ip_output.c revision 84102
1/* 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 34 * $FreeBSD: head/sys/netinet/ip_output.c 84102 2001-09-29 04:34:11Z jlemon $ 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 68static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 69 70#ifdef IPSEC 71#include <netinet6/ipsec.h> 72#include <netkey/key.h> 73#ifdef IPSEC_DEBUG 74#include <netkey/key_debug.h> 75#else 76#define KEYDEBUG(lev,arg) 77#endif 78#endif /*IPSEC*/ 79 80#include <netinet/ip_fw.h> 81 82#ifdef DUMMYNET 83#include <netinet/ip_dummynet.h> 84#endif 85 86#ifdef IPFIREWALL_FORWARD_DEBUG 87#define print_ip(a) printf("%ld.%ld.%ld.%ld",(ntohl(a.s_addr)>>24)&0xFF,\ 88 (ntohl(a.s_addr)>>16)&0xFF,\ 89 (ntohl(a.s_addr)>>8)&0xFF,\ 90 (ntohl(a.s_addr))&0xFF); 91#endif 92 93u_short ip_id; 94 95static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *)); 96static struct ifnet *ip_multicast_if __P((struct in_addr *, int *)); 97static void ip_mloopback 98 __P((struct ifnet *, struct mbuf *, struct sockaddr_in *, int)); 99static int ip_getmoptions 100 __P((struct sockopt *, struct ip_moptions *)); 101static int ip_pcbopts __P((int, struct mbuf **, struct mbuf *)); 102static int ip_setmoptions 103 __P((struct sockopt *, struct ip_moptions **)); 104 105int ip_optcopy __P((struct ip *, struct ip *)); 106 107 108extern struct protosw inetsw[]; 109 110/* 111 * IP output. The packet in mbuf chain m contains a skeletal IP 112 * header (with len, off, ttl, proto, tos, src, dst). 113 * The mbuf chain containing the packet will be freed. 114 * The mbuf opt, if present, will not be freed. 115 */ 116int 117ip_output(m0, opt, ro, flags, imo) 118 struct mbuf *m0; 119 struct mbuf *opt; 120 struct route *ro; 121 int flags; 122 struct ip_moptions *imo; 123{ 124 struct ip *ip, *mhip; 125 struct ifnet *ifp; 126 struct mbuf *m = m0; 127 int hlen = sizeof (struct ip); 128 int len, off, error = 0; 129 struct sockaddr_in *dst; 130 struct in_ifaddr *ia; 131 int isbroadcast, sw_csum; 132#ifdef IPSEC 133 struct route iproute; 134 struct socket *so = NULL; 135 struct secpolicy *sp = NULL; 136#endif 137 u_int16_t divert_cookie; /* firewall cookie */ 138#ifdef PFIL_HOOKS 139 struct packet_filter_hook *pfh; 140 struct mbuf *m1; 141 int rv; 142#endif /* PFIL_HOOKS */ 143#ifdef IPFIREWALL_FORWARD 144 int fwd_rewrite_src = 0; 145#endif 146 struct ip_fw *rule = NULL; 147 148#ifdef IPDIVERT 149 /* Get and reset firewall cookie */ 150 divert_cookie = ip_divert_cookie; 151 ip_divert_cookie = 0; 152#else 153 divert_cookie = 0; 154#endif 155 156#if defined(IPFIREWALL) && defined(DUMMYNET) 157 /* 158 * dummynet packet are prepended a vestigial mbuf with 159 * m_type = MT_DUMMYNET and m_data pointing to the matching 160 * rule. 161 */ 162 if (m->m_type == MT_DUMMYNET) { 163 /* 164 * the packet was already tagged, so part of the 165 * processing was already done, and we need to go down. 166 * Get parameters from the header. 167 */ 168 rule = (struct ip_fw *)(m->m_data) ; 169 opt = NULL ; 170 ro = & ( ((struct dn_pkt *)m)->ro ) ; 171 imo = NULL ; 172 dst = ((struct dn_pkt *)m)->dn_dst ; 173 ifp = ((struct dn_pkt *)m)->ifp ; 174 flags = ((struct dn_pkt *)m)->flags ; 175 176 m0 = m = m->m_next ; 177#ifdef IPSEC 178 so = ipsec_getsocket(m); 179 (void)ipsec_setsocket(m, NULL); 180#endif 181 ip = mtod(m, struct ip *); 182 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ; 183 ia = ifatoia(ro->ro_rt->rt_ifa); 184 goto sendit; 185 } else 186 rule = NULL ; 187#endif 188#ifdef IPSEC 189 so = ipsec_getsocket(m); 190 (void)ipsec_setsocket(m, NULL); 191#endif 192 193#ifdef DIAGNOSTIC 194 if ((m->m_flags & M_PKTHDR) == 0) 195 panic("ip_output no HDR"); 196 if (!ro) 197 panic("ip_output no route, proto = %d", 198 mtod(m, struct ip *)->ip_p); 199#endif 200 if (opt) { 201 m = ip_insertoptions(m, opt, &len); 202 hlen = len; 203 } 204 ip = mtod(m, struct ip *); 205 /* 206 * Fill in IP header. 207 */ 208 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 209 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2); 210 ip->ip_off &= IP_DF; 211#ifdef RANDOM_IP_ID 212 ip->ip_id = ip_randomid(); 213#else 214 ip->ip_id = htons(ip_id++); 215#endif 216 ipstat.ips_localout++; 217 } else { 218 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 219 } 220 221 dst = (struct sockaddr_in *)&ro->ro_dst; 222 /* 223 * If there is a cached route, 224 * check that it is to the same destination 225 * and is still up. If not, free it and try again. 226 */ 227 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 228 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 229 RTFREE(ro->ro_rt); 230 ro->ro_rt = (struct rtentry *)0; 231 } 232 if (ro->ro_rt == 0) { 233 dst->sin_family = AF_INET; 234 dst->sin_len = sizeof(*dst); 235 dst->sin_addr = ip->ip_dst; 236 } 237 /* 238 * If routing to interface only, 239 * short circuit routing lookup. 240 */ 241 if (flags & IP_ROUTETOIF) { 242 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 243 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 244 ipstat.ips_noroute++; 245 error = ENETUNREACH; 246 goto bad; 247 } 248 ifp = ia->ia_ifp; 249 ip->ip_ttl = 1; 250 isbroadcast = in_broadcast(dst->sin_addr, ifp); 251 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 252 imo != NULL && imo->imo_multicast_ifp != NULL) { 253 /* 254 * Bypass the normal routing lookup for multicast 255 * packets if the interface is specified. 256 */ 257 ifp = imo->imo_multicast_ifp; 258 IFP_TO_IA(ifp, ia); 259 isbroadcast = 0; /* fool gcc */ 260 } else { 261 /* 262 * If this is the case, we probably don't want to allocate 263 * a protocol-cloned route since we didn't get one from the 264 * ULP. This lets TCP do its thing, while not burdening 265 * forwarding or ICMP with the overhead of cloning a route. 266 * Of course, we still want to do any cloning requested by 267 * the link layer, as this is probably required in all cases 268 * for correct operation (as it is for ARP). 269 */ 270 if (ro->ro_rt == 0) 271 rtalloc_ign(ro, RTF_PRCLONING); 272 if (ro->ro_rt == 0) { 273 ipstat.ips_noroute++; 274 error = EHOSTUNREACH; 275 goto bad; 276 } 277 ia = ifatoia(ro->ro_rt->rt_ifa); 278 ifp = ro->ro_rt->rt_ifp; 279 ro->ro_rt->rt_use++; 280 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 281 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 282 if (ro->ro_rt->rt_flags & RTF_HOST) 283 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 284 else 285 isbroadcast = in_broadcast(dst->sin_addr, ifp); 286 } 287 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 288 struct in_multi *inm; 289 290 m->m_flags |= M_MCAST; 291 /* 292 * IP destination address is multicast. Make sure "dst" 293 * still points to the address in "ro". (It may have been 294 * changed to point to a gateway address, above.) 