modules are not compiled in, then 782 * we should't have to check the corresponding condition 783 * (because the ipfw control socket should not accept 784 * unsupported rules), but better play safe and drop 785 * packets in case of doubt. 786 */ 787 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) { 788 if (m) 789 m_freem(m); 790 error = EACCES; 791 goto done; 792 } 793 ip = mtod(m, struct ip *); 794 if (off == 0 && dst == old) /* common case */ 795 goto pass; 796 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 797 /* 798 * pass the pkt to dummynet. Need to include 799 * pipe number, m, ifp, ro, dst because these are 800 * not recomputed in the next pass. 801 * All other parameters have been already used and 802 * so they are not needed anymore. 803 * XXX note: if the ifp or ro entry are deleted 804 * while a pkt is in dummynet, we are in trouble! 805 */ 806 args.ro = ro; 807 args.dst = dst; 808 args.flags = flags; 809 810 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, 811 &args); 812 goto done; 813 } 814 #ifdef IPDIVERT 815 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 816 struct mbuf *clone = NULL; 817 818 /* Clone packet if we're doing a 'tee' */ 819 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 820 clone = m_dup(m, M_DONTWAIT); 821 822 /* 823 * XXX 824 * delayed checksums are not currently compatible 825 * with divert sockets. 826 */ 827 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 828 in_delayed_cksum(m); 829 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 830 } 831 832 /* Restore packet header fields to original values */ 833 ip->ip_len = htons(ip->ip_len); 834 ip->ip_off = htons(ip->ip_off); 835 836 /* Deliver packet to divert input routine */ 837 divert_packet(m, 0, off & 0xffff, args.divert_rule); 838 839 /* If 'tee', continue with original packet */ 840 if (clone != NULL) { 841 m = clone; 842 ip = mtod(m, struct ip *); 843 goto pass; 844 } 845 goto done; 846 } 847 #endif 848 849 /* IPFIREWALL_FORWARD */ 850 /* 851 * Check dst to make sure it is directly reachable on the 852 * interface we previously thought it was. 853 * If it isn't (which may be likely in some situations) we have 854 * to re-route it (ie, find a route for the next-hop and the 855 * associated interface) and set them here. This is nested 856 * forwarding which in most cases is undesirable, except where 857 * such control is nigh impossible. So we do it here. 858 * And I'm babbling. 859 */ 860 if (off == 0 && old != dst) { /* FORWARD, dst has changed */ 861 #if 0 862 /* 863 * XXX To improve readability, this block should be 864 * changed into a function call as below: 865 */ 866 error = ip_ipforward(&m, &dst, &ifp); 867 if (error) 868 goto bad; 869 if (m == NULL) /* ip_input consumed the mbuf */ 870 goto done; 871 #else 872 struct in_ifaddr *ia; 873 874 /* 875 * XXX sro_fwd below is static, and a pointer 876 * to it gets passed to routines downstream. 877 * This could have surprisingly bad results in 878 * practice, because its content is overwritten 879 * by subsequent packets. 880 */ 881 /* There must be a better way to do this next line... */ 882 static struct route sro_fwd; 883 struct route *ro_fwd = &sro_fwd; 884 885 #if 0 886 print_ip("IPFIREWALL_FORWARD: New dst ip: ", 887 dst->sin_addr, "\n"); 888 #endif 889 890 /* 891 * We need to figure out if we have been forwarded 892 * to a local socket. If so, then we should somehow 893 * "loop back" to ip_input, and get directed to the 894 * PCB as if we had received this packet. This is 895 * because it may be dificult to identify the packets 896 * you want to forward until they are being output 897 * and have selected an interface. (e.g. locally 898 * initiated packets) If we used the loopback inteface, 899 * we would not be able to control what happens 900 * as the packet runs through ip_input() as 901 * it is done through an ISR. 902 */ 903 LIST_FOREACH(ia, 904 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 905 /* 906 * If the addr to forward to is one 907 * of ours, we pretend to 908 * be the destination for this packet. 909 */ 910 if (IA_SIN(ia)->sin_addr.s_addr == 911 dst->sin_addr.s_addr) 912 break; 913 } 914 if (ia) { /* tell ip_input "dont filter" */ 915 struct m_hdr tag; 916 917 tag.mh_type = MT_TAG; 918 tag.mh_flags = PACKET_TAG_IPFORWARD; 919 tag.mh_data = (caddr_t)args.next_hop; 920 tag.mh_next = m; 921 922 if (m->m_pkthdr.rcvif == NULL) 923 m->m_pkthdr.rcvif = ifunit("lo0"); 924 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 925 m->m_pkthdr.csum_flags |= 926 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 927 m0->m_pkthdr.csum_data = 0xffff; 928 } 929 m->m_pkthdr.csum_flags |= 930 CSUM_IP_CHECKED | CSUM_IP_VALID; 931 ip->ip_len = htons(ip->ip_len); 932 ip->ip_off = htons(ip->ip_off); 933 ip_input((struct mbuf *)&tag); 934 goto done; 935 } 936 /* Some of the logic for this was 937 * nicked from above. 