ip_output.c revision 68150
117683Spst/* 239291Sfenner * Copyright (c) 1982, 1986, 1988, 1990, 1993 317683Spst * The Regents of the University of California. All rights reserved. 417683Spst * 517683Spst * Redistribution and use in source and binary forms, with or without 617683Spst * modification, are permitted provided that the following conditions 717683Spst * are met: 817683Spst * 1. Redistributions of source code must retain the above copyright 917683Spst * notice, this list of conditions and the following disclaimer. 1017683Spst * 2. Redistributions in binary form must reproduce the above copyright 1117683Spst * notice, this list of conditions and the following disclaimer in the 1217683Spst * documentation and/or other materials provided with the distribution. 1317683Spst * 3. All advertising materials mentioning features or use of this software 1417683Spst * must display the following acknowledgement: 1517683Spst * This product includes software developed by the University of 1617683Spst * California, Berkeley and its contributors. 1717683Spst * 4. Neither the name of the University nor the names of its contributors 1817683Spst * may be used to endorse or promote products derived from this software 1917683Spst * without specific prior written permission. 2017683Spst * 2117683Spst * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 2226175Sfenner * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 2339291Sfenner * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 2417683Spst * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 2517683Spst * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 2617683Spst * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 2717683Spst * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 2817683Spst * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 2917683Spst * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 3017683Spst * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 3117683Spst * SUCH DAMAGE. 3217683Spst * 3317683Spst * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 3417683Spst * $FreeBSD: head/sys/netinet/ip_output.c 68150 2000-11-01 01:59:28Z joe $ 3517683Spst */ 3617683Spst 3717683Spst#define _IP_VHL 3817683Spst 3917683Spst#include "opt_ipfw.h" 4017683Spst#include "opt_ipdn.h" 4117683Spst#include "opt_ipdivert.h" 4217683Spst#include "opt_ipfilter.h" 4317683Spst#include "opt_ipsec.h" 4417683Spst#include "opt_pfil_hooks.h" 4517683Spst 4617683Spst#include <sys/param.h> 4717683Spst#include <sys/systm.h> 4817683Spst#include <sys/kernel.h> 4917683Spst#include <sys/malloc.h> 5017683Spst#include <sys/mbuf.h> 5117683Spst#include <sys/protosw.h> 5217683Spst#include <sys/socket.h> 5317683Spst#include <sys/socketvar.h> 5417683Spst 5517683Spst#include <net/if.h> 5617683Spst#include <net/route.h> 5717683Spst 5817683Spst#include <netinet/in.h> 5917683Spst#include <netinet/in_systm.h> 6017683Spst#include <netinet/ip.h> 6117683Spst#include <netinet/in_pcb.h> 6217683Spst#include <netinet/in_var.h> 6317683Spst#include <netinet/ip_var.h> 6417683Spst 6517683Spst#include "faith.h" 6617683Spst 6717683Spst#ifdef vax 6817683Spst#include <machine/mtpr.h> 6917683Spst#endif 7017683Spst#include <machine/in_cksum.h> 7117683Spst 7217683Spststatic MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 7317683Spst 7417683Spst#ifdef IPSEC 7517683Spst#include <netinet6/ipsec.h> 7617683Spst#include <netkey/key.h> 7717683Spst#ifdef IPSEC_DEBUG 7817683Spst#include <netkey/key_debug.h> 7917683Spst#else 8017683Spst#define KEYDEBUG(lev,arg) 8117683Spst#endif 8217683Spst#endif /*IPSEC*/ 8317683Spst 8417683Spst#include <netinet/ip_fw.h> 8517683Spst 8617683Spst#ifdef DUMMYNET 8717683Spst#include <netinet/ip_dummynet.h> 8817683Spst#endif 8917683Spst 9017683Spst#ifdef IPFIREWALL_FORWARD_DEBUG 9117683Spst#define print_ip(a) printf("%ld.%ld.%ld.%ld",(ntohl(a.s_addr)>>24)&0xFF,\ 9217683Spst (ntohl(a.s_addr)>>16)&0xFF,\ 9317683Spst (ntohl(a.s_addr)>>8)&0xFF,\ 9417683Spst (ntohl(a.s_addr))&0xFF); 9517683Spst#endif 9617683Spst 9717683Spstu_short ip_id; 9817683Spst 9917683Spststatic struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *)); 10017683Spststatic void ip_mloopback 10117683Spst __P((struct ifnet *, struct mbuf *, struct sockaddr_in *, int)); 10217683Spststatic int ip_getmoptions 10317683Spst __P((struct sockopt *, struct ip_moptions *)); 10417683Spststatic int ip_pcbopts __P((int, struct mbuf **, struct mbuf *)); 10517683Spststatic int ip_setmoptions 10617683Spst __P((struct sockopt *, struct ip_moptions **)); 10717683Spst 10817683Spstint ip_optcopy __P((struct ip *, struct ip *)); 10917683Spst 11017683Spst 11117683Spstextern struct protosw inetsw[]; 11217683Spst 11317683Spst/* 11417683Spst * IP output. The packet in mbuf chain m contains a skeletal IP 11517683Spst * header (with len, off, ttl, proto, tos, src, dst). 11617683Spst * The mbuf chain containing the packet will be freed. 11717683Spst * The mbuf opt, if present, will not be freed. 11817683Spst */ 11917683Spstint 12017683Spstip_output(m0, opt, ro, flags, imo) 12117683Spst struct mbuf *m0; 12217683Spst struct mbuf *opt; 12317683Spst struct route *ro; 12417683Spst int flags; 12517683Spst struct ip_moptions *imo; 12617683Spst{ 12717683Spst struct ip *ip, *mhip; 12817683Spst struct ifnet *ifp; 12917683Spst struct mbuf *m = m0; 13017683Spst int hlen = sizeof (struct ip); 13117683Spst int len, off, error = 0; 13217683Spst struct sockaddr_in *dst; 13317683Spst struct in_ifaddr *ia; 13417683Spst int isbroadcast, sw_csum; 13517683Spst#ifdef IPSEC 13617683Spst struct route iproute; 13717683Spst struct socket *so = NULL; 13817683Spst struct secpolicy *sp = NULL; 13917683Spst#endif 14017683Spst u_int16_t divert_cookie; /* firewall cookie */ 14117683Spst#ifdef PFIL_HOOKS 14217683Spst struct packet_filter_hook *pfh; 14317683Spst struct mbuf *m1; 14417683Spst int rv; 14517683Spst#endif /* PFIL_HOOKS */ 14617683Spst#ifdef IPFIREWALL_FORWARD 14717683Spst int fwd_rewrite_src = 0; 14817683Spst#endif 14917683Spst struct ip_fw_chain *rule = NULL; 15017683Spst 15117683Spst#ifdef IPDIVERT 15217683Spst /* Get and reset firewall cookie */ 15317683Spst divert_cookie = ip_divert_cookie; 15417683Spst ip_divert_cookie = 0; 15517683Spst#else 15617683Spst divert_cookie = 0; 15717683Spst#endif 15817683Spst 15917683Spst#if defined(IPFIREWALL) && defined(DUMMYNET) 16017683Spst /* 16117683Spst * dummynet packet are prepended a vestigial mbuf with 16217683Spst * m_type = MT_DUMMYNET and m_data pointing to the matching 16317683Spst * rule. 16417683Spst */ 16517683Spst if (m->m_type == MT_DUMMYNET) { 16617683Spst /* 16717683Spst * the packet was already tagged, so part of the 16817683Spst * processing was already done, and we need to go down. 16917683Spst * Get parameters from the header. 