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 #define _IP_VHL 38 39 #include "opt_ipfw.h" 40 #include "opt_ipdn.h" 41 #include "opt_ipdivert.h" 42 #include "opt_ipfilter.h" 43 #include "opt_ipsec.h" 44 #include "opt_pfil_hooks.h" 45 #include "opt_random_ip_id.h" 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/kernel.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 68 static 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 #include <netinet/ip_fw.h> 81 #include <netinet/ip_dummynet.h> 82 83 #ifdef IPFIREWALL_FORWARD_DEBUG 84 #define print_ip(a) printf("%ld.%ld.%ld.%ld",(ntohl(a.s_addr)>>24)&0xFF,\ 85 (ntohl(a.s_addr)>>16)&0xFF,\ 86 (ntohl(a.s_addr)>>8)&0xFF,\ 87 (ntohl(a.s_addr))&0xFF); 88 #endif 89 90 u_short ip_id; 91 92 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *)); 93 static struct ifnet *ip_multicast_if __P((struct in_addr *, int *)); 94 static void ip_mloopback 95 __P((struct ifnet *, struct mbuf *, struct sockaddr_in *, int)); 96 static int ip_getmoptions 97 __P((struct sockopt *, struct ip_moptions *)); 98 static int ip_pcbopts __P((int, struct mbuf **, struct mbuf *)); 99 static int ip_setmoptions 100 __P((struct sockopt *, struct ip_moptions **)); 101 102 int ip_optcopy __P((struct ip *, struct ip *)); 103 104 105 extern struct protosw inetsw[]; 106 107 /* 108 * IP output. The packet in mbuf chain m contains a skeletal IP 109 * header (with len, off, ttl, proto, tos, src, dst). 110 * The mbuf chain containing the packet will be freed. 111 * The mbuf opt, if present, will not be freed. 112 */ 113 int 114 ip_output(m0, opt, ro, flags, imo) 115 struct mbuf *m0; 116 struct mbuf *opt; 117 struct route *ro; 118 int flags; 119 struct ip_moptions *imo; 120 { 121 struct ip *ip, *mhip; 122 struct ifnet *ifp; 123 struct mbuf *m = m0; 124 int hlen = sizeof (struct ip); 125 int len, off, error = 0; 126 struct route iproute; 127 struct sockaddr_in *dst; 128 struct in_ifaddr *ia; 129 int isbroadcast, sw_csum; 130 #ifdef IPSEC 131 struct socket *so = NULL; 132 struct secpolicy *sp = NULL; 133 #endif 134 u_int16_t divert_cookie; /* firewall cookie */ 135 #ifdef PFIL_HOOKS 136 struct packet_filter_hook *pfh; 137 struct mbuf *m1; 138 int rv; 139 #endif /* PFIL_HOOKS */ 140 #ifdef IPFIREWALL_FORWARD 141 int fwd_rewrite_src = 0; 142 #endif 143 struct ip_fw *rule = NULL; 144 145 #ifdef IPDIVERT 146 /* Get and reset firewall cookie */ 147 divert_cookie = ip_divert_cookie; 148 ip_divert_cookie = 0; 149 #else 150 divert_cookie = 0; 151 #endif 152 153 /* 154 * dummynet packet are prepended a vestigial mbuf with 155 * m_type = MT_DUMMYNET and m_data pointing to the matching 156 * rule. 157 */ 158 if (m->m_type == MT_DUMMYNET) { 159 /* 160 * the packet was already tagged, so part of the 161 * processing was already done, and we need to go down. 162 * Get parameters from the header. 163 */ 164 rule = (struct ip_fw *)(m->m_data) ; 165 opt = NULL ; 166 ro = & ( ((struct dn_pkt *)m)->ro ) ; 167 imo = NULL ; 168 dst = ((struct dn_pkt *)m)->dn_dst ; 169 ifp = ((struct dn_pkt *)m)->ifp ; 170 flags = ((struct dn_pkt *)m)->flags ; 171 172 m0 = m = m->m_next ; 173 #ifdef IPSEC 174 so = ipsec_getsocket(m); 175 (void)ipsec_setsocket(m, NULL); 176 #endif 177 ip = mtod(m, struct ip *); 178 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ; 179 ia = ifatoia(ro->ro_rt->rt_ifa); 180 goto sendit; 181 } else 182 rule = NULL ; 183 #ifdef IPSEC 184 so = ipsec_getsocket(m); 185 (void)ipsec_setsocket(m, NULL); 186 #endif 187 188 #ifdef DIAGNOSTIC 189 if ((m->m_flags & M_PKTHDR) == 0) 190 panic("ip_output no HDR"); 191 #endif 192 if (opt) { 193 m = ip_insertoptions(m, opt, &len); 194 hlen = len; 195 } 196 ip = mtod(m, struct ip *); 197 /* 198 * Fill in IP header. 199 */ 200 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 201 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2); 202 ip->ip_off &= IP_DF; 203 #ifdef RANDOM_IP_ID 204 ip->ip_id = ip_randomid(); 205 #else 206 ip->ip_id = htons(ip_id++); 207 #endif 208 ipstat.ips_localout++; 209 } else { 210 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 211 } 212 213 /* Route packet. */ 214 if (ro == NULL) { 215 ro = &iproute; 216 bzero(ro, sizeof(*ro)); 217 } 218 dst = (struct sockaddr_in *)&ro->ro_dst; 219 /* 220 * If there is a cached route, 221 * check that it is to the same destination 222 * and is still up. If not, free it and try again. 223 */ 224 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 225 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 226 RTFREE(ro->ro_rt); 227 ro->ro_rt = (struct rtentry *)0; 228 } 229 if (ro->ro_rt == 0) { 230 dst->sin_family = AF_INET; 231 dst->sin_len = sizeof(*dst); 232 dst->sin_addr = ip->ip_dst; 233 } 234 /* 235 * If routing to interface only, 236 * short circuit routing lookup. 237 */ 238 if (flags & IP_ROUTETOIF) { 239 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 240 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 241 ipstat.ips_noroute++; 242 error = ENETUNREACH; 243 goto bad; 244 } 245 ifp = ia->ia_ifp; 246 ip->ip_ttl = 1; 247 isbroadcast = in_broadcast(dst->sin_addr, ifp); 248 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 249 imo != NULL && imo->imo_multicast_ifp != NULL) { 250 /* 251 * Bypass the normal routing lookup for multicast 252 * packets if the interface is specified. 253 */ 254 ifp = imo->imo_multicast_ifp; 255 IFP_TO_IA(ifp, ia); 256 isbroadcast = 0; /* fool gcc */ 257 } else { 258 /* 259 * If this is the case, we probably don't want to allocate 260 * a protocol-cloned route since we didn't get one from the 261 * ULP. This lets TCP do its thing, while not burdening 262 * forwarding or ICMP with the overhead of cloning a route. 263 * Of course, we still want to do any cloning requested by 264 * the link layer, as this is probably required in all cases 265 * for correct operation (as it is for ARP). 266 */ 267 if (ro->ro_rt == 0) 268 rtalloc_ign(ro, RTF_PRCLONING); 269 if (ro->ro_rt == 0) { 270 ipstat.ips_noroute++; 271 error = EHOSTUNREACH; 272 goto bad; 273 } 274 ia = ifatoia(ro->ro_rt->rt_ifa); 275 ifp = ro->ro_rt->rt_ifp; 276 ro->ro_rt->rt_use++; 277 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 278 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 279 if (ro->ro_rt->rt_flags & RTF_HOST) 280 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 281 else 282 isbroadcast = in_broadcast(dst->sin_addr, ifp); 283 } 284 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 285 struct in_multi *inm; 286 287 m->m_flags |= M_MCAST; 288 /* 289 * IP destination address is multicast. Make sure "dst" 290 * still points to the address in "ro". (It may have been 291 * changed to point to a gateway address, above.) 