938 * 939 * This rewrites the cached route in a local PCB. 940 * Is this what we want to do? 941 */ 942 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 943 944 ro_fwd->ro_rt = 0; 945 rtalloc_ign(ro_fwd, RTF_PRCLONING); 946 947 if (ro_fwd->ro_rt == 0) { 948 ipstat.ips_noroute++; 949 error = EHOSTUNREACH; 950 goto bad; 951 } 952 953 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 954 ifp = ro_fwd->ro_rt->rt_ifp; 955 ro_fwd->ro_rt->rt_use++; 956 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 957 dst = (struct sockaddr_in *) 958 ro_fwd->ro_rt->rt_gateway; 959 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 960 isbroadcast = 961 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 962 else 963 isbroadcast = in_broadcast(dst->sin_addr, ifp); 964 if (ro->ro_rt) 965 RTFREE(ro->ro_rt); 966 ro->ro_rt = ro_fwd->ro_rt; 967 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 968 969 #endif /* ... block to be put into a function */ 970 /* 971 * If we added a default src ip earlier, 972 * which would have been gotten from the-then 973 * interface, do it again, from the new one. 974 */ 975 if (src_was_INADDR_ANY) 976 ip->ip_src = IA_SIN(ia)->sin_addr; 977 goto pass ; 978 } 979 980 /* 981 * if we get here, none of the above matches, and 982 * we have to drop the pkt 983 */ 984 m_freem(m); 985 error = EACCES; /* not sure this is the right error msg */ 986 goto done; 987 } 988 989 pass: 990 /* 127/8 must not appear on wire - RFC1122. */ 991 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 992 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 993 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 994 ipstat.ips_badaddr++; 995 error = EADDRNOTAVAIL; 996 goto bad; 997 } 998 } 999 1000 m->m_pkthdr.csum_flags |= CSUM_IP; 1001 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 1002 if (sw_csum & CSUM_DELAY_DATA) { 1003 in_delayed_cksum(m); 1004 sw_csum &= ~CSUM_DELAY_DATA; 1005 } 1006 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 1007 1008 /* 1009 * If small enough for interface, or the interface will take 1010 * care of the fragmentation for us, can just send directly. 1011 */ 1012 if (ip->ip_len <= ifp->if_mtu || ifp->if_hwassist & CSUM_FRAGMENT) { 1013 ip->ip_len = htons(ip->ip_len); 1014 ip->ip_off = htons(ip->ip_off); 1015 ip->ip_sum = 0; 1016 if (sw_csum & CSUM_DELAY_IP) 1017 ip->ip_sum = in_cksum(m, hlen); 1018 1019 /* Record statistics for this interface address. */ 1020 if (!(flags & IP_FORWARDING) && ia) { 1021 ia->ia_ifa.if_opackets++; 1022 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1023 } 1024 1025 #ifdef IPSEC 1026 /* clean ipsec history once it goes out of the node */ 1027 ipsec_delaux(m); 1028 #endif 1029 1030 #ifdef MBUF_STRESS_TEST 1031 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 1032 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 1033 #endif 1034 error = (*ifp->if_output)(ifp, m, 1035 (struct sockaddr *)dst, ro->ro_rt); 1036 goto done; 1037 } 1038 1039 if (ip->ip_off & IP_DF) { 1040 error = EMSGSIZE; 1041 /* 1042 * This case can happen if the user changed the MTU 1043 * of an interface after enabling IP on it. Because 1044 * most netifs don't keep track of routes pointing to 1045 * them, there is no way for one to update all its 1046 * routes when the MTU is changed. 1047 */ 1048 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 1049 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) && 1050 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 1051 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 1052 } 1053 ipstat.ips_cantfrag++; 1054 goto bad; 1055 } 1056 1057 /* 1058 * Too large for interface; fragment if possible. If successful, 1059 * on return, m will point to a list of packets to be sent. 1060 */ 1061 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 1062 if (error) 1063 goto bad; 1064 for (; m; m = m0) { 1065 m0 = m->m_nextpkt; 1066 m->m_nextpkt = 0; 1067 #ifdef IPSEC 1068 /* clean ipsec history once it goes out of the node */ 1069 ipsec_delaux(m); 1070 #endif 1071 if (error == 0) { 1072 /* Record statistics for this interface address. */ 1073 if (ia != NULL) { 1074 ia->ia_ifa.if_opackets++; 1075 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1076 } 1077 1078 error = (*ifp->if_output)(ifp, m, 1079 (struct sockaddr *)dst, ro->ro_rt); 1080 } else 1081 m_freem(m); 1082 } 1083 1084 if (error == 0) 1085 ipstat.ips_fragmented++; 1086 1087 done: 1088 #ifdef IPSEC 1089 if (ro == &iproute && ro->ro_rt) { 1090 RTFREE(ro->ro_rt); 1091 ro->ro_rt = NULL; 1092 } 1093 if (sp != NULL) { 1094 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1095 printf("DP ip_output call free SP:%p\n", sp)); 1096 key_freesp(sp); 1097 } 1098 #endif 1099 #ifdef FAST_IPSEC 1100 if (ro == &iproute && ro->ro_rt) { 1101 RTFREE(ro->ro_rt); 1102 ro->ro_rt = NULL; 1103 } 1104 if (sp != NULL) 1105 KEY_FREESP(&sp); 1106 #endif 1107 return (error); 1108 bad: 1109 m_freem(m); 1110 goto done; 1111 } 1112 1113 /* 1114 * Create a chain of fragments which fit the given mtu. m_frag points to the 1115 * mbuf to be fragmented; on return it points to the chain with the fragments. 