295 */ 296 dst = (struct sockaddr_in *)&ro->ro_dst; 297 /* 298 * See if the caller provided any multicast options 299 */ 300 if (imo != NULL) { 301 ip->ip_ttl = imo->imo_multicast_ttl; 302 if (imo->imo_multicast_vif != -1) 303 ip->ip_src.s_addr = 304 ip_mcast_src(imo->imo_multicast_vif); 305 } else 306 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 307 /* 308 * Confirm that the outgoing interface supports multicast. 309 */ 310 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 311 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 312 ipstat.ips_noroute++; 313 error = ENETUNREACH; 314 goto bad; 315 } 316 } 317 /* 318 * If source address not specified yet, use address 319 * of outgoing interface. 320 */ 321 if (ip->ip_src.s_addr == INADDR_ANY) { 322 /* Interface may have no addresses. */ 323 if (ia != NULL) 324 ip->ip_src = IA_SIN(ia)->sin_addr; 325 } 326 327 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 328 if (inm != NULL && 329 (imo == NULL || imo->imo_multicast_loop)) { 330 /* 331 * If we belong to the destination multicast group 332 * on the outgoing interface, and the caller did not 333 * forbid loopback, loop back a copy. 334 */ 335 ip_mloopback(ifp, m, dst, hlen); 336 } 337 else { 338 /* 339 * If we are acting as a multicast router, perform 340 * multicast forwarding as if the packet had just 341 * arrived on the interface to which we are about 342 * to send. The multicast forwarding function 343 * recursively calls this function, using the 344 * IP_FORWARDING flag to prevent infinite recursion. 345 * 346 * Multicasts that are looped back by ip_mloopback(), 347 * above, will be forwarded by the ip_input() routine, 348 * if necessary. 349 */ 350 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 351 /* 352 * Check if rsvp daemon is running. If not, don't 353 * set ip_moptions. This ensures that the packet 354 * is multicast and not just sent down one link 355 * as prescribed by rsvpd. 356 */ 357 if (!rsvp_on) 358 imo = NULL; 359 if (ip_mforward(ip, ifp, m, imo) != 0) { 360 m_freem(m); 361 goto done; 362 } 363 } 364 } 365 366 /* 367 * Multicasts with a time-to-live of zero may be looped- 368 * back, above, but must not be transmitted on a network. 369 * Also, multicasts addressed to the loopback interface 370 * are not sent -- the above call to ip_mloopback() will 371 * loop back a copy if this host actually belongs to the 372 * destination group on the loopback interface. 373 */ 374 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 375 m_freem(m); 376 goto done; 377 } 378 379 goto sendit; 380 } 381#ifndef notdef 382 /* 383 * If source address not specified yet, use address 384 * of outgoing interface. 385 */ 386 if (ip->ip_src.s_addr == INADDR_ANY) { 387 /* Interface may have no addresses. */ 388 if (ia != NULL) { 389 ip->ip_src = IA_SIN(ia)->sin_addr; 390#ifdef IPFIREWALL_FORWARD 391 /* Keep note that we did this - if the firewall changes 392 * the next-hop, our interface may change, changing the 393 * default source IP. It's a shame so much effort happens 394 * twice. Oh well. 395 */ 396 fwd_rewrite_src++; 397#endif /* IPFIREWALL_FORWARD */ 398 } 399 } 400#endif /* notdef */ 401 /* 402 * Verify that we have any chance at all of being able to queue 403 * the packet or packet fragments 404 */ 405 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 406 ifp->if_snd.ifq_maxlen) { 407 error = ENOBUFS; 408 goto bad; 409 } 410 411 /* 412 * Look for broadcast address and 413 * and verify user is allowed to send 414 * such a packet. 415 */ 416 if (isbroadcast) { 417 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 418 error = EADDRNOTAVAIL; 419 goto bad; 420 } 421 if ((flags & IP_ALLOWBROADCAST) == 0) { 422 error = EACCES; 423 goto bad; 424 } 425 /* don't allow broadcast messages to be fragmented */ 426 if ((u_short)ip->ip_len > ifp->if_mtu) { 427 error = EMSGSIZE; 428 goto bad; 429 } 430 m->m_flags |= M_BCAST; 431 } else { 432 m->m_flags &= ~M_BCAST; 433 } 434 435sendit: 436#ifdef IPSEC 437 /* get SP for this packet */ 438 if (so == NULL) 439 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 440 else 441 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 442 443 if (sp == NULL) { 444 ipsecstat.out_inval++; 445 goto bad; 446 } 447 448 error = 0; 449 450 /* check policy */ 451 switch (sp->policy) { 452 case IPSEC_POLICY_DISCARD: 453 /* 454 * This packet is just discarded. 455 */ 456 ipsecstat.out_polvio++; 457 goto bad; 458 459 case IPSEC_POLICY_BYPASS: 460 case IPSEC_POLICY_NONE: 461 /* no need to do IPsec. */ 462 goto skip_ipsec; 463 464 case IPSEC_POLICY_IPSEC: 465 if (sp->req == NULL) { 466 /* acquire a policy */ 467 error = key_spdacquire(sp); 468 goto bad; 469 } 470 break; 471 472 case IPSEC_POLICY_ENTRUST: 473 default: 474 printf("ip_output: Invalid policy found. %d\n", sp->policy); 475 } 476 { 477 struct ipsec_output_state state; 478 bzero(&state, sizeof(state)); 479 state.m = m; 480 if (flags & IP_ROUTETOIF) { 481 state.ro = &iproute; 482 bzero(&iproute, sizeof(iproute)); 483 } else 484 state.ro = ro; 485 state.dst = (struct sockaddr *)dst; 486 487 ip->ip_sum = 0; 488 489 /* 490 * XXX 491 * delayed checksums are not currently compatible with IPsec 492 */ 493 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 494 in_delayed_cksum(m); 495 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 496 } 497 498 HTONS(ip->ip_len); 499 HTONS(ip->ip_off); 500 501 error = ipsec4_output(&state, sp, flags); 502 503 m = state.m; 504 if (flags & IP_ROUTETOIF) { 505 /* 506 * if we have tunnel mode SA, we may need to ignore 507 * IP_ROUTETOIF. 