17017683Spst */ 17117683Spst rule = (struct ip_fw_chain *)(m->m_data) ; 17217683Spst opt = NULL ; 17317683Spst ro = & ( ((struct dn_pkt *)m)->ro ) ; 17417683Spst imo = NULL ; 17517683Spst dst = ((struct dn_pkt *)m)->dn_dst ; 17617683Spst ifp = ((struct dn_pkt *)m)->ifp ; 17717683Spst flags = ((struct dn_pkt *)m)->flags ; 17817683Spst 17917683Spst m0 = m = m->m_next ; 18017683Spst#ifdef IPSEC 18117683Spst so = ipsec_getsocket(m); 18217683Spst ipsec_setsocket(m, NULL); 18317683Spst#endif 18417683Spst ip = mtod(m, struct ip *); 18517683Spst hlen = IP_VHL_HL(ip->ip_vhl) << 2 ; 18617683Spst goto sendit; 18739291Sfenner } else 18839291Sfenner rule = NULL ; 18939291Sfenner#endif 19039291Sfenner#ifdef IPSEC 19139291Sfenner so = ipsec_getsocket(m); 19239291Sfenner ipsec_setsocket(m, NULL); 19339291Sfenner#endif 19439291Sfenner 19517683Spst#ifdef DIAGNOSTIC 19617683Spst if ((m->m_flags & M_PKTHDR) == 0) 19717683Spst panic("ip_output no HDR"); 19817683Spst if (!ro) 19917683Spst panic("ip_output no route, proto = %d", 20017683Spst mtod(m, struct ip *)->ip_p); 20117683Spst#endif 20217683Spst if (opt) { 20317683Spst m = ip_insertoptions(m, opt, &len); 20417683Spst hlen = len; 20539291Sfenner } 20639291Sfenner ip = mtod(m, struct ip *); 20739291Sfenner /* 20839291Sfenner * Fill in IP header. 20939291Sfenner */ 21039291Sfenner if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 21139291Sfenner ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2); 21239291Sfenner ip->ip_off &= IP_DF; 21339291Sfenner ip->ip_id = htons(ip_id++); 21439291Sfenner ipstat.ips_localout++; 21539291Sfenner } else { 21639291Sfenner hlen = IP_VHL_HL(ip->ip_vhl) << 2; 21739291Sfenner } 21817683Spst 21917683Spst dst = (struct sockaddr_in *)&ro->ro_dst; 22017683Spst /* 22117683Spst * If there is a cached route, 22217683Spst * check that it is to the same destination 22317683Spst * and is still up. If not, free it and try again. 22417683Spst */ 22517683Spst if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 22617683Spst dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 22717683Spst RTFREE(ro->ro_rt); 22817683Spst ro->ro_rt = (struct rtentry *)0; 22917683Spst } 23017683Spst if (ro->ro_rt == 0) { 23117683Spst dst->sin_family = AF_INET; 23217683Spst dst->sin_len = sizeof(*dst); 23317683Spst dst->sin_addr = ip->ip_dst; 23417683Spst } 23517683Spst /* 23617683Spst * If routing to interface only, 23717683Spst * short circuit routing lookup. 23817683Spst */ 23917683Spst#define ifatoia(ifa) ((struct in_ifaddr *)(ifa)) 24017683Spst#define sintosa(sin) ((struct sockaddr *)(sin)) 24117683Spst if (flags & IP_ROUTETOIF) { 24217683Spst if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 24317683Spst (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 24417683Spst ipstat.ips_noroute++; 24517683Spst error = ENETUNREACH; 24617683Spst goto bad; 24717683Spst } 24817683Spst ifp = ia->ia_ifp; 24917683Spst ip->ip_ttl = 1; 25017683Spst isbroadcast = in_broadcast(dst->sin_addr, ifp); 25117683Spst } else { 25217683Spst /* 25317683Spst * If this is the case, we probably don't want to allocate 25417683Spst * a protocol-cloned route since we didn't get one from the 25517683Spst * ULP. This lets TCP do its thing, while not burdening 25617683Spst * forwarding or ICMP with the overhead of cloning a route. 25717683Spst * Of course, we still want to do any cloning requested by 25817683Spst * the link layer, as this is probably required in all cases 25917683Spst * for correct operation (as it is for ARP). 26017683Spst */ 26117683Spst if (ro->ro_rt == 0) 26217683Spst rtalloc_ign(ro, RTF_PRCLONING); 26317683Spst if (ro->ro_rt == 0) { 264 ipstat.ips_noroute++; 265 error = EHOSTUNREACH; 266 goto bad; 267 } 268 ia = ifatoia(ro->ro_rt->rt_ifa); 269 ifp = ro->ro_rt->rt_ifp; 270 ro->ro_rt->rt_use++; 271 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 272 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 273 if (ro->ro_rt->rt_flags & RTF_HOST) 274 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 275 else 276 isbroadcast = in_broadcast(dst->sin_addr, ifp); 277 } 278 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 279 struct in_multi *inm; 280 281 m->m_flags |= M_MCAST; 282 /* 283 * IP destination address is multicast. Make sure "dst" 284 * still points to the address in "ro". (It may have been 285 * changed to point to a gateway address, above.) 286 */ 287 dst = (struct sockaddr_in *)&ro->ro_dst; 288 /* 289 * See if the caller provided any multicast options 290 */ 291 if (imo != NULL) { 292 ip->ip_ttl = imo->imo_multicast_ttl; 293 if (imo->imo_multicast_ifp != NULL) 294 ifp = imo->imo_multicast_ifp; 295 if (imo->imo_multicast_vif != -1) 296 ip->ip_src.s_addr = 297 ip_mcast_src(imo->imo_multicast_vif); 298 } else 299 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 300 /* 301 * Confirm that the outgoing interface supports multicast. 302 */ 303 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 304 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 305 ipstat.ips_noroute++; 306 error = ENETUNREACH; 307 goto bad; 308 } 309 } 310 /* 311 * If source address not specified yet, use address 312 * of outgoing interface. 313 */ 314 if (ip->ip_src.s_addr == INADDR_ANY) { 315 register struct in_ifaddr *ia1; 316 317 for (ia1 = in_ifaddrhead.tqh_first; ia1; 318 ia1 = ia1->ia_link.tqe_next) 319 if (ia1->ia_ifp == ifp) { 320 ip->ip_src = IA_SIN(ia1)->sin_addr; 321 break; 322 } 323 } 324 325 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 326 if (inm != NULL && 327 (imo == NULL || imo->imo_multicast_loop)) { 328 /* 329 * If we belong to the destination multicast group 330 * on the outgoing interface, and the caller did not 331 * forbid loopback, loop back a copy. 332 */ 333 ip_mloopback(ifp, m, dst, hlen); 334 } 335 else { 336 /* 337 * If we are acting as a multicast router, perform 338 * multicast forwarding as if the packet had just 339 * arrived on the interface to which we are about 340 * to send. The multicast forwarding function 341 * recursively calls this function, using the 342 * IP_FORWARDING flag to prevent infinite recursion. 343 * 344 * Multicasts that are looped back by ip_mloopback(), 345 * above, will be forwarded by the ip_input() routine, 346 * if necessary. 347 */ 348 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 349 /* 350 * Check if rsvp daemon is running. If not, don't 351 * set ip_moptions. This ensures that the packet 352 * is multicast and not just sent down one link 353 * as prescribed by rsvpd. 354 */ 355 if (!rsvp_on) 356 imo = NULL; 357 if (ip_mforward(ip, ifp, m, imo) != 0) { 358 m_freem(m); 359 goto done; 360 } 361 } 362 } 363 364 /* 365 * Multicasts with a time-to-live of zero may be looped- 366 * back, above, but must not be transmitted on a network. 