1116 * Return 0 if no error. If error, m_frag may contain a partially built 1117 * chain of fragments that should be freed by the caller. 1118 * 1119 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 1120 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 1121 */ 1122 int 1123 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 1124 u_long if_hwassist_flags, int sw_csum) 1125 { 1126 int error = 0; 1127 int hlen = ip->ip_hl << 2; 1128 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 1129 int off; 1130 struct mbuf *m0 = *m_frag; /* the original packet */ 1131 int firstlen; 1132 struct mbuf **mnext; 1133 int nfrags; 1134 1135 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 1136 ipstat.ips_cantfrag++; 1137 return EMSGSIZE; 1138 } 1139 1140 /* 1141 * Must be able to put at least 8 bytes per fragment. 1142 */ 1143 if (len < 8) 1144 return EMSGSIZE; 1145 1146 /* 1147 * If the interface will not calculate checksums on 1148 * fragmented packets, then do it here. 1149 */ 1150 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 1151 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 1152 in_delayed_cksum(m0); 1153 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1154 } 1155 1156 if (len > PAGE_SIZE) { 1157 /* 1158 * Fragment large datagrams such that each segment 1159 * contains a multiple of PAGE_SIZE amount of data, 1160 * plus headers. This enables a receiver to perform 1161 * page-flipping zero-copy optimizations. 1162 * 1163 * XXX When does this help given that sender and receiver 1164 * could have different page sizes, and also mtu could 1165 * be less than the receiver's page size ? 1166 */ 1167 int newlen; 1168 struct mbuf *m; 1169 1170 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 1171 off += m->m_len; 1172 1173 /* 1174 * firstlen (off - hlen) must be aligned on an 1175 * 8-byte boundary 1176 */ 1177 if (off < hlen) 1178 goto smart_frag_failure; 1179 off = ((off - hlen) & ~7) + hlen; 1180 newlen = (~PAGE_MASK) & mtu; 1181 if ((newlen + sizeof (struct ip)) > mtu) { 1182 /* we failed, go back the default */ 1183 smart_frag_failure: 1184 newlen = len; 1185 off = hlen + len; 1186 } 1187 len = newlen; 1188 1189 } else { 1190 off = hlen + len; 1191 } 1192 1193 firstlen = off - hlen; 1194 mnext = &m0->m_nextpkt; /* pointer to next packet */ 1195 1196 /* 1197 * Loop through length of segment after first fragment, 1198 * make new header and copy data of each part and link onto chain. 1199 * Here, m0 is the original packet, m is the fragment being created. 1200 * The fragments are linked off the m_nextpkt of the original 1201 * packet, which after processing serves as the first fragment. 1202 */ 1203 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 1204 struct ip *mhip; /* ip header on the fragment */ 1205 struct mbuf *m; 1206 int mhlen = sizeof (struct ip); 1207 1208 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1209 if (m == 0) { 1210 error = ENOBUFS; 1211 ipstat.ips_odropped++; 1212 goto done; 1213 } 1214 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 1215 /* 1216 * In the first mbuf, leave room for the link header, then 1217 * copy the original IP header including options. The payload 1218 * goes into an additional mbuf chain returned by m_copy(). 1219 */ 1220 m->m_data += max_linkhdr; 1221 mhip = mtod(m, struct ip *); 1222 *mhip = *ip; 1223 if (hlen > sizeof (struct ip)) { 1224 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 1225 mhip->ip_v = IPVERSION; 1226 mhip->ip_hl = mhlen >> 2; 1227 } 1228 m->m_len = mhlen; 1229 /* XXX do we need to add ip->ip_off below ? */ 1230 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 1231 if (off + len >= ip->ip_len) { /* last fragment */ 1232 len = ip->ip_len - off; 1233 m->m_flags |= M_LASTFRAG; 1234 } else 1235 mhip->ip_off |= IP_MF; 1236 mhip->ip_len = htons((u_short)(len + mhlen)); 1237 m->m_next = m_copy(m0, off, len); 1238 if (m->m_next == 0) { /* copy failed */ 1239 m_free(m); 1240 error = ENOBUFS; /* ??? */ 1241 ipstat.ips_odropped++; 1242 goto done; 1243 } 1244 m->m_pkthdr.len = mhlen + len; 1245 m->m_pkthdr.rcvif = (struct ifnet *)0; 1246 #ifdef MAC 1247 mac_create_fragment(m0, m); 1248 #endif 1249 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 1250 mhip->ip_off = htons(mhip->ip_off); 1251 mhip->ip_sum = 0; 1252 if (sw_csum & CSUM_DELAY_IP) 1253 mhip->ip_sum = in_cksum(m, mhlen); 1254 *mnext = m; 1255 mnext = &m->m_nextpkt; 1256 } 1257 ipstat.ips_ofragments += nfrags; 1258 1259 /* set first marker for fragment chain */ 1260 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 1261 m0->m_pkthdr.csum_data = nfrags; 1262 1263 /* 1264 * Update first fragment by trimming what's been copied out 1265 * and updating header. 