292 */ 293 dst = (struct sockaddr_in *)&ro->ro_dst; 294 /* 295 * See if the caller provided any multicast options 296 */ 297 if (imo != NULL) { 298 ip->ip_ttl = imo->imo_multicast_ttl; 299 if (imo->imo_multicast_vif != -1) 300 ip->ip_src.s_addr = 301 ip_mcast_src(imo->imo_multicast_vif); 302 } else 303 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 304 /* 305 * Confirm that the outgoing interface supports multicast. 306 */ 307 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 308 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 309 ipstat.ips_noroute++; 310 error = ENETUNREACH; 311 goto bad; 312 } 313 } 314 /* 315 * If source address not specified yet, use address 316 * of outgoing interface. 317 */ 318 if (ip->ip_src.s_addr == INADDR_ANY) { 319 /* Interface may have no addresses. */ 320 if (ia != NULL) 321 ip->ip_src = IA_SIN(ia)->sin_addr; 322 } 323 324 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 325 if (inm != NULL && 326 (imo == NULL || imo->imo_multicast_loop)) { 327 /* 328 * If we belong to the destination multicast group 329 * on the outgoing interface, and the caller did not 330 * forbid loopback, loop back a copy. 331 */ 332 ip_mloopback(ifp, m, dst, hlen); 333 } 334 else { 335 /* 336 * If we are acting as a multicast router, perform 337 * multicast forwarding as if the packet had just 338 * arrived on the interface to which we are about 339 * to send. The multicast forwarding function 340 * recursively calls this function, using the 341 * IP_FORWARDING flag to prevent infinite recursion. 342 * 343 * Multicasts that are looped back by ip_mloopback(), 344 * above, will be forwarded by the ip_input() routine, 345 * if necessary. 346 */ 347 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 348 /* 349 * Check if rsvp daemon is running. If not, don't 350 * set ip_moptions. This ensures that the packet 351 * is multicast and not just sent down one link 352 * as prescribed by rsvpd. 353 */ 354 if (!rsvp_on) 355 imo = NULL; 356 if (ip_mforward(ip, ifp, m, imo) != 0) { 357 m_freem(m); 358 goto done; 359 } 360 } 361 } 362 363 /* 364 * Multicasts with a time-to-live of zero may be looped- 365 * back, above, but must not be transmitted on a network. 366 * Also, multicasts addressed to the loopback interface 367 * are not sent -- the above call to ip_mloopback() will 368 * loop back a copy if this host actually belongs to the 369 * destination group on the loopback interface. 370 */ 371 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 372 m_freem(m); 373 goto done; 374 } 375 376 goto sendit; 377 } 378 #ifndef notdef 379 /* 380 * If source address not specified yet, use address 381 * of outgoing interface. 382 */ 383 if (ip->ip_src.s_addr == INADDR_ANY) { 384 /* Interface may have no addresses. */ 385 if (ia != NULL) { 386 ip->ip_src = IA_SIN(ia)->sin_addr; 387 #ifdef IPFIREWALL_FORWARD 388 /* Keep note that we did this - if the firewall changes 389 * the next-hop, our interface may change, changing the 390 * default source IP. It's a shame so much effort happens 391 * twice. Oh well. 392 */ 393 fwd_rewrite_src++; 394 #endif /* IPFIREWALL_FORWARD */ 395 } 396 } 397 #endif /* notdef */ 398 /* 399 * Verify that we have any chance at all of being able to queue 400 * the packet or packet fragments 401 */ 402 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 403 ifp->if_snd.ifq_maxlen) { 404 error = ENOBUFS; 405 ipstat.ips_odropped++; 406 goto bad; 407 } 408 409 /* 410 * Look for broadcast address and 411 * and verify user is allowed to send 412 * such a packet. 413 */ 414 if (isbroadcast) { 415 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 416 error = EADDRNOTAVAIL; 417 goto bad; 418 } 419 if ((flags & IP_ALLOWBROADCAST) == 0) { 420 error = EACCES; 421 goto bad; 422 } 423 /* don't allow broadcast messages to be fragmented */ 424 if ((u_short)ip->ip_len > ifp->if_mtu) { 425 error = EMSGSIZE; 426 goto bad; 427 } 428 m->m_flags |= M_BCAST; 429 } else { 430 m->m_flags &= ~M_BCAST; 431 } 432 433 sendit: 434 #ifdef IPSEC 435 /* get SP for this packet */ 436 if (so == NULL) 437 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 438 else 439 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 440 441 if (sp == NULL) { 442 ipsecstat.out_inval++; 443 goto bad; 444 } 445 446 error = 0; 447 448 /* check policy */ 449 switch (sp->policy) { 450 case IPSEC_POLICY_DISCARD: 451 /* 452 * This packet is just discarded. 453 */ 454 ipsecstat.out_polvio++; 455 goto bad; 456 457 case IPSEC_POLICY_BYPASS: 458 case IPSEC_POLICY_NONE: 459 /* no need to do IPsec. */ 460 goto skip_ipsec; 461 462 case IPSEC_POLICY_IPSEC: 463 if (sp->req == NULL) { 464 /* acquire a policy */ 465 error = key_spdacquire(sp); 466 goto bad; 467 } 468 break; 469 470 case IPSEC_POLICY_ENTRUST: 471 default: 472 printf("ip_output: Invalid policy found. %d\n", sp->policy); 473 } 474 { 475 struct ipsec_output_state state; 476 bzero(&state, sizeof(state)); 477 state.m = m; 478 if (flags & IP_ROUTETOIF) { 479 state.ro = &iproute; 480 bzero(&iproute, sizeof(iproute)); 481 } else 482 state.ro = ro; 483 state.dst = (struct sockaddr *)dst; 484 485 ip->ip_sum = 0; 486 487 /* 488 * XXX 489 * delayed checksums are not currently compatible with IPsec 490 */ 491 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 492 in_delayed_cksum(m); 493 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 494 } 495 496 HTONS(ip->ip_len); 497 HTONS(ip->ip_off); 498 499 error = ipsec4_output(&state, sp, flags); 500 501 m = state.m; 502 if (flags & IP_ROUTETOIF) { 503 /* 504 * if we have tunnel mode SA, we may need to ignore 505 * IP_ROUTETOIF. 