367 * Also, multicasts addressed to the loopback interface 368 * are not sent -- the above call to ip_mloopback() will 369 * loop back a copy if this host actually belongs to the 370 * destination group on the loopback interface. 371 */ 372 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 373 m_freem(m); 374 goto done; 375 } 376 377 goto sendit; 378 } 379#ifndef notdef 380 /* 381 * If source address not specified yet, use address 382 * of outgoing interface. 383 */ 384 if (ip->ip_src.s_addr == INADDR_ANY) { 385 ip->ip_src = IA_SIN(ia)->sin_addr; 386#ifdef IPFIREWALL_FORWARD 387 /* Keep note that we did this - if the firewall changes 388 * the next-hop, our interface may change, changing the 389 * default source IP. It's a shame so much effort happens 390 * twice. Oh well. 391 */ 392 fwd_rewrite_src++; 393#endif /* IPFIREWALL_FORWARD */ 394 } 395#endif /* notdef */ 396 /* 397 * Verify that we have any chance at all of being able to queue 398 * the packet or packet fragments 399 */ 400 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 401 ifp->if_snd.ifq_maxlen) { 402 error = ENOBUFS; 403 goto bad; 404 } 405 406 /* 407 * Look for broadcast address and 408 * and verify user is allowed to send 409 * such a packet. 410 */ 411 if (isbroadcast) { 412 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 413 error = EADDRNOTAVAIL; 414 goto bad; 415 } 416 if ((flags & IP_ALLOWBROADCAST) == 0) { 417 error = EACCES; 418 goto bad; 419 } 420 /* don't allow broadcast messages to be fragmented */ 421 if ((u_short)ip->ip_len > ifp->if_mtu) { 422 error = EMSGSIZE; 423 goto bad; 424 } 425 m->m_flags |= M_BCAST; 426 } else { 427 m->m_flags &= ~M_BCAST; 428 } 429 430sendit: 431 /* 432 * IpHack's section. 433 * - Xlate: translate packet's addr/port (NAT). 434 * - Firewall: deny/allow/etc. 435 * - Wrap: fake packet's addr/port <unimpl.> 436 * - Encapsulate: put it in another IP and send out. <unimp.> 437 */ 438#ifdef PFIL_HOOKS 439 /* 440 * Run through list of hooks for output packets. 441 */ 442 m1 = m; 443 pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 444 for (; pfh; pfh = pfh->pfil_link.tqe_next) 445 if (pfh->pfil_func) { 446 rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1); 447 if (rv) { 448 error = EHOSTUNREACH; 449 goto done; 450 } 451 m = m1; 452 if (m == NULL) 453 goto done; 454 ip = mtod(m, struct ip *); 455 } 456#endif /* PFIL_HOOKS */ 457 458 /* 459 * Check with the firewall... 460 */ 461 if (fw_enable && ip_fw_chk_ptr) { 462 struct sockaddr_in *old = dst; 463 464 off = (*ip_fw_chk_ptr)(&ip, 465 hlen, ifp, &divert_cookie, &m, &rule, &dst); 466 /* 467 * On return we must do the following: 468 * m == NULL -> drop the pkt 469 * 1<=off<= 0xffff -> DIVERT 470 * (off & 0x10000) -> send to a DUMMYNET pipe 471 * (off & 0x20000) -> TEE the packet 472 * dst != old -> IPFIREWALL_FORWARD 473 * off==0, dst==old -> accept 474 * If some of the above modules is not compiled in, then 475 * we should't have to check the corresponding condition 476 * (because the ipfw control socket should not accept 477 * unsupported rules), but better play safe and drop 478 * packets in case of doubt. 479 */ 480 if (!m) { /* firewall said to reject */ 481 error = EACCES; 482 goto done; 483 } 484 if (off == 0 && dst == old) /* common case */ 485 goto pass ; 486#ifdef DUMMYNET 487 if ((off & IP_FW_PORT_DYNT_FLAG) != 0) { 488 /* 489 * pass the pkt to dummynet. Need to include 490 * pipe number, m, ifp, ro, dst because these are 491 * not recomputed in the next pass. 492 * All other parameters have been already used and 493 * so they are not needed anymore. 494 * XXX note: if the ifp or ro entry are deleted 495 * while a pkt is in dummynet, we are in trouble! 496 */ 497 dummynet_io(off & 0xffff, DN_TO_IP_OUT, m,ifp,ro,dst,rule, 498 flags); 499 goto done; 500 } 501#endif 502#ifdef IPDIVERT 503 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 504 struct mbuf *clone = NULL; 505 506 /* Clone packet if we're doing a 'tee' */ 507 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 508 clone = m_dup(m, M_DONTWAIT); 509 510 /* 511 * XXX 512 * delayed checksums are not currently compatible 513 * with divert sockets. 514 */ 515 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 516 in_delayed_cksum(m); 517 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 518 } 519 520 /* Restore packet header fields to original values */ 521 HTONS(ip->ip_len); 522 HTONS(ip->ip_off); 523 524 /* Deliver packet to divert input routine */ 525 ip_divert_cookie = divert_cookie; 526 divert_packet(m, 0, off & 0xffff); 527 528 /* If 'tee', continue with original packet */ 529 if (clone != NULL) { 530 m = clone; 531 ip = mtod(m, struct ip *); 532 goto pass; 533 } 534 goto done; 535 } 536#endif 537 538#ifdef IPFIREWALL_FORWARD 539 /* Here we check dst to make sure it's directly reachable on the 540 * interface we previously thought it was. 541 * If it isn't (which may be likely in some situations) we have 542 * to re-route it (ie, find a route for the next-hop and the 543 * associated interface) and set them here. This is nested 544 * forwarding which in most cases is undesirable, except where 545 * such control is nigh impossible. So we do it here. 546 * And I'm babbling. 547 */ 548 if (off == 0 && old != dst) { 549 struct in_ifaddr *ia; 550 551 /* It's changed... */ 552 /* There must be a better way to do this next line... */ 553 static struct route sro_fwd, *ro_fwd = &sro_fwd; 554#ifdef IPFIREWALL_FORWARD_DEBUG 555 printf("IPFIREWALL_FORWARD: New dst ip: "); 556 print_ip(dst->sin_addr); 557 printf("\n"); 558#endif 559 /* 560 * We need to figure out if we have been forwarded 561 * to a local socket. If so then we should somehow 562 * "loop back" to ip_input, and get directed to the 563 * PCB as if we had received this packet. This is 564 * because it may be dificult to identify the packets 565 * you want to forward until they are being output 566 * and have selected an interface. (e.g. locally 567 * initiated packets) If we used the loopback inteface, 568 * we would not be able to control what happens 569 * as the packet runs through ip_input() as 570 * it is done through a ISR. 571 */ 572 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 573 ia = TAILQ_NEXT(ia, ia_link)) { 574 /* 575 * If the addr to forward to is one 576 * of ours, we pretend to 577 * be the destination for this packet. 578 */ 579 if (IA_SIN(ia)->sin_addr.s_addr == 580 dst->sin_addr.s_addr) 581 break; 582 } 583 if (ia) { 584 /* tell ip_input "dont filter" */ 585 ip_fw_fwd_addr = dst; 586 if (m->m_pkthdr.rcvif == NULL) 587 m->m_pkthdr.rcvif = ifunit("lo0"); 588 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 589 m->m_pkthdr.csum_flags |= 590 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 591 m0->m_pkthdr.csum_data = 0xffff; 592 } 593 m->m_pkthdr.csum_flags |= 594 CSUM_IP_CHECKED | CSUM_IP_VALID; 595 HTONS(ip->ip_len); 596 HTONS(ip->ip_off); 597 ip_input(m); 598 goto done; 599 } 600 /* Some of the logic for this was 601 * nicked from above. 