508 */ 509 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 510 flags &= ~IP_ROUTETOIF; 511 ro = state.ro; 512 } 513 } else 514 ro = state.ro; 515 dst = (struct sockaddr_in *)state.dst; 516 if (error) { 517 /* mbuf is already reclaimed in ipsec4_output. */ 518 m0 = NULL; 519 switch (error) { 520 case EHOSTUNREACH: 521 case ENETUNREACH: 522 case EMSGSIZE: 523 case ENOBUFS: 524 case ENOMEM: 525 break; 526 default: 527 printf("ip4_output (ipsec): error code %d\n", error); 528 /*fall through*/ 529 case ENOENT: 530 /* don't show these error codes to the user */ 531 error = 0; 532 break; 533 } 534 goto bad; 535 } 536 } 537 538 /* be sure to update variables that are affected by ipsec4_output() */ 539 ip = mtod(m, struct ip *); 540#ifdef _IP_VHL 541 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 542#else 543 hlen = ip->ip_hl << 2; 544#endif 545 if (ro->ro_rt == NULL) { 546 if ((flags & IP_ROUTETOIF) == 0) { 547 printf("ip_output: " 548 "can't update route after IPsec processing\n"); 549 error = EHOSTUNREACH; /*XXX*/ 550 goto bad; 551 } 552 } else { 553 ia = ifatoia(ro->ro_rt->rt_ifa); 554 ifp = ro->ro_rt->rt_ifp; 555 } 556 557 /* make it flipped, again. */ 558 NTOHS(ip->ip_len); 559 NTOHS(ip->ip_off); 560skip_ipsec: 561#endif /*IPSEC*/ 562 563 /* 564 * IpHack's section. 565 * - Xlate: translate packet's addr/port (NAT). 566 * - Firewall: deny/allow/etc. 567 * - Wrap: fake packet's addr/port <unimpl.> 568 * - Encapsulate: put it in another IP and send out. <unimp.> 569 */ 570#ifdef PFIL_HOOKS 571 /* 572 * Run through list of hooks for output packets. 573 */ 574 m1 = m; 575 pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 576 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 577 if (pfh->pfil_func) { 578 rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1); 579 if (rv) { 580 error = EHOSTUNREACH; 581 goto done; 582 } 583 m = m1; 584 if (m == NULL) 585 goto done; 586 ip = mtod(m, struct ip *); 587 } 588#endif /* PFIL_HOOKS */ 589 590 /* 591 * Check with the firewall... 592 */ 593 if (fw_enable && ip_fw_chk_ptr) { 594 struct sockaddr_in *old = dst; 595 596 off = (*ip_fw_chk_ptr)(&ip, 597 hlen, ifp, &divert_cookie, &m, &rule, &dst); 598 /* 599 * On return we must do the following: 600 * m == NULL -> drop the pkt (old interface, deprecated) 601 * (off & 0x40000) -> drop the pkt (new interface) 602 * 1<=off<= 0xffff -> DIVERT 603 * (off & 0x10000) -> send to a DUMMYNET pipe 604 * (off & 0x20000) -> TEE the packet 605 * dst != old -> IPFIREWALL_FORWARD 606 * off==0, dst==old -> accept 607 * If some of the above modules is not compiled in, then 608 * we should't have to check the corresponding condition 609 * (because the ipfw control socket should not accept 610 * unsupported rules), but better play safe and drop 611 * packets in case of doubt. 612 */ 613 if (off & IP_FW_PORT_DENY_FLAG) { /* XXX new interface-denied */ 614 if (m) 615 m_freem(m); 616 error = EACCES ; 617 goto done; 618 } 619 if (!m) { /* firewall said to reject */ 620 static int __debug=10; 621 if (__debug >0) { 622 printf("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#ifdef DUMMYNET 631 if ((off & IP_FW_PORT_DYNT_FLAG) != 0) { 632 /* 633 * pass the pkt to dummynet. Need to include 634 * pipe number, m, ifp, ro, dst because these are 635 * not recomputed in the next pass. 636 * All other parameters have been already used and 637 * so they are not needed anymore. 638 * XXX note: if the ifp or ro entry are deleted 639 * while a pkt is in dummynet, we are in trouble! 640 */ 641 error = dummynet_io(off & 0xffff, DN_TO_IP_OUT, m, 642 ifp,ro,dst,rule, flags); 643 goto done; 644 } 645#endif 646#ifdef IPDIVERT 647 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 648 struct mbuf *clone = NULL; 649 650 /* Clone packet if we're doing a 'tee' */ 651 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 652 clone = m_dup(m, M_DONTWAIT); 653 654 /* 655 * XXX 656 * delayed checksums are not currently compatible 657 * with divert sockets. 658 */ 659 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 660 in_delayed_cksum(m); 661 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 662 } 663 664 /* Restore packet header fields to original values */ 665 HTONS(ip->ip_len); 666 HTONS(ip->ip_off); 667 668 /* Deliver packet to divert input routine */ 669 ip_divert_cookie = divert_cookie; 670 divert_packet(m, 0, off & 0xffff); 671 672 /* If 'tee', continue with original packet */ 673 if (clone != NULL) { 674 m = clone; 675 ip = mtod(m, struct ip *); 676 goto pass; 677 } 678 goto done; 679 } 680#endif 681 682#ifdef IPFIREWALL_FORWARD 683 /* Here we check dst to make sure it's directly reachable on the 684 * interface we previously thought it was. 685 * If it isn't (which may be likely in some situations) we have 686 * to re-route it (ie, find a route for the next-hop and the 687 * associated interface) and set them here. This is nested 688 * forwarding which in most cases is undesirable, except where 689 * such control is nigh impossible. So we do it here. 690 * And I'm babbling. 691 */ 692 if (off == 0 && old != dst) { 693 struct in_ifaddr *ia; 694 695 /* It's changed... */ 696 /* There must be a better way to do this next line... */ 697 static struct route sro_fwd, *ro_fwd = &sro_fwd; 698#ifdef IPFIREWALL_FORWARD_DEBUG 699 printf("IPFIREWALL_FORWARD: New dst ip: "); 700 print_ip(dst->sin_addr); 701 printf("\n"); 702#endif 703 /* 704 * We need to figure out if we have been forwarded 705 * to a local socket. If so then we should somehow 706 * "loop back" to ip_input, and get directed to the 707 * PCB as if we had received this packet. This is 708 * because it may be dificult to identify the packets 709 * you want to forward until they are being output 710 * and have selected an interface. (e.g. locally 711 * initiated packets) If we used the loopback inteface, 712 * we would not be able to control what happens 713 * as the packet runs through ip_input() as 714 * it is done through a ISR. 715 */ 716 LIST_FOREACH(ia, 717 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 718 /* 719 * If the addr to forward to is one 720 * of ours, we pretend to 721 * be the destination for this packet. 722 */ 723 if (IA_SIN(ia)->sin_addr.s_addr == 724 dst->sin_addr.s_addr) 725 break; 726 } 727 if (ia) { 728 /* tell ip_input "dont filter" */ 729 ip_fw_fwd_addr = dst; 730 if (m->m_pkthdr.rcvif == NULL) 731 m->m_pkthdr.rcvif = ifunit("lo0"); 732 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 733 m->m_pkthdr.csum_flags |= 734 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 735 m0->m_pkthdr.csum_data = 0xffff; 736 } 737 m->m_pkthdr.csum_flags |= 738 CSUM_IP_CHECKED | CSUM_IP_VALID; 739 HTONS(ip->ip_len); 740 HTONS(ip->ip_off); 741 ip_input(m); 742 goto done; 743 } 744 /* Some of the logic for this was 745 * nicked from above. 