506 */ 507 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 508 flags &= ~IP_ROUTETOIF; 509 ro = state.ro; 510 } 511 } else 512 ro = state.ro; 513 dst = (struct sockaddr_in *)state.dst; 514 if (error) { 515 /* mbuf is already reclaimed in ipsec4_output. */ 516 m0 = NULL; 517 switch (error) { 518 case EHOSTUNREACH: 519 case ENETUNREACH: 520 case EMSGSIZE: 521 case ENOBUFS: 522 case ENOMEM: 523 break; 524 default: 525 printf("ip4_output (ipsec): error code %d\n", error); 526 /*fall through*/ 527 case ENOENT: 528 /* don't show these error codes to the user */ 529 error = 0; 530 break; 531 } 532 goto bad; 533 } 534 } 535 536 /* be sure to update variables that are affected by ipsec4_output() */ 537 ip = mtod(m, struct ip *); 538 #ifdef _IP_VHL 539 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 540 #else 541 hlen = ip->ip_hl << 2; 542 #endif 543 if (ro->ro_rt == NULL) { 544 if ((flags & IP_ROUTETOIF) == 0) { 545 printf("ip_output: " 546 "can't update route after IPsec processing\n"); 547 error = EHOSTUNREACH; /*XXX*/ 548 goto bad; 549 } 550 } else { 551 ia = ifatoia(ro->ro_rt->rt_ifa); 552 ifp = ro->ro_rt->rt_ifp; 553 } 554 555 /* make it flipped, again. */ 556 NTOHS(ip->ip_len); 557 NTOHS(ip->ip_off); 558 skip_ipsec: 559 #endif /*IPSEC*/ 560 561 /* 562 * IpHack's section. 563 * - Xlate: translate packet's addr/port (NAT). 564 * - Firewall: deny/allow/etc. 565 * - Wrap: fake packet's addr/port <unimpl.> 566 * - Encapsulate: put it in another IP and send out. <unimp.> 567 */ 568 #ifdef PFIL_HOOKS 569 /* 570 * Run through list of hooks for output packets. 571 */ 572 m1 = m; 573 pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 574 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 575 if (pfh->pfil_func) { 576 rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1); 577 if (rv) { 578 error = EHOSTUNREACH; 579 goto done; 580 } 581 m = m1; 582 if (m == NULL) 583 goto done; 584 ip = mtod(m, struct ip *); 585 } 586 #endif /* PFIL_HOOKS */ 587 588 /* 589 * Check with the firewall... 590 */ 591 if (fw_enable && IPFW_LOADED) { 592 struct sockaddr_in *old = dst; 593 594 off = ip_fw_chk_ptr(&ip, 595 hlen, ifp, &divert_cookie, &m, &rule, &dst); 596 /* 597 * On return we must do the following: 598 * m == NULL -> drop the pkt (old interface, deprecated) 599 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 600 * 1<=off<= 0xffff -> DIVERT 601 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 602 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 603 * dst != old -> IPFIREWALL_FORWARD 604 * off==0, dst==old -> accept 605 * If some of the above modules are not compiled in, then 606 * we should't have to check the corresponding condition 607 * (because the ipfw control socket should not accept 608 * unsupported rules), but better play safe and drop 609 * packets in case of doubt. 610 */ 611 if (off & IP_FW_PORT_DENY_FLAG) { /* XXX new interface-denied */ 612 if (m) 613 m_freem(m); 614 error = EACCES; 615 goto done; 616 } 617 if (!m) { /* firewall said to reject */ 618 static int __debug=10; 619 620 if (__debug > 0) { 621 printf( 622 "firewall returns NULL, please update!\n"); 623 __debug--; 624 } 625 error = EACCES; 626 goto done; 627 } 628 if (off == 0 && dst == old) /* common case */ 629 goto pass; 630 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 631 /* 632 * pass the pkt to dummynet. Need to include 633 * pipe number, m, ifp, ro, dst because these are 634 * not recomputed in the next pass. 635 * All other parameters have been already used and 636 * so they are not needed anymore. 637 * XXX note: if the ifp or ro entry are deleted 638 * while a pkt is in dummynet, we are in trouble! 639 */ 640 error = ip_dn_io_ptr(off & 0xffff, DN_TO_IP_OUT, m, 641 ifp, ro, dst, rule, flags); 642 goto done; 643 } 644 #ifdef IPDIVERT 645 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 646 struct mbuf *clone = NULL; 647 648 /* Clone packet if we're doing a 'tee' */ 649 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 650 clone = m_dup(m, M_DONTWAIT); 651 652 /* 653 * XXX 654 * delayed checksums are not currently compatible 655 * with divert sockets. 656 */ 657 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 658 in_delayed_cksum(m); 659 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 660 } 661 662 /* Restore packet header fields to original values */ 663 HTONS(ip->ip_len); 664 HTONS(ip->ip_off); 665 666 /* Deliver packet to divert input routine */ 667 ip_divert_cookie = divert_cookie; 668 divert_packet(m, 0, off & 0xffff); 669 670 /* If 'tee', continue with original packet */ 671 if (clone != NULL) { 672 m = clone; 673 ip = mtod(m, struct ip *); 674 goto pass; 675 } 676 goto done; 677 } 678 #endif 679 680 #ifdef IPFIREWALL_FORWARD 681 /* Here we check dst to make sure it's directly reachable on the 682 * interface we previously thought it was. 683 * If it isn't (which may be likely in some situations) we have 684 * to re-route it (ie, find a route for the next-hop and the 685 * associated interface) and set them here. This is nested 686 * forwarding which in most cases is undesirable, except where 687 * such control is nigh impossible. So we do it here. 688 * And I'm babbling. 689 */ 690 if (off == 0 && old != dst) { 691 struct in_ifaddr *ia; 692 693 /* It's changed... */ 694 /* There must be a better way to do this next line... */ 695 static struct route sro_fwd, *ro_fwd = &sro_fwd; 696 #ifdef IPFIREWALL_FORWARD_DEBUG 697 printf("IPFIREWALL_FORWARD: New dst ip: "); 698 print_ip(dst->sin_addr); 699 printf("\n"); 700 #endif 701 /* 702 * We need to figure out if we have been forwarded 703 * to a local socket. If so then we should somehow 704 * "loop back" to ip_input, and get directed to the 705 * PCB as if we had received this packet. This is 706 * because it may be dificult to identify the packets 707 * you want to forward until they are being output 708 * and have selected an interface. (e.g. locally 709 * initiated packets) If we used the loopback inteface, 710 * we would not be able to control what happens 711 * as the packet runs through ip_input() as 712 * it is done through a ISR. 713 */ 714 LIST_FOREACH(ia, 715 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 716 /* 717 * If the addr to forward to is one 718 * of ours, we pretend to 719 * be the destination for this packet. 720 */ 721 if (IA_SIN(ia)->sin_addr.s_addr == 722 dst->sin_addr.s_addr) 723 break; 724 } 725 if (ia) { 726 /* tell ip_input "dont filter" */ 727 ip_fw_fwd_addr = dst; 728 if (m->m_pkthdr.rcvif == NULL) 729 m->m_pkthdr.rcvif = ifunit("lo0"); 730 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 731 m->m_pkthdr.csum_flags |= 732 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 733 m0->m_pkthdr.csum_data = 0xffff; 734 } 735 m->m_pkthdr.csum_flags |= 736 CSUM_IP_CHECKED | CSUM_IP_VALID; 737 HTONS(ip->ip_len); 738 HTONS(ip->ip_off); 739 ip_input(m); 740 goto done; 741 } 742 /* Some of the logic for this was 743 * nicked from above. 