602 * 603 * This rewrites the cached route in a local PCB. 604 * Is this what we want to do? 605 */ 606 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 607 608 ro_fwd->ro_rt = 0; 609 rtalloc_ign(ro_fwd, RTF_PRCLONING); 610 611 if (ro_fwd->ro_rt == 0) { 612 ipstat.ips_noroute++; 613 error = EHOSTUNREACH; 614 goto bad; 615 } 616 617 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 618 ifp = ro_fwd->ro_rt->rt_ifp; 619 ro_fwd->ro_rt->rt_use++; 620 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 621 dst = (struct sockaddr_in *)ro_fwd->ro_rt->rt_gateway; 622 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 623 isbroadcast = 624 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 625 else 626 isbroadcast = in_broadcast(dst->sin_addr, ifp); 627 RTFREE(ro->ro_rt); 628 ro->ro_rt = ro_fwd->ro_rt; 629 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 630 631 /* 632 * If we added a default src ip earlier, 633 * which would have been gotten from the-then 634 * interface, do it again, from the new one. 635 */ 636 if (fwd_rewrite_src) 637 ip->ip_src = IA_SIN(ia)->sin_addr; 638 goto pass ; 639 } 640#endif /* IPFIREWALL_FORWARD */ 641 /* 642 * if we get here, none of the above matches, and 643 * we have to drop the pkt 644 */ 645 m_freem(m); 646 error = EACCES; /* not sure this is the right error msg */ 647 goto done; 648 } 649 650pass: 651#ifdef IPSEC 652 /* get SP for this packet */ 653 if (so == NULL) 654 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 655 else 656 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 657 658 if (sp == NULL) { 659 ipsecstat.out_inval++; 660 goto bad; 661 } 662 663 error = 0; 664 665 /* check policy */ 666 switch (sp->policy) { 667 case IPSEC_POLICY_DISCARD: 668 /* 669 * This packet is just discarded. 670 */ 671 ipsecstat.out_polvio++; 672 goto bad; 673 674 case IPSEC_POLICY_BYPASS: 675 case IPSEC_POLICY_NONE: 676 /* no need to do IPsec. */ 677 goto skip_ipsec; 678 679 case IPSEC_POLICY_IPSEC: 680 if (sp->req == NULL) { 681 /* XXX should be panic ? */ 682 printf("ip_output: No IPsec request specified.\n"); 683 error = EINVAL; 684 goto bad; 685 } 686 break; 687 688 case IPSEC_POLICY_ENTRUST: 689 default: 690 printf("ip_output: Invalid policy found. %d\n", sp->policy); 691 } 692 { 693 struct ipsec_output_state state; 694 bzero(&state, sizeof(state)); 695 state.m = m; 696 if (flags & IP_ROUTETOIF) { 697 state.ro = &iproute; 698 bzero(&iproute, sizeof(iproute)); 699 } else 700 state.ro = ro; 701 state.dst = (struct sockaddr *)dst; 702 703 ip->ip_sum = 0; 704 705 /* 706 * XXX 707 * delayed checksums are not currently compatible with IPsec 708 */ 709 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 710 in_delayed_cksum(m); 711 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 712 } 713 714 HTONS(ip->ip_len); 715 HTONS(ip->ip_off); 716 717 error = ipsec4_output(&state, sp, flags); 718 719 m = state.m; 720 if (flags & IP_ROUTETOIF) { 721 /* 722 * if we have tunnel mode SA, we may need to ignore 723 * IP_ROUTETOIF. 724 */ 725 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 726 flags &= ~IP_ROUTETOIF; 727 ro = state.ro; 728 } 729 } else 730 ro = state.ro; 731 dst = (struct sockaddr_in *)state.dst; 732 if (error) { 733 /* mbuf is already reclaimed in ipsec4_output. */ 734 m0 = NULL; 735 switch (error) { 736 case EHOSTUNREACH: 737 case ENETUNREACH: 738 case EMSGSIZE: 739 case ENOBUFS: 740 case ENOMEM: 741 break; 742 default: 743 printf("ip4_output (ipsec): error code %d\n", error); 744 /*fall through*/ 745 case ENOENT: 746 /* don't show these error codes to the user */ 747 error = 0; 748 break; 749 } 750 goto bad; 751 } 752 } 753 754 /* be sure to update variables that are affected by ipsec4_output() */ 755 ip = mtod(m, struct ip *); 756#ifdef _IP_VHL 757 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 758#else 759 hlen = ip->ip_hl << 2; 760#endif 761 if (ro->ro_rt == NULL) { 762 if ((flags & IP_ROUTETOIF) == 0) { 763 printf("ip_output: " 764 "can't update route after IPsec processing\n"); 765 error = EHOSTUNREACH; /*XXX*/ 766 goto bad; 767 } 768 } else { 769 ia = ifatoia(ro->ro_rt->rt_ifa); 770 ifp = ro->ro_rt->rt_ifp; 771 } 772 773 /* make it flipped, again. */ 774 NTOHS(ip->ip_len); 775 NTOHS(ip->ip_off); 776skip_ipsec: 777#endif /*IPSEC*/ 778 779 sw_csum = m->m_pkthdr.csum_flags | CSUM_IP; 780 m->m_pkthdr.csum_flags = sw_csum & ifp->if_hwassist; 781 sw_csum &= ~ifp->if_hwassist; 782 if (sw_csum & CSUM_DELAY_DATA) { 783 in_delayed_cksum(m); 784 sw_csum &= ~CSUM_DELAY_DATA; 785 } 786 787 /* 788 * If small enough for interface, or the interface will take 789 * care of the fragmentation for us, can just send directly. 790 */ 791 if ((u_short)ip->ip_len <= ifp->if_mtu || 792 ifp->if_hwassist & CSUM_FRAGMENT) { 793 HTONS(ip->ip_len); 794 HTONS(ip->ip_off); 795 ip->ip_sum = 0; 796 if (sw_csum & CSUM_DELAY_IP) { 797 if (ip->ip_vhl == IP_VHL_BORING) { 798 ip->ip_sum = in_cksum_hdr(ip); 799 } else { 800 ip->ip_sum = in_cksum(m, hlen); 801 } 802 } 803 804 /* Record statistics for this interface address. */ 805 if (!(flags & IP_FORWARDING)) { 806 ia->ia_ifa.if_opackets++; 807 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 808 } 809 810 error = (*ifp->if_output)(ifp, m, 811 (struct sockaddr *)dst, ro->ro_rt); 812 goto done; 813 } 814 /* 815 * Too large for interface; fragment if possible. 816 * Must be able to put at least 8 bytes per fragment. 817 */ 818 if (ip->ip_off & IP_DF) { 819 error = EMSGSIZE; 820 /* 821 * This case can happen if the user changed the MTU 822 * of an interface after enabling IP on it. Because 823 * most netifs don't keep track of routes pointing to 824 * them, there is no way for one to update all its 825 * routes when the MTU is changed. 826 */ 827 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 828 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 829 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 830 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 831 } 832 ipstat.ips_cantfrag++; 833 goto bad; 834 } 835 len = (ifp->if_mtu - hlen) &~ 7; 836 if (len < 8) { 837 error = EMSGSIZE; 838 goto bad; 839 } 840 841 /* 842 * if the interface will not calculate checksums on 843 * fragmented packets, then do it here. 844 */ 845 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 846 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 847 in_delayed_cksum(m); 848 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 849 } 850 851 { 852 int mhlen, firstlen = len; 853 struct mbuf **mnext = &m->m_nextpkt; 854 int nfrags = 1; 855 856 /* 857 * Loop through length of segment after first fragment, 858 * make new header and copy data of each part and link onto chain. 