746 * 747 * This rewrites the cached route in a local PCB. 748 * Is this what we want to do? 749 */ 750 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 751 752 ro_fwd->ro_rt = 0; 753 rtalloc_ign(ro_fwd, RTF_PRCLONING); 754 755 if (ro_fwd->ro_rt == 0) { 756 ipstat.ips_noroute++; 757 error = EHOSTUNREACH; 758 goto bad; 759 } 760 761 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 762 ifp = ro_fwd->ro_rt->rt_ifp; 763 ro_fwd->ro_rt->rt_use++; 764 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 765 dst = (struct sockaddr_in *)ro_fwd->ro_rt->rt_gateway; 766 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 767 isbroadcast = 768 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 769 else 770 isbroadcast = in_broadcast(dst->sin_addr, ifp); 771 RTFREE(ro->ro_rt); 772 ro->ro_rt = ro_fwd->ro_rt; 773 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 774 775 /* 776 * If we added a default src ip earlier, 777 * which would have been gotten from the-then 778 * interface, do it again, from the new one. 779 */ 780 if (fwd_rewrite_src) 781 ip->ip_src = IA_SIN(ia)->sin_addr; 782 goto pass ; 783 } 784#endif /* IPFIREWALL_FORWARD */ 785 /* 786 * if we get here, none of the above matches, and 787 * we have to drop the pkt 788 */ 789 m_freem(m); 790 error = EACCES; /* not sure this is the right error msg */ 791 goto done; 792 } 793 794pass: 795 m->m_pkthdr.csum_flags |= CSUM_IP; 796 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 797 if (sw_csum & CSUM_DELAY_DATA) { 798 in_delayed_cksum(m); 799 sw_csum &= ~CSUM_DELAY_DATA; 800 } 801 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 802 803 /* 804 * If small enough for interface, or the interface will take 805 * care of the fragmentation for us, can just send directly. 806 */ 807 if ((u_short)ip->ip_len <= ifp->if_mtu || 808 ifp->if_hwassist & CSUM_FRAGMENT) { 809 HTONS(ip->ip_len); 810 HTONS(ip->ip_off); 811 ip->ip_sum = 0; 812 if (sw_csum & CSUM_DELAY_IP) { 813 if (ip->ip_vhl == IP_VHL_BORING) { 814 ip->ip_sum = in_cksum_hdr(ip); 815 } else { 816 ip->ip_sum = in_cksum(m, hlen); 817 } 818 } 819 820 /* Record statistics for this interface address. */ 821 if (!(flags & IP_FORWARDING) && ia) { 822 ia->ia_ifa.if_opackets++; 823 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 824 } 825 826#ifdef IPSEC 827 /* clean ipsec history once it goes out of the node */ 828 ipsec_delaux(m); 829#endif 830 831 error = (*ifp->if_output)(ifp, m, 832 (struct sockaddr *)dst, ro->ro_rt); 833 goto done; 834 } 835 /* 836 * Too large for interface; fragment if possible. 837 * Must be able to put at least 8 bytes per fragment. 838 */ 839 if (ip->ip_off & IP_DF) { 840 error = EMSGSIZE; 841 /* 842 * This case can happen if the user changed the MTU 843 * of an interface after enabling IP on it. Because 844 * most netifs don't keep track of routes pointing to 845 * them, there is no way for one to update all its 846 * routes when the MTU is changed. 847 */ 848 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 849 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 850 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 851 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 852 } 853 ipstat.ips_cantfrag++; 854 goto bad; 855 } 856 len = (ifp->if_mtu - hlen) &~ 7; 857 if (len < 8) { 858 error = EMSGSIZE; 859 goto bad; 860 } 861 862 /* 863 * if the interface will not calculate checksums on 864 * fragmented packets, then do it here. 865 */ 866 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 867 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 868 in_delayed_cksum(m); 869 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 870 } 871 872 { 873 int mhlen, firstlen = len; 874 struct mbuf **mnext = &m->m_nextpkt; 875 int nfrags = 1; 876 877 /* 878 * Loop through length of segment after first fragment, 879 * make new header and copy data of each part and link onto chain. 880 */ 881 m0 = m; 882 mhlen = sizeof (struct ip); 883 for (off = hlen + len; off < (u_short)ip->ip_len; off += len) { 884 MGETHDR(m, M_DONTWAIT, MT_HEADER); 885 if (m == 0) { 886 error = ENOBUFS; 887 ipstat.ips_odropped++; 888 goto sendorfree; 889 } 890 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 891 m->m_data += max_linkhdr; 892 mhip = mtod(m, struct ip *); 893 *mhip = *ip; 894 if (hlen > sizeof (struct ip)) { 895 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 896 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2); 897 } 898 m->m_len = mhlen; 899 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 900 if (off + len >= (u_short)ip->ip_len) 901 len = (u_short)ip->ip_len - off; 902 else 903 mhip->ip_off |= IP_MF; 904 mhip->ip_len = htons((u_short)(len + mhlen)); 905 m->m_next = m_copy(m0, off, len); 906 if (m->m_next == 0) { 907 (void) m_free(m); 908 error = ENOBUFS; /* ??? */ 909 ipstat.ips_odropped++; 910 goto sendorfree; 911 } 912 m->m_pkthdr.len = mhlen + len; 913 m->m_pkthdr.rcvif = (struct ifnet *)0; 914 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 915 HTONS(mhip->ip_off); 916 mhip->ip_sum = 0; 917 if (sw_csum & CSUM_DELAY_IP) { 918 if (mhip->ip_vhl == IP_VHL_BORING) { 919 mhip->ip_sum = in_cksum_hdr(mhip); 920 } else { 921 mhip->ip_sum = in_cksum(m, mhlen); 922 } 923 } 924 *mnext = m; 925 mnext = &m->m_nextpkt; 926 nfrags++; 927 } 928 ipstat.ips_ofragments += nfrags; 929 930 /* set first/last markers for fragment chain */ 931 m->m_flags |= M_LASTFRAG; 932 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 933 m0->m_pkthdr.csum_data = nfrags; 934 935 /* 936 * Update first fragment by trimming what's been copied out 937 * and updating header, then send each fragment (in order). 