744 * 745 * This rewrites the cached route in a local PCB. 746 * Is this what we want to do? 747 */ 748 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 749 750 ro_fwd->ro_rt = 0; 751 rtalloc_ign(ro_fwd, RTF_PRCLONING); 752 753 if (ro_fwd->ro_rt == 0) { 754 ipstat.ips_noroute++; 755 error = EHOSTUNREACH; 756 goto bad; 757 } 758 759 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 760 ifp = ro_fwd->ro_rt->rt_ifp; 761 ro_fwd->ro_rt->rt_use++; 762 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 763 dst = (struct sockaddr_in *)ro_fwd->ro_rt->rt_gateway; 764 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 765 isbroadcast = 766 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 767 else 768 isbroadcast = in_broadcast(dst->sin_addr, ifp); 769 RTFREE(ro->ro_rt); 770 ro->ro_rt = ro_fwd->ro_rt; 771 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 772 773 /* 774 * If we added a default src ip earlier, 775 * which would have been gotten from the-then 776 * interface, do it again, from the new one. 777 */ 778 if (fwd_rewrite_src) 779 ip->ip_src = IA_SIN(ia)->sin_addr; 780 goto pass ; 781 } 782 #endif /* IPFIREWALL_FORWARD */ 783 /* 784 * if we get here, none of the above matches, and 785 * we have to drop the pkt 786 */ 787 m_freem(m); 788 error = EACCES; /* not sure this is the right error msg */ 789 goto done; 790 } 791 792 pass: 793 m->m_pkthdr.csum_flags |= CSUM_IP; 794 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 795 if (sw_csum & CSUM_DELAY_DATA) { 796 in_delayed_cksum(m); 797 sw_csum &= ~CSUM_DELAY_DATA; 798 } 799 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 800 801 /* 802 * If small enough for interface, or the interface will take 803 * care of the fragmentation for us, can just send directly. 804 */ 805 if ((u_short)ip->ip_len <= ifp->if_mtu || 806 ifp->if_hwassist & CSUM_FRAGMENT) { 807 HTONS(ip->ip_len); 808 HTONS(ip->ip_off); 809 ip->ip_sum = 0; 810 if (sw_csum & CSUM_DELAY_IP) { 811 if (ip->ip_vhl == IP_VHL_BORING) { 812 ip->ip_sum = in_cksum_hdr(ip); 813 } else { 814 ip->ip_sum = in_cksum(m, hlen); 815 } 816 } 817 818 /* Record statistics for this interface address. */ 819 if (!(flags & IP_FORWARDING) && ia) { 820 ia->ia_ifa.if_opackets++; 821 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 822 } 823 824 #ifdef IPSEC 825 /* clean ipsec history once it goes out of the node */ 826 ipsec_delaux(m); 827 #endif 828 829 error = (*ifp->if_output)(ifp, m, 830 (struct sockaddr *)dst, ro->ro_rt); 831 goto done; 832 } 833 /* 834 * Too large for interface; fragment if possible. 835 * Must be able to put at least 8 bytes per fragment. 836 */ 837 if (ip->ip_off & IP_DF) { 838 error = EMSGSIZE; 839 /* 840 * This case can happen if the user changed the MTU 841 * of an interface after enabling IP on it. Because 842 * most netifs don't keep track of routes pointing to 843 * them, there is no way for one to update all its 844 * routes when the MTU is changed. 845 */ 846 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 847 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 848 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 849 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 850 } 851 ipstat.ips_cantfrag++; 852 goto bad; 853 } 854 len = (ifp->if_mtu - hlen) &~ 7; 855 if (len < 8) { 856 error = EMSGSIZE; 857 goto bad; 858 } 859 860 /* 861 * if the interface will not calculate checksums on 862 * fragmented packets, then do it here. 863 */ 864 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 865 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 866 in_delayed_cksum(m); 867 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 868 } 869 870 { 871 int mhlen, firstlen = len; 872 struct mbuf **mnext = &m->m_nextpkt; 873 int nfrags = 1; 874 875 /* 876 * Loop through length of segment after first fragment, 877 * make new header and copy data of each part and link onto chain. 878 */ 879 m0 = m; 880 mhlen = sizeof (struct ip); 881 for (off = hlen + len; off < (u_short)ip->ip_len; off += len) { 882 MGETHDR(m, M_DONTWAIT, MT_HEADER); 883 if (m == 0) { 884 error = ENOBUFS; 885 ipstat.ips_odropped++; 886 goto sendorfree; 887 } 888 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 889 m->m_data += max_linkhdr; 890 mhip = mtod(m, struct ip *); 891 *mhip = *ip; 892 if (hlen > sizeof (struct ip)) { 893 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 894 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2); 895 } 896 m->m_len = mhlen; 897 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 898 if (off + len >= (u_short)ip->ip_len) 899 len = (u_short)ip->ip_len - off; 900 else 901 mhip->ip_off |= IP_MF; 902 mhip->ip_len = htons((u_short)(len + mhlen)); 903 m->m_next = m_copy(m0, off, len); 904 if (m->m_next == 0) { 905 (void) m_free(m); 906 error = ENOBUFS; /* ??? */ 907 ipstat.ips_odropped++; 908 goto sendorfree; 909 } 910 m->m_pkthdr.len = mhlen + len; 911 m->m_pkthdr.rcvif = (struct ifnet *)0; 912 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 913 HTONS(mhip->ip_off); 914 mhip->ip_sum = 0; 915 if (sw_csum & CSUM_DELAY_IP) { 916 if (mhip->ip_vhl == IP_VHL_BORING) { 917 mhip->ip_sum = in_cksum_hdr(mhip); 918 } else { 919 mhip->ip_sum = in_cksum(m, mhlen); 920 } 921 } 922 *mnext = m; 923 mnext = &m->m_nextpkt; 924 nfrags++; 925 } 926 ipstat.ips_ofragments += nfrags; 927 928 /* set first/last markers for fragment chain */ 929 m->m_flags |= M_LASTFRAG; 930 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 931 m0->m_pkthdr.csum_data = nfrags; 932 933 /* 934 * Update first fragment by trimming what's been copied out 935 * and updating header, then send each fragment (in order). 