859 */ 860 m0 = m; 861 mhlen = sizeof (struct ip); 862 for (off = hlen + len; off < (u_short)ip->ip_len; off += len) { 863 MGETHDR(m, M_DONTWAIT, MT_HEADER); 864 if (m == 0) { 865 error = ENOBUFS; 866 ipstat.ips_odropped++; 867 goto sendorfree; 868 } 869 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 870 m->m_data += max_linkhdr; 871 mhip = mtod(m, struct ip *); 872 *mhip = *ip; 873 if (hlen > sizeof (struct ip)) { 874 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 875 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2); 876 } 877 m->m_len = mhlen; 878 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 879 if (off + len >= (u_short)ip->ip_len) 880 len = (u_short)ip->ip_len - off; 881 else 882 mhip->ip_off |= IP_MF; 883 mhip->ip_len = htons((u_short)(len + mhlen)); 884 m->m_next = m_copy(m0, off, len); 885 if (m->m_next == 0) { 886 (void) m_free(m); 887 error = ENOBUFS; /* ??? */ 888 ipstat.ips_odropped++; 889 goto sendorfree; 890 } 891 m->m_pkthdr.len = mhlen + len; 892 m->m_pkthdr.rcvif = (struct ifnet *)0; 893 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 894 HTONS(mhip->ip_off); 895 mhip->ip_sum = 0; 896 if (sw_csum & CSUM_DELAY_IP) { 897 if (mhip->ip_vhl == IP_VHL_BORING) { 898 mhip->ip_sum = in_cksum_hdr(mhip); 899 } else { 900 mhip->ip_sum = in_cksum(m, mhlen); 901 } 902 } 903 *mnext = m; 904 mnext = &m->m_nextpkt; 905 nfrags++; 906 } 907 ipstat.ips_ofragments += nfrags; 908 909 /* set first/last markers for fragment chain */ 910 m->m_flags |= M_LASTFRAG; 911 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 912 m0->m_pkthdr.csum_data = nfrags; 913 914 /* 915 * Update first fragment by trimming what's been copied out 916 * and updating header, then send each fragment (in order). 917 */ 918 m = m0; 919 m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 920 m->m_pkthdr.len = hlen + firstlen; 921 ip->ip_len = htons((u_short)m->m_pkthdr.len); 922 ip->ip_off |= IP_MF; 923 HTONS(ip->ip_off); 924 ip->ip_sum = 0; 925 if (sw_csum & CSUM_DELAY_IP) { 926 if (ip->ip_vhl == IP_VHL_BORING) { 927 ip->ip_sum = in_cksum_hdr(ip); 928 } else { 929 ip->ip_sum = in_cksum(m, hlen); 930 } 931 } 932sendorfree: 933 for (m = m0; m; m = m0) { 934 m0 = m->m_nextpkt; 935 m->m_nextpkt = 0; 936 if (error == 0) { 937 /* Record statistics for this interface address. */ 938 ia->ia_ifa.if_opackets++; 939 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 940 941 error = (*ifp->if_output)(ifp, m, 942 (struct sockaddr *)dst, ro->ro_rt); 943 } else 944 m_freem(m); 945 } 946 947 if (error == 0) 948 ipstat.ips_fragmented++; 949 } 950done: 951#ifdef IPSEC 952 if (ro == &iproute && ro->ro_rt) { 953 RTFREE(ro->ro_rt); 954 ro->ro_rt = NULL; 955 } 956 if (sp != NULL) { 957 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 958 printf("DP ip_output call free SP:%p\n", sp)); 959 key_freesp(sp); 960 } 961#endif /* IPSEC */ 962 return (error); 963bad: 964 m_freem(m0); 965 goto done; 966} 967 968void 969in_delayed_cksum(struct mbuf *m) 970{ 971 struct ip *ip; 972 u_short csum, offset; 973 974 ip = mtod(m, struct ip *); 975 offset = IP_VHL_HL(ip->ip_vhl) << 2 ; 976 csum = in_cksum_skip(m, ip->ip_len, offset); 977 offset += m->m_pkthdr.csum_data; /* checksum offset */ 978 979 if (offset + sizeof(u_short) > m->m_len) { 980 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 981 m->m_len, offset, ip->ip_p); 982 /* 983 * XXX 984 * this shouldn't happen, but if it does, the 985 * correct behavior may be to insert the checksum 986 * in the existing chain instead of rearranging it. 987 */ 988 m = m_pullup(m, offset + sizeof(u_short)); 989 } 990 *(u_short *)(m->m_data + offset) = csum; 991} 992 993/* 994 * Insert IP options into preformed packet. 995 * Adjust IP destination as required for IP source routing, 996 * as indicated by a non-zero in_addr at the start of the options. 997 * 998 * XXX This routine assumes that the packet has no options in place. 999 */ 1000static struct mbuf * 1001ip_insertoptions(m, opt, phlen) 1002 register struct mbuf *m; 1003 struct mbuf *opt; 1004 int *phlen; 1005{ 1006 register struct ipoption *p = mtod(opt, struct ipoption *); 1007 struct mbuf *n; 1008 register struct ip *ip = mtod(m, struct ip *); 1009 unsigned optlen; 1010 1011 optlen = opt->m_len - sizeof(p->ipopt_dst); 1012 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) 1013 return (m); /* XXX should fail */ 1014 if (p->ipopt_dst.s_addr) 1015 ip->ip_dst = p->ipopt_dst; 1016 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1017 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1018 if (n == 0) 1019 return (m); 1020 n->m_pkthdr.rcvif = (struct ifnet *)0; 1021 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1022 m->m_len -= sizeof(struct ip); 1023 m->m_data += sizeof(struct ip); 1024 n->m_next = m; 1025 m = n; 1026 m->m_len = optlen + sizeof(struct ip); 1027 m->m_data += max_linkhdr; 1028 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1029 } else { 1030 m->m_data -= optlen; 1031 m->m_len += optlen; 1032 m->m_pkthdr.len += optlen; 1033 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1034 } 1035 ip = mtod(m, struct ip *); 1036 bcopy(p->ipopt_list, ip + 1, optlen); 1037 *phlen = sizeof(struct ip) + optlen; 1038 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2); 1039 ip->ip_len += optlen; 1040 return (m); 1041} 1042 1043/* 1044 * Copy options from ip to jp, 1045 * omitting those not copied during fragmentation. 1046 */ 1047int 1048ip_optcopy(ip, jp) 1049 struct ip *ip, *jp; 1050{ 1051 register u_char *cp, *dp; 1052 int opt, optlen, cnt; 1053 1054 cp = (u_char *)(ip + 1); 1055 dp = (u_char *)(jp + 1); 1056 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1057 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1058 opt = cp[0]; 1059 if (opt == IPOPT_EOL) 1060 break; 1061 if (opt == IPOPT_NOP) { 1062 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1063 *dp++ = IPOPT_NOP; 1064 optlen = 1; 1065 continue; 1066 } 1067#ifdef DIAGNOSTIC 1068 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1069 panic("malformed IPv4 option passed to ip_optcopy"); 1070#endif 1071 optlen = cp[IPOPT_OLEN]; 1072#ifdef DIAGNOSTIC 1073 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1074 panic("malformed IPv4 option passed to ip_optcopy"); 1075#endif 1076 /* bogus lengths should have been caught by ip_dooptions */ 1077 if (optlen > cnt) 1078 optlen = cnt; 1079 if (IPOPT_COPIED(opt)) { 1080 bcopy(cp, dp, optlen); 1081 dp += optlen; 1082 } 1083 } 1084 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1085 *dp++ = IPOPT_EOL; 1086 return (optlen); 1087} 1088 1089/* 1090 * IP socket option processing. 