938 */ 939 m = m0; 940 m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 941 m->m_pkthdr.len = hlen + firstlen; 942 ip->ip_len = htons((u_short)m->m_pkthdr.len); 943 ip->ip_off |= IP_MF; 944 HTONS(ip->ip_off); 945 ip->ip_sum = 0; 946 if (sw_csum & CSUM_DELAY_IP) { 947 if (ip->ip_vhl == IP_VHL_BORING) { 948 ip->ip_sum = in_cksum_hdr(ip); 949 } else { 950 ip->ip_sum = in_cksum(m, hlen); 951 } 952 } 953sendorfree: 954 for (m = m0; m; m = m0) { 955 m0 = m->m_nextpkt; 956 m->m_nextpkt = 0; 957#ifdef IPSEC 958 /* clean ipsec history once it goes out of the node */ 959 ipsec_delaux(m); 960#endif 961 if (error == 0) { 962 /* Record statistics for this interface address. */ 963 if (ia != NULL) { 964 ia->ia_ifa.if_opackets++; 965 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 966 } 967 968 error = (*ifp->if_output)(ifp, m, 969 (struct sockaddr *)dst, ro->ro_rt); 970 } else 971 m_freem(m); 972 } 973 974 if (error == 0) 975 ipstat.ips_fragmented++; 976 } 977done: 978#ifdef IPSEC 979 if (ro == &iproute && ro->ro_rt) { 980 RTFREE(ro->ro_rt); 981 ro->ro_rt = NULL; 982 } 983 if (sp != NULL) { 984 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 985 printf("DP ip_output call free SP:%p\n", sp)); 986 key_freesp(sp); 987 } 988#endif /* IPSEC */ 989 return (error); 990bad: 991 m_freem(m0); 992 goto done; 993} 994 995void 996in_delayed_cksum(struct mbuf *m) 997{ 998 struct ip *ip; 999 u_short csum, offset; 1000 1001 ip = mtod(m, struct ip *); 1002 offset = IP_VHL_HL(ip->ip_vhl) << 2 ; 1003 csum = in_cksum_skip(m, ip->ip_len, offset); 1004 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1005 csum = 0xffff; 1006 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1007 1008 if (offset + sizeof(u_short) > m->m_len) { 1009 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1010 m->m_len, offset, ip->ip_p); 1011 /* 1012 * XXX 1013 * this shouldn't happen, but if it does, the 1014 * correct behavior may be to insert the checksum 1015 * in the existing chain instead of rearranging it. 1016 */ 1017 m = m_pullup(m, offset + sizeof(u_short)); 1018 } 1019 *(u_short *)(m->m_data + offset) = csum; 1020} 1021 1022/* 1023 * Insert IP options into preformed packet. 1024 * Adjust IP destination as required for IP source routing, 1025 * as indicated by a non-zero in_addr at the start of the options. 1026 * 1027 * XXX This routine assumes that the packet has no options in place. 1028 */ 1029static struct mbuf * 1030ip_insertoptions(m, opt, phlen) 1031 register struct mbuf *m; 1032 struct mbuf *opt; 1033 int *phlen; 1034{ 1035 register struct ipoption *p = mtod(opt, struct ipoption *); 1036 struct mbuf *n; 1037 register struct ip *ip = mtod(m, struct ip *); 1038 unsigned optlen; 1039 1040 optlen = opt->m_len - sizeof(p->ipopt_dst); 1041 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) 1042 return (m); /* XXX should fail */ 1043 if (p->ipopt_dst.s_addr) 1044 ip->ip_dst = p->ipopt_dst; 1045 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1046 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1047 if (n == 0) 1048 return (m); 1049 n->m_pkthdr.rcvif = (struct ifnet *)0; 1050 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1051 m->m_len -= sizeof(struct ip); 1052 m->m_data += sizeof(struct ip); 1053 n->m_next = m; 1054 m = n; 1055 m->m_len = optlen + sizeof(struct ip); 1056 m->m_data += max_linkhdr; 1057 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1058 } else { 1059 m->m_data -= optlen; 1060 m->m_len += optlen; 1061 m->m_pkthdr.len += optlen; 1062 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1063 } 1064 ip = mtod(m, struct ip *); 1065 bcopy(p->ipopt_list, ip + 1, optlen); 1066 *phlen = sizeof(struct ip) + optlen; 1067 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2); 1068 ip->ip_len += optlen; 1069 return (m); 1070} 1071 1072/* 1073 * Copy options from ip to jp, 1074 * omitting those not copied during fragmentation. 1075 */ 1076int 1077ip_optcopy(ip, jp) 1078 struct ip *ip, *jp; 1079{ 1080 register u_char *cp, *dp; 1081 int opt, optlen, cnt; 1082 1083 cp = (u_char *)(ip + 1); 1084 dp = (u_char *)(jp + 1); 1085 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1086 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1087 opt = cp[0]; 1088 if (opt == IPOPT_EOL) 1089 break; 1090 if (opt == IPOPT_NOP) { 1091 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1092 *dp++ = IPOPT_NOP; 1093 optlen = 1; 1094 continue; 1095 } 1096#ifdef DIAGNOSTIC 1097 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1098 panic("malformed IPv4 option passed to ip_optcopy"); 1099#endif 1100 optlen = cp[IPOPT_OLEN]; 1101#ifdef DIAGNOSTIC 1102 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1103 panic("malformed IPv4 option passed to ip_optcopy"); 1104#endif 1105 /* bogus lengths should have been caught by ip_dooptions */ 1106 if (optlen > cnt) 1107 optlen = cnt; 1108 if (IPOPT_COPIED(opt)) { 1109 bcopy(cp, dp, optlen); 1110 dp += optlen; 1111 } 1112 } 1113 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1114 *dp++ = IPOPT_EOL; 1115 return (optlen); 1116} 1117 1118/* 1119 * IP socket option processing. 1120 */ 1121int 1122ip_ctloutput(so, sopt) 1123 struct socket *so; 1124 struct sockopt *sopt; 1125{ 1126 struct inpcb *inp = sotoinpcb(so); 1127 int error, optval; 1128 1129 error = optval = 0; 1130 if (sopt->sopt_level != IPPROTO_IP) { 1131 return (EINVAL); 1132 } 1133 1134 switch (sopt->sopt_dir) { 1135 case SOPT_SET: 1136 switch (sopt->sopt_name) { 1137 case IP_OPTIONS: 1138#ifdef notyet 1139 case IP_RETOPTS: 1140#endif 1141 { 1142 struct mbuf *m; 1143 if (sopt->sopt_valsize > MLEN) { 1144 error = EMSGSIZE; 1145 break; 1146 } 1147 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1148 if (m == 0) { 1149 error = ENOBUFS; 1150 break; 1151 } 1152 m->m_len = sopt->sopt_valsize; 1153 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1154 m->m_len); 1155 1156 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1157 m)); 1158 } 1159 1160 case IP_TOS: 1161 case IP_TTL: 1162 case IP_RECVOPTS: 1163 case IP_RECVRETOPTS: 1164 case IP_RECVDSTADDR: 1165 case IP_RECVIF: 1166 case IP_FAITH: 1167 error = sooptcopyin(sopt, &optval, sizeof optval, 1168 sizeof optval); 1169 if (error) 1170 break; 1171 1172 switch (sopt->sopt_name) { 1173 case IP_TOS: 1174 inp->inp_ip_tos = optval; 1175 break; 1176 1177 case IP_TTL: 1178 inp->inp_ip_ttl = optval; 1179 break; 1180#define OPTSET(bit) \ 1181 if (optval) \ 1182 inp->inp_flags |= bit; \ 1183 else \ 1184 inp->inp_flags &= ~bit; 1185 1186 case IP_RECVOPTS: 1187 OPTSET(INP_RECVOPTS); 1188 break; 1189 1190 case IP_RECVRETOPTS: 1191 OPTSET(INP_RECVRETOPTS); 1192 break; 1193 1194 case IP_RECVDSTADDR: 1195 OPTSET(INP_RECVDSTADDR); 1196 break; 1197 1198 case IP_RECVIF: 1199 OPTSET(INP_RECVIF); 1200 break; 1201 1202 case IP_FAITH: 1203 OPTSET(INP_FAITH); 1204 break; 1205 } 1206 break; 1207#undef OPTSET 1208 1209 case IP_MULTICAST_IF: 1210 case IP_MULTICAST_VIF: 1211 case IP_MULTICAST_TTL: 1212 case IP_MULTICAST_LOOP: 1213 case IP_ADD_MEMBERSHIP: 1214 case