936 */ 937 m = m0; 938 m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 939 m->m_pkthdr.len = hlen + firstlen; 940 ip->ip_len = htons((u_short)m->m_pkthdr.len); 941 ip->ip_off |= IP_MF; 942 HTONS(ip->ip_off); 943 ip->ip_sum = 0; 944 if (sw_csum & CSUM_DELAY_IP) { 945 if (ip->ip_vhl == IP_VHL_BORING) { 946 ip->ip_sum = in_cksum_hdr(ip); 947 } else { 948 ip->ip_sum = in_cksum(m, hlen); 949 } 950 } 951 sendorfree: 952 for (m = m0; m; m = m0) { 953 m0 = m->m_nextpkt; 954 m->m_nextpkt = 0; 955 #ifdef IPSEC 956 /* clean ipsec history once it goes out of the node */ 957 ipsec_delaux(m); 958 #endif 959 if (error == 0) { 960 /* Record statistics for this interface address. */ 961 if (ia != NULL) { 962 ia->ia_ifa.if_opackets++; 963 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 964 } 965 966 error = (*ifp->if_output)(ifp, m, 967 (struct sockaddr *)dst, ro->ro_rt); 968 } else 969 m_freem(m); 970 } 971 972 if (error == 0) 973 ipstat.ips_fragmented++; 974 } 975 done: 976 #ifdef IPSEC 977 if (ro == &iproute && ro->ro_rt) { 978 RTFREE(ro->ro_rt); 979 ro->ro_rt = NULL; 980 } 981 if (sp != NULL) { 982 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 983 printf("DP ip_output call free SP:%p\n", sp)); 984 key_freesp(sp); 985 } 986 #endif /* IPSEC */ 987 return (error); 988 bad: 989 m_freem(m); 990 goto done; 991 } 992 993 void 994 in_delayed_cksum(struct mbuf *m) 995 { 996 struct ip *ip; 997 u_short csum, offset; 998 999 ip = mtod(m, struct ip *); 1000 offset = IP_VHL_HL(ip->ip_vhl) << 2 ; 1001 csum = in_cksum_skip(m, ip->ip_len, offset); 1002 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1003 csum = 0xffff; 1004 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1005 1006 if (offset + sizeof(u_short) > m->m_len) { 1007 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1008 m->m_len, offset, ip->ip_p); 1009 /* 1010 * XXX 1011 * this shouldn't happen, but if it does, the 1012 * correct behavior may be to insert the checksum 1013 * in the existing chain instead of rearranging it. 1014 */ 1015 m = m_pullup(m, offset + sizeof(u_short)); 1016 } 1017 *(u_short *)(m->m_data + offset) = csum; 1018 } 1019 1020 /* 1021 * Insert IP options into preformed packet. 1022 * Adjust IP destination as required for IP source routing, 1023 * as indicated by a non-zero in_addr at the start of the options. 1024 * 1025 * XXX This routine assumes that the packet has no options in place. 1026 */ 1027 static struct mbuf * 1028 ip_insertoptions(m, opt, phlen) 1029 register struct mbuf *m; 1030 struct mbuf *opt; 1031 int *phlen; 1032 { 1033 register struct ipoption *p = mtod(opt, struct ipoption *); 1034 struct mbuf *n; 1035 register struct ip *ip = mtod(m, struct ip *); 1036 unsigned optlen; 1037 1038 optlen = opt->m_len - sizeof(p->ipopt_dst); 1039 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) 1040 return (m); /* XXX should fail */ 1041 if (p->ipopt_dst.s_addr) 1042 ip->ip_dst = p->ipopt_dst; 1043 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1044 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1045 if (n == 0) 1046 return (m); 1047 n->m_pkthdr.rcvif = (struct ifnet *)0; 1048 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1049 m->m_len -= sizeof(struct ip); 1050 m->m_data += sizeof(struct ip); 1051 n->m_next = m; 1052 m = n; 1053 m->m_len = optlen + sizeof(struct ip); 1054 m->m_data += max_linkhdr; 1055 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1056 } else { 1057 m->m_data -= optlen; 1058 m->m_len += optlen; 1059 m->m_pkthdr.len += optlen; 1060 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1061 } 1062 ip = mtod(m, struct ip *); 1063 bcopy(p->ipopt_list, ip + 1, optlen); 1064 *phlen = sizeof(struct ip) + optlen; 1065 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2); 1066 ip->ip_len += optlen; 1067 return (m); 1068 } 1069 1070 /* 1071 * Copy options from ip to jp, 1072 * omitting those not copied during fragmentation. 1073 */ 1074 int 1075 ip_optcopy(ip, jp) 1076 struct ip *ip, *jp; 1077 { 1078 register u_char *cp, *dp; 1079 int opt, optlen, cnt; 1080 1081 cp = (u_char *)(ip + 1); 1082 dp = (u_char *)(jp + 1); 1083 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1084 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1085 opt = cp[0]; 1086 if (opt == IPOPT_EOL) 1087 break; 1088 if (opt == IPOPT_NOP) { 1089 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1090 *dp++ = IPOPT_NOP; 1091 optlen = 1; 1092 continue; 1093 } 1094 #ifdef DIAGNOSTIC 1095 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1096 panic("malformed IPv4 option passed to ip_optcopy"); 1097 #endif 1098 optlen = cp[IPOPT_OLEN]; 1099 #ifdef DIAGNOSTIC 1100 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1101 panic("malformed IPv4 option passed to ip_optcopy"); 1102 #endif 1103 /* bogus lengths should have been caught by ip_dooptions */ 1104 if (optlen > cnt) 1105 optlen = cnt; 1106 if (IPOPT_COPIED(opt)) { 1107 bcopy(cp, dp, optlen); 1108 dp += optlen; 1109 } 1110 } 1111 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1112 *dp++ = IPOPT_EOL; 1113 return (optlen); 1114 } 1115 1116 /* 1117 * IP socket option processing. 1118 */ 1119 int 1120 ip_ctloutput(so, sopt) 1121 struct socket *so; 1122 struct sockopt *sopt; 1123 { 1124 struct inpcb *inp = sotoinpcb(so); 1125 int error, optval; 1126 1127 error = optval = 0; 1128 if (sopt->sopt_level != IPPROTO_IP) { 1129 return (EINVAL); 1130 } 1131 1132 switch (sopt->sopt_dir) { 1133 case SOPT_SET: 1134 switch (sopt->sopt_name) { 1135 case IP_OPTIONS: 1136 #ifdef notyet 1137 case IP_RETOPTS: 1138 #endif 1139 { 1140 struct mbuf *m; 1141 if (sopt->sopt_valsize > MLEN) { 1142 error = EMSGSIZE; 1143 break; 1144 } 1145 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1146 if (m == 0) { 1147 error = ENOBUFS; 1148 break; 1149 } 1150 m->m_len = sopt->sopt_valsize; 1151 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1152 m->m_len); 1153 1154 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1155 m)); 1156 } 1157 1158 case IP_TOS: 1159 case IP_TTL: 1160 case IP_RECVOPTS: 1161 case IP_RECVRETOPTS: 1162 case IP_RECVDSTADDR: 1163 case IP_RECVIF: 1164 case IP_FAITH: 1165 error = sooptcopyin(sopt, &optval, sizeof optval, 1166 sizeof optval); 1167 if (error) 1168 break; 1169 1170 switch (sopt->sopt_name) { 1171 case IP_TOS: 1172 inp->inp_ip_tos = optval; 1173 break; 1174 1175 case IP_TTL: 1176 inp->inp_ip_ttl = optval; 1177 break; 1178 #define OPTSET(bit) \ 1179 if (optval) \ 1180 inp->inp_flags |= bit; \ 1181 else \ 1182 inp->inp_flags &= ~bit; 1183 1184 case IP_RECVOPTS: 1185 OPTSET(INP_RECVOPTS); 1186 break; 1187 1188 case IP_RECVRETOPTS: 1189 OPTSET(INP_RECVRETOPTS); 1190 break; 1191 1192 case IP_RECVDSTADDR: 1193 OPTSET(INP_RECVDSTADDR); 1194 break; 1195 1196 case IP_RECVIF: 1197 OPTSET(INP_RECVIF); 1198 break; 1199 1200 case IP_FAITH: 1201 OPTSET(INP_FAITH); 1202 break; 1203 } 1204 break; 1205 #undef OPTSET 1206 1207 case IP_MULTICAST_IF: 1208 case IP_MULTICAST_VIF: 1209 case IP_MULTICAST_TTL: 1210 case IP_MULTICAST_LOOP: 1211 case