1091 */ 1092int 1093ip_ctloutput(so, sopt) 1094 struct socket *so; 1095 struct sockopt *sopt; 1096{ 1097 struct inpcb *inp = sotoinpcb(so); 1098 int error, optval; 1099 1100 error = optval = 0; 1101 if (sopt->sopt_level != IPPROTO_IP) { 1102 return (EINVAL); 1103 } 1104 1105 switch (sopt->sopt_dir) { 1106 case SOPT_SET: 1107 switch (sopt->sopt_name) { 1108 case IP_OPTIONS: 1109#ifdef notyet 1110 case IP_RETOPTS: 1111#endif 1112 { 1113 struct mbuf *m; 1114 if (sopt->sopt_valsize > MLEN) { 1115 error = EMSGSIZE; 1116 break; 1117 } 1118 MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_HEADER); 1119 if (m == 0) { 1120 error = ENOBUFS; 1121 break; 1122 } 1123 m->m_len = sopt->sopt_valsize; 1124 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1125 m->m_len); 1126 1127 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1128 m)); 1129 } 1130 1131 case IP_TOS: 1132 case IP_TTL: 1133 case IP_RECVOPTS: 1134 case IP_RECVRETOPTS: 1135 case IP_RECVDSTADDR: 1136 case IP_RECVIF: 1137#if defined(NFAITH) && NFAITH > 0 1138 case IP_FAITH: 1139#endif 1140 error = sooptcopyin(sopt, &optval, sizeof optval, 1141 sizeof optval); 1142 if (error) 1143 break; 1144 1145 switch (sopt->sopt_name) { 1146 case IP_TOS: 1147 inp->inp_ip_tos = optval; 1148 break; 1149 1150 case IP_TTL: 1151 inp->inp_ip_ttl = optval; 1152 break; 1153#define OPTSET(bit) \ 1154 if (optval) \ 1155 inp->inp_flags |= bit; \ 1156 else \ 1157 inp->inp_flags &= ~bit; 1158 1159 case IP_RECVOPTS: 1160 OPTSET(INP_RECVOPTS); 1161 break; 1162 1163 case IP_RECVRETOPTS: 1164 OPTSET(INP_RECVRETOPTS); 1165 break; 1166 1167 case IP_RECVDSTADDR: 1168 OPTSET(INP_RECVDSTADDR); 1169 break; 1170 1171 case IP_RECVIF: 1172 OPTSET(INP_RECVIF); 1173 break; 1174 1175#if defined(NFAITH) && NFAITH > 0 1176 case IP_FAITH: 1177 OPTSET(INP_FAITH); 1178 break; 1179#endif 1180 } 1181 break; 1182#undef OPTSET 1183 1184 case IP_MULTICAST_IF: 1185 case IP_MULTICAST_VIF: 1186 case IP_MULTICAST_TTL: 1187 case IP_MULTICAST_LOOP: 1188 case IP_ADD_MEMBERSHIP: 1189 case IP_DROP_MEMBERSHIP: 1190 error = ip_setmoptions(sopt, &inp->inp_moptions); 1191 break; 1192 1193 case IP_PORTRANGE: 1194 error = sooptcopyin(sopt, &optval, sizeof optval, 1195 sizeof optval); 1196 if (error) 1197 break; 1198 1199 switch (optval) { 1200 case IP_PORTRANGE_DEFAULT: 1201 inp->inp_flags &= ~(INP_LOWPORT); 1202 inp->inp_flags &= ~(INP_HIGHPORT); 1203 break; 1204 1205 case IP_PORTRANGE_HIGH: 1206 inp->inp_flags &= ~(INP_LOWPORT); 1207 inp->inp_flags |= INP_HIGHPORT; 1208 break; 1209 1210 case IP_PORTRANGE_LOW: 1211 inp->inp_flags &= ~(INP_HIGHPORT); 1212 inp->inp_flags |= INP_LOWPORT; 1213 break; 1214 1215 default: 1216 error = EINVAL; 1217 break; 1218 } 1219 break; 1220 1221#ifdef IPSEC 1222 case IP_IPSEC_POLICY: 1223 { 1224 caddr_t req; 1225 size_t len = 0; 1226 int priv; 1227 struct mbuf *m; 1228 int optname; 1229 1230 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1231 break; 1232 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1233 break; 1234 priv = (sopt->sopt_p != NULL && 1235 suser(sopt->sopt_p) != 0) ? 0 : 1; 1236 req = mtod(m, caddr_t); 1237 len = m->m_len; 1238 optname = sopt->sopt_name; 1239 error = ipsec4_set_policy(inp, optname, req, len, priv); 1240 m_freem(m); 1241 break; 1242 } 1243#endif /*IPSEC*/ 1244 1245 default: 1246 error = ENOPROTOOPT; 1247 break; 1248 } 1249 break; 1250 1251 case SOPT_GET: 1252 switch (sopt->sopt_name) { 1253 case IP_OPTIONS: 1254 case IP_RETOPTS: 1255 if (inp->inp_options) 1256 error = sooptcopyout(sopt, 1257 mtod(inp->inp_options, 1258 char *), 1259 inp->inp_options->m_len); 1260 else 1261 sopt->sopt_valsize = 0; 1262 break; 1263 1264 case IP_TOS: 1265 case IP_TTL: 1266 case IP_RECVOPTS: 1267 case IP_RECVRETOPTS: 1268 case IP_RECVDSTADDR: 1269 case IP_RECVIF: 1270 case IP_PORTRANGE: 1271#if defined(NFAITH) && NFAITH > 0 1272 case IP_FAITH: 1273#endif 1274 switch (sopt->sopt_name) { 1275 1276 case IP_TOS: 1277 optval = inp->inp_ip_tos; 1278 break; 1279 1280 case IP_TTL: 1281 optval = inp->inp_ip_ttl; 1282 break; 1283 1284#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1285 1286 case IP_RECVOPTS: 1287 optval = OPTBIT(INP_RECVOPTS); 1288 break; 1289 1290 case IP_RECVRETOPTS: 1291 optval = OPTBIT(INP_RECVRETOPTS); 1292 break; 1293 1294 case IP_RECVDSTADDR: 1295 optval = OPTBIT(INP_RECVDSTADDR); 1296 break; 1297 1298 case IP_RECVIF: 1299 optval = OPTBIT(INP_RECVIF); 1300 break; 1301 1302 case IP_PORTRANGE: 1303 if (inp->inp_flags & INP_HIGHPORT) 1304 optval = IP_PORTRANGE_HIGH; 1305 else if (inp->inp_flags & INP_LOWPORT) 1306 optval = IP_PORTRANGE_LOW; 1307 else 1308 optval = 0; 1309 break; 1310 1311#if defined(NFAITH) && NFAITH > 0 1312 case IP_FAITH: 1313 optval = OPTBIT(INP_FAITH); 1314 break; 1315#endif 1316 } 1317 error = sooptcopyout(sopt, &optval, sizeof optval); 1318 break; 1319 1320 case IP_MULTICAST_IF: 1321 case IP_MULTICAST_VIF: 1322 case IP_MULTICAST_TTL: 1323 case IP_MULTICAST_LOOP: 1324 case IP_ADD_MEMBERSHIP: 1325 case IP_DROP_MEMBERSHIP: 1326 error = ip_getmoptions(sopt, inp->inp_moptions); 1327 break; 1328 1329#ifdef IPSEC 1330 case IP_IPSEC_POLICY: 1331 { 1332 struct mbuf *m = NULL; 1333 caddr_t req = NULL; 1334 size_t len = 0; 1335 1336 if (m != 0) { 1337 req = mtod(m, caddr_t); 1338 len = m->m_len; 1339 } 1340 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1341 if (error == 0) 1342 error = soopt_mcopyout(sopt, m); /* XXX */ 1343 if (error == 0) 1344 m_freem(m); 1345 break; 1346 } 1347#endif /*IPSEC*/ 1348 1349 default: 1350 error = ENOPROTOOPT; 1351 break; 1352 } 1353 break; 1354 } 1355 return (error); 1356} 1357 1358/* 1359 * Set up IP options in pcb for insertion in output packets. 1360 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1361 * with destination address if source routed. 1362 */ 1363static int 1364ip_pcbopts(optname, pcbopt, m) 1365 int optname; 1366 struct mbuf **pcbopt; 1367 register struct mbuf *m; 1368{ 1369 register int cnt, optlen; 1370 register u_char *cp; 1371 u_char opt; 1372 1373 /* turn off any old options */ 1374 if (*pcbopt) 1375 (void)m_free(*pcbopt); 1376 *pcbopt = 0; 1377 if (m == (struct mbuf *)0 || m->m_len == 0) { 1378 /* 1379 * Only turning off any previous options. 1380 */ 1381 if (m) 1382 (void)m_free(m); 1383 return (0); 1384 } 1385 1386#ifndef vax 1387 if (m->m_len % sizeof(int32_t)) 1388 goto bad; 1389#endif 1390 /* 1391 * IP first-hop destination address will be stored before 1392 * actual options; move other options back 1393 * and clear it when none present. 