1266 */ 1267 m_adj(m0, hlen + firstlen - ip->ip_len); 1268 m0->m_pkthdr.len = hlen + firstlen; 1269 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 1270 ip->ip_off |= IP_MF; 1271 ip->ip_off = htons(ip->ip_off); 1272 ip->ip_sum = 0; 1273 if (sw_csum & CSUM_DELAY_IP) 1274 ip->ip_sum = in_cksum(m0, hlen); 1275 1276 done: 1277 *m_frag = m0; 1278 return error; 1279 } 1280 1281 void 1282 in_delayed_cksum(struct mbuf *m) 1283 { 1284 struct ip *ip; 1285 u_short csum, offset; 1286 1287 ip = mtod(m, struct ip *); 1288 offset = ip->ip_hl << 2 ; 1289 csum = in_cksum_skip(m, ip->ip_len, offset); 1290 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1291 csum = 0xffff; 1292 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1293 1294 if (offset + sizeof(u_short) > m->m_len) { 1295 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1296 m->m_len, offset, ip->ip_p); 1297 /* 1298 * XXX 1299 * this shouldn't happen, but if it does, the 1300 * correct behavior may be to insert the checksum 1301 * in the existing chain instead of rearranging it. 1302 */ 1303 m = m_pullup(m, offset + sizeof(u_short)); 1304 } 1305 *(u_short *)(m->m_data + offset) = csum; 1306 } 1307 1308 /* 1309 * Insert IP options into preformed packet. 1310 * Adjust IP destination as required for IP source routing, 1311 * as indicated by a non-zero in_addr at the start of the options. 1312 * 1313 * XXX This routine assumes that the packet has no options in place. 1314 */ 1315 static struct mbuf * 1316 ip_insertoptions(m, opt, phlen) 1317 register struct mbuf *m; 1318 struct mbuf *opt; 1319 int *phlen; 1320 { 1321 register struct ipoption *p = mtod(opt, struct ipoption *); 1322 struct mbuf *n; 1323 register struct ip *ip = mtod(m, struct ip *); 1324 unsigned optlen; 1325 1326 optlen = opt->m_len - sizeof(p->ipopt_dst); 1327 if (optlen + ip->ip_len > IP_MAXPACKET) { 1328 *phlen = 0; 1329 return (m); /* XXX should fail */ 1330 } 1331 if (p->ipopt_dst.s_addr) 1332 ip->ip_dst = p->ipopt_dst; 1333 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1334 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1335 if (n == 0) { 1336 *phlen = 0; 1337 return (m); 1338 } 1339 n->m_pkthdr.rcvif = (struct ifnet *)0; 1340 #ifdef MAC 1341 mac_create_mbuf_from_mbuf(m, n); 1342 #endif 1343 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1344 m->m_len -= sizeof(struct ip); 1345 m->m_data += sizeof(struct ip); 1346 n->m_next = m; 1347 m = n; 1348 m->m_len = optlen + sizeof(struct ip); 1349 m->m_data += max_linkhdr; 1350 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1351 } else { 1352 m->m_data -= optlen; 1353 m->m_len += optlen; 1354 m->m_pkthdr.len += optlen; 1355 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1356 } 1357 ip = mtod(m, struct ip *); 1358 bcopy(p->ipopt_list, ip + 1, optlen); 1359 *phlen = sizeof(struct ip) + optlen; 1360 ip->ip_v = IPVERSION; 1361 ip->ip_hl = *phlen >> 2; 1362 ip->ip_len += optlen; 1363 return (m); 1364 } 1365 1366 /* 1367 * Copy options from ip to jp, 1368 * omitting those not copied during fragmentation. 1369 */ 1370 int 1371 ip_optcopy(ip, jp) 1372 struct ip *ip, *jp; 1373 { 1374 register u_char *cp, *dp; 1375 int opt, optlen, cnt; 1376 1377 cp = (u_char *)(ip + 1); 1378 dp = (u_char *)(jp + 1); 1379 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1380 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1381 opt = cp[0]; 1382 if (opt == IPOPT_EOL) 1383 break; 1384 if (opt == IPOPT_NOP) { 1385 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1386 *dp++ = IPOPT_NOP; 1387 optlen = 1; 1388 continue; 1389 } 1390 1391 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1392 ("ip_optcopy: malformed ipv4 option")); 1393 optlen = cp[IPOPT_OLEN]; 1394 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1395 ("ip_optcopy: malformed ipv4 option")); 1396 1397 /* bogus lengths should have been caught by ip_dooptions */ 1398 if (optlen > cnt) 1399 optlen = cnt; 1400 if (IPOPT_COPIED(opt)) { 1401 bcopy(cp, dp, optlen); 1402 dp += optlen; 1403 } 1404 } 1405 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1406 *dp++ = IPOPT_EOL; 1407 return (optlen); 1408 } 1409 1410 /* 1411 * IP socket option processing. 1412 */ 1413 int 1414 ip_ctloutput(so, sopt) 1415 struct socket *so; 1416 struct sockopt *sopt; 1417 { 1418 struct inpcb *inp = sotoinpcb(so); 1419 int error, optval; 1420 1421 error = optval = 0; 1422 if (sopt->sopt_level != IPPROTO_IP) { 1423 return (EINVAL); 1424 } 1425 1426 switch (sopt->sopt_dir) { 1427 case SOPT_SET: 1428 switch (sopt->sopt_name) { 1429 case IP_OPTIONS: 1430 #ifdef notyet 1431 case IP_RETOPTS: 1432 #endif 1433 { 1434 struct mbuf *m; 1435 if (sopt->sopt_valsize > MLEN) { 1436 error = EMSGSIZE; 1437 break; 1438 } 1439 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1440 if (m == 0) { 1441 error = ENOBUFS; 1442 break; 1443 } 1444 m->m_len = sopt->sopt_valsize; 1445 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1446 m->m_len); 1447 1448 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1449 m)); 1450 } 1451 1452 case IP_TOS: 1453 case IP_TTL: 1454 case IP_RECVOPTS: 1455 case IP_RECVRETOPTS: 1456 case IP_RECVDSTADDR: 1457 case IP_RECVTTL: 1458 case IP_RECVIF: 1459 case IP_FAITH: 1460 case IP_ONESBCAST: 1461 error = sooptcopyin(sopt, &optval, sizeof optval, 1462 sizeof optval); 1463 if (error) 1464 break; 1465 1466 switch (sopt->sopt_name) { 1467 case IP_TOS: 1468 inp->inp_ip_tos = optval; 1469 break; 1470 1471 case IP_TTL: 1472 inp->inp_ip_ttl = optval; 1473 break; 1474 #define OPTSET(bit) \ 1475 if (optval) \ 1476 inp->inp_flags |= bit; \ 1477 else \ 1478 inp->inp_flags &= ~bit; 1479 1480 case IP_RECVOPTS: 1481 OPTSET(INP_RECVOPTS); 1482 break; 1483 1484 case IP_RECVRETOPTS: 1485 OPTSET(INP_RECVRETOPTS); 1486 break; 1487 1488 case IP_RECVDSTADDR: 1489 OPTSET(INP_RECVDSTADDR); 1490 break; 1491 1492 case IP_RECVTTL: 1493 OPTSET(INP_RECVTTL); 1494 break; 1495 1496 case IP_RECVIF: 1497 OPTSET(INP_RECVIF); 1498 break; 1499 1500 case IP_FAITH: 1501 OPTSET(INP_FAITH); 1502 break; 1503 1504 case IP_ONESBCAST: 1505 OPTSET(INP_ONESBCAST); 1506 break; 1507 } 1508 break; 1509 #undef OPTSET 1510 1511 case IP_MULTICAST_IF: 1512 case IP_MULTICAST_VIF: 1513 case IP_MULTICAST_TTL: 1514 case IP_MULTICAST_LOOP: 1515 case IP_ADD_MEMBERSHIP: 1516 case IP_DROP_MEMBERSHIP: 1517 error = ip_setmoptions(sopt, &inp->inp_moptions); 1518 break; 1519 1520 case IP_PORTRANGE: 1521 error = sooptcopyin(sopt, &optval, sizeof optval, 1522 sizeof optval); 1523 if (error) 1524 break; 1525 1526 switch (optval) { 1527 case IP_PORTRANGE_DEFAULT: 1528 inp->inp_flags &= ~(INP_LOWPORT); 1529 inp->inp_flags &= ~(INP_HIGHPORT); 1530 break; 1531 1532 case IP_PORTRANGE_HIGH: 1533 inp->inp_flags &= ~(INP_LOWPORT); 1534 inp->inp_flags |= INP_HIGHPORT; 1535 break; 1536 1537 case IP_PORTRANGE_LOW: 1538 inp->inp_flags &= ~(INP_HIGHPORT); 1539 inp->inp_flags |= INP_LOWPORT; 1540 break; 1541 1542 default: 1543 error = EINVAL; 1544 break; 1545 } 1546 break; 1547 1548 #if defined(IPSEC) || defined(FAST_IPSEC) 1549 case IP_IPSEC_POLICY: 1550 { 1551 caddr_t req; 1552 size_t len = 0; 1553 int priv; 1554 struct mbuf *m; 1555 int optname; 1556 1557 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1558 break; 1559 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1560 break; 1561 priv = (sopt->sopt_td != NULL && 1562 suser(sopt->sopt_td) != 0) ? 0 : 1; 1563 req = mtod(m, caddr_t); 1564 len = m->m_len; 1565 optname = sopt->sopt_name; 1566 error = ipsec4_set_policy(inp, optname, req, len, priv); 1567 m_freem(m); 1568 break; 1569 } 1570 #endif /*IPSEC*/ 1571 1572 default: 1573 error = ENOPROTOOPT; 1574 break; 1575 } 1576 break; 1577 1578 case SOPT_GET: 1579 switch (sopt->sopt_name) { 1580 case IP_OPTIONS: 1581 case IP_RETOPTS: 1582 if (inp->inp_options) 1583 error = sooptcopyout(sopt, 1584 mtod(inp->inp_options, 1585 char *), 1586 inp->inp_options->m_len); 1587 else 1588 sopt->sopt_valsize = 0; 1589 break; 1590 1591 case IP_TOS: 1592 case IP_TTL: 1593 case IP_RECVOPTS: 1594 case IP_RECVRETOPTS: 1595 case IP_RECVDSTADDR: 1596 case IP_RECVTTL: 1597 case IP_RECVIF: 1598 case IP_PORTRANGE: 1599 case IP_FAITH: 1600 case IP_ONESBCAST: 1601 switch (sopt->sopt_name) { 1602 1603 case IP_TOS: 1604 optval = inp->inp_ip_tos; 1605 break; 1606 1607 case IP_TTL: 1608 optval = inp->inp_ip_ttl; 1609 break; 1610 1611 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1612 1613 case IP_RECVOPTS: 1614 optval = OPTBIT(INP_RECVOPTS); 1615 break; 1616 1617 case IP_RECVRETOPTS: 1618 optval = OPTBIT(INP_RECVRETOPTS); 1619 break; 1620 1621 case IP_RECVDSTADDR: 1622 optval = OPTBIT(INP_RECVDSTADDR); 1623 break; 1624 1625 case IP_RECVTTL: 1626 optval = OPTBIT(INP_RECVTTL); 1627 break; 1628 1629 case IP_RECVIF: 1630 optval = OPTBIT(INP_RECVIF); 1631 break; 1632 1633 case IP_PORTRANGE: 1634 if (inp->inp_flags & INP_HIGHPORT) 1635 optval = IP_PORTRANGE_HIGH; 1636 else if (inp->inp_flags & INP_LOWPORT) 1637 optval = IP_PORTRANGE_LOW; 1638 else 1639 optval = 0; 1640 break; 1641 1642 case IP_FAITH: 1643 optval = OPTBIT(INP_FAITH); 1644 break; 1645 1646 case IP_ONESBCAST: 1647 optval = OPTBIT(INP_ONESBCAST); 1648 break; 1649 } 1650 error = sooptcopyout(sopt, &optval, sizeof optval); 1651 break; 1652 1653 case IP_MULTICAST_IF: 1654 case IP_MULTICAST_VIF: 1655 case IP_MULTICAST_TTL: 1656 case IP_MULTICAST_LOOP: 1657 case IP_ADD_MEMBERSHIP: 1658 case IP_DROP_MEMBERSHIP: 1659 error = ip_getmoptions(sopt, inp->inp_moptions); 1660 break; 1661 1662 #if defined(IPSEC) || defined(FAST_IPSEC) 1663 case IP_IPSEC_POLICY: 1664 { 1665 struct mbuf *m = NULL; 1666 caddr_t req = NULL; 1667 size_t len = 0; 1668 1669 if (m != 0) { 1670 req = mtod(m, caddr_t); 1671 len = m->m_len; 1672 } 1673 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1674 if (error == 0) 1675 error = soopt_mcopyout(sopt, m); /* XXX */ 1676 if (error == 0) 1677 m_freem(m); 1678 break; 1679 } 1680 #endif /*IPSEC*/ 1681 1682 default: 1683 error = ENOPROTOOPT; 1684 break; 1685 } 1686 break; 1687 } 1688 return (error); 1689 } 1690 1691 /* 1692 * Set up IP options in pcb for insertion in output packets. 1693 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1694 * with destination address if source routed. 1695 */ 1696 static int 1697 ip_pcbopts(optname, pcbopt, m) 1698 int optname; 1699 struct mbuf **pcbopt; 1700 register struct mbuf *m; 1701 { 1702 register int cnt, optlen; 1703 register u_char *cp; 1704 u_char opt; 1705 1706 /* turn off any old options */ 1707 if (*pcbopt) 1708 (void)m_free(*pcbopt); 1709 *pcbopt = 0; 1710 if (m == (struct mbuf *)0 || m->m_len == 0) { 1711 /* 1712 * Only turning off any previous options. 