IP_ADD_MEMBERSHIP: 1212 case IP_DROP_MEMBERSHIP: 1213 error = ip_setmoptions(sopt, &inp->inp_moptions); 1214 break; 1215 1216 case IP_PORTRANGE: 1217 error = sooptcopyin(sopt, &optval, sizeof optval, 1218 sizeof optval); 1219 if (error) 1220 break; 1221 1222 switch (optval) { 1223 case IP_PORTRANGE_DEFAULT: 1224 inp->inp_flags &= ~(INP_LOWPORT); 1225 inp->inp_flags &= ~(INP_HIGHPORT); 1226 break; 1227 1228 case IP_PORTRANGE_HIGH: 1229 inp->inp_flags &= ~(INP_LOWPORT); 1230 inp->inp_flags |= INP_HIGHPORT; 1231 break; 1232 1233 case IP_PORTRANGE_LOW: 1234 inp->inp_flags &= ~(INP_HIGHPORT); 1235 inp->inp_flags |= INP_LOWPORT; 1236 break; 1237 1238 default: 1239 error = EINVAL; 1240 break; 1241 } 1242 break; 1243 1244 #ifdef IPSEC 1245 case IP_IPSEC_POLICY: 1246 { 1247 caddr_t req; 1248 size_t len = 0; 1249 int priv; 1250 struct mbuf *m; 1251 int optname; 1252 1253 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1254 break; 1255 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1256 break; 1257 priv = (sopt->sopt_td != NULL && 1258 suser_td(sopt->sopt_td) != 0) ? 0 : 1; 1259 req = mtod(m, caddr_t); 1260 len = m->m_len; 1261 optname = sopt->sopt_name; 1262 error = ipsec4_set_policy(inp, optname, req, len, priv); 1263 m_freem(m); 1264 break; 1265 } 1266 #endif /*IPSEC*/ 1267 1268 default: 1269 error = ENOPROTOOPT; 1270 break; 1271 } 1272 break; 1273 1274 case SOPT_GET: 1275 switch (sopt->sopt_name) { 1276 case IP_OPTIONS: 1277 case IP_RETOPTS: 1278 if (inp->inp_options) 1279 error = sooptcopyout(sopt, 1280 mtod(inp->inp_options, 1281 char *), 1282 inp->inp_options->m_len); 1283 else 1284 sopt->sopt_valsize = 0; 1285 break; 1286 1287 case IP_TOS: 1288 case IP_TTL: 1289 case IP_RECVOPTS: 1290 case IP_RECVRETOPTS: 1291 case IP_RECVDSTADDR: 1292 case IP_RECVIF: 1293 case IP_PORTRANGE: 1294 case IP_FAITH: 1295 switch (sopt->sopt_name) { 1296 1297 case IP_TOS: 1298 optval = inp->inp_ip_tos; 1299 break; 1300 1301 case IP_TTL: 1302 optval = inp->inp_ip_ttl; 1303 break; 1304 1305 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1306 1307 case IP_RECVOPTS: 1308 optval = OPTBIT(INP_RECVOPTS); 1309 break; 1310 1311 case IP_RECVRETOPTS: 1312 optval = OPTBIT(INP_RECVRETOPTS); 1313 break; 1314 1315 case IP_RECVDSTADDR: 1316 optval = OPTBIT(INP_RECVDSTADDR); 1317 break; 1318 1319 case IP_RECVIF: 1320 optval = OPTBIT(INP_RECVIF); 1321 break; 1322 1323 case IP_PORTRANGE: 1324 if (inp->inp_flags & INP_HIGHPORT) 1325 optval = IP_PORTRANGE_HIGH; 1326 else if (inp->inp_flags & INP_LOWPORT) 1327 optval = IP_PORTRANGE_LOW; 1328 else 1329 optval = 0; 1330 break; 1331 1332 case IP_FAITH: 1333 optval = OPTBIT(INP_FAITH); 1334 break; 1335 } 1336 error = sooptcopyout(sopt, &optval, sizeof optval); 1337 break; 1338 1339 case IP_MULTICAST_IF: 1340 case IP_MULTICAST_VIF: 1341 case IP_MULTICAST_TTL: 1342 case IP_MULTICAST_LOOP: 1343 case IP_ADD_MEMBERSHIP: 1344 case IP_DROP_MEMBERSHIP: 1345 error = ip_getmoptions(sopt, inp->inp_moptions); 1346 break; 1347 1348 #ifdef IPSEC 1349 case IP_IPSEC_POLICY: 1350 { 1351 struct mbuf *m = NULL; 1352 caddr_t req = NULL; 1353 size_t len = 0; 1354 1355 if (m != 0) { 1356 req = mtod(m, caddr_t); 1357 len = m->m_len; 1358 } 1359 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1360 if (error == 0) 1361 error = soopt_mcopyout(sopt, m); /* XXX */ 1362 if (error == 0) 1363 m_freem(m); 1364 break; 1365 } 1366 #endif /*IPSEC*/ 1367 1368 default: 1369 error = ENOPROTOOPT; 1370 break; 1371 } 1372 break; 1373 } 1374 return (error); 1375 } 1376 1377 /* 1378 * Set up IP options in pcb for insertion in output packets. 1379 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1380 * with destination address if source routed. 1381 */ 1382 static int 1383 ip_pcbopts(optname, pcbopt, m) 1384 int optname; 1385 struct mbuf **pcbopt; 1386 register struct mbuf *m; 1387 { 1388 register int cnt, optlen; 1389 register u_char *cp; 1390 u_char opt; 1391 1392 /* turn off any old options */ 1393 if (*pcbopt) 1394 (void)m_free(*pcbopt); 1395 *pcbopt = 0; 1396 if (m == (struct mbuf *)0 || m->m_len == 0) { 1397 /* 1398 * Only turning off any previous options. 1399 */ 1400 if (m) 1401 (void)m_free(m); 1402 return (0); 1403 } 1404 1405 if (m->m_len % sizeof(int32_t)) 1406 goto bad; 1407 /* 1408 * IP first-hop destination address will be stored before 1409 * actual options; move other options back 1410 * and clear it when none present. 1411 */ 1412 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1413 goto bad; 1414 cnt = m->m_len; 1415 m->m_len += sizeof(struct in_addr); 1416 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1417 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1418 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1419 1420 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1421 opt = cp[IPOPT_OPTVAL]; 1422 if (opt == IPOPT_EOL) 1423 break; 1424 if (opt == IPOPT_NOP) 1425 optlen = 1; 1426 else { 1427 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1428 goto bad; 1429 optlen = cp[IPOPT_OLEN]; 1430 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1431 goto bad; 1432 } 1433 switch (opt) { 1434 1435 default: 1436 break; 1437 1438 case IPOPT_LSRR: 1439 case IPOPT_SSRR: 1440 /* 1441 * user process specifies route as: 1442 * ->A->B->C->D 1443 * D must be our final destination (but we can't 1444 * check that since we may not have connected yet). 1445 * A is first hop destination, which doesn't appear in 1446 * actual IP option, but is stored before the options. 1447 */ 1448 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1449 goto bad; 1450 m->m_len -= sizeof(struct in_addr); 1451 cnt -= sizeof(struct in_addr); 1452 optlen -= sizeof(struct in_addr); 1453 cp[IPOPT_OLEN] = optlen; 1454 /* 1455 * Move first hop before start of options. 1456 */ 1457 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1458 sizeof(struct in_addr)); 1459 /* 1460 * Then copy rest of options back 1461 * to close up the deleted entry. 1462 */ 1463 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1464 sizeof(struct in_addr)), 1465 (caddr_t)&cp[IPOPT_OFFSET+1], 1466 (unsigned)cnt + sizeof(struct in_addr)); 1467 break; 1468 } 1469 } 1470 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1471 goto bad; 1472 *pcbopt = m; 1473 return (0); 1474 1475 bad: 1476 (void)m_free(m); 1477 return (EINVAL); 1478 } 1479 1480 /* 1481 * XXX 1482 * The whole multicast option thing needs to be re-thought. 