1394 */ 1395 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1396 goto bad; 1397 cnt = m->m_len; 1398 m->m_len += sizeof(struct in_addr); 1399 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1400 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1401 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1402 1403 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1404 opt = cp[IPOPT_OPTVAL]; 1405 if (opt == IPOPT_EOL) 1406 break; 1407 if (opt == IPOPT_NOP) 1408 optlen = 1; 1409 else { 1410 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1411 goto bad; 1412 optlen = cp[IPOPT_OLEN]; 1413 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1414 goto bad; 1415 } 1416 switch (opt) { 1417 1418 default: 1419 break; 1420 1421 case IPOPT_LSRR: 1422 case IPOPT_SSRR: 1423 /* 1424 * user process specifies route as: 1425 * ->A->B->C->D 1426 * D must be our final destination (but we can't 1427 * check that since we may not have connected yet). 1428 * A is first hop destination, which doesn't appear in 1429 * actual IP option, but is stored before the options. 1430 */ 1431 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1432 goto bad; 1433 m->m_len -= sizeof(struct in_addr); 1434 cnt -= sizeof(struct in_addr); 1435 optlen -= sizeof(struct in_addr); 1436 cp[IPOPT_OLEN] = optlen; 1437 /* 1438 * Move first hop before start of options. 1439 */ 1440 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1441 sizeof(struct in_addr)); 1442 /* 1443 * Then copy rest of options back 1444 * to close up the deleted entry. 1445 */ 1446 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1447 sizeof(struct in_addr)), 1448 (caddr_t)&cp[IPOPT_OFFSET+1], 1449 (unsigned)cnt + sizeof(struct in_addr)); 1450 break; 1451 } 1452 } 1453 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1454 goto bad; 1455 *pcbopt = m; 1456 return (0); 1457 1458bad: 1459 (void)m_free(m); 1460 return (EINVAL); 1461} 1462 1463/* 1464 * XXX 1465 * The whole multicast option thing needs to be re-thought. 1466 * Several of these options are equally applicable to non-multicast 1467 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1468 * standard option (IP_TTL). 1469 */ 1470/* 1471 * Set the IP multicast options in response to user setsockopt(). 1472 */ 1473static int 1474ip_setmoptions(sopt, imop) 1475 struct sockopt *sopt; 1476 struct ip_moptions **imop; 1477{ 1478 int error = 0; 1479 int i; 1480 struct in_addr addr; 1481 struct ip_mreq mreq; 1482 struct ifnet *ifp; 1483 struct ip_moptions *imo = *imop; 1484 struct route ro; 1485 struct sockaddr_in *dst; 1486 int s; 1487 1488 if (imo == NULL) { 1489 /* 1490 * No multicast option buffer attached to the pcb; 1491 * allocate one and initialize to default values. 1492 */ 1493 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1494 M_WAITOK); 1495 1496 if (imo == NULL) 1497 return (ENOBUFS); 1498 *imop = imo; 1499 imo->imo_multicast_ifp = NULL; 1500 imo->imo_multicast_vif = -1; 1501 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1502 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1503 imo->imo_num_memberships = 0; 1504 } 1505 1506 switch (sopt->sopt_name) { 1507 /* store an index number for the vif you wanna use in the send */ 1508 case IP_MULTICAST_VIF: 1509 if (legal_vif_num == 0) { 1510 error = EOPNOTSUPP; 1511 break; 1512 } 1513 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1514 if (error) 1515 break; 1516 if (!legal_vif_num(i) && (i != -1)) { 1517 error = EINVAL; 1518 break; 1519 } 1520 imo->imo_multicast_vif = i; 1521 break; 1522 1523 case IP_MULTICAST_IF: 1524 /* 1525 * Select the interface for outgoing multicast packets. 1526 */ 1527 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1528 if (error) 1529 break; 1530 /* 1531 * INADDR_ANY is used to remove a previous selection. 1532 * When no interface is selected, a default one is 1533 * chosen every time a multicast packet is sent. 1534 */ 1535 if (addr.s_addr == INADDR_ANY) { 1536 imo->imo_multicast_ifp = NULL; 1537 break; 1538 } 1539 /* 1540 * The selected interface is identified by its local 1541 * IP address. Find the interface and confirm that 1542 * it supports multicasting. 1543 */ 1544 s = splimp(); 1545 INADDR_TO_IFP(addr, ifp); 1546 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1547 splx(s); 1548 error = EADDRNOTAVAIL; 1549 break; 1550 } 1551 imo->imo_multicast_ifp = ifp; 1552 splx(s); 1553 break; 1554 1555 case IP_MULTICAST_TTL: 1556 /* 1557 * Set the IP time-to-live for outgoing multicast packets. 1558 * The original multicast API required a char argument, 1559 * which is inconsistent with the rest of the socket API. 1560 * We allow either a char or an int. 1561 */ 1562 if (sopt->sopt_valsize == 1) { 1563 u_char ttl; 1564 error = sooptcopyin(sopt, &ttl, 1, 1); 1565 if (error) 1566 break; 1567 imo->imo_multicast_ttl = ttl; 1568 } else { 1569 u_int ttl; 1570 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1571 sizeof ttl); 1572 if (error) 1573 break; 1574 if (ttl > 255) 1575 error = EINVAL; 1576 else 1577 imo->imo_multicast_ttl = ttl; 1578 } 1579 break; 1580 1581 case IP_MULTICAST_LOOP: 1582 /* 1583 * Set the loopback flag for outgoing multicast packets. 1584 * Must be zero or one. The original multicast API required a 1585 * char argument, which is inconsistent with the rest 1586 * of the socket API. We allow either a char or an int. 1587 */ 1588 if (sopt->sopt_valsize == 1) { 1589 u_char loop; 1590 error = sooptcopyin(sopt, &loop, 1, 1); 1591 if (error) 1592 break; 1593 imo->imo_multicast_loop = !!loop; 1594 } else { 1595 u_int loop; 1596 error = sooptcopyin(sopt, &loop, sizeof loop, 1597 sizeof loop); 1598 if (error) 1599 break; 1600 imo->imo_multicast_loop = !!loop; 1601 } 1602 break; 1603 1604 case IP_ADD_MEMBERSHIP: 1605 /* 1606 * Add a multicast group membership. 1607 * Group must be a valid IP multicast address. 1608 */ 1609 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1610 if (error) 1611 break; 1612 1613 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1614 error = EINVAL; 1615 break; 1616 } 1617 s = splimp(); 1618 /* 1619 * If no interface address was provided, use the interface of 1620 * the route to the given multicast address. 1621 */ 1622 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1623 bzero((caddr_t)&ro, sizeof(ro)); 1624 dst = (struct sockaddr_in *)&ro.ro_dst; 1625 dst->sin_len = sizeof(*dst); 1626 dst->sin_family = AF_INET; 1627 dst->sin_addr = mreq.imr_multiaddr; 1628 rtalloc(&ro); 1629 if (ro.ro_rt == NULL) { 1630 error = EADDRNOTAVAIL; 1631 splx(s); 1632 break; 1633 } 1634 ifp = ro.ro_rt->rt_ifp; 1635 rtfree(ro.ro_rt); 1636 } 1637 else { 1638 INADDR_TO_IFP(mreq.imr_interface, ifp); 1639 } 1640 1641 /* 1642 * See if we found an interface, and confirm that it 1643 * supports multicast. 