1713 */ 1714 if (m) 1715 (void)m_free(m); 1716 return (0); 1717 } 1718 1719 if (m->m_len % sizeof(int32_t)) 1720 goto bad; 1721 /* 1722 * IP first-hop destination address will be stored before 1723 * actual options; move other options back 1724 * and clear it when none present. 1725 */ 1726 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1727 goto bad; 1728 cnt = m->m_len; 1729 m->m_len += sizeof(struct in_addr); 1730 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1731 bcopy(mtod(m, void *), cp, (unsigned)cnt); 1732 bzero(mtod(m, void *), sizeof(struct in_addr)); 1733 1734 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1735 opt = cp[IPOPT_OPTVAL]; 1736 if (opt == IPOPT_EOL) 1737 break; 1738 if (opt == IPOPT_NOP) 1739 optlen = 1; 1740 else { 1741 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1742 goto bad; 1743 optlen = cp[IPOPT_OLEN]; 1744 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1745 goto bad; 1746 } 1747 switch (opt) { 1748 1749 default: 1750 break; 1751 1752 case IPOPT_LSRR: 1753 case IPOPT_SSRR: 1754 /* 1755 * user process specifies route as: 1756 * ->A->B->C->D 1757 * D must be our final destination (but we can't 1758 * check that since we may not have connected yet). 1759 * A is first hop destination, which doesn't appear in 1760 * actual IP option, but is stored before the options. 1761 */ 1762 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1763 goto bad; 1764 m->m_len -= sizeof(struct in_addr); 1765 cnt -= sizeof(struct in_addr); 1766 optlen -= sizeof(struct in_addr); 1767 cp[IPOPT_OLEN] = optlen; 1768 /* 1769 * Move first hop before start of options. 1770 */ 1771 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1772 sizeof(struct in_addr)); 1773 /* 1774 * Then copy rest of options back 1775 * to close up the deleted entry. 1776 */ 1777 bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1778 &cp[IPOPT_OFFSET+1], 1779 (unsigned)cnt + sizeof(struct in_addr)); 1780 break; 1781 } 1782 } 1783 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1784 goto bad; 1785 *pcbopt = m; 1786 return (0); 1787 1788 bad: 1789 (void)m_free(m); 1790 return (EINVAL); 1791 } 1792 1793 /* 1794 * XXX 1795 * The whole multicast option thing needs to be re-thought. 1796 * Several of these options are equally applicable to non-multicast 1797 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1798 * standard option (IP_TTL). 1799 */ 1800 1801 /* 1802 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1803 */ 1804 static struct ifnet * 1805 ip_multicast_if(a, ifindexp) 1806 struct in_addr *a; 1807 int *ifindexp; 1808 { 1809 int ifindex; 1810 struct ifnet *ifp; 1811 1812 if (ifindexp) 1813 *ifindexp = 0; 1814 if (ntohl(a->s_addr) >> 24 == 0) { 1815 ifindex = ntohl(a->s_addr) & 0xffffff; 1816 if (ifindex < 0 || if_index < ifindex) 1817 return NULL; 1818 ifp = ifnet_byindex(ifindex); 1819 if (ifindexp) 1820 *ifindexp = ifindex; 1821 } else { 1822 INADDR_TO_IFP(*a, ifp); 1823 } 1824 return ifp; 1825 } 1826 1827 /* 1828 * Set the IP multicast options in response to user setsockopt(). 1829 */ 1830 static int 1831 ip_setmoptions(sopt, imop) 1832 struct sockopt *sopt; 1833 struct ip_moptions **imop; 1834 { 1835 int error = 0; 1836 int i; 1837 struct in_addr addr; 1838 struct ip_mreq mreq; 1839 struct ifnet *ifp; 1840 struct ip_moptions *imo = *imop; 1841 struct route ro; 1842 struct sockaddr_in *dst; 1843 int ifindex; 1844 int s; 1845 1846 if (imo == NULL) { 1847 /* 1848 * No multicast option buffer attached to the pcb; 1849 * allocate one and initialize to default values. 1850 */ 1851 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1852 M_WAITOK); 1853 1854 if (imo == NULL) 1855 return (ENOBUFS); 1856 *imop = imo; 1857 imo->imo_multicast_ifp = NULL; 1858 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1859 imo->imo_multicast_vif = -1; 1860 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1861 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1862 imo->imo_num_memberships = 0; 1863 } 1864 1865 switch (sopt->sopt_name) { 1866 /* store an index number for the vif you wanna use in the send */ 1867 case IP_MULTICAST_VIF: 1868 if (legal_vif_num == 0) { 1869 error = EOPNOTSUPP; 1870 break; 1871 } 1872 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1873 if (error) 1874 break; 1875 if (!legal_vif_num(i) && (i != -1)) { 1876 error = EINVAL; 1877 break; 1878 } 1879 imo->imo_multicast_vif = i; 1880 break; 1881 1882 case IP_MULTICAST_IF: 1883 /* 1884 * Select the interface for outgoing multicast packets. 1885 */ 1886 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1887 if (error) 1888 break; 1889 /* 1890 * INADDR_ANY is used to remove a previous selection. 