IP_DROP_MEMBERSHIP: 1215 error = ip_setmoptions(sopt, &inp->inp_moptions); 1216 break; 1217 1218 case IP_PORTRANGE: 1219 error = sooptcopyin(sopt, &optval, sizeof optval, 1220 sizeof optval); 1221 if (error) 1222 break; 1223 1224 switch (optval) { 1225 case IP_PORTRANGE_DEFAULT: 1226 inp->inp_flags &= ~(INP_LOWPORT); 1227 inp->inp_flags &= ~(INP_HIGHPORT); 1228 break; 1229 1230 case IP_PORTRANGE_HIGH: 1231 inp->inp_flags &= ~(INP_LOWPORT); 1232 inp->inp_flags |= INP_HIGHPORT; 1233 break; 1234 1235 case IP_PORTRANGE_LOW: 1236 inp->inp_flags &= ~(INP_HIGHPORT); 1237 inp->inp_flags |= INP_LOWPORT; 1238 break; 1239 1240 default: 1241 error = EINVAL; 1242 break; 1243 } 1244 break; 1245 1246#ifdef IPSEC 1247 case IP_IPSEC_POLICY: 1248 { 1249 caddr_t req; 1250 size_t len = 0; 1251 int priv; 1252 struct mbuf *m; 1253 int optname; 1254 1255 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1256 break; 1257 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1258 break; 1259 priv = (sopt->sopt_td != NULL && 1260 suser_td(sopt->sopt_td) != 0) ? 0 : 1; 1261 req = mtod(m, caddr_t); 1262 len = m->m_len; 1263 optname = sopt->sopt_name; 1264 error = ipsec4_set_policy(inp, optname, req, len, priv); 1265 m_freem(m); 1266 break; 1267 } 1268#endif /*IPSEC*/ 1269 1270 default: 1271 error = ENOPROTOOPT; 1272 break; 1273 } 1274 break; 1275 1276 case SOPT_GET: 1277 switch (sopt->sopt_name) { 1278 case IP_OPTIONS: 1279 case IP_RETOPTS: 1280 if (inp->inp_options) 1281 error = sooptcopyout(sopt, 1282 mtod(inp->inp_options, 1283 char *), 1284 inp->inp_options->m_len); 1285 else 1286 sopt->sopt_valsize = 0; 1287 break; 1288 1289 case IP_TOS: 1290 case IP_TTL: 1291 case IP_RECVOPTS: 1292 case IP_RECVRETOPTS: 1293 case IP_RECVDSTADDR: 1294 case IP_RECVIF: 1295 case IP_PORTRANGE: 1296 case IP_FAITH: 1297 switch (sopt->sopt_name) { 1298 1299 case IP_TOS: 1300 optval = inp->inp_ip_tos; 1301 break; 1302 1303 case IP_TTL: 1304 optval = inp->inp_ip_ttl; 1305 break; 1306 1307#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1308 1309 case IP_RECVOPTS: 1310 optval = OPTBIT(INP_RECVOPTS); 1311 break; 1312 1313 case IP_RECVRETOPTS: 1314 optval = OPTBIT(INP_RECVRETOPTS); 1315 break; 1316 1317 case IP_RECVDSTADDR: 1318 optval = OPTBIT(INP_RECVDSTADDR); 1319 break; 1320 1321 case IP_RECVIF: 1322 optval = OPTBIT(INP_RECVIF); 1323 break; 1324 1325 case IP_PORTRANGE: 1326 if (inp->inp_flags & INP_HIGHPORT) 1327 optval = IP_PORTRANGE_HIGH; 1328 else if (inp->inp_flags & INP_LOWPORT) 1329 optval = IP_PORTRANGE_LOW; 1330 else 1331 optval = 0; 1332 break; 1333 1334 case IP_FAITH: 1335 optval = OPTBIT(INP_FAITH); 1336 break; 1337 } 1338 error = sooptcopyout(sopt, &optval, sizeof optval); 1339 break; 1340 1341 case IP_MULTICAST_IF: 1342 case IP_MULTICAST_VIF: 1343 case IP_MULTICAST_TTL: 1344 case IP_MULTICAST_LOOP: 1345 case IP_ADD_MEMBERSHIP: 1346 case IP_DROP_MEMBERSHIP: 1347 error = ip_getmoptions(sopt, inp->inp_moptions); 1348 break; 1349 1350#ifdef IPSEC 1351 case IP_IPSEC_POLICY: 1352 { 1353 struct mbuf *m = NULL; 1354 caddr_t req = NULL; 1355 size_t len = 0; 1356 1357 if (m != 0) { 1358 req = mtod(m, caddr_t); 1359 len = m->m_len; 1360 } 1361 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1362 if (error == 0) 1363 error = soopt_mcopyout(sopt, m); /* XXX */ 1364 if (error == 0) 1365 m_freem(m); 1366 break; 1367 } 1368#endif /*IPSEC*/ 1369 1370 default: 1371 error = ENOPROTOOPT; 1372 break; 1373 } 1374 break; 1375 } 1376 return (error); 1377} 1378 1379/* 1380 * Set up IP options in pcb for insertion in output packets. 1381 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1382 * with destination address if source routed. 1383 */ 1384static int 1385ip_pcbopts(optname, pcbopt, m) 1386 int optname; 1387 struct mbuf **pcbopt; 1388 register struct mbuf *m; 1389{ 1390 register int cnt, optlen; 1391 register u_char *cp; 1392 u_char opt; 1393 1394 /* turn off any old options */ 1395 if (*pcbopt) 1396 (void)m_free(*pcbopt); 1397 *pcbopt = 0; 1398 if (m == (struct mbuf *)0 || m->m_len == 0) { 1399 /* 1400 * Only turning off any previous options. 1401 */ 1402 if (m) 1403 (void)m_free(m); 1404 return (0); 1405 } 1406 1407 if (m->m_len % sizeof(int32_t)) 1408 goto bad; 1409 /* 1410 * IP first-hop destination address will be stored before 1411 * actual options; move other options back 1412 * and clear it when none present. 1413 */ 1414 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1415 goto bad; 1416 cnt = m->m_len; 1417 m->m_len += sizeof(struct in_addr); 1418 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1419 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1420 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1421 1422 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1423 opt = cp[IPOPT_OPTVAL]; 1424 if (opt == IPOPT_EOL) 1425 break; 1426 if (opt == IPOPT_NOP) 1427 optlen = 1; 1428 else { 1429 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1430 goto bad; 1431 optlen = cp[IPOPT_OLEN]; 1432 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1433 goto bad; 1434 } 1435 switch (opt) { 1436 1437 default: 1438 break; 1439 1440 case IPOPT_LSRR: 1441 case IPOPT_SSRR: 1442 /* 1443 * user process specifies route as: 1444 * ->A->B->C->D 1445 * D must be our final destination (but we can't 1446 * check that since we may not have connected yet). 1447 * A is first hop destination, which doesn't appear in 1448 * actual IP option, but is stored before the options. 1449 */ 1450 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1451 goto bad; 1452 m->m_len -= sizeof(struct in_addr); 1453 cnt -= sizeof(struct in_addr); 1454 optlen -= sizeof(struct in_addr); 1455 cp[IPOPT_OLEN] = optlen; 1456 /* 1457 * Move first hop before start of options. 1458 */ 1459 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1460 sizeof(struct in_addr)); 1461 /* 1462 * Then copy rest of options back 1463 * to close up the deleted entry. 1464 */ 1465 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1466 sizeof(struct in_addr)), 1467 (caddr_t)&cp[IPOPT_OFFSET+1], 1468 (unsigned)cnt + sizeof(struct in_addr)); 1469 break; 1470 } 1471 } 1472 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1473 goto bad; 1474 *pcbopt = m; 1475 return (0); 1476 1477bad: 1478 (void)m_free(m); 1479 return (EINVAL); 1480} 1481 1482/* 1483 * XXX 1484 * The whole multicast option thing needs to be re-thought. 