1644 */ 1645 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1646 error = EADDRNOTAVAIL; 1647 splx(s); 1648 break; 1649 } 1650 /* 1651 * See if the membership already exists or if all the 1652 * membership slots are full. 1653 */ 1654 for (i = 0; i < imo->imo_num_memberships; ++i) { 1655 if (imo->imo_membership[i]->inm_ifp == ifp && 1656 imo->imo_membership[i]->inm_addr.s_addr 1657 == mreq.imr_multiaddr.s_addr) 1658 break; 1659 } 1660 if (i < imo->imo_num_memberships) { 1661 error = EADDRINUSE; 1662 splx(s); 1663 break; 1664 } 1665 if (i == IP_MAX_MEMBERSHIPS) { 1666 error = ETOOMANYREFS; 1667 splx(s); 1668 break; 1669 } 1670 /* 1671 * Everything looks good; add a new record to the multicast 1672 * address list for the given interface. 1673 */ 1674 if ((imo->imo_membership[i] = 1675 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1676 error = ENOBUFS; 1677 splx(s); 1678 break; 1679 } 1680 ++imo->imo_num_memberships; 1681 splx(s); 1682 break; 1683 1684 case IP_DROP_MEMBERSHIP: 1685 /* 1686 * Drop a multicast group membership. 1687 * Group must be a valid IP multicast address. 1688 */ 1689 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1690 if (error) 1691 break; 1692 1693 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1694 error = EINVAL; 1695 break; 1696 } 1697 1698 s = splimp(); 1699 /* 1700 * If an interface address was specified, get a pointer 1701 * to its ifnet structure. 1702 */ 1703 if (mreq.imr_interface.s_addr == INADDR_ANY) 1704 ifp = NULL; 1705 else { 1706 INADDR_TO_IFP(mreq.imr_interface, ifp); 1707 if (ifp == NULL) { 1708 error = EADDRNOTAVAIL; 1709 splx(s); 1710 break; 1711 } 1712 } 1713 /* 1714 * Find the membership in the membership array. 1715 */ 1716 for (i = 0; i < imo->imo_num_memberships; ++i) { 1717 if ((ifp == NULL || 1718 imo->imo_membership[i]->inm_ifp == ifp) && 1719 imo->imo_membership[i]->inm_addr.s_addr == 1720 mreq.imr_multiaddr.s_addr) 1721 break; 1722 } 1723 if (i == imo->imo_num_memberships) { 1724 error = EADDRNOTAVAIL; 1725 splx(s); 1726 break; 1727 } 1728 /* 1729 * Give up the multicast address record to which the 1730 * membership points. 1731 */ 1732 in_delmulti(imo->imo_membership[i]); 1733 /* 1734 * Remove the gap in the membership array. 1735 */ 1736 for (++i; i < imo->imo_num_memberships; ++i) 1737 imo->imo_membership[i-1] = imo->imo_membership[i]; 1738 --imo->imo_num_memberships; 1739 splx(s); 1740 break; 1741 1742 default: 1743 error = EOPNOTSUPP; 1744 break; 1745 } 1746 1747 /* 1748 * If all options have default values, no need to keep the mbuf. 1749 */ 1750 if (imo->imo_multicast_ifp == NULL && 1751 imo->imo_multicast_vif == -1 && 1752 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1753 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1754 imo->imo_num_memberships == 0) { 1755 free(*imop, M_IPMOPTS); 1756 *imop = NULL; 1757 } 1758 1759 return (error); 1760} 1761 1762/* 1763 * Return the IP multicast options in response to user getsockopt(). 1764 */ 1765static int 1766ip_getmoptions(sopt, imo) 1767 struct sockopt *sopt; 1768 register struct ip_moptions *imo; 1769{ 1770 struct in_addr addr; 1771 struct in_ifaddr *ia; 1772 int error, optval; 1773 u_char coptval; 1774 1775 error = 0; 1776 switch (sopt->sopt_name) { 1777 case IP_MULTICAST_VIF: 1778 if (imo != NULL) 1779 optval = imo->imo_multicast_vif; 1780 else 1781 optval = -1; 1782 error = sooptcopyout(sopt, &optval, sizeof optval); 1783 break; 1784 1785 case IP_MULTICAST_IF: 1786 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1787 addr.s_addr = INADDR_ANY; 1788 else { 1789 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1790 addr.s_addr = (ia == NULL) ? INADDR_ANY 1791 : IA_SIN(ia)->sin_addr.s_addr; 1792 } 1793 error = sooptcopyout(sopt, &addr, sizeof addr); 1794 break; 1795 1796 case IP_MULTICAST_TTL: 1797 if (imo == 0) 1798 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1799 else 1800 optval = coptval = imo->imo_multicast_ttl; 1801 if (sopt->sopt_valsize == 1) 1802 error = sooptcopyout(sopt, &coptval, 1); 1803 else 1804 error = sooptcopyout(sopt, &optval, sizeof optval); 1805 break; 1806 1807 case IP_MULTICAST_LOOP: 1808 if (imo == 0) 1809 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1810 else 1811 optval = coptval = imo->imo_multicast_loop; 1812 if (sopt->sopt_valsize == 1) 1813 error = sooptcopyout(sopt, &coptval, 1); 1814 else 1815 error = sooptcopyout(sopt, &optval, sizeof optval); 1816 break; 1817 1818 default: 1819 error = ENOPROTOOPT; 1820 break; 1821 } 1822 return (error); 1823} 1824 1825/* 1826 * Discard the IP multicast options. 1827 */ 1828void 1829ip_freemoptions(imo) 1830 register struct ip_moptions *imo; 1831{ 1832 register int i; 1833 1834 if (imo != NULL) { 1835 for (i = 0; i < imo->imo_num_memberships; ++i) 1836 in_delmulti(imo->imo_membership[i]); 1837 free(imo, M_IPMOPTS); 1838 } 1839} 1840 1841/* 1842 * Routine called from ip_output() to loop back a copy of an IP multicast 1843 * packet to the input queue of a specified interface. Note that this 1844 * calls the output routine of the loopback "driver", but with an interface 1845 * pointer that might NOT be a loopback interface -- evil, but easier than 1846 * replicating that code here. 1847 */ 1848static void 1849ip_mloopback(ifp, m, dst, hlen) 1850 struct ifnet *ifp; 1851 register struct mbuf *m; 1852 register struct sockaddr_in *dst; 1853 int hlen; 1854{ 1855 register struct ip *ip; 1856 struct mbuf *copym; 1857 1858 copym = m_copy(m, 0, M_COPYALL); 1859 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1860 copym = m_pullup(copym, hlen); 1861 if (copym != NULL) { 1862 /* 1863 * We don't bother to fragment if the IP length is greater 1864 * than the interface's MTU. Can this possibly matter? 1865 */ 1866 ip = mtod(copym, struct ip *); 1867 HTONS(ip->ip_len); 1868 HTONS(ip->ip_off); 1869 ip->ip_sum = 0; 1870 if (ip->ip_vhl == IP_VHL_BORING) { 1871 ip->ip_sum = in_cksum_hdr(ip); 1872 } else { 1873 ip->ip_sum = in_cksum(copym, hlen); 1874 } 1875 /* 1876 * NB: 1877 * It's not clear whether there are any lingering 1878 * reentrancy problems in other areas which might 1879 * be exposed by using ip_input directly (in 1880 * particular, everything which modifies the packet 1881 * in-place). Yet another option is using the 1882 * protosw directly to deliver the looped back 1883 * packet. For the moment, we'll err on the side 1884 * of safety by using if_simloop(). 1885 */ 1886#if 1 /* XXX */ 1887 if (dst->sin_family != AF_INET) { 1888 printf("ip_mloopback: bad address family %d\n", 1889 dst->sin_family); 1890 dst->sin_family = AF_INET; 1891 } 1892#endif 1893 1894#ifdef notdef 1895 copym->m_pkthdr.rcvif = ifp; 1896 ip_input(copym); 1897#else 1898 /* if the checksum hasn't been computed, mark it as valid */ 1899 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1900 copym->m_pkthdr.csum_flags |= 1901 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1902 copym->m_pkthdr.csum_data = 0xffff; 1903 } 1904 if_simloop(ifp, copym, dst->sin_family, 0); 1905#endif 1906 } 1907} 1908