1891 * When no interface is selected, a default one is 1892 * chosen every time a multicast packet is sent. 1893 */ 1894 if (addr.s_addr == INADDR_ANY) { 1895 imo->imo_multicast_ifp = NULL; 1896 break; 1897 } 1898 /* 1899 * The selected interface is identified by its local 1900 * IP address. Find the interface and confirm that 1901 * it supports multicasting. 1902 */ 1903 s = splimp(); 1904 ifp = ip_multicast_if(&addr, &ifindex); 1905 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1906 splx(s); 1907 error = EADDRNOTAVAIL; 1908 break; 1909 } 1910 imo->imo_multicast_ifp = ifp; 1911 if (ifindex) 1912 imo->imo_multicast_addr = addr; 1913 else 1914 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1915 splx(s); 1916 break; 1917 1918 case IP_MULTICAST_TTL: 1919 /* 1920 * Set the IP time-to-live for outgoing multicast packets. 1921 * The original multicast API required a char argument, 1922 * which is inconsistent with the rest of the socket API. 1923 * We allow either a char or an int. 1924 */ 1925 if (sopt->sopt_valsize == 1) { 1926 u_char ttl; 1927 error = sooptcopyin(sopt, &ttl, 1, 1); 1928 if (error) 1929 break; 1930 imo->imo_multicast_ttl = ttl; 1931 } else { 1932 u_int ttl; 1933 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1934 sizeof ttl); 1935 if (error) 1936 break; 1937 if (ttl > 255) 1938 error = EINVAL; 1939 else 1940 imo->imo_multicast_ttl = ttl; 1941 } 1942 break; 1943 1944 case IP_MULTICAST_LOOP: 1945 /* 1946 * Set the loopback flag for outgoing multicast packets. 1947 * Must be zero or one. The original multicast API required a 1948 * char argument, which is inconsistent with the rest 1949 * of the socket API. We allow either a char or an int. 1950 */ 1951 if (sopt->sopt_valsize == 1) { 1952 u_char loop; 1953 error = sooptcopyin(sopt, &loop, 1, 1); 1954 if (error) 1955 break; 1956 imo->imo_multicast_loop = !!loop; 1957 } else { 1958 u_int loop; 1959 error = sooptcopyin(sopt, &loop, sizeof loop, 1960 sizeof loop); 1961 if (error) 1962 break; 1963 imo->imo_multicast_loop = !!loop; 1964 } 1965 break; 1966 1967 case IP_ADD_MEMBERSHIP: 1968 /* 1969 * Add a multicast group membership. 1970 * Group must be a valid IP multicast address. 1971 */ 1972 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1973 if (error) 1974 break; 1975 1976 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1977 error = EINVAL; 1978 break; 1979 } 1980 s = splimp(); 1981 /* 1982 * If no interface address was provided, use the interface of 1983 * the route to the given multicast address. 1984 */ 1985 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1986 bzero((caddr_t)&ro, sizeof(ro)); 1987 dst = (struct sockaddr_in *)&ro.ro_dst; 1988 dst->sin_len = sizeof(*dst); 1989 dst->sin_family = AF_INET; 1990 dst->sin_addr = mreq.imr_multiaddr; 1991 rtalloc(&ro); 1992 if (ro.ro_rt == NULL) { 1993 error = EADDRNOTAVAIL; 1994 splx(s); 1995 break; 1996 } 1997 ifp = ro.ro_rt->rt_ifp; 1998 RTFREE(ro.ro_rt); 1999 } 2000 else { 2001 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 2002 } 2003 2004 /* 2005 * See if we found an interface, and confirm that it 2006 * supports multicast. 2007 */ 2008 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2009 error = EADDRNOTAVAIL; 2010 splx(s); 2011 break; 2012 } 2013 /* 2014 * See if the membership already exists or if all the 2015 * membership slots are full. 2016 */ 2017 for (i = 0; i < imo->imo_num_memberships; ++i) { 2018 if (imo->imo_membership[i]->inm_ifp == ifp && 2019 imo->imo_membership[i]->inm_addr.s_addr 2020 == mreq.imr_multiaddr.s_addr) 2021 break; 2022 } 2023 if (i < imo->imo_num_memberships) { 2024 error = EADDRINUSE; 2025 splx(s); 2026 break; 2027 } 2028 if (i == IP_MAX_MEMBERSHIPS) { 2029 error = ETOOMANYREFS; 2030 splx(s); 2031 break; 2032 } 2033 /* 2034 * Everything looks good; add a new record to the multicast 2035 * address list for the given interface. 2036 */ 2037 if ((imo->imo_membership[i] = 2038 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 2039 error = ENOBUFS; 2040 splx(s); 2041 break; 2042 } 2043 ++imo->imo_num_memberships; 2044 splx(s); 2045 break; 2046 2047 case IP_DROP_MEMBERSHIP: 2048 /* 2049 * Drop a multicast group membership. 2050 * Group must be a valid IP multicast address. 2051 */ 2052 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 2053 if (error) 2054 break; 2055 2056 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 2057 error = EINVAL; 2058 break; 2059 } 2060 2061 s = splimp(); 2062 /* 2063 * If an interface address was specified, get a pointer 2064 * to its ifnet structure. 2065 */ 2066 if (mreq.imr_interface.s_addr == INADDR_ANY) 2067 ifp = NULL; 2068 else { 2069 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 2070 if (ifp == NULL) { 2071 error = EADDRNOTAVAIL; 2072 splx(s); 2073 break; 2074 } 2075 } 2076 /* 2077 * Find the membership in the membership array. 2078 */ 2079 for (i = 0; i < imo->imo_num_memberships; ++i) { 2080 if ((ifp == NULL || 2081 imo->imo_membership[i]->inm_ifp == ifp) && 2082 imo->imo_membership[i]->inm_addr.s_addr == 2083 mreq.imr_multiaddr.s_addr) 2084 break; 2085 } 2086 if (i == imo->imo_num_memberships) { 2087 error = EADDRNOTAVAIL; 2088 splx(s); 2089 break; 2090 } 2091 /* 2092 * Give up the multicast address record to which the 2093 * membership points. 