1485 * Several of these options are equally applicable to non-multicast 1486 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1487 * standard option (IP_TTL). 1488 */ 1489 1490/* 1491 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1492 */ 1493static struct ifnet * 1494ip_multicast_if(a, ifindexp) 1495 struct in_addr *a; 1496 int *ifindexp; 1497{ 1498 int ifindex; 1499 struct ifnet *ifp; 1500 1501 if (ifindexp) 1502 *ifindexp = 0; 1503 if (ntohl(a->s_addr) >> 24 == 0) { 1504 ifindex = ntohl(a->s_addr) & 0xffffff; 1505 if (ifindex < 0 || if_index < ifindex) 1506 return NULL; 1507 ifp = ifnet_byindex(ifindex); 1508 if (ifindexp) 1509 *ifindexp = ifindex; 1510 } else { 1511 INADDR_TO_IFP(*a, ifp); 1512 } 1513 return ifp; 1514} 1515 1516/* 1517 * Set the IP multicast options in response to user setsockopt(). 1518 */ 1519static int 1520ip_setmoptions(sopt, imop) 1521 struct sockopt *sopt; 1522 struct ip_moptions **imop; 1523{ 1524 int error = 0; 1525 int i; 1526 struct in_addr addr; 1527 struct ip_mreq mreq; 1528 struct ifnet *ifp; 1529 struct ip_moptions *imo = *imop; 1530 struct route ro; 1531 struct sockaddr_in *dst; 1532 int ifindex; 1533 int s; 1534 1535 if (imo == NULL) { 1536 /* 1537 * No multicast option buffer attached to the pcb; 1538 * allocate one and initialize to default values. 1539 */ 1540 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1541 M_WAITOK); 1542 1543 if (imo == NULL) 1544 return (ENOBUFS); 1545 *imop = imo; 1546 imo->imo_multicast_ifp = NULL; 1547 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1548 imo->imo_multicast_vif = -1; 1549 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1550 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1551 imo->imo_num_memberships = 0; 1552 } 1553 1554 switch (sopt->sopt_name) { 1555 /* store an index number for the vif you wanna use in the send */ 1556 case IP_MULTICAST_VIF: 1557 if (legal_vif_num == 0) { 1558 error = EOPNOTSUPP; 1559 break; 1560 } 1561 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1562 if (error) 1563 break; 1564 if (!legal_vif_num(i) && (i != -1)) { 1565 error = EINVAL; 1566 break; 1567 } 1568 imo->imo_multicast_vif = i; 1569 break; 1570 1571 case IP_MULTICAST_IF: 1572 /* 1573 * Select the interface for outgoing multicast packets. 1574 */ 1575 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1576 if (error) 1577 break; 1578 /* 1579 * INADDR_ANY is used to remove a previous selection. 1580 * When no interface is selected, a default one is 1581 * chosen every time a multicast packet is sent. 1582 */ 1583 if (addr.s_addr == INADDR_ANY) { 1584 imo->imo_multicast_ifp = NULL; 1585 break; 1586 } 1587 /* 1588 * The selected interface is identified by its local 1589 * IP address. Find the interface and confirm that 1590 * it supports multicasting. 1591 */ 1592 s = splimp(); 1593 ifp = ip_multicast_if(&addr, &ifindex); 1594 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1595 splx(s); 1596 error = EADDRNOTAVAIL; 1597 break; 1598 } 1599 imo->imo_multicast_ifp = ifp; 1600 if (ifindex) 1601 imo->imo_multicast_addr = addr; 1602 else 1603 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1604 splx(s); 1605 break; 1606 1607 case IP_MULTICAST_TTL: 1608 /* 1609 * Set the IP time-to-live for outgoing multicast packets. 1610 * The original multicast API required a char argument, 1611 * which is inconsistent with the rest of the socket API. 1612 * We allow either a char or an int. 1613 */ 1614 if (sopt->sopt_valsize == 1) { 1615 u_char ttl; 1616 error = sooptcopyin(sopt, &ttl, 1, 1); 1617 if (error) 1618 break; 1619 imo->imo_multicast_ttl = ttl; 1620 } else { 1621 u_int ttl; 1622 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1623 sizeof ttl); 1624 if (error) 1625 break; 1626 if (ttl > 255) 1627 error = EINVAL; 1628 else 1629 imo->imo_multicast_ttl = ttl; 1630 } 1631 break; 1632 1633 case IP_MULTICAST_LOOP: 1634 /* 1635 * Set the loopback flag for outgoing multicast packets. 1636 * Must be zero or one. The original multicast API required a 1637 * char argument, which is inconsistent with the rest 1638 * of the socket API. We allow either a char or an int. 1639 */ 1640 if (sopt->sopt_valsize == 1) { 1641 u_char loop; 1642 error = sooptcopyin(sopt, &loop, 1, 1); 1643 if (error) 1644 break; 1645 imo->imo_multicast_loop = !!loop; 1646 } else { 1647 u_int loop; 1648 error = sooptcopyin(sopt, &loop, sizeof loop, 1649 sizeof loop); 1650 if (error) 1651 break; 1652 imo->imo_multicast_loop = !!loop; 1653 } 1654 break; 1655 1656 case IP_ADD_MEMBERSHIP: 1657 /* 1658 * Add a multicast group membership. 1659 * Group must be a valid IP multicast address. 1660 */ 1661 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1662 if (error) 1663 break; 1664 1665 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1666 error = EINVAL; 1667 break; 1668 } 1669 s = splimp(); 1670 /* 1671 * If no interface address was provided, use the interface of 1672 * the route to the given multicast address. 1673 */ 1674 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1675 bzero((caddr_t)&ro, sizeof(ro)); 1676 dst = (struct sockaddr_in *)&ro.ro_dst; 1677 dst->sin_len = sizeof(*dst); 1678 dst->sin_family = AF_INET; 1679 dst->sin_addr = mreq.imr_multiaddr; 1680 rtalloc(&ro); 1681 if (ro.ro_rt == NULL) { 1682 error = EADDRNOTAVAIL; 1683 splx(s); 1684 break; 1685 } 1686 ifp = ro.ro_rt->rt_ifp; 1687 rtfree(ro.ro_rt); 1688 } 1689 else { 1690 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1691 } 1692 1693 /* 1694 * See if we found an interface, and confirm that it 1695 * supports multicast. 1696 */ 1697 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1698 error = EADDRNOTAVAIL; 1699 splx(s); 1700 break; 1701 } 1702 /* 1703 * See if the membership already exists or if all the 1704 * membership slots are full. 1705 */ 1706 for (i = 0; i < imo->imo_num_memberships; ++i) { 1707 if (imo->imo_membership[i]->inm_ifp == ifp && 1708 imo->imo_membership[i]->inm_addr.s_addr 1709 == mreq.imr_multiaddr.s_addr) 1710 break; 1711 } 1712 if (i < imo->imo_num_memberships) { 1713 error = EADDRINUSE; 1714 splx(s); 1715 break; 1716 } 1717 if (i == IP_MAX_MEMBERSHIPS) { 1718 error = ETOOMANYREFS; 1719 splx(s); 1720 break; 1721 } 1722 /* 1723 * Everything looks good; add a new record to the multicast 1724 * address list for the given interface. 