1483 * Several of these options are equally applicable to non-multicast 1484 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1485 * standard option (IP_TTL). 1486 */ 1487 1488 /* 1489 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1490 */ 1491 static struct ifnet * 1492 ip_multicast_if(a, ifindexp) 1493 struct in_addr *a; 1494 int *ifindexp; 1495 { 1496 int ifindex; 1497 struct ifnet *ifp; 1498 1499 if (ifindexp) 1500 *ifindexp = 0; 1501 if (ntohl(a->s_addr) >> 24 == 0) { 1502 ifindex = ntohl(a->s_addr) & 0xffffff; 1503 if (ifindex < 0 || if_index < ifindex) 1504 return NULL; 1505 ifp = ifnet_byindex(ifindex); 1506 if (ifindexp) 1507 *ifindexp = ifindex; 1508 } else { 1509 INADDR_TO_IFP(*a, ifp); 1510 } 1511 return ifp; 1512 } 1513 1514 /* 1515 * Set the IP multicast options in response to user setsockopt(). 1516 */ 1517 static int 1518 ip_setmoptions(sopt, imop) 1519 struct sockopt *sopt; 1520 struct ip_moptions **imop; 1521 { 1522 int error = 0; 1523 int i; 1524 struct in_addr addr; 1525 struct ip_mreq mreq; 1526 struct ifnet *ifp; 1527 struct ip_moptions *imo = *imop; 1528 struct route ro; 1529 struct sockaddr_in *dst; 1530 int ifindex; 1531 int s; 1532 1533 if (imo == NULL) { 1534 /* 1535 * No multicast option buffer attached to the pcb; 1536 * allocate one and initialize to default values. 1537 */ 1538 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1539 M_WAITOK); 1540 1541 if (imo == NULL) 1542 return (ENOBUFS); 1543 *imop = imo; 1544 imo->imo_multicast_ifp = NULL; 1545 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1546 imo->imo_multicast_vif = -1; 1547 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1548 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1549 imo->imo_num_memberships = 0; 1550 } 1551 1552 switch (sopt->sopt_name) { 1553 /* store an index number for the vif you wanna use in the send */ 1554 case IP_MULTICAST_VIF: 1555 if (legal_vif_num == 0) { 1556 error = EOPNOTSUPP; 1557 break; 1558 } 1559 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1560 if (error) 1561 break; 1562 if (!legal_vif_num(i) && (i != -1)) { 1563 error = EINVAL; 1564 break; 1565 } 1566 imo->imo_multicast_vif = i; 1567 break; 1568 1569 case IP_MULTICAST_IF: 1570 /* 1571 * Select the interface for outgoing multicast packets. 1572 */ 1573 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1574 if (error) 1575 break; 1576 /* 1577 * INADDR_ANY is used to remove a previous selection. 1578 * When no interface is selected, a default one is 1579 * chosen every time a multicast packet is sent. 1580 */ 1581 if (addr.s_addr == INADDR_ANY) { 1582 imo->imo_multicast_ifp = NULL; 1583 break; 1584 } 1585 /* 1586 * The selected interface is identified by its local 1587 * IP address. Find the interface and confirm that 1588 * it supports multicasting. 1589 */ 1590 s = splimp(); 1591 ifp = ip_multicast_if(&addr, &ifindex); 1592 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1593 splx(s); 1594 error = EADDRNOTAVAIL; 1595 break; 1596 } 1597 imo->imo_multicast_ifp = ifp; 1598 if (ifindex) 1599 imo->imo_multicast_addr = addr; 1600 else 1601 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1602 splx(s); 1603 break; 1604 1605 case IP_MULTICAST_TTL: 1606 /* 1607 * Set the IP time-to-live for outgoing multicast packets. 1608 * The original multicast API required a char argument, 1609 * which is inconsistent with the rest of the socket API. 1610 * We allow either a char or an int. 1611 */ 1612 if (sopt->sopt_valsize == 1) { 1613 u_char ttl; 1614 error = sooptcopyin(sopt, &ttl, 1, 1); 1615 if (error) 1616 break; 1617 imo->imo_multicast_ttl = ttl; 1618 } else { 1619 u_int ttl; 1620 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1621 sizeof ttl); 1622 if (error) 1623 break; 1624 if (ttl > 255) 1625 error = EINVAL; 1626 else 1627 imo->imo_multicast_ttl = ttl; 1628 } 1629 break; 1630 1631 case IP_MULTICAST_LOOP: 1632 /* 1633 * Set the loopback flag for outgoing multicast packets. 1634 * Must be zero or one. The original multicast API required a 1635 * char argument, which is inconsistent with the rest 1636 * of the socket API. We allow either a char or an int. 1637 */ 1638 if (sopt->sopt_valsize == 1) { 1639 u_char loop; 1640 error = sooptcopyin(sopt, &loop, 1, 1); 1641 if (error) 1642 break; 1643 imo->imo_multicast_loop = !!loop; 1644 } else { 1645 u_int loop; 1646 error = sooptcopyin(sopt, &loop, sizeof loop, 1647 sizeof loop); 1648 if (error) 1649 break; 1650 imo->imo_multicast_loop = !!loop; 1651 } 1652 break; 1653 1654 case IP_ADD_MEMBERSHIP: 1655 /* 1656 * Add a multicast group membership. 1657 * Group must be a valid IP multicast address. 1658 */ 1659 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1660 if (error) 1661 break; 1662 1663 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1664 error = EINVAL; 1665 break; 1666 } 1667 s = splimp(); 1668 /* 1669 * If no interface address was provided, use the interface of 1670 * the route to the given multicast address. 1671 */ 1672 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1673 bzero((caddr_t)&ro, sizeof(ro)); 1674 dst = (struct sockaddr_in *)&ro.ro_dst; 1675 dst->sin_len = sizeof(*dst); 1676 dst->sin_family = AF_INET; 1677 dst->sin_addr = mreq.imr_multiaddr; 1678 rtalloc(&ro); 1679 if (ro.ro_rt == NULL) { 1680 error = EADDRNOTAVAIL; 1681 splx(s); 1682 break; 1683 } 1684 ifp = ro.ro_rt->rt_ifp; 1685 rtfree(ro.ro_rt); 1686 } 1687 else { 1688 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1689 } 1690 1691 /* 1692 * See if we found an interface, and confirm that it 1693 * supports multicast. 1694 */ 1695 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1696 error = EADDRNOTAVAIL; 1697 splx(s); 1698 break; 1699 } 1700 /* 1701 * See if the membership already exists or if all the 1702 * membership slots are full. 1703 */ 1704 for (i = 0; i < imo->imo_num_memberships; ++i) { 1705 if (imo->imo_membership[i]->inm_ifp == ifp && 1706 imo->imo_membership[i]->inm_addr.s_addr 1707 == mreq.imr_multiaddr.s_addr) 1708 break; 1709 } 1710 if (i < imo->imo_num_memberships) { 1711 error = EADDRINUSE; 1712 splx(s); 1713 break; 1714 } 1715 if (i == IP_MAX_MEMBERSHIPS) { 1716 error = ETOOMANYREFS; 1717 splx(s); 1718 break; 1719 } 1720 /* 1721 * Everything looks good; add a new record to the multicast 1722 * address list for the given interface. 