2094 */ 2095 in_delmulti(imo->imo_membership[i]); 2096 /* 2097 * Remove the gap in the membership array. 2098 */ 2099 for (++i; i < imo->imo_num_memberships; ++i) 2100 imo->imo_membership[i-1] = imo->imo_membership[i]; 2101 --imo->imo_num_memberships; 2102 splx(s); 2103 break; 2104 2105 default: 2106 error = EOPNOTSUPP; 2107 break; 2108 } 2109 2110 /* 2111 * If all options have default values, no need to keep the mbuf. 2112 */ 2113 if (imo->imo_multicast_ifp == NULL && 2114 imo->imo_multicast_vif == -1 && 2115 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2116 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2117 imo->imo_num_memberships == 0) { 2118 free(*imop, M_IPMOPTS); 2119 *imop = NULL; 2120 } 2121 2122 return (error); 2123 } 2124 2125 /* 2126 * Return the IP multicast options in response to user getsockopt(). 2127 */ 2128 static int 2129 ip_getmoptions(sopt, imo) 2130 struct sockopt *sopt; 2131 register struct ip_moptions *imo; 2132 { 2133 struct in_addr addr; 2134 struct in_ifaddr *ia; 2135 int error, optval; 2136 u_char coptval; 2137 2138 error = 0; 2139 switch (sopt->sopt_name) { 2140 case IP_MULTICAST_VIF: 2141 if (imo != NULL) 2142 optval = imo->imo_multicast_vif; 2143 else 2144 optval = -1; 2145 error = sooptcopyout(sopt, &optval, sizeof optval); 2146 break; 2147 2148 case IP_MULTICAST_IF: 2149 if (imo == NULL || imo->imo_multicast_ifp == NULL) 2150 addr.s_addr = INADDR_ANY; 2151 else if (imo->imo_multicast_addr.s_addr) { 2152 /* return the value user has set */ 2153 addr = imo->imo_multicast_addr; 2154 } else { 2155 IFP_TO_IA(imo->imo_multicast_ifp, ia); 2156 addr.s_addr = (ia == NULL) ? INADDR_ANY 2157 : IA_SIN(ia)->sin_addr.s_addr; 2158 } 2159 error = sooptcopyout(sopt, &addr, sizeof addr); 2160 break; 2161 2162 case IP_MULTICAST_TTL: 2163 if (imo == 0) 2164 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 2165 else 2166 optval = coptval = imo->imo_multicast_ttl; 2167 if (sopt->sopt_valsize == 1) 2168 error = sooptcopyout(sopt, &coptval, 1); 2169 else 2170 error = sooptcopyout(sopt, &optval, sizeof optval); 2171 break; 2172 2173 case IP_MULTICAST_LOOP: 2174 if (imo == 0) 2175 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 2176 else 2177 optval = coptval = imo->imo_multicast_loop; 2178 if (sopt->sopt_valsize == 1) 2179 error = sooptcopyout(sopt, &coptval, 1); 2180 else 2181 error = sooptcopyout(sopt, &optval, sizeof optval); 2182 break; 2183 2184 default: 2185 error = ENOPROTOOPT; 2186 break; 2187 } 2188 return (error); 2189 } 2190 2191 /* 2192 * Discard the IP multicast options. 2193 */ 2194 void 2195 ip_freemoptions(imo) 2196 register struct ip_moptions *imo; 2197 { 2198 register int i; 2199 2200 if (imo != NULL) { 2201 for (i = 0; i < imo->imo_num_memberships; ++i) 2202 in_delmulti(imo->imo_membership[i]); 2203 free(imo, M_IPMOPTS); 2204 } 2205 } 2206 2207 /* 2208 * Routine called from ip_output() to loop back a copy of an IP multicast 2209 * packet to the input queue of a specified interface. Note that this 2210 * calls the output routine of the loopback "driver", but with an interface 2211 * pointer that might NOT be a loopback interface -- evil, but easier than 2212 * replicating that code here. 2213 */ 2214 static void 2215 ip_mloopback(ifp, m, dst, hlen) 2216 struct ifnet *ifp; 2217 register struct mbuf *m; 2218 register struct sockaddr_in *dst; 2219 int hlen; 2220 { 2221 register struct ip *ip; 2222 struct mbuf *copym; 2223 2224 copym = m_copy(m, 0, M_COPYALL); 2225 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 2226 copym = m_pullup(copym, hlen); 2227 if (copym != NULL) { 2228 /* 2229 * We don't bother to fragment if the IP length is greater 2230 * than the interface's MTU. Can this possibly matter? 2231 */ 2232 ip = mtod(copym, struct ip *); 2233 ip->ip_len = htons(ip->ip_len); 2234 ip->ip_off = htons(ip->ip_off); 2235 ip->ip_sum = 0; 2236 ip->ip_sum = in_cksum(copym, hlen); 2237 /* 2238 * NB: 2239 * It's not clear whether there are any lingering 2240 * reentrancy problems in other areas which might 2241 * be exposed by using ip_input directly (in 2242 * particular, everything which modifies the packet 2243 * in-place). Yet another option is using the 2244 * protosw directly to deliver the looped back 2245 * packet. For the moment, we'll err on the side 2246 * of safety by using if_simloop(). 2247 */ 2248 #if 1 /* XXX */ 2249 if (dst->sin_family != AF_INET) { 2250 printf("ip_mloopback: bad address family %d\n", 2251 dst->sin_family); 2252 dst->sin_family = AF_INET; 2253 } 2254 #endif 2255 2256 #ifdef notdef 2257 copym->m_pkthdr.rcvif = ifp; 2258 ip_input(copym); 2259 #else 2260 /* if the checksum hasn't been computed, mark it as valid */ 2261 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2262 copym->m_pkthdr.csum_flags |= 2263 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 2264 copym->m_pkthdr.csum_data = 0xffff; 2265 } 2266 if_simloop(ifp, copym, dst->sin_family, 0); 2267 #endif 2268 } 2269 } 2270