1725 */ 1726 if ((imo->imo_membership[i] = 1727 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1728 error = ENOBUFS; 1729 splx(s); 1730 break; 1731 } 1732 ++imo->imo_num_memberships; 1733 splx(s); 1734 break; 1735 1736 case IP_DROP_MEMBERSHIP: 1737 /* 1738 * Drop a multicast group membership. 1739 * Group must be a valid IP multicast address. 1740 */ 1741 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1742 if (error) 1743 break; 1744 1745 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1746 error = EINVAL; 1747 break; 1748 } 1749 1750 s = splimp(); 1751 /* 1752 * If an interface address was specified, get a pointer 1753 * to its ifnet structure. 1754 */ 1755 if (mreq.imr_interface.s_addr == INADDR_ANY) 1756 ifp = NULL; 1757 else { 1758 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1759 if (ifp == NULL) { 1760 error = EADDRNOTAVAIL; 1761 splx(s); 1762 break; 1763 } 1764 } 1765 /* 1766 * Find the membership in the membership array. 1767 */ 1768 for (i = 0; i < imo->imo_num_memberships; ++i) { 1769 if ((ifp == NULL || 1770 imo->imo_membership[i]->inm_ifp == ifp) && 1771 imo->imo_membership[i]->inm_addr.s_addr == 1772 mreq.imr_multiaddr.s_addr) 1773 break; 1774 } 1775 if (i == imo->imo_num_memberships) { 1776 error = EADDRNOTAVAIL; 1777 splx(s); 1778 break; 1779 } 1780 /* 1781 * Give up the multicast address record to which the 1782 * membership points. 1783 */ 1784 in_delmulti(imo->imo_membership[i]); 1785 /* 1786 * Remove the gap in the membership array. 1787 */ 1788 for (++i; i < imo->imo_num_memberships; ++i) 1789 imo->imo_membership[i-1] = imo->imo_membership[i]; 1790 --imo->imo_num_memberships; 1791 splx(s); 1792 break; 1793 1794 default: 1795 error = EOPNOTSUPP; 1796 break; 1797 } 1798 1799 /* 1800 * If all options have default values, no need to keep the mbuf. 1801 */ 1802 if (imo->imo_multicast_ifp == NULL && 1803 imo->imo_multicast_vif == -1 && 1804 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1805 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1806 imo->imo_num_memberships == 0) { 1807 free(*imop, M_IPMOPTS); 1808 *imop = NULL; 1809 } 1810 1811 return (error); 1812} 1813 1814/* 1815 * Return the IP multicast options in response to user getsockopt(). 1816 */ 1817static int 1818ip_getmoptions(sopt, imo) 1819 struct sockopt *sopt; 1820 register struct ip_moptions *imo; 1821{ 1822 struct in_addr addr; 1823 struct in_ifaddr *ia; 1824 int error, optval; 1825 u_char coptval; 1826 1827 error = 0; 1828 switch (sopt->sopt_name) { 1829 case IP_MULTICAST_VIF: 1830 if (imo != NULL) 1831 optval = imo->imo_multicast_vif; 1832 else 1833 optval = -1; 1834 error = sooptcopyout(sopt, &optval, sizeof optval); 1835 break; 1836 1837 case IP_MULTICAST_IF: 1838 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1839 addr.s_addr = INADDR_ANY; 1840 else if (imo->imo_multicast_addr.s_addr) { 1841 /* return the value user has set */ 1842 addr = imo->imo_multicast_addr; 1843 } else { 1844 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1845 addr.s_addr = (ia == NULL) ? INADDR_ANY 1846 : IA_SIN(ia)->sin_addr.s_addr; 1847 } 1848 error = sooptcopyout(sopt, &addr, sizeof addr); 1849 break; 1850 1851 case IP_MULTICAST_TTL: 1852 if (imo == 0) 1853 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1854 else 1855 optval = coptval = imo->imo_multicast_ttl; 1856 if (sopt->sopt_valsize == 1) 1857 error = sooptcopyout(sopt, &coptval, 1); 1858 else 1859 error = sooptcopyout(sopt, &optval, sizeof optval); 1860 break; 1861 1862 case IP_MULTICAST_LOOP: 1863 if (imo == 0) 1864 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1865 else 1866 optval = coptval = imo->imo_multicast_loop; 1867 if (sopt->sopt_valsize == 1) 1868 error = sooptcopyout(sopt, &coptval, 1); 1869 else 1870 error = sooptcopyout(sopt, &optval, sizeof optval); 1871 break; 1872 1873 default: 1874 error = ENOPROTOOPT; 1875 break; 1876 } 1877 return (error); 1878} 1879 1880/* 1881 * Discard the IP multicast options. 1882 */ 1883void 1884ip_freemoptions(imo) 1885 register struct ip_moptions *imo; 1886{ 1887 register int i; 1888 1889 if (imo != NULL) { 1890 for (i = 0; i < imo->imo_num_memberships; ++i) 1891 in_delmulti(imo->imo_membership[i]); 1892 free(imo, M_IPMOPTS); 1893 } 1894} 1895 1896/* 1897 * Routine called from ip_output() to loop back a copy of an IP multicast 1898 * packet to the input queue of a specified interface. Note that this 1899 * calls the output routine of the loopback "driver", but with an interface 1900 * pointer that might NOT be a loopback interface -- evil, but easier than 1901 * replicating that code here. 1902 */ 1903static void 1904ip_mloopback(ifp, m, dst, hlen) 1905 struct ifnet *ifp; 1906 register struct mbuf *m; 1907 register struct sockaddr_in *dst; 1908 int hlen; 1909{ 1910 register struct ip *ip; 1911 struct mbuf *copym; 1912 1913 copym = m_copy(m, 0, M_COPYALL); 1914 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1915 copym = m_pullup(copym, hlen); 1916 if (copym != NULL) { 1917 /* 1918 * We don't bother to fragment if the IP length is greater 1919 * than the interface's MTU. Can this possibly matter? 1920 */ 1921 ip = mtod(copym, struct ip *); 1922 HTONS(ip->ip_len); 1923 HTONS(ip->ip_off); 1924 ip->ip_sum = 0; 1925 if (ip->ip_vhl == IP_VHL_BORING) { 1926 ip->ip_sum = in_cksum_hdr(ip); 1927 } else { 1928 ip->ip_sum = in_cksum(copym, hlen); 1929 } 1930 /* 1931 * NB: 1932 * It's not clear whether there are any lingering 1933 * reentrancy problems in other areas which might 1934 * be exposed by using ip_input directly (in 1935 * particular, everything which modifies the packet 1936 * in-place). Yet another option is using the 1937 * protosw directly to deliver the looped back 1938 * packet. For the moment, we'll err on the side 1939 * of safety by using if_simloop(). 1940 */ 1941#if 1 /* XXX */ 1942 if (dst->sin_family != AF_INET) { 1943 printf("ip_mloopback: bad address family %d\n", 1944 dst->sin_family); 1945 dst->sin_family = AF_INET; 1946 } 1947#endif 1948 1949#ifdef notdef 1950 copym->m_pkthdr.rcvif = ifp; 1951 ip_input(copym); 1952#else 1953 /* if the checksum hasn't been computed, mark it as valid */ 1954 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1955 copym->m_pkthdr.csum_flags |= 1956 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1957 copym->m_pkthdr.csum_data = 0xffff; 1958 } 1959 if_simloop(ifp, copym, dst->sin_family, 0); 1960#endif 1961 } 1962} 1963