1723 */ 1724 if ((imo->imo_membership[i] = 1725 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1726 error = ENOBUFS; 1727 splx(s); 1728 break; 1729 } 1730 ++imo->imo_num_memberships; 1731 splx(s); 1732 break; 1733 1734 case IP_DROP_MEMBERSHIP: 1735 /* 1736 * Drop a multicast group membership. 1737 * Group must be a valid IP multicast address. 1738 */ 1739 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1740 if (error) 1741 break; 1742 1743 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1744 error = EINVAL; 1745 break; 1746 } 1747 1748 s = splimp(); 1749 /* 1750 * If an interface address was specified, get a pointer 1751 * to its ifnet structure. 1752 */ 1753 if (mreq.imr_interface.s_addr == INADDR_ANY) 1754 ifp = NULL; 1755 else { 1756 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1757 if (ifp == NULL) { 1758 error = EADDRNOTAVAIL; 1759 splx(s); 1760 break; 1761 } 1762 } 1763 /* 1764 * Find the membership in the membership array. 1765 */ 1766 for (i = 0; i < imo->imo_num_memberships; ++i) { 1767 if ((ifp == NULL || 1768 imo->imo_membership[i]->inm_ifp == ifp) && 1769 imo->imo_membership[i]->inm_addr.s_addr == 1770 mreq.imr_multiaddr.s_addr) 1771 break; 1772 } 1773 if (i == imo->imo_num_memberships) { 1774 error = EADDRNOTAVAIL; 1775 splx(s); 1776 break; 1777 } 1778 /* 1779 * Give up the multicast address record to which the 1780 * membership points. 1781 */ 1782 in_delmulti(imo->imo_membership[i]); 1783 /* 1784 * Remove the gap in the membership array. 1785 */ 1786 for (++i; i < imo->imo_num_memberships; ++i) 1787 imo->imo_membership[i-1] = imo->imo_membership[i]; 1788 --imo->imo_num_memberships; 1789 splx(s); 1790 break; 1791 1792 default: 1793 error = EOPNOTSUPP; 1794 break; 1795 } 1796 1797 /* 1798 * If all options have default values, no need to keep the mbuf. 1799 */ 1800 if (imo->imo_multicast_ifp == NULL && 1801 imo->imo_multicast_vif == -1 && 1802 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1803 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1804 imo->imo_num_memberships == 0) { 1805 free(*imop, M_IPMOPTS); 1806 *imop = NULL; 1807 } 1808 1809 return (error); 1810 } 1811 1812 /* 1813 * Return the IP multicast options in response to user getsockopt(). 1814 */ 1815 static int 1816 ip_getmoptions(sopt, imo) 1817 struct sockopt *sopt; 1818 register struct ip_moptions *imo; 1819 { 1820 struct in_addr addr; 1821 struct in_ifaddr *ia; 1822 int error, optval; 1823 u_char coptval; 1824 1825 error = 0; 1826 switch (sopt->sopt_name) { 1827 case IP_MULTICAST_VIF: 1828 if (imo != NULL) 1829 optval = imo->imo_multicast_vif; 1830 else 1831 optval = -1; 1832 error = sooptcopyout(sopt, &optval, sizeof optval); 1833 break; 1834 1835 case IP_MULTICAST_IF: 1836 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1837 addr.s_addr = INADDR_ANY; 1838 else if (imo->imo_multicast_addr.s_addr) { 1839 /* return the value user has set */ 1840 addr = imo->imo_multicast_addr; 1841 } else { 1842 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1843 addr.s_addr = (ia == NULL) ? INADDR_ANY 1844 : IA_SIN(ia)->sin_addr.s_addr; 1845 } 1846 error = sooptcopyout(sopt, &addr, sizeof addr); 1847 break; 1848 1849 case IP_MULTICAST_TTL: 1850 if (imo == 0) 1851 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1852 else 1853 optval = coptval = imo->imo_multicast_ttl; 1854 if (sopt->sopt_valsize == 1) 1855 error = sooptcopyout(sopt, &coptval, 1); 1856 else 1857 error = sooptcopyout(sopt, &optval, sizeof optval); 1858 break; 1859 1860 case IP_MULTICAST_LOOP: 1861 if (imo == 0) 1862 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1863 else 1864 optval = coptval = imo->imo_multicast_loop; 1865 if (sopt->sopt_valsize == 1) 1866 error = sooptcopyout(sopt, &coptval, 1); 1867 else 1868 error = sooptcopyout(sopt, &optval, sizeof optval); 1869 break; 1870 1871 default: 1872 error = ENOPROTOOPT; 1873 break; 1874 } 1875 return (error); 1876 } 1877 1878 /* 1879 * Discard the IP multicast options. 1880 */ 1881 void 1882 ip_freemoptions(imo) 1883 register struct ip_moptions *imo; 1884 { 1885 register int i; 1886 1887 if (imo != NULL) { 1888 for (i = 0; i < imo->imo_num_memberships; ++i) 1889 in_delmulti(imo->imo_membership[i]); 1890 free(imo, M_IPMOPTS); 1891 } 1892 } 1893 1894 /* 1895 * Routine called from ip_output() to loop back a copy of an IP multicast 1896 * packet to the input queue of a specified interface. Note that this 1897 * calls the output routine of the loopback "driver", but with an interface 1898 * pointer that might NOT be a loopback interface -- evil, but easier than 1899 * replicating that code here. 1900 */ 1901 static void 1902 ip_mloopback(ifp, m, dst, hlen) 1903 struct ifnet *ifp; 1904 register struct mbuf *m; 1905 register struct sockaddr_in *dst; 1906 int hlen; 1907 { 1908 register struct ip *ip; 1909 struct mbuf *copym; 1910 1911 copym = m_copy(m, 0, M_COPYALL); 1912 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1913 copym = m_pullup(copym, hlen); 1914 if (copym != NULL) { 1915 /* 1916 * We don't bother to fragment if the IP length is greater 1917 * than the interface's MTU. Can this possibly matter? 1918 */ 1919 ip = mtod(copym, struct ip *); 1920 HTONS(ip->ip_len); 1921 HTONS(ip->ip_off); 1922 ip->ip_sum = 0; 1923 if (ip->ip_vhl == IP_VHL_BORING) { 1924 ip->ip_sum = in_cksum_hdr(ip); 1925 } else { 1926 ip->ip_sum = in_cksum(copym, hlen); 1927 } 1928 /* 1929 * NB: 1930 * It's not clear whether there are any lingering 1931 * reentrancy problems in other areas which might 1932 * be exposed by using ip_input directly (in 1933 * particular, everything which modifies the packet 1934 * in-place). Yet another option is using the 1935 * protosw directly to deliver the looped back 1936 * packet. For the moment, we'll err on the side 1937 * of safety by using if_simloop(). 1938 */ 1939 #if 1 /* XXX */ 1940 if (dst->sin_family != AF_INET) { 1941 printf("ip_mloopback: bad address family %d\n", 1942 dst->sin_family); 1943 dst->sin_family = AF_INET; 1944 } 1945 #endif 1946 1947 #ifdef notdef 1948 copym->m_pkthdr.rcvif = ifp; 1949 ip_input(copym); 1950 #else 1951 /* if the checksum hasn't been computed, mark it as valid */ 1952 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1953 copym->m_pkthdr.csum_flags |= 1954 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1955 copym->m_pkthdr.csum_data = 0xffff; 1956 } 1957 if_simloop(ifp, copym, dst->sin_family, 0); 1958 #endif 1959 } 1960 } 1961